cros-ec IIO ABI test regressions

[hardboprobot]

https://linux.kernelci.org/test/case/id/63624a16c49f70a70ce7db69/
Seen on volteer, octopus, sona and kevin. The testcases that fail are test_cros_ec_accel_iio_abi and test_cros_ec_gyro_iio_abi.

The test last worked with tag-chrome-platform-for-v5.20. The same accelerometers and gyroscopes seem to be probed and installed both in v5.19 and v6.0 (https://lava.collabora.dev/scheduler/job/7919810, https://lava.collabora.dev/scheduler/job/7919787) but the test_cros_ec_accel_iio_abi test fails only for v6.0 onwards.

Additional traces in both 5.19: https://lava.collabora.dev/scheduler/job/7920004 and 6.0 https://lava.collabora.dev/scheduler/job/7920009 show that the IIO ABI has changed slightly, for instance, the deviceX/location attribute has dissappeared. That's probably why the test is failing, and it'll stay like that unless it's updated to reflect the current ABI status.

[penvirus]

Hi,

I was looking at one of the failure.

I noticed that the log:
02:53:28.771308 Marking unfinished test run as failed
02:53:28.771622 test_cros_ec_abi: pass
test_cros_ec_accel_iio_abi: fail
test_cros_ec_accel_iio_data_is_valid: skip
test_cros_ec_battery_abi: pass
test_cros_ec_chardev: pass
test_cros_ec_extcon_usbc_abi: skip
test_cros_ec_gyro_iio_abi: fail

It looks like the test didn't finish in some specific duration. I thought it was timed out.

But, by reading the test_cros_ec_accel_iio_abi() and the sysfs_check_attributes_exists(), I failed to figure out how it could be timed out.

Did I mis-interpret the log?

Or, did it stuck when querying the EC features in is_feature_supported?

[hardboprobot]

I don't see any test getting stuck, if you check the timeline they all run instantaneously. But I agree the output is a bit hard to interpret, it's not clear enough. Part of the reason is that these tests swallow the skip reason for skipped tests.

Anyway, I think the root cause is pretty well understood in this case. For failed tests, it's because of missing sysfs attributes. For skipped tests, it's because of missing devices (which is not an issue).

I'll see if I can improve the logging for these tests so that the error messages are properly traced in the log.

[hardboprobot]

The IIO ABI for cros-ec sensors was changed in this commit: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h=v6.1&id=7cbb6681d7e5b88688234ad370e027a9346ff7a9

I'll adapt the tests to comply with it. Documentation/ABI/testing/sysfs-bus-iio-cros-ec is out of sync as well and it should be updated.

[hardboprobot]

Fixed in hardboprobot/cros-ec-tests@f6c0dbf

[hardboprobot]

Related KernelCI-core PR: kernelci/kernelci-core#1572