Date: Thu, 3 Mar 2005 10:14:21 -0500 (EST) From: Daniel Eischen <deischen@freebsd.org> To: Andriy Tkachuk <andrit@ukr.net> Cc: David Xu <davidxu@freebsd.org> Subject: Re: patch for threads/76690 - critical - fork hang in child for-lc_r Message-ID: <Pine.GSO.4.43.0503031008130.2058-201000@sea.ntplx.net> In-Reply-To: <012901c51fd3$131e23b0$090210ac@BORJA>
next in thread | previous in thread | raw e-mail | index | archive | help
This message is in MIME format. The first part should be readable text,
while the remaining parts are likely unreadable without MIME-aware tools.
Send mail to mime@docserver.cac.washington.edu for more info.
------=_NextPart_000_0126_01C52001.298C64D0
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
Content-ID: <Pine.GSO.4.43.0503031008132.2058@sea.ntplx.net>
On Thu, 3 Mar 2005, Andriy Tkachuk wrote:
> > > > Hmm, libc_r and libpthread handle spinlock differently which malloc
> > > > uses to protect itself, some real world benchmarks are better than
> > this.
>
> yes , you right, David. one have to check __isthreaded before
> firing _SPINLOCK. there will be nothing wrong, because
>
> static spinlock_t thread_lock = _SPINLOCK_INITIALIZER;
>
> initialyzed regardless __isthreaded in malloc.c but
> for optimization probably it is worth to add this check.
> Take a look on updated patch.
>
> btw: i don't see the unlock in child in libpthread. there must be two
> unlocks
I told you in previous mail, _kse_single_thread() calls
_thr_spinlock_init(). The malloc lock is not the only
lock used in libc, so the safe way to make sure libc
is in a clean state after a fork is to reinitialize all
the locks used by libc, not just the malloc lock. libc
really shouldn't try to use any locks unless __isthreaded
is true, so after a fork() it shouldn't really matter
what state the locks are in.
--
DE
------=_NextPart_000_0126_01C52001.298C64D0
Content-Type: APPLICATION/OCTET-STREAM; NAME="libc_r-76690pr.patch2"
Content-Transfer-Encoding: QUOTED-PRINTABLE
Content-ID: <Pine.GSO.4.43.0503031008133.2058@sea.ntplx.net>
Content-Description:
Content-Disposition: ATTACHMENT; FILENAME="libc_r-76690pr.patch2"
diff -r -u lib/libc_r/uthread/uthread_fork.c =
lib.patched/libc_r/uthread/uthread_fork.c=0A=
--- lib/libc_r/uthread/uthread_fork.c Fri Dec 10 09:06:45 2004=0A=
+++ lib.patched/libc_r/uthread/uthread_fork.c Thu Mar 3 14:50:06 2005=0A=
@@ -68,8 +68,15 @@=0A=
af->prepare();=0A=
}=0A=
=0A=
+ extern spinlock_t *__malloc_lock;=0A=
+ if (__isthreaded && __malloc_lock !=3D NULL)=0A=
+ _SPINLOCK(__malloc_lock);=0A=
+=0A=
/* Fork a new process: */=0A=
if ((ret =3D __sys_fork()) !=3D 0) {=0A=
+ if (__isthreaded && __malloc_lock !=3D NULL)=0A=
+ _SPINUNLOCK(__malloc_lock);=0A=
+=0A=
/* Run down atfork parent handlers. */=0A=
TAILQ_FOREACH(af, &_atfork_list, qe) {=0A=
if (af->parent !=3D NULL)=0A=
@@ -78,6 +85,9 @@=0A=
_pthread_mutex_unlock(&_atfork_mutex);=0A=
=0A=
} else {=0A=
+ if (__isthreaded && __malloc_lock !=3D NULL)=0A=
+ _SPINUNLOCK(__malloc_lock);=0A=
+=0A=
/* Close the pthread kernel pipe: */=0A=
__sys_close(_thread_kern_pipe[0]);=0A=
__sys_close(_thread_kern_pipe[1]);=0A=
------=_NextPart_000_0126_01C52001.298C64D0
Content-Type: TEXT/PLAIN; CHARSET=us-ascii
Content-ID: <Pine.GSO.4.43.0503031008134.2058@sea.ntplx.net>
Content-Description:
Content-Disposition: INLINE
_______________________________________________
freebsd-threads@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-threads
To unsubscribe, send any mail to "freebsd-threads-unsubscribe@freebsd.org"
------=_NextPart_000_0126_01C52001.298C64D0--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.GSO.4.43.0503031008130.2058-201000>