binder: signal epoll threads of self-work
authorCarlos Llamas <cmllamas@google.com>
Wed, 31 Jan 2024 21:53:46 +0000 (21:53 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 31 Jan 2024 22:08:28 +0000 (14:08 -0800)
In (e)poll mode, threads often depend on I/O events to determine when
data is ready for consumption. Within binder, a thread may initiate a
command via BINDER_WRITE_READ without a read buffer and then make use
of epoll_wait() or similar to consume any responses afterwards.

It is then crucial that epoll threads are signaled via wakeup when they
queue their own work. Otherwise, they risk waiting indefinitely for an
event leaving their work unhandled. What is worse, subsequent commands
won't trigger a wakeup either as the thread has pending work.

Fixes: 457b9a6f09f0 ("Staging: android: add binder driver")
Cc: Arve Hjønnevåg <arve@android.com>
Cc: Martijn Coenen <maco@android.com>
Cc: Alice Ryhl <aliceryhl@google.com>
Cc: Steven Moreland <smoreland@google.com>
Cc: stable@vger.kernel.org # v4.19+
Signed-off-by: Carlos Llamas <cmllamas@google.com>
Link: https://lore.kernel.org/r/20240131215347.1808751-1-cmllamas@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/android/binder.c

index 8dd23b19e99731ce1ce5f5d6062366aece0dd415..eca24f41556df04ac61747e05aace9622fbcc580 100644 (file)
@@ -478,6 +478,16 @@ binder_enqueue_thread_work_ilocked(struct binder_thread *thread,
 {
        WARN_ON(!list_empty(&thread->waiting_thread_node));
        binder_enqueue_work_ilocked(work, &thread->todo);
+
+       /* (e)poll-based threads require an explicit wakeup signal when
+        * queuing their own work; they rely on these events to consume
+        * messages without I/O block. Without it, threads risk waiting
+        * indefinitely without handling the work.
+        */
+       if (thread->looper & BINDER_LOOPER_STATE_POLL &&
+           thread->pid == current->pid && !thread->process_todo)
+               wake_up_interruptible_sync(&thread->wait);
+
        thread->process_todo = true;
 }