Fixed mutex leaking in Linux: pj_mutex_destroy() will fail to destroy the mutex if the mutex is currently being held. Unfortunately it is quite common in the libraries to acquire the mutex first before destroying it, causing mutexes to leak. The solution is to retry unlocking/destroying the mutex several times (currently 4) while pthread_mutex_destroy() returns EBUSY

git-svn-id: http://svn.pjsip.org/repos/pjproject/trunk@708 74dad513-b988-da41-8d7b-12977e46ad98

1 files changed, 14 insertions, 2 deletions

diff --git a/pjlib/src/pj/os_core_unix.c b/pjlib/src/pj/os_core_unix.c
index 2b640cbc..03aa8f86 100644
--- a/pjlib/src/pj/os_core_unix.c
+++ b/pjlib/src/pj/os_core_unix.c
@@ -1101,7 +1101,9 @@ PJ_DEF(pj_status_t) pj_mutex_trylock(pj_mutex_t *mutex)
  */
 PJ_DEF(pj_status_t) pj_mutex_destroy(pj_mutex_t *mutex)
 {
+    enum { RETRY = 4 };
     int status;
+    unsigned retry;
 
     PJ_CHECK_STACK();
     PJ_ASSERT_RETURN(mutex, PJ_EINVAL);
@@ -1109,11 +1111,21 @@ PJ_DEF(pj_status_t) pj_mutex_destroy(pj_mutex_t *mutex)
 #if PJ_HAS_THREADS
     PJ_LOG(6,(mutex->obj_name, "Mutex destroyed by thread %s",
 			       pj_thread_this()->obj_name));
-    status = pthread_mutex_destroy( &mutex->mutex );
+
+    for (retry=0; retry<RETRY; ++retry) {
+	status = pthread_mutex_destroy( &mutex->mutex );
+	if (status == PJ_SUCCESS)
+	    break;
+	else if (retry<RETRY-1 && status == EBUSY)
+	    pthread_mutex_unlock(&mutex->mutex);
+    }
+
     if (status == 0)
 	return PJ_SUCCESS;
-    else
+    else {
+	pj_assert(!"Error destroying pthread_mutex");
 	return PJ_RETURN_OS_ERROR(status);
+    }
 #else
     pj_assert( mutex == (pj_mutex_t*)1 );
     status = PJ_SUCCESS;