lyind commented on a change in pull request #257: GUACAMOLE-958: Avoid
race/deadlock in guacd_timed_client_free()
URL: https://github.com/apache/guacamole-server/pull/257#discussion_r378807355
##########
File path: src/guacd/proc.c
##########
@@ -267,24 +267,24 @@ static int guacd_timed_client_free(guac_client* client,
int timeout) {
.tv_nsec = current_time.tv_usec * 1000
};
- /* Free the client in a separate thread, so we can time the free operation
*/
- if (pthread_create(&client_free_thread, NULL,
- guacd_client_free_thread, &free_operation))
- return 1;
-
/* The mutex associated with the pthread conditional and flag MUST be
* acquired before attempting to wait for the condition */
if (pthread_mutex_lock(&free_operation.completed_mutex))
return 1;
- /* Wait a finite amount of time for the free operation to finish */
- if (pthread_cond_timedwait(&free_operation.completed_cond,
- &free_operation.completed_mutex, &deadline))
- return 1;
+ /* Free the client in a separate thread, so we can time the free operation
*/
+ if (!pthread_create(&client_free_thread, NULL,
+ guacd_client_free_thread, &free_operation)) {
+
+ /* Wait a finite amount of time for the free operation to finish */
+ (void) pthread_cond_timedwait(&free_operation.completed_cond,
Review comment:
I also noted that statement on the man page and thought about it carefully.
I think that the while-loop is not necessary for our use-case:
`pthread_mutex_lock()` (and a few other functions) are implicit memory
barriers which will enforce synchronization of the state variable
(`free_operation.completed`).
Thus I chose to avoid introducing unnecessary code complexity.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
