futex: Fix barrier comment
authorDavidlohr Bueso <dave@stgolabs.net>
Wed, 6 Feb 2019 18:56:02 +0000 (10:56 -0800)
committerThomas Gleixner <tglx@linutronix.de>
Fri, 8 Feb 2019 12:00:35 +0000 (13:00 +0100)
The current comment for the barrier that guarantees that waiter increment
is always before taking the hb spinlock (barrier (A)) needs to be fixed as
it is misplaced.

This is obviously referring to hb_waiters_inc, which is a full barrier.

Reported-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190206185602.949-1-dave@stgolabs.net
kernel/futex.c

index fdd312d..5ec2473 100644 (file)
@@ -2221,11 +2221,11 @@ static inline struct futex_hash_bucket *queue_lock(struct futex_q *q)
         * decrement the counter at queue_unlock() when some error has
         * occurred and we don't end up adding the task to the list.
         */
-       hb_waiters_inc(hb);
+       hb_waiters_inc(hb); /* implies smp_mb(); (A) */
 
        q->lock_ptr = &hb->lock;
 
-       spin_lock(&hb->lock); /* implies smp_mb(); (A) */
+       spin_lock(&hb->lock);
        return hb;
 }