[PATCH i-g-t v3 02/41] lib/vkms: Add minimal VKMS library and test device default files

[Louis Chauvet]

On 8/7/25 9:48 AM, José Expósito wrote:
> Hi everyone,
> 
> On Thu, Aug 07, 2025 at 10:10:17AM +0530, Karthik B S wrote:
>>
>> On 8/5/2025 9:59 PM, Kamil Konieczny wrote:
>>> Hi José,
>>> On 2025-08-05 at 11:20:09 +0200, José Expósito wrote:
>>>> Hi Louis,
>>>>
>>>> On Wed, Jul 16, 2025 at 11:22:38AM +0200, Louis Chauvet wrote:
>>>>>
>>>>> Le 15/07/2025 à 12:24, José Expósito a écrit :
>>>>>> Create a library containing helpers for creating VKMS devices and
>>>>>> configuring them dynamically using configfs.
>>>>>> For the moment, add the minimal number of helpers to be able to start
>>>>>> testing VKMS's configfs support: Create device, destroy device and
>>>>>> destroy all devices.
>>>>>>
>>>>>> Also, include the simplest possible test using those helpers (checking
>>>>>> the device's default files) and the scaffolding required to generate
>>>>>> the documentation.
>>>>>>
>>>>>> Co-developed-by: Jim Shargo <jshargo at chromium.org>
>>>>>> Signed-off-by: Jim Shargo <jshargo at chromium.org>
>>>>>> Co-developed-by: Marius Vlad <marius.vlad at collabora.com>
>>>>>> Signed-off-by: Marius Vlad <marius.vlad at collabora.com>
>>>>>> Signed-off-by: José Expósito <jose.exposito89 at gmail.com>
>>>>>> ---
>>>>>>     docs/testplan/meson.build        |   7 +-
>>>>>>     lib/igt_vkms.c                   | 207 +++++++++++++++++++++++++++++++
>>>>>>     lib/igt_vkms.h                   |  27 ++++
>>>>>>     lib/meson.build                  |   1 +
>>>>>>     meson.build                      |   8 ++
>>>>>>     tests/meson.build                |   2 +
>>>>>>     tests/vkms/meson.build           |  13 ++
>>>>>>     tests/vkms/vkms_configfs.c       | 132 ++++++++++++++++++++
>>>>>>     tests/vkms/vkms_test_config.json |  72 +++++++++++
>>>>>>     9 files changed, 467 insertions(+), 2 deletions(-)
>>>>>>     create mode 100644 lib/igt_vkms.c
>>>>>>     create mode 100644 lib/igt_vkms.h
>>>>>>     create mode 100644 tests/vkms/meson.build
>>>>>>     create mode 100644 tests/vkms/vkms_configfs.c
>>>>>>     create mode 100644 tests/vkms/vkms_test_config.json
>>>>>>
>>>>>> [...]
>>>>>> diff --git a/lib/igt_vkms.c b/lib/igt_vkms.c
>>>>>> new file mode 100644
>>>>>> index 000000000..fa41f741e
>>>>>> --- /dev/null
>>>>>> +++ b/lib/igt_vkms.c
>>>>>> [...]
>>>>>> +static int detach_pipeline_items(const char *path, const struct stat *info,
>>>>>> +				 const int typeflag, struct FTW *pathinfo)
>>>>> As there will be a new "color pipeline" in DRM, can you rename this set of
>>>>> functions to avoid confusion? I don't have preference for the name, what do
>>>>> you think about "dc" for display controller?
>>> Could you give a link to lore.kernel.org for this 'color pipeline'?
>>>
>>>>> (ditto for the rest of the series)
>>>>>
>>>>> With this modification:
>>>>> Reviewed-by: Louis Chauvet <louis.chauvet at bootlin.com>
>>>> The "pipeline" here refers to the display pipeline as in the docs:
>>>> https://docs.kernel.org/gpu/drm-kms.html#overview
>>>>
>>>> Given that the VKMS driver is a display driver and that "enum vkms_pipeline_item"
>>>> contains those entites, I find easy to understand what we are talking about.
>>>>
>>>> What do you think about leaving the display pipeline as "pipeline" and, once we
>>>> add the color pipeline, use "color_pipeline"?
>>> These looks resonable. Karthik, could you comment on this?
>>
>> Yes, this looks good.
> 
> Thanks for your feedback. I sent v4 with all the requested changes,
> but at least for the moment, I didn't rename the pipeline:
> https://lore.kernel.org/igt-dev/20250807074550.6543-1-jose.exposito89@gmail.com/

I am ok to keep pipeline, don't worry!

I proposed this because I was reading about color pipeline just before 
reviewing your series and it was a bit surprising to read pipeline twice 
with different meaning.

I will finish to review the v4 next week.

Thanks a lot for your work!
Louis Chauvet

> Thanks to everyone for their reviews!
> Jose
> 
>> Thanks,
>> Karthik.B.S
>>>
>>>> I can rename all instances of "pipeline" to "display_pipeline" (with a lot of work,
>>>> 100 occurences in igt_vkms.c in 41 patches), but I find it too verbose for little
>>>> clarification.
>>> Well, you can use sed script for this but I am afraid that there
>>> are many other places in igt with 'pipeline' meaning
>>> 'display pipeline', so it would be a lot more work to do.
>>>
>>> Regards,
>>> Kamil
>>>
>>>> Jose
>>>>
>>>>>> [...]
>>>>> -- 
>>>>> Louis Chauvet, Bootlin
>>>>> Embedded Linux and Kernel engineering
>>>>> https://bootlin.com