Re: [PATCH i-g-t] tests/kms_atomic_transition: Fix retry-path resources in modeset tests

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

Hi pranay,

On 24-06-2026 00:12, Pranay Samala wrote:
> When a bandwidth fallback retry happens, resources from
> previous attempt may not be released, leading to accumulation
> of allocated objects across retries.
> 
> Add cleanup before retry to ensure clean starting state by
> releasing resources and removing allocated framebuffers,
> preventing resource buildup.
> 
> Signed-off-by: Pranay Samala <pranay.samala@intel.com>
> ---
>   tests/kms_atomic_transition.c | 10 ++++++++++
>   1 file changed, 10 insertions(+)
> 
> diff --git a/tests/kms_atomic_transition.c b/tests/kms_atomic_transition.c
> index 58cadc560..37dc77c28 100644
> --- a/tests/kms_atomic_transition.c
> +++ b/tests/kms_atomic_transition.c
> @@ -1026,6 +1026,16 @@ retry:
>   		bool found = igt_override_all_active_output_modes_to_fit_bw(&data->display);
>   		igt_require_f(found, "No valid mode combo found.\n");
>   
> +		if (is_intel_device(data->drm_fd)) {
> +			for_each_crtc(&data->display, crtc) {
> +				igt_pipe_crc_free(data->pipe_crcs[crtc->crtc_index]);
> +				data->pipe_crcs[crtc->crtc_index] = NULL;
> +			}
> +		}
> +
> +		igt_remove_fb(data->drm_fd, &data->fbs[0]);
> +		igt_remove_fb(data->drm_fd, &data->fbs[1]);
> +

At the beginning of the retry label, the height and width are not reset, 
since the igt_override_all_active_output_modes_to_fit_bw() reduces 
modes, the newly created FBs may still use the old (larger) dimensions 
from the first attempt. Consider resetting to 0 and this is not related 
to this patch. could be a follow up.

Rest LGTM.
Reviewed-by: Santhosh Reddy Guddati <santhosh.reddy.guddati@intel.com>



>   		goto retry;
>   	}
>   	igt_display_commit2(&data->display, COMMIT_ATOMIC);