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 */
61 CHECK_WRONG_HEADER = -1,
62 CHECK_SEEK_FAILED = -2,
65 static int check(AVIOContext *pb, int64_t pos, uint32_t *header);
69 static int mp3_read_probe(AVProbeData *p)
71 int max_frames, first_frames = 0;
74 int framesizes, max_framesizes;
76 const uint8_t *buf, *buf0, *buf2, *end;
79 end = p->buf + p->buf_size - sizeof(uint32_t);
80 while(buf0 < end && !*buf0)
87 for(; buf < end; buf= buf2+1) {
89 for(framesizes = frames = 0; buf2 < end; frames++) {
92 header = AV_RB32(buf2);
93 ret = avpriv_mpegaudio_decode_header(&h, header);
97 framesizes += h.frame_size;
99 max_frames = FFMAX(max_frames, frames);
100 max_framesizes = FFMAX(max_framesizes, framesizes);
102 first_frames= frames;
103 if (buf2 == end + sizeof(uint32_t))
107 // keep this in sync with ac3 probe, both need to avoid
108 // issues with MPEG-files!
109 if (first_frames>=7) return AVPROBE_SCORE_EXTENSION + 1;
110 else if(max_frames>200 && p->buf_size < 2*max_framesizes)return AVPROBE_SCORE_EXTENSION;
111 else if(max_frames>=4 && p->buf_size < 2*max_framesizes) return AVPROBE_SCORE_EXTENSION / 2;
112 else if(ff_id3v2_match(buf0, ID3v2_DEFAULT_MAGIC) && 2*ff_id3v2_tag_len(buf0) >= p->buf_size)
113 return p->buf_size < PROBE_BUF_MAX ? AVPROBE_SCORE_EXTENSION / 4 : AVPROBE_SCORE_EXTENSION - 2;
114 else if(first_frames > 1 && whole_used) return 5;
115 else if(max_frames>=1 && p->buf_size < 10*max_framesizes) return 1;
117 //mpegps_mp3_unrecognized_format.mpg has max_frames=3
120 static void read_xing_toc(AVFormatContext *s, int64_t filesize, int64_t duration)
123 MP3DecContext *mp3 = s->priv_data;
124 int fast_seek = s->flags & AVFMT_FLAG_FAST_SEEK;
125 int fill_index = (mp3->usetoc || fast_seek) && duration > 0;
128 !(filesize = avio_size(s->pb))) {
129 av_log(s, AV_LOG_WARNING, "Cannot determine file size, skipping TOC table.\n");
133 for (i = 0; i < XING_TOC_COUNT; i++) {
134 uint8_t b = avio_r8(s->pb);
136 av_add_index_entry(s->streams[0],
137 av_rescale(b, filesize, 256),
138 av_rescale(i, duration, XING_TOC_COUNT),
139 0, 0, AVINDEX_KEYFRAME);
145 static void mp3_parse_info_tag(AVFormatContext *s, AVStream *st,
146 MPADecodeHeader *c, uint32_t spf)
148 #define LAST_BITS(k, n) ((k) & ((1 << (n)) - 1))
149 #define MIDDLE_BITS(k, m, n) LAST_BITS((k) >> (m), ((n) - (m) + 1))
157 int32_t r_gain = INT32_MIN, a_gain = INT32_MIN;
159 MP3DecContext *mp3 = s->priv_data;
160 static const int64_t xing_offtbl[2][2] = {{32, 17}, {17,9}};
161 uint64_t fsize = avio_size(s->pb);
162 fsize = fsize >= avio_tell(s->pb) ? fsize - avio_tell(s->pb) : 0;
164 /* Check for Xing / Info tag */
165 avio_skip(s->pb, xing_offtbl[c->lsf == 1][c->nb_channels == 1]);
166 v = avio_rb32(s->pb);
167 mp3->is_cbr = v == MKBETAG('I', 'n', 'f', 'o');
168 if (v != MKBETAG('X', 'i', 'n', 'g') && !mp3->is_cbr)
171 v = avio_rb32(s->pb);
172 if (v & XING_FLAG_FRAMES)
173 mp3->frames = avio_rb32(s->pb);
174 if (v & XING_FLAG_SIZE)
175 mp3->header_filesize = avio_rb32(s->pb);
176 if (fsize && mp3->header_filesize) {
178 min = FFMIN(fsize, mp3->header_filesize);
179 delta = FFMAX(fsize, mp3->header_filesize) - min;
180 if (fsize > mp3->header_filesize && delta > min >> 4) {
182 av_log(s, AV_LOG_WARNING,
183 "invalid concatenated file detected - using bitrate for duration\n");
184 } else if (delta > min >> 4) {
185 av_log(s, AV_LOG_WARNING,
186 "filesize and duration do not match (growing file?)\n");
189 if (v & XING_FLAG_TOC)
190 read_xing_toc(s, mp3->header_filesize, av_rescale_q(mp3->frames,
191 (AVRational){spf, c->sample_rate},
194 if (v & XING_FLAC_QSCALE)
197 /* Encoder short version string */
198 memset(version, 0, sizeof(version));
199 avio_read(s->pb, version, 9);
201 /* Info Tag revision + VBR method */
204 /* Lowpass filter value */
207 /* ReplayGain peak */
208 v = avio_rb32(s->pb);
209 peak = av_rescale(v, 100000, 1 << 23);
211 /* Radio ReplayGain */
212 v = avio_rb16(s->pb);
214 if (MIDDLE_BITS(v, 13, 15) == 1) {
215 r_gain = MIDDLE_BITS(v, 0, 8) * 10000;
221 /* Audiophile ReplayGain */
222 v = avio_rb16(s->pb);
224 if (MIDDLE_BITS(v, 13, 15) == 2) {
225 a_gain = MIDDLE_BITS(v, 0, 8) * 10000;
231 /* Encoding flags + ATH Type */
234 /* if ABR {specified bitrate} else {minimal bitrate} */
239 if(AV_RB32(version) == MKBETAG('L', 'A', 'M', 'E')
240 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'f')
241 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'c')
244 mp3->start_pad = v>>12;
245 mp3-> end_pad = v&4095;
246 st->start_skip_samples = mp3->start_pad + 528 + 1;
248 st->first_discard_sample = -mp3->end_pad + 528 + 1 + mp3->frames * (int64_t)spf;
249 st->last_discard_sample = mp3->frames * (int64_t)spf;
252 st->start_time = av_rescale_q(st->start_skip_samples,
253 (AVRational){1, c->sample_rate},
255 av_log(s, AV_LOG_DEBUG, "pad %d %d\n", mp3->start_pad, mp3-> end_pad);
264 /* Preset and surround info */
274 crc = ffio_get_checksum(s->pb);
275 v = avio_rb16(s->pb);
278 ff_replaygain_export_raw(st, r_gain, peak, a_gain, 0);
279 av_dict_set(&st->metadata, "encoder", version, 0);
283 static void mp3_parse_vbri_tag(AVFormatContext *s, AVStream *st, int64_t base)
286 MP3DecContext *mp3 = s->priv_data;
288 /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */
289 avio_seek(s->pb, base + 4 + 32, SEEK_SET);
290 v = avio_rb32(s->pb);
291 if (v == MKBETAG('V', 'B', 'R', 'I')) {
292 /* Check tag version */
293 if (avio_rb16(s->pb) == 1) {
294 /* skip delay and quality */
296 mp3->header_filesize = avio_rb32(s->pb);
297 mp3->frames = avio_rb32(s->pb);
303 * Try to find Xing/Info/VBRI tags and compute duration from info therein
305 static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
310 MP3DecContext *mp3 = s->priv_data;
313 ffio_init_checksum(s->pb, ff_crcA001_update, 0);
315 v = avio_rb32(s->pb);
317 ret = avpriv_mpegaudio_decode_header(&c, v);
321 vbrtag_size = c.frame_size;
325 spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
328 mp3->header_filesize = 0;
330 mp3_parse_info_tag(s, st, &c, spf);
331 mp3_parse_vbri_tag(s, st, base);
333 if (!mp3->frames && !mp3->header_filesize)
336 /* Skip the vbr tag frame */
337 avio_seek(s->pb, base + vbrtag_size, SEEK_SET);
340 st->duration = av_rescale_q(mp3->frames, (AVRational){spf, c.sample_rate},
342 if (mp3->header_filesize && mp3->frames && !mp3->is_cbr)
343 st->codecpar->bit_rate = av_rescale(mp3->header_filesize, 8 * c.sample_rate, mp3->frames * (int64_t)spf);
348 static int mp3_read_header(AVFormatContext *s)
350 MP3DecContext *mp3 = s->priv_data;
356 s->metadata = s->internal->id3v2_meta;
357 s->internal->id3v2_meta = NULL;
359 st = avformat_new_stream(s, NULL);
361 return AVERROR(ENOMEM);
363 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
364 st->codecpar->codec_id = AV_CODEC_ID_MP3;
365 st->need_parsing = AVSTREAM_PARSE_FULL_RAW;
368 // lcm of all mp3 sample rates
369 avpriv_set_pts_info(st, 64, 1, 14112000);
372 off = avio_tell(s->pb);
374 if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX))
377 if(s->pb->seekable & AVIO_SEEKABLE_NORMAL)
378 mp3->filesize = avio_size(s->pb);
380 if (mp3_parse_vbr_tags(s, st, off) < 0)
381 avio_seek(s->pb, off, SEEK_SET);
383 ret = ff_replaygain_export(st, s->metadata);
387 off = avio_tell(s->pb);
388 for (i = 0; i < 64 * 1024; i++) {
389 uint32_t header, header2;
392 ffio_ensure_seekback(s->pb, i + 1024 + 4);
393 frame_size = check(s->pb, off + i, &header);
394 if (frame_size > 0) {
395 ret = avio_seek(s->pb, off, SEEK_SET);
398 ffio_ensure_seekback(s->pb, i + 1024 + frame_size + 4);
399 ret = check(s->pb, off + i + frame_size, &header2);
401 (header & SAME_HEADER_MASK) == (header2 & SAME_HEADER_MASK))
403 av_log(s, i > 0 ? AV_LOG_INFO : AV_LOG_VERBOSE, "Skipping %d bytes of junk at %"PRId64".\n", i, off);
404 ret = avio_seek(s->pb, off + i, SEEK_SET);
408 } else if (ret == CHECK_SEEK_FAILED) {
409 av_log(s, AV_LOG_ERROR, "Invalid frame size (%d): Could not seek to %"PRId64".\n", frame_size, off + i + frame_size);
410 return AVERROR(EINVAL);
412 } else if (frame_size == CHECK_SEEK_FAILED) {
413 av_log(s, AV_LOG_ERROR, "Failed to read frame size: Could not seek to %"PRId64".\n", (int64_t) (i + 1024 + frame_size + 4));
414 return AVERROR(EINVAL);
416 ret = avio_seek(s->pb, off, SEEK_SET);
421 // the seek index is relative to the end of the xing vbr headers
422 for (i = 0; i < st->nb_index_entries; i++)
423 st->index_entries[i].pos += avio_tell(s->pb);
425 /* the parameters will be extracted from the compressed bitstream */
429 #define MP3_PACKET_SIZE 1024
431 static int mp3_read_packet(AVFormatContext *s, AVPacket *pkt)
433 MP3DecContext *mp3 = s->priv_data;
437 size= MP3_PACKET_SIZE;
438 pos = avio_tell(s->pb);
439 if(mp3->filesize > ID3v1_TAG_SIZE && pos < mp3->filesize)
440 size= FFMIN(size, mp3->filesize - pos);
442 ret= av_get_packet(s->pb, pkt, size);
449 pkt->flags &= ~AV_PKT_FLAG_CORRUPT;
450 pkt->stream_index = 0;
455 #define SEEK_WINDOW 4096
457 static int check(AVIOContext *pb, int64_t pos, uint32_t *ret_header)
459 int64_t ret = avio_seek(pb, pos, SEEK_SET);
460 uint8_t header_buf[4];
464 return CHECK_SEEK_FAILED;
466 ret = avio_read(pb, &header_buf[0], 4);
467 /* We should always find four bytes for a valid mpa header. */
469 return CHECK_SEEK_FAILED;
471 header = AV_RB32(&header_buf[0]);
472 if (ff_mpa_check_header(header) < 0)
473 return CHECK_WRONG_HEADER;
474 if (avpriv_mpegaudio_decode_header(&sd, header) == 1)
475 return CHECK_WRONG_HEADER;
478 *ret_header = header;
479 return sd.frame_size;
482 static int64_t mp3_sync(AVFormatContext *s, int64_t target_pos, int flags)
484 int dir = (flags&AVSEEK_FLAG_BACKWARD) ? -1 : 1;
486 int best_score, i, j;
489 avio_seek(s->pb, FFMAX(target_pos - SEEK_WINDOW, 0), SEEK_SET);
490 ret = avio_seek(s->pb, target_pos, SEEK_SET);
495 best_pos = target_pos;
497 for(i=0; i<SEEK_WINDOW; i++) {
498 int64_t pos = target_pos + (dir > 0 ? i - SEEK_WINDOW/4 : -i);
499 int64_t candidate = -1;
505 for(j=0; j<MIN_VALID; j++) {
506 ret = check(s->pb, pos, NULL);
508 if (ret == CHECK_WRONG_HEADER) {
510 } else if (ret == CHECK_SEEK_FAILED) {
511 av_log(s, AV_LOG_ERROR, "Could not seek to %"PRId64".\n", pos);
512 return AVERROR(EINVAL);
515 if ((target_pos - pos)*dir <= 0 && FFABS(MIN_VALID/2-j) < score) {
517 score = FFABS(MIN_VALID/2-j);
521 if (best_score > score && j == MIN_VALID) {
522 best_pos = candidate;
529 return avio_seek(s->pb, best_pos, SEEK_SET);
532 static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp,
535 MP3DecContext *mp3 = s->priv_data;
536 AVIndexEntry *ie, ie1;
537 AVStream *st = s->streams[0];
539 int fast_seek = s->flags & AVFMT_FLAG_FAST_SEEK;
540 int64_t filesize = mp3->header_filesize;
543 int64_t size = avio_size(s->pb);
544 if (size > 0 && size > s->internal->data_offset)
545 filesize = size - s->internal->data_offset;
548 if (mp3->xing_toc && (mp3->usetoc || (fast_seek && !mp3->is_cbr))) {
549 int64_t ret = av_index_search_timestamp(st, timestamp, flags);
551 // NOTE: The MP3 TOC is not a precise lookup table. Accuracy is worse
553 av_log(s, AV_LOG_WARNING, "Using MP3 TOC to seek; may be imprecise.\n");
558 ie = &st->index_entries[ret];
559 } else if (fast_seek && st->duration > 0 && filesize > 0) {
561 av_log(s, AV_LOG_WARNING, "Using scaling to seek VBR MP3; may be imprecise.\n");
564 timestamp = av_clip64(timestamp, 0, st->duration);
565 ie->timestamp = timestamp;
566 ie->pos = av_rescale(timestamp, filesize, st->duration) + s->internal->data_offset;
568 return -1; // generic index code
571 best_pos = mp3_sync(s, ie->pos, flags);
575 if (mp3->is_cbr && ie == &ie1 && mp3->frames) {
576 int frame_duration = av_rescale(st->duration, 1, mp3->frames);
577 ie1.timestamp = frame_duration * av_rescale(best_pos - s->internal->data_offset, mp3->frames, mp3->header_filesize);
580 ff_update_cur_dts(s, st, ie->timestamp);
584 static const AVOption options[] = {
585 { "usetoc", "use table of contents", offsetof(MP3DecContext, usetoc), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_DECODING_PARAM},
589 static const AVClass demuxer_class = {
591 .item_name = av_default_item_name,
593 .version = LIBAVUTIL_VERSION_INT,
594 .category = AV_CLASS_CATEGORY_DEMUXER,
597 AVInputFormat ff_mp3_demuxer = {
599 .long_name = NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"),
600 .read_probe = mp3_read_probe,
601 .read_header = mp3_read_header,
602 .read_packet = mp3_read_packet,
603 .read_seek = mp3_seek,
604 .priv_data_size = sizeof(MP3DecContext),
605 .flags = AVFMT_GENERIC_INDEX,
606 .extensions = "mp2,mp3,m2a,mpa", /* XXX: use probe */
607 .priv_class = &demuxer_class,