Re: + exitc-call-proc_exit_connector-after-exit_state-is-set.patch added to -mm tree

[Oleg Nesterov <oleg@redhat.com>]

On 02/27, Guillaume Morin wrote:
>
> On 27 Feb 17:47, Oleg Nesterov wrote:
>
> > Nevermind. Please consider this trivial example:
> >
> > 	tfunc(void *)
> > 	{
> > 		for (;;)
> > 			pause();
> > 	}
> >
> > 	int main(void)
> > 	{
> > 		pthread_create(tfunc);
> > 		pthread_exit();
> > 	}
> >
> > The main thread can exit and call proc_exit_connector() before
> > register_interest_for_pid(), but WNOHANG obviously can't succeed.
>
> What matters is not the exit message of the main thread but the exit
> message from the last threaded dying.  In your example, it's fine that
> waitpid fails since the process is still around.  If you kill it, the
> connector will get a connector message for the thread you created in
> main().

Yes sure. But how can you know that you should take this sub-thread
into account and this is the last thread?

OK... you can probably look at every PROC_EVENT_EXIT which has the
right ->process_tgid and use wait(WNOHANG) every time to detect the
group exit.

OK, thanks, I seem to understand now. Initially I wrongly thought
that your point is literally "tgid should me reapable right after
PROC_EVENT_EXIT with this patch".

Thanks!

Oleg.