[TaskGroup] Reenable test and fix memory issue #67392
Merged
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
|
@swift-ci please smoke test |
al45tair
approved these changes
Jul 19, 2023
ktoso
added
test suite
al45tair
reviewed
Jul 19, 2023
test/Concurrency/Runtime/async_taskgroup_discarding_dontLeak_class_error.swift
Outdated
Show resolved
Hide resolved
ktoso
commented
Jul 19, 2023
test/Concurrency/Runtime/async_taskgroup_discarding_dontLeak_class_error.swift
Outdated
Show resolved
Hide resolved
|
@swift-ci please smoke test |
etcwilde
reviewed
Jul 19, 2023
test/Concurrency/Runtime/async_taskgroup_discarding_dontLeak_class_error.swift
Show resolved
Hide resolved
storing the crash here for reference |
|
@swift-ci please smoke test linux |
ktoso
force-pushed
the
wip-reenable-async_taskgroup_discarding_dontLeak_class_error.swift
branch
from
42f4b8c to
eeb2610
Compare
|
@swift-ci please smoke test linux |
ktoso
requested review from
tshortli,
hborla,
slavapestov and
xedin
as code owners
|
@swift-ci please test |
ktoso
changed the title
ktoso
changed the title
ktoso
changed the title
tomerd
reviewed
Aug 3, 2023
Is the PR description out-of-date? Isn't an important correctness fix in the runtime included? |
Yes, updated to CCC style with all the details from #67700 now |
released memory. Because the waitAll() group method is called before the group returns from withTaskGroup and is used to ensure all tasks have completed before we destroy the task group, this method is then immediately followed by calling TaskGroup::destroy. If we, as previously, first resume the waiting task and then attempt to unlock the lock held by the group, we are in an unsafe situation with racing the task group destroy() and the unlock(). Therefore, we must release the lock before we resume the waiting task.
ktoso
force-pushed
the
wip-reenable-async_taskgroup_discarding_dontLeak_class_error.swift
branch
from
95d5d08 to
d01abcb
Compare
ktoso
changed the title
ktoso
added
concurrency runtime
|
@swift-ci please smoke test |
|
Rebased, this has only the necessary change right now. |
|
@swift-ci please smoke test Windows |
ktoso
changed the title
|
@swift-ci please smoke test Linux (CI/github connection issues) |
|
@swift-ci please smoke test Linux |
1 similar comment
|
@swift-ci please smoke test Linux |
|
@swift-ci please smoke test macOS |
|
Failed Tests (1): |
|
@swift-ci please smoke test macOS |
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.
Description: This re-enables a task group test and fixes a memory issue where we realized we must release the group lock before resuming the waiting task. If we don't do this, we're racing the group's destroy() (coming from the parent task, waiting on waitAll with the unlock()), risking calling unlock() on already released memory - very rarely causing crashes due to this. It is hard to reproduce this issue, due to the race appearing very rarely, but it is consistently reproducible in some environments.
Risk: Low, the fix reorders an unlock() with the waitingTask resume which cannot cause any other consistency issues as there are no other actions taken after the resume. Arguably, this always should have been release + run order.
Reward: Medium, it seems this crash does not manifest frequently in real applications, however it is a real issue we need to fix.
Review by: @al45tair @erik
Testing: CI testing, confirmed the fix on specially prepared environments.
Radar: rdar://112759360 rdar://110474311