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

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

On Tue, Nov 11, 2008 at 06:20:50PM +0100, Oleg Nesterov wrote:
> On 11/11, Dmitry Torokhov wrote:
> >
> > On Fri, Nov 07, 2008 at 04:43:25PM +0100, Oleg Nesterov wrote:
> > >
> > > 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.
> 
> Yes, queue_delayed_work() acts as a barrier for the work->func(), otherwise
> almost any code which uses wqs is broken.
> 
> But let me repeat, if queue_delayed_work() fails becuase this work is
> already queued we (in this particular case) need mb(), not wmb(). Or
> atkbd_schedule_event_work() can miss a bit in ->event_mask. So I think
> this wmb() is misleading.

Could you please explain why wmb() is not enough and full mb() is
needed in this case? I thought that if write happens before we decide
whether to schedule event_work or not it would be enough.

> And unneeded because queue_work() implies mb(),
> but this is not really documented.
> 

It would be great if we can get it documented and then i'd drop *mb()
from atkbd.

Thanks.

-- 
Dmitry