add missing #includes
[libav.git] / libavformat / matroska.c
1 /*
2 * Matroska file demuxer (no muxer yet)
3 * Copyright (c) 2003-2004 The ffmpeg Project
4 *
5 * This library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2 of the License, or (at your option) any later version.
9 *
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
14 *
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with this library; if not, write to the Free Software
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 */
19
20 /**
21 * @file matroska.c
22 * Matroska file demuxer
23 * by Ronald Bultje <rbultje@ronald.bitfreak.net>
24 * with a little help from Moritz Bunkus <moritz@bunkus.org>
25 * Specs available on the matroska project page:
26 * http://www.matroska.org/.
27 */
28
29 #include "avformat.h"
30 /* For codec_get_bmp_id and codec_get_wav_id. */
31 #include "avi.h"
32 #include "intfloat_readwrite.h"
33
34 /* EBML version supported */
35 #define EBML_VERSION 1
36
37 /* top-level master-IDs */
38 #define EBML_ID_HEADER 0x1A45DFA3
39
40 /* IDs in the HEADER master */
41 #define EBML_ID_EBMLVERSION 0x4286
42 #define EBML_ID_EBMLREADVERSION 0x42F7
43 #define EBML_ID_EBMLMAXIDLENGTH 0x42F2
44 #define EBML_ID_EBMLMAXSIZELENGTH 0x42F3
45 #define EBML_ID_DOCTYPE 0x4282
46 #define EBML_ID_DOCTYPEVERSION 0x4287
47 #define EBML_ID_DOCTYPEREADVERSION 0x4285
48
49 /* general EBML types */
50 #define EBML_ID_VOID 0xEC
51
52 /*
53 * Matroska element IDs. max. 32-bit.
54 */
55
56 /* toplevel segment */
57 #define MATROSKA_ID_SEGMENT 0x18538067
58
59 /* matroska top-level master IDs */
60 #define MATROSKA_ID_INFO 0x1549A966
61 #define MATROSKA_ID_TRACKS 0x1654AE6B
62 #define MATROSKA_ID_CUES 0x1C53BB6B
63 #define MATROSKA_ID_TAGS 0x1254C367
64 #define MATROSKA_ID_SEEKHEAD 0x114D9B74
65 #define MATROSKA_ID_CLUSTER 0x1F43B675
66
67 /* IDs in the info master */
68 #define MATROSKA_ID_TIMECODESCALE 0x2AD7B1
69 #define MATROSKA_ID_DURATION 0x4489
70 #define MATROSKA_ID_WRITINGAPP 0x5741
71 #define MATROSKA_ID_MUXINGAPP 0x4D80
72 #define MATROSKA_ID_DATEUTC 0x4461
73
74 /* ID in the tracks master */
75 #define MATROSKA_ID_TRACKENTRY 0xAE
76
77 /* IDs in the trackentry master */
78 #define MATROSKA_ID_TRACKNUMBER 0xD7
79 #define MATROSKA_ID_TRACKUID 0x73C5
80 #define MATROSKA_ID_TRACKTYPE 0x83
81 #define MATROSKA_ID_TRACKAUDIO 0xE1
82 #define MATROSKA_ID_TRACKVIDEO 0xE0
83 #define MATROSKA_ID_CODECID 0x86
84 #define MATROSKA_ID_CODECPRIVATE 0x63A2
85 #define MATROSKA_ID_CODECNAME 0x258688
86 #define MATROSKA_ID_CODECINFOURL 0x3B4040
87 #define MATROSKA_ID_CODECDOWNLOADURL 0x26B240
88 #define MATROSKA_ID_TRACKNAME 0x536E
89 #define MATROSKA_ID_TRACKLANGUAGE 0x22B59C
90 #define MATROSKA_ID_TRACKFLAGENABLED 0xB9
91 #define MATROSKA_ID_TRACKFLAGDEFAULT 0x88
92 #define MATROSKA_ID_TRACKFLAGLACING 0x9C
93 #define MATROSKA_ID_TRACKMINCACHE 0x6DE7
94 #define MATROSKA_ID_TRACKMAXCACHE 0x6DF8
95 #define MATROSKA_ID_TRACKDEFAULTDURATION 0x23E383
96
97 /* IDs in the trackvideo master */
98 #define MATROSKA_ID_VIDEOFRAMERATE 0x2383E3
99 #define MATROSKA_ID_VIDEODISPLAYWIDTH 0x54B0
100 #define MATROSKA_ID_VIDEODISPLAYHEIGHT 0x54BA
101 #define MATROSKA_ID_VIDEOPIXELWIDTH 0xB0
102 #define MATROSKA_ID_VIDEOPIXELHEIGHT 0xBA
103 #define MATROSKA_ID_VIDEOFLAGINTERLACED 0x9A
104 #define MATROSKA_ID_VIDEOSTEREOMODE 0x53B9
105 #define MATROSKA_ID_VIDEOASPECTRATIO 0x54B3
106 #define MATROSKA_ID_VIDEOCOLOURSPACE 0x2EB524
107
108 /* IDs in the trackaudio master */
109 #define MATROSKA_ID_AUDIOSAMPLINGFREQ 0xB5
110 #define MATROSKA_ID_AUDIOBITDEPTH 0x6264
111 #define MATROSKA_ID_AUDIOCHANNELS 0x9F
112
113 /* ID in the cues master */
114 #define MATROSKA_ID_POINTENTRY 0xBB
115
116 /* IDs in the pointentry master */
117 #define MATROSKA_ID_CUETIME 0xB3
118 #define MATROSKA_ID_CUETRACKPOSITION 0xB7
119
120 /* IDs in the cuetrackposition master */
121 #define MATROSKA_ID_CUETRACK 0xF7
122 #define MATROSKA_ID_CUECLUSTERPOSITION 0xF1
123
124 /* IDs in the tags master */
125 /* TODO */
126
127 /* IDs in the seekhead master */
128 #define MATROSKA_ID_SEEKENTRY 0x4DBB
129
130 /* IDs in the seekpoint master */
131 #define MATROSKA_ID_SEEKID 0x53AB
132 #define MATROSKA_ID_SEEKPOSITION 0x53AC
133
134 /* IDs in the cluster master */
135 #define MATROSKA_ID_CLUSTERTIMECODE 0xE7
136 #define MATROSKA_ID_BLOCKGROUP 0xA0
137
138 /* IDs in the blockgroup master */
139 #define MATROSKA_ID_BLOCK 0xA1
140 #define MATROSKA_ID_BLOCKDURATION 0x9B
141 #define MATROSKA_ID_BLOCKREFERENCE 0xFB
142
143 typedef enum {
144 MATROSKA_TRACK_TYPE_VIDEO = 0x1,
145 MATROSKA_TRACK_TYPE_AUDIO = 0x2,
146 MATROSKA_TRACK_TYPE_COMPLEX = 0x3,
147 MATROSKA_TRACK_TYPE_LOGO = 0x10,
148 MATROSKA_TRACK_TYPE_SUBTITLE = 0x11,
149 MATROSKA_TRACK_TYPE_CONTROL = 0x20,
150 } MatroskaTrackType;
151
152 typedef enum {
153 MATROSKA_EYE_MODE_MONO = 0x0,
154 MATROSKA_EYE_MODE_RIGHT = 0x1,
155 MATROSKA_EYE_MODE_LEFT = 0x2,
156 MATROSKA_EYE_MODE_BOTH = 0x3,
157 } MatroskaEyeMode;
158
159 typedef enum {
160 MATROSKA_ASPECT_RATIO_MODE_FREE = 0x0,
161 MATROSKA_ASPECT_RATIO_MODE_KEEP = 0x1,
162 MATROSKA_ASPECT_RATIO_MODE_FIXED = 0x2,
163 } MatroskaAspectRatioMode;
164
165 /*
166 * These aren't in any way "matroska-form" things,
167 * it's just something I use in the muxer/demuxer.
168 */
169
170 typedef enum {
171 MATROSKA_TRACK_ENABLED = (1<<0),
172 MATROSKA_TRACK_DEFAULT = (1<<1),
173 MATROSKA_TRACK_LACING = (1<<2),
174 MATROSKA_TRACK_SHIFT = (1<<16)
175 } MatroskaTrackFlags;
176
177 typedef enum {
178 MATROSKA_VIDEOTRACK_INTERLACED = (MATROSKA_TRACK_SHIFT<<0)
179 } MatroskaVideoTrackFlags;
180
181 /*
182 * Matroska Codec IDs. Strings.
183 */
184
185 #define MATROSKA_CODEC_ID_VIDEO_VFW_FOURCC "V_MS/VFW/FOURCC"
186 #define MATROSKA_CODEC_ID_VIDEO_UNCOMPRESSED "V_UNCOMPRESSED"
187 #define MATROSKA_CODEC_ID_VIDEO_MPEG4_SP "V_MPEG4/ISO/SP"
188 #define MATROSKA_CODEC_ID_VIDEO_MPEG4_ASP "V_MPEG4/ISO/ASP"
189 #define MATROSKA_CODEC_ID_VIDEO_MPEG4_AP "V_MPEG4/ISO/AP"
190 #define MATROSKA_CODEC_ID_VIDEO_MPEG4_AVC "V_MPEG4/ISO/AVC"
191 #define MATROSKA_CODEC_ID_VIDEO_MSMPEG4V3 "V_MPEG4/MS/V3"
192 #define MATROSKA_CODEC_ID_VIDEO_MPEG1 "V_MPEG1"
193 #define MATROSKA_CODEC_ID_VIDEO_MPEG2 "V_MPEG2"
194 #define MATROSKA_CODEC_ID_VIDEO_MJPEG "V_MJPEG"
195 /* TODO: Real/Quicktime */
196
197 #define MATROSKA_CODEC_ID_AUDIO_ACM "A_MS/ACM"
198 #define MATROSKA_CODEC_ID_AUDIO_MPEG1_L1 "A_MPEG/L1"
199 #define MATROSKA_CODEC_ID_AUDIO_MPEG1_L2 "A_MPEG/L2"
200 #define MATROSKA_CODEC_ID_AUDIO_MPEG1_L3 "A_MPEG/L3"
201 #define MATROSKA_CODEC_ID_AUDIO_PCM_INT_BE "A_PCM/INT/BIG"
202 #define MATROSKA_CODEC_ID_AUDIO_PCM_INT_LE "A_PCM/INT/LIT"
203 #define MATROSKA_CODEC_ID_AUDIO_PCM_FLOAT "A_PCM/FLOAT/IEEE"
204 #define MATROSKA_CODEC_ID_AUDIO_AC3 "A_AC3"
205 #define MATROSKA_CODEC_ID_AUDIO_DTS "A_DTS"
206 #define MATROSKA_CODEC_ID_AUDIO_VORBIS "A_VORBIS"
207 #define MATROSKA_CODEC_ID_AUDIO_ACM "A_MS/ACM"
208 #define MATROSKA_CODEC_ID_AUDIO_MPEG2 "A_AAC/MPEG2/"
209 #define MATROSKA_CODEC_ID_AUDIO_MPEG4 "A_AAC/MPEG4/"
210 /* TODO: AC3-9/10 (?), Real, Musepack, Quicktime */
211
212 /* max. depth in the EBML tree structure */
213 #define EBML_MAX_DEPTH 16
214
215 typedef struct Track {
216 MatroskaTrackType type;
217
218 /* Unique track number and track ID. stream_index is the index that
219 * the calling app uses for this track. */
220 uint32_t num,
221 uid,
222 stream_index;
223
224 char *name,
225 *language;
226
227 char *codec_id,
228 *codec_name;
229
230 unsigned char *codec_priv;
231 int codec_priv_size;
232
233 int64_t default_duration;
234 MatroskaTrackFlags flags;
235 } MatroskaTrack;
236
237 typedef struct MatroskaVideoTrack {
238 MatroskaTrack track;
239
240 int pixel_width,
241 pixel_height,
242 display_width,
243 display_height;
244
245 uint32_t fourcc;
246
247 MatroskaAspectRatioMode ar_mode;
248 MatroskaEyeMode eye_mode;
249
250 //..
251 } MatroskaVideoTrack;
252
253 typedef struct MatroskaAudioTrack {
254 MatroskaTrack track;
255
256 int channels,
257 bitdepth,
258 samplerate;
259 //..
260 } MatroskaAudioTrack;
261
262 typedef struct MatroskaSubtitleTrack {
263 MatroskaTrack track;
264
265 //..
266 } MatroskaSubtitleTrack;
267
268 typedef struct MatroskaLevel {
269 uint64_t start, length;
270 } MatroskaLevel;
271
272 typedef struct MatroskaDemuxIndex {
273 uint64_t pos; /* of the corresponding *cluster*! */
274 uint16_t track; /* reference to 'num' */
275 uint64_t time; /* in nanoseconds */
276 } MatroskaDemuxIndex;
277
278 typedef struct MatroskaDemuxContext {
279 AVFormatContext *ctx;
280
281 /* ebml stuff */
282 int num_levels;
283 MatroskaLevel levels[EBML_MAX_DEPTH];
284 int level_up;
285
286 /* matroska stuff */
287 char *writing_app,
288 *muxing_app;
289 int64_t created;
290
291 /* timescale in the file */
292 int64_t time_scale;
293
294 /* length, position (time, ns) */
295 int64_t duration,
296 pos;
297
298 /* num_streams is the number of streams that av_new_stream() was called
299 * for ( = that are available to the calling program). */
300 int num_tracks, num_streams;
301 MatroskaTrack *tracks[MAX_STREAMS];
302
303 /* cache for ID peeking */
304 uint32_t peek_id;
305
306 /* byte position of the segment inside the stream */
307 offset_t segment_start;
308
309 /* The packet queue. */
310 AVPacket **packets;
311 int num_packets;
312
313 /* have we already parse metadata/cues/clusters? */
314 int metadata_parsed,
315 index_parsed,
316 done;
317
318 /* The index for seeking. */
319 int num_indexes;
320 MatroskaDemuxIndex *index;
321 } MatroskaDemuxContext;
322
323 /*
324 * The first few functions handle EBML file parsing. The rest
325 * is the document interpretation. Matroska really just is a
326 * EBML file.
327 */
328
329 /*
330 * Return: the amount of levels in the hierarchy that the
331 * current element lies higher than the previous one.
332 * The opposite isn't done - that's auto-done using master
333 * element reading.
334 */
335
336 static int
337 ebml_read_element_level_up (MatroskaDemuxContext *matroska)
338 {
339 ByteIOContext *pb = &matroska->ctx->pb;
340 offset_t pos = url_ftell(pb);
341 int num = 0;
342
343 while (matroska->num_levels > 0) {
344 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
345
346 if (pos >= level->start + level->length) {
347 matroska->num_levels--;
348 num++;
349 } else {
350 break;
351 }
352 }
353
354 return num;
355 }
356
357 /*
358 * Read: an "EBML number", which is defined as a variable-length
359 * array of bytes. The first byte indicates the length by giving a
360 * number of 0-bits followed by a one. The position of the first
361 * "one" bit inside the first byte indicates the length of this
362 * number.
363 * Returns: num. of bytes read. < 0 on error.
364 */
365
366 static int
367 ebml_read_num (MatroskaDemuxContext *matroska,
368 int max_size,
369 uint64_t *number)
370 {
371 ByteIOContext *pb = &matroska->ctx->pb;
372 int len_mask = 0x80, read = 1, n = 1;
373 int64_t total = 0;
374
375 /* the first byte tells us the length in bytes - get_byte() can normally
376 * return 0, but since that's not a valid first ebmlID byte, we can
377 * use it safely here to catch EOS. */
378 if (!(total = get_byte(pb))) {
379 /* we might encounter EOS here */
380 if (!url_feof(pb)) {
381 offset_t pos = url_ftell(pb);
382 av_log(matroska->ctx, AV_LOG_ERROR,
383 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
384 pos, pos);
385 }
386 return AVERROR_IO; /* EOS or actual I/O error */
387 }
388
389 /* get the length of the EBML number */
390 while (read <= max_size && !(total & len_mask)) {
391 read++;
392 len_mask >>= 1;
393 }
394 if (read > max_size) {
395 offset_t pos = url_ftell(pb) - 1;
396 av_log(matroska->ctx, AV_LOG_ERROR,
397 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
398 (uint8_t) total, pos, pos);
399 return AVERROR_INVALIDDATA;
400 }
401
402 /* read out length */
403 total &= ~len_mask;
404 while (n++ < read)
405 total = (total << 8) | get_byte(pb);
406
407 *number = total;
408
409 return read;
410 }
411
412 /*
413 * Read: the element content data ID.
414 * Return: the number of bytes read or < 0 on error.
415 */
416
417 static int
418 ebml_read_element_id (MatroskaDemuxContext *matroska,
419 uint32_t *id,
420 int *level_up)
421 {
422 int read;
423 uint64_t total;
424
425 /* if we re-call this, use our cached ID */
426 if (matroska->peek_id != 0) {
427 if (level_up)
428 *level_up = 0;
429 *id = matroska->peek_id;
430 return 0;
431 }
432
433 /* read out the "EBML number", include tag in ID */
434 if ((read = ebml_read_num(matroska, 4, &total)) < 0)
435 return read;
436 *id = matroska->peek_id = total | (1 << (read * 7));
437
438 /* level tracking */
439 if (level_up)
440 *level_up = ebml_read_element_level_up(matroska);
441
442 return read;
443 }
444
445 /*
446 * Read: element content length.
447 * Return: the number of bytes read or < 0 on error.
448 */
449
450 static int
451 ebml_read_element_length (MatroskaDemuxContext *matroska,
452 uint64_t *length)
453 {
454 /* clear cache since we're now beyond that data point */
455 matroska->peek_id = 0;
456
457 /* read out the "EBML number", include tag in ID */
458 return ebml_read_num(matroska, 8, length);
459 }
460
461 /*
462 * Return: the ID of the next element, or 0 on error.
463 * Level_up contains the amount of levels that this
464 * next element lies higher than the previous one.
465 */
466
467 static uint32_t
468 ebml_peek_id (MatroskaDemuxContext *matroska,
469 int *level_up)
470 {
471 uint32_t id;
472
473 assert(level_up != NULL);
474
475 if (ebml_read_element_id(matroska, &id, level_up) < 0)
476 return 0;
477
478 return id;
479 }
480
481 /*
482 * Seek to a given offset.
483 * 0 is success, -1 is failure.
484 */
485
486 static int
487 ebml_read_seek (MatroskaDemuxContext *matroska,
488 offset_t offset)
489 {
490 ByteIOContext *pb = &matroska->ctx->pb;
491
492 /* clear ID cache, if any */
493 matroska->peek_id = 0;
494
495 return (url_fseek(pb, offset, SEEK_SET) == offset) ? 0 : -1;
496 }
497
498 /*
499 * Skip the next element.
500 * 0 is success, -1 is failure.
501 */
502
503 static int
504 ebml_read_skip (MatroskaDemuxContext *matroska)
505 {
506 ByteIOContext *pb = &matroska->ctx->pb;
507 uint32_t id;
508 uint64_t length;
509 int res;
510
511 if ((res = ebml_read_element_id(matroska, &id, NULL)) < 0 ||
512 (res = ebml_read_element_length(matroska, &length)) < 0)
513 return res;
514
515 url_fskip(pb, length);
516
517 return 0;
518 }
519
520 /*
521 * Read the next element as an unsigned int.
522 * 0 is success, < 0 is failure.
523 */
524
525 static int
526 ebml_read_uint (MatroskaDemuxContext *matroska,
527 uint32_t *id,
528 uint64_t *num)
529 {
530 ByteIOContext *pb = &matroska->ctx->pb;
531 int n = 0, size, res;
532 uint64_t rlength;
533
534 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
535 (res = ebml_read_element_length(matroska, &rlength)) < 0)
536 return res;
537 size = rlength;
538 if (size < 1 || size > 8) {
539 offset_t pos = url_ftell(pb);
540 av_log(matroska->ctx, AV_LOG_ERROR,
541 "Invalid uint element size %d at position %"PRId64" (0x%"PRIx64")\n",
542 size, pos, pos);
543 return AVERROR_INVALIDDATA;
544 }
545
546 /* big-endian ordening; build up number */
547 *num = 0;
548 while (n++ < size)
549 *num = (*num << 8) | get_byte(pb);
550
551 return 0;
552 }
553
554 /*
555 * Read the next element as a signed int.
556 * 0 is success, < 0 is failure.
557 */
558
559 static int
560 ebml_read_sint (MatroskaDemuxContext *matroska,
561 uint32_t *id,
562 int64_t *num)
563 {
564 ByteIOContext *pb = &matroska->ctx->pb;
565 int size, n = 1, negative = 0, res;
566 uint64_t rlength;
567
568 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
569 (res = ebml_read_element_length(matroska, &rlength)) < 0)
570 return res;
571 size = rlength;
572 if (size < 1 || size > 8) {
573 offset_t pos = url_ftell(pb);
574 av_log(matroska->ctx, AV_LOG_ERROR,
575 "Invalid sint element size %d at position %"PRId64" (0x%"PRIx64")\n",
576 size, pos, pos);
577 return AVERROR_INVALIDDATA;
578 }
579 if ((*num = get_byte(pb)) & 0x80) {
580 negative = 1;
581 *num &= ~0x80;
582 }
583 *num = 0;
584 while (n++ < size)
585 *num = (*num << 8) | get_byte(pb);
586
587 /* make signed */
588 if (negative)
589 *num = *num - (1LL << ((8 * size) - 1));
590
591 return 0;
592 }
593
594 /*
595 * Read the next element as a float.
596 * 0 is success, < 0 is failure.
597 */
598
599 static int
600 ebml_read_float (MatroskaDemuxContext *matroska,
601 uint32_t *id,
602 double *num)
603 {
604 ByteIOContext *pb = &matroska->ctx->pb;
605 int size, res;
606 uint64_t rlength;
607
608 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
609 (res = ebml_read_element_length(matroska, &rlength)) < 0)
610 return res;
611 size = rlength;
612
613 if (size == 4) {
614 *num= av_int2flt(get_be32(pb));
615 } else if(size==8){
616 *num= av_int2dbl(get_be64(pb));
617 } else if(size==10){
618 av_log(matroska->ctx, AV_LOG_ERROR,
619 "FIXME! 10-byte floats unimplemented\n");
620 return AVERROR_UNKNOWN;
621 } else{
622 offset_t pos = url_ftell(pb);
623 av_log(matroska->ctx, AV_LOG_ERROR,
624 "Invalid float element size %d at position %"PRIu64" (0x%"PRIx64")\n",
625 size, pos, pos);
626 return AVERROR_INVALIDDATA;
627 }
628
629 return 0;
630 }
631
632 /*
633 * Read the next element as an ASCII string.
634 * 0 is success, < 0 is failure.
635 */
636
637 static int
638 ebml_read_ascii (MatroskaDemuxContext *matroska,
639 uint32_t *id,
640 char **str)
641 {
642 ByteIOContext *pb = &matroska->ctx->pb;
643 int size, res;
644 uint64_t rlength;
645
646 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
647 (res = ebml_read_element_length(matroska, &rlength)) < 0)
648 return res;
649 size = rlength;
650
651 /* ebml strings are usually not 0-terminated, so we allocate one
652 * byte more, read the string and NULL-terminate it ourselves. */
653 if (size < 0 || !(*str = av_malloc(size + 1))) {
654 av_log(matroska->ctx, AV_LOG_ERROR, "Memory allocation failed\n");
655 return AVERROR_NOMEM;
656 }
657 if (get_buffer(pb, (uint8_t *) *str, size) != size) {
658 offset_t pos = url_ftell(pb);
659 av_log(matroska->ctx, AV_LOG_ERROR,
660 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
661 return AVERROR_IO;
662 }
663 (*str)[size] = '\0';
664
665 return 0;
666 }
667
668 /*
669 * Read the next element as a UTF-8 string.
670 * 0 is success, < 0 is failure.
671 */
672
673 static int
674 ebml_read_utf8 (MatroskaDemuxContext *matroska,
675 uint32_t *id,
676 char **str)
677 {
678 return ebml_read_ascii(matroska, id, str);
679 }
680
681 /*
682 * Read the next element as a date (nanoseconds since 1/1/2000).
683 * 0 is success, < 0 is failure.
684 */
685
686 static int
687 ebml_read_date (MatroskaDemuxContext *matroska,
688 uint32_t *id,
689 int64_t *date)
690 {
691 return ebml_read_sint(matroska, id, date);
692 }
693
694 /*
695 * Read the next element, but only the header. The contents
696 * are supposed to be sub-elements which can be read separately.
697 * 0 is success, < 0 is failure.
698 */
699
700 static int
701 ebml_read_master (MatroskaDemuxContext *matroska,
702 uint32_t *id)
703 {
704 ByteIOContext *pb = &matroska->ctx->pb;
705 uint64_t length;
706 MatroskaLevel *level;
707 int res;
708
709 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
710 (res = ebml_read_element_length(matroska, &length)) < 0)
711 return res;
712
713 /* protect... (Heaven forbids that the '>' is true) */
714 if (matroska->num_levels >= EBML_MAX_DEPTH) {
715 av_log(matroska->ctx, AV_LOG_ERROR,
716 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
717 return AVERROR_NOTSUPP;
718 }
719
720 /* remember level */
721 level = &matroska->levels[matroska->num_levels++];
722 level->start = url_ftell(pb);
723 level->length = length;
724
725 return 0;
726 }
727
728 /*
729 * Read the next element as binary data.
730 * 0 is success, < 0 is failure.
731 */
732
733 static int
734 ebml_read_binary (MatroskaDemuxContext *matroska,
735 uint32_t *id,
736 uint8_t **binary,
737 int *size)
738 {
739 ByteIOContext *pb = &matroska->ctx->pb;
740 uint64_t rlength;
741 int res;
742
743 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
744 (res = ebml_read_element_length(matroska, &rlength)) < 0)
745 return res;
746 *size = rlength;
747
748 if (!(*binary = av_malloc(*size))) {
749 av_log(matroska->ctx, AV_LOG_ERROR,
750 "Memory allocation error\n");
751 return AVERROR_NOMEM;
752 }
753
754 if (get_buffer(pb, *binary, *size) != *size) {
755 offset_t pos = url_ftell(pb);
756 av_log(matroska->ctx, AV_LOG_ERROR,
757 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
758 return AVERROR_IO;
759 }
760
761 return 0;
762 }
763
764 /*
765 * Read signed/unsigned "EBML" numbers.
766 * Return: number of bytes processed, < 0 on error.
767 * XXX: use ebml_read_num().
768 */
769
770 static int
771 matroska_ebmlnum_uint (uint8_t *data,
772 uint32_t size,
773 uint64_t *num)
774 {
775 int len_mask = 0x80, read = 1, n = 1, num_ffs = 0;
776 uint64_t total;
777
778 if (size <= 0)
779 return AVERROR_INVALIDDATA;
780
781 total = data[0];
782 while (read <= 8 && !(total & len_mask)) {
783 read++;
784 len_mask >>= 1;
785 }
786 if (read > 8)
787 return AVERROR_INVALIDDATA;
788
789 if ((total &= (len_mask - 1)) == len_mask - 1)
790 num_ffs++;
791 if (size < read)
792 return AVERROR_INVALIDDATA;
793 while (n < read) {
794 if (data[n] == 0xff)
795 num_ffs++;
796 total = (total << 8) | data[n];
797 n++;
798 }
799
800 if (read == num_ffs)
801 *num = (uint64_t)-1;
802 else
803 *num = total;
804
805 return read;
806 }
807
808 /*
809 * Same as above, but signed.
810 */
811
812 static int
813 matroska_ebmlnum_sint (uint8_t *data,
814 uint32_t size,
815 int64_t *num)
816 {
817 uint64_t unum;
818 int res;
819
820 /* read as unsigned number first */
821 if ((res = matroska_ebmlnum_uint(data, size, &unum)) < 0)
822 return res;
823
824 /* make signed (weird way) */
825 if (unum == (uint64_t)-1)
826 *num = INT64_MAX;
827 else
828 *num = unum - ((1LL << ((7 * res) - 1)) - 1);
829
830 return res;
831 }
832
833 /*
834 * Read an EBML header.
835 * 0 is success, < 0 is failure.
836 */
837
838 static int
839 ebml_read_header (MatroskaDemuxContext *matroska,
840 char **doctype,
841 int *version)
842 {
843 uint32_t id;
844 int level_up, res = 0;
845
846 /* default init */
847 if (doctype)
848 *doctype = NULL;
849 if (version)
850 *version = 1;
851
852 if (!(id = ebml_peek_id(matroska, &level_up)) ||
853 level_up != 0 || id != EBML_ID_HEADER) {
854 av_log(matroska->ctx, AV_LOG_ERROR,
855 "This is not an EBML file (id=0x%x/0x%x)\n", id, EBML_ID_HEADER);
856 return AVERROR_INVALIDDATA;
857 }
858 if ((res = ebml_read_master(matroska, &id)) < 0)
859 return res;
860
861 while (res == 0) {
862 if (!(id = ebml_peek_id(matroska, &level_up)))
863 return AVERROR_IO;
864
865 /* end-of-header */
866 if (level_up)
867 break;
868
869 switch (id) {
870 /* is our read version uptodate? */
871 case EBML_ID_EBMLREADVERSION: {
872 uint64_t num;
873
874 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
875 return res;
876 if (num > EBML_VERSION) {
877 av_log(matroska->ctx, AV_LOG_ERROR,
878 "EBML version %"PRIu64" (> %d) is not supported\n",
879 num, EBML_VERSION);
880 return AVERROR_INVALIDDATA;
881 }
882 break;
883 }
884
885 /* we only handle 8 byte lengths at max */
886 case EBML_ID_EBMLMAXSIZELENGTH: {
887 uint64_t num;
888
889 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
890 return res;
891 if (num > sizeof(uint64_t)) {
892 av_log(matroska->ctx, AV_LOG_ERROR,
893 "Integers of size %"PRIu64" (> %zd) not supported\n",
894 num, sizeof(uint64_t));
895 return AVERROR_INVALIDDATA;
896 }
897 break;
898 }
899
900 /* we handle 4 byte IDs at max */
901 case EBML_ID_EBMLMAXIDLENGTH: {
902 uint64_t num;
903
904 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
905 return res;
906 if (num > sizeof(uint32_t)) {
907 av_log(matroska->ctx, AV_LOG_ERROR,
908 "IDs of size %"PRIu64" (> %zu) not supported\n",
909 num, sizeof(uint32_t));
910 return AVERROR_INVALIDDATA;
911 }
912 break;
913 }
914
915 case EBML_ID_DOCTYPE: {
916 char *text;
917
918 if ((res = ebml_read_ascii(matroska, &id, &text)) < 0)
919 return res;
920 if (doctype) {
921 if (*doctype)
922 av_free(*doctype);
923 *doctype = text;
924 } else
925 av_free(text);
926 break;
927 }
928
929 case EBML_ID_DOCTYPEREADVERSION: {
930 uint64_t num;
931
932 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
933 return res;
934 if (version)
935 *version = num;
936 break;
937 }
938
939 default:
940 av_log(matroska->ctx, AV_LOG_INFO,
941 "Unknown data type 0x%x in EBML header", id);
942 /* pass-through */
943
944 case EBML_ID_VOID:
945 /* we ignore these two, as they don't tell us anything we
946 * care about */
947 case EBML_ID_EBMLVERSION:
948 case EBML_ID_DOCTYPEVERSION:
949 res = ebml_read_skip (matroska);
950 break;
951 }
952 }
953
954 return 0;
955 }
956
957 /*
958 * Put one packet in an application-supplied AVPacket struct.
959 * Returns 0 on success or -1 on failure.
960 */
961
962 static int
963 matroska_deliver_packet (MatroskaDemuxContext *matroska,
964 AVPacket *pkt)
965 {
966 if (matroska->num_packets > 0) {
967 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
968 av_free(matroska->packets[0]);
969 if (matroska->num_packets > 1) {
970 memmove(&matroska->packets[0], &matroska->packets[1],
971 (matroska->num_packets - 1) * sizeof(AVPacket *));
972 matroska->packets =
973 av_realloc(matroska->packets, (matroska->num_packets - 1) *
974 sizeof(AVPacket *));
975 } else {
976 av_free(matroska->packets);
977 matroska->packets = NULL;
978 }
979 matroska->num_packets--;
980 return 0;
981 }
982
983 return -1;
984 }
985
986 /*
987 * Put a packet into our internal queue. Will be delivered to the
988 * user/application during the next get_packet() call.
989 */
990
991 static void
992 matroska_queue_packet (MatroskaDemuxContext *matroska,
993 AVPacket *pkt)
994 {
995 matroska->packets =
996 av_realloc(matroska->packets, (matroska->num_packets + 1) *
997 sizeof(AVPacket *));
998 matroska->packets[matroska->num_packets] = pkt;
999 matroska->num_packets++;
1000 }
1001
1002 /*
1003 * Autodetecting...
1004 */
1005
1006 static int
1007 matroska_probe (AVProbeData *p)
1008 {
1009 uint64_t total = 0;
1010 int len_mask = 0x80, size = 1, n = 1;
1011 uint8_t probe_data[] = { 'm', 'a', 't', 'r', 'o', 's', 'k', 'a' };
1012
1013 if (p->buf_size < 5)
1014 return 0;
1015
1016 /* ebml header? */
1017 if ((p->buf[0] << 24 | p->buf[1] << 16 |
1018 p->buf[2] << 8 | p->buf[3]) != EBML_ID_HEADER)
1019 return 0;
1020
1021 /* length of header */
1022 total = p->buf[4];
1023 while (size <= 8 && !(total & len_mask)) {
1024 size++;
1025 len_mask >>= 1;
1026 }
1027 if (size > 8)
1028 return 0;
1029 total &= (len_mask - 1);
1030 while (n < size)
1031 total = (total << 8) | p->buf[4 + n++];
1032
1033 /* does the probe data contain the whole header? */
1034 if (p->buf_size < 4 + size + total)
1035 return 0;
1036
1037 /* the header must contain the document type 'matroska'. For now,
1038 * we don't parse the whole header but simply check for the
1039 * availability of that array of characters inside the header.
1040 * Not fully fool-proof, but good enough. */
1041 for (n = 4 + size; n < 4 + size + total - sizeof(probe_data); n++)
1042 if (!memcmp (&p->buf[n], probe_data, sizeof(probe_data)))
1043 return AVPROBE_SCORE_MAX;
1044
1045 return 0;
1046 }
1047
1048 /*
1049 * From here on, it's all XML-style DTD stuff... Needs no comments.
1050 */
1051
1052 static int
1053 matroska_parse_info (MatroskaDemuxContext *matroska)
1054 {
1055 int res = 0;
1056 uint32_t id;
1057
1058 av_log(matroska->ctx, AV_LOG_DEBUG, "Parsing info...\n");
1059
1060 while (res == 0) {
1061 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1062 res = AVERROR_IO;
1063 break;
1064 } else if (matroska->level_up) {
1065 matroska->level_up--;
1066 break;
1067 }
1068
1069 switch (id) {
1070 /* cluster timecode */
1071 case MATROSKA_ID_TIMECODESCALE: {
1072 uint64_t num;
1073 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1074 break;
1075 matroska->time_scale = num;
1076 break;
1077 }
1078
1079 case MATROSKA_ID_DURATION: {
1080 double num;
1081 if ((res = ebml_read_float(matroska, &id, &num)) < 0)
1082 break;
1083 matroska->duration = num * matroska->time_scale;
1084 break;
1085 }
1086
1087 case MATROSKA_ID_WRITINGAPP: {
1088 char *text;
1089 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1090 break;
1091 matroska->writing_app = text;
1092 break;
1093 }
1094
1095 case MATROSKA_ID_MUXINGAPP: {
1096 char *text;
1097 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1098 break;
1099 matroska->muxing_app = text;
1100 break;
1101 }
1102
1103 case MATROSKA_ID_DATEUTC: {
1104 int64_t time;
1105 if ((res = ebml_read_date(matroska, &id, &time)) < 0)
1106 break;
1107 matroska->created = time;
1108 break;
1109 }
1110
1111 default:
1112 av_log(matroska->ctx, AV_LOG_INFO,
1113 "Unknown entry 0x%x in info header\n", id);
1114 /* fall-through */
1115
1116 case EBML_ID_VOID:
1117 res = ebml_read_skip(matroska);
1118 break;
1119 }
1120
1121 if (matroska->level_up) {
1122 matroska->level_up--;
1123 break;
1124 }
1125 }
1126
1127 return res;
1128 }
1129
1130 static int
1131 matroska_add_stream (MatroskaDemuxContext *matroska)
1132 {
1133 int res = 0;
1134 uint32_t id;
1135 MatroskaTrack *track;
1136
1137 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing track, adding stream..,\n");
1138
1139 /* Allocate a generic track. As soon as we know its type we'll realloc. */
1140 track = av_mallocz(sizeof(MatroskaTrack));
1141 matroska->num_tracks++;
1142
1143 /* start with the master */
1144 if ((res = ebml_read_master(matroska, &id)) < 0)
1145 return res;
1146
1147 /* try reading the trackentry headers */
1148 while (res == 0) {
1149 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1150 res = AVERROR_IO;
1151 break;
1152 } else if (matroska->level_up > 0) {
1153 matroska->level_up--;
1154 break;
1155 }
1156
1157 switch (id) {
1158 /* track number (unique stream ID) */
1159 case MATROSKA_ID_TRACKNUMBER: {
1160 uint64_t num;
1161 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1162 break;
1163 track->num = num;
1164 break;
1165 }
1166
1167 /* track UID (unique identifier) */
1168 case MATROSKA_ID_TRACKUID: {
1169 uint64_t num;
1170 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1171 break;
1172 track->uid = num;
1173 break;
1174 }
1175
1176 /* track type (video, audio, combined, subtitle, etc.) */
1177 case MATROSKA_ID_TRACKTYPE: {
1178 uint64_t num;
1179 if (track->type != 0) {
1180 av_log(matroska->ctx, AV_LOG_INFO,
1181 "More than one tracktype in an entry - skip\n");
1182 break;
1183 }
1184 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1185 break;
1186 track->type = num;
1187
1188 /* ok, so we're actually going to reallocate this thing */
1189 switch (track->type) {
1190 case MATROSKA_TRACK_TYPE_VIDEO:
1191 track = (MatroskaTrack *)
1192 av_realloc(track, sizeof(MatroskaVideoTrack));
1193 break;
1194 case MATROSKA_TRACK_TYPE_AUDIO:
1195 track = (MatroskaTrack *)
1196 av_realloc(track, sizeof(MatroskaAudioTrack));
1197 ((MatroskaAudioTrack *)track)->channels = 1;
1198 ((MatroskaAudioTrack *)track)->samplerate = 8000;
1199 break;
1200 case MATROSKA_TRACK_TYPE_SUBTITLE:
1201 track = (MatroskaTrack *)
1202 av_realloc(track, sizeof(MatroskaSubtitleTrack));
1203 break;
1204 case MATROSKA_TRACK_TYPE_COMPLEX:
1205 case MATROSKA_TRACK_TYPE_LOGO:
1206 case MATROSKA_TRACK_TYPE_CONTROL:
1207 default:
1208 av_log(matroska->ctx, AV_LOG_INFO,
1209 "Unknown or unsupported track type 0x%x\n",
1210 track->type);
1211 track->type = 0;
1212 break;
1213 }
1214 matroska->tracks[matroska->num_tracks - 1] = track;
1215 break;
1216 }
1217
1218 /* tracktype specific stuff for video */
1219 case MATROSKA_ID_TRACKVIDEO: {
1220 MatroskaVideoTrack *videotrack;
1221 if (track->type != MATROSKA_TRACK_TYPE_VIDEO) {
1222 av_log(matroska->ctx, AV_LOG_INFO,
1223 "video data in non-video track - ignoring\n");
1224 res = AVERROR_INVALIDDATA;
1225 break;
1226 } else if ((res = ebml_read_master(matroska, &id)) < 0)
1227 break;
1228 videotrack = (MatroskaVideoTrack *)track;
1229
1230 while (res == 0) {
1231 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1232 res = AVERROR_IO;
1233 break;
1234 } else if (matroska->level_up > 0) {
1235 matroska->level_up--;
1236 break;
1237 }
1238
1239 switch (id) {
1240 /* fixme, this should be one-up, but I get it here */
1241 case MATROSKA_ID_TRACKDEFAULTDURATION: {
1242 uint64_t num;
1243 if ((res = ebml_read_uint (matroska, &id,
1244 &num)) < 0)
1245 break;
1246 track->default_duration = num;
1247 break;
1248 }
1249
1250 /* video framerate */
1251 case MATROSKA_ID_VIDEOFRAMERATE: {
1252 double num;
1253 if ((res = ebml_read_float(matroska, &id,
1254 &num)) < 0)
1255 break;
1256 track->default_duration = 1000000000 * (1. / num);
1257 break;
1258 }
1259
1260 /* width of the size to display the video at */
1261 case MATROSKA_ID_VIDEODISPLAYWIDTH: {
1262 uint64_t num;
1263 if ((res = ebml_read_uint(matroska, &id,
1264 &num)) < 0)
1265 break;
1266 videotrack->display_width = num;
1267 break;
1268 }
1269
1270 /* height of the size to display the video at */
1271 case MATROSKA_ID_VIDEODISPLAYHEIGHT: {
1272 uint64_t num;
1273 if ((res = ebml_read_uint(matroska, &id,
1274 &num)) < 0)
1275 break;
1276 videotrack->display_height = num;
1277 break;
1278 }
1279
1280 /* width of the video in the file */
1281 case MATROSKA_ID_VIDEOPIXELWIDTH: {
1282 uint64_t num;
1283 if ((res = ebml_read_uint(matroska, &id,
1284 &num)) < 0)
1285 break;
1286 videotrack->pixel_width = num;
1287 break;
1288 }
1289
1290 /* height of the video in the file */
1291 case MATROSKA_ID_VIDEOPIXELHEIGHT: {
1292 uint64_t num;
1293 if ((res = ebml_read_uint(matroska, &id,
1294 &num)) < 0)
1295 break;
1296 videotrack->pixel_height = num;
1297 break;
1298 }
1299
1300 /* whether the video is interlaced */
1301 case MATROSKA_ID_VIDEOFLAGINTERLACED: {
1302 uint64_t num;
1303 if ((res = ebml_read_uint(matroska, &id,
1304 &num)) < 0)
1305 break;
1306 if (num)
1307 track->flags |=
1308 MATROSKA_VIDEOTRACK_INTERLACED;
1309 else
1310 track->flags &=
1311 ~MATROSKA_VIDEOTRACK_INTERLACED;
1312 break;
1313 }
1314
1315 /* stereo mode (whether the video has two streams,
1316 * where one is for the left eye and the other for
1317 * the right eye, which creates a 3D-like
1318 * effect) */
1319 case MATROSKA_ID_VIDEOSTEREOMODE: {
1320 uint64_t num;
1321 if ((res = ebml_read_uint(matroska, &id,
1322 &num)) < 0)
1323 break;
1324 if (num != MATROSKA_EYE_MODE_MONO &&
1325 num != MATROSKA_EYE_MODE_LEFT &&
1326 num != MATROSKA_EYE_MODE_RIGHT &&
1327 num != MATROSKA_EYE_MODE_BOTH) {
1328 av_log(matroska->ctx, AV_LOG_INFO,
1329 "Ignoring unknown eye mode 0x%x\n",
1330 (uint32_t) num);
1331 break;
1332 }
1333 videotrack->eye_mode = num;
1334 break;
1335 }
1336
1337 /* aspect ratio behaviour */
1338 case MATROSKA_ID_VIDEOASPECTRATIO: {
1339 uint64_t num;
1340 if ((res = ebml_read_uint(matroska, &id,
1341 &num)) < 0)
1342 break;
1343 if (num != MATROSKA_ASPECT_RATIO_MODE_FREE &&
1344 num != MATROSKA_ASPECT_RATIO_MODE_KEEP &&
1345 num != MATROSKA_ASPECT_RATIO_MODE_FIXED) {
1346 av_log(matroska->ctx, AV_LOG_INFO,
1347 "Ignoring unknown aspect ratio 0x%x\n",
1348 (uint32_t) num);
1349 break;
1350 }
1351 videotrack->ar_mode = num;
1352 break;
1353 }
1354
1355 /* colourspace (only matters for raw video)
1356 * fourcc */
1357 case MATROSKA_ID_VIDEOCOLOURSPACE: {
1358 uint64_t num;
1359 if ((res = ebml_read_uint(matroska, &id,
1360 &num)) < 0)
1361 break;
1362 videotrack->fourcc = num;
1363 break;
1364 }
1365
1366 default:
1367 av_log(matroska->ctx, AV_LOG_INFO,
1368 "Unknown video track header entry "
1369 "0x%x - ignoring\n", id);
1370 /* pass-through */
1371
1372 case EBML_ID_VOID:
1373 res = ebml_read_skip(matroska);
1374 break;
1375 }
1376
1377 if (matroska->level_up) {
1378 matroska->level_up--;
1379 break;
1380 }
1381 }
1382 break;
1383 }
1384
1385 /* tracktype specific stuff for audio */
1386 case MATROSKA_ID_TRACKAUDIO: {
1387 MatroskaAudioTrack *audiotrack;
1388 if (track->type != MATROSKA_TRACK_TYPE_AUDIO) {
1389 av_log(matroska->ctx, AV_LOG_INFO,
1390 "audio data in non-audio track - ignoring\n");
1391 res = AVERROR_INVALIDDATA;
1392 break;
1393 } else if ((res = ebml_read_master(matroska, &id)) < 0)
1394 break;
1395 audiotrack = (MatroskaAudioTrack *)track;
1396
1397 while (res == 0) {
1398 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1399 res = AVERROR_IO;
1400 break;
1401 } else if (matroska->level_up > 0) {
1402 matroska->level_up--;
1403 break;
1404 }
1405
1406 switch (id) {
1407 /* samplerate */
1408 case MATROSKA_ID_AUDIOSAMPLINGFREQ: {
1409 double num;
1410 if ((res = ebml_read_float(matroska, &id,
1411 &num)) < 0)
1412 break;
1413 audiotrack->samplerate = num;
1414 break;
1415 }
1416
1417 /* bitdepth */
1418 case MATROSKA_ID_AUDIOBITDEPTH: {
1419 uint64_t num;
1420 if ((res = ebml_read_uint(matroska, &id,
1421 &num)) < 0)
1422 break;
1423 audiotrack->bitdepth = num;
1424 break;
1425 }
1426
1427 /* channels */
1428 case MATROSKA_ID_AUDIOCHANNELS: {
1429 uint64_t num;
1430 if ((res = ebml_read_uint(matroska, &id,
1431 &num)) < 0)
1432 break;
1433 audiotrack->channels = num;
1434 break;
1435 }
1436
1437 default:
1438 av_log(matroska->ctx, AV_LOG_INFO,
1439 "Unknown audio track header entry "
1440 "0x%x - ignoring\n", id);
1441 /* pass-through */
1442
1443 case EBML_ID_VOID:
1444 res = ebml_read_skip(matroska);
1445 break;
1446 }
1447
1448 if (matroska->level_up) {
1449 matroska->level_up--;
1450 break;
1451 }
1452 }
1453 break;
1454 }
1455
1456 /* codec identifier */
1457 case MATROSKA_ID_CODECID: {
1458 char *text;
1459 if ((res = ebml_read_ascii(matroska, &id, &text)) < 0)
1460 break;
1461 track->codec_id = text;
1462 break;
1463 }
1464
1465 /* codec private data */
1466 case MATROSKA_ID_CODECPRIVATE: {
1467 uint8_t *data;
1468 int size;
1469 if ((res = ebml_read_binary(matroska, &id, &data, &size) < 0))
1470 break;
1471 track->codec_priv = data;
1472 track->codec_priv_size = size;
1473 break;
1474 }
1475
1476 /* name of the codec */
1477 case MATROSKA_ID_CODECNAME: {
1478 char *text;
1479 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1480 break;
1481 track->codec_name = text;
1482 break;
1483 }
1484
1485 /* name of this track */
1486 case MATROSKA_ID_TRACKNAME: {
1487 char *text;
1488 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1489 break;
1490 track->name = text;
1491 break;
1492 }
1493
1494 /* language (matters for audio/subtitles, mostly) */
1495 case MATROSKA_ID_TRACKLANGUAGE: {
1496 char *text;
1497 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1498 break;
1499 track->language = text;
1500 break;
1501 }
1502
1503 /* whether this is actually used */
1504 case MATROSKA_ID_TRACKFLAGENABLED: {
1505 uint64_t num;
1506 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1507 break;
1508 if (num)
1509 track->flags |= MATROSKA_TRACK_ENABLED;
1510 else
1511 track->flags &= ~MATROSKA_TRACK_ENABLED;
1512 break;
1513 }
1514
1515 /* whether it's the default for this track type */
1516 case MATROSKA_ID_TRACKFLAGDEFAULT: {
1517 uint64_t num;
1518 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1519 break;
1520 if (num)
1521 track->flags |= MATROSKA_TRACK_DEFAULT;
1522 else
1523 track->flags &= ~MATROSKA_TRACK_DEFAULT;
1524 break;
1525 }
1526
1527 /* lacing (like MPEG, where blocks don't end/start on frame
1528 * boundaries) */
1529 case MATROSKA_ID_TRACKFLAGLACING: {
1530 uint64_t num;
1531 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1532 break;
1533 if (num)
1534 track->flags |= MATROSKA_TRACK_LACING;
1535 else
1536 track->flags &= ~MATROSKA_TRACK_LACING;
1537 break;
1538 }
1539
1540 /* default length (in time) of one data block in this track */
1541 case MATROSKA_ID_TRACKDEFAULTDURATION: {
1542 uint64_t num;
1543 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1544 break;
1545 track->default_duration = num;
1546 break;
1547 }
1548
1549 default:
1550 av_log(matroska->ctx, AV_LOG_INFO,
1551 "Unknown track header entry 0x%x - ignoring\n", id);
1552 /* pass-through */
1553
1554 case EBML_ID_VOID:
1555 /* we ignore these because they're nothing useful. */
1556 case MATROSKA_ID_CODECINFOURL:
1557 case MATROSKA_ID_CODECDOWNLOADURL:
1558 case MATROSKA_ID_TRACKMINCACHE:
1559 case MATROSKA_ID_TRACKMAXCACHE:
1560 res = ebml_read_skip(matroska);
1561 break;
1562 }
1563
1564 if (matroska->level_up) {
1565 matroska->level_up--;
1566 break;
1567 }
1568 }
1569
1570 return res;
1571 }
1572
1573 static int
1574 matroska_parse_tracks (MatroskaDemuxContext *matroska)
1575 {
1576 int res = 0;
1577 uint32_t id;
1578
1579 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing tracks...\n");
1580
1581 while (res == 0) {
1582 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1583 res = AVERROR_IO;
1584 break;
1585 } else if (matroska->level_up) {
1586 matroska->level_up--;
1587 break;
1588 }
1589
1590 switch (id) {
1591 /* one track within the "all-tracks" header */
1592 case MATROSKA_ID_TRACKENTRY:
1593 res = matroska_add_stream(matroska);
1594 break;
1595
1596 default:
1597 av_log(matroska->ctx, AV_LOG_INFO,
1598 "Unknown entry 0x%x in track header\n", id);
1599 /* fall-through */
1600
1601 case EBML_ID_VOID:
1602 res = ebml_read_skip(matroska);
1603 break;
1604 }
1605
1606 if (matroska->level_up) {
1607 matroska->level_up--;
1608 break;
1609 }
1610 }
1611
1612 return res;
1613 }
1614
1615 static int
1616 matroska_parse_index (MatroskaDemuxContext *matroska)
1617 {
1618 int res = 0;
1619 uint32_t id;
1620 MatroskaDemuxIndex idx;
1621
1622 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing index...\n");
1623
1624 while (res == 0) {
1625 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1626 res = AVERROR_IO;
1627 break;
1628 } else if (matroska->level_up) {
1629 matroska->level_up--;
1630 break;
1631 }
1632
1633 switch (id) {
1634 /* one single index entry ('point') */
1635 case MATROSKA_ID_POINTENTRY:
1636 if ((res = ebml_read_master(matroska, &id)) < 0)
1637 break;
1638
1639 /* in the end, we hope to fill one entry with a
1640 * timestamp, a file position and a tracknum */
1641 idx.pos = (uint64_t) -1;
1642 idx.time = (uint64_t) -1;
1643 idx.track = (uint16_t) -1;
1644
1645 while (res == 0) {
1646 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1647 res = AVERROR_IO;
1648 break;
1649 } else if (matroska->level_up) {
1650 matroska->level_up--;
1651 break;
1652 }
1653
1654 switch (id) {
1655 /* one single index entry ('point') */
1656 case MATROSKA_ID_CUETIME: {
1657 int64_t time;
1658 if ((res = ebml_read_uint(matroska, &id,
1659 &time)) < 0)
1660 break;
1661 idx.time = time * matroska->time_scale;
1662 break;
1663 }
1664
1665 /* position in the file + track to which it
1666 * belongs */
1667 case MATROSKA_ID_CUETRACKPOSITION:
1668 if ((res = ebml_read_master(matroska, &id)) < 0)
1669 break;
1670
1671 while (res == 0) {
1672 if (!(id = ebml_peek_id (matroska,
1673 &matroska->level_up))) {
1674 res = AVERROR_IO;
1675 break;
1676 } else if (matroska->level_up) {
1677 matroska->level_up--;
1678 break;
1679 }
1680
1681 switch (id) {
1682 /* track number */
1683 case MATROSKA_ID_CUETRACK: {
1684 uint64_t num;
1685 if ((res = ebml_read_uint(matroska,
1686 &id, &num)) < 0)
1687 break;
1688 idx.track = num;
1689 break;
1690 }
1691
1692 /* position in file */
1693 case MATROSKA_ID_CUECLUSTERPOSITION: {
1694 uint64_t num;
1695 if ((res = ebml_read_uint(matroska,
1696 &id, &num)) < 0)
1697 break;
1698 idx.pos = num;
1699 break;
1700 }
1701
1702 default:
1703 av_log(matroska->ctx, AV_LOG_INFO,
1704 "Unknown entry 0x%x in "
1705 "CuesTrackPositions\n", id);
1706 /* fall-through */
1707
1708 case EBML_ID_VOID:
1709 res = ebml_read_skip(matroska);
1710 break;
1711 }
1712
1713 if (matroska->level_up) {
1714 matroska->level_up--;
1715 break;
1716 }
1717 }
1718
1719 break;
1720
1721 default:
1722 av_log(matroska->ctx, AV_LOG_INFO,
1723 "Unknown entry 0x%x in cuespoint "
1724 "index\n", id);
1725 /* fall-through */
1726
1727 case EBML_ID_VOID:
1728 res = ebml_read_skip(matroska);
1729 break;
1730 }
1731
1732 if (matroska->level_up) {
1733 matroska->level_up--;
1734 break;
1735 }
1736 }
1737
1738 /* so let's see if we got what we wanted */
1739 if (idx.pos != (uint64_t) -1 &&
1740 idx.time != (uint64_t) -1 &&
1741 idx.track != (uint16_t) -1) {
1742 if (matroska->num_indexes % 32 == 0) {
1743 /* re-allocate bigger index */
1744 matroska->index =
1745 av_realloc(matroska->index,
1746 (matroska->num_indexes + 32) *
1747 sizeof(MatroskaDemuxIndex));
1748 }
1749 matroska->index[matroska->num_indexes] = idx;
1750 matroska->num_indexes++;
1751 }
1752 break;
1753
1754 default:
1755 av_log(matroska->ctx, AV_LOG_INFO,
1756 "Unknown entry 0x%x in cues header\n", id);
1757 /* fall-through */
1758
1759 case EBML_ID_VOID:
1760 res = ebml_read_skip(matroska);
1761 break;
1762 }
1763
1764 if (matroska->level_up) {
1765 matroska->level_up--;
1766 break;
1767 }
1768 }
1769
1770 return res;
1771 }
1772
1773 static int
1774 matroska_parse_metadata (MatroskaDemuxContext *matroska)
1775 {
1776 int res = 0;
1777 uint32_t id;
1778
1779 while (res == 0) {
1780 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1781 res = AVERROR_IO;
1782 break;
1783 } else if (matroska->level_up) {
1784 matroska->level_up--;
1785 break;
1786 }
1787
1788 switch (id) {
1789 /* Hm, this is unsupported... */
1790 default:
1791 av_log(matroska->ctx, AV_LOG_INFO,
1792 "Unknown entry 0x%x in metadata 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_seekhead (MatroskaDemuxContext *matroska)
1811 {
1812 int res = 0;
1813 uint32_t id;
1814
1815 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing seekhead...\n");
1816
1817 while (res == 0) {
1818 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1819 res = AVERROR_IO;
1820 break;
1821 } else if (matroska->level_up) {
1822 matroska->level_up--;
1823 break;
1824 }
1825
1826 switch (id) {
1827 case MATROSKA_ID_SEEKENTRY: {
1828 uint32_t seek_id = 0, peek_id_cache = 0;
1829 uint64_t seek_pos = (uint64_t) -1, t;
1830
1831 if ((res = ebml_read_master(matroska, &id)) < 0)
1832 break;
1833
1834 while (res == 0) {
1835 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1836 res = AVERROR_IO;
1837 break;
1838 } else if (matroska->level_up) {
1839 matroska->level_up--;
1840 break;
1841 }
1842
1843 switch (id) {
1844 case MATROSKA_ID_SEEKID:
1845 res = ebml_read_uint(matroska, &id, &t);
1846 seek_id = t;
1847 break;
1848
1849 case MATROSKA_ID_SEEKPOSITION:
1850 res = ebml_read_uint(matroska, &id, &seek_pos);
1851 break;
1852
1853 default:
1854 av_log(matroska->ctx, AV_LOG_INFO,
1855 "Unknown seekhead ID 0x%x\n", id);
1856 /* fall-through */
1857
1858 case EBML_ID_VOID:
1859 res = ebml_read_skip(matroska);
1860 break;
1861 }
1862
1863 if (matroska->level_up) {
1864 matroska->level_up--;
1865 break;
1866 }
1867 }
1868
1869 if (!seek_id || seek_pos == (uint64_t) -1) {
1870 av_log(matroska->ctx, AV_LOG_INFO,
1871 "Incomplete seekhead entry (0x%x/%"PRIu64")\n",
1872 seek_id, seek_pos);
1873 break;
1874 }
1875
1876 switch (seek_id) {
1877 case MATROSKA_ID_CUES:
1878 case MATROSKA_ID_TAGS: {
1879 uint32_t level_up = matroska->level_up;
1880 offset_t before_pos;
1881 uint64_t length;
1882 MatroskaLevel level;
1883
1884 /* remember the peeked ID and the current position */
1885 peek_id_cache = matroska->peek_id;
1886 before_pos = url_ftell(&matroska->ctx->pb);
1887
1888 /* seek */
1889 if ((res = ebml_read_seek(matroska, seek_pos +
1890 matroska->segment_start)) < 0)
1891 return res;
1892
1893 /* we don't want to lose our seekhead level, so we add
1894 * a dummy. This is a crude hack. */
1895 if (matroska->num_levels == EBML_MAX_DEPTH) {
1896 av_log(matroska->ctx, AV_LOG_INFO,
1897 "Max EBML element depth (%d) reached, "
1898 "cannot parse further.\n", EBML_MAX_DEPTH);
1899 return AVERROR_UNKNOWN;
1900 }
1901
1902 level.start = 0;
1903 level.length = (uint64_t)-1;
1904 matroska->levels[matroska->num_levels] = level;
1905 matroska->num_levels++;
1906
1907 /* check ID */
1908 if (!(id = ebml_peek_id (matroska,
1909 &matroska->level_up)))
1910 break;
1911 if (id != seek_id) {
1912 av_log(matroska->ctx, AV_LOG_INFO,
1913 "We looked for ID=0x%x but got "
1914 "ID=0x%x (pos=%"PRIu64")",
1915 seek_id, id, seek_pos +
1916 matroska->segment_start);
1917 goto finish;
1918 }
1919
1920 /* read master + parse */
1921 if ((res = ebml_read_master(matroska, &id)) < 0)
1922 break;
1923 switch (id) {
1924 case MATROSKA_ID_CUES:
1925 if (!(res = matroska_parse_index(matroska)) ||
1926 url_feof(&matroska->ctx->pb)) {
1927 matroska->index_parsed = 1;
1928 res = 0;
1929 }
1930 break;
1931 case MATROSKA_ID_TAGS:
1932 if (!(res = matroska_parse_metadata(matroska)) ||
1933 url_feof(&matroska->ctx->pb)) {
1934 matroska->metadata_parsed = 1;
1935 res = 0;
1936 }
1937 break;
1938 }
1939 if (res < 0)
1940 break;
1941
1942 finish:
1943 /* remove dummy level */
1944 while (matroska->num_levels) {
1945 matroska->num_levels--;
1946 length =
1947 matroska->levels[matroska->num_levels].length;
1948 if (length == (uint64_t)-1)
1949 break;
1950 }
1951
1952 /* seek back */
1953 if ((res = ebml_read_seek(matroska, before_pos)) < 0)
1954 return res;
1955 matroska->peek_id = peek_id_cache;
1956 matroska->level_up = level_up;
1957 break;
1958 }
1959
1960 default:
1961 av_log(matroska->ctx, AV_LOG_INFO,
1962 "Ignoring seekhead entry for ID=0x%x\n",
1963 seek_id);
1964 break;
1965 }
1966
1967 break;
1968 }
1969
1970 default:
1971 av_log(matroska->ctx, AV_LOG_INFO,
1972 "Unknown seekhead ID 0x%x\n", id);
1973 /* fall-through */
1974
1975 case EBML_ID_VOID:
1976 res = ebml_read_skip(matroska);
1977 break;
1978 }
1979
1980 if (matroska->level_up) {
1981 matroska->level_up--;
1982 break;
1983 }
1984 }
1985
1986 return res;
1987 }
1988
1989 static int
1990 matroska_read_header (AVFormatContext *s,
1991 AVFormatParameters *ap)
1992 {
1993 MatroskaDemuxContext *matroska = s->priv_data;
1994 char *doctype;
1995 int version, last_level, res = 0;
1996 uint32_t id;
1997
1998 matroska->ctx = s;
1999
2000 /* First read the EBML header. */
2001 doctype = NULL;
2002 if ((res = ebml_read_header(matroska, &doctype, &version)) < 0)
2003 return res;
2004 if ((doctype == NULL) || strcmp(doctype, "matroska")) {
2005 av_log(matroska->ctx, AV_LOG_ERROR,
2006 "Wrong EBML doctype ('%s' != 'matroska').\n",
2007 doctype ? doctype : "(none)");
2008 if (doctype)
2009 av_free(doctype);
2010 return AVERROR_NOFMT;
2011 }
2012 av_free(doctype);
2013 if (version != 1) {
2014 av_log(matroska->ctx, AV_LOG_ERROR,
2015 "Matroska demuxer version 1 too old for file version %d\n",
2016 version);
2017 return AVERROR_NOFMT;
2018 }
2019
2020 /* The next thing is a segment. */
2021 while (1) {
2022 if (!(id = ebml_peek_id(matroska, &last_level)))
2023 return AVERROR_IO;
2024 if (id == MATROSKA_ID_SEGMENT)
2025 break;
2026
2027 /* oi! */
2028 av_log(matroska->ctx, AV_LOG_INFO,
2029 "Expected a Segment ID (0x%x), but received 0x%x!\n",
2030 MATROSKA_ID_SEGMENT, id);
2031 if ((res = ebml_read_skip(matroska)) < 0)
2032 return res;
2033 }
2034
2035 /* We now have a Matroska segment.
2036 * Seeks are from the beginning of the segment,
2037 * after the segment ID/length. */
2038 if ((res = ebml_read_master(matroska, &id)) < 0)
2039 return res;
2040 matroska->segment_start = url_ftell(&s->pb);
2041
2042 matroska->time_scale = 1000000;
2043 /* we've found our segment, start reading the different contents in here */
2044 while (res == 0) {
2045 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2046 res = AVERROR_IO;
2047 break;
2048 } else if (matroska->level_up) {
2049 matroska->level_up--;
2050 break;
2051 }
2052
2053 switch (id) {
2054 /* stream info */
2055 case MATROSKA_ID_INFO: {
2056 if ((res = ebml_read_master(matroska, &id)) < 0)
2057 break;
2058 res = matroska_parse_info(matroska);
2059 break;
2060 }
2061
2062 /* track info headers */
2063 case MATROSKA_ID_TRACKS: {
2064 if ((res = ebml_read_master(matroska, &id)) < 0)
2065 break;
2066 res = matroska_parse_tracks(matroska);
2067 break;
2068 }
2069
2070 /* stream index */
2071 case MATROSKA_ID_CUES: {
2072 if (!matroska->index_parsed) {
2073 if ((res = ebml_read_master(matroska, &id)) < 0)
2074 break;
2075 res = matroska_parse_index(matroska);
2076 } else
2077 res = ebml_read_skip(matroska);
2078 break;
2079 }
2080
2081 /* metadata */
2082 case MATROSKA_ID_TAGS: {
2083 if (!matroska->metadata_parsed) {
2084 if ((res = ebml_read_master(matroska, &id)) < 0)
2085 break;
2086 res = matroska_parse_metadata(matroska);
2087 } else
2088 res = ebml_read_skip(matroska);
2089 break;
2090 }
2091
2092 /* file index (if seekable, seek to Cues/Tags to parse it) */
2093 case MATROSKA_ID_SEEKHEAD: {
2094 if ((res = ebml_read_master(matroska, &id)) < 0)
2095 break;
2096 res = matroska_parse_seekhead(matroska);
2097 break;
2098 }
2099
2100 case MATROSKA_ID_CLUSTER: {
2101 /* Do not read the master - this will be done in the next
2102 * call to matroska_read_packet. */
2103 res = 1;
2104 break;
2105 }
2106
2107 default:
2108 av_log(matroska->ctx, AV_LOG_INFO,
2109 "Unknown matroska file header ID 0x%x\n", id);
2110 /* fall-through */
2111
2112 case EBML_ID_VOID:
2113 res = ebml_read_skip(matroska);
2114 break;
2115 }
2116
2117 if (matroska->level_up) {
2118 matroska->level_up--;
2119 break;
2120 }
2121 }
2122
2123 if (res < 0)
2124 return res;
2125
2126 /* Have we found a cluster? */
2127 if (res == 1) {
2128 int i;
2129 enum CodecID codec_id;
2130 MatroskaTrack *track;
2131 AVStream *st;
2132 void *extradata = NULL;
2133 int extradata_size = 0;
2134
2135 for (i = 0; i < matroska->num_tracks; i++) {
2136 track = matroska->tracks[i];
2137
2138 /* libavformat does not really support subtitles.
2139 * Also apply some sanity checks. */
2140 if ((track->type == MATROSKA_TRACK_TYPE_SUBTITLE) ||
2141 (track->codec_id == NULL))
2142 continue;
2143
2144 /* Set the FourCC from the CodecID. */
2145 /* This is the MS compatibility mode which stores a
2146 * BITMAPINFOHEADER in the CodecPrivate. */
2147 if (!strcmp(track->codec_id,
2148 MATROSKA_CODEC_ID_VIDEO_VFW_FOURCC) &&
2149 (track->codec_priv_size >= 40) &&
2150 (track->codec_priv != NULL)) {
2151 unsigned char *p;
2152
2153 /* Offset of biCompression. Stored in LE. */
2154 p = (unsigned char *)track->codec_priv + 16;
2155 ((MatroskaVideoTrack *)track)->fourcc = (p[3] << 24) |
2156 (p[2] << 16) | (p[1] << 8) | p[0];
2157 codec_id = codec_get_bmp_id(((MatroskaVideoTrack *)track)->fourcc);
2158
2159 } else if (!strcmp(track->codec_id,
2160 MATROSKA_CODEC_ID_VIDEO_MPEG4_SP) ||
2161 !strcmp(track->codec_id,
2162 MATROSKA_CODEC_ID_VIDEO_MPEG4_ASP) ||
2163 !strcmp(track->codec_id,
2164 MATROSKA_CODEC_ID_VIDEO_MPEG4_AP))
2165 codec_id = CODEC_ID_MPEG4;
2166 else if (!strcmp(track->codec_id,
2167 MATROSKA_CODEC_ID_VIDEO_MPEG4_AVC))
2168 codec_id = CODEC_ID_H264;
2169 /* else if (!strcmp(track->codec_id, */
2170 /* MATROSKA_CODEC_ID_VIDEO_UNCOMPRESSED)) */
2171 /* codec_id = CODEC_ID_???; */
2172 else if (!strcmp(track->codec_id,
2173 MATROSKA_CODEC_ID_VIDEO_MSMPEG4V3))
2174 codec_id = CODEC_ID_MSMPEG4V3;
2175 else if (!strcmp(track->codec_id,
2176 MATROSKA_CODEC_ID_VIDEO_MPEG1) ||
2177 !strcmp(track->codec_id,
2178 MATROSKA_CODEC_ID_VIDEO_MPEG2))
2179 codec_id = CODEC_ID_MPEG2VIDEO;
2180
2181 /* This is the MS compatibility mode which stores a
2182 * WAVEFORMATEX in the CodecPrivate. */
2183 else if (!strcmp(track->codec_id,
2184 MATROSKA_CODEC_ID_AUDIO_ACM) &&
2185 (track->codec_priv_size >= 18) &&
2186 (track->codec_priv != NULL)) {
2187 unsigned char *p;
2188 uint16_t tag;
2189
2190 /* Offset of wFormatTag. Stored in LE. */
2191 p = (unsigned char *)track->codec_priv;
2192 tag = (p[1] << 8) | p[0];
2193 codec_id = codec_get_wav_id(tag);
2194
2195 } else if (!strcmp(track->codec_id,
2196 MATROSKA_CODEC_ID_AUDIO_MPEG1_L1) ||
2197 !strcmp(track->codec_id,
2198 MATROSKA_CODEC_ID_AUDIO_MPEG1_L2) ||
2199 !strcmp(track->codec_id,
2200 MATROSKA_CODEC_ID_AUDIO_MPEG1_L3))
2201 codec_id = CODEC_ID_MP3;
2202 else if (!strcmp(track->codec_id,
2203 MATROSKA_CODEC_ID_AUDIO_PCM_INT_BE))
2204 codec_id = CODEC_ID_PCM_U16BE;
2205 else if (!strcmp(track->codec_id,
2206 MATROSKA_CODEC_ID_AUDIO_PCM_INT_LE))
2207 codec_id = CODEC_ID_PCM_U16LE;
2208 /* else if (!strcmp(track->codec_id, */
2209 /* MATROSKA_CODEC_ID_AUDIO_PCM_FLOAT)) */
2210 /* codec_id = CODEC_ID_PCM_???; */
2211 else if (!strcmp(track->codec_id,
2212 MATROSKA_CODEC_ID_AUDIO_AC3))
2213 codec_id = CODEC_ID_AC3;
2214 else if (!strcmp(track->codec_id,
2215 MATROSKA_CODEC_ID_AUDIO_DTS))
2216 codec_id = CODEC_ID_DTS;
2217 /* No such codec id so far. */
2218 /* else if (!strcmp(track->codec_id, */
2219 /* MATROSKA_CODEC_ID_AUDIO_DTS)) */
2220 /* codec_id = CODEC_ID_DTS; */
2221 else if (!strcmp(track->codec_id,
2222 MATROSKA_CODEC_ID_AUDIO_VORBIS)) {
2223 extradata_size = track->codec_priv_size;
2224 if(extradata_size) {
2225 extradata = av_malloc(extradata_size);
2226 if(extradata == NULL)
2227 return AVERROR_NOMEM;
2228 memcpy(extradata, track->codec_priv, extradata_size);
2229 }
2230 codec_id = CODEC_ID_VORBIS;
2231 } else if (!strcmp(track->codec_id,
2232 MATROSKA_CODEC_ID_AUDIO_MPEG2) ||
2233 !strcmp(track->codec_id,
2234 MATROSKA_CODEC_ID_AUDIO_MPEG4))
2235 codec_id = CODEC_ID_AAC;
2236 else
2237 codec_id = CODEC_ID_NONE;
2238
2239 if (codec_id == CODEC_ID_NONE) {
2240 av_log(matroska->ctx, AV_LOG_INFO,
2241 "Unknown/unsupported CodecID %s.\n",
2242 track->codec_id);
2243 }
2244
2245 track->stream_index = matroska->num_streams;
2246
2247 matroska->num_streams++;
2248 st = av_new_stream(s, track->stream_index);
2249 if (st == NULL)
2250 return AVERROR_NOMEM;
2251 av_set_pts_info(st, 24, 1, 1000); /* 24 bit pts in ms */
2252
2253 st->codec->codec_id = codec_id;
2254
2255 if(extradata){
2256 st->codec->extradata = extradata;
2257 st->codec->extradata_size = extradata_size;
2258 } else if(track->codec_priv && track->codec_priv_size > 0){
2259 st->codec->extradata = av_malloc(track->codec_priv_size);
2260 if(st->codec->extradata == NULL)
2261 return AVERROR_NOMEM;
2262 st->codec->extradata_size = track->codec_priv_size;
2263 memcpy(st->codec->extradata, track->codec_priv,
2264 track->codec_priv_size);
2265 }
2266
2267 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2268 MatroskaVideoTrack *videotrack = (MatroskaVideoTrack *)track;
2269
2270 st->codec->codec_type = CODEC_TYPE_VIDEO;
2271 st->codec->codec_tag = videotrack->fourcc;
2272 st->codec->width = videotrack->pixel_width;
2273 st->codec->height = videotrack->pixel_height;
2274 if (videotrack->display_width == 0)
2275 st->codec->sample_aspect_ratio.num =
2276 videotrack->pixel_width;
2277 else
2278 st->codec->sample_aspect_ratio.num =
2279 videotrack->display_width;
2280 if (videotrack->display_height == 0)
2281 st->codec->sample_aspect_ratio.num =
2282 videotrack->pixel_height;
2283 else
2284 st->codec->sample_aspect_ratio.num =
2285 videotrack->display_height;
2286
2287 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2288 MatroskaAudioTrack *audiotrack = (MatroskaAudioTrack *)track;
2289
2290 st->codec->codec_type = CODEC_TYPE_AUDIO;
2291 st->codec->sample_rate = audiotrack->samplerate;
2292 st->codec->channels = audiotrack->channels;
2293 }
2294
2295 /* What do we do with private data? E.g. for Vorbis. */
2296 }
2297 }
2298
2299 return 0;
2300 }
2301
2302 static int
2303 matroska_find_track_by_num (MatroskaDemuxContext *matroska,
2304 int num)
2305 {
2306 int i;
2307
2308 for (i = 0; i < matroska->num_tracks; i++)
2309 if (matroska->tracks[i]->num == num)
2310 return i;
2311
2312 return -1;
2313 }
2314
2315 static int
2316 matroska_parse_blockgroup (MatroskaDemuxContext *matroska,
2317 uint64_t cluster_time)
2318 {
2319 int res = 0;
2320 uint32_t id;
2321 AVPacket *pkt;
2322 int is_keyframe = PKT_FLAG_KEY, last_num_packets = matroska->num_packets;
2323
2324 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing blockgroup...\n");
2325
2326 while (res == 0) {
2327 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2328 res = AVERROR_IO;
2329 break;
2330 } else if (matroska->level_up) {
2331 matroska->level_up--;
2332 break;
2333 }
2334
2335 switch (id) {
2336 /* one block inside the group. Note, block parsing is one
2337 * of the harder things, so this code is a bit complicated.
2338 * See http://www.matroska.org/ for documentation. */
2339 case MATROSKA_ID_BLOCK: {
2340 uint8_t *data, *origdata;
2341 int size;
2342 uint64_t time;
2343 uint32_t *lace_size = NULL;
2344 int n, track, flags, laces = 0;
2345 uint64_t num;
2346 int64_t pos= url_ftell(&matroska->ctx->pb);
2347
2348 if ((res = ebml_read_binary(matroska, &id, &data, &size)) < 0)
2349 break;
2350 origdata = data;
2351
2352 /* first byte(s): blocknum */
2353 if ((n = matroska_ebmlnum_uint(data, size, &num)) < 0) {
2354 av_log(matroska->ctx, AV_LOG_ERROR,
2355 "EBML block data error\n");
2356 av_free(origdata);
2357 break;
2358 }
2359 data += n;
2360 size -= n;
2361
2362 /* fetch track from num */
2363 track = matroska_find_track_by_num(matroska, num);
2364 if (size <= 3 || track < 0 || track >= matroska->num_tracks) {
2365 av_log(matroska->ctx, AV_LOG_INFO,
2366 "Invalid stream %d or size %u\n", track, size);
2367 av_free(origdata);
2368 break;
2369 }
2370 if(matroska->ctx->streams[ matroska->tracks[track]->stream_index ]->discard >= AVDISCARD_ALL){
2371 av_free(origdata);
2372 break;
2373 }
2374
2375 /* time (relative to cluster time) */
2376 time = ((data[0] << 8) | data[1]) * matroska->time_scale;
2377 data += 2;
2378 size -= 2;
2379 flags = *data;
2380 data += 1;
2381 size -= 1;
2382 switch ((flags & 0x06) >> 1) {
2383 case 0x0: /* no lacing */
2384 laces = 1;
2385 lace_size = av_mallocz(sizeof(int));
2386 lace_size[0] = size;
2387 break;
2388
2389 case 0x1: /* xiph lacing */
2390 case 0x2: /* fixed-size lacing */
2391 case 0x3: /* EBML lacing */
2392 if (size == 0) {
2393 res = -1;
2394 break;
2395 }
2396 laces = (*data) + 1;
2397 data += 1;
2398 size -= 1;
2399 lace_size = av_mallocz(laces * sizeof(int));
2400
2401 switch ((flags & 0x06) >> 1) {
2402 case 0x1: /* xiph lacing */ {
2403 uint8_t temp;
2404 uint32_t total = 0;
2405 for (n = 0; res == 0 && n < laces - 1; n++) {
2406 while (1) {
2407 if (size == 0) {
2408 res = -1;
2409 break;
2410 }
2411 temp = *data;
2412 lace_size[n] += temp;
2413 data += 1;
2414 size -= 1;
2415 if (temp != 0xff)
2416 break;
2417 }
2418 total += lace_size[n];
2419 }
2420 lace_size[n] = size - total;
2421 break;
2422 }
2423
2424 case 0x2: /* fixed-size lacing */
2425 for (n = 0; n < laces; n++)
2426 lace_size[n] = size / laces;
2427 break;
2428
2429 case 0x3: /* EBML lacing */ {
2430 uint32_t total;
2431 n = matroska_ebmlnum_uint(data, size, &num);
2432 if (n < 0) {
2433 av_log(matroska->ctx, AV_LOG_INFO,
2434 "EBML block data error\n");
2435 break;
2436 }
2437 data += n;
2438 size -= n;
2439 total = lace_size[0] = num;
2440 for (n = 1; res == 0 && n < laces - 1; n++) {
2441 int64_t snum;
2442 int r;
2443 r = matroska_ebmlnum_sint (data, size,
2444 &snum);
2445 if (r < 0) {
2446 av_log(matroska->ctx, AV_LOG_INFO,
2447 "EBML block data error\n");
2448 break;
2449 }
2450 data += r;
2451 size -= r;
2452 lace_size[n] = lace_size[n - 1] + snum;
2453 total += lace_size[n];
2454 }
2455 lace_size[n] = size - total;
2456 break;
2457 }
2458 }
2459 break;
2460 }
2461
2462 if (res == 0) {
2463 for (n = 0; n < laces; n++) {
2464 uint64_t timecode = 0;
2465
2466 pkt = av_mallocz(sizeof(AVPacket));
2467 /* XXX: prevent data copy... */
2468 if (av_new_packet(pkt,lace_size[n]) < 0) {
2469 res = AVERROR_NOMEM;
2470 break;
2471 }
2472 if (cluster_time != (uint64_t)-1) {
2473 if (time < 0 && (-time) > cluster_time)
2474 timecode = cluster_time;
2475 else
2476 timecode = cluster_time + time;
2477 }
2478 /* FIXME: duration */
2479
2480 memcpy(pkt->data, data, lace_size[n]);
2481 data += lace_size[n];
2482 if (n == 0)
2483 pkt->flags = is_keyframe;
2484 pkt->stream_index =
2485 matroska->tracks[track]->stream_index;
2486
2487 pkt->pts = timecode / 1000000; /* ns to ms */
2488 pkt->pos= pos;
2489
2490 matroska_queue_packet(matroska, pkt);
2491 }
2492 }
2493
2494 av_free(lace_size);
2495 av_free(origdata);
2496 break;
2497 }
2498
2499 case MATROSKA_ID_BLOCKDURATION: {
2500 uint64_t num;
2501 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
2502 break;
2503 av_log(matroska->ctx, AV_LOG_INFO,
2504 "FIXME: implement support for BlockDuration\n");
2505 break;
2506 }
2507
2508 case MATROSKA_ID_BLOCKREFERENCE:
2509 /* We've found a reference, so not even the first frame in
2510 * the lace is a key frame. */
2511 is_keyframe = 0;
2512 if (last_num_packets != matroska->num_packets)
2513 matroska->packets[last_num_packets]->flags = 0;
2514 res = ebml_read_skip(matroska);
2515 break;
2516
2517 default:
2518 av_log(matroska->ctx, AV_LOG_INFO,
2519 "Unknown entry 0x%x in blockgroup data\n", id);
2520 /* fall-through */
2521
2522 case EBML_ID_VOID:
2523 res = ebml_read_skip(matroska);
2524 break;
2525 }
2526
2527 if (matroska->level_up) {
2528 matroska->level_up--;
2529 break;
2530 }
2531 }
2532
2533 return res;
2534 }
2535
2536 static int
2537 matroska_parse_cluster (MatroskaDemuxContext *matroska)
2538 {
2539 int res = 0;
2540 uint32_t id;
2541 uint64_t cluster_time = 0;
2542
2543 av_log(matroska->ctx, AV_LOG_DEBUG,
2544 "parsing cluster at %"PRId64"\n", url_ftell(&matroska->ctx->pb));
2545
2546 while (res == 0) {
2547 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2548 res = AVERROR_IO;
2549 break;
2550 } else if (matroska->level_up) {
2551 matroska->level_up--;
2552 break;
2553 }
2554
2555 switch (id) {
2556 /* cluster timecode */
2557 case MATROSKA_ID_CLUSTERTIMECODE: {
2558 uint64_t num;
2559 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
2560 break;
2561 cluster_time = num * matroska->time_scale;
2562 break;
2563 }
2564
2565 /* a group of blocks inside a cluster */
2566 case MATROSKA_ID_BLOCKGROUP:
2567 if ((res = ebml_read_master(matroska, &id)) < 0)
2568 break;
2569 res = matroska_parse_blockgroup(matroska, cluster_time);
2570 break;
2571
2572 default:
2573 av_log(matroska->ctx, AV_LOG_INFO,
2574 "Unknown entry 0x%x in cluster data\n", id);
2575 /* fall-through */
2576
2577 case EBML_ID_VOID:
2578 res = ebml_read_skip(matroska);
2579 break;
2580 }
2581
2582 if (matroska->level_up) {
2583 matroska->level_up--;
2584 break;
2585 }
2586 }
2587
2588 return res;
2589 }
2590
2591 static int
2592 matroska_read_packet (AVFormatContext *s,
2593 AVPacket *pkt)
2594 {
2595 MatroskaDemuxContext *matroska = s->priv_data;
2596 int res = 0;
2597 uint32_t id;
2598
2599 /* Do we still have a packet queued? */
2600 if (matroska_deliver_packet(matroska, pkt) == 0)
2601 return 0;
2602
2603 /* Have we already reached the end? */
2604 if (matroska->done)
2605 return AVERROR_IO;
2606
2607 while (res == 0) {
2608 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2609 res = AVERROR_IO;
2610 break;
2611 } else if (matroska->level_up) {
2612 matroska->level_up--;
2613 break;
2614 }
2615
2616 switch (id) {
2617 case MATROSKA_ID_CLUSTER:
2618 if ((res = ebml_read_master(matroska, &id)) < 0)
2619 break;
2620 if ((res = matroska_parse_cluster(matroska)) == 0)
2621 res = 1; /* Parsed one cluster, let's get out. */
2622 break;
2623
2624 default:
2625 case EBML_ID_VOID:
2626 res = ebml_read_skip(matroska);
2627 break;
2628 }
2629
2630 if (matroska->level_up) {
2631 matroska->level_up--;
2632 break;
2633 }
2634 }
2635
2636 if (res == -1)
2637 matroska->done = 1;
2638
2639 return matroska_deliver_packet(matroska, pkt);
2640 }
2641
2642 static int
2643 matroska_read_close (AVFormatContext *s)
2644 {
2645 MatroskaDemuxContext *matroska = s->priv_data;
2646 int n = 0;
2647
2648 if (matroska->writing_app)
2649 av_free(matroska->writing_app);
2650 if (matroska->muxing_app)
2651 av_free(matroska->muxing_app);
2652 if (matroska->index)
2653 av_free(matroska->index);
2654
2655 if (matroska->packets != NULL) {
2656 for (n = 0; n < matroska->num_packets; n++) {
2657 av_free_packet(matroska->packets[n]);
2658 av_free(matroska->packets[n]);
2659 }
2660 av_free(matroska->packets);
2661 }
2662
2663 for (n = 0; n < matroska->num_tracks; n++) {
2664 MatroskaTrack *track = matroska->tracks[n];
2665 if (track->codec_id)
2666 av_free(track->codec_id);
2667 if (track->codec_name)
2668 av_free(track->codec_name);
2669 if (track->codec_priv)
2670 av_free(track->codec_priv);
2671 if (track->name)
2672 av_free(track->name);
2673 if (track->language)
2674 av_free(track->language);
2675
2676 av_free(track);
2677 }
2678
2679 for (n = 0; n < s->nb_streams; n++) {
2680 av_free(s->streams[n]->codec->extradata);
2681 }
2682
2683 memset(matroska, 0, sizeof(MatroskaDemuxContext));
2684
2685 return 0;
2686 }
2687
2688 static AVInputFormat matroska_iformat = {
2689 "matroska",
2690 "Matroska file format",
2691 sizeof(MatroskaDemuxContext),
2692 matroska_probe,
2693 matroska_read_header,
2694 matroska_read_packet,
2695 matroska_read_close,
2696 };
2697
2698 int
2699 matroska_init(void)
2700 {
2701 av_register_input_format(&matroska_iformat);
2702 return 0;
2703 }