Fix UB in cancelled tasks #578
Merged
Conversation
This file contains 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
This diff fixes #577 by remembering which continuation _would_ have been resumed if the deferred stop request had completed first. Later, when the deferred stop request completes, resume the remembered continuation (which might be `unhandled_done()`) instead of uncondtionally resuming the primary continuation.
facebook-github-bot
added
the
CLA Signed
janondrusek
approved these changes
Oct 25, 2023
ispeters
added a commit
that referenced
this pull request
Oct 26, 2023
As reported in #577, #578 didn't fix the UB in cancelled tasks. There were two bugs in that "fix": 1. we were writing the coroutine handle to resume *after* synchronizing between the two racing operation states, leading to a data race when the deferred stop source completed second; and 2. it's wrong to call `continuation_.done()` to store the coroutine handle to be resumed later because `done()` completes the operation with `set_done()` as a side effect. This diff fixes both issues. Instead of storing a coroutine handle to be resumed, we store an enum value that describes which coroutine handle to resume (so we don't eagerly invoke `done()`), and we do the store before synchronizing to eliminate the data race. I've added a unit test that fails with a TSAN-detected data race without the fix, and the fix silences the TSAN error. I also clang-formatted `task.hpp` and the modified test file.
ispeters
added a commit
that referenced
this pull request
Oct 27, 2023
As reported in #577, #578 didn't fix the UB in cancelled tasks. There were two bugs in that "fix": 1. we were writing the coroutine handle to resume *after* synchronizing between the two racing operation states, leading to a data race when the deferred stop source completed second; and 2. it's wrong to call `continuation_.done()` to store the coroutine handle to be resumed later because `done()` completes the operation with `set_done()` as a side effect. This diff fixes both issues. Instead of storing a coroutine handle to be resumed, we store an enum value that describes which coroutine handle to resume (so we don't eagerly invoke `done()`), and we do the store before synchronizing to eliminate the data race. I've added a unit test that fails with a TSAN-detected data race without the fix, and the fix silences the TSAN error. I also clang-formatted `task.hpp` and the modified test file.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This diff fixes #577 by remembering which continuation would have been resumed if the deferred stop request had completed first. Later, when the deferred stop request completes, resume the remembered continuation (which might be
unhandled_done()) instead of uncondtionally resuming the primary continuation.