Date: Sun, 4 Feb 2001 02:09:29 -0800 (PST) From: jml@cubical.fi To: freebsd-gnats-submit@FreeBSD.org Subject: ports/24845: linuxthreads does not detect failed rfork Message-ID: <200102041009.f14A9Tg98764@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 24845
>Category: ports
>Synopsis: linuxthreads does not detect failed rfork
>Confidential: no
>Severity: non-critical
>Priority: medium
>Responsible: freebsd-ports
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Sun Feb 04 02:10:01 PST 2001
>Closed-Date:
>Last-Modified:
>Originator: Juha-Matti Liukkonen
>Release: 4.2-RELEASE
>Organization:
Cubical Solutions Ltd
>Environment:
FreeBSD snafu.intra.net 4.2-RELEASE FreeBSD 4.2-RELEASE #0: Sat Feb 3 20:12:33 EET 2001
root@:/usr/src/sys/compile/SERVER i386
>Description:
When a new thread is created by the manager thread calling __clone()
and rfork() fails (eg. proc limit reached), the manager thread enters
busy loop.
>How-To-Repeat:
A small program:
void* func(void* arg) { sleep(100); return 0; }
main() {
struct rlimit rl = { 20, 0 };
pthread_t pt;
setrlimit(RLIMIT_NPROC, &rl);
for (;;) pthread_create(&pt, 0, func, 0);
}
Observe "top" or "ps l" output. Issue sysctl -w kern.corefile="%N-%P.core"
to get per-thread cores. Get a core dump with "kill -TRAP
<manager-pid>; kill <main-pid>"; load the manager's core and say where.
>Fix:
In clone.S, the execution seems to go to the error branch correctly,
but the subsequent return from __clone throws the manager thread back
to its entry point.
Adding "call _exit" before the .cerror jmp exits the manager thread
in the error situation -- maybe the stack gets corrupted somehow to
cause the ret from the .cerror macro to fail?
>Release-Note:
>Audit-Trail:
>Unformatted:
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-ports" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200102041009.f14A9Tg98764>