Plumb VFS2 inside the Sentry

- Added fsbridge package with interface that can be used to open
  and read from VFS1 and VFS2 files.
- Converted ELF loader to use fsbridge
- Added VFS2 types to FSContext
- Added vfs.MountNamespace to ThreadGroup

Updates #1623

PiperOrigin-RevId: 295183950

1 files changed, 27 insertions, 0 deletions

diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go
index 981e8c7fe..a3443ff21 100644
--- a/pkg/sentry/kernel/task.go
+++ b/pkg/sentry/kernel/task.go
@@ -424,6 +424,11 @@ type Task struct {
 	// abstractSockets is protected by mu.
 	abstractSockets *AbstractSocketNamespace
 
+	// mountNamespaceVFS2 is the task's mount namespace.
+	//
+	// It is protected by mu. It is owned by the task goroutine.
+	mountNamespaceVFS2 *vfs.MountNamespace
+
 	// parentDeathSignal is sent to this task's thread group when its parent exits.
 	//
 	// parentDeathSignal is protected by mu.
@@ -638,6 +643,11 @@ func (t *Task) Value(key interface{}) interface{} {
 		return int32(t.ThreadGroup().ID())
 	case fs.CtxRoot:
 		return t.fsContext.RootDirectory()
+	case vfs.CtxRoot:
+		return t.fsContext.RootDirectoryVFS2()
+	case vfs.CtxMountNamespace:
+		t.mountNamespaceVFS2.IncRef()
+		return t.mountNamespaceVFS2
 	case fs.CtxDirentCacheLimiter:
 		return t.k.DirentCacheLimiter
 	case inet.CtxStack:
@@ -701,6 +711,14 @@ func (t *Task) SyscallRestartBlock() SyscallRestartBlock {
 // Preconditions: The caller must be running on the task goroutine, or t.mu
 // must be locked.
 func (t *Task) IsChrooted() bool {
+	if VFS2Enabled {
+		realRoot := t.mountNamespaceVFS2.Root()
+		defer realRoot.DecRef()
+		root := t.fsContext.RootDirectoryVFS2()
+		defer root.DecRef()
+		return root != realRoot
+	}
+
 	realRoot := t.tg.mounts.Root()
 	defer realRoot.DecRef()
 	root := t.fsContext.RootDirectory()
@@ -796,6 +814,15 @@ func (t *Task) MountNamespace() *fs.MountNamespace {
 	return t.tg.mounts
 }
 
+// MountNamespaceVFS2 returns t's MountNamespace. A reference is taken on the
+// returned mount namespace.
+func (t *Task) MountNamespaceVFS2() *vfs.MountNamespace {
+	t.mu.Lock()
+	defer t.mu.Unlock()
+	t.mountNamespaceVFS2.IncRef()
+	return t.mountNamespaceVFS2
+}
+
 // AbstractSockets returns t's AbstractSocketNamespace.
 func (t *Task) AbstractSockets() *AbstractSocketNamespace {
 	return t.abstractSockets