?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
PM / sleep: Asynchronous threads for suspend_noirq
[linux/fpc-iii.git] / drivers / media / platform / ti-vpe / vpe.c
blob1296c5386231e25700a0f95e05a6f0a97d72e5fa
1 /*
2 * TI VPE mem2mem driver, based on the virtual v4l2-mem2mem example driver
3 *
4 * Copyright (c) 2013 Texas Instruments Inc.
5 * David Griego, <dagriego@biglakesoftware.com>
6 * Dale Farnsworth, <dale@farnsworth.org>
7 * Archit Taneja, <archit@ti.com>
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 * Based on the virtual v4l2-mem2mem example device
14 *
15 * This program is free software; you can redistribute it and/or modify it
16 * under the terms of the GNU General Public License version 2 as published by
17 * the Free Software Foundation
18 */
19
20 #include <linux/delay.h>
21 #include <linux/dma-mapping.h>
22 #include <linux/err.h>
23 #include <linux/fs.h>
24 #include <linux/interrupt.h>
25 #include <linux/io.h>
26 #include <linux/ioctl.h>
27 #include <linux/module.h>
28 #include <linux/platform_device.h>
29 #include <linux/pm_runtime.h>
30 #include <linux/sched.h>
31 #include <linux/slab.h>
32 #include <linux/videodev2.h>
33 #include <linux/log2.h>
34
35 #include <media/v4l2-common.h>
36 #include <media/v4l2-ctrls.h>
37 #include <media/v4l2-device.h>
38 #include <media/v4l2-event.h>
39 #include <media/v4l2-ioctl.h>
40 #include <media/v4l2-mem2mem.h>
41 #include <media/videobuf2-core.h>
42 #include <media/videobuf2-dma-contig.h>
43
44 #include "vpdma.h"
45 #include "vpe_regs.h"
46 #include "sc.h"
47 #include "csc.h"
48
49 #define VPE_MODULE_NAME "vpe"
50
51 /* minimum and maximum frame sizes */
52 #define MIN_W 128
53 #define MIN_H 128
54 #define MAX_W 1920
55 #define MAX_H 1080
56
57 /* required alignments */
58 #define S_ALIGN 0 /* multiple of 1 */
59 #define H_ALIGN 1 /* multiple of 2 */
60
61 /* flags that indicate a format can be used for capture/output */
62 #define VPE_FMT_TYPE_CAPTURE (1 << 0)
63 #define VPE_FMT_TYPE_OUTPUT (1 << 1)
64
65 /* used as plane indices */
66 #define VPE_MAX_PLANES 2
67 #define VPE_LUMA 0
68 #define VPE_CHROMA 1
69
70 /* per m2m context info */
71 #define VPE_MAX_SRC_BUFS 3 /* need 3 src fields to de-interlace */
72
73 #define VPE_DEF_BUFS_PER_JOB 1 /* default one buffer per batch job */
74
75 /*
76 * each VPE context can need up to 3 config desciptors, 7 input descriptors,
77 * 3 output descriptors, and 10 control descriptors
78 */
79 #define VPE_DESC_LIST_SIZE (10 * VPDMA_DTD_DESC_SIZE + \
80 13 * VPDMA_CFD_CTD_DESC_SIZE)
81
82 #define vpe_dbg(vpedev, fmt, arg...) \
83 dev_dbg((vpedev)->v4l2_dev.dev, fmt, ##arg)
84 #define vpe_err(vpedev, fmt, arg...) \
85 dev_err((vpedev)->v4l2_dev.dev, fmt, ##arg)
86
87 struct vpe_us_coeffs {
88 unsigned short anchor_fid0_c0;
89 unsigned short anchor_fid0_c1;
90 unsigned short anchor_fid0_c2;
91 unsigned short anchor_fid0_c3;
92 unsigned short interp_fid0_c0;
93 unsigned short interp_fid0_c1;
94 unsigned short interp_fid0_c2;
95 unsigned short interp_fid0_c3;
96 unsigned short anchor_fid1_c0;
97 unsigned short anchor_fid1_c1;
98 unsigned short anchor_fid1_c2;
99 unsigned short anchor_fid1_c3;
100 unsigned short interp_fid1_c0;
101 unsigned short interp_fid1_c1;
102 unsigned short interp_fid1_c2;
103 unsigned short interp_fid1_c3;
104 };
105
106 /*
107 * Default upsampler coefficients
108 */
109 static const struct vpe_us_coeffs us_coeffs[] = {
110 {
111 /* Coefficients for progressive input */
112 0x00C8, 0x0348, 0x0018, 0x3FD8, 0x3FB8, 0x0378, 0x00E8, 0x3FE8,
113 0x00C8, 0x0348, 0x0018, 0x3FD8, 0x3FB8, 0x0378, 0x00E8, 0x3FE8,
114 },
115 {
116 /* Coefficients for Top Field Interlaced input */
117 0x0051, 0x03D5, 0x3FE3, 0x3FF7, 0x3FB5, 0x02E9, 0x018F, 0x3FD3,
118 /* Coefficients for Bottom Field Interlaced input */
119 0x016B, 0x0247, 0x00B1, 0x3F9D, 0x3FCF, 0x03DB, 0x005D, 0x3FF9,
120 },
121 };
122
123 /*
124 * the following registers are for configuring some of the parameters of the
125 * motion and edge detection blocks inside DEI, these generally remain the same,
126 * these could be passed later via userspace if some one needs to tweak these.
127 */
128 struct vpe_dei_regs {
129 unsigned long mdt_spacial_freq_thr_reg; /* VPE_DEI_REG2 */
130 unsigned long edi_config_reg; /* VPE_DEI_REG3 */
131 unsigned long edi_lut_reg0; /* VPE_DEI_REG4 */
132 unsigned long edi_lut_reg1; /* VPE_DEI_REG5 */
133 unsigned long edi_lut_reg2; /* VPE_DEI_REG6 */
134 unsigned long edi_lut_reg3; /* VPE_DEI_REG7 */
135 };
136
137 /*
138 * default expert DEI register values, unlikely to be modified.
139 */
140 static const struct vpe_dei_regs dei_regs = {
141 0x020C0804u,
142 0x0118100Fu,
143 0x08040200u,
144 0x1010100Cu,
145 0x10101010u,
146 0x10101010u,
147 };
148
149 /*
150 * The port_data structure contains per-port data.
151 */
152 struct vpe_port_data {
153 enum vpdma_channel channel; /* VPDMA channel */
154 u8 vb_index; /* input frame f, f-1, f-2 index */
155 u8 vb_part; /* plane index for co-panar formats */
156 };
157
158 /*
159 * Define indices into the port_data tables
160 */
161 #define VPE_PORT_LUMA1_IN 0
162 #define VPE_PORT_CHROMA1_IN 1
163 #define VPE_PORT_LUMA2_IN 2
164 #define VPE_PORT_CHROMA2_IN 3
165 #define VPE_PORT_LUMA3_IN 4
166 #define VPE_PORT_CHROMA3_IN 5
167 #define VPE_PORT_MV_IN 6
168 #define VPE_PORT_MV_OUT 7
169 #define VPE_PORT_LUMA_OUT 8
170 #define VPE_PORT_CHROMA_OUT 9
171 #define VPE_PORT_RGB_OUT 10
172
173 static const struct vpe_port_data port_data[11] = {
174 [VPE_PORT_LUMA1_IN] = {
175 .channel = VPE_CHAN_LUMA1_IN,
176 .vb_index = 0,
177 .vb_part = VPE_LUMA,
178 },
179 [VPE_PORT_CHROMA1_IN] = {
180 .channel = VPE_CHAN_CHROMA1_IN,
181 .vb_index = 0,
182 .vb_part = VPE_CHROMA,
183 },
184 [VPE_PORT_LUMA2_IN] = {
185 .channel = VPE_CHAN_LUMA2_IN,
186 .vb_index = 1,
187 .vb_part = VPE_LUMA,
188 },
189 [VPE_PORT_CHROMA2_IN] = {
190 .channel = VPE_CHAN_CHROMA2_IN,
191 .vb_index = 1,
192 .vb_part = VPE_CHROMA,
193 },
194 [VPE_PORT_LUMA3_IN] = {
195 .channel = VPE_CHAN_LUMA3_IN,
196 .vb_index = 2,
197 .vb_part = VPE_LUMA,
198 },
199 [VPE_PORT_CHROMA3_IN] = {
200 .channel = VPE_CHAN_CHROMA3_IN,
201 .vb_index = 2,
202 .vb_part = VPE_CHROMA,
203 },
204 [VPE_PORT_MV_IN] = {
205 .channel = VPE_CHAN_MV_IN,
206 },
207 [VPE_PORT_MV_OUT] = {
208 .channel = VPE_CHAN_MV_OUT,
209 },
210 [VPE_PORT_LUMA_OUT] = {
211 .channel = VPE_CHAN_LUMA_OUT,
212 .vb_part = VPE_LUMA,
213 },
214 [VPE_PORT_CHROMA_OUT] = {
215 .channel = VPE_CHAN_CHROMA_OUT,
216 .vb_part = VPE_CHROMA,
217 },
218 [VPE_PORT_RGB_OUT] = {
219 .channel = VPE_CHAN_RGB_OUT,
220 .vb_part = VPE_LUMA,
221 },
222 };
223
224
225 /* driver info for each of the supported video formats */
226 struct vpe_fmt {
227 char *name; /* human-readable name */
228 u32 fourcc; /* standard format identifier */
229 u8 types; /* CAPTURE and/or OUTPUT */
230 u8 coplanar; /* set for unpacked Luma and Chroma */
231 /* vpdma format info for each plane */
232 struct vpdma_data_format const *vpdma_fmt[VPE_MAX_PLANES];
233 };
234
235 static struct vpe_fmt vpe_formats[] = {
236 {
237 .name = "YUV 422 co-planar",
238 .fourcc = V4L2_PIX_FMT_NV16,
239 .types = VPE_FMT_TYPE_CAPTURE | VPE_FMT_TYPE_OUTPUT,
240 .coplanar = 1,
241 .vpdma_fmt = { &vpdma_yuv_fmts[VPDMA_DATA_FMT_Y444],
242 &vpdma_yuv_fmts[VPDMA_DATA_FMT_C444],
243 },
244 },
245 {
246 .name = "YUV 420 co-planar",
247 .fourcc = V4L2_PIX_FMT_NV12,
248 .types = VPE_FMT_TYPE_CAPTURE | VPE_FMT_TYPE_OUTPUT,
249 .coplanar = 1,
250 .vpdma_fmt = { &vpdma_yuv_fmts[VPDMA_DATA_FMT_Y420],
251 &vpdma_yuv_fmts[VPDMA_DATA_FMT_C420],
252 },
253 },
254 {
255 .name = "YUYV 422 packed",
256 .fourcc = V4L2_PIX_FMT_YUYV,
257 .types = VPE_FMT_TYPE_CAPTURE | VPE_FMT_TYPE_OUTPUT,
258 .coplanar = 0,
259 .vpdma_fmt = { &vpdma_yuv_fmts[VPDMA_DATA_FMT_YC422],
260 },
261 },
262 {
263 .name = "UYVY 422 packed",
264 .fourcc = V4L2_PIX_FMT_UYVY,
265 .types = VPE_FMT_TYPE_CAPTURE | VPE_FMT_TYPE_OUTPUT,
266 .coplanar = 0,
267 .vpdma_fmt = { &vpdma_yuv_fmts[VPDMA_DATA_FMT_CY422],
268 },
269 },
270 {
271 .name = "RGB888 packed",
272 .fourcc = V4L2_PIX_FMT_RGB24,
273 .types = VPE_FMT_TYPE_CAPTURE,
274 .coplanar = 0,
275 .vpdma_fmt = { &vpdma_rgb_fmts[VPDMA_DATA_FMT_RGB24],
276 },
277 },
278 {
279 .name = "ARGB32",
280 .fourcc = V4L2_PIX_FMT_RGB32,
281 .types = VPE_FMT_TYPE_CAPTURE,
282 .coplanar = 0,
283 .vpdma_fmt = { &vpdma_rgb_fmts[VPDMA_DATA_FMT_ARGB32],
284 },
285 },
286 {
287 .name = "BGR888 packed",
288 .fourcc = V4L2_PIX_FMT_BGR24,
289 .types = VPE_FMT_TYPE_CAPTURE,
290 .coplanar = 0,
291 .vpdma_fmt = { &vpdma_rgb_fmts[VPDMA_DATA_FMT_BGR24],
292 },
293 },
294 {
295 .name = "ABGR32",
296 .fourcc = V4L2_PIX_FMT_BGR32,
297 .types = VPE_FMT_TYPE_CAPTURE,
298 .coplanar = 0,
299 .vpdma_fmt = { &vpdma_rgb_fmts[VPDMA_DATA_FMT_ABGR32],
300 },
301 },
302 };
303
304 /*
305 * per-queue, driver-specific private data.
306 * there is one source queue and one destination queue for each m2m context.
307 */
308 struct vpe_q_data {
309 unsigned int width; /* frame width */
310 unsigned int height; /* frame height */
311 unsigned int bytesperline[VPE_MAX_PLANES]; /* bytes per line in memory */
312 enum v4l2_colorspace colorspace;
313 enum v4l2_field field; /* supported field value */
314 unsigned int flags;
315 unsigned int sizeimage[VPE_MAX_PLANES]; /* image size in memory */
316 struct v4l2_rect c_rect; /* crop/compose rectangle */
317 struct vpe_fmt *fmt; /* format info */
318 };
319
320 /* vpe_q_data flag bits */
321 #define Q_DATA_FRAME_1D (1 << 0)
322 #define Q_DATA_MODE_TILED (1 << 1)
323 #define Q_DATA_INTERLACED (1 << 2)
324
325 enum {
326 Q_DATA_SRC = 0,
327 Q_DATA_DST = 1,
328 };
329
330 /* find our format description corresponding to the passed v4l2_format */
331 static struct vpe_fmt *find_format(struct v4l2_format *f)
332 {
333 struct vpe_fmt *fmt;
334 unsigned int k;
335
336 for (k = 0; k < ARRAY_SIZE(vpe_formats); k++) {
337 fmt = &vpe_formats[k];
338 if (fmt->fourcc == f->fmt.pix.pixelformat)
339 return fmt;
340 }
341
342 return NULL;
343 }
344
345 /*
346 * there is one vpe_dev structure in the driver, it is shared by
347 * all instances.
348 */
349 struct vpe_dev {
350 struct v4l2_device v4l2_dev;
351 struct video_device vfd;
352 struct v4l2_m2m_dev *m2m_dev;
353
354 atomic_t num_instances; /* count of driver instances */
355 dma_addr_t loaded_mmrs; /* shadow mmrs in device */
356 struct mutex dev_mutex;
357 spinlock_t lock;
358
359 int irq;
360 void __iomem *base;
361 struct resource *res;
362
363 struct vb2_alloc_ctx *alloc_ctx;
364 struct vpdma_data *vpdma; /* vpdma data handle */
365 struct sc_data *sc; /* scaler data handle */
366 struct csc_data *csc; /* csc data handle */
367 };
368
369 /*
370 * There is one vpe_ctx structure for each m2m context.
371 */
372 struct vpe_ctx {
373 struct v4l2_fh fh;
374 struct vpe_dev *dev;
375 struct v4l2_m2m_ctx *m2m_ctx;
376 struct v4l2_ctrl_handler hdl;
377
378 unsigned int field; /* current field */
379 unsigned int sequence; /* current frame/field seq */
380 unsigned int aborting; /* abort after next irq */
381
382 unsigned int bufs_per_job; /* input buffers per batch */
383 unsigned int bufs_completed; /* bufs done in this batch */
384
385 struct vpe_q_data q_data[2]; /* src & dst queue data */
386 struct vb2_buffer *src_vbs[VPE_MAX_SRC_BUFS];
387 struct vb2_buffer *dst_vb;
388
389 dma_addr_t mv_buf_dma[2]; /* dma addrs of motion vector in/out bufs */
390 void *mv_buf[2]; /* virtual addrs of motion vector bufs */
391 size_t mv_buf_size; /* current motion vector buffer size */
392 struct vpdma_buf mmr_adb; /* shadow reg addr/data block */
393 struct vpdma_buf sc_coeff_h; /* h coeff buffer */
394 struct vpdma_buf sc_coeff_v; /* v coeff buffer */
395 struct vpdma_desc_list desc_list; /* DMA descriptor list */
396
397 bool deinterlacing; /* using de-interlacer */
398 bool load_mmrs; /* have new shadow reg values */
399
400 unsigned int src_mv_buf_selector;
401 };
402
403
404 /*
405 * M2M devices get 2 queues.
406 * Return the queue given the type.
407 */
408 static struct vpe_q_data *get_q_data(struct vpe_ctx *ctx,
409 enum v4l2_buf_type type)
410 {
411 switch (type) {
412 case V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE:
413 return &ctx->q_data[Q_DATA_SRC];
414 case V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE:
415 return &ctx->q_data[Q_DATA_DST];
416 default:
417 BUG();
418 }
419 return NULL;
420 }
421
422 static u32 read_reg(struct vpe_dev *dev, int offset)
423 {
424 return ioread32(dev->base + offset);
425 }
426
427 static void write_reg(struct vpe_dev *dev, int offset, u32 value)
428 {
429 iowrite32(value, dev->base + offset);
430 }
431
432 /* register field read/write helpers */
433 static int get_field(u32 value, u32 mask, int shift)
434 {
435 return (value & (mask << shift)) >> shift;
436 }
437
438 static int read_field_reg(struct vpe_dev *dev, int offset, u32 mask, int shift)
439 {
440 return get_field(read_reg(dev, offset), mask, shift);
441 }
442
443 static void write_field(u32 *valp, u32 field, u32 mask, int shift)
444 {
445 u32 val = *valp;
446
447 val &= ~(mask << shift);
448 val |= (field & mask) << shift;
449 *valp = val;
450 }
451
452 static void write_field_reg(struct vpe_dev *dev, int offset, u32 field,
453 u32 mask, int shift)
454 {
455 u32 val = read_reg(dev, offset);
456
457 write_field(&val, field, mask, shift);
458
459 write_reg(dev, offset, val);
460 }
461
462 /*
463 * DMA address/data block for the shadow registers
464 */
465 struct vpe_mmr_adb {
466 struct vpdma_adb_hdr out_fmt_hdr;
467 u32 out_fmt_reg[1];
468 u32 out_fmt_pad[3];
469 struct vpdma_adb_hdr us1_hdr;
470 u32 us1_regs[8];
471 struct vpdma_adb_hdr us2_hdr;
472 u32 us2_regs[8];
473 struct vpdma_adb_hdr us3_hdr;
474 u32 us3_regs[8];
475 struct vpdma_adb_hdr dei_hdr;
476 u32 dei_regs[8];
477 struct vpdma_adb_hdr sc_hdr0;
478 u32 sc_regs0[7];
479 u32 sc_pad0[1];
480 struct vpdma_adb_hdr sc_hdr8;
481 u32 sc_regs8[6];
482 u32 sc_pad8[2];
483 struct vpdma_adb_hdr sc_hdr17;
484 u32 sc_regs17[9];
485 u32 sc_pad17[3];
486 struct vpdma_adb_hdr csc_hdr;
487 u32 csc_regs[6];
488 u32 csc_pad[2];
489 };
490
491 #define GET_OFFSET_TOP(ctx, obj, reg) \
492 ((obj)->res->start - ctx->dev->res->start + reg)
493
494 #define VPE_SET_MMR_ADB_HDR(ctx, hdr, regs, offset_a) \
495 VPDMA_SET_MMR_ADB_HDR(ctx->mmr_adb, vpe_mmr_adb, hdr, regs, offset_a)
496 /*
497 * Set the headers for all of the address/data block structures.
498 */
499 static void init_adb_hdrs(struct vpe_ctx *ctx)
500 {
501 VPE_SET_MMR_ADB_HDR(ctx, out_fmt_hdr, out_fmt_reg, VPE_CLK_FORMAT_SELECT);
502 VPE_SET_MMR_ADB_HDR(ctx, us1_hdr, us1_regs, VPE_US1_R0);
503 VPE_SET_MMR_ADB_HDR(ctx, us2_hdr, us2_regs, VPE_US2_R0);
504 VPE_SET_MMR_ADB_HDR(ctx, us3_hdr, us3_regs, VPE_US3_R0);
505 VPE_SET_MMR_ADB_HDR(ctx, dei_hdr, dei_regs, VPE_DEI_FRAME_SIZE);
506 VPE_SET_MMR_ADB_HDR(ctx, sc_hdr0, sc_regs0,
507 GET_OFFSET_TOP(ctx, ctx->dev->sc, CFG_SC0));
508 VPE_SET_MMR_ADB_HDR(ctx, sc_hdr8, sc_regs8,
509 GET_OFFSET_TOP(ctx, ctx->dev->sc, CFG_SC8));
510 VPE_SET_MMR_ADB_HDR(ctx, sc_hdr17, sc_regs17,
511 GET_OFFSET_TOP(ctx, ctx->dev->sc, CFG_SC17));
512 VPE_SET_MMR_ADB_HDR(ctx, csc_hdr, csc_regs,
513 GET_OFFSET_TOP(ctx, ctx->dev->csc, CSC_CSC00));
514 };
515
516 /*
517 * Allocate or re-allocate the motion vector DMA buffers
518 * There are two buffers, one for input and one for output.
519 * However, the roles are reversed after each field is processed.
520 * In other words, after each field is processed, the previous
521 * output (dst) MV buffer becomes the new input (src) MV buffer.
522 */
523 static int realloc_mv_buffers(struct vpe_ctx *ctx, size_t size)
524 {
525 struct device *dev = ctx->dev->v4l2_dev.dev;
526
527 if (ctx->mv_buf_size == size)
528 return 0;
529
530 if (ctx->mv_buf[0])
531 dma_free_coherent(dev, ctx->mv_buf_size, ctx->mv_buf[0],
532 ctx->mv_buf_dma[0]);
533
534 if (ctx->mv_buf[1])
535 dma_free_coherent(dev, ctx->mv_buf_size, ctx->mv_buf[1],
536 ctx->mv_buf_dma[1]);
537
538 if (size == 0)
539 return 0;
540
541 ctx->mv_buf[0] = dma_alloc_coherent(dev, size, &ctx->mv_buf_dma[0],
542 GFP_KERNEL);
543 if (!ctx->mv_buf[0]) {
544 vpe_err(ctx->dev, "failed to allocate motion vector buffer\n");
545 return -ENOMEM;
546 }
547
548 ctx->mv_buf[1] = dma_alloc_coherent(dev, size, &ctx->mv_buf_dma[1],
549 GFP_KERNEL);
550 if (!ctx->mv_buf[1]) {
551 vpe_err(ctx->dev, "failed to allocate motion vector buffer\n");
552 dma_free_coherent(dev, size, ctx->mv_buf[0],
553 ctx->mv_buf_dma[0]);
554
555 return -ENOMEM;
556 }
557
558 ctx->mv_buf_size = size;
559 ctx->src_mv_buf_selector = 0;
560
561 return 0;
562 }
563
564 static void free_mv_buffers(struct vpe_ctx *ctx)
565 {
566 realloc_mv_buffers(ctx, 0);
567 }
568
569 /*
570 * While de-interlacing, we keep the two most recent input buffers
571 * around. This function frees those two buffers when we have
572 * finished processing the current stream.
573 */
574 static void free_vbs(struct vpe_ctx *ctx)
575 {
576 struct vpe_dev *dev = ctx->dev;
577 unsigned long flags;
578
579 if (ctx->src_vbs[2] == NULL)
580 return;
581
582 spin_lock_irqsave(&dev->lock, flags);
583 if (ctx->src_vbs[2]) {
584 v4l2_m2m_buf_done(ctx->src_vbs[2], VB2_BUF_STATE_DONE);
585 v4l2_m2m_buf_done(ctx->src_vbs[1], VB2_BUF_STATE_DONE);
586 }
587 spin_unlock_irqrestore(&dev->lock, flags);
588 }
589
590 /*
591 * Enable or disable the VPE clocks
592 */
593 static void vpe_set_clock_enable(struct vpe_dev *dev, bool on)
594 {
595 u32 val = 0;
596
597 if (on)
598 val = VPE_DATA_PATH_CLK_ENABLE | VPE_VPEDMA_CLK_ENABLE;
599 write_reg(dev, VPE_CLK_ENABLE, val);
600 }
601
602 static void vpe_top_reset(struct vpe_dev *dev)
603 {
604
605 write_field_reg(dev, VPE_CLK_RESET, 1, VPE_DATA_PATH_CLK_RESET_MASK,
606 VPE_DATA_PATH_CLK_RESET_SHIFT);
607
608 usleep_range(100, 150);
609
610 write_field_reg(dev, VPE_CLK_RESET, 0, VPE_DATA_PATH_CLK_RESET_MASK,
611 VPE_DATA_PATH_CLK_RESET_SHIFT);
612 }
613
614 static void vpe_top_vpdma_reset(struct vpe_dev *dev)
615 {
616 write_field_reg(dev, VPE_CLK_RESET, 1, VPE_VPDMA_CLK_RESET_MASK,
617 VPE_VPDMA_CLK_RESET_SHIFT);
618
619 usleep_range(100, 150);
620
621 write_field_reg(dev, VPE_CLK_RESET, 0, VPE_VPDMA_CLK_RESET_MASK,
622 VPE_VPDMA_CLK_RESET_SHIFT);
623 }
624
625 /*
626 * Load the correct of upsampler coefficients into the shadow MMRs
627 */
628 static void set_us_coefficients(struct vpe_ctx *ctx)
629 {
630 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
631 struct vpe_q_data *s_q_data = &ctx->q_data[Q_DATA_SRC];
632 u32 *us1_reg = &mmr_adb->us1_regs[0];
633 u32 *us2_reg = &mmr_adb->us2_regs[0];
634 u32 *us3_reg = &mmr_adb->us3_regs[0];
635 const unsigned short *cp, *end_cp;
636
637 cp = &us_coeffs[0].anchor_fid0_c0;
638
639 if (s_q_data->flags & Q_DATA_INTERLACED) /* interlaced */
640 cp += sizeof(us_coeffs[0]) / sizeof(*cp);
641
642 end_cp = cp + sizeof(us_coeffs[0]) / sizeof(*cp);
643
644 while (cp < end_cp) {
645 write_field(us1_reg, *cp++, VPE_US_C0_MASK, VPE_US_C0_SHIFT);
646 write_field(us1_reg, *cp++, VPE_US_C1_MASK, VPE_US_C1_SHIFT);
647 *us2_reg++ = *us1_reg;
648 *us3_reg++ = *us1_reg++;
649 }
650 ctx->load_mmrs = true;
651 }
652
653 /*
654 * Set the upsampler config mode and the VPDMA line mode in the shadow MMRs.
655 */
656 static void set_cfg_and_line_modes(struct vpe_ctx *ctx)
657 {
658 struct vpe_fmt *fmt = ctx->q_data[Q_DATA_SRC].fmt;
659 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
660 u32 *us1_reg0 = &mmr_adb->us1_regs[0];
661 u32 *us2_reg0 = &mmr_adb->us2_regs[0];
662 u32 *us3_reg0 = &mmr_adb->us3_regs[0];
663 int line_mode = 1;
664 int cfg_mode = 1;
665
666 /*
667 * Cfg Mode 0: YUV420 source, enable upsampler, DEI is de-interlacing.
668 * Cfg Mode 1: YUV422 source, disable upsampler, DEI is de-interlacing.
669 */
670
671 if (fmt->fourcc == V4L2_PIX_FMT_NV12) {
672 cfg_mode = 0;
673 line_mode = 0; /* double lines to line buffer */
674 }
675
676 write_field(us1_reg0, cfg_mode, VPE_US_MODE_MASK, VPE_US_MODE_SHIFT);
677 write_field(us2_reg0, cfg_mode, VPE_US_MODE_MASK, VPE_US_MODE_SHIFT);
678 write_field(us3_reg0, cfg_mode, VPE_US_MODE_MASK, VPE_US_MODE_SHIFT);
679
680 /* regs for now */
681 vpdma_set_line_mode(ctx->dev->vpdma, line_mode, VPE_CHAN_CHROMA1_IN);
682 vpdma_set_line_mode(ctx->dev->vpdma, line_mode, VPE_CHAN_CHROMA2_IN);
683 vpdma_set_line_mode(ctx->dev->vpdma, line_mode, VPE_CHAN_CHROMA3_IN);
684
685 /* frame start for input luma */
686 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
687 VPE_CHAN_LUMA1_IN);
688 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
689 VPE_CHAN_LUMA2_IN);
690 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
691 VPE_CHAN_LUMA3_IN);
692
693 /* frame start for input chroma */
694 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
695 VPE_CHAN_CHROMA1_IN);
696 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
697 VPE_CHAN_CHROMA2_IN);
698 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
699 VPE_CHAN_CHROMA3_IN);
700
701 /* frame start for MV in client */
702 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
703 VPE_CHAN_MV_IN);
704
705 ctx->load_mmrs = true;
706 }
707
708 /*
709 * Set the shadow registers that are modified when the source
710 * format changes.
711 */
712 static void set_src_registers(struct vpe_ctx *ctx)
713 {
714 set_us_coefficients(ctx);
715 }
716
717 /*
718 * Set the shadow registers that are modified when the destination
719 * format changes.
720 */
721 static void set_dst_registers(struct vpe_ctx *ctx)
722 {
723 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
724 enum v4l2_colorspace clrspc = ctx->q_data[Q_DATA_DST].colorspace;
725 struct vpe_fmt *fmt = ctx->q_data[Q_DATA_DST].fmt;
726 u32 val = 0;
727
728 if (clrspc == V4L2_COLORSPACE_SRGB)
729 val |= VPE_RGB_OUT_SELECT;
730 else if (fmt->fourcc == V4L2_PIX_FMT_NV16)
731 val |= VPE_COLOR_SEPARATE_422;
732
733 /*
734 * the source of CHR_DS and CSC is always the scaler, irrespective of
735 * whether it's used or not
736 */
737 val |= VPE_DS_SRC_DEI_SCALER | VPE_CSC_SRC_DEI_SCALER;
738
739 if (fmt->fourcc != V4L2_PIX_FMT_NV12)
740 val |= VPE_DS_BYPASS;
741
742 mmr_adb->out_fmt_reg[0] = val;
743
744 ctx->load_mmrs = true;
745 }
746
747 /*
748 * Set the de-interlacer shadow register values
749 */
750 static void set_dei_regs(struct vpe_ctx *ctx)
751 {
752 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
753 struct vpe_q_data *s_q_data = &ctx->q_data[Q_DATA_SRC];
754 unsigned int src_h = s_q_data->c_rect.height;
755 unsigned int src_w = s_q_data->c_rect.width;
756 u32 *dei_mmr0 = &mmr_adb->dei_regs[0];
757 bool deinterlace = true;
758 u32 val = 0;
759
760 /*
761 * according to TRM, we should set DEI in progressive bypass mode when
762 * the input content is progressive, however, DEI is bypassed correctly
763 * for both progressive and interlace content in interlace bypass mode.
764 * It has been recommended not to use progressive bypass mode.
765 */
766 if ((!ctx->deinterlacing && (s_q_data->flags & Q_DATA_INTERLACED)) ||
767 !(s_q_data->flags & Q_DATA_INTERLACED)) {
768 deinterlace = false;
769 val = VPE_DEI_INTERLACE_BYPASS;
770 }
771
772 src_h = deinterlace ? src_h * 2 : src_h;
773
774 val |= (src_h << VPE_DEI_HEIGHT_SHIFT) |
775 (src_w << VPE_DEI_WIDTH_SHIFT) |
776 VPE_DEI_FIELD_FLUSH;
777
778 *dei_mmr0 = val;
779
780 ctx->load_mmrs = true;
781 }
782
783 static void set_dei_shadow_registers(struct vpe_ctx *ctx)
784 {
785 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
786 u32 *dei_mmr = &mmr_adb->dei_regs[0];
787 const struct vpe_dei_regs *cur = &dei_regs;
788
789 dei_mmr[2] = cur->mdt_spacial_freq_thr_reg;
790 dei_mmr[3] = cur->edi_config_reg;
791 dei_mmr[4] = cur->edi_lut_reg0;
792 dei_mmr[5] = cur->edi_lut_reg1;
793 dei_mmr[6] = cur->edi_lut_reg2;
794 dei_mmr[7] = cur->edi_lut_reg3;
795
796 ctx->load_mmrs = true;
797 }
798
799 /*
800 * Set the shadow registers whose values are modified when either the
801 * source or destination format is changed.
802 */
803 static int set_srcdst_params(struct vpe_ctx *ctx)
804 {
805 struct vpe_q_data *s_q_data = &ctx->q_data[Q_DATA_SRC];
806 struct vpe_q_data *d_q_data = &ctx->q_data[Q_DATA_DST];
807 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
808 unsigned int src_w = s_q_data->c_rect.width;
809 unsigned int src_h = s_q_data->c_rect.height;
810 unsigned int dst_w = d_q_data->c_rect.width;
811 unsigned int dst_h = d_q_data->c_rect.height;
812 size_t mv_buf_size;
813 int ret;
814
815 ctx->sequence = 0;
816 ctx->field = V4L2_FIELD_TOP;
817
818 if ((s_q_data->flags & Q_DATA_INTERLACED) &&
819 !(d_q_data->flags & Q_DATA_INTERLACED)) {
820 int bytes_per_line;
821 const struct vpdma_data_format *mv =
822 &vpdma_misc_fmts[VPDMA_DATA_FMT_MV];
823
824 /*
825 * we make sure that the source image has a 16 byte aligned
826 * stride, we need to do the same for the motion vector buffer
827 * by aligning it's stride to the next 16 byte boundry. this
828 * extra space will not be used by the de-interlacer, but will
829 * ensure that vpdma operates correctly
830 */
831 bytes_per_line = ALIGN((s_q_data->width * mv->depth) >> 3,
832 VPDMA_STRIDE_ALIGN);
833 mv_buf_size = bytes_per_line * s_q_data->height;
834
835 ctx->deinterlacing = 1;
836 src_h <<= 1;
837 } else {
838 ctx->deinterlacing = 0;
839 mv_buf_size = 0;
840 }
841
842 free_vbs(ctx);
843
844 ret = realloc_mv_buffers(ctx, mv_buf_size);
845 if (ret)
846 return ret;
847
848 set_cfg_and_line_modes(ctx);
849 set_dei_regs(ctx);
850
851 csc_set_coeff(ctx->dev->csc, &mmr_adb->csc_regs[0],
852 s_q_data->colorspace, d_q_data->colorspace);
853
854 sc_set_hs_coeffs(ctx->dev->sc, ctx->sc_coeff_h.addr, src_w, dst_w);
855 sc_set_vs_coeffs(ctx->dev->sc, ctx->sc_coeff_v.addr, src_h, dst_h);
856
857 sc_config_scaler(ctx->dev->sc, &mmr_adb->sc_regs0[0],
858 &mmr_adb->sc_regs8[0], &mmr_adb->sc_regs17[0],
859 src_w, src_h, dst_w, dst_h);
860
861 return 0;
862 }
863
864 /*
865 * Return the vpe_ctx structure for a given struct file
866 */
867 static struct vpe_ctx *file2ctx(struct file *file)
868 {
869 return container_of(file->private_data, struct vpe_ctx, fh);
870 }
871
872 /*
873 * mem2mem callbacks
874 */
875
876 /**
877 * job_ready() - check whether an instance is ready to be scheduled to run
878 */
879 static int job_ready(void *priv)
880 {
881 struct vpe_ctx *ctx = priv;
882 int needed = ctx->bufs_per_job;
883
884 if (ctx->deinterlacing && ctx->src_vbs[2] == NULL)
885 needed += 2; /* need additional two most recent fields */
886
887 if (v4l2_m2m_num_src_bufs_ready(ctx->m2m_ctx) < needed)
888 return 0;
889
890 return 1;
891 }
892
893 static void job_abort(void *priv)
894 {
895 struct vpe_ctx *ctx = priv;
896
897 /* Will cancel the transaction in the next interrupt handler */
898 ctx->aborting = 1;
899 }
900
901 /*
902 * Lock access to the device
903 */
904 static void vpe_lock(void *priv)
905 {
906 struct vpe_ctx *ctx = priv;
907 struct vpe_dev *dev = ctx->dev;
908 mutex_lock(&dev->dev_mutex);
909 }
910
911 static void vpe_unlock(void *priv)
912 {
913 struct vpe_ctx *ctx = priv;
914 struct vpe_dev *dev = ctx->dev;
915 mutex_unlock(&dev->dev_mutex);
916 }
917
918 static void vpe_dump_regs(struct vpe_dev *dev)
919 {
920 #define DUMPREG(r) vpe_dbg(dev, "%-35s %08x\n", #r, read_reg(dev, VPE_##r))
921
922 vpe_dbg(dev, "VPE Registers:\n");
923
924 DUMPREG(PID);
925 DUMPREG(SYSCONFIG);
926 DUMPREG(INT0_STATUS0_RAW);
927 DUMPREG(INT0_STATUS0);
928 DUMPREG(INT0_ENABLE0);
929 DUMPREG(INT0_STATUS1_RAW);
930 DUMPREG(INT0_STATUS1);
931 DUMPREG(INT0_ENABLE1);
932 DUMPREG(CLK_ENABLE);
933 DUMPREG(CLK_RESET);
934 DUMPREG(CLK_FORMAT_SELECT);
935 DUMPREG(CLK_RANGE_MAP);
936 DUMPREG(US1_R0);
937 DUMPREG(US1_R1);
938 DUMPREG(US1_R2);
939 DUMPREG(US1_R3);
940 DUMPREG(US1_R4);
941 DUMPREG(US1_R5);
942 DUMPREG(US1_R6);
943 DUMPREG(US1_R7);
944 DUMPREG(US2_R0);
945 DUMPREG(US2_R1);
946 DUMPREG(US2_R2);
947 DUMPREG(US2_R3);
948 DUMPREG(US2_R4);
949 DUMPREG(US2_R5);
950 DUMPREG(US2_R6);
951 DUMPREG(US2_R7);
952 DUMPREG(US3_R0);
953 DUMPREG(US3_R1);
954 DUMPREG(US3_R2);
955 DUMPREG(US3_R3);
956 DUMPREG(US3_R4);
957 DUMPREG(US3_R5);
958 DUMPREG(US3_R6);
959 DUMPREG(US3_R7);
960 DUMPREG(DEI_FRAME_SIZE);
961 DUMPREG(MDT_BYPASS);
962 DUMPREG(MDT_SF_THRESHOLD);
963 DUMPREG(EDI_CONFIG);
964 DUMPREG(DEI_EDI_LUT_R0);
965 DUMPREG(DEI_EDI_LUT_R1);
966 DUMPREG(DEI_EDI_LUT_R2);
967 DUMPREG(DEI_EDI_LUT_R3);
968 DUMPREG(DEI_FMD_WINDOW_R0);
969 DUMPREG(DEI_FMD_WINDOW_R1);
970 DUMPREG(DEI_FMD_CONTROL_R0);
971 DUMPREG(DEI_FMD_CONTROL_R1);
972 DUMPREG(DEI_FMD_STATUS_R0);
973 DUMPREG(DEI_FMD_STATUS_R1);
974 DUMPREG(DEI_FMD_STATUS_R2);
975 #undef DUMPREG
976
977 sc_dump_regs(dev->sc);
978 csc_dump_regs(dev->csc);
979 }
980
981 static void add_out_dtd(struct vpe_ctx *ctx, int port)
982 {
983 struct vpe_q_data *q_data = &ctx->q_data[Q_DATA_DST];
984 const struct vpe_port_data *p_data = &port_data[port];
985 struct vb2_buffer *vb = ctx->dst_vb;
986 struct v4l2_rect *c_rect = &q_data->c_rect;
987 struct vpe_fmt *fmt = q_data->fmt;
988 const struct vpdma_data_format *vpdma_fmt;
989 int mv_buf_selector = !ctx->src_mv_buf_selector;
990 dma_addr_t dma_addr;
991 u32 flags = 0;
992
993 if (port == VPE_PORT_MV_OUT) {
994 vpdma_fmt = &vpdma_misc_fmts[VPDMA_DATA_FMT_MV];
995 dma_addr = ctx->mv_buf_dma[mv_buf_selector];
996 } else {
997 /* to incorporate interleaved formats */
998 int plane = fmt->coplanar ? p_data->vb_part : 0;
999
1000 vpdma_fmt = fmt->vpdma_fmt[plane];
1001 dma_addr = vb2_dma_contig_plane_dma_addr(vb, plane);
1002 if (!dma_addr) {
1003 vpe_err(ctx->dev,
1004 "acquiring output buffer(%d) dma_addr failed\n",
1005 port);
1006 return;
1007 }
1008 }
1009
1010 if (q_data->flags & Q_DATA_FRAME_1D)
1011 flags |= VPDMA_DATA_FRAME_1D;
1012 if (q_data->flags & Q_DATA_MODE_TILED)
1013 flags |= VPDMA_DATA_MODE_TILED;
1014
1015 vpdma_add_out_dtd(&ctx->desc_list, c_rect, vpdma_fmt, dma_addr,
1016 p_data->channel, flags);
1017 }
1018
1019 static void add_in_dtd(struct vpe_ctx *ctx, int port)
1020 {
1021 struct vpe_q_data *q_data = &ctx->q_data[Q_DATA_SRC];
1022 const struct vpe_port_data *p_data = &port_data[port];
1023 struct vb2_buffer *vb = ctx->src_vbs[p_data->vb_index];
1024 struct v4l2_rect *c_rect = &q_data->c_rect;
1025 struct vpe_fmt *fmt = q_data->fmt;
1026 const struct vpdma_data_format *vpdma_fmt;
1027 int mv_buf_selector = ctx->src_mv_buf_selector;
1028 int field = vb->v4l2_buf.field == V4L2_FIELD_BOTTOM;
1029 dma_addr_t dma_addr;
1030 u32 flags = 0;
1031
1032 if (port == VPE_PORT_MV_IN) {
1033 vpdma_fmt = &vpdma_misc_fmts[VPDMA_DATA_FMT_MV];
1034 dma_addr = ctx->mv_buf_dma[mv_buf_selector];
1035 } else {
1036 /* to incorporate interleaved formats */
1037 int plane = fmt->coplanar ? p_data->vb_part : 0;
1038
1039 vpdma_fmt = fmt->vpdma_fmt[plane];
1040
1041 dma_addr = vb2_dma_contig_plane_dma_addr(vb, plane);
1042 if (!dma_addr) {
1043 vpe_err(ctx->dev,
1044 "acquiring input buffer(%d) dma_addr failed\n",
1045 port);
1046 return;
1047 }
1048 }
1049
1050 if (q_data->flags & Q_DATA_FRAME_1D)
1051 flags |= VPDMA_DATA_FRAME_1D;
1052 if (q_data->flags & Q_DATA_MODE_TILED)
1053 flags |= VPDMA_DATA_MODE_TILED;
1054
1055 vpdma_add_in_dtd(&ctx->desc_list, q_data->width, q_data->height,
1056 c_rect, vpdma_fmt, dma_addr, p_data->channel, field, flags);
1057 }
1058
1059 /*
1060 * Enable the expected IRQ sources
1061 */
1062 static void enable_irqs(struct vpe_ctx *ctx)
1063 {
1064 write_reg(ctx->dev, VPE_INT0_ENABLE0_SET, VPE_INT0_LIST0_COMPLETE);
1065 write_reg(ctx->dev, VPE_INT0_ENABLE1_SET, VPE_DEI_ERROR_INT |
1066 VPE_DS1_UV_ERROR_INT);
1067
1068 vpdma_enable_list_complete_irq(ctx->dev->vpdma, 0, true);
1069 }
1070
1071 static void disable_irqs(struct vpe_ctx *ctx)
1072 {
1073 write_reg(ctx->dev, VPE_INT0_ENABLE0_CLR, 0xffffffff);
1074 write_reg(ctx->dev, VPE_INT0_ENABLE1_CLR, 0xffffffff);
1075
1076 vpdma_enable_list_complete_irq(ctx->dev->vpdma, 0, false);
1077 }
1078
1079 /* device_run() - prepares and starts the device
1080 *
1081 * This function is only called when both the source and destination
1082 * buffers are in place.
1083 */
1084 static void device_run(void *priv)
1085 {
1086 struct vpe_ctx *ctx = priv;
1087 struct sc_data *sc = ctx->dev->sc;
1088 struct vpe_q_data *d_q_data = &ctx->q_data[Q_DATA_DST];
1089
1090 if (ctx->deinterlacing && ctx->src_vbs[2] == NULL) {
1091 ctx->src_vbs[2] = v4l2_m2m_src_buf_remove(ctx->m2m_ctx);
1092 WARN_ON(ctx->src_vbs[2] == NULL);
1093 ctx->src_vbs[1] = v4l2_m2m_src_buf_remove(ctx->m2m_ctx);
1094 WARN_ON(ctx->src_vbs[1] == NULL);
1095 }
1096
1097 ctx->src_vbs[0] = v4l2_m2m_src_buf_remove(ctx->m2m_ctx);
1098 WARN_ON(ctx->src_vbs[0] == NULL);
1099 ctx->dst_vb = v4l2_m2m_dst_buf_remove(ctx->m2m_ctx);
1100 WARN_ON(ctx->dst_vb == NULL);
1101
1102 /* config descriptors */
1103 if (ctx->dev->loaded_mmrs != ctx->mmr_adb.dma_addr || ctx->load_mmrs) {
1104 vpdma_map_desc_buf(ctx->dev->vpdma, &ctx->mmr_adb);
1105 vpdma_add_cfd_adb(&ctx->desc_list, CFD_MMR_CLIENT, &ctx->mmr_adb);
1106 ctx->dev->loaded_mmrs = ctx->mmr_adb.dma_addr;
1107 ctx->load_mmrs = false;
1108 }
1109
1110 if (sc->loaded_coeff_h != ctx->sc_coeff_h.dma_addr ||
1111 sc->load_coeff_h) {
1112 vpdma_map_desc_buf(ctx->dev->vpdma, &ctx->sc_coeff_h);
1113 vpdma_add_cfd_block(&ctx->desc_list, CFD_SC_CLIENT,
1114 &ctx->sc_coeff_h, 0);
1115
1116 sc->loaded_coeff_h = ctx->sc_coeff_h.dma_addr;
1117 sc->load_coeff_h = false;
1118 }
1119
1120 if (sc->loaded_coeff_v != ctx->sc_coeff_v.dma_addr ||
1121 sc->load_coeff_v) {
1122 vpdma_map_desc_buf(ctx->dev->vpdma, &ctx->sc_coeff_v);
1123 vpdma_add_cfd_block(&ctx->desc_list, CFD_SC_CLIENT,
1124 &ctx->sc_coeff_v, SC_COEF_SRAM_SIZE >> 4);
1125
1126 sc->loaded_coeff_v = ctx->sc_coeff_v.dma_addr;
1127 sc->load_coeff_v = false;
1128 }
1129
1130 /* output data descriptors */
1131 if (ctx->deinterlacing)
1132 add_out_dtd(ctx, VPE_PORT_MV_OUT);
1133
1134 if (d_q_data->colorspace == V4L2_COLORSPACE_SRGB) {
1135 add_out_dtd(ctx, VPE_PORT_RGB_OUT);
1136 } else {
1137 add_out_dtd(ctx, VPE_PORT_LUMA_OUT);
1138 if (d_q_data->fmt->coplanar)
1139 add_out_dtd(ctx, VPE_PORT_CHROMA_OUT);
1140 }
1141
1142 /* input data descriptors */
1143 if (ctx->deinterlacing) {
1144 add_in_dtd(ctx, VPE_PORT_LUMA3_IN);
1145 add_in_dtd(ctx, VPE_PORT_CHROMA3_IN);
1146
1147 add_in_dtd(ctx, VPE_PORT_LUMA2_IN);
1148 add_in_dtd(ctx, VPE_PORT_CHROMA2_IN);
1149 }
1150
1151 add_in_dtd(ctx, VPE_PORT_LUMA1_IN);
1152 add_in_dtd(ctx, VPE_PORT_CHROMA1_IN);
1153
1154 if (ctx->deinterlacing)
1155 add_in_dtd(ctx, VPE_PORT_MV_IN);
1156
1157 /* sync on channel control descriptors for input ports */
1158 vpdma_add_sync_on_channel_ctd(&ctx->desc_list, VPE_CHAN_LUMA1_IN);
1159 vpdma_add_sync_on_channel_ctd(&ctx->desc_list, VPE_CHAN_CHROMA1_IN);
1160
1161 if (ctx->deinterlacing) {
1162 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1163 VPE_CHAN_LUMA2_IN);
1164 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1165 VPE_CHAN_CHROMA2_IN);
1166
1167 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1168 VPE_CHAN_LUMA3_IN);
1169 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1170 VPE_CHAN_CHROMA3_IN);
1171
1172 vpdma_add_sync_on_channel_ctd(&ctx->desc_list, VPE_CHAN_MV_IN);
1173 }
1174
1175 /* sync on channel control descriptors for output ports */
1176 if (d_q_data->colorspace == V4L2_COLORSPACE_SRGB) {
1177 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1178 VPE_CHAN_RGB_OUT);
1179 } else {
1180 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1181 VPE_CHAN_LUMA_OUT);
1182 if (d_q_data->fmt->coplanar)
1183 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1184 VPE_CHAN_CHROMA_OUT);
1185 }
1186
1187 if (ctx->deinterlacing)
1188 vpdma_add_sync_on_channel_ctd(&ctx->desc_list, VPE_CHAN_MV_OUT);
1189
1190 enable_irqs(ctx);
1191
1192 vpdma_map_desc_buf(ctx->dev->vpdma, &ctx->desc_list.buf);
1193 vpdma_submit_descs(ctx->dev->vpdma, &ctx->desc_list);
1194 }
1195
1196 static void dei_error(struct vpe_ctx *ctx)
1197 {
1198 dev_warn(ctx->dev->v4l2_dev.dev,
1199 "received DEI error interrupt\n");
1200 }
1201
1202 static void ds1_uv_error(struct vpe_ctx *ctx)
1203 {
1204 dev_warn(ctx->dev->v4l2_dev.dev,
1205 "received downsampler error interrupt\n");
1206 }
1207
1208 static irqreturn_t vpe_irq(int irq_vpe, void *data)
1209 {
1210 struct vpe_dev *dev = (struct vpe_dev *)data;
1211 struct vpe_ctx *ctx;
1212 struct vpe_q_data *d_q_data;
1213 struct vb2_buffer *s_vb, *d_vb;
1214 struct v4l2_buffer *s_buf, *d_buf;
1215 unsigned long flags;
1216 u32 irqst0, irqst1;
1217
1218 irqst0 = read_reg(dev, VPE_INT0_STATUS0);
1219 if (irqst0) {
1220 write_reg(dev, VPE_INT0_STATUS0_CLR, irqst0);
1221 vpe_dbg(dev, "INT0_STATUS0 = 0x%08x\n", irqst0);
1222 }
1223
1224 irqst1 = read_reg(dev, VPE_INT0_STATUS1);
1225 if (irqst1) {
1226 write_reg(dev, VPE_INT0_STATUS1_CLR, irqst1);
1227 vpe_dbg(dev, "INT0_STATUS1 = 0x%08x\n", irqst1);
1228 }
1229
1230 ctx = v4l2_m2m_get_curr_priv(dev->m2m_dev);
1231 if (!ctx) {
1232 vpe_err(dev, "instance released before end of transaction\n");
1233 goto handled;
1234 }
1235
1236 if (irqst1) {
1237 if (irqst1 & VPE_DEI_ERROR_INT) {
1238 irqst1 &= ~VPE_DEI_ERROR_INT;
1239 dei_error(ctx);
1240 }
1241 if (irqst1 & VPE_DS1_UV_ERROR_INT) {
1242 irqst1 &= ~VPE_DS1_UV_ERROR_INT;
1243 ds1_uv_error(ctx);
1244 }
1245 }
1246
1247 if (irqst0) {
1248 if (irqst0 & VPE_INT0_LIST0_COMPLETE)
1249 vpdma_clear_list_stat(ctx->dev->vpdma);
1250
1251 irqst0 &= ~(VPE_INT0_LIST0_COMPLETE);
1252 }
1253
1254 if (irqst0 | irqst1) {
1255 dev_warn(dev->v4l2_dev.dev, "Unexpected interrupt: "
1256 "INT0_STATUS0 = 0x%08x, INT0_STATUS1 = 0x%08x\n",
1257 irqst0, irqst1);
1258 }
1259
1260 disable_irqs(ctx);
1261
1262 vpdma_unmap_desc_buf(dev->vpdma, &ctx->desc_list.buf);
1263 vpdma_unmap_desc_buf(dev->vpdma, &ctx->mmr_adb);
1264 vpdma_unmap_desc_buf(dev->vpdma, &ctx->sc_coeff_h);
1265 vpdma_unmap_desc_buf(dev->vpdma, &ctx->sc_coeff_v);
1266
1267 vpdma_reset_desc_list(&ctx->desc_list);
1268
1269 /* the previous dst mv buffer becomes the next src mv buffer */
1270 ctx->src_mv_buf_selector = !ctx->src_mv_buf_selector;
1271
1272 if (ctx->aborting)
1273 goto finished;
1274
1275 s_vb = ctx->src_vbs[0];
1276 d_vb = ctx->dst_vb;
1277 s_buf = &s_vb->v4l2_buf;
1278 d_buf = &d_vb->v4l2_buf;
1279
1280 d_buf->timestamp = s_buf->timestamp;
1281 if (s_buf->flags & V4L2_BUF_FLAG_TIMECODE) {
1282 d_buf->flags |= V4L2_BUF_FLAG_TIMECODE;
1283 d_buf->timecode = s_buf->timecode;
1284 }
1285 d_buf->sequence = ctx->sequence;
1286 d_buf->field = ctx->field;
1287
1288 d_q_data = &ctx->q_data[Q_DATA_DST];
1289 if (d_q_data->flags & Q_DATA_INTERLACED) {
1290 if (ctx->field == V4L2_FIELD_BOTTOM) {
1291 ctx->sequence++;
1292 ctx->field = V4L2_FIELD_TOP;
1293 } else {
1294 WARN_ON(ctx->field != V4L2_FIELD_TOP);
1295 ctx->field = V4L2_FIELD_BOTTOM;
1296 }
1297 } else {
1298 ctx->sequence++;
1299 }
1300
1301 if (ctx->deinterlacing)
1302 s_vb = ctx->src_vbs[2];
1303
1304 spin_lock_irqsave(&dev->lock, flags);
1305 v4l2_m2m_buf_done(s_vb, VB2_BUF_STATE_DONE);
1306 v4l2_m2m_buf_done(d_vb, VB2_BUF_STATE_DONE);
1307 spin_unlock_irqrestore(&dev->lock, flags);
1308
1309 if (ctx->deinterlacing) {
1310 ctx->src_vbs[2] = ctx->src_vbs[1];
1311 ctx->src_vbs[1] = ctx->src_vbs[0];
1312 }
1313
1314 ctx->bufs_completed++;
1315 if (ctx->bufs_completed < ctx->bufs_per_job) {
1316 device_run(ctx);
1317 goto handled;
1318 }
1319
1320 finished:
1321 vpe_dbg(ctx->dev, "finishing transaction\n");
1322 ctx->bufs_completed = 0;
1323 v4l2_m2m_job_finish(dev->m2m_dev, ctx->m2m_ctx);
1324 handled:
1325 return IRQ_HANDLED;
1326 }
1327
1328 /*
1329 * video ioctls
1330 */
1331 static int vpe_querycap(struct file *file, void *priv,
1332 struct v4l2_capability *cap)
1333 {
1334 strncpy(cap->driver, VPE_MODULE_NAME, sizeof(cap->driver) - 1);
1335 strncpy(cap->card, VPE_MODULE_NAME, sizeof(cap->card) - 1);
1336 strlcpy(cap->bus_info, VPE_MODULE_NAME, sizeof(cap->bus_info));
1337 cap->device_caps = V4L2_CAP_VIDEO_M2M | V4L2_CAP_STREAMING;
1338 cap->capabilities = cap->device_caps | V4L2_CAP_DEVICE_CAPS;
1339 return 0;
1340 }
1341
1342 static int __enum_fmt(struct v4l2_fmtdesc *f, u32 type)
1343 {
1344 int i, index;
1345 struct vpe_fmt *fmt = NULL;
1346
1347 index = 0;
1348 for (i = 0; i < ARRAY_SIZE(vpe_formats); ++i) {
1349 if (vpe_formats[i].types & type) {
1350 if (index == f->index) {
1351 fmt = &vpe_formats[i];
1352 break;
1353 }
1354 index++;
1355 }
1356 }
1357
1358 if (!fmt)
1359 return -EINVAL;
1360
1361 strncpy(f->description, fmt->name, sizeof(f->description) - 1);
1362 f->pixelformat = fmt->fourcc;
1363 return 0;
1364 }
1365
1366 static int vpe_enum_fmt(struct file *file, void *priv,
1367 struct v4l2_fmtdesc *f)
1368 {
1369 if (V4L2_TYPE_IS_OUTPUT(f->type))
1370 return __enum_fmt(f, VPE_FMT_TYPE_OUTPUT);
1371
1372 return __enum_fmt(f, VPE_FMT_TYPE_CAPTURE);
1373 }
1374
1375 static int vpe_g_fmt(struct file *file, void *priv, struct v4l2_format *f)
1376 {
1377 struct v4l2_pix_format_mplane *pix = &f->fmt.pix_mp;
1378 struct vpe_ctx *ctx = file2ctx(file);
1379 struct vb2_queue *vq;
1380 struct vpe_q_data *q_data;
1381 int i;
1382
1383 vq = v4l2_m2m_get_vq(ctx->m2m_ctx, f->type);
1384 if (!vq)
1385 return -EINVAL;
1386
1387 q_data = get_q_data(ctx, f->type);
1388
1389 pix->width = q_data->width;
1390 pix->height = q_data->height;
1391 pix->pixelformat = q_data->fmt->fourcc;
1392 pix->field = q_data->field;
1393
1394 if (V4L2_TYPE_IS_OUTPUT(f->type)) {
1395 pix->colorspace = q_data->colorspace;
1396 } else {
1397 struct vpe_q_data *s_q_data;
1398
1399 /* get colorspace from the source queue */
1400 s_q_data = get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE);
1401
1402 pix->colorspace = s_q_data->colorspace;
1403 }
1404
1405 pix->num_planes = q_data->fmt->coplanar ? 2 : 1;
1406
1407 for (i = 0; i < pix->num_planes; i++) {
1408 pix->plane_fmt[i].bytesperline = q_data->bytesperline[i];
1409 pix->plane_fmt[i].sizeimage = q_data->sizeimage[i];
1410 }
1411
1412 return 0;
1413 }
1414
1415 static int __vpe_try_fmt(struct vpe_ctx *ctx, struct v4l2_format *f,
1416 struct vpe_fmt *fmt, int type)
1417 {
1418 struct v4l2_pix_format_mplane *pix = &f->fmt.pix_mp;
1419 struct v4l2_plane_pix_format *plane_fmt;
1420 unsigned int w_align;
1421 int i, depth, depth_bytes;
1422
1423 if (!fmt || !(fmt->types & type)) {
1424 vpe_err(ctx->dev, "Fourcc format (0x%08x) invalid.\n",
1425 pix->pixelformat);
1426 return -EINVAL;
1427 }
1428
1429 if (pix->field != V4L2_FIELD_NONE && pix->field != V4L2_FIELD_ALTERNATE)
1430 pix->field = V4L2_FIELD_NONE;
1431
1432 depth = fmt->vpdma_fmt[VPE_LUMA]->depth;
1433
1434 /*
1435 * the line stride should 16 byte aligned for VPDMA to work, based on
1436 * the bytes per pixel, figure out how much the width should be aligned
1437 * to make sure line stride is 16 byte aligned
1438 */
1439 depth_bytes = depth >> 3;
1440
1441 if (depth_bytes == 3)
1442 /*
1443 * if bpp is 3(as in some RGB formats), the pixel width doesn't
1444 * really help in ensuring line stride is 16 byte aligned
1445 */
1446 w_align = 4;
1447 else
1448 /*
1449 * for the remainder bpp(4, 2 and 1), the pixel width alignment
1450 * can ensure a line stride alignment of 16 bytes. For example,
1451 * if bpp is 2, then the line stride can be 16 byte aligned if
1452 * the width is 8 byte aligned
1453 */
1454 w_align = order_base_2(VPDMA_DESC_ALIGN / depth_bytes);
1455
1456 v4l_bound_align_image(&pix->width, MIN_W, MAX_W, w_align,
1457 &pix->height, MIN_H, MAX_H, H_ALIGN,
1458 S_ALIGN);
1459
1460 pix->num_planes = fmt->coplanar ? 2 : 1;
1461 pix->pixelformat = fmt->fourcc;
1462
1463 if (!pix->colorspace) {
1464 if (fmt->fourcc == V4L2_PIX_FMT_RGB24 ||
1465 fmt->fourcc == V4L2_PIX_FMT_BGR24 ||
1466 fmt->fourcc == V4L2_PIX_FMT_RGB32 ||
1467 fmt->fourcc == V4L2_PIX_FMT_BGR32) {
1468 pix->colorspace = V4L2_COLORSPACE_SRGB;
1469 } else {
1470 if (pix->height > 1280) /* HD */
1471 pix->colorspace = V4L2_COLORSPACE_REC709;
1472 else /* SD */
1473 pix->colorspace = V4L2_COLORSPACE_SMPTE170M;
1474 }
1475 }
1476
1477 for (i = 0; i < pix->num_planes; i++) {
1478 plane_fmt = &pix->plane_fmt[i];
1479 depth = fmt->vpdma_fmt[i]->depth;
1480
1481 if (i == VPE_LUMA)
1482 plane_fmt->bytesperline = (pix->width * depth) >> 3;
1483 else
1484 plane_fmt->bytesperline = pix->width;
1485
1486 plane_fmt->sizeimage =
1487 (pix->height * pix->width * depth) >> 3;
1488 }
1489
1490 return 0;
1491 }
1492
1493 static int vpe_try_fmt(struct file *file, void *priv, struct v4l2_format *f)
1494 {
1495 struct vpe_ctx *ctx = file2ctx(file);
1496 struct vpe_fmt *fmt = find_format(f);
1497
1498 if (V4L2_TYPE_IS_OUTPUT(f->type))
1499 return __vpe_try_fmt(ctx, f, fmt, VPE_FMT_TYPE_OUTPUT);
1500 else
1501 return __vpe_try_fmt(ctx, f, fmt, VPE_FMT_TYPE_CAPTURE);
1502 }
1503
1504 static int __vpe_s_fmt(struct vpe_ctx *ctx, struct v4l2_format *f)
1505 {
1506 struct v4l2_pix_format_mplane *pix = &f->fmt.pix_mp;
1507 struct v4l2_plane_pix_format *plane_fmt;
1508 struct vpe_q_data *q_data;
1509 struct vb2_queue *vq;
1510 int i;
1511
1512 vq = v4l2_m2m_get_vq(ctx->m2m_ctx, f->type);
1513 if (!vq)
1514 return -EINVAL;
1515
1516 if (vb2_is_busy(vq)) {
1517 vpe_err(ctx->dev, "queue busy\n");
1518 return -EBUSY;
1519 }
1520
1521 q_data = get_q_data(ctx, f->type);
1522 if (!q_data)
1523 return -EINVAL;
1524
1525 q_data->fmt = find_format(f);
1526 q_data->width = pix->width;
1527 q_data->height = pix->height;
1528 q_data->colorspace = pix->colorspace;
1529 q_data->field = pix->field;
1530
1531 for (i = 0; i < pix->num_planes; i++) {
1532 plane_fmt = &pix->plane_fmt[i];
1533
1534 q_data->bytesperline[i] = plane_fmt->bytesperline;
1535 q_data->sizeimage[i] = plane_fmt->sizeimage;
1536 }
1537
1538 q_data->c_rect.left = 0;
1539 q_data->c_rect.top = 0;
1540 q_data->c_rect.width = q_data->width;
1541 q_data->c_rect.height = q_data->height;
1542
1543 if (q_data->field == V4L2_FIELD_ALTERNATE)
1544 q_data->flags |= Q_DATA_INTERLACED;
1545 else
1546 q_data->flags &= ~Q_DATA_INTERLACED;
1547
1548 vpe_dbg(ctx->dev, "Setting format for type %d, wxh: %dx%d, fmt: %d bpl_y %d",
1549 f->type, q_data->width, q_data->height, q_data->fmt->fourcc,
1550 q_data->bytesperline[VPE_LUMA]);
1551 if (q_data->fmt->coplanar)
1552 vpe_dbg(ctx->dev, " bpl_uv %d\n",
1553 q_data->bytesperline[VPE_CHROMA]);
1554
1555 return 0;
1556 }
1557
1558 static int vpe_s_fmt(struct file *file, void *priv, struct v4l2_format *f)
1559 {
1560 int ret;
1561 struct vpe_ctx *ctx = file2ctx(file);
1562
1563 ret = vpe_try_fmt(file, priv, f);
1564 if (ret)
1565 return ret;
1566
1567 ret = __vpe_s_fmt(ctx, f);
1568 if (ret)
1569 return ret;
1570
1571 if (V4L2_TYPE_IS_OUTPUT(f->type))
1572 set_src_registers(ctx);
1573 else
1574 set_dst_registers(ctx);
1575
1576 return set_srcdst_params(ctx);
1577 }
1578
1579 static int vpe_reqbufs(struct file *file, void *priv,
1580 struct v4l2_requestbuffers *reqbufs)
1581 {
1582 struct vpe_ctx *ctx = file2ctx(file);
1583
1584 return v4l2_m2m_reqbufs(file, ctx->m2m_ctx, reqbufs);
1585 }
1586
1587 static int vpe_querybuf(struct file *file, void *priv, struct v4l2_buffer *buf)
1588 {
1589 struct vpe_ctx *ctx = file2ctx(file);
1590
1591 return v4l2_m2m_querybuf(file, ctx->m2m_ctx, buf);
1592 }
1593
1594 static int vpe_qbuf(struct file *file, void *priv, struct v4l2_buffer *buf)
1595 {
1596 struct vpe_ctx *ctx = file2ctx(file);
1597
1598 return v4l2_m2m_qbuf(file, ctx->m2m_ctx, buf);
1599 }
1600
1601 static int vpe_dqbuf(struct file *file, void *priv, struct v4l2_buffer *buf)
1602 {
1603 struct vpe_ctx *ctx = file2ctx(file);
1604
1605 return v4l2_m2m_dqbuf(file, ctx->m2m_ctx, buf);
1606 }
1607
1608 static int vpe_streamon(struct file *file, void *priv, enum v4l2_buf_type type)
1609 {
1610 struct vpe_ctx *ctx = file2ctx(file);
1611
1612 return v4l2_m2m_streamon(file, ctx->m2m_ctx, type);
1613 }
1614
1615 static int vpe_streamoff(struct file *file, void *priv, enum v4l2_buf_type type)
1616 {
1617 struct vpe_ctx *ctx = file2ctx(file);
1618
1619 vpe_dump_regs(ctx->dev);
1620 vpdma_dump_regs(ctx->dev->vpdma);
1621
1622 return v4l2_m2m_streamoff(file, ctx->m2m_ctx, type);
1623 }
1624
1625 /*
1626 * defines number of buffers/frames a context can process with VPE before
1627 * switching to a different context. default value is 1 buffer per context
1628 */
1629 #define V4L2_CID_VPE_BUFS_PER_JOB (V4L2_CID_USER_TI_VPE_BASE + 0)
1630
1631 static int vpe_s_ctrl(struct v4l2_ctrl *ctrl)
1632 {
1633 struct vpe_ctx *ctx =
1634 container_of(ctrl->handler, struct vpe_ctx, hdl);
1635
1636 switch (ctrl->id) {
1637 case V4L2_CID_VPE_BUFS_PER_JOB:
1638 ctx->bufs_per_job = ctrl->val;
1639 break;
1640
1641 default:
1642 vpe_err(ctx->dev, "Invalid control\n");
1643 return -EINVAL;
1644 }
1645
1646 return 0;
1647 }
1648
1649 static const struct v4l2_ctrl_ops vpe_ctrl_ops = {
1650 .s_ctrl = vpe_s_ctrl,
1651 };
1652
1653 static const struct v4l2_ioctl_ops vpe_ioctl_ops = {
1654 .vidioc_querycap = vpe_querycap,
1655
1656 .vidioc_enum_fmt_vid_cap_mplane = vpe_enum_fmt,
1657 .vidioc_g_fmt_vid_cap_mplane = vpe_g_fmt,
1658 .vidioc_try_fmt_vid_cap_mplane = vpe_try_fmt,
1659 .vidioc_s_fmt_vid_cap_mplane = vpe_s_fmt,
1660
1661 .vidioc_enum_fmt_vid_out_mplane = vpe_enum_fmt,
1662 .vidioc_g_fmt_vid_out_mplane = vpe_g_fmt,
1663 .vidioc_try_fmt_vid_out_mplane = vpe_try_fmt,
1664 .vidioc_s_fmt_vid_out_mplane = vpe_s_fmt,
1665
1666 .vidioc_reqbufs = vpe_reqbufs,
1667 .vidioc_querybuf = vpe_querybuf,
1668
1669 .vidioc_qbuf = vpe_qbuf,
1670 .vidioc_dqbuf = vpe_dqbuf,
1671
1672 .vidioc_streamon = vpe_streamon,
1673 .vidioc_streamoff = vpe_streamoff,
1674 .vidioc_subscribe_event = v4l2_ctrl_subscribe_event,
1675 .vidioc_unsubscribe_event = v4l2_event_unsubscribe,
1676 };
1677
1678 /*
1679 * Queue operations
1680 */
1681 static int vpe_queue_setup(struct vb2_queue *vq,
1682 const struct v4l2_format *fmt,
1683 unsigned int *nbuffers, unsigned int *nplanes,
1684 unsigned int sizes[], void *alloc_ctxs[])
1685 {
1686 int i;
1687 struct vpe_ctx *ctx = vb2_get_drv_priv(vq);
1688 struct vpe_q_data *q_data;
1689
1690 q_data = get_q_data(ctx, vq->type);
1691
1692 *nplanes = q_data->fmt->coplanar ? 2 : 1;
1693
1694 for (i = 0; i < *nplanes; i++) {
1695 sizes[i] = q_data->sizeimage[i];
1696 alloc_ctxs[i] = ctx->dev->alloc_ctx;
1697 }
1698
1699 vpe_dbg(ctx->dev, "get %d buffer(s) of size %d", *nbuffers,
1700 sizes[VPE_LUMA]);
1701 if (q_data->fmt->coplanar)
1702 vpe_dbg(ctx->dev, " and %d\n", sizes[VPE_CHROMA]);
1703
1704 return 0;
1705 }
1706
1707 static int vpe_buf_prepare(struct vb2_buffer *vb)
1708 {
1709 struct vpe_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
1710 struct vpe_q_data *q_data;
1711 int i, num_planes;
1712
1713 vpe_dbg(ctx->dev, "type: %d\n", vb->vb2_queue->type);
1714
1715 q_data = get_q_data(ctx, vb->vb2_queue->type);
1716 num_planes = q_data->fmt->coplanar ? 2 : 1;
1717
1718 for (i = 0; i < num_planes; i++) {
1719 if (vb2_plane_size(vb, i) < q_data->sizeimage[i]) {
1720 vpe_err(ctx->dev,
1721 "data will not fit into plane (%lu < %lu)\n",
1722 vb2_plane_size(vb, i),
1723 (long) q_data->sizeimage[i]);
1724 return -EINVAL;
1725 }
1726 }
1727
1728 for (i = 0; i < num_planes; i++)
1729 vb2_set_plane_payload(vb, i, q_data->sizeimage[i]);
1730
1731 return 0;
1732 }
1733
1734 static void vpe_buf_queue(struct vb2_buffer *vb)
1735 {
1736 struct vpe_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
1737 v4l2_m2m_buf_queue(ctx->m2m_ctx, vb);
1738 }
1739
1740 static void vpe_wait_prepare(struct vb2_queue *q)
1741 {
1742 struct vpe_ctx *ctx = vb2_get_drv_priv(q);
1743 vpe_unlock(ctx);
1744 }
1745
1746 static void vpe_wait_finish(struct vb2_queue *q)
1747 {
1748 struct vpe_ctx *ctx = vb2_get_drv_priv(q);
1749 vpe_lock(ctx);
1750 }
1751
1752 static struct vb2_ops vpe_qops = {
1753 .queue_setup = vpe_queue_setup,
1754 .buf_prepare = vpe_buf_prepare,
1755 .buf_queue = vpe_buf_queue,
1756 .wait_prepare = vpe_wait_prepare,
1757 .wait_finish = vpe_wait_finish,
1758 };
1759
1760 static int queue_init(void *priv, struct vb2_queue *src_vq,
1761 struct vb2_queue *dst_vq)
1762 {
1763 struct vpe_ctx *ctx = priv;
1764 int ret;
1765
1766 memset(src_vq, 0, sizeof(*src_vq));
1767 src_vq->type = V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE;
1768 src_vq->io_modes = VB2_MMAP;
1769 src_vq->drv_priv = ctx;
1770 src_vq->buf_struct_size = sizeof(struct v4l2_m2m_buffer);
1771 src_vq->ops = &vpe_qops;
1772 src_vq->mem_ops = &vb2_dma_contig_memops;
1773 src_vq->timestamp_type = V4L2_BUF_FLAG_TIMESTAMP_COPY;
1774
1775 ret = vb2_queue_init(src_vq);
1776 if (ret)
1777 return ret;
1778
1779 memset(dst_vq, 0, sizeof(*dst_vq));
1780 dst_vq->type = V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE;
1781 dst_vq->io_modes = VB2_MMAP;
1782 dst_vq->drv_priv = ctx;
1783 dst_vq->buf_struct_size = sizeof(struct v4l2_m2m_buffer);
1784 dst_vq->ops = &vpe_qops;
1785 dst_vq->mem_ops = &vb2_dma_contig_memops;
1786 dst_vq->timestamp_type = V4L2_BUF_FLAG_TIMESTAMP_COPY;
1787
1788 return vb2_queue_init(dst_vq);
1789 }
1790
1791 static const struct v4l2_ctrl_config vpe_bufs_per_job = {
1792 .ops = &vpe_ctrl_ops,
1793 .id = V4L2_CID_VPE_BUFS_PER_JOB,
1794 .name = "Buffers Per Transaction",
1795 .type = V4L2_CTRL_TYPE_INTEGER,
1796 .def = VPE_DEF_BUFS_PER_JOB,
1797 .min = 1,
1798 .max = VIDEO_MAX_FRAME,
1799 .step = 1,
1800 };
1801
1802 /*
1803 * File operations
1804 */
1805 static int vpe_open(struct file *file)
1806 {
1807 struct vpe_dev *dev = video_drvdata(file);
1808 struct vpe_ctx *ctx = NULL;
1809 struct vpe_q_data *s_q_data;
1810 struct v4l2_ctrl_handler *hdl;
1811 int ret;
1812
1813 vpe_dbg(dev, "vpe_open\n");
1814
1815 if (!dev->vpdma->ready) {
1816 vpe_err(dev, "vpdma firmware not loaded\n");
1817 return -ENODEV;
1818 }
1819
1820 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
1821 if (!ctx)
1822 return -ENOMEM;
1823
1824 ctx->dev = dev;
1825
1826 if (mutex_lock_interruptible(&dev->dev_mutex)) {
1827 ret = -ERESTARTSYS;
1828 goto free_ctx;
1829 }
1830
1831 ret = vpdma_create_desc_list(&ctx->desc_list, VPE_DESC_LIST_SIZE,
1832 VPDMA_LIST_TYPE_NORMAL);
1833 if (ret != 0)
1834 goto unlock;
1835
1836 ret = vpdma_alloc_desc_buf(&ctx->mmr_adb, sizeof(struct vpe_mmr_adb));
1837 if (ret != 0)
1838 goto free_desc_list;
1839
1840 ret = vpdma_alloc_desc_buf(&ctx->sc_coeff_h, SC_COEF_SRAM_SIZE);
1841 if (ret != 0)
1842 goto free_mmr_adb;
1843
1844 ret = vpdma_alloc_desc_buf(&ctx->sc_coeff_v, SC_COEF_SRAM_SIZE);
1845 if (ret != 0)
1846 goto free_sc_h;
1847
1848 init_adb_hdrs(ctx);
1849
1850 v4l2_fh_init(&ctx->fh, video_devdata(file));
1851 file->private_data = &ctx->fh;
1852
1853 hdl = &ctx->hdl;
1854 v4l2_ctrl_handler_init(hdl, 1);
1855 v4l2_ctrl_new_custom(hdl, &vpe_bufs_per_job, NULL);
1856 if (hdl->error) {
1857 ret = hdl->error;
1858 goto exit_fh;
1859 }
1860 ctx->fh.ctrl_handler = hdl;
1861 v4l2_ctrl_handler_setup(hdl);
1862
1863 s_q_data = &ctx->q_data[Q_DATA_SRC];
1864 s_q_data->fmt = &vpe_formats[2];
1865 s_q_data->width = 1920;
1866 s_q_data->height = 1080;
1867 s_q_data->sizeimage[VPE_LUMA] = (s_q_data->width * s_q_data->height *
1868 s_q_data->fmt->vpdma_fmt[VPE_LUMA]->depth) >> 3;
1869 s_q_data->colorspace = V4L2_COLORSPACE_SMPTE170M;
1870 s_q_data->field = V4L2_FIELD_NONE;
1871 s_q_data->c_rect.left = 0;
1872 s_q_data->c_rect.top = 0;
1873 s_q_data->c_rect.width = s_q_data->width;
1874 s_q_data->c_rect.height = s_q_data->height;
1875 s_q_data->flags = 0;
1876
1877 ctx->q_data[Q_DATA_DST] = *s_q_data;
1878
1879 set_dei_shadow_registers(ctx);
1880 set_src_registers(ctx);
1881 set_dst_registers(ctx);
1882 ret = set_srcdst_params(ctx);
1883 if (ret)
1884 goto exit_fh;
1885
1886 ctx->m2m_ctx = v4l2_m2m_ctx_init(dev->m2m_dev, ctx, &queue_init);
1887
1888 if (IS_ERR(ctx->m2m_ctx)) {
1889 ret = PTR_ERR(ctx->m2m_ctx);
1890 goto exit_fh;
1891 }
1892
1893 v4l2_fh_add(&ctx->fh);
1894
1895 /*
1896 * for now, just report the creation of the first instance, we can later
1897 * optimize the driver to enable or disable clocks when the first
1898 * instance is created or the last instance released
1899 */
1900 if (atomic_inc_return(&dev->num_instances) == 1)
1901 vpe_dbg(dev, "first instance created\n");
1902
1903 ctx->bufs_per_job = VPE_DEF_BUFS_PER_JOB;
1904
1905 ctx->load_mmrs = true;
1906
1907 vpe_dbg(dev, "created instance %p, m2m_ctx: %p\n",
1908 ctx, ctx->m2m_ctx);
1909
1910 mutex_unlock(&dev->dev_mutex);
1911
1912 return 0;
1913 exit_fh:
1914 v4l2_ctrl_handler_free(hdl);
1915 v4l2_fh_exit(&ctx->fh);
1916 vpdma_free_desc_buf(&ctx->sc_coeff_v);
1917 free_sc_h:
1918 vpdma_free_desc_buf(&ctx->sc_coeff_h);
1919 free_mmr_adb:
1920 vpdma_free_desc_buf(&ctx->mmr_adb);
1921 free_desc_list:
1922 vpdma_free_desc_list(&ctx->desc_list);
1923 unlock:
1924 mutex_unlock(&dev->dev_mutex);
1925 free_ctx:
1926 kfree(ctx);
1927 return ret;
1928 }
1929
1930 static int vpe_release(struct file *file)
1931 {
1932 struct vpe_dev *dev = video_drvdata(file);
1933 struct vpe_ctx *ctx = file2ctx(file);
1934
1935 vpe_dbg(dev, "releasing instance %p\n", ctx);
1936
1937 mutex_lock(&dev->dev_mutex);
1938 free_vbs(ctx);
1939 free_mv_buffers(ctx);
1940 vpdma_free_desc_list(&ctx->desc_list);
1941 vpdma_free_desc_buf(&ctx->mmr_adb);
1942
1943 v4l2_fh_del(&ctx->fh);
1944 v4l2_fh_exit(&ctx->fh);
1945 v4l2_ctrl_handler_free(&ctx->hdl);
1946 v4l2_m2m_ctx_release(ctx->m2m_ctx);
1947
1948 kfree(ctx);
1949
1950 /*
1951 * for now, just report the release of the last instance, we can later
1952 * optimize the driver to enable or disable clocks when the first
1953 * instance is created or the last instance released
1954 */
1955 if (atomic_dec_return(&dev->num_instances) == 0)
1956 vpe_dbg(dev, "last instance released\n");
1957
1958 mutex_unlock(&dev->dev_mutex);
1959
1960 return 0;
1961 }
1962
1963 static unsigned int vpe_poll(struct file *file,
1964 struct poll_table_struct *wait)
1965 {
1966 struct vpe_ctx *ctx = file2ctx(file);
1967 struct vpe_dev *dev = ctx->dev;
1968 int ret;
1969
1970 mutex_lock(&dev->dev_mutex);
1971 ret = v4l2_m2m_poll(file, ctx->m2m_ctx, wait);
1972 mutex_unlock(&dev->dev_mutex);
1973 return ret;
1974 }
1975
1976 static int vpe_mmap(struct file *file, struct vm_area_struct *vma)
1977 {
1978 struct vpe_ctx *ctx = file2ctx(file);
1979 struct vpe_dev *dev = ctx->dev;
1980 int ret;
1981
1982 if (mutex_lock_interruptible(&dev->dev_mutex))
1983 return -ERESTARTSYS;
1984 ret = v4l2_m2m_mmap(file, ctx->m2m_ctx, vma);
1985 mutex_unlock(&dev->dev_mutex);
1986 return ret;
1987 }
1988
1989 static const struct v4l2_file_operations vpe_fops = {
1990 .owner = THIS_MODULE,
1991 .open = vpe_open,
1992 .release = vpe_release,
1993 .poll = vpe_poll,
1994 .unlocked_ioctl = video_ioctl2,
1995 .mmap = vpe_mmap,
1996 };
1997
1998 static struct video_device vpe_videodev = {
1999 .name = VPE_MODULE_NAME,
2000 .fops = &vpe_fops,
2001 .ioctl_ops = &vpe_ioctl_ops,
2002 .minor = -1,
2003 .release = video_device_release,
2004 .vfl_dir = VFL_DIR_M2M,
2005 };
2006
2007 static struct v4l2_m2m_ops m2m_ops = {
2008 .device_run = device_run,
2009 .job_ready = job_ready,
2010 .job_abort = job_abort,
2011 .lock = vpe_lock,
2012 .unlock = vpe_unlock,
2013 };
2014
2015 static int vpe_runtime_get(struct platform_device *pdev)
2016 {
2017 int r;
2018
2019 dev_dbg(&pdev->dev, "vpe_runtime_get\n");
2020
2021 r = pm_runtime_get_sync(&pdev->dev);
2022 WARN_ON(r < 0);
2023 return r < 0 ? r : 0;
2024 }
2025
2026 static void vpe_runtime_put(struct platform_device *pdev)
2027 {
2028
2029 int r;
2030
2031 dev_dbg(&pdev->dev, "vpe_runtime_put\n");
2032
2033 r = pm_runtime_put_sync(&pdev->dev);
2034 WARN_ON(r < 0 && r != -ENOSYS);
2035 }
2036
2037 static int vpe_probe(struct platform_device *pdev)
2038 {
2039 struct vpe_dev *dev;
2040 struct video_device *vfd;
2041 int ret, irq, func;
2042
2043 dev = devm_kzalloc(&pdev->dev, sizeof(*dev), GFP_KERNEL);
2044 if (!dev)
2045 return -ENOMEM;
2046
2047 spin_lock_init(&dev->lock);
2048
2049 ret = v4l2_device_register(&pdev->dev, &dev->v4l2_dev);
2050 if (ret)
2051 return ret;
2052
2053 atomic_set(&dev->num_instances, 0);
2054 mutex_init(&dev->dev_mutex);
2055
2056 dev->res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
2057 "vpe_top");
2058 /*
2059 * HACK: we get resource info from device tree in the form of a list of
2060 * VPE sub blocks, the driver currently uses only the base of vpe_top
2061 * for register access, the driver should be changed later to access
2062 * registers based on the sub block base addresses
2063 */
2064 dev->base = devm_ioremap(&pdev->dev, dev->res->start, SZ_32K);
2065 if (!dev->base) {
2066 ret = -ENOMEM;
2067 goto v4l2_dev_unreg;
2068 }
2069
2070 irq = platform_get_irq(pdev, 0);
2071 ret = devm_request_irq(&pdev->dev, irq, vpe_irq, 0, VPE_MODULE_NAME,
2072 dev);
2073 if (ret)
2074 goto v4l2_dev_unreg;
2075
2076 platform_set_drvdata(pdev, dev);
2077
2078 dev->alloc_ctx = vb2_dma_contig_init_ctx(&pdev->dev);
2079 if (IS_ERR(dev->alloc_ctx)) {
2080 vpe_err(dev, "Failed to alloc vb2 context\n");
2081 ret = PTR_ERR(dev->alloc_ctx);
2082 goto v4l2_dev_unreg;
2083 }
2084
2085 dev->m2m_dev = v4l2_m2m_init(&m2m_ops);
2086 if (IS_ERR(dev->m2m_dev)) {
2087 vpe_err(dev, "Failed to init mem2mem device\n");
2088 ret = PTR_ERR(dev->m2m_dev);
2089 goto rel_ctx;
2090 }
2091
2092 pm_runtime_enable(&pdev->dev);
2093
2094 ret = vpe_runtime_get(pdev);
2095 if (ret)
2096 goto rel_m2m;
2097
2098 /* Perform clk enable followed by reset */
2099 vpe_set_clock_enable(dev, 1);
2100
2101 vpe_top_reset(dev);
2102
2103 func = read_field_reg(dev, VPE_PID, VPE_PID_FUNC_MASK,
2104 VPE_PID_FUNC_SHIFT);
2105 vpe_dbg(dev, "VPE PID function %x\n", func);
2106
2107 vpe_top_vpdma_reset(dev);
2108
2109 dev->sc = sc_create(pdev);
2110 if (IS_ERR(dev->sc)) {
2111 ret = PTR_ERR(dev->sc);
2112 goto runtime_put;
2113 }
2114
2115 dev->csc = csc_create(pdev);
2116 if (IS_ERR(dev->csc)) {
2117 ret = PTR_ERR(dev->csc);
2118 goto runtime_put;
2119 }
2120
2121 dev->vpdma = vpdma_create(pdev);
2122 if (IS_ERR(dev->vpdma)) {
2123 ret = PTR_ERR(dev->vpdma);
2124 goto runtime_put;
2125 }
2126
2127 vfd = &dev->vfd;
2128 *vfd = vpe_videodev;
2129 vfd->lock = &dev->dev_mutex;
2130 vfd->v4l2_dev = &dev->v4l2_dev;
2131
2132 ret = video_register_device(vfd, VFL_TYPE_GRABBER, 0);
2133 if (ret) {
2134 vpe_err(dev, "Failed to register video device\n");
2135 goto runtime_put;
2136 }
2137
2138 video_set_drvdata(vfd, dev);
2139 snprintf(vfd->name, sizeof(vfd->name), "%s", vpe_videodev.name);
2140 dev_info(dev->v4l2_dev.dev, "Device registered as /dev/video%d\n",
2141 vfd->num);
2142
2143 return 0;
2144
2145 runtime_put:
2146 vpe_runtime_put(pdev);
2147 rel_m2m:
2148 pm_runtime_disable(&pdev->dev);
2149 v4l2_m2m_release(dev->m2m_dev);
2150 rel_ctx:
2151 vb2_dma_contig_cleanup_ctx(dev->alloc_ctx);
2152 v4l2_dev_unreg:
2153 v4l2_device_unregister(&dev->v4l2_dev);
2154
2155 return ret;
2156 }
2157
2158 static int vpe_remove(struct platform_device *pdev)
2159 {
2160 struct vpe_dev *dev =
2161 (struct vpe_dev *) platform_get_drvdata(pdev);
2162
2163 v4l2_info(&dev->v4l2_dev, "Removing " VPE_MODULE_NAME);
2164
2165 v4l2_m2m_release(dev->m2m_dev);
2166 video_unregister_device(&dev->vfd);
2167 v4l2_device_unregister(&dev->v4l2_dev);
2168 vb2_dma_contig_cleanup_ctx(dev->alloc_ctx);
2169
2170 vpe_set_clock_enable(dev, 0);
2171 vpe_runtime_put(pdev);
2172 pm_runtime_disable(&pdev->dev);
2173
2174 return 0;
2175 }
2176
2177 #if defined(CONFIG_OF)
2178 static const struct of_device_id vpe_of_match[] = {
2179 {
2180 .compatible = "ti,vpe",
2181 },
2182 {},
2183 };
2184 #else
2185 #define vpe_of_match NULL
2186 #endif
2187
2188 static struct platform_driver vpe_pdrv = {
2189 .probe = vpe_probe,
2190 .remove = vpe_remove,
2191 .driver = {
2192 .name = VPE_MODULE_NAME,
2193 .owner = THIS_MODULE,
2194 .of_match_table = vpe_of_match,
2195 },
2196 };
2197
2198 module_platform_driver(vpe_pdrv);
2199
2200 MODULE_DESCRIPTION("TI VPE driver");
2201 MODULE_AUTHOR("Dale Farnsworth, <dale@farnsworth.org>");
2202 MODULE_LICENSE("GPL");
Linux with added FPC-III support