pybind/rados: fix object lifetime issues and other bugs in aio #7778
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
aio_read: The reference to ret_s begins existing at PyBytes_FromStringAndSize and is handed over to the callback if rados_aio_read succeeds. This creates a lot of subtle scenarios where it might not be XDECREFed (e.g. if a KeyboardInterrupt arrives at the wrong time). Instead, store the pointer to that buffer in the completion object, and hand over responsibility for the XDECREF to it. This guarantees that the "special" reference to this incomplete object will be released when the completion object is deallocated. Also make sure we don't try to _PyBytes_Resize with a negative length. Add a failure case to the aio_read test in test_rados.py Completion: the wrapper methods weren't being called, which prevents the completion objects from being freed until the Ioctx is. Fix this and add a refcount check to the aio_read test. Signed-off-by: Hector Martin <marcan@marcan.st>
Tracking completions is critical for memory safety - if the aio function succeeds, the completion must be tracked. However, if a KeyboardInterrupt or similar arrives between the call and the tracking, the completion will not be tracked. Fix this by tracking the completion before the aio call, and explicitly cleaning up in the failure case. This leaves the opposite problem, where an unexpected exception (other than simple error return from the aio function) will cause the completion to not be freed until the Ioctx is destroyed, but that is a relatively minor issue. Signed-off-by: Hector Martin <marcan@marcan.st>
|
That looks good. |
|
Looks good to me as well. Thanks for the detailed explanations, they make these sorts of fixes much easier to read. |
jdurgin
added a commit
that referenced
this pull request
Mar 1, 2016
pybind/rados: fix object lifetime issues and other bugs in aio Reviewed-by: Mehdi Abaakouk <sileht@redhat.com> Reviewed-by: Josh Durgin <jdurgin@redhat.com>
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 fixes some subtle and some not so subtle object lifetime issues in the aio completion implementation in rados.pyx.
WRT the comment at the end of the second commit, where we still have a case of completion objects living longer than they should if an exception arrives at the wrong time (not likely to be a problem in practice, but worth fixing): I'm working on aio support for the rbd bindings, and am trying out a different approach to keeping the completion objects alive. If that works out I'll switch the rados bindings to the same approach and that should fix the problem.
cc @jdurgin @sileht