2 * Matroska file demuxer (no muxer yet)
3 * Copyright (c) 2003-2004 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 * Specs available on the matroska project page:
28 * http://www.matroska.org/.
32 /* For codec_get_id(). */
36 #include "libavcodec/mpeg4audio.h"
37 #include "libavutil/intfloat_readwrite.h"
38 #include "libavutil/avstring.h"
39 #include "libavutil/lzo.h"
59 typedef const struct EbmlSyntax {
68 const struct EbmlSyntax *n;
88 uint64_t doctype_version;
94 } MatroskaTrackCompression;
99 MatroskaTrackCompression compression;
100 } MatroskaTrackEncoding;
104 uint64_t display_width;
105 uint64_t display_height;
106 uint64_t pixel_width;
107 uint64_t pixel_height;
109 } MatroskaTrackVideo;
113 double out_samplerate;
117 /* real audio header (extracted from extradata) */
125 } MatroskaTrackAudio;
134 uint64_t default_duration;
135 uint64_t flag_default;
136 MatroskaTrackVideo video;
137 MatroskaTrackAudio audio;
147 } MatroskaAttachement;
177 AVFormatContext *ctx;
181 MatroskaLevel levels[EBML_MAX_DEPTH];
188 EbmlList attachments;
193 /* num_streams is the number of streams that av_new_stream() was called
194 * for ( = that are available to the calling program). */
197 /* cache for ID peeking */
200 /* byte position of the segment inside the stream */
201 offset_t segment_start;
203 /* The packet queue. */
210 /* What to skip before effectively reading a packet. */
211 int skip_to_keyframe;
212 AVStream *skip_to_stream;
213 } MatroskaDemuxContext;
226 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(*x))
228 static EbmlSyntax ebml_header[] = {
229 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
230 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
231 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
232 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
233 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
234 { EBML_ID_EBMLVERSION, EBML_NONE },
235 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
236 { EBML_ID_VOID, 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 },
253 { EBML_ID_VOID, EBML_NONE },
257 static EbmlSyntax matroska_track_video[] = {
258 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
259 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
260 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
261 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
262 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
263 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
264 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
265 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_NONE },
266 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
267 { EBML_ID_VOID, EBML_NONE },
271 static EbmlSyntax matroska_track_audio[] = {
272 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
273 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
274 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
275 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
276 { EBML_ID_VOID, EBML_NONE },
280 static EbmlSyntax matroska_track_encoding_compression[] = {
281 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
282 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
283 { EBML_ID_VOID, EBML_NONE },
287 static EbmlSyntax matroska_track_encoding[] = {
288 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
289 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
290 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
291 { EBML_ID_VOID, EBML_NONE },
295 static EbmlSyntax matroska_track_encodings[] = {
296 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
297 { EBML_ID_VOID, EBML_NONE },
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 { EBML_ID_VOID, EBML_NONE },
328 static EbmlSyntax matroska_tracks[] = {
329 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
330 { EBML_ID_VOID, EBML_NONE },
334 static EbmlSyntax matroska_attachment[] = {
335 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
336 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
337 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
338 { MATROSKA_ID_FILEUID, EBML_NONE },
339 { EBML_ID_VOID, EBML_NONE },
343 static EbmlSyntax matroska_attachments[] = {
344 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
345 { EBML_ID_VOID, EBML_NONE },
349 static EbmlSyntax matroska_chapter_display[] = {
350 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
351 { EBML_ID_VOID, 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 { EBML_ID_VOID, 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 { EBML_ID_VOID, EBML_NONE },
374 static EbmlSyntax matroska_chapters[] = {
375 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
376 { EBML_ID_VOID, EBML_NONE },
380 static EbmlSyntax matroska_index_pos[] = {
381 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
382 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
383 { EBML_ID_VOID, EBML_NONE },
387 static EbmlSyntax matroska_index_entry[] = {
388 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
389 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
390 { EBML_ID_VOID, EBML_NONE },
394 static EbmlSyntax matroska_index[] = {
395 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
396 { EBML_ID_VOID, EBML_NONE },
400 static EbmlSyntax matroska_tags[] = {
401 { EBML_ID_VOID, EBML_NONE },
405 static EbmlSyntax matroska_seekhead_entry[] = {
406 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
407 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
408 { EBML_ID_VOID, EBML_NONE },
412 static EbmlSyntax matroska_seekhead[] = {
413 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
414 { EBML_ID_VOID, EBML_NONE },
418 static EbmlSyntax matroska_segment[] = {
419 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
420 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
421 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
422 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
423 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
424 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
425 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
426 { MATROSKA_ID_CLUSTER, EBML_STOP, 0, offsetof(MatroskaDemuxContext,has_cluster_id) },
427 { EBML_ID_VOID, EBML_NONE },
431 static EbmlSyntax matroska_segments[] = {
432 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
436 static EbmlSyntax matroska_blockgroup[] = {
437 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
438 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
439 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration), {.u=AV_NOPTS_VALUE} },
440 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
441 { EBML_ID_VOID, EBML_NONE },
445 static EbmlSyntax matroska_cluster[] = {
446 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
447 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
448 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
449 { EBML_ID_VOID, EBML_NONE },
453 static EbmlSyntax matroska_clusters[] = {
454 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
459 * Return: the amount of levels in the hierarchy that the
460 * current element lies higher than the previous one.
461 * The opposite isn't done - that's auto-done using master
465 ebml_read_element_level_up (MatroskaDemuxContext *matroska)
467 ByteIOContext *pb = matroska->ctx->pb;
468 offset_t pos = url_ftell(pb);
471 while (matroska->num_levels > 0) {
472 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
474 if (pos >= level->start + level->length) {
475 matroska->num_levels--;
486 * Read: an "EBML number", which is defined as a variable-length
487 * array of bytes. The first byte indicates the length by giving a
488 * number of 0-bits followed by a one. The position of the first
489 * "one" bit inside the first byte indicates the length of this
491 * Returns: num. of bytes read. < 0 on error.
494 ebml_read_num (MatroskaDemuxContext *matroska,
498 ByteIOContext *pb = matroska->ctx->pb;
499 int len_mask = 0x80, read = 1, n = 1;
502 /* the first byte tells us the length in bytes - get_byte() can normally
503 * return 0, but since that's not a valid first ebmlID byte, we can
504 * use it safely here to catch EOS. */
505 if (!(total = get_byte(pb))) {
506 /* we might encounter EOS here */
508 offset_t pos = url_ftell(pb);
509 av_log(matroska->ctx, AV_LOG_ERROR,
510 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
513 return AVERROR(EIO); /* EOS or actual I/O error */
516 /* get the length of the EBML number */
517 while (read <= max_size && !(total & len_mask)) {
521 if (read > max_size) {
522 offset_t pos = url_ftell(pb) - 1;
523 av_log(matroska->ctx, AV_LOG_ERROR,
524 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
525 (uint8_t) total, pos, pos);
526 return AVERROR_INVALIDDATA;
529 /* read out length */
532 total = (total << 8) | get_byte(pb);
540 * Read: the element content data ID.
541 * Return: the number of bytes read or < 0 on error.
544 ebml_read_element_id (MatroskaDemuxContext *matroska,
551 /* if we re-call this, use our cached ID */
552 if (matroska->peek_id != 0) {
553 *id = matroska->peek_id;
557 /* read out the "EBML number", include tag in ID */
558 if ((read = ebml_read_num(matroska, 4, &total)) < 0)
560 *id = matroska->peek_id = total | (1 << (read * 7));
566 * Read: element content length.
567 * Return: the number of bytes read or < 0 on error.
570 ebml_read_element_length (MatroskaDemuxContext *matroska,
573 /* clear cache since we're now beyond that data point */
574 matroska->peek_id = 0;
576 /* read out the "EBML number", include tag in ID */
577 return ebml_read_num(matroska, 8, length);
581 * Return: the ID of the next element, or 0 on error.
582 * Level_up contains the amount of levels that this
583 * next element lies higher than the previous one.
586 ebml_peek_id (MatroskaDemuxContext *matroska,
592 res = ebml_read_element_id(matroska, &id, NULL);
596 if (res > 0 && level_up)
597 *level_up = ebml_read_element_level_up(matroska);
603 * Seek to a given offset.
604 * 0 is success, -1 is failure.
607 ebml_read_seek (MatroskaDemuxContext *matroska,
610 ByteIOContext *pb = matroska->ctx->pb;
612 /* clear ID cache, if any */
613 matroska->peek_id = 0;
615 return (url_fseek(pb, offset, SEEK_SET) == offset) ? 0 : -1;
619 * Skip the next element.
620 * 0 is success, -1 is failure.
623 ebml_read_skip (MatroskaDemuxContext *matroska)
625 ByteIOContext *pb = matroska->ctx->pb;
630 if ((res = ebml_read_element_id(matroska, &id, NULL)) < 0 ||
631 (res = ebml_read_element_length(matroska, &length)) < 0)
634 url_fskip(pb, length);
640 * Read the next element as an unsigned int.
641 * 0 is success, < 0 is failure.
644 ebml_read_uint (MatroskaDemuxContext *matroska,
648 ByteIOContext *pb = matroska->ctx->pb;
649 int n = 0, size, res;
652 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
653 (res = ebml_read_element_length(matroska, &rlength)) < 0)
656 if (size < 1 || size > 8) {
657 offset_t pos = url_ftell(pb);
658 av_log(matroska->ctx, AV_LOG_ERROR,
659 "Invalid uint element size %d at position %"PRId64" (0x%"PRIx64")\n",
661 return AVERROR_INVALIDDATA;
664 /* big-endian ordening; build up number */
667 *num = (*num << 8) | get_byte(pb);
673 * Read the next element as a float.
674 * 0 is success, < 0 is failure.
677 ebml_read_float (MatroskaDemuxContext *matroska,
681 ByteIOContext *pb = matroska->ctx->pb;
685 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
686 (res = ebml_read_element_length(matroska, &rlength)) < 0)
691 *num= av_int2flt(get_be32(pb));
693 *num= av_int2dbl(get_be64(pb));
695 offset_t pos = url_ftell(pb);
696 av_log(matroska->ctx, AV_LOG_ERROR,
697 "Invalid float element size %d at position %"PRIu64" (0x%"PRIx64")\n",
699 return AVERROR_INVALIDDATA;
706 * Read the next element as an ASCII string.
707 * 0 is success, < 0 is failure.
710 ebml_read_ascii (MatroskaDemuxContext *matroska,
714 ByteIOContext *pb = matroska->ctx->pb;
718 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
719 (res = ebml_read_element_length(matroska, &rlength)) < 0)
723 /* ebml strings are usually not 0-terminated, so we allocate one
724 * byte more, read the string and NULL-terminate it ourselves. */
725 if (size < 0 || !(*str = av_malloc(size + 1))) {
726 av_log(matroska->ctx, AV_LOG_ERROR, "Memory allocation failed\n");
727 return AVERROR(ENOMEM);
729 if (get_buffer(pb, (uint8_t *) *str, size) != size) {
730 offset_t pos = url_ftell(pb);
731 av_log(matroska->ctx, AV_LOG_ERROR,
732 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
742 * Read the next element, but only the header. The contents
743 * are supposed to be sub-elements which can be read separately.
744 * 0 is success, < 0 is failure.
747 ebml_read_master (MatroskaDemuxContext *matroska,
750 ByteIOContext *pb = matroska->ctx->pb;
752 MatroskaLevel *level;
755 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
756 (res = ebml_read_element_length(matroska, &length)) < 0)
759 if (matroska->num_levels >= EBML_MAX_DEPTH) {
760 av_log(matroska->ctx, AV_LOG_ERROR,
761 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
762 return AVERROR(ENOSYS);
765 level = &matroska->levels[matroska->num_levels++];
766 level->start = url_ftell(pb);
767 level->length = length;
773 * Read the next element as binary data.
774 * 0 is success, < 0 is failure.
777 ebml_read_binary (MatroskaDemuxContext *matroska,
782 ByteIOContext *pb = matroska->ctx->pb;
786 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
787 (res = ebml_read_element_length(matroska, &rlength)) < 0)
791 if (!(*binary = av_malloc(*size))) {
792 av_log(matroska->ctx, AV_LOG_ERROR,
793 "Memory allocation error\n");
794 return AVERROR(ENOMEM);
797 if (get_buffer(pb, *binary, *size) != *size) {
798 offset_t pos = url_ftell(pb);
799 av_log(matroska->ctx, AV_LOG_ERROR,
800 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
808 * Read signed/unsigned "EBML" numbers.
809 * Return: number of bytes processed, < 0 on error.
810 * XXX: use ebml_read_num().
813 matroska_ebmlnum_uint (uint8_t *data,
817 int len_mask = 0x80, read = 1, n = 1, num_ffs = 0;
821 return AVERROR_INVALIDDATA;
824 while (read <= 8 && !(total & len_mask)) {
829 return AVERROR_INVALIDDATA;
831 if ((total &= (len_mask - 1)) == len_mask - 1)
834 return AVERROR_INVALIDDATA;
838 total = (total << 8) | data[n];
851 * Same as above, but signed.
854 matroska_ebmlnum_sint (uint8_t *data,
861 /* read as unsigned number first */
862 if ((res = matroska_ebmlnum_uint(data, size, &unum)) < 0)
865 /* make signed (weird way) */
866 if (unum == (uint64_t)-1)
869 *num = unum - ((1LL << ((7 * res) - 1)) - 1);
875 static MatroskaTrack *
876 matroska_find_track_by_num (MatroskaDemuxContext *matroska,
879 MatroskaTrack *tracks = matroska->tracks.elem;
882 for (i=0; i < matroska->tracks.nb_elem; i++)
883 if (tracks[i].num == num)
886 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
892 * Put one packet in an application-supplied AVPacket struct.
893 * Returns 0 on success or -1 on failure.
896 matroska_deliver_packet (MatroskaDemuxContext *matroska,
899 if (matroska->num_packets > 0) {
900 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
901 av_free(matroska->packets[0]);
902 if (matroska->num_packets > 1) {
903 memmove(&matroska->packets[0], &matroska->packets[1],
904 (matroska->num_packets - 1) * sizeof(AVPacket *));
906 av_realloc(matroska->packets, (matroska->num_packets - 1) *
909 av_freep(&matroska->packets);
911 matroska->num_packets--;
919 * Put a packet into our internal queue. Will be delivered to the
920 * user/application during the next get_packet() call.
923 matroska_queue_packet (MatroskaDemuxContext *matroska,
927 av_realloc(matroska->packets, (matroska->num_packets + 1) *
929 matroska->packets[matroska->num_packets] = pkt;
930 matroska->num_packets++;
934 * Free all packets in our internal queue.
937 matroska_clear_queue (MatroskaDemuxContext *matroska)
939 if (matroska->packets) {
941 for (n = 0; n < matroska->num_packets; n++) {
942 av_free_packet(matroska->packets[n]);
943 av_free(matroska->packets[n]);
945 av_free(matroska->packets);
946 matroska->packets = NULL;
947 matroska->num_packets = 0;
956 matroska_probe (AVProbeData *p)
959 int len_mask = 0x80, size = 1, n = 1;
960 uint8_t probe_data[] = { 'm', 'a', 't', 'r', 'o', 's', 'k', 'a' };
963 if (AV_RB32(p->buf) != EBML_ID_HEADER)
966 /* length of header */
968 while (size <= 8 && !(total & len_mask)) {
974 total &= (len_mask - 1);
976 total = (total << 8) | p->buf[4 + n++];
978 /* does the probe data contain the whole header? */
979 if (p->buf_size < 4 + size + total)
982 /* the header must contain the document type 'matroska'. For now,
983 * we don't parse the whole header but simply check for the
984 * availability of that array of characters inside the header.
985 * Not fully fool-proof, but good enough. */
986 for (n = 4 + size; n <= 4 + size + total - sizeof(probe_data); n++)
987 if (!memcmp (&p->buf[n], probe_data, sizeof(probe_data)))
988 return AVPROBE_SCORE_MAX;
993 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
994 void *data, uint32_t expected_id, int once);
996 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
997 EbmlSyntax *syntax, void *data)
999 uint32_t id = syntax->id;
1003 data = (char *)data + syntax->data_offset;
1004 if (syntax->list_elem_size) {
1005 EbmlList *list = data;
1006 list->elem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
1007 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
1008 memset(data, 0, syntax->list_elem_size);
1013 switch (syntax->type) {
1014 case EBML_UINT: return ebml_read_uint (matroska, &id, data);
1015 case EBML_FLOAT: return ebml_read_float(matroska, &id, data);
1017 case EBML_UTF8: av_free(*(char **)data);
1018 return ebml_read_ascii(matroska, &id, data);
1019 case EBML_BIN: av_free(bin->data);
1020 bin->pos = url_ftell(matroska->ctx->pb);
1021 return ebml_read_binary(matroska, &id, &bin->data,
1023 case EBML_NEST: if ((res=ebml_read_master(matroska, &id)) < 0)
1025 if (id == MATROSKA_ID_SEGMENT)
1026 matroska->segment_start = url_ftell(matroska->ctx->pb);
1027 return ebml_parse(matroska, syntax->def.n, data, 0, 0);
1028 case EBML_PASS: return ebml_parse(matroska, syntax->def.n, data, 0, 1);
1029 case EBML_STOP: *(int *)data = 1; return 1;
1030 default: return ebml_read_skip(matroska);
1034 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1035 uint32_t id, void *data)
1038 for (i=0; syntax[i].id; i++)
1039 if (id == syntax[i].id)
1042 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
1043 return ebml_parse_elem(matroska, &syntax[i], data);
1046 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1047 void *data, uint32_t expected_id, int once)
1052 for (i=0; syntax[i].id; i++)
1053 switch (syntax[i].type) {
1055 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
1058 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
1062 *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
1067 res = ebml_read_master(matroska, &id);
1068 if (id != expected_id)
1069 return AVERROR_INVALIDDATA;
1070 if (id == MATROSKA_ID_SEGMENT)
1071 matroska->segment_start = url_ftell(matroska->ctx->pb);
1075 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1078 } else if (matroska->level_up) {
1079 matroska->level_up--;
1083 res = ebml_parse_id(matroska, syntax, id, data);
1091 static void ebml_free(EbmlSyntax *syntax, void *data)
1094 for (i=0; syntax[i].id; i++) {
1095 void *data_off = (char *)data + syntax[i].data_offset;
1096 switch (syntax[i].type) {
1098 case EBML_UTF8: av_freep(data_off); break;
1099 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
1101 if (syntax[i].list_elem_size) {
1102 EbmlList *list = data_off;
1103 char *ptr = list->elem;
1104 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
1105 ebml_free(syntax[i].def.n, ptr);
1106 av_free(list->elem);
1108 ebml_free(syntax[i].def.n, data_off);
1115 matroska_decode_buffer(uint8_t** buf, int* buf_size, MatroskaTrack *track)
1117 MatroskaTrackEncoding *encodings = track->encodings.elem;
1118 uint8_t* data = *buf;
1119 int isize = *buf_size;
1120 uint8_t* pkt_data = NULL;
1121 int pkt_size = isize;
1125 switch (encodings[0].compression.algo) {
1126 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1127 return encodings[0].compression.settings.size;
1128 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1130 olen = pkt_size *= 3;
1131 pkt_data = av_realloc(pkt_data,
1132 pkt_size+LZO_OUTPUT_PADDING);
1133 result = lzo1x_decode(pkt_data, &olen, data, &isize);
1134 } while (result==LZO_OUTPUT_FULL && pkt_size<10000000);
1140 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
1141 z_stream zstream = {0};
1142 if (inflateInit(&zstream) != Z_OK)
1144 zstream.next_in = data;
1145 zstream.avail_in = isize;
1148 pkt_data = av_realloc(pkt_data, pkt_size);
1149 zstream.avail_out = pkt_size - zstream.total_out;
1150 zstream.next_out = pkt_data + zstream.total_out;
1151 result = inflate(&zstream, Z_NO_FLUSH);
1152 } while (result==Z_OK && pkt_size<10000000);
1153 pkt_size = zstream.total_out;
1154 inflateEnd(&zstream);
1155 if (result != Z_STREAM_END)
1161 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
1162 bz_stream bzstream = {0};
1163 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1165 bzstream.next_in = data;
1166 bzstream.avail_in = isize;
1169 pkt_data = av_realloc(pkt_data, pkt_size);
1170 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1171 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1172 result = BZ2_bzDecompress(&bzstream);
1173 } while (result==BZ_OK && pkt_size<10000000);
1174 pkt_size = bzstream.total_out_lo32;
1175 BZ2_bzDecompressEnd(&bzstream);
1176 if (result != BZ_STREAM_END)
1184 *buf_size = pkt_size;
1192 matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1194 EbmlList *seekhead_list = &matroska->seekhead;
1195 MatroskaSeekhead *seekhead = seekhead_list->elem;
1196 uint32_t peek_id_cache = matroska->peek_id;
1197 uint32_t level_up = matroska->level_up;
1198 offset_t before_pos = url_ftell(matroska->ctx->pb);
1199 MatroskaLevel level;
1202 for (i=0; i<seekhead_list->nb_elem; i++) {
1203 if (seekhead[i].pos <= before_pos
1204 || seekhead[i].id == MATROSKA_ID_SEEKHEAD
1205 || seekhead[i].id == MATROSKA_ID_CLUSTER)
1209 if (ebml_read_seek(matroska,
1210 seekhead[i].pos+matroska->segment_start) < 0)
1213 /* we don't want to lose our seekhead level, so we add
1214 * a dummy. This is a crude hack. */
1215 if (matroska->num_levels == EBML_MAX_DEPTH) {
1216 av_log(matroska->ctx, AV_LOG_INFO,
1217 "Max EBML element depth (%d) reached, "
1218 "cannot parse further.\n", EBML_MAX_DEPTH);
1223 level.length = (uint64_t)-1;
1224 matroska->levels[matroska->num_levels] = level;
1225 matroska->num_levels++;
1227 ebml_parse_id(matroska, matroska_segment, seekhead[i].id, matroska);
1229 /* remove dummy level */
1230 while (matroska->num_levels) {
1231 uint64_t length = matroska->levels[--matroska->num_levels].length;
1232 if (length == (uint64_t)-1)
1238 ebml_read_seek(matroska, before_pos);
1239 matroska->peek_id = peek_id_cache;
1240 matroska->level_up = level_up;
1244 matroska_aac_profile (char *codec_id)
1246 static const char *aac_profiles[] = { "MAIN", "LC", "SSR" };
1249 for (profile=0; profile<ARRAY_SIZE(aac_profiles); profile++)
1250 if (strstr(codec_id, aac_profiles[profile]))
1256 matroska_aac_sri (int samplerate)
1260 for (sri=0; sri<ARRAY_SIZE(ff_mpeg4audio_sample_rates); sri++)
1261 if (ff_mpeg4audio_sample_rates[sri] == samplerate)
1267 matroska_read_header (AVFormatContext *s,
1268 AVFormatParameters *ap)
1270 MatroskaDemuxContext *matroska = s->priv_data;
1271 EbmlList *attachements_list = &matroska->attachments;
1272 MatroskaAttachement *attachements;
1273 EbmlList *chapters_list = &matroska->chapters;
1274 MatroskaChapter *chapters;
1275 MatroskaTrack *tracks;
1276 EbmlList *index_list;
1277 MatroskaIndex *index;
1284 /* First read the EBML header. */
1285 if (ebml_parse(matroska, ebml_syntax, &ebml, 0, 1)
1286 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1287 || ebml.id_length > sizeof(uint32_t) || strcmp(ebml.doctype, "matroska")
1288 || ebml.doctype_version > 2) {
1289 av_log(matroska->ctx, AV_LOG_ERROR,
1290 "EBML header using unsupported features\n"
1291 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1292 ebml.version, ebml.doctype, ebml.doctype_version);
1293 return AVERROR_NOFMT;
1295 ebml_free(ebml_syntax, &ebml);
1297 /* The next thing is a segment. */
1298 if (ebml_parse(matroska, matroska_segments, matroska, 0, 1) < 0)
1300 matroska_execute_seekhead(matroska);
1302 if (matroska->duration)
1303 matroska->ctx->duration = matroska->duration * matroska->time_scale
1304 * 1000 / AV_TIME_BASE;
1305 if (matroska->title)
1306 strncpy(matroska->ctx->title, matroska->title,
1307 sizeof(matroska->ctx->title)-1);
1309 tracks = matroska->tracks.elem;
1310 for (i=0; i < matroska->tracks.nb_elem; i++) {
1311 MatroskaTrack *track = &tracks[i];
1312 enum CodecID codec_id = CODEC_ID_NONE;
1313 EbmlList *encodings_list = &tracks->encodings;
1314 MatroskaTrackEncoding *encodings = encodings_list->elem;
1315 uint8_t *extradata = NULL;
1316 int extradata_size = 0;
1317 int extradata_offset = 0;
1319 /* Apply some sanity checks. */
1320 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1321 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1322 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1323 av_log(matroska->ctx, AV_LOG_INFO,
1324 "Unknown or unsupported track type %"PRIu64"\n",
1328 if (track->codec_id == NULL)
1331 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1332 if (!track->default_duration)
1333 track->default_duration = 1000000000/track->video.frame_rate;
1334 if (!track->video.display_width)
1335 track->video.display_width = track->video.pixel_width;
1336 if (!track->video.display_height)
1337 track->video.display_height = track->video.pixel_height;
1338 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1339 if (!track->audio.out_samplerate)
1340 track->audio.out_samplerate = track->audio.samplerate;
1342 if (encodings_list->nb_elem > 1) {
1343 av_log(matroska->ctx, AV_LOG_ERROR,
1344 "Multiple combined encodings no supported");
1345 } else if (encodings_list->nb_elem == 1) {
1346 if (encodings[0].type ||
1347 (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1349 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1352 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1354 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
1355 encodings[0].scope = 0;
1356 av_log(matroska->ctx, AV_LOG_ERROR,
1357 "Unsupported encoding type");
1358 } else if (track->codec_priv.size && encodings[0].scope&2) {
1359 uint8_t *codec_priv = track->codec_priv.data;
1360 int offset = matroska_decode_buffer(&track->codec_priv.data,
1361 &track->codec_priv.size,
1364 track->codec_priv.data = NULL;
1365 track->codec_priv.size = 0;
1366 av_log(matroska->ctx, AV_LOG_ERROR,
1367 "Failed to decode codec private data\n");
1368 } else if (offset > 0) {
1369 track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
1370 memcpy(track->codec_priv.data,
1371 encodings[0].compression.settings.data, offset);
1372 memcpy(track->codec_priv.data+offset, codec_priv,
1373 track->codec_priv.size);
1374 track->codec_priv.size += offset;
1376 if (codec_priv != track->codec_priv.data)
1377 av_free(codec_priv);
1381 for(j=0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++){
1382 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1383 strlen(ff_mkv_codec_tags[j].str))){
1384 codec_id= ff_mkv_codec_tags[j].id;
1389 st = track->stream = av_new_stream(s, matroska->num_streams++);
1391 return AVERROR(ENOMEM);
1393 if (!strcmp(track->codec_id, MATROSKA_CODEC_ID_VIDEO_VFW_FOURCC)
1394 && track->codec_priv.size >= 40
1395 && track->codec_priv.data != NULL) {
1396 track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1397 codec_id = codec_get_id(codec_bmp_tags, track->video.fourcc);
1398 } else if (!strcmp(track->codec_id, MATROSKA_CODEC_ID_AUDIO_ACM)
1399 && track->codec_priv.size >= 18
1400 && track->codec_priv.data != NULL) {
1401 uint16_t tag = AV_RL16(track->codec_priv.data);
1402 codec_id = codec_get_id(codec_wav_tags, tag);
1403 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1404 && (track->codec_priv.size >= 86)
1405 && (track->codec_priv.data != NULL)) {
1406 track->video.fourcc = AV_RL32(track->codec_priv.data);
1407 codec_id=codec_get_id(codec_movvideo_tags, track->video.fourcc);
1408 } else if (codec_id == CODEC_ID_AAC && !track->codec_priv.size) {
1409 int profile = matroska_aac_profile(track->codec_id);
1410 int sri = matroska_aac_sri(track->audio.samplerate);
1411 extradata = av_malloc(5);
1412 if (extradata == NULL)
1413 return AVERROR(ENOMEM);
1414 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1415 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1416 if (strstr(track->codec_id, "SBR")) {
1417 sri = matroska_aac_sri(track->audio.out_samplerate);
1418 extradata[2] = 0x56;
1419 extradata[3] = 0xE5;
1420 extradata[4] = 0x80 | (sri<<3);
1424 } else if (codec_id == CODEC_ID_TTA) {
1426 extradata_size = 30;
1427 extradata = av_mallocz(extradata_size);
1428 if (extradata == NULL)
1429 return AVERROR(ENOMEM);
1430 init_put_byte(&b, extradata, extradata_size, 1,
1431 NULL, NULL, NULL, NULL);
1432 put_buffer(&b, "TTA1", 4);
1434 put_le16(&b, track->audio.channels);
1435 put_le16(&b, track->audio.bitdepth);
1436 put_le32(&b, track->audio.out_samplerate);
1437 put_le32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1438 } else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
1439 codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
1440 extradata_offset = 26;
1441 track->codec_priv.size -= extradata_offset;
1442 } else if (codec_id == CODEC_ID_RA_144) {
1443 track->audio.out_samplerate = 8000;
1444 track->audio.channels = 1;
1445 } else if (codec_id == CODEC_ID_RA_288 || codec_id == CODEC_ID_COOK ||
1446 codec_id == CODEC_ID_ATRAC3) {
1449 init_put_byte(&b, track->codec_priv.data,track->codec_priv.size,
1450 0, NULL, NULL, NULL, NULL);
1452 track->audio.coded_framesize = get_be32(&b);
1454 track->audio.sub_packet_h = get_be16(&b);
1455 track->audio.frame_size = get_be16(&b);
1456 track->audio.sub_packet_size = get_be16(&b);
1457 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1458 if (codec_id == CODEC_ID_RA_288) {
1459 st->codec->block_align = track->audio.coded_framesize;
1460 track->codec_priv.size = 0;
1462 st->codec->block_align = track->audio.sub_packet_size;
1463 extradata_offset = 78;
1464 track->codec_priv.size -= extradata_offset;
1468 if (codec_id == CODEC_ID_NONE)
1469 av_log(matroska->ctx, AV_LOG_INFO,
1470 "Unknown/unsupported CodecID %s.\n", track->codec_id);
1472 av_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1474 st->codec->codec_id = codec_id;
1476 if (strcmp(track->language, "und"))
1477 av_strlcpy(st->language, track->language, 4);
1479 if (track->flag_default)
1480 st->disposition |= AV_DISPOSITION_DEFAULT;
1482 if (track->default_duration)
1483 av_reduce(&st->codec->time_base.num, &st->codec->time_base.den,
1484 track->default_duration, 1000000000, 30000);
1487 st->codec->extradata = extradata;
1488 st->codec->extradata_size = extradata_size;
1489 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1490 st->codec->extradata = av_malloc(track->codec_priv.size);
1491 if(st->codec->extradata == NULL)
1492 return AVERROR(ENOMEM);
1493 st->codec->extradata_size = track->codec_priv.size;
1494 memcpy(st->codec->extradata,
1495 track->codec_priv.data + extradata_offset,
1496 track->codec_priv.size);
1499 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1500 st->codec->codec_type = CODEC_TYPE_VIDEO;
1501 st->codec->codec_tag = track->video.fourcc;
1502 st->codec->width = track->video.pixel_width;
1503 st->codec->height = track->video.pixel_height;
1504 av_reduce(&st->codec->sample_aspect_ratio.num,
1505 &st->codec->sample_aspect_ratio.den,
1506 st->codec->height * track->video.display_width,
1507 st->codec-> width * track->video.display_height,
1509 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1510 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1511 st->codec->codec_type = CODEC_TYPE_AUDIO;
1512 st->codec->sample_rate = track->audio.out_samplerate;
1513 st->codec->channels = track->audio.channels;
1514 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1515 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
1519 attachements = attachements_list->elem;
1520 for (j=0; j<attachements_list->nb_elem; j++) {
1521 if (!(attachements[j].filename && attachements[j].mime &&
1522 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1523 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1525 AVStream *st = av_new_stream(s, matroska->num_streams++);
1528 st->filename = av_strdup(attachements[j].filename);
1529 st->codec->codec_id = CODEC_ID_NONE;
1530 st->codec->codec_type = CODEC_TYPE_ATTACHMENT;
1531 st->codec->extradata = av_malloc(attachements[j].bin.size);
1532 if(st->codec->extradata == NULL)
1534 st->codec->extradata_size = attachements[j].bin.size;
1535 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1537 for (i=0; ff_mkv_mime_tags[i].id != CODEC_ID_NONE; i++) {
1538 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1539 strlen(ff_mkv_mime_tags[i].str))) {
1540 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1547 chapters = chapters_list->elem;
1548 for (i=0; i<chapters_list->nb_elem; i++)
1549 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid)
1550 ff_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1551 chapters[i].start, chapters[i].end,
1554 index_list = &matroska->index;
1555 index = index_list->elem;
1556 for (i=0; i<index_list->nb_elem; i++) {
1557 EbmlList *pos_list = &index[i].pos;
1558 MatroskaIndexPos *pos = pos_list->elem;
1559 for (j=0; j<pos_list->nb_elem; j++) {
1560 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1562 if (track && track->stream)
1563 av_add_index_entry(track->stream,
1564 pos[j].pos + matroska->segment_start,
1565 index[i].time*matroska->time_scale/AV_TIME_BASE,
1566 0, 0, AVINDEX_KEYFRAME);
1574 matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data, int size,
1575 int64_t pos, uint64_t cluster_time, uint64_t duration,
1578 MatroskaTrack *track;
1583 uint32_t *lace_size = NULL;
1584 int n, flags, laces = 0;
1587 if ((n = matroska_ebmlnum_uint(data, size, &num)) < 0) {
1588 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
1594 track = matroska_find_track_by_num(matroska, num);
1595 if (size <= 3 || !track || !track->stream) {
1596 av_log(matroska->ctx, AV_LOG_INFO,
1597 "Invalid stream %"PRIu64" or size %u\n", num, size);
1601 if (st->discard >= AVDISCARD_ALL)
1603 if (duration == AV_NOPTS_VALUE)
1604 duration = track->default_duration / matroska->time_scale;
1606 block_time = AV_RB16(data);
1610 if (is_keyframe == -1)
1611 is_keyframe = flags & 0x80 ? PKT_FLAG_KEY : 0;
1613 if (matroska->skip_to_keyframe) {
1614 if (!is_keyframe || st != matroska->skip_to_stream)
1616 matroska->skip_to_keyframe = 0;
1619 switch ((flags & 0x06) >> 1) {
1620 case 0x0: /* no lacing */
1622 lace_size = av_mallocz(sizeof(int));
1623 lace_size[0] = size;
1626 case 0x1: /* xiph lacing */
1627 case 0x2: /* fixed-size lacing */
1628 case 0x3: /* EBML lacing */
1629 assert(size>0); // size <=3 is checked before size-=3 above
1630 laces = (*data) + 1;
1633 lace_size = av_mallocz(laces * sizeof(int));
1635 switch ((flags & 0x06) >> 1) {
1636 case 0x1: /* xiph lacing */ {
1639 for (n = 0; res == 0 && n < laces - 1; n++) {
1646 lace_size[n] += temp;
1652 total += lace_size[n];
1654 lace_size[n] = size - total;
1658 case 0x2: /* fixed-size lacing */
1659 for (n = 0; n < laces; n++)
1660 lace_size[n] = size / laces;
1663 case 0x3: /* EBML lacing */ {
1665 n = matroska_ebmlnum_uint(data, size, &num);
1667 av_log(matroska->ctx, AV_LOG_INFO,
1668 "EBML block data error\n");
1673 total = lace_size[0] = num;
1674 for (n = 1; res == 0 && n < laces - 1; n++) {
1677 r = matroska_ebmlnum_sint (data, size, &snum);
1679 av_log(matroska->ctx, AV_LOG_INFO,
1680 "EBML block data error\n");
1685 lace_size[n] = lace_size[n - 1] + snum;
1686 total += lace_size[n];
1688 lace_size[n] = size - total;
1696 uint64_t timecode = AV_NOPTS_VALUE;
1698 if (cluster_time != (uint64_t)-1
1699 && (block_time >= 0 || cluster_time >= -block_time))
1700 timecode = cluster_time + block_time;
1702 for (n = 0; n < laces; n++) {
1703 if (st->codec->codec_id == CODEC_ID_RA_288 ||
1704 st->codec->codec_id == CODEC_ID_COOK ||
1705 st->codec->codec_id == CODEC_ID_ATRAC3) {
1706 int a = st->codec->block_align;
1707 int sps = track->audio.sub_packet_size;
1708 int cfs = track->audio.coded_framesize;
1709 int h = track->audio.sub_packet_h;
1710 int y = track->audio.sub_packet_cnt;
1711 int w = track->audio.frame_size;
1714 if (!track->audio.pkt_cnt) {
1715 if (st->codec->codec_id == CODEC_ID_RA_288)
1716 for (x=0; x<h/2; x++)
1717 memcpy(track->audio.buf+x*2*w+y*cfs,
1720 for (x=0; x<w/sps; x++)
1721 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
1723 if (++track->audio.sub_packet_cnt >= h) {
1724 track->audio.sub_packet_cnt = 0;
1725 track->audio.pkt_cnt = h*w / a;
1728 while (track->audio.pkt_cnt) {
1729 pkt = av_mallocz(sizeof(AVPacket));
1730 av_new_packet(pkt, a);
1731 memcpy(pkt->data, track->audio.buf
1732 + a * (h*w / a - track->audio.pkt_cnt--), a);
1734 pkt->stream_index = st->index;
1735 matroska_queue_packet(matroska, pkt);
1738 MatroskaTrackEncoding *encodings = track->encodings.elem;
1739 int offset = 0, pkt_size = lace_size[n];
1740 uint8_t *pkt_data = data;
1742 if (encodings && encodings->scope & 1) {
1743 offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
1748 pkt = av_mallocz(sizeof(AVPacket));
1749 /* XXX: prevent data copy... */
1750 if (av_new_packet(pkt, pkt_size+offset) < 0) {
1752 res = AVERROR(ENOMEM);
1757 memcpy (pkt->data, encodings->compression.settings.data, offset);
1758 memcpy (pkt->data+offset, pkt_data, pkt_size);
1760 if (pkt_data != data)
1764 pkt->flags = is_keyframe;
1765 pkt->stream_index = st->index;
1767 pkt->pts = timecode;
1769 pkt->duration = duration;
1771 matroska_queue_packet(matroska, pkt);
1774 if (timecode != AV_NOPTS_VALUE)
1775 timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
1776 data += lace_size[n];
1785 matroska_parse_cluster (MatroskaDemuxContext *matroska)
1787 MatroskaCluster cluster = { 0 };
1788 EbmlList *blocks_list;
1789 MatroskaBlock *blocks;
1790 int i, res = ebml_parse(matroska, matroska_clusters, &cluster, 0, 1);
1791 blocks_list = &cluster.blocks;
1792 blocks = blocks_list->elem;
1793 for (i=0; !res && i<blocks_list->nb_elem; i++)
1794 if (blocks[i].bin.size > 0)
1795 res=matroska_parse_block(matroska,
1796 blocks[i].bin.data, blocks[i].bin.size,
1797 blocks[i].bin.pos, cluster.timecode,
1798 blocks[i].duration, !blocks[i].reference);
1799 ebml_free(matroska_cluster, &cluster);
1804 matroska_read_packet (AVFormatContext *s,
1807 MatroskaDemuxContext *matroska = s->priv_data;
1809 while (matroska_deliver_packet(matroska, pkt)) {
1811 return AVERROR(EIO);
1812 if (matroska_parse_cluster(matroska) < 0)
1820 matroska_read_seek (AVFormatContext *s, int stream_index, int64_t timestamp,
1823 MatroskaDemuxContext *matroska = s->priv_data;
1824 AVStream *st = s->streams[stream_index];
1827 index = av_index_search_timestamp(st, timestamp, flags);
1831 matroska_clear_queue(matroska);
1833 url_fseek(s->pb, st->index_entries[index].pos, SEEK_SET);
1834 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
1835 matroska->skip_to_stream = st;
1836 matroska->peek_id = 0;
1837 av_update_cur_dts(s, st, st->index_entries[index].timestamp);
1842 matroska_read_close (AVFormatContext *s)
1844 MatroskaDemuxContext *matroska = s->priv_data;
1845 MatroskaTrack *tracks = matroska->tracks.elem;
1848 matroska_clear_queue(matroska);
1850 for (n=0; n < matroska->tracks.nb_elem; n++)
1851 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
1852 av_free(tracks[n].audio.buf);
1853 ebml_free(matroska_segment, matroska);
1858 AVInputFormat matroska_demuxer = {
1860 NULL_IF_CONFIG_SMALL("Matroska file format"),
1861 sizeof(MatroskaDemuxContext),
1863 matroska_read_header,
1864 matroska_read_packet,
1865 matroska_read_close,