Re: [PATCH] Fix for Bug in PI exit code

[Ingo Molnar <mingo@elte.hu>]

* Dinakar Guniguntala <dino@in.ibm.com> wrote:

> We were seeing oopses like below a lot when using PI mutexes

> ===============================================================================
> After a lot of debugging we found that this is caused due to the following race.
> PM is a PI mutex, A and B are RT threads
> 
>         Thread A (RT)                  Thread B (RT)
>             |
>             v
>     pthread_mutex_lock (PM)                 |
>     (glibc) got mutex                       v
>          do work                   pthread_mutex_lock (PM)
>                                    rt_mutex_timed_lock
> 
>           EINTR                    EINTR (Process gets aborted)
> 
>          do_exit                   lock(pi_mutex->lock->wait_lock)
>     exit_pi_state_list             clear_waiters
>     lock(hb->lock)
>     pi_state->owner = NULL         unlock(pi_mutex->lock->wait_lock)
>     rt_mutex_unlock(pi_mutex)      lock(hb->lock) (blocks)
>     unlock(hb->lock)               unblock -> free_pi_state
>     continue exit processing       doesn't expect pi_state->owner to be NULL
>                                    Panic
> 
> The patch attached below seems to make this problem go away. This has 
> been stress tested quite a bit in the past 24 hours. Does it look sane 
> to you ??

yeah, makes sense. Thanks, applied.

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