7 files changed, 45 insertions, 23 deletions
diff --git a/pkg/sentry/kernel/auth/credentials.go b/pkg/sentry/kernel/auth/credentials.go
index e057d2c6d..6862f2ef5 100644
--- a/pkg/sentry/kernel/auth/credentials.go
+++ b/pkg/sentry/kernel/auth/credentials.go
@@ -232,3 +232,31 @@ func (c *Credentials) UseGID(gid GID) (KGID, error) {
 	}
 	return NoID, syserror.EPERM
 }
+
+// SetUID translates the provided uid to the root user namespace and updates c's
+// uids to it. This performs no permissions or capabilities checks, the caller
+// is responsible for ensuring the calling context is permitted to modify c.
+func (c *Credentials) SetUID(uid UID) error {
+	kuid := c.UserNamespace.MapToKUID(uid)
+	if !kuid.Ok() {
+		return syserror.EINVAL
+	}
+	c.RealKUID = kuid
+	c.EffectiveKUID = kuid
+	c.SavedKUID = kuid
+	return nil
+}
+
+// SetGID translates the provided gid to the root user namespace and updates c's
+// gids to it. This performs no permissions or capabilities checks, the caller
+// is responsible for ensuring the calling context is permitted to modify c.
+func (c *Credentials) SetGID(gid GID) error {
+	kgid := c.UserNamespace.MapToKGID(gid)
+	if !kgid.Ok() {
+		return syserror.EINVAL
+	}
+	c.RealKGID = kgid
+	c.EffectiveKGID = kgid
+	c.SavedKGID = kgid
+	return nil
+}
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/kernel/task_exec.go b/pkg/sentry/kernel/task_exec.go
index 00c425cca..9b69f3cbe 100644
--- a/pkg/sentry/kernel/task_exec.go
+++ b/pkg/sentry/kernel/task_exec.go
@@ -198,6 +198,10 @@ func (r *runSyscallAfterExecStop) execute(t *Task) taskRunState {
 	t.tg.oldRSeqCritical.Store(&OldRSeqCriticalRegion{})
 	t.tg.pidns.owner.mu.Unlock()
 
+	oldFDTable := t.fdTable
+	t.fdTable = t.fdTable.Fork()
+	oldFDTable.DecRef()
+
 	// Remove FDs with the CloseOnExec flag set.
 	t.fdTable.RemoveIf(func(_ *fs.File, _ *vfs.FileDescription, flags FDFlags) bool {
 		return flags.CloseOnExec
diff --git a/pkg/sentry/kernel/timekeeper.go b/pkg/sentry/kernel/timekeeper.go
index da0ea7bb5..0adf25691 100644
--- a/pkg/sentry/kernel/timekeeper.go
+++ b/pkg/sentry/kernel/timekeeper.go
@@ -186,6 +186,7 @@ func (t *Timekeeper) startUpdater() {
 	timer := time.NewTicker(sentrytime.ApproxUpdateInterval)
 	t.wg.Add(1)
 	go func() { // S/R-SAFE: stopped during save.
+		defer t.wg.Done()
 		for {
 			// Start with an update immediately, so the clocks are
 			// ready ASAP.
@@ -220,7 +221,6 @@ func (t *Timekeeper) startUpdater() {
 			select {
 			case <-timer.C:
 			case <-t.stop:
-				t.wg.Done()
 				return
 			}
 		}