Re: zombie with CLONE_THREAD

[Andrea Arcangeli <andrea@suse.de>]

On Thu, Jul 01, 2004 at 08:49:10AM -0700, Linus Torvalds wrote:
> 
> 
> On Thu, 1 Jul 2004, Roland McGrath wrote:
> > 
> > > .. since this information should be available anyway (we'll have woken up 
> > > the tracer, and the tracer will see that the child is gone by simply 
> > > seeing the ESRCH errorcode from ptrace).
> > 
> > When did you wake up the tracer?  I don't see how that happened.
> 
> exit_notify() will inform the tracer:
> 
>         if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
>                 int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
>                 do_notify_parent(tsk, signal);
>         } else if (tsk->ptrace) {
> ***             do_notify_parent(tsk, SIGCHLD);   *****
>         }
> 
> so this should catch it. It even gets the pid of the child in the siginfo 
> structure if it really wants to see that..

it will get the wakeup, but that doesn't mean wait4 will return if
it's waiting for all childs. Now I don't know but with strace it may
even be ok by luck, but it certainly changes the semantics of
wait4/ptrace at least a little and I could remotely imagine if somebody
wants a wait4 to return when a ptraced child exited, if there are other
"regular" children. That's why I admitted it can be considered a feature
and not a completely worthless effort to leave self-reaping tasks as
zombies if they're ptraced.