Re: Linux 3.6-rc4

[Peter Zijlstra <peterz@infradead.org>]

On Fri, 2012-09-07 at 11:39 -0700, Linus Torvalds wrote:
> Al? Please look into this. I'm not entirely sure what's going on, but
> lockdep complains about this:
> 
>  Possible interrupt unsafe locking scenario:
> 
>        CPU0                    CPU1
>        ----                    ----
>   lock(&(&p->alloc_lock)->rlock);
>                                local_irq_disable();
>                                lock(&(&new_timer->it_lock)->rlock);
>                                lock(tasklist_lock);
>   <Interrupt>
>     lock(&(&new_timer->it_lock)->rlock);
> 
>  *** DEADLOCK ***
> 
> and it looks real. IOW, if I read that right, we have the task_lock ->
> it_lock dependency through exit_itimers(), and then we have the
> tasklist_lock -> task_lock dependency everywhere else. So now it_lock
> -> tasklist_lock becomes a deadlock. 

Agreed, I've got the following series from Oleg queued to solve this:

 http://marc.info/?l=linux-kernel&m=134600821828491&w=2