Re: [PATCH] atkbd: cancel delayed work before freeing its structure

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

On Fri, Nov 07, 2008 at 04:43:25PM +0100, Oleg Nesterov wrote:
> On 11/05, Jiri Pirko wrote:
> >
> > diff --git a/drivers/input/keyboard/atkbd.c b/drivers/input/keyboard/atkbd.c
> > index 22016ca..f3bbf49 100644
> > --- a/drivers/input/keyboard/atkbd.c
> > +++ b/drivers/input/keyboard/atkbd.c
> > @@ -824,7 +824,7 @@ static void atkbd_disconnect(struct serio *serio)
> >  	atkbd_disable(atkbd);
> >
> >  	/* make sure we don't have a command in flight */
> > -	flush_scheduled_work();
> > +	cancel_delayed_work_sync(&atkbd->event_work);
> 
> Ping. Dmitry, could you take a look?
> 

Applied, thank you.

> 
> While we are here, what is the reason for atkbd_schedule_event_work()->wmb() ?
> It looks absolutely bogus. Is it for atkbd_event_work() ? In that case it
> is not needed, it must see all previous STOREs because both queue_work() and
> run_workqueue() take cwq->lock. And in any case,
> test_and_set_bit(WORK_STRUCT_PENDING) implies mb().

I wanted to be sure that event_mask is set before we schedule event_work
and I don't want to rely on details of queue_delayed_work
implementation. If the fact that queue_delayed_work acts as a barrier
would be listed part of its published spec I would gladly remove wmb()
from atkbd.

> If schedule_delayed_work()
> fails we can race with the soon-to-be-executed atkbd_event_work(), in that
> case that test_and_set_bit() + test_and_clear_bit(->event_mask) save us,
> but wmb() can't help again.
> 
> Another question is why do we need ->event_mutex? OK, it can serialize
> multiple instances of atkbd_event_work() running on the different CPUs,
> but in that case atkbd_reconnect() needs this lock too? It also calls
> atkbd_set_repeat_rate/atkbd_set_leds.

Probably, I will need to thiknk about it a bit more.

-- 
Dmitry