[PATCH i-g-t v2 1/2] tests/kms_async_flips: Split subtest to handle linear buffers separately

[Karthik B S]

Hi Santhosh,

On 4/16/2025 8:33 PM, Santhosh Reddy Guddati wrote:
> As linear buffers are considered corner cases, split the existing subtests
> run_with_modifiers to handle as a separate subtest.
>
> V2:
> Rename existing subtest to tiled modifier and make the tiled subtest
> intel specific, fix linear modifier subtest to avoid skip on non intel
> platforms (Karthik).
>
> Signed-off-by: Santhosh Reddy Guddati <santhosh.reddy.guddati at intel.com>
> ---
>   tests/kms_async_flips.c | 53 ++++++++++++++++++++++++++++++++++-------
>   1 file changed, 45 insertions(+), 8 deletions(-)
>
> diff --git a/tests/kms_async_flips.c b/tests/kms_async_flips.c
> index 031417ed8..a3defe604 100644
> --- a/tests/kms_async_flips.c
> +++ b/tests/kms_async_flips.c
> @@ -45,7 +45,7 @@
>    * Description: Verify the async flip functionality and the fps during async flips
>    *              Alternate between sync and async flips
>    *
> - * SUBTEST: async-flip-with-page-flip-events
> + * SUBTEST: async-flip-with-page-flip-events-tiled
>    * Description: Verify the async flip functionality and the fps during async flips
>    *              Wait for page flip events in between successive asynchronous flips
>    *
> @@ -66,7 +66,7 @@
>    * SUBTEST: alternate-sync-async-flip-atomic
>    * Description: Verify the async flip functionality and the fps during async flips using atomic path
>    *
> - * SUBTEST: async-flip-with-page-flip-events-atomic
> + * SUBTEST: async-flip-with-page-flip-events-tiled-atomic
>    * Description: Wait for page flip events in between successive asynchronous flips using atomic path
>    *
>    * SUBTEST: test-time-stamp-atomic
> @@ -88,6 +88,14 @@
>    *
>    * SUBTEST: overlay-atomic
>    * Description: Verify overlay planes with async flips in atomic API
> + *
> + * SUBTEST: async-flip-with-page-flip-events-linear
> + * Description: Verify the async flip functionality and the fps during async flips
> + *		with linear modifier
> + *
> + * SUBTEST: async-flip-with-page-flip-events-linear-atomic
> + * Description: Verify the async flip functionality and the fps during async flips
> + *		with linear modifier in Atomic API
>    */
>   
>   #define CURSOR_POS 128
> @@ -128,6 +136,7 @@ typedef struct {
>   	struct buf_ops *bops;
>   	bool atomic_path;
>   	bool overlay_path;
> +	bool linear_modifier;
>   } data_t;
>   
>   static void flip_handler(int fd_, unsigned int sequence, unsigned int tv_sec,
> @@ -266,6 +275,7 @@ static void test_init_ops(data_t *data)
>   	data->alternate_sync_async = false;
>   	data->atomic_path = false;
>   	data->overlay_path = false;
> +	data->linear_modifier = false;
>   }
>   
>   static void test_init_fbs(data_t *data)
> @@ -792,15 +802,23 @@ static void run_test_with_modifiers(data_t *data, void (*test)(data_t *))
>   		test_init(data);
>   
>   		for (int i = 0; i < data->plane->format_mod_count; i++) {
> +			uint64_t modifier = data->plane->modifiers[i];
> +
>   			if (data->plane->formats[i] != DRM_FORMAT_XRGB8888)
>   				continue;
>   
> +			if (data->linear_modifier && modifier != DRM_FORMAT_MOD_LINEAR)
> +				continue;
> +
> +			if (!data->linear_modifier && modifier == DRM_FORMAT_MOD_LINEAR)
> +				continue;
> +
>   			data->allow_fail = true;
> -			data->modifier = data->plane->modifiers[i];
> +			data->modifier = modifier;
>   
>   			igt_dynamic_f("pipe-%s-%s-%s", kmstest_pipe_name(data->pipe),
>   				      data->output->name,
> -				      igt_fb_modifier_name(data->modifier)) {
> +				      igt_fb_modifier_name(modifier)) {
>   				      /*
>   				       * FIXME: joiner+async flip is busted currently in KMD.
>   				       * Remove this check once the issues are fixed in KMD.
> @@ -844,19 +862,38 @@ igt_main
>   			require_monotonic_timestamp(data.drm_fd);
>   
>   		igt_describe("Wait for page flip events in between successive asynchronous flips");
> -		igt_subtest_with_dynamic("async-flip-with-page-flip-events") {
> +		igt_subtest_with_dynamic("async-flip-with-page-flip-events-tiled") {
> +			test_init_ops(&data);
> +			igt_require(is_intel_device(data.drm_fd));
> +			run_test_with_modifiers(&data, test_async_flip);
> +		}
> +
> +		igt_describe("Wait for page flip events in between successive "
> +			     "asynchronous flips using atomic path");
> +		igt_subtest_with_dynamic("async-flip-with-page-flip-events-tiled-atomic") {
> +			test_init_ops(&data);
> +			data.atomic_path = true;
> +			igt_require(is_intel_device(data.drm_fd));
> +			run_test_with_modifiers(&data, test_async_flip);
> +		}
> +
> +		igt_describe("Wait for page flip events in between successive asynchronous "
> +			     "flips with linear modifiers");
> +		igt_subtest_with_dynamic("async-flip-with-page-flip-events-linear") {
>   			test_init_ops(&data);
> +			data.linear_modifier = true;
>   			if (is_intel_device(data.drm_fd))
>   				run_test_with_modifiers(&data, test_async_flip);
>   			else
>   				run_test(&data, test_async_flip);

With this current change in test design, we could just use the run_test 
function itself for linear. This would eliminate the looping for 
non-linear modifiers which we're anyway skipping eventually. Also, 
inside run_test_with_modifiers we can reduce the checks as we would then 
be using that function exclusively for tiled modifiers only.

In run_test we would just need this one change.

data->modifier = data.linear_modifier? DRM_FORMAT_MOD_LINEAR: 
default_modifier(data);

Thanks,
Karthik.B.S
>   		}
>   
> -		igt_describe("Wait for page flip events in between successive "
> -			     "asynchronous flips using atomic path");
> -		igt_subtest_with_dynamic("async-flip-with-page-flip-events-atomic") {
> +		igt_describe("Wait for page flip events in between successive asynchronous "
> +			     "flips using atomic path with linear modifiers");
> +		igt_subtest_with_dynamic("async-flip-with-page-flip-events-linear-atomic") {
>   			test_init_ops(&data);
>   			data.atomic_path = true;
> +			data.linear_modifier = true;
>   			if (is_intel_device(data.drm_fd))
>   				run_test_with_modifiers(&data, test_async_flip);
>   			else