FFmpeg: libavformat/cafdec.c Source File

00001 /*
00002  * Core Audio Format demuxer
00003  * Copyright (c) 2007 Justin Ruggles
00004  * Copyright (c) 2009 Peter Ross
00005  *
00006  * This file is part of FFmpeg.
00007  *
00008  * FFmpeg is free software; you can redistribute it and/or
00009  * modify it under the terms of the GNU Lesser General Public
00010  * License as published by the Free Software Foundation; either
00011  * version 2.1 of the License, or (at your option) any later version.
00012  *
00013  * FFmpeg is distributed in the hope that it will be useful,
00014  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00015  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00016  * Lesser General Public License for more details.
00017  *
00018  * You should have received a copy of the GNU Lesser General Public
00019  * License along with FFmpeg; if not, write to the Free Software
00020  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00021  */
00022 
00028 #include "avformat.h"
00029 #include "internal.h"
00030 #include "riff.h"
00031 #include "isom.h"
00032 #include "libavutil/intreadwrite.h"
00033 #include "libavutil/intfloat.h"
00034 #include "libavutil/dict.h"
00035 #include "caf.h"
00036 
00037 typedef struct {
00038     int bytes_per_packet;           
00039     int frames_per_packet;          
00040     int64_t num_bytes;              
00041 
00042     int64_t packet_cnt;             
00043     int64_t frame_cnt;              
00044 
00045     int64_t data_start;             
00046     int64_t data_size;              
00047 } CaffContext;
00048 
00049 static int probe(AVProbeData *p)
00050 {
00051     if (AV_RB32(p->buf) == MKBETAG('c','a','f','f') && AV_RB16(&p->buf[4]) == 1)
00052         return AVPROBE_SCORE_MAX;
00053     return 0;
00054 }
00055 
00057 static int read_desc_chunk(AVFormatContext *s)
00058 {
00059     AVIOContext *pb = s->pb;
00060     CaffContext *caf  = s->priv_data;
00061     AVStream *st;
00062     int flags;
00063 
00064     /* new audio stream */
00065     st = avformat_new_stream(s, NULL);
00066     if (!st)
00067         return AVERROR(ENOMEM);
00068 
00069     /* parse format description */
00070     st->codec->codec_type  = AVMEDIA_TYPE_AUDIO;
00071     st->codec->sample_rate = av_int2double(avio_rb64(pb));
00072     st->codec->codec_tag   = avio_rl32(pb);
00073     flags = avio_rb32(pb);
00074     caf->bytes_per_packet  = avio_rb32(pb);
00075     st->codec->block_align = caf->bytes_per_packet;
00076     caf->frames_per_packet = avio_rb32(pb);
00077     st->codec->channels    = avio_rb32(pb);
00078     st->codec->bits_per_coded_sample = avio_rb32(pb);
00079 
00080     /* calculate bit rate for constant size packets */
00081     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
00082         st->codec->bit_rate = (uint64_t)st->codec->sample_rate * (uint64_t)caf->bytes_per_packet * 8
00083                               / (uint64_t)caf->frames_per_packet;
00084     } else {
00085         st->codec->bit_rate = 0;
00086     }
00087 
00088     /* determine codec */
00089     if (st->codec->codec_tag == MKTAG('l','p','c','m'))
00090         st->codec->codec_id = ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, (flags ^ 0x2) | 0x4);
00091     else
00092         st->codec->codec_id = ff_codec_get_id(ff_codec_caf_tags, st->codec->codec_tag);
00093     return 0;
00094 }
00095 
00097 static int read_kuki_chunk(AVFormatContext *s, int64_t size)
00098 {
00099     AVIOContext *pb = s->pb;
00100     AVStream *st      = s->streams[0];
00101 
00102     if (size < 0 || size > INT_MAX - FF_INPUT_BUFFER_PADDING_SIZE)
00103         return -1;
00104 
00105     if (st->codec->codec_id == CODEC_ID_AAC) {
00106         /* The magic cookie format for AAC is an mp4 esds atom.
00107            The lavc AAC decoder requires the data from the codec specific
00108            description as extradata input. */
00109         int strt, skip;
00110         MOVAtom atom;
00111 
00112         strt = avio_tell(pb);
00113         ff_mov_read_esds(s, pb, atom);
00114         skip = size - (avio_tell(pb) - strt);
00115         if (skip < 0 || !st->codec->extradata ||
00116             st->codec->codec_id != CODEC_ID_AAC) {
00117             av_log(s, AV_LOG_ERROR, "invalid AAC magic cookie\n");
00118             return AVERROR_INVALIDDATA;
00119         }
00120         avio_skip(pb, skip);
00121     } else if (st->codec->codec_id == CODEC_ID_ALAC) {
00122 #define ALAC_PREAMBLE 12
00123 #define ALAC_HEADER   36
00124 #define ALAC_NEW_KUKI 24
00125         if (size == ALAC_NEW_KUKI) {
00126             st->codec->extradata = av_mallocz(ALAC_HEADER + FF_INPUT_BUFFER_PADDING_SIZE);
00127             if (!st->codec->extradata)
00128                 return AVERROR(ENOMEM);
00129             memcpy(st->codec->extradata, "\0\0\0\24alac", 8);
00130             avio_read(pb, st->codec->extradata + ALAC_HEADER - ALAC_NEW_KUKI, ALAC_NEW_KUKI);
00131             st->codec->extradata_size = ALAC_HEADER;
00132         } else {
00133             if (size < ALAC_PREAMBLE + ALAC_HEADER) {
00134                 av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
00135                 avio_skip(pb, size);
00136                 return AVERROR_INVALIDDATA;
00137             }
00138             avio_skip(pb, ALAC_PREAMBLE);
00139             st->codec->extradata = av_mallocz(ALAC_HEADER + FF_INPUT_BUFFER_PADDING_SIZE);
00140             if (!st->codec->extradata)
00141                 return AVERROR(ENOMEM);
00142             avio_read(pb, st->codec->extradata, ALAC_HEADER);
00143             st->codec->extradata_size = ALAC_HEADER;
00144             avio_skip(pb, size - ALAC_PREAMBLE - ALAC_HEADER);
00145         }
00146     } else {
00147         st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
00148         if (!st->codec->extradata)
00149             return AVERROR(ENOMEM);
00150         avio_read(pb, st->codec->extradata, size);
00151         st->codec->extradata_size = size;
00152     }
00153 
00154     return 0;
00155 }
00156 
00158 static int read_pakt_chunk(AVFormatContext *s, int64_t size)
00159 {
00160     AVIOContext *pb = s->pb;
00161     AVStream *st      = s->streams[0];
00162     CaffContext *caf  = s->priv_data;
00163     int64_t pos = 0, ccount;
00164     int num_packets, i;
00165 
00166     ccount = avio_tell(pb);
00167 
00168     num_packets = avio_rb64(pb);
00169     if (num_packets < 0 || INT32_MAX / sizeof(AVIndexEntry) < num_packets)
00170         return AVERROR_INVALIDDATA;
00171 
00172     st->nb_frames  = avio_rb64(pb); /* valid frames */
00173     st->nb_frames += avio_rb32(pb); /* priming frames */
00174     st->nb_frames += avio_rb32(pb); /* remainder frames */
00175 
00176     st->duration = 0;
00177     for (i = 0; i < num_packets; i++) {
00178         av_add_index_entry(s->streams[0], pos, st->duration, 0, 0, AVINDEX_KEYFRAME);
00179         pos += caf->bytes_per_packet ? caf->bytes_per_packet : ff_mp4_read_descr_len(pb);
00180         st->duration += caf->frames_per_packet ? caf->frames_per_packet : ff_mp4_read_descr_len(pb);
00181     }
00182 
00183     if (avio_tell(pb) - ccount != size) {
00184         av_log(s, AV_LOG_ERROR, "error reading packet table\n");
00185         return -1;
00186     }
00187 
00188     caf->num_bytes = pos;
00189     return 0;
00190 }
00191 
00193 static void read_info_chunk(AVFormatContext *s, int64_t size)
00194 {
00195     AVIOContext *pb = s->pb;
00196     unsigned int i;
00197     unsigned int nb_entries = avio_rb32(pb);
00198     for (i = 0; i < nb_entries; i++) {
00199         char key[32];
00200         char value[1024];
00201         avio_get_str(pb, INT_MAX, key, sizeof(key));
00202         avio_get_str(pb, INT_MAX, value, sizeof(value));
00203         av_dict_set(&s->metadata, key, value, 0);
00204     }
00205 }
00206 
00207 static int read_header(AVFormatContext *s)
00208 {
00209     AVIOContext *pb = s->pb;
00210     CaffContext *caf  = s->priv_data;
00211     AVStream *st;
00212     uint32_t tag = 0;
00213     int found_data, ret;
00214     int64_t size;
00215 
00216     avio_skip(pb, 8); /* magic, version, file flags */
00217 
00218     /* audio description chunk */
00219     if (avio_rb32(pb) != MKBETAG('d','e','s','c')) {
00220         av_log(s, AV_LOG_ERROR, "desc chunk not present\n");
00221         return AVERROR_INVALIDDATA;
00222     }
00223     size = avio_rb64(pb);
00224     if (size != 32)
00225         return AVERROR_INVALIDDATA;
00226 
00227     ret = read_desc_chunk(s);
00228     if (ret)
00229         return ret;
00230     st = s->streams[0];
00231 
00232     /* parse each chunk */
00233     found_data = 0;
00234     while (!url_feof(pb)) {
00235 
00236         /* stop at data chunk if seeking is not supported or
00237            data chunk size is unknown */
00238         if (found_data && (caf->data_size < 0 || !pb->seekable))
00239             break;
00240 
00241         tag  = avio_rb32(pb);
00242         size = avio_rb64(pb);
00243         if (url_feof(pb))
00244             break;
00245 
00246         switch (tag) {
00247         case MKBETAG('d','a','t','a'):
00248             avio_skip(pb, 4); /* edit count */
00249             caf->data_start = avio_tell(pb);
00250             caf->data_size  = size < 0 ? -1 : size - 4;
00251             if (caf->data_size > 0 && pb->seekable)
00252                 avio_skip(pb, caf->data_size);
00253             found_data = 1;
00254             break;
00255 
00256         /* magic cookie chunk */
00257         case MKBETAG('k','u','k','i'):
00258             if (read_kuki_chunk(s, size))
00259                 return AVERROR_INVALIDDATA;
00260             break;
00261 
00262         /* packet table chunk */
00263         case MKBETAG('p','a','k','t'):
00264             if (read_pakt_chunk(s, size))
00265                 return AVERROR_INVALIDDATA;
00266             break;
00267 
00268         case MKBETAG('i','n','f','o'):
00269             read_info_chunk(s, size);
00270             break;
00271 
00272         case MKBETAG('c','h','a','n'):
00273             if (size < 12)
00274                 return AVERROR_INVALIDDATA;
00275             ff_mov_read_chan(s, size, st->codec);
00276             break;
00277 
00278         default:
00279 #define _(x) ((x) >= ' ' ? (x) : ' ')
00280             av_log(s, AV_LOG_WARNING, "skipping CAF chunk: %08X (%c%c%c%c), size %"PRId64"\n",
00281                 tag, _(tag>>24), _((tag>>16)&0xFF), _((tag>>8)&0xFF), _(tag&0xFF), size);
00282 #undef _
00283         case MKBETAG('f','r','e','e'):
00284             if (size < 0)
00285                 return AVERROR_INVALIDDATA;
00286             avio_skip(pb, size);
00287             break;
00288         }
00289     }
00290 
00291     if (!found_data)
00292         return AVERROR_INVALIDDATA;
00293 
00294     if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
00295         if (caf->data_size > 0)
00296             st->nb_frames = (caf->data_size / caf->bytes_per_packet) * caf->frames_per_packet;
00297     } else if (st->nb_index_entries) {
00298         st->codec->bit_rate = st->codec->sample_rate * caf->data_size * 8 /
00299                               st->duration;
00300     } else {
00301         av_log(s, AV_LOG_ERROR, "Missing packet table. It is required when "
00302                                 "block size or frame size are variable.\n");
00303         return AVERROR_INVALIDDATA;
00304     }
00305 
00306     avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
00307     st->start_time = 0;
00308 
00309     /* position the stream at the start of data */
00310     if (caf->data_size >= 0)
00311         avio_seek(pb, caf->data_start, SEEK_SET);
00312 
00313     return 0;
00314 }
00315 
00316 #define CAF_MAX_PKT_SIZE 4096
00317 
00318 static int read_packet(AVFormatContext *s, AVPacket *pkt)
00319 {
00320     AVIOContext *pb = s->pb;
00321     AVStream *st      = s->streams[0];
00322     CaffContext *caf  = s->priv_data;
00323     int res, pkt_size = 0, pkt_frames = 0;
00324     int64_t left      = CAF_MAX_PKT_SIZE;
00325 
00326     if (url_feof(pb))
00327         return AVERROR(EIO);
00328 
00329     /* don't read past end of data chunk */
00330     if (caf->data_size > 0) {
00331         left = (caf->data_start + caf->data_size) - avio_tell(pb);
00332         if (left <= 0)
00333             return AVERROR(EIO);
00334     }
00335 
00336     pkt_frames = caf->frames_per_packet;
00337     pkt_size   = caf->bytes_per_packet;
00338 
00339     if (pkt_size > 0 && pkt_frames == 1) {
00340         pkt_size   = (CAF_MAX_PKT_SIZE / pkt_size) * pkt_size;
00341         pkt_size   = FFMIN(pkt_size, left);
00342         pkt_frames = pkt_size / caf->bytes_per_packet;
00343     } else if (st->nb_index_entries) {
00344         if (caf->packet_cnt < st->nb_index_entries - 1) {
00345             pkt_size   = st->index_entries[caf->packet_cnt + 1].pos       - st->index_entries[caf->packet_cnt].pos;
00346             pkt_frames = st->index_entries[caf->packet_cnt + 1].timestamp - st->index_entries[caf->packet_cnt].timestamp;
00347         } else if (caf->packet_cnt == st->nb_index_entries - 1) {
00348             pkt_size   = caf->num_bytes - st->index_entries[caf->packet_cnt].pos;
00349             pkt_frames = st->duration   - st->index_entries[caf->packet_cnt].timestamp;
00350         } else {
00351             return AVERROR(EIO);
00352         }
00353     }
00354 
00355     if (pkt_size == 0 || pkt_frames == 0 || pkt_size > left)
00356         return AVERROR(EIO);
00357 
00358     res = av_get_packet(pb, pkt, pkt_size);
00359     if (res < 0)
00360         return res;
00361 
00362     pkt->size           = res;
00363     pkt->stream_index   = 0;
00364     pkt->dts = pkt->pts = caf->frame_cnt;
00365 
00366     caf->packet_cnt++;
00367     caf->frame_cnt += pkt_frames;
00368 
00369     return 0;
00370 }
00371 
00372 static int read_seek(AVFormatContext *s, int stream_index,
00373                      int64_t timestamp, int flags)
00374 {
00375     AVStream *st = s->streams[0];
00376     CaffContext *caf = s->priv_data;
00377     int64_t pos, packet_cnt, frame_cnt;
00378 
00379     timestamp = FFMAX(timestamp, 0);
00380 
00381     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
00382         /* calculate new byte position based on target frame position */
00383         pos = caf->bytes_per_packet * timestamp / caf->frames_per_packet;
00384         if (caf->data_size > 0)
00385             pos = FFMIN(pos, caf->data_size);
00386         packet_cnt = pos / caf->bytes_per_packet;
00387         frame_cnt  = caf->frames_per_packet * packet_cnt;
00388     } else if (st->nb_index_entries) {
00389         packet_cnt = av_index_search_timestamp(st, timestamp, flags);
00390         frame_cnt  = st->index_entries[packet_cnt].timestamp;
00391         pos        = st->index_entries[packet_cnt].pos;
00392     } else {
00393         return -1;
00394     }
00395 
00396     if (avio_seek(s->pb, pos + caf->data_start, SEEK_SET) < 0)
00397         return -1;
00398 
00399     caf->packet_cnt = packet_cnt;
00400     caf->frame_cnt  = frame_cnt;
00401 
00402     return 0;
00403 }
00404 
00405 AVInputFormat ff_caf_demuxer = {
00406     .name           = "caf",
00407     .long_name      = NULL_IF_CONFIG_SMALL("Apple Core Audio Format"),
00408     .priv_data_size = sizeof(CaffContext),
00409     .read_probe     = probe,
00410     .read_header    = read_header,
00411     .read_packet    = read_packet,
00412     .read_seek      = read_seek,
00413     .codec_tag      = (const AVCodecTag*[]){ ff_codec_caf_tags, 0 },
00414 };