check if current block contains a B frame and gives this info to parse_block()
[ffmpeg.git] / libavformat / matroska.c
1 /*
2  * Matroska file demuxer (no muxer yet)
3  * Copyright (c) 2003-2004 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 matroska.c
24  * Matroska file demuxer
25  * by Ronald Bultje <rbultje@ronald.bitfreak.net>
26  * with a little help from Moritz Bunkus <moritz@bunkus.org>
27  * Specs available on the matroska project page:
28  * http://www.matroska.org/.
29  */
30
31 #include "avformat.h"
32 /* For codec_get_id(). */
33 #include "riff.h"
34 #include "intfloat_readwrite.h"
35
36 /* EBML version supported */
37 #define EBML_VERSION 1
38
39 /* top-level master-IDs */
40 #define EBML_ID_HEADER             0x1A45DFA3
41
42 /* IDs in the HEADER master */
43 #define EBML_ID_EBMLVERSION        0x4286
44 #define EBML_ID_EBMLREADVERSION    0x42F7
45 #define EBML_ID_EBMLMAXIDLENGTH    0x42F2
46 #define EBML_ID_EBMLMAXSIZELENGTH  0x42F3
47 #define EBML_ID_DOCTYPE            0x4282
48 #define EBML_ID_DOCTYPEVERSION     0x4287
49 #define EBML_ID_DOCTYPEREADVERSION 0x4285
50
51 /* general EBML types */
52 #define EBML_ID_VOID               0xEC
53
54 /*
55  * Matroska element IDs. max. 32-bit.
56  */
57
58 /* toplevel segment */
59 #define MATROSKA_ID_SEGMENT    0x18538067
60
61 /* matroska top-level master IDs */
62 #define MATROSKA_ID_INFO       0x1549A966
63 #define MATROSKA_ID_TRACKS     0x1654AE6B
64 #define MATROSKA_ID_CUES       0x1C53BB6B
65 #define MATROSKA_ID_TAGS       0x1254C367
66 #define MATROSKA_ID_SEEKHEAD   0x114D9B74
67 #define MATROSKA_ID_CLUSTER    0x1F43B675
68
69 /* IDs in the info master */
70 #define MATROSKA_ID_TIMECODESCALE 0x2AD7B1
71 #define MATROSKA_ID_DURATION   0x4489
72 #define MATROSKA_ID_TITLE      0x7BA9
73 #define MATROSKA_ID_WRITINGAPP 0x5741
74 #define MATROSKA_ID_MUXINGAPP  0x4D80
75 #define MATROSKA_ID_DATEUTC    0x4461
76
77 /* ID in the tracks master */
78 #define MATROSKA_ID_TRACKENTRY 0xAE
79
80 /* IDs in the trackentry master */
81 #define MATROSKA_ID_TRACKNUMBER 0xD7
82 #define MATROSKA_ID_TRACKUID   0x73C5
83 #define MATROSKA_ID_TRACKTYPE  0x83
84 #define MATROSKA_ID_TRACKAUDIO 0xE1
85 #define MATROSKA_ID_TRACKVIDEO 0xE0
86 #define MATROSKA_ID_CODECID    0x86
87 #define MATROSKA_ID_CODECPRIVATE 0x63A2
88 #define MATROSKA_ID_CODECNAME  0x258688
89 #define MATROSKA_ID_CODECINFOURL 0x3B4040
90 #define MATROSKA_ID_CODECDOWNLOADURL 0x26B240
91 #define MATROSKA_ID_TRACKNAME  0x536E
92 #define MATROSKA_ID_TRACKLANGUAGE 0x22B59C
93 #define MATROSKA_ID_TRACKFLAGENABLED 0xB9
94 #define MATROSKA_ID_TRACKFLAGDEFAULT 0x88
95 #define MATROSKA_ID_TRACKFLAGLACING 0x9C
96 #define MATROSKA_ID_TRACKMINCACHE 0x6DE7
97 #define MATROSKA_ID_TRACKMAXCACHE 0x6DF8
98 #define MATROSKA_ID_TRACKDEFAULTDURATION 0x23E383
99
100 /* IDs in the trackvideo master */
101 #define MATROSKA_ID_VIDEOFRAMERATE 0x2383E3
102 #define MATROSKA_ID_VIDEODISPLAYWIDTH 0x54B0
103 #define MATROSKA_ID_VIDEODISPLAYHEIGHT 0x54BA
104 #define MATROSKA_ID_VIDEOPIXELWIDTH 0xB0
105 #define MATROSKA_ID_VIDEOPIXELHEIGHT 0xBA
106 #define MATROSKA_ID_VIDEOFLAGINTERLACED 0x9A
107 #define MATROSKA_ID_VIDEOSTEREOMODE 0x53B9
108 #define MATROSKA_ID_VIDEOASPECTRATIO 0x54B3
109 #define MATROSKA_ID_VIDEOCOLOURSPACE 0x2EB524
110
111 /* IDs in the trackaudio master */
112 #define MATROSKA_ID_AUDIOSAMPLINGFREQ 0xB5
113 #define MATROSKA_ID_AUDIOOUTSAMPLINGFREQ 0x78B5
114
115 #define MATROSKA_ID_AUDIOBITDEPTH 0x6264
116 #define MATROSKA_ID_AUDIOCHANNELS 0x9F
117
118 /* ID in the cues master */
119 #define MATROSKA_ID_POINTENTRY 0xBB
120
121 /* IDs in the pointentry master */
122 #define MATROSKA_ID_CUETIME    0xB3
123 #define MATROSKA_ID_CUETRACKPOSITION 0xB7
124
125 /* IDs in the cuetrackposition master */
126 #define MATROSKA_ID_CUETRACK   0xF7
127 #define MATROSKA_ID_CUECLUSTERPOSITION 0xF1
128
129 /* IDs in the tags master */
130 /* TODO */
131
132 /* IDs in the seekhead master */
133 #define MATROSKA_ID_SEEKENTRY  0x4DBB
134
135 /* IDs in the seekpoint master */
136 #define MATROSKA_ID_SEEKID     0x53AB
137 #define MATROSKA_ID_SEEKPOSITION 0x53AC
138
139 /* IDs in the cluster master */
140 #define MATROSKA_ID_CLUSTERTIMECODE 0xE7
141 #define MATROSKA_ID_BLOCKGROUP 0xA0
142 #define MATROSKA_ID_SIMPLEBLOCK 0xA3
143
144 /* IDs in the blockgroup master */
145 #define MATROSKA_ID_BLOCK      0xA1
146 #define MATROSKA_ID_BLOCKDURATION 0x9B
147 #define MATROSKA_ID_BLOCKREFERENCE 0xFB
148
149 typedef enum {
150   MATROSKA_TRACK_TYPE_VIDEO    = 0x1,
151   MATROSKA_TRACK_TYPE_AUDIO    = 0x2,
152   MATROSKA_TRACK_TYPE_COMPLEX  = 0x3,
153   MATROSKA_TRACK_TYPE_LOGO     = 0x10,
154   MATROSKA_TRACK_TYPE_SUBTITLE = 0x11,
155   MATROSKA_TRACK_TYPE_CONTROL  = 0x20,
156 } MatroskaTrackType;
157
158 typedef enum {
159   MATROSKA_EYE_MODE_MONO  = 0x0,
160   MATROSKA_EYE_MODE_RIGHT = 0x1,
161   MATROSKA_EYE_MODE_LEFT  = 0x2,
162   MATROSKA_EYE_MODE_BOTH  = 0x3,
163 } MatroskaEyeMode;
164
165 typedef enum {
166   MATROSKA_ASPECT_RATIO_MODE_FREE  = 0x0,
167   MATROSKA_ASPECT_RATIO_MODE_KEEP  = 0x1,
168   MATROSKA_ASPECT_RATIO_MODE_FIXED = 0x2,
169 } MatroskaAspectRatioMode;
170
171 /*
172  * These aren't in any way "matroska-form" things,
173  * it's just something I use in the muxer/demuxer.
174  */
175
176 typedef enum {
177   MATROSKA_TRACK_ENABLED = (1<<0),
178   MATROSKA_TRACK_DEFAULT = (1<<1),
179   MATROSKA_TRACK_LACING  = (1<<2),
180   MATROSKA_TRACK_REAL_V  = (1<<4),
181   MATROSKA_TRACK_SHIFT   = (1<<16)
182 } MatroskaTrackFlags;
183
184 typedef enum {
185   MATROSKA_VIDEOTRACK_INTERLACED = (MATROSKA_TRACK_SHIFT<<0)
186 } MatroskaVideoTrackFlags;
187
188 /*
189  * Matroska Codec IDs. Strings.
190  */
191
192 typedef struct CodecTags{
193     const char *str;
194     enum CodecID id;
195 }CodecTags;
196
197 #define MATROSKA_CODEC_ID_VIDEO_VFW_FOURCC   "V_MS/VFW/FOURCC"
198 #define MATROSKA_CODEC_ID_AUDIO_ACM          "A_MS/ACM"
199
200 static CodecTags codec_tags[]={
201 //    {"V_MS/VFW/FOURCC"  , CODEC_ID_NONE},
202     {"V_UNCOMPRESSED"   , CODEC_ID_RAWVIDEO},
203     {"V_MPEG4/ISO/SP"   , CODEC_ID_MPEG4},
204     {"V_MPEG4/ISO/ASP"  , CODEC_ID_MPEG4},
205     {"V_MPEG4/ISO/AP"   , CODEC_ID_MPEG4},
206     {"V_MPEG4/ISO/AVC"  , CODEC_ID_H264},
207     {"V_MPEG4/MS/V3"    , CODEC_ID_MSMPEG4V3},
208     {"V_MPEG1"          , CODEC_ID_MPEG1VIDEO},
209     {"V_MPEG2"          , CODEC_ID_MPEG2VIDEO},
210     {"V_MJPEG"          , CODEC_ID_MJPEG},
211     {"V_REAL/RV10"      , CODEC_ID_RV10},
212     {"V_REAL/RV20"      , CODEC_ID_RV20},
213     {"V_REAL/RV30"      , CODEC_ID_RV30},
214     {"V_REAL/RV40"      , CODEC_ID_RV40},
215 /* TODO: Real/Quicktime */
216
217 //    {"A_MS/ACM"         , CODEC_ID_NONE},
218     {"A_MPEG/L1"        , CODEC_ID_MP3},
219     {"A_MPEG/L2"        , CODEC_ID_MP3},
220     {"A_MPEG/L3"        , CODEC_ID_MP3},
221     {"A_PCM/INT/BIG"    , CODEC_ID_PCM_U16BE},
222     {"A_PCM/INT/LIT"    , CODEC_ID_PCM_U16LE},
223 //    {"A_PCM/FLOAT/IEEE" , CODEC_ID_NONE},
224     {"A_AC3"            , CODEC_ID_AC3},
225     {"A_DTS"            , CODEC_ID_DTS},
226     {"A_VORBIS"         , CODEC_ID_VORBIS},
227     {"A_AAC"            , CODEC_ID_AAC},
228     {"A_FLAC"           , CODEC_ID_FLAC},
229     {"A_WAVPACK4"       , CODEC_ID_WAVPACK},
230     {"A_TTA1"           , CODEC_ID_TTA},
231     {NULL               , CODEC_ID_NONE}
232 /* TODO: AC3-9/10 (?), Real, Musepack, Quicktime */
233 };
234
235 /* max. depth in the EBML tree structure */
236 #define EBML_MAX_DEPTH 16
237
238 typedef struct Track {
239     MatroskaTrackType type;
240
241     /* Unique track number and track ID. stream_index is the index that
242      * the calling app uses for this track. */
243     uint32_t num,
244         uid,
245         stream_index;
246
247     char *name,
248         *language;
249
250     char *codec_id,
251         *codec_name;
252
253     unsigned char *codec_priv;
254     int codec_priv_size;
255
256     int64_t default_duration;
257     MatroskaTrackFlags flags;
258 } MatroskaTrack;
259
260 typedef struct MatroskaVideoTrack {
261     MatroskaTrack track;
262
263     int pixel_width,
264         pixel_height,
265         display_width,
266         display_height;
267
268     uint32_t fourcc;
269
270     MatroskaAspectRatioMode ar_mode;
271     MatroskaEyeMode eye_mode;
272
273     //..
274 } MatroskaVideoTrack;
275
276 typedef struct MatroskaAudioTrack {
277     MatroskaTrack track;
278
279     int channels,
280         bitdepth,
281         internal_samplerate,
282         samplerate;
283     //..
284 } MatroskaAudioTrack;
285
286 typedef struct MatroskaSubtitleTrack {
287     MatroskaTrack track;
288
289     //..
290 } MatroskaSubtitleTrack;
291
292 #define MAX_TRACK_SIZE (FFMAX(FFMAX(sizeof(MatroskaVideoTrack), \
293                                     sizeof(MatroskaAudioTrack)), \
294                                     sizeof(MatroskaSubtitleTrack)))
295
296 typedef struct MatroskaLevel {
297     uint64_t start, length;
298 } MatroskaLevel;
299
300 typedef struct MatroskaDemuxIndex {
301   uint64_t        pos;   /* of the corresponding *cluster*! */
302   uint16_t        track; /* reference to 'num' */
303   uint64_t        time;  /* in nanoseconds */
304 } MatroskaDemuxIndex;
305
306 typedef struct MatroskaDemuxContext {
307     AVFormatContext *ctx;
308
309     /* ebml stuff */
310     int num_levels;
311     MatroskaLevel levels[EBML_MAX_DEPTH];
312     int level_up;
313
314     /* matroska stuff */
315     char *writing_app,
316         *muxing_app;
317     int64_t created;
318
319     /* timescale in the file */
320     int64_t time_scale;
321
322     /* position (time, ns) */
323     int64_t pos;
324
325     /* num_streams is the number of streams that av_new_stream() was called
326      * for ( = that are available to the calling program). */
327     int num_tracks, num_streams;
328     MatroskaTrack *tracks[MAX_STREAMS];
329
330     /* cache for ID peeking */
331     uint32_t peek_id;
332
333     /* byte position of the segment inside the stream */
334     offset_t segment_start;
335
336     /* The packet queue. */
337     AVPacket **packets;
338     int num_packets;
339
340     /* have we already parse metadata/cues/clusters? */
341     int metadata_parsed,
342         index_parsed,
343         done;
344
345     /* The index for seeking. */
346     int num_indexes;
347     MatroskaDemuxIndex *index;
348 } MatroskaDemuxContext;
349
350 /*
351  * The first few functions handle EBML file parsing. The rest
352  * is the document interpretation. Matroska really just is a
353  * EBML file.
354  */
355
356 /*
357  * Return: the amount of levels in the hierarchy that the
358  * current element lies higher than the previous one.
359  * The opposite isn't done - that's auto-done using master
360  * element reading.
361  */
362
363 static int
364 ebml_read_element_level_up (MatroskaDemuxContext *matroska)
365 {
366     ByteIOContext *pb = &matroska->ctx->pb;
367     offset_t pos = url_ftell(pb);
368     int num = 0;
369
370     while (matroska->num_levels > 0) {
371         MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
372
373         if (pos >= level->start + level->length) {
374             matroska->num_levels--;
375             num++;
376         } else {
377             break;
378         }
379     }
380
381     return num;
382 }
383
384 /*
385  * Read: an "EBML number", which is defined as a variable-length
386  * array of bytes. The first byte indicates the length by giving a
387  * number of 0-bits followed by a one. The position of the first
388  * "one" bit inside the first byte indicates the length of this
389  * number.
390  * Returns: num. of bytes read. < 0 on error.
391  */
392
393 static int
394 ebml_read_num (MatroskaDemuxContext *matroska,
395                int                   max_size,
396                uint64_t             *number)
397 {
398     ByteIOContext *pb = &matroska->ctx->pb;
399     int len_mask = 0x80, read = 1, n = 1;
400     int64_t total = 0;
401
402     /* the first byte tells us the length in bytes - get_byte() can normally
403      * return 0, but since that's not a valid first ebmlID byte, we can
404      * use it safely here to catch EOS. */
405     if (!(total = get_byte(pb))) {
406         /* we might encounter EOS here */
407         if (!url_feof(pb)) {
408             offset_t pos = url_ftell(pb);
409             av_log(matroska->ctx, AV_LOG_ERROR,
410                    "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
411                    pos, pos);
412         }
413         return AVERROR_IO; /* EOS or actual I/O error */
414     }
415
416     /* get the length of the EBML number */
417     while (read <= max_size && !(total & len_mask)) {
418         read++;
419         len_mask >>= 1;
420     }
421     if (read > max_size) {
422         offset_t pos = url_ftell(pb) - 1;
423         av_log(matroska->ctx, AV_LOG_ERROR,
424                "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
425                (uint8_t) total, pos, pos);
426         return AVERROR_INVALIDDATA;
427     }
428
429     /* read out length */
430     total &= ~len_mask;
431     while (n++ < read)
432         total = (total << 8) | get_byte(pb);
433
434     *number = total;
435
436     return read;
437 }
438
439 /*
440  * Read: the element content data ID.
441  * Return: the number of bytes read or < 0 on error.
442  */
443
444 static int
445 ebml_read_element_id (MatroskaDemuxContext *matroska,
446                       uint32_t             *id,
447                       int                  *level_up)
448 {
449     int read;
450     uint64_t total;
451
452     /* if we re-call this, use our cached ID */
453     if (matroska->peek_id != 0) {
454         if (level_up)
455             *level_up = 0;
456         *id = matroska->peek_id;
457         return 0;
458     }
459
460     /* read out the "EBML number", include tag in ID */
461     if ((read = ebml_read_num(matroska, 4, &total)) < 0)
462         return read;
463     *id = matroska->peek_id  = total | (1 << (read * 7));
464
465     /* level tracking */
466     if (level_up)
467         *level_up = ebml_read_element_level_up(matroska);
468
469     return read;
470 }
471
472 /*
473  * Read: element content length.
474  * Return: the number of bytes read or < 0 on error.
475  */
476
477 static int
478 ebml_read_element_length (MatroskaDemuxContext *matroska,
479                           uint64_t             *length)
480 {
481     /* clear cache since we're now beyond that data point */
482     matroska->peek_id = 0;
483
484     /* read out the "EBML number", include tag in ID */
485     return ebml_read_num(matroska, 8, length);
486 }
487
488 /*
489  * Return: the ID of the next element, or 0 on error.
490  * Level_up contains the amount of levels that this
491  * next element lies higher than the previous one.
492  */
493
494 static uint32_t
495 ebml_peek_id (MatroskaDemuxContext *matroska,
496               int                  *level_up)
497 {
498     uint32_t id;
499
500     assert(level_up != NULL);
501
502     if (ebml_read_element_id(matroska, &id, level_up) < 0)
503         return 0;
504
505     return id;
506 }
507
508 /*
509  * Seek to a given offset.
510  * 0 is success, -1 is failure.
511  */
512
513 static int
514 ebml_read_seek (MatroskaDemuxContext *matroska,
515                 offset_t              offset)
516 {
517     ByteIOContext *pb = &matroska->ctx->pb;
518
519     /* clear ID cache, if any */
520     matroska->peek_id = 0;
521
522     return (url_fseek(pb, offset, SEEK_SET) == offset) ? 0 : -1;
523 }
524
525 /*
526  * Skip the next element.
527  * 0 is success, -1 is failure.
528  */
529
530 static int
531 ebml_read_skip (MatroskaDemuxContext *matroska)
532 {
533     ByteIOContext *pb = &matroska->ctx->pb;
534     uint32_t id;
535     uint64_t length;
536     int res;
537
538     if ((res = ebml_read_element_id(matroska, &id, NULL)) < 0 ||
539         (res = ebml_read_element_length(matroska, &length)) < 0)
540         return res;
541
542     url_fskip(pb, length);
543
544     return 0;
545 }
546
547 /*
548  * Read the next element as an unsigned int.
549  * 0 is success, < 0 is failure.
550  */
551
552 static int
553 ebml_read_uint (MatroskaDemuxContext *matroska,
554                 uint32_t             *id,
555                 uint64_t             *num)
556 {
557     ByteIOContext *pb = &matroska->ctx->pb;
558     int n = 0, size, res;
559     uint64_t rlength;
560
561     if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
562         (res = ebml_read_element_length(matroska, &rlength)) < 0)
563         return res;
564     size = rlength;
565     if (size < 1 || size > 8) {
566         offset_t pos = url_ftell(pb);
567         av_log(matroska->ctx, AV_LOG_ERROR,
568                "Invalid uint element size %d at position %"PRId64" (0x%"PRIx64")\n",
569                 size, pos, pos);
570         return AVERROR_INVALIDDATA;
571     }
572
573     /* big-endian ordening; build up number */
574     *num = 0;
575     while (n++ < size)
576         *num = (*num << 8) | get_byte(pb);
577
578     return 0;
579 }
580
581 /*
582  * Read the next element as a signed int.
583  * 0 is success, < 0 is failure.
584  */
585
586 static int
587 ebml_read_sint (MatroskaDemuxContext *matroska,
588                 uint32_t             *id,
589                 int64_t              *num)
590 {
591     ByteIOContext *pb = &matroska->ctx->pb;
592     int size, n = 1, negative = 0, res;
593     uint64_t rlength;
594
595     if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
596         (res = ebml_read_element_length(matroska, &rlength)) < 0)
597         return res;
598     size = rlength;
599     if (size < 1 || size > 8) {
600         offset_t pos = url_ftell(pb);
601         av_log(matroska->ctx, AV_LOG_ERROR,
602                "Invalid sint element size %d at position %"PRId64" (0x%"PRIx64")\n",
603                 size, pos, pos);
604         return AVERROR_INVALIDDATA;
605     }
606     if ((*num = get_byte(pb)) & 0x80) {
607         negative = 1;
608         *num &= ~0x80;
609     }
610     while (n++ < size)
611         *num = (*num << 8) | get_byte(pb);
612
613     /* make signed */
614     if (negative)
615         *num = *num - (1LL << ((8 * size) - 1));
616
617     return 0;
618 }
619
620 /*
621  * Read the next element as a float.
622  * 0 is success, < 0 is failure.
623  */
624
625 static int
626 ebml_read_float (MatroskaDemuxContext *matroska,
627                  uint32_t             *id,
628                  double               *num)
629 {
630     ByteIOContext *pb = &matroska->ctx->pb;
631     int size, res;
632     uint64_t rlength;
633
634     if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
635         (res = ebml_read_element_length(matroska, &rlength)) < 0)
636         return res;
637     size = rlength;
638
639     if (size == 4) {
640         *num= av_int2flt(get_be32(pb));
641     } else if(size==8){
642         *num= av_int2dbl(get_be64(pb));
643     } else{
644         offset_t pos = url_ftell(pb);
645         av_log(matroska->ctx, AV_LOG_ERROR,
646                "Invalid float element size %d at position %"PRIu64" (0x%"PRIx64")\n",
647                size, pos, pos);
648         return AVERROR_INVALIDDATA;
649     }
650
651     return 0;
652 }
653
654 /*
655  * Read the next element as an ASCII string.
656  * 0 is success, < 0 is failure.
657  */
658
659 static int
660 ebml_read_ascii (MatroskaDemuxContext *matroska,
661                  uint32_t             *id,
662                  char                **str)
663 {
664     ByteIOContext *pb = &matroska->ctx->pb;
665     int size, res;
666     uint64_t rlength;
667
668     if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
669         (res = ebml_read_element_length(matroska, &rlength)) < 0)
670         return res;
671     size = rlength;
672
673     /* ebml strings are usually not 0-terminated, so we allocate one
674      * byte more, read the string and NULL-terminate it ourselves. */
675     if (size < 0 || !(*str = av_malloc(size + 1))) {
676         av_log(matroska->ctx, AV_LOG_ERROR, "Memory allocation failed\n");
677         return AVERROR_NOMEM;
678     }
679     if (get_buffer(pb, (uint8_t *) *str, size) != size) {
680         offset_t pos = url_ftell(pb);
681         av_log(matroska->ctx, AV_LOG_ERROR,
682                "Read error at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
683         return AVERROR_IO;
684     }
685     (*str)[size] = '\0';
686
687     return 0;
688 }
689
690 /*
691  * Read the next element as a UTF-8 string.
692  * 0 is success, < 0 is failure.
693  */
694
695 static int
696 ebml_read_utf8 (MatroskaDemuxContext *matroska,
697                 uint32_t             *id,
698                 char                **str)
699 {
700   return ebml_read_ascii(matroska, id, str);
701 }
702
703 /*
704  * Read the next element as a date (nanoseconds since 1/1/2000).
705  * 0 is success, < 0 is failure.
706  */
707
708 static int
709 ebml_read_date (MatroskaDemuxContext *matroska,
710                 uint32_t             *id,
711                 int64_t              *date)
712 {
713   return ebml_read_sint(matroska, id, date);
714 }
715
716 /*
717  * Read the next element, but only the header. The contents
718  * are supposed to be sub-elements which can be read separately.
719  * 0 is success, < 0 is failure.
720  */
721
722 static int
723 ebml_read_master (MatroskaDemuxContext *matroska,
724                   uint32_t             *id)
725 {
726     ByteIOContext *pb = &matroska->ctx->pb;
727     uint64_t length;
728     MatroskaLevel *level;
729     int res;
730
731     if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
732         (res = ebml_read_element_length(matroska, &length)) < 0)
733         return res;
734
735     /* protect... (Heaven forbids that the '>' is true) */
736     if (matroska->num_levels >= EBML_MAX_DEPTH) {
737         av_log(matroska->ctx, AV_LOG_ERROR,
738                "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
739         return AVERROR_NOTSUPP;
740     }
741
742     /* remember level */
743     level = &matroska->levels[matroska->num_levels++];
744     level->start = url_ftell(pb);
745     level->length = length;
746
747     return 0;
748 }
749
750 /*
751  * Read the next element as binary data.
752  * 0 is success, < 0 is failure.
753  */
754
755 static int
756 ebml_read_binary (MatroskaDemuxContext *matroska,
757                   uint32_t             *id,
758                   uint8_t             **binary,
759                   int                  *size)
760 {
761     ByteIOContext *pb = &matroska->ctx->pb;
762     uint64_t rlength;
763     int res;
764
765     if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
766         (res = ebml_read_element_length(matroska, &rlength)) < 0)
767         return res;
768     *size = rlength;
769
770     if (!(*binary = av_malloc(*size))) {
771         av_log(matroska->ctx, AV_LOG_ERROR,
772                "Memory allocation error\n");
773         return AVERROR_NOMEM;
774     }
775
776     if (get_buffer(pb, *binary, *size) != *size) {
777         offset_t pos = url_ftell(pb);
778         av_log(matroska->ctx, AV_LOG_ERROR,
779                "Read error at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
780         return AVERROR_IO;
781     }
782
783     return 0;
784 }
785
786 /*
787  * Read signed/unsigned "EBML" numbers.
788  * Return: number of bytes processed, < 0 on error.
789  * XXX: use ebml_read_num().
790  */
791
792 static int
793 matroska_ebmlnum_uint (uint8_t  *data,
794                        uint32_t  size,
795                        uint64_t *num)
796 {
797     int len_mask = 0x80, read = 1, n = 1, num_ffs = 0;
798     uint64_t total;
799
800     if (size <= 0)
801         return AVERROR_INVALIDDATA;
802
803     total = data[0];
804     while (read <= 8 && !(total & len_mask)) {
805         read++;
806         len_mask >>= 1;
807     }
808     if (read > 8)
809         return AVERROR_INVALIDDATA;
810
811     if ((total &= (len_mask - 1)) == len_mask - 1)
812         num_ffs++;
813     if (size < read)
814         return AVERROR_INVALIDDATA;
815     while (n < read) {
816         if (data[n] == 0xff)
817             num_ffs++;
818         total = (total << 8) | data[n];
819         n++;
820     }
821
822     if (read == num_ffs)
823         *num = (uint64_t)-1;
824     else
825         *num = total;
826
827     return read;
828 }
829
830 /*
831  * Same as above, but signed.
832  */
833
834 static int
835 matroska_ebmlnum_sint (uint8_t  *data,
836                        uint32_t  size,
837                        int64_t  *num)
838 {
839     uint64_t unum;
840     int res;
841
842     /* read as unsigned number first */
843     if ((res = matroska_ebmlnum_uint(data, size, &unum)) < 0)
844         return res;
845
846     /* make signed (weird way) */
847     if (unum == (uint64_t)-1)
848         *num = INT64_MAX;
849     else
850         *num = unum - ((1LL << ((7 * res) - 1)) - 1);
851
852     return res;
853 }
854
855 /*
856  * Read an EBML header.
857  * 0 is success, < 0 is failure.
858  */
859
860 static int
861 ebml_read_header (MatroskaDemuxContext *matroska,
862                   char                **doctype,
863                   int                  *version)
864 {
865     uint32_t id;
866     int level_up, res = 0;
867
868     /* default init */
869     if (doctype)
870         *doctype = NULL;
871     if (version)
872         *version = 1;
873
874     if (!(id = ebml_peek_id(matroska, &level_up)) ||
875         level_up != 0 || id != EBML_ID_HEADER) {
876         av_log(matroska->ctx, AV_LOG_ERROR,
877                "This is not an EBML file (id=0x%x/0x%x)\n", id, EBML_ID_HEADER);
878         return AVERROR_INVALIDDATA;
879     }
880     if ((res = ebml_read_master(matroska, &id)) < 0)
881         return res;
882
883     while (res == 0) {
884         if (!(id = ebml_peek_id(matroska, &level_up)))
885             return AVERROR_IO;
886
887         /* end-of-header */
888         if (level_up)
889             break;
890
891         switch (id) {
892             /* is our read version uptodate? */
893             case EBML_ID_EBMLREADVERSION: {
894                 uint64_t num;
895
896                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
897                     return res;
898                 if (num > EBML_VERSION) {
899                     av_log(matroska->ctx, AV_LOG_ERROR,
900                            "EBML version %"PRIu64" (> %d) is not supported\n",
901                            num, EBML_VERSION);
902                     return AVERROR_INVALIDDATA;
903                 }
904                 break;
905             }
906
907             /* we only handle 8 byte lengths at max */
908             case EBML_ID_EBMLMAXSIZELENGTH: {
909                 uint64_t num;
910
911                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
912                     return res;
913                 if (num > sizeof(uint64_t)) {
914                     av_log(matroska->ctx, AV_LOG_ERROR,
915                            "Integers of size %"PRIu64" (> %zd) not supported\n",
916                            num, sizeof(uint64_t));
917                     return AVERROR_INVALIDDATA;
918                 }
919                 break;
920             }
921
922             /* we handle 4 byte IDs at max */
923             case EBML_ID_EBMLMAXIDLENGTH: {
924                 uint64_t num;
925
926                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
927                     return res;
928                 if (num > sizeof(uint32_t)) {
929                     av_log(matroska->ctx, AV_LOG_ERROR,
930                            "IDs of size %"PRIu64" (> %zu) not supported\n",
931                             num, sizeof(uint32_t));
932                     return AVERROR_INVALIDDATA;
933                 }
934                 break;
935             }
936
937             case EBML_ID_DOCTYPE: {
938                 char *text;
939
940                 if ((res = ebml_read_ascii(matroska, &id, &text)) < 0)
941                     return res;
942                 if (doctype) {
943                     if (*doctype)
944                         av_free(*doctype);
945                     *doctype = text;
946                 } else
947                     av_free(text);
948                 break;
949             }
950
951             case EBML_ID_DOCTYPEREADVERSION: {
952                 uint64_t num;
953
954                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
955                     return res;
956                 if (version)
957                     *version = num;
958                 break;
959             }
960
961             default:
962                 av_log(matroska->ctx, AV_LOG_INFO,
963                        "Unknown data type 0x%x in EBML header", id);
964                 /* pass-through */
965
966             case EBML_ID_VOID:
967             /* we ignore these two, as they don't tell us anything we
968              * care about */
969             case EBML_ID_EBMLVERSION:
970             case EBML_ID_DOCTYPEVERSION:
971                 res = ebml_read_skip (matroska);
972                 break;
973         }
974     }
975
976     return 0;
977 }
978
979 /*
980  * Put one packet in an application-supplied AVPacket struct.
981  * Returns 0 on success or -1 on failure.
982  */
983
984 static int
985 matroska_deliver_packet (MatroskaDemuxContext *matroska,
986                          AVPacket             *pkt)
987 {
988     if (matroska->num_packets > 0) {
989         memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
990         av_free(matroska->packets[0]);
991         if (matroska->num_packets > 1) {
992             memmove(&matroska->packets[0], &matroska->packets[1],
993                     (matroska->num_packets - 1) * sizeof(AVPacket *));
994             matroska->packets =
995                 av_realloc(matroska->packets, (matroska->num_packets - 1) *
996                            sizeof(AVPacket *));
997         } else {
998             av_freep(&matroska->packets);
999         }
1000         matroska->num_packets--;
1001         return 0;
1002     }
1003
1004     return -1;
1005 }
1006
1007 /*
1008  * Put a packet into our internal queue. Will be delivered to the
1009  * user/application during the next get_packet() call.
1010  */
1011
1012 static void
1013 matroska_queue_packet (MatroskaDemuxContext *matroska,
1014                        AVPacket             *pkt)
1015 {
1016     matroska->packets =
1017         av_realloc(matroska->packets, (matroska->num_packets + 1) *
1018                    sizeof(AVPacket *));
1019     matroska->packets[matroska->num_packets] = pkt;
1020     matroska->num_packets++;
1021 }
1022
1023 /*
1024  * Autodetecting...
1025  */
1026
1027 static int
1028 matroska_probe (AVProbeData *p)
1029 {
1030     uint64_t total = 0;
1031     int len_mask = 0x80, size = 1, n = 1;
1032     uint8_t probe_data[] = { 'm', 'a', 't', 'r', 'o', 's', 'k', 'a' };
1033
1034     if (p->buf_size < 5)
1035         return 0;
1036
1037     /* ebml header? */
1038     if ((p->buf[0] << 24 | p->buf[1] << 16 |
1039          p->buf[2] << 8 | p->buf[3]) != EBML_ID_HEADER)
1040         return 0;
1041
1042     /* length of header */
1043     total = p->buf[4];
1044     while (size <= 8 && !(total & len_mask)) {
1045         size++;
1046         len_mask >>= 1;
1047     }
1048     if (size > 8)
1049       return 0;
1050     total &= (len_mask - 1);
1051     while (n < size)
1052         total = (total << 8) | p->buf[4 + n++];
1053
1054     /* does the probe data contain the whole header? */
1055     if (p->buf_size < 4 + size + total)
1056       return 0;
1057
1058     /* the header must contain the document type 'matroska'. For now,
1059      * we don't parse the whole header but simply check for the
1060      * availability of that array of characters inside the header.
1061      * Not fully fool-proof, but good enough. */
1062     for (n = 4 + size; n <= 4 + size + total - sizeof(probe_data); n++)
1063         if (!memcmp (&p->buf[n], probe_data, sizeof(probe_data)))
1064             return AVPROBE_SCORE_MAX;
1065
1066     return 0;
1067 }
1068
1069 /*
1070  * From here on, it's all XML-style DTD stuff... Needs no comments.
1071  */
1072
1073 static int
1074 matroska_parse_info (MatroskaDemuxContext *matroska)
1075 {
1076     int res = 0;
1077     uint32_t id;
1078
1079     av_log(matroska->ctx, AV_LOG_DEBUG, "Parsing info...\n");
1080
1081     while (res == 0) {
1082         if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1083             res = AVERROR_IO;
1084             break;
1085         } else if (matroska->level_up) {
1086             matroska->level_up--;
1087             break;
1088         }
1089
1090         switch (id) {
1091             /* cluster timecode */
1092             case MATROSKA_ID_TIMECODESCALE: {
1093                 uint64_t num;
1094                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1095                     break;
1096                 matroska->time_scale = num;
1097                 break;
1098             }
1099
1100             case MATROSKA_ID_DURATION: {
1101                 double num;
1102                 if ((res = ebml_read_float(matroska, &id, &num)) < 0)
1103                     break;
1104                 matroska->ctx->duration = num * matroska->time_scale * 1000 / AV_TIME_BASE;
1105                 break;
1106             }
1107
1108             case MATROSKA_ID_TITLE: {
1109                 char *text;
1110                 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1111                     break;
1112                 strncpy(matroska->ctx->title, text,
1113                         sizeof(matroska->ctx->title)-1);
1114                 av_free(text);
1115                 break;
1116             }
1117
1118             case MATROSKA_ID_WRITINGAPP: {
1119                 char *text;
1120                 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1121                     break;
1122                 matroska->writing_app = text;
1123                 break;
1124             }
1125
1126             case MATROSKA_ID_MUXINGAPP: {
1127                 char *text;
1128                 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1129                     break;
1130                 matroska->muxing_app = text;
1131                 break;
1132             }
1133
1134             case MATROSKA_ID_DATEUTC: {
1135                 int64_t time;
1136                 if ((res = ebml_read_date(matroska, &id, &time)) < 0)
1137                     break;
1138                 matroska->created = time;
1139                 break;
1140             }
1141
1142             default:
1143                 av_log(matroska->ctx, AV_LOG_INFO,
1144                        "Unknown entry 0x%x in info header\n", id);
1145                 /* fall-through */
1146
1147             case EBML_ID_VOID:
1148                 res = ebml_read_skip(matroska);
1149                 break;
1150         }
1151
1152         if (matroska->level_up) {
1153             matroska->level_up--;
1154             break;
1155         }
1156     }
1157
1158     return res;
1159 }
1160
1161 static int
1162 matroska_add_stream (MatroskaDemuxContext *matroska)
1163 {
1164     int res = 0;
1165     uint32_t id;
1166     MatroskaTrack *track;
1167
1168     av_log(matroska->ctx, AV_LOG_DEBUG, "parsing track, adding stream..,\n");
1169
1170     /* Allocate a generic track. As soon as we know its type we'll realloc. */
1171     track = av_mallocz(MAX_TRACK_SIZE);
1172     matroska->num_tracks++;
1173
1174     /* start with the master */
1175     if ((res = ebml_read_master(matroska, &id)) < 0)
1176         return res;
1177
1178     /* try reading the trackentry headers */
1179     while (res == 0) {
1180         if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1181             res = AVERROR_IO;
1182             break;
1183         } else if (matroska->level_up > 0) {
1184             matroska->level_up--;
1185             break;
1186         }
1187
1188         switch (id) {
1189             /* track number (unique stream ID) */
1190             case MATROSKA_ID_TRACKNUMBER: {
1191                 uint64_t num;
1192                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1193                     break;
1194                 track->num = num;
1195                 break;
1196             }
1197
1198             /* track UID (unique identifier) */
1199             case MATROSKA_ID_TRACKUID: {
1200                 uint64_t num;
1201                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1202                     break;
1203                 track->uid = num;
1204                 break;
1205             }
1206
1207             /* track type (video, audio, combined, subtitle, etc.) */
1208             case MATROSKA_ID_TRACKTYPE: {
1209                 uint64_t num;
1210                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1211                     break;
1212                 if (track->type && track->type != num) {
1213                     av_log(matroska->ctx, AV_LOG_INFO,
1214                            "More than one tracktype in an entry - skip\n");
1215                     break;
1216                 }
1217                 track->type = num;
1218
1219                 switch (track->type) {
1220                     case MATROSKA_TRACK_TYPE_VIDEO:
1221                     case MATROSKA_TRACK_TYPE_AUDIO:
1222                     case MATROSKA_TRACK_TYPE_SUBTITLE:
1223                         break;
1224                     case MATROSKA_TRACK_TYPE_COMPLEX:
1225                     case MATROSKA_TRACK_TYPE_LOGO:
1226                     case MATROSKA_TRACK_TYPE_CONTROL:
1227                     default:
1228                         av_log(matroska->ctx, AV_LOG_INFO,
1229                                "Unknown or unsupported track type 0x%x\n",
1230                                track->type);
1231                         track->type = 0;
1232                         break;
1233                 }
1234                 matroska->tracks[matroska->num_tracks - 1] = track;
1235                 break;
1236             }
1237
1238             /* tracktype specific stuff for video */
1239             case MATROSKA_ID_TRACKVIDEO: {
1240                 MatroskaVideoTrack *videotrack;
1241                 if (!track->type)
1242                     track->type = MATROSKA_TRACK_TYPE_VIDEO;
1243                 if (track->type != MATROSKA_TRACK_TYPE_VIDEO) {
1244                     av_log(matroska->ctx, AV_LOG_INFO,
1245                            "video data in non-video track - ignoring\n");
1246                     res = AVERROR_INVALIDDATA;
1247                     break;
1248                 } else if ((res = ebml_read_master(matroska, &id)) < 0)
1249                     break;
1250                 videotrack = (MatroskaVideoTrack *)track;
1251
1252                 while (res == 0) {
1253                     if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1254                         res = AVERROR_IO;
1255                         break;
1256                     } else if (matroska->level_up > 0) {
1257                         matroska->level_up--;
1258                         break;
1259                     }
1260
1261                     switch (id) {
1262                         /* fixme, this should be one-up, but I get it here */
1263                         case MATROSKA_ID_TRACKDEFAULTDURATION: {
1264                             uint64_t num;
1265                             if ((res = ebml_read_uint (matroska, &id,
1266                                                        &num)) < 0)
1267                                 break;
1268                             track->default_duration = num;
1269                             break;
1270                         }
1271
1272                         /* video framerate */
1273                         case MATROSKA_ID_VIDEOFRAMERATE: {
1274                             double num;
1275                             if ((res = ebml_read_float(matroska, &id,
1276                                                        &num)) < 0)
1277                                 break;
1278                             track->default_duration = 1000000000 * (1. / num);
1279                             break;
1280                         }
1281
1282                         /* width of the size to display the video at */
1283                         case MATROSKA_ID_VIDEODISPLAYWIDTH: {
1284                             uint64_t num;
1285                             if ((res = ebml_read_uint(matroska, &id,
1286                                                       &num)) < 0)
1287                                 break;
1288                             videotrack->display_width = num;
1289                             break;
1290                         }
1291
1292                         /* height of the size to display the video at */
1293                         case MATROSKA_ID_VIDEODISPLAYHEIGHT: {
1294                             uint64_t num;
1295                             if ((res = ebml_read_uint(matroska, &id,
1296                                                       &num)) < 0)
1297                                 break;
1298                             videotrack->display_height = num;
1299                             break;
1300                         }
1301
1302                         /* width of the video in the file */
1303                         case MATROSKA_ID_VIDEOPIXELWIDTH: {
1304                             uint64_t num;
1305                             if ((res = ebml_read_uint(matroska, &id,
1306                                                       &num)) < 0)
1307                                 break;
1308                             videotrack->pixel_width = num;
1309                             break;
1310                         }
1311
1312                         /* height of the video in the file */
1313                         case MATROSKA_ID_VIDEOPIXELHEIGHT: {
1314                             uint64_t num;
1315                             if ((res = ebml_read_uint(matroska, &id,
1316                                                       &num)) < 0)
1317                                 break;
1318                             videotrack->pixel_height = num;
1319                             break;
1320                         }
1321
1322                         /* whether the video is interlaced */
1323                         case MATROSKA_ID_VIDEOFLAGINTERLACED: {
1324                             uint64_t num;
1325                             if ((res = ebml_read_uint(matroska, &id,
1326                                                       &num)) < 0)
1327                                 break;
1328                             if (num)
1329                                 track->flags |=
1330                                     MATROSKA_VIDEOTRACK_INTERLACED;
1331                             else
1332                                 track->flags &=
1333                                     ~MATROSKA_VIDEOTRACK_INTERLACED;
1334                             break;
1335                         }
1336
1337                         /* stereo mode (whether the video has two streams,
1338                          * where one is for the left eye and the other for
1339                          * the right eye, which creates a 3D-like
1340                          * effect) */
1341                         case MATROSKA_ID_VIDEOSTEREOMODE: {
1342                             uint64_t num;
1343                             if ((res = ebml_read_uint(matroska, &id,
1344                                                       &num)) < 0)
1345                                 break;
1346                             if (num != MATROSKA_EYE_MODE_MONO &&
1347                                 num != MATROSKA_EYE_MODE_LEFT &&
1348                                 num != MATROSKA_EYE_MODE_RIGHT &&
1349                                 num != MATROSKA_EYE_MODE_BOTH) {
1350                                 av_log(matroska->ctx, AV_LOG_INFO,
1351                                        "Ignoring unknown eye mode 0x%x\n",
1352                                        (uint32_t) num);
1353                                 break;
1354                             }
1355                             videotrack->eye_mode = num;
1356                             break;
1357                         }
1358
1359                         /* aspect ratio behaviour */
1360                         case MATROSKA_ID_VIDEOASPECTRATIO: {
1361                             uint64_t num;
1362                             if ((res = ebml_read_uint(matroska, &id,
1363                                                       &num)) < 0)
1364                                 break;
1365                             if (num != MATROSKA_ASPECT_RATIO_MODE_FREE &&
1366                                 num != MATROSKA_ASPECT_RATIO_MODE_KEEP &&
1367                                 num != MATROSKA_ASPECT_RATIO_MODE_FIXED) {
1368                                 av_log(matroska->ctx, AV_LOG_INFO,
1369                                        "Ignoring unknown aspect ratio 0x%x\n",
1370                                        (uint32_t) num);
1371                                 break;
1372                             }
1373                             videotrack->ar_mode = num;
1374                             break;
1375                         }
1376
1377                         /* colourspace (only matters for raw video)
1378                          * fourcc */
1379                         case MATROSKA_ID_VIDEOCOLOURSPACE: {
1380                             uint64_t num;
1381                             if ((res = ebml_read_uint(matroska, &id,
1382                                                       &num)) < 0)
1383                                 break;
1384                             videotrack->fourcc = num;
1385                             break;
1386                         }
1387
1388                         default:
1389                             av_log(matroska->ctx, AV_LOG_INFO,
1390                                    "Unknown video track header entry "
1391                                    "0x%x - ignoring\n", id);
1392                             /* pass-through */
1393
1394                         case EBML_ID_VOID:
1395                             res = ebml_read_skip(matroska);
1396                             break;
1397                     }
1398
1399                     if (matroska->level_up) {
1400                         matroska->level_up--;
1401                         break;
1402                     }
1403                 }
1404                 break;
1405             }
1406
1407             /* tracktype specific stuff for audio */
1408             case MATROSKA_ID_TRACKAUDIO: {
1409                 MatroskaAudioTrack *audiotrack;
1410                 if (!track->type)
1411                     track->type = MATROSKA_TRACK_TYPE_AUDIO;
1412                 if (track->type != MATROSKA_TRACK_TYPE_AUDIO) {
1413                     av_log(matroska->ctx, AV_LOG_INFO,
1414                            "audio data in non-audio track - ignoring\n");
1415                     res = AVERROR_INVALIDDATA;
1416                     break;
1417                 } else if ((res = ebml_read_master(matroska, &id)) < 0)
1418                     break;
1419                 audiotrack = (MatroskaAudioTrack *)track;
1420                 audiotrack->channels = 1;
1421                 audiotrack->samplerate = 8000;
1422
1423                 while (res == 0) {
1424                     if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1425                         res = AVERROR_IO;
1426                         break;
1427                     } else if (matroska->level_up > 0) {
1428                         matroska->level_up--;
1429                         break;
1430                     }
1431
1432                     switch (id) {
1433                         /* samplerate */
1434                         case MATROSKA_ID_AUDIOSAMPLINGFREQ: {
1435                             double num;
1436                             if ((res = ebml_read_float(matroska, &id,
1437                                                        &num)) < 0)
1438                                 break;
1439                             audiotrack->internal_samplerate =
1440                             audiotrack->samplerate = num;
1441                             break;
1442                         }
1443
1444                         case MATROSKA_ID_AUDIOOUTSAMPLINGFREQ: {
1445                             double num;
1446                             if ((res = ebml_read_float(matroska, &id,
1447                                                        &num)) < 0)
1448                                 break;
1449                             audiotrack->samplerate = num;
1450                             break;
1451                         }
1452
1453                             /* bitdepth */
1454                         case MATROSKA_ID_AUDIOBITDEPTH: {
1455                             uint64_t num;
1456                             if ((res = ebml_read_uint(matroska, &id,
1457                                                       &num)) < 0)
1458                                 break;
1459                             audiotrack->bitdepth = num;
1460                             break;
1461                         }
1462
1463                             /* channels */
1464                         case MATROSKA_ID_AUDIOCHANNELS: {
1465                             uint64_t num;
1466                             if ((res = ebml_read_uint(matroska, &id,
1467                                                       &num)) < 0)
1468                                 break;
1469                             audiotrack->channels = num;
1470                             break;
1471                         }
1472
1473                         default:
1474                             av_log(matroska->ctx, AV_LOG_INFO,
1475                                    "Unknown audio track header entry "
1476                                    "0x%x - ignoring\n", id);
1477                             /* pass-through */
1478
1479                         case EBML_ID_VOID:
1480                             res = ebml_read_skip(matroska);
1481                             break;
1482                     }
1483
1484                     if (matroska->level_up) {
1485                         matroska->level_up--;
1486                         break;
1487                     }
1488                 }
1489                 break;
1490             }
1491
1492                 /* codec identifier */
1493             case MATROSKA_ID_CODECID: {
1494                 char *text;
1495                 if ((res = ebml_read_ascii(matroska, &id, &text)) < 0)
1496                     break;
1497                 track->codec_id = text;
1498                 break;
1499             }
1500
1501                 /* codec private data */
1502             case MATROSKA_ID_CODECPRIVATE: {
1503                 uint8_t *data;
1504                 int size;
1505                 if ((res = ebml_read_binary(matroska, &id, &data, &size) < 0))
1506                     break;
1507                 track->codec_priv = data;
1508                 track->codec_priv_size = size;
1509                 break;
1510             }
1511
1512                 /* name of the codec */
1513             case MATROSKA_ID_CODECNAME: {
1514                 char *text;
1515                 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1516                     break;
1517                 track->codec_name = text;
1518                 break;
1519             }
1520
1521                 /* name of this track */
1522             case MATROSKA_ID_TRACKNAME: {
1523                 char *text;
1524                 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1525                     break;
1526                 track->name = text;
1527                 break;
1528             }
1529
1530                 /* language (matters for audio/subtitles, mostly) */
1531             case MATROSKA_ID_TRACKLANGUAGE: {
1532                 char *text;
1533                 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1534                     break;
1535                 track->language = text;
1536                 break;
1537             }
1538
1539                 /* whether this is actually used */
1540             case MATROSKA_ID_TRACKFLAGENABLED: {
1541                 uint64_t num;
1542                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1543                     break;
1544                 if (num)
1545                     track->flags |= MATROSKA_TRACK_ENABLED;
1546                 else
1547                     track->flags &= ~MATROSKA_TRACK_ENABLED;
1548                 break;
1549             }
1550
1551                 /* whether it's the default for this track type */
1552             case MATROSKA_ID_TRACKFLAGDEFAULT: {
1553                 uint64_t num;
1554                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1555                     break;
1556                 if (num)
1557                     track->flags |= MATROSKA_TRACK_DEFAULT;
1558                 else
1559                     track->flags &= ~MATROSKA_TRACK_DEFAULT;
1560                 break;
1561             }
1562
1563                 /* lacing (like MPEG, where blocks don't end/start on frame
1564                  * boundaries) */
1565             case MATROSKA_ID_TRACKFLAGLACING: {
1566                 uint64_t num;
1567                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1568                     break;
1569                 if (num)
1570                     track->flags |= MATROSKA_TRACK_LACING;
1571                 else
1572                     track->flags &= ~MATROSKA_TRACK_LACING;
1573                 break;
1574             }
1575
1576                 /* default length (in time) of one data block in this track */
1577             case MATROSKA_ID_TRACKDEFAULTDURATION: {
1578                 uint64_t num;
1579                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1580                     break;
1581                 track->default_duration = num;
1582                 break;
1583             }
1584
1585             default:
1586                 av_log(matroska->ctx, AV_LOG_INFO,
1587                        "Unknown track header entry 0x%x - ignoring\n", id);
1588                 /* pass-through */
1589
1590             case EBML_ID_VOID:
1591             /* we ignore these because they're nothing useful. */
1592             case MATROSKA_ID_CODECINFOURL:
1593             case MATROSKA_ID_CODECDOWNLOADURL:
1594             case MATROSKA_ID_TRACKMINCACHE:
1595             case MATROSKA_ID_TRACKMAXCACHE:
1596                 res = ebml_read_skip(matroska);
1597                 break;
1598         }
1599
1600         if (matroska->level_up) {
1601             matroska->level_up--;
1602             break;
1603         }
1604     }
1605
1606     return res;
1607 }
1608
1609 static int
1610 matroska_parse_tracks (MatroskaDemuxContext *matroska)
1611 {
1612     int res = 0;
1613     uint32_t id;
1614
1615     av_log(matroska->ctx, AV_LOG_DEBUG, "parsing tracks...\n");
1616
1617     while (res == 0) {
1618         if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1619             res = AVERROR_IO;
1620             break;
1621         } else if (matroska->level_up) {
1622             matroska->level_up--;
1623             break;
1624         }
1625
1626         switch (id) {
1627             /* one track within the "all-tracks" header */
1628             case MATROSKA_ID_TRACKENTRY:
1629                 res = matroska_add_stream(matroska);
1630                 break;
1631
1632             default:
1633                 av_log(matroska->ctx, AV_LOG_INFO,
1634                        "Unknown entry 0x%x in track header\n", id);
1635                 /* fall-through */
1636
1637             case EBML_ID_VOID:
1638                 res = ebml_read_skip(matroska);
1639                 break;
1640         }
1641
1642         if (matroska->level_up) {
1643             matroska->level_up--;
1644             break;
1645         }
1646     }
1647
1648     return res;
1649 }
1650
1651 static int
1652 matroska_parse_index (MatroskaDemuxContext *matroska)
1653 {
1654     int res = 0;
1655     uint32_t id;
1656     MatroskaDemuxIndex idx;
1657
1658     av_log(matroska->ctx, AV_LOG_DEBUG, "parsing index...\n");
1659
1660     while (res == 0) {
1661         if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1662             res = AVERROR_IO;
1663             break;
1664         } else if (matroska->level_up) {
1665             matroska->level_up--;
1666             break;
1667         }
1668
1669         switch (id) {
1670             /* one single index entry ('point') */
1671             case MATROSKA_ID_POINTENTRY:
1672                 if ((res = ebml_read_master(matroska, &id)) < 0)
1673                     break;
1674
1675                 /* in the end, we hope to fill one entry with a
1676                  * timestamp, a file position and a tracknum */
1677                 idx.pos   = (uint64_t) -1;
1678                 idx.time  = (uint64_t) -1;
1679                 idx.track = (uint16_t) -1;
1680
1681                 while (res == 0) {
1682                     if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1683                         res = AVERROR_IO;
1684                         break;
1685                     } else if (matroska->level_up) {
1686                         matroska->level_up--;
1687                         break;
1688                     }
1689
1690                     switch (id) {
1691                         /* one single index entry ('point') */
1692                         case MATROSKA_ID_CUETIME: {
1693                             uint64_t time;
1694                             if ((res = ebml_read_uint(matroska, &id,
1695                                                       &time)) < 0)
1696                                 break;
1697                             idx.time = time * matroska->time_scale;
1698                             break;
1699                         }
1700
1701                         /* position in the file + track to which it
1702                          * belongs */
1703                         case MATROSKA_ID_CUETRACKPOSITION:
1704                             if ((res = ebml_read_master(matroska, &id)) < 0)
1705                                 break;
1706
1707                             while (res == 0) {
1708                                 if (!(id = ebml_peek_id (matroska,
1709                                                     &matroska->level_up))) {
1710                                     res = AVERROR_IO;
1711                                     break;
1712                                 } else if (matroska->level_up) {
1713                                     matroska->level_up--;
1714                                     break;
1715                                 }
1716
1717                                 switch (id) {
1718                                     /* track number */
1719                                     case MATROSKA_ID_CUETRACK: {
1720                                         uint64_t num;
1721                                         if ((res = ebml_read_uint(matroska,
1722                                                           &id, &num)) < 0)
1723                                             break;
1724                                         idx.track = num;
1725                                         break;
1726                                     }
1727
1728                                         /* position in file */
1729                                     case MATROSKA_ID_CUECLUSTERPOSITION: {
1730                                         uint64_t num;
1731                                         if ((res = ebml_read_uint(matroska,
1732                                                           &id, &num)) < 0)
1733                                             break;
1734                                         idx.pos = num;
1735                                         break;
1736                                     }
1737
1738                                     default:
1739                                         av_log(matroska->ctx, AV_LOG_INFO,
1740                                                "Unknown entry 0x%x in "
1741                                                "CuesTrackPositions\n", id);
1742                                         /* fall-through */
1743
1744                                     case EBML_ID_VOID:
1745                                         res = ebml_read_skip(matroska);
1746                                         break;
1747                                 }
1748
1749                                 if (matroska->level_up) {
1750                                     matroska->level_up--;
1751                                     break;
1752                                 }
1753                             }
1754
1755                             break;
1756
1757                         default:
1758                             av_log(matroska->ctx, AV_LOG_INFO,
1759                                    "Unknown entry 0x%x in cuespoint "
1760                                    "index\n", id);
1761                             /* fall-through */
1762
1763                         case EBML_ID_VOID:
1764                             res = ebml_read_skip(matroska);
1765                             break;
1766                     }
1767
1768                     if (matroska->level_up) {
1769                         matroska->level_up--;
1770                         break;
1771                     }
1772                 }
1773
1774                 /* so let's see if we got what we wanted */
1775                 if (idx.pos   != (uint64_t) -1 &&
1776                     idx.time  != (uint64_t) -1 &&
1777                     idx.track != (uint16_t) -1) {
1778                     if (matroska->num_indexes % 32 == 0) {
1779                         /* re-allocate bigger index */
1780                         matroska->index =
1781                             av_realloc(matroska->index,
1782                                        (matroska->num_indexes + 32) *
1783                                        sizeof(MatroskaDemuxIndex));
1784                     }
1785                     matroska->index[matroska->num_indexes] = idx;
1786                     matroska->num_indexes++;
1787                 }
1788                 break;
1789
1790             default:
1791                 av_log(matroska->ctx, AV_LOG_INFO,
1792                        "Unknown entry 0x%x in cues header\n", id);
1793                 /* fall-through */
1794
1795             case EBML_ID_VOID:
1796                 res = ebml_read_skip(matroska);
1797                 break;
1798         }
1799
1800         if (matroska->level_up) {
1801             matroska->level_up--;
1802             break;
1803         }
1804     }
1805
1806     return res;
1807 }
1808
1809 static int
1810 matroska_parse_metadata (MatroskaDemuxContext *matroska)
1811 {
1812     int res = 0;
1813     uint32_t id;
1814
1815     while (res == 0) {
1816         if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1817             res = AVERROR_IO;
1818             break;
1819         } else if (matroska->level_up) {
1820             matroska->level_up--;
1821             break;
1822         }
1823
1824         switch (id) {
1825             /* Hm, this is unsupported... */
1826             default:
1827                 av_log(matroska->ctx, AV_LOG_INFO,
1828                        "Unknown entry 0x%x in metadata header\n", id);
1829                 /* fall-through */
1830
1831             case EBML_ID_VOID:
1832                 res = ebml_read_skip(matroska);
1833                 break;
1834         }
1835
1836         if (matroska->level_up) {
1837             matroska->level_up--;
1838             break;
1839         }
1840     }
1841
1842     return res;
1843 }
1844
1845 static int
1846 matroska_parse_seekhead (MatroskaDemuxContext *matroska)
1847 {
1848     int res = 0;
1849     uint32_t id;
1850
1851     av_log(matroska->ctx, AV_LOG_DEBUG, "parsing seekhead...\n");
1852
1853     while (res == 0) {
1854         if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1855             res = AVERROR_IO;
1856             break;
1857         } else if (matroska->level_up) {
1858             matroska->level_up--;
1859             break;
1860         }
1861
1862         switch (id) {
1863             case MATROSKA_ID_SEEKENTRY: {
1864                 uint32_t seek_id = 0, peek_id_cache = 0;
1865                 uint64_t seek_pos = (uint64_t) -1, t;
1866
1867                 if ((res = ebml_read_master(matroska, &id)) < 0)
1868                     break;
1869
1870                 while (res == 0) {
1871                     if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1872                         res = AVERROR_IO;
1873                         break;
1874                     } else if (matroska->level_up) {
1875                         matroska->level_up--;
1876                         break;
1877                     }
1878
1879                     switch (id) {
1880                         case MATROSKA_ID_SEEKID:
1881                             res = ebml_read_uint(matroska, &id, &t);
1882                             seek_id = t;
1883                             break;
1884
1885                         case MATROSKA_ID_SEEKPOSITION:
1886                             res = ebml_read_uint(matroska, &id, &seek_pos);
1887                             break;
1888
1889                         default:
1890                             av_log(matroska->ctx, AV_LOG_INFO,
1891                                    "Unknown seekhead ID 0x%x\n", id);
1892                             /* fall-through */
1893
1894                         case EBML_ID_VOID:
1895                             res = ebml_read_skip(matroska);
1896                             break;
1897                     }
1898
1899                     if (matroska->level_up) {
1900                         matroska->level_up--;
1901                         break;
1902                     }
1903                 }
1904
1905                 if (!seek_id || seek_pos == (uint64_t) -1) {
1906                     av_log(matroska->ctx, AV_LOG_INFO,
1907                            "Incomplete seekhead entry (0x%x/%"PRIu64")\n",
1908                            seek_id, seek_pos);
1909                     break;
1910                 }
1911
1912                 switch (seek_id) {
1913                     case MATROSKA_ID_CUES:
1914                     case MATROSKA_ID_TAGS: {
1915                         uint32_t level_up = matroska->level_up;
1916                         offset_t before_pos;
1917                         uint64_t length;
1918                         MatroskaLevel level;
1919
1920                         /* remember the peeked ID and the current position */
1921                         peek_id_cache = matroska->peek_id;
1922                         before_pos = url_ftell(&matroska->ctx->pb);
1923
1924                         /* seek */
1925                         if ((res = ebml_read_seek(matroska, seek_pos +
1926                                                matroska->segment_start)) < 0)
1927                             return res;
1928
1929                         /* we don't want to lose our seekhead level, so we add
1930                          * a dummy. This is a crude hack. */
1931                         if (matroska->num_levels == EBML_MAX_DEPTH) {
1932                             av_log(matroska->ctx, AV_LOG_INFO,
1933                                    "Max EBML element depth (%d) reached, "
1934                                    "cannot parse further.\n", EBML_MAX_DEPTH);
1935                             return AVERROR_UNKNOWN;
1936                         }
1937
1938                         level.start = 0;
1939                         level.length = (uint64_t)-1;
1940                         matroska->levels[matroska->num_levels] = level;
1941                         matroska->num_levels++;
1942
1943                         /* check ID */
1944                         if (!(id = ebml_peek_id (matroska,
1945                                                  &matroska->level_up)))
1946                             goto finish;
1947                         if (id != seek_id) {
1948                             av_log(matroska->ctx, AV_LOG_INFO,
1949                                    "We looked for ID=0x%x but got "
1950                                    "ID=0x%x (pos=%"PRIu64")",
1951                                    seek_id, id, seek_pos +
1952                                    matroska->segment_start);
1953                             goto finish;
1954                         }
1955
1956                         /* read master + parse */
1957                         if ((res = ebml_read_master(matroska, &id)) < 0)
1958                             goto finish;
1959                         switch (id) {
1960                             case MATROSKA_ID_CUES:
1961                                 if (!(res = matroska_parse_index(matroska)) ||
1962                                     url_feof(&matroska->ctx->pb)) {
1963                                     matroska->index_parsed = 1;
1964                                     res = 0;
1965                                 }
1966                                 break;
1967                             case MATROSKA_ID_TAGS:
1968                                 if (!(res = matroska_parse_metadata(matroska)) ||
1969                                    url_feof(&matroska->ctx->pb)) {
1970                                     matroska->metadata_parsed = 1;
1971                                     res = 0;
1972                                 }
1973                                 break;
1974                         }
1975
1976                     finish:
1977                         /* remove dummy level */
1978                         while (matroska->num_levels) {
1979                             matroska->num_levels--;
1980                             length =
1981                                 matroska->levels[matroska->num_levels].length;
1982                             if (length == (uint64_t)-1)
1983                                 break;
1984                         }
1985
1986                         /* seek back */
1987                         if ((res = ebml_read_seek(matroska, before_pos)) < 0)
1988                             return res;
1989                         matroska->peek_id = peek_id_cache;
1990                         matroska->level_up = level_up;
1991                         break;
1992                     }
1993
1994                     default:
1995                         av_log(matroska->ctx, AV_LOG_INFO,
1996                                "Ignoring seekhead entry for ID=0x%x\n",
1997                                seek_id);
1998                         break;
1999                 }
2000
2001                 break;
2002             }
2003
2004             default:
2005                 av_log(matroska->ctx, AV_LOG_INFO,
2006                        "Unknown seekhead ID 0x%x\n", id);
2007                 /* fall-through */
2008
2009             case EBML_ID_VOID:
2010                 res = ebml_read_skip(matroska);
2011                 break;
2012         }
2013
2014         if (matroska->level_up) {
2015             matroska->level_up--;
2016             break;
2017         }
2018     }
2019
2020     return res;
2021 }
2022
2023 #define ARRAY_SIZE(x)  (sizeof(x)/sizeof(*x))
2024
2025 static int
2026 matroska_aac_profile (char *codec_id)
2027 {
2028     static const char *aac_profiles[] = {
2029         "MAIN", "LC", "SSR"
2030     };
2031     int profile;
2032
2033     for (profile=0; profile<ARRAY_SIZE(aac_profiles); profile++)
2034         if (strstr(codec_id, aac_profiles[profile]))
2035             break;
2036     return profile + 1;
2037 }
2038
2039 static int
2040 matroska_aac_sri (int samplerate)
2041 {
2042     static const int aac_sample_rates[] = {
2043         96000, 88200, 64000, 48000, 44100, 32000,
2044         24000, 22050, 16000, 12000, 11025,  8000,
2045     };
2046     int sri;
2047
2048     for (sri=0; sri<ARRAY_SIZE(aac_sample_rates); sri++)
2049         if (aac_sample_rates[sri] == samplerate)
2050             break;
2051     return sri;
2052 }
2053
2054 static int
2055 matroska_read_header (AVFormatContext    *s,
2056                       AVFormatParameters *ap)
2057 {
2058     MatroskaDemuxContext *matroska = s->priv_data;
2059     char *doctype;
2060     int version, last_level, res = 0;
2061     uint32_t id;
2062
2063     matroska->ctx = s;
2064
2065     /* First read the EBML header. */
2066     doctype = NULL;
2067     if ((res = ebml_read_header(matroska, &doctype, &version)) < 0)
2068         return res;
2069     if ((doctype == NULL) || strcmp(doctype, "matroska")) {
2070         av_log(matroska->ctx, AV_LOG_ERROR,
2071                "Wrong EBML doctype ('%s' != 'matroska').\n",
2072                doctype ? doctype : "(none)");
2073         if (doctype)
2074             av_free(doctype);
2075         return AVERROR_NOFMT;
2076     }
2077     av_free(doctype);
2078     if (version > 2) {
2079         av_log(matroska->ctx, AV_LOG_ERROR,
2080                "Matroska demuxer version 2 too old for file version %d\n",
2081                version);
2082         return AVERROR_NOFMT;
2083     }
2084
2085     /* The next thing is a segment. */
2086     while (1) {
2087         if (!(id = ebml_peek_id(matroska, &last_level)))
2088             return AVERROR_IO;
2089         if (id == MATROSKA_ID_SEGMENT)
2090             break;
2091
2092         /* oi! */
2093         av_log(matroska->ctx, AV_LOG_INFO,
2094                "Expected a Segment ID (0x%x), but received 0x%x!\n",
2095                MATROSKA_ID_SEGMENT, id);
2096         if ((res = ebml_read_skip(matroska)) < 0)
2097             return res;
2098     }
2099
2100     /* We now have a Matroska segment.
2101      * Seeks are from the beginning of the segment,
2102      * after the segment ID/length. */
2103     if ((res = ebml_read_master(matroska, &id)) < 0)
2104         return res;
2105     matroska->segment_start = url_ftell(&s->pb);
2106
2107     matroska->time_scale = 1000000;
2108     /* we've found our segment, start reading the different contents in here */
2109     while (res == 0) {
2110         if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2111             res = AVERROR_IO;
2112             break;
2113         } else if (matroska->level_up) {
2114             matroska->level_up--;
2115             break;
2116         }
2117
2118         switch (id) {
2119             /* stream info */
2120             case MATROSKA_ID_INFO: {
2121                 if ((res = ebml_read_master(matroska, &id)) < 0)
2122                     break;
2123                 res = matroska_parse_info(matroska);
2124                 break;
2125             }
2126
2127             /* track info headers */
2128             case MATROSKA_ID_TRACKS: {
2129                 if ((res = ebml_read_master(matroska, &id)) < 0)
2130                     break;
2131                 res = matroska_parse_tracks(matroska);
2132                 break;
2133             }
2134
2135             /* stream index */
2136             case MATROSKA_ID_CUES: {
2137                 if (!matroska->index_parsed) {
2138                     if ((res = ebml_read_master(matroska, &id)) < 0)
2139                         break;
2140                     res = matroska_parse_index(matroska);
2141                 } else
2142                     res = ebml_read_skip(matroska);
2143                 break;
2144             }
2145
2146             /* metadata */
2147             case MATROSKA_ID_TAGS: {
2148                 if (!matroska->metadata_parsed) {
2149                     if ((res = ebml_read_master(matroska, &id)) < 0)
2150                         break;
2151                     res = matroska_parse_metadata(matroska);
2152                 } else
2153                     res = ebml_read_skip(matroska);
2154                 break;
2155             }
2156
2157             /* file index (if seekable, seek to Cues/Tags to parse it) */
2158             case MATROSKA_ID_SEEKHEAD: {
2159                 if ((res = ebml_read_master(matroska, &id)) < 0)
2160                     break;
2161                 res = matroska_parse_seekhead(matroska);
2162                 break;
2163             }
2164
2165             case MATROSKA_ID_CLUSTER: {
2166                 /* Do not read the master - this will be done in the next
2167                  * call to matroska_read_packet. */
2168                 res = 1;
2169                 break;
2170             }
2171
2172             default:
2173                 av_log(matroska->ctx, AV_LOG_INFO,
2174                        "Unknown matroska file header ID 0x%x\n", id);
2175             /* fall-through */
2176
2177             case EBML_ID_VOID:
2178                 res = ebml_read_skip(matroska);
2179                 break;
2180         }
2181
2182         if (matroska->level_up) {
2183             matroska->level_up--;
2184             break;
2185         }
2186     }
2187
2188     /* Have we found a cluster? */
2189     if (ebml_peek_id(matroska, NULL) == MATROSKA_ID_CLUSTER) {
2190         int i, j;
2191         MatroskaTrack *track;
2192         AVStream *st;
2193
2194         for (i = 0; i < matroska->num_tracks; i++) {
2195             enum CodecID codec_id = CODEC_ID_NONE;
2196             uint8_t *extradata = NULL;
2197             int extradata_size = 0;
2198             int extradata_offset = 0;
2199             track = matroska->tracks[i];
2200
2201             /* libavformat does not really support subtitles.
2202              * Also apply some sanity checks. */
2203             if ((track->type == MATROSKA_TRACK_TYPE_SUBTITLE) ||
2204                 (track->codec_id == NULL))
2205                 continue;
2206
2207             for(j=0; codec_tags[j].str; j++){
2208                 if(!strncmp(codec_tags[j].str, track->codec_id,
2209                             strlen(codec_tags[j].str))){
2210                     codec_id= codec_tags[j].id;
2211                     break;
2212                 }
2213             }
2214
2215             /* Set the FourCC from the CodecID. */
2216             /* This is the MS compatibility mode which stores a
2217              * BITMAPINFOHEADER in the CodecPrivate. */
2218             if (!strcmp(track->codec_id,
2219                         MATROSKA_CODEC_ID_VIDEO_VFW_FOURCC) &&
2220                 (track->codec_priv_size >= 40) &&
2221                 (track->codec_priv != NULL)) {
2222                 unsigned char *p;
2223
2224                 /* Offset of biCompression. Stored in LE. */
2225                 p = (unsigned char *)track->codec_priv + 16;
2226                 ((MatroskaVideoTrack *)track)->fourcc = (p[3] << 24) |
2227                                  (p[2] << 16) | (p[1] << 8) | p[0];
2228                 codec_id = codec_get_id(codec_bmp_tags, ((MatroskaVideoTrack *)track)->fourcc);
2229
2230             }
2231
2232             /* This is the MS compatibility mode which stores a
2233              * WAVEFORMATEX in the CodecPrivate. */
2234             else if (!strcmp(track->codec_id,
2235                              MATROSKA_CODEC_ID_AUDIO_ACM) &&
2236                 (track->codec_priv_size >= 18) &&
2237                 (track->codec_priv != NULL)) {
2238                 unsigned char *p;
2239                 uint16_t tag;
2240
2241                 /* Offset of wFormatTag. Stored in LE. */
2242                 p = (unsigned char *)track->codec_priv;
2243                 tag = (p[1] << 8) | p[0];
2244                 codec_id = codec_get_id(codec_wav_tags, tag);
2245
2246             }
2247
2248             else if (codec_id == CODEC_ID_AAC && !track->codec_priv_size) {
2249                 MatroskaAudioTrack *audiotrack = (MatroskaAudioTrack *) track;
2250                 int profile = matroska_aac_profile(track->codec_id);
2251                 int sri = matroska_aac_sri(audiotrack->internal_samplerate);
2252                 extradata = av_malloc(5);
2253                 if (extradata == NULL)
2254                     return AVERROR_NOMEM;
2255                 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
2256                 extradata[1] = ((sri&0x01) << 7) | (audiotrack->channels<<3);
2257                 if (strstr(track->codec_id, "SBR")) {
2258                     sri = matroska_aac_sri(audiotrack->samplerate);
2259                     extradata[2] = 0x56;
2260                     extradata[3] = 0xE5;
2261                     extradata[4] = 0x80 | (sri<<3);
2262                     extradata_size = 5;
2263                 } else {
2264                     extradata_size = 2;
2265                 }
2266             }
2267
2268             else if (codec_id == CODEC_ID_TTA) {
2269                 MatroskaAudioTrack *audiotrack = (MatroskaAudioTrack *) track;
2270                 ByteIOContext b;
2271                 extradata_size = 30;
2272                 extradata = av_mallocz(extradata_size);
2273                 if (extradata == NULL)
2274                     return AVERROR_NOMEM;
2275                 init_put_byte(&b, extradata, extradata_size, 1,
2276                               NULL, NULL, NULL, NULL);
2277                 put_buffer(&b, (uint8_t *) "TTA1", 4);
2278                 put_le16(&b, 1);
2279                 put_le16(&b, audiotrack->channels);
2280                 put_le16(&b, audiotrack->bitdepth);
2281                 put_le32(&b, audiotrack->samplerate);
2282                 put_le32(&b, matroska->ctx->duration * audiotrack->samplerate);
2283             }
2284
2285             else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
2286                      codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
2287                 extradata_offset = 26;
2288                 track->codec_priv_size -= extradata_offset;
2289                 track->flags |= MATROSKA_TRACK_REAL_V;
2290             }
2291
2292             if (codec_id == CODEC_ID_NONE) {
2293                 av_log(matroska->ctx, AV_LOG_INFO,
2294                        "Unknown/unsupported CodecID %s.\n",
2295                        track->codec_id);
2296             }
2297
2298             track->stream_index = matroska->num_streams;
2299
2300             matroska->num_streams++;
2301             st = av_new_stream(s, track->stream_index);
2302             if (st == NULL)
2303                 return AVERROR_NOMEM;
2304             av_set_pts_info(st, 64, matroska->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
2305
2306             st->codec->codec_id = codec_id;
2307
2308             if (track->default_duration)
2309                 av_reduce(&st->codec->time_base.num, &st->codec->time_base.den,
2310                           track->default_duration, 1000000000, 30000);
2311
2312             if(extradata){
2313                 st->codec->extradata = extradata;
2314                 st->codec->extradata_size = extradata_size;
2315             } else if(track->codec_priv && track->codec_priv_size > 0){
2316                 st->codec->extradata = av_malloc(track->codec_priv_size);
2317                 if(st->codec->extradata == NULL)
2318                     return AVERROR_NOMEM;
2319                 st->codec->extradata_size = track->codec_priv_size;
2320                 memcpy(st->codec->extradata,track->codec_priv+extradata_offset,
2321                        track->codec_priv_size);
2322             }
2323
2324             if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2325                 MatroskaVideoTrack *videotrack = (MatroskaVideoTrack *)track;
2326
2327                 st->codec->codec_type = CODEC_TYPE_VIDEO;
2328                 st->codec->codec_tag = videotrack->fourcc;
2329                 st->codec->width = videotrack->pixel_width;
2330                 st->codec->height = videotrack->pixel_height;
2331                 if (videotrack->display_width == 0)
2332                     videotrack->display_width= videotrack->pixel_width;
2333                 if (videotrack->display_height == 0)
2334                     videotrack->display_height= videotrack->pixel_height;
2335                 av_reduce(&st->codec->sample_aspect_ratio.num,
2336                           &st->codec->sample_aspect_ratio.den,
2337                           st->codec->height * videotrack->display_width,
2338                           st->codec-> width * videotrack->display_height,
2339                           255);
2340             } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2341                 MatroskaAudioTrack *audiotrack = (MatroskaAudioTrack *)track;
2342
2343                 st->codec->codec_type = CODEC_TYPE_AUDIO;
2344                 st->codec->sample_rate = audiotrack->samplerate;
2345                 st->codec->channels = audiotrack->channels;
2346             } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2347                 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
2348             }
2349
2350             /* What do we do with private data? E.g. for Vorbis. */
2351         }
2352         res = 0;
2353     }
2354
2355     return res;
2356 }
2357
2358 static int
2359 matroska_find_track_by_num (MatroskaDemuxContext *matroska,
2360                             int                   num)
2361 {
2362     int i;
2363
2364     for (i = 0; i < matroska->num_tracks; i++)
2365         if (matroska->tracks[i]->num == num)
2366             return i;
2367
2368     return -1;
2369 }
2370
2371 static inline int
2372 rv_offset(uint8_t *data, int slice, int slices)
2373 {
2374     return AV_RL32(data+8*slice+4) + 8*slices;
2375 }
2376
2377 static int
2378 matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data, int size,
2379                      int64_t pos, uint64_t cluster_time,
2380                      int is_keyframe, int is_bframe,
2381                      int *ptrack, AVPacket **ppkt)
2382 {
2383     int res = 0;
2384     int track;
2385     AVPacket *pkt;
2386     uint8_t *origdata = data;
2387     int16_t block_time;
2388     uint32_t *lace_size = NULL;
2389     int n, flags, laces = 0;
2390     uint64_t num;
2391
2392     /* first byte(s): tracknum */
2393     if ((n = matroska_ebmlnum_uint(data, size, &num)) < 0) {
2394         av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2395         av_free(origdata);
2396         return res;
2397     }
2398     data += n;
2399     size -= n;
2400
2401     /* fetch track from num */
2402     track = matroska_find_track_by_num(matroska, num);
2403     if (ptrack)  *ptrack = track;
2404     if (size <= 3 || track < 0 || track >= matroska->num_tracks) {
2405         av_log(matroska->ctx, AV_LOG_INFO,
2406                "Invalid stream %d or size %u\n", track, size);
2407         av_free(origdata);
2408         return res;
2409     }
2410     if(matroska->ctx->streams[ matroska->tracks[track]->stream_index ]->discard >= AVDISCARD_ALL){
2411         av_free(origdata);
2412         return res;
2413     }
2414
2415     /* block_time (relative to cluster time) */
2416     block_time = (data[0] << 8) | data[1];
2417     data += 2;
2418     size -= 2;
2419     flags = *data;
2420     data += 1;
2421     size -= 1;
2422     if (is_keyframe == -1)
2423         is_keyframe = flags & 1 ? PKT_FLAG_KEY : 0;
2424     switch ((flags & 0x06) >> 1) {
2425         case 0x0: /* no lacing */
2426             laces = 1;
2427             lace_size = av_mallocz(sizeof(int));
2428             lace_size[0] = size;
2429             break;
2430
2431         case 0x1: /* xiph lacing */
2432         case 0x2: /* fixed-size lacing */
2433         case 0x3: /* EBML lacing */
2434             if (size == 0) {
2435                 res = -1;
2436                 break;
2437             }
2438             laces = (*data) + 1;
2439             data += 1;
2440             size -= 1;
2441             lace_size = av_mallocz(laces * sizeof(int));
2442
2443             switch ((flags & 0x06) >> 1) {
2444                 case 0x1: /* xiph lacing */ {
2445                     uint8_t temp;
2446                     uint32_t total = 0;
2447                     for (n = 0; res == 0 && n < laces - 1; n++) {
2448                         while (1) {
2449                             if (size == 0) {
2450                                 res = -1;
2451                                 break;
2452                             }
2453                             temp = *data;
2454                             lace_size[n] += temp;
2455                             data += 1;
2456                             size -= 1;
2457                             if (temp != 0xff)
2458                                 break;
2459                         }
2460                         total += lace_size[n];
2461                     }
2462                     lace_size[n] = size - total;
2463                     break;
2464                 }
2465
2466                 case 0x2: /* fixed-size lacing */
2467                     for (n = 0; n < laces; n++)
2468                         lace_size[n] = size / laces;
2469                     break;
2470
2471                 case 0x3: /* EBML lacing */ {
2472                     uint32_t total;
2473                     n = matroska_ebmlnum_uint(data, size, &num);
2474                     if (n < 0) {
2475                         av_log(matroska->ctx, AV_LOG_INFO,
2476                                "EBML block data error\n");
2477                         break;
2478                     }
2479                     data += n;
2480                     size -= n;
2481                     total = lace_size[0] = num;
2482                     for (n = 1; res == 0 && n < laces - 1; n++) {
2483                         int64_t snum;
2484                         int r;
2485                         r = matroska_ebmlnum_sint (data, size, &snum);
2486                         if (r < 0) {
2487                             av_log(matroska->ctx, AV_LOG_INFO,
2488                                    "EBML block data error\n");
2489                             break;
2490                         }
2491                         data += r;
2492                         size -= r;
2493                         lace_size[n] = lace_size[n - 1] + snum;
2494                         total += lace_size[n];
2495                     }
2496                     lace_size[n] = size - total;
2497                     break;
2498                 }
2499             }
2500             break;
2501     }
2502
2503     if (res == 0) {
2504         int real_v = matroska->tracks[track]->flags & MATROSKA_TRACK_REAL_V;
2505         for (n = 0; n < laces; n++) {
2506             uint64_t timecode = AV_NOPTS_VALUE;
2507             int slice, slices = 1;
2508
2509             if (real_v) {
2510                 slices = *data++ + 1;
2511                 lace_size[n]--;
2512             }
2513             if (cluster_time != (uint64_t)-1 && n == 0) {
2514                 if (cluster_time + block_time >= 0)
2515                     timecode = (cluster_time + block_time) * matroska->time_scale;
2516             }
2517             /* FIXME: duration */
2518
2519             for (slice=0; slice<slices; slice++) {
2520                 int slice_size, slice_offset = 0;
2521                 if (real_v)
2522                     slice_offset = rv_offset(data, slice, slices);
2523                 if (slice+1 == slices)
2524                     slice_size = lace_size[n] - slice_offset;
2525                 else
2526                     slice_size = rv_offset(data, slice+1, slices) - slice_offset;
2527                 pkt = av_mallocz(sizeof(AVPacket));
2528                 if (ppkt)  *ppkt = pkt;
2529                 /* XXX: prevent data copy... */
2530                 if (av_new_packet(pkt, slice_size) < 0) {
2531                     res = AVERROR_NOMEM;
2532                     n = laces-1;
2533                     break;
2534                 }
2535                 memcpy (pkt->data, data+slice_offset, slice_size);
2536
2537                 if (n == 0)
2538                     pkt->flags = is_keyframe;
2539                 pkt->stream_index = matroska->tracks[track]->stream_index;
2540
2541                 pkt->pts = timecode;
2542                 pkt->pos = pos;
2543
2544                 matroska_queue_packet(matroska, pkt);
2545             }
2546             data += lace_size[n];
2547         }
2548     }
2549
2550     av_free(lace_size);
2551     av_free(origdata);
2552     return res;
2553 }
2554
2555 static int
2556 matroska_parse_blockgroup (MatroskaDemuxContext *matroska,
2557                            uint64_t              cluster_time)
2558 {
2559     int res = 0;
2560     uint32_t id;
2561     AVPacket *pkt = NULL;
2562     int is_bframe = 0;
2563     int is_keyframe = PKT_FLAG_KEY, last_num_packets = matroska->num_packets;
2564     uint64_t duration = AV_NOPTS_VALUE;
2565     int track = -1;
2566     uint8_t *data;
2567     int size = 0;
2568     int64_t pos = 0;
2569
2570     av_log(matroska->ctx, AV_LOG_DEBUG, "parsing blockgroup...\n");
2571
2572     while (res == 0) {
2573         if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2574             res = AVERROR_IO;
2575             break;
2576         } else if (matroska->level_up) {
2577             matroska->level_up--;
2578             break;
2579         }
2580
2581         switch (id) {
2582             /* one block inside the group. Note, block parsing is one
2583              * of the harder things, so this code is a bit complicated.
2584              * See http://www.matroska.org/ for documentation. */
2585             case MATROSKA_ID_BLOCK: {
2586                 pos = url_ftell(&matroska->ctx->pb);
2587                 res = ebml_read_binary(matroska, &id, &data, &size);
2588                 break;
2589             }
2590
2591             case MATROSKA_ID_BLOCKDURATION: {
2592                 if ((res = ebml_read_uint(matroska, &id, &duration)) < 0)
2593                     break;
2594                 break;
2595             }
2596
2597             case MATROSKA_ID_BLOCKREFERENCE: {
2598                 int64_t num;
2599                 /* We've found a reference, so not even the first frame in
2600                  * the lace is a key frame. */
2601                 is_keyframe = 0;
2602                 if (last_num_packets != matroska->num_packets)
2603                     matroska->packets[last_num_packets]->flags = 0;
2604                 if ((res = ebml_read_sint(matroska, &id, &num)) < 0)
2605                     break;
2606                 if (num > 0)
2607                     is_bframe = 1;
2608                 break;
2609             }
2610
2611             default:
2612                 av_log(matroska->ctx, AV_LOG_INFO,
2613                        "Unknown entry 0x%x in blockgroup data\n", id);
2614                 /* fall-through */
2615
2616             case EBML_ID_VOID:
2617                 res = ebml_read_skip(matroska);
2618                 break;
2619         }
2620
2621         if (matroska->level_up) {
2622             matroska->level_up--;
2623             break;
2624         }
2625     }
2626
2627     if (res)
2628         return res;
2629
2630     if (size > 0)
2631         res = matroska_parse_block(matroska, data, size, pos, cluster_time,
2632                                    is_keyframe, is_bframe, &track, &pkt);
2633
2634     if (pkt)
2635     {
2636         if (duration != AV_NOPTS_VALUE)
2637             pkt->duration = duration;
2638         else if (track >= 0 && track < matroska->num_tracks)
2639             pkt->duration = matroska->tracks[track]->default_duration / matroska->time_scale;
2640     }
2641
2642     return res;
2643 }
2644
2645 static int
2646 matroska_parse_cluster (MatroskaDemuxContext *matroska)
2647 {
2648     int res = 0;
2649     uint32_t id;
2650     uint64_t cluster_time = 0;
2651     uint8_t *data;
2652     int64_t pos;
2653     int size;
2654
2655     av_log(matroska->ctx, AV_LOG_DEBUG,
2656            "parsing cluster at %"PRId64"\n", url_ftell(&matroska->ctx->pb));
2657
2658     while (res == 0) {
2659         if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2660             res = AVERROR_IO;
2661             break;
2662         } else if (matroska->level_up) {
2663             matroska->level_up--;
2664             break;
2665         }
2666
2667         switch (id) {
2668             /* cluster timecode */
2669             case MATROSKA_ID_CLUSTERTIMECODE: {
2670                 uint64_t num;
2671                 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
2672                     break;
2673                 cluster_time = num;
2674                 break;
2675             }
2676
2677                 /* a group of blocks inside a cluster */
2678             case MATROSKA_ID_BLOCKGROUP:
2679                 if ((res = ebml_read_master(matroska, &id)) < 0)
2680                     break;
2681                 res = matroska_parse_blockgroup(matroska, cluster_time);
2682                 break;
2683
2684             case MATROSKA_ID_SIMPLEBLOCK:
2685                 pos = url_ftell(&matroska->ctx->pb);
2686                 res = ebml_read_binary(matroska, &id, &data, &size);
2687                 if (res == 0)
2688                     res = matroska_parse_block(matroska, data, size, pos,
2689                                                cluster_time, -1,0, NULL, NULL);
2690                 break;
2691
2692             default:
2693                 av_log(matroska->ctx, AV_LOG_INFO,
2694                        "Unknown entry 0x%x in cluster data\n", id);
2695                 /* fall-through */
2696
2697             case EBML_ID_VOID:
2698                 res = ebml_read_skip(matroska);
2699                 break;
2700         }
2701
2702         if (matroska->level_up) {
2703             matroska->level_up--;
2704             break;
2705         }
2706     }
2707
2708     return res;
2709 }
2710
2711 static int
2712 matroska_read_packet (AVFormatContext *s,
2713                       AVPacket        *pkt)
2714 {
2715     MatroskaDemuxContext *matroska = s->priv_data;
2716     int res = 0;
2717     uint32_t id;
2718
2719     /* Read stream until we have a packet queued. */
2720     while (matroska_deliver_packet(matroska, pkt)) {
2721
2722         /* Have we already reached the end? */
2723         if (matroska->done)
2724             return AVERROR_IO;
2725
2726         while (res == 0) {
2727             if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2728                 return AVERROR_IO;
2729             } else if (matroska->level_up) {
2730                 matroska->level_up--;
2731                 break;
2732             }
2733
2734             switch (id) {
2735                 case MATROSKA_ID_CLUSTER:
2736                     if ((res = ebml_read_master(matroska, &id)) < 0)
2737                         break;
2738                     if ((res = matroska_parse_cluster(matroska)) == 0)
2739                         res = 1; /* Parsed one cluster, let's get out. */
2740                     break;
2741
2742                 default:
2743                 case EBML_ID_VOID:
2744                     res = ebml_read_skip(matroska);
2745                     break;
2746             }
2747
2748             if (matroska->level_up) {
2749                 matroska->level_up--;
2750                 break;
2751             }
2752         }
2753
2754         if (res == -1)
2755             matroska->done = 1;
2756     }
2757
2758     return 0;
2759 }
2760
2761 static int
2762 matroska_read_close (AVFormatContext *s)
2763 {
2764     MatroskaDemuxContext *matroska = s->priv_data;
2765     int n = 0;
2766
2767     av_free(matroska->writing_app);
2768     av_free(matroska->muxing_app);
2769     av_free(matroska->index);
2770
2771     if (matroska->packets != NULL) {
2772         for (n = 0; n < matroska->num_packets; n++) {
2773             av_free_packet(matroska->packets[n]);
2774             av_free(matroska->packets[n]);
2775         }
2776         av_free(matroska->packets);
2777     }
2778
2779     for (n = 0; n < matroska->num_tracks; n++) {
2780         MatroskaTrack *track = matroska->tracks[n];
2781         av_free(track->codec_id);
2782         av_free(track->codec_name);
2783         av_free(track->codec_priv);
2784         av_free(track->name);
2785         av_free(track->language);
2786
2787         av_free(track);
2788     }
2789
2790     return 0;
2791 }
2792
2793 AVInputFormat matroska_demuxer = {
2794     "matroska",
2795     "Matroska file format",
2796     sizeof(MatroskaDemuxContext),
2797     matroska_probe,
2798     matroska_read_header,
2799     matroska_read_packet,
2800     matroska_read_close,
2801 };