2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The FFmpeg Project
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * Matroska file demuxer
25 * 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/.
34 /* For ff_codec_get_id(). */
39 #include "libavcodec/mpeg4audio.h"
40 #include "libavutil/intfloat_readwrite.h"
41 #include "libavutil/intreadwrite.h"
42 #include "libavutil/avstring.h"
43 #include "libavutil/lzo.h"
63 typedef const struct EbmlSyntax {
72 const struct EbmlSyntax *n;
92 uint64_t doctype_version;
98 } MatroskaTrackCompression;
103 MatroskaTrackCompression compression;
104 } MatroskaTrackEncoding;
108 uint64_t display_width;
109 uint64_t display_height;
110 uint64_t pixel_width;
111 uint64_t pixel_height;
113 } MatroskaTrackVideo;
117 double out_samplerate;
121 /* real audio header (extracted from extradata) */
129 } MatroskaTrackAudio;
140 uint64_t default_duration;
141 uint64_t flag_default;
142 MatroskaTrackVideo video;
143 MatroskaTrackAudio audio;
147 int64_t end_timecode;
158 } MatroskaAttachement;
196 MatroskaTagTarget target;
211 AVFormatContext *ctx;
215 MatroskaLevel levels[EBML_MAX_DEPTH];
222 EbmlList attachments;
228 /* byte position of the segment inside the stream */
229 int64_t segment_start;
231 /* the packet queue */
239 /* What to skip before effectively reading a packet. */
240 int skip_to_keyframe;
241 uint64_t skip_to_timecode;
242 } MatroskaDemuxContext;
256 static EbmlSyntax ebml_header[] = {
257 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
258 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
259 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
260 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
261 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
262 { EBML_ID_EBMLVERSION, EBML_NONE },
263 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
267 static EbmlSyntax ebml_syntax[] = {
268 { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
272 static EbmlSyntax matroska_info[] = {
273 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
274 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
275 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
276 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
277 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
278 { MATROSKA_ID_DATEUTC, EBML_NONE },
279 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
283 static EbmlSyntax matroska_track_video[] = {
284 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
285 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
286 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
287 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
288 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
289 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
290 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
291 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
292 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
293 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
294 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
295 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
296 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_NONE },
297 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
301 static EbmlSyntax matroska_track_audio[] = {
302 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
303 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
304 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
305 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
309 static EbmlSyntax matroska_track_encoding_compression[] = {
310 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
311 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
315 static EbmlSyntax matroska_track_encoding[] = {
316 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
317 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
318 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
319 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
323 static EbmlSyntax matroska_track_encodings[] = {
324 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
328 static EbmlSyntax matroska_track[] = {
329 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
330 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack,name) },
331 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
332 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
333 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
334 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
335 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
336 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
337 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
338 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
339 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
340 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
341 { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
342 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
343 { MATROSKA_ID_TRACKFLAGFORCED, EBML_NONE },
344 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
345 { MATROSKA_ID_CODECNAME, EBML_NONE },
346 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
347 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
348 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
349 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
350 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
351 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
355 static EbmlSyntax matroska_tracks[] = {
356 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
360 static EbmlSyntax matroska_attachment[] = {
361 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
362 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
363 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
364 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
365 { MATROSKA_ID_FILEDESC, EBML_NONE },
369 static EbmlSyntax matroska_attachments[] = {
370 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
374 static EbmlSyntax matroska_chapter_display[] = {
375 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
376 { MATROSKA_ID_CHAPLANG, EBML_NONE },
380 static EbmlSyntax matroska_chapter_entry[] = {
381 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
382 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
383 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
384 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
385 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
386 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
387 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
388 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
392 static EbmlSyntax matroska_chapter[] = {
393 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
394 { MATROSKA_ID_EDITIONUID, EBML_NONE },
395 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
396 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
397 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
401 static EbmlSyntax matroska_chapters[] = {
402 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
406 static EbmlSyntax matroska_index_pos[] = {
407 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
408 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
409 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
413 static EbmlSyntax matroska_index_entry[] = {
414 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
415 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
419 static EbmlSyntax matroska_index[] = {
420 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
424 static EbmlSyntax matroska_simpletag[] = {
425 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
426 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
427 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag,lang), {.s="und"} },
428 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag,def) },
429 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
433 static EbmlSyntax matroska_tagtargets[] = {
434 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget,type) },
435 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
436 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
437 { MATROSKA_ID_TAGTARGETS_CHAPTERUID,EBML_UINT, 0, offsetof(MatroskaTagTarget,chapteruid) },
438 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
442 static EbmlSyntax matroska_tag[] = {
443 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
444 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
448 static EbmlSyntax matroska_tags[] = {
449 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
453 static EbmlSyntax matroska_seekhead_entry[] = {
454 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
455 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
459 static EbmlSyntax matroska_seekhead[] = {
460 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
464 static EbmlSyntax matroska_segment[] = {
465 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
466 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
467 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
468 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
469 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
470 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
471 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
472 { MATROSKA_ID_CLUSTER, EBML_STOP, 0, offsetof(MatroskaDemuxContext,has_cluster_id) },
476 static EbmlSyntax matroska_segments[] = {
477 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
481 static EbmlSyntax matroska_blockgroup[] = {
482 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
483 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
484 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration), {.u=AV_NOPTS_VALUE} },
485 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
486 { 1, EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
490 static EbmlSyntax matroska_cluster[] = {
491 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
492 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
493 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
494 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
495 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
499 static EbmlSyntax matroska_clusters[] = {
500 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
501 { MATROSKA_ID_INFO, EBML_NONE },
502 { MATROSKA_ID_CUES, EBML_NONE },
503 { MATROSKA_ID_TAGS, EBML_NONE },
504 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
509 * Return: Whether we reached the end of a level in the hierarchy or not.
511 static int ebml_level_end(MatroskaDemuxContext *matroska)
513 ByteIOContext *pb = matroska->ctx->pb;
514 int64_t pos = url_ftell(pb);
516 if (matroska->num_levels > 0) {
517 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
518 if (pos - level->start >= level->length) {
519 matroska->num_levels--;
527 * Read: an "EBML number", which is defined as a variable-length
528 * array of bytes. The first byte indicates the length by giving a
529 * number of 0-bits followed by a one. The position of the first
530 * "one" bit inside the first byte indicates the length of this
532 * Returns: number of bytes read, < 0 on error
534 static int ebml_read_num(MatroskaDemuxContext *matroska, ByteIOContext *pb,
535 int max_size, uint64_t *number)
537 int len_mask = 0x80, read = 1, n = 1;
540 /* The first byte tells us the length in bytes - get_byte() can normally
541 * return 0, but since that's not a valid first ebmlID byte, we can
542 * use it safely here to catch EOS. */
543 if (!(total = get_byte(pb))) {
544 /* we might encounter EOS here */
546 int64_t pos = url_ftell(pb);
547 av_log(matroska->ctx, AV_LOG_ERROR,
548 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
551 return AVERROR(EIO); /* EOS or actual I/O error */
554 /* get the length of the EBML number */
555 while (read <= max_size && !(total & len_mask)) {
559 if (read > max_size) {
560 int64_t pos = url_ftell(pb) - 1;
561 av_log(matroska->ctx, AV_LOG_ERROR,
562 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
563 (uint8_t) total, pos, pos);
564 return AVERROR_INVALIDDATA;
567 /* read out length */
570 total = (total << 8) | get_byte(pb);
578 * Read the next element as an unsigned int.
579 * 0 is success, < 0 is failure.
581 static int ebml_read_uint(ByteIOContext *pb, int size, uint64_t *num)
585 if (size < 1 || size > 8)
586 return AVERROR_INVALIDDATA;
588 /* big-endian ordering; build up number */
591 *num = (*num << 8) | get_byte(pb);
597 * Read the next element as a float.
598 * 0 is success, < 0 is failure.
600 static int ebml_read_float(ByteIOContext *pb, int size, double *num)
603 *num= av_int2flt(get_be32(pb));
605 *num= av_int2dbl(get_be64(pb));
607 return AVERROR_INVALIDDATA;
613 * Read the next element as an ASCII string.
614 * 0 is success, < 0 is failure.
616 static int ebml_read_ascii(ByteIOContext *pb, int size, char **str)
619 /* EBML strings are usually not 0-terminated, so we allocate one
620 * byte more, read the string and NULL-terminate it ourselves. */
621 if (!(*str = av_malloc(size + 1)))
622 return AVERROR(ENOMEM);
623 if (get_buffer(pb, (uint8_t *) *str, size) != size) {
633 * Read the next element as binary data.
634 * 0 is success, < 0 is failure.
636 static int ebml_read_binary(ByteIOContext *pb, int length, EbmlBin *bin)
639 if (!(bin->data = av_malloc(length)))
640 return AVERROR(ENOMEM);
643 bin->pos = url_ftell(pb);
644 if (get_buffer(pb, bin->data, length) != length)
651 * Read the next element, but only the header. The contents
652 * are supposed to be sub-elements which can be read separately.
653 * 0 is success, < 0 is failure.
655 static int ebml_read_master(MatroskaDemuxContext *matroska, int length)
657 ByteIOContext *pb = matroska->ctx->pb;
658 MatroskaLevel *level;
660 if (matroska->num_levels >= EBML_MAX_DEPTH) {
661 av_log(matroska->ctx, AV_LOG_ERROR,
662 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
663 return AVERROR(ENOSYS);
666 level = &matroska->levels[matroska->num_levels++];
667 level->start = url_ftell(pb);
668 level->length = length;
674 * Read signed/unsigned "EBML" numbers.
675 * Return: number of bytes processed, < 0 on error
677 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
678 uint8_t *data, uint32_t size, uint64_t *num)
681 init_put_byte(&pb, data, size, 0, NULL, NULL, NULL, NULL);
682 return ebml_read_num(matroska, &pb, 8, num);
686 * Same as above, but signed.
688 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
689 uint8_t *data, uint32_t size, int64_t *num)
694 /* read as unsigned number first */
695 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
698 /* make signed (weird way) */
699 *num = unum - ((1LL << (7*res - 1)) - 1);
704 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
705 EbmlSyntax *syntax, void *data);
707 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
708 uint32_t id, void *data)
711 for (i=0; syntax[i].id; i++)
712 if (id == syntax[i].id)
714 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32)
715 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
716 return ebml_parse_elem(matroska, &syntax[i], data);
719 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
723 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
725 return res < 0 ? res : ebml_parse_id(matroska, syntax, id, data);
728 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
733 for (i=0; syntax[i].id; i++)
734 switch (syntax[i].type) {
736 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
739 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
743 *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
747 while (!res && !ebml_level_end(matroska))
748 res = ebml_parse(matroska, syntax, data);
753 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
754 EbmlSyntax *syntax, void *data)
756 ByteIOContext *pb = matroska->ctx->pb;
757 uint32_t id = syntax->id;
761 data = (char *)data + syntax->data_offset;
762 if (syntax->list_elem_size) {
763 EbmlList *list = data;
764 list->elem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
765 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
766 memset(data, 0, syntax->list_elem_size);
770 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP)
771 if ((res = ebml_read_num(matroska, pb, 8, &length)) < 0)
774 switch (syntax->type) {
775 case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
776 case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
778 case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
779 case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
780 case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
782 if (id == MATROSKA_ID_SEGMENT)
783 matroska->segment_start = url_ftell(matroska->ctx->pb);
784 return ebml_parse_nest(matroska, syntax->def.n, data);
785 case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
786 case EBML_STOP: *(int *)data = 1; return 1;
787 default: return url_fseek(pb,length,SEEK_CUR)<0 ? AVERROR(EIO) : 0;
789 if (res == AVERROR_INVALIDDATA)
790 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
791 else if (res == AVERROR(EIO))
792 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
796 static void ebml_free(EbmlSyntax *syntax, void *data)
799 for (i=0; syntax[i].id; i++) {
800 void *data_off = (char *)data + syntax[i].data_offset;
801 switch (syntax[i].type) {
803 case EBML_UTF8: av_freep(data_off); break;
804 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
806 if (syntax[i].list_elem_size) {
807 EbmlList *list = data_off;
808 char *ptr = list->elem;
809 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
810 ebml_free(syntax[i].def.n, ptr);
813 ebml_free(syntax[i].def.n, data_off);
823 static int matroska_probe(AVProbeData *p)
826 int len_mask = 0x80, size = 1, n = 1;
827 static const char probe_data[] = "matroska";
830 if (AV_RB32(p->buf) != EBML_ID_HEADER)
833 /* length of header */
835 while (size <= 8 && !(total & len_mask)) {
841 total &= (len_mask - 1);
843 total = (total << 8) | p->buf[4 + n++];
845 /* Does the probe data contain the whole header? */
846 if (p->buf_size < 4 + size + total)
849 /* The header must contain the document type 'matroska'. For now,
850 * we don't parse the whole header but simply check for the
851 * availability of that array of characters inside the header.
852 * Not fully fool-proof, but good enough. */
853 for (n = 4+size; n <= 4+size+total-(sizeof(probe_data)-1); n++)
854 if (!memcmp(p->buf+n, probe_data, sizeof(probe_data)-1))
855 return AVPROBE_SCORE_MAX;
860 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
863 MatroskaTrack *tracks = matroska->tracks.elem;
866 for (i=0; i < matroska->tracks.nb_elem; i++)
867 if (tracks[i].num == num)
870 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
874 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
875 MatroskaTrack *track)
877 MatroskaTrackEncoding *encodings = track->encodings.elem;
878 uint8_t* data = *buf;
879 int isize = *buf_size;
880 uint8_t* pkt_data = NULL;
881 int pkt_size = isize;
885 switch (encodings[0].compression.algo) {
886 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
887 return encodings[0].compression.settings.size;
888 case MATROSKA_TRACK_ENCODING_COMP_LZO:
890 olen = pkt_size *= 3;
891 pkt_data = av_realloc(pkt_data, pkt_size+AV_LZO_OUTPUT_PADDING);
892 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
893 } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
899 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
900 z_stream zstream = {0};
901 if (inflateInit(&zstream) != Z_OK)
903 zstream.next_in = data;
904 zstream.avail_in = isize;
907 pkt_data = av_realloc(pkt_data, pkt_size);
908 zstream.avail_out = pkt_size - zstream.total_out;
909 zstream.next_out = pkt_data + zstream.total_out;
910 result = inflate(&zstream, Z_NO_FLUSH);
911 } while (result==Z_OK && pkt_size<10000000);
912 pkt_size = zstream.total_out;
913 inflateEnd(&zstream);
914 if (result != Z_STREAM_END)
920 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
921 bz_stream bzstream = {0};
922 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
924 bzstream.next_in = data;
925 bzstream.avail_in = isize;
928 pkt_data = av_realloc(pkt_data, pkt_size);
929 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
930 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
931 result = BZ2_bzDecompress(&bzstream);
932 } while (result==BZ_OK && pkt_size<10000000);
933 pkt_size = bzstream.total_out_lo32;
934 BZ2_bzDecompressEnd(&bzstream);
935 if (result != BZ_STREAM_END)
945 *buf_size = pkt_size;
952 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
953 AVPacket *pkt, uint64_t display_duration)
955 char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
956 for (; *ptr!=',' && ptr<end-1; ptr++);
959 for (; *ptr!=',' && ptr<end-1; ptr++);
961 int64_t end_pts = pkt->pts + display_duration;
962 int sc = matroska->time_scale * pkt->pts / 10000000;
963 int ec = matroska->time_scale * end_pts / 10000000;
964 int sh, sm, ss, eh, em, es, len;
965 sh = sc/360000; sc -= 360000*sh;
966 sm = sc/ 6000; sc -= 6000*sm;
967 ss = sc/ 100; sc -= 100*ss;
968 eh = ec/360000; ec -= 360000*eh;
969 em = ec/ 6000; ec -= 6000*em;
970 es = ec/ 100; ec -= 100*es;
972 len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
973 if (!(line = av_malloc(len)))
975 snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
976 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
979 pkt->size = strlen(line);
983 static void matroska_merge_packets(AVPacket *out, AVPacket *in)
985 out->data = av_realloc(out->data, out->size+in->size);
986 memcpy(out->data+out->size, in->data, in->size);
987 out->size += in->size;
988 av_destruct_packet(in);
992 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
993 AVMetadata **metadata, char *prefix)
995 MatroskaTag *tags = list->elem;
999 for (i=0; i < list->nb_elem; i++) {
1000 const char *lang = strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1001 if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1002 else av_strlcpy(key, tags[i].name, sizeof(key));
1003 if (tags[i].def || !lang) {
1004 av_metadata_set2(metadata, key, tags[i].string, 0);
1005 if (tags[i].sub.nb_elem)
1006 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1009 av_strlcat(key, "-", sizeof(key));
1010 av_strlcat(key, lang, sizeof(key));
1011 av_metadata_set2(metadata, key, tags[i].string, 0);
1012 if (tags[i].sub.nb_elem)
1013 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1018 static void matroska_convert_tags(AVFormatContext *s)
1020 MatroskaDemuxContext *matroska = s->priv_data;
1021 MatroskaTags *tags = matroska->tags.elem;
1024 for (i=0; i < matroska->tags.nb_elem; i++) {
1025 if (tags[i].target.attachuid) {
1026 MatroskaAttachement *attachment = matroska->attachments.elem;
1027 for (j=0; j<matroska->attachments.nb_elem; j++)
1028 if (attachment[j].uid == tags[i].target.attachuid)
1029 matroska_convert_tag(s, &tags[i].tag,
1030 &attachment[j].stream->metadata, NULL);
1031 } else if (tags[i].target.chapteruid) {
1032 MatroskaChapter *chapter = matroska->chapters.elem;
1033 for (j=0; j<matroska->chapters.nb_elem; j++)
1034 if (chapter[j].uid == tags[i].target.chapteruid)
1035 matroska_convert_tag(s, &tags[i].tag,
1036 &chapter[j].chapter->metadata, NULL);
1037 } else if (tags[i].target.trackuid) {
1038 MatroskaTrack *track = matroska->tracks.elem;
1039 for (j=0; j<matroska->tracks.nb_elem; j++)
1040 if (track[j].uid == tags[i].target.trackuid)
1041 matroska_convert_tag(s, &tags[i].tag,
1042 &track[j].stream->metadata, NULL);
1044 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1045 tags[i].target.type);
1050 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1052 EbmlList *seekhead_list = &matroska->seekhead;
1053 MatroskaSeekhead *seekhead = seekhead_list->elem;
1054 uint32_t level_up = matroska->level_up;
1055 int64_t before_pos = url_ftell(matroska->ctx->pb);
1056 MatroskaLevel level;
1059 for (i=0; i<seekhead_list->nb_elem; i++) {
1060 int64_t offset = seekhead[i].pos + matroska->segment_start;
1062 if (seekhead[i].pos <= before_pos
1063 || seekhead[i].id == MATROSKA_ID_SEEKHEAD
1064 || seekhead[i].id == MATROSKA_ID_CLUSTER)
1068 if (url_fseek(matroska->ctx->pb, offset, SEEK_SET) != offset)
1071 /* We don't want to lose our seekhead level, so we add
1072 * a dummy. This is a crude hack. */
1073 if (matroska->num_levels == EBML_MAX_DEPTH) {
1074 av_log(matroska->ctx, AV_LOG_INFO,
1075 "Max EBML element depth (%d) reached, "
1076 "cannot parse further.\n", EBML_MAX_DEPTH);
1081 level.length = (uint64_t)-1;
1082 matroska->levels[matroska->num_levels] = level;
1083 matroska->num_levels++;
1085 ebml_parse(matroska, matroska_segment, matroska);
1087 /* remove dummy level */
1088 while (matroska->num_levels) {
1089 uint64_t length = matroska->levels[--matroska->num_levels].length;
1090 if (length == (uint64_t)-1)
1096 url_fseek(matroska->ctx->pb, before_pos, SEEK_SET);
1097 matroska->level_up = level_up;
1100 static int matroska_aac_profile(char *codec_id)
1102 static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1105 for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1106 if (strstr(codec_id, aac_profiles[profile]))
1111 static int matroska_aac_sri(int samplerate)
1115 for (sri=0; sri<FF_ARRAY_ELEMS(ff_mpeg4audio_sample_rates); sri++)
1116 if (ff_mpeg4audio_sample_rates[sri] == samplerate)
1121 static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
1123 MatroskaDemuxContext *matroska = s->priv_data;
1124 EbmlList *attachements_list = &matroska->attachments;
1125 MatroskaAttachement *attachements;
1126 EbmlList *chapters_list = &matroska->chapters;
1127 MatroskaChapter *chapters;
1128 MatroskaTrack *tracks;
1129 EbmlList *index_list;
1130 MatroskaIndex *index;
1131 int index_scale = 1;
1132 uint64_t max_start = 0;
1139 /* First read the EBML header. */
1140 if (ebml_parse(matroska, ebml_syntax, &ebml)
1141 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1142 || ebml.id_length > sizeof(uint32_t) || strcmp(ebml.doctype, "matroska")
1143 || ebml.doctype_version > 2) {
1144 av_log(matroska->ctx, AV_LOG_ERROR,
1145 "EBML header using unsupported features\n"
1146 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1147 ebml.version, ebml.doctype, ebml.doctype_version);
1148 return AVERROR_PATCHWELCOME;
1150 ebml_free(ebml_syntax, &ebml);
1152 /* The next thing is a segment. */
1153 if (ebml_parse(matroska, matroska_segments, matroska) < 0)
1155 matroska_execute_seekhead(matroska);
1157 if (matroska->duration)
1158 matroska->ctx->duration = matroska->duration * matroska->time_scale
1159 * 1000 / AV_TIME_BASE;
1160 av_metadata_set2(&s->metadata, "title", matroska->title, 0);
1162 tracks = matroska->tracks.elem;
1163 for (i=0; i < matroska->tracks.nb_elem; i++) {
1164 MatroskaTrack *track = &tracks[i];
1165 enum CodecID codec_id = CODEC_ID_NONE;
1166 EbmlList *encodings_list = &tracks->encodings;
1167 MatroskaTrackEncoding *encodings = encodings_list->elem;
1168 uint8_t *extradata = NULL;
1169 int extradata_size = 0;
1170 int extradata_offset = 0;
1173 /* Apply some sanity checks. */
1174 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1175 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1176 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1177 av_log(matroska->ctx, AV_LOG_INFO,
1178 "Unknown or unsupported track type %"PRIu64"\n",
1182 if (track->codec_id == NULL)
1185 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1186 if (!track->default_duration)
1187 track->default_duration = 1000000000/track->video.frame_rate;
1188 if (!track->video.display_width)
1189 track->video.display_width = track->video.pixel_width;
1190 if (!track->video.display_height)
1191 track->video.display_height = track->video.pixel_height;
1192 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1193 if (!track->audio.out_samplerate)
1194 track->audio.out_samplerate = track->audio.samplerate;
1196 if (encodings_list->nb_elem > 1) {
1197 av_log(matroska->ctx, AV_LOG_ERROR,
1198 "Multiple combined encodings no supported");
1199 } else if (encodings_list->nb_elem == 1) {
1200 if (encodings[0].type ||
1201 (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1203 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1206 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1208 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
1209 encodings[0].scope = 0;
1210 av_log(matroska->ctx, AV_LOG_ERROR,
1211 "Unsupported encoding type");
1212 } else if (track->codec_priv.size && encodings[0].scope&2) {
1213 uint8_t *codec_priv = track->codec_priv.data;
1214 int offset = matroska_decode_buffer(&track->codec_priv.data,
1215 &track->codec_priv.size,
1218 track->codec_priv.data = NULL;
1219 track->codec_priv.size = 0;
1220 av_log(matroska->ctx, AV_LOG_ERROR,
1221 "Failed to decode codec private data\n");
1222 } else if (offset > 0) {
1223 track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
1224 memcpy(track->codec_priv.data,
1225 encodings[0].compression.settings.data, offset);
1226 memcpy(track->codec_priv.data+offset, codec_priv,
1227 track->codec_priv.size);
1228 track->codec_priv.size += offset;
1230 if (codec_priv != track->codec_priv.data)
1231 av_free(codec_priv);
1235 for(j=0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++){
1236 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1237 strlen(ff_mkv_codec_tags[j].str))){
1238 codec_id= ff_mkv_codec_tags[j].id;
1243 st = track->stream = av_new_stream(s, 0);
1245 return AVERROR(ENOMEM);
1247 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1248 && track->codec_priv.size >= 40
1249 && track->codec_priv.data != NULL) {
1250 track->ms_compat = 1;
1251 track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1252 codec_id = ff_codec_get_id(ff_codec_bmp_tags, track->video.fourcc);
1253 extradata_offset = 40;
1254 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1255 && track->codec_priv.size >= 14
1256 && track->codec_priv.data != NULL) {
1257 init_put_byte(&b, track->codec_priv.data, track->codec_priv.size,
1258 URL_RDONLY, NULL, NULL, NULL, NULL);
1259 ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1260 codec_id = st->codec->codec_id;
1261 extradata_offset = FFMIN(track->codec_priv.size, 18);
1262 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1263 && (track->codec_priv.size >= 86)
1264 && (track->codec_priv.data != NULL)) {
1265 track->video.fourcc = AV_RL32(track->codec_priv.data);
1266 codec_id=ff_codec_get_id(codec_movvideo_tags, track->video.fourcc);
1267 } else if (codec_id == CODEC_ID_PCM_S16BE) {
1268 switch (track->audio.bitdepth) {
1269 case 8: codec_id = CODEC_ID_PCM_U8; break;
1270 case 24: codec_id = CODEC_ID_PCM_S24BE; break;
1271 case 32: codec_id = CODEC_ID_PCM_S32BE; break;
1273 } else if (codec_id == CODEC_ID_PCM_S16LE) {
1274 switch (track->audio.bitdepth) {
1275 case 8: codec_id = CODEC_ID_PCM_U8; break;
1276 case 24: codec_id = CODEC_ID_PCM_S24LE; break;
1277 case 32: codec_id = CODEC_ID_PCM_S32LE; break;
1279 } else if (codec_id==CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1280 codec_id = CODEC_ID_PCM_F64LE;
1281 } else if (codec_id == CODEC_ID_AAC && !track->codec_priv.size) {
1282 int profile = matroska_aac_profile(track->codec_id);
1283 int sri = matroska_aac_sri(track->audio.samplerate);
1284 extradata = av_malloc(5);
1285 if (extradata == NULL)
1286 return AVERROR(ENOMEM);
1287 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1288 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1289 if (strstr(track->codec_id, "SBR")) {
1290 sri = matroska_aac_sri(track->audio.out_samplerate);
1291 extradata[2] = 0x56;
1292 extradata[3] = 0xE5;
1293 extradata[4] = 0x80 | (sri<<3);
1297 } else if (codec_id == CODEC_ID_TTA) {
1298 extradata_size = 30;
1299 extradata = av_mallocz(extradata_size);
1300 if (extradata == NULL)
1301 return AVERROR(ENOMEM);
1302 init_put_byte(&b, extradata, extradata_size, 1,
1303 NULL, NULL, NULL, NULL);
1304 put_buffer(&b, "TTA1", 4);
1306 put_le16(&b, track->audio.channels);
1307 put_le16(&b, track->audio.bitdepth);
1308 put_le32(&b, track->audio.out_samplerate);
1309 put_le32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1310 } else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
1311 codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
1312 extradata_offset = 26;
1313 } else if (codec_id == CODEC_ID_RA_144) {
1314 track->audio.out_samplerate = 8000;
1315 track->audio.channels = 1;
1316 } else if (codec_id == CODEC_ID_RA_288 || codec_id == CODEC_ID_COOK ||
1317 codec_id == CODEC_ID_ATRAC3 || codec_id == CODEC_ID_SIPR) {
1319 init_put_byte(&b, track->codec_priv.data,track->codec_priv.size,
1320 0, NULL, NULL, NULL, NULL);
1322 flavor = get_be16(&b);
1323 track->audio.coded_framesize = get_be32(&b);
1325 track->audio.sub_packet_h = get_be16(&b);
1326 track->audio.frame_size = get_be16(&b);
1327 track->audio.sub_packet_size = get_be16(&b);
1328 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1329 if (codec_id == CODEC_ID_RA_288) {
1330 st->codec->block_align = track->audio.coded_framesize;
1331 track->codec_priv.size = 0;
1333 if (codec_id == CODEC_ID_SIPR && flavor < 4) {
1334 const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1335 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1336 st->codec->bit_rate = sipr_bit_rate[flavor];
1338 st->codec->block_align = track->audio.sub_packet_size;
1339 extradata_offset = 78;
1342 track->codec_priv.size -= extradata_offset;
1344 if (codec_id == CODEC_ID_NONE)
1345 av_log(matroska->ctx, AV_LOG_INFO,
1346 "Unknown/unsupported CodecID %s.\n", track->codec_id);
1348 if (track->time_scale < 0.01)
1349 track->time_scale = 1.0;
1350 av_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1352 st->codec->codec_id = codec_id;
1354 if (strcmp(track->language, "und"))
1355 av_metadata_set2(&st->metadata, "language", track->language, 0);
1356 av_metadata_set2(&st->metadata, "title", track->name, 0);
1358 if (track->flag_default)
1359 st->disposition |= AV_DISPOSITION_DEFAULT;
1361 if (track->default_duration)
1362 av_reduce(&st->codec->time_base.num, &st->codec->time_base.den,
1363 track->default_duration, 1000000000, 30000);
1365 if (!st->codec->extradata) {
1367 st->codec->extradata = extradata;
1368 st->codec->extradata_size = extradata_size;
1369 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1370 st->codec->extradata = av_mallocz(track->codec_priv.size +
1371 FF_INPUT_BUFFER_PADDING_SIZE);
1372 if(st->codec->extradata == NULL)
1373 return AVERROR(ENOMEM);
1374 st->codec->extradata_size = track->codec_priv.size;
1375 memcpy(st->codec->extradata,
1376 track->codec_priv.data + extradata_offset,
1377 track->codec_priv.size);
1381 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1382 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1383 st->codec->codec_tag = track->video.fourcc;
1384 st->codec->width = track->video.pixel_width;
1385 st->codec->height = track->video.pixel_height;
1386 av_reduce(&st->sample_aspect_ratio.num,
1387 &st->sample_aspect_ratio.den,
1388 st->codec->height * track->video.display_width,
1389 st->codec-> width * track->video.display_height,
1391 if (st->codec->codec_id != CODEC_ID_H264)
1392 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1393 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1394 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1395 st->codec->sample_rate = track->audio.out_samplerate;
1396 st->codec->channels = track->audio.channels;
1397 if (st->codec->codec_id != CODEC_ID_AAC)
1398 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1399 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1400 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1404 attachements = attachements_list->elem;
1405 for (j=0; j<attachements_list->nb_elem; j++) {
1406 if (!(attachements[j].filename && attachements[j].mime &&
1407 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1408 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1410 AVStream *st = av_new_stream(s, 0);
1413 av_metadata_set2(&st->metadata, "filename",attachements[j].filename, 0);
1414 st->codec->codec_id = CODEC_ID_NONE;
1415 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
1416 st->codec->extradata = av_malloc(attachements[j].bin.size);
1417 if(st->codec->extradata == NULL)
1419 st->codec->extradata_size = attachements[j].bin.size;
1420 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1422 for (i=0; ff_mkv_mime_tags[i].id != CODEC_ID_NONE; i++) {
1423 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1424 strlen(ff_mkv_mime_tags[i].str))) {
1425 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1429 attachements[j].stream = st;
1433 chapters = chapters_list->elem;
1434 for (i=0; i<chapters_list->nb_elem; i++)
1435 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1436 && (max_start==0 || chapters[i].start > max_start)) {
1437 chapters[i].chapter =
1438 ff_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1439 chapters[i].start, chapters[i].end,
1441 av_metadata_set2(&chapters[i].chapter->metadata,
1442 "title", chapters[i].title, 0);
1443 max_start = chapters[i].start;
1446 index_list = &matroska->index;
1447 index = index_list->elem;
1448 if (index_list->nb_elem
1449 && index[0].time > 100000000000000/matroska->time_scale) {
1450 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1451 index_scale = matroska->time_scale;
1453 for (i=0; i<index_list->nb_elem; i++) {
1454 EbmlList *pos_list = &index[i].pos;
1455 MatroskaIndexPos *pos = pos_list->elem;
1456 for (j=0; j<pos_list->nb_elem; j++) {
1457 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1459 if (track && track->stream)
1460 av_add_index_entry(track->stream,
1461 pos[j].pos + matroska->segment_start,
1462 index[i].time/index_scale, 0, 0,
1467 matroska_convert_tags(s);
1473 * Put one packet in an application-supplied AVPacket struct.
1474 * Returns 0 on success or -1 on failure.
1476 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1479 if (matroska->num_packets > 0) {
1480 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1481 av_free(matroska->packets[0]);
1482 if (matroska->num_packets > 1) {
1483 memmove(&matroska->packets[0], &matroska->packets[1],
1484 (matroska->num_packets - 1) * sizeof(AVPacket *));
1486 av_realloc(matroska->packets, (matroska->num_packets - 1) *
1487 sizeof(AVPacket *));
1489 av_freep(&matroska->packets);
1491 matroska->num_packets--;
1499 * Free all packets in our internal queue.
1501 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1503 if (matroska->packets) {
1505 for (n = 0; n < matroska->num_packets; n++) {
1506 av_free_packet(matroska->packets[n]);
1507 av_free(matroska->packets[n]);
1509 av_freep(&matroska->packets);
1510 matroska->num_packets = 0;
1514 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
1515 int size, int64_t pos, uint64_t cluster_time,
1516 uint64_t duration, int is_keyframe,
1517 int64_t cluster_pos)
1519 uint64_t timecode = AV_NOPTS_VALUE;
1520 MatroskaTrack *track;
1525 uint32_t *lace_size = NULL;
1526 int n, flags, laces = 0;
1529 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
1530 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
1536 track = matroska_find_track_by_num(matroska, num);
1537 if (size <= 3 || !track || !track->stream) {
1538 av_log(matroska->ctx, AV_LOG_INFO,
1539 "Invalid stream %"PRIu64" or size %u\n", num, size);
1543 if (st->discard >= AVDISCARD_ALL)
1545 if (duration == AV_NOPTS_VALUE)
1546 duration = track->default_duration / matroska->time_scale;
1548 block_time = AV_RB16(data);
1552 if (is_keyframe == -1)
1553 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
1555 if (cluster_time != (uint64_t)-1
1556 && (block_time >= 0 || cluster_time >= -block_time)) {
1557 timecode = cluster_time + block_time;
1558 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
1559 && timecode < track->end_timecode)
1560 is_keyframe = 0; /* overlapping subtitles are not key frame */
1562 av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
1563 track->end_timecode = FFMAX(track->end_timecode, timecode+duration);
1566 if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1567 if (!is_keyframe || timecode < matroska->skip_to_timecode)
1569 matroska->skip_to_keyframe = 0;
1572 switch ((flags & 0x06) >> 1) {
1573 case 0x0: /* no lacing */
1575 lace_size = av_mallocz(sizeof(int));
1576 lace_size[0] = size;
1579 case 0x1: /* Xiph lacing */
1580 case 0x2: /* fixed-size lacing */
1581 case 0x3: /* EBML lacing */
1582 assert(size>0); // size <=3 is checked before size-=3 above
1583 laces = (*data) + 1;
1586 lace_size = av_mallocz(laces * sizeof(int));
1588 switch ((flags & 0x06) >> 1) {
1589 case 0x1: /* Xiph lacing */ {
1592 for (n = 0; res == 0 && n < laces - 1; n++) {
1599 lace_size[n] += temp;
1605 total += lace_size[n];
1607 lace_size[n] = size - total;
1611 case 0x2: /* fixed-size lacing */
1612 for (n = 0; n < laces; n++)
1613 lace_size[n] = size / laces;
1616 case 0x3: /* EBML lacing */ {
1618 n = matroska_ebmlnum_uint(matroska, data, size, &num);
1620 av_log(matroska->ctx, AV_LOG_INFO,
1621 "EBML block data error\n");
1626 total = lace_size[0] = num;
1627 for (n = 1; res == 0 && n < laces - 1; n++) {
1630 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1632 av_log(matroska->ctx, AV_LOG_INFO,
1633 "EBML block data error\n");
1638 lace_size[n] = lace_size[n - 1] + snum;
1639 total += lace_size[n];
1641 lace_size[n] = size - total;
1649 for (n = 0; n < laces; n++) {
1650 if ((st->codec->codec_id == CODEC_ID_RA_288 ||
1651 st->codec->codec_id == CODEC_ID_COOK ||
1652 st->codec->codec_id == CODEC_ID_SIPR ||
1653 st->codec->codec_id == CODEC_ID_ATRAC3) &&
1654 st->codec->block_align && track->audio.sub_packet_size) {
1655 int a = st->codec->block_align;
1656 int sps = track->audio.sub_packet_size;
1657 int cfs = track->audio.coded_framesize;
1658 int h = track->audio.sub_packet_h;
1659 int y = track->audio.sub_packet_cnt;
1660 int w = track->audio.frame_size;
1663 if (!track->audio.pkt_cnt) {
1664 if (st->codec->codec_id == CODEC_ID_RA_288)
1665 for (x=0; x<h/2; x++)
1666 memcpy(track->audio.buf+x*2*w+y*cfs,
1668 else if (st->codec->codec_id == CODEC_ID_SIPR)
1669 memcpy(track->audio.buf + y*w, data, w);
1671 for (x=0; x<w/sps; x++)
1672 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
1674 if (++track->audio.sub_packet_cnt >= h) {
1675 if (st->codec->codec_id == CODEC_ID_SIPR)
1676 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
1677 track->audio.sub_packet_cnt = 0;
1678 track->audio.pkt_cnt = h*w / a;
1681 while (track->audio.pkt_cnt) {
1682 pkt = av_mallocz(sizeof(AVPacket));
1683 av_new_packet(pkt, a);
1684 memcpy(pkt->data, track->audio.buf
1685 + a * (h*w / a - track->audio.pkt_cnt--), a);
1687 pkt->stream_index = st->index;
1688 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1691 MatroskaTrackEncoding *encodings = track->encodings.elem;
1692 int offset = 0, pkt_size = lace_size[n];
1693 uint8_t *pkt_data = data;
1695 if (lace_size[n] > size) {
1696 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
1700 if (encodings && encodings->scope & 1) {
1701 offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
1706 pkt = av_mallocz(sizeof(AVPacket));
1707 /* XXX: prevent data copy... */
1708 if (av_new_packet(pkt, pkt_size+offset) < 0) {
1710 res = AVERROR(ENOMEM);
1714 memcpy (pkt->data, encodings->compression.settings.data, offset);
1715 memcpy (pkt->data+offset, pkt_data, pkt_size);
1717 if (pkt_data != data)
1721 pkt->flags = is_keyframe;
1722 pkt->stream_index = st->index;
1724 if (track->ms_compat)
1725 pkt->dts = timecode;
1727 pkt->pts = timecode;
1729 if (st->codec->codec_id == CODEC_ID_TEXT)
1730 pkt->convergence_duration = duration;
1731 else if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE)
1732 pkt->duration = duration;
1734 if (st->codec->codec_id == CODEC_ID_SSA)
1735 matroska_fix_ass_packet(matroska, pkt, duration);
1737 if (matroska->prev_pkt &&
1738 timecode != AV_NOPTS_VALUE &&
1739 matroska->prev_pkt->pts == timecode &&
1740 matroska->prev_pkt->stream_index == st->index)
1741 matroska_merge_packets(matroska->prev_pkt, pkt);
1743 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1744 matroska->prev_pkt = pkt;
1748 if (timecode != AV_NOPTS_VALUE)
1749 timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
1750 data += lace_size[n];
1751 size -= lace_size[n];
1759 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
1761 MatroskaCluster cluster = { 0 };
1762 EbmlList *blocks_list;
1763 MatroskaBlock *blocks;
1765 int64_t pos = url_ftell(matroska->ctx->pb);
1766 matroska->prev_pkt = NULL;
1767 if (matroska->has_cluster_id){
1768 /* For the first cluster we parse, its ID was already read as
1769 part of matroska_read_header(), so don't read it again */
1770 res = ebml_parse_id(matroska, matroska_clusters,
1771 MATROSKA_ID_CLUSTER, &cluster);
1772 pos -= 4; /* sizeof the ID which was already read */
1773 matroska->has_cluster_id = 0;
1775 res = ebml_parse(matroska, matroska_clusters, &cluster);
1776 blocks_list = &cluster.blocks;
1777 blocks = blocks_list->elem;
1778 for (i=0; i<blocks_list->nb_elem; i++)
1779 if (blocks[i].bin.size > 0) {
1780 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
1781 res=matroska_parse_block(matroska,
1782 blocks[i].bin.data, blocks[i].bin.size,
1783 blocks[i].bin.pos, cluster.timecode,
1784 blocks[i].duration, is_keyframe,
1787 ebml_free(matroska_cluster, &cluster);
1788 if (res < 0) matroska->done = 1;
1792 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
1794 MatroskaDemuxContext *matroska = s->priv_data;
1796 while (matroska_deliver_packet(matroska, pkt)) {
1799 matroska_parse_cluster(matroska);
1805 static int matroska_read_seek(AVFormatContext *s, int stream_index,
1806 int64_t timestamp, int flags)
1808 MatroskaDemuxContext *matroska = s->priv_data;
1809 MatroskaTrack *tracks = matroska->tracks.elem;
1810 AVStream *st = s->streams[stream_index];
1811 int i, index, index_sub, index_min;
1813 if (!st->nb_index_entries)
1815 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
1817 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1818 url_fseek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
1819 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1820 matroska_clear_queue(matroska);
1821 if (matroska_parse_cluster(matroska) < 0)
1826 matroska_clear_queue(matroska);
1831 for (i=0; i < matroska->tracks.nb_elem; i++) {
1832 tracks[i].end_timecode = 0;
1833 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
1834 && !tracks[i].stream->discard != AVDISCARD_ALL) {
1835 index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
1837 && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
1838 && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
1839 index_min = index_sub;
1843 url_fseek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
1844 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
1845 matroska->skip_to_timecode = st->index_entries[index].timestamp;
1847 av_update_cur_dts(s, st, st->index_entries[index].timestamp);
1851 static int matroska_read_close(AVFormatContext *s)
1853 MatroskaDemuxContext *matroska = s->priv_data;
1854 MatroskaTrack *tracks = matroska->tracks.elem;
1857 matroska_clear_queue(matroska);
1859 for (n=0; n < matroska->tracks.nb_elem; n++)
1860 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
1861 av_free(tracks[n].audio.buf);
1862 ebml_free(matroska_segment, matroska);
1867 AVInputFormat matroska_demuxer = {
1869 NULL_IF_CONFIG_SMALL("Matroska file format"),
1870 sizeof(MatroskaDemuxContext),
1872 matroska_read_header,
1873 matroska_read_packet,
1874 matroska_read_close,
1876 .metadata_conv = ff_mkv_metadata_conv,