fs/epoll: release refs count and add TLS cleanup handler to release epoll fd reference on thread exit #17741
+44
−0
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…close When an epoll fd is closed, the file descriptors in the teardown and oneshot lists were not being properly dereferenced, leading to fd leaks. This fix ensures that all fds in the teardown and oneshot lists are properly released via file_put() during epoll_do_close(), matching the behavior for fds in the setup list. Impact: - Prevents fd leaks when epoll fd is closed - Ensures proper cleanup of all tracked file descriptors - Critical for applications using EPOLLONESHOT or fd removal operations Signed-off-by: dongjiuzhu1 <[email protected]>
…read exit When a thread is terminated via pthread_exit() while blocked in epoll_wait(), the file reference taken at the beginning of epoll_wait() is not properly released, leading to resource leaks. Problem scenario found during libuv test: 1. Echo server thread is blocked in epoll_wait() 2. Main task sends pthread_kill signal to the server thread 3. Signal handler calls pthread_exit() to terminate the thread 4. epoll_wait() is interrupted before reaching the file_put() call 5. The epoll fd reference count remains elevated 6. epoll_do_close() is never called, leaving fds in internal queues 7. File descriptors leak Solution: Register a TLS (Thread Local Storage) cleanup handler using tls_cleanup_push() at the beginning of epoll_wait() blocking section. This ensures that if the thread exits abnormally (via pthread_exit, pthread_cancel, etc.), the cleanup handler (epoll_cleanup) will be called automatically to release the file reference via file_put(). The cleanup handler is properly paired with tls_cleanup_pop() when epoll_wait() completes normally, ensuring the handler is only invoked on abnormal exit. This fix is applied to both epoll_wait() code paths (with and without extended mode) to ensure consistent behavior. Impact: - Prevents epoll fd reference count leaks on thread cancellation - Ensures proper cleanup even when epoll_wait() is interrupted by pthread_exit - Critical for multi-threaded applications using signals and thread termination - Works together with previous fix for teardown/oneshot list cleanup Signed-off-by: dongjiuzhu1 <[email protected]>
github-actions
bot
added
Area: File System
xiaoxiang781216
approved these changes
Dec 30, 2025
acassis
approved these changes
Dec 31, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Note: Please adhere to Contributing Guidelines.
Summary
patch 1:
When an epoll fd is closed, the file descriptors in the teardown and
oneshot lists were not being properly dereferenced, leading to fd leaks.
This fix ensures that all fds in the teardown and oneshot lists are
properly released via file_put() during epoll_do_close(), matching the
behavior for fds in the setup list.
patch 2
When a thread is terminated via pthread_exit() while blocked in epoll_wait(),
the file reference taken at the beginning of epoll_wait() is not properly
released, leading to resource leaks.
Impact
patch1
patch2
Testing
uv_run_test tcp_ping_pong
