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 Libav.
   7  *
   8  * Libav 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  * Libav 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 Libav; 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 "avformat.h"
  29 #include "internal.h"
  30 #include "isom.h"
  31 #include "mov_chan.h"
  32 #include "libavutil/intreadwrite.h"
  33 #include "libavutil/intfloat.h"
  34 #include "libavutil/dict.h"
  35 #include "caf.h"
  36
  37 typedef struct {
  38     int bytes_per_packet;           ///< bytes in a packet, or 0 if variable
  39     int frames_per_packet;          ///< frames in a packet, or 0 if variable
  40     int64_t num_bytes;              ///< total number of bytes in stream
  41
  42     int64_t packet_cnt;             ///< packet counter
  43     int64_t frame_cnt;              ///< frame counter
  44
  45     int64_t data_start;             ///< data start position, in bytes
  46     int64_t data_size;              ///< raw data size, in bytes
  47 } CaffContext;
  48
  49 static int probe(AVProbeData *p)
  50 {
  51     if (AV_RB32(p->buf) == MKBETAG('c','a','f','f') && AV_RB16(&p->buf[4]) == 1)
  52         return AVPROBE_SCORE_MAX;
  53     return 0;
  54 }
  55
  56 /** Read audio description chunk */
  57 static int read_desc_chunk(AVFormatContext *s)
  58 {
  59     AVIOContext *pb = s->pb;
  60     CaffContext *caf  = s->priv_data;
  61     AVStream *st;
  62     int flags;
  63
  64     /* new audio stream */
  65     st = avformat_new_stream(s, NULL);
  66     if (!st)
  67         return AVERROR(ENOMEM);
  68
  69     /* parse format description */
  70     st->codec->codec_type  = AVMEDIA_TYPE_AUDIO;
  71     st->codec->sample_rate = av_int2double(avio_rb64(pb));
  72     st->codec->codec_tag   = avio_rb32(pb);
  73     flags = avio_rb32(pb);
  74     caf->bytes_per_packet  = avio_rb32(pb);
  75     st->codec->block_align = caf->bytes_per_packet;
  76     caf->frames_per_packet = avio_rb32(pb);
  77     st->codec->channels    = avio_rb32(pb);
  78     st->codec->bits_per_coded_sample = avio_rb32(pb);
  79
  80     /* calculate bit rate for constant size packets */
  81     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
  82         st->codec->bit_rate = (uint64_t)st->codec->sample_rate * (uint64_t)caf->bytes_per_packet * 8
  83                               / (uint64_t)caf->frames_per_packet;
  84     } else {
  85         st->codec->bit_rate = 0;
  86     }
  87
  88     /* determine codec */
  89     if (st->codec->codec_tag == MKBETAG('l','p','c','m'))
  90         st->codec->codec_id = ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, (flags ^ 0x2) | 0x4);
  91     else
  92         st->codec->codec_id = ff_codec_get_id(ff_codec_caf_tags, st->codec->codec_tag);
  93     return 0;
  94 }
  95
  96 /** Read magic cookie chunk */
  97 static int read_kuki_chunk(AVFormatContext *s, int64_t size)
  98 {
  99     AVIOContext *pb = s->pb;
 100     AVStream *st      = s->streams[0];
 101
 102     if (size < 0 || size > INT_MAX - FF_INPUT_BUFFER_PADDING_SIZE)
 103         return -1;
 104
 105     if (st->codec->codec_id == AV_CODEC_ID_AAC) {
 106         /* The magic cookie format for AAC is an mp4 esds atom.
 107            The lavc AAC decoder requires the data from the codec specific
 108            description as extradata input. */
 109         int strt, skip;
 110         MOVAtom atom;
 111
 112         strt = avio_tell(pb);
 113         ff_mov_read_esds(s, pb, atom);
 114         skip = size - (avio_tell(pb) - strt);
 115         if (skip < 0 || !st->codec->extradata ||
 116             st->codec->codec_id != AV_CODEC_ID_AAC) {
 117             av_log(s, AV_LOG_ERROR, "invalid AAC magic cookie\n");
 118             return AVERROR_INVALIDDATA;
 119         }
 120         avio_skip(pb, skip);
 121     } else if (st->codec->codec_id == AV_CODEC_ID_ALAC) {
 122 #define ALAC_PREAMBLE 12
 123 #define ALAC_HEADER   36
 124 #define ALAC_NEW_KUKI 24
 125         uint8_t preamble[12];
 126         if (size < ALAC_NEW_KUKI) {
 127             av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
 128             avio_skip(pb, size);
 129             return AVERROR_INVALIDDATA;
 130         }
 131         avio_read(pb, preamble, ALAC_PREAMBLE);
 132
 133         st->codec->extradata = av_mallocz(ALAC_HEADER + FF_INPUT_BUFFER_PADDING_SIZE);
 134         if (!st->codec->extradata)
 135             return AVERROR(ENOMEM);
 136
 137         /* For the old style cookie, we skip 12 bytes, then read 36 bytes.
 138          * The new style cookie only contains the last 24 bytes of what was
 139          * 36 bytes in the old style cookie, so we fabricate the first 12 bytes
 140          * in that case to maintain compatibility. */
 141         if (!memcmp(&preamble[4], "frmaalac", 8)) {
 142             if (size < ALAC_PREAMBLE + ALAC_HEADER) {
 143                 av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
 144                 av_freep(&st->codec->extradata);
 145                 return AVERROR_INVALIDDATA;
 146             }
 147             avio_read(pb, st->codec->extradata, ALAC_HEADER);
 148             avio_skip(pb, size - ALAC_PREAMBLE - ALAC_HEADER);
 149         } else {
 150             AV_WB32(st->codec->extradata, 36);
 151             memcpy(&st->codec->extradata[4], "alac", 4);
 152             AV_WB32(&st->codec->extradata[8], 0);
 153             memcpy(&st->codec->extradata[12], preamble, 12);
 154             avio_read(pb, &st->codec->extradata[24], ALAC_NEW_KUKI - 12);
 155             avio_skip(pb, size - ALAC_NEW_KUKI);
 156         }
 157         st->codec->extradata_size = ALAC_HEADER;
 158     } else {
 159         st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
 160         if (!st->codec->extradata)
 161             return AVERROR(ENOMEM);
 162         avio_read(pb, st->codec->extradata, size);
 163         st->codec->extradata_size = size;
 164     }
 165
 166     return 0;
 167 }
 168
 169 /** Read packet table chunk */
 170 static int read_pakt_chunk(AVFormatContext *s, int64_t size)
 171 {
 172     AVIOContext *pb = s->pb;
 173     AVStream *st      = s->streams[0];
 174     CaffContext *caf  = s->priv_data;
 175     int64_t pos = 0, ccount, num_packets;
 176     int i;
 177
 178     ccount = avio_tell(pb);
 179
 180     num_packets = avio_rb64(pb);
 181     if (num_packets < 0 || INT32_MAX / sizeof(AVIndexEntry) < num_packets)
 182         return AVERROR_INVALIDDATA;
 183
 184     st->nb_frames  = avio_rb64(pb); /* valid frames */
 185     st->nb_frames += avio_rb32(pb); /* priming frames */
 186     st->nb_frames += avio_rb32(pb); /* remainder frames */
 187
 188     st->duration = 0;
 189     for (i = 0; i < num_packets; i++) {
 190         av_add_index_entry(s->streams[0], pos, st->duration, 0, 0, AVINDEX_KEYFRAME);
 191         pos += caf->bytes_per_packet ? caf->bytes_per_packet : ff_mp4_read_descr_len(pb);
 192         st->duration += caf->frames_per_packet ? caf->frames_per_packet : ff_mp4_read_descr_len(pb);
 193     }
 194
 195     if (avio_tell(pb) - ccount > size) {
 196         av_log(s, AV_LOG_ERROR, "error reading packet table\n");
 197         return AVERROR_INVALIDDATA;
 198     }
 199     avio_skip(pb, ccount + size - avio_tell(pb));
 200
 201     caf->num_bytes = pos;
 202     return 0;
 203 }
 204
 205 /** Read information chunk */
 206 static void read_info_chunk(AVFormatContext *s, int64_t size)
 207 {
 208     AVIOContext *pb = s->pb;
 209     unsigned int i;
 210     unsigned int nb_entries = avio_rb32(pb);
 211     for (i = 0; i < nb_entries; i++) {
 212         char key[32];
 213         char value[1024];
 214         avio_get_str(pb, INT_MAX, key,   sizeof(key));
 215         avio_get_str(pb, INT_MAX, value, sizeof(value));
 216         av_dict_set(&s->metadata, key, value, 0);
 217     }
 218 }
 219
 220 static int read_header(AVFormatContext *s)
 221 {
 222     AVIOContext *pb = s->pb;
 223     CaffContext *caf  = s->priv_data;
 224     AVStream *st;
 225     uint32_t tag = 0;
 226     int found_data, ret;
 227     int64_t size;
 228
 229     avio_skip(pb, 8); /* magic, version, file flags */
 230
 231     /* audio description chunk */
 232     if (avio_rb32(pb) != MKBETAG('d','e','s','c')) {
 233         av_log(s, AV_LOG_ERROR, "desc chunk not present\n");
 234         return AVERROR_INVALIDDATA;
 235     }
 236     size = avio_rb64(pb);
 237     if (size != 32)
 238         return AVERROR_INVALIDDATA;
 239
 240     ret = read_desc_chunk(s);
 241     if (ret)
 242         return ret;
 243     st = s->streams[0];
 244
 245     /* parse each chunk */
 246     found_data = 0;
 247     while (!pb->eof_reached) {
 248
 249         /* stop at data chunk if seeking is not supported or
 250            data chunk size is unknown */
 251         if (found_data && (caf->data_size < 0 || !pb->seekable))
 252             break;
 253
 254         tag  = avio_rb32(pb);
 255         size = avio_rb64(pb);
 256         if (pb->eof_reached)
 257             break;
 258
 259         switch (tag) {
 260         case MKBETAG('d','a','t','a'):
 261             avio_skip(pb, 4); /* edit count */
 262             caf->data_start = avio_tell(pb);
 263             caf->data_size  = size < 0 ? -1 : size - 4;
 264             if (caf->data_size > 0 && pb->seekable)
 265                 avio_skip(pb, caf->data_size);
 266             found_data = 1;
 267             break;
 268
 269         case MKBETAG('c','h','a','n'):
 270             if ((ret = ff_mov_read_chan(s, s->pb, st, size)) < 0)
 271                 return ret;
 272             break;
 273
 274         /* magic cookie chunk */
 275         case MKBETAG('k','u','k','i'):
 276             if (read_kuki_chunk(s, size))
 277                 return AVERROR_INVALIDDATA;
 278             break;
 279
 280         /* packet table chunk */
 281         case MKBETAG('p','a','k','t'):
 282             if (read_pakt_chunk(s, size))
 283                 return AVERROR_INVALIDDATA;
 284             break;
 285
 286         case MKBETAG('i','n','f','o'):
 287             read_info_chunk(s, size);
 288             break;
 289
 290         default:
 291 #define _(x) ((x) >= ' ' ? (x) : ' ')
 292             av_log(s, AV_LOG_WARNING, "skipping CAF chunk: %08X (%c%c%c%c)\n",
 293                 tag, _(tag>>24), _((tag>>16)&0xFF), _((tag>>8)&0xFF), _(tag&0xFF));
 294 #undef _
 295         case MKBETAG('f','r','e','e'):
 296             if (size < 0)
 297                 return AVERROR_INVALIDDATA;
 298             avio_skip(pb, size);
 299             break;
 300         }
 301     }
 302
 303     if (!found_data)
 304         return AVERROR_INVALIDDATA;
 305
 306     if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
 307         if (caf->data_size > 0)
 308             st->nb_frames = (caf->data_size / caf->bytes_per_packet) * caf->frames_per_packet;
 309     } else if (st->nb_index_entries) {
 310         st->codec->bit_rate = st->codec->sample_rate * caf->data_size * 8 /
 311                               st->duration;
 312     } else {
 313         av_log(s, AV_LOG_ERROR, "Missing packet table. It is required when "
 314                                 "block size or frame size are variable.\n");
 315         return AVERROR_INVALIDDATA;
 316     }
 317
 318     avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
 319     st->start_time = 0;
 320
 321     /* position the stream at the start of data */
 322     if (caf->data_size >= 0)
 323         avio_seek(pb, caf->data_start, SEEK_SET);
 324
 325     return 0;
 326 }
 327
 328 #define CAF_MAX_PKT_SIZE 4096
 329
 330 static int read_packet(AVFormatContext *s, AVPacket *pkt)
 331 {
 332     AVIOContext *pb = s->pb;
 333     AVStream *st      = s->streams[0];
 334     CaffContext *caf  = s->priv_data;
 335     int res, pkt_size = 0, pkt_frames = 0;
 336     int64_t left      = CAF_MAX_PKT_SIZE;
 337
 338     if (pb->eof_reached)
 339         return AVERROR(EIO);
 340
 341     /* don't read past end of data chunk */
 342     if (caf->data_size > 0) {
 343         left = (caf->data_start + caf->data_size) - avio_tell(pb);
 344         if (left <= 0)
 345             return AVERROR(EIO);
 346     }
 347
 348     pkt_frames = caf->frames_per_packet;
 349     pkt_size   = caf->bytes_per_packet;
 350
 351     if (pkt_size > 0 && pkt_frames == 1) {
 352         pkt_size   = (CAF_MAX_PKT_SIZE / pkt_size) * pkt_size;
 353         pkt_size   = FFMIN(pkt_size, left);
 354         pkt_frames = pkt_size / caf->bytes_per_packet;
 355     } else if (st->nb_index_entries) {
 356         if (caf->packet_cnt < st->nb_index_entries - 1) {
 357             pkt_size   = st->index_entries[caf->packet_cnt + 1].pos       - st->index_entries[caf->packet_cnt].pos;
 358             pkt_frames = st->index_entries[caf->packet_cnt + 1].timestamp - st->index_entries[caf->packet_cnt].timestamp;
 359         } else if (caf->packet_cnt == st->nb_index_entries - 1) {
 360             pkt_size   = caf->num_bytes - st->index_entries[caf->packet_cnt].pos;
 361             pkt_frames = st->duration   - st->index_entries[caf->packet_cnt].timestamp;
 362         } else {
 363             return AVERROR(EIO);
 364         }
 365     }
 366
 367     if (pkt_size == 0 || pkt_frames == 0 || pkt_size > left)
 368         return AVERROR(EIO);
 369
 370     res = av_get_packet(pb, pkt, pkt_size);
 371     if (res < 0)
 372         return res;
 373
 374     pkt->size           = res;
 375     pkt->stream_index   = 0;
 376     pkt->dts = pkt->pts = caf->frame_cnt;
 377
 378     caf->packet_cnt++;
 379     caf->frame_cnt += pkt_frames;
 380
 381     return 0;
 382 }
 383
 384 static int read_seek(AVFormatContext *s, int stream_index,
 385                      int64_t timestamp, int flags)
 386 {
 387     AVStream *st = s->streams[0];
 388     CaffContext *caf = s->priv_data;
 389     int64_t pos, packet_cnt, frame_cnt;
 390
 391     timestamp = FFMAX(timestamp, 0);
 392
 393     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
 394         /* calculate new byte position based on target frame position */
 395         pos = caf->bytes_per_packet * timestamp / caf->frames_per_packet;
 396         if (caf->data_size > 0)
 397             pos = FFMIN(pos, caf->data_size);
 398         packet_cnt = pos / caf->bytes_per_packet;
 399         frame_cnt  = caf->frames_per_packet * packet_cnt;
 400     } else if (st->nb_index_entries) {
 401         packet_cnt = av_index_search_timestamp(st, timestamp, flags);
 402         frame_cnt  = st->index_entries[packet_cnt].timestamp;
 403         pos        = st->index_entries[packet_cnt].pos;
 404     } else {
 405         return -1;
 406     }
 407
 408     if (avio_seek(s->pb, pos + caf->data_start, SEEK_SET) < 0)
 409         return -1;
 410
 411     caf->packet_cnt = packet_cnt;
 412     caf->frame_cnt  = frame_cnt;
 413
 414     return 0;
 415 }
 416
 417 AVInputFormat ff_caf_demuxer = {
 418     .name           = "caf",
 419     .long_name      = NULL_IF_CONFIG_SMALL("Apple CAF (Core Audio Format)"),
 420     .priv_data_size = sizeof(CaffContext),
 421     .read_probe     = probe,
 422     .read_header    = read_header,
 423     .read_packet    = read_packet,
 424     .read_seek      = read_seek,
 425     .codec_tag      = (const AVCodecTag* const []){ ff_codec_caf_tags, 0 },
 426 };