FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
oggenc.c
Go to the documentation of this file.
1 /*
2  * Ogg muxer
3  * Copyright (c) 2007 Baptiste Coudurier <baptiste dot coudurier at free dot fr>
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/crc.h"
23 #include "libavutil/opt.h"
24 #include "libavutil/mathematics.h"
25 #include "libavutil/opt.h"
26 #include "libavutil/random_seed.h"
27 #include "libavcodec/xiph.h"
28 #include "libavcodec/bytestream.h"
29 #include "libavcodec/flac.h"
30 #include "avformat.h"
31 #include "avio_internal.h"
32 #include "internal.h"
33 #include "vorbiscomment.h"
34 
35 #define MAX_PAGE_SIZE 65025
36 
37 typedef struct {
38  int64_t start_granule;
39  int64_t granule;
45  uint16_t size;
46 } OGGPage;
47 
48 typedef struct {
49  unsigned page_counter;
50  uint8_t *header[3];
51  int header_len[3];
52  /** for theora granule */
53  int kfgshift;
54  int64_t last_kf_pts;
55  int vrev;
56  int eos;
57  unsigned page_count; ///< number of page buffered
58  OGGPage page; ///< current page
59  unsigned serial_num; ///< serial number
60  int64_t last_granule; ///< last packet granule
62 
63 typedef struct OGGPageList {
65  struct OGGPageList *next;
66 } OGGPageList;
67 
68 typedef struct {
69  const AVClass *class;
71  int pref_size; ///< preferred page size (0 => fill all segments)
72  int64_t pref_duration; ///< preferred page duration (0 => fill all segments)
73 } OGGContext;
74 
75 #define OFFSET(x) offsetof(OGGContext, x)
76 #define PARAM AV_OPT_FLAG_ENCODING_PARAM
77 
78 static const AVOption options[] = {
79  { "oggpagesize", "Set preferred Ogg page size.",
80  offsetof(OGGContext, pref_size), AV_OPT_TYPE_INT, {.i64 = 0}, 0, MAX_PAGE_SIZE, AV_OPT_FLAG_ENCODING_PARAM},
81  { "pagesize", "preferred page size in bytes (deprecated)",
82  OFFSET(pref_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, MAX_PAGE_SIZE, PARAM },
83  { "page_duration", "preferred page duration, in microseconds",
84  OFFSET(pref_duration), AV_OPT_TYPE_INT64, { .i64 = 1000000 }, 0, INT64_MAX, PARAM },
85  { NULL },
86 };
87 
88 static const AVClass ogg_muxer_class = {
89  .class_name = "Ogg muxer",
90  .item_name = av_default_item_name,
91  .option = options,
92  .version = LIBAVUTIL_VERSION_INT,
93 };
94 
95 
96 static void ogg_update_checksum(AVFormatContext *s, AVIOContext *pb, int64_t crc_offset)
97 {
98  int64_t pos = avio_tell(pb);
99  uint32_t checksum = ffio_get_checksum(pb);
100  avio_seek(pb, crc_offset, SEEK_SET);
101  avio_wb32(pb, checksum);
102  avio_seek(pb, pos, SEEK_SET);
103 }
104 
105 static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
106 {
107  OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
108  AVIOContext *pb;
109  int64_t crc_offset;
110  int ret, size;
111  uint8_t *buf;
112 
113  ret = avio_open_dyn_buf(&pb);
114  if (ret < 0)
115  return ret;
117  ffio_wfourcc(pb, "OggS");
118  avio_w8(pb, 0);
119  avio_w8(pb, page->flags | extra_flags);
120  avio_wl64(pb, page->granule);
121  avio_wl32(pb, oggstream->serial_num);
122  avio_wl32(pb, oggstream->page_counter++);
123  crc_offset = avio_tell(pb);
124  avio_wl32(pb, 0); // crc
125  avio_w8(pb, page->segments_count);
126  avio_write(pb, page->segments, page->segments_count);
127  avio_write(pb, page->data, page->size);
128 
129  ogg_update_checksum(s, pb, crc_offset);
130  avio_flush(pb);
131 
132  size = avio_close_dyn_buf(pb, &buf);
133  if (size < 0)
134  return size;
135 
136  avio_write(s->pb, buf, size);
137  avio_flush(s->pb);
138  av_free(buf);
139  oggstream->page_count--;
140  return 0;
141 }
142 
143 static int ogg_key_granule(OGGStreamContext *oggstream, int64_t granule)
144 {
145  return oggstream->kfgshift && !(granule & ((1<<oggstream->kfgshift)-1));
146 }
147 
148 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, int64_t granule)
149 {
150  if (oggstream->kfgshift)
151  return (granule>>oggstream->kfgshift) +
152  (granule & ((1<<oggstream->kfgshift)-1));
153  else
154  return granule;
155 }
156 
158 {
159  AVStream *st2 = s->streams[next->stream_index];
160  AVStream *st = s->streams[page->stream_index];
161  int64_t next_granule, cur_granule;
162 
163  if (next->granule == -1 || page->granule == -1)
164  return 0;
165 
166  next_granule = av_rescale_q(ogg_granule_to_timestamp(st2->priv_data, next->granule),
167  st2->time_base, AV_TIME_BASE_Q);
168  cur_granule = av_rescale_q(ogg_granule_to_timestamp(st->priv_data, page->granule),
169  st ->time_base, AV_TIME_BASE_Q);
170  return next_granule > cur_granule;
171 }
172 
173 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
174 {
175  oggstream->page.granule = -1;
176  oggstream->page.flags = 0;
177  oggstream->page.segments_count = 0;
178  oggstream->page.size = 0;
179  return 0;
180 }
181 
183 {
184  OGGContext *ogg = s->priv_data;
185  OGGPageList **p = &ogg->page_list;
186  OGGPageList *l = av_mallocz(sizeof(*l));
187 
188  if (!l)
189  return AVERROR(ENOMEM);
190  l->page = oggstream->page;
191 
192  oggstream->page.start_granule = oggstream->page.granule;
193  oggstream->page_count++;
194  ogg_reset_cur_page(oggstream);
195 
196  while (*p) {
197  if (ogg_compare_granule(s, &(*p)->page, &l->page))
198  break;
199  p = &(*p)->next;
200  }
201  l->next = *p;
202  *p = l;
203 
204  return 0;
205 }
206 
208  uint8_t *data, unsigned size, int64_t granule,
209  int header)
210 {
211  OGGStreamContext *oggstream = st->priv_data;
212  OGGContext *ogg = s->priv_data;
213  int total_segments = size / 255 + 1;
214  uint8_t *p = data;
215  int i, segments, len, flush = 0;
216 
217  // Handles VFR by flushing page because this frame needs to have a timestamp
218  // For theora, keyframes also need to have a timestamp to correctly mark
219  // them as such, otherwise seeking will not work correctly at the very
220  // least with old libogg versions.
221  // Do not try to flush header packets though, that will create broken files.
222  if (st->codec->codec_id == AV_CODEC_ID_THEORA && !header &&
223  (ogg_granule_to_timestamp(oggstream, granule) >
224  ogg_granule_to_timestamp(oggstream, oggstream->last_granule) + 1 ||
225  ogg_key_granule(oggstream, granule))) {
226  if (oggstream->page.granule != -1)
227  ogg_buffer_page(s, oggstream);
228  flush = 1;
229  }
230 
231  // avoid a continued page
232  if (!header && oggstream->page.size > 0 &&
233  MAX_PAGE_SIZE - oggstream->page.size < size) {
234  ogg_buffer_page(s, oggstream);
235  }
236 
237  for (i = 0; i < total_segments; ) {
238  OGGPage *page = &oggstream->page;
239 
240  segments = FFMIN(total_segments - i, 255 - page->segments_count);
241 
242  if (i && !page->segments_count)
243  page->flags |= 1; // continued packet
244 
245  memset(page->segments+page->segments_count, 255, segments - 1);
246  page->segments_count += segments - 1;
247 
248  len = FFMIN(size, segments*255);
249  page->segments[page->segments_count++] = len - (segments-1)*255;
250  memcpy(page->data+page->size, p, len);
251  p += len;
252  size -= len;
253  i += segments;
254  page->size += len;
255 
256  if (i == total_segments)
257  page->granule = granule;
258 
259  if (!header) {
260  AVStream *st = s->streams[page->stream_index];
261 
262  int64_t start = av_rescale_q(page->start_granule, st->time_base,
264  int64_t next = av_rescale_q(page->granule, st->time_base,
266 
267  if (page->segments_count == 255 ||
268  (ogg->pref_size > 0 && page->size >= ogg->pref_size) ||
269  (ogg->pref_duration > 0 && next - start >= ogg->pref_duration)) {
270  ogg_buffer_page(s, oggstream);
271  }
272  }
273  }
274 
275  if (flush && oggstream->page.granule != -1)
276  ogg_buffer_page(s, oggstream);
277 
278  return 0;
279 }
280 
281 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
282  int *header_len, AVDictionary **m, int framing_bit)
283 {
284  const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
285  int size;
286  uint8_t *p, *p0;
287  unsigned int count;
288 
290 
291  size = offset + ff_vorbiscomment_length(*m, vendor, &count) + framing_bit;
292  p = av_mallocz(size);
293  if (!p)
294  return NULL;
295  p0 = p;
296 
297  p += offset;
298  ff_vorbiscomment_write(&p, m, vendor, count);
299  if (framing_bit)
300  bytestream_put_byte(&p, 1);
301 
302  *header_len = size;
303  return p0;
304 }
305 
307  OGGStreamContext *oggstream, int bitexact,
308  AVDictionary **m)
309 {
310  enum FLACExtradataFormat format;
311  uint8_t *streaminfo;
312  uint8_t *p;
313 
314  if (!avpriv_flac_is_extradata_valid(avctx, &format, &streaminfo))
315  return -1;
316 
317  // first packet: STREAMINFO
318  oggstream->header_len[0] = 51;
319  oggstream->header[0] = av_mallocz(51); // per ogg flac specs
320  p = oggstream->header[0];
321  if (!p)
322  return AVERROR(ENOMEM);
323  bytestream_put_byte(&p, 0x7F);
324  bytestream_put_buffer(&p, "FLAC", 4);
325  bytestream_put_byte(&p, 1); // major version
326  bytestream_put_byte(&p, 0); // minor version
327  bytestream_put_be16(&p, 1); // headers packets without this one
328  bytestream_put_buffer(&p, "fLaC", 4);
329  bytestream_put_byte(&p, 0x00); // streaminfo
330  bytestream_put_be24(&p, 34);
332 
333  // second packet: VorbisComment
334  p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
335  if (!p)
336  return AVERROR(ENOMEM);
337  oggstream->header[1] = p;
338  bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
339  bytestream_put_be24(&p, oggstream->header_len[1] - 4);
340 
341  return 0;
342 }
343 
344 #define SPEEX_HEADER_SIZE 80
345 
347  OGGStreamContext *oggstream, int bitexact,
348  AVDictionary **m)
349 {
350  uint8_t *p;
351 
352  if (avctx->extradata_size < SPEEX_HEADER_SIZE)
353  return -1;
354 
355  // first packet: Speex header
357  if (!p)
358  return AVERROR(ENOMEM);
359  oggstream->header[0] = p;
360  oggstream->header_len[0] = SPEEX_HEADER_SIZE;
362  AV_WL32(&oggstream->header[0][68], 0); // set extra_headers to 0
363 
364  // second packet: VorbisComment
365  p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
366  if (!p)
367  return AVERROR(ENOMEM);
368  oggstream->header[1] = p;
369 
370  return 0;
371 }
372 
373 #define OPUS_HEADER_SIZE 19
374 
376  OGGStreamContext *oggstream, int bitexact,
377  AVDictionary **m)
378 {
379  uint8_t *p;
380 
381  if (avctx->extradata_size < OPUS_HEADER_SIZE)
382  return -1;
383 
384  /* first packet: Opus header */
385  p = av_mallocz(avctx->extradata_size);
386  if (!p)
387  return AVERROR(ENOMEM);
388  oggstream->header[0] = p;
389  oggstream->header_len[0] = avctx->extradata_size;
390  bytestream_put_buffer(&p, avctx->extradata, avctx->extradata_size);
391 
392  /* second packet: VorbisComment */
393  p = ogg_write_vorbiscomment(8, bitexact, &oggstream->header_len[1], m, 0);
394  if (!p)
395  return AVERROR(ENOMEM);
396  oggstream->header[1] = p;
397  bytestream_put_buffer(&p, "OpusTags", 8);
398 
399  return 0;
400 }
401 
403 {
404  OGGContext *ogg = s->priv_data;
405  OGGStreamContext *oggstream = NULL;
406  int i, j;
407 
408  if (ogg->pref_size)
409  av_log(s, AV_LOG_WARNING, "The pagesize option is deprecated\n");
410 
411  for (i = 0; i < s->nb_streams; i++) {
412  AVStream *st = s->streams[i];
413  unsigned serial_num = i;
414 
415  if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
416  if (st->codec->codec_id == AV_CODEC_ID_OPUS)
417  /* Opus requires a fixed 48kHz clock */
418  avpriv_set_pts_info(st, 64, 1, 48000);
419  else
420  avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
421  } else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
423  if (st->codec->codec_id != AV_CODEC_ID_VORBIS &&
425  st->codec->codec_id != AV_CODEC_ID_SPEEX &&
426  st->codec->codec_id != AV_CODEC_ID_FLAC &&
427  st->codec->codec_id != AV_CODEC_ID_OPUS) {
428  av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
429  return -1;
430  }
431 
432  if (!st->codec->extradata || !st->codec->extradata_size) {
433  av_log(s, AV_LOG_ERROR, "No extradata present\n");
434  return -1;
435  }
436  oggstream = av_mallocz(sizeof(*oggstream));
437  oggstream->page.stream_index = i;
438 
439  if (!(st->codec->flags & CODEC_FLAG_BITEXACT))
440  do {
441  serial_num = av_get_random_seed();
442  for (j = 0; j < i; j++) {
443  OGGStreamContext *sc = s->streams[j]->priv_data;
444  if (serial_num == sc->serial_num)
445  break;
446  }
447  } while (j < i);
448  oggstream->serial_num = serial_num;
449 
450  st->priv_data = oggstream;
451  if (st->codec->codec_id == AV_CODEC_ID_FLAC) {
452  int err = ogg_build_flac_headers(st->codec, oggstream,
454  &s->metadata);
455  if (err) {
456  av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
457  av_freep(&st->priv_data);
458  return err;
459  }
460  } else if (st->codec->codec_id == AV_CODEC_ID_SPEEX) {
461  int err = ogg_build_speex_headers(st->codec, oggstream,
463  &s->metadata);
464  if (err) {
465  av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
466  av_freep(&st->priv_data);
467  return err;
468  }
469  } else if (st->codec->codec_id == AV_CODEC_ID_OPUS) {
470  int err = ogg_build_opus_headers(st->codec, oggstream,
472  &s->metadata);
473  if (err) {
474  av_log(s, AV_LOG_ERROR, "Error writing Opus headers\n");
475  av_freep(&st->priv_data);
476  return err;
477  }
478  } else {
479  uint8_t *p;
480  const char *cstr = st->codec->codec_id == AV_CODEC_ID_VORBIS ? "vorbis" : "theora";
481  int header_type = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 3 : 0x81;
482  int framing_bit = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 1 : 0;
483 
485  st->codec->codec_id == AV_CODEC_ID_VORBIS ? 30 : 42,
486  oggstream->header, oggstream->header_len) < 0) {
487  av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
488  av_freep(&st->priv_data);
489  return -1;
490  }
491 
493  &oggstream->header_len[1], &s->metadata,
494  framing_bit);
495  oggstream->header[1] = p;
496  if (!p)
497  return AVERROR(ENOMEM);
498 
499  bytestream_put_byte(&p, header_type);
500  bytestream_put_buffer(&p, cstr, 6);
501 
502  if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
503  /** KFGSHIFT is the width of the less significant section of the granule position
504  The less significant section is the frame count since the last keyframe */
505  oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
506  oggstream->vrev = oggstream->header[0][9];
507  av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
508  oggstream->kfgshift, oggstream->vrev);
509  }
510  }
511  }
512 
513  for (j = 0; j < s->nb_streams; j++) {
514  OGGStreamContext *oggstream = s->streams[j]->priv_data;
515  ogg_buffer_data(s, s->streams[j], oggstream->header[0],
516  oggstream->header_len[0], 0, 1);
517  oggstream->page.flags |= 2; // bos
518  ogg_buffer_page(s, oggstream);
519  }
520  for (j = 0; j < s->nb_streams; j++) {
521  AVStream *st = s->streams[j];
522  OGGStreamContext *oggstream = st->priv_data;
523  for (i = 1; i < 3; i++) {
524  if (oggstream->header_len[i])
525  ogg_buffer_data(s, st, oggstream->header[i],
526  oggstream->header_len[i], 0, 1);
527  }
528  ogg_buffer_page(s, oggstream);
529  }
530 
531  oggstream->page.start_granule = AV_NOPTS_VALUE;
532 
533  return 0;
534 }
535 
537 {
538  OGGContext *ogg = s->priv_data;
539  OGGPageList *next, *p;
540 
541  if (!ogg->page_list)
542  return;
543 
544  for (p = ogg->page_list; p; ) {
545  OGGStreamContext *oggstream =
547  if (oggstream->page_count < 2 && !flush)
548  break;
549  ogg_write_page(s, &p->page,
550  flush && oggstream->page_count == 1 ? 4 : 0); // eos
551  next = p->next;
552  av_freep(&p);
553  p = next;
554  }
555  ogg->page_list = p;
556 }
557 
559 {
560  AVStream *st = s->streams[pkt->stream_index];
561  OGGStreamContext *oggstream = st->priv_data;
562  int ret;
563  int64_t granule;
564 
565  if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
566  int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
567  int pframe_count;
568  if (pkt->flags & AV_PKT_FLAG_KEY)
569  oggstream->last_kf_pts = pts;
570  pframe_count = pts - oggstream->last_kf_pts;
571  // prevent frame count from overflow if key frame flag is not set
572  if (pframe_count >= (1<<oggstream->kfgshift)) {
573  oggstream->last_kf_pts += pframe_count;
574  pframe_count = 0;
575  }
576  granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
577  } else if (st->codec->codec_id == AV_CODEC_ID_OPUS)
578  granule = pkt->pts + pkt->duration + av_rescale_q(st->codec->delay, (AVRational){ 1, st->codec->sample_rate }, st->time_base);
579  else
580  granule = pkt->pts + pkt->duration;
581 
582  if (oggstream->page.start_granule == AV_NOPTS_VALUE)
583  oggstream->page.start_granule = pkt->pts;
584 
585  ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule, 0);
586  if (ret < 0)
587  return ret;
588 
589  ogg_write_pages(s, 0);
590 
591  oggstream->last_granule = granule;
592 
593  return 0;
594 }
595 
597 {
598  int i;
599 
600  /* flush current page if needed */
601  for (i = 0; i < s->nb_streams; i++) {
602  OGGStreamContext *oggstream = s->streams[i]->priv_data;
603 
604  if (oggstream->page.size > 0)
605  ogg_buffer_page(s, oggstream);
606  }
607 
608  ogg_write_pages(s, 1);
609 
610  for (i = 0; i < s->nb_streams; i++) {
611  AVStream *st = s->streams[i];
612  OGGStreamContext *oggstream = st->priv_data;
613  if (st->codec->codec_id == AV_CODEC_ID_FLAC ||
614  st->codec->codec_id == AV_CODEC_ID_SPEEX ||
615  st->codec->codec_id == AV_CODEC_ID_OPUS) {
616  av_freep(&oggstream->header[0]);
617  }
618  av_freep(&oggstream->header[1]);
619  av_freep(&st->priv_data);
620  }
621  return 0;
622 }
623 
625  .name = "ogg",
626  .long_name = NULL_IF_CONFIG_SMALL("Ogg"),
627  .mime_type = "application/ogg",
628  .extensions = "ogg,ogv,spx,opus",
629  .priv_data_size = sizeof(OGGContext),
630  .audio_codec = AV_CODEC_ID_FLAC,
631  .video_codec = AV_CODEC_ID_THEORA,
636  .priv_class = &ogg_muxer_class,
637 };