erobot opened a new pull request #8765: URL: https://github.com/apache/pulsar/pull/8765
### Motivation BlockingQueue has race condition that can cause threads waiting forever in multithreading environment. ProducerImpl uses BlockingQueue as pendingMessagesQueue_ and can be blocked forever at it. This PR fixes race condition in BlockingQueue. #### Race condition details https://github.com/apache/pulsar/blob/91e2f832178d9ffd5d78161145d895910296c2d9/pulsar-client-cpp/lib/BlockingQueue.h#L172-L185 Use BlockingQueue::Pop as example, its procedure is: 1. lock 2. check wasFull and then change queue state 3. unlock 4. if wasFull, notify one thread waiting at queueFullCondition Race condition sequence: 1. queue is full and there are multiple threads waiting on queueFullCondition 2. thread A call Pop, lock, wasFull is true, unlock -> queue has one free space 3. thread B call Pop, lock, wasFull is false, unlock -> queue has two free spaces 4. thread A notify one thread waiting at queueFullCondition 5. queue is no loger full again 6. result: except one thread is notified by A, other threads waiting on queueFullCondition are waiting forever ### Modifications * Use notify_all instead of notify_one to notify threads waiting on condition variables Reason: Currently only notify threads when queue is full or empty. After unlock, other threads may change queue state, so thread to notify condition can not determine how queue state changed and should use notify_all in case of more then one change occured. ### Verifying this change - Add a test case BlockingQueueTest.testPushPopRace to test concurrent push and pop ### Does this pull request potentially affect one of the following parts: - Dependencies (does it add or upgrade a dependency): (no) - The public API: (no) - The schema: (no) - The default values of configurations: (no) - The wire protocol: (no) - The rest endpoints: (no) - The admin cli options: (no) - Anything that affects deployment: (no) ### Documentation - Does this pull request introduce a new feature? (no) ---------------------------------------------------------------- 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]
