lavc: add a native Opus decoder.
[ffmpeg.git] / libavformat / oggenc.c
1 /*
2  * Ogg muxer
3  * Copyright (c) 2007 Baptiste Coudurier <baptiste dot coudurier at free dot fr>
4  *
5  * This file is part of Libav.
6  *
7  * Libav 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  * Libav 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 Libav; 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;
41     int stream_index;
42     uint8_t flags;
43     uint8_t segments_count;
44     uint8_t segments[255];
45     uint8_t data[MAX_PAGE_SIZE];
46     uint16_t size;
47 } OGGPage;
48
49 typedef struct {
50     unsigned page_counter;
51     uint8_t *header[3];
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
62 } OGGStreamContext;
63
64 typedef struct OGGPageList {
65     OGGPage page;
66     struct OGGPageList *next;
67 } OGGPageList;
68
69 typedef struct {
70     const AVClass *class;
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)
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     { "pagesize", "preferred page size in bytes (deprecated)",
81         OFFSET(pref_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, MAX_PAGE_SIZE, PARAM },
82     { "page_duration", "preferred page duration, in microseconds",
83         OFFSET(pref_duration), AV_OPT_TYPE_INT, { .i64 = 1000000 }, 0, INT64_MAX, PARAM },
84     { NULL },
85 };
86
87 static const AVClass ogg_muxer_class = {
88     .class_name = "Ogg muxer",
89     .item_name  = av_default_item_name,
90     .option     = options,
91     .version    = LIBAVUTIL_VERSION_INT,
92 };
93
94
95 static void ogg_update_checksum(AVFormatContext *s, AVIOContext *pb, int64_t crc_offset)
96 {
97     int64_t pos = avio_tell(pb);
98     uint32_t checksum = ffio_get_checksum(pb);
99     avio_seek(pb, crc_offset, SEEK_SET);
100     avio_wb32(pb, checksum);
101     avio_seek(pb, pos, SEEK_SET);
102 }
103
104 static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
105 {
106     OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
107     AVIOContext *pb;
108     int64_t crc_offset;
109     int ret, size;
110     uint8_t *buf;
111
112     ret = avio_open_dyn_buf(&pb);
113     if (ret < 0)
114         return ret;
115     ffio_init_checksum(pb, ff_crc04C11DB7_update, 0);
116     ffio_wfourcc(pb, "OggS");
117     avio_w8(pb, 0);
118     avio_w8(pb, page->flags | extra_flags);
119     avio_wl64(pb, page->granule);
120     avio_wl32(pb, oggstream->serial_num);
121     avio_wl32(pb, oggstream->page_counter++);
122     crc_offset = avio_tell(pb);
123     avio_wl32(pb, 0); // crc
124     avio_w8(pb, page->segments_count);
125     avio_write(pb, page->segments, page->segments_count);
126     avio_write(pb, page->data, page->size);
127
128     ogg_update_checksum(s, pb, crc_offset);
129     avio_flush(pb);
130
131     size = avio_close_dyn_buf(pb, &buf);
132     if (size < 0)
133         return size;
134
135     avio_write(s->pb, buf, size);
136     avio_flush(s->pb);
137     av_free(buf);
138     oggstream->page_count--;
139     return 0;
140 }
141
142 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, int64_t granule)
143 {
144     if (oggstream->kfgshift)
145         return (granule>>oggstream->kfgshift) +
146             (granule & ((1<<oggstream->kfgshift)-1));
147     else
148         return granule;
149 }
150
151 static int ogg_compare_granule(AVFormatContext *s, OGGPage *next, OGGPage *page)
152 {
153     AVStream *st2 = s->streams[next->stream_index];
154     AVStream *st  = s->streams[page->stream_index];
155     int64_t next_granule, cur_granule;
156
157     if (next->granule == -1 || page->granule == -1)
158         return 0;
159
160     next_granule = av_rescale_q(ogg_granule_to_timestamp(st2->priv_data, next->granule),
161                                 st2->time_base, AV_TIME_BASE_Q);
162     cur_granule  = av_rescale_q(ogg_granule_to_timestamp(st->priv_data, page->granule),
163                                 st ->time_base, AV_TIME_BASE_Q);
164     return next_granule > cur_granule;
165 }
166
167 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
168 {
169     oggstream->page.granule = -1;
170     oggstream->page.flags = 0;
171     oggstream->page.segments_count = 0;
172     oggstream->page.size = 0;
173     return 0;
174 }
175
176 static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
177 {
178     OGGContext *ogg = s->priv_data;
179     OGGPageList **p = &ogg->page_list;
180     OGGPageList *l = av_mallocz(sizeof(*l));
181
182     if (!l)
183         return AVERROR(ENOMEM);
184     l->page = oggstream->page;
185
186     oggstream->page.start_granule = oggstream->page.granule;
187     oggstream->page_count++;
188     ogg_reset_cur_page(oggstream);
189
190     while (*p) {
191         if (ogg_compare_granule(s, &(*p)->page, &l->page))
192             break;
193         p = &(*p)->next;
194     }
195     l->next = *p;
196     *p = l;
197
198     return 0;
199 }
200
201 static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
202                            uint8_t *data, unsigned size, int64_t granule,
203                            int header)
204 {
205     OGGStreamContext *oggstream = st->priv_data;
206     OGGContext *ogg = s->priv_data;
207     int total_segments = size / 255 + 1;
208     uint8_t *p = data;
209     int i, segments, len, flush = 0;
210
211     // Handles VFR by flushing page because this frame needs to have a timestamp
212     if (st->codec->codec_id == AV_CODEC_ID_THEORA && !header &&
213         ogg_granule_to_timestamp(oggstream, granule) >
214         ogg_granule_to_timestamp(oggstream, oggstream->last_granule) + 1) {
215         if (oggstream->page.granule != -1)
216             ogg_buffer_page(s, oggstream);
217         flush = 1;
218     }
219
220     // avoid a continued page
221     if (!header && oggstream->page.size > 0 &&
222         MAX_PAGE_SIZE - oggstream->page.size < size) {
223         ogg_buffer_page(s, oggstream);
224     }
225
226     for (i = 0; i < total_segments; ) {
227         OGGPage *page = &oggstream->page;
228
229         segments = FFMIN(total_segments - i, 255 - page->segments_count);
230
231         if (i && !page->segments_count)
232             page->flags |= 1; // continued packet
233
234         memset(page->segments+page->segments_count, 255, segments - 1);
235         page->segments_count += segments - 1;
236
237         len = FFMIN(size, segments*255);
238         page->segments[page->segments_count++] = len - (segments-1)*255;
239         memcpy(page->data+page->size, p, len);
240         p += len;
241         size -= len;
242         i += segments;
243         page->size += len;
244
245         if (i == total_segments)
246             page->granule = granule;
247
248         if (!header) {
249             AVStream *st = s->streams[page->stream_index];
250
251             int64_t start = av_rescale_q(page->start_granule, st->time_base,
252                                          AV_TIME_BASE_Q);
253             int64_t next  = av_rescale_q(page->granule, st->time_base,
254                                          AV_TIME_BASE_Q);
255
256             if (page->segments_count == 255 ||
257                 (ogg->pref_size     > 0 && page->size   >= ogg->pref_size) ||
258                 (ogg->pref_duration > 0 && next - start >= ogg->pref_duration)) {
259                 ogg_buffer_page(s, oggstream);
260             }
261         }
262     }
263
264     if (flush && oggstream->page.granule != -1)
265         ogg_buffer_page(s, oggstream);
266
267     return 0;
268 }
269
270 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
271                                         int *header_len, AVDictionary **m, int framing_bit)
272 {
273     const char *vendor = bitexact ? "Libav" : LIBAVFORMAT_IDENT;
274     int size;
275     uint8_t *p, *p0;
276     unsigned int count;
277
278     ff_metadata_conv(m, ff_vorbiscomment_metadata_conv, NULL);
279
280     size = offset + ff_vorbiscomment_length(*m, vendor, &count) + framing_bit;
281     p = av_mallocz(size);
282     if (!p)
283         return NULL;
284     p0 = p;
285
286     p += offset;
287     ff_vorbiscomment_write(&p, m, vendor, count);
288     if (framing_bit)
289         bytestream_put_byte(&p, 1);
290
291     *header_len = size;
292     return p0;
293 }
294
295 static int ogg_build_flac_headers(AVCodecContext *avctx,
296                                   OGGStreamContext *oggstream, int bitexact,
297                                   AVDictionary **m)
298 {
299     enum FLACExtradataFormat format;
300     uint8_t *streaminfo;
301     uint8_t *p;
302
303     if (!avpriv_flac_is_extradata_valid(avctx, &format, &streaminfo))
304         return -1;
305
306     // first packet: STREAMINFO
307     oggstream->header_len[0] = 51;
308     oggstream->header[0] = av_mallocz(51); // per ogg flac specs
309     p = oggstream->header[0];
310     if (!p)
311         return AVERROR(ENOMEM);
312     bytestream_put_byte(&p, 0x7F);
313     bytestream_put_buffer(&p, "FLAC", 4);
314     bytestream_put_byte(&p, 1); // major version
315     bytestream_put_byte(&p, 0); // minor version
316     bytestream_put_be16(&p, 1); // headers packets without this one
317     bytestream_put_buffer(&p, "fLaC", 4);
318     bytestream_put_byte(&p, 0x00); // streaminfo
319     bytestream_put_be24(&p, 34);
320     bytestream_put_buffer(&p, streaminfo, FLAC_STREAMINFO_SIZE);
321
322     // second packet: VorbisComment
323     p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
324     if (!p)
325         return AVERROR(ENOMEM);
326     oggstream->header[1] = p;
327     bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
328     bytestream_put_be24(&p, oggstream->header_len[1] - 4);
329
330     return 0;
331 }
332
333 #define SPEEX_HEADER_SIZE 80
334
335 static int ogg_build_speex_headers(AVCodecContext *avctx,
336                                    OGGStreamContext *oggstream, int bitexact,
337                                    AVDictionary **m)
338 {
339     uint8_t *p;
340
341     if (avctx->extradata_size < SPEEX_HEADER_SIZE)
342         return -1;
343
344     // first packet: Speex header
345     p = av_mallocz(SPEEX_HEADER_SIZE);
346     if (!p)
347         return AVERROR(ENOMEM);
348     oggstream->header[0] = p;
349     oggstream->header_len[0] = SPEEX_HEADER_SIZE;
350     bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
351     AV_WL32(&oggstream->header[0][68], 0);  // set extra_headers to 0
352
353     // second packet: VorbisComment
354     p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
355     if (!p)
356         return AVERROR(ENOMEM);
357     oggstream->header[1] = p;
358
359     return 0;
360 }
361
362 #define OPUS_HEADER_SIZE 19
363
364 static int ogg_build_opus_headers(AVCodecContext *avctx,
365                                   OGGStreamContext *oggstream, int bitexact,
366                                   AVDictionary **m)
367 {
368     uint8_t *p;
369
370     if (avctx->extradata_size < OPUS_HEADER_SIZE)
371         return -1;
372
373     /* first packet: Opus header */
374     p = av_mallocz(avctx->extradata_size);
375     if (!p)
376         return AVERROR(ENOMEM);
377     oggstream->header[0] = p;
378     oggstream->header_len[0] = avctx->extradata_size;
379     bytestream_put_buffer(&p, avctx->extradata, avctx->extradata_size);
380
381     /* second packet: VorbisComment */
382     p = ogg_write_vorbiscomment(8, bitexact, &oggstream->header_len[1], m, 0);
383     if (!p)
384         return AVERROR(ENOMEM);
385     oggstream->header[1] = p;
386     bytestream_put_buffer(&p, "OpusTags", 8);
387
388     return 0;
389 }
390
391 static int ogg_write_header(AVFormatContext *s)
392 {
393     OGGContext *ogg = s->priv_data;
394     OGGStreamContext *oggstream;
395     int i, j;
396
397     if (ogg->pref_size)
398         av_log(s, AV_LOG_WARNING, "The pagesize option is deprecated\n");
399
400     for (i = 0; i < s->nb_streams; i++) {
401         AVStream *st = s->streams[i];
402         unsigned serial_num = i;
403
404         if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
405             if (st->codec->codec_id == AV_CODEC_ID_OPUS)
406                 /* Opus requires a fixed 48kHz clock */
407                 avpriv_set_pts_info(st, 64, 1, 48000);
408             else
409                 avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
410         else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
411             avpriv_set_pts_info(st, 64, st->codec->time_base.num, st->codec->time_base.den);
412         if (st->codec->codec_id != AV_CODEC_ID_VORBIS &&
413             st->codec->codec_id != AV_CODEC_ID_THEORA &&
414             st->codec->codec_id != AV_CODEC_ID_SPEEX  &&
415             st->codec->codec_id != AV_CODEC_ID_FLAC   &&
416             st->codec->codec_id != AV_CODEC_ID_OPUS) {
417             av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
418             return -1;
419         }
420
421         if (!st->codec->extradata || !st->codec->extradata_size) {
422             av_log(s, AV_LOG_ERROR, "No extradata present\n");
423             return -1;
424         }
425         oggstream = av_mallocz(sizeof(*oggstream));
426         oggstream->page.stream_index = i;
427
428         if (!(st->codec->flags & CODEC_FLAG_BITEXACT))
429             do {
430                 serial_num = av_get_random_seed();
431                 for (j = 0; j < i; j++) {
432                     OGGStreamContext *sc = s->streams[j]->priv_data;
433                     if (serial_num == sc->serial_num)
434                         break;
435                 }
436             } while (j < i);
437         oggstream->serial_num = serial_num;
438
439         st->priv_data = oggstream;
440         if (st->codec->codec_id == AV_CODEC_ID_FLAC) {
441             int err = ogg_build_flac_headers(st->codec, oggstream,
442                                              st->codec->flags & CODEC_FLAG_BITEXACT,
443                                              &s->metadata);
444             if (err) {
445                 av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
446                 av_freep(&st->priv_data);
447                 return err;
448             }
449         } else if (st->codec->codec_id == AV_CODEC_ID_SPEEX) {
450             int err = ogg_build_speex_headers(st->codec, oggstream,
451                                               st->codec->flags & CODEC_FLAG_BITEXACT,
452                                               &s->metadata);
453             if (err) {
454                 av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
455                 av_freep(&st->priv_data);
456                 return err;
457             }
458         } else if (st->codec->codec_id == AV_CODEC_ID_OPUS) {
459             int err = ogg_build_opus_headers(st->codec, oggstream,
460                                              st->codec->flags & CODEC_FLAG_BITEXACT,
461                                              &s->metadata);
462             if (err) {
463                 av_log(s, AV_LOG_ERROR, "Error writing Opus headers\n");
464                 av_freep(&st->priv_data);
465                 return err;
466             }
467         } else {
468             uint8_t *p;
469             const char *cstr = st->codec->codec_id == AV_CODEC_ID_VORBIS ? "vorbis" : "theora";
470             int header_type = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 3 : 0x81;
471             int framing_bit = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 1 : 0;
472
473             if (avpriv_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
474                                       st->codec->codec_id == AV_CODEC_ID_VORBIS ? 30 : 42,
475                                       oggstream->header, oggstream->header_len) < 0) {
476                 av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
477                 av_freep(&st->priv_data);
478                 return -1;
479             }
480
481             p = ogg_write_vorbiscomment(7, st->codec->flags & CODEC_FLAG_BITEXACT,
482                                         &oggstream->header_len[1], &s->metadata,
483                                         framing_bit);
484             oggstream->header[1] = p;
485             if (!p)
486                 return AVERROR(ENOMEM);
487
488             bytestream_put_byte(&p, header_type);
489             bytestream_put_buffer(&p, cstr, 6);
490
491             if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
492                 /** KFGSHIFT is the width of the less significant section of the granule position
493                     The less significant section is the frame count since the last keyframe */
494                 oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
495                 oggstream->vrev = oggstream->header[0][9];
496                 av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
497                        oggstream->kfgshift, oggstream->vrev);
498             }
499         }
500     }
501
502     for (j = 0; j < s->nb_streams; j++) {
503         OGGStreamContext *oggstream = s->streams[j]->priv_data;
504         ogg_buffer_data(s, s->streams[j], oggstream->header[0],
505                         oggstream->header_len[0], 0, 1);
506         oggstream->page.flags |= 2; // bos
507         ogg_buffer_page(s, oggstream);
508     }
509     for (j = 0; j < s->nb_streams; j++) {
510         AVStream *st = s->streams[j];
511         OGGStreamContext *oggstream = st->priv_data;
512         for (i = 1; i < 3; i++) {
513             if (oggstream && oggstream->header_len[i])
514                 ogg_buffer_data(s, st, oggstream->header[i],
515                                 oggstream->header_len[i], 0, 1);
516         }
517         ogg_buffer_page(s, oggstream);
518     }
519
520     oggstream->page.start_granule = AV_NOPTS_VALUE;
521
522     return 0;
523 }
524
525 static void ogg_write_pages(AVFormatContext *s, int flush)
526 {
527     OGGContext *ogg = s->priv_data;
528     OGGPageList *next, *p;
529
530     if (!ogg->page_list)
531         return;
532
533     for (p = ogg->page_list; p; ) {
534         OGGStreamContext *oggstream =
535             s->streams[p->page.stream_index]->priv_data;
536         if (oggstream->page_count < 2 && !flush)
537             break;
538         ogg_write_page(s, &p->page,
539                        flush && oggstream->page_count == 1 ? 4 : 0); // eos
540         next = p->next;
541         av_freep(&p);
542         p = next;
543     }
544     ogg->page_list = p;
545 }
546
547 static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
548 {
549     AVStream *st = s->streams[pkt->stream_index];
550     OGGStreamContext *oggstream = st->priv_data;
551     int ret;
552     int64_t granule;
553
554     if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
555         int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
556         int pframe_count;
557         if (pkt->flags & AV_PKT_FLAG_KEY)
558             oggstream->last_kf_pts = pts;
559         pframe_count = pts - oggstream->last_kf_pts;
560         // prevent frame count from overflow if key frame flag is not set
561         if (pframe_count >= (1<<oggstream->kfgshift)) {
562             oggstream->last_kf_pts += pframe_count;
563             pframe_count = 0;
564         }
565         granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
566     } else if (st->codec->codec_id == AV_CODEC_ID_OPUS)
567         granule = pkt->pts + pkt->duration + av_rescale_q(st->codec->delay, (AVRational){ 1, st->codec->sample_rate }, st->time_base);
568     else
569         granule = pkt->pts + pkt->duration;
570
571     if (oggstream->page.start_granule == AV_NOPTS_VALUE)
572         oggstream->page.start_granule = pkt->pts;
573
574     ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule, 0);
575     if (ret < 0)
576         return ret;
577
578     ogg_write_pages(s, 0);
579
580     oggstream->last_granule = granule;
581
582     return 0;
583 }
584
585 static int ogg_write_trailer(AVFormatContext *s)
586 {
587     int i;
588
589     /* flush current page if needed */
590     for (i = 0; i < s->nb_streams; i++) {
591         OGGStreamContext *oggstream = s->streams[i]->priv_data;
592
593         if (oggstream->page.size > 0)
594             ogg_buffer_page(s, oggstream);
595     }
596
597     ogg_write_pages(s, 1);
598
599     for (i = 0; i < s->nb_streams; i++) {
600         AVStream *st = s->streams[i];
601         OGGStreamContext *oggstream = st->priv_data;
602         if (st->codec->codec_id == AV_CODEC_ID_FLAC ||
603             st->codec->codec_id == AV_CODEC_ID_SPEEX ||
604             st->codec->codec_id == AV_CODEC_ID_OPUS) {
605             av_free(oggstream->header[0]);
606         }
607         av_freep(&oggstream->header[1]);
608         av_freep(&st->priv_data);
609     }
610     return 0;
611 }
612
613 AVOutputFormat ff_ogg_muxer = {
614     .name              = "ogg",
615     .long_name         = NULL_IF_CONFIG_SMALL("Ogg"),
616     .mime_type         = "application/ogg",
617     .extensions        = "ogg,ogv,spx,opus",
618     .priv_data_size    = sizeof(OGGContext),
619     .audio_codec       = CONFIG_LIBVORBIS_ENCODER ?
620                          AV_CODEC_ID_VORBIS : AV_CODEC_ID_FLAC,
621     .video_codec       = AV_CODEC_ID_THEORA,
622     .write_header      = ogg_write_header,
623     .write_packet      = ogg_write_packet,
624     .write_trailer     = ogg_write_trailer,
625     .flags             = AVFMT_TS_NEGATIVE,
626     .priv_class        = &ogg_muxer_class,
627 };