Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

gctest hangs rarely on Linux if compiled with TSan #236

Open
ivmai opened this issue Sep 10, 2018 · 18 comments
Open

gctest hangs rarely on Linux if compiled with TSan #236

ivmai opened this issue Sep 10, 2018 · 18 comments

Comments

@ivmai
Copy link
Owner

ivmai commented Sep 10, 2018

Build link: https://travis-ci.org/ivmai/bdwgc/jobs/425865116

Source master (or 8.0.0)
Host: Linux/x64

How to build and run: ./configure --disable-parallel-mark && make check CFLAGS_EXTRA="-fsanitize=thread -D NO_CANCEL_SAFE -D NO_INCREMENTAL -D USE_SPIN_LOCK -fno-omit-frame-pointer -D NTHREADS=15"
@ivmai
Copy link
Owner Author

ivmai commented Apr 23, 2019

Reproduced on latest master (a8c5ee4)
Build link: https://travis-ci.org/ivmai/bdwgc/jobs/523219610

@ivmai
Copy link
Owner Author

ivmai commented Aug 6, 2021

Reproduced on latest master (dd1d0bc)
Build link: https://travis-ci.com/github/ivmai/bdwgc/jobs/529655035

@ivmai ivmai changed the title gctest hangs sometimes if compiled with TSan gctest hangs frequently on Linux if compiled with TSan Mar 16, 2022
@ivmai ivmai mentioned this issue Mar 25, 2022
Closed
@ivmai ivmai changed the title gctest hangs frequently on Linux if compiled with TSan gctest hangs rarely on Linux if compiled with TSan Mar 25, 2022
@ivmai
Copy link
Owner Author

ivmai commented Mar 25, 2022

Source: master (7aa23f4)

Build: https://app.travis-ci.com/github/ivmai/bdwgc/jobs/564340350
gctest hang.

@ivmai
Copy link
Owner Author

ivmai commented Mar 31, 2022

Build: https://app.travis-ci.com/github/ivmai/bdwgc/jobs/565405309
Source: master (0eb8d2f)

@ivmai
Copy link
Owner Author

ivmai commented Apr 1, 2022

Build: https://app.travis-ci.com/github/ivmai/bdwgc/jobs/565562062
Source: master (e34da92)

ivmai added a commit that referenced this issue Apr 2, 2022
@ivmai
Avoid hang in usleep during signals resend in child process if TSan
ba0616b 
Issue #236 (bdwgc).

If a multi-threaded process has been forked, then TSan (as of now)
cannot reasonably function in the child, e.g. usleep() may hang
because some internal lock is not released at fork.  The current
solution is just to disable signals resend in the child process
if there are threads not survived during fork.

* include/private/pthread_stop_world.h [CAN_HANDLE_FORK
&& THREAD_SANITIZER && SIGNAL_BASED_STOP_WORLD] (GC_retry_signals):
Declare GC_EXTERN variable.
* pthread_stop_world.c [CAN_HANDLE_FORK && THREAD_SANITIZER]
(GC_retry_signals): Define as GC_INNER (instead of STATIC).
* pthread_stop_world.c (GC_retry_signals): Always initialize to FALSE;
move comment to GC_stop_init.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& !NO_RETRY_SIGNALS] (GC_stop_init): Set GC_retry_signals to TRUE.
* pthread_support.c [CAN_HANDLE_FORK] (GC_remove_all_threads_but_me):
Update comment; change return type to GC_bool; define, set and return
removed local variable.
* pthread_support.c [CAN_HANDLE_FORK] (fork_child_proc): Define
threads_removed local variable and set it to the result of
GC_remove_all_threads_but_me().
* pthread_support.c [CAN_HANDLE_FORK && THREAD_SANITIZER
&& SIGNAL_BASED_STOP_WORLD] (fork_child_proc): If threads_removed or
GC_parallel then GC_retry_signals set to FALSE; add comment.
@ivmai
Copy link
Owner Author

ivmai commented Apr 2, 2022 •
edited
Loading

gctest still hangs sometimes (even after commit
ba0616b), e.g. https://app.travis-ci.com/github/ivmai/bdwgc/jobs/565659019

@ivmai
Copy link
Owner Author

ivmai commented Apr 3, 2022 •
edited
Loading

Reproduced even w/o GC_ENABLE_SUSPEND_THREAD - should be fixed by 09dd6a6

ivmai added a commit that referenced this issue Apr 3, 2022
@ivmai
Revert "Avoid hang in usleep during signals resend in child if TSan"
e2c9316 
This reverts commit ba0616b.

Issue #236 (bdwgc).

Because threads creation in the child process is not supported
properly by ThreadSanitizer anyway.
@ivmai
Copy link
Owner Author

ivmai commented Apr 4, 2022

Related issue #181

ivmai added a commit that referenced this issue Apr 11, 2022
@ivmai
Avoid potential signal loss before sigsuspend in suspend_handler if TSan
9bf0bb0 
Issue #236 (bdwgc).

As comment between sem_post() and sigsuspend() says GC_sig_thr_restart
signal should be masked at that point otherwise there could be a race.
Thus, this commit removes pthread_sigmask(SIG_UNBLOCK) call before
sem_post() one in GC_suspend_handler_inner.

* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& THREAD_SANITIZER] (GC_suspend_handler_inner): Remove "set" local
variable; do not call sigemptyset(), pthread_sigmask(SIG_UNBLOCK),
sigaddset().
ivmai added a commit that referenced this issue Apr 11, 2022
@ivmai
Do not use usleep between signals resend if ThreadSanitizer
09dd6a6 
Issue #236 (bdwgc).

For a reason, usleep() hangs trying to acquire some TSan lock when
called resend_lost_signals().

* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && THREAD_SANITIZER]
(GC_usleep): Use sched_yield() in a loop instead of usleep() or
nanosleep(); update comment.
@ivmai
Copy link
Owner Author

ivmai commented Apr 11, 2022

Latest build: https://app.travis-ci.com/github/ivmai/bdwgc/jobs/566687404
Source: master (09dd6a6)

ivmai added a commit that referenced this issue Apr 20, 2022
@ivmai
Fix hang in select() called from suspend signal handler if TSan
c207ad8 
Issue #236 (bdwgc).

Previously select() was used to sleep in the suspend signal handler
while the thread is manually suspended.  This is changed to use
sigsuspend() instead.  (But select() is still used for a reason when
the thread is self-suspended.)

* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL] (GC_stop_count):
Update comment.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_handler_inner): Remove
calls of GC_store_stack_ptr(), sem_post(), GC_suspend_self_inner(),
RESTORE_CANCEL() (dedicated to manual thread suspend).
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD && E2K] (GC_suspend_handler_inner): Remove
backing_store_end and backing_store_ptr set and clear dedicated to
manual thread suspend.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_handler_inner): Repeat
sigsuspend() while suspend_cnt&1 and me->stop_info.ext_suspend_cnt is
not updated.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_self_inner): Add
DISABLE_CANCEL() and RESTORE_CANCEL(); refine TODO item.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD && DEBUG_THREADS] (GC_suspend_self_inner):
Log "suspend self" and "resume self" events.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_thread): Rename
saved_stop_count local variable to next_stop_count; add assertion
that self thread is not suspended; replace sem_wait() in a loop to
suspend_restart_barrier(1); increment GC_stop_count on exit of the
function as well (instead of restore).
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_resume_thread): Add assertion that
GC_stop_count is odd; call raise_signal(GC_sig_thr_restart) and
suspend_restart_barrier(1).
@ivmai
Copy link
Owner Author

ivmai commented Apr 20, 2022

The issue is still reproduced rarely, e.g.:

(gdb) t 18
#0  0x00000000004a0067 in __sanitizer::internal_sched_yield() ()
#1  0x000000000047ec45 in __tsan::Mutex::Lock() ()
#2  0x00000000004957a7 in __tsan::MetaMap::GetAndLock(__tsan::ThreadState*, unsigned long, unsigned long, bool, bool) ()
#3  0x000000000048dad5 in __tsan::MutexUnlock(__tsan::ThreadState*, unsigned long, unsigned long, unsigned int) ()
#4  0x0000000000443ca0 in pthread_mutex_unlock ()
#5  0x00000000004ea041 in GC_suspend_thread (thread=<optimized out>) at pthread_stop_world.c:650
#6  0x00000000004b82be in fork_a_thread () at tests/gctest.c:657


(gdb) t 23
#0  0x00007f7f5a5524ba in __GI___sigsuspend (set=0xfed438 <suspend_handler_mask>)
    at ../sysdeps/unix/sysv/linux/sigsuspend.c:26
#1  0x000000000042c485 in sigsuspend ()
#2  0x00000000004e9a8a in GC_suspend_handler_inner (dummy=<optimized out>, context=<optimized out>)
    at pthread_stop_world.c:396
#3  0x00000000004e98ae in GC_suspend_sigaction (sig=<optimized out>, info=<optimized out>, context=<optimized out>)
    at pthread_stop_world.c:276
#4  0x000000000042c96c in __tsan::CallUserSignalHandler(__tsan::ThreadState*, bool, bool, bool, int, __sanitizer::__sanitizer_siginfo*, void*) ()
#5  0x000000000046459c in rtl_sigaction(int, __sanitizer::__sanitizer_siginfo*, void*) ()
#6  <signal handler called>
#7  0x0000000000499217 in __sanitizer::DD::MutexBeforeLock(__sanitizer::DDCallback*, __sanitizer::DDMutex*, bool) ()
#8  0x000000000048d210 in __tsan::MutexPreLock(__tsan::ThreadState*, unsigned long, unsigned long, unsigned int) ()
#9  0x0000000000443b10 in pthread_mutex_lock ()
#10 0x00000000004e8fae in GC_generic_lock (lock=0xfec6d0 <GC_allocate_ml>) at pthread_support.c:2354
#11 0x00000000004e4b03 in GC_lock () at pthread_support.c:2454

Unclear how to workaround this.

ivmai added a commit that referenced this issue Apr 21, 2022
@ivmai
Avoid potential signal loss before sigsuspend in suspend_handler if TSan
9964b6b 
Issue #236 (bdwgc).

As comment between sem_post() and sigsuspend() says GC_sig_thr_restart
signal should be masked at that point otherwise there could be a race.
Thus, this commit removes pthread_sigmask(SIG_UNBLOCK) call before
sem_post() one in GC_suspend_handler_inner.

* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& THREAD_SANITIZER] (GC_suspend_handler_inner): Remove "set" local
variable; do not call sigemptyset(), pthread_sigmask(SIG_UNBLOCK),
sigaddset().
ivmai added a commit that referenced this issue Apr 21, 2022
@ivmai
Do not use usleep between signals resend if ThreadSanitizer
54f957f 
Issue #236 (bdwgc).

For a reason, usleep() hangs trying to acquire some TSan lock when
called resend_lost_signals().

* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && THREAD_SANITIZER]
(GC_usleep): Use sched_yield() in a loop instead of usleep() or
nanosleep(); update comment.
ivmai added a commit that referenced this issue Apr 21, 2022
@ivmai
Fix hang in select() called from suspend signal handler if TSan
94eb525 
(a cherry-pick of commits c207ad8, cae46fb from 'master')

Issue #236 (bdwgc).

Previously select() was used to sleep in the suspend signal handler
while the thread is manually suspended.  This is changed to use
sigsuspend() instead.  (But select() is still used for a reason when
the thread is self-suspended.)

* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_handler_inner): Remove
calls of GC_store_stack_ptr(), sem_post(), GC_suspend_self_inner(),
RESTORE_CANCEL() (dedicated to manual thread suspend).
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD && E2K] (GC_suspend_handler_inner): Remove
backing_store_end and backing_store_ptr set and clear dedicated to
manual thread suspend.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_handler_inner): Do not return
quickly on a duplicate signal if suspend_cnt&1; repeat sigsuspend()
while suspend_cnt&1 and me->stop_info.ext_suspend_cnt is not updated.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_self_inner): Add
DISABLE_CANCEL() and RESTORE_CANCEL(); refine TODO item.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD && DEBUG_THREADS] (GC_suspend_self_inner):
Log "suspend self" and "resume self" events.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_thread): Add assertion
that self thread is not suspended; replace sem_wait() in a loop to
suspend_restart_barrier(1).
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_resume_thread): Call
raise_signal(GC_sig_thr_restart) and suspend_restart_barrier(1).
@ivmai
Copy link
Owner Author

ivmai commented Apr 23, 2022

Latest build (hang): https://app.travis-ci.com/github/ivmai/bdwgc/jobs/567979958
Source: master (fd4e579)

@ivmai ivmai mentioned this issue Apr 24, 2022
Open
ivmai added a commit that referenced this issue Apr 24, 2022
@ivmai
Fix hang in select() called from suspend signal handler if TSan
386aa6b 
(a cherry-pick of commit 94eb525 from 'release-8_2')

Issue #236 (bdwgc).

Previously select() was used to sleep in the suspend signal handler
while the thread is manually suspended.  This is changed to use
sigsuspend() instead.  (But select() is still used for a reason when
the thread is self-suspended.)

* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_handler_inner): Remove
calls of GC_store_stack_ptr(), sem_post(), GC_suspend_self_inner(),
RESTORE_CANCEL() (dedicated to manual thread suspend).
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_handler_inner): Do not return
quickly on a duplicate signal if suspend_cnt&1; repeat sigsuspend()
while suspend_cnt&1 and me->stop_info.ext_suspend_cnt is not updated.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_self_inner): Add
DISABLE_CANCEL() and RESTORE_CANCEL(); refine TODO item.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD && DEBUG_THREADS] (GC_suspend_self_inner):
Log "suspend self" and "resume self" events.
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_suspend_thread): Add assertion
that self thread is not suspended; replace sem_wait() in a loop to
suspend_restart_barrier(1).
* pthread_stop_world.c [!GC_OPENBSD_UTHREADS && !NACL
&& GC_ENABLE_SUSPEND_THREAD] (GC_resume_thread): Call
raise_signal(GC_sig_thr_restart) and suspend_restart_barrier(1).
@ivmai
Copy link
Owner Author

ivmai commented Jun 14, 2022

Latest build: https://app.travis-ci.com/github/ivmai/bdwgc/jobs/573526977
Source: b75f1aa (master)

@ivmai
Copy link
Owner Author

ivmai commented Jul 15, 2022

Build: https://app.travis-ci.com/github/ivmai/bdwgc/jobs/576835719
Source: release-8_2 (4919305)

@ivmai
Copy link
Owner Author

ivmai commented Nov 16, 2022

Build: https://app.travis-ci.com/github/ivmai/bdwgc/jobs/588637564
Source: release-8_2 (2b342c4)

@ivmai
Copy link
Owner Author

ivmai commented Apr 7, 2023

Cannot reproduce locally on this config (as of commit 6dfd81e on master):
./configure --enable-werror --enable-gc-assertions --disable-shared
make -j check CFLAGS_EXTRA="-O0 -g -fsanitize=thread -fno-omit-frame-pointer"
GC_PRINT_STATS=1 gdb ./gctest

@ivmai
Copy link
Owner Author

ivmai commented Apr 27, 2023

Reproduced locally.
Source: master (7334cc9)
Host: Linux 5.15.0-69-generic / x86_64
Compiler: gcc (Ubuntu 11.3.0-1ubuntu1~22.04) 11.3.0
How to build: ./configure --enable-static && make -j check CFLAGS_EXTRA="-O3 -fsanitize=thread -fno-omit-frame-pointer"
Connect to gctest process: LD_LIBRARY_PATH=.libs gdb .libs/gctest -pid <pid>

See stack trace of thread 18: __tsan::MutexUnlock

(gdb) t 1
[Switching to thread 1 (Thread 0x7f732b238340 (LWP 2947513))]
#0  futex_wait (private=0, expected=2, futex_word=0x7f732b5f7400 <GC_allocate_ml>) at ../sysdeps/nptl/futex-internal.h:146
146     ../sysdeps/nptl/futex-internal.h: No such file or directory.
(gdb) bt
#0  futex_wait (private=0, expected=2, futex_word=0x7f732b5f7400 <GC_allocate_ml>) at ../sysdeps/nptl/futex-internal.h:146
#1  __GI___lll_lock_wait (futex=futex@entry=0x7f732b5f7400 <GC_allocate_ml>, private=0) at ./nptl/lowlevellock.c:49
#2  0x00007f732b3bb082 in lll_mutex_lock_optimized (mutex=0x7f732b5f7400 <GC_allocate_ml>) at ./nptl/pthread_mutex_lock.c:48
#3  ___pthread_mutex_lock (mutex=0x7f732b5f7400 <GC_allocate_ml>) at ./nptl/pthread_mutex_lock.c:93
#4  0x00007f732b64c994 in __interceptor_pthread_mutex_lock (m=m@entry=0x7f732b5f7400 <GC_allocate_ml>) at ../../../../src/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:4242
#5  0x00007f732b5a0ae3 in GC_generic_lock (lock=0x7f732b5f7400 <GC_allocate_ml>) at pthread_support.c:2608
#6  0x00007f732b5a1ac4 in GC_lock () at pthread_support.c:2718
#7  0x00007f732b58bb45 in GC_malloc_kind_global (lb=lb@entry=937, k=k@entry=1) at malloc.c:319
#8  0x00007f732b5a441b in GC_malloc_kind (bytes=bytes@entry=937, kind=kind@entry=1) at thread_local_alloc.c:187
#9  0x00007f732b58bbf7 in GC_malloc (lb=lb@entry=937) at malloc.c:363
#10 0x00005632ff0b8c5a in cons (x=0x7f732490ae40, y=0x7f731ffc9000) at tests/gctest.c:320
#11 0x00005632ff0b906e in reverse1 (y=0x7f731ffc9000, x=0x7f7327be3c00) at tests/gctest.c:449
#12 reverse1 (x=<optimized out>, y=<optimized out>) at tests/gctest.c:449
#13 0x00005632ff0b90d6 in reverse1 (y=<optimized out>, x=0x7f7327beb000) at tests/gctest.c:449
#14 reverse1 (y=<optimized out>, x=<optimized out>) at tests/gctest.c:449
#15 reverse1 (y=<optimized out>, x=<optimized out>) at tests/gctest.c:449
#16 reverse1 (x=<optimized out>, y=<optimized out>) at tests/gctest.c:449
#17 0x00005632ff0bb1de in reverse1 (y=<optimized out>, x=0x7f73249ac800) at tests/gctest.c:449
#18 reverse1 (y=<optimized out>, x=<optimized out>) at tests/gctest.c:449
#19 reverse1 (y=<optimized out>, x=<optimized out>) at tests/gctest.c:449
#20 reverse1 (y=<optimized out>, x=<optimized out>) at tests/gctest.c:449
#21 reverse1 (y=<optimized out>, x=<optimized out>) at tests/gctest.c:449
#22 reverse1 (y=<optimized out>, x=<optimized out>) at tests/gctest.c:449
#23 reverse1 (y=0x0, x=<optimized out>) at tests/gctest.c:449
#24 reverse (x=<optimized out>) at tests/gctest.c:458
#25 reverse_test_inner (data=data@entry=0x1) at tests/gctest.c:944
#26 0x00007f732b5a2818 in GC_call_with_gc_active (fn=fn@entry=0x5632ff0ba780 <reverse_test_inner>, client_data=client_data@entry=0x1) at pthread_support.c:2034
#27 0x00005632ff0bb491 in reverse_test_inner (data=0x0) at tests/gctest.c:847
#28 0x00007f732b5a230b in GC_do_blocking_inner (data=0x7ffdcafe28a0 "\200\247\v\377\062V", context=<optimized out>) at pthread_support.c:1854
#29 0x00007f732b58a79a in GC_with_callee_saves_pushed (fn=0x7f732b5a2230 <GC_do_blocking_inner>, arg=<optimized out>, arg@entry=0x7ffdcafe28a0 "\200\247\v\377\062V") at mach_dep.c:421
#30 0x00007f732b59544e in GC_do_blocking (fn=fn@entry=0x5632ff0ba780 <reverse_test_inner>, client_data=client_data@entry=0x0) at misc.c:2314
#31 0x00005632ff0bd418 in reverse_test () at tests/gctest.c:979
#32 run_one_test () at tests/gctest.c:1845
#33 0x00005632ff0b815e in main () at tests/gctest.c:2643

(gdb) t 18
[Switching to thread 18 (Thread 0x7f7320ff8640 (LWP 2947796))]
#0  __sanitizer::internal_sched_yield () at ../../../../src/libsanitizer/sanitizer_common/sanitizer_syscall_linux_x86_64.inc:19
19      ../../../../src/libsanitizer/sanitizer_common/sanitizer_syscall_linux_x86_64.inc: No such file or directory.
(gdb) bt
#0  __sanitizer::internal_sched_yield () at ../../../../src/libsanitizer/sanitizer_common/sanitizer_syscall_linux_x86_64.inc:19
#1  0x00007f732b686bd5 in __tsan::Backoff::Do (this=<synthetic pointer>) at ../../../../src/libsanitizer/tsan/tsan_mutex.cpp:194
#2  __tsan::Mutex::Lock (this=0x7f732a530008) at ../../../../src/libsanitizer/tsan/tsan_mutex.cpp:234
#3  0x00007f732b6a15b5 in __tsan::MetaMap::GetAndLock (this=0x7f732b728408 <__tsan::ctx_placeholder+8>, thr=thr@entry=0x7f7320fb8ac0, pc=pc@entry=140132625895417,
    addr=addr@entry=140132625642496, write_lock=write_lock@entry=true, create=create@entry=true) at ../../../../src/libsanitizer/tsan/tsan_sync.cpp:226
#4  0x00007f732b6a19c3 in __tsan::MetaMap::GetOrCreateAndLock (this=<optimized out>, thr=thr@entry=0x7f7320fb8ac0, pc=pc@entry=140132625895417, addr=addr@entry=140132625642496,
    write_lock=write_lock@entry=true) at ../../../../src/libsanitizer/tsan/tsan_sync.cpp:198
#5  0x00007f732b699963 in __tsan::MutexUnlock (thr=thr@entry=0x7f7320fb8ac0, pc=pc@entry=140132625895417, addr=addr@entry=140132625642496, flagz=flagz@entry=0)
    at ../../../../src/libsanitizer/tsan/tsan_rtl_mutex.cpp:217
#6  0x00007f732b63507b in __interceptor_pthread_mutex_unlock (m=0x7f732b5f7400 <GC_allocate_ml>) at ../../../../src/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:4255
#7  0x00007f732b5a53fc in GC_suspend_thread (thread=<optimized out>) at pthread_stop_world.c:646
#8  0x00005632ff0ba48c in fork_a_thread () at tests/gctest.c:734
#9  0x00005632ff0bb211 in reverse_test_inner (data=data@entry=0x1) at tests/gctest.c:937
#10 0x00007f732b5a2818 in GC_call_with_gc_active (fn=fn@entry=0x5632ff0ba780 <reverse_test_inner>, client_data=client_data@entry=0x1) at pthread_support.c:2034
#11 0x00005632ff0bb491 in reverse_test_inner (data=0x0) at tests/gctest.c:847
#12 0x00007f732b5a230b in GC_do_blocking_inner (data=0x7f7320fb7f30 "\200\247\v\377\062V", context=<optimized out>) at pthread_support.c:1854
#13 0x00007f732b58a79a in GC_with_callee_saves_pushed (fn=0x7f732b5a2230 <GC_do_blocking_inner>, arg=<optimized out>, arg@entry=0x7f7320fb7f30 "\200\247\v\377\062V") at mach_dep.c:421
#14 0x00007f732b59544e in GC_do_blocking (fn=fn@entry=0x5632ff0ba780 <reverse_test_inner>, client_data=client_data@entry=0x0) at misc.c:2314
#15 0x00005632ff0bd418 in reverse_test () at tests/gctest.c:979
#16 run_one_test () at tests/gctest.c:1845
#17 0x00005632ff0bd90c in thr_run_one_test (arg=<optimized out>) at tests/gctest.c:2535
#18 0x00007f732b5a05d8 in GC_pthread_start_inner (sb=sb@entry=0x7f7320fb81b0, arg=arg@entry=0x7ffdcafe29c0) at pthread_start.c:55
#19 0x00007f732b5953a7 in GC_call_with_stack_base (fn=0x7f732b5a0530 <GC_pthread_start_inner>, arg=arg@entry=0x7ffdcafe29c0) at misc.c:2180
#20 0x00007f732b5a09f9 in GC_pthread_start (arg=0x7ffdcafe29c0) at pthread_support.c:2425
#21 0x00007f732b627600 in __tsan_thread_start_func (arg=0x7ffdcafe28b0) at ../../../../src/libsanitizer/tsan/tsan_interceptors_posix.cpp:959
#22 0x00007f732b3b7b43 in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:442
#23 0x00007f732b449a00 in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81

(gdb) t 23
[Switching to thread 23 (Thread 0x7f731cf0e640 (LWP 2949801))]
#0  0x00007f732b36580a in __GI___sigsuspend (set=0x7f732b5f8140 <suspend_handler_mask>) at ../sysdeps/unix/sysv/linux/sigsuspend.c:26
26      ../sysdeps/unix/sysv/linux/sigsuspend.c: No such file or directory.
(gdb) bt
#0  0x00007f732b36580a in __GI___sigsuspend (set=0x7f732b5f8140 <suspend_handler_mask>) at ../sysdeps/unix/sysv/linux/sigsuspend.c:26
#1  0x00007f732b62bfb3 in __interceptor_sigsuspend (mask=mask@entry=0x7f732b5f8140 <suspend_handler_mask>) at ../../../../src/libsanitizer/tsan/tsan_interceptors_posix.cpp:1872
#2  0x00007f732b5a4a48 in GC_suspend_handler_inner (dummy=0x0, context=<optimized out>) at pthread_stop_world.c:398
#3  GC_suspend_sigaction (sig=<optimized out>, info=<optimized out>, context=<optimized out>) at pthread_stop_world.c:276
#4  0x00007f732b6267f7 in __tsan::CallUserSignalHandler (thr=thr@entry=0x7f731ceceac0, sync=<optimized out>, acquire=acquire@entry=true, sigact=sigact@entry=true, sig=sig@entry=31,
    info=info@entry=0x7f731cecd170, uctx=0x7f731cecd040) at ../../../../src/libsanitizer/tsan/tsan_interceptors_posix.cpp:1922
#5  0x00007f732b628df5 in rtl_generic_sighandler (ctx=0x7f731cecd040, info=0x7f731cecd170, sig=31, sigact=true) at ../../../../src/libsanitizer/tsan/tsan_interceptors_posix.cpp:2016
#6  rtl_sigaction (sig=<optimized out>, info=0x7f731cecd170, ctx=0x7f731cecd040) at ../../../../src/libsanitizer/tsan/tsan_interceptors_posix.cpp:2041
#7  <signal handler called>
#8  0x00007f732b698473 in __tsan::MutexPreLock (thr=thr@entry=0x7f731ceceac0, pc=pc@entry=140132625991945, addr=addr@entry=140132625642496, flagz=flagz@entry=0)
    at ../../../../src/libsanitizer/tsan/tsan_rtl_mutex.cpp:149
#9  0x00007f732b64c98b in __interceptor_pthread_mutex_lock (m=m@entry=0x7f732b5f7400 <GC_allocate_ml>) at ../../../../src/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:4241
#10 0x00007f732b5a0ae3 in GC_generic_lock (lock=0x7f732b5f7400 <GC_allocate_ml>) at pthread_support.c:2608
#11 0x00007f732b5a1ac4 in GC_lock () at pthread_support.c:2718
#12 0x00007f732b58bb45 in GC_malloc_kind_global (lb=lb@entry=16, k=k@entry=0) at malloc.c:319
#13 0x00007f732b5a441b in GC_malloc_kind (bytes=bytes@entry=16, kind=kind@entry=0) at thread_local_alloc.c:187
#14 0x00007f732b58bba4 in GC_malloc_atomic (lb=lb@entry=16) at malloc.c:357
#15 0x00005632ff0b9348 in small_cons_leaf (x=10) at tests/gctest.c:401
#16 ints (low=<optimized out>, up=10) at tests/gctest.c:507
#17 0x00005632ff0b933b in ints (low=<optimized out>, up=10) at tests/gctest.c:507
#18 0x00005632ff0b933b in ints (low=<optimized out>, up=10) at tests/gctest.c:507
#19 0x00005632ff0b933b in ints (low=<optimized out>, up=10) at tests/gctest.c:507
#20 0x00005632ff0b933b in ints (low=<optimized out>, up=10) at tests/gctest.c:507
#21 0x00005632ff0b933b in ints (low=<optimized out>, up=10) at tests/gctest.c:507
#22 0x00005632ff0b933b in ints (low=<optimized out>, up=10) at tests/gctest.c:507
#23 0x00005632ff0b933b in ints (low=<optimized out>, up=10) at tests/gctest.c:507
#24 0x00005632ff0b9f6a in ints (up=10, low=2) at tests/gctest.c:507
#25 ints (low=1, up=10) at tests/gctest.c:507
#26 tiny_reverse_test_inner () at tests/gctest.c:645
#27 0x00005632ff0ba323 in tiny_reverse_test (p_resumed=0x0) at tests/gctest.c:671
#28 0x00007f732b5a05d8 in GC_pthread_start_inner (sb=sb@entry=0x7f731cece1b0, arg=arg@entry=0x7f7320fb7770) at pthread_start.c:55
#29 0x00007f732b5953a7 in GC_call_with_stack_base (fn=0x7f732b5a0530 <GC_pthread_start_inner>, arg=arg@entry=0x7f7320fb7770) at misc.c:2180
#30 0x00007f732b5a09f9 in GC_pthread_start (arg=0x7f7320fb7770) at pthread_support.c:2425
#31 0x00007f732b627600 in __tsan_thread_start_func (arg=0x7f7320fb7660) at ../../../../src/libsanitizer/tsan/tsan_interceptors_posix.cpp:959
#32 0x00007f732b3b7b43 in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:442
#33 0x00007f732b449a00 in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81

@ivmai
Copy link
Owner Author

ivmai commented Apr 28, 2023

Source: release-8_2 (bef858c)
Build 1: https://app.travis-ci.com/github/ivmai/bdwgc/jobs/601194806
Build 2: https://app.travis-ci.com/github/ivmai/bdwgc/jobs/601194807
Host: Ubuntu 16.04.7 LTS / x64
Compiler: clang version 7.0.0 (tags/RELEASE_700/final)
Config 1: CFLAGS_EXTRA="-fsanitize=thread -D NO_CANCEL_SAFE -D NO_INCREMENTAL -D USE_SPIN_LOCK -fno-omit-frame-pointer -D TEST_FORK_WITHOUT_ATFORK" CONF_OPTIONS="--enable-gc-assertions --enable-gc-debug --enable-handle-fork=manual --enable-large-config --with-libatomic-ops=no"
Config 2: CFLAGS_EXTRA="-fsanitize=thread -D NO_CANCEL_SAFE -D NO_INCREMENTAL -D USE_SPIN_LOCK -fno-omit-frame-pointer -D NTHREADS=15" CONF_OPTIONS="--disable-parallel-mark"

@ivmai
Copy link
Owner Author

ivmai commented Sep 10, 2023

Build: https://app.travis-ci.com/github/ivmai/bdwgc/jobs/609522449
Source: release-8_2 (399005e)
Host: Ubuntu 16.04.7 LTS / amd64
Compiler: clang version 7.0.0 (tags/RELEASE_700/final)
Config: CFLAGS_EXTRA="-fsanitize=thread -D NO_CANCEL_SAFE -D NO_INCREMENTAL -D USE_SPIN_LOCK -fno-omit-frame-pointer -D NTHREADS=15" CONF_OPTIONS="--disable-parallel-mark"

@ivmai ivmai mentioned this issue Oct 16, 2024
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@ivmai