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 */
59 static int mp3_read_probe(AVProbeData *p)
61 int max_frames, first_frames = 0;
64 const uint8_t *buf, *buf0, *buf2, *end;
65 AVCodecContext *avctx = avcodec_alloc_context3(NULL);
68 return AVERROR(ENOMEM);
71 end = p->buf + p->buf_size - sizeof(uint32_t);
72 while(buf0 < end && !*buf0)
78 for(; buf < end; buf= buf2+1) {
80 if(ff_mpa_check_header(AV_RB32(buf2)))
83 for(frames = 0; buf2 < end; frames++) {
85 header = AV_RB32(buf2);
86 fsize = avpriv_mpa_decode_header(avctx, header,
87 &dummy, &dummy, &dummy, &dummy);
92 max_frames = FFMAX(max_frames, frames);
96 avcodec_free_context(&avctx);
97 // keep this in sync with ac3 probe, both need to avoid
98 // issues with MPEG-files!
99 if (first_frames>=4) return AVPROBE_SCORE_EXTENSION + 1;
100 else if(max_frames>200)return AVPROBE_SCORE_EXTENSION;
101 else if(max_frames>=4 && max_frames >= p->buf_size/10000) return AVPROBE_SCORE_EXTENSION / 2;
102 else if(ff_id3v2_match(buf0, ID3v2_DEFAULT_MAGIC) && 2*ff_id3v2_tag_len(buf0) >= p->buf_size)
103 return p->buf_size < PROBE_BUF_MAX ? AVPROBE_SCORE_EXTENSION / 4 : AVPROBE_SCORE_EXTENSION - 2;
104 else if(max_frames>=1 && max_frames >= p->buf_size/10000) return 1;
106 //mpegps_mp3_unrecognized_format.mpg has max_frames=3
109 static void read_xing_toc(AVFormatContext *s, int64_t filesize, int64_t duration)
112 MP3DecContext *mp3 = s->priv_data;
113 int fill_index = mp3->usetoc == 1 && duration > 0;
116 !(filesize = avio_size(s->pb))) {
117 av_log(s, AV_LOG_WARNING, "Cannot determine file size, skipping TOC table.\n");
121 for (i = 0; i < XING_TOC_COUNT; i++) {
122 uint8_t b = avio_r8(s->pb);
124 av_add_index_entry(s->streams[0],
125 av_rescale(b, filesize, 256),
126 av_rescale(i, duration, XING_TOC_COUNT),
127 0, 0, AVINDEX_KEYFRAME);
133 static void mp3_parse_info_tag(AVFormatContext *s, AVStream *st,
134 MPADecodeHeader *c, uint32_t spf)
136 #define LAST_BITS(k, n) ((k) & ((1 << (n)) - 1))
137 #define MIDDLE_BITS(k, m, n) LAST_BITS((k) >> (m), ((n) - (m)))
145 int32_t r_gain = INT32_MIN, a_gain = INT32_MIN;
147 MP3DecContext *mp3 = s->priv_data;
148 static const int64_t xing_offtbl[2][2] = {{32, 17}, {17,9}};
149 uint64_t fsize = avio_size(s->pb);
150 fsize = fsize >= avio_tell(s->pb) ? fsize - avio_tell(s->pb) : 0;
152 /* Check for Xing / Info tag */
153 avio_skip(s->pb, xing_offtbl[c->lsf == 1][c->nb_channels == 1]);
154 v = avio_rb32(s->pb);
155 mp3->is_cbr = v == MKBETAG('I', 'n', 'f', 'o');
156 if (v != MKBETAG('X', 'i', 'n', 'g') && !mp3->is_cbr)
159 v = avio_rb32(s->pb);
160 if (v & XING_FLAG_FRAMES)
161 mp3->frames = avio_rb32(s->pb);
162 if (v & XING_FLAG_SIZE)
163 mp3->header_filesize = avio_rb32(s->pb);
164 if (fsize && mp3->header_filesize) {
166 min = FFMIN(fsize, mp3->header_filesize);
167 delta = FFMAX(fsize, mp3->header_filesize) - min;
168 if (fsize > mp3->header_filesize && delta > min >> 4) {
170 av_log(s, AV_LOG_WARNING,
171 "invalid concatenated file detected - using bitrate for duration\n");
172 } else if (delta > min >> 4) {
173 av_log(s, AV_LOG_WARNING,
174 "filesize and duration do not match (growing file?)\n");
177 if (v & XING_FLAG_TOC)
178 read_xing_toc(s, mp3->header_filesize, av_rescale_q(mp3->frames,
179 (AVRational){spf, c->sample_rate},
182 if (v & XING_FLAC_QSCALE)
185 /* Encoder short version string */
186 memset(version, 0, sizeof(version));
187 avio_read(s->pb, version, 9);
189 /* Info Tag revision + VBR method */
192 /* Lowpass filter value */
195 /* ReplayGain peak */
196 v = avio_rb32(s->pb);
197 peak = av_rescale(v, 100000, 1 << 23);
199 /* Radio ReplayGain */
200 v = avio_rb16(s->pb);
202 if (MIDDLE_BITS(v, 13, 15) == 1) {
203 r_gain = MIDDLE_BITS(v, 0, 8) * 10000;
209 /* Audiophile ReplayGain */
210 v = avio_rb16(s->pb);
212 if (MIDDLE_BITS(v, 13, 15) == 2) {
213 a_gain = MIDDLE_BITS(v, 0, 8) * 10000;
219 /* Encoding flags + ATH Type */
222 /* if ABR {specified bitrate} else {minimal bitrate} */
227 if(AV_RB32(version) == MKBETAG('L', 'A', 'M', 'E')
228 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'f')
229 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'c')
232 mp3->start_pad = v>>12;
233 mp3-> end_pad = v&4095;
234 st->start_skip_samples = mp3->start_pad + 528 + 1;
236 st->first_discard_sample = -mp3->end_pad + 528 + 1 + mp3->frames * (int64_t)spf;
237 st->last_discard_sample = mp3->frames * (int64_t)spf;
240 st->start_time = av_rescale_q(st->start_skip_samples,
241 (AVRational){1, c->sample_rate},
243 av_log(s, AV_LOG_DEBUG, "pad %d %d\n", mp3->start_pad, mp3-> end_pad);
252 /* Preset and surround info */
262 crc = ffio_get_checksum(s->pb);
263 v = avio_rb16(s->pb);
266 ff_replaygain_export_raw(st, r_gain, peak, a_gain, 0);
267 av_dict_set(&st->metadata, "encoder", version, 0);
271 static void mp3_parse_vbri_tag(AVFormatContext *s, AVStream *st, int64_t base)
274 MP3DecContext *mp3 = s->priv_data;
276 /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */
277 avio_seek(s->pb, base + 4 + 32, SEEK_SET);
278 v = avio_rb32(s->pb);
279 if (v == MKBETAG('V', 'B', 'R', 'I')) {
280 /* Check tag version */
281 if (avio_rb16(s->pb) == 1) {
282 /* skip delay and quality */
284 mp3->header_filesize = avio_rb32(s->pb);
285 mp3->frames = avio_rb32(s->pb);
291 * Try to find Xing/Info/VBRI tags and compute duration from info therein
293 static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
298 MP3DecContext *mp3 = s->priv_data;
300 ffio_init_checksum(s->pb, ff_crcA001_update, 0);
302 v = avio_rb32(s->pb);
303 if(ff_mpa_check_header(v) < 0)
306 if (avpriv_mpegaudio_decode_header(&c, v) == 0)
307 vbrtag_size = c.frame_size;
311 spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
314 mp3->header_filesize = 0;
316 mp3_parse_info_tag(s, st, &c, spf);
317 mp3_parse_vbri_tag(s, st, base);
319 if (!mp3->frames && !mp3->header_filesize)
322 /* Skip the vbr tag frame */
323 avio_seek(s->pb, base + vbrtag_size, SEEK_SET);
326 st->duration = av_rescale_q(mp3->frames, (AVRational){spf, c.sample_rate},
328 if (mp3->header_filesize && mp3->frames && !mp3->is_cbr)
329 st->codec->bit_rate = av_rescale(mp3->header_filesize, 8 * c.sample_rate, mp3->frames * (int64_t)spf);
334 static int mp3_read_header(AVFormatContext *s)
336 MP3DecContext *mp3 = s->priv_data;
343 mp3->usetoc = (s->flags & AVFMT_FLAG_FAST_SEEK) ? 0 : 2;
345 st = avformat_new_stream(s, NULL);
347 return AVERROR(ENOMEM);
349 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
350 st->codec->codec_id = AV_CODEC_ID_MP3;
351 st->need_parsing = AVSTREAM_PARSE_FULL_RAW;
354 // lcm of all mp3 sample rates
355 avpriv_set_pts_info(st, 64, 1, 14112000);
358 off = avio_tell(s->pb);
360 if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX))
364 mp3->filesize = avio_size(s->pb);
366 if (mp3_parse_vbr_tags(s, st, off) < 0)
367 avio_seek(s->pb, off, SEEK_SET);
369 ret = ff_replaygain_export(st, s->metadata);
373 // the seek index is relative to the end of the xing vbr headers
374 for (i = 0; i < st->nb_index_entries; i++)
375 st->index_entries[i].pos += avio_tell(s->pb);
377 /* the parameters will be extracted from the compressed bitstream */
381 #define MP3_PACKET_SIZE 1024
383 static int mp3_read_packet(AVFormatContext *s, AVPacket *pkt)
385 MP3DecContext *mp3 = s->priv_data;
389 size= MP3_PACKET_SIZE;
390 pos = avio_tell(s->pb);
391 if(mp3->filesize > ID3v1_TAG_SIZE && pos < mp3->filesize)
392 size= FFMIN(size, mp3->filesize - pos);
394 ret= av_get_packet(s->pb, pkt, size);
401 pkt->flags &= ~AV_PKT_FLAG_CORRUPT;
402 pkt->stream_index = 0;
404 if (ret >= ID3v1_TAG_SIZE &&
405 memcmp(&pkt->data[ret - ID3v1_TAG_SIZE], "TAG", 3) == 0)
406 ret -= ID3v1_TAG_SIZE;
408 /* note: we need to modify the packet size here to handle the last
414 static int check(AVFormatContext *s, int64_t pos)
416 int64_t ret = avio_seek(s->pb, pos, SEEK_SET);
421 header = avio_rb32(s->pb);
422 if (ff_mpa_check_header(header) < 0)
424 if (avpriv_mpegaudio_decode_header(&sd, header) == 1)
426 return sd.frame_size;
429 static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp,
432 MP3DecContext *mp3 = s->priv_data;
433 AVIndexEntry *ie, ie1;
434 AVStream *st = s->streams[0];
435 int64_t ret = av_index_search_timestamp(st, timestamp, flags);
437 int dir = (flags&AVSEEK_FLAG_BACKWARD) ? -1 : 1;
441 if (mp3->usetoc == 2)
442 return -1; // generic index code
445 && (mp3->usetoc == 0 || !mp3->xing_toc)
447 && mp3->header_filesize > s->internal->data_offset
450 timestamp = av_clip64(timestamp, 0, st->duration);
451 ie->timestamp = timestamp;
452 ie->pos = av_rescale(timestamp, mp3->header_filesize, st->duration) + s->internal->data_offset;
453 } else if (mp3->xing_toc) {
457 ie = &st->index_entries[ret];
462 avio_seek(s->pb, FFMAX(ie->pos - 4096, 0), SEEK_SET);
463 ret = avio_seek(s->pb, ie->pos, SEEK_SET);
470 for(i=0; i<4096; i++) {
471 int64_t pos = ie->pos + (dir > 0 ? i - 1024 : -i);
472 int64_t candidate = -1;
478 for(j=0; j<MIN_VALID; j++) {
482 if ((ie->pos - pos)*dir <= 0 && abs(MIN_VALID/2-j) < score) {
484 score = abs(MIN_VALID/2-j);
488 if (best_score > score && j == MIN_VALID) {
489 best_pos = candidate;
496 ret = avio_seek(s->pb, best_pos, SEEK_SET);
500 if (mp3->is_cbr && ie == &ie1) {
501 int frame_duration = av_rescale(st->duration, 1, mp3->frames);
502 ie1.timestamp = frame_duration * av_rescale(best_pos - s->internal->data_offset, mp3->frames, mp3->header_filesize);
505 ff_update_cur_dts(s, st, ie->timestamp);
509 static const AVOption options[] = {
510 { "usetoc", "use table of contents", offsetof(MP3DecContext, usetoc), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2, AV_OPT_FLAG_DECODING_PARAM},
514 static const AVClass demuxer_class = {
516 .item_name = av_default_item_name,
518 .version = LIBAVUTIL_VERSION_INT,
519 .category = AV_CLASS_CATEGORY_DEMUXER,
522 AVInputFormat ff_mp3_demuxer = {
524 .long_name = NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"),
525 .read_probe = mp3_read_probe,
526 .read_header = mp3_read_header,
527 .read_packet = mp3_read_packet,
528 .read_seek = mp3_seek,
529 .priv_data_size = sizeof(MP3DecContext),
530 .flags = AVFMT_GENERIC_INDEX,
531 .extensions = "mp2,mp3,m2a,mpa", /* XXX: use probe */
532 .priv_class = &demuxer_class,