Re: 5.8-rc*: kernel BUG at kernel/signal.c:1917

[Oleg Nesterov <oleg@redhat.com>]

Peter,

Let me add another note. TASK_TRACED/TASK_STOPPED was always protected by
->siglock. In particular, ttwu(__TASK_TRACED) must be always called with
->siglock held. That is why ptrace_freeze_traced() assumes it can safely
do s/TASK_TRACED/__TASK_TRACED/ under spin_lock(siglock).

Can this change race with

		if (signal_pending_state(prev->state, prev)) {
			prev->state = TASK_RUNNING;
		}

in __schedule() ? Hopefully not, signal-state is protected by siglock too.

So I think this logic was correct even if it doesn't look nice. But "doesn't
look nice" is true for the whole ptrace code ;)

On 07/20, Oleg Nesterov wrote:
>
> On 07/20, Jiri Slaby wrote:
> >
> > You tackled it, we cherry-picked dbfb089d360 to our kernels. Ccing more
> > people.
>
> Thanks... so with this patch __schedule() does
>
> 	prev_state = prev->state;
>
> 	...
>
> 	if (!preempt && prev_state && prev_state == prev->state) {
> 		if (signal_pending_state(prev_state, prev)) {
> 			prev->state = TASK_RUNNING;
> 		} else {
>
> and ptrace_freeze_traced() can change ->state in between. This means
> that this task can return from __schedule() with ->state != RUNNING,
> this can explain BUG_ON(task_is_stopped_or_traced) in do_notify_parent()
> you reported.
>
> Oleg.