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);
1081 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1083 olen = pkt_size *= 3;
1084 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1086 result = AVERROR(ENOMEM);
1089 pkt_data = newpktdata;
1090 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1091 } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
1093 result = AVERROR_INVALIDDATA;
1099 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
1100 z_stream zstream = {0};
1101 if (inflateInit(&zstream) != Z_OK)
1103 zstream.next_in = data;
1104 zstream.avail_in = isize;
1107 newpktdata = av_realloc(pkt_data, pkt_size);
1109 inflateEnd(&zstream);
1112 pkt_data = newpktdata;
1113 zstream.avail_out = pkt_size - zstream.total_out;
1114 zstream.next_out = pkt_data + zstream.total_out;
1116 result = inflate(&zstream, Z_NO_FLUSH);
1118 result = Z_MEM_ERROR;
1119 } while (result==Z_OK && pkt_size<10000000);
1120 pkt_size = zstream.total_out;
1121 inflateEnd(&zstream);
1122 if (result != Z_STREAM_END) {
1123 if (result == Z_MEM_ERROR)
1124 result = AVERROR(ENOMEM);
1126 result = AVERROR_INVALIDDATA;
1133 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
1134 bz_stream bzstream = {0};
1135 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1137 bzstream.next_in = data;
1138 bzstream.avail_in = isize;
1141 newpktdata = av_realloc(pkt_data, pkt_size);
1143 BZ2_bzDecompressEnd(&bzstream);
1146 pkt_data = newpktdata;
1147 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1148 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1150 result = BZ2_bzDecompress(&bzstream);
1152 result = BZ_MEM_ERROR;
1153 } while (result==BZ_OK && pkt_size<10000000);
1154 pkt_size = bzstream.total_out_lo32;
1155 BZ2_bzDecompressEnd(&bzstream);
1156 if (result != BZ_STREAM_END) {
1157 if (result == BZ_MEM_ERROR)
1158 result = AVERROR(ENOMEM);
1160 result = AVERROR_INVALIDDATA;
1167 return AVERROR_INVALIDDATA;
1171 *buf_size = pkt_size;
1178 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
1179 AVPacket *pkt, uint64_t display_duration)
1181 char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
1182 for (; *ptr!=',' && ptr<end-1; ptr++);
1186 for (; *ptr!=',' && ptr<end-1; ptr++);
1188 int64_t end_pts = pkt->pts + display_duration;
1189 int sc = matroska->time_scale * pkt->pts / 10000000;
1190 int ec = matroska->time_scale * end_pts / 10000000;
1191 int sh, sm, ss, eh, em, es, len;
1192 sh = sc/360000; sc -= 360000*sh;
1193 sm = sc/ 6000; sc -= 6000*sm;
1194 ss = sc/ 100; sc -= 100*ss;
1195 eh = ec/360000; ec -= 360000*eh;
1196 em = ec/ 6000; ec -= 6000*em;
1197 es = ec/ 100; ec -= 100*es;
1199 len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
1200 if (!(line = av_malloc(len)))
1202 snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
1203 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
1206 pkt->size = strlen(line);
1210 static int matroska_merge_packets(AVPacket *out, AVPacket *in)
1212 int ret = av_grow_packet(out, in->size);
1215 memcpy(out->data + out->size - in->size, in->data, in->size);
1216 av_destruct_packet(in);
1221 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1222 AVDictionary **metadata, char *prefix)
1224 MatroskaTag *tags = list->elem;
1228 for (i=0; i < list->nb_elem; i++) {
1229 const char *lang= (tags[i].lang && strcmp(tags[i].lang, "und")) ? tags[i].lang : NULL;
1231 if (!tags[i].name) {
1232 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1235 if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1236 else av_strlcpy(key, tags[i].name, sizeof(key));
1237 if (tags[i].def || !lang) {
1238 av_dict_set(metadata, key, tags[i].string, 0);
1239 if (tags[i].sub.nb_elem)
1240 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1243 av_strlcat(key, "-", sizeof(key));
1244 av_strlcat(key, lang, sizeof(key));
1245 av_dict_set(metadata, key, tags[i].string, 0);
1246 if (tags[i].sub.nb_elem)
1247 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1250 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1253 static void matroska_convert_tags(AVFormatContext *s)
1255 MatroskaDemuxContext *matroska = s->priv_data;
1256 MatroskaTags *tags = matroska->tags.elem;
1259 for (i=0; i < matroska->tags.nb_elem; i++) {
1260 if (tags[i].target.attachuid) {
1261 MatroskaAttachement *attachment = matroska->attachments.elem;
1262 for (j=0; j<matroska->attachments.nb_elem; j++)
1263 if (attachment[j].uid == tags[i].target.attachuid
1264 && attachment[j].stream)
1265 matroska_convert_tag(s, &tags[i].tag,
1266 &attachment[j].stream->metadata, NULL);
1267 } else if (tags[i].target.chapteruid) {
1268 MatroskaChapter *chapter = matroska->chapters.elem;
1269 for (j=0; j<matroska->chapters.nb_elem; j++)
1270 if (chapter[j].uid == tags[i].target.chapteruid
1271 && chapter[j].chapter)
1272 matroska_convert_tag(s, &tags[i].tag,
1273 &chapter[j].chapter->metadata, NULL);
1274 } else if (tags[i].target.trackuid) {
1275 MatroskaTrack *track = matroska->tracks.elem;
1276 for (j=0; j<matroska->tracks.nb_elem; j++)
1277 if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1278 matroska_convert_tag(s, &tags[i].tag,
1279 &track[j].stream->metadata, NULL);
1281 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1282 tags[i].target.type);
1287 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska, int idx)
1289 EbmlList *seekhead_list = &matroska->seekhead;
1290 MatroskaSeekhead *seekhead = seekhead_list->elem;
1291 uint32_t level_up = matroska->level_up;
1292 int64_t before_pos = avio_tell(matroska->ctx->pb);
1293 uint32_t saved_id = matroska->current_id;
1294 MatroskaLevel level;
1298 if (idx >= seekhead_list->nb_elem
1299 || seekhead[idx].id == MATROSKA_ID_SEEKHEAD
1300 || seekhead[idx].id == MATROSKA_ID_CLUSTER)
1304 offset = seekhead[idx].pos + matroska->segment_start;
1305 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1306 /* We don't want to lose our seekhead level, so we add
1307 * a dummy. This is a crude hack. */
1308 if (matroska->num_levels == EBML_MAX_DEPTH) {
1309 av_log(matroska->ctx, AV_LOG_INFO,
1310 "Max EBML element depth (%d) reached, "
1311 "cannot parse further.\n", EBML_MAX_DEPTH);
1312 ret = AVERROR_INVALIDDATA;
1315 level.length = (uint64_t)-1;
1316 matroska->levels[matroska->num_levels] = level;
1317 matroska->num_levels++;
1318 matroska->current_id = 0;
1320 ret = ebml_parse(matroska, matroska_segment, matroska);
1322 /* remove dummy level */
1323 while (matroska->num_levels) {
1324 uint64_t length = matroska->levels[--matroska->num_levels].length;
1325 if (length == (uint64_t)-1)
1331 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1332 matroska->level_up = level_up;
1333 matroska->current_id = saved_id;
1338 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1340 EbmlList *seekhead_list = &matroska->seekhead;
1341 int64_t before_pos = avio_tell(matroska->ctx->pb);
1344 // we should not do any seeking in the streaming case
1345 if (!matroska->ctx->pb->seekable ||
1346 (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1349 for (i = 0; i < seekhead_list->nb_elem; i++) {
1350 MatroskaSeekhead *seekhead = seekhead_list->elem;
1351 if (seekhead[i].pos <= before_pos)
1354 // defer cues parsing until we actually need cue data.
1355 if (seekhead[i].id == MATROSKA_ID_CUES) {
1356 matroska->cues_parsing_deferred = 1;
1360 if (matroska_parse_seekhead_entry(matroska, i) < 0) {
1361 // mark index as broken
1362 matroska->cues_parsing_deferred = -1;
1368 static void matroska_add_index_entries(MatroskaDemuxContext *matroska) {
1369 EbmlList *index_list;
1370 MatroskaIndex *index;
1371 int index_scale = 1;
1374 index_list = &matroska->index;
1375 index = index_list->elem;
1376 if (index_list->nb_elem
1377 && index[0].time > 1E14/matroska->time_scale) {
1378 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1379 index_scale = matroska->time_scale;
1381 for (i = 0; i < index_list->nb_elem; i++) {
1382 EbmlList *pos_list = &index[i].pos;
1383 MatroskaIndexPos *pos = pos_list->elem;
1384 for (j = 0; j < pos_list->nb_elem; j++) {
1385 MatroskaTrack *track = matroska_find_track_by_num(matroska, pos[j].track);
1386 if (track && track->stream)
1387 av_add_index_entry(track->stream,
1388 pos[j].pos + matroska->segment_start,
1389 index[i].time/index_scale, 0, 0,
1395 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1396 EbmlList *seekhead_list = &matroska->seekhead;
1397 MatroskaSeekhead *seekhead = seekhead_list->elem;
1400 for (i = 0; i < seekhead_list->nb_elem; i++)
1401 if (seekhead[i].id == MATROSKA_ID_CUES)
1403 assert(i <= seekhead_list->nb_elem);
1405 if (matroska_parse_seekhead_entry(matroska, i) < 0)
1406 matroska->cues_parsing_deferred = -1;
1407 matroska_add_index_entries(matroska);
1410 static int matroska_aac_profile(char *codec_id)
1412 static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1415 for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1416 if (strstr(codec_id, aac_profiles[profile]))
1421 static int matroska_aac_sri(int samplerate)
1425 for (sri=0; sri<FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1426 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1431 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1434 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1435 time_t creation_time = date_utc / 1000000000 + 978307200;
1436 struct tm *ptm = gmtime(&creation_time);
1438 strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm);
1439 av_dict_set(metadata, "creation_time", buffer, 0);
1442 static int matroska_read_header(AVFormatContext *s)
1444 MatroskaDemuxContext *matroska = s->priv_data;
1445 EbmlList *attachements_list = &matroska->attachments;
1446 MatroskaAttachement *attachements;
1447 EbmlList *chapters_list = &matroska->chapters;
1448 MatroskaChapter *chapters;
1449 MatroskaTrack *tracks;
1450 uint64_t max_start = 0;
1458 /* First read the EBML header. */
1459 if (ebml_parse(matroska, ebml_syntax, &ebml)
1460 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1461 || ebml.id_length > sizeof(uint32_t) || ebml.doctype_version > 3 || !ebml.doctype) {
1462 av_log(matroska->ctx, AV_LOG_ERROR,
1463 "EBML header using unsupported features\n"
1464 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1465 ebml.version, ebml.doctype, ebml.doctype_version);
1466 ebml_free(ebml_syntax, &ebml);
1467 return AVERROR_PATCHWELCOME;
1468 } else if (ebml.doctype_version == 3) {
1469 av_log(matroska->ctx, AV_LOG_WARNING,
1470 "EBML header using unsupported features\n"
1471 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1472 ebml.version, ebml.doctype, ebml.doctype_version);
1474 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
1475 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
1477 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
1478 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
1479 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
1480 ebml_free(ebml_syntax, &ebml);
1481 return AVERROR_INVALIDDATA;
1484 ebml_free(ebml_syntax, &ebml);
1486 /* The next thing is a segment. */
1487 pos = avio_tell(matroska->ctx->pb);
1488 res = ebml_parse(matroska, matroska_segments, matroska);
1489 // try resyncing until we find a EBML_STOP type element.
1491 res = matroska_resync(matroska, pos);
1494 pos = avio_tell(matroska->ctx->pb);
1495 res = ebml_parse(matroska, matroska_segment, matroska);
1497 matroska_execute_seekhead(matroska);
1499 if (!matroska->time_scale)
1500 matroska->time_scale = 1000000;
1501 if (matroska->duration)
1502 matroska->ctx->duration = matroska->duration * matroska->time_scale
1503 * 1000 / AV_TIME_BASE;
1504 av_dict_set(&s->metadata, "title", matroska->title, 0);
1506 if (matroska->date_utc.size == 8)
1507 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
1509 tracks = matroska->tracks.elem;
1510 for (i=0; i < matroska->tracks.nb_elem; i++) {
1511 MatroskaTrack *track = &tracks[i];
1512 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1513 EbmlList *encodings_list = &track->encodings;
1514 MatroskaTrackEncoding *encodings = encodings_list->elem;
1515 uint8_t *extradata = NULL;
1516 int extradata_size = 0;
1517 int extradata_offset = 0;
1518 uint32_t fourcc = 0;
1521 /* Apply some sanity checks. */
1522 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1523 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1524 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1525 av_log(matroska->ctx, AV_LOG_INFO,
1526 "Unknown or unsupported track type %"PRIu64"\n",
1530 if (track->codec_id == NULL)
1533 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1534 if (!track->default_duration && track->video.frame_rate > 0)
1535 track->default_duration = 1000000000/track->video.frame_rate;
1536 if (!track->video.display_width)
1537 track->video.display_width = track->video.pixel_width;
1538 if (!track->video.display_height)
1539 track->video.display_height = track->video.pixel_height;
1540 if (track->video.color_space.size == 4)
1541 fourcc = AV_RL32(track->video.color_space.data);
1542 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1543 if (!track->audio.out_samplerate)
1544 track->audio.out_samplerate = track->audio.samplerate;
1546 if (encodings_list->nb_elem > 1) {
1547 av_log(matroska->ctx, AV_LOG_ERROR,
1548 "Multiple combined encodings not supported");
1549 } else if (encodings_list->nb_elem == 1) {
1550 if (encodings[0].type ||
1551 (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1553 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1556 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1558 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
1559 encodings[0].scope = 0;
1560 av_log(matroska->ctx, AV_LOG_ERROR,
1561 "Unsupported encoding type");
1562 } else if (track->codec_priv.size && encodings[0].scope&2) {
1563 uint8_t *codec_priv = track->codec_priv.data;
1564 int ret = matroska_decode_buffer(&track->codec_priv.data,
1565 &track->codec_priv.size,
1568 track->codec_priv.data = NULL;
1569 track->codec_priv.size = 0;
1570 av_log(matroska->ctx, AV_LOG_ERROR,
1571 "Failed to decode codec private data\n");
1574 if (codec_priv != track->codec_priv.data)
1575 av_free(codec_priv);
1579 for(j=0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++){
1580 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1581 strlen(ff_mkv_codec_tags[j].str))){
1582 codec_id= ff_mkv_codec_tags[j].id;
1587 st = track->stream = avformat_new_stream(s, NULL);
1589 return AVERROR(ENOMEM);
1591 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1592 && track->codec_priv.size >= 40
1593 && track->codec_priv.data != NULL) {
1594 track->ms_compat = 1;
1595 fourcc = AV_RL32(track->codec_priv.data + 16);
1596 codec_id = ff_codec_get_id(ff_codec_bmp_tags, fourcc);
1597 extradata_offset = 40;
1598 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1599 && track->codec_priv.size >= 14
1600 && track->codec_priv.data != NULL) {
1602 ffio_init_context(&b, track->codec_priv.data, track->codec_priv.size,
1603 AVIO_FLAG_READ, NULL, NULL, NULL, NULL);
1604 ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1607 codec_id = st->codec->codec_id;
1608 extradata_offset = FFMIN(track->codec_priv.size, 18);
1609 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1610 && (track->codec_priv.size >= 86)
1611 && (track->codec_priv.data != NULL)) {
1612 fourcc = AV_RL32(track->codec_priv.data);
1613 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1614 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX-12) {
1615 /* Only ALAC's magic cookie is stored in Matroska's track headers.
1616 Create the "atom size", "tag", and "tag version" fields the
1617 decoder expects manually. */
1618 extradata_size = 12 + track->codec_priv.size;
1619 extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1620 if (extradata == NULL)
1621 return AVERROR(ENOMEM);
1622 AV_WB32(extradata, extradata_size);
1623 memcpy(&extradata[4], "alac", 4);
1624 AV_WB32(&extradata[8], 0);
1625 memcpy(&extradata[12], track->codec_priv.data, track->codec_priv.size);
1626 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1627 switch (track->audio.bitdepth) {
1628 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1629 case 24: codec_id = AV_CODEC_ID_PCM_S24BE; break;
1630 case 32: codec_id = AV_CODEC_ID_PCM_S32BE; break;
1632 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1633 switch (track->audio.bitdepth) {
1634 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1635 case 24: codec_id = AV_CODEC_ID_PCM_S24LE; break;
1636 case 32: codec_id = AV_CODEC_ID_PCM_S32LE; break;
1638 } else if (codec_id==AV_CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1639 codec_id = AV_CODEC_ID_PCM_F64LE;
1640 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1641 int profile = matroska_aac_profile(track->codec_id);
1642 int sri = matroska_aac_sri(track->audio.samplerate);
1643 extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1644 if (extradata == NULL)
1645 return AVERROR(ENOMEM);
1646 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1647 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1648 if (strstr(track->codec_id, "SBR")) {
1649 sri = matroska_aac_sri(track->audio.out_samplerate);
1650 extradata[2] = 0x56;
1651 extradata[3] = 0xE5;
1652 extradata[4] = 0x80 | (sri<<3);
1656 } else if (codec_id == AV_CODEC_ID_TTA) {
1657 extradata_size = 30;
1658 extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1659 if (extradata == NULL)
1660 return AVERROR(ENOMEM);
1661 ffio_init_context(&b, extradata, extradata_size, 1,
1662 NULL, NULL, NULL, NULL);
1663 avio_write(&b, "TTA1", 4);
1665 avio_wl16(&b, track->audio.channels);
1666 avio_wl16(&b, track->audio.bitdepth);
1667 avio_wl32(&b, track->audio.out_samplerate);
1668 avio_wl32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1669 } else if (codec_id == AV_CODEC_ID_RV10 || codec_id == AV_CODEC_ID_RV20 ||
1670 codec_id == AV_CODEC_ID_RV30 || codec_id == AV_CODEC_ID_RV40) {
1671 extradata_offset = 26;
1672 } else if (codec_id == AV_CODEC_ID_RA_144) {
1673 track->audio.out_samplerate = 8000;
1674 track->audio.channels = 1;
1675 } else if ((codec_id == AV_CODEC_ID_RA_288 || codec_id == AV_CODEC_ID_COOK ||
1676 codec_id == AV_CODEC_ID_ATRAC3 || codec_id == AV_CODEC_ID_SIPR)
1677 && track->codec_priv.data) {
1680 ffio_init_context(&b, track->codec_priv.data,track->codec_priv.size,
1681 0, NULL, NULL, NULL, NULL);
1683 flavor = avio_rb16(&b);
1684 track->audio.coded_framesize = avio_rb32(&b);
1686 track->audio.sub_packet_h = avio_rb16(&b);
1687 track->audio.frame_size = avio_rb16(&b);
1688 track->audio.sub_packet_size = avio_rb16(&b);
1689 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1690 if (codec_id == AV_CODEC_ID_RA_288) {
1691 st->codec->block_align = track->audio.coded_framesize;
1692 track->codec_priv.size = 0;
1694 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1695 const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1696 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1697 st->codec->bit_rate = sipr_bit_rate[flavor];
1699 st->codec->block_align = track->audio.sub_packet_size;
1700 extradata_offset = 78;
1703 track->codec_priv.size -= extradata_offset;
1705 if (codec_id == AV_CODEC_ID_NONE)
1706 av_log(matroska->ctx, AV_LOG_INFO,
1707 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1709 if (track->time_scale < 0.01)
1710 track->time_scale = 1.0;
1711 avpriv_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1713 st->codec->codec_id = codec_id;
1715 if (strcmp(track->language, "und"))
1716 av_dict_set(&st->metadata, "language", track->language, 0);
1717 av_dict_set(&st->metadata, "title", track->name, 0);
1719 if (track->flag_default)
1720 st->disposition |= AV_DISPOSITION_DEFAULT;
1721 if (track->flag_forced)
1722 st->disposition |= AV_DISPOSITION_FORCED;
1724 if (!st->codec->extradata) {
1726 st->codec->extradata = extradata;
1727 st->codec->extradata_size = extradata_size;
1728 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1729 st->codec->extradata = av_mallocz(track->codec_priv.size +
1730 FF_INPUT_BUFFER_PADDING_SIZE);
1731 if(st->codec->extradata == NULL)
1732 return AVERROR(ENOMEM);
1733 st->codec->extradata_size = track->codec_priv.size;
1734 memcpy(st->codec->extradata,
1735 track->codec_priv.data + extradata_offset,
1736 track->codec_priv.size);
1740 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1741 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
1743 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1744 st->codec->codec_tag = fourcc;
1745 st->codec->width = track->video.pixel_width;
1746 st->codec->height = track->video.pixel_height;
1747 av_reduce(&st->sample_aspect_ratio.num,
1748 &st->sample_aspect_ratio.den,
1749 st->codec->height * track->video.display_width,
1750 st->codec-> width * track->video.display_height,
1752 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1753 if (track->default_duration) {
1754 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1755 1000000000, track->default_duration, 30000);
1756 #if FF_API_R_FRAME_RATE
1757 st->r_frame_rate = st->avg_frame_rate;
1761 /* export stereo mode flag as metadata tag */
1762 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREO_MODE_COUNT)
1763 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
1765 /* if we have virtual track, mark the real tracks */
1766 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
1768 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
1770 snprintf(buf, sizeof(buf), "%s_%d",
1771 ff_matroska_video_stereo_plane[planes[j].type], i);
1772 for (k=0; k < matroska->tracks.nb_elem; k++)
1773 if (planes[j].uid == tracks[k].uid) {
1774 av_dict_set(&s->streams[k]->metadata,
1775 "stereo_mode", buf, 0);
1779 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1780 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1781 st->codec->sample_rate = track->audio.out_samplerate;
1782 st->codec->channels = track->audio.channels;
1783 if (st->codec->codec_id != AV_CODEC_ID_AAC)
1784 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1785 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1786 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1787 if (st->codec->codec_id == AV_CODEC_ID_SSA)
1788 matroska->contains_ssa = 1;
1792 attachements = attachements_list->elem;
1793 for (j=0; j<attachements_list->nb_elem; j++) {
1794 if (!(attachements[j].filename && attachements[j].mime &&
1795 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1796 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1798 AVStream *st = avformat_new_stream(s, NULL);
1801 av_dict_set(&st->metadata, "filename",attachements[j].filename, 0);
1802 av_dict_set(&st->metadata, "mimetype", attachements[j].mime, 0);
1803 st->codec->codec_id = AV_CODEC_ID_NONE;
1804 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
1805 st->codec->extradata = av_malloc(attachements[j].bin.size + FF_INPUT_BUFFER_PADDING_SIZE);
1806 if(st->codec->extradata == NULL)
1808 st->codec->extradata_size = attachements[j].bin.size;
1809 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1811 for (i=0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
1812 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1813 strlen(ff_mkv_mime_tags[i].str))) {
1814 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1818 attachements[j].stream = st;
1822 chapters = chapters_list->elem;
1823 for (i=0; i<chapters_list->nb_elem; i++)
1824 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1825 && (max_start==0 || chapters[i].start > max_start)) {
1826 chapters[i].chapter =
1827 avpriv_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1828 chapters[i].start, chapters[i].end,
1830 av_dict_set(&chapters[i].chapter->metadata,
1831 "title", chapters[i].title, 0);
1832 max_start = chapters[i].start;
1835 matroska_add_index_entries(matroska);
1837 matroska_convert_tags(s);
1843 * Put one packet in an application-supplied AVPacket struct.
1844 * Returns 0 on success or -1 on failure.
1846 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1849 if (matroska->num_packets > 0) {
1850 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1851 av_free(matroska->packets[0]);
1852 if (matroska->num_packets > 1) {
1854 memmove(&matroska->packets[0], &matroska->packets[1],
1855 (matroska->num_packets - 1) * sizeof(AVPacket *));
1856 newpackets = av_realloc(matroska->packets,
1857 (matroska->num_packets - 1) * sizeof(AVPacket *));
1859 matroska->packets = newpackets;
1861 av_freep(&matroska->packets);
1862 matroska->prev_pkt = NULL;
1864 matroska->num_packets--;
1872 * Free all packets in our internal queue.
1874 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1876 if (matroska->packets) {
1878 for (n = 0; n < matroska->num_packets; n++) {
1879 av_free_packet(matroska->packets[n]);
1880 av_free(matroska->packets[n]);
1882 av_freep(&matroska->packets);
1883 matroska->num_packets = 0;
1887 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
1889 uint32_t **lace_buf, int *laces)
1892 uint8_t *data = *buf;
1893 uint32_t *lace_size;
1897 *lace_buf = av_mallocz(sizeof(int));
1899 return AVERROR(ENOMEM);
1901 *lace_buf[0] = size;
1905 av_assert0(size > 0);
1909 lace_size = av_mallocz(*laces * sizeof(int));
1911 return AVERROR(ENOMEM);
1914 case 0x1: /* Xiph lacing */ {
1917 for (n = 0; res == 0 && n < *laces - 1; n++) {
1924 lace_size[n] += temp;
1930 total += lace_size[n];
1932 if (size <= total) {
1933 res = AVERROR_INVALIDDATA;
1937 lace_size[n] = size - total;
1941 case 0x2: /* fixed-size lacing */
1942 if (size % (*laces)) {
1943 res = AVERROR_INVALIDDATA;
1946 for (n = 0; n < *laces; n++)
1947 lace_size[n] = size / *laces;
1950 case 0x3: /* EBML lacing */ {
1953 n = matroska_ebmlnum_uint(matroska, data, size, &num);
1955 av_log(matroska->ctx, AV_LOG_INFO,
1956 "EBML block data error\n");
1962 total = lace_size[0] = num;
1963 for (n = 1; res == 0 && n < *laces - 1; n++) {
1966 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1968 av_log(matroska->ctx, AV_LOG_INFO,
1969 "EBML block data error\n");
1975 lace_size[n] = lace_size[n - 1] + snum;
1976 total += lace_size[n];
1978 if (size <= total) {
1979 res = AVERROR_INVALIDDATA;
1982 lace_size[*laces - 1] = size - total;
1988 *lace_buf = lace_size;
1993 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
1994 MatroskaTrack *track,
1996 uint8_t *data, int size,
2000 int a = st->codec->block_align;
2001 int sps = track->audio.sub_packet_size;
2002 int cfs = track->audio.coded_framesize;
2003 int h = track->audio.sub_packet_h;
2004 int y = track->audio.sub_packet_cnt;
2005 int w = track->audio.frame_size;
2008 if (!track->audio.pkt_cnt) {
2009 if (track->audio.sub_packet_cnt == 0)
2010 track->audio.buf_timecode = timecode;
2011 if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2012 if (size < cfs * h / 2) {
2013 av_log(matroska->ctx, AV_LOG_ERROR,
2014 "Corrupt int4 RM-style audio packet size\n");
2015 return AVERROR_INVALIDDATA;
2017 for (x=0; x<h/2; x++)
2018 memcpy(track->audio.buf+x*2*w+y*cfs,
2020 } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2022 av_log(matroska->ctx, AV_LOG_ERROR,
2023 "Corrupt sipr RM-style audio packet size\n");
2024 return AVERROR_INVALIDDATA;
2026 memcpy(track->audio.buf + y*w, data, w);
2028 if (size < sps * w / sps) {
2029 av_log(matroska->ctx, AV_LOG_ERROR,
2030 "Corrupt generic RM-style audio packet size\n");
2031 return AVERROR_INVALIDDATA;
2033 for (x=0; x<w/sps; x++)
2034 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
2037 if (++track->audio.sub_packet_cnt >= h) {
2038 if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2039 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2040 track->audio.sub_packet_cnt = 0;
2041 track->audio.pkt_cnt = h*w / a;
2045 while (track->audio.pkt_cnt) {
2046 AVPacket *pkt = av_mallocz(sizeof(AVPacket));
2047 av_new_packet(pkt, a);
2048 memcpy(pkt->data, track->audio.buf
2049 + a * (h*w / a - track->audio.pkt_cnt--), a);
2050 pkt->pts = track->audio.buf_timecode;
2051 track->audio.buf_timecode = AV_NOPTS_VALUE;
2053 pkt->stream_index = st->index;
2054 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2059 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2060 MatroskaTrack *track,
2062 uint8_t *data, int pkt_size,
2063 uint64_t timecode, uint64_t lace_duration,
2064 int64_t pos, int is_keyframe)
2066 MatroskaTrackEncoding *encodings = track->encodings.elem;
2067 uint8_t *pkt_data = data;
2068 int offset = 0, res;
2071 if (encodings && encodings->scope & 1) {
2072 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2077 if (st->codec->codec_id == AV_CODEC_ID_PRORES)
2080 pkt = av_mallocz(sizeof(AVPacket));
2081 /* XXX: prevent data copy... */
2082 if (av_new_packet(pkt, pkt_size + offset) < 0) {
2084 return AVERROR(ENOMEM);
2087 if (st->codec->codec_id == AV_CODEC_ID_PRORES) {
2088 uint8_t *buf = pkt->data;
2089 bytestream_put_be32(&buf, pkt_size);
2090 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2093 memcpy(pkt->data + offset, pkt_data, pkt_size);
2095 if (pkt_data != data)
2098 pkt->flags = is_keyframe;
2099 pkt->stream_index = st->index;
2101 if (track->ms_compat)
2102 pkt->dts = timecode;
2104 pkt->pts = timecode;
2106 if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2108 * For backward compatibility.
2109 * Historically, we have put subtitle duration
2110 * in convergence_duration, on the off chance
2111 * that the time_scale is less than 1us, which
2112 * could result in a 32bit overflow on the
2113 * normal duration field.
2115 pkt->convergence_duration = lace_duration;
2118 if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE ||
2119 lace_duration <= INT_MAX) {
2121 * For non subtitle tracks, just store the duration
2124 * If it's a subtitle track and duration value does
2125 * not overflow a uint32, then also store it normally.
2127 pkt->duration = lace_duration;
2130 if (st->codec->codec_id == AV_CODEC_ID_SSA)
2131 matroska_fix_ass_packet(matroska, pkt, lace_duration);
2133 if (matroska->prev_pkt &&
2134 timecode != AV_NOPTS_VALUE &&
2135 matroska->prev_pkt->pts == timecode &&
2136 matroska->prev_pkt->stream_index == st->index &&
2137 st->codec->codec_id == AV_CODEC_ID_SSA)
2138 matroska_merge_packets(matroska->prev_pkt, pkt);
2140 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2141 matroska->prev_pkt = pkt;
2147 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
2148 int size, int64_t pos, uint64_t cluster_time,
2149 uint64_t block_duration, int is_keyframe,
2150 int64_t cluster_pos)
2152 uint64_t timecode = AV_NOPTS_VALUE;
2153 MatroskaTrack *track;
2157 uint32_t *lace_size = NULL;
2158 int n, flags, laces = 0;
2161 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2162 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2168 track = matroska_find_track_by_num(matroska, num);
2169 if (!track || !track->stream) {
2170 av_log(matroska->ctx, AV_LOG_INFO,
2171 "Invalid stream %"PRIu64" or size %u\n", num, size);
2172 return AVERROR_INVALIDDATA;
2173 } else if (size <= 3)
2176 if (st->discard >= AVDISCARD_ALL)
2178 av_assert1(block_duration != AV_NOPTS_VALUE);
2180 block_time = AV_RB16(data);
2184 if (is_keyframe == -1)
2185 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2187 if (cluster_time != (uint64_t)-1
2188 && (block_time >= 0 || cluster_time >= -block_time)) {
2189 timecode = cluster_time + block_time;
2190 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
2191 && timecode < track->end_timecode)
2192 is_keyframe = 0; /* overlapping subtitles are not key frame */
2194 av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
2197 if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2198 if (timecode < matroska->skip_to_timecode)
2200 if (!st->skip_to_keyframe) {
2201 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2202 matroska->skip_to_keyframe = 0;
2205 matroska->skip_to_keyframe = 0;
2208 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
2209 &lace_size, &laces);
2214 if (!block_duration)
2215 block_duration = track->default_duration * laces / matroska->time_scale;
2217 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
2218 track->end_timecode =
2219 FFMAX(track->end_timecode, timecode + block_duration);
2221 for (n = 0; n < laces; n++) {
2222 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
2224 if (lace_size[n] > size) {
2225 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2229 if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2230 st->codec->codec_id == AV_CODEC_ID_COOK ||
2231 st->codec->codec_id == AV_CODEC_ID_SIPR ||
2232 st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2233 st->codec->block_align && track->audio.sub_packet_size) {
2235 res = matroska_parse_rm_audio(matroska, track, st, data, size,
2241 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2242 timecode, lace_duration,
2243 pos, !n? is_keyframe : 0);
2248 if (timecode != AV_NOPTS_VALUE)
2249 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2250 data += lace_size[n];
2251 size -= lace_size[n];
2259 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
2261 EbmlList *blocks_list;
2262 MatroskaBlock *blocks;
2264 res = ebml_parse(matroska,
2265 matroska_cluster_incremental_parsing,
2266 &matroska->current_cluster);
2269 if (matroska->current_cluster_pos)
2270 ebml_level_end(matroska);
2271 ebml_free(matroska_cluster, &matroska->current_cluster);
2272 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2273 matroska->current_cluster_num_blocks = 0;
2274 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2275 matroska->prev_pkt = NULL;
2276 /* sizeof the ID which was already read */
2277 if (matroska->current_id)
2278 matroska->current_cluster_pos -= 4;
2279 res = ebml_parse(matroska,
2280 matroska_clusters_incremental,
2281 &matroska->current_cluster);
2282 /* Try parsing the block again. */
2284 res = ebml_parse(matroska,
2285 matroska_cluster_incremental_parsing,
2286 &matroska->current_cluster);
2290 matroska->current_cluster_num_blocks <
2291 matroska->current_cluster.blocks.nb_elem) {
2292 blocks_list = &matroska->current_cluster.blocks;
2293 blocks = blocks_list->elem;
2295 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2296 i = blocks_list->nb_elem - 1;
2297 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2298 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2299 if (!blocks[i].non_simple)
2300 blocks[i].duration = 0;
2301 res = matroska_parse_block(matroska,
2302 blocks[i].bin.data, blocks[i].bin.size,
2304 matroska->current_cluster.timecode,
2305 blocks[i].duration, is_keyframe,
2306 matroska->current_cluster_pos);
2310 if (res < 0) matroska->done = 1;
2314 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
2316 MatroskaCluster cluster = { 0 };
2317 EbmlList *blocks_list;
2318 MatroskaBlock *blocks;
2321 if (!matroska->contains_ssa)
2322 return matroska_parse_cluster_incremental(matroska);
2323 pos = avio_tell(matroska->ctx->pb);
2324 matroska->prev_pkt = NULL;
2325 if (matroska->current_id)
2326 pos -= 4; /* sizeof the ID which was already read */
2327 res = ebml_parse(matroska, matroska_clusters, &cluster);
2328 blocks_list = &cluster.blocks;
2329 blocks = blocks_list->elem;
2330 for (i=0; i<blocks_list->nb_elem; i++)
2331 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2332 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2333 res=matroska_parse_block(matroska,
2334 blocks[i].bin.data, blocks[i].bin.size,
2335 blocks[i].bin.pos, cluster.timecode,
2336 blocks[i].duration, is_keyframe,
2339 ebml_free(matroska_cluster, &cluster);
2343 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
2345 MatroskaDemuxContext *matroska = s->priv_data;
2347 while (matroska_deliver_packet(matroska, pkt)) {
2348 int64_t pos = avio_tell(matroska->ctx->pb);
2351 if (matroska_parse_cluster(matroska) < 0)
2352 matroska_resync(matroska, pos);
2358 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2359 int64_t timestamp, int flags)
2361 MatroskaDemuxContext *matroska = s->priv_data;
2362 MatroskaTrack *tracks = matroska->tracks.elem;
2363 AVStream *st = s->streams[stream_index];
2364 int i, index, index_sub, index_min;
2366 /* Parse the CUES now since we need the index data to seek. */
2367 if (matroska->cues_parsing_deferred > 0) {
2368 matroska->cues_parsing_deferred = 0;
2369 matroska_parse_cues(matroska);
2372 if (!st->nb_index_entries)
2374 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2376 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2377 avio_seek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
2378 matroska->current_id = 0;
2379 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2380 matroska->prev_pkt = NULL;
2381 matroska_clear_queue(matroska);
2382 if (matroska_parse_cluster(matroska) < 0)
2387 matroska_clear_queue(matroska);
2388 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
2392 for (i=0; i < matroska->tracks.nb_elem; i++) {
2393 tracks[i].audio.pkt_cnt = 0;
2394 tracks[i].audio.sub_packet_cnt = 0;
2395 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2396 tracks[i].end_timecode = 0;
2397 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
2398 && !tracks[i].stream->discard != AVDISCARD_ALL) {
2399 index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
2401 && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
2402 && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
2403 index_min = index_sub;
2407 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2408 matroska->current_id = 0;
2409 st->skip_to_keyframe =
2410 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
2411 matroska->skip_to_timecode = st->index_entries[index].timestamp;
2413 matroska->num_levels = 0;
2414 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2417 // slightly hackish but allows proper fallback to
2418 // the generic seeking code.
2419 matroska_clear_queue(matroska);
2420 matroska->current_id = 0;
2421 st->skip_to_keyframe =
2422 matroska->skip_to_keyframe = 0;
2424 matroska->num_levels = 0;
2428 static int matroska_read_close(AVFormatContext *s)
2430 MatroskaDemuxContext *matroska = s->priv_data;
2431 MatroskaTrack *tracks = matroska->tracks.elem;
2434 matroska_clear_queue(matroska);
2436 for (n=0; n < matroska->tracks.nb_elem; n++)
2437 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2438 av_free(tracks[n].audio.buf);
2439 ebml_free(matroska_cluster, &matroska->current_cluster);
2440 ebml_free(matroska_segment, matroska);
2445 AVInputFormat ff_matroska_demuxer = {
2446 .name = "matroska,webm",
2447 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
2448 .priv_data_size = sizeof(MatroskaDemuxContext),
2449 .read_probe = matroska_probe,
2450 .read_header = matroska_read_header,
2451 .read_packet = matroska_read_packet,
2452 .read_close = matroska_read_close,
2453 .read_seek = matroska_read_seek,