Re: [path][rfc] add PR_DETACH prctl command [3/3]

[Oleg Nesterov <oleg@redhat.com>]

On 04/19, Oleg Nesterov wrote:
>
> I'll try to check these patches from the correctness pov tomorrow,
> but to be honest I hope someone will nack them before I start ;)

OK, I briefly looked at 3/3. It looks certainly wrong.

	notif = do_signal_parent(...);
	if (notif != DEATH_REAP) {
	....

do_signal_parent() must not return DEATH_REAP (this means that
leader->exit_signal becomes -1), but this can happen and this is bug.

I can be wrong, but iirc this was already discussed, probably when
you sent the very first version which had the same problem. That is
why (in particular) do_notify_parent() does

	BUG_ON(tsk->group_leader != tsk || !thread_group_empty(tsk))

but you simply removed this check to hide the problem.



Also. I didn't actually read the patch yet, but iiuc: if a task T does
PR_DETACH and then exits, init can't reap it until the old parent does
wait. This can confuse the poor admin, he can see the zombies with
ppid = 1 and there is no way to understand why.

Oleg.