summaryrefslogtreecommitdiffhomepage
path: root/pkg/sentry
diff options
context:
space:
mode:
Diffstat (limited to 'pkg/sentry')
-rw-r--r--pkg/sentry/arch/arch_aarch64.go10
-rw-r--r--pkg/sentry/control/BUILD4
-rw-r--r--pkg/sentry/control/proc.go84
-rw-r--r--pkg/sentry/devices/memdev/full.go1
-rw-r--r--pkg/sentry/devices/memdev/null.go1
-rw-r--r--pkg/sentry/devices/memdev/random.go1
-rw-r--r--pkg/sentry/devices/memdev/zero.go1
-rw-r--r--pkg/sentry/fs/file.go2
-rw-r--r--pkg/sentry/fs/gofer/inode.go2
-rw-r--r--pkg/sentry/fs/host/inode.go3
-rw-r--r--pkg/sentry/fs/lock/lock.go41
-rw-r--r--pkg/sentry/fs/lock/lock_set_functions.go8
-rw-r--r--pkg/sentry/fs/lock/lock_test.go111
-rw-r--r--pkg/sentry/fs/user/BUILD1
-rw-r--r--pkg/sentry/fs/user/path.go57
-rw-r--r--pkg/sentry/fsimpl/devpts/BUILD1
-rw-r--r--pkg/sentry/fsimpl/devpts/devpts.go5
-rw-r--r--pkg/sentry/fsimpl/devpts/master.go5
-rw-r--r--pkg/sentry/fsimpl/devpts/slave.go5
-rw-r--r--pkg/sentry/fsimpl/eventfd/eventfd.go1
-rw-r--r--pkg/sentry/fsimpl/ext/BUILD1
-rw-r--r--pkg/sentry/fsimpl/ext/file_description.go1
-rw-r--r--pkg/sentry/fsimpl/ext/inode.go6
-rw-r--r--pkg/sentry/fsimpl/ext/regular_file.go1
-rw-r--r--pkg/sentry/fsimpl/gofer/BUILD2
-rw-r--r--pkg/sentry/fsimpl/gofer/filesystem.go9
-rw-r--r--pkg/sentry/fsimpl/gofer/gofer.go23
-rw-r--r--pkg/sentry/fsimpl/gofer/special_file.go4
-rw-r--r--pkg/sentry/fsimpl/host/BUILD1
-rw-r--r--pkg/sentry/fsimpl/host/host.go8
-rw-r--r--pkg/sentry/fsimpl/kernfs/BUILD2
-rw-r--r--pkg/sentry/fsimpl/kernfs/dynamic_bytes_file.go10
-rw-r--r--pkg/sentry/fsimpl/kernfs/fd_impl_util.go9
-rw-r--r--pkg/sentry/fsimpl/kernfs/inode_impl_util.go5
-rw-r--r--pkg/sentry/fsimpl/kernfs/kernfs_test.go13
-rw-r--r--pkg/sentry/fsimpl/pipefs/BUILD1
-rw-r--r--pkg/sentry/fsimpl/pipefs/pipefs.go6
-rw-r--r--pkg/sentry/fsimpl/proc/BUILD1
-rw-r--r--pkg/sentry/fsimpl/proc/subtasks.go5
-rw-r--r--pkg/sentry/fsimpl/proc/task.go5
-rw-r--r--pkg/sentry/fsimpl/proc/task_fds.go7
-rw-r--r--pkg/sentry/fsimpl/proc/task_files.go10
-rw-r--r--pkg/sentry/fsimpl/proc/tasks.go5
-rw-r--r--pkg/sentry/fsimpl/signalfd/signalfd.go1
-rw-r--r--pkg/sentry/fsimpl/sys/BUILD1
-rw-r--r--pkg/sentry/fsimpl/sys/sys.go7
-rw-r--r--pkg/sentry/fsimpl/timerfd/timerfd.go1
-rw-r--r--pkg/sentry/fsimpl/tmpfs/filesystem.go6
-rw-r--r--pkg/sentry/fsimpl/tmpfs/regular_file.go23
-rw-r--r--pkg/sentry/fsimpl/tmpfs/tmpfs.go41
-rw-r--r--pkg/sentry/kernel/fd_table.go15
-rw-r--r--pkg/sentry/kernel/kernel.go5
-rw-r--r--pkg/sentry/kernel/pipe/BUILD1
-rw-r--r--pkg/sentry/kernel/pipe/vfs.go13
-rw-r--r--pkg/sentry/loader/vdso.go2
-rw-r--r--pkg/sentry/pgalloc/pgalloc.go11
-rw-r--r--pkg/sentry/platform/kvm/machine_arm64.go2
-rw-r--r--pkg/sentry/socket/hostinet/BUILD1
-rw-r--r--pkg/sentry/socket/hostinet/socket_vfs2.go3
-rw-r--r--pkg/sentry/socket/netlink/BUILD1
-rw-r--r--pkg/sentry/socket/netlink/socket_vfs2.go8
-rw-r--r--pkg/sentry/socket/netstack/BUILD1
-rw-r--r--pkg/sentry/socket/netstack/netstack.go35
-rw-r--r--pkg/sentry/socket/netstack/netstack_vfs2.go3
-rw-r--r--pkg/sentry/socket/unix/BUILD1
-rw-r--r--pkg/sentry/socket/unix/unix.go26
-rw-r--r--pkg/sentry/socket/unix/unix_vfs2.go7
-rw-r--r--pkg/sentry/syscalls/linux/sys_file.go39
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/BUILD2
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/fd.go2
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/lock.go64
-rw-r--r--pkg/sentry/syscalls/linux/vfs2/vfs2.go2
-rw-r--r--pkg/sentry/time/muldiv_arm64.s3
-rw-r--r--pkg/sentry/time/parameters_test.go15
-rw-r--r--pkg/sentry/vfs/BUILD1
-rw-r--r--pkg/sentry/vfs/epoll.go1
-rw-r--r--pkg/sentry/vfs/file_description.go25
-rw-r--r--pkg/sentry/vfs/file_description_impl_util.go80
-rw-r--r--pkg/sentry/vfs/file_description_impl_util_test.go1
-rw-r--r--pkg/sentry/vfs/inotify.go1
80 files changed, 540 insertions, 390 deletions
diff --git a/pkg/sentry/arch/arch_aarch64.go b/pkg/sentry/arch/arch_aarch64.go
index 343f81f59..daba8b172 100644
--- a/pkg/sentry/arch/arch_aarch64.go
+++ b/pkg/sentry/arch/arch_aarch64.go
@@ -17,7 +17,6 @@
package arch
import (
- "encoding/binary"
"fmt"
"io"
@@ -49,9 +48,14 @@ const ARMTrapFlag = uint64(1) << 21
type aarch64FPState []byte
// initAarch64FPState sets up initial state.
+//
+// Related code in Linux kernel: fpsimd_flush_thread().
+// FPCR = FPCR_RM_RN (0x0 << 22).
+//
+// Currently, aarch64FPState is only a space of 0x210 length for fpstate.
+// The fp head is useless in sentry/ptrace/kvm.
+//
func initAarch64FPState(data aarch64FPState) {
- binary.LittleEndian.PutUint32(data, fpsimdMagic)
- binary.LittleEndian.PutUint32(data[4:], fpsimdContextSize)
}
func newAarch64FPStateSlice() []byte {
diff --git a/pkg/sentry/control/BUILD b/pkg/sentry/control/BUILD
index 0c9a62f0d..2c5d14be5 100644
--- a/pkg/sentry/control/BUILD
+++ b/pkg/sentry/control/BUILD
@@ -16,15 +16,12 @@ go_library(
],
deps = [
"//pkg/abi/linux",
- "//pkg/context",
"//pkg/fd",
- "//pkg/fspath",
"//pkg/log",
"//pkg/sentry/fdimport",
"//pkg/sentry/fs",
"//pkg/sentry/fs/host",
"//pkg/sentry/fs/user",
- "//pkg/sentry/fsbridge",
"//pkg/sentry/fsimpl/host",
"//pkg/sentry/kernel",
"//pkg/sentry/kernel/auth",
@@ -36,7 +33,6 @@ go_library(
"//pkg/sentry/vfs",
"//pkg/sentry/watchdog",
"//pkg/sync",
- "//pkg/syserror",
"//pkg/tcpip/link/sniffer",
"//pkg/urpc",
"@org_golang_x_sys//unix:go_default_library",
diff --git a/pkg/sentry/control/proc.go b/pkg/sentry/control/proc.go
index 8767430b7..1bae7cfaf 100644
--- a/pkg/sentry/control/proc.go
+++ b/pkg/sentry/control/proc.go
@@ -25,13 +25,10 @@ import (
"golang.org/x/sys/unix"
"gvisor.dev/gvisor/pkg/abi/linux"
- "gvisor.dev/gvisor/pkg/context"
- "gvisor.dev/gvisor/pkg/fspath"
"gvisor.dev/gvisor/pkg/sentry/fdimport"
"gvisor.dev/gvisor/pkg/sentry/fs"
"gvisor.dev/gvisor/pkg/sentry/fs/host"
"gvisor.dev/gvisor/pkg/sentry/fs/user"
- "gvisor.dev/gvisor/pkg/sentry/fsbridge"
hostvfs2 "gvisor.dev/gvisor/pkg/sentry/fsimpl/host"
"gvisor.dev/gvisor/pkg/sentry/kernel"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
@@ -39,7 +36,6 @@ import (
"gvisor.dev/gvisor/pkg/sentry/limits"
"gvisor.dev/gvisor/pkg/sentry/usage"
"gvisor.dev/gvisor/pkg/sentry/vfs"
- "gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/urpc"
)
@@ -107,6 +103,9 @@ type ExecArgs struct {
// String prints the arguments as a string.
func (args ExecArgs) String() string {
+ if len(args.Argv) == 0 {
+ return args.Filename
+ }
a := make([]string, len(args.Argv))
copy(a, args.Argv)
if args.Filename != "" {
@@ -179,37 +178,30 @@ func (proc *Proc) execAsync(args *ExecArgs) (*kernel.ThreadGroup, kernel.ThreadI
}
ctx := initArgs.NewContext(proc.Kernel)
- if initArgs.Filename == "" {
- if kernel.VFS2Enabled {
- // Get the full path to the filename from the PATH env variable.
- if initArgs.MountNamespaceVFS2 == nil {
- // Set initArgs so that 'ctx' returns the namespace.
- //
- // MountNamespaceVFS2 adds a reference to the namespace, which is
- // transferred to the new process.
- initArgs.MountNamespaceVFS2 = proc.Kernel.GlobalInit().Leader().MountNamespaceVFS2()
- }
- file, err := getExecutableFD(ctx, creds, proc.Kernel.VFS(), initArgs.MountNamespaceVFS2, initArgs.Envv, initArgs.WorkingDirectory, initArgs.Argv[0])
- if err != nil {
- return nil, 0, nil, nil, fmt.Errorf("error finding executable %q in environment %v: %v", initArgs.Argv[0], initArgs.Envv, err)
- }
- initArgs.File = fsbridge.NewVFSFile(file)
- } else {
- if initArgs.MountNamespace == nil {
- // Set initArgs so that 'ctx' returns the namespace.
- initArgs.MountNamespace = proc.Kernel.GlobalInit().Leader().MountNamespace()
+ if kernel.VFS2Enabled {
+ // Get the full path to the filename from the PATH env variable.
+ if initArgs.MountNamespaceVFS2 == nil {
+ // Set initArgs so that 'ctx' returns the namespace.
+ //
+ // MountNamespaceVFS2 adds a reference to the namespace, which is
+ // transferred to the new process.
+ initArgs.MountNamespaceVFS2 = proc.Kernel.GlobalInit().Leader().MountNamespaceVFS2()
+ }
+ } else {
+ if initArgs.MountNamespace == nil {
+ // Set initArgs so that 'ctx' returns the namespace.
+ initArgs.MountNamespace = proc.Kernel.GlobalInit().Leader().MountNamespace()
- // initArgs must hold a reference on MountNamespace, which will
- // be donated to the new process in CreateProcess.
- initArgs.MountNamespace.IncRef()
- }
- f, err := user.ResolveExecutablePath(ctx, creds, initArgs.MountNamespace, initArgs.Envv, initArgs.WorkingDirectory, initArgs.Argv[0])
- if err != nil {
- return nil, 0, nil, nil, fmt.Errorf("error finding executable %q in PATH %v: %v", initArgs.Argv[0], initArgs.Envv, err)
- }
- initArgs.Filename = f
+ // initArgs must hold a reference on MountNamespace, which will
+ // be donated to the new process in CreateProcess.
+ initArgs.MountNamespace.IncRef()
}
}
+ resolved, err := user.ResolveExecutablePath(ctx, &initArgs)
+ if err != nil {
+ return nil, 0, nil, nil, err
+ }
+ initArgs.Filename = resolved
fds := make([]int, len(args.FilePayload.Files))
for i, file := range args.FilePayload.Files {
@@ -422,31 +414,3 @@ func ttyName(tty *kernel.TTY) string {
}
return fmt.Sprintf("pts/%d", tty.Index)
}
-
-// getExecutableFD resolves the given executable name and returns a
-// vfs.FileDescription for the executable file.
-func getExecutableFD(ctx context.Context, creds *auth.Credentials, vfsObj *vfs.VirtualFilesystem, mns *vfs.MountNamespace, envv []string, wd, name string) (*vfs.FileDescription, error) {
- path, err := user.ResolveExecutablePathVFS2(ctx, creds, mns, envv, wd, name)
- if err != nil {
- return nil, err
- }
-
- root := vfs.RootFromContext(ctx)
- defer root.DecRef()
-
- pop := vfs.PathOperation{
- Root: root,
- Start: root, // binPath is absolute, Start can be anything.
- Path: fspath.Parse(path),
- FollowFinalSymlink: true,
- }
- opts := &vfs.OpenOptions{
- Flags: linux.O_RDONLY,
- FileExec: true,
- }
- f, err := vfsObj.OpenAt(ctx, creds, &pop, opts)
- if err == syserror.ENOENT || err == syserror.EACCES {
- return nil, nil
- }
- return f, err
-}
diff --git a/pkg/sentry/devices/memdev/full.go b/pkg/sentry/devices/memdev/full.go
index c7e197691..af66fe4dc 100644
--- a/pkg/sentry/devices/memdev/full.go
+++ b/pkg/sentry/devices/memdev/full.go
@@ -42,6 +42,7 @@ type fullFD struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
vfs.DentryMetadataFileDescriptionImpl
+ vfs.NoLockFD
}
// Release implements vfs.FileDescriptionImpl.Release.
diff --git a/pkg/sentry/devices/memdev/null.go b/pkg/sentry/devices/memdev/null.go
index 33d060d02..92d3d71be 100644
--- a/pkg/sentry/devices/memdev/null.go
+++ b/pkg/sentry/devices/memdev/null.go
@@ -43,6 +43,7 @@ type nullFD struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
vfs.DentryMetadataFileDescriptionImpl
+ vfs.NoLockFD
}
// Release implements vfs.FileDescriptionImpl.Release.
diff --git a/pkg/sentry/devices/memdev/random.go b/pkg/sentry/devices/memdev/random.go
index acfa23149..6b81da5ef 100644
--- a/pkg/sentry/devices/memdev/random.go
+++ b/pkg/sentry/devices/memdev/random.go
@@ -48,6 +48,7 @@ type randomFD struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
vfs.DentryMetadataFileDescriptionImpl
+ vfs.NoLockFD
// off is the "file offset". off is accessed using atomic memory
// operations.
diff --git a/pkg/sentry/devices/memdev/zero.go b/pkg/sentry/devices/memdev/zero.go
index 3b1372b9e..c6f15054d 100644
--- a/pkg/sentry/devices/memdev/zero.go
+++ b/pkg/sentry/devices/memdev/zero.go
@@ -44,6 +44,7 @@ type zeroFD struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
vfs.DentryMetadataFileDescriptionImpl
+ vfs.NoLockFD
}
// Release implements vfs.FileDescriptionImpl.Release.
diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go
index 2a278fbe3..ca41520b4 100644
--- a/pkg/sentry/fs/file.go
+++ b/pkg/sentry/fs/file.go
@@ -146,7 +146,7 @@ func (f *File) DecRef() {
f.DecRefWithDestructor(func() {
// Drop BSD style locks.
lockRng := lock.LockRange{Start: 0, End: lock.LockEOF}
- f.Dirent.Inode.LockCtx.BSD.UnlockRegion(lock.UniqueID(f.UniqueID), lockRng)
+ f.Dirent.Inode.LockCtx.BSD.UnlockRegion(f, lockRng)
// Release resources held by the FileOperations.
f.FileOperations.Release()
diff --git a/pkg/sentry/fs/gofer/inode.go b/pkg/sentry/fs/gofer/inode.go
index a016c896e..51d7368a1 100644
--- a/pkg/sentry/fs/gofer/inode.go
+++ b/pkg/sentry/fs/gofer/inode.go
@@ -640,7 +640,7 @@ func (i *inodeOperations) Allocate(ctx context.Context, inode *fs.Inode, offset,
// WriteOut implements fs.InodeOperations.WriteOut.
func (i *inodeOperations) WriteOut(ctx context.Context, inode *fs.Inode) error {
- if !i.session().cachePolicy.cacheUAttrs(inode) {
+ if inode.MountSource.Flags.ReadOnly || !i.session().cachePolicy.cacheUAttrs(inode) {
return nil
}
diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go
index 62f1246aa..fbfba1b58 100644
--- a/pkg/sentry/fs/host/inode.go
+++ b/pkg/sentry/fs/host/inode.go
@@ -368,6 +368,9 @@ func (i *inodeOperations) Allocate(ctx context.Context, inode *fs.Inode, offset,
// WriteOut implements fs.InodeOperations.WriteOut.
func (i *inodeOperations) WriteOut(ctx context.Context, inode *fs.Inode) error {
+ if inode.MountSource.Flags.ReadOnly {
+ return nil
+ }
// Have we been using host kernel metadata caches?
if !inode.MountSource.Flags.ForcePageCache || !canMap(inode) {
// Then the metadata is already up to date on the host.
diff --git a/pkg/sentry/fs/lock/lock.go b/pkg/sentry/fs/lock/lock.go
index 926538d90..8a5d9c7eb 100644
--- a/pkg/sentry/fs/lock/lock.go
+++ b/pkg/sentry/fs/lock/lock.go
@@ -62,7 +62,7 @@ import (
type LockType int
// UniqueID is a unique identifier of the holder of a regional file lock.
-type UniqueID uint64
+type UniqueID interface{}
const (
// ReadLock describes a POSIX regional file lock to be taken
@@ -98,12 +98,7 @@ type Lock struct {
// If len(Readers) > 0 then HasWriter must be false.
Readers map[UniqueID]bool
- // HasWriter indicates that this is a write lock held by a single
- // UniqueID.
- HasWriter bool
-
- // Writer is only valid if HasWriter is true. It identifies a
- // single write lock holder.
+ // Writer holds the writer unique ID. It's nil if there are no writers.
Writer UniqueID
}
@@ -186,7 +181,6 @@ func makeLock(uid UniqueID, t LockType) Lock {
case ReadLock:
value.Readers[uid] = true
case WriteLock:
- value.HasWriter = true
value.Writer = uid
default:
panic(fmt.Sprintf("makeLock: invalid lock type %d", t))
@@ -196,10 +190,7 @@ func makeLock(uid UniqueID, t LockType) Lock {
// isHeld returns true if uid is a holder of Lock.
func (l Lock) isHeld(uid UniqueID) bool {
- if l.HasWriter && l.Writer == uid {
- return true
- }
- return l.Readers[uid]
+ return l.Writer == uid || l.Readers[uid]
}
// lock sets uid as a holder of a typed lock on Lock.
@@ -214,20 +205,20 @@ func (l *Lock) lock(uid UniqueID, t LockType) {
}
// We cannot downgrade a write lock to a read lock unless the
// uid is the same.
- if l.HasWriter {
+ if l.Writer != nil {
if l.Writer != uid {
panic(fmt.Sprintf("lock: cannot downgrade write lock to read lock for uid %d, writer is %d", uid, l.Writer))
}
// Ensure that there is only one reader if upgrading.
l.Readers = make(map[UniqueID]bool)
// Ensure that there is no longer a writer.
- l.HasWriter = false
+ l.Writer = nil
}
l.Readers[uid] = true
return
case WriteLock:
// If we are already the writer, then this is a no-op.
- if l.HasWriter && l.Writer == uid {
+ if l.Writer == uid {
return
}
// We can only upgrade a read lock to a write lock if there
@@ -243,7 +234,6 @@ func (l *Lock) lock(uid UniqueID, t LockType) {
}
// Ensure that there is only a writer.
l.Readers = make(map[UniqueID]bool)
- l.HasWriter = true
l.Writer = uid
default:
panic(fmt.Sprintf("lock: invalid lock type %d", t))
@@ -277,9 +267,8 @@ func (l LockSet) canLock(uid UniqueID, t LockType, r LockRange) bool {
switch t {
case ReadLock:
return l.lockable(r, func(value Lock) bool {
- // If there is no writer, there's no problem adding
- // another reader.
- if !value.HasWriter {
+ // If there is no writer, there's no problem adding another reader.
+ if value.Writer == nil {
return true
}
// If there is a writer, then it must be the same uid
@@ -289,10 +278,9 @@ func (l LockSet) canLock(uid UniqueID, t LockType, r LockRange) bool {
case WriteLock:
return l.lockable(r, func(value Lock) bool {
// If there are only readers.
- if !value.HasWriter {
- // Then this uid can only take a write lock if
- // this is a private upgrade, meaning that the
- // only reader is uid.
+ if value.Writer == nil {
+ // Then this uid can only take a write lock if this is a private
+ // upgrade, meaning that the only reader is uid.
return len(value.Readers) == 1 && value.Readers[uid]
}
// If the uid is already a writer on this region, then
@@ -304,7 +292,8 @@ func (l LockSet) canLock(uid UniqueID, t LockType, r LockRange) bool {
}
}
-// lock returns true if uid took a lock of type t on the entire range of LockRange.
+// lock returns true if uid took a lock of type t on the entire range of
+// LockRange.
//
// Preconditions: r.Start <= r.End (will panic otherwise).
func (l *LockSet) lock(uid UniqueID, t LockType, r LockRange) bool {
@@ -339,7 +328,7 @@ func (l *LockSet) lock(uid UniqueID, t LockType, r LockRange) bool {
seg, _ = l.SplitUnchecked(seg, r.End)
}
- // Set the lock on the segment. This is guaranteed to
+ // Set the lock on the segment. This is guaranteed to
// always be safe, given canLock above.
value := seg.ValuePtr()
value.lock(uid, t)
@@ -386,7 +375,7 @@ func (l *LockSet) unlock(uid UniqueID, r LockRange) {
value := seg.Value()
var remove bool
- if value.HasWriter && value.Writer == uid {
+ if value.Writer == uid {
// If we are unlocking a writer, then since there can
// only ever be one writer and no readers, then this
// lock should always be removed from the set.
diff --git a/pkg/sentry/fs/lock/lock_set_functions.go b/pkg/sentry/fs/lock/lock_set_functions.go
index 8a3ace0c1..50a16e662 100644
--- a/pkg/sentry/fs/lock/lock_set_functions.go
+++ b/pkg/sentry/fs/lock/lock_set_functions.go
@@ -44,14 +44,9 @@ func (lockSetFunctions) Merge(r1 LockRange, val1 Lock, r2 LockRange, val2 Lock)
return Lock{}, false
}
}
- if val1.HasWriter != val2.HasWriter {
+ if val1.Writer != val2.Writer {
return Lock{}, false
}
- if val1.HasWriter {
- if val1.Writer != val2.Writer {
- return Lock{}, false
- }
- }
return val1, true
}
@@ -62,7 +57,6 @@ func (lockSetFunctions) Split(r LockRange, val Lock, split uint64) (Lock, Lock)
for k, v := range val.Readers {
val0.Readers[k] = v
}
- val0.HasWriter = val.HasWriter
val0.Writer = val.Writer
return val, val0
diff --git a/pkg/sentry/fs/lock/lock_test.go b/pkg/sentry/fs/lock/lock_test.go
index ba002aeb7..fad90984b 100644
--- a/pkg/sentry/fs/lock/lock_test.go
+++ b/pkg/sentry/fs/lock/lock_test.go
@@ -42,9 +42,6 @@ func equals(e0, e1 []entry) bool {
if !reflect.DeepEqual(e0[i].LockRange, e1[i].LockRange) {
return false
}
- if e0[i].Lock.HasWriter != e1[i].Lock.HasWriter {
- return false
- }
if e0[i].Lock.Writer != e1[i].Lock.Writer {
return false
}
@@ -105,7 +102,7 @@ func TestCanLock(t *testing.T) {
LockRange: LockRange{2048, 3072},
},
{
- Lock: Lock{HasWriter: true, Writer: 1},
+ Lock: Lock{Writer: 1},
LockRange: LockRange{3072, 4096},
},
})
@@ -241,7 +238,7 @@ func TestSetLock(t *testing.T) {
// 0 max uint64
after: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, LockEOF},
},
},
@@ -254,7 +251,7 @@ func TestSetLock(t *testing.T) {
// 0 max uint64
before: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, LockEOF},
},
},
@@ -273,7 +270,7 @@ func TestSetLock(t *testing.T) {
LockRange: LockRange{0, 4096},
},
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{4096, LockEOF},
},
},
@@ -301,7 +298,7 @@ func TestSetLock(t *testing.T) {
// 0 4096 max uint64
after: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, 4096},
},
{
@@ -318,7 +315,7 @@ func TestSetLock(t *testing.T) {
// 0 max uint64
before: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, LockEOF},
},
},
@@ -550,7 +547,7 @@ func TestSetLock(t *testing.T) {
LockRange: LockRange{0, 1024},
},
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{1024, 4096},
},
{
@@ -594,7 +591,7 @@ func TestSetLock(t *testing.T) {
LockRange: LockRange{0, 1024},
},
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{1024, 3072},
},
{
@@ -633,7 +630,7 @@ func TestSetLock(t *testing.T) {
// 0 1024 2048 4096 max uint64
before: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, 1024},
},
{
@@ -663,11 +660,11 @@ func TestSetLock(t *testing.T) {
// 0 1024 max uint64
after: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, 1024},
},
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{1024, LockEOF},
},
},
@@ -675,28 +672,30 @@ func TestSetLock(t *testing.T) {
}
for _, test := range tests {
- l := fill(test.before)
+ t.Run(test.name, func(t *testing.T) {
+ l := fill(test.before)
- r := LockRange{Start: test.start, End: test.end}
- success := l.lock(test.uid, test.lockType, r)
- var got []entry
- for seg := l.FirstSegment(); seg.Ok(); seg = seg.NextSegment() {
- got = append(got, entry{
- Lock: seg.Value(),
- LockRange: seg.Range(),
- })
- }
+ r := LockRange{Start: test.start, End: test.end}
+ success := l.lock(test.uid, test.lockType, r)
+ var got []entry
+ for seg := l.FirstSegment(); seg.Ok(); seg = seg.NextSegment() {
+ got = append(got, entry{
+ Lock: seg.Value(),
+ LockRange: seg.Range(),
+ })
+ }
- if success != test.success {
- t.Errorf("%s: setlock(%v, %+v, %d, %d) got success %v, want %v", test.name, test.before, r, test.uid, test.lockType, success, test.success)
- continue
- }
+ if success != test.success {
+ t.Errorf("setlock(%v, %+v, %d, %d) got success %v, want %v", test.before, r, test.uid, test.lockType, success, test.success)
+ return
+ }
- if success {
- if !equals(got, test.after) {
- t.Errorf("%s: got set %+v, want %+v", test.name, got, test.after)
+ if success {
+ if !equals(got, test.after) {
+ t.Errorf("got set %+v, want %+v", got, test.after)
+ }
}
- }
+ })
}
}
@@ -782,7 +781,7 @@ func TestUnlock(t *testing.T) {
// 0 max uint64
before: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, LockEOF},
},
},
@@ -824,7 +823,7 @@ func TestUnlock(t *testing.T) {
// 0 max uint64
before: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, LockEOF},
},
},
@@ -837,7 +836,7 @@ func TestUnlock(t *testing.T) {
// 0 4096 max uint64
after: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{4096, LockEOF},
},
},
@@ -876,7 +875,7 @@ func TestUnlock(t *testing.T) {
// 0 max uint64
before: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, LockEOF},
},
},
@@ -889,7 +888,7 @@ func TestUnlock(t *testing.T) {
// 0 4096
after: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, 4096},
},
},
@@ -906,7 +905,7 @@ func TestUnlock(t *testing.T) {
LockRange: LockRange{0, 1024},
},
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{1024, 4096},
},
{
@@ -974,7 +973,7 @@ func TestUnlock(t *testing.T) {
// 0 1024 4096 max uint64
before: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, 1024},
},
{
@@ -991,7 +990,7 @@ func TestUnlock(t *testing.T) {
// 0 8 4096 max uint64
after: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, 8},
},
{
@@ -1008,7 +1007,7 @@ func TestUnlock(t *testing.T) {
// 0 1024 4096 max uint64
before: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, 1024},
},
{
@@ -1025,7 +1024,7 @@ func TestUnlock(t *testing.T) {
// 0 1024 4096 8192 max uint64
after: []entry{
{
- Lock: Lock{HasWriter: true, Writer: 0},
+ Lock: Lock{Writer: 0},
LockRange: LockRange{0, 1024},
},
{
@@ -1041,19 +1040,21 @@ func TestUnlock(t *testing.T) {
}
for _, test := range tests {
- l := fill(test.before)
+ t.Run(test.name, func(t *testing.T) {
+ l := fill(test.before)
- r := LockRange{Start: test.start, End: test.end}
- l.unlock(test.uid, r)
- var got []entry
- for seg := l.FirstSegment(); seg.Ok(); seg = seg.NextSegment() {
- got = append(got, entry{
- Lock: seg.Value(),
- LockRange: seg.Range(),
- })
- }
- if !equals(got, test.after) {
- t.Errorf("%s: got set %+v, want %+v", test.name, got, test.after)
- }
+ r := LockRange{Start: test.start, End: test.end}
+ l.unlock(test.uid, r)
+ var got []entry
+ for seg := l.FirstSegment(); seg.Ok(); seg = seg.NextSegment() {
+ got = append(got, entry{
+ Lock: seg.Value(),
+ LockRange: seg.Range(),
+ })
+ }
+ if !equals(got, test.after) {
+ t.Errorf("got set %+v, want %+v", got, test.after)
+ }
+ })
}
}
diff --git a/pkg/sentry/fs/user/BUILD b/pkg/sentry/fs/user/BUILD
index bd5dac373..66e949c95 100644
--- a/pkg/sentry/fs/user/BUILD
+++ b/pkg/sentry/fs/user/BUILD
@@ -15,6 +15,7 @@ go_library(
"//pkg/fspath",
"//pkg/log",
"//pkg/sentry/fs",
+ "//pkg/sentry/kernel",
"//pkg/sentry/kernel/auth",
"//pkg/sentry/vfs",
"//pkg/syserror",
diff --git a/pkg/sentry/fs/user/path.go b/pkg/sentry/fs/user/path.go
index fbd4547a7..397e96045 100644
--- a/pkg/sentry/fs/user/path.go
+++ b/pkg/sentry/fs/user/path.go
@@ -24,6 +24,7 @@ import (
"gvisor.dev/gvisor/pkg/fspath"
"gvisor.dev/gvisor/pkg/log"
"gvisor.dev/gvisor/pkg/sentry/fs"
+ "gvisor.dev/gvisor/pkg/sentry/kernel"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
"gvisor.dev/gvisor/pkg/syserror"
@@ -31,7 +32,15 @@ import (
// ResolveExecutablePath resolves the given executable name given the working
// dir and environment.
-func ResolveExecutablePath(ctx context.Context, creds *auth.Credentials, mns *fs.MountNamespace, envv []string, wd, name string) (string, error) {
+func ResolveExecutablePath(ctx context.Context, args *kernel.CreateProcessArgs) (string, error) {
+ name := args.Filename
+ if len(name) == 0 {
+ if len(args.Argv) == 0 {
+ return "", fmt.Errorf("no filename or command provided")
+ }
+ name = args.Argv[0]
+ }
+
// Absolute paths can be used directly.
if path.IsAbs(name) {
return name, nil
@@ -40,6 +49,7 @@ func ResolveExecutablePath(ctx context.Context, creds *auth.Credentials, mns *fs
// Paths with '/' in them should be joined to the working directory, or
// to the root if working directory is not set.
if strings.IndexByte(name, '/') > 0 {
+ wd := args.WorkingDirectory
if wd == "" {
wd = "/"
}
@@ -49,10 +59,24 @@ func ResolveExecutablePath(ctx context.Context, creds *auth.Credentials, mns *fs
return path.Join(wd, name), nil
}
- // Otherwise, We must lookup the name in the paths, starting from the
- // calling context's root directory.
- paths := getPath(envv)
+ // Otherwise, We must lookup the name in the paths.
+ paths := getPath(args.Envv)
+ if kernel.VFS2Enabled {
+ f, err := resolveVFS2(ctx, args.Credentials, args.MountNamespaceVFS2, paths, name)
+ if err != nil {
+ return "", fmt.Errorf("error finding executable %q in PATH %v: %v", name, paths, err)
+ }
+ return f, nil
+ }
+
+ f, err := resolve(ctx, args.MountNamespace, paths, name)
+ if err != nil {
+ return "", fmt.Errorf("error finding executable %q in PATH %v: %v", name, paths, err)
+ }
+ return f, nil
+}
+func resolve(ctx context.Context, mns *fs.MountNamespace, paths []string, name string) (string, error) {
root := fs.RootFromContext(ctx)
if root == nil {
// Caller has no root. Don't bother traversing anything.
@@ -95,30 +119,7 @@ func ResolveExecutablePath(ctx context.Context, creds *auth.Credentials, mns *fs
return "", syserror.ENOENT
}
-// ResolveExecutablePathVFS2 resolves the given executable name given the
-// working dir and environment.
-func ResolveExecutablePathVFS2(ctx context.Context, creds *auth.Credentials, mns *vfs.MountNamespace, envv []string, wd, name string) (string, error) {
- // Absolute paths can be used directly.
- if path.IsAbs(name) {
- return name, nil
- }
-
- // Paths with '/' in them should be joined to the working directory, or
- // to the root if working directory is not set.
- if strings.IndexByte(name, '/') > 0 {
- if wd == "" {
- wd = "/"
- }
- if !path.IsAbs(wd) {
- return "", fmt.Errorf("working directory %q must be absolute", wd)
- }
- return path.Join(wd, name), nil
- }
-
- // Otherwise, We must lookup the name in the paths, starting from the
- // calling context's root directory.
- paths := getPath(envv)
-
+func resolveVFS2(ctx context.Context, creds *auth.Credentials, mns *vfs.MountNamespace, paths []string, name string) (string, error) {
root := mns.Root()
defer root.DecRef()
for _, p := range paths {
diff --git a/pkg/sentry/fsimpl/devpts/BUILD b/pkg/sentry/fsimpl/devpts/BUILD
index 585764223..cf440dce8 100644
--- a/pkg/sentry/fsimpl/devpts/BUILD
+++ b/pkg/sentry/fsimpl/devpts/BUILD
@@ -23,6 +23,7 @@ go_library(
"//pkg/sentry/kernel/auth",
"//pkg/sentry/unimpl",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/sync",
"//pkg/syserror",
"//pkg/usermem",
diff --git a/pkg/sentry/fsimpl/devpts/devpts.go b/pkg/sentry/fsimpl/devpts/devpts.go
index c03c65445..9b0e0cca2 100644
--- a/pkg/sentry/fsimpl/devpts/devpts.go
+++ b/pkg/sentry/fsimpl/devpts/devpts.go
@@ -28,6 +28,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/fsimpl/kernfs"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
)
@@ -116,6 +117,8 @@ type rootInode struct {
kernfs.InodeNotSymlink
kernfs.OrderedChildren
+ locks lock.FileLocks
+
// Keep a reference to this inode's dentry.
dentry kernfs.Dentry
@@ -183,7 +186,7 @@ func (i *rootInode) masterClose(t *Terminal) {
// Open implements kernfs.Inode.Open.
func (i *rootInode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
- fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &i.OrderedChildren, &opts)
+ fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &i.OrderedChildren, &i.locks, &opts)
if err != nil {
return nil, err
}
diff --git a/pkg/sentry/fsimpl/devpts/master.go b/pkg/sentry/fsimpl/devpts/master.go
index 7a7ce5d81..1d22adbe3 100644
--- a/pkg/sentry/fsimpl/devpts/master.go
+++ b/pkg/sentry/fsimpl/devpts/master.go
@@ -22,6 +22,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/unimpl"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
"gvisor.dev/gvisor/pkg/waiter"
@@ -34,6 +35,8 @@ type masterInode struct {
kernfs.InodeNotDirectory
kernfs.InodeNotSymlink
+ locks lock.FileLocks
+
// Keep a reference to this inode's dentry.
dentry kernfs.Dentry
@@ -55,6 +58,7 @@ func (mi *masterInode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vf
inode: mi,
t: t,
}
+ fd.LockFD.Init(&mi.locks)
if err := fd.vfsfd.Init(fd, opts.Flags, rp.Mount(), vfsd, &vfs.FileDescriptionOptions{}); err != nil {
mi.DecRef()
return nil, err
@@ -85,6 +89,7 @@ func (mi *masterInode) SetStat(ctx context.Context, vfsfs *vfs.Filesystem, creds
type masterFileDescription struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
+ vfs.LockFD
inode *masterInode
t *Terminal
diff --git a/pkg/sentry/fsimpl/devpts/slave.go b/pkg/sentry/fsimpl/devpts/slave.go
index 526cd406c..7fe475080 100644
--- a/pkg/sentry/fsimpl/devpts/slave.go
+++ b/pkg/sentry/fsimpl/devpts/slave.go
@@ -21,6 +21,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/fsimpl/kernfs"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
"gvisor.dev/gvisor/pkg/waiter"
@@ -33,6 +34,8 @@ type slaveInode struct {
kernfs.InodeNotDirectory
kernfs.InodeNotSymlink
+ locks lock.FileLocks
+
// Keep a reference to this inode's dentry.
dentry kernfs.Dentry
@@ -51,6 +54,7 @@ func (si *slaveInode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs
fd := &slaveFileDescription{
inode: si,
}
+ fd.LockFD.Init(&si.locks)
if err := fd.vfsfd.Init(fd, opts.Flags, rp.Mount(), vfsd, &vfs.FileDescriptionOptions{}); err != nil {
si.DecRef()
return nil, err
@@ -91,6 +95,7 @@ func (si *slaveInode) SetStat(ctx context.Context, vfsfs *vfs.Filesystem, creds
type slaveFileDescription struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
+ vfs.LockFD
inode *slaveInode
}
diff --git a/pkg/sentry/fsimpl/eventfd/eventfd.go b/pkg/sentry/fsimpl/eventfd/eventfd.go
index c573d7935..d12d78b84 100644
--- a/pkg/sentry/fsimpl/eventfd/eventfd.go
+++ b/pkg/sentry/fsimpl/eventfd/eventfd.go
@@ -37,6 +37,7 @@ type EventFileDescription struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
vfs.DentryMetadataFileDescriptionImpl
+ vfs.NoLockFD
// queue is used to notify interested parties when the event object
// becomes readable or writable.
diff --git a/pkg/sentry/fsimpl/ext/BUILD b/pkg/sentry/fsimpl/ext/BUILD
index ff861d0fe..973fa0def 100644
--- a/pkg/sentry/fsimpl/ext/BUILD
+++ b/pkg/sentry/fsimpl/ext/BUILD
@@ -60,6 +60,7 @@ go_library(
"//pkg/sentry/socket/unix/transport",
"//pkg/sentry/syscalls/linux",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/sync",
"//pkg/syserror",
"//pkg/usermem",
diff --git a/pkg/sentry/fsimpl/ext/file_description.go b/pkg/sentry/fsimpl/ext/file_description.go
index 92f7da40d..90b086468 100644
--- a/pkg/sentry/fsimpl/ext/file_description.go
+++ b/pkg/sentry/fsimpl/ext/file_description.go
@@ -26,6 +26,7 @@ import (
type fileDescription struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
+ vfs.LockFD
}
func (fd *fileDescription) filesystem() *filesystem {
diff --git a/pkg/sentry/fsimpl/ext/inode.go b/pkg/sentry/fsimpl/ext/inode.go
index 485f86f4b..e4b434b13 100644
--- a/pkg/sentry/fsimpl/ext/inode.go
+++ b/pkg/sentry/fsimpl/ext/inode.go
@@ -22,6 +22,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/fsimpl/ext/disklayout"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
)
@@ -54,6 +55,8 @@ type inode struct {
// diskInode gives us access to the inode struct on disk. Immutable.
diskInode disklayout.Inode
+ locks lock.FileLocks
+
// This is immutable. The first field of the implementations must have inode
// as the first field to ensure temporality.
impl interface{}
@@ -157,6 +160,7 @@ func (in *inode) open(rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts *vfs.OpenOpt
switch in.impl.(type) {
case *regularFile:
var fd regularFileFD
+ fd.LockFD.Init(&in.locks)
if err := fd.vfsfd.Init(&fd, opts.Flags, mnt, vfsd, &vfs.FileDescriptionOptions{}); err != nil {
return nil, err
}
@@ -168,6 +172,7 @@ func (in *inode) open(rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts *vfs.OpenOpt
return nil, syserror.EISDIR
}
var fd directoryFD
+ fd.LockFD.Init(&in.locks)
if err := fd.vfsfd.Init(&fd, opts.Flags, mnt, vfsd, &vfs.FileDescriptionOptions{}); err != nil {
return nil, err
}
@@ -178,6 +183,7 @@ func (in *inode) open(rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts *vfs.OpenOpt
return nil, syserror.ELOOP
}
var fd symlinkFD
+ fd.LockFD.Init(&in.locks)
fd.vfsfd.Init(&fd, opts.Flags, mnt, vfsd, &vfs.FileDescriptionOptions{})
return &fd.vfsfd, nil
default:
diff --git a/pkg/sentry/fsimpl/ext/regular_file.go b/pkg/sentry/fsimpl/ext/regular_file.go
index 30135ddb0..f7015c44f 100644
--- a/pkg/sentry/fsimpl/ext/regular_file.go
+++ b/pkg/sentry/fsimpl/ext/regular_file.go
@@ -77,6 +77,7 @@ func (in *inode) isRegular() bool {
// vfs.FileDescriptionImpl.
type regularFileFD struct {
fileDescription
+ vfs.LockFD
// off is the file offset. off is accessed using atomic memory operations.
off int64
diff --git a/pkg/sentry/fsimpl/gofer/BUILD b/pkg/sentry/fsimpl/gofer/BUILD
index f5f35a3bc..5cdeeaeb5 100644
--- a/pkg/sentry/fsimpl/gofer/BUILD
+++ b/pkg/sentry/fsimpl/gofer/BUILD
@@ -54,6 +54,7 @@ go_library(
"//pkg/p9",
"//pkg/safemem",
"//pkg/sentry/fs/fsutil",
+ "//pkg/sentry/fs/lock",
"//pkg/sentry/fsimpl/host",
"//pkg/sentry/hostfd",
"//pkg/sentry/kernel",
@@ -68,6 +69,7 @@ go_library(
"//pkg/sentry/socket/unix/transport",
"//pkg/sentry/usage",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/syserr",
"//pkg/syserror",
"//pkg/unet",
diff --git a/pkg/sentry/fsimpl/gofer/filesystem.go b/pkg/sentry/fsimpl/gofer/filesystem.go
index 36e0e1856..40933b74b 100644
--- a/pkg/sentry/fsimpl/gofer/filesystem.go
+++ b/pkg/sentry/fsimpl/gofer/filesystem.go
@@ -801,6 +801,7 @@ func (d *dentry) openLocked(ctx context.Context, rp *vfs.ResolvingPath, opts *vf
return nil, err
}
fd := &regularFileFD{}
+ fd.LockFD.Init(&d.locks)
if err := fd.vfsfd.Init(fd, opts.Flags, mnt, &d.vfsd, &vfs.FileDescriptionOptions{
AllowDirectIO: true,
}); err != nil {
@@ -826,6 +827,7 @@ func (d *dentry) openLocked(ctx context.Context, rp *vfs.ResolvingPath, opts *vf
}
}
fd := &directoryFD{}
+ fd.LockFD.Init(&d.locks)
if err := fd.vfsfd.Init(fd, opts.Flags, mnt, &d.vfsd, &vfs.FileDescriptionOptions{}); err != nil {
return nil, err
}
@@ -842,7 +844,7 @@ func (d *dentry) openLocked(ctx context.Context, rp *vfs.ResolvingPath, opts *vf
}
case linux.S_IFIFO:
if d.isSynthetic() {
- return d.pipe.Open(ctx, mnt, &d.vfsd, opts.Flags)
+ return d.pipe.Open(ctx, mnt, &d.vfsd, opts.Flags, &d.locks)
}
}
return d.openSpecialFileLocked(ctx, mnt, opts)
@@ -902,7 +904,7 @@ retry:
return nil, err
}
}
- fd, err := newSpecialFileFD(h, mnt, d, opts.Flags)
+ fd, err := newSpecialFileFD(h, mnt, d, &d.locks, opts.Flags)
if err != nil {
h.close(ctx)
return nil, err
@@ -989,6 +991,7 @@ func (d *dentry) createAndOpenChildLocked(ctx context.Context, rp *vfs.Resolving
var childVFSFD *vfs.FileDescription
if useRegularFileFD {
fd := &regularFileFD{}
+ fd.LockFD.Init(&child.locks)
if err := fd.vfsfd.Init(fd, opts.Flags, mnt, &child.vfsd, &vfs.FileDescriptionOptions{
AllowDirectIO: true,
}); err != nil {
@@ -1003,7 +1006,7 @@ func (d *dentry) createAndOpenChildLocked(ctx context.Context, rp *vfs.Resolving
if fdobj != nil {
h.fd = int32(fdobj.Release())
}
- fd, err := newSpecialFileFD(h, mnt, child, opts.Flags)
+ fd, err := newSpecialFileFD(h, mnt, child, &d.locks, opts.Flags)
if err != nil {
h.close(ctx)
return nil, err
diff --git a/pkg/sentry/fsimpl/gofer/gofer.go b/pkg/sentry/fsimpl/gofer/gofer.go
index 3f3bd56f0..0d88a328e 100644
--- a/pkg/sentry/fsimpl/gofer/gofer.go
+++ b/pkg/sentry/fsimpl/gofer/gofer.go
@@ -45,6 +45,7 @@ import (
"gvisor.dev/gvisor/pkg/log"
"gvisor.dev/gvisor/pkg/p9"
"gvisor.dev/gvisor/pkg/sentry/fs/fsutil"
+ fslock "gvisor.dev/gvisor/pkg/sentry/fs/lock"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/kernel/pipe"
ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time"
@@ -52,6 +53,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/pgalloc"
"gvisor.dev/gvisor/pkg/sentry/socket/unix/transport"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/unet"
"gvisor.dev/gvisor/pkg/usermem"
@@ -662,6 +664,8 @@ type dentry struct {
// If this dentry represents a synthetic named pipe, pipe is the pipe
// endpoint bound to this file.
pipe *pipe.VFSPipe
+
+ locks lock.FileLocks
}
// dentryAttrMask returns a p9.AttrMask enabling all attributes used by the
@@ -1366,6 +1370,9 @@ func (d *dentry) decLinks() {
type fileDescription struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
+ vfs.LockFD
+
+ lockLogging sync.Once
}
func (fd *fileDescription) filesystem() *filesystem {
@@ -1416,3 +1423,19 @@ func (fd *fileDescription) Setxattr(ctx context.Context, opts vfs.SetxattrOption
func (fd *fileDescription) Removexattr(ctx context.Context, name string) error {
return fd.dentry().removexattr(ctx, auth.CredentialsFromContext(ctx), name)
}
+
+// LockBSD implements vfs.FileDescriptionImpl.LockBSD.
+func (fd *fileDescription) LockBSD(ctx context.Context, uid fslock.UniqueID, t fslock.LockType, block fslock.Blocker) error {
+ fd.lockLogging.Do(func() {
+ log.Infof("File lock using gofer file handled internally.")
+ })
+ return fd.LockFD.LockBSD(ctx, uid, t, block)
+}
+
+// LockPOSIX implements vfs.FileDescriptionImpl.LockPOSIX.
+func (fd *fileDescription) LockPOSIX(ctx context.Context, uid fslock.UniqueID, t fslock.LockType, rng fslock.LockRange, block fslock.Blocker) error {
+ fd.lockLogging.Do(func() {
+ log.Infof("Range lock using gofer file handled internally.")
+ })
+ return fd.LockFD.LockPOSIX(ctx, uid, t, rng, block)
+}
diff --git a/pkg/sentry/fsimpl/gofer/special_file.go b/pkg/sentry/fsimpl/gofer/special_file.go
index ff6126b87..289efdd25 100644
--- a/pkg/sentry/fsimpl/gofer/special_file.go
+++ b/pkg/sentry/fsimpl/gofer/special_file.go
@@ -22,6 +22,7 @@ import (
"gvisor.dev/gvisor/pkg/fdnotifier"
"gvisor.dev/gvisor/pkg/safemem"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
"gvisor.dev/gvisor/pkg/waiter"
@@ -51,7 +52,7 @@ type specialFileFD struct {
off int64
}
-func newSpecialFileFD(h handle, mnt *vfs.Mount, d *dentry, flags uint32) (*specialFileFD, error) {
+func newSpecialFileFD(h handle, mnt *vfs.Mount, d *dentry, locks *lock.FileLocks, flags uint32) (*specialFileFD, error) {
ftype := d.fileType()
seekable := ftype == linux.S_IFREG
mayBlock := ftype == linux.S_IFIFO || ftype == linux.S_IFSOCK
@@ -60,6 +61,7 @@ func newSpecialFileFD(h handle, mnt *vfs.Mount, d *dentry, flags uint32) (*speci
seekable: seekable,
mayBlock: mayBlock,
}
+ fd.LockFD.Init(locks)
if mayBlock && h.fd >= 0 {
if err := fdnotifier.AddFD(h.fd, &fd.queue); err != nil {
return nil, err
diff --git a/pkg/sentry/fsimpl/host/BUILD b/pkg/sentry/fsimpl/host/BUILD
index ca0fe6d2b..54f16ad63 100644
--- a/pkg/sentry/fsimpl/host/BUILD
+++ b/pkg/sentry/fsimpl/host/BUILD
@@ -39,6 +39,7 @@ go_library(
"//pkg/sentry/unimpl",
"//pkg/sentry/uniqueid",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/sync",
"//pkg/syserr",
"//pkg/syserror",
diff --git a/pkg/sentry/fsimpl/host/host.go b/pkg/sentry/fsimpl/host/host.go
index 18b127521..5ec5100b8 100644
--- a/pkg/sentry/fsimpl/host/host.go
+++ b/pkg/sentry/fsimpl/host/host.go
@@ -34,6 +34,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/memmap"
unixsocket "gvisor.dev/gvisor/pkg/sentry/socket/unix"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
@@ -182,6 +183,8 @@ type inode struct {
kernfs.InodeNotDirectory
kernfs.InodeNotSymlink
+ locks lock.FileLocks
+
// When the reference count reaches zero, the host fd is closed.
refs.AtomicRefCount
@@ -468,7 +471,7 @@ func (i *inode) open(ctx context.Context, d *vfs.Dentry, mnt *vfs.Mount, flags u
return nil, err
}
// Currently, we only allow Unix sockets to be imported.
- return unixsocket.NewFileDescription(ep, ep.Type(), flags, mnt, d)
+ return unixsocket.NewFileDescription(ep, ep.Type(), flags, mnt, d, &i.locks)
}
// TODO(gvisor.dev/issue/1672): Whitelist specific file types here, so that
@@ -478,6 +481,7 @@ func (i *inode) open(ctx context.Context, d *vfs.Dentry, mnt *vfs.Mount, flags u
fileDescription: fileDescription{inode: i},
termios: linux.DefaultSlaveTermios,
}
+ fd.LockFD.Init(&i.locks)
vfsfd := &fd.vfsfd
if err := vfsfd.Init(fd, flags, mnt, d, &vfs.FileDescriptionOptions{}); err != nil {
return nil, err
@@ -486,6 +490,7 @@ func (i *inode) open(ctx context.Context, d *vfs.Dentry, mnt *vfs.Mount, flags u
}
fd := &fileDescription{inode: i}
+ fd.LockFD.Init(&i.locks)
vfsfd := &fd.vfsfd
if err := vfsfd.Init(fd, flags, mnt, d, &vfs.FileDescriptionOptions{}); err != nil {
return nil, err
@@ -497,6 +502,7 @@ func (i *inode) open(ctx context.Context, d *vfs.Dentry, mnt *vfs.Mount, flags u
type fileDescription struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
+ vfs.LockFD
// inode is vfsfd.Dentry().Impl().(*kernfs.Dentry).Inode().(*inode), but
// cached to reduce indirections and casting. fileDescription does not hold
diff --git a/pkg/sentry/fsimpl/kernfs/BUILD b/pkg/sentry/fsimpl/kernfs/BUILD
index ef34cb28a..0299dbde9 100644
--- a/pkg/sentry/fsimpl/kernfs/BUILD
+++ b/pkg/sentry/fsimpl/kernfs/BUILD
@@ -49,6 +49,7 @@ go_library(
"//pkg/sentry/memmap",
"//pkg/sentry/socket/unix/transport",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/sync",
"//pkg/syserror",
"//pkg/usermem",
@@ -67,6 +68,7 @@ go_test(
"//pkg/sentry/fsimpl/testutil",
"//pkg/sentry/kernel/auth",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/syserror",
"//pkg/usermem",
"@com_github_google_go-cmp//cmp:go_default_library",
diff --git a/pkg/sentry/fsimpl/kernfs/dynamic_bytes_file.go b/pkg/sentry/fsimpl/kernfs/dynamic_bytes_file.go
index 1568a9d49..6418de0a3 100644
--- a/pkg/sentry/fsimpl/kernfs/dynamic_bytes_file.go
+++ b/pkg/sentry/fsimpl/kernfs/dynamic_bytes_file.go
@@ -21,6 +21,7 @@ import (
"gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
)
@@ -38,7 +39,8 @@ type DynamicBytesFile struct {
InodeNotDirectory
InodeNotSymlink
- data vfs.DynamicBytesSource
+ locks lock.FileLocks
+ data vfs.DynamicBytesSource
}
var _ Inode = (*DynamicBytesFile)(nil)
@@ -55,7 +57,7 @@ func (f *DynamicBytesFile) Init(creds *auth.Credentials, devMajor, devMinor uint
// Open implements Inode.Open.
func (f *DynamicBytesFile) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
fd := &DynamicBytesFD{}
- if err := fd.Init(rp.Mount(), vfsd, f.data, opts.Flags); err != nil {
+ if err := fd.Init(rp.Mount(), vfsd, f.data, &f.locks, opts.Flags); err != nil {
return nil, err
}
return &fd.vfsfd, nil
@@ -77,13 +79,15 @@ func (*DynamicBytesFile) SetStat(context.Context, *vfs.Filesystem, *auth.Credent
type DynamicBytesFD struct {
vfs.FileDescriptionDefaultImpl
vfs.DynamicBytesFileDescriptionImpl
+ vfs.LockFD
vfsfd vfs.FileDescription
inode Inode
}
// Init initializes a DynamicBytesFD.
-func (fd *DynamicBytesFD) Init(m *vfs.Mount, d *vfs.Dentry, data vfs.DynamicBytesSource, flags uint32) error {
+func (fd *DynamicBytesFD) Init(m *vfs.Mount, d *vfs.Dentry, data vfs.DynamicBytesSource, locks *lock.FileLocks, flags uint32) error {
+ fd.LockFD.Init(locks)
if err := fd.vfsfd.Init(fd, flags, m, d, &vfs.FileDescriptionOptions{}); err != nil {
return err
}
diff --git a/pkg/sentry/fsimpl/kernfs/fd_impl_util.go b/pkg/sentry/fsimpl/kernfs/fd_impl_util.go
index 8284e76a7..33a5968ca 100644
--- a/pkg/sentry/fsimpl/kernfs/fd_impl_util.go
+++ b/pkg/sentry/fsimpl/kernfs/fd_impl_util.go
@@ -22,6 +22,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/memmap"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
@@ -42,6 +43,7 @@ import (
type GenericDirectoryFD struct {
vfs.FileDescriptionDefaultImpl
vfs.DirectoryFileDescriptionDefaultImpl
+ vfs.LockFD
vfsfd vfs.FileDescription
children *OrderedChildren
@@ -55,9 +57,9 @@ type GenericDirectoryFD struct {
// NewGenericDirectoryFD creates a new GenericDirectoryFD and returns its
// dentry.
-func NewGenericDirectoryFD(m *vfs.Mount, d *vfs.Dentry, children *OrderedChildren, opts *vfs.OpenOptions) (*GenericDirectoryFD, error) {
+func NewGenericDirectoryFD(m *vfs.Mount, d *vfs.Dentry, children *OrderedChildren, locks *lock.FileLocks, opts *vfs.OpenOptions) (*GenericDirectoryFD, error) {
fd := &GenericDirectoryFD{}
- if err := fd.Init(children, opts); err != nil {
+ if err := fd.Init(children, locks, opts); err != nil {
return nil, err
}
if err := fd.vfsfd.Init(fd, opts.Flags, m, d, &vfs.FileDescriptionOptions{}); err != nil {
@@ -69,11 +71,12 @@ func NewGenericDirectoryFD(m *vfs.Mount, d *vfs.Dentry, children *OrderedChildre
// Init initializes a GenericDirectoryFD. Use it when overriding
// GenericDirectoryFD. Caller must call fd.VFSFileDescription.Init() with the
// correct implementation.
-func (fd *GenericDirectoryFD) Init(children *OrderedChildren, opts *vfs.OpenOptions) error {
+func (fd *GenericDirectoryFD) Init(children *OrderedChildren, locks *lock.FileLocks, opts *vfs.OpenOptions) error {
if vfs.AccessTypesForOpenFlags(opts)&vfs.MayWrite != 0 {
// Can't open directories for writing.
return syserror.EISDIR
}
+ fd.LockFD.Init(locks)
fd.children = children
return nil
}
diff --git a/pkg/sentry/fsimpl/kernfs/inode_impl_util.go b/pkg/sentry/fsimpl/kernfs/inode_impl_util.go
index 982daa2e6..0e4927215 100644
--- a/pkg/sentry/fsimpl/kernfs/inode_impl_util.go
+++ b/pkg/sentry/fsimpl/kernfs/inode_impl_util.go
@@ -23,6 +23,7 @@ import (
"gvisor.dev/gvisor/pkg/refs"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserror"
)
@@ -555,6 +556,8 @@ type StaticDirectory struct {
InodeAttrs
InodeNoDynamicLookup
OrderedChildren
+
+ locks lock.FileLocks
}
var _ Inode = (*StaticDirectory)(nil)
@@ -584,7 +587,7 @@ func (s *StaticDirectory) Init(creds *auth.Credentials, devMajor, devMinor uint3
// Open implements kernfs.Inode.
func (s *StaticDirectory) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
- fd, err := NewGenericDirectoryFD(rp.Mount(), vfsd, &s.OrderedChildren, &opts)
+ fd, err := NewGenericDirectoryFD(rp.Mount(), vfsd, &s.OrderedChildren, &s.locks, &opts)
if err != nil {
return nil, err
}
diff --git a/pkg/sentry/fsimpl/kernfs/kernfs_test.go b/pkg/sentry/fsimpl/kernfs/kernfs_test.go
index 412cf6ac9..6749facf7 100644
--- a/pkg/sentry/fsimpl/kernfs/kernfs_test.go
+++ b/pkg/sentry/fsimpl/kernfs/kernfs_test.go
@@ -27,6 +27,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/fsimpl/testutil"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
)
@@ -100,8 +101,10 @@ type readonlyDir struct {
kernfs.InodeNotSymlink
kernfs.InodeNoDynamicLookup
kernfs.InodeDirectoryNoNewChildren
-
kernfs.OrderedChildren
+
+ locks lock.FileLocks
+
dentry kernfs.Dentry
}
@@ -117,7 +120,7 @@ func (fs *filesystem) newReadonlyDir(creds *auth.Credentials, mode linux.FileMod
}
func (d *readonlyDir) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
- fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &d.OrderedChildren, &opts)
+ fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &d.OrderedChildren, &d.locks, &opts)
if err != nil {
return nil, err
}
@@ -128,10 +131,12 @@ type dir struct {
attrs
kernfs.InodeNotSymlink
kernfs.InodeNoDynamicLookup
+ kernfs.OrderedChildren
+
+ locks lock.FileLocks
fs *filesystem
dentry kernfs.Dentry
- kernfs.OrderedChildren
}
func (fs *filesystem) newDir(creds *auth.Credentials, mode linux.FileMode, contents map[string]*kernfs.Dentry) *kernfs.Dentry {
@@ -147,7 +152,7 @@ func (fs *filesystem) newDir(creds *auth.Credentials, mode linux.FileMode, conte
}
func (d *dir) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
- fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &d.OrderedChildren, &opts)
+ fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &d.OrderedChildren, &d.locks, &opts)
if err != nil {
return nil, err
}
diff --git a/pkg/sentry/fsimpl/pipefs/BUILD b/pkg/sentry/fsimpl/pipefs/BUILD
index 5950a2d59..c618dbe6c 100644
--- a/pkg/sentry/fsimpl/pipefs/BUILD
+++ b/pkg/sentry/fsimpl/pipefs/BUILD
@@ -15,6 +15,7 @@ go_library(
"//pkg/sentry/kernel/pipe",
"//pkg/sentry/kernel/time",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/syserror",
"//pkg/usermem",
],
diff --git a/pkg/sentry/fsimpl/pipefs/pipefs.go b/pkg/sentry/fsimpl/pipefs/pipefs.go
index cab771211..e4dabaa33 100644
--- a/pkg/sentry/fsimpl/pipefs/pipefs.go
+++ b/pkg/sentry/fsimpl/pipefs/pipefs.go
@@ -27,6 +27,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/kernel/pipe"
ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
)
@@ -81,7 +82,8 @@ type inode struct {
kernfs.InodeNotSymlink
kernfs.InodeNoopRefCount
- pipe *pipe.VFSPipe
+ locks lock.FileLocks
+ pipe *pipe.VFSPipe
ino uint64
uid auth.KUID
@@ -147,7 +149,7 @@ func (i *inode) SetStat(ctx context.Context, vfsfs *vfs.Filesystem, creds *auth.
// Open implements kernfs.Inode.Open.
func (i *inode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
- return i.pipe.Open(ctx, rp.Mount(), vfsd, opts.Flags)
+ return i.pipe.Open(ctx, rp.Mount(), vfsd, opts.Flags, &i.locks)
}
// NewConnectedPipeFDs returns a pair of FileDescriptions representing the read
diff --git a/pkg/sentry/fsimpl/proc/BUILD b/pkg/sentry/fsimpl/proc/BUILD
index 17c1342b5..351ba4ee9 100644
--- a/pkg/sentry/fsimpl/proc/BUILD
+++ b/pkg/sentry/fsimpl/proc/BUILD
@@ -35,6 +35,7 @@ go_library(
"//pkg/sentry/socket/unix/transport",
"//pkg/sentry/usage",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/syserror",
"//pkg/tcpip/header",
"//pkg/usermem",
diff --git a/pkg/sentry/fsimpl/proc/subtasks.go b/pkg/sentry/fsimpl/proc/subtasks.go
index 36a911db4..e2cdb7ee9 100644
--- a/pkg/sentry/fsimpl/proc/subtasks.go
+++ b/pkg/sentry/fsimpl/proc/subtasks.go
@@ -24,6 +24,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/kernel"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
)
@@ -37,6 +38,8 @@ type subtasksInode struct {
kernfs.OrderedChildren
kernfs.AlwaysValid
+ locks lock.FileLocks
+
fs *filesystem
task *kernel.Task
pidns *kernel.PIDNamespace
@@ -153,7 +156,7 @@ func (fd *subtasksFD) SetStat(ctx context.Context, opts vfs.SetStatOptions) erro
// Open implements kernfs.Inode.
func (i *subtasksInode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
fd := &subtasksFD{task: i.task}
- if err := fd.Init(&i.OrderedChildren, &opts); err != nil {
+ if err := fd.Init(&i.OrderedChildren, &i.locks, &opts); err != nil {
return nil, err
}
if err := fd.VFSFileDescription().Init(fd, opts.Flags, rp.Mount(), vfsd, &vfs.FileDescriptionOptions{}); err != nil {
diff --git a/pkg/sentry/fsimpl/proc/task.go b/pkg/sentry/fsimpl/proc/task.go
index 482055db1..44078a765 100644
--- a/pkg/sentry/fsimpl/proc/task.go
+++ b/pkg/sentry/fsimpl/proc/task.go
@@ -25,6 +25,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/mm"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
)
@@ -38,6 +39,8 @@ type taskInode struct {
kernfs.InodeAttrs
kernfs.OrderedChildren
+ locks lock.FileLocks
+
task *kernel.Task
}
@@ -103,7 +106,7 @@ func (i *taskInode) Valid(ctx context.Context) bool {
// Open implements kernfs.Inode.
func (i *taskInode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
- fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &i.OrderedChildren, &opts)
+ fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &i.OrderedChildren, &i.locks, &opts)
if err != nil {
return nil, err
}
diff --git a/pkg/sentry/fsimpl/proc/task_fds.go b/pkg/sentry/fsimpl/proc/task_fds.go
index 44ccc9e4a..ef6c1d04f 100644
--- a/pkg/sentry/fsimpl/proc/task_fds.go
+++ b/pkg/sentry/fsimpl/proc/task_fds.go
@@ -27,6 +27,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/kernel"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
)
@@ -53,6 +54,8 @@ func taskFDExists(t *kernel.Task, fd int32) bool {
}
type fdDir struct {
+ locks lock.FileLocks
+
fs *filesystem
task *kernel.Task
@@ -143,7 +146,7 @@ func (i *fdDirInode) Lookup(ctx context.Context, name string) (*vfs.Dentry, erro
// Open implements kernfs.Inode.
func (i *fdDirInode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
- fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &i.OrderedChildren, &opts)
+ fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &i.OrderedChildren, &i.locks, &opts)
if err != nil {
return nil, err
}
@@ -270,7 +273,7 @@ func (i *fdInfoDirInode) Lookup(ctx context.Context, name string) (*vfs.Dentry,
// Open implements kernfs.Inode.
func (i *fdInfoDirInode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
- fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &i.OrderedChildren, &opts)
+ fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &i.OrderedChildren, &i.locks, &opts)
if err != nil {
return nil, err
}
diff --git a/pkg/sentry/fsimpl/proc/task_files.go b/pkg/sentry/fsimpl/proc/task_files.go
index 2f297e48a..e5eaa91cd 100644
--- a/pkg/sentry/fsimpl/proc/task_files.go
+++ b/pkg/sentry/fsimpl/proc/task_files.go
@@ -30,6 +30,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/mm"
"gvisor.dev/gvisor/pkg/sentry/usage"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
)
@@ -775,6 +776,8 @@ type namespaceInode struct {
kernfs.InodeNoopRefCount
kernfs.InodeNotDirectory
kernfs.InodeNotSymlink
+
+ locks lock.FileLocks
}
var _ kernfs.Inode = (*namespaceInode)(nil)
@@ -791,6 +794,7 @@ func (i *namespaceInode) Init(creds *auth.Credentials, devMajor, devMinor uint32
func (i *namespaceInode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
fd := &namespaceFD{inode: i}
i.IncRef()
+ fd.LockFD.Init(&i.locks)
if err := fd.vfsfd.Init(fd, opts.Flags, rp.Mount(), vfsd, &vfs.FileDescriptionOptions{}); err != nil {
return nil, err
}
@@ -801,6 +805,7 @@ func (i *namespaceInode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *
// /proc/[pid]/ns/*.
type namespaceFD struct {
vfs.FileDescriptionDefaultImpl
+ vfs.LockFD
vfsfd vfs.FileDescription
inode *namespaceInode
@@ -825,8 +830,3 @@ func (fd *namespaceFD) SetStat(ctx context.Context, opts vfs.SetStatOptions) err
func (fd *namespaceFD) Release() {
fd.inode.DecRef()
}
-
-// OnClose implements FileDescriptionImpl.
-func (*namespaceFD) OnClose(context.Context) error {
- return nil
-}
diff --git a/pkg/sentry/fsimpl/proc/tasks.go b/pkg/sentry/fsimpl/proc/tasks.go
index b51d43954..58c8b9d05 100644
--- a/pkg/sentry/fsimpl/proc/tasks.go
+++ b/pkg/sentry/fsimpl/proc/tasks.go
@@ -25,6 +25,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/kernel"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
)
@@ -43,6 +44,8 @@ type tasksInode struct {
kernfs.OrderedChildren
kernfs.AlwaysValid
+ locks lock.FileLocks
+
fs *filesystem
pidns *kernel.PIDNamespace
@@ -197,7 +200,7 @@ func (i *tasksInode) IterDirents(ctx context.Context, cb vfs.IterDirentsCallback
// Open implements kernfs.Inode.
func (i *tasksInode) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
- fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &i.OrderedChildren, &opts)
+ fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &i.OrderedChildren, &i.locks, &opts)
if err != nil {
return nil, err
}
diff --git a/pkg/sentry/fsimpl/signalfd/signalfd.go b/pkg/sentry/fsimpl/signalfd/signalfd.go
index d29ef3f83..242ba9b5d 100644
--- a/pkg/sentry/fsimpl/signalfd/signalfd.go
+++ b/pkg/sentry/fsimpl/signalfd/signalfd.go
@@ -31,6 +31,7 @@ type SignalFileDescription struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
vfs.DentryMetadataFileDescriptionImpl
+ vfs.NoLockFD
// target is the original signal target task.
//
diff --git a/pkg/sentry/fsimpl/sys/BUILD b/pkg/sentry/fsimpl/sys/BUILD
index a741e2bb6..237f17def 100644
--- a/pkg/sentry/fsimpl/sys/BUILD
+++ b/pkg/sentry/fsimpl/sys/BUILD
@@ -15,6 +15,7 @@ go_library(
"//pkg/sentry/kernel",
"//pkg/sentry/kernel/auth",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/syserror",
],
)
diff --git a/pkg/sentry/fsimpl/sys/sys.go b/pkg/sentry/fsimpl/sys/sys.go
index 0af373604..b84463d3a 100644
--- a/pkg/sentry/fsimpl/sys/sys.go
+++ b/pkg/sentry/fsimpl/sys/sys.go
@@ -25,6 +25,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/kernel"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserror"
)
@@ -98,8 +99,10 @@ type dir struct {
kernfs.InodeNoDynamicLookup
kernfs.InodeNotSymlink
kernfs.InodeDirectoryNoNewChildren
-
kernfs.OrderedChildren
+
+ locks lock.FileLocks
+
dentry kernfs.Dentry
}
@@ -121,7 +124,7 @@ func (*dir) SetStat(context.Context, *vfs.Filesystem, *auth.Credentials, vfs.Set
// Open implements kernfs.Inode.Open.
func (d *dir) Open(ctx context.Context, rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error) {
- fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &d.OrderedChildren, &opts)
+ fd, err := kernfs.NewGenericDirectoryFD(rp.Mount(), vfsd, &d.OrderedChildren, &d.locks, &opts)
if err != nil {
return nil, err
}
diff --git a/pkg/sentry/fsimpl/timerfd/timerfd.go b/pkg/sentry/fsimpl/timerfd/timerfd.go
index 60c92d626..2dc90d484 100644
--- a/pkg/sentry/fsimpl/timerfd/timerfd.go
+++ b/pkg/sentry/fsimpl/timerfd/timerfd.go
@@ -32,6 +32,7 @@ type TimerFileDescription struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
vfs.DentryMetadataFileDescriptionImpl
+ vfs.NoLockFD
events waiter.Queue
timer *ktime.Timer
diff --git a/pkg/sentry/fsimpl/tmpfs/filesystem.go b/pkg/sentry/fsimpl/tmpfs/filesystem.go
index e801680e8..72399b321 100644
--- a/pkg/sentry/fsimpl/tmpfs/filesystem.go
+++ b/pkg/sentry/fsimpl/tmpfs/filesystem.go
@@ -399,6 +399,7 @@ func (d *dentry) open(ctx context.Context, rp *vfs.ResolvingPath, opts *vfs.Open
switch impl := d.inode.impl.(type) {
case *regularFile:
var fd regularFileFD
+ fd.LockFD.Init(&d.inode.locks)
if err := fd.vfsfd.Init(&fd, opts.Flags, rp.Mount(), &d.vfsd, &vfs.FileDescriptionOptions{}); err != nil {
return nil, err
}
@@ -414,15 +415,16 @@ func (d *dentry) open(ctx context.Context, rp *vfs.ResolvingPath, opts *vfs.Open
return nil, syserror.EISDIR
}
var fd directoryFD
+ fd.LockFD.Init(&d.inode.locks)
if err := fd.vfsfd.Init(&fd, opts.Flags, rp.Mount(), &d.vfsd, &vfs.FileDescriptionOptions{}); err != nil {
return nil, err
}
return &fd.vfsfd, nil
case *symlink:
- // Can't open symlinks without O_PATH (which is unimplemented).
+ // TODO(gvisor.dev/issue/2782): Can't open symlinks without O_PATH.
return nil, syserror.ELOOP
case *namedPipe:
- return impl.pipe.Open(ctx, rp.Mount(), &d.vfsd, opts.Flags)
+ return impl.pipe.Open(ctx, rp.Mount(), &d.vfsd, opts.Flags, &d.inode.locks)
case *deviceFile:
return rp.VirtualFilesystem().OpenDeviceSpecialFile(ctx, rp.Mount(), &d.vfsd, impl.kind, impl.major, impl.minor, opts)
case *socketFile:
diff --git a/pkg/sentry/fsimpl/tmpfs/regular_file.go b/pkg/sentry/fsimpl/tmpfs/regular_file.go
index 4f2ae04d2..77447b32c 100644
--- a/pkg/sentry/fsimpl/tmpfs/regular_file.go
+++ b/pkg/sentry/fsimpl/tmpfs/regular_file.go
@@ -25,7 +25,6 @@ import (
"gvisor.dev/gvisor/pkg/safemem"
"gvisor.dev/gvisor/pkg/sentry/fs"
"gvisor.dev/gvisor/pkg/sentry/fs/fsutil"
- "gvisor.dev/gvisor/pkg/sentry/fs/lock"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/memmap"
"gvisor.dev/gvisor/pkg/sentry/pgalloc"
@@ -366,28 +365,6 @@ func (fd *regularFileFD) Sync(ctx context.Context) error {
return nil
}
-// LockBSD implements vfs.FileDescriptionImpl.LockBSD.
-func (fd *regularFileFD) LockBSD(ctx context.Context, uid lock.UniqueID, t lock.LockType, block lock.Blocker) error {
- return fd.inode().lockBSD(uid, t, block)
-}
-
-// UnlockBSD implements vfs.FileDescriptionImpl.UnlockBSD.
-func (fd *regularFileFD) UnlockBSD(ctx context.Context, uid lock.UniqueID) error {
- fd.inode().unlockBSD(uid)
- return nil
-}
-
-// LockPOSIX implements vfs.FileDescriptionImpl.LockPOSIX.
-func (fd *regularFileFD) LockPOSIX(ctx context.Context, uid lock.UniqueID, t lock.LockType, rng lock.LockRange, block lock.Blocker) error {
- return fd.inode().lockPOSIX(uid, t, rng, block)
-}
-
-// UnlockPOSIX implements vfs.FileDescriptionImpl.UnlockPOSIX.
-func (fd *regularFileFD) UnlockPOSIX(ctx context.Context, uid lock.UniqueID, rng lock.LockRange) error {
- fd.inode().unlockPOSIX(uid, rng)
- return nil
-}
-
// ConfigureMMap implements vfs.FileDescriptionImpl.ConfigureMMap.
func (fd *regularFileFD) ConfigureMMap(ctx context.Context, opts *memmap.MMapOpts) error {
file := fd.inode().impl.(*regularFile)
diff --git a/pkg/sentry/fsimpl/tmpfs/tmpfs.go b/pkg/sentry/fsimpl/tmpfs/tmpfs.go
index 7ce1b86c7..71a7522af 100644
--- a/pkg/sentry/fsimpl/tmpfs/tmpfs.go
+++ b/pkg/sentry/fsimpl/tmpfs/tmpfs.go
@@ -36,7 +36,6 @@ import (
"gvisor.dev/gvisor/pkg/abi/linux"
"gvisor.dev/gvisor/pkg/context"
- fslock "gvisor.dev/gvisor/pkg/sentry/fs/lock"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
"gvisor.dev/gvisor/pkg/sentry/kernel/time"
"gvisor.dev/gvisor/pkg/sentry/pgalloc"
@@ -311,7 +310,6 @@ type inode struct {
ctime int64 // nanoseconds
mtime int64 // nanoseconds
- // Advisory file locks, which lock at the inode level.
locks lock.FileLocks
// Inotify watches for this inode.
@@ -539,44 +537,6 @@ func (i *inode) setStat(ctx context.Context, creds *auth.Credentials, stat *linu
return nil
}
-// TODO(gvisor.dev/issue/1480): support file locking for file types other than regular.
-func (i *inode) lockBSD(uid fslock.UniqueID, t fslock.LockType, block fslock.Blocker) error {
- switch i.impl.(type) {
- case *regularFile:
- return i.locks.LockBSD(uid, t, block)
- }
- return syserror.EBADF
-}
-
-// TODO(gvisor.dev/issue/1480): support file locking for file types other than regular.
-func (i *inode) unlockBSD(uid fslock.UniqueID) error {
- switch i.impl.(type) {
- case *regularFile:
- i.locks.UnlockBSD(uid)
- return nil
- }
- return syserror.EBADF
-}
-
-// TODO(gvisor.dev/issue/1480): support file locking for file types other than regular.
-func (i *inode) lockPOSIX(uid fslock.UniqueID, t fslock.LockType, rng fslock.LockRange, block fslock.Blocker) error {
- switch i.impl.(type) {
- case *regularFile:
- return i.locks.LockPOSIX(uid, t, rng, block)
- }
- return syserror.EBADF
-}
-
-// TODO(gvisor.dev/issue/1480): support file locking for file types other than regular.
-func (i *inode) unlockPOSIX(uid fslock.UniqueID, rng fslock.LockRange) error {
- switch i.impl.(type) {
- case *regularFile:
- i.locks.UnlockPOSIX(uid, rng)
- return nil
- }
- return syserror.EBADF
-}
-
// allocatedBlocksForSize returns the number of 512B blocks needed to
// accommodate the given size in bytes, as appropriate for struct
// stat::st_blocks and struct statx::stx_blocks. (Note that this 512B block
@@ -708,6 +668,7 @@ func (i *inode) userXattrSupported() bool {
type fileDescription struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
+ vfs.LockFD
}
func (fd *fileDescription) filesystem() *filesystem {
diff --git a/pkg/sentry/kernel/fd_table.go b/pkg/sentry/kernel/fd_table.go
index dbfcef0fa..b35afafe3 100644
--- a/pkg/sentry/kernel/fd_table.go
+++ b/pkg/sentry/kernel/fd_table.go
@@ -80,9 +80,6 @@ type FDTable struct {
refs.AtomicRefCount
k *Kernel
- // uid is a unique identifier.
- uid uint64
-
// mu protects below.
mu sync.Mutex `state:"nosave"`
@@ -130,7 +127,7 @@ func (f *FDTable) loadDescriptorTable(m map[int32]descriptor) {
// drop drops the table reference.
func (f *FDTable) drop(file *fs.File) {
// Release locks.
- file.Dirent.Inode.LockCtx.Posix.UnlockRegion(lock.UniqueID(f.uid), lock.LockRange{0, lock.LockEOF})
+ file.Dirent.Inode.LockCtx.Posix.UnlockRegion(f, lock.LockRange{0, lock.LockEOF})
// Send inotify events.
d := file.Dirent
@@ -164,17 +161,9 @@ func (f *FDTable) dropVFS2(file *vfs.FileDescription) {
file.DecRef()
}
-// ID returns a unique identifier for this FDTable.
-func (f *FDTable) ID() uint64 {
- return f.uid
-}
-
// NewFDTable allocates a new FDTable that may be used by tasks in k.
func (k *Kernel) NewFDTable() *FDTable {
- f := &FDTable{
- k: k,
- uid: atomic.AddUint64(&k.fdMapUids, 1),
- }
+ f := &FDTable{k: k}
f.init()
return f
}
diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go
index 5efeb3767..bcbeb6a39 100644
--- a/pkg/sentry/kernel/kernel.go
+++ b/pkg/sentry/kernel/kernel.go
@@ -194,11 +194,6 @@ type Kernel struct {
// cpuClockTickerSetting is protected by runningTasksMu.
cpuClockTickerSetting ktime.Setting
- // fdMapUids is an ever-increasing counter for generating FDTable uids.
- //
- // fdMapUids is mutable, and is accessed using atomic memory operations.
- fdMapUids uint64
-
// uniqueID is used to generate unique identifiers.
//
// uniqueID is mutable, and is accessed using atomic memory operations.
diff --git a/pkg/sentry/kernel/pipe/BUILD b/pkg/sentry/kernel/pipe/BUILD
index 7bfa9075a..0db546b98 100644
--- a/pkg/sentry/kernel/pipe/BUILD
+++ b/pkg/sentry/kernel/pipe/BUILD
@@ -27,6 +27,7 @@ go_library(
"//pkg/sentry/fs",
"//pkg/sentry/fs/fsutil",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/sync",
"//pkg/syserror",
"//pkg/usermem",
diff --git a/pkg/sentry/kernel/pipe/vfs.go b/pkg/sentry/kernel/pipe/vfs.go
index 2602bed72..c0e9ee1f4 100644
--- a/pkg/sentry/kernel/pipe/vfs.go
+++ b/pkg/sentry/kernel/pipe/vfs.go
@@ -21,6 +21,7 @@ import (
"gvisor.dev/gvisor/pkg/safemem"
"gvisor.dev/gvisor/pkg/sentry/arch"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
@@ -61,11 +62,13 @@ func NewVFSPipe(isNamed bool, sizeBytes, atomicIOBytes int64) *VFSPipe {
//
// Preconditions: statusFlags should not contain an open access mode.
func (vp *VFSPipe) ReaderWriterPair(mnt *vfs.Mount, vfsd *vfs.Dentry, statusFlags uint32) (*vfs.FileDescription, *vfs.FileDescription) {
- return vp.newFD(mnt, vfsd, linux.O_RDONLY|statusFlags), vp.newFD(mnt, vfsd, linux.O_WRONLY|statusFlags)
+ // Connected pipes share the same locks.
+ locks := &lock.FileLocks{}
+ return vp.newFD(mnt, vfsd, linux.O_RDONLY|statusFlags, locks), vp.newFD(mnt, vfsd, linux.O_WRONLY|statusFlags, locks)
}
// Open opens the pipe represented by vp.
-func (vp *VFSPipe) Open(ctx context.Context, mnt *vfs.Mount, vfsd *vfs.Dentry, statusFlags uint32) (*vfs.FileDescription, error) {
+func (vp *VFSPipe) Open(ctx context.Context, mnt *vfs.Mount, vfsd *vfs.Dentry, statusFlags uint32, locks *lock.FileLocks) (*vfs.FileDescription, error) {
vp.mu.Lock()
defer vp.mu.Unlock()
@@ -75,7 +78,7 @@ func (vp *VFSPipe) Open(ctx context.Context, mnt *vfs.Mount, vfsd *vfs.Dentry, s
return nil, syserror.EINVAL
}
- fd := vp.newFD(mnt, vfsd, statusFlags)
+ fd := vp.newFD(mnt, vfsd, statusFlags, locks)
// Named pipes have special blocking semantics during open:
//
@@ -127,10 +130,11 @@ func (vp *VFSPipe) Open(ctx context.Context, mnt *vfs.Mount, vfsd *vfs.Dentry, s
}
// Preconditions: vp.mu must be held.
-func (vp *VFSPipe) newFD(mnt *vfs.Mount, vfsd *vfs.Dentry, statusFlags uint32) *vfs.FileDescription {
+func (vp *VFSPipe) newFD(mnt *vfs.Mount, vfsd *vfs.Dentry, statusFlags uint32, locks *lock.FileLocks) *vfs.FileDescription {
fd := &VFSPipeFD{
pipe: &vp.pipe,
}
+ fd.LockFD.Init(locks)
fd.vfsfd.Init(fd, statusFlags, mnt, vfsd, &vfs.FileDescriptionOptions{
DenyPRead: true,
DenyPWrite: true,
@@ -159,6 +163,7 @@ type VFSPipeFD struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
vfs.DentryMetadataFileDescriptionImpl
+ vfs.LockFD
pipe *Pipe
}
diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go
index 00977fc08..165869028 100644
--- a/pkg/sentry/loader/vdso.go
+++ b/pkg/sentry/loader/vdso.go
@@ -75,7 +75,7 @@ var _ fs.FileOperations = (*byteReader)(nil)
// newByteReaderFile creates a fake file to read data from.
//
-// TODO(gvisor.dev/issue/1623): Convert to VFS2.
+// TODO(gvisor.dev/issue/2921): Convert to VFS2.
func newByteReaderFile(ctx context.Context, data []byte) *fs.File {
// Create a fake inode.
inode := fs.NewInode(
diff --git a/pkg/sentry/pgalloc/pgalloc.go b/pkg/sentry/pgalloc/pgalloc.go
index c8d9facc2..46f19d218 100644
--- a/pkg/sentry/pgalloc/pgalloc.go
+++ b/pkg/sentry/pgalloc/pgalloc.go
@@ -393,16 +393,17 @@ func (f *MemoryFile) Allocate(length uint64, kind usage.MemoryKind) (platform.Fi
return platform.FileRange{}, syserror.ENOMEM
}
- // Expand the file if needed. Note that findAvailableRange will
- // appropriately double the fileSize when required.
+ // Expand the file if needed.
if int64(fr.End) > f.fileSize {
- if err := f.file.Truncate(int64(fr.End)); err != nil {
+ // Round the new file size up to be chunk-aligned.
+ newFileSize := (int64(fr.End) + chunkMask) &^ chunkMask
+ if err := f.file.Truncate(newFileSize); err != nil {
return platform.FileRange{}, err
}
- f.fileSize = int64(fr.End)
+ f.fileSize = newFileSize
f.mappingsMu.Lock()
oldMappings := f.mappings.Load().([]uintptr)
- newMappings := make([]uintptr, f.fileSize>>chunkShift)
+ newMappings := make([]uintptr, newFileSize>>chunkShift)
copy(newMappings, oldMappings)
f.mappings.Store(newMappings)
f.mappingsMu.Unlock()
diff --git a/pkg/sentry/platform/kvm/machine_arm64.go b/pkg/sentry/platform/kvm/machine_arm64.go
index 750751aa3..f3bf973de 100644
--- a/pkg/sentry/platform/kvm/machine_arm64.go
+++ b/pkg/sentry/platform/kvm/machine_arm64.go
@@ -106,7 +106,7 @@ func (m *machine) dropPageTables(pt *pagetables.PageTables) {
defer m.mu.Unlock()
// Clear from all PCIDs.
- for _, c := range m.vCPUs {
+ for _, c := range m.vCPUsByID {
if c.PCIDs != nil {
c.PCIDs.Drop(pt)
}
diff --git a/pkg/sentry/socket/hostinet/BUILD b/pkg/sentry/socket/hostinet/BUILD
index e82d6cd1e..60c9896fc 100644
--- a/pkg/sentry/socket/hostinet/BUILD
+++ b/pkg/sentry/socket/hostinet/BUILD
@@ -34,6 +34,7 @@ go_library(
"//pkg/sentry/socket",
"//pkg/sentry/socket/control",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/syserr",
"//pkg/syserror",
"//pkg/tcpip/stack",
diff --git a/pkg/sentry/socket/hostinet/socket_vfs2.go b/pkg/sentry/socket/hostinet/socket_vfs2.go
index 677743113..027add1fd 100644
--- a/pkg/sentry/socket/hostinet/socket_vfs2.go
+++ b/pkg/sentry/socket/hostinet/socket_vfs2.go
@@ -26,6 +26,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/kernel"
"gvisor.dev/gvisor/pkg/sentry/socket"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserr"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
@@ -35,6 +36,7 @@ import (
type socketVFS2 struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
+ vfs.LockFD
// We store metadata for hostinet sockets internally. Technically, we should
// access metadata (e.g. through stat, chmod) on the host for correctness,
@@ -59,6 +61,7 @@ func newVFS2Socket(t *kernel.Task, family int, stype linux.SockType, protocol in
fd: fd,
},
}
+ s.LockFD.Init(&lock.FileLocks{})
if err := fdnotifier.AddFD(int32(fd), &s.queue); err != nil {
return nil, syserr.FromError(err)
}
diff --git a/pkg/sentry/socket/netlink/BUILD b/pkg/sentry/socket/netlink/BUILD
index 7212d8644..420e573c9 100644
--- a/pkg/sentry/socket/netlink/BUILD
+++ b/pkg/sentry/socket/netlink/BUILD
@@ -29,6 +29,7 @@ go_library(
"//pkg/sentry/socket/unix",
"//pkg/sentry/socket/unix/transport",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/sync",
"//pkg/syserr",
"//pkg/syserror",
diff --git a/pkg/sentry/socket/netlink/socket_vfs2.go b/pkg/sentry/socket/netlink/socket_vfs2.go
index b854bf990..8bfee5193 100644
--- a/pkg/sentry/socket/netlink/socket_vfs2.go
+++ b/pkg/sentry/socket/netlink/socket_vfs2.go
@@ -23,6 +23,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/socket/unix"
"gvisor.dev/gvisor/pkg/sentry/socket/unix/transport"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserr"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/tcpip"
@@ -40,6 +41,7 @@ type SocketVFS2 struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
vfs.DentryMetadataFileDescriptionImpl
+ vfs.LockFD
socketOpsCommon
}
@@ -66,7 +68,7 @@ func NewVFS2(t *kernel.Task, skType linux.SockType, protocol Protocol) (*SocketV
return nil, err
}
- return &SocketVFS2{
+ fd := &SocketVFS2{
socketOpsCommon: socketOpsCommon{
ports: t.Kernel().NetlinkPorts(),
protocol: protocol,
@@ -75,7 +77,9 @@ func NewVFS2(t *kernel.Task, skType linux.SockType, protocol Protocol) (*SocketV
connection: connection,
sendBufferSize: defaultSendBufferSize,
},
- }, nil
+ }
+ fd.LockFD.Init(&lock.FileLocks{})
+ return fd, nil
}
// Readiness implements waiter.Waitable.Readiness.
diff --git a/pkg/sentry/socket/netstack/BUILD b/pkg/sentry/socket/netstack/BUILD
index 8f0f5466e..0f592ecc3 100644
--- a/pkg/sentry/socket/netstack/BUILD
+++ b/pkg/sentry/socket/netstack/BUILD
@@ -37,6 +37,7 @@ go_library(
"//pkg/sentry/socket/netfilter",
"//pkg/sentry/unimpl",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/sync",
"//pkg/syserr",
"//pkg/syserror",
diff --git a/pkg/sentry/socket/netstack/netstack.go b/pkg/sentry/socket/netstack/netstack.go
index e1e0c5931..738277391 100644
--- a/pkg/sentry/socket/netstack/netstack.go
+++ b/pkg/sentry/socket/netstack/netstack.go
@@ -1246,6 +1246,18 @@ func getSockOptTCP(t *kernel.Task, ep commonEndpoint, name, outLen int) (interfa
return int32(time.Duration(v) / time.Second), nil
+ case linux.TCP_KEEPCNT:
+ if outLen < sizeOfInt32 {
+ return nil, syserr.ErrInvalidArgument
+ }
+
+ v, err := ep.GetSockOptInt(tcpip.KeepaliveCountOption)
+ if err != nil {
+ return nil, syserr.TranslateNetstackError(err)
+ }
+
+ return int32(v), nil
+
case linux.TCP_USER_TIMEOUT:
if outLen < sizeOfInt32 {
return nil, syserr.ErrInvalidArgument
@@ -1786,6 +1798,17 @@ func setSockOptTCP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *
}
return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.KeepaliveIntervalOption(time.Second * time.Duration(v))))
+ case linux.TCP_KEEPCNT:
+ if len(optVal) < sizeOfInt32 {
+ return syserr.ErrInvalidArgument
+ }
+
+ v := usermem.ByteOrder.Uint32(optVal)
+ if v < 1 || v > linux.MAX_TCP_KEEPCNT {
+ return syserr.ErrInvalidArgument
+ }
+ return syserr.TranslateNetstackError(ep.SetSockOptInt(tcpip.KeepaliveCountOption, int(v)))
+
case linux.TCP_USER_TIMEOUT:
if len(optVal) < sizeOfInt32 {
return syserr.ErrInvalidArgument
@@ -2115,30 +2138,20 @@ func emitUnimplementedEventTCP(t *kernel.Task, name int) {
switch name {
case linux.TCP_CONGESTION,
linux.TCP_CORK,
- linux.TCP_DEFER_ACCEPT,
linux.TCP_FASTOPEN,
linux.TCP_FASTOPEN_CONNECT,
linux.TCP_FASTOPEN_KEY,
linux.TCP_FASTOPEN_NO_COOKIE,
- linux.TCP_KEEPCNT,
- linux.TCP_KEEPIDLE,
- linux.TCP_KEEPINTVL,
- linux.TCP_LINGER2,
- linux.TCP_MAXSEG,
linux.TCP_QUEUE_SEQ,
- linux.TCP_QUICKACK,
linux.TCP_REPAIR,
linux.TCP_REPAIR_QUEUE,
linux.TCP_REPAIR_WINDOW,
linux.TCP_SAVED_SYN,
linux.TCP_SAVE_SYN,
- linux.TCP_SYNCNT,
linux.TCP_THIN_DUPACK,
linux.TCP_THIN_LINEAR_TIMEOUTS,
linux.TCP_TIMESTAMP,
- linux.TCP_ULP,
- linux.TCP_USER_TIMEOUT,
- linux.TCP_WINDOW_CLAMP:
+ linux.TCP_ULP:
t.Kernel().EmitUnimplementedEvent(t)
}
diff --git a/pkg/sentry/socket/netstack/netstack_vfs2.go b/pkg/sentry/socket/netstack/netstack_vfs2.go
index fcd8013c0..1412a4810 100644
--- a/pkg/sentry/socket/netstack/netstack_vfs2.go
+++ b/pkg/sentry/socket/netstack/netstack_vfs2.go
@@ -25,6 +25,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/socket"
"gvisor.dev/gvisor/pkg/sentry/socket/netfilter"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserr"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/tcpip"
@@ -38,6 +39,7 @@ type SocketVFS2 struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
vfs.DentryMetadataFileDescriptionImpl
+ vfs.LockFD
socketOpsCommon
}
@@ -64,6 +66,7 @@ func NewVFS2(t *kernel.Task, family int, skType linux.SockType, protocol int, qu
protocol: protocol,
},
}
+ s.LockFD.Init(&lock.FileLocks{})
vfsfd := &s.vfsfd
if err := vfsfd.Init(s, linux.O_RDWR, mnt, d, &vfs.FileDescriptionOptions{
DenyPRead: true,
diff --git a/pkg/sentry/socket/unix/BUILD b/pkg/sentry/socket/unix/BUILD
index de2cc4bdf..7d4cc80fe 100644
--- a/pkg/sentry/socket/unix/BUILD
+++ b/pkg/sentry/socket/unix/BUILD
@@ -29,6 +29,7 @@ go_library(
"//pkg/sentry/socket/netstack",
"//pkg/sentry/socket/unix/transport",
"//pkg/sentry/vfs",
+ "//pkg/sentry/vfs/lock",
"//pkg/syserr",
"//pkg/syserror",
"//pkg/tcpip",
diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go
index c4c9db81b..4bb2b6ff4 100644
--- a/pkg/sentry/socket/unix/unix.go
+++ b/pkg/sentry/socket/unix/unix.go
@@ -459,15 +459,25 @@ func (s *socketOpsCommon) SendMsg(t *kernel.Task, src usermem.IOSequence, to []b
To: nil,
}
if len(to) > 0 {
- ep, err := extractEndpoint(t, to)
- if err != nil {
- return 0, err
- }
- defer ep.Release()
- w.To = ep
+ switch s.stype {
+ case linux.SOCK_SEQPACKET:
+ to = nil
+ case linux.SOCK_STREAM:
+ if s.State() == linux.SS_CONNECTED {
+ return 0, syserr.ErrAlreadyConnected
+ }
+ return 0, syserr.ErrNotSupported
+ default:
+ ep, err := extractEndpoint(t, to)
+ if err != nil {
+ return 0, err
+ }
+ defer ep.Release()
+ w.To = ep
- if ep.Passcred() && w.Control.Credentials == nil {
- w.Control.Credentials = control.MakeCreds(t)
+ if ep.Passcred() && w.Control.Credentials == nil {
+ w.Control.Credentials = control.MakeCreds(t)
+ }
}
}
diff --git a/pkg/sentry/socket/unix/unix_vfs2.go b/pkg/sentry/socket/unix/unix_vfs2.go
index 45e109361..8c32371a2 100644
--- a/pkg/sentry/socket/unix/unix_vfs2.go
+++ b/pkg/sentry/socket/unix/unix_vfs2.go
@@ -26,6 +26,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/socket/netstack"
"gvisor.dev/gvisor/pkg/sentry/socket/unix/transport"
"gvisor.dev/gvisor/pkg/sentry/vfs"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/syserr"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/tcpip"
@@ -39,6 +40,7 @@ type SocketVFS2 struct {
vfsfd vfs.FileDescription
vfs.FileDescriptionDefaultImpl
vfs.DentryMetadataFileDescriptionImpl
+ vfs.LockFD
socketOpsCommon
}
@@ -51,7 +53,7 @@ func NewSockfsFile(t *kernel.Task, ep transport.Endpoint, stype linux.SockType)
mnt := t.Kernel().SocketMount()
d := sockfs.NewDentry(t.Credentials(), mnt)
- fd, err := NewFileDescription(ep, stype, linux.O_RDWR, mnt, d)
+ fd, err := NewFileDescription(ep, stype, linux.O_RDWR, mnt, d, &lock.FileLocks{})
if err != nil {
return nil, syserr.FromError(err)
}
@@ -60,7 +62,7 @@ func NewSockfsFile(t *kernel.Task, ep transport.Endpoint, stype linux.SockType)
// NewFileDescription creates and returns a socket file description
// corresponding to the given mount and dentry.
-func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint32, mnt *vfs.Mount, d *vfs.Dentry) (*vfs.FileDescription, error) {
+func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint32, mnt *vfs.Mount, d *vfs.Dentry, locks *lock.FileLocks) (*vfs.FileDescription, error) {
// You can create AF_UNIX, SOCK_RAW sockets. They're the same as
// SOCK_DGRAM and don't require CAP_NET_RAW.
if stype == linux.SOCK_RAW {
@@ -73,6 +75,7 @@ func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint3
stype: stype,
},
}
+ sock.LockFD.Init(locks)
vfsfd := &sock.vfsfd
if err := vfsfd.Init(sock, flags, mnt, d, &vfs.FileDescriptionOptions{
DenyPRead: true,
diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go
index 35a98212a..8347617bd 100644
--- a/pkg/sentry/syscalls/linux/sys_file.go
+++ b/pkg/sentry/syscalls/linux/sys_file.go
@@ -998,9 +998,6 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
return 0, nil, err
}
- // The lock uid is that of the Task's FDTable.
- lockUniqueID := lock.UniqueID(t.FDTable().ID())
-
// These locks don't block; execute the non-blocking operation using the inode's lock
// context directly.
switch flock.Type {
@@ -1010,12 +1007,12 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
}
if cmd == linux.F_SETLK {
// Non-blocking lock, provide a nil lock.Blocker.
- if !file.Dirent.Inode.LockCtx.Posix.LockRegion(lockUniqueID, lock.ReadLock, rng, nil) {
+ if !file.Dirent.Inode.LockCtx.Posix.LockRegion(t.FDTable(), lock.ReadLock, rng, nil) {
return 0, nil, syserror.EAGAIN
}
} else {
// Blocking lock, pass in the task to satisfy the lock.Blocker interface.
- if !file.Dirent.Inode.LockCtx.Posix.LockRegion(lockUniqueID, lock.ReadLock, rng, t) {
+ if !file.Dirent.Inode.LockCtx.Posix.LockRegion(t.FDTable(), lock.ReadLock, rng, t) {
return 0, nil, syserror.EINTR
}
}
@@ -1026,18 +1023,18 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
}
if cmd == linux.F_SETLK {
// Non-blocking lock, provide a nil lock.Blocker.
- if !file.Dirent.Inode.LockCtx.Posix.LockRegion(lockUniqueID, lock.WriteLock, rng, nil) {
+ if !file.Dirent.Inode.LockCtx.Posix.LockRegion(t.FDTable(), lock.WriteLock, rng, nil) {
return 0, nil, syserror.EAGAIN
}
} else {
// Blocking lock, pass in the task to satisfy the lock.Blocker interface.
- if !file.Dirent.Inode.LockCtx.Posix.LockRegion(lockUniqueID, lock.WriteLock, rng, t) {
+ if !file.Dirent.Inode.LockCtx.Posix.LockRegion(t.FDTable(), lock.WriteLock, rng, t) {
return 0, nil, syserror.EINTR
}
}
return 0, nil, nil
case linux.F_UNLCK:
- file.Dirent.Inode.LockCtx.Posix.UnlockRegion(lockUniqueID, rng)
+ file.Dirent.Inode.LockCtx.Posix.UnlockRegion(t.FDTable(), rng)
return 0, nil, nil
default:
return 0, nil, syserror.EINVAL
@@ -2157,22 +2154,6 @@ func Flock(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
nonblocking := operation&linux.LOCK_NB != 0
operation &^= linux.LOCK_NB
- // flock(2):
- // Locks created by flock() are associated with an open file table entry. This means that
- // duplicate file descriptors (created by, for example, fork(2) or dup(2)) refer to the
- // same lock, and this lock may be modified or released using any of these descriptors. Furthermore,
- // the lock is released either by an explicit LOCK_UN operation on any of these duplicate
- // descriptors, or when all such descriptors have been closed.
- //
- // If a process uses open(2) (or similar) to obtain more than one descriptor for the same file,
- // these descriptors are treated independently by flock(). An attempt to lock the file using
- // one of these file descriptors may be denied by a lock that the calling process has already placed via
- // another descriptor.
- //
- // We use the File UniqueID as the lock UniqueID because it needs to reference the same lock across dup(2)
- // and fork(2).
- lockUniqueID := lock.UniqueID(file.UniqueID)
-
// A BSD style lock spans the entire file.
rng := lock.LockRange{
Start: 0,
@@ -2183,29 +2164,29 @@ func Flock(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
case linux.LOCK_EX:
if nonblocking {
// Since we're nonblocking we pass a nil lock.Blocker implementation.
- if !file.Dirent.Inode.LockCtx.BSD.LockRegion(lockUniqueID, lock.WriteLock, rng, nil) {
+ if !file.Dirent.Inode.LockCtx.BSD.LockRegion(file, lock.WriteLock, rng, nil) {
return 0, nil, syserror.EWOULDBLOCK
}
} else {
// Because we're blocking we will pass the task to satisfy the lock.Blocker interface.
- if !file.Dirent.Inode.LockCtx.BSD.LockRegion(lockUniqueID, lock.WriteLock, rng, t) {
+ if !file.Dirent.Inode.LockCtx.BSD.LockRegion(file, lock.WriteLock, rng, t) {
return 0, nil, syserror.EINTR
}
}
case linux.LOCK_SH:
if nonblocking {
// Since we're nonblocking we pass a nil lock.Blocker implementation.
- if !file.Dirent.Inode.LockCtx.BSD.LockRegion(lockUniqueID, lock.ReadLock, rng, nil) {
+ if !file.Dirent.Inode.LockCtx.BSD.LockRegion(file, lock.ReadLock, rng, nil) {
return 0, nil, syserror.EWOULDBLOCK
}
} else {
// Because we're blocking we will pass the task to satisfy the lock.Blocker interface.
- if !file.Dirent.Inode.LockCtx.BSD.LockRegion(lockUniqueID, lock.ReadLock, rng, t) {
+ if !file.Dirent.Inode.LockCtx.BSD.LockRegion(file, lock.ReadLock, rng, t) {
return 0, nil, syserror.EINTR
}
}
case linux.LOCK_UN:
- file.Dirent.Inode.LockCtx.BSD.UnlockRegion(lockUniqueID, rng)
+ file.Dirent.Inode.LockCtx.BSD.UnlockRegion(file, rng)
default:
// flock(2): EINVAL operation is invalid.
return 0, nil, syserror.EINVAL
diff --git a/pkg/sentry/syscalls/linux/vfs2/BUILD b/pkg/sentry/syscalls/linux/vfs2/BUILD
index c0d005247..9f93f4354 100644
--- a/pkg/sentry/syscalls/linux/vfs2/BUILD
+++ b/pkg/sentry/syscalls/linux/vfs2/BUILD
@@ -14,6 +14,7 @@ go_library(
"getdents.go",
"inotify.go",
"ioctl.go",
+ "lock.go",
"memfd.go",
"mmap.go",
"mount.go",
@@ -42,6 +43,7 @@ go_library(
"//pkg/fspath",
"//pkg/gohacks",
"//pkg/sentry/arch",
+ "//pkg/sentry/fs/lock",
"//pkg/sentry/fsbridge",
"//pkg/sentry/fsimpl/eventfd",
"//pkg/sentry/fsimpl/pipefs",
diff --git a/pkg/sentry/syscalls/linux/vfs2/fd.go b/pkg/sentry/syscalls/linux/vfs2/fd.go
index ca0f7fd1e..6006758a5 100644
--- a/pkg/sentry/syscalls/linux/vfs2/fd.go
+++ b/pkg/sentry/syscalls/linux/vfs2/fd.go
@@ -168,7 +168,7 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall
err := tmpfs.AddSeals(file, args[2].Uint())
return 0, nil, err
default:
- // TODO(gvisor.dev/issue/1623): Everything else is not yet supported.
+ // TODO(gvisor.dev/issue/2920): Everything else is not yet supported.
return 0, nil, syserror.EINVAL
}
}
diff --git a/pkg/sentry/syscalls/linux/vfs2/lock.go b/pkg/sentry/syscalls/linux/vfs2/lock.go
new file mode 100644
index 000000000..bf19028c4
--- /dev/null
+++ b/pkg/sentry/syscalls/linux/vfs2/lock.go
@@ -0,0 +1,64 @@
+// Copyright 2020 The 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 vfs2
+
+import (
+ "gvisor.dev/gvisor/pkg/abi/linux"
+ "gvisor.dev/gvisor/pkg/sentry/arch"
+ "gvisor.dev/gvisor/pkg/sentry/fs/lock"
+ "gvisor.dev/gvisor/pkg/sentry/kernel"
+ "gvisor.dev/gvisor/pkg/syserror"
+)
+
+// Flock implements linux syscall flock(2).
+func Flock(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) {
+ fd := args[0].Int()
+ operation := args[1].Int()
+
+ file := t.GetFileVFS2(fd)
+ if file == nil {
+ // flock(2): EBADF fd is not an open file descriptor.
+ return 0, nil, syserror.EBADF
+ }
+ defer file.DecRef()
+
+ nonblocking := operation&linux.LOCK_NB != 0
+ operation &^= linux.LOCK_NB
+
+ var blocker lock.Blocker
+ if !nonblocking {
+ blocker = t
+ }
+
+ switch operation {
+ case linux.LOCK_EX:
+ if err := file.LockBSD(t, lock.WriteLock, blocker); err != nil {
+ return 0, nil, err
+ }
+ case linux.LOCK_SH:
+ if err := file.LockBSD(t, lock.ReadLock, blocker); err != nil {
+ return 0, nil, err
+ }
+ case linux.LOCK_UN:
+ if err := file.UnlockBSD(t); err != nil {
+ return 0, nil, err
+ }
+ default:
+ // flock(2): EINVAL operation is invalid.
+ return 0, nil, syserror.EINVAL
+ }
+
+ return 0, nil, nil
+}
diff --git a/pkg/sentry/syscalls/linux/vfs2/vfs2.go b/pkg/sentry/syscalls/linux/vfs2/vfs2.go
index 7b6e7571a..954c82f97 100644
--- a/pkg/sentry/syscalls/linux/vfs2/vfs2.go
+++ b/pkg/sentry/syscalls/linux/vfs2/vfs2.go
@@ -62,7 +62,7 @@ func Override() {
s.Table[55] = syscalls.Supported("getsockopt", GetSockOpt)
s.Table[59] = syscalls.Supported("execve", Execve)
s.Table[72] = syscalls.Supported("fcntl", Fcntl)
- delete(s.Table, 73) // flock
+ s.Table[73] = syscalls.Supported("fcntl", Flock)
s.Table[74] = syscalls.Supported("fsync", Fsync)
s.Table[75] = syscalls.Supported("fdatasync", Fdatasync)
s.Table[76] = syscalls.Supported("truncate", Truncate)
diff --git a/pkg/sentry/time/muldiv_arm64.s b/pkg/sentry/time/muldiv_arm64.s
index 5ad57a8a3..8afc62d53 100644
--- a/pkg/sentry/time/muldiv_arm64.s
+++ b/pkg/sentry/time/muldiv_arm64.s
@@ -12,12 +12,15 @@
// See the License for the specific language governing permissions and
// limitations under the License.
+#include "funcdata.h"
#include "textflag.h"
// Documentation is available in parameters.go.
//
// func muldiv64(value, multiplier, divisor uint64) (uint64, bool)
TEXT ·muldiv64(SB),NOSPLIT,$40-33
+ GO_ARGS
+ NO_LOCAL_POINTERS
MOVD value+0(FP), R0
MOVD multiplier+8(FP), R1
MOVD divisor+16(FP), R2
diff --git a/pkg/sentry/time/parameters_test.go b/pkg/sentry/time/parameters_test.go
index e1b9084ac..0ce1257f6 100644
--- a/pkg/sentry/time/parameters_test.go
+++ b/pkg/sentry/time/parameters_test.go
@@ -484,3 +484,18 @@ func TestMulDivOverflow(t *testing.T) {
})
}
}
+
+func BenchmarkMuldiv64(b *testing.B) {
+ var v uint64 = math.MaxUint64
+ for i := uint64(1); i <= 1000000; i++ {
+ mult := uint64(1000000000)
+ div := i * mult
+ res, ok := muldiv64(v, mult, div)
+ if !ok {
+ b.Errorf("Result of %v * %v / %v ok got false want true", v, mult, div)
+ }
+ if want := v / i; res != want {
+ b.Errorf("Bad result of %v * %v / %v: got %v, want %v", v, mult, div, res, want)
+ }
+ }
+}
diff --git a/pkg/sentry/vfs/BUILD b/pkg/sentry/vfs/BUILD
index 774cc66cc..16d9f3a28 100644
--- a/pkg/sentry/vfs/BUILD
+++ b/pkg/sentry/vfs/BUILD
@@ -72,6 +72,7 @@ go_library(
"//pkg/sentry/memmap",
"//pkg/sentry/socket/unix/transport",
"//pkg/sentry/uniqueid",
+ "//pkg/sentry/vfs/lock",
"//pkg/sync",
"//pkg/syserror",
"//pkg/usermem",
diff --git a/pkg/sentry/vfs/epoll.go b/pkg/sentry/vfs/epoll.go
index 8297f964b..599c3131c 100644
--- a/pkg/sentry/vfs/epoll.go
+++ b/pkg/sentry/vfs/epoll.go
@@ -31,6 +31,7 @@ type EpollInstance struct {
vfsfd FileDescription
FileDescriptionDefaultImpl
DentryMetadataFileDescriptionImpl
+ NoLockFD
// q holds waiters on this EpollInstance.
q waiter.Queue
diff --git a/pkg/sentry/vfs/file_description.go b/pkg/sentry/vfs/file_description.go
index bb294563d..97b9b18d7 100644
--- a/pkg/sentry/vfs/file_description.go
+++ b/pkg/sentry/vfs/file_description.go
@@ -73,6 +73,8 @@ type FileDescription struct {
// writable is analogous to Linux's FMODE_WRITE.
writable bool
+ usedLockBSD uint32
+
// impl is the FileDescriptionImpl associated with this Filesystem. impl is
// immutable. This should be the last field in FileDescription.
impl FileDescriptionImpl
@@ -175,6 +177,12 @@ func (fd *FileDescription) DecRef() {
}
ep.interestMu.Unlock()
}
+
+ // If BSD locks were used, release any lock that it may have acquired.
+ if atomic.LoadUint32(&fd.usedLockBSD) != 0 {
+ fd.impl.UnlockBSD(context.Background(), fd)
+ }
+
// Release implementation resources.
fd.impl.Release()
if fd.writable {
@@ -420,13 +428,9 @@ type FileDescriptionImpl interface {
Removexattr(ctx context.Context, name string) error
// LockBSD tries to acquire a BSD-style advisory file lock.
- //
- // TODO(gvisor.dev/issue/1480): BSD-style file locking
LockBSD(ctx context.Context, uid lock.UniqueID, t lock.LockType, block lock.Blocker) error
- // LockBSD releases a BSD-style advisory file lock.
- //
- // TODO(gvisor.dev/issue/1480): BSD-style file locking
+ // UnlockBSD releases a BSD-style advisory file lock.
UnlockBSD(ctx context.Context, uid lock.UniqueID) error
// LockPOSIX tries to acquire a POSIX-style advisory file lock.
@@ -736,3 +740,14 @@ func (fd *FileDescription) InodeID() uint64 {
func (fd *FileDescription) Msync(ctx context.Context, mr memmap.MappableRange) error {
return fd.Sync(ctx)
}
+
+// LockBSD tries to acquire a BSD-style advisory file lock.
+func (fd *FileDescription) LockBSD(ctx context.Context, lockType lock.LockType, blocker lock.Blocker) error {
+ atomic.StoreUint32(&fd.usedLockBSD, 1)
+ return fd.impl.LockBSD(ctx, fd, lockType, blocker)
+}
+
+// UnlockBSD releases a BSD-style advisory file lock.
+func (fd *FileDescription) UnlockBSD(ctx context.Context) error {
+ return fd.impl.UnlockBSD(ctx, fd)
+}
diff --git a/pkg/sentry/vfs/file_description_impl_util.go b/pkg/sentry/vfs/file_description_impl_util.go
index f4c111926..af7213dfd 100644
--- a/pkg/sentry/vfs/file_description_impl_util.go
+++ b/pkg/sentry/vfs/file_description_impl_util.go
@@ -21,8 +21,9 @@ import (
"gvisor.dev/gvisor/pkg/abi/linux"
"gvisor.dev/gvisor/pkg/context"
"gvisor.dev/gvisor/pkg/sentry/arch"
- "gvisor.dev/gvisor/pkg/sentry/fs/lock"
+ fslock "gvisor.dev/gvisor/pkg/sentry/fs/lock"
"gvisor.dev/gvisor/pkg/sentry/memmap"
+ "gvisor.dev/gvisor/pkg/sentry/vfs/lock"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
@@ -153,26 +154,6 @@ func (FileDescriptionDefaultImpl) Removexattr(ctx context.Context, name string)
return syserror.ENOTSUP
}
-// LockBSD implements FileDescriptionImpl.LockBSD.
-func (FileDescriptionDefaultImpl) LockBSD(ctx context.Context, uid lock.UniqueID, t lock.LockType, block lock.Blocker) error {
- return syserror.EBADF
-}
-
-// UnlockBSD implements FileDescriptionImpl.UnlockBSD.
-func (FileDescriptionDefaultImpl) UnlockBSD(ctx context.Context, uid lock.UniqueID) error {
- return syserror.EBADF
-}
-
-// LockPOSIX implements FileDescriptionImpl.LockPOSIX.
-func (FileDescriptionDefaultImpl) LockPOSIX(ctx context.Context, uid lock.UniqueID, t lock.LockType, rng lock.LockRange, block lock.Blocker) error {
- return syserror.EBADF
-}
-
-// UnlockPOSIX implements FileDescriptionImpl.UnlockPOSIX.
-func (FileDescriptionDefaultImpl) UnlockPOSIX(ctx context.Context, uid lock.UniqueID, rng lock.LockRange) error {
- return syserror.EBADF
-}
-
// DirectoryFileDescriptionDefaultImpl may be embedded by implementations of
// FileDescriptionImpl that always represent directories to obtain
// implementations of non-directory I/O methods that return EISDIR.
@@ -384,3 +365,60 @@ func GenericConfigureMMap(fd *FileDescription, m memmap.Mappable, opts *memmap.M
fd.IncRef()
return nil
}
+
+// LockFD may be used by most implementations of FileDescriptionImpl.Lock*
+// functions. Caller must call Init().
+type LockFD struct {
+ locks *lock.FileLocks
+}
+
+// Init initializes fd with FileLocks to use.
+func (fd *LockFD) Init(locks *lock.FileLocks) {
+ fd.locks = locks
+}
+
+// LockBSD implements vfs.FileDescriptionImpl.LockBSD.
+func (fd *LockFD) LockBSD(ctx context.Context, uid fslock.UniqueID, t fslock.LockType, block fslock.Blocker) error {
+ return fd.locks.LockBSD(uid, t, block)
+}
+
+// UnlockBSD implements vfs.FileDescriptionImpl.UnlockBSD.
+func (fd *LockFD) UnlockBSD(ctx context.Context, uid fslock.UniqueID) error {
+ fd.locks.UnlockBSD(uid)
+ return nil
+}
+
+// LockPOSIX implements vfs.FileDescriptionImpl.LockPOSIX.
+func (fd *LockFD) LockPOSIX(ctx context.Context, uid fslock.UniqueID, t fslock.LockType, rng fslock.LockRange, block fslock.Blocker) error {
+ return fd.locks.LockPOSIX(uid, t, rng, block)
+}
+
+// UnlockPOSIX implements vfs.FileDescriptionImpl.UnlockPOSIX.
+func (fd *LockFD) UnlockPOSIX(ctx context.Context, uid fslock.UniqueID, rng fslock.LockRange) error {
+ fd.locks.UnlockPOSIX(uid, rng)
+ return nil
+}
+
+// NoLockFD implements Lock*/Unlock* portion of FileDescriptionImpl interface
+// returning ENOLCK.
+type NoLockFD struct{}
+
+// LockBSD implements vfs.FileDescriptionImpl.LockBSD.
+func (NoLockFD) LockBSD(ctx context.Context, uid fslock.UniqueID, t fslock.LockType, block fslock.Blocker) error {
+ return syserror.ENOLCK
+}
+
+// UnlockBSD implements vfs.FileDescriptionImpl.UnlockBSD.
+func (NoLockFD) UnlockBSD(ctx context.Context, uid fslock.UniqueID) error {
+ return syserror.ENOLCK
+}
+
+// LockPOSIX implements vfs.FileDescriptionImpl.LockPOSIX.
+func (NoLockFD) LockPOSIX(ctx context.Context, uid fslock.UniqueID, t fslock.LockType, rng fslock.LockRange, block fslock.Blocker) error {
+ return syserror.ENOLCK
+}
+
+// UnlockPOSIX implements vfs.FileDescriptionImpl.UnlockPOSIX.
+func (NoLockFD) UnlockPOSIX(ctx context.Context, uid fslock.UniqueID, rng fslock.LockRange) error {
+ return syserror.ENOLCK
+}
diff --git a/pkg/sentry/vfs/file_description_impl_util_test.go b/pkg/sentry/vfs/file_description_impl_util_test.go
index 3a75d4d62..5061f6ac9 100644
--- a/pkg/sentry/vfs/file_description_impl_util_test.go
+++ b/pkg/sentry/vfs/file_description_impl_util_test.go
@@ -33,6 +33,7 @@ import (
type fileDescription struct {
vfsfd FileDescription
FileDescriptionDefaultImpl
+ NoLockFD
}
// genCount contains the number of times its DynamicBytesSource.Generate()
diff --git a/pkg/sentry/vfs/inotify.go b/pkg/sentry/vfs/inotify.go
index 05a3051a4..7fa7d2d0c 100644
--- a/pkg/sentry/vfs/inotify.go
+++ b/pkg/sentry/vfs/inotify.go
@@ -57,6 +57,7 @@ type Inotify struct {
vfsfd FileDescription
FileDescriptionDefaultImpl
DentryMetadataFileDescriptionImpl
+ NoLockFD
// Unique identifier for this inotify instance. We don't just reuse the
// inotify fd because fds can be duped. These should not be exposed to the