summaryrefslogtreecommitdiffhomepage
path: root/pkg/sentry/kernel
diff options
context:
space:
mode:
Diffstat (limited to 'pkg/sentry/kernel')
-rw-r--r--pkg/sentry/kernel/BUILD294
-rw-r--r--pkg/sentry/kernel/README.md108
-rw-r--r--pkg/sentry/kernel/auth/BUILD69
-rw-r--r--pkg/sentry/kernel/auth/atomicptr_credentials_unsafe.go37
-rw-r--r--pkg/sentry/kernel/auth/auth_state_autogen.go266
-rw-r--r--pkg/sentry/kernel/auth/auth_unsafe_state_autogen.go35
-rw-r--r--pkg/sentry/kernel/auth/id_map_range.go62
-rw-r--r--pkg/sentry/kernel/auth/id_map_set.go1643
-rw-r--r--pkg/sentry/kernel/contexttest/BUILD17
-rw-r--r--pkg/sentry/kernel/contexttest/contexttest.go40
-rw-r--r--pkg/sentry/kernel/epoll/BUILD51
-rw-r--r--pkg/sentry/kernel/epoll/epoll_list.go193
-rw-r--r--pkg/sentry/kernel/epoll/epoll_state_autogen.go182
-rw-r--r--pkg/sentry/kernel/epoll/epoll_test.go55
-rw-r--r--pkg/sentry/kernel/eventfd/BUILD33
-rw-r--r--pkg/sentry/kernel/eventfd/eventfd_state_autogen.go43
-rw-r--r--pkg/sentry/kernel/eventfd/eventfd_test.go78
-rw-r--r--pkg/sentry/kernel/fasync/BUILD18
-rw-r--r--pkg/sentry/kernel/fasync/fasync_state_autogen.go49
-rw-r--r--pkg/sentry/kernel/fd_table_refs.go117
-rw-r--r--pkg/sentry/kernel/fd_table_test.go228
-rw-r--r--pkg/sentry/kernel/fs_context_refs.go117
-rw-r--r--pkg/sentry/kernel/futex/BUILD58
-rw-r--r--pkg/sentry/kernel/futex/atomicptr_bucket_unsafe.go37
-rw-r--r--pkg/sentry/kernel/futex/futex_state_autogen.go114
-rw-r--r--pkg/sentry/kernel/futex/futex_test.go536
-rw-r--r--pkg/sentry/kernel/futex/futex_unsafe_state_autogen.go35
-rw-r--r--pkg/sentry/kernel/futex/waiter_list.go193
-rw-r--r--pkg/sentry/kernel/g3doc/run_states.dot99
-rw-r--r--pkg/sentry/kernel/g3doc/run_states.pngbin234152 -> 0 bytes
-rw-r--r--pkg/sentry/kernel/kernel_amd64_state_autogen.go5
-rw-r--r--pkg/sentry/kernel/kernel_arm64_state_autogen.go5
-rw-r--r--pkg/sentry/kernel/kernel_opts_state_autogen.go30
-rw-r--r--pkg/sentry/kernel/kernel_state_autogen.go2326
-rw-r--r--pkg/sentry/kernel/kernel_unsafe_state_autogen.go3
-rw-r--r--pkg/sentry/kernel/memevent/BUILD24
-rw-r--r--pkg/sentry/kernel/memevent/memevent_state_autogen.go3
-rw-r--r--pkg/sentry/kernel/memevent/memory_events.proto29
-rw-r--r--pkg/sentry/kernel/memevent/memory_events_go_proto/memory_events.pb.go88
-rw-r--r--pkg/sentry/kernel/pending_signals_list.go193
-rw-r--r--pkg/sentry/kernel/pipe/BUILD54
-rw-r--r--pkg/sentry/kernel/pipe/node_test.go320
-rw-r--r--pkg/sentry/kernel/pipe/pipe_state_autogen.go151
-rw-r--r--pkg/sentry/kernel/pipe/pipe_test.go139
-rw-r--r--pkg/sentry/kernel/pipe/pipe_unsafe_state_autogen.go3
-rw-r--r--pkg/sentry/kernel/process_group_list.go193
-rw-r--r--pkg/sentry/kernel/process_group_refs.go117
-rw-r--r--pkg/sentry/kernel/sched/BUILD19
-rw-r--r--pkg/sentry/kernel/sched/cpuset_test.go44
-rw-r--r--pkg/sentry/kernel/sched/sched_state_autogen.go3
-rw-r--r--pkg/sentry/kernel/semaphore/BUILD49
-rw-r--r--pkg/sentry/kernel/semaphore/semaphore_state_autogen.go205
-rw-r--r--pkg/sentry/kernel/semaphore/semaphore_test.go172
-rw-r--r--pkg/sentry/kernel/semaphore/waiter_list.go193
-rw-r--r--pkg/sentry/kernel/seqatomic_taskgoroutineschedinfo_unsafe.go49
-rw-r--r--pkg/sentry/kernel/session_list.go193
-rw-r--r--pkg/sentry/kernel/session_refs.go117
-rw-r--r--pkg/sentry/kernel/shm/BUILD41
-rw-r--r--pkg/sentry/kernel/shm/shm_refs.go117
-rw-r--r--pkg/sentry/kernel/shm/shm_state_autogen.go142
-rw-r--r--pkg/sentry/kernel/signalfd/BUILD22
-rw-r--r--pkg/sentry/kernel/signalfd/signalfd_state_autogen.go37
-rw-r--r--pkg/sentry/kernel/socket_list.go193
-rw-r--r--pkg/sentry/kernel/table_test.go110
-rw-r--r--pkg/sentry/kernel/task_list.go193
-rw-r--r--pkg/sentry/kernel/task_test.go69
-rw-r--r--pkg/sentry/kernel/time/BUILD20
-rw-r--r--pkg/sentry/kernel/time/time_state_autogen.go97
-rw-r--r--pkg/sentry/kernel/timekeeper_test.go156
-rw-r--r--pkg/sentry/kernel/uncaught_signal.proto37
-rw-r--r--pkg/sentry/kernel/uncaught_signal_go_proto/uncaught_signal.pb.go119
71 files changed, 7898 insertions, 2989 deletions
diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD
deleted file mode 100644
index d436daab4..000000000
--- a/pkg/sentry/kernel/BUILD
+++ /dev/null
@@ -1,294 +0,0 @@
-load("//tools:defs.bzl", "go_library", "go_test", "proto_library")
-load("//tools/go_generics:defs.bzl", "go_template_instance")
-
-package(licenses = ["notice"])
-
-go_template_instance(
- name = "pending_signals_list",
- out = "pending_signals_list.go",
- package = "kernel",
- prefix = "pendingSignal",
- template = "//pkg/ilist:generic_list",
- types = {
- "Element": "*pendingSignal",
- "Linker": "*pendingSignal",
- },
-)
-
-go_template_instance(
- name = "process_group_list",
- out = "process_group_list.go",
- package = "kernel",
- prefix = "processGroup",
- template = "//pkg/ilist:generic_list",
- types = {
- "Element": "*ProcessGroup",
- "Linker": "*ProcessGroup",
- },
-)
-
-go_template_instance(
- name = "seqatomic_taskgoroutineschedinfo",
- out = "seqatomic_taskgoroutineschedinfo_unsafe.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 = {
- "Element": "*Session",
- "Linker": "*Session",
- },
-)
-
-go_template_instance(
- name = "task_list",
- out = "task_list.go",
- package = "kernel",
- prefix = "task",
- template = "//pkg/ilist:generic_list",
- types = {
- "Element": "*Task",
- "Linker": "*Task",
- },
-)
-
-go_template_instance(
- name = "socket_list",
- out = "socket_list.go",
- package = "kernel",
- prefix = "socket",
- template = "//pkg/ilist:generic_list",
- types = {
- "Element": "*SocketEntry",
- "Linker": "*SocketEntry",
- },
-)
-
-go_template_instance(
- name = "fd_table_refs",
- out = "fd_table_refs.go",
- package = "kernel",
- prefix = "FDTable",
- template = "//pkg/refs_vfs2:refs_template",
- types = {
- "T": "FDTable",
- },
-)
-
-go_template_instance(
- name = "fs_context_refs",
- out = "fs_context_refs.go",
- package = "kernel",
- prefix = "FSContext",
- template = "//pkg/refs_vfs2:refs_template",
- types = {
- "T": "FSContext",
- },
-)
-
-go_template_instance(
- name = "process_group_refs",
- out = "process_group_refs.go",
- package = "kernel",
- prefix = "ProcessGroup",
- template = "//pkg/refs_vfs2:refs_template",
- types = {
- "T": "ProcessGroup",
- },
-)
-
-go_template_instance(
- name = "session_refs",
- out = "session_refs.go",
- package = "kernel",
- prefix = "Session",
- template = "//pkg/refs_vfs2:refs_template",
- types = {
- "T": "Session",
- },
-)
-
-proto_library(
- name = "uncaught_signal",
- srcs = ["uncaught_signal.proto"],
- visibility = ["//visibility:public"],
- deps = ["//pkg/sentry/arch:registers_proto"],
-)
-
-go_library(
- name = "kernel",
- srcs = [
- "abstract_socket_namespace.go",
- "aio.go",
- "context.go",
- "fd_table.go",
- "fd_table_refs.go",
- "fd_table_unsafe.go",
- "fs_context.go",
- "fs_context_refs.go",
- "ipc_namespace.go",
- "kcov.go",
- "kcov_unsafe.go",
- "kernel.go",
- "kernel_opts.go",
- "kernel_state.go",
- "pending_signals.go",
- "pending_signals_list.go",
- "pending_signals_state.go",
- "posixtimer.go",
- "process_group_list.go",
- "process_group_refs.go",
- "ptrace.go",
- "ptrace_amd64.go",
- "ptrace_arm64.go",
- "rseq.go",
- "seccomp.go",
- "seqatomic_taskgoroutineschedinfo_unsafe.go",
- "session_list.go",
- "session_refs.go",
- "sessions.go",
- "signal.go",
- "signal_handlers.go",
- "socket_list.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_futex.go",
- "task_identity.go",
- "task_list.go",
- "task_log.go",
- "task_net.go",
- "task_run.go",
- "task_sched.go",
- "task_signals.go",
- "task_start.go",
- "task_stop.go",
- "task_syscall.go",
- "task_usermem.go",
- "task_work.go",
- "thread_group.go",
- "threads.go",
- "timekeeper.go",
- "timekeeper_state.go",
- "tty.go",
- "uts_namespace.go",
- "vdso.go",
- "version.go",
- ],
- imports = [
- "gvisor.dev/gvisor/pkg/bpf",
- "gvisor.dev/gvisor/pkg/sentry/device",
- "gvisor.dev/gvisor/pkg/tcpip",
- ],
- visibility = ["//:sandbox"],
- deps = [
- ":uncaught_signal_go_proto",
- "//pkg/abi",
- "//pkg/abi/linux",
- "//pkg/amutex",
- "//pkg/binary",
- "//pkg/bits",
- "//pkg/bpf",
- "//pkg/context",
- "//pkg/coverage",
- "//pkg/cpuid",
- "//pkg/eventchannel",
- "//pkg/fspath",
- "//pkg/log",
- "//pkg/metric",
- "//pkg/refs",
- "//pkg/refs_vfs2",
- "//pkg/safemem",
- "//pkg/secio",
- "//pkg/sentry/arch",
- "//pkg/sentry/device",
- "//pkg/sentry/fs",
- "//pkg/sentry/fs/lock",
- "//pkg/sentry/fs/timerfd",
- "//pkg/sentry/fsbridge",
- "//pkg/sentry/fsimpl/kernfs",
- "//pkg/sentry/fsimpl/pipefs",
- "//pkg/sentry/fsimpl/sockfs",
- "//pkg/sentry/fsimpl/timerfd",
- "//pkg/sentry/fsimpl/tmpfs",
- "//pkg/sentry/hostcpu",
- "//pkg/sentry/inet",
- "//pkg/sentry/kernel/auth",
- "//pkg/sentry/kernel/epoll",
- "//pkg/sentry/kernel/futex",
- "//pkg/sentry/kernel/sched",
- "//pkg/sentry/kernel/semaphore",
- "//pkg/sentry/kernel/shm",
- "//pkg/sentry/kernel/time",
- "//pkg/sentry/limits",
- "//pkg/sentry/loader",
- "//pkg/sentry/memmap",
- "//pkg/sentry/mm",
- "//pkg/sentry/pgalloc",
- "//pkg/sentry/platform",
- "//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/vfs",
- "//pkg/state",
- "//pkg/state/statefile",
- "//pkg/state/wire",
- "//pkg/sync",
- "//pkg/syserr",
- "//pkg/syserror",
- "//pkg/tcpip",
- "//pkg/tcpip/stack",
- "//pkg/usermem",
- "//pkg/waiter",
- "//tools/go_marshal/marshal",
- ],
-)
-
-go_test(
- name = "kernel_test",
- size = "small",
- srcs = [
- "fd_table_test.go",
- "table_test.go",
- "task_test.go",
- "timekeeper_test.go",
- ],
- library = ":kernel",
- deps = [
- "//pkg/abi",
- "//pkg/context",
- "//pkg/sentry/arch",
- "//pkg/sentry/contexttest",
- "//pkg/sentry/fs",
- "//pkg/sentry/fs/filetest",
- "//pkg/sentry/kernel/sched",
- "//pkg/sentry/limits",
- "//pkg/sentry/pgalloc",
- "//pkg/sentry/time",
- "//pkg/sentry/usage",
- "//pkg/sync",
- "//pkg/syserror",
- "//pkg/usermem",
- ],
-)
diff --git a/pkg/sentry/kernel/README.md b/pkg/sentry/kernel/README.md
deleted file mode 100644
index 427311be8..000000000
--- a/pkg/sentry/kernel/README.md
+++ /dev/null
@@ -1,108 +0,0 @@
-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 -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
-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/auth/BUILD b/pkg/sentry/kernel/auth/BUILD
deleted file mode 100644
index 2bc49483a..000000000
--- a/pkg/sentry/kernel/auth/BUILD
+++ /dev/null
@@ -1,69 +0,0 @@
-load("//tools:defs.bzl", "go_library")
-load("//tools/go_generics:defs.bzl", "go_template_instance")
-
-package(licenses = ["notice"])
-
-go_template_instance(
- name = "atomicptr_credentials",
- out = "atomicptr_credentials_unsafe.go",
- package = "auth",
- suffix = "Credentials",
- template = "//pkg/sync:generic_atomicptr",
- types = {
- "Value": "Credentials",
- },
-)
-
-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 = [
- "atomicptr_credentials_unsafe.go",
- "auth.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",
- ],
- visibility = ["//pkg/sentry:internal"],
- deps = [
- "//pkg/abi/linux",
- "//pkg/bits",
- "//pkg/context",
- "//pkg/log",
- "//pkg/sync",
- "//pkg/syserror",
- ],
-)
diff --git a/pkg/sentry/kernel/auth/atomicptr_credentials_unsafe.go b/pkg/sentry/kernel/auth/atomicptr_credentials_unsafe.go
new file mode 100644
index 000000000..4535c958f
--- /dev/null
+++ b/pkg/sentry/kernel/auth/atomicptr_credentials_unsafe.go
@@ -0,0 +1,37 @@
+package auth
+
+import (
+ "sync/atomic"
+ "unsafe"
+)
+
+// 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.
+//
+// +stateify savable
+type AtomicPtrCredentials struct {
+ ptr unsafe.Pointer `state:".(*Credentials)"`
+}
+
+func (p *AtomicPtrCredentials) savePtr() *Credentials {
+ return p.Load()
+}
+
+func (p *AtomicPtrCredentials) loadPtr(v *Credentials) {
+ p.Store(v)
+}
+
+// Load returns the value set by the most recent Store. It returns nil if there
+// has been no previous call to Store.
+func (p *AtomicPtrCredentials) Load() *Credentials {
+ return (*Credentials)(atomic.LoadPointer(&p.ptr))
+}
+
+// Store sets the value returned by Load to x.
+func (p *AtomicPtrCredentials) Store(x *Credentials) {
+ atomic.StorePointer(&p.ptr, (unsafe.Pointer)(x))
+}
diff --git a/pkg/sentry/kernel/auth/auth_state_autogen.go b/pkg/sentry/kernel/auth/auth_state_autogen.go
new file mode 100644
index 000000000..7a710d06d
--- /dev/null
+++ b/pkg/sentry/kernel/auth/auth_state_autogen.go
@@ -0,0 +1,266 @@
+// automatically generated by stateify.
+
+package auth
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *Credentials) StateTypeName() string {
+ return "pkg/sentry/kernel/auth.Credentials"
+}
+
+func (x *Credentials) StateFields() []string {
+ return []string{
+ "RealKUID",
+ "EffectiveKUID",
+ "SavedKUID",
+ "RealKGID",
+ "EffectiveKGID",
+ "SavedKGID",
+ "ExtraKGIDs",
+ "PermittedCaps",
+ "InheritableCaps",
+ "EffectiveCaps",
+ "BoundingCaps",
+ "KeepCaps",
+ "UserNamespace",
+ }
+}
+
+func (x *Credentials) beforeSave() {}
+
+func (x *Credentials) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.RealKUID)
+ m.Save(1, &x.EffectiveKUID)
+ m.Save(2, &x.SavedKUID)
+ m.Save(3, &x.RealKGID)
+ m.Save(4, &x.EffectiveKGID)
+ m.Save(5, &x.SavedKGID)
+ m.Save(6, &x.ExtraKGIDs)
+ m.Save(7, &x.PermittedCaps)
+ m.Save(8, &x.InheritableCaps)
+ m.Save(9, &x.EffectiveCaps)
+ m.Save(10, &x.BoundingCaps)
+ m.Save(11, &x.KeepCaps)
+ m.Save(12, &x.UserNamespace)
+}
+
+func (x *Credentials) afterLoad() {}
+
+func (x *Credentials) StateLoad(m state.Source) {
+ m.Load(0, &x.RealKUID)
+ m.Load(1, &x.EffectiveKUID)
+ m.Load(2, &x.SavedKUID)
+ m.Load(3, &x.RealKGID)
+ m.Load(4, &x.EffectiveKGID)
+ m.Load(5, &x.SavedKGID)
+ m.Load(6, &x.ExtraKGIDs)
+ m.Load(7, &x.PermittedCaps)
+ m.Load(8, &x.InheritableCaps)
+ m.Load(9, &x.EffectiveCaps)
+ m.Load(10, &x.BoundingCaps)
+ m.Load(11, &x.KeepCaps)
+ m.Load(12, &x.UserNamespace)
+}
+
+func (x *IDMapEntry) StateTypeName() string {
+ return "pkg/sentry/kernel/auth.IDMapEntry"
+}
+
+func (x *IDMapEntry) StateFields() []string {
+ return []string{
+ "FirstID",
+ "FirstParentID",
+ "Length",
+ }
+}
+
+func (x *IDMapEntry) beforeSave() {}
+
+func (x *IDMapEntry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.FirstID)
+ m.Save(1, &x.FirstParentID)
+ m.Save(2, &x.Length)
+}
+
+func (x *IDMapEntry) afterLoad() {}
+
+func (x *IDMapEntry) StateLoad(m state.Source) {
+ m.Load(0, &x.FirstID)
+ m.Load(1, &x.FirstParentID)
+ m.Load(2, &x.Length)
+}
+
+func (x *idMapRange) StateTypeName() string {
+ return "pkg/sentry/kernel/auth.idMapRange"
+}
+
+func (x *idMapRange) StateFields() []string {
+ return []string{
+ "Start",
+ "End",
+ }
+}
+
+func (x *idMapRange) beforeSave() {}
+
+func (x *idMapRange) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.Start)
+ m.Save(1, &x.End)
+}
+
+func (x *idMapRange) afterLoad() {}
+
+func (x *idMapRange) StateLoad(m state.Source) {
+ m.Load(0, &x.Start)
+ m.Load(1, &x.End)
+}
+
+func (x *idMapSet) StateTypeName() string {
+ return "pkg/sentry/kernel/auth.idMapSet"
+}
+
+func (x *idMapSet) StateFields() []string {
+ return []string{
+ "root",
+ }
+}
+
+func (x *idMapSet) beforeSave() {}
+
+func (x *idMapSet) StateSave(m state.Sink) {
+ x.beforeSave()
+ var root *idMapSegmentDataSlices = x.saveRoot()
+ m.SaveValue(0, root)
+}
+
+func (x *idMapSet) afterLoad() {}
+
+func (x *idMapSet) StateLoad(m state.Source) {
+ m.LoadValue(0, new(*idMapSegmentDataSlices), func(y interface{}) { x.loadRoot(y.(*idMapSegmentDataSlices)) })
+}
+
+func (x *idMapnode) StateTypeName() string {
+ return "pkg/sentry/kernel/auth.idMapnode"
+}
+
+func (x *idMapnode) StateFields() []string {
+ return []string{
+ "nrSegments",
+ "parent",
+ "parentIndex",
+ "hasChildren",
+ "maxGap",
+ "keys",
+ "values",
+ "children",
+ }
+}
+
+func (x *idMapnode) beforeSave() {}
+
+func (x *idMapnode) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.nrSegments)
+ m.Save(1, &x.parent)
+ m.Save(2, &x.parentIndex)
+ m.Save(3, &x.hasChildren)
+ m.Save(4, &x.maxGap)
+ m.Save(5, &x.keys)
+ m.Save(6, &x.values)
+ m.Save(7, &x.children)
+}
+
+func (x *idMapnode) afterLoad() {}
+
+func (x *idMapnode) StateLoad(m state.Source) {
+ m.Load(0, &x.nrSegments)
+ m.Load(1, &x.parent)
+ m.Load(2, &x.parentIndex)
+ m.Load(3, &x.hasChildren)
+ m.Load(4, &x.maxGap)
+ m.Load(5, &x.keys)
+ m.Load(6, &x.values)
+ m.Load(7, &x.children)
+}
+
+func (x *idMapSegmentDataSlices) StateTypeName() string {
+ return "pkg/sentry/kernel/auth.idMapSegmentDataSlices"
+}
+
+func (x *idMapSegmentDataSlices) StateFields() []string {
+ return []string{
+ "Start",
+ "End",
+ "Values",
+ }
+}
+
+func (x *idMapSegmentDataSlices) beforeSave() {}
+
+func (x *idMapSegmentDataSlices) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.Start)
+ m.Save(1, &x.End)
+ m.Save(2, &x.Values)
+}
+
+func (x *idMapSegmentDataSlices) afterLoad() {}
+
+func (x *idMapSegmentDataSlices) StateLoad(m state.Source) {
+ m.Load(0, &x.Start)
+ m.Load(1, &x.End)
+ m.Load(2, &x.Values)
+}
+
+func (x *UserNamespace) StateTypeName() string {
+ return "pkg/sentry/kernel/auth.UserNamespace"
+}
+
+func (x *UserNamespace) StateFields() []string {
+ return []string{
+ "parent",
+ "owner",
+ "uidMapFromParent",
+ "uidMapToParent",
+ "gidMapFromParent",
+ "gidMapToParent",
+ }
+}
+
+func (x *UserNamespace) beforeSave() {}
+
+func (x *UserNamespace) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.parent)
+ m.Save(1, &x.owner)
+ m.Save(2, &x.uidMapFromParent)
+ m.Save(3, &x.uidMapToParent)
+ m.Save(4, &x.gidMapFromParent)
+ m.Save(5, &x.gidMapToParent)
+}
+
+func (x *UserNamespace) afterLoad() {}
+
+func (x *UserNamespace) StateLoad(m state.Source) {
+ m.Load(0, &x.parent)
+ m.Load(1, &x.owner)
+ m.Load(2, &x.uidMapFromParent)
+ m.Load(3, &x.uidMapToParent)
+ m.Load(4, &x.gidMapFromParent)
+ m.Load(5, &x.gidMapToParent)
+}
+
+func init() {
+ state.Register((*Credentials)(nil))
+ state.Register((*IDMapEntry)(nil))
+ state.Register((*idMapRange)(nil))
+ state.Register((*idMapSet)(nil))
+ state.Register((*idMapnode)(nil))
+ state.Register((*idMapSegmentDataSlices)(nil))
+ state.Register((*UserNamespace)(nil))
+}
diff --git a/pkg/sentry/kernel/auth/auth_unsafe_state_autogen.go b/pkg/sentry/kernel/auth/auth_unsafe_state_autogen.go
new file mode 100644
index 000000000..0a3507dd5
--- /dev/null
+++ b/pkg/sentry/kernel/auth/auth_unsafe_state_autogen.go
@@ -0,0 +1,35 @@
+// automatically generated by stateify.
+
+package auth
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *AtomicPtrCredentials) StateTypeName() string {
+ return "pkg/sentry/kernel/auth.AtomicPtrCredentials"
+}
+
+func (x *AtomicPtrCredentials) StateFields() []string {
+ return []string{
+ "ptr",
+ }
+}
+
+func (x *AtomicPtrCredentials) beforeSave() {}
+
+func (x *AtomicPtrCredentials) StateSave(m state.Sink) {
+ x.beforeSave()
+ var ptr *Credentials = x.savePtr()
+ m.SaveValue(0, ptr)
+}
+
+func (x *AtomicPtrCredentials) afterLoad() {}
+
+func (x *AtomicPtrCredentials) StateLoad(m state.Source) {
+ m.LoadValue(0, new(*Credentials), func(y interface{}) { x.loadPtr(y.(*Credentials)) })
+}
+
+func init() {
+ state.Register((*AtomicPtrCredentials)(nil))
+}
diff --git a/pkg/sentry/kernel/auth/id_map_range.go b/pkg/sentry/kernel/auth/id_map_range.go
new file mode 100644
index 000000000..833fa3518
--- /dev/null
+++ b/pkg/sentry/kernel/auth/id_map_range.go
@@ -0,0 +1,62 @@
+package auth
+
+// A Range represents a contiguous range of T.
+//
+// +stateify savable
+type idMapRange struct {
+ // Start is the inclusive start of the range.
+ Start uint32
+
+ // End is the exclusive end of the range.
+ End uint32
+}
+
+// WellFormed returns true if r.Start <= r.End. All other methods on a Range
+// require that the Range is well-formed.
+func (r idMapRange) WellFormed() bool {
+ return r.Start <= r.End
+}
+
+// Length returns the length of the range.
+func (r idMapRange) Length() uint32 {
+ return r.End - r.Start
+}
+
+// Contains returns true if r contains x.
+func (r idMapRange) Contains(x uint32) bool {
+ return r.Start <= x && x < r.End
+}
+
+// Overlaps returns true if r and r2 overlap.
+func (r idMapRange) Overlaps(r2 idMapRange) bool {
+ return r.Start < r2.End && r2.Start < r.End
+}
+
+// IsSupersetOf returns true if r is a superset of r2; that is, the range r2 is
+// contained within r.
+func (r idMapRange) IsSupersetOf(r2 idMapRange) bool {
+ return r.Start <= r2.Start && r.End >= r2.End
+}
+
+// Intersect returns a range consisting of the intersection between r and r2.
+// If r and r2 do not overlap, Intersect returns a range with unspecified
+// bounds, but for which Length() == 0.
+func (r idMapRange) Intersect(r2 idMapRange) idMapRange {
+ if r.Start < r2.Start {
+ r.Start = r2.Start
+ }
+ if r.End > r2.End {
+ r.End = r2.End
+ }
+ if r.End < r.Start {
+ r.End = r.Start
+ }
+ return r
+}
+
+// CanSplitAt returns true if it is legal to split a segment spanning the range
+// r at x; that is, splitting at x would produce two ranges, both of which have
+// non-zero length.
+func (r idMapRange) CanSplitAt(x uint32) bool {
+ return r.Contains(x) && r.Start < x
+}
diff --git a/pkg/sentry/kernel/auth/id_map_set.go b/pkg/sentry/kernel/auth/id_map_set.go
new file mode 100644
index 000000000..d8a05ce46
--- /dev/null
+++ b/pkg/sentry/kernel/auth/id_map_set.go
@@ -0,0 +1,1643 @@
+package auth
+
+import (
+ "bytes"
+ "fmt"
+)
+
+// trackGaps is an optional parameter.
+//
+// If trackGaps is 1, the Set will track maximum gap size recursively,
+// enabling the GapIterator.{Prev,Next}LargeEnoughGap functions. In this
+// case, Key must be an unsigned integer.
+//
+// trackGaps must be 0 or 1.
+const idMaptrackGaps = 0
+
+var _ = uint8(idMaptrackGaps << 7) // Will fail if not zero or one.
+
+// dynamicGap is a type that disappears if trackGaps is 0.
+type idMapdynamicGap [idMaptrackGaps]uint32
+
+// Get returns the value of the gap.
+//
+// Precondition: trackGaps must be non-zero.
+func (d *idMapdynamicGap) Get() uint32 {
+ return d[:][0]
+}
+
+// Set sets the value of the gap.
+//
+// Precondition: trackGaps must be non-zero.
+func (d *idMapdynamicGap) Set(v uint32) {
+ d[:][0] = v
+}
+
+const (
+ // minDegree is the minimum degree of an internal node in a Set B-tree.
+ //
+ // - Any non-root node has at least minDegree-1 segments.
+ //
+ // - Any non-root internal (non-leaf) node has at least minDegree children.
+ //
+ // - The root node may have fewer than minDegree-1 segments, but it may
+ // only have 0 segments if the tree is empty.
+ //
+ // Our implementation requires minDegree >= 3. Higher values of minDegree
+ // usually improve performance, but increase memory usage for small sets.
+ idMapminDegree = 3
+
+ idMapmaxDegree = 2 * idMapminDegree
+)
+
+// 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 idMapSet struct {
+ root idMapnode `state:".(*idMapSegmentDataSlices)"`
+}
+
+// IsEmpty returns true if the set contains no segments.
+func (s *idMapSet) IsEmpty() bool {
+ return s.root.nrSegments == 0
+}
+
+// IsEmptyRange returns true iff no segments in the set overlap the given
+// range. This is semantically equivalent to s.SpanRange(r) == 0, but may be
+// more efficient.
+func (s *idMapSet) IsEmptyRange(r idMapRange) bool {
+ switch {
+ case r.Length() < 0:
+ panic(fmt.Sprintf("invalid range %v", r))
+ case r.Length() == 0:
+ return true
+ }
+ _, gap := s.Find(r.Start)
+ if !gap.Ok() {
+ return false
+ }
+ return r.End <= gap.End()
+}
+
+// Span returns the total size of all segments in the set.
+func (s *idMapSet) Span() uint32 {
+ var sz uint32
+ for seg := s.FirstSegment(); seg.Ok(); seg = seg.NextSegment() {
+ sz += seg.Range().Length()
+ }
+ return sz
+}
+
+// SpanRange returns the total size of the intersection of segments in the set
+// with the given range.
+func (s *idMapSet) SpanRange(r idMapRange) uint32 {
+ switch {
+ case r.Length() < 0:
+ panic(fmt.Sprintf("invalid range %v", r))
+ case r.Length() == 0:
+ return 0
+ }
+ var sz uint32
+ for seg := s.LowerBoundSegment(r.Start); seg.Ok() && seg.Start() < r.End; seg = seg.NextSegment() {
+ sz += seg.Range().Intersect(r).Length()
+ }
+ return sz
+}
+
+// FirstSegment returns the first segment in the set. If the set is empty,
+// FirstSegment returns a terminal iterator.
+func (s *idMapSet) FirstSegment() idMapIterator {
+ if s.root.nrSegments == 0 {
+ return idMapIterator{}
+ }
+ return s.root.firstSegment()
+}
+
+// LastSegment returns the last segment in the set. If the set is empty,
+// LastSegment returns a terminal iterator.
+func (s *idMapSet) LastSegment() idMapIterator {
+ if s.root.nrSegments == 0 {
+ return idMapIterator{}
+ }
+ return s.root.lastSegment()
+}
+
+// FirstGap returns the first gap in the set.
+func (s *idMapSet) FirstGap() idMapGapIterator {
+ n := &s.root
+ for n.hasChildren {
+ n = n.children[0]
+ }
+ return idMapGapIterator{n, 0}
+}
+
+// LastGap returns the last gap in the set.
+func (s *idMapSet) LastGap() idMapGapIterator {
+ n := &s.root
+ for n.hasChildren {
+ n = n.children[n.nrSegments]
+ }
+ return idMapGapIterator{n, n.nrSegments}
+}
+
+// Find returns the segment or gap whose range contains the given key. If a
+// segment is found, the returned Iterator is non-terminal and the
+// returned GapIterator is terminal. Otherwise, the returned Iterator is
+// terminal and the returned GapIterator is non-terminal.
+func (s *idMapSet) Find(key uint32) (idMapIterator, idMapGapIterator) {
+ n := &s.root
+ for {
+
+ lower := 0
+ upper := n.nrSegments
+ for lower < upper {
+ i := lower + (upper-lower)/2
+ if r := n.keys[i]; key < r.End {
+ if key >= r.Start {
+ return idMapIterator{n, i}, idMapGapIterator{}
+ }
+ upper = i
+ } else {
+ lower = i + 1
+ }
+ }
+ i := lower
+ if !n.hasChildren {
+ return idMapIterator{}, idMapGapIterator{n, i}
+ }
+ n = n.children[i]
+ }
+}
+
+// FindSegment returns the segment whose range contains the given key. If no
+// such segment exists, FindSegment returns a terminal iterator.
+func (s *idMapSet) FindSegment(key uint32) idMapIterator {
+ seg, _ := s.Find(key)
+ return seg
+}
+
+// LowerBoundSegment returns the segment with the lowest range that contains a
+// key greater than or equal to min. If no such segment exists,
+// LowerBoundSegment returns a terminal iterator.
+func (s *idMapSet) LowerBoundSegment(min uint32) idMapIterator {
+ seg, gap := s.Find(min)
+ if seg.Ok() {
+ return seg
+ }
+ return gap.NextSegment()
+}
+
+// UpperBoundSegment returns the segment with the highest range that contains a
+// key less than or equal to max. If no such segment exists, UpperBoundSegment
+// returns a terminal iterator.
+func (s *idMapSet) UpperBoundSegment(max uint32) idMapIterator {
+ seg, gap := s.Find(max)
+ if seg.Ok() {
+ return seg
+ }
+ return gap.PrevSegment()
+}
+
+// FindGap returns the gap containing the given key. If no such gap exists
+// (i.e. the set contains a segment containing that key), FindGap returns a
+// terminal iterator.
+func (s *idMapSet) FindGap(key uint32) idMapGapIterator {
+ _, gap := s.Find(key)
+ return gap
+}
+
+// LowerBoundGap returns the gap with the lowest range that is greater than or
+// equal to min.
+func (s *idMapSet) LowerBoundGap(min uint32) idMapGapIterator {
+ seg, gap := s.Find(min)
+ if gap.Ok() {
+ return gap
+ }
+ return seg.NextGap()
+}
+
+// UpperBoundGap returns the gap with the highest range that is less than or
+// equal to max.
+func (s *idMapSet) UpperBoundGap(max uint32) idMapGapIterator {
+ seg, gap := s.Find(max)
+ if gap.Ok() {
+ return gap
+ }
+ return seg.PrevGap()
+}
+
+// Add inserts the given segment into the set and returns true. If the new
+// segment can be merged with adjacent segments, Add will do so. If the new
+// segment would overlap an existing segment, Add returns false. If Add
+// succeeds, all existing iterators are invalidated.
+func (s *idMapSet) Add(r idMapRange, val uint32) bool {
+ if r.Length() <= 0 {
+ panic(fmt.Sprintf("invalid segment range %v", r))
+ }
+ gap := s.FindGap(r.Start)
+ if !gap.Ok() {
+ return false
+ }
+ if r.End > gap.End() {
+ return false
+ }
+ s.Insert(gap, r, val)
+ return true
+}
+
+// AddWithoutMerging inserts the given segment into the set and returns true.
+// If it would overlap an existing segment, AddWithoutMerging does nothing and
+// returns false. If AddWithoutMerging succeeds, all existing iterators are
+// invalidated.
+func (s *idMapSet) AddWithoutMerging(r idMapRange, val uint32) bool {
+ if r.Length() <= 0 {
+ panic(fmt.Sprintf("invalid segment range %v", r))
+ }
+ gap := s.FindGap(r.Start)
+ if !gap.Ok() {
+ return false
+ }
+ if r.End > gap.End() {
+ return false
+ }
+ s.InsertWithoutMergingUnchecked(gap, r, val)
+ return true
+}
+
+// Insert inserts the given segment into the given gap. If the new segment can
+// be merged with adjacent segments, Insert will do so. Insert returns an
+// iterator to the segment containing the inserted value (which may have been
+// merged with other values). All existing iterators (including gap, but not
+// including the returned iterator) are invalidated.
+//
+// If the gap cannot accommodate the segment, or if r is invalid, Insert panics.
+//
+// Insert is semantically equivalent to a InsertWithoutMerging followed by a
+// Merge, but may be more efficient. Note that there is no unchecked variant of
+// Insert since Insert must retrieve and inspect gap's predecessor and
+// successor segments regardless.
+func (s *idMapSet) Insert(gap idMapGapIterator, r idMapRange, val uint32) idMapIterator {
+ if r.Length() <= 0 {
+ panic(fmt.Sprintf("invalid segment range %v", r))
+ }
+ prev, next := gap.PrevSegment(), gap.NextSegment()
+ if prev.Ok() && prev.End() > r.Start {
+ panic(fmt.Sprintf("new segment %v overlaps predecessor %v", r, prev.Range()))
+ }
+ if next.Ok() && next.Start() < r.End {
+ panic(fmt.Sprintf("new segment %v overlaps successor %v", r, next.Range()))
+ }
+ if prev.Ok() && prev.End() == r.Start {
+ if mval, ok := (idMapFunctions{}).Merge(prev.Range(), prev.Value(), r, val); ok {
+ shrinkMaxGap := idMaptrackGaps != 0 && gap.Range().Length() == gap.node.maxGap.Get()
+ prev.SetEndUnchecked(r.End)
+ prev.SetValue(mval)
+ if shrinkMaxGap {
+ gap.node.updateMaxGapLeaf()
+ }
+ if next.Ok() && next.Start() == r.End {
+ val = mval
+ if mval, ok := (idMapFunctions{}).Merge(prev.Range(), val, next.Range(), next.Value()); ok {
+ prev.SetEndUnchecked(next.End())
+ prev.SetValue(mval)
+ return s.Remove(next).PrevSegment()
+ }
+ }
+ return prev
+ }
+ }
+ if next.Ok() && next.Start() == r.End {
+ if mval, ok := (idMapFunctions{}).Merge(r, val, next.Range(), next.Value()); ok {
+ shrinkMaxGap := idMaptrackGaps != 0 && gap.Range().Length() == gap.node.maxGap.Get()
+ next.SetStartUnchecked(r.Start)
+ next.SetValue(mval)
+ if shrinkMaxGap {
+ gap.node.updateMaxGapLeaf()
+ }
+ return next
+ }
+ }
+
+ return s.InsertWithoutMergingUnchecked(gap, r, val)
+}
+
+// InsertWithoutMerging inserts the given segment into the given gap and
+// returns an iterator to the inserted segment. All existing iterators
+// (including gap, but not including the returned iterator) are invalidated.
+//
+// If the gap cannot accommodate the segment, or if r is invalid,
+// InsertWithoutMerging panics.
+func (s *idMapSet) InsertWithoutMerging(gap idMapGapIterator, r idMapRange, val uint32) idMapIterator {
+ if r.Length() <= 0 {
+ panic(fmt.Sprintf("invalid segment range %v", r))
+ }
+ if gr := gap.Range(); !gr.IsSupersetOf(r) {
+ panic(fmt.Sprintf("cannot insert segment range %v into gap range %v", r, gr))
+ }
+ return s.InsertWithoutMergingUnchecked(gap, r, val)
+}
+
+// InsertWithoutMergingUnchecked inserts the given segment into the given gap
+// and returns an iterator to the inserted segment. All existing iterators
+// (including gap, but not including the returned iterator) are invalidated.
+//
+// Preconditions:
+// * r.Start >= gap.Start().
+// * r.End <= gap.End().
+func (s *idMapSet) InsertWithoutMergingUnchecked(gap idMapGapIterator, r idMapRange, val uint32) idMapIterator {
+ gap = gap.node.rebalanceBeforeInsert(gap)
+ splitMaxGap := idMaptrackGaps != 0 && (gap.node.nrSegments == 0 || gap.Range().Length() == gap.node.maxGap.Get())
+ copy(gap.node.keys[gap.index+1:], gap.node.keys[gap.index:gap.node.nrSegments])
+ copy(gap.node.values[gap.index+1:], gap.node.values[gap.index:gap.node.nrSegments])
+ gap.node.keys[gap.index] = r
+ gap.node.values[gap.index] = val
+ gap.node.nrSegments++
+ if splitMaxGap {
+ gap.node.updateMaxGapLeaf()
+ }
+ return idMapIterator{gap.node, gap.index}
+}
+
+// Remove removes the given segment and returns an iterator to the vacated gap.
+// All existing iterators (including seg, but not including the returned
+// iterator) are invalidated.
+func (s *idMapSet) Remove(seg idMapIterator) idMapGapIterator {
+
+ if seg.node.hasChildren {
+
+ victim := seg.PrevSegment()
+
+ seg.SetRangeUnchecked(victim.Range())
+ seg.SetValue(victim.Value())
+
+ nextAdjacentNode := seg.NextSegment().node
+ if idMaptrackGaps != 0 {
+ nextAdjacentNode.updateMaxGapLeaf()
+ }
+ return s.Remove(victim).NextGap()
+ }
+ copy(seg.node.keys[seg.index:], seg.node.keys[seg.index+1:seg.node.nrSegments])
+ copy(seg.node.values[seg.index:], seg.node.values[seg.index+1:seg.node.nrSegments])
+ idMapFunctions{}.ClearValue(&seg.node.values[seg.node.nrSegments-1])
+ seg.node.nrSegments--
+ if idMaptrackGaps != 0 {
+ seg.node.updateMaxGapLeaf()
+ }
+ return seg.node.rebalanceAfterRemove(idMapGapIterator{seg.node, seg.index})
+}
+
+// RemoveAll removes all segments from the set. All existing iterators are
+// invalidated.
+func (s *idMapSet) RemoveAll() {
+ s.root = idMapnode{}
+}
+
+// RemoveRange removes all segments in the given range. An iterator to the
+// newly formed gap is returned, and all existing iterators are invalidated.
+func (s *idMapSet) RemoveRange(r idMapRange) idMapGapIterator {
+ seg, gap := s.Find(r.Start)
+ if seg.Ok() {
+ seg = s.Isolate(seg, r)
+ gap = s.Remove(seg)
+ }
+ for seg = gap.NextSegment(); seg.Ok() && seg.Start() < r.End; seg = gap.NextSegment() {
+ seg = s.Isolate(seg, r)
+ gap = s.Remove(seg)
+ }
+ return gap
+}
+
+// Merge attempts to merge two neighboring segments. If successful, Merge
+// returns an iterator to the merged segment, and all existing iterators are
+// invalidated. Otherwise, Merge returns a terminal iterator.
+//
+// If first is not the predecessor of second, Merge panics.
+func (s *idMapSet) Merge(first, second idMapIterator) idMapIterator {
+ if first.NextSegment() != second {
+ panic(fmt.Sprintf("attempt to merge non-neighboring segments %v, %v", first.Range(), second.Range()))
+ }
+ return s.MergeUnchecked(first, second)
+}
+
+// MergeUnchecked attempts to merge two neighboring segments. If successful,
+// MergeUnchecked returns an iterator to the merged segment, and all existing
+// iterators are invalidated. Otherwise, MergeUnchecked returns a terminal
+// iterator.
+//
+// Precondition: first is the predecessor of second: first.NextSegment() ==
+// second, first == second.PrevSegment().
+func (s *idMapSet) MergeUnchecked(first, second idMapIterator) idMapIterator {
+ if first.End() == second.Start() {
+ if mval, ok := (idMapFunctions{}).Merge(first.Range(), first.Value(), second.Range(), second.Value()); ok {
+
+ first.SetEndUnchecked(second.End())
+ first.SetValue(mval)
+
+ return s.Remove(second).PrevSegment()
+ }
+ }
+ return idMapIterator{}
+}
+
+// MergeAll attempts to merge all adjacent segments in the set. All existing
+// iterators are invalidated.
+func (s *idMapSet) MergeAll() {
+ seg := s.FirstSegment()
+ if !seg.Ok() {
+ return
+ }
+ next := seg.NextSegment()
+ for next.Ok() {
+ if mseg := s.MergeUnchecked(seg, next); mseg.Ok() {
+ seg, next = mseg, mseg.NextSegment()
+ } else {
+ seg, next = next, next.NextSegment()
+ }
+ }
+}
+
+// MergeRange attempts to merge all adjacent segments that contain a key in the
+// specific range. All existing iterators are invalidated.
+func (s *idMapSet) MergeRange(r idMapRange) {
+ seg := s.LowerBoundSegment(r.Start)
+ if !seg.Ok() {
+ return
+ }
+ next := seg.NextSegment()
+ for next.Ok() && next.Range().Start < r.End {
+ if mseg := s.MergeUnchecked(seg, next); mseg.Ok() {
+ seg, next = mseg, mseg.NextSegment()
+ } else {
+ seg, next = next, next.NextSegment()
+ }
+ }
+}
+
+// MergeAdjacent attempts to merge the segment containing r.Start with its
+// predecessor, and the segment containing r.End-1 with its successor.
+func (s *idMapSet) MergeAdjacent(r idMapRange) {
+ first := s.FindSegment(r.Start)
+ if first.Ok() {
+ if prev := first.PrevSegment(); prev.Ok() {
+ s.Merge(prev, first)
+ }
+ }
+ last := s.FindSegment(r.End - 1)
+ if last.Ok() {
+ if next := last.NextSegment(); next.Ok() {
+ s.Merge(last, next)
+ }
+ }
+}
+
+// Split splits the given segment at the given key and returns iterators to the
+// two resulting segments. All existing iterators (including seg, but not
+// including the returned iterators) are invalidated.
+//
+// If the segment cannot be split at split (because split is at the start or
+// end of the segment's range, so splitting would produce a segment with zero
+// length, or because split falls outside the segment's range altogether),
+// Split panics.
+func (s *idMapSet) Split(seg idMapIterator, split uint32) (idMapIterator, idMapIterator) {
+ if !seg.Range().CanSplitAt(split) {
+ panic(fmt.Sprintf("can't split %v at %v", seg.Range(), split))
+ }
+ return s.SplitUnchecked(seg, split)
+}
+
+// SplitUnchecked splits the given segment at the given key and returns
+// iterators to the two resulting segments. All existing iterators (including
+// seg, but not including the returned iterators) are invalidated.
+//
+// Preconditions: seg.Start() < key < seg.End().
+func (s *idMapSet) SplitUnchecked(seg idMapIterator, split uint32) (idMapIterator, idMapIterator) {
+ val1, val2 := (idMapFunctions{}).Split(seg.Range(), seg.Value(), split)
+ end2 := seg.End()
+ seg.SetEndUnchecked(split)
+ seg.SetValue(val1)
+ seg2 := s.InsertWithoutMergingUnchecked(seg.NextGap(), idMapRange{split, end2}, val2)
+
+ return seg2.PrevSegment(), seg2
+}
+
+// SplitAt splits the segment straddling split, if one exists. SplitAt returns
+// true if a segment was split and false otherwise. If SplitAt splits a
+// segment, all existing iterators are invalidated.
+func (s *idMapSet) SplitAt(split uint32) bool {
+ if seg := s.FindSegment(split); seg.Ok() && seg.Range().CanSplitAt(split) {
+ s.SplitUnchecked(seg, split)
+ return true
+ }
+ return false
+}
+
+// Isolate ensures that the given segment's range does not escape r by
+// splitting at r.Start and r.End if necessary, and returns an updated iterator
+// to the bounded segment. All existing iterators (including seg, but not
+// including the returned iterators) are invalidated.
+func (s *idMapSet) Isolate(seg idMapIterator, r idMapRange) idMapIterator {
+ if seg.Range().CanSplitAt(r.Start) {
+ _, seg = s.SplitUnchecked(seg, r.Start)
+ }
+ if seg.Range().CanSplitAt(r.End) {
+ seg, _ = s.SplitUnchecked(seg, r.End)
+ }
+ return seg
+}
+
+// ApplyContiguous applies a function to a contiguous range of segments,
+// splitting if necessary. The function is applied until the first gap is
+// encountered, at which point the gap is returned. If the function is applied
+// across the entire range, a terminal gap is returned. All existing iterators
+// are invalidated.
+//
+// N.B. The Iterator must not be invalidated by the function.
+func (s *idMapSet) ApplyContiguous(r idMapRange, fn func(seg idMapIterator)) idMapGapIterator {
+ seg, gap := s.Find(r.Start)
+ if !seg.Ok() {
+ return gap
+ }
+ for {
+ seg = s.Isolate(seg, r)
+ fn(seg)
+ if seg.End() >= r.End {
+ return idMapGapIterator{}
+ }
+ gap = seg.NextGap()
+ if !gap.IsEmpty() {
+ return gap
+ }
+ seg = gap.NextSegment()
+ if !seg.Ok() {
+
+ return idMapGapIterator{}
+ }
+ }
+}
+
+// +stateify savable
+type idMapnode struct {
+ // An internal binary tree node looks like:
+ //
+ // K
+ // / \
+ // Cl Cr
+ //
+ // where all keys in the subtree rooted by Cl (the left subtree) are less
+ // than K (the key of the parent node), and all keys in the subtree rooted
+ // by Cr (the right subtree) are greater than K.
+ //
+ // An internal B-tree node's indexes work out to look like:
+ //
+ // K0 K1 K2 ... Kn-1
+ // / \/ \/ \ ... / \
+ // C0 C1 C2 C3 ... Cn-1 Cn
+ //
+ // where n is nrSegments.
+ nrSegments int
+
+ // parent is a pointer to this node's parent. If this node is root, parent
+ // is nil.
+ parent *idMapnode
+
+ // parentIndex is the index of this node in parent.children.
+ parentIndex int
+
+ // Flag for internal nodes that is technically redundant with "children[0]
+ // != nil", but is stored in the first cache line. "hasChildren" rather
+ // than "isLeaf" because false must be the correct value for an empty root.
+ hasChildren bool
+
+ // The longest gap within this node. If the node is a leaf, it's simply the
+ // maximum gap among all the (nrSegments+1) gaps formed by its nrSegments keys
+ // including the 0th and nrSegments-th gap possibly shared with its upper-level
+ // nodes; if it's a non-leaf node, it's the max of all children's maxGap.
+ maxGap idMapdynamicGap
+
+ // Nodes store keys and values in separate arrays to maximize locality in
+ // the common case (scanning keys for lookup).
+ keys [idMapmaxDegree - 1]idMapRange
+ values [idMapmaxDegree - 1]uint32
+ children [idMapmaxDegree]*idMapnode
+}
+
+// firstSegment returns the first segment in the subtree rooted by n.
+//
+// Preconditions: n.nrSegments != 0.
+func (n *idMapnode) firstSegment() idMapIterator {
+ for n.hasChildren {
+ n = n.children[0]
+ }
+ return idMapIterator{n, 0}
+}
+
+// lastSegment returns the last segment in the subtree rooted by n.
+//
+// Preconditions: n.nrSegments != 0.
+func (n *idMapnode) lastSegment() idMapIterator {
+ for n.hasChildren {
+ n = n.children[n.nrSegments]
+ }
+ return idMapIterator{n, n.nrSegments - 1}
+}
+
+func (n *idMapnode) prevSibling() *idMapnode {
+ if n.parent == nil || n.parentIndex == 0 {
+ return nil
+ }
+ return n.parent.children[n.parentIndex-1]
+}
+
+func (n *idMapnode) nextSibling() *idMapnode {
+ if n.parent == nil || n.parentIndex == n.parent.nrSegments {
+ return nil
+ }
+ return n.parent.children[n.parentIndex+1]
+}
+
+// rebalanceBeforeInsert splits n and its ancestors if they are full, as
+// required for insertion, and returns an updated iterator to the position
+// represented by gap.
+func (n *idMapnode) rebalanceBeforeInsert(gap idMapGapIterator) idMapGapIterator {
+ if n.nrSegments < idMapmaxDegree-1 {
+ return gap
+ }
+ if n.parent != nil {
+ gap = n.parent.rebalanceBeforeInsert(gap)
+ }
+ if n.parent == nil {
+
+ left := &idMapnode{
+ nrSegments: idMapminDegree - 1,
+ parent: n,
+ parentIndex: 0,
+ hasChildren: n.hasChildren,
+ }
+ right := &idMapnode{
+ nrSegments: idMapminDegree - 1,
+ parent: n,
+ parentIndex: 1,
+ hasChildren: n.hasChildren,
+ }
+ copy(left.keys[:idMapminDegree-1], n.keys[:idMapminDegree-1])
+ copy(left.values[:idMapminDegree-1], n.values[:idMapminDegree-1])
+ copy(right.keys[:idMapminDegree-1], n.keys[idMapminDegree:])
+ copy(right.values[:idMapminDegree-1], n.values[idMapminDegree:])
+ n.keys[0], n.values[0] = n.keys[idMapminDegree-1], n.values[idMapminDegree-1]
+ idMapzeroValueSlice(n.values[1:])
+ if n.hasChildren {
+ copy(left.children[:idMapminDegree], n.children[:idMapminDegree])
+ copy(right.children[:idMapminDegree], n.children[idMapminDegree:])
+ idMapzeroNodeSlice(n.children[2:])
+ for i := 0; i < idMapminDegree; i++ {
+ left.children[i].parent = left
+ left.children[i].parentIndex = i
+ right.children[i].parent = right
+ right.children[i].parentIndex = i
+ }
+ }
+ n.nrSegments = 1
+ n.hasChildren = true
+ n.children[0] = left
+ n.children[1] = right
+
+ if idMaptrackGaps != 0 {
+ left.updateMaxGapLocal()
+ right.updateMaxGapLocal()
+ }
+ if gap.node != n {
+ return gap
+ }
+ if gap.index < idMapminDegree {
+ return idMapGapIterator{left, gap.index}
+ }
+ return idMapGapIterator{right, gap.index - idMapminDegree}
+ }
+
+ copy(n.parent.keys[n.parentIndex+1:], n.parent.keys[n.parentIndex:n.parent.nrSegments])
+ copy(n.parent.values[n.parentIndex+1:], n.parent.values[n.parentIndex:n.parent.nrSegments])
+ n.parent.keys[n.parentIndex], n.parent.values[n.parentIndex] = n.keys[idMapminDegree-1], n.values[idMapminDegree-1]
+ copy(n.parent.children[n.parentIndex+2:], n.parent.children[n.parentIndex+1:n.parent.nrSegments+1])
+ for i := n.parentIndex + 2; i < n.parent.nrSegments+2; i++ {
+ n.parent.children[i].parentIndex = i
+ }
+ sibling := &idMapnode{
+ nrSegments: idMapminDegree - 1,
+ parent: n.parent,
+ parentIndex: n.parentIndex + 1,
+ hasChildren: n.hasChildren,
+ }
+ n.parent.children[n.parentIndex+1] = sibling
+ n.parent.nrSegments++
+ copy(sibling.keys[:idMapminDegree-1], n.keys[idMapminDegree:])
+ copy(sibling.values[:idMapminDegree-1], n.values[idMapminDegree:])
+ idMapzeroValueSlice(n.values[idMapminDegree-1:])
+ if n.hasChildren {
+ copy(sibling.children[:idMapminDegree], n.children[idMapminDegree:])
+ idMapzeroNodeSlice(n.children[idMapminDegree:])
+ for i := 0; i < idMapminDegree; i++ {
+ sibling.children[i].parent = sibling
+ sibling.children[i].parentIndex = i
+ }
+ }
+ n.nrSegments = idMapminDegree - 1
+
+ if idMaptrackGaps != 0 {
+ n.updateMaxGapLocal()
+ sibling.updateMaxGapLocal()
+ }
+
+ if gap.node != n {
+ return gap
+ }
+ if gap.index < idMapminDegree {
+ return gap
+ }
+ return idMapGapIterator{sibling, gap.index - idMapminDegree}
+}
+
+// rebalanceAfterRemove "unsplits" n and its ancestors if they are deficient
+// (contain fewer segments than required by B-tree invariants), as required for
+// removal, and returns an updated iterator to the position represented by gap.
+//
+// Precondition: n is the only node in the tree that may currently violate a
+// B-tree invariant.
+func (n *idMapnode) rebalanceAfterRemove(gap idMapGapIterator) idMapGapIterator {
+ for {
+ if n.nrSegments >= idMapminDegree-1 {
+ return gap
+ }
+ if n.parent == nil {
+
+ return gap
+ }
+
+ if sibling := n.prevSibling(); sibling != nil && sibling.nrSegments >= idMapminDegree {
+ copy(n.keys[1:], n.keys[:n.nrSegments])
+ copy(n.values[1:], n.values[:n.nrSegments])
+ n.keys[0] = n.parent.keys[n.parentIndex-1]
+ n.values[0] = n.parent.values[n.parentIndex-1]
+ n.parent.keys[n.parentIndex-1] = sibling.keys[sibling.nrSegments-1]
+ n.parent.values[n.parentIndex-1] = sibling.values[sibling.nrSegments-1]
+ idMapFunctions{}.ClearValue(&sibling.values[sibling.nrSegments-1])
+ if n.hasChildren {
+ copy(n.children[1:], n.children[:n.nrSegments+1])
+ n.children[0] = sibling.children[sibling.nrSegments]
+ sibling.children[sibling.nrSegments] = nil
+ n.children[0].parent = n
+ n.children[0].parentIndex = 0
+ for i := 1; i < n.nrSegments+2; i++ {
+ n.children[i].parentIndex = i
+ }
+ }
+ n.nrSegments++
+ sibling.nrSegments--
+
+ if idMaptrackGaps != 0 {
+ n.updateMaxGapLocal()
+ sibling.updateMaxGapLocal()
+ }
+ if gap.node == sibling && gap.index == sibling.nrSegments {
+ return idMapGapIterator{n, 0}
+ }
+ if gap.node == n {
+ return idMapGapIterator{n, gap.index + 1}
+ }
+ return gap
+ }
+ if sibling := n.nextSibling(); sibling != nil && sibling.nrSegments >= idMapminDegree {
+ n.keys[n.nrSegments] = n.parent.keys[n.parentIndex]
+ n.values[n.nrSegments] = n.parent.values[n.parentIndex]
+ n.parent.keys[n.parentIndex] = sibling.keys[0]
+ n.parent.values[n.parentIndex] = sibling.values[0]
+ copy(sibling.keys[:sibling.nrSegments-1], sibling.keys[1:])
+ copy(sibling.values[:sibling.nrSegments-1], sibling.values[1:])
+ idMapFunctions{}.ClearValue(&sibling.values[sibling.nrSegments-1])
+ if n.hasChildren {
+ n.children[n.nrSegments+1] = sibling.children[0]
+ copy(sibling.children[:sibling.nrSegments], sibling.children[1:])
+ sibling.children[sibling.nrSegments] = nil
+ n.children[n.nrSegments+1].parent = n
+ n.children[n.nrSegments+1].parentIndex = n.nrSegments + 1
+ for i := 0; i < sibling.nrSegments; i++ {
+ sibling.children[i].parentIndex = i
+ }
+ }
+ n.nrSegments++
+ sibling.nrSegments--
+
+ if idMaptrackGaps != 0 {
+ n.updateMaxGapLocal()
+ sibling.updateMaxGapLocal()
+ }
+ if gap.node == sibling {
+ if gap.index == 0 {
+ return idMapGapIterator{n, n.nrSegments}
+ }
+ return idMapGapIterator{sibling, gap.index - 1}
+ }
+ return gap
+ }
+
+ p := n.parent
+ if p.nrSegments == 1 {
+
+ left, right := p.children[0], p.children[1]
+ p.nrSegments = left.nrSegments + right.nrSegments + 1
+ p.hasChildren = left.hasChildren
+ p.keys[left.nrSegments] = p.keys[0]
+ p.values[left.nrSegments] = p.values[0]
+ copy(p.keys[:left.nrSegments], left.keys[:left.nrSegments])
+ copy(p.values[:left.nrSegments], left.values[:left.nrSegments])
+ copy(p.keys[left.nrSegments+1:], right.keys[:right.nrSegments])
+ copy(p.values[left.nrSegments+1:], right.values[:right.nrSegments])
+ if left.hasChildren {
+ copy(p.children[:left.nrSegments+1], left.children[:left.nrSegments+1])
+ copy(p.children[left.nrSegments+1:], right.children[:right.nrSegments+1])
+ for i := 0; i < p.nrSegments+1; i++ {
+ p.children[i].parent = p
+ p.children[i].parentIndex = i
+ }
+ } else {
+ p.children[0] = nil
+ p.children[1] = nil
+ }
+
+ if gap.node == left {
+ return idMapGapIterator{p, gap.index}
+ }
+ if gap.node == right {
+ return idMapGapIterator{p, gap.index + left.nrSegments + 1}
+ }
+ return gap
+ }
+ // Merge n and either sibling, along with the segment separating the
+ // two, into whichever of the two nodes comes first. This is the
+ // reverse of the non-root splitting case in
+ // node.rebalanceBeforeInsert.
+ var left, right *idMapnode
+ if n.parentIndex > 0 {
+ left = n.prevSibling()
+ right = n
+ } else {
+ left = n
+ right = n.nextSibling()
+ }
+
+ if gap.node == right {
+ gap = idMapGapIterator{left, gap.index + left.nrSegments + 1}
+ }
+ left.keys[left.nrSegments] = p.keys[left.parentIndex]
+ left.values[left.nrSegments] = p.values[left.parentIndex]
+ copy(left.keys[left.nrSegments+1:], right.keys[:right.nrSegments])
+ copy(left.values[left.nrSegments+1:], right.values[:right.nrSegments])
+ if left.hasChildren {
+ copy(left.children[left.nrSegments+1:], right.children[:right.nrSegments+1])
+ for i := left.nrSegments + 1; i < left.nrSegments+right.nrSegments+2; i++ {
+ left.children[i].parent = left
+ left.children[i].parentIndex = i
+ }
+ }
+ left.nrSegments += right.nrSegments + 1
+ copy(p.keys[left.parentIndex:], p.keys[left.parentIndex+1:p.nrSegments])
+ copy(p.values[left.parentIndex:], p.values[left.parentIndex+1:p.nrSegments])
+ idMapFunctions{}.ClearValue(&p.values[p.nrSegments-1])
+ copy(p.children[left.parentIndex+1:], p.children[left.parentIndex+2:p.nrSegments+1])
+ for i := 0; i < p.nrSegments; i++ {
+ p.children[i].parentIndex = i
+ }
+ p.children[p.nrSegments] = nil
+ p.nrSegments--
+
+ if idMaptrackGaps != 0 {
+ left.updateMaxGapLocal()
+ }
+
+ n = p
+ }
+}
+
+// updateMaxGapLeaf updates maxGap bottom-up from the calling leaf until no
+// necessary update.
+//
+// Preconditions: n must be a leaf node, trackGaps must be 1.
+func (n *idMapnode) updateMaxGapLeaf() {
+ if n.hasChildren {
+ panic(fmt.Sprintf("updateMaxGapLeaf should always be called on leaf node: %v", n))
+ }
+ max := n.calculateMaxGapLeaf()
+ if max == n.maxGap.Get() {
+
+ return
+ }
+ oldMax := n.maxGap.Get()
+ n.maxGap.Set(max)
+ if max > oldMax {
+
+ for p := n.parent; p != nil; p = p.parent {
+ if p.maxGap.Get() >= max {
+
+ break
+ }
+
+ p.maxGap.Set(max)
+ }
+ return
+ }
+
+ for p := n.parent; p != nil; p = p.parent {
+ if p.maxGap.Get() > oldMax {
+
+ break
+ }
+
+ parentNewMax := p.calculateMaxGapInternal()
+ if p.maxGap.Get() == parentNewMax {
+
+ break
+ }
+
+ p.maxGap.Set(parentNewMax)
+ }
+}
+
+// updateMaxGapLocal updates maxGap of the calling node solely with no
+// propagation to ancestor nodes.
+//
+// Precondition: trackGaps must be 1.
+func (n *idMapnode) updateMaxGapLocal() {
+ if !n.hasChildren {
+
+ n.maxGap.Set(n.calculateMaxGapLeaf())
+ } else {
+
+ n.maxGap.Set(n.calculateMaxGapInternal())
+ }
+}
+
+// calculateMaxGapLeaf iterates the gaps within a leaf node and calculate the
+// max.
+//
+// Preconditions: n must be a leaf node.
+func (n *idMapnode) calculateMaxGapLeaf() uint32 {
+ max := idMapGapIterator{n, 0}.Range().Length()
+ for i := 1; i <= n.nrSegments; i++ {
+ if current := (idMapGapIterator{n, i}).Range().Length(); current > max {
+ max = current
+ }
+ }
+ return max
+}
+
+// calculateMaxGapInternal iterates children's maxGap within an internal node n
+// and calculate the max.
+//
+// Preconditions: n must be a non-leaf node.
+func (n *idMapnode) calculateMaxGapInternal() uint32 {
+ max := n.children[0].maxGap.Get()
+ for i := 1; i <= n.nrSegments; i++ {
+ if current := n.children[i].maxGap.Get(); current > max {
+ max = current
+ }
+ }
+ return max
+}
+
+// searchFirstLargeEnoughGap returns the first gap having at least minSize length
+// in the subtree rooted by n. If not found, return a terminal gap iterator.
+func (n *idMapnode) searchFirstLargeEnoughGap(minSize uint32) idMapGapIterator {
+ if n.maxGap.Get() < minSize {
+ return idMapGapIterator{}
+ }
+ if n.hasChildren {
+ for i := 0; i <= n.nrSegments; i++ {
+ if largeEnoughGap := n.children[i].searchFirstLargeEnoughGap(minSize); largeEnoughGap.Ok() {
+ return largeEnoughGap
+ }
+ }
+ } else {
+ for i := 0; i <= n.nrSegments; i++ {
+ currentGap := idMapGapIterator{n, i}
+ if currentGap.Range().Length() >= minSize {
+ return currentGap
+ }
+ }
+ }
+ panic(fmt.Sprintf("invalid maxGap in %v", n))
+}
+
+// searchLastLargeEnoughGap returns the last gap having at least minSize length
+// in the subtree rooted by n. If not found, return a terminal gap iterator.
+func (n *idMapnode) searchLastLargeEnoughGap(minSize uint32) idMapGapIterator {
+ if n.maxGap.Get() < minSize {
+ return idMapGapIterator{}
+ }
+ if n.hasChildren {
+ for i := n.nrSegments; i >= 0; i-- {
+ if largeEnoughGap := n.children[i].searchLastLargeEnoughGap(minSize); largeEnoughGap.Ok() {
+ return largeEnoughGap
+ }
+ }
+ } else {
+ for i := n.nrSegments; i >= 0; i-- {
+ currentGap := idMapGapIterator{n, i}
+ if currentGap.Range().Length() >= minSize {
+ return currentGap
+ }
+ }
+ }
+ panic(fmt.Sprintf("invalid maxGap in %v", n))
+}
+
+// A Iterator is conceptually one of:
+//
+// - A pointer to a segment in a set; or
+//
+// - A terminal iterator, which is a sentinel indicating that the end of
+// iteration has been reached.
+//
+// Iterators are copyable values and are meaningfully equality-comparable. The
+// zero value of Iterator is a terminal iterator.
+//
+// Unless otherwise specified, any mutation of a set invalidates all existing
+// iterators into the set.
+type idMapIterator struct {
+ // node is the node containing the iterated segment. If the iterator is
+ // terminal, node is nil.
+ node *idMapnode
+
+ // index is the index of the segment in node.keys/values.
+ index int
+}
+
+// Ok returns true if the iterator is not terminal. All other methods are only
+// valid for non-terminal iterators.
+func (seg idMapIterator) Ok() bool {
+ return seg.node != nil
+}
+
+// Range returns the iterated segment's range key.
+func (seg idMapIterator) Range() idMapRange {
+ return seg.node.keys[seg.index]
+}
+
+// Start is equivalent to Range().Start, but should be preferred if only the
+// start of the range is needed.
+func (seg idMapIterator) Start() uint32 {
+ return seg.node.keys[seg.index].Start
+}
+
+// End is equivalent to Range().End, but should be preferred if only the end of
+// the range is needed.
+func (seg idMapIterator) End() uint32 {
+ return seg.node.keys[seg.index].End
+}
+
+// SetRangeUnchecked mutates the iterated segment's range key. This operation
+// does not invalidate any iterators.
+//
+// Preconditions:
+// * r.Length() > 0.
+// * The new range must not overlap an existing one:
+// * If seg.NextSegment().Ok(), then r.end <= seg.NextSegment().Start().
+// * If seg.PrevSegment().Ok(), then r.start >= seg.PrevSegment().End().
+func (seg idMapIterator) SetRangeUnchecked(r idMapRange) {
+ seg.node.keys[seg.index] = r
+}
+
+// SetRange mutates the iterated segment's range key. If the new range would
+// cause the iterated segment to overlap another segment, or if the new range
+// is invalid, SetRange panics. This operation does not invalidate any
+// iterators.
+func (seg idMapIterator) SetRange(r idMapRange) {
+ if r.Length() <= 0 {
+ panic(fmt.Sprintf("invalid segment range %v", r))
+ }
+ if prev := seg.PrevSegment(); prev.Ok() && r.Start < prev.End() {
+ panic(fmt.Sprintf("new segment range %v overlaps segment range %v", r, prev.Range()))
+ }
+ if next := seg.NextSegment(); next.Ok() && r.End > next.Start() {
+ panic(fmt.Sprintf("new segment range %v overlaps segment range %v", r, next.Range()))
+ }
+ seg.SetRangeUnchecked(r)
+}
+
+// SetStartUnchecked mutates the iterated segment's start. This operation does
+// not invalidate any iterators.
+//
+// Preconditions: The new start must be valid:
+// * start < seg.End()
+// * If seg.PrevSegment().Ok(), then start >= seg.PrevSegment().End().
+func (seg idMapIterator) SetStartUnchecked(start uint32) {
+ seg.node.keys[seg.index].Start = start
+}
+
+// SetStart mutates the iterated segment's start. If the new start value would
+// cause the iterated segment to overlap another segment, or would result in an
+// invalid range, SetStart panics. This operation does not invalidate any
+// iterators.
+func (seg idMapIterator) SetStart(start uint32) {
+ if start >= seg.End() {
+ panic(fmt.Sprintf("new start %v would invalidate segment range %v", start, seg.Range()))
+ }
+ if prev := seg.PrevSegment(); prev.Ok() && start < prev.End() {
+ panic(fmt.Sprintf("new start %v would cause segment range %v to overlap segment range %v", start, seg.Range(), prev.Range()))
+ }
+ seg.SetStartUnchecked(start)
+}
+
+// SetEndUnchecked mutates the iterated segment's end. This operation does not
+// invalidate any iterators.
+//
+// Preconditions: The new end must be valid:
+// * end > seg.Start().
+// * If seg.NextSegment().Ok(), then end <= seg.NextSegment().Start().
+func (seg idMapIterator) SetEndUnchecked(end uint32) {
+ seg.node.keys[seg.index].End = end
+}
+
+// SetEnd mutates the iterated segment's end. If the new end value would cause
+// the iterated segment to overlap another segment, or would result in an
+// invalid range, SetEnd panics. This operation does not invalidate any
+// iterators.
+func (seg idMapIterator) SetEnd(end uint32) {
+ if end <= seg.Start() {
+ panic(fmt.Sprintf("new end %v would invalidate segment range %v", end, seg.Range()))
+ }
+ if next := seg.NextSegment(); next.Ok() && end > next.Start() {
+ panic(fmt.Sprintf("new end %v would cause segment range %v to overlap segment range %v", end, seg.Range(), next.Range()))
+ }
+ seg.SetEndUnchecked(end)
+}
+
+// Value returns a copy of the iterated segment's value.
+func (seg idMapIterator) Value() uint32 {
+ return seg.node.values[seg.index]
+}
+
+// ValuePtr returns a pointer to the iterated segment's value. The pointer is
+// invalidated if the iterator is invalidated. This operation does not
+// invalidate any iterators.
+func (seg idMapIterator) ValuePtr() *uint32 {
+ return &seg.node.values[seg.index]
+}
+
+// SetValue mutates the iterated segment's value. This operation does not
+// invalidate any iterators.
+func (seg idMapIterator) SetValue(val uint32) {
+ seg.node.values[seg.index] = val
+}
+
+// PrevSegment returns the iterated segment's predecessor. If there is no
+// preceding segment, PrevSegment returns a terminal iterator.
+func (seg idMapIterator) PrevSegment() idMapIterator {
+ if seg.node.hasChildren {
+ return seg.node.children[seg.index].lastSegment()
+ }
+ if seg.index > 0 {
+ return idMapIterator{seg.node, seg.index - 1}
+ }
+ if seg.node.parent == nil {
+ return idMapIterator{}
+ }
+ return idMapsegmentBeforePosition(seg.node.parent, seg.node.parentIndex)
+}
+
+// NextSegment returns the iterated segment's successor. If there is no
+// succeeding segment, NextSegment returns a terminal iterator.
+func (seg idMapIterator) NextSegment() idMapIterator {
+ if seg.node.hasChildren {
+ return seg.node.children[seg.index+1].firstSegment()
+ }
+ if seg.index < seg.node.nrSegments-1 {
+ return idMapIterator{seg.node, seg.index + 1}
+ }
+ if seg.node.parent == nil {
+ return idMapIterator{}
+ }
+ return idMapsegmentAfterPosition(seg.node.parent, seg.node.parentIndex)
+}
+
+// PrevGap returns the gap immediately before the iterated segment.
+func (seg idMapIterator) PrevGap() idMapGapIterator {
+ if seg.node.hasChildren {
+
+ return seg.node.children[seg.index].lastSegment().NextGap()
+ }
+ return idMapGapIterator{seg.node, seg.index}
+}
+
+// NextGap returns the gap immediately after the iterated segment.
+func (seg idMapIterator) NextGap() idMapGapIterator {
+ if seg.node.hasChildren {
+ return seg.node.children[seg.index+1].firstSegment().PrevGap()
+ }
+ return idMapGapIterator{seg.node, seg.index + 1}
+}
+
+// PrevNonEmpty returns the iterated segment's predecessor if it is adjacent,
+// or the gap before the iterated segment otherwise. If seg.Start() ==
+// Functions.MinKey(), PrevNonEmpty will return two terminal iterators.
+// Otherwise, exactly one of the iterators returned by PrevNonEmpty will be
+// non-terminal.
+func (seg idMapIterator) PrevNonEmpty() (idMapIterator, idMapGapIterator) {
+ gap := seg.PrevGap()
+ if gap.Range().Length() != 0 {
+ return idMapIterator{}, gap
+ }
+ return gap.PrevSegment(), idMapGapIterator{}
+}
+
+// NextNonEmpty returns the iterated segment's successor if it is adjacent, or
+// the gap after the iterated segment otherwise. If seg.End() ==
+// Functions.MaxKey(), NextNonEmpty will return two terminal iterators.
+// Otherwise, exactly one of the iterators returned by NextNonEmpty will be
+// non-terminal.
+func (seg idMapIterator) NextNonEmpty() (idMapIterator, idMapGapIterator) {
+ gap := seg.NextGap()
+ if gap.Range().Length() != 0 {
+ return idMapIterator{}, gap
+ }
+ return gap.NextSegment(), idMapGapIterator{}
+}
+
+// A GapIterator is conceptually one of:
+//
+// - A pointer to a position between two segments, before the first segment, or
+// after the last segment in a set, called a *gap*; or
+//
+// - A terminal iterator, which is a sentinel indicating that the end of
+// iteration has been reached.
+//
+// Note that the gap between two adjacent segments exists (iterators to it are
+// non-terminal), but has a length of zero. GapIterator.IsEmpty returns true
+// for such gaps. An empty set contains a single gap, spanning the entire range
+// of the set's keys.
+//
+// GapIterators are copyable values and are meaningfully equality-comparable.
+// The zero value of GapIterator is a terminal iterator.
+//
+// Unless otherwise specified, any mutation of a set invalidates all existing
+// iterators into the set.
+type idMapGapIterator struct {
+ // The representation of a GapIterator is identical to that of an Iterator,
+ // except that index corresponds to positions between segments in the same
+ // way as for node.children (see comment for node.nrSegments).
+ node *idMapnode
+ index int
+}
+
+// Ok returns true if the iterator is not terminal. All other methods are only
+// valid for non-terminal iterators.
+func (gap idMapGapIterator) Ok() bool {
+ return gap.node != nil
+}
+
+// Range returns the range spanned by the iterated gap.
+func (gap idMapGapIterator) Range() idMapRange {
+ return idMapRange{gap.Start(), gap.End()}
+}
+
+// Start is equivalent to Range().Start, but should be preferred if only the
+// start of the range is needed.
+func (gap idMapGapIterator) Start() uint32 {
+ if ps := gap.PrevSegment(); ps.Ok() {
+ return ps.End()
+ }
+ return idMapFunctions{}.MinKey()
+}
+
+// End is equivalent to Range().End, but should be preferred if only the end of
+// the range is needed.
+func (gap idMapGapIterator) End() uint32 {
+ if ns := gap.NextSegment(); ns.Ok() {
+ return ns.Start()
+ }
+ return idMapFunctions{}.MaxKey()
+}
+
+// IsEmpty returns true if the iterated gap is empty (that is, the "gap" is
+// between two adjacent segments.)
+func (gap idMapGapIterator) IsEmpty() bool {
+ return gap.Range().Length() == 0
+}
+
+// PrevSegment returns the segment immediately before the iterated gap. If no
+// such segment exists, PrevSegment returns a terminal iterator.
+func (gap idMapGapIterator) PrevSegment() idMapIterator {
+ return idMapsegmentBeforePosition(gap.node, gap.index)
+}
+
+// NextSegment returns the segment immediately after the iterated gap. If no
+// such segment exists, NextSegment returns a terminal iterator.
+func (gap idMapGapIterator) NextSegment() idMapIterator {
+ return idMapsegmentAfterPosition(gap.node, gap.index)
+}
+
+// PrevGap returns the iterated gap's predecessor. If no such gap exists,
+// PrevGap returns a terminal iterator.
+func (gap idMapGapIterator) PrevGap() idMapGapIterator {
+ seg := gap.PrevSegment()
+ if !seg.Ok() {
+ return idMapGapIterator{}
+ }
+ return seg.PrevGap()
+}
+
+// NextGap returns the iterated gap's successor. If no such gap exists, NextGap
+// returns a terminal iterator.
+func (gap idMapGapIterator) NextGap() idMapGapIterator {
+ seg := gap.NextSegment()
+ if !seg.Ok() {
+ return idMapGapIterator{}
+ }
+ return seg.NextGap()
+}
+
+// NextLargeEnoughGap returns the iterated gap's first next gap with larger
+// length than minSize. If not found, return a terminal gap iterator (does NOT
+// include this gap itself).
+//
+// Precondition: trackGaps must be 1.
+func (gap idMapGapIterator) NextLargeEnoughGap(minSize uint32) idMapGapIterator {
+ if idMaptrackGaps != 1 {
+ panic("set is not tracking gaps")
+ }
+ if gap.node != nil && gap.node.hasChildren && gap.index == gap.node.nrSegments {
+
+ gap.node = gap.NextSegment().node
+ gap.index = 0
+ return gap.nextLargeEnoughGapHelper(minSize)
+ }
+ return gap.nextLargeEnoughGapHelper(minSize)
+}
+
+// nextLargeEnoughGapHelper is the helper function used by NextLargeEnoughGap
+// to do the real recursions.
+//
+// Preconditions: gap is NOT the trailing gap of a non-leaf node.
+func (gap idMapGapIterator) nextLargeEnoughGapHelper(minSize uint32) idMapGapIterator {
+
+ for gap.node != nil &&
+ (gap.node.maxGap.Get() < minSize || (!gap.node.hasChildren && gap.index == gap.node.nrSegments)) {
+ gap.node, gap.index = gap.node.parent, gap.node.parentIndex
+ }
+
+ if gap.node == nil {
+ return idMapGapIterator{}
+ }
+
+ gap.index++
+ for gap.index <= gap.node.nrSegments {
+ if gap.node.hasChildren {
+ if largeEnoughGap := gap.node.children[gap.index].searchFirstLargeEnoughGap(minSize); largeEnoughGap.Ok() {
+ return largeEnoughGap
+ }
+ } else {
+ if gap.Range().Length() >= minSize {
+ return gap
+ }
+ }
+ gap.index++
+ }
+ gap.node, gap.index = gap.node.parent, gap.node.parentIndex
+ if gap.node != nil && gap.index == gap.node.nrSegments {
+
+ gap.node, gap.index = gap.node.parent, gap.node.parentIndex
+ }
+ return gap.nextLargeEnoughGapHelper(minSize)
+}
+
+// PrevLargeEnoughGap returns the iterated gap's first prev gap with larger or
+// equal length than minSize. If not found, return a terminal gap iterator
+// (does NOT include this gap itself).
+//
+// Precondition: trackGaps must be 1.
+func (gap idMapGapIterator) PrevLargeEnoughGap(minSize uint32) idMapGapIterator {
+ if idMaptrackGaps != 1 {
+ panic("set is not tracking gaps")
+ }
+ if gap.node != nil && gap.node.hasChildren && gap.index == 0 {
+
+ gap.node = gap.PrevSegment().node
+ gap.index = gap.node.nrSegments
+ return gap.prevLargeEnoughGapHelper(minSize)
+ }
+ return gap.prevLargeEnoughGapHelper(minSize)
+}
+
+// prevLargeEnoughGapHelper is the helper function used by PrevLargeEnoughGap
+// to do the real recursions.
+//
+// Preconditions: gap is NOT the first gap of a non-leaf node.
+func (gap idMapGapIterator) prevLargeEnoughGapHelper(minSize uint32) idMapGapIterator {
+
+ for gap.node != nil &&
+ (gap.node.maxGap.Get() < minSize || (!gap.node.hasChildren && gap.index == 0)) {
+ gap.node, gap.index = gap.node.parent, gap.node.parentIndex
+ }
+
+ if gap.node == nil {
+ return idMapGapIterator{}
+ }
+
+ gap.index--
+ for gap.index >= 0 {
+ if gap.node.hasChildren {
+ if largeEnoughGap := gap.node.children[gap.index].searchLastLargeEnoughGap(minSize); largeEnoughGap.Ok() {
+ return largeEnoughGap
+ }
+ } else {
+ if gap.Range().Length() >= minSize {
+ return gap
+ }
+ }
+ gap.index--
+ }
+ gap.node, gap.index = gap.node.parent, gap.node.parentIndex
+ if gap.node != nil && gap.index == 0 {
+
+ gap.node, gap.index = gap.node.parent, gap.node.parentIndex
+ }
+ return gap.prevLargeEnoughGapHelper(minSize)
+}
+
+// segmentBeforePosition returns the predecessor segment of the position given
+// by n.children[i], which may or may not contain a child. If no such segment
+// exists, segmentBeforePosition returns a terminal iterator.
+func idMapsegmentBeforePosition(n *idMapnode, i int) idMapIterator {
+ for i == 0 {
+ if n.parent == nil {
+ return idMapIterator{}
+ }
+ n, i = n.parent, n.parentIndex
+ }
+ return idMapIterator{n, i - 1}
+}
+
+// segmentAfterPosition returns the successor segment of the position given by
+// n.children[i], which may or may not contain a child. If no such segment
+// exists, segmentAfterPosition returns a terminal iterator.
+func idMapsegmentAfterPosition(n *idMapnode, i int) idMapIterator {
+ for i == n.nrSegments {
+ if n.parent == nil {
+ return idMapIterator{}
+ }
+ n, i = n.parent, n.parentIndex
+ }
+ return idMapIterator{n, i}
+}
+
+func idMapzeroValueSlice(slice []uint32) {
+
+ for i := range slice {
+ idMapFunctions{}.ClearValue(&slice[i])
+ }
+}
+
+func idMapzeroNodeSlice(slice []*idMapnode) {
+ for i := range slice {
+ slice[i] = nil
+ }
+}
+
+// String stringifies a Set for debugging.
+func (s *idMapSet) String() string {
+ return s.root.String()
+}
+
+// String stringifies a node (and all of its children) for debugging.
+func (n *idMapnode) String() string {
+ var buf bytes.Buffer
+ n.writeDebugString(&buf, "")
+ return buf.String()
+}
+
+func (n *idMapnode) writeDebugString(buf *bytes.Buffer, prefix string) {
+ if n.hasChildren != (n.nrSegments > 0 && n.children[0] != nil) {
+ buf.WriteString(prefix)
+ buf.WriteString(fmt.Sprintf("WARNING: inconsistent value of hasChildren: got %v, want %v\n", n.hasChildren, !n.hasChildren))
+ }
+ for i := 0; i < n.nrSegments; i++ {
+ if child := n.children[i]; child != nil {
+ cprefix := fmt.Sprintf("%s- % 3d ", prefix, i)
+ if child.parent != n || child.parentIndex != i {
+ buf.WriteString(cprefix)
+ buf.WriteString(fmt.Sprintf("WARNING: inconsistent linkage to parent: got (%p, %d), want (%p, %d)\n", child.parent, child.parentIndex, n, i))
+ }
+ child.writeDebugString(buf, fmt.Sprintf("%s- % 3d ", prefix, i))
+ }
+ buf.WriteString(prefix)
+ if n.hasChildren {
+ if idMaptrackGaps != 0 {
+ buf.WriteString(fmt.Sprintf("- % 3d: %v => %v, maxGap: %d\n", i, n.keys[i], n.values[i], n.maxGap.Get()))
+ } else {
+ buf.WriteString(fmt.Sprintf("- % 3d: %v => %v\n", i, n.keys[i], n.values[i]))
+ }
+ } else {
+ buf.WriteString(fmt.Sprintf("- % 3d: %v => %v\n", i, n.keys[i], n.values[i]))
+ }
+ }
+ if child := n.children[n.nrSegments]; child != nil {
+ child.writeDebugString(buf, fmt.Sprintf("%s- % 3d ", prefix, n.nrSegments))
+ }
+}
+
+// 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 idMapSegmentDataSlices struct {
+ Start []uint32
+ End []uint32
+ Values []uint32
+}
+
+// ExportSortedSlice returns a copy of all segments in the given set, in ascending
+// key order.
+func (s *idMapSet) ExportSortedSlices() *idMapSegmentDataSlices {
+ var sds idMapSegmentDataSlices
+ for seg := s.FirstSegment(); seg.Ok(); seg = seg.NextSegment() {
+ sds.Start = append(sds.Start, seg.Start())
+ sds.End = append(sds.End, seg.End())
+ sds.Values = append(sds.Values, seg.Value())
+ }
+ sds.Start = sds.Start[:len(sds.Start):len(sds.Start)]
+ sds.End = sds.End[:len(sds.End):len(sds.End)]
+ sds.Values = sds.Values[:len(sds.Values):len(sds.Values)]
+ return &sds
+}
+
+// ImportSortedSlice initializes the given set from the given slice.
+//
+// Preconditions:
+// * s must be empty.
+// * sds must represent a valid set (the segments in sds must have valid
+// lengths that do not overlap).
+// * The segments in sds must be sorted in ascending key order.
+func (s *idMapSet) ImportSortedSlices(sds *idMapSegmentDataSlices) error {
+ if !s.IsEmpty() {
+ return fmt.Errorf("cannot import into non-empty set %v", s)
+ }
+ gap := s.FirstGap()
+ for i := range sds.Start {
+ r := idMapRange{sds.Start[i], sds.End[i]}
+ if !gap.Range().IsSupersetOf(r) {
+ return fmt.Errorf("segment overlaps a preceding segment or is incorrectly sorted: [%d, %d) => %v", sds.Start[i], sds.End[i], sds.Values[i])
+ }
+ gap = s.InsertWithoutMerging(gap, r, sds.Values[i]).NextGap()
+ }
+ return nil
+}
+
+// segmentTestCheck returns an error if s is incorrectly sorted, does not
+// contain exactly expectedSegments segments, or contains a segment which
+// fails the passed check.
+//
+// This should be used only for testing, and has been added to this package for
+// templating convenience.
+func (s *idMapSet) segmentTestCheck(expectedSegments int, segFunc func(int, idMapRange, uint32) error) error {
+ havePrev := false
+ prev := uint32(0)
+ nrSegments := 0
+ for seg := s.FirstSegment(); seg.Ok(); seg = seg.NextSegment() {
+ next := seg.Start()
+ if havePrev && prev >= next {
+ return fmt.Errorf("incorrect order: key %d (segment %d) >= key %d (segment %d)", prev, nrSegments-1, next, nrSegments)
+ }
+ if segFunc != nil {
+ if err := segFunc(nrSegments, seg.Range(), seg.Value()); err != nil {
+ return err
+ }
+ }
+ prev = next
+ havePrev = true
+ nrSegments++
+ }
+ if nrSegments != expectedSegments {
+ return fmt.Errorf("incorrect number of segments: got %d, wanted %d", nrSegments, expectedSegments)
+ }
+ return nil
+}
+
+// countSegments counts the number of segments in the set.
+//
+// Similar to Check, this should only be used for testing.
+func (s *idMapSet) countSegments() (segments int) {
+ for seg := s.FirstSegment(); seg.Ok(); seg = seg.NextSegment() {
+ segments++
+ }
+ return segments
+}
+func (s *idMapSet) saveRoot() *idMapSegmentDataSlices {
+ return s.ExportSortedSlices()
+}
+
+func (s *idMapSet) loadRoot(sds *idMapSegmentDataSlices) {
+ if err := s.ImportSortedSlices(sds); err != nil {
+ panic(err)
+ }
+}
diff --git a/pkg/sentry/kernel/contexttest/BUILD b/pkg/sentry/kernel/contexttest/BUILD
deleted file mode 100644
index 9d26392c0..000000000
--- a/pkg/sentry/kernel/contexttest/BUILD
+++ /dev/null
@@ -1,17 +0,0 @@
-load("//tools:defs.bzl", "go_library")
-
-package(licenses = ["notice"])
-
-go_library(
- name = "contexttest",
- testonly = 1,
- srcs = ["contexttest.go"],
- visibility = ["//pkg/sentry:internal"],
- deps = [
- "//pkg/context",
- "//pkg/sentry/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
deleted file mode 100644
index 22c340e56..000000000
--- a/pkg/sentry/kernel/contexttest/contexttest.go
+++ /dev/null
@@ -1,40 +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 contexttest provides a test context.Context which includes
-// a dummy kernel pointing to a valid platform.
-package contexttest
-
-import (
- "testing"
-
- "gvisor.dev/gvisor/pkg/context"
- "gvisor.dev/gvisor/pkg/sentry/contexttest"
- "gvisor.dev/gvisor/pkg/sentry/kernel"
- "gvisor.dev/gvisor/pkg/sentry/pgalloc"
- "gvisor.dev/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),
- }
- k.SetMemoryFile(pgalloc.MemoryFileFromContext(ctx))
- ctx.(*contexttest.TestContext).RegisterValue(kernel.CtxKernel, k)
- return ctx
-}
diff --git a/pkg/sentry/kernel/epoll/BUILD b/pkg/sentry/kernel/epoll/BUILD
deleted file mode 100644
index 75eedd5a2..000000000
--- a/pkg/sentry/kernel/epoll/BUILD
+++ /dev/null
@@ -1,51 +0,0 @@
-load("//tools:defs.bzl", "go_library", "go_test")
-load("//tools/go_generics:defs.bzl", "go_template_instance")
-
-package(licenses = ["notice"])
-
-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",
- ],
- visibility = ["//pkg/sentry:internal"],
- deps = [
- "//pkg/abi/linux",
- "//pkg/context",
- "//pkg/refs",
- "//pkg/sentry/fs",
- "//pkg/sentry/fs/anon",
- "//pkg/sentry/fs/fsutil",
- "//pkg/sync",
- "//pkg/usermem",
- "//pkg/waiter",
- ],
-)
-
-go_test(
- name = "epoll_test",
- size = "small",
- srcs = [
- "epoll_test.go",
- ],
- library = ":epoll",
- deps = [
- "//pkg/sentry/contexttest",
- "//pkg/sentry/fs/filetest",
- "//pkg/waiter",
- ],
-)
diff --git a/pkg/sentry/kernel/epoll/epoll_list.go b/pkg/sentry/kernel/epoll/epoll_list.go
new file mode 100644
index 000000000..56d05c770
--- /dev/null
+++ b/pkg/sentry/kernel/epoll/epoll_list.go
@@ -0,0 +1,193 @@
+package epoll
+
+// 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 pollEntryElementMapper struct{}
+
+// linkerFor maps an Element to a Linker.
+//
+// This default implementation should be inlined.
+//
+//go:nosplit
+func (pollEntryElementMapper) linkerFor(elem *pollEntry) *pollEntry { 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.
+//
+// The zero value for List is an empty list ready to use.
+//
+// To iterate over a list (where l is a List):
+// for e := l.Front(); e != nil; e = e.Next() {
+// // do something with e.
+// }
+//
+// +stateify savable
+type pollEntryList struct {
+ head *pollEntry
+ tail *pollEntry
+}
+
+// Reset resets list l to the empty state.
+func (l *pollEntryList) Reset() {
+ l.head = nil
+ l.tail = nil
+}
+
+// Empty returns true iff the list is empty.
+func (l *pollEntryList) Empty() bool {
+ return l.head == nil
+}
+
+// Front returns the first element of list l or nil.
+func (l *pollEntryList) Front() *pollEntry {
+ return l.head
+}
+
+// Back returns the last element of list l or nil.
+func (l *pollEntryList) Back() *pollEntry {
+ return l.tail
+}
+
+// Len returns the number of elements in the list.
+//
+// NOTE: This is an O(n) operation.
+func (l *pollEntryList) Len() (count int) {
+ for e := l.Front(); e != nil; e = (pollEntryElementMapper{}.linkerFor(e)).Next() {
+ count++
+ }
+ return count
+}
+
+// PushFront inserts the element e at the front of list l.
+func (l *pollEntryList) PushFront(e *pollEntry) {
+ linker := pollEntryElementMapper{}.linkerFor(e)
+ linker.SetNext(l.head)
+ linker.SetPrev(nil)
+ if l.head != nil {
+ pollEntryElementMapper{}.linkerFor(l.head).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+
+ l.head = e
+}
+
+// PushBack inserts the element e at the back of list l.
+func (l *pollEntryList) PushBack(e *pollEntry) {
+ linker := pollEntryElementMapper{}.linkerFor(e)
+ linker.SetNext(nil)
+ linker.SetPrev(l.tail)
+ if l.tail != nil {
+ pollEntryElementMapper{}.linkerFor(l.tail).SetNext(e)
+ } else {
+ l.head = e
+ }
+
+ l.tail = e
+}
+
+// PushBackList inserts list m at the end of list l, emptying m.
+func (l *pollEntryList) PushBackList(m *pollEntryList) {
+ if l.head == nil {
+ l.head = m.head
+ l.tail = m.tail
+ } else if m.head != nil {
+ pollEntryElementMapper{}.linkerFor(l.tail).SetNext(m.head)
+ pollEntryElementMapper{}.linkerFor(m.head).SetPrev(l.tail)
+
+ l.tail = m.tail
+ }
+ m.head = nil
+ m.tail = nil
+}
+
+// InsertAfter inserts e after b.
+func (l *pollEntryList) InsertAfter(b, e *pollEntry) {
+ bLinker := pollEntryElementMapper{}.linkerFor(b)
+ eLinker := pollEntryElementMapper{}.linkerFor(e)
+
+ a := bLinker.Next()
+
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ bLinker.SetNext(e)
+
+ if a != nil {
+ pollEntryElementMapper{}.linkerFor(a).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+}
+
+// InsertBefore inserts e before a.
+func (l *pollEntryList) InsertBefore(a, e *pollEntry) {
+ aLinker := pollEntryElementMapper{}.linkerFor(a)
+ eLinker := pollEntryElementMapper{}.linkerFor(e)
+
+ b := aLinker.Prev()
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ aLinker.SetPrev(e)
+
+ if b != nil {
+ pollEntryElementMapper{}.linkerFor(b).SetNext(e)
+ } else {
+ l.head = e
+ }
+}
+
+// Remove removes e from l.
+func (l *pollEntryList) Remove(e *pollEntry) {
+ linker := pollEntryElementMapper{}.linkerFor(e)
+ prev := linker.Prev()
+ next := linker.Next()
+
+ if prev != nil {
+ pollEntryElementMapper{}.linkerFor(prev).SetNext(next)
+ } else if l.head == e {
+ l.head = next
+ }
+
+ if next != nil {
+ pollEntryElementMapper{}.linkerFor(next).SetPrev(prev)
+ } else if l.tail == e {
+ l.tail = prev
+ }
+
+ linker.SetNext(nil)
+ linker.SetPrev(nil)
+}
+
+// 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 pollEntryEntry struct {
+ next *pollEntry
+ prev *pollEntry
+}
+
+// Next returns the entry that follows e in the list.
+func (e *pollEntryEntry) Next() *pollEntry {
+ return e.next
+}
+
+// Prev returns the entry that precedes e in the list.
+func (e *pollEntryEntry) Prev() *pollEntry {
+ return e.prev
+}
+
+// SetNext assigns 'entry' as the entry that follows e in the list.
+func (e *pollEntryEntry) SetNext(elem *pollEntry) {
+ e.next = elem
+}
+
+// SetPrev assigns 'entry' as the entry that precedes e in the list.
+func (e *pollEntryEntry) SetPrev(elem *pollEntry) {
+ e.prev = elem
+}
diff --git a/pkg/sentry/kernel/epoll/epoll_state_autogen.go b/pkg/sentry/kernel/epoll/epoll_state_autogen.go
new file mode 100644
index 000000000..3b9d76751
--- /dev/null
+++ b/pkg/sentry/kernel/epoll/epoll_state_autogen.go
@@ -0,0 +1,182 @@
+// automatically generated by stateify.
+
+package epoll
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *FileIdentifier) StateTypeName() string {
+ return "pkg/sentry/kernel/epoll.FileIdentifier"
+}
+
+func (x *FileIdentifier) StateFields() []string {
+ return []string{
+ "File",
+ "Fd",
+ }
+}
+
+func (x *FileIdentifier) beforeSave() {}
+
+func (x *FileIdentifier) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.File)
+ m.Save(1, &x.Fd)
+}
+
+func (x *FileIdentifier) afterLoad() {}
+
+func (x *FileIdentifier) StateLoad(m state.Source) {
+ m.LoadWait(0, &x.File)
+ m.Load(1, &x.Fd)
+}
+
+func (x *pollEntry) StateTypeName() string {
+ return "pkg/sentry/kernel/epoll.pollEntry"
+}
+
+func (x *pollEntry) StateFields() []string {
+ return []string{
+ "pollEntryEntry",
+ "id",
+ "userData",
+ "mask",
+ "flags",
+ "epoll",
+ }
+}
+
+func (x *pollEntry) beforeSave() {}
+
+func (x *pollEntry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.pollEntryEntry)
+ m.Save(1, &x.id)
+ m.Save(2, &x.userData)
+ m.Save(3, &x.mask)
+ m.Save(4, &x.flags)
+ m.Save(5, &x.epoll)
+}
+
+func (x *pollEntry) StateLoad(m state.Source) {
+ m.Load(0, &x.pollEntryEntry)
+ m.LoadWait(1, &x.id)
+ m.Load(2, &x.userData)
+ m.Load(3, &x.mask)
+ m.Load(4, &x.flags)
+ m.Load(5, &x.epoll)
+ m.AfterLoad(x.afterLoad)
+}
+
+func (x *EventPoll) StateTypeName() string {
+ return "pkg/sentry/kernel/epoll.EventPoll"
+}
+
+func (x *EventPoll) StateFields() []string {
+ return []string{
+ "files",
+ "readyList",
+ "waitingList",
+ "disabledList",
+ }
+}
+
+func (x *EventPoll) beforeSave() {}
+
+func (x *EventPoll) StateSave(m state.Sink) {
+ x.beforeSave()
+ if !state.IsZeroValue(&x.FilePipeSeek) {
+ state.Failf("FilePipeSeek is %#v, expected zero", &x.FilePipeSeek)
+ }
+ if !state.IsZeroValue(&x.FileNotDirReaddir) {
+ state.Failf("FileNotDirReaddir is %#v, expected zero", &x.FileNotDirReaddir)
+ }
+ if !state.IsZeroValue(&x.FileNoFsync) {
+ state.Failf("FileNoFsync is %#v, expected zero", &x.FileNoFsync)
+ }
+ if !state.IsZeroValue(&x.FileNoopFlush) {
+ state.Failf("FileNoopFlush is %#v, expected zero", &x.FileNoopFlush)
+ }
+ if !state.IsZeroValue(&x.FileNoIoctl) {
+ state.Failf("FileNoIoctl is %#v, expected zero", &x.FileNoIoctl)
+ }
+ if !state.IsZeroValue(&x.FileNoMMap) {
+ state.Failf("FileNoMMap is %#v, expected zero", &x.FileNoMMap)
+ }
+ if !state.IsZeroValue(&x.Queue) {
+ state.Failf("Queue is %#v, expected zero", &x.Queue)
+ }
+ m.Save(0, &x.files)
+ m.Save(1, &x.readyList)
+ m.Save(2, &x.waitingList)
+ m.Save(3, &x.disabledList)
+}
+
+func (x *EventPoll) StateLoad(m state.Source) {
+ m.Load(0, &x.files)
+ m.Load(1, &x.readyList)
+ m.Load(2, &x.waitingList)
+ m.Load(3, &x.disabledList)
+ m.AfterLoad(x.afterLoad)
+}
+
+func (x *pollEntryList) StateTypeName() string {
+ return "pkg/sentry/kernel/epoll.pollEntryList"
+}
+
+func (x *pollEntryList) StateFields() []string {
+ return []string{
+ "head",
+ "tail",
+ }
+}
+
+func (x *pollEntryList) beforeSave() {}
+
+func (x *pollEntryList) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.head)
+ m.Save(1, &x.tail)
+}
+
+func (x *pollEntryList) afterLoad() {}
+
+func (x *pollEntryList) StateLoad(m state.Source) {
+ m.Load(0, &x.head)
+ m.Load(1, &x.tail)
+}
+
+func (x *pollEntryEntry) StateTypeName() string {
+ return "pkg/sentry/kernel/epoll.pollEntryEntry"
+}
+
+func (x *pollEntryEntry) StateFields() []string {
+ return []string{
+ "next",
+ "prev",
+ }
+}
+
+func (x *pollEntryEntry) beforeSave() {}
+
+func (x *pollEntryEntry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.next)
+ m.Save(1, &x.prev)
+}
+
+func (x *pollEntryEntry) afterLoad() {}
+
+func (x *pollEntryEntry) StateLoad(m state.Source) {
+ m.Load(0, &x.next)
+ m.Load(1, &x.prev)
+}
+
+func init() {
+ state.Register((*FileIdentifier)(nil))
+ state.Register((*pollEntry)(nil))
+ state.Register((*EventPoll)(nil))
+ state.Register((*pollEntryList)(nil))
+ state.Register((*pollEntryEntry)(nil))
+}
diff --git a/pkg/sentry/kernel/epoll/epoll_test.go b/pkg/sentry/kernel/epoll/epoll_test.go
deleted file mode 100644
index 55b505593..000000000
--- a/pkg/sentry/kernel/epoll/epoll_test.go
+++ /dev/null
@@ -1,55 +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 epoll
-
-import (
- "testing"
-
- "gvisor.dev/gvisor/pkg/sentry/contexttest"
- "gvisor.dev/gvisor/pkg/sentry/fs/filetest"
- "gvisor.dev/gvisor/pkg/waiter"
-)
-
-func TestFileDestroyed(t *testing.T) {
- f := filetest.NewTestFile(t)
- id := FileIdentifier{f, 12}
-
- ctx := contexttest.Context(t)
- efile := NewEventPoll(ctx)
- 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(ctx)
-
- 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
deleted file mode 100644
index 9983a32e5..000000000
--- a/pkg/sentry/kernel/eventfd/BUILD
+++ /dev/null
@@ -1,33 +0,0 @@
-load("//tools:defs.bzl", "go_library", "go_test")
-
-package(licenses = ["notice"])
-
-go_library(
- name = "eventfd",
- srcs = ["eventfd.go"],
- visibility = ["//pkg/sentry:internal"],
- deps = [
- "//pkg/abi/linux",
- "//pkg/context",
- "//pkg/fdnotifier",
- "//pkg/sentry/fs",
- "//pkg/sentry/fs/anon",
- "//pkg/sentry/fs/fsutil",
- "//pkg/sync",
- "//pkg/syserror",
- "//pkg/usermem",
- "//pkg/waiter",
- ],
-)
-
-go_test(
- name = "eventfd_test",
- size = "small",
- srcs = ["eventfd_test.go"],
- library = ":eventfd",
- deps = [
- "//pkg/sentry/contexttest",
- "//pkg/usermem",
- "//pkg/waiter",
- ],
-)
diff --git a/pkg/sentry/kernel/eventfd/eventfd_state_autogen.go b/pkg/sentry/kernel/eventfd/eventfd_state_autogen.go
new file mode 100644
index 000000000..72a285daa
--- /dev/null
+++ b/pkg/sentry/kernel/eventfd/eventfd_state_autogen.go
@@ -0,0 +1,43 @@
+// automatically generated by stateify.
+
+package eventfd
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *EventOperations) StateTypeName() string {
+ return "pkg/sentry/kernel/eventfd.EventOperations"
+}
+
+func (x *EventOperations) StateFields() []string {
+ return []string{
+ "val",
+ "semMode",
+ "hostfd",
+ }
+}
+
+func (x *EventOperations) beforeSave() {}
+
+func (x *EventOperations) StateSave(m state.Sink) {
+ x.beforeSave()
+ if !state.IsZeroValue(&x.wq) {
+ state.Failf("wq is %#v, expected zero", &x.wq)
+ }
+ m.Save(0, &x.val)
+ m.Save(1, &x.semMode)
+ m.Save(2, &x.hostfd)
+}
+
+func (x *EventOperations) afterLoad() {}
+
+func (x *EventOperations) StateLoad(m state.Source) {
+ m.Load(0, &x.val)
+ m.Load(1, &x.semMode)
+ m.Load(2, &x.hostfd)
+}
+
+func init() {
+ state.Register((*EventOperations)(nil))
+}
diff --git a/pkg/sentry/kernel/eventfd/eventfd_test.go b/pkg/sentry/kernel/eventfd/eventfd_test.go
deleted file mode 100644
index 9b4892f74..000000000
--- a/pkg/sentry/kernel/eventfd/eventfd_test.go
+++ /dev/null
@@ -1,78 +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 eventfd
-
-import (
- "testing"
-
- "gvisor.dev/gvisor/pkg/sentry/contexttest"
- "gvisor.dev/gvisor/pkg/usermem"
- "gvisor.dev/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/fasync/BUILD b/pkg/sentry/kernel/fasync/BUILD
deleted file mode 100644
index 2b3955598..000000000
--- a/pkg/sentry/kernel/fasync/BUILD
+++ /dev/null
@@ -1,18 +0,0 @@
-load("//tools:defs.bzl", "go_library")
-
-package(licenses = ["notice"])
-
-go_library(
- name = "fasync",
- srcs = ["fasync.go"],
- visibility = ["//:sandbox"],
- deps = [
- "//pkg/abi/linux",
- "//pkg/sentry/fs",
- "//pkg/sentry/kernel",
- "//pkg/sentry/kernel/auth",
- "//pkg/sentry/vfs",
- "//pkg/sync",
- "//pkg/waiter",
- ],
-)
diff --git a/pkg/sentry/kernel/fasync/fasync_state_autogen.go b/pkg/sentry/kernel/fasync/fasync_state_autogen.go
new file mode 100644
index 000000000..8da4a1b44
--- /dev/null
+++ b/pkg/sentry/kernel/fasync/fasync_state_autogen.go
@@ -0,0 +1,49 @@
+// automatically generated by stateify.
+
+package fasync
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *FileAsync) StateTypeName() string {
+ return "pkg/sentry/kernel/fasync.FileAsync"
+}
+
+func (x *FileAsync) StateFields() []string {
+ return []string{
+ "e",
+ "requester",
+ "registered",
+ "recipientPG",
+ "recipientTG",
+ "recipientT",
+ }
+}
+
+func (x *FileAsync) beforeSave() {}
+
+func (x *FileAsync) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.e)
+ m.Save(1, &x.requester)
+ m.Save(2, &x.registered)
+ m.Save(3, &x.recipientPG)
+ m.Save(4, &x.recipientTG)
+ m.Save(5, &x.recipientT)
+}
+
+func (x *FileAsync) afterLoad() {}
+
+func (x *FileAsync) StateLoad(m state.Source) {
+ m.Load(0, &x.e)
+ m.Load(1, &x.requester)
+ m.Load(2, &x.registered)
+ m.Load(3, &x.recipientPG)
+ m.Load(4, &x.recipientTG)
+ m.Load(5, &x.recipientT)
+}
+
+func init() {
+ state.Register((*FileAsync)(nil))
+}
diff --git a/pkg/sentry/kernel/fd_table_refs.go b/pkg/sentry/kernel/fd_table_refs.go
new file mode 100644
index 000000000..a630289c9
--- /dev/null
+++ b/pkg/sentry/kernel/fd_table_refs.go
@@ -0,0 +1,117 @@
+package kernel
+
+import (
+ "fmt"
+ "gvisor.dev/gvisor/pkg/log"
+ refs_vfs1 "gvisor.dev/gvisor/pkg/refs"
+ "runtime"
+ "sync/atomic"
+)
+
+// ownerType is used to customize logging. Note that we use a pointer to T so
+// that we do not copy the entire object when passed as a format parameter.
+var FDTableownerType *FDTable
+
+// Refs implements refs.RefCounter. It keeps a reference count using atomic
+// operations and calls the destructor when the count reaches zero.
+//
+// Note that the number of references is actually refCount + 1 so that a default
+// zero-value Refs object contains one reference.
+//
+// TODO(gvisor.dev/issue/1486): Store stack traces when leak check is enabled in
+// a map with 16-bit hashes, and store the hash in the top 16 bits of refCount.
+// This will allow us to add stack trace information to the leak messages
+// without growing the size of Refs.
+//
+// +stateify savable
+type FDTableRefs struct {
+ // refCount is composed of two fields:
+ //
+ // [32-bit speculative references]:[32-bit real references]
+ //
+ // Speculative references are used for TryIncRef, to avoid a CompareAndSwap
+ // loop. See IncRef, DecRef and TryIncRef for details of how these fields are
+ // used.
+ refCount int64
+}
+
+func (r *FDTableRefs) finalize() {
+ var note string
+ switch refs_vfs1.GetLeakMode() {
+ case refs_vfs1.NoLeakChecking:
+ return
+ case refs_vfs1.UninitializedLeakChecking:
+ note = "(Leak checker uninitialized): "
+ }
+ if n := r.ReadRefs(); n != 0 {
+ log.Warningf("%sRefs %p owned by %T garbage collected with ref count of %d (want 0)", note, r, FDTableownerType, n)
+ }
+}
+
+// EnableLeakCheck checks for reference leaks when Refs gets garbage collected.
+func (r *FDTableRefs) EnableLeakCheck() {
+ if refs_vfs1.GetLeakMode() != refs_vfs1.NoLeakChecking {
+ runtime.SetFinalizer(r, (*FDTableRefs).finalize)
+ }
+}
+
+// ReadRefs returns the current number of references. The returned count is
+// inherently racy and is unsafe to use without external synchronization.
+func (r *FDTableRefs) ReadRefs() int64 {
+
+ return atomic.LoadInt64(&r.refCount) + 1
+}
+
+// IncRef implements refs.RefCounter.IncRef.
+//
+//go:nosplit
+func (r *FDTableRefs) IncRef() {
+ if v := atomic.AddInt64(&r.refCount, 1); v <= 0 {
+ panic(fmt.Sprintf("Incrementing non-positive ref count %p owned by %T", r, FDTableownerType))
+ }
+}
+
+// TryIncRef implements refs.RefCounter.TryIncRef.
+//
+// To do this safely without a loop, a speculative reference is first acquired
+// on the object. This allows multiple concurrent TryIncRef calls to distinguish
+// other TryIncRef calls from genuine references held.
+//
+//go:nosplit
+func (r *FDTableRefs) TryIncRef() bool {
+ const speculativeRef = 1 << 32
+ v := atomic.AddInt64(&r.refCount, speculativeRef)
+ if int32(v) < 0 {
+
+ atomic.AddInt64(&r.refCount, -speculativeRef)
+ return false
+ }
+
+ atomic.AddInt64(&r.refCount, -speculativeRef+1)
+ return true
+}
+
+// DecRef implements refs.RefCounter.DecRef.
+//
+// Note that speculative references are counted here. Since they were added
+// prior to real references reaching zero, they will successfully convert to
+// real references. In other words, we see speculative references only in the
+// following case:
+//
+// A: TryIncRef [speculative increase => sees non-negative references]
+// B: DecRef [real decrease]
+// A: TryIncRef [transform speculative to real]
+//
+//go:nosplit
+func (r *FDTableRefs) DecRef(destroy func()) {
+ switch v := atomic.AddInt64(&r.refCount, -1); {
+ case v < -1:
+ panic(fmt.Sprintf("Decrementing non-positive ref count %p, owned by %T", r, FDTableownerType))
+
+ case v == -1:
+
+ if destroy != nil {
+ destroy()
+ }
+ }
+}
diff --git a/pkg/sentry/kernel/fd_table_test.go b/pkg/sentry/kernel/fd_table_test.go
deleted file mode 100644
index bf5460083..000000000
--- a/pkg/sentry/kernel/fd_table_test.go
+++ /dev/null
@@ -1,228 +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 kernel
-
-import (
- "runtime"
- "testing"
-
- "gvisor.dev/gvisor/pkg/context"
- "gvisor.dev/gvisor/pkg/sentry/contexttest"
- "gvisor.dev/gvisor/pkg/sentry/fs"
- "gvisor.dev/gvisor/pkg/sentry/fs/filetest"
- "gvisor.dev/gvisor/pkg/sentry/limits"
- "gvisor.dev/gvisor/pkg/sync"
-)
-
-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 runTest(t testing.TB, fn func(ctx context.Context, fdTable *FDTable, file *fs.File, limitSet *limits.LimitSet)) {
- t.Helper() // Don't show in stacks.
-
- // Create the limits and context.
- limitSet := limits.NewLimitSet()
- limitSet.Set(limits.NumberOfFiles, limits.Limit{maxFD, maxFD}, true)
- ctx := contexttest.WithLimitSet(contexttest.Context(t), limitSet)
-
- // Create a test file.;
- file := filetest.NewTestFile(t)
-
- // Create the table.
- fdTable := new(FDTable)
- fdTable.init()
-
- // Run the test.
- fn(ctx, fdTable, file, limitSet)
-}
-
-// TestFDTableMany allocates maxFD FDs, i.e. maxes out the FDTable, 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 TestFDTableMany(t *testing.T) {
- runTest(t, func(ctx context.Context, fdTable *FDTable, file *fs.File, _ *limits.LimitSet) {
- for i := 0; i < maxFD; i++ {
- if _, err := fdTable.NewFDs(ctx, 0, []*fs.File{file}, FDFlags{}); err != nil {
- t.Fatalf("Allocated %v FDs but wanted to allocate %v", i, maxFD)
- }
- }
-
- if _, err := fdTable.NewFDs(ctx, 0, []*fs.File{file}, FDFlags{}); err == nil {
- t.Fatalf("fdTable.NewFDs(0, r) in full map: got nil, wanted error")
- }
-
- if err := fdTable.NewFDAt(ctx, 1, file, FDFlags{}); err != nil {
- t.Fatalf("fdTable.NewFDAt(1, r, FDFlags{}): got %v, wanted nil", err)
- }
-
- i := int32(2)
- fdTable.Remove(ctx, i)
- if fds, err := fdTable.NewFDs(ctx, 0, []*fs.File{file}, FDFlags{}); err != nil || fds[0] != i {
- t.Fatalf("Allocated %v FDs but wanted to allocate %v: %v", i, maxFD, err)
- }
- })
-}
-
-func TestFDTableOverLimit(t *testing.T) {
- runTest(t, func(ctx context.Context, fdTable *FDTable, file *fs.File, _ *limits.LimitSet) {
- if _, err := fdTable.NewFDs(ctx, maxFD, []*fs.File{file}, FDFlags{}); err == nil {
- t.Fatalf("fdTable.NewFDs(maxFD, f): got nil, wanted error")
- }
-
- if _, err := fdTable.NewFDs(ctx, maxFD-2, []*fs.File{file, file, file}, FDFlags{}); err == nil {
- t.Fatalf("fdTable.NewFDs(maxFD-2, {f,f,f}): got nil, wanted error")
- }
-
- if fds, err := fdTable.NewFDs(ctx, maxFD-3, []*fs.File{file, file, file}, FDFlags{}); err != nil {
- t.Fatalf("fdTable.NewFDs(maxFD-3, {f,f,f}): got %v, wanted nil", err)
- } else {
- for _, fd := range fds {
- fdTable.Remove(ctx, fd)
- }
- }
-
- if fds, err := fdTable.NewFDs(ctx, maxFD-1, []*fs.File{file}, FDFlags{}); err != nil || fds[0] != maxFD-1 {
- t.Fatalf("fdTable.NewFDAt(1, r, FDFlags{}): got %v, wanted nil", err)
- }
-
- if fds, err := fdTable.NewFDs(ctx, 0, []*fs.File{file}, FDFlags{}); err != nil {
- t.Fatalf("Adding an FD to a resized map: got %v, want nil", err)
- } else if len(fds) != 1 || fds[0] != 0 {
- t.Fatalf("Added an FD to a resized map: got %v, want {1}", fds)
- }
- })
-}
-
-// TestFDTable 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 TestFDTable(t *testing.T) {
- runTest(t, func(ctx context.Context, fdTable *FDTable, file *fs.File, limitSet *limits.LimitSet) {
- // Cap the limit at one.
- limitSet.Set(limits.NumberOfFiles, limits.Limit{1, maxFD}, true)
-
- if _, err := fdTable.NewFDs(ctx, 0, []*fs.File{file}, FDFlags{}); err != nil {
- t.Fatalf("Adding an FD to an empty 1-size map: got %v, want nil", err)
- }
-
- if _, err := fdTable.NewFDs(ctx, 0, []*fs.File{file}, FDFlags{}); err == nil {
- t.Fatalf("Adding an FD to a filled 1-size map: got nil, wanted an error")
- }
-
- // Remove the previous limit.
- limitSet.Set(limits.NumberOfFiles, limits.Limit{maxFD, maxFD}, true)
-
- if fds, err := fdTable.NewFDs(ctx, 0, []*fs.File{file}, FDFlags{}); err != nil {
- t.Fatalf("Adding an FD to a resized map: got %v, want nil", err)
- } else if len(fds) != 1 || fds[0] != 1 {
- t.Fatalf("Added an FD to a resized map: got %v, want {1}", fds)
- }
-
- if err := fdTable.NewFDAt(ctx, 1, file, FDFlags{}); err != nil {
- t.Fatalf("Replacing FD 1 via fdTable.NewFDAt(1, r, FDFlags{}): got %v, wanted nil", err)
- }
-
- if err := fdTable.NewFDAt(ctx, maxFD+1, file, FDFlags{}); err == nil {
- t.Fatalf("Using an FD that was too large via fdTable.NewFDAt(%v, r, FDFlags{}): got nil, wanted an error", maxFD+1)
- }
-
- if ref, _ := fdTable.Get(1); ref == nil {
- t.Fatalf("fdTable.Get(1): got nil, wanted %v", file)
- }
-
- if ref, _ := fdTable.Get(2); ref != nil {
- t.Fatalf("fdTable.Get(2): got a %v, wanted nil", ref)
- }
-
- ref, _ := fdTable.Remove(ctx, 1)
- if ref == nil {
- t.Fatalf("fdTable.Remove(1) for an existing FD: failed, want success")
- }
- ref.DecRef(ctx)
-
- if ref, _ := fdTable.Remove(ctx, 1); ref != nil {
- t.Fatalf("r.Remove(1) for a removed FD: got success, want failure")
- }
- })
-}
-
-func TestDescriptorFlags(t *testing.T) {
- runTest(t, func(ctx context.Context, fdTable *FDTable, file *fs.File, _ *limits.LimitSet) {
- if err := fdTable.NewFDAt(ctx, 2, file, FDFlags{CloseOnExec: true}); err != nil {
- t.Fatalf("fdTable.NewFDAt(2, r, FDFlags{}): got %v, wanted nil", err)
- }
-
- newFile, flags := fdTable.Get(2)
- if newFile == nil {
- t.Fatalf("fdTable.Get(2): got a %v, wanted nil", newFile)
- }
-
- if !flags.CloseOnExec {
- t.Fatalf("new File flags %v don't match original %d\n", flags, 0)
- }
- })
-}
-
-func BenchmarkFDLookupAndDecRef(b *testing.B) {
- b.StopTimer() // Setup.
-
- runTest(b, func(ctx context.Context, fdTable *FDTable, file *fs.File, _ *limits.LimitSet) {
- fds, err := fdTable.NewFDs(ctx, 0, []*fs.File{file, file, file, file, file}, FDFlags{})
- if err != nil {
- b.Fatalf("fdTable.NewFDs: got %v, wanted nil", err)
- }
-
- b.StartTimer() // Benchmark.
- for i := 0; i < b.N; i++ {
- tf, _ := fdTable.Get(fds[i%len(fds)])
- tf.DecRef(ctx)
- }
- })
-}
-
-func BenchmarkFDLookupAndDecRefConcurrent(b *testing.B) {
- b.StopTimer() // Setup.
-
- runTest(b, func(ctx context.Context, fdTable *FDTable, file *fs.File, _ *limits.LimitSet) {
- fds, err := fdTable.NewFDs(ctx, 0, []*fs.File{file, file, file, file, file}, FDFlags{})
- if err != nil {
- b.Fatalf("fdTable.NewFDs: got %v, wanted nil", err)
- }
-
- concurrency := runtime.GOMAXPROCS(0)
- if concurrency < 4 {
- concurrency = 4
- }
- each := b.N / concurrency
-
- b.StartTimer() // Benchmark.
- var wg sync.WaitGroup
- for i := 0; i < concurrency; i++ {
- wg.Add(1)
- go func() {
- defer wg.Done()
- for i := 0; i < each; i++ {
- tf, _ := fdTable.Get(fds[i%len(fds)])
- tf.DecRef(ctx)
- }
- }()
- }
- wg.Wait()
- })
-}
diff --git a/pkg/sentry/kernel/fs_context_refs.go b/pkg/sentry/kernel/fs_context_refs.go
new file mode 100644
index 000000000..e8bb1e6ee
--- /dev/null
+++ b/pkg/sentry/kernel/fs_context_refs.go
@@ -0,0 +1,117 @@
+package kernel
+
+import (
+ "fmt"
+ "gvisor.dev/gvisor/pkg/log"
+ refs_vfs1 "gvisor.dev/gvisor/pkg/refs"
+ "runtime"
+ "sync/atomic"
+)
+
+// ownerType is used to customize logging. Note that we use a pointer to T so
+// that we do not copy the entire object when passed as a format parameter.
+var FSContextownerType *FSContext
+
+// Refs implements refs.RefCounter. It keeps a reference count using atomic
+// operations and calls the destructor when the count reaches zero.
+//
+// Note that the number of references is actually refCount + 1 so that a default
+// zero-value Refs object contains one reference.
+//
+// TODO(gvisor.dev/issue/1486): Store stack traces when leak check is enabled in
+// a map with 16-bit hashes, and store the hash in the top 16 bits of refCount.
+// This will allow us to add stack trace information to the leak messages
+// without growing the size of Refs.
+//
+// +stateify savable
+type FSContextRefs struct {
+ // refCount is composed of two fields:
+ //
+ // [32-bit speculative references]:[32-bit real references]
+ //
+ // Speculative references are used for TryIncRef, to avoid a CompareAndSwap
+ // loop. See IncRef, DecRef and TryIncRef for details of how these fields are
+ // used.
+ refCount int64
+}
+
+func (r *FSContextRefs) finalize() {
+ var note string
+ switch refs_vfs1.GetLeakMode() {
+ case refs_vfs1.NoLeakChecking:
+ return
+ case refs_vfs1.UninitializedLeakChecking:
+ note = "(Leak checker uninitialized): "
+ }
+ if n := r.ReadRefs(); n != 0 {
+ log.Warningf("%sRefs %p owned by %T garbage collected with ref count of %d (want 0)", note, r, FSContextownerType, n)
+ }
+}
+
+// EnableLeakCheck checks for reference leaks when Refs gets garbage collected.
+func (r *FSContextRefs) EnableLeakCheck() {
+ if refs_vfs1.GetLeakMode() != refs_vfs1.NoLeakChecking {
+ runtime.SetFinalizer(r, (*FSContextRefs).finalize)
+ }
+}
+
+// ReadRefs returns the current number of references. The returned count is
+// inherently racy and is unsafe to use without external synchronization.
+func (r *FSContextRefs) ReadRefs() int64 {
+
+ return atomic.LoadInt64(&r.refCount) + 1
+}
+
+// IncRef implements refs.RefCounter.IncRef.
+//
+//go:nosplit
+func (r *FSContextRefs) IncRef() {
+ if v := atomic.AddInt64(&r.refCount, 1); v <= 0 {
+ panic(fmt.Sprintf("Incrementing non-positive ref count %p owned by %T", r, FSContextownerType))
+ }
+}
+
+// TryIncRef implements refs.RefCounter.TryIncRef.
+//
+// To do this safely without a loop, a speculative reference is first acquired
+// on the object. This allows multiple concurrent TryIncRef calls to distinguish
+// other TryIncRef calls from genuine references held.
+//
+//go:nosplit
+func (r *FSContextRefs) TryIncRef() bool {
+ const speculativeRef = 1 << 32
+ v := atomic.AddInt64(&r.refCount, speculativeRef)
+ if int32(v) < 0 {
+
+ atomic.AddInt64(&r.refCount, -speculativeRef)
+ return false
+ }
+
+ atomic.AddInt64(&r.refCount, -speculativeRef+1)
+ return true
+}
+
+// DecRef implements refs.RefCounter.DecRef.
+//
+// Note that speculative references are counted here. Since they were added
+// prior to real references reaching zero, they will successfully convert to
+// real references. In other words, we see speculative references only in the
+// following case:
+//
+// A: TryIncRef [speculative increase => sees non-negative references]
+// B: DecRef [real decrease]
+// A: TryIncRef [transform speculative to real]
+//
+//go:nosplit
+func (r *FSContextRefs) DecRef(destroy func()) {
+ switch v := atomic.AddInt64(&r.refCount, -1); {
+ case v < -1:
+ panic(fmt.Sprintf("Decrementing non-positive ref count %p, owned by %T", r, FSContextownerType))
+
+ case v == -1:
+
+ if destroy != nil {
+ destroy()
+ }
+ }
+}
diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD
deleted file mode 100644
index daa2dae76..000000000
--- a/pkg/sentry/kernel/futex/BUILD
+++ /dev/null
@@ -1,58 +0,0 @@
-load("//tools:defs.bzl", "go_library", "go_test")
-load("//tools/go_generics:defs.bzl", "go_template_instance")
-
-package(licenses = ["notice"])
-
-go_template_instance(
- name = "atomicptr_bucket",
- out = "atomicptr_bucket_unsafe.go",
- package = "futex",
- suffix = "Bucket",
- template = "//pkg/sync:generic_atomicptr",
- types = {
- "Value": "bucket",
- },
-)
-
-go_template_instance(
- name = "waiter_list",
- out = "waiter_list.go",
- package = "futex",
- prefix = "waiter",
- template = "//pkg/ilist:generic_list",
- types = {
- "Element": "*Waiter",
- "Linker": "*Waiter",
- },
-)
-
-go_library(
- name = "futex",
- srcs = [
- "atomicptr_bucket_unsafe.go",
- "futex.go",
- "waiter_list.go",
- ],
- visibility = ["//pkg/sentry:internal"],
- deps = [
- "//pkg/abi/linux",
- "//pkg/context",
- "//pkg/log",
- "//pkg/sentry/memmap",
- "//pkg/sync",
- "//pkg/syserror",
- "//pkg/usermem",
- ],
-)
-
-go_test(
- name = "futex_test",
- size = "small",
- srcs = ["futex_test.go"],
- library = ":futex",
- deps = [
- "//pkg/context",
- "//pkg/sync",
- "//pkg/usermem",
- ],
-)
diff --git a/pkg/sentry/kernel/futex/atomicptr_bucket_unsafe.go b/pkg/sentry/kernel/futex/atomicptr_bucket_unsafe.go
new file mode 100644
index 000000000..d3fdf09b0
--- /dev/null
+++ b/pkg/sentry/kernel/futex/atomicptr_bucket_unsafe.go
@@ -0,0 +1,37 @@
+package futex
+
+import (
+ "sync/atomic"
+ "unsafe"
+)
+
+// 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.
+//
+// +stateify savable
+type AtomicPtrBucket struct {
+ ptr unsafe.Pointer `state:".(*bucket)"`
+}
+
+func (p *AtomicPtrBucket) savePtr() *bucket {
+ return p.Load()
+}
+
+func (p *AtomicPtrBucket) loadPtr(v *bucket) {
+ p.Store(v)
+}
+
+// Load returns the value set by the most recent Store. It returns nil if there
+// has been no previous call to Store.
+func (p *AtomicPtrBucket) Load() *bucket {
+ return (*bucket)(atomic.LoadPointer(&p.ptr))
+}
+
+// Store sets the value returned by Load to x.
+func (p *AtomicPtrBucket) Store(x *bucket) {
+ atomic.StorePointer(&p.ptr, (unsafe.Pointer)(x))
+}
diff --git a/pkg/sentry/kernel/futex/futex_state_autogen.go b/pkg/sentry/kernel/futex/futex_state_autogen.go
new file mode 100644
index 000000000..13e9a97d0
--- /dev/null
+++ b/pkg/sentry/kernel/futex/futex_state_autogen.go
@@ -0,0 +1,114 @@
+// automatically generated by stateify.
+
+package futex
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *bucket) StateTypeName() string {
+ return "pkg/sentry/kernel/futex.bucket"
+}
+
+func (x *bucket) StateFields() []string {
+ return []string{}
+}
+
+func (x *bucket) beforeSave() {}
+
+func (x *bucket) StateSave(m state.Sink) {
+ x.beforeSave()
+ if !state.IsZeroValue(&x.waiters) {
+ state.Failf("waiters is %#v, expected zero", &x.waiters)
+ }
+}
+
+func (x *bucket) afterLoad() {}
+
+func (x *bucket) StateLoad(m state.Source) {
+}
+
+func (x *Manager) StateTypeName() string {
+ return "pkg/sentry/kernel/futex.Manager"
+}
+
+func (x *Manager) StateFields() []string {
+ return []string{
+ "sharedBucket",
+ }
+}
+
+func (x *Manager) beforeSave() {}
+
+func (x *Manager) StateSave(m state.Sink) {
+ x.beforeSave()
+ if !state.IsZeroValue(&x.privateBuckets) {
+ state.Failf("privateBuckets is %#v, expected zero", &x.privateBuckets)
+ }
+ m.Save(0, &x.sharedBucket)
+}
+
+func (x *Manager) afterLoad() {}
+
+func (x *Manager) StateLoad(m state.Source) {
+ m.Load(0, &x.sharedBucket)
+}
+
+func (x *waiterList) StateTypeName() string {
+ return "pkg/sentry/kernel/futex.waiterList"
+}
+
+func (x *waiterList) StateFields() []string {
+ return []string{
+ "head",
+ "tail",
+ }
+}
+
+func (x *waiterList) beforeSave() {}
+
+func (x *waiterList) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.head)
+ m.Save(1, &x.tail)
+}
+
+func (x *waiterList) afterLoad() {}
+
+func (x *waiterList) StateLoad(m state.Source) {
+ m.Load(0, &x.head)
+ m.Load(1, &x.tail)
+}
+
+func (x *waiterEntry) StateTypeName() string {
+ return "pkg/sentry/kernel/futex.waiterEntry"
+}
+
+func (x *waiterEntry) StateFields() []string {
+ return []string{
+ "next",
+ "prev",
+ }
+}
+
+func (x *waiterEntry) beforeSave() {}
+
+func (x *waiterEntry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.next)
+ m.Save(1, &x.prev)
+}
+
+func (x *waiterEntry) afterLoad() {}
+
+func (x *waiterEntry) StateLoad(m state.Source) {
+ m.Load(0, &x.next)
+ m.Load(1, &x.prev)
+}
+
+func init() {
+ state.Register((*bucket)(nil))
+ state.Register((*Manager)(nil))
+ state.Register((*waiterList)(nil))
+ state.Register((*waiterEntry)(nil))
+}
diff --git a/pkg/sentry/kernel/futex/futex_test.go b/pkg/sentry/kernel/futex/futex_test.go
deleted file mode 100644
index d0128c548..000000000
--- a/pkg/sentry/kernel/futex/futex_test.go
+++ /dev/null
@@ -1,536 +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 futex
-
-import (
- "math"
- "runtime"
- "sync/atomic"
- "syscall"
- "testing"
- "unsafe"
-
- "gvisor.dev/gvisor/pkg/context"
- "gvisor.dev/gvisor/pkg/sync"
- "gvisor.dev/gvisor/pkg/usermem"
-)
-
-// 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 struct {
- context.Context
- data []byte
-}
-
-const sizeofInt32 = 4
-
-func newTestData(size uint) testData {
- return testData{
- data: make([]byte, size),
- }
-}
-
-func (t testData) SwapUint32(addr usermem.Addr, new uint32) (uint32, error) {
- val := atomic.SwapUint32((*uint32)(unsafe.Pointer(&t.data[addr])), new)
- return val, nil
-}
-
-func (t testData) CompareAndSwapUint32(addr usermem.Addr, old, new uint32) (uint32, error) {
- if atomic.CompareAndSwapUint32((*uint32)(unsafe.Pointer(&t.data[addr])), old, new) {
- return old, nil
- }
- return atomic.LoadUint32((*uint32)(unsafe.Pointer(&t.data[addr]))), nil
-}
-
-func (t testData) LoadUint32(addr usermem.Addr) (uint32, error) {
- return atomic.LoadUint32((*uint32)(unsafe.Pointer(&t.data[addr]))), nil
-}
-
-func (t testData) GetSharedKey(addr usermem.Addr) (Key, error) {
- return Key{
- Kind: KindSharedMappable,
- Offset: uint64(addr),
- }, nil
-}
-
-func futexKind(private bool) string {
- if private {
- return "private"
- }
- return "shared"
-}
-
-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, ta, addr, private, val, bitmask); err != nil {
- t.Fatalf("WaitPrepare failed: %v", err)
- }
- return w
-}
-
-func TestFutexWake(t *testing.T) {
- 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, d)
-
- // 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)
- }
-
- // Expect the waiter to have been woken.
- if !w.woken() {
- t.Error("waiter not woken")
- }
- })
- }
-}
-
-func TestFutexWakeBitmask(t *testing.T) {
- 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, d)
-
- // 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)
- }
-
- // Expect the waiter to still be waiting.
- if w.woken() {
- t.Error("waiter woken unexpectedly")
- }
-
- // 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)
- }
-
- // Expect that the waiter was woken.
- if !w.woken() {
- t.Error("waiter not woken")
- }
- })
- }
-}
-
-func TestFutexWakeTwo(t *testing.T) {
- 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], d)
- }
-
- // 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)
- }
-
- // 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) {
- 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, d)
- w2 := newPreparedTestWaiter(t, m, d, 1*sizeofInt32, private, 0, ^uint32(0))
- defer m.WaitComplete(w2, d)
-
- // 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)
- }
-
- // 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) {
- 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) {
- 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, d)
- w2 := newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0))
- defer m.WaitComplete(w2, d)
-
- // 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)
- }
-
- // 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) {
- 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, d)
- w2 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0))
- defer m.WaitComplete(w2, d)
-
- // 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)
- }
-
- // 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) {
- 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, d)
- w2 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0))
- defer m.WaitComplete(w2, d)
-
- // 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)
- }
-
- // 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) {
- 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, d)
- w2 := newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0))
- defer m.WaitComplete(w2, d)
-
- // Add two waiters on address sizeofInt32.
- w3 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0))
- defer m.WaitComplete(w3, d)
- w4 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0))
- defer m.WaitComplete(w4, d)
-
- // 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)
- }
-
- // 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")
- }
- })
- }
-}
-
-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, d)
- w2 := newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0))
- defer m.WaitComplete(w2, d)
-
- // Add two waiters on address sizeofInt32.
- w3 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0))
- defer m.WaitComplete(w3, d)
- w4 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0))
- defer m.WaitComplete(w4, d)
-
- // 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)
- }
-
- // 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")
- }
- })
- }
-}
-
-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], d)
- }
-
- // 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)
- }
-
- // 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 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], d)
- }
-
- // 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)
- }
-
- // 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)
- }
- })
- }
-}
-
-const (
- testMutexSize = sizeofInt32
- testMutexLocked uint32 = 1
- testMutexUnlocked uint32 = 0
-)
-
-// 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 usermem.Addr
- d testData
- m *Manager
-}
-
-func newTestMutex(addr usermem.Addr, 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.data[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, 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, t.d)
- }
-}
-
-// 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.data[t.a])), testMutexUnlocked)
-
- // Notify all waiters.
- t.m.Wake(t.d, t.a, true, ^uint32(0), math.MaxInt32)
-}
-
-// 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 TestMutexStress(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
- }
-}
diff --git a/pkg/sentry/kernel/futex/futex_unsafe_state_autogen.go b/pkg/sentry/kernel/futex/futex_unsafe_state_autogen.go
new file mode 100644
index 000000000..244cdb502
--- /dev/null
+++ b/pkg/sentry/kernel/futex/futex_unsafe_state_autogen.go
@@ -0,0 +1,35 @@
+// automatically generated by stateify.
+
+package futex
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *AtomicPtrBucket) StateTypeName() string {
+ return "pkg/sentry/kernel/futex.AtomicPtrBucket"
+}
+
+func (x *AtomicPtrBucket) StateFields() []string {
+ return []string{
+ "ptr",
+ }
+}
+
+func (x *AtomicPtrBucket) beforeSave() {}
+
+func (x *AtomicPtrBucket) StateSave(m state.Sink) {
+ x.beforeSave()
+ var ptr *bucket = x.savePtr()
+ m.SaveValue(0, ptr)
+}
+
+func (x *AtomicPtrBucket) afterLoad() {}
+
+func (x *AtomicPtrBucket) StateLoad(m state.Source) {
+ m.LoadValue(0, new(*bucket), func(y interface{}) { x.loadPtr(y.(*bucket)) })
+}
+
+func init() {
+ state.Register((*AtomicPtrBucket)(nil))
+}
diff --git a/pkg/sentry/kernel/futex/waiter_list.go b/pkg/sentry/kernel/futex/waiter_list.go
new file mode 100644
index 000000000..ab6997b26
--- /dev/null
+++ b/pkg/sentry/kernel/futex/waiter_list.go
@@ -0,0 +1,193 @@
+package futex
+
+// 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 waiterElementMapper struct{}
+
+// linkerFor maps an Element to a Linker.
+//
+// This default implementation should be inlined.
+//
+//go:nosplit
+func (waiterElementMapper) linkerFor(elem *Waiter) *Waiter { 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.
+//
+// The zero value for List is an empty list ready to use.
+//
+// To iterate over a list (where l is a List):
+// for e := l.Front(); e != nil; e = e.Next() {
+// // do something with e.
+// }
+//
+// +stateify savable
+type waiterList struct {
+ head *Waiter
+ tail *Waiter
+}
+
+// Reset resets list l to the empty state.
+func (l *waiterList) Reset() {
+ l.head = nil
+ l.tail = nil
+}
+
+// Empty returns true iff the list is empty.
+func (l *waiterList) Empty() bool {
+ return l.head == nil
+}
+
+// Front returns the first element of list l or nil.
+func (l *waiterList) Front() *Waiter {
+ return l.head
+}
+
+// Back returns the last element of list l or nil.
+func (l *waiterList) Back() *Waiter {
+ return l.tail
+}
+
+// Len returns the number of elements in the list.
+//
+// NOTE: This is an O(n) operation.
+func (l *waiterList) Len() (count int) {
+ for e := l.Front(); e != nil; e = (waiterElementMapper{}.linkerFor(e)).Next() {
+ count++
+ }
+ return count
+}
+
+// PushFront inserts the element e at the front of list l.
+func (l *waiterList) PushFront(e *Waiter) {
+ linker := waiterElementMapper{}.linkerFor(e)
+ linker.SetNext(l.head)
+ linker.SetPrev(nil)
+ if l.head != nil {
+ waiterElementMapper{}.linkerFor(l.head).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+
+ l.head = e
+}
+
+// PushBack inserts the element e at the back of list l.
+func (l *waiterList) PushBack(e *Waiter) {
+ linker := waiterElementMapper{}.linkerFor(e)
+ linker.SetNext(nil)
+ linker.SetPrev(l.tail)
+ if l.tail != nil {
+ waiterElementMapper{}.linkerFor(l.tail).SetNext(e)
+ } else {
+ l.head = e
+ }
+
+ l.tail = e
+}
+
+// PushBackList inserts list m at the end of list l, emptying m.
+func (l *waiterList) PushBackList(m *waiterList) {
+ if l.head == nil {
+ l.head = m.head
+ l.tail = m.tail
+ } else if m.head != nil {
+ waiterElementMapper{}.linkerFor(l.tail).SetNext(m.head)
+ waiterElementMapper{}.linkerFor(m.head).SetPrev(l.tail)
+
+ l.tail = m.tail
+ }
+ m.head = nil
+ m.tail = nil
+}
+
+// InsertAfter inserts e after b.
+func (l *waiterList) InsertAfter(b, e *Waiter) {
+ bLinker := waiterElementMapper{}.linkerFor(b)
+ eLinker := waiterElementMapper{}.linkerFor(e)
+
+ a := bLinker.Next()
+
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ bLinker.SetNext(e)
+
+ if a != nil {
+ waiterElementMapper{}.linkerFor(a).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+}
+
+// InsertBefore inserts e before a.
+func (l *waiterList) InsertBefore(a, e *Waiter) {
+ aLinker := waiterElementMapper{}.linkerFor(a)
+ eLinker := waiterElementMapper{}.linkerFor(e)
+
+ b := aLinker.Prev()
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ aLinker.SetPrev(e)
+
+ if b != nil {
+ waiterElementMapper{}.linkerFor(b).SetNext(e)
+ } else {
+ l.head = e
+ }
+}
+
+// Remove removes e from l.
+func (l *waiterList) Remove(e *Waiter) {
+ linker := waiterElementMapper{}.linkerFor(e)
+ prev := linker.Prev()
+ next := linker.Next()
+
+ if prev != nil {
+ waiterElementMapper{}.linkerFor(prev).SetNext(next)
+ } else if l.head == e {
+ l.head = next
+ }
+
+ if next != nil {
+ waiterElementMapper{}.linkerFor(next).SetPrev(prev)
+ } else if l.tail == e {
+ l.tail = prev
+ }
+
+ linker.SetNext(nil)
+ linker.SetPrev(nil)
+}
+
+// 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 waiterEntry struct {
+ next *Waiter
+ prev *Waiter
+}
+
+// Next returns the entry that follows e in the list.
+func (e *waiterEntry) Next() *Waiter {
+ return e.next
+}
+
+// Prev returns the entry that precedes e in the list.
+func (e *waiterEntry) Prev() *Waiter {
+ return e.prev
+}
+
+// SetNext assigns 'entry' as the entry that follows e in the list.
+func (e *waiterEntry) SetNext(elem *Waiter) {
+ e.next = elem
+}
+
+// SetPrev assigns 'entry' as the entry that precedes e in the list.
+func (e *waiterEntry) SetPrev(elem *Waiter) {
+ e.prev = elem
+}
diff --git a/pkg/sentry/kernel/g3doc/run_states.dot b/pkg/sentry/kernel/g3doc/run_states.dot
deleted file mode 100644
index 7861fe1f5..000000000
--- a/pkg/sentry/kernel/g3doc/run_states.dot
+++ /dev/null
@@ -1,99 +0,0 @@
-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/g3doc/run_states.png b/pkg/sentry/kernel/g3doc/run_states.png
deleted file mode 100644
index b63b60f02..000000000
--- a/pkg/sentry/kernel/g3doc/run_states.png
+++ /dev/null
Binary files differ
diff --git a/pkg/sentry/kernel/kernel_amd64_state_autogen.go b/pkg/sentry/kernel/kernel_amd64_state_autogen.go
new file mode 100644
index 000000000..12de47ad0
--- /dev/null
+++ b/pkg/sentry/kernel/kernel_amd64_state_autogen.go
@@ -0,0 +1,5 @@
+// automatically generated by stateify.
+
+// +build amd64
+
+package kernel
diff --git a/pkg/sentry/kernel/kernel_arm64_state_autogen.go b/pkg/sentry/kernel/kernel_arm64_state_autogen.go
new file mode 100644
index 000000000..3c040d283
--- /dev/null
+++ b/pkg/sentry/kernel/kernel_arm64_state_autogen.go
@@ -0,0 +1,5 @@
+// automatically generated by stateify.
+
+// +build arm64
+
+package kernel
diff --git a/pkg/sentry/kernel/kernel_opts_state_autogen.go b/pkg/sentry/kernel/kernel_opts_state_autogen.go
new file mode 100644
index 000000000..a01604159
--- /dev/null
+++ b/pkg/sentry/kernel/kernel_opts_state_autogen.go
@@ -0,0 +1,30 @@
+// automatically generated by stateify.
+
+package kernel
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *SpecialOpts) StateTypeName() string {
+ return "pkg/sentry/kernel.SpecialOpts"
+}
+
+func (x *SpecialOpts) StateFields() []string {
+ return []string{}
+}
+
+func (x *SpecialOpts) beforeSave() {}
+
+func (x *SpecialOpts) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *SpecialOpts) afterLoad() {}
+
+func (x *SpecialOpts) StateLoad(m state.Source) {
+}
+
+func init() {
+ state.Register((*SpecialOpts)(nil))
+}
diff --git a/pkg/sentry/kernel/kernel_state_autogen.go b/pkg/sentry/kernel/kernel_state_autogen.go
new file mode 100644
index 000000000..d0ff135d7
--- /dev/null
+++ b/pkg/sentry/kernel/kernel_state_autogen.go
@@ -0,0 +1,2326 @@
+// automatically generated by stateify.
+
+package kernel
+
+import (
+ "gvisor.dev/gvisor/pkg/bpf"
+ "gvisor.dev/gvisor/pkg/sentry/device"
+ "gvisor.dev/gvisor/pkg/state"
+ "gvisor.dev/gvisor/pkg/tcpip"
+)
+
+func (x *abstractEndpoint) StateTypeName() string {
+ return "pkg/sentry/kernel.abstractEndpoint"
+}
+
+func (x *abstractEndpoint) StateFields() []string {
+ return []string{
+ "ep",
+ "socket",
+ "name",
+ "ns",
+ }
+}
+
+func (x *abstractEndpoint) beforeSave() {}
+
+func (x *abstractEndpoint) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.ep)
+ m.Save(1, &x.socket)
+ m.Save(2, &x.name)
+ m.Save(3, &x.ns)
+}
+
+func (x *abstractEndpoint) afterLoad() {}
+
+func (x *abstractEndpoint) StateLoad(m state.Source) {
+ m.Load(0, &x.ep)
+ m.Load(1, &x.socket)
+ m.Load(2, &x.name)
+ m.Load(3, &x.ns)
+}
+
+func (x *AbstractSocketNamespace) StateTypeName() string {
+ return "pkg/sentry/kernel.AbstractSocketNamespace"
+}
+
+func (x *AbstractSocketNamespace) StateFields() []string {
+ return []string{
+ "endpoints",
+ }
+}
+
+func (x *AbstractSocketNamespace) beforeSave() {}
+
+func (x *AbstractSocketNamespace) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.endpoints)
+}
+
+func (x *AbstractSocketNamespace) afterLoad() {}
+
+func (x *AbstractSocketNamespace) StateLoad(m state.Source) {
+ m.Load(0, &x.endpoints)
+}
+
+func (x *FDFlags) StateTypeName() string {
+ return "pkg/sentry/kernel.FDFlags"
+}
+
+func (x *FDFlags) StateFields() []string {
+ return []string{
+ "CloseOnExec",
+ }
+}
+
+func (x *FDFlags) beforeSave() {}
+
+func (x *FDFlags) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.CloseOnExec)
+}
+
+func (x *FDFlags) afterLoad() {}
+
+func (x *FDFlags) StateLoad(m state.Source) {
+ m.Load(0, &x.CloseOnExec)
+}
+
+func (x *descriptor) StateTypeName() string {
+ return "pkg/sentry/kernel.descriptor"
+}
+
+func (x *descriptor) StateFields() []string {
+ return []string{
+ "file",
+ "fileVFS2",
+ "flags",
+ }
+}
+
+func (x *descriptor) beforeSave() {}
+
+func (x *descriptor) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.file)
+ m.Save(1, &x.fileVFS2)
+ m.Save(2, &x.flags)
+}
+
+func (x *descriptor) afterLoad() {}
+
+func (x *descriptor) StateLoad(m state.Source) {
+ m.Load(0, &x.file)
+ m.Load(1, &x.fileVFS2)
+ m.Load(2, &x.flags)
+}
+
+func (x *FDTable) StateTypeName() string {
+ return "pkg/sentry/kernel.FDTable"
+}
+
+func (x *FDTable) StateFields() []string {
+ return []string{
+ "FDTableRefs",
+ "k",
+ "next",
+ "used",
+ "descriptorTable",
+ }
+}
+
+func (x *FDTable) beforeSave() {}
+
+func (x *FDTable) StateSave(m state.Sink) {
+ x.beforeSave()
+ var descriptorTable map[int32]descriptor = x.saveDescriptorTable()
+ m.SaveValue(4, descriptorTable)
+ m.Save(0, &x.FDTableRefs)
+ m.Save(1, &x.k)
+ m.Save(2, &x.next)
+ m.Save(3, &x.used)
+}
+
+func (x *FDTable) afterLoad() {}
+
+func (x *FDTable) StateLoad(m state.Source) {
+ m.Load(0, &x.FDTableRefs)
+ m.Load(1, &x.k)
+ m.Load(2, &x.next)
+ m.Load(3, &x.used)
+ m.LoadValue(4, new(map[int32]descriptor), func(y interface{}) { x.loadDescriptorTable(y.(map[int32]descriptor)) })
+}
+
+func (x *FDTableRefs) StateTypeName() string {
+ return "pkg/sentry/kernel.FDTableRefs"
+}
+
+func (x *FDTableRefs) StateFields() []string {
+ return []string{
+ "refCount",
+ }
+}
+
+func (x *FDTableRefs) beforeSave() {}
+
+func (x *FDTableRefs) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.refCount)
+}
+
+func (x *FDTableRefs) afterLoad() {}
+
+func (x *FDTableRefs) StateLoad(m state.Source) {
+ m.Load(0, &x.refCount)
+}
+
+func (x *FSContext) StateTypeName() string {
+ return "pkg/sentry/kernel.FSContext"
+}
+
+func (x *FSContext) StateFields() []string {
+ return []string{
+ "FSContextRefs",
+ "root",
+ "rootVFS2",
+ "cwd",
+ "cwdVFS2",
+ "umask",
+ }
+}
+
+func (x *FSContext) beforeSave() {}
+
+func (x *FSContext) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.FSContextRefs)
+ m.Save(1, &x.root)
+ m.Save(2, &x.rootVFS2)
+ m.Save(3, &x.cwd)
+ m.Save(4, &x.cwdVFS2)
+ m.Save(5, &x.umask)
+}
+
+func (x *FSContext) afterLoad() {}
+
+func (x *FSContext) StateLoad(m state.Source) {
+ m.Load(0, &x.FSContextRefs)
+ m.Load(1, &x.root)
+ m.Load(2, &x.rootVFS2)
+ m.Load(3, &x.cwd)
+ m.Load(4, &x.cwdVFS2)
+ m.Load(5, &x.umask)
+}
+
+func (x *FSContextRefs) StateTypeName() string {
+ return "pkg/sentry/kernel.FSContextRefs"
+}
+
+func (x *FSContextRefs) StateFields() []string {
+ return []string{
+ "refCount",
+ }
+}
+
+func (x *FSContextRefs) beforeSave() {}
+
+func (x *FSContextRefs) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.refCount)
+}
+
+func (x *FSContextRefs) afterLoad() {}
+
+func (x *FSContextRefs) StateLoad(m state.Source) {
+ m.Load(0, &x.refCount)
+}
+
+func (x *IPCNamespace) StateTypeName() string {
+ return "pkg/sentry/kernel.IPCNamespace"
+}
+
+func (x *IPCNamespace) StateFields() []string {
+ return []string{
+ "userNS",
+ "semaphores",
+ "shms",
+ }
+}
+
+func (x *IPCNamespace) beforeSave() {}
+
+func (x *IPCNamespace) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.userNS)
+ m.Save(1, &x.semaphores)
+ m.Save(2, &x.shms)
+}
+
+func (x *IPCNamespace) afterLoad() {}
+
+func (x *IPCNamespace) StateLoad(m state.Source) {
+ m.Load(0, &x.userNS)
+ m.Load(1, &x.semaphores)
+ m.Load(2, &x.shms)
+}
+
+func (x *Kernel) StateTypeName() string {
+ return "pkg/sentry/kernel.Kernel"
+}
+
+func (x *Kernel) StateFields() []string {
+ return []string{
+ "featureSet",
+ "timekeeper",
+ "tasks",
+ "rootUserNamespace",
+ "rootNetworkNamespace",
+ "applicationCores",
+ "useHostCores",
+ "extraAuxv",
+ "vdso",
+ "rootUTSNamespace",
+ "rootIPCNamespace",
+ "rootAbstractSocketNamespace",
+ "futexes",
+ "globalInit",
+ "realtimeClock",
+ "monotonicClock",
+ "syslog",
+ "runningTasks",
+ "cpuClock",
+ "cpuClockTickerDisabled",
+ "cpuClockTickerSetting",
+ "uniqueID",
+ "nextInotifyCookie",
+ "netlinkPorts",
+ "danglingEndpoints",
+ "sockets",
+ "nextSocketEntry",
+ "deviceRegistry",
+ "DirentCacheLimiter",
+ "SpecialOpts",
+ "vfs",
+ "hostMount",
+ "pipeMount",
+ "shmMount",
+ "socketMount",
+ "SleepForAddressSpaceActivation",
+ }
+}
+
+func (x *Kernel) beforeSave() {}
+
+func (x *Kernel) StateSave(m state.Sink) {
+ x.beforeSave()
+ var danglingEndpoints []tcpip.Endpoint = x.saveDanglingEndpoints()
+ m.SaveValue(24, danglingEndpoints)
+ var deviceRegistry *device.Registry = x.saveDeviceRegistry()
+ m.SaveValue(27, deviceRegistry)
+ m.Save(0, &x.featureSet)
+ m.Save(1, &x.timekeeper)
+ m.Save(2, &x.tasks)
+ m.Save(3, &x.rootUserNamespace)
+ m.Save(4, &x.rootNetworkNamespace)
+ m.Save(5, &x.applicationCores)
+ m.Save(6, &x.useHostCores)
+ m.Save(7, &x.extraAuxv)
+ m.Save(8, &x.vdso)
+ m.Save(9, &x.rootUTSNamespace)
+ m.Save(10, &x.rootIPCNamespace)
+ m.Save(11, &x.rootAbstractSocketNamespace)
+ m.Save(12, &x.futexes)
+ m.Save(13, &x.globalInit)
+ m.Save(14, &x.realtimeClock)
+ m.Save(15, &x.monotonicClock)
+ m.Save(16, &x.syslog)
+ m.Save(17, &x.runningTasks)
+ m.Save(18, &x.cpuClock)
+ m.Save(19, &x.cpuClockTickerDisabled)
+ m.Save(20, &x.cpuClockTickerSetting)
+ m.Save(21, &x.uniqueID)
+ m.Save(22, &x.nextInotifyCookie)
+ m.Save(23, &x.netlinkPorts)
+ m.Save(25, &x.sockets)
+ m.Save(26, &x.nextSocketEntry)
+ m.Save(28, &x.DirentCacheLimiter)
+ m.Save(29, &x.SpecialOpts)
+ m.Save(30, &x.vfs)
+ m.Save(31, &x.hostMount)
+ m.Save(32, &x.pipeMount)
+ m.Save(33, &x.shmMount)
+ m.Save(34, &x.socketMount)
+ m.Save(35, &x.SleepForAddressSpaceActivation)
+}
+
+func (x *Kernel) afterLoad() {}
+
+func (x *Kernel) StateLoad(m state.Source) {
+ m.Load(0, &x.featureSet)
+ m.Load(1, &x.timekeeper)
+ m.Load(2, &x.tasks)
+ m.Load(3, &x.rootUserNamespace)
+ m.Load(4, &x.rootNetworkNamespace)
+ m.Load(5, &x.applicationCores)
+ m.Load(6, &x.useHostCores)
+ m.Load(7, &x.extraAuxv)
+ m.Load(8, &x.vdso)
+ m.Load(9, &x.rootUTSNamespace)
+ m.Load(10, &x.rootIPCNamespace)
+ m.Load(11, &x.rootAbstractSocketNamespace)
+ m.Load(12, &x.futexes)
+ m.Load(13, &x.globalInit)
+ m.Load(14, &x.realtimeClock)
+ m.Load(15, &x.monotonicClock)
+ m.Load(16, &x.syslog)
+ m.Load(17, &x.runningTasks)
+ m.Load(18, &x.cpuClock)
+ m.Load(19, &x.cpuClockTickerDisabled)
+ m.Load(20, &x.cpuClockTickerSetting)
+ m.Load(21, &x.uniqueID)
+ m.Load(22, &x.nextInotifyCookie)
+ m.Load(23, &x.netlinkPorts)
+ m.Load(25, &x.sockets)
+ m.Load(26, &x.nextSocketEntry)
+ m.Load(28, &x.DirentCacheLimiter)
+ m.Load(29, &x.SpecialOpts)
+ m.Load(30, &x.vfs)
+ m.Load(31, &x.hostMount)
+ m.Load(32, &x.pipeMount)
+ m.Load(33, &x.shmMount)
+ m.Load(34, &x.socketMount)
+ m.Load(35, &x.SleepForAddressSpaceActivation)
+ m.LoadValue(24, new([]tcpip.Endpoint), func(y interface{}) { x.loadDanglingEndpoints(y.([]tcpip.Endpoint)) })
+ m.LoadValue(27, new(*device.Registry), func(y interface{}) { x.loadDeviceRegistry(y.(*device.Registry)) })
+}
+
+func (x *SocketEntry) StateTypeName() string {
+ return "pkg/sentry/kernel.SocketEntry"
+}
+
+func (x *SocketEntry) StateFields() []string {
+ return []string{
+ "socketEntry",
+ "k",
+ "Sock",
+ "SockVFS2",
+ "ID",
+ }
+}
+
+func (x *SocketEntry) beforeSave() {}
+
+func (x *SocketEntry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.socketEntry)
+ m.Save(1, &x.k)
+ m.Save(2, &x.Sock)
+ m.Save(3, &x.SockVFS2)
+ m.Save(4, &x.ID)
+}
+
+func (x *SocketEntry) afterLoad() {}
+
+func (x *SocketEntry) StateLoad(m state.Source) {
+ m.Load(0, &x.socketEntry)
+ m.Load(1, &x.k)
+ m.Load(2, &x.Sock)
+ m.Load(3, &x.SockVFS2)
+ m.Load(4, &x.ID)
+}
+
+func (x *pendingSignals) StateTypeName() string {
+ return "pkg/sentry/kernel.pendingSignals"
+}
+
+func (x *pendingSignals) StateFields() []string {
+ return []string{
+ "signals",
+ }
+}
+
+func (x *pendingSignals) beforeSave() {}
+
+func (x *pendingSignals) StateSave(m state.Sink) {
+ x.beforeSave()
+ var signals []savedPendingSignal = x.saveSignals()
+ m.SaveValue(0, signals)
+}
+
+func (x *pendingSignals) afterLoad() {}
+
+func (x *pendingSignals) StateLoad(m state.Source) {
+ m.LoadValue(0, new([]savedPendingSignal), func(y interface{}) { x.loadSignals(y.([]savedPendingSignal)) })
+}
+
+func (x *pendingSignalQueue) StateTypeName() string {
+ return "pkg/sentry/kernel.pendingSignalQueue"
+}
+
+func (x *pendingSignalQueue) StateFields() []string {
+ return []string{
+ "pendingSignalList",
+ "length",
+ }
+}
+
+func (x *pendingSignalQueue) beforeSave() {}
+
+func (x *pendingSignalQueue) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.pendingSignalList)
+ m.Save(1, &x.length)
+}
+
+func (x *pendingSignalQueue) afterLoad() {}
+
+func (x *pendingSignalQueue) StateLoad(m state.Source) {
+ m.Load(0, &x.pendingSignalList)
+ m.Load(1, &x.length)
+}
+
+func (x *pendingSignal) StateTypeName() string {
+ return "pkg/sentry/kernel.pendingSignal"
+}
+
+func (x *pendingSignal) StateFields() []string {
+ return []string{
+ "pendingSignalEntry",
+ "SignalInfo",
+ "timer",
+ }
+}
+
+func (x *pendingSignal) beforeSave() {}
+
+func (x *pendingSignal) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.pendingSignalEntry)
+ m.Save(1, &x.SignalInfo)
+ m.Save(2, &x.timer)
+}
+
+func (x *pendingSignal) afterLoad() {}
+
+func (x *pendingSignal) StateLoad(m state.Source) {
+ m.Load(0, &x.pendingSignalEntry)
+ m.Load(1, &x.SignalInfo)
+ m.Load(2, &x.timer)
+}
+
+func (x *pendingSignalList) StateTypeName() string {
+ return "pkg/sentry/kernel.pendingSignalList"
+}
+
+func (x *pendingSignalList) StateFields() []string {
+ return []string{
+ "head",
+ "tail",
+ }
+}
+
+func (x *pendingSignalList) beforeSave() {}
+
+func (x *pendingSignalList) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.head)
+ m.Save(1, &x.tail)
+}
+
+func (x *pendingSignalList) afterLoad() {}
+
+func (x *pendingSignalList) StateLoad(m state.Source) {
+ m.Load(0, &x.head)
+ m.Load(1, &x.tail)
+}
+
+func (x *pendingSignalEntry) StateTypeName() string {
+ return "pkg/sentry/kernel.pendingSignalEntry"
+}
+
+func (x *pendingSignalEntry) StateFields() []string {
+ return []string{
+ "next",
+ "prev",
+ }
+}
+
+func (x *pendingSignalEntry) beforeSave() {}
+
+func (x *pendingSignalEntry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.next)
+ m.Save(1, &x.prev)
+}
+
+func (x *pendingSignalEntry) afterLoad() {}
+
+func (x *pendingSignalEntry) StateLoad(m state.Source) {
+ m.Load(0, &x.next)
+ m.Load(1, &x.prev)
+}
+
+func (x *savedPendingSignal) StateTypeName() string {
+ return "pkg/sentry/kernel.savedPendingSignal"
+}
+
+func (x *savedPendingSignal) StateFields() []string {
+ return []string{
+ "si",
+ "timer",
+ }
+}
+
+func (x *savedPendingSignal) beforeSave() {}
+
+func (x *savedPendingSignal) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.si)
+ m.Save(1, &x.timer)
+}
+
+func (x *savedPendingSignal) afterLoad() {}
+
+func (x *savedPendingSignal) StateLoad(m state.Source) {
+ m.Load(0, &x.si)
+ m.Load(1, &x.timer)
+}
+
+func (x *IntervalTimer) StateTypeName() string {
+ return "pkg/sentry/kernel.IntervalTimer"
+}
+
+func (x *IntervalTimer) StateFields() []string {
+ return []string{
+ "timer",
+ "target",
+ "signo",
+ "id",
+ "sigval",
+ "group",
+ "sigpending",
+ "sigorphan",
+ "overrunCur",
+ "overrunLast",
+ }
+}
+
+func (x *IntervalTimer) beforeSave() {}
+
+func (x *IntervalTimer) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.timer)
+ m.Save(1, &x.target)
+ m.Save(2, &x.signo)
+ m.Save(3, &x.id)
+ m.Save(4, &x.sigval)
+ m.Save(5, &x.group)
+ m.Save(6, &x.sigpending)
+ m.Save(7, &x.sigorphan)
+ m.Save(8, &x.overrunCur)
+ m.Save(9, &x.overrunLast)
+}
+
+func (x *IntervalTimer) afterLoad() {}
+
+func (x *IntervalTimer) StateLoad(m state.Source) {
+ m.Load(0, &x.timer)
+ m.Load(1, &x.target)
+ m.Load(2, &x.signo)
+ m.Load(3, &x.id)
+ m.Load(4, &x.sigval)
+ m.Load(5, &x.group)
+ m.Load(6, &x.sigpending)
+ m.Load(7, &x.sigorphan)
+ m.Load(8, &x.overrunCur)
+ m.Load(9, &x.overrunLast)
+}
+
+func (x *processGroupList) StateTypeName() string {
+ return "pkg/sentry/kernel.processGroupList"
+}
+
+func (x *processGroupList) StateFields() []string {
+ return []string{
+ "head",
+ "tail",
+ }
+}
+
+func (x *processGroupList) beforeSave() {}
+
+func (x *processGroupList) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.head)
+ m.Save(1, &x.tail)
+}
+
+func (x *processGroupList) afterLoad() {}
+
+func (x *processGroupList) StateLoad(m state.Source) {
+ m.Load(0, &x.head)
+ m.Load(1, &x.tail)
+}
+
+func (x *processGroupEntry) StateTypeName() string {
+ return "pkg/sentry/kernel.processGroupEntry"
+}
+
+func (x *processGroupEntry) StateFields() []string {
+ return []string{
+ "next",
+ "prev",
+ }
+}
+
+func (x *processGroupEntry) beforeSave() {}
+
+func (x *processGroupEntry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.next)
+ m.Save(1, &x.prev)
+}
+
+func (x *processGroupEntry) afterLoad() {}
+
+func (x *processGroupEntry) StateLoad(m state.Source) {
+ m.Load(0, &x.next)
+ m.Load(1, &x.prev)
+}
+
+func (x *ProcessGroupRefs) StateTypeName() string {
+ return "pkg/sentry/kernel.ProcessGroupRefs"
+}
+
+func (x *ProcessGroupRefs) StateFields() []string {
+ return []string{
+ "refCount",
+ }
+}
+
+func (x *ProcessGroupRefs) beforeSave() {}
+
+func (x *ProcessGroupRefs) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.refCount)
+}
+
+func (x *ProcessGroupRefs) afterLoad() {}
+
+func (x *ProcessGroupRefs) StateLoad(m state.Source) {
+ m.Load(0, &x.refCount)
+}
+
+func (x *ptraceOptions) StateTypeName() string {
+ return "pkg/sentry/kernel.ptraceOptions"
+}
+
+func (x *ptraceOptions) StateFields() []string {
+ return []string{
+ "ExitKill",
+ "SysGood",
+ "TraceClone",
+ "TraceExec",
+ "TraceExit",
+ "TraceFork",
+ "TraceSeccomp",
+ "TraceVfork",
+ "TraceVforkDone",
+ }
+}
+
+func (x *ptraceOptions) beforeSave() {}
+
+func (x *ptraceOptions) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.ExitKill)
+ m.Save(1, &x.SysGood)
+ m.Save(2, &x.TraceClone)
+ m.Save(3, &x.TraceExec)
+ m.Save(4, &x.TraceExit)
+ m.Save(5, &x.TraceFork)
+ m.Save(6, &x.TraceSeccomp)
+ m.Save(7, &x.TraceVfork)
+ m.Save(8, &x.TraceVforkDone)
+}
+
+func (x *ptraceOptions) afterLoad() {}
+
+func (x *ptraceOptions) StateLoad(m state.Source) {
+ m.Load(0, &x.ExitKill)
+ m.Load(1, &x.SysGood)
+ m.Load(2, &x.TraceClone)
+ m.Load(3, &x.TraceExec)
+ m.Load(4, &x.TraceExit)
+ m.Load(5, &x.TraceFork)
+ m.Load(6, &x.TraceSeccomp)
+ m.Load(7, &x.TraceVfork)
+ m.Load(8, &x.TraceVforkDone)
+}
+
+func (x *ptraceStop) StateTypeName() string {
+ return "pkg/sentry/kernel.ptraceStop"
+}
+
+func (x *ptraceStop) StateFields() []string {
+ return []string{
+ "frozen",
+ "listen",
+ }
+}
+
+func (x *ptraceStop) beforeSave() {}
+
+func (x *ptraceStop) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.frozen)
+ m.Save(1, &x.listen)
+}
+
+func (x *ptraceStop) afterLoad() {}
+
+func (x *ptraceStop) StateLoad(m state.Source) {
+ m.Load(0, &x.frozen)
+ m.Load(1, &x.listen)
+}
+
+func (x *OldRSeqCriticalRegion) StateTypeName() string {
+ return "pkg/sentry/kernel.OldRSeqCriticalRegion"
+}
+
+func (x *OldRSeqCriticalRegion) StateFields() []string {
+ return []string{
+ "CriticalSection",
+ "Restart",
+ }
+}
+
+func (x *OldRSeqCriticalRegion) beforeSave() {}
+
+func (x *OldRSeqCriticalRegion) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.CriticalSection)
+ m.Save(1, &x.Restart)
+}
+
+func (x *OldRSeqCriticalRegion) afterLoad() {}
+
+func (x *OldRSeqCriticalRegion) StateLoad(m state.Source) {
+ m.Load(0, &x.CriticalSection)
+ m.Load(1, &x.Restart)
+}
+
+func (x *sessionList) StateTypeName() string {
+ return "pkg/sentry/kernel.sessionList"
+}
+
+func (x *sessionList) StateFields() []string {
+ return []string{
+ "head",
+ "tail",
+ }
+}
+
+func (x *sessionList) beforeSave() {}
+
+func (x *sessionList) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.head)
+ m.Save(1, &x.tail)
+}
+
+func (x *sessionList) afterLoad() {}
+
+func (x *sessionList) StateLoad(m state.Source) {
+ m.Load(0, &x.head)
+ m.Load(1, &x.tail)
+}
+
+func (x *sessionEntry) StateTypeName() string {
+ return "pkg/sentry/kernel.sessionEntry"
+}
+
+func (x *sessionEntry) StateFields() []string {
+ return []string{
+ "next",
+ "prev",
+ }
+}
+
+func (x *sessionEntry) beforeSave() {}
+
+func (x *sessionEntry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.next)
+ m.Save(1, &x.prev)
+}
+
+func (x *sessionEntry) afterLoad() {}
+
+func (x *sessionEntry) StateLoad(m state.Source) {
+ m.Load(0, &x.next)
+ m.Load(1, &x.prev)
+}
+
+func (x *SessionRefs) StateTypeName() string {
+ return "pkg/sentry/kernel.SessionRefs"
+}
+
+func (x *SessionRefs) StateFields() []string {
+ return []string{
+ "refCount",
+ }
+}
+
+func (x *SessionRefs) beforeSave() {}
+
+func (x *SessionRefs) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.refCount)
+}
+
+func (x *SessionRefs) afterLoad() {}
+
+func (x *SessionRefs) StateLoad(m state.Source) {
+ m.Load(0, &x.refCount)
+}
+
+func (x *Session) StateTypeName() string {
+ return "pkg/sentry/kernel.Session"
+}
+
+func (x *Session) StateFields() []string {
+ return []string{
+ "SessionRefs",
+ "leader",
+ "id",
+ "foreground",
+ "processGroups",
+ "sessionEntry",
+ }
+}
+
+func (x *Session) beforeSave() {}
+
+func (x *Session) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.SessionRefs)
+ m.Save(1, &x.leader)
+ m.Save(2, &x.id)
+ m.Save(3, &x.foreground)
+ m.Save(4, &x.processGroups)
+ m.Save(5, &x.sessionEntry)
+}
+
+func (x *Session) afterLoad() {}
+
+func (x *Session) StateLoad(m state.Source) {
+ m.Load(0, &x.SessionRefs)
+ m.Load(1, &x.leader)
+ m.Load(2, &x.id)
+ m.Load(3, &x.foreground)
+ m.Load(4, &x.processGroups)
+ m.Load(5, &x.sessionEntry)
+}
+
+func (x *ProcessGroup) StateTypeName() string {
+ return "pkg/sentry/kernel.ProcessGroup"
+}
+
+func (x *ProcessGroup) StateFields() []string {
+ return []string{
+ "refs",
+ "originator",
+ "id",
+ "session",
+ "ancestors",
+ "processGroupEntry",
+ }
+}
+
+func (x *ProcessGroup) beforeSave() {}
+
+func (x *ProcessGroup) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.refs)
+ m.Save(1, &x.originator)
+ m.Save(2, &x.id)
+ m.Save(3, &x.session)
+ m.Save(4, &x.ancestors)
+ m.Save(5, &x.processGroupEntry)
+}
+
+func (x *ProcessGroup) afterLoad() {}
+
+func (x *ProcessGroup) StateLoad(m state.Source) {
+ m.Load(0, &x.refs)
+ m.Load(1, &x.originator)
+ m.Load(2, &x.id)
+ m.Load(3, &x.session)
+ m.Load(4, &x.ancestors)
+ m.Load(5, &x.processGroupEntry)
+}
+
+func (x *SignalHandlers) StateTypeName() string {
+ return "pkg/sentry/kernel.SignalHandlers"
+}
+
+func (x *SignalHandlers) StateFields() []string {
+ return []string{
+ "actions",
+ }
+}
+
+func (x *SignalHandlers) beforeSave() {}
+
+func (x *SignalHandlers) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.actions)
+}
+
+func (x *SignalHandlers) afterLoad() {}
+
+func (x *SignalHandlers) StateLoad(m state.Source) {
+ m.Load(0, &x.actions)
+}
+
+func (x *socketList) StateTypeName() string {
+ return "pkg/sentry/kernel.socketList"
+}
+
+func (x *socketList) StateFields() []string {
+ return []string{
+ "head",
+ "tail",
+ }
+}
+
+func (x *socketList) beforeSave() {}
+
+func (x *socketList) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.head)
+ m.Save(1, &x.tail)
+}
+
+func (x *socketList) afterLoad() {}
+
+func (x *socketList) StateLoad(m state.Source) {
+ m.Load(0, &x.head)
+ m.Load(1, &x.tail)
+}
+
+func (x *socketEntry) StateTypeName() string {
+ return "pkg/sentry/kernel.socketEntry"
+}
+
+func (x *socketEntry) StateFields() []string {
+ return []string{
+ "next",
+ "prev",
+ }
+}
+
+func (x *socketEntry) beforeSave() {}
+
+func (x *socketEntry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.next)
+ m.Save(1, &x.prev)
+}
+
+func (x *socketEntry) afterLoad() {}
+
+func (x *socketEntry) StateLoad(m state.Source) {
+ m.Load(0, &x.next)
+ m.Load(1, &x.prev)
+}
+
+func (x *syscallTableInfo) StateTypeName() string {
+ return "pkg/sentry/kernel.syscallTableInfo"
+}
+
+func (x *syscallTableInfo) StateFields() []string {
+ return []string{
+ "OS",
+ "Arch",
+ }
+}
+
+func (x *syscallTableInfo) beforeSave() {}
+
+func (x *syscallTableInfo) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.OS)
+ m.Save(1, &x.Arch)
+}
+
+func (x *syscallTableInfo) afterLoad() {}
+
+func (x *syscallTableInfo) StateLoad(m state.Source) {
+ m.Load(0, &x.OS)
+ m.Load(1, &x.Arch)
+}
+
+func (x *syslog) StateTypeName() string {
+ return "pkg/sentry/kernel.syslog"
+}
+
+func (x *syslog) StateFields() []string {
+ return []string{
+ "msg",
+ }
+}
+
+func (x *syslog) beforeSave() {}
+
+func (x *syslog) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.msg)
+}
+
+func (x *syslog) afterLoad() {}
+
+func (x *syslog) StateLoad(m state.Source) {
+ m.Load(0, &x.msg)
+}
+
+func (x *Task) StateTypeName() string {
+ return "pkg/sentry/kernel.Task"
+}
+
+func (x *Task) StateFields() []string {
+ return []string{
+ "taskNode",
+ "runState",
+ "taskWorkCount",
+ "taskWork",
+ "haveSyscallReturn",
+ "gosched",
+ "yieldCount",
+ "pendingSignals",
+ "signalMask",
+ "realSignalMask",
+ "haveSavedSignalMask",
+ "savedSignalMask",
+ "signalStack",
+ "groupStopPending",
+ "groupStopAcknowledged",
+ "trapStopPending",
+ "trapNotifyPending",
+ "stop",
+ "exitStatus",
+ "syscallRestartBlock",
+ "k",
+ "containerID",
+ "tc",
+ "fsContext",
+ "fdTable",
+ "vforkParent",
+ "exitState",
+ "exitTracerNotified",
+ "exitTracerAcked",
+ "exitParentNotified",
+ "exitParentAcked",
+ "ptraceTracer",
+ "ptraceTracees",
+ "ptraceSeized",
+ "ptraceOpts",
+ "ptraceSyscallMode",
+ "ptraceSinglestep",
+ "ptraceCode",
+ "ptraceSiginfo",
+ "ptraceEventMsg",
+ "ioUsage",
+ "creds",
+ "utsns",
+ "ipcns",
+ "abstractSockets",
+ "mountNamespaceVFS2",
+ "parentDeathSignal",
+ "syscallFilters",
+ "cleartid",
+ "allowedCPUMask",
+ "cpu",
+ "niceness",
+ "numaPolicy",
+ "numaNodeMask",
+ "netns",
+ "rseqCPU",
+ "oldRSeqCPUAddr",
+ "rseqAddr",
+ "rseqSignature",
+ "robustList",
+ "startTime",
+ "kcov",
+ }
+}
+
+func (x *Task) beforeSave() {}
+
+func (x *Task) StateSave(m state.Sink) {
+ x.beforeSave()
+ if !state.IsZeroValue(&x.signalQueue) {
+ state.Failf("signalQueue is %#v, expected zero", &x.signalQueue)
+ }
+ var ptraceTracer *Task = x.savePtraceTracer()
+ m.SaveValue(31, ptraceTracer)
+ var syscallFilters []bpf.Program = x.saveSyscallFilters()
+ m.SaveValue(47, syscallFilters)
+ m.Save(0, &x.taskNode)
+ m.Save(1, &x.runState)
+ m.Save(2, &x.taskWorkCount)
+ m.Save(3, &x.taskWork)
+ m.Save(4, &x.haveSyscallReturn)
+ m.Save(5, &x.gosched)
+ m.Save(6, &x.yieldCount)
+ m.Save(7, &x.pendingSignals)
+ m.Save(8, &x.signalMask)
+ m.Save(9, &x.realSignalMask)
+ m.Save(10, &x.haveSavedSignalMask)
+ m.Save(11, &x.savedSignalMask)
+ m.Save(12, &x.signalStack)
+ m.Save(13, &x.groupStopPending)
+ m.Save(14, &x.groupStopAcknowledged)
+ m.Save(15, &x.trapStopPending)
+ m.Save(16, &x.trapNotifyPending)
+ m.Save(17, &x.stop)
+ m.Save(18, &x.exitStatus)
+ m.Save(19, &x.syscallRestartBlock)
+ m.Save(20, &x.k)
+ m.Save(21, &x.containerID)
+ m.Save(22, &x.tc)
+ m.Save(23, &x.fsContext)
+ m.Save(24, &x.fdTable)
+ m.Save(25, &x.vforkParent)
+ m.Save(26, &x.exitState)
+ m.Save(27, &x.exitTracerNotified)
+ m.Save(28, &x.exitTracerAcked)
+ m.Save(29, &x.exitParentNotified)
+ m.Save(30, &x.exitParentAcked)
+ m.Save(32, &x.ptraceTracees)
+ m.Save(33, &x.ptraceSeized)
+ m.Save(34, &x.ptraceOpts)
+ m.Save(35, &x.ptraceSyscallMode)
+ m.Save(36, &x.ptraceSinglestep)
+ m.Save(37, &x.ptraceCode)
+ m.Save(38, &x.ptraceSiginfo)
+ m.Save(39, &x.ptraceEventMsg)
+ m.Save(40, &x.ioUsage)
+ m.Save(41, &x.creds)
+ m.Save(42, &x.utsns)
+ m.Save(43, &x.ipcns)
+ m.Save(44, &x.abstractSockets)
+ m.Save(45, &x.mountNamespaceVFS2)
+ m.Save(46, &x.parentDeathSignal)
+ m.Save(48, &x.cleartid)
+ m.Save(49, &x.allowedCPUMask)
+ m.Save(50, &x.cpu)
+ m.Save(51, &x.niceness)
+ m.Save(52, &x.numaPolicy)
+ m.Save(53, &x.numaNodeMask)
+ m.Save(54, &x.netns)
+ m.Save(55, &x.rseqCPU)
+ m.Save(56, &x.oldRSeqCPUAddr)
+ m.Save(57, &x.rseqAddr)
+ m.Save(58, &x.rseqSignature)
+ m.Save(59, &x.robustList)
+ m.Save(60, &x.startTime)
+ m.Save(61, &x.kcov)
+}
+
+func (x *Task) StateLoad(m state.Source) {
+ m.Load(0, &x.taskNode)
+ m.Load(1, &x.runState)
+ m.Load(2, &x.taskWorkCount)
+ m.Load(3, &x.taskWork)
+ m.Load(4, &x.haveSyscallReturn)
+ m.Load(5, &x.gosched)
+ m.Load(6, &x.yieldCount)
+ m.Load(7, &x.pendingSignals)
+ m.Load(8, &x.signalMask)
+ m.Load(9, &x.realSignalMask)
+ m.Load(10, &x.haveSavedSignalMask)
+ m.Load(11, &x.savedSignalMask)
+ m.Load(12, &x.signalStack)
+ m.Load(13, &x.groupStopPending)
+ m.Load(14, &x.groupStopAcknowledged)
+ m.Load(15, &x.trapStopPending)
+ m.Load(16, &x.trapNotifyPending)
+ m.Load(17, &x.stop)
+ m.Load(18, &x.exitStatus)
+ m.Load(19, &x.syscallRestartBlock)
+ m.Load(20, &x.k)
+ m.Load(21, &x.containerID)
+ m.Load(22, &x.tc)
+ m.Load(23, &x.fsContext)
+ m.Load(24, &x.fdTable)
+ m.Load(25, &x.vforkParent)
+ m.Load(26, &x.exitState)
+ m.Load(27, &x.exitTracerNotified)
+ m.Load(28, &x.exitTracerAcked)
+ m.Load(29, &x.exitParentNotified)
+ m.Load(30, &x.exitParentAcked)
+ m.Load(32, &x.ptraceTracees)
+ m.Load(33, &x.ptraceSeized)
+ m.Load(34, &x.ptraceOpts)
+ m.Load(35, &x.ptraceSyscallMode)
+ m.Load(36, &x.ptraceSinglestep)
+ m.Load(37, &x.ptraceCode)
+ m.Load(38, &x.ptraceSiginfo)
+ m.Load(39, &x.ptraceEventMsg)
+ m.Load(40, &x.ioUsage)
+ m.Load(41, &x.creds)
+ m.Load(42, &x.utsns)
+ m.Load(43, &x.ipcns)
+ m.Load(44, &x.abstractSockets)
+ m.Load(45, &x.mountNamespaceVFS2)
+ m.Load(46, &x.parentDeathSignal)
+ m.Load(48, &x.cleartid)
+ m.Load(49, &x.allowedCPUMask)
+ m.Load(50, &x.cpu)
+ m.Load(51, &x.niceness)
+ m.Load(52, &x.numaPolicy)
+ m.Load(53, &x.numaNodeMask)
+ m.Load(54, &x.netns)
+ m.Load(55, &x.rseqCPU)
+ m.Load(56, &x.oldRSeqCPUAddr)
+ m.Load(57, &x.rseqAddr)
+ m.Load(58, &x.rseqSignature)
+ m.Load(59, &x.robustList)
+ m.Load(60, &x.startTime)
+ m.Load(61, &x.kcov)
+ m.LoadValue(31, new(*Task), func(y interface{}) { x.loadPtraceTracer(y.(*Task)) })
+ m.LoadValue(47, new([]bpf.Program), func(y interface{}) { x.loadSyscallFilters(y.([]bpf.Program)) })
+ m.AfterLoad(x.afterLoad)
+}
+
+func (x *runSyscallAfterPtraceEventClone) StateTypeName() string {
+ return "pkg/sentry/kernel.runSyscallAfterPtraceEventClone"
+}
+
+func (x *runSyscallAfterPtraceEventClone) StateFields() []string {
+ return []string{
+ "vforkChild",
+ "vforkChildTID",
+ }
+}
+
+func (x *runSyscallAfterPtraceEventClone) beforeSave() {}
+
+func (x *runSyscallAfterPtraceEventClone) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.vforkChild)
+ m.Save(1, &x.vforkChildTID)
+}
+
+func (x *runSyscallAfterPtraceEventClone) afterLoad() {}
+
+func (x *runSyscallAfterPtraceEventClone) StateLoad(m state.Source) {
+ m.Load(0, &x.vforkChild)
+ m.Load(1, &x.vforkChildTID)
+}
+
+func (x *runSyscallAfterVforkStop) StateTypeName() string {
+ return "pkg/sentry/kernel.runSyscallAfterVforkStop"
+}
+
+func (x *runSyscallAfterVforkStop) StateFields() []string {
+ return []string{
+ "childTID",
+ }
+}
+
+func (x *runSyscallAfterVforkStop) beforeSave() {}
+
+func (x *runSyscallAfterVforkStop) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.childTID)
+}
+
+func (x *runSyscallAfterVforkStop) afterLoad() {}
+
+func (x *runSyscallAfterVforkStop) StateLoad(m state.Source) {
+ m.Load(0, &x.childTID)
+}
+
+func (x *vforkStop) StateTypeName() string {
+ return "pkg/sentry/kernel.vforkStop"
+}
+
+func (x *vforkStop) StateFields() []string {
+ return []string{}
+}
+
+func (x *vforkStop) beforeSave() {}
+
+func (x *vforkStop) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *vforkStop) afterLoad() {}
+
+func (x *vforkStop) StateLoad(m state.Source) {
+}
+
+func (x *TaskContext) StateTypeName() string {
+ return "pkg/sentry/kernel.TaskContext"
+}
+
+func (x *TaskContext) StateFields() []string {
+ return []string{
+ "Name",
+ "Arch",
+ "MemoryManager",
+ "fu",
+ "st",
+ }
+}
+
+func (x *TaskContext) beforeSave() {}
+
+func (x *TaskContext) StateSave(m state.Sink) {
+ x.beforeSave()
+ var st syscallTableInfo = x.saveSt()
+ m.SaveValue(4, st)
+ m.Save(0, &x.Name)
+ m.Save(1, &x.Arch)
+ m.Save(2, &x.MemoryManager)
+ m.Save(3, &x.fu)
+}
+
+func (x *TaskContext) afterLoad() {}
+
+func (x *TaskContext) StateLoad(m state.Source) {
+ m.Load(0, &x.Name)
+ m.Load(1, &x.Arch)
+ m.Load(2, &x.MemoryManager)
+ m.Load(3, &x.fu)
+ m.LoadValue(4, new(syscallTableInfo), func(y interface{}) { x.loadSt(y.(syscallTableInfo)) })
+}
+
+func (x *execStop) StateTypeName() string {
+ return "pkg/sentry/kernel.execStop"
+}
+
+func (x *execStop) StateFields() []string {
+ return []string{}
+}
+
+func (x *execStop) beforeSave() {}
+
+func (x *execStop) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *execStop) afterLoad() {}
+
+func (x *execStop) StateLoad(m state.Source) {
+}
+
+func (x *runSyscallAfterExecStop) StateTypeName() string {
+ return "pkg/sentry/kernel.runSyscallAfterExecStop"
+}
+
+func (x *runSyscallAfterExecStop) StateFields() []string {
+ return []string{
+ "tc",
+ }
+}
+
+func (x *runSyscallAfterExecStop) beforeSave() {}
+
+func (x *runSyscallAfterExecStop) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.tc)
+}
+
+func (x *runSyscallAfterExecStop) afterLoad() {}
+
+func (x *runSyscallAfterExecStop) StateLoad(m state.Source) {
+ m.Load(0, &x.tc)
+}
+
+func (x *ExitStatus) StateTypeName() string {
+ return "pkg/sentry/kernel.ExitStatus"
+}
+
+func (x *ExitStatus) StateFields() []string {
+ return []string{
+ "Code",
+ "Signo",
+ }
+}
+
+func (x *ExitStatus) beforeSave() {}
+
+func (x *ExitStatus) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.Code)
+ m.Save(1, &x.Signo)
+}
+
+func (x *ExitStatus) afterLoad() {}
+
+func (x *ExitStatus) StateLoad(m state.Source) {
+ m.Load(0, &x.Code)
+ m.Load(1, &x.Signo)
+}
+
+func (x *runExit) StateTypeName() string {
+ return "pkg/sentry/kernel.runExit"
+}
+
+func (x *runExit) StateFields() []string {
+ return []string{}
+}
+
+func (x *runExit) beforeSave() {}
+
+func (x *runExit) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *runExit) afterLoad() {}
+
+func (x *runExit) StateLoad(m state.Source) {
+}
+
+func (x *runExitMain) StateTypeName() string {
+ return "pkg/sentry/kernel.runExitMain"
+}
+
+func (x *runExitMain) StateFields() []string {
+ return []string{}
+}
+
+func (x *runExitMain) beforeSave() {}
+
+func (x *runExitMain) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *runExitMain) afterLoad() {}
+
+func (x *runExitMain) StateLoad(m state.Source) {
+}
+
+func (x *runExitNotify) StateTypeName() string {
+ return "pkg/sentry/kernel.runExitNotify"
+}
+
+func (x *runExitNotify) StateFields() []string {
+ return []string{}
+}
+
+func (x *runExitNotify) beforeSave() {}
+
+func (x *runExitNotify) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *runExitNotify) afterLoad() {}
+
+func (x *runExitNotify) StateLoad(m state.Source) {
+}
+
+func (x *taskList) StateTypeName() string {
+ return "pkg/sentry/kernel.taskList"
+}
+
+func (x *taskList) StateFields() []string {
+ return []string{
+ "head",
+ "tail",
+ }
+}
+
+func (x *taskList) beforeSave() {}
+
+func (x *taskList) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.head)
+ m.Save(1, &x.tail)
+}
+
+func (x *taskList) afterLoad() {}
+
+func (x *taskList) StateLoad(m state.Source) {
+ m.Load(0, &x.head)
+ m.Load(1, &x.tail)
+}
+
+func (x *taskEntry) StateTypeName() string {
+ return "pkg/sentry/kernel.taskEntry"
+}
+
+func (x *taskEntry) StateFields() []string {
+ return []string{
+ "next",
+ "prev",
+ }
+}
+
+func (x *taskEntry) beforeSave() {}
+
+func (x *taskEntry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.next)
+ m.Save(1, &x.prev)
+}
+
+func (x *taskEntry) afterLoad() {}
+
+func (x *taskEntry) StateLoad(m state.Source) {
+ m.Load(0, &x.next)
+ m.Load(1, &x.prev)
+}
+
+func (x *runApp) StateTypeName() string {
+ return "pkg/sentry/kernel.runApp"
+}
+
+func (x *runApp) StateFields() []string {
+ return []string{}
+}
+
+func (x *runApp) beforeSave() {}
+
+func (x *runApp) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *runApp) afterLoad() {}
+
+func (x *runApp) StateLoad(m state.Source) {
+}
+
+func (x *TaskGoroutineSchedInfo) StateTypeName() string {
+ return "pkg/sentry/kernel.TaskGoroutineSchedInfo"
+}
+
+func (x *TaskGoroutineSchedInfo) StateFields() []string {
+ return []string{
+ "Timestamp",
+ "State",
+ "UserTicks",
+ "SysTicks",
+ }
+}
+
+func (x *TaskGoroutineSchedInfo) beforeSave() {}
+
+func (x *TaskGoroutineSchedInfo) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.Timestamp)
+ m.Save(1, &x.State)
+ m.Save(2, &x.UserTicks)
+ m.Save(3, &x.SysTicks)
+}
+
+func (x *TaskGoroutineSchedInfo) afterLoad() {}
+
+func (x *TaskGoroutineSchedInfo) StateLoad(m state.Source) {
+ m.Load(0, &x.Timestamp)
+ m.Load(1, &x.State)
+ m.Load(2, &x.UserTicks)
+ m.Load(3, &x.SysTicks)
+}
+
+func (x *taskClock) StateTypeName() string {
+ return "pkg/sentry/kernel.taskClock"
+}
+
+func (x *taskClock) StateFields() []string {
+ return []string{
+ "t",
+ "includeSys",
+ }
+}
+
+func (x *taskClock) beforeSave() {}
+
+func (x *taskClock) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.t)
+ m.Save(1, &x.includeSys)
+}
+
+func (x *taskClock) afterLoad() {}
+
+func (x *taskClock) StateLoad(m state.Source) {
+ m.Load(0, &x.t)
+ m.Load(1, &x.includeSys)
+}
+
+func (x *tgClock) StateTypeName() string {
+ return "pkg/sentry/kernel.tgClock"
+}
+
+func (x *tgClock) StateFields() []string {
+ return []string{
+ "tg",
+ "includeSys",
+ }
+}
+
+func (x *tgClock) beforeSave() {}
+
+func (x *tgClock) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.tg)
+ m.Save(1, &x.includeSys)
+}
+
+func (x *tgClock) afterLoad() {}
+
+func (x *tgClock) StateLoad(m state.Source) {
+ m.Load(0, &x.tg)
+ m.Load(1, &x.includeSys)
+}
+
+func (x *groupStop) StateTypeName() string {
+ return "pkg/sentry/kernel.groupStop"
+}
+
+func (x *groupStop) StateFields() []string {
+ return []string{}
+}
+
+func (x *groupStop) beforeSave() {}
+
+func (x *groupStop) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *groupStop) afterLoad() {}
+
+func (x *groupStop) StateLoad(m state.Source) {
+}
+
+func (x *runInterrupt) StateTypeName() string {
+ return "pkg/sentry/kernel.runInterrupt"
+}
+
+func (x *runInterrupt) StateFields() []string {
+ return []string{}
+}
+
+func (x *runInterrupt) beforeSave() {}
+
+func (x *runInterrupt) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *runInterrupt) afterLoad() {}
+
+func (x *runInterrupt) StateLoad(m state.Source) {
+}
+
+func (x *runInterruptAfterSignalDeliveryStop) StateTypeName() string {
+ return "pkg/sentry/kernel.runInterruptAfterSignalDeliveryStop"
+}
+
+func (x *runInterruptAfterSignalDeliveryStop) StateFields() []string {
+ return []string{}
+}
+
+func (x *runInterruptAfterSignalDeliveryStop) beforeSave() {}
+
+func (x *runInterruptAfterSignalDeliveryStop) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *runInterruptAfterSignalDeliveryStop) afterLoad() {}
+
+func (x *runInterruptAfterSignalDeliveryStop) StateLoad(m state.Source) {
+}
+
+func (x *runSyscallAfterSyscallEnterStop) StateTypeName() string {
+ return "pkg/sentry/kernel.runSyscallAfterSyscallEnterStop"
+}
+
+func (x *runSyscallAfterSyscallEnterStop) StateFields() []string {
+ return []string{}
+}
+
+func (x *runSyscallAfterSyscallEnterStop) beforeSave() {}
+
+func (x *runSyscallAfterSyscallEnterStop) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *runSyscallAfterSyscallEnterStop) afterLoad() {}
+
+func (x *runSyscallAfterSyscallEnterStop) StateLoad(m state.Source) {
+}
+
+func (x *runSyscallAfterSysemuStop) StateTypeName() string {
+ return "pkg/sentry/kernel.runSyscallAfterSysemuStop"
+}
+
+func (x *runSyscallAfterSysemuStop) StateFields() []string {
+ return []string{}
+}
+
+func (x *runSyscallAfterSysemuStop) beforeSave() {}
+
+func (x *runSyscallAfterSysemuStop) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *runSyscallAfterSysemuStop) afterLoad() {}
+
+func (x *runSyscallAfterSysemuStop) StateLoad(m state.Source) {
+}
+
+func (x *runSyscallReinvoke) StateTypeName() string {
+ return "pkg/sentry/kernel.runSyscallReinvoke"
+}
+
+func (x *runSyscallReinvoke) StateFields() []string {
+ return []string{}
+}
+
+func (x *runSyscallReinvoke) beforeSave() {}
+
+func (x *runSyscallReinvoke) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *runSyscallReinvoke) afterLoad() {}
+
+func (x *runSyscallReinvoke) StateLoad(m state.Source) {
+}
+
+func (x *runSyscallExit) StateTypeName() string {
+ return "pkg/sentry/kernel.runSyscallExit"
+}
+
+func (x *runSyscallExit) StateFields() []string {
+ return []string{}
+}
+
+func (x *runSyscallExit) beforeSave() {}
+
+func (x *runSyscallExit) StateSave(m state.Sink) {
+ x.beforeSave()
+}
+
+func (x *runSyscallExit) afterLoad() {}
+
+func (x *runSyscallExit) StateLoad(m state.Source) {
+}
+
+func (x *ThreadGroup) StateTypeName() string {
+ return "pkg/sentry/kernel.ThreadGroup"
+}
+
+func (x *ThreadGroup) StateFields() []string {
+ return []string{
+ "threadGroupNode",
+ "signalHandlers",
+ "pendingSignals",
+ "groupStopDequeued",
+ "groupStopSignal",
+ "groupStopPendingCount",
+ "groupStopComplete",
+ "groupStopWaitable",
+ "groupContNotify",
+ "groupContInterrupted",
+ "groupContWaitable",
+ "exiting",
+ "exitStatus",
+ "terminationSignal",
+ "itimerRealTimer",
+ "itimerVirtSetting",
+ "itimerProfSetting",
+ "rlimitCPUSoftSetting",
+ "cpuTimersEnabled",
+ "timers",
+ "nextTimerID",
+ "exitedCPUStats",
+ "childCPUStats",
+ "ioUsage",
+ "maxRSS",
+ "childMaxRSS",
+ "limits",
+ "processGroup",
+ "execed",
+ "oldRSeqCritical",
+ "mounts",
+ "tty",
+ "oomScoreAdj",
+ }
+}
+
+func (x *ThreadGroup) beforeSave() {}
+
+func (x *ThreadGroup) StateSave(m state.Sink) {
+ x.beforeSave()
+ var oldRSeqCritical *OldRSeqCriticalRegion = x.saveOldRSeqCritical()
+ m.SaveValue(29, oldRSeqCritical)
+ m.Save(0, &x.threadGroupNode)
+ m.Save(1, &x.signalHandlers)
+ m.Save(2, &x.pendingSignals)
+ m.Save(3, &x.groupStopDequeued)
+ m.Save(4, &x.groupStopSignal)
+ m.Save(5, &x.groupStopPendingCount)
+ m.Save(6, &x.groupStopComplete)
+ m.Save(7, &x.groupStopWaitable)
+ m.Save(8, &x.groupContNotify)
+ m.Save(9, &x.groupContInterrupted)
+ m.Save(10, &x.groupContWaitable)
+ m.Save(11, &x.exiting)
+ m.Save(12, &x.exitStatus)
+ m.Save(13, &x.terminationSignal)
+ m.Save(14, &x.itimerRealTimer)
+ m.Save(15, &x.itimerVirtSetting)
+ m.Save(16, &x.itimerProfSetting)
+ m.Save(17, &x.rlimitCPUSoftSetting)
+ m.Save(18, &x.cpuTimersEnabled)
+ m.Save(19, &x.timers)
+ m.Save(20, &x.nextTimerID)
+ m.Save(21, &x.exitedCPUStats)
+ m.Save(22, &x.childCPUStats)
+ m.Save(23, &x.ioUsage)
+ m.Save(24, &x.maxRSS)
+ m.Save(25, &x.childMaxRSS)
+ m.Save(26, &x.limits)
+ m.Save(27, &x.processGroup)
+ m.Save(28, &x.execed)
+ m.Save(30, &x.mounts)
+ m.Save(31, &x.tty)
+ m.Save(32, &x.oomScoreAdj)
+}
+
+func (x *ThreadGroup) afterLoad() {}
+
+func (x *ThreadGroup) StateLoad(m state.Source) {
+ m.Load(0, &x.threadGroupNode)
+ m.Load(1, &x.signalHandlers)
+ m.Load(2, &x.pendingSignals)
+ m.Load(3, &x.groupStopDequeued)
+ m.Load(4, &x.groupStopSignal)
+ m.Load(5, &x.groupStopPendingCount)
+ m.Load(6, &x.groupStopComplete)
+ m.Load(7, &x.groupStopWaitable)
+ m.Load(8, &x.groupContNotify)
+ m.Load(9, &x.groupContInterrupted)
+ m.Load(10, &x.groupContWaitable)
+ m.Load(11, &x.exiting)
+ m.Load(12, &x.exitStatus)
+ m.Load(13, &x.terminationSignal)
+ m.Load(14, &x.itimerRealTimer)
+ m.Load(15, &x.itimerVirtSetting)
+ m.Load(16, &x.itimerProfSetting)
+ m.Load(17, &x.rlimitCPUSoftSetting)
+ m.Load(18, &x.cpuTimersEnabled)
+ m.Load(19, &x.timers)
+ m.Load(20, &x.nextTimerID)
+ m.Load(21, &x.exitedCPUStats)
+ m.Load(22, &x.childCPUStats)
+ m.Load(23, &x.ioUsage)
+ m.Load(24, &x.maxRSS)
+ m.Load(25, &x.childMaxRSS)
+ m.Load(26, &x.limits)
+ m.Load(27, &x.processGroup)
+ m.Load(28, &x.execed)
+ m.Load(30, &x.mounts)
+ m.Load(31, &x.tty)
+ m.Load(32, &x.oomScoreAdj)
+ m.LoadValue(29, new(*OldRSeqCriticalRegion), func(y interface{}) { x.loadOldRSeqCritical(y.(*OldRSeqCriticalRegion)) })
+}
+
+func (x *itimerRealListener) StateTypeName() string {
+ return "pkg/sentry/kernel.itimerRealListener"
+}
+
+func (x *itimerRealListener) StateFields() []string {
+ return []string{
+ "tg",
+ }
+}
+
+func (x *itimerRealListener) beforeSave() {}
+
+func (x *itimerRealListener) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.tg)
+}
+
+func (x *itimerRealListener) afterLoad() {}
+
+func (x *itimerRealListener) StateLoad(m state.Source) {
+ m.Load(0, &x.tg)
+}
+
+func (x *TaskSet) StateTypeName() string {
+ return "pkg/sentry/kernel.TaskSet"
+}
+
+func (x *TaskSet) StateFields() []string {
+ return []string{
+ "Root",
+ "sessions",
+ }
+}
+
+func (x *TaskSet) beforeSave() {}
+
+func (x *TaskSet) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.Root)
+ m.Save(1, &x.sessions)
+}
+
+func (x *TaskSet) afterLoad() {}
+
+func (x *TaskSet) StateLoad(m state.Source) {
+ m.Load(0, &x.Root)
+ m.Load(1, &x.sessions)
+}
+
+func (x *PIDNamespace) StateTypeName() string {
+ return "pkg/sentry/kernel.PIDNamespace"
+}
+
+func (x *PIDNamespace) StateFields() []string {
+ return []string{
+ "owner",
+ "parent",
+ "userns",
+ "last",
+ "tasks",
+ "tids",
+ "tgids",
+ "sessions",
+ "sids",
+ "processGroups",
+ "pgids",
+ "exiting",
+ }
+}
+
+func (x *PIDNamespace) beforeSave() {}
+
+func (x *PIDNamespace) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.owner)
+ m.Save(1, &x.parent)
+ m.Save(2, &x.userns)
+ m.Save(3, &x.last)
+ m.Save(4, &x.tasks)
+ m.Save(5, &x.tids)
+ m.Save(6, &x.tgids)
+ m.Save(7, &x.sessions)
+ m.Save(8, &x.sids)
+ m.Save(9, &x.processGroups)
+ m.Save(10, &x.pgids)
+ m.Save(11, &x.exiting)
+}
+
+func (x *PIDNamespace) afterLoad() {}
+
+func (x *PIDNamespace) StateLoad(m state.Source) {
+ m.Load(0, &x.owner)
+ m.Load(1, &x.parent)
+ m.Load(2, &x.userns)
+ m.Load(3, &x.last)
+ m.Load(4, &x.tasks)
+ m.Load(5, &x.tids)
+ m.Load(6, &x.tgids)
+ m.Load(7, &x.sessions)
+ m.Load(8, &x.sids)
+ m.Load(9, &x.processGroups)
+ m.Load(10, &x.pgids)
+ m.Load(11, &x.exiting)
+}
+
+func (x *threadGroupNode) StateTypeName() string {
+ return "pkg/sentry/kernel.threadGroupNode"
+}
+
+func (x *threadGroupNode) StateFields() []string {
+ return []string{
+ "pidns",
+ "leader",
+ "execing",
+ "tasks",
+ "tasksCount",
+ "liveTasks",
+ "activeTasks",
+ }
+}
+
+func (x *threadGroupNode) beforeSave() {}
+
+func (x *threadGroupNode) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.pidns)
+ m.Save(1, &x.leader)
+ m.Save(2, &x.execing)
+ m.Save(3, &x.tasks)
+ m.Save(4, &x.tasksCount)
+ m.Save(5, &x.liveTasks)
+ m.Save(6, &x.activeTasks)
+}
+
+func (x *threadGroupNode) afterLoad() {}
+
+func (x *threadGroupNode) StateLoad(m state.Source) {
+ m.Load(0, &x.pidns)
+ m.Load(1, &x.leader)
+ m.Load(2, &x.execing)
+ m.Load(3, &x.tasks)
+ m.Load(4, &x.tasksCount)
+ m.Load(5, &x.liveTasks)
+ m.Load(6, &x.activeTasks)
+}
+
+func (x *taskNode) StateTypeName() string {
+ return "pkg/sentry/kernel.taskNode"
+}
+
+func (x *taskNode) StateFields() []string {
+ return []string{
+ "tg",
+ "taskEntry",
+ "parent",
+ "children",
+ "childPIDNamespace",
+ }
+}
+
+func (x *taskNode) beforeSave() {}
+
+func (x *taskNode) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.tg)
+ m.Save(1, &x.taskEntry)
+ m.Save(2, &x.parent)
+ m.Save(3, &x.children)
+ m.Save(4, &x.childPIDNamespace)
+}
+
+func (x *taskNode) afterLoad() {}
+
+func (x *taskNode) StateLoad(m state.Source) {
+ m.LoadWait(0, &x.tg)
+ m.Load(1, &x.taskEntry)
+ m.Load(2, &x.parent)
+ m.Load(3, &x.children)
+ m.Load(4, &x.childPIDNamespace)
+}
+
+func (x *Timekeeper) StateTypeName() string {
+ return "pkg/sentry/kernel.Timekeeper"
+}
+
+func (x *Timekeeper) StateFields() []string {
+ return []string{
+ "bootTime",
+ "saveMonotonic",
+ "saveRealtime",
+ "params",
+ }
+}
+
+func (x *Timekeeper) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.bootTime)
+ m.Save(1, &x.saveMonotonic)
+ m.Save(2, &x.saveRealtime)
+ m.Save(3, &x.params)
+}
+
+func (x *Timekeeper) StateLoad(m state.Source) {
+ m.Load(0, &x.bootTime)
+ m.Load(1, &x.saveMonotonic)
+ m.Load(2, &x.saveRealtime)
+ m.Load(3, &x.params)
+ m.AfterLoad(x.afterLoad)
+}
+
+func (x *timekeeperClock) StateTypeName() string {
+ return "pkg/sentry/kernel.timekeeperClock"
+}
+
+func (x *timekeeperClock) StateFields() []string {
+ return []string{
+ "tk",
+ "c",
+ }
+}
+
+func (x *timekeeperClock) beforeSave() {}
+
+func (x *timekeeperClock) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.tk)
+ m.Save(1, &x.c)
+}
+
+func (x *timekeeperClock) afterLoad() {}
+
+func (x *timekeeperClock) StateLoad(m state.Source) {
+ m.Load(0, &x.tk)
+ m.Load(1, &x.c)
+}
+
+func (x *TTY) StateTypeName() string {
+ return "pkg/sentry/kernel.TTY"
+}
+
+func (x *TTY) StateFields() []string {
+ return []string{
+ "Index",
+ "tg",
+ }
+}
+
+func (x *TTY) beforeSave() {}
+
+func (x *TTY) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.Index)
+ m.Save(1, &x.tg)
+}
+
+func (x *TTY) afterLoad() {}
+
+func (x *TTY) StateLoad(m state.Source) {
+ m.Load(0, &x.Index)
+ m.Load(1, &x.tg)
+}
+
+func (x *UTSNamespace) StateTypeName() string {
+ return "pkg/sentry/kernel.UTSNamespace"
+}
+
+func (x *UTSNamespace) StateFields() []string {
+ return []string{
+ "hostName",
+ "domainName",
+ "userns",
+ }
+}
+
+func (x *UTSNamespace) beforeSave() {}
+
+func (x *UTSNamespace) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.hostName)
+ m.Save(1, &x.domainName)
+ m.Save(2, &x.userns)
+}
+
+func (x *UTSNamespace) afterLoad() {}
+
+func (x *UTSNamespace) StateLoad(m state.Source) {
+ m.Load(0, &x.hostName)
+ m.Load(1, &x.domainName)
+ m.Load(2, &x.userns)
+}
+
+func (x *VDSOParamPage) StateTypeName() string {
+ return "pkg/sentry/kernel.VDSOParamPage"
+}
+
+func (x *VDSOParamPage) StateFields() []string {
+ return []string{
+ "mfp",
+ "fr",
+ "seq",
+ }
+}
+
+func (x *VDSOParamPage) beforeSave() {}
+
+func (x *VDSOParamPage) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.mfp)
+ m.Save(1, &x.fr)
+ m.Save(2, &x.seq)
+}
+
+func (x *VDSOParamPage) afterLoad() {}
+
+func (x *VDSOParamPage) StateLoad(m state.Source) {
+ m.Load(0, &x.mfp)
+ m.Load(1, &x.fr)
+ m.Load(2, &x.seq)
+}
+
+func init() {
+ state.Register((*abstractEndpoint)(nil))
+ state.Register((*AbstractSocketNamespace)(nil))
+ state.Register((*FDFlags)(nil))
+ state.Register((*descriptor)(nil))
+ state.Register((*FDTable)(nil))
+ state.Register((*FDTableRefs)(nil))
+ state.Register((*FSContext)(nil))
+ state.Register((*FSContextRefs)(nil))
+ state.Register((*IPCNamespace)(nil))
+ state.Register((*Kernel)(nil))
+ state.Register((*SocketEntry)(nil))
+ state.Register((*pendingSignals)(nil))
+ state.Register((*pendingSignalQueue)(nil))
+ state.Register((*pendingSignal)(nil))
+ state.Register((*pendingSignalList)(nil))
+ state.Register((*pendingSignalEntry)(nil))
+ state.Register((*savedPendingSignal)(nil))
+ state.Register((*IntervalTimer)(nil))
+ state.Register((*processGroupList)(nil))
+ state.Register((*processGroupEntry)(nil))
+ state.Register((*ProcessGroupRefs)(nil))
+ state.Register((*ptraceOptions)(nil))
+ state.Register((*ptraceStop)(nil))
+ state.Register((*OldRSeqCriticalRegion)(nil))
+ state.Register((*sessionList)(nil))
+ state.Register((*sessionEntry)(nil))
+ state.Register((*SessionRefs)(nil))
+ state.Register((*Session)(nil))
+ state.Register((*ProcessGroup)(nil))
+ state.Register((*SignalHandlers)(nil))
+ state.Register((*socketList)(nil))
+ state.Register((*socketEntry)(nil))
+ state.Register((*syscallTableInfo)(nil))
+ state.Register((*syslog)(nil))
+ state.Register((*Task)(nil))
+ state.Register((*runSyscallAfterPtraceEventClone)(nil))
+ state.Register((*runSyscallAfterVforkStop)(nil))
+ state.Register((*vforkStop)(nil))
+ state.Register((*TaskContext)(nil))
+ state.Register((*execStop)(nil))
+ state.Register((*runSyscallAfterExecStop)(nil))
+ state.Register((*ExitStatus)(nil))
+ state.Register((*runExit)(nil))
+ state.Register((*runExitMain)(nil))
+ state.Register((*runExitNotify)(nil))
+ state.Register((*taskList)(nil))
+ state.Register((*taskEntry)(nil))
+ state.Register((*runApp)(nil))
+ state.Register((*TaskGoroutineSchedInfo)(nil))
+ state.Register((*taskClock)(nil))
+ state.Register((*tgClock)(nil))
+ state.Register((*groupStop)(nil))
+ state.Register((*runInterrupt)(nil))
+ state.Register((*runInterruptAfterSignalDeliveryStop)(nil))
+ state.Register((*runSyscallAfterSyscallEnterStop)(nil))
+ state.Register((*runSyscallAfterSysemuStop)(nil))
+ state.Register((*runSyscallReinvoke)(nil))
+ state.Register((*runSyscallExit)(nil))
+ state.Register((*ThreadGroup)(nil))
+ state.Register((*itimerRealListener)(nil))
+ state.Register((*TaskSet)(nil))
+ state.Register((*PIDNamespace)(nil))
+ state.Register((*threadGroupNode)(nil))
+ state.Register((*taskNode)(nil))
+ state.Register((*Timekeeper)(nil))
+ state.Register((*timekeeperClock)(nil))
+ state.Register((*TTY)(nil))
+ state.Register((*UTSNamespace)(nil))
+ state.Register((*VDSOParamPage)(nil))
+}
diff --git a/pkg/sentry/kernel/kernel_unsafe_state_autogen.go b/pkg/sentry/kernel/kernel_unsafe_state_autogen.go
new file mode 100644
index 000000000..12130bf74
--- /dev/null
+++ b/pkg/sentry/kernel/kernel_unsafe_state_autogen.go
@@ -0,0 +1,3 @@
+// automatically generated by stateify.
+
+package kernel
diff --git a/pkg/sentry/kernel/memevent/BUILD b/pkg/sentry/kernel/memevent/BUILD
deleted file mode 100644
index 4486848d2..000000000
--- a/pkg/sentry/kernel/memevent/BUILD
+++ /dev/null
@@ -1,24 +0,0 @@
-load("//tools:defs.bzl", "go_library", "proto_library")
-
-package(licenses = ["notice"])
-
-go_library(
- name = "memevent",
- srcs = ["memory_events.go"],
- visibility = ["//:sandbox"],
- deps = [
- ":memory_events_go_proto",
- "//pkg/eventchannel",
- "//pkg/log",
- "//pkg/metric",
- "//pkg/sentry/kernel",
- "//pkg/sentry/usage",
- "//pkg/sync",
- ],
-)
-
-proto_library(
- name = "memory_events",
- srcs = ["memory_events.proto"],
- visibility = ["//visibility:public"],
-)
diff --git a/pkg/sentry/kernel/memevent/memevent_state_autogen.go b/pkg/sentry/kernel/memevent/memevent_state_autogen.go
new file mode 100644
index 000000000..4a1679fa9
--- /dev/null
+++ b/pkg/sentry/kernel/memevent/memevent_state_autogen.go
@@ -0,0 +1,3 @@
+// automatically generated by stateify.
+
+package memevent
diff --git a/pkg/sentry/kernel/memevent/memory_events.proto b/pkg/sentry/kernel/memevent/memory_events.proto
deleted file mode 100644
index bf8029ff5..000000000
--- a/pkg/sentry/kernel/memevent/memory_events.proto
+++ /dev/null
@@ -1,29 +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.
-
-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;
-
- // 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;
-}
diff --git a/pkg/sentry/kernel/memevent/memory_events_go_proto/memory_events.pb.go b/pkg/sentry/kernel/memevent/memory_events_go_proto/memory_events.pb.go
new file mode 100644
index 000000000..f8b857fa9
--- /dev/null
+++ b/pkg/sentry/kernel/memevent/memory_events_go_proto/memory_events.pb.go
@@ -0,0 +1,88 @@
+// Code generated by protoc-gen-go. DO NOT EDIT.
+// source: pkg/sentry/kernel/memevent/memory_events.proto
+
+package gvisor
+
+import (
+ fmt "fmt"
+ proto "github.com/golang/protobuf/proto"
+ math "math"
+)
+
+// Reference imports to suppress errors if they are not otherwise used.
+var _ = proto.Marshal
+var _ = fmt.Errorf
+var _ = math.Inf
+
+// This is a compile-time assertion to ensure that this generated file
+// is compatible with the proto package it is being compiled against.
+// A compilation error at this line likely means your copy of the
+// proto package needs to be updated.
+const _ = proto.ProtoPackageIsVersion3 // please upgrade the proto package
+
+type MemoryUsageEvent struct {
+ Total uint64 `protobuf:"varint,1,opt,name=total,proto3" json:"total,omitempty"`
+ Mapped uint64 `protobuf:"varint,2,opt,name=mapped,proto3" json:"mapped,omitempty"`
+ XXX_NoUnkeyedLiteral struct{} `json:"-"`
+ XXX_unrecognized []byte `json:"-"`
+ XXX_sizecache int32 `json:"-"`
+}
+
+func (m *MemoryUsageEvent) Reset() { *m = MemoryUsageEvent{} }
+func (m *MemoryUsageEvent) String() string { return proto.CompactTextString(m) }
+func (*MemoryUsageEvent) ProtoMessage() {}
+func (*MemoryUsageEvent) Descriptor() ([]byte, []int) {
+ return fileDescriptor_cd85fc8d1130e4b0, []int{0}
+}
+
+func (m *MemoryUsageEvent) XXX_Unmarshal(b []byte) error {
+ return xxx_messageInfo_MemoryUsageEvent.Unmarshal(m, b)
+}
+func (m *MemoryUsageEvent) XXX_Marshal(b []byte, deterministic bool) ([]byte, error) {
+ return xxx_messageInfo_MemoryUsageEvent.Marshal(b, m, deterministic)
+}
+func (m *MemoryUsageEvent) XXX_Merge(src proto.Message) {
+ xxx_messageInfo_MemoryUsageEvent.Merge(m, src)
+}
+func (m *MemoryUsageEvent) XXX_Size() int {
+ return xxx_messageInfo_MemoryUsageEvent.Size(m)
+}
+func (m *MemoryUsageEvent) XXX_DiscardUnknown() {
+ xxx_messageInfo_MemoryUsageEvent.DiscardUnknown(m)
+}
+
+var xxx_messageInfo_MemoryUsageEvent proto.InternalMessageInfo
+
+func (m *MemoryUsageEvent) GetTotal() uint64 {
+ if m != nil {
+ return m.Total
+ }
+ return 0
+}
+
+func (m *MemoryUsageEvent) GetMapped() uint64 {
+ if m != nil {
+ return m.Mapped
+ }
+ return 0
+}
+
+func init() {
+ proto.RegisterType((*MemoryUsageEvent)(nil), "gvisor.MemoryUsageEvent")
+}
+
+func init() {
+ proto.RegisterFile("pkg/sentry/kernel/memevent/memory_events.proto", fileDescriptor_cd85fc8d1130e4b0)
+}
+
+var fileDescriptor_cd85fc8d1130e4b0 = []byte{
+ // 128 bytes of a gzipped FileDescriptorProto
+ 0x1f, 0x8b, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0xff, 0xe2, 0xd2, 0x2b, 0xc8, 0x4e, 0xd7,
+ 0x2f, 0x4e, 0xcd, 0x2b, 0x29, 0xaa, 0xd4, 0xcf, 0x4e, 0x2d, 0xca, 0x4b, 0xcd, 0xd1, 0xcf, 0x4d,
+ 0xcd, 0x4d, 0x2d, 0x4b, 0xcd, 0x2b, 0x01, 0x31, 0xf2, 0x8b, 0x2a, 0xe3, 0xc1, 0x9c, 0x62, 0xbd,
+ 0x82, 0xa2, 0xfc, 0x92, 0x7c, 0x21, 0xb6, 0xf4, 0xb2, 0xcc, 0xe2, 0xfc, 0x22, 0x25, 0x07, 0x2e,
+ 0x01, 0x5f, 0xb0, 0x74, 0x68, 0x71, 0x62, 0x7a, 0xaa, 0x2b, 0x48, 0x89, 0x90, 0x08, 0x17, 0x6b,
+ 0x49, 0x7e, 0x49, 0x62, 0x8e, 0x04, 0xa3, 0x02, 0xa3, 0x06, 0x4b, 0x10, 0x84, 0x23, 0x24, 0xc6,
+ 0xc5, 0x96, 0x9b, 0x58, 0x50, 0x90, 0x9a, 0x22, 0xc1, 0x04, 0x16, 0x86, 0xf2, 0x92, 0xd8, 0xc0,
+ 0x06, 0x1a, 0x03, 0x02, 0x00, 0x00, 0xff, 0xff, 0x99, 0x31, 0x2f, 0x9d, 0x82, 0x00, 0x00, 0x00,
+}
diff --git a/pkg/sentry/kernel/pending_signals_list.go b/pkg/sentry/kernel/pending_signals_list.go
new file mode 100644
index 000000000..d0062087a
--- /dev/null
+++ b/pkg/sentry/kernel/pending_signals_list.go
@@ -0,0 +1,193 @@
+package kernel
+
+// 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 pendingSignalElementMapper struct{}
+
+// linkerFor maps an Element to a Linker.
+//
+// This default implementation should be inlined.
+//
+//go:nosplit
+func (pendingSignalElementMapper) linkerFor(elem *pendingSignal) *pendingSignal { 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.
+//
+// The zero value for List is an empty list ready to use.
+//
+// To iterate over a list (where l is a List):
+// for e := l.Front(); e != nil; e = e.Next() {
+// // do something with e.
+// }
+//
+// +stateify savable
+type pendingSignalList struct {
+ head *pendingSignal
+ tail *pendingSignal
+}
+
+// Reset resets list l to the empty state.
+func (l *pendingSignalList) Reset() {
+ l.head = nil
+ l.tail = nil
+}
+
+// Empty returns true iff the list is empty.
+func (l *pendingSignalList) Empty() bool {
+ return l.head == nil
+}
+
+// Front returns the first element of list l or nil.
+func (l *pendingSignalList) Front() *pendingSignal {
+ return l.head
+}
+
+// Back returns the last element of list l or nil.
+func (l *pendingSignalList) Back() *pendingSignal {
+ return l.tail
+}
+
+// Len returns the number of elements in the list.
+//
+// NOTE: This is an O(n) operation.
+func (l *pendingSignalList) Len() (count int) {
+ for e := l.Front(); e != nil; e = (pendingSignalElementMapper{}.linkerFor(e)).Next() {
+ count++
+ }
+ return count
+}
+
+// PushFront inserts the element e at the front of list l.
+func (l *pendingSignalList) PushFront(e *pendingSignal) {
+ linker := pendingSignalElementMapper{}.linkerFor(e)
+ linker.SetNext(l.head)
+ linker.SetPrev(nil)
+ if l.head != nil {
+ pendingSignalElementMapper{}.linkerFor(l.head).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+
+ l.head = e
+}
+
+// PushBack inserts the element e at the back of list l.
+func (l *pendingSignalList) PushBack(e *pendingSignal) {
+ linker := pendingSignalElementMapper{}.linkerFor(e)
+ linker.SetNext(nil)
+ linker.SetPrev(l.tail)
+ if l.tail != nil {
+ pendingSignalElementMapper{}.linkerFor(l.tail).SetNext(e)
+ } else {
+ l.head = e
+ }
+
+ l.tail = e
+}
+
+// PushBackList inserts list m at the end of list l, emptying m.
+func (l *pendingSignalList) PushBackList(m *pendingSignalList) {
+ if l.head == nil {
+ l.head = m.head
+ l.tail = m.tail
+ } else if m.head != nil {
+ pendingSignalElementMapper{}.linkerFor(l.tail).SetNext(m.head)
+ pendingSignalElementMapper{}.linkerFor(m.head).SetPrev(l.tail)
+
+ l.tail = m.tail
+ }
+ m.head = nil
+ m.tail = nil
+}
+
+// InsertAfter inserts e after b.
+func (l *pendingSignalList) InsertAfter(b, e *pendingSignal) {
+ bLinker := pendingSignalElementMapper{}.linkerFor(b)
+ eLinker := pendingSignalElementMapper{}.linkerFor(e)
+
+ a := bLinker.Next()
+
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ bLinker.SetNext(e)
+
+ if a != nil {
+ pendingSignalElementMapper{}.linkerFor(a).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+}
+
+// InsertBefore inserts e before a.
+func (l *pendingSignalList) InsertBefore(a, e *pendingSignal) {
+ aLinker := pendingSignalElementMapper{}.linkerFor(a)
+ eLinker := pendingSignalElementMapper{}.linkerFor(e)
+
+ b := aLinker.Prev()
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ aLinker.SetPrev(e)
+
+ if b != nil {
+ pendingSignalElementMapper{}.linkerFor(b).SetNext(e)
+ } else {
+ l.head = e
+ }
+}
+
+// Remove removes e from l.
+func (l *pendingSignalList) Remove(e *pendingSignal) {
+ linker := pendingSignalElementMapper{}.linkerFor(e)
+ prev := linker.Prev()
+ next := linker.Next()
+
+ if prev != nil {
+ pendingSignalElementMapper{}.linkerFor(prev).SetNext(next)
+ } else if l.head == e {
+ l.head = next
+ }
+
+ if next != nil {
+ pendingSignalElementMapper{}.linkerFor(next).SetPrev(prev)
+ } else if l.tail == e {
+ l.tail = prev
+ }
+
+ linker.SetNext(nil)
+ linker.SetPrev(nil)
+}
+
+// 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 pendingSignalEntry struct {
+ next *pendingSignal
+ prev *pendingSignal
+}
+
+// Next returns the entry that follows e in the list.
+func (e *pendingSignalEntry) Next() *pendingSignal {
+ return e.next
+}
+
+// Prev returns the entry that precedes e in the list.
+func (e *pendingSignalEntry) Prev() *pendingSignal {
+ return e.prev
+}
+
+// SetNext assigns 'entry' as the entry that follows e in the list.
+func (e *pendingSignalEntry) SetNext(elem *pendingSignal) {
+ e.next = elem
+}
+
+// SetPrev assigns 'entry' as the entry that precedes e in the list.
+func (e *pendingSignalEntry) SetPrev(elem *pendingSignal) {
+ e.prev = elem
+}
diff --git a/pkg/sentry/kernel/pipe/BUILD b/pkg/sentry/kernel/pipe/BUILD
deleted file mode 100644
index 449643118..000000000
--- a/pkg/sentry/kernel/pipe/BUILD
+++ /dev/null
@@ -1,54 +0,0 @@
-load("//tools:defs.bzl", "go_library", "go_test")
-
-package(licenses = ["notice"])
-
-go_library(
- name = "pipe",
- srcs = [
- "device.go",
- "node.go",
- "pipe.go",
- "pipe_unsafe.go",
- "pipe_util.go",
- "reader.go",
- "reader_writer.go",
- "vfs.go",
- "writer.go",
- ],
- visibility = ["//pkg/sentry:internal"],
- deps = [
- "//pkg/abi/linux",
- "//pkg/amutex",
- "//pkg/buffer",
- "//pkg/context",
- "//pkg/safemem",
- "//pkg/sentry/arch",
- "//pkg/sentry/device",
- "//pkg/sentry/fs",
- "//pkg/sentry/fs/fsutil",
- "//pkg/sentry/fs/lock",
- "//pkg/sentry/vfs",
- "//pkg/sync",
- "//pkg/syserror",
- "//pkg/usermem",
- "//pkg/waiter",
- ],
-)
-
-go_test(
- name = "pipe_test",
- size = "small",
- srcs = [
- "node_test.go",
- "pipe_test.go",
- ],
- library = ":pipe",
- deps = [
- "//pkg/context",
- "//pkg/sentry/contexttest",
- "//pkg/sentry/fs",
- "//pkg/syserror",
- "//pkg/usermem",
- "//pkg/waiter",
- ],
-)
diff --git a/pkg/sentry/kernel/pipe/node_test.go b/pkg/sentry/kernel/pipe/node_test.go
deleted file mode 100644
index ce0db5583..000000000
--- a/pkg/sentry/kernel/pipe/node_test.go
+++ /dev/null
@@ -1,320 +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
-
-import (
- "testing"
- "time"
-
- "gvisor.dev/gvisor/pkg/context"
- "gvisor.dev/gvisor/pkg/sentry/contexttest"
- "gvisor.dev/gvisor/pkg/sentry/fs"
- "gvisor.dev/gvisor/pkg/syserror"
- "gvisor.dev/gvisor/pkg/usermem"
-)
-
-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{}{}
-}
-
-func (s *sleeper) Interrupted() bool {
- return len(s.ch) != 0
-}
-
-type openResult struct {
- *fs.File
- 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) {
- inode := fs.NewMockInode(ctx, fs.NewMockMountSource(nil), fs.StableAttr{Type: fs.Pipe})
- d := fs.NewDirent(ctx, inode, "pipe")
- file, err := n.GetFile(ctx, d, 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) {
- inode := fs.NewMockInode(ctx, fs.NewMockMountSource(nil), fs.StableAttr{Type: fs.Pipe})
- d := fs.NewDirent(ctx, inode, "pipe")
- file, err := n.GetFile(ctx, d, flags)
- if resChan != nil {
- resChan <- openResult{file, err}
- }
- return file, err
-}
-
-func newNamedPipe(t *testing.T) *Pipe {
- return NewPipe(true, DefaultPipeSize, usermem.PageSize)
-}
-
-func newAnonPipe(t *testing.T) *Pipe {
- return NewPipe(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) {
- ctx := newSleeperContext(t)
- f := NewInodeOperations(ctx, perms, newNamedPipe(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) {
- ctx := newSleeperContext(t)
- f := NewInodeOperations(ctx, perms, newNamedPipe(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) {
- ctx := newSleeperContext(t)
- f := NewInodeOperations(ctx, perms, newNamedPipe(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) {
- ctx := newSleeperContext(t)
- f := NewInodeOperations(ctx, perms, newNamedPipe(t))
-
- rFile, _ := testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true, NonBlocking: true}, nil)
- rFile.DecRef(ctx)
-
- 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) {
- 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
- // 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) {
- ctx := newSleeperContext(t)
- f := NewInodeOperations(ctx, perms, newNamedPipe(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) {
- ctx := newSleeperContext(t)
- f := NewInodeOperations(ctx, perms, newNamedPipe(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) {
- ctx := newSleeperContext(t)
- f := NewInodeOperations(ctx, perms, newNamedPipe(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 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 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)
- }
-}
-
-func TestNonBlockingReadOpenWithWriter(t *testing.T) {
- ctx := newSleeperContext(t)
- f := NewInodeOperations(ctx, perms, newNamedPipe(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) {
- ctx := newSleeperContext(t)
- f := NewInodeOperations(ctx, perms, newNamedPipe(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) {
- 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)
- }
-}
-
-func TestAnonWriteOpen(t *testing.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_state_autogen.go b/pkg/sentry/kernel/pipe/pipe_state_autogen.go
new file mode 100644
index 000000000..df55abce6
--- /dev/null
+++ b/pkg/sentry/kernel/pipe/pipe_state_autogen.go
@@ -0,0 +1,151 @@
+// automatically generated by stateify.
+
+package pipe
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *inodeOperations) StateTypeName() string {
+ return "pkg/sentry/kernel/pipe.inodeOperations"
+}
+
+func (x *inodeOperations) StateFields() []string {
+ return []string{
+ "InodeSimpleAttributes",
+ "p",
+ }
+}
+
+func (x *inodeOperations) beforeSave() {}
+
+func (x *inodeOperations) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.InodeSimpleAttributes)
+ m.Save(1, &x.p)
+}
+
+func (x *inodeOperations) afterLoad() {}
+
+func (x *inodeOperations) StateLoad(m state.Source) {
+ m.Load(0, &x.InodeSimpleAttributes)
+ m.Load(1, &x.p)
+}
+
+func (x *Pipe) StateTypeName() string {
+ return "pkg/sentry/kernel/pipe.Pipe"
+}
+
+func (x *Pipe) StateFields() []string {
+ return []string{
+ "isNamed",
+ "atomicIOBytes",
+ "readers",
+ "writers",
+ "view",
+ "max",
+ "hadWriter",
+ }
+}
+
+func (x *Pipe) beforeSave() {}
+
+func (x *Pipe) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.isNamed)
+ m.Save(1, &x.atomicIOBytes)
+ m.Save(2, &x.readers)
+ m.Save(3, &x.writers)
+ m.Save(4, &x.view)
+ m.Save(5, &x.max)
+ m.Save(6, &x.hadWriter)
+}
+
+func (x *Pipe) afterLoad() {}
+
+func (x *Pipe) StateLoad(m state.Source) {
+ m.Load(0, &x.isNamed)
+ m.Load(1, &x.atomicIOBytes)
+ m.Load(2, &x.readers)
+ m.Load(3, &x.writers)
+ m.Load(4, &x.view)
+ m.Load(5, &x.max)
+ m.Load(6, &x.hadWriter)
+}
+
+func (x *Reader) StateTypeName() string {
+ return "pkg/sentry/kernel/pipe.Reader"
+}
+
+func (x *Reader) StateFields() []string {
+ return []string{
+ "ReaderWriter",
+ }
+}
+
+func (x *Reader) beforeSave() {}
+
+func (x *Reader) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.ReaderWriter)
+}
+
+func (x *Reader) afterLoad() {}
+
+func (x *Reader) StateLoad(m state.Source) {
+ m.Load(0, &x.ReaderWriter)
+}
+
+func (x *ReaderWriter) StateTypeName() string {
+ return "pkg/sentry/kernel/pipe.ReaderWriter"
+}
+
+func (x *ReaderWriter) StateFields() []string {
+ return []string{
+ "Pipe",
+ }
+}
+
+func (x *ReaderWriter) beforeSave() {}
+
+func (x *ReaderWriter) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.Pipe)
+}
+
+func (x *ReaderWriter) afterLoad() {}
+
+func (x *ReaderWriter) StateLoad(m state.Source) {
+ m.Load(0, &x.Pipe)
+}
+
+func (x *Writer) StateTypeName() string {
+ return "pkg/sentry/kernel/pipe.Writer"
+}
+
+func (x *Writer) StateFields() []string {
+ return []string{
+ "ReaderWriter",
+ }
+}
+
+func (x *Writer) beforeSave() {}
+
+func (x *Writer) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.ReaderWriter)
+}
+
+func (x *Writer) afterLoad() {}
+
+func (x *Writer) StateLoad(m state.Source) {
+ m.Load(0, &x.ReaderWriter)
+}
+
+func init() {
+ state.Register((*inodeOperations)(nil))
+ state.Register((*Pipe)(nil))
+ state.Register((*Reader)(nil))
+ state.Register((*ReaderWriter)(nil))
+ state.Register((*Writer)(nil))
+}
diff --git a/pkg/sentry/kernel/pipe/pipe_test.go b/pkg/sentry/kernel/pipe/pipe_test.go
deleted file mode 100644
index fe97e9800..000000000
--- a/pkg/sentry/kernel/pipe/pipe_test.go
+++ /dev/null
@@ -1,139 +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
-
-import (
- "bytes"
- "testing"
-
- "gvisor.dev/gvisor/pkg/sentry/contexttest"
- "gvisor.dev/gvisor/pkg/syserror"
- "gvisor.dev/gvisor/pkg/usermem"
- "gvisor.dev/gvisor/pkg/waiter"
-)
-
-func TestPipeRW(t *testing.T) {
- ctx := contexttest.Context(t)
- r, w := NewConnectedPipe(ctx, 65536, 4096)
- defer r.DecRef(ctx)
- defer w.DecRef(ctx)
-
- 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(ctx)
- defer w.DecRef(ctx)
-
- 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
- const capacity = MinimumPipeSize
-
- ctx := contexttest.Context(t)
- r, w := NewConnectedPipe(ctx, capacity, atomicIOBytes)
- defer r.DecRef(ctx)
- defer w.DecRef(ctx)
-
- msg := make([]byte, capacity+1)
- n, err := w.Writev(ctx, usermem.BytesIOSequence(msg))
- if wantN, wantErr := int64(capacity), 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(ctx)
- defer w.DecRef(ctx)
-
- 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/pipe_unsafe_state_autogen.go b/pkg/sentry/kernel/pipe/pipe_unsafe_state_autogen.go
new file mode 100644
index 000000000..d3b40feb4
--- /dev/null
+++ b/pkg/sentry/kernel/pipe/pipe_unsafe_state_autogen.go
@@ -0,0 +1,3 @@
+// automatically generated by stateify.
+
+package pipe
diff --git a/pkg/sentry/kernel/process_group_list.go b/pkg/sentry/kernel/process_group_list.go
new file mode 100644
index 000000000..bc2df5bce
--- /dev/null
+++ b/pkg/sentry/kernel/process_group_list.go
@@ -0,0 +1,193 @@
+package kernel
+
+// 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 processGroupElementMapper struct{}
+
+// linkerFor maps an Element to a Linker.
+//
+// This default implementation should be inlined.
+//
+//go:nosplit
+func (processGroupElementMapper) linkerFor(elem *ProcessGroup) *ProcessGroup { 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.
+//
+// The zero value for List is an empty list ready to use.
+//
+// To iterate over a list (where l is a List):
+// for e := l.Front(); e != nil; e = e.Next() {
+// // do something with e.
+// }
+//
+// +stateify savable
+type processGroupList struct {
+ head *ProcessGroup
+ tail *ProcessGroup
+}
+
+// Reset resets list l to the empty state.
+func (l *processGroupList) Reset() {
+ l.head = nil
+ l.tail = nil
+}
+
+// Empty returns true iff the list is empty.
+func (l *processGroupList) Empty() bool {
+ return l.head == nil
+}
+
+// Front returns the first element of list l or nil.
+func (l *processGroupList) Front() *ProcessGroup {
+ return l.head
+}
+
+// Back returns the last element of list l or nil.
+func (l *processGroupList) Back() *ProcessGroup {
+ return l.tail
+}
+
+// Len returns the number of elements in the list.
+//
+// NOTE: This is an O(n) operation.
+func (l *processGroupList) Len() (count int) {
+ for e := l.Front(); e != nil; e = (processGroupElementMapper{}.linkerFor(e)).Next() {
+ count++
+ }
+ return count
+}
+
+// PushFront inserts the element e at the front of list l.
+func (l *processGroupList) PushFront(e *ProcessGroup) {
+ linker := processGroupElementMapper{}.linkerFor(e)
+ linker.SetNext(l.head)
+ linker.SetPrev(nil)
+ if l.head != nil {
+ processGroupElementMapper{}.linkerFor(l.head).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+
+ l.head = e
+}
+
+// PushBack inserts the element e at the back of list l.
+func (l *processGroupList) PushBack(e *ProcessGroup) {
+ linker := processGroupElementMapper{}.linkerFor(e)
+ linker.SetNext(nil)
+ linker.SetPrev(l.tail)
+ if l.tail != nil {
+ processGroupElementMapper{}.linkerFor(l.tail).SetNext(e)
+ } else {
+ l.head = e
+ }
+
+ l.tail = e
+}
+
+// PushBackList inserts list m at the end of list l, emptying m.
+func (l *processGroupList) PushBackList(m *processGroupList) {
+ if l.head == nil {
+ l.head = m.head
+ l.tail = m.tail
+ } else if m.head != nil {
+ processGroupElementMapper{}.linkerFor(l.tail).SetNext(m.head)
+ processGroupElementMapper{}.linkerFor(m.head).SetPrev(l.tail)
+
+ l.tail = m.tail
+ }
+ m.head = nil
+ m.tail = nil
+}
+
+// InsertAfter inserts e after b.
+func (l *processGroupList) InsertAfter(b, e *ProcessGroup) {
+ bLinker := processGroupElementMapper{}.linkerFor(b)
+ eLinker := processGroupElementMapper{}.linkerFor(e)
+
+ a := bLinker.Next()
+
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ bLinker.SetNext(e)
+
+ if a != nil {
+ processGroupElementMapper{}.linkerFor(a).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+}
+
+// InsertBefore inserts e before a.
+func (l *processGroupList) InsertBefore(a, e *ProcessGroup) {
+ aLinker := processGroupElementMapper{}.linkerFor(a)
+ eLinker := processGroupElementMapper{}.linkerFor(e)
+
+ b := aLinker.Prev()
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ aLinker.SetPrev(e)
+
+ if b != nil {
+ processGroupElementMapper{}.linkerFor(b).SetNext(e)
+ } else {
+ l.head = e
+ }
+}
+
+// Remove removes e from l.
+func (l *processGroupList) Remove(e *ProcessGroup) {
+ linker := processGroupElementMapper{}.linkerFor(e)
+ prev := linker.Prev()
+ next := linker.Next()
+
+ if prev != nil {
+ processGroupElementMapper{}.linkerFor(prev).SetNext(next)
+ } else if l.head == e {
+ l.head = next
+ }
+
+ if next != nil {
+ processGroupElementMapper{}.linkerFor(next).SetPrev(prev)
+ } else if l.tail == e {
+ l.tail = prev
+ }
+
+ linker.SetNext(nil)
+ linker.SetPrev(nil)
+}
+
+// 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 processGroupEntry struct {
+ next *ProcessGroup
+ prev *ProcessGroup
+}
+
+// Next returns the entry that follows e in the list.
+func (e *processGroupEntry) Next() *ProcessGroup {
+ return e.next
+}
+
+// Prev returns the entry that precedes e in the list.
+func (e *processGroupEntry) Prev() *ProcessGroup {
+ return e.prev
+}
+
+// SetNext assigns 'entry' as the entry that follows e in the list.
+func (e *processGroupEntry) SetNext(elem *ProcessGroup) {
+ e.next = elem
+}
+
+// SetPrev assigns 'entry' as the entry that precedes e in the list.
+func (e *processGroupEntry) SetPrev(elem *ProcessGroup) {
+ e.prev = elem
+}
diff --git a/pkg/sentry/kernel/process_group_refs.go b/pkg/sentry/kernel/process_group_refs.go
new file mode 100644
index 000000000..4b257d548
--- /dev/null
+++ b/pkg/sentry/kernel/process_group_refs.go
@@ -0,0 +1,117 @@
+package kernel
+
+import (
+ "fmt"
+ "gvisor.dev/gvisor/pkg/log"
+ refs_vfs1 "gvisor.dev/gvisor/pkg/refs"
+ "runtime"
+ "sync/atomic"
+)
+
+// ownerType is used to customize logging. Note that we use a pointer to T so
+// that we do not copy the entire object when passed as a format parameter.
+var ProcessGroupownerType *ProcessGroup
+
+// Refs implements refs.RefCounter. It keeps a reference count using atomic
+// operations and calls the destructor when the count reaches zero.
+//
+// Note that the number of references is actually refCount + 1 so that a default
+// zero-value Refs object contains one reference.
+//
+// TODO(gvisor.dev/issue/1486): Store stack traces when leak check is enabled in
+// a map with 16-bit hashes, and store the hash in the top 16 bits of refCount.
+// This will allow us to add stack trace information to the leak messages
+// without growing the size of Refs.
+//
+// +stateify savable
+type ProcessGroupRefs struct {
+ // refCount is composed of two fields:
+ //
+ // [32-bit speculative references]:[32-bit real references]
+ //
+ // Speculative references are used for TryIncRef, to avoid a CompareAndSwap
+ // loop. See IncRef, DecRef and TryIncRef for details of how these fields are
+ // used.
+ refCount int64
+}
+
+func (r *ProcessGroupRefs) finalize() {
+ var note string
+ switch refs_vfs1.GetLeakMode() {
+ case refs_vfs1.NoLeakChecking:
+ return
+ case refs_vfs1.UninitializedLeakChecking:
+ note = "(Leak checker uninitialized): "
+ }
+ if n := r.ReadRefs(); n != 0 {
+ log.Warningf("%sRefs %p owned by %T garbage collected with ref count of %d (want 0)", note, r, ProcessGroupownerType, n)
+ }
+}
+
+// EnableLeakCheck checks for reference leaks when Refs gets garbage collected.
+func (r *ProcessGroupRefs) EnableLeakCheck() {
+ if refs_vfs1.GetLeakMode() != refs_vfs1.NoLeakChecking {
+ runtime.SetFinalizer(r, (*ProcessGroupRefs).finalize)
+ }
+}
+
+// ReadRefs returns the current number of references. The returned count is
+// inherently racy and is unsafe to use without external synchronization.
+func (r *ProcessGroupRefs) ReadRefs() int64 {
+
+ return atomic.LoadInt64(&r.refCount) + 1
+}
+
+// IncRef implements refs.RefCounter.IncRef.
+//
+//go:nosplit
+func (r *ProcessGroupRefs) IncRef() {
+ if v := atomic.AddInt64(&r.refCount, 1); v <= 0 {
+ panic(fmt.Sprintf("Incrementing non-positive ref count %p owned by %T", r, ProcessGroupownerType))
+ }
+}
+
+// TryIncRef implements refs.RefCounter.TryIncRef.
+//
+// To do this safely without a loop, a speculative reference is first acquired
+// on the object. This allows multiple concurrent TryIncRef calls to distinguish
+// other TryIncRef calls from genuine references held.
+//
+//go:nosplit
+func (r *ProcessGroupRefs) TryIncRef() bool {
+ const speculativeRef = 1 << 32
+ v := atomic.AddInt64(&r.refCount, speculativeRef)
+ if int32(v) < 0 {
+
+ atomic.AddInt64(&r.refCount, -speculativeRef)
+ return false
+ }
+
+ atomic.AddInt64(&r.refCount, -speculativeRef+1)
+ return true
+}
+
+// DecRef implements refs.RefCounter.DecRef.
+//
+// Note that speculative references are counted here. Since they were added
+// prior to real references reaching zero, they will successfully convert to
+// real references. In other words, we see speculative references only in the
+// following case:
+//
+// A: TryIncRef [speculative increase => sees non-negative references]
+// B: DecRef [real decrease]
+// A: TryIncRef [transform speculative to real]
+//
+//go:nosplit
+func (r *ProcessGroupRefs) DecRef(destroy func()) {
+ switch v := atomic.AddInt64(&r.refCount, -1); {
+ case v < -1:
+ panic(fmt.Sprintf("Decrementing non-positive ref count %p, owned by %T", r, ProcessGroupownerType))
+
+ case v == -1:
+
+ if destroy != nil {
+ destroy()
+ }
+ }
+}
diff --git a/pkg/sentry/kernel/sched/BUILD b/pkg/sentry/kernel/sched/BUILD
deleted file mode 100644
index 1b82e087b..000000000
--- a/pkg/sentry/kernel/sched/BUILD
+++ /dev/null
@@ -1,19 +0,0 @@
-load("//tools:defs.bzl", "go_library", "go_test")
-
-package(licenses = ["notice"])
-
-go_library(
- name = "sched",
- srcs = [
- "cpuset.go",
- "sched.go",
- ],
- visibility = ["//pkg/sentry:internal"],
-)
-
-go_test(
- name = "sched_test",
- size = "small",
- srcs = ["cpuset_test.go"],
- library = ":sched",
-)
diff --git a/pkg/sentry/kernel/sched/cpuset_test.go b/pkg/sentry/kernel/sched/cpuset_test.go
deleted file mode 100644
index 3af9f1197..000000000
--- a/pkg/sentry/kernel/sched/cpuset_test.go
+++ /dev/null
@@ -1,44 +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 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_state_autogen.go b/pkg/sentry/kernel/sched/sched_state_autogen.go
new file mode 100644
index 000000000..9705ca79d
--- /dev/null
+++ b/pkg/sentry/kernel/sched/sched_state_autogen.go
@@ -0,0 +1,3 @@
+// automatically generated by stateify.
+
+package sched
diff --git a/pkg/sentry/kernel/semaphore/BUILD b/pkg/sentry/kernel/semaphore/BUILD
deleted file mode 100644
index 65e5427c1..000000000
--- a/pkg/sentry/kernel/semaphore/BUILD
+++ /dev/null
@@ -1,49 +0,0 @@
-load("//tools:defs.bzl", "go_library", "go_test")
-load("//tools/go_generics:defs.bzl", "go_template_instance")
-
-package(licenses = ["notice"])
-
-go_template_instance(
- name = "waiter_list",
- out = "waiter_list.go",
- package = "semaphore",
- prefix = "waiter",
- template = "//pkg/ilist:generic_list",
- types = {
- "Element": "*waiter",
- "Linker": "*waiter",
- },
-)
-
-go_library(
- name = "semaphore",
- srcs = [
- "semaphore.go",
- "waiter_list.go",
- ],
- visibility = ["//pkg/sentry:internal"],
- deps = [
- "//pkg/abi/linux",
- "//pkg/context",
- "//pkg/log",
- "//pkg/sentry/fs",
- "//pkg/sentry/kernel/auth",
- "//pkg/sentry/kernel/time",
- "//pkg/sync",
- "//pkg/syserror",
- ],
-)
-
-go_test(
- name = "semaphore_test",
- size = "small",
- srcs = ["semaphore_test.go"],
- library = ":semaphore",
- deps = [
- "//pkg/abi/linux",
- "//pkg/context",
- "//pkg/sentry/contexttest",
- "//pkg/sentry/kernel/auth",
- "//pkg/syserror",
- ],
-)
diff --git a/pkg/sentry/kernel/semaphore/semaphore_state_autogen.go b/pkg/sentry/kernel/semaphore/semaphore_state_autogen.go
new file mode 100644
index 000000000..fe5adfbee
--- /dev/null
+++ b/pkg/sentry/kernel/semaphore/semaphore_state_autogen.go
@@ -0,0 +1,205 @@
+// automatically generated by stateify.
+
+package semaphore
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *Registry) StateTypeName() string {
+ return "pkg/sentry/kernel/semaphore.Registry"
+}
+
+func (x *Registry) StateFields() []string {
+ return []string{
+ "userNS",
+ "semaphores",
+ "lastIDUsed",
+ }
+}
+
+func (x *Registry) beforeSave() {}
+
+func (x *Registry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.userNS)
+ m.Save(1, &x.semaphores)
+ m.Save(2, &x.lastIDUsed)
+}
+
+func (x *Registry) afterLoad() {}
+
+func (x *Registry) StateLoad(m state.Source) {
+ m.Load(0, &x.userNS)
+ m.Load(1, &x.semaphores)
+ m.Load(2, &x.lastIDUsed)
+}
+
+func (x *Set) StateTypeName() string {
+ return "pkg/sentry/kernel/semaphore.Set"
+}
+
+func (x *Set) StateFields() []string {
+ return []string{
+ "registry",
+ "ID",
+ "key",
+ "creator",
+ "owner",
+ "perms",
+ "opTime",
+ "changeTime",
+ "sems",
+ "dead",
+ }
+}
+
+func (x *Set) beforeSave() {}
+
+func (x *Set) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.registry)
+ m.Save(1, &x.ID)
+ m.Save(2, &x.key)
+ m.Save(3, &x.creator)
+ m.Save(4, &x.owner)
+ m.Save(5, &x.perms)
+ m.Save(6, &x.opTime)
+ m.Save(7, &x.changeTime)
+ m.Save(8, &x.sems)
+ m.Save(9, &x.dead)
+}
+
+func (x *Set) afterLoad() {}
+
+func (x *Set) StateLoad(m state.Source) {
+ m.Load(0, &x.registry)
+ m.Load(1, &x.ID)
+ m.Load(2, &x.key)
+ m.Load(3, &x.creator)
+ m.Load(4, &x.owner)
+ m.Load(5, &x.perms)
+ m.Load(6, &x.opTime)
+ m.Load(7, &x.changeTime)
+ m.Load(8, &x.sems)
+ m.Load(9, &x.dead)
+}
+
+func (x *sem) StateTypeName() string {
+ return "pkg/sentry/kernel/semaphore.sem"
+}
+
+func (x *sem) StateFields() []string {
+ return []string{
+ "value",
+ "pid",
+ }
+}
+
+func (x *sem) beforeSave() {}
+
+func (x *sem) StateSave(m state.Sink) {
+ x.beforeSave()
+ if !state.IsZeroValue(&x.waiters) {
+ state.Failf("waiters is %#v, expected zero", &x.waiters)
+ }
+ m.Save(0, &x.value)
+ m.Save(1, &x.pid)
+}
+
+func (x *sem) afterLoad() {}
+
+func (x *sem) StateLoad(m state.Source) {
+ m.Load(0, &x.value)
+ m.Load(1, &x.pid)
+}
+
+func (x *waiter) StateTypeName() string {
+ return "pkg/sentry/kernel/semaphore.waiter"
+}
+
+func (x *waiter) StateFields() []string {
+ return []string{
+ "waiterEntry",
+ "value",
+ "ch",
+ }
+}
+
+func (x *waiter) beforeSave() {}
+
+func (x *waiter) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.waiterEntry)
+ m.Save(1, &x.value)
+ m.Save(2, &x.ch)
+}
+
+func (x *waiter) afterLoad() {}
+
+func (x *waiter) StateLoad(m state.Source) {
+ m.Load(0, &x.waiterEntry)
+ m.Load(1, &x.value)
+ m.Load(2, &x.ch)
+}
+
+func (x *waiterList) StateTypeName() string {
+ return "pkg/sentry/kernel/semaphore.waiterList"
+}
+
+func (x *waiterList) StateFields() []string {
+ return []string{
+ "head",
+ "tail",
+ }
+}
+
+func (x *waiterList) beforeSave() {}
+
+func (x *waiterList) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.head)
+ m.Save(1, &x.tail)
+}
+
+func (x *waiterList) afterLoad() {}
+
+func (x *waiterList) StateLoad(m state.Source) {
+ m.Load(0, &x.head)
+ m.Load(1, &x.tail)
+}
+
+func (x *waiterEntry) StateTypeName() string {
+ return "pkg/sentry/kernel/semaphore.waiterEntry"
+}
+
+func (x *waiterEntry) StateFields() []string {
+ return []string{
+ "next",
+ "prev",
+ }
+}
+
+func (x *waiterEntry) beforeSave() {}
+
+func (x *waiterEntry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.next)
+ m.Save(1, &x.prev)
+}
+
+func (x *waiterEntry) afterLoad() {}
+
+func (x *waiterEntry) StateLoad(m state.Source) {
+ m.Load(0, &x.next)
+ m.Load(1, &x.prev)
+}
+
+func init() {
+ state.Register((*Registry)(nil))
+ state.Register((*Set)(nil))
+ state.Register((*sem)(nil))
+ state.Register((*waiter)(nil))
+ state.Register((*waiterList)(nil))
+ state.Register((*waiterEntry)(nil))
+}
diff --git a/pkg/sentry/kernel/semaphore/semaphore_test.go b/pkg/sentry/kernel/semaphore/semaphore_test.go
deleted file mode 100644
index e47acefdf..000000000
--- a/pkg/sentry/kernel/semaphore/semaphore_test.go
+++ /dev/null
@@ -1,172 +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 semaphore
-
-import (
- "testing"
-
- "gvisor.dev/gvisor/pkg/abi/linux"
- "gvisor.dev/gvisor/pkg/context"
- "gvisor.dev/gvisor/pkg/sentry/contexttest"
- "gvisor.dev/gvisor/pkg/sentry/kernel/auth"
- "gvisor.dev/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, 123)
- 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{
- {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{
- {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{
- {SemOp: 1},
- }
- executeOps(ctx, t, set, ops, false)
-
- ops[0].SemOp = -2
- ops[0].SemFlg = linux.IPC_NOWAIT
- 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, 123); 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(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)
- }
- 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{
- {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/semaphore/waiter_list.go b/pkg/sentry/kernel/semaphore/waiter_list.go
new file mode 100644
index 000000000..007bf67b0
--- /dev/null
+++ b/pkg/sentry/kernel/semaphore/waiter_list.go
@@ -0,0 +1,193 @@
+package semaphore
+
+// 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 waiterElementMapper struct{}
+
+// linkerFor maps an Element to a Linker.
+//
+// This default implementation should be inlined.
+//
+//go:nosplit
+func (waiterElementMapper) linkerFor(elem *waiter) *waiter { 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.
+//
+// The zero value for List is an empty list ready to use.
+//
+// To iterate over a list (where l is a List):
+// for e := l.Front(); e != nil; e = e.Next() {
+// // do something with e.
+// }
+//
+// +stateify savable
+type waiterList struct {
+ head *waiter
+ tail *waiter
+}
+
+// Reset resets list l to the empty state.
+func (l *waiterList) Reset() {
+ l.head = nil
+ l.tail = nil
+}
+
+// Empty returns true iff the list is empty.
+func (l *waiterList) Empty() bool {
+ return l.head == nil
+}
+
+// Front returns the first element of list l or nil.
+func (l *waiterList) Front() *waiter {
+ return l.head
+}
+
+// Back returns the last element of list l or nil.
+func (l *waiterList) Back() *waiter {
+ return l.tail
+}
+
+// Len returns the number of elements in the list.
+//
+// NOTE: This is an O(n) operation.
+func (l *waiterList) Len() (count int) {
+ for e := l.Front(); e != nil; e = (waiterElementMapper{}.linkerFor(e)).Next() {
+ count++
+ }
+ return count
+}
+
+// PushFront inserts the element e at the front of list l.
+func (l *waiterList) PushFront(e *waiter) {
+ linker := waiterElementMapper{}.linkerFor(e)
+ linker.SetNext(l.head)
+ linker.SetPrev(nil)
+ if l.head != nil {
+ waiterElementMapper{}.linkerFor(l.head).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+
+ l.head = e
+}
+
+// PushBack inserts the element e at the back of list l.
+func (l *waiterList) PushBack(e *waiter) {
+ linker := waiterElementMapper{}.linkerFor(e)
+ linker.SetNext(nil)
+ linker.SetPrev(l.tail)
+ if l.tail != nil {
+ waiterElementMapper{}.linkerFor(l.tail).SetNext(e)
+ } else {
+ l.head = e
+ }
+
+ l.tail = e
+}
+
+// PushBackList inserts list m at the end of list l, emptying m.
+func (l *waiterList) PushBackList(m *waiterList) {
+ if l.head == nil {
+ l.head = m.head
+ l.tail = m.tail
+ } else if m.head != nil {
+ waiterElementMapper{}.linkerFor(l.tail).SetNext(m.head)
+ waiterElementMapper{}.linkerFor(m.head).SetPrev(l.tail)
+
+ l.tail = m.tail
+ }
+ m.head = nil
+ m.tail = nil
+}
+
+// InsertAfter inserts e after b.
+func (l *waiterList) InsertAfter(b, e *waiter) {
+ bLinker := waiterElementMapper{}.linkerFor(b)
+ eLinker := waiterElementMapper{}.linkerFor(e)
+
+ a := bLinker.Next()
+
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ bLinker.SetNext(e)
+
+ if a != nil {
+ waiterElementMapper{}.linkerFor(a).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+}
+
+// InsertBefore inserts e before a.
+func (l *waiterList) InsertBefore(a, e *waiter) {
+ aLinker := waiterElementMapper{}.linkerFor(a)
+ eLinker := waiterElementMapper{}.linkerFor(e)
+
+ b := aLinker.Prev()
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ aLinker.SetPrev(e)
+
+ if b != nil {
+ waiterElementMapper{}.linkerFor(b).SetNext(e)
+ } else {
+ l.head = e
+ }
+}
+
+// Remove removes e from l.
+func (l *waiterList) Remove(e *waiter) {
+ linker := waiterElementMapper{}.linkerFor(e)
+ prev := linker.Prev()
+ next := linker.Next()
+
+ if prev != nil {
+ waiterElementMapper{}.linkerFor(prev).SetNext(next)
+ } else if l.head == e {
+ l.head = next
+ }
+
+ if next != nil {
+ waiterElementMapper{}.linkerFor(next).SetPrev(prev)
+ } else if l.tail == e {
+ l.tail = prev
+ }
+
+ linker.SetNext(nil)
+ linker.SetPrev(nil)
+}
+
+// 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 waiterEntry struct {
+ next *waiter
+ prev *waiter
+}
+
+// Next returns the entry that follows e in the list.
+func (e *waiterEntry) Next() *waiter {
+ return e.next
+}
+
+// Prev returns the entry that precedes e in the list.
+func (e *waiterEntry) Prev() *waiter {
+ return e.prev
+}
+
+// SetNext assigns 'entry' as the entry that follows e in the list.
+func (e *waiterEntry) SetNext(elem *waiter) {
+ e.next = elem
+}
+
+// SetPrev assigns 'entry' as the entry that precedes e in the list.
+func (e *waiterEntry) SetPrev(elem *waiter) {
+ e.prev = elem
+}
diff --git a/pkg/sentry/kernel/seqatomic_taskgoroutineschedinfo_unsafe.go b/pkg/sentry/kernel/seqatomic_taskgoroutineschedinfo_unsafe.go
new file mode 100644
index 000000000..a0f2fe45c
--- /dev/null
+++ b/pkg/sentry/kernel/seqatomic_taskgoroutineschedinfo_unsafe.go
@@ -0,0 +1,49 @@
+package kernel
+
+import (
+ "unsafe"
+
+ "fmt"
+ "gvisor.dev/gvisor/pkg/sync"
+ "reflect"
+ "strings"
+)
+
+// SeqAtomicLoad returns a copy of *ptr, ensuring that the read does not race
+// with any writer critical sections in seq.
+//
+//go:nosplit
+func SeqAtomicLoadTaskGoroutineSchedInfo(seq *sync.SeqCount, ptr *TaskGoroutineSchedInfo) TaskGoroutineSchedInfo {
+ for {
+ if val, ok := SeqAtomicTryLoadTaskGoroutineSchedInfo(seq, seq.BeginRead(), ptr); ok {
+ return val
+ }
+ }
+}
+
+// SeqAtomicTryLoad returns a copy of *ptr while in a reader critical section
+// in seq initiated by a call to seq.BeginRead() that returned epoch. If the
+// read would race with a writer critical section, SeqAtomicTryLoad returns
+// (unspecified, false).
+//
+//go:nosplit
+func SeqAtomicTryLoadTaskGoroutineSchedInfo(seq *sync.SeqCount, epoch sync.SeqCountEpoch, ptr *TaskGoroutineSchedInfo) (val TaskGoroutineSchedInfo, ok bool) {
+ if sync.RaceEnabled {
+
+ sync.Memmove(unsafe.Pointer(&val), unsafe.Pointer(ptr), unsafe.Sizeof(val))
+ } else {
+
+ val = *ptr
+ }
+ ok = seq.ReadOk(epoch)
+ return
+}
+
+func initTaskGoroutineSchedInfo() {
+ var val TaskGoroutineSchedInfo
+ typ := reflect.TypeOf(val)
+ name := typ.Name()
+ if ptrs := sync.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/sentry/kernel/session_list.go b/pkg/sentry/kernel/session_list.go
new file mode 100644
index 000000000..c328b2526
--- /dev/null
+++ b/pkg/sentry/kernel/session_list.go
@@ -0,0 +1,193 @@
+package kernel
+
+// 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 sessionElementMapper struct{}
+
+// linkerFor maps an Element to a Linker.
+//
+// This default implementation should be inlined.
+//
+//go:nosplit
+func (sessionElementMapper) linkerFor(elem *Session) *Session { 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.
+//
+// The zero value for List is an empty list ready to use.
+//
+// To iterate over a list (where l is a List):
+// for e := l.Front(); e != nil; e = e.Next() {
+// // do something with e.
+// }
+//
+// +stateify savable
+type sessionList struct {
+ head *Session
+ tail *Session
+}
+
+// Reset resets list l to the empty state.
+func (l *sessionList) Reset() {
+ l.head = nil
+ l.tail = nil
+}
+
+// Empty returns true iff the list is empty.
+func (l *sessionList) Empty() bool {
+ return l.head == nil
+}
+
+// Front returns the first element of list l or nil.
+func (l *sessionList) Front() *Session {
+ return l.head
+}
+
+// Back returns the last element of list l or nil.
+func (l *sessionList) Back() *Session {
+ return l.tail
+}
+
+// Len returns the number of elements in the list.
+//
+// NOTE: This is an O(n) operation.
+func (l *sessionList) Len() (count int) {
+ for e := l.Front(); e != nil; e = (sessionElementMapper{}.linkerFor(e)).Next() {
+ count++
+ }
+ return count
+}
+
+// PushFront inserts the element e at the front of list l.
+func (l *sessionList) PushFront(e *Session) {
+ linker := sessionElementMapper{}.linkerFor(e)
+ linker.SetNext(l.head)
+ linker.SetPrev(nil)
+ if l.head != nil {
+ sessionElementMapper{}.linkerFor(l.head).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+
+ l.head = e
+}
+
+// PushBack inserts the element e at the back of list l.
+func (l *sessionList) PushBack(e *Session) {
+ linker := sessionElementMapper{}.linkerFor(e)
+ linker.SetNext(nil)
+ linker.SetPrev(l.tail)
+ if l.tail != nil {
+ sessionElementMapper{}.linkerFor(l.tail).SetNext(e)
+ } else {
+ l.head = e
+ }
+
+ l.tail = e
+}
+
+// PushBackList inserts list m at the end of list l, emptying m.
+func (l *sessionList) PushBackList(m *sessionList) {
+ if l.head == nil {
+ l.head = m.head
+ l.tail = m.tail
+ } else if m.head != nil {
+ sessionElementMapper{}.linkerFor(l.tail).SetNext(m.head)
+ sessionElementMapper{}.linkerFor(m.head).SetPrev(l.tail)
+
+ l.tail = m.tail
+ }
+ m.head = nil
+ m.tail = nil
+}
+
+// InsertAfter inserts e after b.
+func (l *sessionList) InsertAfter(b, e *Session) {
+ bLinker := sessionElementMapper{}.linkerFor(b)
+ eLinker := sessionElementMapper{}.linkerFor(e)
+
+ a := bLinker.Next()
+
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ bLinker.SetNext(e)
+
+ if a != nil {
+ sessionElementMapper{}.linkerFor(a).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+}
+
+// InsertBefore inserts e before a.
+func (l *sessionList) InsertBefore(a, e *Session) {
+ aLinker := sessionElementMapper{}.linkerFor(a)
+ eLinker := sessionElementMapper{}.linkerFor(e)
+
+ b := aLinker.Prev()
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ aLinker.SetPrev(e)
+
+ if b != nil {
+ sessionElementMapper{}.linkerFor(b).SetNext(e)
+ } else {
+ l.head = e
+ }
+}
+
+// Remove removes e from l.
+func (l *sessionList) Remove(e *Session) {
+ linker := sessionElementMapper{}.linkerFor(e)
+ prev := linker.Prev()
+ next := linker.Next()
+
+ if prev != nil {
+ sessionElementMapper{}.linkerFor(prev).SetNext(next)
+ } else if l.head == e {
+ l.head = next
+ }
+
+ if next != nil {
+ sessionElementMapper{}.linkerFor(next).SetPrev(prev)
+ } else if l.tail == e {
+ l.tail = prev
+ }
+
+ linker.SetNext(nil)
+ linker.SetPrev(nil)
+}
+
+// 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 sessionEntry struct {
+ next *Session
+ prev *Session
+}
+
+// Next returns the entry that follows e in the list.
+func (e *sessionEntry) Next() *Session {
+ return e.next
+}
+
+// Prev returns the entry that precedes e in the list.
+func (e *sessionEntry) Prev() *Session {
+ return e.prev
+}
+
+// SetNext assigns 'entry' as the entry that follows e in the list.
+func (e *sessionEntry) SetNext(elem *Session) {
+ e.next = elem
+}
+
+// SetPrev assigns 'entry' as the entry that precedes e in the list.
+func (e *sessionEntry) SetPrev(elem *Session) {
+ e.prev = elem
+}
diff --git a/pkg/sentry/kernel/session_refs.go b/pkg/sentry/kernel/session_refs.go
new file mode 100644
index 000000000..204fdd060
--- /dev/null
+++ b/pkg/sentry/kernel/session_refs.go
@@ -0,0 +1,117 @@
+package kernel
+
+import (
+ "fmt"
+ "gvisor.dev/gvisor/pkg/log"
+ refs_vfs1 "gvisor.dev/gvisor/pkg/refs"
+ "runtime"
+ "sync/atomic"
+)
+
+// ownerType is used to customize logging. Note that we use a pointer to T so
+// that we do not copy the entire object when passed as a format parameter.
+var SessionownerType *Session
+
+// Refs implements refs.RefCounter. It keeps a reference count using atomic
+// operations and calls the destructor when the count reaches zero.
+//
+// Note that the number of references is actually refCount + 1 so that a default
+// zero-value Refs object contains one reference.
+//
+// TODO(gvisor.dev/issue/1486): Store stack traces when leak check is enabled in
+// a map with 16-bit hashes, and store the hash in the top 16 bits of refCount.
+// This will allow us to add stack trace information to the leak messages
+// without growing the size of Refs.
+//
+// +stateify savable
+type SessionRefs struct {
+ // refCount is composed of two fields:
+ //
+ // [32-bit speculative references]:[32-bit real references]
+ //
+ // Speculative references are used for TryIncRef, to avoid a CompareAndSwap
+ // loop. See IncRef, DecRef and TryIncRef for details of how these fields are
+ // used.
+ refCount int64
+}
+
+func (r *SessionRefs) finalize() {
+ var note string
+ switch refs_vfs1.GetLeakMode() {
+ case refs_vfs1.NoLeakChecking:
+ return
+ case refs_vfs1.UninitializedLeakChecking:
+ note = "(Leak checker uninitialized): "
+ }
+ if n := r.ReadRefs(); n != 0 {
+ log.Warningf("%sRefs %p owned by %T garbage collected with ref count of %d (want 0)", note, r, SessionownerType, n)
+ }
+}
+
+// EnableLeakCheck checks for reference leaks when Refs gets garbage collected.
+func (r *SessionRefs) EnableLeakCheck() {
+ if refs_vfs1.GetLeakMode() != refs_vfs1.NoLeakChecking {
+ runtime.SetFinalizer(r, (*SessionRefs).finalize)
+ }
+}
+
+// ReadRefs returns the current number of references. The returned count is
+// inherently racy and is unsafe to use without external synchronization.
+func (r *SessionRefs) ReadRefs() int64 {
+
+ return atomic.LoadInt64(&r.refCount) + 1
+}
+
+// IncRef implements refs.RefCounter.IncRef.
+//
+//go:nosplit
+func (r *SessionRefs) IncRef() {
+ if v := atomic.AddInt64(&r.refCount, 1); v <= 0 {
+ panic(fmt.Sprintf("Incrementing non-positive ref count %p owned by %T", r, SessionownerType))
+ }
+}
+
+// TryIncRef implements refs.RefCounter.TryIncRef.
+//
+// To do this safely without a loop, a speculative reference is first acquired
+// on the object. This allows multiple concurrent TryIncRef calls to distinguish
+// other TryIncRef calls from genuine references held.
+//
+//go:nosplit
+func (r *SessionRefs) TryIncRef() bool {
+ const speculativeRef = 1 << 32
+ v := atomic.AddInt64(&r.refCount, speculativeRef)
+ if int32(v) < 0 {
+
+ atomic.AddInt64(&r.refCount, -speculativeRef)
+ return false
+ }
+
+ atomic.AddInt64(&r.refCount, -speculativeRef+1)
+ return true
+}
+
+// DecRef implements refs.RefCounter.DecRef.
+//
+// Note that speculative references are counted here. Since they were added
+// prior to real references reaching zero, they will successfully convert to
+// real references. In other words, we see speculative references only in the
+// following case:
+//
+// A: TryIncRef [speculative increase => sees non-negative references]
+// B: DecRef [real decrease]
+// A: TryIncRef [transform speculative to real]
+//
+//go:nosplit
+func (r *SessionRefs) DecRef(destroy func()) {
+ switch v := atomic.AddInt64(&r.refCount, -1); {
+ case v < -1:
+ panic(fmt.Sprintf("Decrementing non-positive ref count %p, owned by %T", r, SessionownerType))
+
+ case v == -1:
+
+ if destroy != nil {
+ destroy()
+ }
+ }
+}
diff --git a/pkg/sentry/kernel/shm/BUILD b/pkg/sentry/kernel/shm/BUILD
deleted file mode 100644
index b7e4b480d..000000000
--- a/pkg/sentry/kernel/shm/BUILD
+++ /dev/null
@@ -1,41 +0,0 @@
-load("//tools:defs.bzl", "go_library")
-load("//tools/go_generics:defs.bzl", "go_template_instance")
-
-package(licenses = ["notice"])
-
-go_template_instance(
- name = "shm_refs",
- out = "shm_refs.go",
- package = "shm",
- prefix = "Shm",
- template = "//pkg/refs_vfs2:refs_template",
- types = {
- "T": "Shm",
- },
-)
-
-go_library(
- name = "shm",
- srcs = [
- "device.go",
- "shm.go",
- "shm_refs.go",
- ],
- visibility = ["//pkg/sentry:internal"],
- deps = [
- "//pkg/abi/linux",
- "//pkg/context",
- "//pkg/log",
- "//pkg/refs",
- "//pkg/sentry/device",
- "//pkg/sentry/fs",
- "//pkg/sentry/kernel/auth",
- "//pkg/sentry/kernel/time",
- "//pkg/sentry/memmap",
- "//pkg/sentry/pgalloc",
- "//pkg/sentry/usage",
- "//pkg/sync",
- "//pkg/syserror",
- "//pkg/usermem",
- ],
-)
diff --git a/pkg/sentry/kernel/shm/shm_refs.go b/pkg/sentry/kernel/shm/shm_refs.go
new file mode 100644
index 000000000..4bffdd0b3
--- /dev/null
+++ b/pkg/sentry/kernel/shm/shm_refs.go
@@ -0,0 +1,117 @@
+package shm
+
+import (
+ "fmt"
+ "gvisor.dev/gvisor/pkg/log"
+ refs_vfs1 "gvisor.dev/gvisor/pkg/refs"
+ "runtime"
+ "sync/atomic"
+)
+
+// ownerType is used to customize logging. Note that we use a pointer to T so
+// that we do not copy the entire object when passed as a format parameter.
+var ShmownerType *Shm
+
+// Refs implements refs.RefCounter. It keeps a reference count using atomic
+// operations and calls the destructor when the count reaches zero.
+//
+// Note that the number of references is actually refCount + 1 so that a default
+// zero-value Refs object contains one reference.
+//
+// TODO(gvisor.dev/issue/1486): Store stack traces when leak check is enabled in
+// a map with 16-bit hashes, and store the hash in the top 16 bits of refCount.
+// This will allow us to add stack trace information to the leak messages
+// without growing the size of Refs.
+//
+// +stateify savable
+type ShmRefs struct {
+ // refCount is composed of two fields:
+ //
+ // [32-bit speculative references]:[32-bit real references]
+ //
+ // Speculative references are used for TryIncRef, to avoid a CompareAndSwap
+ // loop. See IncRef, DecRef and TryIncRef for details of how these fields are
+ // used.
+ refCount int64
+}
+
+func (r *ShmRefs) finalize() {
+ var note string
+ switch refs_vfs1.GetLeakMode() {
+ case refs_vfs1.NoLeakChecking:
+ return
+ case refs_vfs1.UninitializedLeakChecking:
+ note = "(Leak checker uninitialized): "
+ }
+ if n := r.ReadRefs(); n != 0 {
+ log.Warningf("%sRefs %p owned by %T garbage collected with ref count of %d (want 0)", note, r, ShmownerType, n)
+ }
+}
+
+// EnableLeakCheck checks for reference leaks when Refs gets garbage collected.
+func (r *ShmRefs) EnableLeakCheck() {
+ if refs_vfs1.GetLeakMode() != refs_vfs1.NoLeakChecking {
+ runtime.SetFinalizer(r, (*ShmRefs).finalize)
+ }
+}
+
+// ReadRefs returns the current number of references. The returned count is
+// inherently racy and is unsafe to use without external synchronization.
+func (r *ShmRefs) ReadRefs() int64 {
+
+ return atomic.LoadInt64(&r.refCount) + 1
+}
+
+// IncRef implements refs.RefCounter.IncRef.
+//
+//go:nosplit
+func (r *ShmRefs) IncRef() {
+ if v := atomic.AddInt64(&r.refCount, 1); v <= 0 {
+ panic(fmt.Sprintf("Incrementing non-positive ref count %p owned by %T", r, ShmownerType))
+ }
+}
+
+// TryIncRef implements refs.RefCounter.TryIncRef.
+//
+// To do this safely without a loop, a speculative reference is first acquired
+// on the object. This allows multiple concurrent TryIncRef calls to distinguish
+// other TryIncRef calls from genuine references held.
+//
+//go:nosplit
+func (r *ShmRefs) TryIncRef() bool {
+ const speculativeRef = 1 << 32
+ v := atomic.AddInt64(&r.refCount, speculativeRef)
+ if int32(v) < 0 {
+
+ atomic.AddInt64(&r.refCount, -speculativeRef)
+ return false
+ }
+
+ atomic.AddInt64(&r.refCount, -speculativeRef+1)
+ return true
+}
+
+// DecRef implements refs.RefCounter.DecRef.
+//
+// Note that speculative references are counted here. Since they were added
+// prior to real references reaching zero, they will successfully convert to
+// real references. In other words, we see speculative references only in the
+// following case:
+//
+// A: TryIncRef [speculative increase => sees non-negative references]
+// B: DecRef [real decrease]
+// A: TryIncRef [transform speculative to real]
+//
+//go:nosplit
+func (r *ShmRefs) DecRef(destroy func()) {
+ switch v := atomic.AddInt64(&r.refCount, -1); {
+ case v < -1:
+ panic(fmt.Sprintf("Decrementing non-positive ref count %p, owned by %T", r, ShmownerType))
+
+ case v == -1:
+
+ if destroy != nil {
+ destroy()
+ }
+ }
+}
diff --git a/pkg/sentry/kernel/shm/shm_state_autogen.go b/pkg/sentry/kernel/shm/shm_state_autogen.go
new file mode 100644
index 000000000..fa80353b6
--- /dev/null
+++ b/pkg/sentry/kernel/shm/shm_state_autogen.go
@@ -0,0 +1,142 @@
+// automatically generated by stateify.
+
+package shm
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *Registry) StateTypeName() string {
+ return "pkg/sentry/kernel/shm.Registry"
+}
+
+func (x *Registry) StateFields() []string {
+ return []string{
+ "userNS",
+ "shms",
+ "keysToShms",
+ "totalPages",
+ "lastIDUsed",
+ }
+}
+
+func (x *Registry) beforeSave() {}
+
+func (x *Registry) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.userNS)
+ m.Save(1, &x.shms)
+ m.Save(2, &x.keysToShms)
+ m.Save(3, &x.totalPages)
+ m.Save(4, &x.lastIDUsed)
+}
+
+func (x *Registry) afterLoad() {}
+
+func (x *Registry) StateLoad(m state.Source) {
+ m.Load(0, &x.userNS)
+ m.Load(1, &x.shms)
+ m.Load(2, &x.keysToShms)
+ m.Load(3, &x.totalPages)
+ m.Load(4, &x.lastIDUsed)
+}
+
+func (x *Shm) StateTypeName() string {
+ return "pkg/sentry/kernel/shm.Shm"
+}
+
+func (x *Shm) StateFields() []string {
+ return []string{
+ "ShmRefs",
+ "mfp",
+ "registry",
+ "ID",
+ "creator",
+ "size",
+ "effectiveSize",
+ "fr",
+ "key",
+ "perms",
+ "owner",
+ "attachTime",
+ "detachTime",
+ "changeTime",
+ "creatorPID",
+ "lastAttachDetachPID",
+ "pendingDestruction",
+ }
+}
+
+func (x *Shm) beforeSave() {}
+
+func (x *Shm) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.ShmRefs)
+ m.Save(1, &x.mfp)
+ m.Save(2, &x.registry)
+ m.Save(3, &x.ID)
+ m.Save(4, &x.creator)
+ m.Save(5, &x.size)
+ m.Save(6, &x.effectiveSize)
+ m.Save(7, &x.fr)
+ m.Save(8, &x.key)
+ m.Save(9, &x.perms)
+ m.Save(10, &x.owner)
+ m.Save(11, &x.attachTime)
+ m.Save(12, &x.detachTime)
+ m.Save(13, &x.changeTime)
+ m.Save(14, &x.creatorPID)
+ m.Save(15, &x.lastAttachDetachPID)
+ m.Save(16, &x.pendingDestruction)
+}
+
+func (x *Shm) afterLoad() {}
+
+func (x *Shm) StateLoad(m state.Source) {
+ m.Load(0, &x.ShmRefs)
+ m.Load(1, &x.mfp)
+ m.Load(2, &x.registry)
+ m.Load(3, &x.ID)
+ m.Load(4, &x.creator)
+ m.Load(5, &x.size)
+ m.Load(6, &x.effectiveSize)
+ m.Load(7, &x.fr)
+ m.Load(8, &x.key)
+ m.Load(9, &x.perms)
+ m.Load(10, &x.owner)
+ m.Load(11, &x.attachTime)
+ m.Load(12, &x.detachTime)
+ m.Load(13, &x.changeTime)
+ m.Load(14, &x.creatorPID)
+ m.Load(15, &x.lastAttachDetachPID)
+ m.Load(16, &x.pendingDestruction)
+}
+
+func (x *ShmRefs) StateTypeName() string {
+ return "pkg/sentry/kernel/shm.ShmRefs"
+}
+
+func (x *ShmRefs) StateFields() []string {
+ return []string{
+ "refCount",
+ }
+}
+
+func (x *ShmRefs) beforeSave() {}
+
+func (x *ShmRefs) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.refCount)
+}
+
+func (x *ShmRefs) afterLoad() {}
+
+func (x *ShmRefs) StateLoad(m state.Source) {
+ m.Load(0, &x.refCount)
+}
+
+func init() {
+ state.Register((*Registry)(nil))
+ state.Register((*Shm)(nil))
+ state.Register((*ShmRefs)(nil))
+}
diff --git a/pkg/sentry/kernel/signalfd/BUILD b/pkg/sentry/kernel/signalfd/BUILD
deleted file mode 100644
index 3eb78e91b..000000000
--- a/pkg/sentry/kernel/signalfd/BUILD
+++ /dev/null
@@ -1,22 +0,0 @@
-load("//tools:defs.bzl", "go_library")
-
-licenses(["notice"])
-
-go_library(
- name = "signalfd",
- srcs = ["signalfd.go"],
- visibility = ["//pkg/sentry:internal"],
- deps = [
- "//pkg/abi/linux",
- "//pkg/binary",
- "//pkg/context",
- "//pkg/sentry/fs",
- "//pkg/sentry/fs/anon",
- "//pkg/sentry/fs/fsutil",
- "//pkg/sentry/kernel",
- "//pkg/sync",
- "//pkg/syserror",
- "//pkg/usermem",
- "//pkg/waiter",
- ],
-)
diff --git a/pkg/sentry/kernel/signalfd/signalfd_state_autogen.go b/pkg/sentry/kernel/signalfd/signalfd_state_autogen.go
new file mode 100644
index 000000000..089a35b60
--- /dev/null
+++ b/pkg/sentry/kernel/signalfd/signalfd_state_autogen.go
@@ -0,0 +1,37 @@
+// automatically generated by stateify.
+
+package signalfd
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *SignalOperations) StateTypeName() string {
+ return "pkg/sentry/kernel/signalfd.SignalOperations"
+}
+
+func (x *SignalOperations) StateFields() []string {
+ return []string{
+ "target",
+ "mask",
+ }
+}
+
+func (x *SignalOperations) beforeSave() {}
+
+func (x *SignalOperations) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.target)
+ m.Save(1, &x.mask)
+}
+
+func (x *SignalOperations) afterLoad() {}
+
+func (x *SignalOperations) StateLoad(m state.Source) {
+ m.Load(0, &x.target)
+ m.Load(1, &x.mask)
+}
+
+func init() {
+ state.Register((*SignalOperations)(nil))
+}
diff --git a/pkg/sentry/kernel/socket_list.go b/pkg/sentry/kernel/socket_list.go
new file mode 100644
index 000000000..d2d4307a1
--- /dev/null
+++ b/pkg/sentry/kernel/socket_list.go
@@ -0,0 +1,193 @@
+package kernel
+
+// 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 socketElementMapper struct{}
+
+// linkerFor maps an Element to a Linker.
+//
+// This default implementation should be inlined.
+//
+//go:nosplit
+func (socketElementMapper) linkerFor(elem *SocketEntry) *SocketEntry { 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.
+//
+// The zero value for List is an empty list ready to use.
+//
+// To iterate over a list (where l is a List):
+// for e := l.Front(); e != nil; e = e.Next() {
+// // do something with e.
+// }
+//
+// +stateify savable
+type socketList struct {
+ head *SocketEntry
+ tail *SocketEntry
+}
+
+// Reset resets list l to the empty state.
+func (l *socketList) Reset() {
+ l.head = nil
+ l.tail = nil
+}
+
+// Empty returns true iff the list is empty.
+func (l *socketList) Empty() bool {
+ return l.head == nil
+}
+
+// Front returns the first element of list l or nil.
+func (l *socketList) Front() *SocketEntry {
+ return l.head
+}
+
+// Back returns the last element of list l or nil.
+func (l *socketList) Back() *SocketEntry {
+ return l.tail
+}
+
+// Len returns the number of elements in the list.
+//
+// NOTE: This is an O(n) operation.
+func (l *socketList) Len() (count int) {
+ for e := l.Front(); e != nil; e = (socketElementMapper{}.linkerFor(e)).Next() {
+ count++
+ }
+ return count
+}
+
+// PushFront inserts the element e at the front of list l.
+func (l *socketList) PushFront(e *SocketEntry) {
+ linker := socketElementMapper{}.linkerFor(e)
+ linker.SetNext(l.head)
+ linker.SetPrev(nil)
+ if l.head != nil {
+ socketElementMapper{}.linkerFor(l.head).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+
+ l.head = e
+}
+
+// PushBack inserts the element e at the back of list l.
+func (l *socketList) PushBack(e *SocketEntry) {
+ linker := socketElementMapper{}.linkerFor(e)
+ linker.SetNext(nil)
+ linker.SetPrev(l.tail)
+ if l.tail != nil {
+ socketElementMapper{}.linkerFor(l.tail).SetNext(e)
+ } else {
+ l.head = e
+ }
+
+ l.tail = e
+}
+
+// PushBackList inserts list m at the end of list l, emptying m.
+func (l *socketList) PushBackList(m *socketList) {
+ if l.head == nil {
+ l.head = m.head
+ l.tail = m.tail
+ } else if m.head != nil {
+ socketElementMapper{}.linkerFor(l.tail).SetNext(m.head)
+ socketElementMapper{}.linkerFor(m.head).SetPrev(l.tail)
+
+ l.tail = m.tail
+ }
+ m.head = nil
+ m.tail = nil
+}
+
+// InsertAfter inserts e after b.
+func (l *socketList) InsertAfter(b, e *SocketEntry) {
+ bLinker := socketElementMapper{}.linkerFor(b)
+ eLinker := socketElementMapper{}.linkerFor(e)
+
+ a := bLinker.Next()
+
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ bLinker.SetNext(e)
+
+ if a != nil {
+ socketElementMapper{}.linkerFor(a).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+}
+
+// InsertBefore inserts e before a.
+func (l *socketList) InsertBefore(a, e *SocketEntry) {
+ aLinker := socketElementMapper{}.linkerFor(a)
+ eLinker := socketElementMapper{}.linkerFor(e)
+
+ b := aLinker.Prev()
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ aLinker.SetPrev(e)
+
+ if b != nil {
+ socketElementMapper{}.linkerFor(b).SetNext(e)
+ } else {
+ l.head = e
+ }
+}
+
+// Remove removes e from l.
+func (l *socketList) Remove(e *SocketEntry) {
+ linker := socketElementMapper{}.linkerFor(e)
+ prev := linker.Prev()
+ next := linker.Next()
+
+ if prev != nil {
+ socketElementMapper{}.linkerFor(prev).SetNext(next)
+ } else if l.head == e {
+ l.head = next
+ }
+
+ if next != nil {
+ socketElementMapper{}.linkerFor(next).SetPrev(prev)
+ } else if l.tail == e {
+ l.tail = prev
+ }
+
+ linker.SetNext(nil)
+ linker.SetPrev(nil)
+}
+
+// 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 socketEntry struct {
+ next *SocketEntry
+ prev *SocketEntry
+}
+
+// Next returns the entry that follows e in the list.
+func (e *socketEntry) Next() *SocketEntry {
+ return e.next
+}
+
+// Prev returns the entry that precedes e in the list.
+func (e *socketEntry) Prev() *SocketEntry {
+ return e.prev
+}
+
+// SetNext assigns 'entry' as the entry that follows e in the list.
+func (e *socketEntry) SetNext(elem *SocketEntry) {
+ e.next = elem
+}
+
+// SetPrev assigns 'entry' as the entry that precedes e in the list.
+func (e *socketEntry) SetPrev(elem *SocketEntry) {
+ e.prev = elem
+}
diff --git a/pkg/sentry/kernel/table_test.go b/pkg/sentry/kernel/table_test.go
deleted file mode 100644
index 32cf47e05..000000000
--- a/pkg/sentry/kernel/table_test.go
+++ /dev/null
@@ -1,110 +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 kernel
-
-import (
- "testing"
-
- "gvisor.dev/gvisor/pkg/abi"
- "gvisor.dev/gvisor/pkg/sentry/arch"
-)
-
-const (
- maxTestSyscall = 1000
-)
-
-func createSyscallTable() *SyscallTable {
- m := make(map[uintptr]Syscall)
- for i := uintptr(0); i <= maxTestSyscall; i++ {
- j := i
- m[i] = Syscall{
- Fn: 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_list.go b/pkg/sentry/kernel/task_list.go
new file mode 100644
index 000000000..f0c501845
--- /dev/null
+++ b/pkg/sentry/kernel/task_list.go
@@ -0,0 +1,193 @@
+package kernel
+
+// 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 taskElementMapper struct{}
+
+// linkerFor maps an Element to a Linker.
+//
+// This default implementation should be inlined.
+//
+//go:nosplit
+func (taskElementMapper) linkerFor(elem *Task) *Task { 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.
+//
+// The zero value for List is an empty list ready to use.
+//
+// To iterate over a list (where l is a List):
+// for e := l.Front(); e != nil; e = e.Next() {
+// // do something with e.
+// }
+//
+// +stateify savable
+type taskList struct {
+ head *Task
+ tail *Task
+}
+
+// Reset resets list l to the empty state.
+func (l *taskList) Reset() {
+ l.head = nil
+ l.tail = nil
+}
+
+// Empty returns true iff the list is empty.
+func (l *taskList) Empty() bool {
+ return l.head == nil
+}
+
+// Front returns the first element of list l or nil.
+func (l *taskList) Front() *Task {
+ return l.head
+}
+
+// Back returns the last element of list l or nil.
+func (l *taskList) Back() *Task {
+ return l.tail
+}
+
+// Len returns the number of elements in the list.
+//
+// NOTE: This is an O(n) operation.
+func (l *taskList) Len() (count int) {
+ for e := l.Front(); e != nil; e = (taskElementMapper{}.linkerFor(e)).Next() {
+ count++
+ }
+ return count
+}
+
+// PushFront inserts the element e at the front of list l.
+func (l *taskList) PushFront(e *Task) {
+ linker := taskElementMapper{}.linkerFor(e)
+ linker.SetNext(l.head)
+ linker.SetPrev(nil)
+ if l.head != nil {
+ taskElementMapper{}.linkerFor(l.head).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+
+ l.head = e
+}
+
+// PushBack inserts the element e at the back of list l.
+func (l *taskList) PushBack(e *Task) {
+ linker := taskElementMapper{}.linkerFor(e)
+ linker.SetNext(nil)
+ linker.SetPrev(l.tail)
+ if l.tail != nil {
+ taskElementMapper{}.linkerFor(l.tail).SetNext(e)
+ } else {
+ l.head = e
+ }
+
+ l.tail = e
+}
+
+// PushBackList inserts list m at the end of list l, emptying m.
+func (l *taskList) PushBackList(m *taskList) {
+ if l.head == nil {
+ l.head = m.head
+ l.tail = m.tail
+ } else if m.head != nil {
+ taskElementMapper{}.linkerFor(l.tail).SetNext(m.head)
+ taskElementMapper{}.linkerFor(m.head).SetPrev(l.tail)
+
+ l.tail = m.tail
+ }
+ m.head = nil
+ m.tail = nil
+}
+
+// InsertAfter inserts e after b.
+func (l *taskList) InsertAfter(b, e *Task) {
+ bLinker := taskElementMapper{}.linkerFor(b)
+ eLinker := taskElementMapper{}.linkerFor(e)
+
+ a := bLinker.Next()
+
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ bLinker.SetNext(e)
+
+ if a != nil {
+ taskElementMapper{}.linkerFor(a).SetPrev(e)
+ } else {
+ l.tail = e
+ }
+}
+
+// InsertBefore inserts e before a.
+func (l *taskList) InsertBefore(a, e *Task) {
+ aLinker := taskElementMapper{}.linkerFor(a)
+ eLinker := taskElementMapper{}.linkerFor(e)
+
+ b := aLinker.Prev()
+ eLinker.SetNext(a)
+ eLinker.SetPrev(b)
+ aLinker.SetPrev(e)
+
+ if b != nil {
+ taskElementMapper{}.linkerFor(b).SetNext(e)
+ } else {
+ l.head = e
+ }
+}
+
+// Remove removes e from l.
+func (l *taskList) Remove(e *Task) {
+ linker := taskElementMapper{}.linkerFor(e)
+ prev := linker.Prev()
+ next := linker.Next()
+
+ if prev != nil {
+ taskElementMapper{}.linkerFor(prev).SetNext(next)
+ } else if l.head == e {
+ l.head = next
+ }
+
+ if next != nil {
+ taskElementMapper{}.linkerFor(next).SetPrev(prev)
+ } else if l.tail == e {
+ l.tail = prev
+ }
+
+ linker.SetNext(nil)
+ linker.SetPrev(nil)
+}
+
+// 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 taskEntry struct {
+ next *Task
+ prev *Task
+}
+
+// Next returns the entry that follows e in the list.
+func (e *taskEntry) Next() *Task {
+ return e.next
+}
+
+// Prev returns the entry that precedes e in the list.
+func (e *taskEntry) Prev() *Task {
+ return e.prev
+}
+
+// SetNext assigns 'entry' as the entry that follows e in the list.
+func (e *taskEntry) SetNext(elem *Task) {
+ e.next = elem
+}
+
+// SetPrev assigns 'entry' as the entry that precedes e in the list.
+func (e *taskEntry) SetPrev(elem *Task) {
+ e.prev = elem
+}
diff --git a/pkg/sentry/kernel/task_test.go b/pkg/sentry/kernel/task_test.go
deleted file mode 100644
index cfcde9a7a..000000000
--- a/pkg/sentry/kernel/task_test.go
+++ /dev/null
@@ -1,69 +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 kernel
-
-import (
- "testing"
-
- "gvisor.dev/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/time/BUILD b/pkg/sentry/kernel/time/BUILD
deleted file mode 100644
index 2817aa3ba..000000000
--- a/pkg/sentry/kernel/time/BUILD
+++ /dev/null
@@ -1,20 +0,0 @@
-load("//tools:defs.bzl", "go_library")
-
-package(licenses = ["notice"])
-
-go_library(
- name = "time",
- srcs = [
- "context.go",
- "tcpip.go",
- "time.go",
- ],
- visibility = ["//pkg/sentry:internal"],
- deps = [
- "//pkg/abi/linux",
- "//pkg/context",
- "//pkg/sync",
- "//pkg/syserror",
- "//pkg/waiter",
- ],
-)
diff --git a/pkg/sentry/kernel/time/time_state_autogen.go b/pkg/sentry/kernel/time/time_state_autogen.go
new file mode 100644
index 000000000..4fb37df9a
--- /dev/null
+++ b/pkg/sentry/kernel/time/time_state_autogen.go
@@ -0,0 +1,97 @@
+// automatically generated by stateify.
+
+package time
+
+import (
+ "gvisor.dev/gvisor/pkg/state"
+)
+
+func (x *Time) StateTypeName() string {
+ return "pkg/sentry/kernel/time.Time"
+}
+
+func (x *Time) StateFields() []string {
+ return []string{
+ "ns",
+ }
+}
+
+func (x *Time) beforeSave() {}
+
+func (x *Time) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.ns)
+}
+
+func (x *Time) afterLoad() {}
+
+func (x *Time) StateLoad(m state.Source) {
+ m.Load(0, &x.ns)
+}
+
+func (x *Setting) StateTypeName() string {
+ return "pkg/sentry/kernel/time.Setting"
+}
+
+func (x *Setting) StateFields() []string {
+ return []string{
+ "Enabled",
+ "Next",
+ "Period",
+ }
+}
+
+func (x *Setting) beforeSave() {}
+
+func (x *Setting) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.Enabled)
+ m.Save(1, &x.Next)
+ m.Save(2, &x.Period)
+}
+
+func (x *Setting) afterLoad() {}
+
+func (x *Setting) StateLoad(m state.Source) {
+ m.Load(0, &x.Enabled)
+ m.Load(1, &x.Next)
+ m.Load(2, &x.Period)
+}
+
+func (x *Timer) StateTypeName() string {
+ return "pkg/sentry/kernel/time.Timer"
+}
+
+func (x *Timer) StateFields() []string {
+ return []string{
+ "clock",
+ "listener",
+ "setting",
+ "paused",
+ }
+}
+
+func (x *Timer) beforeSave() {}
+
+func (x *Timer) StateSave(m state.Sink) {
+ x.beforeSave()
+ m.Save(0, &x.clock)
+ m.Save(1, &x.listener)
+ m.Save(2, &x.setting)
+ m.Save(3, &x.paused)
+}
+
+func (x *Timer) afterLoad() {}
+
+func (x *Timer) StateLoad(m state.Source) {
+ m.Load(0, &x.clock)
+ m.Load(1, &x.listener)
+ m.Load(2, &x.setting)
+ m.Load(3, &x.paused)
+}
+
+func init() {
+ state.Register((*Time)(nil))
+ state.Register((*Setting)(nil))
+ state.Register((*Timer)(nil))
+}
diff --git a/pkg/sentry/kernel/timekeeper_test.go b/pkg/sentry/kernel/timekeeper_test.go
deleted file mode 100644
index cf2f7ca72..000000000
--- a/pkg/sentry/kernel/timekeeper_test.go
+++ /dev/null
@@ -1,156 +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 kernel
-
-import (
- "testing"
-
- "gvisor.dev/gvisor/pkg/sentry/contexttest"
- "gvisor.dev/gvisor/pkg/sentry/pgalloc"
- sentrytime "gvisor.dev/gvisor/pkg/sentry/time"
- "gvisor.dev/gvisor/pkg/sentry/usage"
- "gvisor.dev/gvisor/pkg/syserror"
- "gvisor.dev/gvisor/pkg/usermem"
-)
-
-// 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)
- 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(mfp, 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 = make(chan struct{})
- 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 = make(chan struct{})
- 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/uncaught_signal.proto b/pkg/sentry/kernel/uncaught_signal.proto
deleted file mode 100644
index 0bdb062cb..000000000
--- a/pkg/sentry/kernel/uncaught_signal.proto
+++ /dev/null
@@ -1,37 +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.
-
-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;
-}
diff --git a/pkg/sentry/kernel/uncaught_signal_go_proto/uncaught_signal.pb.go b/pkg/sentry/kernel/uncaught_signal_go_proto/uncaught_signal.pb.go
new file mode 100644
index 000000000..822e549ab
--- /dev/null
+++ b/pkg/sentry/kernel/uncaught_signal_go_proto/uncaught_signal.pb.go
@@ -0,0 +1,119 @@
+// Code generated by protoc-gen-go. DO NOT EDIT.
+// source: pkg/sentry/kernel/uncaught_signal.proto
+
+package gvisor
+
+import (
+ fmt "fmt"
+ proto "github.com/golang/protobuf/proto"
+ registers_go_proto "gvisor.dev/gvisor/pkg/sentry/arch/registers_go_proto"
+ math "math"
+)
+
+// Reference imports to suppress errors if they are not otherwise used.
+var _ = proto.Marshal
+var _ = fmt.Errorf
+var _ = math.Inf
+
+// This is a compile-time assertion to ensure that this generated file
+// is compatible with the proto package it is being compiled against.
+// A compilation error at this line likely means your copy of the
+// proto package needs to be updated.
+const _ = proto.ProtoPackageIsVersion3 // please upgrade the proto package
+
+type UncaughtSignal struct {
+ Tid int32 `protobuf:"varint,1,opt,name=tid,proto3" json:"tid,omitempty"`
+ Pid int32 `protobuf:"varint,2,opt,name=pid,proto3" json:"pid,omitempty"`
+ Registers *registers_go_proto.Registers `protobuf:"bytes,3,opt,name=registers,proto3" json:"registers,omitempty"`
+ SignalNumber int32 `protobuf:"varint,4,opt,name=signal_number,json=signalNumber,proto3" json:"signal_number,omitempty"`
+ FaultAddr uint64 `protobuf:"varint,5,opt,name=fault_addr,json=faultAddr,proto3" json:"fault_addr,omitempty"`
+ XXX_NoUnkeyedLiteral struct{} `json:"-"`
+ XXX_unrecognized []byte `json:"-"`
+ XXX_sizecache int32 `json:"-"`
+}
+
+func (m *UncaughtSignal) Reset() { *m = UncaughtSignal{} }
+func (m *UncaughtSignal) String() string { return proto.CompactTextString(m) }
+func (*UncaughtSignal) ProtoMessage() {}
+func (*UncaughtSignal) Descriptor() ([]byte, []int) {
+ return fileDescriptor_5ca9e03e13704688, []int{0}
+}
+
+func (m *UncaughtSignal) XXX_Unmarshal(b []byte) error {
+ return xxx_messageInfo_UncaughtSignal.Unmarshal(m, b)
+}
+func (m *UncaughtSignal) XXX_Marshal(b []byte, deterministic bool) ([]byte, error) {
+ return xxx_messageInfo_UncaughtSignal.Marshal(b, m, deterministic)
+}
+func (m *UncaughtSignal) XXX_Merge(src proto.Message) {
+ xxx_messageInfo_UncaughtSignal.Merge(m, src)
+}
+func (m *UncaughtSignal) XXX_Size() int {
+ return xxx_messageInfo_UncaughtSignal.Size(m)
+}
+func (m *UncaughtSignal) XXX_DiscardUnknown() {
+ xxx_messageInfo_UncaughtSignal.DiscardUnknown(m)
+}
+
+var xxx_messageInfo_UncaughtSignal proto.InternalMessageInfo
+
+func (m *UncaughtSignal) GetTid() int32 {
+ if m != nil {
+ return m.Tid
+ }
+ return 0
+}
+
+func (m *UncaughtSignal) GetPid() int32 {
+ if m != nil {
+ return m.Pid
+ }
+ return 0
+}
+
+func (m *UncaughtSignal) GetRegisters() *registers_go_proto.Registers {
+ if m != nil {
+ return m.Registers
+ }
+ return nil
+}
+
+func (m *UncaughtSignal) GetSignalNumber() int32 {
+ if m != nil {
+ return m.SignalNumber
+ }
+ return 0
+}
+
+func (m *UncaughtSignal) GetFaultAddr() uint64 {
+ if m != nil {
+ return m.FaultAddr
+ }
+ return 0
+}
+
+func init() {
+ proto.RegisterType((*UncaughtSignal)(nil), "gvisor.UncaughtSignal")
+}
+
+func init() {
+ proto.RegisterFile("pkg/sentry/kernel/uncaught_signal.proto", fileDescriptor_5ca9e03e13704688)
+}
+
+var fileDescriptor_5ca9e03e13704688 = []byte{
+ // 210 bytes of a gzipped FileDescriptorProto
+ 0x1f, 0x8b, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0xff, 0x4c, 0x8e, 0x4d, 0x4a, 0xc6, 0x30,
+ 0x10, 0x86, 0x89, 0xfd, 0x81, 0xc6, 0x1f, 0x34, 0xab, 0x20, 0x88, 0x45, 0x17, 0x76, 0xd5, 0x80,
+ 0x9e, 0xc0, 0x0b, 0xb8, 0x88, 0xb8, 0x2e, 0x69, 0x13, 0xd3, 0xd0, 0x9a, 0x86, 0x49, 0x22, 0x78,
+ 0x24, 0x6f, 0x29, 0x4d, 0xd4, 0xef, 0xdb, 0x0d, 0xcf, 0xbc, 0xf3, 0xcc, 0x8b, 0x1f, 0xdc, 0xa2,
+ 0x99, 0x57, 0x36, 0xc0, 0x17, 0x5b, 0x14, 0x58, 0xb5, 0xb2, 0x68, 0x27, 0x11, 0xf5, 0x1c, 0x06,
+ 0x6f, 0xb4, 0x15, 0x6b, 0xef, 0x60, 0x0b, 0x1b, 0xa9, 0xf5, 0xa7, 0xf1, 0x1b, 0x5c, 0xdf, 0x1e,
+ 0x1d, 0x08, 0x98, 0x66, 0x06, 0x4a, 0x1b, 0x1f, 0x14, 0xf8, 0x1c, 0xbc, 0xfb, 0x46, 0xf8, 0xe2,
+ 0xed, 0x57, 0xf1, 0x9a, 0x0c, 0xe4, 0x12, 0x17, 0xc1, 0x48, 0x8a, 0x5a, 0xd4, 0x55, 0x7c, 0x1f,
+ 0x77, 0xe2, 0x8c, 0xa4, 0x27, 0x99, 0x38, 0x23, 0x09, 0xc3, 0xcd, 0xbf, 0x89, 0x16, 0x2d, 0xea,
+ 0x4e, 0x1f, 0xaf, 0xfa, 0xfc, 0xb3, 0xe7, 0x7f, 0x0b, 0x7e, 0xc8, 0x90, 0x7b, 0x7c, 0x9e, 0x0b,
+ 0x0e, 0x36, 0x7e, 0x8c, 0x0a, 0x68, 0x99, 0x64, 0x67, 0x19, 0xbe, 0x24, 0x46, 0x6e, 0x30, 0x7e,
+ 0x17, 0x71, 0x0d, 0x83, 0x90, 0x12, 0x68, 0xd5, 0xa2, 0xae, 0xe4, 0x4d, 0x22, 0xcf, 0x52, 0xc2,
+ 0x58, 0xa7, 0xca, 0x4f, 0x3f, 0x01, 0x00, 0x00, 0xff, 0xff, 0xfd, 0x62, 0x54, 0xdf, 0x06, 0x01,
+ 0x00, 0x00,
+}