by Vignesh Raman

[permalink] [raw]

Subject: Re: [PATCH v4 00/11] drm/ci: Add support for GPU and display testing

Hi Helen,

On 07/03/24 19:32, Helen Koike wrote:
>
>
> On 06/03/2024 00:06, Vignesh Raman wrote:
>> Some ARM SOCs have a separate display controller and GPU, each with
>> different drivers. For mediatek mt8173, the GPU driver is powervr,
>> and the display driver is mediatek. In the case of mediatek mt8183,
>> the GPU driver is panfrost, and the display driver is mediatek.
>> With rockchip rk3288/rk3399, the GPU driver is panfrost, while the
>> display driver is rockchip. For amlogic meson, the GPU driver is
>> panfrost, and the display driver is meson.
>>
>> IGT tests run various tests with different xfails and can test both
>> GPU devices and KMS/display devices. Currently, in drm-ci for MediaTek,
>> Rockchip, and Amlogic Meson platforms, only the GPU driver is tested.
>> This leads to incomplete coverage since the display is never tested on
>> these platforms. This commit series adds support in drm-ci to run tests
>> for both GPU and display drivers for MediaTek, Rockchip, and Amlogic
>> Meson platforms.
>>
>> Uprev mesa and IGT in drm-ci and add amd, v3d, vc4 and vgem specific
>> tests to testlist. Have testlist.txt per driver and include a base
>> testlist so that the driver specific tests will run only on those
>> hardware
>> and add testlists to the MAINTAINERS file.
>>
>> This series also includes patch to add vkms testing to drm-ci.
>>
>> Working pipeline link,
>> https://gitlab.freedesktop.org/vigneshraman/linux/-/pipelines/1119859
>
> Hi Vignesh, thanks for this patchset.
>
> Just to confirm, the new convention is:
>
> stages are named per DRIVER_NAME

Yes correct.

> jobs are named DRIVER_NAME:GPU_VERSION

Yes

> xfails files and testlist files are named DRIVER_NAME-<xfails-suffix>.txt

Base testlist is named as testlist.txt
driver specific testlist is named as testlist-$DRIVER_NAME.txt

xfails files are named as $DRIVER_NAME-$GPU_VERSION-<xfails-suffix>.txt
(mediatek-mt8183-fails.txt)

>
> Looks good.
>
> I just have some minor comments/questions in the series, please check.

Will address those and send v5.

Regards,
Vignesh

2024-03-15 12:03:49

by Helen Koike

[permalink] [raw]

Subject: Re: [PATCH v4 09/11] drm/ci: rockchip: Refactor existing rockchip jobs

On 15/03/2024 08:18, Vignesh Raman wrote:
> Hi Helen,
>
> On 07/03/24 19:32, Helen Koike wrote:
>>
>>
>> On 06/03/2024 00:06, Vignesh Raman wrote:
>>> For rockchip rk3288 and rk3399, the display driver is rockchip.
>>> Currently, in drm-ci for rockchip, only the display driver is
>>> tested. Refactor the existing rockchip jobs so that gpu driver
>>> testing jobs can be added later and update xfails accordingly.
>>>
>>> Signed-off-by: Vignesh Raman <[email protected]>
>>> ---
>>>
>>> v2:
>>>    - Refactor the patch to rename job to indicate display driver
>>> testing,
>>>      rename the existing xfail files.
>>>
>>> v3:
>>>    - Add the job name in GPU_VERSION and use it for xfail file names
>>>      instead of using DRIVER_NAME. Also update xfails.
>>>
>>> v4:
>>>    - Remove the display suffix in job and rename xfails accordingly.
>>>      Remove the change adding job name in GPU_VERSION.
>>>
>>> ---
>>> drivers/gpu/drm/ci/test.yml                   | 36 ++++++++----
>>> .../drm/ci/xfails/rockchip-rk3288-fails.txt   | 58 ++++++-------------
>>> .../drm/ci/xfails/rockchip-rk3288-flakes.txt | 20 +++++++
>>> .../drm/ci/xfails/rockchip-rk3288-skips.txt   | 54 ++---------------
>>> .../drm/ci/xfails/rockchip-rk3399-fails.txt   | 38 ++++++------
>>> .../drm/ci/xfails/rockchip-rk3399-flakes.txt | 28 +++++++--
>>> .../drm/ci/xfails/rockchip-rk3399-skips.txt   | 5 +-
>>> 7 files changed, 110 insertions(+), 129 deletions(-)
>>> create mode 100644
>>> drivers/gpu/drm/ci/xfails/rockchip-rk3288-flakes.txt
>>>
>>> diff --git a/drivers/gpu/drm/ci/test.yml b/drivers/gpu/drm/ci/test.yml
>>> index 6ae6398b3d88..831e580e6dfd 100644
>>> --- a/drivers/gpu/drm/ci/test.yml
>>> +++ b/drivers/gpu/drm/ci/test.yml
>>> @@ -153,33 +153,45 @@ msm:sdm845:
>>>     script:
>>>       - ./install/bare-metal/cros-servo.sh
>>> -rockchip:rk3288:
>>> - extends:
>>> -    - .lava-igt:arm32
>>> +.rockchip:
>>>     stage: rockchip
>>>     variables:
>>> -    DRIVER_NAME: rockchip
>>
>> Looks like it make sense to keep DRIVER_NAME here, no?
>>
>>> -    DEVICE_TYPE: rk3288-veyron-jaq
>>>       DTB: ${DEVICE_TYPE}
>>>       BOOT_METHOD: depthcharge
>>> +
>>> +.rk3288:
>>> + extends:
>>> +    - .lava-igt:arm32
>>> +    - .rockchip
>>
>> Maybe, instead of extending .rockchip here, make .rockchip tied to the
>> DRIVER_NAME and .rk3288 tied to the GPU_VERSION, and on
>> rockchip:rk3288 you can extend both .rockchip and .rk3288, what do you
>> think?
>> So rockchip:rk3399 you can extend .rockchip and .rk3399.
>>
>> and in the panfrost one you can have a .panfrost (that can extend
>> .rockchip if they are the same definition).
>>
>> I feel it becomes less confusing, what do you think?
>>
>> I would even add some prefix or suffix to make it less confusing, like
>> .driver-rockchip and .gpu-rk3288 for instance, making it a bit more
>> intuitive and helping our future selves :)
>
> Thanks for the suggestion. This can be done. Should we do it only
> for rockchip jobs or others also (meson, mediatek) ?

I guess we could keep the same patter for all the make sense.

Helen

>
>>
>>> + variables:
>>> +    DEVICE_TYPE: rk3288-veyron-jaq
>>>       KERNEL_IMAGE_TYPE: "zimage"
>>> -    GPU_VERSION: rk3288
>>
>> Looks like it make sense to keep GPU_VERSION here, no? Same comment
>> for .rk3399.
> Yes, will fix this.
>
> Regards,
> Vignesh