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 "id3v2.h"
  38 #include "internal.h"
  39 #include "metadata.h"
  40 #include "pcm.h"
  41 #include "riff.h"
  42 #include "w64.h"
  43 #include "spdif.h"
  44
  45 typedef struct WAVDemuxContext {
  46     const AVClass *class;
  47     int64_t data_end;
  48     int w64;
  49     int64_t smv_data_ofs;
  50     int smv_block_size;
  51     int smv_frames_per_jpeg;
  52     int smv_block;
  53     int smv_last_stream;
  54     int smv_eof;
  55     int audio_eof;
  56     int ignore_length;
  57     int spdif;
  58     int smv_cur_pt;
  59     int smv_given_first;
  60     int unaligned; // e.g. if an odd number of bytes ID3 tag was prepended
  61     int rifx; // RIFX: integer byte order for parameters is big endian
  62 } WAVDemuxContext;
  63
  64 static void set_spdif(AVFormatContext *s, WAVDemuxContext *wav)
  65 {
  66     if (CONFIG_SPDIF_DEMUXER && s->streams[0]->codecpar->codec_tag == 1) {
  67         enum AVCodecID codec;
  68         int len = 1<<16;
  69         int ret = ffio_ensure_seekback(s->pb, len);
  70
  71         if (ret >= 0) {
  72             uint8_t *buf = av_malloc(len + AV_INPUT_BUFFER_PADDING_SIZE);
  73             if (!buf) {
  74                 ret = AVERROR(ENOMEM);
  75             } else {
  76                 int64_t pos = avio_tell(s->pb);
  77                 len = ret = avio_read(s->pb, buf, len);
  78                 if (len >= 0) {
  79                     ret = ff_spdif_probe(buf, len, &codec);
  80                     if (ret > AVPROBE_SCORE_EXTENSION) {
  81                         s->streams[0]->codecpar->codec_id = codec;
  82                         wav->spdif = 1;
  83                     }
  84                 }
  85                 avio_seek(s->pb, pos, SEEK_SET);
  86                 av_free(buf);
  87             }
  88         }
  89
  90         if (ret < 0)
  91             av_log(s, AV_LOG_WARNING, "Cannot check for SPDIF\n");
  92     }
  93 }
  94
  95 #if CONFIG_WAV_DEMUXER
  96
  97 static int64_t next_tag(AVIOContext *pb, uint32_t *tag, int big_endian)
  98 {
  99     *tag = avio_rl32(pb);
 100     if (!big_endian) {
 101         return avio_rl32(pb);
 102     } else {
 103         return avio_rb32(pb);
 104     }
 105 }
 106
 107 /* RIFF chunks are always at even offsets relative to where they start. */
 108 static int64_t wav_seek_tag(WAVDemuxContext * wav, AVIOContext *s, int64_t offset, int whence)
 109 {
 110     offset += offset < INT64_MAX && offset + wav->unaligned & 1;
 111
 112     return avio_seek(s, offset, whence);
 113 }
 114
 115 /* return the size of the found tag */
 116 static int64_t find_tag(WAVDemuxContext * wav, AVIOContext *pb, uint32_t tag1)
 117 {
 118     unsigned int tag;
 119     int64_t size;
 120
 121     for (;;) {
 122         if (avio_feof(pb))
 123             return AVERROR_EOF;
 124         size = next_tag(pb, &tag, wav->rifx);
 125         if (tag == tag1)
 126             break;
 127         wav_seek_tag(wav, pb, size, SEEK_CUR);
 128     }
 129     return size;
 130 }
 131
 132 static int wav_probe(const AVProbeData *p)
 133 {
 134     /* check file header */
 135     if (p->buf_size <= 32)
 136         return 0;
 137     if (!memcmp(p->buf + 8, "WAVE", 4)) {
 138         if (!memcmp(p->buf, "RIFF", 4) || !memcmp(p->buf, "RIFX", 4))
 139             /* Since the ACT demuxer has a standard WAV header at the top of
 140              * its own, the returned score is decreased to avoid a probe
 141              * conflict between ACT and WAV. */
 142             return AVPROBE_SCORE_MAX - 1;
 143         else if (!memcmp(p->buf,      "RF64", 4) &&
 144                  !memcmp(p->buf + 12, "ds64", 4))
 145             return AVPROBE_SCORE_MAX;
 146     }
 147     return 0;
 148 }
 149
 150 static void handle_stream_probing(AVStream *st)
 151 {
 152     if (st->codecpar->codec_id == AV_CODEC_ID_PCM_S16LE) {
 153         st->request_probe = AVPROBE_SCORE_EXTENSION;
 154         st->probe_packets = FFMIN(st->probe_packets, 32);
 155     }
 156 }
 157
 158 static int wav_parse_fmt_tag(AVFormatContext *s, int64_t size, AVStream **st)
 159 {
 160     AVIOContext *pb = s->pb;
 161     WAVDemuxContext *wav = s->priv_data;
 162     int ret;
 163
 164     /* parse fmt header */
 165     *st = avformat_new_stream(s, NULL);
 166     if (!*st)
 167         return AVERROR(ENOMEM);
 168
 169     ret = ff_get_wav_header(s, pb, (*st)->codecpar, size, wav->rifx);
 170     if (ret < 0)
 171         return ret;
 172     handle_stream_probing(*st);
 173
 174     (*st)->need_parsing = AVSTREAM_PARSE_FULL_RAW;
 175
 176     avpriv_set_pts_info(*st, 64, 1, (*st)->codecpar->sample_rate);
 177
 178     return 0;
 179 }
 180
 181 static int wav_parse_xma2_tag(AVFormatContext *s, int64_t size, AVStream **st)
 182 {
 183     AVIOContext *pb = s->pb;
 184     int version, num_streams, i, channels = 0, ret;
 185
 186     if (size < 36)
 187         return AVERROR_INVALIDDATA;
 188
 189     *st = avformat_new_stream(s, NULL);
 190     if (!*st)
 191         return AVERROR(ENOMEM);
 192
 193     (*st)->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
 194     (*st)->codecpar->codec_id   = AV_CODEC_ID_XMA2;
 195     (*st)->need_parsing         = AVSTREAM_PARSE_FULL_RAW;
 196
 197     version = avio_r8(pb);
 198     if (version != 3 && version != 4)
 199         return AVERROR_INVALIDDATA;
 200     num_streams = avio_r8(pb);
 201     if (size != (32 + ((version==3)?0:8) + 4*num_streams))
 202         return AVERROR_INVALIDDATA;
 203     avio_skip(pb, 10);
 204     (*st)->codecpar->sample_rate = avio_rb32(pb);
 205     if (version == 4)
 206         avio_skip(pb, 8);
 207     avio_skip(pb, 4);
 208     (*st)->duration = avio_rb32(pb);
 209     avio_skip(pb, 8);
 210
 211     for (i = 0; i < num_streams; i++) {
 212         channels += avio_r8(pb);
 213         avio_skip(pb, 3);
 214     }
 215     (*st)->codecpar->channels = channels;
 216
 217     if ((*st)->codecpar->channels <= 0 || (*st)->codecpar->sample_rate <= 0)
 218         return AVERROR_INVALIDDATA;
 219
 220     avpriv_set_pts_info(*st, 64, 1, (*st)->codecpar->sample_rate);
 221
 222     avio_seek(pb, -size, SEEK_CUR);
 223     if ((ret = ff_get_extradata(s, (*st)->codecpar, pb, size)) < 0)
 224         return ret;
 225
 226     return 0;
 227 }
 228
 229 static inline int wav_parse_bext_string(AVFormatContext *s, const char *key,
 230                                         int length)
 231 {
 232     char temp[257];
 233     int ret;
 234
 235     av_assert0(length < sizeof(temp));
 236     if ((ret = avio_read(s->pb, temp, length)) != length)
 237         return ret < 0 ? ret : AVERROR_INVALIDDATA;
 238
 239     temp[length] = 0;
 240
 241     if (strlen(temp))
 242         return av_dict_set(&s->metadata, key, temp, 0);
 243
 244     return 0;
 245 }
 246
 247 static int wav_parse_bext_tag(AVFormatContext *s, int64_t size)
 248 {
 249     char temp[131], *coding_history;
 250     int ret, x;
 251     uint64_t time_reference;
 252     int64_t umid_parts[8], umid_mask = 0;
 253
 254     if ((ret = wav_parse_bext_string(s, "description", 256)) < 0 ||
 255         (ret = wav_parse_bext_string(s, "originator", 32)) < 0 ||
 256         (ret = wav_parse_bext_string(s, "originator_reference", 32)) < 0 ||
 257         (ret = wav_parse_bext_string(s, "origination_date", 10)) < 0 ||
 258         (ret = wav_parse_bext_string(s, "origination_time", 8)) < 0)
 259         return ret;
 260
 261     time_reference = avio_rl64(s->pb);
 262     snprintf(temp, sizeof(temp), "%"PRIu64, time_reference);
 263     if ((ret = av_dict_set(&s->metadata, "time_reference", temp, 0)) < 0)
 264         return ret;
 265
 266     /* check if version is >= 1, in which case an UMID may be present */
 267     if (avio_rl16(s->pb) >= 1) {
 268         for (x = 0; x < 8; x++)
 269             umid_mask |= umid_parts[x] = avio_rb64(s->pb);
 270
 271         if (umid_mask) {
 272             /* the string formatting below is per SMPTE 330M-2004 Annex C */
 273             if (umid_parts[4] == 0 && umid_parts[5] == 0 &&
 274                 umid_parts[6] == 0 && umid_parts[7] == 0) {
 275                 /* basic UMID */
 276                 snprintf(temp, sizeof(temp),
 277                          "0x%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64,
 278                          umid_parts[0], umid_parts[1],
 279                          umid_parts[2], umid_parts[3]);
 280             } else {
 281                 /* extended UMID */
 282                 snprintf(temp, sizeof(temp),
 283                          "0x%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64
 284                          "%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64,
 285                          umid_parts[0], umid_parts[1],
 286                          umid_parts[2], umid_parts[3],
 287                          umid_parts[4], umid_parts[5],
 288                          umid_parts[6], umid_parts[7]);
 289             }
 290
 291             if ((ret = av_dict_set(&s->metadata, "umid", temp, 0)) < 0)
 292                 return ret;
 293         }
 294
 295         avio_skip(s->pb, 190);
 296     } else
 297         avio_skip(s->pb, 254);
 298
 299     if (size > 602) {
 300         /* CodingHistory present */
 301         size -= 602;
 302
 303         if (!(coding_history = av_malloc(size + 1)))
 304             return AVERROR(ENOMEM);
 305
 306         if ((ret = avio_read(s->pb, coding_history, size)) != size) {
 307             av_free(coding_history);
 308             return ret < 0 ? ret : AVERROR_INVALIDDATA;
 309         }
 310
 311         coding_history[size] = 0;
 312         if ((ret = av_dict_set(&s->metadata, "coding_history", coding_history,
 313                                AV_DICT_DONT_STRDUP_VAL)) < 0)
 314             return ret;
 315     }
 316
 317     return 0;
 318 }
 319
 320 static const AVMetadataConv wav_metadata_conv[] = {
 321     { "description",      "comment"       },
 322     { "originator",       "encoded_by"    },
 323     { "origination_date", "date"          },
 324     { "origination_time", "creation_time" },
 325     { 0 },
 326 };
 327
 328 /* wav input */
 329 static int wav_read_header(AVFormatContext *s)
 330 {
 331     int64_t size, av_uninit(data_size);
 332     int64_t sample_count = 0;
 333     int rf64 = 0;
 334     uint32_t tag;
 335     AVIOContext *pb      = s->pb;
 336     AVStream *st         = NULL;
 337     WAVDemuxContext *wav = s->priv_data;
 338     int ret, got_fmt = 0, got_xma2 = 0;
 339     int64_t next_tag_ofs, data_ofs = -1;
 340
 341     wav->unaligned = avio_tell(s->pb) & 1;
 342
 343     wav->smv_data_ofs = -1;
 344
 345     /* read chunk ID */
 346     tag = avio_rl32(pb);
 347     switch (tag) {
 348     case MKTAG('R', 'I', 'F', 'F'):
 349         break;
 350     case MKTAG('R', 'I', 'F', 'X'):
 351         wav->rifx = 1;
 352         break;
 353     case MKTAG('R', 'F', '6', '4'):
 354         rf64 = 1;
 355         break;
 356     default:
 357         av_log(s, AV_LOG_ERROR, "invalid start code %s in RIFF header\n",
 358                av_fourcc2str(tag));
 359         return AVERROR_INVALIDDATA;
 360     }
 361
 362     /* read chunk size */
 363     avio_rl32(pb);
 364
 365     /* read format */
 366     if (avio_rl32(pb) != MKTAG('W', 'A', 'V', 'E')) {
 367         av_log(s, AV_LOG_ERROR, "invalid format in RIFF header\n");
 368         return AVERROR_INVALIDDATA;
 369     }
 370
 371     if (rf64) {
 372         if (avio_rl32(pb) != MKTAG('d', 's', '6', '4'))
 373             return AVERROR_INVALIDDATA;
 374         size = avio_rl32(pb);
 375         if (size < 24)
 376             return AVERROR_INVALIDDATA;
 377         avio_rl64(pb); /* RIFF size */
 378
 379         data_size    = avio_rl64(pb);
 380         sample_count = avio_rl64(pb);
 381
 382         if (data_size < 0 || sample_count < 0) {
 383             av_log(s, AV_LOG_ERROR, "negative data_size and/or sample_count in "
 384                    "ds64: data_size = %"PRId64", sample_count = %"PRId64"\n",
 385                    data_size, sample_count);
 386             return AVERROR_INVALIDDATA;
 387         }
 388         avio_skip(pb, size - 24); /* skip rest of ds64 chunk */
 389
 390     }
 391
 392     for (;;) {
 393         AVStream *vst;
 394         size         = next_tag(pb, &tag, wav->rifx);
 395         next_tag_ofs = avio_tell(pb) + size;
 396
 397         if (avio_feof(pb))
 398             break;
 399
 400         switch (tag) {
 401         case MKTAG('f', 'm', 't', ' '):
 402             /* only parse the first 'fmt ' tag found */
 403             if (!got_xma2 && !got_fmt && (ret = wav_parse_fmt_tag(s, size, &st)) < 0) {
 404                 return ret;
 405             } else if (got_fmt)
 406                 av_log(s, AV_LOG_WARNING, "found more than one 'fmt ' tag\n");
 407
 408             got_fmt = 1;
 409             break;
 410         case MKTAG('X', 'M', 'A', '2'):
 411             /* only parse the first 'XMA2' tag found */
 412             if (!got_fmt && !got_xma2 && (ret = wav_parse_xma2_tag(s, size, &st)) < 0) {
 413                 return ret;
 414             } else if (got_xma2)
 415                 av_log(s, AV_LOG_WARNING, "found more than one 'XMA2' tag\n");
 416
 417             got_xma2 = 1;
 418             break;
 419         case MKTAG('d', 'a', 't', 'a'):
 420             if (!(pb->seekable & AVIO_SEEKABLE_NORMAL) && !got_fmt && !got_xma2) {
 421                 av_log(s, AV_LOG_ERROR,
 422                        "found no 'fmt ' tag before the 'data' tag\n");
 423                 return AVERROR_INVALIDDATA;
 424             }
 425
 426             if (rf64) {
 427                 next_tag_ofs = wav->data_end = av_sat_add64(avio_tell(pb), data_size);
 428             } else if (size != 0xFFFFFFFF) {
 429                 data_size    = size;
 430                 next_tag_ofs = wav->data_end = size ? next_tag_ofs : INT64_MAX;
 431             } else {
 432                 av_log(s, AV_LOG_WARNING, "Ignoring maximum wav data size, "
 433                        "file may be invalid\n");
 434                 data_size    = 0;
 435                 next_tag_ofs = wav->data_end = INT64_MAX;
 436             }
 437
 438             data_ofs = avio_tell(pb);
 439
 440             /* don't look for footer metadata if we can't seek or if we don't
 441              * know where the data tag ends
 442              */
 443             if (!(pb->seekable & AVIO_SEEKABLE_NORMAL) || (!rf64 && !size))
 444                 goto break_loop;
 445             break;
 446         case MKTAG('f', 'a', 'c', 't'):
 447             if (!sample_count)
 448                 sample_count = (!wav->rifx ? avio_rl32(pb) : avio_rb32(pb));
 449             break;
 450         case MKTAG('b', 'e', 'x', 't'):
 451             if ((ret = wav_parse_bext_tag(s, size)) < 0)
 452                 return ret;
 453             break;
 454         case MKTAG('S','M','V','0'):
 455             if (!got_fmt) {
 456                 av_log(s, AV_LOG_ERROR, "found no 'fmt ' tag before the 'SMV0' tag\n");
 457                 return AVERROR_INVALIDDATA;
 458             }
 459             // SMV file, a wav file with video appended.
 460             if (size != MKTAG('0','2','0','0')) {
 461                 av_log(s, AV_LOG_ERROR, "Unknown SMV version found\n");
 462                 goto break_loop;
 463             }
 464             av_log(s, AV_LOG_DEBUG, "Found SMV data\n");
 465             wav->smv_given_first = 0;
 466             vst = avformat_new_stream(s, NULL);
 467             if (!vst)
 468                 return AVERROR(ENOMEM);
 469             avio_r8(pb);
 470             vst->id = 1;
 471             vst->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
 472             vst->codecpar->codec_id = AV_CODEC_ID_SMVJPEG;
 473             vst->codecpar->width  = avio_rl24(pb);
 474             vst->codecpar->height = avio_rl24(pb);
 475             if ((ret = ff_alloc_extradata(vst->codecpar, 4)) < 0) {
 476                 av_log(s, AV_LOG_ERROR, "Could not allocate extradata.\n");
 477                 return ret;
 478             }
 479             size = avio_rl24(pb);
 480             wav->smv_data_ofs = avio_tell(pb) + (size - 5) * 3;
 481             avio_rl24(pb);
 482             wav->smv_block_size = avio_rl24(pb);
 483             if (!wav->smv_block_size)
 484                 return AVERROR_INVALIDDATA;
 485             avpriv_set_pts_info(vst, 32, 1, avio_rl24(pb));
 486             vst->duration = avio_rl24(pb);
 487             avio_rl24(pb);
 488             avio_rl24(pb);
 489             wav->smv_frames_per_jpeg = avio_rl24(pb);
 490             if (wav->smv_frames_per_jpeg > 65536) {
 491                 av_log(s, AV_LOG_ERROR, "too many frames per jpeg\n");
 492                 return AVERROR_INVALIDDATA;
 493             }
 494             AV_WL32(vst->codecpar->extradata, wav->smv_frames_per_jpeg);
 495             wav->smv_cur_pt = 0;
 496             goto break_loop;
 497         case MKTAG('L', 'I', 'S', 'T'):
 498             if (size < 4) {
 499                 av_log(s, AV_LOG_ERROR, "too short LIST tag\n");
 500                 return AVERROR_INVALIDDATA;
 501             }
 502             switch (avio_rl32(pb)) {
 503             case MKTAG('I', 'N', 'F', 'O'):
 504                 ff_read_riff_info(s, size - 4);
 505             }
 506             break;
 507         case MKTAG('I', 'D', '3', ' '):
 508         case MKTAG('i', 'd', '3', ' '): {
 509             ID3v2ExtraMeta *id3v2_extra_meta = NULL;
 510             ff_id3v2_read_dict(pb, &s->internal->id3v2_meta, ID3v2_DEFAULT_MAGIC, &id3v2_extra_meta);
 511             if (id3v2_extra_meta) {
 512                 ff_id3v2_parse_apic(s, id3v2_extra_meta);
 513                 ff_id3v2_parse_chapters(s, id3v2_extra_meta);
 514                 ff_id3v2_parse_priv(s, id3v2_extra_meta);
 515             }
 516             ff_id3v2_free_extra_meta(&id3v2_extra_meta);
 517             }
 518             break;
 519         }
 520
 521         /* seek to next tag unless we know that we'll run into EOF */
 522         if ((avio_size(pb) > 0 && next_tag_ofs >= avio_size(pb)) ||
 523             wav_seek_tag(wav, pb, next_tag_ofs, SEEK_SET) < 0) {
 524             break;
 525         }
 526     }
 527
 528 break_loop:
 529     if (!got_fmt && !got_xma2) {
 530         av_log(s, AV_LOG_ERROR, "no 'fmt ' or 'XMA2' tag found\n");
 531         return AVERROR_INVALIDDATA;
 532     }
 533
 534     if (data_ofs < 0) {
 535         av_log(s, AV_LOG_ERROR, "no 'data' tag found\n");
 536         return AVERROR_INVALIDDATA;
 537     }
 538
 539     avio_seek(pb, data_ofs, SEEK_SET);
 540
 541     if (data_size > (INT64_MAX>>3)) {
 542         av_log(s, AV_LOG_WARNING, "Data size %"PRId64" is too large\n", data_size);
 543         data_size = 0;
 544     }
 545
 546     if (   st->codecpar->bit_rate > 0 && data_size > 0
 547         && st->codecpar->sample_rate > 0
 548         && sample_count > 0 && st->codecpar->channels > 1
 549         && sample_count % st->codecpar->channels == 0) {
 550         if (fabs(8.0 * data_size * st->codecpar->channels * st->codecpar->sample_rate /
 551             sample_count /st->codecpar->bit_rate - 1.0) < 0.3)
 552             sample_count /= st->codecpar->channels;
 553     }
 554
 555     if (   data_size > 0 && sample_count && st->codecpar->channels
 556         && (data_size << 3) / sample_count / st->codecpar->channels > st->codecpar->bits_per_coded_sample  + 1) {
 557         av_log(s, AV_LOG_WARNING, "ignoring wrong sample_count %"PRId64"\n", sample_count);
 558         sample_count = 0;
 559     }
 560
 561     /* G.729 hack (for Ticket4577)
 562      * FIXME: Come up with cleaner, more general solution */
 563     if (st->codecpar->codec_id == AV_CODEC_ID_G729 && sample_count && (data_size << 3) > sample_count) {
 564         av_log(s, AV_LOG_WARNING, "ignoring wrong sample_count %"PRId64"\n", sample_count);
 565         sample_count = 0;
 566     }
 567
 568     if (!sample_count || av_get_exact_bits_per_sample(st->codecpar->codec_id) > 0)
 569         if (   st->codecpar->channels
 570             && data_size
 571             && av_get_bits_per_sample(st->codecpar->codec_id)
 572             && wav->data_end <= avio_size(pb))
 573             sample_count = (data_size << 3)
 574                                   /
 575                 (st->codecpar->channels * (uint64_t)av_get_bits_per_sample(st->codecpar->codec_id));
 576
 577     if (sample_count)
 578         st->duration = sample_count;
 579
 580     if (st->codecpar->codec_id == AV_CODEC_ID_PCM_S32LE &&
 581         st->codecpar->block_align == st->codecpar->channels * 4 &&
 582         st->codecpar->bits_per_coded_sample == 32 &&
 583         st->codecpar->extradata_size == 2 &&
 584         AV_RL16(st->codecpar->extradata) == 1) {
 585         st->codecpar->codec_id = AV_CODEC_ID_PCM_F16LE;
 586         st->codecpar->bits_per_coded_sample = 16;
 587     } else if (st->codecpar->codec_id == AV_CODEC_ID_PCM_S24LE &&
 588                st->codecpar->block_align == st->codecpar->channels * 4 &&
 589                st->codecpar->bits_per_coded_sample == 24) {
 590         st->codecpar->codec_id = AV_CODEC_ID_PCM_F24LE;
 591     } else if (st->codecpar->codec_id == AV_CODEC_ID_XMA1 ||
 592                st->codecpar->codec_id == AV_CODEC_ID_XMA2) {
 593         st->codecpar->block_align = 2048;
 594     } else if (st->codecpar->codec_id == AV_CODEC_ID_ADPCM_MS && st->codecpar->channels > 2 &&
 595                st->codecpar->block_align < INT_MAX / st->codecpar->channels) {
 596         st->codecpar->block_align *= st->codecpar->channels;
 597     }
 598
 599     ff_metadata_conv_ctx(s, NULL, wav_metadata_conv);
 600     ff_metadata_conv_ctx(s, NULL, ff_riff_info_conv);
 601
 602     set_spdif(s, wav);
 603
 604     return 0;
 605 }
 606
 607 /**
 608  * Find chunk with w64 GUID by skipping over other chunks.
 609  * @return the size of the found chunk
 610  */
 611 static int64_t find_guid(AVIOContext *pb, const uint8_t guid1[16])
 612 {
 613     uint8_t guid[16];
 614     int64_t size;
 615
 616     while (!avio_feof(pb)) {
 617         avio_read(pb, guid, 16);
 618         size = avio_rl64(pb);
 619         if (size <= 24 || size > INT64_MAX - 8)
 620             return AVERROR_INVALIDDATA;
 621         if (!memcmp(guid, guid1, 16))
 622             return size;
 623         avio_skip(pb, FFALIGN(size, INT64_C(8)) - 24);
 624     }
 625     return AVERROR_EOF;
 626 }
 627
 628 #define MAX_SIZE 4096
 629
 630 static int wav_read_packet(AVFormatContext *s, AVPacket *pkt)
 631 {
 632     int ret, size;
 633     int64_t left;
 634     AVStream *st;
 635     WAVDemuxContext *wav = s->priv_data;
 636
 637     if (CONFIG_SPDIF_DEMUXER && wav->spdif == 1)
 638         return ff_spdif_read_packet(s, pkt);
 639
 640     if (wav->smv_data_ofs > 0) {
 641         int64_t audio_dts, video_dts;
 642 smv_retry:
 643         audio_dts = (int32_t)s->streams[0]->cur_dts;
 644         video_dts = (int32_t)s->streams[1]->cur_dts;
 645
 646         if (audio_dts != AV_NOPTS_VALUE && video_dts != AV_NOPTS_VALUE) {
 647             /*We always return a video frame first to get the pixel format first*/
 648             wav->smv_last_stream = wav->smv_given_first ?
 649                 av_compare_ts(video_dts, s->streams[1]->time_base,
 650                               audio_dts, s->streams[0]->time_base) > 0 : 0;
 651             wav->smv_given_first = 1;
 652         }
 653         wav->smv_last_stream = !wav->smv_last_stream;
 654         wav->smv_last_stream |= wav->audio_eof;
 655         wav->smv_last_stream &= !wav->smv_eof;
 656         if (wav->smv_last_stream) {
 657             uint64_t old_pos = avio_tell(s->pb);
 658             uint64_t new_pos = wav->smv_data_ofs +
 659                 wav->smv_block * (int64_t)wav->smv_block_size;
 660             if (avio_seek(s->pb, new_pos, SEEK_SET) < 0) {
 661                 ret = AVERROR_EOF;
 662                 goto smv_out;
 663             }
 664             size = avio_rl24(s->pb);
 665             if (size > wav->smv_block_size) {
 666                 ret = AVERROR_EOF;
 667                 goto smv_out;
 668             }
 669             ret  = av_get_packet(s->pb, pkt, size);
 670             if (ret < 0)
 671                 goto smv_out;
 672             pkt->pos -= 3;
 673             pkt->pts = wav->smv_block * wav->smv_frames_per_jpeg + wav->smv_cur_pt;
 674             wav->smv_cur_pt++;
 675             if (wav->smv_frames_per_jpeg > 0)
 676                 wav->smv_cur_pt %= wav->smv_frames_per_jpeg;
 677             if (!wav->smv_cur_pt)
 678                 wav->smv_block++;
 679
 680             pkt->stream_index = 1;
 681 smv_out:
 682             avio_seek(s->pb, old_pos, SEEK_SET);
 683             if (ret == AVERROR_EOF) {
 684                 wav->smv_eof = 1;
 685                 goto smv_retry;
 686             }
 687             return ret;
 688         }
 689     }
 690
 691     st = s->streams[0];
 692
 693     left = wav->data_end - avio_tell(s->pb);
 694     if (wav->ignore_length)
 695         left = INT_MAX;
 696     if (left <= 0) {
 697         if (CONFIG_W64_DEMUXER && wav->w64)
 698             left = find_guid(s->pb, ff_w64_guid_data) - 24;
 699         else
 700             left = find_tag(wav, s->pb, MKTAG('d', 'a', 't', 'a'));
 701         if (left < 0) {
 702             wav->audio_eof = 1;
 703             if (wav->smv_data_ofs > 0 && !wav->smv_eof)
 704                 goto smv_retry;
 705             return AVERROR_EOF;
 706         }
 707         if (INT64_MAX - left < avio_tell(s->pb))
 708             return AVERROR_INVALIDDATA;
 709         wav->data_end = avio_tell(s->pb) + left;
 710     }
 711
 712     size = MAX_SIZE;
 713     if (st->codecpar->block_align > 1) {
 714         if (size < st->codecpar->block_align)
 715             size = st->codecpar->block_align;
 716         size = (size / st->codecpar->block_align) * st->codecpar->block_align;
 717     }
 718     size = FFMIN(size, left);
 719     ret  = av_get_packet(s->pb, pkt, size);
 720     if (ret < 0)
 721         return ret;
 722     pkt->stream_index = 0;
 723
 724     return ret;
 725 }
 726
 727 static int wav_read_seek(AVFormatContext *s,
 728                          int stream_index, int64_t timestamp, int flags)
 729 {
 730     WAVDemuxContext *wav = s->priv_data;
 731     AVStream *st;
 732     wav->smv_eof = 0;
 733     wav->audio_eof = 0;
 734     if (wav->smv_data_ofs > 0) {
 735         int64_t smv_timestamp = timestamp;
 736         if (stream_index == 0)
 737             smv_timestamp = av_rescale_q(timestamp, s->streams[0]->time_base, s->streams[1]->time_base);
 738         else
 739             timestamp = av_rescale_q(smv_timestamp, s->streams[1]->time_base, s->streams[0]->time_base);
 740         if (wav->smv_frames_per_jpeg > 0) {
 741             wav->smv_block = smv_timestamp / wav->smv_frames_per_jpeg;
 742             wav->smv_cur_pt = smv_timestamp % wav->smv_frames_per_jpeg;
 743         }
 744     }
 745
 746     st = s->streams[0];
 747     switch (st->codecpar->codec_id) {
 748     case AV_CODEC_ID_MP2:
 749     case AV_CODEC_ID_MP3:
 750     case AV_CODEC_ID_AC3:
 751     case AV_CODEC_ID_DTS:
 752     case AV_CODEC_ID_XMA2:
 753         /* use generic seeking with dynamically generated indexes */
 754         return -1;
 755     default:
 756         break;
 757     }
 758     return ff_pcm_read_seek(s, stream_index, timestamp, flags);
 759 }
 760
 761 #define OFFSET(x) offsetof(WAVDemuxContext, x)
 762 #define DEC AV_OPT_FLAG_DECODING_PARAM
 763 static const AVOption demux_options[] = {
 764     { "ignore_length", "Ignore length", OFFSET(ignore_length), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, DEC },
 765     { NULL },
 766 };
 767
 768 static const AVClass wav_demuxer_class = {
 769     .class_name = "WAV demuxer",
 770     .item_name  = av_default_item_name,
 771     .option     = demux_options,
 772     .version    = LIBAVUTIL_VERSION_INT,
 773 };
 774 AVInputFormat ff_wav_demuxer = {
 775     .name           = "wav",
 776     .long_name      = NULL_IF_CONFIG_SMALL("WAV / WAVE (Waveform Audio)"),
 777     .priv_data_size = sizeof(WAVDemuxContext),
 778     .read_probe     = wav_probe,
 779     .read_header    = wav_read_header,
 780     .read_packet    = wav_read_packet,
 781     .read_seek      = wav_read_seek,
 782     .flags          = AVFMT_GENERIC_INDEX,
 783     .codec_tag      = (const AVCodecTag * const []) { ff_codec_wav_tags,  0 },
 784     .priv_class     = &wav_demuxer_class,
 785 };
 786 #endif /* CONFIG_WAV_DEMUXER */
 787
 788 #if CONFIG_W64_DEMUXER
 789 static int w64_probe(const AVProbeData *p)
 790 {
 791     if (p->buf_size <= 40)
 792         return 0;
 793     if (!memcmp(p->buf,      ff_w64_guid_riff, 16) &&
 794         !memcmp(p->buf + 24, ff_w64_guid_wave, 16))
 795         return AVPROBE_SCORE_MAX;
 796     else
 797         return 0;
 798 }
 799
 800 static int w64_read_header(AVFormatContext *s)
 801 {
 802     int64_t size, data_ofs = 0;
 803     AVIOContext *pb      = s->pb;
 804     WAVDemuxContext *wav = s->priv_data;
 805     AVStream *st;
 806     uint8_t guid[16];
 807     int ret;
 808
 809     if (avio_read(pb, guid, 16) != 16 || memcmp(guid, ff_w64_guid_riff, 16))
 810         return AVERROR_INVALIDDATA;
 811
 812     /* riff + wave + fmt + sizes */
 813     if (avio_rl64(pb) < 16 + 8 + 16 + 8 + 16 + 8)
 814         return AVERROR_INVALIDDATA;
 815
 816     avio_read(pb, guid, 16);
 817     if (memcmp(guid, ff_w64_guid_wave, 16)) {
 818         av_log(s, AV_LOG_ERROR, "could not find wave guid\n");
 819         return AVERROR_INVALIDDATA;
 820     }
 821
 822     wav->w64 = 1;
 823
 824     st = avformat_new_stream(s, NULL);
 825     if (!st)
 826         return AVERROR(ENOMEM);
 827
 828     while (!avio_feof(pb)) {
 829         if (avio_read(pb, guid, 16) != 16)
 830             break;
 831         size = avio_rl64(pb);
 832         if (size <= 24 || INT64_MAX - size < avio_tell(pb))
 833             return AVERROR_INVALIDDATA;
 834
 835         if (!memcmp(guid, ff_w64_guid_fmt, 16)) {
 836             /* subtract chunk header size - normal wav file doesn't count it */
 837             ret = ff_get_wav_header(s, pb, st->codecpar, size - 24, 0);
 838             if (ret < 0)
 839                 return ret;
 840             avio_skip(pb, FFALIGN(size, INT64_C(8)) - size);
 841
 842             avpriv_set_pts_info(st, 64, 1, st->codecpar->sample_rate);
 843         } else if (!memcmp(guid, ff_w64_guid_fact, 16)) {
 844             int64_t samples;
 845
 846             samples = avio_rl64(pb);
 847             if (samples > 0)
 848                 st->duration = samples;
 849             avio_skip(pb, FFALIGN(size, INT64_C(8)) - 32);
 850         } else if (!memcmp(guid, ff_w64_guid_data, 16)) {
 851             wav->data_end = avio_tell(pb) + size - 24;
 852
 853             data_ofs = avio_tell(pb);
 854             if (!(pb->seekable & AVIO_SEEKABLE_NORMAL))
 855                 break;
 856
 857             avio_skip(pb, size - 24);
 858         } else if (!memcmp(guid, ff_w64_guid_summarylist, 16)) {
 859             int64_t start, end, cur;
 860             uint32_t count, chunk_size, i;
 861             int64_t filesize  = avio_size(s->pb);
 862
 863             start = avio_tell(pb);
 864             end = start + FFALIGN(size, INT64_C(8)) - 24;
 865             count = avio_rl32(pb);
 866
 867             for (i = 0; i < count; i++) {
 868                 char chunk_key[5], *value;
 869
 870                 if (avio_feof(pb) || (cur = avio_tell(pb)) < 0 || cur > end - 8 /* = tag + size */)
 871                     break;
 872
 873                 chunk_key[4] = 0;
 874                 avio_read(pb, chunk_key, 4);
 875                 chunk_size = avio_rl32(pb);
 876                 if (chunk_size == UINT32_MAX || (filesize >= 0 && chunk_size > filesize))
 877                     return AVERROR_INVALIDDATA;
 878
 879                 value = av_mallocz(chunk_size + 1);
 880                 if (!value)
 881                     return AVERROR(ENOMEM);
 882
 883                 ret = avio_get_str16le(pb, chunk_size, value, chunk_size);
 884                 if (ret < 0) {
 885                     av_free(value);
 886                     return ret;
 887                 }
 888                 avio_skip(pb, chunk_size - ret);
 889
 890                 av_dict_set(&s->metadata, chunk_key, value, AV_DICT_DONT_STRDUP_VAL);
 891             }
 892
 893             avio_skip(pb, end - avio_tell(pb));
 894         } else {
 895             av_log(s, AV_LOG_DEBUG, "unknown guid: "FF_PRI_GUID"\n", FF_ARG_GUID(guid));
 896             avio_skip(pb, FFALIGN(size, INT64_C(8)) - 24);
 897         }
 898     }
 899
 900     if (!data_ofs)
 901         return AVERROR_EOF;
 902
 903     ff_metadata_conv_ctx(s, NULL, wav_metadata_conv);
 904     ff_metadata_conv_ctx(s, NULL, ff_riff_info_conv);
 905
 906     handle_stream_probing(st);
 907     st->need_parsing = AVSTREAM_PARSE_FULL_RAW;
 908
 909     avio_seek(pb, data_ofs, SEEK_SET);
 910
 911     set_spdif(s, wav);
 912
 913     return 0;
 914 }
 915
 916 AVInputFormat ff_w64_demuxer = {
 917     .name           = "w64",
 918     .long_name      = NULL_IF_CONFIG_SMALL("Sony Wave64"),
 919     .priv_data_size = sizeof(WAVDemuxContext),
 920     .read_probe     = w64_probe,
 921     .read_header    = w64_read_header,
 922     .read_packet    = wav_read_packet,
 923     .read_seek      = wav_read_seek,
 924     .flags          = AVFMT_GENERIC_INDEX,
 925     .codec_tag      = (const AVCodecTag * const []) { ff_codec_wav_tags, 0 },
 926 };
 927 #endif /* CONFIG_W64_DEMUXER */