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/bytestream.h"
41 #include "libavcodec/mpeg4audio.h"
42 #include "libavutil/base64.h"
43 #include "libavutil/intfloat.h"
44 #include "libavutil/intreadwrite.h"
45 #include "libavutil/avstring.h"
46 #include "libavutil/lzo.h"
47 #include "libavutil/dict.h"
69 typedef const struct EbmlSyntax {
78 const struct EbmlSyntax *n;
98 uint64_t doctype_version;
104 } MatroskaTrackCompression;
109 } MatroskaTrackEncryption;
114 MatroskaTrackCompression compression;
115 MatroskaTrackEncryption encryption;
116 } MatroskaTrackEncoding;
120 uint64_t display_width;
121 uint64_t display_height;
122 uint64_t pixel_width;
123 uint64_t pixel_height;
125 uint64_t stereo_mode;
127 } MatroskaTrackVideo;
131 double out_samplerate;
135 /* real audio header (extracted from extradata) */
142 uint64_t buf_timecode;
144 } MatroskaTrackAudio;
149 } MatroskaTrackPlane;
152 EbmlList combine_planes;
153 } MatroskaTrackOperation;
164 uint64_t default_duration;
165 uint64_t flag_default;
166 uint64_t flag_forced;
167 uint64_t codec_delay;
168 uint64_t seek_preroll;
169 MatroskaTrackVideo video;
170 MatroskaTrackAudio audio;
171 MatroskaTrackOperation operation;
175 int64_t end_timecode;
177 uint64_t max_block_additional_id;
187 } MatroskaAttachement;
225 MatroskaTagTarget target;
245 AVFormatContext *ctx;
249 MatroskaLevel levels[EBML_MAX_DEPTH];
258 EbmlList attachments;
264 /* byte position of the segment inside the stream */
265 int64_t segment_start;
267 /* the packet queue */
274 /* What to skip before effectively reading a packet. */
275 int skip_to_keyframe;
276 uint64_t skip_to_timecode;
278 /* File has a CUES element, but we defer parsing until it is needed. */
279 int cues_parsing_deferred;
281 int current_cluster_num_blocks;
282 int64_t current_cluster_pos;
283 MatroskaCluster current_cluster;
285 /* File has SSA subtitles which prevent incremental cluster parsing. */
287 } MatroskaDemuxContext;
294 uint64_t additional_id;
296 int64_t discard_padding;
299 static EbmlSyntax ebml_header[] = {
300 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
301 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
302 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
303 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
304 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
305 { EBML_ID_EBMLVERSION, EBML_NONE },
306 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
310 static EbmlSyntax ebml_syntax[] = {
311 { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
315 static EbmlSyntax matroska_info[] = {
316 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
317 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
318 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
319 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
320 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
321 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext,date_utc) },
322 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
326 static EbmlSyntax matroska_track_video[] = {
327 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
328 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width), {.u=-1} },
329 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height), {.u=-1} },
330 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
331 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
332 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo,color_space) },
333 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,stereo_mode) },
334 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,alpha_mode) },
335 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
336 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
337 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
338 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
339 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
340 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
341 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
345 static EbmlSyntax matroska_track_audio[] = {
346 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
347 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
348 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
349 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
353 static EbmlSyntax matroska_track_encoding_compression[] = {
354 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
355 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
359 static EbmlSyntax matroska_track_encoding_encryption[] = {
360 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u=0} },
361 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
362 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
363 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
364 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
365 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
366 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
369 static EbmlSyntax matroska_track_encoding[] = {
370 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
371 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
372 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
373 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding,encryption), {.n=matroska_track_encoding_encryption} },
374 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
378 static EbmlSyntax matroska_track_encodings[] = {
379 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
383 static EbmlSyntax matroska_track_plane[] = {
384 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
385 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
389 static EbmlSyntax matroska_track_combine_planes[] = {
390 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n=matroska_track_plane} },
394 static EbmlSyntax matroska_track_operation[] = {
395 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n=matroska_track_combine_planes} },
399 static EbmlSyntax matroska_track[] = {
400 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
401 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack,name) },
402 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
403 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
404 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
405 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
406 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
407 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
408 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
409 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
410 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack,flag_forced), {.u=0} },
411 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
412 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
413 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack,operation), {.n=matroska_track_operation} },
414 { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
415 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack,max_block_additional_id) },
416 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack,codec_delay) },
417 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack,seek_preroll) },
418 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
419 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
420 { MATROSKA_ID_CODECNAME, EBML_NONE },
421 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
422 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
423 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
424 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
425 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
429 static EbmlSyntax matroska_tracks[] = {
430 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
434 static EbmlSyntax matroska_attachment[] = {
435 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
436 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
437 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
438 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
439 { MATROSKA_ID_FILEDESC, EBML_NONE },
443 static EbmlSyntax matroska_attachments[] = {
444 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
448 static EbmlSyntax matroska_chapter_display[] = {
449 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
450 { MATROSKA_ID_CHAPLANG, EBML_NONE },
454 static EbmlSyntax matroska_chapter_entry[] = {
455 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
456 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
457 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
458 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
459 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
460 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
461 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
462 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
466 static EbmlSyntax matroska_chapter[] = {
467 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
468 { MATROSKA_ID_EDITIONUID, EBML_NONE },
469 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
470 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
471 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
475 static EbmlSyntax matroska_chapters[] = {
476 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
480 static EbmlSyntax matroska_index_pos[] = {
481 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
482 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
483 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
484 { MATROSKA_ID_CUEDURATION, EBML_NONE },
485 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
489 static EbmlSyntax matroska_index_entry[] = {
490 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
491 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
495 static EbmlSyntax matroska_index[] = {
496 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
500 static EbmlSyntax matroska_simpletag[] = {
501 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
502 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
503 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag,lang), {.s="und"} },
504 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag,def) },
505 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag,def) },
506 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
510 static EbmlSyntax matroska_tagtargets[] = {
511 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget,type) },
512 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
513 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
514 { MATROSKA_ID_TAGTARGETS_CHAPTERUID,EBML_UINT, 0, offsetof(MatroskaTagTarget,chapteruid) },
515 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
519 static EbmlSyntax matroska_tag[] = {
520 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
521 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
525 static EbmlSyntax matroska_tags[] = {
526 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
530 static EbmlSyntax matroska_seekhead_entry[] = {
531 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
532 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
536 static EbmlSyntax matroska_seekhead[] = {
537 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
541 static EbmlSyntax matroska_segment[] = {
542 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
543 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
544 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
545 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
546 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
547 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
548 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
549 { MATROSKA_ID_CLUSTER, EBML_STOP },
553 static EbmlSyntax matroska_segments[] = {
554 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
558 static EbmlSyntax matroska_blockmore[] = {
559 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
560 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
564 static EbmlSyntax matroska_blockadditions[] = {
565 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n=matroska_blockmore} },
569 static EbmlSyntax matroska_blockgroup[] = {
570 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
571 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, {.n=matroska_blockadditions} },
572 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
573 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration) },
574 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock,discard_padding) },
575 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock,reference) },
576 { MATROSKA_ID_CODECSTATE, EBML_NONE },
577 { 1, EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
581 static EbmlSyntax matroska_cluster[] = {
582 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
583 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
584 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
585 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
586 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
590 static EbmlSyntax matroska_clusters[] = {
591 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
592 { MATROSKA_ID_INFO, EBML_NONE },
593 { MATROSKA_ID_CUES, EBML_NONE },
594 { MATROSKA_ID_TAGS, EBML_NONE },
595 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
599 static EbmlSyntax matroska_cluster_incremental_parsing[] = {
600 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
601 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
602 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
603 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
604 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
605 { MATROSKA_ID_INFO, EBML_NONE },
606 { MATROSKA_ID_CUES, EBML_NONE },
607 { MATROSKA_ID_TAGS, EBML_NONE },
608 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
609 { MATROSKA_ID_CLUSTER, EBML_STOP },
613 static EbmlSyntax matroska_cluster_incremental[] = {
614 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
615 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
616 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
617 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
618 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
622 static EbmlSyntax matroska_clusters_incremental[] = {
623 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster_incremental} },
624 { MATROSKA_ID_INFO, EBML_NONE },
625 { MATROSKA_ID_CUES, EBML_NONE },
626 { MATROSKA_ID_TAGS, EBML_NONE },
627 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
631 static const char *const matroska_doctypes[] = { "matroska", "webm" };
633 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
635 AVIOContext *pb = matroska->ctx->pb;
637 matroska->current_id = 0;
638 matroska->num_levels = 0;
640 /* seek to next position to resync from */
641 if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
646 // try to find a toplevel element
647 while (!url_feof(pb)) {
648 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
649 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
650 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
651 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
652 matroska->current_id = id;
655 id = (id << 8) | avio_r8(pb);
663 * Return: Whether we reached the end of a level in the hierarchy or not.
665 static int ebml_level_end(MatroskaDemuxContext *matroska)
667 AVIOContext *pb = matroska->ctx->pb;
668 int64_t pos = avio_tell(pb);
670 if (matroska->num_levels > 0) {
671 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
672 if (pos - level->start >= level->length || matroska->current_id) {
673 matroska->num_levels--;
681 * Read: an "EBML number", which is defined as a variable-length
682 * array of bytes. The first byte indicates the length by giving a
683 * number of 0-bits followed by a one. The position of the first
684 * "one" bit inside the first byte indicates the length of this
686 * Returns: number of bytes read, < 0 on error
688 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
689 int max_size, uint64_t *number)
694 /* The first byte tells us the length in bytes - avio_r8() can normally
695 * return 0, but since that's not a valid first ebmlID byte, we can
696 * use it safely here to catch EOS. */
697 if (!(total = avio_r8(pb))) {
698 /* we might encounter EOS here */
700 int64_t pos = avio_tell(pb);
701 av_log(matroska->ctx, AV_LOG_ERROR,
702 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
704 return pb->error ? pb->error : AVERROR(EIO);
709 /* get the length of the EBML number */
710 read = 8 - ff_log2_tab[total];
711 if (read > max_size) {
712 int64_t pos = avio_tell(pb) - 1;
713 av_log(matroska->ctx, AV_LOG_ERROR,
714 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
715 (uint8_t) total, pos, pos);
716 return AVERROR_INVALIDDATA;
719 /* read out length */
720 total ^= 1 << ff_log2_tab[total];
722 total = (total << 8) | avio_r8(pb);
730 * Read a EBML length value.
731 * This needs special handling for the "unknown length" case which has multiple
734 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
737 int res = ebml_read_num(matroska, pb, 8, number);
738 if (res > 0 && *number + 1 == 1ULL << (7 * res))
739 *number = 0xffffffffffffffULL;
744 * Read the next element as an unsigned int.
745 * 0 is success, < 0 is failure.
747 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
752 return AVERROR_INVALIDDATA;
754 /* big-endian ordering; build up number */
757 *num = (*num << 8) | avio_r8(pb);
763 * Read the next element as a signed int.
764 * 0 is success, < 0 is failure.
766 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
771 return AVERROR_INVALIDDATA;
776 *num = sign_extend(avio_r8(pb), 8);
778 /* big-endian ordering; build up number */
780 *num = (*num << 8) | avio_r8(pb);
787 * Read the next element as a float.
788 * 0 is success, < 0 is failure.
790 static int ebml_read_float(AVIOContext *pb, int size, double *num)
794 } else if (size == 4) {
795 *num = av_int2float(avio_rb32(pb));
796 } else if (size == 8){
797 *num = av_int2double(avio_rb64(pb));
799 return AVERROR_INVALIDDATA;
805 * Read the next element as an ASCII string.
806 * 0 is success, < 0 is failure.
808 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
812 /* EBML strings are usually not 0-terminated, so we allocate one
813 * byte more, read the string and NULL-terminate it ourselves. */
814 if (!(res = av_malloc(size + 1)))
815 return AVERROR(ENOMEM);
816 if (avio_read(pb, (uint8_t *) res, size) != size) {
828 * Read the next element as binary data.
829 * 0 is success, < 0 is failure.
831 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
833 av_fast_padded_malloc(&bin->data, &bin->size, length);
835 return AVERROR(ENOMEM);
838 bin->pos = avio_tell(pb);
839 if (avio_read(pb, bin->data, length) != length) {
840 av_freep(&bin->data);
849 * Read the next element, but only the header. The contents
850 * are supposed to be sub-elements which can be read separately.
851 * 0 is success, < 0 is failure.
853 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
855 AVIOContext *pb = matroska->ctx->pb;
856 MatroskaLevel *level;
858 if (matroska->num_levels >= EBML_MAX_DEPTH) {
859 av_log(matroska->ctx, AV_LOG_ERROR,
860 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
861 return AVERROR(ENOSYS);
864 level = &matroska->levels[matroska->num_levels++];
865 level->start = avio_tell(pb);
866 level->length = length;
872 * Read signed/unsigned "EBML" numbers.
873 * Return: number of bytes processed, < 0 on error
875 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
876 uint8_t *data, uint32_t size, uint64_t *num)
879 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
880 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
884 * Same as above, but signed.
886 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
887 uint8_t *data, uint32_t size, int64_t *num)
892 /* read as unsigned number first */
893 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
896 /* make signed (weird way) */
897 *num = unum - ((1LL << (7*res - 1)) - 1);
902 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
903 EbmlSyntax *syntax, void *data);
905 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
906 uint32_t id, void *data)
909 for (i=0; syntax[i].id; i++)
910 if (id == syntax[i].id)
912 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
913 matroska->num_levels > 0 &&
914 matroska->levels[matroska->num_levels-1].length == 0xffffffffffffff)
915 return 0; // we reached the end of an unknown size cluster
916 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
917 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
918 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
919 return AVERROR_INVALIDDATA;
921 return ebml_parse_elem(matroska, &syntax[i], data);
924 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
927 if (!matroska->current_id) {
929 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
932 matroska->current_id = id | 1 << 7*res;
934 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
937 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
942 for (i=0; syntax[i].id; i++)
943 switch (syntax[i].type) {
945 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
948 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
952 // the default may be NULL
953 if (syntax[i].def.s) {
954 uint8_t **dst = (uint8_t**)((uint8_t*)data + syntax[i].data_offset);
955 *dst = av_strdup(syntax[i].def.s);
957 return AVERROR(ENOMEM);
962 while (!res && !ebml_level_end(matroska))
963 res = ebml_parse(matroska, syntax, data);
968 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
969 EbmlSyntax *syntax, void *data)
971 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
974 // max. 16 MB for strings
975 [EBML_STR] = 0x1000000,
976 [EBML_UTF8] = 0x1000000,
977 // max. 256 MB for binary data
978 [EBML_BIN] = 0x10000000,
979 // no limits for anything else
981 AVIOContext *pb = matroska->ctx->pb;
982 uint32_t id = syntax->id;
987 data = (char *)data + syntax->data_offset;
988 if (syntax->list_elem_size) {
989 EbmlList *list = data;
990 newelem = av_realloc_array(list->elem, list->nb_elem+1, syntax->list_elem_size);
992 return AVERROR(ENOMEM);
993 list->elem = newelem;
994 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
995 memset(data, 0, syntax->list_elem_size);
999 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
1000 matroska->current_id = 0;
1001 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1003 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1004 av_log(matroska->ctx, AV_LOG_ERROR,
1005 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1006 length, max_lengths[syntax->type], syntax->type);
1007 return AVERROR_INVALIDDATA;
1011 switch (syntax->type) {
1012 case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
1013 case EBML_SINT: res = ebml_read_sint (pb, length, data); break;
1014 case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
1016 case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
1017 case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
1018 case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
1020 if (id == MATROSKA_ID_SEGMENT)
1021 matroska->segment_start = avio_tell(matroska->ctx->pb);
1022 return ebml_parse_nest(matroska, syntax->def.n, data);
1023 case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
1024 case EBML_STOP: return 1;
1026 if(ffio_limit(pb, length) != length)
1027 return AVERROR(EIO);
1028 return avio_skip(pb,length)<0 ? AVERROR(EIO) : 0;
1030 if (res == AVERROR_INVALIDDATA)
1031 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1032 else if (res == AVERROR(EIO))
1033 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1037 static void ebml_free(EbmlSyntax *syntax, void *data)
1040 for (i=0; syntax[i].id; i++) {
1041 void *data_off = (char *)data + syntax[i].data_offset;
1042 switch (syntax[i].type) {
1044 case EBML_UTF8: av_freep(data_off); break;
1045 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
1047 if (syntax[i].list_elem_size) {
1048 EbmlList *list = data_off;
1049 char *ptr = list->elem;
1050 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
1051 ebml_free(syntax[i].def.n, ptr);
1052 av_free(list->elem);
1054 ebml_free(syntax[i].def.n, data_off);
1064 static int matroska_probe(AVProbeData *p)
1067 int len_mask = 0x80, size = 1, n = 1, i;
1070 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1073 /* length of header */
1075 while (size <= 8 && !(total & len_mask)) {
1081 total &= (len_mask - 1);
1083 total = (total << 8) | p->buf[4 + n++];
1085 /* Does the probe data contain the whole header? */
1086 if (p->buf_size < 4 + size + total)
1089 /* The header should contain a known document type. For now,
1090 * we don't parse the whole header but simply check for the
1091 * availability of that array of characters inside the header.
1092 * Not fully fool-proof, but good enough. */
1093 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1094 int probelen = strlen(matroska_doctypes[i]);
1095 if (total < probelen)
1097 for (n = 4+size; n <= 4+size+total-probelen; n++)
1098 if (!memcmp(p->buf+n, matroska_doctypes[i], probelen))
1099 return AVPROBE_SCORE_MAX;
1102 // probably valid EBML header but no recognized doctype
1103 return AVPROBE_SCORE_EXTENSION;
1106 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1109 MatroskaTrack *tracks = matroska->tracks.elem;
1112 for (i=0; i < matroska->tracks.nb_elem; i++)
1113 if (tracks[i].num == num)
1116 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1120 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
1121 MatroskaTrack *track)
1123 MatroskaTrackEncoding *encodings = track->encodings.elem;
1124 uint8_t* data = *buf;
1125 int isize = *buf_size;
1126 uint8_t* pkt_data = NULL;
1127 uint8_t av_unused *newpktdata;
1128 int pkt_size = isize;
1132 if (pkt_size >= 10000000U)
1133 return AVERROR_INVALIDDATA;
1135 switch (encodings[0].compression.algo) {
1136 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP: {
1137 int header_size = encodings[0].compression.settings.size;
1138 uint8_t *header = encodings[0].compression.settings.data;
1140 if (header_size && !header) {
1141 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1148 pkt_size = isize + header_size;
1149 pkt_data = av_malloc(pkt_size);
1151 return AVERROR(ENOMEM);
1153 memcpy(pkt_data, header, header_size);
1154 memcpy(pkt_data + header_size, data, isize);
1158 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1160 olen = pkt_size *= 3;
1161 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1163 result = AVERROR(ENOMEM);
1166 pkt_data = newpktdata;
1167 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1168 } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
1170 result = AVERROR_INVALIDDATA;
1177 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
1178 z_stream zstream = {0};
1179 if (inflateInit(&zstream) != Z_OK)
1181 zstream.next_in = data;
1182 zstream.avail_in = isize;
1185 newpktdata = av_realloc(pkt_data, pkt_size);
1187 inflateEnd(&zstream);
1190 pkt_data = newpktdata;
1191 zstream.avail_out = pkt_size - zstream.total_out;
1192 zstream.next_out = pkt_data + zstream.total_out;
1194 result = inflate(&zstream, Z_NO_FLUSH);
1196 result = Z_MEM_ERROR;
1197 } while (result==Z_OK && pkt_size<10000000);
1198 pkt_size = zstream.total_out;
1199 inflateEnd(&zstream);
1200 if (result != Z_STREAM_END) {
1201 if (result == Z_MEM_ERROR)
1202 result = AVERROR(ENOMEM);
1204 result = AVERROR_INVALIDDATA;
1211 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
1212 bz_stream bzstream = {0};
1213 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1215 bzstream.next_in = data;
1216 bzstream.avail_in = isize;
1219 newpktdata = av_realloc(pkt_data, pkt_size);
1221 BZ2_bzDecompressEnd(&bzstream);
1224 pkt_data = newpktdata;
1225 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1226 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1228 result = BZ2_bzDecompress(&bzstream);
1230 result = BZ_MEM_ERROR;
1231 } while (result==BZ_OK && pkt_size<10000000);
1232 pkt_size = bzstream.total_out_lo32;
1233 BZ2_bzDecompressEnd(&bzstream);
1234 if (result != BZ_STREAM_END) {
1235 if (result == BZ_MEM_ERROR)
1236 result = AVERROR(ENOMEM);
1238 result = AVERROR_INVALIDDATA;
1245 return AVERROR_INVALIDDATA;
1249 *buf_size = pkt_size;
1257 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
1258 AVPacket *pkt, uint64_t display_duration)
1261 char *layer, *ptr = pkt->data, *end = ptr+pkt->size;
1262 for (; *ptr!=',' && ptr<end-1; ptr++);
1266 for (; *ptr!=',' && ptr<end-1; ptr++);
1268 int64_t end_pts = pkt->pts + display_duration;
1269 int sc = matroska->time_scale * pkt->pts / 10000000;
1270 int ec = matroska->time_scale * end_pts / 10000000;
1271 int sh, sm, ss, eh, em, es, len;
1272 sh = sc/360000; sc -= 360000*sh;
1273 sm = sc/ 6000; sc -= 6000*sm;
1274 ss = sc/ 100; sc -= 100*ss;
1275 eh = ec/360000; ec -= 360000*eh;
1276 em = ec/ 6000; ec -= 6000*em;
1277 es = ec/ 100; ec -= 100*es;
1279 len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
1280 if (!(line = av_buffer_alloc(len)))
1282 snprintf(line->data, len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
1283 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
1284 av_buffer_unref(&pkt->buf);
1286 pkt->data = line->data;
1287 pkt->size = strlen(line->data);
1291 static int matroska_merge_packets(AVPacket *out, AVPacket *in)
1293 int ret = av_grow_packet(out, in->size);
1297 memcpy(out->data + out->size - in->size, in->data, in->size);
1305 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1306 AVDictionary **metadata, char *prefix)
1308 MatroskaTag *tags = list->elem;
1312 for (i=0; i < list->nb_elem; i++) {
1313 const char *lang = tags[i].lang && strcmp(tags[i].lang, "und") ?
1314 tags[i].lang : NULL;
1316 if (!tags[i].name) {
1317 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1320 if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1321 else av_strlcpy(key, tags[i].name, sizeof(key));
1322 if (tags[i].def || !lang) {
1323 av_dict_set(metadata, key, tags[i].string, 0);
1324 if (tags[i].sub.nb_elem)
1325 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1328 av_strlcat(key, "-", sizeof(key));
1329 av_strlcat(key, lang, sizeof(key));
1330 av_dict_set(metadata, key, tags[i].string, 0);
1331 if (tags[i].sub.nb_elem)
1332 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1335 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1338 static void matroska_convert_tags(AVFormatContext *s)
1340 MatroskaDemuxContext *matroska = s->priv_data;
1341 MatroskaTags *tags = matroska->tags.elem;
1344 for (i=0; i < matroska->tags.nb_elem; i++) {
1345 if (tags[i].target.attachuid) {
1346 MatroskaAttachement *attachment = matroska->attachments.elem;
1347 for (j=0; j<matroska->attachments.nb_elem; j++)
1348 if (attachment[j].uid == tags[i].target.attachuid
1349 && attachment[j].stream)
1350 matroska_convert_tag(s, &tags[i].tag,
1351 &attachment[j].stream->metadata, NULL);
1352 } else if (tags[i].target.chapteruid) {
1353 MatroskaChapter *chapter = matroska->chapters.elem;
1354 for (j=0; j<matroska->chapters.nb_elem; j++)
1355 if (chapter[j].uid == tags[i].target.chapteruid
1356 && chapter[j].chapter)
1357 matroska_convert_tag(s, &tags[i].tag,
1358 &chapter[j].chapter->metadata, NULL);
1359 } else if (tags[i].target.trackuid) {
1360 MatroskaTrack *track = matroska->tracks.elem;
1361 for (j=0; j<matroska->tracks.nb_elem; j++)
1362 if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1363 matroska_convert_tag(s, &tags[i].tag,
1364 &track[j].stream->metadata, NULL);
1366 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1367 tags[i].target.type);
1372 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska, int idx)
1374 EbmlList *seekhead_list = &matroska->seekhead;
1375 MatroskaSeekhead *seekhead = seekhead_list->elem;
1376 uint32_t level_up = matroska->level_up;
1377 int64_t before_pos = avio_tell(matroska->ctx->pb);
1378 uint32_t saved_id = matroska->current_id;
1379 MatroskaLevel level;
1383 if (idx >= seekhead_list->nb_elem
1384 || seekhead[idx].id == MATROSKA_ID_SEEKHEAD
1385 || seekhead[idx].id == MATROSKA_ID_CLUSTER)
1389 offset = seekhead[idx].pos + matroska->segment_start;
1390 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1391 /* We don't want to lose our seekhead level, so we add
1392 * a dummy. This is a crude hack. */
1393 if (matroska->num_levels == EBML_MAX_DEPTH) {
1394 av_log(matroska->ctx, AV_LOG_INFO,
1395 "Max EBML element depth (%d) reached, "
1396 "cannot parse further.\n", EBML_MAX_DEPTH);
1397 ret = AVERROR_INVALIDDATA;
1400 level.length = (uint64_t)-1;
1401 matroska->levels[matroska->num_levels] = level;
1402 matroska->num_levels++;
1403 matroska->current_id = 0;
1405 ret = ebml_parse(matroska, matroska_segment, matroska);
1407 /* remove dummy level */
1408 while (matroska->num_levels) {
1409 uint64_t length = matroska->levels[--matroska->num_levels].length;
1410 if (length == (uint64_t)-1)
1416 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1417 matroska->level_up = level_up;
1418 matroska->current_id = saved_id;
1423 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1425 EbmlList *seekhead_list = &matroska->seekhead;
1426 int64_t before_pos = avio_tell(matroska->ctx->pb);
1429 // we should not do any seeking in the streaming case
1430 if (!matroska->ctx->pb->seekable ||
1431 (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1434 for (i = 0; i < seekhead_list->nb_elem; i++) {
1435 MatroskaSeekhead *seekhead = seekhead_list->elem;
1436 if (seekhead[i].pos <= before_pos)
1439 // defer cues parsing until we actually need cue data.
1440 if (seekhead[i].id == MATROSKA_ID_CUES) {
1441 matroska->cues_parsing_deferred = 1;
1445 if (matroska_parse_seekhead_entry(matroska, i) < 0) {
1446 // mark index as broken
1447 matroska->cues_parsing_deferred = -1;
1453 static void matroska_add_index_entries(MatroskaDemuxContext *matroska) {
1454 EbmlList *index_list;
1455 MatroskaIndex *index;
1456 int index_scale = 1;
1459 index_list = &matroska->index;
1460 index = index_list->elem;
1461 if (index_list->nb_elem
1462 && index[0].time > 1E14/matroska->time_scale) {
1463 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1464 index_scale = matroska->time_scale;
1466 for (i = 0; i < index_list->nb_elem; i++) {
1467 EbmlList *pos_list = &index[i].pos;
1468 MatroskaIndexPos *pos = pos_list->elem;
1469 for (j = 0; j < pos_list->nb_elem; j++) {
1470 MatroskaTrack *track = matroska_find_track_by_num(matroska, pos[j].track);
1471 if (track && track->stream)
1472 av_add_index_entry(track->stream,
1473 pos[j].pos + matroska->segment_start,
1474 index[i].time/index_scale, 0, 0,
1480 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1481 EbmlList *seekhead_list = &matroska->seekhead;
1482 MatroskaSeekhead *seekhead = seekhead_list->elem;
1485 for (i = 0; i < seekhead_list->nb_elem; i++)
1486 if (seekhead[i].id == MATROSKA_ID_CUES)
1488 av_assert1(i <= seekhead_list->nb_elem);
1490 if (matroska_parse_seekhead_entry(matroska, i) < 0)
1491 matroska->cues_parsing_deferred = -1;
1492 matroska_add_index_entries(matroska);
1495 static int matroska_aac_profile(char *codec_id)
1497 static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1500 for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1501 if (strstr(codec_id, aac_profiles[profile]))
1506 static int matroska_aac_sri(int samplerate)
1510 for (sri=0; sri<FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1511 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1516 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1519 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1520 time_t creation_time = date_utc / 1000000000 + 978307200;
1521 struct tm *ptm = gmtime(&creation_time);
1523 strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm);
1524 av_dict_set(metadata, "creation_time", buffer, 0);
1527 static int matroska_read_header(AVFormatContext *s)
1529 MatroskaDemuxContext *matroska = s->priv_data;
1530 EbmlList *attachements_list = &matroska->attachments;
1531 MatroskaAttachement *attachements;
1532 EbmlList *chapters_list = &matroska->chapters;
1533 MatroskaChapter *chapters;
1534 MatroskaTrack *tracks;
1535 uint64_t max_start = 0;
1543 /* First read the EBML header. */
1544 if (ebml_parse(matroska, ebml_syntax, &ebml)
1545 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1546 || ebml.id_length > sizeof(uint32_t) || ebml.doctype_version > 3 || !ebml.doctype) {
1547 av_log(matroska->ctx, AV_LOG_ERROR,
1548 "EBML header using unsupported features\n"
1549 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1550 ebml.version, ebml.doctype, ebml.doctype_version);
1551 ebml_free(ebml_syntax, &ebml);
1552 return AVERROR_PATCHWELCOME;
1553 } else if (ebml.doctype_version == 3) {
1554 av_log(matroska->ctx, AV_LOG_WARNING,
1555 "EBML header using unsupported features\n"
1556 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1557 ebml.version, ebml.doctype, ebml.doctype_version);
1559 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
1560 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
1562 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
1563 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
1564 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
1565 ebml_free(ebml_syntax, &ebml);
1566 return AVERROR_INVALIDDATA;
1569 ebml_free(ebml_syntax, &ebml);
1571 /* The next thing is a segment. */
1572 pos = avio_tell(matroska->ctx->pb);
1573 res = ebml_parse(matroska, matroska_segments, matroska);
1574 // try resyncing until we find a EBML_STOP type element.
1576 res = matroska_resync(matroska, pos);
1579 pos = avio_tell(matroska->ctx->pb);
1580 res = ebml_parse(matroska, matroska_segment, matroska);
1582 matroska_execute_seekhead(matroska);
1584 if (!matroska->time_scale)
1585 matroska->time_scale = 1000000;
1586 if (matroska->duration)
1587 matroska->ctx->duration = matroska->duration * matroska->time_scale
1588 * 1000 / AV_TIME_BASE;
1589 av_dict_set(&s->metadata, "title", matroska->title, 0);
1591 if (matroska->date_utc.size == 8)
1592 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
1594 tracks = matroska->tracks.elem;
1595 for (i=0; i < matroska->tracks.nb_elem; i++) {
1596 MatroskaTrack *track = &tracks[i];
1597 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1598 EbmlList *encodings_list = &track->encodings;
1599 MatroskaTrackEncoding *encodings = encodings_list->elem;
1600 uint8_t *extradata = NULL;
1601 int extradata_size = 0;
1602 int extradata_offset = 0;
1603 uint32_t fourcc = 0;
1605 char* key_id_base64 = NULL;
1607 /* Apply some sanity checks. */
1608 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1609 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1610 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
1611 track->type != MATROSKA_TRACK_TYPE_METADATA) {
1612 av_log(matroska->ctx, AV_LOG_INFO,
1613 "Unknown or unsupported track type %"PRIu64"\n",
1617 if (track->codec_id == NULL)
1620 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1621 if (!track->default_duration && track->video.frame_rate > 0)
1622 track->default_duration = 1000000000/track->video.frame_rate;
1623 if (track->video.display_width == -1)
1624 track->video.display_width = track->video.pixel_width;
1625 if (track->video.display_height == -1)
1626 track->video.display_height = track->video.pixel_height;
1627 if (track->video.color_space.size == 4)
1628 fourcc = AV_RL32(track->video.color_space.data);
1629 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1630 if (!track->audio.out_samplerate)
1631 track->audio.out_samplerate = track->audio.samplerate;
1633 if (encodings_list->nb_elem > 1) {
1634 av_log(matroska->ctx, AV_LOG_ERROR,
1635 "Multiple combined encodings not supported");
1636 } else if (encodings_list->nb_elem == 1) {
1637 if (encodings[0].type) {
1638 if (encodings[0].encryption.key_id.size > 0) {
1639 /* Save the encryption key id to be stored later as a
1641 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
1642 key_id_base64 = av_malloc(b64_size);
1643 if (key_id_base64 == NULL)
1644 return AVERROR(ENOMEM);
1646 av_base64_encode(key_id_base64, b64_size,
1647 encodings[0].encryption.key_id.data,
1648 encodings[0].encryption.key_id.size);
1650 encodings[0].scope = 0;
1651 av_log(matroska->ctx, AV_LOG_ERROR,
1652 "Unsupported encoding type");
1656 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1659 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1662 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
1664 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
1665 encodings[0].scope = 0;
1666 av_log(matroska->ctx, AV_LOG_ERROR,
1667 "Unsupported encoding type");
1668 } else if (track->codec_priv.size && encodings[0].scope&2) {
1669 uint8_t *codec_priv = track->codec_priv.data;
1670 int ret = matroska_decode_buffer(&track->codec_priv.data,
1671 &track->codec_priv.size,
1674 track->codec_priv.data = NULL;
1675 track->codec_priv.size = 0;
1676 av_log(matroska->ctx, AV_LOG_ERROR,
1677 "Failed to decode codec private data\n");
1680 if (codec_priv != track->codec_priv.data)
1681 av_free(codec_priv);
1685 for(j=0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++){
1686 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1687 strlen(ff_mkv_codec_tags[j].str))){
1688 codec_id= ff_mkv_codec_tags[j].id;
1693 st = track->stream = avformat_new_stream(s, NULL);
1695 av_free(key_id_base64);
1696 return AVERROR(ENOMEM);
1699 if (key_id_base64) {
1700 /* export encryption key id as base64 metadata tag */
1701 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
1702 av_freep(&key_id_base64);
1705 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1706 && track->codec_priv.size >= 40
1707 && track->codec_priv.data != NULL) {
1708 track->ms_compat = 1;
1709 fourcc = AV_RL32(track->codec_priv.data + 16);
1710 codec_id = ff_codec_get_id(ff_codec_bmp_tags, fourcc);
1711 extradata_offset = 40;
1712 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1713 && track->codec_priv.size >= 14
1714 && track->codec_priv.data != NULL) {
1716 ffio_init_context(&b, track->codec_priv.data, track->codec_priv.size,
1717 0, NULL, NULL, NULL, NULL);
1718 ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1721 codec_id = st->codec->codec_id;
1722 extradata_offset = FFMIN(track->codec_priv.size, 18);
1723 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1724 && (track->codec_priv.size >= 86)
1725 && (track->codec_priv.data != NULL)) {
1726 fourcc = AV_RL32(track->codec_priv.data);
1727 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1728 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1729 switch (track->audio.bitdepth) {
1730 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1731 case 24: codec_id = AV_CODEC_ID_PCM_S24BE; break;
1732 case 32: codec_id = AV_CODEC_ID_PCM_S32BE; break;
1734 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1735 switch (track->audio.bitdepth) {
1736 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1737 case 24: codec_id = AV_CODEC_ID_PCM_S24LE; break;
1738 case 32: codec_id = AV_CODEC_ID_PCM_S32LE; break;
1740 } else if (codec_id==AV_CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1741 codec_id = AV_CODEC_ID_PCM_F64LE;
1742 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1743 int profile = matroska_aac_profile(track->codec_id);
1744 int sri = matroska_aac_sri(track->audio.samplerate);
1745 extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1746 if (extradata == NULL)
1747 return AVERROR(ENOMEM);
1748 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1749 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1750 if (strstr(track->codec_id, "SBR")) {
1751 sri = matroska_aac_sri(track->audio.out_samplerate);
1752 extradata[2] = 0x56;
1753 extradata[3] = 0xE5;
1754 extradata[4] = 0x80 | (sri<<3);
1758 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - FF_INPUT_BUFFER_PADDING_SIZE) {
1759 /* Only ALAC's magic cookie is stored in Matroska's track headers.
1760 Create the "atom size", "tag", and "tag version" fields the
1761 decoder expects manually. */
1762 extradata_size = 12 + track->codec_priv.size;
1763 extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1764 if (extradata == NULL)
1765 return AVERROR(ENOMEM);
1766 AV_WB32(extradata, extradata_size);
1767 memcpy(&extradata[4], "alac", 4);
1768 AV_WB32(&extradata[8], 0);
1769 memcpy(&extradata[12], track->codec_priv.data,
1770 track->codec_priv.size);
1771 } else if (codec_id == AV_CODEC_ID_TTA) {
1772 extradata_size = 30;
1773 extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1774 if (extradata == NULL)
1775 return AVERROR(ENOMEM);
1776 ffio_init_context(&b, extradata, extradata_size, 1,
1777 NULL, NULL, NULL, NULL);
1778 avio_write(&b, "TTA1", 4);
1780 avio_wl16(&b, track->audio.channels);
1781 avio_wl16(&b, track->audio.bitdepth);
1782 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
1783 return AVERROR_INVALIDDATA;
1784 avio_wl32(&b, track->audio.out_samplerate);
1785 avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale), track->audio.out_samplerate, AV_TIME_BASE * 1000));
1786 } else if (codec_id == AV_CODEC_ID_RV10 || codec_id == AV_CODEC_ID_RV20 ||
1787 codec_id == AV_CODEC_ID_RV30 || codec_id == AV_CODEC_ID_RV40) {
1788 extradata_offset = 26;
1789 } else if (codec_id == AV_CODEC_ID_RA_144) {
1790 track->audio.out_samplerate = 8000;
1791 track->audio.channels = 1;
1792 } else if ((codec_id == AV_CODEC_ID_RA_288 || codec_id == AV_CODEC_ID_COOK ||
1793 codec_id == AV_CODEC_ID_ATRAC3 || codec_id == AV_CODEC_ID_SIPR)
1794 && track->codec_priv.data) {
1797 ffio_init_context(&b, track->codec_priv.data,track->codec_priv.size,
1798 0, NULL, NULL, NULL, NULL);
1800 flavor = avio_rb16(&b);
1801 track->audio.coded_framesize = avio_rb32(&b);
1803 track->audio.sub_packet_h = avio_rb16(&b);
1804 track->audio.frame_size = avio_rb16(&b);
1805 track->audio.sub_packet_size = avio_rb16(&b);
1806 if (flavor <= 0 || track->audio.coded_framesize <= 0 ||
1807 track->audio.sub_packet_h <= 0 || track->audio.frame_size <= 0 ||
1808 track->audio.sub_packet_size <= 0)
1809 return AVERROR_INVALIDDATA;
1810 track->audio.buf = av_malloc_array(track->audio.sub_packet_h, track->audio.frame_size);
1811 if (!track->audio.buf)
1812 return AVERROR(ENOMEM);
1813 if (codec_id == AV_CODEC_ID_RA_288) {
1814 st->codec->block_align = track->audio.coded_framesize;
1815 track->codec_priv.size = 0;
1817 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1818 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1819 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1820 st->codec->bit_rate = sipr_bit_rate[flavor];
1822 st->codec->block_align = track->audio.sub_packet_size;
1823 extradata_offset = 78;
1826 track->codec_priv.size -= extradata_offset;
1828 if (codec_id == AV_CODEC_ID_NONE)
1829 av_log(matroska->ctx, AV_LOG_INFO,
1830 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1832 if (track->time_scale < 0.01)
1833 track->time_scale = 1.0;
1834 avpriv_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1836 st->codec->codec_id = codec_id;
1838 if (strcmp(track->language, "und"))
1839 av_dict_set(&st->metadata, "language", track->language, 0);
1840 av_dict_set(&st->metadata, "title", track->name, 0);
1842 if (track->flag_default)
1843 st->disposition |= AV_DISPOSITION_DEFAULT;
1844 if (track->flag_forced)
1845 st->disposition |= AV_DISPOSITION_FORCED;
1847 if (!st->codec->extradata) {
1849 st->codec->extradata = extradata;
1850 st->codec->extradata_size = extradata_size;
1851 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1852 if (ff_alloc_extradata(st->codec, track->codec_priv.size))
1853 return AVERROR(ENOMEM);
1854 memcpy(st->codec->extradata,
1855 track->codec_priv.data + extradata_offset,
1856 track->codec_priv.size);
1860 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1861 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
1863 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1864 st->codec->codec_tag = fourcc;
1865 st->codec->width = track->video.pixel_width;
1866 st->codec->height = track->video.pixel_height;
1867 av_reduce(&st->sample_aspect_ratio.num,
1868 &st->sample_aspect_ratio.den,
1869 st->codec->height * track->video.display_width,
1870 st->codec-> width * track->video.display_height,
1872 if (st->codec->codec_id != AV_CODEC_ID_HEVC)
1873 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1875 if (track->default_duration) {
1876 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1877 1000000000, track->default_duration, 30000);
1878 #if FF_API_R_FRAME_RATE
1879 if (st->avg_frame_rate.num < st->avg_frame_rate.den * 1000L)
1880 st->r_frame_rate = st->avg_frame_rate;
1884 /* export stereo mode flag as metadata tag */
1885 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREO_MODE_COUNT)
1886 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
1888 /* export alpha mode flag as metadata tag */
1889 if (track->video.alpha_mode)
1890 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
1892 /* if we have virtual track, mark the real tracks */
1893 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
1895 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
1897 snprintf(buf, sizeof(buf), "%s_%d",
1898 ff_matroska_video_stereo_plane[planes[j].type], i);
1899 for (k=0; k < matroska->tracks.nb_elem; k++)
1900 if (planes[j].uid == tracks[k].uid) {
1901 av_dict_set(&s->streams[k]->metadata,
1902 "stereo_mode", buf, 0);
1906 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1907 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1908 st->codec->sample_rate = track->audio.out_samplerate;
1909 st->codec->channels = track->audio.channels;
1910 st->codec->bits_per_coded_sample = track->audio.bitdepth;
1911 if (st->codec->codec_id != AV_CODEC_ID_AAC)
1912 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1913 if (track->codec_delay > 0) {
1914 st->codec->delay = av_rescale_q(track->codec_delay,
1915 (AVRational){1, 1000000000},
1916 (AVRational){1, st->codec->sample_rate});
1918 if (track->seek_preroll > 0) {
1919 av_codec_set_seek_preroll(st->codec,
1920 av_rescale_q(track->seek_preroll,
1921 (AVRational){1, 1000000000},
1922 (AVRational){1, st->codec->sample_rate}));
1924 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
1925 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1927 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
1928 st->disposition |= AV_DISPOSITION_CAPTIONS;
1929 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
1930 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
1931 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
1932 st->disposition |= AV_DISPOSITION_METADATA;
1934 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1935 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1937 if (st->codec->codec_id == AV_CODEC_ID_SSA ||
1938 st->codec->codec_id == AV_CODEC_ID_ASS)
1940 if (st->codec->codec_id == AV_CODEC_ID_ASS)
1942 matroska->contains_ssa = 1;
1946 attachements = attachements_list->elem;
1947 for (j=0; j<attachements_list->nb_elem; j++) {
1948 if (!(attachements[j].filename && attachements[j].mime &&
1949 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1950 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1952 AVStream *st = avformat_new_stream(s, NULL);
1955 av_dict_set(&st->metadata, "filename",attachements[j].filename, 0);
1956 av_dict_set(&st->metadata, "mimetype", attachements[j].mime, 0);
1957 st->codec->codec_id = AV_CODEC_ID_NONE;
1958 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
1959 if (ff_alloc_extradata(st->codec, attachements[j].bin.size))
1961 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1963 for (i=0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
1964 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1965 strlen(ff_mkv_mime_tags[i].str))) {
1966 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1970 attachements[j].stream = st;
1974 chapters = chapters_list->elem;
1975 for (i=0; i<chapters_list->nb_elem; i++)
1976 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1977 && (max_start==0 || chapters[i].start > max_start)) {
1978 chapters[i].chapter =
1979 avpriv_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1980 chapters[i].start, chapters[i].end,
1982 av_dict_set(&chapters[i].chapter->metadata,
1983 "title", chapters[i].title, 0);
1984 max_start = chapters[i].start;
1987 matroska_add_index_entries(matroska);
1989 matroska_convert_tags(s);
1995 * Put one packet in an application-supplied AVPacket struct.
1996 * Returns 0 on success or -1 on failure.
1998 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2001 if (matroska->num_packets > 0) {
2002 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2003 av_free(matroska->packets[0]);
2004 if (matroska->num_packets > 1) {
2006 memmove(&matroska->packets[0], &matroska->packets[1],
2007 (matroska->num_packets - 1) * sizeof(AVPacket *));
2008 newpackets = av_realloc(matroska->packets,
2009 (matroska->num_packets - 1) * sizeof(AVPacket *));
2011 matroska->packets = newpackets;
2013 av_freep(&matroska->packets);
2014 matroska->prev_pkt = NULL;
2016 matroska->num_packets--;
2024 * Free all packets in our internal queue.
2026 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2028 matroska->prev_pkt = NULL;
2029 if (matroska->packets) {
2031 for (n = 0; n < matroska->num_packets; n++) {
2032 av_free_packet(matroska->packets[n]);
2033 av_free(matroska->packets[n]);
2035 av_freep(&matroska->packets);
2036 matroska->num_packets = 0;
2040 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2041 int* buf_size, int type,
2042 uint32_t **lace_buf, int *laces)
2044 int res = 0, n, size = *buf_size;
2045 uint8_t *data = *buf;
2046 uint32_t *lace_size;
2050 *lace_buf = av_mallocz(sizeof(int));
2052 return AVERROR(ENOMEM);
2054 *lace_buf[0] = size;
2058 av_assert0(size > 0);
2062 lace_size = av_mallocz(*laces * sizeof(int));
2064 return AVERROR(ENOMEM);
2067 case 0x1: /* Xiph lacing */ {
2070 for (n = 0; res == 0 && n < *laces - 1; n++) {
2072 if (size <= total) {
2073 res = AVERROR_INVALIDDATA;
2078 lace_size[n] += temp;
2085 if (size <= total) {
2086 res = AVERROR_INVALIDDATA;
2090 lace_size[n] = size - total;
2094 case 0x2: /* fixed-size lacing */
2095 if (size % (*laces)) {
2096 res = AVERROR_INVALIDDATA;
2099 for (n = 0; n < *laces; n++)
2100 lace_size[n] = size / *laces;
2103 case 0x3: /* EBML lacing */ {
2106 n = matroska_ebmlnum_uint(matroska, data, size, &num);
2107 if (n < 0 || num > INT_MAX) {
2108 av_log(matroska->ctx, AV_LOG_INFO,
2109 "EBML block data error\n");
2110 res = n<0 ? n : AVERROR_INVALIDDATA;
2115 total = lace_size[0] = num;
2116 for (n = 1; res == 0 && n < *laces - 1; n++) {
2119 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2120 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
2121 av_log(matroska->ctx, AV_LOG_INFO,
2122 "EBML block data error\n");
2123 res = r<0 ? r : AVERROR_INVALIDDATA;
2128 lace_size[n] = lace_size[n - 1] + snum;
2129 total += lace_size[n];
2131 if (size <= total) {
2132 res = AVERROR_INVALIDDATA;
2135 lace_size[*laces - 1] = size - total;
2141 *lace_buf = lace_size;
2147 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2148 MatroskaTrack *track,
2150 uint8_t *data, int size,
2154 int a = st->codec->block_align;
2155 int sps = track->audio.sub_packet_size;
2156 int cfs = track->audio.coded_framesize;
2157 int h = track->audio.sub_packet_h;
2158 int y = track->audio.sub_packet_cnt;
2159 int w = track->audio.frame_size;
2162 if (!track->audio.pkt_cnt) {
2163 if (track->audio.sub_packet_cnt == 0)
2164 track->audio.buf_timecode = timecode;
2165 if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2166 if (size < cfs * h / 2) {
2167 av_log(matroska->ctx, AV_LOG_ERROR,
2168 "Corrupt int4 RM-style audio packet size\n");
2169 return AVERROR_INVALIDDATA;
2171 for (x=0; x<h/2; x++)
2172 memcpy(track->audio.buf+x*2*w+y*cfs,
2174 } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2176 av_log(matroska->ctx, AV_LOG_ERROR,
2177 "Corrupt sipr RM-style audio packet size\n");
2178 return AVERROR_INVALIDDATA;
2180 memcpy(track->audio.buf + y*w, data, w);
2182 if (size < sps * w / sps || h<=0) {
2183 av_log(matroska->ctx, AV_LOG_ERROR,
2184 "Corrupt generic RM-style audio packet size\n");
2185 return AVERROR_INVALIDDATA;
2187 for (x=0; x<w/sps; x++)
2188 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
2191 if (++track->audio.sub_packet_cnt >= h) {
2192 if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2193 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2194 track->audio.sub_packet_cnt = 0;
2195 track->audio.pkt_cnt = h*w / a;
2199 while (track->audio.pkt_cnt) {
2200 AVPacket *pkt = NULL;
2201 if (!(pkt = av_mallocz(sizeof(AVPacket))) || av_new_packet(pkt, a) < 0){
2203 return AVERROR(ENOMEM);
2205 memcpy(pkt->data, track->audio.buf
2206 + a * (h*w / a - track->audio.pkt_cnt--), a);
2207 pkt->pts = track->audio.buf_timecode;
2208 track->audio.buf_timecode = AV_NOPTS_VALUE;
2210 pkt->stream_index = st->index;
2211 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2217 /* reconstruct full wavpack blocks from mangled matroska ones */
2218 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
2219 uint8_t **pdst, int *size)
2221 uint8_t *dst = NULL;
2226 int ret, offset = 0;
2228 if (srclen < 12 || track->stream->codec->extradata_size < 2)
2229 return AVERROR_INVALIDDATA;
2231 ver = AV_RL16(track->stream->codec->extradata);
2233 samples = AV_RL32(src);
2237 while (srclen >= 8) {
2242 uint32_t flags = AV_RL32(src);
2243 uint32_t crc = AV_RL32(src + 4);
2247 multiblock = (flags & 0x1800) != 0x1800;
2250 ret = AVERROR_INVALIDDATA;
2253 blocksize = AV_RL32(src);
2259 if (blocksize > srclen) {
2260 ret = AVERROR_INVALIDDATA;
2264 tmp = av_realloc(dst, dstlen + blocksize + 32);
2266 ret = AVERROR(ENOMEM);
2270 dstlen += blocksize + 32;
2272 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2273 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2274 AV_WL16(dst + offset + 8, ver); // version
2275 AV_WL16(dst + offset + 10, 0); // track/index_no
2276 AV_WL32(dst + offset + 12, 0); // total samples
2277 AV_WL32(dst + offset + 16, 0); // block index
2278 AV_WL32(dst + offset + 20, samples); // number of samples
2279 AV_WL32(dst + offset + 24, flags); // flags
2280 AV_WL32(dst + offset + 28, crc); // crc
2281 memcpy (dst + offset + 32, src, blocksize); // block data
2284 srclen -= blocksize;
2285 offset += blocksize + 32;
2298 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
2299 MatroskaTrack *track,
2301 uint8_t *data, int data_len,
2307 uint8_t *id, *settings, *text, *buf;
2308 int id_len, settings_len, text_len;
2313 return AVERROR_INVALIDDATA;
2316 q = data + data_len;
2321 if (*p == '\r' || *p == '\n') {
2330 if (p >= q || *p != '\n')
2331 return AVERROR_INVALIDDATA;
2337 if (*p == '\r' || *p == '\n') {
2338 settings_len = p - settings;
2346 if (p >= q || *p != '\n')
2347 return AVERROR_INVALIDDATA;
2352 while (text_len > 0) {
2353 const int len = text_len - 1;
2354 const uint8_t c = p[len];
2355 if (c != '\r' && c != '\n')
2361 return AVERROR_INVALIDDATA;
2363 pkt = av_mallocz(sizeof(*pkt));
2364 err = av_new_packet(pkt, text_len);
2367 return AVERROR(err);
2370 memcpy(pkt->data, text, text_len);
2373 buf = av_packet_new_side_data(pkt,
2374 AV_PKT_DATA_WEBVTT_IDENTIFIER,
2378 return AVERROR(ENOMEM);
2380 memcpy(buf, id, id_len);
2383 if (settings_len > 0) {
2384 buf = av_packet_new_side_data(pkt,
2385 AV_PKT_DATA_WEBVTT_SETTINGS,
2389 return AVERROR(ENOMEM);
2391 memcpy(buf, settings, settings_len);
2394 // Do we need this for subtitles?
2395 // pkt->flags = AV_PKT_FLAG_KEY;
2397 pkt->stream_index = st->index;
2398 pkt->pts = timecode;
2400 // Do we need this for subtitles?
2401 // pkt->dts = timecode;
2403 pkt->duration = duration;
2406 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2407 matroska->prev_pkt = pkt;
2412 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2413 MatroskaTrack *track,
2415 uint8_t *data, int pkt_size,
2416 uint64_t timecode, uint64_t lace_duration,
2417 int64_t pos, int is_keyframe,
2418 uint8_t *additional, uint64_t additional_id, int additional_size,
2419 int64_t discard_padding)
2421 MatroskaTrackEncoding *encodings = track->encodings.elem;
2422 uint8_t *pkt_data = data;
2423 int offset = 0, res;
2426 if (encodings && !encodings->type && encodings->scope & 1) {
2427 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2432 if (st->codec->codec_id == AV_CODEC_ID_WAVPACK) {
2434 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
2436 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing a wavpack block.\n");
2439 if (pkt_data != data)
2440 av_freep(&pkt_data);
2444 if (st->codec->codec_id == AV_CODEC_ID_PRORES)
2447 pkt = av_mallocz(sizeof(AVPacket));
2448 /* XXX: prevent data copy... */
2449 if (av_new_packet(pkt, pkt_size + offset) < 0) {
2451 res = AVERROR(ENOMEM);
2455 if (st->codec->codec_id == AV_CODEC_ID_PRORES) {
2456 uint8_t *buf = pkt->data;
2457 bytestream_put_be32(&buf, pkt_size);
2458 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2461 memcpy(pkt->data + offset, pkt_data, pkt_size);
2463 if (pkt_data != data)
2464 av_freep(&pkt_data);
2466 pkt->flags = is_keyframe;
2467 pkt->stream_index = st->index;
2469 if (additional_size > 0) {
2470 uint8_t *side_data = av_packet_new_side_data(pkt,
2471 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
2472 additional_size + 8);
2473 if(side_data == NULL) {
2474 av_free_packet(pkt);
2476 return AVERROR(ENOMEM);
2478 AV_WB64(side_data, additional_id);
2479 memcpy(side_data + 8, additional, additional_size);
2482 if (discard_padding) {
2483 uint8_t *side_data = av_packet_new_side_data(pkt,
2484 AV_PKT_DATA_SKIP_SAMPLES,
2486 if(side_data == NULL) {
2487 av_free_packet(pkt);
2489 return AVERROR(ENOMEM);
2491 AV_WL32(side_data, 0);
2492 AV_WL32(side_data + 4, av_rescale_q(discard_padding,
2493 (AVRational){1, 1000000000},
2494 (AVRational){1, st->codec->sample_rate}));
2497 if (track->ms_compat)
2498 pkt->dts = timecode;
2500 pkt->pts = timecode;
2502 if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2504 * For backward compatibility.
2505 * Historically, we have put subtitle duration
2506 * in convergence_duration, on the off chance
2507 * that the time_scale is less than 1us, which
2508 * could result in a 32bit overflow on the
2509 * normal duration field.
2511 pkt->convergence_duration = lace_duration;
2514 if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE ||
2515 lace_duration <= INT_MAX) {
2517 * For non subtitle tracks, just store the duration
2520 * If it's a subtitle track and duration value does
2521 * not overflow a uint32, then also store it normally.
2523 pkt->duration = lace_duration;
2527 if (st->codec->codec_id == AV_CODEC_ID_SSA)
2528 matroska_fix_ass_packet(matroska, pkt, lace_duration);
2530 if (matroska->prev_pkt &&
2531 timecode != AV_NOPTS_VALUE &&
2532 matroska->prev_pkt->pts == timecode &&
2533 matroska->prev_pkt->stream_index == st->index &&
2534 st->codec->codec_id == AV_CODEC_ID_SSA)
2535 matroska_merge_packets(matroska->prev_pkt, pkt);
2537 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2538 matroska->prev_pkt = pkt;
2541 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2542 matroska->prev_pkt = pkt;
2547 if (pkt_data != data)
2548 av_freep(&pkt_data);
2552 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
2553 int size, int64_t pos, uint64_t cluster_time,
2554 uint64_t block_duration, int is_keyframe,
2555 uint8_t *additional, uint64_t additional_id, int additional_size,
2556 int64_t cluster_pos, int64_t discard_padding)
2558 uint64_t timecode = AV_NOPTS_VALUE;
2559 MatroskaTrack *track;
2563 uint32_t *lace_size = NULL;
2564 int n, flags, laces = 0;
2566 int trust_default_duration = 1;
2568 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2569 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2575 track = matroska_find_track_by_num(matroska, num);
2576 if (!track || !track->stream) {
2577 av_log(matroska->ctx, AV_LOG_INFO,
2578 "Invalid stream %"PRIu64" or size %u\n", num, size);
2579 return AVERROR_INVALIDDATA;
2580 } else if (size <= 3)
2583 if (st->discard >= AVDISCARD_ALL)
2585 av_assert1(block_duration != AV_NOPTS_VALUE);
2587 block_time = sign_extend(AV_RB16(data), 16);
2591 if (is_keyframe == -1)
2592 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2594 if (cluster_time != (uint64_t)-1
2595 && (block_time >= 0 || cluster_time >= -block_time)) {
2596 timecode = cluster_time + block_time;
2597 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
2598 && timecode < track->end_timecode)
2599 is_keyframe = 0; /* overlapping subtitles are not key frame */
2601 av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
2604 if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2605 if (timecode < matroska->skip_to_timecode)
2608 matroska->skip_to_keyframe = 0;
2609 else if (!st->skip_to_keyframe) {
2610 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2611 matroska->skip_to_keyframe = 0;
2615 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
2616 &lace_size, &laces);
2621 if (track->audio.samplerate == 8000) {
2622 // If this is needed for more codecs, then add them here
2623 if (st->codec->codec_id == AV_CODEC_ID_AC3) {
2624 if(track->audio.samplerate != st->codec->sample_rate || !st->codec->frame_size)
2625 trust_default_duration = 0;
2629 if (!block_duration && trust_default_duration)
2630 block_duration = track->default_duration * laces / matroska->time_scale;
2632 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
2633 track->end_timecode =
2634 FFMAX(track->end_timecode, timecode + block_duration);
2636 for (n = 0; n < laces; n++) {
2637 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
2639 if (lace_size[n] > size) {
2640 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2644 if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2645 st->codec->codec_id == AV_CODEC_ID_COOK ||
2646 st->codec->codec_id == AV_CODEC_ID_SIPR ||
2647 st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2648 st->codec->block_align && track->audio.sub_packet_size) {
2650 res = matroska_parse_rm_audio(matroska, track, st, data,
2656 } else if (st->codec->codec_id == AV_CODEC_ID_WEBVTT) {
2657 res = matroska_parse_webvtt(matroska, track, st,
2659 timecode, lace_duration,
2665 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2666 timecode, lace_duration,
2667 pos, !n? is_keyframe : 0,
2668 additional, additional_id, additional_size,
2674 if (timecode != AV_NOPTS_VALUE)
2675 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2676 data += lace_size[n];
2677 size -= lace_size[n];
2685 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
2687 EbmlList *blocks_list;
2688 MatroskaBlock *blocks;
2690 res = ebml_parse(matroska,
2691 matroska_cluster_incremental_parsing,
2692 &matroska->current_cluster);
2695 if (matroska->current_cluster_pos)
2696 ebml_level_end(matroska);
2697 ebml_free(matroska_cluster, &matroska->current_cluster);
2698 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2699 matroska->current_cluster_num_blocks = 0;
2700 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2701 matroska->prev_pkt = NULL;
2702 /* sizeof the ID which was already read */
2703 if (matroska->current_id)
2704 matroska->current_cluster_pos -= 4;
2705 res = ebml_parse(matroska,
2706 matroska_clusters_incremental,
2707 &matroska->current_cluster);
2708 /* Try parsing the block again. */
2710 res = ebml_parse(matroska,
2711 matroska_cluster_incremental_parsing,
2712 &matroska->current_cluster);
2716 matroska->current_cluster_num_blocks <
2717 matroska->current_cluster.blocks.nb_elem) {
2718 blocks_list = &matroska->current_cluster.blocks;
2719 blocks = blocks_list->elem;
2721 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2722 i = blocks_list->nb_elem - 1;
2723 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2724 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2725 uint8_t* additional = blocks[i].additional.size > 0 ?
2726 blocks[i].additional.data : NULL;
2727 if (!blocks[i].non_simple)
2728 blocks[i].duration = 0;
2729 res = matroska_parse_block(matroska,
2730 blocks[i].bin.data, blocks[i].bin.size,
2732 matroska->current_cluster.timecode,
2733 blocks[i].duration, is_keyframe,
2734 additional, blocks[i].additional_id,
2735 blocks[i].additional.size,
2736 matroska->current_cluster_pos,
2737 blocks[i].discard_padding);
2744 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
2746 MatroskaCluster cluster = { 0 };
2747 EbmlList *blocks_list;
2748 MatroskaBlock *blocks;
2751 if (!matroska->contains_ssa)
2752 return matroska_parse_cluster_incremental(matroska);
2753 pos = avio_tell(matroska->ctx->pb);
2754 matroska->prev_pkt = NULL;
2755 if (matroska->current_id)
2756 pos -= 4; /* sizeof the ID which was already read */
2757 res = ebml_parse(matroska, matroska_clusters, &cluster);
2758 blocks_list = &cluster.blocks;
2759 blocks = blocks_list->elem;
2760 for (i=0; i<blocks_list->nb_elem; i++)
2761 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2762 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2763 res=matroska_parse_block(matroska,
2764 blocks[i].bin.data, blocks[i].bin.size,
2765 blocks[i].bin.pos, cluster.timecode,
2766 blocks[i].duration, is_keyframe, NULL, 0, 0,
2767 pos, blocks[i].discard_padding);
2769 ebml_free(matroska_cluster, &cluster);
2773 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
2775 MatroskaDemuxContext *matroska = s->priv_data;
2777 while (matroska_deliver_packet(matroska, pkt)) {
2778 int64_t pos = avio_tell(matroska->ctx->pb);
2781 if (matroska_parse_cluster(matroska) < 0)
2782 matroska_resync(matroska, pos);
2788 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2789 int64_t timestamp, int flags)
2791 MatroskaDemuxContext *matroska = s->priv_data;
2792 MatroskaTrack *tracks = matroska->tracks.elem;
2793 AVStream *st = s->streams[stream_index];
2794 int i, index, index_sub, index_min;
2796 /* Parse the CUES now since we need the index data to seek. */
2797 if (matroska->cues_parsing_deferred > 0) {
2798 matroska->cues_parsing_deferred = 0;
2799 matroska_parse_cues(matroska);
2802 if (!st->nb_index_entries)
2804 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2806 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2807 avio_seek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
2808 matroska->current_id = 0;
2809 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2810 matroska_clear_queue(matroska);
2811 if (matroska_parse_cluster(matroska) < 0)
2816 matroska_clear_queue(matroska);
2817 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
2821 for (i=0; i < matroska->tracks.nb_elem; i++) {
2822 tracks[i].audio.pkt_cnt = 0;
2823 tracks[i].audio.sub_packet_cnt = 0;
2824 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2825 tracks[i].end_timecode = 0;
2826 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
2827 && tracks[i].stream->discard != AVDISCARD_ALL) {
2828 index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
2829 while(index_sub >= 0
2831 && tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos
2832 && st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
2837 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2838 matroska->current_id = 0;
2839 if (flags & AVSEEK_FLAG_ANY) {
2840 st->skip_to_keyframe = 0;
2841 matroska->skip_to_timecode = timestamp;
2843 st->skip_to_keyframe = 1;
2844 matroska->skip_to_timecode = st->index_entries[index].timestamp;
2846 matroska->skip_to_keyframe = 1;
2848 matroska->num_levels = 0;
2849 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2852 // slightly hackish but allows proper fallback to
2853 // the generic seeking code.
2854 matroska_clear_queue(matroska);
2855 matroska->current_id = 0;
2856 st->skip_to_keyframe =
2857 matroska->skip_to_keyframe = 0;
2859 matroska->num_levels = 0;
2863 static int matroska_read_close(AVFormatContext *s)
2865 MatroskaDemuxContext *matroska = s->priv_data;
2866 MatroskaTrack *tracks = matroska->tracks.elem;
2869 matroska_clear_queue(matroska);
2871 for (n=0; n < matroska->tracks.nb_elem; n++)
2872 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2873 av_free(tracks[n].audio.buf);
2874 ebml_free(matroska_cluster, &matroska->current_cluster);
2875 ebml_free(matroska_segment, matroska);
2880 AVInputFormat ff_matroska_demuxer = {
2881 .name = "matroska,webm",
2882 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
2883 .priv_data_size = sizeof(MatroskaDemuxContext),
2884 .read_probe = matroska_probe,
2885 .read_header = matroska_read_header,
2886 .read_packet = matroska_read_packet,
2887 .read_close = matroska_read_close,
2888 .read_seek = matroska_read_seek,