1 /*
2 * Ut Video decoder
3 * Copyright (c) 2011 Konstantin Shishkov
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 /**
23 * @file
24 * Ut Video decoder
25 */
26
27 #include <inttypes.h>
28 #include <stdlib.h>
29
30 #define CACHED_BITSTREAM_READER !ARCH_X86_32
31 #define UNCHECKED_BITSTREAM_READER 1
32
33 #include "libavutil/intreadwrite.h"
34 #include "libavutil/pixdesc.h"
35 #include "avcodec.h"
36 #include "bswapdsp.h"
37 #include "bytestream.h"
38 #include "get_bits.h"
39 #include "internal.h"
40 #include "thread.h"
41 #include "utvideo.h"
42
43 typedef struct HuffEntry {
44 uint8_t len;
45 uint16_t sym;
46 } HuffEntry;
47
build_huff(UtvideoContext * c,const uint8_t * src,VLC * vlc,int * fsym,unsigned nb_elems)48 static int build_huff(UtvideoContext *c, const uint8_t *src, VLC *vlc,
49 int *fsym, unsigned nb_elems)
50 {
51 int i;
52 HuffEntry he[1024];
53 uint8_t bits[1024];
54 uint16_t codes_count[33] = { 0 };
55
56 *fsym = -1;
57 for (i = 0; i < nb_elems; i++) {
58 if (src[i] == 0) {
59 *fsym = i;
60 return 0;
61 } else if (src[i] == 255) {
62 bits[i] = 0;
63 } else if (src[i] <= 32) {
64 bits[i] = src[i];
65 } else
66 return AVERROR_INVALIDDATA;
67
68 codes_count[bits[i]]++;
69 }
70 if (codes_count[0] == nb_elems)
71 return AVERROR_INVALIDDATA;
72
73 /* For Ut Video, longer codes are to the left of the tree and
74 * for codes with the same length the symbol is descending from
75 * left to right. So after the next loop --codes_count[i] will
76 * be the index of the first (lowest) symbol of length i when
77 * indexed by the position in the tree with left nodes being first. */
78 for (int i = 31; i >= 0; i--)
79 codes_count[i] += codes_count[i + 1];
80
81 for (unsigned i = 0; i < nb_elems; i++)
82 he[--codes_count[bits[i]]] = (HuffEntry) { bits[i], i };
83
84 #define VLC_BITS 11
85 return ff_init_vlc_from_lengths(vlc, VLC_BITS, codes_count[0],
86 &he[0].len, sizeof(*he),
87 &he[0].sym, sizeof(*he), 2, 0, 0, c->avctx);
88 }
89
decode_plane10(UtvideoContext * c,int plane_no,uint16_t * dst,ptrdiff_t stride,int width,int height,const uint8_t * src,const uint8_t * huff,int use_pred)90 static int decode_plane10(UtvideoContext *c, int plane_no,
91 uint16_t *dst, ptrdiff_t stride,
92 int width, int height,
93 const uint8_t *src, const uint8_t *huff,
94 int use_pred)
95 {
96 int i, j, slice, pix, ret;
97 int sstart, send;
98 VLC vlc;
99 GetBitContext gb;
100 int prev, fsym;
101
102 if ((ret = build_huff(c, huff, &vlc, &fsym, 1024)) < 0) {
103 av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n");
104 return ret;
105 }
106 if (fsym >= 0) { // build_huff reported a symbol to fill slices with
107 send = 0;
108 for (slice = 0; slice < c->slices; slice++) {
109 uint16_t *dest;
110
111 sstart = send;
112 send = (height * (slice + 1) / c->slices);
113 dest = dst + sstart * stride;
114
115 prev = 0x200;
116 for (j = sstart; j < send; j++) {
117 for (i = 0; i < width; i++) {
118 pix = fsym;
119 if (use_pred) {
120 prev += pix;
121 prev &= 0x3FF;
122 pix = prev;
123 }
124 dest[i] = pix;
125 }
126 dest += stride;
127 }
128 }
129 return 0;
130 }
131
132 send = 0;
133 for (slice = 0; slice < c->slices; slice++) {
134 uint16_t *dest;
135 int slice_data_start, slice_data_end, slice_size;
136
137 sstart = send;
138 send = (height * (slice + 1) / c->slices);
139 dest = dst + sstart * stride;
140
141 // slice offset and size validation was done earlier
142 slice_data_start = slice ? AV_RL32(src + slice * 4 - 4) : 0;
143 slice_data_end = AV_RL32(src + slice * 4);
144 slice_size = slice_data_end - slice_data_start;
145
146 if (!slice_size) {
147 av_log(c->avctx, AV_LOG_ERROR, "Plane has more than one symbol "
148 "yet a slice has a length of zero.\n");
149 goto fail;
150 }
151
152 memset(c->slice_bits + slice_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
153 c->bdsp.bswap_buf((uint32_t *) c->slice_bits,
154 (uint32_t *)(src + slice_data_start + c->slices * 4),
155 (slice_data_end - slice_data_start + 3) >> 2);
156 init_get_bits(&gb, c->slice_bits, slice_size * 8);
157
158 prev = 0x200;
159 for (j = sstart; j < send; j++) {
160 for (i = 0; i < width; i++) {
161 pix = get_vlc2(&gb, vlc.table, VLC_BITS, 3);
162 if (pix < 0) {
163 av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
164 goto fail;
165 }
166 if (use_pred) {
167 prev += pix;
168 prev &= 0x3FF;
169 pix = prev;
170 }
171 dest[i] = pix;
172 }
173 dest += stride;
174 if (get_bits_left(&gb) < 0) {
175 av_log(c->avctx, AV_LOG_ERROR,
176 "Slice decoding ran out of bits\n");
177 goto fail;
178 }
179 }
180 if (get_bits_left(&gb) > 32)
181 av_log(c->avctx, AV_LOG_WARNING,
182 "%d bits left after decoding slice\n", get_bits_left(&gb));
183 }
184
185 ff_free_vlc(&vlc);
186
187 return 0;
188 fail:
189 ff_free_vlc(&vlc);
190 return AVERROR_INVALIDDATA;
191 }
192
compute_cmask(int plane_no,int interlaced,enum AVPixelFormat pix_fmt)193 static int compute_cmask(int plane_no, int interlaced, enum AVPixelFormat pix_fmt)
194 {
195 const int is_luma = (pix_fmt == AV_PIX_FMT_YUV420P) && !plane_no;
196
197 if (interlaced)
198 return ~(1 + 2 * is_luma);
199
200 return ~is_luma;
201 }
202
decode_plane(UtvideoContext * c,int plane_no,uint8_t * dst,ptrdiff_t stride,int width,int height,const uint8_t * src,int use_pred)203 static int decode_plane(UtvideoContext *c, int plane_no,
204 uint8_t *dst, ptrdiff_t stride,
205 int width, int height,
206 const uint8_t *src, int use_pred)
207 {
208 int i, j, slice, pix;
209 int sstart, send;
210 VLC vlc;
211 GetBitContext gb;
212 int ret, prev, fsym;
213 const int cmask = compute_cmask(plane_no, c->interlaced, c->avctx->pix_fmt);
214
215 if (c->pack) {
216 send = 0;
217 for (slice = 0; slice < c->slices; slice++) {
218 GetBitContext cbit, pbit;
219 uint8_t *dest, *p;
220
221 ret = init_get_bits8_le(&cbit, c->control_stream[plane_no][slice], c->control_stream_size[plane_no][slice]);
222 if (ret < 0)
223 return ret;
224
225 ret = init_get_bits8_le(&pbit, c->packed_stream[plane_no][slice], c->packed_stream_size[plane_no][slice]);
226 if (ret < 0)
227 return ret;
228
229 sstart = send;
230 send = (height * (slice + 1) / c->slices) & cmask;
231 dest = dst + sstart * stride;
232
233 if (3 * ((dst + send * stride - dest + 7)/8) > get_bits_left(&cbit))
234 return AVERROR_INVALIDDATA;
235
236 for (p = dest; p < dst + send * stride; p += 8) {
237 int bits = get_bits_le(&cbit, 3);
238
239 if (bits == 0) {
240 *(uint64_t *) p = 0;
241 } else {
242 uint32_t sub = 0x80 >> (8 - (bits + 1)), add;
243 int k;
244
245 if ((bits + 1) * 8 > get_bits_left(&pbit))
246 return AVERROR_INVALIDDATA;
247
248 for (k = 0; k < 8; k++) {
249
250 p[k] = get_bits_le(&pbit, bits + 1);
251 add = (~p[k] & sub) << (8 - bits);
252 p[k] -= sub;
253 p[k] += add;
254 }
255 }
256 }
257 }
258
259 return 0;
260 }
261
262 if (build_huff(c, src, &vlc, &fsym, 256)) {
263 av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n");
264 return AVERROR_INVALIDDATA;
265 }
266 if (fsym >= 0) { // build_huff reported a symbol to fill slices with
267 send = 0;
268 for (slice = 0; slice < c->slices; slice++) {
269 uint8_t *dest;
270
271 sstart = send;
272 send = (height * (slice + 1) / c->slices) & cmask;
273 dest = dst + sstart * stride;
274
275 prev = 0x80;
276 for (j = sstart; j < send; j++) {
277 for (i = 0; i < width; i++) {
278 pix = fsym;
279 if (use_pred) {
280 prev += (unsigned)pix;
281 pix = prev;
282 }
283 dest[i] = pix;
284 }
285 dest += stride;
286 }
287 }
288 return 0;
289 }
290
291 src += 256;
292
293 send = 0;
294 for (slice = 0; slice < c->slices; slice++) {
295 uint8_t *dest;
296 int slice_data_start, slice_data_end, slice_size;
297
298 sstart = send;
299 send = (height * (slice + 1) / c->slices) & cmask;
300 dest = dst + sstart * stride;
301
302 // slice offset and size validation was done earlier
303 slice_data_start = slice ? AV_RL32(src + slice * 4 - 4) : 0;
304 slice_data_end = AV_RL32(src + slice * 4);
305 slice_size = slice_data_end - slice_data_start;
306
307 if (!slice_size) {
308 av_log(c->avctx, AV_LOG_ERROR, "Plane has more than one symbol "
309 "yet a slice has a length of zero.\n");
310 goto fail;
311 }
312
313 memset(c->slice_bits + slice_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
314 c->bdsp.bswap_buf((uint32_t *) c->slice_bits,
315 (uint32_t *)(src + slice_data_start + c->slices * 4),
316 (slice_data_end - slice_data_start + 3) >> 2);
317 init_get_bits(&gb, c->slice_bits, slice_size * 8);
318
319 prev = 0x80;
320 for (j = sstart; j < send; j++) {
321 for (i = 0; i < width; i++) {
322 pix = get_vlc2(&gb, vlc.table, VLC_BITS, 3);
323 if (pix < 0) {
324 av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
325 goto fail;
326 }
327 if (use_pred) {
328 prev += pix;
329 pix = prev;
330 }
331 dest[i] = pix;
332 }
333 if (get_bits_left(&gb) < 0) {
334 av_log(c->avctx, AV_LOG_ERROR,
335 "Slice decoding ran out of bits\n");
336 goto fail;
337 }
338 dest += stride;
339 }
340 if (get_bits_left(&gb) > 32)
341 av_log(c->avctx, AV_LOG_WARNING,
342 "%d bits left after decoding slice\n", get_bits_left(&gb));
343 }
344
345 ff_free_vlc(&vlc);
346
347 return 0;
348 fail:
349 ff_free_vlc(&vlc);
350 return AVERROR_INVALIDDATA;
351 }
352
353 #undef A
354 #undef B
355 #undef C
356
restore_median_planar(UtvideoContext * c,uint8_t * src,ptrdiff_t stride,int width,int height,int slices,int rmode)357 static void restore_median_planar(UtvideoContext *c, uint8_t *src, ptrdiff_t stride,
358 int width, int height, int slices, int rmode)
359 {
360 int i, j, slice;
361 int A, B, C;
362 uint8_t *bsrc;
363 int slice_start, slice_height;
364 const int cmask = ~rmode;
365
366 for (slice = 0; slice < slices; slice++) {
367 slice_start = ((slice * height) / slices) & cmask;
368 slice_height = ((((slice + 1) * height) / slices) & cmask) -
369 slice_start;
370
371 if (!slice_height)
372 continue;
373 bsrc = src + slice_start * stride;
374
375 // first line - left neighbour prediction
376 bsrc[0] += 0x80;
377 c->llviddsp.add_left_pred(bsrc, bsrc, width, 0);
378 bsrc += stride;
379 if (slice_height <= 1)
380 continue;
381 // second line - first element has top prediction, the rest uses median
382 C = bsrc[-stride];
383 bsrc[0] += C;
384 A = bsrc[0];
385 for (i = 1; i < FFMIN(width, 16); i++) { /* scalar loop (DSP need align 16) */
386 B = bsrc[i - stride];
387 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
388 C = B;
389 A = bsrc[i];
390 }
391 if (width > 16)
392 c->llviddsp.add_median_pred(bsrc + 16, bsrc - stride + 16,
393 bsrc + 16, width - 16, &A, &B);
394
395 bsrc += stride;
396 // the rest of lines use continuous median prediction
397 for (j = 2; j < slice_height; j++) {
398 c->llviddsp.add_median_pred(bsrc, bsrc - stride,
399 bsrc, width, &A, &B);
400 bsrc += stride;
401 }
402 }
403 }
404
405 /* UtVideo interlaced mode treats every two lines as a single one,
406 * so restoring function should take care of possible padding between
407 * two parts of the same "line".
408 */
restore_median_planar_il(UtvideoContext * c,uint8_t * src,ptrdiff_t stride,int width,int height,int slices,int rmode)409 static void restore_median_planar_il(UtvideoContext *c, uint8_t *src, ptrdiff_t stride,
410 int width, int height, int slices, int rmode)
411 {
412 int i, j, slice;
413 int A, B, C;
414 uint8_t *bsrc;
415 int slice_start, slice_height;
416 const int cmask = ~(rmode ? 3 : 1);
417 const ptrdiff_t stride2 = stride << 1;
418
419 for (slice = 0; slice < slices; slice++) {
420 slice_start = ((slice * height) / slices) & cmask;
421 slice_height = ((((slice + 1) * height) / slices) & cmask) -
422 slice_start;
423 slice_height >>= 1;
424 if (!slice_height)
425 continue;
426
427 bsrc = src + slice_start * stride;
428
429 // first line - left neighbour prediction
430 bsrc[0] += 0x80;
431 A = c->llviddsp.add_left_pred(bsrc, bsrc, width, 0);
432 c->llviddsp.add_left_pred(bsrc + stride, bsrc + stride, width, A);
433 bsrc += stride2;
434 if (slice_height <= 1)
435 continue;
436 // second line - first element has top prediction, the rest uses median
437 C = bsrc[-stride2];
438 bsrc[0] += C;
439 A = bsrc[0];
440 for (i = 1; i < FFMIN(width, 16); i++) { /* scalar loop (DSP need align 16) */
441 B = bsrc[i - stride2];
442 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
443 C = B;
444 A = bsrc[i];
445 }
446 if (width > 16)
447 c->llviddsp.add_median_pred(bsrc + 16, bsrc - stride2 + 16,
448 bsrc + 16, width - 16, &A, &B);
449
450 c->llviddsp.add_median_pred(bsrc + stride, bsrc - stride,
451 bsrc + stride, width, &A, &B);
452 bsrc += stride2;
453 // the rest of lines use continuous median prediction
454 for (j = 2; j < slice_height; j++) {
455 c->llviddsp.add_median_pred(bsrc, bsrc - stride2,
456 bsrc, width, &A, &B);
457 c->llviddsp.add_median_pred(bsrc + stride, bsrc - stride,
458 bsrc + stride, width, &A, &B);
459 bsrc += stride2;
460 }
461 }
462 }
463
restore_gradient_planar(UtvideoContext * c,uint8_t * src,ptrdiff_t stride,int width,int height,int slices,int rmode)464 static void restore_gradient_planar(UtvideoContext *c, uint8_t *src, ptrdiff_t stride,
465 int width, int height, int slices, int rmode)
466 {
467 int i, j, slice;
468 int A, B, C;
469 uint8_t *bsrc;
470 int slice_start, slice_height;
471 const int cmask = ~rmode;
472 int min_width = FFMIN(width, 32);
473
474 for (slice = 0; slice < slices; slice++) {
475 slice_start = ((slice * height) / slices) & cmask;
476 slice_height = ((((slice + 1) * height) / slices) & cmask) -
477 slice_start;
478
479 if (!slice_height)
480 continue;
481 bsrc = src + slice_start * stride;
482
483 // first line - left neighbour prediction
484 bsrc[0] += 0x80;
485 c->llviddsp.add_left_pred(bsrc, bsrc, width, 0);
486 bsrc += stride;
487 if (slice_height <= 1)
488 continue;
489 for (j = 1; j < slice_height; j++) {
490 // second line - first element has top prediction, the rest uses gradient
491 bsrc[0] = (bsrc[0] + bsrc[-stride]) & 0xFF;
492 for (i = 1; i < min_width; i++) { /* dsp need align 32 */
493 A = bsrc[i - stride];
494 B = bsrc[i - (stride + 1)];
495 C = bsrc[i - 1];
496 bsrc[i] = (A - B + C + bsrc[i]) & 0xFF;
497 }
498 if (width > 32)
499 c->llviddsp.add_gradient_pred(bsrc + 32, stride, width - 32);
500 bsrc += stride;
501 }
502 }
503 }
504
restore_gradient_planar_il(UtvideoContext * c,uint8_t * src,ptrdiff_t stride,int width,int height,int slices,int rmode)505 static void restore_gradient_planar_il(UtvideoContext *c, uint8_t *src, ptrdiff_t stride,
506 int width, int height, int slices, int rmode)
507 {
508 int i, j, slice;
509 int A, B, C;
510 uint8_t *bsrc;
511 int slice_start, slice_height;
512 const int cmask = ~(rmode ? 3 : 1);
513 const ptrdiff_t stride2 = stride << 1;
514 int min_width = FFMIN(width, 32);
515
516 for (slice = 0; slice < slices; slice++) {
517 slice_start = ((slice * height) / slices) & cmask;
518 slice_height = ((((slice + 1) * height) / slices) & cmask) -
519 slice_start;
520 slice_height >>= 1;
521 if (!slice_height)
522 continue;
523
524 bsrc = src + slice_start * stride;
525
526 // first line - left neighbour prediction
527 bsrc[0] += 0x80;
528 A = c->llviddsp.add_left_pred(bsrc, bsrc, width, 0);
529 c->llviddsp.add_left_pred(bsrc + stride, bsrc + stride, width, A);
530 bsrc += stride2;
531 if (slice_height <= 1)
532 continue;
533 for (j = 1; j < slice_height; j++) {
534 // second line - first element has top prediction, the rest uses gradient
535 bsrc[0] = (bsrc[0] + bsrc[-stride2]) & 0xFF;
536 for (i = 1; i < min_width; i++) { /* dsp need align 32 */
537 A = bsrc[i - stride2];
538 B = bsrc[i - (stride2 + 1)];
539 C = bsrc[i - 1];
540 bsrc[i] = (A - B + C + bsrc[i]) & 0xFF;
541 }
542 if (width > 32)
543 c->llviddsp.add_gradient_pred(bsrc + 32, stride2, width - 32);
544
545 A = bsrc[-stride];
546 B = bsrc[-(1 + stride + stride - width)];
547 C = bsrc[width - 1];
548 bsrc[stride] = (A - B + C + bsrc[stride]) & 0xFF;
549 for (i = 1; i < width; i++) {
550 A = bsrc[i - stride];
551 B = bsrc[i - (1 + stride)];
552 C = bsrc[i - 1 + stride];
553 bsrc[i + stride] = (A - B + C + bsrc[i + stride]) & 0xFF;
554 }
555 bsrc += stride2;
556 }
557 }
558 }
559
decode_frame(AVCodecContext * avctx,void * data,int * got_frame,AVPacket * avpkt)560 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
561 AVPacket *avpkt)
562 {
563 const uint8_t *buf = avpkt->data;
564 int buf_size = avpkt->size;
565 UtvideoContext *c = avctx->priv_data;
566 int i, j;
567 const uint8_t *plane_start[5];
568 int plane_size, max_slice_size = 0, slice_start, slice_end, slice_size;
569 int ret;
570 GetByteContext gb;
571 ThreadFrame frame = { .f = data };
572
573 if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0)
574 return ret;
575
576 /* parse plane structure to get frame flags and validate slice offsets */
577 bytestream2_init(&gb, buf, buf_size);
578
579 if (c->pack) {
580 const uint8_t *packed_stream;
581 const uint8_t *control_stream;
582 GetByteContext pb;
583 uint32_t nb_cbs;
584 int left;
585
586 c->frame_info = PRED_GRADIENT << 8;
587
588 if (bytestream2_get_byte(&gb) != 1)
589 return AVERROR_INVALIDDATA;
590 bytestream2_skip(&gb, 3);
591 c->offset = bytestream2_get_le32(&gb);
592
593 if (buf_size <= c->offset + 8LL)
594 return AVERROR_INVALIDDATA;
595
596 bytestream2_init(&pb, buf + 8 + c->offset, buf_size - 8 - c->offset);
597
598 nb_cbs = bytestream2_get_le32(&pb);
599 if (nb_cbs > c->offset)
600 return AVERROR_INVALIDDATA;
601
602 packed_stream = buf + 8;
603 control_stream = packed_stream + (c->offset - nb_cbs);
604 left = control_stream - packed_stream;
605
606 for (i = 0; i < c->planes; i++) {
607 for (j = 0; j < c->slices; j++) {
608 c->packed_stream[i][j] = packed_stream;
609 c->packed_stream_size[i][j] = bytestream2_get_le32(&pb);
610 if (c->packed_stream_size[i][j] > left)
611 return AVERROR_INVALIDDATA;
612 left -= c->packed_stream_size[i][j];
613 packed_stream += c->packed_stream_size[i][j];
614 }
615 }
616
617 left = buf + buf_size - control_stream;
618
619 for (i = 0; i < c->planes; i++) {
620 for (j = 0; j < c->slices; j++) {
621 c->control_stream[i][j] = control_stream;
622 c->control_stream_size[i][j] = bytestream2_get_le32(&pb);
623 if (c->control_stream_size[i][j] > left)
624 return AVERROR_INVALIDDATA;
625 left -= c->control_stream_size[i][j];
626 control_stream += c->control_stream_size[i][j];
627 }
628 }
629 } else if (c->pro) {
630 if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
631 av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
632 return AVERROR_INVALIDDATA;
633 }
634 c->frame_info = bytestream2_get_le32u(&gb);
635 c->slices = ((c->frame_info >> 16) & 0xff) + 1;
636 for (i = 0; i < c->planes; i++) {
637 plane_start[i] = gb.buffer;
638 if (bytestream2_get_bytes_left(&gb) < 1024 + 4 * c->slices) {
639 av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
640 return AVERROR_INVALIDDATA;
641 }
642 slice_start = 0;
643 slice_end = 0;
644 for (j = 0; j < c->slices; j++) {
645 slice_end = bytestream2_get_le32u(&gb);
646 if (slice_end < 0 || slice_end < slice_start ||
647 bytestream2_get_bytes_left(&gb) < slice_end + 1024LL) {
648 av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
649 return AVERROR_INVALIDDATA;
650 }
651 slice_size = slice_end - slice_start;
652 slice_start = slice_end;
653 max_slice_size = FFMAX(max_slice_size, slice_size);
654 }
655 plane_size = slice_end;
656 bytestream2_skipu(&gb, plane_size);
657 bytestream2_skipu(&gb, 1024);
658 }
659 plane_start[c->planes] = gb.buffer;
660 } else {
661 for (i = 0; i < c->planes; i++) {
662 plane_start[i] = gb.buffer;
663 if (bytestream2_get_bytes_left(&gb) < 256 + 4 * c->slices) {
664 av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
665 return AVERROR_INVALIDDATA;
666 }
667 bytestream2_skipu(&gb, 256);
668 slice_start = 0;
669 slice_end = 0;
670 for (j = 0; j < c->slices; j++) {
671 slice_end = bytestream2_get_le32u(&gb);
672 if (slice_end < 0 || slice_end < slice_start ||
673 bytestream2_get_bytes_left(&gb) < slice_end) {
674 av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
675 return AVERROR_INVALIDDATA;
676 }
677 slice_size = slice_end - slice_start;
678 slice_start = slice_end;
679 max_slice_size = FFMAX(max_slice_size, slice_size);
680 }
681 plane_size = slice_end;
682 bytestream2_skipu(&gb, plane_size);
683 }
684 plane_start[c->planes] = gb.buffer;
685 if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
686 av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
687 return AVERROR_INVALIDDATA;
688 }
689 c->frame_info = bytestream2_get_le32u(&gb);
690 }
691 av_log(avctx, AV_LOG_DEBUG, "frame information flags %"PRIX32"\n",
692 c->frame_info);
693
694 c->frame_pred = (c->frame_info >> 8) & 3;
695
696 max_slice_size += 4*avctx->width;
697
698 if (!c->pack) {
699 av_fast_malloc(&c->slice_bits, &c->slice_bits_size,
700 max_slice_size + AV_INPUT_BUFFER_PADDING_SIZE);
701
702 if (!c->slice_bits) {
703 av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer\n");
704 return AVERROR(ENOMEM);
705 }
706 }
707
708 switch (c->avctx->pix_fmt) {
709 case AV_PIX_FMT_GBRP:
710 case AV_PIX_FMT_GBRAP:
711 for (i = 0; i < c->planes; i++) {
712 ret = decode_plane(c, i, frame.f->data[i],
713 frame.f->linesize[i], avctx->width,
714 avctx->height, plane_start[i],
715 c->frame_pred == PRED_LEFT);
716 if (ret)
717 return ret;
718 if (c->frame_pred == PRED_MEDIAN) {
719 if (!c->interlaced) {
720 restore_median_planar(c, frame.f->data[i],
721 frame.f->linesize[i], avctx->width,
722 avctx->height, c->slices, 0);
723 } else {
724 restore_median_planar_il(c, frame.f->data[i],
725 frame.f->linesize[i],
726 avctx->width, avctx->height, c->slices,
727 0);
728 }
729 } else if (c->frame_pred == PRED_GRADIENT) {
730 if (!c->interlaced) {
731 restore_gradient_planar(c, frame.f->data[i],
732 frame.f->linesize[i], avctx->width,
733 avctx->height, c->slices, 0);
734 } else {
735 restore_gradient_planar_il(c, frame.f->data[i],
736 frame.f->linesize[i],
737 avctx->width, avctx->height, c->slices,
738 0);
739 }
740 }
741 }
742 c->utdsp.restore_rgb_planes(frame.f->data[2], frame.f->data[0], frame.f->data[1],
743 frame.f->linesize[2], frame.f->linesize[0], frame.f->linesize[1],
744 avctx->width, avctx->height);
745 break;
746 case AV_PIX_FMT_GBRAP10:
747 case AV_PIX_FMT_GBRP10:
748 for (i = 0; i < c->planes; i++) {
749 ret = decode_plane10(c, i, (uint16_t *)frame.f->data[i],
750 frame.f->linesize[i] / 2, avctx->width,
751 avctx->height, plane_start[i],
752 plane_start[i + 1] - 1024,
753 c->frame_pred == PRED_LEFT);
754 if (ret)
755 return ret;
756 }
757 c->utdsp.restore_rgb_planes10((uint16_t *)frame.f->data[2], (uint16_t *)frame.f->data[0], (uint16_t *)frame.f->data[1],
758 frame.f->linesize[2] / 2, frame.f->linesize[0] / 2, frame.f->linesize[1] / 2,
759 avctx->width, avctx->height);
760 break;
761 case AV_PIX_FMT_YUV420P:
762 for (i = 0; i < 3; i++) {
763 ret = decode_plane(c, i, frame.f->data[i], frame.f->linesize[i],
764 avctx->width >> !!i, avctx->height >> !!i,
765 plane_start[i], c->frame_pred == PRED_LEFT);
766 if (ret)
767 return ret;
768 if (c->frame_pred == PRED_MEDIAN) {
769 if (!c->interlaced) {
770 restore_median_planar(c, frame.f->data[i], frame.f->linesize[i],
771 avctx->width >> !!i, avctx->height >> !!i,
772 c->slices, !i);
773 } else {
774 restore_median_planar_il(c, frame.f->data[i], frame.f->linesize[i],
775 avctx->width >> !!i,
776 avctx->height >> !!i,
777 c->slices, !i);
778 }
779 } else if (c->frame_pred == PRED_GRADIENT) {
780 if (!c->interlaced) {
781 restore_gradient_planar(c, frame.f->data[i], frame.f->linesize[i],
782 avctx->width >> !!i, avctx->height >> !!i,
783 c->slices, !i);
784 } else {
785 restore_gradient_planar_il(c, frame.f->data[i], frame.f->linesize[i],
786 avctx->width >> !!i,
787 avctx->height >> !!i,
788 c->slices, !i);
789 }
790 }
791 }
792 break;
793 case AV_PIX_FMT_YUV422P:
794 for (i = 0; i < 3; i++) {
795 ret = decode_plane(c, i, frame.f->data[i], frame.f->linesize[i],
796 avctx->width >> !!i, avctx->height,
797 plane_start[i], c->frame_pred == PRED_LEFT);
798 if (ret)
799 return ret;
800 if (c->frame_pred == PRED_MEDIAN) {
801 if (!c->interlaced) {
802 restore_median_planar(c, frame.f->data[i], frame.f->linesize[i],
803 avctx->width >> !!i, avctx->height,
804 c->slices, 0);
805 } else {
806 restore_median_planar_il(c, frame.f->data[i], frame.f->linesize[i],
807 avctx->width >> !!i, avctx->height,
808 c->slices, 0);
809 }
810 } else if (c->frame_pred == PRED_GRADIENT) {
811 if (!c->interlaced) {
812 restore_gradient_planar(c, frame.f->data[i], frame.f->linesize[i],
813 avctx->width >> !!i, avctx->height,
814 c->slices, 0);
815 } else {
816 restore_gradient_planar_il(c, frame.f->data[i], frame.f->linesize[i],
817 avctx->width >> !!i, avctx->height,
818 c->slices, 0);
819 }
820 }
821 }
822 break;
823 case AV_PIX_FMT_YUV444P:
824 for (i = 0; i < 3; i++) {
825 ret = decode_plane(c, i, frame.f->data[i], frame.f->linesize[i],
826 avctx->width, avctx->height,
827 plane_start[i], c->frame_pred == PRED_LEFT);
828 if (ret)
829 return ret;
830 if (c->frame_pred == PRED_MEDIAN) {
831 if (!c->interlaced) {
832 restore_median_planar(c, frame.f->data[i], frame.f->linesize[i],
833 avctx->width, avctx->height,
834 c->slices, 0);
835 } else {
836 restore_median_planar_il(c, frame.f->data[i], frame.f->linesize[i],
837 avctx->width, avctx->height,
838 c->slices, 0);
839 }
840 } else if (c->frame_pred == PRED_GRADIENT) {
841 if (!c->interlaced) {
842 restore_gradient_planar(c, frame.f->data[i], frame.f->linesize[i],
843 avctx->width, avctx->height,
844 c->slices, 0);
845 } else {
846 restore_gradient_planar_il(c, frame.f->data[i], frame.f->linesize[i],
847 avctx->width, avctx->height,
848 c->slices, 0);
849 }
850 }
851 }
852 break;
853 case AV_PIX_FMT_YUV420P10:
854 for (i = 0; i < 3; i++) {
855 ret = decode_plane10(c, i, (uint16_t *)frame.f->data[i], frame.f->linesize[i] / 2,
856 avctx->width >> !!i, avctx->height >> !!i,
857 plane_start[i], plane_start[i + 1] - 1024, c->frame_pred == PRED_LEFT);
858 if (ret)
859 return ret;
860 }
861 break;
862 case AV_PIX_FMT_YUV422P10:
863 for (i = 0; i < 3; i++) {
864 ret = decode_plane10(c, i, (uint16_t *)frame.f->data[i], frame.f->linesize[i] / 2,
865 avctx->width >> !!i, avctx->height,
866 plane_start[i], plane_start[i + 1] - 1024, c->frame_pred == PRED_LEFT);
867 if (ret)
868 return ret;
869 }
870 break;
871 }
872
873 frame.f->key_frame = 1;
874 frame.f->pict_type = AV_PICTURE_TYPE_I;
875 frame.f->interlaced_frame = !!c->interlaced;
876
877 *got_frame = 1;
878
879 /* always report that the buffer was completely consumed */
880 return buf_size;
881 }
882
decode_init(AVCodecContext * avctx)883 static av_cold int decode_init(AVCodecContext *avctx)
884 {
885 UtvideoContext * const c = avctx->priv_data;
886 int h_shift, v_shift;
887
888 c->avctx = avctx;
889
890 ff_utvideodsp_init(&c->utdsp);
891 ff_bswapdsp_init(&c->bdsp);
892 ff_llviddsp_init(&c->llviddsp);
893
894 c->slice_bits_size = 0;
895
896 switch (avctx->codec_tag) {
897 case MKTAG('U', 'L', 'R', 'G'):
898 c->planes = 3;
899 avctx->pix_fmt = AV_PIX_FMT_GBRP;
900 break;
901 case MKTAG('U', 'L', 'R', 'A'):
902 c->planes = 4;
903 avctx->pix_fmt = AV_PIX_FMT_GBRAP;
904 break;
905 case MKTAG('U', 'L', 'Y', '0'):
906 c->planes = 3;
907 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
908 avctx->colorspace = AVCOL_SPC_BT470BG;
909 break;
910 case MKTAG('U', 'L', 'Y', '2'):
911 c->planes = 3;
912 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
913 avctx->colorspace = AVCOL_SPC_BT470BG;
914 break;
915 case MKTAG('U', 'L', 'Y', '4'):
916 c->planes = 3;
917 avctx->pix_fmt = AV_PIX_FMT_YUV444P;
918 avctx->colorspace = AVCOL_SPC_BT470BG;
919 break;
920 case MKTAG('U', 'Q', 'Y', '0'):
921 c->planes = 3;
922 c->pro = 1;
923 avctx->pix_fmt = AV_PIX_FMT_YUV420P10;
924 break;
925 case MKTAG('U', 'Q', 'Y', '2'):
926 c->planes = 3;
927 c->pro = 1;
928 avctx->pix_fmt = AV_PIX_FMT_YUV422P10;
929 break;
930 case MKTAG('U', 'Q', 'R', 'G'):
931 c->planes = 3;
932 c->pro = 1;
933 avctx->pix_fmt = AV_PIX_FMT_GBRP10;
934 break;
935 case MKTAG('U', 'Q', 'R', 'A'):
936 c->planes = 4;
937 c->pro = 1;
938 avctx->pix_fmt = AV_PIX_FMT_GBRAP10;
939 break;
940 case MKTAG('U', 'L', 'H', '0'):
941 c->planes = 3;
942 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
943 avctx->colorspace = AVCOL_SPC_BT709;
944 break;
945 case MKTAG('U', 'L', 'H', '2'):
946 c->planes = 3;
947 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
948 avctx->colorspace = AVCOL_SPC_BT709;
949 break;
950 case MKTAG('U', 'L', 'H', '4'):
951 c->planes = 3;
952 avctx->pix_fmt = AV_PIX_FMT_YUV444P;
953 avctx->colorspace = AVCOL_SPC_BT709;
954 break;
955 case MKTAG('U', 'M', 'Y', '2'):
956 c->planes = 3;
957 c->pack = 1;
958 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
959 avctx->colorspace = AVCOL_SPC_BT470BG;
960 break;
961 case MKTAG('U', 'M', 'H', '2'):
962 c->planes = 3;
963 c->pack = 1;
964 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
965 avctx->colorspace = AVCOL_SPC_BT709;
966 break;
967 case MKTAG('U', 'M', 'Y', '4'):
968 c->planes = 3;
969 c->pack = 1;
970 avctx->pix_fmt = AV_PIX_FMT_YUV444P;
971 avctx->colorspace = AVCOL_SPC_BT470BG;
972 break;
973 case MKTAG('U', 'M', 'H', '4'):
974 c->planes = 3;
975 c->pack = 1;
976 avctx->pix_fmt = AV_PIX_FMT_YUV444P;
977 avctx->colorspace = AVCOL_SPC_BT709;
978 break;
979 case MKTAG('U', 'M', 'R', 'G'):
980 c->planes = 3;
981 c->pack = 1;
982 avctx->pix_fmt = AV_PIX_FMT_GBRP;
983 break;
984 case MKTAG('U', 'M', 'R', 'A'):
985 c->planes = 4;
986 c->pack = 1;
987 avctx->pix_fmt = AV_PIX_FMT_GBRAP;
988 break;
989 default:
990 av_log(avctx, AV_LOG_ERROR, "Unknown Ut Video FOURCC provided (%08X)\n",
991 avctx->codec_tag);
992 return AVERROR_INVALIDDATA;
993 }
994
995 av_pix_fmt_get_chroma_sub_sample(avctx->pix_fmt, &h_shift, &v_shift);
996 if ((avctx->width & ((1<<h_shift)-1)) ||
997 (avctx->height & ((1<<v_shift)-1))) {
998 avpriv_request_sample(avctx, "Odd dimensions");
999 return AVERROR_PATCHWELCOME;
1000 }
1001
1002 if (c->pack && avctx->extradata_size >= 16) {
1003 av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
1004 avctx->extradata[3], avctx->extradata[2],
1005 avctx->extradata[1], avctx->extradata[0]);
1006 av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
1007 AV_RB32(avctx->extradata + 4));
1008 c->compression = avctx->extradata[8];
1009 if (c->compression != 2)
1010 avpriv_request_sample(avctx, "Unknown compression type");
1011 c->slices = avctx->extradata[9] + 1;
1012 } else if (!c->pro && avctx->extradata_size >= 16) {
1013 av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
1014 avctx->extradata[3], avctx->extradata[2],
1015 avctx->extradata[1], avctx->extradata[0]);
1016 av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
1017 AV_RB32(avctx->extradata + 4));
1018 c->frame_info_size = AV_RL32(avctx->extradata + 8);
1019 c->flags = AV_RL32(avctx->extradata + 12);
1020
1021 if (c->frame_info_size != 4)
1022 avpriv_request_sample(avctx, "Frame info not 4 bytes");
1023 av_log(avctx, AV_LOG_DEBUG, "Encoding parameters %08"PRIX32"\n", c->flags);
1024 c->slices = (c->flags >> 24) + 1;
1025 c->compression = c->flags & 1;
1026 c->interlaced = c->flags & 0x800;
1027 } else if (c->pro && avctx->extradata_size == 8) {
1028 av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
1029 avctx->extradata[3], avctx->extradata[2],
1030 avctx->extradata[1], avctx->extradata[0]);
1031 av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
1032 AV_RB32(avctx->extradata + 4));
1033 c->interlaced = 0;
1034 c->frame_info_size = 4;
1035 } else {
1036 av_log(avctx, AV_LOG_ERROR,
1037 "Insufficient extradata size %d, should be at least 16\n",
1038 avctx->extradata_size);
1039 return AVERROR_INVALIDDATA;
1040 }
1041
1042 return 0;
1043 }
1044
decode_end(AVCodecContext * avctx)1045 static av_cold int decode_end(AVCodecContext *avctx)
1046 {
1047 UtvideoContext * const c = avctx->priv_data;
1048
1049 av_freep(&c->slice_bits);
1050
1051 return 0;
1052 }
1053
1054 AVCodec ff_utvideo_decoder = {
1055 .name = "utvideo",
1056 .long_name = NULL_IF_CONFIG_SMALL("Ut Video"),
1057 .type = AVMEDIA_TYPE_VIDEO,
1058 .id = AV_CODEC_ID_UTVIDEO,
1059 .priv_data_size = sizeof(UtvideoContext),
1060 .init = decode_init,
1061 .close = decode_end,
1062 .decode = decode_frame,
1063 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,
1064 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE,
1065 };
1066