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++) {
92 header = AV_RB32(buf2);
93 ret = avpriv_mpegaudio_decode_header(&h, header);
97 available = FFMIN(h.frame_size, end - buf2);
98 for (buf3 = buf2 + 4; buf3 < buf2 + available; buf3++) {
99 uint32_t next_sync = AV_RB32(buf3);
100 header_emu += (next_sync & MP3_MASK) == (header & MP3_MASK);
104 framesizes += h.frame_size;
105 if (available < h.frame_size) {
109 buf2 += h.frame_size;
111 max_frames = FFMAX(max_frames, frames);
112 max_framesizes = FFMAX(max_framesizes, framesizes);
114 first_frames= frames;
115 if (buf2 == end + sizeof(uint32_t))
119 // keep this in sync with ac3 probe, both need to avoid
120 // issues with MPEG-files!
121 if (first_frames>=7) return AVPROBE_SCORE_EXTENSION + 1;
122 else if (max_frames>200 && p->buf_size < 2*max_framesizes)return AVPROBE_SCORE_EXTENSION;
123 else if (max_frames>=4 && p->buf_size < 2*max_framesizes) return AVPROBE_SCORE_EXTENSION / 2;
124 else if (ff_id3v2_match(buf0, ID3v2_DEFAULT_MAGIC) && 2*ff_id3v2_tag_len(buf0) >= p->buf_size)
125 return p->buf_size < PROBE_BUF_MAX ? AVPROBE_SCORE_EXTENSION / 4 : AVPROBE_SCORE_EXTENSION - 2;
126 else if (first_frames > 1 && whole_used) return 5;
127 else if (max_frames>=1 && p->buf_size < 10*max_framesizes) return 1;
129 //mpegps_mp3_unrecognized_format.mpg has max_frames=3
132 static void read_xing_toc(AVFormatContext *s, int64_t filesize, int64_t duration)
135 MP3DecContext *mp3 = s->priv_data;
136 int fast_seek = s->flags & AVFMT_FLAG_FAST_SEEK;
137 int fill_index = (mp3->usetoc || fast_seek) && duration > 0;
140 !(filesize = avio_size(s->pb))) {
141 av_log(s, AV_LOG_WARNING, "Cannot determine file size, skipping TOC table.\n");
145 for (i = 0; i < XING_TOC_COUNT; i++) {
146 uint8_t b = avio_r8(s->pb);
148 av_add_index_entry(s->streams[0],
149 av_rescale(b, filesize, 256),
150 av_rescale(i, duration, XING_TOC_COUNT),
151 0, 0, AVINDEX_KEYFRAME);
157 static void mp3_parse_info_tag(AVFormatContext *s, AVStream *st,
158 MPADecodeHeader *c, uint32_t spf)
160 #define LAST_BITS(k, n) ((k) & ((1 << (n)) - 1))
161 #define MIDDLE_BITS(k, m, n) LAST_BITS((k) >> (m), ((n) - (m) + 1))
169 int32_t r_gain = INT32_MIN, a_gain = INT32_MIN;
171 MP3DecContext *mp3 = s->priv_data;
172 static const int64_t xing_offtbl[2][2] = {{32, 17}, {17,9}};
173 uint64_t fsize = avio_size(s->pb);
174 fsize = fsize >= avio_tell(s->pb) ? fsize - avio_tell(s->pb) : 0;
176 /* Check for Xing / Info tag */
177 avio_skip(s->pb, xing_offtbl[c->lsf == 1][c->nb_channels == 1]);
178 v = avio_rb32(s->pb);
179 mp3->is_cbr = v == MKBETAG('I', 'n', 'f', 'o');
180 if (v != MKBETAG('X', 'i', 'n', 'g') && !mp3->is_cbr)
183 v = avio_rb32(s->pb);
184 if (v & XING_FLAG_FRAMES)
185 mp3->frames = avio_rb32(s->pb);
186 if (v & XING_FLAG_SIZE)
187 mp3->header_filesize = avio_rb32(s->pb);
188 if (fsize && mp3->header_filesize) {
190 min = FFMIN(fsize, mp3->header_filesize);
191 delta = FFMAX(fsize, mp3->header_filesize) - min;
192 if (fsize > mp3->header_filesize && delta > min >> 4) {
194 av_log(s, AV_LOG_WARNING,
195 "invalid concatenated file detected - using bitrate for duration\n");
196 } else if (delta > min >> 4) {
197 av_log(s, AV_LOG_WARNING,
198 "filesize and duration do not match (growing file?)\n");
201 if (v & XING_FLAG_TOC)
202 read_xing_toc(s, mp3->header_filesize, av_rescale_q(mp3->frames,
203 (AVRational){spf, c->sample_rate},
206 if (v & XING_FLAC_QSCALE)
209 /* Encoder short version string */
210 memset(version, 0, sizeof(version));
211 avio_read(s->pb, version, 9);
213 /* Info Tag revision + VBR method */
216 /* Lowpass filter value */
219 /* ReplayGain peak */
220 v = avio_rb32(s->pb);
221 peak = av_rescale(v, 100000, 1 << 23);
223 /* Radio ReplayGain */
224 v = avio_rb16(s->pb);
226 if (MIDDLE_BITS(v, 13, 15) == 1) {
227 r_gain = MIDDLE_BITS(v, 0, 8) * 10000;
233 /* Audiophile ReplayGain */
234 v = avio_rb16(s->pb);
236 if (MIDDLE_BITS(v, 13, 15) == 2) {
237 a_gain = MIDDLE_BITS(v, 0, 8) * 10000;
243 /* Encoding flags + ATH Type */
246 /* if ABR {specified bitrate} else {minimal bitrate} */
250 v = avio_rb24(s->pb);
251 if (AV_RB32(version) == MKBETAG('L', 'A', 'M', 'E')
252 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'f')
253 || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'c')
256 mp3->start_pad = v>>12;
257 mp3-> end_pad = v&4095;
258 st->start_skip_samples = mp3->start_pad + 528 + 1;
260 st->first_discard_sample = -mp3->end_pad + 528 + 1 + mp3->frames * (int64_t)spf;
261 st->last_discard_sample = mp3->frames * (int64_t)spf;
264 st->start_time = av_rescale_q(st->start_skip_samples,
265 (AVRational){1, c->sample_rate},
267 av_log(s, AV_LOG_DEBUG, "pad %d %d\n", mp3->start_pad, mp3-> end_pad);
276 /* Preset and surround info */
286 crc = ffio_get_checksum(s->pb);
287 v = avio_rb16(s->pb);
290 ff_replaygain_export_raw(st, r_gain, peak, a_gain, 0);
291 av_dict_set(&st->metadata, "encoder", version, 0);
295 static void mp3_parse_vbri_tag(AVFormatContext *s, AVStream *st, int64_t base)
298 MP3DecContext *mp3 = s->priv_data;
300 /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */
301 avio_seek(s->pb, base + 4 + 32, SEEK_SET);
302 v = avio_rb32(s->pb);
303 if (v == MKBETAG('V', 'B', 'R', 'I')) {
304 /* Check tag version */
305 if (avio_rb16(s->pb) == 1) {
306 /* skip delay and quality */
308 mp3->header_filesize = avio_rb32(s->pb);
309 mp3->frames = avio_rb32(s->pb);
315 * Try to find Xing/Info/VBRI tags and compute duration from info therein
317 static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
322 MP3DecContext *mp3 = s->priv_data;
325 ffio_init_checksum(s->pb, ff_crcA001_update, 0);
327 v = avio_rb32(s->pb);
329 ret = avpriv_mpegaudio_decode_header(&c, v);
333 vbrtag_size = c.frame_size;
337 spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
340 mp3->header_filesize = 0;
342 mp3_parse_info_tag(s, st, &c, spf);
343 mp3_parse_vbri_tag(s, st, base);
345 if (!mp3->frames && !mp3->header_filesize)
348 /* Skip the vbr tag frame */
349 avio_seek(s->pb, base + vbrtag_size, SEEK_SET);
352 st->duration = av_rescale_q(mp3->frames, (AVRational){spf, c.sample_rate},
354 if (mp3->header_filesize && mp3->frames && !mp3->is_cbr)
355 st->codecpar->bit_rate = av_rescale(mp3->header_filesize, 8 * c.sample_rate, mp3->frames * (int64_t)spf);
360 static int mp3_read_header(AVFormatContext *s)
362 MP3DecContext *mp3 = s->priv_data;
368 s->metadata = s->internal->id3v2_meta;
369 s->internal->id3v2_meta = NULL;
371 st = avformat_new_stream(s, NULL);
373 return AVERROR(ENOMEM);
375 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
376 st->codecpar->codec_id = AV_CODEC_ID_MP3;
377 st->need_parsing = AVSTREAM_PARSE_FULL_RAW;
380 // lcm of all mp3 sample rates
381 avpriv_set_pts_info(st, 64, 1, 14112000);
384 off = avio_tell(s->pb);
386 if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX))
389 if (s->pb->seekable & AVIO_SEEKABLE_NORMAL)
390 mp3->filesize = avio_size(s->pb);
392 if (mp3_parse_vbr_tags(s, st, off) < 0)
393 avio_seek(s->pb, off, SEEK_SET);
395 ret = ff_replaygain_export(st, s->metadata);
399 off = avio_tell(s->pb);
400 for (i = 0; i < 64 * 1024; i++) {
401 uint32_t header, header2;
404 ffio_ensure_seekback(s->pb, i + 1024 + 4);
405 frame_size = check(s->pb, off + i, &header);
406 if (frame_size > 0) {
407 ret = avio_seek(s->pb, off, SEEK_SET);
410 ffio_ensure_seekback(s->pb, i + 1024 + frame_size + 4);
411 ret = check(s->pb, off + i + frame_size, &header2);
413 (header & MP3_MASK) == (header2 & MP3_MASK))
415 av_log(s, i > 0 ? AV_LOG_INFO : AV_LOG_VERBOSE, "Skipping %d bytes of junk at %"PRId64".\n", i, off);
416 ret = avio_seek(s->pb, off + i, SEEK_SET);
420 } else if (ret == CHECK_SEEK_FAILED) {
421 av_log(s, AV_LOG_ERROR, "Invalid frame size (%d): Could not seek to %"PRId64".\n", frame_size, off + i + frame_size);
422 return AVERROR(EINVAL);
424 } else if (frame_size == CHECK_SEEK_FAILED) {
425 av_log(s, AV_LOG_ERROR, "Failed to read frame size: Could not seek to %"PRId64".\n", (int64_t) (i + 1024 + frame_size + 4));
426 return AVERROR(EINVAL);
428 ret = avio_seek(s->pb, off, SEEK_SET);
433 // the seek index is relative to the end of the xing vbr headers
434 for (i = 0; i < st->nb_index_entries; i++)
435 st->index_entries[i].pos += avio_tell(s->pb);
437 /* the parameters will be extracted from the compressed bitstream */
441 #define MP3_PACKET_SIZE 1024
443 static int mp3_read_packet(AVFormatContext *s, AVPacket *pkt)
445 MP3DecContext *mp3 = s->priv_data;
449 size = MP3_PACKET_SIZE;
450 pos = avio_tell(s->pb);
451 if (mp3->filesize > ID3v1_TAG_SIZE && pos < mp3->filesize)
452 size= FFMIN(size, mp3->filesize - pos);
454 ret = av_get_packet(s->pb, pkt, size);
461 pkt->flags &= ~AV_PKT_FLAG_CORRUPT;
462 pkt->stream_index = 0;
467 #define SEEK_WINDOW 4096
469 static int check(AVIOContext *pb, int64_t pos, uint32_t *ret_header)
471 int64_t ret = avio_seek(pb, pos, SEEK_SET);
472 uint8_t header_buf[4];
476 return CHECK_SEEK_FAILED;
478 ret = avio_read(pb, &header_buf[0], 4);
479 /* We should always find four bytes for a valid mpa header. */
481 return CHECK_SEEK_FAILED;
483 header = AV_RB32(&header_buf[0]);
484 if (ff_mpa_check_header(header) < 0)
485 return CHECK_WRONG_HEADER;
486 if (avpriv_mpegaudio_decode_header(&sd, header) == 1)
487 return CHECK_WRONG_HEADER;
490 *ret_header = header;
491 return sd.frame_size;
494 static int64_t mp3_sync(AVFormatContext *s, int64_t target_pos, int flags)
496 int dir = (flags&AVSEEK_FLAG_BACKWARD) ? -1 : 1;
498 int best_score, i, j;
501 avio_seek(s->pb, FFMAX(target_pos - SEEK_WINDOW, 0), SEEK_SET);
502 ret = avio_seek(s->pb, target_pos, SEEK_SET);
507 best_pos = target_pos;
509 for (i = 0; i < SEEK_WINDOW; i++) {
510 int64_t pos = target_pos + (dir > 0 ? i - SEEK_WINDOW/4 : -i);
511 int64_t candidate = -1;
517 for (j = 0; j < MIN_VALID; j++) {
518 ret = check(s->pb, pos, NULL);
520 if (ret == CHECK_WRONG_HEADER) {
522 } else if (ret == CHECK_SEEK_FAILED) {
523 av_log(s, AV_LOG_ERROR, "Could not seek to %"PRId64".\n", pos);
524 return AVERROR(EINVAL);
527 if ((target_pos - pos)*dir <= 0 && FFABS(MIN_VALID/2-j) < score) {
529 score = FFABS(MIN_VALID/2-j);
533 if (best_score > score && j == MIN_VALID) {
534 best_pos = candidate;
541 return avio_seek(s->pb, best_pos, SEEK_SET);
544 static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp,
547 MP3DecContext *mp3 = s->priv_data;
548 AVIndexEntry *ie, ie1;
549 AVStream *st = s->streams[0];
551 int fast_seek = s->flags & AVFMT_FLAG_FAST_SEEK;
552 int64_t filesize = mp3->header_filesize;
555 int64_t size = avio_size(s->pb);
556 if (size > 0 && size > s->internal->data_offset)
557 filesize = size - s->internal->data_offset;
560 if (mp3->xing_toc && (mp3->usetoc || (fast_seek && !mp3->is_cbr))) {
561 int64_t ret = av_index_search_timestamp(st, timestamp, flags);
563 // NOTE: The MP3 TOC is not a precise lookup table. Accuracy is worse
565 av_log(s, AV_LOG_WARNING, "Using MP3 TOC to seek; may be imprecise.\n");
570 ie = &st->index_entries[ret];
571 } else if (fast_seek && st->duration > 0 && filesize > 0) {
573 av_log(s, AV_LOG_WARNING, "Using scaling to seek VBR MP3; may be imprecise.\n");
576 timestamp = av_clip64(timestamp, 0, st->duration);
577 ie->timestamp = timestamp;
578 ie->pos = av_rescale(timestamp, filesize, st->duration) + s->internal->data_offset;
580 return -1; // generic index code
583 best_pos = mp3_sync(s, ie->pos, flags);
587 if (mp3->is_cbr && ie == &ie1 && mp3->frames) {
588 int frame_duration = av_rescale(st->duration, 1, mp3->frames);
589 ie1.timestamp = frame_duration * av_rescale(best_pos - s->internal->data_offset, mp3->frames, mp3->header_filesize);
592 ff_update_cur_dts(s, st, ie->timestamp);
596 static const AVOption options[] = {
597 { "usetoc", "use table of contents", offsetof(MP3DecContext, usetoc), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_DECODING_PARAM},
601 static const AVClass demuxer_class = {
603 .item_name = av_default_item_name,
605 .version = LIBAVUTIL_VERSION_INT,
606 .category = AV_CLASS_CATEGORY_DEMUXER,
609 AVInputFormat ff_mp3_demuxer = {
611 .long_name = NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"),
612 .read_probe = mp3_read_probe,
613 .read_header = mp3_read_header,
614 .read_packet = mp3_read_packet,
615 .read_seek = mp3_seek,
616 .priv_data_size = sizeof(MP3DecContext),
617 .flags = AVFMT_GENERIC_INDEX,
618 .extensions = "mp2,mp3,m2a,mpa", /* XXX: use probe */
619 .priv_class = &demuxer_class,