[PATCH v8 00/11] drm: Add new pixel formats for Xilinx Zynqmp

[PATCH v8 00/11] drm: Add new pixel formats for Xilinx Zynqmp

[Tomi Valkeinen]

Add new DRM pixel formats and add support for those in the Xilinx zynqmp
display driver.

All other formats except XVUY2101010 are already supported in upstream
gstreamer, but gstreamer's kmssink does not have the support yet, as it
obviously cannot support the formats without kernel having the formats.

Xilinx has support for these formats in their BSP kernel, and Xilinx has
a branch here, adding the support to gstreamer kmssink:

https://github.com/Xilinx/gst-plugins-bad.git xlnx-rebase-v1.18.5

New formats added:

DRM_FORMAT_Y8
- 8-bit Y-only
- fourcc: "GREY"
- gstreamer: GRAY8

DRM_FORMAT_Y10_P32
- 10-bit Y-only, three pixels packed into 32-bits
- fourcc: "YPA4"
- gstreamer: GRAY10_LE32

DRM_FORMAT_XV15
- Like NV12, but with 10-bit components
- fourcc: "XV15"
- gstreamer: NV12_10LE32

DRM_FORMAT_XV20
- Like NV16, but with 10-bit components
- fourcc: "XV20"
- gstreamer: NV16_10LE32

DRM_FORMAT_X403
- 10-bit planar 4:4:4, with three samples packed into 32-bits
- fourcc: "X403"
- gstreamer: Y444_10LE32

XVUY2101010
- 10-bit 4:4:4, one pixel in 32 bits
- fourcc: "XY30"

Some notes:

I know the 8-bit greyscale format has been discussed before, and the
guidance was to use DRM_FORMAT_R8. While I'm not totally against that, I
would argue that adding DRM_FORMAT_Y8 makes sense, as:

1) We can mark it as 'is_yuv' in the drm_format_info, and this can help
   the drivers handle e.g. full/limited range. Probably some hardware
   handles grayscale as a value used for all RGB components, in which case
   R8 makes sense, but when the hardware handles the Y-only pixels as YCbCr,
   where Cb and Cr are "neutral", it makes more sense to consider the
   format as an YUV format rather than RGB.

2) We can have the same fourcc as in v4l2. While not strictly necessary,
   it's a constant source of confusion when the fourccs differ.

3) It (possibly) makes more sense for the user to use Y8/GREY format
   instead of R8, as, in my experience, the documentation usually refers
   to gray(scale) format or Y-only format.

As we add new Y-only formats, it makes sense to have similar terms, so
we need to adjust the Y10_P32 format name accordingly.

I have made some adjustments to the formats compared to the Xilinx's
branch. E.g. The DRM_FORMAT_Y10_P32 format in Xilinx's kmssink uses
fourcc "Y10 ", and DRM_FORMAT_Y10. I didn't like those, as the format is
a packed format, three 10-bit pixels in a 32-bit container, and I think
Y10 means a 10-bit pixel in a 16-bit container.

Generally speaking, if someone has good ideas for the format define
names or fourccs, speak up, as it's not easy to invent good names =).
That said, keeping them the same as in the Xilinx trees will, of course,
be slightly easier for the users of Xilinx platforms.

I made WIP additions to modetest to support most of these formats,
partially based on Xilinx's code:

https://github.com/tomba/libdrm.git xilinx

A few thoughts about that:

modetest uses bo_create_dumb(), and as highlighted in recent discussions
in the kernel list [1], dumb buffers are only for RGB formats. They may
work for non-RGB formats, but that's platform specific. None of the
formats I add here are RGB formats. Do we want to go this way with
modetest?

I also feel that the current structure of modetest is not well suited to
more complicated formats. Both the buffer allocation is a bit more
difficult (see "Add virtual_width and pixels_per_container"), and the
drawing is complicated (see, e.g., "Add support for DRM_FORMAT_XV15 &
DRM_FORMAT_XV20").

I have recently added support for these Xilinx formats to both kms++ [2] and
pykms/pixutils [3][4] (WIP), and it's not been easy... But I have to say I
think I like the template based version in kms++. That won't work in
modetest, of course, but a non-templated version might be implementable,
but probably much slower.

In any case, I slighly feel it's not worth merging the modetest patches
I have for these formats: they complicate the code quite a bit, break
the RGB-only rule, and I'm not sure if there really are (m)any users. If
we want to add support to modetest, I think a bigger rewrite of the test
pattern code might be in order.

[1] https://lore.kernel.org/all/20250109150310.219442-26-tzimmermann%40suse.de/
[2] git@github.com:tomba/kmsxx.git xilinx
[3] git@github.com:tomba/pykms.git xilinx
[4] git@github.com:tomba/pixutils.git xilinx

Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
Changes in v8:
- Expand the "drm/fourcc: Add DRM_FORMAT_Y8" commit description to
  explain the rationale
- Add comment to "drm: xlnx: zynqmp: Add support for Y8 and Y10_P32"
  explainig the Y-only matrix
- Remove extra blank line
- Link to v7: https://lore.kernel.org/r/20251201-xilinx-formats-v7-0-1e1558adfefc@ideasonboard.com

Changes in v7:
- Added Reviewed-bys
- Rebased on v6.18
- Link to v6: https://lore.kernel.org/r/20251001-xilinx-formats-v6-0-014b076b542a@ideasonboard.com

Changes in v6:
- Added tags for reviews
- Rebased on v6.17
- Link to v5: https://lore.kernel.org/r/20250425-xilinx-formats-v5-0-c74263231630@ideasonboard.com

Changes in v5:
- Add comment about Y-only formats, clarifying how the display pipeline
  handles them (they're handled as YCbCr, with Cb and Cr as "neutral")
- Clarify X403 format in the patch description
- Set unused Y-only CSC offsets to 0 (instead of 0x1800).
- Add R-bs
- Link to v4: https://lore.kernel.org/r/20250326-xilinx-formats-v4-0-322a300c6d72@ideasonboard.com

Changes in v4:
- Reformat the drm_format_info entries a bit
- Calculate block size only once in drm_format_info_bpp()
- Declare local variables in separate lines
- Add review tags
- Fix commit message referring to Y10_LE32 (should be Y10_P32)
- Link to v3: https://lore.kernel.org/r/20250212-xilinx-formats-v3-0-90d0fe106995@ideasonboard.com

Changes in v3:
- Drop "drm: xlnx: zynqmp: Fix max dma segment size". It is already
  pushed.
- Add XVUY2101010 format.
- Rename DRM_FORMAT_Y10_LE32 to DRM_FORMAT_Y10_P32.
- Link to v2: https://lore.kernel.org/r/20250115-xilinx-formats-v2-0-160327ca652a@ideasonboard.com

Changes in v2:
- I noticed V4L2 already has fourcc Y10P, referring to MIPI-style packed
  Y10 format. So I changed Y10_LE32 fourcc to YPA4. If logic has any
  relevance here, P means packed, A means 10, 4 means "in 4 bytes".
- Added tags to "Fix max dma segment size" patch
- Updated description for "Add warning for bad bpp"
- Link to v1: https://lore.kernel.org/r/20241204-xilinx-formats-v1-0-0bf2c5147db1@ideasonboard.com

---
Tomi Valkeinen (11):
      drm/fourcc: Add warning for bad bpp
      drm/fourcc: Add DRM_FORMAT_XV15/XV20
      drm/fourcc: Add DRM_FORMAT_Y8
      drm/fourcc: Add DRM_FORMAT_Y10_P32
      drm/fourcc: Add DRM_FORMAT_X403
      drm/fourcc: Add DRM_FORMAT_XVUY2101010
      drm: xlnx: zynqmp: Use drm helpers when calculating buffer sizes
      drm: xlnx: zynqmp: Add support for XV15 & XV20
      drm: xlnx: zynqmp: Add support for Y8 and Y10_P32
      drm: xlnx: zynqmp: Add support for X403
      drm: xlnx: zynqmp: Add support for XVUY2101010

 drivers/gpu/drm/drm_fourcc.c       | 28 +++++++++++++++++--
 drivers/gpu/drm/xlnx/zynqmp_disp.c | 56 +++++++++++++++++++++++++++++++++++---
 include/uapi/drm/drm_fourcc.h      | 28 +++++++++++++++++++
 3 files changed, 105 insertions(+), 7 deletions(-)
---
base-commit: 7d0a66e4bb9081d75c82ec4957c50034cb0ea449
change-id: 20241120-xilinx-formats-f71901621833

Best regards,
-- 
Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>

[PATCH v8 01/11] drm/fourcc: Add warning for bad bpp

[Tomi Valkeinen]

drm_format_info_bpp() cannot be used for formats which do not have an
integer bits-per-pixel in a pixel block.

E.g. DRM_FORMAT_XV15's (not yet in upstream) plane 0 has three 10-bit
pixels (Y components), and two padding bits, in a 4 byte block. That is
10.666... bits per pixel when considering the whole 4 byte block, which
is what drm_format_info_bpp() does. Thus a driver that supports such
formats cannot use drm_format_info_bpp(),

It is a driver bug if this happens, but so handle wrong calls by
printing a warning and returning 0.

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
 drivers/gpu/drm/drm_fourcc.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/drm_fourcc.c b/drivers/gpu/drm/drm_fourcc.c
index e0d533611040..e662aea9d105 100644
--- a/drivers/gpu/drm/drm_fourcc.c
+++ b/drivers/gpu/drm/drm_fourcc.c
@@ -491,12 +491,20 @@ EXPORT_SYMBOL(drm_format_info_block_height);
  */
 unsigned int drm_format_info_bpp(const struct drm_format_info *info, int plane)
 {
+	unsigned int block_size;
+
 	if (!info || plane < 0 || plane >= info->num_planes)
 		return 0;
 
-	return info->char_per_block[plane] * 8 /
-	       (drm_format_info_block_width(info, plane) *
-		drm_format_info_block_height(info, plane));
+	block_size = drm_format_info_block_width(info, plane) *
+		     drm_format_info_block_height(info, plane);
+
+	if (info->char_per_block[plane] * 8 % block_size) {
+		pr_warn("unable to return an integer bpp\n");
+		return 0;
+	}
+
+	return info->char_per_block[plane] * 8 / block_size;
 }
 EXPORT_SYMBOL(drm_format_info_bpp);
 

-- 
2.43.0

[PATCH v8 02/11] drm/fourcc: Add DRM_FORMAT_XV15/XV20

[Tomi Valkeinen]

Add two new pixel formats:

DRM_FORMAT_XV15 ("XV15")
DRM_FORMAT_XV20 ("XV20")

The formats are 2 plane 10 bit per component YCbCr, with the XV15 2x2
subsampled whereas XV20 is 2x1 subsampled.

Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Vishal Sagar <vishal.sagar@amd.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
 drivers/gpu/drm/drm_fourcc.c  | 6 ++++++
 include/uapi/drm/drm_fourcc.h | 8 ++++++++
 2 files changed, 14 insertions(+)

diff --git a/drivers/gpu/drm/drm_fourcc.c b/drivers/gpu/drm/drm_fourcc.c
index e662aea9d105..b22ef86428a1 100644
--- a/drivers/gpu/drm/drm_fourcc.c
+++ b/drivers/gpu/drm/drm_fourcc.c
@@ -381,6 +381,12 @@ const struct drm_format_info *__drm_format_info(u32 format)
 		{ .format = DRM_FORMAT_S416,            .depth = 0,  .num_planes = 3,
 		  .char_per_block = { 2, 2, 2 }, .block_w = { 1, 1, 1 }, .block_h = { 1, 1, 1 },
 		  .hsub = 1, .vsub = 1, .is_yuv = true},
+		{ .format = DRM_FORMAT_XV15,		.depth = 0,  .num_planes = 2,
+		  .char_per_block = { 4, 8, 0 }, .block_w = { 3, 3, 0 }, .block_h = { 1, 1, 0 },
+		  .hsub = 2, .vsub = 2, .is_yuv = true },
+		{ .format = DRM_FORMAT_XV20,		.depth = 0,  .num_planes = 2,
+		  .char_per_block = { 4, 8, 0 }, .block_w = { 3, 3, 0 }, .block_h = { 1, 1, 0 },
+		  .hsub = 2, .vsub = 1, .is_yuv = true },
 	};
 
 	unsigned int i;
diff --git a/include/uapi/drm/drm_fourcc.h b/include/uapi/drm/drm_fourcc.h
index e527b24bd824..6c786701238e 100644
--- a/include/uapi/drm/drm_fourcc.h
+++ b/include/uapi/drm/drm_fourcc.h
@@ -322,6 +322,14 @@ extern "C" {
 #define DRM_FORMAT_RGB565_A8	fourcc_code('R', '5', 'A', '8')
 #define DRM_FORMAT_BGR565_A8	fourcc_code('B', '5', 'A', '8')
 
+/*
+ * 2 plane 10 bit per component YCrCb
+ * index 0 = Y plane, [31:0] x:Y2:Y1:Y0 2:10:10:10 little endian
+ * index 1 = Cb:Cr plane, [63:0] x:Cr2:Cb2:Cr1:x:Cb1:Cr0:Cb0 2:10:10:10:2:10:10:10 little endian
+ */
+#define DRM_FORMAT_XV15		fourcc_code('X', 'V', '1', '5') /* 2x2 subsampled Cr:Cb plane 2:10:10:10 */
+#define DRM_FORMAT_XV20		fourcc_code('X', 'V', '2', '0') /* 2x1 subsampled Cr:Cb plane 2:10:10:10 */
+
 /*
  * 2 plane YCbCr
  * index 0 = Y plane, [7:0] Y

-- 
2.43.0

[PATCH v8 03/11] drm/fourcc: Add DRM_FORMAT_Y8

[Tomi Valkeinen]

Add greyscale Y8 format.

The 8-bit greyscale format has been discussed before, and the earlier
guidance was to use DRM_FORMAT_R8, as a single-channel 8-bit pixel.

However, adding DRM_FORMAT_Y8 makes sense, we can mark it as 'is_yuv' in
the drm_format_info, and this can help the drivers handle e.g.
full/limited range. This will distinguish two single-channel formats:
R8, which is a RGB format with the same value for all components, and
Y8, which is a Y-only YCbCr format, with Cb and Cr being neutral.

Acked-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Reviewed-by: Vishal Sagar <vishal.sagar@amd.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
 drivers/gpu/drm/drm_fourcc.c  | 1 +
 include/uapi/drm/drm_fourcc.h | 9 +++++++++
 2 files changed, 10 insertions(+)

diff --git a/drivers/gpu/drm/drm_fourcc.c b/drivers/gpu/drm/drm_fourcc.c
index b22ef86428a1..a39b9d7a5b62 100644
--- a/drivers/gpu/drm/drm_fourcc.c
+++ b/drivers/gpu/drm/drm_fourcc.c
@@ -275,6 +275,7 @@ const struct drm_format_info *__drm_format_info(u32 format)
 		{ .format = DRM_FORMAT_YVU422,		.depth = 0,  .num_planes = 3, .cpp = { 1, 1, 1 }, .hsub = 2, .vsub = 1, .is_yuv = true },
 		{ .format = DRM_FORMAT_YUV444,		.depth = 0,  .num_planes = 3, .cpp = { 1, 1, 1 }, .hsub = 1, .vsub = 1, .is_yuv = true },
 		{ .format = DRM_FORMAT_YVU444,		.depth = 0,  .num_planes = 3, .cpp = { 1, 1, 1 }, .hsub = 1, .vsub = 1, .is_yuv = true },
+		{ .format = DRM_FORMAT_Y8,		.depth = 8,  .num_planes = 1, .cpp = { 1, 0, 0 }, .hsub = 1, .vsub = 1, .is_yuv = true },
 		{ .format = DRM_FORMAT_NV12,		.depth = 0,  .num_planes = 2, .cpp = { 1, 2, 0 }, .hsub = 2, .vsub = 2, .is_yuv = true },
 		{ .format = DRM_FORMAT_NV21,		.depth = 0,  .num_planes = 2, .cpp = { 1, 2, 0 }, .hsub = 2, .vsub = 2, .is_yuv = true },
 		{ .format = DRM_FORMAT_NV16,		.depth = 0,  .num_planes = 2, .cpp = { 1, 2, 0 }, .hsub = 2, .vsub = 1, .is_yuv = true },
diff --git a/include/uapi/drm/drm_fourcc.h b/include/uapi/drm/drm_fourcc.h
index 6c786701238e..e4451668499a 100644
--- a/include/uapi/drm/drm_fourcc.h
+++ b/include/uapi/drm/drm_fourcc.h
@@ -459,6 +459,15 @@ extern "C" {
 #define DRM_FORMAT_YUV444	fourcc_code('Y', 'U', '2', '4') /* non-subsampled Cb (1) and Cr (2) planes */
 #define DRM_FORMAT_YVU444	fourcc_code('Y', 'V', '2', '4') /* non-subsampled Cr (1) and Cb (2) planes */
 
+/*
+ * Y-only (greyscale) formats
+ *
+ * The Y-only formats are handled similarly to the YCbCr formats in the display
+ * pipeline, with the Cb and Cr implicitly neutral (0.0 in nominal values). This
+ * also means that COLOR_RANGE property applies to the Y-only formats.
+ */
+
+#define DRM_FORMAT_Y8		fourcc_code('G', 'R', 'E', 'Y')  /* 8-bit Y-only */
 
 /*
  * Format Modifiers:

-- 
2.43.0

[PATCH v8 04/11] drm/fourcc: Add DRM_FORMAT_Y10_P32

[Tomi Valkeinen]

Add Y10_P32, a 10 bit greyscale format, with 3 pixels packed into
32-bit container.

The fourcc for the format is 'YPA4', which comes from Y - Y only, P -
packed, A - 10 (as in 0xA), 4 - 4 bytes.

Reviewed-by: Vishal Sagar <vishal.sagar@amd.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
 drivers/gpu/drm/drm_fourcc.c  | 3 +++
 include/uapi/drm/drm_fourcc.h | 1 +
 2 files changed, 4 insertions(+)

diff --git a/drivers/gpu/drm/drm_fourcc.c b/drivers/gpu/drm/drm_fourcc.c
index a39b9d7a5b62..0d222f6c1a30 100644
--- a/drivers/gpu/drm/drm_fourcc.c
+++ b/drivers/gpu/drm/drm_fourcc.c
@@ -388,6 +388,9 @@ const struct drm_format_info *__drm_format_info(u32 format)
 		{ .format = DRM_FORMAT_XV20,		.depth = 0,  .num_planes = 2,
 		  .char_per_block = { 4, 8, 0 }, .block_w = { 3, 3, 0 }, .block_h = { 1, 1, 0 },
 		  .hsub = 2, .vsub = 1, .is_yuv = true },
+		{ .format = DRM_FORMAT_Y10_P32,		.depth = 0,  .num_planes = 1,
+		  .char_per_block = { 4, 0, 0 }, .block_w = { 3, 0, 0 }, .block_h = { 1, 0, 0 },
+		  .hsub = 1, .vsub = 1, .is_yuv = true },
 	};
 
 	unsigned int i;
diff --git a/include/uapi/drm/drm_fourcc.h b/include/uapi/drm/drm_fourcc.h
index e4451668499a..bf6a41462af5 100644
--- a/include/uapi/drm/drm_fourcc.h
+++ b/include/uapi/drm/drm_fourcc.h
@@ -468,6 +468,7 @@ extern "C" {
  */
 
 #define DRM_FORMAT_Y8		fourcc_code('G', 'R', 'E', 'Y')  /* 8-bit Y-only */
+#define DRM_FORMAT_Y10_P32	fourcc_code('Y', 'P', 'A', '4')  /* [31:0] x:Y2:Y1:Y0 2:10:10:10 little endian */
 
 /*
  * Format Modifiers:

-- 
2.43.0

[PATCH v8 05/11] drm/fourcc: Add DRM_FORMAT_X403

[Tomi Valkeinen]

Add X403, a 3 plane 10 bits per component non-subsampled YCbCr format.

Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Reviewed-by: Vishal Sagar <vishal.sagar@amd.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
 drivers/gpu/drm/drm_fourcc.c  | 3 +++
 include/uapi/drm/drm_fourcc.h | 9 +++++++++
 2 files changed, 12 insertions(+)

diff --git a/drivers/gpu/drm/drm_fourcc.c b/drivers/gpu/drm/drm_fourcc.c
index 0d222f6c1a30..ab643dedd6d4 100644
--- a/drivers/gpu/drm/drm_fourcc.c
+++ b/drivers/gpu/drm/drm_fourcc.c
@@ -391,6 +391,9 @@ const struct drm_format_info *__drm_format_info(u32 format)
 		{ .format = DRM_FORMAT_Y10_P32,		.depth = 0,  .num_planes = 1,
 		  .char_per_block = { 4, 0, 0 }, .block_w = { 3, 0, 0 }, .block_h = { 1, 0, 0 },
 		  .hsub = 1, .vsub = 1, .is_yuv = true },
+		{ .format = DRM_FORMAT_X403,		.depth = 0,  .num_planes = 3,
+		  .char_per_block = { 4, 4, 4 }, .block_w = { 3, 3, 3 }, .block_h = { 1, 1, 1 },
+		  .hsub = 1, .vsub = 1, .is_yuv = true },
 	};
 
 	unsigned int i;
diff --git a/include/uapi/drm/drm_fourcc.h b/include/uapi/drm/drm_fourcc.h
index bf6a41462af5..0128398b7936 100644
--- a/include/uapi/drm/drm_fourcc.h
+++ b/include/uapi/drm/drm_fourcc.h
@@ -403,6 +403,15 @@ extern "C" {
  */
 #define DRM_FORMAT_Q401		fourcc_code('Q', '4', '0', '1')
 
+/*
+ * 3 plane non-subsampled (444) YCbCr
+ * 10 bpc, 30 bits per sample image data in a single contiguous buffer.
+ * index 0: Y plane,  [31:0] x:Y2:Y1:Y0    [2:10:10:10] little endian
+ * index 1: Cb plane, [31:0] x:Cb2:Cb1:Cb0 [2:10:10:10] little endian
+ * index 2: Cr plane, [31:0] x:Cr2:Cr1:Cr0 [2:10:10:10] little endian
+ */
+#define DRM_FORMAT_X403		fourcc_code('X', '4', '0', '3')
+
 /*
  * 3 plane YCbCr LSB aligned
  * In order to use these formats in a similar fashion to MSB aligned ones

-- 
2.43.0

[PATCH v8 06/11] drm/fourcc: Add DRM_FORMAT_XVUY2101010

[Tomi Valkeinen]

Add XVUY2101010, a 10 bits per component YCbCr format in a 32 bit
container.

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Vishal Sagar <vishal.sagar@amd.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
 drivers/gpu/drm/drm_fourcc.c  | 1 +
 include/uapi/drm/drm_fourcc.h | 1 +
 2 files changed, 2 insertions(+)

diff --git a/drivers/gpu/drm/drm_fourcc.c b/drivers/gpu/drm/drm_fourcc.c
index ab643dedd6d4..a736df2de3fc 100644
--- a/drivers/gpu/drm/drm_fourcc.c
+++ b/drivers/gpu/drm/drm_fourcc.c
@@ -288,6 +288,7 @@ const struct drm_format_info *__drm_format_info(u32 format)
 		{ .format = DRM_FORMAT_VYUY,		.depth = 0,  .num_planes = 1, .cpp = { 2, 0, 0 }, .hsub = 2, .vsub = 1, .is_yuv = true },
 		{ .format = DRM_FORMAT_XYUV8888,	.depth = 0,  .num_planes = 1, .cpp = { 4, 0, 0 }, .hsub = 1, .vsub = 1, .is_yuv = true },
 		{ .format = DRM_FORMAT_VUY888,          .depth = 0,  .num_planes = 1, .cpp = { 3, 0, 0 }, .hsub = 1, .vsub = 1, .is_yuv = true },
+		{ .format = DRM_FORMAT_XVUY2101010,     .depth = 0,  .num_planes = 1, .cpp = { 4, 0, 0 }, .hsub = 1, .vsub = 1, .is_yuv = true },
 		{ .format = DRM_FORMAT_AYUV,		.depth = 0,  .num_planes = 1, .cpp = { 4, 0, 0 }, .hsub = 1, .vsub = 1, .has_alpha = true, .is_yuv = true },
 		{ .format = DRM_FORMAT_Y210,            .depth = 0,  .num_planes = 1, .cpp = { 4, 0, 0 }, .hsub = 2, .vsub = 1, .is_yuv = true },
 		{ .format = DRM_FORMAT_Y212,            .depth = 0,  .num_planes = 1, .cpp = { 4, 0, 0 }, .hsub = 2, .vsub = 1, .is_yuv = true },
diff --git a/include/uapi/drm/drm_fourcc.h b/include/uapi/drm/drm_fourcc.h
index 0128398b7936..13e3c57046be 100644
--- a/include/uapi/drm/drm_fourcc.h
+++ b/include/uapi/drm/drm_fourcc.h
@@ -264,6 +264,7 @@ extern "C" {
 #define DRM_FORMAT_XVUY8888	fourcc_code('X', 'V', 'U', 'Y') /* [31:0] X:Cr:Cb:Y 8:8:8:8 little endian */
 #define DRM_FORMAT_VUY888	fourcc_code('V', 'U', '2', '4') /* [23:0] Cr:Cb:Y 8:8:8 little endian */
 #define DRM_FORMAT_VUY101010	fourcc_code('V', 'U', '3', '0') /* Y followed by U then V, 10:10:10. Non-linear modifier only */
+#define DRM_FORMAT_XVUY2101010	fourcc_code('X', 'Y', '3', '0') /* [31:0] x:Cr:Cb:Y 2:10:10:10 little endian */
 
 /*
  * packed Y2xx indicate for each component, xx valid data occupy msb

-- 
2.43.0

[PATCH v8 07/11] drm: xlnx: zynqmp: Use drm helpers when calculating buffer sizes

[Tomi Valkeinen]

Use drm helpers, drm_format_info_plane_width(),
drm_format_info_plane_height() and drm_format_info_min_pitch() to
calculate sizes for the DMA.

This cleans up the code, but also makes it possible to support more
complex formats (like XV15, XV20).

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
 drivers/gpu/drm/xlnx/zynqmp_disp.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/xlnx/zynqmp_disp.c b/drivers/gpu/drm/xlnx/zynqmp_disp.c
index 80d1e499a18d..b9883ea2d03e 100644
--- a/drivers/gpu/drm/xlnx/zynqmp_disp.c
+++ b/drivers/gpu/drm/xlnx/zynqmp_disp.c
@@ -1116,16 +1116,19 @@ int zynqmp_disp_layer_update(struct zynqmp_disp_layer *layer,
 		return 0;
 
 	for (i = 0; i < info->num_planes; i++) {
-		unsigned int width = state->crtc_w / (i ? info->hsub : 1);
-		unsigned int height = state->crtc_h / (i ? info->vsub : 1);
 		struct zynqmp_disp_layer_dma *dma = &layer->dmas[i];
 		struct dma_async_tx_descriptor *desc;
 		dma_addr_t dma_addr;
+		unsigned int width;
+		unsigned int height;
+
+		width = drm_format_info_plane_width(info, state->crtc_w, i);
+		height = drm_format_info_plane_height(info, state->crtc_h, i);
 
 		dma_addr = drm_fb_dma_get_gem_addr(state->fb, state, i);
 
 		dma->xt.numf = height;
-		dma->sgl.size = width * info->cpp[i];
+		dma->sgl.size = drm_format_info_min_pitch(info, i, width);
 		dma->sgl.icg = state->fb->pitches[i] - dma->sgl.size;
 		dma->xt.src_start = dma_addr;
 		dma->xt.frame_size = 1;

-- 
2.43.0

[PATCH v8 08/11] drm: xlnx: zynqmp: Add support for XV15 & XV20

[Tomi Valkeinen]

Add support for XV15 & XV20 formats.

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Vishal Sagar <vishal.sagar@amd.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
 drivers/gpu/drm/xlnx/zynqmp_disp.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/drivers/gpu/drm/xlnx/zynqmp_disp.c b/drivers/gpu/drm/xlnx/zynqmp_disp.c
index b9883ea2d03e..1dc77f2e4262 100644
--- a/drivers/gpu/drm/xlnx/zynqmp_disp.c
+++ b/drivers/gpu/drm/xlnx/zynqmp_disp.c
@@ -297,6 +297,16 @@ static const struct zynqmp_disp_format avbuf_vid_fmts[] = {
 		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_YV16CI_420,
 		.swap		= true,
 		.sf		= scaling_factors_888,
+	}, {
+		.drm_fmt	= DRM_FORMAT_XV15,
+		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_YV16CI_420_10,
+		.swap		= false,
+		.sf		= scaling_factors_101010,
+	}, {
+		.drm_fmt	= DRM_FORMAT_XV20,
+		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_YV16CI_10,
+		.swap		= false,
+		.sf		= scaling_factors_101010,
 	},
 };
 

-- 
2.43.0

[PATCH v8 09/11] drm: xlnx: zynqmp: Add support for Y8 and Y10_P32

[Tomi Valkeinen]

Add support for Y8 and Y10_P32 formats. We also need to add new csc
matrices for the y-only formats.

Reviewed-by: Vishal Sagar <vishal.sagar@amd.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
 drivers/gpu/drm/xlnx/zynqmp_disp.c | 27 ++++++++++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/xlnx/zynqmp_disp.c b/drivers/gpu/drm/xlnx/zynqmp_disp.c
index 1dc77f2e4262..57bb6d1dd10a 100644
--- a/drivers/gpu/drm/xlnx/zynqmp_disp.c
+++ b/drivers/gpu/drm/xlnx/zynqmp_disp.c
@@ -307,6 +307,16 @@ static const struct zynqmp_disp_format avbuf_vid_fmts[] = {
 		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_YV16CI_10,
 		.swap		= false,
 		.sf		= scaling_factors_101010,
+	}, {
+		.drm_fmt	= DRM_FORMAT_Y8,
+		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_MONO,
+		.swap		= false,
+		.sf		= scaling_factors_888,
+	}, {
+		.drm_fmt	= DRM_FORMAT_Y10_P32,
+		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_YONLY_10,
+		.swap		= false,
+		.sf		= scaling_factors_101010,
 	},
 };
 
@@ -697,6 +707,17 @@ static const u32 csc_sdtv_to_rgb_offsets[] = {
 	0x0, 0x1800, 0x1800
 };
 
+/* In Y-only mode the single Y channel is on the third column  */
+static const u16 csc_sdtv_to_rgb_yonly_matrix[] = {
+	0x0, 0x0, 0x1000,
+	0x0, 0x0, 0x1000,
+	0x0, 0x0, 0x1000,
+};
+
+static const u32 csc_sdtv_to_rgb_yonly_offsets[] = {
+	0x0, 0x0, 0x0
+};
+
 /**
  * zynqmp_disp_blend_set_output_format - Set the output format of the blender
  * @disp: Display controller
@@ -846,7 +867,11 @@ static void zynqmp_disp_blend_layer_enable(struct zynqmp_disp *disp,
 				ZYNQMP_DISP_V_BLEND_LAYER_CONTROL(layer->id),
 				val);
 
-	if (layer->drm_fmt->is_yuv) {
+	if (layer->drm_fmt->format == DRM_FORMAT_Y8 ||
+	    layer->drm_fmt->format == DRM_FORMAT_Y10_P32) {
+		coeffs = csc_sdtv_to_rgb_yonly_matrix;
+		offsets = csc_sdtv_to_rgb_yonly_offsets;
+	} else if (layer->drm_fmt->is_yuv) {
 		coeffs = csc_sdtv_to_rgb_matrix;
 		offsets = csc_sdtv_to_rgb_offsets;
 	} else {

-- 
2.43.0

[PATCH v8 10/11] drm: xlnx: zynqmp: Add support for X403

[Tomi Valkeinen]

Add support for X403 format.

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Vishal Sagar <vishal.sagar@amd.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
 drivers/gpu/drm/xlnx/zynqmp_disp.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/gpu/drm/xlnx/zynqmp_disp.c b/drivers/gpu/drm/xlnx/zynqmp_disp.c
index 57bb6d1dd10a..98105d1c4456 100644
--- a/drivers/gpu/drm/xlnx/zynqmp_disp.c
+++ b/drivers/gpu/drm/xlnx/zynqmp_disp.c
@@ -317,6 +317,11 @@ static const struct zynqmp_disp_format avbuf_vid_fmts[] = {
 		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_YONLY_10,
 		.swap		= false,
 		.sf		= scaling_factors_101010,
+	}, {
+		.drm_fmt	= DRM_FORMAT_X403,
+		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_YV24_10,
+		.swap		= false,
+		.sf		= scaling_factors_101010,
 	},
 };
 

-- 
2.43.0

[PATCH v8 11/11] drm: xlnx: zynqmp: Add support for XVUY2101010

[Tomi Valkeinen]

Add support for XVUY2101010 format.

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Vishal Sagar <vishal.sagar@amd.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
 drivers/gpu/drm/xlnx/zynqmp_disp.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/gpu/drm/xlnx/zynqmp_disp.c b/drivers/gpu/drm/xlnx/zynqmp_disp.c
index 98105d1c4456..a00a57c6dcca 100644
--- a/drivers/gpu/drm/xlnx/zynqmp_disp.c
+++ b/drivers/gpu/drm/xlnx/zynqmp_disp.c
@@ -322,6 +322,11 @@ static const struct zynqmp_disp_format avbuf_vid_fmts[] = {
 		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_YV24_10,
 		.swap		= false,
 		.sf		= scaling_factors_101010,
+	}, {
+		.drm_fmt	= DRM_FORMAT_XVUY2101010,
+		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_YUV444_10,
+		.swap		= false,
+		.sf		= scaling_factors_101010,
 	},
 };
 

-- 
2.43.0

Re: [PATCH v8 03/11] drm/fourcc: Add DRM_FORMAT_Y8

[Laurent Pinchart]

On Wed, Jan 28, 2026 at 07:25:28PM +0200, Tomi Valkeinen wrote:
> Add greyscale Y8 format.
> 
> The 8-bit greyscale format has been discussed before, and the earlier
> guidance was to use DRM_FORMAT_R8, as a single-channel 8-bit pixel.
> 
> However, adding DRM_FORMAT_Y8 makes sense, we can mark it as 'is_yuv' in
> the drm_format_info, and this can help the drivers handle e.g.
> full/limited range. This will distinguish two single-channel formats:
> R8, which is a RGB format with the same value for all components, and
> Y8, which is a Y-only YCbCr format, with Cb and Cr being neutral.
> 
> Acked-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
> Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.com>
> Reviewed-by: Vishal Sagar <vishal.sagar@amd.com>
> Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>

> ---
>  drivers/gpu/drm/drm_fourcc.c  | 1 +
>  include/uapi/drm/drm_fourcc.h | 9 +++++++++
>  2 files changed, 10 insertions(+)
> 
> diff --git a/drivers/gpu/drm/drm_fourcc.c b/drivers/gpu/drm/drm_fourcc.c
> index b22ef86428a1..a39b9d7a5b62 100644
> --- a/drivers/gpu/drm/drm_fourcc.c
> +++ b/drivers/gpu/drm/drm_fourcc.c
> @@ -275,6 +275,7 @@ const struct drm_format_info *__drm_format_info(u32 format)
>  		{ .format = DRM_FORMAT_YVU422,		.depth = 0,  .num_planes = 3, .cpp = { 1, 1, 1 }, .hsub = 2, .vsub = 1, .is_yuv = true },
>  		{ .format = DRM_FORMAT_YUV444,		.depth = 0,  .num_planes = 3, .cpp = { 1, 1, 1 }, .hsub = 1, .vsub = 1, .is_yuv = true },
>  		{ .format = DRM_FORMAT_YVU444,		.depth = 0,  .num_planes = 3, .cpp = { 1, 1, 1 }, .hsub = 1, .vsub = 1, .is_yuv = true },
> +		{ .format = DRM_FORMAT_Y8,		.depth = 8,  .num_planes = 1, .cpp = { 1, 0, 0 }, .hsub = 1, .vsub = 1, .is_yuv = true },
>  		{ .format = DRM_FORMAT_NV12,		.depth = 0,  .num_planes = 2, .cpp = { 1, 2, 0 }, .hsub = 2, .vsub = 2, .is_yuv = true },
>  		{ .format = DRM_FORMAT_NV21,		.depth = 0,  .num_planes = 2, .cpp = { 1, 2, 0 }, .hsub = 2, .vsub = 2, .is_yuv = true },
>  		{ .format = DRM_FORMAT_NV16,		.depth = 0,  .num_planes = 2, .cpp = { 1, 2, 0 }, .hsub = 2, .vsub = 1, .is_yuv = true },
> diff --git a/include/uapi/drm/drm_fourcc.h b/include/uapi/drm/drm_fourcc.h
> index 6c786701238e..e4451668499a 100644
> --- a/include/uapi/drm/drm_fourcc.h
> +++ b/include/uapi/drm/drm_fourcc.h
> @@ -459,6 +459,15 @@ extern "C" {
>  #define DRM_FORMAT_YUV444	fourcc_code('Y', 'U', '2', '4') /* non-subsampled Cb (1) and Cr (2) planes */
>  #define DRM_FORMAT_YVU444	fourcc_code('Y', 'V', '2', '4') /* non-subsampled Cr (1) and Cb (2) planes */
>  
> +/*
> + * Y-only (greyscale) formats
> + *
> + * The Y-only formats are handled similarly to the YCbCr formats in the display
> + * pipeline, with the Cb and Cr implicitly neutral (0.0 in nominal values). This
> + * also means that COLOR_RANGE property applies to the Y-only formats.
> + */
> +
> +#define DRM_FORMAT_Y8		fourcc_code('G', 'R', 'E', 'Y')  /* 8-bit Y-only */
>  
>  /*
>   * Format Modifiers:

-- 
Regards,

Laurent Pinchart

Re: [PATCH v8 09/11] drm: xlnx: zynqmp: Add support for Y8 and Y10_P32

[Laurent Pinchart]

On Wed, Jan 28, 2026 at 07:25:34PM +0200, Tomi Valkeinen wrote:
> Add support for Y8 and Y10_P32 formats. We also need to add new csc
> matrices for the y-only formats.
> 
> Reviewed-by: Vishal Sagar <vishal.sagar@amd.com>
> Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>

> ---
>  drivers/gpu/drm/xlnx/zynqmp_disp.c | 27 ++++++++++++++++++++++++++-
>  1 file changed, 26 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/xlnx/zynqmp_disp.c b/drivers/gpu/drm/xlnx/zynqmp_disp.c
> index 1dc77f2e4262..57bb6d1dd10a 100644
> --- a/drivers/gpu/drm/xlnx/zynqmp_disp.c
> +++ b/drivers/gpu/drm/xlnx/zynqmp_disp.c
> @@ -307,6 +307,16 @@ static const struct zynqmp_disp_format avbuf_vid_fmts[] = {
>  		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_YV16CI_10,
>  		.swap		= false,
>  		.sf		= scaling_factors_101010,
> +	}, {
> +		.drm_fmt	= DRM_FORMAT_Y8,
> +		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_MONO,
> +		.swap		= false,
> +		.sf		= scaling_factors_888,
> +	}, {
> +		.drm_fmt	= DRM_FORMAT_Y10_P32,
> +		.buf_fmt	= ZYNQMP_DISP_AV_BUF_FMT_NL_VID_YONLY_10,
> +		.swap		= false,
> +		.sf		= scaling_factors_101010,
>  	},
>  };
>  
> @@ -697,6 +707,17 @@ static const u32 csc_sdtv_to_rgb_offsets[] = {
>  	0x0, 0x1800, 0x1800
>  };
>  
> +/* In Y-only mode the single Y channel is on the third column  */
> +static const u16 csc_sdtv_to_rgb_yonly_matrix[] = {
> +	0x0, 0x0, 0x1000,
> +	0x0, 0x0, 0x1000,
> +	0x0, 0x0, 0x1000,
> +};
> +
> +static const u32 csc_sdtv_to_rgb_yonly_offsets[] = {
> +	0x0, 0x0, 0x0
> +};
> +
>  /**
>   * zynqmp_disp_blend_set_output_format - Set the output format of the blender
>   * @disp: Display controller
> @@ -846,7 +867,11 @@ static void zynqmp_disp_blend_layer_enable(struct zynqmp_disp *disp,
>  				ZYNQMP_DISP_V_BLEND_LAYER_CONTROL(layer->id),
>  				val);
>  
> -	if (layer->drm_fmt->is_yuv) {
> +	if (layer->drm_fmt->format == DRM_FORMAT_Y8 ||
> +	    layer->drm_fmt->format == DRM_FORMAT_Y10_P32) {
> +		coeffs = csc_sdtv_to_rgb_yonly_matrix;
> +		offsets = csc_sdtv_to_rgb_yonly_offsets;
> +	} else if (layer->drm_fmt->is_yuv) {
>  		coeffs = csc_sdtv_to_rgb_matrix;
>  		offsets = csc_sdtv_to_rgb_offsets;
>  	} else {

-- 
Regards,

Laurent Pinchart

Re: [PATCH v8 00/11] drm: Add new pixel formats for Xilinx Zynqmp

[Tomi Valkeinen]

Hi Maarten, Maxime, Thomas,

We've got multiple R-b's for all of the patches. Can this be merged? Or
give an ack for the fourcc ones and I can push via drm-misc.

 Tomi

On 28/01/2026 19:25, Tomi Valkeinen wrote:
> Add new DRM pixel formats and add support for those in the Xilinx zynqmp
> display driver.
> 
> All other formats except XVUY2101010 are already supported in upstream
> gstreamer, but gstreamer's kmssink does not have the support yet, as it
> obviously cannot support the formats without kernel having the formats.
> 
> Xilinx has support for these formats in their BSP kernel, and Xilinx has
> a branch here, adding the support to gstreamer kmssink:
> 
> https://github.com/Xilinx/gst-plugins-bad.git xlnx-rebase-v1.18.5
> 
> New formats added:
> 
> DRM_FORMAT_Y8
> - 8-bit Y-only
> - fourcc: "GREY"
> - gstreamer: GRAY8
> 
> DRM_FORMAT_Y10_P32
> - 10-bit Y-only, three pixels packed into 32-bits
> - fourcc: "YPA4"
> - gstreamer: GRAY10_LE32
> 
> DRM_FORMAT_XV15
> - Like NV12, but with 10-bit components
> - fourcc: "XV15"
> - gstreamer: NV12_10LE32
> 
> DRM_FORMAT_XV20
> - Like NV16, but with 10-bit components
> - fourcc: "XV20"
> - gstreamer: NV16_10LE32
> 
> DRM_FORMAT_X403
> - 10-bit planar 4:4:4, with three samples packed into 32-bits
> - fourcc: "X403"
> - gstreamer: Y444_10LE32
> 
> XVUY2101010
> - 10-bit 4:4:4, one pixel in 32 bits
> - fourcc: "XY30"
> 
> Some notes:
> 
> I know the 8-bit greyscale format has been discussed before, and the
> guidance was to use DRM_FORMAT_R8. While I'm not totally against that, I
> would argue that adding DRM_FORMAT_Y8 makes sense, as:
> 
> 1) We can mark it as 'is_yuv' in the drm_format_info, and this can help
>    the drivers handle e.g. full/limited range. Probably some hardware
>    handles grayscale as a value used for all RGB components, in which case
>    R8 makes sense, but when the hardware handles the Y-only pixels as YCbCr,
>    where Cb and Cr are "neutral", it makes more sense to consider the
>    format as an YUV format rather than RGB.
> 
> 2) We can have the same fourcc as in v4l2. While not strictly necessary,
>    it's a constant source of confusion when the fourccs differ.
> 
> 3) It (possibly) makes more sense for the user to use Y8/GREY format
>    instead of R8, as, in my experience, the documentation usually refers
>    to gray(scale) format or Y-only format.
> 
> As we add new Y-only formats, it makes sense to have similar terms, so
> we need to adjust the Y10_P32 format name accordingly.
> 
> I have made some adjustments to the formats compared to the Xilinx's
> branch. E.g. The DRM_FORMAT_Y10_P32 format in Xilinx's kmssink uses
> fourcc "Y10 ", and DRM_FORMAT_Y10. I didn't like those, as the format is
> a packed format, three 10-bit pixels in a 32-bit container, and I think
> Y10 means a 10-bit pixel in a 16-bit container.
> 
> Generally speaking, if someone has good ideas for the format define
> names or fourccs, speak up, as it's not easy to invent good names =).
> That said, keeping them the same as in the Xilinx trees will, of course,
> be slightly easier for the users of Xilinx platforms.
> 
> I made WIP additions to modetest to support most of these formats,
> partially based on Xilinx's code:
> 
> https://github.com/tomba/libdrm.git xilinx
> 
> A few thoughts about that:
> 
> modetest uses bo_create_dumb(), and as highlighted in recent discussions
> in the kernel list [1], dumb buffers are only for RGB formats. They may
> work for non-RGB formats, but that's platform specific. None of the
> formats I add here are RGB formats. Do we want to go this way with
> modetest?
> 
> I also feel that the current structure of modetest is not well suited to
> more complicated formats. Both the buffer allocation is a bit more
> difficult (see "Add virtual_width and pixels_per_container"), and the
> drawing is complicated (see, e.g., "Add support for DRM_FORMAT_XV15 &
> DRM_FORMAT_XV20").
> 
> I have recently added support for these Xilinx formats to both kms++ [2] and
> pykms/pixutils [3][4] (WIP), and it's not been easy... But I have to say I
> think I like the template based version in kms++. That won't work in
> modetest, of course, but a non-templated version might be implementable,
> but probably much slower.
> 
> In any case, I slighly feel it's not worth merging the modetest patches
> I have for these formats: they complicate the code quite a bit, break
> the RGB-only rule, and I'm not sure if there really are (m)any users. If
> we want to add support to modetest, I think a bigger rewrite of the test
> pattern code might be in order.
> 
> [1] https://lore.kernel.org/all/20250109150310.219442-26-tzimmermann%40suse.de/
> [2] git@github.com:tomba/kmsxx.git xilinx
> [3] git@github.com:tomba/pykms.git xilinx
> [4] git@github.com:tomba/pixutils.git xilinx
> 
> Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
> ---
> Changes in v8:
> - Expand the "drm/fourcc: Add DRM_FORMAT_Y8" commit description to
>   explain the rationale
> - Add comment to "drm: xlnx: zynqmp: Add support for Y8 and Y10_P32"
>   explainig the Y-only matrix
> - Remove extra blank line
> - Link to v7: https://lore.kernel.org/r/20251201-xilinx-formats-v7-0-1e1558adfefc@ideasonboard.com
> 
> Changes in v7:
> - Added Reviewed-bys
> - Rebased on v6.18
> - Link to v6: https://lore.kernel.org/r/20251001-xilinx-formats-v6-0-014b076b542a@ideasonboard.com
> 
> Changes in v6:
> - Added tags for reviews
> - Rebased on v6.17
> - Link to v5: https://lore.kernel.org/r/20250425-xilinx-formats-v5-0-c74263231630@ideasonboard.com
> 
> Changes in v5:
> - Add comment about Y-only formats, clarifying how the display pipeline
>   handles them (they're handled as YCbCr, with Cb and Cr as "neutral")
> - Clarify X403 format in the patch description
> - Set unused Y-only CSC offsets to 0 (instead of 0x1800).
> - Add R-bs
> - Link to v4: https://lore.kernel.org/r/20250326-xilinx-formats-v4-0-322a300c6d72@ideasonboard.com
> 
> Changes in v4:
> - Reformat the drm_format_info entries a bit
> - Calculate block size only once in drm_format_info_bpp()
> - Declare local variables in separate lines
> - Add review tags
> - Fix commit message referring to Y10_LE32 (should be Y10_P32)
> - Link to v3: https://lore.kernel.org/r/20250212-xilinx-formats-v3-0-90d0fe106995@ideasonboard.com
> 
> Changes in v3:
> - Drop "drm: xlnx: zynqmp: Fix max dma segment size". It is already
>   pushed.
> - Add XVUY2101010 format.
> - Rename DRM_FORMAT_Y10_LE32 to DRM_FORMAT_Y10_P32.
> - Link to v2: https://lore.kernel.org/r/20250115-xilinx-formats-v2-0-160327ca652a@ideasonboard.com
> 
> Changes in v2:
> - I noticed V4L2 already has fourcc Y10P, referring to MIPI-style packed
>   Y10 format. So I changed Y10_LE32 fourcc to YPA4. If logic has any
>   relevance here, P means packed, A means 10, 4 means "in 4 bytes".
> - Added tags to "Fix max dma segment size" patch
> - Updated description for "Add warning for bad bpp"
> - Link to v1: https://lore.kernel.org/r/20241204-xilinx-formats-v1-0-0bf2c5147db1@ideasonboard.com
> 
> ---
> Tomi Valkeinen (11):
>       drm/fourcc: Add warning for bad bpp
>       drm/fourcc: Add DRM_FORMAT_XV15/XV20
>       drm/fourcc: Add DRM_FORMAT_Y8
>       drm/fourcc: Add DRM_FORMAT_Y10_P32
>       drm/fourcc: Add DRM_FORMAT_X403
>       drm/fourcc: Add DRM_FORMAT_XVUY2101010
>       drm: xlnx: zynqmp: Use drm helpers when calculating buffer sizes
>       drm: xlnx: zynqmp: Add support for XV15 & XV20
>       drm: xlnx: zynqmp: Add support for Y8 and Y10_P32
>       drm: xlnx: zynqmp: Add support for X403
>       drm: xlnx: zynqmp: Add support for XVUY2101010
> 
>  drivers/gpu/drm/drm_fourcc.c       | 28 +++++++++++++++++--
>  drivers/gpu/drm/xlnx/zynqmp_disp.c | 56 +++++++++++++++++++++++++++++++++++---
>  include/uapi/drm/drm_fourcc.h      | 28 +++++++++++++++++++
>  3 files changed, 105 insertions(+), 7 deletions(-)
> ---
> base-commit: 7d0a66e4bb9081d75c82ec4957c50034cb0ea449
> change-id: 20241120-xilinx-formats-f71901621833
> 
> Best regards,

Re: [PATCH v8 04/11] drm/fourcc: Add DRM_FORMAT_Y10_P32

[Simon Ser]

On Wednesday, January 28th, 2026 at 18:25, Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> wrote:

> Add Y10_P32, a 10 bit greyscale format, with 3 pixels packed into
> 32-bit container.
> 
> The fourcc for the format is 'YPA4', which comes from Y - Y only, P -
> packed, A - 10 (as in 0xA), 4 - 4 bytes.

I know we aren't super consistent about DRM format names, but… this _P32
suffix doesn't make sense to me: we never had it before, and a lot of
DRM formats are packed (in the Vulkan sense [1]), so I'm not sure why
this one would need it.

What's special about this one is that it's the first (AFAIK) which has
3 pixels per block. Some YCbCr formats are sub-sampled, but always use a
square block.

I would suggest something like YYYX1010102. We use the "channels
followed by bits per component" pattern elsewhere.

[1]: https://docs.vulkan.org/spec/latest/chapters/formats.html#formats-packed

Re: [PATCH v8 02/11] drm/fourcc: Add DRM_FORMAT_XV15/XV20

[Simon Ser]

What is the difference between DRM_FORMAT_XV15 and DRM_FORMAT_P030?

Re: [PATCH v8 06/11] drm/fourcc: Add DRM_FORMAT_XVUY2101010

[Simon Ser]

Reviewed-by: Simon Ser <contact@emersion.fr>

Re: [PATCH v8 02/11] drm/fourcc: Add DRM_FORMAT_XV15/XV20

[Tomi Valkeinen]

Hi Simon, Dave,

On 19/03/2026 17:15, Simon Ser wrote:
> What is the difference between DRM_FORMAT_XV15 and DRM_FORMAT_P030?

Good question. I thought the Cr & Cb are swapped between those formats.

The VC4 driver uses HVS_PIXEL_ORDER_XYCBCR define, and the comment in
drm_fourcc.h starts with "YCbCr420". So, CbCr. But then the comments
then continue with CrCb.

XV15/20 isn't super clear either, starting with "YCrCb" but then "Cb:Cr
plane".

The components in the plane descriptions look identical though, hinting
they are the same. So to be sure, I tested it:

I had test pattern support for XV15/20 in pykms, which I can test on a
Xilinx board. P030 is a bit more difficult, as on RPi5 it requires
DRM_FORMAT_MOD_BROADCOM_SAND128. But with a linear -> SAND128 converter,
I was able to test linear NV12 converted to NV12+SAND128 and shown on
the screen (just for reference), and similarly linear XV15 ->
P030+SAND128. And, indeed, the linear XV15 test pattern shows correctly
on screen when converted to SAND128. So afaics XV15 is indeed identical
to P030, and I can drop XV15.

This then brings up the question about XV20. That format doesn't exist
yet, but should it then be named similarly to P030? However, I have no
idea what P030 means, so I don't know what a 2x1 subsampled P030 would
be called...

Thoughts/ideas? Or just keep XV20?

 Tomi

Re: [PATCH v8 04/11] drm/fourcc: Add DRM_FORMAT_Y10_P32

[Tomi Valkeinen]

Hi,

On 19/03/2026 16:59, Simon Ser wrote:
> On Wednesday, January 28th, 2026 at 18:25, Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> wrote:
> 
>> Add Y10_P32, a 10 bit greyscale format, with 3 pixels packed into
>> 32-bit container.
>>
>> The fourcc for the format is 'YPA4', which comes from Y - Y only, P -
>> packed, A - 10 (as in 0xA), 4 - 4 bytes.
> 
> I know we aren't super consistent about DRM format names, but… this _P32
> suffix doesn't make sense to me: we never had it before, and a lot of
> DRM formats are packed (in the Vulkan sense [1]), so I'm not sure why
> this one would need it.
> 
> What's special about this one is that it's the first (AFAIK) which has
> 3 pixels per block. Some YCbCr formats are sub-sampled, but always use a
> square block.
> 
> I would suggest something like YYYX1010102. We use the "channels
> followed by bits per component" pattern elsewhere.
I'm ok with that. But wouldn't it be XYYY2101010 instead?

And then we just wait for someone to add a x:Y0:Y1:Y2 format (instead of
x:Y2:Y1:Y0). Should we be more specific, and already make this one
XY2Y1Y0_2101010...

 Tomi

Re: [PATCH v8 04/11] drm/fourcc: Add DRM_FORMAT_Y10_P32

[Simon Ser]

On Friday, March 20th, 2026 at 11:48, Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> wrote:

> Hi,
> 
> On 19/03/2026 16:59, Simon Ser wrote:
> > On Wednesday, January 28th, 2026 at 18:25, Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> wrote:
> >
> >> Add Y10_P32, a 10 bit greyscale format, with 3 pixels packed into
> >> 32-bit container.
> >>
> >> The fourcc for the format is 'YPA4', which comes from Y - Y only, P -
> >> packed, A - 10 (as in 0xA), 4 - 4 bytes.
> >
> > I know we aren't super consistent about DRM format names, but… this _P32
> > suffix doesn't make sense to me: we never had it before, and a lot of
> > DRM formats are packed (in the Vulkan sense [1]), so I'm not sure why
> > this one would need it.
> >
> > What's special about this one is that it's the first (AFAIK) which has
> > 3 pixels per block. Some YCbCr formats are sub-sampled, but always use a
> > square block.
> >
> > I would suggest something like YYYX1010102. We use the "channels
> > followed by bits per component" pattern elsewhere.
> I'm ok with that. But wouldn't it be XYYY2101010 instead?

Oh yes, indeed.

> And then we just wait for someone to add a x:Y0:Y1:Y2 format (instead of
> x:Y2:Y1:Y0). Should we be more specific, and already make this one
> XY2Y1Y0_2101010...

Looking at all existing YCbCr formats, they all use the Y2-Y1-Y0 ordering,
none of them use the reverse. So I'd say it's safe enough to just leave it
out and treat it as the default. We can think of a new pattern if/when we
hit the reverse case.

Re: [PATCH v8 02/11] drm/fourcc: Add DRM_FORMAT_XV15/XV20

[Simon Ser]

On Friday, March 20th, 2026 at 11:37, Tomi Valkeinen <tomi.valkeinen@ideasonboard.com> wrote:

> On 19/03/2026 17:15, Simon Ser wrote:
> > What is the difference between DRM_FORMAT_XV15 and DRM_FORMAT_P030?
> 
> Good question. I thought the Cr & Cb are swapped between those formats.
> 
> The VC4 driver uses HVS_PIXEL_ORDER_XYCBCR define, and the comment in
> drm_fourcc.h starts with "YCbCr420". So, CbCr. But then the comments
> then continue with CrCb.
> 
> XV15/20 isn't super clear either, starting with "YCrCb" but then "Cb:Cr
> plane".
> 
> The components in the plane descriptions look identical though, hinting
> they are the same. So to be sure, I tested it:
> 
> I had test pattern support for XV15/20 in pykms, which I can test on a
> Xilinx board. P030 is a bit more difficult, as on RPi5 it requires
> DRM_FORMAT_MOD_BROADCOM_SAND128. But with a linear -> SAND128 converter,
> I was able to test linear NV12 converted to NV12+SAND128 and shown on
> the screen (just for reference), and similarly linear XV15 ->
> P030+SAND128. And, indeed, the linear XV15 test pattern shows correctly
> on screen when converted to SAND128. So afaics XV15 is indeed identical
> to P030, and I can drop XV15.
> 
> This then brings up the question about XV20. That format doesn't exist
> yet, but should it then be named similarly to P030? However, I have no
> idea what P030 means, so I don't know what a 2x1 subsampled P030 would
> be called...
> 
> Thoughts/ideas? Or just keep XV20?

I've replied with some ideas on the new version's thread :)