Re: [PATCH]CPU hotplug breaks wake_up_new_task

[Nick Piggin <nickpiggin@yahoo.com.au>]

Srivatsa Vaddagiri wrote:
> On Tue, May 31, 2005 at 07:29:39AM +0000, Shaohua Li wrote:
>
> > Hi,
> > There is a race condition at wake_up_new_task at CPU hotplug case.
> > Say do_fork
> >        copy_process (which sets new forked task's current cpu, cpu_allowed)
> >                <-------- the new forked task's current cpu is offline
> >        wake_up_new_task
> > wake_up_new_task will put the forked task into a dead cpu.
>
>
> This was noticed/fixed long back. Apparently somebody has reintroduced
> the bug. The simple fix for this race is:

If you're looking at the -mm tree, then I think I must
have reintroduced the bug. It needs a comment.

> --- kernel/fork.c.org	2005-05-31 14:57:15.000000000 +0530
> +++ kernel/fork.c	2005-05-31 15:07:20.000000000 +0530
> @@ -1024,8 +1024,7 @@ static task_t *copy_process(unsigned lon
>  	 * parent's CPU). This avoids alot of nasty races.
>  	 */
>  	p->cpus_allowed = current->cpus_allowed;
> -	if (unlikely(!cpu_isset(task_cpu(p), p->cpus_allowed)))
> -		set_task_cpu(p, smp_processor_id());
> +	set_task_cpu(p, smp_processor_id());
>  
>  	/*
>  	 * Check for pending SIGKILL! The new thread should not be allowed
>
> Could you test and check if it avoids whatever problem you are seeing?

And this patch will break balance-on-fork.
How about conditionally setting task_cpu if the task's current
CPU is offline?

Thanks,
Nick

--
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com 
-
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/