Remove support for not creating local MRs #796
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Outdated suggestions cannot be applied.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.
Commits 27548d4 and a1ac0f5 fixed providers that required memory registrations when FI_MR_LOCAL was set, but also broke providers that clear FI_MR_LOCAL (such as HPE's provider), because I did not account for the mr_local handling in the send/recv transport.
We don't have a great way to test that case from AWS, the vast majority of transfers will either be HMEM (which creates an MR anyway) or control messages (which have a freelist which creates an MR), and the Libfabric specification allows passing an MR descriptor to a local operation even if the provider clears the FI_MR_LOCAL bit. Therefore, the best path forward seems to be removing the code to skip registration if FI_MR_LOCAL is cleared, and always creating an MR.
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.