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"
36 /* For ff_codec_get_id(). */
41 #include "libavcodec/bytestream.h"
42 #include "libavcodec/mpeg4audio.h"
43 #include "libavutil/intfloat.h"
44 #include "libavutil/intreadwrite.h"
45 #include "libavutil/avstring.h"
46 #include "libavutil/lzo.h"
47 #include "libavutil/dict.h"
68 typedef const struct EbmlSyntax {
77 const struct EbmlSyntax *n;
97 uint64_t doctype_version;
103 } MatroskaTrackCompression;
108 MatroskaTrackCompression compression;
109 } MatroskaTrackEncoding;
113 uint64_t display_width;
114 uint64_t display_height;
115 uint64_t pixel_width;
116 uint64_t pixel_height;
118 uint64_t stereo_mode;
119 } MatroskaTrackVideo;
123 double out_samplerate;
127 /* real audio header (extracted from extradata) */
134 uint64_t buf_timecode;
136 } MatroskaTrackAudio;
141 } MatroskaTrackPlane;
144 EbmlList combine_planes;
145 } MatroskaTrackOperation;
156 uint64_t default_duration;
157 uint64_t flag_default;
158 uint64_t flag_forced;
159 MatroskaTrackVideo video;
160 MatroskaTrackAudio audio;
161 MatroskaTrackOperation operation;
165 int64_t end_timecode;
167 uint64_t max_block_additional_id;
177 } MatroskaAttachement;
215 MatroskaTagTarget target;
235 AVFormatContext *ctx;
239 MatroskaLevel levels[EBML_MAX_DEPTH];
248 EbmlList attachments;
254 /* byte position of the segment inside the stream */
255 int64_t segment_start;
257 /* the packet queue */
264 /* What to skip before effectively reading a packet. */
265 int skip_to_keyframe;
266 uint64_t skip_to_timecode;
268 /* File has a CUES element, but we defer parsing until it is needed. */
269 int cues_parsing_deferred;
271 int current_cluster_num_blocks;
272 int64_t current_cluster_pos;
273 MatroskaCluster current_cluster;
275 /* File has SSA subtitles which prevent incremental cluster parsing. */
277 } MatroskaDemuxContext;
284 uint64_t additional_id;
288 static EbmlSyntax ebml_header[] = {
289 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
290 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
291 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
292 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
293 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
294 { EBML_ID_EBMLVERSION, EBML_NONE },
295 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
299 static EbmlSyntax ebml_syntax[] = {
300 { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
304 static EbmlSyntax matroska_info[] = {
305 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
306 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
307 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
308 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
309 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
310 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext,date_utc) },
311 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
315 static EbmlSyntax matroska_track_video[] = {
316 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
317 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
318 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
319 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
320 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
321 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo,color_space) },
322 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,stereo_mode) },
323 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
324 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
325 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
326 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
327 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
328 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
329 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
333 static EbmlSyntax matroska_track_audio[] = {
334 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
335 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
336 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
337 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
341 static EbmlSyntax matroska_track_encoding_compression[] = {
342 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
343 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
347 static EbmlSyntax matroska_track_encoding[] = {
348 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
349 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
350 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
351 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
355 static EbmlSyntax matroska_track_encodings[] = {
356 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
360 static EbmlSyntax matroska_track_plane[] = {
361 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
362 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
366 static EbmlSyntax matroska_track_combine_planes[] = {
367 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n=matroska_track_plane} },
371 static EbmlSyntax matroska_track_operation[] = {
372 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n=matroska_track_combine_planes} },
376 static EbmlSyntax matroska_track[] = {
377 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
378 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack,name) },
379 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
380 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
381 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
382 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
383 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
384 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
385 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
386 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
387 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack,flag_forced), {.u=0} },
388 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
389 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
390 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack,operation), {.n=matroska_track_operation} },
391 { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
392 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack,max_block_additional_id) },
393 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
394 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
395 { MATROSKA_ID_CODECNAME, EBML_NONE },
396 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
397 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
398 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
399 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
400 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
404 static EbmlSyntax matroska_tracks[] = {
405 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
409 static EbmlSyntax matroska_attachment[] = {
410 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
411 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
412 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
413 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
414 { MATROSKA_ID_FILEDESC, EBML_NONE },
418 static EbmlSyntax matroska_attachments[] = {
419 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
423 static EbmlSyntax matroska_chapter_display[] = {
424 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
425 { MATROSKA_ID_CHAPLANG, EBML_NONE },
429 static EbmlSyntax matroska_chapter_entry[] = {
430 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
431 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
432 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
433 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
434 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
435 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
436 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
437 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
441 static EbmlSyntax matroska_chapter[] = {
442 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
443 { MATROSKA_ID_EDITIONUID, EBML_NONE },
444 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
445 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
446 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
450 static EbmlSyntax matroska_chapters[] = {
451 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
455 static EbmlSyntax matroska_index_pos[] = {
456 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
457 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
458 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
462 static EbmlSyntax matroska_index_entry[] = {
463 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
464 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
468 static EbmlSyntax matroska_index[] = {
469 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
473 static EbmlSyntax matroska_simpletag[] = {
474 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
475 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
476 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag,lang), {.s="und"} },
477 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag,def) },
478 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag,def) },
479 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
483 static EbmlSyntax matroska_tagtargets[] = {
484 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget,type) },
485 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
486 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
487 { MATROSKA_ID_TAGTARGETS_CHAPTERUID,EBML_UINT, 0, offsetof(MatroskaTagTarget,chapteruid) },
488 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
492 static EbmlSyntax matroska_tag[] = {
493 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
494 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
498 static EbmlSyntax matroska_tags[] = {
499 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
503 static EbmlSyntax matroska_seekhead_entry[] = {
504 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
505 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
509 static EbmlSyntax matroska_seekhead[] = {
510 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
514 static EbmlSyntax matroska_segment[] = {
515 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
516 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
517 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
518 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
519 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
520 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
521 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
522 { MATROSKA_ID_CLUSTER, EBML_STOP },
526 static EbmlSyntax matroska_segments[] = {
527 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
531 static EbmlSyntax matroska_blockmore[] = {
532 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
533 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
537 static EbmlSyntax matroska_blockadditions[] = {
538 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n=matroska_blockmore} },
542 static EbmlSyntax matroska_blockgroup[] = {
543 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
544 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, {.n=matroska_blockadditions} },
545 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
546 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration) },
547 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
548 { 1, EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
552 static EbmlSyntax matroska_cluster[] = {
553 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
554 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
555 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
556 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
557 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
561 static EbmlSyntax matroska_clusters[] = {
562 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
563 { MATROSKA_ID_INFO, EBML_NONE },
564 { MATROSKA_ID_CUES, EBML_NONE },
565 { MATROSKA_ID_TAGS, EBML_NONE },
566 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
570 static EbmlSyntax matroska_cluster_incremental_parsing[] = {
571 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
572 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
573 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
574 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
575 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
576 { MATROSKA_ID_INFO, EBML_NONE },
577 { MATROSKA_ID_CUES, EBML_NONE },
578 { MATROSKA_ID_TAGS, EBML_NONE },
579 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
580 { MATROSKA_ID_CLUSTER, EBML_STOP },
584 static EbmlSyntax matroska_cluster_incremental[] = {
585 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
586 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
587 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
588 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
589 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
593 static EbmlSyntax matroska_clusters_incremental[] = {
594 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster_incremental} },
595 { MATROSKA_ID_INFO, EBML_NONE },
596 { MATROSKA_ID_CUES, EBML_NONE },
597 { MATROSKA_ID_TAGS, EBML_NONE },
598 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
602 static const char *const matroska_doctypes[] = { "matroska", "webm" };
604 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
606 AVIOContext *pb = matroska->ctx->pb;
608 matroska->current_id = 0;
609 matroska->num_levels = 0;
611 // seek to next position to resync from
612 if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0 || avio_tell(pb) <= last_pos)
617 // try to find a toplevel element
618 while (!url_feof(pb)) {
619 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
620 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
621 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
622 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS)
624 matroska->current_id = id;
627 id = (id << 8) | avio_r8(pb);
635 * Return: Whether we reached the end of a level in the hierarchy or not.
637 static int ebml_level_end(MatroskaDemuxContext *matroska)
639 AVIOContext *pb = matroska->ctx->pb;
640 int64_t pos = avio_tell(pb);
642 if (matroska->num_levels > 0) {
643 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
644 if (pos - level->start >= level->length || matroska->current_id) {
645 matroska->num_levels--;
653 * Read: an "EBML number", which is defined as a variable-length
654 * array of bytes. The first byte indicates the length by giving a
655 * number of 0-bits followed by a one. The position of the first
656 * "one" bit inside the first byte indicates the length of this
658 * Returns: number of bytes read, < 0 on error
660 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
661 int max_size, uint64_t *number)
666 /* The first byte tells us the length in bytes - avio_r8() can normally
667 * return 0, but since that's not a valid first ebmlID byte, we can
668 * use it safely here to catch EOS. */
669 if (!(total = avio_r8(pb))) {
670 /* we might encounter EOS here */
672 int64_t pos = avio_tell(pb);
673 av_log(matroska->ctx, AV_LOG_ERROR,
674 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
676 return pb->error ? pb->error : AVERROR(EIO);
681 /* get the length of the EBML number */
682 read = 8 - ff_log2_tab[total];
683 if (read > max_size) {
684 int64_t pos = avio_tell(pb) - 1;
685 av_log(matroska->ctx, AV_LOG_ERROR,
686 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
687 (uint8_t) total, pos, pos);
688 return AVERROR_INVALIDDATA;
691 /* read out length */
692 total ^= 1 << ff_log2_tab[total];
694 total = (total << 8) | avio_r8(pb);
702 * Read a EBML length value.
703 * This needs special handling for the "unknown length" case which has multiple
706 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
709 int res = ebml_read_num(matroska, pb, 8, number);
710 if (res > 0 && *number + 1 == 1ULL << (7 * res))
711 *number = 0xffffffffffffffULL;
716 * Read the next element as an unsigned int.
717 * 0 is success, < 0 is failure.
719 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
724 return AVERROR_INVALIDDATA;
726 /* big-endian ordering; build up number */
729 *num = (*num << 8) | avio_r8(pb);
735 * Read the next element as a float.
736 * 0 is success, < 0 is failure.
738 static int ebml_read_float(AVIOContext *pb, int size, double *num)
742 } else if (size == 4) {
743 *num = av_int2float(avio_rb32(pb));
744 } else if (size == 8){
745 *num = av_int2double(avio_rb64(pb));
747 return AVERROR_INVALIDDATA;
753 * Read the next element as an ASCII string.
754 * 0 is success, < 0 is failure.
756 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
760 /* EBML strings are usually not 0-terminated, so we allocate one
761 * byte more, read the string and NULL-terminate it ourselves. */
762 if (!(res = av_malloc(size + 1)))
763 return AVERROR(ENOMEM);
764 if (avio_read(pb, (uint8_t *) res, size) != size) {
776 * Read the next element as binary data.
777 * 0 is success, < 0 is failure.
779 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
781 av_fast_padded_malloc(&bin->data, &bin->size, length);
783 return AVERROR(ENOMEM);
786 bin->pos = avio_tell(pb);
787 if (avio_read(pb, bin->data, length) != length) {
788 av_freep(&bin->data);
797 * Read the next element, but only the header. The contents
798 * are supposed to be sub-elements which can be read separately.
799 * 0 is success, < 0 is failure.
801 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
803 AVIOContext *pb = matroska->ctx->pb;
804 MatroskaLevel *level;
806 if (matroska->num_levels >= EBML_MAX_DEPTH) {
807 av_log(matroska->ctx, AV_LOG_ERROR,
808 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
809 return AVERROR(ENOSYS);
812 level = &matroska->levels[matroska->num_levels++];
813 level->start = avio_tell(pb);
814 level->length = length;
820 * Read signed/unsigned "EBML" numbers.
821 * Return: number of bytes processed, < 0 on error
823 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
824 uint8_t *data, uint32_t size, uint64_t *num)
827 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
828 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
832 * Same as above, but signed.
834 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
835 uint8_t *data, uint32_t size, int64_t *num)
840 /* read as unsigned number first */
841 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
844 /* make signed (weird way) */
845 *num = unum - ((1LL << (7*res - 1)) - 1);
850 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
851 EbmlSyntax *syntax, void *data);
853 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
854 uint32_t id, void *data)
857 for (i=0; syntax[i].id; i++)
858 if (id == syntax[i].id)
860 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
861 matroska->num_levels > 0 &&
862 matroska->levels[matroska->num_levels-1].length == 0xffffffffffffff)
863 return 0; // we reached the end of an unknown size cluster
864 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
865 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
866 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
867 return AVERROR_INVALIDDATA;
869 return ebml_parse_elem(matroska, &syntax[i], data);
872 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
875 if (!matroska->current_id) {
877 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
880 matroska->current_id = id | 1 << 7*res;
882 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
885 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
890 for (i=0; syntax[i].id; i++)
891 switch (syntax[i].type) {
893 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
896 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
900 *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
904 while (!res && !ebml_level_end(matroska))
905 res = ebml_parse(matroska, syntax, data);
910 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
911 EbmlSyntax *syntax, void *data)
913 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
916 // max. 16 MB for strings
917 [EBML_STR] = 0x1000000,
918 [EBML_UTF8] = 0x1000000,
919 // max. 256 MB for binary data
920 [EBML_BIN] = 0x10000000,
921 // no limits for anything else
923 AVIOContext *pb = matroska->ctx->pb;
924 uint32_t id = syntax->id;
929 data = (char *)data + syntax->data_offset;
930 if (syntax->list_elem_size) {
931 EbmlList *list = data;
932 newelem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
934 return AVERROR(ENOMEM);
935 list->elem = newelem;
936 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
937 memset(data, 0, syntax->list_elem_size);
941 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
942 matroska->current_id = 0;
943 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
945 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
946 av_log(matroska->ctx, AV_LOG_ERROR,
947 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
948 length, max_lengths[syntax->type], syntax->type);
949 return AVERROR_INVALIDDATA;
953 switch (syntax->type) {
954 case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
955 case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
957 case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
958 case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
959 case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
961 if (id == MATROSKA_ID_SEGMENT)
962 matroska->segment_start = avio_tell(matroska->ctx->pb);
963 return ebml_parse_nest(matroska, syntax->def.n, data);
964 case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
965 case EBML_STOP: return 1;
967 if(ffio_limit(pb, length) != length)
969 return avio_skip(pb,length)<0 ? AVERROR(EIO) : 0;
971 if (res == AVERROR_INVALIDDATA)
972 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
973 else if (res == AVERROR(EIO))
974 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
978 static void ebml_free(EbmlSyntax *syntax, void *data)
981 for (i=0; syntax[i].id; i++) {
982 void *data_off = (char *)data + syntax[i].data_offset;
983 switch (syntax[i].type) {
985 case EBML_UTF8: av_freep(data_off); break;
986 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
988 if (syntax[i].list_elem_size) {
989 EbmlList *list = data_off;
990 char *ptr = list->elem;
991 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
992 ebml_free(syntax[i].def.n, ptr);
995 ebml_free(syntax[i].def.n, data_off);
1005 static int matroska_probe(AVProbeData *p)
1008 int len_mask = 0x80, size = 1, n = 1, i;
1011 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1014 /* length of header */
1016 while (size <= 8 && !(total & len_mask)) {
1022 total &= (len_mask - 1);
1024 total = (total << 8) | p->buf[4 + n++];
1026 /* Does the probe data contain the whole header? */
1027 if (p->buf_size < 4 + size + total)
1030 /* The header should contain a known document type. For now,
1031 * we don't parse the whole header but simply check for the
1032 * availability of that array of characters inside the header.
1033 * Not fully fool-proof, but good enough. */
1034 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1035 int probelen = strlen(matroska_doctypes[i]);
1036 if (total < probelen)
1038 for (n = 4+size; n <= 4+size+total-probelen; n++)
1039 if (!memcmp(p->buf+n, matroska_doctypes[i], probelen))
1040 return AVPROBE_SCORE_MAX;
1043 // probably valid EBML header but no recognized doctype
1044 return AVPROBE_SCORE_MAX/2;
1047 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1050 MatroskaTrack *tracks = matroska->tracks.elem;
1053 for (i=0; i < matroska->tracks.nb_elem; i++)
1054 if (tracks[i].num == num)
1057 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1061 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
1062 MatroskaTrack *track)
1064 MatroskaTrackEncoding *encodings = track->encodings.elem;
1065 uint8_t* data = *buf;
1066 int isize = *buf_size;
1067 uint8_t* pkt_data = NULL;
1068 uint8_t av_unused *newpktdata;
1069 int pkt_size = isize;
1073 if (pkt_size >= 10000000U)
1074 return AVERROR_INVALIDDATA;
1076 switch (encodings[0].compression.algo) {
1077 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP: {
1078 int header_size = encodings[0].compression.settings.size;
1079 uint8_t *header = encodings[0].compression.settings.data;
1081 if (header_size && !header) {
1082 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1089 pkt_size = isize + header_size;
1090 pkt_data = av_malloc(pkt_size);
1092 return AVERROR(ENOMEM);
1094 memcpy(pkt_data, header, header_size);
1095 memcpy(pkt_data + header_size, data, isize);
1099 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1101 olen = pkt_size *= 3;
1102 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1104 result = AVERROR(ENOMEM);
1107 pkt_data = newpktdata;
1108 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1109 } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
1111 result = AVERROR_INVALIDDATA;
1118 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
1119 z_stream zstream = {0};
1120 if (inflateInit(&zstream) != Z_OK)
1122 zstream.next_in = data;
1123 zstream.avail_in = isize;
1126 newpktdata = av_realloc(pkt_data, pkt_size);
1128 inflateEnd(&zstream);
1131 pkt_data = newpktdata;
1132 zstream.avail_out = pkt_size - zstream.total_out;
1133 zstream.next_out = pkt_data + zstream.total_out;
1135 result = inflate(&zstream, Z_NO_FLUSH);
1137 result = Z_MEM_ERROR;
1138 } while (result==Z_OK && pkt_size<10000000);
1139 pkt_size = zstream.total_out;
1140 inflateEnd(&zstream);
1141 if (result != Z_STREAM_END) {
1142 if (result == Z_MEM_ERROR)
1143 result = AVERROR(ENOMEM);
1145 result = AVERROR_INVALIDDATA;
1152 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
1153 bz_stream bzstream = {0};
1154 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1156 bzstream.next_in = data;
1157 bzstream.avail_in = isize;
1160 newpktdata = av_realloc(pkt_data, pkt_size);
1162 BZ2_bzDecompressEnd(&bzstream);
1165 pkt_data = newpktdata;
1166 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1167 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1169 result = BZ2_bzDecompress(&bzstream);
1171 result = BZ_MEM_ERROR;
1172 } while (result==BZ_OK && pkt_size<10000000);
1173 pkt_size = bzstream.total_out_lo32;
1174 BZ2_bzDecompressEnd(&bzstream);
1175 if (result != BZ_STREAM_END) {
1176 if (result == BZ_MEM_ERROR)
1177 result = AVERROR(ENOMEM);
1179 result = AVERROR_INVALIDDATA;
1186 return AVERROR_INVALIDDATA;
1190 *buf_size = pkt_size;
1197 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
1198 AVPacket *pkt, uint64_t display_duration)
1200 char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
1201 for (; *ptr!=',' && ptr<end-1; ptr++);
1205 for (; *ptr!=',' && ptr<end-1; ptr++);
1207 int64_t end_pts = pkt->pts + display_duration;
1208 int sc = matroska->time_scale * pkt->pts / 10000000;
1209 int ec = matroska->time_scale * end_pts / 10000000;
1210 int sh, sm, ss, eh, em, es, len;
1211 sh = sc/360000; sc -= 360000*sh;
1212 sm = sc/ 6000; sc -= 6000*sm;
1213 ss = sc/ 100; sc -= 100*ss;
1214 eh = ec/360000; ec -= 360000*eh;
1215 em = ec/ 6000; ec -= 6000*em;
1216 es = ec/ 100; ec -= 100*es;
1218 len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
1219 if (!(line = av_malloc(len)))
1221 snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
1222 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
1225 pkt->size = strlen(line);
1229 static int matroska_merge_packets(AVPacket *out, AVPacket *in)
1231 int ret = av_grow_packet(out, in->size);
1235 memcpy(out->data + out->size - in->size, in->data, in->size);
1242 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1243 AVDictionary **metadata, char *prefix)
1245 MatroskaTag *tags = list->elem;
1249 for (i=0; i < list->nb_elem; i++) {
1250 const char *lang= (tags[i].lang && strcmp(tags[i].lang, "und")) ? tags[i].lang : NULL;
1252 if (!tags[i].name) {
1253 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1256 if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1257 else av_strlcpy(key, tags[i].name, sizeof(key));
1258 if (tags[i].def || !lang) {
1259 av_dict_set(metadata, key, tags[i].string, 0);
1260 if (tags[i].sub.nb_elem)
1261 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1264 av_strlcat(key, "-", sizeof(key));
1265 av_strlcat(key, lang, sizeof(key));
1266 av_dict_set(metadata, key, tags[i].string, 0);
1267 if (tags[i].sub.nb_elem)
1268 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1271 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1274 static void matroska_convert_tags(AVFormatContext *s)
1276 MatroskaDemuxContext *matroska = s->priv_data;
1277 MatroskaTags *tags = matroska->tags.elem;
1280 for (i=0; i < matroska->tags.nb_elem; i++) {
1281 if (tags[i].target.attachuid) {
1282 MatroskaAttachement *attachment = matroska->attachments.elem;
1283 for (j=0; j<matroska->attachments.nb_elem; j++)
1284 if (attachment[j].uid == tags[i].target.attachuid
1285 && attachment[j].stream)
1286 matroska_convert_tag(s, &tags[i].tag,
1287 &attachment[j].stream->metadata, NULL);
1288 } else if (tags[i].target.chapteruid) {
1289 MatroskaChapter *chapter = matroska->chapters.elem;
1290 for (j=0; j<matroska->chapters.nb_elem; j++)
1291 if (chapter[j].uid == tags[i].target.chapteruid
1292 && chapter[j].chapter)
1293 matroska_convert_tag(s, &tags[i].tag,
1294 &chapter[j].chapter->metadata, NULL);
1295 } else if (tags[i].target.trackuid) {
1296 MatroskaTrack *track = matroska->tracks.elem;
1297 for (j=0; j<matroska->tracks.nb_elem; j++)
1298 if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1299 matroska_convert_tag(s, &tags[i].tag,
1300 &track[j].stream->metadata, NULL);
1302 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1303 tags[i].target.type);
1308 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska, int idx)
1310 EbmlList *seekhead_list = &matroska->seekhead;
1311 MatroskaSeekhead *seekhead = seekhead_list->elem;
1312 uint32_t level_up = matroska->level_up;
1313 int64_t before_pos = avio_tell(matroska->ctx->pb);
1314 uint32_t saved_id = matroska->current_id;
1315 MatroskaLevel level;
1319 if (idx >= seekhead_list->nb_elem
1320 || seekhead[idx].id == MATROSKA_ID_SEEKHEAD
1321 || seekhead[idx].id == MATROSKA_ID_CLUSTER)
1325 offset = seekhead[idx].pos + matroska->segment_start;
1326 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1327 /* We don't want to lose our seekhead level, so we add
1328 * a dummy. This is a crude hack. */
1329 if (matroska->num_levels == EBML_MAX_DEPTH) {
1330 av_log(matroska->ctx, AV_LOG_INFO,
1331 "Max EBML element depth (%d) reached, "
1332 "cannot parse further.\n", EBML_MAX_DEPTH);
1333 ret = AVERROR_INVALIDDATA;
1336 level.length = (uint64_t)-1;
1337 matroska->levels[matroska->num_levels] = level;
1338 matroska->num_levels++;
1339 matroska->current_id = 0;
1341 ret = ebml_parse(matroska, matroska_segment, matroska);
1343 /* remove dummy level */
1344 while (matroska->num_levels) {
1345 uint64_t length = matroska->levels[--matroska->num_levels].length;
1346 if (length == (uint64_t)-1)
1352 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1353 matroska->level_up = level_up;
1354 matroska->current_id = saved_id;
1359 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1361 EbmlList *seekhead_list = &matroska->seekhead;
1362 int64_t before_pos = avio_tell(matroska->ctx->pb);
1365 // we should not do any seeking in the streaming case
1366 if (!matroska->ctx->pb->seekable ||
1367 (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1370 for (i = 0; i < seekhead_list->nb_elem; i++) {
1371 MatroskaSeekhead *seekhead = seekhead_list->elem;
1372 if (seekhead[i].pos <= before_pos)
1375 // defer cues parsing until we actually need cue data.
1376 if (seekhead[i].id == MATROSKA_ID_CUES) {
1377 matroska->cues_parsing_deferred = 1;
1381 if (matroska_parse_seekhead_entry(matroska, i) < 0) {
1382 // mark index as broken
1383 matroska->cues_parsing_deferred = -1;
1389 static void matroska_add_index_entries(MatroskaDemuxContext *matroska) {
1390 EbmlList *index_list;
1391 MatroskaIndex *index;
1392 int index_scale = 1;
1395 index_list = &matroska->index;
1396 index = index_list->elem;
1397 if (index_list->nb_elem
1398 && index[0].time > 1E14/matroska->time_scale) {
1399 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1400 index_scale = matroska->time_scale;
1402 for (i = 0; i < index_list->nb_elem; i++) {
1403 EbmlList *pos_list = &index[i].pos;
1404 MatroskaIndexPos *pos = pos_list->elem;
1405 for (j = 0; j < pos_list->nb_elem; j++) {
1406 MatroskaTrack *track = matroska_find_track_by_num(matroska, pos[j].track);
1407 if (track && track->stream)
1408 av_add_index_entry(track->stream,
1409 pos[j].pos + matroska->segment_start,
1410 index[i].time/index_scale, 0, 0,
1416 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1417 EbmlList *seekhead_list = &matroska->seekhead;
1418 MatroskaSeekhead *seekhead = seekhead_list->elem;
1421 for (i = 0; i < seekhead_list->nb_elem; i++)
1422 if (seekhead[i].id == MATROSKA_ID_CUES)
1424 assert(i <= seekhead_list->nb_elem);
1426 if (matroska_parse_seekhead_entry(matroska, i) < 0)
1427 matroska->cues_parsing_deferred = -1;
1428 matroska_add_index_entries(matroska);
1431 static int matroska_aac_profile(char *codec_id)
1433 static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1436 for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1437 if (strstr(codec_id, aac_profiles[profile]))
1442 static int matroska_aac_sri(int samplerate)
1446 for (sri=0; sri<FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1447 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1452 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1455 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1456 time_t creation_time = date_utc / 1000000000 + 978307200;
1457 struct tm *ptm = gmtime(&creation_time);
1459 strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm);
1460 av_dict_set(metadata, "creation_time", buffer, 0);
1463 static int matroska_read_header(AVFormatContext *s)
1465 MatroskaDemuxContext *matroska = s->priv_data;
1466 EbmlList *attachements_list = &matroska->attachments;
1467 MatroskaAttachement *attachements;
1468 EbmlList *chapters_list = &matroska->chapters;
1469 MatroskaChapter *chapters;
1470 MatroskaTrack *tracks;
1471 uint64_t max_start = 0;
1479 /* First read the EBML header. */
1480 if (ebml_parse(matroska, ebml_syntax, &ebml)
1481 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1482 || ebml.id_length > sizeof(uint32_t) || ebml.doctype_version > 3 || !ebml.doctype) {
1483 av_log(matroska->ctx, AV_LOG_ERROR,
1484 "EBML header using unsupported features\n"
1485 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1486 ebml.version, ebml.doctype, ebml.doctype_version);
1487 ebml_free(ebml_syntax, &ebml);
1488 return AVERROR_PATCHWELCOME;
1489 } else if (ebml.doctype_version == 3) {
1490 av_log(matroska->ctx, AV_LOG_WARNING,
1491 "EBML header using unsupported features\n"
1492 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1493 ebml.version, ebml.doctype, ebml.doctype_version);
1495 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
1496 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
1498 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
1499 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
1500 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
1501 ebml_free(ebml_syntax, &ebml);
1502 return AVERROR_INVALIDDATA;
1505 ebml_free(ebml_syntax, &ebml);
1507 /* The next thing is a segment. */
1508 pos = avio_tell(matroska->ctx->pb);
1509 res = ebml_parse(matroska, matroska_segments, matroska);
1510 // try resyncing until we find a EBML_STOP type element.
1512 res = matroska_resync(matroska, pos);
1515 pos = avio_tell(matroska->ctx->pb);
1516 res = ebml_parse(matroska, matroska_segment, matroska);
1518 matroska_execute_seekhead(matroska);
1520 if (!matroska->time_scale)
1521 matroska->time_scale = 1000000;
1522 if (matroska->duration)
1523 matroska->ctx->duration = matroska->duration * matroska->time_scale
1524 * 1000 / AV_TIME_BASE;
1525 av_dict_set(&s->metadata, "title", matroska->title, 0);
1527 if (matroska->date_utc.size == 8)
1528 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
1530 tracks = matroska->tracks.elem;
1531 for (i=0; i < matroska->tracks.nb_elem; i++) {
1532 MatroskaTrack *track = &tracks[i];
1533 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1534 EbmlList *encodings_list = &track->encodings;
1535 MatroskaTrackEncoding *encodings = encodings_list->elem;
1536 uint8_t *extradata = NULL;
1537 int extradata_size = 0;
1538 int extradata_offset = 0;
1539 uint32_t fourcc = 0;
1542 /* Apply some sanity checks. */
1543 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1544 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1545 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1546 av_log(matroska->ctx, AV_LOG_INFO,
1547 "Unknown or unsupported track type %"PRIu64"\n",
1551 if (track->codec_id == NULL)
1554 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1555 if (!track->default_duration && track->video.frame_rate > 0)
1556 track->default_duration = 1000000000/track->video.frame_rate;
1557 if (!track->video.display_width)
1558 track->video.display_width = track->video.pixel_width;
1559 if (!track->video.display_height)
1560 track->video.display_height = track->video.pixel_height;
1561 if (track->video.color_space.size == 4)
1562 fourcc = AV_RL32(track->video.color_space.data);
1563 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1564 if (!track->audio.out_samplerate)
1565 track->audio.out_samplerate = track->audio.samplerate;
1567 if (encodings_list->nb_elem > 1) {
1568 av_log(matroska->ctx, AV_LOG_ERROR,
1569 "Multiple combined encodings not supported");
1570 } else if (encodings_list->nb_elem == 1) {
1571 if (encodings[0].type ||
1574 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1577 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1580 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
1582 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP)) {
1583 encodings[0].scope = 0;
1584 av_log(matroska->ctx, AV_LOG_ERROR,
1585 "Unsupported encoding type");
1586 } else if (track->codec_priv.size && encodings[0].scope&2) {
1587 uint8_t *codec_priv = track->codec_priv.data;
1588 int ret = matroska_decode_buffer(&track->codec_priv.data,
1589 &track->codec_priv.size,
1592 track->codec_priv.data = NULL;
1593 track->codec_priv.size = 0;
1594 av_log(matroska->ctx, AV_LOG_ERROR,
1595 "Failed to decode codec private data\n");
1598 if (codec_priv != track->codec_priv.data)
1599 av_free(codec_priv);
1603 for(j=0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++){
1604 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1605 strlen(ff_mkv_codec_tags[j].str))){
1606 codec_id= ff_mkv_codec_tags[j].id;
1611 st = track->stream = avformat_new_stream(s, NULL);
1613 return AVERROR(ENOMEM);
1615 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1616 && track->codec_priv.size >= 40
1617 && track->codec_priv.data != NULL) {
1618 track->ms_compat = 1;
1619 fourcc = AV_RL32(track->codec_priv.data + 16);
1620 codec_id = ff_codec_get_id(ff_codec_bmp_tags, fourcc);
1621 extradata_offset = 40;
1622 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1623 && track->codec_priv.size >= 14
1624 && track->codec_priv.data != NULL) {
1626 ffio_init_context(&b, track->codec_priv.data, track->codec_priv.size,
1627 0, NULL, NULL, NULL, NULL);
1628 ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1631 codec_id = st->codec->codec_id;
1632 extradata_offset = FFMIN(track->codec_priv.size, 18);
1633 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1634 && (track->codec_priv.size >= 86)
1635 && (track->codec_priv.data != NULL)) {
1636 fourcc = AV_RL32(track->codec_priv.data);
1637 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1638 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX-12) {
1639 /* Only ALAC's magic cookie is stored in Matroska's track headers.
1640 Create the "atom size", "tag", and "tag version" fields the
1641 decoder expects manually. */
1642 extradata_size = 12 + track->codec_priv.size;
1643 extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1644 if (extradata == NULL)
1645 return AVERROR(ENOMEM);
1646 AV_WB32(extradata, extradata_size);
1647 memcpy(&extradata[4], "alac", 4);
1648 AV_WB32(&extradata[8], 0);
1649 memcpy(&extradata[12], track->codec_priv.data, track->codec_priv.size);
1650 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1651 switch (track->audio.bitdepth) {
1652 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1653 case 24: codec_id = AV_CODEC_ID_PCM_S24BE; break;
1654 case 32: codec_id = AV_CODEC_ID_PCM_S32BE; break;
1656 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1657 switch (track->audio.bitdepth) {
1658 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1659 case 24: codec_id = AV_CODEC_ID_PCM_S24LE; break;
1660 case 32: codec_id = AV_CODEC_ID_PCM_S32LE; break;
1662 } else if (codec_id==AV_CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1663 codec_id = AV_CODEC_ID_PCM_F64LE;
1664 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1665 int profile = matroska_aac_profile(track->codec_id);
1666 int sri = matroska_aac_sri(track->audio.samplerate);
1667 extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1668 if (extradata == NULL)
1669 return AVERROR(ENOMEM);
1670 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1671 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1672 if (strstr(track->codec_id, "SBR")) {
1673 sri = matroska_aac_sri(track->audio.out_samplerate);
1674 extradata[2] = 0x56;
1675 extradata[3] = 0xE5;
1676 extradata[4] = 0x80 | (sri<<3);
1680 } else if (codec_id == AV_CODEC_ID_TTA) {
1681 extradata_size = 30;
1682 extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1683 if (extradata == NULL)
1684 return AVERROR(ENOMEM);
1685 ffio_init_context(&b, extradata, extradata_size, 1,
1686 NULL, NULL, NULL, NULL);
1687 avio_write(&b, "TTA1", 4);
1689 avio_wl16(&b, track->audio.channels);
1690 avio_wl16(&b, track->audio.bitdepth);
1691 avio_wl32(&b, track->audio.out_samplerate);
1692 avio_wl32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1693 } else if (codec_id == AV_CODEC_ID_RV10 || codec_id == AV_CODEC_ID_RV20 ||
1694 codec_id == AV_CODEC_ID_RV30 || codec_id == AV_CODEC_ID_RV40) {
1695 extradata_offset = 26;
1696 } else if (codec_id == AV_CODEC_ID_RA_144) {
1697 track->audio.out_samplerate = 8000;
1698 track->audio.channels = 1;
1699 } else if ((codec_id == AV_CODEC_ID_RA_288 || codec_id == AV_CODEC_ID_COOK ||
1700 codec_id == AV_CODEC_ID_ATRAC3 || codec_id == AV_CODEC_ID_SIPR)
1701 && track->codec_priv.data) {
1704 ffio_init_context(&b, track->codec_priv.data,track->codec_priv.size,
1705 0, NULL, NULL, NULL, NULL);
1707 flavor = avio_rb16(&b);
1708 track->audio.coded_framesize = avio_rb32(&b);
1710 track->audio.sub_packet_h = avio_rb16(&b);
1711 track->audio.frame_size = avio_rb16(&b);
1712 track->audio.sub_packet_size = avio_rb16(&b);
1713 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1714 if (codec_id == AV_CODEC_ID_RA_288) {
1715 st->codec->block_align = track->audio.coded_framesize;
1716 track->codec_priv.size = 0;
1718 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1719 const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1720 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1721 st->codec->bit_rate = sipr_bit_rate[flavor];
1723 st->codec->block_align = track->audio.sub_packet_size;
1724 extradata_offset = 78;
1727 track->codec_priv.size -= extradata_offset;
1729 if (codec_id == AV_CODEC_ID_NONE)
1730 av_log(matroska->ctx, AV_LOG_INFO,
1731 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1733 if (track->time_scale < 0.01)
1734 track->time_scale = 1.0;
1735 avpriv_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1737 st->codec->codec_id = codec_id;
1739 if (strcmp(track->language, "und"))
1740 av_dict_set(&st->metadata, "language", track->language, 0);
1741 av_dict_set(&st->metadata, "title", track->name, 0);
1743 if (track->flag_default)
1744 st->disposition |= AV_DISPOSITION_DEFAULT;
1745 if (track->flag_forced)
1746 st->disposition |= AV_DISPOSITION_FORCED;
1748 if (!st->codec->extradata) {
1750 st->codec->extradata = extradata;
1751 st->codec->extradata_size = extradata_size;
1752 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1753 st->codec->extradata = av_mallocz(track->codec_priv.size +
1754 FF_INPUT_BUFFER_PADDING_SIZE);
1755 if(st->codec->extradata == NULL)
1756 return AVERROR(ENOMEM);
1757 st->codec->extradata_size = track->codec_priv.size;
1758 memcpy(st->codec->extradata,
1759 track->codec_priv.data + extradata_offset,
1760 track->codec_priv.size);
1764 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1765 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
1767 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1768 st->codec->codec_tag = fourcc;
1769 st->codec->width = track->video.pixel_width;
1770 st->codec->height = track->video.pixel_height;
1771 av_reduce(&st->sample_aspect_ratio.num,
1772 &st->sample_aspect_ratio.den,
1773 st->codec->height * track->video.display_width,
1774 st->codec-> width * track->video.display_height,
1776 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1777 if (track->default_duration) {
1778 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1779 1000000000, track->default_duration, 30000);
1780 #if FF_API_R_FRAME_RATE
1781 st->r_frame_rate = st->avg_frame_rate;
1785 /* export stereo mode flag as metadata tag */
1786 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREO_MODE_COUNT)
1787 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
1789 /* if we have virtual track, mark the real tracks */
1790 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
1792 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
1794 snprintf(buf, sizeof(buf), "%s_%d",
1795 ff_matroska_video_stereo_plane[planes[j].type], i);
1796 for (k=0; k < matroska->tracks.nb_elem; k++)
1797 if (planes[j].uid == tracks[k].uid) {
1798 av_dict_set(&s->streams[k]->metadata,
1799 "stereo_mode", buf, 0);
1803 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1804 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1805 st->codec->sample_rate = track->audio.out_samplerate;
1806 st->codec->channels = track->audio.channels;
1807 st->codec->bits_per_coded_sample = track->audio.bitdepth;
1808 if (st->codec->codec_id != AV_CODEC_ID_AAC)
1809 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1810 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1811 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1812 if (st->codec->codec_id == AV_CODEC_ID_SSA)
1813 matroska->contains_ssa = 1;
1817 attachements = attachements_list->elem;
1818 for (j=0; j<attachements_list->nb_elem; j++) {
1819 if (!(attachements[j].filename && attachements[j].mime &&
1820 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1821 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1823 AVStream *st = avformat_new_stream(s, NULL);
1826 av_dict_set(&st->metadata, "filename",attachements[j].filename, 0);
1827 av_dict_set(&st->metadata, "mimetype", attachements[j].mime, 0);
1828 st->codec->codec_id = AV_CODEC_ID_NONE;
1829 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
1830 st->codec->extradata = av_malloc(attachements[j].bin.size + FF_INPUT_BUFFER_PADDING_SIZE);
1831 if(st->codec->extradata == NULL)
1833 st->codec->extradata_size = attachements[j].bin.size;
1834 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1836 for (i=0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
1837 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1838 strlen(ff_mkv_mime_tags[i].str))) {
1839 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1843 attachements[j].stream = st;
1847 chapters = chapters_list->elem;
1848 for (i=0; i<chapters_list->nb_elem; i++)
1849 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1850 && (max_start==0 || chapters[i].start > max_start)) {
1851 chapters[i].chapter =
1852 avpriv_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1853 chapters[i].start, chapters[i].end,
1855 av_dict_set(&chapters[i].chapter->metadata,
1856 "title", chapters[i].title, 0);
1857 max_start = chapters[i].start;
1860 matroska_add_index_entries(matroska);
1862 matroska_convert_tags(s);
1868 * Put one packet in an application-supplied AVPacket struct.
1869 * Returns 0 on success or -1 on failure.
1871 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1874 if (matroska->num_packets > 0) {
1875 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1876 av_free(matroska->packets[0]);
1877 if (matroska->num_packets > 1) {
1879 memmove(&matroska->packets[0], &matroska->packets[1],
1880 (matroska->num_packets - 1) * sizeof(AVPacket *));
1881 newpackets = av_realloc(matroska->packets,
1882 (matroska->num_packets - 1) * sizeof(AVPacket *));
1884 matroska->packets = newpackets;
1886 av_freep(&matroska->packets);
1887 matroska->prev_pkt = NULL;
1889 matroska->num_packets--;
1897 * Free all packets in our internal queue.
1899 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1901 matroska->prev_pkt = NULL;
1902 if (matroska->packets) {
1904 for (n = 0; n < matroska->num_packets; n++) {
1905 av_free_packet(matroska->packets[n]);
1906 av_free(matroska->packets[n]);
1908 av_freep(&matroska->packets);
1909 matroska->num_packets = 0;
1913 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
1915 uint32_t **lace_buf, int *laces)
1918 uint8_t *data = *buf;
1919 uint32_t *lace_size;
1923 *lace_buf = av_mallocz(sizeof(int));
1925 return AVERROR(ENOMEM);
1927 *lace_buf[0] = size;
1931 av_assert0(size > 0);
1935 lace_size = av_mallocz(*laces * sizeof(int));
1937 return AVERROR(ENOMEM);
1940 case 0x1: /* Xiph lacing */ {
1943 for (n = 0; res == 0 && n < *laces - 1; n++) {
1950 lace_size[n] += temp;
1956 total += lace_size[n];
1958 if (size <= total) {
1959 res = AVERROR_INVALIDDATA;
1963 lace_size[n] = size - total;
1967 case 0x2: /* fixed-size lacing */
1968 if (size % (*laces)) {
1969 res = AVERROR_INVALIDDATA;
1972 for (n = 0; n < *laces; n++)
1973 lace_size[n] = size / *laces;
1976 case 0x3: /* EBML lacing */ {
1979 n = matroska_ebmlnum_uint(matroska, data, size, &num);
1981 av_log(matroska->ctx, AV_LOG_INFO,
1982 "EBML block data error\n");
1988 total = lace_size[0] = num;
1989 for (n = 1; res == 0 && n < *laces - 1; n++) {
1992 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1994 av_log(matroska->ctx, AV_LOG_INFO,
1995 "EBML block data error\n");
2001 lace_size[n] = lace_size[n - 1] + snum;
2002 total += lace_size[n];
2004 if (size <= total) {
2005 res = AVERROR_INVALIDDATA;
2008 lace_size[*laces - 1] = size - total;
2014 *lace_buf = lace_size;
2019 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2020 MatroskaTrack *track,
2022 uint8_t *data, int size,
2026 int a = st->codec->block_align;
2027 int sps = track->audio.sub_packet_size;
2028 int cfs = track->audio.coded_framesize;
2029 int h = track->audio.sub_packet_h;
2030 int y = track->audio.sub_packet_cnt;
2031 int w = track->audio.frame_size;
2034 if (!track->audio.pkt_cnt) {
2035 if (track->audio.sub_packet_cnt == 0)
2036 track->audio.buf_timecode = timecode;
2037 if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2038 if (size < cfs * h / 2) {
2039 av_log(matroska->ctx, AV_LOG_ERROR,
2040 "Corrupt int4 RM-style audio packet size\n");
2041 return AVERROR_INVALIDDATA;
2043 for (x=0; x<h/2; x++)
2044 memcpy(track->audio.buf+x*2*w+y*cfs,
2046 } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2048 av_log(matroska->ctx, AV_LOG_ERROR,
2049 "Corrupt sipr RM-style audio packet size\n");
2050 return AVERROR_INVALIDDATA;
2052 memcpy(track->audio.buf + y*w, data, w);
2054 if (size < sps * w / sps || h<=0) {
2055 av_log(matroska->ctx, AV_LOG_ERROR,
2056 "Corrupt generic RM-style audio packet size\n");
2057 return AVERROR_INVALIDDATA;
2059 for (x=0; x<w/sps; x++)
2060 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
2063 if (++track->audio.sub_packet_cnt >= h) {
2064 if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2065 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2066 track->audio.sub_packet_cnt = 0;
2067 track->audio.pkt_cnt = h*w / a;
2071 while (track->audio.pkt_cnt) {
2072 AVPacket *pkt = NULL;
2073 if (!(pkt = av_mallocz(sizeof(AVPacket))) || av_new_packet(pkt, a) < 0){
2075 return AVERROR(ENOMEM);
2077 memcpy(pkt->data, track->audio.buf
2078 + a * (h*w / a - track->audio.pkt_cnt--), a);
2079 pkt->pts = track->audio.buf_timecode;
2080 track->audio.buf_timecode = AV_NOPTS_VALUE;
2082 pkt->stream_index = st->index;
2083 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2088 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2089 MatroskaTrack *track,
2091 uint8_t *data, int pkt_size,
2092 uint64_t timecode, uint64_t lace_duration,
2093 int64_t pos, int is_keyframe,
2094 uint8_t *additional, uint64_t additional_id, int additional_size)
2096 MatroskaTrackEncoding *encodings = track->encodings.elem;
2097 uint8_t *pkt_data = data;
2098 int offset = 0, res;
2101 if (encodings && encodings->scope & 1) {
2102 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2107 if (st->codec->codec_id == AV_CODEC_ID_PRORES)
2110 pkt = av_mallocz(sizeof(AVPacket));
2111 /* XXX: prevent data copy... */
2112 if (av_new_packet(pkt, pkt_size + offset) < 0) {
2114 return AVERROR(ENOMEM);
2117 if (st->codec->codec_id == AV_CODEC_ID_PRORES) {
2118 uint8_t *buf = pkt->data;
2119 bytestream_put_be32(&buf, pkt_size);
2120 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2123 memcpy(pkt->data + offset, pkt_data, pkt_size);
2125 if (pkt_data != data)
2128 pkt->flags = is_keyframe;
2129 pkt->stream_index = st->index;
2131 if (additional_size > 0) {
2132 uint8_t *side_data = av_packet_new_side_data(pkt,
2133 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
2134 additional_size + sizeof(additional_id));
2135 uint8_t additional_id_buf[8];
2136 if(side_data == NULL) {
2137 return AVERROR(ENOMEM);
2139 AV_WB64(additional_id_buf, additional_id);
2140 memcpy(side_data, additional_id_buf, 8);
2141 memcpy(side_data + 8, additional, additional_size);
2144 if (track->ms_compat)
2145 pkt->dts = timecode;
2147 pkt->pts = timecode;
2149 if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2151 * For backward compatibility.
2152 * Historically, we have put subtitle duration
2153 * in convergence_duration, on the off chance
2154 * that the time_scale is less than 1us, which
2155 * could result in a 32bit overflow on the
2156 * normal duration field.
2158 pkt->convergence_duration = lace_duration;
2161 if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE ||
2162 lace_duration <= INT_MAX) {
2164 * For non subtitle tracks, just store the duration
2167 * If it's a subtitle track and duration value does
2168 * not overflow a uint32, then also store it normally.
2170 pkt->duration = lace_duration;
2173 if (st->codec->codec_id == AV_CODEC_ID_SSA)
2174 matroska_fix_ass_packet(matroska, pkt, lace_duration);
2176 if (matroska->prev_pkt &&
2177 timecode != AV_NOPTS_VALUE &&
2178 matroska->prev_pkt->pts == timecode &&
2179 matroska->prev_pkt->stream_index == st->index &&
2180 st->codec->codec_id == AV_CODEC_ID_SSA)
2181 matroska_merge_packets(matroska->prev_pkt, pkt);
2183 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2184 matroska->prev_pkt = pkt;
2190 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
2191 int size, int64_t pos, uint64_t cluster_time,
2192 uint64_t block_duration, int is_keyframe,
2193 uint8_t *additional, uint64_t additional_id, int additional_size,
2194 int64_t cluster_pos)
2196 uint64_t timecode = AV_NOPTS_VALUE;
2197 MatroskaTrack *track;
2201 uint32_t *lace_size = NULL;
2202 int n, flags, laces = 0;
2205 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2206 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2212 track = matroska_find_track_by_num(matroska, num);
2213 if (!track || !track->stream) {
2214 av_log(matroska->ctx, AV_LOG_INFO,
2215 "Invalid stream %"PRIu64" or size %u\n", num, size);
2216 return AVERROR_INVALIDDATA;
2217 } else if (size <= 3)
2220 if (st->discard >= AVDISCARD_ALL)
2222 av_assert1(block_duration != AV_NOPTS_VALUE);
2224 block_time = AV_RB16(data);
2228 if (is_keyframe == -1)
2229 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2231 if (cluster_time != (uint64_t)-1
2232 && (block_time >= 0 || cluster_time >= -block_time)) {
2233 timecode = cluster_time + block_time;
2234 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
2235 && timecode < track->end_timecode)
2236 is_keyframe = 0; /* overlapping subtitles are not key frame */
2238 av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
2241 if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2242 if (timecode < matroska->skip_to_timecode)
2244 if (!st->skip_to_keyframe) {
2245 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2246 matroska->skip_to_keyframe = 0;
2249 matroska->skip_to_keyframe = 0;
2252 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
2253 &lace_size, &laces);
2258 if (!block_duration)
2259 block_duration = track->default_duration * laces / matroska->time_scale;
2261 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
2262 track->end_timecode =
2263 FFMAX(track->end_timecode, timecode + block_duration);
2265 for (n = 0; n < laces; n++) {
2266 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
2268 if (lace_size[n] > size) {
2269 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2273 if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2274 st->codec->codec_id == AV_CODEC_ID_COOK ||
2275 st->codec->codec_id == AV_CODEC_ID_SIPR ||
2276 st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2277 st->codec->block_align && track->audio.sub_packet_size) {
2279 res = matroska_parse_rm_audio(matroska, track, st, data, size,
2285 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2286 timecode, lace_duration,
2287 pos, !n? is_keyframe : 0,
2288 additional, additional_id, additional_size);
2293 if (timecode != AV_NOPTS_VALUE)
2294 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2295 data += lace_size[n];
2296 size -= lace_size[n];
2304 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
2306 EbmlList *blocks_list;
2307 MatroskaBlock *blocks;
2309 res = ebml_parse(matroska,
2310 matroska_cluster_incremental_parsing,
2311 &matroska->current_cluster);
2314 if (matroska->current_cluster_pos)
2315 ebml_level_end(matroska);
2316 ebml_free(matroska_cluster, &matroska->current_cluster);
2317 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2318 matroska->current_cluster_num_blocks = 0;
2319 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2320 matroska->prev_pkt = NULL;
2321 /* sizeof the ID which was already read */
2322 if (matroska->current_id)
2323 matroska->current_cluster_pos -= 4;
2324 res = ebml_parse(matroska,
2325 matroska_clusters_incremental,
2326 &matroska->current_cluster);
2327 /* Try parsing the block again. */
2329 res = ebml_parse(matroska,
2330 matroska_cluster_incremental_parsing,
2331 &matroska->current_cluster);
2335 matroska->current_cluster_num_blocks <
2336 matroska->current_cluster.blocks.nb_elem) {
2337 blocks_list = &matroska->current_cluster.blocks;
2338 blocks = blocks_list->elem;
2340 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2341 i = blocks_list->nb_elem - 1;
2342 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2343 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2344 uint8_t* additional = blocks[i].additional.size > 0 ?
2345 blocks[i].additional.data : NULL;
2346 if (!blocks[i].non_simple)
2347 blocks[i].duration = 0;
2348 res = matroska_parse_block(matroska,
2349 blocks[i].bin.data, blocks[i].bin.size,
2351 matroska->current_cluster.timecode,
2352 blocks[i].duration, is_keyframe,
2353 additional, blocks[i].additional_id,
2354 blocks[i].additional.size,
2355 matroska->current_cluster_pos);
2359 if (res < 0) matroska->done = 1;
2363 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
2365 MatroskaCluster cluster = { 0 };
2366 EbmlList *blocks_list;
2367 MatroskaBlock *blocks;
2370 if (!matroska->contains_ssa)
2371 return matroska_parse_cluster_incremental(matroska);
2372 pos = avio_tell(matroska->ctx->pb);
2373 matroska->prev_pkt = NULL;
2374 if (matroska->current_id)
2375 pos -= 4; /* sizeof the ID which was already read */
2376 res = ebml_parse(matroska, matroska_clusters, &cluster);
2377 blocks_list = &cluster.blocks;
2378 blocks = blocks_list->elem;
2379 for (i=0; i<blocks_list->nb_elem; i++)
2380 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2381 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2382 res=matroska_parse_block(matroska,
2383 blocks[i].bin.data, blocks[i].bin.size,
2384 blocks[i].bin.pos, cluster.timecode,
2385 blocks[i].duration, is_keyframe, NULL, 0, 0,
2388 ebml_free(matroska_cluster, &cluster);
2392 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
2394 MatroskaDemuxContext *matroska = s->priv_data;
2396 while (matroska_deliver_packet(matroska, pkt)) {
2397 int64_t pos = avio_tell(matroska->ctx->pb);
2400 if (matroska_parse_cluster(matroska) < 0)
2401 matroska_resync(matroska, pos);
2407 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2408 int64_t timestamp, int flags)
2410 MatroskaDemuxContext *matroska = s->priv_data;
2411 MatroskaTrack *tracks = matroska->tracks.elem;
2412 AVStream *st = s->streams[stream_index];
2413 int i, index, index_sub, index_min;
2415 /* Parse the CUES now since we need the index data to seek. */
2416 if (matroska->cues_parsing_deferred > 0) {
2417 matroska->cues_parsing_deferred = 0;
2418 matroska_parse_cues(matroska);
2421 if (!st->nb_index_entries)
2423 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2425 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2426 avio_seek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
2427 matroska->current_id = 0;
2428 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2429 matroska_clear_queue(matroska);
2430 if (matroska_parse_cluster(matroska) < 0)
2435 matroska_clear_queue(matroska);
2436 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
2440 for (i=0; i < matroska->tracks.nb_elem; i++) {
2441 tracks[i].audio.pkt_cnt = 0;
2442 tracks[i].audio.sub_packet_cnt = 0;
2443 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2444 tracks[i].end_timecode = 0;
2445 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
2446 && tracks[i].stream->discard != AVDISCARD_ALL) {
2447 index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
2449 && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
2450 && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
2451 index_min = index_sub;
2455 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2456 matroska->current_id = 0;
2457 if (flags & AVSEEK_FLAG_ANY) {
2458 st->skip_to_keyframe = 0;
2459 matroska->skip_to_timecode = timestamp;
2461 st->skip_to_keyframe = 1;
2462 matroska->skip_to_timecode = st->index_entries[index].timestamp;
2464 matroska->skip_to_keyframe = 1;
2466 matroska->num_levels = 0;
2467 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2470 // slightly hackish but allows proper fallback to
2471 // the generic seeking code.
2472 matroska_clear_queue(matroska);
2473 matroska->current_id = 0;
2474 st->skip_to_keyframe =
2475 matroska->skip_to_keyframe = 0;
2477 matroska->num_levels = 0;
2481 static int matroska_read_close(AVFormatContext *s)
2483 MatroskaDemuxContext *matroska = s->priv_data;
2484 MatroskaTrack *tracks = matroska->tracks.elem;
2487 matroska_clear_queue(matroska);
2489 for (n=0; n < matroska->tracks.nb_elem; n++)
2490 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2491 av_free(tracks[n].audio.buf);
2492 ebml_free(matroska_cluster, &matroska->current_cluster);
2493 ebml_free(matroska_segment, matroska);
2498 AVInputFormat ff_matroska_demuxer = {
2499 .name = "matroska,webm",
2500 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
2501 .priv_data_size = sizeof(MatroskaDemuxContext),
2502 .read_probe = matroska_probe,
2503 .read_header = matroska_read_header,
2504 .read_packet = matroska_read_packet,
2505 .read_close = matroska_read_close,
2506 .read_seek = matroska_read_seek,