3 * Copyright (c) 2007 Baptiste Coudurier <baptiste dot coudurier at free dot fr>
5 * This file is part of FFmpeg.
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.
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.
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
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"
31 #include "avio_internal.h"
33 #include "vorbiscomment.h"
35 #define MAX_PAGE_SIZE 65025
38 int64_t start_granule;
42 uint8_t segments_count;
43 uint8_t segments[255];
44 uint8_t data[MAX_PAGE_SIZE];
49 unsigned page_counter;
52 /** for theora granule */
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
63 typedef struct OGGPageList {
65 struct OGGPageList *next;
70 OGGPageList *page_list;
71 int pref_size; ///< preferred page size (0 => fill all segments)
72 int64_t pref_duration; ///< preferred page duration (0 => fill all segments)
75 #define OFFSET(x) offsetof(OGGContext, x)
76 #define PARAM AV_OPT_FLAG_ENCODING_PARAM
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_INT, { .i64 = 1000000 }, 0, INT64_MAX, PARAM },
88 static const AVClass ogg_muxer_class = {
89 .class_name = "Ogg muxer",
90 .item_name = av_default_item_name,
92 .version = LIBAVUTIL_VERSION_INT,
96 static void ogg_update_checksum(AVFormatContext *s, AVIOContext *pb, int64_t crc_offset)
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);
105 static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
107 OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
113 ret = avio_open_dyn_buf(&pb);
116 ffio_init_checksum(pb, ff_crc04C11DB7_update, 0);
117 ffio_wfourcc(pb, "OggS");
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);
129 ogg_update_checksum(s, pb, crc_offset);
132 size = avio_close_dyn_buf(pb, &buf);
136 avio_write(s->pb, buf, size);
139 oggstream->page_count--;
143 static int ogg_key_granule(OGGStreamContext *oggstream, int64_t granule)
145 return oggstream->kfgshift && !(granule & ((1<<oggstream->kfgshift)-1));
148 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, int64_t granule)
150 if (oggstream->kfgshift)
151 return (granule>>oggstream->kfgshift) +
152 (granule & ((1<<oggstream->kfgshift)-1));
157 static int ogg_compare_granule(AVFormatContext *s, OGGPage *next, OGGPage *page)
159 AVStream *st2 = s->streams[next->stream_index];
160 AVStream *st = s->streams[page->stream_index];
161 int64_t next_granule, cur_granule;
163 if (next->granule == -1 || page->granule == -1)
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;
173 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
175 oggstream->page.granule = -1;
176 oggstream->page.flags = 0;
177 oggstream->page.segments_count = 0;
178 oggstream->page.size = 0;
182 static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
184 OGGContext *ogg = s->priv_data;
185 OGGPageList **p = &ogg->page_list;
186 OGGPageList *l = av_mallocz(sizeof(*l));
189 return AVERROR(ENOMEM);
190 l->page = oggstream->page;
192 oggstream->page.start_granule = oggstream->page.granule;
193 oggstream->page_count++;
194 ogg_reset_cur_page(oggstream);
197 if (ogg_compare_granule(s, &(*p)->page, &l->page))
207 static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
208 uint8_t *data, unsigned size, int64_t granule,
211 OGGStreamContext *oggstream = st->priv_data;
212 OGGContext *ogg = s->priv_data;
213 int total_segments = size / 255 + 1;
215 int i, segments, len, flush = 0;
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);
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);
237 for (i = 0; i < total_segments; ) {
238 OGGPage *page = &oggstream->page;
240 segments = FFMIN(total_segments - i, 255 - page->segments_count);
242 if (i && !page->segments_count)
243 page->flags |= 1; // continued packet
245 memset(page->segments+page->segments_count, 255, segments - 1);
246 page->segments_count += segments - 1;
248 len = FFMIN(size, segments*255);
249 page->segments[page->segments_count++] = len - (segments-1)*255;
250 memcpy(page->data+page->size, p, len);
256 if (i == total_segments)
257 page->granule = granule;
260 AVStream *st = s->streams[page->stream_index];
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,
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);
275 if (flush && oggstream->page.granule != -1)
276 ogg_buffer_page(s, oggstream);
281 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
282 int *header_len, AVDictionary **m, int framing_bit)
284 const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
289 ff_metadata_conv(m, ff_vorbiscomment_metadata_conv, NULL);
291 size = offset + ff_vorbiscomment_length(*m, vendor, &count) + framing_bit;
292 p = av_mallocz(size);
298 ff_vorbiscomment_write(&p, m, vendor, count);
300 bytestream_put_byte(&p, 1);
306 static int ogg_build_flac_headers(AVCodecContext *avctx,
307 OGGStreamContext *oggstream, int bitexact,
310 enum FLACExtradataFormat format;
314 if (!avpriv_flac_is_extradata_valid(avctx, &format, &streaminfo))
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];
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);
331 bytestream_put_buffer(&p, streaminfo, FLAC_STREAMINFO_SIZE);
333 // second packet: VorbisComment
334 p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
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);
344 #define SPEEX_HEADER_SIZE 80
346 static int ogg_build_speex_headers(AVCodecContext *avctx,
347 OGGStreamContext *oggstream, int bitexact,
352 if (avctx->extradata_size < SPEEX_HEADER_SIZE)
355 // first packet: Speex header
356 p = av_mallocz(SPEEX_HEADER_SIZE);
358 return AVERROR(ENOMEM);
359 oggstream->header[0] = p;
360 oggstream->header_len[0] = SPEEX_HEADER_SIZE;
361 bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
362 AV_WL32(&oggstream->header[0][68], 0); // set extra_headers to 0
364 // second packet: VorbisComment
365 p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
367 return AVERROR(ENOMEM);
368 oggstream->header[1] = p;
373 #define OPUS_HEADER_SIZE 19
375 static int ogg_build_opus_headers(AVCodecContext *avctx,
376 OGGStreamContext *oggstream, int bitexact,
381 if (avctx->extradata_size < OPUS_HEADER_SIZE)
384 /* first packet: Opus header */
385 p = av_mallocz(avctx->extradata_size);
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);
392 /* second packet: VorbisComment */
393 p = ogg_write_vorbiscomment(8, bitexact, &oggstream->header_len[1], m, 0);
395 return AVERROR(ENOMEM);
396 oggstream->header[1] = p;
397 bytestream_put_buffer(&p, "OpusTags", 8);
402 static int ogg_write_header(AVFormatContext *s)
404 OGGContext *ogg = s->priv_data;
405 OGGStreamContext *oggstream;
409 av_log(s, AV_LOG_WARNING, "The pagesize option is deprecated\n");
411 for (i = 0; i < s->nb_streams; i++) {
412 AVStream *st = s->streams[i];
413 unsigned serial_num = i;
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);
420 avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
421 } else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
422 avpriv_set_pts_info(st, 64, st->codec->time_base.num, st->codec->time_base.den);
423 if (st->codec->codec_id != AV_CODEC_ID_VORBIS &&
424 st->codec->codec_id != AV_CODEC_ID_THEORA &&
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);
432 if (!st->codec->extradata || !st->codec->extradata_size) {
433 av_log(s, AV_LOG_ERROR, "No extradata present\n");
436 oggstream = av_mallocz(sizeof(*oggstream));
437 oggstream->page.stream_index = i;
439 if (!(st->codec->flags & CODEC_FLAG_BITEXACT))
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)
448 oggstream->serial_num = serial_num;
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,
453 st->codec->flags & CODEC_FLAG_BITEXACT,
456 av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
457 av_freep(&st->priv_data);
460 } else if (st->codec->codec_id == AV_CODEC_ID_SPEEX) {
461 int err = ogg_build_speex_headers(st->codec, oggstream,
462 st->codec->flags & CODEC_FLAG_BITEXACT,
465 av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
466 av_freep(&st->priv_data);
469 } else if (st->codec->codec_id == AV_CODEC_ID_OPUS) {
470 int err = ogg_build_opus_headers(st->codec, oggstream,
471 st->codec->flags & CODEC_FLAG_BITEXACT,
474 av_log(s, AV_LOG_ERROR, "Error writing Opus headers\n");
475 av_freep(&st->priv_data);
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;
484 if (avpriv_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
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);
492 p = ogg_write_vorbiscomment(7, st->codec->flags & CODEC_FLAG_BITEXACT,
493 &oggstream->header_len[1], &s->metadata,
495 oggstream->header[1] = p;
497 return AVERROR(ENOMEM);
499 bytestream_put_byte(&p, header_type);
500 bytestream_put_buffer(&p, cstr, 6);
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);
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);
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);
528 ogg_buffer_page(s, oggstream);
531 oggstream->page.start_granule = AV_NOPTS_VALUE;
536 static void ogg_write_pages(AVFormatContext *s, int flush)
538 OGGContext *ogg = s->priv_data;
539 OGGPageList *next, *p;
544 for (p = ogg->page_list; p; ) {
545 OGGStreamContext *oggstream =
546 s->streams[p->page.stream_index]->priv_data;
547 if (oggstream->page_count < 2 && !flush)
549 ogg_write_page(s, &p->page,
550 flush && oggstream->page_count == 1 ? 4 : 0); // eos
558 static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
560 AVStream *st = s->streams[pkt->stream_index];
561 OGGStreamContext *oggstream = st->priv_data;
565 if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
566 int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
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;
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);
580 granule = pkt->pts + pkt->duration;
582 if (oggstream->page.start_granule == AV_NOPTS_VALUE)
583 oggstream->page.start_granule = pkt->pts;
585 ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule, 0);
589 ogg_write_pages(s, 0);
591 oggstream->last_granule = granule;
596 static int ogg_write_trailer(AVFormatContext *s)
600 /* flush current page if needed */
601 for (i = 0; i < s->nb_streams; i++) {
602 OGGStreamContext *oggstream = s->streams[i]->priv_data;
604 if (oggstream->page.size > 0)
605 ogg_buffer_page(s, oggstream);
608 ogg_write_pages(s, 1);
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]);
618 av_freep(&oggstream->header[1]);
619 av_freep(&st->priv_data);
624 AVOutputFormat ff_ogg_muxer = {
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,
632 .write_header = ogg_write_header,
633 .write_packet = ogg_write_packet,
634 .write_trailer = ogg_write_trailer,
635 .priv_class = &ogg_muxer_class,