Re: [PATCH v2] locking/semaphore: Use wake_q to wake up processes outside lock critical section

[Waiman Long]

On 9/20/23 20:05, John Stultz wrote:
> On Fri, Sep 9, 2022 at 12:28 PM Waiman Long <longman@xxxxxxxxxx> wrote:
> > It was found that a circular lock dependency can happen with the
> > following locking sequence:
> >
> >     +--> (console_sem).lock --> &p->pi_lock --> &rq->__lock --+
> >     |                                                         |
> >     +---------------------------------------------------------+
> >
> > The &p->pi_lock --> &rq->__lock sequence is very common in all the
> > task_rq_lock() calls.
> Thanks for sending this out! I've been hitting these lockdep warningns
> a lot recently, particularly if I have any debug printks/WARN_ONs in
> the scheduler that trip, so I'm eager to get a fix for this!
>
> That said, running with your patch, I'm seeing bootup hang pretty
> close to where init starts when I've had a fair amount of debug
> printks go off. It's odd because the lockup detectors don't seem to
> fire.
> I'll try to debug further, but wanted to give you a heads up. Let me
> know if you have any suggestions.

Thanks for testing this patch. This was not merged because Peter thought 
the merging of atomic console would probably make this not an issue at 
all. We are close to getting the atomic console merged. Let's see if 
this is really the case.

Cheers,
Longman