Re: [patch 1/5] Fix timeout bug in rtmutex in 2.6.18-rt

[Esben Nielsen <nielsen.esben@googlemail.com>]

On Fri, 10 Nov 2006, Ingo Molnar wrote:

> on SMP i had to remove this assert:
>
> Index: linux/kernel/rtmutex.c
> ===================================================================
> --- linux.orig/kernel/rtmutex.c
> +++ linux/kernel/rtmutex.c
> @@ -823,7 +823,7 @@ rt_spin_lock_slowunlock(struct rt_mutex
> 	}
> 	wakeup_next_waiter(lock, 1);
> 	spin_unlock_irqrestore(&lock->wait_lock, flags);
> -	BUG_ON(current->boosting_prio != MAX_PRIO);
> +//	BUG_ON(current->boosting_prio != MAX_PRIO);
> 	/* Undo pi boosting.when necessary */
> 	rt_mutex_adjust_prio(current);
> }
>
> because it triggered almost immediately after bootup.


Hmm, boosting_prio should be MAX_PRIO at that point. boosting_prio should
only be different from MAX_PRIO while a task is trying to lock a mutex - 
i.e. while it is in a *_slowlock() function.

On which kernel have you applied the patch?


I am looking at 2.6.18-rt6 and trying to see if there is a race conditions.
I can't see it that easily (or I would probably have seen it already :-). 
task->boosting_prio is not set when task->pi_lock is not taken. And then 
it is only set to something else than MAX_PRIO if task->pi_blocked_on is 
set.


> 	Ingo

Esben
-
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/