rust_binder: clear freeze listener on node removal
Subsystems (3)
Alice Ryhl
maintainerPATCH v2 1/1 rust_binder: clear freeze listener on node removal
Generally userspace is supposed to explicitly clear freeze listeners before they drop the refcount on the node ref to zero, but there's nothing forcing that. Currently, in this scenario the freeze listener remains in the freeze_listeners rbtree and in the remote node's freeze listener list, even though the ref for which the listener is registered is gone. This could potentially lead to a memory leak due to a refcount cycle. Thus, remove the freeze listener in this scenario.
---
This series is based on top of:
https://lore.kernel.org/all/[email protected]/
---
Changes in v2:
- Rebase on v3 of parent series.
- Link to v1: https://lore.kernel.org/r/20260609-remove-freeze-on-remove-node-v1-1-f67f3b9bfeb8@google.com
--- drivers/android/binder/freeze.rs | 11 +++++++++-- drivers/android/binder/node.rs | 6 ++++-- drivers/android/binder/process.rs | 13 +++++++++++-- 3 files changed, 24 insertions(+), 6 deletions(-)
drivers/android/binder/freeze.rs +9 -2
@@ -154,10 +154,17 @@ fn debug_print(&self, m: &SeqFile, prefix: &str, _tprefix: &str) -> Result<()> { }impl FreezeListener { -pub (crate )fn on_process_exit (&self, proc: &Arc <Process >) { +/// Called when this freeze listener is cleared abnormally. +/// +/// This occurs either because the process exited or because the process dropped its last +/// refcount on the node ref without explicitly removing the freeze listener first. +/// +/// The returned `KVVec` is just a value that should be dropped outside of the lock. +pub (crate )fn on_process_cleanup (&self, proc: &Process ) ->KVVec <Arc <Process >> {if !self.is_clearing { - self.node.remove_freeze_listener (proc); +return self.node.remove_freeze_listener (proc); } +KVVec ::new () } }
drivers/android/binder/node.rs +4 -2
@@ -679,11 +679,13 @@ pub(crate) fn add_freeze_listener(Ok (Ok (())) } -pub (crate )fn remove_freeze_listener (&self, p: &Arc <Process >) ->KVVec <Arc <Process >> { +pub (crate )fn remove_freeze_listener (&self, p: &Process ) ->KVVec <Arc <Process >> {let mut guard = self.owner.inner.lock ();let inner = self.inner.access_mut (&mut guard);let len = inner.freeze_list.len (); - inner.freeze_list.retain (|proc| !Arc ::ptr_eq (proc, p)); + inner + .freeze_list + .retain (|proc| !core::ptr::eq ::<Process >(&**proc, p));if len == inner.freeze_list.len () {pr_warn! ("Could not remove freeze listener for {}\n" ,
drivers/android/binder/process.rs +11 -2
@@ -950,6 +950,8 @@ pub(crate) fn update_ref(// increment references on itself. let _to_free_by_handle;let _to_free_by_node; +let _to_free_freeze_listener; +let _to_free_freeze_listener_cleanup;let mut refs = self.node_refs.lock ();if let Some (info) = refs.by_handle.get_mut (&handle) {if info.node_ref ().update (inc, strong) {
@@ -965,8 +967,15 @@ pub(crate) fn update_ref(// SAFETY: We are removing the `NodeRefInfo` from the right node. unsafe { info.node_ref2 ().node.remove_node_info (info) }; -let id = info.node_ref ().node.global_id (); + +if let Some (freeze) = *info.freeze () { +if let Some (fl) = refs.freeze_listeners.remove (&freeze) { + _to_free_freeze_listener_cleanup = fl.on_process_cleanup (&self); + _to_free_freeze_listener = fl; + } + } + _to_free_by_handle = refs.by_handle.remove_node (&handle); _to_free_by_node = refs.by_node.remove_node (&id); refs.handle_is_present.release_id (handleas usize );
@@ -1391,7 +1400,7 @@ fn deferred_release(self: Arc<Self>) {// Clean up freeze listeners. let freeze_listeners =take (&mut self.node_refs.lock ().freeze_listeners);for listenerin freeze_listeners.values () { - listener.on_process_exit (&self); + listener.on_process_cleanup (&self); }drop (freeze_listeners);
---
base-commit: 3bc831df9ee16fceee851872315161377ca1417d
change-id: 20260609-remove-freeze-on-remove-node-30e72ff6b46e prerequisite-change-id:20260608-binder-noderefs-spin-3a0ec0589043:v3
prerequisite-patch-id: 4984d542e1da65da603c302f0bccb423867f71c2
prerequisite-patch-id: e1ae5f73b329080c0fcea1d2509c085e105f67c3
prerequisite-patch-id: 63f2092c6fd9dd54b7adf93be8d5670d8490f401
prerequisite-patch-id: 5811aeb4ca435a1d4d0bb347d4de9f3b2d91f814
prerequisite-patch-id: 556e45c697f9888c8446245bbcf478422991fddf
prerequisite-patch-id: f567ba2263108b12871013309ae7c482d500eeae
Best regards,
Signature · 1 line
Alice Ryhl <[email protected]>
Greg Kroah-Hartman
maintainerQuoted 8 lines · Alice Ryhl
Generally userspace is supposed to explicitly clear freeze listeners before they drop the refcount on the node ref to zero, but there's nothing forcing that. Currently, in this scenario the freeze listener remains in the
freeze_listenersrbtree and in the remote node's freeze listener list, even though the ref for which the listener is registered is gone. This could potentially lead to a memory leak due to a refcount cycle. Thus, remove the freeze listener in this scenario.
Quoted 6 lines
Fixes eafedbc7c050 ("rust_binder: add Rust Binder driver")Signed-off-by Alice Ryhl <[email protected]>---
This series is based on top of:
https://lore.kernel.org/all/[email protected]/
Hm, but that's not a bugfix series, so I can't take this patch now for 7.2-final. Do you want to redo this one or wait for 7.3-rc1?
thanks,
greg k-h
Alice Ryhl
maintainerQuoted 17 lines · Greg Kroah-Hartman
Quoted 8 lines · Alice Ryhl
Generally userspace is supposed to explicitly clear freeze listeners before they drop the refcount on the node ref to zero, but there's nothing forcing that. Currently, in this scenario the freeze listener remains in the
freeze_listenersrbtree and in the remote node's freeze listener list, even though the ref for which the listener is registered is gone. This could potentially lead to a memory leak due to a refcount cycle. Thus, remove the freeze listener in this scenario.Quoted 6 lines
Fixes eafedbc7c050 ("rust_binder: add Rust Binder driver")Signed-off-by Alice Ryhl <[email protected]>---
This series is based on top of:
https://lore.kernel.org/all/[email protected]/
Quoted 2 lines
Hm, but that's not a bugfix series, so I can't take this patch now for 7.2-final. Do you want to redo this one or wait for 7.3-rc1?
I can reorder them.
Alice
Greg Kroah-Hartman
maintainerQuoted 19 lines · Alice Ryhl
Quoted 17 lines · Greg Kroah-Hartman
Quoted 14 lines · Alice Ryhl
Generally userspace is supposed to explicitly clear freeze listeners before they drop the refcount on the node ref to zero, but there's nothing forcing that. Currently, in this scenario the freeze listener remains in the
freeze_listenersrbtree and in the remote node's freeze listener list, even though the ref for which the listener is registered is gone. This could potentially lead to a memory leak due to a refcount cycle. Thus, remove the freeze listener in this scenario.Fixes eafedbc7c050 ("rust_binder: add Rust Binder driver")Signed-off-by Alice Ryhl <[email protected]>---
This series is based on top of:
https://lore.kernel.org/all/[email protected]/
Quoted 4 lines
Quoted 2 lines
Hm, but that's not a bugfix series, so I can't take this patch now for 7.2-final. Do you want to redo this one or wait for 7.3-rc1?
I can reorder them.
Great, thanks!