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