Skip to contentSkip to conversation

selftests/sched_ext: testing on BPF CI

DISCUSSION Ihor Solodrai 3 participants 5 messages Jan 29, 2025 – May 3, 2025 BPF

lore
Share & tools
lore

Ihor Solodrai

author reviewer
#0 lore raw

Hi Tejun, Andrea.

I tested a couple of variants of bpf-next + sched_ext source tree, just sharing the results.

I found a working state: BPF CI pipeline ran successfully twice (that's 8 build + run of selftests/sched_ext/runner in total).

Working state requires most patches between sched_ext/master and sched_ext/for-6.14-fixes [1], and also the patch
"tools/sched_ext: Receive updates from SCX repo" [2]

On plain bpf-next the dsp_local_on test fails [3]. Without the patch [2] there is a build error [4]: missing SCX_ENUM_INIT definition.

We probably don't want to enable selftests/sched_ext on BPF CI with that many "temporary" patches. I suggest to wait until all of this is merged upstream.

You can check the full list of patches here:
https://github.com/kernel-patches/vmtest/pull/332/files

[1] https://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext.git/log/?h=for-6.14-fixes
[2] https://lore.kernel.org/all/[email protected]/
[3] https://github.com/kernel-patches/vmtest/actions/runs/13019837022
[4] https://github.com/kernel-patches/vmtest/actions/runs/13020458479

#1 lore raw

Hi Ihor,

Quoted 4 lines · Ihor Solodrai

Hi Tejun, Andrea.

I tested a couple of variants of bpf-next + sched_ext source tree, just sharing the results.

Thanks for testing!

Quoted 3 lines

I found a working state: BPF CI pipeline ran successfully twice (that's 8 build + run of selftests/sched_ext/runner in total).

Ok.

Quoted 8 lines

Working state requires most patches between sched_ext/master and sched_ext/for-6.14-fixes [1], and also the patch
"tools/sched_ext: Receive updates from SCX repo" [2]

On plain bpf-next the dsp_local_on test fails [3]. Without the patch [2] there is a build error [4]: missing SCX_ENUM_INIT definition.

We definitely need all the patches in sched_ext/for-6.14-fixes. I think once Tejun sends the PR and we land the for-6.14-fixes upstream we should reach a stable state with the sched_ext selftests. I don't have any other additional pending fix at the moment.

Quoted 4 lines

We probably don't want to enable selftests/sched_ext on BPF CI with that many "temporary" patches. I suggest to wait until all of this is merged upstream.

Sounds reasonable to me. Tejun?

Quoted 8 lines

You can check the full list of patches here:
https://github.com/kernel-patches/vmtest/pull/332/files

[1] https://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext.git/log/?h=for-6.14-fixes
[2] https://lore.kernel.org/all/[email protected]/
[3] https://github.com/kernel-patches/vmtest/actions/runs/13019837022
[4] https://github.com/kernel-patches/vmtest/actions/runs/13020458479

Thanks,
-Andrea

#2 lore raw
Quoted 31 lines · Andrea Righi

Hi Ihor,

Quoted 4 lines · Ihor Solodrai

Hi Tejun, Andrea.

I tested a couple of variants of bpf-next + sched_ext source tree, just sharing the results.

Thanks for testing!

Quoted 3 lines

I found a working state: BPF CI pipeline ran successfully twice (that's 8 build + run of selftests/sched_ext/runner in total).

Ok.

Quoted 8 lines

Working state requires most patches between sched_ext/master and sched_ext/for-6.14-fixes [1], and also the patch
"tools/sched_ext: Receive updates from SCX repo" [2]

On plain bpf-next the dsp_local_on test fails [3]. Without the patch [2] there is a build error [4]: missing SCX_ENUM_INIT definition.

We definitely need all the patches in sched_ext/for-6.14-fixes. I think once Tejun sends the PR and we land the for-6.14-fixes upstream we should reach a stable state with the sched_ext selftests. I don't have any other additional pending fix at the moment.

Quoted 1 line
Quoted 5 lines
Quoted 3 lines

We probably don't want to enable selftests/sched_ext on BPF CI with that many "temporary" patches. I suggest to wait until all of this is merged upstream.

Sounds reasonable to me. Tejun?

Sure.

Thanks.

Signature · 1 line
tejun
author reviewer
#3 lore raw
In reply to Ihor Solodrai
Quoted 12 lines · Ihor Solodrai

Hi Tejun, Andrea.

I tested a couple of variants of bpf-next + sched_ext source tree, just sharing the results.

I found a working state: BPF CI pipeline ran successfully twice (that's 8 build + run of selftests/sched_ext/runner in total).

Working state requires most patches between sched_ext/master and sched_ext/for-6.14-fixes [1], and also the patch
"tools/sched_ext: Receive updates from SCX repo" [2]

Quoted 8 lines

On plain bpf-next the dsp_local_on test fails [3]. Without the patch [2] there is a build error [4]: missing SCX_ENUM_INIT definition.

We probably don't want to enable selftests/sched_ext on BPF CI with that many "temporary" patches. I suggest to wait until all of this is merged upstream.

Hi everyone. I tried enabling sched_ext selftests on CI today, and there are no issues on bpf-next tip (f263336a41da).

https://github.com/kernel-patches/vmtest/actions/runs/14802453691

If there are no objections, I'm going to push this to BPF CI on Monday.

As a reminder, this means that selftests/sched_ext test runner will be built and executed for pending BPF patches, and BPF CI pipeline will fail in case of problems there.

Quoted 7 lines

You can check the full list of patches here:
https://github.com/kernel-patches/vmtest/pull/332/files

[1] https://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext.git/log/?h=for-6.14-fixes
[2] https://lore.kernel.org/all/[email protected]/
[3] https://github.com/kernel-patches/vmtest/actions/runs/13019837022
[4] https://github.com/kernel-patches/vmtest/actions/runs/13020458479

#4 lore raw

Hi Ihor,

Quoted 25 lines · Ihor Solodrai
Quoted 12 lines · Ihor Solodrai

Hi Tejun, Andrea.

I tested a couple of variants of bpf-next + sched_ext source tree, just sharing the results.

I found a working state: BPF CI pipeline ran successfully twice (that's 8 build + run of selftests/sched_ext/runner in total).

Working state requires most patches between sched_ext/master and sched_ext/for-6.14-fixes [1], and also the patch
"tools/sched_ext: Receive updates from SCX repo" [2]

Quoted 8 lines

On plain bpf-next the dsp_local_on test fails [3]. Without the patch [2] there is a build error [4]: missing SCX_ENUM_INIT definition.

We probably don't want to enable selftests/sched_ext on BPF CI with that many "temporary" patches. I suggest to wait until all of this is merged upstream.

Hi everyone. I tried enabling sched_ext selftests on CI today, and there are no issues on bpf-next tip (f263336a41da).

Quoted 7 lines

https://github.com/kernel-patches/vmtest/actions/runs/14802453691

If there are no objections, I'm going to push this to BPF CI on Monday.

As a reminder, this means that selftests/sched_ext test runner will be built and executed for pending BPF patches, and BPF CI pipeline will fail in case of problems there.

I don't have any objection. The sched_ext selftests are in a stable state now and I think it's a good idea to run them periodically against the latest BPF patches to catch potential breakages early.

Thanks,
-Andrea

Quoted 9 lines
Quoted 8 lines

You can check the full list of patches here:
https://github.com/kernel-patches/vmtest/pull/332/files

[1] https://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext.git/log/?h=for-6.14-fixes
[2] https://lore.kernel.org/all/[email protected]/
[3] https://github.com/kernel-patches/vmtest/actions/runs/13019837022
[4] https://github.com/kernel-patches/vmtest/actions/runs/13020458479