Re: [PATCH 3/6] workqueue: use @pool instead of @gcwq or @cpu where applicable

[Tejun Heo <tj@xxxxxxxxxx>]

Hello, Tony.

On Tue, Jul 10, 2012 at 04:30:36PM -0700, Tony Luck wrote:
> On Mon, Jul 9, 2012 at 11:41 AM, Tejun Heo <tj@xxxxxxxxxx> wrote:
> > @@ -1234,7 +1235,7 @@ static void worker_enter_idle(struct worker *worker)
> >          */
> >         WARN_ON_ONCE(gcwq->trustee_state == TRUSTEE_DONE &&
> >                      pool->nr_workers == pool->nr_idle &&
> > -                    atomic_read(get_gcwq_nr_running(gcwq->cpu)));
> > +                    atomic_read(get_pool_nr_running(pool)));
> >  }
> 
> Just had this WARN_ON_ONCE trigger on ia64 booting next-20120710. I
> haven't bisected ... just noticed  that two patches in this series tinker
> with lines in this check. next-20120706 didn't generate the WARN.

Sorry about the delay.  The warning is spurious.  As now there are
multiple pools, nr_running check should be done before
pool->nr_workers check.  Will post fix soon.

Thank you.

-- 
tejun