https://bugs.kde.org/show_bug.cgi?id=407764
Bug ID: 407764
Summary: drd cond_post_wait gets wrong (?) condition on s390x
z13 system
Product: valgrind
Version: unspecified
Platform: Other
OS: Linux
Status: REPORTED
Severity: normal
Priority: NOR
Component: drd
Assignee: bvanass...@acm.org
Reporter: m...@klomp.org
CC: ar...@linux.ibm.com
Target Milestone: ---
This might be specific to RHEL8 which is build for s390x z13. I am not seeing
this on Fedora 30 on s390x (which defaults to an older s390x instruction set).
With the following program on Fedora 30 s390x:
$ rpm -q valgrind glibc binutils gcc; uname -r
valgrind-3.15.0-1.fc30.s390x
glibc-2.29-12.fc30.s390x
binutils-2.31.1-29.fc30.s390x
gcc-9.1.1-1.fc30.s390x
5.0.16-200.fc29.s390x
$ cat pthread_cond.c
// gcc -g -lpthread -o pthread_cond pthread_cond.c
#include <stdio.h>
#include <pthread.h>
#include <stdio.h>
static pthread_mutex_t mutex;
static pthread_cond_t cond;
static pthread_t thread;
static int ready = 0;
static void *ThreadFunction(void *ptr)
{
pthread_mutex_lock(&mutex);
ready = 1;
pthread_cond_signal(&cond);
pthread_mutex_unlock(&mutex);
return NULL;
}
int main()
{
pthread_mutex_init(&mutex, NULL);
pthread_cond_init(&cond, NULL);
pthread_mutex_lock(&mutex);
printf ("Starting thread...\n");
pthread_create(&thread, NULL, ThreadFunction, (void*) NULL);
printf ("Waiting to be ready...\n");
while (!ready) {
pthread_cond_wait(&cond, &mutex);
}
printf ("ready!\n");
pthread_mutex_unlock(&mutex);
printf ("Waiting for thread to finish...\n");
pthread_join(thread, NULL);
pthread_mutex_destroy(&mutex);
printf("finished\n");
return 0;
}
$ gcc -g -lpthread -o pthread_cond pthread_cond.c
$ valgrind -q --tool=drd --trace-cond=yes ./pthread_cond
==65261== [1] cond_init cond 0x1002090
Starting thread...
==65261== [1] cond_init cond 0x1ffefff720
==65261== [1] cond_pre_wait cond 0x1ffefff720
==65261== [2] cond_signal cond 0x1ffefff720
==65261== [1] cond_post_wait cond 0x1ffefff720
==65261== [1] cond_destroy cond 0x1ffefff720
Waiting to be ready...
==65261== [1] cond_pre_wait cond 0x1002090
==65261== [2] cond_signal cond 0x1002090
==65261== [1] cond_post_wait cond 0x1002090
ready!
Waiting for thread to finish...
finished
But on RHEL8:
# rpm -q valgrind glibc binutils gcc; uname -r
valgrind-3.14.0-11.el8.s390x
glibc-2.28-53.el8.s390x
binutils-2.30-54.el8.s390x
gcc-8.2.1-3.5.el8.s390x
4.18.0-83.el8.s390x
# valgrind -q --tool=drd --trace-cond=yes ./pthread_cond
==14549== [1] cond_init cond 0x1002090
Starting thread...
==14549== [1] cond_init cond 0x1ffefffd00
==14549== [1] cond_pre_wait cond 0x1ffefffd00
==14549== [2] cond_signal cond 0x1ffefffd00
==14549== [1] cond_post_wait cond 0x4a2a580
==14549== condition variable has been destroyed while being waited upon: cond
0x4a2a580, mutex 0x0 locked by thread 0
==14549== at 0x483C9CE: pthread_cond_wait_intercept
(drd_pthread_intercepts.c:1086)
==14549== by 0x483C9CE: pthread_cond_wait@* (drd_pthread_intercepts.c:1091)
==14549== by 0x4838849: vgDrd_sema_down (drd_pthread_intercepts.c:290)
==14549== by 0x4838849: pthread_create_intercept
(drd_pthread_intercepts.c:573)
==14549== by 0x4838849: pthread_create@* (drd_pthread_intercepts.c:584)
==14549==
==14549== [1] cond_destroy cond 0x1ffefffd00
==14549== destruction of condition variable being waited upon: cond
0x1ffefffd00
==14549== at 0x483C552: pthread_cond_destroy_intercept
(drd_pthread_intercepts.c:1063)
==14549== by 0x483C552: pthread_cond_destroy@*
(drd_pthread_intercepts.c:1073)
==14549== by 0x48387BF: vgDrd_sema_destroy (drd_pthread_intercepts.c:283)
==14549== by 0x48387BF: pthread_create_intercept
(drd_pthread_intercepts.c:576)
==14549== by 0x48387BF: pthread_create@* (drd_pthread_intercepts.c:584)
==14549== cond 0x1ffefffd00 was first observed at:
==14549== at 0x483C0AA: pthread_cond_init_intercept
(drd_pthread_intercepts.c:1022)
==14549== by 0x483C0AA: pthread_cond_init@* (drd_pthread_intercepts.c:1030)
==14549== by 0x483810D: vgDrd_sema_init (drd_pthread_intercepts.c:276)
==14549== by 0x48385D9: pthread_create_intercept
(drd_pthread_intercepts.c:541)
==14549== by 0x48385D9: pthread_create@* (drd_pthread_intercepts.c:584)
==14549==
==14549== Destroying condition variable that is being waited upon: cond
0x1ffefffd00, mutex 0x0 locked by thread 0
==14549== at 0x483C552: pthread_cond_destroy_intercept
(drd_pthread_intercepts.c:1063)
==14549== by 0x483C552: pthread_cond_destroy@*
(drd_pthread_intercepts.c:1073)
==14549== by 0x48387BF: vgDrd_sema_destroy (drd_pthread_intercepts.c:283)
==14549== by 0x48387BF: pthread_create_intercept
(drd_pthread_intercepts.c:576)
==14549== by 0x48387BF: pthread_create@* (drd_pthread_intercepts.c:584)
==14549==
Waiting to be ready...
==14549== [1] cond_pre_wait cond 0x1002090
==14549== [2] cond_signal cond 0x1002090
==14549== [1] cond_post_wait cond 0x1002090
ready!
Waiting for thread to finish...
finished
--
You are receiving this mail because:
You are watching all bug changes.
