fix a bug when Linux kernel is busy

This commit is contained in:
laurent
2018-09-20 16:21:37 +02:00
parent d8472426c4
commit e047162803
@@ -116,15 +116,14 @@ extern "C" {
pthread_mutex_lock (&t->queue_cond_lock); pthread_mutex_lock (&t->queue_cond_lock);
int ret=itti_send_msg_to_task_locked(destination_task_id, instance, message); int ret=itti_send_msg_to_task_locked(destination_task_id, instance, message);
// while ( t->message_queue.size()>0 && t->admin.func != NULL ) while ( t->message_queue.size()>0 && t->admin.func != NULL ) {
if ( t->message_queue.size()>0 && t->admin.func != NULL ) {
if (t->message_queue.size()>1) if (t->message_queue.size()>1)
LOG_W(TMR,"queue in no thread mode is %ld\n", t->message_queue.size()); LOG_W(TMR,"queue in no thread mode is %ld\n", t->message_queue.size());
pthread_mutex_unlock (&t->queue_cond_lock); pthread_mutex_unlock (&t->queue_cond_lock);
t->admin.func(NULL); t->admin.func(NULL);
} else { pthread_mutex_lock (&t->queue_cond_lock);
pthread_mutex_unlock (&t->queue_cond_lock);
} }
pthread_mutex_unlock (&t->queue_cond_lock);
return ret; return ret;
} }
@@ -198,6 +197,8 @@ extern "C" {
pthread_mutex_unlock(&t->queue_cond_lock); pthread_mutex_unlock(&t->queue_cond_lock);
LOG_D(TMR,"enter blocking wait for %s\n", itti_get_task_name(task_id)); LOG_D(TMR,"enter blocking wait for %s\n", itti_get_task_name(task_id));
t->nb_events = epoll_wait(t->epoll_fd,t->events,t->nb_fd_epoll, epoll_timeout); t->nb_events = epoll_wait(t->epoll_fd,t->events,t->nb_fd_epoll, epoll_timeout);
if ( t->nb_events < 0 && (errno == EINTR || errno == EAGAIN ) )
pthread_mutex_lock(&t->queue_cond_lock);
} while (t->nb_events < 0 && (errno == EINTR || errno == EAGAIN ) ); } while (t->nb_events < 0 && (errno == EINTR || errno == EAGAIN ) );
AssertFatal (t->nb_events >=0, AssertFatal (t->nb_events >=0,