Fix deadlock in procfs

Kernfs provides an internal mechanism to defer calls to `DecRef()` because
on the last reference `Filesystem.mu` must be held and most places that
need to call `DecRef()` are inside the lock. The same can be true for
filesystems that extend kernfs. procfs needs to look up files and `DecRef()`
them inside the `kernfs.Filesystem.mu`. If the files happen to be procfs
files, it can deadlock trying to decrement if it's the last reference.
This change extends the mechanism to external callers to defer DecRefs
to `vfs.FileDescription` and `vfs.VirtualDentries`.

PiperOrigin-RevId: 384361647

1 files changed, 4 insertions, 4 deletions

diff --git a/pkg/sentry/kernel/abstract_socket_namespace.go b/pkg/sentry/kernel/abstract_socket_namespace.go
index d100e58d7..5d86a04f3 100644
--- a/pkg/sentry/kernel/abstract_socket_namespace.go
+++ b/pkg/sentry/kernel/abstract_socket_namespace.go
@@ -27,7 +27,7 @@ import (
 // +stateify savable
 type abstractEndpoint struct {
 	ep     transport.BoundEndpoint
-	socket refsvfs2.RefCounter
+	socket refsvfs2.TryRefCounter
 	name   string
 	ns     *AbstractSocketNamespace
 }
@@ -57,7 +57,7 @@ func NewAbstractSocketNamespace() *AbstractSocketNamespace {
 // its backing socket.
 type boundEndpoint struct {
 	transport.BoundEndpoint
-	socket refsvfs2.RefCounter
+	socket refsvfs2.TryRefCounter
 }
 
 // Release implements transport.BoundEndpoint.Release.
@@ -89,7 +89,7 @@ func (a *AbstractSocketNamespace) BoundEndpoint(name string) transport.BoundEndp
 //
 // When the last reference managed by socket is dropped, ep may be removed from the
 // namespace.
-func (a *AbstractSocketNamespace) Bind(ctx context.Context, name string, ep transport.BoundEndpoint, socket refsvfs2.RefCounter) error {
+func (a *AbstractSocketNamespace) Bind(ctx context.Context, name string, ep transport.BoundEndpoint, socket refsvfs2.TryRefCounter) error {
 	a.mu.Lock()
 	defer a.mu.Unlock()
 
@@ -109,7 +109,7 @@ func (a *AbstractSocketNamespace) Bind(ctx context.Context, name string, ep tran
 
 // Remove removes the specified socket at name from the abstract socket
 // namespace, if it has not yet been replaced.
-func (a *AbstractSocketNamespace) Remove(name string, socket refsvfs2.RefCounter) {
+func (a *AbstractSocketNamespace) Remove(name string, socket refsvfs2.TryRefCounter) {
 	a.mu.Lock()
 	defer a.mu.Unlock()