btl/openib: fix rdmacm locking bug #1837
Conversation
|
@jladd-mlnx The rdmacm fix exposed this long standing bug. Caught by Jenkins. |
hjelmn
force-pushed
the
rdmacm_fix
branch
2 times, most recently
from
bbb321c to
6dcb09e
Compare
|
Build Failed with XL compiler! Please review the log, and get in touch if you have questions. |
This commit fixes a long standing bug in rdmacm. It is required that the thread that calls mca_btl_openib_endpoint_cpc_complete holds the endpoint lock. This was not the case for rdmacm. This causes debug builds to abort. This change also required changing mca_btl_openib_endpoint_send_cts to require the endpoint lock to be held when calling. Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
|
Build Failed with XL compiler! Please review the log, and get in touch if you have questions. Gist: https://gist.github.com/95bdd56fe7f803ed1ecde6bc9a777222 |
|
@jjhursey Failure seems unrelated to this PR: |
|
Build Failed with GNU compiler! Please review the log, and get in touch if you have questions. Gist: https://gist.github.com/2d03072a40e4bff2c07b7a8053ee4365 |
|
Yeah I think that something odd on the machine. I'm turning off the 'run example' part of the CI test for now, while I diagnose. Let's run that again and see if it passes. bot:ibm:retest |
|
Got past rdmacm! Lost it on the threading bug. |
|
Build Failed with XL compiler! Please review the log, and get in touch if you have questions. Gist: https://gist.github.com/ibm-ompi/2746fecf92fda4d80ff1b80b23edbfc7 |
3 participants
This commit fixes a long standing bug in rdmacm. It is required that
the thread that calls mca_btl_openib_endpoint_cpc_complete holds the
endpoint lock. This was not the case for rdmacm. This causes debug
builds to abort.
Signed-off-by: Nathan Hjelm hjelmn@lanl.gov