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 "isom.h"
00031 #include "mov_chan.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 == AV_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 != AV_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 == AV_CODEC_ID_ALAC) {
00122 #define ALAC_PREAMBLE 12
00123 #define ALAC_HEADER   36
00124 #define ALAC_NEW_KUKI 24
00125         uint8_t preamble[12];
00126         if (size < ALAC_NEW_KUKI) {
00127             av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
00128             avio_skip(pb, size);
00129             return AVERROR_INVALIDDATA;
00130         }
00131         avio_read(pb, preamble, ALAC_PREAMBLE);
00132 
00133         st->codec->extradata = av_mallocz(ALAC_HEADER + FF_INPUT_BUFFER_PADDING_SIZE);
00134         if (!st->codec->extradata)
00135             return AVERROR(ENOMEM);
00136 
00137         /* For the old style cookie, we skip 12 bytes, then read 36 bytes.
00138          * The new style cookie only contains the last 24 bytes of what was
00139          * 36 bytes in the old style cookie, so we fabricate the first 12 bytes
00140          * in that case to maintain compatibility. */
00141         if (!memcmp(&preamble[4], "frmaalac", 8)) {
00142             if (size < ALAC_PREAMBLE + ALAC_HEADER) {
00143                 av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
00144                 av_freep(&st->codec->extradata);
00145                 return AVERROR_INVALIDDATA;
00146             }
00147             avio_read(pb, st->codec->extradata, ALAC_HEADER);
00148             avio_skip(pb, size - ALAC_PREAMBLE - ALAC_HEADER);
00149         } else {
00150             AV_WB32(st->codec->extradata, 36);
00151             memcpy(&st->codec->extradata[4], "alac", 4);
00152             AV_WB32(&st->codec->extradata[8], 0);
00153             memcpy(&st->codec->extradata[12], preamble, 12);
00154             avio_read(pb, &st->codec->extradata[24], ALAC_NEW_KUKI - 12);
00155             avio_skip(pb, size - ALAC_NEW_KUKI);
00156         }
00157         st->codec->extradata_size = ALAC_HEADER;
00158     } else {
00159         st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
00160         if (!st->codec->extradata)
00161             return AVERROR(ENOMEM);
00162         avio_read(pb, st->codec->extradata, size);
00163         st->codec->extradata_size = size;
00164     }
00165 
00166     return 0;
00167 }
00168 
00170 static int read_pakt_chunk(AVFormatContext *s, int64_t size)
00171 {
00172     AVIOContext *pb = s->pb;
00173     AVStream *st      = s->streams[0];
00174     CaffContext *caf  = s->priv_data;
00175     int64_t pos = 0, ccount, num_packets;
00176     int i;
00177 
00178     ccount = avio_tell(pb);
00179 
00180     num_packets = avio_rb64(pb);
00181     if (num_packets < 0 || INT32_MAX / sizeof(AVIndexEntry) < num_packets)
00182         return AVERROR_INVALIDDATA;
00183 
00184     st->nb_frames  = avio_rb64(pb); /* valid frames */
00185     st->nb_frames += avio_rb32(pb); /* priming frames */
00186     st->nb_frames += avio_rb32(pb); /* remainder frames */
00187 
00188     st->duration = 0;
00189     for (i = 0; i < num_packets; i++) {
00190         av_add_index_entry(s->streams[0], pos, st->duration, 0, 0, AVINDEX_KEYFRAME);
00191         pos += caf->bytes_per_packet ? caf->bytes_per_packet : ff_mp4_read_descr_len(pb);
00192         st->duration += caf->frames_per_packet ? caf->frames_per_packet : ff_mp4_read_descr_len(pb);
00193     }
00194 
00195     if (avio_tell(pb) - ccount > size) {
00196         av_log(s, AV_LOG_ERROR, "error reading packet table\n");
00197         return AVERROR_INVALIDDATA;
00198     }
00199     avio_skip(pb, ccount + size - avio_tell(pb));
00200 
00201     caf->num_bytes = pos;
00202     return 0;
00203 }
00204 
00206 static void read_info_chunk(AVFormatContext *s, int64_t size)
00207 {
00208     AVIOContext *pb = s->pb;
00209     unsigned int i;
00210     unsigned int nb_entries = avio_rb32(pb);
00211     for (i = 0; i < nb_entries; i++) {
00212         char key[32];
00213         char value[1024];
00214         avio_get_str(pb, INT_MAX, key, sizeof(key));
00215         avio_get_str(pb, INT_MAX, value, sizeof(value));
00216         av_dict_set(&s->metadata, key, value, 0);
00217     }
00218 }
00219 
00220 static int read_header(AVFormatContext *s)
00221 {
00222     AVIOContext *pb = s->pb;
00223     CaffContext *caf  = s->priv_data;
00224     AVStream *st;
00225     uint32_t tag = 0;
00226     int found_data, ret;
00227     int64_t size, pos;
00228 
00229     avio_skip(pb, 8); /* magic, version, file flags */
00230 
00231     /* audio description chunk */
00232     if (avio_rb32(pb) != MKBETAG('d','e','s','c')) {
00233         av_log(s, AV_LOG_ERROR, "desc chunk not present\n");
00234         return AVERROR_INVALIDDATA;
00235     }
00236     size = avio_rb64(pb);
00237     if (size != 32)
00238         return AVERROR_INVALIDDATA;
00239 
00240     ret = read_desc_chunk(s);
00241     if (ret)
00242         return ret;
00243     st = s->streams[0];
00244 
00245     /* parse each chunk */
00246     found_data = 0;
00247     while (!url_feof(pb)) {
00248 
00249         /* stop at data chunk if seeking is not supported or
00250            data chunk size is unknown */
00251         if (found_data && (caf->data_size < 0 || !pb->seekable))
00252             break;
00253 
00254         tag  = avio_rb32(pb);
00255         size = avio_rb64(pb);
00256         pos  = avio_tell(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             break;
00300         }
00301 
00302         if (size > 0) {
00303             if (pos + size < pos)
00304                 return AVERROR_INVALIDDATA;
00305             avio_skip(pb, FFMAX(0, pos + size - avio_tell(pb)));
00306         }
00307     }
00308 
00309     if (!found_data)
00310         return AVERROR_INVALIDDATA;
00311 
00312     if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
00313         if (caf->data_size > 0)
00314             st->nb_frames = (caf->data_size / caf->bytes_per_packet) * caf->frames_per_packet;
00315     } else if (st->nb_index_entries && st->duration > 0) {
00316         st->codec->bit_rate = st->codec->sample_rate * caf->data_size * 8 /
00317                               st->duration;
00318     } else {
00319         av_log(s, AV_LOG_ERROR, "Missing packet table. It is required when "
00320                                 "block size or frame size are variable.\n");
00321         return AVERROR_INVALIDDATA;
00322     }
00323 
00324     avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
00325     st->start_time = 0;
00326 
00327     /* position the stream at the start of data */
00328     if (caf->data_size >= 0)
00329         avio_seek(pb, caf->data_start, SEEK_SET);
00330 
00331     return 0;
00332 }
00333 
00334 #define CAF_MAX_PKT_SIZE 4096
00335 
00336 static int read_packet(AVFormatContext *s, AVPacket *pkt)
00337 {
00338     AVIOContext *pb = s->pb;
00339     AVStream *st      = s->streams[0];
00340     CaffContext *caf  = s->priv_data;
00341     int res, pkt_size = 0, pkt_frames = 0;
00342     int64_t left      = CAF_MAX_PKT_SIZE;
00343 
00344     if (url_feof(pb))
00345         return AVERROR_EOF;
00346 
00347     /* don't read past end of data chunk */
00348     if (caf->data_size > 0) {
00349         left = (caf->data_start + caf->data_size) - avio_tell(pb);
00350         if (!left)
00351             return AVERROR_EOF;
00352         if (left < 0)
00353             return AVERROR(EIO);
00354     }
00355 
00356     pkt_frames = caf->frames_per_packet;
00357     pkt_size   = caf->bytes_per_packet;
00358 
00359     if (pkt_size > 0 && pkt_frames == 1) {
00360         pkt_size   = (CAF_MAX_PKT_SIZE / pkt_size) * pkt_size;
00361         pkt_size   = FFMIN(pkt_size, left);
00362         pkt_frames = pkt_size / caf->bytes_per_packet;
00363     } else if (st->nb_index_entries) {
00364         if (caf->packet_cnt < st->nb_index_entries - 1) {
00365             pkt_size   = st->index_entries[caf->packet_cnt + 1].pos       - st->index_entries[caf->packet_cnt].pos;
00366             pkt_frames = st->index_entries[caf->packet_cnt + 1].timestamp - st->index_entries[caf->packet_cnt].timestamp;
00367         } else if (caf->packet_cnt == st->nb_index_entries - 1) {
00368             pkt_size   = caf->num_bytes - st->index_entries[caf->packet_cnt].pos;
00369             pkt_frames = st->duration   - st->index_entries[caf->packet_cnt].timestamp;
00370         } else {
00371             return AVERROR(EIO);
00372         }
00373     }
00374 
00375     if (pkt_size == 0 || pkt_frames == 0 || pkt_size > left)
00376         return AVERROR(EIO);
00377 
00378     res = av_get_packet(pb, pkt, pkt_size);
00379     if (res < 0)
00380         return res;
00381 
00382     pkt->size           = res;
00383     pkt->stream_index   = 0;
00384     pkt->dts = pkt->pts = caf->frame_cnt;
00385 
00386     caf->packet_cnt++;
00387     caf->frame_cnt += pkt_frames;
00388 
00389     return 0;
00390 }
00391 
00392 static int read_seek(AVFormatContext *s, int stream_index,
00393                      int64_t timestamp, int flags)
00394 {
00395     AVStream *st = s->streams[0];
00396     CaffContext *caf = s->priv_data;
00397     int64_t pos, packet_cnt, frame_cnt;
00398 
00399     timestamp = FFMAX(timestamp, 0);
00400 
00401     if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
00402         /* calculate new byte position based on target frame position */
00403         pos = caf->bytes_per_packet * (timestamp / caf->frames_per_packet);
00404         if (caf->data_size > 0)
00405             pos = FFMIN(pos, caf->data_size);
00406         packet_cnt = pos / caf->bytes_per_packet;
00407         frame_cnt  = caf->frames_per_packet * packet_cnt;
00408     } else if (st->nb_index_entries) {
00409         packet_cnt = av_index_search_timestamp(st, timestamp, flags);
00410         frame_cnt  = st->index_entries[packet_cnt].timestamp;
00411         pos        = st->index_entries[packet_cnt].pos;
00412     } else {
00413         return -1;
00414     }
00415 
00416     if (avio_seek(s->pb, pos + caf->data_start, SEEK_SET) < 0)
00417         return -1;
00418 
00419     caf->packet_cnt = packet_cnt;
00420     caf->frame_cnt  = frame_cnt;
00421 
00422     return 0;
00423 }
00424 
00425 AVInputFormat ff_caf_demuxer = {
00426     .name           = "caf",
00427     .long_name      = NULL_IF_CONFIG_SMALL("Apple CAF (Core Audio Format)"),
00428     .priv_data_size = sizeof(CaffContext),
00429     .read_probe     = probe,
00430     .read_header    = read_header,
00431     .read_packet    = read_packet,
00432     .read_seek      = read_seek,
00433     .codec_tag      = (const AVCodecTag* const []){ ff_codec_caf_tags, 0 },
00434 };