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
23 * @file libavformat/matroskadec.c
24 * Matroska file demuxer
25 * by Ronald Bultje <rbultje@ronald.bitfreak.net>
26 * with a little help from Moritz Bunkus <moritz@bunkus.org>
27 * totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28 * Specs available on the Matroska project page: http://www.matroska.org/.
33 /* For ff_codec_get_id(). */
37 #include "libavcodec/mpeg4audio.h"
38 #include "libavutil/intfloat_readwrite.h"
39 #include "libavutil/intreadwrite.h"
40 #include "libavutil/avstring.h"
41 #include "libavutil/lzo.h"
61 typedef const struct EbmlSyntax {
70 const struct EbmlSyntax *n;
90 uint64_t doctype_version;
96 } MatroskaTrackCompression;
101 MatroskaTrackCompression compression;
102 } MatroskaTrackEncoding;
106 uint64_t display_width;
107 uint64_t display_height;
108 uint64_t pixel_width;
109 uint64_t pixel_height;
111 } MatroskaTrackVideo;
115 double out_samplerate;
119 /* real audio header (extracted from extradata) */
127 } MatroskaTrackAudio;
138 uint64_t default_duration;
139 uint64_t flag_default;
140 MatroskaTrackVideo video;
141 MatroskaTrackAudio audio;
145 int64_t end_timecode;
155 } MatroskaAttachement;
193 MatroskaTagTarget target;
208 AVFormatContext *ctx;
212 MatroskaLevel levels[EBML_MAX_DEPTH];
219 EbmlList attachments;
225 /* byte position of the segment inside the stream */
226 int64_t segment_start;
228 /* the packet queue */
236 /* What to skip before effectively reading a packet. */
237 int skip_to_keyframe;
238 uint64_t skip_to_timecode;
239 } MatroskaDemuxContext;
253 static EbmlSyntax ebml_header[] = {
254 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
255 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
256 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
257 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
258 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
259 { EBML_ID_EBMLVERSION, EBML_NONE },
260 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
264 static EbmlSyntax ebml_syntax[] = {
265 { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
269 static EbmlSyntax matroska_info[] = {
270 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
271 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
272 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
273 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
274 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
275 { MATROSKA_ID_DATEUTC, EBML_NONE },
276 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
280 static EbmlSyntax matroska_track_video[] = {
281 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
282 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
283 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
284 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
285 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
286 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
287 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
288 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
289 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
290 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
291 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
292 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
293 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_NONE },
294 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
298 static EbmlSyntax matroska_track_audio[] = {
299 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
300 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
301 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
302 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
306 static EbmlSyntax matroska_track_encoding_compression[] = {
307 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
308 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
312 static EbmlSyntax matroska_track_encoding[] = {
313 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
314 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
315 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
316 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
320 static EbmlSyntax matroska_track_encodings[] = {
321 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
325 static EbmlSyntax matroska_track[] = {
326 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
327 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack,name) },
328 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
329 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
330 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
331 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
332 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
333 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
334 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
335 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
336 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
337 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
338 { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
339 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
340 { MATROSKA_ID_TRACKFLAGFORCED, EBML_NONE },
341 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
342 { MATROSKA_ID_CODECNAME, EBML_NONE },
343 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
344 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
345 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
346 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
347 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
348 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
352 static EbmlSyntax matroska_tracks[] = {
353 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
357 static EbmlSyntax matroska_attachment[] = {
358 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
359 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
360 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
361 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
362 { MATROSKA_ID_FILEDESC, EBML_NONE },
366 static EbmlSyntax matroska_attachments[] = {
367 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
371 static EbmlSyntax matroska_chapter_display[] = {
372 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
373 { MATROSKA_ID_CHAPLANG, EBML_NONE },
377 static EbmlSyntax matroska_chapter_entry[] = {
378 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
379 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
380 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
381 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
382 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
383 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
384 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
385 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
389 static EbmlSyntax matroska_chapter[] = {
390 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
391 { MATROSKA_ID_EDITIONUID, EBML_NONE },
392 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
393 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
394 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
398 static EbmlSyntax matroska_chapters[] = {
399 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
403 static EbmlSyntax matroska_index_pos[] = {
404 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
405 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
406 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
410 static EbmlSyntax matroska_index_entry[] = {
411 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
412 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
416 static EbmlSyntax matroska_index[] = {
417 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
421 static EbmlSyntax matroska_simpletag[] = {
422 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
423 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
424 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag,lang), {.s="und"} },
425 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag,def) },
426 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
430 static EbmlSyntax matroska_tagtargets[] = {
431 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget,type) },
432 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
433 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
434 { MATROSKA_ID_TAGTARGETS_CHAPTERUID,EBML_UINT, 0, offsetof(MatroskaTagTarget,chapteruid) },
435 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
439 static EbmlSyntax matroska_tag[] = {
440 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
441 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
445 static EbmlSyntax matroska_tags[] = {
446 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
450 static EbmlSyntax matroska_seekhead_entry[] = {
451 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
452 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
456 static EbmlSyntax matroska_seekhead[] = {
457 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
461 static EbmlSyntax matroska_segment[] = {
462 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
463 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
464 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
465 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
466 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
467 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
468 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
469 { MATROSKA_ID_CLUSTER, EBML_STOP, 0, offsetof(MatroskaDemuxContext,has_cluster_id) },
473 static EbmlSyntax matroska_segments[] = {
474 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
478 static EbmlSyntax matroska_blockgroup[] = {
479 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
480 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
481 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration), {.u=AV_NOPTS_VALUE} },
482 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
483 { 1, EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
487 static EbmlSyntax matroska_cluster[] = {
488 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
489 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
490 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
491 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
492 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
496 static EbmlSyntax matroska_clusters[] = {
497 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
498 { MATROSKA_ID_INFO, EBML_NONE },
499 { MATROSKA_ID_CUES, EBML_NONE },
500 { MATROSKA_ID_TAGS, EBML_NONE },
501 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
506 * Return: Whether we reached the end of a level in the hierarchy or not.
508 static int ebml_level_end(MatroskaDemuxContext *matroska)
510 ByteIOContext *pb = matroska->ctx->pb;
511 int64_t pos = url_ftell(pb);
513 if (matroska->num_levels > 0) {
514 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
515 if (pos - level->start >= level->length) {
516 matroska->num_levels--;
524 * Read: an "EBML number", which is defined as a variable-length
525 * array of bytes. The first byte indicates the length by giving a
526 * number of 0-bits followed by a one. The position of the first
527 * "one" bit inside the first byte indicates the length of this
529 * Returns: number of bytes read, < 0 on error
531 static int ebml_read_num(MatroskaDemuxContext *matroska, ByteIOContext *pb,
532 int max_size, uint64_t *number)
534 int len_mask = 0x80, read = 1, n = 1;
537 /* The first byte tells us the length in bytes - get_byte() can normally
538 * return 0, but since that's not a valid first ebmlID byte, we can
539 * use it safely here to catch EOS. */
540 if (!(total = get_byte(pb))) {
541 /* we might encounter EOS here */
543 int64_t pos = url_ftell(pb);
544 av_log(matroska->ctx, AV_LOG_ERROR,
545 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
548 return AVERROR(EIO); /* EOS or actual I/O error */
551 /* get the length of the EBML number */
552 while (read <= max_size && !(total & len_mask)) {
556 if (read > max_size) {
557 int64_t pos = url_ftell(pb) - 1;
558 av_log(matroska->ctx, AV_LOG_ERROR,
559 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
560 (uint8_t) total, pos, pos);
561 return AVERROR_INVALIDDATA;
564 /* read out length */
567 total = (total << 8) | get_byte(pb);
575 * Read the next element as an unsigned int.
576 * 0 is success, < 0 is failure.
578 static int ebml_read_uint(ByteIOContext *pb, int size, uint64_t *num)
582 if (size < 1 || size > 8)
583 return AVERROR_INVALIDDATA;
585 /* big-endian ordering; build up number */
588 *num = (*num << 8) | get_byte(pb);
594 * Read the next element as a float.
595 * 0 is success, < 0 is failure.
597 static int ebml_read_float(ByteIOContext *pb, int size, double *num)
600 *num= av_int2flt(get_be32(pb));
602 *num= av_int2dbl(get_be64(pb));
604 return AVERROR_INVALIDDATA;
610 * Read the next element as an ASCII string.
611 * 0 is success, < 0 is failure.
613 static int ebml_read_ascii(ByteIOContext *pb, int size, char **str)
616 /* EBML strings are usually not 0-terminated, so we allocate one
617 * byte more, read the string and NULL-terminate it ourselves. */
618 if (!(*str = av_malloc(size + 1)))
619 return AVERROR(ENOMEM);
620 if (get_buffer(pb, (uint8_t *) *str, size) != size) {
630 * Read the next element as binary data.
631 * 0 is success, < 0 is failure.
633 static int ebml_read_binary(ByteIOContext *pb, int length, EbmlBin *bin)
636 if (!(bin->data = av_malloc(length)))
637 return AVERROR(ENOMEM);
640 bin->pos = url_ftell(pb);
641 if (get_buffer(pb, bin->data, length) != length)
648 * Read the next element, but only the header. The contents
649 * are supposed to be sub-elements which can be read separately.
650 * 0 is success, < 0 is failure.
652 static int ebml_read_master(MatroskaDemuxContext *matroska, int length)
654 ByteIOContext *pb = matroska->ctx->pb;
655 MatroskaLevel *level;
657 if (matroska->num_levels >= EBML_MAX_DEPTH) {
658 av_log(matroska->ctx, AV_LOG_ERROR,
659 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
660 return AVERROR(ENOSYS);
663 level = &matroska->levels[matroska->num_levels++];
664 level->start = url_ftell(pb);
665 level->length = length;
671 * Read signed/unsigned "EBML" numbers.
672 * Return: number of bytes processed, < 0 on error
674 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
675 uint8_t *data, uint32_t size, uint64_t *num)
678 init_put_byte(&pb, data, size, 0, NULL, NULL, NULL, NULL);
679 return ebml_read_num(matroska, &pb, 8, num);
683 * Same as above, but signed.
685 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
686 uint8_t *data, uint32_t size, int64_t *num)
691 /* read as unsigned number first */
692 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
695 /* make signed (weird way) */
696 *num = unum - ((1LL << (7*res - 1)) - 1);
701 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
702 EbmlSyntax *syntax, void *data);
704 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
705 uint32_t id, void *data)
708 for (i=0; syntax[i].id; i++)
709 if (id == syntax[i].id)
711 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32)
712 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
713 return ebml_parse_elem(matroska, &syntax[i], data);
716 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
720 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
722 return res < 0 ? res : ebml_parse_id(matroska, syntax, id, data);
725 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
730 for (i=0; syntax[i].id; i++)
731 switch (syntax[i].type) {
733 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
736 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
740 *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
744 while (!res && !ebml_level_end(matroska))
745 res = ebml_parse(matroska, syntax, data);
750 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
751 EbmlSyntax *syntax, void *data)
753 ByteIOContext *pb = matroska->ctx->pb;
754 uint32_t id = syntax->id;
758 data = (char *)data + syntax->data_offset;
759 if (syntax->list_elem_size) {
760 EbmlList *list = data;
761 list->elem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
762 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
763 memset(data, 0, syntax->list_elem_size);
767 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP)
768 if ((res = ebml_read_num(matroska, pb, 8, &length)) < 0)
771 switch (syntax->type) {
772 case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
773 case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
775 case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
776 case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
777 case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
779 if (id == MATROSKA_ID_SEGMENT)
780 matroska->segment_start = url_ftell(matroska->ctx->pb);
781 return ebml_parse_nest(matroska, syntax->def.n, data);
782 case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
783 case EBML_STOP: *(int *)data = 1; return 1;
784 default: return url_fseek(pb,length,SEEK_CUR)<0 ? AVERROR(EIO) : 0;
786 if (res == AVERROR_INVALIDDATA)
787 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
788 else if (res == AVERROR(EIO))
789 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
793 static void ebml_free(EbmlSyntax *syntax, void *data)
796 for (i=0; syntax[i].id; i++) {
797 void *data_off = (char *)data + syntax[i].data_offset;
798 switch (syntax[i].type) {
800 case EBML_UTF8: av_freep(data_off); break;
801 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
803 if (syntax[i].list_elem_size) {
804 EbmlList *list = data_off;
805 char *ptr = list->elem;
806 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
807 ebml_free(syntax[i].def.n, ptr);
810 ebml_free(syntax[i].def.n, data_off);
820 static int matroska_probe(AVProbeData *p)
823 int len_mask = 0x80, size = 1, n = 1;
824 static const char probe_data[] = "matroska";
827 if (AV_RB32(p->buf) != EBML_ID_HEADER)
830 /* length of header */
832 while (size <= 8 && !(total & len_mask)) {
838 total &= (len_mask - 1);
840 total = (total << 8) | p->buf[4 + n++];
842 /* Does the probe data contain the whole header? */
843 if (p->buf_size < 4 + size + total)
846 /* The header must contain the document type 'matroska'. For now,
847 * we don't parse the whole header but simply check for the
848 * availability of that array of characters inside the header.
849 * Not fully fool-proof, but good enough. */
850 for (n = 4+size; n <= 4+size+total-(sizeof(probe_data)-1); n++)
851 if (!memcmp(p->buf+n, probe_data, sizeof(probe_data)-1))
852 return AVPROBE_SCORE_MAX;
857 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
860 MatroskaTrack *tracks = matroska->tracks.elem;
863 for (i=0; i < matroska->tracks.nb_elem; i++)
864 if (tracks[i].num == num)
867 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
871 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
872 MatroskaTrack *track)
874 MatroskaTrackEncoding *encodings = track->encodings.elem;
875 uint8_t* data = *buf;
876 int isize = *buf_size;
877 uint8_t* pkt_data = NULL;
878 int pkt_size = isize;
882 switch (encodings[0].compression.algo) {
883 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
884 return encodings[0].compression.settings.size;
885 case MATROSKA_TRACK_ENCODING_COMP_LZO:
887 olen = pkt_size *= 3;
888 pkt_data = av_realloc(pkt_data, pkt_size+AV_LZO_OUTPUT_PADDING);
889 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
890 } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
896 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
897 z_stream zstream = {0};
898 if (inflateInit(&zstream) != Z_OK)
900 zstream.next_in = data;
901 zstream.avail_in = isize;
904 pkt_data = av_realloc(pkt_data, pkt_size);
905 zstream.avail_out = pkt_size - zstream.total_out;
906 zstream.next_out = pkt_data + zstream.total_out;
907 result = inflate(&zstream, Z_NO_FLUSH);
908 } while (result==Z_OK && pkt_size<10000000);
909 pkt_size = zstream.total_out;
910 inflateEnd(&zstream);
911 if (result != Z_STREAM_END)
917 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
918 bz_stream bzstream = {0};
919 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
921 bzstream.next_in = data;
922 bzstream.avail_in = isize;
925 pkt_data = av_realloc(pkt_data, pkt_size);
926 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
927 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
928 result = BZ2_bzDecompress(&bzstream);
929 } while (result==BZ_OK && pkt_size<10000000);
930 pkt_size = bzstream.total_out_lo32;
931 BZ2_bzDecompressEnd(&bzstream);
932 if (result != BZ_STREAM_END)
942 *buf_size = pkt_size;
949 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
950 AVPacket *pkt, uint64_t display_duration)
952 char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
953 for (; *ptr!=',' && ptr<end-1; ptr++);
956 for (; *ptr!=',' && ptr<end-1; ptr++);
958 int64_t end_pts = pkt->pts + display_duration;
959 int sc = matroska->time_scale * pkt->pts / 10000000;
960 int ec = matroska->time_scale * end_pts / 10000000;
961 int sh, sm, ss, eh, em, es, len;
962 sh = sc/360000; sc -= 360000*sh;
963 sm = sc/ 6000; sc -= 6000*sm;
964 ss = sc/ 100; sc -= 100*ss;
965 eh = ec/360000; ec -= 360000*eh;
966 em = ec/ 6000; ec -= 6000*em;
967 es = ec/ 100; ec -= 100*es;
969 len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
970 if (!(line = av_malloc(len)))
972 snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
973 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
976 pkt->size = strlen(line);
980 static void matroska_merge_packets(AVPacket *out, AVPacket *in)
982 out->data = av_realloc(out->data, out->size+in->size);
983 memcpy(out->data+out->size, in->data, in->size);
984 out->size += in->size;
985 av_destruct_packet(in);
989 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
990 AVMetadata **metadata, char *prefix)
992 MatroskaTag *tags = list->elem;
996 for (i=0; i < list->nb_elem; i++) {
997 const char *lang = strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
998 if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
999 else av_strlcpy(key, tags[i].name, sizeof(key));
1000 if (tags[i].def || !lang) {
1001 av_metadata_set(metadata, key, tags[i].string);
1002 if (tags[i].sub.nb_elem)
1003 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1006 av_strlcat(key, "-", sizeof(key));
1007 av_strlcat(key, lang, sizeof(key));
1008 av_metadata_set(metadata, key, tags[i].string);
1009 if (tags[i].sub.nb_elem)
1010 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1015 static void matroska_convert_tags(AVFormatContext *s)
1017 MatroskaDemuxContext *matroska = s->priv_data;
1018 MatroskaTags *tags = matroska->tags.elem;
1021 for (i=0; i < matroska->tags.nb_elem; i++) {
1022 if (tags[i].target.attachuid) {
1023 MatroskaAttachement *attachment = matroska->attachments.elem;
1024 for (j=0; j<matroska->attachments.nb_elem; j++)
1025 if (attachment[j].uid == tags[i].target.attachuid)
1026 matroska_convert_tag(s, &tags[i].tag,
1027 &attachment[j].stream->metadata, NULL);
1028 } else if (tags[i].target.chapteruid) {
1029 MatroskaChapter *chapter = matroska->chapters.elem;
1030 for (j=0; j<matroska->chapters.nb_elem; j++)
1031 if (chapter[j].uid == tags[i].target.chapteruid)
1032 matroska_convert_tag(s, &tags[i].tag,
1033 &chapter[j].chapter->metadata, NULL);
1034 } else if (tags[i].target.trackuid) {
1035 MatroskaTrack *track = matroska->tracks.elem;
1036 for (j=0; j<matroska->tracks.nb_elem; j++)
1037 if (track[j].uid == tags[i].target.trackuid)
1038 matroska_convert_tag(s, &tags[i].tag,
1039 &track[j].stream->metadata, NULL);
1041 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1042 tags[i].target.type);
1047 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1049 EbmlList *seekhead_list = &matroska->seekhead;
1050 MatroskaSeekhead *seekhead = seekhead_list->elem;
1051 uint32_t level_up = matroska->level_up;
1052 int64_t before_pos = url_ftell(matroska->ctx->pb);
1053 MatroskaLevel level;
1056 for (i=0; i<seekhead_list->nb_elem; i++) {
1057 int64_t offset = seekhead[i].pos + matroska->segment_start;
1059 if (seekhead[i].pos <= before_pos
1060 || seekhead[i].id == MATROSKA_ID_SEEKHEAD
1061 || seekhead[i].id == MATROSKA_ID_CLUSTER)
1065 if (url_fseek(matroska->ctx->pb, offset, SEEK_SET) != offset)
1068 /* We don't want to lose our seekhead level, so we add
1069 * a dummy. This is a crude hack. */
1070 if (matroska->num_levels == EBML_MAX_DEPTH) {
1071 av_log(matroska->ctx, AV_LOG_INFO,
1072 "Max EBML element depth (%d) reached, "
1073 "cannot parse further.\n", EBML_MAX_DEPTH);
1078 level.length = (uint64_t)-1;
1079 matroska->levels[matroska->num_levels] = level;
1080 matroska->num_levels++;
1082 ebml_parse(matroska, matroska_segment, matroska);
1084 /* remove dummy level */
1085 while (matroska->num_levels) {
1086 uint64_t length = matroska->levels[--matroska->num_levels].length;
1087 if (length == (uint64_t)-1)
1093 url_fseek(matroska->ctx->pb, before_pos, SEEK_SET);
1094 matroska->level_up = level_up;
1097 static int matroska_aac_profile(char *codec_id)
1099 static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1102 for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1103 if (strstr(codec_id, aac_profiles[profile]))
1108 static int matroska_aac_sri(int samplerate)
1112 for (sri=0; sri<FF_ARRAY_ELEMS(ff_mpeg4audio_sample_rates); sri++)
1113 if (ff_mpeg4audio_sample_rates[sri] == samplerate)
1118 static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
1120 MatroskaDemuxContext *matroska = s->priv_data;
1121 EbmlList *attachements_list = &matroska->attachments;
1122 MatroskaAttachement *attachements;
1123 EbmlList *chapters_list = &matroska->chapters;
1124 MatroskaChapter *chapters;
1125 MatroskaTrack *tracks;
1126 EbmlList *index_list;
1127 MatroskaIndex *index;
1128 int index_scale = 1;
1129 uint64_t max_start = 0;
1136 /* First read the EBML header. */
1137 if (ebml_parse(matroska, ebml_syntax, &ebml)
1138 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1139 || ebml.id_length > sizeof(uint32_t) || strcmp(ebml.doctype, "matroska")
1140 || ebml.doctype_version > 2) {
1141 av_log(matroska->ctx, AV_LOG_ERROR,
1142 "EBML header using unsupported features\n"
1143 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1144 ebml.version, ebml.doctype, ebml.doctype_version);
1145 return AVERROR_NOFMT;
1147 ebml_free(ebml_syntax, &ebml);
1149 /* The next thing is a segment. */
1150 if (ebml_parse(matroska, matroska_segments, matroska) < 0)
1152 matroska_execute_seekhead(matroska);
1154 if (matroska->duration)
1155 matroska->ctx->duration = matroska->duration * matroska->time_scale
1156 * 1000 / AV_TIME_BASE;
1157 av_metadata_set(&s->metadata, "title", matroska->title);
1159 tracks = matroska->tracks.elem;
1160 for (i=0; i < matroska->tracks.nb_elem; i++) {
1161 MatroskaTrack *track = &tracks[i];
1162 enum CodecID codec_id = CODEC_ID_NONE;
1163 EbmlList *encodings_list = &tracks->encodings;
1164 MatroskaTrackEncoding *encodings = encodings_list->elem;
1165 uint8_t *extradata = NULL;
1166 int extradata_size = 0;
1167 int extradata_offset = 0;
1170 /* Apply some sanity checks. */
1171 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1172 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1173 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1174 av_log(matroska->ctx, AV_LOG_INFO,
1175 "Unknown or unsupported track type %"PRIu64"\n",
1179 if (track->codec_id == NULL)
1182 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1183 if (!track->default_duration)
1184 track->default_duration = 1000000000/track->video.frame_rate;
1185 if (!track->video.display_width)
1186 track->video.display_width = track->video.pixel_width;
1187 if (!track->video.display_height)
1188 track->video.display_height = track->video.pixel_height;
1189 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1190 if (!track->audio.out_samplerate)
1191 track->audio.out_samplerate = track->audio.samplerate;
1193 if (encodings_list->nb_elem > 1) {
1194 av_log(matroska->ctx, AV_LOG_ERROR,
1195 "Multiple combined encodings no supported");
1196 } else if (encodings_list->nb_elem == 1) {
1197 if (encodings[0].type ||
1198 (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1200 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1203 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1205 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
1206 encodings[0].scope = 0;
1207 av_log(matroska->ctx, AV_LOG_ERROR,
1208 "Unsupported encoding type");
1209 } else if (track->codec_priv.size && encodings[0].scope&2) {
1210 uint8_t *codec_priv = track->codec_priv.data;
1211 int offset = matroska_decode_buffer(&track->codec_priv.data,
1212 &track->codec_priv.size,
1215 track->codec_priv.data = NULL;
1216 track->codec_priv.size = 0;
1217 av_log(matroska->ctx, AV_LOG_ERROR,
1218 "Failed to decode codec private data\n");
1219 } else if (offset > 0) {
1220 track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
1221 memcpy(track->codec_priv.data,
1222 encodings[0].compression.settings.data, offset);
1223 memcpy(track->codec_priv.data+offset, codec_priv,
1224 track->codec_priv.size);
1225 track->codec_priv.size += offset;
1227 if (codec_priv != track->codec_priv.data)
1228 av_free(codec_priv);
1232 for(j=0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++){
1233 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1234 strlen(ff_mkv_codec_tags[j].str))){
1235 codec_id= ff_mkv_codec_tags[j].id;
1240 st = track->stream = av_new_stream(s, 0);
1242 return AVERROR(ENOMEM);
1244 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1245 && track->codec_priv.size >= 40
1246 && track->codec_priv.data != NULL) {
1247 track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1248 codec_id = ff_codec_get_id(ff_codec_bmp_tags, track->video.fourcc);
1249 extradata_offset = 40;
1250 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1251 && track->codec_priv.size >= 14
1252 && track->codec_priv.data != NULL) {
1253 init_put_byte(&b, track->codec_priv.data, track->codec_priv.size,
1254 URL_RDONLY, NULL, NULL, NULL, NULL);
1255 ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1256 codec_id = st->codec->codec_id;
1257 extradata_offset = FFMIN(track->codec_priv.size, 18);
1258 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1259 && (track->codec_priv.size >= 86)
1260 && (track->codec_priv.data != NULL)) {
1261 track->video.fourcc = AV_RL32(track->codec_priv.data);
1262 codec_id=ff_codec_get_id(codec_movvideo_tags, track->video.fourcc);
1263 } else if (codec_id == CODEC_ID_PCM_S16BE) {
1264 switch (track->audio.bitdepth) {
1265 case 8: codec_id = CODEC_ID_PCM_U8; break;
1266 case 24: codec_id = CODEC_ID_PCM_S24BE; break;
1267 case 32: codec_id = CODEC_ID_PCM_S32BE; break;
1269 } else if (codec_id == CODEC_ID_PCM_S16LE) {
1270 switch (track->audio.bitdepth) {
1271 case 8: codec_id = CODEC_ID_PCM_U8; break;
1272 case 24: codec_id = CODEC_ID_PCM_S24LE; break;
1273 case 32: codec_id = CODEC_ID_PCM_S32LE; break;
1275 } else if (codec_id==CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1276 codec_id = CODEC_ID_PCM_F64LE;
1277 } else if (codec_id == CODEC_ID_AAC && !track->codec_priv.size) {
1278 int profile = matroska_aac_profile(track->codec_id);
1279 int sri = matroska_aac_sri(track->audio.samplerate);
1280 extradata = av_malloc(5);
1281 if (extradata == NULL)
1282 return AVERROR(ENOMEM);
1283 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1284 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1285 if (strstr(track->codec_id, "SBR")) {
1286 sri = matroska_aac_sri(track->audio.out_samplerate);
1287 extradata[2] = 0x56;
1288 extradata[3] = 0xE5;
1289 extradata[4] = 0x80 | (sri<<3);
1293 } else if (codec_id == CODEC_ID_TTA) {
1294 extradata_size = 30;
1295 extradata = av_mallocz(extradata_size);
1296 if (extradata == NULL)
1297 return AVERROR(ENOMEM);
1298 init_put_byte(&b, extradata, extradata_size, 1,
1299 NULL, NULL, NULL, NULL);
1300 put_buffer(&b, "TTA1", 4);
1302 put_le16(&b, track->audio.channels);
1303 put_le16(&b, track->audio.bitdepth);
1304 put_le32(&b, track->audio.out_samplerate);
1305 put_le32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1306 } else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
1307 codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
1308 extradata_offset = 26;
1309 } else if (codec_id == CODEC_ID_RA_144) {
1310 track->audio.out_samplerate = 8000;
1311 track->audio.channels = 1;
1312 } else if (codec_id == CODEC_ID_RA_288 || codec_id == CODEC_ID_COOK ||
1313 codec_id == CODEC_ID_ATRAC3) {
1314 init_put_byte(&b, track->codec_priv.data,track->codec_priv.size,
1315 0, NULL, NULL, NULL, NULL);
1317 track->audio.coded_framesize = get_be32(&b);
1319 track->audio.sub_packet_h = get_be16(&b);
1320 track->audio.frame_size = get_be16(&b);
1321 track->audio.sub_packet_size = get_be16(&b);
1322 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1323 if (codec_id == CODEC_ID_RA_288) {
1324 st->codec->block_align = track->audio.coded_framesize;
1325 track->codec_priv.size = 0;
1327 st->codec->block_align = track->audio.sub_packet_size;
1328 extradata_offset = 78;
1331 track->codec_priv.size -= extradata_offset;
1333 if (codec_id == CODEC_ID_NONE)
1334 av_log(matroska->ctx, AV_LOG_INFO,
1335 "Unknown/unsupported CodecID %s.\n", track->codec_id);
1337 if (track->time_scale < 0.01)
1338 track->time_scale = 1.0;
1339 av_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1341 st->codec->codec_id = codec_id;
1343 if (strcmp(track->language, "und"))
1344 av_metadata_set(&st->metadata, "language", track->language);
1345 av_metadata_set(&st->metadata, "description", track->name);
1347 if (track->flag_default)
1348 st->disposition |= AV_DISPOSITION_DEFAULT;
1350 if (track->default_duration)
1351 av_reduce(&st->codec->time_base.num, &st->codec->time_base.den,
1352 track->default_duration, 1000000000, 30000);
1354 if (!st->codec->extradata) {
1356 st->codec->extradata = extradata;
1357 st->codec->extradata_size = extradata_size;
1358 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1359 st->codec->extradata = av_mallocz(track->codec_priv.size +
1360 FF_INPUT_BUFFER_PADDING_SIZE);
1361 if(st->codec->extradata == NULL)
1362 return AVERROR(ENOMEM);
1363 st->codec->extradata_size = track->codec_priv.size;
1364 memcpy(st->codec->extradata,
1365 track->codec_priv.data + extradata_offset,
1366 track->codec_priv.size);
1370 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1371 st->codec->codec_type = CODEC_TYPE_VIDEO;
1372 st->codec->codec_tag = track->video.fourcc;
1373 st->codec->width = track->video.pixel_width;
1374 st->codec->height = track->video.pixel_height;
1375 av_reduce(&st->sample_aspect_ratio.num,
1376 &st->sample_aspect_ratio.den,
1377 st->codec->height * track->video.display_width,
1378 st->codec-> width * track->video.display_height,
1380 if (st->codec->codec_id != CODEC_ID_H264)
1381 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1382 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1383 st->codec->codec_type = CODEC_TYPE_AUDIO;
1384 st->codec->sample_rate = track->audio.out_samplerate;
1385 st->codec->channels = track->audio.channels;
1386 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1387 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
1391 attachements = attachements_list->elem;
1392 for (j=0; j<attachements_list->nb_elem; j++) {
1393 if (!(attachements[j].filename && attachements[j].mime &&
1394 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1395 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1397 AVStream *st = av_new_stream(s, 0);
1400 av_metadata_set(&st->metadata, "filename",attachements[j].filename);
1401 st->codec->codec_id = CODEC_ID_NONE;
1402 st->codec->codec_type = CODEC_TYPE_ATTACHMENT;
1403 st->codec->extradata = av_malloc(attachements[j].bin.size);
1404 if(st->codec->extradata == NULL)
1406 st->codec->extradata_size = attachements[j].bin.size;
1407 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1409 for (i=0; ff_mkv_mime_tags[i].id != CODEC_ID_NONE; i++) {
1410 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1411 strlen(ff_mkv_mime_tags[i].str))) {
1412 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1416 attachements[j].stream = st;
1420 chapters = chapters_list->elem;
1421 for (i=0; i<chapters_list->nb_elem; i++)
1422 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1423 && (max_start==0 || chapters[i].start > max_start)) {
1424 chapters[i].chapter =
1425 ff_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1426 chapters[i].start, chapters[i].end,
1428 av_metadata_set(&chapters[i].chapter->metadata,
1429 "title", chapters[i].title);
1430 max_start = chapters[i].start;
1433 index_list = &matroska->index;
1434 index = index_list->elem;
1435 if (index_list->nb_elem
1436 && index[0].time > 100000000000000/matroska->time_scale) {
1437 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1438 index_scale = matroska->time_scale;
1440 for (i=0; i<index_list->nb_elem; i++) {
1441 EbmlList *pos_list = &index[i].pos;
1442 MatroskaIndexPos *pos = pos_list->elem;
1443 for (j=0; j<pos_list->nb_elem; j++) {
1444 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1446 if (track && track->stream)
1447 av_add_index_entry(track->stream,
1448 pos[j].pos + matroska->segment_start,
1449 index[i].time/index_scale, 0, 0,
1454 matroska_convert_tags(s);
1460 * Put one packet in an application-supplied AVPacket struct.
1461 * Returns 0 on success or -1 on failure.
1463 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1466 if (matroska->num_packets > 0) {
1467 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1468 av_free(matroska->packets[0]);
1469 if (matroska->num_packets > 1) {
1470 memmove(&matroska->packets[0], &matroska->packets[1],
1471 (matroska->num_packets - 1) * sizeof(AVPacket *));
1473 av_realloc(matroska->packets, (matroska->num_packets - 1) *
1474 sizeof(AVPacket *));
1476 av_freep(&matroska->packets);
1478 matroska->num_packets--;
1486 * Free all packets in our internal queue.
1488 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1490 if (matroska->packets) {
1492 for (n = 0; n < matroska->num_packets; n++) {
1493 av_free_packet(matroska->packets[n]);
1494 av_free(matroska->packets[n]);
1496 av_freep(&matroska->packets);
1497 matroska->num_packets = 0;
1501 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
1502 int size, int64_t pos, uint64_t cluster_time,
1503 uint64_t duration, int is_keyframe,
1504 int64_t cluster_pos)
1506 uint64_t timecode = AV_NOPTS_VALUE;
1507 MatroskaTrack *track;
1512 uint32_t *lace_size = NULL;
1513 int n, flags, laces = 0;
1516 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
1517 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
1523 track = matroska_find_track_by_num(matroska, num);
1524 if (size <= 3 || !track || !track->stream) {
1525 av_log(matroska->ctx, AV_LOG_INFO,
1526 "Invalid stream %"PRIu64" or size %u\n", num, size);
1530 if (st->discard >= AVDISCARD_ALL)
1532 if (duration == AV_NOPTS_VALUE)
1533 duration = track->default_duration / matroska->time_scale;
1535 block_time = AV_RB16(data);
1539 if (is_keyframe == -1)
1540 is_keyframe = flags & 0x80 ? PKT_FLAG_KEY : 0;
1542 if (cluster_time != (uint64_t)-1
1543 && (block_time >= 0 || cluster_time >= -block_time)) {
1544 timecode = cluster_time + block_time;
1545 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
1546 && timecode < track->end_timecode)
1547 is_keyframe = 0; /* overlapping subtitles are not key frame */
1549 av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
1550 track->end_timecode = FFMAX(track->end_timecode, timecode+duration);
1553 if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1554 if (!is_keyframe || timecode < matroska->skip_to_timecode)
1556 matroska->skip_to_keyframe = 0;
1559 switch ((flags & 0x06) >> 1) {
1560 case 0x0: /* no lacing */
1562 lace_size = av_mallocz(sizeof(int));
1563 lace_size[0] = size;
1566 case 0x1: /* Xiph lacing */
1567 case 0x2: /* fixed-size lacing */
1568 case 0x3: /* EBML lacing */
1569 assert(size>0); // size <=3 is checked before size-=3 above
1570 laces = (*data) + 1;
1573 lace_size = av_mallocz(laces * sizeof(int));
1575 switch ((flags & 0x06) >> 1) {
1576 case 0x1: /* Xiph lacing */ {
1579 for (n = 0; res == 0 && n < laces - 1; n++) {
1586 lace_size[n] += temp;
1592 total += lace_size[n];
1594 lace_size[n] = size - total;
1598 case 0x2: /* fixed-size lacing */
1599 for (n = 0; n < laces; n++)
1600 lace_size[n] = size / laces;
1603 case 0x3: /* EBML lacing */ {
1605 n = matroska_ebmlnum_uint(matroska, data, size, &num);
1607 av_log(matroska->ctx, AV_LOG_INFO,
1608 "EBML block data error\n");
1613 total = lace_size[0] = num;
1614 for (n = 1; res == 0 && n < laces - 1; n++) {
1617 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1619 av_log(matroska->ctx, AV_LOG_INFO,
1620 "EBML block data error\n");
1625 lace_size[n] = lace_size[n - 1] + snum;
1626 total += lace_size[n];
1628 lace_size[n] = size - total;
1636 for (n = 0; n < laces; n++) {
1637 if ((st->codec->codec_id == CODEC_ID_RA_288 ||
1638 st->codec->codec_id == CODEC_ID_COOK ||
1639 st->codec->codec_id == CODEC_ID_ATRAC3) &&
1640 st->codec->block_align && track->audio.sub_packet_size) {
1641 int a = st->codec->block_align;
1642 int sps = track->audio.sub_packet_size;
1643 int cfs = track->audio.coded_framesize;
1644 int h = track->audio.sub_packet_h;
1645 int y = track->audio.sub_packet_cnt;
1646 int w = track->audio.frame_size;
1649 if (!track->audio.pkt_cnt) {
1650 if (st->codec->codec_id == CODEC_ID_RA_288)
1651 for (x=0; x<h/2; x++)
1652 memcpy(track->audio.buf+x*2*w+y*cfs,
1655 for (x=0; x<w/sps; x++)
1656 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
1658 if (++track->audio.sub_packet_cnt >= h) {
1659 track->audio.sub_packet_cnt = 0;
1660 track->audio.pkt_cnt = h*w / a;
1663 while (track->audio.pkt_cnt) {
1664 pkt = av_mallocz(sizeof(AVPacket));
1665 av_new_packet(pkt, a);
1666 memcpy(pkt->data, track->audio.buf
1667 + a * (h*w / a - track->audio.pkt_cnt--), a);
1669 pkt->stream_index = st->index;
1670 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1673 MatroskaTrackEncoding *encodings = track->encodings.elem;
1674 int offset = 0, pkt_size = lace_size[n];
1675 uint8_t *pkt_data = data;
1677 if (encodings && encodings->scope & 1) {
1678 offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
1683 pkt = av_mallocz(sizeof(AVPacket));
1684 /* XXX: prevent data copy... */
1685 if (av_new_packet(pkt, pkt_size+offset) < 0) {
1687 res = AVERROR(ENOMEM);
1691 memcpy (pkt->data, encodings->compression.settings.data, offset);
1692 memcpy (pkt->data+offset, pkt_data, pkt_size);
1694 if (pkt_data != data)
1698 pkt->flags = is_keyframe;
1699 pkt->stream_index = st->index;
1701 pkt->pts = timecode;
1703 if (st->codec->codec_id == CODEC_ID_TEXT)
1704 pkt->convergence_duration = duration;
1705 else if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE)
1706 pkt->duration = duration;
1708 if (st->codec->codec_id == CODEC_ID_SSA)
1709 matroska_fix_ass_packet(matroska, pkt, duration);
1711 if (matroska->prev_pkt &&
1712 timecode != AV_NOPTS_VALUE &&
1713 matroska->prev_pkt->pts == timecode &&
1714 matroska->prev_pkt->stream_index == st->index)
1715 matroska_merge_packets(matroska->prev_pkt, pkt);
1717 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1718 matroska->prev_pkt = pkt;
1722 if (timecode != AV_NOPTS_VALUE)
1723 timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
1724 data += lace_size[n];
1732 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
1734 MatroskaCluster cluster = { 0 };
1735 EbmlList *blocks_list;
1736 MatroskaBlock *blocks;
1738 int64_t pos = url_ftell(matroska->ctx->pb);
1739 matroska->prev_pkt = NULL;
1740 if (matroska->has_cluster_id){
1741 /* For the first cluster we parse, its ID was already read as
1742 part of matroska_read_header(), so don't read it again */
1743 res = ebml_parse_id(matroska, matroska_clusters,
1744 MATROSKA_ID_CLUSTER, &cluster);
1745 pos -= 4; /* sizeof the ID which was already read */
1746 matroska->has_cluster_id = 0;
1748 res = ebml_parse(matroska, matroska_clusters, &cluster);
1749 blocks_list = &cluster.blocks;
1750 blocks = blocks_list->elem;
1751 for (i=0; i<blocks_list->nb_elem; i++)
1752 if (blocks[i].bin.size > 0) {
1753 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
1754 res=matroska_parse_block(matroska,
1755 blocks[i].bin.data, blocks[i].bin.size,
1756 blocks[i].bin.pos, cluster.timecode,
1757 blocks[i].duration, is_keyframe,
1760 ebml_free(matroska_cluster, &cluster);
1761 if (res < 0) matroska->done = 1;
1765 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
1767 MatroskaDemuxContext *matroska = s->priv_data;
1769 while (matroska_deliver_packet(matroska, pkt)) {
1772 matroska_parse_cluster(matroska);
1778 static int matroska_read_seek(AVFormatContext *s, int stream_index,
1779 int64_t timestamp, int flags)
1781 MatroskaDemuxContext *matroska = s->priv_data;
1782 MatroskaTrack *tracks = matroska->tracks.elem;
1783 AVStream *st = s->streams[stream_index];
1784 int i, index, index_sub, index_min;
1786 if (!st->nb_index_entries)
1788 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
1790 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1791 url_fseek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
1792 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1793 matroska_clear_queue(matroska);
1794 if (matroska_parse_cluster(matroska) < 0)
1799 matroska_clear_queue(matroska);
1804 for (i=0; i < matroska->tracks.nb_elem; i++) {
1805 tracks[i].end_timecode = 0;
1806 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
1807 && !tracks[i].stream->discard != AVDISCARD_ALL) {
1808 index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
1810 && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
1811 && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
1812 index_min = index_sub;
1816 url_fseek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
1817 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
1818 matroska->skip_to_timecode = st->index_entries[index].timestamp;
1820 av_update_cur_dts(s, st, st->index_entries[index].timestamp);
1824 static int matroska_read_close(AVFormatContext *s)
1826 MatroskaDemuxContext *matroska = s->priv_data;
1827 MatroskaTrack *tracks = matroska->tracks.elem;
1830 matroska_clear_queue(matroska);
1832 for (n=0; n < matroska->tracks.nb_elem; n++)
1833 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
1834 av_free(tracks[n].audio.buf);
1835 ebml_free(matroska_segment, matroska);
1840 AVInputFormat matroska_demuxer = {
1842 NULL_IF_CONFIG_SMALL("Matroska file format"),
1843 sizeof(MatroskaDemuxContext),
1845 matroska_read_header,
1846 matroska_read_packet,
1847 matroska_read_close,
1849 .metadata_conv = ff_mkv_metadata_conv,