Bug 79526
| Summary: | A memory leak in pthread_rwlock_init() | ||
|---|---|---|---|
| Product: | Base System | Reporter: | Kimmo Mustonen <tzimmo> |
| Component: | threads | Assignee: | freebsd-threads (Nobody) <threads> |
| Status: | Closed FIXED | ||
| Severity: | Affects Only Me | ||
| Priority: | Normal | ||
| Version: | 5.3-RELEASE | ||
| Hardware: | Any | ||
| OS: | Any | ||
On Mon, Apr 04, 2005 at 03:36:02PM +0300, Kimmo Mustonen wrote: > Haven't looked at the sources, but probably easy to fix. > I suppose pthread_rwlock_destroy() just does not free all the > resources pthread_rwlock_init() allocates. Can you try this patch? prwlock was being allocated by malloc inside pthread_rwlock_init(), but never freed. --- lib/libpthread/thread/thr_rwlock.c.orig Mon Apr 4 16:22:16 2005 +++ lib/libpthread/thread/thr_rwlock.c Mon Apr 4 16:23:26 2005 @@ -131,6 +131,7 @@ prwlock->blocked_writers = 0; *rwlock = prwlock; + free(prwlock); } } } -- Craig Rodrigues rodrigc@crodrigues.org On Mon, 4 Apr 2005, Craig Rodrigues wrote:
> Can you try this patch?
> prwlock was being allocated by malloc inside pthread_rwlock_init(),
> but never freed.
> *rwlock = prwlock;
> + free(prwlock);
First, this patch is broken. It sets the pointer to a datastructure but
frees the contents of the structure immediately after that, later causing
referencing to deallocated memory and crashing.
The valgrind version did not crash but still leaked memory after this
(broken) patch. I didn't realise valgrind has its own pthread library.
Thus, the problem is most probably in the valgrind implementation of
pthread library and the actual pthread library does not leak. So I don't
think there's a need for a patch in pthread.
I hope this problem will disappear when the new valgrind using native
pthread library is ported to FreeBSD.
Regards,
Kimmo Mustonen
State Changed From-To: open->closed It is not a bug of thread library, it should be fixed in valgrind ports. |
Calling pthread_rwlock_init() leaks memory each time it is called. Example code ---8<---8<--- #include <pthread.h> /* start using valgrind: valgrind --tool=memcheck -v --leak-check=yes --show-reachable=yes \ --num-callers=100 ./leak_pthread_1 */ int main(void) { int i; pthread_rwlock_t lock; for (i = 0; i < 100; i++) { pthread_rwlock_init(&lock, NULL); pthread_rwlock_destroy(&lock); } return 0; } ---8<---8<--- and valgrind report ---8<---8<--- ==801== 2012 bytes in 301 blocks are still reachable in loss record 1 of 1 ==801== at 0x3C033183: malloc (in /usr/local/lib/valgrind/vgpreload_memcheck.so) ==801== by 0x3C03A8D0: (within /usr/local/lib/valgrind/libpthread.so.1) ==801== by 0x3C03BF49: pthread_mutex_init (in /usr/local/lib/valgrind/libpthread.so.1) ==801== by 0x3C03BFA9: pthread_mutex_lock (in /usr/local/lib/valgrind/libpthread.so.1) ==801== by 0x3C03F174: (within /usr/local/lib/valgrind/libpthread.so.1) ==801== by 0x3C03F2D9: pthread_rwlock_init (in /usr/local/lib/valgrind/libpthread.so.1) ==801== by 0x8048609: main (leak_pthread_1.c:14) ==801== ---8<---8<--- Fix: Haven't looked at the sources, but probably easy to fix. I suppose pthread_rwlock_destroy() just does not free all the resources pthread_rwlock_init() allocates. How-To-Repeat: Run the example code using valgrind.