2b17154539102e44c42b190b15c9b7aed958d254
[libav.git] / libavformat / matroskaenc.c
1 /*
2 * Matroska muxer
3 * Copyright (c) 2007 David Conrad
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 "avformat.h"
23 #include "riff.h"
24 #include "isom.h"
25 #include "matroska.h"
26 #include "avc.h"
27 #include "flacenc.h"
28 #include "libavutil/intreadwrite.h"
29 #include "libavutil/md5.h"
30 #include "libavcodec/xiph.h"
31 #include "libavcodec/mpeg4audio.h"
32
33 typedef struct ebml_master {
34 int64_t pos; ///< absolute offset in the file where the master's elements start
35 int sizebytes; ///< how many bytes were reserved for the size
36 } ebml_master;
37
38 typedef struct mkv_seekhead_entry {
39 unsigned int elementid;
40 uint64_t segmentpos;
41 } mkv_seekhead_entry;
42
43 typedef struct mkv_seekhead {
44 int64_t filepos;
45 int64_t segment_offset; ///< the file offset to the beginning of the segment
46 int reserved_size; ///< -1 if appending to file
47 int max_entries;
48 mkv_seekhead_entry *entries;
49 int num_entries;
50 } mkv_seekhead;
51
52 typedef struct {
53 uint64_t pts;
54 int tracknum;
55 int64_t cluster_pos; ///< file offset of the cluster containing the block
56 } mkv_cuepoint;
57
58 typedef struct {
59 int64_t segment_offset;
60 mkv_cuepoint *entries;
61 int num_entries;
62 } mkv_cues;
63
64 typedef struct {
65 int write_dts;
66 } mkv_track;
67
68 typedef struct MatroskaMuxContext {
69 ByteIOContext *dyn_bc;
70 ebml_master segment;
71 int64_t segment_offset;
72 int64_t segment_uid;
73 ebml_master cluster;
74 int64_t cluster_pos; ///< file offset of the current cluster
75 int64_t cluster_pts;
76 int64_t duration_offset;
77 int64_t duration;
78 mkv_seekhead *main_seekhead;
79 mkv_seekhead *cluster_seekhead;
80 mkv_cues *cues;
81 mkv_track *tracks;
82
83 struct AVMD5 *md5_ctx;
84 } MatroskaMuxContext;
85
86
87 /** 2 bytes * 3 for EBML IDs, 3 1-byte EBML lengths, 8 bytes for 64 bit
88 * offset, 4 bytes for target EBML ID */
89 #define MAX_SEEKENTRY_SIZE 21
90
91 /** per-cuepoint-track - 3 1-byte EBML IDs, 3 1-byte EBML sizes, 2
92 * 8-byte uint max */
93 #define MAX_CUETRACKPOS_SIZE 22
94
95 /** per-cuepoint - 2 1-byte EBML IDs, 2 1-byte EBML sizes, 8-byte uint max */
96 #define MAX_CUEPOINT_SIZE(num_tracks) 12 + MAX_CUETRACKPOS_SIZE*num_tracks
97
98
99 static int ebml_id_size(unsigned int id)
100 {
101 return (av_log2(id+1)-1)/7+1;
102 }
103
104 static void put_ebml_id(ByteIOContext *pb, unsigned int id)
105 {
106 int i = ebml_id_size(id);
107 while (i--)
108 put_byte(pb, id >> (i*8));
109 }
110
111 /**
112 * Write an EBML size meaning "unknown size".
113 *
114 * @param bytes The number of bytes the size should occupy (maximum: 8).
115 */
116 static void put_ebml_size_unknown(ByteIOContext *pb, int bytes)
117 {
118 assert(bytes <= 8);
119 put_byte(pb, 0x1ff >> bytes);
120 while (--bytes)
121 put_byte(pb, 0xff);
122 }
123
124 /**
125 * Calculate how many bytes are needed to represent a given number in EBML.
126 */
127 static int ebml_num_size(uint64_t num)
128 {
129 int bytes = 1;
130 while ((num+1) >> bytes*7) bytes++;
131 return bytes;
132 }
133
134 /**
135 * Write a number in EBML variable length format.
136 *
137 * @param bytes The number of bytes that need to be used to write the number.
138 * If zero, any number of bytes can be used.
139 */
140 static void put_ebml_num(ByteIOContext *pb, uint64_t num, int bytes)
141 {
142 int i, needed_bytes = ebml_num_size(num);
143
144 // sizes larger than this are currently undefined in EBML
145 assert(num < (1ULL<<56)-1);
146
147 if (bytes == 0)
148 // don't care how many bytes are used, so use the min
149 bytes = needed_bytes;
150 // the bytes needed to write the given size would exceed the bytes
151 // that we need to use, so write unknown size. This shouldn't happen.
152 assert(bytes >= needed_bytes);
153
154 num |= 1ULL << bytes*7;
155 for (i = bytes - 1; i >= 0; i--)
156 put_byte(pb, num >> i*8);
157 }
158
159 static void put_ebml_uint(ByteIOContext *pb, unsigned int elementid, uint64_t val)
160 {
161 int i, bytes = 1;
162 uint64_t tmp = val;
163 while (tmp>>=8) bytes++;
164
165 put_ebml_id(pb, elementid);
166 put_ebml_num(pb, bytes, 0);
167 for (i = bytes - 1; i >= 0; i--)
168 put_byte(pb, val >> i*8);
169 }
170
171 static void put_ebml_float(ByteIOContext *pb, unsigned int elementid, double val)
172 {
173 put_ebml_id(pb, elementid);
174 put_ebml_num(pb, 8, 0);
175 put_be64(pb, av_dbl2int(val));
176 }
177
178 static void put_ebml_binary(ByteIOContext *pb, unsigned int elementid,
179 const uint8_t *buf, int size)
180 {
181 put_ebml_id(pb, elementid);
182 put_ebml_num(pb, size, 0);
183 put_buffer(pb, buf, size);
184 }
185
186 static void put_ebml_string(ByteIOContext *pb, unsigned int elementid, const char *str)
187 {
188 put_ebml_binary(pb, elementid, str, strlen(str));
189 }
190
191 /**
192 * Writes a void element of a given size. Useful for reserving space in
193 * the file to be written to later.
194 *
195 * @param size The number of bytes to reserve, which must be at least 2.
196 */
197 static void put_ebml_void(ByteIOContext *pb, uint64_t size)
198 {
199 int64_t currentpos = url_ftell(pb);
200
201 assert(size >= 2);
202
203 put_ebml_id(pb, EBML_ID_VOID);
204 // we need to subtract the length needed to store the size from the
205 // size we need to reserve so 2 cases, we use 8 bytes to store the
206 // size if possible, 1 byte otherwise
207 if (size < 10)
208 put_ebml_num(pb, size-1, 0);
209 else
210 put_ebml_num(pb, size-9, 8);
211 while(url_ftell(pb) < currentpos + size)
212 put_byte(pb, 0);
213 }
214
215 static ebml_master start_ebml_master(ByteIOContext *pb, unsigned int elementid, uint64_t expectedsize)
216 {
217 int bytes = expectedsize ? ebml_num_size(expectedsize) : 8;
218 put_ebml_id(pb, elementid);
219 put_ebml_size_unknown(pb, bytes);
220 return (ebml_master){ url_ftell(pb), bytes };
221 }
222
223 static void end_ebml_master(ByteIOContext *pb, ebml_master master)
224 {
225 int64_t pos = url_ftell(pb);
226
227 if (url_fseek(pb, master.pos - master.sizebytes, SEEK_SET) < 0)
228 return;
229 put_ebml_num(pb, pos - master.pos, master.sizebytes);
230 url_fseek(pb, pos, SEEK_SET);
231 }
232
233 static void put_xiph_size(ByteIOContext *pb, int size)
234 {
235 int i;
236 for (i = 0; i < size / 255; i++)
237 put_byte(pb, 255);
238 put_byte(pb, size % 255);
239 }
240
241 /**
242 * Initialize a mkv_seekhead element to be ready to index level 1 Matroska
243 * elements. If a maximum number of elements is specified, enough space
244 * will be reserved at the current file location to write a seek head of
245 * that size.
246 *
247 * @param segment_offset The absolute offset to the position in the file
248 * where the segment begins.
249 * @param numelements The maximum number of elements that will be indexed
250 * by this seek head, 0 if unlimited.
251 */
252 static mkv_seekhead * mkv_start_seekhead(ByteIOContext *pb, int64_t segment_offset, int numelements)
253 {
254 mkv_seekhead *new_seekhead = av_mallocz(sizeof(mkv_seekhead));
255 if (new_seekhead == NULL)
256 return NULL;
257
258 new_seekhead->segment_offset = segment_offset;
259
260 if (numelements > 0) {
261 new_seekhead->filepos = url_ftell(pb);
262 // 21 bytes max for a seek entry, 10 bytes max for the SeekHead ID
263 // and size, and 3 bytes to guarantee that an EBML void element
264 // will fit afterwards
265 new_seekhead->reserved_size = numelements * MAX_SEEKENTRY_SIZE + 13;
266 new_seekhead->max_entries = numelements;
267 put_ebml_void(pb, new_seekhead->reserved_size);
268 }
269 return new_seekhead;
270 }
271
272 static int mkv_add_seekhead_entry(mkv_seekhead *seekhead, unsigned int elementid, uint64_t filepos)
273 {
274 mkv_seekhead_entry *entries = seekhead->entries;
275
276 // don't store more elements than we reserved space for
277 if (seekhead->max_entries > 0 && seekhead->max_entries <= seekhead->num_entries)
278 return -1;
279
280 entries = av_realloc(entries, (seekhead->num_entries + 1) * sizeof(mkv_seekhead_entry));
281 if (entries == NULL)
282 return AVERROR(ENOMEM);
283
284 entries[seekhead->num_entries ].elementid = elementid;
285 entries[seekhead->num_entries++].segmentpos = filepos - seekhead->segment_offset;
286
287 seekhead->entries = entries;
288 return 0;
289 }
290
291 /**
292 * Write the seek head to the file and free it. If a maximum number of
293 * elements was specified to mkv_start_seekhead(), the seek head will
294 * be written at the location reserved for it. Otherwise, it is written
295 * at the current location in the file.
296 *
297 * @return The file offset where the seekhead was written,
298 * -1 if an error occurred.
299 */
300 static int64_t mkv_write_seekhead(ByteIOContext *pb, mkv_seekhead *seekhead)
301 {
302 ebml_master metaseek, seekentry;
303 int64_t currentpos;
304 int i;
305
306 currentpos = url_ftell(pb);
307
308 if (seekhead->reserved_size > 0)
309 if (url_fseek(pb, seekhead->filepos, SEEK_SET) < 0)
310 return -1;
311
312 metaseek = start_ebml_master(pb, MATROSKA_ID_SEEKHEAD, seekhead->reserved_size);
313 for (i = 0; i < seekhead->num_entries; i++) {
314 mkv_seekhead_entry *entry = &seekhead->entries[i];
315
316 seekentry = start_ebml_master(pb, MATROSKA_ID_SEEKENTRY, MAX_SEEKENTRY_SIZE);
317
318 put_ebml_id(pb, MATROSKA_ID_SEEKID);
319 put_ebml_num(pb, ebml_id_size(entry->elementid), 0);
320 put_ebml_id(pb, entry->elementid);
321
322 put_ebml_uint(pb, MATROSKA_ID_SEEKPOSITION, entry->segmentpos);
323 end_ebml_master(pb, seekentry);
324 }
325 end_ebml_master(pb, metaseek);
326
327 if (seekhead->reserved_size > 0) {
328 uint64_t remaining = seekhead->filepos + seekhead->reserved_size - url_ftell(pb);
329 put_ebml_void(pb, remaining);
330 url_fseek(pb, currentpos, SEEK_SET);
331
332 currentpos = seekhead->filepos;
333 }
334 av_free(seekhead->entries);
335 av_free(seekhead);
336
337 return currentpos;
338 }
339
340 static mkv_cues * mkv_start_cues(int64_t segment_offset)
341 {
342 mkv_cues *cues = av_mallocz(sizeof(mkv_cues));
343 if (cues == NULL)
344 return NULL;
345
346 cues->segment_offset = segment_offset;
347 return cues;
348 }
349
350 static int mkv_add_cuepoint(mkv_cues *cues, int stream, int64_t ts, int64_t cluster_pos)
351 {
352 mkv_cuepoint *entries = cues->entries;
353
354 entries = av_realloc(entries, (cues->num_entries + 1) * sizeof(mkv_cuepoint));
355 if (entries == NULL)
356 return AVERROR(ENOMEM);
357
358 if (ts < 0)
359 return 0;
360
361 entries[cues->num_entries ].pts = ts;
362 entries[cues->num_entries ].tracknum = stream + 1;
363 entries[cues->num_entries++].cluster_pos = cluster_pos - cues->segment_offset;
364
365 cues->entries = entries;
366 return 0;
367 }
368
369 static int64_t mkv_write_cues(ByteIOContext *pb, mkv_cues *cues, int num_tracks)
370 {
371 ebml_master cues_element;
372 int64_t currentpos;
373 int i, j;
374
375 currentpos = url_ftell(pb);
376 cues_element = start_ebml_master(pb, MATROSKA_ID_CUES, 0);
377
378 for (i = 0; i < cues->num_entries; i++) {
379 ebml_master cuepoint, track_positions;
380 mkv_cuepoint *entry = &cues->entries[i];
381 uint64_t pts = entry->pts;
382
383 cuepoint = start_ebml_master(pb, MATROSKA_ID_POINTENTRY, MAX_CUEPOINT_SIZE(num_tracks));
384 put_ebml_uint(pb, MATROSKA_ID_CUETIME, pts);
385
386 // put all the entries from different tracks that have the exact same
387 // timestamp into the same CuePoint
388 for (j = 0; j < cues->num_entries - i && entry[j].pts == pts; j++) {
389 track_positions = start_ebml_master(pb, MATROSKA_ID_CUETRACKPOSITION, MAX_CUETRACKPOS_SIZE);
390 put_ebml_uint(pb, MATROSKA_ID_CUETRACK , entry[j].tracknum );
391 put_ebml_uint(pb, MATROSKA_ID_CUECLUSTERPOSITION, entry[j].cluster_pos);
392 end_ebml_master(pb, track_positions);
393 }
394 i += j - 1;
395 end_ebml_master(pb, cuepoint);
396 }
397 end_ebml_master(pb, cues_element);
398
399 av_free(cues->entries);
400 av_free(cues);
401 return currentpos;
402 }
403
404 static int put_xiph_codecpriv(AVFormatContext *s, ByteIOContext *pb, AVCodecContext *codec)
405 {
406 uint8_t *header_start[3];
407 int header_len[3];
408 int first_header_size;
409 int j;
410
411 if (codec->codec_id == CODEC_ID_VORBIS)
412 first_header_size = 30;
413 else
414 first_header_size = 42;
415
416 if (ff_split_xiph_headers(codec->extradata, codec->extradata_size,
417 first_header_size, header_start, header_len) < 0) {
418 av_log(s, AV_LOG_ERROR, "Extradata corrupt.\n");
419 return -1;
420 }
421
422 put_byte(pb, 2); // number packets - 1
423 for (j = 0; j < 2; j++) {
424 put_xiph_size(pb, header_len[j]);
425 }
426 for (j = 0; j < 3; j++)
427 put_buffer(pb, header_start[j], header_len[j]);
428
429 return 0;
430 }
431
432 static void get_aac_sample_rates(AVFormatContext *s, AVCodecContext *codec, int *sample_rate, int *output_sample_rate)
433 {
434 int sri;
435
436 if (codec->extradata_size < 2) {
437 av_log(s, AV_LOG_WARNING, "No AAC extradata, unable to determine samplerate.\n");
438 return;
439 }
440
441 sri = ((codec->extradata[0] << 1) & 0xE) | (codec->extradata[1] >> 7);
442 if (sri > 12) {
443 av_log(s, AV_LOG_WARNING, "AAC samplerate index out of bounds\n");
444 return;
445 }
446 *sample_rate = ff_mpeg4audio_sample_rates[sri];
447
448 // if sbr, get output sample rate as well
449 if (codec->extradata_size == 5) {
450 sri = (codec->extradata[4] >> 3) & 0xF;
451 if (sri > 12) {
452 av_log(s, AV_LOG_WARNING, "AAC output samplerate index out of bounds\n");
453 return;
454 }
455 *output_sample_rate = ff_mpeg4audio_sample_rates[sri];
456 }
457 }
458
459 static int mkv_write_codecprivate(AVFormatContext *s, ByteIOContext *pb, AVCodecContext *codec, int native_id, int qt_id)
460 {
461 ByteIOContext *dyn_cp;
462 uint8_t *codecpriv;
463 int ret, codecpriv_size;
464
465 ret = url_open_dyn_buf(&dyn_cp);
466 if(ret < 0)
467 return ret;
468
469 if (native_id) {
470 if (codec->codec_id == CODEC_ID_VORBIS || codec->codec_id == CODEC_ID_THEORA)
471 ret = put_xiph_codecpriv(s, dyn_cp, codec);
472 else if (codec->codec_id == CODEC_ID_FLAC)
473 ret = ff_flac_write_header(dyn_cp, codec, 1);
474 else if (codec->codec_id == CODEC_ID_H264)
475 ret = ff_isom_write_avcc(dyn_cp, codec->extradata, codec->extradata_size);
476 else if (codec->extradata_size)
477 put_buffer(dyn_cp, codec->extradata, codec->extradata_size);
478 } else if (codec->codec_type == CODEC_TYPE_VIDEO) {
479 if (qt_id) {
480 if (!codec->codec_tag)
481 codec->codec_tag = ff_codec_get_tag(codec_movvideo_tags, codec->codec_id);
482 if (codec->extradata_size)
483 put_buffer(dyn_cp, codec->extradata, codec->extradata_size);
484 } else {
485 if (!codec->codec_tag)
486 codec->codec_tag = ff_codec_get_tag(ff_codec_bmp_tags, codec->codec_id);
487 if (!codec->codec_tag) {
488 av_log(s, AV_LOG_ERROR, "No bmp codec ID found.\n");
489 ret = -1;
490 }
491
492 ff_put_bmp_header(dyn_cp, codec, ff_codec_bmp_tags, 0);
493 }
494
495 } else if (codec->codec_type == CODEC_TYPE_AUDIO) {
496 if (!codec->codec_tag)
497 codec->codec_tag = ff_codec_get_tag(ff_codec_wav_tags, codec->codec_id);
498 if (!codec->codec_tag) {
499 av_log(s, AV_LOG_ERROR, "No wav codec ID found.\n");
500 ret = -1;
501 }
502
503 ff_put_wav_header(dyn_cp, codec);
504 }
505
506 codecpriv_size = url_close_dyn_buf(dyn_cp, &codecpriv);
507 if (codecpriv_size)
508 put_ebml_binary(pb, MATROSKA_ID_CODECPRIVATE, codecpriv, codecpriv_size);
509 av_free(codecpriv);
510 return ret;
511 }
512
513 static int mkv_write_tracks(AVFormatContext *s)
514 {
515 MatroskaMuxContext *mkv = s->priv_data;
516 ByteIOContext *pb = s->pb;
517 ebml_master tracks;
518 int i, j, ret;
519
520 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_TRACKS, url_ftell(pb));
521 if (ret < 0) return ret;
522
523 tracks = start_ebml_master(pb, MATROSKA_ID_TRACKS, 0);
524 for (i = 0; i < s->nb_streams; i++) {
525 AVStream *st = s->streams[i];
526 AVCodecContext *codec = st->codec;
527 ebml_master subinfo, track;
528 int native_id = 0;
529 int qt_id = 0;
530 int bit_depth = av_get_bits_per_sample(codec->codec_id);
531 int sample_rate = codec->sample_rate;
532 int output_sample_rate = 0;
533 AVMetadataTag *tag;
534
535 if (!bit_depth)
536 bit_depth = av_get_bits_per_sample_format(codec->sample_fmt);
537
538 if (codec->codec_id == CODEC_ID_AAC)
539 get_aac_sample_rates(s, codec, &sample_rate, &output_sample_rate);
540
541 track = start_ebml_master(pb, MATROSKA_ID_TRACKENTRY, 0);
542 put_ebml_uint (pb, MATROSKA_ID_TRACKNUMBER , i + 1);
543 put_ebml_uint (pb, MATROSKA_ID_TRACKUID , i + 1);
544 put_ebml_uint (pb, MATROSKA_ID_TRACKFLAGLACING , 0); // no lacing (yet)
545 put_ebml_float(pb, MATROSKA_ID_TRACKTIMECODESCALE, 1.0);
546
547 if ((tag = av_metadata_get(st->metadata, "title", NULL, 0)))
548 put_ebml_string(pb, MATROSKA_ID_TRACKNAME, tag->value);
549 tag = av_metadata_get(st->metadata, "language", NULL, 0);
550 put_ebml_string(pb, MATROSKA_ID_TRACKLANGUAGE, tag ? tag->value:"und");
551
552 if (st->disposition)
553 put_ebml_uint(pb, MATROSKA_ID_TRACKFLAGDEFAULT, !!(st->disposition & AV_DISPOSITION_DEFAULT));
554
555 // look for a codec ID string specific to mkv to use,
556 // if none are found, use AVI codes
557 for (j = 0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++) {
558 if (ff_mkv_codec_tags[j].id == codec->codec_id) {
559 put_ebml_string(pb, MATROSKA_ID_CODECID, ff_mkv_codec_tags[j].str);
560 native_id = 1;
561 break;
562 }
563 }
564
565 switch (codec->codec_type) {
566 case CODEC_TYPE_VIDEO:
567 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_VIDEO);
568
569 if (!native_id &&
570 ff_codec_get_tag(codec_movvideo_tags, codec->codec_id) &&
571 (!ff_codec_get_tag(ff_codec_bmp_tags, codec->codec_id)
572 || codec->codec_id == CODEC_ID_SVQ1
573 || codec->codec_id == CODEC_ID_SVQ3
574 || codec->codec_id == CODEC_ID_CINEPAK))
575 qt_id = 1;
576
577 if (qt_id)
578 put_ebml_string(pb, MATROSKA_ID_CODECID, "V_QUICKTIME");
579 else if (!native_id) {
580 // if there is no mkv-specific codec ID, use VFW mode
581 put_ebml_string(pb, MATROSKA_ID_CODECID, "V_MS/VFW/FOURCC");
582 mkv->tracks[i].write_dts = 1;
583 }
584
585 subinfo = start_ebml_master(pb, MATROSKA_ID_TRACKVIDEO, 0);
586 // XXX: interlace flag?
587 put_ebml_uint (pb, MATROSKA_ID_VIDEOPIXELWIDTH , codec->width);
588 put_ebml_uint (pb, MATROSKA_ID_VIDEOPIXELHEIGHT, codec->height);
589 if (st->sample_aspect_ratio.num) {
590 int d_width = codec->width*av_q2d(st->sample_aspect_ratio);
591 put_ebml_uint(pb, MATROSKA_ID_VIDEODISPLAYWIDTH , d_width);
592 put_ebml_uint(pb, MATROSKA_ID_VIDEODISPLAYHEIGHT, codec->height);
593 }
594 end_ebml_master(pb, subinfo);
595 break;
596
597 case CODEC_TYPE_AUDIO:
598 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_AUDIO);
599
600 if (!native_id)
601 // no mkv-specific ID, use ACM mode
602 put_ebml_string(pb, MATROSKA_ID_CODECID, "A_MS/ACM");
603
604 subinfo = start_ebml_master(pb, MATROSKA_ID_TRACKAUDIO, 0);
605 put_ebml_uint (pb, MATROSKA_ID_AUDIOCHANNELS , codec->channels);
606 put_ebml_float (pb, MATROSKA_ID_AUDIOSAMPLINGFREQ, sample_rate);
607 if (output_sample_rate)
608 put_ebml_float(pb, MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, output_sample_rate);
609 if (bit_depth)
610 put_ebml_uint(pb, MATROSKA_ID_AUDIOBITDEPTH, bit_depth);
611 end_ebml_master(pb, subinfo);
612 break;
613
614 case CODEC_TYPE_SUBTITLE:
615 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_SUBTITLE);
616 break;
617 default:
618 av_log(s, AV_LOG_ERROR, "Only audio, video, and subtitles are supported for Matroska.");
619 break;
620 }
621 ret = mkv_write_codecprivate(s, pb, codec, native_id, qt_id);
622 if (ret < 0) return ret;
623
624 end_ebml_master(pb, track);
625
626 // ms precision is the de-facto standard timescale for mkv files
627 av_set_pts_info(st, 64, 1, 1000);
628 }
629 end_ebml_master(pb, tracks);
630 return 0;
631 }
632
633 static int mkv_write_chapters(AVFormatContext *s)
634 {
635 MatroskaMuxContext *mkv = s->priv_data;
636 ByteIOContext *pb = s->pb;
637 ebml_master chapters, editionentry;
638 AVRational scale = {1, 1E9};
639 int i, ret;
640
641 if (!s->nb_chapters)
642 return 0;
643
644 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_CHAPTERS, url_ftell(pb));
645 if (ret < 0) return ret;
646
647 chapters = start_ebml_master(pb, MATROSKA_ID_CHAPTERS , 0);
648 editionentry = start_ebml_master(pb, MATROSKA_ID_EDITIONENTRY, 0);
649 put_ebml_uint(pb, MATROSKA_ID_EDITIONFLAGDEFAULT, 1);
650 put_ebml_uint(pb, MATROSKA_ID_EDITIONFLAGHIDDEN , 0);
651 for (i = 0; i < s->nb_chapters; i++) {
652 ebml_master chapteratom, chapterdisplay;
653 AVChapter *c = s->chapters[i];
654 AVMetadataTag *t = NULL;
655
656 chapteratom = start_ebml_master(pb, MATROSKA_ID_CHAPTERATOM, 0);
657 put_ebml_uint(pb, MATROSKA_ID_CHAPTERUID, c->id);
658 put_ebml_uint(pb, MATROSKA_ID_CHAPTERTIMESTART,
659 av_rescale_q(c->start, c->time_base, scale));
660 put_ebml_uint(pb, MATROSKA_ID_CHAPTERTIMEEND,
661 av_rescale_q(c->end, c->time_base, scale));
662 put_ebml_uint(pb, MATROSKA_ID_CHAPTERFLAGHIDDEN , 0);
663 put_ebml_uint(pb, MATROSKA_ID_CHAPTERFLAGENABLED, 1);
664 if ((t = av_metadata_get(c->metadata, "title", NULL, 0))) {
665 chapterdisplay = start_ebml_master(pb, MATROSKA_ID_CHAPTERDISPLAY, 0);
666 put_ebml_string(pb, MATROSKA_ID_CHAPSTRING, t->value);
667 put_ebml_string(pb, MATROSKA_ID_CHAPLANG , "und");
668 end_ebml_master(pb, chapterdisplay);
669 }
670 end_ebml_master(pb, chapteratom);
671 }
672 end_ebml_master(pb, editionentry);
673 end_ebml_master(pb, chapters);
674 return 0;
675 }
676
677 static int mkv_write_header(AVFormatContext *s)
678 {
679 MatroskaMuxContext *mkv = s->priv_data;
680 ByteIOContext *pb = s->pb;
681 ebml_master ebml_header, segment_info;
682 AVMetadataTag *tag;
683 int ret;
684
685 mkv->md5_ctx = av_mallocz(av_md5_size);
686 av_md5_init(mkv->md5_ctx);
687 mkv->tracks = av_mallocz(s->nb_streams * sizeof(*mkv->tracks));
688
689 ebml_header = start_ebml_master(pb, EBML_ID_HEADER, 0);
690 put_ebml_uint (pb, EBML_ID_EBMLVERSION , 1);
691 put_ebml_uint (pb, EBML_ID_EBMLREADVERSION , 1);
692 put_ebml_uint (pb, EBML_ID_EBMLMAXIDLENGTH , 4);
693 put_ebml_uint (pb, EBML_ID_EBMLMAXSIZELENGTH , 8);
694 put_ebml_string (pb, EBML_ID_DOCTYPE , "matroska");
695 put_ebml_uint (pb, EBML_ID_DOCTYPEVERSION , 2);
696 put_ebml_uint (pb, EBML_ID_DOCTYPEREADVERSION , 2);
697 end_ebml_master(pb, ebml_header);
698
699 mkv->segment = start_ebml_master(pb, MATROSKA_ID_SEGMENT, 0);
700 mkv->segment_offset = url_ftell(pb);
701
702 // we write 2 seek heads - one at the end of the file to point to each
703 // cluster, and one at the beginning to point to all other level one
704 // elements (including the seek head at the end of the file), which
705 // isn't more than 10 elements if we only write one of each other
706 // currently defined level 1 element
707 mkv->main_seekhead = mkv_start_seekhead(pb, mkv->segment_offset, 10);
708 mkv->cluster_seekhead = mkv_start_seekhead(pb, mkv->segment_offset, 0);
709 if (mkv->main_seekhead == NULL || mkv->cluster_seekhead == NULL)
710 return AVERROR(ENOMEM);
711
712 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_INFO, url_ftell(pb));
713 if (ret < 0) return ret;
714
715 segment_info = start_ebml_master(pb, MATROSKA_ID_INFO, 0);
716 put_ebml_uint(pb, MATROSKA_ID_TIMECODESCALE, 1000000);
717 if ((tag = av_metadata_get(s->metadata, "title", NULL, 0)))
718 put_ebml_string(pb, MATROSKA_ID_TITLE, tag->value);
719 if (!(s->streams[0]->codec->flags & CODEC_FLAG_BITEXACT)) {
720 put_ebml_string(pb, MATROSKA_ID_MUXINGAPP , LIBAVFORMAT_IDENT);
721 put_ebml_string(pb, MATROSKA_ID_WRITINGAPP, LIBAVFORMAT_IDENT);
722
723 // reserve space to write the segment UID later
724 mkv->segment_uid = url_ftell(pb);
725 put_ebml_void(pb, 19);
726 }
727
728 // reserve space for the duration
729 mkv->duration = 0;
730 mkv->duration_offset = url_ftell(pb);
731 put_ebml_void(pb, 11); // assumes double-precision float to be written
732 end_ebml_master(pb, segment_info);
733
734 ret = mkv_write_tracks(s);
735 if (ret < 0) return ret;
736
737 ret = mkv_write_chapters(s);
738 if (ret < 0) return ret;
739
740 if (url_is_streamed(s->pb))
741 mkv_write_seekhead(pb, mkv->main_seekhead);
742
743 mkv->cues = mkv_start_cues(mkv->segment_offset);
744 if (mkv->cues == NULL)
745 return AVERROR(ENOMEM);
746
747 put_flush_packet(pb);
748 return 0;
749 }
750
751 static int mkv_blockgroup_size(int pkt_size)
752 {
753 int size = pkt_size + 4;
754 size += ebml_num_size(size);
755 size += 2; // EBML ID for block and block duration
756 size += 8; // max size of block duration
757 size += ebml_num_size(size);
758 size += 1; // blockgroup EBML ID
759 return size;
760 }
761
762 static int ass_get_duration(const uint8_t *p)
763 {
764 int sh, sm, ss, sc, eh, em, es, ec;
765 uint64_t start, end;
766
767 if (sscanf(p, "%*[^,],%d:%d:%d%*c%d,%d:%d:%d%*c%d",
768 &sh, &sm, &ss, &sc, &eh, &em, &es, &ec) != 8)
769 return 0;
770 start = 3600000*sh + 60000*sm + 1000*ss + 10*sc;
771 end = 3600000*eh + 60000*em + 1000*es + 10*ec;
772 return end - start;
773 }
774
775 static int mkv_write_ass_blocks(AVFormatContext *s, ByteIOContext *pb, AVPacket *pkt)
776 {
777 MatroskaMuxContext *mkv = s->priv_data;
778 int i, layer = 0, max_duration = 0, size, line_size, data_size = pkt->size;
779 uint8_t *start, *end, *data = pkt->data;
780 ebml_master blockgroup;
781 char buffer[2048];
782
783 while (data_size) {
784 int duration = ass_get_duration(data);
785 max_duration = FFMAX(duration, max_duration);
786 end = memchr(data, '\n', data_size);
787 size = line_size = end ? end-data+1 : data_size;
788 size -= end ? (end[-1]=='\r')+1 : 0;
789 start = data;
790 for (i=0; i<3; i++, start++)
791 if (!(start = memchr(start, ',', size-(start-data))))
792 return max_duration;
793 size -= start - data;
794 sscanf(data, "Dialogue: %d,", &layer);
795 i = snprintf(buffer, sizeof(buffer), "%"PRId64",%d,",
796 s->streams[pkt->stream_index]->nb_frames++, layer);
797 size = FFMIN(i+size, sizeof(buffer));
798 memcpy(buffer+i, start, size-i);
799
800 av_log(s, AV_LOG_DEBUG, "Writing block at offset %" PRIu64 ", size %d, "
801 "pts %" PRId64 ", duration %d\n",
802 url_ftell(pb), size, pkt->pts, duration);
803 blockgroup = start_ebml_master(pb, MATROSKA_ID_BLOCKGROUP, mkv_blockgroup_size(size));
804 put_ebml_id(pb, MATROSKA_ID_BLOCK);
805 put_ebml_num(pb, size+4, 0);
806 put_byte(pb, 0x80 | (pkt->stream_index + 1)); // this assumes stream_index is less than 126
807 put_be16(pb, pkt->pts - mkv->cluster_pts);
808 put_byte(pb, 0);
809 put_buffer(pb, buffer, size);
810 put_ebml_uint(pb, MATROSKA_ID_BLOCKDURATION, duration);
811 end_ebml_master(pb, blockgroup);
812
813 data += line_size;
814 data_size -= line_size;
815 }
816
817 return max_duration;
818 }
819
820 static void mkv_write_block(AVFormatContext *s, ByteIOContext *pb,
821 unsigned int blockid, AVPacket *pkt, int flags)
822 {
823 MatroskaMuxContext *mkv = s->priv_data;
824 AVCodecContext *codec = s->streams[pkt->stream_index]->codec;
825 uint8_t *data = NULL;
826 int size = pkt->size;
827 int64_t ts = mkv->tracks[pkt->stream_index].write_dts ? pkt->dts : pkt->pts;
828
829 av_log(s, AV_LOG_DEBUG, "Writing block at offset %" PRIu64 ", size %d, "
830 "pts %" PRId64 ", dts %" PRId64 ", duration %d, flags %d\n",
831 url_ftell(pb), pkt->size, pkt->pts, pkt->dts, pkt->duration, flags);
832 if (codec->codec_id == CODEC_ID_H264 && codec->extradata_size > 0 &&
833 (AV_RB24(codec->extradata) == 1 || AV_RB32(codec->extradata) == 1))
834 ff_avc_parse_nal_units_buf(pkt->data, &data, &size);
835 else
836 data = pkt->data;
837 put_ebml_id(pb, blockid);
838 put_ebml_num(pb, size+4, 0);
839 put_byte(pb, 0x80 | (pkt->stream_index + 1)); // this assumes stream_index is less than 126
840 put_be16(pb, ts - mkv->cluster_pts);
841 put_byte(pb, flags);
842 put_buffer(pb, data, size);
843 if (data != pkt->data)
844 av_free(data);
845 }
846
847 static void mkv_flush_dynbuf(AVFormatContext *s)
848 {
849 MatroskaMuxContext *mkv = s->priv_data;
850 int bufsize;
851 uint8_t *dyn_buf;
852
853 if (!mkv->dyn_bc)
854 return;
855
856 bufsize = url_close_dyn_buf(mkv->dyn_bc, &dyn_buf);
857 put_buffer(s->pb, dyn_buf, bufsize);
858 av_free(dyn_buf);
859 mkv->dyn_bc = NULL;
860 }
861
862 static int mkv_write_packet(AVFormatContext *s, AVPacket *pkt)
863 {
864 MatroskaMuxContext *mkv = s->priv_data;
865 ByteIOContext *pb = s->pb;
866 AVCodecContext *codec = s->streams[pkt->stream_index]->codec;
867 int keyframe = !!(pkt->flags & PKT_FLAG_KEY);
868 int duration = pkt->duration;
869 int ret;
870 int64_t ts = mkv->tracks[pkt->stream_index].write_dts ? pkt->dts : pkt->pts;
871
872 if (ts == AV_NOPTS_VALUE) {
873 av_log(s, AV_LOG_ERROR, "Can't write packet with unknown timestamp\n");
874 return AVERROR(EINVAL);
875 }
876
877 if (url_is_streamed(s->pb)) {
878 if (!mkv->dyn_bc)
879 url_open_dyn_buf(&mkv->dyn_bc);
880 pb = mkv->dyn_bc;
881 }
882
883 if (!mkv->cluster_pos) {
884 ret = mkv_add_seekhead_entry(mkv->cluster_seekhead, MATROSKA_ID_CLUSTER, url_ftell(pb));
885 if (ret < 0) return ret;
886
887 mkv->cluster_pos = url_ftell(s->pb);
888 mkv->cluster = start_ebml_master(pb, MATROSKA_ID_CLUSTER, 0);
889 put_ebml_uint(pb, MATROSKA_ID_CLUSTERTIMECODE, FFMAX(0, ts));
890 mkv->cluster_pts = FFMAX(0, ts);
891 av_md5_update(mkv->md5_ctx, pkt->data, FFMIN(200, pkt->size));
892 }
893
894 if (codec->codec_type != CODEC_TYPE_SUBTITLE) {
895 mkv_write_block(s, pb, MATROSKA_ID_SIMPLEBLOCK, pkt, keyframe << 7);
896 } else if (codec->codec_id == CODEC_ID_SSA) {
897 duration = mkv_write_ass_blocks(s, pb, pkt);
898 } else {
899 ebml_master blockgroup = start_ebml_master(pb, MATROSKA_ID_BLOCKGROUP, mkv_blockgroup_size(pkt->size));
900 duration = pkt->convergence_duration;
901 mkv_write_block(s, pb, MATROSKA_ID_BLOCK, pkt, 0);
902 put_ebml_uint(pb, MATROSKA_ID_BLOCKDURATION, duration);
903 end_ebml_master(pb, blockgroup);
904 }
905
906 if (codec->codec_type == CODEC_TYPE_VIDEO && keyframe) {
907 ret = mkv_add_cuepoint(mkv->cues, pkt->stream_index, ts, mkv->cluster_pos);
908 if (ret < 0) return ret;
909 }
910
911 // start a new cluster every 5 MB or 5 sec, or 32k / 1 sec for streaming
912 if ((url_is_streamed(s->pb) && (url_ftell(pb) > 32*1024 || ts > mkv->cluster_pts + 1000))
913 || url_ftell(pb) > mkv->cluster_pos + 5*1024*1024 || ts > mkv->cluster_pts + 5000) {
914 av_log(s, AV_LOG_DEBUG, "Starting new cluster at offset %" PRIu64
915 " bytes, pts %" PRIu64 "\n", url_ftell(pb), ts);
916 end_ebml_master(pb, mkv->cluster);
917 mkv->cluster_pos = 0;
918 if (mkv->dyn_bc)
919 mkv_flush_dynbuf(s);
920 }
921
922 mkv->duration = FFMAX(mkv->duration, ts + duration);
923 return 0;
924 }
925
926 static int mkv_write_trailer(AVFormatContext *s)
927 {
928 MatroskaMuxContext *mkv = s->priv_data;
929 ByteIOContext *pb = s->pb;
930 int64_t currentpos, second_seekhead, cuespos;
931 int ret;
932
933 if (mkv->dyn_bc) {
934 end_ebml_master(mkv->dyn_bc, mkv->cluster);
935 mkv_flush_dynbuf(s);
936 } else if (mkv->cluster_pos) {
937 end_ebml_master(pb, mkv->cluster);
938 }
939
940 if (!url_is_streamed(pb)) {
941 cuespos = mkv_write_cues(pb, mkv->cues, s->nb_streams);
942 second_seekhead = mkv_write_seekhead(pb, mkv->cluster_seekhead);
943
944 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_CUES , cuespos);
945 if (ret < 0) return ret;
946 if (second_seekhead >= 0) {
947 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_SEEKHEAD, second_seekhead);
948 if (ret < 0) return ret;
949 }
950 mkv_write_seekhead(pb, mkv->main_seekhead);
951
952 // update the duration
953 av_log(s, AV_LOG_DEBUG, "end duration = %" PRIu64 "\n", mkv->duration);
954 currentpos = url_ftell(pb);
955 url_fseek(pb, mkv->duration_offset, SEEK_SET);
956 put_ebml_float(pb, MATROSKA_ID_DURATION, mkv->duration);
957
958 // write the md5sum of some frames as the segment UID
959 if (!(s->streams[0]->codec->flags & CODEC_FLAG_BITEXACT)) {
960 uint8_t segment_uid[16];
961 av_md5_final(mkv->md5_ctx, segment_uid);
962 url_fseek(pb, mkv->segment_uid, SEEK_SET);
963 put_ebml_binary(pb, MATROSKA_ID_SEGMENTUID, segment_uid, 16);
964 }
965 url_fseek(pb, currentpos, SEEK_SET);
966 }
967
968 end_ebml_master(pb, mkv->segment);
969 av_free(mkv->md5_ctx);
970 av_free(mkv->tracks);
971 put_flush_packet(pb);
972 return 0;
973 }
974
975 AVOutputFormat matroska_muxer = {
976 "matroska",
977 NULL_IF_CONFIG_SMALL("Matroska file format"),
978 "video/x-matroska",
979 "mkv",
980 sizeof(MatroskaMuxContext),
981 CODEC_ID_MP2,
982 CODEC_ID_MPEG4,
983 mkv_write_header,
984 mkv_write_packet,
985 mkv_write_trailer,
986 .flags = AVFMT_GLOBALHEADER | AVFMT_VARIABLE_FPS,
987 .codec_tag = (const AVCodecTag* const []){ff_codec_bmp_tags, ff_codec_wav_tags, 0},
988 .subtitle_codec = CODEC_ID_TEXT,
989 };
990
991 AVOutputFormat matroska_audio_muxer = {
992 "matroska",
993 NULL_IF_CONFIG_SMALL("Matroska file format"),
994 "audio/x-matroska",
995 "mka",
996 sizeof(MatroskaMuxContext),
997 CODEC_ID_MP2,
998 CODEC_ID_NONE,
999 mkv_write_header,
1000 mkv_write_packet,
1001 mkv_write_trailer,
1002 .flags = AVFMT_GLOBALHEADER,
1003 .codec_tag = (const AVCodecTag* const []){ff_codec_wav_tags, 0},
1004 };