Fix various bugs in get/put fallbacks in pml/ob1 #12817
Merged
+30
−21
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
…allback Under a number of circumstances it may be necessary to abandon an RDMA get in ob1. In some cases it falls back to put but it may fall back to using send/recv. If that happens then we may either crash or leak RDMA fragments because they are still attached to the send request. Debug builds will crash due to a check on rdma_frag when they are returned. This CL fixes the flaw by releasing any rdma fragment when sceduling sends. Signed-off-by: Nathan Hjelm <hjelmn@google.com>
If a put or get operation fails it may later be retried by mca_pml_ob1_process_pending_rdma which increments retries on each new attempt. There is a flaw in the code where both the put and get failures also increment this counter leading to it giving up twice as fast. This commit removes the increments on the put and get failures. Signed-off-by: Nathan Hjelm <hjelmn@google.com>
…n failure The mca_pml_ob1_recv_request_get_frag_failed method is responsible for returning or queueing the fragment but mca_pml_ob1_rget_completion was freeing it unconditionally. This will lead to a double return of the fragment to the free list and may lead to other errors if the fragment was queued for retry. This commit fixes the issue by only returning the fragment if it did not fail. Signed-off-by: Nathan Hjelm <hjelmn@google.com>
bosilca
approved these changes
Sep 20, 2024
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Outdated suggestions cannot be applied.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.
These were discovered when using pml/ob1 with btl/uct on a new system. I have not been working on Open MPI in some time so it is hard to tell exactly when these bugs were introduced but they could be quite old. See the commit messages for more details.
Fixes #10545