btl/ugni: actually make the endpoint lock recursive

[hjelmn]

Signed-off-by: Nathan Hjelm hjelmn@lanl.gov

(cherry picked from commit open-mpi/ompi@83062db)

Signed-off-by: Nathan Hjelm hjelmn@lanl.gov

[hjelmn]

:bot🏷️bug
:bot:milestone:v2.0.1
:bot:assign: @hppritcha

Missed this when transcribing the patch.

[jsquyres]

@hjelmn What bug does this solve?

[hjelmn]

A bug I introduced. If MPI_THREAD_MULTIPLE is on it deadlocks on the first connection. Couldn't test after transcribing because none of our open Cray systems were up. Hit the bug immediately once I got to test.

[hjelmn]

Half of the fix made it. See https://github.com/open-mpi/ompi-release/blob/v2.x/opal/mca/btl/ugni/btl_ugni_endpoint.h#L37

[hjelmn]

Also, this does not require an rc2 IMO.

[lanl-ompi]

Test FAILed.

[lanl-ompi]

Test FAILed.

[jsquyres]

@hppritcha Assuming you can review, I'm ok with this.

[hjelmn]

LANL jenkins blew up.

:bot:lanl:retest

[lanl-ompi]

Test FAILed.

[lanl-ompi]

Test FAILed.

[hjelmn]

@hppritcha LANL jenkins looks like it is busted.

[mellanox-github]

Test FAILed.
See http://bgate.mellanox.com/jenkins/job/gh-ompi-release-pr/2122/ for details.

[hjelmn]

And now the mellanox Jenkins is belly-up.

00:48:21.114 --2016-08-24 20:56:33--  (try: 2)  http://www.mcs.anl.gov/~thakur/thread-tests/thread-tests-1.1.tar.gz
00:48:21.114 Connecting to www.mcs.anl.gov (www.mcs.anl.gov)|140.221.6.95|:80... failed: Connection timed out.
00:50:28.370 Retrying.

[hjelmn]

bot:mellanox:retest
bot:lanl:retest

[lanl-ompi]

Test FAILed.

[mellanox-github]

Test PASSed.
See http://bgate.mellanox.com/jenkins/job/gh-ompi-release-pr/2124/ for details.

[hppritcha]

bot:lanl:retest

[lanl-ompi]

Test FAILed.

[lanl-ompi]

Test FAILed.

[hppritcha]

UH network problems hitting of some sort.
@edgargabriel if you have the chance could you see if the cluster at UH is having external connectivity issues?

[hppritcha]

lets wait for CI modulo the mess up at UH then merge.
👍

[edgargabriel]

I can login onto the node without issues, but I know that there were some network issues the last two days.

[jsquyres]

bot:ibm:retest

[hppritcha]

@edgargabriel says UH is messed up while school starts up so we'll ignore dlopen and distcheck tests for this PR.

[edgargabriel]

not objecting to the decision, but I am still a bit surprised that it 'hangs' since both of my pr's that I filed today got through that point without major issues.

[hppritcha]

bot:lanl:retest
@jsquyres i vote for merging in and seeing how mtt goes over the weekend