Re: [PATCH] secure unlock_task_sighand() call

[Oleg Nesterov <oleg@redhat.com>]

On 12/21, Linus Torvalds wrote:
>
> We have coredump serialization in exit_mm() that I think *should* make
> this all ok - if we still see p->mm matching our mm,

Yes. And the comment says:

	lock_task_sighand(p)
	must be used. Since p->mm != NULL and we hold ->mmap_sem
	it can't fail.

IOW, this task can't pass exit_mm() and thus lock_task_sighand() can't
fail.

> >                 do {
> >                         if (p->mm) {
> >                                 if (unlikely(p->mm == mm)) {
> > -                                       lock_task_sighand(p, &flags);
> > -                                       nr += zap_process(p, exit_code);
> > -                                       unlock_task_sighand(p, &flags);
> > +                                       if (lock_task_sighand(p, &flags) {
> > +                                               nr += zap_process(p, exit_code);
> > +                                               unlock_task_sighand(p, &flags);
> > +                                       }

I too do not think this is needed. But perhaps BUG_ON() make sense.

Note: just in case, this has another problem: while_each_thread() is racy.
We already have the initial fixes in -mm, this code (as other users)
should be converted to use for_each_thread(), I'll send the patch(es).

Oleg.