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
39 int64_t start_granule;
43 uint8_t segments_count;
44 uint8_t segments[255];
45 uint8_t data[MAX_PAGE_SIZE];
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;
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)
76 #define OFFSET(x) offsetof(OGGContext, x)
77 #define PARAM AV_OPT_FLAG_ENCODING_PARAM
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 },
89 static void ogg_update_checksum(AVFormatContext *s, AVIOContext *pb, int64_t crc_offset)
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);
98 static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
100 OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
106 ret = avio_open_dyn_buf(&pb);
109 ffio_init_checksum(pb, ff_crc04C11DB7_update, 0);
110 ffio_wfourcc(pb, "OggS");
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);
122 ogg_update_checksum(s, pb, crc_offset);
125 size = avio_close_dyn_buf(pb, &buf);
129 avio_write(s->pb, buf, size);
132 oggstream->page_count--;
136 static int ogg_key_granule(OGGStreamContext *oggstream, int64_t granule)
138 return oggstream->kfgshift && !(granule & ((1<<oggstream->kfgshift)-1));
141 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, int64_t granule)
143 if (oggstream->kfgshift)
144 return (granule>>oggstream->kfgshift) +
145 (granule & ((1<<oggstream->kfgshift)-1));
150 static int ogg_compare_granule(AVFormatContext *s, OGGPage *next, OGGPage *page)
152 AVStream *st2 = s->streams[next->stream_index];
153 AVStream *st = s->streams[page->stream_index];
154 int64_t next_granule, cur_granule;
156 if (next->granule == -1 || page->granule == -1)
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;
166 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
168 oggstream->page.granule = -1;
169 oggstream->page.flags = 0;
170 oggstream->page.segments_count = 0;
171 oggstream->page.size = 0;
175 static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
177 OGGContext *ogg = s->priv_data;
178 OGGPageList **p = &ogg->page_list;
179 OGGPageList *l = av_mallocz(sizeof(*l));
182 return AVERROR(ENOMEM);
183 l->page = oggstream->page;
185 oggstream->page.start_granule = oggstream->page.granule;
186 oggstream->page_count++;
187 ogg_reset_cur_page(oggstream);
190 if (ogg_compare_granule(s, &(*p)->page, &l->page))
200 static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
201 uint8_t *data, unsigned size, int64_t granule,
204 OGGStreamContext *oggstream = st->priv_data;
205 OGGContext *ogg = s->priv_data;
206 int total_segments = size / 255 + 1;
208 int i, segments, len, flush = 0;
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);
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);
230 for (i = 0; i < total_segments; ) {
231 OGGPage *page = &oggstream->page;
233 segments = FFMIN(total_segments - i, 255 - page->segments_count);
235 if (i && !page->segments_count)
236 page->flags |= 1; // continued packet
238 memset(page->segments+page->segments_count, 255, segments - 1);
239 page->segments_count += segments - 1;
241 len = FFMIN(size, segments*255);
242 page->segments[page->segments_count++] = len - (segments-1)*255;
243 memcpy(page->data+page->size, p, len);
249 if (i == total_segments)
250 page->granule = granule;
253 AVStream *st = s->streams[page->stream_index];
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,
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);
268 if (flush && oggstream->page.granule != -1)
269 ogg_buffer_page(s, oggstream);
274 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
275 int *header_len, AVDictionary **m, int framing_bit)
277 const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
281 ff_metadata_conv(m, ff_vorbiscomment_metadata_conv, NULL);
283 size = offset + ff_vorbiscomment_length(*m, vendor) + framing_bit;
284 p = av_mallocz(size);
290 ff_vorbiscomment_write(&p, m, vendor);
292 bytestream_put_byte(&p, 1);
298 static int ogg_build_flac_headers(AVCodecContext *avctx,
299 OGGStreamContext *oggstream, int bitexact,
302 enum FLACExtradataFormat format;
306 if (!avpriv_flac_is_extradata_valid(avctx, &format, &streaminfo))
309 // first packet: STREAMINFO
310 oggstream->header_len[0] = 51;
311 oggstream->header[0] = av_mallocz(51); // per ogg flac specs
312 p = oggstream->header[0];
314 return AVERROR(ENOMEM);
315 bytestream_put_byte(&p, 0x7F);
316 bytestream_put_buffer(&p, "FLAC", 4);
317 bytestream_put_byte(&p, 1); // major version
318 bytestream_put_byte(&p, 0); // minor version
319 bytestream_put_be16(&p, 1); // headers packets without this one
320 bytestream_put_buffer(&p, "fLaC", 4);
321 bytestream_put_byte(&p, 0x00); // streaminfo
322 bytestream_put_be24(&p, 34);
323 bytestream_put_buffer(&p, streaminfo, FLAC_STREAMINFO_SIZE);
325 // second packet: VorbisComment
326 p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
328 return AVERROR(ENOMEM);
329 oggstream->header[1] = p;
330 bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
331 bytestream_put_be24(&p, oggstream->header_len[1] - 4);
336 #define SPEEX_HEADER_SIZE 80
338 static int ogg_build_speex_headers(AVCodecContext *avctx,
339 OGGStreamContext *oggstream, int bitexact,
344 if (avctx->extradata_size < SPEEX_HEADER_SIZE)
347 // first packet: Speex header
348 p = av_mallocz(SPEEX_HEADER_SIZE);
350 return AVERROR(ENOMEM);
351 oggstream->header[0] = p;
352 oggstream->header_len[0] = SPEEX_HEADER_SIZE;
353 bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
354 AV_WL32(&oggstream->header[0][68], 0); // set extra_headers to 0
356 // second packet: VorbisComment
357 p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
359 return AVERROR(ENOMEM);
360 oggstream->header[1] = p;
365 #define OPUS_HEADER_SIZE 19
367 static int ogg_build_opus_headers(AVCodecContext *avctx,
368 OGGStreamContext *oggstream, int bitexact,
373 if (avctx->extradata_size < OPUS_HEADER_SIZE)
376 /* first packet: Opus header */
377 p = av_mallocz(avctx->extradata_size);
379 return AVERROR(ENOMEM);
380 oggstream->header[0] = p;
381 oggstream->header_len[0] = avctx->extradata_size;
382 bytestream_put_buffer(&p, avctx->extradata, avctx->extradata_size);
384 /* second packet: VorbisComment */
385 p = ogg_write_vorbiscomment(8, bitexact, &oggstream->header_len[1], m, 0);
387 return AVERROR(ENOMEM);
388 oggstream->header[1] = p;
389 bytestream_put_buffer(&p, "OpusTags", 8);
394 static void ogg_write_pages(AVFormatContext *s, int flush)
396 OGGContext *ogg = s->priv_data;
397 OGGPageList *next, *p;
402 for (p = ogg->page_list; p; ) {
403 OGGStreamContext *oggstream =
404 s->streams[p->page.stream_index]->priv_data;
405 if (oggstream->page_count < 2 && !flush)
407 ogg_write_page(s, &p->page,
408 flush == 1 && oggstream->page_count == 1 ? 4 : 0); // eos
416 static int ogg_write_header(AVFormatContext *s)
418 OGGContext *ogg = s->priv_data;
419 OGGStreamContext *oggstream = NULL;
423 av_log(s, AV_LOG_WARNING, "The pagesize option is deprecated\n");
425 for (i = 0; i < s->nb_streams; i++) {
426 AVStream *st = s->streams[i];
427 unsigned serial_num = i;
429 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
430 if (st->codec->codec_id == AV_CODEC_ID_OPUS)
431 /* Opus requires a fixed 48kHz clock */
432 avpriv_set_pts_info(st, 64, 1, 48000);
434 avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
435 } else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
436 avpriv_set_pts_info(st, 64, st->codec->time_base.num, st->codec->time_base.den);
437 if (st->codec->codec_id != AV_CODEC_ID_VORBIS &&
438 st->codec->codec_id != AV_CODEC_ID_THEORA &&
439 st->codec->codec_id != AV_CODEC_ID_SPEEX &&
440 st->codec->codec_id != AV_CODEC_ID_FLAC &&
441 st->codec->codec_id != AV_CODEC_ID_OPUS) {
442 av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
446 if (!st->codec->extradata || !st->codec->extradata_size) {
447 av_log(s, AV_LOG_ERROR, "No extradata present\n");
450 oggstream = av_mallocz(sizeof(*oggstream));
452 return AVERROR(ENOMEM);
454 oggstream->page.stream_index = i;
456 if (!(s->flags & AVFMT_FLAG_BITEXACT))
458 serial_num = av_get_random_seed();
459 for (j = 0; j < i; j++) {
460 OGGStreamContext *sc = s->streams[j]->priv_data;
461 if (serial_num == sc->serial_num)
465 oggstream->serial_num = serial_num;
467 av_dict_copy(&st->metadata, s->metadata, AV_DICT_DONT_OVERWRITE);
469 st->priv_data = oggstream;
470 if (st->codec->codec_id == AV_CODEC_ID_FLAC) {
471 int err = ogg_build_flac_headers(st->codec, oggstream,
472 s->flags & AVFMT_FLAG_BITEXACT,
475 av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
476 av_freep(&st->priv_data);
479 } else if (st->codec->codec_id == AV_CODEC_ID_SPEEX) {
480 int err = ogg_build_speex_headers(st->codec, oggstream,
481 s->flags & AVFMT_FLAG_BITEXACT,
484 av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
485 av_freep(&st->priv_data);
488 } else if (st->codec->codec_id == AV_CODEC_ID_OPUS) {
489 int err = ogg_build_opus_headers(st->codec, oggstream,
490 s->flags & AVFMT_FLAG_BITEXACT,
493 av_log(s, AV_LOG_ERROR, "Error writing Opus headers\n");
494 av_freep(&st->priv_data);
499 const char *cstr = st->codec->codec_id == AV_CODEC_ID_VORBIS ? "vorbis" : "theora";
500 int header_type = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 3 : 0x81;
501 int framing_bit = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 1 : 0;
503 if (avpriv_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
504 st->codec->codec_id == AV_CODEC_ID_VORBIS ? 30 : 42,
505 oggstream->header, oggstream->header_len) < 0) {
506 av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
507 av_freep(&st->priv_data);
511 p = ogg_write_vorbiscomment(7, s->flags & AVFMT_FLAG_BITEXACT,
512 &oggstream->header_len[1], &st->metadata,
514 oggstream->header[1] = p;
516 return AVERROR(ENOMEM);
518 bytestream_put_byte(&p, header_type);
519 bytestream_put_buffer(&p, cstr, 6);
521 if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
522 /** KFGSHIFT is the width of the less significant section of the granule position
523 The less significant section is the frame count since the last keyframe */
524 oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
525 oggstream->vrev = oggstream->header[0][9];
526 av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
527 oggstream->kfgshift, oggstream->vrev);
532 for (j = 0; j < s->nb_streams; j++) {
533 OGGStreamContext *oggstream = s->streams[j]->priv_data;
534 ogg_buffer_data(s, s->streams[j], oggstream->header[0],
535 oggstream->header_len[0], 0, 1);
536 oggstream->page.flags |= 2; // bos
537 ogg_buffer_page(s, oggstream);
539 for (j = 0; j < s->nb_streams; j++) {
540 AVStream *st = s->streams[j];
541 OGGStreamContext *oggstream = st->priv_data;
542 for (i = 1; i < 3; i++) {
543 if (oggstream->header_len[i])
544 ogg_buffer_data(s, st, oggstream->header[i],
545 oggstream->header_len[i], 0, 1);
547 ogg_buffer_page(s, oggstream);
550 oggstream->page.start_granule = AV_NOPTS_VALUE;
552 ogg_write_pages(s, 2);
557 static int ogg_write_packet_internal(AVFormatContext *s, AVPacket *pkt)
559 AVStream *st = s->streams[pkt->stream_index];
560 OGGStreamContext *oggstream = st->priv_data;
564 if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
565 int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
567 if (pkt->flags & AV_PKT_FLAG_KEY)
568 oggstream->last_kf_pts = pts;
569 pframe_count = pts - oggstream->last_kf_pts;
570 // prevent frame count from overflow if key frame flag is not set
571 if (pframe_count >= (1<<oggstream->kfgshift)) {
572 oggstream->last_kf_pts += pframe_count;
575 granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
576 } else if (st->codec->codec_id == AV_CODEC_ID_OPUS)
577 granule = pkt->pts + pkt->duration + av_rescale_q(st->codec->delay, (AVRational){ 1, st->codec->sample_rate }, st->time_base);
579 granule = pkt->pts + pkt->duration;
581 if (oggstream->page.start_granule == AV_NOPTS_VALUE)
582 oggstream->page.start_granule = pkt->pts;
584 ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule, 0);
588 ogg_write_pages(s, 0);
590 oggstream->last_granule = granule;
595 static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
600 return ogg_write_packet_internal(s, pkt);
602 for (i = 0; i < s->nb_streams; i++) {
603 OGGStreamContext *oggstream = s->streams[i]->priv_data;
604 if (oggstream->page.segments_count)
605 ogg_buffer_page(s, oggstream);
608 ogg_write_pages(s, 2);
612 static int ogg_write_trailer(AVFormatContext *s)
616 /* flush current page if needed */
617 for (i = 0; i < s->nb_streams; i++) {
618 OGGStreamContext *oggstream = s->streams[i]->priv_data;
620 if (oggstream->page.size > 0)
621 ogg_buffer_page(s, oggstream);
624 ogg_write_pages(s, 1);
626 for (i = 0; i < s->nb_streams; i++) {
627 AVStream *st = s->streams[i];
628 OGGStreamContext *oggstream = st->priv_data;
629 if (st->codec->codec_id == AV_CODEC_ID_FLAC ||
630 st->codec->codec_id == AV_CODEC_ID_SPEEX ||
631 st->codec->codec_id == AV_CODEC_ID_OPUS) {
632 av_freep(&oggstream->header[0]);
634 av_freep(&oggstream->header[1]);
635 av_freep(&st->priv_data);
641 static const AVClass ogg_muxer_class = {
642 .class_name = "Ogg muxer",
643 .item_name = av_default_item_name,
645 .version = LIBAVUTIL_VERSION_INT,
648 AVOutputFormat ff_ogg_muxer = {
650 .long_name = NULL_IF_CONFIG_SMALL("Ogg"),
651 .mime_type = "application/ogg",
652 .extensions = "ogg,ogv"
653 #if !CONFIG_SPEEX_MUXER
656 #if !CONFIG_OPUS_MUXER
660 .priv_data_size = sizeof(OGGContext),
661 .audio_codec = CONFIG_LIBVORBIS_ENCODER ?
662 AV_CODEC_ID_VORBIS : AV_CODEC_ID_FLAC,
663 .video_codec = AV_CODEC_ID_THEORA,
664 .write_header = ogg_write_header,
665 .write_packet = ogg_write_packet,
666 .write_trailer = ogg_write_trailer,
667 .flags = AVFMT_TS_NEGATIVE | AVFMT_ALLOW_FLUSH,
668 .priv_class = &ogg_muxer_class,
673 static const AVClass oga_muxer_class = {
674 .class_name = "Ogg audio muxer",
675 .item_name = av_default_item_name,
677 .version = LIBAVUTIL_VERSION_INT,
680 AVOutputFormat ff_oga_muxer = {
682 .long_name = NULL_IF_CONFIG_SMALL("Ogg audio"),
683 .mime_type = "audio/ogg",
685 .priv_data_size = sizeof(OGGContext),
686 .audio_codec = AV_CODEC_ID_VORBIS,
687 .video_codec = AV_CODEC_ID_NONE,
688 .write_header = ogg_write_header,
689 .write_packet = ogg_write_packet,
690 .write_trailer = ogg_write_trailer,
691 .flags = AVFMT_TS_NEGATIVE,
692 .priv_class = &oga_muxer_class,
696 #if CONFIG_SPEEX_MUXER
697 static const AVClass speex_muxer_class = {
698 .class_name = "Speex muxer",
699 .item_name = av_default_item_name,
701 .version = LIBAVUTIL_VERSION_INT,
704 AVOutputFormat ff_speex_muxer = {
706 .long_name = NULL_IF_CONFIG_SMALL("Speex"),
707 .mime_type = "audio/ogg",
709 .priv_data_size = sizeof(OGGContext),
710 .audio_codec = AV_CODEC_ID_SPEEX,
711 .video_codec = AV_CODEC_ID_NONE,
712 .write_header = ogg_write_header,
713 .write_packet = ogg_write_packet,
714 .write_trailer = ogg_write_trailer,
715 .flags = AVFMT_TS_NEGATIVE,
716 .priv_class = &speex_muxer_class,
720 #if CONFIG_OPUS_MUXER
721 static const AVClass opus_muxer_class = {
722 .class_name = "Opus muxer",
723 .item_name = av_default_item_name,
725 .version = LIBAVUTIL_VERSION_INT,
728 AVOutputFormat ff_opus_muxer = {
730 .long_name = NULL_IF_CONFIG_SMALL("Opus"),
731 .mime_type = "audio/ogg",
732 .extensions = "opus",
733 .priv_data_size = sizeof(OGGContext),
734 .audio_codec = AV_CODEC_ID_OPUS,
735 .video_codec = AV_CODEC_ID_NONE,
736 .write_header = ogg_write_header,
737 .write_packet = ogg_write_packet,
738 .write_trailer = ogg_write_trailer,
739 .flags = AVFMT_TS_NEGATIVE,
740 .priv_class = &opus_muxer_class,