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
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"
32 #include "avio_internal.h"
34 #include "vorbiscomment.h"
36 #define MAX_PAGE_SIZE 65025
38 typedef struct OGGPage {
39 int64_t start_granule;
43 uint8_t segments_count;
44 uint8_t segments[255];
45 uint8_t data[MAX_PAGE_SIZE];
49 typedef struct OGGStreamContext {
50 unsigned page_counter;
53 /** for theora granule */
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
64 typedef struct OGGPageList {
66 struct OGGPageList *next;
69 typedef struct OGGContext {
71 OGGPageList *page_list;
72 int pref_size; ///< preferred page size (0 => fill all segments)
73 int64_t pref_duration; ///< preferred page duration (0 => fill all segments)
77 #define OFFSET(x) offsetof(OGGContext, x)
78 #define PARAM AV_OPT_FLAG_ENCODING_PARAM
80 static const AVOption options[] = {
81 { "serial_offset", "serial number offset",
82 OFFSET(serial_offset), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, PARAM },
83 { "oggpagesize", "Set preferred Ogg page size.",
84 OFFSET(pref_size), AV_OPT_TYPE_INT, {.i64 = 0}, 0, MAX_PAGE_SIZE, PARAM},
85 { "pagesize", "preferred page size in bytes (deprecated)",
86 OFFSET(pref_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, MAX_PAGE_SIZE, PARAM },
87 { "page_duration", "preferred page duration, in microseconds",
88 OFFSET(pref_duration), AV_OPT_TYPE_INT64, { .i64 = 1000000 }, 0, INT64_MAX, PARAM },
92 #define OGG_CLASS(flavor, name)\
93 static const AVClass flavor ## _muxer_class = {\
94 .class_name = #name " muxer",\
95 .item_name = av_default_item_name,\
97 .version = LIBAVUTIL_VERSION_INT,\
100 static void ogg_update_checksum(AVFormatContext *s, AVIOContext *pb, int64_t crc_offset)
102 int64_t pos = avio_tell(pb);
103 uint32_t checksum = ffio_get_checksum(pb);
104 avio_seek(pb, crc_offset, SEEK_SET);
105 avio_wb32(pb, checksum);
106 avio_seek(pb, pos, SEEK_SET);
109 static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
111 OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
117 ret = avio_open_dyn_buf(&pb);
120 ffio_init_checksum(pb, ff_crc04C11DB7_update, 0);
121 ffio_wfourcc(pb, "OggS");
123 avio_w8(pb, page->flags | extra_flags);
124 avio_wl64(pb, page->granule);
125 avio_wl32(pb, oggstream->serial_num);
126 avio_wl32(pb, oggstream->page_counter++);
127 crc_offset = avio_tell(pb);
128 avio_wl32(pb, 0); // crc
129 avio_w8(pb, page->segments_count);
130 avio_write(pb, page->segments, page->segments_count);
131 avio_write(pb, page->data, page->size);
133 ogg_update_checksum(s, pb, crc_offset);
136 size = avio_close_dyn_buf(pb, &buf);
140 avio_write(s->pb, buf, size);
143 oggstream->page_count--;
147 static int ogg_key_granule(OGGStreamContext *oggstream, int64_t granule)
149 return oggstream->kfgshift && !(granule & ((1<<oggstream->kfgshift)-1));
152 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, int64_t granule)
154 if (oggstream->kfgshift)
155 return (granule>>oggstream->kfgshift) +
156 (granule & ((1<<oggstream->kfgshift)-1));
161 static int ogg_compare_granule(AVFormatContext *s, OGGPage *next, OGGPage *page)
163 AVStream *st2 = s->streams[next->stream_index];
164 AVStream *st = s->streams[page->stream_index];
165 int64_t next_granule, cur_granule;
167 if (next->granule == -1 || page->granule == -1)
170 next_granule = av_rescale_q(ogg_granule_to_timestamp(st2->priv_data, next->granule),
171 st2->time_base, AV_TIME_BASE_Q);
172 cur_granule = av_rescale_q(ogg_granule_to_timestamp(st->priv_data, page->granule),
173 st ->time_base, AV_TIME_BASE_Q);
174 return next_granule > cur_granule;
177 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
179 oggstream->page.granule = -1;
180 oggstream->page.flags = 0;
181 oggstream->page.segments_count = 0;
182 oggstream->page.size = 0;
186 static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
188 OGGContext *ogg = s->priv_data;
189 OGGPageList **p = &ogg->page_list;
190 OGGPageList *l = av_mallocz(sizeof(*l));
193 return AVERROR(ENOMEM);
194 l->page = oggstream->page;
196 oggstream->page.start_granule = oggstream->page.granule;
197 oggstream->page_count++;
198 ogg_reset_cur_page(oggstream);
201 if (ogg_compare_granule(s, &(*p)->page, &l->page))
211 static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
212 uint8_t *data, unsigned size, int64_t granule,
215 OGGStreamContext *oggstream = st->priv_data;
216 OGGContext *ogg = s->priv_data;
217 int total_segments = size / 255 + 1;
219 int i, segments, len, flush = 0;
221 // Handles VFR by flushing page because this frame needs to have a timestamp
222 // For theora, keyframes also need to have a timestamp to correctly mark
223 // them as such, otherwise seeking will not work correctly at the very
224 // least with old libogg versions.
225 // Do not try to flush header packets though, that will create broken files.
226 if (st->codec->codec_id == AV_CODEC_ID_THEORA && !header &&
227 (ogg_granule_to_timestamp(oggstream, granule) >
228 ogg_granule_to_timestamp(oggstream, oggstream->last_granule) + 1 ||
229 ogg_key_granule(oggstream, granule))) {
230 if (oggstream->page.granule != -1)
231 ogg_buffer_page(s, oggstream);
235 // avoid a continued page
236 if (!header && oggstream->page.size > 0 &&
237 MAX_PAGE_SIZE - oggstream->page.size < size) {
238 ogg_buffer_page(s, oggstream);
241 for (i = 0; i < total_segments; ) {
242 OGGPage *page = &oggstream->page;
244 segments = FFMIN(total_segments - i, 255 - page->segments_count);
246 if (i && !page->segments_count)
247 page->flags |= 1; // continued packet
249 memset(page->segments+page->segments_count, 255, segments - 1);
250 page->segments_count += segments - 1;
252 len = FFMIN(size, segments*255);
253 page->segments[page->segments_count++] = len - (segments-1)*255;
254 memcpy(page->data+page->size, p, len);
260 if (i == total_segments)
261 page->granule = granule;
264 AVStream *st = s->streams[page->stream_index];
266 int64_t start = av_rescale_q(page->start_granule, st->time_base,
268 int64_t next = av_rescale_q(page->granule, st->time_base,
271 if (page->segments_count == 255 ||
272 (ogg->pref_size > 0 && page->size >= ogg->pref_size) ||
273 (ogg->pref_duration > 0 && next - start >= ogg->pref_duration)) {
274 ogg_buffer_page(s, oggstream);
279 if (flush && oggstream->page.granule != -1)
280 ogg_buffer_page(s, oggstream);
285 static uint8_t *ogg_write_vorbiscomment(int64_t offset, int bitexact,
286 int *header_len, AVDictionary **m, int framing_bit)
288 const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
292 ff_metadata_conv(m, ff_vorbiscomment_metadata_conv, NULL);
294 size = offset + ff_vorbiscomment_length(*m, vendor) + framing_bit;
297 p = av_mallocz(size);
303 ff_vorbiscomment_write(&p, m, vendor);
305 bytestream_put_byte(&p, 1);
311 static int ogg_build_flac_headers(AVCodecContext *avctx,
312 OGGStreamContext *oggstream, int bitexact,
317 if (avctx->extradata_size < FLAC_STREAMINFO_SIZE)
318 return AVERROR(EINVAL);
320 // first packet: STREAMINFO
321 oggstream->header_len[0] = 51;
322 oggstream->header[0] = av_mallocz(51); // per ogg flac specs
323 p = oggstream->header[0];
325 return AVERROR(ENOMEM);
326 bytestream_put_byte(&p, 0x7F);
327 bytestream_put_buffer(&p, "FLAC", 4);
328 bytestream_put_byte(&p, 1); // major version
329 bytestream_put_byte(&p, 0); // minor version
330 bytestream_put_be16(&p, 1); // headers packets without this one
331 bytestream_put_buffer(&p, "fLaC", 4);
332 bytestream_put_byte(&p, 0x00); // streaminfo
333 bytestream_put_be24(&p, 34);
334 bytestream_put_buffer(&p, avctx->extradata, FLAC_STREAMINFO_SIZE);
336 // second packet: VorbisComment
337 p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
339 return AVERROR(ENOMEM);
340 oggstream->header[1] = p;
341 bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
342 bytestream_put_be24(&p, oggstream->header_len[1] - 4);
347 #define SPEEX_HEADER_SIZE 80
349 static int ogg_build_speex_headers(AVCodecContext *avctx,
350 OGGStreamContext *oggstream, int bitexact,
355 if (avctx->extradata_size < SPEEX_HEADER_SIZE)
356 return AVERROR_INVALIDDATA;
358 // first packet: Speex header
359 p = av_mallocz(SPEEX_HEADER_SIZE);
361 return AVERROR(ENOMEM);
362 oggstream->header[0] = p;
363 oggstream->header_len[0] = SPEEX_HEADER_SIZE;
364 bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
365 AV_WL32(&oggstream->header[0][68], 0); // set extra_headers to 0
367 // second packet: VorbisComment
368 p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
370 return AVERROR(ENOMEM);
371 oggstream->header[1] = p;
376 #define OPUS_HEADER_SIZE 19
378 static int ogg_build_opus_headers(AVCodecContext *avctx,
379 OGGStreamContext *oggstream, int bitexact,
384 if (avctx->extradata_size < OPUS_HEADER_SIZE)
385 return AVERROR_INVALIDDATA;
387 /* first packet: Opus header */
388 p = av_mallocz(avctx->extradata_size);
390 return AVERROR(ENOMEM);
391 oggstream->header[0] = p;
392 oggstream->header_len[0] = avctx->extradata_size;
393 bytestream_put_buffer(&p, avctx->extradata, avctx->extradata_size);
395 /* second packet: VorbisComment */
396 p = ogg_write_vorbiscomment(8, bitexact, &oggstream->header_len[1], m, 0);
398 return AVERROR(ENOMEM);
399 oggstream->header[1] = p;
400 bytestream_put_buffer(&p, "OpusTags", 8);
405 static void ogg_write_pages(AVFormatContext *s, int flush)
407 OGGContext *ogg = s->priv_data;
408 OGGPageList *next, *p;
413 for (p = ogg->page_list; p; ) {
414 OGGStreamContext *oggstream =
415 s->streams[p->page.stream_index]->priv_data;
416 if (oggstream->page_count < 2 && !flush)
418 ogg_write_page(s, &p->page,
419 flush == 1 && oggstream->page_count == 1 ? 4 : 0); // eos
427 static int ogg_write_header(AVFormatContext *s)
429 OGGContext *ogg = s->priv_data;
430 OGGStreamContext *oggstream = NULL;
434 av_log(s, AV_LOG_WARNING, "The pagesize option is deprecated\n");
436 for (i = 0; i < s->nb_streams; i++) {
437 AVStream *st = s->streams[i];
438 unsigned serial_num = i + ogg->serial_offset;
440 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
441 if (st->codec->codec_id == AV_CODEC_ID_OPUS)
442 /* Opus requires a fixed 48kHz clock */
443 avpriv_set_pts_info(st, 64, 1, 48000);
445 avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
448 if (st->codec->codec_id != AV_CODEC_ID_VORBIS &&
449 st->codec->codec_id != AV_CODEC_ID_THEORA &&
450 st->codec->codec_id != AV_CODEC_ID_SPEEX &&
451 st->codec->codec_id != AV_CODEC_ID_FLAC &&
452 st->codec->codec_id != AV_CODEC_ID_OPUS) {
453 av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
454 return AVERROR(EINVAL);
457 if (!st->codec->extradata || !st->codec->extradata_size) {
458 av_log(s, AV_LOG_ERROR, "No extradata present\n");
459 return AVERROR_INVALIDDATA;
461 oggstream = av_mallocz(sizeof(*oggstream));
463 return AVERROR(ENOMEM);
465 oggstream->page.stream_index = i;
467 if (!(s->flags & AVFMT_FLAG_BITEXACT))
469 serial_num = av_get_random_seed();
470 for (j = 0; j < i; j++) {
471 OGGStreamContext *sc = s->streams[j]->priv_data;
472 if (serial_num == sc->serial_num)
476 oggstream->serial_num = serial_num;
478 av_dict_copy(&st->metadata, s->metadata, AV_DICT_DONT_OVERWRITE);
480 st->priv_data = oggstream;
481 if (st->codec->codec_id == AV_CODEC_ID_FLAC) {
482 int err = ogg_build_flac_headers(st->codec, oggstream,
483 s->flags & AVFMT_FLAG_BITEXACT,
486 av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
487 av_freep(&st->priv_data);
490 } else if (st->codec->codec_id == AV_CODEC_ID_SPEEX) {
491 int err = ogg_build_speex_headers(st->codec, oggstream,
492 s->flags & AVFMT_FLAG_BITEXACT,
495 av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
496 av_freep(&st->priv_data);
499 } else if (st->codec->codec_id == AV_CODEC_ID_OPUS) {
500 int err = ogg_build_opus_headers(st->codec, oggstream,
501 s->flags & AVFMT_FLAG_BITEXACT,
504 av_log(s, AV_LOG_ERROR, "Error writing Opus headers\n");
505 av_freep(&st->priv_data);
510 const char *cstr = st->codec->codec_id == AV_CODEC_ID_VORBIS ? "vorbis" : "theora";
511 int header_type = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 3 : 0x81;
512 int framing_bit = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 1 : 0;
514 if (avpriv_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
515 st->codec->codec_id == AV_CODEC_ID_VORBIS ? 30 : 42,
516 (const uint8_t**)oggstream->header, oggstream->header_len) < 0) {
517 av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
518 av_freep(&st->priv_data);
519 return AVERROR_INVALIDDATA;
522 p = ogg_write_vorbiscomment(7, s->flags & AVFMT_FLAG_BITEXACT,
523 &oggstream->header_len[1], &st->metadata,
525 oggstream->header[1] = p;
527 return AVERROR(ENOMEM);
529 bytestream_put_byte(&p, header_type);
530 bytestream_put_buffer(&p, cstr, 6);
532 if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
533 /** KFGSHIFT is the width of the less significant section of the granule position
534 The less significant section is the frame count since the last keyframe */
535 oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
536 oggstream->vrev = oggstream->header[0][9];
537 av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
538 oggstream->kfgshift, oggstream->vrev);
543 for (j = 0; j < s->nb_streams; j++) {
544 OGGStreamContext *oggstream = s->streams[j]->priv_data;
545 ogg_buffer_data(s, s->streams[j], oggstream->header[0],
546 oggstream->header_len[0], 0, 1);
547 oggstream->page.flags |= 2; // bos
548 ogg_buffer_page(s, oggstream);
550 for (j = 0; j < s->nb_streams; j++) {
551 AVStream *st = s->streams[j];
552 OGGStreamContext *oggstream = st->priv_data;
553 for (i = 1; i < 3; i++) {
554 if (oggstream->header_len[i])
555 ogg_buffer_data(s, st, oggstream->header[i],
556 oggstream->header_len[i], 0, 1);
558 ogg_buffer_page(s, oggstream);
561 oggstream->page.start_granule = AV_NOPTS_VALUE;
563 ogg_write_pages(s, 2);
568 static int ogg_write_packet_internal(AVFormatContext *s, AVPacket *pkt)
570 AVStream *st = s->streams[pkt->stream_index];
571 OGGStreamContext *oggstream = st->priv_data;
575 if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
576 int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
578 if (pkt->flags & AV_PKT_FLAG_KEY)
579 oggstream->last_kf_pts = pts;
580 pframe_count = pts - oggstream->last_kf_pts;
581 // prevent frame count from overflow if key frame flag is not set
582 if (pframe_count >= (1<<oggstream->kfgshift)) {
583 oggstream->last_kf_pts += pframe_count;
586 granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
587 } else if (st->codec->codec_id == AV_CODEC_ID_OPUS)
588 granule = pkt->pts + pkt->duration +
589 av_rescale_q(st->codec->initial_padding,
590 (AVRational){ 1, st->codec->sample_rate },
593 granule = pkt->pts + pkt->duration;
595 if (oggstream->page.start_granule == AV_NOPTS_VALUE)
596 oggstream->page.start_granule = pkt->pts;
598 ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule, 0);
602 ogg_write_pages(s, 0);
604 oggstream->last_granule = granule;
609 static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
614 return ogg_write_packet_internal(s, pkt);
616 for (i = 0; i < s->nb_streams; i++) {
617 OGGStreamContext *oggstream = s->streams[i]->priv_data;
618 if (oggstream->page.segments_count)
619 ogg_buffer_page(s, oggstream);
622 ogg_write_pages(s, 2);
626 static int ogg_write_trailer(AVFormatContext *s)
630 /* flush current page if needed */
631 for (i = 0; i < s->nb_streams; i++) {
632 OGGStreamContext *oggstream = s->streams[i]->priv_data;
634 if (oggstream->page.size > 0)
635 ogg_buffer_page(s, oggstream);
638 ogg_write_pages(s, 1);
640 for (i = 0; i < s->nb_streams; i++) {
641 AVStream *st = s->streams[i];
642 OGGStreamContext *oggstream = st->priv_data;
643 if (st->codec->codec_id == AV_CODEC_ID_FLAC ||
644 st->codec->codec_id == AV_CODEC_ID_SPEEX ||
645 st->codec->codec_id == AV_CODEC_ID_OPUS) {
646 av_freep(&oggstream->header[0]);
648 av_freep(&oggstream->header[1]);
649 av_freep(&st->priv_data);
656 AVOutputFormat ff_ogg_muxer = {
658 .long_name = NULL_IF_CONFIG_SMALL("Ogg"),
659 .mime_type = "application/ogg",
660 .extensions = "ogg,ogv"
661 #if !CONFIG_SPX_MUXER
664 #if !CONFIG_OPUS_MUXER
668 .priv_data_size = sizeof(OGGContext),
669 .audio_codec = CONFIG_LIBVORBIS_ENCODER ?
670 AV_CODEC_ID_VORBIS : AV_CODEC_ID_FLAC,
671 .video_codec = AV_CODEC_ID_THEORA,
672 .write_header = ogg_write_header,
673 .write_packet = ogg_write_packet,
674 .write_trailer = ogg_write_trailer,
675 .flags = AVFMT_TS_NEGATIVE | AVFMT_ALLOW_FLUSH,
676 .priv_class = &ogg_muxer_class,
681 OGG_CLASS(oga, Ogg audio)
682 AVOutputFormat ff_oga_muxer = {
684 .long_name = NULL_IF_CONFIG_SMALL("Ogg Audio"),
685 .mime_type = "audio/ogg",
687 .priv_data_size = sizeof(OGGContext),
688 .audio_codec = CONFIG_LIBVORBIS_ENCODER ?
689 AV_CODEC_ID_VORBIS : AV_CODEC_ID_FLAC,
690 .write_header = ogg_write_header,
691 .write_packet = ogg_write_packet,
692 .write_trailer = ogg_write_trailer,
693 .flags = AVFMT_TS_NEGATIVE | AVFMT_ALLOW_FLUSH,
694 .priv_class = &oga_muxer_class,
699 OGG_CLASS(spx, Ogg Speex)
700 AVOutputFormat ff_spx_muxer = {
702 .long_name = NULL_IF_CONFIG_SMALL("Ogg Speex"),
703 .mime_type = "audio/ogg",
705 .priv_data_size = sizeof(OGGContext),
706 .audio_codec = AV_CODEC_ID_SPEEX,
707 .write_header = ogg_write_header,
708 .write_packet = ogg_write_packet,
709 .write_trailer = ogg_write_trailer,
710 .flags = AVFMT_TS_NEGATIVE | AVFMT_ALLOW_FLUSH,
711 .priv_class = &spx_muxer_class,
715 #if CONFIG_OPUS_MUXER
716 OGG_CLASS(opus, Ogg Opus)
717 AVOutputFormat ff_opus_muxer = {
719 .long_name = NULL_IF_CONFIG_SMALL("Ogg Opus"),
720 .mime_type = "audio/ogg",
721 .extensions = "opus",
722 .priv_data_size = sizeof(OGGContext),
723 .audio_codec = AV_CODEC_ID_OPUS,
724 .write_header = ogg_write_header,
725 .write_packet = ogg_write_packet,
726 .write_trailer = ogg_write_trailer,
727 .flags = AVFMT_TS_NEGATIVE | AVFMT_ALLOW_FLUSH,
728 .priv_class = &opus_muxer_class,