Re: [patch] O(1) sys_exit(), threading, scalable-exit-2.5.31-A6

[Daniel Jacobowitz <dan@debian.org>]

On Tue, Aug 20, 2002 at 04:36:19PM +0200, Ingo Molnar wrote:
> 
> the attached patch ontop of BK-curr fixes the ptrace wait4() anomaly that
> can be observed in any previous Linux kernel i could get my hands at. So
> in fact ->ptrace_children, besides being a speedup, also fixed a bug that
> couldnt be fixed in any satisfactory way before.
> 
> 	Ingo
> 
> --- linux/kernel/exit.c.orig	Tue Aug 20 16:28:57 2002
> +++ linux/kernel/exit.c	Tue Aug 20 16:30:13 2002
> @@ -731,7 +731,7 @@
>  		tsk = next_thread(tsk);
>  	} while (tsk != current);
>  	read_unlock(&tasklist_lock);
> -	if (flag) {
> +	if (flag || !list_empty(&current->ptrace_children)) {
>  		retval = 0;
>  		if (options & WNOHANG)
>  			goto end_wait4;

Ingo,

At this point your ptrace changes have completely broken both
_TRACEME/exec and _ATTACH debugging.  If an attached process finishes
while a debugger is attached, its parent no longer gets the proper wait
result for it:

wait4(-1, [WIFEXITED(s) && WEXITSTATUS(s) == 0], WNOHANG|WUNTRACED, NULL) = 478
wait4(-1, [WIFEXITED(s) && WEXITSTATUS(s) == 0], WNOHANG|WUNTRACED, NULL) = 478
wait4(-1, [WIFEXITED(s) && WEXITSTATUS(s) == 0], WNOHANG|WUNTRACED, NULL) = 478
wait4(-1, [WIFEXITED(s) && WEXITSTATUS(s) == 0], WNOHANG|WUNTRACED, NULL) = 478
wait4(-1, [WIFEXITED(s) && WEXITSTATUS(s) == 0], WNOHANG|WUNTRACED, NULL) = 478
wait4(-1, [WIFEXITED(s) && WEXITSTATUS(s) == 0], WNOHANG|WUNTRACED, NULL) = 478

etc.  It is never removed from the list.  _TRACEME/exec debugging
appears to have the same problem but it's harder to tell, since one can
not strace GDB in 2.5 without the patch I posted here two weeks ago. 
If you don't have a chance to look at this I'll investigate later
today.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer