Misc VFS2 fixes

- Fix defer operation ordering in kernfs.Filesystem.AccessAt()
- Add AT_NULL entry in proc/pid/auvx
- Fix line padding in /proc/pid/maps
- Fix linux_dirent serialization for getdents(2)
- Remove file creation flags from vfs.FileDescription.statusFlags()

Updates #1193, #1035

PiperOrigin-RevId: 307704159

pkg/sentry/fs/proc/task.go

@@ -73,8 +73,7 @@ func checkTaskState(t *kernel.Task) error {
 type taskDir struct {
 	ramfs.Dir
 
-	t     *kernel.Task
+	t *kernel.Task
-	pidns *kernel.PIDNamespace
 }
 
 var _ fs.InodeOperations = (*taskDir)(nil)

pkg/sentry/fsimpl/kernfs/filesystem.go

@@ -246,8 +246,8 @@ func (fs *Filesystem) Sync(ctx context.Context) error {
 // AccessAt implements vfs.Filesystem.Impl.AccessAt.
 func (fs *Filesystem) AccessAt(ctx context.Context, rp *vfs.ResolvingPath, creds *auth.Credentials, ats vfs.AccessTypes) error {
 	fs.mu.RLock()
-	defer fs.mu.RUnlock()
 	defer fs.processDeferredDecRefs()
+	defer fs.mu.RUnlock()
 
 	_, inode, err := fs.walkExistingLocked(ctx, rp)
 	if err != nil {

pkg/sentry/fsimpl/proc/task_files.go

@@ -111,17 +111,18 @@ func (d *auxvData) Generate(ctx context.Context, buf *bytes.Buffer) error {
 	}
 	defer m.DecUsers(ctx)
 
-	// Space for buffer with AT_NULL (0) terminator at the end.
 	auxv := m.Auxv()
+	// Space for buffer with AT_NULL (0) terminator at the end.
 	buf.Grow((len(auxv) + 1) * 16)
 	for _, e := range auxv {
-		var tmp [8]byte
+		var tmp [16]byte
-		usermem.ByteOrder.PutUint64(tmp[:], e.Key)
+		usermem.ByteOrder.PutUint64(tmp[:8], e.Key)
-		buf.Write(tmp[:])
+		usermem.ByteOrder.PutUint64(tmp[8:], uint64(e.Value))
-
-		usermem.ByteOrder.PutUint64(tmp[:], uint64(e.Value))
 		buf.Write(tmp[:])
 	}
+	var atNull [16]byte
+	buf.Write(atNull[:])
+
 	return nil
 }
 

pkg/sentry/fsimpl/proc/tasks_sys.go

@@ -39,7 +39,7 @@ func newSysDir(root *auth.Credentials, inoGen InoGenerator, k *kernel.Kernel) *k
 			"shmmni":   newDentry(root, inoGen.NextIno(), 0444, shmData(linux.SHMMNI)),
 		}),
 		"vm": kernfs.NewStaticDir(root, inoGen.NextIno(), 0555, map[string]*kernfs.Dentry{
-			"mmap_min_addr":     newDentry(root, inoGen.NextIno(), 0444, &mmapMinAddrData{}),
+			"mmap_min_addr":     newDentry(root, inoGen.NextIno(), 0444, &mmapMinAddrData{k: k}),
 			"overcommit_memory": newDentry(root, inoGen.NextIno(), 0444, newStaticFile("0\n")),
 		}),
 		"net": newSysNetDir(root, inoGen, k),

pkg/sentry/mm/procfs.go

@@ -148,7 +148,7 @@ func (mm *MemoryManager) appendVMAMapsEntryLocked(ctx context.Context, vseg vmaI
 
 	// Do not include the guard page: fs/proc/task_mmu.c:show_map_vma() =>
 	// stack_guard_page_start().
-	fmt.Fprintf(b, "%08x-%08x %s%s %08x %02x:%02x %d ",
+	lineLen, _ := fmt.Fprintf(b, "%08x-%08x %s%s %08x %02x:%02x %d ",
 		vseg.Start(), vseg.End(), vma.realPerms, private, vma.off, devMajor, devMinor, ino)
 
 	// Figure out our filename or hint.
@@ -165,7 +165,7 @@ func (mm *MemoryManager) appendVMAMapsEntryLocked(ctx context.Context, vseg vmaI
 	}
 	if s != "" {
 		// Per linux, we pad until the 74th character.
-		if pad := 73 - b.Len(); pad > 0 {
+		if pad := 73 - lineLen; pad > 0 {
 			b.WriteString(strings.Repeat(" ", pad))
 		}
 		b.WriteString(s)

pkg/sentry/syscalls/linux/vfs2/getdents.go

@@ -130,7 +130,7 @@ func (cb *getdentsCallback) Handle(dirent vfs.Dirent) error {
 		if cb.t.Arch().Width() != 8 {
 			panic(fmt.Sprintf("unsupported sizeof(unsigned long): %d", cb.t.Arch().Width()))
 		}
-		size := 8 + 8 + 2 + 1 + 1 + 1 + len(dirent.Name)
+		size := 8 + 8 + 2 + 1 + 1 + len(dirent.Name)
 		size = (size + 7) &^ 7 // round up to multiple of sizeof(long)
 		if size > cb.remaining {
 			return syserror.EINVAL
@@ -143,11 +143,11 @@ func (cb *getdentsCallback) Handle(dirent vfs.Dirent) error {
 		// Zero out all remaining bytes in buf, including the NUL terminator
 		// after dirent.Name and the zero padding byte between the name and
 		// dirent type.
-		bufTail := buf[18+len(dirent.Name):]
+		bufTail := buf[18+len(dirent.Name) : size-1]
 		for i := range bufTail {
 			bufTail[i] = 0
 		}
-		bufTail[2] = dirent.Type
+		buf[size-1] = dirent.Type
 	}
 	n, err := cb.t.CopyOutBytes(cb.addr, buf)
 	if err != nil {

pkg/sentry/vfs/file_description.go

@@ -111,10 +111,10 @@ type FileDescriptionOptions struct {
 }
 
 // Init must be called before first use of fd. If it succeeds, it takes
-// references on mnt and d. statusFlags is the initial file description status
+// references on mnt and d. flags is the initial file description flags, which
-// flags, which is usually the full set of flags passed to open(2).
+// is usually the full set of flags passed to open(2).
-func (fd *FileDescription) Init(impl FileDescriptionImpl, statusFlags uint32, mnt *Mount, d *Dentry, opts *FileDescriptionOptions) error {
+func (fd *FileDescription) Init(impl FileDescriptionImpl, flags uint32, mnt *Mount, d *Dentry, opts *FileDescriptionOptions) error {
-	writable := MayWriteFileWithOpenFlags(statusFlags)
+	writable := MayWriteFileWithOpenFlags(flags)
 	if writable {
 		if err := mnt.CheckBeginWrite(); err != nil {
 			return err
@@ -122,7 +122,10 @@ func (fd *FileDescription) Init(impl FileDescriptionImpl, statusFlags uint32, mn
 	}
 
 	fd.refs = 1
-	fd.statusFlags = statusFlags
+
+	// Remove "file creation flags" to mirror the behavior from file.f_flags in
+	// fs/open.c:do_dentry_open
+	fd.statusFlags = flags &^ (linux.O_CREAT | linux.O_EXCL | linux.O_NOCTTY | linux.O_TRUNC)
 	fd.vd = VirtualDentry{
 		mount:  mnt,
 		dentry: d,
@@ -130,7 +133,7 @@ func (fd *FileDescription) Init(impl FileDescriptionImpl, statusFlags uint32, mn
 	mnt.IncRef()
 	d.IncRef()
 	fd.opts = *opts
-	fd.readable = MayReadFileWithOpenFlags(statusFlags)
+	fd.readable = MayReadFileWithOpenFlags(flags)
 	fd.writable = writable
 	fd.impl = impl
 	return nil