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 <stdint.h>
23 
24 #include "libavutil/crc.h"
25 #include "libavutil/mathematics.h"
26 #include "libavutil/opt.h"
27 #include "libavutil/random_seed.h"
28 #include "libavcodec/xiph.h"
29 #include "libavcodec/bytestream.h"
30 #include "libavcodec/flac.h"
31 #include "avformat.h"
32 #include "avio_internal.h"
33 #include "internal.h"
34 #include "vorbiscomment.h"
35 
36 #define MAX_PAGE_SIZE 65025
37 
38 typedef struct {
39  int64_t start_granule;
40  int64_t granule;
46  uint16_t size;
47 } OGGPage;
48 
49 typedef struct {
50  unsigned page_counter;
52  int header_len[3];
53  /** for theora granule */
54  int kfgshift;
55  int64_t last_kf_pts;
56  int vrev;
57  int eos;
58  unsigned page_count; ///< number of page buffered
59  OGGPage page; ///< current page
60  unsigned serial_num; ///< serial number
61  int64_t last_granule; ///< last packet granule
63 
64 typedef struct OGGPageList {
66  struct OGGPageList *next;
67 } OGGPageList;
68 
69 typedef struct {
70  const AVClass *class;
72  int pref_size; ///< preferred page size (0 => fill all segments)
73  int64_t pref_duration; ///< preferred page duration (0 => fill all segments)
74 } OGGContext;
75 
76 #define OFFSET(x) offsetof(OGGContext, x)
77 #define PARAM AV_OPT_FLAG_ENCODING_PARAM
78 
79 static const AVOption options[] = {
80  { "oggpagesize", "Set preferred Ogg page size.",
81  offsetof(OGGContext, pref_size), AV_OPT_TYPE_INT, {.i64 = 0}, 0, MAX_PAGE_SIZE, AV_OPT_FLAG_ENCODING_PARAM},
82  { "pagesize", "preferred page size in bytes (deprecated)",
83  OFFSET(pref_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, MAX_PAGE_SIZE, PARAM },
84  { "page_duration", "preferred page duration, in microseconds",
85  OFFSET(pref_duration), AV_OPT_TYPE_INT64, { .i64 = 1000000 }, 0, INT64_MAX, PARAM },
86  { NULL },
87 };
88 
89 static void ogg_update_checksum(AVFormatContext *s, AVIOContext *pb, int64_t crc_offset)
90 {
91  int64_t pos = avio_tell(pb);
92  uint32_t checksum = ffio_get_checksum(pb);
93  avio_seek(pb, crc_offset, SEEK_SET);
94  avio_wb32(pb, checksum);
95  avio_seek(pb, pos, SEEK_SET);
96 }
97 
98 static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
99 {
100  OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
101  AVIOContext *pb;
102  int64_t crc_offset;
103  int ret, size;
104  uint8_t *buf;
105 
106  ret = avio_open_dyn_buf(&pb);
107  if (ret < 0)
108  return ret;
110  ffio_wfourcc(pb, "OggS");
111  avio_w8(pb, 0);
112  avio_w8(pb, page->flags | extra_flags);
113  avio_wl64(pb, page->granule);
114  avio_wl32(pb, oggstream->serial_num);
115  avio_wl32(pb, oggstream->page_counter++);
116  crc_offset = avio_tell(pb);
117  avio_wl32(pb, 0); // crc
118  avio_w8(pb, page->segments_count);
119  avio_write(pb, page->segments, page->segments_count);
120  avio_write(pb, page->data, page->size);
121 
122  ogg_update_checksum(s, pb, crc_offset);
123  avio_flush(pb);
124 
125  size = avio_close_dyn_buf(pb, &buf);
126  if (size < 0)
127  return size;
128 
129  avio_write(s->pb, buf, size);
130  avio_flush(s->pb);
131  av_free(buf);
132  oggstream->page_count--;
133  return 0;
134 }
135 
136 static int ogg_key_granule(OGGStreamContext *oggstream, int64_t granule)
137 {
138  return oggstream->kfgshift && !(granule & ((1<<oggstream->kfgshift)-1));
139 }
140 
141 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, int64_t granule)
142 {
143  if (oggstream->kfgshift)
144  return (granule>>oggstream->kfgshift) +
145  (granule & ((1<<oggstream->kfgshift)-1));
146  else
147  return granule;
148 }
149 
151 {
152  AVStream *st2 = s->streams[next->stream_index];
153  AVStream *st = s->streams[page->stream_index];
154  int64_t next_granule, cur_granule;
155 
156  if (next->granule == -1 || page->granule == -1)
157  return 0;
158 
159  next_granule = av_rescale_q(ogg_granule_to_timestamp(st2->priv_data, next->granule),
160  st2->time_base, AV_TIME_BASE_Q);
161  cur_granule = av_rescale_q(ogg_granule_to_timestamp(st->priv_data, page->granule),
162  st ->time_base, AV_TIME_BASE_Q);
163  return next_granule > cur_granule;
164 }
165 
166 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
167 {
168  oggstream->page.granule = -1;
169  oggstream->page.flags = 0;
170  oggstream->page.segments_count = 0;
171  oggstream->page.size = 0;
172  return 0;
173 }
174 
176 {
177  OGGContext *ogg = s->priv_data;
178  OGGPageList **p = &ogg->page_list;
179  OGGPageList *l = av_mallocz(sizeof(*l));
180 
181  if (!l)
182  return AVERROR(ENOMEM);
183  l->page = oggstream->page;
184 
185  oggstream->page.start_granule = oggstream->page.granule;
186  oggstream->page_count++;
187  ogg_reset_cur_page(oggstream);
188 
189  while (*p) {
190  if (ogg_compare_granule(s, &(*p)->page, &l->page))
191  break;
192  p = &(*p)->next;
193  }
194  l->next = *p;
195  *p = l;
196 
197  return 0;
198 }
199 
201  uint8_t *data, unsigned size, int64_t granule,
202  int header)
203 {
204  OGGStreamContext *oggstream = st->priv_data;
205  OGGContext *ogg = s->priv_data;
206  int total_segments = size / 255 + 1;
207  uint8_t *p = data;
208  int i, segments, len, flush = 0;
209 
210  // Handles VFR by flushing page because this frame needs to have a timestamp
211  // For theora, keyframes also need to have a timestamp to correctly mark
212  // them as such, otherwise seeking will not work correctly at the very
213  // least with old libogg versions.
214  // Do not try to flush header packets though, that will create broken files.
215  if (st->codec->codec_id == AV_CODEC_ID_THEORA && !header &&
216  (ogg_granule_to_timestamp(oggstream, granule) >
217  ogg_granule_to_timestamp(oggstream, oggstream->last_granule) + 1 ||
218  ogg_key_granule(oggstream, granule))) {
219  if (oggstream->page.granule != -1)
220  ogg_buffer_page(s, oggstream);
221  flush = 1;
222  }
223 
224  // avoid a continued page
225  if (!header && oggstream->page.size > 0 &&
226  MAX_PAGE_SIZE - oggstream->page.size < size) {
227  ogg_buffer_page(s, oggstream);
228  }
229 
230  for (i = 0; i < total_segments; ) {
231  OGGPage *page = &oggstream->page;
232 
233  segments = FFMIN(total_segments - i, 255 - page->segments_count);
234 
235  if (i && !page->segments_count)
236  page->flags |= 1; // continued packet
237 
238  memset(page->segments+page->segments_count, 255, segments - 1);
239  page->segments_count += segments - 1;
240 
241  len = FFMIN(size, segments*255);
242  page->segments[page->segments_count++] = len - (segments-1)*255;
243  memcpy(page->data+page->size, p, len);
244  p += len;
245  size -= len;
246  i += segments;
247  page->size += len;
248 
249  if (i == total_segments)
250  page->granule = granule;
251 
252  if (!header) {
253  AVStream *st = s->streams[page->stream_index];
254 
255  int64_t start = av_rescale_q(page->start_granule, st->time_base,
257  int64_t next = av_rescale_q(page->granule, st->time_base,
259 
260  if (page->segments_count == 255 ||
261  (ogg->pref_size > 0 && page->size >= ogg->pref_size) ||
262  (ogg->pref_duration > 0 && next - start >= ogg->pref_duration)) {
263  ogg_buffer_page(s, oggstream);
264  }
265  }
266  }
267 
268  if (flush && oggstream->page.granule != -1)
269  ogg_buffer_page(s, oggstream);
270 
271  return 0;
272 }
273 
274 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
275  int *header_len, AVDictionary **m, int framing_bit)
276 {
277  const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
278  int size;
279  uint8_t *p, *p0;
280  unsigned int count;
281 
283 
284  size = offset + ff_vorbiscomment_length(*m, vendor, &count) + framing_bit;
285  p = av_mallocz(size);
286  if (!p)
287  return NULL;
288  p0 = p;
289 
290  p += offset;
291  ff_vorbiscomment_write(&p, m, vendor, count);
292  if (framing_bit)
293  bytestream_put_byte(&p, 1);
294 
295  *header_len = size;
296  return p0;
297 }
298 
300  OGGStreamContext *oggstream, int bitexact,
301  AVDictionary **m)
302 {
303  enum FLACExtradataFormat format;
304  uint8_t *streaminfo;
305  uint8_t *p;
306 
307  if (!avpriv_flac_is_extradata_valid(avctx, &format, &streaminfo))
308  return -1;
309 
310  // first packet: STREAMINFO
311  oggstream->header_len[0] = 51;
312  oggstream->header[0] = av_mallocz(51); // per ogg flac specs
313  p = oggstream->header[0];
314  if (!p)
315  return AVERROR(ENOMEM);
316  bytestream_put_byte(&p, 0x7F);
317  bytestream_put_buffer(&p, "FLAC", 4);
318  bytestream_put_byte(&p, 1); // major version
319  bytestream_put_byte(&p, 0); // minor version
320  bytestream_put_be16(&p, 1); // headers packets without this one
321  bytestream_put_buffer(&p, "fLaC", 4);
322  bytestream_put_byte(&p, 0x00); // streaminfo
323  bytestream_put_be24(&p, 34);
325 
326  // second packet: VorbisComment
327  p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
328  if (!p)
329  return AVERROR(ENOMEM);
330  oggstream->header[1] = p;
331  bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
332  bytestream_put_be24(&p, oggstream->header_len[1] - 4);
333 
334  return 0;
335 }
336 
337 #define SPEEX_HEADER_SIZE 80
338 
340  OGGStreamContext *oggstream, int bitexact,
341  AVDictionary **m)
342 {
343  uint8_t *p;
344 
345  if (avctx->extradata_size < SPEEX_HEADER_SIZE)
346  return -1;
347 
348  // first packet: Speex header
350  if (!p)
351  return AVERROR(ENOMEM);
352  oggstream->header[0] = p;
353  oggstream->header_len[0] = SPEEX_HEADER_SIZE;
355  AV_WL32(&oggstream->header[0][68], 0); // set extra_headers to 0
356 
357  // second packet: VorbisComment
358  p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
359  if (!p)
360  return AVERROR(ENOMEM);
361  oggstream->header[1] = p;
362 
363  return 0;
364 }
365 
366 #define OPUS_HEADER_SIZE 19
367 
369  OGGStreamContext *oggstream, int bitexact,
370  AVDictionary **m)
371 {
372  uint8_t *p;
373 
374  if (avctx->extradata_size < OPUS_HEADER_SIZE)
375  return -1;
376 
377  /* first packet: Opus header */
378  p = av_mallocz(avctx->extradata_size);
379  if (!p)
380  return AVERROR(ENOMEM);
381  oggstream->header[0] = p;
382  oggstream->header_len[0] = avctx->extradata_size;
383  bytestream_put_buffer(&p, avctx->extradata, avctx->extradata_size);
384 
385  /* second packet: VorbisComment */
386  p = ogg_write_vorbiscomment(8, bitexact, &oggstream->header_len[1], m, 0);
387  if (!p)
388  return AVERROR(ENOMEM);
389  oggstream->header[1] = p;
390  bytestream_put_buffer(&p, "OpusTags", 8);
391 
392  return 0;
393 }
394 
396 {
397  OGGContext *ogg = s->priv_data;
398  OGGStreamContext *oggstream = NULL;
399  int i, j;
400 
401  if (ogg->pref_size)
402  av_log(s, AV_LOG_WARNING, "The pagesize option is deprecated\n");
403 
404  for (i = 0; i < s->nb_streams; i++) {
405  AVStream *st = s->streams[i];
406  unsigned serial_num = i;
407 
408  if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
409  if (st->codec->codec_id == AV_CODEC_ID_OPUS)
410  /* Opus requires a fixed 48kHz clock */
411  avpriv_set_pts_info(st, 64, 1, 48000);
412  else
413  avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
414  } else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
416  if (st->codec->codec_id != AV_CODEC_ID_VORBIS &&
418  st->codec->codec_id != AV_CODEC_ID_SPEEX &&
419  st->codec->codec_id != AV_CODEC_ID_FLAC &&
420  st->codec->codec_id != AV_CODEC_ID_OPUS) {
421  av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
422  return -1;
423  }
424 
425  if (!st->codec->extradata || !st->codec->extradata_size) {
426  av_log(s, AV_LOG_ERROR, "No extradata present\n");
427  return -1;
428  }
429  oggstream = av_mallocz(sizeof(*oggstream));
430  oggstream->page.stream_index = i;
431 
432  if (!(st->codec->flags & CODEC_FLAG_BITEXACT))
433  do {
434  serial_num = av_get_random_seed();
435  for (j = 0; j < i; j++) {
436  OGGStreamContext *sc = s->streams[j]->priv_data;
437  if (serial_num == sc->serial_num)
438  break;
439  }
440  } while (j < i);
441  oggstream->serial_num = serial_num;
442 
444 
445  st->priv_data = oggstream;
446  if (st->codec->codec_id == AV_CODEC_ID_FLAC) {
447  int err = ogg_build_flac_headers(st->codec, oggstream,
449  &st->metadata);
450  if (err) {
451  av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
452  av_freep(&st->priv_data);
453  return err;
454  }
455  } else if (st->codec->codec_id == AV_CODEC_ID_SPEEX) {
456  int err = ogg_build_speex_headers(st->codec, oggstream,
458  &st->metadata);
459  if (err) {
460  av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
461  av_freep(&st->priv_data);
462  return err;
463  }
464  } else if (st->codec->codec_id == AV_CODEC_ID_OPUS) {
465  int err = ogg_build_opus_headers(st->codec, oggstream,
467  &st->metadata);
468  if (err) {
469  av_log(s, AV_LOG_ERROR, "Error writing Opus headers\n");
470  av_freep(&st->priv_data);
471  return err;
472  }
473  } else {
474  uint8_t *p;
475  const char *cstr = st->codec->codec_id == AV_CODEC_ID_VORBIS ? "vorbis" : "theora";
476  int header_type = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 3 : 0x81;
477  int framing_bit = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 1 : 0;
478 
480  st->codec->codec_id == AV_CODEC_ID_VORBIS ? 30 : 42,
481  oggstream->header, oggstream->header_len) < 0) {
482  av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
483  av_freep(&st->priv_data);
484  return -1;
485  }
486 
488  &oggstream->header_len[1], &st->metadata,
489  framing_bit);
490  oggstream->header[1] = p;
491  if (!p)
492  return AVERROR(ENOMEM);
493 
494  bytestream_put_byte(&p, header_type);
495  bytestream_put_buffer(&p, cstr, 6);
496 
497  if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
498  /** KFGSHIFT is the width of the less significant section of the granule position
499  The less significant section is the frame count since the last keyframe */
500  oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
501  oggstream->vrev = oggstream->header[0][9];
502  av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
503  oggstream->kfgshift, oggstream->vrev);
504  }
505  }
506  }
507 
508  for (j = 0; j < s->nb_streams; j++) {
509  OGGStreamContext *oggstream = s->streams[j]->priv_data;
510  ogg_buffer_data(s, s->streams[j], oggstream->header[0],
511  oggstream->header_len[0], 0, 1);
512  oggstream->page.flags |= 2; // bos
513  ogg_buffer_page(s, oggstream);
514  }
515  for (j = 0; j < s->nb_streams; j++) {
516  AVStream *st = s->streams[j];
517  OGGStreamContext *oggstream = st->priv_data;
518  for (i = 1; i < 3; i++) {
519  if (oggstream->header_len[i])
520  ogg_buffer_data(s, st, oggstream->header[i],
521  oggstream->header_len[i], 0, 1);
522  }
523  ogg_buffer_page(s, oggstream);
524  }
525 
526  oggstream->page.start_granule = AV_NOPTS_VALUE;
527 
528  return 0;
529 }
530 
532 {
533  OGGContext *ogg = s->priv_data;
534  OGGPageList *next, *p;
535 
536  if (!ogg->page_list)
537  return;
538 
539  for (p = ogg->page_list; p; ) {
540  OGGStreamContext *oggstream =
542  if (oggstream->page_count < 2 && !flush)
543  break;
544  ogg_write_page(s, &p->page,
545  flush && oggstream->page_count == 1 ? 4 : 0); // eos
546  next = p->next;
547  av_freep(&p);
548  p = next;
549  }
550  ogg->page_list = p;
551 }
552 
554 {
555  AVStream *st = s->streams[pkt->stream_index];
556  OGGStreamContext *oggstream = st->priv_data;
557  int ret;
558  int64_t granule;
559 
560  if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
561  int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
562  int pframe_count;
563  if (pkt->flags & AV_PKT_FLAG_KEY)
564  oggstream->last_kf_pts = pts;
565  pframe_count = pts - oggstream->last_kf_pts;
566  // prevent frame count from overflow if key frame flag is not set
567  if (pframe_count >= (1<<oggstream->kfgshift)) {
568  oggstream->last_kf_pts += pframe_count;
569  pframe_count = 0;
570  }
571  granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
572  } else if (st->codec->codec_id == AV_CODEC_ID_OPUS)
573  granule = pkt->pts + pkt->duration + av_rescale_q(st->codec->delay, (AVRational){ 1, st->codec->sample_rate }, st->time_base);
574  else
575  granule = pkt->pts + pkt->duration;
576 
577  if (oggstream->page.start_granule == AV_NOPTS_VALUE)
578  oggstream->page.start_granule = pkt->pts;
579 
580  ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule, 0);
581  if (ret < 0)
582  return ret;
583 
584  ogg_write_pages(s, 0);
585 
586  oggstream->last_granule = granule;
587 
588  return 0;
589 }
590 
592 {
593  int i;
594 
595  /* flush current page if needed */
596  for (i = 0; i < s->nb_streams; i++) {
597  OGGStreamContext *oggstream = s->streams[i]->priv_data;
598 
599  if (oggstream->page.size > 0)
600  ogg_buffer_page(s, oggstream);
601  }
602 
603  ogg_write_pages(s, 1);
604 
605  for (i = 0; i < s->nb_streams; i++) {
606  AVStream *st = s->streams[i];
607  OGGStreamContext *oggstream = st->priv_data;
608  if (st->codec->codec_id == AV_CODEC_ID_FLAC ||
609  st->codec->codec_id == AV_CODEC_ID_SPEEX ||
610  st->codec->codec_id == AV_CODEC_ID_OPUS) {
611  av_freep(&oggstream->header[0]);
612  }
613  av_freep(&oggstream->header[1]);
614  av_freep(&st->priv_data);
615  }
616  return 0;
617 }
618 
619 #if CONFIG_OGG_MUXER
620 static const AVClass ogg_muxer_class = {
621  .class_name = "Ogg muxer",
622  .item_name = av_default_item_name,
623  .option = options,
624  .version = LIBAVUTIL_VERSION_INT,
625 };
626 
627 AVOutputFormat ff_ogg_muxer = {
628  .name = "ogg",
629  .long_name = NULL_IF_CONFIG_SMALL("Ogg"),
630  .mime_type = "application/ogg",
631  .extensions = "ogg,ogv"
632 #if !CONFIG_SPEEX_MUXER
633  ",spx"
634 #endif
635 #if !CONFIG_OPUS_MUXER
636  ",opus"
637 #endif
638  ,
639  .priv_data_size = sizeof(OGGContext),
640  .audio_codec = CONFIG_LIBVORBIS_ENCODER ?
642  .video_codec = AV_CODEC_ID_THEORA,
647  .priv_class = &ogg_muxer_class,
648 };
649 #endif
650 
651 #if CONFIG_SPEEX_MUXER
652 static const AVClass speex_muxer_class = {
653  .class_name = "Speex muxer",
654  .item_name = av_default_item_name,
655  .option = options,
656  .version = LIBAVUTIL_VERSION_INT,
657 };
658 
659 AVOutputFormat ff_speex_muxer = {
660  .name = "speex",
661  .long_name = NULL_IF_CONFIG_SMALL("Speex"),
662  .mime_type = "audio/ogg",
663  .extensions = "spx",
664  .priv_data_size = sizeof(OGGContext),
665  .audio_codec = AV_CODEC_ID_SPEEX,
666  .video_codec = AV_CODEC_ID_NONE,
671  .priv_class = &speex_muxer_class,
672 };
673 #endif
674 
675 #if CONFIG_OPUS_MUXER
676 static const AVClass opus_muxer_class = {
677  .class_name = "Opus muxer",
678  .item_name = av_default_item_name,
679  .option = options,
680  .version = LIBAVUTIL_VERSION_INT,
681 };
682 
683 AVOutputFormat ff_opus_muxer = {
684  .name = "opus",
685  .long_name = NULL_IF_CONFIG_SMALL("Opus"),
686  .mime_type = "audio/ogg",
687  .extensions = "opus",
688  .priv_data_size = sizeof(OGGContext),
689  .audio_codec = AV_CODEC_ID_OPUS,
690  .video_codec = AV_CODEC_ID_NONE,
695  .priv_class = &opus_muxer_class,
696 };
697 #endif