3 * Copyright (c) 2003 Fabrice Bellard
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 #include "libavutil/opt.h"
23 #include "libavutil/avstring.h"
24 #include "libavutil/intreadwrite.h"
25 #include "libavutil/crc.h"
26 #include "libavutil/dict.h"
27 #include "libavutil/mathematics.h"
30 #include "avio_internal.h"
33 #include "replaygain.h"
35 #include "libavcodec/avcodec.h"
36 #include "libavcodec/mpegaudiodecheader.h"
38 #define XING_FLAG_FRAMES 0x01
39 #define XING_FLAG_SIZE 0x02
40 #define XING_FLAG_TOC 0x04
41 #define XING_FLAC_QSCALE 0x08
43 #define XING_TOC_COUNT 100
52 unsigned frames; /* Total number of frames in file */
53 unsigned header_filesize; /* Total number of bytes in the stream */
57 static int check(AVFormatContext *s, int64_t pos);
61 static int mp3_read_probe(AVProbeData *p)
63 int max_frames, first_frames = 0;
66 const uint8_t *buf, *buf0, *buf2, *end;
67 AVCodecContext *avctx = avcodec_alloc_context3(NULL);
70 return AVERROR(ENOMEM);
73 end = p->buf + p->buf_size - sizeof(uint32_t);
74 while(buf0 < end && !*buf0)
80 for(; buf < end; buf= buf2+1) {
82 if(ff_mpa_check_header(AV_RB32(buf2)))
85 for(frames = 0; buf2 < end; frames++) {
87 header = AV_RB32(buf2);
88 fsize = avpriv_mpa_decode_header(avctx, header,
89 &dummy, &dummy, &dummy, &dummy);
94 max_frames = FFMAX(max_frames, frames);
98 avcodec_free_context(&avctx);
99 // keep this in sync with ac3 probe, both need to avoid
100 // issues with MPEG-files!
101 if (first_frames>=4) return AVPROBE_SCORE_EXTENSION + 1;
102 else if(max_frames>200)return AVPROBE_SCORE_EXTENSION;
103 else if(max_frames>=4 && max_frames >= p->buf_size/10000) return AVPROBE_SCORE_EXTENSION / 2;
104 else if(ff_id3v2_match(buf0, ID3v2_DEFAULT_MAGIC) && 2*ff_id3v2_tag_len(buf0) >= p->buf_size)
105 return p->buf_size < PROBE_BUF_MAX ? AVPROBE_SCORE_EXTENSION / 4 : AVPROBE_SCORE_EXTENSION - 2;
106 else if(max_frames>=1 && max_frames >= p->buf_size/10000) return 1;
108 //mpegps_mp3_unrecognized_format.mpg has max_frames=3
111 static void read_xing_toc(AVFormatContext *s, int64_t filesize, int64_t duration)
114 MP3DecContext *mp3 = s->priv_data;
115 int fill_index = mp3->usetoc == 1 && duration > 0;
118 !(filesize = avio_size(s->pb))) {
119 av_log(s, AV_LOG_WARNING, "Cannot determine file size, skipping TOC table.\n");
123 for (i = 0; i < XING_TOC_COUNT; i++) {
124 uint8_t b = avio_r8(s->pb);
126 av_add_index_entry(s->streams[0],
127 av_rescale(b, filesize, 256),
128 av_rescale(i, duration, XING_TOC_COUNT),
129 0, 0, AVINDEX_KEYFRAME);
135 static void mp3_parse_info_tag(AVFormatContext *s, AVStream *st,
136 MPADecodeHeader *c, uint32_t spf)
138 #define LAST_BITS(k, n) ((k) & ((1 << (n)) - 1))
139 #define MIDDLE_BITS(k, m, n) LAST_BITS((k) >> (m), ((n) - (m)))
147 int32_t r_gain = INT32_MIN, a_gain = INT32_MIN;
149 MP3DecContext *mp3 = s->priv_data;
150 static const int64_t xing_offtbl[2][2] = {{32, 17}, {17,9}};
151 uint64_t fsize = avio_size(s->pb);
152 fsize = fsize >= avio_tell(s->pb) ? fsize - avio_tell(s->pb) : 0;
154 /* Check for Xing / Info tag */
155 avio_skip(s->pb, xing_offtbl[c->lsf == 1][c->nb_channels == 1]);
156 v = avio_rb32(s->pb);
157 mp3->is_cbr = v == MKBETAG('I', 'n', 'f', 'o');
158 if (v != MKBETAG('X', 'i', 'n', 'g') && !mp3->is_cbr)
161 v = avio_rb32(s->pb);
162 if (v & XING_FLAG_FRAMES)
163 mp3->frames = avio_rb32(s->pb);
164 if (v & XING_FLAG_SIZE)
165 mp3->header_filesize = avio_rb32(s->pb);
166 if (fsize && mp3->header_filesize) {
168 min = FFMIN(fsize, mp3->header_filesize);
169 delta = FFMAX(fsize, mp3->header_filesize) - min;
170 if (fsize > mp3->header_filesize && delta > min >> 4) {
172 av_log(s, AV_LOG_WARNING,
173 "invalid concatenated file detected - using bitrate for duration\n");
174 } else if (delta > min >> 4) {
175 av_log(s, AV_LOG_WARNING,
176 "filesize and duration do not match (growing file?)\n");
179 if (v & XING_FLAG_TOC)
180 read_xing_toc(s, mp3->header_filesize, av_rescale_q(mp3->frames,
181 (AVRational){spf, c->sample_rate},
184 if (v & XING_FLAC_QSCALE)
187 /* Encoder short version string */
188 memset(version, 0, sizeof(version));
189 avio_read(s->pb, version, 9);
191 /* Info Tag revision + VBR method */
194 /* Lowpass filter value */
197 /* ReplayGain peak */
198 v = avio_rb32(s->pb);
199 peak = av_rescale(v, 100000, 1 << 23);
201 /* Radio ReplayGain */
202 v = avio_rb16(s->pb);
204 if (MIDDLE_BITS(v, 13, 15) == 1) {
205 r_gain = MIDDLE_BITS(v, 0, 8) * 10000;
211 /* Audiophile ReplayGain */
212 v = avio_rb16(s->pb);
214 if (MIDDLE_BITS(v, 13, 15) == 2) {
215 a_gain = MIDDLE_BITS(v, 0, 8) * 10000;
221 /* Encoding flags + ATH Type */
224 /* if ABR {specified bitrate} else {minimal bitrate} */
229 if(AV_RB32(version) == MKBETAG('L', 'A', 'M', 'E')
230 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'f')
231 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'c')
234 mp3->start_pad = v>>12;
235 mp3-> end_pad = v&4095;
236 st->start_skip_samples = mp3->start_pad + 528 + 1;
238 st->first_discard_sample = -mp3->end_pad + 528 + 1 + mp3->frames * (int64_t)spf;
239 st->last_discard_sample = mp3->frames * (int64_t)spf;
242 st->start_time = av_rescale_q(st->start_skip_samples,
243 (AVRational){1, c->sample_rate},
245 av_log(s, AV_LOG_DEBUG, "pad %d %d\n", mp3->start_pad, mp3-> end_pad);
254 /* Preset and surround info */
264 crc = ffio_get_checksum(s->pb);
265 v = avio_rb16(s->pb);
268 ff_replaygain_export_raw(st, r_gain, peak, a_gain, 0);
269 av_dict_set(&st->metadata, "encoder", version, 0);
273 static void mp3_parse_vbri_tag(AVFormatContext *s, AVStream *st, int64_t base)
276 MP3DecContext *mp3 = s->priv_data;
278 /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */
279 avio_seek(s->pb, base + 4 + 32, SEEK_SET);
280 v = avio_rb32(s->pb);
281 if (v == MKBETAG('V', 'B', 'R', 'I')) {
282 /* Check tag version */
283 if (avio_rb16(s->pb) == 1) {
284 /* skip delay and quality */
286 mp3->header_filesize = avio_rb32(s->pb);
287 mp3->frames = avio_rb32(s->pb);
293 * Try to find Xing/Info/VBRI tags and compute duration from info therein
295 static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
300 MP3DecContext *mp3 = s->priv_data;
302 ffio_init_checksum(s->pb, ff_crcA001_update, 0);
304 v = avio_rb32(s->pb);
305 if(ff_mpa_check_header(v) < 0)
308 if (avpriv_mpegaudio_decode_header(&c, v) == 0)
309 vbrtag_size = c.frame_size;
313 spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
316 mp3->header_filesize = 0;
318 mp3_parse_info_tag(s, st, &c, spf);
319 mp3_parse_vbri_tag(s, st, base);
321 if (!mp3->frames && !mp3->header_filesize)
324 /* Skip the vbr tag frame */
325 avio_seek(s->pb, base + vbrtag_size, SEEK_SET);
328 st->duration = av_rescale_q(mp3->frames, (AVRational){spf, c.sample_rate},
330 if (mp3->header_filesize && mp3->frames && !mp3->is_cbr)
331 st->codec->bit_rate = av_rescale(mp3->header_filesize, 8 * c.sample_rate, mp3->frames * (int64_t)spf);
336 static int mp3_read_header(AVFormatContext *s)
338 MP3DecContext *mp3 = s->priv_data;
345 mp3->usetoc = (s->flags & AVFMT_FLAG_FAST_SEEK) ? 0 : 2;
347 st = avformat_new_stream(s, NULL);
349 return AVERROR(ENOMEM);
351 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
352 st->codec->codec_id = AV_CODEC_ID_MP3;
353 st->need_parsing = AVSTREAM_PARSE_FULL_RAW;
356 // lcm of all mp3 sample rates
357 avpriv_set_pts_info(st, 64, 1, 14112000);
360 off = avio_tell(s->pb);
362 if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX))
366 mp3->filesize = avio_size(s->pb);
368 if (mp3_parse_vbr_tags(s, st, off) < 0)
369 avio_seek(s->pb, off, SEEK_SET);
371 ret = ff_replaygain_export(st, s->metadata);
375 off = avio_tell(s->pb);
376 for (i = 0; i < 64 * 1024; i++) {
378 ffio_ensure_seekback(s->pb, i + 1024 + 4);
379 if (check(s, off + i) >= 0) {
380 av_log(s, AV_LOG_INFO, "Skipping %d bytes of junk at %"PRId64".\n", i, off);
381 avio_seek(s->pb, off + i, SEEK_SET);
384 avio_seek(s->pb, off, SEEK_SET);
387 // the seek index is relative to the end of the xing vbr headers
388 for (i = 0; i < st->nb_index_entries; i++)
389 st->index_entries[i].pos += avio_tell(s->pb);
391 /* the parameters will be extracted from the compressed bitstream */
395 #define MP3_PACKET_SIZE 1024
397 static int mp3_read_packet(AVFormatContext *s, AVPacket *pkt)
399 MP3DecContext *mp3 = s->priv_data;
403 size= MP3_PACKET_SIZE;
404 pos = avio_tell(s->pb);
405 if(mp3->filesize > ID3v1_TAG_SIZE && pos < mp3->filesize)
406 size= FFMIN(size, mp3->filesize - pos);
408 ret= av_get_packet(s->pb, pkt, size);
415 pkt->flags &= ~AV_PKT_FLAG_CORRUPT;
416 pkt->stream_index = 0;
418 if (ret >= ID3v1_TAG_SIZE &&
419 memcmp(&pkt->data[ret - ID3v1_TAG_SIZE], "TAG", 3) == 0)
420 ret -= ID3v1_TAG_SIZE;
422 /* note: we need to modify the packet size here to handle the last
428 static int check(AVFormatContext *s, int64_t pos)
430 int64_t ret = avio_seek(s->pb, pos, SEEK_SET);
435 header = avio_rb32(s->pb);
436 if (ff_mpa_check_header(header) < 0)
438 if (avpriv_mpegaudio_decode_header(&sd, header) == 1)
440 return sd.frame_size;
443 static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp,
446 MP3DecContext *mp3 = s->priv_data;
447 AVIndexEntry *ie, ie1;
448 AVStream *st = s->streams[0];
449 int64_t ret = av_index_search_timestamp(st, timestamp, flags);
451 int dir = (flags&AVSEEK_FLAG_BACKWARD) ? -1 : 1;
455 if (mp3->usetoc == 2)
456 return -1; // generic index code
459 && (mp3->usetoc == 0 || !mp3->xing_toc)
461 && mp3->header_filesize > s->internal->data_offset
464 timestamp = av_clip64(timestamp, 0, st->duration);
465 ie->timestamp = timestamp;
466 ie->pos = av_rescale(timestamp, mp3->header_filesize, st->duration) + s->internal->data_offset;
467 } else if (mp3->xing_toc) {
471 ie = &st->index_entries[ret];
476 avio_seek(s->pb, FFMAX(ie->pos - 4096, 0), SEEK_SET);
477 ret = avio_seek(s->pb, ie->pos, SEEK_SET);
484 for(i=0; i<4096; i++) {
485 int64_t pos = ie->pos + (dir > 0 ? i - 1024 : -i);
486 int64_t candidate = -1;
492 for(j=0; j<MIN_VALID; j++) {
496 if ((ie->pos - pos)*dir <= 0 && abs(MIN_VALID/2-j) < score) {
498 score = abs(MIN_VALID/2-j);
502 if (best_score > score && j == MIN_VALID) {
503 best_pos = candidate;
510 ret = avio_seek(s->pb, best_pos, SEEK_SET);
514 if (mp3->is_cbr && ie == &ie1) {
515 int frame_duration = av_rescale(st->duration, 1, mp3->frames);
516 ie1.timestamp = frame_duration * av_rescale(best_pos - s->internal->data_offset, mp3->frames, mp3->header_filesize);
519 ff_update_cur_dts(s, st, ie->timestamp);
523 static const AVOption options[] = {
524 { "usetoc", "use table of contents", offsetof(MP3DecContext, usetoc), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2, AV_OPT_FLAG_DECODING_PARAM},
528 static const AVClass demuxer_class = {
530 .item_name = av_default_item_name,
532 .version = LIBAVUTIL_VERSION_INT,
533 .category = AV_CLASS_CATEGORY_DEMUXER,
536 AVInputFormat ff_mp3_demuxer = {
538 .long_name = NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"),
539 .read_probe = mp3_read_probe,
540 .read_header = mp3_read_header,
541 .read_packet = mp3_read_packet,
542 .read_seek = mp3_seek,
543 .priv_data_size = sizeof(MP3DecContext),
544 .flags = AVFMT_GENERIC_INDEX,
545 .extensions = "mp2,mp3,m2a,mpa", /* XXX: use probe */
546 .priv_class = &demuxer_class,