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(). */
38 #include "libavcodec/mpeg4audio.h"
39 #include "libavutil/intfloat_readwrite.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavutil/avstring.h"
42 #include "libavutil/lzo.h"
62 typedef const struct EbmlSyntax {
71 const struct EbmlSyntax *n;
91 uint64_t doctype_version;
97 } MatroskaTrackCompression;
102 MatroskaTrackCompression compression;
103 } MatroskaTrackEncoding;
107 uint64_t display_width;
108 uint64_t display_height;
109 uint64_t pixel_width;
110 uint64_t pixel_height;
112 } MatroskaTrackVideo;
116 double out_samplerate;
120 /* real audio header (extracted from extradata) */
128 } MatroskaTrackAudio;
139 uint64_t default_duration;
140 uint64_t flag_default;
141 MatroskaTrackVideo video;
142 MatroskaTrackAudio audio;
146 int64_t end_timecode;
157 } MatroskaAttachement;
195 MatroskaTagTarget target;
210 AVFormatContext *ctx;
214 MatroskaLevel levels[EBML_MAX_DEPTH];
221 EbmlList attachments;
227 /* byte position of the segment inside the stream */
228 int64_t segment_start;
230 /* the packet queue */
238 /* What to skip before effectively reading a packet. */
239 int skip_to_keyframe;
240 uint64_t skip_to_timecode;
241 } MatroskaDemuxContext;
255 static EbmlSyntax ebml_header[] = {
256 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
257 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
258 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
259 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
260 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
261 { EBML_ID_EBMLVERSION, EBML_NONE },
262 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
266 static EbmlSyntax ebml_syntax[] = {
267 { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
271 static EbmlSyntax matroska_info[] = {
272 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
273 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
274 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
275 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
276 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
277 { MATROSKA_ID_DATEUTC, EBML_NONE },
278 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
282 static EbmlSyntax matroska_track_video[] = {
283 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
284 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
285 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
286 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
287 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
288 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
289 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
290 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
291 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
292 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
293 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
294 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
295 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_NONE },
296 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
300 static EbmlSyntax matroska_track_audio[] = {
301 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
302 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
303 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
304 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
308 static EbmlSyntax matroska_track_encoding_compression[] = {
309 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
310 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
314 static EbmlSyntax matroska_track_encoding[] = {
315 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
316 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
317 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
318 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
322 static EbmlSyntax matroska_track_encodings[] = {
323 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
327 static EbmlSyntax matroska_track[] = {
328 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
329 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack,name) },
330 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
331 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
332 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
333 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
334 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
335 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
336 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
337 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
338 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
339 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
340 { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
341 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
342 { MATROSKA_ID_TRACKFLAGFORCED, EBML_NONE },
343 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
344 { MATROSKA_ID_CODECNAME, EBML_NONE },
345 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
346 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
347 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
348 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
349 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
350 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
354 static EbmlSyntax matroska_tracks[] = {
355 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
359 static EbmlSyntax matroska_attachment[] = {
360 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
361 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
362 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
363 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
364 { MATROSKA_ID_FILEDESC, EBML_NONE },
368 static EbmlSyntax matroska_attachments[] = {
369 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
373 static EbmlSyntax matroska_chapter_display[] = {
374 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
375 { MATROSKA_ID_CHAPLANG, EBML_NONE },
379 static EbmlSyntax matroska_chapter_entry[] = {
380 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
381 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
382 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
383 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
384 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
385 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
386 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
387 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
391 static EbmlSyntax matroska_chapter[] = {
392 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
393 { MATROSKA_ID_EDITIONUID, EBML_NONE },
394 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
395 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
396 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
400 static EbmlSyntax matroska_chapters[] = {
401 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
405 static EbmlSyntax matroska_index_pos[] = {
406 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
407 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
408 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
412 static EbmlSyntax matroska_index_entry[] = {
413 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
414 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
418 static EbmlSyntax matroska_index[] = {
419 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
423 static EbmlSyntax matroska_simpletag[] = {
424 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
425 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
426 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag,lang), {.s="und"} },
427 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag,def) },
428 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
432 static EbmlSyntax matroska_tagtargets[] = {
433 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget,type) },
434 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
435 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
436 { MATROSKA_ID_TAGTARGETS_CHAPTERUID,EBML_UINT, 0, offsetof(MatroskaTagTarget,chapteruid) },
437 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
441 static EbmlSyntax matroska_tag[] = {
442 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
443 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
447 static EbmlSyntax matroska_tags[] = {
448 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
452 static EbmlSyntax matroska_seekhead_entry[] = {
453 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
454 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
458 static EbmlSyntax matroska_seekhead[] = {
459 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
463 static EbmlSyntax matroska_segment[] = {
464 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
465 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
466 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
467 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
468 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
469 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
470 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
471 { MATROSKA_ID_CLUSTER, EBML_STOP, 0, offsetof(MatroskaDemuxContext,has_cluster_id) },
475 static EbmlSyntax matroska_segments[] = {
476 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
480 static EbmlSyntax matroska_blockgroup[] = {
481 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
482 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
483 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration), {.u=AV_NOPTS_VALUE} },
484 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
485 { 1, EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
489 static EbmlSyntax matroska_cluster[] = {
490 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
491 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
492 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
493 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
494 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
498 static EbmlSyntax matroska_clusters[] = {
499 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
500 { MATROSKA_ID_INFO, EBML_NONE },
501 { MATROSKA_ID_CUES, EBML_NONE },
502 { MATROSKA_ID_TAGS, EBML_NONE },
503 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
508 * Return: Whether we reached the end of a level in the hierarchy or not.
510 static int ebml_level_end(MatroskaDemuxContext *matroska)
512 ByteIOContext *pb = matroska->ctx->pb;
513 int64_t pos = url_ftell(pb);
515 if (matroska->num_levels > 0) {
516 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
517 if (pos - level->start >= level->length) {
518 matroska->num_levels--;
526 * Read: an "EBML number", which is defined as a variable-length
527 * array of bytes. The first byte indicates the length by giving a
528 * number of 0-bits followed by a one. The position of the first
529 * "one" bit inside the first byte indicates the length of this
531 * Returns: number of bytes read, < 0 on error
533 static int ebml_read_num(MatroskaDemuxContext *matroska, ByteIOContext *pb,
534 int max_size, uint64_t *number)
536 int len_mask = 0x80, read = 1, n = 1;
539 /* The first byte tells us the length in bytes - get_byte() can normally
540 * return 0, but since that's not a valid first ebmlID byte, we can
541 * use it safely here to catch EOS. */
542 if (!(total = get_byte(pb))) {
543 /* we might encounter EOS here */
545 int64_t pos = url_ftell(pb);
546 av_log(matroska->ctx, AV_LOG_ERROR,
547 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
550 return AVERROR(EIO); /* EOS or actual I/O error */
553 /* get the length of the EBML number */
554 while (read <= max_size && !(total & len_mask)) {
558 if (read > max_size) {
559 int64_t pos = url_ftell(pb) - 1;
560 av_log(matroska->ctx, AV_LOG_ERROR,
561 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
562 (uint8_t) total, pos, pos);
563 return AVERROR_INVALIDDATA;
566 /* read out length */
569 total = (total << 8) | get_byte(pb);
577 * Read the next element as an unsigned int.
578 * 0 is success, < 0 is failure.
580 static int ebml_read_uint(ByteIOContext *pb, int size, uint64_t *num)
584 if (size < 1 || size > 8)
585 return AVERROR_INVALIDDATA;
587 /* big-endian ordering; build up number */
590 *num = (*num << 8) | get_byte(pb);
596 * Read the next element as a float.
597 * 0 is success, < 0 is failure.
599 static int ebml_read_float(ByteIOContext *pb, int size, double *num)
602 *num= av_int2flt(get_be32(pb));
604 *num= av_int2dbl(get_be64(pb));
606 return AVERROR_INVALIDDATA;
612 * Read the next element as an ASCII string.
613 * 0 is success, < 0 is failure.
615 static int ebml_read_ascii(ByteIOContext *pb, int size, char **str)
618 /* EBML strings are usually not 0-terminated, so we allocate one
619 * byte more, read the string and NULL-terminate it ourselves. */
620 if (!(*str = av_malloc(size + 1)))
621 return AVERROR(ENOMEM);
622 if (get_buffer(pb, (uint8_t *) *str, size) != size) {
632 * Read the next element as binary data.
633 * 0 is success, < 0 is failure.
635 static int ebml_read_binary(ByteIOContext *pb, int length, EbmlBin *bin)
638 if (!(bin->data = av_malloc(length)))
639 return AVERROR(ENOMEM);
642 bin->pos = url_ftell(pb);
643 if (get_buffer(pb, bin->data, length) != length)
650 * Read the next element, but only the header. The contents
651 * are supposed to be sub-elements which can be read separately.
652 * 0 is success, < 0 is failure.
654 static int ebml_read_master(MatroskaDemuxContext *matroska, int length)
656 ByteIOContext *pb = matroska->ctx->pb;
657 MatroskaLevel *level;
659 if (matroska->num_levels >= EBML_MAX_DEPTH) {
660 av_log(matroska->ctx, AV_LOG_ERROR,
661 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
662 return AVERROR(ENOSYS);
665 level = &matroska->levels[matroska->num_levels++];
666 level->start = url_ftell(pb);
667 level->length = length;
673 * Read signed/unsigned "EBML" numbers.
674 * Return: number of bytes processed, < 0 on error
676 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
677 uint8_t *data, uint32_t size, uint64_t *num)
680 init_put_byte(&pb, data, size, 0, NULL, NULL, NULL, NULL);
681 return ebml_read_num(matroska, &pb, 8, num);
685 * Same as above, but signed.
687 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
688 uint8_t *data, uint32_t size, int64_t *num)
693 /* read as unsigned number first */
694 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
697 /* make signed (weird way) */
698 *num = unum - ((1LL << (7*res - 1)) - 1);
703 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
704 EbmlSyntax *syntax, void *data);
706 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
707 uint32_t id, void *data)
710 for (i=0; syntax[i].id; i++)
711 if (id == syntax[i].id)
713 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32)
714 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
715 return ebml_parse_elem(matroska, &syntax[i], data);
718 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
722 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
724 return res < 0 ? res : ebml_parse_id(matroska, syntax, id, data);
727 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
732 for (i=0; syntax[i].id; i++)
733 switch (syntax[i].type) {
735 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
738 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
742 *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
746 while (!res && !ebml_level_end(matroska))
747 res = ebml_parse(matroska, syntax, data);
752 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
753 EbmlSyntax *syntax, void *data)
755 ByteIOContext *pb = matroska->ctx->pb;
756 uint32_t id = syntax->id;
760 data = (char *)data + syntax->data_offset;
761 if (syntax->list_elem_size) {
762 EbmlList *list = data;
763 list->elem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
764 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
765 memset(data, 0, syntax->list_elem_size);
769 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP)
770 if ((res = ebml_read_num(matroska, pb, 8, &length)) < 0)
773 switch (syntax->type) {
774 case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
775 case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
777 case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
778 case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
779 case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
781 if (id == MATROSKA_ID_SEGMENT)
782 matroska->segment_start = url_ftell(matroska->ctx->pb);
783 return ebml_parse_nest(matroska, syntax->def.n, data);
784 case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
785 case EBML_STOP: *(int *)data = 1; return 1;
786 default: return url_fseek(pb,length,SEEK_CUR)<0 ? AVERROR(EIO) : 0;
788 if (res == AVERROR_INVALIDDATA)
789 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
790 else if (res == AVERROR(EIO))
791 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
795 static void ebml_free(EbmlSyntax *syntax, void *data)
798 for (i=0; syntax[i].id; i++) {
799 void *data_off = (char *)data + syntax[i].data_offset;
800 switch (syntax[i].type) {
802 case EBML_UTF8: av_freep(data_off); break;
803 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
805 if (syntax[i].list_elem_size) {
806 EbmlList *list = data_off;
807 char *ptr = list->elem;
808 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
809 ebml_free(syntax[i].def.n, ptr);
812 ebml_free(syntax[i].def.n, data_off);
822 static int matroska_probe(AVProbeData *p)
825 int len_mask = 0x80, size = 1, n = 1;
826 static const char probe_data[] = "matroska";
829 if (AV_RB32(p->buf) != EBML_ID_HEADER)
832 /* length of header */
834 while (size <= 8 && !(total & len_mask)) {
840 total &= (len_mask - 1);
842 total = (total << 8) | p->buf[4 + n++];
844 /* Does the probe data contain the whole header? */
845 if (p->buf_size < 4 + size + total)
848 /* The header must contain the document type 'matroska'. For now,
849 * we don't parse the whole header but simply check for the
850 * availability of that array of characters inside the header.
851 * Not fully fool-proof, but good enough. */
852 for (n = 4+size; n <= 4+size+total-(sizeof(probe_data)-1); n++)
853 if (!memcmp(p->buf+n, probe_data, sizeof(probe_data)-1))
854 return AVPROBE_SCORE_MAX;
859 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
862 MatroskaTrack *tracks = matroska->tracks.elem;
865 for (i=0; i < matroska->tracks.nb_elem; i++)
866 if (tracks[i].num == num)
869 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
873 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
874 MatroskaTrack *track)
876 MatroskaTrackEncoding *encodings = track->encodings.elem;
877 uint8_t* data = *buf;
878 int isize = *buf_size;
879 uint8_t* pkt_data = NULL;
880 int pkt_size = isize;
884 switch (encodings[0].compression.algo) {
885 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
886 return encodings[0].compression.settings.size;
887 case MATROSKA_TRACK_ENCODING_COMP_LZO:
889 olen = pkt_size *= 3;
890 pkt_data = av_realloc(pkt_data, pkt_size+AV_LZO_OUTPUT_PADDING);
891 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
892 } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
898 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
899 z_stream zstream = {0};
900 if (inflateInit(&zstream) != Z_OK)
902 zstream.next_in = data;
903 zstream.avail_in = isize;
906 pkt_data = av_realloc(pkt_data, pkt_size);
907 zstream.avail_out = pkt_size - zstream.total_out;
908 zstream.next_out = pkt_data + zstream.total_out;
909 result = inflate(&zstream, Z_NO_FLUSH);
910 } while (result==Z_OK && pkt_size<10000000);
911 pkt_size = zstream.total_out;
912 inflateEnd(&zstream);
913 if (result != Z_STREAM_END)
919 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
920 bz_stream bzstream = {0};
921 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
923 bzstream.next_in = data;
924 bzstream.avail_in = isize;
927 pkt_data = av_realloc(pkt_data, pkt_size);
928 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
929 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
930 result = BZ2_bzDecompress(&bzstream);
931 } while (result==BZ_OK && pkt_size<10000000);
932 pkt_size = bzstream.total_out_lo32;
933 BZ2_bzDecompressEnd(&bzstream);
934 if (result != BZ_STREAM_END)
944 *buf_size = pkt_size;
951 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
952 AVPacket *pkt, uint64_t display_duration)
954 char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
955 for (; *ptr!=',' && ptr<end-1; ptr++);
958 for (; *ptr!=',' && ptr<end-1; ptr++);
960 int64_t end_pts = pkt->pts + display_duration;
961 int sc = matroska->time_scale * pkt->pts / 10000000;
962 int ec = matroska->time_scale * end_pts / 10000000;
963 int sh, sm, ss, eh, em, es, len;
964 sh = sc/360000; sc -= 360000*sh;
965 sm = sc/ 6000; sc -= 6000*sm;
966 ss = sc/ 100; sc -= 100*ss;
967 eh = ec/360000; ec -= 360000*eh;
968 em = ec/ 6000; ec -= 6000*em;
969 es = ec/ 100; ec -= 100*es;
971 len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
972 if (!(line = av_malloc(len)))
974 snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
975 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
978 pkt->size = strlen(line);
982 static void matroska_merge_packets(AVPacket *out, AVPacket *in)
984 out->data = av_realloc(out->data, out->size+in->size);
985 memcpy(out->data+out->size, in->data, in->size);
986 out->size += in->size;
987 av_destruct_packet(in);
991 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
992 AVMetadata **metadata, char *prefix)
994 MatroskaTag *tags = list->elem;
998 for (i=0; i < list->nb_elem; i++) {
999 const char *lang = strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1000 if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1001 else av_strlcpy(key, tags[i].name, sizeof(key));
1002 if (tags[i].def || !lang) {
1003 av_metadata_set(metadata, key, tags[i].string);
1004 if (tags[i].sub.nb_elem)
1005 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1008 av_strlcat(key, "-", sizeof(key));
1009 av_strlcat(key, lang, sizeof(key));
1010 av_metadata_set(metadata, key, tags[i].string);
1011 if (tags[i].sub.nb_elem)
1012 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1017 static void matroska_convert_tags(AVFormatContext *s)
1019 MatroskaDemuxContext *matroska = s->priv_data;
1020 MatroskaTags *tags = matroska->tags.elem;
1023 for (i=0; i < matroska->tags.nb_elem; i++) {
1024 if (tags[i].target.attachuid) {
1025 MatroskaAttachement *attachment = matroska->attachments.elem;
1026 for (j=0; j<matroska->attachments.nb_elem; j++)
1027 if (attachment[j].uid == tags[i].target.attachuid)
1028 matroska_convert_tag(s, &tags[i].tag,
1029 &attachment[j].stream->metadata, NULL);
1030 } else if (tags[i].target.chapteruid) {
1031 MatroskaChapter *chapter = matroska->chapters.elem;
1032 for (j=0; j<matroska->chapters.nb_elem; j++)
1033 if (chapter[j].uid == tags[i].target.chapteruid)
1034 matroska_convert_tag(s, &tags[i].tag,
1035 &chapter[j].chapter->metadata, NULL);
1036 } else if (tags[i].target.trackuid) {
1037 MatroskaTrack *track = matroska->tracks.elem;
1038 for (j=0; j<matroska->tracks.nb_elem; j++)
1039 if (track[j].uid == tags[i].target.trackuid)
1040 matroska_convert_tag(s, &tags[i].tag,
1041 &track[j].stream->metadata, NULL);
1043 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1044 tags[i].target.type);
1049 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1051 EbmlList *seekhead_list = &matroska->seekhead;
1052 MatroskaSeekhead *seekhead = seekhead_list->elem;
1053 uint32_t level_up = matroska->level_up;
1054 int64_t before_pos = url_ftell(matroska->ctx->pb);
1055 MatroskaLevel level;
1058 for (i=0; i<seekhead_list->nb_elem; i++) {
1059 int64_t offset = seekhead[i].pos + matroska->segment_start;
1061 if (seekhead[i].pos <= before_pos
1062 || seekhead[i].id == MATROSKA_ID_SEEKHEAD
1063 || seekhead[i].id == MATROSKA_ID_CLUSTER)
1067 if (url_fseek(matroska->ctx->pb, offset, SEEK_SET) != offset)
1070 /* We don't want to lose our seekhead level, so we add
1071 * a dummy. This is a crude hack. */
1072 if (matroska->num_levels == EBML_MAX_DEPTH) {
1073 av_log(matroska->ctx, AV_LOG_INFO,
1074 "Max EBML element depth (%d) reached, "
1075 "cannot parse further.\n", EBML_MAX_DEPTH);
1080 level.length = (uint64_t)-1;
1081 matroska->levels[matroska->num_levels] = level;
1082 matroska->num_levels++;
1084 ebml_parse(matroska, matroska_segment, matroska);
1086 /* remove dummy level */
1087 while (matroska->num_levels) {
1088 uint64_t length = matroska->levels[--matroska->num_levels].length;
1089 if (length == (uint64_t)-1)
1095 url_fseek(matroska->ctx->pb, before_pos, SEEK_SET);
1096 matroska->level_up = level_up;
1099 static int matroska_aac_profile(char *codec_id)
1101 static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1104 for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1105 if (strstr(codec_id, aac_profiles[profile]))
1110 static int matroska_aac_sri(int samplerate)
1114 for (sri=0; sri<FF_ARRAY_ELEMS(ff_mpeg4audio_sample_rates); sri++)
1115 if (ff_mpeg4audio_sample_rates[sri] == samplerate)
1120 static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
1122 MatroskaDemuxContext *matroska = s->priv_data;
1123 EbmlList *attachements_list = &matroska->attachments;
1124 MatroskaAttachement *attachements;
1125 EbmlList *chapters_list = &matroska->chapters;
1126 MatroskaChapter *chapters;
1127 MatroskaTrack *tracks;
1128 EbmlList *index_list;
1129 MatroskaIndex *index;
1130 int index_scale = 1;
1131 uint64_t max_start = 0;
1138 /* First read the EBML header. */
1139 if (ebml_parse(matroska, ebml_syntax, &ebml)
1140 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1141 || ebml.id_length > sizeof(uint32_t) || strcmp(ebml.doctype, "matroska")
1142 || ebml.doctype_version > 2) {
1143 av_log(matroska->ctx, AV_LOG_ERROR,
1144 "EBML header using unsupported features\n"
1145 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1146 ebml.version, ebml.doctype, ebml.doctype_version);
1147 return AVERROR_PATCHWELCOME;
1149 ebml_free(ebml_syntax, &ebml);
1151 /* The next thing is a segment. */
1152 if (ebml_parse(matroska, matroska_segments, matroska) < 0)
1154 matroska_execute_seekhead(matroska);
1156 if (matroska->duration)
1157 matroska->ctx->duration = matroska->duration * matroska->time_scale
1158 * 1000 / AV_TIME_BASE;
1159 av_metadata_set(&s->metadata, "title", matroska->title);
1161 tracks = matroska->tracks.elem;
1162 for (i=0; i < matroska->tracks.nb_elem; i++) {
1163 MatroskaTrack *track = &tracks[i];
1164 enum CodecID codec_id = CODEC_ID_NONE;
1165 EbmlList *encodings_list = &tracks->encodings;
1166 MatroskaTrackEncoding *encodings = encodings_list->elem;
1167 uint8_t *extradata = NULL;
1168 int extradata_size = 0;
1169 int extradata_offset = 0;
1172 /* Apply some sanity checks. */
1173 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1174 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1175 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1176 av_log(matroska->ctx, AV_LOG_INFO,
1177 "Unknown or unsupported track type %"PRIu64"\n",
1181 if (track->codec_id == NULL)
1184 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1185 if (!track->default_duration)
1186 track->default_duration = 1000000000/track->video.frame_rate;
1187 if (!track->video.display_width)
1188 track->video.display_width = track->video.pixel_width;
1189 if (!track->video.display_height)
1190 track->video.display_height = track->video.pixel_height;
1191 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1192 if (!track->audio.out_samplerate)
1193 track->audio.out_samplerate = track->audio.samplerate;
1195 if (encodings_list->nb_elem > 1) {
1196 av_log(matroska->ctx, AV_LOG_ERROR,
1197 "Multiple combined encodings no supported");
1198 } else if (encodings_list->nb_elem == 1) {
1199 if (encodings[0].type ||
1200 (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1202 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1205 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1207 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
1208 encodings[0].scope = 0;
1209 av_log(matroska->ctx, AV_LOG_ERROR,
1210 "Unsupported encoding type");
1211 } else if (track->codec_priv.size && encodings[0].scope&2) {
1212 uint8_t *codec_priv = track->codec_priv.data;
1213 int offset = matroska_decode_buffer(&track->codec_priv.data,
1214 &track->codec_priv.size,
1217 track->codec_priv.data = NULL;
1218 track->codec_priv.size = 0;
1219 av_log(matroska->ctx, AV_LOG_ERROR,
1220 "Failed to decode codec private data\n");
1221 } else if (offset > 0) {
1222 track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
1223 memcpy(track->codec_priv.data,
1224 encodings[0].compression.settings.data, offset);
1225 memcpy(track->codec_priv.data+offset, codec_priv,
1226 track->codec_priv.size);
1227 track->codec_priv.size += offset;
1229 if (codec_priv != track->codec_priv.data)
1230 av_free(codec_priv);
1234 for(j=0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++){
1235 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1236 strlen(ff_mkv_codec_tags[j].str))){
1237 codec_id= ff_mkv_codec_tags[j].id;
1242 st = track->stream = av_new_stream(s, 0);
1244 return AVERROR(ENOMEM);
1246 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1247 && track->codec_priv.size >= 40
1248 && track->codec_priv.data != NULL) {
1249 track->ms_compat = 1;
1250 track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1251 codec_id = ff_codec_get_id(ff_codec_bmp_tags, track->video.fourcc);
1252 extradata_offset = 40;
1253 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1254 && track->codec_priv.size >= 14
1255 && track->codec_priv.data != NULL) {
1256 init_put_byte(&b, track->codec_priv.data, track->codec_priv.size,
1257 URL_RDONLY, NULL, NULL, NULL, NULL);
1258 ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1259 codec_id = st->codec->codec_id;
1260 extradata_offset = FFMIN(track->codec_priv.size, 18);
1261 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1262 && (track->codec_priv.size >= 86)
1263 && (track->codec_priv.data != NULL)) {
1264 track->video.fourcc = AV_RL32(track->codec_priv.data);
1265 codec_id=ff_codec_get_id(codec_movvideo_tags, track->video.fourcc);
1266 } else if (codec_id == CODEC_ID_PCM_S16BE) {
1267 switch (track->audio.bitdepth) {
1268 case 8: codec_id = CODEC_ID_PCM_U8; break;
1269 case 24: codec_id = CODEC_ID_PCM_S24BE; break;
1270 case 32: codec_id = CODEC_ID_PCM_S32BE; break;
1272 } else if (codec_id == CODEC_ID_PCM_S16LE) {
1273 switch (track->audio.bitdepth) {
1274 case 8: codec_id = CODEC_ID_PCM_U8; break;
1275 case 24: codec_id = CODEC_ID_PCM_S24LE; break;
1276 case 32: codec_id = CODEC_ID_PCM_S32LE; break;
1278 } else if (codec_id==CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1279 codec_id = CODEC_ID_PCM_F64LE;
1280 } else if (codec_id == CODEC_ID_AAC && !track->codec_priv.size) {
1281 int profile = matroska_aac_profile(track->codec_id);
1282 int sri = matroska_aac_sri(track->audio.samplerate);
1283 extradata = av_malloc(5);
1284 if (extradata == NULL)
1285 return AVERROR(ENOMEM);
1286 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1287 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1288 if (strstr(track->codec_id, "SBR")) {
1289 sri = matroska_aac_sri(track->audio.out_samplerate);
1290 extradata[2] = 0x56;
1291 extradata[3] = 0xE5;
1292 extradata[4] = 0x80 | (sri<<3);
1296 } else if (codec_id == CODEC_ID_TTA) {
1297 extradata_size = 30;
1298 extradata = av_mallocz(extradata_size);
1299 if (extradata == NULL)
1300 return AVERROR(ENOMEM);
1301 init_put_byte(&b, extradata, extradata_size, 1,
1302 NULL, NULL, NULL, NULL);
1303 put_buffer(&b, "TTA1", 4);
1305 put_le16(&b, track->audio.channels);
1306 put_le16(&b, track->audio.bitdepth);
1307 put_le32(&b, track->audio.out_samplerate);
1308 put_le32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1309 } else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
1310 codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
1311 extradata_offset = 26;
1312 } else if (codec_id == CODEC_ID_RA_144) {
1313 track->audio.out_samplerate = 8000;
1314 track->audio.channels = 1;
1315 } else if (codec_id == CODEC_ID_RA_288 || codec_id == CODEC_ID_COOK ||
1316 codec_id == CODEC_ID_ATRAC3 || codec_id == CODEC_ID_SIPR) {
1318 init_put_byte(&b, track->codec_priv.data,track->codec_priv.size,
1319 0, NULL, NULL, NULL, NULL);
1321 flavor = get_be16(&b);
1322 track->audio.coded_framesize = get_be32(&b);
1324 track->audio.sub_packet_h = get_be16(&b);
1325 track->audio.frame_size = get_be16(&b);
1326 track->audio.sub_packet_size = get_be16(&b);
1327 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1328 if (codec_id == CODEC_ID_RA_288) {
1329 st->codec->block_align = track->audio.coded_framesize;
1330 track->codec_priv.size = 0;
1332 if (codec_id == CODEC_ID_SIPR && flavor < 4) {
1333 const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1334 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1335 st->codec->bit_rate = sipr_bit_rate[flavor];
1337 st->codec->block_align = track->audio.sub_packet_size;
1338 extradata_offset = 78;
1341 track->codec_priv.size -= extradata_offset;
1343 if (codec_id == CODEC_ID_NONE)
1344 av_log(matroska->ctx, AV_LOG_INFO,
1345 "Unknown/unsupported CodecID %s.\n", track->codec_id);
1347 if (track->time_scale < 0.01)
1348 track->time_scale = 1.0;
1349 av_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1351 st->codec->codec_id = codec_id;
1353 if (strcmp(track->language, "und"))
1354 av_metadata_set(&st->metadata, "language", track->language);
1355 av_metadata_set(&st->metadata, "title", track->name);
1357 if (track->flag_default)
1358 st->disposition |= AV_DISPOSITION_DEFAULT;
1360 if (track->default_duration)
1361 av_reduce(&st->codec->time_base.num, &st->codec->time_base.den,
1362 track->default_duration, 1000000000, 30000);
1364 if (!st->codec->extradata) {
1366 st->codec->extradata = extradata;
1367 st->codec->extradata_size = extradata_size;
1368 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1369 st->codec->extradata = av_mallocz(track->codec_priv.size +
1370 FF_INPUT_BUFFER_PADDING_SIZE);
1371 if(st->codec->extradata == NULL)
1372 return AVERROR(ENOMEM);
1373 st->codec->extradata_size = track->codec_priv.size;
1374 memcpy(st->codec->extradata,
1375 track->codec_priv.data + extradata_offset,
1376 track->codec_priv.size);
1380 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1381 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1382 st->codec->codec_tag = track->video.fourcc;
1383 st->codec->width = track->video.pixel_width;
1384 st->codec->height = track->video.pixel_height;
1385 av_reduce(&st->sample_aspect_ratio.num,
1386 &st->sample_aspect_ratio.den,
1387 st->codec->height * track->video.display_width,
1388 st->codec-> width * track->video.display_height,
1390 if (st->codec->codec_id != CODEC_ID_H264)
1391 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1392 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1393 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1394 st->codec->sample_rate = track->audio.out_samplerate;
1395 st->codec->channels = track->audio.channels;
1396 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1397 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1401 attachements = attachements_list->elem;
1402 for (j=0; j<attachements_list->nb_elem; j++) {
1403 if (!(attachements[j].filename && attachements[j].mime &&
1404 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1405 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1407 AVStream *st = av_new_stream(s, 0);
1410 av_metadata_set(&st->metadata, "filename",attachements[j].filename);
1411 st->codec->codec_id = CODEC_ID_NONE;
1412 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
1413 st->codec->extradata = av_malloc(attachements[j].bin.size);
1414 if(st->codec->extradata == NULL)
1416 st->codec->extradata_size = attachements[j].bin.size;
1417 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1419 for (i=0; ff_mkv_mime_tags[i].id != CODEC_ID_NONE; i++) {
1420 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1421 strlen(ff_mkv_mime_tags[i].str))) {
1422 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1426 attachements[j].stream = st;
1430 chapters = chapters_list->elem;
1431 for (i=0; i<chapters_list->nb_elem; i++)
1432 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1433 && (max_start==0 || chapters[i].start > max_start)) {
1434 chapters[i].chapter =
1435 ff_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1436 chapters[i].start, chapters[i].end,
1438 av_metadata_set(&chapters[i].chapter->metadata,
1439 "title", chapters[i].title);
1440 max_start = chapters[i].start;
1443 index_list = &matroska->index;
1444 index = index_list->elem;
1445 if (index_list->nb_elem
1446 && index[0].time > 100000000000000/matroska->time_scale) {
1447 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1448 index_scale = matroska->time_scale;
1450 for (i=0; i<index_list->nb_elem; i++) {
1451 EbmlList *pos_list = &index[i].pos;
1452 MatroskaIndexPos *pos = pos_list->elem;
1453 for (j=0; j<pos_list->nb_elem; j++) {
1454 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1456 if (track && track->stream)
1457 av_add_index_entry(track->stream,
1458 pos[j].pos + matroska->segment_start,
1459 index[i].time/index_scale, 0, 0,
1464 matroska_convert_tags(s);
1470 * Put one packet in an application-supplied AVPacket struct.
1471 * Returns 0 on success or -1 on failure.
1473 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1476 if (matroska->num_packets > 0) {
1477 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1478 av_free(matroska->packets[0]);
1479 if (matroska->num_packets > 1) {
1480 memmove(&matroska->packets[0], &matroska->packets[1],
1481 (matroska->num_packets - 1) * sizeof(AVPacket *));
1483 av_realloc(matroska->packets, (matroska->num_packets - 1) *
1484 sizeof(AVPacket *));
1486 av_freep(&matroska->packets);
1488 matroska->num_packets--;
1496 * Free all packets in our internal queue.
1498 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1500 if (matroska->packets) {
1502 for (n = 0; n < matroska->num_packets; n++) {
1503 av_free_packet(matroska->packets[n]);
1504 av_free(matroska->packets[n]);
1506 av_freep(&matroska->packets);
1507 matroska->num_packets = 0;
1511 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
1512 int size, int64_t pos, uint64_t cluster_time,
1513 uint64_t duration, int is_keyframe,
1514 int64_t cluster_pos)
1516 uint64_t timecode = AV_NOPTS_VALUE;
1517 MatroskaTrack *track;
1522 uint32_t *lace_size = NULL;
1523 int n, flags, laces = 0;
1526 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
1527 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
1533 track = matroska_find_track_by_num(matroska, num);
1534 if (size <= 3 || !track || !track->stream) {
1535 av_log(matroska->ctx, AV_LOG_INFO,
1536 "Invalid stream %"PRIu64" or size %u\n", num, size);
1540 if (st->discard >= AVDISCARD_ALL)
1542 if (duration == AV_NOPTS_VALUE)
1543 duration = track->default_duration / matroska->time_scale;
1545 block_time = AV_RB16(data);
1549 if (is_keyframe == -1)
1550 is_keyframe = flags & 0x80 ? PKT_FLAG_KEY : 0;
1552 if (cluster_time != (uint64_t)-1
1553 && (block_time >= 0 || cluster_time >= -block_time)) {
1554 timecode = cluster_time + block_time;
1555 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
1556 && timecode < track->end_timecode)
1557 is_keyframe = 0; /* overlapping subtitles are not key frame */
1559 av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
1560 track->end_timecode = FFMAX(track->end_timecode, timecode+duration);
1563 if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1564 if (!is_keyframe || timecode < matroska->skip_to_timecode)
1566 matroska->skip_to_keyframe = 0;
1569 switch ((flags & 0x06) >> 1) {
1570 case 0x0: /* no lacing */
1572 lace_size = av_mallocz(sizeof(int));
1573 lace_size[0] = size;
1576 case 0x1: /* Xiph lacing */
1577 case 0x2: /* fixed-size lacing */
1578 case 0x3: /* EBML lacing */
1579 assert(size>0); // size <=3 is checked before size-=3 above
1580 laces = (*data) + 1;
1583 lace_size = av_mallocz(laces * sizeof(int));
1585 switch ((flags & 0x06) >> 1) {
1586 case 0x1: /* Xiph lacing */ {
1589 for (n = 0; res == 0 && n < laces - 1; n++) {
1596 lace_size[n] += temp;
1602 total += lace_size[n];
1604 lace_size[n] = size - total;
1608 case 0x2: /* fixed-size lacing */
1609 for (n = 0; n < laces; n++)
1610 lace_size[n] = size / laces;
1613 case 0x3: /* EBML lacing */ {
1615 n = matroska_ebmlnum_uint(matroska, data, size, &num);
1617 av_log(matroska->ctx, AV_LOG_INFO,
1618 "EBML block data error\n");
1623 total = lace_size[0] = num;
1624 for (n = 1; res == 0 && n < laces - 1; n++) {
1627 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1629 av_log(matroska->ctx, AV_LOG_INFO,
1630 "EBML block data error\n");
1635 lace_size[n] = lace_size[n - 1] + snum;
1636 total += lace_size[n];
1638 lace_size[n] = size - total;
1646 for (n = 0; n < laces; n++) {
1647 if ((st->codec->codec_id == CODEC_ID_RA_288 ||
1648 st->codec->codec_id == CODEC_ID_COOK ||
1649 st->codec->codec_id == CODEC_ID_SIPR ||
1650 st->codec->codec_id == CODEC_ID_ATRAC3) &&
1651 st->codec->block_align && track->audio.sub_packet_size) {
1652 int a = st->codec->block_align;
1653 int sps = track->audio.sub_packet_size;
1654 int cfs = track->audio.coded_framesize;
1655 int h = track->audio.sub_packet_h;
1656 int y = track->audio.sub_packet_cnt;
1657 int w = track->audio.frame_size;
1660 if (!track->audio.pkt_cnt) {
1661 if (st->codec->codec_id == CODEC_ID_RA_288)
1662 for (x=0; x<h/2; x++)
1663 memcpy(track->audio.buf+x*2*w+y*cfs,
1665 else if (st->codec->codec_id == CODEC_ID_SIPR)
1666 memcpy(track->audio.buf + y*w, data, w);
1668 for (x=0; x<w/sps; x++)
1669 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
1671 if (++track->audio.sub_packet_cnt >= h) {
1672 if (st->codec->codec_id == CODEC_ID_SIPR)
1673 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
1674 track->audio.sub_packet_cnt = 0;
1675 track->audio.pkt_cnt = h*w / a;
1678 while (track->audio.pkt_cnt) {
1679 pkt = av_mallocz(sizeof(AVPacket));
1680 av_new_packet(pkt, a);
1681 memcpy(pkt->data, track->audio.buf
1682 + a * (h*w / a - track->audio.pkt_cnt--), a);
1684 pkt->stream_index = st->index;
1685 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1688 MatroskaTrackEncoding *encodings = track->encodings.elem;
1689 int offset = 0, pkt_size = lace_size[n];
1690 uint8_t *pkt_data = data;
1692 if (lace_size[n] > size) {
1693 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
1697 if (encodings && encodings->scope & 1) {
1698 offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
1703 pkt = av_mallocz(sizeof(AVPacket));
1704 /* XXX: prevent data copy... */
1705 if (av_new_packet(pkt, pkt_size+offset) < 0) {
1707 res = AVERROR(ENOMEM);
1711 memcpy (pkt->data, encodings->compression.settings.data, offset);
1712 memcpy (pkt->data+offset, pkt_data, pkt_size);
1714 if (pkt_data != data)
1718 pkt->flags = is_keyframe;
1719 pkt->stream_index = st->index;
1721 if (track->ms_compat)
1722 pkt->dts = timecode;
1724 pkt->pts = timecode;
1726 if (st->codec->codec_id == CODEC_ID_TEXT)
1727 pkt->convergence_duration = duration;
1728 else if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE)
1729 pkt->duration = duration;
1731 if (st->codec->codec_id == CODEC_ID_SSA)
1732 matroska_fix_ass_packet(matroska, pkt, duration);
1734 if (matroska->prev_pkt &&
1735 timecode != AV_NOPTS_VALUE &&
1736 matroska->prev_pkt->pts == timecode &&
1737 matroska->prev_pkt->stream_index == st->index)
1738 matroska_merge_packets(matroska->prev_pkt, pkt);
1740 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1741 matroska->prev_pkt = pkt;
1745 if (timecode != AV_NOPTS_VALUE)
1746 timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
1747 data += lace_size[n];
1748 size -= lace_size[n];
1756 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
1758 MatroskaCluster cluster = { 0 };
1759 EbmlList *blocks_list;
1760 MatroskaBlock *blocks;
1762 int64_t pos = url_ftell(matroska->ctx->pb);
1763 matroska->prev_pkt = NULL;
1764 if (matroska->has_cluster_id){
1765 /* For the first cluster we parse, its ID was already read as
1766 part of matroska_read_header(), so don't read it again */
1767 res = ebml_parse_id(matroska, matroska_clusters,
1768 MATROSKA_ID_CLUSTER, &cluster);
1769 pos -= 4; /* sizeof the ID which was already read */
1770 matroska->has_cluster_id = 0;
1772 res = ebml_parse(matroska, matroska_clusters, &cluster);
1773 blocks_list = &cluster.blocks;
1774 blocks = blocks_list->elem;
1775 for (i=0; i<blocks_list->nb_elem; i++)
1776 if (blocks[i].bin.size > 0) {
1777 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
1778 res=matroska_parse_block(matroska,
1779 blocks[i].bin.data, blocks[i].bin.size,
1780 blocks[i].bin.pos, cluster.timecode,
1781 blocks[i].duration, is_keyframe,
1784 ebml_free(matroska_cluster, &cluster);
1785 if (res < 0) matroska->done = 1;
1789 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
1791 MatroskaDemuxContext *matroska = s->priv_data;
1793 while (matroska_deliver_packet(matroska, pkt)) {
1796 matroska_parse_cluster(matroska);
1802 static int matroska_read_seek(AVFormatContext *s, int stream_index,
1803 int64_t timestamp, int flags)
1805 MatroskaDemuxContext *matroska = s->priv_data;
1806 MatroskaTrack *tracks = matroska->tracks.elem;
1807 AVStream *st = s->streams[stream_index];
1808 int i, index, index_sub, index_min;
1810 if (!st->nb_index_entries)
1812 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
1814 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1815 url_fseek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
1816 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1817 matroska_clear_queue(matroska);
1818 if (matroska_parse_cluster(matroska) < 0)
1823 matroska_clear_queue(matroska);
1828 for (i=0; i < matroska->tracks.nb_elem; i++) {
1829 tracks[i].end_timecode = 0;
1830 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
1831 && !tracks[i].stream->discard != AVDISCARD_ALL) {
1832 index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
1834 && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
1835 && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
1836 index_min = index_sub;
1840 url_fseek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
1841 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
1842 matroska->skip_to_timecode = st->index_entries[index].timestamp;
1844 av_update_cur_dts(s, st, st->index_entries[index].timestamp);
1848 static int matroska_read_close(AVFormatContext *s)
1850 MatroskaDemuxContext *matroska = s->priv_data;
1851 MatroskaTrack *tracks = matroska->tracks.elem;
1854 matroska_clear_queue(matroska);
1856 for (n=0; n < matroska->tracks.nb_elem; n++)
1857 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
1858 av_free(tracks[n].audio.buf);
1859 ebml_free(matroska_segment, matroska);
1864 AVInputFormat matroska_demuxer = {
1866 NULL_IF_CONFIG_SMALL("Matroska file format"),
1867 sizeof(MatroskaDemuxContext),
1869 matroska_read_header,
1870 matroska_read_packet,
1871 matroska_read_close,
1873 .metadata_conv = ff_mkv_metadata_conv,