futex: Add memory barrier commentary to futex_wait_queue_me()
authorDarren Hart <dvhltc@us.ibm.com>
Thu, 24 Sep 2009 18:54:47 +0000 (11:54 -0700)
committerIngo Molnar <mingo@elte.hu>
Thu, 24 Sep 2009 20:30:10 +0000 (22:30 +0200)
The memory barrier semantics of futex_wait_queue_me() are
non-obvious. Add some commentary to try and clarify it.

Signed-off-by: Darren Hart <dvhltc@us.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Dinakar Guniguntala <dino@in.ibm.com>
Cc: John Stultz <johnstul@us.ibm.com>
LKML-Reference: <20090924185447.694.38948.stgit@Aeon>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
kernel/futex.c

index 463af2e..b911adc 100644 (file)
@@ -1656,6 +1656,12 @@ out:
 static void futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q *q,
                                struct hrtimer_sleeper *timeout)
 {
+       /*
+        * The task state is guaranteed to be set before another task can
+        * wake it. set_current_state() is implemented using set_mb() and
+        * queue_me() calls spin_unlock() upon completion, both serializing
+        * access to the hash list and forcing another memory barrier.
+        */
        set_current_state(TASK_INTERRUPTIBLE);
        queue_me(q, hb);