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 "riff.h"
00030 #include "isom.h"
00031 #include "libavutil/intreadwrite.h"
00032 #include "caf.h"
00033 
00034 typedef struct {
00035     int bytes_per_packet;           
00036     int frames_per_packet;          
00037     int64_t num_bytes;              
00038 
00039     int64_t packet_cnt;             
00040     int64_t frame_cnt;              
00041 
00042     int64_t data_start;             
00043     int64_t data_size;              
00044 } CaffContext;
00045 
00046 static int probe(AVProbeData *p)
00047 {
00048     if (AV_RB32(p->buf) == MKBETAG('c','a','f','f') && AV_RB16(&p->buf[4]) == 1)
00049         return AVPROBE_SCORE_MAX;
00050     return 0;
00051 }
00052 
00054 static int read_desc_chunk(AVFormatContext *s)
00055 {
00056     ByteIOContext *pb = s->pb;
00057     CaffContext *caf  = s->priv_data;
00058     AVStream *st;
00059     int flags;
00060 
00061     /* new audio stream */
00062     st = av_new_stream(s, 0);
00063     if (!st)
00064         return AVERROR(ENOMEM);
00065 
00066     /* parse format description */
00067     st->codec->codec_type  = AVMEDIA_TYPE_AUDIO;
00068     st->codec->sample_rate = av_int2dbl(get_be64(pb));
00069     st->codec->codec_tag   = get_be32(pb);
00070     flags = get_be32(pb);
00071     caf->bytes_per_packet  = get_be32(pb);
00072     st->codec->block_align = caf->bytes_per_packet;
00073     caf->frames_per_packet = get_be32(pb);
00074     st->codec->channels    = get_be32(pb);
00075     st->codec->bits_per_coded_sample = get_be32(pb);
00076 
00077     /* calculate bit rate for constant size packets */
00078     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
00079         st->codec->bit_rate = (uint64_t)st->codec->sample_rate * (uint64_t)caf->bytes_per_packet * 8
00080                               / (uint64_t)caf->frames_per_packet;
00081     } else {
00082         st->codec->bit_rate = 0;
00083     }
00084 
00085     /* determine codec */
00086     if (st->codec->codec_tag == MKBETAG('l','p','c','m'))
00087         st->codec->codec_id = ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, (flags ^ 0x2) | 0x4);
00088     else
00089         st->codec->codec_id = ff_codec_get_id(ff_codec_caf_tags, st->codec->codec_tag);
00090     return 0;
00091 }
00092 
00094 static int read_kuki_chunk(AVFormatContext *s, int64_t size)
00095 {
00096     ByteIOContext *pb = s->pb;
00097     AVStream *st      = s->streams[0];
00098 
00099     if (size < 0 || size > INT_MAX - FF_INPUT_BUFFER_PADDING_SIZE)
00100         return -1;
00101 
00102     if (st->codec->codec_id == CODEC_ID_AAC) {
00103         /* The magic cookie format for AAC is an mp4 esds atom.
00104            The lavc AAC decoder requires the data from the codec specific
00105            description as extradata input. */
00106         int strt, skip;
00107         MOVAtom atom;
00108 
00109         strt = url_ftell(pb);
00110         ff_mov_read_esds(s, pb, atom);
00111         skip = size - (url_ftell(pb) - strt);
00112         if (skip < 0 || !st->codec->extradata ||
00113             st->codec->codec_id != CODEC_ID_AAC) {
00114             av_log(s, AV_LOG_ERROR, "invalid AAC magic cookie\n");
00115             return AVERROR_INVALIDDATA;
00116         }
00117         url_fskip(pb, skip);
00118     } else {
00119         st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
00120         if (!st->codec->extradata)
00121             return AVERROR(ENOMEM);
00122         get_buffer(pb, st->codec->extradata, size);
00123         st->codec->extradata_size = size;
00124     }
00125 
00126     return 0;
00127 }
00128 
00130 static int read_pakt_chunk(AVFormatContext *s, int64_t size)
00131 {
00132     ByteIOContext *pb = s->pb;
00133     AVStream *st      = s->streams[0];
00134     CaffContext *caf  = s->priv_data;
00135     int64_t pos = 0, ccount;
00136     int num_packets, i;
00137 
00138     ccount = url_ftell(pb);
00139 
00140     num_packets = get_be64(pb);
00141     if (num_packets < 0 || INT32_MAX / sizeof(AVIndexEntry) < num_packets)
00142         return AVERROR_INVALIDDATA;
00143 
00144     st->nb_frames  = get_be64(pb); /* valid frames */
00145     st->nb_frames += get_be32(pb); /* priming frames */
00146     st->nb_frames += get_be32(pb); /* remainder frames */
00147 
00148     st->duration = 0;
00149     for (i = 0; i < num_packets; i++) {
00150         av_add_index_entry(s->streams[0], pos, st->duration, 0, 0, AVINDEX_KEYFRAME);
00151         pos += caf->bytes_per_packet ? caf->bytes_per_packet : ff_mp4_read_descr_len(pb);
00152         st->duration += caf->frames_per_packet ? caf->frames_per_packet : ff_mp4_read_descr_len(pb);
00153     }
00154 
00155     if (url_ftell(pb) - ccount != size) {
00156         av_log(s, AV_LOG_ERROR, "error reading packet table\n");
00157         return -1;
00158     }
00159 
00160     caf->num_bytes = pos;
00161     return 0;
00162 }
00163 
00165 static void read_info_chunk(AVFormatContext *s, int64_t size)
00166 {
00167     ByteIOContext *pb = s->pb;
00168     unsigned int i;
00169     unsigned int nb_entries = get_be32(pb);
00170     for (i = 0; i < nb_entries; i++) {
00171         char key[32];
00172         char value[1024];
00173         get_strz(pb, key, sizeof(key));
00174         get_strz(pb, value, sizeof(value));
00175         av_metadata_set2(&s->metadata, key, value, 0);
00176     }
00177 }
00178 
00179 static int read_header(AVFormatContext *s,
00180                        AVFormatParameters *ap)
00181 {
00182     ByteIOContext *pb = s->pb;
00183     CaffContext *caf  = s->priv_data;
00184     AVStream *st;
00185     uint32_t tag = 0;
00186     int found_data, ret;
00187     int64_t size;
00188 
00189     url_fskip(pb, 8); /* magic, version, file flags */
00190 
00191     /* audio description chunk */
00192     if (get_be32(pb) != MKBETAG('d','e','s','c')) {
00193         av_log(s, AV_LOG_ERROR, "desc chunk not present\n");
00194         return AVERROR_INVALIDDATA;
00195     }
00196     size = get_be64(pb);
00197     if (size != 32)
00198         return AVERROR_INVALIDDATA;
00199 
00200     ret = read_desc_chunk(s);
00201     if (ret)
00202         return ret;
00203     st = s->streams[0];
00204 
00205     /* parse each chunk */
00206     found_data = 0;
00207     while (!url_feof(pb)) {
00208 
00209         /* stop at data chunk if seeking is not supported or
00210            data chunk size is unknown */
00211         if (found_data && (caf->data_size < 0 || url_is_streamed(pb)))
00212             break;
00213 
00214         tag  = get_be32(pb);
00215         size = get_be64(pb);
00216         if (url_feof(pb))
00217             break;
00218 
00219         switch (tag) {
00220         case MKBETAG('d','a','t','a'):
00221             url_fskip(pb, 4); /* edit count */
00222             caf->data_start = url_ftell(pb);
00223             caf->data_size  = size < 0 ? -1 : size - 4;
00224             if (caf->data_size > 0 && !url_is_streamed(pb))
00225                 url_fskip(pb, caf->data_size);
00226             found_data = 1;
00227             break;
00228 
00229         /* magic cookie chunk */
00230         case MKBETAG('k','u','k','i'):
00231             if (read_kuki_chunk(s, size))
00232                 return AVERROR_INVALIDDATA;
00233             break;
00234 
00235         /* packet table chunk */
00236         case MKBETAG('p','a','k','t'):
00237             if (read_pakt_chunk(s, size))
00238                 return AVERROR_INVALIDDATA;
00239             break;
00240 
00241         case MKBETAG('i','n','f','o'):
00242             read_info_chunk(s, size);
00243             break;
00244 
00245         default:
00246 #define _(x) ((x) >= ' ' ? (x) : ' ')
00247             av_log(s, AV_LOG_WARNING, "skipping CAF chunk: %08X (%c%c%c%c)\n",
00248                 tag, _(tag>>24), _((tag>>16)&0xFF), _((tag>>8)&0xFF), _(tag&0xFF));
00249 #undef _
00250         case MKBETAG('f','r','e','e'):
00251             if (size < 0)
00252                 return AVERROR_INVALIDDATA;
00253             url_fskip(pb, size);
00254             break;
00255         }
00256     }
00257 
00258     if (!found_data)
00259         return AVERROR_INVALIDDATA;
00260 
00261     if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
00262         if (caf->data_size > 0)
00263             st->nb_frames = (caf->data_size / caf->bytes_per_packet) * caf->frames_per_packet;
00264     } else if (st->nb_index_entries) {
00265         st->codec->bit_rate = st->codec->sample_rate * caf->data_size * 8 /
00266                               st->duration;
00267     } else {
00268         av_log(s, AV_LOG_ERROR, "Missing packet table. It is required when "
00269                                 "block size or frame size are variable.\n");
00270         return AVERROR_INVALIDDATA;
00271     }
00272     s->file_size = url_fsize(pb);
00273     s->file_size = FFMAX(0, s->file_size);
00274 
00275     av_set_pts_info(st, 64, 1, st->codec->sample_rate);
00276     st->start_time = 0;
00277 
00278     /* position the stream at the start of data */
00279     if (caf->data_size >= 0)
00280         url_fseek(pb, caf->data_start, SEEK_SET);
00281 
00282     return 0;
00283 }
00284 
00285 #define CAF_MAX_PKT_SIZE 4096
00286 
00287 static int read_packet(AVFormatContext *s, AVPacket *pkt)
00288 {
00289     ByteIOContext *pb = s->pb;
00290     AVStream *st      = s->streams[0];
00291     CaffContext *caf  = s->priv_data;
00292     int res, pkt_size = 0, pkt_frames = 0;
00293     int64_t left      = CAF_MAX_PKT_SIZE;
00294 
00295     if (url_feof(pb))
00296         return AVERROR(EIO);
00297 
00298     /* don't read past end of data chunk */
00299     if (caf->data_size > 0) {
00300         left = (caf->data_start + caf->data_size) - url_ftell(pb);
00301         if (left <= 0)
00302             return AVERROR(EIO);
00303     }
00304 
00305     pkt_frames = caf->frames_per_packet;
00306     pkt_size   = caf->bytes_per_packet;
00307 
00308     if (pkt_size > 0 && pkt_frames == 1) {
00309         pkt_size   = (CAF_MAX_PKT_SIZE / pkt_size) * pkt_size;
00310         pkt_size   = FFMIN(pkt_size, left);
00311         pkt_frames = pkt_size / caf->bytes_per_packet;
00312     } else if (st->nb_index_entries) {
00313         if (caf->packet_cnt < st->nb_index_entries - 1) {
00314             pkt_size   = st->index_entries[caf->packet_cnt + 1].pos       - st->index_entries[caf->packet_cnt].pos;
00315             pkt_frames = st->index_entries[caf->packet_cnt + 1].timestamp - st->index_entries[caf->packet_cnt].timestamp;
00316         } else if (caf->packet_cnt == st->nb_index_entries - 1) {
00317             pkt_size   = caf->num_bytes - st->index_entries[caf->packet_cnt].pos;
00318             pkt_frames = st->duration   - st->index_entries[caf->packet_cnt].timestamp;
00319         } else {
00320             return AVERROR(EIO);
00321         }
00322     }
00323 
00324     if (pkt_size == 0 || pkt_frames == 0 || pkt_size > left)
00325         return AVERROR(EIO);
00326 
00327     res = av_get_packet(pb, pkt, pkt_size);
00328     if (res < 0)
00329         return res;
00330 
00331     pkt->size           = res;
00332     pkt->stream_index   = 0;
00333     pkt->dts = pkt->pts = caf->frame_cnt;
00334 
00335     caf->packet_cnt++;
00336     caf->frame_cnt += pkt_frames;
00337 
00338     return 0;
00339 }
00340 
00341 static int read_seek(AVFormatContext *s, int stream_index,
00342                      int64_t timestamp, int flags)
00343 {
00344     AVStream *st = s->streams[0];
00345     CaffContext *caf = s->priv_data;
00346     int64_t pos;
00347 
00348     timestamp = FFMAX(timestamp, 0);
00349 
00350     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
00351         /* calculate new byte position based on target frame position */
00352         pos = caf->bytes_per_packet * timestamp / caf->frames_per_packet;
00353         if (caf->data_size > 0)
00354             pos = FFMIN(pos, caf->data_size);
00355         caf->packet_cnt = pos / caf->bytes_per_packet;
00356         caf->frame_cnt  = caf->frames_per_packet * caf->packet_cnt;
00357     } else if (st->nb_index_entries) {
00358         caf->packet_cnt = av_index_search_timestamp(st, timestamp, flags);
00359         caf->frame_cnt  = st->index_entries[caf->packet_cnt].timestamp;
00360         pos             = st->index_entries[caf->packet_cnt].pos;
00361     } else {
00362         return -1;
00363     }
00364 
00365     url_fseek(s->pb, pos + caf->data_start, SEEK_SET);
00366     return 0;
00367 }
00368 
00369 AVInputFormat caf_demuxer = {
00370     "caf",
00371     NULL_IF_CONFIG_SMALL("Apple Core Audio Format"),
00372     sizeof(CaffContext),
00373     probe,
00374     read_header,
00375     read_packet,
00376     NULL,
00377     read_seek,
00378     .codec_tag = (const AVCodecTag*[]){ff_codec_caf_tags, 0},
00379 };