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
53 unsigned frames; /* Total number of frames in file */
54 unsigned header_filesize; /* Total number of bytes in the stream */
59 CHECK_WRONG_HEADER = -1,
60 CHECK_SEEK_FAILED = -2,
63 static int check(AVIOContext *pb, int64_t pos, uint32_t *header);
67 static int mp3_read_probe(const AVProbeData *p)
69 int max_frames, first_frames = 0;
72 int framesizes, max_framesizes;
74 const uint8_t *buf, *buf0, *buf2, *buf3, *end;
77 end = p->buf + p->buf_size - sizeof(uint32_t);
78 while (buf0 < end && !*buf0)
85 for (; buf < end; buf = buf2+1) {
87 for (framesizes = frames = 0; buf2 < end; frames++) {
91 header = AV_RB32(buf2);
92 ret = avpriv_mpegaudio_decode_header(&h, header);
93 if (ret != 0 || end - buf2 < h.frame_size)
96 for (buf3 = buf2 + 4; buf3 < buf2 + h.frame_size; buf3++) {
97 uint32_t next_sync = AV_RB32(buf3);
98 header_emu += (next_sync & MP3_MASK) == (header & MP3_MASK);
102 buf2 += h.frame_size;
103 framesizes += h.frame_size;
105 max_frames = FFMAX(max_frames, frames);
106 max_framesizes = FFMAX(max_framesizes, framesizes);
108 first_frames= frames;
109 if (buf2 == end + sizeof(uint32_t))
113 // keep this in sync with ac3 probe, both need to avoid
114 // issues with MPEG-files!
115 if (first_frames>=7) return AVPROBE_SCORE_EXTENSION + 1;
116 else if (max_frames>200 && p->buf_size < 2*max_framesizes)return AVPROBE_SCORE_EXTENSION;
117 else if (max_frames>=4 && p->buf_size < 2*max_framesizes) return AVPROBE_SCORE_EXTENSION / 2;
118 else if (ff_id3v2_match(buf0, ID3v2_DEFAULT_MAGIC) && 2*ff_id3v2_tag_len(buf0) >= p->buf_size)
119 return p->buf_size < PROBE_BUF_MAX ? AVPROBE_SCORE_EXTENSION / 4 : AVPROBE_SCORE_EXTENSION - 2;
120 else if (first_frames > 1 && whole_used) return 5;
121 else if (max_frames>=1 && p->buf_size < 10*max_framesizes) return 1;
123 //mpegps_mp3_unrecognized_format.mpg has max_frames=3
126 static void read_xing_toc(AVFormatContext *s, int64_t filesize, int64_t duration)
129 MP3DecContext *mp3 = s->priv_data;
130 int fast_seek = s->flags & AVFMT_FLAG_FAST_SEEK;
131 int fill_index = (mp3->usetoc || fast_seek) && duration > 0;
134 !(filesize = avio_size(s->pb))) {
135 av_log(s, AV_LOG_WARNING, "Cannot determine file size, skipping TOC table.\n");
139 for (i = 0; i < XING_TOC_COUNT; i++) {
140 uint8_t b = avio_r8(s->pb);
142 av_add_index_entry(s->streams[0],
143 av_rescale(b, filesize, 256),
144 av_rescale(i, duration, XING_TOC_COUNT),
145 0, 0, AVINDEX_KEYFRAME);
151 static void mp3_parse_info_tag(AVFormatContext *s, AVStream *st,
152 MPADecodeHeader *c, uint32_t spf)
154 #define LAST_BITS(k, n) ((k) & ((1 << (n)) - 1))
155 #define MIDDLE_BITS(k, m, n) LAST_BITS((k) >> (m), ((n) - (m) + 1))
163 int32_t r_gain = INT32_MIN, a_gain = INT32_MIN;
165 MP3DecContext *mp3 = s->priv_data;
166 static const int64_t xing_offtbl[2][2] = {{32, 17}, {17,9}};
167 uint64_t fsize = avio_size(s->pb);
168 fsize = fsize >= avio_tell(s->pb) ? fsize - avio_tell(s->pb) : 0;
170 /* Check for Xing / Info tag */
171 avio_skip(s->pb, xing_offtbl[c->lsf == 1][c->nb_channels == 1]);
172 v = avio_rb32(s->pb);
173 mp3->is_cbr = v == MKBETAG('I', 'n', 'f', 'o');
174 if (v != MKBETAG('X', 'i', 'n', 'g') && !mp3->is_cbr)
177 v = avio_rb32(s->pb);
178 if (v & XING_FLAG_FRAMES)
179 mp3->frames = avio_rb32(s->pb);
180 if (v & XING_FLAG_SIZE)
181 mp3->header_filesize = avio_rb32(s->pb);
182 if (fsize && mp3->header_filesize) {
184 min = FFMIN(fsize, mp3->header_filesize);
185 delta = FFMAX(fsize, mp3->header_filesize) - min;
186 if (fsize > mp3->header_filesize && delta > min >> 4) {
188 av_log(s, AV_LOG_WARNING,
189 "invalid concatenated file detected - using bitrate for duration\n");
190 } else if (delta > min >> 4) {
191 av_log(s, AV_LOG_WARNING,
192 "filesize and duration do not match (growing file?)\n");
195 if (v & XING_FLAG_TOC)
196 read_xing_toc(s, mp3->header_filesize, av_rescale_q(mp3->frames,
197 (AVRational){spf, c->sample_rate},
200 if (v & XING_FLAC_QSCALE)
203 /* Encoder short version string */
204 memset(version, 0, sizeof(version));
205 avio_read(s->pb, version, 9);
207 /* Info Tag revision + VBR method */
210 /* Lowpass filter value */
213 /* ReplayGain peak */
214 v = avio_rb32(s->pb);
215 peak = av_rescale(v, 100000, 1 << 23);
217 /* Radio ReplayGain */
218 v = avio_rb16(s->pb);
220 if (MIDDLE_BITS(v, 13, 15) == 1) {
221 r_gain = MIDDLE_BITS(v, 0, 8) * 10000;
227 /* Audiophile ReplayGain */
228 v = avio_rb16(s->pb);
230 if (MIDDLE_BITS(v, 13, 15) == 2) {
231 a_gain = MIDDLE_BITS(v, 0, 8) * 10000;
237 /* Encoding flags + ATH Type */
240 /* if ABR {specified bitrate} else {minimal bitrate} */
244 v = avio_rb24(s->pb);
245 if (AV_RB32(version) == MKBETAG('L', 'A', 'M', 'E')
246 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'f')
247 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'c')
250 mp3->start_pad = v>>12;
251 mp3-> end_pad = v&4095;
252 st->start_skip_samples = mp3->start_pad + 528 + 1;
254 st->first_discard_sample = -mp3->end_pad + 528 + 1 + mp3->frames * (int64_t)spf;
255 st->last_discard_sample = mp3->frames * (int64_t)spf;
258 st->start_time = av_rescale_q(st->start_skip_samples,
259 (AVRational){1, c->sample_rate},
261 av_log(s, AV_LOG_DEBUG, "pad %d %d\n", mp3->start_pad, mp3-> end_pad);
270 /* Preset and surround info */
280 crc = ffio_get_checksum(s->pb);
281 v = avio_rb16(s->pb);
284 ff_replaygain_export_raw(st, r_gain, peak, a_gain, 0);
285 av_dict_set(&st->metadata, "encoder", version, 0);
289 static void mp3_parse_vbri_tag(AVFormatContext *s, AVStream *st, int64_t base)
292 MP3DecContext *mp3 = s->priv_data;
294 /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */
295 avio_seek(s->pb, base + 4 + 32, SEEK_SET);
296 v = avio_rb32(s->pb);
297 if (v == MKBETAG('V', 'B', 'R', 'I')) {
298 /* Check tag version */
299 if (avio_rb16(s->pb) == 1) {
300 /* skip delay and quality */
302 mp3->header_filesize = avio_rb32(s->pb);
303 mp3->frames = avio_rb32(s->pb);
309 * Try to find Xing/Info/VBRI tags and compute duration from info therein
311 static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
316 MP3DecContext *mp3 = s->priv_data;
319 ffio_init_checksum(s->pb, ff_crcA001_update, 0);
321 v = avio_rb32(s->pb);
323 ret = avpriv_mpegaudio_decode_header(&c, v);
327 vbrtag_size = c.frame_size;
331 spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
334 mp3->header_filesize = 0;
336 mp3_parse_info_tag(s, st, &c, spf);
337 mp3_parse_vbri_tag(s, st, base);
339 if (!mp3->frames && !mp3->header_filesize)
342 /* Skip the vbr tag frame */
343 avio_seek(s->pb, base + vbrtag_size, SEEK_SET);
346 st->duration = av_rescale_q(mp3->frames, (AVRational){spf, c.sample_rate},
348 if (mp3->header_filesize && mp3->frames && !mp3->is_cbr)
349 st->codecpar->bit_rate = av_rescale(mp3->header_filesize, 8 * c.sample_rate, mp3->frames * (int64_t)spf);
354 static int mp3_read_header(AVFormatContext *s)
356 MP3DecContext *mp3 = s->priv_data;
362 s->metadata = s->internal->id3v2_meta;
363 s->internal->id3v2_meta = NULL;
365 st = avformat_new_stream(s, NULL);
367 return AVERROR(ENOMEM);
369 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
370 st->codecpar->codec_id = AV_CODEC_ID_MP3;
371 st->need_parsing = AVSTREAM_PARSE_FULL_RAW;
374 // lcm of all mp3 sample rates
375 avpriv_set_pts_info(st, 64, 1, 14112000);
378 off = avio_tell(s->pb);
380 if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX))
383 if (s->pb->seekable & AVIO_SEEKABLE_NORMAL)
384 mp3->filesize = avio_size(s->pb);
386 if (mp3_parse_vbr_tags(s, st, off) < 0)
387 avio_seek(s->pb, off, SEEK_SET);
389 ret = ff_replaygain_export(st, s->metadata);
393 off = avio_tell(s->pb);
394 for (i = 0; i < 64 * 1024; i++) {
395 uint32_t header, header2;
398 ffio_ensure_seekback(s->pb, i + 1024 + 4);
399 frame_size = check(s->pb, off + i, &header);
400 if (frame_size > 0) {
401 ret = avio_seek(s->pb, off, SEEK_SET);
404 ffio_ensure_seekback(s->pb, i + 1024 + frame_size + 4);
405 ret = check(s->pb, off + i + frame_size, &header2);
407 (header & MP3_MASK) == (header2 & MP3_MASK))
409 av_log(s, i > 0 ? AV_LOG_INFO : AV_LOG_VERBOSE, "Skipping %d bytes of junk at %"PRId64".\n", i, off);
410 ret = avio_seek(s->pb, off + i, SEEK_SET);
414 } else if (ret == CHECK_SEEK_FAILED) {
415 av_log(s, AV_LOG_ERROR, "Invalid frame size (%d): Could not seek to %"PRId64".\n", frame_size, off + i + frame_size);
416 return AVERROR(EINVAL);
418 } else if (frame_size == CHECK_SEEK_FAILED) {
419 av_log(s, AV_LOG_ERROR, "Failed to read frame size: Could not seek to %"PRId64".\n", (int64_t) (i + 1024 + frame_size + 4));
420 return AVERROR(EINVAL);
422 ret = avio_seek(s->pb, off, SEEK_SET);
427 // the seek index is relative to the end of the xing vbr headers
428 for (i = 0; i < st->nb_index_entries; i++)
429 st->index_entries[i].pos += avio_tell(s->pb);
431 /* the parameters will be extracted from the compressed bitstream */
435 #define MP3_PACKET_SIZE 1024
437 static int mp3_read_packet(AVFormatContext *s, AVPacket *pkt)
439 MP3DecContext *mp3 = s->priv_data;
443 size = MP3_PACKET_SIZE;
444 pos = avio_tell(s->pb);
445 if (mp3->filesize > ID3v1_TAG_SIZE && pos < mp3->filesize)
446 size= FFMIN(size, mp3->filesize - pos);
448 ret = av_get_packet(s->pb, pkt, size);
455 pkt->flags &= ~AV_PKT_FLAG_CORRUPT;
456 pkt->stream_index = 0;
461 #define SEEK_WINDOW 4096
463 static int check(AVIOContext *pb, int64_t pos, uint32_t *ret_header)
465 int64_t ret = avio_seek(pb, pos, SEEK_SET);
466 uint8_t header_buf[4];
470 return CHECK_SEEK_FAILED;
472 ret = avio_read(pb, &header_buf[0], 4);
473 /* We should always find four bytes for a valid mpa header. */
475 return CHECK_SEEK_FAILED;
477 header = AV_RB32(&header_buf[0]);
478 if (ff_mpa_check_header(header) < 0)
479 return CHECK_WRONG_HEADER;
480 if (avpriv_mpegaudio_decode_header(&sd, header) == 1)
481 return CHECK_WRONG_HEADER;
484 *ret_header = header;
485 return sd.frame_size;
488 static int64_t mp3_sync(AVFormatContext *s, int64_t target_pos, int flags)
490 int dir = (flags&AVSEEK_FLAG_BACKWARD) ? -1 : 1;
492 int best_score, i, j;
495 avio_seek(s->pb, FFMAX(target_pos - SEEK_WINDOW, 0), SEEK_SET);
496 ret = avio_seek(s->pb, target_pos, SEEK_SET);
501 best_pos = target_pos;
503 for (i = 0; i < SEEK_WINDOW; i++) {
504 int64_t pos = target_pos + (dir > 0 ? i - SEEK_WINDOW/4 : -i);
505 int64_t candidate = -1;
511 for (j = 0; j < MIN_VALID; j++) {
512 ret = check(s->pb, pos, NULL);
514 if (ret == CHECK_WRONG_HEADER) {
516 } else if (ret == CHECK_SEEK_FAILED) {
517 av_log(s, AV_LOG_ERROR, "Could not seek to %"PRId64".\n", pos);
518 return AVERROR(EINVAL);
521 if ((target_pos - pos)*dir <= 0 && FFABS(MIN_VALID/2-j) < score) {
523 score = FFABS(MIN_VALID/2-j);
527 if (best_score > score && j == MIN_VALID) {
528 best_pos = candidate;
535 return avio_seek(s->pb, best_pos, SEEK_SET);
538 static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp,
541 MP3DecContext *mp3 = s->priv_data;
542 AVIndexEntry *ie, ie1;
543 AVStream *st = s->streams[0];
545 int fast_seek = s->flags & AVFMT_FLAG_FAST_SEEK;
546 int64_t filesize = mp3->header_filesize;
549 int64_t size = avio_size(s->pb);
550 if (size > 0 && size > s->internal->data_offset)
551 filesize = size - s->internal->data_offset;
554 if (mp3->xing_toc && (mp3->usetoc || (fast_seek && !mp3->is_cbr))) {
555 int64_t ret = av_index_search_timestamp(st, timestamp, flags);
557 // NOTE: The MP3 TOC is not a precise lookup table. Accuracy is worse
559 av_log(s, AV_LOG_WARNING, "Using MP3 TOC to seek; may be imprecise.\n");
564 ie = &st->index_entries[ret];
565 } else if (fast_seek && st->duration > 0 && filesize > 0) {
567 av_log(s, AV_LOG_WARNING, "Using scaling to seek VBR MP3; may be imprecise.\n");
570 timestamp = av_clip64(timestamp, 0, st->duration);
571 ie->timestamp = timestamp;
572 ie->pos = av_rescale(timestamp, filesize, st->duration) + s->internal->data_offset;
574 return -1; // generic index code
577 best_pos = mp3_sync(s, ie->pos, flags);
581 if (mp3->is_cbr && ie == &ie1 && mp3->frames) {
582 int frame_duration = av_rescale(st->duration, 1, mp3->frames);
583 ie1.timestamp = frame_duration * av_rescale(best_pos - s->internal->data_offset, mp3->frames, mp3->header_filesize);
586 ff_update_cur_dts(s, st, ie->timestamp);
590 static const AVOption options[] = {
591 { "usetoc", "use table of contents", offsetof(MP3DecContext, usetoc), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_DECODING_PARAM},
595 static const AVClass demuxer_class = {
597 .item_name = av_default_item_name,
599 .version = LIBAVUTIL_VERSION_INT,
600 .category = AV_CLASS_CATEGORY_DEMUXER,
603 AVInputFormat ff_mp3_demuxer = {
605 .long_name = NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"),
606 .read_probe = mp3_read_probe,
607 .read_header = mp3_read_header,
608 .read_packet = mp3_read_packet,
609 .read_seek = mp3_seek,
610 .priv_data_size = sizeof(MP3DecContext),
611 .flags = AVFMT_GENERIC_INDEX,
612 .extensions = "mp2,mp3,m2a,mpa", /* XXX: use probe */
613 .priv_class = &demuxer_class,