matroskadec.c
Go to the documentation of this file.
1 /*
2  * Matroska file demuxer
3  * Copyright (c) 2003-2008 The Libav Project
4  *
5  * This file is part of Libav.
6  *
7  * Libav 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.
11  *
12  * Libav 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.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with Libav; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
31 #include <stdio.h>
32 #include "avformat.h"
33 #include "internal.h"
34 #include "avio_internal.h"
35 /* For ff_codec_get_id(). */
36 #include "riff.h"
37 #include "isom.h"
38 #include "rm.h"
39 #include "matroska.h"
40 #include "libavcodec/mpeg4audio.h"
41 #include "libavutil/intfloat.h"
42 #include "libavutil/intreadwrite.h"
43 #include "libavutil/avstring.h"
44 #include "libavutil/lzo.h"
45 #include "libavutil/dict.h"
46 #if CONFIG_ZLIB
47 #include <zlib.h>
48 #endif
49 #if CONFIG_BZLIB
50 #include <bzlib.h>
51 #endif
52 
53 typedef enum {
64 } EbmlType;
65 
66 typedef const struct EbmlSyntax {
67  uint32_t id;
71  union {
72  uint64_t u;
73  double f;
74  const char *s;
75  const struct EbmlSyntax *n;
76  } def;
77 } EbmlSyntax;
78 
79 typedef struct {
80  int nb_elem;
81  void *elem;
82 } EbmlList;
83 
84 typedef struct {
85  int size;
86  uint8_t *data;
87  int64_t pos;
88 } EbmlBin;
89 
90 typedef struct {
91  uint64_t version;
92  uint64_t max_size;
93  uint64_t id_length;
94  char *doctype;
95  uint64_t doctype_version;
96 } Ebml;
97 
98 typedef struct {
99  uint64_t algo;
102 
103 typedef struct {
104  uint64_t scope;
105  uint64_t type;
108 
109 typedef struct {
110  double frame_rate;
111  uint64_t display_width;
112  uint64_t display_height;
113  uint64_t pixel_width;
114  uint64_t pixel_height;
115  uint64_t fourcc;
117 
118 typedef struct {
119  double samplerate;
121  uint64_t bitdepth;
122  uint64_t channels;
123 
124  /* real audio header (extracted from extradata) */
130  int pkt_cnt;
131  uint64_t buf_timecode;
132  uint8_t *buf;
134 
135 typedef struct {
136  uint64_t num;
137  uint64_t uid;
138  uint64_t type;
139  char *name;
140  char *codec_id;
142  char *language;
143  double time_scale;
145  uint64_t flag_default;
146  uint64_t flag_forced;
150 
152  int64_t end_timecode;
154 } MatroskaTrack;
155 
156 typedef struct {
157  uint64_t uid;
158  char *filename;
159  char *mime;
161 
164 
165 typedef struct {
166  uint64_t start;
167  uint64_t end;
168  uint64_t uid;
169  char *title;
170 
173 
174 typedef struct {
175  uint64_t track;
176  uint64_t pos;
178 
179 typedef struct {
180  uint64_t time;
182 } MatroskaIndex;
183 
184 typedef struct {
185  char *name;
186  char *string;
187  char *lang;
188  uint64_t def;
190 } MatroskaTag;
191 
192 typedef struct {
193  char *type;
194  uint64_t typevalue;
195  uint64_t trackuid;
196  uint64_t chapteruid;
197  uint64_t attachuid;
199 
200 typedef struct {
203 } MatroskaTags;
204 
205 typedef struct {
206  uint64_t id;
207  uint64_t pos;
209 
210 typedef struct {
211  uint64_t start;
212  uint64_t length;
213 } MatroskaLevel;
214 
215 typedef struct {
217 
218  /* EBML stuff */
221  int level_up;
222  uint32_t current_id;
223 
224  uint64_t time_scale;
225  double duration;
226  char *title;
233 
234  /* byte position of the segment inside the stream */
235  int64_t segment_start;
236 
237  /* the packet queue */
241 
242  int done;
243 
244  /* What to skip before effectively reading a packet. */
247 
248  /* File has a CUES element, but we defer parsing until it is needed. */
251 
252 typedef struct {
253  uint64_t duration;
254  int64_t reference;
255  uint64_t non_simple;
257 } MatroskaBlock;
258 
259 typedef struct {
260  uint64_t timecode;
263 
265  { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
266  { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
267  { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
268  { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
269  { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
272  { 0 }
273 };
274 
276  { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
277  { 0 }
278 };
279 
281  { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
283  { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
288  { 0 }
289 };
290 
293  { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
294  { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
295  { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
296  { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
297  { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
306  { 0 }
307 };
308 
310  { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
311  { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
313  { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
314  { 0 }
315 };
316 
320  { 0 }
321 };
322 
324  { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
325  { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
326  { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
328  { 0 }
329 };
330 
332  { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
333  { 0 }
334 };
335 
337  { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
339  { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
340  { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
342  { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
343  { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
344  { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
345  { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
346  { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
347  { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack,flag_forced), {.u=0} },
348  { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
349  { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
350  { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
360  { 0 }
361 };
362 
364  { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
365  { 0 }
366 };
367 
369  { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
370  { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
371  { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
372  { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
374  { 0 }
375 };
376 
378  { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
379  { 0 }
380 };
381 
383  { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
385  { 0 }
386 };
387 
391  { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
392  { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
397  { 0 }
398 };
399 
401  { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
406  { 0 }
407 };
408 
410  { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
411  { 0 }
412 };
413 
415  { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
418  { 0 }
419 };
420 
422  { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
423  { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
424  { 0 }
425 };
426 
428  { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
429  { 0 }
430 };
431 
433  { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
434  { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
435  { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag,lang), {.s="und"} },
436  { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag,def) },
437  { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag,def) },
438  { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
439  { 0 }
440 };
441 
444  { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
445  { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
447  { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
448  { 0 }
449 };
450 
452  { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
453  { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
454  { 0 }
455 };
456 
458  { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
459  { 0 }
460 };
461 
463  { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
464  { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
465  { 0 }
466 };
467 
469  { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
470  { 0 }
471 };
472 
474  { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
475  { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
476  { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
477  { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
478  { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
479  { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
480  { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
482  { 0 }
483 };
484 
486  { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
487  { 0 }
488 };
489 
491  { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
492  { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
494  { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
495  { 1, EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
496  { 0 }
497 };
498 
500  { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
501  { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
502  { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
505  { 0 }
506 };
507 
509  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
514  { 0 }
515 };
516 
517 static const char *matroska_doctypes[] = { "matroska", "webm" };
518 
519 /*
520  * Return: Whether we reached the end of a level in the hierarchy or not.
521  */
523 {
524  AVIOContext *pb = matroska->ctx->pb;
525  int64_t pos = avio_tell(pb);
526 
527  if (matroska->num_levels > 0) {
528  MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
529  if (pos - level->start >= level->length || matroska->current_id) {
530  matroska->num_levels--;
531  return 1;
532  }
533  }
534  return 0;
535 }
536 
537 /*
538  * Read: an "EBML number", which is defined as a variable-length
539  * array of bytes. The first byte indicates the length by giving a
540  * number of 0-bits followed by a one. The position of the first
541  * "one" bit inside the first byte indicates the length of this
542  * number.
543  * Returns: number of bytes read, < 0 on error
544  */
546  int max_size, uint64_t *number)
547 {
548  int read = 1, n = 1;
549  uint64_t total = 0;
550 
551  /* The first byte tells us the length in bytes - avio_r8() can normally
552  * return 0, but since that's not a valid first ebmlID byte, we can
553  * use it safely here to catch EOS. */
554  if (!(total = avio_r8(pb))) {
555  /* we might encounter EOS here */
556  if (!pb->eof_reached) {
557  int64_t pos = avio_tell(pb);
558  av_log(matroska->ctx, AV_LOG_ERROR,
559  "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
560  pos, pos);
561  }
562  return AVERROR(EIO); /* EOS or actual I/O error */
563  }
564 
565  /* get the length of the EBML number */
566  read = 8 - ff_log2_tab[total];
567  if (read > max_size) {
568  int64_t pos = avio_tell(pb) - 1;
569  av_log(matroska->ctx, AV_LOG_ERROR,
570  "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
571  (uint8_t) total, pos, pos);
572  return AVERROR_INVALIDDATA;
573  }
574 
575  /* read out length */
576  total ^= 1 << ff_log2_tab[total];
577  while (n++ < read)
578  total = (total << 8) | avio_r8(pb);
579 
580  *number = total;
581 
582  return read;
583 }
584 
591  uint64_t *number)
592 {
593  int res = ebml_read_num(matroska, pb, 8, number);
594  if (res > 0 && *number + 1 == 1ULL << (7 * res))
595  *number = 0xffffffffffffffULL;
596  return res;
597 }
598 
599 /*
600  * Read the next element as an unsigned int.
601  * 0 is success, < 0 is failure.
602  */
603 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
604 {
605  int n = 0;
606 
607  if (size > 8)
608  return AVERROR_INVALIDDATA;
609 
610  /* big-endian ordering; build up number */
611  *num = 0;
612  while (n++ < size)
613  *num = (*num << 8) | avio_r8(pb);
614 
615  return 0;
616 }
617 
618 /*
619  * Read the next element as a float.
620  * 0 is success, < 0 is failure.
621  */
622 static int ebml_read_float(AVIOContext *pb, int size, double *num)
623 {
624  if (size == 0) {
625  *num = 0;
626  } else if (size == 4) {
627  *num = av_int2float(avio_rb32(pb));
628  } else if (size == 8){
629  *num = av_int2double(avio_rb64(pb));
630  } else
631  return AVERROR_INVALIDDATA;
632 
633  return 0;
634 }
635 
636 /*
637  * Read the next element as an ASCII string.
638  * 0 is success, < 0 is failure.
639  */
640 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
641 {
642  char *res;
643 
644  /* EBML strings are usually not 0-terminated, so we allocate one
645  * byte more, read the string and NULL-terminate it ourselves. */
646  if (!(res = av_malloc(size + 1)))
647  return AVERROR(ENOMEM);
648  if (avio_read(pb, (uint8_t *) res, size) != size) {
649  av_free(res);
650  return AVERROR(EIO);
651  }
652  (res)[size] = '\0';
653  av_free(*str);
654  *str = res;
655 
656  return 0;
657 }
658 
659 /*
660  * Read the next element as binary data.
661  * 0 is success, < 0 is failure.
662  */
663 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
664 {
665  av_free(bin->data);
666  if (!(bin->data = av_malloc(length)))
667  return AVERROR(ENOMEM);
668 
669  bin->size = length;
670  bin->pos = avio_tell(pb);
671  if (avio_read(pb, bin->data, length) != length) {
672  av_freep(&bin->data);
673  return AVERROR(EIO);
674  }
675 
676  return 0;
677 }
678 
679 /*
680  * Read the next element, but only the header. The contents
681  * are supposed to be sub-elements which can be read separately.
682  * 0 is success, < 0 is failure.
683  */
684 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
685 {
686  AVIOContext *pb = matroska->ctx->pb;
688 
689  if (matroska->num_levels >= EBML_MAX_DEPTH) {
690  av_log(matroska->ctx, AV_LOG_ERROR,
691  "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
692  return AVERROR(ENOSYS);
693  }
694 
695  level = &matroska->levels[matroska->num_levels++];
696  level->start = avio_tell(pb);
697  level->length = length;
698 
699  return 0;
700 }
701 
702 /*
703  * Read signed/unsigned "EBML" numbers.
704  * Return: number of bytes processed, < 0 on error
705  */
707  uint8_t *data, uint32_t size, uint64_t *num)
708 {
709  AVIOContext pb;
710  ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
711  return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
712 }
713 
714 /*
715  * Same as above, but signed.
716  */
718  uint8_t *data, uint32_t size, int64_t *num)
719 {
720  uint64_t unum;
721  int res;
722 
723  /* read as unsigned number first */
724  if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
725  return res;
726 
727  /* make signed (weird way) */
728  *num = unum - ((1LL << (7*res - 1)) - 1);
729 
730  return res;
731 }
732 
733 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
734  EbmlSyntax *syntax, void *data);
735 
736 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
737  uint32_t id, void *data)
738 {
739  int i;
740  for (i=0; syntax[i].id; i++)
741  if (id == syntax[i].id)
742  break;
743  if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
744  matroska->num_levels > 0 &&
745  matroska->levels[matroska->num_levels-1].length == 0xffffffffffffff)
746  return 0; // we reached the end of an unknown size cluster
747  if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32)
748  av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
749  return ebml_parse_elem(matroska, &syntax[i], data);
750 }
751 
752 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
753  void *data)
754 {
755  if (!matroska->current_id) {
756  uint64_t id;
757  int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
758  if (res < 0)
759  return res;
760  matroska->current_id = id | 1 << 7*res;
761  }
762  return ebml_parse_id(matroska, syntax, matroska->current_id, data);
763 }
764 
765 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
766  void *data)
767 {
768  int i, res = 0;
769 
770  for (i=0; syntax[i].id; i++)
771  switch (syntax[i].type) {
772  case EBML_UINT:
773  *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
774  break;
775  case EBML_FLOAT:
776  *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
777  break;
778  case EBML_STR:
779  case EBML_UTF8:
780  *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
781  break;
782  }
783 
784  while (!res && !ebml_level_end(matroska))
785  res = ebml_parse(matroska, syntax, data);
786 
787  return res;
788 }
789 
791  EbmlSyntax *syntax, void *data)
792 {
793  static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
794  [EBML_UINT] = 8,
795  [EBML_FLOAT] = 8,
796  // max. 16 MB for strings
797  [EBML_STR] = 0x1000000,
798  [EBML_UTF8] = 0x1000000,
799  // max. 256 MB for binary data
800  [EBML_BIN] = 0x10000000,
801  // no limits for anything else
802  };
803  AVIOContext *pb = matroska->ctx->pb;
804  uint32_t id = syntax->id;
805  uint64_t length;
806  int res;
807  void *newelem;
808 
809  data = (char *)data + syntax->data_offset;
810  if (syntax->list_elem_size) {
811  EbmlList *list = data;
812  newelem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
813  if (!newelem)
814  return AVERROR(ENOMEM);
815  list->elem = newelem;
816  data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
817  memset(data, 0, syntax->list_elem_size);
818  list->nb_elem++;
819  }
820 
821  if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
822  matroska->current_id = 0;
823  if ((res = ebml_read_length(matroska, pb, &length)) < 0)
824  return res;
825  if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
826  av_log(matroska->ctx, AV_LOG_ERROR,
827  "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
828  length, max_lengths[syntax->type], syntax->type);
829  return AVERROR_INVALIDDATA;
830  }
831  }
832 
833  switch (syntax->type) {
834  case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
835  case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
836  case EBML_STR:
837  case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
838  case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
839  case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
840  return res;
841  if (id == MATROSKA_ID_SEGMENT)
842  matroska->segment_start = avio_tell(matroska->ctx->pb);
843  return ebml_parse_nest(matroska, syntax->def.n, data);
844  case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
845  case EBML_STOP: return 1;
846  default: return avio_skip(pb,length)<0 ? AVERROR(EIO) : 0;
847  }
848  if (res == AVERROR_INVALIDDATA)
849  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
850  else if (res == AVERROR(EIO))
851  av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
852  return res;
853 }
854 
855 static void ebml_free(EbmlSyntax *syntax, void *data)
856 {
857  int i, j;
858  for (i=0; syntax[i].id; i++) {
859  void *data_off = (char *)data + syntax[i].data_offset;
860  switch (syntax[i].type) {
861  case EBML_STR:
862  case EBML_UTF8: av_freep(data_off); break;
863  case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
864  case EBML_NEST:
865  if (syntax[i].list_elem_size) {
866  EbmlList *list = data_off;
867  char *ptr = list->elem;
868  for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
869  ebml_free(syntax[i].def.n, ptr);
870  av_free(list->elem);
871  } else
872  ebml_free(syntax[i].def.n, data_off);
873  default: break;
874  }
875  }
876 }
877 
878 
879 /*
880  * Autodetecting...
881  */
883 {
884  uint64_t total = 0;
885  int len_mask = 0x80, size = 1, n = 1, i;
886 
887  /* EBML header? */
888  if (AV_RB32(p->buf) != EBML_ID_HEADER)
889  return 0;
890 
891  /* length of header */
892  total = p->buf[4];
893  while (size <= 8 && !(total & len_mask)) {
894  size++;
895  len_mask >>= 1;
896  }
897  if (size > 8)
898  return 0;
899  total &= (len_mask - 1);
900  while (n < size)
901  total = (total << 8) | p->buf[4 + n++];
902 
903  /* Does the probe data contain the whole header? */
904  if (p->buf_size < 4 + size + total)
905  return 0;
906 
907  /* The header should contain a known document type. For now,
908  * we don't parse the whole header but simply check for the
909  * availability of that array of characters inside the header.
910  * Not fully fool-proof, but good enough. */
911  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
912  int probelen = strlen(matroska_doctypes[i]);
913  if (total < probelen)
914  continue;
915  for (n = 4+size; n <= 4+size+total-probelen; n++)
916  if (!memcmp(p->buf+n, matroska_doctypes[i], probelen))
917  return AVPROBE_SCORE_MAX;
918  }
919 
920  // probably valid EBML header but no recognized doctype
921  return AVPROBE_SCORE_MAX/2;
922 }
923 
925  int num)
926 {
927  MatroskaTrack *tracks = matroska->tracks.elem;
928  int i;
929 
930  for (i=0; i < matroska->tracks.nb_elem; i++)
931  if (tracks[i].num == num)
932  return &tracks[i];
933 
934  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
935  return NULL;
936 }
937 
938 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
939  MatroskaTrack *track)
940 {
941  MatroskaTrackEncoding *encodings = track->encodings.elem;
942  uint8_t* data = *buf;
943  int isize = *buf_size;
944  uint8_t* pkt_data = NULL;
945  uint8_t* newpktdata;
946  int pkt_size = isize;
947  int result = 0;
948  int olen;
949 
950  if (pkt_size >= 10000000)
951  return -1;
952 
953  switch (encodings[0].compression.algo) {
955  return encodings[0].compression.settings.size;
957  do {
958  olen = pkt_size *= 3;
959  pkt_data = av_realloc(pkt_data, pkt_size+AV_LZO_OUTPUT_PADDING);
960  result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
961  } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
962  if (result)
963  goto failed;
964  pkt_size -= olen;
965  break;
966 #if CONFIG_ZLIB
968  z_stream zstream = {0};
969  if (inflateInit(&zstream) != Z_OK)
970  return -1;
971  zstream.next_in = data;
972  zstream.avail_in = isize;
973  do {
974  pkt_size *= 3;
975  newpktdata = av_realloc(pkt_data, pkt_size);
976  if (!newpktdata) {
977  inflateEnd(&zstream);
978  goto failed;
979  }
980  pkt_data = newpktdata;
981  zstream.avail_out = pkt_size - zstream.total_out;
982  zstream.next_out = pkt_data + zstream.total_out;
983  result = inflate(&zstream, Z_NO_FLUSH);
984  } while (result==Z_OK && pkt_size<10000000);
985  pkt_size = zstream.total_out;
986  inflateEnd(&zstream);
987  if (result != Z_STREAM_END)
988  goto failed;
989  break;
990  }
991 #endif
992 #if CONFIG_BZLIB
994  bz_stream bzstream = {0};
995  if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
996  return -1;
997  bzstream.next_in = data;
998  bzstream.avail_in = isize;
999  do {
1000  pkt_size *= 3;
1001  newpktdata = av_realloc(pkt_data, pkt_size);
1002  if (!newpktdata) {
1003  BZ2_bzDecompressEnd(&bzstream);
1004  goto failed;
1005  }
1006  pkt_data = newpktdata;
1007  bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1008  bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1009  result = BZ2_bzDecompress(&bzstream);
1010  } while (result==BZ_OK && pkt_size<10000000);
1011  pkt_size = bzstream.total_out_lo32;
1012  BZ2_bzDecompressEnd(&bzstream);
1013  if (result != BZ_STREAM_END)
1014  goto failed;
1015  break;
1016  }
1017 #endif
1018  default:
1019  return -1;
1020  }
1021 
1022  *buf = pkt_data;
1023  *buf_size = pkt_size;
1024  return 0;
1025  failed:
1026  av_free(pkt_data);
1027  return -1;
1028 }
1029 
1031  AVPacket *pkt, uint64_t display_duration)
1032 {
1033  char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
1034  for (; *ptr!=',' && ptr<end-1; ptr++);
1035  if (*ptr == ',')
1036  layer = ++ptr;
1037  for (; *ptr!=',' && ptr<end-1; ptr++);
1038  if (*ptr == ',') {
1039  int64_t end_pts = pkt->pts + display_duration;
1040  int sc = matroska->time_scale * pkt->pts / 10000000;
1041  int ec = matroska->time_scale * end_pts / 10000000;
1042  int sh, sm, ss, eh, em, es, len;
1043  sh = sc/360000; sc -= 360000*sh;
1044  sm = sc/ 6000; sc -= 6000*sm;
1045  ss = sc/ 100; sc -= 100*ss;
1046  eh = ec/360000; ec -= 360000*eh;
1047  em = ec/ 6000; ec -= 6000*em;
1048  es = ec/ 100; ec -= 100*es;
1049  *ptr++ = '\0';
1050  len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
1051  if (!(line = av_malloc(len)))
1052  return;
1053  snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
1054  layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
1055  av_free(pkt->data);
1056  pkt->data = line;
1057  pkt->size = strlen(line);
1058  }
1059 }
1060 
1062 {
1063  void *newdata = av_realloc(out->data, out->size+in->size);
1064  if (!newdata)
1065  return AVERROR(ENOMEM);
1066  out->data = newdata;
1067  memcpy(out->data+out->size, in->data, in->size);
1068  out->size += in->size;
1069  av_destruct_packet(in);
1070  av_free(in);
1071  return 0;
1072 }
1073 
1075  AVDictionary **metadata, char *prefix)
1076 {
1077  MatroskaTag *tags = list->elem;
1078  char key[1024];
1079  int i;
1080 
1081  for (i=0; i < list->nb_elem; i++) {
1082  const char *lang = tags[i].lang && strcmp(tags[i].lang, "und") ?
1083  tags[i].lang : NULL;
1084 
1085  if (!tags[i].name) {
1086  av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1087  continue;
1088  }
1089  if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1090  else av_strlcpy(key, tags[i].name, sizeof(key));
1091  if (tags[i].def || !lang) {
1092  av_dict_set(metadata, key, tags[i].string, 0);
1093  if (tags[i].sub.nb_elem)
1094  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1095  }
1096  if (lang) {
1097  av_strlcat(key, "-", sizeof(key));
1098  av_strlcat(key, lang, sizeof(key));
1099  av_dict_set(metadata, key, tags[i].string, 0);
1100  if (tags[i].sub.nb_elem)
1101  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1102  }
1103  }
1105 }
1106 
1108 {
1109  MatroskaDemuxContext *matroska = s->priv_data;
1110  MatroskaTags *tags = matroska->tags.elem;
1111  int i, j;
1112 
1113  for (i=0; i < matroska->tags.nb_elem; i++) {
1114  if (tags[i].target.attachuid) {
1115  MatroskaAttachement *attachment = matroska->attachments.elem;
1116  for (j=0; j<matroska->attachments.nb_elem; j++)
1117  if (attachment[j].uid == tags[i].target.attachuid
1118  && attachment[j].stream)
1119  matroska_convert_tag(s, &tags[i].tag,
1120  &attachment[j].stream->metadata, NULL);
1121  } else if (tags[i].target.chapteruid) {
1122  MatroskaChapter *chapter = matroska->chapters.elem;
1123  for (j=0; j<matroska->chapters.nb_elem; j++)
1124  if (chapter[j].uid == tags[i].target.chapteruid
1125  && chapter[j].chapter)
1126  matroska_convert_tag(s, &tags[i].tag,
1127  &chapter[j].chapter->metadata, NULL);
1128  } else if (tags[i].target.trackuid) {
1129  MatroskaTrack *track = matroska->tracks.elem;
1130  for (j=0; j<matroska->tracks.nb_elem; j++)
1131  if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1132  matroska_convert_tag(s, &tags[i].tag,
1133  &track[j].stream->metadata, NULL);
1134  } else {
1135  matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1136  tags[i].target.type);
1137  }
1138  }
1139 }
1140 
1142 {
1143  EbmlList *seekhead_list = &matroska->seekhead;
1144  MatroskaSeekhead *seekhead = seekhead_list->elem;
1145  uint32_t level_up = matroska->level_up;
1146  int64_t before_pos = avio_tell(matroska->ctx->pb);
1147  uint32_t saved_id = matroska->current_id;
1149  int64_t offset;
1150  int ret = 0;
1151 
1152  if (idx >= seekhead_list->nb_elem
1153  || seekhead[idx].id == MATROSKA_ID_SEEKHEAD
1154  || seekhead[idx].id == MATROSKA_ID_CLUSTER)
1155  return 0;
1156 
1157  /* seek */
1158  offset = seekhead[idx].pos + matroska->segment_start;
1159  if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1160  /* We don't want to lose our seekhead level, so we add
1161  * a dummy. This is a crude hack. */
1162  if (matroska->num_levels == EBML_MAX_DEPTH) {
1163  av_log(matroska->ctx, AV_LOG_INFO,
1164  "Max EBML element depth (%d) reached, "
1165  "cannot parse further.\n", EBML_MAX_DEPTH);
1166  ret = AVERROR_INVALIDDATA;
1167  } else {
1168  level.start = 0;
1169  level.length = (uint64_t)-1;
1170  matroska->levels[matroska->num_levels] = level;
1171  matroska->num_levels++;
1172  matroska->current_id = 0;
1173 
1174  ret = ebml_parse(matroska, matroska_segment, matroska);
1175 
1176  /* remove dummy level */
1177  while (matroska->num_levels) {
1178  uint64_t length = matroska->levels[--matroska->num_levels].length;
1179  if (length == (uint64_t)-1)
1180  break;
1181  }
1182  }
1183  }
1184  /* seek back */
1185  avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1186  matroska->level_up = level_up;
1187  matroska->current_id = saved_id;
1188 
1189  return ret;
1190 }
1191 
1193 {
1194  EbmlList *seekhead_list = &matroska->seekhead;
1195  int64_t before_pos = avio_tell(matroska->ctx->pb);
1196  int i;
1197 
1198  // we should not do any seeking in the streaming case
1199  if (!matroska->ctx->pb->seekable ||
1200  (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1201  return;
1202 
1203  for (i = 0; i < seekhead_list->nb_elem; i++) {
1204  MatroskaSeekhead *seekhead = seekhead_list->elem;
1205  if (seekhead[i].pos <= before_pos)
1206  continue;
1207 
1208  // defer cues parsing until we actually need cue data.
1209  if (seekhead[i].id == MATROSKA_ID_CUES) {
1210  matroska->cues_parsing_deferred = 1;
1211  continue;
1212  }
1213 
1214  if (matroska_parse_seekhead_entry(matroska, i) < 0)
1215  break;
1216  }
1217 }
1218 
1220  EbmlList *seekhead_list = &matroska->seekhead;
1221  MatroskaSeekhead *seekhead = seekhead_list->elem;
1222  EbmlList *index_list;
1224  int index_scale = 1;
1225  int i, j;
1226 
1227  for (i = 0; i < seekhead_list->nb_elem; i++)
1228  if (seekhead[i].id == MATROSKA_ID_CUES)
1229  break;
1230  assert(i <= seekhead_list->nb_elem);
1231 
1232  matroska_parse_seekhead_entry(matroska, i);
1233 
1234  index_list = &matroska->index;
1235  index = index_list->elem;
1236  if (index_list->nb_elem
1237  && index[0].time > 1E14/matroska->time_scale) {
1238  av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1239  index_scale = matroska->time_scale;
1240  }
1241  for (i = 0; i < index_list->nb_elem; i++) {
1242  EbmlList *pos_list = &index[i].pos;
1243  MatroskaIndexPos *pos = pos_list->elem;
1244  for (j = 0; j < pos_list->nb_elem; j++) {
1245  MatroskaTrack *track = matroska_find_track_by_num(matroska, pos[j].track);
1246  if (track && track->stream)
1247  av_add_index_entry(track->stream,
1248  pos[j].pos + matroska->segment_start,
1249  index[i].time/index_scale, 0, 0,
1251  }
1252  }
1253 }
1254 
1256 {
1257  static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1258  int profile;
1259 
1260  for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1261  if (strstr(codec_id, aac_profiles[profile]))
1262  break;
1263  return profile + 1;
1264 }
1265 
1266 static int matroska_aac_sri(int samplerate)
1267 {
1268  int sri;
1269 
1270  for (sri=0; sri<FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1271  if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1272  break;
1273  return sri;
1274 }
1275 
1277 {
1278  MatroskaDemuxContext *matroska = s->priv_data;
1279  EbmlList *attachements_list = &matroska->attachments;
1280  MatroskaAttachement *attachements;
1281  EbmlList *chapters_list = &matroska->chapters;
1282  MatroskaChapter *chapters;
1283  MatroskaTrack *tracks;
1284  uint64_t max_start = 0;
1285  Ebml ebml = { 0 };
1286  AVStream *st;
1287  int i, j, res;
1288 
1289  matroska->ctx = s;
1290 
1291  /* First read the EBML header. */
1292  if (ebml_parse(matroska, ebml_syntax, &ebml)
1293  || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1294  || ebml.id_length > sizeof(uint32_t) || ebml.doctype_version > 2) {
1295  av_log(matroska->ctx, AV_LOG_ERROR,
1296  "EBML header using unsupported features\n"
1297  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1298  ebml.version, ebml.doctype, ebml.doctype_version);
1299  ebml_free(ebml_syntax, &ebml);
1300  return AVERROR_PATCHWELCOME;
1301  }
1302  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
1303  if (!strcmp(ebml.doctype, matroska_doctypes[i]))
1304  break;
1305  if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
1306  av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
1307  }
1308  ebml_free(ebml_syntax, &ebml);
1309 
1310  /* The next thing is a segment. */
1311  if ((res = ebml_parse(matroska, matroska_segments, matroska)) < 0)
1312  return res;
1313  matroska_execute_seekhead(matroska);
1314 
1315  if (!matroska->time_scale)
1316  matroska->time_scale = 1000000;
1317  if (matroska->duration)
1318  matroska->ctx->duration = matroska->duration * matroska->time_scale
1319  * 1000 / AV_TIME_BASE;
1320  av_dict_set(&s->metadata, "title", matroska->title, 0);
1321 
1322  tracks = matroska->tracks.elem;
1323  for (i=0; i < matroska->tracks.nb_elem; i++) {
1324  MatroskaTrack *track = &tracks[i];
1326  EbmlList *encodings_list = &tracks->encodings;
1327  MatroskaTrackEncoding *encodings = encodings_list->elem;
1328  uint8_t *extradata = NULL;
1329  int extradata_size = 0;
1330  int extradata_offset = 0;
1331  AVIOContext b;
1332 
1333  /* Apply some sanity checks. */
1334  if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1335  track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1336  track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1337  av_log(matroska->ctx, AV_LOG_INFO,
1338  "Unknown or unsupported track type %"PRIu64"\n",
1339  track->type);
1340  continue;
1341  }
1342  if (track->codec_id == NULL)
1343  continue;
1344 
1345  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1346  if (!track->default_duration)
1347  track->default_duration = 1000000000/track->video.frame_rate;
1348  if (!track->video.display_width)
1349  track->video.display_width = track->video.pixel_width;
1350  if (!track->video.display_height)
1351  track->video.display_height = track->video.pixel_height;
1352  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1353  if (!track->audio.out_samplerate)
1354  track->audio.out_samplerate = track->audio.samplerate;
1355  }
1356  if (encodings_list->nb_elem > 1) {
1357  av_log(matroska->ctx, AV_LOG_ERROR,
1358  "Multiple combined encodings not supported");
1359  } else if (encodings_list->nb_elem == 1) {
1360  if (encodings[0].type ||
1361  (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1362 #if CONFIG_ZLIB
1364 #endif
1365 #if CONFIG_BZLIB
1367 #endif
1369  encodings[0].scope = 0;
1370  av_log(matroska->ctx, AV_LOG_ERROR,
1371  "Unsupported encoding type");
1372  } else if (track->codec_priv.size && encodings[0].scope&2) {
1373  uint8_t *codec_priv = track->codec_priv.data;
1374  int offset = matroska_decode_buffer(&track->codec_priv.data,
1375  &track->codec_priv.size,
1376  track);
1377  if (offset < 0) {
1378  track->codec_priv.data = NULL;
1379  track->codec_priv.size = 0;
1380  av_log(matroska->ctx, AV_LOG_ERROR,
1381  "Failed to decode codec private data\n");
1382  } else if (offset > 0) {
1383  track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
1384  memcpy(track->codec_priv.data,
1385  encodings[0].compression.settings.data, offset);
1386  memcpy(track->codec_priv.data+offset, codec_priv,
1387  track->codec_priv.size);
1388  track->codec_priv.size += offset;
1389  }
1390  if (codec_priv != track->codec_priv.data)
1391  av_free(codec_priv);
1392  }
1393  }
1394 
1395  for(j=0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++){
1396  if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1397  strlen(ff_mkv_codec_tags[j].str))){
1398  codec_id= ff_mkv_codec_tags[j].id;
1399  break;
1400  }
1401  }
1402 
1403  st = track->stream = avformat_new_stream(s, NULL);
1404  if (st == NULL)
1405  return AVERROR(ENOMEM);
1406 
1407  if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1408  && track->codec_priv.size >= 40
1409  && track->codec_priv.data != NULL) {
1410  track->ms_compat = 1;
1411  track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1412  codec_id = ff_codec_get_id(ff_codec_bmp_tags, track->video.fourcc);
1413  extradata_offset = 40;
1414  } else if (!strcmp(track->codec_id, "A_MS/ACM")
1415  && track->codec_priv.size >= 14
1416  && track->codec_priv.data != NULL) {
1417  int ret;
1418  ffio_init_context(&b, track->codec_priv.data, track->codec_priv.size,
1419  0, NULL, NULL, NULL, NULL);
1420  ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1421  if (ret < 0)
1422  return ret;
1423  codec_id = st->codec->codec_id;
1424  extradata_offset = FFMIN(track->codec_priv.size, 18);
1425  } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1426  && (track->codec_priv.size >= 86)
1427  && (track->codec_priv.data != NULL)) {
1428  track->video.fourcc = AV_RL32(track->codec_priv.data);
1429  codec_id=ff_codec_get_id(codec_movvideo_tags, track->video.fourcc);
1430  } else if (codec_id == CODEC_ID_PCM_S16BE) {
1431  switch (track->audio.bitdepth) {
1432  case 8: codec_id = CODEC_ID_PCM_U8; break;
1433  case 24: codec_id = CODEC_ID_PCM_S24BE; break;
1434  case 32: codec_id = CODEC_ID_PCM_S32BE; break;
1435  }
1436  } else if (codec_id == CODEC_ID_PCM_S16LE) {
1437  switch (track->audio.bitdepth) {
1438  case 8: codec_id = CODEC_ID_PCM_U8; break;
1439  case 24: codec_id = CODEC_ID_PCM_S24LE; break;
1440  case 32: codec_id = CODEC_ID_PCM_S32LE; break;
1441  }
1442  } else if (codec_id==CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1443  codec_id = CODEC_ID_PCM_F64LE;
1444  } else if (codec_id == CODEC_ID_AAC && !track->codec_priv.size) {
1445  int profile = matroska_aac_profile(track->codec_id);
1446  int sri = matroska_aac_sri(track->audio.samplerate);
1447  extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1448  if (extradata == NULL)
1449  return AVERROR(ENOMEM);
1450  extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1451  extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1452  if (strstr(track->codec_id, "SBR")) {
1453  sri = matroska_aac_sri(track->audio.out_samplerate);
1454  extradata[2] = 0x56;
1455  extradata[3] = 0xE5;
1456  extradata[4] = 0x80 | (sri<<3);
1457  extradata_size = 5;
1458  } else
1459  extradata_size = 2;
1460  } else if (codec_id == CODEC_ID_TTA) {
1461  extradata_size = 30;
1462  extradata = av_mallocz(extradata_size);
1463  if (extradata == NULL)
1464  return AVERROR(ENOMEM);
1465  ffio_init_context(&b, extradata, extradata_size, 1,
1466  NULL, NULL, NULL, NULL);
1467  avio_write(&b, "TTA1", 4);
1468  avio_wl16(&b, 1);
1469  avio_wl16(&b, track->audio.channels);
1470  avio_wl16(&b, track->audio.bitdepth);
1471  avio_wl32(&b, track->audio.out_samplerate);
1472  avio_wl32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1473  } else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
1474  codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
1475  extradata_offset = 26;
1476  } else if (codec_id == CODEC_ID_RA_144) {
1477  track->audio.out_samplerate = 8000;
1478  track->audio.channels = 1;
1479  } else if (codec_id == CODEC_ID_RA_288 || codec_id == CODEC_ID_COOK ||
1480  codec_id == CODEC_ID_ATRAC3 || codec_id == CODEC_ID_SIPR) {
1481  int flavor;
1482  ffio_init_context(&b, track->codec_priv.data,track->codec_priv.size,
1483  0, NULL, NULL, NULL, NULL);
1484  avio_skip(&b, 22);
1485  flavor = avio_rb16(&b);
1486  track->audio.coded_framesize = avio_rb32(&b);
1487  avio_skip(&b, 12);
1488  track->audio.sub_packet_h = avio_rb16(&b);
1489  track->audio.frame_size = avio_rb16(&b);
1490  track->audio.sub_packet_size = avio_rb16(&b);
1491  track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1492  if (codec_id == CODEC_ID_RA_288) {
1493  st->codec->block_align = track->audio.coded_framesize;
1494  track->codec_priv.size = 0;
1495  } else {
1496  if (codec_id == CODEC_ID_SIPR && flavor < 4) {
1497  const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1498  track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1499  st->codec->bit_rate = sipr_bit_rate[flavor];
1500  }
1501  st->codec->block_align = track->audio.sub_packet_size;
1502  extradata_offset = 78;
1503  }
1504  }
1505  track->codec_priv.size -= extradata_offset;
1506 
1507  if (codec_id == CODEC_ID_NONE)
1508  av_log(matroska->ctx, AV_LOG_INFO,
1509  "Unknown/unsupported CodecID %s.\n", track->codec_id);
1510 
1511  if (track->time_scale < 0.01)
1512  track->time_scale = 1.0;
1513  avpriv_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1514 
1515  st->codec->codec_id = codec_id;
1516  st->start_time = 0;
1517  if (strcmp(track->language, "und"))
1518  av_dict_set(&st->metadata, "language", track->language, 0);
1519  av_dict_set(&st->metadata, "title", track->name, 0);
1520 
1521  if (track->flag_default)
1523  if (track->flag_forced)
1525 
1526  if (!st->codec->extradata) {
1527  if(extradata){
1528  st->codec->extradata = extradata;
1529  st->codec->extradata_size = extradata_size;
1530  } else if(track->codec_priv.data && track->codec_priv.size > 0){
1531  st->codec->extradata = av_mallocz(track->codec_priv.size +
1533  if(st->codec->extradata == NULL)
1534  return AVERROR(ENOMEM);
1535  st->codec->extradata_size = track->codec_priv.size;
1536  memcpy(st->codec->extradata,
1537  track->codec_priv.data + extradata_offset,
1538  track->codec_priv.size);
1539  }
1540  }
1541 
1542  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1544  st->codec->codec_tag = track->video.fourcc;
1545  st->codec->width = track->video.pixel_width;
1546  st->codec->height = track->video.pixel_height;
1548  &st->sample_aspect_ratio.den,
1549  st->codec->height * track->video.display_width,
1550  st->codec-> width * track->video.display_height,
1551  255);
1552  if (st->codec->codec_id != CODEC_ID_H264)
1554  if (track->default_duration)
1555  st->avg_frame_rate = av_d2q(1000000000.0/track->default_duration, INT_MAX);
1556  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1558  st->codec->sample_rate = track->audio.out_samplerate;
1559  st->codec->channels = track->audio.channels;
1560  if (st->codec->codec_id != CODEC_ID_AAC)
1562  } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1564  }
1565  }
1566 
1567  attachements = attachements_list->elem;
1568  for (j=0; j<attachements_list->nb_elem; j++) {
1569  if (!(attachements[j].filename && attachements[j].mime &&
1570  attachements[j].bin.data && attachements[j].bin.size > 0)) {
1571  av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1572  } else {
1573  AVStream *st = avformat_new_stream(s, NULL);
1574  if (st == NULL)
1575  break;
1576  av_dict_set(&st->metadata, "filename",attachements[j].filename, 0);
1577  av_dict_set(&st->metadata, "mimetype", attachements[j].mime, 0);
1578  st->codec->codec_id = CODEC_ID_NONE;
1580  st->codec->extradata = av_malloc(attachements[j].bin.size);
1581  if(st->codec->extradata == NULL)
1582  break;
1583  st->codec->extradata_size = attachements[j].bin.size;
1584  memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1585 
1586  for (i=0; ff_mkv_mime_tags[i].id != CODEC_ID_NONE; i++) {
1587  if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1588  strlen(ff_mkv_mime_tags[i].str))) {
1589  st->codec->codec_id = ff_mkv_mime_tags[i].id;
1590  break;
1591  }
1592  }
1593  attachements[j].stream = st;
1594  }
1595  }
1596 
1597  chapters = chapters_list->elem;
1598  for (i=0; i<chapters_list->nb_elem; i++)
1599  if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1600  && (max_start==0 || chapters[i].start > max_start)) {
1601  chapters[i].chapter =
1602  avpriv_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1603  chapters[i].start, chapters[i].end,
1604  chapters[i].title);
1605  av_dict_set(&chapters[i].chapter->metadata,
1606  "title", chapters[i].title, 0);
1607  max_start = chapters[i].start;
1608  }
1609 
1611 
1612  return 0;
1613 }
1614 
1615 /*
1616  * Put one packet in an application-supplied AVPacket struct.
1617  * Returns 0 on success or -1 on failure.
1618  */
1620  AVPacket *pkt)
1621 {
1622  if (matroska->num_packets > 0) {
1623  memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1624  av_free(matroska->packets[0]);
1625  if (matroska->num_packets > 1) {
1626  void *newpackets;
1627  memmove(&matroska->packets[0], &matroska->packets[1],
1628  (matroska->num_packets - 1) * sizeof(AVPacket *));
1629  newpackets = av_realloc(matroska->packets,
1630  (matroska->num_packets - 1) * sizeof(AVPacket *));
1631  if (newpackets)
1632  matroska->packets = newpackets;
1633  } else {
1634  av_freep(&matroska->packets);
1635  }
1636  matroska->num_packets--;
1637  return 0;
1638  }
1639 
1640  return -1;
1641 }
1642 
1643 /*
1644  * Free all packets in our internal queue.
1645  */
1647 {
1648  if (matroska->packets) {
1649  int n;
1650  for (n = 0; n < matroska->num_packets; n++) {
1651  av_free_packet(matroska->packets[n]);
1652  av_free(matroska->packets[n]);
1653  }
1654  av_freep(&matroska->packets);
1655  matroska->num_packets = 0;
1656  }
1657 }
1658 
1659 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
1660  int size, int64_t pos, uint64_t cluster_time,
1661  uint64_t duration, int is_keyframe,
1662  int64_t cluster_pos)
1663 {
1664  uint64_t timecode = AV_NOPTS_VALUE;
1665  MatroskaTrack *track;
1666  int res = 0;
1667  AVStream *st;
1668  AVPacket *pkt;
1669  int16_t block_time;
1670  uint32_t *lace_size = NULL;
1671  int n, flags, laces = 0;
1672  uint64_t num;
1673 
1674  if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
1675  av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
1676  return res;
1677  }
1678  data += n;
1679  size -= n;
1680 
1681  track = matroska_find_track_by_num(matroska, num);
1682  if (!track || !track->stream) {
1683  av_log(matroska->ctx, AV_LOG_INFO,
1684  "Invalid stream %"PRIu64" or size %u\n", num, size);
1685  return AVERROR_INVALIDDATA;
1686  } else if (size <= 3)
1687  return 0;
1688  st = track->stream;
1689  if (st->discard >= AVDISCARD_ALL)
1690  return res;
1691  if (duration == AV_NOPTS_VALUE)
1692  duration = track->default_duration / matroska->time_scale;
1693 
1694  block_time = AV_RB16(data);
1695  data += 2;
1696  flags = *data++;
1697  size -= 3;
1698  if (is_keyframe == -1)
1699  is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
1700 
1701  if (cluster_time != (uint64_t)-1
1702  && (block_time >= 0 || cluster_time >= -block_time)) {
1703  timecode = cluster_time + block_time;
1704  if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
1705  && timecode < track->end_timecode)
1706  is_keyframe = 0; /* overlapping subtitles are not key frame */
1707  if (is_keyframe)
1708  av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
1709  track->end_timecode = FFMAX(track->end_timecode, timecode+duration);
1710  }
1711 
1712  if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1713  if (!is_keyframe || timecode < matroska->skip_to_timecode)
1714  return res;
1715  matroska->skip_to_keyframe = 0;
1716  }
1717 
1718  switch ((flags & 0x06) >> 1) {
1719  case 0x0: /* no lacing */
1720  laces = 1;
1721  lace_size = av_mallocz(sizeof(int));
1722  lace_size[0] = size;
1723  break;
1724 
1725  case 0x1: /* Xiph lacing */
1726  case 0x2: /* fixed-size lacing */
1727  case 0x3: /* EBML lacing */
1728  assert(size>0); // size <=3 is checked before size-=3 above
1729  laces = (*data) + 1;
1730  data += 1;
1731  size -= 1;
1732  lace_size = av_mallocz(laces * sizeof(int));
1733 
1734  switch ((flags & 0x06) >> 1) {
1735  case 0x1: /* Xiph lacing */ {
1736  uint8_t temp;
1737  uint32_t total = 0;
1738  for (n = 0; res == 0 && n < laces - 1; n++) {
1739  while (1) {
1740  if (size == 0) {
1741  res = -1;
1742  break;
1743  }
1744  temp = *data;
1745  lace_size[n] += temp;
1746  data += 1;
1747  size -= 1;
1748  if (temp != 0xff)
1749  break;
1750  }
1751  total += lace_size[n];
1752  }
1753  lace_size[n] = size - total;
1754  break;
1755  }
1756 
1757  case 0x2: /* fixed-size lacing */
1758  for (n = 0; n < laces; n++)
1759  lace_size[n] = size / laces;
1760  break;
1761 
1762  case 0x3: /* EBML lacing */ {
1763  uint32_t total;
1764  n = matroska_ebmlnum_uint(matroska, data, size, &num);
1765  if (n < 0) {
1766  av_log(matroska->ctx, AV_LOG_INFO,
1767  "EBML block data error\n");
1768  break;
1769  }
1770  data += n;
1771  size -= n;
1772  total = lace_size[0] = num;
1773  for (n = 1; res == 0 && n < laces - 1; n++) {
1774  int64_t snum;
1775  int r;
1776  r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1777  if (r < 0) {
1778  av_log(matroska->ctx, AV_LOG_INFO,
1779  "EBML block data error\n");
1780  break;
1781  }
1782  data += r;
1783  size -= r;
1784  lace_size[n] = lace_size[n - 1] + snum;
1785  total += lace_size[n];
1786  }
1787  lace_size[laces - 1] = size - total;
1788  break;
1789  }
1790  }
1791  break;
1792  }
1793 
1794  if (res == 0) {
1795  for (n = 0; n < laces; n++) {
1796  if ((st->codec->codec_id == CODEC_ID_RA_288 ||
1797  st->codec->codec_id == CODEC_ID_COOK ||
1798  st->codec->codec_id == CODEC_ID_SIPR ||
1799  st->codec->codec_id == CODEC_ID_ATRAC3) &&
1800  st->codec->block_align && track->audio.sub_packet_size) {
1801  int a = st->codec->block_align;
1802  int sps = track->audio.sub_packet_size;
1803  int cfs = track->audio.coded_framesize;
1804  int h = track->audio.sub_packet_h;
1805  int y = track->audio.sub_packet_cnt;
1806  int w = track->audio.frame_size;
1807  int x;
1808 
1809  if (!track->audio.pkt_cnt) {
1810  if (track->audio.sub_packet_cnt == 0)
1811  track->audio.buf_timecode = timecode;
1812  if (st->codec->codec_id == CODEC_ID_RA_288) {
1813  if (size < cfs * h / 2) {
1814  av_log(matroska->ctx, AV_LOG_ERROR,
1815  "Corrupt int4 RM-style audio packet size\n");
1816  return AVERROR_INVALIDDATA;
1817  }
1818  for (x=0; x<h/2; x++)
1819  memcpy(track->audio.buf+x*2*w+y*cfs,
1820  data+x*cfs, cfs);
1821  } else if (st->codec->codec_id == CODEC_ID_SIPR) {
1822  if (size < w) {
1823  av_log(matroska->ctx, AV_LOG_ERROR,
1824  "Corrupt sipr RM-style audio packet size\n");
1825  return AVERROR_INVALIDDATA;
1826  }
1827  memcpy(track->audio.buf + y*w, data, w);
1828  } else {
1829  if (size < sps * w / sps) {
1830  av_log(matroska->ctx, AV_LOG_ERROR,
1831  "Corrupt generic RM-style audio packet size\n");
1832  return AVERROR_INVALIDDATA;
1833  }
1834  for (x=0; x<w/sps; x++)
1835  memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
1836  }
1837 
1838  if (++track->audio.sub_packet_cnt >= h) {
1839  if (st->codec->codec_id == CODEC_ID_SIPR)
1840  ff_rm_reorder_sipr_data(track->audio.buf, h, w);
1841  track->audio.sub_packet_cnt = 0;
1842  track->audio.pkt_cnt = h*w / a;
1843  }
1844  }
1845  while (track->audio.pkt_cnt) {
1846  pkt = av_mallocz(sizeof(AVPacket));
1847  av_new_packet(pkt, a);
1848  memcpy(pkt->data, track->audio.buf
1849  + a * (h*w / a - track->audio.pkt_cnt--), a);
1850  pkt->pts = track->audio.buf_timecode;
1852  pkt->pos = pos;
1853  pkt->stream_index = st->index;
1854  dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1855  }
1856  } else {
1857  MatroskaTrackEncoding *encodings = track->encodings.elem;
1858  int offset = 0, pkt_size = lace_size[n];
1859  uint8_t *pkt_data = data;
1860 
1861  if (pkt_size > size) {
1862  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
1863  break;
1864  }
1865 
1866  if (encodings && encodings->scope & 1) {
1867  offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
1868  if (offset < 0)
1869  continue;
1870  }
1871 
1872  pkt = av_mallocz(sizeof(AVPacket));
1873  /* XXX: prevent data copy... */
1874  if (av_new_packet(pkt, pkt_size+offset) < 0) {
1875  av_free(pkt);
1876  res = AVERROR(ENOMEM);
1877  break;
1878  }
1879  if (offset)
1880  memcpy (pkt->data, encodings->compression.settings.data, offset);
1881  memcpy (pkt->data+offset, pkt_data, pkt_size);
1882 
1883  if (pkt_data != data)
1884  av_free(pkt_data);
1885 
1886  if (n == 0)
1887  pkt->flags = is_keyframe;
1888  pkt->stream_index = st->index;
1889 
1890  if (track->ms_compat)
1891  pkt->dts = timecode;
1892  else
1893  pkt->pts = timecode;
1894  pkt->pos = pos;
1895  if (st->codec->codec_id == CODEC_ID_TEXT)
1897  else if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE)
1898  pkt->duration = duration;
1899 
1900  if (st->codec->codec_id == CODEC_ID_SSA)
1901  matroska_fix_ass_packet(matroska, pkt, duration);
1902 
1903  if (matroska->prev_pkt &&
1904  timecode != AV_NOPTS_VALUE &&
1905  matroska->prev_pkt->pts == timecode &&
1906  matroska->prev_pkt->stream_index == st->index &&
1907  st->codec->codec_id == CODEC_ID_SSA)
1908  matroska_merge_packets(matroska->prev_pkt, pkt);
1909  else {
1910  dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1911  matroska->prev_pkt = pkt;
1912  }
1913  }
1914 
1915  if (timecode != AV_NOPTS_VALUE)
1916  timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
1917  data += lace_size[n];
1918  size -= lace_size[n];
1919  }
1920  }
1921 
1922  av_free(lace_size);
1923  return res;
1924 }
1925 
1927 {
1928  MatroskaCluster cluster = { 0 };
1929  EbmlList *blocks_list;
1930  MatroskaBlock *blocks;
1931  int i, res;
1932  int64_t pos = avio_tell(matroska->ctx->pb);
1933  matroska->prev_pkt = NULL;
1934  if (matroska->current_id)
1935  pos -= 4; /* sizeof the ID which was already read */
1936  res = ebml_parse(matroska, matroska_clusters, &cluster);
1937  blocks_list = &cluster.blocks;
1938  blocks = blocks_list->elem;
1939  for (i=0; i<blocks_list->nb_elem && !res; i++)
1940  if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
1941  int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
1942  if (!blocks[i].non_simple)
1943  blocks[i].duration = AV_NOPTS_VALUE;
1944  res=matroska_parse_block(matroska,
1945  blocks[i].bin.data, blocks[i].bin.size,
1946  blocks[i].bin.pos, cluster.timecode,
1947  blocks[i].duration, is_keyframe,
1948  pos);
1949  }
1950  ebml_free(matroska_cluster, &cluster);
1951  if (res < 0) matroska->done = 1;
1952  return res;
1953 }
1954 
1956 {
1957  MatroskaDemuxContext *matroska = s->priv_data;
1958  int ret = 0;
1959 
1960  while (!ret && matroska_deliver_packet(matroska, pkt)) {
1961  if (matroska->done)
1962  return AVERROR_EOF;
1963  ret = matroska_parse_cluster(matroska);
1964  }
1965 
1966  return ret;
1967 }
1968 
1969 static int matroska_read_seek(AVFormatContext *s, int stream_index,
1970  int64_t timestamp, int flags)
1971 {
1972  MatroskaDemuxContext *matroska = s->priv_data;
1973  MatroskaTrack *tracks = matroska->tracks.elem;
1974  AVStream *st = s->streams[stream_index];
1975  int i, index, index_sub, index_min;
1976 
1977  /* Parse the CUES now since we need the index data to seek. */
1978  if (matroska->cues_parsing_deferred) {
1979  matroska_parse_cues(matroska);
1980  matroska->cues_parsing_deferred = 0;
1981  }
1982 
1983  if (!st->nb_index_entries)
1984  return 0;
1985  timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
1986 
1987  if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1988  avio_seek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
1989  matroska->current_id = 0;
1990  while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1991  matroska_clear_queue(matroska);
1992  if (matroska_parse_cluster(matroska) < 0)
1993  break;
1994  }
1995  }
1996 
1997  matroska_clear_queue(matroska);
1998  if (index < 0)
1999  return 0;
2000 
2001  index_min = index;
2002  for (i=0; i < matroska->tracks.nb_elem; i++) {
2003  tracks[i].audio.pkt_cnt = 0;
2004  tracks[i].audio.sub_packet_cnt = 0;
2005  tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2006  tracks[i].end_timecode = 0;
2007  if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
2008  && !tracks[i].stream->discard != AVDISCARD_ALL) {
2009  index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
2010  if (index_sub >= 0
2011  && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
2012  && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
2013  index_min = index_sub;
2014  }
2015  }
2016 
2017  avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2018  matroska->current_id = 0;
2019  matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
2020  matroska->skip_to_timecode = st->index_entries[index].timestamp;
2021  matroska->done = 0;
2022  ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2023  return 0;
2024 }
2025 
2027 {
2028  MatroskaDemuxContext *matroska = s->priv_data;
2029  MatroskaTrack *tracks = matroska->tracks.elem;
2030  int n;
2031 
2032  matroska_clear_queue(matroska);
2033 
2034  for (n=0; n < matroska->tracks.nb_elem; n++)
2035  if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2036  av_free(tracks[n].audio.buf);
2037  ebml_free(matroska_segment, matroska);
2038 
2039  return 0;
2040 }
2041 
2043  .name = "matroska,webm",
2044  .long_name = NULL_IF_CONFIG_SMALL("Matroska/WebM file format"),
2045  .priv_data_size = sizeof(MatroskaDemuxContext),
2051 };