1 // SPDX-License-Identifier: GPL-2.0+
2 /*
3  * i.MX Pixel Pipeline (PXP) mem-to-mem scaler/CSC/rotator driver
4  *
5  * Copyright (c) 2018 Pengutronix, Philipp Zabel
6  *
7  * based on vim2m
8  *
9  * Copyright (c) 2009-2010 Samsung Electronics Co., Ltd.
10  * Pawel Osciak, <pawel@osciak.com>
11  * Marek Szyprowski, <m.szyprowski@samsung.com>
12  */
13 #include <linux/clk.h>
14 #include <linux/delay.h>
15 #include <linux/dma-mapping.h>
16 #include <linux/interrupt.h>
17 #include <linux/io.h>
18 #include <linux/iopoll.h>
19 #include <linux/module.h>
20 #include <linux/of.h>
21 #include <linux/sched.h>
22 #include <linux/slab.h>
23 
24 #include <linux/platform_device.h>
25 #include <media/v4l2-mem2mem.h>
26 #include <media/v4l2-device.h>
27 #include <media/v4l2-ioctl.h>
28 #include <media/v4l2-ctrls.h>
29 #include <media/v4l2-event.h>
30 #include <media/videobuf2-dma-contig.h>
31 
32 #include "imx-pxp.h"
33 
34 static unsigned int debug;
35 module_param(debug, uint, 0644);
36 MODULE_PARM_DESC(debug, "activates debug info");
37 
38 #define MIN_W 8
39 #define MIN_H 8
40 #define MAX_W 4096
41 #define MAX_H 4096
42 #define ALIGN_W 3 /* 8x8 pixel blocks */
43 #define ALIGN_H 3
44 
45 /* Flags that indicate a format can be used for capture/output */
46 #define MEM2MEM_CAPTURE	(1 << 0)
47 #define MEM2MEM_OUTPUT	(1 << 1)
48 
49 #define MEM2MEM_NAME		"pxp"
50 
51 /* Flags that indicate processing mode */
52 #define MEM2MEM_HFLIP	(1 << 0)
53 #define MEM2MEM_VFLIP	(1 << 1)
54 
55 #define dprintk(dev, fmt, arg...) \
56 	v4l2_dbg(1, debug, &dev->v4l2_dev, "%s: " fmt, __func__, ## arg)
57 
58 struct pxp_fmt {
59 	u32	fourcc;
60 	int	depth;
61 	/* Types the format can be used for */
62 	u32	types;
63 };
64 
65 static struct pxp_fmt formats[] = {
66 	{
67 		.fourcc	= V4L2_PIX_FMT_XBGR32,
68 		.depth	= 32,
69 		/* Both capture and output format */
70 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
71 	}, {
72 		.fourcc	= V4L2_PIX_FMT_ABGR32,
73 		.depth	= 32,
74 		/* Capture-only format */
75 		.types	= MEM2MEM_CAPTURE,
76 	}, {
77 		.fourcc	= V4L2_PIX_FMT_BGR24,
78 		.depth	= 24,
79 		.types	= MEM2MEM_CAPTURE,
80 	}, {
81 		.fourcc	= V4L2_PIX_FMT_RGB565,
82 		.depth	= 16,
83 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
84 	}, {
85 		.fourcc	= V4L2_PIX_FMT_RGB555,
86 		.depth	= 16,
87 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
88 	}, {
89 		.fourcc = V4L2_PIX_FMT_RGB444,
90 		.depth	= 16,
91 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
92 	}, {
93 		.fourcc = V4L2_PIX_FMT_VUYA32,
94 		.depth	= 32,
95 		.types	= MEM2MEM_CAPTURE,
96 	}, {
97 		.fourcc = V4L2_PIX_FMT_VUYX32,
98 		.depth	= 32,
99 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
100 	}, {
101 		.fourcc = V4L2_PIX_FMT_UYVY,
102 		.depth	= 16,
103 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
104 	}, {
105 		.fourcc = V4L2_PIX_FMT_YUYV,
106 		.depth	= 16,
107 		/* Output-only format */
108 		.types	= MEM2MEM_OUTPUT,
109 	}, {
110 		.fourcc = V4L2_PIX_FMT_VYUY,
111 		.depth	= 16,
112 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
113 	}, {
114 		.fourcc = V4L2_PIX_FMT_YVYU,
115 		.depth	= 16,
116 		.types	= MEM2MEM_OUTPUT,
117 	}, {
118 		.fourcc = V4L2_PIX_FMT_GREY,
119 		.depth	= 8,
120 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
121 	}, {
122 		.fourcc = V4L2_PIX_FMT_Y4,
123 		.depth	= 4,
124 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
125 	}, {
126 		.fourcc = V4L2_PIX_FMT_NV16,
127 		.depth	= 16,
128 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
129 	}, {
130 		.fourcc = V4L2_PIX_FMT_NV12,
131 		.depth	= 12,
132 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
133 	}, {
134 		.fourcc = V4L2_PIX_FMT_NV21,
135 		.depth	= 12,
136 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
137 	}, {
138 		.fourcc = V4L2_PIX_FMT_NV61,
139 		.depth	= 16,
140 		.types	= MEM2MEM_CAPTURE | MEM2MEM_OUTPUT,
141 	}, {
142 		.fourcc = V4L2_PIX_FMT_YUV422P,
143 		.depth	= 16,
144 		.types	= MEM2MEM_OUTPUT,
145 	}, {
146 		.fourcc = V4L2_PIX_FMT_YUV420,
147 		.depth	= 12,
148 		.types	= MEM2MEM_OUTPUT,
149 	},
150 };
151 
152 #define NUM_FORMATS ARRAY_SIZE(formats)
153 
154 /* Per-queue, driver-specific private data */
155 struct pxp_q_data {
156 	unsigned int		width;
157 	unsigned int		height;
158 	unsigned int		bytesperline;
159 	unsigned int		sizeimage;
160 	unsigned int		sequence;
161 	struct pxp_fmt		*fmt;
162 	enum v4l2_ycbcr_encoding ycbcr_enc;
163 	enum v4l2_quantization	quant;
164 };
165 
166 enum {
167 	V4L2_M2M_SRC = 0,
168 	V4L2_M2M_DST = 1,
169 };
170 
171 static struct pxp_fmt *find_format(struct v4l2_format *f)
172 {
173 	struct pxp_fmt *fmt;
174 	unsigned int k;
175 
176 	for (k = 0; k < NUM_FORMATS; k++) {
177 		fmt = &formats[k];
178 		if (fmt->fourcc == f->fmt.pix.pixelformat)
179 			break;
180 	}
181 
182 	if (k == NUM_FORMATS)
183 		return NULL;
184 
185 	return &formats[k];
186 }
187 
188 struct pxp_dev {
189 	struct v4l2_device	v4l2_dev;
190 	struct video_device	vfd;
191 
192 	struct clk		*clk;
193 	void __iomem		*mmio;
194 
195 	atomic_t		num_inst;
196 	struct mutex		dev_mutex;
197 	spinlock_t		irqlock;
198 
199 	struct v4l2_m2m_dev	*m2m_dev;
200 };
201 
202 struct pxp_ctx {
203 	struct v4l2_fh		fh;
204 	struct pxp_dev	*dev;
205 
206 	struct v4l2_ctrl_handler hdl;
207 
208 	/* Abort requested by m2m */
209 	int			aborting;
210 
211 	/* Processing mode */
212 	int			mode;
213 	u8			alpha_component;
214 	u8			rotation;
215 
216 	enum v4l2_colorspace	colorspace;
217 	enum v4l2_xfer_func	xfer_func;
218 
219 	/* Source and destination queue data */
220 	struct pxp_q_data   q_data[2];
221 };
222 
223 static inline struct pxp_ctx *file2ctx(struct file *file)
224 {
225 	return container_of(file->private_data, struct pxp_ctx, fh);
226 }
227 
228 static struct pxp_q_data *get_q_data(struct pxp_ctx *ctx,
229 					 enum v4l2_buf_type type)
230 {
231 	if (type == V4L2_BUF_TYPE_VIDEO_OUTPUT)
232 		return &ctx->q_data[V4L2_M2M_SRC];
233 	else
234 		return &ctx->q_data[V4L2_M2M_DST];
235 }
236 
237 static u32 pxp_v4l2_pix_fmt_to_ps_format(u32 v4l2_pix_fmt)
238 {
239 	switch (v4l2_pix_fmt) {
240 	case V4L2_PIX_FMT_XBGR32:  return BV_PXP_PS_CTRL_FORMAT__RGB888;
241 	case V4L2_PIX_FMT_RGB555:  return BV_PXP_PS_CTRL_FORMAT__RGB555;
242 	case V4L2_PIX_FMT_RGB444:  return BV_PXP_PS_CTRL_FORMAT__RGB444;
243 	case V4L2_PIX_FMT_RGB565:  return BV_PXP_PS_CTRL_FORMAT__RGB565;
244 	case V4L2_PIX_FMT_VUYX32:  return BV_PXP_PS_CTRL_FORMAT__YUV1P444;
245 	case V4L2_PIX_FMT_UYVY:    return BV_PXP_PS_CTRL_FORMAT__UYVY1P422;
246 	case V4L2_PIX_FMT_YUYV:    return BM_PXP_PS_CTRL_WB_SWAP |
247 					  BV_PXP_PS_CTRL_FORMAT__UYVY1P422;
248 	case V4L2_PIX_FMT_VYUY:    return BV_PXP_PS_CTRL_FORMAT__VYUY1P422;
249 	case V4L2_PIX_FMT_YVYU:    return BM_PXP_PS_CTRL_WB_SWAP |
250 					  BV_PXP_PS_CTRL_FORMAT__VYUY1P422;
251 	case V4L2_PIX_FMT_GREY:    return BV_PXP_PS_CTRL_FORMAT__Y8;
252 	default:
253 	case V4L2_PIX_FMT_Y4:      return BV_PXP_PS_CTRL_FORMAT__Y4;
254 	case V4L2_PIX_FMT_NV16:    return BV_PXP_PS_CTRL_FORMAT__YUV2P422;
255 	case V4L2_PIX_FMT_NV12:    return BV_PXP_PS_CTRL_FORMAT__YUV2P420;
256 	case V4L2_PIX_FMT_NV21:    return BV_PXP_PS_CTRL_FORMAT__YVU2P420;
257 	case V4L2_PIX_FMT_NV61:    return BV_PXP_PS_CTRL_FORMAT__YVU2P422;
258 	case V4L2_PIX_FMT_YUV422P: return BV_PXP_PS_CTRL_FORMAT__YUV422;
259 	case V4L2_PIX_FMT_YUV420:  return BV_PXP_PS_CTRL_FORMAT__YUV420;
260 	}
261 }
262 
263 static u32 pxp_v4l2_pix_fmt_to_out_format(u32 v4l2_pix_fmt)
264 {
265 	switch (v4l2_pix_fmt) {
266 	case V4L2_PIX_FMT_XBGR32:   return BV_PXP_OUT_CTRL_FORMAT__RGB888;
267 	case V4L2_PIX_FMT_ABGR32:   return BV_PXP_OUT_CTRL_FORMAT__ARGB8888;
268 	case V4L2_PIX_FMT_BGR24:    return BV_PXP_OUT_CTRL_FORMAT__RGB888P;
269 	/* Missing V4L2 pixel formats for ARGB1555 and ARGB4444 */
270 	case V4L2_PIX_FMT_RGB555:   return BV_PXP_OUT_CTRL_FORMAT__RGB555;
271 	case V4L2_PIX_FMT_RGB444:   return BV_PXP_OUT_CTRL_FORMAT__RGB444;
272 	case V4L2_PIX_FMT_RGB565:   return BV_PXP_OUT_CTRL_FORMAT__RGB565;
273 	case V4L2_PIX_FMT_VUYA32:
274 	case V4L2_PIX_FMT_VUYX32:   return BV_PXP_OUT_CTRL_FORMAT__YUV1P444;
275 	case V4L2_PIX_FMT_UYVY:     return BV_PXP_OUT_CTRL_FORMAT__UYVY1P422;
276 	case V4L2_PIX_FMT_VYUY:     return BV_PXP_OUT_CTRL_FORMAT__VYUY1P422;
277 	case V4L2_PIX_FMT_GREY:     return BV_PXP_OUT_CTRL_FORMAT__Y8;
278 	default:
279 	case V4L2_PIX_FMT_Y4:       return BV_PXP_OUT_CTRL_FORMAT__Y4;
280 	case V4L2_PIX_FMT_NV16:     return BV_PXP_OUT_CTRL_FORMAT__YUV2P422;
281 	case V4L2_PIX_FMT_NV12:     return BV_PXP_OUT_CTRL_FORMAT__YUV2P420;
282 	case V4L2_PIX_FMT_NV61:     return BV_PXP_OUT_CTRL_FORMAT__YVU2P422;
283 	case V4L2_PIX_FMT_NV21:     return BV_PXP_OUT_CTRL_FORMAT__YVU2P420;
284 	}
285 }
286 
287 static bool pxp_v4l2_pix_fmt_is_yuv(u32 v4l2_pix_fmt)
288 {
289 	switch (v4l2_pix_fmt) {
290 	case V4L2_PIX_FMT_VUYA32:
291 	case V4L2_PIX_FMT_VUYX32:
292 	case V4L2_PIX_FMT_UYVY:
293 	case V4L2_PIX_FMT_YUYV:
294 	case V4L2_PIX_FMT_VYUY:
295 	case V4L2_PIX_FMT_YVYU:
296 	case V4L2_PIX_FMT_NV16:
297 	case V4L2_PIX_FMT_NV12:
298 	case V4L2_PIX_FMT_NV61:
299 	case V4L2_PIX_FMT_NV21:
300 	case V4L2_PIX_FMT_YUV420:
301 	case V4L2_PIX_FMT_YUV422P:
302 	case V4L2_PIX_FMT_GREY:
303 	case V4L2_PIX_FMT_Y4:
304 		return true;
305 	default:
306 		return false;
307 	}
308 }
309 
310 static void pxp_setup_csc(struct pxp_ctx *ctx)
311 {
312 	struct pxp_dev *dev = ctx->dev;
313 	enum v4l2_ycbcr_encoding ycbcr_enc;
314 	enum v4l2_quantization quantization;
315 
316 	if (pxp_v4l2_pix_fmt_is_yuv(ctx->q_data[V4L2_M2M_SRC].fmt->fourcc) &&
317 	    !pxp_v4l2_pix_fmt_is_yuv(ctx->q_data[V4L2_M2M_DST].fmt->fourcc)) {
318 		/*
319 		 * CSC1 YUV/YCbCr to RGB conversion is implemented as follows:
320 		 *
321 		 * |R|   |C0 0  C1|   |Y  + Yoffset |
322 		 * |G| = |C0 C3 C2| * |Cb + UVoffset|
323 		 * |B|   |C0 C4 0 |   |Cr + UVoffset|
324 		 *
325 		 * Results are clamped to 0..255.
326 		 *
327 		 * BT.601 limited range:
328 		 *
329 		 * |R|   |1.1644  0.0000  1.5960|   |Y  - 16 |
330 		 * |G| = |1.1644 -0.3917 -0.8129| * |Cb - 128|
331 		 * |B|   |1.1644  2.0172  0.0000|   |Cr - 128|
332 		 */
333 		static const u32 csc1_coef_bt601_lim[3] = {
334 			BM_PXP_CSC1_COEF0_YCBCR_MODE |
335 			BF_PXP_CSC1_COEF0_C0(0x12a) |	/*  1.1641 (-0.03 %) */
336 			BF_PXP_CSC1_COEF0_UV_OFFSET(-128) |
337 			BF_PXP_CSC1_COEF0_Y_OFFSET(-16),
338 			BF_PXP_CSC1_COEF1_C1(0x198) |	/*  1.5938 (-0.23 %) */
339 			BF_PXP_CSC1_COEF1_C4(0x204),	/*  2.0156 (-0.16 %) */
340 			BF_PXP_CSC1_COEF2_C2(0x730) |	/* -0.8125 (+0.04 %) */
341 			BF_PXP_CSC1_COEF2_C3(0x79c),	/* -0.3906 (+0.11 %) */
342 		};
343 		/*
344 		 * BT.601 full range:
345 		 *
346 		 * |R|   |1.0000  0.0000  1.4020|   |Y  + 0  |
347 		 * |G| = |1.0000 -0.3441 -0.7141| * |Cb - 128|
348 		 * |B|   |1.0000  1.7720  0.0000|   |Cr - 128|
349 		 */
350 		static const u32 csc1_coef_bt601_full[3] = {
351 			BM_PXP_CSC1_COEF0_YCBCR_MODE |
352 			BF_PXP_CSC1_COEF0_C0(0x100) |	/*  1.0000 (+0.00 %) */
353 			BF_PXP_CSC1_COEF0_UV_OFFSET(-128) |
354 			BF_PXP_CSC1_COEF0_Y_OFFSET(0),
355 			BF_PXP_CSC1_COEF1_C1(0x166) |	/*  1.3984 (-0.36 %) */
356 			BF_PXP_CSC1_COEF1_C4(0x1c5),	/*  1.7695 (-0.25 %) */
357 			BF_PXP_CSC1_COEF2_C2(0x74a) |	/* -0.7109 (+0.32 %) */
358 			BF_PXP_CSC1_COEF2_C3(0x7a8),	/* -0.3438 (+0.04 %) */
359 		};
360 		/*
361 		 * Rec.709 limited range:
362 		 *
363 		 * |R|   |1.1644  0.0000  1.7927|   |Y  - 16 |
364 		 * |G| = |1.1644 -0.2132 -0.5329| * |Cb - 128|
365 		 * |B|   |1.1644  2.1124  0.0000|   |Cr - 128|
366 		 */
367 		static const u32 csc1_coef_rec709_lim[3] = {
368 			BM_PXP_CSC1_COEF0_YCBCR_MODE |
369 			BF_PXP_CSC1_COEF0_C0(0x12a) |	/*  1.1641 (-0.03 %) */
370 			BF_PXP_CSC1_COEF0_UV_OFFSET(-128) |
371 			BF_PXP_CSC1_COEF0_Y_OFFSET(-16),
372 			BF_PXP_CSC1_COEF1_C1(0x1ca) |	/*  1.7891 (-0.37 %) */
373 			BF_PXP_CSC1_COEF1_C4(0x21c),	/*  2.1094 (-0.30 %) */
374 			BF_PXP_CSC1_COEF2_C2(0x778) |	/* -0.5312 (+0.16 %) */
375 			BF_PXP_CSC1_COEF2_C3(0x7ca),	/* -0.2109 (+0.23 %) */
376 		};
377 		/*
378 		 * Rec.709 full range:
379 		 *
380 		 * |R|   |1.0000  0.0000  1.5748|   |Y  + 0  |
381 		 * |G| = |1.0000 -0.1873 -0.4681| * |Cb - 128|
382 		 * |B|   |1.0000  1.8556  0.0000|   |Cr - 128|
383 		 */
384 		static const u32 csc1_coef_rec709_full[3] = {
385 			BM_PXP_CSC1_COEF0_YCBCR_MODE |
386 			BF_PXP_CSC1_COEF0_C0(0x100) |	/*  1.0000 (+0.00 %) */
387 			BF_PXP_CSC1_COEF0_UV_OFFSET(-128) |
388 			BF_PXP_CSC1_COEF0_Y_OFFSET(0),
389 			BF_PXP_CSC1_COEF1_C1(0x193) |	/*  1.5742 (-0.06 %) */
390 			BF_PXP_CSC1_COEF1_C4(0x1db),	/*  1.8555 (-0.01 %) */
391 			BF_PXP_CSC1_COEF2_C2(0x789) |	/* -0.4648 (+0.33 %) */
392 			BF_PXP_CSC1_COEF2_C3(0x7d1),	/* -0.1836 (+0.37 %) */
393 		};
394 		/*
395 		 * BT.2020 limited range:
396 		 *
397 		 * |R|   |1.1644  0.0000  1.6787|   |Y  - 16 |
398 		 * |G| = |1.1644 -0.1874 -0.6505| * |Cb - 128|
399 		 * |B|   |1.1644  2.1418  0.0000|   |Cr - 128|
400 		 */
401 		static const u32 csc1_coef_bt2020_lim[3] = {
402 			BM_PXP_CSC1_COEF0_YCBCR_MODE |
403 			BF_PXP_CSC1_COEF0_C0(0x12a) |	/*  1.1641 (-0.03 %) */
404 			BF_PXP_CSC1_COEF0_UV_OFFSET(-128) |
405 			BF_PXP_CSC1_COEF0_Y_OFFSET(-16),
406 			BF_PXP_CSC1_COEF1_C1(0x1ad) |	/*  1.6758 (-0.29 %) */
407 			BF_PXP_CSC1_COEF1_C4(0x224),	/*  2.1406 (-0.11 %) */
408 			BF_PXP_CSC1_COEF2_C2(0x75a) |	/* -0.6484 (+0.20 %) */
409 			BF_PXP_CSC1_COEF2_C3(0x7d1),	/* -0.1836 (+0.38 %) */
410 		};
411 		/*
412 		 * BT.2020 full range:
413 		 *
414 		 * |R|   |1.0000  0.0000  1.4746|   |Y  + 0  |
415 		 * |G| = |1.0000 -0.1646 -0.5714| * |Cb - 128|
416 		 * |B|   |1.0000  1.8814  0.0000|   |Cr - 128|
417 		 */
418 		static const u32 csc1_coef_bt2020_full[3] = {
419 			BM_PXP_CSC1_COEF0_YCBCR_MODE |
420 			BF_PXP_CSC1_COEF0_C0(0x100) |	/*  1.0000 (+0.00 %) */
421 			BF_PXP_CSC1_COEF0_UV_OFFSET(-128) |
422 			BF_PXP_CSC1_COEF0_Y_OFFSET(0),
423 			BF_PXP_CSC1_COEF1_C1(0x179) |	/*  1.4727 (-0.19 %) */
424 			BF_PXP_CSC1_COEF1_C4(0x1e1),	/*  1.8789 (-0.25 %) */
425 			BF_PXP_CSC1_COEF2_C2(0x76e) |	/* -0.5703 (+0.11 %) */
426 			BF_PXP_CSC1_COEF2_C3(0x7d6),	/* -0.1641 (+0.05 %) */
427 		};
428 		/*
429 		 * SMPTE 240m limited range:
430 		 *
431 		 * |R|   |1.1644  0.0000  1.7937|   |Y  - 16 |
432 		 * |G| = |1.1644 -0.2565 -0.5427| * |Cb - 128|
433 		 * |B|   |1.1644  2.0798  0.0000|   |Cr - 128|
434 		 */
435 		static const u32 csc1_coef_smpte240m_lim[3] = {
436 			BM_PXP_CSC1_COEF0_YCBCR_MODE |
437 			BF_PXP_CSC1_COEF0_C0(0x12a) |	/*  1.1641 (-0.03 %) */
438 			BF_PXP_CSC1_COEF0_UV_OFFSET(-128) |
439 			BF_PXP_CSC1_COEF0_Y_OFFSET(-16),
440 			BF_PXP_CSC1_COEF1_C1(0x1cb) |	/*  1.7930 (-0.07 %) */
441 			BF_PXP_CSC1_COEF1_C4(0x214),	/*  2.0781 (-0.17 %) */
442 			BF_PXP_CSC1_COEF2_C2(0x776) |	/* -0.5391 (+0.36 %) */
443 			BF_PXP_CSC1_COEF2_C3(0x7bf),	/* -0.2539 (+0.26 %) */
444 		};
445 		/*
446 		 * SMPTE 240m full range:
447 		 *
448 		 * |R|   |1.0000  0.0000  1.5756|   |Y  + 0  |
449 		 * |G| = |1.0000 -0.2253 -0.4767| * |Cb - 128|
450 		 * |B|   |1.0000  1.8270  0.0000|   |Cr - 128|
451 		 */
452 		static const u32 csc1_coef_smpte240m_full[3] = {
453 			BM_PXP_CSC1_COEF0_YCBCR_MODE |
454 			BF_PXP_CSC1_COEF0_C0(0x100) |	/*  1.0000 (+0.00 %) */
455 			BF_PXP_CSC1_COEF0_UV_OFFSET(-128) |
456 			BF_PXP_CSC1_COEF0_Y_OFFSET(0),
457 			BF_PXP_CSC1_COEF1_C1(0x193) |	/*  1.5742 (-0.14 %) */
458 			BF_PXP_CSC1_COEF1_C4(0x1d3),	/*  1.8242 (-0.28 %) */
459 			BF_PXP_CSC1_COEF2_C2(0x786) |	/* -0.4766 (+0.01 %) */
460 			BF_PXP_CSC1_COEF2_C3(0x7c7),	/* -0.2227 (+0.26 %) */
461 		};
462 		const u32 *csc1_coef;
463 
464 		ycbcr_enc = ctx->q_data[V4L2_M2M_SRC].ycbcr_enc;
465 		quantization = ctx->q_data[V4L2_M2M_SRC].quant;
466 
467 		if (ycbcr_enc == V4L2_YCBCR_ENC_601) {
468 			if (quantization == V4L2_QUANTIZATION_FULL_RANGE)
469 				csc1_coef = csc1_coef_bt601_full;
470 			else
471 				csc1_coef = csc1_coef_bt601_lim;
472 		} else if (ycbcr_enc == V4L2_YCBCR_ENC_709) {
473 			if (quantization == V4L2_QUANTIZATION_FULL_RANGE)
474 				csc1_coef = csc1_coef_rec709_full;
475 			else
476 				csc1_coef = csc1_coef_rec709_lim;
477 		} else if (ycbcr_enc == V4L2_YCBCR_ENC_BT2020) {
478 			if (quantization == V4L2_QUANTIZATION_FULL_RANGE)
479 				csc1_coef = csc1_coef_bt2020_full;
480 			else
481 				csc1_coef = csc1_coef_bt2020_lim;
482 		} else {
483 			if (quantization == V4L2_QUANTIZATION_FULL_RANGE)
484 				csc1_coef = csc1_coef_smpte240m_full;
485 			else
486 				csc1_coef = csc1_coef_smpte240m_lim;
487 		}
488 
489 		writel(csc1_coef[0], dev->mmio + HW_PXP_CSC1_COEF0);
490 		writel(csc1_coef[1], dev->mmio + HW_PXP_CSC1_COEF1);
491 		writel(csc1_coef[2], dev->mmio + HW_PXP_CSC1_COEF2);
492 	} else {
493 		writel(BM_PXP_CSC1_COEF0_BYPASS, dev->mmio + HW_PXP_CSC1_COEF0);
494 	}
495 
496 	if (!pxp_v4l2_pix_fmt_is_yuv(ctx->q_data[V4L2_M2M_SRC].fmt->fourcc) &&
497 	    pxp_v4l2_pix_fmt_is_yuv(ctx->q_data[V4L2_M2M_DST].fmt->fourcc)) {
498 		/*
499 		 * CSC2 RGB to YUV/YCbCr conversion is implemented as follows:
500 		 *
501 		 * |Y |   |A1 A2 A3|   |R|   |D1|
502 		 * |Cb| = |B1 B2 B3| * |G| + |D2|
503 		 * |Cr|   |C1 C2 C3|   |B|   |D3|
504 		 *
505 		 * Results are clamped to 0..255.
506 		 *
507 		 * BT.601 limited range:
508 		 *
509 		 * |Y |   | 0.2568  0.5041  0.0979|   |R|   |16 |
510 		 * |Cb| = |-0.1482 -0.2910  0.4392| * |G| + |128|
511 		 * |Cr|   | 0.4392  0.4392 -0.3678|   |B|   |128|
512 		 */
513 		static const u32 csc2_coef_bt601_lim[6] = {
514 			BF_PXP_CSC2_COEF0_A2(0x081) |	/*  0.5039 (-0.02 %) */
515 			BF_PXP_CSC2_COEF0_A1(0x041),	/*  0.2539 (-0.29 %) */
516 			BF_PXP_CSC2_COEF1_B1(0x7db) |	/* -0.1445 (+0.37 %) */
517 			BF_PXP_CSC2_COEF1_A3(0x019),	/*  0.0977 (-0.02 %) */
518 			BF_PXP_CSC2_COEF2_B3(0x070) |	/*  0.4375 (-0.17 %) */
519 			BF_PXP_CSC2_COEF2_B2(0x7b6),	/* -0.2891 (+0.20 %) */
520 			BF_PXP_CSC2_COEF3_C2(0x7a2) |	/* -0.3672 (+0.06 %) */
521 			BF_PXP_CSC2_COEF3_C1(0x070),	/*  0.4375 (-0.17 %) */
522 			BF_PXP_CSC2_COEF4_D1(16) |
523 			BF_PXP_CSC2_COEF4_C3(0x7ee),	/* -0.0703 (+0.11 %) */
524 			BF_PXP_CSC2_COEF5_D3(128) |
525 			BF_PXP_CSC2_COEF5_D2(128),
526 		};
527 		/*
528 		 * BT.601 full range:
529 		 *
530 		 * |Y |   | 0.2990  0.5870  0.1140|   |R|   |0  |
531 		 * |Cb| = |-0.1687 -0.3313  0.5000| * |G| + |128|
532 		 * |Cr|   | 0.5000  0.5000 -0.4187|   |B|   |128|
533 		 */
534 		static const u32 csc2_coef_bt601_full[6] = {
535 			BF_PXP_CSC2_COEF0_A2(0x096) |	/*  0.5859 (-0.11 %) */
536 			BF_PXP_CSC2_COEF0_A1(0x04c),	/*  0.2969 (-0.21 %) */
537 			BF_PXP_CSC2_COEF1_B1(0x7d5) |	/* -0.1680 (+0.07 %) */
538 			BF_PXP_CSC2_COEF1_A3(0x01d),	/*  0.1133 (-0.07 %) */
539 			BF_PXP_CSC2_COEF2_B3(0x080) |	/*  0.5000 (+0.00 %) */
540 			BF_PXP_CSC2_COEF2_B2(0x7ac),	/* -0.3281 (+0.32 %) */
541 			BF_PXP_CSC2_COEF3_C2(0x795) |	/* -0.4180 (+0.07 %) */
542 			BF_PXP_CSC2_COEF3_C1(0x080),	/*  0.5000 (+0.00 %) */
543 			BF_PXP_CSC2_COEF4_D1(0) |
544 			BF_PXP_CSC2_COEF4_C3(0x7ec),	/* -0.0781 (+0.32 %) */
545 			BF_PXP_CSC2_COEF5_D3(128) |
546 			BF_PXP_CSC2_COEF5_D2(128),
547 		};
548 		/*
549 		 * Rec.709 limited range:
550 		 *
551 		 * |Y |   | 0.1826  0.6142  0.0620|   |R|   |16 |
552 		 * |Cb| = |-0.1007 -0.3385  0.4392| * |G| + |128|
553 		 * |Cr|   | 0.4392  0.4392 -0.3990|   |B|   |128|
554 		 */
555 		static const u32 csc2_coef_rec709_lim[6] = {
556 			BF_PXP_CSC2_COEF0_A2(0x09d) |	/*  0.6133 (-0.09 %) */
557 			BF_PXP_CSC2_COEF0_A1(0x02e),	/*  0.1797 (-0.29 %) */
558 			BF_PXP_CSC2_COEF1_B1(0x7e7) |	/* -0.0977 (+0.30 %) */
559 			BF_PXP_CSC2_COEF1_A3(0x00f),	/*  0.0586 (-0.34 %) */
560 			BF_PXP_CSC2_COEF2_B3(0x070) |	/*  0.4375 (-0.17 %) */
561 			BF_PXP_CSC2_COEF2_B2(0x7aa),	/* -0.3359 (+0.26 %) */
562 			BF_PXP_CSC2_COEF3_C2(0x79a) |	/* -0.3984 (+0.05 %) */
563 			BF_PXP_CSC2_COEF3_C1(0x070),	/*  0.4375 (-0.17 %) */
564 			BF_PXP_CSC2_COEF4_D1(16) |
565 			BF_PXP_CSC2_COEF4_C3(0x7f6),	/* -0.0391 (+0.12 %) */
566 			BF_PXP_CSC2_COEF5_D3(128) |
567 			BF_PXP_CSC2_COEF5_D2(128),
568 		};
569 		/*
570 		 * Rec.709 full range:
571 		 *
572 		 * |Y |   | 0.2126  0.7152  0.0722|   |R|   |0  |
573 		 * |Cb| = |-0.1146 -0.3854  0.5000| * |G| + |128|
574 		 * |Cr|   | 0.5000  0.5000 -0.4542|   |B|   |128|
575 		 */
576 		static const u32 csc2_coef_rec709_full[6] = {
577 			BF_PXP_CSC2_COEF0_A2(0x0b7) |	/*  0.7148 (-0.04 %) */
578 			BF_PXP_CSC2_COEF0_A1(0x036),	/*  0.2109 (-0.17 %) */
579 			BF_PXP_CSC2_COEF1_B1(0x7e3) |	/* -0.1133 (+0.13 %) */
580 			BF_PXP_CSC2_COEF1_A3(0x012),	/*  0.0703 (-0.19 %) */
581 			BF_PXP_CSC2_COEF2_B3(0x080) |	/*  0.5000 (+0.00 %) */
582 			BF_PXP_CSC2_COEF2_B2(0x79e),	/* -0.3828 (+0.26 %) */
583 			BF_PXP_CSC2_COEF3_C2(0x78c) |	/* -0.4531 (+0.11 %) */
584 			BF_PXP_CSC2_COEF3_C1(0x080),	/*  0.5000 (+0.00 %) */
585 			BF_PXP_CSC2_COEF4_D1(0) |
586 			BF_PXP_CSC2_COEF4_C3(0x7f5),	/* -0.0430 (+0.28 %) */
587 			BF_PXP_CSC2_COEF5_D3(128) |
588 			BF_PXP_CSC2_COEF5_D2(128),
589 		};
590 		/*
591 		 * BT.2020 limited range:
592 		 *
593 		 * |Y |   | 0.2256  0.5823  0.0509|   |R|   |16 |
594 		 * |Cb| = |-0.1226 -0.3166  0.4392| * |G| + |128|
595 		 * |Cr|   | 0.4392  0.4392 -0.4039|   |B|   |128|
596 		 */
597 		static const u32 csc2_coef_bt2020_lim[6] = {
598 			BF_PXP_CSC2_COEF0_A2(0x095) |	/*  0.5820 (-0.03 %) */
599 			BF_PXP_CSC2_COEF0_A1(0x039),	/*  0.2227 (-0.30 %) */
600 			BF_PXP_CSC2_COEF1_B1(0x7e1) |	/* -0.1211 (+0.15 %) */
601 			BF_PXP_CSC2_COEF1_A3(0x00d),	/*  0.0508 (-0.01 %) */
602 			BF_PXP_CSC2_COEF2_B3(0x070) |	/*  0.4375 (-0.17 %) */
603 			BF_PXP_CSC2_COEF2_B2(0x7af),	/* -0.3164 (+0.02 %) */
604 			BF_PXP_CSC2_COEF3_C2(0x799) |	/* -0.4023 (+0.16 %) */
605 			BF_PXP_CSC2_COEF3_C1(0x070),	/*  0.4375 (-0.17 %) */
606 			BF_PXP_CSC2_COEF4_D1(16) |
607 			BF_PXP_CSC2_COEF4_C3(0x7f7),	/* -0.0352 (+0.02 %) */
608 			BF_PXP_CSC2_COEF5_D3(128) |
609 			BF_PXP_CSC2_COEF5_D2(128),
610 		};
611 		/*
612 		 * BT.2020 full range:
613 		 *
614 		 * |Y |   | 0.2627  0.6780  0.0593|   |R|   |0  |
615 		 * |Cb| = |-0.1396 -0.3604  0.5000| * |G| + |128|
616 		 * |Cr|   | 0.5000  0.5000 -0.4598|   |B|   |128|
617 		 */
618 		static const u32 csc2_coef_bt2020_full[6] = {
619 			BF_PXP_CSC2_COEF0_A2(0x0ad) |	/*  0.6758 (-0.22 %) */
620 			BF_PXP_CSC2_COEF0_A1(0x043),	/*  0.2617 (-0.10 %) */
621 			BF_PXP_CSC2_COEF1_B1(0x7dd) |	/* -0.1367 (+0.29 %) */
622 			BF_PXP_CSC2_COEF1_A3(0x00f),	/*  0.0586 (-0.07 %) */
623 			BF_PXP_CSC2_COEF2_B3(0x080) |	/*  0.5000 (+0.00 %) */
624 			BF_PXP_CSC2_COEF2_B2(0x7a4),	/* -0.3594 (+0.10 %) */
625 			BF_PXP_CSC2_COEF3_C2(0x78b) |	/* -0.4570 (+0.28 %) */
626 			BF_PXP_CSC2_COEF3_C1(0x080),	/*  0.5000 (+0.00 %) */
627 			BF_PXP_CSC2_COEF4_D1(0) |
628 			BF_PXP_CSC2_COEF4_C3(0x7f6),	/* -0.0391 (+0.11 %) */
629 			BF_PXP_CSC2_COEF5_D3(128) |
630 			BF_PXP_CSC2_COEF5_D2(128),
631 		};
632 		/*
633 		 * SMPTE 240m limited range:
634 		 *
635 		 * |Y |   | 0.1821  0.6020  0.0747|   |R|   |16 |
636 		 * |Cb| = |-0.1019 -0.3373  0.4392| * |G| + |128|
637 		 * |Cr|   | 0.4392  0.4392 -0.3909|   |B|   |128|
638 		 */
639 		static const u32 csc2_coef_smpte240m_lim[6] = {
640 			BF_PXP_CSC2_COEF0_A2(0x09a) |	/*  0.6016 (-0.05 %) */
641 			BF_PXP_CSC2_COEF0_A1(0x02e),	/*  0.1797 (-0.24 %) */
642 			BF_PXP_CSC2_COEF1_B1(0x7e6) |	/* -0.1016 (+0.03 %) */
643 			BF_PXP_CSC2_COEF1_A3(0x013),	/*  0.0742 (-0.05 %) */
644 			BF_PXP_CSC2_COEF2_B3(0x070) |	/*  0.4375 (-0.17 %) */
645 			BF_PXP_CSC2_COEF2_B2(0x7aa),	/* -0.3359 (+0.14 %) */
646 			BF_PXP_CSC2_COEF3_C2(0x79c) |	/* -0.3906 (+0.03 %) */
647 			BF_PXP_CSC2_COEF3_C1(0x070),	/*  0.4375 (-0.17 %) */
648 			BF_PXP_CSC2_COEF4_D1(16) |
649 			BF_PXP_CSC2_COEF4_C3(0x7f4),	/* -0.0469 (+0.14 %) */
650 			BF_PXP_CSC2_COEF5_D3(128) |
651 			BF_PXP_CSC2_COEF5_D2(128),
652 		};
653 		/*
654 		 * SMPTE 240m full range:
655 		 *
656 		 * |Y |   | 0.2120  0.7010  0.0870|   |R|   |0  |
657 		 * |Cb| = |-0.1160 -0.3840  0.5000| * |G| + |128|
658 		 * |Cr|   | 0.5000  0.5000 -0.4450|   |B|   |128|
659 		 */
660 		static const u32 csc2_coef_smpte240m_full[6] = {
661 			BF_PXP_CSC2_COEF0_A2(0x0b3) |	/*  0.6992 (-0.18 %) */
662 			BF_PXP_CSC2_COEF0_A1(0x036),	/*  0.2109 (-0.11 %) */
663 			BF_PXP_CSC2_COEF1_B1(0x7e3) |	/* -0.1133 (+0.27 %) */
664 			BF_PXP_CSC2_COEF1_A3(0x016),	/*  0.0859 (-0.11 %) */
665 			BF_PXP_CSC2_COEF2_B3(0x080) |	/*  0.5000 (+0.00 %) */
666 			BF_PXP_CSC2_COEF2_B2(0x79e),	/* -0.3828 (+0.12 %) */
667 			BF_PXP_CSC2_COEF3_C2(0x78f) |	/* -0.4414 (+0.36 %) */
668 			BF_PXP_CSC2_COEF3_C1(0x080),	/*  0.5000 (+0.00 %) */
669 			BF_PXP_CSC2_COEF4_D1(0) |
670 			BF_PXP_CSC2_COEF4_C3(0x7f2),	/* -0.0547 (+0.03 %) */
671 			BF_PXP_CSC2_COEF5_D3(128) |
672 			BF_PXP_CSC2_COEF5_D2(128),
673 		};
674 		const u32 *csc2_coef;
675 		u32 csc2_ctrl;
676 
677 		ycbcr_enc = ctx->q_data[V4L2_M2M_DST].ycbcr_enc;
678 		quantization = ctx->q_data[V4L2_M2M_DST].quant;
679 
680 		if (ycbcr_enc == V4L2_YCBCR_ENC_601) {
681 			if (quantization == V4L2_QUANTIZATION_FULL_RANGE)
682 				csc2_coef = csc2_coef_bt601_full;
683 			else
684 				csc2_coef = csc2_coef_bt601_lim;
685 		} else if (ycbcr_enc == V4L2_YCBCR_ENC_709) {
686 			if (quantization == V4L2_QUANTIZATION_FULL_RANGE)
687 				csc2_coef = csc2_coef_rec709_full;
688 			else
689 				csc2_coef = csc2_coef_rec709_lim;
690 		} else if (ycbcr_enc == V4L2_YCBCR_ENC_BT2020) {
691 			if (quantization == V4L2_QUANTIZATION_FULL_RANGE)
692 				csc2_coef = csc2_coef_bt2020_full;
693 			else
694 				csc2_coef = csc2_coef_bt2020_lim;
695 		} else {
696 			if (quantization == V4L2_QUANTIZATION_FULL_RANGE)
697 				csc2_coef = csc2_coef_smpte240m_full;
698 			else
699 				csc2_coef = csc2_coef_smpte240m_lim;
700 		}
701 		if (quantization == V4L2_QUANTIZATION_FULL_RANGE) {
702 			csc2_ctrl = BV_PXP_CSC2_CTRL_CSC_MODE__RGB2YUV <<
703 				    BP_PXP_CSC2_CTRL_CSC_MODE;
704 		} else {
705 			csc2_ctrl = BV_PXP_CSC2_CTRL_CSC_MODE__RGB2YCbCr <<
706 				    BP_PXP_CSC2_CTRL_CSC_MODE;
707 		}
708 
709 		writel(csc2_ctrl, dev->mmio + HW_PXP_CSC2_CTRL);
710 		writel(csc2_coef[0], dev->mmio + HW_PXP_CSC2_COEF0);
711 		writel(csc2_coef[1], dev->mmio + HW_PXP_CSC2_COEF1);
712 		writel(csc2_coef[2], dev->mmio + HW_PXP_CSC2_COEF2);
713 		writel(csc2_coef[3], dev->mmio + HW_PXP_CSC2_COEF3);
714 		writel(csc2_coef[4], dev->mmio + HW_PXP_CSC2_COEF4);
715 		writel(csc2_coef[5], dev->mmio + HW_PXP_CSC2_COEF5);
716 	} else {
717 		writel(BM_PXP_CSC2_CTRL_BYPASS, dev->mmio + HW_PXP_CSC2_CTRL);
718 	}
719 }
720 
721 static int pxp_start(struct pxp_ctx *ctx, struct vb2_v4l2_buffer *in_vb,
722 		     struct vb2_v4l2_buffer *out_vb)
723 {
724 	struct pxp_dev *dev = ctx->dev;
725 	struct pxp_q_data *q_data;
726 	u32 src_width, src_height, src_stride, src_fourcc;
727 	u32 dst_width, dst_height, dst_stride, dst_fourcc;
728 	dma_addr_t p_in, p_out;
729 	u32 ctrl, out_ctrl, out_buf, out_buf2, out_pitch, out_lrc, out_ps_ulc;
730 	u32 out_ps_lrc;
731 	u32 ps_ctrl, ps_buf, ps_ubuf, ps_vbuf, ps_pitch, ps_scale, ps_offset;
732 	u32 as_ulc, as_lrc;
733 	u32 y_size;
734 	u32 decx, decy, xscale, yscale;
735 
736 	q_data = get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_OUTPUT);
737 
738 	src_width = ctx->q_data[V4L2_M2M_SRC].width;
739 	dst_width = ctx->q_data[V4L2_M2M_DST].width;
740 	src_height = ctx->q_data[V4L2_M2M_SRC].height;
741 	dst_height = ctx->q_data[V4L2_M2M_DST].height;
742 	src_stride = ctx->q_data[V4L2_M2M_SRC].bytesperline;
743 	dst_stride = ctx->q_data[V4L2_M2M_DST].bytesperline;
744 	src_fourcc = ctx->q_data[V4L2_M2M_SRC].fmt->fourcc;
745 	dst_fourcc = ctx->q_data[V4L2_M2M_DST].fmt->fourcc;
746 
747 	p_in = vb2_dma_contig_plane_dma_addr(&in_vb->vb2_buf, 0);
748 	p_out = vb2_dma_contig_plane_dma_addr(&out_vb->vb2_buf, 0);
749 
750 	if (!p_in || !p_out) {
751 		v4l2_err(&dev->v4l2_dev,
752 			 "Acquiring DMA addresses of buffers failed\n");
753 		return -EFAULT;
754 	}
755 
756 	out_vb->sequence =
757 		get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_CAPTURE)->sequence++;
758 	in_vb->sequence = q_data->sequence++;
759 	out_vb->vb2_buf.timestamp = in_vb->vb2_buf.timestamp;
760 
761 	if (in_vb->flags & V4L2_BUF_FLAG_TIMECODE)
762 		out_vb->timecode = in_vb->timecode;
763 	out_vb->field = in_vb->field;
764 	out_vb->flags = in_vb->flags &
765 		(V4L2_BUF_FLAG_TIMECODE |
766 		 V4L2_BUF_FLAG_KEYFRAME |
767 		 V4L2_BUF_FLAG_PFRAME |
768 		 V4L2_BUF_FLAG_BFRAME |
769 		 V4L2_BUF_FLAG_TSTAMP_SRC_MASK);
770 
771 	/* 8x8 block size */
772 	ctrl = BF_PXP_CTRL_VFLIP0(!!(ctx->mode & MEM2MEM_VFLIP)) |
773 	       BF_PXP_CTRL_HFLIP0(!!(ctx->mode & MEM2MEM_HFLIP)) |
774 	       BF_PXP_CTRL_ROTATE0(ctx->rotation);
775 	/* Always write alpha value as V4L2_CID_ALPHA_COMPONENT */
776 	out_ctrl = BF_PXP_OUT_CTRL_ALPHA(ctx->alpha_component) |
777 		   BF_PXP_OUT_CTRL_ALPHA_OUTPUT(1) |
778 		   pxp_v4l2_pix_fmt_to_out_format(dst_fourcc);
779 	out_buf = p_out;
780 
781 	if (ctx->rotation == BV_PXP_CTRL_ROTATE0__ROT_90 ||
782 	    ctx->rotation == BV_PXP_CTRL_ROTATE0__ROT_270)
783 		swap(dst_width, dst_height);
784 
785 	switch (dst_fourcc) {
786 	case V4L2_PIX_FMT_NV12:
787 	case V4L2_PIX_FMT_NV21:
788 	case V4L2_PIX_FMT_NV16:
789 	case V4L2_PIX_FMT_NV61:
790 		out_buf2 = out_buf + dst_stride * dst_height;
791 		break;
792 	default:
793 		out_buf2 = 0;
794 	}
795 
796 	out_pitch = BF_PXP_OUT_PITCH_PITCH(dst_stride);
797 	out_lrc = BF_PXP_OUT_LRC_X(dst_width - 1) |
798 		  BF_PXP_OUT_LRC_Y(dst_height - 1);
799 	/* PS covers whole output */
800 	out_ps_ulc = BF_PXP_OUT_PS_ULC_X(0) | BF_PXP_OUT_PS_ULC_Y(0);
801 	out_ps_lrc = BF_PXP_OUT_PS_LRC_X(dst_width - 1) |
802 		     BF_PXP_OUT_PS_LRC_Y(dst_height - 1);
803 	/* no AS */
804 	as_ulc = BF_PXP_OUT_AS_ULC_X(1) | BF_PXP_OUT_AS_ULC_Y(1);
805 	as_lrc = BF_PXP_OUT_AS_LRC_X(0) | BF_PXP_OUT_AS_LRC_Y(0);
806 
807 	decx = (src_width <= dst_width) ? 0 : ilog2(src_width / dst_width);
808 	decy = (src_height <= dst_height) ? 0 : ilog2(src_height / dst_height);
809 	ps_ctrl = BF_PXP_PS_CTRL_DECX(decx) | BF_PXP_PS_CTRL_DECY(decy) |
810 		  pxp_v4l2_pix_fmt_to_ps_format(src_fourcc);
811 	ps_buf = p_in;
812 	y_size = src_stride * src_height;
813 	switch (src_fourcc) {
814 	case V4L2_PIX_FMT_YUV420:
815 		ps_ubuf = ps_buf + y_size;
816 		ps_vbuf = ps_ubuf + y_size / 4;
817 		break;
818 	case V4L2_PIX_FMT_YUV422P:
819 		ps_ubuf = ps_buf + y_size;
820 		ps_vbuf = ps_ubuf + y_size / 2;
821 		break;
822 	case V4L2_PIX_FMT_NV12:
823 	case V4L2_PIX_FMT_NV21:
824 	case V4L2_PIX_FMT_NV16:
825 	case V4L2_PIX_FMT_NV61:
826 		ps_ubuf = ps_buf + y_size;
827 		ps_vbuf = 0;
828 		break;
829 	case V4L2_PIX_FMT_GREY:
830 	case V4L2_PIX_FMT_Y4:
831 		ps_ubuf = 0;
832 		/* In grayscale mode, ps_vbuf contents are reused as CbCr */
833 		ps_vbuf = 0x8080;
834 		break;
835 	default:
836 		ps_ubuf = 0;
837 		ps_vbuf = 0;
838 		break;
839 	}
840 	ps_pitch = BF_PXP_PS_PITCH_PITCH(src_stride);
841 	if (decx) {
842 		xscale = (src_width >> decx) * 0x1000 / dst_width;
843 	} else {
844 		switch (src_fourcc) {
845 		case V4L2_PIX_FMT_UYVY:
846 		case V4L2_PIX_FMT_YUYV:
847 		case V4L2_PIX_FMT_VYUY:
848 		case V4L2_PIX_FMT_YVYU:
849 		case V4L2_PIX_FMT_NV16:
850 		case V4L2_PIX_FMT_NV12:
851 		case V4L2_PIX_FMT_NV21:
852 		case V4L2_PIX_FMT_NV61:
853 		case V4L2_PIX_FMT_YUV422P:
854 		case V4L2_PIX_FMT_YUV420:
855 			/*
856 			 * This avoids sampling past the right edge for
857 			 * horizontally chroma subsampled formats.
858 			 */
859 			xscale = (src_width - 2) * 0x1000 / (dst_width - 1);
860 			break;
861 		default:
862 			xscale = (src_width - 1) * 0x1000 / (dst_width - 1);
863 			break;
864 		}
865 	}
866 	if (decy)
867 		yscale = (src_height >> decy) * 0x1000 / dst_height;
868 	else
869 		yscale = (src_height - 1) * 0x1000 / (dst_height - 1);
870 	ps_scale = BF_PXP_PS_SCALE_YSCALE(yscale) |
871 		   BF_PXP_PS_SCALE_XSCALE(xscale);
872 	ps_offset = BF_PXP_PS_OFFSET_YOFFSET(0) | BF_PXP_PS_OFFSET_XOFFSET(0);
873 
874 	writel(ctrl, dev->mmio + HW_PXP_CTRL);
875 	/* skip STAT */
876 	writel(out_ctrl, dev->mmio + HW_PXP_OUT_CTRL);
877 	writel(out_buf, dev->mmio + HW_PXP_OUT_BUF);
878 	writel(out_buf2, dev->mmio + HW_PXP_OUT_BUF2);
879 	writel(out_pitch, dev->mmio + HW_PXP_OUT_PITCH);
880 	writel(out_lrc, dev->mmio + HW_PXP_OUT_LRC);
881 	writel(out_ps_ulc, dev->mmio + HW_PXP_OUT_PS_ULC);
882 	writel(out_ps_lrc, dev->mmio + HW_PXP_OUT_PS_LRC);
883 	writel(as_ulc, dev->mmio + HW_PXP_OUT_AS_ULC);
884 	writel(as_lrc, dev->mmio + HW_PXP_OUT_AS_LRC);
885 	writel(ps_ctrl, dev->mmio + HW_PXP_PS_CTRL);
886 	writel(ps_buf, dev->mmio + HW_PXP_PS_BUF);
887 	writel(ps_ubuf, dev->mmio + HW_PXP_PS_UBUF);
888 	writel(ps_vbuf, dev->mmio + HW_PXP_PS_VBUF);
889 	writel(ps_pitch, dev->mmio + HW_PXP_PS_PITCH);
890 	writel(0x00ffffff, dev->mmio + HW_PXP_PS_BACKGROUND_0);
891 	writel(ps_scale, dev->mmio + HW_PXP_PS_SCALE);
892 	writel(ps_offset, dev->mmio + HW_PXP_PS_OFFSET);
893 	/* disable processed surface color keying */
894 	writel(0x00ffffff, dev->mmio + HW_PXP_PS_CLRKEYLOW_0);
895 	writel(0x00000000, dev->mmio + HW_PXP_PS_CLRKEYHIGH_0);
896 
897 	/* disable alpha surface color keying */
898 	writel(0x00ffffff, dev->mmio + HW_PXP_AS_CLRKEYLOW_0);
899 	writel(0x00000000, dev->mmio + HW_PXP_AS_CLRKEYHIGH_0);
900 
901 	/* setup CSC */
902 	pxp_setup_csc(ctx);
903 
904 	/* bypass LUT */
905 	writel(BM_PXP_LUT_CTRL_BYPASS, dev->mmio + HW_PXP_LUT_CTRL);
906 
907 	writel(BF_PXP_DATA_PATH_CTRL0_MUX15_SEL(0)|
908 	       BF_PXP_DATA_PATH_CTRL0_MUX14_SEL(1)|
909 	       BF_PXP_DATA_PATH_CTRL0_MUX13_SEL(0)|
910 	       BF_PXP_DATA_PATH_CTRL0_MUX12_SEL(0)|
911 	       BF_PXP_DATA_PATH_CTRL0_MUX11_SEL(0)|
912 	       BF_PXP_DATA_PATH_CTRL0_MUX10_SEL(0)|
913 	       BF_PXP_DATA_PATH_CTRL0_MUX9_SEL(1)|
914 	       BF_PXP_DATA_PATH_CTRL0_MUX8_SEL(0)|
915 	       BF_PXP_DATA_PATH_CTRL0_MUX7_SEL(0)|
916 	       BF_PXP_DATA_PATH_CTRL0_MUX6_SEL(0)|
917 	       BF_PXP_DATA_PATH_CTRL0_MUX5_SEL(0)|
918 	       BF_PXP_DATA_PATH_CTRL0_MUX4_SEL(0)|
919 	       BF_PXP_DATA_PATH_CTRL0_MUX3_SEL(0)|
920 	       BF_PXP_DATA_PATH_CTRL0_MUX2_SEL(0)|
921 	       BF_PXP_DATA_PATH_CTRL0_MUX1_SEL(0)|
922 	       BF_PXP_DATA_PATH_CTRL0_MUX0_SEL(0),
923 	       dev->mmio + HW_PXP_DATA_PATH_CTRL0);
924 	writel(BF_PXP_DATA_PATH_CTRL1_MUX17_SEL(1) |
925 	       BF_PXP_DATA_PATH_CTRL1_MUX16_SEL(1),
926 	       dev->mmio + HW_PXP_DATA_PATH_CTRL1);
927 
928 	writel(0xffff, dev->mmio + HW_PXP_IRQ_MASK);
929 
930 	/* ungate, enable PS/AS/OUT and PXP operation */
931 	writel(BM_PXP_CTRL_IRQ_ENABLE, dev->mmio + HW_PXP_CTRL_SET);
932 	writel(BM_PXP_CTRL_ENABLE | BM_PXP_CTRL_ENABLE_CSC2 |
933 	       BM_PXP_CTRL_ENABLE_LUT | BM_PXP_CTRL_ENABLE_ROTATE0 |
934 	       BM_PXP_CTRL_ENABLE_PS_AS_OUT, dev->mmio + HW_PXP_CTRL_SET);
935 
936 	return 0;
937 }
938 
939 static void pxp_job_finish(struct pxp_dev *dev)
940 {
941 	struct pxp_ctx *curr_ctx;
942 	struct vb2_v4l2_buffer *src_vb, *dst_vb;
943 	unsigned long flags;
944 
945 	curr_ctx = v4l2_m2m_get_curr_priv(dev->m2m_dev);
946 
947 	if (curr_ctx == NULL) {
948 		pr_err("Instance released before the end of transaction\n");
949 		return;
950 	}
951 
952 	src_vb = v4l2_m2m_src_buf_remove(curr_ctx->fh.m2m_ctx);
953 	dst_vb = v4l2_m2m_dst_buf_remove(curr_ctx->fh.m2m_ctx);
954 
955 	spin_lock_irqsave(&dev->irqlock, flags);
956 	v4l2_m2m_buf_done(src_vb, VB2_BUF_STATE_DONE);
957 	v4l2_m2m_buf_done(dst_vb, VB2_BUF_STATE_DONE);
958 	spin_unlock_irqrestore(&dev->irqlock, flags);
959 
960 	dprintk(curr_ctx->dev, "Finishing transaction\n");
961 	v4l2_m2m_job_finish(dev->m2m_dev, curr_ctx->fh.m2m_ctx);
962 }
963 
964 /*
965  * mem2mem callbacks
966  */
967 static void pxp_device_run(void *priv)
968 {
969 	struct pxp_ctx *ctx = priv;
970 	struct vb2_v4l2_buffer *src_buf, *dst_buf;
971 
972 	src_buf = v4l2_m2m_next_src_buf(ctx->fh.m2m_ctx);
973 	dst_buf = v4l2_m2m_next_dst_buf(ctx->fh.m2m_ctx);
974 
975 	pxp_start(ctx, src_buf, dst_buf);
976 }
977 
978 static int pxp_job_ready(void *priv)
979 {
980 	struct pxp_ctx *ctx = priv;
981 
982 	if (v4l2_m2m_num_src_bufs_ready(ctx->fh.m2m_ctx) < 1 ||
983 	    v4l2_m2m_num_dst_bufs_ready(ctx->fh.m2m_ctx) < 1) {
984 		dprintk(ctx->dev, "Not enough buffers available\n");
985 		return 0;
986 	}
987 
988 	return 1;
989 }
990 
991 static void pxp_job_abort(void *priv)
992 {
993 	struct pxp_ctx *ctx = priv;
994 
995 	/* Will cancel the transaction in the next interrupt handler */
996 	ctx->aborting = 1;
997 }
998 
999 /*
1000  * interrupt handler
1001  */
1002 static irqreturn_t pxp_irq_handler(int irq, void *dev_id)
1003 {
1004 	struct pxp_dev *dev = dev_id;
1005 	u32 stat;
1006 
1007 	stat = readl(dev->mmio + HW_PXP_STAT);
1008 
1009 	if (stat & BM_PXP_STAT_IRQ0) {
1010 		/* we expect x = 0, y = height, irq0 = 1 */
1011 		if (stat & ~(BM_PXP_STAT_BLOCKX | BM_PXP_STAT_BLOCKY |
1012 			     BM_PXP_STAT_IRQ0))
1013 			dprintk(dev, "%s: stat = 0x%08x\n", __func__, stat);
1014 		writel(BM_PXP_STAT_IRQ0, dev->mmio + HW_PXP_STAT_CLR);
1015 
1016 		pxp_job_finish(dev);
1017 	} else {
1018 		u32 irq = readl(dev->mmio + HW_PXP_IRQ);
1019 
1020 		dprintk(dev, "%s: stat = 0x%08x\n", __func__, stat);
1021 		dprintk(dev, "%s: irq = 0x%08x\n", __func__, irq);
1022 
1023 		writel(irq, dev->mmio + HW_PXP_IRQ_CLR);
1024 	}
1025 
1026 	return IRQ_HANDLED;
1027 }
1028 
1029 /*
1030  * video ioctls
1031  */
1032 static int pxp_querycap(struct file *file, void *priv,
1033 			   struct v4l2_capability *cap)
1034 {
1035 	strscpy(cap->driver, MEM2MEM_NAME, sizeof(cap->driver));
1036 	strscpy(cap->card, MEM2MEM_NAME, sizeof(cap->card));
1037 	snprintf(cap->bus_info, sizeof(cap->bus_info),
1038 			"platform:%s", MEM2MEM_NAME);
1039 	return 0;
1040 }
1041 
1042 static int pxp_enum_fmt(struct v4l2_fmtdesc *f, u32 type)
1043 {
1044 	int i, num;
1045 	struct pxp_fmt *fmt;
1046 
1047 	num = 0;
1048 
1049 	for (i = 0; i < NUM_FORMATS; ++i) {
1050 		if (formats[i].types & type) {
1051 			/* index-th format of type type found ? */
1052 			if (num == f->index)
1053 				break;
1054 			/*
1055 			 * Correct type but haven't reached our index yet,
1056 			 * just increment per-type index
1057 			 */
1058 			++num;
1059 		}
1060 	}
1061 
1062 	if (i < NUM_FORMATS) {
1063 		/* Format found */
1064 		fmt = &formats[i];
1065 		f->pixelformat = fmt->fourcc;
1066 		return 0;
1067 	}
1068 
1069 	/* Format not found */
1070 	return -EINVAL;
1071 }
1072 
1073 static int pxp_enum_fmt_vid_cap(struct file *file, void *priv,
1074 				struct v4l2_fmtdesc *f)
1075 {
1076 	return pxp_enum_fmt(f, MEM2MEM_CAPTURE);
1077 }
1078 
1079 static int pxp_enum_fmt_vid_out(struct file *file, void *priv,
1080 				struct v4l2_fmtdesc *f)
1081 {
1082 	return pxp_enum_fmt(f, MEM2MEM_OUTPUT);
1083 }
1084 
1085 static int pxp_g_fmt(struct pxp_ctx *ctx, struct v4l2_format *f)
1086 {
1087 	struct vb2_queue *vq;
1088 	struct pxp_q_data *q_data;
1089 
1090 	vq = v4l2_m2m_get_vq(ctx->fh.m2m_ctx, f->type);
1091 	if (!vq)
1092 		return -EINVAL;
1093 
1094 	q_data = get_q_data(ctx, f->type);
1095 
1096 	f->fmt.pix.width	= q_data->width;
1097 	f->fmt.pix.height	= q_data->height;
1098 	f->fmt.pix.field	= V4L2_FIELD_NONE;
1099 	f->fmt.pix.pixelformat	= q_data->fmt->fourcc;
1100 	f->fmt.pix.bytesperline	= q_data->bytesperline;
1101 	f->fmt.pix.sizeimage	= q_data->sizeimage;
1102 	f->fmt.pix.colorspace	= ctx->colorspace;
1103 	f->fmt.pix.xfer_func	= ctx->xfer_func;
1104 	f->fmt.pix.ycbcr_enc	= q_data->ycbcr_enc;
1105 	f->fmt.pix.quantization	= q_data->quant;
1106 
1107 	return 0;
1108 }
1109 
1110 static int pxp_g_fmt_vid_out(struct file *file, void *priv,
1111 				struct v4l2_format *f)
1112 {
1113 	return pxp_g_fmt(file2ctx(file), f);
1114 }
1115 
1116 static int pxp_g_fmt_vid_cap(struct file *file, void *priv,
1117 				struct v4l2_format *f)
1118 {
1119 	return pxp_g_fmt(file2ctx(file), f);
1120 }
1121 
1122 static inline u32 pxp_bytesperline(struct pxp_fmt *fmt, u32 width)
1123 {
1124 	switch (fmt->fourcc) {
1125 	case V4L2_PIX_FMT_YUV420:
1126 	case V4L2_PIX_FMT_NV12:
1127 	case V4L2_PIX_FMT_NV21:
1128 	case V4L2_PIX_FMT_YUV422P:
1129 	case V4L2_PIX_FMT_NV16:
1130 	case V4L2_PIX_FMT_NV61:
1131 		return width;
1132 	default:
1133 		return (width * fmt->depth) >> 3;
1134 	}
1135 }
1136 
1137 static inline u32 pxp_sizeimage(struct pxp_fmt *fmt, u32 width, u32 height)
1138 {
1139 	return (fmt->depth * width * height) >> 3;
1140 }
1141 
1142 static int pxp_try_fmt(struct v4l2_format *f, struct pxp_fmt *fmt)
1143 {
1144 	v4l_bound_align_image(&f->fmt.pix.width, MIN_W, MAX_W, ALIGN_W,
1145 			      &f->fmt.pix.height, MIN_H, MAX_H, ALIGN_H, 0);
1146 
1147 	f->fmt.pix.bytesperline = pxp_bytesperline(fmt, f->fmt.pix.width);
1148 	f->fmt.pix.sizeimage = pxp_sizeimage(fmt, f->fmt.pix.width,
1149 					     f->fmt.pix.height);
1150 	f->fmt.pix.field = V4L2_FIELD_NONE;
1151 
1152 	return 0;
1153 }
1154 
1155 static void
1156 pxp_fixup_colorimetry_cap(struct pxp_ctx *ctx, u32 dst_fourcc,
1157 			  enum v4l2_ycbcr_encoding *ycbcr_enc,
1158 			  enum v4l2_quantization *quantization)
1159 {
1160 	bool dst_is_yuv = pxp_v4l2_pix_fmt_is_yuv(dst_fourcc);
1161 
1162 	if (pxp_v4l2_pix_fmt_is_yuv(ctx->q_data[V4L2_M2M_SRC].fmt->fourcc) ==
1163 	    dst_is_yuv) {
1164 		/*
1165 		 * There is no support for conversion between different YCbCr
1166 		 * encodings or between RGB limited and full range.
1167 		 */
1168 		*ycbcr_enc = ctx->q_data[V4L2_M2M_SRC].ycbcr_enc;
1169 		*quantization = ctx->q_data[V4L2_M2M_SRC].quant;
1170 	} else {
1171 		*ycbcr_enc = V4L2_MAP_YCBCR_ENC_DEFAULT(ctx->colorspace);
1172 		*quantization = V4L2_MAP_QUANTIZATION_DEFAULT(!dst_is_yuv,
1173 							      ctx->colorspace,
1174 							      *ycbcr_enc);
1175 	}
1176 }
1177 
1178 static int pxp_try_fmt_vid_cap(struct file *file, void *priv,
1179 			       struct v4l2_format *f)
1180 {
1181 	struct pxp_fmt *fmt;
1182 	struct pxp_ctx *ctx = file2ctx(file);
1183 
1184 	fmt = find_format(f);
1185 	if (!fmt) {
1186 		f->fmt.pix.pixelformat = formats[0].fourcc;
1187 		fmt = find_format(f);
1188 	}
1189 	if (!(fmt->types & MEM2MEM_CAPTURE)) {
1190 		v4l2_err(&ctx->dev->v4l2_dev,
1191 			 "Fourcc format (0x%08x) invalid.\n",
1192 			 f->fmt.pix.pixelformat);
1193 		return -EINVAL;
1194 	}
1195 
1196 	f->fmt.pix.colorspace = ctx->colorspace;
1197 	f->fmt.pix.xfer_func = ctx->xfer_func;
1198 
1199 	pxp_fixup_colorimetry_cap(ctx, fmt->fourcc,
1200 				  &f->fmt.pix.ycbcr_enc,
1201 				  &f->fmt.pix.quantization);
1202 
1203 	return pxp_try_fmt(f, fmt);
1204 }
1205 
1206 static int pxp_try_fmt_vid_out(struct file *file, void *priv,
1207 			       struct v4l2_format *f)
1208 {
1209 	struct pxp_fmt *fmt;
1210 	struct pxp_ctx *ctx = file2ctx(file);
1211 
1212 	fmt = find_format(f);
1213 	if (!fmt) {
1214 		f->fmt.pix.pixelformat = formats[0].fourcc;
1215 		fmt = find_format(f);
1216 	}
1217 	if (!(fmt->types & MEM2MEM_OUTPUT)) {
1218 		v4l2_err(&ctx->dev->v4l2_dev,
1219 			 "Fourcc format (0x%08x) invalid.\n",
1220 			 f->fmt.pix.pixelformat);
1221 		return -EINVAL;
1222 	}
1223 
1224 	if (!f->fmt.pix.colorspace)
1225 		f->fmt.pix.colorspace = V4L2_COLORSPACE_REC709;
1226 
1227 	return pxp_try_fmt(f, fmt);
1228 }
1229 
1230 static int pxp_s_fmt(struct pxp_ctx *ctx, struct v4l2_format *f)
1231 {
1232 	struct pxp_q_data *q_data;
1233 	struct vb2_queue *vq;
1234 
1235 	vq = v4l2_m2m_get_vq(ctx->fh.m2m_ctx, f->type);
1236 	if (!vq)
1237 		return -EINVAL;
1238 
1239 	q_data = get_q_data(ctx, f->type);
1240 	if (!q_data)
1241 		return -EINVAL;
1242 
1243 	if (vb2_is_busy(vq)) {
1244 		v4l2_err(&ctx->dev->v4l2_dev, "%s queue busy\n", __func__);
1245 		return -EBUSY;
1246 	}
1247 
1248 	q_data->fmt		= find_format(f);
1249 	q_data->width		= f->fmt.pix.width;
1250 	q_data->height		= f->fmt.pix.height;
1251 	q_data->bytesperline	= f->fmt.pix.bytesperline;
1252 	q_data->sizeimage	= f->fmt.pix.sizeimage;
1253 
1254 	dprintk(ctx->dev,
1255 		"Setting format for type %d, wxh: %dx%d, fmt: %d\n",
1256 		f->type, q_data->width, q_data->height, q_data->fmt->fourcc);
1257 
1258 	return 0;
1259 }
1260 
1261 static int pxp_s_fmt_vid_cap(struct file *file, void *priv,
1262 			     struct v4l2_format *f)
1263 {
1264 	struct pxp_ctx *ctx = file2ctx(file);
1265 	int ret;
1266 
1267 	ret = pxp_try_fmt_vid_cap(file, priv, f);
1268 	if (ret)
1269 		return ret;
1270 
1271 	ret = pxp_s_fmt(file2ctx(file), f);
1272 	if (ret)
1273 		return ret;
1274 
1275 	ctx->q_data[V4L2_M2M_DST].ycbcr_enc = f->fmt.pix.ycbcr_enc;
1276 	ctx->q_data[V4L2_M2M_DST].quant = f->fmt.pix.quantization;
1277 
1278 	return 0;
1279 }
1280 
1281 static int pxp_s_fmt_vid_out(struct file *file, void *priv,
1282 			     struct v4l2_format *f)
1283 {
1284 	struct pxp_ctx *ctx = file2ctx(file);
1285 	int ret;
1286 
1287 	ret = pxp_try_fmt_vid_out(file, priv, f);
1288 	if (ret)
1289 		return ret;
1290 
1291 	ret = pxp_s_fmt(file2ctx(file), f);
1292 	if (ret)
1293 		return ret;
1294 
1295 	ctx->colorspace = f->fmt.pix.colorspace;
1296 	ctx->xfer_func = f->fmt.pix.xfer_func;
1297 	ctx->q_data[V4L2_M2M_SRC].ycbcr_enc = f->fmt.pix.ycbcr_enc;
1298 	ctx->q_data[V4L2_M2M_SRC].quant = f->fmt.pix.quantization;
1299 
1300 	pxp_fixup_colorimetry_cap(ctx, ctx->q_data[V4L2_M2M_DST].fmt->fourcc,
1301 				  &ctx->q_data[V4L2_M2M_DST].ycbcr_enc,
1302 				  &ctx->q_data[V4L2_M2M_DST].quant);
1303 
1304 	return 0;
1305 }
1306 
1307 static u8 pxp_degrees_to_rot_mode(u32 degrees)
1308 {
1309 	switch (degrees) {
1310 	case 90:
1311 		return BV_PXP_CTRL_ROTATE0__ROT_90;
1312 	case 180:
1313 		return BV_PXP_CTRL_ROTATE0__ROT_180;
1314 	case 270:
1315 		return BV_PXP_CTRL_ROTATE0__ROT_270;
1316 	case 0:
1317 	default:
1318 		return BV_PXP_CTRL_ROTATE0__ROT_0;
1319 	}
1320 }
1321 
1322 static int pxp_s_ctrl(struct v4l2_ctrl *ctrl)
1323 {
1324 	struct pxp_ctx *ctx =
1325 		container_of(ctrl->handler, struct pxp_ctx, hdl);
1326 
1327 	switch (ctrl->id) {
1328 	case V4L2_CID_HFLIP:
1329 		if (ctrl->val)
1330 			ctx->mode |= MEM2MEM_HFLIP;
1331 		else
1332 			ctx->mode &= ~MEM2MEM_HFLIP;
1333 		break;
1334 
1335 	case V4L2_CID_VFLIP:
1336 		if (ctrl->val)
1337 			ctx->mode |= MEM2MEM_VFLIP;
1338 		else
1339 			ctx->mode &= ~MEM2MEM_VFLIP;
1340 		break;
1341 
1342 	case V4L2_CID_ROTATE:
1343 		ctx->rotation = pxp_degrees_to_rot_mode(ctrl->val);
1344 		break;
1345 
1346 	case V4L2_CID_ALPHA_COMPONENT:
1347 		ctx->alpha_component = ctrl->val;
1348 		break;
1349 
1350 	default:
1351 		v4l2_err(&ctx->dev->v4l2_dev, "Invalid control\n");
1352 		return -EINVAL;
1353 	}
1354 
1355 	return 0;
1356 }
1357 
1358 static const struct v4l2_ctrl_ops pxp_ctrl_ops = {
1359 	.s_ctrl = pxp_s_ctrl,
1360 };
1361 
1362 static const struct v4l2_ioctl_ops pxp_ioctl_ops = {
1363 	.vidioc_querycap	= pxp_querycap,
1364 
1365 	.vidioc_enum_fmt_vid_cap = pxp_enum_fmt_vid_cap,
1366 	.vidioc_g_fmt_vid_cap	= pxp_g_fmt_vid_cap,
1367 	.vidioc_try_fmt_vid_cap	= pxp_try_fmt_vid_cap,
1368 	.vidioc_s_fmt_vid_cap	= pxp_s_fmt_vid_cap,
1369 
1370 	.vidioc_enum_fmt_vid_out = pxp_enum_fmt_vid_out,
1371 	.vidioc_g_fmt_vid_out	= pxp_g_fmt_vid_out,
1372 	.vidioc_try_fmt_vid_out	= pxp_try_fmt_vid_out,
1373 	.vidioc_s_fmt_vid_out	= pxp_s_fmt_vid_out,
1374 
1375 	.vidioc_reqbufs		= v4l2_m2m_ioctl_reqbufs,
1376 	.vidioc_querybuf	= v4l2_m2m_ioctl_querybuf,
1377 	.vidioc_qbuf		= v4l2_m2m_ioctl_qbuf,
1378 	.vidioc_dqbuf		= v4l2_m2m_ioctl_dqbuf,
1379 	.vidioc_prepare_buf	= v4l2_m2m_ioctl_prepare_buf,
1380 	.vidioc_create_bufs	= v4l2_m2m_ioctl_create_bufs,
1381 	.vidioc_expbuf		= v4l2_m2m_ioctl_expbuf,
1382 
1383 	.vidioc_streamon	= v4l2_m2m_ioctl_streamon,
1384 	.vidioc_streamoff	= v4l2_m2m_ioctl_streamoff,
1385 
1386 	.vidioc_subscribe_event = v4l2_ctrl_subscribe_event,
1387 	.vidioc_unsubscribe_event = v4l2_event_unsubscribe,
1388 };
1389 
1390 /*
1391  * Queue operations
1392  */
1393 static int pxp_queue_setup(struct vb2_queue *vq,
1394 			   unsigned int *nbuffers, unsigned int *nplanes,
1395 			   unsigned int sizes[], struct device *alloc_devs[])
1396 {
1397 	struct pxp_ctx *ctx = vb2_get_drv_priv(vq);
1398 	struct pxp_q_data *q_data;
1399 	unsigned int size, count = *nbuffers;
1400 
1401 	q_data = get_q_data(ctx, vq->type);
1402 
1403 	size = q_data->sizeimage;
1404 
1405 	*nbuffers = count;
1406 
1407 	if (*nplanes)
1408 		return sizes[0] < size ? -EINVAL : 0;
1409 
1410 	*nplanes = 1;
1411 	sizes[0] = size;
1412 
1413 	dprintk(ctx->dev, "get %d buffer(s) of size %d each.\n", count, size);
1414 
1415 	return 0;
1416 }
1417 
1418 static int pxp_buf_prepare(struct vb2_buffer *vb)
1419 {
1420 	struct vb2_v4l2_buffer *vbuf = to_vb2_v4l2_buffer(vb);
1421 	struct pxp_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
1422 	struct pxp_dev *dev = ctx->dev;
1423 	struct pxp_q_data *q_data;
1424 
1425 	dprintk(ctx->dev, "type: %d\n", vb->vb2_queue->type);
1426 
1427 	q_data = get_q_data(ctx, vb->vb2_queue->type);
1428 	if (V4L2_TYPE_IS_OUTPUT(vb->vb2_queue->type)) {
1429 		if (vbuf->field == V4L2_FIELD_ANY)
1430 			vbuf->field = V4L2_FIELD_NONE;
1431 		if (vbuf->field != V4L2_FIELD_NONE) {
1432 			dprintk(dev, "%s field isn't supported\n", __func__);
1433 			return -EINVAL;
1434 		}
1435 	}
1436 
1437 	if (vb2_plane_size(vb, 0) < q_data->sizeimage) {
1438 		dprintk(dev, "%s data will not fit into plane (%lu < %lu)\n",
1439 			__func__, vb2_plane_size(vb, 0),
1440 			(long)q_data->sizeimage);
1441 		return -EINVAL;
1442 	}
1443 
1444 	vb2_set_plane_payload(vb, 0, q_data->sizeimage);
1445 
1446 	return 0;
1447 }
1448 
1449 static void pxp_buf_queue(struct vb2_buffer *vb)
1450 {
1451 	struct vb2_v4l2_buffer *vbuf = to_vb2_v4l2_buffer(vb);
1452 	struct pxp_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
1453 
1454 	v4l2_m2m_buf_queue(ctx->fh.m2m_ctx, vbuf);
1455 }
1456 
1457 static int pxp_start_streaming(struct vb2_queue *q, unsigned int count)
1458 {
1459 	struct pxp_ctx *ctx = vb2_get_drv_priv(q);
1460 	struct pxp_q_data *q_data = get_q_data(ctx, q->type);
1461 
1462 	q_data->sequence = 0;
1463 	return 0;
1464 }
1465 
1466 static void pxp_stop_streaming(struct vb2_queue *q)
1467 {
1468 	struct pxp_ctx *ctx = vb2_get_drv_priv(q);
1469 	struct vb2_v4l2_buffer *vbuf;
1470 	unsigned long flags;
1471 
1472 	for (;;) {
1473 		if (V4L2_TYPE_IS_OUTPUT(q->type))
1474 			vbuf = v4l2_m2m_src_buf_remove(ctx->fh.m2m_ctx);
1475 		else
1476 			vbuf = v4l2_m2m_dst_buf_remove(ctx->fh.m2m_ctx);
1477 		if (vbuf == NULL)
1478 			return;
1479 		spin_lock_irqsave(&ctx->dev->irqlock, flags);
1480 		v4l2_m2m_buf_done(vbuf, VB2_BUF_STATE_ERROR);
1481 		spin_unlock_irqrestore(&ctx->dev->irqlock, flags);
1482 	}
1483 }
1484 
1485 static const struct vb2_ops pxp_qops = {
1486 	.queue_setup	 = pxp_queue_setup,
1487 	.buf_prepare	 = pxp_buf_prepare,
1488 	.buf_queue	 = pxp_buf_queue,
1489 	.start_streaming = pxp_start_streaming,
1490 	.stop_streaming  = pxp_stop_streaming,
1491 	.wait_prepare	 = vb2_ops_wait_prepare,
1492 	.wait_finish	 = vb2_ops_wait_finish,
1493 };
1494 
1495 static int queue_init(void *priv, struct vb2_queue *src_vq,
1496 		      struct vb2_queue *dst_vq)
1497 {
1498 	struct pxp_ctx *ctx = priv;
1499 	int ret;
1500 
1501 	src_vq->type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
1502 	src_vq->io_modes = VB2_MMAP | VB2_DMABUF;
1503 	src_vq->drv_priv = ctx;
1504 	src_vq->buf_struct_size = sizeof(struct v4l2_m2m_buffer);
1505 	src_vq->ops = &pxp_qops;
1506 	src_vq->mem_ops = &vb2_dma_contig_memops;
1507 	src_vq->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_COPY;
1508 	src_vq->lock = &ctx->dev->dev_mutex;
1509 	src_vq->dev = ctx->dev->v4l2_dev.dev;
1510 
1511 	ret = vb2_queue_init(src_vq);
1512 	if (ret)
1513 		return ret;
1514 
1515 	dst_vq->type = V4L2_BUF_TYPE_VIDEO_CAPTURE;
1516 	dst_vq->io_modes = VB2_MMAP | VB2_DMABUF;
1517 	dst_vq->drv_priv = ctx;
1518 	dst_vq->buf_struct_size = sizeof(struct v4l2_m2m_buffer);
1519 	dst_vq->ops = &pxp_qops;
1520 	dst_vq->mem_ops = &vb2_dma_contig_memops;
1521 	dst_vq->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_COPY;
1522 	dst_vq->lock = &ctx->dev->dev_mutex;
1523 	dst_vq->dev = ctx->dev->v4l2_dev.dev;
1524 
1525 	return vb2_queue_init(dst_vq);
1526 }
1527 
1528 /*
1529  * File operations
1530  */
1531 static int pxp_open(struct file *file)
1532 {
1533 	struct pxp_dev *dev = video_drvdata(file);
1534 	struct pxp_ctx *ctx = NULL;
1535 	struct v4l2_ctrl_handler *hdl;
1536 	int rc = 0;
1537 
1538 	if (mutex_lock_interruptible(&dev->dev_mutex))
1539 		return -ERESTARTSYS;
1540 	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
1541 	if (!ctx) {
1542 		rc = -ENOMEM;
1543 		goto open_unlock;
1544 	}
1545 
1546 	v4l2_fh_init(&ctx->fh, video_devdata(file));
1547 	file->private_data = &ctx->fh;
1548 	ctx->dev = dev;
1549 	hdl = &ctx->hdl;
1550 	v4l2_ctrl_handler_init(hdl, 4);
1551 	v4l2_ctrl_new_std(hdl, &pxp_ctrl_ops, V4L2_CID_HFLIP, 0, 1, 1, 0);
1552 	v4l2_ctrl_new_std(hdl, &pxp_ctrl_ops, V4L2_CID_VFLIP, 0, 1, 1, 0);
1553 	v4l2_ctrl_new_std(hdl, &pxp_ctrl_ops, V4L2_CID_ROTATE, 0, 270, 90, 0);
1554 	v4l2_ctrl_new_std(hdl, &pxp_ctrl_ops, V4L2_CID_ALPHA_COMPONENT,
1555 			  0, 255, 1, 255);
1556 	if (hdl->error) {
1557 		rc = hdl->error;
1558 		v4l2_ctrl_handler_free(hdl);
1559 		kfree(ctx);
1560 		goto open_unlock;
1561 	}
1562 	ctx->fh.ctrl_handler = hdl;
1563 	v4l2_ctrl_handler_setup(hdl);
1564 
1565 	ctx->q_data[V4L2_M2M_SRC].fmt = &formats[0];
1566 	ctx->q_data[V4L2_M2M_SRC].width = 640;
1567 	ctx->q_data[V4L2_M2M_SRC].height = 480;
1568 	ctx->q_data[V4L2_M2M_SRC].bytesperline =
1569 		pxp_bytesperline(&formats[0], 640);
1570 	ctx->q_data[V4L2_M2M_SRC].sizeimage =
1571 		pxp_sizeimage(&formats[0], 640, 480);
1572 	ctx->q_data[V4L2_M2M_DST] = ctx->q_data[V4L2_M2M_SRC];
1573 	ctx->colorspace = V4L2_COLORSPACE_REC709;
1574 
1575 	ctx->fh.m2m_ctx = v4l2_m2m_ctx_init(dev->m2m_dev, ctx, &queue_init);
1576 
1577 	if (IS_ERR(ctx->fh.m2m_ctx)) {
1578 		rc = PTR_ERR(ctx->fh.m2m_ctx);
1579 
1580 		v4l2_ctrl_handler_free(hdl);
1581 		v4l2_fh_exit(&ctx->fh);
1582 		kfree(ctx);
1583 		goto open_unlock;
1584 	}
1585 
1586 	v4l2_fh_add(&ctx->fh);
1587 	atomic_inc(&dev->num_inst);
1588 
1589 	dprintk(dev, "Created instance: %p, m2m_ctx: %p\n",
1590 		ctx, ctx->fh.m2m_ctx);
1591 
1592 open_unlock:
1593 	mutex_unlock(&dev->dev_mutex);
1594 	return rc;
1595 }
1596 
1597 static int pxp_release(struct file *file)
1598 {
1599 	struct pxp_dev *dev = video_drvdata(file);
1600 	struct pxp_ctx *ctx = file2ctx(file);
1601 
1602 	dprintk(dev, "Releasing instance %p\n", ctx);
1603 
1604 	v4l2_fh_del(&ctx->fh);
1605 	v4l2_fh_exit(&ctx->fh);
1606 	v4l2_ctrl_handler_free(&ctx->hdl);
1607 	mutex_lock(&dev->dev_mutex);
1608 	v4l2_m2m_ctx_release(ctx->fh.m2m_ctx);
1609 	mutex_unlock(&dev->dev_mutex);
1610 	kfree(ctx);
1611 
1612 	atomic_dec(&dev->num_inst);
1613 
1614 	return 0;
1615 }
1616 
1617 static const struct v4l2_file_operations pxp_fops = {
1618 	.owner		= THIS_MODULE,
1619 	.open		= pxp_open,
1620 	.release	= pxp_release,
1621 	.poll		= v4l2_m2m_fop_poll,
1622 	.unlocked_ioctl	= video_ioctl2,
1623 	.mmap		= v4l2_m2m_fop_mmap,
1624 };
1625 
1626 static const struct video_device pxp_videodev = {
1627 	.name		= MEM2MEM_NAME,
1628 	.vfl_dir	= VFL_DIR_M2M,
1629 	.fops		= &pxp_fops,
1630 	.device_caps	= V4L2_CAP_VIDEO_M2M | V4L2_CAP_STREAMING,
1631 	.ioctl_ops	= &pxp_ioctl_ops,
1632 	.minor		= -1,
1633 	.release	= video_device_release_empty,
1634 };
1635 
1636 static const struct v4l2_m2m_ops m2m_ops = {
1637 	.device_run	= pxp_device_run,
1638 	.job_ready	= pxp_job_ready,
1639 	.job_abort	= pxp_job_abort,
1640 };
1641 
1642 static int pxp_soft_reset(struct pxp_dev *dev)
1643 {
1644 	int ret;
1645 	u32 val;
1646 
1647 	writel(BM_PXP_CTRL_SFTRST, dev->mmio + HW_PXP_CTRL_CLR);
1648 	writel(BM_PXP_CTRL_CLKGATE, dev->mmio + HW_PXP_CTRL_CLR);
1649 
1650 	writel(BM_PXP_CTRL_SFTRST, dev->mmio + HW_PXP_CTRL_SET);
1651 
1652 	ret = readl_poll_timeout(dev->mmio + HW_PXP_CTRL, val,
1653 				 val & BM_PXP_CTRL_CLKGATE, 0, 100);
1654 	if (ret < 0)
1655 		return ret;
1656 
1657 	writel(BM_PXP_CTRL_SFTRST, dev->mmio + HW_PXP_CTRL_CLR);
1658 	writel(BM_PXP_CTRL_CLKGATE, dev->mmio + HW_PXP_CTRL_CLR);
1659 
1660 	return 0;
1661 }
1662 
1663 static int pxp_probe(struct platform_device *pdev)
1664 {
1665 	struct pxp_dev *dev;
1666 	struct video_device *vfd;
1667 	int irq;
1668 	int ret;
1669 
1670 	dev = devm_kzalloc(&pdev->dev, sizeof(*dev), GFP_KERNEL);
1671 	if (!dev)
1672 		return -ENOMEM;
1673 
1674 	dev->clk = devm_clk_get(&pdev->dev, "axi");
1675 	if (IS_ERR(dev->clk)) {
1676 		ret = PTR_ERR(dev->clk);
1677 		dev_err(&pdev->dev, "Failed to get clk: %d\n", ret);
1678 		return ret;
1679 	}
1680 
1681 	dev->mmio = devm_platform_ioremap_resource(pdev, 0);
1682 	if (IS_ERR(dev->mmio))
1683 		return PTR_ERR(dev->mmio);
1684 
1685 	irq = platform_get_irq(pdev, 0);
1686 	if (irq < 0)
1687 		return irq;
1688 
1689 	spin_lock_init(&dev->irqlock);
1690 
1691 	ret = devm_request_threaded_irq(&pdev->dev, irq, NULL, pxp_irq_handler,
1692 			IRQF_ONESHOT, dev_name(&pdev->dev), dev);
1693 	if (ret < 0) {
1694 		dev_err(&pdev->dev, "Failed to request irq: %d\n", ret);
1695 		return ret;
1696 	}
1697 
1698 	ret = clk_prepare_enable(dev->clk);
1699 	if (ret < 0)
1700 		return ret;
1701 
1702 	ret = pxp_soft_reset(dev);
1703 	if (ret < 0) {
1704 		dev_err(&pdev->dev, "PXP reset timeout: %d\n", ret);
1705 		goto err_clk;
1706 	}
1707 
1708 	ret = v4l2_device_register(&pdev->dev, &dev->v4l2_dev);
1709 	if (ret)
1710 		goto err_clk;
1711 
1712 	atomic_set(&dev->num_inst, 0);
1713 	mutex_init(&dev->dev_mutex);
1714 
1715 	dev->vfd = pxp_videodev;
1716 	vfd = &dev->vfd;
1717 	vfd->lock = &dev->dev_mutex;
1718 	vfd->v4l2_dev = &dev->v4l2_dev;
1719 
1720 	video_set_drvdata(vfd, dev);
1721 	snprintf(vfd->name, sizeof(vfd->name), "%s", pxp_videodev.name);
1722 	v4l2_info(&dev->v4l2_dev,
1723 			"Device registered as /dev/video%d\n", vfd->num);
1724 
1725 	platform_set_drvdata(pdev, dev);
1726 
1727 	dev->m2m_dev = v4l2_m2m_init(&m2m_ops);
1728 	if (IS_ERR(dev->m2m_dev)) {
1729 		v4l2_err(&dev->v4l2_dev, "Failed to init mem2mem device\n");
1730 		ret = PTR_ERR(dev->m2m_dev);
1731 		goto err_v4l2;
1732 	}
1733 
1734 	ret = video_register_device(vfd, VFL_TYPE_VIDEO, 0);
1735 	if (ret) {
1736 		v4l2_err(&dev->v4l2_dev, "Failed to register video device\n");
1737 		goto err_m2m;
1738 	}
1739 
1740 	return 0;
1741 
1742 err_m2m:
1743 	v4l2_m2m_release(dev->m2m_dev);
1744 err_v4l2:
1745 	v4l2_device_unregister(&dev->v4l2_dev);
1746 err_clk:
1747 	clk_disable_unprepare(dev->clk);
1748 
1749 	return ret;
1750 }
1751 
1752 static int pxp_remove(struct platform_device *pdev)
1753 {
1754 	struct pxp_dev *dev = platform_get_drvdata(pdev);
1755 
1756 	writel(BM_PXP_CTRL_CLKGATE, dev->mmio + HW_PXP_CTRL_SET);
1757 	writel(BM_PXP_CTRL_SFTRST, dev->mmio + HW_PXP_CTRL_SET);
1758 
1759 	clk_disable_unprepare(dev->clk);
1760 
1761 	v4l2_info(&dev->v4l2_dev, "Removing " MEM2MEM_NAME);
1762 	video_unregister_device(&dev->vfd);
1763 	v4l2_m2m_release(dev->m2m_dev);
1764 	v4l2_device_unregister(&dev->v4l2_dev);
1765 
1766 	return 0;
1767 }
1768 
1769 static const struct of_device_id pxp_dt_ids[] = {
1770 	{ .compatible = "fsl,imx6ull-pxp", .data = NULL },
1771 	{ },
1772 };
1773 MODULE_DEVICE_TABLE(of, pxp_dt_ids);
1774 
1775 static struct platform_driver pxp_driver = {
1776 	.probe		= pxp_probe,
1777 	.remove		= pxp_remove,
1778 	.driver		= {
1779 		.name	= MEM2MEM_NAME,
1780 		.of_match_table = pxp_dt_ids,
1781 	},
1782 };
1783 
1784 module_platform_driver(pxp_driver);
1785 
1786 MODULE_DESCRIPTION("i.MX PXP mem2mem scaler/CSC/rotator");
1787 MODULE_AUTHOR("Philipp Zabel <kernel@pengutronix.de>");
1788 MODULE_LICENSE("GPL");
1789