libavformat/cafdec.c

   1 /*
   2  * Core Audio Format demuxer
   3  * Copyright (c) 2007 Justin Ruggles
   4  * Copyright (c) 2009 Peter Ross
   5  *
   6  * This file is part of FFmpeg.
   7  *
   8  * FFmpeg is free software; you can redistribute it and/or
   9  * modify it under the terms of the GNU Lesser General Public
  10  * License as published by the Free Software Foundation; either
  11  * version 2.1 of the License, or (at your option) any later version.
  12  *
  13  * FFmpeg is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  16  * Lesser General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU Lesser General Public
  19  * License along with FFmpeg; if not, write to the Free Software
  20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  21  */
  22
  23 /**
  24  * @file
  25  * Core Audio Format demuxer
  26  */
  27
  28 #include <inttypes.h>
  29
  30 #include "avformat.h"
  31 #include "avio_internal.h"
  32 #include "demux.h"
  33 #include "internal.h"
  34 #include "isom.h"
  35 #include "mov_chan.h"
  36 #include "libavcodec/flac.h"
  37 #include "libavutil/intreadwrite.h"
  38 #include "libavutil/intfloat.h"
  39 #include "libavutil/dict.h"
  40 #include "libavutil/mem.h"
  41 #include "caf.h"
  42
  43 typedef struct CafContext {
  44     int bytes_per_packet;           ///< bytes in a packet, or 0 if variable
  45     int frames_per_packet;          ///< frames in a packet, or 0 if variable
  46     int64_t num_bytes;              ///< total number of bytes in stream
  47
  48     int64_t packet_cnt;             ///< packet counter
  49     int64_t frame_cnt;              ///< frame counter
  50
  51     int64_t data_start;             ///< data start position, in bytes
  52     int64_t data_size;              ///< raw data size, in bytes
  53 } CafContext;
  54
  55 static int probe(const AVProbeData *p)
  56 {
  57     if (AV_RB32(p->buf) != MKBETAG('c','a','f','f'))
  58         return 0;
  59     if (AV_RB16(&p->buf[4]) != 1)
  60         return 0;
  61     if (AV_RB32(p->buf + 8) != MKBETAG('d','e','s','c'))
  62         return 0;
  63     if (AV_RB64(p->buf + 12) != 32)
  64         return 0;
  65     return AVPROBE_SCORE_MAX;
  66 }
  67
  68 /** Read audio description chunk */
  69 static int read_desc_chunk(AVFormatContext *s)
  70 {
  71     AVIOContext *pb = s->pb;
  72     CafContext *caf = s->priv_data;
  73     AVStream *st;
  74     int flags;
  75
  76     /* new audio stream */
  77     st = avformat_new_stream(s, NULL);
  78     if (!st)
  79         return AVERROR(ENOMEM);
  80
  81     /* parse format description */
  82     st->codecpar->codec_type  = AVMEDIA_TYPE_AUDIO;
  83     st->codecpar->sample_rate = av_clipd(av_int2double(avio_rb64(pb)), 0, INT_MAX);
  84     st->codecpar->codec_tag   = avio_rl32(pb);
  85     flags = avio_rb32(pb);
  86     caf->bytes_per_packet  = avio_rb32(pb);
  87     st->codecpar->block_align = caf->bytes_per_packet;
  88     caf->frames_per_packet = avio_rb32(pb);
  89     st->codecpar->ch_layout.nb_channels = avio_rb32(pb);
  90     st->codecpar->bits_per_coded_sample = avio_rb32(pb);
  91
  92     if (caf->bytes_per_packet < 0 || caf->frames_per_packet < 0 || st->codecpar->ch_layout.nb_channels < 0)
  93         return AVERROR_INVALIDDATA;
  94
  95     /* calculate bit rate for constant size packets */
  96     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
  97         st->codecpar->bit_rate = (uint64_t)st->codecpar->sample_rate * (uint64_t)caf->bytes_per_packet * 8
  98                                  / (uint64_t)caf->frames_per_packet;
  99     } else {
 100         st->codecpar->bit_rate = 0;
 101     }
 102
 103     /* determine codec */
 104     if (st->codecpar->codec_tag == MKTAG('l','p','c','m'))
 105         st->codecpar->codec_id = ff_mov_get_lpcm_codec_id(st->codecpar->bits_per_coded_sample, (flags ^ 0x2) | 0x4);
 106     else
 107         st->codecpar->codec_id = ff_codec_get_id(ff_codec_caf_tags, st->codecpar->codec_tag);
 108     return 0;
 109 }
 110
 111 /** Read magic cookie chunk */
 112 static int read_kuki_chunk(AVFormatContext *s, int64_t size)
 113 {
 114     AVIOContext *pb = s->pb;
 115     AVStream *st      = s->streams[0];
 116     int ret;
 117
 118     if (size < 0 || size > INT_MAX - AV_INPUT_BUFFER_PADDING_SIZE)
 119         return -1;
 120
 121     if (st->codecpar->codec_id == AV_CODEC_ID_AAC) {
 122         /* The magic cookie format for AAC is an mp4 esds atom.
 123            The lavc AAC decoder requires the data from the codec specific
 124            description as extradata input. */
 125         int strt, skip;
 126
 127         strt = avio_tell(pb);
 128         ff_mov_read_esds(s, pb);
 129         skip = size - (avio_tell(pb) - strt);
 130         if (skip < 0 || !st->codecpar->extradata ||
 131             st->codecpar->codec_id != AV_CODEC_ID_AAC) {
 132             av_log(s, AV_LOG_ERROR, "invalid AAC magic cookie\n");
 133             return AVERROR_INVALIDDATA;
 134         }
 135         avio_skip(pb, skip);
 136     } else if (st->codecpar->codec_id == AV_CODEC_ID_ALAC) {
 137 #define ALAC_PREAMBLE 12
 138 #define ALAC_HEADER   36
 139 #define ALAC_NEW_KUKI 24
 140         uint8_t preamble[12];
 141         if (size < ALAC_NEW_KUKI) {
 142             av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
 143             avio_skip(pb, size);
 144             return AVERROR_INVALIDDATA;
 145         }
 146         if ((ret = ffio_read_size(pb, preamble, ALAC_PREAMBLE)) < 0) {
 147             av_log(s, AV_LOG_ERROR, "failed to read preamble\n");
 148             return ret;
 149         }
 150
 151         if ((ret = ff_alloc_extradata(st->codecpar, ALAC_HEADER)) < 0)
 152             return ret;
 153
 154         /* For the old style cookie, we skip 12 bytes, then read 36 bytes.
 155          * The new style cookie only contains the last 24 bytes of what was
 156          * 36 bytes in the old style cookie, so we fabricate the first 12 bytes
 157          * in that case to maintain compatibility. */
 158         if (!memcmp(&preamble[4], "frmaalac", 8)) {
 159             if (size < ALAC_PREAMBLE + ALAC_HEADER) {
 160                 av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
 161                 av_freep(&st->codecpar->extradata);
 162                 return AVERROR_INVALIDDATA;
 163             }
 164             if (avio_read(pb, st->codecpar->extradata, ALAC_HEADER) != ALAC_HEADER) {
 165                 av_log(s, AV_LOG_ERROR, "failed to read kuki header\n");
 166                 av_freep(&st->codecpar->extradata);
 167                 return AVERROR_INVALIDDATA;
 168             }
 169             avio_skip(pb, size - ALAC_PREAMBLE - ALAC_HEADER);
 170         } else {
 171             AV_WB32(st->codecpar->extradata, 36);
 172             memcpy(&st->codecpar->extradata[4], "alac", 4);
 173             AV_WB32(&st->codecpar->extradata[8], 0);
 174             memcpy(&st->codecpar->extradata[12], preamble, 12);
 175             if (avio_read(pb, &st->codecpar->extradata[24], ALAC_NEW_KUKI - 12) != ALAC_NEW_KUKI - 12) {
 176                 av_log(s, AV_LOG_ERROR, "failed to read new kuki header\n");
 177                 av_freep(&st->codecpar->extradata);
 178                 return AVERROR_INVALIDDATA;
 179             }
 180             avio_skip(pb, size - ALAC_NEW_KUKI);
 181         }
 182     } else if (st->codecpar->codec_id == AV_CODEC_ID_FLAC) {
 183         int last, type, flac_metadata_size;
 184         uint8_t buf[4];
 185         /* The magic cookie format for FLAC consists mostly of an mp4 dfLa atom. */
 186         if (size < (16 + FLAC_STREAMINFO_SIZE)) {
 187             av_log(s, AV_LOG_ERROR, "invalid FLAC magic cookie\n");
 188             return AVERROR_INVALIDDATA;
 189         }
 190         /* Check cookie version. */
 191         if (avio_r8(pb) != 0) {
 192             av_log(s, AV_LOG_ERROR, "unknown FLAC magic cookie\n");
 193             return AVERROR_INVALIDDATA;
 194         }
 195         avio_rb24(pb); /* Flags */
 196         /* read dfLa fourcc */
 197         if (avio_read(pb, buf, 4) != 4) {
 198             av_log(s, AV_LOG_ERROR, "failed to read FLAC magic cookie\n");
 199             return pb->error < 0 ? pb->error : AVERROR_INVALIDDATA;
 200         }
 201         if (memcmp(buf, "dfLa", 4)) {
 202             av_log(s, AV_LOG_ERROR, "invalid FLAC magic cookie\n");
 203             return AVERROR_INVALIDDATA;
 204         }
 205         /* Check dfLa version. */
 206         if (avio_r8(pb) != 0) {
 207             av_log(s, AV_LOG_ERROR, "unknown dfLa version\n");
 208             return AVERROR_INVALIDDATA;
 209         }
 210         avio_rb24(pb); /* Flags */
 211         if (avio_read(pb, buf, sizeof(buf)) != sizeof(buf)) {
 212             av_log(s, AV_LOG_ERROR, "failed to read FLAC metadata block header\n");
 213             return pb->error < 0 ? pb->error : AVERROR_INVALIDDATA;
 214         }
 215         flac_parse_block_header(buf, &last, &type, &flac_metadata_size);
 216         if (type != FLAC_METADATA_TYPE_STREAMINFO || flac_metadata_size != FLAC_STREAMINFO_SIZE) {
 217             av_log(s, AV_LOG_ERROR, "STREAMINFO must be first FLACMetadataBlock\n");
 218             return AVERROR_INVALIDDATA;
 219         }
 220         ret = ff_get_extradata(s, st->codecpar, pb, FLAC_STREAMINFO_SIZE);
 221         if (ret < 0)
 222             return ret;
 223         if (!last)
 224             av_log(s, AV_LOG_WARNING, "non-STREAMINFO FLACMetadataBlock(s) ignored\n");
 225     } else if (st->codecpar->codec_id == AV_CODEC_ID_OPUS) {
 226         // The data layout for Opus is currently unknown, so we do not export
 227         // extradata at all. Multichannel streams are not supported.
 228         if (st->codecpar->ch_layout.nb_channels > 2) {
 229             avpriv_request_sample(s, "multichannel Opus in CAF");
 230             return AVERROR_PATCHWELCOME;
 231         }
 232         avio_skip(pb, size);
 233     } else if ((ret = ff_get_extradata(s, st->codecpar, pb, size)) < 0) {
 234         return ret;
 235     }
 236
 237     return 0;
 238 }
 239
 240 /** Read packet table chunk */
 241 static int read_pakt_chunk(AVFormatContext *s, int64_t size)
 242 {
 243     AVIOContext *pb = s->pb;
 244     AVStream *st      = s->streams[0];
 245     CafContext *caf   = s->priv_data;
 246     int64_t pos = 0, ccount, num_packets;
 247     int i;
 248     int ret;
 249
 250     ccount = avio_tell(pb);
 251
 252     num_packets = avio_rb64(pb);
 253     if (num_packets < 0 || INT32_MAX / sizeof(AVIndexEntry) < num_packets)
 254         return AVERROR_INVALIDDATA;
 255
 256     st->nb_frames  = avio_rb64(pb); /* valid frames */
 257     st->nb_frames += avio_rb32(pb); /* priming frames */
 258     st->nb_frames += avio_rb32(pb); /* remainder frames */
 259
 260     if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
 261         st->duration = caf->frames_per_packet * num_packets;
 262         pos          = caf-> bytes_per_packet * num_packets;
 263     } else {
 264         st->duration = 0;
 265         for (i = 0; i < num_packets; i++) {
 266             if (avio_feof(pb))
 267                 return AVERROR_INVALIDDATA;
 268             ret = av_add_index_entry(s->streams[0], pos, st->duration, 0, 0, AVINDEX_KEYFRAME);
 269             if (ret < 0)
 270                 return ret;
 271             pos += caf->bytes_per_packet ? caf->bytes_per_packet : ff_mp4_read_descr_len(pb);
 272             st->duration += caf->frames_per_packet ? caf->frames_per_packet : ff_mp4_read_descr_len(pb);
 273         }
 274     }
 275
 276     if (avio_tell(pb) - ccount > size || size > INT64_MAX - ccount) {
 277         av_log(s, AV_LOG_ERROR, "error reading packet table\n");
 278         return AVERROR_INVALIDDATA;
 279     }
 280     avio_seek(pb, ccount + size, SEEK_SET);
 281
 282     caf->num_bytes = pos;
 283     return 0;
 284 }
 285
 286 /** Read information chunk */
 287 static void read_info_chunk(AVFormatContext *s, int64_t size)
 288 {
 289     AVIOContext *pb = s->pb;
 290     unsigned int i;
 291     unsigned int nb_entries = avio_rb32(pb);
 292     for (i = 0; i < nb_entries && !avio_feof(pb); i++) {
 293         char key[32];
 294         char value[1024];
 295         avio_get_str(pb, INT_MAX, key, sizeof(key));
 296         avio_get_str(pb, INT_MAX, value, sizeof(value));
 297         if (!*key)
 298             continue;
 299         av_dict_set(&s->metadata, key, value, 0);
 300     }
 301 }
 302
 303 static int read_header(AVFormatContext *s)
 304 {
 305     AVIOContext *pb = s->pb;
 306     CafContext *caf = s->priv_data;
 307     AVStream *st;
 308     uint32_t tag = 0;
 309     int found_data, ret;
 310     int64_t size, pos;
 311
 312     avio_skip(pb, 8); /* magic, version, file flags */
 313
 314     /* audio description chunk */
 315     if (avio_rb32(pb) != MKBETAG('d','e','s','c')) {
 316         av_log(s, AV_LOG_ERROR, "desc chunk not present\n");
 317         return AVERROR_INVALIDDATA;
 318     }
 319     size = avio_rb64(pb);
 320     if (size != 32)
 321         return AVERROR_INVALIDDATA;
 322
 323     ret = read_desc_chunk(s);
 324     if (ret)
 325         return ret;
 326     st = s->streams[0];
 327
 328     /* parse each chunk */
 329     found_data = 0;
 330     while (!avio_feof(pb)) {
 331
 332         /* stop at data chunk if seeking is not supported or
 333            data chunk size is unknown */
 334         if (found_data && (caf->data_size < 0 || !(pb->seekable & AVIO_SEEKABLE_NORMAL)))
 335             break;
 336
 337         tag  = avio_rb32(pb);
 338         size = avio_rb64(pb);
 339         pos  = avio_tell(pb);
 340         if (avio_feof(pb))
 341             break;
 342
 343         switch (tag) {
 344         case MKBETAG('d','a','t','a'):
 345             avio_skip(pb, 4); /* edit count */
 346             caf->data_start = avio_tell(pb);
 347             caf->data_size  = size < 0 ? -1 : size - 4;
 348             if (caf->data_start < 0 || caf->data_size > INT64_MAX - caf->data_start)
 349                 return AVERROR_INVALIDDATA;
 350
 351             if (caf->data_size > 0 && (pb->seekable & AVIO_SEEKABLE_NORMAL))
 352                 avio_skip(pb, caf->data_size);
 353             found_data = 1;
 354             break;
 355
 356         case MKBETAG('c','h','a','n'):
 357             if ((ret = ff_mov_read_chan(s, s->pb, st, size)) < 0)
 358                 return ret;
 359             break;
 360
 361         /* magic cookie chunk */
 362         case MKBETAG('k','u','k','i'):
 363             if (read_kuki_chunk(s, size))
 364                 return AVERROR_INVALIDDATA;
 365             break;
 366
 367         /* packet table chunk */
 368         case MKBETAG('p','a','k','t'):
 369             if (read_pakt_chunk(s, size))
 370                 return AVERROR_INVALIDDATA;
 371             break;
 372
 373         case MKBETAG('i','n','f','o'):
 374             read_info_chunk(s, size);
 375             break;
 376
 377         default:
 378             av_log(s, AV_LOG_WARNING,
 379                    "skipping CAF chunk: %08"PRIX32" (%s), size %"PRId64"\n",
 380                    tag, av_fourcc2str(av_bswap32(tag)), size);
 381         case MKBETAG('f','r','e','e'):
 382             if (size < 0 && found_data)
 383                 goto found_data;
 384             if (size < 0)
 385                 return AVERROR_INVALIDDATA;
 386             break;
 387         }
 388
 389         if (size > 0 && (pb->seekable & AVIO_SEEKABLE_NORMAL)) {
 390             if (pos > INT64_MAX - size)
 391                 return AVERROR_INVALIDDATA;
 392             avio_seek(pb, pos + size, SEEK_SET);
 393         }
 394     }
 395
 396     if (!found_data)
 397         return AVERROR_INVALIDDATA;
 398
 399 found_data:
 400     if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
 401         if (caf->data_size > 0 && caf->data_size / caf->bytes_per_packet < INT64_MAX / caf->frames_per_packet)
 402             st->nb_frames = (caf->data_size / caf->bytes_per_packet) * caf->frames_per_packet;
 403     } else if (ffstream(st)->nb_index_entries && st->duration > 0) {
 404         if (st->codecpar->sample_rate && caf->data_size / st->duration > INT64_MAX / st->codecpar->sample_rate / 8) {
 405             av_log(s, AV_LOG_ERROR, "Overflow during bit rate calculation %d * 8 * %"PRId64"\n",
 406                    st->codecpar->sample_rate, caf->data_size / st->duration);
 407             return AVERROR_INVALIDDATA;
 408         }
 409         st->codecpar->bit_rate = st->codecpar->sample_rate * 8LL *
 410                                  (caf->data_size / st->duration);
 411     } else {
 412         av_log(s, AV_LOG_ERROR, "Missing packet table. It is required when "
 413                                 "block size or frame size are variable.\n");
 414         return AVERROR_INVALIDDATA;
 415     }
 416
 417     avpriv_set_pts_info(st, 64, 1, st->codecpar->sample_rate);
 418     st->start_time = 0;
 419
 420     /* position the stream at the start of data */
 421     if (caf->data_size >= 0)
 422         avio_seek(pb, caf->data_start, SEEK_SET);
 423
 424     return 0;
 425 }
 426
 427 #define CAF_MAX_PKT_SIZE 4096
 428
 429 static int read_packet(AVFormatContext *s, AVPacket *pkt)
 430 {
 431     AVIOContext *pb = s->pb;
 432     AVStream *st      = s->streams[0];
 433     FFStream *const sti = ffstream(st);
 434     CafContext *caf   = s->priv_data;
 435     int res, pkt_size = 0, pkt_frames = 0;
 436     int64_t left      = CAF_MAX_PKT_SIZE;
 437
 438     if (avio_feof(pb))
 439         return AVERROR_EOF;
 440
 441     /* don't read past end of data chunk */
 442     if (caf->data_size > 0) {
 443         left = (caf->data_start + caf->data_size) - avio_tell(pb);
 444         if (!left)
 445             return AVERROR_EOF;
 446         if (left < 0)
 447             return AVERROR_INVALIDDATA;
 448     }
 449
 450     pkt_frames = caf->frames_per_packet;
 451     pkt_size   = caf->bytes_per_packet;
 452
 453     if (pkt_size > 0 && pkt_frames == 1) {
 454         pkt_size   = (CAF_MAX_PKT_SIZE / pkt_size) * pkt_size;
 455         pkt_size   = FFMIN(pkt_size, left);
 456         pkt_frames = pkt_size / caf->bytes_per_packet;
 457     } else if (sti->nb_index_entries) {
 458         if (caf->packet_cnt < sti->nb_index_entries - 1) {
 459             pkt_size   = sti->index_entries[caf->packet_cnt + 1].pos       - sti->index_entries[caf->packet_cnt].pos;
 460             pkt_frames = sti->index_entries[caf->packet_cnt + 1].timestamp - sti->index_entries[caf->packet_cnt].timestamp;
 461         } else if (caf->packet_cnt == sti->nb_index_entries - 1) {
 462             pkt_size   = caf->num_bytes - sti->index_entries[caf->packet_cnt].pos;
 463             pkt_frames = st->duration   - sti->index_entries[caf->packet_cnt].timestamp;
 464         } else {
 465             return AVERROR_INVALIDDATA;
 466         }
 467     }
 468
 469     if (pkt_size == 0 || pkt_frames == 0 || pkt_size > left)
 470         return AVERROR_INVALIDDATA;
 471
 472     res = av_get_packet(pb, pkt, pkt_size);
 473     if (res < 0)
 474         return res;
 475
 476     pkt->size           = res;
 477     pkt->stream_index   = 0;
 478     pkt->dts = pkt->pts = caf->frame_cnt;
 479
 480     caf->packet_cnt++;
 481     caf->frame_cnt += pkt_frames;
 482
 483     return 0;
 484 }
 485
 486 static int read_seek(AVFormatContext *s, int stream_index,
 487                      int64_t timestamp, int flags)
 488 {
 489     AVStream *st = s->streams[0];
 490     FFStream *const sti = ffstream(st);
 491     CafContext *caf = s->priv_data;
 492     int64_t pos, packet_cnt, frame_cnt;
 493
 494     timestamp = FFMAX(timestamp, 0);
 495
 496     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
 497         /* calculate new byte position based on target frame position */
 498         pos = caf->bytes_per_packet * (timestamp / caf->frames_per_packet);
 499         if (caf->data_size > 0)
 500             pos = FFMIN(pos, caf->data_size);
 501         packet_cnt = pos / caf->bytes_per_packet;
 502         frame_cnt  = caf->frames_per_packet * packet_cnt;
 503     } else if (sti->nb_index_entries) {
 504         packet_cnt = av_index_search_timestamp(st, timestamp, flags);
 505         frame_cnt  = sti->index_entries[packet_cnt].timestamp;
 506         pos        = sti->index_entries[packet_cnt].pos;
 507     } else {
 508         return -1;
 509     }
 510
 511     if (avio_seek(s->pb, pos + caf->data_start, SEEK_SET) < 0)
 512         return -1;
 513
 514     caf->packet_cnt = packet_cnt;
 515     caf->frame_cnt  = frame_cnt;
 516
 517     return 0;
 518 }
 519
 520 const FFInputFormat ff_caf_demuxer = {
 521     .p.name         = "caf",
 522     .p.long_name    = NULL_IF_CONFIG_SMALL("Apple CAF (Core Audio Format)"),
 523     .p.codec_tag    = ff_caf_codec_tags_list,
 524     .priv_data_size = sizeof(CafContext),
 525     .read_probe     = probe,
 526     .read_header    = read_header,
 527     .read_packet    = read_packet,
 528     .read_seek      = read_seek,
 529 };