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
45 #define SAME_HEADER_MASK \
46 (0xffe00000 | (3 << 17) | (3 << 10) | (3 << 19))
55 unsigned frames; /* Total number of frames in file */
56 unsigned header_filesize; /* Total number of bytes in the stream */
60 static int check(AVIOContext *pb, int64_t pos, uint32_t *header);
64 static int mp3_read_probe(AVProbeData *p)
66 int max_frames, first_frames = 0;
69 const uint8_t *buf, *buf0, *buf2, *end;
70 AVCodecContext *avctx = avcodec_alloc_context3(NULL);
73 return AVERROR(ENOMEM);
76 end = p->buf + p->buf_size - sizeof(uint32_t);
77 while(buf0 < end && !*buf0)
83 for(; buf < end; buf= buf2+1) {
85 if(ff_mpa_check_header(AV_RB32(buf2)))
88 for(frames = 0; buf2 < end; frames++) {
90 header = AV_RB32(buf2);
91 fsize = avpriv_mpa_decode_header(avctx, header,
92 &dummy, &dummy, &dummy, &dummy);
97 max_frames = FFMAX(max_frames, frames);
101 avcodec_free_context(&avctx);
102 // keep this in sync with ac3 probe, both need to avoid
103 // issues with MPEG-files!
104 if (first_frames>=7) return AVPROBE_SCORE_EXTENSION + 1;
105 else if(max_frames>200)return AVPROBE_SCORE_EXTENSION;
106 else if(max_frames>=4 && max_frames >= p->buf_size/10000) return AVPROBE_SCORE_EXTENSION / 2;
107 else if(ff_id3v2_match(buf0, ID3v2_DEFAULT_MAGIC) && 2*ff_id3v2_tag_len(buf0) >= p->buf_size)
108 return p->buf_size < PROBE_BUF_MAX ? AVPROBE_SCORE_EXTENSION / 4 : AVPROBE_SCORE_EXTENSION - 2;
109 else if(max_frames>=1 && max_frames >= p->buf_size/10000) return 1;
111 //mpegps_mp3_unrecognized_format.mpg has max_frames=3
114 static void read_xing_toc(AVFormatContext *s, int64_t filesize, int64_t duration)
117 MP3DecContext *mp3 = s->priv_data;
118 int fill_index = mp3->usetoc == 1 && duration > 0;
121 !(filesize = avio_size(s->pb))) {
122 av_log(s, AV_LOG_WARNING, "Cannot determine file size, skipping TOC table.\n");
126 for (i = 0; i < XING_TOC_COUNT; i++) {
127 uint8_t b = avio_r8(s->pb);
129 av_add_index_entry(s->streams[0],
130 av_rescale(b, filesize, 256),
131 av_rescale(i, duration, XING_TOC_COUNT),
132 0, 0, AVINDEX_KEYFRAME);
138 static void mp3_parse_info_tag(AVFormatContext *s, AVStream *st,
139 MPADecodeHeader *c, uint32_t spf)
141 #define LAST_BITS(k, n) ((k) & ((1 << (n)) - 1))
142 #define MIDDLE_BITS(k, m, n) LAST_BITS((k) >> (m), ((n) - (m)))
150 int32_t r_gain = INT32_MIN, a_gain = INT32_MIN;
152 MP3DecContext *mp3 = s->priv_data;
153 static const int64_t xing_offtbl[2][2] = {{32, 17}, {17,9}};
154 uint64_t fsize = avio_size(s->pb);
155 fsize = fsize >= avio_tell(s->pb) ? fsize - avio_tell(s->pb) : 0;
157 /* Check for Xing / Info tag */
158 avio_skip(s->pb, xing_offtbl[c->lsf == 1][c->nb_channels == 1]);
159 v = avio_rb32(s->pb);
160 mp3->is_cbr = v == MKBETAG('I', 'n', 'f', 'o');
161 if (v != MKBETAG('X', 'i', 'n', 'g') && !mp3->is_cbr)
164 v = avio_rb32(s->pb);
165 if (v & XING_FLAG_FRAMES)
166 mp3->frames = avio_rb32(s->pb);
167 if (v & XING_FLAG_SIZE)
168 mp3->header_filesize = avio_rb32(s->pb);
169 if (fsize && mp3->header_filesize) {
171 min = FFMIN(fsize, mp3->header_filesize);
172 delta = FFMAX(fsize, mp3->header_filesize) - min;
173 if (fsize > mp3->header_filesize && delta > min >> 4) {
175 av_log(s, AV_LOG_WARNING,
176 "invalid concatenated file detected - using bitrate for duration\n");
177 } else if (delta > min >> 4) {
178 av_log(s, AV_LOG_WARNING,
179 "filesize and duration do not match (growing file?)\n");
182 if (v & XING_FLAG_TOC)
183 read_xing_toc(s, mp3->header_filesize, av_rescale_q(mp3->frames,
184 (AVRational){spf, c->sample_rate},
187 if (v & XING_FLAC_QSCALE)
190 /* Encoder short version string */
191 memset(version, 0, sizeof(version));
192 avio_read(s->pb, version, 9);
194 /* Info Tag revision + VBR method */
197 /* Lowpass filter value */
200 /* ReplayGain peak */
201 v = avio_rb32(s->pb);
202 peak = av_rescale(v, 100000, 1 << 23);
204 /* Radio ReplayGain */
205 v = avio_rb16(s->pb);
207 if (MIDDLE_BITS(v, 13, 15) == 1) {
208 r_gain = MIDDLE_BITS(v, 0, 8) * 10000;
214 /* Audiophile ReplayGain */
215 v = avio_rb16(s->pb);
217 if (MIDDLE_BITS(v, 13, 15) == 2) {
218 a_gain = MIDDLE_BITS(v, 0, 8) * 10000;
224 /* Encoding flags + ATH Type */
227 /* if ABR {specified bitrate} else {minimal bitrate} */
232 if(AV_RB32(version) == MKBETAG('L', 'A', 'M', 'E')
233 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'f')
234 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'c')
237 mp3->start_pad = v>>12;
238 mp3-> end_pad = v&4095;
239 st->start_skip_samples = mp3->start_pad + 528 + 1;
241 st->first_discard_sample = -mp3->end_pad + 528 + 1 + mp3->frames * (int64_t)spf;
242 st->last_discard_sample = mp3->frames * (int64_t)spf;
245 st->start_time = av_rescale_q(st->start_skip_samples,
246 (AVRational){1, c->sample_rate},
248 av_log(s, AV_LOG_DEBUG, "pad %d %d\n", mp3->start_pad, mp3-> end_pad);
257 /* Preset and surround info */
267 crc = ffio_get_checksum(s->pb);
268 v = avio_rb16(s->pb);
271 ff_replaygain_export_raw(st, r_gain, peak, a_gain, 0);
272 av_dict_set(&st->metadata, "encoder", version, 0);
276 static void mp3_parse_vbri_tag(AVFormatContext *s, AVStream *st, int64_t base)
279 MP3DecContext *mp3 = s->priv_data;
281 /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */
282 avio_seek(s->pb, base + 4 + 32, SEEK_SET);
283 v = avio_rb32(s->pb);
284 if (v == MKBETAG('V', 'B', 'R', 'I')) {
285 /* Check tag version */
286 if (avio_rb16(s->pb) == 1) {
287 /* skip delay and quality */
289 mp3->header_filesize = avio_rb32(s->pb);
290 mp3->frames = avio_rb32(s->pb);
296 * Try to find Xing/Info/VBRI tags and compute duration from info therein
298 static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
303 MP3DecContext *mp3 = s->priv_data;
305 ffio_init_checksum(s->pb, ff_crcA001_update, 0);
307 v = avio_rb32(s->pb);
308 if(ff_mpa_check_header(v) < 0)
311 if (avpriv_mpegaudio_decode_header(&c, v) == 0)
312 vbrtag_size = c.frame_size;
316 spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
319 mp3->header_filesize = 0;
321 mp3_parse_info_tag(s, st, &c, spf);
322 mp3_parse_vbri_tag(s, st, base);
324 if (!mp3->frames && !mp3->header_filesize)
327 /* Skip the vbr tag frame */
328 avio_seek(s->pb, base + vbrtag_size, SEEK_SET);
331 st->duration = av_rescale_q(mp3->frames, (AVRational){spf, c.sample_rate},
333 if (mp3->header_filesize && mp3->frames && !mp3->is_cbr)
334 st->codec->bit_rate = av_rescale(mp3->header_filesize, 8 * c.sample_rate, mp3->frames * (int64_t)spf);
339 static int mp3_read_header(AVFormatContext *s)
341 MP3DecContext *mp3 = s->priv_data;
348 mp3->usetoc = (s->flags & AVFMT_FLAG_FAST_SEEK) ? 1 : 2;
350 st = avformat_new_stream(s, NULL);
352 return AVERROR(ENOMEM);
354 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
355 st->codec->codec_id = AV_CODEC_ID_MP3;
356 st->need_parsing = AVSTREAM_PARSE_FULL_RAW;
359 // lcm of all mp3 sample rates
360 avpriv_set_pts_info(st, 64, 1, 14112000);
363 off = avio_tell(s->pb);
365 if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX))
369 mp3->filesize = avio_size(s->pb);
371 if (mp3_parse_vbr_tags(s, st, off) < 0)
372 avio_seek(s->pb, off, SEEK_SET);
374 ret = ff_replaygain_export(st, s->metadata);
378 off = avio_tell(s->pb);
379 for (i = 0; i < 64 * 1024; i++) {
380 uint32_t header, header2;
383 ffio_ensure_seekback(s->pb, i + 1024 + 4);
384 frame_size = check(s->pb, off + i, &header);
385 if (frame_size > 0) {
386 avio_seek(s->pb, off, SEEK_SET);
387 ffio_ensure_seekback(s->pb, i + 1024 + frame_size + 4);
388 if (check(s->pb, off + i + frame_size, &header2) >= 0 &&
389 (header & SAME_HEADER_MASK) == (header2 & SAME_HEADER_MASK))
391 av_log(s, AV_LOG_INFO, "Skipping %d bytes of junk at %"PRId64".\n", i, off);
392 avio_seek(s->pb, off + i, SEEK_SET);
396 avio_seek(s->pb, off, SEEK_SET);
399 // the seek index is relative to the end of the xing vbr headers
400 for (i = 0; i < st->nb_index_entries; i++)
401 st->index_entries[i].pos += avio_tell(s->pb);
403 /* the parameters will be extracted from the compressed bitstream */
407 #define MP3_PACKET_SIZE 1024
409 static int mp3_read_packet(AVFormatContext *s, AVPacket *pkt)
411 MP3DecContext *mp3 = s->priv_data;
415 size= MP3_PACKET_SIZE;
416 pos = avio_tell(s->pb);
417 if(mp3->filesize > ID3v1_TAG_SIZE && pos < mp3->filesize)
418 size= FFMIN(size, mp3->filesize - pos);
420 ret= av_get_packet(s->pb, pkt, size);
427 pkt->flags &= ~AV_PKT_FLAG_CORRUPT;
428 pkt->stream_index = 0;
433 #define SEEK_WINDOW 4096
435 static int check(AVIOContext *pb, int64_t pos, uint32_t *ret_header)
437 int64_t ret = avio_seek(pb, pos, SEEK_SET);
443 header = avio_rb32(pb);
444 if (ff_mpa_check_header(header) < 0)
446 if (avpriv_mpegaudio_decode_header(&sd, header) == 1)
450 *ret_header = header;
451 return sd.frame_size;
454 static int64_t mp3_sync(AVFormatContext *s, int64_t target_pos, int flags)
456 int dir = (flags&AVSEEK_FLAG_BACKWARD) ? -1 : 1;
458 int best_score, i, j;
461 avio_seek(s->pb, FFMAX(target_pos - SEEK_WINDOW, 0), SEEK_SET);
462 ret = avio_seek(s->pb, target_pos, SEEK_SET);
467 best_pos = target_pos;
469 for(i=0; i<SEEK_WINDOW; i++) {
470 int64_t pos = target_pos + (dir > 0 ? i - SEEK_WINDOW/4 : -i);
471 int64_t candidate = -1;
477 for(j=0; j<MIN_VALID; j++) {
478 ret = check(s->pb, pos, NULL);
481 if ((target_pos - pos)*dir <= 0 && abs(MIN_VALID/2-j) < score) {
483 score = abs(MIN_VALID/2-j);
487 if (best_score > score && j == MIN_VALID) {
488 best_pos = candidate;
495 return avio_seek(s->pb, best_pos, SEEK_SET);
498 static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp,
501 MP3DecContext *mp3 = s->priv_data;
502 AVIndexEntry *ie, ie1;
503 AVStream *st = s->streams[0];
504 int64_t ret = av_index_search_timestamp(st, timestamp, flags);
506 int fast_seek = (s->flags & AVFMT_FLAG_FAST_SEEK) ? 1 : 0;
507 int64_t filesize = mp3->header_filesize;
509 if (mp3->usetoc == 2)
510 return -1; // generic index code
513 int64_t size = avio_size(s->pb);
514 if (size > 0 && size > s->internal->data_offset)
515 filesize = size - s->internal->data_offset;
518 if ( (mp3->is_cbr || fast_seek)
519 && (mp3->usetoc == 0 || !mp3->xing_toc)
523 timestamp = av_clip64(timestamp, 0, st->duration);
524 ie->timestamp = timestamp;
525 ie->pos = av_rescale(timestamp, filesize, st->duration) + s->internal->data_offset;
526 } else if (mp3->xing_toc) {
530 ie = &st->index_entries[ret];
535 best_pos = mp3_sync(s, ie->pos, flags);
539 if (mp3->is_cbr && ie == &ie1 && mp3->frames) {
540 int frame_duration = av_rescale(st->duration, 1, mp3->frames);
541 ie1.timestamp = frame_duration * av_rescale(best_pos - s->internal->data_offset, mp3->frames, mp3->header_filesize);
544 ff_update_cur_dts(s, st, ie->timestamp);
548 static const AVOption options[] = {
549 { "usetoc", "use table of contents", offsetof(MP3DecContext, usetoc), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2, AV_OPT_FLAG_DECODING_PARAM},
553 static const AVClass demuxer_class = {
555 .item_name = av_default_item_name,
557 .version = LIBAVUTIL_VERSION_INT,
558 .category = AV_CLASS_CATEGORY_DEMUXER,
561 AVInputFormat ff_mp3_demuxer = {
563 .long_name = NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"),
564 .read_probe = mp3_read_probe,
565 .read_header = mp3_read_header,
566 .read_packet = mp3_read_packet,
567 .read_seek = mp3_seek,
568 .priv_data_size = sizeof(MP3DecContext),
569 .flags = AVFMT_GENERIC_INDEX,
570 .extensions = "mp2,mp3,m2a,mpa", /* XXX: use probe */
571 .priv_class = &demuxer_class,