On Wed, Jul 23, 2008 at 08:46:58PM +0200, Ivo van Doorn wrote:
> On Wednesday 23 July 2008, Henrique de Moraes Holschuh wrote:
> > Limit the number of rfkill-input global operations per second. It lacked
> > the limiter that non-global operations (state changes) had. This way, a
> > rogue input event generator cannot force rfkill-input to hog the workqueue
> > too much.
> >
> > Rework the limiter code so that newer state change requests (rfkill input
> > events) will override older ones that haven't been acted upon yet. It used
> > to ignore new ones that were past the rate limit.
This was done to deal with potential jitter from button devices.
In the proposed implementation, if button bounces and generates 2 or
more press/release pairs and we manage to schedule and execure first
task (which is scheduled with delay 0) before the last press/release
arrives we will schedule an extra work and toggle the switch again as
far as I can see.
> >
> > @@ -132,24 +147,22 @@ static void rfkill_schedule_toggle(struct rfkill_task *task)
> > {
> > unsigned long flags;
> >
> > - if (unlikely(work_pending(&rfkill_global_task.work)))
> > + if (unlikely(delayed_work_pending(&rfkill_global_task.dwork)))
What if the global task comletes executing right at this moment?
> > return;
> >
> > spin_lock_irqsave(&task->lock, flags);
> > -
> > - if (time_after(jiffies, task->last + msecs_to_jiffies(200))) {
> > - task->desired_state =
> > - rfkill_state_complement(task->desired_state);
> > + task->desired_state = rfkill_state_complement(task->desired_state);
> > + if (likely(!delayed_work_pending(&task->dwork))) {
> > + schedule_delayed_work(&task->dwork,
> > + rfkill_ratelimit(task->last));
> > task->last = jiffies;
> > - schedule_work(&task->work);
> > }
> > -
> > spin_unlock_irqrestore(&task->lock, flags);
> > }
> >
--
Dmitry
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
