oneapi-src / oneapi-src/unified-memory-framework
Race Condition Between Thread Destructor and Pool Destroy in Scalable Pool
Open
Nobody has claimed this yet.
bug
high priority
- Dominant language
- C
- Stars
- 98
- Forks
- 48
- Avg merge
- 3d 20h
- Merged PRs (30d)
- 6
Description
A race condition occurs when the scalable pool is destroyed while another thread is performing Thread-Local Storage (TLS) destruction. This issue arises only if one thread is in its TLS destructor and another thread calls umfPoolDestroy. There is no problem when a thread is destructed after the pool has been destroyed.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by tracing scalable pool destruction through umfPoolDestroy and the thread-local storage destructor path. Reproduce concurrent pool destruction and thread teardown, then verify that the race is eliminated without breaking normal destruction after the pool has been destroyed.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- c
- Domain
- backend
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100