Re: workqueue destruction BUG_ON

[Tejun Heo <tj@kernel.org>]

Hello,

On 08/24/2010 03:17 PM, Johannes Berg wrote:
> On Tue, 2010-08-24 at 15:07 +0200, Tejun Heo wrote:
> 
>>> [  500.874185] ------------[ cut here ]------------
>>> [  500.875212] kernel BUG at kernel/workqueue.c:2849!
>>
>> Are you sure you're running the patched kernel?  With the patch
>> applied, the BUG_ON() wouldn't be on line 2849 (on both rc1 and 2).
> 
> Yes:
> 
> void destroy_workqueue(struct workqueue_struct *wq)
> {
>         unsigned int cpu;
>    
>         wq->flags |= WQ_DYING;
>         flush_workqueue(wq);
>   
>         /*
>          * wq list is used to freeze wq, remove from list after
>          * flushing is complete in case freeze races us.
>          */
>         spin_lock(&workqueue_lock);
>         list_del(&wq->list);
>         spin_unlock(&workqueue_lock);
> 
>         /* sanity check */
>         for_each_cwq_cpu(cpu, wq) {
>                 struct cpu_workqueue_struct *cwq = get_cwq(cpu, wq);
>                 int i;
> 
>                 for (i = 0; i < WORK_NR_COLORS; i++)
>                         BUG_ON(cwq->nr_in_flight[i]);
> 2849:           BUG_ON(cwq->nr_active);
>                 BUG_ON(!list_empty(&cwq->delayed_works));
> 
> 
> Applying the patch reported some offset, but the kernel is just rc1 +
> wireless stuff.

I see, thanks for verifying.  I probably got confused about the line
number.  Hmm... weird.  I'll prep further debug patch but can you
please tell me what you did to trigger the bug?

Thanks.

-- 
tejun