Libav
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 
41 static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
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_freep(&tt);
160  av_freep(&ct);
161  return AVERROR(ENOMEM);
162  }
163  if ((ret = av_base64_decode(pict, ct, vl)) > 0)
164  ret = ff_flac_parse_picture(as, pict, ret);
165  av_freep(&tt);
166  av_freep(&ct);
167  av_freep(&pict);
168  if (ret < 0) {
169  av_log(as, AV_LOG_WARNING, "Failed to parse cover art block.\n");
170  continue;
171  }
172  } else if (!ogm_chapter(as, tt, ct)) {
173  updates++;
174  av_dict_set(m, tt, ct,
177  }
178  }
179  }
180 
181  if (p != end)
182  av_log(as, AV_LOG_INFO,
183  "%ti bytes of comment header remain\n", end - p);
184  if (n > 0)
185  av_log(as, AV_LOG_INFO,
186  "truncated comment header, %i comments not found\n", n);
187 
189 
190  return updates;
191 }
192 
193 /*
194  * Parse the vorbis header
195  *
196  * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec
197  * [vorbis_version] = read 32 bits as unsigned integer | Not used
198  * [audio_channels] = read 8 bit integer as unsigned | Used
199  * [audio_sample_rate] = read 32 bits as unsigned integer | Used
200  * [bitrate_maximum] = read 32 bits as signed integer | Not used yet
201  * [bitrate_nominal] = read 32 bits as signed integer | Not used yet
202  * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate
203  * [blocksize_0] = read 4 bits as unsigned integer | Not Used
204  * [blocksize_1] = read 4 bits as unsigned integer | Not Used
205  * [framing_flag] = read one bit | Not Used
206  */
207 
209  unsigned int len[3];
210  unsigned char *packet[3];
212  int64_t final_pts;
214 };
215 
217  struct oggvorbis_private *priv,
218  uint8_t **buf)
219 {
220  int i, offset, len, err;
221  unsigned char *ptr;
222 
223  len = priv->len[0] + priv->len[1] + priv->len[2];
224  ptr = *buf = av_mallocz(len + len / 255 + 64);
225  if (!ptr)
226  return AVERROR(ENOMEM);
227 
228  ptr[0] = 2;
229  offset = 1;
230  offset += av_xiphlacing(&ptr[offset], priv->len[0]);
231  offset += av_xiphlacing(&ptr[offset], priv->len[1]);
232  for (i = 0; i < 3; i++) {
233  memcpy(&ptr[offset], priv->packet[i], priv->len[i]);
234  offset += priv->len[i];
235  av_freep(&priv->packet[i]);
236  }
237  if ((err = av_reallocp(buf, offset + FF_INPUT_BUFFER_PADDING_SIZE)) < 0)
238  return err;
239  return offset;
240 }
241 
242 static void vorbis_cleanup(AVFormatContext *s, int idx)
243 {
244  struct ogg *ogg = s->priv_data;
245  struct ogg_stream *os = ogg->streams + idx;
246  struct oggvorbis_private *priv = os->private;
247  int i;
248  if (os->private)
249  for (i = 0; i < 3; i++)
250  av_freep(&priv->packet[i]);
251 }
252 
253 static int vorbis_header(AVFormatContext *s, int idx)
254 {
255  struct ogg *ogg = s->priv_data;
256  AVStream *st = s->streams[idx];
257  struct ogg_stream *os = ogg->streams + idx;
258  struct oggvorbis_private *priv;
259  int pkt_type = os->buf[os->pstart];
260 
261  if (!os->private) {
262  os->private = av_mallocz(sizeof(struct oggvorbis_private));
263  if (!os->private)
264  return AVERROR(ENOMEM);
265  }
266 
267  if (!(pkt_type & 1))
268  return 0;
269 
270  if (os->psize < 1 || pkt_type > 5)
271  return AVERROR_INVALIDDATA;
272 
273  priv = os->private;
274 
275  if (priv->packet[pkt_type >> 1])
276  return AVERROR_INVALIDDATA;
277  if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1])
278  return AVERROR_INVALIDDATA;
279 
280  priv->len[pkt_type >> 1] = os->psize;
281  priv->packet[pkt_type >> 1] = av_mallocz(os->psize);
282  if (!priv->packet[pkt_type >> 1])
283  return AVERROR(ENOMEM);
284  memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize);
285  if (os->buf[os->pstart] == 1) {
286  const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
287  unsigned blocksize, bs0, bs1;
288  int srate;
289 
290  if (os->psize != 30)
291  return AVERROR_INVALIDDATA;
292 
293  if (bytestream_get_le32(&p) != 0) /* vorbis_version */
294  return AVERROR_INVALIDDATA;
295 
296  st->codec->channels = bytestream_get_byte(&p);
297  srate = bytestream_get_le32(&p);
298  p += 4; // skip maximum bitrate
299  st->codec->bit_rate = bytestream_get_le32(&p); // nominal bitrate
300  p += 4; // skip minimum bitrate
301 
302  blocksize = bytestream_get_byte(&p);
303  bs0 = blocksize & 15;
304  bs1 = blocksize >> 4;
305 
306  if (bs0 > bs1)
307  return AVERROR_INVALIDDATA;
308  if (bs0 < 6 || bs1 > 13)
309  return AVERROR_INVALIDDATA;
310 
311  if (bytestream_get_byte(&p) != 1) /* framing_flag */
312  return AVERROR_INVALIDDATA;
313 
316 
317  if (srate > 0) {
318  st->codec->sample_rate = srate;
319  avpriv_set_pts_info(st, 64, 1, srate);
320  }
321  } else if (os->buf[os->pstart] == 3) {
322  if (os->psize > 8 &&
323  ff_vorbis_stream_comment(s, st, os->buf + os->pstart + 7,
324  os->psize - 8) >= 0) {
325  unsigned new_len;
326 
327  int ret = ff_replaygain_export(st, st->metadata);
328  if (ret < 0)
329  return ret;
330 
331  // drop all metadata we parsed and which is not required by libvorbis
332  new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
333  if (new_len >= 16 && new_len < os->psize) {
334  AV_WL32(priv->packet[1] + new_len - 5, 0);
335  priv->packet[1][new_len - 1] = 1;
336  priv->len[1] = new_len;
337  }
338  }
339  } else {
340  int ret = fixup_vorbis_headers(s, priv, &st->codec->extradata);
341  if (ret < 0) {
342  st->codec->extradata_size = 0;
343  return ret;
344  }
345  st->codec->extradata_size = ret;
346  if ((ret = avpriv_vorbis_parse_extradata(st->codec, &priv->vp))) {
347  av_freep(&st->codec->extradata);
348  st->codec->extradata_size = 0;
349  return ret;
350  }
351  }
352 
353  return 1;
354 }
355 
356 static int vorbis_packet(AVFormatContext *s, int idx)
357 {
358  struct ogg *ogg = s->priv_data;
359  struct ogg_stream *os = ogg->streams + idx;
360  struct oggvorbis_private *priv = os->private;
361  int duration;
362 
363  /* first packet handling
364  * here we parse the duration of each packet in the first page and compare
365  * the total duration to the page granule to find the encoder delay and
366  * set the first timestamp */
367  if (!os->lastpts) {
368  int seg;
369  uint8_t *last_pkt = os->buf + os->pstart;
370  uint8_t *next_pkt = last_pkt;
371  int first_duration = 0;
372 
374  duration = 0;
375  for (seg = 0; seg < os->nsegs; seg++) {
376  if (os->segments[seg] < 255) {
377  int d = avpriv_vorbis_parse_frame(&priv->vp, last_pkt, 1);
378  if (d < 0) {
379  duration = os->granule;
380  break;
381  }
382  if (!duration)
383  first_duration = d;
384  duration += d;
385  last_pkt = next_pkt + os->segments[seg];
386  }
387  next_pkt += os->segments[seg];
388  }
389  os->lastpts =
390  os->lastdts = os->granule - duration;
391  s->streams[idx]->start_time = os->lastpts + first_duration;
392  if (s->streams[idx]->duration)
393  s->streams[idx]->duration -= s->streams[idx]->start_time;
394  s->streams[idx]->cur_dts = AV_NOPTS_VALUE;
395  priv->final_pts = AV_NOPTS_VALUE;
397  }
398 
399  /* parse packet duration */
400  if (os->psize > 0) {
401  duration = avpriv_vorbis_parse_frame(&priv->vp, os->buf + os->pstart, 1);
402  if (duration <= 0) {
404  return 0;
405  }
406  os->pduration = duration;
407  }
408 
409  /* final packet handling
410  * here we save the pts of the first packet in the final page, sum up all
411  * packet durations in the final page except for the last one, and compare
412  * to the page granule to find the duration of the final packet */
413  if (os->flags & OGG_FLAG_EOS) {
414  if (os->lastpts != AV_NOPTS_VALUE) {
415  priv->final_pts = os->lastpts;
416  priv->final_duration = 0;
417  }
418  if (os->segp == os->nsegs)
419  os->pduration = os->granule - priv->final_pts - priv->final_duration;
420  priv->final_duration += os->pduration;
421  }
422 
423  return 0;
424 }
425 
426 const struct ogg_codec ff_vorbis_codec = {
427  .magic = "\001vorbis",
428  .magicsize = 7,
429  .header = vorbis_header,
430  .packet = vorbis_packet,
431  .cleanup = vorbis_cleanup,
432  .nb_header = 3,
433 };
unsigned int nb_chapters
Number of chapters in AVChapter array.
Definition: avformat.h:1119
int ff_vorbis_stream_comment(AVFormatContext *as, AVStream *st, const uint8_t *buf, int size)
void * av_malloc(size_t size)
Allocate a block of size bytes with alignment suitable for all memory accesses (including vectors if ...
Definition: mem.c:62
#define AVERROR_INVALIDDATA
Invalid data found when processing input.
Definition: error.h:54
static int vorbis_header(AVFormatContext *s, int idx)
#define AVSTREAM_EVENT_FLAG_METADATA_UPDATED
The call resulted in updated metadata.
Definition: avformat.h:819
int size
VorbisParseContext vp
Copyright (C) 2005 Michael Ahlberg, Måns Rullgård.
Definition: oggdec.h:31
unsigned int pflags
Definition: oggdec.h:67
int avpriv_vorbis_parse_frame(VorbisParseContext *s, const uint8_t *buf, int buf_size)
Get the duration for a Vorbis packet.
#define AV_LOG_WARNING
Something somehow does not look correct.
Definition: log.h:129
void avpriv_set_pts_info(AVStream *s, int pts_wrap_bits, unsigned int pts_num, unsigned int pts_den)
Set the time base and wrapping info for a given stream.
Definition: utils.c:2829
static int fixup_vorbis_headers(AVFormatContext *as, struct oggvorbis_private *priv, uint8_t **buf)
int event_flags
Flags for the user to detect events happening on the stream.
Definition: avformat.h:818
void avpriv_vorbis_parse_reset(VorbisParseContext *s)
int flags
Definition: oggdec.h:76
int ff_vorbis_comment(AVFormatContext *as, AVDictionary **m, const uint8_t *buf, int size, int parse_picture)
AVDictionary * metadata
Definition: avformat.h:909
unsigned char * packet[3]
static int64_t duration
Definition: avplay.c:246
int64_t lastpts
Definition: oggdec.h:72
#define AV_DICT_DONT_STRDUP_KEY
Take ownership of a key that's been allocated with av_malloc() and children.
Definition: dict.h:63
void av_freep(void *arg)
Free a memory block which has been allocated with av_malloc(z)() or av_realloc() and set the pointer ...
Definition: mem.c:198
AVChapter * avpriv_new_chapter(AVFormatContext *s, int id, AVRational time_base, int64_t start, int64_t end, const char *title)
Add a new chapter.
Definition: utils.c:2601
Format I/O context.
Definition: avformat.h:922
unsigned int psize
Definition: oggdec.h:66
int64_t cur_dts
Definition: avformat.h:851
Public dictionary API.
uint8_t
int id
unique ID to identify the chapter
Definition: avformat.h:906
#define AV_WL32(p, d)
Definition: intreadwrite.h:255
uint8_t * extradata
some codecs need / can use extradata like Huffman tables.
Definition: avcodec.h:1164
AVStream ** streams
A list of all streams in the file.
Definition: avformat.h:990
Vorbis audio parser.
int av_reallocp(void *ptr, size_t size)
Allocate or reallocate a block of memory.
Definition: mem.c:140
bitstream reader API header.
int ff_replaygain_export(AVStream *st, AVDictionary *metadata)
Parse replaygain tags and export them as per-stream side data.
Definition: replaygain.c:110
static void vorbis_cleanup(AVFormatContext *s, int idx)
const AVMetadataConv ff_vorbiscomment_metadata_conv[]
VorbisComment metadata conversion mapping.
Definition: vorbiscomment.c:33
void av_free(void *ptr)
Free a memory block which has been allocated with av_malloc(z)() or av_realloc(). ...
Definition: mem.c:186
#define AVERROR(e)
Definition: error.h:43
#define OGG_FLAG_EOS
Definition: oggdec.h:106
AVChapter ** chapters
Definition: avformat.h:1120
void av_log(void *avcl, int level, const char *fmt,...)
Definition: log.c:169
uint8_t segments[255]
Definition: oggdec.h:80
AVCodecContext * codec
Codec context associated with this stream.
Definition: avformat.h:718
#define FF_INPUT_BUFFER_PADDING_SIZE
Required number of additionally allocated bytes at the end of the input bitstream for decoding...
Definition: avcodec.h:531
int bit_rate
the average bitrate
Definition: avcodec.h:1114
uint64_t granule
Definition: oggdec.h:70
#define AV_DICT_DONT_STRDUP_VAL
Take ownership of a value that's been allocated with av_malloc() and chilren.
Definition: dict.h:66
unsigned int pstart
Definition: oggdec.h:65
struct ogg_stream * streams
Definition: oggdec.h:97
int segp
Definition: oggdec.h:79
static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
int ff_flac_parse_picture(AVFormatContext *s, uint8_t *buf, int buf_size)
Definition: flac_picture.c:26
#define AV_RL32
Definition: intreadwrite.h:146
AVDictionary * metadata
Definition: avformat.h:771
Stream structure.
Definition: avformat.h:699
NULL
Definition: eval.c:55
#define AV_LOG_INFO
Standard information.
Definition: log.h:134
enum AVMediaType codec_type
Definition: avcodec.h:1058
unsigned int pduration
Definition: oggdec.h:68
int nsegs
Definition: oggdec.h:79
enum AVCodecID codec_id
Definition: avcodec.h:1067
unsigned int len[3]
int sample_rate
samples per second
Definition: avcodec.h:1807
int extradata_size
Definition: avcodec.h:1165
void * private
Definition: oggdec.h:85
int av_dict_set(AVDictionary **pm, const char *key, const char *value, int flags)
Set the given entry in *pm, overwriting an existing entry.
Definition: dict.c:68
unsigned int av_xiphlacing(unsigned char *s, unsigned int v)
Encode extradata length to a buffer.
Definition: utils.c:2237
rational number numerator/denominator
Definition: rational.h:43
byte swapping routines
int64_t lastdts
Definition: oggdec.h:73
const int8_t * magic
Definition: oggdec.h:32
int64_t duration
Decoding: duration of the stream, in stream time base.
Definition: avformat.h:756
static int av_toupper(int c)
Locale-independent conversion of ASCII characters to uppercase.
Definition: avstring.h:172
uint8_t * buf
Definition: oggdec.h:62
Main libavformat public API header.
int64_t start_time
Decoding: pts of the first frame of the stream, in stream time base.
Definition: avformat.h:749
static int vorbis_packet(AVFormatContext *s, int idx)
#define AV_PKT_FLAG_CORRUPT
The packet content is corrupted.
Definition: avcodec.h:1020
void ff_metadata_conv(AVDictionary **pm, const AVMetadataConv *d_conv, const AVMetadataConv *s_conv)
Definition: metadata.c:26
Definition: oggdec.h:96
int len
int channels
number of audio channels
Definition: avcodec.h:1808
void * priv_data
Format private data.
Definition: avformat.h:950
const struct ogg_codec ff_vorbis_codec
int av_base64_decode(uint8_t *out, const char *in, int out_size)
Decode a base64-encoded string.
Definition: base64.c:45
int avpriv_vorbis_parse_extradata(AVCodecContext *avctx, VorbisParseContext *s)
Initialize the Vorbis parser using headers in the extradata.
void * av_mallocz(size_t size)
Allocate a block of size bytes with alignment suitable for all memory accesses (including vectors if ...
Definition: mem.c:205
#define AV_NOPTS_VALUE
Undefined timestamp value.
Definition: avutil.h:228