summaryrefslogtreecommitdiff
path: root/libavcodec/wavpack.c (plain)
blob: 2bda3599a8448e05e6c733bf69d7192f1eaa1a4f
1/*
2 * WavPack lossless audio decoder
3 * Copyright (c) 2006,2011 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#include "libavutil/channel_layout.h"
23
24#define BITSTREAM_READER_LE
25#include "avcodec.h"
26#include "bytestream.h"
27#include "get_bits.h"
28#include "internal.h"
29#include "thread.h"
30#include "unary.h"
31#include "wavpack.h"
32
33/**
34 * @file
35 * WavPack lossless audio decoder
36 */
37
38typedef struct SavedContext {
39 int offset;
40 int size;
41 int bits_used;
42 uint32_t crc;
43} SavedContext;
44
45typedef struct WavpackFrameContext {
46 AVCodecContext *avctx;
47 int frame_flags;
48 int stereo, stereo_in;
49 int joint;
50 uint32_t CRC;
51 GetBitContext gb;
52 int got_extra_bits;
53 uint32_t crc_extra_bits;
54 GetBitContext gb_extra_bits;
55 int data_size; // in bits
56 int samples;
57 int terms;
58 Decorr decorr[MAX_TERMS];
59 int zero, one, zeroes;
60 int extra_bits;
61 int and, or, shift;
62 int post_shift;
63 int hybrid, hybrid_bitrate;
64 int hybrid_maxclip, hybrid_minclip;
65 int float_flag;
66 int float_shift;
67 int float_max_exp;
68 WvChannel ch[2];
69 int pos;
70 SavedContext sc, extra_sc;
71} WavpackFrameContext;
72
73#define WV_MAX_FRAME_DECODERS 14
74
75typedef struct WavpackContext {
76 AVCodecContext *avctx;
77
78 WavpackFrameContext *fdec[WV_MAX_FRAME_DECODERS];
79 int fdec_num;
80
81 int block;
82 int samples;
83 int ch_offset;
84} WavpackContext;
85
86#define LEVEL_DECAY(a) (((a) + 0x80) >> 8)
87
88static av_always_inline int get_tail(GetBitContext *gb, int k)
89{
90 int p, e, res;
91
92 if (k < 1)
93 return 0;
94 p = av_log2(k);
95 e = (1 << (p + 1)) - k - 1;
96 res = get_bitsz(gb, p);
97 if (res >= e)
98 res = (res << 1) - e + get_bits1(gb);
99 return res;
100}
101
102static int update_error_limit(WavpackFrameContext *ctx)
103{
104 int i, br[2], sl[2];
105
106 for (i = 0; i <= ctx->stereo_in; i++) {
107 if (ctx->ch[i].bitrate_acc > UINT_MAX - ctx->ch[i].bitrate_delta)
108 return AVERROR_INVALIDDATA;
109 ctx->ch[i].bitrate_acc += ctx->ch[i].bitrate_delta;
110 br[i] = ctx->ch[i].bitrate_acc >> 16;
111 sl[i] = LEVEL_DECAY(ctx->ch[i].slow_level);
112 }
113 if (ctx->stereo_in && ctx->hybrid_bitrate) {
114 int balance = (sl[1] - sl[0] + br[1] + 1) >> 1;
115 if (balance > br[0]) {
116 br[1] = br[0] << 1;
117 br[0] = 0;
118 } else if (-balance > br[0]) {
119 br[0] <<= 1;
120 br[1] = 0;
121 } else {
122 br[1] = br[0] + balance;
123 br[0] = br[0] - balance;
124 }
125 }
126 for (i = 0; i <= ctx->stereo_in; i++) {
127 if (ctx->hybrid_bitrate) {
128 if (sl[i] - br[i] > -0x100)
129 ctx->ch[i].error_limit = wp_exp2(sl[i] - br[i] + 0x100);
130 else
131 ctx->ch[i].error_limit = 0;
132 } else {
133 ctx->ch[i].error_limit = wp_exp2(br[i]);
134 }
135 }
136
137 return 0;
138}
139
140static int wv_get_value(WavpackFrameContext *ctx, GetBitContext *gb,
141 int channel, int *last)
142{
143 int t, t2;
144 int sign, base, add, ret;
145 WvChannel *c = &ctx->ch[channel];
146
147 *last = 0;
148
149 if ((ctx->ch[0].median[0] < 2U) && (ctx->ch[1].median[0] < 2U) &&
150 !ctx->zero && !ctx->one) {
151 if (ctx->zeroes) {
152 ctx->zeroes--;
153 if (ctx->zeroes) {
154 c->slow_level -= LEVEL_DECAY(c->slow_level);
155 return 0;
156 }
157 } else {
158 t = get_unary_0_33(gb);
159 if (t >= 2) {
160 if (get_bits_left(gb) < t - 1)
161 goto error;
162 t = get_bits_long(gb, t - 1) | (1 << (t - 1));
163 } else {
164 if (get_bits_left(gb) < 0)
165 goto error;
166 }
167 ctx->zeroes = t;
168 if (ctx->zeroes) {
169 memset(ctx->ch[0].median, 0, sizeof(ctx->ch[0].median));
170 memset(ctx->ch[1].median, 0, sizeof(ctx->ch[1].median));
171 c->slow_level -= LEVEL_DECAY(c->slow_level);
172 return 0;
173 }
174 }
175 }
176
177 if (ctx->zero) {
178 t = 0;
179 ctx->zero = 0;
180 } else {
181 t = get_unary_0_33(gb);
182 if (get_bits_left(gb) < 0)
183 goto error;
184 if (t == 16) {
185 t2 = get_unary_0_33(gb);
186 if (t2 < 2) {
187 if (get_bits_left(gb) < 0)
188 goto error;
189 t += t2;
190 } else {
191 if (get_bits_left(gb) < t2 - 1)
192 goto error;
193 t += get_bits_long(gb, t2 - 1) | (1 << (t2 - 1));
194 }
195 }
196
197 if (ctx->one) {
198 ctx->one = t & 1;
199 t = (t >> 1) + 1;
200 } else {
201 ctx->one = t & 1;
202 t >>= 1;
203 }
204 ctx->zero = !ctx->one;
205 }
206
207 if (ctx->hybrid && !channel) {
208 if (update_error_limit(ctx) < 0)
209 goto error;
210 }
211
212 if (!t) {
213 base = 0;
214 add = GET_MED(0) - 1;
215 DEC_MED(0);
216 } else if (t == 1) {
217 base = GET_MED(0);
218 add = GET_MED(1) - 1;
219 INC_MED(0);
220 DEC_MED(1);
221 } else if (t == 2) {
222 base = GET_MED(0) + GET_MED(1);
223 add = GET_MED(2) - 1;
224 INC_MED(0);
225 INC_MED(1);
226 DEC_MED(2);
227 } else {
228 base = GET_MED(0) + GET_MED(1) + GET_MED(2) * (t - 2);
229 add = GET_MED(2) - 1;
230 INC_MED(0);
231 INC_MED(1);
232 INC_MED(2);
233 }
234 if (!c->error_limit) {
235 if (add >= 0x2000000U) {
236 av_log(ctx->avctx, AV_LOG_ERROR, "k %d is too large\n", add);
237 goto error;
238 }
239 ret = base + get_tail(gb, add);
240 if (get_bits_left(gb) <= 0)
241 goto error;
242 } else {
243 int mid = (base * 2 + add + 1) >> 1;
244 while (add > c->error_limit) {
245 if (get_bits_left(gb) <= 0)
246 goto error;
247 if (get_bits1(gb)) {
248 add -= (mid - base);
249 base = mid;
250 } else
251 add = mid - base - 1;
252 mid = (base * 2 + add + 1) >> 1;
253 }
254 ret = mid;
255 }
256 sign = get_bits1(gb);
257 if (ctx->hybrid_bitrate)
258 c->slow_level += wp_log2(ret) - LEVEL_DECAY(c->slow_level);
259 return sign ? ~ret : ret;
260
261error:
262 ret = get_bits_left(gb);
263 if (ret <= 0) {
264 av_log(ctx->avctx, AV_LOG_ERROR, "Too few bits (%d) left\n", ret);
265 }
266 *last = 1;
267 return 0;
268}
269
270static inline int wv_get_value_integer(WavpackFrameContext *s, uint32_t *crc,
271 unsigned S)
272{
273 unsigned bit;
274
275 if (s->extra_bits) {
276 S *= 1 << s->extra_bits;
277
278 if (s->got_extra_bits &&
279 get_bits_left(&s->gb_extra_bits) >= s->extra_bits) {
280 S |= get_bits_long(&s->gb_extra_bits, s->extra_bits);
281 *crc = *crc * 9 + (S & 0xffff) * 3 + ((unsigned)S >> 16);
282 }
283 }
284
285 bit = (S & s->and) | s->or;
286 bit = ((S + bit) << s->shift) - bit;
287
288 if (s->hybrid)
289 bit = av_clip(bit, s->hybrid_minclip, s->hybrid_maxclip);
290
291 return bit << s->post_shift;
292}
293
294static float wv_get_value_float(WavpackFrameContext *s, uint32_t *crc, int S)
295{
296 union {
297 float f;
298 uint32_t u;
299 } value;
300
301 unsigned int sign;
302 int exp = s->float_max_exp;
303
304 if (s->got_extra_bits) {
305 const int max_bits = 1 + 23 + 8 + 1;
306 const int left_bits = get_bits_left(&s->gb_extra_bits);
307
308 if (left_bits + 8 * AV_INPUT_BUFFER_PADDING_SIZE < max_bits)
309 return 0.0;
310 }
311
312 if (S) {
313 S <<= s->float_shift;
314 sign = S < 0;
315 if (sign)
316 S = -S;
317 if (S >= 0x1000000) {
318 if (s->got_extra_bits && get_bits1(&s->gb_extra_bits))
319 S = get_bits(&s->gb_extra_bits, 23);
320 else
321 S = 0;
322 exp = 255;
323 } else if (exp) {
324 int shift = 23 - av_log2(S);
325 exp = s->float_max_exp;
326 if (exp <= shift)
327 shift = --exp;
328 exp -= shift;
329
330 if (shift) {
331 S <<= shift;
332 if ((s->float_flag & WV_FLT_SHIFT_ONES) ||
333 (s->got_extra_bits &&
334 (s->float_flag & WV_FLT_SHIFT_SAME) &&
335 get_bits1(&s->gb_extra_bits))) {
336 S |= (1 << shift) - 1;
337 } else if (s->got_extra_bits &&
338 (s->float_flag & WV_FLT_SHIFT_SENT)) {
339 S |= get_bits(&s->gb_extra_bits, shift);
340 }
341 }
342 } else {
343 exp = s->float_max_exp;
344 }
345 S &= 0x7fffff;
346 } else {
347 sign = 0;
348 exp = 0;
349 if (s->got_extra_bits && (s->float_flag & WV_FLT_ZERO_SENT)) {
350 if (get_bits1(&s->gb_extra_bits)) {
351 S = get_bits(&s->gb_extra_bits, 23);
352 if (s->float_max_exp >= 25)
353 exp = get_bits(&s->gb_extra_bits, 8);
354 sign = get_bits1(&s->gb_extra_bits);
355 } else {
356 if (s->float_flag & WV_FLT_ZERO_SIGN)
357 sign = get_bits1(&s->gb_extra_bits);
358 }
359 }
360 }
361
362 *crc = *crc * 27 + S * 9 + exp * 3 + sign;
363
364 value.u = (sign << 31) | (exp << 23) | S;
365 return value.f;
366}
367
368static void wv_reset_saved_context(WavpackFrameContext *s)
369{
370 s->pos = 0;
371 s->sc.crc = s->extra_sc.crc = 0xFFFFFFFF;
372}
373
374static inline int wv_check_crc(WavpackFrameContext *s, uint32_t crc,
375 uint32_t crc_extra_bits)
376{
377 if (crc != s->CRC) {
378 av_log(s->avctx, AV_LOG_ERROR, "CRC error\n");
379 return AVERROR_INVALIDDATA;
380 }
381 if (s->got_extra_bits && crc_extra_bits != s->crc_extra_bits) {
382 av_log(s->avctx, AV_LOG_ERROR, "Extra bits CRC error\n");
383 return AVERROR_INVALIDDATA;
384 }
385
386 return 0;
387}
388
389static inline int wv_unpack_stereo(WavpackFrameContext *s, GetBitContext *gb,
390 void *dst_l, void *dst_r, const int type)
391{
392 int i, j, count = 0;
393 int last, t;
394 int A, B, L, L2, R, R2;
395 int pos = s->pos;
396 uint32_t crc = s->sc.crc;
397 uint32_t crc_extra_bits = s->extra_sc.crc;
398 int16_t *dst16_l = dst_l;
399 int16_t *dst16_r = dst_r;
400 int32_t *dst32_l = dst_l;
401 int32_t *dst32_r = dst_r;
402 float *dstfl_l = dst_l;
403 float *dstfl_r = dst_r;
404
405 s->one = s->zero = s->zeroes = 0;
406 do {
407 L = wv_get_value(s, gb, 0, &last);
408 if (last)
409 break;
410 R = wv_get_value(s, gb, 1, &last);
411 if (last)
412 break;
413 for (i = 0; i < s->terms; i++) {
414 t = s->decorr[i].value;
415 if (t > 0) {
416 if (t > 8) {
417 if (t & 1) {
418 A = 2U * s->decorr[i].samplesA[0] - s->decorr[i].samplesA[1];
419 B = 2U * s->decorr[i].samplesB[0] - s->decorr[i].samplesB[1];
420 } else {
421 A = (int)(3U * s->decorr[i].samplesA[0] - s->decorr[i].samplesA[1]) >> 1;
422 B = (int)(3U * s->decorr[i].samplesB[0] - s->decorr[i].samplesB[1]) >> 1;
423 }
424 s->decorr[i].samplesA[1] = s->decorr[i].samplesA[0];
425 s->decorr[i].samplesB[1] = s->decorr[i].samplesB[0];
426 j = 0;
427 } else {
428 A = s->decorr[i].samplesA[pos];
429 B = s->decorr[i].samplesB[pos];
430 j = (pos + t) & 7;
431 }
432 if (type != AV_SAMPLE_FMT_S16P) {
433 L2 = L + ((s->decorr[i].weightA * (int64_t)A + 512) >> 10);
434 R2 = R + ((s->decorr[i].weightB * (int64_t)B + 512) >> 10);
435 } else {
436 L2 = L + ((s->decorr[i].weightA * A + 512) >> 10);
437 R2 = R + ((s->decorr[i].weightB * B + 512) >> 10);
438 }
439 if (A && L)
440 s->decorr[i].weightA -= ((((L ^ A) >> 30) & 2) - 1) * s->decorr[i].delta;
441 if (B && R)
442 s->decorr[i].weightB -= ((((R ^ B) >> 30) & 2) - 1) * s->decorr[i].delta;
443 s->decorr[i].samplesA[j] = L = L2;
444 s->decorr[i].samplesB[j] = R = R2;
445 } else if (t == -1) {
446 if (type != AV_SAMPLE_FMT_S16P)
447 L2 = L + ((s->decorr[i].weightA * (int64_t)s->decorr[i].samplesA[0] + 512) >> 10);
448 else
449 L2 = L + ((s->decorr[i].weightA * s->decorr[i].samplesA[0] + 512) >> 10);
450 UPDATE_WEIGHT_CLIP(s->decorr[i].weightA, s->decorr[i].delta, s->decorr[i].samplesA[0], L);
451 L = L2;
452 if (type != AV_SAMPLE_FMT_S16P)
453 R2 = R + ((s->decorr[i].weightB * (int64_t)L2 + 512) >> 10);
454 else
455 R2 = R + ((s->decorr[i].weightB * L2 + 512) >> 10);
456 UPDATE_WEIGHT_CLIP(s->decorr[i].weightB, s->decorr[i].delta, L2, R);
457 R = R2;
458 s->decorr[i].samplesA[0] = R;
459 } else {
460 if (type != AV_SAMPLE_FMT_S16P)
461 R2 = R + ((s->decorr[i].weightB * (int64_t)s->decorr[i].samplesB[0] + 512) >> 10);
462 else
463 R2 = R + ((s->decorr[i].weightB * s->decorr[i].samplesB[0] + 512) >> 10);
464 UPDATE_WEIGHT_CLIP(s->decorr[i].weightB, s->decorr[i].delta, s->decorr[i].samplesB[0], R);
465 R = R2;
466
467 if (t == -3) {
468 R2 = s->decorr[i].samplesA[0];
469 s->decorr[i].samplesA[0] = R;
470 }
471
472 if (type != AV_SAMPLE_FMT_S16P)
473 L2 = L + ((s->decorr[i].weightA * (int64_t)R2 + 512) >> 10);
474 else
475 L2 = L + ((s->decorr[i].weightA * R2 + 512) >> 10);
476 UPDATE_WEIGHT_CLIP(s->decorr[i].weightA, s->decorr[i].delta, R2, L);
477 L = L2;
478 s->decorr[i].samplesB[0] = L;
479 }
480 }
481
482 if (type == AV_SAMPLE_FMT_S16P) {
483 if (FFABS(L) + FFABS(R) > (1<<19)) {
484 av_log(s->avctx, AV_LOG_ERROR, "sample %d %d too large\n", L, R);
485 return AVERROR_INVALIDDATA;
486 }
487 }
488
489 pos = (pos + 1) & 7;
490 if (s->joint)
491 L += (unsigned)(R -= (unsigned)(L >> 1));
492 crc = (crc * 3 + L) * 3 + R;
493
494 if (type == AV_SAMPLE_FMT_FLTP) {
495 *dstfl_l++ = wv_get_value_float(s, &crc_extra_bits, L);
496 *dstfl_r++ = wv_get_value_float(s, &crc_extra_bits, R);
497 } else if (type == AV_SAMPLE_FMT_S32P) {
498 *dst32_l++ = wv_get_value_integer(s, &crc_extra_bits, L);
499 *dst32_r++ = wv_get_value_integer(s, &crc_extra_bits, R);
500 } else {
501 *dst16_l++ = wv_get_value_integer(s, &crc_extra_bits, L);
502 *dst16_r++ = wv_get_value_integer(s, &crc_extra_bits, R);
503 }
504 count++;
505 } while (!last && count < s->samples);
506
507 wv_reset_saved_context(s);
508
509 if (last && count < s->samples) {
510 int size = av_get_bytes_per_sample(type);
511 memset((uint8_t*)dst_l + count*size, 0, (s->samples-count)*size);
512 memset((uint8_t*)dst_r + count*size, 0, (s->samples-count)*size);
513 }
514
515 if ((s->avctx->err_recognition & AV_EF_CRCCHECK) &&
516 wv_check_crc(s, crc, crc_extra_bits))
517 return AVERROR_INVALIDDATA;
518
519 return 0;
520}
521
522static inline int wv_unpack_mono(WavpackFrameContext *s, GetBitContext *gb,
523 void *dst, const int type)
524{
525 int i, j, count = 0;
526 int last, t;
527 int A, S, T;
528 int pos = s->pos;
529 uint32_t crc = s->sc.crc;
530 uint32_t crc_extra_bits = s->extra_sc.crc;
531 int16_t *dst16 = dst;
532 int32_t *dst32 = dst;
533 float *dstfl = dst;
534
535 s->one = s->zero = s->zeroes = 0;
536 do {
537 T = wv_get_value(s, gb, 0, &last);
538 S = 0;
539 if (last)
540 break;
541 for (i = 0; i < s->terms; i++) {
542 t = s->decorr[i].value;
543 if (t > 8) {
544 if (t & 1)
545 A = 2 * s->decorr[i].samplesA[0] - s->decorr[i].samplesA[1];
546 else
547 A = (3 * s->decorr[i].samplesA[0] - s->decorr[i].samplesA[1]) >> 1;
548 s->decorr[i].samplesA[1] = s->decorr[i].samplesA[0];
549 j = 0;
550 } else {
551 A = s->decorr[i].samplesA[pos];
552 j = (pos + t) & 7;
553 }
554 if (type != AV_SAMPLE_FMT_S16P)
555 S = T + ((s->decorr[i].weightA * (int64_t)A + 512) >> 10);
556 else
557 S = T + ((s->decorr[i].weightA * A + 512) >> 10);
558 if (A && T)
559 s->decorr[i].weightA -= ((((T ^ A) >> 30) & 2) - 1) * s->decorr[i].delta;
560 s->decorr[i].samplesA[j] = T = S;
561 }
562 pos = (pos + 1) & 7;
563 crc = crc * 3 + S;
564
565 if (type == AV_SAMPLE_FMT_FLTP) {
566 *dstfl++ = wv_get_value_float(s, &crc_extra_bits, S);
567 } else if (type == AV_SAMPLE_FMT_S32P) {
568 *dst32++ = wv_get_value_integer(s, &crc_extra_bits, S);
569 } else {
570 *dst16++ = wv_get_value_integer(s, &crc_extra_bits, S);
571 }
572 count++;
573 } while (!last && count < s->samples);
574
575 wv_reset_saved_context(s);
576
577 if (last && count < s->samples) {
578 int size = av_get_bytes_per_sample(type);
579 memset((uint8_t*)dst + count*size, 0, (s->samples-count)*size);
580 }
581
582 if (s->avctx->err_recognition & AV_EF_CRCCHECK) {
583 int ret = wv_check_crc(s, crc, crc_extra_bits);
584 if (ret < 0 && s->avctx->err_recognition & AV_EF_EXPLODE)
585 return ret;
586 }
587
588 return 0;
589}
590
591static av_cold int wv_alloc_frame_context(WavpackContext *c)
592{
593 if (c->fdec_num == WV_MAX_FRAME_DECODERS)
594 return -1;
595
596 c->fdec[c->fdec_num] = av_mallocz(sizeof(**c->fdec));
597 if (!c->fdec[c->fdec_num])
598 return -1;
599 c->fdec_num++;
600 c->fdec[c->fdec_num - 1]->avctx = c->avctx;
601 wv_reset_saved_context(c->fdec[c->fdec_num - 1]);
602
603 return 0;
604}
605
606#if HAVE_THREADS
607static int init_thread_copy(AVCodecContext *avctx)
608{
609 WavpackContext *s = avctx->priv_data;
610 s->avctx = avctx;
611 return 0;
612}
613#endif
614
615static av_cold int wavpack_decode_init(AVCodecContext *avctx)
616{
617 WavpackContext *s = avctx->priv_data;
618
619 s->avctx = avctx;
620
621 s->fdec_num = 0;
622
623 return 0;
624}
625
626static av_cold int wavpack_decode_end(AVCodecContext *avctx)
627{
628 WavpackContext *s = avctx->priv_data;
629 int i;
630
631 for (i = 0; i < s->fdec_num; i++)
632 av_freep(&s->fdec[i]);
633 s->fdec_num = 0;
634
635 return 0;
636}
637
638static int wavpack_decode_block(AVCodecContext *avctx, int block_no,
639 AVFrame *frame, const uint8_t *buf, int buf_size)
640{
641 WavpackContext *wc = avctx->priv_data;
642 ThreadFrame tframe = { .f = frame };
643 WavpackFrameContext *s;
644 GetByteContext gb;
645 void *samples_l = NULL, *samples_r = NULL;
646 int ret;
647 int got_terms = 0, got_weights = 0, got_samples = 0,
648 got_entropy = 0, got_bs = 0, got_float = 0, got_hybrid = 0;
649 int i, j, id, size, ssize, weights, t;
650 int bpp, chan = 0, chmask = 0, orig_bpp, sample_rate = 0;
651 int multiblock;
652
653 if (block_no >= wc->fdec_num && wv_alloc_frame_context(wc) < 0) {
654 av_log(avctx, AV_LOG_ERROR, "Error creating frame decode context\n");
655 return AVERROR_INVALIDDATA;
656 }
657
658 s = wc->fdec[block_no];
659 if (!s) {
660 av_log(avctx, AV_LOG_ERROR, "Context for block %d is not present\n",
661 block_no);
662 return AVERROR_INVALIDDATA;
663 }
664
665 memset(s->decorr, 0, MAX_TERMS * sizeof(Decorr));
666 memset(s->ch, 0, sizeof(s->ch));
667 s->extra_bits = 0;
668 s->and = s->or = s->shift = 0;
669 s->got_extra_bits = 0;
670
671 bytestream2_init(&gb, buf, buf_size);
672
673 s->samples = bytestream2_get_le32(&gb);
674 if (s->samples != wc->samples) {
675 av_log(avctx, AV_LOG_ERROR, "Mismatching number of samples in "
676 "a sequence: %d and %d\n", wc->samples, s->samples);
677 return AVERROR_INVALIDDATA;
678 }
679 s->frame_flags = bytestream2_get_le32(&gb);
680 bpp = av_get_bytes_per_sample(avctx->sample_fmt);
681 orig_bpp = ((s->frame_flags & 0x03) + 1) << 3;
682 multiblock = (s->frame_flags & WV_SINGLE_BLOCK) != WV_SINGLE_BLOCK;
683
684 s->stereo = !(s->frame_flags & WV_MONO);
685 s->stereo_in = (s->frame_flags & WV_FALSE_STEREO) ? 0 : s->stereo;
686 s->joint = s->frame_flags & WV_JOINT_STEREO;
687 s->hybrid = s->frame_flags & WV_HYBRID_MODE;
688 s->hybrid_bitrate = s->frame_flags & WV_HYBRID_BITRATE;
689 s->post_shift = bpp * 8 - orig_bpp + ((s->frame_flags >> 13) & 0x1f);
690 if (s->post_shift < 0 || s->post_shift > 31) {
691 return AVERROR_INVALIDDATA;
692 }
693 s->hybrid_maxclip = ((1LL << (orig_bpp - 1)) - 1);
694 s->hybrid_minclip = ((-1UL << (orig_bpp - 1)));
695 s->CRC = bytestream2_get_le32(&gb);
696
697 // parse metadata blocks
698 while (bytestream2_get_bytes_left(&gb)) {
699 id = bytestream2_get_byte(&gb);
700 size = bytestream2_get_byte(&gb);
701 if (id & WP_IDF_LONG) {
702 size |= (bytestream2_get_byte(&gb)) << 8;
703 size |= (bytestream2_get_byte(&gb)) << 16;
704 }
705 size <<= 1; // size is specified in words
706 ssize = size;
707 if (id & WP_IDF_ODD)
708 size--;
709 if (size < 0) {
710 av_log(avctx, AV_LOG_ERROR,
711 "Got incorrect block %02X with size %i\n", id, size);
712 break;
713 }
714 if (bytestream2_get_bytes_left(&gb) < ssize) {
715 av_log(avctx, AV_LOG_ERROR,
716 "Block size %i is out of bounds\n", size);
717 break;
718 }
719 switch (id & WP_IDF_MASK) {
720 case WP_ID_DECTERMS:
721 if (size > MAX_TERMS) {
722 av_log(avctx, AV_LOG_ERROR, "Too many decorrelation terms\n");
723 s->terms = 0;
724 bytestream2_skip(&gb, ssize);
725 continue;
726 }
727 s->terms = size;
728 for (i = 0; i < s->terms; i++) {
729 uint8_t val = bytestream2_get_byte(&gb);
730 s->decorr[s->terms - i - 1].value = (val & 0x1F) - 5;
731 s->decorr[s->terms - i - 1].delta = val >> 5;
732 }
733 got_terms = 1;
734 break;
735 case WP_ID_DECWEIGHTS:
736 if (!got_terms) {
737 av_log(avctx, AV_LOG_ERROR, "No decorrelation terms met\n");
738 continue;
739 }
740 weights = size >> s->stereo_in;
741 if (weights > MAX_TERMS || weights > s->terms) {
742 av_log(avctx, AV_LOG_ERROR, "Too many decorrelation weights\n");
743 bytestream2_skip(&gb, ssize);
744 continue;
745 }
746 for (i = 0; i < weights; i++) {
747 t = (int8_t)bytestream2_get_byte(&gb);
748 s->decorr[s->terms - i - 1].weightA = t * (1 << 3);
749 if (s->decorr[s->terms - i - 1].weightA > 0)
750 s->decorr[s->terms - i - 1].weightA +=
751 (s->decorr[s->terms - i - 1].weightA + 64) >> 7;
752 if (s->stereo_in) {
753 t = (int8_t)bytestream2_get_byte(&gb);
754 s->decorr[s->terms - i - 1].weightB = t * (1 << 3);
755 if (s->decorr[s->terms - i - 1].weightB > 0)
756 s->decorr[s->terms - i - 1].weightB +=
757 (s->decorr[s->terms - i - 1].weightB + 64) >> 7;
758 }
759 }
760 got_weights = 1;
761 break;
762 case WP_ID_DECSAMPLES:
763 if (!got_terms) {
764 av_log(avctx, AV_LOG_ERROR, "No decorrelation terms met\n");
765 continue;
766 }
767 t = 0;
768 for (i = s->terms - 1; (i >= 0) && (t < size); i--) {
769 if (s->decorr[i].value > 8) {
770 s->decorr[i].samplesA[0] =
771 wp_exp2(bytestream2_get_le16(&gb));
772 s->decorr[i].samplesA[1] =
773 wp_exp2(bytestream2_get_le16(&gb));
774
775 if (s->stereo_in) {
776 s->decorr[i].samplesB[0] =
777 wp_exp2(bytestream2_get_le16(&gb));
778 s->decorr[i].samplesB[1] =
779 wp_exp2(bytestream2_get_le16(&gb));
780 t += 4;
781 }
782 t += 4;
783 } else if (s->decorr[i].value < 0) {
784 s->decorr[i].samplesA[0] =
785 wp_exp2(bytestream2_get_le16(&gb));
786 s->decorr[i].samplesB[0] =
787 wp_exp2(bytestream2_get_le16(&gb));
788 t += 4;
789 } else {
790 for (j = 0; j < s->decorr[i].value; j++) {
791 s->decorr[i].samplesA[j] =
792 wp_exp2(bytestream2_get_le16(&gb));
793 if (s->stereo_in) {
794 s->decorr[i].samplesB[j] =
795 wp_exp2(bytestream2_get_le16(&gb));
796 }
797 }
798 t += s->decorr[i].value * 2 * (s->stereo_in + 1);
799 }
800 }
801 got_samples = 1;
802 break;
803 case WP_ID_ENTROPY:
804 if (size != 6 * (s->stereo_in + 1)) {
805 av_log(avctx, AV_LOG_ERROR,
806 "Entropy vars size should be %i, got %i.\n",
807 6 * (s->stereo_in + 1), size);
808 bytestream2_skip(&gb, ssize);
809 continue;
810 }
811 for (j = 0; j <= s->stereo_in; j++)
812 for (i = 0; i < 3; i++) {
813 s->ch[j].median[i] = wp_exp2(bytestream2_get_le16(&gb));
814 }
815 got_entropy = 1;
816 break;
817 case WP_ID_HYBRID:
818 if (s->hybrid_bitrate) {
819 for (i = 0; i <= s->stereo_in; i++) {
820 s->ch[i].slow_level = wp_exp2(bytestream2_get_le16(&gb));
821 size -= 2;
822 }
823 }
824 for (i = 0; i < (s->stereo_in + 1); i++) {
825 s->ch[i].bitrate_acc = bytestream2_get_le16(&gb) << 16;
826 size -= 2;
827 }
828 if (size > 0) {
829 for (i = 0; i < (s->stereo_in + 1); i++) {
830 s->ch[i].bitrate_delta =
831 wp_exp2((int16_t)bytestream2_get_le16(&gb));
832 }
833 } else {
834 for (i = 0; i < (s->stereo_in + 1); i++)
835 s->ch[i].bitrate_delta = 0;
836 }
837 got_hybrid = 1;
838 break;
839 case WP_ID_INT32INFO: {
840 uint8_t val[4];
841 if (size != 4) {
842 av_log(avctx, AV_LOG_ERROR,
843 "Invalid INT32INFO, size = %i\n",
844 size);
845 bytestream2_skip(&gb, ssize - 4);
846 continue;
847 }
848 bytestream2_get_buffer(&gb, val, 4);
849 if (val[0] > 31) {
850 av_log(avctx, AV_LOG_ERROR,
851 "Invalid INT32INFO, extra_bits = %d (> 32)\n", val[0]);
852 continue;
853 } else if (val[0]) {
854 s->extra_bits = val[0];
855 } else if (val[1]) {
856 s->shift = val[1];
857 } else if (val[2]) {
858 s->and = s->or = 1;
859 s->shift = val[2];
860 } else if (val[3]) {
861 s->and = 1;
862 s->shift = val[3];
863 }
864 if (s->shift > 31) {
865 av_log(avctx, AV_LOG_ERROR,
866 "Invalid INT32INFO, shift = %d (> 31)\n", s->shift);
867 s->and = s->or = s->shift = 0;
868 continue;
869 }
870 /* original WavPack decoder forces 32-bit lossy sound to be treated
871 * as 24-bit one in order to have proper clipping */
872 if (s->hybrid && bpp == 4 && s->post_shift < 8 && s->shift > 8) {
873 s->post_shift += 8;
874 s->shift -= 8;
875 s->hybrid_maxclip >>= 8;
876 s->hybrid_minclip >>= 8;
877 }
878 break;
879 }
880 case WP_ID_FLOATINFO:
881 if (size != 4) {
882 av_log(avctx, AV_LOG_ERROR,
883 "Invalid FLOATINFO, size = %i\n", size);
884 bytestream2_skip(&gb, ssize);
885 continue;
886 }
887 s->float_flag = bytestream2_get_byte(&gb);
888 s->float_shift = bytestream2_get_byte(&gb);
889 s->float_max_exp = bytestream2_get_byte(&gb);
890 got_float = 1;
891 bytestream2_skip(&gb, 1);
892 break;
893 case WP_ID_DATA:
894 s->sc.offset = bytestream2_tell(&gb);
895 s->sc.size = size * 8;
896 if ((ret = init_get_bits8(&s->gb, gb.buffer, size)) < 0)
897 return ret;
898 s->data_size = size * 8;
899 bytestream2_skip(&gb, size);
900 got_bs = 1;
901 break;
902 case WP_ID_EXTRABITS:
903 if (size <= 4) {
904 av_log(avctx, AV_LOG_ERROR, "Invalid EXTRABITS, size = %i\n",
905 size);
906 bytestream2_skip(&gb, size);
907 continue;
908 }
909 s->extra_sc.offset = bytestream2_tell(&gb);
910 s->extra_sc.size = size * 8;
911 if ((ret = init_get_bits8(&s->gb_extra_bits, gb.buffer, size)) < 0)
912 return ret;
913 s->crc_extra_bits = get_bits_long(&s->gb_extra_bits, 32);
914 bytestream2_skip(&gb, size);
915 s->got_extra_bits = 1;
916 break;
917 case WP_ID_CHANINFO:
918 if (size <= 1) {
919 av_log(avctx, AV_LOG_ERROR,
920 "Insufficient channel information\n");
921 return AVERROR_INVALIDDATA;
922 }
923 chan = bytestream2_get_byte(&gb);
924 switch (size - 2) {
925 case 0:
926 chmask = bytestream2_get_byte(&gb);
927 break;
928 case 1:
929 chmask = bytestream2_get_le16(&gb);
930 break;
931 case 2:
932 chmask = bytestream2_get_le24(&gb);
933 break;
934 case 3:
935 chmask = bytestream2_get_le32(&gb);
936 break;
937 case 5:
938 size = bytestream2_get_byte(&gb);
939 if (avctx->channels != size)
940 av_log(avctx, AV_LOG_WARNING, "%i channels signalled"
941 " instead of %i.\n", size, avctx->channels);
942 chan |= (bytestream2_get_byte(&gb) & 0xF) << 8;
943 chmask = bytestream2_get_le16(&gb);
944 break;
945 default:
946 av_log(avctx, AV_LOG_ERROR, "Invalid channel info size %d\n",
947 size);
948 chan = avctx->channels;
949 chmask = avctx->channel_layout;
950 }
951 break;
952 case WP_ID_SAMPLE_RATE:
953 if (size != 3) {
954 av_log(avctx, AV_LOG_ERROR, "Invalid custom sample rate.\n");
955 return AVERROR_INVALIDDATA;
956 }
957 sample_rate = bytestream2_get_le24(&gb);
958 break;
959 default:
960 bytestream2_skip(&gb, size);
961 }
962 if (id & WP_IDF_ODD)
963 bytestream2_skip(&gb, 1);
964 }
965
966 if (!got_terms) {
967 av_log(avctx, AV_LOG_ERROR, "No block with decorrelation terms\n");
968 return AVERROR_INVALIDDATA;
969 }
970 if (!got_weights) {
971 av_log(avctx, AV_LOG_ERROR, "No block with decorrelation weights\n");
972 return AVERROR_INVALIDDATA;
973 }
974 if (!got_samples) {
975 av_log(avctx, AV_LOG_ERROR, "No block with decorrelation samples\n");
976 return AVERROR_INVALIDDATA;
977 }
978 if (!got_entropy) {
979 av_log(avctx, AV_LOG_ERROR, "No block with entropy info\n");
980 return AVERROR_INVALIDDATA;
981 }
982 if (s->hybrid && !got_hybrid) {
983 av_log(avctx, AV_LOG_ERROR, "Hybrid config not found\n");
984 return AVERROR_INVALIDDATA;
985 }
986 if (!got_bs) {
987 av_log(avctx, AV_LOG_ERROR, "Packed samples not found\n");
988 return AVERROR_INVALIDDATA;
989 }
990 if (!got_float && avctx->sample_fmt == AV_SAMPLE_FMT_FLTP) {
991 av_log(avctx, AV_LOG_ERROR, "Float information not found\n");
992 return AVERROR_INVALIDDATA;
993 }
994 if (s->got_extra_bits && avctx->sample_fmt != AV_SAMPLE_FMT_FLTP) {
995 const int size = get_bits_left(&s->gb_extra_bits);
996 const int wanted = s->samples * s->extra_bits << s->stereo_in;
997 if (size < wanted) {
998 av_log(avctx, AV_LOG_ERROR, "Too small EXTRABITS\n");
999 s->got_extra_bits = 0;
1000 }
1001 }
1002
1003 if (!wc->ch_offset) {
1004 int sr = (s->frame_flags >> 23) & 0xf;
1005 if (sr == 0xf) {
1006 if (!sample_rate) {
1007 av_log(avctx, AV_LOG_ERROR, "Custom sample rate missing.\n");
1008 return AVERROR_INVALIDDATA;
1009 }
1010 avctx->sample_rate = sample_rate;
1011 } else
1012 avctx->sample_rate = wv_rates[sr];
1013
1014 if (multiblock) {
1015 if (chan)
1016 avctx->channels = chan;
1017 if (chmask)
1018 avctx->channel_layout = chmask;
1019 } else {
1020 avctx->channels = s->stereo ? 2 : 1;
1021 avctx->channel_layout = s->stereo ? AV_CH_LAYOUT_STEREO :
1022 AV_CH_LAYOUT_MONO;
1023 }
1024
1025 /* get output buffer */
1026 frame->nb_samples = s->samples + 1;
1027 if ((ret = ff_thread_get_buffer(avctx, &tframe, 0)) < 0)
1028 return ret;
1029 frame->nb_samples = s->samples;
1030 }
1031
1032 if (wc->ch_offset + s->stereo >= avctx->channels) {
1033 av_log(avctx, AV_LOG_WARNING, "Too many channels coded in a packet.\n");
1034 return ((avctx->err_recognition & AV_EF_EXPLODE) || !wc->ch_offset) ? AVERROR_INVALIDDATA : 0;
1035 }
1036
1037 samples_l = frame->extended_data[wc->ch_offset];
1038 if (s->stereo)
1039 samples_r = frame->extended_data[wc->ch_offset + 1];
1040
1041 wc->ch_offset += 1 + s->stereo;
1042
1043 if (s->stereo_in) {
1044 ret = wv_unpack_stereo(s, &s->gb, samples_l, samples_r, avctx->sample_fmt);
1045 if (ret < 0)
1046 return ret;
1047 } else {
1048 ret = wv_unpack_mono(s, &s->gb, samples_l, avctx->sample_fmt);
1049 if (ret < 0)
1050 return ret;
1051
1052 if (s->stereo)
1053 memcpy(samples_r, samples_l, bpp * s->samples);
1054 }
1055
1056 return 0;
1057}
1058
1059static void wavpack_decode_flush(AVCodecContext *avctx)
1060{
1061 WavpackContext *s = avctx->priv_data;
1062 int i;
1063
1064 for (i = 0; i < s->fdec_num; i++)
1065 wv_reset_saved_context(s->fdec[i]);
1066}
1067
1068static int wavpack_decode_frame(AVCodecContext *avctx, void *data,
1069 int *got_frame_ptr, AVPacket *avpkt)
1070{
1071 WavpackContext *s = avctx->priv_data;
1072 const uint8_t *buf = avpkt->data;
1073 int buf_size = avpkt->size;
1074 AVFrame *frame = data;
1075 int frame_size, ret, frame_flags;
1076
1077 if (avpkt->size <= WV_HEADER_SIZE)
1078 return AVERROR_INVALIDDATA;
1079
1080 s->block = 0;
1081 s->ch_offset = 0;
1082
1083 /* determine number of samples */
1084 s->samples = AV_RL32(buf + 20);
1085 frame_flags = AV_RL32(buf + 24);
1086 if (s->samples <= 0 || s->samples > WV_MAX_SAMPLES) {
1087 av_log(avctx, AV_LOG_ERROR, "Invalid number of samples: %d\n",
1088 s->samples);
1089 return AVERROR_INVALIDDATA;
1090 }
1091
1092 if (frame_flags & 0x80) {
1093 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
1094 } else if ((frame_flags & 0x03) <= 1) {
1095 avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
1096 } else {
1097 avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
1098 avctx->bits_per_raw_sample = ((frame_flags & 0x03) + 1) << 3;
1099 }
1100
1101 while (buf_size > 0) {
1102 if (buf_size <= WV_HEADER_SIZE)
1103 break;
1104 frame_size = AV_RL32(buf + 4) - 12;
1105 buf += 20;
1106 buf_size -= 20;
1107 if (frame_size <= 0 || frame_size > buf_size) {
1108 av_log(avctx, AV_LOG_ERROR,
1109 "Block %d has invalid size (size %d vs. %d bytes left)\n",
1110 s->block, frame_size, buf_size);
1111 wavpack_decode_flush(avctx);
1112 return AVERROR_INVALIDDATA;
1113 }
1114 if ((ret = wavpack_decode_block(avctx, s->block,
1115 frame, buf, frame_size)) < 0) {
1116 wavpack_decode_flush(avctx);
1117 return ret;
1118 }
1119 s->block++;
1120 buf += frame_size;
1121 buf_size -= frame_size;
1122 }
1123
1124 if (s->ch_offset != avctx->channels) {
1125 av_log(avctx, AV_LOG_ERROR, "Not enough channels coded in a packet.\n");
1126 return AVERROR_INVALIDDATA;
1127 }
1128
1129 *got_frame_ptr = 1;
1130
1131 return avpkt->size;
1132}
1133
1134AVCodec ff_wavpack_decoder = {
1135 .name = "wavpack",
1136 .long_name = NULL_IF_CONFIG_SMALL("WavPack"),
1137 .type = AVMEDIA_TYPE_AUDIO,
1138 .id = AV_CODEC_ID_WAVPACK,
1139 .priv_data_size = sizeof(WavpackContext),
1140 .init = wavpack_decode_init,
1141 .close = wavpack_decode_end,
1142 .decode = wavpack_decode_frame,
1143 .flush = wavpack_decode_flush,
1144 .init_thread_copy = ONLY_IF_THREADS_ENABLED(init_thread_copy),
1145 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,
1146};
1147
generated by cgit at 2024-05-17 08:31:38 (GMT)