Fix segmentation fault on canceled requests that are not awaited #153
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
Calling `InflightRequests::cancelAll()` does not guarantee the requests to be canceled immediately, just like any other request we must check their status before releasing the `ucxx::Request` object. To prevent the user from releasing a request that has not yet completed and may still call the completion callback we must make sure we still keep a reference to it until its status is set. With this change we now ensure both `ucxx::Endpoint` and `ucxx::Worker` will only release references after those requests that have issued cancelation complete.
It is important that all requests have a valid status before objects can be destroyed, thus we should cancel it if all references to the Cython `UCXRequest` object have been dropped but the request has not completed yet.
pentschev
added
bug
wence-
reviewed
Jan 8, 2024
Although not perfect, the new name is closer to the `size()` method name.
|
Thanks for reviewing this one too, @wence- ! |
|
/merge |
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Outdated suggestions cannot be applied.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.
Calling
InflightRequests::cancelAll()does not guarantee the requests to be canceled immediately, just like any other request we must check their status before releasing theucxx::Requestobject. To prevent the user from releasing a request that has not yet completed and may still call the completion callback we must make sure we still keep a referenceto it until its status is set. With this change we now ensure both
ucxx::Endpointanducxx::Workerwill only release references after those requests that have issued cancelation complete.It is also important that all requests have a valid status before destroying the object, thus we should cancel a request if all references to the Cython
UCXRequestobject have been dropped but the request has not completed yet.Additionally use
std::unique_ptr<InflightRequests>inucxx::Workerand reenabletest_ucxx_unreachable.