Statistics
| Branch: | Revision:

ffmpeg / libavformat / matroskadec.c @ f7cf0f31

History | View | Annotate | Download (68.4 KB)

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
 * by Ronald Bultje <rbultje@ronald.bitfreak.net>
26
 * with a little help from Moritz Bunkus <moritz@bunkus.org>
27
 * totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28
 * Specs available on the Matroska project page: http://www.matroska.org/.
29
 */
30

    
31
#include <stdio.h>
32
#include "avformat.h"
33
#include "internal.h"
34
/* For ff_codec_get_id(). */
35
#include "riff.h"
36
#include "isom.h"
37
#include "rm.h"
38
#include "matroska.h"
39
#include "libavcodec/mpeg4audio.h"
40
#include "libavutil/intfloat_readwrite.h"
41
#include "libavutil/intreadwrite.h"
42
#include "libavutil/avstring.h"
43
#include "libavutil/lzo.h"
44
#if CONFIG_ZLIB
45
#include <zlib.h>
46
#endif
47
#if CONFIG_BZLIB
48
#include <bzlib.h>
49
#endif
50

    
51
typedef enum {
52
    EBML_NONE,
53
    EBML_UINT,
54
    EBML_FLOAT,
55
    EBML_STR,
56
    EBML_UTF8,
57
    EBML_BIN,
58
    EBML_NEST,
59
    EBML_PASS,
60
    EBML_STOP,
61
} EbmlType;
62

    
63
typedef const struct EbmlSyntax {
64
    uint32_t id;
65
    EbmlType type;
66
    int list_elem_size;
67
    int data_offset;
68
    union {
69
        uint64_t    u;
70
        double      f;
71
        const char *s;
72
        const struct EbmlSyntax *n;
73
    } def;
74
} EbmlSyntax;
75

    
76
typedef struct {
77
    int nb_elem;
78
    void *elem;
79
} EbmlList;
80

    
81
typedef struct {
82
    int      size;
83
    uint8_t *data;
84
    int64_t  pos;
85
} EbmlBin;
86

    
87
typedef struct {
88
    uint64_t version;
89
    uint64_t max_size;
90
    uint64_t id_length;
91
    char    *doctype;
92
    uint64_t doctype_version;
93
} Ebml;
94

    
95
typedef struct {
96
    uint64_t algo;
97
    EbmlBin  settings;
98
} MatroskaTrackCompression;
99

    
100
typedef struct {
101
    uint64_t scope;
102
    uint64_t type;
103
    MatroskaTrackCompression compression;
104
} MatroskaTrackEncoding;
105

    
106
typedef struct {
107
    double   frame_rate;
108
    uint64_t display_width;
109
    uint64_t display_height;
110
    uint64_t pixel_width;
111
    uint64_t pixel_height;
112
    uint64_t fourcc;
113
} MatroskaTrackVideo;
114

    
115
typedef struct {
116
    double   samplerate;
117
    double   out_samplerate;
118
    uint64_t bitdepth;
119
    uint64_t channels;
120

    
121
    /* real audio header (extracted from extradata) */
122
    int      coded_framesize;
123
    int      sub_packet_h;
124
    int      frame_size;
125
    int      sub_packet_size;
126
    int      sub_packet_cnt;
127
    int      pkt_cnt;
128
    uint8_t *buf;
129
} MatroskaTrackAudio;
130

    
131
typedef struct {
132
    uint64_t num;
133
    uint64_t uid;
134
    uint64_t type;
135
    char    *name;
136
    char    *codec_id;
137
    EbmlBin  codec_priv;
138
    char    *language;
139
    double time_scale;
140
    uint64_t default_duration;
141
    uint64_t flag_default;
142
    uint64_t flag_forced;
143
    MatroskaTrackVideo video;
144
    MatroskaTrackAudio audio;
145
    EbmlList encodings;
146

    
147
    AVStream *stream;
148
    int64_t end_timecode;
149
    int ms_compat;
150
} MatroskaTrack;
151

    
152
typedef struct {
153
    uint64_t uid;
154
    char *filename;
155
    char *mime;
156
    EbmlBin bin;
157

    
158
    AVStream *stream;
159
} MatroskaAttachement;
160

    
161
typedef struct {
162
    uint64_t start;
163
    uint64_t end;
164
    uint64_t uid;
165
    char    *title;
166

    
167
    AVChapter *chapter;
168
} MatroskaChapter;
169

    
170
typedef struct {
171
    uint64_t track;
172
    uint64_t pos;
173
} MatroskaIndexPos;
174

    
175
typedef struct {
176
    uint64_t time;
177
    EbmlList pos;
178
} MatroskaIndex;
179

    
180
typedef struct {
181
    char *name;
182
    char *string;
183
    char *lang;
184
    uint64_t def;
185
    EbmlList sub;
186
} MatroskaTag;
187

    
188
typedef struct {
189
    char    *type;
190
    uint64_t typevalue;
191
    uint64_t trackuid;
192
    uint64_t chapteruid;
193
    uint64_t attachuid;
194
} MatroskaTagTarget;
195

    
196
typedef struct {
197
    MatroskaTagTarget target;
198
    EbmlList tag;
199
} MatroskaTags;
200

    
201
typedef struct {
202
    uint64_t id;
203
    uint64_t pos;
204
} MatroskaSeekhead;
205

    
206
typedef struct {
207
    uint64_t start;
208
    uint64_t length;
209
} MatroskaLevel;
210

    
211
typedef struct {
212
    AVFormatContext *ctx;
213

    
214
    /* EBML stuff */
215
    int num_levels;
216
    MatroskaLevel levels[EBML_MAX_DEPTH];
217
    int level_up;
218
    uint32_t current_id;
219

    
220
    uint64_t time_scale;
221
    double   duration;
222
    char    *title;
223
    EbmlList tracks;
224
    EbmlList attachments;
225
    EbmlList chapters;
226
    EbmlList index;
227
    EbmlList tags;
228
    EbmlList seekhead;
229

    
230
    /* byte position of the segment inside the stream */
231
    int64_t segment_start;
232

    
233
    /* the packet queue */
234
    AVPacket **packets;
235
    int num_packets;
236
    AVPacket *prev_pkt;
237

    
238
    int done;
239

    
240
    /* What to skip before effectively reading a packet. */
241
    int skip_to_keyframe;
242
    uint64_t skip_to_timecode;
243
} MatroskaDemuxContext;
244

    
245
typedef struct {
246
    uint64_t duration;
247
    int64_t  reference;
248
    uint64_t non_simple;
249
    EbmlBin  bin;
250
} MatroskaBlock;
251

    
252
typedef struct {
253
    uint64_t timecode;
254
    EbmlList blocks;
255
} MatroskaCluster;
256

    
257
static EbmlSyntax ebml_header[] = {
258
    { EBML_ID_EBMLREADVERSION,        EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
259
    { EBML_ID_EBMLMAXSIZELENGTH,      EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
260
    { EBML_ID_EBMLMAXIDLENGTH,        EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
261
    { EBML_ID_DOCTYPE,                EBML_STR,  0, offsetof(Ebml,doctype), {.s="(none)"} },
262
    { EBML_ID_DOCTYPEREADVERSION,     EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
263
    { EBML_ID_EBMLVERSION,            EBML_NONE },
264
    { EBML_ID_DOCTYPEVERSION,         EBML_NONE },
265
    { 0 }
266
};
267

    
268
static EbmlSyntax ebml_syntax[] = {
269
    { EBML_ID_HEADER,                 EBML_NEST, 0, 0, {.n=ebml_header} },
270
    { 0 }
271
};
272

    
273
static EbmlSyntax matroska_info[] = {
274
    { MATROSKA_ID_TIMECODESCALE,      EBML_UINT,  0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
275
    { MATROSKA_ID_DURATION,           EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
276
    { MATROSKA_ID_TITLE,              EBML_UTF8,  0, offsetof(MatroskaDemuxContext,title) },
277
    { MATROSKA_ID_WRITINGAPP,         EBML_NONE },
278
    { MATROSKA_ID_MUXINGAPP,          EBML_NONE },
279
    { MATROSKA_ID_DATEUTC,            EBML_NONE },
280
    { MATROSKA_ID_SEGMENTUID,         EBML_NONE },
281
    { 0 }
282
};
283

    
284
static EbmlSyntax matroska_track_video[] = {
285
    { MATROSKA_ID_VIDEOFRAMERATE,     EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
286
    { MATROSKA_ID_VIDEODISPLAYWIDTH,  EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
287
    { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
288
    { MATROSKA_ID_VIDEOPIXELWIDTH,    EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
289
    { MATROSKA_ID_VIDEOPIXELHEIGHT,   EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
290
    { MATROSKA_ID_VIDEOCOLORSPACE,    EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
291
    { MATROSKA_ID_VIDEOPIXELCROPB,    EBML_NONE },
292
    { MATROSKA_ID_VIDEOPIXELCROPT,    EBML_NONE },
293
    { MATROSKA_ID_VIDEOPIXELCROPL,    EBML_NONE },
294
    { MATROSKA_ID_VIDEOPIXELCROPR,    EBML_NONE },
295
    { MATROSKA_ID_VIDEODISPLAYUNIT,   EBML_NONE },
296
    { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
297
    { MATROSKA_ID_VIDEOSTEREOMODE,    EBML_NONE },
298
    { MATROSKA_ID_VIDEOASPECTRATIO,   EBML_NONE },
299
    { 0 }
300
};
301

    
302
static EbmlSyntax matroska_track_audio[] = {
303
    { MATROSKA_ID_AUDIOSAMPLINGFREQ,  EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
304
    { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
305
    { MATROSKA_ID_AUDIOBITDEPTH,      EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
306
    { MATROSKA_ID_AUDIOCHANNELS,      EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
307
    { 0 }
308
};
309

    
310
static EbmlSyntax matroska_track_encoding_compression[] = {
311
    { MATROSKA_ID_ENCODINGCOMPALGO,   EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
312
    { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
313
    { 0 }
314
};
315

    
316
static EbmlSyntax matroska_track_encoding[] = {
317
    { MATROSKA_ID_ENCODINGSCOPE,      EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
318
    { MATROSKA_ID_ENCODINGTYPE,       EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
319
    { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
320
    { MATROSKA_ID_ENCODINGORDER,      EBML_NONE },
321
    { 0 }
322
};
323

    
324
static EbmlSyntax matroska_track_encodings[] = {
325
    { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
326
    { 0 }
327
};
328

    
329
static EbmlSyntax matroska_track[] = {
330
    { MATROSKA_ID_TRACKNUMBER,          EBML_UINT, 0, offsetof(MatroskaTrack,num) },
331
    { MATROSKA_ID_TRACKNAME,            EBML_UTF8, 0, offsetof(MatroskaTrack,name) },
332
    { MATROSKA_ID_TRACKUID,             EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
333
    { MATROSKA_ID_TRACKTYPE,            EBML_UINT, 0, offsetof(MatroskaTrack,type) },
334
    { MATROSKA_ID_CODECID,              EBML_STR,  0, offsetof(MatroskaTrack,codec_id) },
335
    { MATROSKA_ID_CODECPRIVATE,         EBML_BIN,  0, offsetof(MatroskaTrack,codec_priv) },
336
    { MATROSKA_ID_TRACKLANGUAGE,        EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
337
    { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
338
    { MATROSKA_ID_TRACKTIMECODESCALE,   EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
339
    { MATROSKA_ID_TRACKFLAGDEFAULT,     EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
340
    { MATROSKA_ID_TRACKFLAGFORCED,      EBML_UINT, 0, offsetof(MatroskaTrack,flag_forced), {.u=0} },
341
    { MATROSKA_ID_TRACKVIDEO,           EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
342
    { MATROSKA_ID_TRACKAUDIO,           EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
343
    { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
344
    { MATROSKA_ID_TRACKFLAGENABLED,     EBML_NONE },
345
    { MATROSKA_ID_TRACKFLAGLACING,      EBML_NONE },
346
    { MATROSKA_ID_CODECNAME,            EBML_NONE },
347
    { MATROSKA_ID_CODECDECODEALL,       EBML_NONE },
348
    { MATROSKA_ID_CODECINFOURL,         EBML_NONE },
349
    { MATROSKA_ID_CODECDOWNLOADURL,     EBML_NONE },
350
    { MATROSKA_ID_TRACKMINCACHE,        EBML_NONE },
351
    { MATROSKA_ID_TRACKMAXCACHE,        EBML_NONE },
352
    { MATROSKA_ID_TRACKMAXBLKADDID,     EBML_NONE },
353
    { 0 }
354
};
355

    
356
static EbmlSyntax matroska_tracks[] = {
357
    { MATROSKA_ID_TRACKENTRY,         EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
358
    { 0 }
359
};
360

    
361
static EbmlSyntax matroska_attachment[] = {
362
    { MATROSKA_ID_FILEUID,            EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
363
    { MATROSKA_ID_FILENAME,           EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
364
    { MATROSKA_ID_FILEMIMETYPE,       EBML_STR,  0, offsetof(MatroskaAttachement,mime) },
365
    { MATROSKA_ID_FILEDATA,           EBML_BIN,  0, offsetof(MatroskaAttachement,bin) },
366
    { MATROSKA_ID_FILEDESC,           EBML_NONE },
367
    { 0 }
368
};
369

    
370
static EbmlSyntax matroska_attachments[] = {
371
    { MATROSKA_ID_ATTACHEDFILE,       EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
372
    { 0 }
373
};
374

    
375
static EbmlSyntax matroska_chapter_display[] = {
376
    { MATROSKA_ID_CHAPSTRING,         EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
377
    { MATROSKA_ID_CHAPLANG,           EBML_NONE },
378
    { 0 }
379
};
380

    
381
static EbmlSyntax matroska_chapter_entry[] = {
382
    { MATROSKA_ID_CHAPTERTIMESTART,   EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
383
    { MATROSKA_ID_CHAPTERTIMEEND,     EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
384
    { MATROSKA_ID_CHAPTERUID,         EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
385
    { MATROSKA_ID_CHAPTERDISPLAY,     EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
386
    { MATROSKA_ID_CHAPTERFLAGHIDDEN,  EBML_NONE },
387
    { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
388
    { MATROSKA_ID_CHAPTERPHYSEQUIV,   EBML_NONE },
389
    { MATROSKA_ID_CHAPTERATOM,        EBML_NONE },
390
    { 0 }
391
};
392

    
393
static EbmlSyntax matroska_chapter[] = {
394
    { MATROSKA_ID_CHAPTERATOM,        EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
395
    { MATROSKA_ID_EDITIONUID,         EBML_NONE },
396
    { MATROSKA_ID_EDITIONFLAGHIDDEN,  EBML_NONE },
397
    { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
398
    { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
399
    { 0 }
400
};
401

    
402
static EbmlSyntax matroska_chapters[] = {
403
    { MATROSKA_ID_EDITIONENTRY,       EBML_NEST, 0, 0, {.n=matroska_chapter} },
404
    { 0 }
405
};
406

    
407
static EbmlSyntax matroska_index_pos[] = {
408
    { MATROSKA_ID_CUETRACK,           EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
409
    { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos)   },
410
    { MATROSKA_ID_CUEBLOCKNUMBER,     EBML_NONE },
411
    { 0 }
412
};
413

    
414
static EbmlSyntax matroska_index_entry[] = {
415
    { MATROSKA_ID_CUETIME,            EBML_UINT, 0, offsetof(MatroskaIndex,time) },
416
    { MATROSKA_ID_CUETRACKPOSITION,   EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
417
    { 0 }
418
};
419

    
420
static EbmlSyntax matroska_index[] = {
421
    { MATROSKA_ID_POINTENTRY,         EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
422
    { 0 }
423
};
424

    
425
static EbmlSyntax matroska_simpletag[] = {
426
    { MATROSKA_ID_TAGNAME,            EBML_UTF8, 0, offsetof(MatroskaTag,name) },
427
    { MATROSKA_ID_TAGSTRING,          EBML_UTF8, 0, offsetof(MatroskaTag,string) },
428
    { MATROSKA_ID_TAGLANG,            EBML_STR,  0, offsetof(MatroskaTag,lang), {.s="und"} },
429
    { MATROSKA_ID_TAGDEFAULT,         EBML_UINT, 0, offsetof(MatroskaTag,def) },
430
    { MATROSKA_ID_TAGDEFAULT_BUG,     EBML_UINT, 0, offsetof(MatroskaTag,def) },
431
    { MATROSKA_ID_SIMPLETAG,          EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
432
    { 0 }
433
};
434

    
435
static EbmlSyntax matroska_tagtargets[] = {
436
    { MATROSKA_ID_TAGTARGETS_TYPE,      EBML_STR,  0, offsetof(MatroskaTagTarget,type) },
437
    { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
438
    { MATROSKA_ID_TAGTARGETS_TRACKUID,  EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
439
    { MATROSKA_ID_TAGTARGETS_CHAPTERUID,EBML_UINT, 0, offsetof(MatroskaTagTarget,chapteruid) },
440
    { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
441
    { 0 }
442
};
443

    
444
static EbmlSyntax matroska_tag[] = {
445
    { MATROSKA_ID_SIMPLETAG,          EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
446
    { MATROSKA_ID_TAGTARGETS,         EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
447
    { 0 }
448
};
449

    
450
static EbmlSyntax matroska_tags[] = {
451
    { MATROSKA_ID_TAG,                EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
452
    { 0 }
453
};
454

    
455
static EbmlSyntax matroska_seekhead_entry[] = {
456
    { MATROSKA_ID_SEEKID,             EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
457
    { MATROSKA_ID_SEEKPOSITION,       EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
458
    { 0 }
459
};
460

    
461
static EbmlSyntax matroska_seekhead[] = {
462
    { MATROSKA_ID_SEEKENTRY,          EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
463
    { 0 }
464
};
465

    
466
static EbmlSyntax matroska_segment[] = {
467
    { MATROSKA_ID_INFO,           EBML_NEST, 0, 0, {.n=matroska_info       } },
468
    { MATROSKA_ID_TRACKS,         EBML_NEST, 0, 0, {.n=matroska_tracks     } },
469
    { MATROSKA_ID_ATTACHMENTS,    EBML_NEST, 0, 0, {.n=matroska_attachments} },
470
    { MATROSKA_ID_CHAPTERS,       EBML_NEST, 0, 0, {.n=matroska_chapters   } },
471
    { MATROSKA_ID_CUES,           EBML_NEST, 0, 0, {.n=matroska_index      } },
472
    { MATROSKA_ID_TAGS,           EBML_NEST, 0, 0, {.n=matroska_tags       } },
473
    { MATROSKA_ID_SEEKHEAD,       EBML_NEST, 0, 0, {.n=matroska_seekhead   } },
474
    { MATROSKA_ID_CLUSTER,        EBML_STOP },
475
    { 0 }
476
};
477

    
478
static EbmlSyntax matroska_segments[] = {
479
    { MATROSKA_ID_SEGMENT,        EBML_NEST, 0, 0, {.n=matroska_segment    } },
480
    { 0 }
481
};
482

    
483
static EbmlSyntax matroska_blockgroup[] = {
484
    { MATROSKA_ID_BLOCK,          EBML_BIN,  0, offsetof(MatroskaBlock,bin) },
485
    { MATROSKA_ID_SIMPLEBLOCK,    EBML_BIN,  0, offsetof(MatroskaBlock,bin) },
486
    { MATROSKA_ID_BLOCKDURATION,  EBML_UINT, 0, offsetof(MatroskaBlock,duration), {.u=AV_NOPTS_VALUE} },
487
    { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
488
    { 1,                          EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
489
    { 0 }
490
};
491

    
492
static EbmlSyntax matroska_cluster[] = {
493
    { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
494
    { MATROSKA_ID_BLOCKGROUP,     EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
495
    { MATROSKA_ID_SIMPLEBLOCK,    EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
496
    { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
497
    { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
498
    { 0 }
499
};
500

    
501
static EbmlSyntax matroska_clusters[] = {
502
    { MATROSKA_ID_CLUSTER,        EBML_NEST, 0, 0, {.n=matroska_cluster} },
503
    { MATROSKA_ID_INFO,           EBML_NONE },
504
    { MATROSKA_ID_CUES,           EBML_NONE },
505
    { MATROSKA_ID_TAGS,           EBML_NONE },
506
    { MATROSKA_ID_SEEKHEAD,       EBML_NONE },
507
    { 0 }
508
};
509

    
510
static const char *matroska_doctypes[] = { "matroska", "webm" };
511

    
512
/*
513
 * Return: Whether we reached the end of a level in the hierarchy or not.
514
 */
515
static int ebml_level_end(MatroskaDemuxContext *matroska)
516
{
517
    ByteIOContext *pb = matroska->ctx->pb;
518
    int64_t pos = url_ftell(pb);
519

    
520
    if (matroska->num_levels > 0) {
521
        MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
522
        if (pos - level->start >= level->length || matroska->current_id) {
523
            matroska->num_levels--;
524
            return 1;
525
        }
526
    }
527
    return 0;
528
}
529

    
530
/*
531
 * Read: an "EBML number", which is defined as a variable-length
532
 * array of bytes. The first byte indicates the length by giving a
533
 * number of 0-bits followed by a one. The position of the first
534
 * "one" bit inside the first byte indicates the length of this
535
 * number.
536
 * Returns: number of bytes read, < 0 on error
537
 */
538
static int ebml_read_num(MatroskaDemuxContext *matroska, ByteIOContext *pb,
539
                         int max_size, uint64_t *number)
540
{
541
    int len_mask = 0x80, read = 1, n = 1;
542
    int64_t total = 0;
543

    
544
    /* The first byte tells us the length in bytes - get_byte() can normally
545
     * return 0, but since that's not a valid first ebmlID byte, we can
546
     * use it safely here to catch EOS. */
547
    if (!(total = get_byte(pb))) {
548
        /* we might encounter EOS here */
549
        if (!url_feof(pb)) {
550
            int64_t pos = url_ftell(pb);
551
            av_log(matroska->ctx, AV_LOG_ERROR,
552
                   "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
553
                   pos, pos);
554
        }
555
        return AVERROR(EIO); /* EOS or actual I/O error */
556
    }
557

    
558
    /* get the length of the EBML number */
559
    while (read <= max_size && !(total & len_mask)) {
560
        read++;
561
        len_mask >>= 1;
562
    }
563
    if (read > max_size) {
564
        int64_t pos = url_ftell(pb) - 1;
565
        av_log(matroska->ctx, AV_LOG_ERROR,
566
               "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
567
               (uint8_t) total, pos, pos);
568
        return AVERROR_INVALIDDATA;
569
    }
570

    
571
    /* read out length */
572
    total &= ~len_mask;
573
    while (n++ < read)
574
        total = (total << 8) | get_byte(pb);
575

    
576
    *number = total;
577

    
578
    return read;
579
}
580

    
581
/*
582
 * Read the next element as an unsigned int.
583
 * 0 is success, < 0 is failure.
584
 */
585
static int ebml_read_uint(ByteIOContext *pb, int size, uint64_t *num)
586
{
587
    int n = 0;
588

    
589
    if (size < 1 || size > 8)
590
        return AVERROR_INVALIDDATA;
591

    
592
    /* big-endian ordering; build up number */
593
    *num = 0;
594
    while (n++ < size)
595
        *num = (*num << 8) | get_byte(pb);
596

    
597
    return 0;
598
}
599

    
600
/*
601
 * Read the next element as a float.
602
 * 0 is success, < 0 is failure.
603
 */
604
static int ebml_read_float(ByteIOContext *pb, int size, double *num)
605
{
606
    if (size == 4) {
607
        *num= av_int2flt(get_be32(pb));
608
    } else if(size==8){
609
        *num= av_int2dbl(get_be64(pb));
610
    } else
611
        return AVERROR_INVALIDDATA;
612

    
613
    return 0;
614
}
615

    
616
/*
617
 * Read the next element as an ASCII string.
618
 * 0 is success, < 0 is failure.
619
 */
620
static int ebml_read_ascii(ByteIOContext *pb, int size, char **str)
621
{
622
    av_free(*str);
623
    /* EBML strings are usually not 0-terminated, so we allocate one
624
     * byte more, read the string and NULL-terminate it ourselves. */
625
    if (!(*str = av_malloc(size + 1)))
626
        return AVERROR(ENOMEM);
627
    if (get_buffer(pb, (uint8_t *) *str, size) != size) {
628
        av_freep(str);
629
        return AVERROR(EIO);
630
    }
631
    (*str)[size] = '\0';
632

    
633
    return 0;
634
}
635

    
636
/*
637
 * Read the next element as binary data.
638
 * 0 is success, < 0 is failure.
639
 */
640
static int ebml_read_binary(ByteIOContext *pb, int length, EbmlBin *bin)
641
{
642
    av_free(bin->data);
643
    if (!(bin->data = av_malloc(length)))
644
        return AVERROR(ENOMEM);
645

    
646
    bin->size = length;
647
    bin->pos  = url_ftell(pb);
648
    if (get_buffer(pb, bin->data, length) != length) {
649
        av_freep(&bin->data);
650
        return AVERROR(EIO);
651
    }
652

    
653
    return 0;
654
}
655

    
656
/*
657
 * Read the next element, but only the header. The contents
658
 * are supposed to be sub-elements which can be read separately.
659
 * 0 is success, < 0 is failure.
660
 */
661
static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
662
{
663
    ByteIOContext *pb = matroska->ctx->pb;
664
    MatroskaLevel *level;
665

    
666
    if (matroska->num_levels >= EBML_MAX_DEPTH) {
667
        av_log(matroska->ctx, AV_LOG_ERROR,
668
               "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
669
        return AVERROR(ENOSYS);
670
    }
671

    
672
    level = &matroska->levels[matroska->num_levels++];
673
    level->start = url_ftell(pb);
674
    level->length = length;
675

    
676
    return 0;
677
}
678

    
679
/*
680
 * Read signed/unsigned "EBML" numbers.
681
 * Return: number of bytes processed, < 0 on error
682
 */
683
static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
684
                                 uint8_t *data, uint32_t size, uint64_t *num)
685
{
686
    ByteIOContext pb;
687
    init_put_byte(&pb, data, size, 0, NULL, NULL, NULL, NULL);
688
    return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
689
}
690

    
691
/*
692
 * Same as above, but signed.
693
 */
694
static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
695
                                 uint8_t *data, uint32_t size, int64_t *num)
696
{
697
    uint64_t unum;
698
    int res;
699

    
700
    /* read as unsigned number first */
701
    if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
702
        return res;
703

    
704
    /* make signed (weird way) */
705
    *num = unum - ((1LL << (7*res - 1)) - 1);
706

    
707
    return res;
708
}
709

    
710
static int ebml_parse_elem(MatroskaDemuxContext *matroska,
711
                           EbmlSyntax *syntax, void *data);
712

    
713
static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
714
                         uint32_t id, void *data)
715
{
716
    int i;
717
    for (i=0; syntax[i].id; i++)
718
        if (id == syntax[i].id)
719
            break;
720
    if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
721
        matroska->num_levels > 0 &&
722
        matroska->levels[matroska->num_levels-1].length == 0xffffffffffffff)
723
        return 0;  // we reached the end of an unknown size cluster
724
    if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32)
725
        av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
726
    return ebml_parse_elem(matroska, &syntax[i], data);
727
}
728

    
729
static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
730
                      void *data)
731
{
732
    if (!matroska->current_id) {
733
        uint64_t id;
734
        int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
735
        if (res < 0)
736
            return res;
737
        matroska->current_id = id | 1 << 7*res;
738
    }
739
    return ebml_parse_id(matroska, syntax, matroska->current_id, data);
740
}
741

    
742
static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
743
                           void *data)
744
{
745
    int i, res = 0;
746

    
747
    for (i=0; syntax[i].id; i++)
748
        switch (syntax[i].type) {
749
        case EBML_UINT:
750
            *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
751
            break;
752
        case EBML_FLOAT:
753
            *(double   *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
754
            break;
755
        case EBML_STR:
756
        case EBML_UTF8:
757
            *(char    **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
758
            break;
759
        }
760

    
761
    while (!res && !ebml_level_end(matroska))
762
        res = ebml_parse(matroska, syntax, data);
763

    
764
    return res;
765
}
766

    
767
static int ebml_parse_elem(MatroskaDemuxContext *matroska,
768
                           EbmlSyntax *syntax, void *data)
769
{
770
    ByteIOContext *pb = matroska->ctx->pb;
771
    uint32_t id = syntax->id;
772
    uint64_t length;
773
    int res;
774

    
775
    data = (char *)data + syntax->data_offset;
776
    if (syntax->list_elem_size) {
777
        EbmlList *list = data;
778
        list->elem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
779
        data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
780
        memset(data, 0, syntax->list_elem_size);
781
        list->nb_elem++;
782
    }
783

    
784
    if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
785
        matroska->current_id = 0;
786
        if ((res = ebml_read_num(matroska, pb, 8, &length)) < 0)
787
            return res;
788
    }
789

    
790
    switch (syntax->type) {
791
    case EBML_UINT:  res = ebml_read_uint  (pb, length, data);  break;
792
    case EBML_FLOAT: res = ebml_read_float (pb, length, data);  break;
793
    case EBML_STR:
794
    case EBML_UTF8:  res = ebml_read_ascii (pb, length, data);  break;
795
    case EBML_BIN:   res = ebml_read_binary(pb, length, data);  break;
796
    case EBML_NEST:  if ((res=ebml_read_master(matroska, length)) < 0)
797
                         return res;
798
                     if (id == MATROSKA_ID_SEGMENT)
799
                         matroska->segment_start = url_ftell(matroska->ctx->pb);
800
                     return ebml_parse_nest(matroska, syntax->def.n, data);
801
    case EBML_PASS:  return ebml_parse_id(matroska, syntax->def.n, id, data);
802
    case EBML_STOP:  return 1;
803
    default:         return url_fseek(pb,length,SEEK_CUR)<0 ? AVERROR(EIO) : 0;
804
    }
805
    if (res == AVERROR_INVALIDDATA)
806
        av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
807
    else if (res == AVERROR(EIO))
808
        av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
809
    return res;
810
}
811

    
812
static void ebml_free(EbmlSyntax *syntax, void *data)
813
{
814
    int i, j;
815
    for (i=0; syntax[i].id; i++) {
816
        void *data_off = (char *)data + syntax[i].data_offset;
817
        switch (syntax[i].type) {
818
        case EBML_STR:
819
        case EBML_UTF8:  av_freep(data_off);                      break;
820
        case EBML_BIN:   av_freep(&((EbmlBin *)data_off)->data);  break;
821
        case EBML_NEST:
822
            if (syntax[i].list_elem_size) {
823
                EbmlList *list = data_off;
824
                char *ptr = list->elem;
825
                for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
826
                    ebml_free(syntax[i].def.n, ptr);
827
                av_free(list->elem);
828
            } else
829
                ebml_free(syntax[i].def.n, data_off);
830
        default:  break;
831
        }
832
    }
833
}
834

    
835

    
836
/*
837
 * Autodetecting...
838
 */
839
static int matroska_probe(AVProbeData *p)
840
{
841
    uint64_t total = 0;
842
    int len_mask = 0x80, size = 1, n = 1, i;
843

    
844
    /* EBML header? */
845
    if (AV_RB32(p->buf) != EBML_ID_HEADER)
846
        return 0;
847

    
848
    /* length of header */
849
    total = p->buf[4];
850
    while (size <= 8 && !(total & len_mask)) {
851
        size++;
852
        len_mask >>= 1;
853
    }
854
    if (size > 8)
855
      return 0;
856
    total &= (len_mask - 1);
857
    while (n < size)
858
        total = (total << 8) | p->buf[4 + n++];
859

    
860
    /* Does the probe data contain the whole header? */
861
    if (p->buf_size < 4 + size + total)
862
      return 0;
863

    
864
    /* The header should contain a known document type. For now,
865
     * we don't parse the whole header but simply check for the
866
     * availability of that array of characters inside the header.
867
     * Not fully fool-proof, but good enough. */
868
    for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
869
        int probelen = strlen(matroska_doctypes[i]);
870
        for (n = 4+size; n <= 4+size+total-probelen; n++)
871
            if (!memcmp(p->buf+n, matroska_doctypes[i], probelen))
872
                return AVPROBE_SCORE_MAX;
873
    }
874

    
875
    // probably valid EBML header but no recognized doctype
876
    return AVPROBE_SCORE_MAX/2;
877
}
878

    
879
static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
880
                                                 int num)
881
{
882
    MatroskaTrack *tracks = matroska->tracks.elem;
883
    int i;
884

    
885
    for (i=0; i < matroska->tracks.nb_elem; i++)
886
        if (tracks[i].num == num)
887
            return &tracks[i];
888

    
889
    av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
890
    return NULL;
891
}
892

    
893
static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
894
                                  MatroskaTrack *track)
895
{
896
    MatroskaTrackEncoding *encodings = track->encodings.elem;
897
    uint8_t* data = *buf;
898
    int isize = *buf_size;
899
    uint8_t* pkt_data = NULL;
900
    int pkt_size = isize;
901
    int result = 0;
902
    int olen;
903

    
904
    switch (encodings[0].compression.algo) {
905
    case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
906
        return encodings[0].compression.settings.size;
907
    case MATROSKA_TRACK_ENCODING_COMP_LZO:
908
        do {
909
            olen = pkt_size *= 3;
910
            pkt_data = av_realloc(pkt_data, pkt_size+AV_LZO_OUTPUT_PADDING);
911
            result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
912
        } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
913
        if (result)
914
            goto failed;
915
        pkt_size -= olen;
916
        break;
917
#if CONFIG_ZLIB
918
    case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
919
        z_stream zstream = {0};
920
        if (inflateInit(&zstream) != Z_OK)
921
            return -1;
922
        zstream.next_in = data;
923
        zstream.avail_in = isize;
924
        do {
925
            pkt_size *= 3;
926
            pkt_data = av_realloc(pkt_data, pkt_size);
927
            zstream.avail_out = pkt_size - zstream.total_out;
928
            zstream.next_out = pkt_data + zstream.total_out;
929
            result = inflate(&zstream, Z_NO_FLUSH);
930
        } while (result==Z_OK && pkt_size<10000000);
931
        pkt_size = zstream.total_out;
932
        inflateEnd(&zstream);
933
        if (result != Z_STREAM_END)
934
            goto failed;
935
        break;
936
    }
937
#endif
938
#if CONFIG_BZLIB
939
    case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
940
        bz_stream bzstream = {0};
941
        if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
942
            return -1;
943
        bzstream.next_in = data;
944
        bzstream.avail_in = isize;
945
        do {
946
            pkt_size *= 3;
947
            pkt_data = av_realloc(pkt_data, pkt_size);
948
            bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
949
            bzstream.next_out = pkt_data + bzstream.total_out_lo32;
950
            result = BZ2_bzDecompress(&bzstream);
951
        } while (result==BZ_OK && pkt_size<10000000);
952
        pkt_size = bzstream.total_out_lo32;
953
        BZ2_bzDecompressEnd(&bzstream);
954
        if (result != BZ_STREAM_END)
955
            goto failed;
956
        break;
957
    }
958
#endif
959
    default:
960
        return -1;
961
    }
962

    
963
    *buf = pkt_data;
964
    *buf_size = pkt_size;
965
    return 0;
966
 failed:
967
    av_free(pkt_data);
968
    return -1;
969
}
970

    
971
static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
972
                                    AVPacket *pkt, uint64_t display_duration)
973
{
974
    char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
975
    for (; *ptr!=',' && ptr<end-1; ptr++);
976
    if (*ptr == ',')
977
        layer = ++ptr;
978
    for (; *ptr!=',' && ptr<end-1; ptr++);
979
    if (*ptr == ',') {
980
        int64_t end_pts = pkt->pts + display_duration;
981
        int sc = matroska->time_scale * pkt->pts / 10000000;
982
        int ec = matroska->time_scale * end_pts  / 10000000;
983
        int sh, sm, ss, eh, em, es, len;
984
        sh = sc/360000;  sc -= 360000*sh;
985
        sm = sc/  6000;  sc -=   6000*sm;
986
        ss = sc/   100;  sc -=    100*ss;
987
        eh = ec/360000;  ec -= 360000*eh;
988
        em = ec/  6000;  ec -=   6000*em;
989
        es = ec/   100;  ec -=    100*es;
990
        *ptr++ = '\0';
991
        len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
992
        if (!(line = av_malloc(len)))
993
            return;
994
        snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
995
                 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
996
        av_free(pkt->data);
997
        pkt->data = line;
998
        pkt->size = strlen(line);
999
    }
1000
}
1001

    
1002
static void matroska_merge_packets(AVPacket *out, AVPacket *in)
1003
{
1004
    out->data = av_realloc(out->data, out->size+in->size);
1005
    memcpy(out->data+out->size, in->data, in->size);
1006
    out->size += in->size;
1007
    av_destruct_packet(in);
1008
    av_free(in);
1009
}
1010

    
1011
static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1012
                                 AVMetadata **metadata, char *prefix)
1013
{
1014
    MatroskaTag *tags = list->elem;
1015
    char key[1024];
1016
    int i;
1017

    
1018
    for (i=0; i < list->nb_elem; i++) {
1019
        const char *lang = strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1020
        if (prefix)  snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1021
        else         av_strlcpy(key, tags[i].name, sizeof(key));
1022
        if (tags[i].def || !lang) {
1023
        av_metadata_set2(metadata, key, tags[i].string, 0);
1024
        if (tags[i].sub.nb_elem)
1025
            matroska_convert_tag(s, &tags[i].sub, metadata, key);
1026
        }
1027
        if (lang) {
1028
            av_strlcat(key, "-", sizeof(key));
1029
            av_strlcat(key, lang, sizeof(key));
1030
            av_metadata_set2(metadata, key, tags[i].string, 0);
1031
            if (tags[i].sub.nb_elem)
1032
                matroska_convert_tag(s, &tags[i].sub, metadata, key);
1033
        }
1034
    }
1035
}
1036

    
1037
static void matroska_convert_tags(AVFormatContext *s)
1038
{
1039
    MatroskaDemuxContext *matroska = s->priv_data;
1040
    MatroskaTags *tags = matroska->tags.elem;
1041
    int i, j;
1042

    
1043
    for (i=0; i < matroska->tags.nb_elem; i++) {
1044
        if (tags[i].target.attachuid) {
1045
            MatroskaAttachement *attachment = matroska->attachments.elem;
1046
            for (j=0; j<matroska->attachments.nb_elem; j++)
1047
                if (attachment[j].uid == tags[i].target.attachuid)
1048
                    matroska_convert_tag(s, &tags[i].tag,
1049
                                         &attachment[j].stream->metadata, NULL);
1050
        } else if (tags[i].target.chapteruid) {
1051
            MatroskaChapter *chapter = matroska->chapters.elem;
1052
            for (j=0; j<matroska->chapters.nb_elem; j++)
1053
                if (chapter[j].uid == tags[i].target.chapteruid)
1054
                    matroska_convert_tag(s, &tags[i].tag,
1055
                                         &chapter[j].chapter->metadata, NULL);
1056
        } else if (tags[i].target.trackuid) {
1057
            MatroskaTrack *track = matroska->tracks.elem;
1058
            for (j=0; j<matroska->tracks.nb_elem; j++)
1059
                if (track[j].uid == tags[i].target.trackuid)
1060
                    matroska_convert_tag(s, &tags[i].tag,
1061
                                         &track[j].stream->metadata, NULL);
1062
        } else {
1063
            matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1064
                                 tags[i].target.type);
1065
        }
1066
    }
1067
}
1068

    
1069
static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1070
{
1071
    EbmlList *seekhead_list = &matroska->seekhead;
1072
    MatroskaSeekhead *seekhead = seekhead_list->elem;
1073
    uint32_t level_up = matroska->level_up;
1074
    int64_t before_pos = url_ftell(matroska->ctx->pb);
1075
    uint32_t saved_id = matroska->current_id;
1076
    MatroskaLevel level;
1077
    int i;
1078

    
1079
    // we should not do any seeking in the streaming case
1080
    if (url_is_streamed(matroska->ctx->pb) ||
1081
        (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1082
        return;
1083

    
1084
    for (i=0; i<seekhead_list->nb_elem; i++) {
1085
        int64_t offset = seekhead[i].pos + matroska->segment_start;
1086

    
1087
        if (seekhead[i].pos <= before_pos
1088
            || seekhead[i].id == MATROSKA_ID_SEEKHEAD
1089
            || seekhead[i].id == MATROSKA_ID_CLUSTER)
1090
            continue;
1091

    
1092
        /* seek */
1093
        if (url_fseek(matroska->ctx->pb, offset, SEEK_SET) != offset)
1094
            continue;
1095

    
1096
        /* We don't want to lose our seekhead level, so we add
1097
         * a dummy. This is a crude hack. */
1098
        if (matroska->num_levels == EBML_MAX_DEPTH) {
1099
            av_log(matroska->ctx, AV_LOG_INFO,
1100
                   "Max EBML element depth (%d) reached, "
1101
                   "cannot parse further.\n", EBML_MAX_DEPTH);
1102
            break;
1103
        }
1104

    
1105
        level.start = 0;
1106
        level.length = (uint64_t)-1;
1107
        matroska->levels[matroska->num_levels] = level;
1108
        matroska->num_levels++;
1109
        matroska->current_id = 0;
1110

    
1111
        ebml_parse(matroska, matroska_segment, matroska);
1112

    
1113
        /* remove dummy level */
1114
        while (matroska->num_levels) {
1115
            uint64_t length = matroska->levels[--matroska->num_levels].length;
1116
            if (length == (uint64_t)-1)
1117
                break;
1118
        }
1119
    }
1120

    
1121
    /* seek back */
1122
    url_fseek(matroska->ctx->pb, before_pos, SEEK_SET);
1123
    matroska->level_up = level_up;
1124
    matroska->current_id = saved_id;
1125
}
1126

    
1127
static int matroska_aac_profile(char *codec_id)
1128
{
1129
    static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
1130
    int profile;
1131

    
1132
    for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
1133
        if (strstr(codec_id, aac_profiles[profile]))
1134
            break;
1135
    return profile + 1;
1136
}
1137

    
1138
static int matroska_aac_sri(int samplerate)
1139
{
1140
    int sri;
1141

    
1142
    for (sri=0; sri<FF_ARRAY_ELEMS(ff_mpeg4audio_sample_rates); sri++)
1143
        if (ff_mpeg4audio_sample_rates[sri] == samplerate)
1144
            break;
1145
    return sri;
1146
}
1147

    
1148
static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
1149
{
1150
    MatroskaDemuxContext *matroska = s->priv_data;
1151
    EbmlList *attachements_list = &matroska->attachments;
1152
    MatroskaAttachement *attachements;
1153
    EbmlList *chapters_list = &matroska->chapters;
1154
    MatroskaChapter *chapters;
1155
    MatroskaTrack *tracks;
1156
    EbmlList *index_list;
1157
    MatroskaIndex *index;
1158
    int index_scale = 1;
1159
    uint64_t max_start = 0;
1160
    Ebml ebml = { 0 };
1161
    AVStream *st;
1162
    int i, j, res;
1163

    
1164
    matroska->ctx = s;
1165

    
1166
    /* First read the EBML header. */
1167
    if (ebml_parse(matroska, ebml_syntax, &ebml)
1168
        || ebml.version > EBML_VERSION       || ebml.max_size > sizeof(uint64_t)
1169
        || ebml.id_length > sizeof(uint32_t) || ebml.doctype_version > 2) {
1170
        av_log(matroska->ctx, AV_LOG_ERROR,
1171
               "EBML header using unsupported features\n"
1172
               "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1173
               ebml.version, ebml.doctype, ebml.doctype_version);
1174
        ebml_free(ebml_syntax, &ebml);
1175
        return AVERROR_PATCHWELCOME;
1176
    }
1177
    for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
1178
        if (!strcmp(ebml.doctype, matroska_doctypes[i]))
1179
            break;
1180
    if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
1181
        av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
1182
    }
1183
    ebml_free(ebml_syntax, &ebml);
1184

    
1185
    /* The next thing is a segment. */
1186
    if ((res = ebml_parse(matroska, matroska_segments, matroska)) < 0)
1187
        return res;
1188
    matroska_execute_seekhead(matroska);
1189

    
1190
    if (!matroska->time_scale)
1191
        matroska->time_scale = 1000000;
1192
    if (matroska->duration)
1193
        matroska->ctx->duration = matroska->duration * matroska->time_scale
1194
                                  * 1000 / AV_TIME_BASE;
1195
    av_metadata_set2(&s->metadata, "title", matroska->title, 0);
1196

    
1197
    tracks = matroska->tracks.elem;
1198
    for (i=0; i < matroska->tracks.nb_elem; i++) {
1199
        MatroskaTrack *track = &tracks[i];
1200
        enum CodecID codec_id = CODEC_ID_NONE;
1201
        EbmlList *encodings_list = &tracks->encodings;
1202
        MatroskaTrackEncoding *encodings = encodings_list->elem;
1203
        uint8_t *extradata = NULL;
1204
        int extradata_size = 0;
1205
        int extradata_offset = 0;
1206
        ByteIOContext b;
1207

    
1208
        /* Apply some sanity checks. */
1209
        if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1210
            track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1211
            track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1212
            av_log(matroska->ctx, AV_LOG_INFO,
1213
                   "Unknown or unsupported track type %"PRIu64"\n",
1214
                   track->type);
1215
            continue;
1216
        }
1217
        if (track->codec_id == NULL)
1218
            continue;
1219

    
1220
        if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1221
            if (!track->default_duration)
1222
                track->default_duration = 1000000000/track->video.frame_rate;
1223
            if (!track->video.display_width)
1224
                track->video.display_width = track->video.pixel_width;
1225
            if (!track->video.display_height)
1226
                track->video.display_height = track->video.pixel_height;
1227
        } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1228
            if (!track->audio.out_samplerate)
1229
                track->audio.out_samplerate = track->audio.samplerate;
1230
        }
1231
        if (encodings_list->nb_elem > 1) {
1232
            av_log(matroska->ctx, AV_LOG_ERROR,
1233
                   "Multiple combined encodings no supported");
1234
        } else if (encodings_list->nb_elem == 1) {
1235
            if (encodings[0].type ||
1236
                (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1237
#if CONFIG_ZLIB
1238
                 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1239
#endif
1240
#if CONFIG_BZLIB
1241
                 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1242
#endif
1243
                 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
1244
                encodings[0].scope = 0;
1245
                av_log(matroska->ctx, AV_LOG_ERROR,
1246
                       "Unsupported encoding type");
1247
            } else if (track->codec_priv.size && encodings[0].scope&2) {
1248
                uint8_t *codec_priv = track->codec_priv.data;
1249
                int offset = matroska_decode_buffer(&track->codec_priv.data,
1250
                                                    &track->codec_priv.size,
1251
                                                    track);
1252
                if (offset < 0) {
1253
                    track->codec_priv.data = NULL;
1254
                    track->codec_priv.size = 0;
1255
                    av_log(matroska->ctx, AV_LOG_ERROR,
1256
                           "Failed to decode codec private data\n");
1257
                } else if (offset > 0) {
1258
                    track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
1259
                    memcpy(track->codec_priv.data,
1260
                           encodings[0].compression.settings.data, offset);
1261
                    memcpy(track->codec_priv.data+offset, codec_priv,
1262
                           track->codec_priv.size);
1263
                    track->codec_priv.size += offset;
1264
                }
1265
                if (codec_priv != track->codec_priv.data)
1266
                    av_free(codec_priv);
1267
            }
1268
        }
1269

    
1270
        for(j=0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++){
1271
            if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1272
                        strlen(ff_mkv_codec_tags[j].str))){
1273
                codec_id= ff_mkv_codec_tags[j].id;
1274
                break;
1275
            }
1276
        }
1277

    
1278
        st = track->stream = av_new_stream(s, 0);
1279
        if (st == NULL)
1280
            return AVERROR(ENOMEM);
1281

    
1282
        if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1283
            && track->codec_priv.size >= 40
1284
            && track->codec_priv.data != NULL) {
1285
            track->ms_compat = 1;
1286
            track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1287
            codec_id = ff_codec_get_id(ff_codec_bmp_tags, track->video.fourcc);
1288
            extradata_offset = 40;
1289
        } else if (!strcmp(track->codec_id, "A_MS/ACM")
1290
                   && track->codec_priv.size >= 14
1291
                   && track->codec_priv.data != NULL) {
1292
            init_put_byte(&b, track->codec_priv.data, track->codec_priv.size,
1293
                          URL_RDONLY, NULL, NULL, NULL, NULL);
1294
            ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1295
            codec_id = st->codec->codec_id;
1296
            extradata_offset = FFMIN(track->codec_priv.size, 18);
1297
        } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1298
                   && (track->codec_priv.size >= 86)
1299
                   && (track->codec_priv.data != NULL)) {
1300
            track->video.fourcc = AV_RL32(track->codec_priv.data);
1301
            codec_id=ff_codec_get_id(codec_movvideo_tags, track->video.fourcc);
1302
        } else if (codec_id == CODEC_ID_PCM_S16BE) {
1303
            switch (track->audio.bitdepth) {
1304
            case  8:  codec_id = CODEC_ID_PCM_U8;     break;
1305
            case 24:  codec_id = CODEC_ID_PCM_S24BE;  break;
1306
            case 32:  codec_id = CODEC_ID_PCM_S32BE;  break;
1307
            }
1308
        } else if (codec_id == CODEC_ID_PCM_S16LE) {
1309
            switch (track->audio.bitdepth) {
1310
            case  8:  codec_id = CODEC_ID_PCM_U8;     break;
1311
            case 24:  codec_id = CODEC_ID_PCM_S24LE;  break;
1312
            case 32:  codec_id = CODEC_ID_PCM_S32LE;  break;
1313
            }
1314
        } else if (codec_id==CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1315
            codec_id = CODEC_ID_PCM_F64LE;
1316
        } else if (codec_id == CODEC_ID_AAC && !track->codec_priv.size) {
1317
            int profile = matroska_aac_profile(track->codec_id);
1318
            int sri = matroska_aac_sri(track->audio.samplerate);
1319
            extradata = av_malloc(5);
1320
            if (extradata == NULL)
1321
                return AVERROR(ENOMEM);
1322
            extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1323
            extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1324
            if (strstr(track->codec_id, "SBR")) {
1325
                sri = matroska_aac_sri(track->audio.out_samplerate);
1326
                extradata[2] = 0x56;
1327
                extradata[3] = 0xE5;
1328
                extradata[4] = 0x80 | (sri<<3);
1329
                extradata_size = 5;
1330
            } else
1331
                extradata_size = 2;
1332
        } else if (codec_id == CODEC_ID_TTA) {
1333
            extradata_size = 30;
1334
            extradata = av_mallocz(extradata_size);
1335
            if (extradata == NULL)
1336
                return AVERROR(ENOMEM);
1337
            init_put_byte(&b, extradata, extradata_size, 1,
1338
                          NULL, NULL, NULL, NULL);
1339
            put_buffer(&b, "TTA1", 4);
1340
            put_le16(&b, 1);
1341
            put_le16(&b, track->audio.channels);
1342
            put_le16(&b, track->audio.bitdepth);
1343
            put_le32(&b, track->audio.out_samplerate);
1344
            put_le32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1345
        } else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
1346
                   codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
1347
            extradata_offset = 26;
1348
        } else if (codec_id == CODEC_ID_RA_144) {
1349
            track->audio.out_samplerate = 8000;
1350
            track->audio.channels = 1;
1351
        } else if (codec_id == CODEC_ID_RA_288 || codec_id == CODEC_ID_COOK ||
1352
                   codec_id == CODEC_ID_ATRAC3 || codec_id == CODEC_ID_SIPR) {
1353
            int flavor;
1354
            init_put_byte(&b, track->codec_priv.data,track->codec_priv.size,
1355
                          0, NULL, NULL, NULL, NULL);
1356
            url_fskip(&b, 22);
1357
            flavor                       = get_be16(&b);
1358
            track->audio.coded_framesize = get_be32(&b);
1359
            url_fskip(&b, 12);
1360
            track->audio.sub_packet_h    = get_be16(&b);
1361
            track->audio.frame_size      = get_be16(&b);
1362
            track->audio.sub_packet_size = get_be16(&b);
1363
            track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1364
            if (codec_id == CODEC_ID_RA_288) {
1365
                st->codec->block_align = track->audio.coded_framesize;
1366
                track->codec_priv.size = 0;
1367
            } else {
1368
                if (codec_id == CODEC_ID_SIPR && flavor < 4) {
1369
                    const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1370
                    track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1371
                    st->codec->bit_rate = sipr_bit_rate[flavor];
1372
                }
1373
                st->codec->block_align = track->audio.sub_packet_size;
1374
                extradata_offset = 78;
1375
            }
1376
        }
1377
        track->codec_priv.size -= extradata_offset;
1378

    
1379
        if (codec_id == CODEC_ID_NONE)
1380
            av_log(matroska->ctx, AV_LOG_INFO,
1381
                   "Unknown/unsupported CodecID %s.\n", track->codec_id);
1382

    
1383
        if (track->time_scale < 0.01)
1384
            track->time_scale = 1.0;
1385
        av_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1386

    
1387
        st->codec->codec_id = codec_id;
1388
        st->start_time = 0;
1389
        if (strcmp(track->language, "und"))
1390
            av_metadata_set2(&st->metadata, "language", track->language, 0);
1391
        av_metadata_set2(&st->metadata, "title", track->name, 0);
1392

    
1393
        if (track->flag_default)
1394
            st->disposition |= AV_DISPOSITION_DEFAULT;
1395
        if (track->flag_forced)
1396
            st->disposition |= AV_DISPOSITION_FORCED;
1397

    
1398
        if (track->default_duration)
1399
            av_reduce(&st->codec->time_base.num, &st->codec->time_base.den,
1400
                      track->default_duration, 1000000000, 30000);
1401

    
1402
        if (!st->codec->extradata) {
1403
            if(extradata){
1404
                st->codec->extradata = extradata;
1405
                st->codec->extradata_size = extradata_size;
1406
            } else if(track->codec_priv.data && track->codec_priv.size > 0){
1407
                st->codec->extradata = av_mallocz(track->codec_priv.size +
1408
                                                  FF_INPUT_BUFFER_PADDING_SIZE);
1409
                if(st->codec->extradata == NULL)
1410
                    return AVERROR(ENOMEM);
1411
                st->codec->extradata_size = track->codec_priv.size;
1412
                memcpy(st->codec->extradata,
1413
                       track->codec_priv.data + extradata_offset,
1414
                       track->codec_priv.size);
1415
            }
1416
        }
1417

    
1418
        if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1419
            st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1420
            st->codec->codec_tag  = track->video.fourcc;
1421
            st->codec->width  = track->video.pixel_width;
1422
            st->codec->height = track->video.pixel_height;
1423
            av_reduce(&st->sample_aspect_ratio.num,
1424
                      &st->sample_aspect_ratio.den,
1425
                      st->codec->height * track->video.display_width,
1426
                      st->codec-> width * track->video.display_height,
1427
                      255);
1428
            if (st->codec->codec_id != CODEC_ID_H264)
1429
            st->need_parsing = AVSTREAM_PARSE_HEADERS;
1430
            if (track->default_duration)
1431
                st->avg_frame_rate = av_d2q(1000000000.0/track->default_duration, INT_MAX);
1432
        } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1433
            st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1434
            st->codec->sample_rate = track->audio.out_samplerate;
1435
            st->codec->channels = track->audio.channels;
1436
            if (st->codec->codec_id != CODEC_ID_AAC)
1437
            st->need_parsing = AVSTREAM_PARSE_HEADERS;
1438
        } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1439
            st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1440
        }
1441
    }
1442

    
1443
    attachements = attachements_list->elem;
1444
    for (j=0; j<attachements_list->nb_elem; j++) {
1445
        if (!(attachements[j].filename && attachements[j].mime &&
1446
              attachements[j].bin.data && attachements[j].bin.size > 0)) {
1447
            av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1448
        } else {
1449
            AVStream *st = av_new_stream(s, 0);
1450
            if (st == NULL)
1451
                break;
1452
            av_metadata_set2(&st->metadata, "filename",attachements[j].filename, 0);
1453
            st->codec->codec_id = CODEC_ID_NONE;
1454
            st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
1455
            st->codec->extradata  = av_malloc(attachements[j].bin.size);
1456
            if(st->codec->extradata == NULL)
1457
                break;
1458
            st->codec->extradata_size = attachements[j].bin.size;
1459
            memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1460

    
1461
            for (i=0; ff_mkv_mime_tags[i].id != CODEC_ID_NONE; i++) {
1462
                if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1463
                             strlen(ff_mkv_mime_tags[i].str))) {
1464
                    st->codec->codec_id = ff_mkv_mime_tags[i].id;
1465
                    break;
1466
                }
1467
            }
1468
            attachements[j].stream = st;
1469
        }
1470
    }
1471

    
1472
    chapters = chapters_list->elem;
1473
    for (i=0; i<chapters_list->nb_elem; i++)
1474
        if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
1475
            && (max_start==0 || chapters[i].start > max_start)) {
1476
            chapters[i].chapter =
1477
            ff_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1478
                           chapters[i].start, chapters[i].end,
1479
                           chapters[i].title);
1480
            av_metadata_set2(&chapters[i].chapter->metadata,
1481
                             "title", chapters[i].title, 0);
1482
            max_start = chapters[i].start;
1483
        }
1484

    
1485
    index_list = &matroska->index;
1486
    index = index_list->elem;
1487
    if (index_list->nb_elem
1488
        && index[0].time > 100000000000000/matroska->time_scale) {
1489
        av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1490
        index_scale = matroska->time_scale;
1491
    }
1492
    for (i=0; i<index_list->nb_elem; i++) {
1493
        EbmlList *pos_list = &index[i].pos;
1494
        MatroskaIndexPos *pos = pos_list->elem;
1495
        for (j=0; j<pos_list->nb_elem; j++) {
1496
            MatroskaTrack *track = matroska_find_track_by_num(matroska,
1497
                                                              pos[j].track);
1498
            if (track && track->stream)
1499
                av_add_index_entry(track->stream,
1500
                                   pos[j].pos + matroska->segment_start,
1501
                                   index[i].time/index_scale, 0, 0,
1502
                                   AVINDEX_KEYFRAME);
1503
        }
1504
    }
1505

    
1506
    matroska_convert_tags(s);
1507

    
1508
    return 0;
1509
}
1510

    
1511
/*
1512
 * Put one packet in an application-supplied AVPacket struct.
1513
 * Returns 0 on success or -1 on failure.
1514
 */
1515
static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1516
                                   AVPacket *pkt)
1517
{
1518
    if (matroska->num_packets > 0) {
1519
        memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1520
        av_free(matroska->packets[0]);
1521
        if (matroska->num_packets > 1) {
1522
            memmove(&matroska->packets[0], &matroska->packets[1],
1523
                    (matroska->num_packets - 1) * sizeof(AVPacket *));
1524
            matroska->packets =
1525
                av_realloc(matroska->packets, (matroska->num_packets - 1) *
1526
                           sizeof(AVPacket *));
1527
        } else {
1528
            av_freep(&matroska->packets);
1529
        }
1530
        matroska->num_packets--;
1531
        return 0;
1532
    }
1533

    
1534
    return -1;
1535
}
1536

    
1537
/*
1538
 * Free all packets in our internal queue.
1539
 */
1540
static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1541
{
1542
    if (matroska->packets) {
1543
        int n;
1544
        for (n = 0; n < matroska->num_packets; n++) {
1545
            av_free_packet(matroska->packets[n]);
1546
            av_free(matroska->packets[n]);
1547
        }
1548
        av_freep(&matroska->packets);
1549
        matroska->num_packets = 0;
1550
    }
1551
}
1552

    
1553
static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
1554
                                int size, int64_t pos, uint64_t cluster_time,
1555
                                uint64_t duration, int is_keyframe,
1556
                                int64_t cluster_pos)
1557
{
1558
    uint64_t timecode = AV_NOPTS_VALUE;
1559
    MatroskaTrack *track;
1560
    int res = 0;
1561
    AVStream *st;
1562
    AVPacket *pkt;
1563
    int16_t block_time;
1564
    uint32_t *lace_size = NULL;
1565
    int n, flags, laces = 0;
1566
    uint64_t num;
1567

    
1568
    if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
1569
        av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
1570
        return res;
1571
    }
1572
    data += n;
1573
    size -= n;
1574

    
1575
    track = matroska_find_track_by_num(matroska, num);
1576
    if (size <= 3 || !track || !track->stream) {
1577
        av_log(matroska->ctx, AV_LOG_INFO,
1578
               "Invalid stream %"PRIu64" or size %u\n", num, size);
1579
        return res;
1580
    }
1581
    st = track->stream;
1582
    if (st->discard >= AVDISCARD_ALL)
1583
        return res;
1584
    if (duration == AV_NOPTS_VALUE)
1585
        duration = track->default_duration / matroska->time_scale;
1586

    
1587
    block_time = AV_RB16(data);
1588
    data += 2;
1589
    flags = *data++;
1590
    size -= 3;
1591
    if (is_keyframe == -1)
1592
        is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
1593

    
1594
    if (cluster_time != (uint64_t)-1
1595
        && (block_time >= 0 || cluster_time >= -block_time)) {
1596
        timecode = cluster_time + block_time;
1597
        if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
1598
            && timecode < track->end_timecode)
1599
            is_keyframe = 0;  /* overlapping subtitles are not key frame */
1600
        if (is_keyframe)
1601
            av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
1602
        track->end_timecode = FFMAX(track->end_timecode, timecode+duration);
1603
    }
1604

    
1605
    if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1606
        if (!is_keyframe || timecode < matroska->skip_to_timecode)
1607
            return res;
1608
        matroska->skip_to_keyframe = 0;
1609
    }
1610

    
1611
    switch ((flags & 0x06) >> 1) {
1612
        case 0x0: /* no lacing */
1613
            laces = 1;
1614
            lace_size = av_mallocz(sizeof(int));
1615
            lace_size[0] = size;
1616
            break;
1617

    
1618
        case 0x1: /* Xiph lacing */
1619
        case 0x2: /* fixed-size lacing */
1620
        case 0x3: /* EBML lacing */
1621
            assert(size>0); // size <=3 is checked before size-=3 above
1622
            laces = (*data) + 1;
1623
            data += 1;
1624
            size -= 1;
1625
            lace_size = av_mallocz(laces * sizeof(int));
1626

    
1627
            switch ((flags & 0x06) >> 1) {
1628
                case 0x1: /* Xiph lacing */ {
1629
                    uint8_t temp;
1630
                    uint32_t total = 0;
1631
                    for (n = 0; res == 0 && n < laces - 1; n++) {
1632
                        while (1) {
1633
                            if (size == 0) {
1634
                                res = -1;
1635
                                break;
1636
                            }
1637
                            temp = *data;
1638
                            lace_size[n] += temp;
1639
                            data += 1;
1640
                            size -= 1;
1641
                            if (temp != 0xff)
1642
                                break;
1643
                        }
1644
                        total += lace_size[n];
1645
                    }
1646
                    lace_size[n] = size - total;
1647
                    break;
1648
                }
1649

    
1650
                case 0x2: /* fixed-size lacing */
1651
                    for (n = 0; n < laces; n++)
1652
                        lace_size[n] = size / laces;
1653
                    break;
1654

    
1655
                case 0x3: /* EBML lacing */ {
1656
                    uint32_t total;
1657
                    n = matroska_ebmlnum_uint(matroska, data, size, &num);
1658
                    if (n < 0) {
1659
                        av_log(matroska->ctx, AV_LOG_INFO,
1660
                               "EBML block data error\n");
1661
                        break;
1662
                    }
1663
                    data += n;
1664
                    size -= n;
1665
                    total = lace_size[0] = num;
1666
                    for (n = 1; res == 0 && n < laces - 1; n++) {
1667
                        int64_t snum;
1668
                        int r;
1669
                        r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1670
                        if (r < 0) {
1671
                            av_log(matroska->ctx, AV_LOG_INFO,
1672
                                   "EBML block data error\n");
1673
                            break;
1674
                        }
1675
                        data += r;
1676
                        size -= r;
1677
                        lace_size[n] = lace_size[n - 1] + snum;
1678
                        total += lace_size[n];
1679
                    }
1680
                    lace_size[n] = size - total;
1681
                    break;
1682
                }
1683
            }
1684
            break;
1685
    }
1686

    
1687
    if (res == 0) {
1688
        for (n = 0; n < laces; n++) {
1689
            if ((st->codec->codec_id == CODEC_ID_RA_288 ||
1690
                 st->codec->codec_id == CODEC_ID_COOK ||
1691
                 st->codec->codec_id == CODEC_ID_SIPR ||
1692
                 st->codec->codec_id == CODEC_ID_ATRAC3) &&
1693
                 st->codec->block_align && track->audio.sub_packet_size) {
1694
                int a = st->codec->block_align;
1695
                int sps = track->audio.sub_packet_size;
1696
                int cfs = track->audio.coded_framesize;
1697
                int h = track->audio.sub_packet_h;
1698
                int y = track->audio.sub_packet_cnt;
1699
                int w = track->audio.frame_size;
1700
                int x;
1701

    
1702
                if (!track->audio.pkt_cnt) {
1703
                    if (st->codec->codec_id == CODEC_ID_RA_288)
1704
                        for (x=0; x<h/2; x++)
1705
                            memcpy(track->audio.buf+x*2*w+y*cfs,
1706
                                   data+x*cfs, cfs);
1707
                    else if (st->codec->codec_id == CODEC_ID_SIPR)
1708
                        memcpy(track->audio.buf + y*w, data, w);
1709
                    else
1710
                        for (x=0; x<w/sps; x++)
1711
                            memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
1712

    
1713
                    if (++track->audio.sub_packet_cnt >= h) {
1714
                        if (st->codec->codec_id == CODEC_ID_SIPR)
1715
                            ff_rm_reorder_sipr_data(track->audio.buf, h, w);
1716
                        track->audio.sub_packet_cnt = 0;
1717
                        track->audio.pkt_cnt = h*w / a;
1718
                    }
1719
                }
1720
                while (track->audio.pkt_cnt) {
1721
                    pkt = av_mallocz(sizeof(AVPacket));
1722
                    av_new_packet(pkt, a);
1723
                    memcpy(pkt->data, track->audio.buf
1724
                           + a * (h*w / a - track->audio.pkt_cnt--), a);
1725
                    pkt->pos = pos;
1726
                    pkt->stream_index = st->index;
1727
                    dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1728
                }
1729
            } else {
1730
                MatroskaTrackEncoding *encodings = track->encodings.elem;
1731
                int offset = 0, pkt_size = lace_size[n];
1732
                uint8_t *pkt_data = data;
1733

    
1734
                if (lace_size[n] > size) {
1735
                    av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
1736
                    break;
1737
                }
1738

    
1739
                if (encodings && encodings->scope & 1) {
1740
                    offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
1741
                    if (offset < 0)
1742
                        continue;
1743
                }
1744

    
1745
                pkt = av_mallocz(sizeof(AVPacket));
1746
                /* XXX: prevent data copy... */
1747
                if (av_new_packet(pkt, pkt_size+offset) < 0) {
1748
                    av_free(pkt);
1749
                    res = AVERROR(ENOMEM);
1750
                    break;
1751
                }
1752
                if (offset)
1753
                    memcpy (pkt->data, encodings->compression.settings.data, offset);
1754
                memcpy (pkt->data+offset, pkt_data, pkt_size);
1755

    
1756
                if (pkt_data != data)
1757
                    av_free(pkt_data);
1758

    
1759
                if (n == 0)
1760
                    pkt->flags = is_keyframe;
1761
                pkt->stream_index = st->index;
1762

    
1763
                if (track->ms_compat)
1764
                    pkt->dts = timecode;
1765
                else
1766
                    pkt->pts = timecode;
1767
                pkt->pos = pos;
1768
                if (st->codec->codec_id == CODEC_ID_TEXT)
1769
                    pkt->convergence_duration = duration;
1770
                else if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE)
1771
                    pkt->duration = duration;
1772

    
1773
                if (st->codec->codec_id == CODEC_ID_SSA)
1774
                    matroska_fix_ass_packet(matroska, pkt, duration);
1775

    
1776
                if (matroska->prev_pkt &&
1777
                    timecode != AV_NOPTS_VALUE &&
1778
                    matroska->prev_pkt->pts == timecode &&
1779
                    matroska->prev_pkt->stream_index == st->index &&
1780
                    st->codec->codec_id == CODEC_ID_SSA)
1781
                    matroska_merge_packets(matroska->prev_pkt, pkt);
1782
                else {
1783
                    dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1784
                    matroska->prev_pkt = pkt;
1785
                }
1786
            }
1787

    
1788
            if (timecode != AV_NOPTS_VALUE)
1789
                timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
1790
            data += lace_size[n];
1791
            size -= lace_size[n];
1792
        }
1793
    }
1794

    
1795
    av_free(lace_size);
1796
    return res;
1797
}
1798

    
1799
static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
1800
{
1801
    MatroskaCluster cluster = { 0 };
1802
    EbmlList *blocks_list;
1803
    MatroskaBlock *blocks;
1804
    int i, res;
1805
    int64_t pos = url_ftell(matroska->ctx->pb);
1806
    matroska->prev_pkt = NULL;
1807
    if (matroska->current_id)
1808
        pos -= 4;  /* sizeof the ID which was already read */
1809
    res = ebml_parse(matroska, matroska_clusters, &cluster);
1810
    blocks_list = &cluster.blocks;
1811
    blocks = blocks_list->elem;
1812
    for (i=0; i<blocks_list->nb_elem; i++)
1813
        if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
1814
            int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
1815
            res=matroska_parse_block(matroska,
1816
                                     blocks[i].bin.data, blocks[i].bin.size,
1817
                                     blocks[i].bin.pos,  cluster.timecode,
1818
                                     blocks[i].duration, is_keyframe,
1819
                                     pos);
1820
        }
1821
    ebml_free(matroska_cluster, &cluster);
1822
    if (res < 0)  matroska->done = 1;
1823
    return res;
1824
}
1825

    
1826
static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
1827
{
1828
    MatroskaDemuxContext *matroska = s->priv_data;
1829

    
1830
    while (matroska_deliver_packet(matroska, pkt)) {
1831
        if (matroska->done)
1832
            return AVERROR_EOF;
1833
        matroska_parse_cluster(matroska);
1834
    }
1835

    
1836
    return 0;
1837
}
1838

    
1839
static int matroska_read_seek(AVFormatContext *s, int stream_index,
1840
                              int64_t timestamp, int flags)
1841
{
1842
    MatroskaDemuxContext *matroska = s->priv_data;
1843
    MatroskaTrack *tracks = matroska->tracks.elem;
1844
    AVStream *st = s->streams[stream_index];
1845
    int i, index, index_sub, index_min;
1846

    
1847
    if (!st->nb_index_entries)
1848
        return 0;
1849
    timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
1850

    
1851
    if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1852
        url_fseek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
1853
        while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
1854
            matroska_clear_queue(matroska);
1855
            if (matroska_parse_cluster(matroska) < 0)
1856
                break;
1857
        }
1858
    }
1859

    
1860
    matroska_clear_queue(matroska);
1861
    if (index < 0)
1862
        return 0;
1863

    
1864
    index_min = index;
1865
    for (i=0; i < matroska->tracks.nb_elem; i++) {
1866
        tracks[i].end_timecode = 0;
1867
        if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
1868
            && !tracks[i].stream->discard != AVDISCARD_ALL) {
1869
            index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
1870
            if (index_sub >= 0
1871
                && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
1872
                && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
1873
                index_min = index_sub;
1874
        }
1875
    }
1876

    
1877
    url_fseek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
1878
    matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
1879
    matroska->skip_to_timecode = st->index_entries[index].timestamp;
1880
    matroska->done = 0;
1881
    av_update_cur_dts(s, st, st->index_entries[index].timestamp);
1882
    return 0;
1883
}
1884

    
1885
static int matroska_read_close(AVFormatContext *s)
1886
{
1887
    MatroskaDemuxContext *matroska = s->priv_data;
1888
    MatroskaTrack *tracks = matroska->tracks.elem;
1889
    int n;
1890

    
1891
    matroska_clear_queue(matroska);
1892

    
1893
    for (n=0; n < matroska->tracks.nb_elem; n++)
1894
        if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
1895
            av_free(tracks[n].audio.buf);
1896
    ebml_free(matroska_segment, matroska);
1897

    
1898
    return 0;
1899
}
1900

    
1901
AVInputFormat matroska_demuxer = {
1902
    "matroska,webm",
1903
    NULL_IF_CONFIG_SMALL("Matroska/WebM file format"),
1904
    sizeof(MatroskaDemuxContext),
1905
    matroska_probe,
1906
    matroska_read_header,
1907
    matroska_read_packet,
1908
    matroska_read_close,
1909
    matroska_read_seek,
1910
    .metadata_conv = ff_mkv_metadata_conv,
1911
};