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 
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 
58 static void rpf_configure_stream(struct vsp1_entity *entity,
59 				 struct vsp1_pipeline *pipe,
60 				 struct vsp1_dl_list *dl,
61 				 struct vsp1_dl_body *dlb)
62 {
63 	struct vsp1_rwpf *rpf = to_rwpf(&entity->subdev);
64 	const struct vsp1_format_info *fmtinfo = rpf->fmtinfo;
65 	const struct v4l2_pix_format_mplane *format = &rpf->format;
66 	const struct v4l2_mbus_framefmt *source_format;
67 	const struct v4l2_mbus_framefmt *sink_format;
68 	unsigned int left = 0;
69 	unsigned int top = 0;
70 	u32 pstride;
71 	u32 infmt;
72 
73 	/* Stride */
74 	pstride = format->plane_fmt[0].bytesperline
75 		<< VI6_RPF_SRCM_PSTRIDE_Y_SHIFT;
76 	if (format->num_planes > 1)
77 		pstride |= format->plane_fmt[1].bytesperline
78 			<< VI6_RPF_SRCM_PSTRIDE_C_SHIFT;
79 
80 	/*
81 	 * pstride has both STRIDE_Y and STRIDE_C, but multiplying the whole
82 	 * of pstride by 2 is conveniently OK here as we are multiplying both
83 	 * values.
84 	 */
85 	if (pipe->interlaced)
86 		pstride *= 2;
87 
88 	vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_PSTRIDE, pstride);
89 
90 	/* Format */
91 	sink_format = vsp1_entity_get_pad_format(&rpf->entity,
92 						 rpf->entity.config,
93 						 RWPF_PAD_SINK);
94 	source_format = vsp1_entity_get_pad_format(&rpf->entity,
95 						   rpf->entity.config,
96 						   RWPF_PAD_SOURCE);
97 
98 	infmt = VI6_RPF_INFMT_CIPM
99 	      | (fmtinfo->hwfmt << VI6_RPF_INFMT_RDFMT_SHIFT);
100 
101 	if (fmtinfo->swap_yc)
102 		infmt |= VI6_RPF_INFMT_SPYCS;
103 	if (fmtinfo->swap_uv)
104 		infmt |= VI6_RPF_INFMT_SPUVS;
105 
106 	if (sink_format->code != source_format->code)
107 		infmt |= VI6_RPF_INFMT_CSC;
108 
109 	vsp1_rpf_write(rpf, dlb, VI6_RPF_INFMT, infmt);
110 	vsp1_rpf_write(rpf, dlb, VI6_RPF_DSWAP, fmtinfo->swap);
111 
112 	if (entity->vsp1->info->gen == 4) {
113 		u32 ext_infmt0;
114 		u32 ext_infmt1;
115 		u32 ext_infmt2;
116 
117 		switch (fmtinfo->fourcc) {
118 		case V4L2_PIX_FMT_RGBX1010102:
119 			ext_infmt0 = VI6_RPF_EXT_INFMT0_BYPP_M1_RGB10;
120 			ext_infmt1 = VI6_RPF_EXT_INFMT1_PACK_CPOS(0, 10, 20, 0);
121 			ext_infmt2 = VI6_RPF_EXT_INFMT2_PACK_CLEN(10, 10, 10, 0);
122 			break;
123 
124 		case V4L2_PIX_FMT_RGBA1010102:
125 			ext_infmt0 = VI6_RPF_EXT_INFMT0_BYPP_M1_RGB10;
126 			ext_infmt1 = VI6_RPF_EXT_INFMT1_PACK_CPOS(0, 10, 20, 30);
127 			ext_infmt2 = VI6_RPF_EXT_INFMT2_PACK_CLEN(10, 10, 10, 2);
128 			break;
129 
130 		case V4L2_PIX_FMT_ARGB2101010:
131 			ext_infmt0 = VI6_RPF_EXT_INFMT0_BYPP_M1_RGB10;
132 			ext_infmt1 = VI6_RPF_EXT_INFMT1_PACK_CPOS(2, 12, 22, 0);
133 			ext_infmt2 = VI6_RPF_EXT_INFMT2_PACK_CLEN(10, 10, 10, 2);
134 			break;
135 
136 		case V4L2_PIX_FMT_Y210:
137 			ext_infmt0 = VI6_RPF_EXT_INFMT0_F2B |
138 				     VI6_RPF_EXT_INFMT0_IPBD_Y_10 |
139 				     VI6_RPF_EXT_INFMT0_IPBD_C_10;
140 			ext_infmt1 = 0x0;
141 			ext_infmt2 = 0x0;
142 			break;
143 
144 		case V4L2_PIX_FMT_Y212:
145 			ext_infmt0 = VI6_RPF_EXT_INFMT0_F2B |
146 				     VI6_RPF_EXT_INFMT0_IPBD_Y_12 |
147 				     VI6_RPF_EXT_INFMT0_IPBD_C_12;
148 			ext_infmt1 = 0x0;
149 			ext_infmt2 = 0x0;
150 			break;
151 
152 		default:
153 			ext_infmt0 = 0;
154 			ext_infmt1 = 0;
155 			ext_infmt2 = 0;
156 			break;
157 		}
158 
159 		vsp1_rpf_write(rpf, dlb, VI6_RPF_EXT_INFMT0, ext_infmt0);
160 		vsp1_rpf_write(rpf, dlb, VI6_RPF_EXT_INFMT1, ext_infmt1);
161 		vsp1_rpf_write(rpf, dlb, VI6_RPF_EXT_INFMT2, ext_infmt2);
162 	}
163 
164 	/* Output location. */
165 	if (pipe->brx) {
166 		const struct v4l2_rect *compose;
167 
168 		compose = vsp1_entity_get_pad_selection(pipe->brx,
169 							pipe->brx->config,
170 							rpf->brx_input,
171 							V4L2_SEL_TGT_COMPOSE);
172 		left = compose->left;
173 		top = compose->top;
174 	}
175 
176 	if (pipe->interlaced)
177 		top /= 2;
178 
179 	vsp1_rpf_write(rpf, dlb, VI6_RPF_LOC,
180 		       (left << VI6_RPF_LOC_HCOORD_SHIFT) |
181 		       (top << VI6_RPF_LOC_VCOORD_SHIFT));
182 
183 	/*
184 	 * On Gen2 use the alpha channel (extended to 8 bits) when available or
185 	 * a fixed alpha value set through the V4L2_CID_ALPHA_COMPONENT control
186 	 * otherwise.
187 	 *
188 	 * The Gen3+ RPF has extended alpha capability and can both multiply the
189 	 * alpha channel by a fixed global alpha value, and multiply the pixel
190 	 * components to convert the input to premultiplied alpha.
191 	 *
192 	 * As alpha premultiplication is available in the BRx for both Gen2 and
193 	 * Gen3+ we handle it there and use the Gen3 alpha multiplier for global
194 	 * alpha multiplication only. This however prevents conversion to
195 	 * premultiplied alpha if no BRx is present in the pipeline. If that use
196 	 * case turns out to be useful we will revisit the implementation (for
197 	 * Gen3 only).
198 	 *
199 	 * We enable alpha multiplication on Gen3+ using the fixed alpha value
200 	 * set through the V4L2_CID_ALPHA_COMPONENT control when the input
201 	 * contains an alpha channel. On Gen2 the global alpha is ignored in
202 	 * that case.
203 	 *
204 	 * In all cases, disable color keying.
205 	 */
206 	vsp1_rpf_write(rpf, dlb, VI6_RPF_ALPH_SEL, VI6_RPF_ALPH_SEL_AEXT_EXT |
207 		       (fmtinfo->alpha ? VI6_RPF_ALPH_SEL_ASEL_PACKED
208 				       : VI6_RPF_ALPH_SEL_ASEL_FIXED));
209 
210 	if (entity->vsp1->info->gen >= 3) {
211 		u32 mult;
212 
213 		if (fmtinfo->alpha) {
214 			/*
215 			 * When the input contains an alpha channel enable the
216 			 * alpha multiplier. If the input is premultiplied we
217 			 * need to multiply both the alpha channel and the pixel
218 			 * components by the global alpha value to keep them
219 			 * premultiplied. Otherwise multiply the alpha channel
220 			 * only.
221 			 */
222 			bool premultiplied = format->flags
223 					   & V4L2_PIX_FMT_FLAG_PREMUL_ALPHA;
224 
225 			mult = VI6_RPF_MULT_ALPHA_A_MMD_RATIO
226 			     | (premultiplied ?
227 				VI6_RPF_MULT_ALPHA_P_MMD_RATIO :
228 				VI6_RPF_MULT_ALPHA_P_MMD_NONE);
229 		} else {
230 			/*
231 			 * When the input doesn't contain an alpha channel the
232 			 * global alpha value is applied in the unpacking unit,
233 			 * the alpha multiplier isn't needed and must be
234 			 * disabled.
235 			 */
236 			mult = VI6_RPF_MULT_ALPHA_A_MMD_NONE
237 			     | VI6_RPF_MULT_ALPHA_P_MMD_NONE;
238 		}
239 
240 		rpf->mult_alpha = mult;
241 	}
242 
243 	vsp1_rpf_write(rpf, dlb, VI6_RPF_MSK_CTRL, 0);
244 	vsp1_rpf_write(rpf, dlb, VI6_RPF_CKEY_CTRL, 0);
245 
246 }
247 
248 static void vsp1_rpf_configure_autofld(struct vsp1_rwpf *rpf,
249 				       struct vsp1_dl_list *dl)
250 {
251 	const struct v4l2_pix_format_mplane *format = &rpf->format;
252 	struct vsp1_dl_ext_cmd *cmd;
253 	struct vsp1_extcmd_auto_fld_body *auto_fld;
254 	u32 offset_y, offset_c;
255 
256 	cmd = vsp1_dl_get_pre_cmd(dl);
257 	if (WARN_ONCE(!cmd, "Failed to obtain an autofld cmd"))
258 		return;
259 
260 	/* Re-index our auto_fld to match the current RPF. */
261 	auto_fld = cmd->data;
262 	auto_fld = &auto_fld[rpf->entity.index];
263 
264 	auto_fld->top_y0 = rpf->mem.addr[0];
265 	auto_fld->top_c0 = rpf->mem.addr[1];
266 	auto_fld->top_c1 = rpf->mem.addr[2];
267 
268 	offset_y = format->plane_fmt[0].bytesperline;
269 	offset_c = format->plane_fmt[1].bytesperline;
270 
271 	auto_fld->bottom_y0 = rpf->mem.addr[0] + offset_y;
272 	auto_fld->bottom_c0 = rpf->mem.addr[1] + offset_c;
273 	auto_fld->bottom_c1 = rpf->mem.addr[2] + offset_c;
274 
275 	cmd->flags |= VI6_DL_EXT_AUTOFLD_INT | BIT(16 + rpf->entity.index);
276 }
277 
278 static void rpf_configure_frame(struct vsp1_entity *entity,
279 				struct vsp1_pipeline *pipe,
280 				struct vsp1_dl_list *dl,
281 				struct vsp1_dl_body *dlb)
282 {
283 	struct vsp1_rwpf *rpf = to_rwpf(&entity->subdev);
284 
285 	vsp1_rpf_write(rpf, dlb, VI6_RPF_VRTCOL_SET,
286 		       rpf->alpha << VI6_RPF_VRTCOL_SET_LAYA_SHIFT);
287 	vsp1_rpf_write(rpf, dlb, VI6_RPF_MULT_ALPHA, rpf->mult_alpha |
288 		       (rpf->alpha << VI6_RPF_MULT_ALPHA_RATIO_SHIFT));
289 
290 	vsp1_pipeline_propagate_alpha(pipe, dlb, rpf->alpha);
291 }
292 
293 static void rpf_configure_partition(struct vsp1_entity *entity,
294 				    struct vsp1_pipeline *pipe,
295 				    struct vsp1_dl_list *dl,
296 				    struct vsp1_dl_body *dlb)
297 {
298 	struct vsp1_rwpf *rpf = to_rwpf(&entity->subdev);
299 	struct vsp1_rwpf_memory mem = rpf->mem;
300 	struct vsp1_device *vsp1 = rpf->entity.vsp1;
301 	const struct vsp1_format_info *fmtinfo = rpf->fmtinfo;
302 	const struct v4l2_pix_format_mplane *format = &rpf->format;
303 	struct v4l2_rect crop;
304 
305 	/*
306 	 * Source size and crop offsets.
307 	 *
308 	 * The crop offsets correspond to the location of the crop
309 	 * rectangle top left corner in the plane buffer. Only two
310 	 * offsets are needed, as planes 2 and 3 always have identical
311 	 * strides.
312 	 */
313 	crop = *vsp1_rwpf_get_crop(rpf, rpf->entity.config);
314 
315 	/*
316 	 * Partition Algorithm Control
317 	 *
318 	 * The partition algorithm can split this frame into multiple
319 	 * slices. We must scale our partition window based on the pipe
320 	 * configuration to match the destination partition window.
321 	 * To achieve this, we adjust our crop to provide a 'sub-crop'
322 	 * matching the expected partition window. Only 'left' and
323 	 * 'width' need to be adjusted.
324 	 */
325 	if (pipe->partitions > 1) {
326 		crop.width = pipe->partition->rpf.width;
327 		crop.left += pipe->partition->rpf.left;
328 	}
329 
330 	if (pipe->interlaced) {
331 		crop.height = round_down(crop.height / 2, fmtinfo->vsub);
332 		crop.top = round_down(crop.top / 2, fmtinfo->vsub);
333 	}
334 
335 	vsp1_rpf_write(rpf, dlb, VI6_RPF_SRC_BSIZE,
336 		       (crop.width << VI6_RPF_SRC_BSIZE_BHSIZE_SHIFT) |
337 		       (crop.height << VI6_RPF_SRC_BSIZE_BVSIZE_SHIFT));
338 	vsp1_rpf_write(rpf, dlb, VI6_RPF_SRC_ESIZE,
339 		       (crop.width << VI6_RPF_SRC_ESIZE_EHSIZE_SHIFT) |
340 		       (crop.height << VI6_RPF_SRC_ESIZE_EVSIZE_SHIFT));
341 
342 	mem.addr[0] += crop.top * format->plane_fmt[0].bytesperline
343 		     + crop.left * fmtinfo->bpp[0] / 8;
344 
345 	if (format->num_planes > 1) {
346 		unsigned int bpl = format->plane_fmt[1].bytesperline;
347 		unsigned int offset;
348 
349 		offset = crop.top / fmtinfo->vsub * bpl
350 		       + crop.left / fmtinfo->hsub * fmtinfo->bpp[1] / 8;
351 		mem.addr[1] += offset;
352 		mem.addr[2] += offset;
353 	}
354 
355 	/*
356 	 * On Gen3+ hardware the SPUVS bit has no effect on 3-planar
357 	 * formats. Swap the U and V planes manually in that case.
358 	 */
359 	if (vsp1->info->gen >= 3 && format->num_planes == 3 &&
360 	    fmtinfo->swap_uv)
361 		swap(mem.addr[1], mem.addr[2]);
362 
363 	/*
364 	 * Interlaced pipelines will use the extended pre-cmd to process
365 	 * SRCM_ADDR_{Y,C0,C1}.
366 	 */
367 	if (pipe->interlaced) {
368 		vsp1_rpf_configure_autofld(rpf, dl);
369 	} else {
370 		vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_Y, mem.addr[0]);
371 		vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_C0, mem.addr[1]);
372 		vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_C1, mem.addr[2]);
373 	}
374 }
375 
376 static void rpf_partition(struct vsp1_entity *entity,
377 			  struct vsp1_pipeline *pipe,
378 			  struct vsp1_partition *partition,
379 			  unsigned int partition_idx,
380 			  struct vsp1_partition_window *window)
381 {
382 	partition->rpf = *window;
383 }
384 
385 static const struct vsp1_entity_operations rpf_entity_ops = {
386 	.configure_stream = rpf_configure_stream,
387 	.configure_frame = rpf_configure_frame,
388 	.configure_partition = rpf_configure_partition,
389 	.partition = rpf_partition,
390 };
391 
392 /* -----------------------------------------------------------------------------
393  * Initialization and Cleanup
394  */
395 
396 struct vsp1_rwpf *vsp1_rpf_create(struct vsp1_device *vsp1, unsigned int index)
397 {
398 	struct vsp1_rwpf *rpf;
399 	char name[6];
400 	int ret;
401 
402 	rpf = devm_kzalloc(vsp1->dev, sizeof(*rpf), GFP_KERNEL);
403 	if (rpf == NULL)
404 		return ERR_PTR(-ENOMEM);
405 
406 	rpf->max_width = RPF_MAX_WIDTH;
407 	rpf->max_height = RPF_MAX_HEIGHT;
408 
409 	rpf->entity.ops = &rpf_entity_ops;
410 	rpf->entity.type = VSP1_ENTITY_RPF;
411 	rpf->entity.index = index;
412 
413 	sprintf(name, "rpf.%u", index);
414 	ret = vsp1_entity_init(vsp1, &rpf->entity, name, 2, &rpf_ops,
415 			       MEDIA_ENT_F_PROC_VIDEO_PIXEL_FORMATTER);
416 	if (ret < 0)
417 		return ERR_PTR(ret);
418 
419 	/* Initialize the control handler. */
420 	ret = vsp1_rwpf_init_ctrls(rpf, 0);
421 	if (ret < 0) {
422 		dev_err(vsp1->dev, "rpf%u: failed to initialize controls\n",
423 			index);
424 		goto error;
425 	}
426 
427 	v4l2_ctrl_handler_setup(&rpf->ctrls);
428 
429 	return rpf;
430 
431 error:
432 	vsp1_entity_destroy(&rpf->entity);
433 	return ERR_PTR(ret);
434 }
435