You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[TGL][SDW][RT711] Initialization not complete, timed out in suspend-resume test #2599
Description
In daily report 962, several TGL device show below error message in suspend-resume test case, include TGL_RVP_SDW, TGLU_ANDR_SDCA and TGL_OLPMLK_RT715 and CML_OLP_RT711
[ 8611.896412] kernel: rt711 sdw:0:25d:711:0: Initialization not complete, timed out
[ 8611.896427] kernel: PM: dpm_run_callback(): rt711_dev_resume+0x0/0x90 [snd_soc_rt711] returns -110
[ 8611.896433] kernel: PM: Device sdw:0:25d:711:0 failed to resume: error -110
Recipe
Kernel Branch: topic/sof-dev
Kernel Commit: 25012df
SOF Branch: master
SOF Commit: cc1cd8a
That's not good... Do we know what has changed since the last case that worked? We haven't seen those issues for weeks, it's surprising that they pop-up again.
I think we should firstly find a easier way to reproduce it. I can only see this issue in run-all-tests.sh. I can't see it if I run ./check-suspend-resume.sh only.
@bardliao it's reproduced in other CI tests.
I agree that it's not clear why that happens only when we run multiple tests, but the problem does exist :-)
@plbossart Yep, my point is that if we want to figure out when did the issue start from, we need a easier way to reproduce it. But you have fixed it by #2603 :)
I am not happy though, I'd really want to know what causes this test to fail in the sequence - something must interfere. it's very very odd that the suspend-resume tests work if there is nothing done prior.
This only happens after d5dcce0 99aed92 and two iterations of kmod test.
@bardliao can you submit to sof-test a test change or test parameter/configuration that reproduces more easily? It doesn't have to be in CI if it takes long but it should be at least in nightly tests.
Yep, my point is that if we want to figure out when did the issue start from, we need a easier way to reproduce it. But you have fixed it by #2603 :)
Reverting fixes locally is required to "test the test". If the reverts are too complicated then backtracking is another option.
lower priority to P2 as it is not a usual end user scenario with kmod remove/install.
No. This breaks our CI and needs to be fixed as P1. no exceptions, sorry. You are sending a really bad message here.
Thanks for your feedback @plbossart . That was the decision from the bug scrub meeting. FWIW, I have no objection to mark it P1 since Bard and you who actually works on it agree. :)
Activity
plbossart commentedon Dec 4, 2020
That's not good... Do we know what has changed since the last case that worked? We haven't seen those issues for weeks, it's surprising that they pop-up again.
@bardliao @RanderWang FYI.
bardliao commentedon Dec 7, 2020
I think we should firstly find a easier way to reproduce it. I can only see this issue in run-all-tests.sh. I can't see it if I run ./check-suspend-resume.sh only.
plbossart commentedon Dec 7, 2020
@bardliao it's reproduced in other CI tests.
I agree that it's not clear why that happens only when we run multiple tests, but the problem does exist :-)
bardliao commentedon Dec 8, 2020
@plbossart Yep, my point is that if we want to figure out when did the issue start from, we need a easier way to reproduce it. But you have fixed it by #2603 :)
plbossart commentedon Dec 8, 2020
I am not happy though, I'd really want to know what causes this test to fail in the sequence - something must interfere. it's very very odd that the suspend-resume tests work if there is nothing done prior.
keyonjie commentedon Dec 15, 2020
lower priority to P2 as it is not a usual end user scenario with kmod remove/install.
bardliao commentedon Dec 15, 2020
This only happens after d5dcce0c414fcbfe4c2037b66ac69ea5f9b3f75cand 99aed92 and two iterations of kmod test.
plbossart commentedon Dec 15, 2020
No. This breaks our CI and needs to be fixed as P1. no exceptions, sorry. You are sending a really bad message here.
marc-hb commentedon Dec 17, 2020
@bardliao can you submit to sof-test a test change or test parameter/configuration that reproduces more easily? It doesn't have to be in CI if it takes long but it should be at least in nightly tests.
Reverting fixes locally is required to "test the test". If the reverts are too complicated then backtracking is another option.
keyonjie commentedon Dec 21, 2020
Thanks for your feedback @plbossart . That was the decision from the bug scrub meeting. FWIW, I have no objection to mark it P1 since Bard and you who actually works on it agree. :)
keyonjie commentedon Dec 21, 2020
assign it to @bardliao who actually already work on it.
keyonjie commentedon Dec 21, 2020
will PR fix this issue: #2635
mengdonglin commentedon Dec 22, 2020
Closing as PR #2635 was merged.
marc-hb commentedon Jan 4, 2021
I hope you meant "as #2635 was merged and the tests have been consistently passing" :-)
Either way, re-opening based on the most recent comments in #2635, more specifically: "we don't need this" and "please prepare a revert".