rwsem: avoid taking wait_lock in rwsem_down_write_failed
authorMichel Lespinasse <walken@google.com>
Tue, 7 May 2013 13:45:56 +0000 (06:45 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 7 May 2013 14:20:16 +0000 (07:20 -0700)
In rwsem_down_write_failed(), if there are active locks after we wake up
(i.e.  the lock got stolen from us), skip taking the wait_lock and go
back to sleep immediately.

Signed-off-by: Michel Lespinasse <walken@google.com>
Reviewed-by: Peter Hurley <peter@hurleysoftware.com>
Acked-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
lib/rwsem.c

index 64c2dc0..edf3d9c 100644 (file)
@@ -214,8 +214,8 @@ struct rw_semaphore __sched *rwsem_down_write_failed(struct rw_semaphore *sem)
                sem = __rwsem_do_wake(sem, RWSEM_WAKE_READ_OWNED);
 
        /* wait until we successfully acquire the lock */
+       set_task_state(tsk, TASK_UNINTERRUPTIBLE);
        while (true) {
-               set_task_state(tsk, TASK_UNINTERRUPTIBLE);
 
                /* Try acquiring the write lock. */
                count = RWSEM_ACTIVE_WRITE_BIAS;
@@ -226,7 +226,13 @@ struct rw_semaphore __sched *rwsem_down_write_failed(struct rw_semaphore *sem)
                        break;
 
                raw_spin_unlock_irq(&sem->wait_lock);
-               schedule();
+
+               /* Block until there are no active lockers. */
+               do {
+                       schedule();
+                       set_task_state(tsk, TASK_UNINTERRUPTIBLE);
+               } while (sem->count & RWSEM_ACTIVE_MASK);
+
                raw_spin_lock_irq(&sem->wait_lock);
        }