e78d062cbea17af6a1472f0ced1a8c10e540ebb1
[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/random_seed.h"
30 #include "libavutil/lfg.h"
31 #include "libavcodec/xiph.h"
32 #include "libavcodec/mpeg4audio.h"
33
34 typedef struct ebml_master {
35 int64_t pos; ///< absolute offset in the file where the master's elements start
36 int sizebytes; ///< how many bytes were reserved for the size
37 } ebml_master;
38
39 typedef struct mkv_seekhead_entry {
40 unsigned int elementid;
41 uint64_t segmentpos;
42 } mkv_seekhead_entry;
43
44 typedef struct mkv_seekhead {
45 int64_t filepos;
46 int64_t segment_offset; ///< the file offset to the beginning of the segment
47 int reserved_size; ///< -1 if appending to file
48 int max_entries;
49 mkv_seekhead_entry *entries;
50 int num_entries;
51 } mkv_seekhead;
52
53 typedef struct {
54 uint64_t pts;
55 int tracknum;
56 int64_t cluster_pos; ///< file offset of the cluster containing the block
57 } mkv_cuepoint;
58
59 typedef struct {
60 int64_t segment_offset;
61 mkv_cuepoint *entries;
62 int num_entries;
63 } mkv_cues;
64
65 typedef struct {
66 int write_dts;
67 } mkv_track;
68
69 #define MODE_MATROSKAv2 0x01
70 #define MODE_WEBM 0x02
71
72 typedef struct MatroskaMuxContext {
73 int mode;
74 ByteIOContext *dyn_bc;
75 ebml_master segment;
76 int64_t segment_offset;
77 ebml_master cluster;
78 int64_t cluster_pos; ///< file offset of the current cluster
79 int64_t cluster_pts;
80 int64_t duration_offset;
81 int64_t duration;
82 mkv_seekhead *main_seekhead;
83 mkv_cues *cues;
84 mkv_track *tracks;
85
86 AVLFG lfg;
87 unsigned int audio_buffer_size;
88 AVPacket cur_audio_pkt;
89 } MatroskaMuxContext;
90
91
92 /** 2 bytes * 3 for EBML IDs, 3 1-byte EBML lengths, 8 bytes for 64 bit
93 * offset, 4 bytes for target EBML ID */
94 #define MAX_SEEKENTRY_SIZE 21
95
96 /** per-cuepoint-track - 3 1-byte EBML IDs, 3 1-byte EBML sizes, 2
97 * 8-byte uint max */
98 #define MAX_CUETRACKPOS_SIZE 22
99
100 /** per-cuepoint - 2 1-byte EBML IDs, 2 1-byte EBML sizes, 8-byte uint max */
101 #define MAX_CUEPOINT_SIZE(num_tracks) 12 + MAX_CUETRACKPOS_SIZE*num_tracks
102
103
104 static int ebml_id_size(unsigned int id)
105 {
106 return (av_log2(id+1)-1)/7+1;
107 }
108
109 static void put_ebml_id(ByteIOContext *pb, unsigned int id)
110 {
111 int i = ebml_id_size(id);
112 while (i--)
113 put_byte(pb, id >> (i*8));
114 }
115
116 /**
117 * Write an EBML size meaning "unknown size".
118 *
119 * @param bytes The number of bytes the size should occupy (maximum: 8).
120 */
121 static void put_ebml_size_unknown(ByteIOContext *pb, int bytes)
122 {
123 assert(bytes <= 8);
124 put_byte(pb, 0x1ff >> bytes);
125 while (--bytes)
126 put_byte(pb, 0xff);
127 }
128
129 /**
130 * Calculate how many bytes are needed to represent a given number in EBML.
131 */
132 static int ebml_num_size(uint64_t num)
133 {
134 int bytes = 1;
135 while ((num+1) >> bytes*7) bytes++;
136 return bytes;
137 }
138
139 /**
140 * Write a number in EBML variable length format.
141 *
142 * @param bytes The number of bytes that need to be used to write the number.
143 * If zero, any number of bytes can be used.
144 */
145 static void put_ebml_num(ByteIOContext *pb, uint64_t num, int bytes)
146 {
147 int i, needed_bytes = ebml_num_size(num);
148
149 // sizes larger than this are currently undefined in EBML
150 assert(num < (1ULL<<56)-1);
151
152 if (bytes == 0)
153 // don't care how many bytes are used, so use the min
154 bytes = needed_bytes;
155 // the bytes needed to write the given size would exceed the bytes
156 // that we need to use, so write unknown size. This shouldn't happen.
157 assert(bytes >= needed_bytes);
158
159 num |= 1ULL << bytes*7;
160 for (i = bytes - 1; i >= 0; i--)
161 put_byte(pb, num >> i*8);
162 }
163
164 static void put_ebml_uint(ByteIOContext *pb, unsigned int elementid, uint64_t val)
165 {
166 int i, bytes = 1;
167 uint64_t tmp = val;
168 while (tmp>>=8) bytes++;
169
170 put_ebml_id(pb, elementid);
171 put_ebml_num(pb, bytes, 0);
172 for (i = bytes - 1; i >= 0; i--)
173 put_byte(pb, val >> i*8);
174 }
175
176 static void put_ebml_float(ByteIOContext *pb, unsigned int elementid, double val)
177 {
178 put_ebml_id(pb, elementid);
179 put_ebml_num(pb, 8, 0);
180 put_be64(pb, av_dbl2int(val));
181 }
182
183 static void put_ebml_binary(ByteIOContext *pb, unsigned int elementid,
184 const void *buf, int size)
185 {
186 put_ebml_id(pb, elementid);
187 put_ebml_num(pb, size, 0);
188 put_buffer(pb, buf, size);
189 }
190
191 static void put_ebml_string(ByteIOContext *pb, unsigned int elementid, const char *str)
192 {
193 put_ebml_binary(pb, elementid, str, strlen(str));
194 }
195
196 /**
197 * Writes a void element of a given size. Useful for reserving space in
198 * the file to be written to later.
199 *
200 * @param size The number of bytes to reserve, which must be at least 2.
201 */
202 static void put_ebml_void(ByteIOContext *pb, uint64_t size)
203 {
204 int64_t currentpos = url_ftell(pb);
205
206 assert(size >= 2);
207
208 put_ebml_id(pb, EBML_ID_VOID);
209 // we need to subtract the length needed to store the size from the
210 // size we need to reserve so 2 cases, we use 8 bytes to store the
211 // size if possible, 1 byte otherwise
212 if (size < 10)
213 put_ebml_num(pb, size-1, 0);
214 else
215 put_ebml_num(pb, size-9, 8);
216 while(url_ftell(pb) < currentpos + size)
217 put_byte(pb, 0);
218 }
219
220 static ebml_master start_ebml_master(ByteIOContext *pb, unsigned int elementid, uint64_t expectedsize)
221 {
222 int bytes = expectedsize ? ebml_num_size(expectedsize) : 8;
223 put_ebml_id(pb, elementid);
224 put_ebml_size_unknown(pb, bytes);
225 return (ebml_master){ url_ftell(pb), bytes };
226 }
227
228 static void end_ebml_master(ByteIOContext *pb, ebml_master master)
229 {
230 int64_t pos = url_ftell(pb);
231
232 if (url_fseek(pb, master.pos - master.sizebytes, SEEK_SET) < 0)
233 return;
234 put_ebml_num(pb, pos - master.pos, master.sizebytes);
235 url_fseek(pb, pos, SEEK_SET);
236 }
237
238 static void put_xiph_size(ByteIOContext *pb, int size)
239 {
240 int i;
241 for (i = 0; i < size / 255; i++)
242 put_byte(pb, 255);
243 put_byte(pb, size % 255);
244 }
245
246 /**
247 * Initialize a mkv_seekhead element to be ready to index level 1 Matroska
248 * elements. If a maximum number of elements is specified, enough space
249 * will be reserved at the current file location to write a seek head of
250 * that size.
251 *
252 * @param segment_offset The absolute offset to the position in the file
253 * where the segment begins.
254 * @param numelements The maximum number of elements that will be indexed
255 * by this seek head, 0 if unlimited.
256 */
257 static mkv_seekhead * mkv_start_seekhead(ByteIOContext *pb, int64_t segment_offset, int numelements)
258 {
259 mkv_seekhead *new_seekhead = av_mallocz(sizeof(mkv_seekhead));
260 if (new_seekhead == NULL)
261 return NULL;
262
263 new_seekhead->segment_offset = segment_offset;
264
265 if (numelements > 0) {
266 new_seekhead->filepos = url_ftell(pb);
267 // 21 bytes max for a seek entry, 10 bytes max for the SeekHead ID
268 // and size, and 3 bytes to guarantee that an EBML void element
269 // will fit afterwards
270 new_seekhead->reserved_size = numelements * MAX_SEEKENTRY_SIZE + 13;
271 new_seekhead->max_entries = numelements;
272 put_ebml_void(pb, new_seekhead->reserved_size);
273 }
274 return new_seekhead;
275 }
276
277 static int mkv_add_seekhead_entry(mkv_seekhead *seekhead, unsigned int elementid, uint64_t filepos)
278 {
279 mkv_seekhead_entry *entries = seekhead->entries;
280
281 // don't store more elements than we reserved space for
282 if (seekhead->max_entries > 0 && seekhead->max_entries <= seekhead->num_entries)
283 return -1;
284
285 entries = av_realloc(entries, (seekhead->num_entries + 1) * sizeof(mkv_seekhead_entry));
286 if (entries == NULL)
287 return AVERROR(ENOMEM);
288
289 entries[seekhead->num_entries ].elementid = elementid;
290 entries[seekhead->num_entries++].segmentpos = filepos - seekhead->segment_offset;
291
292 seekhead->entries = entries;
293 return 0;
294 }
295
296 /**
297 * Write the seek head to the file and free it. If a maximum number of
298 * elements was specified to mkv_start_seekhead(), the seek head will
299 * be written at the location reserved for it. Otherwise, it is written
300 * at the current location in the file.
301 *
302 * @return The file offset where the seekhead was written,
303 * -1 if an error occurred.
304 */
305 static int64_t mkv_write_seekhead(ByteIOContext *pb, mkv_seekhead *seekhead)
306 {
307 ebml_master metaseek, seekentry;
308 int64_t currentpos;
309 int i;
310
311 currentpos = url_ftell(pb);
312
313 if (seekhead->reserved_size > 0)
314 if (url_fseek(pb, seekhead->filepos, SEEK_SET) < 0)
315 return -1;
316
317 metaseek = start_ebml_master(pb, MATROSKA_ID_SEEKHEAD, seekhead->reserved_size);
318 for (i = 0; i < seekhead->num_entries; i++) {
319 mkv_seekhead_entry *entry = &seekhead->entries[i];
320
321 seekentry = start_ebml_master(pb, MATROSKA_ID_SEEKENTRY, MAX_SEEKENTRY_SIZE);
322
323 put_ebml_id(pb, MATROSKA_ID_SEEKID);
324 put_ebml_num(pb, ebml_id_size(entry->elementid), 0);
325 put_ebml_id(pb, entry->elementid);
326
327 put_ebml_uint(pb, MATROSKA_ID_SEEKPOSITION, entry->segmentpos);
328 end_ebml_master(pb, seekentry);
329 }
330 end_ebml_master(pb, metaseek);
331
332 if (seekhead->reserved_size > 0) {
333 uint64_t remaining = seekhead->filepos + seekhead->reserved_size - url_ftell(pb);
334 put_ebml_void(pb, remaining);
335 url_fseek(pb, currentpos, SEEK_SET);
336
337 currentpos = seekhead->filepos;
338 }
339 av_free(seekhead->entries);
340 av_free(seekhead);
341
342 return currentpos;
343 }
344
345 static mkv_cues * mkv_start_cues(int64_t segment_offset)
346 {
347 mkv_cues *cues = av_mallocz(sizeof(mkv_cues));
348 if (cues == NULL)
349 return NULL;
350
351 cues->segment_offset = segment_offset;
352 return cues;
353 }
354
355 static int mkv_add_cuepoint(mkv_cues *cues, int stream, int64_t ts, int64_t cluster_pos)
356 {
357 mkv_cuepoint *entries = cues->entries;
358
359 entries = av_realloc(entries, (cues->num_entries + 1) * sizeof(mkv_cuepoint));
360 if (entries == NULL)
361 return AVERROR(ENOMEM);
362
363 if (ts < 0)
364 return 0;
365
366 entries[cues->num_entries ].pts = ts;
367 entries[cues->num_entries ].tracknum = stream + 1;
368 entries[cues->num_entries++].cluster_pos = cluster_pos - cues->segment_offset;
369
370 cues->entries = entries;
371 return 0;
372 }
373
374 static int64_t mkv_write_cues(ByteIOContext *pb, mkv_cues *cues, int num_tracks)
375 {
376 ebml_master cues_element;
377 int64_t currentpos;
378 int i, j;
379
380 currentpos = url_ftell(pb);
381 cues_element = start_ebml_master(pb, MATROSKA_ID_CUES, 0);
382
383 for (i = 0; i < cues->num_entries; i++) {
384 ebml_master cuepoint, track_positions;
385 mkv_cuepoint *entry = &cues->entries[i];
386 uint64_t pts = entry->pts;
387
388 cuepoint = start_ebml_master(pb, MATROSKA_ID_POINTENTRY, MAX_CUEPOINT_SIZE(num_tracks));
389 put_ebml_uint(pb, MATROSKA_ID_CUETIME, pts);
390
391 // put all the entries from different tracks that have the exact same
392 // timestamp into the same CuePoint
393 for (j = 0; j < cues->num_entries - i && entry[j].pts == pts; j++) {
394 track_positions = start_ebml_master(pb, MATROSKA_ID_CUETRACKPOSITION, MAX_CUETRACKPOS_SIZE);
395 put_ebml_uint(pb, MATROSKA_ID_CUETRACK , entry[j].tracknum );
396 put_ebml_uint(pb, MATROSKA_ID_CUECLUSTERPOSITION, entry[j].cluster_pos);
397 end_ebml_master(pb, track_positions);
398 }
399 i += j - 1;
400 end_ebml_master(pb, cuepoint);
401 }
402 end_ebml_master(pb, cues_element);
403
404 av_free(cues->entries);
405 av_free(cues);
406 return currentpos;
407 }
408
409 static int put_xiph_codecpriv(AVFormatContext *s, ByteIOContext *pb, AVCodecContext *codec)
410 {
411 uint8_t *header_start[3];
412 int header_len[3];
413 int first_header_size;
414 int j;
415
416 if (codec->codec_id == CODEC_ID_VORBIS)
417 first_header_size = 30;
418 else
419 first_header_size = 42;
420
421 if (ff_split_xiph_headers(codec->extradata, codec->extradata_size,
422 first_header_size, header_start, header_len) < 0) {
423 av_log(s, AV_LOG_ERROR, "Extradata corrupt.\n");
424 return -1;
425 }
426
427 put_byte(pb, 2); // number packets - 1
428 for (j = 0; j < 2; j++) {
429 put_xiph_size(pb, header_len[j]);
430 }
431 for (j = 0; j < 3; j++)
432 put_buffer(pb, header_start[j], header_len[j]);
433
434 return 0;
435 }
436
437 static void get_aac_sample_rates(AVFormatContext *s, AVCodecContext *codec, int *sample_rate, int *output_sample_rate)
438 {
439 int sri;
440
441 if (codec->extradata_size < 2) {
442 av_log(s, AV_LOG_WARNING, "No AAC extradata, unable to determine samplerate.\n");
443 return;
444 }
445
446 sri = ((codec->extradata[0] << 1) & 0xE) | (codec->extradata[1] >> 7);
447 if (sri > 12) {
448 av_log(s, AV_LOG_WARNING, "AAC samplerate index out of bounds\n");
449 return;
450 }
451 *sample_rate = ff_mpeg4audio_sample_rates[sri];
452
453 // if sbr, get output sample rate as well
454 if (codec->extradata_size == 5) {
455 sri = (codec->extradata[4] >> 3) & 0xF;
456 if (sri > 12) {
457 av_log(s, AV_LOG_WARNING, "AAC output samplerate index out of bounds\n");
458 return;
459 }
460 *output_sample_rate = ff_mpeg4audio_sample_rates[sri];
461 }
462 }
463
464 static int mkv_write_codecprivate(AVFormatContext *s, ByteIOContext *pb, AVCodecContext *codec, int native_id, int qt_id)
465 {
466 ByteIOContext *dyn_cp;
467 uint8_t *codecpriv;
468 int ret, codecpriv_size;
469
470 ret = url_open_dyn_buf(&dyn_cp);
471 if(ret < 0)
472 return ret;
473
474 if (native_id) {
475 if (codec->codec_id == CODEC_ID_VORBIS || codec->codec_id == CODEC_ID_THEORA)
476 ret = put_xiph_codecpriv(s, dyn_cp, codec);
477 else if (codec->codec_id == CODEC_ID_FLAC)
478 ret = ff_flac_write_header(dyn_cp, codec, 1);
479 else if (codec->codec_id == CODEC_ID_H264)
480 ret = ff_isom_write_avcc(dyn_cp, codec->extradata, codec->extradata_size);
481 else if (codec->extradata_size)
482 put_buffer(dyn_cp, codec->extradata, codec->extradata_size);
483 } else if (codec->codec_type == AVMEDIA_TYPE_VIDEO) {
484 if (qt_id) {
485 if (!codec->codec_tag)
486 codec->codec_tag = ff_codec_get_tag(codec_movvideo_tags, codec->codec_id);
487 if (codec->extradata_size)
488 put_buffer(dyn_cp, codec->extradata, codec->extradata_size);
489 } else {
490 if (!codec->codec_tag)
491 codec->codec_tag = ff_codec_get_tag(ff_codec_bmp_tags, codec->codec_id);
492 if (!codec->codec_tag) {
493 av_log(s, AV_LOG_ERROR, "No bmp codec ID found.\n");
494 ret = -1;
495 }
496
497 ff_put_bmp_header(dyn_cp, codec, ff_codec_bmp_tags, 0);
498 }
499
500 } else if (codec->codec_type == AVMEDIA_TYPE_AUDIO) {
501 unsigned int tag;
502 tag = ff_codec_get_tag(ff_codec_wav_tags, codec->codec_id);
503 if (!tag) {
504 av_log(s, AV_LOG_ERROR, "No wav codec ID found.\n");
505 ret = -1;
506 }
507 if (!codec->codec_tag)
508 codec->codec_tag = tag;
509
510 ff_put_wav_header(dyn_cp, codec);
511 }
512
513 codecpriv_size = url_close_dyn_buf(dyn_cp, &codecpriv);
514 if (codecpriv_size)
515 put_ebml_binary(pb, MATROSKA_ID_CODECPRIVATE, codecpriv, codecpriv_size);
516 av_free(codecpriv);
517 return ret;
518 }
519
520 static int mkv_write_tracks(AVFormatContext *s)
521 {
522 MatroskaMuxContext *mkv = s->priv_data;
523 ByteIOContext *pb = s->pb;
524 ebml_master tracks;
525 int i, j, ret;
526
527 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_TRACKS, url_ftell(pb));
528 if (ret < 0) return ret;
529
530 tracks = start_ebml_master(pb, MATROSKA_ID_TRACKS, 0);
531 for (i = 0; i < s->nb_streams; i++) {
532 AVStream *st = s->streams[i];
533 AVCodecContext *codec = st->codec;
534 ebml_master subinfo, track;
535 int native_id = 0;
536 int qt_id = 0;
537 int bit_depth = av_get_bits_per_sample(codec->codec_id);
538 int sample_rate = codec->sample_rate;
539 int output_sample_rate = 0;
540 AVMetadataTag *tag;
541
542 if (!bit_depth)
543 bit_depth = av_get_bits_per_sample_format(codec->sample_fmt);
544
545 if (codec->codec_id == CODEC_ID_AAC)
546 get_aac_sample_rates(s, codec, &sample_rate, &output_sample_rate);
547
548 track = start_ebml_master(pb, MATROSKA_ID_TRACKENTRY, 0);
549 put_ebml_uint (pb, MATROSKA_ID_TRACKNUMBER , i + 1);
550 put_ebml_uint (pb, MATROSKA_ID_TRACKUID , i + 1);
551 put_ebml_uint (pb, MATROSKA_ID_TRACKFLAGLACING , 0); // no lacing (yet)
552
553 if ((tag = av_metadata_get(st->metadata, "title", NULL, 0)))
554 put_ebml_string(pb, MATROSKA_ID_TRACKNAME, tag->value);
555 tag = av_metadata_get(st->metadata, "language", NULL, 0);
556 put_ebml_string(pb, MATROSKA_ID_TRACKLANGUAGE, tag ? tag->value:"und");
557
558 if (st->disposition)
559 put_ebml_uint(pb, MATROSKA_ID_TRACKFLAGDEFAULT, !!(st->disposition & AV_DISPOSITION_DEFAULT));
560
561 // look for a codec ID string specific to mkv to use,
562 // if none are found, use AVI codes
563 for (j = 0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++) {
564 if (ff_mkv_codec_tags[j].id == codec->codec_id) {
565 put_ebml_string(pb, MATROSKA_ID_CODECID, ff_mkv_codec_tags[j].str);
566 native_id = 1;
567 break;
568 }
569 }
570
571 if (mkv->mode == MODE_WEBM && !(codec->codec_id == CODEC_ID_VP8 ||
572 codec->codec_id == CODEC_ID_VORBIS)) {
573 av_log(s, AV_LOG_ERROR,
574 "Only VP8 video and Vorbis audio are supported for WebM.\n");
575 return AVERROR(EINVAL);
576 }
577
578 switch (codec->codec_type) {
579 case AVMEDIA_TYPE_VIDEO:
580 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_VIDEO);
581 put_ebml_uint(pb, MATROSKA_ID_TRACKDEFAULTDURATION, av_q2d(codec->time_base)*1E9);
582
583 if (!native_id &&
584 ff_codec_get_tag(codec_movvideo_tags, codec->codec_id) &&
585 (!ff_codec_get_tag(ff_codec_bmp_tags, codec->codec_id)
586 || codec->codec_id == CODEC_ID_SVQ1
587 || codec->codec_id == CODEC_ID_SVQ3
588 || codec->codec_id == CODEC_ID_CINEPAK))
589 qt_id = 1;
590
591 if (qt_id)
592 put_ebml_string(pb, MATROSKA_ID_CODECID, "V_QUICKTIME");
593 else if (!native_id) {
594 // if there is no mkv-specific codec ID, use VFW mode
595 put_ebml_string(pb, MATROSKA_ID_CODECID, "V_MS/VFW/FOURCC");
596 mkv->tracks[i].write_dts = 1;
597 }
598
599 subinfo = start_ebml_master(pb, MATROSKA_ID_TRACKVIDEO, 0);
600 // XXX: interlace flag?
601 put_ebml_uint (pb, MATROSKA_ID_VIDEOPIXELWIDTH , codec->width);
602 put_ebml_uint (pb, MATROSKA_ID_VIDEOPIXELHEIGHT, codec->height);
603 if (st->sample_aspect_ratio.num) {
604 int d_width = codec->width*av_q2d(st->sample_aspect_ratio);
605 put_ebml_uint(pb, MATROSKA_ID_VIDEODISPLAYWIDTH , d_width);
606 put_ebml_uint(pb, MATROSKA_ID_VIDEODISPLAYHEIGHT, codec->height);
607 }
608 end_ebml_master(pb, subinfo);
609 break;
610
611 case AVMEDIA_TYPE_AUDIO:
612 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_AUDIO);
613
614 if (!native_id)
615 // no mkv-specific ID, use ACM mode
616 put_ebml_string(pb, MATROSKA_ID_CODECID, "A_MS/ACM");
617
618 subinfo = start_ebml_master(pb, MATROSKA_ID_TRACKAUDIO, 0);
619 put_ebml_uint (pb, MATROSKA_ID_AUDIOCHANNELS , codec->channels);
620 put_ebml_float (pb, MATROSKA_ID_AUDIOSAMPLINGFREQ, sample_rate);
621 if (output_sample_rate)
622 put_ebml_float(pb, MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, output_sample_rate);
623 if (bit_depth)
624 put_ebml_uint(pb, MATROSKA_ID_AUDIOBITDEPTH, bit_depth);
625 end_ebml_master(pb, subinfo);
626 break;
627
628 case AVMEDIA_TYPE_SUBTITLE:
629 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_SUBTITLE);
630 break;
631 default:
632 av_log(s, AV_LOG_ERROR, "Only audio, video, and subtitles are supported for Matroska.");
633 break;
634 }
635 ret = mkv_write_codecprivate(s, pb, codec, native_id, qt_id);
636 if (ret < 0) return ret;
637
638 end_ebml_master(pb, track);
639
640 // ms precision is the de-facto standard timescale for mkv files
641 av_set_pts_info(st, 64, 1, 1000);
642 }
643 end_ebml_master(pb, tracks);
644 return 0;
645 }
646
647 static int mkv_write_chapters(AVFormatContext *s)
648 {
649 MatroskaMuxContext *mkv = s->priv_data;
650 ByteIOContext *pb = s->pb;
651 ebml_master chapters, editionentry;
652 AVRational scale = {1, 1E9};
653 int i, ret;
654
655 if (!s->nb_chapters)
656 return 0;
657
658 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_CHAPTERS, url_ftell(pb));
659 if (ret < 0) return ret;
660
661 chapters = start_ebml_master(pb, MATROSKA_ID_CHAPTERS , 0);
662 editionentry = start_ebml_master(pb, MATROSKA_ID_EDITIONENTRY, 0);
663 put_ebml_uint(pb, MATROSKA_ID_EDITIONFLAGDEFAULT, 1);
664 put_ebml_uint(pb, MATROSKA_ID_EDITIONFLAGHIDDEN , 0);
665 for (i = 0; i < s->nb_chapters; i++) {
666 ebml_master chapteratom, chapterdisplay;
667 AVChapter *c = s->chapters[i];
668 AVMetadataTag *t = NULL;
669
670 chapteratom = start_ebml_master(pb, MATROSKA_ID_CHAPTERATOM, 0);
671 put_ebml_uint(pb, MATROSKA_ID_CHAPTERUID, c->id);
672 put_ebml_uint(pb, MATROSKA_ID_CHAPTERTIMESTART,
673 av_rescale_q(c->start, c->time_base, scale));
674 put_ebml_uint(pb, MATROSKA_ID_CHAPTERTIMEEND,
675 av_rescale_q(c->end, c->time_base, scale));
676 put_ebml_uint(pb, MATROSKA_ID_CHAPTERFLAGHIDDEN , 0);
677 put_ebml_uint(pb, MATROSKA_ID_CHAPTERFLAGENABLED, 1);
678 if ((t = av_metadata_get(c->metadata, "title", NULL, 0))) {
679 chapterdisplay = start_ebml_master(pb, MATROSKA_ID_CHAPTERDISPLAY, 0);
680 put_ebml_string(pb, MATROSKA_ID_CHAPSTRING, t->value);
681 put_ebml_string(pb, MATROSKA_ID_CHAPLANG , "und");
682 end_ebml_master(pb, chapterdisplay);
683 }
684 end_ebml_master(pb, chapteratom);
685 }
686 end_ebml_master(pb, editionentry);
687 end_ebml_master(pb, chapters);
688 return 0;
689 }
690
691 static int mkv_write_header(AVFormatContext *s)
692 {
693 MatroskaMuxContext *mkv = s->priv_data;
694 ByteIOContext *pb = s->pb;
695 ebml_master ebml_header, segment_info;
696 AVMetadataTag *tag;
697 int ret, i;
698
699 if (!strcmp(s->oformat->name, "webm")) mkv->mode = MODE_WEBM;
700 else mkv->mode = MODE_MATROSKAv2;
701
702 av_lfg_init(&mkv->lfg, av_get_random_seed());
703
704 mkv->tracks = av_mallocz(s->nb_streams * sizeof(*mkv->tracks));
705 if (!mkv->tracks)
706 return AVERROR(ENOMEM);
707
708 ebml_header = start_ebml_master(pb, EBML_ID_HEADER, 0);
709 put_ebml_uint (pb, EBML_ID_EBMLVERSION , 1);
710 put_ebml_uint (pb, EBML_ID_EBMLREADVERSION , 1);
711 put_ebml_uint (pb, EBML_ID_EBMLMAXIDLENGTH , 4);
712 put_ebml_uint (pb, EBML_ID_EBMLMAXSIZELENGTH , 8);
713 put_ebml_string (pb, EBML_ID_DOCTYPE , s->oformat->name);
714 put_ebml_uint (pb, EBML_ID_DOCTYPEVERSION , 2);
715 put_ebml_uint (pb, EBML_ID_DOCTYPEREADVERSION , 2);
716 end_ebml_master(pb, ebml_header);
717
718 mkv->segment = start_ebml_master(pb, MATROSKA_ID_SEGMENT, 0);
719 mkv->segment_offset = url_ftell(pb);
720
721 // we write 2 seek heads - one at the end of the file to point to each
722 // cluster, and one at the beginning to point to all other level one
723 // elements (including the seek head at the end of the file), which
724 // isn't more than 10 elements if we only write one of each other
725 // currently defined level 1 element
726 mkv->main_seekhead = mkv_start_seekhead(pb, mkv->segment_offset, 10);
727 if (!mkv->main_seekhead)
728 return AVERROR(ENOMEM);
729
730 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_INFO, url_ftell(pb));
731 if (ret < 0) return ret;
732
733 segment_info = start_ebml_master(pb, MATROSKA_ID_INFO, 0);
734 put_ebml_uint(pb, MATROSKA_ID_TIMECODESCALE, 1000000);
735 if ((tag = av_metadata_get(s->metadata, "title", NULL, 0)))
736 put_ebml_string(pb, MATROSKA_ID_TITLE, tag->value);
737 if (!(s->streams[0]->codec->flags & CODEC_FLAG_BITEXACT)) {
738 uint32_t segment_uid[4];
739 for (i = 0; i < 4; i++)
740 segment_uid[i] = av_lfg_get(&mkv->lfg);
741
742 put_ebml_string(pb, MATROSKA_ID_MUXINGAPP , LIBAVFORMAT_IDENT);
743 put_ebml_string(pb, MATROSKA_ID_WRITINGAPP, LIBAVFORMAT_IDENT);
744 put_ebml_binary(pb, MATROSKA_ID_SEGMENTUID, segment_uid, 16);
745 }
746
747 // reserve space for the duration
748 mkv->duration = 0;
749 mkv->duration_offset = url_ftell(pb);
750 put_ebml_void(pb, 11); // assumes double-precision float to be written
751 end_ebml_master(pb, segment_info);
752
753 ret = mkv_write_tracks(s);
754 if (ret < 0) return ret;
755
756 if (mkv->mode != MODE_WEBM) {
757 ret = mkv_write_chapters(s);
758 if (ret < 0) return ret;
759 }
760
761 if (url_is_streamed(s->pb))
762 mkv_write_seekhead(pb, mkv->main_seekhead);
763
764 mkv->cues = mkv_start_cues(mkv->segment_offset);
765 if (mkv->cues == NULL)
766 return AVERROR(ENOMEM);
767
768 av_init_packet(&mkv->cur_audio_pkt);
769 mkv->cur_audio_pkt.size = 0;
770 mkv->audio_buffer_size = 0;
771
772 put_flush_packet(pb);
773 return 0;
774 }
775
776 static int mkv_blockgroup_size(int pkt_size)
777 {
778 int size = pkt_size + 4;
779 size += ebml_num_size(size);
780 size += 2; // EBML ID for block and block duration
781 size += 8; // max size of block duration
782 size += ebml_num_size(size);
783 size += 1; // blockgroup EBML ID
784 return size;
785 }
786
787 static int ass_get_duration(const uint8_t *p)
788 {
789 int sh, sm, ss, sc, eh, em, es, ec;
790 uint64_t start, end;
791
792 if (sscanf(p, "%*[^,],%d:%d:%d%*c%d,%d:%d:%d%*c%d",
793 &sh, &sm, &ss, &sc, &eh, &em, &es, &ec) != 8)
794 return 0;
795 start = 3600000*sh + 60000*sm + 1000*ss + 10*sc;
796 end = 3600000*eh + 60000*em + 1000*es + 10*ec;
797 return end - start;
798 }
799
800 static int mkv_write_ass_blocks(AVFormatContext *s, ByteIOContext *pb, AVPacket *pkt)
801 {
802 MatroskaMuxContext *mkv = s->priv_data;
803 int i, layer = 0, max_duration = 0, size, line_size, data_size = pkt->size;
804 uint8_t *start, *end, *data = pkt->data;
805 ebml_master blockgroup;
806 char buffer[2048];
807
808 while (data_size) {
809 int duration = ass_get_duration(data);
810 max_duration = FFMAX(duration, max_duration);
811 end = memchr(data, '\n', data_size);
812 size = line_size = end ? end-data+1 : data_size;
813 size -= end ? (end[-1]=='\r')+1 : 0;
814 start = data;
815 for (i=0; i<3; i++, start++)
816 if (!(start = memchr(start, ',', size-(start-data))))
817 return max_duration;
818 size -= start - data;
819 sscanf(data, "Dialogue: %d,", &layer);
820 i = snprintf(buffer, sizeof(buffer), "%"PRId64",%d,",
821 s->streams[pkt->stream_index]->nb_frames++, layer);
822 size = FFMIN(i+size, sizeof(buffer));
823 memcpy(buffer+i, start, size-i);
824
825 av_log(s, AV_LOG_DEBUG, "Writing block at offset %" PRIu64 ", size %d, "
826 "pts %" PRId64 ", duration %d\n",
827 url_ftell(pb), size, pkt->pts, duration);
828 blockgroup = start_ebml_master(pb, MATROSKA_ID_BLOCKGROUP, mkv_blockgroup_size(size));
829 put_ebml_id(pb, MATROSKA_ID_BLOCK);
830 put_ebml_num(pb, size+4, 0);
831 put_byte(pb, 0x80 | (pkt->stream_index + 1)); // this assumes stream_index is less than 126
832 put_be16(pb, pkt->pts - mkv->cluster_pts);
833 put_byte(pb, 0);
834 put_buffer(pb, buffer, size);
835 put_ebml_uint(pb, MATROSKA_ID_BLOCKDURATION, duration);
836 end_ebml_master(pb, blockgroup);
837
838 data += line_size;
839 data_size -= line_size;
840 }
841
842 return max_duration;
843 }
844
845 static void mkv_write_block(AVFormatContext *s, ByteIOContext *pb,
846 unsigned int blockid, AVPacket *pkt, int flags)
847 {
848 MatroskaMuxContext *mkv = s->priv_data;
849 AVCodecContext *codec = s->streams[pkt->stream_index]->codec;
850 uint8_t *data = NULL;
851 int size = pkt->size;
852 int64_t ts = mkv->tracks[pkt->stream_index].write_dts ? pkt->dts : pkt->pts;
853
854 av_log(s, AV_LOG_DEBUG, "Writing block at offset %" PRIu64 ", size %d, "
855 "pts %" PRId64 ", dts %" PRId64 ", duration %d, flags %d\n",
856 url_ftell(pb), pkt->size, pkt->pts, pkt->dts, pkt->duration, flags);
857 if (codec->codec_id == CODEC_ID_H264 && codec->extradata_size > 0 &&
858 (AV_RB24(codec->extradata) == 1 || AV_RB32(codec->extradata) == 1))
859 ff_avc_parse_nal_units_buf(pkt->data, &data, &size);
860 else
861 data = pkt->data;
862 put_ebml_id(pb, blockid);
863 put_ebml_num(pb, size+4, 0);
864 put_byte(pb, 0x80 | (pkt->stream_index + 1)); // this assumes stream_index is less than 126
865 put_be16(pb, ts - mkv->cluster_pts);
866 put_byte(pb, flags);
867 put_buffer(pb, data, size);
868 if (data != pkt->data)
869 av_free(data);
870 }
871
872 static void mkv_flush_dynbuf(AVFormatContext *s)
873 {
874 MatroskaMuxContext *mkv = s->priv_data;
875 int bufsize;
876 uint8_t *dyn_buf;
877
878 if (!mkv->dyn_bc)
879 return;
880
881 bufsize = url_close_dyn_buf(mkv->dyn_bc, &dyn_buf);
882 put_buffer(s->pb, dyn_buf, bufsize);
883 av_free(dyn_buf);
884 mkv->dyn_bc = NULL;
885 }
886
887 static int mkv_write_packet_internal(AVFormatContext *s, AVPacket *pkt)
888 {
889 MatroskaMuxContext *mkv = s->priv_data;
890 ByteIOContext *pb = s->pb;
891 AVCodecContext *codec = s->streams[pkt->stream_index]->codec;
892 int keyframe = !!(pkt->flags & AV_PKT_FLAG_KEY);
893 int duration = pkt->duration;
894 int ret;
895 int64_t ts = mkv->tracks[pkt->stream_index].write_dts ? pkt->dts : pkt->pts;
896
897 if (ts == AV_NOPTS_VALUE) {
898 av_log(s, AV_LOG_ERROR, "Can't write packet with unknown timestamp\n");
899 return AVERROR(EINVAL);
900 }
901
902 if (url_is_streamed(s->pb)) {
903 if (!mkv->dyn_bc)
904 url_open_dyn_buf(&mkv->dyn_bc);
905 pb = mkv->dyn_bc;
906 }
907
908 if (!mkv->cluster_pos) {
909 mkv->cluster_pos = url_ftell(s->pb);
910 mkv->cluster = start_ebml_master(pb, MATROSKA_ID_CLUSTER, 0);
911 put_ebml_uint(pb, MATROSKA_ID_CLUSTERTIMECODE, FFMAX(0, ts));
912 mkv->cluster_pts = FFMAX(0, ts);
913 }
914
915 if (codec->codec_type != AVMEDIA_TYPE_SUBTITLE) {
916 mkv_write_block(s, pb, MATROSKA_ID_SIMPLEBLOCK, pkt, keyframe << 7);
917 } else if (codec->codec_id == CODEC_ID_SSA) {
918 duration = mkv_write_ass_blocks(s, pb, pkt);
919 } else {
920 ebml_master blockgroup = start_ebml_master(pb, MATROSKA_ID_BLOCKGROUP, mkv_blockgroup_size(pkt->size));
921 duration = pkt->convergence_duration;
922 mkv_write_block(s, pb, MATROSKA_ID_BLOCK, pkt, 0);
923 put_ebml_uint(pb, MATROSKA_ID_BLOCKDURATION, duration);
924 end_ebml_master(pb, blockgroup);
925 }
926
927 if (codec->codec_type == AVMEDIA_TYPE_VIDEO && keyframe) {
928 ret = mkv_add_cuepoint(mkv->cues, pkt->stream_index, ts, mkv->cluster_pos);
929 if (ret < 0) return ret;
930 }
931
932 mkv->duration = FFMAX(mkv->duration, ts + duration);
933 return 0;
934 }
935
936 static int mkv_copy_packet(MatroskaMuxContext *mkv, const AVPacket *pkt)
937 {
938 uint8_t *data = mkv->cur_audio_pkt.data;
939 mkv->cur_audio_pkt = *pkt;
940 mkv->cur_audio_pkt.data = av_fast_realloc(data, &mkv->audio_buffer_size, pkt->size);
941 if (!mkv->cur_audio_pkt.data)
942 return AVERROR(ENOMEM);
943
944 memcpy(mkv->cur_audio_pkt.data, pkt->data, pkt->size);
945 mkv->cur_audio_pkt.size = pkt->size;
946 return 0;
947 }
948
949 static int mkv_write_packet(AVFormatContext *s, AVPacket *pkt)
950 {
951 MatroskaMuxContext *mkv = s->priv_data;
952 ByteIOContext *pb = url_is_streamed(s->pb) ? mkv->dyn_bc : s->pb;
953 AVCodecContext *codec = s->streams[pkt->stream_index]->codec;
954 int ret, keyframe = !!(pkt->flags & AV_PKT_FLAG_KEY);
955 int64_t ts = mkv->tracks[pkt->stream_index].write_dts ? pkt->dts : pkt->pts;
956 int cluster_size = url_ftell(pb) - (url_is_streamed(s->pb) ? 0 : mkv->cluster_pos);
957
958 // start a new cluster every 5 MB or 5 sec, or 32k / 1 sec for streaming or
959 // after 4k and on a keyframe
960 if (mkv->cluster_pos &&
961 ((url_is_streamed(s->pb) && (cluster_size > 32*1024 || ts > mkv->cluster_pts + 1000))
962 || cluster_size > 5*1024*1024 || ts > mkv->cluster_pts + 5000
963 || (codec->codec_type == AVMEDIA_TYPE_VIDEO && keyframe && cluster_size > 4*1024))) {
964 av_log(s, AV_LOG_DEBUG, "Starting new cluster at offset %" PRIu64
965 " bytes, pts %" PRIu64 "\n", url_ftell(pb), ts);
966 end_ebml_master(pb, mkv->cluster);
967 mkv->cluster_pos = 0;
968 if (mkv->dyn_bc)
969 mkv_flush_dynbuf(s);
970 }
971
972 // check if we have an audio packet cached
973 if (mkv->cur_audio_pkt.size > 0) {
974 ret = mkv_write_packet_internal(s, &mkv->cur_audio_pkt);
975 mkv->cur_audio_pkt.size = 0;
976 if (ret < 0) {
977 av_log(s, AV_LOG_ERROR, "Could not write cached audio packet ret:%d\n", ret);
978 return ret;
979 }
980 }
981
982 // buffer an audio packet to ensure the packet containing the video
983 // keyframe's timecode is contained in the same cluster for WebM
984 if (codec->codec_type == AVMEDIA_TYPE_AUDIO)
985 ret = mkv_copy_packet(mkv, pkt);
986 else
987 ret = mkv_write_packet_internal(s, pkt);
988 return ret;
989 }
990
991 static int mkv_write_trailer(AVFormatContext *s)
992 {
993 MatroskaMuxContext *mkv = s->priv_data;
994 ByteIOContext *pb = s->pb;
995 int64_t currentpos, cuespos;
996 int ret;
997
998 // check if we have an audio packet cached
999 if (mkv->cur_audio_pkt.size > 0) {
1000 ret = mkv_write_packet_internal(s, &mkv->cur_audio_pkt);
1001 mkv->cur_audio_pkt.size = 0;
1002 if (ret < 0) {
1003 av_log(s, AV_LOG_ERROR, "Could not write cached audio packet ret:%d\n", ret);
1004 return ret;
1005 }
1006 }
1007
1008 if (mkv->dyn_bc) {
1009 end_ebml_master(mkv->dyn_bc, mkv->cluster);
1010 mkv_flush_dynbuf(s);
1011 } else if (mkv->cluster_pos) {
1012 end_ebml_master(pb, mkv->cluster);
1013 }
1014
1015 if (!url_is_streamed(pb)) {
1016 cuespos = mkv_write_cues(pb, mkv->cues, s->nb_streams);
1017
1018 ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_CUES , cuespos);
1019 if (ret < 0) return ret;
1020 mkv_write_seekhead(pb, mkv->main_seekhead);
1021
1022 // update the duration
1023 av_log(s, AV_LOG_DEBUG, "end duration = %" PRIu64 "\n", mkv->duration);
1024 currentpos = url_ftell(pb);
1025 url_fseek(pb, mkv->duration_offset, SEEK_SET);
1026 put_ebml_float(pb, MATROSKA_ID_DURATION, mkv->duration);
1027
1028 url_fseek(pb, currentpos, SEEK_SET);
1029 }
1030
1031 end_ebml_master(pb, mkv->segment);
1032 av_free(mkv->tracks);
1033 av_destruct_packet(&mkv->cur_audio_pkt);
1034 put_flush_packet(pb);
1035 return 0;
1036 }
1037
1038 #if CONFIG_MATROSKA_MUXER
1039 AVOutputFormat matroska_muxer = {
1040 "matroska",
1041 NULL_IF_CONFIG_SMALL("Matroska file format"),
1042 "video/x-matroska",
1043 "mkv",
1044 sizeof(MatroskaMuxContext),
1045 CODEC_ID_MP2,
1046 CODEC_ID_MPEG4,
1047 mkv_write_header,
1048 mkv_write_packet,
1049 mkv_write_trailer,
1050 .flags = AVFMT_GLOBALHEADER | AVFMT_VARIABLE_FPS,
1051 .codec_tag = (const AVCodecTag* const []){ff_codec_bmp_tags, ff_codec_wav_tags, 0},
1052 .subtitle_codec = CODEC_ID_TEXT,
1053 };
1054 #endif
1055
1056 #if CONFIG_WEBM_MUXER
1057 AVOutputFormat webm_muxer = {
1058 "webm",
1059 NULL_IF_CONFIG_SMALL("WebM file format"),
1060 "video/webm",
1061 "webm",
1062 sizeof(MatroskaMuxContext),
1063 CODEC_ID_VORBIS,
1064 CODEC_ID_VP8,
1065 mkv_write_header,
1066 mkv_write_packet,
1067 mkv_write_trailer,
1068 .flags = AVFMT_GLOBALHEADER | AVFMT_VARIABLE_FPS,
1069 };
1070 #endif
1071
1072 #if CONFIG_MATROSKA_AUDIO_MUXER
1073 AVOutputFormat matroska_audio_muxer = {
1074 "matroska",
1075 NULL_IF_CONFIG_SMALL("Matroska file format"),
1076 "audio/x-matroska",
1077 "mka",
1078 sizeof(MatroskaMuxContext),
1079 CODEC_ID_MP2,
1080 CODEC_ID_NONE,
1081 mkv_write_header,
1082 mkv_write_packet,
1083 mkv_write_trailer,
1084 .flags = AVFMT_GLOBALHEADER,
1085 .codec_tag = (const AVCodecTag* const []){ff_codec_wav_tags, 0},
1086 };
1087 #endif