public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
From: "dvyukov at google dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug sanitizer/55561] TSAN: Fortran/OMP yields false positives
Date: Sat, 29 Dec 2012 10:13:00 -0000 [thread overview]
Message-ID: <bug-55561-4-qmWM8GdGBV@http.gcc.gnu.org/bugzilla/> (raw)
In-Reply-To: <bug-55561-4@http.gcc.gnu.org/bugzilla/>
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55561
--- Comment #20 from Dmitry Vyukov <dvyukov at google dot com> 2012-12-29 10:13:00 UTC ---
(In reply to comment #19)
> On Wed, Dec 26, 2012 at 12:23 AM, Joost.VandeVondele at mat dot
> ethz.ch <gcc-bugzilla@gcc.gnu.org> wrote:
> >
> > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55561
> >
> > --- Comment #16 from Joost VandeVondele <Joost.VandeVondele at mat dot ethz.ch> 2012-12-25 20:23:07 UTC ---
> > many things appear to work fine, but seemingly parallel do loops with a dynamic
> > schedule generate warnings in libgomp. I also seem to observe that they are not
> > strictly deterministic, sometimes these warnings happen, sometimes not.
> >
> > Testcase:
> >
> > !$OMP PARALLEL PRIVATE(j)
> >
> > j=OMP_GET_THREAD_NUM()
> >
> > ! no warnings without the dynamic schedule
> > !$OMP DO SCHEDULE(DYNAMIC,2)
> > DO i=1,10
> > ENDDO
> >
> > !$OMP END PARALLEL
> > END
> >
> > Result:
> >
> > vjoost@nanosim-s01.ethz.ch:/data/vjoost/clean/cp2k/cp2k/src> ./a.out
> > vjoost@nanosim-s01.ethz.ch:/data/vjoost/clean/cp2k/cp2k/src> ./a.out
> > vjoost@nanosim-s01.ethz.ch:/data/vjoost/clean/cp2k/cp2k/src> ./a.out
> > vjoost@nanosim-s01.ethz.ch:/data/vjoost/clean/cp2k/cp2k/src> ./a.out
> > ==================
> > WARNING: ThreadSanitizer: data race (pid=35190)
> > Read of size 8 at 0x7d3000027290 by main thread:
> > #0 gomp_iter_dynamic_next
> > /data/vjoost/gnu/gcc_trunk/obj/x86_64-unknown-linux-gnu/libgomp/../../../gcc/libgomp/iter.c:190
> > (libgomp.so.1+0x000000006678)
> > #1 GOMP_loop_dynamic_start
> > /data/vjoost/gnu/gcc_trunk/obj/x86_64-unknown-linux-gnu/libgomp/../../../gcc/libgomp/loop.c:128
> > (libgomp.so.1+0x000000007a03)
> > #2 MAIN__._omp_fn.0 test.f90:0 (exe+0x000000000d7d)
> > #3 MAIN__ test.f90:0 (exe+0x000000000ccb)
> > #4 main ??:0 (exe+0x000000000d1a)
> >
> > Previous write of size 8 at 0x7d3000027290 by thread 1:
> > #0 gomp_loop_init
> > /data/vjoost/gnu/gcc_trunk/obj/x86_64-unknown-linux-gnu/libgomp/../../../gcc/libgomp/loop.c:41
> > (libgomp.so.1+0x000000007a96)
> > #1 MAIN__._omp_fn.0 test.f90:0 (exe+0x000000000d7d)
> > #2 gomp_thread_start
> > /data/vjoost/gnu/gcc_trunk/obj/x86_64-unknown-linux-gnu/libgomp/../../../gcc/libgomp/team.c:116
> > (libgomp.so.1+0x00000000d012)
> >
> > Location is heap block of size 1568 at 0x7d3000027100 allocated by main
> > thread:
> > #0 malloc ??:0 (libtsan.so.0+0x00000001896e)
> > #1 gomp_malloc
> > /data/vjoost/gnu/gcc_trunk/obj/x86_64-unknown-linux-gnu/libgomp/../../../gcc/libgomp/alloc.c:36
> > (libgomp.so.1+0x00000000417a)
> > #2 gomp_new_team
> > /data/vjoost/gnu/gcc_trunk/obj/x86_64-unknown-linux-gnu/libgomp/../../../gcc/libgomp/team.c:145
> > (libgomp.so.1+0x00000000d27a)
> > #3 GOMP_parallel_start
> > /data/vjoost/gnu/gcc_trunk/obj/x86_64-unknown-linux-gnu/libgomp/../../../gcc/libgomp/parallel.c:108
> > (libgomp.so.1+0x00000000afc7)
> > #4 MAIN__ test.f90:0 (exe+0x000000000cc1)
> > #5 main ??:0 (exe+0x000000000d1a)
> >
> > Thread 1 (tid=35191, running) created at:
> > #0 pthread_create ??:0 (libtsan.so.0+0x00000001a868)
> > #1 gomp_team_start
> > /data/vjoost/gnu/gcc_trunk/obj/x86_64-unknown-linux-gnu/libgomp/../../../gcc/libgomp/team.c:440
> > (libgomp.so.1+0x00000000d908)
> > #2 GOMP_parallel_start
> > /data/vjoost/gnu/gcc_trunk/obj/x86_64-unknown-linux-gnu/libgomp/../../../gcc/libgomp/parallel.c:108
> > (libgomp.so.1+0x00000000afd7)
> > #3 MAIN__ test.f90:0 (exe+0x000000000cc1)
> > #4 main ??:0 (exe+0x000000000d1a)
>
>
> Looks like unsafe publication of gomp_work_share data.
>
> Can you show disassembly of
> > #2 MAIN__._omp_fn.0 test.f90:0 (exe+0x000000000d7d)
> ?
> How does it choose between calling gomp_loop_init() and
> GOMP_loop_dynamic_start()?
>
> Humm... do omp generated functions (like MAIN__._omp_fn.0) pass
> through tsan pass? Perhaps it contains some atomic op that tsan does
> not see.
Congratulations! We've found racy unsafe publication in libgomp with
ThreadSanitizer:
gomp_loop_dynamic_start() uses the following functions to synchronize with each
other. As you can see gomp_ptrlock_get() contains fast-and-dead unsafe
fast-path.
libgomp/config/posix/ptrlock.h
static inline void *gomp_ptrlock_get (gomp_ptrlock_t *ptrlock)
{
if (ptrlock->ptr != NULL)
return ptrlock->ptr;
gomp_mutex_lock (&ptrlock->lock);
if (ptrlock->ptr != NULL)
{
gomp_mutex_unlock (&ptrlock->lock);
return ptrlock->ptr;
}
return NULL;
}
static inline void gomp_ptrlock_set (gomp_ptrlock_t *ptrlock, void *ptr)
{
ptrlock->ptr = ptr;
gomp_mutex_unlock (&ptrlock->lock);
}
next prev parent reply other threads:[~2012-12-29 10:13 UTC|newest]
Thread overview: 53+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-12-02 8:16 [Bug sanitizer/55561] New: TSAN crashes for Fortran Joost.VandeVondele at mat dot ethz.ch
2012-12-02 9:21 ` [Bug sanitizer/55561] " dvyukov at google dot com
2012-12-02 9:30 ` Joost.VandeVondele at mat dot ethz.ch
2012-12-02 9:36 ` dvyukov at google dot com
2012-12-02 10:28 ` kcc at gcc dot gnu.org
2012-12-02 21:11 ` jakub at gcc dot gnu.org
2012-12-03 7:42 ` Joost.VandeVondele at mat dot ethz.ch
2012-12-10 12:44 ` Joost.VandeVondele at mat dot ethz.ch
2012-12-10 12:47 ` pinskia at gcc dot gnu.org
2012-12-10 12:53 ` Joost.VandeVondele at mat dot ethz.ch
2012-12-10 12:57 ` dvyukov at google dot com
2012-12-10 12:59 ` Joost.VandeVondele at mat dot ethz.ch
2012-12-10 15:07 ` dvyukov at google dot com
2012-12-10 15:56 ` Joost.VandeVondele at mat dot ethz.ch
2012-12-11 9:48 ` [Bug sanitizer/55561] TSAN: Fortran/OMP yields false positives Joost.VandeVondele at mat dot ethz.ch
2012-12-25 19:30 ` Joost.VandeVondele at mat dot ethz.ch
2012-12-25 20:23 ` Joost.VandeVondele at mat dot ethz.ch
2012-12-26 19:35 ` Joost.VandeVondele at mat dot ethz.ch
2012-12-29 9:33 ` dvyukov at google dot com
2012-12-29 9:38 ` dvyukov at google dot com
2012-12-29 10:13 ` dvyukov at google dot com [this message]
2012-12-29 10:21 ` dvyukov at google dot com
2012-12-30 9:03 ` Joost.VandeVondele at mat dot ethz.ch
2012-12-30 9:58 ` dvyukov at google dot com
2012-12-30 10:11 ` dvyukov at google dot com
2012-12-30 14:53 ` Joost.VandeVondele at mat dot ethz.ch
2012-12-30 17:07 ` dvyukov at google dot com
2012-12-30 19:57 ` Joost.VandeVondele at mat dot ethz.ch
2013-01-01 17:14 ` Joost.VandeVondele at mat dot ethz.ch
2013-01-02 9:09 ` dvyukov at google dot com
2013-01-02 9:44 ` jakub at gcc dot gnu.org
2013-01-02 10:28 ` dvyukov at google dot com
2013-01-07 21:35 ` Joost.VandeVondele at mat dot ethz.ch
2013-01-08 9:17 ` dvyukov at google dot com
2013-01-10 11:27 ` Joost.VandeVondele at mat dot ethz.ch
2013-01-10 11:37 ` jakub at gcc dot gnu.org
2013-01-31 7:43 ` amodra at gmail dot com
2013-01-31 14:29 ` jakub at gcc dot gnu.org
2013-01-31 16:57 ` jakub at gcc dot gnu.org
2013-02-01 20:00 ` dvyukov at google dot com
2013-03-29 8:11 ` [Bug sanitizer/55561] TSAN: provide a TSAN instrumented libgomp Joost.VandeVondele at mat dot ethz.ch
2013-06-03 13:21 ` dvyukov at google dot com
2014-01-21 12:44 ` emil.styrke at gmail dot com
2014-01-22 8:14 ` Joost.VandeVondele at mat dot ethz.ch
2014-05-08 19:49 ` roland at rschulz dot eu
2014-05-09 5:38 ` Joost.VandeVondele at mat dot ethz.ch
2014-05-09 10:26 ` dvyukov at google dot com
2014-05-14 19:17 ` Joost.VandeVondele at mat dot ethz.ch
2014-05-14 19:25 ` dvyukov at google dot com
2014-07-09 6:07 ` roland at rschulz dot eu
2014-07-09 7:12 ` jakub at gcc dot gnu.org
2014-07-09 7:17 ` Joost.VandeVondele at mat dot ethz.ch
2024-02-29 17:27 ` egallager at gcc dot gnu.org
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=bug-55561-4-qmWM8GdGBV@http.gcc.gnu.org/bugzilla/ \
--to=gcc-bugzilla@gcc.gnu.org \
--cc=gcc-bugs@gcc.gnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).