baseline regression in kevin

[r-c-n]

The following bootrr tests failed starting on kernel v5.19-rc1:

• bootrr.cros-ec-accel0-probed
• bootrr.cros-ec-accel1-probed
• bootrr.cros-ec-sensors-gyro0-probed
• bootrr.rockchip-i2s1-probed

Initially discussed in https://lore.kernel.org/chrome-platform/[email protected]/T/#m232b9217365668316bfc59acd6dc5ae2b96d2dc9

The problem seems to be related to how these bootrr tests are defined:

assert_driver_present cros-ec-sensors-driver-present cros-ec-sensors
if kernel_greater_than "5.5"; then
    assert_device_present cros-ec-sensors-accel0-probed cros-ec-sensors cros-ec-accel.11.*
    assert_device_present cros-ec-sensors-accel1-probed cros-ec-sensors cros-ec-accel.13.*
    assert_device_present cros-ec-sensors-gyro0-probed cros-ec-sensors cros-ec-gyro.12.*

    assert_driver_present cros-ec-sensorhub-driver-present cros-ec-sensorhub
    assert_device_present cros-ec-sensorhub-probed cros-ec-sensorhub cros-ec-sensorhub.1.*
else
    assert_device_present cros-ec-sensors-accel0-probed cros-ec-sensors cros-ec-accel.0
    assert_device_present cros-ec-sensors-accel1-probed cros-ec-sensors cros-ec-accel.1
    assert_device_present cros-ec-sensors-gyro0-probed cros-ec-sensors cros-ec-gyro.0
fi

This way of probing the devices doesn't look very future proof, as Brian Norris remarked:

That sounds like we're trying to test kernel implementation details, asynchronous probe race conditions, Makefile / linker ordering, and similar -- not anything that we actually expect to remain stable across kernel versions :(

For instance, in kevin, for v5.19-rc1 onwards these devices are named with a different device index:

/sys/bus/platform/drivers/cros-ec-sensors/cros-ec-gyro.11.auto
/sys/bus/platform/drivers/cros-ec-sensors/cros-ec-accel.12.auto
/sys/bus/platform/drivers/cros-ec-sensors/cros-ec-accel.10.auto

KernelCI dashboard link: https://linux.kernelci.org/test/plan/id/635f585b4cac8cb70ce7db4e/

[computersforpeace]

The i2s1 failure is because of https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b5fbaf7d779f5f02b7f75b080e7707222573be2a

We should just delete that case, and perhaps add a case for the spdif device.

[gctucker]

Note that it's also possible to check the kernel version in bootrr and look for a different device accordingly, if that makes sense in practice.

[crazoes]

@gctucker created this PR for staging kernelci/bootrr#23 as a short term fix. We can avoid checking for the ID in the device name for now.

[padovan]

Old issue. (also we are not tracking kernel test failure/issues through GitHub anymore)