Re: [PATCH 1/3] wait-simple: Introduce the simple waitqueue implementation

[Peter Zijlstra <peterz@infradead.org>]

On Thu, Dec 12, 2013 at 12:18:09PM +0100, Peter Zijlstra wrote:
> On Wed, Dec 11, 2013 at 08:06:37PM -0500, Paul Gortmaker wrote:
> > +unsigned int
> > +__swake_up_locked(struct swait_queue_head *head, unsigned int state,
> > +		  unsigned int num)
> > +{
> > +	struct swaiter *curr, *next;
> > +	int woken = 0;
> > +
> > +	list_for_each_entry_safe(curr, next, &head->task_list, node) {
> > +		if (wake_up_state(curr->task, state)) {
> > +			__swait_dequeue(curr);
> > +			/*
> > +			 * The waiting task can free the waiter as
> > +			 * soon as curr->task = NULL is written,
> > +			 * without taking any locks. A memory barrier
> > +			 * is required here to prevent the following
> > +			 * store to curr->task from getting ahead of
> > +			 * the dequeue operation.
> > +			 */
> > +			smp_wmb();
> > +			curr->task = NULL;
> > +			if (++woken == num)
> > +				break;
> > +		}
> > +	}
> > +	return woken;
> > +}
> > +
> > +unsigned int
> > +__swake_up(struct swait_queue_head *head, unsigned int state, unsigned int num)
> > +{
> > +	unsigned long flags;
> > +	int woken;
> > +
> > +	if (!swaitqueue_active(head))
> > +		return 0;
> > +
> > +	raw_spin_lock_irqsave(&head->lock, flags);
> > +	woken = __swake_up_locked(head, state, num);
> > +	raw_spin_unlock_irqrestore(&head->lock, flags);
> > +	return woken;
> > +}
> > +EXPORT_SYMBOL(__swake_up);
> 
> Urgh, fail. Do not put unbounded loops in raw_spin_lock.
> 
> I think I posted a patch a while back to cure this.

tada!

lkml.kernel.org/r/20131004145625.GN3081@twins.programming.kicks-ass.net