2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The Libav Project
5 * This file is part of Libav.
7 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * Matroska file demuxer
25 * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
26 * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
27 * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28 * @see specs available on the Matroska project page: http://www.matroska.org/
34 #include "avio_internal.h"
35 /* For ff_codec_get_id(). */
40 #include "libavcodec/mpeg4audio.h"
41 #include "libavutil/intfloat.h"
42 #include "libavutil/intreadwrite.h"
43 #include "libavutil/avstring.h"
44 #include "libavutil/lzo.h"
45 #include "libavutil/dict.h"
66 typedef const struct EbmlSyntax {
75 const struct EbmlSyntax *n;
95 uint64_t doctype_version;
101 } MatroskaTrackCompression;
106 MatroskaTrackCompression compression;
107 } MatroskaTrackEncoding;
111 uint64_t display_width;
112 uint64_t display_height;
113 uint64_t pixel_width;
114 uint64_t pixel_height;
116 } MatroskaTrackVideo;
120 double out_samplerate;
124 /* real audio header (extracted from extradata) */
131 uint64_t buf_timecode;
133 } MatroskaTrackAudio;
144 uint64_t default_duration;
145 uint64_t flag_default;
146 uint64_t flag_forced;
147 MatroskaTrackVideo video;
148 MatroskaTrackAudio audio;
152 int64_t end_timecode;
163 } MatroskaAttachement;
201 MatroskaTagTarget target;
221 AVFormatContext *ctx;
225 MatroskaLevel levels[EBML_MAX_DEPTH];
233 EbmlList attachments;
239 /* byte position of the segment inside the stream */
240 int64_t segment_start;
242 /* the packet queue */
249 /* What to skip before effectively reading a packet. */
250 int skip_to_keyframe;
251 uint64_t skip_to_timecode;
253 /* File has a CUES element, but we defer parsing until it is needed. */
254 int cues_parsing_deferred;
256 int current_cluster_num_blocks;
257 int64_t current_cluster_pos;
258 MatroskaCluster current_cluster;
260 /* File has SSA subtitles which prevent incremental cluster parsing. */
262 } MatroskaDemuxContext;
271 static EbmlSyntax ebml_header[] = {
272 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
273 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
274 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
275 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
276 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
277 { EBML_ID_EBMLVERSION, EBML_NONE },
278 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
282 static EbmlSyntax ebml_syntax[] = {
283 { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
287 static EbmlSyntax matroska_info[] = {
288 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
289 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
290 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
291 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
292 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
293 { MATROSKA_ID_DATEUTC, EBML_NONE },
294 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
298 static EbmlSyntax matroska_track_video[] = {
299 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
300 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
301 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
302 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
303 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
304 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
305 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
306 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
307 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
308 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
309 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
310 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
311 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_NONE },
312 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
316 static EbmlSyntax matroska_track_audio[] = {
317 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
318 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
319 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
320 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
324 static EbmlSyntax matroska_track_encoding_compression[] = {
325 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
326 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
330 static EbmlSyntax matroska_track_encoding[] = {
331 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
332 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
333 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
334 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
338 static EbmlSyntax matroska_track_encodings[] = {
339 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
343 static EbmlSyntax matroska_track[] = {
344 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
345 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack,name) },
346 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
347 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
348 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
349 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
350 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
351 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
352 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
353 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
354 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack,flag_forced), {.u=0} },
355 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
356 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
357 { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
358 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
359 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
360 { MATROSKA_ID_CODECNAME, EBML_NONE },
361 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
362 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
363 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
364 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
365 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
366 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
370 static EbmlSyntax matroska_tracks[] = {
371 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
375 static EbmlSyntax matroska_attachment[] = {
376 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
377 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
378 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
379 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
380 { MATROSKA_ID_FILEDESC, EBML_NONE },
384 static EbmlSyntax matroska_attachments[] = {
385 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
389 static EbmlSyntax matroska_chapter_display[] = {
390 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
391 { MATROSKA_ID_CHAPLANG, EBML_NONE },
395 static EbmlSyntax matroska_chapter_entry[] = {
396 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
397 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
398 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
399 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
400 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
401 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
402 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
403 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
407 static EbmlSyntax matroska_chapter[] = {
408 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
409 { MATROSKA_ID_EDITIONUID, EBML_NONE },
410 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
411 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
412 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
416 static EbmlSyntax matroska_chapters[] = {
417 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
421 static EbmlSyntax matroska_index_pos[] = {
422 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
423 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
424 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
428 static EbmlSyntax matroska_index_entry[] = {
429 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
430 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
434 static EbmlSyntax matroska_index[] = {
435 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
439 static EbmlSyntax matroska_simpletag[] = {
440 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
441 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
442 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag,lang), {.s="und"} },
443 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag,def) },
444 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag,def) },
445 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
449 static EbmlSyntax matroska_tagtargets[] = {
450 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget,type) },
451 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
452 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
453 { MATROSKA_ID_TAGTARGETS_CHAPTERUID,EBML_UINT, 0, offsetof(MatroskaTagTarget,chapteruid) },
454 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
458 static EbmlSyntax matroska_tag[] = {
459 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
460 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
464 static EbmlSyntax matroska_tags[] = {
465 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
469 static EbmlSyntax matroska_seekhead_entry[] = {
470 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
471 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
475 static EbmlSyntax matroska_seekhead[] = {
476 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
480 static EbmlSyntax matroska_segment[] = {
481 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
482 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
483 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
484 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
485 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
486 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
487 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
488 { MATROSKA_ID_CLUSTER, EBML_STOP },
492 static EbmlSyntax matroska_segments[] = {
493 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
497 static EbmlSyntax matroska_blockgroup[] = {
498 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
499 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
500 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration), {.u=AV_NOPTS_VALUE} },
501 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
502 { 1, EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
506 static EbmlSyntax matroska_cluster[] = {
507 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
508 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
509 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
510 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
511 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
515 static EbmlSyntax matroska_clusters[] = {
516 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
517 { MATROSKA_ID_INFO, EBML_NONE },
518 { MATROSKA_ID_CUES, EBML_NONE },
519 { MATROSKA_ID_TAGS, EBML_NONE },
520 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
524 static EbmlSyntax matroska_cluster_incremental_parsing[] = {
525 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
526 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
527 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
528 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
529 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
530 { MATROSKA_ID_INFO, EBML_NONE },
531 { MATROSKA_ID_CUES, EBML_NONE },
532 { MATROSKA_ID_TAGS, EBML_NONE },
533 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
534 { MATROSKA_ID_CLUSTER, EBML_STOP },
538 static EbmlSyntax matroska_cluster_incremental[] = {
539 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
540 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
541 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
542 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
543 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
547 static EbmlSyntax matroska_clusters_incremental[] = {
548 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster_incremental} },
549 { MATROSKA_ID_INFO, EBML_NONE },
550 { MATROSKA_ID_CUES, EBML_NONE },
551 { MATROSKA_ID_TAGS, EBML_NONE },
552 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
556 static const char *const matroska_doctypes[] = { "matroska", "webm" };
559 * Return: Whether we reached the end of a level in the hierarchy or not.
561 static int ebml_level_end(MatroskaDemuxContext *matroska)
563 AVIOContext *pb = matroska->ctx->pb;
564 int64_t pos = avio_tell(pb);
566 if (matroska->num_levels > 0) {
567 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
568 if (pos - level->start >= level->length || matroska->current_id) {
569 matroska->num_levels--;
577 * Read: an "EBML number", which is defined as a variable-length
578 * array of bytes. The first byte indicates the length by giving a
579 * number of 0-bits followed by a one. The position of the first
580 * "one" bit inside the first byte indicates the length of this
582 * Returns: number of bytes read, < 0 on error
584 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
585 int max_size, uint64_t *number)
590 /* The first byte tells us the length in bytes - avio_r8() can normally
591 * return 0, but since that's not a valid first ebmlID byte, we can
592 * use it safely here to catch EOS. */
593 if (!(total = avio_r8(pb))) {
594 /* we might encounter EOS here */
595 if (!pb->eof_reached) {
596 int64_t pos = avio_tell(pb);
597 av_log(matroska->ctx, AV_LOG_ERROR,
598 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
600 return pb->error ? pb->error : AVERROR(EIO);
605 /* get the length of the EBML number */
606 read = 8 - ff_log2_tab[total];
607 if (read > max_size) {
608 int64_t pos = avio_tell(pb) - 1;
609 av_log(matroska->ctx, AV_LOG_ERROR,
610 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
611 (uint8_t) total, pos, pos);
612 return AVERROR_INVALIDDATA;
615 /* read out length */
616 total ^= 1 << ff_log2_tab[total];
618 total = (total << 8) | avio_r8(pb);
626 * Read a EBML length value.
627 * This needs special handling for the "unknown length" case which has multiple
630 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
633 int res = ebml_read_num(matroska, pb, 8, number);
634 if (res > 0 && *number + 1 == 1ULL << (7 * res))
635 *number = 0xffffffffffffffULL;
640 * Read the next element as an unsigned int.
641 * 0 is success, < 0 is failure.
643 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
648 return AVERROR_INVALIDDATA;
650 /* big-endian ordering; build up number */
653 *num = (*num << 8) | avio_r8(pb);
659 * Read the next element as a float.
660 * 0 is success, < 0 is failure.
662 static int ebml_read_float(AVIOContext *pb, int size, double *num)
666 } else if (size == 4) {
667 *num = av_int2float(avio_rb32(pb));
668 } else if (size == 8){
669 *num = av_int2double(avio_rb64(pb));
671 return AVERROR_INVALIDDATA;
677 * Read the next element as an ASCII string.
678 * 0 is success, < 0 is failure.
680 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
684 /* EBML strings are usually not 0-terminated, so we allocate one
685 * byte more, read the string and NULL-terminate it ourselves. */
686 if (!(res = av_malloc(size + 1)))
687 return AVERROR(ENOMEM);
688 if (avio_read(pb, (uint8_t *) res, size) != size) {
700 * Read the next element as binary data.
701 * 0 is success, < 0 is failure.
703 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
706 if (!(bin->data = av_malloc(length)))
707 return AVERROR(ENOMEM);
710 bin->pos = avio_tell(pb);
711 if (avio_read(pb, bin->data, length) != length) {
712 av_freep(&bin->data);
720 * Read the next element, but only the header. The contents
721 * are supposed to be sub-elements which can be read separately.
722 * 0 is success, < 0 is failure.
724 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
726 AVIOContext *pb = matroska->ctx->pb;
727 MatroskaLevel *level;
729 if (matroska->num_levels >= EBML_MAX_DEPTH) {
730 av_log(matroska->ctx, AV_LOG_ERROR,
731 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
732 return AVERROR(ENOSYS);
735 level = &matroska->levels[matroska->num_levels++];
736 level->start = avio_tell(pb);
737 level->length = length;
743 * Read signed/unsigned "EBML" numbers.
744 * Return: number of bytes processed, < 0 on error
746 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
747 uint8_t *data, uint32_t size, uint64_t *num)
750 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
751 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
755 * Same as above, but signed.
757 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
758 uint8_t *data, uint32_t size, int64_t *num)
763 /* read as unsigned number first */
764 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
767 /* make signed (weird way) */
768 *num = unum - ((1LL << (7*res - 1)) - 1);
773 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
774 EbmlSyntax *syntax, void *data);
776 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
777 uint32_t id, void *data)
780 for (i=0; syntax[i].id; i++)
781 if (id == syntax[i].id)
783 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
784 matroska->num_levels > 0 &&
785 matroska->levels[matroska->num_levels-1].length == 0xffffffffffffff)
786 return 0; // we reached the end of an unknown size cluster
787 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
788 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
789 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
790 return AVERROR_INVALIDDATA;
792 return ebml_parse_elem(matroska, &syntax[i], data);
795 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
798 if (!matroska->current_id) {
800 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
803 matroska->current_id = id | 1 << 7*res;
805 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
808 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
813 for (i=0; syntax[i].id; i++)
814 switch (syntax[i].type) {
816 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
819 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
823 *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
827 while (!res && !ebml_level_end(matroska))
828 res = ebml_parse(matroska, syntax, data);
833 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
834 EbmlSyntax *syntax, void *data)
836 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
839 // max. 16 MB for strings
840 [EBML_STR] = 0x1000000,
841 [EBML_UTF8] = 0x1000000,
842 // max. 256 MB for binary data
843 [EBML_BIN] = 0x10000000,
844 // no limits for anything else
846 AVIOContext *pb = matroska->ctx->pb;
847 uint32_t id = syntax->id;
852 data = (char *)data + syntax->data_offset;
853 if (syntax->list_elem_size) {
854 EbmlList *list = data;
855 newelem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
857 return AVERROR(ENOMEM);
858 list->elem = newelem;
859 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
860 memset(data, 0, syntax->list_elem_size);
864 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
865 matroska->current_id = 0;
866 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
868 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
869 av_log(matroska->ctx, AV_LOG_ERROR,
870 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
871 length, max_lengths[syntax->type], syntax->type);
872 return AVERROR_INVALIDDATA;
876 switch (syntax->type) {
877 case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
878 case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
880 case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
881 case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
882 case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
884 if (id == MATROSKA_ID_SEGMENT)
885 matroska->segment_start = avio_tell(matroska->ctx->pb);
886 return ebml_parse_nest(matroska, syntax->def.n, data);
887 case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
888 case EBML_STOP: return 1;
889 default: return avio_skip(pb,length)<0 ? AVERROR(EIO) : 0;
891 if (res == AVERROR_INVALIDDATA)
892 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
893 else if (res == AVERROR(EIO))
894 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
898 static void ebml_free(EbmlSyntax *syntax, void *data)
901 for (i=0; syntax[i].id; i++) {
902 void *data_off = (char *)data + syntax[i].data_offset;
903 switch (syntax[i].type) {
905 case EBML_UTF8: av_freep(data_off); break;
906 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
908 if (syntax[i].list_elem_size) {
909 EbmlList *list = data_off;
910 char *ptr = list->elem;
911 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
912 ebml_free(syntax[i].def.n, ptr);
915 ebml_free(syntax[i].def.n, data_off);
925 static int matroska_probe(AVProbeData *p)
928 int len_mask = 0x80, size = 1, n = 1, i;
931 if (AV_RB32(p->buf) != EBML_ID_HEADER)
934 /* length of header */
936 while (size <= 8 && !(total & len_mask)) {
942 total &= (len_mask - 1);
944 total = (total << 8) | p->buf[4 + n++];
946 /* Does the probe data contain the whole header? */
947 if (p->buf_size < 4 + size + total)
950 /* The header should contain a known document type. For now,
951 * we don't parse the whole header but simply check for the
952 * availability of that array of characters inside the header.
953 * Not fully fool-proof, but good enough. */
954 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
955 int probelen = strlen(matroska_doctypes[i]);
956 if (total < probelen)
958 for (n = 4+size; n <= 4+size+total-probelen; n++)
959 if (!memcmp(p->buf+n, matroska_doctypes[i], probelen))
960 return AVPROBE_SCORE_MAX;
963 // probably valid EBML header but no recognized doctype
964 return AVPROBE_SCORE_MAX/2;
967 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
970 MatroskaTrack *tracks = matroska->tracks.elem;
973 for (i=0; i < matroska->tracks.nb_elem; i++)
974 if (tracks[i].num == num)
977 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
981 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
982 MatroskaTrack *track)
984 MatroskaTrackEncoding *encodings = track->encodings.elem;
985 uint8_t* data = *buf;
986 int isize = *buf_size;
987 uint8_t* pkt_data = NULL;
988 uint8_t av_unused *newpktdata;
989 int pkt_size = isize;
993 if (pkt_size >= 10000000)
996 switch (encodings[0].compression.algo) {
997 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
998 return encodings[0].compression.settings.size;
999 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1001 olen = pkt_size *= 3;
1002 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1006 pkt_data = newpktdata;
1007 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1008 } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
1014 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
1015 z_stream zstream = {0};
1016 if (inflateInit(&zstream) != Z_OK)
1018 zstream.next_in = data;
1019 zstream.avail_in = isize;
1022 newpktdata = av_realloc(pkt_data, pkt_size);
1024 inflateEnd(&zstream);
1027 pkt_data = newpktdata;
1028 zstream.avail_out = pkt_size - zstream.total_out;
1029 zstream.next_out = pkt_data + zstream.total_out;
1030 result = inflate(&zstream, Z_NO_FLUSH);
1031 } while (result==Z_OK && pkt_size<10000000);
1032 pkt_size = zstream.total_out;
1033 inflateEnd(&zstream);
1034 if (result != Z_STREAM_END)
1040 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
1041 bz_stream bzstream = {0};
1042 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1044 bzstream.next_in = data;
1045 bzstream.avail_in = isize;
1048 newpktdata = av_realloc(pkt_data, pkt_size);
1050 BZ2_bzDecompressEnd(&bzstream);
1053 pkt_data = newpktdata;
1054 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1055 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1056 result = BZ2_bzDecompress(&bzstream);
1057 } while (result==BZ_OK && pkt_size<10000000);
1058 pkt_size = bzstream.total_out_lo32;
1059 BZ2_bzDecompressEnd(&bzstream);
1060 if (result != BZ_STREAM_END)
1070 *buf_size = pkt_size;
1077 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
1078 AVPacket *pkt, uint64_t display_duration)
1080 char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
1081 for (; *ptr!=',' && ptr<end-1; ptr++);
1084 for (; *ptr!=',' && ptr<end-1; ptr++);
1086 int64_t end_pts = pkt->pts + display_duration;
1087 int sc = matroska->time_scale * pkt->pts / 10000000;
1088 int ec = matroska->time_scale * end_pts / 10000000;
1089 int sh, sm, ss, eh, em, es, len;
1090 sh = sc/360000; sc -= 360000*sh;
1091 sm = sc/ 6000; sc -= 6000*sm;
1092 ss = sc/ 100; sc -= 100*ss;
1093 eh = ec/360000; ec -= 360000*eh;
1094 em = ec/ 6000; ec -= 6000*em;
1095 es = ec/ 100; ec -= 100*es;
1097 len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
1098 if (!(line = av_malloc(len)))
1100 snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
1101 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
1104 pkt->size = strlen(line);
1108 static int matroska_merge_packets(AVPacket *out, AVPacket *in)
1110 void *newdata = av_realloc(out->data, out->size+in->size);
1112 return AVERROR(ENOMEM);
1113 out->data = newdata;
1114 memcpy(out->data+out->size, in->data, in->size);
1115 out->size += in->size;
1116 av_destruct_packet(in);
1121 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1122 AVDictionary **metadata, char *prefix)
1124 MatroskaTag *tags = list->elem;
1128 for (i=0; i < list->nb_elem; i++) {
1129 const char *lang = strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1131 if (!tags[i].name) {
1132 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1135 if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1136 else av_strlcpy(key, tags[i].name, sizeof(key));
1137 if (tags[i].def || !lang) {
1138 av_dict_set(metadata, key, tags[i].string, 0);
1139 if (tags[i].sub.nb_elem)
1140 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1143 av_strlcat(key, "-", sizeof(key));
1144 av_strlcat(key, lang, sizeof(key));
1145 av_dict_set(metadata, key, tags[i].string, 0);
1146 if (tags[i].sub.nb_elem)
1147 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1150 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1153 static void matroska_convert_tags(AVFormatContext *s)
1155 MatroskaDemuxContext *matroska = s->priv_data;
1156 MatroskaTags *tags = matroska->tags.elem;
1159 for (i=0; i < matroska->tags.nb_elem; i++) {
1160 if (tags[i].target.attachuid) {
1161 MatroskaAttachement *attachment = matroska->attachments.elem;
1162 for (j=0; j<matroska->attachments.nb_elem; j++)
1163 if (attachment[j].uid == tags[i].target.attachuid
1164 && attachment[j].stream)
1165 matroska_convert_tag(s, &tags[i].tag,
1166 &attachment[j].stream->metadata, NULL);
1167 } else if (tags[i].target.chapteruid) {
1168 MatroskaChapter *chapter = matroska->chapters.elem;
1169 for (j=0; j<matroska->chapters.nb_elem; j++)
1170 if (chapter[j].uid == tags[i].target.chapteruid
1171 && chapter[j].chapter)
1172 matroska_convert_tag(s, &tags[i].tag,
1173 &chapter[j].chapter->metadata, NULL);
1174 } else if (tags[i].target.trackuid) {
1175 MatroskaTrack *track = matroska->tracks.elem;
1176 for (j=0; j<matroska->tracks.nb_elem; j++)
1177 if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1178 matroska_convert_tag(s, &tags[i].tag,
1179 &track[j].stream->metadata, NULL);
1181 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1182 tags[i].target.type);
1187 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska, int idx)
1189 EbmlList *seekhead_list = &matroska->seekhead;
1190 MatroskaSeekhead *seekhead = seekhead_list->elem;
1191 uint32_t level_up = matroska->level_up;
1192 int64_t before_pos = avio_tell(matroska->ctx->pb);
1193 uint32_t saved_id = matroska->current_id;
1194 MatroskaLevel level;
1198 if (idx >= seekhead_list->nb_elem
1199 || seekhead[idx].id == MATROSKA_ID_SEEKHEAD
1200 || seekhead[idx].id == MATROSKA_ID_CLUSTER)
1204 offset = seekhead[idx].pos + matroska->segment_start;
1205 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1206 /* We don't want to lose our seekhead level, so we add
1207 * a dummy. This is a crude hack. */
1208 if (matroska->num_levels == EBML_MAX_DEPTH) {
1209 av_log(matroska->ctx, AV_LOG_INFO,
1210 "Max EBML element depth (%d) reached, "
1211 "cannot parse further.\n", EBML_MAX_DEPTH);
1212 ret = AVERROR_INVALIDDATA;
1215 level.length = (uint64_t)-1;
1216 matroska->levels[matroska->num_levels] = level;
1217 matroska->num_levels++;
1218 matroska->current_id = 0;
1220 ret = ebml_parse(matroska, matroska_segment, matroska);
1222 /* remove dummy level */
1223 while (matroska->num_levels) {
1224 uint64_t length = matroska->levels[--matroska->num_levels].length;
1225 if (length == (uint64_t)-1)
1231 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1232 matroska->level_up = level_up;
1233 matroska->current_id = saved_id;
1238 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1240 EbmlList *seekhead_list = &matroska->seekhead;
1241 int64_t before_pos = avio_tell(matroska->ctx->pb);
1244 // we should not do any seeking in the streaming case
1245 if (!matroska->ctx->pb->seekable ||
1246 (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1249 for (i = 0; i < seekhead_list->nb_elem; i++) {
1250 MatroskaSeekhead *seekhead = seekhead_list->elem;
1251 if (seekhead[i].pos <= before_pos)
1254 // defer cues parsing until we actually need cue data.
1255 if (seekhead[i].id == MATROSKA_ID_CUES) {
1256 matroska->cues_parsing_deferred = 1;
1260 if (matroska_parse_seekhead_entry(matroska, i) < 0)
1265 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1266 EbmlList *seekhead_list = &matroska->seekhead;
1267 MatroskaSeekhead *seekhead = seekhead_list->elem;
1268 EbmlList *index_list;
1269 MatroskaIndex *index;
1270 int index_scale = 1;
1273 for (i = 0; i < seekhead_list->nb_elem; i++)
1274 if (seekhead[i].id == MATROSKA_ID_CUES)
1276 assert(i <= seekhead_list->nb_elem);
1278 matroska_parse_seekhead_entry(matroska, i);
1280 index_list = &matroska->index;
1281 index = index_list->elem;
1282 if (index_list->nb_elem
1283 && index[0].time > 1E14/matroska->time_scale) {
1284 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1285 index_scale = matroska->time_scale;
1287 for (i = 0; i < index_list->nb_elem; i++) {
1288 EbmlList *pos_list = &index[i].pos;
1289 MatroskaIndexPos *pos = pos_list->elem;
1290 for (j = 0; j < pos_list->nb_elem; j++) {
1291 MatroskaTrack *track = matroska_find_track_by_num(matroska, pos[j].track);
1292 if (track && track->stream)
1293 av_add_index_entry(track->stream,
1294 pos[j].pos + matroska->segment_start,
1295 index[i].time/index_scale, 0, 0,
1301 static int matroska_aac_profile(char *codec_id)
1303 static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1306 for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1307 if (strstr(codec_id, aac_profiles[profile]))
1312 static int matroska_aac_sri(int samplerate)
1316 for (sri=0; sri<FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1317 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1322 static int matroska_read_header(AVFormatContext *s)
1324 MatroskaDemuxContext *matroska = s->priv_data;
1325 EbmlList *attachements_list = &matroska->attachments;
1326 MatroskaAttachement *attachements;
1327 EbmlList *chapters_list = &matroska->chapters;
1328 MatroskaChapter *chapters;
1329 MatroskaTrack *tracks;
1330 uint64_t max_start = 0;
1337 /* First read the EBML header. */
1338 if (ebml_parse(matroska, ebml_syntax, &ebml)
1339 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1340 || ebml.id_length > sizeof(uint32_t) || ebml.doctype_version > 2) {
1341 av_log(matroska->ctx, AV_LOG_ERROR,
1342 "EBML header using unsupported features\n"
1343 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1344 ebml.version, ebml.doctype, ebml.doctype_version);
1345 ebml_free(ebml_syntax, &ebml);
1346 return AVERROR_PATCHWELCOME;
1348 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
1349 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
1351 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
1352 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
1353 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
1354 ebml_free(ebml_syntax, &ebml);
1355 return AVERROR_INVALIDDATA;
1358 ebml_free(ebml_syntax, &ebml);
1360 /* The next thing is a segment. */
1361 if ((res = ebml_parse(matroska, matroska_segments, matroska)) < 0)
1363 matroska_execute_seekhead(matroska);
1365 if (!matroska->time_scale)
1366 matroska->time_scale = 1000000;
1367 if (matroska->duration)
1368 matroska->ctx->duration = matroska->duration * matroska->time_scale
1369 * 1000 / AV_TIME_BASE;
1370 av_dict_set(&s->metadata, "title", matroska->title, 0);
1372 tracks = matroska->tracks.elem;
1373 for (i=0; i < matroska->tracks.nb_elem; i++) {
1374 MatroskaTrack *track = &tracks[i];
1375 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1376 EbmlList *encodings_list = &tracks->encodings;
1377 MatroskaTrackEncoding *encodings = encodings_list->elem;
1378 uint8_t *extradata = NULL;
1379 int extradata_size = 0;
1380 int extradata_offset = 0;
1383 /* Apply some sanity checks. */
1384 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1385 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1386 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1387 av_log(matroska->ctx, AV_LOG_INFO,
1388 "Unknown or unsupported track type %"PRIu64"\n",
1392 if (track->codec_id == NULL)
1395 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1396 if (!track->default_duration && track->video.frame_rate > 0)
1397 track->default_duration = 1000000000/track->video.frame_rate;
1398 if (!track->video.display_width)
1399 track->video.display_width = track->video.pixel_width;
1400 if (!track->video.display_height)
1401 track->video.display_height = track->video.pixel_height;
1402 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1403 if (!track->audio.out_samplerate)
1404 track->audio.out_samplerate = track->audio.samplerate;
1406 if (encodings_list->nb_elem > 1) {
1407 av_log(matroska->ctx, AV_LOG_ERROR,
1408 "Multiple combined encodings not supported");
1409 } else if (encodings_list->nb_elem == 1) {
1410 if (encodings[0].type ||
1411 (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1413 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1416 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1418 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
1419 encodings[0].scope = 0;
1420 av_log(matroska->ctx, AV_LOG_ERROR,
1421 "Unsupported encoding type");
1422 } else if (track->codec_priv.size && encodings[0].scope&2) {
1423 uint8_t *codec_priv = track->codec_priv.data;
1424 int offset = matroska_decode_buffer(&track->codec_priv.data,
1425 &track->codec_priv.size,
1428 track->codec_priv.data = NULL;
1429 track->codec_priv.size = 0;
1430 av_log(matroska->ctx, AV_LOG_ERROR,
1431 "Failed to decode codec private data\n");
1432 } else if (offset > 0) {
1433 track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
1434 memcpy(track->codec_priv.data,
1435 encodings[0].compression.settings.data, offset);
1436 memcpy(track->codec_priv.data+offset, codec_priv,
1437 track->codec_priv.size);
1438 track->codec_priv.size += offset;
1440 if (codec_priv != track->codec_priv.data)
1441 av_free(codec_priv);
1445 for(j=0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++){
1446 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1447 strlen(ff_mkv_codec_tags[j].str))){
1448 codec_id= ff_mkv_codec_tags[j].id;
1453 st = track->stream = avformat_new_stream(s, NULL);
1455 return AVERROR(ENOMEM);
1457 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1458 && track->codec_priv.size >= 40
1459 && track->codec_priv.data != NULL) {
1460 track->ms_compat = 1;
1461 track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1462 codec_id = ff_codec_get_id(ff_codec_bmp_tags, track->video.fourcc);
1463 extradata_offset = 40;
1464 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1465 && track->codec_priv.size >= 14
1466 && track->codec_priv.data != NULL) {
1468 ffio_init_context(&b, track->codec_priv.data, track->codec_priv.size,
1469 AVIO_FLAG_READ, NULL, NULL, NULL, NULL);
1470 ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1473 codec_id = st->codec->codec_id;
1474 extradata_offset = FFMIN(track->codec_priv.size, 18);
1475 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1476 && (track->codec_priv.size >= 86)
1477 && (track->codec_priv.data != NULL)) {
1478 track->video.fourcc = AV_RL32(track->codec_priv.data);
1479 codec_id=ff_codec_get_id(ff_codec_movvideo_tags, track->video.fourcc);
1480 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1481 switch (track->audio.bitdepth) {
1482 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1483 case 24: codec_id = AV_CODEC_ID_PCM_S24BE; break;
1484 case 32: codec_id = AV_CODEC_ID_PCM_S32BE; break;
1486 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1487 switch (track->audio.bitdepth) {
1488 case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1489 case 24: codec_id = AV_CODEC_ID_PCM_S24LE; break;
1490 case 32: codec_id = AV_CODEC_ID_PCM_S32LE; break;
1492 } else if (codec_id==AV_CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1493 codec_id = AV_CODEC_ID_PCM_F64LE;
1494 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1495 int profile = matroska_aac_profile(track->codec_id);
1496 int sri = matroska_aac_sri(track->audio.samplerate);
1497 extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1498 if (extradata == NULL)
1499 return AVERROR(ENOMEM);
1500 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1501 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1502 if (strstr(track->codec_id, "SBR")) {
1503 sri = matroska_aac_sri(track->audio.out_samplerate);
1504 extradata[2] = 0x56;
1505 extradata[3] = 0xE5;
1506 extradata[4] = 0x80 | (sri<<3);
1510 } else if (codec_id == AV_CODEC_ID_TTA) {
1511 extradata_size = 30;
1512 extradata = av_mallocz(extradata_size);
1513 if (extradata == NULL)
1514 return AVERROR(ENOMEM);
1515 ffio_init_context(&b, extradata, extradata_size, 1,
1516 NULL, NULL, NULL, NULL);
1517 avio_write(&b, "TTA1", 4);
1519 avio_wl16(&b, track->audio.channels);
1520 avio_wl16(&b, track->audio.bitdepth);
1521 avio_wl32(&b, track->audio.out_samplerate);
1522 avio_wl32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1523 } else if (codec_id == AV_CODEC_ID_RV10 || codec_id == AV_CODEC_ID_RV20 ||
1524 codec_id == AV_CODEC_ID_RV30 || codec_id == AV_CODEC_ID_RV40) {
1525 extradata_offset = 26;
1526 } else if (codec_id == AV_CODEC_ID_RA_144) {
1527 track->audio.out_samplerate = 8000;
1528 track->audio.channels = 1;
1529 } else if (codec_id == AV_CODEC_ID_RA_288 || codec_id == AV_CODEC_ID_COOK ||
1530 codec_id == AV_CODEC_ID_ATRAC3 || codec_id == AV_CODEC_ID_SIPR) {
1532 ffio_init_context(&b, track->codec_priv.data,track->codec_priv.size,
1533 0, NULL, NULL, NULL, NULL);
1535 flavor = avio_rb16(&b);
1536 track->audio.coded_framesize = avio_rb32(&b);
1538 track->audio.sub_packet_h = avio_rb16(&b);
1539 track->audio.frame_size = avio_rb16(&b);
1540 track->audio.sub_packet_size = avio_rb16(&b);
1541 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1542 if (codec_id == AV_CODEC_ID_RA_288) {
1543 st->codec->block_align = track->audio.coded_framesize;
1544 track->codec_priv.size = 0;
1546 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1547 const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1548 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1549 st->codec->bit_rate = sipr_bit_rate[flavor];
1551 st->codec->block_align = track->audio.sub_packet_size;
1552 extradata_offset = 78;
1555 track->codec_priv.size -= extradata_offset;
1557 if (codec_id == AV_CODEC_ID_NONE)
1558 av_log(matroska->ctx, AV_LOG_INFO,
1559 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1561 if (track->time_scale < 0.01)
1562 track->time_scale = 1.0;
1563 avpriv_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1565 st->codec->codec_id = codec_id;
1567 if (strcmp(track->language, "und"))
1568 av_dict_set(&st->metadata, "language", track->language, 0);
1569 av_dict_set(&st->metadata, "title", track->name, 0);
1571 if (track->flag_default)
1572 st->disposition |= AV_DISPOSITION_DEFAULT;
1573 if (track->flag_forced)
1574 st->disposition |= AV_DISPOSITION_FORCED;
1576 if (!st->codec->extradata) {
1578 st->codec->extradata = extradata;
1579 st->codec->extradata_size = extradata_size;
1580 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1581 st->codec->extradata = av_mallocz(track->codec_priv.size +
1582 FF_INPUT_BUFFER_PADDING_SIZE);
1583 if(st->codec->extradata == NULL)
1584 return AVERROR(ENOMEM);
1585 st->codec->extradata_size = track->codec_priv.size;
1586 memcpy(st->codec->extradata,
1587 track->codec_priv.data + extradata_offset,
1588 track->codec_priv.size);
1592 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1593 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1594 st->codec->codec_tag = track->video.fourcc;
1595 st->codec->width = track->video.pixel_width;
1596 st->codec->height = track->video.pixel_height;
1597 av_reduce(&st->sample_aspect_ratio.num,
1598 &st->sample_aspect_ratio.den,
1599 st->codec->height * track->video.display_width,
1600 st->codec-> width * track->video.display_height,
1602 if (st->codec->codec_id != AV_CODEC_ID_H264)
1603 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1604 if (track->default_duration) {
1605 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1606 1000000000, track->default_duration, 30000);
1607 #if FF_API_R_FRAME_RATE
1608 st->r_frame_rate = st->avg_frame_rate;
1611 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1612 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1613 st->codec->sample_rate = track->audio.out_samplerate;
1614 st->codec->channels = track->audio.channels;
1615 if (st->codec->codec_id != AV_CODEC_ID_AAC)
1616 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1617 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1618 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1619 if (st->codec->codec_id == AV_CODEC_ID_SSA)
1620 matroska->contains_ssa = 1;
1624 attachements = attachements_list->elem;
1625 for (j=0; j<attachements_list->nb_elem; j++) {
1626 if (!(attachements[j].filename && attachements[j].mime &&
1627 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1628 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1630 AVStream *st = avformat_new_stream(s, NULL);
1633 av_dict_set(&st->metadata, "filename",attachements[j].filename, 0);
1634 av_dict_set(&st->metadata, "mimetype", attachements[j].mime, 0);
1635 st->codec->codec_id = AV_CODEC_ID_NONE;
1636 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
1637 st->codec->extradata = av_malloc(attachements[j].bin.size);
1638 if(st->codec->extradata == NULL)
1640 st->codec->extradata_size = attachements[j].bin.size;
1641 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1643 for (i=0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
1644 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1645 strlen(ff_mkv_mime_tags[i].str))) {
1646 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1650 attachements[j].stream = st;
1654 chapters = chapters_list->elem;
1655 for (i=0; i<chapters_list->nb_elem; i++)
1656 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1657 && (max_start==0 || chapters[i].start > max_start)) {
1658 chapters[i].chapter =
1659 avpriv_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1660 chapters[i].start, chapters[i].end,
1662 av_dict_set(&chapters[i].chapter->metadata,
1663 "title", chapters[i].title, 0);
1664 max_start = chapters[i].start;
1667 matroska_convert_tags(s);
1673 * Put one packet in an application-supplied AVPacket struct.
1674 * Returns 0 on success or -1 on failure.
1676 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1679 if (matroska->num_packets > 0) {
1680 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1681 av_free(matroska->packets[0]);
1682 if (matroska->num_packets > 1) {
1684 memmove(&matroska->packets[0], &matroska->packets[1],
1685 (matroska->num_packets - 1) * sizeof(AVPacket *));
1686 newpackets = av_realloc(matroska->packets,
1687 (matroska->num_packets - 1) * sizeof(AVPacket *));
1689 matroska->packets = newpackets;
1691 av_freep(&matroska->packets);
1692 matroska->prev_pkt = NULL;
1694 matroska->num_packets--;
1702 * Free all packets in our internal queue.
1704 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1706 if (matroska->packets) {
1708 for (n = 0; n < matroska->num_packets; n++) {
1709 av_free_packet(matroska->packets[n]);
1710 av_free(matroska->packets[n]);
1712 av_freep(&matroska->packets);
1713 matroska->num_packets = 0;
1717 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
1718 int size, int64_t pos, uint64_t cluster_time,
1719 uint64_t duration, int is_keyframe,
1720 int64_t cluster_pos)
1722 uint64_t timecode = AV_NOPTS_VALUE;
1723 MatroskaTrack *track;
1728 uint32_t *lace_size = NULL;
1729 int n, flags, laces = 0;
1732 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
1733 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
1739 track = matroska_find_track_by_num(matroska, num);
1740 if (!track || !track->stream) {
1741 av_log(matroska->ctx, AV_LOG_INFO,
1742 "Invalid stream %"PRIu64" or size %u\n", num, size);
1743 return AVERROR_INVALIDDATA;
1744 } else if (size <= 3)
1747 if (st->discard >= AVDISCARD_ALL)
1749 if (duration == AV_NOPTS_VALUE)
1750 duration = track->default_duration / matroska->time_scale;
1752 block_time = AV_RB16(data);
1756 if (is_keyframe == -1)
1757 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
1759 if (cluster_time != (uint64_t)-1
1760 && (block_time >= 0 || cluster_time >= -block_time)) {
1761 timecode = cluster_time + block_time;
1762 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
1763 && timecode < track->end_timecode)
1764 is_keyframe = 0; /* overlapping subtitles are not key frame */
1766 av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
1767 track->end_timecode = FFMAX(track->end_timecode, timecode+duration);
1770 if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1771 if (!is_keyframe || timecode < matroska->skip_to_timecode)
1773 matroska->skip_to_keyframe = 0;
1776 switch ((flags & 0x06) >> 1) {
1777 case 0x0: /* no lacing */
1779 lace_size = av_mallocz(sizeof(int));
1780 lace_size[0] = size;
1783 case 0x1: /* Xiph lacing */
1784 case 0x2: /* fixed-size lacing */
1785 case 0x3: /* EBML lacing */
1786 assert(size>0); // size <=3 is checked before size-=3 above
1787 laces = (*data) + 1;
1790 lace_size = av_mallocz(laces * sizeof(int));
1792 switch ((flags & 0x06) >> 1) {
1793 case 0x1: /* Xiph lacing */ {
1796 for (n = 0; res == 0 && n < laces - 1; n++) {
1803 lace_size[n] += temp;
1809 total += lace_size[n];
1811 lace_size[n] = size - total;
1815 case 0x2: /* fixed-size lacing */
1816 for (n = 0; n < laces; n++)
1817 lace_size[n] = size / laces;
1820 case 0x3: /* EBML lacing */ {
1822 n = matroska_ebmlnum_uint(matroska, data, size, &num);
1824 av_log(matroska->ctx, AV_LOG_INFO,
1825 "EBML block data error\n");
1830 total = lace_size[0] = num;
1831 for (n = 1; res == 0 && n < laces - 1; n++) {
1834 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1836 av_log(matroska->ctx, AV_LOG_INFO,
1837 "EBML block data error\n");
1842 lace_size[n] = lace_size[n - 1] + snum;
1843 total += lace_size[n];
1845 lace_size[laces - 1] = size - total;
1853 for (n = 0; n < laces; n++) {
1854 if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
1855 st->codec->codec_id == AV_CODEC_ID_COOK ||
1856 st->codec->codec_id == AV_CODEC_ID_SIPR ||
1857 st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
1858 st->codec->block_align && track->audio.sub_packet_size) {
1859 int a = st->codec->block_align;
1860 int sps = track->audio.sub_packet_size;
1861 int cfs = track->audio.coded_framesize;
1862 int h = track->audio.sub_packet_h;
1863 int y = track->audio.sub_packet_cnt;
1864 int w = track->audio.frame_size;
1867 if (!track->audio.pkt_cnt) {
1868 if (track->audio.sub_packet_cnt == 0)
1869 track->audio.buf_timecode = timecode;
1870 if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
1871 if (size < cfs * h / 2) {
1872 av_log(matroska->ctx, AV_LOG_ERROR,
1873 "Corrupt int4 RM-style audio packet size\n");
1874 res = AVERROR_INVALIDDATA;
1877 for (x=0; x<h/2; x++)
1878 memcpy(track->audio.buf+x*2*w+y*cfs,
1880 } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
1882 av_log(matroska->ctx, AV_LOG_ERROR,
1883 "Corrupt sipr RM-style audio packet size\n");
1884 res = AVERROR_INVALIDDATA;
1887 memcpy(track->audio.buf + y*w, data, w);
1889 if (size < sps * w / sps) {
1890 av_log(matroska->ctx, AV_LOG_ERROR,
1891 "Corrupt generic RM-style audio packet size\n");
1892 res = AVERROR_INVALIDDATA;
1895 for (x=0; x<w/sps; x++)
1896 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
1899 if (++track->audio.sub_packet_cnt >= h) {
1900 if (st->codec->codec_id == AV_CODEC_ID_SIPR)
1901 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
1902 track->audio.sub_packet_cnt = 0;
1903 track->audio.pkt_cnt = h*w / a;
1906 while (track->audio.pkt_cnt) {
1907 pkt = av_mallocz(sizeof(AVPacket));
1908 av_new_packet(pkt, a);
1909 memcpy(pkt->data, track->audio.buf
1910 + a * (h*w / a - track->audio.pkt_cnt--), a);
1911 pkt->pts = track->audio.buf_timecode;
1912 track->audio.buf_timecode = AV_NOPTS_VALUE;
1914 pkt->stream_index = st->index;
1915 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1918 MatroskaTrackEncoding *encodings = track->encodings.elem;
1919 int offset = 0, pkt_size = lace_size[n];
1920 uint8_t *pkt_data = data;
1922 if (pkt_size > size) {
1923 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
1927 if (encodings && encodings->scope & 1) {
1928 offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
1933 pkt = av_mallocz(sizeof(AVPacket));
1934 /* XXX: prevent data copy... */
1935 if (av_new_packet(pkt, pkt_size+offset) < 0) {
1937 res = AVERROR(ENOMEM);
1941 memcpy (pkt->data, encodings->compression.settings.data, offset);
1942 memcpy (pkt->data+offset, pkt_data, pkt_size);
1944 if (pkt_data != data)
1948 pkt->flags = is_keyframe;
1949 pkt->stream_index = st->index;
1951 if (track->ms_compat)
1952 pkt->dts = timecode;
1954 pkt->pts = timecode;
1956 if (st->codec->codec_id == AV_CODEC_ID_TEXT)
1957 pkt->convergence_duration = duration;
1958 else if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE)
1959 pkt->duration = duration;
1961 if (st->codec->codec_id == AV_CODEC_ID_SSA)
1962 matroska_fix_ass_packet(matroska, pkt, duration);
1964 if (matroska->prev_pkt &&
1965 timecode != AV_NOPTS_VALUE &&
1966 matroska->prev_pkt->pts == timecode &&
1967 matroska->prev_pkt->stream_index == st->index &&
1968 st->codec->codec_id == AV_CODEC_ID_SSA)
1969 matroska_merge_packets(matroska->prev_pkt, pkt);
1971 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1972 matroska->prev_pkt = pkt;
1976 if (timecode != AV_NOPTS_VALUE)
1977 timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
1978 data += lace_size[n];
1979 size -= lace_size[n];
1988 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
1990 EbmlList *blocks_list;
1991 MatroskaBlock *blocks;
1993 res = ebml_parse(matroska,
1994 matroska_cluster_incremental_parsing,
1995 &matroska->current_cluster);
1998 if (matroska->current_cluster_pos)
1999 ebml_level_end(matroska);
2000 ebml_free(matroska_cluster, &matroska->current_cluster);
2001 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2002 matroska->current_cluster_num_blocks = 0;
2003 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2004 matroska->prev_pkt = NULL;
2005 /* sizeof the ID which was already read */
2006 if (matroska->current_id)
2007 matroska->current_cluster_pos -= 4;
2008 res = ebml_parse(matroska,
2009 matroska_clusters_incremental,
2010 &matroska->current_cluster);
2011 /* Try parsing the block again. */
2013 res = ebml_parse(matroska,
2014 matroska_cluster_incremental_parsing,
2015 &matroska->current_cluster);
2019 matroska->current_cluster_num_blocks <
2020 matroska->current_cluster.blocks.nb_elem) {
2021 blocks_list = &matroska->current_cluster.blocks;
2022 blocks = blocks_list->elem;
2024 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2025 i = blocks_list->nb_elem - 1;
2026 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2027 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2028 if (!blocks[i].non_simple)
2029 blocks[i].duration = AV_NOPTS_VALUE;
2030 res = matroska_parse_block(matroska,
2031 blocks[i].bin.data, blocks[i].bin.size,
2033 matroska->current_cluster.timecode,
2034 blocks[i].duration, is_keyframe,
2035 matroska->current_cluster_pos);
2039 if (res < 0) matroska->done = 1;
2043 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
2045 MatroskaCluster cluster = { 0 };
2046 EbmlList *blocks_list;
2047 MatroskaBlock *blocks;
2050 if (!matroska->contains_ssa)
2051 return matroska_parse_cluster_incremental(matroska);
2052 pos = avio_tell(matroska->ctx->pb);
2053 matroska->prev_pkt = NULL;
2054 if (matroska->current_id)
2055 pos -= 4; /* sizeof the ID which was already read */
2056 res = ebml_parse(matroska, matroska_clusters, &cluster);
2057 blocks_list = &cluster.blocks;
2058 blocks = blocks_list->elem;
2059 for (i=0; i<blocks_list->nb_elem && !res; i++)
2060 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2061 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2062 if (!blocks[i].non_simple)
2063 blocks[i].duration = AV_NOPTS_VALUE;
2064 res=matroska_parse_block(matroska,
2065 blocks[i].bin.data, blocks[i].bin.size,
2066 blocks[i].bin.pos, cluster.timecode,
2067 blocks[i].duration, is_keyframe,
2070 ebml_free(matroska_cluster, &cluster);
2071 if (res < 0) matroska->done = 1;
2075 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
2077 MatroskaDemuxContext *matroska = s->priv_data;
2080 while (!ret && matroska_deliver_packet(matroska, pkt)) {
2083 ret = matroska_parse_cluster(matroska);
2086 if (ret == AVERROR_INVALIDDATA) {
2087 pkt->flags |= AV_PKT_FLAG_CORRUPT;
2094 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2095 int64_t timestamp, int flags)
2097 MatroskaDemuxContext *matroska = s->priv_data;
2098 MatroskaTrack *tracks = matroska->tracks.elem;
2099 AVStream *st = s->streams[stream_index];
2100 int i, index, index_sub, index_min;
2102 /* Parse the CUES now since we need the index data to seek. */
2103 if (matroska->cues_parsing_deferred) {
2104 matroska_parse_cues(matroska);
2105 matroska->cues_parsing_deferred = 0;
2108 if (!st->nb_index_entries)
2110 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2112 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2113 avio_seek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
2114 matroska->current_id = 0;
2115 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2116 matroska->prev_pkt = NULL;
2117 matroska_clear_queue(matroska);
2118 if (matroska_parse_cluster(matroska) < 0)
2123 matroska_clear_queue(matroska);
2128 for (i=0; i < matroska->tracks.nb_elem; i++) {
2129 tracks[i].audio.pkt_cnt = 0;
2130 tracks[i].audio.sub_packet_cnt = 0;
2131 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2132 tracks[i].end_timecode = 0;
2133 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
2134 && !tracks[i].stream->discard != AVDISCARD_ALL) {
2135 index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
2137 && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
2138 && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
2139 index_min = index_sub;
2143 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2144 matroska->current_id = 0;
2145 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
2146 matroska->skip_to_timecode = st->index_entries[index].timestamp;
2148 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2152 static int matroska_read_close(AVFormatContext *s)
2154 MatroskaDemuxContext *matroska = s->priv_data;
2155 MatroskaTrack *tracks = matroska->tracks.elem;
2158 matroska_clear_queue(matroska);
2160 for (n=0; n < matroska->tracks.nb_elem; n++)
2161 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2162 av_free(tracks[n].audio.buf);
2163 ebml_free(matroska_cluster, &matroska->current_cluster);
2164 ebml_free(matroska_segment, matroska);
2169 AVInputFormat ff_matroska_demuxer = {
2170 .name = "matroska,webm",
2171 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
2172 .priv_data_size = sizeof(MatroskaDemuxContext),
2173 .read_probe = matroska_probe,
2174 .read_header = matroska_read_header,
2175 .read_packet = matroska_read_packet,
2176 .read_close = matroska_read_close,
2177 .read_seek = matroska_read_seek,