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  * BW64 demuxer
  10  *
  11  * This file is part of FFmpeg.
  12  *
  13  * FFmpeg is free software; you can redistribute it and/or
  14  * modify it under the terms of the GNU Lesser General Public
  15  * License as published by the Free Software Foundation; either
  16  * version 2.1 of the License, or (at your option) any later version.
  17  *
  18  * FFmpeg is distributed in the hope that it will be useful,
  19  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  20  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  21  * Lesser General Public License for more details.
  22  *
  23  * You should have received a copy of the GNU Lesser General Public
  24  * License along with FFmpeg; if not, write to the Free Software
  25  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  26  */
  27
  28 #include <stdint.h>
  29
  30 #include "config_components.h"
  31 #include "libavutil/avassert.h"
  32 #include "libavutil/dict.h"
  33 #include "libavutil/intreadwrite.h"
  34 #include "libavutil/log.h"
  35 #include "libavutil/mathematics.h"
  36 #include "libavutil/mem.h"
  37 #include "libavutil/opt.h"
  38 #include "libavcodec/internal.h"
  39 #include "avformat.h"
  40 #include "avio.h"
  41 #include "avio_internal.h"
  42 #include "demux.h"
  43 #include "id3v2.h"
  44 #include "internal.h"
  45 #include "metadata.h"
  46 #include "pcm.h"
  47 #include "riff.h"
  48 #include "w64.h"
  49 #include "spdif.h"
  50
  51 typedef struct WAVDemuxContext {
  52     const AVClass *class;
  53     int64_t data_end;
  54     int w64;
  55     AVStream *vst;
  56     int64_t smv_data_ofs;
  57     int smv_block_size;
  58     int smv_frames_per_jpeg;
  59     int smv_block;
  60     int smv_last_stream;
  61     int smv_eof;
  62     int audio_eof;
  63     int ignore_length;
  64     int max_size;
  65     int spdif;
  66     int smv_given_first;
  67     int unaligned; // e.g. if an odd number of bytes ID3 tag was prepended
  68     int rifx; // RIFX: integer byte order for parameters is big endian
  69 } WAVDemuxContext;
  70
  71 #define OFFSET(x) offsetof(WAVDemuxContext, x)
  72 #define DEC AV_OPT_FLAG_DECODING_PARAM
  73 static const AVOption demux_options[] = {
  74 #define W64_DEMUXER_OPTIONS_OFFSET (1 * CONFIG_WAV_DEMUXER)
  75 #if CONFIG_WAV_DEMUXER
  76     { "ignore_length", "Ignore length", OFFSET(ignore_length), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, DEC },
  77 #endif
  78     { "max_size",      "max size of single packet", OFFSET(max_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1 << 22, DEC },
  79     { NULL },
  80 };
  81
  82 static void set_max_size(AVStream *st, WAVDemuxContext *wav)
  83 {
  84     if (wav->max_size <= 0) {
  85         int max_size = ff_pcm_default_packet_size(st->codecpar);
  86         wav->max_size = max_size < 0 ? 4096 : max_size;
  87     }
  88 }
  89
  90 static void set_spdif(AVFormatContext *s, WAVDemuxContext *wav)
  91 {
  92     if (CONFIG_SPDIF_DEMUXER && s->streams[0]->codecpar->codec_tag == 1) {
  93         enum AVCodecID codec;
  94         int len = 1<<16;
  95         int ret = ffio_ensure_seekback(s->pb, len);
  96
  97         if (ret >= 0) {
  98             uint8_t *buf = av_malloc(len + AV_INPUT_BUFFER_PADDING_SIZE);
  99             if (!buf) {
 100                 ret = AVERROR(ENOMEM);
 101             } else {
 102                 int64_t pos = avio_tell(s->pb);
 103                 len = ret = avio_read(s->pb, buf, len);
 104                 if (len >= 0) {
 105                     ret = ff_spdif_probe(buf, len, &codec);
 106                     if (ret > AVPROBE_SCORE_EXTENSION) {
 107                         s->streams[0]->codecpar->codec_id = codec;
 108                         wav->spdif = 1;
 109                     }
 110                 }
 111                 avio_seek(s->pb, pos, SEEK_SET);
 112                 av_free(buf);
 113             }
 114         }
 115
 116         if (ret < 0)
 117             av_log(s, AV_LOG_WARNING, "Cannot check for SPDIF\n");
 118     }
 119 }
 120
 121 #if CONFIG_WAV_DEMUXER
 122
 123 static int64_t next_tag(AVIOContext *pb, uint32_t *tag, int big_endian)
 124 {
 125     *tag = avio_rl32(pb);
 126     if (!big_endian) {
 127         return avio_rl32(pb);
 128     } else {
 129         return avio_rb32(pb);
 130     }
 131 }
 132
 133 /* RIFF chunks are always at even offsets relative to where they start. */
 134 static int64_t wav_seek_tag(WAVDemuxContext * wav, AVIOContext *s, int64_t offset, int whence)
 135 {
 136     offset += offset < INT64_MAX && offset + wav->unaligned & 1;
 137
 138     return avio_seek(s, offset, whence);
 139 }
 140
 141 /* return the size of the found tag */
 142 static int64_t find_tag(WAVDemuxContext * wav, AVIOContext *pb, uint32_t tag1)
 143 {
 144     unsigned int tag;
 145     int64_t size;
 146
 147     for (;;) {
 148         if (avio_feof(pb))
 149             return AVERROR_EOF;
 150         size = next_tag(pb, &tag, wav->rifx);
 151         if (tag == tag1)
 152             break;
 153         wav_seek_tag(wav, pb, size, SEEK_CUR);
 154     }
 155     return size;
 156 }
 157
 158 static int wav_probe(const AVProbeData *p)
 159 {
 160     /* check file header */
 161     if (p->buf_size <= 32)
 162         return 0;
 163     if (!memcmp(p->buf + 8, "WAVE", 4)) {
 164         if (!memcmp(p->buf, "RIFF", 4) || !memcmp(p->buf, "RIFX", 4))
 165             /* Since the ACT demuxer has a standard WAV header at the top of
 166              * its own, the returned score is decreased to avoid a probe
 167              * conflict between ACT and WAV. */
 168             return AVPROBE_SCORE_MAX - 1;
 169         else if ((!memcmp(p->buf,      "RF64", 4) ||
 170                   !memcmp(p->buf,      "BW64", 4)) &&
 171                  !memcmp(p->buf + 12, "ds64", 4))
 172             return AVPROBE_SCORE_MAX;
 173     }
 174     return 0;
 175 }
 176
 177 static void handle_stream_probing(AVStream *st)
 178 {
 179     if (st->codecpar->codec_id == AV_CODEC_ID_PCM_S16LE) {
 180         FFStream *const sti = ffstream(st);
 181         sti->request_probe = AVPROBE_SCORE_EXTENSION;
 182         sti->probe_packets = FFMIN(sti->probe_packets, 32);
 183     }
 184 }
 185
 186 static int wav_parse_fmt_tag(AVFormatContext *s, int64_t size, AVStream *st)
 187 {
 188     AVIOContext *pb = s->pb;
 189     WAVDemuxContext *wav = s->priv_data;
 190     int ret;
 191
 192     /* parse fmt header */
 193     ret = ff_get_wav_header(s, pb, st->codecpar, size, wav->rifx);
 194     if (ret < 0)
 195         return ret;
 196     handle_stream_probing(st);
 197
 198     ffstream(st)->need_parsing = AVSTREAM_PARSE_FULL_RAW;
 199
 200     avpriv_set_pts_info(st, 64, 1, st->codecpar->sample_rate);
 201
 202     return 0;
 203 }
 204
 205 static int wav_parse_xma2_tag(AVFormatContext *s, int64_t size, AVStream *st)
 206 {
 207     AVIOContext *pb = s->pb;
 208     int version, num_streams, i, channels = 0, ret;
 209
 210     if (size < 36)
 211         return AVERROR_INVALIDDATA;
 212
 213     st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
 214     st->codecpar->codec_id   = AV_CODEC_ID_XMA2;
 215     ffstream(st)->need_parsing = AVSTREAM_PARSE_FULL_RAW;
 216
 217     version = avio_r8(pb);
 218     if (version != 3 && version != 4)
 219         return AVERROR_INVALIDDATA;
 220     num_streams = avio_r8(pb);
 221     if (size != (32 + ((version==3)?0:8) + 4*num_streams))
 222         return AVERROR_INVALIDDATA;
 223     avio_skip(pb, 10);
 224     st->codecpar->sample_rate = avio_rb32(pb);
 225     if (version == 4)
 226         avio_skip(pb, 8);
 227     avio_skip(pb, 4);
 228     st->duration = avio_rb32(pb);
 229     avio_skip(pb, 8);
 230
 231     for (i = 0; i < num_streams; i++) {
 232         channels += avio_r8(pb);
 233         avio_skip(pb, 3);
 234     }
 235     av_channel_layout_uninit(&st->codecpar->ch_layout);
 236     st->codecpar->ch_layout.order       = AV_CHANNEL_ORDER_UNSPEC;
 237     st->codecpar->ch_layout.nb_channels = channels;
 238
 239     if (st->codecpar->ch_layout.nb_channels <= 0 || st->codecpar->sample_rate <= 0)
 240         return AVERROR_INVALIDDATA;
 241
 242     avpriv_set_pts_info(st, 64, 1, st->codecpar->sample_rate);
 243
 244     avio_seek(pb, -size, SEEK_CUR);
 245     if ((ret = ff_get_extradata(s, st->codecpar, pb, size)) < 0)
 246         return ret;
 247
 248     return 0;
 249 }
 250
 251 static inline int wav_parse_bext_string(AVFormatContext *s, const char *key,
 252                                         int length)
 253 {
 254     char temp[257];
 255     int ret;
 256
 257     av_assert0(length < sizeof(temp));
 258     if ((ret = ffio_read_size(s->pb, temp, length)) < 0)
 259         return ret;
 260
 261     temp[length] = 0;
 262
 263     if (strlen(temp))
 264         return av_dict_set(&s->metadata, key, temp, 0);
 265
 266     return 0;
 267 }
 268
 269 static int wav_parse_bext_tag(AVFormatContext *s, int64_t size)
 270 {
 271     char temp[131], *coding_history;
 272     int ret, x;
 273     uint64_t time_reference;
 274     int64_t umid_parts[8], umid_mask = 0;
 275
 276     if ((ret = wav_parse_bext_string(s, "description", 256)) < 0 ||
 277         (ret = wav_parse_bext_string(s, "originator", 32)) < 0 ||
 278         (ret = wav_parse_bext_string(s, "originator_reference", 32)) < 0 ||
 279         (ret = wav_parse_bext_string(s, "origination_date", 10)) < 0 ||
 280         (ret = wav_parse_bext_string(s, "origination_time", 8)) < 0)
 281         return ret;
 282
 283     time_reference = avio_rl64(s->pb);
 284     snprintf(temp, sizeof(temp), "%"PRIu64, time_reference);
 285     if ((ret = av_dict_set(&s->metadata, "time_reference", temp, 0)) < 0)
 286         return ret;
 287
 288     /* check if version is >= 1, in which case an UMID may be present */
 289     if (avio_rl16(s->pb) >= 1) {
 290         for (x = 0; x < 8; x++)
 291             umid_mask |= umid_parts[x] = avio_rb64(s->pb);
 292
 293         if (umid_mask) {
 294             /* the string formatting below is per SMPTE 330M-2004 Annex C */
 295             if (umid_parts[4] == 0 && umid_parts[5] == 0 &&
 296                 umid_parts[6] == 0 && umid_parts[7] == 0) {
 297                 /* basic UMID */
 298                 snprintf(temp, sizeof(temp),
 299                          "0x%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64,
 300                          umid_parts[0], umid_parts[1],
 301                          umid_parts[2], umid_parts[3]);
 302             } else {
 303                 /* extended UMID */
 304                 snprintf(temp, sizeof(temp),
 305                          "0x%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64
 306                          "%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64,
 307                          umid_parts[0], umid_parts[1],
 308                          umid_parts[2], umid_parts[3],
 309                          umid_parts[4], umid_parts[5],
 310                          umid_parts[6], umid_parts[7]);
 311             }
 312
 313             if ((ret = av_dict_set(&s->metadata, "umid", temp, 0)) < 0)
 314                 return ret;
 315         }
 316
 317         avio_skip(s->pb, 190);
 318     } else
 319         avio_skip(s->pb, 254);
 320
 321     if (size > 602) {
 322         /* CodingHistory present */
 323         size -= 602;
 324
 325         if (!(coding_history = av_malloc(size + 1)))
 326             return AVERROR(ENOMEM);
 327
 328         if ((ret = ffio_read_size(s->pb, coding_history, size)) < 0) {
 329             av_free(coding_history);
 330             return ret;
 331         }
 332
 333         coding_history[size] = 0;
 334         if ((ret = av_dict_set(&s->metadata, "coding_history", coding_history,
 335                                AV_DICT_DONT_STRDUP_VAL)) < 0)
 336             return ret;
 337     }
 338
 339     return 0;
 340 }
 341
 342 static const AVMetadataConv wav_metadata_conv[] = {
 343     { "description",      "comment"       },
 344     { "originator",       "encoded_by"    },
 345     { "origination_date", "date"          },
 346     { "origination_time", "creation_time" },
 347     { 0 },
 348 };
 349
 350 /* wav input */
 351 static int wav_read_header(AVFormatContext *s)
 352 {
 353     int64_t size, av_uninit(data_size);
 354     int64_t sample_count = 0;
 355     int rf64 = 0, bw64 = 0;
 356     uint32_t tag;
 357     AVIOContext *pb      = s->pb;
 358     AVStream *st         = NULL;
 359     WAVDemuxContext *wav = s->priv_data;
 360     int ret, got_fmt = 0, got_xma2 = 0;
 361     int64_t next_tag_ofs, data_ofs = -1;
 362
 363     wav->unaligned = avio_tell(s->pb) & 1;
 364
 365     wav->smv_data_ofs = -1;
 366
 367     /* read chunk ID */
 368     tag = avio_rl32(pb);
 369     switch (tag) {
 370     case MKTAG('R', 'I', 'F', 'F'):
 371         break;
 372     case MKTAG('R', 'I', 'F', 'X'):
 373         wav->rifx = 1;
 374         break;
 375     case MKTAG('R', 'F', '6', '4'):
 376         rf64 = 1;
 377         break;
 378     case MKTAG('B', 'W', '6', '4'):
 379         bw64 = 1;
 380         break;
 381     default:
 382         av_log(s, AV_LOG_ERROR, "invalid start code %s in RIFF header\n",
 383                av_fourcc2str(tag));
 384         return AVERROR_INVALIDDATA;
 385     }
 386
 387     /* read chunk size */
 388     avio_rl32(pb);
 389
 390     /* read format */
 391     if (avio_rl32(pb) != MKTAG('W', 'A', 'V', 'E')) {
 392         av_log(s, AV_LOG_ERROR, "invalid format in RIFF header\n");
 393         return AVERROR_INVALIDDATA;
 394     }
 395
 396     if (rf64 || bw64) {
 397         if (avio_rl32(pb) != MKTAG('d', 's', '6', '4'))
 398             return AVERROR_INVALIDDATA;
 399         size = avio_rl32(pb);
 400         if (size < 24)
 401             return AVERROR_INVALIDDATA;
 402         avio_rl64(pb); /* RIFF size */
 403
 404         data_size    = avio_rl64(pb);
 405         sample_count = avio_rl64(pb);
 406
 407         if (data_size < 0 || sample_count < 0) {
 408             av_log(s, AV_LOG_ERROR, "negative data_size and/or sample_count in "
 409                    "ds64: data_size = %"PRId64", sample_count = %"PRId64"\n",
 410                    data_size, sample_count);
 411             return AVERROR_INVALIDDATA;
 412         }
 413         avio_skip(pb, size - 24); /* skip rest of ds64 chunk */
 414
 415     }
 416
 417     /* Create the audio stream now so that its index is always zero */
 418     st = avformat_new_stream(s, NULL);
 419     if (!st)
 420         return AVERROR(ENOMEM);
 421
 422     for (;;) {
 423         AVStream *vst;
 424         size         = next_tag(pb, &tag, wav->rifx);
 425         next_tag_ofs = avio_tell(pb) + size;
 426
 427         if (avio_feof(pb))
 428             break;
 429
 430         switch (tag) {
 431         case MKTAG('f', 'm', 't', ' '):
 432             /* only parse the first 'fmt ' tag found */
 433             if (!got_xma2 && !got_fmt && (ret = wav_parse_fmt_tag(s, size, st)) < 0) {
 434                 return ret;
 435             } else if (got_fmt)
 436                 av_log(s, AV_LOG_WARNING, "found more than one 'fmt ' tag\n");
 437
 438             got_fmt = 1;
 439             break;
 440         case MKTAG('X', 'M', 'A', '2'):
 441             /* only parse the first 'XMA2' tag found */
 442             if (!got_fmt && !got_xma2 && (ret = wav_parse_xma2_tag(s, size, st)) < 0) {
 443                 return ret;
 444             } else if (got_xma2)
 445                 av_log(s, AV_LOG_WARNING, "found more than one 'XMA2' tag\n");
 446
 447             got_xma2 = 1;
 448             break;
 449         case MKTAG('d', 'a', 't', 'a'):
 450             if (!(pb->seekable & AVIO_SEEKABLE_NORMAL) && !got_fmt && !got_xma2) {
 451                 av_log(s, AV_LOG_ERROR,
 452                        "found no 'fmt ' tag before the 'data' tag\n");
 453                 return AVERROR_INVALIDDATA;
 454             }
 455
 456             if (rf64 || bw64) {
 457                 next_tag_ofs = wav->data_end = av_sat_add64(avio_tell(pb), data_size);
 458             } else if (size != 0xFFFFFFFF) {
 459                 data_size    = size;
 460                 next_tag_ofs = wav->data_end = size ? next_tag_ofs : INT64_MAX;
 461             } else {
 462                 av_log(s, AV_LOG_WARNING, "Ignoring maximum wav data size, "
 463                        "file may be invalid\n");
 464                 data_size    = 0;
 465                 next_tag_ofs = wav->data_end = INT64_MAX;
 466             }
 467
 468             data_ofs = avio_tell(pb);
 469
 470             /* don't look for footer metadata if we can't seek or if we don't
 471              * know where the data tag ends
 472              */
 473             if (!(pb->seekable & AVIO_SEEKABLE_NORMAL) || (!(rf64 && !bw64) && !size))
 474                 goto break_loop;
 475             break;
 476         case MKTAG('f', 'a', 'c', 't'):
 477             if (!sample_count)
 478                 sample_count = (!wav->rifx ? avio_rl32(pb) : avio_rb32(pb));
 479             break;
 480         case MKTAG('b', 'e', 'x', 't'):
 481             if ((ret = wav_parse_bext_tag(s, size)) < 0)
 482                 return ret;
 483             break;
 484         case MKTAG('S','M','V','0'):
 485             if (!got_fmt) {
 486                 av_log(s, AV_LOG_ERROR, "found no 'fmt ' tag before the 'SMV0' tag\n");
 487                 return AVERROR_INVALIDDATA;
 488             }
 489             // SMV file, a wav file with video appended.
 490             if (size != MKTAG('0','2','0','0')) {
 491                 av_log(s, AV_LOG_ERROR, "Unknown SMV version found\n");
 492                 goto break_loop;
 493             }
 494             av_log(s, AV_LOG_DEBUG, "Found SMV data\n");
 495             wav->smv_given_first = 0;
 496             vst = avformat_new_stream(s, NULL);
 497             if (!vst)
 498                 return AVERROR(ENOMEM);
 499             wav->vst = vst;
 500             avio_r8(pb);
 501             vst->id = 1;
 502             vst->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
 503             vst->codecpar->codec_id = AV_CODEC_ID_SMVJPEG;
 504             vst->codecpar->width  = avio_rl24(pb);
 505             vst->codecpar->height = avio_rl24(pb);
 506             if ((ret = ff_alloc_extradata(vst->codecpar, 4)) < 0) {
 507                 av_log(s, AV_LOG_ERROR, "Could not allocate extradata.\n");
 508                 return ret;
 509             }
 510             size = avio_rl24(pb);
 511             wav->smv_data_ofs = avio_tell(pb) + (size - 5) * 3;
 512             avio_rl24(pb);
 513             wav->smv_block_size = avio_rl24(pb);
 514             if (!wav->smv_block_size)
 515                 return AVERROR_INVALIDDATA;
 516             avpriv_set_pts_info(vst, 32, 1, avio_rl24(pb));
 517             vst->duration = avio_rl24(pb);
 518             avio_rl24(pb);
 519             avio_rl24(pb);
 520             wav->smv_frames_per_jpeg = avio_rl24(pb);
 521             if (wav->smv_frames_per_jpeg > 65536) {
 522                 av_log(s, AV_LOG_ERROR, "too many frames per jpeg\n");
 523                 return AVERROR_INVALIDDATA;
 524             }
 525             AV_WL32(vst->codecpar->extradata, wav->smv_frames_per_jpeg);
 526             goto break_loop;
 527         case MKTAG('L', 'I', 'S', 'T'):
 528         case MKTAG('l', 'i', 's', 't'):
 529             if (size < 4) {
 530                 av_log(s, AV_LOG_ERROR, "too short LIST tag\n");
 531                 return AVERROR_INVALIDDATA;
 532             }
 533             switch (avio_rl32(pb)) {
 534             case MKTAG('I', 'N', 'F', 'O'):
 535                 ff_read_riff_info(s, size - 4);
 536                 break;
 537             case MKTAG('a', 'd', 't', 'l'):
 538                 if (s->nb_chapters > 0) {
 539                     while (avio_tell(pb) < next_tag_ofs &&
 540                            !avio_feof(pb)) {
 541                         char cue_label[512];
 542                         unsigned id, sub_size;
 543
 544                         if (avio_rl32(pb) != MKTAG('l', 'a', 'b', 'l'))
 545                             break;
 546
 547                         sub_size = avio_rl32(pb);
 548                         if (sub_size < 5)
 549                             break;
 550                         id       = avio_rl32(pb);
 551                         avio_get_str(pb, sub_size - 4, cue_label, sizeof(cue_label));
 552                         avio_skip(pb, avio_tell(pb) & 1);
 553
 554                         for (int i = 0; i < s->nb_chapters; i++) {
 555                             if (s->chapters[i]->id == id) {
 556                                 av_dict_set(&s->chapters[i]->metadata, "title", cue_label, 0);
 557                                 break;
 558                             }
 559                         }
 560                     }
 561                 }
 562                 break;
 563             }
 564             break;
 565         case MKTAG('I', 'D', '3', ' '):
 566         case MKTAG('i', 'd', '3', ' '): {
 567             ID3v2ExtraMeta *id3v2_extra_meta;
 568             ff_id3v2_read(s, ID3v2_DEFAULT_MAGIC, &id3v2_extra_meta, 0);
 569             if (id3v2_extra_meta) {
 570                 ff_id3v2_parse_apic(s, id3v2_extra_meta);
 571                 ff_id3v2_parse_chapters(s, id3v2_extra_meta);
 572                 ff_id3v2_parse_priv(s, id3v2_extra_meta);
 573             }
 574             ff_id3v2_free_extra_meta(&id3v2_extra_meta);
 575             }
 576             break;
 577         case MKTAG('c', 'u', 'e', ' '):
 578             if (size >= 4 && got_fmt && st->codecpar->sample_rate > 0) {
 579                 AVRational tb = {1, st->codecpar->sample_rate};
 580                 unsigned nb_cues = avio_rl32(pb);
 581
 582                 if (size >= nb_cues * 24LL + 4LL) {
 583                     for (int i = 0; i < nb_cues; i++) {
 584                         unsigned offset, id = avio_rl32(pb);
 585
 586                         if (avio_feof(pb))
 587                             return AVERROR_INVALIDDATA;
 588
 589                         avio_skip(pb, 16);
 590                         offset = avio_rl32(pb);
 591
 592                         if (!avpriv_new_chapter(s, id, tb, offset, AV_NOPTS_VALUE, NULL))
 593                             return AVERROR(ENOMEM);
 594                     }
 595                 }
 596             }
 597             break;
 598         }
 599
 600         /* seek to next tag unless we know that we'll run into EOF */
 601         if ((avio_size(pb) > 0 && next_tag_ofs >= avio_size(pb)) ||
 602             wav_seek_tag(wav, pb, next_tag_ofs, SEEK_SET) < 0) {
 603             break;
 604         }
 605     }
 606
 607 break_loop:
 608     if (!got_fmt && !got_xma2) {
 609         av_log(s, AV_LOG_ERROR, "no 'fmt ' or 'XMA2' tag found\n");
 610         return AVERROR_INVALIDDATA;
 611     }
 612
 613     if (data_ofs < 0) {
 614         av_log(s, AV_LOG_ERROR, "no 'data' tag found\n");
 615         return AVERROR_INVALIDDATA;
 616     }
 617
 618     avio_seek(pb, data_ofs, SEEK_SET);
 619
 620     if (data_size > (INT64_MAX>>3)) {
 621         av_log(s, AV_LOG_WARNING, "Data size %"PRId64" is too large\n", data_size);
 622         data_size = 0;
 623     }
 624
 625     if (   st->codecpar->bit_rate > 0 && data_size > 0
 626         && st->codecpar->sample_rate > 0
 627         && sample_count > 0 && st->codecpar->ch_layout.nb_channels > 1
 628         && sample_count % st->codecpar->ch_layout.nb_channels == 0) {
 629         if (fabs(8.0 * data_size * st->codecpar->ch_layout.nb_channels * st->codecpar->sample_rate /
 630             sample_count /st->codecpar->bit_rate - 1.0) < 0.3)
 631             sample_count /= st->codecpar->ch_layout.nb_channels;
 632     }
 633
 634     if (data_size > 0 && sample_count && st->codecpar->ch_layout.nb_channels &&
 635         (data_size << 3) / sample_count / st->codecpar->ch_layout.nb_channels > st->codecpar->bits_per_coded_sample  + 1) {
 636         av_log(s, AV_LOG_WARNING, "ignoring wrong sample_count %"PRId64"\n", sample_count);
 637         sample_count = 0;
 638     }
 639
 640     /* G.729 hack (for Ticket4577)
 641      * FIXME: Come up with cleaner, more general solution */
 642     if (st->codecpar->codec_id == AV_CODEC_ID_G729 && sample_count && (data_size << 3) > sample_count) {
 643         av_log(s, AV_LOG_WARNING, "ignoring wrong sample_count %"PRId64"\n", sample_count);
 644         sample_count = 0;
 645     }
 646
 647     if (!sample_count || av_get_exact_bits_per_sample(st->codecpar->codec_id) > 0)
 648         if (   st->codecpar->ch_layout.nb_channels
 649             && data_size
 650             && av_get_bits_per_sample(st->codecpar->codec_id)
 651             && wav->data_end <= avio_size(pb))
 652             sample_count = (data_size << 3)
 653                                   /
 654                 (st->codecpar->ch_layout.nb_channels * (uint64_t)av_get_bits_per_sample(st->codecpar->codec_id));
 655
 656     if (sample_count)
 657         st->duration = sample_count;
 658
 659     if (st->codecpar->codec_id == AV_CODEC_ID_PCM_S32LE &&
 660         st->codecpar->block_align == st->codecpar->ch_layout.nb_channels * 4 &&
 661         st->codecpar->bits_per_coded_sample == 32 &&
 662         st->codecpar->extradata_size == 2 &&
 663         AV_RL16(st->codecpar->extradata) == 1) {
 664         st->codecpar->codec_id = AV_CODEC_ID_PCM_F16LE;
 665         st->codecpar->bits_per_coded_sample = 16;
 666     } else if (st->codecpar->codec_id == AV_CODEC_ID_PCM_S24LE &&
 667                st->codecpar->block_align == st->codecpar->ch_layout.nb_channels * 4 &&
 668                st->codecpar->bits_per_coded_sample == 24) {
 669         st->codecpar->codec_id = AV_CODEC_ID_PCM_F24LE;
 670     } else if (st->codecpar->codec_id == AV_CODEC_ID_XMA1 ||
 671                st->codecpar->codec_id == AV_CODEC_ID_XMA2) {
 672         st->codecpar->block_align = 2048;
 673     } else if (st->codecpar->codec_id == AV_CODEC_ID_ADPCM_MS && st->codecpar->ch_layout.nb_channels > 2 &&
 674                st->codecpar->block_align < INT_MAX / st->codecpar->ch_layout.nb_channels) {
 675         st->codecpar->block_align *= st->codecpar->ch_layout.nb_channels;
 676     }
 677
 678     ff_metadata_conv_ctx(s, NULL, wav_metadata_conv);
 679     ff_metadata_conv_ctx(s, NULL, ff_riff_info_conv);
 680
 681     set_spdif(s, wav);
 682     set_max_size(st, wav);
 683
 684     return 0;
 685 }
 686
 687 /**
 688  * Find chunk with w64 GUID by skipping over other chunks.
 689  * @return the size of the found chunk
 690  */
 691 static int64_t find_guid(AVIOContext *pb, const uint8_t guid1[16])
 692 {
 693     uint8_t guid[16];
 694     int64_t size;
 695
 696     while (!avio_feof(pb)) {
 697         avio_read(pb, guid, 16);
 698         size = avio_rl64(pb);
 699         if (size <= 24 || size > INT64_MAX - 8)
 700             return AVERROR_INVALIDDATA;
 701         if (!memcmp(guid, guid1, 16))
 702             return size;
 703         avio_skip(pb, FFALIGN(size, INT64_C(8)) - 24);
 704     }
 705     return AVERROR_EOF;
 706 }
 707
 708 static int wav_read_packet(AVFormatContext *s, AVPacket *pkt)
 709 {
 710     int ret, size;
 711     int64_t left;
 712     WAVDemuxContext *wav = s->priv_data;
 713     AVStream *st = s->streams[0];
 714
 715     if (CONFIG_SPDIF_DEMUXER && wav->spdif == 1)
 716         return ff_spdif_read_packet(s, pkt);
 717
 718     if (wav->smv_data_ofs > 0) {
 719         int64_t audio_dts, video_dts;
 720         AVStream *vst = wav->vst;
 721 smv_retry:
 722         audio_dts = (int32_t)ffstream( st)->cur_dts;
 723         video_dts = (int32_t)ffstream(vst)->cur_dts;
 724
 725         if (audio_dts != AV_NOPTS_VALUE && video_dts != AV_NOPTS_VALUE) {
 726             /*We always return a video frame first to get the pixel format first*/
 727             wav->smv_last_stream = wav->smv_given_first ?
 728                 av_compare_ts(video_dts, vst->time_base,
 729                               audio_dts,  st->time_base) > 0 : 0;
 730             wav->smv_given_first = 1;
 731         }
 732         wav->smv_last_stream = !wav->smv_last_stream;
 733         wav->smv_last_stream |= wav->audio_eof;
 734         wav->smv_last_stream &= !wav->smv_eof;
 735         if (wav->smv_last_stream) {
 736             uint64_t old_pos = avio_tell(s->pb);
 737             uint64_t new_pos = wav->smv_data_ofs +
 738                 wav->smv_block * (int64_t)wav->smv_block_size;
 739             if (avio_seek(s->pb, new_pos, SEEK_SET) < 0) {
 740                 ret = AVERROR_EOF;
 741                 goto smv_out;
 742             }
 743             size = avio_rl24(s->pb);
 744             if (size > wav->smv_block_size) {
 745                 ret = AVERROR_EOF;
 746                 goto smv_out;
 747             }
 748             ret  = av_get_packet(s->pb, pkt, size);
 749             if (ret < 0)
 750                 goto smv_out;
 751             pkt->pos -= 3;
 752             pkt->pts = wav->smv_block * wav->smv_frames_per_jpeg;
 753             pkt->duration = wav->smv_frames_per_jpeg;
 754             wav->smv_block++;
 755
 756             pkt->stream_index = vst->index;
 757 smv_out:
 758             avio_seek(s->pb, old_pos, SEEK_SET);
 759             if (ret == AVERROR_EOF) {
 760                 wav->smv_eof = 1;
 761                 goto smv_retry;
 762             }
 763             return ret;
 764         }
 765     }
 766
 767     left = wav->data_end - avio_tell(s->pb);
 768     if (wav->ignore_length)
 769         left = INT_MAX;
 770     if (left <= 0) {
 771         if (CONFIG_W64_DEMUXER && wav->w64)
 772             left = find_guid(s->pb, ff_w64_guid_data) - 24;
 773         else
 774             left = find_tag(wav, s->pb, MKTAG('d', 'a', 't', 'a'));
 775         if (left < 0) {
 776             wav->audio_eof = 1;
 777             if (wav->smv_data_ofs > 0 && !wav->smv_eof)
 778                 goto smv_retry;
 779             return AVERROR_EOF;
 780         }
 781         if (INT64_MAX - left < avio_tell(s->pb))
 782             return AVERROR_INVALIDDATA;
 783         wav->data_end = avio_tell(s->pb) + left;
 784     }
 785
 786     size = wav->max_size;
 787     if (st->codecpar->block_align > 1) {
 788         if (size < st->codecpar->block_align)
 789             size = st->codecpar->block_align;
 790         size = (size / st->codecpar->block_align) * st->codecpar->block_align;
 791     }
 792     size = FFMIN(size, left);
 793     ret  = av_get_packet(s->pb, pkt, size);
 794     if (ret < 0)
 795         return ret;
 796     pkt->stream_index = 0;
 797
 798     return ret;
 799 }
 800
 801 static int wav_read_seek(AVFormatContext *s,
 802                          int stream_index, int64_t timestamp, int flags)
 803 {
 804     WAVDemuxContext *wav = s->priv_data;
 805     AVStream *ast = s->streams[0], *vst = wav->vst;
 806     wav->smv_eof = 0;
 807     wav->audio_eof = 0;
 808
 809     if (stream_index != 0 && (!vst || stream_index != vst->index))
 810         return AVERROR(EINVAL);
 811     if (wav->smv_data_ofs > 0) {
 812         int64_t smv_timestamp = timestamp;
 813         if (stream_index == 0)
 814             smv_timestamp = av_rescale_q(timestamp, ast->time_base, vst->time_base);
 815         else
 816             timestamp = av_rescale_q(smv_timestamp, vst->time_base, ast->time_base);
 817         if (wav->smv_frames_per_jpeg > 0) {
 818             wav->smv_block = smv_timestamp / wav->smv_frames_per_jpeg;
 819         }
 820     }
 821
 822     switch (ast->codecpar->codec_id) {
 823     case AV_CODEC_ID_MP2:
 824     case AV_CODEC_ID_MP3:
 825     case AV_CODEC_ID_AC3:
 826     case AV_CODEC_ID_DTS:
 827     case AV_CODEC_ID_XMA2:
 828         /* use generic seeking with dynamically generated indexes */
 829         return -1;
 830     default:
 831         break;
 832     }
 833     return ff_pcm_read_seek(s, 0, timestamp, flags);
 834 }
 835
 836 static const AVClass wav_demuxer_class = {
 837     .class_name = "WAV demuxer",
 838     .item_name  = av_default_item_name,
 839     .option     = demux_options,
 840     .version    = LIBAVUTIL_VERSION_INT,
 841 };
 842 const FFInputFormat ff_wav_demuxer = {
 843     .p.name         = "wav",
 844     .p.long_name    = NULL_IF_CONFIG_SMALL("WAV / WAVE (Waveform Audio)"),
 845     .p.flags        = AVFMT_GENERIC_INDEX,
 846     .p.codec_tag    = ff_wav_codec_tags_list,
 847     .p.priv_class   = &wav_demuxer_class,
 848     .priv_data_size = sizeof(WAVDemuxContext),
 849     .read_probe     = wav_probe,
 850     .read_header    = wav_read_header,
 851     .read_packet    = wav_read_packet,
 852     .read_seek      = wav_read_seek,
 853 };
 854 #endif /* CONFIG_WAV_DEMUXER */
 855
 856 #if CONFIG_W64_DEMUXER
 857 static int w64_probe(const AVProbeData *p)
 858 {
 859     if (p->buf_size <= 40)
 860         return 0;
 861     if (!memcmp(p->buf,      ff_w64_guid_riff, 16) &&
 862         !memcmp(p->buf + 24, ff_w64_guid_wave, 16))
 863         return AVPROBE_SCORE_MAX;
 864     else
 865         return 0;
 866 }
 867
 868 static int w64_read_header(AVFormatContext *s)
 869 {
 870     int64_t size, data_ofs = 0;
 871     AVIOContext *pb      = s->pb;
 872     WAVDemuxContext *wav = s->priv_data;
 873     AVStream *st;
 874     uint8_t guid[16];
 875     int ret;
 876
 877     if (avio_read(pb, guid, 16) != 16 || memcmp(guid, ff_w64_guid_riff, 16))
 878         return AVERROR_INVALIDDATA;
 879
 880     /* riff + wave + fmt + sizes */
 881     if (avio_rl64(pb) < 16 + 8 + 16 + 8 + 16 + 8)
 882         return AVERROR_INVALIDDATA;
 883
 884     avio_read(pb, guid, 16);
 885     if (memcmp(guid, ff_w64_guid_wave, 16)) {
 886         av_log(s, AV_LOG_ERROR, "could not find wave guid\n");
 887         return AVERROR_INVALIDDATA;
 888     }
 889
 890     wav->w64 = 1;
 891
 892     st = avformat_new_stream(s, NULL);
 893     if (!st)
 894         return AVERROR(ENOMEM);
 895
 896     while (!avio_feof(pb)) {
 897         if (avio_read(pb, guid, 16) != 16)
 898             break;
 899         size = avio_rl64(pb);
 900         if (size <= 24 || INT64_MAX - size < avio_tell(pb)) {
 901             if (data_ofs)
 902                 break;
 903             return AVERROR_INVALIDDATA;
 904         }
 905
 906         if (!memcmp(guid, ff_w64_guid_fmt, 16)) {
 907             /* subtract chunk header size - normal wav file doesn't count it */
 908             ret = ff_get_wav_header(s, pb, st->codecpar, size - 24, 0);
 909             if (ret < 0)
 910                 return ret;
 911             avio_skip(pb, FFALIGN(size, INT64_C(8)) - size);
 912             if (st->codecpar->block_align &&
 913                 st->codecpar->ch_layout.nb_channels < FF_SANE_NB_CHANNELS &&
 914                 st->codecpar->bits_per_coded_sample < 128) {
 915                 int block_align = st->codecpar->block_align;
 916
 917                 block_align = FFMAX(block_align,
 918                                     ((st->codecpar->bits_per_coded_sample + 7) / 8) *
 919                                     st->codecpar->ch_layout.nb_channels);
 920                 if (block_align > st->codecpar->block_align) {
 921                     av_log(s, AV_LOG_WARNING, "invalid block_align: %d, broken file.\n",
 922                            st->codecpar->block_align);
 923                     st->codecpar->block_align = block_align;
 924                 }
 925             }
 926             avpriv_set_pts_info(st, 64, 1, st->codecpar->sample_rate);
 927         } else if (!memcmp(guid, ff_w64_guid_fact, 16)) {
 928             int64_t samples;
 929
 930             samples = avio_rl64(pb);
 931             if (samples > 0)
 932                 st->duration = samples;
 933             avio_skip(pb, FFALIGN(size, INT64_C(8)) - 32);
 934         } else if (!memcmp(guid, ff_w64_guid_data, 16)) {
 935             wav->data_end = avio_tell(pb) + size - 24;
 936
 937             data_ofs = avio_tell(pb);
 938             if (!(pb->seekable & AVIO_SEEKABLE_NORMAL))
 939                 break;
 940
 941             avio_skip(pb, size - 24);
 942         } else if (!memcmp(guid, ff_w64_guid_summarylist, 16)) {
 943             int64_t start, end, cur;
 944             uint32_t count, chunk_size, i;
 945             int64_t filesize  = avio_size(s->pb);
 946
 947             start = avio_tell(pb);
 948             end = start + FFALIGN(size, INT64_C(8)) - 24;
 949             count = avio_rl32(pb);
 950
 951             for (i = 0; i < count; i++) {
 952                 char chunk_key[5], *value;
 953
 954                 if (avio_feof(pb) || (cur = avio_tell(pb)) < 0 || cur > end - 8 /* = tag + size */)
 955                     break;
 956
 957                 chunk_key[4] = 0;
 958                 avio_read(pb, chunk_key, 4);
 959                 chunk_size = avio_rl32(pb);
 960                 if (chunk_size == UINT32_MAX || (filesize >= 0 && chunk_size > filesize))
 961                     return AVERROR_INVALIDDATA;
 962
 963                 value = av_malloc(chunk_size + 1);
 964                 if (!value)
 965                     return AVERROR(ENOMEM);
 966
 967                 ret = avio_get_str16le(pb, chunk_size, value, chunk_size);
 968                 if (ret < 0) {
 969                     av_free(value);
 970                     return ret;
 971                 }
 972                 avio_skip(pb, chunk_size - ret);
 973
 974                 av_dict_set(&s->metadata, chunk_key, value, AV_DICT_DONT_STRDUP_VAL);
 975             }
 976
 977             avio_skip(pb, end - avio_tell(pb));
 978         } else {
 979             av_log(s, AV_LOG_DEBUG, "unknown guid: "FF_PRI_GUID"\n", FF_ARG_GUID(guid));
 980             avio_skip(pb, FFALIGN(size, INT64_C(8)) - 24);
 981         }
 982     }
 983
 984     if (!data_ofs)
 985         return AVERROR_EOF;
 986
 987     ff_metadata_conv_ctx(s, NULL, wav_metadata_conv);
 988     ff_metadata_conv_ctx(s, NULL, ff_riff_info_conv);
 989
 990     handle_stream_probing(st);
 991     ffstream(st)->need_parsing = AVSTREAM_PARSE_FULL_RAW;
 992
 993     avio_seek(pb, data_ofs, SEEK_SET);
 994
 995     set_spdif(s, wav);
 996     set_max_size(st, wav);
 997
 998     return 0;
 999 }
1000
1001 static const AVClass w64_demuxer_class = {
1002     .class_name = "W64 demuxer",
1003     .item_name  = av_default_item_name,
1004     .option     = &demux_options[W64_DEMUXER_OPTIONS_OFFSET],
1005     .version    = LIBAVUTIL_VERSION_INT,
1006 };
1007
1008 const FFInputFormat ff_w64_demuxer = {
1009     .p.name         = "w64",
1010     .p.long_name    = NULL_IF_CONFIG_SMALL("Sony Wave64"),
1011     .p.flags        = AVFMT_GENERIC_INDEX,
1012     .p.codec_tag    = ff_wav_codec_tags_list,
1013     .p.priv_class   = &w64_demuxer_class,
1014     .priv_data_size = sizeof(WAVDemuxContext),
1015     .read_probe     = w64_probe,
1016     .read_header    = w64_read_header,
1017     .read_packet    = wav_read_packet,
1018     .read_seek      = wav_read_seek,
1019 };
1020 #endif /* CONFIG_W64_DEMUXER */