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;
148 } MatroskaAttachement;
184 AVFormatContext *ctx;
188 MatroskaLevel levels[EBML_MAX_DEPTH];
195 EbmlList attachments;
201 /* byte position of the segment inside the stream */
202 offset_t segment_start;
204 /* the packet queue */
211 /* What to skip before effectively reading a packet. */
212 int skip_to_keyframe;
213 AVStream *skip_to_stream;
214 } MatroskaDemuxContext;
227 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(*x))
229 static EbmlSyntax ebml_header[] = {
230 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
231 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
232 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
233 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
234 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
235 { EBML_ID_EBMLVERSION, EBML_NONE },
236 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
240 static EbmlSyntax ebml_syntax[] = {
241 { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
245 static EbmlSyntax matroska_info[] = {
246 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
247 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
248 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
249 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
250 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
251 { MATROSKA_ID_DATEUTC, EBML_NONE },
252 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
256 static EbmlSyntax matroska_track_video[] = {
257 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
258 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
259 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
260 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
261 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
262 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
263 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
264 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
265 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
266 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
267 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
268 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
269 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_NONE },
270 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
274 static EbmlSyntax matroska_track_audio[] = {
275 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
276 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
277 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
278 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
282 static EbmlSyntax matroska_track_encoding_compression[] = {
283 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
284 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
288 static EbmlSyntax matroska_track_encoding[] = {
289 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
290 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
291 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
292 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
296 static EbmlSyntax matroska_track_encodings[] = {
297 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
301 static EbmlSyntax matroska_track[] = {
302 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
303 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
304 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
305 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
306 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
307 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
308 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
309 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
310 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
311 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
312 { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
313 { MATROSKA_ID_TRACKUID, EBML_NONE },
314 { MATROSKA_ID_TRACKNAME, EBML_NONE },
315 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
316 { MATROSKA_ID_TRACKFLAGFORCED, EBML_NONE },
317 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
318 { MATROSKA_ID_CODECNAME, EBML_NONE },
319 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
320 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
321 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
322 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
323 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
324 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
328 static EbmlSyntax matroska_tracks[] = {
329 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
333 static EbmlSyntax matroska_attachment[] = {
334 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
335 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
336 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
337 { MATROSKA_ID_FILEDESC, EBML_NONE },
338 { MATROSKA_ID_FILEUID, EBML_NONE },
342 static EbmlSyntax matroska_attachments[] = {
343 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
347 static EbmlSyntax matroska_chapter_display[] = {
348 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
349 { MATROSKA_ID_CHAPLANG, EBML_NONE },
353 static EbmlSyntax matroska_chapter_entry[] = {
354 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
355 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
356 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
357 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
358 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
359 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
360 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
361 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
365 static EbmlSyntax matroska_chapter[] = {
366 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
367 { MATROSKA_ID_EDITIONUID, EBML_NONE },
368 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
369 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
370 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
374 static EbmlSyntax matroska_chapters[] = {
375 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
379 static EbmlSyntax matroska_index_pos[] = {
380 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
381 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
382 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
386 static EbmlSyntax matroska_index_entry[] = {
387 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
388 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
392 static EbmlSyntax matroska_index[] = {
393 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
397 static EbmlSyntax matroska_simpletag[] = {
398 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
399 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
400 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
401 { MATROSKA_ID_TAGLANG, EBML_NONE },
402 { MATROSKA_ID_TAGDEFAULT, EBML_NONE },
406 static EbmlSyntax matroska_tag[] = {
407 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), 0, {.n=matroska_simpletag} },
408 { MATROSKA_ID_TAGTARGETS, EBML_NONE },
412 static EbmlSyntax matroska_tags[] = {
413 { MATROSKA_ID_TAG, EBML_NEST, 0, offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
417 static EbmlSyntax matroska_seekhead_entry[] = {
418 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
419 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
423 static EbmlSyntax matroska_seekhead[] = {
424 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
428 static EbmlSyntax matroska_segment[] = {
429 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
430 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
431 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
432 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
433 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
434 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
435 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
436 { MATROSKA_ID_CLUSTER, EBML_STOP, 0, offsetof(MatroskaDemuxContext,has_cluster_id) },
440 static EbmlSyntax matroska_segments[] = {
441 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
445 static EbmlSyntax matroska_blockgroup[] = {
446 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
447 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
448 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration), {.u=AV_NOPTS_VALUE} },
449 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
453 static EbmlSyntax matroska_cluster[] = {
454 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
455 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
456 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
457 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
458 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
462 static EbmlSyntax matroska_clusters[] = {
463 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
464 { MATROSKA_ID_INFO, EBML_NONE },
465 { MATROSKA_ID_CUES, EBML_NONE },
466 { MATROSKA_ID_TAGS, EBML_NONE },
467 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
471 #define SIZE_OFF(x) sizeof(((AVFormatContext*)0)->x),offsetof(AVFormatContext,x)
477 { "TITLE", SIZE_OFF(title) },
478 { "ARTIST", SIZE_OFF(author) },
479 { "WRITTEN_BY", SIZE_OFF(author) },
480 { "LEAD_PERFORMER", SIZE_OFF(author) },
481 { "COPYRIGHT", SIZE_OFF(copyright) },
482 { "COMMENT", SIZE_OFF(comment) },
483 { "ALBUM", SIZE_OFF(album) },
484 { "DATE_WRITTEN", SIZE_OFF(year) },
485 { "DATE_RELEASED", SIZE_OFF(year) },
486 { "PART_NUMBER", SIZE_OFF(track) },
487 { "GENRE", SIZE_OFF(genre) },
491 * Return: Whether we reached the end of a level in the hierarchy or not.
493 static int ebml_level_end(MatroskaDemuxContext *matroska)
495 ByteIOContext *pb = matroska->ctx->pb;
496 offset_t pos = url_ftell(pb);
498 if (matroska->num_levels > 0) {
499 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
500 if (pos - level->start >= level->length) {
501 matroska->num_levels--;
509 * Read: an "EBML number", which is defined as a variable-length
510 * array of bytes. The first byte indicates the length by giving a
511 * number of 0-bits followed by a one. The position of the first
512 * "one" bit inside the first byte indicates the length of this
514 * Returns: number of bytes read, < 0 on error
516 static int ebml_read_num(MatroskaDemuxContext *matroska, ByteIOContext *pb,
517 int max_size, uint64_t *number)
519 int len_mask = 0x80, read = 1, n = 1;
522 /* The first byte tells us the length in bytes - get_byte() can normally
523 * return 0, but since that's not a valid first ebmlID byte, we can
524 * use it safely here to catch EOS. */
525 if (!(total = get_byte(pb))) {
526 /* we might encounter EOS here */
528 offset_t pos = url_ftell(pb);
529 av_log(matroska->ctx, AV_LOG_ERROR,
530 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
533 return AVERROR(EIO); /* EOS or actual I/O error */
536 /* get the length of the EBML number */
537 while (read <= max_size && !(total & len_mask)) {
541 if (read > max_size) {
542 offset_t pos = url_ftell(pb) - 1;
543 av_log(matroska->ctx, AV_LOG_ERROR,
544 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
545 (uint8_t) total, pos, pos);
546 return AVERROR_INVALIDDATA;
549 /* read out length */
552 total = (total << 8) | get_byte(pb);
560 * Read the next element as an unsigned int.
561 * 0 is success, < 0 is failure.
563 static int ebml_read_uint(ByteIOContext *pb, int size, uint64_t *num)
567 if (size < 1 || size > 8)
568 return AVERROR_INVALIDDATA;
570 /* big-endian ordering; build up number */
573 *num = (*num << 8) | get_byte(pb);
579 * Read the next element as a float.
580 * 0 is success, < 0 is failure.
582 static int ebml_read_float(ByteIOContext *pb, int size, double *num)
585 *num= av_int2flt(get_be32(pb));
587 *num= av_int2dbl(get_be64(pb));
589 return AVERROR_INVALIDDATA;
595 * Read the next element as an ASCII string.
596 * 0 is success, < 0 is failure.
598 static int ebml_read_ascii(ByteIOContext *pb, int size, char **str)
601 /* EBML strings are usually not 0-terminated, so we allocate one
602 * byte more, read the string and NULL-terminate it ourselves. */
603 if (!(*str = av_malloc(size + 1)))
604 return AVERROR(ENOMEM);
605 if (get_buffer(pb, (uint8_t *) *str, size) != size) {
615 * Read the next element as binary data.
616 * 0 is success, < 0 is failure.
618 static int ebml_read_binary(ByteIOContext *pb, int length, EbmlBin *bin)
621 if (!(bin->data = av_malloc(length)))
622 return AVERROR(ENOMEM);
625 bin->pos = url_ftell(pb);
626 if (get_buffer(pb, bin->data, length) != length)
633 * Read the next element, but only the header. The contents
634 * are supposed to be sub-elements which can be read separately.
635 * 0 is success, < 0 is failure.
637 static int ebml_read_master(MatroskaDemuxContext *matroska, int length)
639 ByteIOContext *pb = matroska->ctx->pb;
640 MatroskaLevel *level;
642 if (matroska->num_levels >= EBML_MAX_DEPTH) {
643 av_log(matroska->ctx, AV_LOG_ERROR,
644 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
645 return AVERROR(ENOSYS);
648 level = &matroska->levels[matroska->num_levels++];
649 level->start = url_ftell(pb);
650 level->length = length;
656 * Read signed/unsigned "EBML" numbers.
657 * Return: number of bytes processed, < 0 on error
659 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
660 uint8_t *data, uint32_t size, uint64_t *num)
663 init_put_byte(&pb, data, size, 0, NULL, NULL, NULL, NULL);
664 return ebml_read_num(matroska, &pb, 8, num);
668 * Same as above, but signed.
670 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
671 uint8_t *data, uint32_t size, int64_t *num)
676 /* read as unsigned number first */
677 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
680 /* make signed (weird way) */
681 *num = unum - ((1LL << (7*res - 1)) - 1);
686 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
687 EbmlSyntax *syntax, void *data);
689 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
690 uint32_t id, void *data)
693 for (i=0; syntax[i].id; i++)
694 if (id == syntax[i].id)
696 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32)
697 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
698 return ebml_parse_elem(matroska, &syntax[i], data);
701 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
705 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
707 return res < 0 ? res : ebml_parse_id(matroska, syntax, id, data);
710 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
715 for (i=0; syntax[i].id; i++)
716 switch (syntax[i].type) {
718 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
721 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
725 *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
729 while (!res && !ebml_level_end(matroska))
730 res = ebml_parse(matroska, syntax, data);
735 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
736 EbmlSyntax *syntax, void *data)
738 ByteIOContext *pb = matroska->ctx->pb;
739 uint32_t id = syntax->id;
743 data = (char *)data + syntax->data_offset;
744 if (syntax->list_elem_size) {
745 EbmlList *list = data;
746 list->elem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
747 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
748 memset(data, 0, syntax->list_elem_size);
752 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP)
753 if ((res = ebml_read_num(matroska, pb, 8, &length)) < 0)
756 switch (syntax->type) {
757 case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
758 case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
760 case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
761 case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
762 case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
764 if (id == MATROSKA_ID_SEGMENT)
765 matroska->segment_start = url_ftell(matroska->ctx->pb);
766 return ebml_parse_nest(matroska, syntax->def.n, data);
767 case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
768 case EBML_STOP: *(int *)data = 1; return 1;
769 default: return url_fseek(pb,length,SEEK_CUR)<0 ? AVERROR(EIO) : 0;
771 if (res == AVERROR_INVALIDDATA)
772 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
773 else if (res == AVERROR(EIO))
774 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
778 static void ebml_free(EbmlSyntax *syntax, void *data)
781 for (i=0; syntax[i].id; i++) {
782 void *data_off = (char *)data + syntax[i].data_offset;
783 switch (syntax[i].type) {
785 case EBML_UTF8: av_freep(data_off); break;
786 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
788 if (syntax[i].list_elem_size) {
789 EbmlList *list = data_off;
790 char *ptr = list->elem;
791 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
792 ebml_free(syntax[i].def.n, ptr);
795 ebml_free(syntax[i].def.n, data_off);
805 static int matroska_probe(AVProbeData *p)
808 int len_mask = 0x80, size = 1, n = 1;
809 static const char probe_data[] = "matroska";
812 if (AV_RB32(p->buf) != EBML_ID_HEADER)
815 /* length of header */
817 while (size <= 8 && !(total & len_mask)) {
823 total &= (len_mask - 1);
825 total = (total << 8) | p->buf[4 + n++];
827 /* Does the probe data contain the whole header? */
828 if (p->buf_size < 4 + size + total)
831 /* The header must contain the document type 'matroska'. For now,
832 * we don't parse the whole header but simply check for the
833 * availability of that array of characters inside the header.
834 * Not fully fool-proof, but good enough. */
835 for (n = 4+size; n <= 4+size+total-(sizeof(probe_data)-1); n++)
836 if (!memcmp(p->buf+n, probe_data, sizeof(probe_data)-1))
837 return AVPROBE_SCORE_MAX;
842 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
845 MatroskaTrack *tracks = matroska->tracks.elem;
848 for (i=0; i < matroska->tracks.nb_elem; i++)
849 if (tracks[i].num == num)
852 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
856 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
857 MatroskaTrack *track)
859 MatroskaTrackEncoding *encodings = track->encodings.elem;
860 uint8_t* data = *buf;
861 int isize = *buf_size;
862 uint8_t* pkt_data = NULL;
863 int pkt_size = isize;
867 switch (encodings[0].compression.algo) {
868 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
869 return encodings[0].compression.settings.size;
870 case MATROSKA_TRACK_ENCODING_COMP_LZO:
872 olen = pkt_size *= 3;
873 pkt_data = av_realloc(pkt_data,
874 pkt_size+LZO_OUTPUT_PADDING);
875 result = lzo1x_decode(pkt_data, &olen, data, &isize);
876 } while (result==LZO_OUTPUT_FULL && pkt_size<10000000);
882 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
883 z_stream zstream = {0};
884 if (inflateInit(&zstream) != Z_OK)
886 zstream.next_in = data;
887 zstream.avail_in = isize;
890 pkt_data = av_realloc(pkt_data, pkt_size);
891 zstream.avail_out = pkt_size - zstream.total_out;
892 zstream.next_out = pkt_data + zstream.total_out;
893 result = inflate(&zstream, Z_NO_FLUSH);
894 } while (result==Z_OK && pkt_size<10000000);
895 pkt_size = zstream.total_out;
896 inflateEnd(&zstream);
897 if (result != Z_STREAM_END)
903 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
904 bz_stream bzstream = {0};
905 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
907 bzstream.next_in = data;
908 bzstream.avail_in = isize;
911 pkt_data = av_realloc(pkt_data, pkt_size);
912 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
913 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
914 result = BZ2_bzDecompress(&bzstream);
915 } while (result==BZ_OK && pkt_size<10000000);
916 pkt_size = bzstream.total_out_lo32;
917 BZ2_bzDecompressEnd(&bzstream);
918 if (result != BZ_STREAM_END)
928 *buf_size = pkt_size;
935 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
938 char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
939 for (; *ptr!=',' && ptr<end-1; ptr++);
942 for (; *ptr!=',' && ptr<end-1; ptr++);
944 int64_t end_pts = pkt->pts + pkt->convergence_duration;
945 int sc = matroska->time_scale * pkt->pts / 10000000;
946 int ec = matroska->time_scale * end_pts / 10000000;
947 int sh, sm, ss, eh, em, es, len;
948 sh = sc/360000; sc -= 360000*sh;
949 sm = sc/ 6000; sc -= 6000*sm;
950 ss = sc/ 100; sc -= 100*ss;
951 eh = ec/360000; ec -= 360000*eh;
952 em = ec/ 6000; ec -= 6000*em;
953 es = ec/ 100; ec -= 100*es;
955 len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
956 if (!(line = av_malloc(len)))
958 snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s",
959 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
962 pkt->size = strlen(line);
966 static void matroska_convert_tags(AVFormatContext *s, EbmlList *list)
968 MatroskaTag *tags = list->elem;
971 for (i=0; i < list->nb_elem; i++) {
972 for (j=0; j < ARRAY_SIZE(metadata); j++){
973 if (!strcmp(tags[i].name, metadata[j].name)) {
974 int *ptr = (int *)((char *)s + metadata[j].offset);
976 if (metadata[j].size > sizeof(int))
977 av_strlcpy((char *)ptr, tags[i].string, metadata[j].size);
979 *ptr = atoi(tags[i].string);
982 if (tags[i].sub.nb_elem)
983 matroska_convert_tags(s, &tags[i].sub);
987 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
989 EbmlList *seekhead_list = &matroska->seekhead;
990 MatroskaSeekhead *seekhead = seekhead_list->elem;
991 uint32_t level_up = matroska->level_up;
992 offset_t before_pos = url_ftell(matroska->ctx->pb);
996 for (i=0; i<seekhead_list->nb_elem; i++) {
997 offset_t offset = seekhead[i].pos + matroska->segment_start;
999 if (seekhead[i].pos <= before_pos
1000 || seekhead[i].id == MATROSKA_ID_SEEKHEAD
1001 || seekhead[i].id == MATROSKA_ID_CLUSTER)
1005 if (url_fseek(matroska->ctx->pb, offset, SEEK_SET) != offset)
1008 /* We don't want to lose our seekhead level, so we add
1009 * a dummy. This is a crude hack. */
1010 if (matroska->num_levels == EBML_MAX_DEPTH) {
1011 av_log(matroska->ctx, AV_LOG_INFO,
1012 "Max EBML element depth (%d) reached, "
1013 "cannot parse further.\n", EBML_MAX_DEPTH);
1018 level.length = (uint64_t)-1;
1019 matroska->levels[matroska->num_levels] = level;
1020 matroska->num_levels++;
1022 ebml_parse(matroska, matroska_segment, matroska);
1024 /* remove dummy level */
1025 while (matroska->num_levels) {
1026 uint64_t length = matroska->levels[--matroska->num_levels].length;
1027 if (length == (uint64_t)-1)
1033 url_fseek(matroska->ctx->pb, before_pos, SEEK_SET);
1034 matroska->level_up = level_up;
1037 static int matroska_aac_profile(char *codec_id)
1039 static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1042 for (profile=0; profile<ARRAY_SIZE(aac_profiles); profile++)
1043 if (strstr(codec_id, aac_profiles[profile]))
1048 static int matroska_aac_sri(int samplerate)
1052 for (sri=0; sri<ARRAY_SIZE(ff_mpeg4audio_sample_rates); sri++)
1053 if (ff_mpeg4audio_sample_rates[sri] == samplerate)
1058 static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
1060 MatroskaDemuxContext *matroska = s->priv_data;
1061 EbmlList *attachements_list = &matroska->attachments;
1062 MatroskaAttachement *attachements;
1063 EbmlList *chapters_list = &matroska->chapters;
1064 MatroskaChapter *chapters;
1065 MatroskaTrack *tracks;
1066 EbmlList *index_list;
1067 MatroskaIndex *index;
1074 /* First read the EBML header. */
1075 if (ebml_parse(matroska, ebml_syntax, &ebml)
1076 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1077 || ebml.id_length > sizeof(uint32_t) || strcmp(ebml.doctype, "matroska")
1078 || ebml.doctype_version > 2) {
1079 av_log(matroska->ctx, AV_LOG_ERROR,
1080 "EBML header using unsupported features\n"
1081 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1082 ebml.version, ebml.doctype, ebml.doctype_version);
1083 return AVERROR_NOFMT;
1085 ebml_free(ebml_syntax, &ebml);
1087 /* The next thing is a segment. */
1088 if (ebml_parse(matroska, matroska_segments, matroska) < 0)
1090 matroska_execute_seekhead(matroska);
1092 if (matroska->duration)
1093 matroska->ctx->duration = matroska->duration * matroska->time_scale
1094 * 1000 / AV_TIME_BASE;
1095 if (matroska->title)
1096 strncpy(matroska->ctx->title, matroska->title,
1097 sizeof(matroska->ctx->title)-1);
1098 matroska_convert_tags(s, &matroska->tags);
1100 tracks = matroska->tracks.elem;
1101 for (i=0; i < matroska->tracks.nb_elem; i++) {
1102 MatroskaTrack *track = &tracks[i];
1103 enum CodecID codec_id = CODEC_ID_NONE;
1104 EbmlList *encodings_list = &tracks->encodings;
1105 MatroskaTrackEncoding *encodings = encodings_list->elem;
1106 uint8_t *extradata = NULL;
1107 int extradata_size = 0;
1108 int extradata_offset = 0;
1110 /* Apply some sanity checks. */
1111 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1112 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1113 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1114 av_log(matroska->ctx, AV_LOG_INFO,
1115 "Unknown or unsupported track type %"PRIu64"\n",
1119 if (track->codec_id == NULL)
1122 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1123 if (!track->default_duration)
1124 track->default_duration = 1000000000/track->video.frame_rate;
1125 if (!track->video.display_width)
1126 track->video.display_width = track->video.pixel_width;
1127 if (!track->video.display_height)
1128 track->video.display_height = track->video.pixel_height;
1129 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1130 if (!track->audio.out_samplerate)
1131 track->audio.out_samplerate = track->audio.samplerate;
1133 if (encodings_list->nb_elem > 1) {
1134 av_log(matroska->ctx, AV_LOG_ERROR,
1135 "Multiple combined encodings no supported");
1136 } else if (encodings_list->nb_elem == 1) {
1137 if (encodings[0].type ||
1138 (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1140 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1143 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1145 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
1146 encodings[0].scope = 0;
1147 av_log(matroska->ctx, AV_LOG_ERROR,
1148 "Unsupported encoding type");
1149 } else if (track->codec_priv.size && encodings[0].scope&2) {
1150 uint8_t *codec_priv = track->codec_priv.data;
1151 int offset = matroska_decode_buffer(&track->codec_priv.data,
1152 &track->codec_priv.size,
1155 track->codec_priv.data = NULL;
1156 track->codec_priv.size = 0;
1157 av_log(matroska->ctx, AV_LOG_ERROR,
1158 "Failed to decode codec private data\n");
1159 } else if (offset > 0) {
1160 track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
1161 memcpy(track->codec_priv.data,
1162 encodings[0].compression.settings.data, offset);
1163 memcpy(track->codec_priv.data+offset, codec_priv,
1164 track->codec_priv.size);
1165 track->codec_priv.size += offset;
1167 if (codec_priv != track->codec_priv.data)
1168 av_free(codec_priv);
1172 for(j=0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++){
1173 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1174 strlen(ff_mkv_codec_tags[j].str))){
1175 codec_id= ff_mkv_codec_tags[j].id;
1180 st = track->stream = av_new_stream(s, 0);
1182 return AVERROR(ENOMEM);
1184 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1185 && track->codec_priv.size >= 40
1186 && track->codec_priv.data != NULL) {
1187 track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1188 codec_id = codec_get_id(codec_bmp_tags, track->video.fourcc);
1189 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1190 && track->codec_priv.size >= 18
1191 && track->codec_priv.data != NULL) {
1192 uint16_t tag = AV_RL16(track->codec_priv.data);
1193 codec_id = codec_get_id(codec_wav_tags, tag);
1194 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1195 && (track->codec_priv.size >= 86)
1196 && (track->codec_priv.data != NULL)) {
1197 track->video.fourcc = AV_RL32(track->codec_priv.data);
1198 codec_id=codec_get_id(codec_movvideo_tags, track->video.fourcc);
1199 } else if (codec_id == CODEC_ID_PCM_S16BE) {
1200 switch (track->audio.bitdepth) {
1201 case 8: codec_id = CODEC_ID_PCM_U8; break;
1202 case 24: codec_id = CODEC_ID_PCM_S24BE; break;
1203 case 32: codec_id = CODEC_ID_PCM_S32BE; break;
1205 } else if (codec_id == CODEC_ID_PCM_S16LE) {
1206 switch (track->audio.bitdepth) {
1207 case 8: codec_id = CODEC_ID_PCM_U8; break;
1208 case 24: codec_id = CODEC_ID_PCM_S24LE; break;
1209 case 32: codec_id = CODEC_ID_PCM_S32LE; break;
1211 } else if (codec_id==CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1212 codec_id = CODEC_ID_PCM_F64LE;
1213 } else if (codec_id == CODEC_ID_AAC && !track->codec_priv.size) {
1214 int profile = matroska_aac_profile(track->codec_id);
1215 int sri = matroska_aac_sri(track->audio.samplerate);
1216 extradata = av_malloc(5);
1217 if (extradata == NULL)
1218 return AVERROR(ENOMEM);
1219 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1220 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1221 if (strstr(track->codec_id, "SBR")) {
1222 sri = matroska_aac_sri(track->audio.out_samplerate);
1223 extradata[2] = 0x56;
1224 extradata[3] = 0xE5;
1225 extradata[4] = 0x80 | (sri<<3);
1229 } else if (codec_id == CODEC_ID_TTA) {
1231 extradata_size = 30;
1232 extradata = av_mallocz(extradata_size);
1233 if (extradata == NULL)
1234 return AVERROR(ENOMEM);
1235 init_put_byte(&b, extradata, extradata_size, 1,
1236 NULL, NULL, NULL, NULL);
1237 put_buffer(&b, "TTA1", 4);
1239 put_le16(&b, track->audio.channels);
1240 put_le16(&b, track->audio.bitdepth);
1241 put_le32(&b, track->audio.out_samplerate);
1242 put_le32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1243 } else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
1244 codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
1245 extradata_offset = 26;
1246 track->codec_priv.size -= extradata_offset;
1247 } else if (codec_id == CODEC_ID_RA_144) {
1248 track->audio.out_samplerate = 8000;
1249 track->audio.channels = 1;
1250 } else if (codec_id == CODEC_ID_RA_288 || codec_id == CODEC_ID_COOK ||
1251 codec_id == CODEC_ID_ATRAC3) {
1254 init_put_byte(&b, track->codec_priv.data,track->codec_priv.size,
1255 0, NULL, NULL, NULL, NULL);
1257 track->audio.coded_framesize = get_be32(&b);
1259 track->audio.sub_packet_h = get_be16(&b);
1260 track->audio.frame_size = get_be16(&b);
1261 track->audio.sub_packet_size = get_be16(&b);
1262 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1263 if (codec_id == CODEC_ID_RA_288) {
1264 st->codec->block_align = track->audio.coded_framesize;
1265 track->codec_priv.size = 0;
1267 st->codec->block_align = track->audio.sub_packet_size;
1268 extradata_offset = 78;
1269 track->codec_priv.size -= extradata_offset;
1273 if (codec_id == CODEC_ID_NONE)
1274 av_log(matroska->ctx, AV_LOG_INFO,
1275 "Unknown/unsupported CodecID %s.\n", track->codec_id);
1277 av_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1279 st->codec->codec_id = codec_id;
1281 if (strcmp(track->language, "und"))
1282 av_strlcpy(st->language, track->language, 4);
1284 if (track->flag_default)
1285 st->disposition |= AV_DISPOSITION_DEFAULT;
1287 if (track->default_duration)
1288 av_reduce(&st->codec->time_base.num, &st->codec->time_base.den,
1289 track->default_duration, 1000000000, 30000);
1292 st->codec->extradata = extradata;
1293 st->codec->extradata_size = extradata_size;
1294 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1295 st->codec->extradata = av_malloc(track->codec_priv.size);
1296 if(st->codec->extradata == NULL)
1297 return AVERROR(ENOMEM);
1298 st->codec->extradata_size = track->codec_priv.size;
1299 memcpy(st->codec->extradata,
1300 track->codec_priv.data + extradata_offset,
1301 track->codec_priv.size);
1304 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1305 st->codec->codec_type = CODEC_TYPE_VIDEO;
1306 st->codec->codec_tag = track->video.fourcc;
1307 st->codec->width = track->video.pixel_width;
1308 st->codec->height = track->video.pixel_height;
1309 av_reduce(&st->sample_aspect_ratio.num,
1310 &st->sample_aspect_ratio.den,
1311 st->codec->height * track->video.display_width,
1312 st->codec-> width * track->video.display_height,
1314 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1315 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1316 st->codec->codec_type = CODEC_TYPE_AUDIO;
1317 st->codec->sample_rate = track->audio.out_samplerate;
1318 st->codec->channels = track->audio.channels;
1319 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1320 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
1324 attachements = attachements_list->elem;
1325 for (j=0; j<attachements_list->nb_elem; j++) {
1326 if (!(attachements[j].filename && attachements[j].mime &&
1327 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1328 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1330 AVStream *st = av_new_stream(s, 0);
1333 st->filename = av_strdup(attachements[j].filename);
1334 st->codec->codec_id = CODEC_ID_NONE;
1335 st->codec->codec_type = CODEC_TYPE_ATTACHMENT;
1336 st->codec->extradata = av_malloc(attachements[j].bin.size);
1337 if(st->codec->extradata == NULL)
1339 st->codec->extradata_size = attachements[j].bin.size;
1340 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1342 for (i=0; ff_mkv_mime_tags[i].id != CODEC_ID_NONE; i++) {
1343 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1344 strlen(ff_mkv_mime_tags[i].str))) {
1345 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1352 chapters = chapters_list->elem;
1353 for (i=0; i<chapters_list->nb_elem; i++)
1354 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid)
1355 ff_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1356 chapters[i].start, chapters[i].end,
1359 index_list = &matroska->index;
1360 index = index_list->elem;
1361 for (i=0; i<index_list->nb_elem; i++) {
1362 EbmlList *pos_list = &index[i].pos;
1363 MatroskaIndexPos *pos = pos_list->elem;
1364 for (j=0; j<pos_list->nb_elem; j++) {
1365 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1367 if (track && track->stream)
1368 av_add_index_entry(track->stream,
1369 pos[j].pos + matroska->segment_start,
1370 index[i].time, 0, 0, AVINDEX_KEYFRAME);
1378 * Put one packet in an application-supplied AVPacket struct.
1379 * Returns 0 on success or -1 on failure.
1381 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1384 if (matroska->num_packets > 0) {
1385 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1386 av_free(matroska->packets[0]);
1387 if (matroska->num_packets > 1) {
1388 memmove(&matroska->packets[0], &matroska->packets[1],
1389 (matroska->num_packets - 1) * sizeof(AVPacket *));
1391 av_realloc(matroska->packets, (matroska->num_packets - 1) *
1392 sizeof(AVPacket *));
1394 av_freep(&matroska->packets);
1396 matroska->num_packets--;
1404 * Free all packets in our internal queue.
1406 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1408 if (matroska->packets) {
1410 for (n = 0; n < matroska->num_packets; n++) {
1411 av_free_packet(matroska->packets[n]);
1412 av_free(matroska->packets[n]);
1414 av_freep(&matroska->packets);
1415 matroska->num_packets = 0;
1419 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
1420 int size, int64_t pos, uint64_t cluster_time,
1421 uint64_t duration, int is_keyframe,
1422 int64_t cluster_pos)
1424 MatroskaTrack *track;
1429 uint32_t *lace_size = NULL;
1430 int n, flags, laces = 0;
1433 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
1434 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
1440 track = matroska_find_track_by_num(matroska, num);
1441 if (size <= 3 || !track || !track->stream) {
1442 av_log(matroska->ctx, AV_LOG_INFO,
1443 "Invalid stream %"PRIu64" or size %u\n", num, size);
1447 if (st->discard >= AVDISCARD_ALL)
1449 if (duration == AV_NOPTS_VALUE)
1450 duration = track->default_duration / matroska->time_scale;
1452 block_time = AV_RB16(data);
1456 if (is_keyframe == -1)
1457 is_keyframe = flags & 0x80 ? PKT_FLAG_KEY : 0;
1459 if (matroska->skip_to_keyframe) {
1460 if (!is_keyframe || st != matroska->skip_to_stream)
1462 matroska->skip_to_keyframe = 0;
1465 switch ((flags & 0x06) >> 1) {
1466 case 0x0: /* no lacing */
1468 lace_size = av_mallocz(sizeof(int));
1469 lace_size[0] = size;
1472 case 0x1: /* Xiph lacing */
1473 case 0x2: /* fixed-size lacing */
1474 case 0x3: /* EBML lacing */
1475 assert(size>0); // size <=3 is checked before size-=3 above
1476 laces = (*data) + 1;
1479 lace_size = av_mallocz(laces * sizeof(int));
1481 switch ((flags & 0x06) >> 1) {
1482 case 0x1: /* Xiph lacing */ {
1485 for (n = 0; res == 0 && n < laces - 1; n++) {
1492 lace_size[n] += temp;
1498 total += lace_size[n];
1500 lace_size[n] = size - total;
1504 case 0x2: /* fixed-size lacing */
1505 for (n = 0; n < laces; n++)
1506 lace_size[n] = size / laces;
1509 case 0x3: /* EBML lacing */ {
1511 n = matroska_ebmlnum_uint(matroska, data, size, &num);
1513 av_log(matroska->ctx, AV_LOG_INFO,
1514 "EBML block data error\n");
1519 total = lace_size[0] = num;
1520 for (n = 1; res == 0 && n < laces - 1; n++) {
1523 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1525 av_log(matroska->ctx, AV_LOG_INFO,
1526 "EBML block data error\n");
1531 lace_size[n] = lace_size[n - 1] + snum;
1532 total += lace_size[n];
1534 lace_size[n] = size - total;
1542 uint64_t timecode = AV_NOPTS_VALUE;
1544 if (cluster_time != (uint64_t)-1
1545 && (block_time >= 0 || cluster_time >= -block_time)) {
1546 timecode = cluster_time + block_time;
1548 av_add_index_entry(st, cluster_pos, timecode,
1549 0, 0, AVINDEX_KEYFRAME);
1552 for (n = 0; n < laces; n++) {
1553 if (st->codec->codec_id == CODEC_ID_RA_288 ||
1554 st->codec->codec_id == CODEC_ID_COOK ||
1555 st->codec->codec_id == CODEC_ID_ATRAC3) {
1556 int a = st->codec->block_align;
1557 int sps = track->audio.sub_packet_size;
1558 int cfs = track->audio.coded_framesize;
1559 int h = track->audio.sub_packet_h;
1560 int y = track->audio.sub_packet_cnt;
1561 int w = track->audio.frame_size;
1564 if (!track->audio.pkt_cnt) {
1565 if (st->codec->codec_id == CODEC_ID_RA_288)
1566 for (x=0; x<h/2; x++)
1567 memcpy(track->audio.buf+x*2*w+y*cfs,
1570 for (x=0; x<w/sps; x++)
1571 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
1573 if (++track->audio.sub_packet_cnt >= h) {
1574 track->audio.sub_packet_cnt = 0;
1575 track->audio.pkt_cnt = h*w / a;
1578 while (track->audio.pkt_cnt) {
1579 pkt = av_mallocz(sizeof(AVPacket));
1580 av_new_packet(pkt, a);
1581 memcpy(pkt->data, track->audio.buf
1582 + a * (h*w / a - track->audio.pkt_cnt--), a);
1584 pkt->stream_index = st->index;
1585 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1588 MatroskaTrackEncoding *encodings = track->encodings.elem;
1589 int offset = 0, pkt_size = lace_size[n];
1590 uint8_t *pkt_data = data;
1592 if (encodings && encodings->scope & 1) {
1593 offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
1598 pkt = av_mallocz(sizeof(AVPacket));
1599 /* XXX: prevent data copy... */
1600 if (av_new_packet(pkt, pkt_size+offset) < 0) {
1602 res = AVERROR(ENOMEM);
1607 memcpy (pkt->data, encodings->compression.settings.data, offset);
1608 memcpy (pkt->data+offset, pkt_data, pkt_size);
1610 if (pkt_data != data)
1614 pkt->flags = is_keyframe;
1615 pkt->stream_index = st->index;
1617 pkt->pts = timecode;
1619 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE)
1620 pkt->convergence_duration = duration;
1622 pkt->duration = duration;
1624 if (st->codec->codec_id == CODEC_ID_SSA)
1625 matroska_fix_ass_packet(matroska, pkt);
1627 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
1630 if (timecode != AV_NOPTS_VALUE)
1631 timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
1632 data += lace_size[n];
1640 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
1642 MatroskaCluster cluster = { 0 };
1643 EbmlList *blocks_list;
1644 MatroskaBlock *blocks;
1646 offset_t pos = url_ftell(matroska->ctx->pb);
1647 if (matroska->has_cluster_id){
1648 /* For the first cluster we parse, its ID was already read as
1649 part of matroska_read_header(), so don't read it again */
1650 res = ebml_parse_id(matroska, matroska_clusters,
1651 MATROSKA_ID_CLUSTER, &cluster);
1652 pos -= 4; /* sizeof the ID which was already read */
1653 matroska->has_cluster_id = 0;
1655 res = ebml_parse(matroska, matroska_clusters, &cluster);
1656 blocks_list = &cluster.blocks;
1657 blocks = blocks_list->elem;
1658 for (i=0; i<blocks_list->nb_elem; i++)
1659 if (blocks[i].bin.size > 0)
1660 res=matroska_parse_block(matroska,
1661 blocks[i].bin.data, blocks[i].bin.size,
1662 blocks[i].bin.pos, cluster.timecode,
1663 blocks[i].duration, !blocks[i].reference,
1665 ebml_free(matroska_cluster, &cluster);
1666 if (res < 0) matroska->done = 1;
1670 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
1672 MatroskaDemuxContext *matroska = s->priv_data;
1674 while (matroska_deliver_packet(matroska, pkt)) {
1676 return AVERROR(EIO);
1677 matroska_parse_cluster(matroska);
1683 static int matroska_read_seek(AVFormatContext *s, int stream_index,
1684 int64_t timestamp, int flags)
1686 MatroskaDemuxContext *matroska = s->priv_data;
1687 AVStream *st = s->streams[stream_index];
1693 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1694 if (st->nb_index_entries)
1695 url_fseek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
1696 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1697 matroska_clear_queue(matroska);
1698 if (matroska_parse_cluster(matroska) < 0)
1703 matroska_clear_queue(matroska);
1707 url_fseek(s->pb, st->index_entries[index].pos, SEEK_SET);
1708 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
1709 matroska->skip_to_stream = st;
1711 av_update_cur_dts(s, st, st->index_entries[index].timestamp);
1715 static int matroska_read_close(AVFormatContext *s)
1717 MatroskaDemuxContext *matroska = s->priv_data;
1718 MatroskaTrack *tracks = matroska->tracks.elem;
1721 matroska_clear_queue(matroska);
1723 for (n=0; n < matroska->tracks.nb_elem; n++)
1724 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
1725 av_free(tracks[n].audio.buf);
1726 ebml_free(matroska_segment, matroska);
1731 AVInputFormat matroska_demuxer = {
1733 NULL_IF_CONFIG_SMALL("Matroska file format"),
1734 sizeof(MatroskaDemuxContext),
1736 matroska_read_header,
1737 matroska_read_packet,
1738 matroska_read_close,