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/intfloat.h"
43 #include "libavutil/intreadwrite.h"
44 #include "libavutil/avstring.h"
45 #include "libavutil/lzo.h"
46 #include "libavutil/dict.h"
67 typedef const struct EbmlSyntax {
76 const struct EbmlSyntax *n;
96 uint64_t doctype_version;
102 } MatroskaTrackCompression;
107 MatroskaTrackCompression compression;
108 } MatroskaTrackEncoding;
112 uint64_t display_width;
113 uint64_t display_height;
114 uint64_t pixel_width;
115 uint64_t pixel_height;
117 uint64_t stereo_mode;
118 } MatroskaTrackVideo;
122 double out_samplerate;
126 /* real audio header (extracted from extradata) */
133 uint64_t buf_timecode;
135 } MatroskaTrackAudio;
140 } MatroskaTrackPlane;
143 EbmlList combine_planes;
144 } MatroskaTrackOperation;
155 uint64_t default_duration;
156 uint64_t flag_default;
157 uint64_t flag_forced;
158 MatroskaTrackVideo video;
159 MatroskaTrackAudio audio;
160 MatroskaTrackOperation operation;
164 int64_t end_timecode;
175 } MatroskaAttachement;
213 MatroskaTagTarget target;
233 AVFormatContext *ctx;
237 MatroskaLevel levels[EBML_MAX_DEPTH];
246 EbmlList attachments;
252 /* byte position of the segment inside the stream */
253 int64_t segment_start;
255 /* the packet queue */
262 /* What to skip before effectively reading a packet. */
263 int skip_to_keyframe;
264 uint64_t skip_to_timecode;
266 /* File has a CUES element, but we defer parsing until it is needed. */
267 int cues_parsing_deferred;
269 int current_cluster_num_blocks;
270 int64_t current_cluster_pos;
271 MatroskaCluster current_cluster;
273 /* File has SSA subtitles which prevent incremental cluster parsing. */
275 } MatroskaDemuxContext;
284 static EbmlSyntax ebml_header[] = {
285 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
286 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
287 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
288 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
289 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
290 { EBML_ID_EBMLVERSION, EBML_NONE },
291 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
295 static EbmlSyntax ebml_syntax[] = {
296 { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
300 static EbmlSyntax matroska_info[] = {
301 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
302 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
303 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
304 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
305 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
306 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext,date_utc) },
307 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
311 static EbmlSyntax matroska_track_video[] = {
312 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
313 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
314 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
315 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
316 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
317 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo,color_space) },
318 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,stereo_mode) },
319 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
320 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
321 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
322 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
323 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
324 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
325 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
329 static EbmlSyntax matroska_track_audio[] = {
330 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
331 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
332 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
333 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
337 static EbmlSyntax matroska_track_encoding_compression[] = {
338 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
339 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
343 static EbmlSyntax matroska_track_encoding[] = {
344 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
345 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
346 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
347 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
351 static EbmlSyntax matroska_track_encodings[] = {
352 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
356 static EbmlSyntax matroska_track_plane[] = {
357 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
358 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
362 static EbmlSyntax matroska_track_combine_planes[] = {
363 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n=matroska_track_plane} },
367 static EbmlSyntax matroska_track_operation[] = {
368 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n=matroska_track_combine_planes} },
372 static EbmlSyntax matroska_track[] = {
373 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
374 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack,name) },
375 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
376 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
377 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
378 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
379 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
380 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
381 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
382 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
383 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack,flag_forced), {.u=0} },
384 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
385 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
386 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack,operation), {.n=matroska_track_operation} },
387 { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
388 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
389 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
390 { MATROSKA_ID_CODECNAME, EBML_NONE },
391 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
392 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
393 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
394 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
395 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
396 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
400 static EbmlSyntax matroska_tracks[] = {
401 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
405 static EbmlSyntax matroska_attachment[] = {
406 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
407 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
408 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
409 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
410 { MATROSKA_ID_FILEDESC, EBML_NONE },
414 static EbmlSyntax matroska_attachments[] = {
415 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
419 static EbmlSyntax matroska_chapter_display[] = {
420 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
421 { MATROSKA_ID_CHAPLANG, EBML_NONE },
425 static EbmlSyntax matroska_chapter_entry[] = {
426 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
427 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
428 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
429 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
430 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
431 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
432 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
433 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
437 static EbmlSyntax matroska_chapter[] = {
438 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
439 { MATROSKA_ID_EDITIONUID, EBML_NONE },
440 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
441 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
442 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
446 static EbmlSyntax matroska_chapters[] = {
447 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
451 static EbmlSyntax matroska_index_pos[] = {
452 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
453 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
454 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
458 static EbmlSyntax matroska_index_entry[] = {
459 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
460 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
464 static EbmlSyntax matroska_index[] = {
465 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
469 static EbmlSyntax matroska_simpletag[] = {
470 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
471 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
472 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag,lang), {.s="und"} },
473 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag,def) },
474 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag,def) },
475 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
479 static EbmlSyntax matroska_tagtargets[] = {
480 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget,type) },
481 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
482 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
483 { MATROSKA_ID_TAGTARGETS_CHAPTERUID,EBML_UINT, 0, offsetof(MatroskaTagTarget,chapteruid) },
484 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
488 static EbmlSyntax matroska_tag[] = {
489 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
490 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
494 static EbmlSyntax matroska_tags[] = {
495 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
499 static EbmlSyntax matroska_seekhead_entry[] = {
500 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
501 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
505 static EbmlSyntax matroska_seekhead[] = {
506 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
510 static EbmlSyntax matroska_segment[] = {
511 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
512 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
513 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
514 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
515 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
516 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
517 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
518 { MATROSKA_ID_CLUSTER, EBML_STOP },
522 static EbmlSyntax matroska_segments[] = {
523 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
527 static EbmlSyntax matroska_blockgroup[] = {
528 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
529 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
530 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration) },
531 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
532 { 1, EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
536 static EbmlSyntax matroska_cluster[] = {
537 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
538 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
539 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
540 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
541 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
545 static EbmlSyntax matroska_clusters[] = {
546 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
547 { MATROSKA_ID_INFO, EBML_NONE },
548 { MATROSKA_ID_CUES, EBML_NONE },
549 { MATROSKA_ID_TAGS, EBML_NONE },
550 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
554 static EbmlSyntax matroska_cluster_incremental_parsing[] = {
555 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
556 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
557 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
558 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
559 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
560 { MATROSKA_ID_INFO, EBML_NONE },
561 { MATROSKA_ID_CUES, EBML_NONE },
562 { MATROSKA_ID_TAGS, EBML_NONE },
563 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
564 { MATROSKA_ID_CLUSTER, EBML_STOP },
568 static EbmlSyntax matroska_cluster_incremental[] = {
569 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
570 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
571 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
572 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
573 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
577 static EbmlSyntax matroska_clusters_incremental[] = {
578 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster_incremental} },
579 { MATROSKA_ID_INFO, EBML_NONE },
580 { MATROSKA_ID_CUES, EBML_NONE },
581 { MATROSKA_ID_TAGS, EBML_NONE },
582 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
586 static const char *const matroska_doctypes[] = { "matroska", "webm" };
588 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
590 AVIOContext *pb = matroska->ctx->pb;
592 matroska->current_id = 0;
593 matroska->num_levels = 0;
595 // seek to next position to resync from
596 if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0 || avio_tell(pb) <= last_pos)
601 // try to find a toplevel element
602 while (!url_feof(pb)) {
603 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
604 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
605 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
606 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS)
608 matroska->current_id = id;
611 id = (id << 8) | avio_r8(pb);
619 * Return: Whether we reached the end of a level in the hierarchy or not.
621 static int ebml_level_end(MatroskaDemuxContext *matroska)
623 AVIOContext *pb = matroska->ctx->pb;
624 int64_t pos = avio_tell(pb);
626 if (matroska->num_levels > 0) {
627 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
628 if (pos - level->start >= level->length || matroska->current_id) {
629 matroska->num_levels--;
637 * Read: an "EBML number", which is defined as a variable-length
638 * array of bytes. The first byte indicates the length by giving a
639 * number of 0-bits followed by a one. The position of the first
640 * "one" bit inside the first byte indicates the length of this
642 * Returns: number of bytes read, < 0 on error
644 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
645 int max_size, uint64_t *number)
650 /* The first byte tells us the length in bytes - avio_r8() can normally
651 * return 0, but since that's not a valid first ebmlID byte, we can
652 * use it safely here to catch EOS. */
653 if (!(total = avio_r8(pb))) {
654 /* we might encounter EOS here */
656 int64_t pos = avio_tell(pb);
657 av_log(matroska->ctx, AV_LOG_ERROR,
658 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
660 return pb->error ? pb->error : AVERROR(EIO);
665 /* get the length of the EBML number */
666 read = 8 - ff_log2_tab[total];
667 if (read > max_size) {
668 int64_t pos = avio_tell(pb) - 1;
669 av_log(matroska->ctx, AV_LOG_ERROR,
670 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
671 (uint8_t) total, pos, pos);
672 return AVERROR_INVALIDDATA;
675 /* read out length */
676 total ^= 1 << ff_log2_tab[total];
678 total = (total << 8) | avio_r8(pb);
686 * Read a EBML length value.
687 * This needs special handling for the "unknown length" case which has multiple
690 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
693 int res = ebml_read_num(matroska, pb, 8, number);
694 if (res > 0 && *number + 1 == 1ULL << (7 * res))
695 *number = 0xffffffffffffffULL;
700 * Read the next element as an unsigned int.
701 * 0 is success, < 0 is failure.
703 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
708 return AVERROR_INVALIDDATA;
710 /* big-endian ordering; build up number */
713 *num = (*num << 8) | avio_r8(pb);
719 * Read the next element as a float.
720 * 0 is success, < 0 is failure.
722 static int ebml_read_float(AVIOContext *pb, int size, double *num)
726 } else if (size == 4) {
727 *num = av_int2float(avio_rb32(pb));
728 } else if (size == 8){
729 *num = av_int2double(avio_rb64(pb));
731 return AVERROR_INVALIDDATA;
737 * Read the next element as an ASCII string.
738 * 0 is success, < 0 is failure.
740 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
744 /* EBML strings are usually not 0-terminated, so we allocate one
745 * byte more, read the string and NULL-terminate it ourselves. */
746 if (!(res = av_malloc(size + 1)))
747 return AVERROR(ENOMEM);
748 if (avio_read(pb, (uint8_t *) res, size) != size) {
760 * Read the next element as binary data.
761 * 0 is success, < 0 is failure.
763 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
766 if (!(bin->data = av_malloc(length)))
767 return AVERROR(ENOMEM);
770 bin->pos = avio_tell(pb);
771 if (avio_read(pb, bin->data, length) != length) {
772 av_freep(&bin->data);
780 * Read the next element, but only the header. The contents
781 * are supposed to be sub-elements which can be read separately.
782 * 0 is success, < 0 is failure.
784 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
786 AVIOContext *pb = matroska->ctx->pb;
787 MatroskaLevel *level;
789 if (matroska->num_levels >= EBML_MAX_DEPTH) {
790 av_log(matroska->ctx, AV_LOG_ERROR,
791 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
792 return AVERROR(ENOSYS);
795 level = &matroska->levels[matroska->num_levels++];
796 level->start = avio_tell(pb);
797 level->length = length;
803 * Read signed/unsigned "EBML" numbers.
804 * Return: number of bytes processed, < 0 on error
806 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
807 uint8_t *data, uint32_t size, uint64_t *num)
810 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
811 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
815 * Same as above, but signed.
817 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
818 uint8_t *data, uint32_t size, int64_t *num)
823 /* read as unsigned number first */
824 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
827 /* make signed (weird way) */
828 *num = unum - ((1LL << (7*res - 1)) - 1);
833 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
834 EbmlSyntax *syntax, void *data);
836 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
837 uint32_t id, void *data)
840 for (i=0; syntax[i].id; i++)
841 if (id == syntax[i].id)
843 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
844 matroska->num_levels > 0 &&
845 matroska->levels[matroska->num_levels-1].length == 0xffffffffffffff)
846 return 0; // we reached the end of an unknown size cluster
847 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
848 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
849 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
850 return AVERROR_INVALIDDATA;
852 return ebml_parse_elem(matroska, &syntax[i], data);
855 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
858 if (!matroska->current_id) {
860 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
863 matroska->current_id = id | 1 << 7*res;
865 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
868 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
873 for (i=0; syntax[i].id; i++)
874 switch (syntax[i].type) {
876 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
879 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
883 *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
887 while (!res && !ebml_level_end(matroska))
888 res = ebml_parse(matroska, syntax, data);
893 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
894 EbmlSyntax *syntax, void *data)
896 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
899 // max. 16 MB for strings
900 [EBML_STR] = 0x1000000,
901 [EBML_UTF8] = 0x1000000,
902 // max. 256 MB for binary data
903 [EBML_BIN] = 0x10000000,
904 // no limits for anything else
906 AVIOContext *pb = matroska->ctx->pb;
907 uint32_t id = syntax->id;
912 data = (char *)data + syntax->data_offset;
913 if (syntax->list_elem_size) {
914 EbmlList *list = data;
915 newelem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
917 return AVERROR(ENOMEM);
918 list->elem = newelem;
919 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
920 memset(data, 0, syntax->list_elem_size);
924 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
925 matroska->current_id = 0;
926 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
928 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
929 av_log(matroska->ctx, AV_LOG_ERROR,
930 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
931 length, max_lengths[syntax->type], syntax->type);
932 return AVERROR_INVALIDDATA;
936 switch (syntax->type) {
937 case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
938 case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
940 case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
941 case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
942 case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
944 if (id == MATROSKA_ID_SEGMENT)
945 matroska->segment_start = avio_tell(matroska->ctx->pb);
946 return ebml_parse_nest(matroska, syntax->def.n, data);
947 case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
948 case EBML_STOP: return 1;
950 if(ffio_limit(pb, length) != length)
952 return avio_skip(pb,length)<0 ? AVERROR(EIO) : 0;
954 if (res == AVERROR_INVALIDDATA)
955 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
956 else if (res == AVERROR(EIO))
957 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
961 static void ebml_free(EbmlSyntax *syntax, void *data)
964 for (i=0; syntax[i].id; i++) {
965 void *data_off = (char *)data + syntax[i].data_offset;
966 switch (syntax[i].type) {
968 case EBML_UTF8: av_freep(data_off); break;
969 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
971 if (syntax[i].list_elem_size) {
972 EbmlList *list = data_off;
973 char *ptr = list->elem;
974 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
975 ebml_free(syntax[i].def.n, ptr);
978 ebml_free(syntax[i].def.n, data_off);
988 static int matroska_probe(AVProbeData *p)
991 int len_mask = 0x80, size = 1, n = 1, i;
994 if (AV_RB32(p->buf) != EBML_ID_HEADER)
997 /* length of header */
999 while (size <= 8 && !(total & len_mask)) {
1005 total &= (len_mask - 1);
1007 total = (total << 8) | p->buf[4 + n++];
1009 /* Does the probe data contain the whole header? */
1010 if (p->buf_size < 4 + size + total)
1013 /* The header should contain a known document type. For now,
1014 * we don't parse the whole header but simply check for the
1015 * availability of that array of characters inside the header.
1016 * Not fully fool-proof, but good enough. */
1017 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1018 int probelen = strlen(matroska_doctypes[i]);
1019 if (total < probelen)
1021 for (n = 4+size; n <= 4+size+total-probelen; n++)
1022 if (!memcmp(p->buf+n, matroska_doctypes[i], probelen))
1023 return AVPROBE_SCORE_MAX;
1026 // probably valid EBML header but no recognized doctype
1027 return AVPROBE_SCORE_MAX/2;
1030 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1033 MatroskaTrack *tracks = matroska->tracks.elem;
1036 for (i=0; i < matroska->tracks.nb_elem; i++)
1037 if (tracks[i].num == num)
1040 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1044 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
1045 MatroskaTrack *track)
1047 MatroskaTrackEncoding *encodings = track->encodings.elem;
1048 uint8_t* data = *buf;
1049 int isize = *buf_size;
1050 uint8_t* pkt_data = NULL;
1051 uint8_t av_unused *newpktdata;
1052 int pkt_size = isize;
1056 if (pkt_size >= 10000000U)
1057 return AVERROR_INVALIDDATA;
1059 switch (encodings[0].compression.algo) {
1060 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP: {
1061 int header_size = encodings[0].compression.settings.size;
1062 uint8_t *header = encodings[0].compression.settings.data;
1064 if (header_size && !header) {
1065 av_log(0, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1072 pkt_size = isize + header_size;
1073 pkt_data = av_malloc(pkt_size);
1075 return AVERROR(ENOMEM);
1077 memcpy(pkt_data, header, header_size);
1078 memcpy(pkt_data + header_size, data, isize);
1082 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1084 olen = pkt_size *= 3;
1085 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1087 result = AVERROR(ENOMEM);
1090 pkt_data = newpktdata;
1091 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1092 } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
1094 result = AVERROR_INVALIDDATA;
1101 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
1102 z_stream zstream = {0};
1103 if (inflateInit(&zstream) != Z_OK)
1105 zstream.next_in = data;
1106 zstream.avail_in = isize;
1109 newpktdata = av_realloc(pkt_data, pkt_size);
1111 inflateEnd(&zstream);
1114 pkt_data = newpktdata;
1115 zstream.avail_out = pkt_size - zstream.total_out;
1116 zstream.next_out = pkt_data + zstream.total_out;
1118 result = inflate(&zstream, Z_NO_FLUSH);
1120 result = Z_MEM_ERROR;
1121 } while (result==Z_OK && pkt_size<10000000);
1122 pkt_size = zstream.total_out;
1123 inflateEnd(&zstream);
1124 if (result != Z_STREAM_END) {
1125 if (result == Z_MEM_ERROR)
1126 result = AVERROR(ENOMEM);
1128 result = AVERROR_INVALIDDATA;
1135 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
1136 bz_stream bzstream = {0};
1137 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1139 bzstream.next_in = data;
1140 bzstream.avail_in = isize;
1143 newpktdata = av_realloc(pkt_data, pkt_size);
1145 BZ2_bzDecompressEnd(&bzstream);
1148 pkt_data = newpktdata;
1149 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1150 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1152 result = BZ2_bzDecompress(&bzstream);
1154 result = BZ_MEM_ERROR;
1155 } while (result==BZ_OK && pkt_size<10000000);
1156 pkt_size = bzstream.total_out_lo32;
1157 BZ2_bzDecompressEnd(&bzstream);
1158 if (result != BZ_STREAM_END) {
1159 if (result == BZ_MEM_ERROR)
1160 result = AVERROR(ENOMEM);
1162 result = AVERROR_INVALIDDATA;
1169 return AVERROR_INVALIDDATA;
1173 *buf_size = pkt_size;
1180 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
1181 AVPacket *pkt, uint64_t display_duration)
1183 char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
1184 for (; *ptr!=',' && ptr<end-1; ptr++);
1188 for (; *ptr!=',' && ptr<end-1; ptr++);
1190 int64_t end_pts = pkt->pts + display_duration;
1191 int sc = matroska->time_scale * pkt->pts / 10000000;
1192 int ec = matroska->time_scale * end_pts / 10000000;
1193 int sh, sm, ss, eh, em, es, len;
1194 sh = sc/360000; sc -= 360000*sh;
1195 sm = sc/ 6000; sc -= 6000*sm;
1196 ss = sc/ 100; sc -= 100*ss;
1197 eh = ec/360000; ec -= 360000*eh;
1198 em = ec/ 6000; ec -= 6000*em;
1199 es = ec/ 100; ec -= 100*es;
1201 len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
1202 if (!(line = av_malloc(len)))
1204 snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
1205 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
1208 pkt->size = strlen(line);
1212 static int matroska_merge_packets(AVPacket *out, AVPacket *in)
1214 int ret = av_grow_packet(out, in->size);
1218 memcpy(out->data + out->size - in->size, in->data, in->size);
1225 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1226 AVDictionary **metadata, char *prefix)
1228 MatroskaTag *tags = list->elem;
1232 for (i=0; i < list->nb_elem; i++) {
1233 const char *lang= (tags[i].lang && strcmp(tags[i].lang, "und")) ? tags[i].lang : NULL;
1235 if (!tags[i].name) {
1236 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1239 if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1240 else av_strlcpy(key, tags[i].name, sizeof(key));
1241 if (tags[i].def || !lang) {
1242 av_dict_set(metadata, key, tags[i].string, 0);
1243 if (tags[i].sub.nb_elem)
1244 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1247 av_strlcat(key, "-", sizeof(key));
1248 av_strlcat(key, lang, sizeof(key));
1249 av_dict_set(metadata, key, tags[i].string, 0);
1250 if (tags[i].sub.nb_elem)
1251 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1254 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1257 static void matroska_convert_tags(AVFormatContext *s)
1259 MatroskaDemuxContext *matroska = s->priv_data;
1260 MatroskaTags *tags = matroska->tags.elem;
1263 for (i=0; i < matroska->tags.nb_elem; i++) {
1264 if (tags[i].target.attachuid) {
1265 MatroskaAttachement *attachment = matroska->attachments.elem;
1266 for (j=0; j<matroska->attachments.nb_elem; j++)
1267 if (attachment[j].uid == tags[i].target.attachuid
1268 && attachment[j].stream)
1269 matroska_convert_tag(s, &tags[i].tag,
1270 &attachment[j].stream->metadata, NULL);
1271 } else if (tags[i].target.chapteruid) {
1272 MatroskaChapter *chapter = matroska->chapters.elem;
1273 for (j=0; j<matroska->chapters.nb_elem; j++)
1274 if (chapter[j].uid == tags[i].target.chapteruid
1275 && chapter[j].chapter)
1276 matroska_convert_tag(s, &tags[i].tag,
1277 &chapter[j].chapter->metadata, NULL);
1278 } else if (tags[i].target.trackuid) {
1279 MatroskaTrack *track = matroska->tracks.elem;
1280 for (j=0; j<matroska->tracks.nb_elem; j++)
1281 if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1282 matroska_convert_tag(s, &tags[i].tag,
1283 &track[j].stream->metadata, NULL);
1285 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1286 tags[i].target.type);
1291 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska, int idx)
1293 EbmlList *seekhead_list = &matroska->seekhead;
1294 MatroskaSeekhead *seekhead = seekhead_list->elem;
1295 uint32_t level_up = matroska->level_up;
1296 int64_t before_pos = avio_tell(matroska->ctx->pb);
1297 uint32_t saved_id = matroska->current_id;
1298 MatroskaLevel level;
1302 if (idx >= seekhead_list->nb_elem
1303 || seekhead[idx].id == MATROSKA_ID_SEEKHEAD
1304 || seekhead[idx].id == MATROSKA_ID_CLUSTER)
1308 offset = seekhead[idx].pos + matroska->segment_start;
1309 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1310 /* We don't want to lose our seekhead level, so we add
1311 * a dummy. This is a crude hack. */
1312 if (matroska->num_levels == EBML_MAX_DEPTH) {
1313 av_log(matroska->ctx, AV_LOG_INFO,
1314 "Max EBML element depth (%d) reached, "
1315 "cannot parse further.\n", EBML_MAX_DEPTH);
1316 ret = AVERROR_INVALIDDATA;
1319 level.length = (uint64_t)-1;
1320 matroska->levels[matroska->num_levels] = level;
1321 matroska->num_levels++;
1322 matroska->current_id = 0;
1324 ret = ebml_parse(matroska, matroska_segment, matroska);
1326 /* remove dummy level */
1327 while (matroska->num_levels) {
1328 uint64_t length = matroska->levels[--matroska->num_levels].length;
1329 if (length == (uint64_t)-1)
1335 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1336 matroska->level_up = level_up;
1337 matroska->current_id = saved_id;
1342 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1344 EbmlList *seekhead_list = &matroska->seekhead;
1345 int64_t before_pos = avio_tell(matroska->ctx->pb);
1348 // we should not do any seeking in the streaming case
1349 if (!matroska->ctx->pb->seekable ||
1350 (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1353 for (i = 0; i < seekhead_list->nb_elem; i++) {
1354 MatroskaSeekhead *seekhead = seekhead_list->elem;
1355 if (seekhead[i].pos <= before_pos)
1358 // defer cues parsing until we actually need cue data.
1359 if (seekhead[i].id == MATROSKA_ID_CUES) {
1360 matroska->cues_parsing_deferred = 1;
1364 if (matroska_parse_seekhead_entry(matroska, i) < 0) {
1365 // mark index as broken
1366 matroska->cues_parsing_deferred = -1;
1372 static void matroska_add_index_entries(MatroskaDemuxContext *matroska) {
1373 EbmlList *index_list;
1374 MatroskaIndex *index;
1375 int index_scale = 1;
1378 index_list = &matroska->index;
1379 index = index_list->elem;
1380 if (index_list->nb_elem
1381 && index[0].time > 1E14/matroska->time_scale) {
1382 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1383 index_scale = matroska->time_scale;
1385 for (i = 0; i < index_list->nb_elem; i++) {
1386 EbmlList *pos_list = &index[i].pos;
1387 MatroskaIndexPos *pos = pos_list->elem;
1388 for (j = 0; j < pos_list->nb_elem; j++) {
1389 MatroskaTrack *track = matroska_find_track_by_num(matroska, pos[j].track);
1390 if (track && track->stream)
1391 av_add_index_entry(track->stream,
1392 pos[j].pos + matroska->segment_start,
1393 index[i].time/index_scale, 0, 0,
1399 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1400 EbmlList *seekhead_list = &matroska->seekhead;
1401 MatroskaSeekhead *seekhead = seekhead_list->elem;
1404 for (i = 0; i < seekhead_list->nb_elem; i++)
1405 if (seekhead[i].id == MATROSKA_ID_CUES)
1407 assert(i <= seekhead_list->nb_elem);
1409 if (matroska_parse_seekhead_entry(matroska, i) < 0)
1410 matroska->cues_parsing_deferred = -1;
1411 matroska_add_index_entries(matroska);
1414 static int matroska_aac_profile(char *codec_id)
1416 static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1419 for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1420 if (strstr(codec_id, aac_profiles[profile]))
1425 static int matroska_aac_sri(int samplerate)
1429 for (sri=0; sri<FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1430 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1435 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1438 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1439 time_t creation_time = date_utc / 1000000000 + 978307200;
1440 struct tm *ptm = gmtime(&creation_time);
1442 strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm);
1443 av_dict_set(metadata, "creation_time", buffer, 0);
1446 static int matroska_read_header(AVFormatContext *s)
1448 MatroskaDemuxContext *matroska = s->priv_data;
1449 EbmlList *attachements_list = &matroska->attachments;
1450 MatroskaAttachement *attachements;
1451 EbmlList *chapters_list = &matroska->chapters;
1452 MatroskaChapter *chapters;
1453 MatroskaTrack *tracks;
1454 uint64_t max_start = 0;
1462 /* First read the EBML header. */
1463 if (ebml_parse(matroska, ebml_syntax, &ebml)
1464 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1465 || ebml.id_length > sizeof(uint32_t) || ebml.doctype_version > 3 || !ebml.doctype) {
1466 av_log(matroska->ctx, AV_LOG_ERROR,
1467 "EBML header using unsupported features\n"
1468 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1469 ebml.version, ebml.doctype, ebml.doctype_version);
1470 ebml_free(ebml_syntax, &ebml);
1471 return AVERROR_PATCHWELCOME;
1472 } else if (ebml.doctype_version == 3) {
1473 av_log(matroska->ctx, AV_LOG_WARNING,
1474 "EBML header using unsupported features\n"
1475 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1476 ebml.version, ebml.doctype, ebml.doctype_version);
1478 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
1479 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
1481 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
1482 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
1483 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
1484 ebml_free(ebml_syntax, &ebml);
1485 return AVERROR_INVALIDDATA;
1488 ebml_free(ebml_syntax, &ebml);
1490 /* The next thing is a segment. */
1491 pos = avio_tell(matroska->ctx->pb);
1492 res = ebml_parse(matroska, matroska_segments, matroska);
1493 // try resyncing until we find a EBML_STOP type element.
1495 res = matroska_resync(matroska, pos);
1498 pos = avio_tell(matroska->ctx->pb);
1499 res = ebml_parse(matroska, matroska_segment, matroska);
1501 matroska_execute_seekhead(matroska);
1503 if (!matroska->time_scale)
1504 matroska->time_scale = 1000000;
1505 if (matroska->duration)
1506 matroska->ctx->duration = matroska->duration * matroska->time_scale
1507 * 1000 / AV_TIME_BASE;
1508 av_dict_set(&s->metadata, "title", matroska->title, 0);
1510 if (matroska->date_utc.size == 8)
1511 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
1513 tracks = matroska->tracks.elem;
1514 for (i=0; i < matroska->tracks.nb_elem; i++) {
1515 MatroskaTrack *track = &tracks[i];
1516 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1517 EbmlList *encodings_list = &track->encodings;
1518 MatroskaTrackEncoding *encodings = encodings_list->elem;
1519 uint8_t *extradata = NULL;
1520 int extradata_size = 0;
1521 int extradata_offset = 0;
1522 uint32_t fourcc = 0;
1525 /* Apply some sanity checks. */
1526 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1527 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1528 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1529 av_log(matroska->ctx, AV_LOG_INFO,
1530 "Unknown or unsupported track type %"PRIu64"\n",
1534 if (track->codec_id == NULL)
1537 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1538 if (!track->default_duration && track->video.frame_rate > 0)
1539 track->default_duration = 1000000000/track->video.frame_rate;
1540 if (!track->video.display_width)
1541 track->video.display_width = track->video.pixel_width;
1542 if (!track->video.display_height)
1543 track->video.display_height = track->video.pixel_height;
1544 if (track->video.color_space.size == 4)
1545 fourcc = AV_RL32(track->video.color_space.data);
1546 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1547 if (!track->audio.out_samplerate)
1548 track->audio.out_samplerate = track->audio.samplerate;
1550 if (encodings_list->nb_elem > 1) {
1551 av_log(matroska->ctx, AV_LOG_ERROR,
1552 "Multiple combined encodings not supported");
1553 } else if (encodings_list->nb_elem == 1) {
1554 if (encodings[0].type ||
1557 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1560 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1563 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
1565 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP)) {
1566 encodings[0].scope = 0;
1567 av_log(matroska->ctx, AV_LOG_ERROR,
1568 "Unsupported encoding type");
1569 } else if (track->codec_priv.size && encodings[0].scope&2) {
1570 uint8_t *codec_priv = track->codec_priv.data;
1571 int ret = matroska_decode_buffer(&track->codec_priv.data,
1572 &track->codec_priv.size,
1575 track->codec_priv.data = NULL;
1576 track->codec_priv.size = 0;
1577 av_log(matroska->ctx, AV_LOG_ERROR,
1578 "Failed to decode codec private data\n");
1581 if (codec_priv != track->codec_priv.data)
1582 av_free(codec_priv);
1586 for(j=0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++){
1587 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1588 strlen(ff_mkv_codec_tags[j].str))){
1589 codec_id= ff_mkv_codec_tags[j].id;
1594 st = track->stream = avformat_new_stream(s, NULL);
1596 return AVERROR(ENOMEM);
1598 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1599 && track->codec_priv.size >= 40
1600 && track->codec_priv.data != NULL) {
1601 track->ms_compat = 1;
1602 fourcc = AV_RL32(track->codec_priv.data + 16);
1603 codec_id = ff_codec_get_id(ff_codec_bmp_tags, fourcc);
1604 extradata_offset = 40;
1605 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1606 && track->codec_priv.size >= 14
1607 && track->codec_priv.data != NULL) {
1609 ffio_init_context(&b, track->codec_priv.data, track->codec_priv.size,
1610 AVIO_FLAG_READ, NULL, NULL, NULL, NULL);
1611 ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1614 codec_id = st->codec->codec_id;
1615 extradata_offset = FFMIN(track->codec_priv.size, 18);
1616 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1617 && (track->codec_priv.size >= 86)
1618 && (track->codec_priv.data != NULL)) {
1619 fourcc = AV_RL32(track->codec_priv.data);
1620 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1621 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX-12) {
1622 /* Only ALAC's magic cookie is stored in Matroska's track headers.
1623 Create the "atom size", "tag", and "tag version" fields the
1624 decoder expects manually. */
1625 extradata_size = 12 + track->codec_priv.size;
1626 extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1627 if (extradata == NULL)
1628 return AVERROR(ENOMEM);
1629 AV_WB32(extradata, extradata_size);
1630 memcpy(&extradata[4], "alac", 4);
1631 AV_WB32(&extradata[8], 0);
1632 memcpy(&extradata[12], track->codec_priv.data, track->codec_priv.size);
1633 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1634 switch (track->audio.bitdepth) {
1635 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1636 case 24: codec_id = AV_CODEC_ID_PCM_S24BE; break;
1637 case 32: codec_id = AV_CODEC_ID_PCM_S32BE; break;
1639 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1640 switch (track->audio.bitdepth) {
1641 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1642 case 24: codec_id = AV_CODEC_ID_PCM_S24LE; break;
1643 case 32: codec_id = AV_CODEC_ID_PCM_S32LE; break;
1645 } else if (codec_id==AV_CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1646 codec_id = AV_CODEC_ID_PCM_F64LE;
1647 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1648 int profile = matroska_aac_profile(track->codec_id);
1649 int sri = matroska_aac_sri(track->audio.samplerate);
1650 extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1651 if (extradata == NULL)
1652 return AVERROR(ENOMEM);
1653 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1654 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1655 if (strstr(track->codec_id, "SBR")) {
1656 sri = matroska_aac_sri(track->audio.out_samplerate);
1657 extradata[2] = 0x56;
1658 extradata[3] = 0xE5;
1659 extradata[4] = 0x80 | (sri<<3);
1663 } else if (codec_id == AV_CODEC_ID_TTA) {
1664 extradata_size = 30;
1665 extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1666 if (extradata == NULL)
1667 return AVERROR(ENOMEM);
1668 ffio_init_context(&b, extradata, extradata_size, 1,
1669 NULL, NULL, NULL, NULL);
1670 avio_write(&b, "TTA1", 4);
1672 avio_wl16(&b, track->audio.channels);
1673 avio_wl16(&b, track->audio.bitdepth);
1674 avio_wl32(&b, track->audio.out_samplerate);
1675 avio_wl32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1676 } else if (codec_id == AV_CODEC_ID_RV10 || codec_id == AV_CODEC_ID_RV20 ||
1677 codec_id == AV_CODEC_ID_RV30 || codec_id == AV_CODEC_ID_RV40) {
1678 extradata_offset = 26;
1679 } else if (codec_id == AV_CODEC_ID_RA_144) {
1680 track->audio.out_samplerate = 8000;
1681 track->audio.channels = 1;
1682 } else if ((codec_id == AV_CODEC_ID_RA_288 || codec_id == AV_CODEC_ID_COOK ||
1683 codec_id == AV_CODEC_ID_ATRAC3 || codec_id == AV_CODEC_ID_SIPR)
1684 && track->codec_priv.data) {
1687 ffio_init_context(&b, track->codec_priv.data,track->codec_priv.size,
1688 0, NULL, NULL, NULL, NULL);
1690 flavor = avio_rb16(&b);
1691 track->audio.coded_framesize = avio_rb32(&b);
1693 track->audio.sub_packet_h = avio_rb16(&b);
1694 track->audio.frame_size = avio_rb16(&b);
1695 track->audio.sub_packet_size = avio_rb16(&b);
1696 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1697 if (codec_id == AV_CODEC_ID_RA_288) {
1698 st->codec->block_align = track->audio.coded_framesize;
1699 track->codec_priv.size = 0;
1701 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1702 const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1703 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1704 st->codec->bit_rate = sipr_bit_rate[flavor];
1706 st->codec->block_align = track->audio.sub_packet_size;
1707 extradata_offset = 78;
1710 track->codec_priv.size -= extradata_offset;
1712 if (codec_id == AV_CODEC_ID_NONE)
1713 av_log(matroska->ctx, AV_LOG_INFO,
1714 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1716 if (track->time_scale < 0.01)
1717 track->time_scale = 1.0;
1718 avpriv_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1720 st->codec->codec_id = codec_id;
1722 if (strcmp(track->language, "und"))
1723 av_dict_set(&st->metadata, "language", track->language, 0);
1724 av_dict_set(&st->metadata, "title", track->name, 0);
1726 if (track->flag_default)
1727 st->disposition |= AV_DISPOSITION_DEFAULT;
1728 if (track->flag_forced)
1729 st->disposition |= AV_DISPOSITION_FORCED;
1731 if (!st->codec->extradata) {
1733 st->codec->extradata = extradata;
1734 st->codec->extradata_size = extradata_size;
1735 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1736 st->codec->extradata = av_mallocz(track->codec_priv.size +
1737 FF_INPUT_BUFFER_PADDING_SIZE);
1738 if(st->codec->extradata == NULL)
1739 return AVERROR(ENOMEM);
1740 st->codec->extradata_size = track->codec_priv.size;
1741 memcpy(st->codec->extradata,
1742 track->codec_priv.data + extradata_offset,
1743 track->codec_priv.size);
1747 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1748 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
1750 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1751 st->codec->codec_tag = fourcc;
1752 st->codec->width = track->video.pixel_width;
1753 st->codec->height = track->video.pixel_height;
1754 av_reduce(&st->sample_aspect_ratio.num,
1755 &st->sample_aspect_ratio.den,
1756 st->codec->height * track->video.display_width,
1757 st->codec-> width * track->video.display_height,
1759 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1760 if (track->default_duration) {
1761 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1762 1000000000, track->default_duration, 30000);
1763 #if FF_API_R_FRAME_RATE
1764 st->r_frame_rate = st->avg_frame_rate;
1768 /* export stereo mode flag as metadata tag */
1769 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREO_MODE_COUNT)
1770 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
1772 /* if we have virtual track, mark the real tracks */
1773 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
1775 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
1777 snprintf(buf, sizeof(buf), "%s_%d",
1778 ff_matroska_video_stereo_plane[planes[j].type], i);
1779 for (k=0; k < matroska->tracks.nb_elem; k++)
1780 if (planes[j].uid == tracks[k].uid) {
1781 av_dict_set(&s->streams[k]->metadata,
1782 "stereo_mode", buf, 0);
1786 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1787 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1788 st->codec->sample_rate = track->audio.out_samplerate;
1789 st->codec->channels = track->audio.channels;
1790 if (st->codec->codec_id != AV_CODEC_ID_AAC)
1791 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1792 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1793 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1794 if (st->codec->codec_id == AV_CODEC_ID_SSA)
1795 matroska->contains_ssa = 1;
1799 attachements = attachements_list->elem;
1800 for (j=0; j<attachements_list->nb_elem; j++) {
1801 if (!(attachements[j].filename && attachements[j].mime &&
1802 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1803 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1805 AVStream *st = avformat_new_stream(s, NULL);
1808 av_dict_set(&st->metadata, "filename",attachements[j].filename, 0);
1809 av_dict_set(&st->metadata, "mimetype", attachements[j].mime, 0);
1810 st->codec->codec_id = AV_CODEC_ID_NONE;
1811 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
1812 st->codec->extradata = av_malloc(attachements[j].bin.size + FF_INPUT_BUFFER_PADDING_SIZE);
1813 if(st->codec->extradata == NULL)
1815 st->codec->extradata_size = attachements[j].bin.size;
1816 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1818 for (i=0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
1819 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1820 strlen(ff_mkv_mime_tags[i].str))) {
1821 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1825 attachements[j].stream = st;
1829 chapters = chapters_list->elem;
1830 for (i=0; i<chapters_list->nb_elem; i++)
1831 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1832 && (max_start==0 || chapters[i].start > max_start)) {
1833 chapters[i].chapter =
1834 avpriv_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1835 chapters[i].start, chapters[i].end,
1837 av_dict_set(&chapters[i].chapter->metadata,
1838 "title", chapters[i].title, 0);
1839 max_start = chapters[i].start;
1842 matroska_add_index_entries(matroska);
1844 matroska_convert_tags(s);
1850 * Put one packet in an application-supplied AVPacket struct.
1851 * Returns 0 on success or -1 on failure.
1853 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1856 if (matroska->num_packets > 0) {
1857 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1858 av_free(matroska->packets[0]);
1859 if (matroska->num_packets > 1) {
1861 memmove(&matroska->packets[0], &matroska->packets[1],
1862 (matroska->num_packets - 1) * sizeof(AVPacket *));
1863 newpackets = av_realloc(matroska->packets,
1864 (matroska->num_packets - 1) * sizeof(AVPacket *));
1866 matroska->packets = newpackets;
1868 av_freep(&matroska->packets);
1869 matroska->prev_pkt = NULL;
1871 matroska->num_packets--;
1879 * Free all packets in our internal queue.
1881 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1883 if (matroska->packets) {
1885 for (n = 0; n < matroska->num_packets; n++) {
1886 av_free_packet(matroska->packets[n]);
1887 av_free(matroska->packets[n]);
1889 av_freep(&matroska->packets);
1890 matroska->num_packets = 0;
1894 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
1896 uint32_t **lace_buf, int *laces)
1899 uint8_t *data = *buf;
1900 uint32_t *lace_size;
1904 *lace_buf = av_mallocz(sizeof(int));
1906 return AVERROR(ENOMEM);
1908 *lace_buf[0] = size;
1912 av_assert0(size > 0);
1916 lace_size = av_mallocz(*laces * sizeof(int));
1918 return AVERROR(ENOMEM);
1921 case 0x1: /* Xiph lacing */ {
1924 for (n = 0; res == 0 && n < *laces - 1; n++) {
1931 lace_size[n] += temp;
1937 total += lace_size[n];
1939 if (size <= total) {
1940 res = AVERROR_INVALIDDATA;
1944 lace_size[n] = size - total;
1948 case 0x2: /* fixed-size lacing */
1949 if (size % (*laces)) {
1950 res = AVERROR_INVALIDDATA;
1953 for (n = 0; n < *laces; n++)
1954 lace_size[n] = size / *laces;
1957 case 0x3: /* EBML lacing */ {
1960 n = matroska_ebmlnum_uint(matroska, data, size, &num);
1962 av_log(matroska->ctx, AV_LOG_INFO,
1963 "EBML block data error\n");
1969 total = lace_size[0] = num;
1970 for (n = 1; res == 0 && n < *laces - 1; n++) {
1973 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1975 av_log(matroska->ctx, AV_LOG_INFO,
1976 "EBML block data error\n");
1982 lace_size[n] = lace_size[n - 1] + snum;
1983 total += lace_size[n];
1985 if (size <= total) {
1986 res = AVERROR_INVALIDDATA;
1989 lace_size[*laces - 1] = size - total;
1995 *lace_buf = lace_size;
2000 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2001 MatroskaTrack *track,
2003 uint8_t *data, int size,
2007 int a = st->codec->block_align;
2008 int sps = track->audio.sub_packet_size;
2009 int cfs = track->audio.coded_framesize;
2010 int h = track->audio.sub_packet_h;
2011 int y = track->audio.sub_packet_cnt;
2012 int w = track->audio.frame_size;
2015 if (!track->audio.pkt_cnt) {
2016 if (track->audio.sub_packet_cnt == 0)
2017 track->audio.buf_timecode = timecode;
2018 if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2019 if (size < cfs * h / 2) {
2020 av_log(matroska->ctx, AV_LOG_ERROR,
2021 "Corrupt int4 RM-style audio packet size\n");
2022 return AVERROR_INVALIDDATA;
2024 for (x=0; x<h/2; x++)
2025 memcpy(track->audio.buf+x*2*w+y*cfs,
2027 } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2029 av_log(matroska->ctx, AV_LOG_ERROR,
2030 "Corrupt sipr RM-style audio packet size\n");
2031 return AVERROR_INVALIDDATA;
2033 memcpy(track->audio.buf + y*w, data, w);
2035 if (size < sps * w / sps) {
2036 av_log(matroska->ctx, AV_LOG_ERROR,
2037 "Corrupt generic RM-style audio packet size\n");
2038 return AVERROR_INVALIDDATA;
2040 for (x=0; x<w/sps; x++)
2041 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
2044 if (++track->audio.sub_packet_cnt >= h) {
2045 if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2046 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2047 track->audio.sub_packet_cnt = 0;
2048 track->audio.pkt_cnt = h*w / a;
2052 while (track->audio.pkt_cnt) {
2053 AVPacket *pkt = NULL;
2054 if (!(pkt = av_mallocz(sizeof(AVPacket))) || av_new_packet(pkt, a) < 0){
2056 return AVERROR(ENOMEM);
2058 memcpy(pkt->data, track->audio.buf
2059 + a * (h*w / a - track->audio.pkt_cnt--), a);
2060 pkt->pts = track->audio.buf_timecode;
2061 track->audio.buf_timecode = AV_NOPTS_VALUE;
2063 pkt->stream_index = st->index;
2064 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2069 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2070 MatroskaTrack *track,
2072 uint8_t *data, int pkt_size,
2073 uint64_t timecode, uint64_t lace_duration,
2074 int64_t pos, int is_keyframe)
2076 MatroskaTrackEncoding *encodings = track->encodings.elem;
2077 uint8_t *pkt_data = data;
2078 int offset = 0, res;
2081 if (encodings && encodings->scope & 1) {
2082 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2087 if (st->codec->codec_id == AV_CODEC_ID_PRORES)
2090 pkt = av_mallocz(sizeof(AVPacket));
2091 /* XXX: prevent data copy... */
2092 if (av_new_packet(pkt, pkt_size + offset) < 0) {
2094 return AVERROR(ENOMEM);
2097 if (st->codec->codec_id == AV_CODEC_ID_PRORES) {
2098 uint8_t *buf = pkt->data;
2099 bytestream_put_be32(&buf, pkt_size);
2100 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2103 memcpy(pkt->data + offset, pkt_data, pkt_size);
2105 if (pkt_data != data)
2108 pkt->flags = is_keyframe;
2109 pkt->stream_index = st->index;
2111 if (track->ms_compat)
2112 pkt->dts = timecode;
2114 pkt->pts = timecode;
2116 if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2118 * For backward compatibility.
2119 * Historically, we have put subtitle duration
2120 * in convergence_duration, on the off chance
2121 * that the time_scale is less than 1us, which
2122 * could result in a 32bit overflow on the
2123 * normal duration field.
2125 pkt->convergence_duration = lace_duration;
2128 if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE ||
2129 lace_duration <= INT_MAX) {
2131 * For non subtitle tracks, just store the duration
2134 * If it's a subtitle track and duration value does
2135 * not overflow a uint32, then also store it normally.
2137 pkt->duration = lace_duration;
2140 if (st->codec->codec_id == AV_CODEC_ID_SSA)
2141 matroska_fix_ass_packet(matroska, pkt, lace_duration);
2143 if (matroska->prev_pkt &&
2144 timecode != AV_NOPTS_VALUE &&
2145 matroska->prev_pkt->pts == timecode &&
2146 matroska->prev_pkt->stream_index == st->index &&
2147 st->codec->codec_id == AV_CODEC_ID_SSA)
2148 matroska_merge_packets(matroska->prev_pkt, pkt);
2150 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2151 matroska->prev_pkt = pkt;
2157 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
2158 int size, int64_t pos, uint64_t cluster_time,
2159 uint64_t block_duration, int is_keyframe,
2160 int64_t cluster_pos)
2162 uint64_t timecode = AV_NOPTS_VALUE;
2163 MatroskaTrack *track;
2167 uint32_t *lace_size = NULL;
2168 int n, flags, laces = 0;
2171 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2172 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2178 track = matroska_find_track_by_num(matroska, num);
2179 if (!track || !track->stream) {
2180 av_log(matroska->ctx, AV_LOG_INFO,
2181 "Invalid stream %"PRIu64" or size %u\n", num, size);
2182 return AVERROR_INVALIDDATA;
2183 } else if (size <= 3)
2186 if (st->discard >= AVDISCARD_ALL)
2188 av_assert1(block_duration != AV_NOPTS_VALUE);
2190 block_time = AV_RB16(data);
2194 if (is_keyframe == -1)
2195 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2197 if (cluster_time != (uint64_t)-1
2198 && (block_time >= 0 || cluster_time >= -block_time)) {
2199 timecode = cluster_time + block_time;
2200 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
2201 && timecode < track->end_timecode)
2202 is_keyframe = 0; /* overlapping subtitles are not key frame */
2204 av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
2207 if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2208 if (timecode < matroska->skip_to_timecode)
2210 if (!st->skip_to_keyframe) {
2211 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2212 matroska->skip_to_keyframe = 0;
2215 matroska->skip_to_keyframe = 0;
2218 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
2219 &lace_size, &laces);
2224 if (!block_duration)
2225 block_duration = track->default_duration * laces / matroska->time_scale;
2227 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
2228 track->end_timecode =
2229 FFMAX(track->end_timecode, timecode + block_duration);
2231 for (n = 0; n < laces; n++) {
2232 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
2234 if (lace_size[n] > size) {
2235 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2239 if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2240 st->codec->codec_id == AV_CODEC_ID_COOK ||
2241 st->codec->codec_id == AV_CODEC_ID_SIPR ||
2242 st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2243 st->codec->block_align && track->audio.sub_packet_size) {
2245 res = matroska_parse_rm_audio(matroska, track, st, data, size,
2251 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2252 timecode, lace_duration,
2253 pos, !n? is_keyframe : 0);
2258 if (timecode != AV_NOPTS_VALUE)
2259 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2260 data += lace_size[n];
2261 size -= lace_size[n];
2269 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
2271 EbmlList *blocks_list;
2272 MatroskaBlock *blocks;
2274 res = ebml_parse(matroska,
2275 matroska_cluster_incremental_parsing,
2276 &matroska->current_cluster);
2279 if (matroska->current_cluster_pos)
2280 ebml_level_end(matroska);
2281 ebml_free(matroska_cluster, &matroska->current_cluster);
2282 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2283 matroska->current_cluster_num_blocks = 0;
2284 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2285 matroska->prev_pkt = NULL;
2286 /* sizeof the ID which was already read */
2287 if (matroska->current_id)
2288 matroska->current_cluster_pos -= 4;
2289 res = ebml_parse(matroska,
2290 matroska_clusters_incremental,
2291 &matroska->current_cluster);
2292 /* Try parsing the block again. */
2294 res = ebml_parse(matroska,
2295 matroska_cluster_incremental_parsing,
2296 &matroska->current_cluster);
2300 matroska->current_cluster_num_blocks <
2301 matroska->current_cluster.blocks.nb_elem) {
2302 blocks_list = &matroska->current_cluster.blocks;
2303 blocks = blocks_list->elem;
2305 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2306 i = blocks_list->nb_elem - 1;
2307 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2308 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2309 if (!blocks[i].non_simple)
2310 blocks[i].duration = 0;
2311 res = matroska_parse_block(matroska,
2312 blocks[i].bin.data, blocks[i].bin.size,
2314 matroska->current_cluster.timecode,
2315 blocks[i].duration, is_keyframe,
2316 matroska->current_cluster_pos);
2320 if (res < 0) matroska->done = 1;
2324 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
2326 MatroskaCluster cluster = { 0 };
2327 EbmlList *blocks_list;
2328 MatroskaBlock *blocks;
2331 if (!matroska->contains_ssa)
2332 return matroska_parse_cluster_incremental(matroska);
2333 pos = avio_tell(matroska->ctx->pb);
2334 matroska->prev_pkt = NULL;
2335 if (matroska->current_id)
2336 pos -= 4; /* sizeof the ID which was already read */
2337 res = ebml_parse(matroska, matroska_clusters, &cluster);
2338 blocks_list = &cluster.blocks;
2339 blocks = blocks_list->elem;
2340 for (i=0; i<blocks_list->nb_elem; i++)
2341 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2342 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2343 res=matroska_parse_block(matroska,
2344 blocks[i].bin.data, blocks[i].bin.size,
2345 blocks[i].bin.pos, cluster.timecode,
2346 blocks[i].duration, is_keyframe,
2349 ebml_free(matroska_cluster, &cluster);
2353 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
2355 MatroskaDemuxContext *matroska = s->priv_data;
2357 while (matroska_deliver_packet(matroska, pkt)) {
2358 int64_t pos = avio_tell(matroska->ctx->pb);
2361 if (matroska_parse_cluster(matroska) < 0)
2362 matroska_resync(matroska, pos);
2368 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2369 int64_t timestamp, int flags)
2371 MatroskaDemuxContext *matroska = s->priv_data;
2372 MatroskaTrack *tracks = matroska->tracks.elem;
2373 AVStream *st = s->streams[stream_index];
2374 int i, index, index_sub, index_min;
2376 /* Parse the CUES now since we need the index data to seek. */
2377 if (matroska->cues_parsing_deferred > 0) {
2378 matroska->cues_parsing_deferred = 0;
2379 matroska_parse_cues(matroska);
2382 if (!st->nb_index_entries)
2384 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2386 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2387 avio_seek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
2388 matroska->current_id = 0;
2389 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2390 matroska->prev_pkt = NULL;
2391 matroska_clear_queue(matroska);
2392 if (matroska_parse_cluster(matroska) < 0)
2397 matroska_clear_queue(matroska);
2398 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
2402 for (i=0; i < matroska->tracks.nb_elem; i++) {
2403 tracks[i].audio.pkt_cnt = 0;
2404 tracks[i].audio.sub_packet_cnt = 0;
2405 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2406 tracks[i].end_timecode = 0;
2407 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
2408 && tracks[i].stream->discard != AVDISCARD_ALL) {
2409 index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
2411 && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
2412 && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
2413 index_min = index_sub;
2417 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2418 matroska->current_id = 0;
2419 st->skip_to_keyframe =
2420 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
2421 matroska->skip_to_timecode = st->index_entries[index].timestamp;
2423 matroska->num_levels = 0;
2424 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2427 // slightly hackish but allows proper fallback to
2428 // the generic seeking code.
2429 matroska_clear_queue(matroska);
2430 matroska->current_id = 0;
2431 st->skip_to_keyframe =
2432 matroska->skip_to_keyframe = 0;
2434 matroska->num_levels = 0;
2438 static int matroska_read_close(AVFormatContext *s)
2440 MatroskaDemuxContext *matroska = s->priv_data;
2441 MatroskaTrack *tracks = matroska->tracks.elem;
2444 matroska_clear_queue(matroska);
2446 for (n=0; n < matroska->tracks.nb_elem; n++)
2447 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2448 av_free(tracks[n].audio.buf);
2449 ebml_free(matroska_cluster, &matroska->current_cluster);
2450 ebml_free(matroska_segment, matroska);
2455 AVInputFormat ff_matroska_demuxer = {
2456 .name = "matroska,webm",
2457 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
2458 .priv_data_size = sizeof(MatroskaDemuxContext),
2459 .read_probe = matroska_probe,
2460 .read_header = matroska_read_header,
2461 .read_packet = matroska_read_packet,
2462 .read_close = matroska_read_close,
2463 .read_seek = matroska_read_seek,