ad4189384f527f115f9b2d5b226e9275837ecd9b
[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 MatroskaMuxContext {
65 ebml_master segment;
66 int64_t segment_offset;
67 int64_t segment_uid;
68 ebml_master cluster;
69 int64_t cluster_pos; ///< file offset of the current cluster
70 uint64_t cluster_pts;
71 int64_t duration_offset;
72 uint64_t duration;
73 mkv_seekhead *main_seekhead;
74 mkv_seekhead *cluster_seekhead;
75 mkv_cues *cues;
76
77 struct AVMD5 *md5_ctx;
78 } MatroskaMuxContext;
79
80
81 /** 2 bytes * 3 for EBML IDs, 3 1-byte EBML lengths, 8 bytes for 64 bit
82 * offset, 4 bytes for target EBML ID */
83 #define MAX_SEEKENTRY_SIZE 21
84
85 /** per-cuepoint-track - 3 1-byte EBML IDs, 3 1-byte EBML sizes, 2
86 * 8-byte uint max */
87 #define MAX_CUETRACKPOS_SIZE 22
88
89 /** per-cuepoint - 2 1-byte EBML IDs, 2 1-byte EBML sizes, 8-byte uint max */
90 #define MAX_CUEPOINT_SIZE(num_tracks) 12 + MAX_CUETRACKPOS_SIZE*num_tracks
91
92
93 static int ebml_id_size(unsigned int id)
94 {
95 return (av_log2(id+1)-1)/7+1;
96 }
97
98 static void put_ebml_id(ByteIOContext *pb, unsigned int id)
99 {
100 int i = ebml_id_size(id);
101 while (i--)
102 put_byte(pb, id >> (i*8));
103 }
104
105 /**
106 * Write an EBML size meaning "unknown size".
107 *
108 * @param bytes The number of bytes the size should occupy (maximum: 8).
109 */
110 static void put_ebml_size_unknown(ByteIOContext *pb, int bytes)
111 {
112 assert(bytes <= 8);
113 put_byte(pb, 0x1ff >> bytes);
114 while (--bytes)
115 put_byte(pb, 0xff);
116 }
117
118 /**
119 * Calculate how many bytes are needed to represent a given number in EBML.
120 */
121 static int ebml_num_size(uint64_t num)
122 {
123 int bytes = 1;
124 while ((num+1) >> bytes*7) bytes++;
125 return bytes;
126 }
127
128 /**
129 * Write a number in EBML variable length format.
130 *
131 * @param bytes The number of bytes that need to be used to write the number.
132 * If zero, any number of bytes can be used.
133 */
134 static void put_ebml_num(ByteIOContext *pb, uint64_t num, int bytes)
135 {
136 int i, needed_bytes = ebml_num_size(num);
137
138 // sizes larger than this are currently undefined in EBML
139 assert(num < (1ULL<<56)-1);
140
141 if (bytes == 0)
142 // don't care how many bytes are used, so use the min
143 bytes = needed_bytes;
144 // the bytes needed to write the given size would exceed the bytes
145 // that we need to use, so write unknown size. This shouldn't happen.
146 assert(bytes >= needed_bytes);
147
148 num |= 1ULL << bytes*7;
149 for (i = bytes - 1; i >= 0; i--)
150 put_byte(pb, num >> i*8);
151 }
152
153 static void put_ebml_uint(ByteIOContext *pb, unsigned int elementid, uint64_t val)
154 {
155 int i, bytes = 1;
156 uint64_t tmp = val;
157 while (tmp>>=8) bytes++;
158
159 put_ebml_id(pb, elementid);
160 put_ebml_num(pb, bytes, 0);
161 for (i = bytes - 1; i >= 0; i--)
162 put_byte(pb, val >> i*8);
163 }
164
165 static void put_ebml_float(ByteIOContext *pb, unsigned int elementid, double val)
166 {
167 put_ebml_id(pb, elementid);
168 put_ebml_num(pb, 8, 0);
169 put_be64(pb, av_dbl2int(val));
170 }
171
172 static void put_ebml_binary(ByteIOContext *pb, unsigned int elementid,
173 const uint8_t *buf, int size)
174 {
175 put_ebml_id(pb, elementid);
176 put_ebml_num(pb, size, 0);
177 put_buffer(pb, buf, size);
178 }
179
180 static void put_ebml_string(ByteIOContext *pb, unsigned int elementid, const char *str)
181 {
182 put_ebml_binary(pb, elementid, str, strlen(str));
183 }
184
185 /**
186 * Writes a void element of a given size. Useful for reserving space in
187 * the file to be written to later.
188 *
189 * @param size The number of bytes to reserve, which must be at least 2.
190 */
191 static void put_ebml_void(ByteIOContext *pb, uint64_t size)
192 {
193 int64_t currentpos = url_ftell(pb);
194
195 assert(size >= 2);
196
197 put_ebml_id(pb, EBML_ID_VOID);
198 // we need to subtract the length needed to store the size from the
199 // size we need to reserve so 2 cases, we use 8 bytes to store the
200 // size if possible, 1 byte otherwise
201 if (size < 10)
202 put_ebml_num(pb, size-1, 0);
203 else
204 put_ebml_num(pb, size-9, 8);
205 while(url_ftell(pb) < currentpos + size)
206 put_byte(pb, 0);
207 }
208
209 static ebml_master start_ebml_master(ByteIOContext *pb, unsigned int elementid, uint64_t expectedsize)
210 {
211 int bytes = expectedsize ? ebml_num_size(expectedsize) : 8;
212 put_ebml_id(pb, elementid);
213 put_ebml_size_unknown(pb, bytes);
214 return (ebml_master){ url_ftell(pb), bytes };
215 }
216
217 static void end_ebml_master(ByteIOContext *pb, ebml_master master)
218 {
219 int64_t pos = url_ftell(pb);
220
221 // leave the unknown size for masters when streaming
222 if (url_is_streamed(pb))
223 return;
224
225 url_fseek(pb, master.pos - master.sizebytes, SEEK_SET);
226 put_ebml_num(pb, pos - master.pos, master.sizebytes);
227 url_fseek(pb, pos, SEEK_SET);
228 }
229
230 static void put_xiph_size(ByteIOContext *pb, int size)
231 {
232 int i;
233 for (i = 0; i < size / 255; i++)
234 put_byte(pb, 255);
235 put_byte(pb, size % 255);
236 }
237
238 /**
239 * Initialize a mkv_seekhead element to be ready to index level 1 Matroska
240 * elements. If a maximum number of elements is specified, enough space
241 * will be reserved at the current file location to write a seek head of
242 * that size.
243 *
244 * @param segment_offset The absolute offset to the position in the file
245 * where the segment begins.
246 * @param numelements The maximum number of elements that will be indexed
247 * by this seek head, 0 if unlimited.
248 */
249 static mkv_seekhead * mkv_start_seekhead(ByteIOContext *pb, int64_t segment_offset, int numelements)
250 {
251 mkv_seekhead *new_seekhead = av_mallocz(sizeof(mkv_seekhead));
252 if (new_seekhead == NULL)
253 return NULL;
254
255 new_seekhead->segment_offset = segment_offset;
256
257 if (numelements > 0) {
258 new_seekhead->filepos = url_ftell(pb);
259 // 21 bytes max for a seek entry, 10 bytes max for the SeekHead ID
260 // and size, and 3 bytes to guarantee that an EBML void element
261 // will fit afterwards
262 new_seekhead->reserved_size = numelements * MAX_SEEKENTRY_SIZE + 13;
263 new_seekhead->max_entries = numelements;
264 put_ebml_void(pb, new_seekhead->reserved_size);
265 }
266 return new_seekhead;
267 }
268
269 static int mkv_add_seekhead_entry(mkv_seekhead *seekhead, unsigned int elementid, uint64_t filepos)
270 {
271 mkv_seekhead_entry *entries = seekhead->entries;
272
273 // don't store more elements than we reserved space for
274 if (seekhead->max_entries > 0 && seekhead->max_entries <= seekhead->num_entries)
275 return -1;
276
277 entries = av_realloc(entries, (seekhead->num_entries + 1) * sizeof(mkv_seekhead_entry));
278 if (entries == NULL)
279 return AVERROR(ENOMEM);
280
281 entries[seekhead->num_entries ].elementid = elementid;
282 entries[seekhead->num_entries++].segmentpos = filepos - seekhead->segment_offset;
283
284 seekhead->entries = entries;
285 return 0;
286 }
287
288 /**
289 * Write the seek head to the file and free it. If a maximum number of
290 * elements was specified to mkv_start_seekhead(), the seek head will
291 * be written at the location reserved for it. Otherwise, it is written
292 * at the current location in the file.
293 *
294 * @return The file offset where the seekhead was written.
295 */
296 static int64_t mkv_write_seekhead(ByteIOContext *pb, mkv_seekhead *seekhead)
297 {
298 ebml_master metaseek, seekentry;
299 int64_t currentpos;
300 int i;
301
302 currentpos = url_ftell(pb);
303
304 if (seekhead->reserved_size > 0)
305 url_fseek(pb, seekhead->filepos, SEEK_SET);
306
307 metaseek = start_ebml_master(pb, MATROSKA_ID_SEEKHEAD, seekhead->reserved_size);
308 for (i = 0; i < seekhead->num_entries; i++) {
309 mkv_seekhead_entry *entry = &seekhead->entries[i];
310
311 seekentry = start_ebml_master(pb, MATROSKA_ID_SEEKENTRY, MAX_SEEKENTRY_SIZE);
312
313 put_ebml_id(pb, MATROSKA_ID_SEEKID);
314 put_ebml_num(pb, ebml_id_size(entry->elementid), 0);
315 put_ebml_id(pb, entry->elementid);
316
317 put_ebml_uint(pb, MATROSKA_ID_SEEKPOSITION, entry->segmentpos);
318 end_ebml_master(pb, seekentry);
319 }
320 end_ebml_master(pb, metaseek);
321
322 if (seekhead->reserved_size > 0) {
323 uint64_t remaining = seekhead->filepos + seekhead->reserved_size - url_ftell(pb);
324 put_ebml_void(pb, remaining);
325 url_fseek(pb, currentpos, SEEK_SET);
326
327 currentpos = seekhead->filepos;
328 }
329 av_free(seekhead->entries);
330 av_free(seekhead);
331
332 return currentpos;
333 }
334
335 static mkv_cues * mkv_start_cues(int64_t segment_offset)
336 {
337 mkv_cues *cues = av_mallocz(sizeof(mkv_cues));
338 if (cues == NULL)
339 return NULL;
340
341 cues->segment_offset = segment_offset;
342 return cues;
343 }
344
345 static int mkv_add_cuepoint(mkv_cues *cues, AVPacket *pkt, int64_t cluster_pos)
346 {
347 mkv_cuepoint *entries = cues->entries;
348
349 entries = av_realloc(entries, (cues->num_entries + 1) * sizeof(mkv_cuepoint));
350 if (entries == NULL)
351 return AVERROR(ENOMEM);
352
353 entries[cues->num_entries ].pts = pkt->pts;
354 entries[cues->num_entries ].tracknum = pkt->stream_index + 1;
355 entries[cues->num_entries++].cluster_pos = cluster_pos - cues->segment_offset;
356
357 cues->entries = entries;
358 return 0;
359 }
360
361 static int64_t mkv_write_cues(ByteIOContext *pb, mkv_cues *cues, int num_tracks)
362 {
363 ebml_master cues_element;
364 int64_t currentpos;
365 int i, j;
366
367 currentpos = url_ftell(pb);
368 cues_element = start_ebml_master(pb, MATROSKA_ID_CUES, 0);
369
370 for (i = 0; i < cues->num_entries; i++) {
371 ebml_master cuepoint, track_positions;
372 mkv_cuepoint *entry = &cues->entries[i];
373 uint64_t pts = entry->pts;
374
375 cuepoint = start_ebml_master(pb, MATROSKA_ID_POINTENTRY, MAX_CUEPOINT_SIZE(num_tracks));
376 put_ebml_uint(pb, MATROSKA_ID_CUETIME, pts);
377
378 // put all the entries from different tracks that have the exact same
379 // timestamp into the same CuePoint
380 for (j = 0; j < cues->num_entries - i && entry[j].pts == pts; j++) {
381 track_positions = start_ebml_master(pb, MATROSKA_ID_CUETRACKPOSITION, MAX_CUETRACKPOS_SIZE);
382 put_ebml_uint(pb, MATROSKA_ID_CUETRACK , entry[j].tracknum );
383 put_ebml_uint(pb, MATROSKA_ID_CUECLUSTERPOSITION, entry[j].cluster_pos);
384 end_ebml_master(pb, track_positions);
385 }
386 i += j - 1;
387 end_ebml_master(pb, cuepoint);
388 }
389 end_ebml_master(pb, cues_element);
390
391 av_free(cues->entries);
392 av_free(cues);
393 return currentpos;
394 }
395
396 static int put_xiph_codecpriv(AVFormatContext *s, ByteIOContext *pb, AVCodecContext *codec)
397 {
398 uint8_t *header_start[3];
399 int header_len[3];
400 int first_header_size;
401 int j;
402
403 if (codec->codec_id == CODEC_ID_VORBIS)
404 first_header_size = 30;
405 else
406 first_header_size = 42;
407
408 if (ff_split_xiph_headers(codec->extradata, codec->extradata_size,
409 first_header_size, header_start, header_len) < 0) {
410 av_log(s, AV_LOG_ERROR, "Extradata corrupt.\n");
411 return -1;
412 }
413
414 put_byte(pb, 2); // number packets - 1
415 for (j = 0; j < 2; j++) {
416 put_xiph_size(pb, header_len[j]);
417 }
418 for (j = 0; j < 3; j++)
419 put_buffer(pb, header_start[j], header_len[j]);
420
421 return 0;
422 }
423
424 static void get_aac_sample_rates(AVFormatContext *s, AVCodecContext *codec, int *sample_rate, int *output_sample_rate)
425 {
426 int sri;
427
428 if (codec->extradata_size < 2) {
429 av_log(s, AV_LOG_WARNING, "No AAC extradata, unable to determine samplerate.\n");
430 return;
431 }
432
433 sri = ((codec->extradata[0] << 1) & 0xE) | (codec->extradata[1] >> 7);
434 if (sri > 12) {
435 av_log(s, AV_LOG_WARNING, "AAC samplerate index out of bounds\n");
436 return;
437 }
438 *sample_rate = ff_mpeg4audio_sample_rates[sri];
439
440 // if sbr, get output sample rate as well
441 if (codec->extradata_size == 5) {
442 sri = (codec->extradata[4] >> 3) & 0xF;
443 if (sri > 12) {
444 av_log(s, AV_LOG_WARNING, "AAC output samplerate index out of bounds\n");
445 return;
446 }
447 *output_sample_rate = ff_mpeg4audio_sample_rates[sri];
448 }
449 }
450
451 static int mkv_write_codecprivate(AVFormatContext *s, ByteIOContext *pb, AVCodecContext *codec, int native_id, int qt_id)
452 {
453 ByteIOContext *dyn_cp;
454 uint8_t *codecpriv;
455 int ret, codecpriv_size;
456
457 ret = url_open_dyn_buf(&dyn_cp);
458 if(ret < 0)
459 return ret;
460
461 if (native_id) {
462 if (codec->codec_id == CODEC_ID_VORBIS || codec->codec_id == CODEC_ID_THEORA)
463 ret = put_xiph_codecpriv(s, dyn_cp, codec);
464 else if (codec->codec_id == CODEC_ID_FLAC)
465 ret = ff_flac_write_header(dyn_cp, codec);
466 else if (codec->codec_id == CODEC_ID_H264)
467 ret = ff_isom_write_avcc(dyn_cp, codec->extradata, codec->extradata_size);
468 else if (codec->extradata_size)
469 put_buffer(dyn_cp, codec->extradata, codec->extradata_size);
470 } else if (codec->codec_type == CODEC_TYPE_VIDEO) {
471 if (qt_id) {
472 if (!codec->codec_tag)
473 codec->codec_tag = ff_codec_get_tag(codec_movvideo_tags, codec->codec_id);
474 if (codec->extradata_size)
475 put_buffer(dyn_cp, codec->extradata, codec->extradata_size);
476 } else {
477 if (!codec->codec_tag)
478 codec->codec_tag = ff_codec_get_tag(ff_codec_bmp_tags, codec->codec_id);
479 if (!codec->codec_tag) {
480 av_log(s, AV_LOG_ERROR, "No bmp codec ID found.");
481 ret = -1;
482 }
483
484 ff_put_bmp_header(dyn_cp, codec, ff_codec_bmp_tags, 0);
485 }
486
487 } else if (codec->codec_type == CODEC_TYPE_AUDIO) {
488 if (!codec->codec_tag)
489 codec->codec_tag = ff_codec_get_tag(ff_codec_wav_tags, codec->codec_id);
490 if (!codec->codec_tag) {
491 av_log(s, AV_LOG_ERROR, "No wav codec ID found.");
492 ret = -1;
493 }
494
495 ff_put_wav_header(dyn_cp, codec);
496 }
497
498 codecpriv_size = url_close_dyn_buf(dyn_cp, &codecpriv);
499 if (codecpriv_size)
500 put_ebml_binary(pb, MATROSKA_ID_CODECPRIVATE, codecpriv, codecpriv_size);
501 av_free(codecpriv);
502 return ret;
503 }
504
505 static int mkv_write_tracks(AVFormatContext *s)
506 {
507 MatroskaMuxContext *mkv = s->priv_data;
508 ByteIOContext *pb = s->pb;
509 ebml_master tracks;
510 int i, j, ret;
511
512 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_TRACKS, url_ftell(pb));
513 if (ret < 0) return ret;
514
515 tracks = start_ebml_master(pb, MATROSKA_ID_TRACKS, 0);
516 for (i = 0; i < s->nb_streams; i++) {
517 AVStream *st = s->streams[i];
518 AVCodecContext *codec = st->codec;
519 ebml_master subinfo, track;
520 int native_id = 0;
521 int qt_id = 0;
522 int bit_depth = av_get_bits_per_sample(codec->codec_id);
523 int sample_rate = codec->sample_rate;
524 int output_sample_rate = 0;
525 AVMetadataTag *tag;
526
527 if (!bit_depth)
528 bit_depth = av_get_bits_per_sample_format(codec->sample_fmt);
529
530 if (codec->codec_id == CODEC_ID_AAC)
531 get_aac_sample_rates(s, codec, &sample_rate, &output_sample_rate);
532
533 track = start_ebml_master(pb, MATROSKA_ID_TRACKENTRY, 0);
534 put_ebml_uint (pb, MATROSKA_ID_TRACKNUMBER , i + 1);
535 put_ebml_uint (pb, MATROSKA_ID_TRACKUID , i + 1);
536 put_ebml_uint (pb, MATROSKA_ID_TRACKFLAGLACING , 0); // no lacing (yet)
537 put_ebml_float(pb, MATROSKA_ID_TRACKTIMECODESCALE, 1.0);
538
539 if ((tag = av_metadata_get(st->metadata, "title", NULL, 0)))
540 put_ebml_string(pb, MATROSKA_ID_TRACKNAME, tag->value);
541 tag = av_metadata_get(st->metadata, "language", NULL, 0);
542 put_ebml_string(pb, MATROSKA_ID_TRACKLANGUAGE, tag ? tag->value:"und");
543
544 if (st->disposition)
545 put_ebml_uint(pb, MATROSKA_ID_TRACKFLAGDEFAULT, !!(st->disposition & AV_DISPOSITION_DEFAULT));
546
547 // look for a codec ID string specific to mkv to use,
548 // if none are found, use AVI codes
549 for (j = 0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++) {
550 if (ff_mkv_codec_tags[j].id == codec->codec_id) {
551 put_ebml_string(pb, MATROSKA_ID_CODECID, ff_mkv_codec_tags[j].str);
552 native_id = 1;
553 break;
554 }
555 }
556
557 switch (codec->codec_type) {
558 case CODEC_TYPE_VIDEO:
559 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_VIDEO);
560
561 if (!native_id &&
562 ff_codec_get_tag(codec_movvideo_tags, codec->codec_id) &&
563 (!ff_codec_get_tag(ff_codec_bmp_tags, codec->codec_id)
564 || codec->codec_id == CODEC_ID_SVQ1
565 || codec->codec_id == CODEC_ID_SVQ3
566 || codec->codec_id == CODEC_ID_CINEPAK))
567 qt_id = 1;
568
569 if (qt_id)
570 put_ebml_string(pb, MATROSKA_ID_CODECID, "V_QUICKTIME");
571 else if (!native_id)
572 // if there is no mkv-specific codec ID, use VFW mode
573 put_ebml_string(pb, MATROSKA_ID_CODECID, "V_MS/VFW/FOURCC");
574
575 subinfo = start_ebml_master(pb, MATROSKA_ID_TRACKVIDEO, 0);
576 // XXX: interlace flag?
577 put_ebml_uint (pb, MATROSKA_ID_VIDEOPIXELWIDTH , codec->width);
578 put_ebml_uint (pb, MATROSKA_ID_VIDEOPIXELHEIGHT, codec->height);
579 if (st->sample_aspect_ratio.num) {
580 int d_width = codec->width*av_q2d(st->sample_aspect_ratio);
581 put_ebml_uint(pb, MATROSKA_ID_VIDEODISPLAYWIDTH , d_width);
582 put_ebml_uint(pb, MATROSKA_ID_VIDEODISPLAYHEIGHT, codec->height);
583 }
584 end_ebml_master(pb, subinfo);
585 break;
586
587 case CODEC_TYPE_AUDIO:
588 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_AUDIO);
589
590 if (!native_id)
591 // no mkv-specific ID, use ACM mode
592 put_ebml_string(pb, MATROSKA_ID_CODECID, "A_MS/ACM");
593
594 subinfo = start_ebml_master(pb, MATROSKA_ID_TRACKAUDIO, 0);
595 put_ebml_uint (pb, MATROSKA_ID_AUDIOCHANNELS , codec->channels);
596 put_ebml_float (pb, MATROSKA_ID_AUDIOSAMPLINGFREQ, sample_rate);
597 if (output_sample_rate)
598 put_ebml_float(pb, MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, output_sample_rate);
599 if (bit_depth)
600 put_ebml_uint(pb, MATROSKA_ID_AUDIOBITDEPTH, bit_depth);
601 end_ebml_master(pb, subinfo);
602 break;
603
604 case CODEC_TYPE_SUBTITLE:
605 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_SUBTITLE);
606 break;
607 default:
608 av_log(s, AV_LOG_ERROR, "Only audio, video, and subtitles are supported for Matroska.");
609 break;
610 }
611 ret = mkv_write_codecprivate(s, pb, codec, native_id, qt_id);
612 if (ret < 0) return ret;
613
614 end_ebml_master(pb, track);
615
616 // ms precision is the de-facto standard timescale for mkv files
617 av_set_pts_info(st, 64, 1, 1000);
618 }
619 end_ebml_master(pb, tracks);
620 return 0;
621 }
622
623 static int mkv_write_chapters(AVFormatContext *s)
624 {
625 MatroskaMuxContext *mkv = s->priv_data;
626 ByteIOContext *pb = s->pb;
627 ebml_master chapters, editionentry;
628 AVRational scale = {1, 1E9};
629 int i, ret;
630
631 if (!s->nb_chapters)
632 return 0;
633
634 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_CHAPTERS, url_ftell(pb));
635 if (ret < 0) return ret;
636
637 chapters = start_ebml_master(pb, MATROSKA_ID_CHAPTERS , 0);
638 editionentry = start_ebml_master(pb, MATROSKA_ID_EDITIONENTRY, 0);
639 put_ebml_uint(pb, MATROSKA_ID_EDITIONFLAGDEFAULT, 1);
640 put_ebml_uint(pb, MATROSKA_ID_EDITIONFLAGHIDDEN , 0);
641 for (i = 0; i < s->nb_chapters; i++) {
642 ebml_master chapteratom, chapterdisplay;
643 AVChapter *c = s->chapters[i];
644 AVMetadataTag *t = NULL;
645
646 chapteratom = start_ebml_master(pb, MATROSKA_ID_CHAPTERATOM, 0);
647 put_ebml_uint(pb, MATROSKA_ID_CHAPTERUID, c->id);
648 put_ebml_uint(pb, MATROSKA_ID_CHAPTERTIMESTART,
649 av_rescale_q(c->start, c->time_base, scale));
650 put_ebml_uint(pb, MATROSKA_ID_CHAPTERTIMEEND,
651 av_rescale_q(c->end, c->time_base, scale));
652 put_ebml_uint(pb, MATROSKA_ID_CHAPTERFLAGHIDDEN , 0);
653 put_ebml_uint(pb, MATROSKA_ID_CHAPTERFLAGENABLED, 1);
654 if ((t = av_metadata_get(c->metadata, "title", NULL, 0))) {
655 chapterdisplay = start_ebml_master(pb, MATROSKA_ID_CHAPTERDISPLAY, 0);
656 put_ebml_string(pb, MATROSKA_ID_CHAPSTRING, t->value);
657 put_ebml_string(pb, MATROSKA_ID_CHAPLANG , "und");
658 end_ebml_master(pb, chapterdisplay);
659 }
660 end_ebml_master(pb, chapteratom);
661 }
662 end_ebml_master(pb, editionentry);
663 end_ebml_master(pb, chapters);
664 return 0;
665 }
666
667 static int mkv_write_header(AVFormatContext *s)
668 {
669 MatroskaMuxContext *mkv = s->priv_data;
670 ByteIOContext *pb = s->pb;
671 ebml_master ebml_header, segment_info;
672 AVMetadataTag *tag;
673 int ret;
674
675 mkv->md5_ctx = av_mallocz(av_md5_size);
676 av_md5_init(mkv->md5_ctx);
677
678 ebml_header = start_ebml_master(pb, EBML_ID_HEADER, 0);
679 put_ebml_uint (pb, EBML_ID_EBMLVERSION , 1);
680 put_ebml_uint (pb, EBML_ID_EBMLREADVERSION , 1);
681 put_ebml_uint (pb, EBML_ID_EBMLMAXIDLENGTH , 4);
682 put_ebml_uint (pb, EBML_ID_EBMLMAXSIZELENGTH , 8);
683 put_ebml_string (pb, EBML_ID_DOCTYPE , "matroska");
684 put_ebml_uint (pb, EBML_ID_DOCTYPEVERSION , 2);
685 put_ebml_uint (pb, EBML_ID_DOCTYPEREADVERSION , 2);
686 end_ebml_master(pb, ebml_header);
687
688 mkv->segment = start_ebml_master(pb, MATROSKA_ID_SEGMENT, 0);
689 mkv->segment_offset = url_ftell(pb);
690
691 // we write 2 seek heads - one at the end of the file to point to each
692 // cluster, and one at the beginning to point to all other level one
693 // elements (including the seek head at the end of the file), which
694 // isn't more than 10 elements if we only write one of each other
695 // currently defined level 1 element
696 mkv->main_seekhead = mkv_start_seekhead(pb, mkv->segment_offset, 10);
697 mkv->cluster_seekhead = mkv_start_seekhead(pb, mkv->segment_offset, 0);
698 if (mkv->main_seekhead == NULL || mkv->cluster_seekhead == NULL)
699 return AVERROR(ENOMEM);
700
701 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_INFO, url_ftell(pb));
702 if (ret < 0) return ret;
703
704 segment_info = start_ebml_master(pb, MATROSKA_ID_INFO, 0);
705 put_ebml_uint(pb, MATROSKA_ID_TIMECODESCALE, 1000000);
706 if ((tag = av_metadata_get(s->metadata, "title", NULL, 0)))
707 put_ebml_string(pb, MATROSKA_ID_TITLE, tag->value);
708 if (!(s->streams[0]->codec->flags & CODEC_FLAG_BITEXACT)) {
709 put_ebml_string(pb, MATROSKA_ID_MUXINGAPP , LIBAVFORMAT_IDENT);
710 put_ebml_string(pb, MATROSKA_ID_WRITINGAPP, LIBAVFORMAT_IDENT);
711
712 // reserve space to write the segment UID later
713 mkv->segment_uid = url_ftell(pb);
714 put_ebml_void(pb, 19);
715 }
716
717 // reserve space for the duration
718 mkv->duration = 0;
719 mkv->duration_offset = url_ftell(pb);
720 put_ebml_void(pb, 11); // assumes double-precision float to be written
721 end_ebml_master(pb, segment_info);
722
723 ret = mkv_write_tracks(s);
724 if (ret < 0) return ret;
725
726 ret = mkv_write_chapters(s);
727 if (ret < 0) return ret;
728
729 ret = mkv_add_seekhead_entry(mkv->cluster_seekhead, MATROSKA_ID_CLUSTER, url_ftell(pb));
730 if (ret < 0) return ret;
731
732 mkv->cluster_pos = url_ftell(pb);
733 mkv->cluster = start_ebml_master(pb, MATROSKA_ID_CLUSTER, 0);
734 put_ebml_uint(pb, MATROSKA_ID_CLUSTERTIMECODE, 0);
735 mkv->cluster_pts = 0;
736
737 mkv->cues = mkv_start_cues(mkv->segment_offset);
738 if (mkv->cues == NULL)
739 return AVERROR(ENOMEM);
740
741 put_flush_packet(pb);
742 return 0;
743 }
744
745 static int mkv_blockgroup_size(int pkt_size)
746 {
747 int size = pkt_size + 4;
748 size += ebml_num_size(size);
749 size += 2; // EBML ID for block and block duration
750 size += 8; // max size of block duration
751 size += ebml_num_size(size);
752 size += 1; // blockgroup EBML ID
753 return size;
754 }
755
756 static int ass_get_duration(const uint8_t *p)
757 {
758 int sh, sm, ss, sc, eh, em, es, ec;
759 uint64_t start, end;
760
761 if (sscanf(p, "%*[^,],%d:%d:%d%*c%d,%d:%d:%d%*c%d",
762 &sh, &sm, &ss, &sc, &eh, &em, &es, &ec) != 8)
763 return 0;
764 start = 3600000*sh + 60000*sm + 1000*ss + 10*sc;
765 end = 3600000*eh + 60000*em + 1000*es + 10*ec;
766 return end - start;
767 }
768
769 static int mkv_write_ass_blocks(AVFormatContext *s, AVPacket *pkt)
770 {
771 MatroskaMuxContext *mkv = s->priv_data;
772 ByteIOContext *pb = s->pb;
773 int i, layer = 0, max_duration = 0, size, line_size, data_size = pkt->size;
774 uint8_t *start, *end, *data = pkt->data;
775 ebml_master blockgroup;
776 char buffer[2048];
777
778 while (data_size) {
779 int duration = ass_get_duration(data);
780 max_duration = FFMAX(duration, max_duration);
781 end = memchr(data, '\n', data_size);
782 size = line_size = end ? end-data+1 : data_size;
783 size -= end ? (end[-1]=='\r')+1 : 0;
784 start = data;
785 for (i=0; i<3; i++, start++)
786 if (!(start = memchr(start, ',', size-(start-data))))
787 return max_duration;
788 size -= start - data;
789 sscanf(data, "Dialogue: %d,", &layer);
790 i = snprintf(buffer, sizeof(buffer), "%"PRId64",%d,",
791 s->streams[pkt->stream_index]->nb_frames++, layer);
792 size = FFMIN(i+size, sizeof(buffer));
793 memcpy(buffer+i, start, size-i);
794
795 av_log(s, AV_LOG_DEBUG, "Writing block at offset %" PRIu64 ", size %d, "
796 "pts %" PRId64 ", duration %d\n",
797 url_ftell(pb), size, pkt->pts, duration);
798 blockgroup = start_ebml_master(pb, MATROSKA_ID_BLOCKGROUP, mkv_blockgroup_size(size));
799 put_ebml_id(pb, MATROSKA_ID_BLOCK);
800 put_ebml_num(pb, size+4, 0);
801 put_byte(pb, 0x80 | (pkt->stream_index + 1)); // this assumes stream_index is less than 126
802 put_be16(pb, pkt->pts - mkv->cluster_pts);
803 put_byte(pb, 0);
804 put_buffer(pb, buffer, size);
805 put_ebml_uint(pb, MATROSKA_ID_BLOCKDURATION, duration);
806 end_ebml_master(pb, blockgroup);
807
808 data += line_size;
809 data_size -= line_size;
810 }
811
812 return max_duration;
813 }
814
815 static void mkv_write_block(AVFormatContext *s, unsigned int blockid, AVPacket *pkt, int flags)
816 {
817 MatroskaMuxContext *mkv = s->priv_data;
818 ByteIOContext *pb = s->pb;
819 AVCodecContext *codec = s->streams[pkt->stream_index]->codec;
820 uint8_t *data = NULL;
821 int size = pkt->size;
822
823 av_log(s, AV_LOG_DEBUG, "Writing block at offset %" PRIu64 ", size %d, "
824 "pts %" PRId64 ", dts %" PRId64 ", duration %d, flags %d\n",
825 url_ftell(pb), pkt->size, pkt->pts, pkt->dts, pkt->duration, flags);
826 if (codec->codec_id == CODEC_ID_H264 && codec->extradata_size > 0 &&
827 (AV_RB24(codec->extradata) == 1 || AV_RB32(codec->extradata) == 1))
828 ff_avc_parse_nal_units_buf(pkt->data, &data, &size);
829 else
830 data = pkt->data;
831 put_ebml_id(pb, blockid);
832 put_ebml_num(pb, size+4, 0);
833 put_byte(pb, 0x80 | (pkt->stream_index + 1)); // this assumes stream_index is less than 126
834 put_be16(pb, pkt->pts - mkv->cluster_pts);
835 put_byte(pb, flags);
836 put_buffer(pb, data, size);
837 if (data != pkt->data)
838 av_free(data);
839 }
840
841 static int mkv_write_packet(AVFormatContext *s, AVPacket *pkt)
842 {
843 MatroskaMuxContext *mkv = s->priv_data;
844 ByteIOContext *pb = s->pb;
845 AVCodecContext *codec = s->streams[pkt->stream_index]->codec;
846 int keyframe = !!(pkt->flags & PKT_FLAG_KEY);
847 int duration = pkt->duration;
848 int ret;
849
850 // start a new cluster every 5 MB or 5 sec
851 if (url_ftell(pb) > mkv->cluster_pos + 5*1024*1024 || pkt->pts > mkv->cluster_pts + 5000) {
852 av_log(s, AV_LOG_DEBUG, "Starting new cluster at offset %" PRIu64
853 " bytes, pts %" PRIu64 "\n", url_ftell(pb), pkt->pts);
854 end_ebml_master(pb, mkv->cluster);
855
856 ret = mkv_add_seekhead_entry(mkv->cluster_seekhead, MATROSKA_ID_CLUSTER, url_ftell(pb));
857 if (ret < 0) return ret;
858
859 mkv->cluster_pos = url_ftell(pb);
860 mkv->cluster = start_ebml_master(pb, MATROSKA_ID_CLUSTER, 0);
861 put_ebml_uint(pb, MATROSKA_ID_CLUSTERTIMECODE, pkt->pts);
862 mkv->cluster_pts = pkt->pts;
863 av_md5_update(mkv->md5_ctx, pkt->data, FFMIN(200, pkt->size));
864 }
865
866 if (codec->codec_type != CODEC_TYPE_SUBTITLE) {
867 mkv_write_block(s, MATROSKA_ID_SIMPLEBLOCK, pkt, keyframe << 7);
868 } else if (codec->codec_id == CODEC_ID_SSA) {
869 duration = mkv_write_ass_blocks(s, pkt);
870 } else {
871 ebml_master blockgroup = start_ebml_master(pb, MATROSKA_ID_BLOCKGROUP, mkv_blockgroup_size(pkt->size));
872 duration = pkt->convergence_duration;
873 mkv_write_block(s, MATROSKA_ID_BLOCK, pkt, 0);
874 put_ebml_uint(pb, MATROSKA_ID_BLOCKDURATION, duration);
875 end_ebml_master(pb, blockgroup);
876 }
877
878 if (codec->codec_type == CODEC_TYPE_VIDEO && keyframe) {
879 ret = mkv_add_cuepoint(mkv->cues, pkt, mkv->cluster_pos);
880 if (ret < 0) return ret;
881 }
882
883 mkv->duration = FFMAX(mkv->duration, pkt->pts + duration);
884 return 0;
885 }
886
887 static int mkv_write_trailer(AVFormatContext *s)
888 {
889 MatroskaMuxContext *mkv = s->priv_data;
890 ByteIOContext *pb = s->pb;
891 int64_t currentpos, second_seekhead, cuespos;
892 int ret;
893
894 end_ebml_master(pb, mkv->cluster);
895
896 if (!url_is_streamed(pb)) {
897 cuespos = mkv_write_cues(pb, mkv->cues, s->nb_streams);
898 second_seekhead = mkv_write_seekhead(pb, mkv->cluster_seekhead);
899
900 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_CUES , cuespos);
901 if (ret < 0) return ret;
902 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_SEEKHEAD, second_seekhead);
903 if (ret < 0) return ret;
904 mkv_write_seekhead(pb, mkv->main_seekhead);
905
906 // update the duration
907 av_log(s, AV_LOG_DEBUG, "end duration = %" PRIu64 "\n", mkv->duration);
908 currentpos = url_ftell(pb);
909 url_fseek(pb, mkv->duration_offset, SEEK_SET);
910 put_ebml_float(pb, MATROSKA_ID_DURATION, mkv->duration);
911
912 // write the md5sum of some frames as the segment UID
913 if (!(s->streams[0]->codec->flags & CODEC_FLAG_BITEXACT)) {
914 uint8_t segment_uid[16];
915 av_md5_final(mkv->md5_ctx, segment_uid);
916 url_fseek(pb, mkv->segment_uid, SEEK_SET);
917 put_ebml_binary(pb, MATROSKA_ID_SEGMENTUID, segment_uid, 16);
918 }
919 url_fseek(pb, currentpos, SEEK_SET);
920 }
921
922 end_ebml_master(pb, mkv->segment);
923 av_free(mkv->md5_ctx);
924 put_flush_packet(pb);
925 return 0;
926 }
927
928 AVOutputFormat matroska_muxer = {
929 "matroska",
930 NULL_IF_CONFIG_SMALL("Matroska file format"),
931 "video/x-matroska",
932 "mkv",
933 sizeof(MatroskaMuxContext),
934 CODEC_ID_MP2,
935 CODEC_ID_MPEG4,
936 mkv_write_header,
937 mkv_write_packet,
938 mkv_write_trailer,
939 .flags = AVFMT_GLOBALHEADER | AVFMT_VARIABLE_FPS,
940 .codec_tag = (const AVCodecTag* const []){ff_codec_bmp_tags, ff_codec_wav_tags, 0},
941 .subtitle_codec = CODEC_ID_TEXT,
942 };
943
944 AVOutputFormat matroska_audio_muxer = {
945 "matroska",
946 NULL_IF_CONFIG_SMALL("Matroska file format"),
947 "audio/x-matroska",
948 "mka",
949 sizeof(MatroskaMuxContext),
950 CODEC_ID_MP2,
951 CODEC_ID_NONE,
952 mkv_write_header,
953 mkv_write_packet,
954 mkv_write_trailer,
955 .flags = AVFMT_GLOBALHEADER,
956 .codec_tag = (const AVCodecTag* const []){ff_codec_wav_tags, 0},
957 };