libavformat/wavdec.c

   1 /*
   2  * WAV demuxer
   3  * Copyright (c) 2001, 2002 Fabrice Bellard
   4  *
   5  * Sony Wave64 demuxer
   6  * RF64 demuxer
   7  * Copyright (c) 2009 Daniel Verkamp
   8  *
   9  * This file is part of FFmpeg.
  10  *
  11  * FFmpeg is free software; you can redistribute it and/or
  12  * modify it under the terms of the GNU Lesser General Public
  13  * License as published by the Free Software Foundation; either
  14  * version 2.1 of the License, or (at your option) any later version.
  15  *
  16  * FFmpeg is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  19  * Lesser General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU Lesser General Public
  22  * License along with FFmpeg; if not, write to the Free Software
  23  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  24  */
  25
  26 #include <stdint.h>
  27
  28 #include "libavutil/avassert.h"
  29 #include "libavutil/dict.h"
  30 #include "libavutil/intreadwrite.h"
  31 #include "libavutil/log.h"
  32 #include "libavutil/mathematics.h"
  33 #include "libavutil/opt.h"
  34 #include "avformat.h"
  35 #include "avio.h"
  36 #include "avio_internal.h"
  37 #include "internal.h"
  38 #include "metadata.h"
  39 #include "pcm.h"
  40 #include "riff.h"
  41 #include "w64.h"
  42 #include "spdif.h"
  43
  44 typedef struct WAVDemuxContext {
  45     const AVClass *class;
  46     int64_t data_end;
  47     int w64;
  48     int64_t smv_data_ofs;
  49     int smv_block_size;
  50     int smv_frames_per_jpeg;
  51     int smv_block;
  52     int smv_last_stream;
  53     int smv_eof;
  54     int audio_eof;
  55     int ignore_length;
  56     int spdif;
  57     int smv_cur_pt;
  58     int smv_given_first;
  59     int unaligned; // e.g. if an odd number of bytes ID3 tag was prepended
  60     int rifx; // RIFX: integer byte order for parameters is big endian
  61 } WAVDemuxContext;
  62
  63 static void set_spdif(AVFormatContext *s, WAVDemuxContext *wav)
  64 {
  65     if (CONFIG_SPDIF_DEMUXER && s->streams[0]->codecpar->codec_tag == 1) {
  66         enum AVCodecID codec;
  67         uint8_t *buf = NULL;
  68         int len = 1<<16;
  69         int ret = ffio_ensure_seekback(s->pb, len);
  70         int64_t pos = avio_tell(s->pb);
  71
  72         if (ret < 0)
  73             goto end;
  74
  75         buf = av_malloc(len + AV_INPUT_BUFFER_PADDING_SIZE);
  76         if (!buf) {
  77             ret = AVERROR(ENOMEM);
  78             goto end;
  79         }
  80
  81         len = ret = avio_read(s->pb, buf, len);
  82         if (ret < 0)
  83             goto end;
  84
  85         ret = ff_spdif_probe(buf, len, &codec);
  86         if (ret > AVPROBE_SCORE_EXTENSION) {
  87             s->streams[0]->codecpar->codec_id = codec;
  88             wav->spdif = 1;
  89         }
  90 end:
  91         avio_seek(s->pb, pos, SEEK_SET);
  92         if (ret < 0)
  93             av_log(s, AV_LOG_WARNING, "Cannot check for SPDIF\n");
  94         av_free(buf);
  95     }
  96 }
  97
  98 #if CONFIG_WAV_DEMUXER
  99
 100 static int64_t next_tag(AVIOContext *pb, uint32_t *tag, int big_endian)
 101 {
 102     *tag = avio_rl32(pb);
 103     if (!big_endian) {
 104         return avio_rl32(pb);
 105     } else {
 106         return avio_rb32(pb);
 107     }
 108 }
 109
 110 /* RIFF chunks are always at even offsets relative to where they start. */
 111 static int64_t wav_seek_tag(WAVDemuxContext * wav, AVIOContext *s, int64_t offset, int whence)
 112 {
 113     offset += offset < INT64_MAX && offset + wav->unaligned & 1;
 114
 115     return avio_seek(s, offset, whence);
 116 }
 117
 118 /* return the size of the found tag */
 119 static int64_t find_tag(WAVDemuxContext * wav, AVIOContext *pb, uint32_t tag1)
 120 {
 121     unsigned int tag;
 122     int64_t size;
 123
 124     for (;;) {
 125         if (avio_feof(pb))
 126             return AVERROR_EOF;
 127         size = next_tag(pb, &tag, wav->rifx);
 128         if (tag == tag1)
 129             break;
 130         wav_seek_tag(wav, pb, size, SEEK_CUR);
 131     }
 132     return size;
 133 }
 134
 135 static int wav_probe(AVProbeData *p)
 136 {
 137     /* check file header */
 138     if (p->buf_size <= 32)
 139         return 0;
 140     if (!memcmp(p->buf + 8, "WAVE", 4)) {
 141         if (!memcmp(p->buf, "RIFF", 4) || !memcmp(p->buf, "RIFX", 4))
 142             /* Since the ACT demuxer has a standard WAV header at the top of
 143              * its own, the returned score is decreased to avoid a probe
 144              * conflict between ACT and WAV. */
 145             return AVPROBE_SCORE_MAX - 1;
 146         else if (!memcmp(p->buf,      "RF64", 4) &&
 147                  !memcmp(p->buf + 12, "ds64", 4))
 148             return AVPROBE_SCORE_MAX;
 149     }
 150     return 0;
 151 }
 152
 153 static void handle_stream_probing(AVStream *st)
 154 {
 155     if (st->codecpar->codec_id == AV_CODEC_ID_PCM_S16LE) {
 156         st->request_probe = AVPROBE_SCORE_EXTENSION;
 157         st->probe_packets = FFMIN(st->probe_packets, 32);
 158     }
 159 }
 160
 161 static int wav_parse_fmt_tag(AVFormatContext *s, int64_t size, AVStream **st)
 162 {
 163     AVIOContext *pb = s->pb;
 164     WAVDemuxContext *wav = s->priv_data;
 165     int ret;
 166
 167     /* parse fmt header */
 168     *st = avformat_new_stream(s, NULL);
 169     if (!*st)
 170         return AVERROR(ENOMEM);
 171
 172     ret = ff_get_wav_header(s, pb, (*st)->codecpar, size, wav->rifx);
 173     if (ret < 0)
 174         return ret;
 175     handle_stream_probing(*st);
 176
 177     (*st)->need_parsing = AVSTREAM_PARSE_FULL_RAW;
 178
 179     avpriv_set_pts_info(*st, 64, 1, (*st)->codecpar->sample_rate);
 180
 181     return 0;
 182 }
 183
 184 static int wav_parse_xma2_tag(AVFormatContext *s, int64_t size, AVStream **st)
 185 {
 186     AVIOContext *pb = s->pb;
 187     int num_streams, i, channels = 0;
 188
 189     if (size < 44)
 190         return AVERROR_INVALIDDATA;
 191
 192     *st = avformat_new_stream(s, NULL);
 193     if (!*st)
 194         return AVERROR(ENOMEM);
 195
 196     (*st)->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
 197     (*st)->codecpar->codec_id   = AV_CODEC_ID_XMA2;
 198     (*st)->need_parsing         = AVSTREAM_PARSE_FULL_RAW;
 199
 200     avio_skip(pb, 1);
 201     num_streams = avio_r8(pb);
 202     if (size < 40 + num_streams * 4)
 203         return AVERROR_INVALIDDATA;
 204     avio_skip(pb, 10);
 205     (*st)->codecpar->sample_rate = avio_rb32(pb);
 206     avio_skip(pb, 12);
 207     (*st)->duration = avio_rb32(pb);
 208     avio_skip(pb, 8);
 209
 210     for (i = 0; i < num_streams; i++) {
 211         channels += avio_r8(pb);
 212         avio_skip(pb, 3);
 213     }
 214     (*st)->codecpar->channels = channels;
 215
 216     if ((*st)->codecpar->channels <= 0 || (*st)->codecpar->sample_rate <= 0)
 217         return AVERROR_INVALIDDATA;
 218
 219     avpriv_set_pts_info(*st, 64, 1, (*st)->codecpar->sample_rate);
 220     if (ff_alloc_extradata((*st)->codecpar, 34))
 221         return AVERROR(ENOMEM);
 222     memset((*st)->codecpar->extradata, 0, 34);
 223
 224     return 0;
 225 }
 226
 227 static inline int wav_parse_bext_string(AVFormatContext *s, const char *key,
 228                                         int length)
 229 {
 230     char temp[257];
 231     int ret;
 232
 233     av_assert0(length < sizeof(temp));
 234     if ((ret = avio_read(s->pb, temp, length)) != length)
 235         return ret < 0 ? ret : AVERROR_INVALIDDATA;
 236
 237     temp[length] = 0;
 238
 239     if (strlen(temp))
 240         return av_dict_set(&s->metadata, key, temp, 0);
 241
 242     return 0;
 243 }
 244
 245 static int wav_parse_bext_tag(AVFormatContext *s, int64_t size)
 246 {
 247     char temp[131], *coding_history;
 248     int ret, x;
 249     uint64_t time_reference;
 250     int64_t umid_parts[8], umid_mask = 0;
 251
 252     if ((ret = wav_parse_bext_string(s, "description", 256)) < 0 ||
 253         (ret = wav_parse_bext_string(s, "originator", 32)) < 0 ||
 254         (ret = wav_parse_bext_string(s, "originator_reference", 32)) < 0 ||
 255         (ret = wav_parse_bext_string(s, "origination_date", 10)) < 0 ||
 256         (ret = wav_parse_bext_string(s, "origination_time", 8)) < 0)
 257         return ret;
 258
 259     time_reference = avio_rl64(s->pb);
 260     snprintf(temp, sizeof(temp), "%"PRIu64, time_reference);
 261     if ((ret = av_dict_set(&s->metadata, "time_reference", temp, 0)) < 0)
 262         return ret;
 263
 264     /* check if version is >= 1, in which case an UMID may be present */
 265     if (avio_rl16(s->pb) >= 1) {
 266         for (x = 0; x < 8; x++)
 267             umid_mask |= umid_parts[x] = avio_rb64(s->pb);
 268
 269         if (umid_mask) {
 270             /* the string formatting below is per SMPTE 330M-2004 Annex C */
 271             if (umid_parts[4] == 0 && umid_parts[5] == 0 &&
 272                 umid_parts[6] == 0 && umid_parts[7] == 0) {
 273                 /* basic UMID */
 274                 snprintf(temp, sizeof(temp),
 275                          "0x%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64,
 276                          umid_parts[0], umid_parts[1],
 277                          umid_parts[2], umid_parts[3]);
 278             } else {
 279                 /* extended UMID */
 280                 snprintf(temp, sizeof(temp),
 281                          "0x%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64
 282                          "%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64,
 283                          umid_parts[0], umid_parts[1],
 284                          umid_parts[2], umid_parts[3],
 285                          umid_parts[4], umid_parts[5],
 286                          umid_parts[6], umid_parts[7]);
 287             }
 288
 289             if ((ret = av_dict_set(&s->metadata, "umid", temp, 0)) < 0)
 290                 return ret;
 291         }
 292
 293         avio_skip(s->pb, 190);
 294     } else
 295         avio_skip(s->pb, 254);
 296
 297     if (size > 602) {
 298         /* CodingHistory present */
 299         size -= 602;
 300
 301         if (!(coding_history = av_malloc(size + 1)))
 302             return AVERROR(ENOMEM);
 303
 304         if ((ret = avio_read(s->pb, coding_history, size)) != size) {
 305             av_free(coding_history);
 306             return ret < 0 ? ret : AVERROR_INVALIDDATA;
 307         }
 308
 309         coding_history[size] = 0;
 310         if ((ret = av_dict_set(&s->metadata, "coding_history", coding_history,
 311                                AV_DICT_DONT_STRDUP_VAL)) < 0)
 312             return ret;
 313     }
 314
 315     return 0;
 316 }
 317
 318 static const AVMetadataConv wav_metadata_conv[] = {
 319     { "description",      "comment"       },
 320     { "originator",       "encoded_by"    },
 321     { "origination_date", "date"          },
 322     { "origination_time", "creation_time" },
 323     { 0 },
 324 };
 325
 326 /* wav input */
 327 static int wav_read_header(AVFormatContext *s)
 328 {
 329     int64_t size, av_uninit(data_size);
 330     int64_t sample_count = 0;
 331     int rf64 = 0;
 332     char start_code[32];
 333     uint32_t tag;
 334     AVIOContext *pb      = s->pb;
 335     AVStream *st         = NULL;
 336     WAVDemuxContext *wav = s->priv_data;
 337     int ret, got_fmt = 0, got_xma2 = 0;
 338     int64_t next_tag_ofs, data_ofs = -1;
 339
 340     wav->unaligned = avio_tell(s->pb) & 1;
 341
 342     wav->smv_data_ofs = -1;
 343
 344     /* read chunk ID */
 345     tag = avio_rl32(pb);
 346     switch (tag) {
 347     case MKTAG('R', 'I', 'F', 'F'):
 348         break;
 349     case MKTAG('R', 'I', 'F', 'X'):
 350         wav->rifx = 1;
 351         break;
 352     case MKTAG('R', 'F', '6', '4'):
 353         rf64 = 1;
 354         break;
 355     default:
 356         av_get_codec_tag_string(start_code, sizeof(start_code), tag);
 357         av_log(s, AV_LOG_ERROR, "invalid start code %s in RIFF header\n", start_code);
 358         return AVERROR_INVALIDDATA;
 359     }
 360
 361     /* read chunk size */
 362     avio_rl32(pb);
 363
 364     /* read format */
 365     if (avio_rl32(pb) != MKTAG('W', 'A', 'V', 'E')) {
 366         av_log(s, AV_LOG_ERROR, "invalid format in RIFF header\n");
 367         return AVERROR_INVALIDDATA;
 368     }
 369
 370     if (rf64) {
 371         if (avio_rl32(pb) != MKTAG('d', 's', '6', '4'))
 372             return AVERROR_INVALIDDATA;
 373         size = avio_rl32(pb);
 374         if (size < 24)
 375             return AVERROR_INVALIDDATA;
 376         avio_rl64(pb); /* RIFF size */
 377
 378         data_size    = avio_rl64(pb);
 379         sample_count = avio_rl64(pb);
 380
 381         if (data_size < 0 || sample_count < 0) {
 382             av_log(s, AV_LOG_ERROR, "negative data_size and/or sample_count in "
 383                    "ds64: data_size = %"PRId64", sample_count = %"PRId64"\n",
 384                    data_size, sample_count);
 385             return AVERROR_INVALIDDATA;
 386         }
 387         avio_skip(pb, size - 24); /* skip rest of ds64 chunk */
 388
 389     }
 390
 391     for (;;) {
 392         AVStream *vst;
 393         size         = next_tag(pb, &tag, wav->rifx);
 394         next_tag_ofs = avio_tell(pb) + size;
 395
 396         if (avio_feof(pb))
 397             break;
 398
 399         switch (tag) {
 400         case MKTAG('f', 'm', 't', ' '):
 401             /* only parse the first 'fmt ' tag found */
 402             if (!got_xma2 && !got_fmt && (ret = wav_parse_fmt_tag(s, size, &st)) < 0) {
 403                 return ret;
 404             } else if (got_fmt)
 405                 av_log(s, AV_LOG_WARNING, "found more than one 'fmt ' tag\n");
 406
 407             got_fmt = 1;
 408             break;
 409         case MKTAG('X', 'M', 'A', '2'):
 410             /* only parse the first 'XMA2' tag found */
 411             if (!got_fmt && !got_xma2 && (ret = wav_parse_xma2_tag(s, size, &st)) < 0) {
 412                 return ret;
 413             } else if (got_xma2)
 414                 av_log(s, AV_LOG_WARNING, "found more than one 'XMA2' tag\n");
 415
 416             got_xma2 = 1;
 417             break;
 418         case MKTAG('d', 'a', 't', 'a'):
 419             if (!pb->seekable && !got_fmt && !got_xma2) {
 420                 av_log(s, AV_LOG_ERROR,
 421                        "found no 'fmt ' tag before the 'data' tag\n");
 422                 return AVERROR_INVALIDDATA;
 423             }
 424
 425             if (rf64) {
 426                 next_tag_ofs = wav->data_end = avio_tell(pb) + data_size;
 427             } else if (size != 0xFFFFFFFF) {
 428                 data_size    = size;
 429                 next_tag_ofs = wav->data_end = size ? next_tag_ofs : INT64_MAX;
 430             } else {
 431                 av_log(s, AV_LOG_WARNING, "Ignoring maximum wav data size, "
 432                        "file may be invalid\n");
 433                 data_size    = 0;
 434                 next_tag_ofs = wav->data_end = INT64_MAX;
 435             }
 436
 437             data_ofs = avio_tell(pb);
 438
 439             /* don't look for footer metadata if we can't seek or if we don't
 440              * know where the data tag ends
 441              */
 442             if (!pb->seekable || (!rf64 && !size))
 443                 goto break_loop;
 444             break;
 445         case MKTAG('f', 'a', 'c', 't'):
 446             if (!sample_count)
 447                 sample_count = (!wav->rifx ? avio_rl32(pb) : avio_rb32(pb));
 448             break;
 449         case MKTAG('b', 'e', 'x', 't'):
 450             if ((ret = wav_parse_bext_tag(s, size)) < 0)
 451                 return ret;
 452             break;
 453         case MKTAG('S','M','V','0'):
 454             if (!got_fmt) {
 455                 av_log(s, AV_LOG_ERROR, "found no 'fmt ' tag before the 'SMV0' tag\n");
 456                 return AVERROR_INVALIDDATA;
 457             }
 458             // SMV file, a wav file with video appended.
 459             if (size != MKTAG('0','2','0','0')) {
 460                 av_log(s, AV_LOG_ERROR, "Unknown SMV version found\n");
 461                 goto break_loop;
 462             }
 463             av_log(s, AV_LOG_DEBUG, "Found SMV data\n");
 464             wav->smv_given_first = 0;
 465             vst = avformat_new_stream(s, NULL);
 466             if (!vst)
 467                 return AVERROR(ENOMEM);
 468             avio_r8(pb);
 469             vst->id = 1;
 470             vst->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
 471             vst->codecpar->codec_id = AV_CODEC_ID_SMVJPEG;
 472             vst->codecpar->width  = avio_rl24(pb);
 473             vst->codecpar->height = avio_rl24(pb);
 474             if (ff_alloc_extradata(vst->codecpar, 4)) {
 475                 av_log(s, AV_LOG_ERROR, "Could not allocate extradata.\n");
 476                 return AVERROR(ENOMEM);
 477             }
 478             size = avio_rl24(pb);
 479             wav->smv_data_ofs = avio_tell(pb) + (size - 5) * 3;
 480             avio_rl24(pb);
 481             wav->smv_block_size = avio_rl24(pb);
 482             if (!wav->smv_block_size)
 483                 return AVERROR_INVALIDDATA;
 484             avpriv_set_pts_info(vst, 32, 1, avio_rl24(pb));
 485             vst->duration = avio_rl24(pb);
 486             avio_rl24(pb);
 487             avio_rl24(pb);
 488             wav->smv_frames_per_jpeg = avio_rl24(pb);
 489             if (wav->smv_frames_per_jpeg > 65536) {
 490                 av_log(s, AV_LOG_ERROR, "too many frames per jpeg\n");
 491                 return AVERROR_INVALIDDATA;
 492             }
 493             AV_WL32(vst->codecpar->extradata, wav->smv_frames_per_jpeg);
 494             wav->smv_cur_pt = 0;
 495             goto break_loop;
 496         case MKTAG('L', 'I', 'S', 'T'):
 497             if (size < 4) {
 498                 av_log(s, AV_LOG_ERROR, "too short LIST tag\n");
 499                 return AVERROR_INVALIDDATA;
 500             }
 501             switch (avio_rl32(pb)) {
 502             case MKTAG('I', 'N', 'F', 'O'):
 503                 ff_read_riff_info(s, size - 4);
 504             }
 505             break;
 506         }
 507
 508         /* seek to next tag unless we know that we'll run into EOF */
 509         if ((avio_size(pb) > 0 && next_tag_ofs >= avio_size(pb)) ||
 510             wav_seek_tag(wav, pb, next_tag_ofs, SEEK_SET) < 0) {
 511             break;
 512         }
 513     }
 514
 515 break_loop:
 516     if (!got_fmt && !got_xma2) {
 517         av_log(s, AV_LOG_ERROR, "no 'fmt ' or 'XMA2' tag found\n");
 518         return AVERROR_INVALIDDATA;
 519     }
 520
 521     if (data_ofs < 0) {
 522         av_log(s, AV_LOG_ERROR, "no 'data' tag found\n");
 523         return AVERROR_INVALIDDATA;
 524     }
 525
 526     avio_seek(pb, data_ofs, SEEK_SET);
 527
 528     if (data_size > (INT64_MAX>>3)) {
 529         av_log(s, AV_LOG_WARNING, "Data size %"PRId64" is too large\n", data_size);
 530         data_size = 0;
 531     }
 532
 533     if (   st->codecpar->bit_rate > 0 && data_size > 0
 534         && st->codecpar->sample_rate > 0
 535         && sample_count > 0 && st->codecpar->channels > 1
 536         && sample_count % st->codecpar->channels == 0) {
 537         if (fabs(8.0 * data_size * st->codecpar->channels * st->codecpar->sample_rate /
 538             sample_count /st->codecpar->bit_rate - 1.0) < 0.3)
 539             sample_count /= st->codecpar->channels;
 540     }
 541
 542     if (   data_size > 0 && sample_count && st->codecpar->channels
 543         && (data_size << 3) / sample_count / st->codecpar->channels > st->codecpar->bits_per_coded_sample  + 1) {
 544         av_log(s, AV_LOG_WARNING, "ignoring wrong sample_count %"PRId64"\n", sample_count);
 545         sample_count = 0;
 546     }
 547
 548     /* G.729 hack (for Ticket4577)
 549      * FIXME: Come up with cleaner, more general solution */
 550     if (st->codecpar->codec_id == AV_CODEC_ID_G729 && sample_count && (data_size << 3) > sample_count) {
 551         av_log(s, AV_LOG_WARNING, "ignoring wrong sample_count %"PRId64"\n", sample_count);
 552         sample_count = 0;
 553     }
 554
 555     if (!sample_count || av_get_exact_bits_per_sample(st->codecpar->codec_id) > 0)
 556         if (   st->codecpar->channels
 557             && data_size
 558             && av_get_bits_per_sample(st->codecpar->codec_id)
 559             && wav->data_end <= avio_size(pb))
 560             sample_count = (data_size << 3)
 561                                   /
 562                 (st->codecpar->channels * (uint64_t)av_get_bits_per_sample(st->codecpar->codec_id));
 563
 564     if (sample_count)
 565         st->duration = sample_count;
 566
 567     ff_metadata_conv_ctx(s, NULL, wav_metadata_conv);
 568     ff_metadata_conv_ctx(s, NULL, ff_riff_info_conv);
 569
 570     set_spdif(s, wav);
 571
 572     return 0;
 573 }
 574
 575 /**
 576  * Find chunk with w64 GUID by skipping over other chunks.
 577  * @return the size of the found chunk
 578  */
 579 static int64_t find_guid(AVIOContext *pb, const uint8_t guid1[16])
 580 {
 581     uint8_t guid[16];
 582     int64_t size;
 583
 584     while (!avio_feof(pb)) {
 585         avio_read(pb, guid, 16);
 586         size = avio_rl64(pb);
 587         if (size <= 24 || size > INT64_MAX - 8)
 588             return AVERROR_INVALIDDATA;
 589         if (!memcmp(guid, guid1, 16))
 590             return size;
 591         avio_skip(pb, FFALIGN(size, INT64_C(8)) - 24);
 592     }
 593     return AVERROR_EOF;
 594 }
 595
 596 #define MAX_SIZE 4096
 597
 598 static int wav_read_packet(AVFormatContext *s, AVPacket *pkt)
 599 {
 600     int ret, size;
 601     int64_t left;
 602     AVStream *st;
 603     WAVDemuxContext *wav = s->priv_data;
 604
 605     if (CONFIG_SPDIF_DEMUXER && wav->spdif == 1)
 606         return ff_spdif_read_packet(s, pkt);
 607
 608     if (wav->smv_data_ofs > 0) {
 609         int64_t audio_dts, video_dts;
 610 smv_retry:
 611         audio_dts = (int32_t)s->streams[0]->cur_dts;
 612         video_dts = (int32_t)s->streams[1]->cur_dts;
 613
 614         if (audio_dts != AV_NOPTS_VALUE && video_dts != AV_NOPTS_VALUE) {
 615             /*We always return a video frame first to get the pixel format first*/
 616             wav->smv_last_stream = wav->smv_given_first ?
 617                 av_compare_ts(video_dts, s->streams[1]->time_base,
 618                               audio_dts, s->streams[0]->time_base) > 0 : 0;
 619             wav->smv_given_first = 1;
 620         }
 621         wav->smv_last_stream = !wav->smv_last_stream;
 622         wav->smv_last_stream |= wav->audio_eof;
 623         wav->smv_last_stream &= !wav->smv_eof;
 624         if (wav->smv_last_stream) {
 625             uint64_t old_pos = avio_tell(s->pb);
 626             uint64_t new_pos = wav->smv_data_ofs +
 627                 wav->smv_block * (int64_t)wav->smv_block_size;
 628             if (avio_seek(s->pb, new_pos, SEEK_SET) < 0) {
 629                 ret = AVERROR_EOF;
 630                 goto smv_out;
 631             }
 632             size = avio_rl24(s->pb);
 633             ret  = av_get_packet(s->pb, pkt, size);
 634             if (ret < 0)
 635                 goto smv_out;
 636             pkt->pos -= 3;
 637             pkt->pts = wav->smv_block * wav->smv_frames_per_jpeg + wav->smv_cur_pt;
 638             wav->smv_cur_pt++;
 639             if (wav->smv_frames_per_jpeg > 0)
 640                 wav->smv_cur_pt %= wav->smv_frames_per_jpeg;
 641             if (!wav->smv_cur_pt)
 642                 wav->smv_block++;
 643
 644             pkt->stream_index = 1;
 645 smv_out:
 646             avio_seek(s->pb, old_pos, SEEK_SET);
 647             if (ret == AVERROR_EOF) {
 648                 wav->smv_eof = 1;
 649                 goto smv_retry;
 650             }
 651             return ret;
 652         }
 653     }
 654
 655     st = s->streams[0];
 656
 657     left = wav->data_end - avio_tell(s->pb);
 658     if (wav->ignore_length)
 659         left = INT_MAX;
 660     if (left <= 0) {
 661         if (CONFIG_W64_DEMUXER && wav->w64)
 662             left = find_guid(s->pb, ff_w64_guid_data) - 24;
 663         else
 664             left = find_tag(wav, s->pb, MKTAG('d', 'a', 't', 'a'));
 665         if (left < 0) {
 666             wav->audio_eof = 1;
 667             if (wav->smv_data_ofs > 0 && !wav->smv_eof)
 668                 goto smv_retry;
 669             return AVERROR_EOF;
 670         }
 671         wav->data_end = avio_tell(s->pb) + left;
 672     }
 673
 674     size = MAX_SIZE;
 675     if (st->codecpar->block_align > 1) {
 676         if (size < st->codecpar->block_align)
 677             size = st->codecpar->block_align;
 678         size = (size / st->codecpar->block_align) * st->codecpar->block_align;
 679     }
 680     size = FFMIN(size, left);
 681     ret  = av_get_packet(s->pb, pkt, size);
 682     if (ret < 0)
 683         return ret;
 684     pkt->stream_index = 0;
 685
 686     return ret;
 687 }
 688
 689 static int wav_read_seek(AVFormatContext *s,
 690                          int stream_index, int64_t timestamp, int flags)
 691 {
 692     WAVDemuxContext *wav = s->priv_data;
 693     AVStream *st;
 694     wav->smv_eof = 0;
 695     wav->audio_eof = 0;
 696     if (wav->smv_data_ofs > 0) {
 697         int64_t smv_timestamp = timestamp;
 698         if (stream_index == 0)
 699             smv_timestamp = av_rescale_q(timestamp, s->streams[0]->time_base, s->streams[1]->time_base);
 700         else
 701             timestamp = av_rescale_q(smv_timestamp, s->streams[1]->time_base, s->streams[0]->time_base);
 702         if (wav->smv_frames_per_jpeg > 0) {
 703             wav->smv_block = smv_timestamp / wav->smv_frames_per_jpeg;
 704             wav->smv_cur_pt = smv_timestamp % wav->smv_frames_per_jpeg;
 705         }
 706     }
 707
 708     st = s->streams[0];
 709     switch (st->codecpar->codec_id) {
 710     case AV_CODEC_ID_MP2:
 711     case AV_CODEC_ID_MP3:
 712     case AV_CODEC_ID_AC3:
 713     case AV_CODEC_ID_DTS:
 714         /* use generic seeking with dynamically generated indexes */
 715         return -1;
 716     default:
 717         break;
 718     }
 719     return ff_pcm_read_seek(s, stream_index, timestamp, flags);
 720 }
 721
 722 #define OFFSET(x) offsetof(WAVDemuxContext, x)
 723 #define DEC AV_OPT_FLAG_DECODING_PARAM
 724 static const AVOption demux_options[] = {
 725     { "ignore_length", "Ignore length", OFFSET(ignore_length), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, DEC },
 726     { NULL },
 727 };
 728
 729 static const AVClass wav_demuxer_class = {
 730     .class_name = "WAV demuxer",
 731     .item_name  = av_default_item_name,
 732     .option     = demux_options,
 733     .version    = LIBAVUTIL_VERSION_INT,
 734 };
 735 AVInputFormat ff_wav_demuxer = {
 736     .name           = "wav",
 737     .long_name      = NULL_IF_CONFIG_SMALL("WAV / WAVE (Waveform Audio)"),
 738     .priv_data_size = sizeof(WAVDemuxContext),
 739     .read_probe     = wav_probe,
 740     .read_header    = wav_read_header,
 741     .read_packet    = wav_read_packet,
 742     .read_seek      = wav_read_seek,
 743     .flags          = AVFMT_GENERIC_INDEX,
 744     .codec_tag      = (const AVCodecTag * const []) { ff_codec_wav_tags,  0 },
 745     .priv_class     = &wav_demuxer_class,
 746 };
 747 #endif /* CONFIG_WAV_DEMUXER */
 748
 749 #if CONFIG_W64_DEMUXER
 750 static int w64_probe(AVProbeData *p)
 751 {
 752     if (p->buf_size <= 40)
 753         return 0;
 754     if (!memcmp(p->buf,      ff_w64_guid_riff, 16) &&
 755         !memcmp(p->buf + 24, ff_w64_guid_wave, 16))
 756         return AVPROBE_SCORE_MAX;
 757     else
 758         return 0;
 759 }
 760
 761 static int w64_read_header(AVFormatContext *s)
 762 {
 763     int64_t size, data_ofs = 0;
 764     AVIOContext *pb      = s->pb;
 765     WAVDemuxContext *wav = s->priv_data;
 766     AVStream *st;
 767     uint8_t guid[16];
 768     int ret;
 769
 770     avio_read(pb, guid, 16);
 771     if (memcmp(guid, ff_w64_guid_riff, 16))
 772         return AVERROR_INVALIDDATA;
 773
 774     /* riff + wave + fmt + sizes */
 775     if (avio_rl64(pb) < 16 + 8 + 16 + 8 + 16 + 8)
 776         return AVERROR_INVALIDDATA;
 777
 778     avio_read(pb, guid, 16);
 779     if (memcmp(guid, ff_w64_guid_wave, 16)) {
 780         av_log(s, AV_LOG_ERROR, "could not find wave guid\n");
 781         return AVERROR_INVALIDDATA;
 782     }
 783
 784     wav->w64 = 1;
 785
 786     st = avformat_new_stream(s, NULL);
 787     if (!st)
 788         return AVERROR(ENOMEM);
 789
 790     while (!avio_feof(pb)) {
 791         if (avio_read(pb, guid, 16) != 16)
 792             break;
 793         size = avio_rl64(pb);
 794         if (size <= 24 || INT64_MAX - size < avio_tell(pb))
 795             return AVERROR_INVALIDDATA;
 796
 797         if (!memcmp(guid, ff_w64_guid_fmt, 16)) {
 798             /* subtract chunk header size - normal wav file doesn't count it */
 799             ret = ff_get_wav_header(s, pb, st->codecpar, size - 24, 0);
 800             if (ret < 0)
 801                 return ret;
 802             avio_skip(pb, FFALIGN(size, INT64_C(8)) - size);
 803
 804             avpriv_set_pts_info(st, 64, 1, st->codecpar->sample_rate);
 805         } else if (!memcmp(guid, ff_w64_guid_fact, 16)) {
 806             int64_t samples;
 807
 808             samples = avio_rl64(pb);
 809             if (samples > 0)
 810                 st->duration = samples;
 811         } else if (!memcmp(guid, ff_w64_guid_data, 16)) {
 812             wav->data_end = avio_tell(pb) + size - 24;
 813
 814             data_ofs = avio_tell(pb);
 815             if (!pb->seekable)
 816                 break;
 817
 818             avio_skip(pb, size - 24);
 819         } else if (!memcmp(guid, ff_w64_guid_summarylist, 16)) {
 820             int64_t start, end, cur;
 821             uint32_t count, chunk_size, i;
 822             int64_t filesize  = avio_size(s->pb);
 823
 824             start = avio_tell(pb);
 825             end = start + FFALIGN(size, INT64_C(8)) - 24;
 826             count = avio_rl32(pb);
 827
 828             for (i = 0; i < count; i++) {
 829                 char chunk_key[5], *value;
 830
 831                 if (avio_feof(pb) || (cur = avio_tell(pb)) < 0 || cur > end - 8 /* = tag + size */)
 832                     break;
 833
 834                 chunk_key[4] = 0;
 835                 avio_read(pb, chunk_key, 4);
 836                 chunk_size = avio_rl32(pb);
 837                 if (chunk_size == UINT32_MAX || (filesize >= 0 && chunk_size > filesize))
 838                     return AVERROR_INVALIDDATA;
 839
 840                 value = av_mallocz(chunk_size + 1);
 841                 if (!value)
 842                     return AVERROR(ENOMEM);
 843
 844                 ret = avio_get_str16le(pb, chunk_size, value, chunk_size);
 845                 if (ret < 0) {
 846                     av_free(value);
 847                     return ret;
 848                 }
 849                 avio_skip(pb, chunk_size - ret);
 850
 851                 av_dict_set(&s->metadata, chunk_key, value, AV_DICT_DONT_STRDUP_VAL);
 852             }
 853
 854             avio_skip(pb, end - avio_tell(pb));
 855         } else {
 856             av_log(s, AV_LOG_DEBUG, "unknown guid: "FF_PRI_GUID"\n", FF_ARG_GUID(guid));
 857             avio_skip(pb, FFALIGN(size, INT64_C(8)) - 24);
 858         }
 859     }
 860
 861     if (!data_ofs)
 862         return AVERROR_EOF;
 863
 864     ff_metadata_conv_ctx(s, NULL, wav_metadata_conv);
 865     ff_metadata_conv_ctx(s, NULL, ff_riff_info_conv);
 866
 867     handle_stream_probing(st);
 868     st->need_parsing = AVSTREAM_PARSE_FULL_RAW;
 869
 870     avio_seek(pb, data_ofs, SEEK_SET);
 871
 872     set_spdif(s, wav);
 873
 874     return 0;
 875 }
 876
 877 AVInputFormat ff_w64_demuxer = {
 878     .name           = "w64",
 879     .long_name      = NULL_IF_CONFIG_SMALL("Sony Wave64"),
 880     .priv_data_size = sizeof(WAVDemuxContext),
 881     .read_probe     = w64_probe,
 882     .read_header    = w64_read_header,
 883     .read_packet    = wav_read_packet,
 884     .read_seek      = wav_read_seek,
 885     .flags          = AVFMT_GENERIC_INDEX,
 886     .codec_tag      = (const AVCodecTag * const []) { ff_codec_wav_tags, 0 },
 887 };
 888 #endif /* CONFIG_W64_DEMUXER */