Export metadata in the generic format. Deprecate old conversion API.
[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 FFmpeg.
6  *
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.
11  *
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.
16  *
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
20  */
21
22 #include "libavutil/crc.h"
23 #include "libavutil/random_seed.h"
24 #include "libavcodec/xiph.h"
25 #include "libavcodec/bytestream.h"
26 #include "libavcodec/flac.h"
27 #include "avformat.h"
28 #include "internal.h"
29 #include "vorbiscomment.h"
30
31 #define MAX_PAGE_SIZE 65025
32
33 typedef struct {
34     int64_t granule;
35     int stream_index;
36     uint8_t flags;
37     uint8_t segments_count;
38     uint8_t segments[255];
39     uint8_t data[MAX_PAGE_SIZE];
40     uint16_t size;
41 } OGGPage;
42
43 typedef struct {
44     unsigned page_counter;
45     uint8_t *header[3];
46     int header_len[3];
47     /** for theora granule */
48     int kfgshift;
49     int64_t last_kf_pts;
50     int vrev;
51     int eos;
52     unsigned page_count; ///< number of page buffered
53     OGGPage page; ///< current page
54     unsigned serial_num; ///< serial number
55 } OGGStreamContext;
56
57 typedef struct OGGPageList {
58     OGGPage page;
59     struct OGGPageList *next;
60 } OGGPageList;
61
62 typedef struct {
63     OGGPageList *page_list;
64 } OGGContext;
65
66 static void ogg_update_checksum(AVFormatContext *s, ByteIOContext *pb, int64_t crc_offset)
67 {
68     int64_t pos = url_ftell(pb);
69     uint32_t checksum = get_checksum(pb);
70     url_fseek(pb, crc_offset, SEEK_SET);
71     put_be32(pb, checksum);
72     url_fseek(pb, pos, SEEK_SET);
73 }
74
75 static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
76 {
77     OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
78     ByteIOContext *pb;
79     int64_t crc_offset;
80     int ret, size;
81     uint8_t *buf;
82
83     ret = url_open_dyn_buf(&pb);
84     if (ret < 0)
85         return ret;
86     init_checksum(pb, ff_crc04C11DB7_update, 0);
87     put_tag(pb, "OggS");
88     put_byte(pb, 0);
89     put_byte(pb, page->flags | extra_flags);
90     put_le64(pb, page->granule);
91     put_le32(pb, oggstream->serial_num);
92     put_le32(pb, oggstream->page_counter++);
93     crc_offset = url_ftell(pb);
94     put_le32(pb, 0); // crc
95     put_byte(pb, page->segments_count);
96     put_buffer(pb, page->segments, page->segments_count);
97     put_buffer(pb, page->data, page->size);
98
99     ogg_update_checksum(s, pb, crc_offset);
100     put_flush_packet(pb);
101
102     size = url_close_dyn_buf(pb, &buf);
103     if (size < 0)
104         return size;
105
106     put_buffer(s->pb, buf, size);
107     put_flush_packet(s->pb);
108     av_free(buf);
109     oggstream->page_count--;
110     return 0;
111 }
112
113 static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, OGGPage *page)
114 {
115     if (oggstream->kfgshift)
116         return (page->granule>>oggstream->kfgshift) +
117             (page->granule & ((1<<oggstream->kfgshift)-1));
118     else
119         return page->granule;
120 }
121
122 static int ogg_compare_granule(AVFormatContext *s, OGGPage *next, OGGPage *page)
123 {
124     AVStream *st2 = s->streams[next->stream_index];
125     AVStream *st  = s->streams[page->stream_index];
126     int64_t next_granule, cur_granule;
127
128     if (next->granule == -1 || page->granule == -1)
129         return 0;
130
131     next_granule = av_rescale_q(ogg_granule_to_timestamp(st2->priv_data, next),
132                                 st2->time_base, AV_TIME_BASE_Q);
133     cur_granule  = av_rescale_q(ogg_granule_to_timestamp(st->priv_data, page),
134                                 st ->time_base, AV_TIME_BASE_Q);
135     return next_granule > cur_granule;
136 }
137
138 static int ogg_reset_cur_page(OGGStreamContext *oggstream)
139 {
140     oggstream->page.granule = -1;
141     oggstream->page.flags = 0;
142     oggstream->page.segments_count = 0;
143     oggstream->page.size = 0;
144     return 0;
145 }
146
147 static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
148 {
149     OGGContext *ogg = s->priv_data;
150     OGGPageList **p = &ogg->page_list;
151     OGGPageList *l = av_mallocz(sizeof(*l));
152
153     if (!l)
154         return AVERROR(ENOMEM);
155     l->page = oggstream->page;
156
157     oggstream->page_count++;
158     ogg_reset_cur_page(oggstream);
159
160     while (*p) {
161         if (ogg_compare_granule(s, &(*p)->page, &l->page))
162             break;
163         p = &(*p)->next;
164     }
165     l->next = *p;
166     *p = l;
167
168     return 0;
169 }
170
171 static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
172                            uint8_t *data, unsigned size, int64_t granule)
173 {
174     OGGStreamContext *oggstream = st->priv_data;
175     int total_segments = size / 255 + 1;
176     uint8_t *p = data;
177     int i, segments, len;
178
179     for (i = 0; i < total_segments; ) {
180         OGGPage *page = &oggstream->page;
181
182         segments = FFMIN(total_segments - i, 255 - page->segments_count);
183
184         if (i && !page->segments_count)
185             page->flags |= 1; // continued packet
186
187         memset(page->segments+page->segments_count, 255, segments - 1);
188         page->segments_count += segments - 1;
189
190         len = FFMIN(size, segments*255);
191         page->segments[page->segments_count++] = len - (segments-1)*255;
192         memcpy(page->data+page->size, p, len);
193         p += len;
194         size -= len;
195         i += segments;
196         page->size += len;
197
198         if (i == total_segments)
199             page->granule = granule;
200
201         if (page->segments_count == 255) {
202             ogg_buffer_page(s, oggstream);
203         }
204     }
205     return 0;
206 }
207
208 static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
209                                         int *header_len, AVMetadata **m, int framing_bit)
210 {
211     const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
212     int size;
213     uint8_t *p, *p0;
214     unsigned int count;
215
216     size = offset + ff_vorbiscomment_length(*m, vendor, &count) + framing_bit;
217     p = av_mallocz(size);
218     if (!p)
219         return NULL;
220     p0 = p;
221
222     p += offset;
223     ff_vorbiscomment_write(&p, m, vendor, count);
224     if (framing_bit)
225         bytestream_put_byte(&p, 1);
226
227     *header_len = size;
228     return p0;
229 }
230
231 static int ogg_build_flac_headers(AVCodecContext *avctx,
232                                   OGGStreamContext *oggstream, int bitexact,
233                                   AVMetadata **m)
234 {
235     enum FLACExtradataFormat format;
236     uint8_t *streaminfo;
237     uint8_t *p;
238
239     if (!ff_flac_is_extradata_valid(avctx, &format, &streaminfo))
240         return -1;
241
242     // first packet: STREAMINFO
243     oggstream->header_len[0] = 51;
244     oggstream->header[0] = av_mallocz(51); // per ogg flac specs
245     p = oggstream->header[0];
246     if (!p)
247         return AVERROR(ENOMEM);
248     bytestream_put_byte(&p, 0x7F);
249     bytestream_put_buffer(&p, "FLAC", 4);
250     bytestream_put_byte(&p, 1); // major version
251     bytestream_put_byte(&p, 0); // minor version
252     bytestream_put_be16(&p, 1); // headers packets without this one
253     bytestream_put_buffer(&p, "fLaC", 4);
254     bytestream_put_byte(&p, 0x00); // streaminfo
255     bytestream_put_be24(&p, 34);
256     bytestream_put_buffer(&p, streaminfo, FLAC_STREAMINFO_SIZE);
257
258     // second packet: VorbisComment
259     p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
260     if (!p)
261         return AVERROR(ENOMEM);
262     oggstream->header[1] = p;
263     bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
264     bytestream_put_be24(&p, oggstream->header_len[1] - 4);
265
266     return 0;
267 }
268
269 #define SPEEX_HEADER_SIZE 80
270
271 static int ogg_build_speex_headers(AVCodecContext *avctx,
272                                    OGGStreamContext *oggstream, int bitexact,
273                                    AVMetadata **m)
274 {
275     uint8_t *p;
276
277     if (avctx->extradata_size < SPEEX_HEADER_SIZE)
278         return -1;
279
280     // first packet: Speex header
281     p = av_mallocz(SPEEX_HEADER_SIZE);
282     if (!p)
283         return AVERROR(ENOMEM);
284     oggstream->header[0] = p;
285     oggstream->header_len[0] = SPEEX_HEADER_SIZE;
286     bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
287     AV_WL32(&oggstream->header[0][68], 0);  // set extra_headers to 0
288
289     // second packet: VorbisComment
290     p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
291     if (!p)
292         return AVERROR(ENOMEM);
293     oggstream->header[1] = p;
294
295     return 0;
296 }
297
298 static int ogg_write_header(AVFormatContext *s)
299 {
300     OGGStreamContext *oggstream;
301     int i, j;
302
303     for (i = 0; i < s->nb_streams; i++) {
304         AVStream *st = s->streams[i];
305         unsigned serial_num = i;
306
307         if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
308             av_set_pts_info(st, 64, 1, st->codec->sample_rate);
309         else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
310             av_set_pts_info(st, 64, st->codec->time_base.num, st->codec->time_base.den);
311         if (st->codec->codec_id != CODEC_ID_VORBIS &&
312             st->codec->codec_id != CODEC_ID_THEORA &&
313             st->codec->codec_id != CODEC_ID_SPEEX  &&
314             st->codec->codec_id != CODEC_ID_FLAC) {
315             av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
316             return -1;
317         }
318
319         if (!st->codec->extradata || !st->codec->extradata_size) {
320             av_log(s, AV_LOG_ERROR, "No extradata present\n");
321             return -1;
322         }
323         oggstream = av_mallocz(sizeof(*oggstream));
324         oggstream->page.stream_index = i;
325
326         if (!(st->codec->flags & CODEC_FLAG_BITEXACT))
327             do {
328                 serial_num = av_get_random_seed();
329                 for (j = 0; j < i; j++) {
330                     OGGStreamContext *sc = s->streams[j]->priv_data;
331                     if (serial_num == sc->serial_num)
332                         break;
333                 }
334             } while (j < i);
335         oggstream->serial_num = serial_num;
336
337         st->priv_data = oggstream;
338         if (st->codec->codec_id == CODEC_ID_FLAC) {
339             int err = ogg_build_flac_headers(st->codec, oggstream,
340                                              st->codec->flags & CODEC_FLAG_BITEXACT,
341                                              &s->metadata);
342             if (err) {
343                 av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
344                 av_freep(&st->priv_data);
345                 return err;
346             }
347         } else if (st->codec->codec_id == CODEC_ID_SPEEX) {
348             int err = ogg_build_speex_headers(st->codec, oggstream,
349                                               st->codec->flags & CODEC_FLAG_BITEXACT,
350                                               &s->metadata);
351             if (err) {
352                 av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
353                 av_freep(&st->priv_data);
354                 return err;
355             }
356         } else {
357             uint8_t *p;
358             const char *cstr = st->codec->codec_id == CODEC_ID_VORBIS ? "vorbis" : "theora";
359             int header_type = st->codec->codec_id == CODEC_ID_VORBIS ? 3 : 0x81;
360             int framing_bit = st->codec->codec_id == CODEC_ID_VORBIS ? 1 : 0;
361
362             if (ff_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
363                                       st->codec->codec_id == CODEC_ID_VORBIS ? 30 : 42,
364                                       oggstream->header, oggstream->header_len) < 0) {
365                 av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
366                 av_freep(&st->priv_data);
367                 return -1;
368             }
369
370             p = ogg_write_vorbiscomment(7, st->codec->flags & CODEC_FLAG_BITEXACT,
371                                         &oggstream->header_len[1], &s->metadata,
372                                         framing_bit);
373             if (!p)
374                 return AVERROR(ENOMEM);
375
376             oggstream->header[1] = p;
377             bytestream_put_byte(&p, header_type);
378             bytestream_put_buffer(&p, cstr, 6);
379
380             if (st->codec->codec_id == CODEC_ID_THEORA) {
381                 /** KFGSHIFT is the width of the less significant section of the granule position
382                     The less significant section is the frame count since the last keyframe */
383                 oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
384                 oggstream->vrev = oggstream->header[0][9];
385                 av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
386                        oggstream->kfgshift, oggstream->vrev);
387             }
388         }
389     }
390
391     for (j = 0; j < s->nb_streams; j++) {
392         OGGStreamContext *oggstream = s->streams[j]->priv_data;
393         ogg_buffer_data(s, s->streams[j], oggstream->header[0],
394                         oggstream->header_len[0], 0);
395         oggstream->page.flags |= 2; // bos
396         ogg_buffer_page(s, oggstream);
397     }
398     for (j = 0; j < s->nb_streams; j++) {
399         AVStream *st = s->streams[j];
400         OGGStreamContext *oggstream = st->priv_data;
401         for (i = 1; i < 3; i++) {
402             if (oggstream && oggstream->header_len[i])
403                 ogg_buffer_data(s, st, oggstream->header[i],
404                                 oggstream->header_len[i], 0);
405         }
406         ogg_buffer_page(s, oggstream);
407     }
408     return 0;
409 }
410
411 static void ogg_write_pages(AVFormatContext *s, int flush)
412 {
413     OGGContext *ogg = s->priv_data;
414     OGGPageList *next, *p;
415
416     if (!ogg->page_list)
417         return;
418
419     for (p = ogg->page_list; p; ) {
420         OGGStreamContext *oggstream =
421             s->streams[p->page.stream_index]->priv_data;
422         if (oggstream->page_count < 2 && !flush)
423             break;
424         ogg_write_page(s, &p->page,
425                        flush && oggstream->page_count == 1 ? 4 : 0); // eos
426         next = p->next;
427         av_freep(&p);
428         p = next;
429     }
430     ogg->page_list = p;
431 }
432
433 static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
434 {
435     AVStream *st = s->streams[pkt->stream_index];
436     OGGStreamContext *oggstream = st->priv_data;
437     int ret;
438     int64_t granule;
439
440     if (st->codec->codec_id == CODEC_ID_THEORA) {
441         int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
442         int pframe_count;
443         if (pkt->flags & AV_PKT_FLAG_KEY)
444             oggstream->last_kf_pts = pts;
445         pframe_count = pts - oggstream->last_kf_pts;
446         // prevent frame count from overflow if key frame flag is not set
447         if (pframe_count >= (1<<oggstream->kfgshift)) {
448             oggstream->last_kf_pts += pframe_count;
449             pframe_count = 0;
450         }
451         granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
452     } else
453         granule = pkt->pts + pkt->duration;
454
455     ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule);
456     if (ret < 0)
457         return ret;
458
459     ogg_write_pages(s, 0);
460
461     return 0;
462 }
463
464 static int ogg_write_trailer(AVFormatContext *s)
465 {
466     int i;
467
468     /* flush current page */
469     for (i = 0; i < s->nb_streams; i++)
470         ogg_buffer_page(s, s->streams[i]->priv_data);
471
472     ogg_write_pages(s, 1);
473
474     for (i = 0; i < s->nb_streams; i++) {
475         AVStream *st = s->streams[i];
476         OGGStreamContext *oggstream = st->priv_data;
477         if (st->codec->codec_id == CODEC_ID_FLAC ||
478             st->codec->codec_id == CODEC_ID_SPEEX) {
479             av_free(oggstream->header[0]);
480             av_free(oggstream->header[1]);
481         }
482         av_freep(&st->priv_data);
483     }
484     return 0;
485 }
486
487 AVOutputFormat ogg_muxer = {
488     "ogg",
489     NULL_IF_CONFIG_SMALL("Ogg"),
490     "application/ogg",
491     "ogg,ogv,spx",
492     sizeof(OGGContext),
493     CODEC_ID_FLAC,
494     CODEC_ID_THEORA,
495     ogg_write_header,
496     ogg_write_packet,
497     ogg_write_trailer,
498 };