Add plumbing for importing fds in VFS2, along with non-socket, non-TTY impl.

In VFS2, imported file descriptors are stored in a kernfs-based filesystem.
Upon calling ImportFD, the host fd can be accessed in two ways:
1. a FileDescription that can be added to the FDTable, and
2. a Dentry in the host.filesystem mount, which we will want to access through
magic symlinks in /proc/[pid]/fd/.

An implementation of the kernfs.Inode interface stores a unique host fd. This
inode can be inserted into file descriptions as well as dentries.

This change also plumbs in three FileDescriptionImpls corresponding to fds for
sockets, TTYs, and other files (only the latter is implemented here).
These implementations will mostly make corresponding syscalls to the host.
Where possible, the logic is ported over from pkg/sentry/fs/host.

Updates #1672

PiperOrigin-RevId: 299417263

5 files changed, 48 insertions, 39 deletions

diff --git a/pkg/sentry/fsimpl/kernfs/dynamic_bytes_file.go b/pkg/sentry/fsimpl/kernfs/dynamic_bytes_file.go
index d092ccb2a..1c026f4d8 100644
--- a/pkg/sentry/fsimpl/kernfs/dynamic_bytes_file.go
+++ b/pkg/sentry/fsimpl/kernfs/dynamic_bytes_file.go
@@ -122,7 +122,7 @@ func (fd *DynamicBytesFD) Release() {}
 // Stat implements vfs.FileDescriptionImpl.Stat.
 func (fd *DynamicBytesFD) Stat(ctx context.Context, opts vfs.StatOptions) (linux.Statx, error) {
 	fs := fd.vfsfd.VirtualDentry().Mount().Filesystem()
-	return fd.inode.Stat(fs), nil
+	return fd.inode.Stat(fs, opts)
 }
 
 // SetStat implements vfs.FileDescriptionImpl.SetStat.
diff --git a/pkg/sentry/fsimpl/kernfs/fd_impl_util.go b/pkg/sentry/fsimpl/kernfs/fd_impl_util.go
index 5650512e0..da821d524 100644
--- a/pkg/sentry/fsimpl/kernfs/fd_impl_util.go
+++ b/pkg/sentry/fsimpl/kernfs/fd_impl_util.go
@@ -107,9 +107,13 @@ func (fd *GenericDirectoryFD) IterDirents(ctx context.Context, cb vfs.IterDirent
 	fs.mu.Lock()
 	defer fs.mu.Unlock()
 
+	opts := vfs.StatOptions{Mask: linux.STATX_INO}
 	// Handle ".".
 	if fd.off == 0 {
-		stat := fd.inode().Stat(vfsFS)
+		stat, err := fd.inode().Stat(vfsFS, opts)
+		if err != nil {
+			return err
+		}
 		dirent := vfs.Dirent{
 			Name:    ".",
 			Type:    linux.DT_DIR,
@@ -125,7 +129,10 @@ func (fd *GenericDirectoryFD) IterDirents(ctx context.Context, cb vfs.IterDirent
 	// Handle "..".
 	if fd.off == 1 {
 		parentInode := vfsd.ParentOrSelf().Impl().(*Dentry).inode
-		stat := parentInode.Stat(vfsFS)
+		stat, err := parentInode.Stat(vfsFS, opts)
+		if err != nil {
+			return err
+		}
 		dirent := vfs.Dirent{
 			Name:    "..",
 			Type:    linux.FileMode(stat.Mode).DirentType(),
@@ -146,7 +153,10 @@ func (fd *GenericDirectoryFD) IterDirents(ctx context.Context, cb vfs.IterDirent
 	childIdx := fd.off - 2
 	for it := fd.children.nthLocked(childIdx); it != nil; it = it.Next() {
 		inode := it.Dentry.Impl().(*Dentry).inode
-		stat := inode.Stat(vfsFS)
+		stat, err := inode.Stat(vfsFS, opts)
+		if err != nil {
+			return err
+		}
 		dirent := vfs.Dirent{
 			Name:    it.Name,
 			Type:    linux.FileMode(stat.Mode).DirentType(),
@@ -190,7 +200,7 @@ func (fd *GenericDirectoryFD) Seek(ctx context.Context, offset int64, whence int
 func (fd *GenericDirectoryFD) Stat(ctx context.Context, opts vfs.StatOptions) (linux.Statx, error) {
 	fs := fd.filesystem()
 	inode := fd.inode()
-	return inode.Stat(fs), nil
+	return inode.Stat(fs, opts)
 }
 
 // SetStat implements vfs.FileDescriptionImpl.SetStat.
diff --git a/pkg/sentry/fsimpl/kernfs/filesystem.go b/pkg/sentry/fsimpl/kernfs/filesystem.go
index 292f58afd..1d7e04ad4 100644
--- a/pkg/sentry/fsimpl/kernfs/filesystem.go
+++ b/pkg/sentry/fsimpl/kernfs/filesystem.go
@@ -12,10 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-// This file implements vfs.FilesystemImpl for kernfs.
-
 package kernfs
 
+// This file implements vfs.FilesystemImpl for kernfs.
+
 import (
 	"fmt"
 
@@ -634,7 +634,7 @@ func (fs *Filesystem) StatAt(ctx context.Context, rp *vfs.ResolvingPath, opts vf
 	if err != nil {
 		return linux.Statx{}, err
 	}
-	return inode.Stat(fs.VFSFilesystem()), nil
+	return inode.Stat(fs.VFSFilesystem(), opts)
 }
 
 // StatFSAt implements vfs.FilesystemImpl.StatFSAt.
diff --git a/pkg/sentry/fsimpl/kernfs/inode_impl_util.go b/pkg/sentry/fsimpl/kernfs/inode_impl_util.go
index 099d70a16..d50018b18 100644
--- a/pkg/sentry/fsimpl/kernfs/inode_impl_util.go
+++ b/pkg/sentry/fsimpl/kernfs/inode_impl_util.go
@@ -36,20 +36,20 @@ type InodeNoopRefCount struct {
 }
 
 // IncRef implements Inode.IncRef.
-func (n *InodeNoopRefCount) IncRef() {
+func (InodeNoopRefCount) IncRef() {
 }
 
 // DecRef implements Inode.DecRef.
-func (n *InodeNoopRefCount) DecRef() {
+func (InodeNoopRefCount) DecRef() {
 }
 
 // TryIncRef implements Inode.TryIncRef.
-func (n *InodeNoopRefCount) TryIncRef() bool {
+func (InodeNoopRefCount) TryIncRef() bool {
 	return true
 }
 
 // Destroy implements Inode.Destroy.
-func (n *InodeNoopRefCount) Destroy() {
+func (InodeNoopRefCount) Destroy() {
 }
 
 // InodeDirectoryNoNewChildren partially implements the Inode interface.
@@ -58,27 +58,27 @@ func (n *InodeNoopRefCount) Destroy() {
 type InodeDirectoryNoNewChildren struct{}
 
 // NewFile implements Inode.NewFile.
-func (*InodeDirectoryNoNewChildren) NewFile(context.Context, string, vfs.OpenOptions) (*vfs.Dentry, error) {
+func (InodeDirectoryNoNewChildren) NewFile(context.Context, string, vfs.OpenOptions) (*vfs.Dentry, error) {
 	return nil, syserror.EPERM
 }
 
 // NewDir implements Inode.NewDir.
-func (*InodeDirectoryNoNewChildren) NewDir(context.Context, string, vfs.MkdirOptions) (*vfs.Dentry, error) {
+func (InodeDirectoryNoNewChildren) NewDir(context.Context, string, vfs.MkdirOptions) (*vfs.Dentry, error) {
 	return nil, syserror.EPERM
 }
 
 // NewLink implements Inode.NewLink.
-func (*InodeDirectoryNoNewChildren) NewLink(context.Context, string, Inode) (*vfs.Dentry, error) {
+func (InodeDirectoryNoNewChildren) NewLink(context.Context, string, Inode) (*vfs.Dentry, error) {
 	return nil, syserror.EPERM
 }
 
 // NewSymlink implements Inode.NewSymlink.
-func (*InodeDirectoryNoNewChildren) NewSymlink(context.Context, string, string) (*vfs.Dentry, error) {
+func (InodeDirectoryNoNewChildren) NewSymlink(context.Context, string, string) (*vfs.Dentry, error) {
 	return nil, syserror.EPERM
 }
 
 // NewNode implements Inode.NewNode.
-func (*InodeDirectoryNoNewChildren) NewNode(context.Context, string, vfs.MknodOptions) (*vfs.Dentry, error) {
+func (InodeDirectoryNoNewChildren) NewNode(context.Context, string, vfs.MknodOptions) (*vfs.Dentry, error) {
 	return nil, syserror.EPERM
 }
 
@@ -90,62 +90,62 @@ type InodeNotDirectory struct {
 }
 
 // HasChildren implements Inode.HasChildren.
-func (*InodeNotDirectory) HasChildren() bool {
+func (InodeNotDirectory) HasChildren() bool {
 	return false
 }
 
 // NewFile implements Inode.NewFile.
-func (*InodeNotDirectory) NewFile(context.Context, string, vfs.OpenOptions) (*vfs.Dentry, error) {
+func (InodeNotDirectory) NewFile(context.Context, string, vfs.OpenOptions) (*vfs.Dentry, error) {
 	panic("NewFile called on non-directory inode")
 }
 
 // NewDir implements Inode.NewDir.
-func (*InodeNotDirectory) NewDir(context.Context, string, vfs.MkdirOptions) (*vfs.Dentry, error) {
+func (InodeNotDirectory) NewDir(context.Context, string, vfs.MkdirOptions) (*vfs.Dentry, error) {
 	panic("NewDir called on non-directory inode")
 }
 
 // NewLink implements Inode.NewLinkink.
-func (*InodeNotDirectory) NewLink(context.Context, string, Inode) (*vfs.Dentry, error) {
+func (InodeNotDirectory) NewLink(context.Context, string, Inode) (*vfs.Dentry, error) {
 	panic("NewLink called on non-directory inode")
 }
 
 // NewSymlink implements Inode.NewSymlink.
-func (*InodeNotDirectory) NewSymlink(context.Context, string, string) (*vfs.Dentry, error) {
+func (InodeNotDirectory) NewSymlink(context.Context, string, string) (*vfs.Dentry, error) {
 	panic("NewSymlink called on non-directory inode")
 }
 
 // NewNode implements Inode.NewNode.
-func (*InodeNotDirectory) NewNode(context.Context, string, vfs.MknodOptions) (*vfs.Dentry, error) {
+func (InodeNotDirectory) NewNode(context.Context, string, vfs.MknodOptions) (*vfs.Dentry, error) {
 	panic("NewNode called on non-directory inode")
 }
 
 // Unlink implements Inode.Unlink.
-func (*InodeNotDirectory) Unlink(context.Context, string, *vfs.Dentry) error {
+func (InodeNotDirectory) Unlink(context.Context, string, *vfs.Dentry) error {
 	panic("Unlink called on non-directory inode")
 }
 
 // RmDir implements Inode.RmDir.
-func (*InodeNotDirectory) RmDir(context.Context, string, *vfs.Dentry) error {
+func (InodeNotDirectory) RmDir(context.Context, string, *vfs.Dentry) error {
 	panic("RmDir called on non-directory inode")
 }
 
 // Rename implements Inode.Rename.
-func (*InodeNotDirectory) Rename(context.Context, string, string, *vfs.Dentry, *vfs.Dentry) (*vfs.Dentry, error) {
+func (InodeNotDirectory) Rename(context.Context, string, string, *vfs.Dentry, *vfs.Dentry) (*vfs.Dentry, error) {
 	panic("Rename called on non-directory inode")
 }
 
 // Lookup implements Inode.Lookup.
-func (*InodeNotDirectory) Lookup(ctx context.Context, name string) (*vfs.Dentry, error) {
+func (InodeNotDirectory) Lookup(ctx context.Context, name string) (*vfs.Dentry, error) {
 	panic("Lookup called on non-directory inode")
 }
 
 // IterDirents implements Inode.IterDirents.
-func (*InodeNotDirectory) IterDirents(ctx context.Context, callback vfs.IterDirentsCallback, offset, relOffset int64) (newOffset int64, err error) {
+func (InodeNotDirectory) IterDirents(ctx context.Context, callback vfs.IterDirentsCallback, offset, relOffset int64) (newOffset int64, err error) {
 	panic("IterDirents called on non-directory inode")
 }
 
 // Valid implements Inode.Valid.
-func (*InodeNotDirectory) Valid(context.Context) bool {
+func (InodeNotDirectory) Valid(context.Context) bool {
 	return true
 }
 
@@ -157,17 +157,17 @@ func (*InodeNotDirectory) Valid(context.Context) bool {
 type InodeNoDynamicLookup struct{}
 
 // Lookup implements Inode.Lookup.
-func (*InodeNoDynamicLookup) Lookup(ctx context.Context, name string) (*vfs.Dentry, error) {
+func (InodeNoDynamicLookup) Lookup(ctx context.Context, name string) (*vfs.Dentry, error) {
 	return nil, syserror.ENOENT
 }
 
 // IterDirents implements Inode.IterDirents.
-func (*InodeNoDynamicLookup) IterDirents(ctx context.Context, callback vfs.IterDirentsCallback, offset, relOffset int64) (int64, error) {
+func (InodeNoDynamicLookup) IterDirents(ctx context.Context, callback vfs.IterDirentsCallback, offset, relOffset int64) (int64, error) {
 	return offset, nil
 }
 
 // Valid implements Inode.Valid.
-func (*InodeNoDynamicLookup) Valid(ctx context.Context) bool {
+func (InodeNoDynamicLookup) Valid(ctx context.Context) bool {
 	return true
 }
 
@@ -177,7 +177,7 @@ func (*InodeNoDynamicLookup) Valid(ctx context.Context) bool {
 type InodeNotSymlink struct{}
 
 // Readlink implements Inode.Readlink.
-func (*InodeNotSymlink) Readlink(context.Context) (string, error) {
+func (InodeNotSymlink) Readlink(context.Context) (string, error) {
 	return "", syserror.EINVAL
 }
 
@@ -219,7 +219,7 @@ func (a *InodeAttrs) Mode() linux.FileMode {
 // Stat partially implements Inode.Stat. Note that this function doesn't provide
 // all the stat fields, and the embedder should consider extending the result
 // with filesystem-specific fields.
-func (a *InodeAttrs) Stat(*vfs.Filesystem) linux.Statx {
+func (a *InodeAttrs) Stat(*vfs.Filesystem, vfs.StatOptions) (linux.Statx, error) {
 	var stat linux.Statx
 	stat.Mask = linux.STATX_TYPE | linux.STATX_MODE | linux.STATX_UID | linux.STATX_GID | linux.STATX_INO | linux.STATX_NLINK
 	stat.Ino = atomic.LoadUint64(&a.ino)
@@ -230,7 +230,7 @@ func (a *InodeAttrs) Stat(*vfs.Filesystem) linux.Statx {
 
 	// TODO: Implement other stat fields like timestamps.
 
-	return stat
+	return stat, nil
 }
 
 // SetStat implements Inode.SetStat.
diff --git a/pkg/sentry/fsimpl/kernfs/kernfs.go b/pkg/sentry/fsimpl/kernfs/kernfs.go
index c74fa999b..a8ab2a2ba 100644
--- a/pkg/sentry/fsimpl/kernfs/kernfs.go
+++ b/pkg/sentry/fsimpl/kernfs/kernfs.go
@@ -176,8 +176,6 @@ type Dentry struct {
 	vfsd  vfs.Dentry
 	inode Inode
 
-	refs uint64
-
 	// flags caches useful information about the dentry from the inode. See the
 	// dflags* consts above. Must be accessed by atomic ops.
 	flags uint32
@@ -302,7 +300,8 @@ type Inode interface {
 	// this inode. The returned file description should hold a reference on the
 	// inode for its lifetime.
 	//
-	// Precondition: !rp.Done(). vfsd.Impl() must be a kernfs Dentry.
+	// Precondition: rp.Done(). vfsd.Impl() must be the kernfs Dentry containing
+	// the inode on which Open() is being called.
 	Open(rp *vfs.ResolvingPath, vfsd *vfs.Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error)
 }
 
@@ -328,7 +327,7 @@ type inodeMetadata interface {
 
 	// Stat returns the metadata for this inode. This corresponds to
 	// vfs.FilesystemImpl.StatAt.
-	Stat(fs *vfs.Filesystem) linux.Statx
+	Stat(fs *vfs.Filesystem, opts vfs.StatOptions) (linux.Statx, error)
 
 	// SetStat updates the metadata for this inode. This corresponds to
 	// vfs.FilesystemImpl.SetStatAt.