Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Import several uct btl updates #12839

Merged
merged 7 commits into from
Oct 3, 2024
Merged

Import several uct btl updates #12839

merged 7 commits into from
Oct 3, 2024

Commits on Oct 1, 2024

  1. pml/ob1: fix potential double return of RDMA fragment on get operatio… 

    …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 <[email protected]>
(cherry picked from commit b7f8cae)
    @hjelmn @bosilca
    hjelmn authored and bosilca committed Oct 1, 2024
    Configuration menu
    Copy the full SHA
    1f5ec3e View commit details
    Browse the repository at this point in the history

  2. pml/ob1: fix double increment of the RDMA frag retry counter 

    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 <[email protected]>
(cherry picked from commit 27efeb9)
    @hjelmn @bosilca
    hjelmn authored and bosilca committed Oct 1, 2024
    Configuration menu
    Copy the full SHA
    40c290e View commit details
    Browse the repository at this point in the history

  3. pml/ob1: ensure RDMA fragments are released in the get -> send/recv f… 

    …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 <[email protected]>
(cherry picked from commit 020e83f)
    @hjelmn @bosilca
    hjelmn authored and bosilca committed Oct 1, 2024
    Configuration menu
    Copy the full SHA
    25862a9 View commit details
    Browse the repository at this point in the history

  4. btl/uct: fix fetching atomics support for osc/rdma 

    The check that requires 32 and 64 bit atomic support is way outdated at this
point. This commit takes the union of the two instead of the intersection
because technically it is up to the btl user to determine which of each type
are supported. This commit also ensures that MCA_BTL_ATOMIC_SUPPORTS_32BIT is
set when 32-bit atomics are available and ensures that the required
MCA_BTL_FLAGS_RDMA_REMOTE_COMPLETION flag is set on the btl.

Signed-off-by: Nathan Hjelm <[email protected]>
(cherry picked from commit 8adf9f5)
    @hjelmn @bosilca
    hjelmn authored and bosilca committed Oct 1, 2024
    Configuration menu
    Copy the full SHA
    18abef5 View commit details
    Browse the repository at this point in the history

  5. Cleanup unused code. 

    Signed-off-by: George Bosilca <[email protected]>
(cherry picked from commit 58400ad)
    @bosilca
    bosilca committed Oct 1, 2024
    Configuration menu
    Copy the full SHA
    3d05ccc View commit details
    Browse the repository at this point in the history

  6. Allow for packing less data than expected. 

    Return the updated amount to allow the upper level to gracefully handle
the case.

Signed-off-by: George Bosilca <[email protected]>
(cherry picked from commit 27514c2)
    @bosilca
    bosilca committed Oct 1, 2024
    Configuration menu
    Copy the full SHA
    8689370 View commit details
    Browse the repository at this point in the history

  7. Return faster if nothing to do. 

    Signed-off-by: George Bosilca <[email protected]>
(cherry picked from commit f29109a)
    @bosilca
    bosilca committed Oct 1, 2024
    Configuration menu
    Copy the full SHA
    15cd497 View commit details
    Browse the repository at this point in the history