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 "libavutil/dict.h"
00033 #include "caf.h"
00034 
00035 typedef struct {
00036     int bytes_per_packet;           
00037     int frames_per_packet;          
00038     int64_t num_bytes;              
00039 
00040     int64_t packet_cnt;             
00041     int64_t frame_cnt;              
00042 
00043     int64_t data_start;             
00044     int64_t data_size;              
00045 } CaffContext;
00046 
00047 static int probe(AVProbeData *p)
00048 {
00049     if (AV_RB32(p->buf) == MKBETAG('c','a','f','f') && AV_RB16(&p->buf[4]) == 1)
00050         return AVPROBE_SCORE_MAX;
00051     return 0;
00052 }
00053 
00055 static int read_desc_chunk(AVFormatContext *s)
00056 {
00057     AVIOContext *pb = s->pb;
00058     CaffContext *caf  = s->priv_data;
00059     AVStream *st;
00060     int flags;
00061 
00062     /* new audio stream */
00063     st = av_new_stream(s, 0);
00064     if (!st)
00065         return AVERROR(ENOMEM);
00066 
00067     /* parse format description */
00068     st->codec->codec_type  = AVMEDIA_TYPE_AUDIO;
00069     st->codec->sample_rate = av_int2dbl(avio_rb64(pb));
00070     st->codec->codec_tag   = avio_rb32(pb);
00071     flags = avio_rb32(pb);
00072     caf->bytes_per_packet  = avio_rb32(pb);
00073     st->codec->block_align = caf->bytes_per_packet;
00074     caf->frames_per_packet = avio_rb32(pb);
00075     st->codec->channels    = avio_rb32(pb);
00076     st->codec->bits_per_coded_sample = avio_rb32(pb);
00077 
00078     /* calculate bit rate for constant size packets */
00079     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
00080         st->codec->bit_rate = (uint64_t)st->codec->sample_rate * (uint64_t)caf->bytes_per_packet * 8
00081                               / (uint64_t)caf->frames_per_packet;
00082     } else {
00083         st->codec->bit_rate = 0;
00084     }
00085 
00086     /* determine codec */
00087     if (st->codec->codec_tag == MKBETAG('l','p','c','m'))
00088         st->codec->codec_id = ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, (flags ^ 0x2) | 0x4);
00089     else
00090         st->codec->codec_id = ff_codec_get_id(ff_codec_caf_tags, st->codec->codec_tag);
00091     return 0;
00092 }
00093 
00095 static int read_kuki_chunk(AVFormatContext *s, int64_t size)
00096 {
00097     AVIOContext *pb = s->pb;
00098     AVStream *st      = s->streams[0];
00099 
00100     if (size < 0 || size > INT_MAX - FF_INPUT_BUFFER_PADDING_SIZE)
00101         return -1;
00102 
00103     if (st->codec->codec_id == CODEC_ID_AAC) {
00104         /* The magic cookie format for AAC is an mp4 esds atom.
00105            The lavc AAC decoder requires the data from the codec specific
00106            description as extradata input. */
00107         int strt, skip;
00108         MOVAtom atom;
00109 
00110         strt = avio_tell(pb);
00111         ff_mov_read_esds(s, pb, atom);
00112         skip = size - (avio_tell(pb) - strt);
00113         if (skip < 0 || !st->codec->extradata ||
00114             st->codec->codec_id != CODEC_ID_AAC) {
00115             av_log(s, AV_LOG_ERROR, "invalid AAC magic cookie\n");
00116             return AVERROR_INVALIDDATA;
00117         }
00118         avio_skip(pb, skip);
00119     } else if (st->codec->codec_id == CODEC_ID_ALAC) {
00120 #define ALAC_PREAMBLE 12
00121 #define ALAC_HEADER   36
00122         if (size < ALAC_PREAMBLE + ALAC_HEADER) {
00123             av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
00124             avio_skip(pb, size);
00125             return AVERROR_INVALIDDATA;
00126         }
00127         avio_skip(pb, ALAC_PREAMBLE);
00128         st->codec->extradata = av_mallocz(ALAC_HEADER + FF_INPUT_BUFFER_PADDING_SIZE);
00129         if (!st->codec->extradata)
00130             return AVERROR(ENOMEM);
00131         avio_read(pb, st->codec->extradata, ALAC_HEADER);
00132         st->codec->extradata_size = ALAC_HEADER;
00133         avio_skip(pb, size - ALAC_PREAMBLE - ALAC_HEADER);
00134     } else {
00135         st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
00136         if (!st->codec->extradata)
00137             return AVERROR(ENOMEM);
00138         avio_read(pb, st->codec->extradata, size);
00139         st->codec->extradata_size = size;
00140     }
00141 
00142     return 0;
00143 }
00144 
00146 static int read_pakt_chunk(AVFormatContext *s, int64_t size)
00147 {
00148     AVIOContext *pb = s->pb;
00149     AVStream *st      = s->streams[0];
00150     CaffContext *caf  = s->priv_data;
00151     int64_t pos = 0, ccount;
00152     int num_packets, i;
00153 
00154     ccount = avio_tell(pb);
00155 
00156     num_packets = avio_rb64(pb);
00157     if (num_packets < 0 || INT32_MAX / sizeof(AVIndexEntry) < num_packets)
00158         return AVERROR_INVALIDDATA;
00159 
00160     st->nb_frames  = avio_rb64(pb); /* valid frames */
00161     st->nb_frames += avio_rb32(pb); /* priming frames */
00162     st->nb_frames += avio_rb32(pb); /* remainder frames */
00163 
00164     st->duration = 0;
00165     for (i = 0; i < num_packets; i++) {
00166         av_add_index_entry(s->streams[0], pos, st->duration, 0, 0, AVINDEX_KEYFRAME);
00167         pos += caf->bytes_per_packet ? caf->bytes_per_packet : ff_mp4_read_descr_len(pb);
00168         st->duration += caf->frames_per_packet ? caf->frames_per_packet : ff_mp4_read_descr_len(pb);
00169     }
00170 
00171     if (avio_tell(pb) - ccount != size) {
00172         av_log(s, AV_LOG_ERROR, "error reading packet table\n");
00173         return -1;
00174     }
00175 
00176     caf->num_bytes = pos;
00177     return 0;
00178 }
00179 
00181 static void read_info_chunk(AVFormatContext *s, int64_t size)
00182 {
00183     AVIOContext *pb = s->pb;
00184     unsigned int i;
00185     unsigned int nb_entries = avio_rb32(pb);
00186     for (i = 0; i < nb_entries; i++) {
00187         char key[32];
00188         char value[1024];
00189         get_strz(pb, key, sizeof(key));
00190         get_strz(pb, value, sizeof(value));
00191         av_dict_set(&s->metadata, key, value, 0);
00192     }
00193 }
00194 
00195 static int read_header(AVFormatContext *s,
00196                        AVFormatParameters *ap)
00197 {
00198     AVIOContext *pb = s->pb;
00199     CaffContext *caf  = s->priv_data;
00200     AVStream *st;
00201     uint32_t tag = 0;
00202     int found_data, ret;
00203     int64_t size;
00204 
00205     avio_skip(pb, 8); /* magic, version, file flags */
00206 
00207     /* audio description chunk */
00208     if (avio_rb32(pb) != MKBETAG('d','e','s','c')) {
00209         av_log(s, AV_LOG_ERROR, "desc chunk not present\n");
00210         return AVERROR_INVALIDDATA;
00211     }
00212     size = avio_rb64(pb);
00213     if (size != 32)
00214         return AVERROR_INVALIDDATA;
00215 
00216     ret = read_desc_chunk(s);
00217     if (ret)
00218         return ret;
00219     st = s->streams[0];
00220 
00221     /* parse each chunk */
00222     found_data = 0;
00223     while (!url_feof(pb)) {
00224 
00225         /* stop at data chunk if seeking is not supported or
00226            data chunk size is unknown */
00227         if (found_data && (caf->data_size < 0 || !pb->seekable))
00228             break;
00229 
00230         tag  = avio_rb32(pb);
00231         size = avio_rb64(pb);
00232         if (url_feof(pb))
00233             break;
00234 
00235         switch (tag) {
00236         case MKBETAG('d','a','t','a'):
00237             avio_skip(pb, 4); /* edit count */
00238             caf->data_start = avio_tell(pb);
00239             caf->data_size  = size < 0 ? -1 : size - 4;
00240             if (caf->data_size > 0 && pb->seekable)
00241                 avio_skip(pb, caf->data_size);
00242             found_data = 1;
00243             break;
00244 
00245         /* magic cookie chunk */
00246         case MKBETAG('k','u','k','i'):
00247             if (read_kuki_chunk(s, size))
00248                 return AVERROR_INVALIDDATA;
00249             break;
00250 
00251         /* packet table chunk */
00252         case MKBETAG('p','a','k','t'):
00253             if (read_pakt_chunk(s, size))
00254                 return AVERROR_INVALIDDATA;
00255             break;
00256 
00257         case MKBETAG('i','n','f','o'):
00258             read_info_chunk(s, size);
00259             break;
00260 
00261         case MKBETAG('c','h','a','n'):
00262             if (size < 12)
00263                 return AVERROR_INVALIDDATA;
00264             ff_mov_read_chan(s, size, st->codec);
00265             break;
00266 
00267         default:
00268 #define _(x) ((x) >= ' ' ? (x) : ' ')
00269             av_log(s, AV_LOG_WARNING, "skipping CAF chunk: %08X (%c%c%c%c), size %"PRId64"\n",
00270                 tag, _(tag>>24), _((tag>>16)&0xFF), _((tag>>8)&0xFF), _(tag&0xFF), size);
00271 #undef _
00272         case MKBETAG('f','r','e','e'):
00273             if (size < 0)
00274                 return AVERROR_INVALIDDATA;
00275             avio_skip(pb, size);
00276             break;
00277         }
00278     }
00279 
00280     if (!found_data)
00281         return AVERROR_INVALIDDATA;
00282 
00283     if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
00284         if (caf->data_size > 0)
00285             st->nb_frames = (caf->data_size / caf->bytes_per_packet) * caf->frames_per_packet;
00286     } else if (st->nb_index_entries) {
00287         st->codec->bit_rate = st->codec->sample_rate * caf->data_size * 8 /
00288                               st->duration;
00289     } else {
00290         av_log(s, AV_LOG_ERROR, "Missing packet table. It is required when "
00291                                 "block size or frame size are variable.\n");
00292         return AVERROR_INVALIDDATA;
00293     }
00294     s->file_size = avio_size(pb);
00295     s->file_size = FFMAX(0, s->file_size);
00296 
00297     av_set_pts_info(st, 64, 1, st->codec->sample_rate);
00298     st->start_time = 0;
00299 
00300     /* position the stream at the start of data */
00301     if (caf->data_size >= 0)
00302         avio_seek(pb, caf->data_start, SEEK_SET);
00303 
00304     return 0;
00305 }
00306 
00307 #define CAF_MAX_PKT_SIZE 4096
00308 
00309 static int read_packet(AVFormatContext *s, AVPacket *pkt)
00310 {
00311     AVIOContext *pb = s->pb;
00312     AVStream *st      = s->streams[0];
00313     CaffContext *caf  = s->priv_data;
00314     int res, pkt_size = 0, pkt_frames = 0;
00315     int64_t left      = CAF_MAX_PKT_SIZE;
00316 
00317     if (url_feof(pb))
00318         return AVERROR(EIO);
00319 
00320     /* don't read past end of data chunk */
00321     if (caf->data_size > 0) {
00322         left = (caf->data_start + caf->data_size) - avio_tell(pb);
00323         if (left <= 0)
00324             return AVERROR(EIO);
00325     }
00326 
00327     pkt_frames = caf->frames_per_packet;
00328     pkt_size   = caf->bytes_per_packet;
00329 
00330     if (pkt_size > 0 && pkt_frames == 1) {
00331         pkt_size   = (CAF_MAX_PKT_SIZE / pkt_size) * pkt_size;
00332         pkt_size   = FFMIN(pkt_size, left);
00333         pkt_frames = pkt_size / caf->bytes_per_packet;
00334     } else if (st->nb_index_entries) {
00335         if (caf->packet_cnt < st->nb_index_entries - 1) {
00336             pkt_size   = st->index_entries[caf->packet_cnt + 1].pos       - st->index_entries[caf->packet_cnt].pos;
00337             pkt_frames = st->index_entries[caf->packet_cnt + 1].timestamp - st->index_entries[caf->packet_cnt].timestamp;
00338         } else if (caf->packet_cnt == st->nb_index_entries - 1) {
00339             pkt_size   = caf->num_bytes - st->index_entries[caf->packet_cnt].pos;
00340             pkt_frames = st->duration   - st->index_entries[caf->packet_cnt].timestamp;
00341         } else {
00342             return AVERROR(EIO);
00343         }
00344     }
00345 
00346     if (pkt_size == 0 || pkt_frames == 0 || pkt_size > left)
00347         return AVERROR(EIO);
00348 
00349     res = av_get_packet(pb, pkt, pkt_size);
00350     if (res < 0)
00351         return res;
00352 
00353     pkt->size           = res;
00354     pkt->stream_index   = 0;
00355     pkt->dts = pkt->pts = caf->frame_cnt;
00356 
00357     caf->packet_cnt++;
00358     caf->frame_cnt += pkt_frames;
00359 
00360     return 0;
00361 }
00362 
00363 static int read_seek(AVFormatContext *s, int stream_index,
00364                      int64_t timestamp, int flags)
00365 {
00366     AVStream *st = s->streams[0];
00367     CaffContext *caf = s->priv_data;
00368     int64_t pos;
00369 
00370     timestamp = FFMAX(timestamp, 0);
00371 
00372     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
00373         /* calculate new byte position based on target frame position */
00374         pos = caf->bytes_per_packet * timestamp / caf->frames_per_packet;
00375         if (caf->data_size > 0)
00376             pos = FFMIN(pos, caf->data_size);
00377         caf->packet_cnt = pos / caf->bytes_per_packet;
00378         caf->frame_cnt  = caf->frames_per_packet * caf->packet_cnt;
00379     } else if (st->nb_index_entries) {
00380         caf->packet_cnt = av_index_search_timestamp(st, timestamp, flags);
00381         caf->frame_cnt  = st->index_entries[caf->packet_cnt].timestamp;
00382         pos             = st->index_entries[caf->packet_cnt].pos;
00383     } else {
00384         return -1;
00385     }
00386 
00387     avio_seek(s->pb, pos + caf->data_start, SEEK_SET);
00388     return 0;
00389 }
00390 
00391 AVInputFormat ff_caf_demuxer = {
00392     "caf",
00393     NULL_IF_CONFIG_SMALL("Apple Core Audio Format"),
00394     sizeof(CaffContext),
00395     probe,
00396     read_header,
00397     read_packet,
00398     NULL,
00399     read_seek,
00400     .codec_tag = (const AVCodecTag*[]){ff_codec_caf_tags, 0},
00401 };