2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The FFmpeg Project
5 * This file is part of FFmpeg.
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.
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.
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
24 * Matroska file demuxer
25 * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
26 * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
27 * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28 * @see specs available on the Matroska project page: http://www.matroska.org/
34 #include "avio_internal.h"
35 /* For ff_codec_get_id(). */
40 #include "libavcodec/mpeg4audio.h"
41 #include "libavutil/intfloat.h"
42 #include "libavutil/intreadwrite.h"
43 #include "libavutil/avstring.h"
44 #include "libavutil/lzo.h"
45 #include "libavutil/dict.h"
66 typedef const struct EbmlSyntax {
75 const struct EbmlSyntax *n;
95 uint64_t doctype_version;
101 } MatroskaTrackCompression;
106 MatroskaTrackCompression compression;
107 } MatroskaTrackEncoding;
111 uint64_t display_width;
112 uint64_t display_height;
113 uint64_t pixel_width;
114 uint64_t pixel_height;
116 uint64_t stereo_mode;
117 } MatroskaTrackVideo;
121 double out_samplerate;
125 /* real audio header (extracted from extradata) */
132 uint64_t buf_timecode;
134 } MatroskaTrackAudio;
139 } MatroskaTrackPlane;
142 EbmlList combine_planes;
143 } MatroskaTrackOperation;
154 uint64_t default_duration;
155 uint64_t flag_default;
156 uint64_t flag_forced;
157 MatroskaTrackVideo video;
158 MatroskaTrackAudio audio;
159 MatroskaTrackOperation operation;
163 int64_t end_timecode;
174 } MatroskaAttachement;
212 MatroskaTagTarget target;
232 AVFormatContext *ctx;
236 MatroskaLevel levels[EBML_MAX_DEPTH];
245 EbmlList attachments;
251 /* byte position of the segment inside the stream */
252 int64_t segment_start;
254 /* the packet queue */
261 /* What to skip before effectively reading a packet. */
262 int skip_to_keyframe;
263 uint64_t skip_to_timecode;
265 /* File has a CUES element, but we defer parsing until it is needed. */
266 int cues_parsing_deferred;
268 int current_cluster_num_blocks;
269 int64_t current_cluster_pos;
270 MatroskaCluster current_cluster;
272 /* File has SSA subtitles which prevent incremental cluster parsing. */
274 } MatroskaDemuxContext;
283 static EbmlSyntax ebml_header[] = {
284 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
285 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
286 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
287 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
288 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
289 { EBML_ID_EBMLVERSION, EBML_NONE },
290 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
294 static EbmlSyntax ebml_syntax[] = {
295 { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
299 static EbmlSyntax matroska_info[] = {
300 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
301 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
302 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
303 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
304 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
305 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext,date_utc) },
306 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
310 static EbmlSyntax matroska_track_video[] = {
311 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
312 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
313 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
314 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
315 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
316 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo,color_space) },
317 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,stereo_mode) },
318 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
319 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
320 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
321 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
322 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
323 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
324 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
328 static EbmlSyntax matroska_track_audio[] = {
329 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
330 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
331 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
332 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
336 static EbmlSyntax matroska_track_encoding_compression[] = {
337 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
338 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
342 static EbmlSyntax matroska_track_encoding[] = {
343 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
344 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
345 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
346 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
350 static EbmlSyntax matroska_track_encodings[] = {
351 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
355 static EbmlSyntax matroska_track_plane[] = {
356 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
357 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
361 static EbmlSyntax matroska_track_combine_planes[] = {
362 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n=matroska_track_plane} },
366 static EbmlSyntax matroska_track_operation[] = {
367 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n=matroska_track_combine_planes} },
371 static EbmlSyntax matroska_track[] = {
372 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
373 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack,name) },
374 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
375 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
376 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
377 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
378 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
379 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
380 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
381 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
382 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack,flag_forced), {.u=0} },
383 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
384 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
385 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack,operation), {.n=matroska_track_operation} },
386 { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
387 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
388 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
389 { MATROSKA_ID_CODECNAME, EBML_NONE },
390 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
391 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
392 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
393 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
394 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
395 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
399 static EbmlSyntax matroska_tracks[] = {
400 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
404 static EbmlSyntax matroska_attachment[] = {
405 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
406 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
407 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
408 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
409 { MATROSKA_ID_FILEDESC, EBML_NONE },
413 static EbmlSyntax matroska_attachments[] = {
414 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
418 static EbmlSyntax matroska_chapter_display[] = {
419 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
420 { MATROSKA_ID_CHAPLANG, EBML_NONE },
424 static EbmlSyntax matroska_chapter_entry[] = {
425 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
426 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
427 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
428 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
429 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
430 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
431 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
432 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
436 static EbmlSyntax matroska_chapter[] = {
437 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
438 { MATROSKA_ID_EDITIONUID, EBML_NONE },
439 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
440 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
441 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
445 static EbmlSyntax matroska_chapters[] = {
446 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
450 static EbmlSyntax matroska_index_pos[] = {
451 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
452 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
453 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
457 static EbmlSyntax matroska_index_entry[] = {
458 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
459 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
463 static EbmlSyntax matroska_index[] = {
464 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
468 static EbmlSyntax matroska_simpletag[] = {
469 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
470 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
471 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag,lang), {.s="und"} },
472 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag,def) },
473 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag,def) },
474 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
478 static EbmlSyntax matroska_tagtargets[] = {
479 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget,type) },
480 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
481 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
482 { MATROSKA_ID_TAGTARGETS_CHAPTERUID,EBML_UINT, 0, offsetof(MatroskaTagTarget,chapteruid) },
483 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
487 static EbmlSyntax matroska_tag[] = {
488 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
489 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
493 static EbmlSyntax matroska_tags[] = {
494 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
498 static EbmlSyntax matroska_seekhead_entry[] = {
499 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
500 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
504 static EbmlSyntax matroska_seekhead[] = {
505 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
509 static EbmlSyntax matroska_segment[] = {
510 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
511 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
512 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
513 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
514 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
515 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
516 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
517 { MATROSKA_ID_CLUSTER, EBML_STOP },
521 static EbmlSyntax matroska_segments[] = {
522 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
526 static EbmlSyntax matroska_blockgroup[] = {
527 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
528 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
529 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration) },
530 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
531 { 1, EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
535 static EbmlSyntax matroska_cluster[] = {
536 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
537 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
538 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
539 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
540 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
544 static EbmlSyntax matroska_clusters[] = {
545 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
546 { MATROSKA_ID_INFO, EBML_NONE },
547 { MATROSKA_ID_CUES, EBML_NONE },
548 { MATROSKA_ID_TAGS, EBML_NONE },
549 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
553 static EbmlSyntax matroska_cluster_incremental_parsing[] = {
554 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
555 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
556 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
557 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
558 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
559 { MATROSKA_ID_INFO, EBML_NONE },
560 { MATROSKA_ID_CUES, EBML_NONE },
561 { MATROSKA_ID_TAGS, EBML_NONE },
562 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
563 { MATROSKA_ID_CLUSTER, EBML_STOP },
567 static EbmlSyntax matroska_cluster_incremental[] = {
568 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
569 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
570 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
571 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
572 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
576 static EbmlSyntax matroska_clusters_incremental[] = {
577 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster_incremental} },
578 { MATROSKA_ID_INFO, EBML_NONE },
579 { MATROSKA_ID_CUES, EBML_NONE },
580 { MATROSKA_ID_TAGS, EBML_NONE },
581 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
585 static const char *const matroska_doctypes[] = { "matroska", "webm" };
587 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
589 AVIOContext *pb = matroska->ctx->pb;
591 matroska->current_id = 0;
592 matroska->num_levels = 0;
594 // seek to next position to resync from
595 if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0 || avio_tell(pb) <= last_pos)
600 // try to find a toplevel element
601 while (!url_feof(pb)) {
602 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
603 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
604 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
605 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS)
607 matroska->current_id = id;
610 id = (id << 8) | avio_r8(pb);
618 * Return: Whether we reached the end of a level in the hierarchy or not.
620 static int ebml_level_end(MatroskaDemuxContext *matroska)
622 AVIOContext *pb = matroska->ctx->pb;
623 int64_t pos = avio_tell(pb);
625 if (matroska->num_levels > 0) {
626 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
627 if (pos - level->start >= level->length || matroska->current_id) {
628 matroska->num_levels--;
636 * Read: an "EBML number", which is defined as a variable-length
637 * array of bytes. The first byte indicates the length by giving a
638 * number of 0-bits followed by a one. The position of the first
639 * "one" bit inside the first byte indicates the length of this
641 * Returns: number of bytes read, < 0 on error
643 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
644 int max_size, uint64_t *number)
649 /* The first byte tells us the length in bytes - avio_r8() can normally
650 * return 0, but since that's not a valid first ebmlID byte, we can
651 * use it safely here to catch EOS. */
652 if (!(total = avio_r8(pb))) {
653 /* we might encounter EOS here */
655 int64_t pos = avio_tell(pb);
656 av_log(matroska->ctx, AV_LOG_ERROR,
657 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
659 return pb->error ? pb->error : AVERROR(EIO);
664 /* get the length of the EBML number */
665 read = 8 - ff_log2_tab[total];
666 if (read > max_size) {
667 int64_t pos = avio_tell(pb) - 1;
668 av_log(matroska->ctx, AV_LOG_ERROR,
669 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
670 (uint8_t) total, pos, pos);
671 return AVERROR_INVALIDDATA;
674 /* read out length */
675 total ^= 1 << ff_log2_tab[total];
677 total = (total << 8) | avio_r8(pb);
685 * Read a EBML length value.
686 * This needs special handling for the "unknown length" case which has multiple
689 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
692 int res = ebml_read_num(matroska, pb, 8, number);
693 if (res > 0 && *number + 1 == 1ULL << (7 * res))
694 *number = 0xffffffffffffffULL;
699 * Read the next element as an unsigned int.
700 * 0 is success, < 0 is failure.
702 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
707 return AVERROR_INVALIDDATA;
709 /* big-endian ordering; build up number */
712 *num = (*num << 8) | avio_r8(pb);
718 * Read the next element as a float.
719 * 0 is success, < 0 is failure.
721 static int ebml_read_float(AVIOContext *pb, int size, double *num)
725 } else if (size == 4) {
726 *num = av_int2float(avio_rb32(pb));
727 } else if (size == 8){
728 *num = av_int2double(avio_rb64(pb));
730 return AVERROR_INVALIDDATA;
736 * Read the next element as an ASCII string.
737 * 0 is success, < 0 is failure.
739 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
743 /* EBML strings are usually not 0-terminated, so we allocate one
744 * byte more, read the string and NULL-terminate it ourselves. */
745 if (!(res = av_malloc(size + 1)))
746 return AVERROR(ENOMEM);
747 if (avio_read(pb, (uint8_t *) res, size) != size) {
759 * Read the next element as binary data.
760 * 0 is success, < 0 is failure.
762 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
765 if (!(bin->data = av_malloc(length)))
766 return AVERROR(ENOMEM);
769 bin->pos = avio_tell(pb);
770 if (avio_read(pb, bin->data, length) != length) {
771 av_freep(&bin->data);
779 * Read the next element, but only the header. The contents
780 * are supposed to be sub-elements which can be read separately.
781 * 0 is success, < 0 is failure.
783 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
785 AVIOContext *pb = matroska->ctx->pb;
786 MatroskaLevel *level;
788 if (matroska->num_levels >= EBML_MAX_DEPTH) {
789 av_log(matroska->ctx, AV_LOG_ERROR,
790 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
791 return AVERROR(ENOSYS);
794 level = &matroska->levels[matroska->num_levels++];
795 level->start = avio_tell(pb);
796 level->length = length;
802 * Read signed/unsigned "EBML" numbers.
803 * Return: number of bytes processed, < 0 on error
805 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
806 uint8_t *data, uint32_t size, uint64_t *num)
809 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
810 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
814 * Same as above, but signed.
816 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
817 uint8_t *data, uint32_t size, int64_t *num)
822 /* read as unsigned number first */
823 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
826 /* make signed (weird way) */
827 *num = unum - ((1LL << (7*res - 1)) - 1);
832 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
833 EbmlSyntax *syntax, void *data);
835 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
836 uint32_t id, void *data)
839 for (i=0; syntax[i].id; i++)
840 if (id == syntax[i].id)
842 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
843 matroska->num_levels > 0 &&
844 matroska->levels[matroska->num_levels-1].length == 0xffffffffffffff)
845 return 0; // we reached the end of an unknown size cluster
846 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
847 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
848 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
849 return AVERROR_INVALIDDATA;
851 return ebml_parse_elem(matroska, &syntax[i], data);
854 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
857 if (!matroska->current_id) {
859 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
862 matroska->current_id = id | 1 << 7*res;
864 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
867 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
872 for (i=0; syntax[i].id; i++)
873 switch (syntax[i].type) {
875 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
878 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
882 *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
886 while (!res && !ebml_level_end(matroska))
887 res = ebml_parse(matroska, syntax, data);
892 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
893 EbmlSyntax *syntax, void *data)
895 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
898 // max. 16 MB for strings
899 [EBML_STR] = 0x1000000,
900 [EBML_UTF8] = 0x1000000,
901 // max. 256 MB for binary data
902 [EBML_BIN] = 0x10000000,
903 // no limits for anything else
905 AVIOContext *pb = matroska->ctx->pb;
906 uint32_t id = syntax->id;
911 data = (char *)data + syntax->data_offset;
912 if (syntax->list_elem_size) {
913 EbmlList *list = data;
914 newelem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
916 return AVERROR(ENOMEM);
917 list->elem = newelem;
918 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
919 memset(data, 0, syntax->list_elem_size);
923 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
924 matroska->current_id = 0;
925 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
927 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
928 av_log(matroska->ctx, AV_LOG_ERROR,
929 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
930 length, max_lengths[syntax->type], syntax->type);
931 return AVERROR_INVALIDDATA;
935 switch (syntax->type) {
936 case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
937 case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
939 case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
940 case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
941 case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
943 if (id == MATROSKA_ID_SEGMENT)
944 matroska->segment_start = avio_tell(matroska->ctx->pb);
945 return ebml_parse_nest(matroska, syntax->def.n, data);
946 case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
947 case EBML_STOP: return 1;
949 if(ffio_limit(pb, length) != length)
951 return avio_skip(pb,length)<0 ? AVERROR(EIO) : 0;
953 if (res == AVERROR_INVALIDDATA)
954 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
955 else if (res == AVERROR(EIO))
956 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
960 static void ebml_free(EbmlSyntax *syntax, void *data)
963 for (i=0; syntax[i].id; i++) {
964 void *data_off = (char *)data + syntax[i].data_offset;
965 switch (syntax[i].type) {
967 case EBML_UTF8: av_freep(data_off); break;
968 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
970 if (syntax[i].list_elem_size) {
971 EbmlList *list = data_off;
972 char *ptr = list->elem;
973 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
974 ebml_free(syntax[i].def.n, ptr);
977 ebml_free(syntax[i].def.n, data_off);
987 static int matroska_probe(AVProbeData *p)
990 int len_mask = 0x80, size = 1, n = 1, i;
993 if (AV_RB32(p->buf) != EBML_ID_HEADER)
996 /* length of header */
998 while (size <= 8 && !(total & len_mask)) {
1004 total &= (len_mask - 1);
1006 total = (total << 8) | p->buf[4 + n++];
1008 /* Does the probe data contain the whole header? */
1009 if (p->buf_size < 4 + size + total)
1012 /* The header should contain a known document type. For now,
1013 * we don't parse the whole header but simply check for the
1014 * availability of that array of characters inside the header.
1015 * Not fully fool-proof, but good enough. */
1016 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1017 int probelen = strlen(matroska_doctypes[i]);
1018 if (total < probelen)
1020 for (n = 4+size; n <= 4+size+total-probelen; n++)
1021 if (!memcmp(p->buf+n, matroska_doctypes[i], probelen))
1022 return AVPROBE_SCORE_MAX;
1025 // probably valid EBML header but no recognized doctype
1026 return AVPROBE_SCORE_MAX/2;
1029 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1032 MatroskaTrack *tracks = matroska->tracks.elem;
1035 for (i=0; i < matroska->tracks.nb_elem; i++)
1036 if (tracks[i].num == num)
1039 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1043 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
1044 MatroskaTrack *track)
1046 MatroskaTrackEncoding *encodings = track->encodings.elem;
1047 uint8_t* data = *buf;
1048 int isize = *buf_size;
1049 uint8_t* pkt_data = NULL;
1050 uint8_t av_unused *newpktdata;
1051 int pkt_size = isize;
1055 if (pkt_size >= 10000000U)
1058 switch (encodings[0].compression.algo) {
1059 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1060 if (encodings[0].compression.settings.size && !encodings[0].compression.settings.data) {
1061 av_log(0, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1064 return encodings[0].compression.settings.size;
1065 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1067 olen = pkt_size *= 3;
1068 pkt_data = av_realloc(pkt_data, pkt_size+AV_LZO_OUTPUT_PADDING);
1069 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1070 } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
1076 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
1077 z_stream zstream = {0};
1078 if (inflateInit(&zstream) != Z_OK)
1080 zstream.next_in = data;
1081 zstream.avail_in = isize;
1084 newpktdata = av_realloc(pkt_data, pkt_size);
1086 inflateEnd(&zstream);
1089 pkt_data = newpktdata;
1090 zstream.avail_out = pkt_size - zstream.total_out;
1091 zstream.next_out = pkt_data + zstream.total_out;
1093 result = inflate(&zstream, Z_NO_FLUSH);
1095 result = Z_MEM_ERROR;
1096 } while (result==Z_OK && pkt_size<10000000);
1097 pkt_size = zstream.total_out;
1098 inflateEnd(&zstream);
1099 if (result != Z_STREAM_END)
1105 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
1106 bz_stream bzstream = {0};
1107 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1109 bzstream.next_in = data;
1110 bzstream.avail_in = isize;
1113 newpktdata = av_realloc(pkt_data, pkt_size);
1115 BZ2_bzDecompressEnd(&bzstream);
1118 pkt_data = newpktdata;
1119 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1120 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1122 result = BZ2_bzDecompress(&bzstream);
1124 result = BZ_MEM_ERROR;
1125 } while (result==BZ_OK && pkt_size<10000000);
1126 pkt_size = bzstream.total_out_lo32;
1127 BZ2_bzDecompressEnd(&bzstream);
1128 if (result != BZ_STREAM_END)
1138 *buf_size = pkt_size;
1145 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
1146 AVPacket *pkt, uint64_t display_duration)
1148 char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
1149 for (; *ptr!=',' && ptr<end-1; ptr++);
1153 for (; *ptr!=',' && ptr<end-1; ptr++);
1155 int64_t end_pts = pkt->pts + display_duration;
1156 int sc = matroska->time_scale * pkt->pts / 10000000;
1157 int ec = matroska->time_scale * end_pts / 10000000;
1158 int sh, sm, ss, eh, em, es, len;
1159 sh = sc/360000; sc -= 360000*sh;
1160 sm = sc/ 6000; sc -= 6000*sm;
1161 ss = sc/ 100; sc -= 100*ss;
1162 eh = ec/360000; ec -= 360000*eh;
1163 em = ec/ 6000; ec -= 6000*em;
1164 es = ec/ 100; ec -= 100*es;
1166 len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
1167 if (!(line = av_malloc(len)))
1169 snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
1170 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
1173 pkt->size = strlen(line);
1177 static int matroska_merge_packets(AVPacket *out, AVPacket *in)
1179 int ret = av_grow_packet(out, in->size);
1182 memcpy(out->data + out->size - in->size, in->data, in->size);
1183 av_destruct_packet(in);
1188 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1189 AVDictionary **metadata, char *prefix)
1191 MatroskaTag *tags = list->elem;
1195 for (i=0; i < list->nb_elem; i++) {
1196 const char *lang= (tags[i].lang && strcmp(tags[i].lang, "und")) ? tags[i].lang : NULL;
1198 if (!tags[i].name) {
1199 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1202 if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1203 else av_strlcpy(key, tags[i].name, sizeof(key));
1204 if (tags[i].def || !lang) {
1205 av_dict_set(metadata, key, tags[i].string, 0);
1206 if (tags[i].sub.nb_elem)
1207 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1210 av_strlcat(key, "-", sizeof(key));
1211 av_strlcat(key, lang, sizeof(key));
1212 av_dict_set(metadata, key, tags[i].string, 0);
1213 if (tags[i].sub.nb_elem)
1214 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1217 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1220 static void matroska_convert_tags(AVFormatContext *s)
1222 MatroskaDemuxContext *matroska = s->priv_data;
1223 MatroskaTags *tags = matroska->tags.elem;
1226 for (i=0; i < matroska->tags.nb_elem; i++) {
1227 if (tags[i].target.attachuid) {
1228 MatroskaAttachement *attachment = matroska->attachments.elem;
1229 for (j=0; j<matroska->attachments.nb_elem; j++)
1230 if (attachment[j].uid == tags[i].target.attachuid
1231 && attachment[j].stream)
1232 matroska_convert_tag(s, &tags[i].tag,
1233 &attachment[j].stream->metadata, NULL);
1234 } else if (tags[i].target.chapteruid) {
1235 MatroskaChapter *chapter = matroska->chapters.elem;
1236 for (j=0; j<matroska->chapters.nb_elem; j++)
1237 if (chapter[j].uid == tags[i].target.chapteruid
1238 && chapter[j].chapter)
1239 matroska_convert_tag(s, &tags[i].tag,
1240 &chapter[j].chapter->metadata, NULL);
1241 } else if (tags[i].target.trackuid) {
1242 MatroskaTrack *track = matroska->tracks.elem;
1243 for (j=0; j<matroska->tracks.nb_elem; j++)
1244 if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1245 matroska_convert_tag(s, &tags[i].tag,
1246 &track[j].stream->metadata, NULL);
1248 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1249 tags[i].target.type);
1254 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska, int idx)
1256 EbmlList *seekhead_list = &matroska->seekhead;
1257 MatroskaSeekhead *seekhead = seekhead_list->elem;
1258 uint32_t level_up = matroska->level_up;
1259 int64_t before_pos = avio_tell(matroska->ctx->pb);
1260 uint32_t saved_id = matroska->current_id;
1261 MatroskaLevel level;
1265 if (idx >= seekhead_list->nb_elem
1266 || seekhead[idx].id == MATROSKA_ID_SEEKHEAD
1267 || seekhead[idx].id == MATROSKA_ID_CLUSTER)
1271 offset = seekhead[idx].pos + matroska->segment_start;
1272 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1273 /* We don't want to lose our seekhead level, so we add
1274 * a dummy. This is a crude hack. */
1275 if (matroska->num_levels == EBML_MAX_DEPTH) {
1276 av_log(matroska->ctx, AV_LOG_INFO,
1277 "Max EBML element depth (%d) reached, "
1278 "cannot parse further.\n", EBML_MAX_DEPTH);
1279 ret = AVERROR_INVALIDDATA;
1282 level.length = (uint64_t)-1;
1283 matroska->levels[matroska->num_levels] = level;
1284 matroska->num_levels++;
1285 matroska->current_id = 0;
1287 ret = ebml_parse(matroska, matroska_segment, matroska);
1289 /* remove dummy level */
1290 while (matroska->num_levels) {
1291 uint64_t length = matroska->levels[--matroska->num_levels].length;
1292 if (length == (uint64_t)-1)
1298 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1299 matroska->level_up = level_up;
1300 matroska->current_id = saved_id;
1305 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1307 EbmlList *seekhead_list = &matroska->seekhead;
1308 int64_t before_pos = avio_tell(matroska->ctx->pb);
1311 // we should not do any seeking in the streaming case
1312 if (!matroska->ctx->pb->seekable ||
1313 (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1316 for (i = 0; i < seekhead_list->nb_elem; i++) {
1317 MatroskaSeekhead *seekhead = seekhead_list->elem;
1318 if (seekhead[i].pos <= before_pos)
1321 // defer cues parsing until we actually need cue data.
1322 if (seekhead[i].id == MATROSKA_ID_CUES) {
1323 matroska->cues_parsing_deferred = 1;
1327 if (matroska_parse_seekhead_entry(matroska, i) < 0) {
1328 // mark index as broken
1329 matroska->cues_parsing_deferred = -1;
1335 static void matroska_add_index_entries(MatroskaDemuxContext *matroska) {
1336 EbmlList *index_list;
1337 MatroskaIndex *index;
1338 int index_scale = 1;
1341 index_list = &matroska->index;
1342 index = index_list->elem;
1343 if (index_list->nb_elem
1344 && index[0].time > 1E14/matroska->time_scale) {
1345 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1346 index_scale = matroska->time_scale;
1348 for (i = 0; i < index_list->nb_elem; i++) {
1349 EbmlList *pos_list = &index[i].pos;
1350 MatroskaIndexPos *pos = pos_list->elem;
1351 for (j = 0; j < pos_list->nb_elem; j++) {
1352 MatroskaTrack *track = matroska_find_track_by_num(matroska, pos[j].track);
1353 if (track && track->stream)
1354 av_add_index_entry(track->stream,
1355 pos[j].pos + matroska->segment_start,
1356 index[i].time/index_scale, 0, 0,
1362 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1363 EbmlList *seekhead_list = &matroska->seekhead;
1364 MatroskaSeekhead *seekhead = seekhead_list->elem;
1367 for (i = 0; i < seekhead_list->nb_elem; i++)
1368 if (seekhead[i].id == MATROSKA_ID_CUES)
1370 assert(i <= seekhead_list->nb_elem);
1372 if (matroska_parse_seekhead_entry(matroska, i) < 0)
1373 matroska->cues_parsing_deferred = -1;
1374 matroska_add_index_entries(matroska);
1377 static int matroska_aac_profile(char *codec_id)
1379 static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1382 for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1383 if (strstr(codec_id, aac_profiles[profile]))
1388 static int matroska_aac_sri(int samplerate)
1392 for (sri=0; sri<FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1393 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1398 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1401 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1402 time_t creation_time = date_utc / 1000000000 + 978307200;
1403 struct tm *ptm = gmtime(&creation_time);
1405 strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm);
1406 av_dict_set(metadata, "creation_time", buffer, 0);
1409 static int matroska_read_header(AVFormatContext *s)
1411 MatroskaDemuxContext *matroska = s->priv_data;
1412 EbmlList *attachements_list = &matroska->attachments;
1413 MatroskaAttachement *attachements;
1414 EbmlList *chapters_list = &matroska->chapters;
1415 MatroskaChapter *chapters;
1416 MatroskaTrack *tracks;
1417 uint64_t max_start = 0;
1425 /* First read the EBML header. */
1426 if (ebml_parse(matroska, ebml_syntax, &ebml)
1427 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1428 || ebml.id_length > sizeof(uint32_t) || ebml.doctype_version > 3 || !ebml.doctype) {
1429 av_log(matroska->ctx, AV_LOG_ERROR,
1430 "EBML header using unsupported features\n"
1431 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1432 ebml.version, ebml.doctype, ebml.doctype_version);
1433 ebml_free(ebml_syntax, &ebml);
1434 return AVERROR_PATCHWELCOME;
1435 } else if (ebml.doctype_version == 3) {
1436 av_log(matroska->ctx, AV_LOG_WARNING,
1437 "EBML header using unsupported features\n"
1438 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1439 ebml.version, ebml.doctype, ebml.doctype_version);
1441 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
1442 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
1444 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
1445 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
1447 ebml_free(ebml_syntax, &ebml);
1449 /* The next thing is a segment. */
1450 pos = avio_tell(matroska->ctx->pb);
1451 res = ebml_parse(matroska, matroska_segments, matroska);
1452 // try resyncing until we find a EBML_STOP type element.
1454 res = matroska_resync(matroska, pos);
1457 pos = avio_tell(matroska->ctx->pb);
1458 res = ebml_parse(matroska, matroska_segment, matroska);
1460 matroska_execute_seekhead(matroska);
1462 if (!matroska->time_scale)
1463 matroska->time_scale = 1000000;
1464 if (matroska->duration)
1465 matroska->ctx->duration = matroska->duration * matroska->time_scale
1466 * 1000 / AV_TIME_BASE;
1467 av_dict_set(&s->metadata, "title", matroska->title, 0);
1469 if (matroska->date_utc.size == 8)
1470 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
1472 tracks = matroska->tracks.elem;
1473 for (i=0; i < matroska->tracks.nb_elem; i++) {
1474 MatroskaTrack *track = &tracks[i];
1475 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1476 EbmlList *encodings_list = &track->encodings;
1477 MatroskaTrackEncoding *encodings = encodings_list->elem;
1478 uint8_t *extradata = NULL;
1479 int extradata_size = 0;
1480 int extradata_offset = 0;
1481 uint32_t fourcc = 0;
1484 /* Apply some sanity checks. */
1485 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1486 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1487 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1488 av_log(matroska->ctx, AV_LOG_INFO,
1489 "Unknown or unsupported track type %"PRIu64"\n",
1493 if (track->codec_id == NULL)
1496 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1497 if (!track->default_duration && track->video.frame_rate > 0)
1498 track->default_duration = 1000000000/track->video.frame_rate;
1499 if (!track->video.display_width)
1500 track->video.display_width = track->video.pixel_width;
1501 if (!track->video.display_height)
1502 track->video.display_height = track->video.pixel_height;
1503 if (track->video.color_space.size == 4)
1504 fourcc = AV_RL32(track->video.color_space.data);
1505 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1506 if (!track->audio.out_samplerate)
1507 track->audio.out_samplerate = track->audio.samplerate;
1509 if (encodings_list->nb_elem > 1) {
1510 av_log(matroska->ctx, AV_LOG_ERROR,
1511 "Multiple combined encodings not supported");
1512 } else if (encodings_list->nb_elem == 1) {
1513 if (encodings[0].type ||
1514 (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1516 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1519 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1521 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
1522 encodings[0].scope = 0;
1523 av_log(matroska->ctx, AV_LOG_ERROR,
1524 "Unsupported encoding type");
1525 } else if (track->codec_priv.size && encodings[0].scope&2) {
1526 uint8_t *codec_priv = track->codec_priv.data;
1527 int offset = matroska_decode_buffer(&track->codec_priv.data,
1528 &track->codec_priv.size,
1531 track->codec_priv.data = NULL;
1532 track->codec_priv.size = 0;
1533 av_log(matroska->ctx, AV_LOG_ERROR,
1534 "Failed to decode codec private data\n");
1535 } else if (offset > 0) {
1536 track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
1537 memcpy(track->codec_priv.data,
1538 encodings[0].compression.settings.data, offset);
1539 memcpy(track->codec_priv.data+offset, codec_priv,
1540 track->codec_priv.size);
1541 track->codec_priv.size += offset;
1543 if (codec_priv != track->codec_priv.data)
1544 av_free(codec_priv);
1548 for(j=0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++){
1549 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1550 strlen(ff_mkv_codec_tags[j].str))){
1551 codec_id= ff_mkv_codec_tags[j].id;
1556 st = track->stream = avformat_new_stream(s, NULL);
1558 return AVERROR(ENOMEM);
1560 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1561 && track->codec_priv.size >= 40
1562 && track->codec_priv.data != NULL) {
1563 track->ms_compat = 1;
1564 fourcc = AV_RL32(track->codec_priv.data + 16);
1565 codec_id = ff_codec_get_id(ff_codec_bmp_tags, fourcc);
1566 extradata_offset = 40;
1567 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1568 && track->codec_priv.size >= 14
1569 && track->codec_priv.data != NULL) {
1571 ffio_init_context(&b, track->codec_priv.data, track->codec_priv.size,
1572 AVIO_FLAG_READ, NULL, NULL, NULL, NULL);
1573 ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1576 codec_id = st->codec->codec_id;
1577 extradata_offset = FFMIN(track->codec_priv.size, 18);
1578 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1579 && (track->codec_priv.size >= 86)
1580 && (track->codec_priv.data != NULL)) {
1581 fourcc = AV_RL32(track->codec_priv.data);
1582 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1583 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX-12) {
1584 /* Only ALAC's magic cookie is stored in Matroska's track headers.
1585 Create the "atom size", "tag", and "tag version" fields the
1586 decoder expects manually. */
1587 extradata_size = 12 + track->codec_priv.size;
1588 extradata = av_mallocz(extradata_size);
1589 if (extradata == NULL)
1590 return AVERROR(ENOMEM);
1591 AV_WB32(extradata, extradata_size);
1592 memcpy(&extradata[4], "alac", 4);
1593 AV_WB32(&extradata[8], 0);
1594 memcpy(&extradata[12], track->codec_priv.data, track->codec_priv.size);
1595 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1596 switch (track->audio.bitdepth) {
1597 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1598 case 24: codec_id = AV_CODEC_ID_PCM_S24BE; break;
1599 case 32: codec_id = AV_CODEC_ID_PCM_S32BE; break;
1601 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1602 switch (track->audio.bitdepth) {
1603 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1604 case 24: codec_id = AV_CODEC_ID_PCM_S24LE; break;
1605 case 32: codec_id = AV_CODEC_ID_PCM_S32LE; break;
1607 } else if (codec_id==AV_CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1608 codec_id = AV_CODEC_ID_PCM_F64LE;
1609 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1610 int profile = matroska_aac_profile(track->codec_id);
1611 int sri = matroska_aac_sri(track->audio.samplerate);
1612 extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1613 if (extradata == NULL)
1614 return AVERROR(ENOMEM);
1615 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1616 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1617 if (strstr(track->codec_id, "SBR")) {
1618 sri = matroska_aac_sri(track->audio.out_samplerate);
1619 extradata[2] = 0x56;
1620 extradata[3] = 0xE5;
1621 extradata[4] = 0x80 | (sri<<3);
1625 } else if (codec_id == AV_CODEC_ID_TTA) {
1626 extradata_size = 30;
1627 extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1628 if (extradata == NULL)
1629 return AVERROR(ENOMEM);
1630 ffio_init_context(&b, extradata, extradata_size, 1,
1631 NULL, NULL, NULL, NULL);
1632 avio_write(&b, "TTA1", 4);
1634 avio_wl16(&b, track->audio.channels);
1635 avio_wl16(&b, track->audio.bitdepth);
1636 avio_wl32(&b, track->audio.out_samplerate);
1637 avio_wl32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1638 } else if (codec_id == AV_CODEC_ID_RV10 || codec_id == AV_CODEC_ID_RV20 ||
1639 codec_id == AV_CODEC_ID_RV30 || codec_id == AV_CODEC_ID_RV40) {
1640 extradata_offset = 26;
1641 } else if (codec_id == AV_CODEC_ID_RA_144) {
1642 track->audio.out_samplerate = 8000;
1643 track->audio.channels = 1;
1644 } else if ((codec_id == AV_CODEC_ID_RA_288 || codec_id == AV_CODEC_ID_COOK ||
1645 codec_id == AV_CODEC_ID_ATRAC3 || codec_id == AV_CODEC_ID_SIPR)
1646 && track->codec_priv.data) {
1649 ffio_init_context(&b, track->codec_priv.data,track->codec_priv.size,
1650 0, NULL, NULL, NULL, NULL);
1652 flavor = avio_rb16(&b);
1653 track->audio.coded_framesize = avio_rb32(&b);
1655 track->audio.sub_packet_h = avio_rb16(&b);
1656 track->audio.frame_size = avio_rb16(&b);
1657 track->audio.sub_packet_size = avio_rb16(&b);
1658 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1659 if (codec_id == AV_CODEC_ID_RA_288) {
1660 st->codec->block_align = track->audio.coded_framesize;
1661 track->codec_priv.size = 0;
1663 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1664 const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1665 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1666 st->codec->bit_rate = sipr_bit_rate[flavor];
1668 st->codec->block_align = track->audio.sub_packet_size;
1669 extradata_offset = 78;
1672 track->codec_priv.size -= extradata_offset;
1674 if (codec_id == AV_CODEC_ID_NONE)
1675 av_log(matroska->ctx, AV_LOG_INFO,
1676 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1678 if (track->time_scale < 0.01)
1679 track->time_scale = 1.0;
1680 avpriv_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1682 st->codec->codec_id = codec_id;
1684 if (strcmp(track->language, "und"))
1685 av_dict_set(&st->metadata, "language", track->language, 0);
1686 av_dict_set(&st->metadata, "title", track->name, 0);
1688 if (track->flag_default)
1689 st->disposition |= AV_DISPOSITION_DEFAULT;
1690 if (track->flag_forced)
1691 st->disposition |= AV_DISPOSITION_FORCED;
1693 if (!st->codec->extradata) {
1695 st->codec->extradata = extradata;
1696 st->codec->extradata_size = extradata_size;
1697 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1698 st->codec->extradata = av_mallocz(track->codec_priv.size +
1699 FF_INPUT_BUFFER_PADDING_SIZE);
1700 if(st->codec->extradata == NULL)
1701 return AVERROR(ENOMEM);
1702 st->codec->extradata_size = track->codec_priv.size;
1703 memcpy(st->codec->extradata,
1704 track->codec_priv.data + extradata_offset,
1705 track->codec_priv.size);
1709 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1710 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
1712 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1713 st->codec->codec_tag = fourcc;
1714 st->codec->width = track->video.pixel_width;
1715 st->codec->height = track->video.pixel_height;
1716 av_reduce(&st->sample_aspect_ratio.num,
1717 &st->sample_aspect_ratio.den,
1718 st->codec->height * track->video.display_width,
1719 st->codec-> width * track->video.display_height,
1721 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1722 if (track->default_duration) {
1723 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1724 1000000000, track->default_duration, 30000);
1725 #if FF_API_R_FRAME_RATE
1726 st->r_frame_rate = st->avg_frame_rate;
1730 /* export stereo mode flag as metadata tag */
1731 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREO_MODE_COUNT)
1732 av_dict_set(&st->metadata, "stereo_mode", matroska_video_stereo_mode[track->video.stereo_mode], 0);
1734 /* if we have virtual track, mark the real tracks */
1735 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
1737 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
1739 snprintf(buf, sizeof(buf), "%s_%d",
1740 matroska_video_stereo_plane[planes[j].type], i);
1741 for (k=0; k < matroska->tracks.nb_elem; k++)
1742 if (planes[j].uid == tracks[k].uid) {
1743 av_dict_set(&s->streams[k]->metadata,
1744 "stereo_mode", buf, 0);
1748 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1749 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1750 st->codec->sample_rate = track->audio.out_samplerate;
1751 st->codec->channels = track->audio.channels;
1752 if (st->codec->codec_id != AV_CODEC_ID_AAC)
1753 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1754 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1755 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1756 if (st->codec->codec_id == AV_CODEC_ID_SSA)
1757 matroska->contains_ssa = 1;
1761 attachements = attachements_list->elem;
1762 for (j=0; j<attachements_list->nb_elem; j++) {
1763 if (!(attachements[j].filename && attachements[j].mime &&
1764 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1765 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1767 AVStream *st = avformat_new_stream(s, NULL);
1770 av_dict_set(&st->metadata, "filename",attachements[j].filename, 0);
1771 av_dict_set(&st->metadata, "mimetype", attachements[j].mime, 0);
1772 st->codec->codec_id = AV_CODEC_ID_NONE;
1773 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
1774 st->codec->extradata = av_malloc(attachements[j].bin.size + FF_INPUT_BUFFER_PADDING_SIZE);
1775 if(st->codec->extradata == NULL)
1777 st->codec->extradata_size = attachements[j].bin.size;
1778 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1780 for (i=0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
1781 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1782 strlen(ff_mkv_mime_tags[i].str))) {
1783 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1787 attachements[j].stream = st;
1791 chapters = chapters_list->elem;
1792 for (i=0; i<chapters_list->nb_elem; i++)
1793 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1794 && (max_start==0 || chapters[i].start > max_start)) {
1795 chapters[i].chapter =
1796 avpriv_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1797 chapters[i].start, chapters[i].end,
1799 av_dict_set(&chapters[i].chapter->metadata,
1800 "title", chapters[i].title, 0);
1801 max_start = chapters[i].start;
1804 matroska_add_index_entries(matroska);
1806 matroska_convert_tags(s);
1812 * Put one packet in an application-supplied AVPacket struct.
1813 * Returns 0 on success or -1 on failure.
1815 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1818 if (matroska->num_packets > 0) {
1819 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1820 av_free(matroska->packets[0]);
1821 if (matroska->num_packets > 1) {
1823 memmove(&matroska->packets[0], &matroska->packets[1],
1824 (matroska->num_packets - 1) * sizeof(AVPacket *));
1825 newpackets = av_realloc(matroska->packets,
1826 (matroska->num_packets - 1) * sizeof(AVPacket *));
1828 matroska->packets = newpackets;
1830 av_freep(&matroska->packets);
1831 matroska->prev_pkt = NULL;
1833 matroska->num_packets--;
1841 * Free all packets in our internal queue.
1843 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1845 if (matroska->packets) {
1847 for (n = 0; n < matroska->num_packets; n++) {
1848 av_free_packet(matroska->packets[n]);
1849 av_free(matroska->packets[n]);
1851 av_freep(&matroska->packets);
1852 matroska->num_packets = 0;
1856 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
1857 int size, int64_t pos, uint64_t cluster_time,
1858 uint64_t duration, int is_keyframe,
1859 int64_t cluster_pos)
1861 uint64_t timecode = AV_NOPTS_VALUE;
1862 MatroskaTrack *track;
1867 uint32_t *lace_size = NULL;
1868 int n, flags, laces = 0;
1871 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
1872 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
1878 track = matroska_find_track_by_num(matroska, num);
1879 if (!track || !track->stream) {
1880 av_log(matroska->ctx, AV_LOG_INFO,
1881 "Invalid stream %"PRIu64" or size %u\n", num, size);
1882 return AVERROR_INVALIDDATA;
1883 } else if (size <= 3)
1886 if (st->discard >= AVDISCARD_ALL)
1888 av_assert1(duration != AV_NOPTS_VALUE);
1890 block_time = AV_RB16(data);
1894 if (is_keyframe == -1)
1895 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
1897 if (cluster_time != (uint64_t)-1
1898 && (block_time >= 0 || cluster_time >= -block_time)) {
1899 timecode = cluster_time + block_time;
1900 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
1901 && timecode < track->end_timecode)
1902 is_keyframe = 0; /* overlapping subtitles are not key frame */
1904 av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
1907 if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1908 if (timecode < matroska->skip_to_timecode)
1910 if (!st->skip_to_keyframe) {
1911 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
1912 matroska->skip_to_keyframe = 0;
1915 matroska->skip_to_keyframe = 0;
1918 switch ((flags & 0x06) >> 1) {
1919 case 0x0: /* no lacing */
1921 lace_size = av_mallocz(sizeof(int));
1922 lace_size[0] = size;
1925 case 0x1: /* Xiph lacing */
1926 case 0x2: /* fixed-size lacing */
1927 case 0x3: /* EBML lacing */
1928 av_assert0(size>0); // size <=3 is checked before size-=3 above
1929 laces = (*data) + 1;
1932 lace_size = av_mallocz(laces * sizeof(int));
1934 switch ((flags & 0x06) >> 1) {
1935 case 0x1: /* Xiph lacing */ {
1938 for (n = 0; res == 0 && n < laces - 1; n++) {
1945 lace_size[n] += temp;
1951 total += lace_size[n];
1953 lace_size[n] = size - total;
1957 case 0x2: /* fixed-size lacing */
1958 for (n = 0; n < laces; n++)
1959 lace_size[n] = size / laces;
1962 case 0x3: /* EBML lacing */ {
1964 n = matroska_ebmlnum_uint(matroska, data, size, &num);
1966 av_log(matroska->ctx, AV_LOG_INFO,
1967 "EBML block data error\n");
1972 total = lace_size[0] = num;
1973 for (n = 1; res == 0 && n < laces - 1; n++) {
1976 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1978 av_log(matroska->ctx, AV_LOG_INFO,
1979 "EBML block data error\n");
1984 lace_size[n] = lace_size[n - 1] + snum;
1985 total += lace_size[n];
1987 lace_size[laces - 1] = size - total;
1996 duration = track->default_duration * laces / matroska->time_scale;
1998 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
1999 track->end_timecode = FFMAX(track->end_timecode, timecode+duration);
2001 for (n = 0; n < laces; n++) {
2002 int64_t lace_duration = duration*(n+1) / laces - duration*n / laces;
2004 if (lace_size[n] > size) {
2005 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2009 if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2010 st->codec->codec_id == AV_CODEC_ID_COOK ||
2011 st->codec->codec_id == AV_CODEC_ID_SIPR ||
2012 st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2013 st->codec->block_align && track->audio.sub_packet_size) {
2014 int a = st->codec->block_align;
2015 int sps = track->audio.sub_packet_size;
2016 int cfs = track->audio.coded_framesize;
2017 int h = track->audio.sub_packet_h;
2018 int y = track->audio.sub_packet_cnt;
2019 int w = track->audio.frame_size;
2022 if (!track->audio.pkt_cnt) {
2023 if (track->audio.sub_packet_cnt == 0)
2024 track->audio.buf_timecode = timecode;
2025 if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2026 if (size < cfs * h / 2) {
2027 av_log(matroska->ctx, AV_LOG_ERROR,
2028 "Corrupt int4 RM-style audio packet size\n");
2029 res = AVERROR_INVALIDDATA;
2032 for (x=0; x<h/2; x++)
2033 memcpy(track->audio.buf+x*2*w+y*cfs,
2035 } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2037 av_log(matroska->ctx, AV_LOG_ERROR,
2038 "Corrupt sipr RM-style audio packet size\n");
2039 res = AVERROR_INVALIDDATA;
2042 memcpy(track->audio.buf + y*w, data, w);
2044 if (size < sps * w / sps) {
2045 av_log(matroska->ctx, AV_LOG_ERROR,
2046 "Corrupt generic RM-style audio packet size\n");
2047 res = AVERROR_INVALIDDATA;
2050 for (x=0; x<w/sps; x++)
2051 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
2054 if (++track->audio.sub_packet_cnt >= h) {
2055 if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2056 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2057 track->audio.sub_packet_cnt = 0;
2058 track->audio.pkt_cnt = h*w / a;
2061 while (track->audio.pkt_cnt) {
2062 pkt = av_mallocz(sizeof(AVPacket));
2063 av_new_packet(pkt, a);
2064 memcpy(pkt->data, track->audio.buf
2065 + a * (h*w / a - track->audio.pkt_cnt--), a);
2066 pkt->pts = track->audio.buf_timecode;
2067 track->audio.buf_timecode = AV_NOPTS_VALUE;
2069 pkt->stream_index = st->index;
2070 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2073 MatroskaTrackEncoding *encodings = track->encodings.elem;
2075 uint32_t pkt_size = lace_size[n];
2076 uint8_t *pkt_data = data;
2078 if (encodings && encodings->scope & 1) {
2079 offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
2082 av_assert0(offset + pkt_size >= pkt_size);
2085 pkt = av_mallocz(sizeof(AVPacket));
2086 /* XXX: prevent data copy... */
2087 if (av_new_packet(pkt, pkt_size+offset) < 0) {
2089 res = AVERROR(ENOMEM);
2093 memcpy (pkt->data, encodings->compression.settings.data, offset);
2094 memcpy (pkt->data+offset, pkt_data, pkt_size);
2096 if (pkt_data != data)
2100 pkt->flags = is_keyframe;
2101 pkt->stream_index = st->index;
2103 if (track->ms_compat)
2104 pkt->dts = timecode;
2106 pkt->pts = timecode;
2108 if (st->codec->codec_id == AV_CODEC_ID_SUBRIP)
2109 pkt->convergence_duration = lace_duration;
2110 else if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE)
2111 pkt->duration = lace_duration;
2113 if (st->codec->codec_id == AV_CODEC_ID_SSA)
2114 matroska_fix_ass_packet(matroska, pkt, lace_duration);
2116 if (matroska->prev_pkt &&
2117 timecode != AV_NOPTS_VALUE &&
2118 matroska->prev_pkt->pts == timecode &&
2119 matroska->prev_pkt->stream_index == st->index &&
2120 st->codec->codec_id == AV_CODEC_ID_SSA)
2121 matroska_merge_packets(matroska->prev_pkt, pkt);
2123 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2124 matroska->prev_pkt = pkt;
2128 if (timecode != AV_NOPTS_VALUE)
2129 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2130 data += lace_size[n];
2131 size -= lace_size[n];
2140 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
2142 EbmlList *blocks_list;
2143 MatroskaBlock *blocks;
2145 res = ebml_parse(matroska,
2146 matroska_cluster_incremental_parsing,
2147 &matroska->current_cluster);
2150 if (matroska->current_cluster_pos)
2151 ebml_level_end(matroska);
2152 ebml_free(matroska_cluster, &matroska->current_cluster);
2153 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2154 matroska->current_cluster_num_blocks = 0;
2155 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2156 matroska->prev_pkt = NULL;
2157 /* sizeof the ID which was already read */
2158 if (matroska->current_id)
2159 matroska->current_cluster_pos -= 4;
2160 res = ebml_parse(matroska,
2161 matroska_clusters_incremental,
2162 &matroska->current_cluster);
2163 /* Try parsing the block again. */
2165 res = ebml_parse(matroska,
2166 matroska_cluster_incremental_parsing,
2167 &matroska->current_cluster);
2171 matroska->current_cluster_num_blocks <
2172 matroska->current_cluster.blocks.nb_elem) {
2173 blocks_list = &matroska->current_cluster.blocks;
2174 blocks = blocks_list->elem;
2176 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2177 i = blocks_list->nb_elem - 1;
2178 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2179 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2180 if (!blocks[i].non_simple)
2181 blocks[i].duration = 0;
2182 res = matroska_parse_block(matroska,
2183 blocks[i].bin.data, blocks[i].bin.size,
2185 matroska->current_cluster.timecode,
2186 blocks[i].duration, is_keyframe,
2187 matroska->current_cluster_pos);
2191 if (res < 0) matroska->done = 1;
2195 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
2197 MatroskaCluster cluster = { 0 };
2198 EbmlList *blocks_list;
2199 MatroskaBlock *blocks;
2202 if (!matroska->contains_ssa)
2203 return matroska_parse_cluster_incremental(matroska);
2204 pos = avio_tell(matroska->ctx->pb);
2205 matroska->prev_pkt = NULL;
2206 if (matroska->current_id)
2207 pos -= 4; /* sizeof the ID which was already read */
2208 res = ebml_parse(matroska, matroska_clusters, &cluster);
2209 blocks_list = &cluster.blocks;
2210 blocks = blocks_list->elem;
2211 for (i=0; i<blocks_list->nb_elem; i++)
2212 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2213 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2214 res=matroska_parse_block(matroska,
2215 blocks[i].bin.data, blocks[i].bin.size,
2216 blocks[i].bin.pos, cluster.timecode,
2217 blocks[i].duration, is_keyframe,
2220 ebml_free(matroska_cluster, &cluster);
2224 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
2226 MatroskaDemuxContext *matroska = s->priv_data;
2228 while (matroska_deliver_packet(matroska, pkt)) {
2229 int64_t pos = avio_tell(matroska->ctx->pb);
2232 if (matroska_parse_cluster(matroska) < 0)
2233 matroska_resync(matroska, pos);
2239 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2240 int64_t timestamp, int flags)
2242 MatroskaDemuxContext *matroska = s->priv_data;
2243 MatroskaTrack *tracks = matroska->tracks.elem;
2244 AVStream *st = s->streams[stream_index];
2245 int i, index, index_sub, index_min;
2247 /* Parse the CUES now since we need the index data to seek. */
2248 if (matroska->cues_parsing_deferred > 0) {
2249 matroska->cues_parsing_deferred = 0;
2250 matroska_parse_cues(matroska);
2253 if (!st->nb_index_entries)
2255 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2257 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2258 avio_seek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
2259 matroska->current_id = 0;
2260 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2261 matroska->prev_pkt = NULL;
2262 matroska_clear_queue(matroska);
2263 if (matroska_parse_cluster(matroska) < 0)
2268 matroska_clear_queue(matroska);
2269 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
2273 for (i=0; i < matroska->tracks.nb_elem; i++) {
2274 tracks[i].audio.pkt_cnt = 0;
2275 tracks[i].audio.sub_packet_cnt = 0;
2276 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2277 tracks[i].end_timecode = 0;
2278 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
2279 && !tracks[i].stream->discard != AVDISCARD_ALL) {
2280 index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
2282 && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
2283 && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
2284 index_min = index_sub;
2288 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2289 matroska->current_id = 0;
2290 st->skip_to_keyframe =
2291 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
2292 matroska->skip_to_timecode = st->index_entries[index].timestamp;
2294 matroska->num_levels = 0;
2295 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2298 // slightly hackish but allows proper fallback to
2299 // the generic seeking code.
2300 matroska_clear_queue(matroska);
2301 matroska->current_id = 0;
2302 st->skip_to_keyframe =
2303 matroska->skip_to_keyframe = 0;
2305 matroska->num_levels = 0;
2309 static int matroska_read_close(AVFormatContext *s)
2311 MatroskaDemuxContext *matroska = s->priv_data;
2312 MatroskaTrack *tracks = matroska->tracks.elem;
2315 matroska_clear_queue(matroska);
2317 for (n=0; n < matroska->tracks.nb_elem; n++)
2318 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2319 av_free(tracks[n].audio.buf);
2320 ebml_free(matroska_cluster, &matroska->current_cluster);
2321 ebml_free(matroska_segment, matroska);
2326 AVInputFormat ff_matroska_demuxer = {
2327 .name = "matroska,webm",
2328 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
2329 .priv_data_size = sizeof(MatroskaDemuxContext),
2330 .read_probe = matroska_probe,
2331 .read_header = matroska_read_header,
2332 .read_packet = matroska_read_packet,
2333 .read_close = matroska_read_close,
2334 .read_seek = matroska_read_seek,