Re: GPF in run_workqueue()/list_del_init(cwq->worklist.next) onresume (was: Re: Help needed: Resume problems in 2.6.32-rc, perhapsrelated to preempt_count leakage in keventd)

[Frederic Weisbecker]

On Mon, Nov 16, 2009 at 12:37:06AM +0100, Frederic Weisbecker wrote:
> On Thu, Nov 12, 2009 at 06:33:00PM +0100, Thomas Gleixner wrote:
> > @@ -145,6 +255,7 @@ static void __queue_work(struct cpu_work
> >  {
> >  	unsigned long flags;
> >  
> > +	debug_work_activate(work);
> >  	spin_lock_irqsave(&cwq->lock, flags);
> >  	insert_work(cwq, work, &cwq->worklist);
> 
> 
> 
> Since you are doing that from insert_wq_barrier too, which
> endpoint is also insert_work(), why not put debug_work_activate
> there instead? Or may be you really prefer to do this outside
> the spinlock (which in off-case is zero-overhead). May be that
> can sleep or?


/me now remembers this path can't sleep since we can queue a
work from anywhere...so I guess this is to not bloat the lock
overhead.

Whatever, this is really a small detail.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/