47c0e98a48e8c9ff5c07094b5df5e587cf0e230a
[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
33 /* EBML version supported */
34 #define EBML_VERSION 1
35
36 /* top-level master-IDs */
37 #define EBML_ID_HEADER 0x1A45DFA3
38
39 /* IDs in the HEADER master */
40 #define EBML_ID_EBMLVERSION 0x4286
41 #define EBML_ID_EBMLREADVERSION 0x42F7
42 #define EBML_ID_EBMLMAXIDLENGTH 0x42F2
43 #define EBML_ID_EBMLMAXSIZELENGTH 0x42F3
44 #define EBML_ID_DOCTYPE 0x4282
45 #define EBML_ID_DOCTYPEVERSION 0x4287
46 #define EBML_ID_DOCTYPEREADVERSION 0x4285
47
48 /* general EBML types */
49 #define EBML_ID_VOID 0xEC
50
51 /*
52 * Matroska element IDs. max. 32-bit.
53 */
54
55 /* toplevel segment */
56 #define MATROSKA_ID_SEGMENT 0x18538067
57
58 /* matroska top-level master IDs */
59 #define MATROSKA_ID_INFO 0x1549A966
60 #define MATROSKA_ID_TRACKS 0x1654AE6B
61 #define MATROSKA_ID_CUES 0x1C53BB6B
62 #define MATROSKA_ID_TAGS 0x1254C367
63 #define MATROSKA_ID_SEEKHEAD 0x114D9B74
64 #define MATROSKA_ID_CLUSTER 0x1F43B675
65
66 /* IDs in the info master */
67 #define MATROSKA_ID_TIMECODESCALE 0x2AD7B1
68 #define MATROSKA_ID_DURATION 0x4489
69 #define MATROSKA_ID_WRITINGAPP 0x5741
70 #define MATROSKA_ID_MUXINGAPP 0x4D80
71 #define MATROSKA_ID_DATEUTC 0x4461
72
73 /* ID in the tracks master */
74 #define MATROSKA_ID_TRACKENTRY 0xAE
75
76 /* IDs in the trackentry master */
77 #define MATROSKA_ID_TRACKNUMBER 0xD7
78 #define MATROSKA_ID_TRACKUID 0x73C5
79 #define MATROSKA_ID_TRACKTYPE 0x83
80 #define MATROSKA_ID_TRACKAUDIO 0xE1
81 #define MATROSKA_ID_TRACKVIDEO 0xE0
82 #define MATROSKA_ID_CODECID 0x86
83 #define MATROSKA_ID_CODECPRIVATE 0x63A2
84 #define MATROSKA_ID_CODECNAME 0x258688
85 #define MATROSKA_ID_CODECINFOURL 0x3B4040
86 #define MATROSKA_ID_CODECDOWNLOADURL 0x26B240
87 #define MATROSKA_ID_TRACKNAME 0x536E
88 #define MATROSKA_ID_TRACKLANGUAGE 0x22B59C
89 #define MATROSKA_ID_TRACKFLAGENABLED 0xB9
90 #define MATROSKA_ID_TRACKFLAGDEFAULT 0x88
91 #define MATROSKA_ID_TRACKFLAGLACING 0x9C
92 #define MATROSKA_ID_TRACKMINCACHE 0x6DE7
93 #define MATROSKA_ID_TRACKMAXCACHE 0x6DF8
94 #define MATROSKA_ID_TRACKDEFAULTDURATION 0x23E383
95
96 /* IDs in the trackvideo master */
97 #define MATROSKA_ID_VIDEOFRAMERATE 0x2383E3
98 #define MATROSKA_ID_VIDEODISPLAYWIDTH 0x54B0
99 #define MATROSKA_ID_VIDEODISPLAYHEIGHT 0x54BA
100 #define MATROSKA_ID_VIDEOPIXELWIDTH 0xB0
101 #define MATROSKA_ID_VIDEOPIXELHEIGHT 0xBA
102 #define MATROSKA_ID_VIDEOFLAGINTERLACED 0x9A
103 #define MATROSKA_ID_VIDEOSTEREOMODE 0x53B9
104 #define MATROSKA_ID_VIDEOASPECTRATIO 0x54B3
105 #define MATROSKA_ID_VIDEOCOLOURSPACE 0x2EB524
106
107 /* IDs in the trackaudio master */
108 #define MATROSKA_ID_AUDIOSAMPLINGFREQ 0xB5
109 #define MATROSKA_ID_AUDIOBITDEPTH 0x6264
110 #define MATROSKA_ID_AUDIOCHANNELS 0x9F
111
112 /* ID in the cues master */
113 #define MATROSKA_ID_POINTENTRY 0xBB
114
115 /* IDs in the pointentry master */
116 #define MATROSKA_ID_CUETIME 0xB3
117 #define MATROSKA_ID_CUETRACKPOSITION 0xB7
118
119 /* IDs in the cuetrackposition master */
120 #define MATROSKA_ID_CUETRACK 0xF7
121 #define MATROSKA_ID_CUECLUSTERPOSITION 0xF1
122
123 /* IDs in the tags master */
124 /* TODO */
125
126 /* IDs in the seekhead master */
127 #define MATROSKA_ID_SEEKENTRY 0x4DBB
128
129 /* IDs in the seekpoint master */
130 #define MATROSKA_ID_SEEKID 0x53AB
131 #define MATROSKA_ID_SEEKPOSITION 0x53AC
132
133 /* IDs in the cluster master */
134 #define MATROSKA_ID_CLUSTERTIMECODE 0xE7
135 #define MATROSKA_ID_BLOCKGROUP 0xA0
136
137 /* IDs in the blockgroup master */
138 #define MATROSKA_ID_BLOCK 0xA1
139 #define MATROSKA_ID_BLOCKDURATION 0x9B
140 #define MATROSKA_ID_BLOCKREFERENCE 0xFB
141
142 typedef enum {
143 MATROSKA_TRACK_TYPE_VIDEO = 0x1,
144 MATROSKA_TRACK_TYPE_AUDIO = 0x2,
145 MATROSKA_TRACK_TYPE_COMPLEX = 0x3,
146 MATROSKA_TRACK_TYPE_LOGO = 0x10,
147 MATROSKA_TRACK_TYPE_SUBTITLE = 0x11,
148 MATROSKA_TRACK_TYPE_CONTROL = 0x20,
149 } MatroskaTrackType;
150
151 typedef enum {
152 MATROSKA_EYE_MODE_MONO = 0x0,
153 MATROSKA_EYE_MODE_RIGHT = 0x1,
154 MATROSKA_EYE_MODE_LEFT = 0x2,
155 MATROSKA_EYE_MODE_BOTH = 0x3,
156 } MatroskaEyeMode;
157
158 typedef enum {
159 MATROSKA_ASPECT_RATIO_MODE_FREE = 0x0,
160 MATROSKA_ASPECT_RATIO_MODE_KEEP = 0x1,
161 MATROSKA_ASPECT_RATIO_MODE_FIXED = 0x2,
162 } MatroskaAspectRatioMode;
163
164 /*
165 * These aren't in any way "matroska-form" things,
166 * it's just something I use in the muxer/demuxer.
167 */
168
169 typedef enum {
170 MATROSKA_TRACK_ENABLED = (1<<0),
171 MATROSKA_TRACK_DEFAULT = (1<<1),
172 MATROSKA_TRACK_LACING = (1<<2),
173 MATROSKA_TRACK_SHIFT = (1<<16)
174 } MatroskaTrackFlags;
175
176 typedef enum {
177 MATROSKA_VIDEOTRACK_INTERLACED = (MATROSKA_TRACK_SHIFT<<0)
178 } MatroskaVideoTrackFlags;
179
180 /*
181 * Matroska Codec IDs. Strings.
182 */
183
184 #define MATROSKA_CODEC_ID_VIDEO_VFW_FOURCC "V_MS/VFW/FOURCC"
185 #define MATROSKA_CODEC_ID_VIDEO_UNCOMPRESSED "V_UNCOMPRESSED"
186 #define MATROSKA_CODEC_ID_VIDEO_MPEG4_SP "V_MPEG4/ISO/SP"
187 #define MATROSKA_CODEC_ID_VIDEO_MPEG4_ASP "V_MPEG4/ISO/ASP"
188 #define MATROSKA_CODEC_ID_VIDEO_MPEG4_AP "V_MPEG4/ISO/AP"
189 #define MATROSKA_CODEC_ID_VIDEO_MPEG4_AVC "V_MPEG4/ISO/AVC"
190 #define MATROSKA_CODEC_ID_VIDEO_MSMPEG4V3 "V_MPEG4/MS/V3"
191 #define MATROSKA_CODEC_ID_VIDEO_MPEG1 "V_MPEG1"
192 #define MATROSKA_CODEC_ID_VIDEO_MPEG2 "V_MPEG2"
193 #define MATROSKA_CODEC_ID_VIDEO_MJPEG "V_MJPEG"
194 /* TODO: Real/Quicktime */
195
196 #define MATROSKA_CODEC_ID_AUDIO_ACM "A_MS/ACM"
197 #define MATROSKA_CODEC_ID_AUDIO_MPEG1_L1 "A_MPEG/L1"
198 #define MATROSKA_CODEC_ID_AUDIO_MPEG1_L2 "A_MPEG/L2"
199 #define MATROSKA_CODEC_ID_AUDIO_MPEG1_L3 "A_MPEG/L3"
200 #define MATROSKA_CODEC_ID_AUDIO_PCM_INT_BE "A_PCM/INT/BIG"
201 #define MATROSKA_CODEC_ID_AUDIO_PCM_INT_LE "A_PCM/INT/LIT"
202 #define MATROSKA_CODEC_ID_AUDIO_PCM_FLOAT "A_PCM/FLOAT/IEEE"
203 #define MATROSKA_CODEC_ID_AUDIO_AC3 "A_AC3"
204 #define MATROSKA_CODEC_ID_AUDIO_DTS "A_DTS"
205 #define MATROSKA_CODEC_ID_AUDIO_VORBIS "A_VORBIS"
206 #define MATROSKA_CODEC_ID_AUDIO_ACM "A_MS/ACM"
207 #define MATROSKA_CODEC_ID_AUDIO_MPEG2 "A_AAC/MPEG2/"
208 #define MATROSKA_CODEC_ID_AUDIO_MPEG4 "A_AAC/MPEG4/"
209 /* TODO: AC3-9/10 (?), Real, Musepack, Quicktime */
210
211 /* max. depth in the EBML tree structure */
212 #define EBML_MAX_DEPTH 16
213
214 typedef struct Track {
215 MatroskaTrackType type;
216
217 /* Unique track number and track ID. stream_index is the index that
218 * the calling app uses for this track. */
219 uint32_t num,
220 uid,
221 stream_index;
222
223 char *name,
224 *language;
225
226 char *codec_id,
227 *codec_name;
228
229 unsigned char *codec_priv;
230 int codec_priv_size;
231
232 int64_t default_duration;
233 MatroskaTrackFlags flags;
234 } MatroskaTrack;
235
236 typedef struct MatroskaVideoTrack {
237 MatroskaTrack track;
238
239 int pixel_width,
240 pixel_height,
241 display_width,
242 display_height;
243
244 uint32_t fourcc;
245
246 MatroskaAspectRatioMode ar_mode;
247 MatroskaEyeMode eye_mode;
248
249 //..
250 } MatroskaVideoTrack;
251
252 typedef struct MatroskaAudioTrack {
253 MatroskaTrack track;
254
255 int channels,
256 bitdepth,
257 samplerate;
258 //..
259 } MatroskaAudioTrack;
260
261 typedef struct MatroskaSubtitleTrack {
262 MatroskaTrack track;
263
264 //..
265 } MatroskaSubtitleTrack;
266
267 typedef struct MatroskaLevel {
268 uint64_t start, length;
269 } MatroskaLevel;
270
271 typedef struct MatroskaDemuxIndex {
272 uint64_t pos; /* of the corresponding *cluster*! */
273 uint16_t track; /* reference to 'num' */
274 uint64_t time; /* in nanoseconds */
275 } MatroskaDemuxIndex;
276
277 typedef struct MatroskaDemuxContext {
278 AVFormatContext *ctx;
279
280 /* ebml stuff */
281 int num_levels;
282 MatroskaLevel levels[EBML_MAX_DEPTH];
283 int level_up;
284
285 /* matroska stuff */
286 char *writing_app,
287 *muxing_app;
288 int64_t created;
289
290 /* timescale in the file */
291 int64_t time_scale;
292
293 /* length, position (time, ns) */
294 int64_t duration,
295 pos;
296
297 /* num_streams is the number of streams that av_new_stream() was called
298 * for ( = that are available to the calling program). */
299 int num_tracks, num_streams;
300 MatroskaTrack *tracks[MAX_STREAMS];
301
302 /* cache for ID peeking */
303 uint32_t peek_id;
304
305 /* byte position of the segment inside the stream */
306 offset_t segment_start;
307
308 /* The packet queue. */
309 AVPacket **packets;
310 int num_packets;
311
312 /* have we already parse metadata/cues/clusters? */
313 int metadata_parsed,
314 index_parsed,
315 done;
316
317 /* The index for seeking. */
318 int num_indexes;
319 MatroskaDemuxIndex *index;
320 } MatroskaDemuxContext;
321
322 /*
323 * The first few functions handle EBML file parsing. The rest
324 * is the document interpretation. Matroska really just is a
325 * EBML file.
326 */
327
328 /*
329 * Return: the amount of levels in the hierarchy that the
330 * current element lies higher than the previous one.
331 * The opposite isn't done - that's auto-done using master
332 * element reading.
333 */
334
335 static int
336 ebml_read_element_level_up (MatroskaDemuxContext *matroska)
337 {
338 ByteIOContext *pb = &matroska->ctx->pb;
339 offset_t pos = url_ftell(pb);
340 int num = 0;
341
342 while (matroska->num_levels > 0) {
343 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
344
345 if (pos >= level->start + level->length) {
346 matroska->num_levels--;
347 num++;
348 } else {
349 break;
350 }
351 }
352
353 return num;
354 }
355
356 /*
357 * Read: an "EBML number", which is defined as a variable-length
358 * array of bytes. The first byte indicates the length by giving a
359 * number of 0-bits followed by a one. The position of the first
360 * "one" bit inside the first byte indicates the length of this
361 * number.
362 * Returns: num. of bytes read. < 0 on error.
363 */
364
365 static int
366 ebml_read_num (MatroskaDemuxContext *matroska,
367 int max_size,
368 uint64_t *number)
369 {
370 ByteIOContext *pb = &matroska->ctx->pb;
371 int len_mask = 0x80, read = 1, n = 1;
372 int64_t total = 0;
373
374 /* the first byte tells us the length in bytes - get_byte() can normally
375 * return 0, but since that's not a valid first ebmlID byte, we can
376 * use it safely here to catch EOS. */
377 if (!(total = get_byte(pb))) {
378 /* we might encounter EOS here */
379 if (!url_feof(pb)) {
380 offset_t pos = url_ftell(pb);
381 av_log(matroska->ctx, AV_LOG_ERROR,
382 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
383 pos, pos);
384 }
385 return AVERROR_IO; /* EOS or actual I/O error */
386 }
387
388 /* get the length of the EBML number */
389 while (read <= max_size && !(total & len_mask)) {
390 read++;
391 len_mask >>= 1;
392 }
393 if (read > max_size) {
394 offset_t pos = url_ftell(pb) - 1;
395 av_log(matroska->ctx, AV_LOG_ERROR,
396 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
397 (uint8_t) total, pos, pos);
398 return AVERROR_INVALIDDATA;
399 }
400
401 /* read out length */
402 total &= ~len_mask;
403 while (n++ < read)
404 total = (total << 8) | get_byte(pb);
405
406 *number = total;
407
408 return read;
409 }
410
411 /*
412 * Read: the element content data ID.
413 * Return: the number of bytes read or < 0 on error.
414 */
415
416 static int
417 ebml_read_element_id (MatroskaDemuxContext *matroska,
418 uint32_t *id,
419 int *level_up)
420 {
421 int read;
422 uint64_t total;
423
424 /* if we re-call this, use our cached ID */
425 if (matroska->peek_id != 0) {
426 if (level_up)
427 *level_up = 0;
428 *id = matroska->peek_id;
429 return 0;
430 }
431
432 /* read out the "EBML number", include tag in ID */
433 if ((read = ebml_read_num(matroska, 4, &total)) < 0)
434 return read;
435 *id = matroska->peek_id = total | (1 << (read * 7));
436
437 /* level tracking */
438 if (level_up)
439 *level_up = ebml_read_element_level_up(matroska);
440
441 return read;
442 }
443
444 /*
445 * Read: element content length.
446 * Return: the number of bytes read or < 0 on error.
447 */
448
449 static int
450 ebml_read_element_length (MatroskaDemuxContext *matroska,
451 uint64_t *length)
452 {
453 /* clear cache since we're now beyond that data point */
454 matroska->peek_id = 0;
455
456 /* read out the "EBML number", include tag in ID */
457 return ebml_read_num(matroska, 8, length);
458 }
459
460 /*
461 * Return: the ID of the next element, or 0 on error.
462 * Level_up contains the amount of levels that this
463 * next element lies higher than the previous one.
464 */
465
466 static uint32_t
467 ebml_peek_id (MatroskaDemuxContext *matroska,
468 int *level_up)
469 {
470 uint32_t id;
471
472 assert(level_up != NULL);
473
474 if (ebml_read_element_id(matroska, &id, level_up) < 0)
475 return 0;
476
477 return id;
478 }
479
480 /*
481 * Seek to a given offset.
482 * 0 is success, -1 is failure.
483 */
484
485 static int
486 ebml_read_seek (MatroskaDemuxContext *matroska,
487 offset_t offset)
488 {
489 ByteIOContext *pb = &matroska->ctx->pb;
490
491 /* clear ID cache, if any */
492 matroska->peek_id = 0;
493
494 return (url_fseek(pb, offset, SEEK_SET) == offset) ? 0 : -1;
495 }
496
497 /*
498 * Skip the next element.
499 * 0 is success, -1 is failure.
500 */
501
502 static int
503 ebml_read_skip (MatroskaDemuxContext *matroska)
504 {
505 ByteIOContext *pb = &matroska->ctx->pb;
506 uint32_t id;
507 uint64_t length;
508 int res;
509
510 if ((res = ebml_read_element_id(matroska, &id, NULL)) < 0 ||
511 (res = ebml_read_element_length(matroska, &length)) < 0)
512 return res;
513
514 url_fskip(pb, length);
515
516 return 0;
517 }
518
519 /*
520 * Read the next element as an unsigned int.
521 * 0 is success, < 0 is failure.
522 */
523
524 static int
525 ebml_read_uint (MatroskaDemuxContext *matroska,
526 uint32_t *id,
527 uint64_t *num)
528 {
529 ByteIOContext *pb = &matroska->ctx->pb;
530 int n = 0, size, res;
531 uint64_t rlength;
532
533 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
534 (res = ebml_read_element_length(matroska, &rlength)) < 0)
535 return res;
536 size = rlength;
537 if (size < 1 || size > 8) {
538 offset_t pos = url_ftell(pb);
539 av_log(matroska->ctx, AV_LOG_ERROR,
540 "Invalid uint element size %d at position %"PRId64" (0x%"PRIx64")\n",
541 size, pos, pos);
542 return AVERROR_INVALIDDATA;
543 }
544
545 /* big-endian ordening; build up number */
546 *num = 0;
547 while (n++ < size)
548 *num = (*num << 8) | get_byte(pb);
549
550 return 0;
551 }
552
553 /*
554 * Read the next element as a signed int.
555 * 0 is success, < 0 is failure.
556 */
557
558 static int
559 ebml_read_sint (MatroskaDemuxContext *matroska,
560 uint32_t *id,
561 int64_t *num)
562 {
563 ByteIOContext *pb = &matroska->ctx->pb;
564 int size, n = 1, negative = 0, res;
565 uint64_t rlength;
566
567 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
568 (res = ebml_read_element_length(matroska, &rlength)) < 0)
569 return res;
570 size = rlength;
571 if (size < 1 || size > 8) {
572 offset_t pos = url_ftell(pb);
573 av_log(matroska->ctx, AV_LOG_ERROR,
574 "Invalid sint element size %d at position %"PRId64" (0x%"PRIx64")\n",
575 size, pos, pos);
576 return AVERROR_INVALIDDATA;
577 }
578 if ((*num = get_byte(pb)) & 0x80) {
579 negative = 1;
580 *num &= ~0x80;
581 }
582 *num = 0;
583 while (n++ < size)
584 *num = (*num << 8) | get_byte(pb);
585
586 /* make signed */
587 if (negative)
588 *num = *num - (1LL << ((8 * size) - 1));
589
590 return 0;
591 }
592
593 /*
594 * Read the next element as a float.
595 * 0 is success, < 0 is failure.
596 */
597
598 static int
599 ebml_read_float (MatroskaDemuxContext *matroska,
600 uint32_t *id,
601 double *num)
602 {
603 ByteIOContext *pb = &matroska->ctx->pb;
604 int size, res;
605 uint64_t rlength;
606
607 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
608 (res = ebml_read_element_length(matroska, &rlength)) < 0)
609 return res;
610 size = rlength;
611
612 if (size == 4) {
613 *num= av_int2flt(get_be32(pb));
614 } else if(size==8){
615 *num= av_int2dbl(get_be64(pb));
616 } else if(size==10){
617 av_log(matroska->ctx, AV_LOG_ERROR,
618 "FIXME! 10-byte floats unimplemented\n");
619 return AVERROR_UNKNOWN;
620 } else{
621 offset_t pos = url_ftell(pb);
622 av_log(matroska->ctx, AV_LOG_ERROR,
623 "Invalid float element size %d at position %"PRIu64" (0x%"PRIx64")\n",
624 size, pos, pos);
625 return AVERROR_INVALIDDATA;
626 }
627
628 return 0;
629 }
630
631 /*
632 * Read the next element as an ASCII string.
633 * 0 is success, < 0 is failure.
634 */
635
636 static int
637 ebml_read_ascii (MatroskaDemuxContext *matroska,
638 uint32_t *id,
639 char **str)
640 {
641 ByteIOContext *pb = &matroska->ctx->pb;
642 int size, res;
643 uint64_t rlength;
644
645 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
646 (res = ebml_read_element_length(matroska, &rlength)) < 0)
647 return res;
648 size = rlength;
649
650 /* ebml strings are usually not 0-terminated, so we allocate one
651 * byte more, read the string and NULL-terminate it ourselves. */
652 if (size < 0 || !(*str = av_malloc(size + 1))) {
653 av_log(matroska->ctx, AV_LOG_ERROR, "Memory allocation failed\n");
654 return AVERROR_NOMEM;
655 }
656 if (get_buffer(pb, (uint8_t *) *str, size) != size) {
657 offset_t pos = url_ftell(pb);
658 av_log(matroska->ctx, AV_LOG_ERROR,
659 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
660 return AVERROR_IO;
661 }
662 (*str)[size] = '\0';
663
664 return 0;
665 }
666
667 /*
668 * Read the next element as a UTF-8 string.
669 * 0 is success, < 0 is failure.
670 */
671
672 static int
673 ebml_read_utf8 (MatroskaDemuxContext *matroska,
674 uint32_t *id,
675 char **str)
676 {
677 return ebml_read_ascii(matroska, id, str);
678 }
679
680 /*
681 * Read the next element as a date (nanoseconds since 1/1/2000).
682 * 0 is success, < 0 is failure.
683 */
684
685 static int
686 ebml_read_date (MatroskaDemuxContext *matroska,
687 uint32_t *id,
688 int64_t *date)
689 {
690 return ebml_read_sint(matroska, id, date);
691 }
692
693 /*
694 * Read the next element, but only the header. The contents
695 * are supposed to be sub-elements which can be read separately.
696 * 0 is success, < 0 is failure.
697 */
698
699 static int
700 ebml_read_master (MatroskaDemuxContext *matroska,
701 uint32_t *id)
702 {
703 ByteIOContext *pb = &matroska->ctx->pb;
704 uint64_t length;
705 MatroskaLevel *level;
706 int res;
707
708 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
709 (res = ebml_read_element_length(matroska, &length)) < 0)
710 return res;
711
712 /* protect... (Heaven forbids that the '>' is true) */
713 if (matroska->num_levels >= EBML_MAX_DEPTH) {
714 av_log(matroska->ctx, AV_LOG_ERROR,
715 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
716 return AVERROR_NOTSUPP;
717 }
718
719 /* remember level */
720 level = &matroska->levels[matroska->num_levels++];
721 level->start = url_ftell(pb);
722 level->length = length;
723
724 return 0;
725 }
726
727 /*
728 * Read the next element as binary data.
729 * 0 is success, < 0 is failure.
730 */
731
732 static int
733 ebml_read_binary (MatroskaDemuxContext *matroska,
734 uint32_t *id,
735 uint8_t **binary,
736 int *size)
737 {
738 ByteIOContext *pb = &matroska->ctx->pb;
739 uint64_t rlength;
740 int res;
741
742 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
743 (res = ebml_read_element_length(matroska, &rlength)) < 0)
744 return res;
745 *size = rlength;
746
747 if (!(*binary = av_malloc(*size))) {
748 av_log(matroska->ctx, AV_LOG_ERROR,
749 "Memory allocation error\n");
750 return AVERROR_NOMEM;
751 }
752
753 if (get_buffer(pb, *binary, *size) != *size) {
754 offset_t pos = url_ftell(pb);
755 av_log(matroska->ctx, AV_LOG_ERROR,
756 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
757 return AVERROR_IO;
758 }
759
760 return 0;
761 }
762
763 /*
764 * Read signed/unsigned "EBML" numbers.
765 * Return: number of bytes processed, < 0 on error.
766 * XXX: use ebml_read_num().
767 */
768
769 static int
770 matroska_ebmlnum_uint (uint8_t *data,
771 uint32_t size,
772 uint64_t *num)
773 {
774 int len_mask = 0x80, read = 1, n = 1, num_ffs = 0;
775 uint64_t total;
776
777 if (size <= 0)
778 return AVERROR_INVALIDDATA;
779
780 total = data[0];
781 while (read <= 8 && !(total & len_mask)) {
782 read++;
783 len_mask >>= 1;
784 }
785 if (read > 8)
786 return AVERROR_INVALIDDATA;
787
788 if ((total &= (len_mask - 1)) == len_mask - 1)
789 num_ffs++;
790 if (size < read)
791 return AVERROR_INVALIDDATA;
792 while (n < read) {
793 if (data[n] == 0xff)
794 num_ffs++;
795 total = (total << 8) | data[n];
796 n++;
797 }
798
799 if (read == num_ffs)
800 *num = (uint64_t)-1;
801 else
802 *num = total;
803
804 return read;
805 }
806
807 /*
808 * Same as above, but signed.
809 */
810
811 static int
812 matroska_ebmlnum_sint (uint8_t *data,
813 uint32_t size,
814 int64_t *num)
815 {
816 uint64_t unum;
817 int res;
818
819 /* read as unsigned number first */
820 if ((res = matroska_ebmlnum_uint(data, size, &unum)) < 0)
821 return res;
822
823 /* make signed (weird way) */
824 if (unum == (uint64_t)-1)
825 *num = INT64_MAX;
826 else
827 *num = unum - ((1LL << ((7 * res) - 1)) - 1);
828
829 return res;
830 }
831
832 /*
833 * Read an EBML header.
834 * 0 is success, < 0 is failure.
835 */
836
837 static int
838 ebml_read_header (MatroskaDemuxContext *matroska,
839 char **doctype,
840 int *version)
841 {
842 uint32_t id;
843 int level_up, res = 0;
844
845 /* default init */
846 if (doctype)
847 *doctype = NULL;
848 if (version)
849 *version = 1;
850
851 if (!(id = ebml_peek_id(matroska, &level_up)) ||
852 level_up != 0 || id != EBML_ID_HEADER) {
853 av_log(matroska->ctx, AV_LOG_ERROR,
854 "This is not an EBML file (id=0x%x/0x%x)\n", id, EBML_ID_HEADER);
855 return AVERROR_INVALIDDATA;
856 }
857 if ((res = ebml_read_master(matroska, &id)) < 0)
858 return res;
859
860 while (res == 0) {
861 if (!(id = ebml_peek_id(matroska, &level_up)))
862 return AVERROR_IO;
863
864 /* end-of-header */
865 if (level_up)
866 break;
867
868 switch (id) {
869 /* is our read version uptodate? */
870 case EBML_ID_EBMLREADVERSION: {
871 uint64_t num;
872
873 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
874 return res;
875 if (num > EBML_VERSION) {
876 av_log(matroska->ctx, AV_LOG_ERROR,
877 "EBML version %"PRIu64" (> %d) is not supported\n",
878 num, EBML_VERSION);
879 return AVERROR_INVALIDDATA;
880 }
881 break;
882 }
883
884 /* we only handle 8 byte lengths at max */
885 case EBML_ID_EBMLMAXSIZELENGTH: {
886 uint64_t num;
887
888 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
889 return res;
890 if (num > sizeof(uint64_t)) {
891 av_log(matroska->ctx, AV_LOG_ERROR,
892 "Integers of size %"PRIu64" (> %zd) not supported\n",
893 num, sizeof(uint64_t));
894 return AVERROR_INVALIDDATA;
895 }
896 break;
897 }
898
899 /* we handle 4 byte IDs at max */
900 case EBML_ID_EBMLMAXIDLENGTH: {
901 uint64_t num;
902
903 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
904 return res;
905 if (num > sizeof(uint32_t)) {
906 av_log(matroska->ctx, AV_LOG_ERROR,
907 "IDs of size %"PRIu64" (> %zu) not supported\n",
908 num, sizeof(uint32_t));
909 return AVERROR_INVALIDDATA;
910 }
911 break;
912 }
913
914 case EBML_ID_DOCTYPE: {
915 char *text;
916
917 if ((res = ebml_read_ascii(matroska, &id, &text)) < 0)
918 return res;
919 if (doctype) {
920 if (*doctype)
921 av_free(*doctype);
922 *doctype = text;
923 } else
924 av_free(text);
925 break;
926 }
927
928 case EBML_ID_DOCTYPEREADVERSION: {
929 uint64_t num;
930
931 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
932 return res;
933 if (version)
934 *version = num;
935 break;
936 }
937
938 default:
939 av_log(matroska->ctx, AV_LOG_INFO,
940 "Unknown data type 0x%x in EBML header", id);
941 /* pass-through */
942
943 case EBML_ID_VOID:
944 /* we ignore these two, as they don't tell us anything we
945 * care about */
946 case EBML_ID_EBMLVERSION:
947 case EBML_ID_DOCTYPEVERSION:
948 res = ebml_read_skip (matroska);
949 break;
950 }
951 }
952
953 return 0;
954 }
955
956 /*
957 * Put one packet in an application-supplied AVPacket struct.
958 * Returns 0 on success or -1 on failure.
959 */
960
961 static int
962 matroska_deliver_packet (MatroskaDemuxContext *matroska,
963 AVPacket *pkt)
964 {
965 if (matroska->num_packets > 0) {
966 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
967 av_free(matroska->packets[0]);
968 if (matroska->num_packets > 1) {
969 memmove(&matroska->packets[0], &matroska->packets[1],
970 (matroska->num_packets - 1) * sizeof(AVPacket *));
971 matroska->packets =
972 av_realloc(matroska->packets, (matroska->num_packets - 1) *
973 sizeof(AVPacket *));
974 } else {
975 av_free(matroska->packets);
976 matroska->packets = NULL;
977 }
978 matroska->num_packets--;
979 return 0;
980 }
981
982 return -1;
983 }
984
985 /*
986 * Put a packet into our internal queue. Will be delivered to the
987 * user/application during the next get_packet() call.
988 */
989
990 static void
991 matroska_queue_packet (MatroskaDemuxContext *matroska,
992 AVPacket *pkt)
993 {
994 matroska->packets =
995 av_realloc(matroska->packets, (matroska->num_packets + 1) *
996 sizeof(AVPacket *));
997 matroska->packets[matroska->num_packets] = pkt;
998 matroska->num_packets++;
999 }
1000
1001 /*
1002 * Autodetecting...
1003 */
1004
1005 static int
1006 matroska_probe (AVProbeData *p)
1007 {
1008 uint64_t total = 0;
1009 int len_mask = 0x80, size = 1, n = 1;
1010 uint8_t probe_data[] = { 'm', 'a', 't', 'r', 'o', 's', 'k', 'a' };
1011
1012 if (p->buf_size < 5)
1013 return 0;
1014
1015 /* ebml header? */
1016 if ((p->buf[0] << 24 | p->buf[1] << 16 |
1017 p->buf[2] << 8 | p->buf[3]) != EBML_ID_HEADER)
1018 return 0;
1019
1020 /* length of header */
1021 total = p->buf[4];
1022 while (size <= 8 && !(total & len_mask)) {
1023 size++;
1024 len_mask >>= 1;
1025 }
1026 if (size > 8)
1027 return 0;
1028 total &= (len_mask - 1);
1029 while (n < size)
1030 total = (total << 8) | p->buf[4 + n++];
1031
1032 /* does the probe data contain the whole header? */
1033 if (p->buf_size < 4 + size + total)
1034 return 0;
1035
1036 /* the header must contain the document type 'matroska'. For now,
1037 * we don't parse the whole header but simply check for the
1038 * availability of that array of characters inside the header.
1039 * Not fully fool-proof, but good enough. */
1040 for (n = 4 + size; n < 4 + size + total - sizeof(probe_data); n++)
1041 if (!memcmp (&p->buf[n], probe_data, sizeof(probe_data)))
1042 return AVPROBE_SCORE_MAX;
1043
1044 return 0;
1045 }
1046
1047 /*
1048 * From here on, it's all XML-style DTD stuff... Needs no comments.
1049 */
1050
1051 static int
1052 matroska_parse_info (MatroskaDemuxContext *matroska)
1053 {
1054 int res = 0;
1055 uint32_t id;
1056
1057 av_log(matroska->ctx, AV_LOG_DEBUG, "Parsing info...\n");
1058
1059 while (res == 0) {
1060 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1061 res = AVERROR_IO;
1062 break;
1063 } else if (matroska->level_up) {
1064 matroska->level_up--;
1065 break;
1066 }
1067
1068 switch (id) {
1069 /* cluster timecode */
1070 case MATROSKA_ID_TIMECODESCALE: {
1071 uint64_t num;
1072 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1073 break;
1074 matroska->time_scale = num;
1075 break;
1076 }
1077
1078 case MATROSKA_ID_DURATION: {
1079 double num;
1080 if ((res = ebml_read_float(matroska, &id, &num)) < 0)
1081 break;
1082 matroska->duration = num * matroska->time_scale;
1083 break;
1084 }
1085
1086 case MATROSKA_ID_WRITINGAPP: {
1087 char *text;
1088 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1089 break;
1090 matroska->writing_app = text;
1091 break;
1092 }
1093
1094 case MATROSKA_ID_MUXINGAPP: {
1095 char *text;
1096 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1097 break;
1098 matroska->muxing_app = text;
1099 break;
1100 }
1101
1102 case MATROSKA_ID_DATEUTC: {
1103 int64_t time;
1104 if ((res = ebml_read_date(matroska, &id, &time)) < 0)
1105 break;
1106 matroska->created = time;
1107 break;
1108 }
1109
1110 default:
1111 av_log(matroska->ctx, AV_LOG_INFO,
1112 "Unknown entry 0x%x in info header\n", id);
1113 /* fall-through */
1114
1115 case EBML_ID_VOID:
1116 res = ebml_read_skip(matroska);
1117 break;
1118 }
1119
1120 if (matroska->level_up) {
1121 matroska->level_up--;
1122 break;
1123 }
1124 }
1125
1126 return res;
1127 }
1128
1129 static int
1130 matroska_add_stream (MatroskaDemuxContext *matroska)
1131 {
1132 int res = 0;
1133 uint32_t id;
1134 MatroskaTrack *track;
1135
1136 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing track, adding stream..,\n");
1137
1138 /* Allocate a generic track. As soon as we know its type we'll realloc. */
1139 track = av_mallocz(sizeof(MatroskaTrack));
1140 matroska->num_tracks++;
1141
1142 /* start with the master */
1143 if ((res = ebml_read_master(matroska, &id)) < 0)
1144 return res;
1145
1146 /* try reading the trackentry headers */
1147 while (res == 0) {
1148 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1149 res = AVERROR_IO;
1150 break;
1151 } else if (matroska->level_up > 0) {
1152 matroska->level_up--;
1153 break;
1154 }
1155
1156 switch (id) {
1157 /* track number (unique stream ID) */
1158 case MATROSKA_ID_TRACKNUMBER: {
1159 uint64_t num;
1160 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1161 break;
1162 track->num = num;
1163 break;
1164 }
1165
1166 /* track UID (unique identifier) */
1167 case MATROSKA_ID_TRACKUID: {
1168 uint64_t num;
1169 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1170 break;
1171 track->uid = num;
1172 break;
1173 }
1174
1175 /* track type (video, audio, combined, subtitle, etc.) */
1176 case MATROSKA_ID_TRACKTYPE: {
1177 uint64_t num;
1178 if (track->type != 0) {
1179 av_log(matroska->ctx, AV_LOG_INFO,
1180 "More than one tracktype in an entry - skip\n");
1181 break;
1182 }
1183 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1184 break;
1185 track->type = num;
1186
1187 /* ok, so we're actually going to reallocate this thing */
1188 switch (track->type) {
1189 case MATROSKA_TRACK_TYPE_VIDEO:
1190 track = (MatroskaTrack *)
1191 av_realloc(track, sizeof(MatroskaVideoTrack));
1192 break;
1193 case MATROSKA_TRACK_TYPE_AUDIO:
1194 track = (MatroskaTrack *)
1195 av_realloc(track, sizeof(MatroskaAudioTrack));
1196 ((MatroskaAudioTrack *)track)->channels = 1;
1197 ((MatroskaAudioTrack *)track)->samplerate = 8000;
1198 break;
1199 case MATROSKA_TRACK_TYPE_SUBTITLE:
1200 track = (MatroskaTrack *)
1201 av_realloc(track, sizeof(MatroskaSubtitleTrack));
1202 break;
1203 case MATROSKA_TRACK_TYPE_COMPLEX:
1204 case MATROSKA_TRACK_TYPE_LOGO:
1205 case MATROSKA_TRACK_TYPE_CONTROL:
1206 default:
1207 av_log(matroska->ctx, AV_LOG_INFO,
1208 "Unknown or unsupported track type 0x%x\n",
1209 track->type);
1210 track->type = 0;
1211 break;
1212 }
1213 matroska->tracks[matroska->num_tracks - 1] = track;
1214 break;
1215 }
1216
1217 /* tracktype specific stuff for video */
1218 case MATROSKA_ID_TRACKVIDEO: {
1219 MatroskaVideoTrack *videotrack;
1220 if (track->type != MATROSKA_TRACK_TYPE_VIDEO) {
1221 av_log(matroska->ctx, AV_LOG_INFO,
1222 "video data in non-video track - ignoring\n");
1223 res = AVERROR_INVALIDDATA;
1224 break;
1225 } else if ((res = ebml_read_master(matroska, &id)) < 0)
1226 break;
1227 videotrack = (MatroskaVideoTrack *)track;
1228
1229 while (res == 0) {
1230 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1231 res = AVERROR_IO;
1232 break;
1233 } else if (matroska->level_up > 0) {
1234 matroska->level_up--;
1235 break;
1236 }
1237
1238 switch (id) {
1239 /* fixme, this should be one-up, but I get it here */
1240 case MATROSKA_ID_TRACKDEFAULTDURATION: {
1241 uint64_t num;
1242 if ((res = ebml_read_uint (matroska, &id,
1243 &num)) < 0)
1244 break;
1245 track->default_duration = num;
1246 break;
1247 }
1248
1249 /* video framerate */
1250 case MATROSKA_ID_VIDEOFRAMERATE: {
1251 double num;
1252 if ((res = ebml_read_float(matroska, &id,
1253 &num)) < 0)
1254 break;
1255 track->default_duration = 1000000000 * (1. / num);
1256 break;
1257 }
1258
1259 /* width of the size to display the video at */
1260 case MATROSKA_ID_VIDEODISPLAYWIDTH: {
1261 uint64_t num;
1262 if ((res = ebml_read_uint(matroska, &id,
1263 &num)) < 0)
1264 break;
1265 videotrack->display_width = num;
1266 break;
1267 }
1268
1269 /* height of the size to display the video at */
1270 case MATROSKA_ID_VIDEODISPLAYHEIGHT: {
1271 uint64_t num;
1272 if ((res = ebml_read_uint(matroska, &id,
1273 &num)) < 0)
1274 break;
1275 videotrack->display_height = num;
1276 break;
1277 }
1278
1279 /* width of the video in the file */
1280 case MATROSKA_ID_VIDEOPIXELWIDTH: {
1281 uint64_t num;
1282 if ((res = ebml_read_uint(matroska, &id,
1283 &num)) < 0)
1284 break;
1285 videotrack->pixel_width = num;
1286 break;
1287 }
1288
1289 /* height of the video in the file */
1290 case MATROSKA_ID_VIDEOPIXELHEIGHT: {
1291 uint64_t num;
1292 if ((res = ebml_read_uint(matroska, &id,
1293 &num)) < 0)
1294 break;
1295 videotrack->pixel_height = num;
1296 break;
1297 }
1298
1299 /* whether the video is interlaced */
1300 case MATROSKA_ID_VIDEOFLAGINTERLACED: {
1301 uint64_t num;
1302 if ((res = ebml_read_uint(matroska, &id,
1303 &num)) < 0)
1304 break;
1305 if (num)
1306 track->flags |=
1307 MATROSKA_VIDEOTRACK_INTERLACED;
1308 else
1309 track->flags &=
1310 ~MATROSKA_VIDEOTRACK_INTERLACED;
1311 break;
1312 }
1313
1314 /* stereo mode (whether the video has two streams,
1315 * where one is for the left eye and the other for
1316 * the right eye, which creates a 3D-like
1317 * effect) */
1318 case MATROSKA_ID_VIDEOSTEREOMODE: {
1319 uint64_t num;
1320 if ((res = ebml_read_uint(matroska, &id,
1321 &num)) < 0)
1322 break;
1323 if (num != MATROSKA_EYE_MODE_MONO &&
1324 num != MATROSKA_EYE_MODE_LEFT &&
1325 num != MATROSKA_EYE_MODE_RIGHT &&
1326 num != MATROSKA_EYE_MODE_BOTH) {
1327 av_log(matroska->ctx, AV_LOG_INFO,
1328 "Ignoring unknown eye mode 0x%x\n",
1329 (uint32_t) num);
1330 break;
1331 }
1332 videotrack->eye_mode = num;
1333 break;
1334 }
1335
1336 /* aspect ratio behaviour */
1337 case MATROSKA_ID_VIDEOASPECTRATIO: {
1338 uint64_t num;
1339 if ((res = ebml_read_uint(matroska, &id,
1340 &num)) < 0)
1341 break;
1342 if (num != MATROSKA_ASPECT_RATIO_MODE_FREE &&
1343 num != MATROSKA_ASPECT_RATIO_MODE_KEEP &&
1344 num != MATROSKA_ASPECT_RATIO_MODE_FIXED) {
1345 av_log(matroska->ctx, AV_LOG_INFO,
1346 "Ignoring unknown aspect ratio 0x%x\n",
1347 (uint32_t) num);
1348 break;
1349 }
1350 videotrack->ar_mode = num;
1351 break;
1352 }
1353
1354 /* colourspace (only matters for raw video)
1355 * fourcc */
1356 case MATROSKA_ID_VIDEOCOLOURSPACE: {
1357 uint64_t num;
1358 if ((res = ebml_read_uint(matroska, &id,
1359 &num)) < 0)
1360 break;
1361 videotrack->fourcc = num;
1362 break;
1363 }
1364
1365 default:
1366 av_log(matroska->ctx, AV_LOG_INFO,
1367 "Unknown video track header entry "
1368 "0x%x - ignoring\n", id);
1369 /* pass-through */
1370
1371 case EBML_ID_VOID:
1372 res = ebml_read_skip(matroska);
1373 break;
1374 }
1375
1376 if (matroska->level_up) {
1377 matroska->level_up--;
1378 break;
1379 }
1380 }
1381 break;
1382 }
1383
1384 /* tracktype specific stuff for audio */
1385 case MATROSKA_ID_TRACKAUDIO: {
1386 MatroskaAudioTrack *audiotrack;
1387 if (track->type != MATROSKA_TRACK_TYPE_AUDIO) {
1388 av_log(matroska->ctx, AV_LOG_INFO,
1389 "audio data in non-audio track - ignoring\n");
1390 res = AVERROR_INVALIDDATA;
1391 break;
1392 } else if ((res = ebml_read_master(matroska, &id)) < 0)
1393 break;
1394 audiotrack = (MatroskaAudioTrack *)track;
1395
1396 while (res == 0) {
1397 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1398 res = AVERROR_IO;
1399 break;
1400 } else if (matroska->level_up > 0) {
1401 matroska->level_up--;
1402 break;
1403 }
1404
1405 switch (id) {
1406 /* samplerate */
1407 case MATROSKA_ID_AUDIOSAMPLINGFREQ: {
1408 double num;
1409 if ((res = ebml_read_float(matroska, &id,
1410 &num)) < 0)
1411 break;
1412 audiotrack->samplerate = num;
1413 break;
1414 }
1415
1416 /* bitdepth */
1417 case MATROSKA_ID_AUDIOBITDEPTH: {
1418 uint64_t num;
1419 if ((res = ebml_read_uint(matroska, &id,
1420 &num)) < 0)
1421 break;
1422 audiotrack->bitdepth = num;
1423 break;
1424 }
1425
1426 /* channels */
1427 case MATROSKA_ID_AUDIOCHANNELS: {
1428 uint64_t num;
1429 if ((res = ebml_read_uint(matroska, &id,
1430 &num)) < 0)
1431 break;
1432 audiotrack->channels = num;
1433 break;
1434 }
1435
1436 default:
1437 av_log(matroska->ctx, AV_LOG_INFO,
1438 "Unknown audio track header entry "
1439 "0x%x - ignoring\n", id);
1440 /* pass-through */
1441
1442 case EBML_ID_VOID:
1443 res = ebml_read_skip(matroska);
1444 break;
1445 }
1446
1447 if (matroska->level_up) {
1448 matroska->level_up--;
1449 break;
1450 }
1451 }
1452 break;
1453 }
1454
1455 /* codec identifier */
1456 case MATROSKA_ID_CODECID: {
1457 char *text;
1458 if ((res = ebml_read_ascii(matroska, &id, &text)) < 0)
1459 break;
1460 track->codec_id = text;
1461 break;
1462 }
1463
1464 /* codec private data */
1465 case MATROSKA_ID_CODECPRIVATE: {
1466 uint8_t *data;
1467 int size;
1468 if ((res = ebml_read_binary(matroska, &id, &data, &size) < 0))
1469 break;
1470 track->codec_priv = data;
1471 track->codec_priv_size = size;
1472 break;
1473 }
1474
1475 /* name of the codec */
1476 case MATROSKA_ID_CODECNAME: {
1477 char *text;
1478 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1479 break;
1480 track->codec_name = text;
1481 break;
1482 }
1483
1484 /* name of this track */
1485 case MATROSKA_ID_TRACKNAME: {
1486 char *text;
1487 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1488 break;
1489 track->name = text;
1490 break;
1491 }
1492
1493 /* language (matters for audio/subtitles, mostly) */
1494 case MATROSKA_ID_TRACKLANGUAGE: {
1495 char *text;
1496 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1497 break;
1498 track->language = text;
1499 break;
1500 }
1501
1502 /* whether this is actually used */
1503 case MATROSKA_ID_TRACKFLAGENABLED: {
1504 uint64_t num;
1505 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1506 break;
1507 if (num)
1508 track->flags |= MATROSKA_TRACK_ENABLED;
1509 else
1510 track->flags &= ~MATROSKA_TRACK_ENABLED;
1511 break;
1512 }
1513
1514 /* whether it's the default for this track type */
1515 case MATROSKA_ID_TRACKFLAGDEFAULT: {
1516 uint64_t num;
1517 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1518 break;
1519 if (num)
1520 track->flags |= MATROSKA_TRACK_DEFAULT;
1521 else
1522 track->flags &= ~MATROSKA_TRACK_DEFAULT;
1523 break;
1524 }
1525
1526 /* lacing (like MPEG, where blocks don't end/start on frame
1527 * boundaries) */
1528 case MATROSKA_ID_TRACKFLAGLACING: {
1529 uint64_t num;
1530 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1531 break;
1532 if (num)
1533 track->flags |= MATROSKA_TRACK_LACING;
1534 else
1535 track->flags &= ~MATROSKA_TRACK_LACING;
1536 break;
1537 }
1538
1539 /* default length (in time) of one data block in this track */
1540 case MATROSKA_ID_TRACKDEFAULTDURATION: {
1541 uint64_t num;
1542 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1543 break;
1544 track->default_duration = num;
1545 break;
1546 }
1547
1548 default:
1549 av_log(matroska->ctx, AV_LOG_INFO,
1550 "Unknown track header entry 0x%x - ignoring\n", id);
1551 /* pass-through */
1552
1553 case EBML_ID_VOID:
1554 /* we ignore these because they're nothing useful. */
1555 case MATROSKA_ID_CODECINFOURL:
1556 case MATROSKA_ID_CODECDOWNLOADURL:
1557 case MATROSKA_ID_TRACKMINCACHE:
1558 case MATROSKA_ID_TRACKMAXCACHE:
1559 res = ebml_read_skip(matroska);
1560 break;
1561 }
1562
1563 if (matroska->level_up) {
1564 matroska->level_up--;
1565 break;
1566 }
1567 }
1568
1569 return res;
1570 }
1571
1572 static int
1573 matroska_parse_tracks (MatroskaDemuxContext *matroska)
1574 {
1575 int res = 0;
1576 uint32_t id;
1577
1578 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing tracks...\n");
1579
1580 while (res == 0) {
1581 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1582 res = AVERROR_IO;
1583 break;
1584 } else if (matroska->level_up) {
1585 matroska->level_up--;
1586 break;
1587 }
1588
1589 switch (id) {
1590 /* one track within the "all-tracks" header */
1591 case MATROSKA_ID_TRACKENTRY:
1592 res = matroska_add_stream(matroska);
1593 break;
1594
1595 default:
1596 av_log(matroska->ctx, AV_LOG_INFO,
1597 "Unknown entry 0x%x in track header\n", id);
1598 /* fall-through */
1599
1600 case EBML_ID_VOID:
1601 res = ebml_read_skip(matroska);
1602 break;
1603 }
1604
1605 if (matroska->level_up) {
1606 matroska->level_up--;
1607 break;
1608 }
1609 }
1610
1611 return res;
1612 }
1613
1614 static int
1615 matroska_parse_index (MatroskaDemuxContext *matroska)
1616 {
1617 int res = 0;
1618 uint32_t id;
1619 MatroskaDemuxIndex idx;
1620
1621 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing index...\n");
1622
1623 while (res == 0) {
1624 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1625 res = AVERROR_IO;
1626 break;
1627 } else if (matroska->level_up) {
1628 matroska->level_up--;
1629 break;
1630 }
1631
1632 switch (id) {
1633 /* one single index entry ('point') */
1634 case MATROSKA_ID_POINTENTRY:
1635 if ((res = ebml_read_master(matroska, &id)) < 0)
1636 break;
1637
1638 /* in the end, we hope to fill one entry with a
1639 * timestamp, a file position and a tracknum */
1640 idx.pos = (uint64_t) -1;
1641 idx.time = (uint64_t) -1;
1642 idx.track = (uint16_t) -1;
1643
1644 while (res == 0) {
1645 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1646 res = AVERROR_IO;
1647 break;
1648 } else if (matroska->level_up) {
1649 matroska->level_up--;
1650 break;
1651 }
1652
1653 switch (id) {
1654 /* one single index entry ('point') */
1655 case MATROSKA_ID_CUETIME: {
1656 int64_t time;
1657 if ((res = ebml_read_uint(matroska, &id,
1658 &time)) < 0)
1659 break;
1660 idx.time = time * matroska->time_scale;
1661 break;
1662 }
1663
1664 /* position in the file + track to which it
1665 * belongs */
1666 case MATROSKA_ID_CUETRACKPOSITION:
1667 if ((res = ebml_read_master(matroska, &id)) < 0)
1668 break;
1669
1670 while (res == 0) {
1671 if (!(id = ebml_peek_id (matroska,
1672 &matroska->level_up))) {
1673 res = AVERROR_IO;
1674 break;
1675 } else if (matroska->level_up) {
1676 matroska->level_up--;
1677 break;
1678 }
1679
1680 switch (id) {
1681 /* track number */
1682 case MATROSKA_ID_CUETRACK: {
1683 uint64_t num;
1684 if ((res = ebml_read_uint(matroska,
1685 &id, &num)) < 0)
1686 break;
1687 idx.track = num;
1688 break;
1689 }
1690
1691 /* position in file */
1692 case MATROSKA_ID_CUECLUSTERPOSITION: {
1693 uint64_t num;
1694 if ((res = ebml_read_uint(matroska,
1695 &id, &num)) < 0)
1696 break;
1697 idx.pos = num;
1698 break;
1699 }
1700
1701 default:
1702 av_log(matroska->ctx, AV_LOG_INFO,
1703 "Unknown entry 0x%x in "
1704 "CuesTrackPositions\n", id);
1705 /* fall-through */
1706
1707 case EBML_ID_VOID:
1708 res = ebml_read_skip(matroska);
1709 break;
1710 }
1711
1712 if (matroska->level_up) {
1713 matroska->level_up--;
1714 break;
1715 }
1716 }
1717
1718 break;
1719
1720 default:
1721 av_log(matroska->ctx, AV_LOG_INFO,
1722 "Unknown entry 0x%x in cuespoint "
1723 "index\n", id);
1724 /* fall-through */
1725
1726 case EBML_ID_VOID:
1727 res = ebml_read_skip(matroska);
1728 break;
1729 }
1730
1731 if (matroska->level_up) {
1732 matroska->level_up--;
1733 break;
1734 }
1735 }
1736
1737 /* so let's see if we got what we wanted */
1738 if (idx.pos != (uint64_t) -1 &&
1739 idx.time != (uint64_t) -1 &&
1740 idx.track != (uint16_t) -1) {
1741 if (matroska->num_indexes % 32 == 0) {
1742 /* re-allocate bigger index */
1743 matroska->index =
1744 av_realloc(matroska->index,
1745 (matroska->num_indexes + 32) *
1746 sizeof(MatroskaDemuxIndex));
1747 }
1748 matroska->index[matroska->num_indexes] = idx;
1749 matroska->num_indexes++;
1750 }
1751 break;
1752
1753 default:
1754 av_log(matroska->ctx, AV_LOG_INFO,
1755 "Unknown entry 0x%x in cues header\n", id);
1756 /* fall-through */
1757
1758 case EBML_ID_VOID:
1759 res = ebml_read_skip(matroska);
1760 break;
1761 }
1762
1763 if (matroska->level_up) {
1764 matroska->level_up--;
1765 break;
1766 }
1767 }
1768
1769 return res;
1770 }
1771
1772 static int
1773 matroska_parse_metadata (MatroskaDemuxContext *matroska)
1774 {
1775 int res = 0;
1776 uint32_t id;
1777
1778 while (res == 0) {
1779 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1780 res = AVERROR_IO;
1781 break;
1782 } else if (matroska->level_up) {
1783 matroska->level_up--;
1784 break;
1785 }
1786
1787 switch (id) {
1788 /* Hm, this is unsupported... */
1789 default:
1790 av_log(matroska->ctx, AV_LOG_INFO,
1791 "Unknown entry 0x%x in metadata header\n", id);
1792 /* fall-through */
1793
1794 case EBML_ID_VOID:
1795 res = ebml_read_skip(matroska);
1796 break;
1797 }
1798
1799 if (matroska->level_up) {
1800 matroska->level_up--;
1801 break;
1802 }
1803 }
1804
1805 return res;
1806 }
1807
1808 static int
1809 matroska_parse_seekhead (MatroskaDemuxContext *matroska)
1810 {
1811 int res = 0;
1812 uint32_t id;
1813
1814 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing seekhead...\n");
1815
1816 while (res == 0) {
1817 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1818 res = AVERROR_IO;
1819 break;
1820 } else if (matroska->level_up) {
1821 matroska->level_up--;
1822 break;
1823 }
1824
1825 switch (id) {
1826 case MATROSKA_ID_SEEKENTRY: {
1827 uint32_t seek_id = 0, peek_id_cache = 0;
1828 uint64_t seek_pos = (uint64_t) -1, t;
1829
1830 if ((res = ebml_read_master(matroska, &id)) < 0)
1831 break;
1832
1833 while (res == 0) {
1834 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1835 res = AVERROR_IO;
1836 break;
1837 } else if (matroska->level_up) {
1838 matroska->level_up--;
1839 break;
1840 }
1841
1842 switch (id) {
1843 case MATROSKA_ID_SEEKID:
1844 res = ebml_read_uint(matroska, &id, &t);
1845 seek_id = t;
1846 break;
1847
1848 case MATROSKA_ID_SEEKPOSITION:
1849 res = ebml_read_uint(matroska, &id, &seek_pos);
1850 break;
1851
1852 default:
1853 av_log(matroska->ctx, AV_LOG_INFO,
1854 "Unknown seekhead ID 0x%x\n", id);
1855 /* fall-through */
1856
1857 case EBML_ID_VOID:
1858 res = ebml_read_skip(matroska);
1859 break;
1860 }
1861
1862 if (matroska->level_up) {
1863 matroska->level_up--;
1864 break;
1865 }
1866 }
1867
1868 if (!seek_id || seek_pos == (uint64_t) -1) {
1869 av_log(matroska->ctx, AV_LOG_INFO,
1870 "Incomplete seekhead entry (0x%x/%"PRIu64")\n",
1871 seek_id, seek_pos);
1872 break;
1873 }
1874
1875 switch (seek_id) {
1876 case MATROSKA_ID_CUES:
1877 case MATROSKA_ID_TAGS: {
1878 uint32_t level_up = matroska->level_up;
1879 offset_t before_pos;
1880 uint64_t length;
1881 MatroskaLevel level;
1882
1883 /* remember the peeked ID and the current position */
1884 peek_id_cache = matroska->peek_id;
1885 before_pos = url_ftell(&matroska->ctx->pb);
1886
1887 /* seek */
1888 if ((res = ebml_read_seek(matroska, seek_pos +
1889 matroska->segment_start)) < 0)
1890 return res;
1891
1892 /* we don't want to lose our seekhead level, so we add
1893 * a dummy. This is a crude hack. */
1894 if (matroska->num_levels == EBML_MAX_DEPTH) {
1895 av_log(matroska->ctx, AV_LOG_INFO,
1896 "Max EBML element depth (%d) reached, "
1897 "cannot parse further.\n", EBML_MAX_DEPTH);
1898 return AVERROR_UNKNOWN;
1899 }
1900
1901 level.start = 0;
1902 level.length = (uint64_t)-1;
1903 matroska->levels[matroska->num_levels] = level;
1904 matroska->num_levels++;
1905
1906 /* check ID */
1907 if (!(id = ebml_peek_id (matroska,
1908 &matroska->level_up)))
1909 break;
1910 if (id != seek_id) {
1911 av_log(matroska->ctx, AV_LOG_INFO,
1912 "We looked for ID=0x%x but got "
1913 "ID=0x%x (pos=%"PRIu64")",
1914 seek_id, id, seek_pos +
1915 matroska->segment_start);
1916 goto finish;
1917 }
1918
1919 /* read master + parse */
1920 if ((res = ebml_read_master(matroska, &id)) < 0)
1921 break;
1922 switch (id) {
1923 case MATROSKA_ID_CUES:
1924 if (!(res = matroska_parse_index(matroska)) ||
1925 url_feof(&matroska->ctx->pb)) {
1926 matroska->index_parsed = 1;
1927 res = 0;
1928 }
1929 break;
1930 case MATROSKA_ID_TAGS:
1931 if (!(res = matroska_parse_metadata(matroska)) ||
1932 url_feof(&matroska->ctx->pb)) {
1933 matroska->metadata_parsed = 1;
1934 res = 0;
1935 }
1936 break;
1937 }
1938 if (res < 0)
1939 break;
1940
1941 finish:
1942 /* remove dummy level */
1943 while (matroska->num_levels) {
1944 matroska->num_levels--;
1945 length =
1946 matroska->levels[matroska->num_levels].length;
1947 if (length == (uint64_t)-1)
1948 break;
1949 }
1950
1951 /* seek back */
1952 if ((res = ebml_read_seek(matroska, before_pos)) < 0)
1953 return res;
1954 matroska->peek_id = peek_id_cache;
1955 matroska->level_up = level_up;
1956 break;
1957 }
1958
1959 default:
1960 av_log(matroska->ctx, AV_LOG_INFO,
1961 "Ignoring seekhead entry for ID=0x%x\n",
1962 seek_id);
1963 break;
1964 }
1965
1966 break;
1967 }
1968
1969 default:
1970 av_log(matroska->ctx, AV_LOG_INFO,
1971 "Unknown seekhead ID 0x%x\n", id);
1972 /* fall-through */
1973
1974 case EBML_ID_VOID:
1975 res = ebml_read_skip(matroska);
1976 break;
1977 }
1978
1979 if (matroska->level_up) {
1980 matroska->level_up--;
1981 break;
1982 }
1983 }
1984
1985 return res;
1986 }
1987
1988 static int
1989 matroska_read_header (AVFormatContext *s,
1990 AVFormatParameters *ap)
1991 {
1992 MatroskaDemuxContext *matroska = s->priv_data;
1993 char *doctype;
1994 int version, last_level, res = 0;
1995 uint32_t id;
1996
1997 matroska->ctx = s;
1998
1999 /* First read the EBML header. */
2000 doctype = NULL;
2001 if ((res = ebml_read_header(matroska, &doctype, &version)) < 0)
2002 return res;
2003 if ((doctype == NULL) || strcmp(doctype, "matroska")) {
2004 av_log(matroska->ctx, AV_LOG_ERROR,
2005 "Wrong EBML doctype ('%s' != 'matroska').\n",
2006 doctype ? doctype : "(none)");
2007 if (doctype)
2008 av_free(doctype);
2009 return AVERROR_NOFMT;
2010 }
2011 av_free(doctype);
2012 if (version != 1) {
2013 av_log(matroska->ctx, AV_LOG_ERROR,
2014 "Matroska demuxer version 1 too old for file version %d\n",
2015 version);
2016 return AVERROR_NOFMT;
2017 }
2018
2019 /* The next thing is a segment. */
2020 while (1) {
2021 if (!(id = ebml_peek_id(matroska, &last_level)))
2022 return AVERROR_IO;
2023 if (id == MATROSKA_ID_SEGMENT)
2024 break;
2025
2026 /* oi! */
2027 av_log(matroska->ctx, AV_LOG_INFO,
2028 "Expected a Segment ID (0x%x), but received 0x%x!\n",
2029 MATROSKA_ID_SEGMENT, id);
2030 if ((res = ebml_read_skip(matroska)) < 0)
2031 return res;
2032 }
2033
2034 /* We now have a Matroska segment.
2035 * Seeks are from the beginning of the segment,
2036 * after the segment ID/length. */
2037 if ((res = ebml_read_master(matroska, &id)) < 0)
2038 return res;
2039 matroska->segment_start = url_ftell(&s->pb);
2040
2041 matroska->time_scale = 1000000;
2042 /* we've found our segment, start reading the different contents in here */
2043 while (res == 0) {
2044 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2045 res = AVERROR_IO;
2046 break;
2047 } else if (matroska->level_up) {
2048 matroska->level_up--;
2049 break;
2050 }
2051
2052 switch (id) {
2053 /* stream info */
2054 case MATROSKA_ID_INFO: {
2055 if ((res = ebml_read_master(matroska, &id)) < 0)
2056 break;
2057 res = matroska_parse_info(matroska);
2058 break;
2059 }
2060
2061 /* track info headers */
2062 case MATROSKA_ID_TRACKS: {
2063 if ((res = ebml_read_master(matroska, &id)) < 0)
2064 break;
2065 res = matroska_parse_tracks(matroska);
2066 break;
2067 }
2068
2069 /* stream index */
2070 case MATROSKA_ID_CUES: {
2071 if (!matroska->index_parsed) {
2072 if ((res = ebml_read_master(matroska, &id)) < 0)
2073 break;
2074 res = matroska_parse_index(matroska);
2075 } else
2076 res = ebml_read_skip(matroska);
2077 break;
2078 }
2079
2080 /* metadata */
2081 case MATROSKA_ID_TAGS: {
2082 if (!matroska->metadata_parsed) {
2083 if ((res = ebml_read_master(matroska, &id)) < 0)
2084 break;
2085 res = matroska_parse_metadata(matroska);
2086 } else
2087 res = ebml_read_skip(matroska);
2088 break;
2089 }
2090
2091 /* file index (if seekable, seek to Cues/Tags to parse it) */
2092 case MATROSKA_ID_SEEKHEAD: {
2093 if ((res = ebml_read_master(matroska, &id)) < 0)
2094 break;
2095 res = matroska_parse_seekhead(matroska);
2096 break;
2097 }
2098
2099 case MATROSKA_ID_CLUSTER: {
2100 /* Do not read the master - this will be done in the next
2101 * call to matroska_read_packet. */
2102 res = 1;
2103 break;
2104 }
2105
2106 default:
2107 av_log(matroska->ctx, AV_LOG_INFO,
2108 "Unknown matroska file header ID 0x%x\n", id);
2109 /* fall-through */
2110
2111 case EBML_ID_VOID:
2112 res = ebml_read_skip(matroska);
2113 break;
2114 }
2115
2116 if (matroska->level_up) {
2117 matroska->level_up--;
2118 break;
2119 }
2120 }
2121
2122 if (res < 0)
2123 return res;
2124
2125 /* Have we found a cluster? */
2126 if (res == 1) {
2127 int i;
2128 enum CodecID codec_id;
2129 MatroskaTrack *track;
2130 AVStream *st;
2131 void *extradata = NULL;
2132 int extradata_size = 0;
2133
2134 for (i = 0; i < matroska->num_tracks; i++) {
2135 track = matroska->tracks[i];
2136
2137 /* libavformat does not really support subtitles.
2138 * Also apply some sanity checks. */
2139 if ((track->type == MATROSKA_TRACK_TYPE_SUBTITLE) ||
2140 (track->codec_id == NULL))
2141 continue;
2142
2143 /* Set the FourCC from the CodecID. */
2144 /* This is the MS compatibility mode which stores a
2145 * BITMAPINFOHEADER in the CodecPrivate. */
2146 if (!strcmp(track->codec_id,
2147 MATROSKA_CODEC_ID_VIDEO_VFW_FOURCC) &&
2148 (track->codec_priv_size >= 40) &&
2149 (track->codec_priv != NULL)) {
2150 unsigned char *p;
2151
2152 /* Offset of biCompression. Stored in LE. */
2153 p = (unsigned char *)track->codec_priv + 16;
2154 ((MatroskaVideoTrack *)track)->fourcc = (p[3] << 24) |
2155 (p[2] << 16) | (p[1] << 8) | p[0];
2156 codec_id = codec_get_bmp_id(((MatroskaVideoTrack *)track)->fourcc);
2157
2158 } else if (!strcmp(track->codec_id,
2159 MATROSKA_CODEC_ID_VIDEO_MPEG4_SP) ||
2160 !strcmp(track->codec_id,
2161 MATROSKA_CODEC_ID_VIDEO_MPEG4_ASP) ||
2162 !strcmp(track->codec_id,
2163 MATROSKA_CODEC_ID_VIDEO_MPEG4_AP))
2164 codec_id = CODEC_ID_MPEG4;
2165 else if (!strcmp(track->codec_id,
2166 MATROSKA_CODEC_ID_VIDEO_MPEG4_AVC))
2167 codec_id = CODEC_ID_H264;
2168 /* else if (!strcmp(track->codec_id, */
2169 /* MATROSKA_CODEC_ID_VIDEO_UNCOMPRESSED)) */
2170 /* codec_id = CODEC_ID_???; */
2171 else if (!strcmp(track->codec_id,
2172 MATROSKA_CODEC_ID_VIDEO_MSMPEG4V3))
2173 codec_id = CODEC_ID_MSMPEG4V3;
2174 else if (!strcmp(track->codec_id,
2175 MATROSKA_CODEC_ID_VIDEO_MPEG1) ||
2176 !strcmp(track->codec_id,
2177 MATROSKA_CODEC_ID_VIDEO_MPEG2))
2178 codec_id = CODEC_ID_MPEG2VIDEO;
2179
2180 /* This is the MS compatibility mode which stores a
2181 * WAVEFORMATEX in the CodecPrivate. */
2182 else if (!strcmp(track->codec_id,
2183 MATROSKA_CODEC_ID_AUDIO_ACM) &&
2184 (track->codec_priv_size >= 18) &&
2185 (track->codec_priv != NULL)) {
2186 unsigned char *p;
2187 uint16_t tag;
2188
2189 /* Offset of wFormatTag. Stored in LE. */
2190 p = (unsigned char *)track->codec_priv;
2191 tag = (p[1] << 8) | p[0];
2192 codec_id = codec_get_wav_id(tag);
2193
2194 } else if (!strcmp(track->codec_id,
2195 MATROSKA_CODEC_ID_AUDIO_MPEG1_L1) ||
2196 !strcmp(track->codec_id,
2197 MATROSKA_CODEC_ID_AUDIO_MPEG1_L2) ||
2198 !strcmp(track->codec_id,
2199 MATROSKA_CODEC_ID_AUDIO_MPEG1_L3))
2200 codec_id = CODEC_ID_MP3;
2201 else if (!strcmp(track->codec_id,
2202 MATROSKA_CODEC_ID_AUDIO_PCM_INT_BE))
2203 codec_id = CODEC_ID_PCM_U16BE;
2204 else if (!strcmp(track->codec_id,
2205 MATROSKA_CODEC_ID_AUDIO_PCM_INT_LE))
2206 codec_id = CODEC_ID_PCM_U16LE;
2207 /* else if (!strcmp(track->codec_id, */
2208 /* MATROSKA_CODEC_ID_AUDIO_PCM_FLOAT)) */
2209 /* codec_id = CODEC_ID_PCM_???; */
2210 else if (!strcmp(track->codec_id,
2211 MATROSKA_CODEC_ID_AUDIO_AC3))
2212 codec_id = CODEC_ID_AC3;
2213 else if (!strcmp(track->codec_id,
2214 MATROSKA_CODEC_ID_AUDIO_DTS))
2215 codec_id = CODEC_ID_DTS;
2216 /* No such codec id so far. */
2217 /* else if (!strcmp(track->codec_id, */
2218 /* MATROSKA_CODEC_ID_AUDIO_DTS)) */
2219 /* codec_id = CODEC_ID_DTS; */
2220 else if (!strcmp(track->codec_id,
2221 MATROSKA_CODEC_ID_AUDIO_VORBIS)) {
2222 extradata_size = track->codec_priv_size;
2223 if(extradata_size) {
2224 extradata = av_malloc(extradata_size);
2225 if(extradata == NULL)
2226 return AVERROR_NOMEM;
2227 memcpy(extradata, track->codec_priv, extradata_size);
2228 }
2229 codec_id = CODEC_ID_VORBIS;
2230 } else if (!strcmp(track->codec_id,
2231 MATROSKA_CODEC_ID_AUDIO_MPEG2) ||
2232 !strcmp(track->codec_id,
2233 MATROSKA_CODEC_ID_AUDIO_MPEG4))
2234 codec_id = CODEC_ID_AAC;
2235 else
2236 codec_id = CODEC_ID_NONE;
2237
2238 if (codec_id == CODEC_ID_NONE) {
2239 av_log(matroska->ctx, AV_LOG_INFO,
2240 "Unknown/unsupported CodecID %s.\n",
2241 track->codec_id);
2242 }
2243
2244 track->stream_index = matroska->num_streams;
2245
2246 matroska->num_streams++;
2247 st = av_new_stream(s, track->stream_index);
2248 if (st == NULL)
2249 return AVERROR_NOMEM;
2250 av_set_pts_info(st, 24, 1, 1000); /* 24 bit pts in ms */
2251
2252 st->codec->codec_id = codec_id;
2253
2254 if(extradata){
2255 st->codec->extradata = extradata;
2256 st->codec->extradata_size = extradata_size;
2257 } else if(track->codec_priv && track->codec_priv_size > 0){
2258 st->codec->extradata = av_malloc(track->codec_priv_size);
2259 if(st->codec->extradata == NULL)
2260 return AVERROR_NOMEM;
2261 st->codec->extradata_size = track->codec_priv_size;
2262 memcpy(st->codec->extradata, track->codec_priv,
2263 track->codec_priv_size);
2264 }
2265
2266 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2267 MatroskaVideoTrack *videotrack = (MatroskaVideoTrack *)track;
2268
2269 st->codec->codec_type = CODEC_TYPE_VIDEO;
2270 st->codec->codec_tag = videotrack->fourcc;
2271 st->codec->width = videotrack->pixel_width;
2272 st->codec->height = videotrack->pixel_height;
2273 if (videotrack->display_width == 0)
2274 st->codec->sample_aspect_ratio.num =
2275 videotrack->pixel_width;
2276 else
2277 st->codec->sample_aspect_ratio.num =
2278 videotrack->display_width;
2279 if (videotrack->display_height == 0)
2280 st->codec->sample_aspect_ratio.num =
2281 videotrack->pixel_height;
2282 else
2283 st->codec->sample_aspect_ratio.num =
2284 videotrack->display_height;
2285
2286 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2287 MatroskaAudioTrack *audiotrack = (MatroskaAudioTrack *)track;
2288
2289 st->codec->codec_type = CODEC_TYPE_AUDIO;
2290 st->codec->sample_rate = audiotrack->samplerate;
2291 st->codec->channels = audiotrack->channels;
2292 }
2293
2294 /* What do we do with private data? E.g. for Vorbis. */
2295 }
2296 }
2297
2298 return 0;
2299 }
2300
2301 static int
2302 matroska_find_track_by_num (MatroskaDemuxContext *matroska,
2303 int num)
2304 {
2305 int i;
2306
2307 for (i = 0; i < matroska->num_tracks; i++)
2308 if (matroska->tracks[i]->num == num)
2309 return i;
2310
2311 return -1;
2312 }
2313
2314 static int
2315 matroska_parse_blockgroup (MatroskaDemuxContext *matroska,
2316 uint64_t cluster_time)
2317 {
2318 int res = 0;
2319 uint32_t id;
2320 AVPacket *pkt;
2321 int is_keyframe = PKT_FLAG_KEY, last_num_packets = matroska->num_packets;
2322
2323 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing blockgroup...\n");
2324
2325 while (res == 0) {
2326 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2327 res = AVERROR_IO;
2328 break;
2329 } else if (matroska->level_up) {
2330 matroska->level_up--;
2331 break;
2332 }
2333
2334 switch (id) {
2335 /* one block inside the group. Note, block parsing is one
2336 * of the harder things, so this code is a bit complicated.
2337 * See http://www.matroska.org/ for documentation. */
2338 case MATROSKA_ID_BLOCK: {
2339 uint8_t *data, *origdata;
2340 int size;
2341 uint64_t time;
2342 uint32_t *lace_size = NULL;
2343 int n, track, flags, laces = 0;
2344 uint64_t num;
2345 int64_t pos= url_ftell(&matroska->ctx->pb);
2346
2347 if ((res = ebml_read_binary(matroska, &id, &data, &size)) < 0)
2348 break;
2349 origdata = data;
2350
2351 /* first byte(s): blocknum */
2352 if ((n = matroska_ebmlnum_uint(data, size, &num)) < 0) {
2353 av_log(matroska->ctx, AV_LOG_ERROR,
2354 "EBML block data error\n");
2355 av_free(origdata);
2356 break;
2357 }
2358 data += n;
2359 size -= n;
2360
2361 /* fetch track from num */
2362 track = matroska_find_track_by_num(matroska, num);
2363 if (size <= 3 || track < 0 || track >= matroska->num_tracks) {
2364 av_log(matroska->ctx, AV_LOG_INFO,
2365 "Invalid stream %d or size %u\n", track, size);
2366 av_free(origdata);
2367 break;
2368 }
2369 if(matroska->ctx->streams[ matroska->tracks[track]->stream_index ]->discard >= AVDISCARD_ALL){
2370 av_free(origdata);
2371 break;
2372 }
2373
2374 /* time (relative to cluster time) */
2375 time = ((data[0] << 8) | data[1]) * matroska->time_scale;
2376 data += 2;
2377 size -= 2;
2378 flags = *data;
2379 data += 1;
2380 size -= 1;
2381 switch ((flags & 0x06) >> 1) {
2382 case 0x0: /* no lacing */
2383 laces = 1;
2384 lace_size = av_mallocz(sizeof(int));
2385 lace_size[0] = size;
2386 break;
2387
2388 case 0x1: /* xiph lacing */
2389 case 0x2: /* fixed-size lacing */
2390 case 0x3: /* EBML lacing */
2391 if (size == 0) {
2392 res = -1;
2393 break;
2394 }
2395 laces = (*data) + 1;
2396 data += 1;
2397 size -= 1;
2398 lace_size = av_mallocz(laces * sizeof(int));
2399
2400 switch ((flags & 0x06) >> 1) {
2401 case 0x1: /* xiph lacing */ {
2402 uint8_t temp;
2403 uint32_t total = 0;
2404 for (n = 0; res == 0 && n < laces - 1; n++) {
2405 while (1) {
2406 if (size == 0) {
2407 res = -1;
2408 break;
2409 }
2410 temp = *data;
2411 lace_size[n] += temp;
2412 data += 1;
2413 size -= 1;
2414 if (temp != 0xff)
2415 break;
2416 }
2417 total += lace_size[n];
2418 }
2419 lace_size[n] = size - total;
2420 break;
2421 }
2422
2423 case 0x2: /* fixed-size lacing */
2424 for (n = 0; n < laces; n++)
2425 lace_size[n] = size / laces;
2426 break;
2427
2428 case 0x3: /* EBML lacing */ {
2429 uint32_t total;
2430 n = matroska_ebmlnum_uint(data, size, &num);
2431 if (n < 0) {
2432 av_log(matroska->ctx, AV_LOG_INFO,
2433 "EBML block data error\n");
2434 break;
2435 }
2436 data += n;
2437 size -= n;
2438 total = lace_size[0] = num;
2439 for (n = 1; res == 0 && n < laces - 1; n++) {
2440 int64_t snum;
2441 int r;
2442 r = matroska_ebmlnum_sint (data, size,
2443 &snum);
2444 if (r < 0) {
2445 av_log(matroska->ctx, AV_LOG_INFO,
2446 "EBML block data error\n");
2447 break;
2448 }
2449 data += r;
2450 size -= r;
2451 lace_size[n] = lace_size[n - 1] + snum;
2452 total += lace_size[n];
2453 }
2454 lace_size[n] = size - total;
2455 break;
2456 }
2457 }
2458 break;
2459 }
2460
2461 if (res == 0) {
2462 for (n = 0; n < laces; n++) {
2463 uint64_t timecode = 0;
2464
2465 pkt = av_mallocz(sizeof(AVPacket));
2466 /* XXX: prevent data copy... */
2467 if (av_new_packet(pkt,lace_size[n]) < 0) {
2468 res = AVERROR_NOMEM;
2469 break;
2470 }
2471 if (cluster_time != (uint64_t)-1) {
2472 if (time < 0 && (-time) > cluster_time)
2473 timecode = cluster_time;
2474 else
2475 timecode = cluster_time + time;
2476 }
2477 /* FIXME: duration */
2478
2479 memcpy(pkt->data, data, lace_size[n]);
2480 data += lace_size[n];
2481 if (n == 0)
2482 pkt->flags = is_keyframe;
2483 pkt->stream_index =
2484 matroska->tracks[track]->stream_index;
2485
2486 pkt->pts = timecode / 1000000; /* ns to ms */
2487 pkt->pos= pos;
2488
2489 matroska_queue_packet(matroska, pkt);
2490 }
2491 }
2492
2493 av_free(lace_size);
2494 av_free(origdata);
2495 break;
2496 }
2497
2498 case MATROSKA_ID_BLOCKDURATION: {
2499 uint64_t num;
2500 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
2501 break;
2502 av_log(matroska->ctx, AV_LOG_INFO,
2503 "FIXME: implement support for BlockDuration\n");
2504 break;
2505 }
2506
2507 case MATROSKA_ID_BLOCKREFERENCE:
2508 /* We've found a reference, so not even the first frame in
2509 * the lace is a key frame. */
2510 is_keyframe = 0;
2511 if (last_num_packets != matroska->num_packets)
2512 matroska->packets[last_num_packets]->flags = 0;
2513 res = ebml_read_skip(matroska);
2514 break;
2515
2516 default:
2517 av_log(matroska->ctx, AV_LOG_INFO,
2518 "Unknown entry 0x%x in blockgroup data\n", id);
2519 /* fall-through */
2520
2521 case EBML_ID_VOID:
2522 res = ebml_read_skip(matroska);
2523 break;
2524 }
2525
2526 if (matroska->level_up) {
2527 matroska->level_up--;
2528 break;
2529 }
2530 }
2531
2532 return res;
2533 }
2534
2535 static int
2536 matroska_parse_cluster (MatroskaDemuxContext *matroska)
2537 {
2538 int res = 0;
2539 uint32_t id;
2540 uint64_t cluster_time = 0;
2541
2542 av_log(matroska->ctx, AV_LOG_DEBUG,
2543 "parsing cluster at %"PRId64"\n", url_ftell(&matroska->ctx->pb));
2544
2545 while (res == 0) {
2546 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2547 res = AVERROR_IO;
2548 break;
2549 } else if (matroska->level_up) {
2550 matroska->level_up--;
2551 break;
2552 }
2553
2554 switch (id) {
2555 /* cluster timecode */
2556 case MATROSKA_ID_CLUSTERTIMECODE: {
2557 uint64_t num;
2558 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
2559 break;
2560 cluster_time = num * matroska->time_scale;
2561 break;
2562 }
2563
2564 /* a group of blocks inside a cluster */
2565 case MATROSKA_ID_BLOCKGROUP:
2566 if ((res = ebml_read_master(matroska, &id)) < 0)
2567 break;
2568 res = matroska_parse_blockgroup(matroska, cluster_time);
2569 break;
2570
2571 default:
2572 av_log(matroska->ctx, AV_LOG_INFO,
2573 "Unknown entry 0x%x in cluster data\n", id);
2574 /* fall-through */
2575
2576 case EBML_ID_VOID:
2577 res = ebml_read_skip(matroska);
2578 break;
2579 }
2580
2581 if (matroska->level_up) {
2582 matroska->level_up--;
2583 break;
2584 }
2585 }
2586
2587 return res;
2588 }
2589
2590 static int
2591 matroska_read_packet (AVFormatContext *s,
2592 AVPacket *pkt)
2593 {
2594 MatroskaDemuxContext *matroska = s->priv_data;
2595 int res = 0;
2596 uint32_t id;
2597
2598 /* Do we still have a packet queued? */
2599 if (matroska_deliver_packet(matroska, pkt) == 0)
2600 return 0;
2601
2602 /* Have we already reached the end? */
2603 if (matroska->done)
2604 return AVERROR_IO;
2605
2606 while (res == 0) {
2607 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2608 res = AVERROR_IO;
2609 break;
2610 } else if (matroska->level_up) {
2611 matroska->level_up--;
2612 break;
2613 }
2614
2615 switch (id) {
2616 case MATROSKA_ID_CLUSTER:
2617 if ((res = ebml_read_master(matroska, &id)) < 0)
2618 break;
2619 if ((res = matroska_parse_cluster(matroska)) == 0)
2620 res = 1; /* Parsed one cluster, let's get out. */
2621 break;
2622
2623 default:
2624 case EBML_ID_VOID:
2625 res = ebml_read_skip(matroska);
2626 break;
2627 }
2628
2629 if (matroska->level_up) {
2630 matroska->level_up--;
2631 break;
2632 }
2633 }
2634
2635 if (res == -1)
2636 matroska->done = 1;
2637
2638 return matroska_deliver_packet(matroska, pkt);
2639 }
2640
2641 static int
2642 matroska_read_close (AVFormatContext *s)
2643 {
2644 MatroskaDemuxContext *matroska = s->priv_data;
2645 int n = 0;
2646
2647 if (matroska->writing_app)
2648 av_free(matroska->writing_app);
2649 if (matroska->muxing_app)
2650 av_free(matroska->muxing_app);
2651 if (matroska->index)
2652 av_free(matroska->index);
2653
2654 if (matroska->packets != NULL) {
2655 for (n = 0; n < matroska->num_packets; n++) {
2656 av_free_packet(matroska->packets[n]);
2657 av_free(matroska->packets[n]);
2658 }
2659 av_free(matroska->packets);
2660 }
2661
2662 for (n = 0; n < matroska->num_tracks; n++) {
2663 MatroskaTrack *track = matroska->tracks[n];
2664 if (track->codec_id)
2665 av_free(track->codec_id);
2666 if (track->codec_name)
2667 av_free(track->codec_name);
2668 if (track->codec_priv)
2669 av_free(track->codec_priv);
2670 if (track->name)
2671 av_free(track->name);
2672 if (track->language)
2673 av_free(track->language);
2674
2675 av_free(track);
2676 }
2677
2678 for (n = 0; n < s->nb_streams; n++) {
2679 av_free(s->streams[n]->codec->extradata);
2680 }
2681
2682 memset(matroska, 0, sizeof(MatroskaDemuxContext));
2683
2684 return 0;
2685 }
2686
2687 static AVInputFormat matroska_iformat = {
2688 "matroska",
2689 "Matroska file format",
2690 sizeof(MatroskaDemuxContext),
2691 matroska_probe,
2692 matroska_read_header,
2693 matroska_read_packet,
2694 matroska_read_close,
2695 };
2696
2697 int
2698 matroska_init(void)
2699 {
2700 av_register_input_format(&matroska_iformat);
2701 return 0;
2702 }