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/.
33 /* For codec_get_id(). */
37 #include "libavcodec/mpeg4audio.h"
38 #include "libavutil/intfloat_readwrite.h"
39 #include "libavutil/avstring.h"
40 #include "libavutil/lzo.h"
60 typedef const struct EbmlSyntax {
69 const struct EbmlSyntax *n;
89 uint64_t doctype_version;
95 } MatroskaTrackCompression;
100 MatroskaTrackCompression compression;
101 } MatroskaTrackEncoding;
105 uint64_t display_width;
106 uint64_t display_height;
107 uint64_t pixel_width;
108 uint64_t pixel_height;
110 } MatroskaTrackVideo;
114 double out_samplerate;
118 /* real audio header (extracted from extradata) */
126 } MatroskaTrackAudio;
135 uint64_t default_duration;
136 uint64_t flag_default;
137 MatroskaTrackVideo video;
138 MatroskaTrackAudio audio;
142 int64_t end_timecode;
149 } MatroskaAttachement;
185 AVFormatContext *ctx;
189 MatroskaLevel levels[EBML_MAX_DEPTH];
196 EbmlList attachments;
202 /* byte position of the segment inside the stream */
203 int64_t segment_start;
205 /* the packet queue */
213 /* What to skip before effectively reading a packet. */
214 int skip_to_keyframe;
215 uint64_t skip_to_timecode;
216 } MatroskaDemuxContext;
229 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(*x))
231 static EbmlSyntax ebml_header[] = {
232 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
233 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
234 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
235 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
236 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
237 { EBML_ID_EBMLVERSION, EBML_NONE },
238 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
242 static EbmlSyntax ebml_syntax[] = {
243 { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
247 static EbmlSyntax matroska_info[] = {
248 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
249 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
250 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
251 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
252 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
253 { MATROSKA_ID_DATEUTC, EBML_NONE },
254 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
258 static EbmlSyntax matroska_track_video[] = {
259 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
260 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
261 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
262 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
263 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
264 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
265 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
266 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
267 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
268 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
269 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
270 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
271 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_NONE },
272 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
276 static EbmlSyntax matroska_track_audio[] = {
277 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
278 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
279 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
280 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
284 static EbmlSyntax matroska_track_encoding_compression[] = {
285 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
286 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
290 static EbmlSyntax matroska_track_encoding[] = {
291 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
292 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
293 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
294 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
298 static EbmlSyntax matroska_track_encodings[] = {
299 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
303 static EbmlSyntax matroska_track[] = {
304 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
305 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
306 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
307 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
308 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
309 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
310 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
311 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
312 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
313 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
314 { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
315 { MATROSKA_ID_TRACKUID, EBML_NONE },
316 { MATROSKA_ID_TRACKNAME, EBML_NONE },
317 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
318 { MATROSKA_ID_TRACKFLAGFORCED, EBML_NONE },
319 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
320 { MATROSKA_ID_CODECNAME, EBML_NONE },
321 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
322 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
323 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
324 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
325 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
326 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
330 static EbmlSyntax matroska_tracks[] = {
331 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
335 static EbmlSyntax matroska_attachment[] = {
336 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
337 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
338 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
339 { MATROSKA_ID_FILEDESC, EBML_NONE },
340 { MATROSKA_ID_FILEUID, EBML_NONE },
344 static EbmlSyntax matroska_attachments[] = {
345 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
349 static EbmlSyntax matroska_chapter_display[] = {
350 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
351 { MATROSKA_ID_CHAPLANG, EBML_NONE },
355 static EbmlSyntax matroska_chapter_entry[] = {
356 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
357 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
358 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
359 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
360 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
361 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
362 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
363 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
367 static EbmlSyntax matroska_chapter[] = {
368 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
369 { MATROSKA_ID_EDITIONUID, EBML_NONE },
370 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
371 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
372 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
376 static EbmlSyntax matroska_chapters[] = {
377 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
381 static EbmlSyntax matroska_index_pos[] = {
382 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
383 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
384 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
388 static EbmlSyntax matroska_index_entry[] = {
389 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
390 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
394 static EbmlSyntax matroska_index[] = {
395 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
399 static EbmlSyntax matroska_simpletag[] = {
400 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
401 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
402 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
403 { MATROSKA_ID_TAGLANG, EBML_NONE },
404 { MATROSKA_ID_TAGDEFAULT, EBML_NONE },
408 static EbmlSyntax matroska_tag[] = {
409 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), 0, {.n=matroska_simpletag} },
410 { MATROSKA_ID_TAGTARGETS, EBML_NONE },
414 static EbmlSyntax matroska_tags[] = {
415 { MATROSKA_ID_TAG, EBML_NEST, 0, offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
419 static EbmlSyntax matroska_seekhead_entry[] = {
420 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
421 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
425 static EbmlSyntax matroska_seekhead[] = {
426 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
430 static EbmlSyntax matroska_segment[] = {
431 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
432 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
433 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
434 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
435 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
436 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
437 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
438 { MATROSKA_ID_CLUSTER, EBML_STOP, 0, offsetof(MatroskaDemuxContext,has_cluster_id) },
442 static EbmlSyntax matroska_segments[] = {
443 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
447 static EbmlSyntax matroska_blockgroup[] = {
448 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
449 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
450 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration), {.u=AV_NOPTS_VALUE} },
451 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
455 static EbmlSyntax matroska_cluster[] = {
456 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
457 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
458 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
459 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
460 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
464 static EbmlSyntax matroska_clusters[] = {
465 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
466 { MATROSKA_ID_INFO, EBML_NONE },
467 { MATROSKA_ID_CUES, EBML_NONE },
468 { MATROSKA_ID_TAGS, EBML_NONE },
469 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
473 #define SIZE_OFF(x) sizeof(((AVFormatContext*)0)->x),offsetof(AVFormatContext,x)
479 { "TITLE", SIZE_OFF(title) },
480 { "ARTIST", SIZE_OFF(author) },
481 { "WRITTEN_BY", SIZE_OFF(author) },
482 { "LEAD_PERFORMER", SIZE_OFF(author) },
483 { "COPYRIGHT", SIZE_OFF(copyright) },
484 { "COMMENT", SIZE_OFF(comment) },
485 { "ALBUM", SIZE_OFF(album) },
486 { "DATE_WRITTEN", SIZE_OFF(year) },
487 { "DATE_RELEASED", SIZE_OFF(year) },
488 { "PART_NUMBER", SIZE_OFF(track) },
489 { "GENRE", SIZE_OFF(genre) },
493 * Return: Whether we reached the end of a level in the hierarchy or not.
495 static int ebml_level_end(MatroskaDemuxContext *matroska)
497 ByteIOContext *pb = matroska->ctx->pb;
498 int64_t pos = url_ftell(pb);
500 if (matroska->num_levels > 0) {
501 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
502 if (pos - level->start >= level->length) {
503 matroska->num_levels--;
511 * Read: an "EBML number", which is defined as a variable-length
512 * array of bytes. The first byte indicates the length by giving a
513 * number of 0-bits followed by a one. The position of the first
514 * "one" bit inside the first byte indicates the length of this
516 * Returns: number of bytes read, < 0 on error
518 static int ebml_read_num(MatroskaDemuxContext *matroska, ByteIOContext *pb,
519 int max_size, uint64_t *number)
521 int len_mask = 0x80, read = 1, n = 1;
524 /* The first byte tells us the length in bytes - get_byte() can normally
525 * return 0, but since that's not a valid first ebmlID byte, we can
526 * use it safely here to catch EOS. */
527 if (!(total = get_byte(pb))) {
528 /* we might encounter EOS here */
530 int64_t pos = url_ftell(pb);
531 av_log(matroska->ctx, AV_LOG_ERROR,
532 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
535 return AVERROR(EIO); /* EOS or actual I/O error */
538 /* get the length of the EBML number */
539 while (read <= max_size && !(total & len_mask)) {
543 if (read > max_size) {
544 int64_t pos = url_ftell(pb) - 1;
545 av_log(matroska->ctx, AV_LOG_ERROR,
546 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
547 (uint8_t) total, pos, pos);
548 return AVERROR_INVALIDDATA;
551 /* read out length */
554 total = (total << 8) | get_byte(pb);
562 * Read the next element as an unsigned int.
563 * 0 is success, < 0 is failure.
565 static int ebml_read_uint(ByteIOContext *pb, int size, uint64_t *num)
569 if (size < 1 || size > 8)
570 return AVERROR_INVALIDDATA;
572 /* big-endian ordering; build up number */
575 *num = (*num << 8) | get_byte(pb);
581 * Read the next element as a float.
582 * 0 is success, < 0 is failure.
584 static int ebml_read_float(ByteIOContext *pb, int size, double *num)
587 *num= av_int2flt(get_be32(pb));
589 *num= av_int2dbl(get_be64(pb));
591 return AVERROR_INVALIDDATA;
597 * Read the next element as an ASCII string.
598 * 0 is success, < 0 is failure.
600 static int ebml_read_ascii(ByteIOContext *pb, int size, char **str)
603 /* EBML strings are usually not 0-terminated, so we allocate one
604 * byte more, read the string and NULL-terminate it ourselves. */
605 if (!(*str = av_malloc(size + 1)))
606 return AVERROR(ENOMEM);
607 if (get_buffer(pb, (uint8_t *) *str, size) != size) {
617 * Read the next element as binary data.
618 * 0 is success, < 0 is failure.
620 static int ebml_read_binary(ByteIOContext *pb, int length, EbmlBin *bin)
623 if (!(bin->data = av_malloc(length)))
624 return AVERROR(ENOMEM);
627 bin->pos = url_ftell(pb);
628 if (get_buffer(pb, bin->data, length) != length)
635 * Read the next element, but only the header. The contents
636 * are supposed to be sub-elements which can be read separately.
637 * 0 is success, < 0 is failure.
639 static int ebml_read_master(MatroskaDemuxContext *matroska, int length)
641 ByteIOContext *pb = matroska->ctx->pb;
642 MatroskaLevel *level;
644 if (matroska->num_levels >= EBML_MAX_DEPTH) {
645 av_log(matroska->ctx, AV_LOG_ERROR,
646 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
647 return AVERROR(ENOSYS);
650 level = &matroska->levels[matroska->num_levels++];
651 level->start = url_ftell(pb);
652 level->length = length;
658 * Read signed/unsigned "EBML" numbers.
659 * Return: number of bytes processed, < 0 on error
661 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
662 uint8_t *data, uint32_t size, uint64_t *num)
665 init_put_byte(&pb, data, size, 0, NULL, NULL, NULL, NULL);
666 return ebml_read_num(matroska, &pb, 8, num);
670 * Same as above, but signed.
672 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
673 uint8_t *data, uint32_t size, int64_t *num)
678 /* read as unsigned number first */
679 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
682 /* make signed (weird way) */
683 *num = unum - ((1LL << (7*res - 1)) - 1);
688 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
689 EbmlSyntax *syntax, void *data);
691 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
692 uint32_t id, void *data)
695 for (i=0; syntax[i].id; i++)
696 if (id == syntax[i].id)
698 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32)
699 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
700 return ebml_parse_elem(matroska, &syntax[i], data);
703 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
707 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
709 return res < 0 ? res : ebml_parse_id(matroska, syntax, id, data);
712 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
717 for (i=0; syntax[i].id; i++)
718 switch (syntax[i].type) {
720 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
723 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
727 *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
731 while (!res && !ebml_level_end(matroska))
732 res = ebml_parse(matroska, syntax, data);
737 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
738 EbmlSyntax *syntax, void *data)
740 ByteIOContext *pb = matroska->ctx->pb;
741 uint32_t id = syntax->id;
745 data = (char *)data + syntax->data_offset;
746 if (syntax->list_elem_size) {
747 EbmlList *list = data;
748 list->elem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
749 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
750 memset(data, 0, syntax->list_elem_size);
754 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP)
755 if ((res = ebml_read_num(matroska, pb, 8, &length)) < 0)
758 switch (syntax->type) {
759 case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
760 case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
762 case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
763 case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
764 case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
766 if (id == MATROSKA_ID_SEGMENT)
767 matroska->segment_start = url_ftell(matroska->ctx->pb);
768 return ebml_parse_nest(matroska, syntax->def.n, data);
769 case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
770 case EBML_STOP: *(int *)data = 1; return 1;
771 default: return url_fseek(pb,length,SEEK_CUR)<0 ? AVERROR(EIO) : 0;
773 if (res == AVERROR_INVALIDDATA)
774 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
775 else if (res == AVERROR(EIO))
776 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
780 static void ebml_free(EbmlSyntax *syntax, void *data)
783 for (i=0; syntax[i].id; i++) {
784 void *data_off = (char *)data + syntax[i].data_offset;
785 switch (syntax[i].type) {
787 case EBML_UTF8: av_freep(data_off); break;
788 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
790 if (syntax[i].list_elem_size) {
791 EbmlList *list = data_off;
792 char *ptr = list->elem;
793 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
794 ebml_free(syntax[i].def.n, ptr);
797 ebml_free(syntax[i].def.n, data_off);
807 static int matroska_probe(AVProbeData *p)
810 int len_mask = 0x80, size = 1, n = 1;
811 static const char probe_data[] = "matroska";
814 if (AV_RB32(p->buf) != EBML_ID_HEADER)
817 /* length of header */
819 while (size <= 8 && !(total & len_mask)) {
825 total &= (len_mask - 1);
827 total = (total << 8) | p->buf[4 + n++];
829 /* Does the probe data contain the whole header? */
830 if (p->buf_size < 4 + size + total)
833 /* The header must contain the document type 'matroska'. For now,
834 * we don't parse the whole header but simply check for the
835 * availability of that array of characters inside the header.
836 * Not fully fool-proof, but good enough. */
837 for (n = 4+size; n <= 4+size+total-(sizeof(probe_data)-1); n++)
838 if (!memcmp(p->buf+n, probe_data, sizeof(probe_data)-1))
839 return AVPROBE_SCORE_MAX;
844 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
847 MatroskaTrack *tracks = matroska->tracks.elem;
850 for (i=0; i < matroska->tracks.nb_elem; i++)
851 if (tracks[i].num == num)
854 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
858 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
859 MatroskaTrack *track)
861 MatroskaTrackEncoding *encodings = track->encodings.elem;
862 uint8_t* data = *buf;
863 int isize = *buf_size;
864 uint8_t* pkt_data = NULL;
865 int pkt_size = isize;
869 switch (encodings[0].compression.algo) {
870 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
871 return encodings[0].compression.settings.size;
872 case MATROSKA_TRACK_ENCODING_COMP_LZO:
874 olen = pkt_size *= 3;
875 pkt_data = av_realloc(pkt_data,
876 pkt_size+LZO_OUTPUT_PADDING);
877 result = lzo1x_decode(pkt_data, &olen, data, &isize);
878 } while (result==LZO_OUTPUT_FULL && pkt_size<10000000);
884 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
885 z_stream zstream = {0};
886 if (inflateInit(&zstream) != Z_OK)
888 zstream.next_in = data;
889 zstream.avail_in = isize;
892 pkt_data = av_realloc(pkt_data, pkt_size);
893 zstream.avail_out = pkt_size - zstream.total_out;
894 zstream.next_out = pkt_data + zstream.total_out;
895 result = inflate(&zstream, Z_NO_FLUSH);
896 } while (result==Z_OK && pkt_size<10000000);
897 pkt_size = zstream.total_out;
898 inflateEnd(&zstream);
899 if (result != Z_STREAM_END)
905 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
906 bz_stream bzstream = {0};
907 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
909 bzstream.next_in = data;
910 bzstream.avail_in = isize;
913 pkt_data = av_realloc(pkt_data, pkt_size);
914 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
915 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
916 result = BZ2_bzDecompress(&bzstream);
917 } while (result==BZ_OK && pkt_size<10000000);
918 pkt_size = bzstream.total_out_lo32;
919 BZ2_bzDecompressEnd(&bzstream);
920 if (result != BZ_STREAM_END)
930 *buf_size = pkt_size;
937 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
938 AVPacket *pkt, uint64_t display_duration)
940 char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
941 for (; *ptr!=',' && ptr<end-1; ptr++);
944 for (; *ptr!=',' && ptr<end-1; ptr++);
946 int64_t end_pts = pkt->pts + display_duration;
947 int sc = matroska->time_scale * pkt->pts / 10000000;
948 int ec = matroska->time_scale * end_pts / 10000000;
949 int sh, sm, ss, eh, em, es, len;
950 sh = sc/360000; sc -= 360000*sh;
951 sm = sc/ 6000; sc -= 6000*sm;
952 ss = sc/ 100; sc -= 100*ss;
953 eh = ec/360000; ec -= 360000*eh;
954 em = ec/ 6000; ec -= 6000*em;
955 es = ec/ 100; ec -= 100*es;
957 len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
958 if (!(line = av_malloc(len)))
960 snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
961 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
964 pkt->size = strlen(line);
968 static void matroska_merge_packets(AVPacket *out, AVPacket *in)
970 out->data = av_realloc(out->data, out->size+in->size);
971 memcpy(out->data+out->size, in->data, in->size);
972 out->size += in->size;
973 av_destruct_packet(in);
977 static void matroska_convert_tags(AVFormatContext *s, EbmlList *list)
979 MatroskaTag *tags = list->elem;
982 for (i=0; i < list->nb_elem; i++) {
983 for (j=0; j < ARRAY_SIZE(metadata); j++){
984 if (!strcmp(tags[i].name, metadata[j].name)) {
985 int *ptr = (int *)((char *)s + metadata[j].offset);
987 if (metadata[j].size > sizeof(int))
988 av_strlcpy((char *)ptr, tags[i].string, metadata[j].size);
990 *ptr = atoi(tags[i].string);
993 if (tags[i].sub.nb_elem)
994 matroska_convert_tags(s, &tags[i].sub);
998 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1000 EbmlList *seekhead_list = &matroska->seekhead;
1001 MatroskaSeekhead *seekhead = seekhead_list->elem;
1002 uint32_t level_up = matroska->level_up;
1003 int64_t before_pos = url_ftell(matroska->ctx->pb);
1004 MatroskaLevel level;
1007 for (i=0; i<seekhead_list->nb_elem; i++) {
1008 int64_t offset = seekhead[i].pos + matroska->segment_start;
1010 if (seekhead[i].pos <= before_pos
1011 || seekhead[i].id == MATROSKA_ID_SEEKHEAD
1012 || seekhead[i].id == MATROSKA_ID_CLUSTER)
1016 if (url_fseek(matroska->ctx->pb, offset, SEEK_SET) != offset)
1019 /* We don't want to lose our seekhead level, so we add
1020 * a dummy. This is a crude hack. */
1021 if (matroska->num_levels == EBML_MAX_DEPTH) {
1022 av_log(matroska->ctx, AV_LOG_INFO,
1023 "Max EBML element depth (%d) reached, "
1024 "cannot parse further.\n", EBML_MAX_DEPTH);
1029 level.length = (uint64_t)-1;
1030 matroska->levels[matroska->num_levels] = level;
1031 matroska->num_levels++;
1033 ebml_parse(matroska, matroska_segment, matroska);
1035 /* remove dummy level */
1036 while (matroska->num_levels) {
1037 uint64_t length = matroska->levels[--matroska->num_levels].length;
1038 if (length == (uint64_t)-1)
1044 url_fseek(matroska->ctx->pb, before_pos, SEEK_SET);
1045 matroska->level_up = level_up;
1048 static int matroska_aac_profile(char *codec_id)
1050 static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1053 for (profile=0; profile<ARRAY_SIZE(aac_profiles); profile++)
1054 if (strstr(codec_id, aac_profiles[profile]))
1059 static int matroska_aac_sri(int samplerate)
1063 for (sri=0; sri<ARRAY_SIZE(ff_mpeg4audio_sample_rates); sri++)
1064 if (ff_mpeg4audio_sample_rates[sri] == samplerate)
1069 static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
1071 MatroskaDemuxContext *matroska = s->priv_data;
1072 EbmlList *attachements_list = &matroska->attachments;
1073 MatroskaAttachement *attachements;
1074 EbmlList *chapters_list = &matroska->chapters;
1075 MatroskaChapter *chapters;
1076 MatroskaTrack *tracks;
1077 EbmlList *index_list;
1078 MatroskaIndex *index;
1085 /* First read the EBML header. */
1086 if (ebml_parse(matroska, ebml_syntax, &ebml)
1087 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1088 || ebml.id_length > sizeof(uint32_t) || strcmp(ebml.doctype, "matroska")
1089 || ebml.doctype_version > 2) {
1090 av_log(matroska->ctx, AV_LOG_ERROR,
1091 "EBML header using unsupported features\n"
1092 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1093 ebml.version, ebml.doctype, ebml.doctype_version);
1094 return AVERROR_NOFMT;
1096 ebml_free(ebml_syntax, &ebml);
1098 /* The next thing is a segment. */
1099 if (ebml_parse(matroska, matroska_segments, matroska) < 0)
1101 matroska_execute_seekhead(matroska);
1103 if (matroska->duration)
1104 matroska->ctx->duration = matroska->duration * matroska->time_scale
1105 * 1000 / AV_TIME_BASE;
1106 if (matroska->title)
1107 strncpy(matroska->ctx->title, matroska->title,
1108 sizeof(matroska->ctx->title)-1);
1109 matroska_convert_tags(s, &matroska->tags);
1111 tracks = matroska->tracks.elem;
1112 for (i=0; i < matroska->tracks.nb_elem; i++) {
1113 MatroskaTrack *track = &tracks[i];
1114 enum CodecID codec_id = CODEC_ID_NONE;
1115 EbmlList *encodings_list = &tracks->encodings;
1116 MatroskaTrackEncoding *encodings = encodings_list->elem;
1117 uint8_t *extradata = NULL;
1118 int extradata_size = 0;
1119 int extradata_offset = 0;
1121 /* Apply some sanity checks. */
1122 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1123 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1124 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1125 av_log(matroska->ctx, AV_LOG_INFO,
1126 "Unknown or unsupported track type %"PRIu64"\n",
1130 if (track->codec_id == NULL)
1133 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1134 if (!track->default_duration)
1135 track->default_duration = 1000000000/track->video.frame_rate;
1136 if (!track->video.display_width)
1137 track->video.display_width = track->video.pixel_width;
1138 if (!track->video.display_height)
1139 track->video.display_height = track->video.pixel_height;
1140 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1141 if (!track->audio.out_samplerate)
1142 track->audio.out_samplerate = track->audio.samplerate;
1144 if (encodings_list->nb_elem > 1) {
1145 av_log(matroska->ctx, AV_LOG_ERROR,
1146 "Multiple combined encodings no supported");
1147 } else if (encodings_list->nb_elem == 1) {
1148 if (encodings[0].type ||
1149 (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1151 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1154 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1156 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
1157 encodings[0].scope = 0;
1158 av_log(matroska->ctx, AV_LOG_ERROR,
1159 "Unsupported encoding type");
1160 } else if (track->codec_priv.size && encodings[0].scope&2) {
1161 uint8_t *codec_priv = track->codec_priv.data;
1162 int offset = matroska_decode_buffer(&track->codec_priv.data,
1163 &track->codec_priv.size,
1166 track->codec_priv.data = NULL;
1167 track->codec_priv.size = 0;
1168 av_log(matroska->ctx, AV_LOG_ERROR,
1169 "Failed to decode codec private data\n");
1170 } else if (offset > 0) {
1171 track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
1172 memcpy(track->codec_priv.data,
1173 encodings[0].compression.settings.data, offset);
1174 memcpy(track->codec_priv.data+offset, codec_priv,
1175 track->codec_priv.size);
1176 track->codec_priv.size += offset;
1178 if (codec_priv != track->codec_priv.data)
1179 av_free(codec_priv);
1183 for(j=0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++){
1184 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1185 strlen(ff_mkv_codec_tags[j].str))){
1186 codec_id= ff_mkv_codec_tags[j].id;
1191 st = track->stream = av_new_stream(s, 0);
1193 return AVERROR(ENOMEM);
1195 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1196 && track->codec_priv.size >= 40
1197 && track->codec_priv.data != NULL) {
1198 track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1199 codec_id = codec_get_id(codec_bmp_tags, track->video.fourcc);
1200 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1201 && track->codec_priv.size >= 18
1202 && track->codec_priv.data != NULL) {
1203 uint16_t tag = AV_RL16(track->codec_priv.data);
1204 codec_id = codec_get_id(codec_wav_tags, tag);
1205 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1206 && (track->codec_priv.size >= 86)
1207 && (track->codec_priv.data != NULL)) {
1208 track->video.fourcc = AV_RL32(track->codec_priv.data);
1209 codec_id=codec_get_id(codec_movvideo_tags, track->video.fourcc);
1210 } else if (codec_id == CODEC_ID_PCM_S16BE) {
1211 switch (track->audio.bitdepth) {
1212 case 8: codec_id = CODEC_ID_PCM_U8; break;
1213 case 24: codec_id = CODEC_ID_PCM_S24BE; break;
1214 case 32: codec_id = CODEC_ID_PCM_S32BE; break;
1216 } else if (codec_id == CODEC_ID_PCM_S16LE) {
1217 switch (track->audio.bitdepth) {
1218 case 8: codec_id = CODEC_ID_PCM_U8; break;
1219 case 24: codec_id = CODEC_ID_PCM_S24LE; break;
1220 case 32: codec_id = CODEC_ID_PCM_S32LE; break;
1222 } else if (codec_id==CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1223 codec_id = CODEC_ID_PCM_F64LE;
1224 } else if (codec_id == CODEC_ID_AAC && !track->codec_priv.size) {
1225 int profile = matroska_aac_profile(track->codec_id);
1226 int sri = matroska_aac_sri(track->audio.samplerate);
1227 extradata = av_malloc(5);
1228 if (extradata == NULL)
1229 return AVERROR(ENOMEM);
1230 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1231 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1232 if (strstr(track->codec_id, "SBR")) {
1233 sri = matroska_aac_sri(track->audio.out_samplerate);
1234 extradata[2] = 0x56;
1235 extradata[3] = 0xE5;
1236 extradata[4] = 0x80 | (sri<<3);
1240 } else if (codec_id == CODEC_ID_TTA) {
1242 extradata_size = 30;
1243 extradata = av_mallocz(extradata_size);
1244 if (extradata == NULL)
1245 return AVERROR(ENOMEM);
1246 init_put_byte(&b, extradata, extradata_size, 1,
1247 NULL, NULL, NULL, NULL);
1248 put_buffer(&b, "TTA1", 4);
1250 put_le16(&b, track->audio.channels);
1251 put_le16(&b, track->audio.bitdepth);
1252 put_le32(&b, track->audio.out_samplerate);
1253 put_le32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1254 } else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
1255 codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
1256 extradata_offset = 26;
1257 track->codec_priv.size -= extradata_offset;
1258 } else if (codec_id == CODEC_ID_RA_144) {
1259 track->audio.out_samplerate = 8000;
1260 track->audio.channels = 1;
1261 } else if (codec_id == CODEC_ID_RA_288 || codec_id == CODEC_ID_COOK ||
1262 codec_id == CODEC_ID_ATRAC3) {
1265 init_put_byte(&b, track->codec_priv.data,track->codec_priv.size,
1266 0, NULL, NULL, NULL, NULL);
1268 track->audio.coded_framesize = get_be32(&b);
1270 track->audio.sub_packet_h = get_be16(&b);
1271 track->audio.frame_size = get_be16(&b);
1272 track->audio.sub_packet_size = get_be16(&b);
1273 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1274 if (codec_id == CODEC_ID_RA_288) {
1275 st->codec->block_align = track->audio.coded_framesize;
1276 track->codec_priv.size = 0;
1278 st->codec->block_align = track->audio.sub_packet_size;
1279 extradata_offset = 78;
1280 track->codec_priv.size -= extradata_offset;
1284 if (codec_id == CODEC_ID_NONE)
1285 av_log(matroska->ctx, AV_LOG_INFO,
1286 "Unknown/unsupported CodecID %s.\n", track->codec_id);
1288 if (track->time_scale < 0.01)
1289 track->time_scale = 1.0;
1290 av_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1292 st->codec->codec_id = codec_id;
1294 if (strcmp(track->language, "und"))
1295 av_strlcpy(st->language, track->language, 4);
1297 if (track->flag_default)
1298 st->disposition |= AV_DISPOSITION_DEFAULT;
1300 if (track->default_duration)
1301 av_reduce(&st->codec->time_base.num, &st->codec->time_base.den,
1302 track->default_duration, 1000000000, 30000);
1305 st->codec->extradata = extradata;
1306 st->codec->extradata_size = extradata_size;
1307 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1308 st->codec->extradata = av_malloc(track->codec_priv.size);
1309 if(st->codec->extradata == NULL)
1310 return AVERROR(ENOMEM);
1311 st->codec->extradata_size = track->codec_priv.size;
1312 memcpy(st->codec->extradata,
1313 track->codec_priv.data + extradata_offset,
1314 track->codec_priv.size);
1317 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1318 st->codec->codec_type = CODEC_TYPE_VIDEO;
1319 st->codec->codec_tag = track->video.fourcc;
1320 st->codec->width = track->video.pixel_width;
1321 st->codec->height = track->video.pixel_height;
1322 av_reduce(&st->sample_aspect_ratio.num,
1323 &st->sample_aspect_ratio.den,
1324 st->codec->height * track->video.display_width,
1325 st->codec-> width * track->video.display_height,
1327 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1328 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1329 st->codec->codec_type = CODEC_TYPE_AUDIO;
1330 st->codec->sample_rate = track->audio.out_samplerate;
1331 st->codec->channels = track->audio.channels;
1332 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1333 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
1337 attachements = attachements_list->elem;
1338 for (j=0; j<attachements_list->nb_elem; j++) {
1339 if (!(attachements[j].filename && attachements[j].mime &&
1340 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1341 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1343 AVStream *st = av_new_stream(s, 0);
1346 st->filename = av_strdup(attachements[j].filename);
1347 st->codec->codec_id = CODEC_ID_NONE;
1348 st->codec->codec_type = CODEC_TYPE_ATTACHMENT;
1349 st->codec->extradata = av_malloc(attachements[j].bin.size);
1350 if(st->codec->extradata == NULL)
1352 st->codec->extradata_size = attachements[j].bin.size;
1353 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1355 for (i=0; ff_mkv_mime_tags[i].id != CODEC_ID_NONE; i++) {
1356 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1357 strlen(ff_mkv_mime_tags[i].str))) {
1358 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1365 chapters = chapters_list->elem;
1366 for (i=0; i<chapters_list->nb_elem; i++)
1367 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid)
1368 ff_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1369 chapters[i].start, chapters[i].end,
1372 index_list = &matroska->index;
1373 index = index_list->elem;
1374 for (i=0; i<index_list->nb_elem; i++) {
1375 EbmlList *pos_list = &index[i].pos;
1376 MatroskaIndexPos *pos = pos_list->elem;
1377 for (j=0; j<pos_list->nb_elem; j++) {
1378 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1380 if (track && track->stream)
1381 av_add_index_entry(track->stream,
1382 pos[j].pos + matroska->segment_start,
1383 index[i].time, 0, 0, AVINDEX_KEYFRAME);
1391 * Put one packet in an application-supplied AVPacket struct.
1392 * Returns 0 on success or -1 on failure.
1394 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1397 if (matroska->num_packets > 0) {
1398 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1399 av_free(matroska->packets[0]);
1400 if (matroska->num_packets > 1) {
1401 memmove(&matroska->packets[0], &matroska->packets[1],
1402 (matroska->num_packets - 1) * sizeof(AVPacket *));
1404 av_realloc(matroska->packets, (matroska->num_packets - 1) *
1405 sizeof(AVPacket *));
1407 av_freep(&matroska->packets);
1409 matroska->num_packets--;
1417 * Free all packets in our internal queue.
1419 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1421 if (matroska->packets) {
1423 for (n = 0; n < matroska->num_packets; n++) {
1424 av_free_packet(matroska->packets[n]);
1425 av_free(matroska->packets[n]);
1427 av_freep(&matroska->packets);
1428 matroska->num_packets = 0;
1432 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
1433 int size, int64_t pos, uint64_t cluster_time,
1434 uint64_t duration, int is_keyframe,
1435 int64_t cluster_pos)
1437 uint64_t timecode = AV_NOPTS_VALUE;
1438 MatroskaTrack *track;
1443 uint32_t *lace_size = NULL;
1444 int n, flags, laces = 0;
1447 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
1448 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
1454 track = matroska_find_track_by_num(matroska, num);
1455 if (size <= 3 || !track || !track->stream) {
1456 av_log(matroska->ctx, AV_LOG_INFO,
1457 "Invalid stream %"PRIu64" or size %u\n", num, size);
1461 if (st->discard >= AVDISCARD_ALL)
1463 if (duration == AV_NOPTS_VALUE)
1464 duration = track->default_duration / matroska->time_scale;
1466 block_time = AV_RB16(data);
1470 if (is_keyframe == -1)
1471 is_keyframe = flags & 0x80 ? PKT_FLAG_KEY : 0;
1473 if (cluster_time != (uint64_t)-1
1474 && (block_time >= 0 || cluster_time >= -block_time)) {
1475 timecode = cluster_time + block_time;
1476 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
1477 && timecode < track->end_timecode)
1478 is_keyframe = 0; /* overlapping subtitles are not key frame */
1480 av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
1481 track->end_timecode = FFMAX(track->end_timecode, timecode+duration);
1484 if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1485 if (!is_keyframe || timecode < matroska->skip_to_timecode)
1487 matroska->skip_to_keyframe = 0;
1490 switch ((flags & 0x06) >> 1) {
1491 case 0x0: /* no lacing */
1493 lace_size = av_mallocz(sizeof(int));
1494 lace_size[0] = size;
1497 case 0x1: /* Xiph lacing */
1498 case 0x2: /* fixed-size lacing */
1499 case 0x3: /* EBML lacing */
1500 assert(size>0); // size <=3 is checked before size-=3 above
1501 laces = (*data) + 1;
1504 lace_size = av_mallocz(laces * sizeof(int));
1506 switch ((flags & 0x06) >> 1) {
1507 case 0x1: /* Xiph lacing */ {
1510 for (n = 0; res == 0 && n < laces - 1; n++) {
1517 lace_size[n] += temp;
1523 total += lace_size[n];
1525 lace_size[n] = size - total;
1529 case 0x2: /* fixed-size lacing */
1530 for (n = 0; n < laces; n++)
1531 lace_size[n] = size / laces;
1534 case 0x3: /* EBML lacing */ {
1536 n = matroska_ebmlnum_uint(matroska, data, size, &num);
1538 av_log(matroska->ctx, AV_LOG_INFO,
1539 "EBML block data error\n");
1544 total = lace_size[0] = num;
1545 for (n = 1; res == 0 && n < laces - 1; n++) {
1548 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1550 av_log(matroska->ctx, AV_LOG_INFO,
1551 "EBML block data error\n");
1556 lace_size[n] = lace_size[n - 1] + snum;
1557 total += lace_size[n];
1559 lace_size[n] = size - total;
1567 for (n = 0; n < laces; n++) {
1568 if (st->codec->codec_id == CODEC_ID_RA_288 ||
1569 st->codec->codec_id == CODEC_ID_COOK ||
1570 st->codec->codec_id == CODEC_ID_ATRAC3) {
1571 int a = st->codec->block_align;
1572 int sps = track->audio.sub_packet_size;
1573 int cfs = track->audio.coded_framesize;
1574 int h = track->audio.sub_packet_h;
1575 int y = track->audio.sub_packet_cnt;
1576 int w = track->audio.frame_size;
1579 if (!track->audio.pkt_cnt) {
1580 if (st->codec->codec_id == CODEC_ID_RA_288)
1581 for (x=0; x<h/2; x++)
1582 memcpy(track->audio.buf+x*2*w+y*cfs,
1585 for (x=0; x<w/sps; x++)
1586 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
1588 if (++track->audio.sub_packet_cnt >= h) {
1589 track->audio.sub_packet_cnt = 0;
1590 track->audio.pkt_cnt = h*w / a;
1593 while (track->audio.pkt_cnt) {
1594 pkt = av_mallocz(sizeof(AVPacket));
1595 av_new_packet(pkt, a);
1596 memcpy(pkt->data, track->audio.buf
1597 + a * (h*w / a - track->audio.pkt_cnt--), a);
1599 pkt->stream_index = st->index;
1600 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1603 MatroskaTrackEncoding *encodings = track->encodings.elem;
1604 int offset = 0, pkt_size = lace_size[n];
1605 uint8_t *pkt_data = data;
1607 if (encodings && encodings->scope & 1) {
1608 offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
1613 pkt = av_mallocz(sizeof(AVPacket));
1614 /* XXX: prevent data copy... */
1615 if (av_new_packet(pkt, pkt_size+offset) < 0) {
1617 res = AVERROR(ENOMEM);
1622 memcpy (pkt->data, encodings->compression.settings.data, offset);
1623 memcpy (pkt->data+offset, pkt_data, pkt_size);
1625 if (pkt_data != data)
1629 pkt->flags = is_keyframe;
1630 pkt->stream_index = st->index;
1632 pkt->pts = timecode;
1634 if (st->codec->codec_id == CODEC_ID_TEXT)
1635 pkt->convergence_duration = duration;
1636 else if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE)
1637 pkt->duration = duration;
1639 if (st->codec->codec_id == CODEC_ID_SSA)
1640 matroska_fix_ass_packet(matroska, pkt, duration);
1642 if (matroska->prev_pkt &&
1643 timecode != AV_NOPTS_VALUE &&
1644 matroska->prev_pkt->pts == timecode &&
1645 matroska->prev_pkt->stream_index == st->index)
1646 matroska_merge_packets(matroska->prev_pkt, pkt);
1648 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1649 matroska->prev_pkt = pkt;
1653 if (timecode != AV_NOPTS_VALUE)
1654 timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
1655 data += lace_size[n];
1663 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
1665 MatroskaCluster cluster = { 0 };
1666 EbmlList *blocks_list;
1667 MatroskaBlock *blocks;
1669 int64_t pos = url_ftell(matroska->ctx->pb);
1670 matroska->prev_pkt = NULL;
1671 if (matroska->has_cluster_id){
1672 /* For the first cluster we parse, its ID was already read as
1673 part of matroska_read_header(), so don't read it again */
1674 res = ebml_parse_id(matroska, matroska_clusters,
1675 MATROSKA_ID_CLUSTER, &cluster);
1676 pos -= 4; /* sizeof the ID which was already read */
1677 matroska->has_cluster_id = 0;
1679 res = ebml_parse(matroska, matroska_clusters, &cluster);
1680 blocks_list = &cluster.blocks;
1681 blocks = blocks_list->elem;
1682 for (i=0; i<blocks_list->nb_elem; i++)
1683 if (blocks[i].bin.size > 0)
1684 res=matroska_parse_block(matroska,
1685 blocks[i].bin.data, blocks[i].bin.size,
1686 blocks[i].bin.pos, cluster.timecode,
1687 blocks[i].duration, !blocks[i].reference,
1689 ebml_free(matroska_cluster, &cluster);
1690 if (res < 0) matroska->done = 1;
1694 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
1696 MatroskaDemuxContext *matroska = s->priv_data;
1698 while (matroska_deliver_packet(matroska, pkt)) {
1700 return AVERROR(EIO);
1701 matroska_parse_cluster(matroska);
1707 static int matroska_read_seek(AVFormatContext *s, int stream_index,
1708 int64_t timestamp, int flags)
1710 MatroskaDemuxContext *matroska = s->priv_data;
1711 MatroskaTrack *tracks = matroska->tracks.elem;
1712 AVStream *st = s->streams[stream_index];
1713 int i, index, index_sub, index_min;
1715 if (!st->nb_index_entries)
1717 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
1719 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1720 url_fseek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
1721 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1722 matroska_clear_queue(matroska);
1723 if (matroska_parse_cluster(matroska) < 0)
1728 matroska_clear_queue(matroska);
1733 for (i=0; i < matroska->tracks.nb_elem; i++) {
1734 tracks[i].end_timecode = 0;
1735 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
1736 && !tracks[i].stream->discard != AVDISCARD_ALL) {
1737 index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
1739 && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
1740 && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
1741 index_min = index_sub;
1745 url_fseek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
1746 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
1747 matroska->skip_to_timecode = st->index_entries[index].timestamp;
1749 av_update_cur_dts(s, st, st->index_entries[index].timestamp);
1753 static int matroska_read_close(AVFormatContext *s)
1755 MatroskaDemuxContext *matroska = s->priv_data;
1756 MatroskaTrack *tracks = matroska->tracks.elem;
1759 matroska_clear_queue(matroska);
1761 for (n=0; n < matroska->tracks.nb_elem; n++)
1762 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
1763 av_free(tracks[n].audio.buf);
1764 ebml_free(matroska_segment, matroska);
1769 AVInputFormat matroska_demuxer = {
1771 NULL_IF_CONFIG_SMALL("Matroska file format"),
1772 sizeof(MatroskaDemuxContext),
1774 matroska_read_header,
1775 matroska_read_packet,
1776 matroska_read_close,