utvideodec: Handle slice_height being zero
[libav.git] / libavcodec / utvideodec.c
1 /*
2 * Ut Video decoder
3 * Copyright (c) 2011 Konstantin Shishkov
4 *
5 * This file is part of Libav.
6 *
7 * Libav 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 * Libav 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 Libav; 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 #include "libavutil/intreadwrite.h"
31 #include "avcodec.h"
32 #include "bswapdsp.h"
33 #include "bytestream.h"
34 #include "get_bits.h"
35 #include "thread.h"
36 #include "utvideo.h"
37
38 static int build_huff(const uint8_t *src, VLC *vlc, int *fsym)
39 {
40 int i;
41 HuffEntry he[256];
42 int last;
43 uint32_t codes[256];
44 uint8_t bits[256];
45 uint8_t syms[256];
46 uint32_t code;
47
48 *fsym = -1;
49 for (i = 0; i < 256; i++) {
50 he[i].sym = i;
51 he[i].len = *src++;
52 }
53 qsort(he, 256, sizeof(*he), ff_ut_huff_cmp_len);
54
55 if (!he[0].len) {
56 *fsym = he[0].sym;
57 return 0;
58 }
59 if (he[0].len > 32)
60 return -1;
61
62 last = 255;
63 while (he[last].len == 255 && last)
64 last--;
65
66 code = 1;
67 for (i = last; i >= 0; i--) {
68 codes[i] = code >> (32 - he[i].len);
69 bits[i] = he[i].len;
70 syms[i] = he[i].sym;
71 code += 0x80000000u >> (he[i].len - 1);
72 }
73
74 return ff_init_vlc_sparse(vlc, FFMIN(he[last].len, 9), last + 1,
75 bits, sizeof(*bits), sizeof(*bits),
76 codes, sizeof(*codes), sizeof(*codes),
77 syms, sizeof(*syms), sizeof(*syms), 0);
78 }
79
80 static int decode_plane(UtvideoContext *c, int plane_no,
81 uint8_t *dst, int step, int stride,
82 int width, int height,
83 const uint8_t *src, int use_pred)
84 {
85 int i, j, slice, pix;
86 int sstart, send;
87 VLC vlc;
88 GetBitContext gb;
89 int prev, fsym;
90 const int cmask = ~(!plane_no && c->avctx->pix_fmt == AV_PIX_FMT_YUV420P);
91
92 if (build_huff(src, &vlc, &fsym)) {
93 av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n");
94 return AVERROR_INVALIDDATA;
95 }
96 if (fsym >= 0) { // build_huff reported a symbol to fill slices with
97 send = 0;
98 for (slice = 0; slice < c->slices; slice++) {
99 uint8_t *dest;
100
101 sstart = send;
102 send = (height * (slice + 1) / c->slices) & cmask;
103 dest = dst + sstart * stride;
104
105 prev = 0x80;
106 for (j = sstart; j < send; j++) {
107 for (i = 0; i < width * step; i += step) {
108 pix = fsym;
109 if (use_pred) {
110 prev += pix;
111 pix = prev;
112 }
113 dest[i] = pix;
114 }
115 dest += stride;
116 }
117 }
118 return 0;
119 }
120
121 src += 256;
122
123 send = 0;
124 for (slice = 0; slice < c->slices; slice++) {
125 uint8_t *dest;
126 int slice_data_start, slice_data_end, slice_size;
127
128 sstart = send;
129 send = (height * (slice + 1) / c->slices) & cmask;
130 dest = dst + sstart * stride;
131
132 // slice offset and size validation was done earlier
133 slice_data_start = slice ? AV_RL32(src + slice * 4 - 4) : 0;
134 slice_data_end = AV_RL32(src + slice * 4);
135 slice_size = slice_data_end - slice_data_start;
136
137 if (!slice_size) {
138 av_log(c->avctx, AV_LOG_ERROR, "Plane has more than one symbol "
139 "yet a slice has a length of zero.\n");
140 goto fail;
141 }
142
143 memcpy(c->slice_bits, src + slice_data_start + c->slices * 4,
144 slice_size);
145 memset(c->slice_bits + slice_size, 0, FF_INPUT_BUFFER_PADDING_SIZE);
146 c->bdsp.bswap_buf((uint32_t *) c->slice_bits,
147 (uint32_t *) c->slice_bits,
148 (slice_data_end - slice_data_start + 3) >> 2);
149 init_get_bits(&gb, c->slice_bits, slice_size * 8);
150
151 prev = 0x80;
152 for (j = sstart; j < send; j++) {
153 for (i = 0; i < width * step; i += step) {
154 if (get_bits_left(&gb) <= 0) {
155 av_log(c->avctx, AV_LOG_ERROR,
156 "Slice decoding ran out of bits\n");
157 goto fail;
158 }
159 pix = get_vlc2(&gb, vlc.table, vlc.bits, 4);
160 if (pix < 0) {
161 av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
162 goto fail;
163 }
164 if (use_pred) {
165 prev += pix;
166 pix = prev;
167 }
168 dest[i] = pix;
169 }
170 dest += stride;
171 }
172 if (get_bits_left(&gb) > 32)
173 av_log(c->avctx, AV_LOG_WARNING,
174 "%d bits left after decoding slice\n", get_bits_left(&gb));
175 }
176
177 ff_free_vlc(&vlc);
178
179 return 0;
180 fail:
181 ff_free_vlc(&vlc);
182 return AVERROR_INVALIDDATA;
183 }
184
185 static void restore_rgb_planes(uint8_t *src, int step, int stride, int width,
186 int height)
187 {
188 int i, j;
189 uint8_t r, g, b;
190
191 for (j = 0; j < height; j++) {
192 for (i = 0; i < width * step; i += step) {
193 r = src[i];
194 g = src[i + 1];
195 b = src[i + 2];
196 src[i] = r + g - 0x80;
197 src[i + 2] = b + g - 0x80;
198 }
199 src += stride;
200 }
201 }
202
203 static void restore_median(uint8_t *src, int step, int stride,
204 int width, int height, int slices, int rmode)
205 {
206 int i, j, slice;
207 int A, B, C;
208 uint8_t *bsrc;
209 int slice_start, slice_height;
210 const int cmask = ~rmode;
211
212 for (slice = 0; slice < slices; slice++) {
213 slice_start = ((slice * height) / slices) & cmask;
214 slice_height = ((((slice + 1) * height) / slices) & cmask) -
215 slice_start;
216 if (!slice_height)
217 continue;
218
219 bsrc = src + slice_start * stride;
220
221 // first line - left neighbour prediction
222 bsrc[0] += 0x80;
223 A = bsrc[0];
224 for (i = step; i < width * step; i += step) {
225 bsrc[i] += A;
226 A = bsrc[i];
227 }
228 bsrc += stride;
229 if (slice_height == 1)
230 continue;
231 // second line - first element has top prediction, the rest uses median
232 C = bsrc[-stride];
233 bsrc[0] += C;
234 A = bsrc[0];
235 for (i = step; i < width * step; i += step) {
236 B = bsrc[i - stride];
237 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
238 C = B;
239 A = bsrc[i];
240 }
241 bsrc += stride;
242 // the rest of lines use continuous median prediction
243 for (j = 2; j < slice_height; j++) {
244 for (i = 0; i < width * step; i += step) {
245 B = bsrc[i - stride];
246 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
247 C = B;
248 A = bsrc[i];
249 }
250 bsrc += stride;
251 }
252 }
253 }
254
255 /* UtVideo interlaced mode treats every two lines as a single one,
256 * so restoring function should take care of possible padding between
257 * two parts of the same "line".
258 */
259 static void restore_median_il(uint8_t *src, int step, int stride,
260 int width, int height, int slices, int rmode)
261 {
262 int i, j, slice;
263 int A, B, C;
264 uint8_t *bsrc;
265 int slice_start, slice_height;
266 const int cmask = ~(rmode ? 3 : 1);
267 const int stride2 = stride << 1;
268
269 for (slice = 0; slice < slices; slice++) {
270 slice_start = ((slice * height) / slices) & cmask;
271 slice_height = ((((slice + 1) * height) / slices) & cmask) -
272 slice_start;
273 slice_height >>= 1;
274 if (!slice_height)
275 continue;
276
277 bsrc = src + slice_start * stride;
278
279 // first line - left neighbour prediction
280 bsrc[0] += 0x80;
281 A = bsrc[0];
282 for (i = step; i < width * step; i += step) {
283 bsrc[i] += A;
284 A = bsrc[i];
285 }
286 for (i = 0; i < width * step; i += step) {
287 bsrc[stride + i] += A;
288 A = bsrc[stride + i];
289 }
290 bsrc += stride2;
291 if (slice_height == 1)
292 continue;
293 // second line - first element has top prediction, the rest uses median
294 C = bsrc[-stride2];
295 bsrc[0] += C;
296 A = bsrc[0];
297 for (i = step; i < width * step; i += step) {
298 B = bsrc[i - stride2];
299 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
300 C = B;
301 A = bsrc[i];
302 }
303 for (i = 0; i < width * step; i += step) {
304 B = bsrc[i - stride];
305 bsrc[stride + i] += mid_pred(A, B, (uint8_t)(A + B - C));
306 C = B;
307 A = bsrc[stride + i];
308 }
309 bsrc += stride2;
310 // the rest of lines use continuous median prediction
311 for (j = 2; j < slice_height; j++) {
312 for (i = 0; i < width * step; i += step) {
313 B = bsrc[i - stride2];
314 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
315 C = B;
316 A = bsrc[i];
317 }
318 for (i = 0; i < width * step; i += step) {
319 B = bsrc[i - stride];
320 bsrc[i + stride] += mid_pred(A, B, (uint8_t)(A + B - C));
321 C = B;
322 A = bsrc[i + stride];
323 }
324 bsrc += stride2;
325 }
326 }
327 }
328
329 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
330 AVPacket *avpkt)
331 {
332 const uint8_t *buf = avpkt->data;
333 int buf_size = avpkt->size;
334 UtvideoContext *c = avctx->priv_data;
335 int i, j;
336 const uint8_t *plane_start[5];
337 int plane_size, max_slice_size = 0, slice_start, slice_end, slice_size;
338 int ret;
339 GetByteContext gb;
340 ThreadFrame frame = { .f = data };
341
342 if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0) {
343 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
344 return ret;
345 }
346
347 ff_thread_finish_setup(avctx);
348
349 /* parse plane structure to get frame flags and validate slice offsets */
350 bytestream2_init(&gb, buf, buf_size);
351 for (i = 0; i < c->planes; i++) {
352 plane_start[i] = gb.buffer;
353 if (bytestream2_get_bytes_left(&gb) < 256 + 4 * c->slices) {
354 av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
355 return AVERROR_INVALIDDATA;
356 }
357 bytestream2_skipu(&gb, 256);
358 slice_start = 0;
359 slice_end = 0;
360 for (j = 0; j < c->slices; j++) {
361 slice_end = bytestream2_get_le32u(&gb);
362 slice_size = slice_end - slice_start;
363 if (slice_end < 0 || slice_size < 0 ||
364 bytestream2_get_bytes_left(&gb) < slice_end) {
365 av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
366 return AVERROR_INVALIDDATA;
367 }
368 slice_start = slice_end;
369 max_slice_size = FFMAX(max_slice_size, slice_size);
370 }
371 plane_size = slice_end;
372 bytestream2_skipu(&gb, plane_size);
373 }
374 plane_start[c->planes] = gb.buffer;
375 if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
376 av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
377 return AVERROR_INVALIDDATA;
378 }
379 c->frame_info = bytestream2_get_le32u(&gb);
380 av_log(avctx, AV_LOG_DEBUG, "frame information flags %"PRIX32"\n",
381 c->frame_info);
382
383 c->frame_pred = (c->frame_info >> 8) & 3;
384
385 if (c->frame_pred == PRED_GRADIENT) {
386 avpriv_request_sample(avctx, "Frame with gradient prediction");
387 return AVERROR_PATCHWELCOME;
388 }
389
390 av_fast_malloc(&c->slice_bits, &c->slice_bits_size,
391 max_slice_size + FF_INPUT_BUFFER_PADDING_SIZE);
392
393 if (!c->slice_bits) {
394 av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer\n");
395 return AVERROR(ENOMEM);
396 }
397
398 switch (c->avctx->pix_fmt) {
399 case AV_PIX_FMT_RGB24:
400 case AV_PIX_FMT_RGBA:
401 for (i = 0; i < c->planes; i++) {
402 ret = decode_plane(c, i, frame.f->data[0] + ff_ut_rgb_order[i],
403 c->planes, frame.f->linesize[0], avctx->width,
404 avctx->height, plane_start[i],
405 c->frame_pred == PRED_LEFT);
406 if (ret)
407 return ret;
408 if (c->frame_pred == PRED_MEDIAN) {
409 if (!c->interlaced) {
410 restore_median(frame.f->data[0] + ff_ut_rgb_order[i],
411 c->planes, frame.f->linesize[0], avctx->width,
412 avctx->height, c->slices, 0);
413 } else {
414 restore_median_il(frame.f->data[0] + ff_ut_rgb_order[i],
415 c->planes, frame.f->linesize[0],
416 avctx->width, avctx->height, c->slices,
417 0);
418 }
419 }
420 }
421 restore_rgb_planes(frame.f->data[0], c->planes, frame.f->linesize[0],
422 avctx->width, avctx->height);
423 break;
424 case AV_PIX_FMT_YUV420P:
425 for (i = 0; i < 3; i++) {
426 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
427 avctx->width >> !!i, avctx->height >> !!i,
428 plane_start[i], c->frame_pred == PRED_LEFT);
429 if (ret)
430 return ret;
431 if (c->frame_pred == PRED_MEDIAN) {
432 if (!c->interlaced) {
433 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
434 avctx->width >> !!i, avctx->height >> !!i,
435 c->slices, !i);
436 } else {
437 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
438 avctx->width >> !!i,
439 avctx->height >> !!i,
440 c->slices, !i);
441 }
442 }
443 }
444 break;
445 case AV_PIX_FMT_YUV422P:
446 for (i = 0; i < 3; i++) {
447 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
448 avctx->width >> !!i, avctx->height,
449 plane_start[i], c->frame_pred == PRED_LEFT);
450 if (ret)
451 return ret;
452 if (c->frame_pred == PRED_MEDIAN) {
453 if (!c->interlaced) {
454 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
455 avctx->width >> !!i, avctx->height,
456 c->slices, 0);
457 } else {
458 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
459 avctx->width >> !!i, avctx->height,
460 c->slices, 0);
461 }
462 }
463 }
464 break;
465 }
466
467 frame.f->key_frame = 1;
468 frame.f->pict_type = AV_PICTURE_TYPE_I;
469 frame.f->interlaced_frame = !!c->interlaced;
470
471 *got_frame = 1;
472
473 /* always report that the buffer was completely consumed */
474 return buf_size;
475 }
476
477 static av_cold int decode_init(AVCodecContext *avctx)
478 {
479 UtvideoContext * const c = avctx->priv_data;
480
481 c->avctx = avctx;
482
483 ff_bswapdsp_init(&c->bdsp);
484
485 if (avctx->extradata_size < 16) {
486 av_log(avctx, AV_LOG_ERROR,
487 "Insufficient extradata size %d, should be at least 16\n",
488 avctx->extradata_size);
489 return AVERROR_INVALIDDATA;
490 }
491
492 av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
493 avctx->extradata[3], avctx->extradata[2],
494 avctx->extradata[1], avctx->extradata[0]);
495 av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
496 AV_RB32(avctx->extradata + 4));
497 c->frame_info_size = AV_RL32(avctx->extradata + 8);
498 c->flags = AV_RL32(avctx->extradata + 12);
499
500 if (c->frame_info_size != 4)
501 avpriv_request_sample(avctx, "Frame info not 4 bytes");
502 av_log(avctx, AV_LOG_DEBUG, "Encoding parameters %08"PRIX32"\n", c->flags);
503 c->slices = (c->flags >> 24) + 1;
504 c->compression = c->flags & 1;
505 c->interlaced = c->flags & 0x800;
506
507 c->slice_bits_size = 0;
508
509 switch (avctx->codec_tag) {
510 case MKTAG('U', 'L', 'R', 'G'):
511 c->planes = 3;
512 avctx->pix_fmt = AV_PIX_FMT_RGB24;
513 break;
514 case MKTAG('U', 'L', 'R', 'A'):
515 c->planes = 4;
516 avctx->pix_fmt = AV_PIX_FMT_RGBA;
517 break;
518 case MKTAG('U', 'L', 'Y', '0'):
519 c->planes = 3;
520 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
521 avctx->colorspace = AVCOL_SPC_BT470BG;
522 break;
523 case MKTAG('U', 'L', 'Y', '2'):
524 c->planes = 3;
525 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
526 avctx->colorspace = AVCOL_SPC_BT470BG;
527 break;
528 case MKTAG('U', 'L', 'H', '0'):
529 c->planes = 3;
530 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
531 avctx->colorspace = AVCOL_SPC_BT709;
532 break;
533 case MKTAG('U', 'L', 'H', '2'):
534 c->planes = 3;
535 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
536 avctx->colorspace = AVCOL_SPC_BT709;
537 break;
538 default:
539 av_log(avctx, AV_LOG_ERROR, "Unknown Ut Video FOURCC provided (%08X)\n",
540 avctx->codec_tag);
541 return AVERROR_INVALIDDATA;
542 }
543
544 return 0;
545 }
546
547 static av_cold int decode_end(AVCodecContext *avctx)
548 {
549 UtvideoContext * const c = avctx->priv_data;
550
551 av_freep(&c->slice_bits);
552
553 return 0;
554 }
555
556 AVCodec ff_utvideo_decoder = {
557 .name = "utvideo",
558 .long_name = NULL_IF_CONFIG_SMALL("Ut Video"),
559 .type = AVMEDIA_TYPE_VIDEO,
560 .id = AV_CODEC_ID_UTVIDEO,
561 .priv_data_size = sizeof(UtvideoContext),
562 .init = decode_init,
563 .close = decode_end,
564 .decode = decode_frame,
565 .capabilities = CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS,
566 };