media: amphion: tell and handle contiguous and non contiguous format

Driver should tell the number of memory planes and component planes.
the amphion vpu support non contiguous planes,
but for compatibility with other device
that only support contiguous planes.
driver can add support for contiguous planes in the same time.
Then the mem_planes can be different from the comp_planes.
driver need to handle buffer according mem_planes and comp_planes.

So driver can support NV12 and NV12M.
Signed-off-by: Ming Qian <ming.qian@nxp.com>
Reviewed-by: Tommaso Merciai <tommaso.merciai@amarulasolutions.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>

drivers/media/platform/amphion/venc.c

@@ -69,13 +69,16 @@ struct venc_frame_t {
 static const struct vpu_format venc_formats[] = {
 	{
 		.pixfmt = V4L2_PIX_FMT_NV12M,
-		.num_planes = 2,
+		.mem_planes = 2,
+		.comp_planes = 2,
 		.type = V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE,
 	},
 	{
 		.pixfmt = V4L2_PIX_FMT_H264,
-		.num_planes = 1,
+		.mem_planes = 1,
+		.comp_planes = 1,
 		.type = V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE,
+		.flags = V4L2_FMT_FLAG_COMPRESSED
 	},
 	{0, 0, 0, 0},
 };
@@ -173,14 +176,14 @@ static int venc_g_fmt(struct file *file, void *fh, struct v4l2_format *f)
 	cur_fmt = vpu_get_format(inst, f->type);
 
 	pixmp->pixelformat = cur_fmt->pixfmt;
-	pixmp->num_planes = cur_fmt->num_planes;
+	pixmp->num_planes = cur_fmt->mem_planes;
 	pixmp->width = cur_fmt->width;
 	pixmp->height = cur_fmt->height;
 	pixmp->field = cur_fmt->field;
 	pixmp->flags = cur_fmt->flags;
 	for (i = 0; i < pixmp->num_planes; i++) {
 		pixmp->plane_fmt[i].bytesperline = cur_fmt->bytesperline[i];
-		pixmp->plane_fmt[i].sizeimage = cur_fmt->sizeimage[i];
+		pixmp->plane_fmt[i].sizeimage = vpu_get_fmt_plane_size(cur_fmt, i);
 	}
 
 	f->fmt.pix_mp.colorspace = venc->params.color.primaries;
@@ -194,8 +197,9 @@ static int venc_g_fmt(struct file *file, void *fh, struct v4l2_format *f)
 static int venc_try_fmt(struct file *file, void *fh, struct v4l2_format *f)
 {
 	struct vpu_inst *inst = to_inst(file);
+	struct vpu_format fmt;
 
-	vpu_try_fmt_common(inst, f);
+	vpu_try_fmt_common(inst, f, &fmt);
 
 	return 0;
 }
@@ -203,12 +207,11 @@ static int venc_try_fmt(struct file *file, void *fh, struct v4l2_format *f)
 static int venc_s_fmt(struct file *file, void *fh, struct v4l2_format *f)
 {
 	struct vpu_inst *inst = to_inst(file);
-	const struct vpu_format *fmt;
+	struct vpu_format fmt;
 	struct vpu_format *cur_fmt;
 	struct vb2_queue *q;
 	struct venc_t *venc = inst->priv;
 	struct v4l2_pix_format_mplane *pix_mp = &f->fmt.pix_mp;
-	int i;
 
 	q = v4l2_m2m_get_vq(inst->fh.m2m_ctx, f->type);
 	if (!q)
@@ -216,24 +219,12 @@ static int venc_s_fmt(struct file *file, void *fh, struct v4l2_format *f)
 	if (vb2_is_busy(q))
 		return -EBUSY;
 
-	fmt = vpu_try_fmt_common(inst, f);
-	if (!fmt)
+	if (vpu_try_fmt_common(inst, f, &fmt))
 		return -EINVAL;
 
 	cur_fmt = vpu_get_format(inst, f->type);
 
-	cur_fmt->pixfmt = fmt->pixfmt;
-	cur_fmt->num_planes = fmt->num_planes;
-	cur_fmt->flags = fmt->flags;
-	cur_fmt->width = pix_mp->width;
-	cur_fmt->height = pix_mp->height;
-	for (i = 0; i < fmt->num_planes; i++) {
-		cur_fmt->sizeimage[i] = pix_mp->plane_fmt[i].sizeimage;
-		cur_fmt->bytesperline[i] = pix_mp->plane_fmt[i].bytesperline;
-	}
-
-	if (pix_mp->field != V4L2_FIELD_ANY)
-		cur_fmt->field = pix_mp->field;
+	memcpy(cur_fmt, &fmt, sizeof(*cur_fmt));
 
 	if (V4L2_TYPE_IS_OUTPUT(f->type)) {
 		venc->params.input_format = cur_fmt->pixfmt;

drivers/media/platform/amphion/vpu.h

@@ -84,7 +84,8 @@ struct vpu_dev {
 
 struct vpu_format {
 	u32 pixfmt;
-	unsigned int num_planes;
+	u32 mem_planes;
+	u32 comp_planes;
 	u32 type;
 	u32 flags;
 	u32 width;
@@ -92,6 +93,7 @@ struct vpu_format {
 	u32 sizeimage[VIDEO_MAX_PLANES];
 	u32 bytesperline[VIDEO_MAX_PLANES];
 	u32 field;
+	u32 sibling;
 };
 
 struct vpu_core_resources {

drivers/media/platform/amphion/vpu_dbg.c

@@ -90,9 +90,9 @@ static int vpu_dbg_instance(struct seq_file *s, void *data)
 			vq->last_buffer_dequeued);
 	if (seq_write(s, str, num))
 		return 0;
-	for (i = 0; i < inst->out_format.num_planes; i++) {
+	for (i = 0; i < inst->out_format.mem_planes; i++) {
 		num = scnprintf(str, sizeof(str), " %d(%d)",
-				inst->out_format.sizeimage[i],
+				vpu_get_fmt_plane_size(&inst->out_format, i),
 				inst->out_format.bytesperline[i]);
 		if (seq_write(s, str, num))
 			return 0;
@@ -114,9 +114,9 @@ static int vpu_dbg_instance(struct seq_file *s, void *data)
 			vq->last_buffer_dequeued);
 	if (seq_write(s, str, num))
 		return 0;
-	for (i = 0; i < inst->cap_format.num_planes; i++) {
+	for (i = 0; i < inst->cap_format.mem_planes; i++) {
 		num = scnprintf(str, sizeof(str), " %d(%d)",
-				inst->cap_format.sizeimage[i],
+				vpu_get_fmt_plane_size(&inst->cap_format, i),
 				inst->cap_format.bytesperline[i]);
 		if (seq_write(s, str, num))
 			return 0;

drivers/media/platform/amphion/vpu_helpers.c

@@ -59,6 +59,36 @@ const struct vpu_format *vpu_helper_find_format(struct vpu_inst *inst, u32 type,
 	return NULL;
 }
 
+const struct vpu_format *vpu_helper_find_sibling(struct vpu_inst *inst, u32 type, u32 pixelfmt)
+{
+	const struct vpu_format *fmt;
+	const struct vpu_format *sibling;
+
+	fmt = vpu_helper_find_format(inst, type, pixelfmt);
+	if (!fmt || !fmt->sibling)
+		return NULL;
+
+	sibling = vpu_helper_find_format(inst, type, fmt->sibling);
+	if (!sibling || sibling->sibling != fmt->pixfmt ||
+	    sibling->comp_planes != fmt->comp_planes)
+		return NULL;
+
+	return sibling;
+}
+
+bool vpu_helper_match_format(struct vpu_inst *inst, u32 type, u32 fmta, u32 fmtb)
+{
+	const struct vpu_format *sibling;
+
+	if (fmta == fmtb)
+		return true;
+
+	sibling = vpu_helper_find_sibling(inst, type, fmta);
+	if (sibling && sibling->pixfmt == fmtb)
+		return true;
+	return false;
+}
+
 const struct vpu_format *vpu_helper_enum_format(struct vpu_inst *inst, u32 type, int index)
 {
 	const struct vpu_format *pfmt;
@@ -123,9 +153,10 @@ static u32 get_nv12_plane_size(u32 width, u32 height, int plane_no,
 	u32 bytesperline;
 	u32 size = 0;
 
-	bytesperline = ALIGN(width, stride);
+	bytesperline = width;
 	if (pbl)
 		bytesperline = max(bytesperline, *pbl);
+	bytesperline = ALIGN(bytesperline, stride);
 	height = ALIGN(height, 2);
 	if (plane_no == 0)
 		size = bytesperline * height;
@@ -148,13 +179,13 @@ static u32 get_tiled_8l128_plane_size(u32 fmt, u32 width, u32 height, int plane_
 
 	if (interlaced)
 		hs++;
-	if (fmt == V4L2_PIX_FMT_NV12M_10BE_8L128)
+	if (fmt == V4L2_PIX_FMT_NV12M_10BE_8L128 || fmt == V4L2_PIX_FMT_NV12_10BE_8L128)
 		bitdepth = 10;
 	bytesperline = DIV_ROUND_UP(width * bitdepth, BITS_PER_BYTE);
-	bytesperline = ALIGN(bytesperline, 1 << ws);
-	bytesperline = ALIGN(bytesperline, stride);
 	if (pbl)
 		bytesperline = max(bytesperline, *pbl);
+	bytesperline = ALIGN(bytesperline, 1 << ws);
+	bytesperline = ALIGN(bytesperline, stride);
 	height = ALIGN(height, 1 << hs);
 	if (plane_no == 0)
 		size = bytesperline * height;
@@ -172,9 +203,10 @@ static u32 get_default_plane_size(u32 width, u32 height, int plane_no,
 	u32 bytesperline;
 	u32 size = 0;
 
-	bytesperline = ALIGN(width, stride);
+	bytesperline = width;
 	if (pbl)
 		bytesperline = max(bytesperline, *pbl);
+	bytesperline = ALIGN(bytesperline, stride);
 	if (plane_no == 0)
 		size = bytesperline * height;
 	if (pbl)
@@ -187,9 +219,12 @@ u32 vpu_helper_get_plane_size(u32 fmt, u32 w, u32 h, int plane_no,
 			      u32 stride, u32 interlaced, u32 *pbl)
 {
 	switch (fmt) {
+	case V4L2_PIX_FMT_NV12:
 	case V4L2_PIX_FMT_NV12M:
 		return get_nv12_plane_size(w, h, plane_no, stride, interlaced, pbl);
+	case V4L2_PIX_FMT_NV12_8L128:
 	case V4L2_PIX_FMT_NV12M_8L128:
+	case V4L2_PIX_FMT_NV12_10BE_8L128:
 	case V4L2_PIX_FMT_NV12M_10BE_8L128:
 		return get_tiled_8l128_plane_size(fmt, w, h, plane_no, stride, interlaced, pbl);
 	default:

drivers/media/platform/amphion/vpu_helpers.h

@@ -14,6 +14,8 @@ struct vpu_pair {
 int vpu_helper_find_in_array_u8(const u8 *array, u32 size, u32 x);
 bool vpu_helper_check_type(struct vpu_inst *inst, u32 type);
 const struct vpu_format *vpu_helper_find_format(struct vpu_inst *inst, u32 type, u32 pixelfmt);
+const struct vpu_format *vpu_helper_find_sibling(struct vpu_inst *inst, u32 type, u32 pixelfmt);
+bool vpu_helper_match_format(struct vpu_inst *inst, u32 type, u32 fmta, u32 fmtb);
 const struct vpu_format *vpu_helper_enum_format(struct vpu_inst *inst, u32 type, int index);
 u32 vpu_helper_valid_frame_width(struct vpu_inst *inst, u32 width);
 u32 vpu_helper_valid_frame_height(struct vpu_inst *inst, u32 height);

drivers/media/platform/amphion/vpu_malone.c

@@ -583,7 +583,8 @@ bool vpu_malone_check_fmt(enum vpu_core_type type, u32 pixelfmt)
 	if (!vpu_imx8q_check_fmt(type, pixelfmt))
 		return false;
 
-	if (pixelfmt == V4L2_PIX_FMT_NV12M_8L128 || pixelfmt == V4L2_PIX_FMT_NV12M_10BE_8L128)
+	if (pixelfmt == V4L2_PIX_FMT_NV12_8L128 || pixelfmt == V4L2_PIX_FMT_NV12_10BE_8L128 ||
+	    pixelfmt == V4L2_PIX_FMT_NV12M_8L128 || pixelfmt == V4L2_PIX_FMT_NV12M_10BE_8L128)
 		return true;
 	if (vpu_malone_format_remap(pixelfmt) == MALONE_FMT_NULL)
 		return false;

drivers/media/platform/amphion/vpu_v4l2.c

@@ -133,51 +133,136 @@ bool vpu_is_source_empty(struct vpu_inst *inst)
 	return true;
 }
 
-const struct vpu_format *vpu_try_fmt_common(struct vpu_inst *inst, struct v4l2_format *f)
+static int vpu_init_format(struct vpu_inst *inst, struct vpu_format *fmt)
+{
+	const struct vpu_format *info;
+
+	info = vpu_helper_find_format(inst, fmt->type, fmt->pixfmt);
+	if (!info) {
+		info = vpu_helper_enum_format(inst, fmt->type, 0);
+		if (!info)
+			return -EINVAL;
+	}
+	memcpy(fmt, info, sizeof(*fmt));
+
+	return 0;
+}
+
+static int vpu_calc_fmt_bytesperline(struct v4l2_format *f, struct vpu_format *fmt)
 {
 	struct v4l2_pix_format_mplane *pixmp = &f->fmt.pix_mp;
-	u32 type = f->type;
+	int i;
+
+	if (fmt->flags & V4L2_FMT_FLAG_COMPRESSED) {
+		for (i = 0; i < fmt->comp_planes; i++)
+			fmt->bytesperline[i] = 0;
+		return 0;
+	}
+	if (pixmp->num_planes == fmt->comp_planes) {
+		for (i = 0; i < fmt->comp_planes; i++)
+			fmt->bytesperline[i] = pixmp->plane_fmt[i].bytesperline;
+		return 0;
+	}
+	if (pixmp->num_planes > 1)
+		return -EINVAL;
+
+	/*amphion vpu only support nv12 and nv12 tiled,
+	 * so the bytesperline of luma and chroma should be same
+	 */
+	for (i = 0; i < fmt->comp_planes; i++)
+		fmt->bytesperline[i] = pixmp->plane_fmt[0].bytesperline;
+
+	return 0;
+}
+
+static int vpu_calc_fmt_sizeimage(struct vpu_inst *inst, struct vpu_format *fmt)
+{
 	u32 stride = 1;
-	u32 bytesperline;
-	u32 sizeimage;
-	const struct vpu_format *fmt;
-	const struct vpu_core_resources *res;
 	int i;
 
-	fmt = vpu_helper_find_format(inst, type, pixmp->pixelformat);
-	if (!fmt) {
-		fmt = vpu_helper_enum_format(inst, type, 0);
-		if (!fmt)
-			return NULL;
-		pixmp->pixelformat = fmt->pixfmt;
+	if (!(fmt->flags & V4L2_FMT_FLAG_COMPRESSED)) {
+		const struct vpu_core_resources *res = vpu_get_resource(inst);
+
+		if (res)
+			stride = res->stride;
 	}
 
-	res = vpu_get_resource(inst);
-	if (res)
-		stride = res->stride;
-	if (pixmp->width)
-		pixmp->width = vpu_helper_valid_frame_width(inst, pixmp->width);
-	if (pixmp->height)
-		pixmp->height = vpu_helper_valid_frame_height(inst, pixmp->height);
+	for (i = 0; i < fmt->comp_planes; i++) {
+		fmt->sizeimage[i] = vpu_helper_get_plane_size(fmt->pixfmt,
+							      fmt->width,
+							      fmt->height,
+							      i,
+							      stride,
+							      fmt->field != V4L2_FIELD_NONE ? 1 : 0,
+							      &fmt->bytesperline[i]);
+		fmt->sizeimage[i] = max_t(u32, fmt->sizeimage[i], PAGE_SIZE);
+		if (fmt->flags & V4L2_FMT_FLAG_COMPRESSED) {
+			fmt->sizeimage[i] = clamp_val(fmt->sizeimage[i], SZ_128K, SZ_8M);
+			fmt->bytesperline[i] = 0;
+		}
+	}
+
+	return 0;
+}
+
+u32 vpu_get_fmt_plane_size(struct vpu_format *fmt, u32 plane_no)
+{
+	u32 size;
+	int i;
+
+	if (plane_no >= fmt->mem_planes)
+		return 0;
+
+	if (fmt->comp_planes == fmt->mem_planes)
+		return fmt->sizeimage[plane_no];
+	if (plane_no < fmt->mem_planes - 1)
+		return fmt->sizeimage[plane_no];
+
+	size = fmt->sizeimage[plane_no];
+	for (i = fmt->mem_planes; i < fmt->comp_planes; i++)
+		size += fmt->sizeimage[i];
+
+	return size;
+}
+
+int vpu_try_fmt_common(struct vpu_inst *inst, struct v4l2_format *f, struct vpu_format *fmt)
+{
+	struct v4l2_pix_format_mplane *pixmp = &f->fmt.pix_mp;
+	int i;
+	int ret;
+
+	fmt->pixfmt = pixmp->pixelformat;
+	fmt->type = f->type;
+	ret = vpu_init_format(inst, fmt);
+	if (ret < 0)
+		return ret;
+
+	fmt->width = pixmp->width;
+	fmt->height = pixmp->height;
+	if (fmt->width)
+		fmt->width = vpu_helper_valid_frame_width(inst, fmt->width);
+	if (fmt->height)
+		fmt->height = vpu_helper_valid_frame_height(inst, fmt->height);
+	fmt->field = pixmp->field == V4L2_FIELD_ANY ? V4L2_FIELD_NONE : pixmp->field;
+	vpu_calc_fmt_bytesperline(f, fmt);
+	vpu_calc_fmt_sizeimage(inst, fmt);
+	if ((fmt->flags & V4L2_FMT_FLAG_COMPRESSED) && pixmp->plane_fmt[0].sizeimage)
+		fmt->sizeimage[0] = clamp_val(pixmp->plane_fmt[0].sizeimage, SZ_128K, SZ_8M);
+
+	pixmp->pixelformat = fmt->pixfmt;
+	pixmp->width = fmt->width;
+	pixmp->height = fmt->height;
 	pixmp->flags = fmt->flags;
-	pixmp->num_planes = fmt->num_planes;
-	if (pixmp->field == V4L2_FIELD_ANY)
-		pixmp->field = V4L2_FIELD_NONE;
+	pixmp->num_planes = fmt->mem_planes;
+	pixmp->field = fmt->field;
+	memset(pixmp->reserved, 0, sizeof(pixmp->reserved));
 	for (i = 0; i < pixmp->num_planes; i++) {
-		bytesperline = max_t(s32, pixmp->plane_fmt[i].bytesperline, 0);
-		sizeimage = vpu_helper_get_plane_size(pixmp->pixelformat,
-						      pixmp->width,
-						      pixmp->height,
-						      i,
-						      stride,
-						      pixmp->field > V4L2_FIELD_NONE ? 1 : 0,
-						      &bytesperline);
-		sizeimage = max_t(s32, pixmp->plane_fmt[i].sizeimage, sizeimage);
-		pixmp->plane_fmt[i].bytesperline = bytesperline;
-		pixmp->plane_fmt[i].sizeimage = sizeimage;
+		pixmp->plane_fmt[i].bytesperline = fmt->bytesperline[i];
+		pixmp->plane_fmt[i].sizeimage = vpu_get_fmt_plane_size(fmt, i);
+		memset(pixmp->plane_fmt[i].reserved, 0, sizeof(pixmp->plane_fmt[i].reserved));
 	}
 
-	return fmt;
+	return 0;
 }
 
 static bool vpu_check_ready(struct vpu_inst *inst, u32 type)
@@ -382,10 +467,10 @@ static int vpu_vb2_queue_setup(struct vb2_queue *vq,
 	cur_fmt = vpu_get_format(inst, vq->type);
 
 	if (*plane_count) {
-		if (*plane_count != cur_fmt->num_planes)
+		if (*plane_count != cur_fmt->mem_planes)
 			return -EINVAL;
-		for (i = 0; i < cur_fmt->num_planes; i++) {
-			if (psize[i] < cur_fmt->sizeimage[i])
+		for (i = 0; i < cur_fmt->mem_planes; i++) {
+			if (psize[i] < vpu_get_fmt_plane_size(cur_fmt, i))
 				return -EINVAL;
 		}
 		return 0;
@@ -395,9 +480,9 @@ static int vpu_vb2_queue_setup(struct vb2_queue *vq,
 		*buf_count = max_t(unsigned int, *buf_count, inst->min_buffer_out);
 	else
 		*buf_count = max_t(unsigned int, *buf_count, inst->min_buffer_cap);
-	*plane_count = cur_fmt->num_planes;
-	for (i = 0; i < cur_fmt->num_planes; i++)
-		psize[i] = cur_fmt->sizeimage[i];
+	*plane_count = cur_fmt->mem_planes;
+	for (i = 0; i < cur_fmt->mem_planes; i++)
+		psize[i] = vpu_get_fmt_plane_size(cur_fmt, i);
 
 	return 0;
 }
@@ -427,8 +512,8 @@ static int vpu_vb2_buf_prepare(struct vb2_buffer *vb)
 	u32 i;
 
 	cur_fmt = vpu_get_format(inst, vb->type);
-	for (i = 0; i < cur_fmt->num_planes; i++) {
-		if (vpu_get_vb_length(vb, i) < cur_fmt->sizeimage[i]) {
+	for (i = 0; i < cur_fmt->mem_planes; i++) {
+		if (vpu_get_vb_length(vb, i) < vpu_get_fmt_plane_size(cur_fmt, i)) {
 			dev_dbg(inst->dev, "[%d] %s buf[%d] is invalid\n",
 				inst->id, vpu_type_name(vb->type), vb->index);
 			vpu_set_buffer_state(vbuf, VPU_BUF_STATE_ERROR);

drivers/media/platform/amphion/vpu_v4l2.h

@@ -16,7 +16,8 @@ unsigned int vpu_get_buffer_state(struct vb2_v4l2_buffer *vbuf);
 int vpu_v4l2_open(struct file *file, struct vpu_inst *inst);
 int vpu_v4l2_close(struct file *file);
 
-const struct vpu_format *vpu_try_fmt_common(struct vpu_inst *inst, struct v4l2_format *f);
+u32 vpu_get_fmt_plane_size(struct vpu_format *fmt, u32 plane_no);
+int vpu_try_fmt_common(struct vpu_inst *inst, struct v4l2_format *f, struct vpu_format *fmt);
 int vpu_process_output_buffer(struct vpu_inst *inst);
 int vpu_process_capture_buffer(struct vpu_inst *inst);
 struct vb2_v4l2_buffer *vpu_next_src_buf(struct vpu_inst *inst);

drivers/media/platform/amphion/vpu_windsor.c

@@ -775,6 +775,8 @@ static int vpu_windsor_fill_yuv_frame(struct vpu_shared_addr *shared,
 				      u32 instance,
 				      struct vb2_buffer *vb)
 {
+	struct vpu_inst *inst = vb2_get_drv_priv(vb->vb2_queue);
+	struct vpu_format *out_fmt;
 	struct vpu_enc_yuv_desc *desc;
 	struct vb2_v4l2_buffer *vbuf;
 
@@ -782,6 +784,7 @@ static int vpu_windsor_fill_yuv_frame(struct vpu_shared_addr *shared,
 		return -EINVAL;
 
 	desc = get_yuv_desc(shared, instance);
+	out_fmt = vpu_get_format(inst, vb->type);
 
 	vbuf = to_vb2_v4l2_buffer(vb);
 	desc->frame_id = vbuf->sequence;
@@ -790,7 +793,10 @@ static int vpu_windsor_fill_yuv_frame(struct vpu_shared_addr *shared,
 	else
 		desc->key_frame = 0;
 	desc->luma_base = vpu_get_vb_phy_addr(vb, 0);
-	desc->chroma_base = vpu_get_vb_phy_addr(vb, 1);
+	if (vb->num_planes > 1)
+		desc->chroma_base = vpu_get_vb_phy_addr(vb, 1);
+	else
+		desc->chroma_base = desc->luma_base + out_fmt->sizeimage[0];
 
 	return 0;
 }