[Intel-gfx] [PATCH] igt/kms_rotation_crc: Add a subtest to validate Y-tiled obj + Y fb modifier

[Vivek Kasireddy]

On Tue, 20 Oct 2015 10:10:20 +0100
Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com> wrote:

> 
> On 20/10/15 02:14, Vivek Kasireddy wrote:
> > Hi Tvrtko,
> >
> > On Mon, 19 Oct 2015 11:20:05 +0100
> > Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com> wrote:
> >
> >>
> >> Hi,
> >>
> >> On 17/10/15 03:47, Vivek Kasireddy wrote:
> >>> The main goal of this subtest is to verify whether flipping a
> >>> framebuffer with a Y fb modifier (90/270 degree rotation) and
> >>> an associated Y-tiled object works or not.
> >>>
> >>> Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> >>> Signed-off-by: Vivek Kasireddy <vivek.kasireddy at intel.com>
> >>> ---
> >>>    tests/kms_rotation_crc.c | 83
> >>> ++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed,
> >>> 83 insertions(+)
> >>>
> >>> diff --git a/tests/kms_rotation_crc.c b/tests/kms_rotation_crc.c
> >>> index cc9847e..bb9aecb 100644
> >>> --- a/tests/kms_rotation_crc.c
> >>> +++ b/tests/kms_rotation_crc.c
> >>> @@ -264,6 +264,83 @@ static void test_plane_rotation(data_t *data,
> >>> enum igt_plane plane_type) igt_require_f(valid_tests, "no valid
> >>> crtc/connector combinations found\n"); }
> >>>
> >>> +static void test_plane_rotation_ytiled_obj(data_t *data, enum
> >>> igt_plane plane_type) +{
> >>> +	igt_display_t *display = &data->display;
> >>> +	uint64_t tiling = LOCAL_I915_FORMAT_MOD_Y_TILED;
> >>> +	uint32_t format = DRM_FORMAT_XRGB8888;
> >>> +	int bpp = igt_drm_format_to_bpp(format);
> >>> +	enum igt_commit_style commit = COMMIT_LEGACY;
> >>> +	int fd = data->gfx_fd;
> >>> +	int valid_tests = 0;
> >>> +	igt_output_t *output;
> >>> +	int ret;
> >>> +
> >>> +	if (plane_type == IGT_PLANE_PRIMARY || plane_type ==
> >>> IGT_PLANE_CURSOR) {
> >>> +		igt_require(data->display.has_universal_planes);
> >>> +		commit = COMMIT_UNIVERSAL;
> >>> +	}
> >>> +
> >>> +	for_each_connected_output(display, output) {
> >>> +		igt_plane_t *plane;
> >>> +		drmModeModeInfo *mode =
> >>> igt_output_get_mode(output);
> >>> +		unsigned int w = mode->hdisplay;
> >>> +		unsigned int h = mode->vdisplay;
> >>> +		unsigned int stride, size;
> >>> +		uint32_t gem_handle;
> >>> +
> >>> +		for (stride = 512; stride < (w * bpp / 8); stride
> >>> *= 2)
> >>> +			;
> >>> +		for (size = 1024*1024; size < stride * h; size *=
> >>> 2)
> >>> +			;
> >>> +
> >>> +		gem_handle = gem_create(fd, size);
> >>> +		ret = __gem_set_tiling(fd, gem_handle,
> >>> I915_TILING_Y, stride);
> >>> +		igt_assert(ret == 0);
> >>> +
> >>> +		do_or_die(__kms_addfb(fd, gem_handle, w, h,
> >>> stride,
> >>> +			  format, tiling,
> >>> LOCAL_DRM_MODE_FB_MODIFIERS,
> >>> +			  &data->fb.fb_id));
> >>> +
> >>> +		data->fb.width = w;
> >>> +		data->fb.height = h;
> >>> +		data->fb.gem_handle = gem_handle;
> >>> +		data->fb.stride = stride;
> >>> +		data->fb.size = size;
> >>> +		data->fb.tiling = tiling;
> >>> +		data->fb.drm_format = format;
> >>> +
> >>> +		plane = igt_output_get_plane(output, plane_type);
> >>> +		igt_require(igt_plane_supports_rotation(plane));
> >>> +
> >>> +		igt_plane_set_fb(plane, NULL);
> >>> +		igt_display_commit(display);
> >>> +
> >>> +		igt_plane_set_rotation(plane, data->rotation);
> >>> +		paint_squares(data, mode, IGT_ROTATION_0, plane);
> >>> +		igt_plane_set_fb(plane, &data->fb);
> >>> +
> >>> +		drmModeObjectSetProperty(fd,
> >>> plane->drm_plane->plane_id,
> >>> +					 DRM_MODE_OBJECT_PLANE,
> >>> +
> >>> plane->rotation_property,
> >>> +					 plane->rotation);
> >>> +		ret = igt_display_try_commit2(display, commit);
> >>> +		igt_assert(ret == 0);
> >>
> >> This manages to trigger the WARN? How come since paint_squares
> >> above will have created the normal VMA?
> >
> > Yes, this does trigger the WARN regardless of whether paint_squares
> > is called or not. I am going to have to find out why the normal VMA
> > was not created when paint_squares was called. And, this subtest
> > does not restore the console cleanly which is something I also have
> > to fix.
> 
> Could be that normal VMA is not created in the Cairo rendering path
> for Y tiled buffers. Since the final rendering is done via a blit and
> if you have PPGTT that would do it I think.
> 
> If PPGTT is always true for gen9 then this could work.

Hi Tvrtko,
Yes, I see that PPGTT is always true for gen9 and perhaps the reason
why the normal VMA is not created. Do you suggest that I get rid of the
call to paint_squares and further simplify the test by just flipping the
fb to only one output?

Thanks and Regards,
Vivek

> 
> Regards,
> 
> Tvrtko