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 #define OGG_CLASS(flavor, name)\
90 static const AVClass flavor ## _muxer_class = {\
91 .class_name = #name " muxer",\
92 .item_name = av_default_item_name,\
94 .version = LIBAVUTIL_VERSION_INT,\
97 static void ogg_update_checksum(AVFormatContext *s, AVIOContext *pb, int64_t crc_offset)
99 int64_t pos = avio_tell(pb);
100 uint32_t checksum = ffio_get_checksum(pb);
101 avio_seek(pb, crc_offset, SEEK_SET);
102 avio_wb32(pb, checksum);
103 avio_seek(pb, pos, SEEK_SET);
106 static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
108 OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
114 ret = avio_open_dyn_buf(&pb);
117 ffio_init_checksum(pb, ff_crc04C11DB7_update, 0);
118 ffio_wfourcc(pb, "OggS");
120 avio_w8(pb, page->flags | extra_flags);
121 avio_wl64(pb, page->granule);
122 avio_wl32(pb, oggstream->serial_num);
123 avio_wl32(pb, oggstream->page_counter++);
124 crc_offset = avio_tell(pb);
125 avio_wl32(pb, 0); // crc
126 avio_w8(pb, page->segments_count);
127 avio_write(pb, page->segments, page->segments_count);
128 avio_write(pb, page->data, page->size);
130 ogg_update_checksum(s, pb, crc_offset);
133 size = avio_close_dyn_buf(pb, &buf);
137 avio_write(s->pb, buf, size);
140 oggstream->page_count--;
144 static int ogg_key_granule(OGGStreamContext *oggstream, int64_t granule)
146 return oggstream->kfgshift && !(granule & ((1<<oggstream->kfgshift)-1));
149 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, int64_t granule)
151 if (oggstream->kfgshift)
152 return (granule>>oggstream->kfgshift) +
153 (granule & ((1<<oggstream->kfgshift)-1));
158 static int ogg_compare_granule(AVFormatContext *s, OGGPage *next, OGGPage *page)
160 AVStream *st2 = s->streams[next->stream_index];
161 AVStream *st = s->streams[page->stream_index];
162 int64_t next_granule, cur_granule;
164 if (next->granule == -1 || page->granule == -1)
167 next_granule = av_rescale_q(ogg_granule_to_timestamp(st2->priv_data, next->granule),
168 st2->time_base, AV_TIME_BASE_Q);
169 cur_granule = av_rescale_q(ogg_granule_to_timestamp(st->priv_data, page->granule),
170 st ->time_base, AV_TIME_BASE_Q);
171 return next_granule > cur_granule;
174 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
176 oggstream->page.granule = -1;
177 oggstream->page.flags = 0;
178 oggstream->page.segments_count = 0;
179 oggstream->page.size = 0;
183 static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
185 OGGContext *ogg = s->priv_data;
186 OGGPageList **p = &ogg->page_list;
187 OGGPageList *l = av_mallocz(sizeof(*l));
190 return AVERROR(ENOMEM);
191 l->page = oggstream->page;
193 oggstream->page.start_granule = oggstream->page.granule;
194 oggstream->page_count++;
195 ogg_reset_cur_page(oggstream);
198 if (ogg_compare_granule(s, &(*p)->page, &l->page))
208 static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
209 uint8_t *data, unsigned size, int64_t granule,
212 OGGStreamContext *oggstream = st->priv_data;
213 OGGContext *ogg = s->priv_data;
214 int total_segments = size / 255 + 1;
216 int i, segments, len, flush = 0;
218 // Handles VFR by flushing page because this frame needs to have a timestamp
219 // For theora, keyframes also need to have a timestamp to correctly mark
220 // them as such, otherwise seeking will not work correctly at the very
221 // least with old libogg versions.
222 // Do not try to flush header packets though, that will create broken files.
223 if (st->codec->codec_id == AV_CODEC_ID_THEORA && !header &&
224 (ogg_granule_to_timestamp(oggstream, granule) >
225 ogg_granule_to_timestamp(oggstream, oggstream->last_granule) + 1 ||
226 ogg_key_granule(oggstream, granule))) {
227 if (oggstream->page.granule != -1)
228 ogg_buffer_page(s, oggstream);
232 // avoid a continued page
233 if (!header && oggstream->page.size > 0 &&
234 MAX_PAGE_SIZE - oggstream->page.size < size) {
235 ogg_buffer_page(s, oggstream);
238 for (i = 0; i < total_segments; ) {
239 OGGPage *page = &oggstream->page;
241 segments = FFMIN(total_segments - i, 255 - page->segments_count);
243 if (i && !page->segments_count)
244 page->flags |= 1; // continued packet
246 memset(page->segments+page->segments_count, 255, segments - 1);
247 page->segments_count += segments - 1;
249 len = FFMIN(size, segments*255);
250 page->segments[page->segments_count++] = len - (segments-1)*255;
251 memcpy(page->data+page->size, p, len);
257 if (i == total_segments)
258 page->granule = granule;
261 AVStream *st = s->streams[page->stream_index];
263 int64_t start = av_rescale_q(page->start_granule, st->time_base,
265 int64_t next = av_rescale_q(page->granule, st->time_base,
268 if (page->segments_count == 255 ||
269 (ogg->pref_size > 0 && page->size >= ogg->pref_size) ||
270 (ogg->pref_duration > 0 && next - start >= ogg->pref_duration)) {
271 ogg_buffer_page(s, oggstream);
276 if (flush && oggstream->page.granule != -1)
277 ogg_buffer_page(s, oggstream);
282 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
283 int *header_len, AVDictionary **m, int framing_bit)
285 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) + framing_bit;
292 p = av_mallocz(size);
298 ff_vorbiscomment_write(&p, m, vendor);
300 bytestream_put_byte(&p, 1);
306 static int ogg_build_flac_headers(AVCodecContext *avctx,
307 OGGStreamContext *oggstream, int bitexact,
312 if (avctx->extradata_size < FLAC_STREAMINFO_SIZE)
313 return AVERROR(EINVAL);
315 // first packet: STREAMINFO
316 oggstream->header_len[0] = 51;
317 oggstream->header[0] = av_mallocz(51); // per ogg flac specs
318 p = oggstream->header[0];
320 return AVERROR(ENOMEM);
321 bytestream_put_byte(&p, 0x7F);
322 bytestream_put_buffer(&p, "FLAC", 4);
323 bytestream_put_byte(&p, 1); // major version
324 bytestream_put_byte(&p, 0); // minor version
325 bytestream_put_be16(&p, 1); // headers packets without this one
326 bytestream_put_buffer(&p, "fLaC", 4);
327 bytestream_put_byte(&p, 0x00); // streaminfo
328 bytestream_put_be24(&p, 34);
329 bytestream_put_buffer(&p, avctx->extradata, FLAC_STREAMINFO_SIZE);
331 // second packet: VorbisComment
332 p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
334 return AVERROR(ENOMEM);
335 oggstream->header[1] = p;
336 bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
337 bytestream_put_be24(&p, oggstream->header_len[1] - 4);
342 #define SPEEX_HEADER_SIZE 80
344 static int ogg_build_speex_headers(AVCodecContext *avctx,
345 OGGStreamContext *oggstream, int bitexact,
350 if (avctx->extradata_size < SPEEX_HEADER_SIZE)
351 return AVERROR_INVALIDDATA;
353 // first packet: Speex header
354 p = av_mallocz(SPEEX_HEADER_SIZE);
356 return AVERROR(ENOMEM);
357 oggstream->header[0] = p;
358 oggstream->header_len[0] = SPEEX_HEADER_SIZE;
359 bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
360 AV_WL32(&oggstream->header[0][68], 0); // set extra_headers to 0
362 // second packet: VorbisComment
363 p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
365 return AVERROR(ENOMEM);
366 oggstream->header[1] = p;
371 #define OPUS_HEADER_SIZE 19
373 static int ogg_build_opus_headers(AVCodecContext *avctx,
374 OGGStreamContext *oggstream, int bitexact,
379 if (avctx->extradata_size < OPUS_HEADER_SIZE)
380 return AVERROR_INVALIDDATA;
382 /* first packet: Opus header */
383 p = av_mallocz(avctx->extradata_size);
385 return AVERROR(ENOMEM);
386 oggstream->header[0] = p;
387 oggstream->header_len[0] = avctx->extradata_size;
388 bytestream_put_buffer(&p, avctx->extradata, avctx->extradata_size);
390 /* second packet: VorbisComment */
391 p = ogg_write_vorbiscomment(8, bitexact, &oggstream->header_len[1], m, 0);
393 return AVERROR(ENOMEM);
394 oggstream->header[1] = p;
395 bytestream_put_buffer(&p, "OpusTags", 8);
400 static void ogg_write_pages(AVFormatContext *s, int flush)
402 OGGContext *ogg = s->priv_data;
403 OGGPageList *next, *p;
408 for (p = ogg->page_list; p; ) {
409 OGGStreamContext *oggstream =
410 s->streams[p->page.stream_index]->priv_data;
411 if (oggstream->page_count < 2 && !flush)
413 ogg_write_page(s, &p->page,
414 flush == 1 && oggstream->page_count == 1 ? 4 : 0); // eos
422 static int ogg_write_header(AVFormatContext *s)
424 OGGContext *ogg = s->priv_data;
425 OGGStreamContext *oggstream = NULL;
429 av_log(s, AV_LOG_WARNING, "The pagesize option is deprecated\n");
431 for (i = 0; i < s->nb_streams; i++) {
432 AVStream *st = s->streams[i];
433 unsigned serial_num = i;
435 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
436 if (st->codec->codec_id == AV_CODEC_ID_OPUS)
437 /* Opus requires a fixed 48kHz clock */
438 avpriv_set_pts_info(st, 64, 1, 48000);
440 avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
443 if (st->codec->codec_id != AV_CODEC_ID_VORBIS &&
444 st->codec->codec_id != AV_CODEC_ID_THEORA &&
445 st->codec->codec_id != AV_CODEC_ID_SPEEX &&
446 st->codec->codec_id != AV_CODEC_ID_FLAC &&
447 st->codec->codec_id != AV_CODEC_ID_OPUS) {
448 av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
449 return AVERROR(EINVAL);
452 if (!st->codec->extradata || !st->codec->extradata_size) {
453 av_log(s, AV_LOG_ERROR, "No extradata present\n");
454 return AVERROR_INVALIDDATA;
456 oggstream = av_mallocz(sizeof(*oggstream));
458 return AVERROR(ENOMEM);
460 oggstream->page.stream_index = i;
462 if (!(s->flags & AVFMT_FLAG_BITEXACT))
464 serial_num = av_get_random_seed();
465 for (j = 0; j < i; j++) {
466 OGGStreamContext *sc = s->streams[j]->priv_data;
467 if (serial_num == sc->serial_num)
471 oggstream->serial_num = serial_num;
473 av_dict_copy(&st->metadata, s->metadata, AV_DICT_DONT_OVERWRITE);
475 st->priv_data = oggstream;
476 if (st->codec->codec_id == AV_CODEC_ID_FLAC) {
477 int err = ogg_build_flac_headers(st->codec, oggstream,
478 s->flags & AVFMT_FLAG_BITEXACT,
481 av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
482 av_freep(&st->priv_data);
485 } else if (st->codec->codec_id == AV_CODEC_ID_SPEEX) {
486 int err = ogg_build_speex_headers(st->codec, oggstream,
487 s->flags & AVFMT_FLAG_BITEXACT,
490 av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
491 av_freep(&st->priv_data);
494 } else if (st->codec->codec_id == AV_CODEC_ID_OPUS) {
495 int err = ogg_build_opus_headers(st->codec, oggstream,
496 s->flags & AVFMT_FLAG_BITEXACT,
499 av_log(s, AV_LOG_ERROR, "Error writing Opus headers\n");
500 av_freep(&st->priv_data);
505 const char *cstr = st->codec->codec_id == AV_CODEC_ID_VORBIS ? "vorbis" : "theora";
506 int header_type = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 3 : 0x81;
507 int framing_bit = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 1 : 0;
509 if (avpriv_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
510 st->codec->codec_id == AV_CODEC_ID_VORBIS ? 30 : 42,
511 (const uint8_t**)oggstream->header, oggstream->header_len) < 0) {
512 av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
513 av_freep(&st->priv_data);
514 return AVERROR_INVALIDDATA;
517 p = ogg_write_vorbiscomment(7, s->flags & AVFMT_FLAG_BITEXACT,
518 &oggstream->header_len[1], &st->metadata,
520 oggstream->header[1] = p;
522 return AVERROR(ENOMEM);
524 bytestream_put_byte(&p, header_type);
525 bytestream_put_buffer(&p, cstr, 6);
527 if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
528 /** KFGSHIFT is the width of the less significant section of the granule position
529 The less significant section is the frame count since the last keyframe */
530 oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
531 oggstream->vrev = oggstream->header[0][9];
532 av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
533 oggstream->kfgshift, oggstream->vrev);
538 for (j = 0; j < s->nb_streams; j++) {
539 OGGStreamContext *oggstream = s->streams[j]->priv_data;
540 ogg_buffer_data(s, s->streams[j], oggstream->header[0],
541 oggstream->header_len[0], 0, 1);
542 oggstream->page.flags |= 2; // bos
543 ogg_buffer_page(s, oggstream);
545 for (j = 0; j < s->nb_streams; j++) {
546 AVStream *st = s->streams[j];
547 OGGStreamContext *oggstream = st->priv_data;
548 for (i = 1; i < 3; i++) {
549 if (oggstream->header_len[i])
550 ogg_buffer_data(s, st, oggstream->header[i],
551 oggstream->header_len[i], 0, 1);
553 ogg_buffer_page(s, oggstream);
556 oggstream->page.start_granule = AV_NOPTS_VALUE;
558 ogg_write_pages(s, 2);
563 static int ogg_write_packet_internal(AVFormatContext *s, AVPacket *pkt)
565 AVStream *st = s->streams[pkt->stream_index];
566 OGGStreamContext *oggstream = st->priv_data;
570 if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
571 int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
573 if (pkt->flags & AV_PKT_FLAG_KEY)
574 oggstream->last_kf_pts = pts;
575 pframe_count = pts - oggstream->last_kf_pts;
576 // prevent frame count from overflow if key frame flag is not set
577 if (pframe_count >= (1<<oggstream->kfgshift)) {
578 oggstream->last_kf_pts += pframe_count;
581 granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
582 } else if (st->codec->codec_id == AV_CODEC_ID_OPUS)
583 granule = pkt->pts + pkt->duration +
584 av_rescale_q(st->codec->initial_padding,
585 (AVRational){ 1, st->codec->sample_rate },
588 granule = pkt->pts + pkt->duration;
590 if (oggstream->page.start_granule == AV_NOPTS_VALUE)
591 oggstream->page.start_granule = pkt->pts;
593 ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule, 0);
597 ogg_write_pages(s, 0);
599 oggstream->last_granule = granule;
604 static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
609 return ogg_write_packet_internal(s, pkt);
611 for (i = 0; i < s->nb_streams; i++) {
612 OGGStreamContext *oggstream = s->streams[i]->priv_data;
613 if (oggstream->page.segments_count)
614 ogg_buffer_page(s, oggstream);
617 ogg_write_pages(s, 2);
621 static int ogg_write_trailer(AVFormatContext *s)
625 /* flush current page if needed */
626 for (i = 0; i < s->nb_streams; i++) {
627 OGGStreamContext *oggstream = s->streams[i]->priv_data;
629 if (oggstream->page.size > 0)
630 ogg_buffer_page(s, oggstream);
633 ogg_write_pages(s, 1);
635 for (i = 0; i < s->nb_streams; i++) {
636 AVStream *st = s->streams[i];
637 OGGStreamContext *oggstream = st->priv_data;
638 if (st->codec->codec_id == AV_CODEC_ID_FLAC ||
639 st->codec->codec_id == AV_CODEC_ID_SPEEX ||
640 st->codec->codec_id == AV_CODEC_ID_OPUS) {
641 av_freep(&oggstream->header[0]);
643 av_freep(&oggstream->header[1]);
644 av_freep(&st->priv_data);
651 AVOutputFormat ff_ogg_muxer = {
653 .long_name = NULL_IF_CONFIG_SMALL("Ogg"),
654 .mime_type = "application/ogg",
655 .extensions = "ogg,ogv"
656 #if !CONFIG_SPX_MUXER
659 #if !CONFIG_OPUS_MUXER
663 .priv_data_size = sizeof(OGGContext),
664 .audio_codec = CONFIG_LIBVORBIS_ENCODER ?
665 AV_CODEC_ID_VORBIS : AV_CODEC_ID_FLAC,
666 .video_codec = AV_CODEC_ID_THEORA,
667 .write_header = ogg_write_header,
668 .write_packet = ogg_write_packet,
669 .write_trailer = ogg_write_trailer,
670 .flags = AVFMT_TS_NEGATIVE | AVFMT_ALLOW_FLUSH,
671 .priv_class = &ogg_muxer_class,
676 OGG_CLASS(oga, Ogg audio)
677 AVOutputFormat ff_oga_muxer = {
679 .long_name = NULL_IF_CONFIG_SMALL("Ogg Audio"),
680 .mime_type = "audio/ogg",
682 .priv_data_size = sizeof(OGGContext),
683 .audio_codec = CONFIG_LIBVORBIS_ENCODER ?
684 AV_CODEC_ID_VORBIS : AV_CODEC_ID_FLAC,
685 .write_header = ogg_write_header,
686 .write_packet = ogg_write_packet,
687 .write_trailer = ogg_write_trailer,
688 .flags = AVFMT_TS_NEGATIVE | AVFMT_ALLOW_FLUSH,
689 .priv_class = &oga_muxer_class,
694 OGG_CLASS(spx, Ogg Speex)
695 AVOutputFormat ff_spx_muxer = {
697 .long_name = NULL_IF_CONFIG_SMALL("Ogg Speex"),
698 .mime_type = "audio/ogg",
700 .priv_data_size = sizeof(OGGContext),
701 .audio_codec = AV_CODEC_ID_SPEEX,
702 .write_header = ogg_write_header,
703 .write_packet = ogg_write_packet,
704 .write_trailer = ogg_write_trailer,
705 .flags = AVFMT_TS_NEGATIVE | AVFMT_ALLOW_FLUSH,
706 .priv_class = &spx_muxer_class,
710 #if CONFIG_OPUS_MUXER
711 OGG_CLASS(opus, Ogg Opus)
712 AVOutputFormat ff_opus_muxer = {
714 .long_name = NULL_IF_CONFIG_SMALL("Ogg Opus"),
715 .mime_type = "audio/ogg",
716 .extensions = "opus",
717 .priv_data_size = sizeof(OGGContext),
718 .audio_codec = AV_CODEC_ID_OPUS,
719 .write_header = ogg_write_header,
720 .write_packet = ogg_write_packet,
721 .write_trailer = ogg_write_trailer,
722 .flags = AVFMT_TS_NEGATIVE | AVFMT_ALLOW_FLUSH,
723 .priv_class = &opus_muxer_class,