FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
oggparsevorbis.c
Go to the documentation of this file.
1 /*
2  * Copyright (C) 2005 Michael Ahlberg, Måns Rullgård
3  *
4  * Permission is hereby granted, free of charge, to any person
5  * obtaining a copy of this software and associated documentation
6  * files (the "Software"), to deal in the Software without
7  * restriction, including without limitation the rights to use, copy,
8  * modify, merge, publish, distribute, sublicense, and/or sell copies
9  * of the Software, and to permit persons to whom the Software is
10  * furnished to do so, subject to the following conditions:
11  *
12  * The above copyright notice and this permission notice shall be
13  * included in all copies or substantial portions of the Software.
14  *
15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
16  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
17  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
18  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
19  * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
20  * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
22  * DEALINGS IN THE SOFTWARE.
23  */
24 
25 #include <stdlib.h>
26 
27 #include "libavutil/avstring.h"
28 #include "libavutil/base64.h"
29 #include "libavutil/bswap.h"
30 #include "libavutil/dict.h"
31 #include "libavcodec/bytestream.h"
32 #include "libavcodec/get_bits.h"
34 #include "avformat.h"
35 #include "flac_picture.h"
36 #include "internal.h"
37 #include "oggdec.h"
38 #include "vorbiscomment.h"
39 #include "replaygain.h"
40 
42 {
43  int i, cnum, h, m, s, ms, keylen = strlen(key);
44  AVChapter *chapter = NULL;
45 
46  if (keylen < 9 || sscanf(key, "CHAPTER%03d", &cnum) != 1)
47  return 0;
48 
49  if (keylen <= 10) {
50  if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4)
51  return 0;
52 
53  avpriv_new_chapter(as, cnum, (AVRational) { 1, 1000 },
54  ms + 1000 * (s + 60 * (m + 60 * h)),
56  av_free(val);
57  } else if (!strcmp(key + keylen - 4, "NAME")) {
58  for (i = 0; i < as->nb_chapters; i++)
59  if (as->chapters[i]->id == cnum) {
60  chapter = as->chapters[i];
61  break;
62  }
63  if (!chapter)
64  return 0;
65 
66  av_dict_set(&chapter->metadata, "title", val, AV_DICT_DONT_STRDUP_VAL);
67  } else
68  return 0;
69 
70  av_free(key);
71  return 1;
72 }
73 
75  const uint8_t *buf, int size)
76 {
77  int updates = ff_vorbis_comment(as, &st->metadata, buf, size, 1);
78 
79  if (updates > 0) {
81  }
82 
83  return updates;
84 }
85 
87  const uint8_t *buf, int size,
88  int parse_picture)
89 {
90  const uint8_t *p = buf;
91  const uint8_t *end = buf + size;
92  int updates = 0;
93  unsigned n, j;
94  int s;
95 
96  /* must have vendor_length and user_comment_list_length */
97  if (size < 8)
98  return AVERROR_INVALIDDATA;
99 
100  s = bytestream_get_le32(&p);
101 
102  if (end - p - 4 < s || s < 0)
103  return AVERROR_INVALIDDATA;
104 
105  p += s;
106 
107  n = bytestream_get_le32(&p);
108 
109  while (end - p >= 4 && n > 0) {
110  const char *t, *v;
111  int tl, vl;
112 
113  s = bytestream_get_le32(&p);
114 
115  if (end - p < s || s < 0)
116  break;
117 
118  t = p;
119  p += s;
120  n--;
121 
122  v = memchr(t, '=', s);
123  if (!v)
124  continue;
125 
126  tl = v - t;
127  vl = s - tl - 1;
128  v++;
129 
130  if (tl && vl) {
131  char *tt, *ct;
132 
133  tt = av_malloc(tl + 1);
134  ct = av_malloc(vl + 1);
135  if (!tt || !ct) {
136  av_freep(&tt);
137  av_freep(&ct);
138  return AVERROR(ENOMEM);
139  }
140 
141  for (j = 0; j < tl; j++)
142  tt[j] = av_toupper(t[j]);
143  tt[tl] = 0;
144 
145  memcpy(ct, v, vl);
146  ct[vl] = 0;
147 
148  /* The format in which the pictures are stored is the FLAC format.
149  * Xiph says: "The binary FLAC picture structure is base64 encoded
150  * and placed within a VorbisComment with the tag name
151  * 'METADATA_BLOCK_PICTURE'. This is the preferred and
152  * recommended way of embedding cover art within VorbisComments."
153  */
154  if (!strcmp(tt, "METADATA_BLOCK_PICTURE") && parse_picture) {
155  int ret;
156  char *pict = av_malloc(vl);
157 
158  if (!pict) {
159  av_log(as, AV_LOG_WARNING, "out-of-memory error. Skipping cover art block.\n");
160  av_freep(&tt);
161  av_freep(&ct);
162  continue;
163  }
164  if ((ret = av_base64_decode(pict, ct, vl)) > 0)
165  ret = ff_flac_parse_picture(as, pict, ret);
166  av_freep(&tt);
167  av_freep(&ct);
168  av_freep(&pict);
169  if (ret < 0) {
170  av_log(as, AV_LOG_WARNING, "Failed to parse cover art block.\n");
171  continue;
172  }
173  } else if (!ogm_chapter(as, tt, ct)) {
174  updates++;
175  if (av_dict_get(*m, tt, NULL, 0)) {
176  av_dict_set(m, tt, ";", AV_DICT_APPEND);
177  }
178  av_dict_set(m, tt, ct,
181  av_freep(&ct);
182  }
183  }
184  }
185 
186  if (p != end)
187  av_log(as, AV_LOG_INFO,
188  "%"PTRDIFF_SPECIFIER" bytes of comment header remain\n", end - p);
189  if (n > 0)
190  av_log(as, AV_LOG_INFO,
191  "truncated comment header, %i comments not found\n", n);
192 
194 
195  return updates;
196 }
197 
198 /*
199  * Parse the vorbis header
200  *
201  * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec
202  * [vorbis_version] = read 32 bits as unsigned integer | Not used
203  * [audio_channels] = read 8 bit integer as unsigned | Used
204  * [audio_sample_rate] = read 32 bits as unsigned integer | Used
205  * [bitrate_maximum] = read 32 bits as signed integer | Not used yet
206  * [bitrate_nominal] = read 32 bits as signed integer | Not used yet
207  * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate
208  * [blocksize_0] = read 4 bits as unsigned integer | Not Used
209  * [blocksize_1] = read 4 bits as unsigned integer | Not Used
210  * [framing_flag] = read one bit | Not Used
211  */
212 
214  unsigned int len[3];
215  unsigned char *packet[3];
217  int64_t final_pts;
219 };
220 
222  struct oggvorbis_private *priv,
223  uint8_t **buf)
224 {
225  int i, offset, len, err;
226  int buf_len;
227  unsigned char *ptr;
228 
229  len = priv->len[0] + priv->len[1] + priv->len[2];
230  buf_len = len + len / 255 + 64;
231  ptr = *buf = av_realloc(NULL, buf_len);
232  if (!ptr)
233  return AVERROR(ENOMEM);
234  memset(*buf, '\0', buf_len);
235 
236  ptr[0] = 2;
237  offset = 1;
238  offset += av_xiphlacing(&ptr[offset], priv->len[0]);
239  offset += av_xiphlacing(&ptr[offset], priv->len[1]);
240  for (i = 0; i < 3; i++) {
241  memcpy(&ptr[offset], priv->packet[i], priv->len[i]);
242  offset += priv->len[i];
243  av_freep(&priv->packet[i]);
244  }
245  if ((err = av_reallocp(buf, offset + FF_INPUT_BUFFER_PADDING_SIZE)) < 0)
246  return err;
247  return offset;
248 }
249 
250 static void vorbis_cleanup(AVFormatContext *s, int idx)
251 {
252  struct ogg *ogg = s->priv_data;
253  struct ogg_stream *os = ogg->streams + idx;
254  struct oggvorbis_private *priv = os->private;
255  int i;
256  if (os->private) {
257  av_vorbis_parse_free(&priv->vp);
258  for (i = 0; i < 3; i++)
259  av_freep(&priv->packet[i]);
260  }
261 }
262 
264 {
265  struct ogg *ogg = s->priv_data;
266  struct ogg_stream *os = ogg->streams + idx;
267  AVStream *st = s->streams[idx];
268  int ret;
269 
270  if (os->psize <= 8)
271  return 0;
272 
273  /* New metadata packet; release old data. */
274  av_dict_free(&st->metadata);
275  ret = ff_vorbis_stream_comment(s, st, os->buf + os->pstart + 7,
276  os->psize - 8);
277  if (ret < 0)
278  return ret;
279 
280  /* Update the metadata if possible. */
281  av_freep(&os->new_metadata);
282  if (st->metadata) {
284  /* Send an empty dictionary to indicate that metadata has been cleared. */
285  } else {
286  os->new_metadata = av_malloc(1);
287  os->new_metadata_size = 0;
288  }
289 
290  return ret;
291 }
292 
293 static int vorbis_header(AVFormatContext *s, int idx)
294 {
295  struct ogg *ogg = s->priv_data;
296  AVStream *st = s->streams[idx];
297  struct ogg_stream *os = ogg->streams + idx;
298  struct oggvorbis_private *priv;
299  int pkt_type = os->buf[os->pstart];
300 
301  if (!os->private) {
302  os->private = av_mallocz(sizeof(struct oggvorbis_private));
303  if (!os->private)
304  return AVERROR(ENOMEM);
305  }
306 
307  priv = os->private;
308 
309  if (!(pkt_type & 1))
310  return priv->vp ? 0 : AVERROR_INVALIDDATA;
311 
312  if (os->psize < 1 || pkt_type > 5)
313  return AVERROR_INVALIDDATA;
314 
315  if (priv->packet[pkt_type >> 1])
316  return AVERROR_INVALIDDATA;
317  if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1])
318  return AVERROR_INVALIDDATA;
319 
320  priv->len[pkt_type >> 1] = os->psize;
321  priv->packet[pkt_type >> 1] = av_mallocz(os->psize);
322  if (!priv->packet[pkt_type >> 1])
323  return AVERROR(ENOMEM);
324  memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize);
325  if (os->buf[os->pstart] == 1) {
326  const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
327  unsigned blocksize, bs0, bs1;
328  int srate;
329  int channels;
330 
331  if (os->psize != 30)
332  return AVERROR_INVALIDDATA;
333 
334  if (bytestream_get_le32(&p) != 0) /* vorbis_version */
335  return AVERROR_INVALIDDATA;
336 
337  channels = bytestream_get_byte(&p);
338  if (st->codec->channels && channels != st->codec->channels) {
339  av_log(s, AV_LOG_ERROR, "Channel change is not supported\n");
340  return AVERROR_PATCHWELCOME;
341  }
342  st->codec->channels = channels;
343  srate = bytestream_get_le32(&p);
344  p += 4; // skip maximum bitrate
345  st->codec->bit_rate = bytestream_get_le32(&p); // nominal bitrate
346  p += 4; // skip minimum bitrate
347 
348  blocksize = bytestream_get_byte(&p);
349  bs0 = blocksize & 15;
350  bs1 = blocksize >> 4;
351 
352  if (bs0 > bs1)
353  return AVERROR_INVALIDDATA;
354  if (bs0 < 6 || bs1 > 13)
355  return AVERROR_INVALIDDATA;
356 
357  if (bytestream_get_byte(&p) != 1) /* framing_flag */
358  return AVERROR_INVALIDDATA;
359 
362 
363  if (srate > 0) {
364  st->codec->sample_rate = srate;
365  avpriv_set_pts_info(st, 64, 1, srate);
366  }
367  } else if (os->buf[os->pstart] == 3) {
368  if (vorbis_update_metadata(s, idx) >= 0 && priv->len[1] > 10) {
369  unsigned new_len;
370 
371  int ret = ff_replaygain_export(st, st->metadata);
372  if (ret < 0)
373  return ret;
374 
375  // drop all metadata we parsed and which is not required by libvorbis
376  new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
377  if (new_len >= 16 && new_len < os->psize) {
378  AV_WL32(priv->packet[1] + new_len - 5, 0);
379  priv->packet[1][new_len - 1] = 1;
380  priv->len[1] = new_len;
381  }
382  }
383  } else {
384  int ret = fixup_vorbis_headers(s, priv, &st->codec->extradata);
385  if (ret < 0) {
386  st->codec->extradata_size = 0;
387  return ret;
388  }
389  st->codec->extradata_size = ret;
390 
392  if (!priv->vp) {
393  av_freep(&st->codec->extradata);
394  st->codec->extradata_size = 0;
395  return AVERROR_UNKNOWN;
396  }
397  }
398 
399  return 1;
400 }
401 
402 static int vorbis_packet(AVFormatContext *s, int idx)
403 {
404  struct ogg *ogg = s->priv_data;
405  struct ogg_stream *os = ogg->streams + idx;
406  struct oggvorbis_private *priv = os->private;
407  int duration, flags = 0;
408 
409  /* first packet handling
410  * here we parse the duration of each packet in the first page and compare
411  * the total duration to the page granule to find the encoder delay and
412  * set the first timestamp */
413  if ((!os->lastpts || os->lastpts == AV_NOPTS_VALUE) && !(os->flags & OGG_FLAG_EOS) && (int64_t)os->granule>=0) {
414  int seg, d;
415  uint8_t *last_pkt = os->buf + os->pstart;
416  uint8_t *next_pkt = last_pkt;
417 
418  av_vorbis_parse_reset(priv->vp);
419  duration = 0;
420  seg = os->segp;
421  d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags);
422  if (d < 0) {
424  return 0;
425  } else if (flags & VORBIS_FLAG_COMMENT) {
426  vorbis_update_metadata(s, idx);
427  flags = 0;
428  }
429  duration += d;
430  last_pkt = next_pkt = next_pkt + os->psize;
431  for (; seg < os->nsegs; seg++) {
432  if (os->segments[seg] < 255) {
433  int d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags);
434  if (d < 0) {
435  duration = os->granule;
436  break;
437  } else if (flags & VORBIS_FLAG_COMMENT) {
438  vorbis_update_metadata(s, idx);
439  flags = 0;
440  }
441  duration += d;
442  last_pkt = next_pkt + os->segments[seg];
443  }
444  next_pkt += os->segments[seg];
445  }
446  os->lastpts =
447  os->lastdts = os->granule - duration;
448 
449  if (!os->granule && duration) //hack to deal with broken files (Ticket3710)
450  os->lastpts = os->lastdts = AV_NOPTS_VALUE;
451 
452  if (s->streams[idx]->start_time == AV_NOPTS_VALUE) {
453  s->streams[idx]->start_time = FFMAX(os->lastpts, 0);
454  if (s->streams[idx]->duration != AV_NOPTS_VALUE)
455  s->streams[idx]->duration -= s->streams[idx]->start_time;
456  }
457  priv->final_pts = AV_NOPTS_VALUE;
458  av_vorbis_parse_reset(priv->vp);
459  }
460 
461  /* parse packet duration */
462  if (os->psize > 0) {
463  duration = av_vorbis_parse_frame_flags(priv->vp, os->buf + os->pstart, 1, &flags);
464  if (duration < 0) {
466  return 0;
467  } else if (flags & VORBIS_FLAG_COMMENT) {
468  vorbis_update_metadata(s, idx);
469  flags = 0;
470  }
471  os->pduration = duration;
472  }
473 
474  /* final packet handling
475  * here we save the pts of the first packet in the final page, sum up all
476  * packet durations in the final page except for the last one, and compare
477  * to the page granule to find the duration of the final packet */
478  if (os->flags & OGG_FLAG_EOS) {
479  if (os->lastpts != AV_NOPTS_VALUE) {
480  priv->final_pts = os->lastpts;
481  priv->final_duration = 0;
482  }
483  if (os->segp == os->nsegs)
484  os->pduration = os->granule - priv->final_pts - priv->final_duration;
485  priv->final_duration += os->pduration;
486  }
487 
488  return 0;
489 }
490 
491 const struct ogg_codec ff_vorbis_codec = {
492  .magic = "\001vorbis",
493  .magicsize = 7,
494  .header = vorbis_header,
495  .packet = vorbis_packet,
496  .cleanup = vorbis_cleanup,
497  .nb_header = 3,
498 };