Re: [PATCH] tests/kms_writeback: refresh to expose writeback output

[Alex Hung <alex.hung@amd.com>]

On 2024-05-13 16:37, Abhinav Kumar wrote:
> 
> 
> On 5/13/2024 3:32 PM, Alex Hung wrote:
>>
>>
>> On 2024-05-13 16:20, Abhinav Kumar wrote:
>>>
>>>
>>> On 5/13/2024 3:06 PM, Alex Hung wrote:
>>>>
>>>>
>>>> On 2024-05-13 15:48, Abhinav Kumar wrote:
>>>>>
>>>>>
>>>>> On 5/13/2024 2:30 PM, Alex Hung wrote:
>>>>>> With dc2d7fb4f978, all kms_writeback subtests are skipped.
>>>>>>
>>>>>> Refresh the outputs to grab all supported connectors,
>>>>>> the writeback output that is just enabled.
>>>>>>
>>>>>
>>>>> Can you please help to explain how the tests are getting skipped 
>>>>> due to setting the client_cap?
>>>>>
>>>>> And also why calling igt_display_reset_outputs() is fixing this issue?
>>>>
>>>> Without the igt_display_reset_outputs(), the result is
>>>>
>>>> IGT-Version: 1.28-gbe9b99928 (x86_64) (Linux: 6.7.0-rc5-99-custom 
>>>> x86_64)
>>>> Using IGT_SRANDOM=1715637543 for randomisation
>>>> Opened device: /dev/dri/card0
>>>> Test requirement not met in function __igt_unique____real_main553, 
>>>> file ../tests/kms_writeback.c:579:
>>>> Test requirement: output
>>>> Last errno: 95, Operation not supported
>>>> Subtest writeback-pixel-formats: SKIP (0.000s)
>>>> Subtest writeback-invalid-parameters: SKIP (0.000s)
>>>> Subtest writeback-fb-id: SKIP (0.000s)
>>>> Subtest writeback-fb-id-XRGB2101010: SKIP (0.000s)
>>>> Subtest writeback-check-output: SKIP (0.000s)
>>>> Subtest writeback-check-output-XRGB2101010: SKIP (0.000s)
>>>>
>>>>
>>>> It looks like the below "output = 
>>>> kms_writeback_get_output(&display)" will return null as it cannot 
>>>> find any connector type with DRM_MODE_CONNECTOR_WRITEBACK.
>>>>
>>>> The igt_display_reset_outputs() is able to re-initialize the output 
>>>> and therefore includes the writeback output.
>>>>
>>>
>>> Thanks for the details, I understand the issue now.
>>>
>>> But rather than calling igt_display_reset_outputs(), I think better 
>>> way will be to set the cap before calling igt_display_require()
>>>
>>> diff --git a/tests/kms_writeback.c b/tests/kms_writeback.c
>>> index f89b2d62fcc9..6ecab46e440b 100644
>>> --- a/tests/kms_writeback.c
>>> +++ b/tests/kms_writeback.c
>>> @@ -567,14 +567,13 @@ igt_main_args("b:c:f:dl", long_options, 
>>> help_str, opt_handler, NULL)
>>>
>>>                  kmstest_set_vt_graphics_mode();
>>>
>>> +               ret = drmSetClientCap(display.drm_fd, 
>>> DRM_CLIENT_CAP_WRITEBACK_CONNECTORS, 1);
>>> +               igt_require_f(!ret, "error setting 
>>> DRM_CLIENT_CAP_WRITEBACK_CONNECTORS\n");
>>> +
>>>                  igt_display_require(&display, display.drm_fd);
>>>
>>> Let me know if that works for you.
>>>
>>
>> This works if DRM_CLIENT_CAP_ATOMIC is also set as below:
>>
>> Is this preferred?
>>
>> --- a/tests/kms_writeback.c
>> +++ b/tests/kms_writeback.c
>> @@ -567,14 +567,18 @@ igt_main_args("b:c:f:dl", long_options, 
>> help_str, opt_handler, NULL)
>>
>>                  kmstest_set_vt_graphics_mode();
>>
>> -               igt_display_require(&display, display.drm_fd);
>> -
>> -               igt_require(display.is_atomic);
>> +               if (drmSetClientCap(display.drm_fd, 
>> DRM_CLIENT_CAP_ATOMIC, 1) == 0)
>> +                       display.is_atomic = 1;
>>
>>                  ret = drmSetClientCap(display.drm_fd, 
>> DRM_CLIENT_CAP_WRITEBACK_CONNECTORS, 1);
>>
>>                  igt_require_f(!ret, "error setting 
>> DRM_CLIENT_CAP_WRITEBACK_CONNECTORS\n");
>>
>> +               igt_display_require(&display, display.drm_fd);
>> +
>> +               igt_require(display.is_atomic);
>> +
>> +
>>
> 
> Yes, looks correct to me because drm driver expects both to be set.
> 
> 357     case DRM_CLIENT_CAP_WRITEBACK_CONNECTORS:
> 358         if (!file_priv->atomic)
> 359             return -EINVAL;
> 360         if (req->value > 1)
> 361             return -EINVAL;
> 362         file_priv->writeback_connectors = req->value;
> 363         break;

Thanks. Let me send a V2

>>
>>
>>
>>>>>
>>>>>> Fixes: dc2d7fb4f978 ("lib/igt_kms: move setting 
>>>>>> DRM_CLIENT_CAP_WRITEBACK_CONNECTORS to kms_writeback")
>>>>>> Signed-off-by: Alex Hung <alex.hung@amd.com>
>>>>>> ---
>>>>>>   tests/kms_writeback.c | 4 +++-
>>>>>>   1 file changed, 3 insertions(+), 1 deletion(-)
>>>>>>
>>>>>> diff --git a/tests/kms_writeback.c b/tests/kms_writeback.c
>>>>>> index f89b2d62f..4c54f3b89 100644
>>>>>> --- a/tests/kms_writeback.c
>>>>>> +++ b/tests/kms_writeback.c
>>>>>> @@ -572,9 +572,11 @@ igt_main_args("b:c:f:dl", long_options, 
>>>>>> help_str, opt_handler, NULL)
>>>>>>           igt_require(display.is_atomic);
>>>>>>           ret = drmSetClientCap(display.drm_fd, 
>>>>>> DRM_CLIENT_CAP_WRITEBACK_CONNECTORS, 1);
>>>>>> -
>>>>>
>>>>> Is this part needed for this patch?
>>>>
>>>> No, but I think they are related and thus grouping them together is 
>>>> easier to read.
>>>>
>>>>>
>>>>>>           igt_require_f(!ret, "error setting 
>>>>>> DRM_CLIENT_CAP_WRITEBACK_CONNECTORS\n");
>>>>>> +        /* Refresh the outputs to grab all supported connectors.*/
>>>>>> +        igt_display_reset_outputs(&display);
>>>>>> +
>>>>>>           output = kms_writeback_get_output(&display);
>>>>>>           igt_require(output);