On Mon, Aug 04, 2014 at 12:30:26PM +0200, Peter Zijlstra wrote:

> +long wait_woken(wait_queue_t *wait, unsigned mode, long timeout)
> +{
> +	set_current_state(mode);
> +	if (!(wait->flags & WQ_FLAG_WOKEN))
> +		timeout = schedule_timeout(timeout);
> +	else
> +		wait->flags &= ~WQ_FLAG_WOKEN;

I just noticed that poll_schedule_timeout() uses set_mb() for clearing
its triggered variable. But I'm not entirely sure I see why..

> +	__set_current_state(TASK_RUNNING);
> +
> +	return timeout;
> +}
> +EXPORT_SYMBOL(wait_woken);
> +
> +int woken_wake_function(wait_queue_t *wait, unsigned mode, int sync, void *key)
> +{
> +	/*
> +	 * Although this function is called under waitqueue lock, LOCK
> +	 * doesn't imply write barrier and the users expect write
> +	 * barrier semantics on wakeup functions.  The following
> +	 * smp_wmb() is equivalent to smp_wmb() in try_to_wake_up()
> +	 * and is paired with set_mb() in wait_woken().
> +	 */
> +	smp_wmb();
> +	wait->flags |= WQ_FLAG_WOKEN;
> +
> +	return default_wake_function(wait, mode, sync, key);
> +}
> +EXPORT_SYMBOL(woken_wake_function);

So possibly we could also use this for poll_schedule_timeout() and
__pollwake().