On Fri, 28 Aug 2026, Thadeu Lima de Souza Cascardo <[email protected]> wrote:
> On Wed, Aug 19, 2026 at 03:07:40PM -0400, Mark Yacoub wrote:
>> Writeback and virtual connectors (e.g. DRM_MODE_CONNECTOR_WRITEBACK and
>> DRM_MODE_CONNECTOR_VIRTUAL) do not represent physical display sinks and
>> cannot be modeset using standard CRTC / atomic commit operations without
>> configuring a writeback output framebuffer (WRITEBACK_FB_ID).
>> 
>> On drivers such as MSM, writeback connectors (e.g. Virtual-1, Virtual-2)
>> report connected status with virtual modes. When generic KMS tests iterate
>> over connected outputs via for_each_connected_output() or during initial
>> display setup in igt_display_init(), attempting standard modesets on these
>> connectors causes failures (-EINVAL) before test execution can begin.
>> 
>> Exclude DRM_MODE_CONNECTOR_WRITEBACK and DRM_MODE_CONNECTOR_VIRTUAL in
>> igt_output_is_connected() so generic KMS tests only target standard
>> physical display outputs, while dedicated writeback tests can continue to
>> iterate all outputs explicitly via for_each_output().
>> 
>> Signed-off-by: Mark Yacoub <[email protected]>
>> ---
>>  lib/igt_kms.h | 4 ++++
>>  1 file changed, 4 insertions(+)
>> 
>> diff --git a/lib/igt_kms.h b/lib/igt_kms.h
>> index 521a03c01..8d4d14847 100644
>> --- a/lib/igt_kms.h
>> +++ b/lib/igt_kms.h
>> @@ -668,6 +668,10 @@ static inline bool igt_output_is_connected(igt_output_t 
>> *output)
>>          !output->config.connector->count_modes)
>>              return false;
>>  
>> +    if (output->config.connector->connector_type == 
>> DRM_MODE_CONNECTOR_WRITEBACK ||
>> +        output->config.connector->connector_type == 
>> DRM_MODE_CONNECTOR_VIRTUAL)
>> +            return false;
>> +
>>      if (output->config.connector->connection == DRM_MODE_CONNECTED)
>>              return true;
>>  
>> -- 
>> 2.55.0.737.g08866a6d13-goog
>> 
>
> This makes a lot of tests (including a lot of kms_*) be skipped on VKMS. Is
> that correct?

I think the commit overlooked the fact that VKMS has virtual connectors
only. Probably needs a revert and a return to the drawing board.

Also, Someone(tm) should run CI with VKMS and IGT. Never tried myself,
but shouldn't that be possible in user-mode Linux?


BR,
Jani.


-- 
Jani Nikula, Intel

Reply via email to