Re: [PATCH i-g-t v1] tests/kms_async_flips: Use lowest mode for crc subtest

["Reddy Guddati, Santhosh" <santhosh.reddy.guddati@intel.com>]

Hi Rama,

On 10-03-2026 20:37, Naladala, Ramanaidu wrote:
> Hi Santhosh,
> 
> On 3/9/2026 1:54 PM, Santhosh Reddy Guddati wrote:
>> Currently The CRC subtest uses mode[0] which maps to preffered high
>> refresh mode. On panels with multiple modes, pick the lower mode to
>> speed up crc test.
>>
>> Signed-off-by: Santhosh Reddy Guddati <santhosh.reddy.guddati@intel.com>
>> ---
>>   tests/kms_async_flips.c | 15 +++++++++++++--
>>   1 file changed, 13 insertions(+), 2 deletions(-)
>>
>> diff --git a/tests/kms_async_flips.c b/tests/kms_async_flips.c
>> index a42407363..8639e2299 100644
>> --- a/tests/kms_async_flips.c
>> +++ b/tests/kms_async_flips.c
>> @@ -825,6 +825,17 @@ static void paint_fb(data_t *data, struct igt_fb 
>> *fb,
>>       }
>>   }
>> +static drmModeModeInfoPtr get_lower_mode_for_crc(igt_output_t *output)
>> +{
>> +    drmModeConnector *connector = output->config.connector;
>> +
>> +    igt_assert(connector->count_modes > 0);
>> +
>> +    igt_sort_connector_modes(connector, sort_drm_modes_by_clk_asc);
>> +
>> +    return &connector->modes[0];
>> +}
>> +
>>   static void test_crc(data_t *data)
>>   {
>>       unsigned int frame = 0;
>> @@ -834,8 +845,8 @@ static void test_crc(data_t *data)
>>       igt_display_commit2(&data->display, data->display.is_atomic ? 
>> COMMIT_ATOMIC : COMMIT_LEGACY);
>> -    /* make things faster by using a smallish mode */
>> -    mode = &data->output->config.connector->modes[0];
>> +    /* make things faster by using a lower bandwidth mode */
>> +    mode = get_lower_mode_for_crc(data->output);
> 
> This optimization is for simulation? if yes, Add a check for simulation 
> then change to lower mode. On Hardware it won't take much time. Why this 
> time optimization required on hardware?
> 
> is any issues seen?

Thanks for the comment.

This change is not for simulation.

The purpose is to fix CRC subtest mode selection on panels with multiple 
modes (especially when there are HRR modes). Earlier we used 
`connector->modes[0]`, which can pick a high-refresh/high-bandwidth mode 
(if modes[0] has highest) and make this test path more sensitive.

Now we select a lower-bandwidth mode using `get_lower_mode_for_crc()`. 
The async flip + CRC validation logic is unchanged. So this is a focused 
fix to make the test stable on HRR panels, not a new behavior change.

In past , we have seen issues related to this and the test was handled 
to adjust the criteria.

https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/8247
> 
>>       width = mode->hdisplay;
>>       height = mode->vdisplay;