Re: [PATCH] fix-flush_workqueue-vs-cpu_dead-race-update

[Oleg Nesterov]

On 01/06, Srivatsa Vaddagiri wrote:
>
> On Sat, Jan 06, 2007 at 06:10:36PM +0300, Oleg Nesterov wrote:
> > Increment hotplug_sequence earlier, under CPU_DOWN_PREPARE. We can't
> > miss the event, the task running flush_workqueue() will be re-scheduled
> > at least once before CPU actually disappears from cpu_online_map.
> 
> Eww ..what happens if flush_workqueue() starts after CPU_DOWN_PREPARE?
                                                 ^^^^^
Stupid me. Thanks.

> CPU_DOWN_PREPARE(8)
> 	hotplug_sequence++ = 10
> 
> 					flush_workqueue()
> 						sequence = 10
> 						flush cpus 1 ....7
> 
> CPU_DEAD(8)
> 	take_over_work(8->1)
> 
> 					return not flushing dead cpu8 (=BUG)

I'll try to do something else tomorrow. Do you see a simple soulution?

The current usage of workqueue_mutex (I mean stable kernel) is broken
and deadlockable. We really need to change it.

Oleg.

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