Re: v2.6.26-rc7: BUG task_struct: Poison overwritten

[Oleg Nesterov <oleg@tv-sign.ru>]

On 06/22, Vegard Nossum wrote:
>
> I was poking around in kernel/sched.c and noticed something odd: In
> migrate_dead(), we have this code:
>
>         /*
>          * Drop lock around migration; if someone else moves it,
>          * that's OK. No task can be added to this CPU, so iteration is
>          * fine.
>          */
>         spin_unlock_irq(&rq->lock);
>         move_task_off_dead_cpu(dead_cpu, p);
>         spin_lock_irq(&rq->lock);
>
> which is fine in itself, I guess. But spin_unlock_irq() will enable
> interrupts. And move_task_off_dead_cpu() has this comment:
>
> /*
>  * Figure out where task on dead CPU should go, use force if necessary.
>  * NOTE: interrupts should be disabled by the caller
>  */
> static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p)
> {
>
> ...but here, interrupts will not be disabled. On the other hand
> __migrate_task_irq() (called by move_task_off_dead_cpu()) calls
> local_irq_disable() itself... What do you think of this? Is the
> comment wrong? Or is there a difference between "interrupts" and
> "local_irq"?

Yes, the comment is wrong, thanks.

It wasn't updated by

"do CPU_DEAD migrating under read_lock(tasklist) instead of write_lock_irq(tasklist)"
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=f7b4cddcc5aca03e80e357360c9424dfba5056c2

Previously move_task_off_dead_cpu() was called under write_lock_irq(tasklist),
and we can't take tasklist for writing without disabling irqs.

If you don't see other problems, could you make a patch to fix the comment?

Oleg.