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 
41 #define XING_TOC_COUNT 100
42 
43 typedef struct {
44  AVClass *class;
45  int64_t filesize;
46  int xing_toc;
47  int start_pad;
48  int end_pad;
49  int usetoc;
50  unsigned frames; /* Total number of frames in file */
51  unsigned header_filesize; /* Total number of bytes in the stream */
52  int is_cbr;
54 
55 /* mp3 read */
56 
58 {
59  int max_frames, first_frames = 0;
60  int fsize, frames, sample_rate;
61  uint32_t header;
62  const uint8_t *buf, *buf0, *buf2, *end;
63  AVCodecContext avctx;
64 
65  buf0 = p->buf;
66  end = p->buf + p->buf_size - sizeof(uint32_t);
67  while(buf0 < end && !*buf0)
68  buf0++;
69 
70  max_frames = 0;
71  buf = buf0;
72 
73  for(; buf < end; buf= buf2+1) {
74  buf2 = buf;
75  if(ff_mpa_check_header(AV_RB32(buf2)))
76  continue;
77 
78  for(frames = 0; buf2 < end; frames++) {
79  header = AV_RB32(buf2);
80  fsize = avpriv_mpa_decode_header(&avctx, header, &sample_rate, &sample_rate, &sample_rate, &sample_rate);
81  if(fsize < 0)
82  break;
83  buf2 += fsize;
84  }
85  max_frames = FFMAX(max_frames, frames);
86  if(buf == buf0)
87  first_frames= frames;
88  }
89  // keep this in sync with ac3 probe, both need to avoid
90  // issues with MPEG-files!
91  if (first_frames>=4) return AVPROBE_SCORE_EXTENSION + 1;
92  else if(max_frames>200)return AVPROBE_SCORE_EXTENSION;
93  else if(max_frames>=4 && max_frames >= p->buf_size/10000) return AVPROBE_SCORE_EXTENSION / 2;
94  else if(ff_id3v2_match(buf0, ID3v2_DEFAULT_MAGIC) && 2*ff_id3v2_tag_len(buf0) >= p->buf_size)
96  else if(max_frames>=1 && max_frames >= p->buf_size/10000) return 1;
97  else return 0;
98 //mpegps_mp3_unrecognized_format.mpg has max_frames=3
99 }
100 
101 static void read_xing_toc(AVFormatContext *s, int64_t filesize, int64_t duration)
102 {
103  int i;
104  MP3DecContext *mp3 = s->priv_data;
105  int fill_index = mp3->usetoc && duration > 0;
106 
107  if (!filesize &&
108  !(filesize = avio_size(s->pb))) {
109  av_log(s, AV_LOG_WARNING, "Cannot determine file size, skipping TOC table.\n");
110  fill_index = 0;
111  }
112 
113  for (i = 0; i < XING_TOC_COUNT; i++) {
114  uint8_t b = avio_r8(s->pb);
115  if (fill_index)
117  av_rescale(b, filesize, 256),
118  av_rescale(i, duration, XING_TOC_COUNT),
119  0, 0, AVINDEX_KEYFRAME);
120  }
121  if (fill_index)
122  mp3->xing_toc = 1;
123 }
124 
126  MPADecodeHeader *c, uint32_t spf)
127 {
128 #define LAST_BITS(k, n) ((k) & ((1 << (n)) - 1))
129 #define MIDDLE_BITS(k, m, n) LAST_BITS((k) >> (m), ((n) - (m)))
130 
131  uint16_t crc;
132  uint32_t v;
133 
134  char version[10];
135 
136  uint32_t peak = 0;
137  int32_t r_gain = INT32_MIN, a_gain = INT32_MIN;
138 
139  MP3DecContext *mp3 = s->priv_data;
140  static const int64_t xing_offtbl[2][2] = {{32, 17}, {17,9}};
141  uint64_t fsize = avio_size(s->pb);
142 
143  /* Check for Xing / Info tag */
144  avio_skip(s->pb, xing_offtbl[c->lsf == 1][c->nb_channels == 1]);
145  v = avio_rb32(s->pb);
146  mp3->is_cbr = v == MKBETAG('I', 'n', 'f', 'o');
147  if (v != MKBETAG('X', 'i', 'n', 'g') && !mp3->is_cbr)
148  return;
149 
150  v = avio_rb32(s->pb);
151  if (v & XING_FLAG_FRAMES)
152  mp3->frames = avio_rb32(s->pb);
153  if (v & XING_FLAG_SIZE)
154  mp3->header_filesize = avio_rb32(s->pb);
155  if (fsize && mp3->header_filesize) {
156  uint64_t min, delta;
157  min = FFMIN(fsize, mp3->header_filesize);
158  delta = FFMAX(fsize, mp3->header_filesize) - min;
159  if (fsize > mp3->header_filesize && delta > min >> 4) {
160  mp3->frames = 0;
161  } else if (delta > min >> 4) {
163  "filesize and duration do not match (growing file?)\n");
164  }
165  }
166  if (v & XING_FLAG_TOC)
168  (AVRational){spf, c->sample_rate},
169  st->time_base));
170  /* VBR quality */
171  if(v & 8)
172  avio_skip(s->pb, 4);
173 
174  /* Encoder short version string */
175  memset(version, 0, sizeof(version));
176  avio_read(s->pb, version, 9);
177 
178  /* Info Tag revision + VBR method */
179  avio_r8(s->pb);
180 
181  /* Lowpass filter value */
182  avio_r8(s->pb);
183 
184  /* ReplayGain peak */
185  v = avio_rb32(s->pb);
186  peak = av_rescale(v, 100000, 1 << 23);
187 
188  /* Radio ReplayGain */
189  v = avio_rb16(s->pb);
190 
191  if (MIDDLE_BITS(v, 13, 15) == 1) {
192  r_gain = MIDDLE_BITS(v, 0, 8) * 10000;
193 
194  if (v & (1 << 9))
195  r_gain *= -1;
196  }
197 
198  /* Audiophile ReplayGain */
199  v = avio_rb16(s->pb);
200 
201  if (MIDDLE_BITS(v, 13, 15) == 2) {
202  a_gain = MIDDLE_BITS(v, 0, 8) * 10000;
203 
204  if (v & (1 << 9))
205  a_gain *= -1;
206  }
207 
208  /* Encoding flags + ATH Type */
209  avio_r8(s->pb);
210 
211  /* if ABR {specified bitrate} else {minimal bitrate} */
212  avio_r8(s->pb);
213 
214  /* Encoder delays */
215  v= avio_rb24(s->pb);
216  if(AV_RB32(version) == MKBETAG('L', 'A', 'M', 'E')
217  || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'f')) {
218 
219  mp3->start_pad = v>>12;
220  mp3-> end_pad = v&4095;
221  st->skip_samples = mp3->start_pad + 528 + 1;
222  if (mp3->frames) {
223  st->first_discard_sample = -mp3->end_pad + 528 + 1 + mp3->frames * (int64_t)spf;
224  st->last_discard_sample = mp3->frames * (int64_t)spf;
225  }
226  if (!st->start_time)
227  st->start_time = av_rescale_q(st->skip_samples,
228  (AVRational){1, c->sample_rate},
229  st->time_base);
230  av_log(s, AV_LOG_DEBUG, "pad %d %d\n", mp3->start_pad, mp3-> end_pad);
231  }
232 
233  /* Misc */
234  avio_r8(s->pb);
235 
236  /* MP3 gain */
237  avio_r8(s->pb);
238 
239  /* Preset and surround info */
240  avio_rb16(s->pb);
241 
242  /* Music length */
243  avio_rb32(s->pb);
244 
245  /* Music CRC */
246  avio_rb16(s->pb);
247 
248  /* Info Tag CRC */
249  crc = ffio_get_checksum(s->pb);
250  v = avio_rb16(s->pb);
251 
252  if (v == crc) {
253  ff_replaygain_export_raw(st, r_gain, peak, a_gain, 0);
254  av_dict_set(&st->metadata, "encoder", version, 0);
255  }
256 }
257 
258 static void mp3_parse_vbri_tag(AVFormatContext *s, AVStream *st, int64_t base)
259 {
260  uint32_t v;
261  MP3DecContext *mp3 = s->priv_data;
262 
263  /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */
264  avio_seek(s->pb, base + 4 + 32, SEEK_SET);
265  v = avio_rb32(s->pb);
266  if (v == MKBETAG('V', 'B', 'R', 'I')) {
267  /* Check tag version */
268  if (avio_rb16(s->pb) == 1) {
269  /* skip delay and quality */
270  avio_skip(s->pb, 4);
271  mp3->header_filesize = avio_rb32(s->pb);
272  mp3->frames = avio_rb32(s->pb);
273  }
274  }
275 }
276 
277 /**
278  * Try to find Xing/Info/VBRI tags and compute duration from info therein
279  */
280 static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
281 {
282  uint32_t v, spf;
284  int vbrtag_size = 0;
285  MP3DecContext *mp3 = s->priv_data;
286 
288 
289  v = avio_rb32(s->pb);
290  if(ff_mpa_check_header(v) < 0)
291  return -1;
292 
293  if (avpriv_mpegaudio_decode_header(&c, v) == 0)
294  vbrtag_size = c.frame_size;
295  if(c.layer != 3)
296  return -1;
297 
298  spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
299 
300  mp3->frames = 0;
301  mp3->header_filesize = 0;
302 
303  mp3_parse_info_tag(s, st, &c, spf);
304  mp3_parse_vbri_tag(s, st, base);
305 
306  if (!mp3->frames && !mp3->header_filesize)
307  return -1;
308 
309  /* Skip the vbr tag frame */
310  avio_seek(s->pb, base + vbrtag_size, SEEK_SET);
311 
312  if (mp3->frames)
313  st->duration = av_rescale_q(mp3->frames, (AVRational){spf, c.sample_rate},
314  st->time_base);
315  if (mp3->header_filesize && mp3->frames && !mp3->is_cbr)
316  st->codec->bit_rate = av_rescale(mp3->header_filesize, 8 * c.sample_rate, mp3->frames * (int64_t)spf);
317 
318  return 0;
319 }
320 
322 {
323  MP3DecContext *mp3 = s->priv_data;
324  AVStream *st;
325  int64_t off;
326  int ret;
327 
328  st = avformat_new_stream(s, NULL);
329  if (!st)
330  return AVERROR(ENOMEM);
331 
335  st->start_time = 0;
336 
337  // lcm of all mp3 sample rates
338  avpriv_set_pts_info(st, 64, 1, 14112000);
339 
340  s->pb->maxsize = -1;
341  off = avio_tell(s->pb);
342 
343  if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX))
344  ff_id3v1_read(s);
345 
346  if(s->pb->seekable)
347  mp3->filesize = avio_size(s->pb);
348 
349  if (mp3_parse_vbr_tags(s, st, off) < 0)
350  avio_seek(s->pb, off, SEEK_SET);
351 
352  ret = ff_replaygain_export(st, s->metadata);
353  if (ret < 0)
354  return ret;
355 
356  /* the parameters will be extracted from the compressed bitstream */
357  return 0;
358 }
359 
360 #define MP3_PACKET_SIZE 1024
361 
363 {
364  MP3DecContext *mp3 = s->priv_data;
365  int ret, size;
366  int64_t pos;
367 
368  size= MP3_PACKET_SIZE;
369  pos = avio_tell(s->pb);
370  if(mp3->filesize > ID3v1_TAG_SIZE && pos < mp3->filesize)
371  size= FFMIN(size, mp3->filesize - pos);
372 
373  ret= av_get_packet(s->pb, pkt, size);
374  if (ret <= 0) {
375  if(ret<0)
376  return ret;
377  return AVERROR_EOF;
378  }
379 
380  pkt->flags &= ~AV_PKT_FLAG_CORRUPT;
381  pkt->stream_index = 0;
382 
383  if (ret >= ID3v1_TAG_SIZE &&
384  memcmp(&pkt->data[ret - ID3v1_TAG_SIZE], "TAG", 3) == 0)
385  ret -= ID3v1_TAG_SIZE;
386 
387  /* note: we need to modify the packet size here to handle the last
388  packet */
389  pkt->size = ret;
390  return ret;
391 }
392 
393 static int check(AVFormatContext *s, int64_t pos)
394 {
395  int64_t ret = avio_seek(s->pb, pos, SEEK_SET);
396  unsigned header;
397  MPADecodeHeader sd;
398  if (ret < 0)
399  return ret;
400  header = avio_rb32(s->pb);
401  if (ff_mpa_check_header(header) < 0)
402  return -1;
403  if (avpriv_mpegaudio_decode_header(&sd, header) == 1)
404  return -1;
405  return sd.frame_size;
406 }
407 
408 static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp,
409  int flags)
410 {
411  MP3DecContext *mp3 = s->priv_data;
412  AVIndexEntry *ie, ie1;
413  AVStream *st = s->streams[0];
414  int64_t ret = av_index_search_timestamp(st, timestamp, flags);
415  int i, j;
416  int dir = (flags&AVSEEK_FLAG_BACKWARD) ? -1 : 1;
417  int64_t best_pos;
418  int best_score;
419 
420  if ( mp3->is_cbr
421  && st->duration > 0
422  && mp3->header_filesize > s->data_offset
423  && mp3->frames) {
424  int64_t filesize = avio_size(s->pb);
425  int64_t duration;
426  if (filesize <= s->data_offset)
427  filesize = mp3->header_filesize;
428  filesize -= s->data_offset;
429  duration = av_rescale(st->duration, filesize, mp3->header_filesize - s->data_offset);
430  ie = &ie1;
431  timestamp = av_clip64(timestamp, 0, duration);
432  ie->timestamp = timestamp;
433  ie->pos = av_rescale(timestamp, filesize, duration) + s->data_offset;
434  } else if (mp3->xing_toc) {
435  if (ret < 0)
436  return ret;
437 
438  ie = &st->index_entries[ret];
439  } else {
440  st->skip_samples = timestamp <= 0 ? mp3->start_pad + 528 + 1 : 0;
441 
442  return -1;
443  }
444 
445  avio_seek(s->pb, FFMAX(ie->pos - 4096, 0), SEEK_SET);
446  ret = avio_seek(s->pb, ie->pos, SEEK_SET);
447  if (ret < 0)
448  return ret;
449 
450 #define MIN_VALID 3
451  best_pos = ie->pos;
452  best_score = 999;
453  for(i=0; i<4096; i++) {
454  int64_t pos = ie->pos + (dir > 0 ? i - 1024 : -i);
455  int64_t candidate = -1;
456  int score = 999;
457  for(j=0; j<MIN_VALID; j++) {
458  ret = check(s, pos);
459  if(ret < 0)
460  break;
461  if ((ie->pos - pos)*dir <= 0 && abs(MIN_VALID/2-j) < score) {
462  candidate = pos;
463  score = abs(MIN_VALID/2-j);
464  }
465  pos += ret;
466  }
467  if (best_score > score && j == MIN_VALID) {
468  best_pos = candidate;
469  best_score = score;
470  if(score == 0)
471  break;
472  }
473  }
474 
475  ret = avio_seek(s->pb, best_pos, SEEK_SET);
476  if (ret < 0)
477  return ret;
478  ff_update_cur_dts(s, st, ie->timestamp);
479  st->skip_samples = ie->timestamp <= 0 ? mp3->start_pad + 528 + 1 : 0;
480  return 0;
481 }
482 
483 static const AVOption options[] = {
484  { "usetoc", "use table of contents", offsetof(MP3DecContext, usetoc), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, AV_OPT_FLAG_DECODING_PARAM},
485  { NULL },
486 };
487 
488 static const AVClass demuxer_class = {
489  .class_name = "mp3",
490  .item_name = av_default_item_name,
491  .option = options,
492  .version = LIBAVUTIL_VERSION_INT,
493  .category = AV_CLASS_CATEGORY_DEMUXER,
494 };
495 
497  .name = "mp3",
498  .long_name = NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"),
499  .read_probe = mp3_read_probe,
500  .read_header = mp3_read_header,
501  .read_packet = mp3_read_packet,
502  .read_seek = mp3_seek,
503  .priv_data_size = sizeof(MP3DecContext),
505  .extensions = "mp2,mp3,m2a,mpa", /* XXX: use probe */
506  .priv_class = &demuxer_class,
507 };