FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
matroskadec.c
Go to the documentation of this file.
1 /*
2  * Matroska file demuxer
3  * Copyright (c) 2003-2008 The FFmpeg Project
4  *
5  * This file is part of FFmpeg.
6  *
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.
11  *
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.
16  *
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
20  */
21 
22 /**
23  * @file
24  * Matroska file demuxer
25  * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
26  * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
27  * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28  * @see specs available on the Matroska project page: http://www.matroska.org/
29  */
30 
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 "rmsipr.h"
39 #include "matroska.h"
40 #include "libavcodec/bytestream.h"
41 #include "libavcodec/mpeg4audio.h"
42 #include "libavutil/base64.h"
43 #include "libavutil/intfloat.h"
44 #include "libavutil/intreadwrite.h"
45 #include "libavutil/avstring.h"
46 #include "libavutil/lzo.h"
47 #include "libavutil/dict.h"
48 #if CONFIG_ZLIB
49 #include <zlib.h>
50 #endif
51 #if CONFIG_BZLIB
52 #include <bzlib.h>
53 #endif
54 
55 typedef enum {
67 } EbmlType;
68 
69 typedef const struct EbmlSyntax {
70  uint32_t id;
74  union {
75  uint64_t u;
76  double f;
77  const char *s;
78  const struct EbmlSyntax *n;
79  } def;
80 } EbmlSyntax;
81 
82 typedef struct {
83  int nb_elem;
84  void *elem;
85 } EbmlList;
86 
87 typedef struct {
88  int size;
90  int64_t pos;
91 } EbmlBin;
92 
93 typedef struct {
94  uint64_t version;
95  uint64_t max_size;
96  uint64_t id_length;
97  char *doctype;
98  uint64_t doctype_version;
99 } Ebml;
100 
101 typedef struct {
102  uint64_t algo;
105 
106 typedef struct {
107  uint64_t algo;
110 
111 typedef struct {
112  uint64_t scope;
113  uint64_t type;
117 
118 typedef struct {
119  double frame_rate;
120  uint64_t display_width;
121  uint64_t display_height;
122  uint64_t pixel_width;
123  uint64_t pixel_height;
125  uint64_t stereo_mode;
126  uint64_t alpha_mode;
128 
129 typedef struct {
130  double samplerate;
132  uint64_t bitdepth;
133  uint64_t channels;
134 
135  /* real audio header (extracted from extradata) */
141  int pkt_cnt;
142  uint64_t buf_timecode;
145 
146 typedef struct {
147  uint64_t uid;
148  uint64_t type;
150 
151 typedef struct {
154 
155 typedef struct {
156  uint64_t num;
157  uint64_t uid;
158  uint64_t type;
159  char *name;
160  char *codec_id;
162  char *language;
163  double time_scale;
165  uint64_t flag_default;
166  uint64_t flag_forced;
167  uint64_t codec_delay;
168  uint64_t seek_preroll;
173 
175  int64_t end_timecode;
178 } MatroskaTrack;
179 
180 typedef struct {
181  uint64_t uid;
182  char *filename;
183  char *mime;
185 
188 
189 typedef struct {
190  uint64_t start;
191  uint64_t end;
192  uint64_t uid;
193  char *title;
194 
197 
198 typedef struct {
199  uint64_t track;
200  uint64_t pos;
202 
203 typedef struct {
204  uint64_t time;
206 } MatroskaIndex;
207 
208 typedef struct {
209  char *name;
210  char *string;
211  char *lang;
212  uint64_t def;
214 } MatroskaTag;
215 
216 typedef struct {
217  char *type;
218  uint64_t typevalue;
219  uint64_t trackuid;
220  uint64_t chapteruid;
221  uint64_t attachuid;
223 
224 typedef struct {
227 } MatroskaTags;
228 
229 typedef struct {
230  uint64_t id;
231  uint64_t pos;
233 
234 typedef struct {
235  uint64_t start;
236  uint64_t length;
237 } MatroskaLevel;
238 
239 typedef struct {
240  uint64_t timecode;
243 
244 typedef struct {
246 
247  /* EBML stuff */
250  int level_up;
251  uint32_t current_id;
252 
253  uint64_t time_scale;
254  double duration;
255  char *title;
263 
264  /* byte position of the segment inside the stream */
265  int64_t segment_start;
266 
267  /* the packet queue */
271 
272  int done;
273 
274  /* What to skip before effectively reading a packet. */
277 
278  /* File has a CUES element, but we defer parsing until it is needed. */
280 
284 
285  /* File has SSA subtitles which prevent incremental cluster parsing. */
288 
289 typedef struct {
290  uint64_t duration;
291  int64_t reference;
292  uint64_t non_simple;
294  uint64_t additional_id;
297 } MatroskaBlock;
298 
300  { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
301  { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
302  { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
303  { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
304  { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
307  { 0 }
308 };
309 
311  { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
312  { 0 }
313 };
314 
316  { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
318  { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
321  { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext,date_utc) },
323  { 0 }
324 };
325 
327  { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
328  { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width), {.u=-1} },
329  { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height), {.u=-1} },
330  { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
331  { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
332  { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo,color_space) },
333  { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,stereo_mode) },
334  { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,alpha_mode) },
342  { 0 }
343 };
344 
346  { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
347  { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
349  { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
350  { 0 }
351 };
352 
356  { 0 }
357 };
358 
367  { 0 }
368 };
370  { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
372  { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
373  { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding,encryption), {.n=matroska_track_encoding_encryption} },
375  { 0 }
376 };
377 
379  { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
380  { 0 }
381 };
382 
386  { 0 }
387 };
388 
390  { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n=matroska_track_plane} },
391  { 0 }
392 };
393 
395  { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n=matroska_track_combine_planes} },
396  { 0 }
397 };
398 
400  { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
402  { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
405  { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
406  { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
407  { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
408  { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
409  { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
410  { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack,flag_forced), {.u=0} },
411  { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
412  { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
413  { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack,operation), {.n=matroska_track_operation} },
414  { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
415  { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack,max_block_additional_id) },
416  { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack,codec_delay) },
417  { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack,seek_preroll) },
426  { 0 }
427 };
428 
430  { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
431  { 0 }
432 };
433 
435  { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
436  { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
437  { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
438  { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
440  { 0 }
441 };
442 
444  { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
445  { 0 }
446 };
447 
449  { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
451  { 0 }
452 };
453 
457  { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
458  { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
463  { 0 }
464 };
465 
467  { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
472  { 0 }
473 };
474 
476  { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
477  { 0 }
478 };
479 
481  { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
486  { 0 }
487 };
488 
490  { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
491  { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
492  { 0 }
493 };
494 
496  { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
497  { 0 }
498 };
499 
501  { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
502  { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
503  { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag,lang), {.s="und"} },
504  { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag,def) },
505  { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag,def) },
506  { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
507  { 0 }
508 };
509 
512  { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
513  { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
515  { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
516  { 0 }
517 };
518 
520  { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
521  { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
522  { 0 }
523 };
524 
526  { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
527  { 0 }
528 };
529 
531  { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
532  { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
533  { 0 }
534 };
535 
537  { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
538  { 0 }
539 };
540 
542  { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
543  { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
544  { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
545  { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
546  { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
547  { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
548  { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
550  { 0 }
551 };
552 
554  { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
555  { 0 }
556 };
557 
559  { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
560  { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
561  { 0 }
562 };
563 
565  { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n=matroska_blockmore} },
566  { 0 }
567 };
568 
570  { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
571  { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, {.n=matroska_blockadditions} },
572  { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
574  { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock,discard_padding) },
575  { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock,reference) },
577  { 1, EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
578  { 0 }
579 };
580 
582  { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
583  { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
584  { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
587  { 0 }
588 };
589 
591  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
596  { 0 }
597 };
598 
600  { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
601  { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
602  { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
610  { 0 }
611 };
612 
614  { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
619  { 0 }
620 };
621 
623  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster_incremental} },
628  { 0 }
629 };
630 
631 static const char *const matroska_doctypes[] = { "matroska", "webm" };
632 
633 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
634 {
635  AVIOContext *pb = matroska->ctx->pb;
636  uint32_t id;
637  matroska->current_id = 0;
638  matroska->num_levels = 0;
639 
640  /* seek to next position to resync from */
641  if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
642  goto eof;
643 
644  id = avio_rb32(pb);
645 
646  // try to find a toplevel element
647  while (!url_feof(pb)) {
648  if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
649  id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
651  id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
652  matroska->current_id = id;
653  return 0;
654  }
655  id = (id << 8) | avio_r8(pb);
656  }
657 eof:
658  matroska->done = 1;
659  return AVERROR_EOF;
660 }
661 
662 /*
663  * Return: Whether we reached the end of a level in the hierarchy or not.
664  */
666 {
667  AVIOContext *pb = matroska->ctx->pb;
668  int64_t pos = avio_tell(pb);
669 
670  if (matroska->num_levels > 0) {
671  MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
672  if (pos - level->start >= level->length || matroska->current_id) {
673  matroska->num_levels--;
674  return 1;
675  }
676  }
677  return 0;
678 }
679 
680 /*
681  * Read: an "EBML number", which is defined as a variable-length
682  * array of bytes. The first byte indicates the length by giving a
683  * number of 0-bits followed by a one. The position of the first
684  * "one" bit inside the first byte indicates the length of this
685  * number.
686  * Returns: number of bytes read, < 0 on error
687  */
689  int max_size, uint64_t *number)
690 {
691  int read = 1, n = 1;
692  uint64_t total = 0;
693 
694  /* The first byte tells us the length in bytes - avio_r8() can normally
695  * return 0, but since that's not a valid first ebmlID byte, we can
696  * use it safely here to catch EOS. */
697  if (!(total = avio_r8(pb))) {
698  /* we might encounter EOS here */
699  if (!url_feof(pb)) {
700  int64_t pos = avio_tell(pb);
701  av_log(matroska->ctx, AV_LOG_ERROR,
702  "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
703  pos, pos);
704  return pb->error ? pb->error : AVERROR(EIO);
705  }
706  return AVERROR_EOF;
707  }
708 
709  /* get the length of the EBML number */
710  read = 8 - ff_log2_tab[total];
711  if (read > max_size) {
712  int64_t pos = avio_tell(pb) - 1;
713  av_log(matroska->ctx, AV_LOG_ERROR,
714  "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
715  (uint8_t) total, pos, pos);
716  return AVERROR_INVALIDDATA;
717  }
718 
719  /* read out length */
720  total ^= 1 << ff_log2_tab[total];
721  while (n++ < read)
722  total = (total << 8) | avio_r8(pb);
723 
724  *number = total;
725 
726  return read;
727 }
728 
729 /**
730  * Read a EBML length value.
731  * This needs special handling for the "unknown length" case which has multiple
732  * encodings.
733  */
735  uint64_t *number)
736 {
737  int res = ebml_read_num(matroska, pb, 8, number);
738  if (res > 0 && *number + 1 == 1ULL << (7 * res))
739  *number = 0xffffffffffffffULL;
740  return res;
741 }
742 
743 /*
744  * Read the next element as an unsigned int.
745  * 0 is success, < 0 is failure.
746  */
747 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
748 {
749  int n = 0;
750 
751  if (size > 8)
752  return AVERROR_INVALIDDATA;
753 
754  /* big-endian ordering; build up number */
755  *num = 0;
756  while (n++ < size)
757  *num = (*num << 8) | avio_r8(pb);
758 
759  return 0;
760 }
761 
762 /*
763  * Read the next element as a signed int.
764  * 0 is success, < 0 is failure.
765  */
766 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
767 {
768  int n = 1;
769 
770  if (size > 8)
771  return AVERROR_INVALIDDATA;
772 
773  if (size == 0) {
774  *num = 0;
775  } else {
776  *num = sign_extend(avio_r8(pb), 8);
777 
778  /* big-endian ordering; build up number */
779  while (n++ < size)
780  *num = (*num << 8) | avio_r8(pb);
781  }
782 
783  return 0;
784 }
785 
786 /*
787  * Read the next element as a float.
788  * 0 is success, < 0 is failure.
789  */
790 static int ebml_read_float(AVIOContext *pb, int size, double *num)
791 {
792  if (size == 0) {
793  *num = 0;
794  } else if (size == 4) {
795  *num = av_int2float(avio_rb32(pb));
796  } else if (size == 8){
797  *num = av_int2double(avio_rb64(pb));
798  } else
799  return AVERROR_INVALIDDATA;
800 
801  return 0;
802 }
803 
804 /*
805  * Read the next element as an ASCII string.
806  * 0 is success, < 0 is failure.
807  */
808 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
809 {
810  char *res;
811 
812  /* EBML strings are usually not 0-terminated, so we allocate one
813  * byte more, read the string and NULL-terminate it ourselves. */
814  if (!(res = av_malloc(size + 1)))
815  return AVERROR(ENOMEM);
816  if (avio_read(pb, (uint8_t *) res, size) != size) {
817  av_free(res);
818  return AVERROR(EIO);
819  }
820  (res)[size] = '\0';
821  av_free(*str);
822  *str = res;
823 
824  return 0;
825 }
826 
827 /*
828  * Read the next element as binary data.
829  * 0 is success, < 0 is failure.
830  */
831 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
832 {
833  av_fast_padded_malloc(&bin->data, &bin->size, length);
834  if (!bin->data)
835  return AVERROR(ENOMEM);
836 
837  bin->size = length;
838  bin->pos = avio_tell(pb);
839  if (avio_read(pb, bin->data, length) != length) {
840  av_freep(&bin->data);
841  bin->size = 0;
842  return AVERROR(EIO);
843  }
844 
845  return 0;
846 }
847 
848 /*
849  * Read the next element, but only the header. The contents
850  * are supposed to be sub-elements which can be read separately.
851  * 0 is success, < 0 is failure.
852  */
853 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
854 {
855  AVIOContext *pb = matroska->ctx->pb;
857 
858  if (matroska->num_levels >= EBML_MAX_DEPTH) {
859  av_log(matroska->ctx, AV_LOG_ERROR,
860  "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
861  return AVERROR(ENOSYS);
862  }
863 
864  level = &matroska->levels[matroska->num_levels++];
865  level->start = avio_tell(pb);
866  level->length = length;
867 
868  return 0;
869 }
870 
871 /*
872  * Read signed/unsigned "EBML" numbers.
873  * Return: number of bytes processed, < 0 on error
874  */
876  uint8_t *data, uint32_t size, uint64_t *num)
877 {
878  AVIOContext pb;
879  ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
880  return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
881 }
882 
883 /*
884  * Same as above, but signed.
885  */
887  uint8_t *data, uint32_t size, int64_t *num)
888 {
889  uint64_t unum;
890  int res;
891 
892  /* read as unsigned number first */
893  if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
894  return res;
895 
896  /* make signed (weird way) */
897  *num = unum - ((1LL << (7*res - 1)) - 1);
898 
899  return res;
900 }
901 
902 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
903  EbmlSyntax *syntax, void *data);
904 
905 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
906  uint32_t id, void *data)
907 {
908  int i;
909  for (i=0; syntax[i].id; i++)
910  if (id == syntax[i].id)
911  break;
912  if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
913  matroska->num_levels > 0 &&
914  matroska->levels[matroska->num_levels-1].length == 0xffffffffffffff)
915  return 0; // we reached the end of an unknown size cluster
916  if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
917  av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
918  if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
919  return AVERROR_INVALIDDATA;
920  }
921  return ebml_parse_elem(matroska, &syntax[i], data);
922 }
923 
924 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
925  void *data)
926 {
927  if (!matroska->current_id) {
928  uint64_t id;
929  int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
930  if (res < 0)
931  return res;
932  matroska->current_id = id | 1 << 7*res;
933  }
934  return ebml_parse_id(matroska, syntax, matroska->current_id, data);
935 }
936 
937 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
938  void *data)
939 {
940  int i, res = 0;
941 
942  for (i=0; syntax[i].id; i++)
943  switch (syntax[i].type) {
944  case EBML_UINT:
945  *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
946  break;
947  case EBML_FLOAT:
948  *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
949  break;
950  case EBML_STR:
951  case EBML_UTF8:
952  // the default may be NULL
953  if (syntax[i].def.s) {
954  uint8_t **dst = (uint8_t**)((uint8_t*)data + syntax[i].data_offset);
955  *dst = av_strdup(syntax[i].def.s);
956  if (!*dst)
957  return AVERROR(ENOMEM);
958  }
959  break;
960  }
961 
962  while (!res && !ebml_level_end(matroska))
963  res = ebml_parse(matroska, syntax, data);
964 
965  return res;
966 }
967 
969  EbmlSyntax *syntax, void *data)
970 {
971  static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
972  [EBML_UINT] = 8,
973  [EBML_FLOAT] = 8,
974  // max. 16 MB for strings
975  [EBML_STR] = 0x1000000,
976  [EBML_UTF8] = 0x1000000,
977  // max. 256 MB for binary data
978  [EBML_BIN] = 0x10000000,
979  // no limits for anything else
980  };
981  AVIOContext *pb = matroska->ctx->pb;
982  uint32_t id = syntax->id;
983  uint64_t length;
984  int res;
985  void *newelem;
986 
987  data = (char *)data + syntax->data_offset;
988  if (syntax->list_elem_size) {
989  EbmlList *list = data;
990  newelem = av_realloc_array(list->elem, list->nb_elem+1, syntax->list_elem_size);
991  if (!newelem)
992  return AVERROR(ENOMEM);
993  list->elem = newelem;
994  data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
995  memset(data, 0, syntax->list_elem_size);
996  list->nb_elem++;
997  }
998 
999  if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
1000  matroska->current_id = 0;
1001  if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1002  return res;
1003  if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1004  av_log(matroska->ctx, AV_LOG_ERROR,
1005  "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1006  length, max_lengths[syntax->type], syntax->type);
1007  return AVERROR_INVALIDDATA;
1008  }
1009  }
1010 
1011  switch (syntax->type) {
1012  case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
1013  case EBML_SINT: res = ebml_read_sint (pb, length, data); break;
1014  case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
1015  case EBML_STR:
1016  case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
1017  case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
1018  case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
1019  return res;
1020  if (id == MATROSKA_ID_SEGMENT)
1021  matroska->segment_start = avio_tell(matroska->ctx->pb);
1022  return ebml_parse_nest(matroska, syntax->def.n, data);
1023  case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
1024  case EBML_STOP: return 1;
1025  default:
1026  if(ffio_limit(pb, length) != length)
1027  return AVERROR(EIO);
1028  return avio_skip(pb,length)<0 ? AVERROR(EIO) : 0;
1029  }
1030  if (res == AVERROR_INVALIDDATA)
1031  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1032  else if (res == AVERROR(EIO))
1033  av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1034  return res;
1035 }
1036 
1037 static void ebml_free(EbmlSyntax *syntax, void *data)
1038 {
1039  int i, j;
1040  for (i=0; syntax[i].id; i++) {
1041  void *data_off = (char *)data + syntax[i].data_offset;
1042  switch (syntax[i].type) {
1043  case EBML_STR:
1044  case EBML_UTF8: av_freep(data_off); break;
1045  case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
1046  case EBML_NEST:
1047  if (syntax[i].list_elem_size) {
1048  EbmlList *list = data_off;
1049  char *ptr = list->elem;
1050  for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
1051  ebml_free(syntax[i].def.n, ptr);
1052  av_free(list->elem);
1053  } else
1054  ebml_free(syntax[i].def.n, data_off);
1055  default: break;
1056  }
1057  }
1058 }
1059 
1060 
1061 /*
1062  * Autodetecting...
1063  */
1065 {
1066  uint64_t total = 0;
1067  int len_mask = 0x80, size = 1, n = 1, i;
1068 
1069  /* EBML header? */
1070  if (AV_RB32(p->buf) != EBML_ID_HEADER)
1071  return 0;
1072 
1073  /* length of header */
1074  total = p->buf[4];
1075  while (size <= 8 && !(total & len_mask)) {
1076  size++;
1077  len_mask >>= 1;
1078  }
1079  if (size > 8)
1080  return 0;
1081  total &= (len_mask - 1);
1082  while (n < size)
1083  total = (total << 8) | p->buf[4 + n++];
1084 
1085  /* Does the probe data contain the whole header? */
1086  if (p->buf_size < 4 + size + total)
1087  return 0;
1088 
1089  /* The header should contain a known document type. For now,
1090  * we don't parse the whole header but simply check for the
1091  * availability of that array of characters inside the header.
1092  * Not fully fool-proof, but good enough. */
1093  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1094  int probelen = strlen(matroska_doctypes[i]);
1095  if (total < probelen)
1096  continue;
1097  for (n = 4+size; n <= 4+size+total-probelen; n++)
1098  if (!memcmp(p->buf+n, matroska_doctypes[i], probelen))
1099  return AVPROBE_SCORE_MAX;
1100  }
1101 
1102  // probably valid EBML header but no recognized doctype
1103  return AVPROBE_SCORE_EXTENSION;
1104 }
1105 
1107  int num)
1108 {
1109  MatroskaTrack *tracks = matroska->tracks.elem;
1110  int i;
1111 
1112  for (i=0; i < matroska->tracks.nb_elem; i++)
1113  if (tracks[i].num == num)
1114  return &tracks[i];
1115 
1116  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1117  return NULL;
1118 }
1119 
1120 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
1121  MatroskaTrack *track)
1122 {
1123  MatroskaTrackEncoding *encodings = track->encodings.elem;
1124  uint8_t* data = *buf;
1125  int isize = *buf_size;
1126  uint8_t* pkt_data = NULL;
1127  uint8_t av_unused *newpktdata;
1128  int pkt_size = isize;
1129  int result = 0;
1130  int olen;
1131 
1132  if (pkt_size >= 10000000U)
1133  return AVERROR_INVALIDDATA;
1134 
1135  switch (encodings[0].compression.algo) {
1137  int header_size = encodings[0].compression.settings.size;
1138  uint8_t *header = encodings[0].compression.settings.data;
1139 
1140  if (header_size && !header) {
1141  av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1142  return -1;
1143  }
1144 
1145  if (!header_size)
1146  return 0;
1147 
1148  pkt_size = isize + header_size;
1149  pkt_data = av_malloc(pkt_size);
1150  if (!pkt_data)
1151  return AVERROR(ENOMEM);
1152 
1153  memcpy(pkt_data, header, header_size);
1154  memcpy(pkt_data + header_size, data, isize);
1155  break;
1156  }
1157 #if CONFIG_LZO
1159  do {
1160  olen = pkt_size *= 3;
1161  newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1162  if (!newpktdata) {
1163  result = AVERROR(ENOMEM);
1164  goto failed;
1165  }
1166  pkt_data = newpktdata;
1167  result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1168  } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
1169  if (result) {
1170  result = AVERROR_INVALIDDATA;
1171  goto failed;
1172  }
1173  pkt_size -= olen;
1174  break;
1175 #endif
1176 #if CONFIG_ZLIB
1178  z_stream zstream = {0};
1179  if (inflateInit(&zstream) != Z_OK)
1180  return -1;
1181  zstream.next_in = data;
1182  zstream.avail_in = isize;
1183  do {
1184  pkt_size *= 3;
1185  newpktdata = av_realloc(pkt_data, pkt_size);
1186  if (!newpktdata) {
1187  inflateEnd(&zstream);
1188  goto failed;
1189  }
1190  pkt_data = newpktdata;
1191  zstream.avail_out = pkt_size - zstream.total_out;
1192  zstream.next_out = pkt_data + zstream.total_out;
1193  if (pkt_data) {
1194  result = inflate(&zstream, Z_NO_FLUSH);
1195  } else
1196  result = Z_MEM_ERROR;
1197  } while (result==Z_OK && pkt_size<10000000);
1198  pkt_size = zstream.total_out;
1199  inflateEnd(&zstream);
1200  if (result != Z_STREAM_END) {
1201  if (result == Z_MEM_ERROR)
1202  result = AVERROR(ENOMEM);
1203  else
1204  result = AVERROR_INVALIDDATA;
1205  goto failed;
1206  }
1207  break;
1208  }
1209 #endif
1210 #if CONFIG_BZLIB
1212  bz_stream bzstream = {0};
1213  if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1214  return -1;
1215  bzstream.next_in = data;
1216  bzstream.avail_in = isize;
1217  do {
1218  pkt_size *= 3;
1219  newpktdata = av_realloc(pkt_data, pkt_size);
1220  if (!newpktdata) {
1221  BZ2_bzDecompressEnd(&bzstream);
1222  goto failed;
1223  }
1224  pkt_data = newpktdata;
1225  bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1226  bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1227  if (pkt_data) {
1228  result = BZ2_bzDecompress(&bzstream);
1229  } else
1230  result = BZ_MEM_ERROR;
1231  } while (result==BZ_OK && pkt_size<10000000);
1232  pkt_size = bzstream.total_out_lo32;
1233  BZ2_bzDecompressEnd(&bzstream);
1234  if (result != BZ_STREAM_END) {
1235  if (result == BZ_MEM_ERROR)
1236  result = AVERROR(ENOMEM);
1237  else
1238  result = AVERROR_INVALIDDATA;
1239  goto failed;
1240  }
1241  break;
1242  }
1243 #endif
1244  default:
1245  return AVERROR_INVALIDDATA;
1246  }
1247 
1248  *buf = pkt_data;
1249  *buf_size = pkt_size;
1250  return 0;
1251  failed:
1252  av_free(pkt_data);
1253  return result;
1254 }
1255 
1256 #if FF_API_ASS_SSA
1257 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
1258  AVPacket *pkt, uint64_t display_duration)
1259 {
1260  AVBufferRef *line;
1261  char *layer, *ptr = pkt->data, *end = ptr+pkt->size;
1262  for (; *ptr!=',' && ptr<end-1; ptr++);
1263  if (*ptr == ',')
1264  ptr++;
1265  layer = ptr;
1266  for (; *ptr!=',' && ptr<end-1; ptr++);
1267  if (*ptr == ',') {
1268  int64_t end_pts = pkt->pts + display_duration;
1269  int sc = matroska->time_scale * pkt->pts / 10000000;
1270  int ec = matroska->time_scale * end_pts / 10000000;
1271  int sh, sm, ss, eh, em, es, len;
1272  sh = sc/360000; sc -= 360000*sh;
1273  sm = sc/ 6000; sc -= 6000*sm;
1274  ss = sc/ 100; sc -= 100*ss;
1275  eh = ec/360000; ec -= 360000*eh;
1276  em = ec/ 6000; ec -= 6000*em;
1277  es = ec/ 100; ec -= 100*es;
1278  *ptr++ = '\0';
1279  len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
1280  if (!(line = av_buffer_alloc(len)))
1281  return;
1282  snprintf(line->data, len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
1283  layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
1284  av_buffer_unref(&pkt->buf);
1285  pkt->buf = line;
1286  pkt->data = line->data;
1287  pkt->size = strlen(line->data);
1288  }
1289 }
1290 
1291 static int matroska_merge_packets(AVPacket *out, AVPacket *in)
1292 {
1293  int ret = av_grow_packet(out, in->size);
1294  if (ret < 0)
1295  return ret;
1296 
1297  memcpy(out->data + out->size - in->size, in->data, in->size);
1298 
1299  av_free_packet(in);
1300  av_free(in);
1301  return 0;
1302 }
1303 #endif
1304 
1306  AVDictionary **metadata, char *prefix)
1307 {
1308  MatroskaTag *tags = list->elem;
1309  char key[1024];
1310  int i;
1311 
1312  for (i=0; i < list->nb_elem; i++) {
1313  const char *lang = tags[i].lang && strcmp(tags[i].lang, "und") ?
1314  tags[i].lang : NULL;
1315 
1316  if (!tags[i].name) {
1317  av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1318  continue;
1319  }
1320  if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1321  else av_strlcpy(key, tags[i].name, sizeof(key));
1322  if (tags[i].def || !lang) {
1323  av_dict_set(metadata, key, tags[i].string, 0);
1324  if (tags[i].sub.nb_elem)
1325  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1326  }
1327  if (lang) {
1328  av_strlcat(key, "-", sizeof(key));
1329  av_strlcat(key, lang, sizeof(key));
1330  av_dict_set(metadata, key, tags[i].string, 0);
1331  if (tags[i].sub.nb_elem)
1332  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1333  }
1334  }
1335  ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1336 }
1337 
1339 {
1340  MatroskaDemuxContext *matroska = s->priv_data;
1341  MatroskaTags *tags = matroska->tags.elem;
1342  int i, j;
1343 
1344  for (i=0; i < matroska->tags.nb_elem; i++) {
1345  if (tags[i].target.attachuid) {
1346  MatroskaAttachement *attachment = matroska->attachments.elem;
1347  for (j=0; j<matroska->attachments.nb_elem; j++)
1348  if (attachment[j].uid == tags[i].target.attachuid
1349  && attachment[j].stream)
1350  matroska_convert_tag(s, &tags[i].tag,
1351  &attachment[j].stream->metadata, NULL);
1352  } else if (tags[i].target.chapteruid) {
1353  MatroskaChapter *chapter = matroska->chapters.elem;
1354  for (j=0; j<matroska->chapters.nb_elem; j++)
1355  if (chapter[j].uid == tags[i].target.chapteruid
1356  && chapter[j].chapter)
1357  matroska_convert_tag(s, &tags[i].tag,
1358  &chapter[j].chapter->metadata, NULL);
1359  } else if (tags[i].target.trackuid) {
1360  MatroskaTrack *track = matroska->tracks.elem;
1361  for (j=0; j<matroska->tracks.nb_elem; j++)
1362  if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1363  matroska_convert_tag(s, &tags[i].tag,
1364  &track[j].stream->metadata, NULL);
1365  } else {
1366  matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1367  tags[i].target.type);
1368  }
1369  }
1370 }
1371 
1373 {
1374  EbmlList *seekhead_list = &matroska->seekhead;
1375  MatroskaSeekhead *seekhead = seekhead_list->elem;
1376  uint32_t level_up = matroska->level_up;
1377  int64_t before_pos = avio_tell(matroska->ctx->pb);
1378  uint32_t saved_id = matroska->current_id;
1380  int64_t offset;
1381  int ret = 0;
1382 
1383  if (idx >= seekhead_list->nb_elem
1384  || seekhead[idx].id == MATROSKA_ID_SEEKHEAD
1385  || seekhead[idx].id == MATROSKA_ID_CLUSTER)
1386  return 0;
1387 
1388  /* seek */
1389  offset = seekhead[idx].pos + matroska->segment_start;
1390  if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1391  /* We don't want to lose our seekhead level, so we add
1392  * a dummy. This is a crude hack. */
1393  if (matroska->num_levels == EBML_MAX_DEPTH) {
1394  av_log(matroska->ctx, AV_LOG_INFO,
1395  "Max EBML element depth (%d) reached, "
1396  "cannot parse further.\n", EBML_MAX_DEPTH);
1397  ret = AVERROR_INVALIDDATA;
1398  } else {
1399  level.start = 0;
1400  level.length = (uint64_t)-1;
1401  matroska->levels[matroska->num_levels] = level;
1402  matroska->num_levels++;
1403  matroska->current_id = 0;
1404 
1405  ret = ebml_parse(matroska, matroska_segment, matroska);
1406 
1407  /* remove dummy level */
1408  while (matroska->num_levels) {
1409  uint64_t length = matroska->levels[--matroska->num_levels].length;
1410  if (length == (uint64_t)-1)
1411  break;
1412  }
1413  }
1414  }
1415  /* seek back */
1416  avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1417  matroska->level_up = level_up;
1418  matroska->current_id = saved_id;
1419 
1420  return ret;
1421 }
1422 
1424 {
1425  EbmlList *seekhead_list = &matroska->seekhead;
1426  int64_t before_pos = avio_tell(matroska->ctx->pb);
1427  int i;
1428 
1429  // we should not do any seeking in the streaming case
1430  if (!matroska->ctx->pb->seekable ||
1431  (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1432  return;
1433 
1434  for (i = 0; i < seekhead_list->nb_elem; i++) {
1435  MatroskaSeekhead *seekhead = seekhead_list->elem;
1436  if (seekhead[i].pos <= before_pos)
1437  continue;
1438 
1439  // defer cues parsing until we actually need cue data.
1440  if (seekhead[i].id == MATROSKA_ID_CUES) {
1441  matroska->cues_parsing_deferred = 1;
1442  continue;
1443  }
1444 
1445  if (matroska_parse_seekhead_entry(matroska, i) < 0) {
1446  // mark index as broken
1447  matroska->cues_parsing_deferred = -1;
1448  break;
1449  }
1450  }
1451 }
1452 
1454  EbmlList *index_list;
1456  int index_scale = 1;
1457  int i, j;
1458 
1459  index_list = &matroska->index;
1460  index = index_list->elem;
1461  if (index_list->nb_elem
1462  && index[0].time > 1E14/matroska->time_scale) {
1463  av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1464  index_scale = matroska->time_scale;
1465  }
1466  for (i = 0; i < index_list->nb_elem; i++) {
1467  EbmlList *pos_list = &index[i].pos;
1468  MatroskaIndexPos *pos = pos_list->elem;
1469  for (j = 0; j < pos_list->nb_elem; j++) {
1470  MatroskaTrack *track = matroska_find_track_by_num(matroska, pos[j].track);
1471  if (track && track->stream)
1472  av_add_index_entry(track->stream,
1473  pos[j].pos + matroska->segment_start,
1474  index[i].time/index_scale, 0, 0,
1476  }
1477  }
1478 }
1479 
1481  EbmlList *seekhead_list = &matroska->seekhead;
1482  MatroskaSeekhead *seekhead = seekhead_list->elem;
1483  int i;
1484 
1485  for (i = 0; i < seekhead_list->nb_elem; i++)
1486  if (seekhead[i].id == MATROSKA_ID_CUES)
1487  break;
1488  av_assert1(i <= seekhead_list->nb_elem);
1489 
1490  if (matroska_parse_seekhead_entry(matroska, i) < 0)
1491  matroska->cues_parsing_deferred = -1;
1492  matroska_add_index_entries(matroska);
1493 }
1494 
1496 {
1497  static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1498  int profile;
1499 
1500  for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1501  if (strstr(codec_id, aac_profiles[profile]))
1502  break;
1503  return profile + 1;
1504 }
1505 
1506 static int matroska_aac_sri(int samplerate)
1507 {
1508  int sri;
1509 
1510  for (sri=0; sri<FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1511  if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1512  break;
1513  return sri;
1514 }
1515 
1516 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1517 {
1518  char buffer[32];
1519  /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1520  time_t creation_time = date_utc / 1000000000 + 978307200;
1521  struct tm *ptm = gmtime(&creation_time);
1522  if (!ptm) return;
1523  strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm);
1524  av_dict_set(metadata, "creation_time", buffer, 0);
1525 }
1526 
1528 {
1529  MatroskaDemuxContext *matroska = s->priv_data;
1530  EbmlList *attachements_list = &matroska->attachments;
1531  MatroskaAttachement *attachements;
1532  EbmlList *chapters_list = &matroska->chapters;
1533  MatroskaChapter *chapters;
1534  MatroskaTrack *tracks;
1535  uint64_t max_start = 0;
1536  int64_t pos;
1537  Ebml ebml = { 0 };
1538  AVStream *st;
1539  int i, j, k, res;
1540 
1541  matroska->ctx = s;
1542 
1543  /* First read the EBML header. */
1544  if (ebml_parse(matroska, ebml_syntax, &ebml)
1545  || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1546  || ebml.id_length > sizeof(uint32_t) || ebml.doctype_version > 3 || !ebml.doctype) {
1547  av_log(matroska->ctx, AV_LOG_ERROR,
1548  "EBML header using unsupported features\n"
1549  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1550  ebml.version, ebml.doctype, ebml.doctype_version);
1551  ebml_free(ebml_syntax, &ebml);
1552  return AVERROR_PATCHWELCOME;
1553  } else if (ebml.doctype_version == 3) {
1554  av_log(matroska->ctx, AV_LOG_WARNING,
1555  "EBML header using unsupported features\n"
1556  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1557  ebml.version, ebml.doctype, ebml.doctype_version);
1558  }
1559  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
1560  if (!strcmp(ebml.doctype, matroska_doctypes[i]))
1561  break;
1562  if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
1563  av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
1564  if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
1565  ebml_free(ebml_syntax, &ebml);
1566  return AVERROR_INVALIDDATA;
1567  }
1568  }
1569  ebml_free(ebml_syntax, &ebml);
1570 
1571  /* The next thing is a segment. */
1572  pos = avio_tell(matroska->ctx->pb);
1573  res = ebml_parse(matroska, matroska_segments, matroska);
1574  // try resyncing until we find a EBML_STOP type element.
1575  while (res != 1) {
1576  res = matroska_resync(matroska, pos);
1577  if (res < 0)
1578  return res;
1579  pos = avio_tell(matroska->ctx->pb);
1580  res = ebml_parse(matroska, matroska_segment, matroska);
1581  }
1582  matroska_execute_seekhead(matroska);
1583 
1584  if (!matroska->time_scale)
1585  matroska->time_scale = 1000000;
1586  if (matroska->duration)
1587  matroska->ctx->duration = matroska->duration * matroska->time_scale
1588  * 1000 / AV_TIME_BASE;
1589  av_dict_set(&s->metadata, "title", matroska->title, 0);
1590 
1591  if (matroska->date_utc.size == 8)
1593 
1594  tracks = matroska->tracks.elem;
1595  for (i=0; i < matroska->tracks.nb_elem; i++) {
1596  MatroskaTrack *track = &tracks[i];
1598  EbmlList *encodings_list = &track->encodings;
1599  MatroskaTrackEncoding *encodings = encodings_list->elem;
1600  uint8_t *extradata = NULL;
1601  int extradata_size = 0;
1602  int extradata_offset = 0;
1603  uint32_t fourcc = 0;
1604  AVIOContext b;
1605  char* key_id_base64 = NULL;
1606 
1607  /* Apply some sanity checks. */
1608  if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1609  track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1610  track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
1611  track->type != MATROSKA_TRACK_TYPE_METADATA) {
1612  av_log(matroska->ctx, AV_LOG_INFO,
1613  "Unknown or unsupported track type %"PRIu64"\n",
1614  track->type);
1615  continue;
1616  }
1617  if (track->codec_id == NULL)
1618  continue;
1619 
1620  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1621  if (!track->default_duration && track->video.frame_rate > 0)
1622  track->default_duration = 1000000000/track->video.frame_rate;
1623  if (track->video.display_width == -1)
1624  track->video.display_width = track->video.pixel_width;
1625  if (track->video.display_height == -1)
1626  track->video.display_height = track->video.pixel_height;
1627  if (track->video.color_space.size == 4)
1628  fourcc = AV_RL32(track->video.color_space.data);
1629  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1630  if (!track->audio.out_samplerate)
1631  track->audio.out_samplerate = track->audio.samplerate;
1632  }
1633  if (encodings_list->nb_elem > 1) {
1634  av_log(matroska->ctx, AV_LOG_ERROR,
1635  "Multiple combined encodings not supported");
1636  } else if (encodings_list->nb_elem == 1) {
1637  if (encodings[0].type) {
1638  if (encodings[0].encryption.key_id.size > 0) {
1639  /* Save the encryption key id to be stored later as a
1640  metadata tag. */
1641  const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
1642  key_id_base64 = av_malloc(b64_size);
1643  if (key_id_base64 == NULL)
1644  return AVERROR(ENOMEM);
1645 
1646  av_base64_encode(key_id_base64, b64_size,
1647  encodings[0].encryption.key_id.data,
1648  encodings[0].encryption.key_id.size);
1649  } else {
1650  encodings[0].scope = 0;
1651  av_log(matroska->ctx, AV_LOG_ERROR,
1652  "Unsupported encoding type");
1653  }
1654  } else if (
1655 #if CONFIG_ZLIB
1656  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1657 #endif
1658 #if CONFIG_BZLIB
1660 #endif
1661 #if CONFIG_LZO
1663 #endif
1665  encodings[0].scope = 0;
1666  av_log(matroska->ctx, AV_LOG_ERROR,
1667  "Unsupported encoding type");
1668  } else if (track->codec_priv.size && encodings[0].scope&2) {
1669  uint8_t *codec_priv = track->codec_priv.data;
1670  int ret = matroska_decode_buffer(&track->codec_priv.data,
1671  &track->codec_priv.size,
1672  track);
1673  if (ret < 0) {
1674  track->codec_priv.data = NULL;
1675  track->codec_priv.size = 0;
1676  av_log(matroska->ctx, AV_LOG_ERROR,
1677  "Failed to decode codec private data\n");
1678  }
1679 
1680  if (codec_priv != track->codec_priv.data)
1681  av_free(codec_priv);
1682  }
1683  }
1684 
1685  for(j=0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++){
1686  if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1687  strlen(ff_mkv_codec_tags[j].str))){
1688  codec_id= ff_mkv_codec_tags[j].id;
1689  break;
1690  }
1691  }
1692 
1693  st = track->stream = avformat_new_stream(s, NULL);
1694  if (st == NULL) {
1695  av_free(key_id_base64);
1696  return AVERROR(ENOMEM);
1697  }
1698 
1699  if (key_id_base64) {
1700  /* export encryption key id as base64 metadata tag */
1701  av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
1702  av_freep(&key_id_base64);
1703  }
1704 
1705  if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1706  && track->codec_priv.size >= 40
1707  && track->codec_priv.data != NULL) {
1708  track->ms_compat = 1;
1709  fourcc = AV_RL32(track->codec_priv.data + 16);
1710  codec_id = ff_codec_get_id(ff_codec_bmp_tags, fourcc);
1711  extradata_offset = 40;
1712  } else if (!strcmp(track->codec_id, "A_MS/ACM")
1713  && track->codec_priv.size >= 14
1714  && track->codec_priv.data != NULL) {
1715  int ret;
1716  ffio_init_context(&b, track->codec_priv.data, track->codec_priv.size,
1717  0, NULL, NULL, NULL, NULL);
1718  ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1719  if (ret < 0)
1720  return ret;
1721  codec_id = st->codec->codec_id;
1722  extradata_offset = FFMIN(track->codec_priv.size, 18);
1723  } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1724  && (track->codec_priv.size >= 86)
1725  && (track->codec_priv.data != NULL)) {
1726  fourcc = AV_RL32(track->codec_priv.data);
1727  codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1728  } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1729  switch (track->audio.bitdepth) {
1730  case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1731  case 24: codec_id = AV_CODEC_ID_PCM_S24BE; break;
1732  case 32: codec_id = AV_CODEC_ID_PCM_S32BE; break;
1733  }
1734  } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1735  switch (track->audio.bitdepth) {
1736  case 8: codec_id = AV_CODEC_ID_PCM_U8; break;
1737  case 24: codec_id = AV_CODEC_ID_PCM_S24LE; break;
1738  case 32: codec_id = AV_CODEC_ID_PCM_S32LE; break;
1739  }
1740  } else if (codec_id==AV_CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1741  codec_id = AV_CODEC_ID_PCM_F64LE;
1742  } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1743  int profile = matroska_aac_profile(track->codec_id);
1744  int sri = matroska_aac_sri(track->audio.samplerate);
1745  extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1746  if (extradata == NULL)
1747  return AVERROR(ENOMEM);
1748  extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1749  extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1750  if (strstr(track->codec_id, "SBR")) {
1751  sri = matroska_aac_sri(track->audio.out_samplerate);
1752  extradata[2] = 0x56;
1753  extradata[3] = 0xE5;
1754  extradata[4] = 0x80 | (sri<<3);
1755  extradata_size = 5;
1756  } else
1757  extradata_size = 2;
1758  } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - FF_INPUT_BUFFER_PADDING_SIZE) {
1759  /* Only ALAC's magic cookie is stored in Matroska's track headers.
1760  Create the "atom size", "tag", and "tag version" fields the
1761  decoder expects manually. */
1762  extradata_size = 12 + track->codec_priv.size;
1763  extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1764  if (extradata == NULL)
1765  return AVERROR(ENOMEM);
1766  AV_WB32(extradata, extradata_size);
1767  memcpy(&extradata[4], "alac", 4);
1768  AV_WB32(&extradata[8], 0);
1769  memcpy(&extradata[12], track->codec_priv.data,
1770  track->codec_priv.size);
1771  } else if (codec_id == AV_CODEC_ID_TTA) {
1772  extradata_size = 30;
1773  extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1774  if (extradata == NULL)
1775  return AVERROR(ENOMEM);
1776  ffio_init_context(&b, extradata, extradata_size, 1,
1777  NULL, NULL, NULL, NULL);
1778  avio_write(&b, "TTA1", 4);
1779  avio_wl16(&b, 1);
1780  avio_wl16(&b, track->audio.channels);
1781  avio_wl16(&b, track->audio.bitdepth);
1782  if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
1783  return AVERROR_INVALIDDATA;
1784  avio_wl32(&b, track->audio.out_samplerate);
1785  avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale), track->audio.out_samplerate, AV_TIME_BASE * 1000));
1786  } else if (codec_id == AV_CODEC_ID_RV10 || codec_id == AV_CODEC_ID_RV20 ||
1787  codec_id == AV_CODEC_ID_RV30 || codec_id == AV_CODEC_ID_RV40) {
1788  extradata_offset = 26;
1789  } else if (codec_id == AV_CODEC_ID_RA_144) {
1790  track->audio.out_samplerate = 8000;
1791  track->audio.channels = 1;
1792  } else if ((codec_id == AV_CODEC_ID_RA_288 || codec_id == AV_CODEC_ID_COOK ||
1793  codec_id == AV_CODEC_ID_ATRAC3 || codec_id == AV_CODEC_ID_SIPR)
1794  && track->codec_priv.data) {
1795  int flavor;
1796 
1797  ffio_init_context(&b, track->codec_priv.data,track->codec_priv.size,
1798  0, NULL, NULL, NULL, NULL);
1799  avio_skip(&b, 22);
1800  flavor = avio_rb16(&b);
1801  track->audio.coded_framesize = avio_rb32(&b);
1802  avio_skip(&b, 12);
1803  track->audio.sub_packet_h = avio_rb16(&b);
1804  track->audio.frame_size = avio_rb16(&b);
1805  track->audio.sub_packet_size = avio_rb16(&b);
1806  if (flavor < 0 || track->audio.coded_framesize <= 0 ||
1807  track->audio.sub_packet_h <= 0 || track->audio.frame_size <= 0 ||
1808  track->audio.sub_packet_size <= 0)
1809  return AVERROR_INVALIDDATA;
1810  track->audio.buf = av_malloc_array(track->audio.sub_packet_h, track->audio.frame_size);
1811  if (!track->audio.buf)
1812  return AVERROR(ENOMEM);
1813  if (codec_id == AV_CODEC_ID_RA_288) {
1814  st->codec->block_align = track->audio.coded_framesize;
1815  track->codec_priv.size = 0;
1816  } else {
1817  if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1818  static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1819  track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1820  st->codec->bit_rate = sipr_bit_rate[flavor];
1821  }
1822  st->codec->block_align = track->audio.sub_packet_size;
1823  extradata_offset = 78;
1824  }
1825  }
1826  track->codec_priv.size -= extradata_offset;
1827 
1828  if (codec_id == AV_CODEC_ID_NONE)
1829  av_log(matroska->ctx, AV_LOG_INFO,
1830  "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1831 
1832  if (track->time_scale < 0.01)
1833  track->time_scale = 1.0;
1834  avpriv_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1835 
1836  st->codec->codec_id = codec_id;
1837  st->start_time = 0;
1838  if (strcmp(track->language, "und"))
1839  av_dict_set(&st->metadata, "language", track->language, 0);
1840  av_dict_set(&st->metadata, "title", track->name, 0);
1841 
1842  if (track->flag_default)
1844  if (track->flag_forced)
1846 
1847  if (!st->codec->extradata) {
1848  if(extradata){
1849  st->codec->extradata = extradata;
1850  st->codec->extradata_size = extradata_size;
1851  } else if(track->codec_priv.data && track->codec_priv.size > 0){
1852  if (ff_alloc_extradata(st->codec, track->codec_priv.size))
1853  return AVERROR(ENOMEM);
1854  memcpy(st->codec->extradata,
1855  track->codec_priv.data + extradata_offset,
1856  track->codec_priv.size);
1857  }
1858  }
1859 
1860  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1862 
1864  st->codec->codec_tag = fourcc;
1865  st->codec->width = track->video.pixel_width;
1866  st->codec->height = track->video.pixel_height;
1868  &st->sample_aspect_ratio.den,
1869  st->codec->height * track->video.display_width,
1870  st->codec-> width * track->video.display_height,
1871  255);
1872  if (st->codec->codec_id != AV_CODEC_ID_HEVC)
1874  if (track->default_duration) {
1876  1000000000, track->default_duration, 30000);
1877 #if FF_API_R_FRAME_RATE
1878  if (st->avg_frame_rate.num < st->avg_frame_rate.den * 1000L)
1879  st->r_frame_rate = st->avg_frame_rate;
1880 #endif
1881  }
1882 
1883  /* export stereo mode flag as metadata tag */
1885  av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
1886 
1887  /* export alpha mode flag as metadata tag */
1888  if (track->video.alpha_mode)
1889  av_dict_set(&st->metadata, "alpha_mode", "1", 0);
1890 
1891  /* if we have virtual track, mark the real tracks */
1892  for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
1893  char buf[32];
1894  if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
1895  continue;
1896  snprintf(buf, sizeof(buf), "%s_%d",
1897  ff_matroska_video_stereo_plane[planes[j].type], i);
1898  for (k=0; k < matroska->tracks.nb_elem; k++)
1899  if (planes[j].uid == tracks[k].uid) {
1900  av_dict_set(&s->streams[k]->metadata,
1901  "stereo_mode", buf, 0);
1902  break;
1903  }
1904  }
1905  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1907  st->codec->sample_rate = track->audio.out_samplerate;
1908  st->codec->channels = track->audio.channels;
1909  if (!st->codec->bits_per_coded_sample)
1910  st->codec->bits_per_coded_sample = track->audio.bitdepth;
1911  if (st->codec->codec_id != AV_CODEC_ID_AAC)
1913  if (track->codec_delay > 0) {
1914  st->codec->delay = av_rescale_q(track->codec_delay,
1915  (AVRational){1, 1000000000},
1916  (AVRational){1, st->codec->sample_rate});
1917  }
1918  if (track->seek_preroll > 0) {
1920  av_rescale_q(track->seek_preroll,
1921  (AVRational){1, 1000000000},
1922  (AVRational){1, st->codec->sample_rate}));
1923  }
1924  } else if (codec_id == AV_CODEC_ID_WEBVTT) {
1925  st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1926 
1927  if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
1928  st->disposition |= AV_DISPOSITION_CAPTIONS;
1929  } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
1930  st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
1931  } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
1932  st->disposition |= AV_DISPOSITION_METADATA;
1933  }
1934  } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1935  st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1936 #if FF_API_ASS_SSA
1937  if (st->codec->codec_id == AV_CODEC_ID_SSA ||
1938  st->codec->codec_id == AV_CODEC_ID_ASS)
1939 #else
1940  if (st->codec->codec_id == AV_CODEC_ID_ASS)
1941 #endif
1942  matroska->contains_ssa = 1;
1943  }
1944  }
1945 
1946  attachements = attachements_list->elem;
1947  for (j=0; j<attachements_list->nb_elem; j++) {
1948  if (!(attachements[j].filename && attachements[j].mime &&
1949  attachements[j].bin.data && attachements[j].bin.size > 0)) {
1950  av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1951  } else {
1952  AVStream *st = avformat_new_stream(s, NULL);
1953  if (st == NULL)
1954  break;
1955  av_dict_set(&st->metadata, "filename",attachements[j].filename, 0);
1956  av_dict_set(&st->metadata, "mimetype", attachements[j].mime, 0);
1959  if (ff_alloc_extradata(st->codec, attachements[j].bin.size))
1960  break;
1961  memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1962 
1963  for (i=0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
1964  if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1965  strlen(ff_mkv_mime_tags[i].str))) {
1966  st->codec->codec_id = ff_mkv_mime_tags[i].id;
1967  break;
1968  }
1969  }
1970  attachements[j].stream = st;
1971  }
1972  }
1973 
1974  chapters = chapters_list->elem;
1975  for (i=0; i<chapters_list->nb_elem; i++)
1976  if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1977  && (max_start==0 || chapters[i].start > max_start)) {
1978  chapters[i].chapter =
1979  avpriv_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1980  chapters[i].start, chapters[i].end,
1981  chapters[i].title);
1982  av_dict_set(&chapters[i].chapter->metadata,
1983  "title", chapters[i].title, 0);
1984  max_start = chapters[i].start;
1985  }
1986 
1987  matroska_add_index_entries(matroska);
1988 
1990 
1991  return 0;
1992 }
1993 
1994 /*
1995  * Put one packet in an application-supplied AVPacket struct.
1996  * Returns 0 on success or -1 on failure.
1997  */
1999  AVPacket *pkt)
2000 {
2001  if (matroska->num_packets > 0) {
2002  memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2003  av_free(matroska->packets[0]);
2004  if (matroska->num_packets > 1) {
2005  void *newpackets;
2006  memmove(&matroska->packets[0], &matroska->packets[1],
2007  (matroska->num_packets - 1) * sizeof(AVPacket *));
2008  newpackets = av_realloc(matroska->packets,
2009  (matroska->num_packets - 1) * sizeof(AVPacket *));
2010  if (newpackets)
2011  matroska->packets = newpackets;
2012  } else {
2013  av_freep(&matroska->packets);
2014  matroska->prev_pkt = NULL;
2015  }
2016  matroska->num_packets--;
2017  return 0;
2018  }
2019 
2020  return -1;
2021 }
2022 
2023 /*
2024  * Free all packets in our internal queue.
2025  */
2027 {
2028  matroska->prev_pkt = NULL;
2029  if (matroska->packets) {
2030  int n;
2031  for (n = 0; n < matroska->num_packets; n++) {
2032  av_free_packet(matroska->packets[n]);
2033  av_free(matroska->packets[n]);
2034  }
2035  av_freep(&matroska->packets);
2036  matroska->num_packets = 0;
2037  }
2038 }
2039 
2041  int* buf_size, int type,
2042  uint32_t **lace_buf, int *laces)
2043 {
2044  int res = 0, n, size = *buf_size;
2045  uint8_t *data = *buf;
2046  uint32_t *lace_size;
2047 
2048  if (!type) {
2049  *laces = 1;
2050  *lace_buf = av_mallocz(sizeof(int));
2051  if (!*lace_buf)
2052  return AVERROR(ENOMEM);
2053 
2054  *lace_buf[0] = size;
2055  return 0;
2056  }
2057 
2058  av_assert0(size > 0);
2059  *laces = *data + 1;
2060  data += 1;
2061  size -= 1;
2062  lace_size = av_mallocz(*laces * sizeof(int));
2063  if (!lace_size)
2064  return AVERROR(ENOMEM);
2065 
2066  switch (type) {
2067  case 0x1: /* Xiph lacing */ {
2068  uint8_t temp;
2069  uint32_t total = 0;
2070  for (n = 0; res == 0 && n < *laces - 1; n++) {
2071  while (1) {
2072  if (size <= total) {
2073  res = AVERROR_INVALIDDATA;
2074  break;
2075  }
2076  temp = *data;
2077  total += temp;
2078  lace_size[n] += temp;
2079  data += 1;
2080  size -= 1;
2081  if (temp != 0xff)
2082  break;
2083  }
2084  }
2085  if (size <= total) {
2086  res = AVERROR_INVALIDDATA;
2087  break;
2088  }
2089 
2090  lace_size[n] = size - total;
2091  break;
2092  }
2093 
2094  case 0x2: /* fixed-size lacing */
2095  if (size % (*laces)) {
2096  res = AVERROR_INVALIDDATA;
2097  break;
2098  }
2099  for (n = 0; n < *laces; n++)
2100  lace_size[n] = size / *laces;
2101  break;
2102 
2103  case 0x3: /* EBML lacing */ {
2104  uint64_t num;
2105  uint64_t total;
2106  n = matroska_ebmlnum_uint(matroska, data, size, &num);
2107  if (n < 0 || num > INT_MAX) {
2108  av_log(matroska->ctx, AV_LOG_INFO,
2109  "EBML block data error\n");
2110  res = n<0 ? n : AVERROR_INVALIDDATA;
2111  break;
2112  }
2113  data += n;
2114  size -= n;
2115  total = lace_size[0] = num;
2116  for (n = 1; res == 0 && n < *laces - 1; n++) {
2117  int64_t snum;
2118  int r;
2119  r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2120  if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
2121  av_log(matroska->ctx, AV_LOG_INFO,
2122  "EBML block data error\n");
2123  res = r<0 ? r : AVERROR_INVALIDDATA;
2124  break;
2125  }
2126  data += r;
2127  size -= r;
2128  lace_size[n] = lace_size[n - 1] + snum;
2129  total += lace_size[n];
2130  }
2131  if (size <= total) {
2132  res = AVERROR_INVALIDDATA;
2133  break;
2134  }
2135  lace_size[*laces - 1] = size - total;
2136  break;
2137  }
2138  }
2139 
2140  *buf = data;
2141  *lace_buf = lace_size;
2142  *buf_size = size;
2143 
2144  return res;
2145 }
2146 
2148  MatroskaTrack *track,
2149  AVStream *st,
2150  uint8_t *data, int size,
2151  uint64_t timecode,
2152  int64_t pos)
2153 {
2154  int a = st->codec->block_align;
2155  int sps = track->audio.sub_packet_size;
2156  int cfs = track->audio.coded_framesize;
2157  int h = track->audio.sub_packet_h;
2158  int y = track->audio.sub_packet_cnt;
2159  int w = track->audio.frame_size;
2160  int x;
2161 
2162  if (!track->audio.pkt_cnt) {
2163  if (track->audio.sub_packet_cnt == 0)
2164  track->audio.buf_timecode = timecode;
2165  if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2166  if (size < cfs * h / 2) {
2167  av_log(matroska->ctx, AV_LOG_ERROR,
2168  "Corrupt int4 RM-style audio packet size\n");
2169  return AVERROR_INVALIDDATA;
2170  }
2171  for (x=0; x<h/2; x++)
2172  memcpy(track->audio.buf+x*2*w+y*cfs,
2173  data+x*cfs, cfs);
2174  } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2175  if (size < w) {
2176  av_log(matroska->ctx, AV_LOG_ERROR,
2177  "Corrupt sipr RM-style audio packet size\n");
2178  return AVERROR_INVALIDDATA;
2179  }
2180  memcpy(track->audio.buf + y*w, data, w);
2181  } else {
2182  if (size < sps * w / sps || h<=0 || w%sps) {
2183  av_log(matroska->ctx, AV_LOG_ERROR,
2184  "Corrupt generic RM-style audio packet size\n");
2185  return AVERROR_INVALIDDATA;
2186  }
2187  for (x=0; x<w/sps; x++)
2188  memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
2189  }
2190 
2191  if (++track->audio.sub_packet_cnt >= h) {
2192  if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2193  ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2194  track->audio.sub_packet_cnt = 0;
2195  track->audio.pkt_cnt = h*w / a;
2196  }
2197  }
2198 
2199  while (track->audio.pkt_cnt) {
2200  AVPacket *pkt = NULL;
2201  if (!(pkt = av_mallocz(sizeof(AVPacket))) || av_new_packet(pkt, a) < 0){
2202  av_free(pkt);
2203  return AVERROR(ENOMEM);
2204  }
2205  memcpy(pkt->data, track->audio.buf
2206  + a * (h*w / a - track->audio.pkt_cnt--), a);
2207  pkt->pts = track->audio.buf_timecode;
2209  pkt->pos = pos;
2210  pkt->stream_index = st->index;
2211  dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2212  }
2213 
2214  return 0;
2215 }
2216 
2217 /* reconstruct full wavpack blocks from mangled matroska ones */
2219  uint8_t **pdst, int *size)
2220 {
2221  uint8_t *dst = NULL;
2222  int dstlen = 0;
2223  int srclen = *size;
2224  uint32_t samples;
2225  uint16_t ver;
2226  int ret, offset = 0;
2227 
2228  if (srclen < 12 || track->stream->codec->extradata_size < 2)
2229  return AVERROR_INVALIDDATA;
2230 
2231  ver = AV_RL16(track->stream->codec->extradata);
2232 
2233  samples = AV_RL32(src);
2234  src += 4;
2235  srclen -= 4;
2236 
2237  while (srclen >= 8) {
2238  int multiblock;
2239  uint32_t blocksize;
2240  uint8_t *tmp;
2241 
2242  uint32_t flags = AV_RL32(src);
2243  uint32_t crc = AV_RL32(src + 4);
2244  src += 8;
2245  srclen -= 8;
2246 
2247  multiblock = (flags & 0x1800) != 0x1800;
2248  if (multiblock) {
2249  if (srclen < 4) {
2250  ret = AVERROR_INVALIDDATA;
2251  goto fail;
2252  }
2253  blocksize = AV_RL32(src);
2254  src += 4;
2255  srclen -= 4;
2256  } else
2257  blocksize = srclen;
2258 
2259  if (blocksize > srclen) {
2260  ret = AVERROR_INVALIDDATA;
2261  goto fail;
2262  }
2263 
2264  tmp = av_realloc(dst, dstlen + blocksize + 32);
2265  if (!tmp) {
2266  ret = AVERROR(ENOMEM);
2267  goto fail;
2268  }
2269  dst = tmp;
2270  dstlen += blocksize + 32;
2271 
2272  AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2273  AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2274  AV_WL16(dst + offset + 8, ver); // version
2275  AV_WL16(dst + offset + 10, 0); // track/index_no
2276  AV_WL32(dst + offset + 12, 0); // total samples
2277  AV_WL32(dst + offset + 16, 0); // block index
2278  AV_WL32(dst + offset + 20, samples); // number of samples
2279  AV_WL32(dst + offset + 24, flags); // flags
2280  AV_WL32(dst + offset + 28, crc); // crc
2281  memcpy (dst + offset + 32, src, blocksize); // block data
2282 
2283  src += blocksize;
2284  srclen -= blocksize;
2285  offset += blocksize + 32;
2286  }
2287 
2288  *pdst = dst;
2289  *size = dstlen;
2290 
2291  return 0;
2292 
2293 fail:
2294  av_freep(&dst);
2295  return ret;
2296 }
2297 
2299  MatroskaTrack *track,
2300  AVStream *st,
2301  uint8_t *data, int data_len,
2302  uint64_t timecode,
2303  uint64_t duration,
2304  int64_t pos)
2305 {
2306  AVPacket *pkt;
2307  uint8_t *id, *settings, *text, *buf;
2308  int id_len, settings_len, text_len;
2309  uint8_t *p, *q;
2310  int err;
2311 
2312  if (data_len <= 0)
2313  return AVERROR_INVALIDDATA;
2314 
2315  p = data;
2316  q = data + data_len;
2317 
2318  id = p;
2319  id_len = -1;
2320  while (p < q) {
2321  if (*p == '\r' || *p == '\n') {
2322  id_len = p - id;
2323  if (*p == '\r')
2324  p++;
2325  break;
2326  }
2327  p++;
2328  }
2329 
2330  if (p >= q || *p != '\n')
2331  return AVERROR_INVALIDDATA;
2332  p++;
2333 
2334  settings = p;
2335  settings_len = -1;
2336  while (p < q) {
2337  if (*p == '\r' || *p == '\n') {
2338  settings_len = p - settings;
2339  if (*p == '\r')
2340  p++;
2341  break;
2342  }
2343  p++;
2344  }
2345 
2346  if (p >= q || *p != '\n')
2347  return AVERROR_INVALIDDATA;
2348  p++;
2349 
2350  text = p;
2351  text_len = q - p;
2352  while (text_len > 0) {
2353  const int len = text_len - 1;
2354  const uint8_t c = p[len];
2355  if (c != '\r' && c != '\n')
2356  break;
2357  text_len = len;
2358  }
2359 
2360  if (text_len <= 0)
2361  return AVERROR_INVALIDDATA;
2362 
2363  pkt = av_mallocz(sizeof(*pkt));
2364  err = av_new_packet(pkt, text_len);
2365  if (err < 0) {
2366  av_free(pkt);
2367  return AVERROR(err);
2368  }
2369 
2370  memcpy(pkt->data, text, text_len);
2371 
2372  if (id_len > 0) {
2373  buf = av_packet_new_side_data(pkt,
2375  id_len);
2376  if (buf == NULL) {
2377  av_free(pkt);
2378  return AVERROR(ENOMEM);
2379  }
2380  memcpy(buf, id, id_len);
2381  }
2382 
2383  if (settings_len > 0) {
2384  buf = av_packet_new_side_data(pkt,
2386  settings_len);
2387  if (buf == NULL) {
2388  av_free(pkt);
2389  return AVERROR(ENOMEM);
2390  }
2391  memcpy(buf, settings, settings_len);
2392  }
2393 
2394  // Do we need this for subtitles?
2395  // pkt->flags = AV_PKT_FLAG_KEY;
2396 
2397  pkt->stream_index = st->index;
2398  pkt->pts = timecode;
2399 
2400  // Do we need this for subtitles?
2401  // pkt->dts = timecode;
2402 
2403  pkt->duration = duration;
2404  pkt->pos = pos;
2405 
2406  dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2407  matroska->prev_pkt = pkt;
2408 
2409  return 0;
2410 }
2411 
2413  MatroskaTrack *track,
2414  AVStream *st,
2415  uint8_t *data, int pkt_size,
2416  uint64_t timecode, uint64_t lace_duration,
2417  int64_t pos, int is_keyframe,
2418  uint8_t *additional, uint64_t additional_id, int additional_size,
2419  int64_t discard_padding)
2420 {
2421  MatroskaTrackEncoding *encodings = track->encodings.elem;
2422  uint8_t *pkt_data = data;
2423  int offset = 0, res;
2424  AVPacket *pkt;
2425 
2426  if (encodings && !encodings->type && encodings->scope & 1) {
2427  res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2428  if (res < 0)
2429  return res;
2430  }
2431 
2432  if (st->codec->codec_id == AV_CODEC_ID_WAVPACK) {
2433  uint8_t *wv_data;
2434  res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
2435  if (res < 0) {
2436  av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing a wavpack block.\n");
2437  goto fail;
2438  }
2439  if (pkt_data != data)
2440  av_freep(&pkt_data);
2441  pkt_data = wv_data;
2442  }
2443 
2444  if (st->codec->codec_id == AV_CODEC_ID_PRORES)
2445  offset = 8;
2446 
2447  pkt = av_mallocz(sizeof(AVPacket));
2448  /* XXX: prevent data copy... */
2449  if (av_new_packet(pkt, pkt_size + offset) < 0) {
2450  av_free(pkt);
2451  res = AVERROR(ENOMEM);
2452  goto fail;
2453  }
2454 
2455  if (st->codec->codec_id == AV_CODEC_ID_PRORES) {
2456  uint8_t *buf = pkt->data;
2457  bytestream_put_be32(&buf, pkt_size);
2458  bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2459  }
2460 
2461  memcpy(pkt->data + offset, pkt_data, pkt_size);
2462 
2463  if (pkt_data != data)
2464  av_freep(&pkt_data);
2465 
2466  pkt->flags = is_keyframe;
2467  pkt->stream_index = st->index;
2468 
2469  if (additional_size > 0) {
2470  uint8_t *side_data = av_packet_new_side_data(pkt,
2472  additional_size + 8);
2473  if(side_data == NULL) {
2474  av_free_packet(pkt);
2475  av_free(pkt);
2476  return AVERROR(ENOMEM);
2477  }
2478  AV_WB64(side_data, additional_id);
2479  memcpy(side_data + 8, additional, additional_size);
2480  }
2481 
2482  if (discard_padding) {
2483  uint8_t *side_data = av_packet_new_side_data(pkt,
2485  10);
2486  if(side_data == NULL) {
2487  av_free_packet(pkt);
2488  av_free(pkt);
2489  return AVERROR(ENOMEM);
2490  }
2491  AV_WL32(side_data, 0);
2492  AV_WL32(side_data + 4, av_rescale_q(discard_padding,
2493  (AVRational){1, 1000000000},
2494  (AVRational){1, st->codec->sample_rate}));
2495  }
2496 
2497  if (track->ms_compat)
2498  pkt->dts = timecode;
2499  else
2500  pkt->pts = timecode;
2501  pkt->pos = pos;
2502  if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2503  /*
2504  * For backward compatibility.
2505  * Historically, we have put subtitle duration
2506  * in convergence_duration, on the off chance
2507  * that the time_scale is less than 1us, which
2508  * could result in a 32bit overflow on the
2509  * normal duration field.
2510  */
2511  pkt->convergence_duration = lace_duration;
2512  }
2513 
2514  if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE ||
2515  lace_duration <= INT_MAX) {
2516  /*
2517  * For non subtitle tracks, just store the duration
2518  * as normal.
2519  *
2520  * If it's a subtitle track and duration value does
2521  * not overflow a uint32, then also store it normally.
2522  */
2523  pkt->duration = lace_duration;
2524  }
2525 
2526 #if FF_API_ASS_SSA
2527  if (st->codec->codec_id == AV_CODEC_ID_SSA)
2528  matroska_fix_ass_packet(matroska, pkt, lace_duration);
2529 
2530  if (matroska->prev_pkt &&
2531  timecode != AV_NOPTS_VALUE &&
2532  matroska->prev_pkt->pts == timecode &&
2533  matroska->prev_pkt->stream_index == st->index &&
2534  st->codec->codec_id == AV_CODEC_ID_SSA)
2535  matroska_merge_packets(matroska->prev_pkt, pkt);
2536  else {
2537  dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
2538  matroska->prev_pkt = pkt;
2539  }
2540 #else
2541  dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2542  matroska->prev_pkt = pkt;
2543 #endif
2544 
2545  return 0;
2546 fail:
2547  if (pkt_data != data)
2548  av_freep(&pkt_data);
2549  return res;
2550 }
2551 
2553  int size, int64_t pos, uint64_t cluster_time,
2554  uint64_t block_duration, int is_keyframe,
2555  uint8_t *additional, uint64_t additional_id, int additional_size,
2556  int64_t cluster_pos, int64_t discard_padding)
2557 {
2558  uint64_t timecode = AV_NOPTS_VALUE;
2559  MatroskaTrack *track;
2560  int res = 0;
2561  AVStream *st;
2562  int16_t block_time;
2563  uint32_t *lace_size = NULL;
2564  int n, flags, laces = 0;
2565  uint64_t num;
2566  int trust_default_duration = 1;
2567 
2568  if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2569  av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2570  return n;
2571  }
2572  data += n;
2573  size -= n;
2574 
2575  track = matroska_find_track_by_num(matroska, num);
2576  if (!track || !track->stream) {
2577  av_log(matroska->ctx, AV_LOG_INFO,
2578  "Invalid stream %"PRIu64" or size %u\n", num, size);
2579  return AVERROR_INVALIDDATA;
2580  } else if (size <= 3)
2581  return 0;
2582  st = track->stream;
2583  if (st->discard >= AVDISCARD_ALL)
2584  return res;
2585  av_assert1(block_duration != AV_NOPTS_VALUE);
2586 
2587  block_time = sign_extend(AV_RB16(data), 16);
2588  data += 2;
2589  flags = *data++;
2590  size -= 3;
2591  if (is_keyframe == -1)
2592  is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2593 
2594  if (cluster_time != (uint64_t)-1
2595  && (block_time >= 0 || cluster_time >= -block_time)) {
2596  timecode = cluster_time + block_time;
2597  if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
2598  && timecode < track->end_timecode)
2599  is_keyframe = 0; /* overlapping subtitles are not key frame */
2600  if (is_keyframe)
2601  av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
2602  }
2603 
2604  if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2605  if (timecode < matroska->skip_to_timecode)
2606  return res;
2607  if (is_keyframe)
2608  matroska->skip_to_keyframe = 0;
2609  else if (!st->skip_to_keyframe) {
2610  av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2611  matroska->skip_to_keyframe = 0;
2612  }
2613  }
2614 
2615  res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
2616  &lace_size, &laces);
2617 
2618  if (res)
2619  goto end;
2620 
2621  if (track->audio.samplerate == 8000) {
2622  // If this is needed for more codecs, then add them here
2623  if (st->codec->codec_id == AV_CODEC_ID_AC3) {
2624  if(track->audio.samplerate != st->codec->sample_rate || !st->codec->frame_size)
2625  trust_default_duration = 0;
2626  }
2627  }
2628 
2629  if (!block_duration && trust_default_duration)
2630  block_duration = track->default_duration * laces / matroska->time_scale;
2631 
2632  if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
2633  track->end_timecode =
2634  FFMAX(track->end_timecode, timecode + block_duration);
2635 
2636  for (n = 0; n < laces; n++) {
2637  int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
2638 
2639  if (lace_size[n] > size) {
2640  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2641  break;
2642  }
2643 
2644  if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2645  st->codec->codec_id == AV_CODEC_ID_COOK ||
2646  st->codec->codec_id == AV_CODEC_ID_SIPR ||
2647  st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2648  st->codec->block_align && track->audio.sub_packet_size) {
2649 
2650  res = matroska_parse_rm_audio(matroska, track, st, data,
2651  lace_size[n],
2652  timecode, pos);
2653  if (res)
2654  goto end;
2655 
2656  } else if (st->codec->codec_id == AV_CODEC_ID_WEBVTT) {
2657  res = matroska_parse_webvtt(matroska, track, st,
2658  data, lace_size[n],
2659  timecode, lace_duration,
2660  pos);
2661  if (res)
2662  goto end;
2663 
2664  } else {
2665  res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2666  timecode, lace_duration,
2667  pos, !n? is_keyframe : 0,
2668  additional, additional_id, additional_size,
2669  discard_padding);
2670  if (res)
2671  goto end;
2672  }
2673 
2674  if (timecode != AV_NOPTS_VALUE)
2675  timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2676  data += lace_size[n];
2677  size -= lace_size[n];
2678  }
2679 
2680 end:
2681  av_free(lace_size);
2682  return res;
2683 }
2684 
2686 {
2687  EbmlList *blocks_list;
2688  MatroskaBlock *blocks;
2689  int i, res;
2690  res = ebml_parse(matroska,
2691  matroska_cluster_incremental_parsing,
2692  &matroska->current_cluster);
2693  if (res == 1) {
2694  /* New Cluster */
2695  if (matroska->current_cluster_pos)
2696  ebml_level_end(matroska);
2697  ebml_free(matroska_cluster, &matroska->current_cluster);
2698  memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2699  matroska->current_cluster_num_blocks = 0;
2700  matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2701  matroska->prev_pkt = NULL;
2702  /* sizeof the ID which was already read */
2703  if (matroska->current_id)
2704  matroska->current_cluster_pos -= 4;
2705  res = ebml_parse(matroska,
2706  matroska_clusters_incremental,
2707  &matroska->current_cluster);
2708  /* Try parsing the block again. */
2709  if (res == 1)
2710  res = ebml_parse(matroska,
2711  matroska_cluster_incremental_parsing,
2712  &matroska->current_cluster);
2713  }
2714 
2715  if (!res &&
2716  matroska->current_cluster_num_blocks <
2717  matroska->current_cluster.blocks.nb_elem) {
2718  blocks_list = &matroska->current_cluster.blocks;
2719  blocks = blocks_list->elem;
2720 
2721  matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2722  i = blocks_list->nb_elem - 1;
2723  if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2724  int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2725  uint8_t* additional = blocks[i].additional.size > 0 ?
2726  blocks[i].additional.data : NULL;
2727  if (!blocks[i].non_simple)
2728  blocks[i].duration = 0;
2729  res = matroska_parse_block(matroska,
2730  blocks[i].bin.data, blocks[i].bin.size,
2731  blocks[i].bin.pos,
2732  matroska->current_cluster.timecode,
2733  blocks[i].duration, is_keyframe,
2734  additional, blocks[i].additional_id,
2735  blocks[i].additional.size,
2736  matroska->current_cluster_pos,
2737  blocks[i].discard_padding);
2738  }
2739  }
2740 
2741  return res;
2742 }
2743 
2745 {
2746  MatroskaCluster cluster = { 0 };
2747  EbmlList *blocks_list;
2748  MatroskaBlock *blocks;
2749  int i, res;
2750  int64_t pos;
2751  if (!matroska->contains_ssa)
2752  return matroska_parse_cluster_incremental(matroska);
2753  pos = avio_tell(matroska->ctx->pb);
2754  matroska->prev_pkt = NULL;
2755  if (matroska->current_id)
2756  pos -= 4; /* sizeof the ID which was already read */
2757  res = ebml_parse(matroska, matroska_clusters, &cluster);
2758  blocks_list = &cluster.blocks;
2759  blocks = blocks_list->elem;
2760  for (i=0; i<blocks_list->nb_elem; i++)
2761  if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2762  int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2763  res=matroska_parse_block(matroska,
2764  blocks[i].bin.data, blocks[i].bin.size,
2765  blocks[i].bin.pos, cluster.timecode,
2766  blocks[i].duration, is_keyframe, NULL, 0, 0,
2767  pos, blocks[i].discard_padding);
2768  }
2769  ebml_free(matroska_cluster, &cluster);
2770  return res;
2771 }
2772 
2774 {
2775  MatroskaDemuxContext *matroska = s->priv_data;
2776 
2777  while (matroska_deliver_packet(matroska, pkt)) {
2778  int64_t pos = avio_tell(matroska->ctx->pb);
2779  if (matroska->done)
2780  return AVERROR_EOF;
2781  if (matroska_parse_cluster(matroska) < 0)
2782  matroska_resync(matroska, pos);
2783  }
2784 
2785  return 0;
2786 }
2787 
2788 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2789  int64_t timestamp, int flags)
2790 {
2791  MatroskaDemuxContext *matroska = s->priv_data;
2792  MatroskaTrack *tracks = matroska->tracks.elem;
2793  AVStream *st = s->streams[stream_index];
2794  int i, index, index_sub, index_min;
2795 
2796  /* Parse the CUES now since we need the index data to seek. */
2797  if (matroska->cues_parsing_deferred > 0) {
2798  matroska->cues_parsing_deferred = 0;
2799  matroska_parse_cues(matroska);
2800  }
2801 
2802  if (!st->nb_index_entries)
2803  goto err;
2804  timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2805 
2806  if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2807  avio_seek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
2808  matroska->current_id = 0;
2809  while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2810  matroska_clear_queue(matroska);
2811  if (matroska_parse_cluster(matroska) < 0)
2812  break;
2813  }
2814  }
2815 
2816  matroska_clear_queue(matroska);
2817  if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
2818  goto err;
2819 
2820  index_min = index;
2821  for (i=0; i < matroska->tracks.nb_elem; i++) {
2822  tracks[i].audio.pkt_cnt = 0;
2823  tracks[i].audio.sub_packet_cnt = 0;
2824  tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2825  tracks[i].end_timecode = 0;
2826  if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
2827  && tracks[i].stream->discard != AVDISCARD_ALL) {
2828  index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
2829  while(index_sub >= 0
2830  && index_min >= 0
2831  && tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos
2832  && st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
2833  index_min--;
2834  }
2835  }
2836 
2837  avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2838  matroska->current_id = 0;
2839  if (flags & AVSEEK_FLAG_ANY) {
2840  st->skip_to_keyframe = 0;
2841  matroska->skip_to_timecode = timestamp;
2842  } else {
2843  st->skip_to_keyframe = 1;
2844  matroska->skip_to_timecode = st->index_entries[index].timestamp;
2845  }
2846  matroska->skip_to_keyframe = 1;
2847  matroska->done = 0;
2848  matroska->num_levels = 0;
2849  ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2850  return 0;
2851 err:
2852  // slightly hackish but allows proper fallback to
2853  // the generic seeking code.
2854  matroska_clear_queue(matroska);
2855  matroska->current_id = 0;
2856  st->skip_to_keyframe =
2857  matroska->skip_to_keyframe = 0;
2858  matroska->done = 0;
2859  matroska->num_levels = 0;
2860  return -1;
2861 }
2862 
2864 {
2865  MatroskaDemuxContext *matroska = s->priv_data;
2866  MatroskaTrack *tracks = matroska->tracks.elem;
2867  int n;
2868 
2869  matroska_clear_queue(matroska);
2870 
2871  for (n=0; n < matroska->tracks.nb_elem; n++)
2872  if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2873  av_free(tracks[n].audio.buf);
2874  ebml_free(matroska_cluster, &matroska->current_cluster);
2875  ebml_free(matroska_segment, matroska);
2876 
2877  return 0;
2878 }
2879 
2881  .name = "matroska,webm",
2882  .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
2883  .priv_data_size = sizeof(MatroskaDemuxContext),
2889 };