summaryrefslogtreecommitdiffhomepage
path: root/pkg/sentry/fs
diff options
context:
space:
mode:
Diffstat (limited to 'pkg/sentry/fs')
-rw-r--r--pkg/sentry/fs/dirent_cache.go2
-rw-r--r--pkg/sentry/fs/host/util.go12
-rw-r--r--pkg/sentry/fs/inotify.go5
-rw-r--r--pkg/sentry/fs/proc/task.go126
4 files changed, 120 insertions, 25 deletions
diff --git a/pkg/sentry/fs/dirent_cache.go b/pkg/sentry/fs/dirent_cache.go
index 25514ace4..33de32c69 100644
--- a/pkg/sentry/fs/dirent_cache.go
+++ b/pkg/sentry/fs/dirent_cache.go
@@ -101,8 +101,6 @@ func (c *DirentCache) remove(d *Dirent) {
panic(fmt.Sprintf("trying to remove %v, which is not in the dirent cache", d))
}
c.list.Remove(d)
- d.SetPrev(nil)
- d.SetNext(nil)
d.DecRef()
c.currentSize--
if c.limit != nil {
diff --git a/pkg/sentry/fs/host/util.go b/pkg/sentry/fs/host/util.go
index e37e687c6..7c60dc1db 100644
--- a/pkg/sentry/fs/host/util.go
+++ b/pkg/sentry/fs/host/util.go
@@ -24,7 +24,7 @@ import (
"gvisor.dev/gvisor/pkg/sentry/device"
"gvisor.dev/gvisor/pkg/sentry/fs"
"gvisor.dev/gvisor/pkg/sentry/kernel/auth"
- ktime "gvisor.dev/gvisor/pkg/sentry/kernel/time"
+ "gvisor.dev/gvisor/pkg/sentry/kernel/time"
"gvisor.dev/gvisor/pkg/syserror"
)
@@ -152,9 +152,9 @@ func unstableAttr(mo *superOperations, s *syscall.Stat_t) fs.UnstableAttr {
Usage: s.Blocks * 512,
Perms: fs.FilePermsFromMode(linux.FileMode(s.Mode)),
Owner: owner(mo, s),
- AccessTime: ktime.FromUnix(s.Atim.Sec, s.Atim.Nsec),
- ModificationTime: ktime.FromUnix(s.Mtim.Sec, s.Mtim.Nsec),
- StatusChangeTime: ktime.FromUnix(s.Ctim.Sec, s.Ctim.Nsec),
+ AccessTime: time.FromUnix(s.Atim.Sec, s.Atim.Nsec),
+ ModificationTime: time.FromUnix(s.Mtim.Sec, s.Mtim.Nsec),
+ StatusChangeTime: time.FromUnix(s.Ctim.Sec, s.Ctim.Nsec),
Links: uint64(s.Nlink),
}
}
@@ -165,6 +165,8 @@ type dirInfo struct {
bufp int // location of next record in buf.
}
+// LINT.IfChange
+
// isBlockError unwraps os errors and checks if they are caused by EAGAIN or
// EWOULDBLOCK. This is so they can be transformed into syserror.ErrWouldBlock.
func isBlockError(err error) bool {
@@ -177,6 +179,8 @@ func isBlockError(err error) bool {
return false
}
+// LINT.ThenChange(../../fsimpl/host/util.go)
+
func hostEffectiveKIDs() (uint32, []uint32, error) {
gids, err := os.Getgroups()
if err != nil {
diff --git a/pkg/sentry/fs/inotify.go b/pkg/sentry/fs/inotify.go
index 928c90aa0..e3a715c1f 100644
--- a/pkg/sentry/fs/inotify.go
+++ b/pkg/sentry/fs/inotify.go
@@ -143,7 +143,10 @@ func (i *Inotify) Read(ctx context.Context, _ *File, dst usermem.IOSequence, _ i
}
var writeLen int64
- for event := i.events.Front(); event != nil; event = event.Next() {
+ for it := i.events.Front(); it != nil; {
+ event := it
+ it = it.Next()
+
// Does the buffer have enough remaining space to hold the event we're
// about to write out?
if dst.NumBytes() < int64(event.sizeOf()) {
diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go
index 8ab8d8a02..4e9b0fc00 100644
--- a/pkg/sentry/fs/proc/task.go
+++ b/pkg/sentry/fs/proc/task.go
@@ -72,24 +72,26 @@ var _ fs.InodeOperations = (*taskDir)(nil)
// newTaskDir creates a new proc task entry.
func (p *proc) newTaskDir(t *kernel.Task, msrc *fs.MountSource, isThreadGroup bool) *fs.Inode {
contents := map[string]*fs.Inode{
- "auxv": newAuxvec(t, msrc),
- "cmdline": newExecArgInode(t, msrc, cmdlineExecArg),
- "comm": newComm(t, msrc),
- "environ": newExecArgInode(t, msrc, environExecArg),
- "exe": newExe(t, msrc),
- "fd": newFdDir(t, msrc),
- "fdinfo": newFdInfoDir(t, msrc),
- "gid_map": newGIDMap(t, msrc),
- "io": newIO(t, msrc, isThreadGroup),
- "maps": newMaps(t, msrc),
- "mountinfo": seqfile.NewSeqFileInode(t, &mountInfoFile{t: t}, msrc),
- "mounts": seqfile.NewSeqFileInode(t, &mountsFile{t: t}, msrc),
- "ns": newNamespaceDir(t, msrc),
- "smaps": newSmaps(t, msrc),
- "stat": newTaskStat(t, msrc, isThreadGroup, p.pidns),
- "statm": newStatm(t, msrc),
- "status": newStatus(t, msrc, p.pidns),
- "uid_map": newUIDMap(t, msrc),
+ "auxv": newAuxvec(t, msrc),
+ "cmdline": newExecArgInode(t, msrc, cmdlineExecArg),
+ "comm": newComm(t, msrc),
+ "environ": newExecArgInode(t, msrc, environExecArg),
+ "exe": newExe(t, msrc),
+ "fd": newFdDir(t, msrc),
+ "fdinfo": newFdInfoDir(t, msrc),
+ "gid_map": newGIDMap(t, msrc),
+ "io": newIO(t, msrc, isThreadGroup),
+ "maps": newMaps(t, msrc),
+ "mountinfo": seqfile.NewSeqFileInode(t, &mountInfoFile{t: t}, msrc),
+ "mounts": seqfile.NewSeqFileInode(t, &mountsFile{t: t}, msrc),
+ "ns": newNamespaceDir(t, msrc),
+ "oom_score": newOOMScore(t, msrc),
+ "oom_score_adj": newOOMScoreAdj(t, msrc),
+ "smaps": newSmaps(t, msrc),
+ "stat": newTaskStat(t, msrc, isThreadGroup, p.pidns),
+ "statm": newStatm(t, msrc),
+ "status": newStatus(t, msrc, p.pidns),
+ "uid_map": newUIDMap(t, msrc),
}
if isThreadGroup {
contents["task"] = p.newSubtasks(t, msrc)
@@ -796,4 +798,92 @@ func (f *auxvecFile) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequenc
return int64(n), err
}
+// newOOMScore returns a oom_score file. It is a stub that always returns 0.
+// TODO(gvisor.dev/issue/1967)
+func newOOMScore(t *kernel.Task, msrc *fs.MountSource) *fs.Inode {
+ return newStaticProcInode(t, msrc, []byte("0\n"))
+}
+
+// oomScoreAdj is a file containing the oom_score adjustment for a task.
+//
+// +stateify savable
+type oomScoreAdj struct {
+ fsutil.SimpleFileInode
+
+ t *kernel.Task
+}
+
+// +stateify savable
+type oomScoreAdjFile struct {
+ fsutil.FileGenericSeek `state:"nosave"`
+ fsutil.FileNoIoctl `state:"nosave"`
+ fsutil.FileNoMMap `state:"nosave"`
+ fsutil.FileNoSplice `state:"nosave"`
+ fsutil.FileNoopFlush `state:"nosave"`
+ fsutil.FileNoopFsync `state:"nosave"`
+ fsutil.FileNoopRelease `state:"nosave"`
+ fsutil.FileNotDirReaddir `state:"nosave"`
+ fsutil.FileUseInodeUnstableAttr `state:"nosave"`
+ waiter.AlwaysReady `state:"nosave"`
+
+ t *kernel.Task
+}
+
+// newOOMScoreAdj returns a oom_score_adj file.
+func newOOMScoreAdj(t *kernel.Task, msrc *fs.MountSource) *fs.Inode {
+ i := &oomScoreAdj{
+ SimpleFileInode: *fsutil.NewSimpleFileInode(t, fs.RootOwner, fs.FilePermsFromMode(0644), linux.PROC_SUPER_MAGIC),
+ t: t,
+ }
+ return newProcInode(t, i, msrc, fs.SpecialFile, t)
+}
+
+// Truncate implements fs.InodeOperations.Truncate. Truncate is called when
+// O_TRUNC is specified for any kind of existing Dirent but is not called via
+// (f)truncate for proc files.
+func (*oomScoreAdj) Truncate(context.Context, *fs.Inode, int64) error {
+ return nil
+}
+
+// GetFile implements fs.InodeOperations.GetFile.
+func (o *oomScoreAdj) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) {
+ return fs.NewFile(ctx, dirent, flags, &oomScoreAdjFile{t: o.t}), nil
+}
+
+// Read implements fs.FileOperations.Read.
+func (f *oomScoreAdjFile) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, offset int64) (int64, error) {
+ if offset != 0 {
+ return 0, io.EOF
+ }
+ adj, err := f.t.OOMScoreAdj()
+ if err != nil {
+ return 0, err
+ }
+ adjBytes := []byte(strconv.FormatInt(int64(adj), 10) + "\n")
+ n, err := dst.CopyOut(ctx, adjBytes)
+ return int64(n), err
+}
+
+// Write implements fs.FileOperations.Write.
+func (f *oomScoreAdjFile) Write(ctx context.Context, _ *fs.File, src usermem.IOSequence, offset int64) (int64, error) {
+ if src.NumBytes() == 0 {
+ return 0, nil
+ }
+
+ // Limit input size so as not to impact performance if input size is large.
+ src = src.TakeFirst(usermem.PageSize - 1)
+
+ var v int32
+ n, err := usermem.CopyInt32StringInVec(ctx, src.IO, src.Addrs, &v, src.Opts)
+ if err != nil {
+ return 0, err
+ }
+
+ if err := f.t.SetOOMScoreAdj(v); err != nil {
+ return 0, err
+ }
+
+ return n, nil
+}
+
// LINT.ThenChange(../../fsimpl/proc/task.go|../../fsimpl/proc/task_files.go)
generated by cgit v1.2.3 (git 2.39.1) at 2025-01-25 11:32:02 +0000