Open plbossart opened 3 years ago
In theory Call Trace
should be caught, see the last lines of sof-test/tools/sof-kernel-log-check.sh
Wait: this is a boot time error, correct? We don't catch boot time errors. This has been discussed in internal issue 466
kernel: [ 4.078863] BUG: key ffff8a9203c92dc0 has not been registered!
kernel: [ 4.078865] ------------[ cut here ]------------
kernel: [ 4.078866] DEBUG_LOCKS_WARN_ON(1)
...
kernel: [ 4.078902] Call Trace:
kernel: [ 4.078906] __kernfs_create_file+0x76/0x100
@marc-hb this happens when the snd-sof-pci driver is probed. This can happen at boot time if there's no blacklist.
If you don't catch such errors at boot time, then the sof-ci framework is broken somehow. If we do NOT detect errors on boot and do NOT prevent probe on boot, then we have a huge gap in our test coverage.
We don't catch boot time errors.
@xiulipan says "it may not that simple, may depend on the test". @xiulipan to investigate.
IMHO it's OK if some tests catch boot time errors, for instance it's OK if tests after the first one in a suite don't look at boot time errors.
@plbossart @marc-hb After checking our old logs, we could got boot time call trace but we did not enable kernel log check in verify test cases. I will send a RFC to enable kernel log check in one of our verify test script.
There's been a lot of journalctl changes, so maybe this has been fixed.
Next step: inject a fake error and make sure this test (and others...) fail.
We have fake_kern_error()
in sof-test but not sure it's similar enough to this DEBUG_LOCKS_WARN_ON
in Intel Daily test 627, the verify-kernel-module-load-probe passes for TGL_UNB_SDW12
However the dmesg contains a big deplock WARN tracked here: https://github.com/thesofproject/linux/issues/2542
The question is: shouldn't sof-test fail this case?
cc:
774