FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
mp3dec.c
Go to the documentation of this file.
1 /*
2  * MP3 demuxer
3  * Copyright (c) 2003 Fabrice Bellard
4  *
5  * This file is part of FFmpeg.
6  *
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.
11  *
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.
16  *
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
20  */
21 
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"
28 #include "avformat.h"
29 #include "internal.h"
30 #include "avio_internal.h"
31 #include "id3v2.h"
32 #include "id3v1.h"
33 #include "replaygain.h"
34 
36 
37 #define XING_FLAG_FRAMES 0x01
38 #define XING_FLAG_SIZE 0x02
39 #define XING_FLAG_TOC 0x04
40 #define XING_FLAC_QSCALE 0x08
41 
42 #define XING_TOC_COUNT 100
43 
44 typedef struct {
45  AVClass *class;
46  int64_t filesize;
47  int xing_toc;
48  int start_pad;
49  int end_pad;
50  int usetoc;
51  unsigned frames; /* Total number of frames in file */
52  unsigned header_filesize; /* Total number of bytes in the stream */
53  int is_cbr;
55 
56 /* mp3 read */
57 
59 {
60  int max_frames, first_frames = 0;
61  int fsize, frames;
62  uint32_t header;
63  const uint8_t *buf, *buf0, *buf2, *end;
64  AVCodecContext avctx;
65 
66  buf0 = p->buf;
67  end = p->buf + p->buf_size - sizeof(uint32_t);
68  while(buf0 < end && !*buf0)
69  buf0++;
70 
71  max_frames = 0;
72  buf = buf0;
73 
74  for(; buf < end; buf= buf2+1) {
75  buf2 = buf;
76  if(ff_mpa_check_header(AV_RB32(buf2)))
77  continue;
78 
79  for(frames = 0; buf2 < end; frames++) {
80  int dummy;
81  header = AV_RB32(buf2);
82  fsize = avpriv_mpa_decode_header(&avctx, header, &dummy, &dummy, &dummy, &dummy);
83  if(fsize < 0)
84  break;
85  buf2 += fsize;
86  }
87  max_frames = FFMAX(max_frames, frames);
88  if(buf == buf0)
89  first_frames= frames;
90  }
91  // keep this in sync with ac3 probe, both need to avoid
92  // issues with MPEG-files!
93  if (first_frames>=4) return AVPROBE_SCORE_EXTENSION + 1;
94  else if(max_frames>200)return AVPROBE_SCORE_EXTENSION;
95  else if(max_frames>=4 && max_frames >= p->buf_size/10000) return AVPROBE_SCORE_EXTENSION / 2;
96  else if(ff_id3v2_match(buf0, ID3v2_DEFAULT_MAGIC) && 2*ff_id3v2_tag_len(buf0) >= p->buf_size)
98  else if(max_frames>=1 && max_frames >= p->buf_size/10000) return 1;
99  else return 0;
100 //mpegps_mp3_unrecognized_format.mpg has max_frames=3
101 }
102 
103 static void read_xing_toc(AVFormatContext *s, int64_t filesize, int64_t duration)
104 {
105  int i;
106  MP3DecContext *mp3 = s->priv_data;
107  int fill_index = mp3->usetoc && duration > 0;
108 
109  if (!filesize &&
110  !(filesize = avio_size(s->pb))) {
111  av_log(s, AV_LOG_WARNING, "Cannot determine file size, skipping TOC table.\n");
112  fill_index = 0;
113  }
114 
115  for (i = 0; i < XING_TOC_COUNT; i++) {
116  uint8_t b = avio_r8(s->pb);
117  if (fill_index)
119  av_rescale(b, filesize, 256),
120  av_rescale(i, duration, XING_TOC_COUNT),
121  0, 0, AVINDEX_KEYFRAME);
122  }
123  if (fill_index)
124  mp3->xing_toc = 1;
125 }
126 
128  MPADecodeHeader *c, uint32_t spf)
129 {
130 #define LAST_BITS(k, n) ((k) & ((1 << (n)) - 1))
131 #define MIDDLE_BITS(k, m, n) LAST_BITS((k) >> (m), ((n) - (m)))
132 
133  uint16_t crc;
134  uint32_t v;
135 
136  char version[10];
137 
138  uint32_t peak = 0;
139  int32_t r_gain = INT32_MIN, a_gain = INT32_MIN;
140 
141  MP3DecContext *mp3 = s->priv_data;
142  static const int64_t xing_offtbl[2][2] = {{32, 17}, {17,9}};
143  uint64_t fsize = avio_size(s->pb);
144 
145  /* Check for Xing / Info tag */
146  avio_skip(s->pb, xing_offtbl[c->lsf == 1][c->nb_channels == 1]);
147  v = avio_rb32(s->pb);
148  mp3->is_cbr = v == MKBETAG('I', 'n', 'f', 'o');
149  if (v != MKBETAG('X', 'i', 'n', 'g') && !mp3->is_cbr)
150  return;
151 
152  v = avio_rb32(s->pb);
153  if (v & XING_FLAG_FRAMES)
154  mp3->frames = avio_rb32(s->pb);
155  if (v & XING_FLAG_SIZE)
156  mp3->header_filesize = avio_rb32(s->pb);
157  if (fsize && mp3->header_filesize) {
158  uint64_t min, delta;
159  min = FFMIN(fsize, mp3->header_filesize);
160  delta = FFMAX(fsize, mp3->header_filesize) - min;
161  if (fsize > mp3->header_filesize && delta > min >> 4) {
162  mp3->frames = 0;
163  } else if (delta > min >> 4) {
165  "filesize and duration do not match (growing file?)\n");
166  }
167  }
168  if (v & XING_FLAG_TOC)
170  (AVRational){spf, c->sample_rate},
171  st->time_base));
172  /* VBR quality */
173  if (v & XING_FLAC_QSCALE)
174  avio_rb32(s->pb);
175 
176  /* Encoder short version string */
177  memset(version, 0, sizeof(version));
178  avio_read(s->pb, version, 9);
179 
180  /* Info Tag revision + VBR method */
181  avio_r8(s->pb);
182 
183  /* Lowpass filter value */
184  avio_r8(s->pb);
185 
186  /* ReplayGain peak */
187  v = avio_rb32(s->pb);
188  peak = av_rescale(v, 100000, 1 << 23);
189 
190  /* Radio ReplayGain */
191  v = avio_rb16(s->pb);
192 
193  if (MIDDLE_BITS(v, 13, 15) == 1) {
194  r_gain = MIDDLE_BITS(v, 0, 8) * 10000;
195 
196  if (v & (1 << 9))
197  r_gain *= -1;
198  }
199 
200  /* Audiophile ReplayGain */
201  v = avio_rb16(s->pb);
202 
203  if (MIDDLE_BITS(v, 13, 15) == 2) {
204  a_gain = MIDDLE_BITS(v, 0, 8) * 10000;
205 
206  if (v & (1 << 9))
207  a_gain *= -1;
208  }
209 
210  /* Encoding flags + ATH Type */
211  avio_r8(s->pb);
212 
213  /* if ABR {specified bitrate} else {minimal bitrate} */
214  avio_r8(s->pb);
215 
216  /* Encoder delays */
217  v= avio_rb24(s->pb);
218  if(AV_RB32(version) == MKBETAG('L', 'A', 'M', 'E')
219  || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'f')
220  || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'c')
221  ) {
222 
223  mp3->start_pad = v>>12;
224  mp3-> end_pad = v&4095;
225  st->skip_samples = mp3->start_pad + 528 + 1;
226  if (mp3->frames) {
227  st->first_discard_sample = -mp3->end_pad + 528 + 1 + mp3->frames * (int64_t)spf;
228  st->last_discard_sample = mp3->frames * (int64_t)spf;
229  }
230  if (!st->start_time)
231  st->start_time = av_rescale_q(st->skip_samples,
232  (AVRational){1, c->sample_rate},
233  st->time_base);
234  av_log(s, AV_LOG_DEBUG, "pad %d %d\n", mp3->start_pad, mp3-> end_pad);
235  }
236 
237  /* Misc */
238  avio_r8(s->pb);
239 
240  /* MP3 gain */
241  avio_r8(s->pb);
242 
243  /* Preset and surround info */
244  avio_rb16(s->pb);
245 
246  /* Music length */
247  avio_rb32(s->pb);
248 
249  /* Music CRC */
250  avio_rb16(s->pb);
251 
252  /* Info Tag CRC */
253  crc = ffio_get_checksum(s->pb);
254  v = avio_rb16(s->pb);
255 
256  if (v == crc) {
257  ff_replaygain_export_raw(st, r_gain, peak, a_gain, 0);
258  av_dict_set(&st->metadata, "encoder", version, 0);
259  }
260 }
261 
262 static void mp3_parse_vbri_tag(AVFormatContext *s, AVStream *st, int64_t base)
263 {
264  uint32_t v;
265  MP3DecContext *mp3 = s->priv_data;
266 
267  /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */
268  avio_seek(s->pb, base + 4 + 32, SEEK_SET);
269  v = avio_rb32(s->pb);
270  if (v == MKBETAG('V', 'B', 'R', 'I')) {
271  /* Check tag version */
272  if (avio_rb16(s->pb) == 1) {
273  /* skip delay and quality */
274  avio_skip(s->pb, 4);
275  mp3->header_filesize = avio_rb32(s->pb);
276  mp3->frames = avio_rb32(s->pb);
277  }
278  }
279 }
280 
281 /**
282  * Try to find Xing/Info/VBRI tags and compute duration from info therein
283  */
284 static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
285 {
286  uint32_t v, spf;
288  int vbrtag_size = 0;
289  MP3DecContext *mp3 = s->priv_data;
290 
292 
293  v = avio_rb32(s->pb);
294  if(ff_mpa_check_header(v) < 0)
295  return -1;
296 
297  if (avpriv_mpegaudio_decode_header(&c, v) == 0)
298  vbrtag_size = c.frame_size;
299  if(c.layer != 3)
300  return -1;
301 
302  spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
303 
304  mp3->frames = 0;
305  mp3->header_filesize = 0;
306 
307  mp3_parse_info_tag(s, st, &c, spf);
308  mp3_parse_vbri_tag(s, st, base);
309 
310  if (!mp3->frames && !mp3->header_filesize)
311  return -1;
312 
313  /* Skip the vbr tag frame */
314  avio_seek(s->pb, base + vbrtag_size, SEEK_SET);
315 
316  if (mp3->frames)
317  st->duration = av_rescale_q(mp3->frames, (AVRational){spf, c.sample_rate},
318  st->time_base);
319  if (mp3->header_filesize && mp3->frames && !mp3->is_cbr)
320  st->codec->bit_rate = av_rescale(mp3->header_filesize, 8 * c.sample_rate, mp3->frames * (int64_t)spf);
321 
322  return 0;
323 }
324 
326 {
327  MP3DecContext *mp3 = s->priv_data;
328  AVStream *st;
329  int64_t off;
330  int ret;
331 
332  st = avformat_new_stream(s, NULL);
333  if (!st)
334  return AVERROR(ENOMEM);
335 
339  st->start_time = 0;
340 
341  // lcm of all mp3 sample rates
342  avpriv_set_pts_info(st, 64, 1, 14112000);
343 
344  s->pb->maxsize = -1;
345  off = avio_tell(s->pb);
346 
347  if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX))
348  ff_id3v1_read(s);
349 
350  if(s->pb->seekable)
351  mp3->filesize = avio_size(s->pb);
352 
353  if (mp3_parse_vbr_tags(s, st, off) < 0)
354  avio_seek(s->pb, off, SEEK_SET);
355 
356  ret = ff_replaygain_export(st, s->metadata);
357  if (ret < 0)
358  return ret;
359 
360  /* the parameters will be extracted from the compressed bitstream */
361  return 0;
362 }
363 
364 #define MP3_PACKET_SIZE 1024
365 
367 {
368  MP3DecContext *mp3 = s->priv_data;
369  int ret, size;
370  int64_t pos;
371 
372  size= MP3_PACKET_SIZE;
373  pos = avio_tell(s->pb);
374  if(mp3->filesize > ID3v1_TAG_SIZE && pos < mp3->filesize)
375  size= FFMIN(size, mp3->filesize - pos);
376 
377  ret= av_get_packet(s->pb, pkt, size);
378  if (ret <= 0) {
379  if(ret<0)
380  return ret;
381  return AVERROR_EOF;
382  }
383 
384  pkt->flags &= ~AV_PKT_FLAG_CORRUPT;
385  pkt->stream_index = 0;
386 
387  if (ret >= ID3v1_TAG_SIZE &&
388  memcmp(&pkt->data[ret - ID3v1_TAG_SIZE], "TAG", 3) == 0)
389  ret -= ID3v1_TAG_SIZE;
390 
391  /* note: we need to modify the packet size here to handle the last
392  packet */
393  pkt->size = ret;
394  return ret;
395 }
396 
397 static int check(AVFormatContext *s, int64_t pos)
398 {
399  int64_t ret = avio_seek(s->pb, pos, SEEK_SET);
400  unsigned header;
401  MPADecodeHeader sd;
402  if (ret < 0)
403  return ret;
404  header = avio_rb32(s->pb);
405  if (ff_mpa_check_header(header) < 0)
406  return -1;
407  if (avpriv_mpegaudio_decode_header(&sd, header) == 1)
408  return -1;
409  return sd.frame_size;
410 }
411 
412 static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp,
413  int flags)
414 {
415  MP3DecContext *mp3 = s->priv_data;
416  AVIndexEntry *ie, ie1;
417  AVStream *st = s->streams[0];
418  int64_t ret = av_index_search_timestamp(st, timestamp, flags);
419  int i, j;
420  int dir = (flags&AVSEEK_FLAG_BACKWARD) ? -1 : 1;
421  int64_t best_pos;
422  int best_score;
423 
424  if ( mp3->is_cbr
425  && st->duration > 0
426  && mp3->header_filesize > s->data_offset
427  && mp3->frames) {
428  int64_t filesize = avio_size(s->pb);
429  int64_t duration;
430  if (filesize <= s->data_offset)
431  filesize = mp3->header_filesize;
432  filesize -= s->data_offset;
433  duration = av_rescale(st->duration, filesize, mp3->header_filesize - s->data_offset);
434  ie = &ie1;
435  timestamp = av_clip64(timestamp, 0, duration);
436  ie->timestamp = timestamp;
437  ie->pos = av_rescale(timestamp, filesize, duration) + s->data_offset;
438  } else if (mp3->xing_toc) {
439  if (ret < 0)
440  return ret;
441 
442  ie = &st->index_entries[ret];
443  } else {
444  st->skip_samples = timestamp <= 0 ? mp3->start_pad + 528 + 1 : 0;
445 
446  return -1;
447  }
448 
449  avio_seek(s->pb, FFMAX(ie->pos - 4096, 0), SEEK_SET);
450  ret = avio_seek(s->pb, ie->pos, SEEK_SET);
451  if (ret < 0)
452  return ret;
453 
454 #define MIN_VALID 3
455  best_pos = ie->pos;
456  best_score = 999;
457  for(i=0; i<4096; i++) {
458  int64_t pos = ie->pos + (dir > 0 ? i - 1024 : -i);
459  int64_t candidate = -1;
460  int score = 999;
461 
462  if (pos < 0)
463  continue;
464 
465  for(j=0; j<MIN_VALID; j++) {
466  ret = check(s, pos);
467  if(ret < 0)
468  break;
469  if ((ie->pos - pos)*dir <= 0 && abs(MIN_VALID/2-j) < score) {
470  candidate = pos;
471  score = abs(MIN_VALID/2-j);
472  }
473  pos += ret;
474  }
475  if (best_score > score && j == MIN_VALID) {
476  best_pos = candidate;
477  best_score = score;
478  if(score == 0)
479  break;
480  }
481  }
482 
483  ret = avio_seek(s->pb, best_pos, SEEK_SET);
484  if (ret < 0)
485  return ret;
486  ff_update_cur_dts(s, st, ie->timestamp);
487  st->skip_samples = ie->timestamp <= 0 ? mp3->start_pad + 528 + 1 : 0;
488  return 0;
489 }
490 
491 static const AVOption options[] = {
492  { "usetoc", "use table of contents", offsetof(MP3DecContext, usetoc), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, AV_OPT_FLAG_DECODING_PARAM},
493  { NULL },
494 };
495 
496 static const AVClass demuxer_class = {
497  .class_name = "mp3",
498  .item_name = av_default_item_name,
499  .option = options,
500  .version = LIBAVUTIL_VERSION_INT,
501  .category = AV_CLASS_CATEGORY_DEMUXER,
502 };
503 
505  .name = "mp3",
506  .long_name = NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"),
507  .read_probe = mp3_read_probe,
508  .read_header = mp3_read_header,
509  .read_packet = mp3_read_packet,
510  .read_seek = mp3_seek,
511  .priv_data_size = sizeof(MP3DecContext),
513  .extensions = "mp2,mp3,m2a,mpa", /* XXX: use probe */
514  .priv_class = &demuxer_class,
515 };