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 "mov_chan.h"
00033 #include "libavutil/intreadwrite.h"
00034 #include "libavutil/intfloat.h"
00035 #include "libavutil/dict.h"
00036 #include "caf.h"
00037 
00038 typedef struct {
00039     int bytes_per_packet;           
00040     int frames_per_packet;          
00041     int64_t num_bytes;              
00042 
00043     int64_t packet_cnt;             
00044     int64_t frame_cnt;              
00045 
00046     int64_t data_start;             
00047     int64_t data_size;              
00048 } CaffContext;
00049 
00050 static int probe(AVProbeData *p)
00051 {
00052     if (AV_RB32(p->buf) == MKBETAG('c','a','f','f') && AV_RB16(&p->buf[4]) == 1)
00053         return AVPROBE_SCORE_MAX;
00054     return 0;
00055 }
00056 
00058 static int read_desc_chunk(AVFormatContext *s)
00059 {
00060     AVIOContext *pb = s->pb;
00061     CaffContext *caf  = s->priv_data;
00062     AVStream *st;
00063     int flags;
00064 
00065     /* new audio stream */
00066     st = avformat_new_stream(s, NULL);
00067     if (!st)
00068         return AVERROR(ENOMEM);
00069 
00070     /* parse format description */
00071     st->codec->codec_type  = AVMEDIA_TYPE_AUDIO;
00072     st->codec->sample_rate = av_int2double(avio_rb64(pb));
00073     st->codec->codec_tag   = avio_rl32(pb);
00074     flags = avio_rb32(pb);
00075     caf->bytes_per_packet  = avio_rb32(pb);
00076     st->codec->block_align = caf->bytes_per_packet;
00077     caf->frames_per_packet = avio_rb32(pb);
00078     st->codec->channels    = avio_rb32(pb);
00079     st->codec->bits_per_coded_sample = avio_rb32(pb);
00080 
00081     /* calculate bit rate for constant size packets */
00082     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
00083         st->codec->bit_rate = (uint64_t)st->codec->sample_rate * (uint64_t)caf->bytes_per_packet * 8
00084                               / (uint64_t)caf->frames_per_packet;
00085     } else {
00086         st->codec->bit_rate = 0;
00087     }
00088 
00089     /* determine codec */
00090     if (st->codec->codec_tag == MKTAG('l','p','c','m'))
00091         st->codec->codec_id = ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, (flags ^ 0x2) | 0x4);
00092     else
00093         st->codec->codec_id = ff_codec_get_id(ff_codec_caf_tags, st->codec->codec_tag);
00094     return 0;
00095 }
00096 
00098 static int read_kuki_chunk(AVFormatContext *s, int64_t size)
00099 {
00100     AVIOContext *pb = s->pb;
00101     AVStream *st      = s->streams[0];
00102 
00103     if (size < 0 || size > INT_MAX - FF_INPUT_BUFFER_PADDING_SIZE)
00104         return -1;
00105 
00106     if (st->codec->codec_id == AV_CODEC_ID_AAC) {
00107         /* The magic cookie format for AAC is an mp4 esds atom.
00108            The lavc AAC decoder requires the data from the codec specific
00109            description as extradata input. */
00110         int strt, skip;
00111         MOVAtom atom;
00112 
00113         strt = avio_tell(pb);
00114         ff_mov_read_esds(s, pb, atom);
00115         skip = size - (avio_tell(pb) - strt);
00116         if (skip < 0 || !st->codec->extradata ||
00117             st->codec->codec_id != AV_CODEC_ID_AAC) {
00118             av_log(s, AV_LOG_ERROR, "invalid AAC magic cookie\n");
00119             return AVERROR_INVALIDDATA;
00120         }
00121         avio_skip(pb, skip);
00122     } else if (st->codec->codec_id == AV_CODEC_ID_ALAC) {
00123 #define ALAC_PREAMBLE 12
00124 #define ALAC_HEADER   36
00125 #define ALAC_NEW_KUKI 24
00126         uint8_t preamble[12];
00127         if (size < ALAC_NEW_KUKI) {
00128             av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
00129             avio_skip(pb, size);
00130             return AVERROR_INVALIDDATA;
00131         }
00132         avio_read(pb, preamble, ALAC_PREAMBLE);
00133 
00134         st->codec->extradata = av_mallocz(ALAC_HEADER + FF_INPUT_BUFFER_PADDING_SIZE);
00135         if (!st->codec->extradata)
00136             return AVERROR(ENOMEM);
00137 
00138         /* For the old style cookie, we skip 12 bytes, then read 36 bytes.
00139          * The new style cookie only contains the last 24 bytes of what was
00140          * 36 bytes in the old style cookie, so we fabricate the first 12 bytes
00141          * in that case to maintain compatibility. */
00142         if (!memcmp(&preamble[4], "frmaalac", 8)) {
00143             if (size < ALAC_PREAMBLE + ALAC_HEADER) {
00144                 av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
00145                 av_freep(&st->codec->extradata);
00146                 return AVERROR_INVALIDDATA;
00147             }
00148             avio_read(pb, st->codec->extradata, ALAC_HEADER);
00149             avio_skip(pb, size - ALAC_PREAMBLE - ALAC_HEADER);
00150         } else {
00151             AV_WB32(st->codec->extradata, 36);
00152             memcpy(&st->codec->extradata[4], "alac", 4);
00153             AV_WB32(&st->codec->extradata[8], 0);
00154             memcpy(&st->codec->extradata[12], preamble, 12);
00155             avio_read(pb, &st->codec->extradata[24], ALAC_NEW_KUKI - 12);
00156             avio_skip(pb, size - ALAC_NEW_KUKI);
00157         }
00158         st->codec->extradata_size = ALAC_HEADER;
00159     } else {
00160         st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
00161         if (!st->codec->extradata)
00162             return AVERROR(ENOMEM);
00163         avio_read(pb, st->codec->extradata, size);
00164         st->codec->extradata_size = size;
00165     }
00166 
00167     return 0;
00168 }
00169 
00171 static int read_pakt_chunk(AVFormatContext *s, int64_t size)
00172 {
00173     AVIOContext *pb = s->pb;
00174     AVStream *st      = s->streams[0];
00175     CaffContext *caf  = s->priv_data;
00176     int64_t pos = 0, ccount, num_packets;
00177     int i;
00178 
00179     ccount = avio_tell(pb);
00180 
00181     num_packets = avio_rb64(pb);
00182     if (num_packets < 0 || INT32_MAX / sizeof(AVIndexEntry) < num_packets)
00183         return AVERROR_INVALIDDATA;
00184 
00185     st->nb_frames  = avio_rb64(pb); /* valid frames */
00186     st->nb_frames += avio_rb32(pb); /* priming frames */
00187     st->nb_frames += avio_rb32(pb); /* remainder frames */
00188 
00189     st->duration = 0;
00190     for (i = 0; i < num_packets; i++) {
00191         av_add_index_entry(s->streams[0], pos, st->duration, 0, 0, AVINDEX_KEYFRAME);
00192         pos += caf->bytes_per_packet ? caf->bytes_per_packet : ff_mp4_read_descr_len(pb);
00193         st->duration += caf->frames_per_packet ? caf->frames_per_packet : ff_mp4_read_descr_len(pb);
00194     }
00195 
00196     if (avio_tell(pb) - ccount > size) {
00197         av_log(s, AV_LOG_ERROR, "error reading packet table\n");
00198         return AVERROR_INVALIDDATA;
00199     }
00200     avio_skip(pb, ccount + size - avio_tell(pb));
00201 
00202     caf->num_bytes = pos;
00203     return 0;
00204 }
00205 
00207 static void read_info_chunk(AVFormatContext *s, int64_t size)
00208 {
00209     AVIOContext *pb = s->pb;
00210     unsigned int i;
00211     unsigned int nb_entries = avio_rb32(pb);
00212     for (i = 0; i < nb_entries; i++) {
00213         char key[32];
00214         char value[1024];
00215         avio_get_str(pb, INT_MAX, key, sizeof(key));
00216         avio_get_str(pb, INT_MAX, value, sizeof(value));
00217         av_dict_set(&s->metadata, key, value, 0);
00218     }
00219 }
00220 
00221 static int read_header(AVFormatContext *s)
00222 {
00223     AVIOContext *pb = s->pb;
00224     CaffContext *caf  = s->priv_data;
00225     AVStream *st;
00226     uint32_t tag = 0;
00227     int found_data, ret;
00228     int64_t size;
00229 
00230     avio_skip(pb, 8); /* magic, version, file flags */
00231 
00232     /* audio description chunk */
00233     if (avio_rb32(pb) != MKBETAG('d','e','s','c')) {
00234         av_log(s, AV_LOG_ERROR, "desc chunk not present\n");
00235         return AVERROR_INVALIDDATA;
00236     }
00237     size = avio_rb64(pb);
00238     if (size != 32)
00239         return AVERROR_INVALIDDATA;
00240 
00241     ret = read_desc_chunk(s);
00242     if (ret)
00243         return ret;
00244     st = s->streams[0];
00245 
00246     /* parse each chunk */
00247     found_data = 0;
00248     while (!url_feof(pb)) {
00249 
00250         /* stop at data chunk if seeking is not supported or
00251            data chunk size is unknown */
00252         if (found_data && (caf->data_size < 0 || !pb->seekable))
00253             break;
00254 
00255         tag  = avio_rb32(pb);
00256         size = avio_rb64(pb);
00257         if (url_feof(pb))
00258             break;
00259 
00260         switch (tag) {
00261         case MKBETAG('d','a','t','a'):
00262             avio_skip(pb, 4); /* edit count */
00263             caf->data_start = avio_tell(pb);
00264             caf->data_size  = size < 0 ? -1 : size - 4;
00265             if (caf->data_size > 0 && pb->seekable)
00266                 avio_skip(pb, caf->data_size);
00267             found_data = 1;
00268             break;
00269 
00270         case MKBETAG('c','h','a','n'):
00271             if ((ret = ff_mov_read_chan(s, s->pb, st, size)) < 0)
00272                 return ret;
00273             break;
00274 
00275         /* magic cookie chunk */
00276         case MKBETAG('k','u','k','i'):
00277             if (read_kuki_chunk(s, size))
00278                 return AVERROR_INVALIDDATA;
00279             break;
00280 
00281         /* packet table chunk */
00282         case MKBETAG('p','a','k','t'):
00283             if (read_pakt_chunk(s, size))
00284                 return AVERROR_INVALIDDATA;
00285             break;
00286 
00287         case MKBETAG('i','n','f','o'):
00288             read_info_chunk(s, size);
00289             break;
00290 
00291         default:
00292 #define _(x) ((x) >= ' ' ? (x) : ' ')
00293             av_log(s, AV_LOG_WARNING, "skipping CAF chunk: %08X (%c%c%c%c), size %"PRId64"\n",
00294                 tag, _(tag>>24), _((tag>>16)&0xFF), _((tag>>8)&0xFF), _(tag&0xFF), size);
00295 #undef _
00296         case MKBETAG('f','r','e','e'):
00297             if (size < 0)
00298                 return AVERROR_INVALIDDATA;
00299             avio_skip(pb, size);
00300             break;
00301         }
00302     }
00303 
00304     if (!found_data)
00305         return AVERROR_INVALIDDATA;
00306 
00307     if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
00308         if (caf->data_size > 0)
00309             st->nb_frames = (caf->data_size / caf->bytes_per_packet) * caf->frames_per_packet;
00310     } else if (st->nb_index_entries) {
00311         st->codec->bit_rate = st->codec->sample_rate * caf->data_size * 8 /
00312                               st->duration;
00313     } else {
00314         av_log(s, AV_LOG_ERROR, "Missing packet table. It is required when "
00315                                 "block size or frame size are variable.\n");
00316         return AVERROR_INVALIDDATA;
00317     }
00318 
00319     avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
00320     st->start_time = 0;
00321 
00322     /* position the stream at the start of data */
00323     if (caf->data_size >= 0)
00324         avio_seek(pb, caf->data_start, SEEK_SET);
00325 
00326     return 0;
00327 }
00328 
00329 #define CAF_MAX_PKT_SIZE 4096
00330 
00331 static int read_packet(AVFormatContext *s, AVPacket *pkt)
00332 {
00333     AVIOContext *pb = s->pb;
00334     AVStream *st      = s->streams[0];
00335     CaffContext *caf  = s->priv_data;
00336     int res, pkt_size = 0, pkt_frames = 0;
00337     int64_t left      = CAF_MAX_PKT_SIZE;
00338 
00339     if (url_feof(pb))
00340         return AVERROR(EIO);
00341 
00342     /* don't read past end of data chunk */
00343     if (caf->data_size > 0) {
00344         left = (caf->data_start + caf->data_size) - avio_tell(pb);
00345         if (left <= 0)
00346             return AVERROR(EIO);
00347     }
00348 
00349     pkt_frames = caf->frames_per_packet;
00350     pkt_size   = caf->bytes_per_packet;
00351 
00352     if (pkt_size > 0 && pkt_frames == 1) {
00353         pkt_size   = (CAF_MAX_PKT_SIZE / pkt_size) * pkt_size;
00354         pkt_size   = FFMIN(pkt_size, left);
00355         pkt_frames = pkt_size / caf->bytes_per_packet;
00356     } else if (st->nb_index_entries) {
00357         if (caf->packet_cnt < st->nb_index_entries - 1) {
00358             pkt_size   = st->index_entries[caf->packet_cnt + 1].pos       - st->index_entries[caf->packet_cnt].pos;
00359             pkt_frames = st->index_entries[caf->packet_cnt + 1].timestamp - st->index_entries[caf->packet_cnt].timestamp;
00360         } else if (caf->packet_cnt == st->nb_index_entries - 1) {
00361             pkt_size   = caf->num_bytes - st->index_entries[caf->packet_cnt].pos;
00362             pkt_frames = st->duration   - st->index_entries[caf->packet_cnt].timestamp;
00363         } else {
00364             return AVERROR(EIO);
00365         }
00366     }
00367 
00368     if (pkt_size == 0 || pkt_frames == 0 || pkt_size > left)
00369         return AVERROR(EIO);
00370 
00371     res = av_get_packet(pb, pkt, pkt_size);
00372     if (res < 0)
00373         return res;
00374 
00375     pkt->size           = res;
00376     pkt->stream_index   = 0;
00377     pkt->dts = pkt->pts = caf->frame_cnt;
00378 
00379     caf->packet_cnt++;
00380     caf->frame_cnt += pkt_frames;
00381 
00382     return 0;
00383 }
00384 
00385 static int read_seek(AVFormatContext *s, int stream_index,
00386                      int64_t timestamp, int flags)
00387 {
00388     AVStream *st = s->streams[0];
00389     CaffContext *caf = s->priv_data;
00390     int64_t pos, packet_cnt, frame_cnt;
00391 
00392     timestamp = FFMAX(timestamp, 0);
00393 
00394     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
00395         /* calculate new byte position based on target frame position */
00396         pos = caf->bytes_per_packet * timestamp / caf->frames_per_packet;
00397         if (caf->data_size > 0)
00398             pos = FFMIN(pos, caf->data_size);
00399         packet_cnt = pos / caf->bytes_per_packet;
00400         frame_cnt  = caf->frames_per_packet * packet_cnt;
00401     } else if (st->nb_index_entries) {
00402         packet_cnt = av_index_search_timestamp(st, timestamp, flags);
00403         frame_cnt  = st->index_entries[packet_cnt].timestamp;
00404         pos        = st->index_entries[packet_cnt].pos;
00405     } else {
00406         return -1;
00407     }
00408 
00409     if (avio_seek(s->pb, pos + caf->data_start, SEEK_SET) < 0)
00410         return -1;
00411 
00412     caf->packet_cnt = packet_cnt;
00413     caf->frame_cnt  = frame_cnt;
00414 
00415     return 0;
00416 }
00417 
00418 AVInputFormat ff_caf_demuxer = {
00419     .name           = "caf",
00420     .long_name      = NULL_IF_CONFIG_SMALL("Apple CAF (Core Audio Format)"),
00421     .priv_data_size = sizeof(CaffContext),
00422     .read_probe     = probe,
00423     .read_header    = read_header,
00424     .read_packet    = read_packet,
00425     .read_seek      = read_seek,
00426     .codec_tag      = (const AVCodecTag* const []){ ff_codec_caf_tags, 0 },
00427 };