Re: [PATCH i-g-t] tests/kms_invalid_mode: Allow clock-too-high test on non-Intel platforms

[Jani Nikula <jani.nikula@intel.com>]

On Tue, 14 Apr 2026, Jason-JH Lin <jason-jh.lin@mediatek.com> wrote:
> The clock-too-high subtest was being skipped on non-Intel platforms
> because igt_get_max_dotclock() returns 0 when reading from
> Intel-specific debugfs fails.
>
> This change allows the test to run on all platforms by:
> - Using a clearly invalid clock value (10 GHz) when max_dotclock is
>   unavailable, which any reasonable driver should reject
> - Restricting bigjoiner/ultrajoiner logic to Intel devices only
>
> This prevents the test from being marked as IGNORED/SKIP on non-Intel
> platforms while maintaining the original test intent of verifying that
> drivers properly reject modes with excessively high clock rates.
>
> Tested on MTK platforms where the test now properly executes and
> verifies invalid clock validation.

I think overall we'll need a framework to ask platform/device specific
things instead of portraying the services as generic, like
igt_get_max_dotclock().

>
> Signed-off-by: Jason-JH Lin <jason-jh.lin@mediatek.com>
> ---
>  tests/kms_invalid_mode.c | 38 +++++++++++++++++++++++---------------
>  1 file changed, 23 insertions(+), 15 deletions(-)
>
> diff --git a/tests/kms_invalid_mode.c b/tests/kms_invalid_mode.c
> index 5edffb649ef4..18d79e59c736 100644
> --- a/tests/kms_invalid_mode.c
> +++ b/tests/kms_invalid_mode.c
> @@ -126,7 +126,13 @@ adjust_mode_clock_too_high(data_t *data, drmModeModeInfoPtr mode)
>  {
>  	int max_dotclock = data->max_dotclock;
>  
> -	igt_require(max_dotclock != 0);
> +	/*
> +	 * If max_dotclock is unavailable (e.g., non-Intel platforms),
> +	 * use an obviously invalid value that any driver should reject.
> +	 * 10 GHz is well beyond any reasonable hardware capability.
> +	 */
> +	if (max_dotclock == 0)
> +		max_dotclock = 10000000;  /* 10 GHz in kHz */

Not a fan of using magic numbers like this.

Perhaps

	if (!max_dotclock) {
		mode->clock = -1;
		return;
	}

would be less magic?

>  
>  	/*
>  	 * FIXME When we have a fixed mode, the kernel will ignore
> @@ -139,21 +145,23 @@ adjust_mode_clock_too_high(data_t *data, drmModeModeInfoPtr mode)
>  	if (has_scaling_mode_prop(data))
>  		return false;
>  
> -	/*
> -	 * Newer platforms can support modes higher than the maximum dot clock
> -	 * by using pipe joiner, so set the mode clock twice that of maximum
> -	 * dot clock;
> -	 */
> -	if (can_bigjoiner(data)) {
> -		igt_info("Platform supports bigjoiner with %s\n",
> -			 data->output->name);
> -		max_dotclock *= 2;
> -	}
> +	if (is_intel_device(data->drm_fd)) {
> +		/*
> +		 * Newer platforms can support modes higher than the maximum dot clock
> +		 * by using pipe joiner, so set the mode clock twice that of maximum
> +		 * dot clock;
> +		 */
> +		if (can_bigjoiner(data)) {
> +			igt_info("Platform supports bigjoiner with %s\n",
> +				 data->output->name);
> +			max_dotclock *= 2;
> +		}
>  
> -	if (can_ultrajoiner(data)) {
> -		igt_info("Platform supports ultrajoiner with %s\n",
> -			 data->output->name);
> -		max_dotclock *= 4;
> +		if (can_ultrajoiner(data)) {
> +			igt_info("Platform supports ultrajoiner with %s\n",
> +				 data->output->name);
> +			max_dotclock *= 4;
> +		}
>  	}
>  
>  	mode->clock = max_dotclock + 1;

-- 
Jani Nikula, Intel