canonical-ubuntu-qa team mailing list archive

[Ian Whitfield <2114518@xxxxxxxxxxxxxxxxxx>]

Public bug reported:

net:busy_poll_test.sh fails on seemingly any GCE instance. It's a kernel
selftest that was added with Plucky, and we're seeing it since the first
Plucky tests on GCE, therefore this is not a regression.

plucky:linux-gcp, noble:linux-gcp-6.14, noble:linux-hwe-6.14 all have
the same test failure where the test will time out after 1m30s. On AWS,
the noble:linux-gcp-6.14 and noble:linux-hwe-6.14 kernels pass the test
in <1 second.

Test output is as follows:
21:44:58 DEBUG| [stdout] # timeout set to 0
21:44:58 DEBUG| [stdout] # selftests: net: busy_poll_test.sh
21:47:13 DEBUG| [stdout] # 2025/06/09 21:47:13 socat[1106646] E connect(8, AF=2 192.168.1.1:48675, 16): Connection timed out
21:47:13 DEBUG| [stdout] # md5sum mismatch
21:47:13 DEBUG| [stdout] # input file md5sum: 00987e5988639ce2decb7b5757c7bc63
21:47:13 DEBUG| [stdout] # output file md5sum: d41d8cd98f00b204e9800998ecf8427e
21:47:13 DEBUG| [stdout] # test_busypoll failed
21:47:13 DEBUG| [stdout] not ok 1 selftests: net: busy_poll_test.sh # exit=1 

The md5sum is taken to verify that a file has completed transfer over a socket, the timeout indicates that the socket dies during or before the transfer.
The line in the busy_poll_test.sh script that times out is:
$ ip netns exec nscl socat -u /tmp/tmp.lFmRMmbet3 TCP:192.168.1.1:48675
Elixir link: https://elixir.bootlin.com/linux/v6.15.1/source/tools/testing/selftests/net/busy_poll_test.sh#L84

Because this issue is present across multiple kernels, and is limited to
GCE hardware, and is not a regression, it can be hinted for now until
the underlying problem is resolved.

** Affects: ubuntu-kernel-tests
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of Canonical
Platform QA Team, which is subscribed to ubuntu-kernel-tests.
https://bugs.launchpad.net/bugs/2114518

Title:
  ubuntu_kselftests_net.net:busy_poll_test.sh fails on Google cloud
  instances

Status in ubuntu-kernel-tests:
  New

Bug description:
  net:busy_poll_test.sh fails on seemingly any GCE instance. It's a
  kernel selftest that was added with Plucky, and we're seeing it since
  the first Plucky tests on GCE, therefore this is not a regression.

  plucky:linux-gcp, noble:linux-gcp-6.14, noble:linux-hwe-6.14 all have
  the same test failure where the test will time out after 1m30s. On
  AWS, the noble:linux-gcp-6.14 and noble:linux-hwe-6.14 kernels pass
  the test in <1 second.

  Test output is as follows:
  21:44:58 DEBUG| [stdout] # timeout set to 0
  21:44:58 DEBUG| [stdout] # selftests: net: busy_poll_test.sh
  21:47:13 DEBUG| [stdout] # 2025/06/09 21:47:13 socat[1106646] E connect(8, AF=2 192.168.1.1:48675, 16): Connection timed out
  21:47:13 DEBUG| [stdout] # md5sum mismatch
  21:47:13 DEBUG| [stdout] # input file md5sum: 00987e5988639ce2decb7b5757c7bc63
  21:47:13 DEBUG| [stdout] # output file md5sum: d41d8cd98f00b204e9800998ecf8427e
  21:47:13 DEBUG| [stdout] # test_busypoll failed
  21:47:13 DEBUG| [stdout] not ok 1 selftests: net: busy_poll_test.sh # exit=1 

  The md5sum is taken to verify that a file has completed transfer over a socket, the timeout indicates that the socket dies during or before the transfer.
  The line in the busy_poll_test.sh script that times out is:
  $ ip netns exec nscl socat -u /tmp/tmp.lFmRMmbet3 TCP:192.168.1.1:48675
  Elixir link: https://elixir.bootlin.com/linux/v6.15.1/source/tools/testing/selftests/net/busy_poll_test.sh#L84

  Because this issue is present across multiple kernels, and is limited
  to GCE hardware, and is not a regression, it can be hinted for now
  until the underlying problem is resolved.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu-kernel-tests/+bug/2114518/+subscriptions