Input: xpad - add support for Xbox1 PDP Camo series gamepad
[linux/fpc-iii.git] / drivers / media / platform / ti-vpe / vpe.c
blob0189f7f7cb03fbeba9d3222f77d090d0cd6e1bb3
1 /*
2 * TI VPE mem2mem driver, based on the virtual v4l2-mem2mem example driver
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>
9 * Copyright (c) 2009-2010 Samsung Electronics Co., Ltd.
10 * Pawel Osciak, <pawel@osciak.com>
11 * Marek Szyprowski, <m.szyprowski@samsung.com>
13 * Based on the virtual v4l2-mem2mem example device
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
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/of.h>
29 #include <linux/platform_device.h>
30 #include <linux/pm_runtime.h>
31 #include <linux/sched.h>
32 #include <linux/slab.h>
33 #include <linux/videodev2.h>
34 #include <linux/log2.h>
35 #include <linux/sizes.h>
37 #include <media/v4l2-common.h>
38 #include <media/v4l2-ctrls.h>
39 #include <media/v4l2-device.h>
40 #include <media/v4l2-event.h>
41 #include <media/v4l2-ioctl.h>
42 #include <media/v4l2-mem2mem.h>
43 #include <media/videobuf2-v4l2.h>
44 #include <media/videobuf2-dma-contig.h>
46 #include "vpdma.h"
47 #include "vpe_regs.h"
48 #include "sc.h"
49 #include "csc.h"
51 #define VPE_MODULE_NAME "vpe"
53 /* minimum and maximum frame sizes */
54 #define MIN_W 32
55 #define MIN_H 32
56 #define MAX_W 1920
57 #define MAX_H 1080
59 /* required alignments */
60 #define S_ALIGN 0 /* multiple of 1 */
61 #define H_ALIGN 1 /* multiple of 2 */
63 /* flags that indicate a format can be used for capture/output */
64 #define VPE_FMT_TYPE_CAPTURE (1 << 0)
65 #define VPE_FMT_TYPE_OUTPUT (1 << 1)
67 /* used as plane indices */
68 #define VPE_MAX_PLANES 2
69 #define VPE_LUMA 0
70 #define VPE_CHROMA 1
72 /* per m2m context info */
73 #define VPE_MAX_SRC_BUFS 3 /* need 3 src fields to de-interlace */
75 #define VPE_DEF_BUFS_PER_JOB 1 /* default one buffer per batch job */
78 * each VPE context can need up to 3 config descriptors, 7 input descriptors,
79 * 3 output descriptors, and 10 control descriptors
81 #define VPE_DESC_LIST_SIZE (10 * VPDMA_DTD_DESC_SIZE + \
82 13 * VPDMA_CFD_CTD_DESC_SIZE)
84 #define vpe_dbg(vpedev, fmt, arg...) \
85 dev_dbg((vpedev)->v4l2_dev.dev, fmt, ##arg)
86 #define vpe_err(vpedev, fmt, arg...) \
87 dev_err((vpedev)->v4l2_dev.dev, fmt, ##arg)
89 struct vpe_us_coeffs {
90 unsigned short anchor_fid0_c0;
91 unsigned short anchor_fid0_c1;
92 unsigned short anchor_fid0_c2;
93 unsigned short anchor_fid0_c3;
94 unsigned short interp_fid0_c0;
95 unsigned short interp_fid0_c1;
96 unsigned short interp_fid0_c2;
97 unsigned short interp_fid0_c3;
98 unsigned short anchor_fid1_c0;
99 unsigned short anchor_fid1_c1;
100 unsigned short anchor_fid1_c2;
101 unsigned short anchor_fid1_c3;
102 unsigned short interp_fid1_c0;
103 unsigned short interp_fid1_c1;
104 unsigned short interp_fid1_c2;
105 unsigned short interp_fid1_c3;
109 * Default upsampler coefficients
111 static const struct vpe_us_coeffs us_coeffs[] = {
113 /* Coefficients for progressive input */
114 0x00C8, 0x0348, 0x0018, 0x3FD8, 0x3FB8, 0x0378, 0x00E8, 0x3FE8,
115 0x00C8, 0x0348, 0x0018, 0x3FD8, 0x3FB8, 0x0378, 0x00E8, 0x3FE8,
118 /* Coefficients for Top Field Interlaced input */
119 0x0051, 0x03D5, 0x3FE3, 0x3FF7, 0x3FB5, 0x02E9, 0x018F, 0x3FD3,
120 /* Coefficients for Bottom Field Interlaced input */
121 0x016B, 0x0247, 0x00B1, 0x3F9D, 0x3FCF, 0x03DB, 0x005D, 0x3FF9,
126 * the following registers are for configuring some of the parameters of the
127 * motion and edge detection blocks inside DEI, these generally remain the same,
128 * these could be passed later via userspace if some one needs to tweak these.
130 struct vpe_dei_regs {
131 unsigned long mdt_spacial_freq_thr_reg; /* VPE_DEI_REG2 */
132 unsigned long edi_config_reg; /* VPE_DEI_REG3 */
133 unsigned long edi_lut_reg0; /* VPE_DEI_REG4 */
134 unsigned long edi_lut_reg1; /* VPE_DEI_REG5 */
135 unsigned long edi_lut_reg2; /* VPE_DEI_REG6 */
136 unsigned long edi_lut_reg3; /* VPE_DEI_REG7 */
140 * default expert DEI register values, unlikely to be modified.
142 static const struct vpe_dei_regs dei_regs = {
143 .mdt_spacial_freq_thr_reg = 0x020C0804u,
144 .edi_config_reg = 0x0118100Fu,
145 .edi_lut_reg0 = 0x08040200u,
146 .edi_lut_reg1 = 0x1010100Cu,
147 .edi_lut_reg2 = 0x10101010u,
148 .edi_lut_reg3 = 0x10101010u,
152 * The port_data structure contains per-port data.
154 struct vpe_port_data {
155 enum vpdma_channel channel; /* VPDMA channel */
156 u8 vb_index; /* input frame f, f-1, f-2 index */
157 u8 vb_part; /* plane index for co-panar formats */
161 * Define indices into the port_data tables
163 #define VPE_PORT_LUMA1_IN 0
164 #define VPE_PORT_CHROMA1_IN 1
165 #define VPE_PORT_LUMA2_IN 2
166 #define VPE_PORT_CHROMA2_IN 3
167 #define VPE_PORT_LUMA3_IN 4
168 #define VPE_PORT_CHROMA3_IN 5
169 #define VPE_PORT_MV_IN 6
170 #define VPE_PORT_MV_OUT 7
171 #define VPE_PORT_LUMA_OUT 8
172 #define VPE_PORT_CHROMA_OUT 9
173 #define VPE_PORT_RGB_OUT 10
175 static const struct vpe_port_data port_data[11] = {
176 [VPE_PORT_LUMA1_IN] = {
177 .channel = VPE_CHAN_LUMA1_IN,
178 .vb_index = 0,
179 .vb_part = VPE_LUMA,
181 [VPE_PORT_CHROMA1_IN] = {
182 .channel = VPE_CHAN_CHROMA1_IN,
183 .vb_index = 0,
184 .vb_part = VPE_CHROMA,
186 [VPE_PORT_LUMA2_IN] = {
187 .channel = VPE_CHAN_LUMA2_IN,
188 .vb_index = 1,
189 .vb_part = VPE_LUMA,
191 [VPE_PORT_CHROMA2_IN] = {
192 .channel = VPE_CHAN_CHROMA2_IN,
193 .vb_index = 1,
194 .vb_part = VPE_CHROMA,
196 [VPE_PORT_LUMA3_IN] = {
197 .channel = VPE_CHAN_LUMA3_IN,
198 .vb_index = 2,
199 .vb_part = VPE_LUMA,
201 [VPE_PORT_CHROMA3_IN] = {
202 .channel = VPE_CHAN_CHROMA3_IN,
203 .vb_index = 2,
204 .vb_part = VPE_CHROMA,
206 [VPE_PORT_MV_IN] = {
207 .channel = VPE_CHAN_MV_IN,
209 [VPE_PORT_MV_OUT] = {
210 .channel = VPE_CHAN_MV_OUT,
212 [VPE_PORT_LUMA_OUT] = {
213 .channel = VPE_CHAN_LUMA_OUT,
214 .vb_part = VPE_LUMA,
216 [VPE_PORT_CHROMA_OUT] = {
217 .channel = VPE_CHAN_CHROMA_OUT,
218 .vb_part = VPE_CHROMA,
220 [VPE_PORT_RGB_OUT] = {
221 .channel = VPE_CHAN_RGB_OUT,
222 .vb_part = VPE_LUMA,
227 /* driver info for each of the supported video formats */
228 struct vpe_fmt {
229 char *name; /* human-readable name */
230 u32 fourcc; /* standard format identifier */
231 u8 types; /* CAPTURE and/or OUTPUT */
232 u8 coplanar; /* set for unpacked Luma and Chroma */
233 /* vpdma format info for each plane */
234 struct vpdma_data_format const *vpdma_fmt[VPE_MAX_PLANES];
237 static struct vpe_fmt vpe_formats[] = {
239 .name = "YUV 422 co-planar",
240 .fourcc = V4L2_PIX_FMT_NV16,
241 .types = VPE_FMT_TYPE_CAPTURE | VPE_FMT_TYPE_OUTPUT,
242 .coplanar = 1,
243 .vpdma_fmt = { &vpdma_yuv_fmts[VPDMA_DATA_FMT_Y444],
244 &vpdma_yuv_fmts[VPDMA_DATA_FMT_C444],
248 .name = "YUV 420 co-planar",
249 .fourcc = V4L2_PIX_FMT_NV12,
250 .types = VPE_FMT_TYPE_CAPTURE | VPE_FMT_TYPE_OUTPUT,
251 .coplanar = 1,
252 .vpdma_fmt = { &vpdma_yuv_fmts[VPDMA_DATA_FMT_Y420],
253 &vpdma_yuv_fmts[VPDMA_DATA_FMT_C420],
257 .name = "YUYV 422 packed",
258 .fourcc = V4L2_PIX_FMT_YUYV,
259 .types = VPE_FMT_TYPE_CAPTURE | VPE_FMT_TYPE_OUTPUT,
260 .coplanar = 0,
261 .vpdma_fmt = { &vpdma_yuv_fmts[VPDMA_DATA_FMT_YC422],
265 .name = "UYVY 422 packed",
266 .fourcc = V4L2_PIX_FMT_UYVY,
267 .types = VPE_FMT_TYPE_CAPTURE | VPE_FMT_TYPE_OUTPUT,
268 .coplanar = 0,
269 .vpdma_fmt = { &vpdma_yuv_fmts[VPDMA_DATA_FMT_CY422],
273 .name = "RGB888 packed",
274 .fourcc = V4L2_PIX_FMT_RGB24,
275 .types = VPE_FMT_TYPE_CAPTURE,
276 .coplanar = 0,
277 .vpdma_fmt = { &vpdma_rgb_fmts[VPDMA_DATA_FMT_RGB24],
281 .name = "ARGB32",
282 .fourcc = V4L2_PIX_FMT_RGB32,
283 .types = VPE_FMT_TYPE_CAPTURE,
284 .coplanar = 0,
285 .vpdma_fmt = { &vpdma_rgb_fmts[VPDMA_DATA_FMT_ARGB32],
289 .name = "BGR888 packed",
290 .fourcc = V4L2_PIX_FMT_BGR24,
291 .types = VPE_FMT_TYPE_CAPTURE,
292 .coplanar = 0,
293 .vpdma_fmt = { &vpdma_rgb_fmts[VPDMA_DATA_FMT_BGR24],
297 .name = "ABGR32",
298 .fourcc = V4L2_PIX_FMT_BGR32,
299 .types = VPE_FMT_TYPE_CAPTURE,
300 .coplanar = 0,
301 .vpdma_fmt = { &vpdma_rgb_fmts[VPDMA_DATA_FMT_ABGR32],
307 * per-queue, driver-specific private data.
308 * there is one source queue and one destination queue for each m2m context.
310 struct vpe_q_data {
311 unsigned int width; /* frame width */
312 unsigned int height; /* frame height */
313 unsigned int bytesperline[VPE_MAX_PLANES]; /* bytes per line in memory */
314 enum v4l2_colorspace colorspace;
315 enum v4l2_field field; /* supported field value */
316 unsigned int flags;
317 unsigned int sizeimage[VPE_MAX_PLANES]; /* image size in memory */
318 struct v4l2_rect c_rect; /* crop/compose rectangle */
319 struct vpe_fmt *fmt; /* format info */
322 /* vpe_q_data flag bits */
323 #define Q_DATA_FRAME_1D (1 << 0)
324 #define Q_DATA_MODE_TILED (1 << 1)
325 #define Q_DATA_INTERLACED (1 << 2)
327 enum {
328 Q_DATA_SRC = 0,
329 Q_DATA_DST = 1,
332 /* find our format description corresponding to the passed v4l2_format */
333 static struct vpe_fmt *find_format(struct v4l2_format *f)
335 struct vpe_fmt *fmt;
336 unsigned int k;
338 for (k = 0; k < ARRAY_SIZE(vpe_formats); k++) {
339 fmt = &vpe_formats[k];
340 if (fmt->fourcc == f->fmt.pix.pixelformat)
341 return fmt;
344 return NULL;
348 * there is one vpe_dev structure in the driver, it is shared by
349 * all instances.
351 struct vpe_dev {
352 struct v4l2_device v4l2_dev;
353 struct video_device vfd;
354 struct v4l2_m2m_dev *m2m_dev;
356 atomic_t num_instances; /* count of driver instances */
357 dma_addr_t loaded_mmrs; /* shadow mmrs in device */
358 struct mutex dev_mutex;
359 spinlock_t lock;
361 int irq;
362 void __iomem *base;
363 struct resource *res;
365 struct vpdma_data *vpdma; /* vpdma data handle */
366 struct sc_data *sc; /* scaler data handle */
367 struct csc_data *csc; /* csc data handle */
371 * There is one vpe_ctx structure for each m2m context.
373 struct vpe_ctx {
374 struct v4l2_fh fh;
375 struct vpe_dev *dev;
376 struct v4l2_ctrl_handler hdl;
378 unsigned int field; /* current field */
379 unsigned int sequence; /* current frame/field seq */
380 unsigned int aborting; /* abort after next irq */
382 unsigned int bufs_per_job; /* input buffers per batch */
383 unsigned int bufs_completed; /* bufs done in this batch */
385 struct vpe_q_data q_data[2]; /* src & dst queue data */
386 struct vb2_v4l2_buffer *src_vbs[VPE_MAX_SRC_BUFS];
387 struct vb2_v4l2_buffer *dst_vb;
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 */
397 bool deinterlacing; /* using de-interlacer */
398 bool load_mmrs; /* have new shadow reg values */
400 unsigned int src_mv_buf_selector;
405 * M2M devices get 2 queues.
406 * Return the queue given the type.
408 static struct vpe_q_data *get_q_data(struct vpe_ctx *ctx,
409 enum v4l2_buf_type type)
411 switch (type) {
412 case V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE:
413 case V4L2_BUF_TYPE_VIDEO_OUTPUT:
414 return &ctx->q_data[Q_DATA_SRC];
415 case V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE:
416 case V4L2_BUF_TYPE_VIDEO_CAPTURE:
417 return &ctx->q_data[Q_DATA_DST];
418 default:
419 BUG();
421 return NULL;
424 static u32 read_reg(struct vpe_dev *dev, int offset)
426 return ioread32(dev->base + offset);
429 static void write_reg(struct vpe_dev *dev, int offset, u32 value)
431 iowrite32(value, dev->base + offset);
434 /* register field read/write helpers */
435 static int get_field(u32 value, u32 mask, int shift)
437 return (value & (mask << shift)) >> shift;
440 static int read_field_reg(struct vpe_dev *dev, int offset, u32 mask, int shift)
442 return get_field(read_reg(dev, offset), mask, shift);
445 static void write_field(u32 *valp, u32 field, u32 mask, int shift)
447 u32 val = *valp;
449 val &= ~(mask << shift);
450 val |= (field & mask) << shift;
451 *valp = val;
454 static void write_field_reg(struct vpe_dev *dev, int offset, u32 field,
455 u32 mask, int shift)
457 u32 val = read_reg(dev, offset);
459 write_field(&val, field, mask, shift);
461 write_reg(dev, offset, val);
465 * DMA address/data block for the shadow registers
467 struct vpe_mmr_adb {
468 struct vpdma_adb_hdr out_fmt_hdr;
469 u32 out_fmt_reg[1];
470 u32 out_fmt_pad[3];
471 struct vpdma_adb_hdr us1_hdr;
472 u32 us1_regs[8];
473 struct vpdma_adb_hdr us2_hdr;
474 u32 us2_regs[8];
475 struct vpdma_adb_hdr us3_hdr;
476 u32 us3_regs[8];
477 struct vpdma_adb_hdr dei_hdr;
478 u32 dei_regs[8];
479 struct vpdma_adb_hdr sc_hdr0;
480 u32 sc_regs0[7];
481 u32 sc_pad0[1];
482 struct vpdma_adb_hdr sc_hdr8;
483 u32 sc_regs8[6];
484 u32 sc_pad8[2];
485 struct vpdma_adb_hdr sc_hdr17;
486 u32 sc_regs17[9];
487 u32 sc_pad17[3];
488 struct vpdma_adb_hdr csc_hdr;
489 u32 csc_regs[6];
490 u32 csc_pad[2];
493 #define GET_OFFSET_TOP(ctx, obj, reg) \
494 ((obj)->res->start - ctx->dev->res->start + reg)
496 #define VPE_SET_MMR_ADB_HDR(ctx, hdr, regs, offset_a) \
497 VPDMA_SET_MMR_ADB_HDR(ctx->mmr_adb, vpe_mmr_adb, hdr, regs, offset_a)
499 * Set the headers for all of the address/data block structures.
501 static void init_adb_hdrs(struct vpe_ctx *ctx)
503 VPE_SET_MMR_ADB_HDR(ctx, out_fmt_hdr, out_fmt_reg, VPE_CLK_FORMAT_SELECT);
504 VPE_SET_MMR_ADB_HDR(ctx, us1_hdr, us1_regs, VPE_US1_R0);
505 VPE_SET_MMR_ADB_HDR(ctx, us2_hdr, us2_regs, VPE_US2_R0);
506 VPE_SET_MMR_ADB_HDR(ctx, us3_hdr, us3_regs, VPE_US3_R0);
507 VPE_SET_MMR_ADB_HDR(ctx, dei_hdr, dei_regs, VPE_DEI_FRAME_SIZE);
508 VPE_SET_MMR_ADB_HDR(ctx, sc_hdr0, sc_regs0,
509 GET_OFFSET_TOP(ctx, ctx->dev->sc, CFG_SC0));
510 VPE_SET_MMR_ADB_HDR(ctx, sc_hdr8, sc_regs8,
511 GET_OFFSET_TOP(ctx, ctx->dev->sc, CFG_SC8));
512 VPE_SET_MMR_ADB_HDR(ctx, sc_hdr17, sc_regs17,
513 GET_OFFSET_TOP(ctx, ctx->dev->sc, CFG_SC17));
514 VPE_SET_MMR_ADB_HDR(ctx, csc_hdr, csc_regs,
515 GET_OFFSET_TOP(ctx, ctx->dev->csc, CSC_CSC00));
519 * Allocate or re-allocate the motion vector DMA buffers
520 * There are two buffers, one for input and one for output.
521 * However, the roles are reversed after each field is processed.
522 * In other words, after each field is processed, the previous
523 * output (dst) MV buffer becomes the new input (src) MV buffer.
525 static int realloc_mv_buffers(struct vpe_ctx *ctx, size_t size)
527 struct device *dev = ctx->dev->v4l2_dev.dev;
529 if (ctx->mv_buf_size == size)
530 return 0;
532 if (ctx->mv_buf[0])
533 dma_free_coherent(dev, ctx->mv_buf_size, ctx->mv_buf[0],
534 ctx->mv_buf_dma[0]);
536 if (ctx->mv_buf[1])
537 dma_free_coherent(dev, ctx->mv_buf_size, ctx->mv_buf[1],
538 ctx->mv_buf_dma[1]);
540 if (size == 0)
541 return 0;
543 ctx->mv_buf[0] = dma_alloc_coherent(dev, size, &ctx->mv_buf_dma[0],
544 GFP_KERNEL);
545 if (!ctx->mv_buf[0]) {
546 vpe_err(ctx->dev, "failed to allocate motion vector buffer\n");
547 return -ENOMEM;
550 ctx->mv_buf[1] = dma_alloc_coherent(dev, size, &ctx->mv_buf_dma[1],
551 GFP_KERNEL);
552 if (!ctx->mv_buf[1]) {
553 vpe_err(ctx->dev, "failed to allocate motion vector buffer\n");
554 dma_free_coherent(dev, size, ctx->mv_buf[0],
555 ctx->mv_buf_dma[0]);
557 return -ENOMEM;
560 ctx->mv_buf_size = size;
561 ctx->src_mv_buf_selector = 0;
563 return 0;
566 static void free_mv_buffers(struct vpe_ctx *ctx)
568 realloc_mv_buffers(ctx, 0);
572 * While de-interlacing, we keep the two most recent input buffers
573 * around. This function frees those two buffers when we have
574 * finished processing the current stream.
576 static void free_vbs(struct vpe_ctx *ctx)
578 struct vpe_dev *dev = ctx->dev;
579 unsigned long flags;
581 if (ctx->src_vbs[2] == NULL)
582 return;
584 spin_lock_irqsave(&dev->lock, flags);
585 if (ctx->src_vbs[2]) {
586 v4l2_m2m_buf_done(ctx->src_vbs[2], VB2_BUF_STATE_DONE);
587 v4l2_m2m_buf_done(ctx->src_vbs[1], VB2_BUF_STATE_DONE);
589 spin_unlock_irqrestore(&dev->lock, flags);
593 * Enable or disable the VPE clocks
595 static void vpe_set_clock_enable(struct vpe_dev *dev, bool on)
597 u32 val = 0;
599 if (on)
600 val = VPE_DATA_PATH_CLK_ENABLE | VPE_VPEDMA_CLK_ENABLE;
601 write_reg(dev, VPE_CLK_ENABLE, val);
604 static void vpe_top_reset(struct vpe_dev *dev)
607 write_field_reg(dev, VPE_CLK_RESET, 1, VPE_DATA_PATH_CLK_RESET_MASK,
608 VPE_DATA_PATH_CLK_RESET_SHIFT);
610 usleep_range(100, 150);
612 write_field_reg(dev, VPE_CLK_RESET, 0, VPE_DATA_PATH_CLK_RESET_MASK,
613 VPE_DATA_PATH_CLK_RESET_SHIFT);
616 static void vpe_top_vpdma_reset(struct vpe_dev *dev)
618 write_field_reg(dev, VPE_CLK_RESET, 1, VPE_VPDMA_CLK_RESET_MASK,
619 VPE_VPDMA_CLK_RESET_SHIFT);
621 usleep_range(100, 150);
623 write_field_reg(dev, VPE_CLK_RESET, 0, VPE_VPDMA_CLK_RESET_MASK,
624 VPE_VPDMA_CLK_RESET_SHIFT);
628 * Load the correct of upsampler coefficients into the shadow MMRs
630 static void set_us_coefficients(struct vpe_ctx *ctx)
632 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
633 struct vpe_q_data *s_q_data = &ctx->q_data[Q_DATA_SRC];
634 u32 *us1_reg = &mmr_adb->us1_regs[0];
635 u32 *us2_reg = &mmr_adb->us2_regs[0];
636 u32 *us3_reg = &mmr_adb->us3_regs[0];
637 const unsigned short *cp, *end_cp;
639 cp = &us_coeffs[0].anchor_fid0_c0;
641 if (s_q_data->flags & Q_DATA_INTERLACED) /* interlaced */
642 cp += sizeof(us_coeffs[0]) / sizeof(*cp);
644 end_cp = cp + sizeof(us_coeffs[0]) / sizeof(*cp);
646 while (cp < end_cp) {
647 write_field(us1_reg, *cp++, VPE_US_C0_MASK, VPE_US_C0_SHIFT);
648 write_field(us1_reg, *cp++, VPE_US_C1_MASK, VPE_US_C1_SHIFT);
649 *us2_reg++ = *us1_reg;
650 *us3_reg++ = *us1_reg++;
652 ctx->load_mmrs = true;
656 * Set the upsampler config mode and the VPDMA line mode in the shadow MMRs.
658 static void set_cfg_and_line_modes(struct vpe_ctx *ctx)
660 struct vpe_fmt *fmt = ctx->q_data[Q_DATA_SRC].fmt;
661 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
662 u32 *us1_reg0 = &mmr_adb->us1_regs[0];
663 u32 *us2_reg0 = &mmr_adb->us2_regs[0];
664 u32 *us3_reg0 = &mmr_adb->us3_regs[0];
665 int line_mode = 1;
666 int cfg_mode = 1;
669 * Cfg Mode 0: YUV420 source, enable upsampler, DEI is de-interlacing.
670 * Cfg Mode 1: YUV422 source, disable upsampler, DEI is de-interlacing.
673 if (fmt->fourcc == V4L2_PIX_FMT_NV12) {
674 cfg_mode = 0;
675 line_mode = 0; /* double lines to line buffer */
678 write_field(us1_reg0, cfg_mode, VPE_US_MODE_MASK, VPE_US_MODE_SHIFT);
679 write_field(us2_reg0, cfg_mode, VPE_US_MODE_MASK, VPE_US_MODE_SHIFT);
680 write_field(us3_reg0, cfg_mode, VPE_US_MODE_MASK, VPE_US_MODE_SHIFT);
682 /* regs for now */
683 vpdma_set_line_mode(ctx->dev->vpdma, line_mode, VPE_CHAN_CHROMA1_IN);
684 vpdma_set_line_mode(ctx->dev->vpdma, line_mode, VPE_CHAN_CHROMA2_IN);
685 vpdma_set_line_mode(ctx->dev->vpdma, line_mode, VPE_CHAN_CHROMA3_IN);
687 /* frame start for input luma */
688 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
689 VPE_CHAN_LUMA1_IN);
690 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
691 VPE_CHAN_LUMA2_IN);
692 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
693 VPE_CHAN_LUMA3_IN);
695 /* frame start for input chroma */
696 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
697 VPE_CHAN_CHROMA1_IN);
698 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
699 VPE_CHAN_CHROMA2_IN);
700 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
701 VPE_CHAN_CHROMA3_IN);
703 /* frame start for MV in client */
704 vpdma_set_frame_start_event(ctx->dev->vpdma, VPDMA_FSEVENT_CHANNEL_ACTIVE,
705 VPE_CHAN_MV_IN);
707 ctx->load_mmrs = true;
711 * Set the shadow registers that are modified when the source
712 * format changes.
714 static void set_src_registers(struct vpe_ctx *ctx)
716 set_us_coefficients(ctx);
720 * Set the shadow registers that are modified when the destination
721 * format changes.
723 static void set_dst_registers(struct vpe_ctx *ctx)
725 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
726 enum v4l2_colorspace clrspc = ctx->q_data[Q_DATA_DST].colorspace;
727 struct vpe_fmt *fmt = ctx->q_data[Q_DATA_DST].fmt;
728 u32 val = 0;
730 if (clrspc == V4L2_COLORSPACE_SRGB)
731 val |= VPE_RGB_OUT_SELECT;
732 else if (fmt->fourcc == V4L2_PIX_FMT_NV16)
733 val |= VPE_COLOR_SEPARATE_422;
736 * the source of CHR_DS and CSC is always the scaler, irrespective of
737 * whether it's used or not
739 val |= VPE_DS_SRC_DEI_SCALER | VPE_CSC_SRC_DEI_SCALER;
741 if (fmt->fourcc != V4L2_PIX_FMT_NV12)
742 val |= VPE_DS_BYPASS;
744 mmr_adb->out_fmt_reg[0] = val;
746 ctx->load_mmrs = true;
750 * Set the de-interlacer shadow register values
752 static void set_dei_regs(struct vpe_ctx *ctx)
754 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
755 struct vpe_q_data *s_q_data = &ctx->q_data[Q_DATA_SRC];
756 unsigned int src_h = s_q_data->c_rect.height;
757 unsigned int src_w = s_q_data->c_rect.width;
758 u32 *dei_mmr0 = &mmr_adb->dei_regs[0];
759 bool deinterlace = true;
760 u32 val = 0;
763 * according to TRM, we should set DEI in progressive bypass mode when
764 * the input content is progressive, however, DEI is bypassed correctly
765 * for both progressive and interlace content in interlace bypass mode.
766 * It has been recommended not to use progressive bypass mode.
768 if ((!ctx->deinterlacing && (s_q_data->flags & Q_DATA_INTERLACED)) ||
769 !(s_q_data->flags & Q_DATA_INTERLACED)) {
770 deinterlace = false;
771 val = VPE_DEI_INTERLACE_BYPASS;
774 src_h = deinterlace ? src_h * 2 : src_h;
776 val |= (src_h << VPE_DEI_HEIGHT_SHIFT) |
777 (src_w << VPE_DEI_WIDTH_SHIFT) |
778 VPE_DEI_FIELD_FLUSH;
780 *dei_mmr0 = val;
782 ctx->load_mmrs = true;
785 static void set_dei_shadow_registers(struct vpe_ctx *ctx)
787 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
788 u32 *dei_mmr = &mmr_adb->dei_regs[0];
789 const struct vpe_dei_regs *cur = &dei_regs;
791 dei_mmr[2] = cur->mdt_spacial_freq_thr_reg;
792 dei_mmr[3] = cur->edi_config_reg;
793 dei_mmr[4] = cur->edi_lut_reg0;
794 dei_mmr[5] = cur->edi_lut_reg1;
795 dei_mmr[6] = cur->edi_lut_reg2;
796 dei_mmr[7] = cur->edi_lut_reg3;
798 ctx->load_mmrs = true;
802 * Set the shadow registers whose values are modified when either the
803 * source or destination format is changed.
805 static int set_srcdst_params(struct vpe_ctx *ctx)
807 struct vpe_q_data *s_q_data = &ctx->q_data[Q_DATA_SRC];
808 struct vpe_q_data *d_q_data = &ctx->q_data[Q_DATA_DST];
809 struct vpe_mmr_adb *mmr_adb = ctx->mmr_adb.addr;
810 unsigned int src_w = s_q_data->c_rect.width;
811 unsigned int src_h = s_q_data->c_rect.height;
812 unsigned int dst_w = d_q_data->c_rect.width;
813 unsigned int dst_h = d_q_data->c_rect.height;
814 size_t mv_buf_size;
815 int ret;
817 ctx->sequence = 0;
818 ctx->field = V4L2_FIELD_TOP;
820 if ((s_q_data->flags & Q_DATA_INTERLACED) &&
821 !(d_q_data->flags & Q_DATA_INTERLACED)) {
822 int bytes_per_line;
823 const struct vpdma_data_format *mv =
824 &vpdma_misc_fmts[VPDMA_DATA_FMT_MV];
827 * we make sure that the source image has a 16 byte aligned
828 * stride, we need to do the same for the motion vector buffer
829 * by aligning it's stride to the next 16 byte boundry. this
830 * extra space will not be used by the de-interlacer, but will
831 * ensure that vpdma operates correctly
833 bytes_per_line = ALIGN((s_q_data->width * mv->depth) >> 3,
834 VPDMA_STRIDE_ALIGN);
835 mv_buf_size = bytes_per_line * s_q_data->height;
837 ctx->deinterlacing = true;
838 src_h <<= 1;
839 } else {
840 ctx->deinterlacing = false;
841 mv_buf_size = 0;
844 free_vbs(ctx);
846 ret = realloc_mv_buffers(ctx, mv_buf_size);
847 if (ret)
848 return ret;
850 set_cfg_and_line_modes(ctx);
851 set_dei_regs(ctx);
853 csc_set_coeff(ctx->dev->csc, &mmr_adb->csc_regs[0],
854 s_q_data->colorspace, d_q_data->colorspace);
856 sc_set_hs_coeffs(ctx->dev->sc, ctx->sc_coeff_h.addr, src_w, dst_w);
857 sc_set_vs_coeffs(ctx->dev->sc, ctx->sc_coeff_v.addr, src_h, dst_h);
859 sc_config_scaler(ctx->dev->sc, &mmr_adb->sc_regs0[0],
860 &mmr_adb->sc_regs8[0], &mmr_adb->sc_regs17[0],
861 src_w, src_h, dst_w, dst_h);
863 return 0;
867 * Return the vpe_ctx structure for a given struct file
869 static struct vpe_ctx *file2ctx(struct file *file)
871 return container_of(file->private_data, struct vpe_ctx, fh);
875 * mem2mem callbacks
879 * job_ready() - check whether an instance is ready to be scheduled to run
881 static int job_ready(void *priv)
883 struct vpe_ctx *ctx = priv;
884 int needed = ctx->bufs_per_job;
886 if (ctx->deinterlacing && ctx->src_vbs[2] == NULL)
887 needed += 2; /* need additional two most recent fields */
889 if (v4l2_m2m_num_src_bufs_ready(ctx->fh.m2m_ctx) < needed)
890 return 0;
892 if (v4l2_m2m_num_dst_bufs_ready(ctx->fh.m2m_ctx) < needed)
893 return 0;
895 return 1;
898 static void job_abort(void *priv)
900 struct vpe_ctx *ctx = priv;
902 /* Will cancel the transaction in the next interrupt handler */
903 ctx->aborting = 1;
907 * Lock access to the device
909 static void vpe_lock(void *priv)
911 struct vpe_ctx *ctx = priv;
912 struct vpe_dev *dev = ctx->dev;
913 mutex_lock(&dev->dev_mutex);
916 static void vpe_unlock(void *priv)
918 struct vpe_ctx *ctx = priv;
919 struct vpe_dev *dev = ctx->dev;
920 mutex_unlock(&dev->dev_mutex);
923 static void vpe_dump_regs(struct vpe_dev *dev)
925 #define DUMPREG(r) vpe_dbg(dev, "%-35s %08x\n", #r, read_reg(dev, VPE_##r))
927 vpe_dbg(dev, "VPE Registers:\n");
929 DUMPREG(PID);
930 DUMPREG(SYSCONFIG);
931 DUMPREG(INT0_STATUS0_RAW);
932 DUMPREG(INT0_STATUS0);
933 DUMPREG(INT0_ENABLE0);
934 DUMPREG(INT0_STATUS1_RAW);
935 DUMPREG(INT0_STATUS1);
936 DUMPREG(INT0_ENABLE1);
937 DUMPREG(CLK_ENABLE);
938 DUMPREG(CLK_RESET);
939 DUMPREG(CLK_FORMAT_SELECT);
940 DUMPREG(CLK_RANGE_MAP);
941 DUMPREG(US1_R0);
942 DUMPREG(US1_R1);
943 DUMPREG(US1_R2);
944 DUMPREG(US1_R3);
945 DUMPREG(US1_R4);
946 DUMPREG(US1_R5);
947 DUMPREG(US1_R6);
948 DUMPREG(US1_R7);
949 DUMPREG(US2_R0);
950 DUMPREG(US2_R1);
951 DUMPREG(US2_R2);
952 DUMPREG(US2_R3);
953 DUMPREG(US2_R4);
954 DUMPREG(US2_R5);
955 DUMPREG(US2_R6);
956 DUMPREG(US2_R7);
957 DUMPREG(US3_R0);
958 DUMPREG(US3_R1);
959 DUMPREG(US3_R2);
960 DUMPREG(US3_R3);
961 DUMPREG(US3_R4);
962 DUMPREG(US3_R5);
963 DUMPREG(US3_R6);
964 DUMPREG(US3_R7);
965 DUMPREG(DEI_FRAME_SIZE);
966 DUMPREG(MDT_BYPASS);
967 DUMPREG(MDT_SF_THRESHOLD);
968 DUMPREG(EDI_CONFIG);
969 DUMPREG(DEI_EDI_LUT_R0);
970 DUMPREG(DEI_EDI_LUT_R1);
971 DUMPREG(DEI_EDI_LUT_R2);
972 DUMPREG(DEI_EDI_LUT_R3);
973 DUMPREG(DEI_FMD_WINDOW_R0);
974 DUMPREG(DEI_FMD_WINDOW_R1);
975 DUMPREG(DEI_FMD_CONTROL_R0);
976 DUMPREG(DEI_FMD_CONTROL_R1);
977 DUMPREG(DEI_FMD_STATUS_R0);
978 DUMPREG(DEI_FMD_STATUS_R1);
979 DUMPREG(DEI_FMD_STATUS_R2);
980 #undef DUMPREG
982 sc_dump_regs(dev->sc);
983 csc_dump_regs(dev->csc);
986 static void add_out_dtd(struct vpe_ctx *ctx, int port)
988 struct vpe_q_data *q_data = &ctx->q_data[Q_DATA_DST];
989 const struct vpe_port_data *p_data = &port_data[port];
990 struct vb2_buffer *vb = &ctx->dst_vb->vb2_buf;
991 struct vpe_fmt *fmt = q_data->fmt;
992 const struct vpdma_data_format *vpdma_fmt;
993 int mv_buf_selector = !ctx->src_mv_buf_selector;
994 dma_addr_t dma_addr;
995 u32 flags = 0;
997 if (port == VPE_PORT_MV_OUT) {
998 vpdma_fmt = &vpdma_misc_fmts[VPDMA_DATA_FMT_MV];
999 dma_addr = ctx->mv_buf_dma[mv_buf_selector];
1000 } else {
1001 /* to incorporate interleaved formats */
1002 int plane = fmt->coplanar ? p_data->vb_part : 0;
1004 vpdma_fmt = fmt->vpdma_fmt[plane];
1005 dma_addr = vb2_dma_contig_plane_dma_addr(vb, plane);
1006 if (!dma_addr) {
1007 vpe_err(ctx->dev,
1008 "acquiring output buffer(%d) dma_addr failed\n",
1009 port);
1010 return;
1014 if (q_data->flags & Q_DATA_FRAME_1D)
1015 flags |= VPDMA_DATA_FRAME_1D;
1016 if (q_data->flags & Q_DATA_MODE_TILED)
1017 flags |= VPDMA_DATA_MODE_TILED;
1019 vpdma_add_out_dtd(&ctx->desc_list, q_data->width, &q_data->c_rect,
1020 vpdma_fmt, dma_addr, p_data->channel, flags);
1023 static void add_in_dtd(struct vpe_ctx *ctx, int port)
1025 struct vpe_q_data *q_data = &ctx->q_data[Q_DATA_SRC];
1026 const struct vpe_port_data *p_data = &port_data[port];
1027 struct vb2_buffer *vb = &ctx->src_vbs[p_data->vb_index]->vb2_buf;
1028 struct vb2_v4l2_buffer *vbuf = to_vb2_v4l2_buffer(vb);
1029 struct vpe_fmt *fmt = q_data->fmt;
1030 const struct vpdma_data_format *vpdma_fmt;
1031 int mv_buf_selector = ctx->src_mv_buf_selector;
1032 int field = vbuf->field == V4L2_FIELD_BOTTOM;
1033 int frame_width, frame_height;
1034 dma_addr_t dma_addr;
1035 u32 flags = 0;
1037 if (port == VPE_PORT_MV_IN) {
1038 vpdma_fmt = &vpdma_misc_fmts[VPDMA_DATA_FMT_MV];
1039 dma_addr = ctx->mv_buf_dma[mv_buf_selector];
1040 } else {
1041 /* to incorporate interleaved formats */
1042 int plane = fmt->coplanar ? p_data->vb_part : 0;
1044 vpdma_fmt = fmt->vpdma_fmt[plane];
1046 dma_addr = vb2_dma_contig_plane_dma_addr(vb, plane);
1047 if (!dma_addr) {
1048 vpe_err(ctx->dev,
1049 "acquiring input buffer(%d) dma_addr failed\n",
1050 port);
1051 return;
1055 if (q_data->flags & Q_DATA_FRAME_1D)
1056 flags |= VPDMA_DATA_FRAME_1D;
1057 if (q_data->flags & Q_DATA_MODE_TILED)
1058 flags |= VPDMA_DATA_MODE_TILED;
1060 frame_width = q_data->c_rect.width;
1061 frame_height = q_data->c_rect.height;
1063 if (p_data->vb_part && fmt->fourcc == V4L2_PIX_FMT_NV12)
1064 frame_height /= 2;
1066 vpdma_add_in_dtd(&ctx->desc_list, q_data->width, &q_data->c_rect,
1067 vpdma_fmt, dma_addr, p_data->channel, field, flags, frame_width,
1068 frame_height, 0, 0);
1072 * Enable the expected IRQ sources
1074 static void enable_irqs(struct vpe_ctx *ctx)
1076 write_reg(ctx->dev, VPE_INT0_ENABLE0_SET, VPE_INT0_LIST0_COMPLETE);
1077 write_reg(ctx->dev, VPE_INT0_ENABLE1_SET, VPE_DEI_ERROR_INT |
1078 VPE_DS1_UV_ERROR_INT);
1080 vpdma_enable_list_complete_irq(ctx->dev->vpdma, 0, true);
1083 static void disable_irqs(struct vpe_ctx *ctx)
1085 write_reg(ctx->dev, VPE_INT0_ENABLE0_CLR, 0xffffffff);
1086 write_reg(ctx->dev, VPE_INT0_ENABLE1_CLR, 0xffffffff);
1088 vpdma_enable_list_complete_irq(ctx->dev->vpdma, 0, false);
1091 /* device_run() - prepares and starts the device
1093 * This function is only called when both the source and destination
1094 * buffers are in place.
1096 static void device_run(void *priv)
1098 struct vpe_ctx *ctx = priv;
1099 struct sc_data *sc = ctx->dev->sc;
1100 struct vpe_q_data *d_q_data = &ctx->q_data[Q_DATA_DST];
1102 if (ctx->deinterlacing && ctx->src_vbs[2] == NULL) {
1103 ctx->src_vbs[2] = v4l2_m2m_src_buf_remove(ctx->fh.m2m_ctx);
1104 WARN_ON(ctx->src_vbs[2] == NULL);
1105 ctx->src_vbs[1] = v4l2_m2m_src_buf_remove(ctx->fh.m2m_ctx);
1106 WARN_ON(ctx->src_vbs[1] == NULL);
1109 ctx->src_vbs[0] = v4l2_m2m_src_buf_remove(ctx->fh.m2m_ctx);
1110 WARN_ON(ctx->src_vbs[0] == NULL);
1111 ctx->dst_vb = v4l2_m2m_dst_buf_remove(ctx->fh.m2m_ctx);
1112 WARN_ON(ctx->dst_vb == NULL);
1114 /* config descriptors */
1115 if (ctx->dev->loaded_mmrs != ctx->mmr_adb.dma_addr || ctx->load_mmrs) {
1116 vpdma_map_desc_buf(ctx->dev->vpdma, &ctx->mmr_adb);
1117 vpdma_add_cfd_adb(&ctx->desc_list, CFD_MMR_CLIENT, &ctx->mmr_adb);
1118 ctx->dev->loaded_mmrs = ctx->mmr_adb.dma_addr;
1119 ctx->load_mmrs = false;
1122 if (sc->loaded_coeff_h != ctx->sc_coeff_h.dma_addr ||
1123 sc->load_coeff_h) {
1124 vpdma_map_desc_buf(ctx->dev->vpdma, &ctx->sc_coeff_h);
1125 vpdma_add_cfd_block(&ctx->desc_list, CFD_SC_CLIENT,
1126 &ctx->sc_coeff_h, 0);
1128 sc->loaded_coeff_h = ctx->sc_coeff_h.dma_addr;
1129 sc->load_coeff_h = false;
1132 if (sc->loaded_coeff_v != ctx->sc_coeff_v.dma_addr ||
1133 sc->load_coeff_v) {
1134 vpdma_map_desc_buf(ctx->dev->vpdma, &ctx->sc_coeff_v);
1135 vpdma_add_cfd_block(&ctx->desc_list, CFD_SC_CLIENT,
1136 &ctx->sc_coeff_v, SC_COEF_SRAM_SIZE >> 4);
1138 sc->loaded_coeff_v = ctx->sc_coeff_v.dma_addr;
1139 sc->load_coeff_v = false;
1142 /* output data descriptors */
1143 if (ctx->deinterlacing)
1144 add_out_dtd(ctx, VPE_PORT_MV_OUT);
1146 if (d_q_data->colorspace == V4L2_COLORSPACE_SRGB) {
1147 add_out_dtd(ctx, VPE_PORT_RGB_OUT);
1148 } else {
1149 add_out_dtd(ctx, VPE_PORT_LUMA_OUT);
1150 if (d_q_data->fmt->coplanar)
1151 add_out_dtd(ctx, VPE_PORT_CHROMA_OUT);
1154 /* input data descriptors */
1155 if (ctx->deinterlacing) {
1156 add_in_dtd(ctx, VPE_PORT_LUMA3_IN);
1157 add_in_dtd(ctx, VPE_PORT_CHROMA3_IN);
1159 add_in_dtd(ctx, VPE_PORT_LUMA2_IN);
1160 add_in_dtd(ctx, VPE_PORT_CHROMA2_IN);
1163 add_in_dtd(ctx, VPE_PORT_LUMA1_IN);
1164 add_in_dtd(ctx, VPE_PORT_CHROMA1_IN);
1166 if (ctx->deinterlacing)
1167 add_in_dtd(ctx, VPE_PORT_MV_IN);
1169 /* sync on channel control descriptors for input ports */
1170 vpdma_add_sync_on_channel_ctd(&ctx->desc_list, VPE_CHAN_LUMA1_IN);
1171 vpdma_add_sync_on_channel_ctd(&ctx->desc_list, VPE_CHAN_CHROMA1_IN);
1173 if (ctx->deinterlacing) {
1174 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1175 VPE_CHAN_LUMA2_IN);
1176 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1177 VPE_CHAN_CHROMA2_IN);
1179 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1180 VPE_CHAN_LUMA3_IN);
1181 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1182 VPE_CHAN_CHROMA3_IN);
1184 vpdma_add_sync_on_channel_ctd(&ctx->desc_list, VPE_CHAN_MV_IN);
1187 /* sync on channel control descriptors for output ports */
1188 if (d_q_data->colorspace == V4L2_COLORSPACE_SRGB) {
1189 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1190 VPE_CHAN_RGB_OUT);
1191 } else {
1192 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1193 VPE_CHAN_LUMA_OUT);
1194 if (d_q_data->fmt->coplanar)
1195 vpdma_add_sync_on_channel_ctd(&ctx->desc_list,
1196 VPE_CHAN_CHROMA_OUT);
1199 if (ctx->deinterlacing)
1200 vpdma_add_sync_on_channel_ctd(&ctx->desc_list, VPE_CHAN_MV_OUT);
1202 enable_irqs(ctx);
1204 vpdma_map_desc_buf(ctx->dev->vpdma, &ctx->desc_list.buf);
1205 vpdma_submit_descs(ctx->dev->vpdma, &ctx->desc_list);
1208 static void dei_error(struct vpe_ctx *ctx)
1210 dev_warn(ctx->dev->v4l2_dev.dev,
1211 "received DEI error interrupt\n");
1214 static void ds1_uv_error(struct vpe_ctx *ctx)
1216 dev_warn(ctx->dev->v4l2_dev.dev,
1217 "received downsampler error interrupt\n");
1220 static irqreturn_t vpe_irq(int irq_vpe, void *data)
1222 struct vpe_dev *dev = (struct vpe_dev *)data;
1223 struct vpe_ctx *ctx;
1224 struct vpe_q_data *d_q_data;
1225 struct vb2_v4l2_buffer *s_vb, *d_vb;
1226 unsigned long flags;
1227 u32 irqst0, irqst1;
1229 irqst0 = read_reg(dev, VPE_INT0_STATUS0);
1230 if (irqst0) {
1231 write_reg(dev, VPE_INT0_STATUS0_CLR, irqst0);
1232 vpe_dbg(dev, "INT0_STATUS0 = 0x%08x\n", irqst0);
1235 irqst1 = read_reg(dev, VPE_INT0_STATUS1);
1236 if (irqst1) {
1237 write_reg(dev, VPE_INT0_STATUS1_CLR, irqst1);
1238 vpe_dbg(dev, "INT0_STATUS1 = 0x%08x\n", irqst1);
1241 ctx = v4l2_m2m_get_curr_priv(dev->m2m_dev);
1242 if (!ctx) {
1243 vpe_err(dev, "instance released before end of transaction\n");
1244 goto handled;
1247 if (irqst1) {
1248 if (irqst1 & VPE_DEI_ERROR_INT) {
1249 irqst1 &= ~VPE_DEI_ERROR_INT;
1250 dei_error(ctx);
1252 if (irqst1 & VPE_DS1_UV_ERROR_INT) {
1253 irqst1 &= ~VPE_DS1_UV_ERROR_INT;
1254 ds1_uv_error(ctx);
1258 if (irqst0) {
1259 if (irqst0 & VPE_INT0_LIST0_COMPLETE)
1260 vpdma_clear_list_stat(ctx->dev->vpdma);
1262 irqst0 &= ~(VPE_INT0_LIST0_COMPLETE);
1265 if (irqst0 | irqst1) {
1266 dev_warn(dev->v4l2_dev.dev, "Unexpected interrupt: "
1267 "INT0_STATUS0 = 0x%08x, INT0_STATUS1 = 0x%08x\n",
1268 irqst0, irqst1);
1271 disable_irqs(ctx);
1273 vpdma_unmap_desc_buf(dev->vpdma, &ctx->desc_list.buf);
1274 vpdma_unmap_desc_buf(dev->vpdma, &ctx->mmr_adb);
1275 vpdma_unmap_desc_buf(dev->vpdma, &ctx->sc_coeff_h);
1276 vpdma_unmap_desc_buf(dev->vpdma, &ctx->sc_coeff_v);
1278 vpdma_reset_desc_list(&ctx->desc_list);
1280 /* the previous dst mv buffer becomes the next src mv buffer */
1281 ctx->src_mv_buf_selector = !ctx->src_mv_buf_selector;
1283 if (ctx->aborting)
1284 goto finished;
1286 s_vb = ctx->src_vbs[0];
1287 d_vb = ctx->dst_vb;
1289 d_vb->flags = s_vb->flags;
1290 d_vb->vb2_buf.timestamp = s_vb->vb2_buf.timestamp;
1292 if (s_vb->flags & V4L2_BUF_FLAG_TIMECODE)
1293 d_vb->timecode = s_vb->timecode;
1295 d_vb->sequence = ctx->sequence;
1297 d_q_data = &ctx->q_data[Q_DATA_DST];
1298 if (d_q_data->flags & Q_DATA_INTERLACED) {
1299 d_vb->field = ctx->field;
1300 if (ctx->field == V4L2_FIELD_BOTTOM) {
1301 ctx->sequence++;
1302 ctx->field = V4L2_FIELD_TOP;
1303 } else {
1304 WARN_ON(ctx->field != V4L2_FIELD_TOP);
1305 ctx->field = V4L2_FIELD_BOTTOM;
1307 } else {
1308 d_vb->field = V4L2_FIELD_NONE;
1309 ctx->sequence++;
1312 if (ctx->deinterlacing)
1313 s_vb = ctx->src_vbs[2];
1315 spin_lock_irqsave(&dev->lock, flags);
1316 v4l2_m2m_buf_done(s_vb, VB2_BUF_STATE_DONE);
1317 v4l2_m2m_buf_done(d_vb, VB2_BUF_STATE_DONE);
1318 spin_unlock_irqrestore(&dev->lock, flags);
1320 if (ctx->deinterlacing) {
1321 ctx->src_vbs[2] = ctx->src_vbs[1];
1322 ctx->src_vbs[1] = ctx->src_vbs[0];
1325 ctx->bufs_completed++;
1326 if (ctx->bufs_completed < ctx->bufs_per_job) {
1327 device_run(ctx);
1328 goto handled;
1331 finished:
1332 vpe_dbg(ctx->dev, "finishing transaction\n");
1333 ctx->bufs_completed = 0;
1334 v4l2_m2m_job_finish(dev->m2m_dev, ctx->fh.m2m_ctx);
1335 handled:
1336 return IRQ_HANDLED;
1340 * video ioctls
1342 static int vpe_querycap(struct file *file, void *priv,
1343 struct v4l2_capability *cap)
1345 strncpy(cap->driver, VPE_MODULE_NAME, sizeof(cap->driver) - 1);
1346 strncpy(cap->card, VPE_MODULE_NAME, sizeof(cap->card) - 1);
1347 snprintf(cap->bus_info, sizeof(cap->bus_info), "platform:%s",
1348 VPE_MODULE_NAME);
1349 cap->device_caps = V4L2_CAP_VIDEO_M2M_MPLANE | V4L2_CAP_STREAMING;
1350 cap->capabilities = cap->device_caps | V4L2_CAP_DEVICE_CAPS;
1351 return 0;
1354 static int __enum_fmt(struct v4l2_fmtdesc *f, u32 type)
1356 int i, index;
1357 struct vpe_fmt *fmt = NULL;
1359 index = 0;
1360 for (i = 0; i < ARRAY_SIZE(vpe_formats); ++i) {
1361 if (vpe_formats[i].types & type) {
1362 if (index == f->index) {
1363 fmt = &vpe_formats[i];
1364 break;
1366 index++;
1370 if (!fmt)
1371 return -EINVAL;
1373 strncpy(f->description, fmt->name, sizeof(f->description) - 1);
1374 f->pixelformat = fmt->fourcc;
1375 return 0;
1378 static int vpe_enum_fmt(struct file *file, void *priv,
1379 struct v4l2_fmtdesc *f)
1381 if (V4L2_TYPE_IS_OUTPUT(f->type))
1382 return __enum_fmt(f, VPE_FMT_TYPE_OUTPUT);
1384 return __enum_fmt(f, VPE_FMT_TYPE_CAPTURE);
1387 static int vpe_g_fmt(struct file *file, void *priv, struct v4l2_format *f)
1389 struct v4l2_pix_format_mplane *pix = &f->fmt.pix_mp;
1390 struct vpe_ctx *ctx = file2ctx(file);
1391 struct vb2_queue *vq;
1392 struct vpe_q_data *q_data;
1393 int i;
1395 vq = v4l2_m2m_get_vq(ctx->fh.m2m_ctx, f->type);
1396 if (!vq)
1397 return -EINVAL;
1399 q_data = get_q_data(ctx, f->type);
1401 pix->width = q_data->width;
1402 pix->height = q_data->height;
1403 pix->pixelformat = q_data->fmt->fourcc;
1404 pix->field = q_data->field;
1406 if (V4L2_TYPE_IS_OUTPUT(f->type)) {
1407 pix->colorspace = q_data->colorspace;
1408 } else {
1409 struct vpe_q_data *s_q_data;
1411 /* get colorspace from the source queue */
1412 s_q_data = get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE);
1414 pix->colorspace = s_q_data->colorspace;
1417 pix->num_planes = q_data->fmt->coplanar ? 2 : 1;
1419 for (i = 0; i < pix->num_planes; i++) {
1420 pix->plane_fmt[i].bytesperline = q_data->bytesperline[i];
1421 pix->plane_fmt[i].sizeimage = q_data->sizeimage[i];
1424 return 0;
1427 static int __vpe_try_fmt(struct vpe_ctx *ctx, struct v4l2_format *f,
1428 struct vpe_fmt *fmt, int type)
1430 struct v4l2_pix_format_mplane *pix = &f->fmt.pix_mp;
1431 struct v4l2_plane_pix_format *plane_fmt;
1432 unsigned int w_align;
1433 int i, depth, depth_bytes;
1435 if (!fmt || !(fmt->types & type)) {
1436 vpe_err(ctx->dev, "Fourcc format (0x%08x) invalid.\n",
1437 pix->pixelformat);
1438 return -EINVAL;
1441 if (pix->field != V4L2_FIELD_NONE && pix->field != V4L2_FIELD_ALTERNATE)
1442 pix->field = V4L2_FIELD_NONE;
1444 depth = fmt->vpdma_fmt[VPE_LUMA]->depth;
1447 * the line stride should 16 byte aligned for VPDMA to work, based on
1448 * the bytes per pixel, figure out how much the width should be aligned
1449 * to make sure line stride is 16 byte aligned
1451 depth_bytes = depth >> 3;
1453 if (depth_bytes == 3)
1455 * if bpp is 3(as in some RGB formats), the pixel width doesn't
1456 * really help in ensuring line stride is 16 byte aligned
1458 w_align = 4;
1459 else
1461 * for the remainder bpp(4, 2 and 1), the pixel width alignment
1462 * can ensure a line stride alignment of 16 bytes. For example,
1463 * if bpp is 2, then the line stride can be 16 byte aligned if
1464 * the width is 8 byte aligned
1466 w_align = order_base_2(VPDMA_DESC_ALIGN / depth_bytes);
1468 v4l_bound_align_image(&pix->width, MIN_W, MAX_W, w_align,
1469 &pix->height, MIN_H, MAX_H, H_ALIGN,
1470 S_ALIGN);
1472 pix->num_planes = fmt->coplanar ? 2 : 1;
1473 pix->pixelformat = fmt->fourcc;
1475 if (!pix->colorspace) {
1476 if (fmt->fourcc == V4L2_PIX_FMT_RGB24 ||
1477 fmt->fourcc == V4L2_PIX_FMT_BGR24 ||
1478 fmt->fourcc == V4L2_PIX_FMT_RGB32 ||
1479 fmt->fourcc == V4L2_PIX_FMT_BGR32) {
1480 pix->colorspace = V4L2_COLORSPACE_SRGB;
1481 } else {
1482 if (pix->height > 1280) /* HD */
1483 pix->colorspace = V4L2_COLORSPACE_REC709;
1484 else /* SD */
1485 pix->colorspace = V4L2_COLORSPACE_SMPTE170M;
1489 memset(pix->reserved, 0, sizeof(pix->reserved));
1490 for (i = 0; i < pix->num_planes; i++) {
1491 plane_fmt = &pix->plane_fmt[i];
1492 depth = fmt->vpdma_fmt[i]->depth;
1494 if (i == VPE_LUMA)
1495 plane_fmt->bytesperline = (pix->width * depth) >> 3;
1496 else
1497 plane_fmt->bytesperline = pix->width;
1499 plane_fmt->sizeimage =
1500 (pix->height * pix->width * depth) >> 3;
1502 memset(plane_fmt->reserved, 0, sizeof(plane_fmt->reserved));
1505 return 0;
1508 static int vpe_try_fmt(struct file *file, void *priv, struct v4l2_format *f)
1510 struct vpe_ctx *ctx = file2ctx(file);
1511 struct vpe_fmt *fmt = find_format(f);
1513 if (V4L2_TYPE_IS_OUTPUT(f->type))
1514 return __vpe_try_fmt(ctx, f, fmt, VPE_FMT_TYPE_OUTPUT);
1515 else
1516 return __vpe_try_fmt(ctx, f, fmt, VPE_FMT_TYPE_CAPTURE);
1519 static int __vpe_s_fmt(struct vpe_ctx *ctx, struct v4l2_format *f)
1521 struct v4l2_pix_format_mplane *pix = &f->fmt.pix_mp;
1522 struct v4l2_plane_pix_format *plane_fmt;
1523 struct vpe_q_data *q_data;
1524 struct vb2_queue *vq;
1525 int i;
1527 vq = v4l2_m2m_get_vq(ctx->fh.m2m_ctx, f->type);
1528 if (!vq)
1529 return -EINVAL;
1531 if (vb2_is_busy(vq)) {
1532 vpe_err(ctx->dev, "queue busy\n");
1533 return -EBUSY;
1536 q_data = get_q_data(ctx, f->type);
1537 if (!q_data)
1538 return -EINVAL;
1540 q_data->fmt = find_format(f);
1541 q_data->width = pix->width;
1542 q_data->height = pix->height;
1543 q_data->colorspace = pix->colorspace;
1544 q_data->field = pix->field;
1546 for (i = 0; i < pix->num_planes; i++) {
1547 plane_fmt = &pix->plane_fmt[i];
1549 q_data->bytesperline[i] = plane_fmt->bytesperline;
1550 q_data->sizeimage[i] = plane_fmt->sizeimage;
1553 q_data->c_rect.left = 0;
1554 q_data->c_rect.top = 0;
1555 q_data->c_rect.width = q_data->width;
1556 q_data->c_rect.height = q_data->height;
1558 if (q_data->field == V4L2_FIELD_ALTERNATE)
1559 q_data->flags |= Q_DATA_INTERLACED;
1560 else
1561 q_data->flags &= ~Q_DATA_INTERLACED;
1563 vpe_dbg(ctx->dev, "Setting format for type %d, wxh: %dx%d, fmt: %d bpl_y %d",
1564 f->type, q_data->width, q_data->height, q_data->fmt->fourcc,
1565 q_data->bytesperline[VPE_LUMA]);
1566 if (q_data->fmt->coplanar)
1567 vpe_dbg(ctx->dev, " bpl_uv %d\n",
1568 q_data->bytesperline[VPE_CHROMA]);
1570 return 0;
1573 static int vpe_s_fmt(struct file *file, void *priv, struct v4l2_format *f)
1575 int ret;
1576 struct vpe_ctx *ctx = file2ctx(file);
1578 ret = vpe_try_fmt(file, priv, f);
1579 if (ret)
1580 return ret;
1582 ret = __vpe_s_fmt(ctx, f);
1583 if (ret)
1584 return ret;
1586 if (V4L2_TYPE_IS_OUTPUT(f->type))
1587 set_src_registers(ctx);
1588 else
1589 set_dst_registers(ctx);
1591 return set_srcdst_params(ctx);
1594 static int __vpe_try_selection(struct vpe_ctx *ctx, struct v4l2_selection *s)
1596 struct vpe_q_data *q_data;
1598 if ((s->type != V4L2_BUF_TYPE_VIDEO_CAPTURE) &&
1599 (s->type != V4L2_BUF_TYPE_VIDEO_OUTPUT))
1600 return -EINVAL;
1602 q_data = get_q_data(ctx, s->type);
1603 if (!q_data)
1604 return -EINVAL;
1606 switch (s->target) {
1607 case V4L2_SEL_TGT_COMPOSE:
1609 * COMPOSE target is only valid for capture buffer type, return
1610 * error for output buffer type
1612 if (s->type == V4L2_BUF_TYPE_VIDEO_OUTPUT)
1613 return -EINVAL;
1614 break;
1615 case V4L2_SEL_TGT_CROP:
1617 * CROP target is only valid for output buffer type, return
1618 * error for capture buffer type
1620 if (s->type == V4L2_BUF_TYPE_VIDEO_CAPTURE)
1621 return -EINVAL;
1622 break;
1624 * bound and default crop/compose targets are invalid targets to
1625 * try/set
1627 default:
1628 return -EINVAL;
1631 if (s->r.top < 0 || s->r.left < 0) {
1632 vpe_err(ctx->dev, "negative values for top and left\n");
1633 s->r.top = s->r.left = 0;
1636 v4l_bound_align_image(&s->r.width, MIN_W, q_data->width, 1,
1637 &s->r.height, MIN_H, q_data->height, H_ALIGN, S_ALIGN);
1639 /* adjust left/top if cropping rectangle is out of bounds */
1640 if (s->r.left + s->r.width > q_data->width)
1641 s->r.left = q_data->width - s->r.width;
1642 if (s->r.top + s->r.height > q_data->height)
1643 s->r.top = q_data->height - s->r.height;
1645 return 0;
1648 static int vpe_g_selection(struct file *file, void *fh,
1649 struct v4l2_selection *s)
1651 struct vpe_ctx *ctx = file2ctx(file);
1652 struct vpe_q_data *q_data;
1653 bool use_c_rect = false;
1655 if ((s->type != V4L2_BUF_TYPE_VIDEO_CAPTURE) &&
1656 (s->type != V4L2_BUF_TYPE_VIDEO_OUTPUT))
1657 return -EINVAL;
1659 q_data = get_q_data(ctx, s->type);
1660 if (!q_data)
1661 return -EINVAL;
1663 switch (s->target) {
1664 case V4L2_SEL_TGT_COMPOSE_DEFAULT:
1665 case V4L2_SEL_TGT_COMPOSE_BOUNDS:
1666 if (s->type == V4L2_BUF_TYPE_VIDEO_OUTPUT)
1667 return -EINVAL;
1668 break;
1669 case V4L2_SEL_TGT_CROP_BOUNDS:
1670 case V4L2_SEL_TGT_CROP_DEFAULT:
1671 if (s->type == V4L2_BUF_TYPE_VIDEO_CAPTURE)
1672 return -EINVAL;
1673 break;
1674 case V4L2_SEL_TGT_COMPOSE:
1675 if (s->type == V4L2_BUF_TYPE_VIDEO_OUTPUT)
1676 return -EINVAL;
1677 use_c_rect = true;
1678 break;
1679 case V4L2_SEL_TGT_CROP:
1680 if (s->type == V4L2_BUF_TYPE_VIDEO_CAPTURE)
1681 return -EINVAL;
1682 use_c_rect = true;
1683 break;
1684 default:
1685 return -EINVAL;
1688 if (use_c_rect) {
1690 * for CROP/COMPOSE target type, return c_rect params from the
1691 * respective buffer type
1693 s->r = q_data->c_rect;
1694 } else {
1696 * for DEFAULT/BOUNDS target type, return width and height from
1697 * S_FMT of the respective buffer type
1699 s->r.left = 0;
1700 s->r.top = 0;
1701 s->r.width = q_data->width;
1702 s->r.height = q_data->height;
1705 return 0;
1709 static int vpe_s_selection(struct file *file, void *fh,
1710 struct v4l2_selection *s)
1712 struct vpe_ctx *ctx = file2ctx(file);
1713 struct vpe_q_data *q_data;
1714 struct v4l2_selection sel = *s;
1715 int ret;
1717 ret = __vpe_try_selection(ctx, &sel);
1718 if (ret)
1719 return ret;
1721 q_data = get_q_data(ctx, sel.type);
1722 if (!q_data)
1723 return -EINVAL;
1725 if ((q_data->c_rect.left == sel.r.left) &&
1726 (q_data->c_rect.top == sel.r.top) &&
1727 (q_data->c_rect.width == sel.r.width) &&
1728 (q_data->c_rect.height == sel.r.height)) {
1729 vpe_dbg(ctx->dev,
1730 "requested crop/compose values are already set\n");
1731 return 0;
1734 q_data->c_rect = sel.r;
1736 return set_srcdst_params(ctx);
1740 * defines number of buffers/frames a context can process with VPE before
1741 * switching to a different context. default value is 1 buffer per context
1743 #define V4L2_CID_VPE_BUFS_PER_JOB (V4L2_CID_USER_TI_VPE_BASE + 0)
1745 static int vpe_s_ctrl(struct v4l2_ctrl *ctrl)
1747 struct vpe_ctx *ctx =
1748 container_of(ctrl->handler, struct vpe_ctx, hdl);
1750 switch (ctrl->id) {
1751 case V4L2_CID_VPE_BUFS_PER_JOB:
1752 ctx->bufs_per_job = ctrl->val;
1753 break;
1755 default:
1756 vpe_err(ctx->dev, "Invalid control\n");
1757 return -EINVAL;
1760 return 0;
1763 static const struct v4l2_ctrl_ops vpe_ctrl_ops = {
1764 .s_ctrl = vpe_s_ctrl,
1767 static const struct v4l2_ioctl_ops vpe_ioctl_ops = {
1768 .vidioc_querycap = vpe_querycap,
1770 .vidioc_enum_fmt_vid_cap_mplane = vpe_enum_fmt,
1771 .vidioc_g_fmt_vid_cap_mplane = vpe_g_fmt,
1772 .vidioc_try_fmt_vid_cap_mplane = vpe_try_fmt,
1773 .vidioc_s_fmt_vid_cap_mplane = vpe_s_fmt,
1775 .vidioc_enum_fmt_vid_out_mplane = vpe_enum_fmt,
1776 .vidioc_g_fmt_vid_out_mplane = vpe_g_fmt,
1777 .vidioc_try_fmt_vid_out_mplane = vpe_try_fmt,
1778 .vidioc_s_fmt_vid_out_mplane = vpe_s_fmt,
1780 .vidioc_g_selection = vpe_g_selection,
1781 .vidioc_s_selection = vpe_s_selection,
1783 .vidioc_reqbufs = v4l2_m2m_ioctl_reqbufs,
1784 .vidioc_querybuf = v4l2_m2m_ioctl_querybuf,
1785 .vidioc_qbuf = v4l2_m2m_ioctl_qbuf,
1786 .vidioc_dqbuf = v4l2_m2m_ioctl_dqbuf,
1787 .vidioc_streamon = v4l2_m2m_ioctl_streamon,
1788 .vidioc_streamoff = v4l2_m2m_ioctl_streamoff,
1790 .vidioc_subscribe_event = v4l2_ctrl_subscribe_event,
1791 .vidioc_unsubscribe_event = v4l2_event_unsubscribe,
1795 * Queue operations
1797 static int vpe_queue_setup(struct vb2_queue *vq,
1798 unsigned int *nbuffers, unsigned int *nplanes,
1799 unsigned int sizes[], struct device *alloc_devs[])
1801 int i;
1802 struct vpe_ctx *ctx = vb2_get_drv_priv(vq);
1803 struct vpe_q_data *q_data;
1805 q_data = get_q_data(ctx, vq->type);
1807 *nplanes = q_data->fmt->coplanar ? 2 : 1;
1809 for (i = 0; i < *nplanes; i++)
1810 sizes[i] = q_data->sizeimage[i];
1812 vpe_dbg(ctx->dev, "get %d buffer(s) of size %d", *nbuffers,
1813 sizes[VPE_LUMA]);
1814 if (q_data->fmt->coplanar)
1815 vpe_dbg(ctx->dev, " and %d\n", sizes[VPE_CHROMA]);
1817 return 0;
1820 static int vpe_buf_prepare(struct vb2_buffer *vb)
1822 struct vb2_v4l2_buffer *vbuf = to_vb2_v4l2_buffer(vb);
1823 struct vpe_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
1824 struct vpe_q_data *q_data;
1825 int i, num_planes;
1827 vpe_dbg(ctx->dev, "type: %d\n", vb->vb2_queue->type);
1829 q_data = get_q_data(ctx, vb->vb2_queue->type);
1830 num_planes = q_data->fmt->coplanar ? 2 : 1;
1832 if (vb->vb2_queue->type == V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE) {
1833 if (!(q_data->flags & Q_DATA_INTERLACED)) {
1834 vbuf->field = V4L2_FIELD_NONE;
1835 } else {
1836 if (vbuf->field != V4L2_FIELD_TOP &&
1837 vbuf->field != V4L2_FIELD_BOTTOM)
1838 return -EINVAL;
1842 for (i = 0; i < num_planes; i++) {
1843 if (vb2_plane_size(vb, i) < q_data->sizeimage[i]) {
1844 vpe_err(ctx->dev,
1845 "data will not fit into plane (%lu < %lu)\n",
1846 vb2_plane_size(vb, i),
1847 (long) q_data->sizeimage[i]);
1848 return -EINVAL;
1852 for (i = 0; i < num_planes; i++)
1853 vb2_set_plane_payload(vb, i, q_data->sizeimage[i]);
1855 return 0;
1858 static void vpe_buf_queue(struct vb2_buffer *vb)
1860 struct vb2_v4l2_buffer *vbuf = to_vb2_v4l2_buffer(vb);
1861 struct vpe_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
1863 v4l2_m2m_buf_queue(ctx->fh.m2m_ctx, vbuf);
1866 static int vpe_start_streaming(struct vb2_queue *q, unsigned int count)
1868 /* currently we do nothing here */
1870 return 0;
1873 static void vpe_stop_streaming(struct vb2_queue *q)
1875 struct vpe_ctx *ctx = vb2_get_drv_priv(q);
1877 vpe_dump_regs(ctx->dev);
1878 vpdma_dump_regs(ctx->dev->vpdma);
1881 static const struct vb2_ops vpe_qops = {
1882 .queue_setup = vpe_queue_setup,
1883 .buf_prepare = vpe_buf_prepare,
1884 .buf_queue = vpe_buf_queue,
1885 .wait_prepare = vb2_ops_wait_prepare,
1886 .wait_finish = vb2_ops_wait_finish,
1887 .start_streaming = vpe_start_streaming,
1888 .stop_streaming = vpe_stop_streaming,
1891 static int queue_init(void *priv, struct vb2_queue *src_vq,
1892 struct vb2_queue *dst_vq)
1894 struct vpe_ctx *ctx = priv;
1895 struct vpe_dev *dev = ctx->dev;
1896 int ret;
1898 memset(src_vq, 0, sizeof(*src_vq));
1899 src_vq->type = V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE;
1900 src_vq->io_modes = VB2_MMAP | VB2_DMABUF;
1901 src_vq->drv_priv = ctx;
1902 src_vq->buf_struct_size = sizeof(struct v4l2_m2m_buffer);
1903 src_vq->ops = &vpe_qops;
1904 src_vq->mem_ops = &vb2_dma_contig_memops;
1905 src_vq->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_COPY;
1906 src_vq->lock = &dev->dev_mutex;
1907 src_vq->dev = dev->v4l2_dev.dev;
1909 ret = vb2_queue_init(src_vq);
1910 if (ret)
1911 return ret;
1913 memset(dst_vq, 0, sizeof(*dst_vq));
1914 dst_vq->type = V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE;
1915 dst_vq->io_modes = VB2_MMAP | VB2_DMABUF;
1916 dst_vq->drv_priv = ctx;
1917 dst_vq->buf_struct_size = sizeof(struct v4l2_m2m_buffer);
1918 dst_vq->ops = &vpe_qops;
1919 dst_vq->mem_ops = &vb2_dma_contig_memops;
1920 dst_vq->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_COPY;
1921 dst_vq->lock = &dev->dev_mutex;
1922 dst_vq->dev = dev->v4l2_dev.dev;
1924 return vb2_queue_init(dst_vq);
1927 static const struct v4l2_ctrl_config vpe_bufs_per_job = {
1928 .ops = &vpe_ctrl_ops,
1929 .id = V4L2_CID_VPE_BUFS_PER_JOB,
1930 .name = "Buffers Per Transaction",
1931 .type = V4L2_CTRL_TYPE_INTEGER,
1932 .def = VPE_DEF_BUFS_PER_JOB,
1933 .min = 1,
1934 .max = VIDEO_MAX_FRAME,
1935 .step = 1,
1939 * File operations
1941 static int vpe_open(struct file *file)
1943 struct vpe_dev *dev = video_drvdata(file);
1944 struct vpe_q_data *s_q_data;
1945 struct v4l2_ctrl_handler *hdl;
1946 struct vpe_ctx *ctx;
1947 int ret;
1949 vpe_dbg(dev, "vpe_open\n");
1951 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
1952 if (!ctx)
1953 return -ENOMEM;
1955 ctx->dev = dev;
1957 if (mutex_lock_interruptible(&dev->dev_mutex)) {
1958 ret = -ERESTARTSYS;
1959 goto free_ctx;
1962 ret = vpdma_create_desc_list(&ctx->desc_list, VPE_DESC_LIST_SIZE,
1963 VPDMA_LIST_TYPE_NORMAL);
1964 if (ret != 0)
1965 goto unlock;
1967 ret = vpdma_alloc_desc_buf(&ctx->mmr_adb, sizeof(struct vpe_mmr_adb));
1968 if (ret != 0)
1969 goto free_desc_list;
1971 ret = vpdma_alloc_desc_buf(&ctx->sc_coeff_h, SC_COEF_SRAM_SIZE);
1972 if (ret != 0)
1973 goto free_mmr_adb;
1975 ret = vpdma_alloc_desc_buf(&ctx->sc_coeff_v, SC_COEF_SRAM_SIZE);
1976 if (ret != 0)
1977 goto free_sc_h;
1979 init_adb_hdrs(ctx);
1981 v4l2_fh_init(&ctx->fh, video_devdata(file));
1982 file->private_data = &ctx->fh;
1984 hdl = &ctx->hdl;
1985 v4l2_ctrl_handler_init(hdl, 1);
1986 v4l2_ctrl_new_custom(hdl, &vpe_bufs_per_job, NULL);
1987 if (hdl->error) {
1988 ret = hdl->error;
1989 goto exit_fh;
1991 ctx->fh.ctrl_handler = hdl;
1992 v4l2_ctrl_handler_setup(hdl);
1994 s_q_data = &ctx->q_data[Q_DATA_SRC];
1995 s_q_data->fmt = &vpe_formats[2];
1996 s_q_data->width = 1920;
1997 s_q_data->height = 1080;
1998 s_q_data->bytesperline[VPE_LUMA] = (s_q_data->width *
1999 s_q_data->fmt->vpdma_fmt[VPE_LUMA]->depth) >> 3;
2000 s_q_data->sizeimage[VPE_LUMA] = (s_q_data->bytesperline[VPE_LUMA] *
2001 s_q_data->height);
2002 s_q_data->colorspace = V4L2_COLORSPACE_REC709;
2003 s_q_data->field = V4L2_FIELD_NONE;
2004 s_q_data->c_rect.left = 0;
2005 s_q_data->c_rect.top = 0;
2006 s_q_data->c_rect.width = s_q_data->width;
2007 s_q_data->c_rect.height = s_q_data->height;
2008 s_q_data->flags = 0;
2010 ctx->q_data[Q_DATA_DST] = *s_q_data;
2012 set_dei_shadow_registers(ctx);
2013 set_src_registers(ctx);
2014 set_dst_registers(ctx);
2015 ret = set_srcdst_params(ctx);
2016 if (ret)
2017 goto exit_fh;
2019 ctx->fh.m2m_ctx = v4l2_m2m_ctx_init(dev->m2m_dev, ctx, &queue_init);
2021 if (IS_ERR(ctx->fh.m2m_ctx)) {
2022 ret = PTR_ERR(ctx->fh.m2m_ctx);
2023 goto exit_fh;
2026 v4l2_fh_add(&ctx->fh);
2029 * for now, just report the creation of the first instance, we can later
2030 * optimize the driver to enable or disable clocks when the first
2031 * instance is created or the last instance released
2033 if (atomic_inc_return(&dev->num_instances) == 1)
2034 vpe_dbg(dev, "first instance created\n");
2036 ctx->bufs_per_job = VPE_DEF_BUFS_PER_JOB;
2038 ctx->load_mmrs = true;
2040 vpe_dbg(dev, "created instance %p, m2m_ctx: %p\n",
2041 ctx, ctx->fh.m2m_ctx);
2043 mutex_unlock(&dev->dev_mutex);
2045 return 0;
2046 exit_fh:
2047 v4l2_ctrl_handler_free(hdl);
2048 v4l2_fh_exit(&ctx->fh);
2049 vpdma_free_desc_buf(&ctx->sc_coeff_v);
2050 free_sc_h:
2051 vpdma_free_desc_buf(&ctx->sc_coeff_h);
2052 free_mmr_adb:
2053 vpdma_free_desc_buf(&ctx->mmr_adb);
2054 free_desc_list:
2055 vpdma_free_desc_list(&ctx->desc_list);
2056 unlock:
2057 mutex_unlock(&dev->dev_mutex);
2058 free_ctx:
2059 kfree(ctx);
2060 return ret;
2063 static int vpe_release(struct file *file)
2065 struct vpe_dev *dev = video_drvdata(file);
2066 struct vpe_ctx *ctx = file2ctx(file);
2068 vpe_dbg(dev, "releasing instance %p\n", ctx);
2070 mutex_lock(&dev->dev_mutex);
2071 free_vbs(ctx);
2072 free_mv_buffers(ctx);
2073 vpdma_free_desc_list(&ctx->desc_list);
2074 vpdma_free_desc_buf(&ctx->mmr_adb);
2076 v4l2_fh_del(&ctx->fh);
2077 v4l2_fh_exit(&ctx->fh);
2078 v4l2_ctrl_handler_free(&ctx->hdl);
2079 v4l2_m2m_ctx_release(ctx->fh.m2m_ctx);
2081 kfree(ctx);
2084 * for now, just report the release of the last instance, we can later
2085 * optimize the driver to enable or disable clocks when the first
2086 * instance is created or the last instance released
2088 if (atomic_dec_return(&dev->num_instances) == 0)
2089 vpe_dbg(dev, "last instance released\n");
2091 mutex_unlock(&dev->dev_mutex);
2093 return 0;
2096 static const struct v4l2_file_operations vpe_fops = {
2097 .owner = THIS_MODULE,
2098 .open = vpe_open,
2099 .release = vpe_release,
2100 .poll = v4l2_m2m_fop_poll,
2101 .unlocked_ioctl = video_ioctl2,
2102 .mmap = v4l2_m2m_fop_mmap,
2105 static struct video_device vpe_videodev = {
2106 .name = VPE_MODULE_NAME,
2107 .fops = &vpe_fops,
2108 .ioctl_ops = &vpe_ioctl_ops,
2109 .minor = -1,
2110 .release = video_device_release_empty,
2111 .vfl_dir = VFL_DIR_M2M,
2114 static struct v4l2_m2m_ops m2m_ops = {
2115 .device_run = device_run,
2116 .job_ready = job_ready,
2117 .job_abort = job_abort,
2118 .lock = vpe_lock,
2119 .unlock = vpe_unlock,
2122 static int vpe_runtime_get(struct platform_device *pdev)
2124 int r;
2126 dev_dbg(&pdev->dev, "vpe_runtime_get\n");
2128 r = pm_runtime_get_sync(&pdev->dev);
2129 WARN_ON(r < 0);
2130 return r < 0 ? r : 0;
2133 static void vpe_runtime_put(struct platform_device *pdev)
2136 int r;
2138 dev_dbg(&pdev->dev, "vpe_runtime_put\n");
2140 r = pm_runtime_put_sync(&pdev->dev);
2141 WARN_ON(r < 0 && r != -ENOSYS);
2144 static void vpe_fw_cb(struct platform_device *pdev)
2146 struct vpe_dev *dev = platform_get_drvdata(pdev);
2147 struct video_device *vfd;
2148 int ret;
2150 vfd = &dev->vfd;
2151 *vfd = vpe_videodev;
2152 vfd->lock = &dev->dev_mutex;
2153 vfd->v4l2_dev = &dev->v4l2_dev;
2155 ret = video_register_device(vfd, VFL_TYPE_GRABBER, 0);
2156 if (ret) {
2157 vpe_err(dev, "Failed to register video device\n");
2159 vpe_set_clock_enable(dev, 0);
2160 vpe_runtime_put(pdev);
2161 pm_runtime_disable(&pdev->dev);
2162 v4l2_m2m_release(dev->m2m_dev);
2163 v4l2_device_unregister(&dev->v4l2_dev);
2165 return;
2168 video_set_drvdata(vfd, dev);
2169 snprintf(vfd->name, sizeof(vfd->name), "%s", vpe_videodev.name);
2170 dev_info(dev->v4l2_dev.dev, "Device registered as /dev/video%d\n",
2171 vfd->num);
2174 static int vpe_probe(struct platform_device *pdev)
2176 struct vpe_dev *dev;
2177 int ret, irq, func;
2179 dev = devm_kzalloc(&pdev->dev, sizeof(*dev), GFP_KERNEL);
2180 if (!dev)
2181 return -ENOMEM;
2183 spin_lock_init(&dev->lock);
2185 ret = v4l2_device_register(&pdev->dev, &dev->v4l2_dev);
2186 if (ret)
2187 return ret;
2189 atomic_set(&dev->num_instances, 0);
2190 mutex_init(&dev->dev_mutex);
2192 dev->res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
2193 "vpe_top");
2195 * HACK: we get resource info from device tree in the form of a list of
2196 * VPE sub blocks, the driver currently uses only the base of vpe_top
2197 * for register access, the driver should be changed later to access
2198 * registers based on the sub block base addresses
2200 dev->base = devm_ioremap(&pdev->dev, dev->res->start, SZ_32K);
2201 if (!dev->base) {
2202 ret = -ENOMEM;
2203 goto v4l2_dev_unreg;
2206 irq = platform_get_irq(pdev, 0);
2207 ret = devm_request_irq(&pdev->dev, irq, vpe_irq, 0, VPE_MODULE_NAME,
2208 dev);
2209 if (ret)
2210 goto v4l2_dev_unreg;
2212 platform_set_drvdata(pdev, dev);
2214 dev->m2m_dev = v4l2_m2m_init(&m2m_ops);
2215 if (IS_ERR(dev->m2m_dev)) {
2216 vpe_err(dev, "Failed to init mem2mem device\n");
2217 ret = PTR_ERR(dev->m2m_dev);
2218 goto v4l2_dev_unreg;
2221 pm_runtime_enable(&pdev->dev);
2223 ret = vpe_runtime_get(pdev);
2224 if (ret)
2225 goto rel_m2m;
2227 /* Perform clk enable followed by reset */
2228 vpe_set_clock_enable(dev, 1);
2230 vpe_top_reset(dev);
2232 func = read_field_reg(dev, VPE_PID, VPE_PID_FUNC_MASK,
2233 VPE_PID_FUNC_SHIFT);
2234 vpe_dbg(dev, "VPE PID function %x\n", func);
2236 vpe_top_vpdma_reset(dev);
2238 dev->sc = sc_create(pdev);
2239 if (IS_ERR(dev->sc)) {
2240 ret = PTR_ERR(dev->sc);
2241 goto runtime_put;
2244 dev->csc = csc_create(pdev);
2245 if (IS_ERR(dev->csc)) {
2246 ret = PTR_ERR(dev->csc);
2247 goto runtime_put;
2250 dev->vpdma = vpdma_create(pdev, vpe_fw_cb);
2251 if (IS_ERR(dev->vpdma)) {
2252 ret = PTR_ERR(dev->vpdma);
2253 goto runtime_put;
2256 return 0;
2258 runtime_put:
2259 vpe_runtime_put(pdev);
2260 rel_m2m:
2261 pm_runtime_disable(&pdev->dev);
2262 v4l2_m2m_release(dev->m2m_dev);
2263 v4l2_dev_unreg:
2264 v4l2_device_unregister(&dev->v4l2_dev);
2266 return ret;
2269 static int vpe_remove(struct platform_device *pdev)
2271 struct vpe_dev *dev = platform_get_drvdata(pdev);
2273 v4l2_info(&dev->v4l2_dev, "Removing " VPE_MODULE_NAME);
2275 v4l2_m2m_release(dev->m2m_dev);
2276 video_unregister_device(&dev->vfd);
2277 v4l2_device_unregister(&dev->v4l2_dev);
2279 vpe_set_clock_enable(dev, 0);
2280 vpe_runtime_put(pdev);
2281 pm_runtime_disable(&pdev->dev);
2283 return 0;
2286 #if defined(CONFIG_OF)
2287 static const struct of_device_id vpe_of_match[] = {
2289 .compatible = "ti,vpe",
2293 #endif
2295 static struct platform_driver vpe_pdrv = {
2296 .probe = vpe_probe,
2297 .remove = vpe_remove,
2298 .driver = {
2299 .name = VPE_MODULE_NAME,
2300 .of_match_table = of_match_ptr(vpe_of_match),
2304 module_platform_driver(vpe_pdrv);
2306 MODULE_DESCRIPTION("TI VPE driver");
2307 MODULE_AUTHOR("Dale Farnsworth, <dale@farnsworth.org>");
2308 MODULE_LICENSE("GPL");