test-configs: Enable landlock selftests #1236
Conversation
@broonie I don't see it, only Le Potato |
|
Thanks Mark! |
config/core/test-configs.yaml
Outdated
| @@ -2061,6 +2067,7 @@ test_configs: | |||
| - baseline-cip-nfs | |||
| - baseline-nfs | |||
| - kselftest-alsa | |||
| - kselftest-landlock | |||
There was a problem hiding this comment.
Ah yes I see this now, sorry. ☕
|
This is currently conflicting with other PRs that need to get merged first, then this one will need to be rebased. |
broonie
force-pushed
the
landlock
branch
2 times, most recently
from
d4d4ab2 to
4300889
Compare
|
What is the status of this PR? I guess it conflicts with most new tests added in the test-configs.yaml file and it will continue until it is merged. |
|
We need to reduce the backlog of older PRs first as some are enabling tests in the same part of the YAML files and should be ready to be merged. |
|
I can't tell what else this is conflicting with? |
|
Well, if that KVM one is blocked on the Baylibre lab it would be good to let this one progress given that there should be fewer blockers. Though I think we could probably merge the KVM one TBH. |
|
Yes I think we can merge it, the BayLibre issue is not a blocker but I would have liked some feedback as to why the jobs are only failing in their lab. Like I put in another comment, if it can't be fixed we can add a filter to stop running it in the BayLibre lab. |
|
Great, thanks. FWIW the error looks like what you see if the NFS server is configured to support v4 but you don't jump through whatever extra hoops v4 auth requires - I limit my NFS server to v3 for similar reasons. |
@montjoie this ^ |
config/core/test-configs.yaml
Outdated
| @@ -2539,6 +2546,7 @@ test_configs: | |||
| - kselftest-cpufreq | |||
| - kselftest-futex | |||
| - kselftest-kvm | |||
| - kselftest-landlock | |||
There was a problem hiding this comment.
Looks like there's some problem:
https://storage.staging.kernelci.org/kernelci/staging-mainline/staging-mainline-20220824.1/arm64/defconfig+kselftest/gcc-10/lab-broonie/kselftest-landlock-meson-gxl-s905x-libretech-cc.html#L1802
The landlock collection is missing from the kselftest tarball, so there's no results.
Also the jobs are all failing now, they could complete until the end of August.
There was a problem hiding this comment.
There's a landlock build issue:
make[3]: Entering directory '/tmp/kci/linux/tools/testing/selftests/landlock'
make[3]: *** No rule to make target '/tmp/kci/linux/build/kselftest/landlock/base_test', needed by 'all'. Stop.
make[3]: Leaving directory '/tmp/kci/linux/tools/testing/selftests/landlock'
There was a problem hiding this comment.
It looks like the job failures aren't specific to landlock - it looks like all arm64 boots with the kselftest custom kernel broke in the same way starting at the same time. The build issue is still a problem of course.
There was a problem hiding this comment.
There's a landlock build issue:
make[3]: Entering directory '/tmp/kci/linux/tools/testing/selftests/landlock' make[3]: *** No rule to make target '/tmp/kci/linux/build/kselftest/landlock/base_test', needed by 'all'. Stop. make[3]: Leaving directory '/tmp/kci/linux/tools/testing/selftests/landlock'
I still cannot reproduce this build issue. Where can we find these logs and the related artifacts?
There was a problem hiding this comment.
Not the same job but the error showing up in:
https://storage.kernelci.org/next/master/next-20220908/arm64/defconfig/gcc-10/logs/kselftest.log
is
make[3]: Entering directory '/tmp/kci/linux/tools/testing/selftests/landlock'
make[3]: *** No rule to make target '/tmp/kci/linux/build/kselftest/landlock/base_test', needed by 'all'. Stop.
make[3]: Leaving directory '/tmp/kci/linux/tools/testing/selftests/landlock'
There was a problem hiding this comment.
Can you confirm that it works @gctucker? Getting your Tested-by tag would help: https://lore.kernel.org/linux-kselftest/[email protected]/
There was a problem hiding this comment.
Lots of platform failed to boot due to a bug in linux-next, I think that's now fixed so let's wait for today's results.
There was a problem hiding this comment.
Still no results as the boards that are supposed to run the test failed to boot...
https://staging.kernelci.org/test/job/kernelci/branch/staging-next/kernel/staging-next-20220923.0/plan/kselftest-landlock/
There was a problem hiding this comment.
Ah it looks like a known issue where some debug network interfaces are brought up as a side-effect of enabling some configs with kselftest builds:
https://storage.staging.kernelci.org/kernelci/staging-next/staging-next-20220923.0/arm64/defconfig+kselftest/gcc-10/lab-baylibre/kselftest-landlock-meson-gxl-s905x-libretech-cc.html#L876
The landlock tests were built fine though:
https://storage.staging.kernelci.org/kernelci/staging-next/staging-next-20220923.0/arm64/defconfig+kselftest/gcc-10/logs/kselftest.log
|
Everything looks good now. Could someone merge this PR? |
|
Any news on this? It's been sitting for an extremely long time at this point with nothing actionable... |
|
The Beaglebone is just a random infra issue that happened to affect that particular run, though it looks like the kselftest config is broken there anyway unfortunately https://lava.sirena.org.uk/scheduler/job/3748. The potato issue does look legit unfortunately though it has run in the past and whatever looks to be nothing to do with the test case - it would really help if we had baseline jobs running for the kselftest kernels but unfortunately we don't seem to. Either way I'm not seeing something that's at all specific to this test case here and this is just getting incredibly frustrating, the actual tests seem fine but the pull request only gets looked at very intermittently and any infra issue at that time means another month or whatever of delay. The build issue that cropped up was legit (though that had been fine earlier...) but otherwise... |
Add a base definition for the landlock selftests. This is a very quick to run testsuite. Signed-off-by: Mark Brown <[email protected]>
|
I keep looking latest results, and it doesnt look like defconfig+kselftest kernel can boot at all on 2 available boards: https://staging.kernelci.org/build/id/63712751901342a358b0d63b/ |
|
@nuclearcat these issues are not related to this PR, but probably to the hardware support which would explain the failed boots. I see |
|
I know, just documenting that i am looking to PR from time to time. Not sure i have enough time and resources to troubleshoot hardware support issue. |
Yes, definitely! I'd really like to run these tests on x86 machines too. |
|
There's a bunch of x86 ChromeBooks in the Collabora labs, hopefully one of them has sufficient capacity to run some more tests? |
|
I will ask for that. |
|
Hi @broonie , @l0kod , i added x86 DUT for landlock and it seems working fine. Please let me know if results are as expected, then i can merge it. |
Enable the landlock selftests, initially just on Le Potato and BeagleBone
Black. This is a very quick to run testsuite.
Signed-off-by: Mark Brown [email protected]