Re: [PATCH i-g-t v3] tests/kms_plane: Skip CRC frame sequence check on MediaTek

Hi Jason-JH,
On 2026-08-15 at 01:28:36 +0800, Jason-JH Lin wrote:
> MediaTek's DRM driver uses an internal per-commit CRC queue that
> reports CRC entries with a frame number offset due to pipeline delay
> (3 vblanks after commit). This causes crc->frame > expected_vblank,
> which fails the exact-match validation in capture_crc().
> 
> Since igt_pipe_crc_get_for_frame() already guarantees crc->frame >=
> vblank via its internal loop, the exact-match check is overly strict
> for queued CRC drivers. Skip it on MediaTek while preserving the
> validation for other platforms.
> 
> Signed-off-by: Jason-JH Lin <[email protected]>

Applied, thank you all!



Regards,
Kamil

> ---
>  tests/kms_plane.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/tests/kms_plane.c b/tests/kms_plane.c
> index 12dfbfe1d82b..14d2d810f1a0 100644
> --- a/tests/kms_plane.c
> +++ b/tests/kms_plane.c
> @@ -766,6 +766,15 @@ static void capture_crc(data_t *data, unsigned int vblank, igt_crc_t *crc)
>  {
>  	igt_pipe_crc_get_for_frame(data->drm_fd, data->pipe_crc, vblank, crc);
>  
> +	/*
> +	 * Platforms with an internal CRC queue (e.g. MediaTek) report CRC
> +	 * entries with a frame number greater than the requested vblank due
> +	 * to pipeline delay. Since igt_pipe_crc_get_for_frame() already
> +	 * guarantees crc->frame >= vblank, skip the exact-match check.
> +	 */
> +	if (is_mtk_device(data->drm_fd))
> +		return;
> +
>  	igt_fail_on_f(!igt_skip_crc_compare && !igt_run_in_simulation() &&
>  		      crc->has_valid_frame && crc->frame != vblank,
>  		      "Got CRC for the wrong frame (got %u, expected %u). CRC buffer overflow?\n",
> -- 
> 2.43.0
>