1 // SPDX-License-Identifier: GPL-2.0+
2 /*
3  * vsp1_rpf.c  --  R-Car VSP1 Read Pixel Formatter
4  *
5  * Copyright (C) 2013-2014 Renesas Electronics Corporation
6  *
7  * Contact: Laurent Pinchart (laurent.pinchart@ideasonboard.com)
8  */
9 
10 #include <linux/device.h>
11 
12 #include <media/v4l2-subdev.h>
13 
14 #include "vsp1.h"
15 #include "vsp1_dl.h"
16 #include "vsp1_pipe.h"
17 #include "vsp1_rwpf.h"
18 #include "vsp1_video.h"
19 
20 #define RPF_MAX_WIDTH				8190
21 #define RPF_MAX_HEIGHT				8190
22 
23 /* Pre extended display list command data structure. */
24 struct vsp1_extcmd_auto_fld_body {
25 	u32 top_y0;
26 	u32 bottom_y0;
27 	u32 top_c0;
28 	u32 bottom_c0;
29 	u32 top_c1;
30 	u32 bottom_c1;
31 	u32 reserved0;
32 	u32 reserved1;
33 } __packed;
34 
35 /* -----------------------------------------------------------------------------
36  * Device Access
37  */
38 
vsp1_rpf_write(struct vsp1_rwpf * rpf,struct vsp1_dl_body * dlb,u32 reg,u32 data)39 static inline void vsp1_rpf_write(struct vsp1_rwpf *rpf,
40 				  struct vsp1_dl_body *dlb, u32 reg, u32 data)
41 {
42 	vsp1_dl_body_write(dlb, reg + rpf->entity.index * VI6_RPF_OFFSET,
43 			       data);
44 }
45 
46 /* -----------------------------------------------------------------------------
47  * V4L2 Subdevice Operations
48  */
49 
50 static const struct v4l2_subdev_ops rpf_ops = {
51 	.pad    = &vsp1_rwpf_pad_ops,
52 };
53 
54 /* -----------------------------------------------------------------------------
55  * VSP1 Entity Operations
56  */
57 
rpf_configure_stream(struct vsp1_entity * entity,struct vsp1_pipeline * pipe,struct vsp1_dl_body * dlb)58 static void rpf_configure_stream(struct vsp1_entity *entity,
59 				 struct vsp1_pipeline *pipe,
60 				 struct vsp1_dl_body *dlb)
61 {
62 	struct vsp1_rwpf *rpf = to_rwpf(&entity->subdev);
63 	const struct vsp1_format_info *fmtinfo = rpf->fmtinfo;
64 	const struct v4l2_pix_format_mplane *format = &rpf->format;
65 	const struct v4l2_mbus_framefmt *source_format;
66 	const struct v4l2_mbus_framefmt *sink_format;
67 	unsigned int left = 0;
68 	unsigned int top = 0;
69 	u32 pstride;
70 	u32 infmt;
71 
72 	/* Stride */
73 	pstride = format->plane_fmt[0].bytesperline
74 		<< VI6_RPF_SRCM_PSTRIDE_Y_SHIFT;
75 	if (format->num_planes > 1)
76 		pstride |= format->plane_fmt[1].bytesperline
77 			<< VI6_RPF_SRCM_PSTRIDE_C_SHIFT;
78 
79 	/*
80 	 * pstride has both STRIDE_Y and STRIDE_C, but multiplying the whole
81 	 * of pstride by 2 is conveniently OK here as we are multiplying both
82 	 * values.
83 	 */
84 	if (pipe->interlaced)
85 		pstride *= 2;
86 
87 	vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_PSTRIDE, pstride);
88 
89 	/* Format */
90 	sink_format = vsp1_entity_get_pad_format(&rpf->entity,
91 						 rpf->entity.config,
92 						 RWPF_PAD_SINK);
93 	source_format = vsp1_entity_get_pad_format(&rpf->entity,
94 						   rpf->entity.config,
95 						   RWPF_PAD_SOURCE);
96 
97 	infmt = VI6_RPF_INFMT_CIPM
98 	      | (fmtinfo->hwfmt << VI6_RPF_INFMT_RDFMT_SHIFT);
99 
100 	if (fmtinfo->swap_yc)
101 		infmt |= VI6_RPF_INFMT_SPYCS;
102 	if (fmtinfo->swap_uv)
103 		infmt |= VI6_RPF_INFMT_SPUVS;
104 
105 	if (sink_format->code != source_format->code)
106 		infmt |= VI6_RPF_INFMT_CSC;
107 
108 	vsp1_rpf_write(rpf, dlb, VI6_RPF_INFMT, infmt);
109 	vsp1_rpf_write(rpf, dlb, VI6_RPF_DSWAP, fmtinfo->swap);
110 
111 	/* Output location */
112 	if (pipe->brx) {
113 		const struct v4l2_rect *compose;
114 
115 		compose = vsp1_entity_get_pad_selection(pipe->brx,
116 							pipe->brx->config,
117 							rpf->brx_input,
118 							V4L2_SEL_TGT_COMPOSE);
119 		left = compose->left;
120 		top = compose->top;
121 	}
122 
123 	if (pipe->interlaced)
124 		top /= 2;
125 
126 	vsp1_rpf_write(rpf, dlb, VI6_RPF_LOC,
127 		       (left << VI6_RPF_LOC_HCOORD_SHIFT) |
128 		       (top << VI6_RPF_LOC_VCOORD_SHIFT));
129 
130 	/*
131 	 * On Gen2 use the alpha channel (extended to 8 bits) when available or
132 	 * a fixed alpha value set through the V4L2_CID_ALPHA_COMPONENT control
133 	 * otherwise.
134 	 *
135 	 * The Gen3 RPF has extended alpha capability and can both multiply the
136 	 * alpha channel by a fixed global alpha value, and multiply the pixel
137 	 * components to convert the input to premultiplied alpha.
138 	 *
139 	 * As alpha premultiplication is available in the BRx for both Gen2 and
140 	 * Gen3 we handle it there and use the Gen3 alpha multiplier for global
141 	 * alpha multiplication only. This however prevents conversion to
142 	 * premultiplied alpha if no BRx is present in the pipeline. If that use
143 	 * case turns out to be useful we will revisit the implementation (for
144 	 * Gen3 only).
145 	 *
146 	 * We enable alpha multiplication on Gen3 using the fixed alpha value
147 	 * set through the V4L2_CID_ALPHA_COMPONENT control when the input
148 	 * contains an alpha channel. On Gen2 the global alpha is ignored in
149 	 * that case.
150 	 *
151 	 * In all cases, disable color keying.
152 	 */
153 	vsp1_rpf_write(rpf, dlb, VI6_RPF_ALPH_SEL, VI6_RPF_ALPH_SEL_AEXT_EXT |
154 		       (fmtinfo->alpha ? VI6_RPF_ALPH_SEL_ASEL_PACKED
155 				       : VI6_RPF_ALPH_SEL_ASEL_FIXED));
156 
157 	if (entity->vsp1->info->gen == 3) {
158 		u32 mult;
159 
160 		if (fmtinfo->alpha) {
161 			/*
162 			 * When the input contains an alpha channel enable the
163 			 * alpha multiplier. If the input is premultiplied we
164 			 * need to multiply both the alpha channel and the pixel
165 			 * components by the global alpha value to keep them
166 			 * premultiplied. Otherwise multiply the alpha channel
167 			 * only.
168 			 */
169 			bool premultiplied = format->flags
170 					   & V4L2_PIX_FMT_FLAG_PREMUL_ALPHA;
171 
172 			mult = VI6_RPF_MULT_ALPHA_A_MMD_RATIO
173 			     | (premultiplied ?
174 				VI6_RPF_MULT_ALPHA_P_MMD_RATIO :
175 				VI6_RPF_MULT_ALPHA_P_MMD_NONE);
176 		} else {
177 			/*
178 			 * When the input doesn't contain an alpha channel the
179 			 * global alpha value is applied in the unpacking unit,
180 			 * the alpha multiplier isn't needed and must be
181 			 * disabled.
182 			 */
183 			mult = VI6_RPF_MULT_ALPHA_A_MMD_NONE
184 			     | VI6_RPF_MULT_ALPHA_P_MMD_NONE;
185 		}
186 
187 		rpf->mult_alpha = mult;
188 	}
189 
190 	vsp1_rpf_write(rpf, dlb, VI6_RPF_MSK_CTRL, 0);
191 	vsp1_rpf_write(rpf, dlb, VI6_RPF_CKEY_CTRL, 0);
192 
193 }
194 
vsp1_rpf_configure_autofld(struct vsp1_rwpf * rpf,struct vsp1_dl_list * dl)195 static void vsp1_rpf_configure_autofld(struct vsp1_rwpf *rpf,
196 				       struct vsp1_dl_list *dl)
197 {
198 	const struct v4l2_pix_format_mplane *format = &rpf->format;
199 	struct vsp1_dl_ext_cmd *cmd;
200 	struct vsp1_extcmd_auto_fld_body *auto_fld;
201 	u32 offset_y, offset_c;
202 
203 	cmd = vsp1_dl_get_pre_cmd(dl);
204 	if (WARN_ONCE(!cmd, "Failed to obtain an autofld cmd"))
205 		return;
206 
207 	/* Re-index our auto_fld to match the current RPF. */
208 	auto_fld = cmd->data;
209 	auto_fld = &auto_fld[rpf->entity.index];
210 
211 	auto_fld->top_y0 = rpf->mem.addr[0];
212 	auto_fld->top_c0 = rpf->mem.addr[1];
213 	auto_fld->top_c1 = rpf->mem.addr[2];
214 
215 	offset_y = format->plane_fmt[0].bytesperline;
216 	offset_c = format->plane_fmt[1].bytesperline;
217 
218 	auto_fld->bottom_y0 = rpf->mem.addr[0] + offset_y;
219 	auto_fld->bottom_c0 = rpf->mem.addr[1] + offset_c;
220 	auto_fld->bottom_c1 = rpf->mem.addr[2] + offset_c;
221 
222 	cmd->flags |= VI6_DL_EXT_AUTOFLD_INT | BIT(16 + rpf->entity.index);
223 }
224 
rpf_configure_frame(struct vsp1_entity * entity,struct vsp1_pipeline * pipe,struct vsp1_dl_list * dl,struct vsp1_dl_body * dlb)225 static void rpf_configure_frame(struct vsp1_entity *entity,
226 				struct vsp1_pipeline *pipe,
227 				struct vsp1_dl_list *dl,
228 				struct vsp1_dl_body *dlb)
229 {
230 	struct vsp1_rwpf *rpf = to_rwpf(&entity->subdev);
231 
232 	vsp1_rpf_write(rpf, dlb, VI6_RPF_VRTCOL_SET,
233 		       rpf->alpha << VI6_RPF_VRTCOL_SET_LAYA_SHIFT);
234 	vsp1_rpf_write(rpf, dlb, VI6_RPF_MULT_ALPHA, rpf->mult_alpha |
235 		       (rpf->alpha << VI6_RPF_MULT_ALPHA_RATIO_SHIFT));
236 
237 	vsp1_pipeline_propagate_alpha(pipe, dlb, rpf->alpha);
238 }
239 
rpf_configure_partition(struct vsp1_entity * entity,struct vsp1_pipeline * pipe,struct vsp1_dl_list * dl,struct vsp1_dl_body * dlb)240 static void rpf_configure_partition(struct vsp1_entity *entity,
241 				    struct vsp1_pipeline *pipe,
242 				    struct vsp1_dl_list *dl,
243 				    struct vsp1_dl_body *dlb)
244 {
245 	struct vsp1_rwpf *rpf = to_rwpf(&entity->subdev);
246 	struct vsp1_rwpf_memory mem = rpf->mem;
247 	struct vsp1_device *vsp1 = rpf->entity.vsp1;
248 	const struct vsp1_format_info *fmtinfo = rpf->fmtinfo;
249 	const struct v4l2_pix_format_mplane *format = &rpf->format;
250 	struct v4l2_rect crop;
251 
252 	/*
253 	 * Source size and crop offsets.
254 	 *
255 	 * The crop offsets correspond to the location of the crop
256 	 * rectangle top left corner in the plane buffer. Only two
257 	 * offsets are needed, as planes 2 and 3 always have identical
258 	 * strides.
259 	 */
260 	crop = *vsp1_rwpf_get_crop(rpf, rpf->entity.config);
261 
262 	/*
263 	 * Partition Algorithm Control
264 	 *
265 	 * The partition algorithm can split this frame into multiple
266 	 * slices. We must scale our partition window based on the pipe
267 	 * configuration to match the destination partition window.
268 	 * To achieve this, we adjust our crop to provide a 'sub-crop'
269 	 * matching the expected partition window. Only 'left' and
270 	 * 'width' need to be adjusted.
271 	 */
272 	if (pipe->partitions > 1) {
273 		crop.width = pipe->partition->rpf.width;
274 		crop.left += pipe->partition->rpf.left;
275 	}
276 
277 	if (pipe->interlaced) {
278 		crop.height = round_down(crop.height / 2, fmtinfo->vsub);
279 		crop.top = round_down(crop.top / 2, fmtinfo->vsub);
280 	}
281 
282 	vsp1_rpf_write(rpf, dlb, VI6_RPF_SRC_BSIZE,
283 		       (crop.width << VI6_RPF_SRC_BSIZE_BHSIZE_SHIFT) |
284 		       (crop.height << VI6_RPF_SRC_BSIZE_BVSIZE_SHIFT));
285 	vsp1_rpf_write(rpf, dlb, VI6_RPF_SRC_ESIZE,
286 		       (crop.width << VI6_RPF_SRC_ESIZE_EHSIZE_SHIFT) |
287 		       (crop.height << VI6_RPF_SRC_ESIZE_EVSIZE_SHIFT));
288 
289 	mem.addr[0] += crop.top * format->plane_fmt[0].bytesperline
290 		     + crop.left * fmtinfo->bpp[0] / 8;
291 
292 	if (format->num_planes > 1) {
293 		unsigned int offset;
294 
295 		offset = crop.top * format->plane_fmt[1].bytesperline
296 		       + crop.left / fmtinfo->hsub
297 		       * fmtinfo->bpp[1] / 8;
298 		mem.addr[1] += offset;
299 		mem.addr[2] += offset;
300 	}
301 
302 	/*
303 	 * On Gen3 hardware the SPUVS bit has no effect on 3-planar
304 	 * formats. Swap the U and V planes manually in that case.
305 	 */
306 	if (vsp1->info->gen == 3 && format->num_planes == 3 &&
307 	    fmtinfo->swap_uv)
308 		swap(mem.addr[1], mem.addr[2]);
309 
310 	/*
311 	 * Interlaced pipelines will use the extended pre-cmd to process
312 	 * SRCM_ADDR_{Y,C0,C1}
313 	 */
314 	if (pipe->interlaced) {
315 		vsp1_rpf_configure_autofld(rpf, dl);
316 	} else {
317 		vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_Y, mem.addr[0]);
318 		vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_C0, mem.addr[1]);
319 		vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_C1, mem.addr[2]);
320 	}
321 }
322 
rpf_partition(struct vsp1_entity * entity,struct vsp1_pipeline * pipe,struct vsp1_partition * partition,unsigned int partition_idx,struct vsp1_partition_window * window)323 static void rpf_partition(struct vsp1_entity *entity,
324 			  struct vsp1_pipeline *pipe,
325 			  struct vsp1_partition *partition,
326 			  unsigned int partition_idx,
327 			  struct vsp1_partition_window *window)
328 {
329 	partition->rpf = *window;
330 }
331 
332 static const struct vsp1_entity_operations rpf_entity_ops = {
333 	.configure_stream = rpf_configure_stream,
334 	.configure_frame = rpf_configure_frame,
335 	.configure_partition = rpf_configure_partition,
336 	.partition = rpf_partition,
337 };
338 
339 /* -----------------------------------------------------------------------------
340  * Initialization and Cleanup
341  */
342 
vsp1_rpf_create(struct vsp1_device * vsp1,unsigned int index)343 struct vsp1_rwpf *vsp1_rpf_create(struct vsp1_device *vsp1, unsigned int index)
344 {
345 	struct vsp1_rwpf *rpf;
346 	char name[6];
347 	int ret;
348 
349 	rpf = devm_kzalloc(vsp1->dev, sizeof(*rpf), GFP_KERNEL);
350 	if (rpf == NULL)
351 		return ERR_PTR(-ENOMEM);
352 
353 	rpf->max_width = RPF_MAX_WIDTH;
354 	rpf->max_height = RPF_MAX_HEIGHT;
355 
356 	rpf->entity.ops = &rpf_entity_ops;
357 	rpf->entity.type = VSP1_ENTITY_RPF;
358 	rpf->entity.index = index;
359 
360 	sprintf(name, "rpf.%u", index);
361 	ret = vsp1_entity_init(vsp1, &rpf->entity, name, 2, &rpf_ops,
362 			       MEDIA_ENT_F_PROC_VIDEO_PIXEL_FORMATTER);
363 	if (ret < 0)
364 		return ERR_PTR(ret);
365 
366 	/* Initialize the control handler. */
367 	ret = vsp1_rwpf_init_ctrls(rpf, 0);
368 	if (ret < 0) {
369 		dev_err(vsp1->dev, "rpf%u: failed to initialize controls\n",
370 			index);
371 		goto error;
372 	}
373 
374 	v4l2_ctrl_handler_setup(&rpf->ctrls);
375 
376 	return rpf;
377 
378 error:
379 	vsp1_entity_destroy(&rpf->entity);
380 	return ERR_PTR(ret);
381 }
382