Re: [PATCH i-g-t] tests/intel/kms_sharpness_filter: Fix stale mode pointer in joiner bypass

[Karthik B S <karthik.b.s@intel.com>]

Hi Pranay,

On 3/23/2026 7:41 PM, Pranay Samala wrote:
> When joiner mode is detected, the returned mode pointer from
> igt_get_non_joiner_mode() points into connector-owned memory
> which can become invalid after subsequent display state changes.
>
> Fix this by copying the mode into local variable for stable
> storage. Skip the output if no non-joiner mode is found.
>
> Fixes: 10dccc7fc82f ("tests/intel/kms_sharpness_filter: Detect and bypass joiner modes")
> Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/7045
> Signed-off-by: Pranay Samala <pranay.samala@intel.com>
> ---
>   tests/intel/kms_sharpness_filter.c | 13 +++++++++++--
>   1 file changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/tests/intel/kms_sharpness_filter.c b/tests/intel/kms_sharpness_filter.c
> index 96fb139ae..c2295c0ee 100644
> --- a/tests/intel/kms_sharpness_filter.c
> +++ b/tests/intel/kms_sharpness_filter.c
> @@ -430,6 +430,7 @@ run_sharpness_filter_test(data_t *data, enum test_type type)
>   	igt_display_t *display = &data->display;
>   	igt_output_t *output;
>   	igt_crtc_t *crtc;
> +	drmModeModeInfo *m, selected_mode;
>   	char name[40];
>   
>   	for_each_connected_output(display, output) {
> @@ -438,7 +439,6 @@ run_sharpness_filter_test(data_t *data, enum test_type type)
>   
>   			data->output = output;
>   			data->crtc = crtc;
> -			data->mode = igt_output_get_mode(data->output);
>   
>   			/*
>   			 * FIXME: Joiner + CASF currently unsupported.
> @@ -446,7 +446,15 @@ run_sharpness_filter_test(data_t *data, enum test_type type)
>   			 * Until then, run on non-joiner mode in joiner configuration.
>   			 */
>   			if (is_joiner_mode(data->drm_fd, data->output)) {
> -				data->mode = igt_get_non_joiner_mode(data->drm_fd, data->output);
> +				m = igt_get_non_joiner_mode(data->drm_fd, data->output);
> +				if (!m) {
> +					igt_info("No non-joiner mode found on %s\n",
> +						 igt_output_name(data->output));
> +					continue;
> +				}
> +
> +				selected_mode = *m;
> +				data->mode = &selected_mode;

Instead of doing this, if we get the mode after setting the CRTC on the 
output and we should be getting the correct mode. Could you please check 
this.

Thanks and Regards,
Karthik.B.S
>   				igt_info("Executing on mode %dx%d@%d\n",
>   					 data->mode->hdisplay,
>   					 data->mode->vdisplay,
> @@ -463,6 +471,7 @@ run_sharpness_filter_test(data_t *data, enum test_type type)
>   
>   			igt_output_set_crtc(data->output,
>   					    data->crtc);
> +			igt_output_override_mode(data->output, data->mode);
>   
>   			if (!intel_pipe_output_combo_valid(display)) {
>   				igt_output_set_crtc(data->output, NULL);