ubuntu-kernel-tests

IPv4 rp_filter tests in net:fib_tests.sh failed on F-5.4

Bug #2019858 reported by Po-Hsu Lin
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
ubuntu-kernel-tests
New
Undecided
Unassigned

Bug Description

Test failed with:
# IPv4 rp_filter tests
 # TEST: rp_filter passes local packets [FAIL]
 # TEST: rp_filter passes loopback packets [FAIL]

This test can be run with:

  sudo ./fib_tests.sh -t rp_filter

This issue does not exist in the last cycle. So it looks like a regression to me.

It can be found on the following kernel in sru-20230417:
* f/aws 5.4.0-1102.110
* f/azure 5.4.0-1108.114
* f/gke 5.4.0-1099.106
* f/gkeop 5.4.0-1069.73
* f/nv-dgx 5.4.0-1029.32

Within the same cycle, this issue does not exist on the following:
* f/linux 5.4.0-149.166 (test passed)
* f/linux 5.4.0-150.167 (test passed)
* f/aws-fips 5.4.0-1102.110+fips1 (bug 2015440 still exist, test skipped)
* f/azure-fips 5.4.0-1108.114+fips1 (bug 2015440 still exist, test skipped)
* f/fips 5.4.0-1077.86 (test passed)
* f/gcp 5.4.0-1105.114 (test passed)
* f/gcp-fips 5.4.0-1105.114+fips1 (test passed)
* f/ibm 5.4.0-1049.54 (test passed)
* f/kvm 5.4.0-1091.97 (unable to test)
* f/oracle 5.4.0-1101.110 (test passed)

See original description
Revision history for this message
Po-Hsu Lin (cypressyew) wrote :

It's because the fib_tests was skipped due to:
https://bugs.launchpad.net/ubuntu-kernel-tests/+bug/2015440

I will modify title and content to not calling it a regression.

summary: - [Potential Regression] IPv4 rp_filter tests in net:fib_tests.sh failed
- on F-5.4
+ IPv4 rp_filter tests in net:fib_tests.sh failed on F-5.4
Po-Hsu Lin (cypressyew)
description: updated
description: updated
Revision history for this message
Po-Hsu Lin (cypressyew) wrote :

Looking back into test history, this test has been removed from sru-20230130 (taking F-aws 5.4.0-1097 for example, the test has been removed since that cycle), in sru-20230102 F-aws 5.4.0-1095.103 passed with this test.

So it looks like there is something wrong between 5.4.0-1095.103 and 5.4.0-1102.110, but because of bug 2015440 I was unable to determine on which point this test has been regressed.

Next step:
Test F-aws in -updates with fix for bug 2015440 applied manually to make sure this is not a regression this cycle.

Po-Hsu Lin (cypressyew)
description: updated
Revision history for this message
Po-Hsu Lin (cypressyew) wrote :

After some debugging work, Jesse found the root cause is because the fib_tests.sh was removed in the closing commit.

Take f/aws for example, it's been removed in:

commit f16d2ff29d31782255c5fe04cf374303298f5065
Author: Tim Gardner <email address hidden>
Date: Sun Apr 23 17:26:31 2023 -0600

    UBUNTU: Ubuntu-aws-5.4.0-1102.110

    Signed-off-by: Tim Gardner <email address hidden>

And for f/azure:
commit e5cd9600405416217de67ab2be410f1077732c71
Author: Tim Gardner <email address hidden>
Date: Mon Apr 24 09:58:24 2023 -0600

    UBUNTU: Ubuntu-azure-5.4.0-1108.114

    Signed-off-by: Tim Gardner <email address hidden>

For f/iot:
commit 292bf55da78f34f7213f5417c8dead0c0ed9b964
Author: Masahiro Yamada <email address hidden>
Date: Wed May 10 16:04:07 2023 +0900

    UBUNTU: Ubuntu-iot-5.4.0-1015.17

    Signed-off-by: Masahiro Yamada <email address hidden>

So when we're testing this, it will be tested with a very old version of fib_tests.sh that exists in the kernel tarball and thus failing like this.

Revision history for this message
Po-Hsu Lin (cypressyew) wrote :

To provide more detailed information about the history of this fib_tests.sh test:

The test was removed in commit 68f18b30 since Ubuntu-5.4.0-143.160
The test was added back in commit a839003a but lack of executable bit since Ubuntu-5.4.0-149.166
The executable bit was added back in commit c9b09cc1 since Ubuntu-5.4.0-149.166

Revision history for this message
Po-Hsu Lin (cypressyew) wrote :

This is hitting focal/linux-gkeop/5.4.0-1076.80 but it's good with focal/linux-gkeop/5.4.0-1078.82

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.
Subscribing...

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.