libdts support by (Benjamin Zores <ben at geexbox dot org>)
[libav.git] / libavformat / matroska.c
CommitLineData
08abe0fd
MN
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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 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
142typedef 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
151typedef 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
158typedef 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
169typedef 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
176typedef 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_MSMPEG4V3 "V_MPEG4/MS/V3"
190#define MATROSKA_CODEC_ID_VIDEO_MPEG1 "V_MPEG1"
191#define MATROSKA_CODEC_ID_VIDEO_MPEG2 "V_MPEG2"
192#define MATROSKA_CODEC_ID_VIDEO_MJPEG "V_MJPEG"
193/* TODO: Real/Quicktime */
194
195#define MATROSKA_CODEC_ID_AUDIO_ACM "A_MS/ACM"
196#define MATROSKA_CODEC_ID_AUDIO_MPEG1_L1 "A_MPEG/L1"
197#define MATROSKA_CODEC_ID_AUDIO_MPEG1_L2 "A_MPEG/L2"
198#define MATROSKA_CODEC_ID_AUDIO_MPEG1_L3 "A_MPEG/L3"
199#define MATROSKA_CODEC_ID_AUDIO_PCM_INT_BE "A_PCM/INT/BIG"
200#define MATROSKA_CODEC_ID_AUDIO_PCM_INT_LE "A_PCM/INT/LIT"
201#define MATROSKA_CODEC_ID_AUDIO_PCM_FLOAT "A_PCM/FLOAT/IEEE"
202#define MATROSKA_CODEC_ID_AUDIO_AC3 "A_AC3"
203#define MATROSKA_CODEC_ID_AUDIO_DTS "A_DTS"
204#define MATROSKA_CODEC_ID_AUDIO_VORBIS "A_VORBIS"
205#define MATROSKA_CODEC_ID_AUDIO_ACM "A_MS/ACM"
206#define MATROSKA_CODEC_ID_AUDIO_MPEG2 "A_AAC/MPEG2/"
207#define MATROSKA_CODEC_ID_AUDIO_MPEG4 "A_AAC/MPEG4/"
208/* TODO: AC3-9/10 (?), Real, Musepack, Quicktime */
209
210/* max. depth in the EBML tree structure */
211#define EBML_MAX_DEPTH 16
212
213typedef struct Track {
214 MatroskaTrackType type;
215
216 /* Unique track number and track ID. stream_index is the index that
217 * the calling app uses for this track. */
218 uint32_t num,
219 uid,
220 stream_index;
221
222 char *name,
223 *language;
224
225 char *codec_id,
226 *codec_name;
227
228 unsigned char *codec_priv;
229 int codec_priv_size;
230
231 int64_t default_duration;
232 MatroskaTrackFlags flags;
233} MatroskaTrack;
234
235typedef struct MatroskaVideoTrack {
236 MatroskaTrack track;
237
238 int pixel_width,
239 pixel_height,
240 display_width,
241 display_height;
242
243 uint32_t fourcc;
244
245 MatroskaAspectRatioMode ar_mode;
246 MatroskaEyeMode eye_mode;
247
248 //..
249} MatroskaVideoTrack;
250
251typedef struct MatroskaAudioTrack {
252 MatroskaTrack track;
253
254 int channels,
255 bitdepth,
256 samplerate;
257 //..
258} MatroskaAudioTrack;
259
260typedef struct MatroskaSubtitleTrack {
261 MatroskaTrack track;
262
263 //..
264} MatroskaSubtitleTrack;
265
266typedef struct MatroskaLevel {
267 uint64_t start, length;
268} MatroskaLevel;
269
270typedef struct MatroskaDemuxIndex {
271 uint64_t pos; /* of the corresponding *cluster*! */
272 uint16_t track; /* reference to 'num' */
273 uint64_t time; /* in nanoseconds */
274} MatroskaDemuxIndex;
275
276typedef struct MatroskaDemuxContext {
277 AVFormatContext *ctx;
278
279 /* ebml stuff */
280 int num_levels;
281 MatroskaLevel levels[EBML_MAX_DEPTH];
282 int level_up;
283
284 /* matroska stuff */
285 char *writing_app,
286 *muxing_app;
287 int64_t created;
288
289 /* timescale in the file */
290 int64_t time_scale;
291
292 /* length, position (time, ns) */
293 int64_t duration,
294 pos;
295
296 /* num_streams is the number of streams that av_new_stream() was called
297 * for ( = that are available to the calling program). */
298 int num_tracks, num_streams;
299 MatroskaTrack *tracks[MAX_STREAMS];
300
301 /* cache for ID peeking */
302 uint32_t peek_id;
303
304 /* byte position of the segment inside the stream */
305 offset_t segment_start;
306
307 /* The packet queue. */
308 AVPacket **packets;
309 int num_packets;
310
311 /* have we already parse metadata/cues/clusters? */
312 int metadata_parsed,
313 index_parsed,
314 done;
315
316 /* The index for seeking. */
317 int num_indexes;
318 MatroskaDemuxIndex *index;
319} MatroskaDemuxContext;
320
321/*
322 * The first few functions handle EBML file parsing. The rest
323 * is the document interpretation. Matroska really just is a
324 * EBML file.
325 */
326
327/*
328 * Return: the amount of levels in the hierarchy that the
329 * current element lies higher than the previous one.
330 * The opposite isn't done - that's auto-done using master
331 * element reading.
332 */
333
334static int
335ebml_read_element_level_up (MatroskaDemuxContext *matroska)
336{
337 ByteIOContext *pb = &matroska->ctx->pb;
338 offset_t pos = url_ftell(pb);
339 int num = 0;
340
341 while (matroska->num_levels > 0) {
342 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
343
344 if (pos >= level->start + level->length) {
345 matroska->num_levels--;
346 num++;
347 } else {
348 break;
349 }
350 }
351
352 return num;
353}
354
355/*
356 * Read: an "EBML number", which is defined as a variable-length
357 * array of bytes. The first byte indicates the length by giving a
358 * number of 0-bits followed by a one. The position of the first
359 * "one" bit inside the first byte indicates the length of this
360 * number.
361 * Returns: num. of bytes read. < 0 on error.
362 */
363
364static int
365ebml_read_num (MatroskaDemuxContext *matroska,
366 int max_size,
367 uint64_t *number)
368{
369 ByteIOContext *pb = &matroska->ctx->pb;
370 int len_mask = 0x80, read = 1, n = 1;
371 int64_t total = 0;
372
373 /* the first byte tells us the length in bytes - get_byte() can normally
374 * return 0, but since that's not a valid first ebmlID byte, we can
375 * use it safely here to catch EOS. */
376 if (!(total = get_byte(pb))) {
377 /* we might encounter EOS here */
378 if (!url_feof(pb)) {
379 offset_t pos = url_ftell(pb);
380 av_log(matroska->ctx, AV_LOG_ERROR,
381 "Read error at pos. %llu (0x%llx)\n",
382 pos, pos);
383 }
384 return AVERROR_IO; /* EOS or actual I/O error */
385 }
386
387 /* get the length of the EBML number */
388 while (read <= max_size && !(total & len_mask)) {
389 read++;
390 len_mask >>= 1;
391 }
392 if (read > max_size) {
393 offset_t pos = url_ftell(pb) - 1;
394 av_log(matroska->ctx, AV_LOG_ERROR,
395 "Invalid EBML number size tag 0x%02x at pos %llu (0x%llx)\n",
396 (uint8_t) total, pos, pos);
397 return AVERROR_INVALIDDATA;
398 }
399
400 /* read out length */
401 total &= ~len_mask;
402 while (n++ < read)
403 total = (total << 8) | get_byte(pb);
404
405 *number = total;
406
407 return read;
408}
409
410/*
411 * Read: the element content data ID.
412 * Return: the number of bytes read or < 0 on error.
413 */
414
415static int
416ebml_read_element_id (MatroskaDemuxContext *matroska,
417 uint32_t *id,
418 int *level_up)
419{
420 int read;
421 uint64_t total;
422
423 /* if we re-call this, use our cached ID */
424 if (matroska->peek_id != 0) {
425 if (level_up)
426 *level_up = 0;
427 *id = matroska->peek_id;
428 return 0;
429 }
430
431 /* read out the "EBML number", include tag in ID */
432 if ((read = ebml_read_num(matroska, 4, &total)) < 0)
433 return read;
434 *id = matroska->peek_id = total | (1 << (read * 7));
435
436 /* level tracking */
437 if (level_up)
438 *level_up = ebml_read_element_level_up(matroska);
439
440 return read;
441}
442
443/*
444 * Read: element content length.
445 * Return: the number of bytes read or < 0 on error.
446 */
447
448static int
449ebml_read_element_length (MatroskaDemuxContext *matroska,
450 uint64_t *length)
451{
452 /* clear cache since we're now beyond that data point */
453 matroska->peek_id = 0;
454
455 /* read out the "EBML number", include tag in ID */
456 return ebml_read_num(matroska, 8, length);
457}
458
459/*
460 * Return: the ID of the next element, or 0 on error.
461 * Level_up contains the amount of levels that this
462 * next element lies higher than the previous one.
463 */
464
465static uint32_t
466ebml_peek_id (MatroskaDemuxContext *matroska,
467 int *level_up)
468{
469 uint32_t id;
470
471 assert(level_up != NULL);
472
473 if (ebml_read_element_id(matroska, &id, level_up) < 0)
474 return 0;
475
476 return id;
477}
478
479/*
480 * Seek to a given offset.
481 * 0 is success, -1 is failure.
482 */
483
484static int
485ebml_read_seek (MatroskaDemuxContext *matroska,
486 offset_t offset)
487{
488 ByteIOContext *pb = &matroska->ctx->pb;
489
490 /* clear ID cache, if any */
491 matroska->peek_id = 0;
492
493 return (url_fseek(pb, offset, SEEK_SET) == offset) ? 0 : -1;
494}
495
496/*
497 * Skip the next element.
498 * 0 is success, -1 is failure.
499 */
500
501static int
502ebml_read_skip (MatroskaDemuxContext *matroska)
503{
504 ByteIOContext *pb = &matroska->ctx->pb;
505 uint32_t id;
506 uint64_t length;
507 int res;
508
509 if ((res = ebml_read_element_id(matroska, &id, NULL)) < 0 ||
510 (res = ebml_read_element_length(matroska, &length)) < 0)
511 return res;
512
513 url_fskip(pb, length);
514
515 return 0;
516}
517
518/*
519 * Read the next element as an unsigned int.
520 * 0 is success, < 0 is failure.
521 */
522
523static int
524ebml_read_uint (MatroskaDemuxContext *matroska,
525 uint32_t *id,
526 uint64_t *num)
527{
528 ByteIOContext *pb = &matroska->ctx->pb;
529 int n = 0, size, res;
530 uint64_t rlength;
531
532 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
533 (res = ebml_read_element_length(matroska, &rlength)) < 0)
534 return res;
535 size = rlength;
536 if (size < 1 || size > 8) {
537 offset_t pos = url_ftell(pb);
538 av_log(matroska->ctx, AV_LOG_ERROR,
539 "Invalid uint element size %d at position %lld (0x%llx)\n",
540 size, pos, pos);
541 return AVERROR_INVALIDDATA;
542 }
543
544 /* big-endian ordening; build up number */
545 *num = 0;
546 while (n++ < size)
547 *num = (*num << 8) | get_byte(pb);
548
549 return 0;
550}
551
552/*
553 * Read the next element as a signed int.
554 * 0 is success, < 0 is failure.
555 */
556
557static int
558ebml_read_sint (MatroskaDemuxContext *matroska,
559 uint32_t *id,
560 int64_t *num)
561{
562 ByteIOContext *pb = &matroska->ctx->pb;
563 int size, n = 1, negative = 0, res;
564 uint64_t rlength;
565
566 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
567 (res = ebml_read_element_length(matroska, &rlength)) < 0)
568 return res;
569 size = rlength;
570 if (size < 1 || size > 8) {
571 offset_t pos = url_ftell(pb);
572 av_log(matroska->ctx, AV_LOG_ERROR,
573 "Invalid sint element size %d at position %lld (0x%llx)\n",
574 size, pos, pos);
575 return AVERROR_INVALIDDATA;
576 }
577 if ((*num = get_byte(pb)) & 0x80) {
578 negative = 1;
579 *num &= ~0x80;
580 }
581 *num = 0;
582 while (n++ < size)
583 *num = (*num << 8) | get_byte(pb);
584
585 /* make signed */
586 if (negative)
587 *num = *num - (1LL << ((8 * size) - 1));
588
589 return 0;
590}
591
592/*
593 * Read the next element as a float.
594 * 0 is success, < 0 is failure.
595 */
596
597static int
598ebml_read_float (MatroskaDemuxContext *matroska,
599 uint32_t *id,
600 double *num)
601{
602 ByteIOContext *pb = &matroska->ctx->pb;
603 int size, res;
604 uint64_t rlength;
605
606 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
607 (res = ebml_read_element_length(matroska, &rlength)) < 0)
608 return res;
609 size = rlength;
610
611 if (size != 4 && size != 8 && size != 10) {
612 offset_t pos = url_ftell(pb);
613 av_log(matroska->ctx, AV_LOG_ERROR,
614 "Invalid float element size %d at position %llu (0x%llx)\n",
615 size, pos, pos);
616 return AVERROR_INVALIDDATA;
617 }
618 if (size == 10) {
619 av_log(matroska->ctx, AV_LOG_ERROR,
620 "FIXME! 10-byte floats unimplemented\n");
621 return AVERROR_UNKNOWN;
622 }
623
624 if (size == 4) {
625 float f;
626
6e046750 627 while (size-- > 0)
08abe0fd 628#ifdef WORDS_BIGENDIAN
6e046750 629 ((uint8_t *) &f)[3 - size] = get_byte(pb);
08abe0fd 630#else
08abe0fd
MN
631 ((uint8_t *) &f)[size] = get_byte(pb);
632#endif
633
634 *num = f;
635 } else {
636 double d;
637
6e046750 638 while (size-- > 0)
08abe0fd 639#ifdef WORDS_BIGENDIAN
6e046750 640 ((uint8_t *) &d)[7 - size] = get_byte(pb);
08abe0fd 641#else
08abe0fd
MN
642 ((uint8_t *) &d)[size] = get_byte(pb);
643#endif
644
645 *num = d;
646 }
647
648 return 0;
649}
650
651/*
652 * Read the next element as an ASCII string.
653 * 0 is success, < 0 is failure.
654 */
655
656static int
657ebml_read_ascii (MatroskaDemuxContext *matroska,
658 uint32_t *id,
659 char **str)
660{
661 ByteIOContext *pb = &matroska->ctx->pb;
662 int size, res;
663 uint64_t rlength;
664
665 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
666 (res = ebml_read_element_length(matroska, &rlength)) < 0)
667 return res;
668 size = rlength;
669
670 /* ebml strings are usually not 0-terminated, so we allocate one
671 * byte more, read the string and NULL-terminate it ourselves. */
672 if (!(*str = av_malloc(size + 1))) {
673 av_log(matroska->ctx, AV_LOG_ERROR, "Memory allocation failed\n");
674 return AVERROR_NOMEM;
675 }
676 if (get_buffer(pb, (uint8_t *) *str, size) != size) {
677 offset_t pos = url_ftell(pb);
678 av_log(matroska->ctx, AV_LOG_ERROR,
679 "Read error at pos. %llu (0x%llx)\n", pos, pos);
680 return AVERROR_IO;
681 }
682 (*str)[size] = '\0';
683
684 return 0;
685}
686
687/*
688 * Read the next element as a UTF-8 string.
689 * 0 is success, < 0 is failure.
690 */
691
692static int
693ebml_read_utf8 (MatroskaDemuxContext *matroska,
694 uint32_t *id,
695 char **str)
696{
697 return ebml_read_ascii(matroska, id, str);
698}
699
700/*
701 * Read the next element as a date (nanoseconds since 1/1/2000).
702 * 0 is success, < 0 is failure.
703 */
704
705static int
706ebml_read_date (MatroskaDemuxContext *matroska,
707 uint32_t *id,
708 int64_t *date)
709{
710 return ebml_read_sint(matroska, id, date);
711}
712
713/*
714 * Read the next element, but only the header. The contents
715 * are supposed to be sub-elements which can be read separately.
716 * 0 is success, < 0 is failure.
717 */
718
719static int
720ebml_read_master (MatroskaDemuxContext *matroska,
721 uint32_t *id)
722{
723 ByteIOContext *pb = &matroska->ctx->pb;
724 uint64_t length;
725 MatroskaLevel *level;
726 int res;
727
728 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
729 (res = ebml_read_element_length(matroska, &length)) < 0)
730 return res;
731
732 /* protect... (Heaven forbids that the '>' is true) */
733 if (matroska->num_levels >= EBML_MAX_DEPTH) {
734 av_log(matroska->ctx, AV_LOG_ERROR,
735 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
736 return AVERROR_NOTSUPP;
737 }
738
739 /* remember level */
740 level = &matroska->levels[matroska->num_levels++];
741 level->start = url_ftell(pb);
742 level->length = length;
743
744 return 0;
745}
746
747/*
748 * Read the next element as binary data.
749 * 0 is success, < 0 is failure.
750 */
751
752static int
753ebml_read_binary (MatroskaDemuxContext *matroska,
754 uint32_t *id,
755 uint8_t **binary,
756 int *size)
757{
758 ByteIOContext *pb = &matroska->ctx->pb;
759 uint64_t rlength;
760 int res;
761
762 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
763 (res = ebml_read_element_length(matroska, &rlength)) < 0)
764 return res;
765 *size = rlength;
766
767 if (!(*binary = av_malloc(*size))) {
768 av_log(matroska->ctx, AV_LOG_ERROR,
769 "Memory allocation error\n");
770 return AVERROR_NOMEM;
771 }
772
773 if (get_buffer(pb, *binary, *size) != *size) {
774 offset_t pos = url_ftell(pb);
775 av_log(matroska->ctx, AV_LOG_ERROR,
776 "Read error at pos. %llu (0x%llx)\n", pos, pos);
777 return AVERROR_IO;
778 }
779
780 return 0;
781}
782
783/*
784 * Read signed/unsigned "EBML" numbers.
785 * Return: number of bytes processed, < 0 on error.
786 * XXX: use ebml_read_num().
787 */
788
789static int
790matroska_ebmlnum_uint (uint8_t *data,
791 uint32_t size,
792 uint64_t *num)
793{
794 int len_mask = 0x80, read = 1, n = 1, num_ffs = 0;
795 uint64_t total;
796
797 if (size <= 0)
798 return AVERROR_INVALIDDATA;
799
800 total = data[0];
801 while (read <= 8 && !(total & len_mask)) {
802 read++;
803 len_mask >>= 1;
804 }
805 if (read > 8)
806 return AVERROR_INVALIDDATA;
807
808 if ((total &= (len_mask - 1)) == len_mask - 1)
809 num_ffs++;
810 if (size < read)
811 return AVERROR_INVALIDDATA;
812 while (n < read) {
813 if (data[n] == 0xff)
814 num_ffs++;
815 total = (total << 8) | data[n];
816 n++;
817 }
818
819 if (!total)
820 return AVERROR_INVALIDDATA;
821
822 if (read == num_ffs)
823 *num = (uint64_t)-1;
824 else
825 *num = total;
826
827 return read;
828}
829
830/*
831 * Same as above, but signed.
832 */
833
834static int
835matroska_ebmlnum_sint (uint8_t *data,
836 uint32_t size,
837 int64_t *num)
838{
839 uint64_t unum;
840 int res;
841
842 /* read as unsigned number first */
843 if ((res = matroska_ebmlnum_uint(data, size, &unum)) < 0)
844 return res;
845
846 /* make signed (weird way) */
847 if (unum == (uint64_t)-1)
848 *num = INT64_MAX;
849 else
850 *num = unum - ((1LL << ((7 * res) - 1)) - 1);
851
852 return res;
853}
854
855/*
856 * Read an EBML header.
857 * 0 is success, < 0 is failure.
858 */
859
860static int
861ebml_read_header (MatroskaDemuxContext *matroska,
862 char **doctype,
863 int *version)
864{
865 uint32_t id;
866 int level_up, res = 0;
867
868 /* default init */
869 if (doctype)
870 *doctype = NULL;
871 if (version)
872 *version = 1;
873
874 if (!(id = ebml_peek_id(matroska, &level_up)) ||
875 level_up != 0 || id != EBML_ID_HEADER) {
876 av_log(matroska->ctx, AV_LOG_ERROR,
877 "This is not an EBML file (id=0x%x/0x%x)\n", id, EBML_ID_HEADER);
878 return AVERROR_INVALIDDATA;
879 }
880 if ((res = ebml_read_master(matroska, &id)) < 0)
881 return res;
882
883 while (res == 0) {
884 if (!(id = ebml_peek_id(matroska, &level_up)))
885 return AVERROR_IO;
886
887 /* end-of-header */
888 if (level_up)
889 break;
890
891 switch (id) {
892 /* is our read version uptodate? */
893 case EBML_ID_EBMLREADVERSION: {
894 uint64_t num;
895
896 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
897 return res;
898 if (num > EBML_VERSION) {
899 av_log(matroska->ctx, AV_LOG_ERROR,
900 "EBML version %llu (> %d) is not supported\n",
901 num, EBML_VERSION);
902 return AVERROR_INVALIDDATA;
903 }
904 break;
905 }
906
907 /* we only handle 8 byte lengths at max */
908 case EBML_ID_EBMLMAXSIZELENGTH: {
909 uint64_t num;
910
911 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
912 return res;
913 if (num > sizeof(uint64_t)) {
914 av_log(matroska->ctx, AV_LOG_ERROR,
915 "Integers of size %llu (> %d) not supported\n",
916 num, sizeof(uint64_t));
917 return AVERROR_INVALIDDATA;
918 }
919 break;
920 }
921
922 /* we handle 4 byte IDs at max */
923 case EBML_ID_EBMLMAXIDLENGTH: {
924 uint64_t num;
925
926 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
927 return res;
928 if (num > sizeof(uint32_t)) {
929 av_log(matroska->ctx, AV_LOG_ERROR,
e9afa2f4 930 "IDs of size %llu (> %u) not supported\n",
08abe0fd
MN
931 num, sizeof(uint32_t));
932 return AVERROR_INVALIDDATA;
933 }
934 break;
935 }
936
937 case EBML_ID_DOCTYPE: {
938 char *text;
939
940 if ((res = ebml_read_ascii(matroska, &id, &text)) < 0)
941 return res;
942 if (doctype) {
943 if (*doctype)
944 av_free(*doctype);
945 *doctype = text;
946 } else
947 av_free(text);
948 break;
949 }
950
951 case EBML_ID_DOCTYPEREADVERSION: {
952 uint64_t num;
953
954 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
955 return res;
956 if (version)
957 *version = num;
958 break;
959 }
960
961 default:
962 av_log(matroska->ctx, AV_LOG_INFO,
963 "Unknown data type 0x%x in EBML header", id);
964 /* pass-through */
965
966 case EBML_ID_VOID:
967 /* we ignore these two, as they don't tell us anything we
968 * care about */
969 case EBML_ID_EBMLVERSION:
970 case EBML_ID_DOCTYPEVERSION:
971 res = ebml_read_skip (matroska);
972 break;
973 }
974 }
975
976 return 0;
977}
978
979/*
980 * Put one packet in an application-supplied AVPacket struct.
981 * Returns 0 on success or -1 on failure.
982 */
983
984static int
985matroska_deliver_packet (MatroskaDemuxContext *matroska,
986 AVPacket *pkt)
987{
988 if (matroska->num_packets > 0) {
989 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
990 av_free(matroska->packets[0]);
991 if (matroska->num_packets > 1) {
992 memmove(&matroska->packets[0], &matroska->packets[1],
993 (matroska->num_packets - 1) * sizeof(AVPacket *));
994 matroska->packets =
995 av_realloc(matroska->packets, (matroska->num_packets - 1) *
996 sizeof(AVPacket *));
997 } else {
998 av_free(matroska->packets);
999 matroska->packets = NULL;
1000 }
1001 matroska->num_packets--;
1002 return 0;
1003 }
1004
1005 return -1;
1006}
1007
1008/*
1009 * Put a packet into our internal queue. Will be delivered to the
1010 * user/application during the next get_packet() call.
1011 */
1012
1013static void
1014matroska_queue_packet (MatroskaDemuxContext *matroska,
1015 AVPacket *pkt)
1016{
1017 matroska->packets =
1018 av_realloc(matroska->packets, (matroska->num_packets + 1) *
1019 sizeof(AVPacket *));
1020 matroska->packets[matroska->num_packets] = pkt;
1021 matroska->num_packets++;
1022}
1023
1024/*
1025 * Autodetecting...
1026 */
1027
1028static int
1029matroska_probe (AVProbeData *p)
1030{
1031 uint64_t total = 0;
1032 int len_mask = 0x80, size = 1, n = 1;
1033 uint8_t probe_data[] = { 'm', 'a', 't', 'r', 'o', 's', 'k', 'a' };
1034
1035 if (p->buf_size < 5)
1036 return 0;
1037
1038 /* ebml header? */
1039 if ((p->buf[0] << 24 | p->buf[1] << 16 |
1040 p->buf[2] << 8 | p->buf[3]) != EBML_ID_HEADER)
1041 return 0;
1042
1043 /* length of header */
1044 total = p->buf[4];
1045 while (size <= 8 && !(total & len_mask)) {
1046 size++;
1047 len_mask >>= 1;
1048 }
1049 if (size > 8)
1050 return 0;
1051 total &= (len_mask - 1);
1052 while (n < size)
1053 total = (total << 8) | p->buf[4 + n++];
1054
1055 /* does the probe data contain the whole header? */
1056 if (p->buf_size < 4 + size + total)
1057 return 0;
1058
1059 /* the header must contain the document type 'matroska'. For now,
1060 * we don't parse the whole header but simply check for the
1061 * availability of that array of characters inside the header.
1062 * Not fully fool-proof, but good enough. */
1063 for (n = 4 + size; n < 4 + size + total - sizeof(probe_data); n++)
1064 if (!memcmp (&p->buf[n], probe_data, sizeof(probe_data)))
1065 return AVPROBE_SCORE_MAX;
1066
1067 return 0;
1068}
1069
1070/*
1071 * From here on, it's all XML-style DTD stuff... Needs no comments.
1072 */
1073
1074static int
1075matroska_parse_info (MatroskaDemuxContext *matroska)
1076{
1077 int res = 0;
1078 uint32_t id;
1079
1080 av_log(matroska->ctx, AV_LOG_DEBUG, "Parsing info...\n");
1081
1082 while (res == 0) {
1083 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1084 res = AVERROR_IO;
1085 break;
1086 } else if (matroska->level_up) {
1087 matroska->level_up--;
1088 break;
1089 }
1090
1091 switch (id) {
1092 /* cluster timecode */
1093 case MATROSKA_ID_TIMECODESCALE: {
1094 uint64_t num;
1095 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1096 break;
1097 matroska->time_scale = num;
1098 break;
1099 }
1100
1101 case MATROSKA_ID_DURATION: {
1102 double num;
1103 if ((res = ebml_read_float(matroska, &id, &num)) < 0)
1104 break;
1105 matroska->duration = num * matroska->time_scale;
1106 break;
1107 }
1108
1109 case MATROSKA_ID_WRITINGAPP: {
1110 char *text;
1111 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1112 break;
1113 matroska->writing_app = text;
1114 break;
1115 }
1116
1117 case MATROSKA_ID_MUXINGAPP: {
1118 char *text;
1119 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1120 break;
1121 matroska->muxing_app = text;
1122 break;
1123 }
1124
1125 case MATROSKA_ID_DATEUTC: {
1126 int64_t time;
1127 if ((res = ebml_read_date(matroska, &id, &time)) < 0)
1128 break;
1129 matroska->created = time;
1130 break;
1131 }
1132
1133 default:
1134 av_log(matroska->ctx, AV_LOG_INFO,
1135 "Unknown entry 0x%x in info header\n", id);
1136 /* fall-through */
1137
1138 case EBML_ID_VOID:
1139 res = ebml_read_skip(matroska);
1140 break;
1141 }
1142
1143 if (matroska->level_up) {
1144 matroska->level_up--;
1145 break;
1146 }
1147 }
1148
1149 return res;
1150}
1151
1152static int
1153matroska_add_stream (MatroskaDemuxContext *matroska)
1154{
1155 int res = 0;
1156 uint32_t id;
1157 MatroskaTrack *track;
1158
1159 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing track, adding stream..,\n");
1160
1161 /* Allocate a generic track. As soon as we know its type we'll realloc. */
1162 track = av_mallocz(sizeof(MatroskaTrack));
1163 matroska->num_tracks++;
1164
1165 /* start with the master */
1166 if ((res = ebml_read_master(matroska, &id)) < 0)
1167 return res;
1168
1169 /* try reading the trackentry headers */
1170 while (res == 0) {
1171 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1172 res = AVERROR_IO;
1173 break;
1174 } else if (matroska->level_up > 0) {
1175 matroska->level_up--;
1176 break;
1177 }
1178
1179 switch (id) {
1180 /* track number (unique stream ID) */
1181 case MATROSKA_ID_TRACKNUMBER: {
1182 uint64_t num;
1183 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1184 break;
1185 track->num = num;
1186 break;
1187 }
1188
1189 /* track UID (unique identifier) */
1190 case MATROSKA_ID_TRACKUID: {
1191 uint64_t num;
1192 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1193 break;
1194 track->uid = num;
1195 break;
1196 }
1197
1198 /* track type (video, audio, combined, subtitle, etc.) */
1199 case MATROSKA_ID_TRACKTYPE: {
1200 uint64_t num;
1201 if (track->type != 0) {
1202 av_log(matroska->ctx, AV_LOG_INFO,
1203 "More than one tracktype in an entry - skip\n");
1204 break;
1205 }
1206 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1207 break;
1208 track->type = num;
1209
1210 /* ok, so we're actually going to reallocate this thing */
1211 switch (track->type) {
1212 case MATROSKA_TRACK_TYPE_VIDEO:
1213 track = (MatroskaTrack *)
1214 av_realloc(track, sizeof(MatroskaVideoTrack));
1215 break;
1216 case MATROSKA_TRACK_TYPE_AUDIO:
1217 track = (MatroskaTrack *)
1218 av_realloc(track, sizeof(MatroskaAudioTrack));
1219 ((MatroskaAudioTrack *)track)->channels = 1;
1220 ((MatroskaAudioTrack *)track)->samplerate = 8000;
1221 break;
1222 case MATROSKA_TRACK_TYPE_SUBTITLE:
1223 track = (MatroskaTrack *)
1224 av_realloc(track, sizeof(MatroskaSubtitleTrack));
1225 break;
1226 case MATROSKA_TRACK_TYPE_COMPLEX:
1227 case MATROSKA_TRACK_TYPE_LOGO:
1228 case MATROSKA_TRACK_TYPE_CONTROL:
1229 default:
1230 av_log(matroska->ctx, AV_LOG_INFO,
1231 "Unknown or unsupported track type 0x%x\n",
1232 track->type);
1233 track->type = 0;
1234 break;
1235 }
1236 matroska->tracks[matroska->num_tracks - 1] = track;
1237 break;
1238 }
1239
1240 /* tracktype specific stuff for video */
1241 case MATROSKA_ID_TRACKVIDEO: {
1242 MatroskaVideoTrack *videotrack;
1243 if (track->type != MATROSKA_TRACK_TYPE_VIDEO) {
1244 av_log(matroska->ctx, AV_LOG_INFO,
1245 "video data in non-video track - ignoring\n");
1246 res = AVERROR_INVALIDDATA;
1247 break;
1248 } else if ((res = ebml_read_master(matroska, &id)) < 0)
1249 break;
1250 videotrack = (MatroskaVideoTrack *)track;
1251
1252 while (res == 0) {
1253 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1254 res = AVERROR_IO;
1255 break;
1256 } else if (matroska->level_up > 0) {
1257 matroska->level_up--;
1258 break;
1259 }
1260
1261 switch (id) {
1262 /* fixme, this should be one-up, but I get it here */
1263 case MATROSKA_ID_TRACKDEFAULTDURATION: {
1264 uint64_t num;
1265 if ((res = ebml_read_uint (matroska, &id,
1266 &num)) < 0)
1267 break;
1268 track->default_duration = num;
1269 break;
1270 }
1271
1272 /* video framerate */
1273 case MATROSKA_ID_VIDEOFRAMERATE: {
1274 double num;
1275 if ((res = ebml_read_float(matroska, &id,
1276 &num)) < 0)
1277 break;
1278 track->default_duration = 1000000000 * (1. / num);
1279 break;
1280 }
1281
1282 /* width of the size to display the video at */
1283 case MATROSKA_ID_VIDEODISPLAYWIDTH: {
1284 uint64_t num;
1285 if ((res = ebml_read_uint(matroska, &id,
1286 &num)) < 0)
1287 break;
1288 videotrack->display_width = num;
1289 break;
1290 }
1291
1292 /* height of the size to display the video at */
1293 case MATROSKA_ID_VIDEODISPLAYHEIGHT: {
1294 uint64_t num;
1295 if ((res = ebml_read_uint(matroska, &id,
1296 &num)) < 0)
1297 break;
1298 videotrack->display_height = num;
1299 break;
1300 }
1301
1302 /* width of the video in the file */
1303 case MATROSKA_ID_VIDEOPIXELWIDTH: {
1304 uint64_t num;
1305 if ((res = ebml_read_uint(matroska, &id,
1306 &num)) < 0)
1307 break;
1308 videotrack->pixel_width = num;
1309 break;
1310 }
1311
1312 /* height of the video in the file */
1313 case MATROSKA_ID_VIDEOPIXELHEIGHT: {
1314 uint64_t num;
1315 if ((res = ebml_read_uint(matroska, &id,
1316 &num)) < 0)
1317 break;
1318 videotrack->pixel_height = num;
1319 break;
1320 }
1321
1322 /* whether the video is interlaced */
1323 case MATROSKA_ID_VIDEOFLAGINTERLACED: {
1324 uint64_t num;
1325 if ((res = ebml_read_uint(matroska, &id,
1326 &num)) < 0)
1327 break;
1328 if (num)
1329 track->flags |=
1330 MATROSKA_VIDEOTRACK_INTERLACED;
1331 else
1332 track->flags &=
1333 ~MATROSKA_VIDEOTRACK_INTERLACED;
1334 break;
1335 }
1336
1337 /* stereo mode (whether the video has two streams,
1338 * where one is for the left eye and the other for
1339 * the right eye, which creates a 3D-like
1340 * effect) */
1341 case MATROSKA_ID_VIDEOSTEREOMODE: {
1342 uint64_t num;
1343 if ((res = ebml_read_uint(matroska, &id,
1344 &num)) < 0)
1345 break;
1346 if (num != MATROSKA_EYE_MODE_MONO &&
1347 num != MATROSKA_EYE_MODE_LEFT &&
1348 num != MATROSKA_EYE_MODE_RIGHT &&
1349 num != MATROSKA_EYE_MODE_BOTH) {
1350 av_log(matroska->ctx, AV_LOG_INFO,
1351 "Ignoring unknown eye mode 0x%x\n",
1352 (uint32_t) num);
1353 break;
1354 }
1355 videotrack->eye_mode = num;
1356 break;
1357 }
1358
1359 /* aspect ratio behaviour */
1360 case MATROSKA_ID_VIDEOASPECTRATIO: {
1361 uint64_t num;
1362 if ((res = ebml_read_uint(matroska, &id,
1363 &num)) < 0)
1364 break;
1365 if (num != MATROSKA_ASPECT_RATIO_MODE_FREE &&
1366 num != MATROSKA_ASPECT_RATIO_MODE_KEEP &&
1367 num != MATROSKA_ASPECT_RATIO_MODE_FIXED) {
1368 av_log(matroska->ctx, AV_LOG_INFO,
1369 "Ignoring unknown aspect ratio 0x%x\n",
1370 (uint32_t) num);
1371 break;
1372 }
1373 videotrack->ar_mode = num;
1374 break;
1375 }
1376
1377 /* colourspace (only matters for raw video)
1378 * fourcc */
1379 case MATROSKA_ID_VIDEOCOLOURSPACE: {
1380 uint64_t num;
1381 if ((res = ebml_read_uint(matroska, &id,
1382 &num)) < 0)
1383 break;
1384 videotrack->fourcc = num;
1385 break;
1386 }
1387
1388 default:
1389 av_log(matroska->ctx, AV_LOG_INFO,
1390 "Unknown video track header entry "
1391 "0x%x - ignoring\n", id);
1392 /* pass-through */
1393
1394 case EBML_ID_VOID:
1395 res = ebml_read_skip(matroska);
1396 break;
1397 }
1398
1399 if (matroska->level_up) {
1400 matroska->level_up--;
1401 break;
1402 }
1403 }
1404 break;
1405 }
1406
1407 /* tracktype specific stuff for audio */
1408 case MATROSKA_ID_TRACKAUDIO: {
1409 MatroskaAudioTrack *audiotrack;
1410 if (track->type != MATROSKA_TRACK_TYPE_AUDIO) {
1411 av_log(matroska->ctx, AV_LOG_INFO,
1412 "audio data in non-audio track - ignoring\n");
1413 res = AVERROR_INVALIDDATA;
1414 break;
1415 } else if ((res = ebml_read_master(matroska, &id)) < 0)
1416 break;
1417 audiotrack = (MatroskaAudioTrack *)track;
1418
1419 while (res == 0) {
1420 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1421 res = AVERROR_IO;
1422 break;
1423 } else if (matroska->level_up > 0) {
1424 matroska->level_up--;
1425 break;
1426 }
1427
1428 switch (id) {
1429 /* samplerate */
1430 case MATROSKA_ID_AUDIOSAMPLINGFREQ: {
1431 double num;
1432 if ((res = ebml_read_float(matroska, &id,
1433 &num)) < 0)
1434 break;
1435 audiotrack->samplerate = num;
1436 break;
1437 }
1438
1439 /* bitdepth */
1440 case MATROSKA_ID_AUDIOBITDEPTH: {
1441 uint64_t num;
1442 if ((res = ebml_read_uint(matroska, &id,
1443 &num)) < 0)
1444 break;
1445 audiotrack->bitdepth = num;
1446 break;
1447 }
1448
1449 /* channels */
1450 case MATROSKA_ID_AUDIOCHANNELS: {
1451 uint64_t num;
1452 if ((res = ebml_read_uint(matroska, &id,
1453 &num)) < 0)
1454 break;
1455 audiotrack->channels = num;
1456 break;
1457 }
1458
1459 default:
1460 av_log(matroska->ctx, AV_LOG_INFO,
1461 "Unknown audio track header entry "
1462 "0x%x - ignoring\n", id);
1463 /* pass-through */
1464
1465 case EBML_ID_VOID:
1466 res = ebml_read_skip(matroska);
1467 break;
1468 }
1469
1470 if (matroska->level_up) {
1471 matroska->level_up--;
1472 break;
1473 }
1474 }
1475 break;
1476 }
1477
1478 /* codec identifier */
1479 case MATROSKA_ID_CODECID: {
1480 char *text;
1481 if ((res = ebml_read_ascii(matroska, &id, &text)) < 0)
1482 break;
1483 track->codec_id = text;
1484 break;
1485 }
1486
1487 /* codec private data */
1488 case MATROSKA_ID_CODECPRIVATE: {
1489 uint8_t *data;
1490 int size;
1491 if ((res = ebml_read_binary(matroska, &id, &data, &size) < 0))
1492 break;
1493 track->codec_priv = data;
1494 track->codec_priv_size = size;
1495 break;
1496 }
1497
1498 /* name of the codec */
1499 case MATROSKA_ID_CODECNAME: {
1500 char *text;
1501 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1502 break;
1503 track->codec_name = text;
1504 break;
1505 }
1506
1507 /* name of this track */
1508 case MATROSKA_ID_TRACKNAME: {
1509 char *text;
1510 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1511 break;
1512 track->name = text;
1513 break;
1514 }
1515
1516 /* language (matters for audio/subtitles, mostly) */
1517 case MATROSKA_ID_TRACKLANGUAGE: {
1518 char *text;
1519 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1520 break;
1521 track->language = text;
1522 break;
1523 }
1524
1525 /* whether this is actually used */
1526 case MATROSKA_ID_TRACKFLAGENABLED: {
1527 uint64_t num;
1528 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1529 break;
1530 if (num)
1531 track->flags |= MATROSKA_TRACK_ENABLED;
1532 else
1533 track->flags &= ~MATROSKA_TRACK_ENABLED;
1534 break;
1535 }
1536
1537 /* whether it's the default for this track type */
1538 case MATROSKA_ID_TRACKFLAGDEFAULT: {
1539 uint64_t num;
1540 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1541 break;
1542 if (num)
1543 track->flags |= MATROSKA_TRACK_DEFAULT;
1544 else
1545 track->flags &= ~MATROSKA_TRACK_DEFAULT;
1546 break;
1547 }
1548
1549 /* lacing (like MPEG, where blocks don't end/start on frame
1550 * boundaries) */
1551 case MATROSKA_ID_TRACKFLAGLACING: {
1552 uint64_t num;
1553 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1554 break;
1555 if (num)
1556 track->flags |= MATROSKA_TRACK_LACING;
1557 else
1558 track->flags &= ~MATROSKA_TRACK_LACING;
1559 break;
1560 }
1561
1562 /* default length (in time) of one data block in this track */
1563 case MATROSKA_ID_TRACKDEFAULTDURATION: {
1564 uint64_t num;
1565 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1566 break;
1567 track->default_duration = num;
1568 break;
1569 }
1570
1571 default:
1572 av_log(matroska->ctx, AV_LOG_INFO,
1573 "Unknown track header entry 0x%x - ignoring\n", id);
1574 /* pass-through */
1575
1576 case EBML_ID_VOID:
1577 /* we ignore these because they're nothing useful. */
1578 case MATROSKA_ID_CODECINFOURL:
1579 case MATROSKA_ID_CODECDOWNLOADURL:
1580 case MATROSKA_ID_TRACKMINCACHE:
1581 case MATROSKA_ID_TRACKMAXCACHE:
1582 res = ebml_read_skip(matroska);
1583 break;
1584 }
1585
1586 if (matroska->level_up) {
1587 matroska->level_up--;
1588 break;
1589 }
1590 }
1591
1592 return res;
1593}
1594
1595static int
1596matroska_parse_tracks (MatroskaDemuxContext *matroska)
1597{
1598 int res = 0;
1599 uint32_t id;
1600
1601 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing tracks...\n");
1602
1603 while (res == 0) {
1604 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1605 res = AVERROR_IO;
1606 break;
1607 } else if (matroska->level_up) {
1608 matroska->level_up--;
1609 break;
1610 }
1611
1612 switch (id) {
1613 /* one track within the "all-tracks" header */
1614 case MATROSKA_ID_TRACKENTRY:
1615 res = matroska_add_stream(matroska);
1616 break;
1617
1618 default:
1619 av_log(matroska->ctx, AV_LOG_INFO,
1620 "Unknown entry 0x%x in track header\n", id);
1621 /* fall-through */
1622
1623 case EBML_ID_VOID:
1624 res = ebml_read_skip(matroska);
1625 break;
1626 }
1627
1628 if (matroska->level_up) {
1629 matroska->level_up--;
1630 break;
1631 }
1632 }
1633
1634 return res;
1635}
1636
1637static int
1638matroska_parse_index (MatroskaDemuxContext *matroska)
1639{
1640 int res = 0;
1641 uint32_t id;
1642 MatroskaDemuxIndex idx;
1643
1644 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing index...\n");
1645
1646 while (res == 0) {
1647 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1648 res = AVERROR_IO;
1649 break;
1650 } else if (matroska->level_up) {
1651 matroska->level_up--;
1652 break;
1653 }
1654
1655 switch (id) {
1656 /* one single index entry ('point') */
1657 case MATROSKA_ID_POINTENTRY:
1658 if ((res = ebml_read_master(matroska, &id)) < 0)
1659 break;
1660
1661 /* in the end, we hope to fill one entry with a
1662 * timestamp, a file position and a tracknum */
1663 idx.pos = (uint64_t) -1;
1664 idx.time = (uint64_t) -1;
1665 idx.track = (uint16_t) -1;
1666
1667 while (res == 0) {
1668 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1669 res = AVERROR_IO;
1670 break;
1671 } else if (matroska->level_up) {
1672 matroska->level_up--;
1673 break;
1674 }
1675
1676 switch (id) {
1677 /* one single index entry ('point') */
1678 case MATROSKA_ID_CUETIME: {
1679 int64_t time;
1680 if ((res = ebml_read_uint(matroska, &id,
1681 &time)) < 0)
1682 break;
1683 idx.time = time * matroska->time_scale;
1684 break;
1685 }
1686
1687 /* position in the file + track to which it
1688 * belongs */
1689 case MATROSKA_ID_CUETRACKPOSITION:
1690 if ((res = ebml_read_master(matroska, &id)) < 0)
1691 break;
1692
1693 while (res == 0) {
1694 if (!(id = ebml_peek_id (matroska,
1695 &matroska->level_up))) {
1696 res = AVERROR_IO;
1697 break;
1698 } else if (matroska->level_up) {
1699 matroska->level_up--;
1700 break;
1701 }
1702
1703 switch (id) {
1704 /* track number */
1705 case MATROSKA_ID_CUETRACK: {
1706 uint64_t num;
1707 if ((res = ebml_read_uint(matroska,
1708 &id, &num)) < 0)
1709 break;
1710 idx.track = num;
1711 break;
1712 }
1713
1714 /* position in file */
1715 case MATROSKA_ID_CUECLUSTERPOSITION: {
1716 uint64_t num;
1717 if ((res = ebml_read_uint(matroska,
1718 &id, &num)) < 0)
1719 break;
1720 idx.pos = num;
1721 break;
1722 }
1723
1724 default:
1725 av_log(matroska->ctx, AV_LOG_INFO,
1726 "Unknown entry 0x%x in "
1727 "CuesTrackPositions\n", id);
1728 /* fall-through */
1729
1730 case EBML_ID_VOID:
1731 res = ebml_read_skip(matroska);
1732 break;
1733 }
1734
1735 if (matroska->level_up) {
1736 matroska->level_up--;
1737 break;
1738 }
1739 }
1740
1741 break;
1742
1743 default:
1744 av_log(matroska->ctx, AV_LOG_INFO,
1745 "Unknown entry 0x%x in cuespoint "
1746 "index\n", id);
1747 /* fall-through */
1748
1749 case EBML_ID_VOID:
1750 res = ebml_read_skip(matroska);
1751 break;
1752 }
1753
1754 if (matroska->level_up) {
1755 matroska->level_up--;
1756 break;
1757 }
1758 }
1759
1760 /* so let's see if we got what we wanted */
1761 if (idx.pos != (uint64_t) -1 &&
1762 idx.time != (uint64_t) -1 &&
1763 idx.track != (uint16_t) -1) {
1764 if (matroska->num_indexes % 32 == 0) {
1765 /* re-allocate bigger index */
1766 matroska->index =
1767 av_realloc(matroska->index,
1768 (matroska->num_indexes + 32) *
1769 sizeof(MatroskaDemuxIndex));
1770 }
1771 matroska->index[matroska->num_indexes] = idx;
1772 matroska->num_indexes++;
1773 }
1774 break;
1775
1776 default:
1777 av_log(matroska->ctx, AV_LOG_INFO,
1778 "Unknown entry 0x%x in cues header\n", id);
1779 /* fall-through */
1780
1781 case EBML_ID_VOID:
1782 res = ebml_read_skip(matroska);
1783 break;
1784 }
1785
1786 if (matroska->level_up) {
1787 matroska->level_up--;
1788 break;
1789 }
1790 }
1791
1792 return res;
1793}
1794
1795static int
1796matroska_parse_metadata (MatroskaDemuxContext *matroska)
1797{
1798 int res = 0;
1799 uint32_t id;
1800
1801 while (res == 0) {
1802 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1803 res = AVERROR_IO;
1804 break;
1805 } else if (matroska->level_up) {
1806 matroska->level_up--;
1807 break;
1808 }
1809
1810 switch (id) {
1811 /* Hm, this is unsupported... */
1812 default:
1813 av_log(matroska->ctx, AV_LOG_INFO,
1814 "Unknown entry 0x%x in metadata header\n", id);
1815 /* fall-through */
1816
1817 case EBML_ID_VOID:
1818 res = ebml_read_skip(matroska);
1819 break;
1820 }
1821
1822 if (matroska->level_up) {
1823 matroska->level_up--;
1824 break;
1825 }
1826 }
1827
1828 return res;
1829}
1830
1831static int
1832matroska_parse_seekhead (MatroskaDemuxContext *matroska)
1833{
1834 int res = 0;
1835 uint32_t id;
1836
1837 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing seekhead...\n");
1838
1839 while (res == 0) {
1840 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1841 res = AVERROR_IO;
1842 break;
1843 } else if (matroska->level_up) {
1844 matroska->level_up--;
1845 break;
1846 }
1847
1848 switch (id) {
1849 case MATROSKA_ID_SEEKENTRY: {
1850 uint32_t seek_id = 0, peek_id_cache = 0;
1851 uint64_t seek_pos = (uint64_t) -1, t;
1852
1853 if ((res = ebml_read_master(matroska, &id)) < 0)
1854 break;
1855
1856 while (res == 0) {
1857 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1858 res = AVERROR_IO;
1859 break;
1860 } else if (matroska->level_up) {
1861 matroska->level_up--;
1862 break;
1863 }
1864
1865 switch (id) {
1866 case MATROSKA_ID_SEEKID:
1867 res = ebml_read_uint(matroska, &id, &t);
1868 seek_id = t;
1869 break;
1870
1871 case MATROSKA_ID_SEEKPOSITION:
1872 res = ebml_read_uint(matroska, &id, &seek_pos);
1873 break;
1874
1875 default:
1876 av_log(matroska->ctx, AV_LOG_INFO,
1877 "Unknown seekhead ID 0x%x\n", id);
1878 /* fall-through */
1879
1880 case EBML_ID_VOID:
1881 res = ebml_read_skip(matroska);
1882 break;
1883 }
1884
1885 if (matroska->level_up) {
1886 matroska->level_up--;
1887 break;
1888 }
1889 }
1890
1891 if (!seek_id || seek_pos == (uint64_t) -1) {
1892 av_log(matroska->ctx, AV_LOG_INFO,
1893 "Incomplete seekhead entry (0x%x/%llu)\n",
1894 seek_id, seek_pos);
1895 break;
1896 }
1897
1898 switch (seek_id) {
1899 case MATROSKA_ID_CUES:
1900 case MATROSKA_ID_TAGS: {
1901 uint32_t level_up = matroska->level_up;
1902 offset_t before_pos;
1903 uint64_t length;
1904 MatroskaLevel level;
1905
1906 /* remember the peeked ID and the current position */
1907 peek_id_cache = matroska->peek_id;
1908 before_pos = url_ftell(&matroska->ctx->pb);
1909
1910 /* seek */
1911 if ((res = ebml_read_seek(matroska, seek_pos +
1912 matroska->segment_start)) < 0)
1913 return res;
1914
1915 /* we don't want to lose our seekhead level, so we add
1916 * a dummy. This is a crude hack. */
1917 if (matroska->num_levels == EBML_MAX_DEPTH) {
1918 av_log(matroska->ctx, AV_LOG_INFO,
1919 "Max EBML element depth (%d) reached, "
1920 "cannot parse further.\n", EBML_MAX_DEPTH);
1921 return AVERROR_UNKNOWN;
1922 }
1923
1924 level.start = 0;
1925 level.length = (uint64_t)-1;
1926 matroska->levels[matroska->num_levels] = level;
1927 matroska->num_levels++;
1928
1929 /* check ID */
1930 if (!(id = ebml_peek_id (matroska,
1931 &matroska->level_up)))
1932 break;
1933 if (id != seek_id) {
1934 av_log(matroska->ctx, AV_LOG_INFO,
1935 "We looked for ID=0x%x but got "
1936 "ID=0x%x (pos=%llu)",
1937 seek_id, id, seek_pos +
1938 matroska->segment_start);
1939 goto finish;
1940 }
1941
1942 /* read master + parse */
1943 if ((res = ebml_read_master(matroska, &id)) < 0)
1944 break;
1945 switch (id) {
1946 case MATROSKA_ID_CUES:
1947 if (!(res = matroska_parse_index(matroska)) ||
1948 url_feof(&matroska->ctx->pb)) {
1949 matroska->index_parsed = 1;
1950 res = 0;
1951 }
1952 break;
1953 case MATROSKA_ID_TAGS:
1954 if (!(res = matroska_parse_metadata(matroska)) ||
1955 url_feof(&matroska->ctx->pb)) {
1956 matroska->metadata_parsed = 1;
1957 res = 0;
1958 }
1959 break;
1960 }
1961 if (res < 0)
1962 break;
1963
1964 finish:
1965 /* remove dummy level */
1966 while (matroska->num_levels) {
1967 matroska->num_levels--;
1968 length =
1969 matroska->levels[matroska->num_levels].length;
1970 if (length == (uint64_t)-1)
1971 break;
1972 }
1973
1974 /* seek back */
1975 if ((res = ebml_read_seek(matroska, before_pos)) < 0)
1976 return res;
1977 matroska->peek_id = peek_id_cache;
1978 matroska->level_up = level_up;
1979 break;
1980 }
1981
1982 default:
1983 av_log(matroska->ctx, AV_LOG_INFO,
1984 "Ignoring seekhead entry for ID=0x%x\n",
1985 seek_id);
1986 break;
1987 }
1988
1989 break;
1990 }
1991
1992 default:
1993 av_log(matroska->ctx, AV_LOG_INFO,
1994 "Unknown seekhead ID 0x%x\n", id);
1995 /* fall-through */
1996
1997 case EBML_ID_VOID:
1998 res = ebml_read_skip(matroska);
1999 break;
2000 }
2001
2002 if (matroska->level_up) {
2003 matroska->level_up--;
2004 break;
2005 }
2006 }
2007
2008 return res;
2009}
2010
2011static int
2012matroska_read_header (AVFormatContext *s,
2013 AVFormatParameters *ap)
2014{
2015 MatroskaDemuxContext *matroska = s->priv_data;
2016 char *doctype;
2017 int version, last_level, res = 0;
2018 uint32_t id;
2019
2020 matroska->ctx = s;
2021
2022 /* First read the EBML header. */
2023 doctype = NULL;
2024 if ((res = ebml_read_header(matroska, &doctype, &version)) < 0)
2025 return res;
2026 if ((doctype == NULL) || strcmp(doctype, "matroska")) {
2027 av_log(matroska->ctx, AV_LOG_ERROR,
2028 "Wrong EBML doctype ('%s' != 'matroska').\n",
2029 doctype ? doctype : "(none)");
2030 if (doctype)
2031 av_free(doctype);
2032 return AVERROR_NOFMT;
2033 }
2034 av_free(doctype);
2035 if (version != 1) {
2036 av_log(matroska->ctx, AV_LOG_ERROR,
2037 "Matroska demuxer version 1 too old for file version %d\n",
2038 version);
2039 return AVERROR_NOFMT;
2040 }
2041
2042 /* The next thing is a segment. */
2043 while (1) {
2044 if (!(id = ebml_peek_id(matroska, &last_level)))
2045 return AVERROR_IO;
2046 if (id == MATROSKA_ID_SEGMENT)
2047 break;
2048
2049 /* oi! */
2050 av_log(matroska->ctx, AV_LOG_INFO,
2051 "Expected a Segment ID (0x%x), but received 0x%x!\n",
2052 MATROSKA_ID_SEGMENT, id);
2053 if ((res = ebml_read_skip(matroska)) < 0)
2054 return res;
2055 }
2056
2057 /* We now have a Matroska segment.
2058 * Seeks are from the beginning of the segment,
2059 * after the segment ID/length. */
2060 if ((res = ebml_read_master(matroska, &id)) < 0)
2061 return res;
2062 matroska->segment_start = url_ftell(&s->pb);
2063
2064 matroska->time_scale = 1000000;
2065 /* we've found our segment, start reading the different contents in here */
2066 while (res == 0) {
2067 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2068 res = AVERROR_IO;
2069 break;
2070 } else if (matroska->level_up) {
2071 matroska->level_up--;
2072 break;
2073 }
2074
2075 switch (id) {
2076 /* stream info */
2077 case MATROSKA_ID_INFO: {
2078 if ((res = ebml_read_master(matroska, &id)) < 0)
2079 break;
2080 res = matroska_parse_info(matroska);
2081 break;
2082 }
2083
2084 /* track info headers */
2085 case MATROSKA_ID_TRACKS: {
2086 if ((res = ebml_read_master(matroska, &id)) < 0)
2087 break;
2088 res = matroska_parse_tracks(matroska);
2089 break;
2090 }
2091
2092 /* stream index */
2093 case MATROSKA_ID_CUES: {
2094 if (!matroska->index_parsed) {
2095 if ((res = ebml_read_master(matroska, &id)) < 0)
2096 break;
2097 res = matroska_parse_index(matroska);
2098 } else
2099 res = ebml_read_skip(matroska);
2100 break;
2101 }
2102
2103 /* metadata */
2104 case MATROSKA_ID_TAGS: {
2105 if (!matroska->metadata_parsed) {
2106 if ((res = ebml_read_master(matroska, &id)) < 0)
2107 break;
2108 res = matroska_parse_metadata(matroska);
2109 } else
2110 res = ebml_read_skip(matroska);
2111 break;
2112 }
2113
2114 /* file index (if seekable, seek to Cues/Tags to parse it) */
2115 case MATROSKA_ID_SEEKHEAD: {
2116 if ((res = ebml_read_master(matroska, &id)) < 0)
2117 break;
2118 res = matroska_parse_seekhead(matroska);
2119 break;
2120 }
2121
2122 case MATROSKA_ID_CLUSTER: {
2123 /* Do not read the master - this will be done in the next
2124 * call to matroska_read_packet. */
2125 res = 1;
2126 break;
2127 }
2128
2129 default:
2130 av_log(matroska->ctx, AV_LOG_INFO,
2131 "Unknown matroska file header ID 0x%x\n", id);
2132 /* fall-through */
2133
2134 case EBML_ID_VOID:
2135 res = ebml_read_skip(matroska);
2136 break;
2137 }
2138
2139 if (matroska->level_up) {
2140 matroska->level_up--;
2141 break;
2142 }
2143 }
2144
2145 if (res < 0)
2146 return res;
2147
2148 /* Have we found a cluster? */
2149 if (res == 1) {
2150 int i;
2151 enum CodecID codec_id;
2152 MatroskaTrack *track;
2153 AVStream *st;
2154
08abe0fd
MN
2155 for (i = 0; i < matroska->num_tracks; i++) {
2156 track = matroska->tracks[i];
2157
2158 /* libavformat does not really support subtitles.
2159 * Also apply some sanity checks. */
2160 if ((track->type == MATROSKA_TRACK_TYPE_SUBTITLE) ||
2161 (track->codec_id == NULL))
2162 continue;
2163
2164 /* Set the FourCC from the CodecID. */
2165 /* This is the MS compatibility mode which stores a
2166 * BITMAPINFOHEADER in the CodecPrivate. */
2167 if (!strcmp(track->codec_id,
2168 MATROSKA_CODEC_ID_VIDEO_VFW_FOURCC) &&
2169 (track->codec_priv_size >= 40) &&
2170 (track->codec_priv != NULL)) {
2171 unsigned char *p;
2172
2173 /* Offset of biCompression. Stored in LE. */
2174 p = (unsigned char *)track->codec_priv + 16;
2175 ((MatroskaVideoTrack *)track)->fourcc = (p[3] << 24) |
2176 (p[2] << 16) | (p[1] << 8) | p[0];
2177 codec_id = codec_get_bmp_id(((MatroskaVideoTrack *)track)->fourcc);
2178
2179 } else if (!strcmp(track->codec_id,
2180 MATROSKA_CODEC_ID_VIDEO_MPEG4_SP) ||
2181 !strcmp(track->codec_id,
2182 MATROSKA_CODEC_ID_VIDEO_MPEG4_ASP) ||
2183 !strcmp(track->codec_id,
2184 MATROSKA_CODEC_ID_VIDEO_MPEG4_AP))
2185 codec_id = CODEC_ID_MPEG4;
2186/* else if (!strcmp(track->codec_id, */
2187/* MATROSKA_CODEC_ID_VIDEO_UNCOMPRESSED)) */
2188/* codec_id = CODEC_ID_???; */
2189 else if (!strcmp(track->codec_id,
2190 MATROSKA_CODEC_ID_VIDEO_MSMPEG4V3))
2191 codec_id = CODEC_ID_MSMPEG4V3;
2192 else if (!strcmp(track->codec_id,
2193 MATROSKA_CODEC_ID_VIDEO_MPEG1) ||
2194 !strcmp(track->codec_id,
2195 MATROSKA_CODEC_ID_VIDEO_MPEG2))
2196 codec_id = CODEC_ID_MPEG2VIDEO;
2197
2198 /* This is the MS compatibility mode which stores a
2199 * WAVEFORMATEX in the CodecPrivate. */
2200 else if (!strcmp(track->codec_id,
2201 MATROSKA_CODEC_ID_AUDIO_ACM) &&
2202 (track->codec_priv_size >= 18) &&
2203 (track->codec_priv != NULL)) {
2204 unsigned char *p;
2205 uint16_t tag;
2206
2207 /* Offset of wFormatTag. Stored in LE. */
2208 p = (unsigned char *)track->codec_priv;
2209 tag = (p[1] << 8) | p[0];
2210 codec_id = codec_get_wav_id(tag);
2211
2212 } else if (!strcmp(track->codec_id,
2213 MATROSKA_CODEC_ID_AUDIO_MPEG1_L1) ||
2214 !strcmp(track->codec_id,
2215 MATROSKA_CODEC_ID_AUDIO_MPEG1_L2) ||
2216 !strcmp(track->codec_id,
2217 MATROSKA_CODEC_ID_AUDIO_MPEG1_L3))
2218 codec_id = CODEC_ID_MP3;
2219 else if (!strcmp(track->codec_id,
2220 MATROSKA_CODEC_ID_AUDIO_PCM_INT_BE))
2221 codec_id = CODEC_ID_PCM_U16BE;
2222 else if (!strcmp(track->codec_id,
2223 MATROSKA_CODEC_ID_AUDIO_PCM_INT_LE))
2224 codec_id = CODEC_ID_PCM_U16LE;
2225/* else if (!strcmp(track->codec_id, */
2226/* MATROSKA_CODEC_ID_AUDIO_PCM_FLOAT)) */
2227/* codec_id = CODEC_ID_PCM_???; */
2228 else if (!strcmp(track->codec_id,
2229 MATROSKA_CODEC_ID_AUDIO_AC3))
2230 codec_id = CODEC_ID_AC3;
23c99253
MN
2231 else if (!strcmp(track->codec_id,
2232 MATROSKA_CODEC_ID_AUDIO_DTS))
2233 codec_id = CODEC_ID_DTS;
08abe0fd
MN
2234 /* No such codec id so far. */
2235/* else if (!strcmp(track->codec_id, */
2236/* MATROSKA_CODEC_ID_AUDIO_DTS)) */
2237/* codec_id = CODEC_ID_DTS; */
2238 else if (!strcmp(track->codec_id,
2239 MATROSKA_CODEC_ID_AUDIO_VORBIS))
2240 codec_id = CODEC_ID_VORBIS;
2241 else if (!strcmp(track->codec_id,
2242 MATROSKA_CODEC_ID_AUDIO_MPEG2) ||
2243 !strcmp(track->codec_id,
2244 MATROSKA_CODEC_ID_AUDIO_MPEG4))
2245 codec_id = CODEC_ID_AAC;
2246 else
2247 codec_id = CODEC_ID_NONE;
2248
2249 if (codec_id == CODEC_ID_NONE) {
2250 av_log(matroska->ctx, AV_LOG_INFO,
2251 "Unknown/unsupported CodecID %s.\n",
2252 track->codec_id);
2253 }
2254
2255 track->stream_index = matroska->num_streams;
2256
2257 matroska->num_streams++;
2258 st = av_new_stream(s, track->stream_index);
2259 if (st == NULL)
2260 return AVERROR_NOMEM;
9ee91c2f 2261 av_set_pts_info(st, 24, 1, 1000); /* 24 bit pts in ms */
08abe0fd
MN
2262
2263 st->codec.codec_id = codec_id;
2264
2265 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2266 MatroskaVideoTrack *videotrack = (MatroskaVideoTrack *)track;
2267
2268 st->codec.codec_type = CODEC_TYPE_VIDEO;
2269 st->codec.codec_tag = videotrack->fourcc;
2270 st->codec.width = videotrack->pixel_width;
2271 st->codec.height = videotrack->pixel_height;
2272 if (videotrack->display_width == 0)
2273 st->codec.sample_aspect_ratio.num =
2274 videotrack->pixel_width;
2275 else
2276 st->codec.sample_aspect_ratio.num =
2277 videotrack->display_width;
2278 if (videotrack->display_height == 0)
2279 st->codec.sample_aspect_ratio.num =
2280 videotrack->pixel_height;
2281 else
2282 st->codec.sample_aspect_ratio.num =
2283 videotrack->display_height;
2284
2285 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2286 MatroskaAudioTrack *audiotrack = (MatroskaAudioTrack *)track;
2287
2288 st->codec.codec_type = CODEC_TYPE_AUDIO;
2289 st->codec.sample_rate = audiotrack->samplerate;
2290 st->codec.channels = audiotrack->channels;
2291 }
2292
2293 /* What do we do with private data? E.g. for Vorbis. */
2294 }
2295 }
2296
2297 return 0;
2298}
2299
2300static int
2301matroska_find_track_by_num (MatroskaDemuxContext *matroska,
2302 int num)
2303{
2304 int i;
2305
2306 for (i = 0; i < matroska->num_tracks; i++)
2307 if (matroska->tracks[i]->num == num)
2308 return i;
2309
2310 return -1;
2311}
2312
2313static int
2314matroska_parse_blockgroup (MatroskaDemuxContext *matroska,
2315 uint64_t cluster_time)
2316{
2317 int res = 0;
2318 uint32_t id;
2319 AVPacket *pkt;
2320 int is_keyframe = PKT_FLAG_KEY, last_num_packets = matroska->num_packets;
2321
2322 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing blockgroup...\n");
2323
2324 while (res == 0) {
2325 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2326 res = AVERROR_IO;
2327 break;
2328 } else if (matroska->level_up) {
2329 matroska->level_up--;
2330 break;
2331 }
2332
2333 switch (id) {
2334 /* one block inside the group. Note, block parsing is one
2335 * of the harder things, so this code is a bit complicated.
2336 * See http://www.matroska.org/ for documentation. */
2337 case MATROSKA_ID_BLOCK: {
2338 uint8_t *data, *origdata;
2339 int size;
2340 uint64_t time;
2341 uint32_t *lace_size = NULL;
2342 int n, track, flags, laces = 0;
2343 uint64_t num;
2344
2345 if ((res = ebml_read_binary(matroska, &id, &data, &size)) < 0)
2346 break;
2347 origdata = data;
2348
2349 /* first byte(s): blocknum */
2350 if ((n = matroska_ebmlnum_uint(data, size, &num)) < 0) {
2351 av_log(matroska->ctx, AV_LOG_ERROR,
2352 "EBML block data error\n");
2353 av_free(origdata);
2354 break;
2355 }
2356 data += n;
2357 size -= n;
2358
2359 /* fetch track from num */
2360 track = matroska_find_track_by_num(matroska, num);
2361 if (size <= 3 || track < 0 || track >= matroska->num_tracks) {
2362 av_log(matroska->ctx, AV_LOG_INFO,
2363 "Invalid stream %d or size %u\n", track, size);
2364 av_free(origdata);
2365 break;
2366 }
2367
2368 /* time (relative to cluster time) */
2369 time = ((data[0] << 8) | data[1]) * matroska->time_scale;
2370 data += 2;
2371 size -= 2;
2372 flags = *data;
2373 data += 1;
2374 size -= 1;
2375 switch ((flags & 0x06) >> 1) {
2376 case 0x0: /* no lacing */
2377 laces = 1;
2378 lace_size = av_mallocz(sizeof(int));
2379 lace_size[0] = size;
2380 break;
2381
2382 case 0x1: /* xiph lacing */
2383 case 0x2: /* fixed-size lacing */
2384 case 0x3: /* EBML lacing */
2385 if (size == 0) {
2386 res = -1;
2387 break;
2388 }
2389 laces = (*data) + 1;
2390 data += 1;
2391 size -= 1;
2392 lace_size = av_mallocz(laces * sizeof(int));
2393
2394 switch ((flags & 0x06) >> 1) {
2395 case 0x1: /* xiph lacing */ {
2396 uint8_t temp;
2397 uint32_t total = 0;
2398 for (n = 0; res == 0 && n < laces - 1; n++) {
2399 while (1) {
2400 if (size == 0) {
2401 res = -1;
2402 break;
2403 }
2404 temp = *data;
2405 lace_size[n] += temp;
2406 data += 1;
2407 size -= 1;
2408 if (temp != 0xff)
2409 break;
2410 }
2411 total += lace_size[n];
2412 }
2413 lace_size[n] = size - total;
2414 break;
2415 }
2416
2417 case 0x2: /* fixed-size lacing */
2418 for (n = 0; n < laces; n++)
2419 lace_size[n] = size / laces;
2420 break;
2421
2422 case 0x3: /* EBML lacing */ {
2423 uint32_t total;
2424 n = matroska_ebmlnum_uint(data, size, &num);
2425 if (n < 0) {
2426 av_log(matroska->ctx, AV_LOG_INFO,
2427 "EBML block data error\n");
2428 break;
2429 }
2430 data += n;
2431 size -= n;
2432 total = lace_size[0] = num;
2433 for (n = 1; res == 0 && n < laces - 1; n++) {
2434 int64_t snum;
2435 int r;
2436 r = matroska_ebmlnum_sint (data, size,
2437 &snum);
2438 if (r < 0) {
2439 av_log(matroska->ctx, AV_LOG_INFO,
2440 "EBML block data error\n");
2441 break;
2442 }
2443 data += r;
2444 size -= r;
2445 lace_size[n] = lace_size[n - 1] + snum;
2446 total += lace_size[n];
2447 }
2448 lace_size[n] = size - total;
2449 break;
2450 }
2451 }
2452 break;
2453 }
2454
2455 if (res == 0) {
2456 for (n = 0; n < laces; n++) {
2457 uint64_t timecode = 0;
2458
2459 pkt = av_mallocz(sizeof(AVPacket));
2460 /* XXX: prevent data copy... */
2461 if (av_new_packet(pkt,lace_size[n]) < 0) {
2462 res = AVERROR_NOMEM;
2463 break;
2464 }
2465 if (cluster_time != (uint64_t)-1) {
2466 if (time < 0 && (-time) > cluster_time)
2467 timecode = cluster_time;
2468 else
2469 timecode = cluster_time + time;
2470 }
2471 /* FIXME: duration */
2472
2473 memcpy(pkt->data, data, lace_size[n]);
2474 data += lace_size[n];
2475 if (n == 0)
2476 pkt->flags = is_keyframe;
2477 pkt->stream_index =
2478 matroska->tracks[track]->stream_index;
2479
2480 pkt->pts = timecode / 1000000; /* ns to ms */
2481
2482 matroska_queue_packet(matroska, pkt);
2483 }
2484 }
2485
2486 av_free(lace_size);
2487 av_free(origdata);
2488 break;
2489 }
2490
2491 case MATROSKA_ID_BLOCKDURATION: {
2492 uint64_t num;
2493 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
2494 break;
2495 av_log(matroska->ctx, AV_LOG_INFO,
2496 "FIXME: implement support for BlockDuration\n");
2497 break;
2498 }
2499
2500 case MATROSKA_ID_BLOCKREFERENCE:
2501 /* We've found a reference, so not even the first frame in
2502 * the lace is a key frame. */
2503 is_keyframe = 0;
2504 if (last_num_packets != matroska->num_packets)
2505 matroska->packets[last_num_packets]->flags = 0;
2506 res = ebml_read_skip(matroska);
2507 break;
2508
2509 default:
2510 av_log(matroska->ctx, AV_LOG_INFO,
2511 "Unknown entry 0x%x in blockgroup data\n", id);
2512 /* fall-through */
2513
2514 case EBML_ID_VOID:
2515 res = ebml_read_skip(matroska);
2516 break;
2517 }
2518
2519 if (matroska->level_up) {
2520 matroska->level_up--;
2521 break;
2522 }
2523 }
2524
2525 return res;
2526}
2527
2528static int
2529matroska_parse_cluster (MatroskaDemuxContext *matroska)
2530{
2531 int res = 0;
2532 uint32_t id;
2533 uint64_t cluster_time = 0;
2534
2535 av_log(matroska->ctx, AV_LOG_DEBUG,
2536 "parsing cluster at %lld\n", url_ftell(&matroska->ctx->pb));
2537
2538 while (res == 0) {
2539 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2540 res = AVERROR_IO;
2541 break;
2542 } else if (matroska->level_up) {
2543 matroska->level_up--;
2544 break;
2545 }
2546
2547 switch (id) {
2548 /* cluster timecode */
2549 case MATROSKA_ID_CLUSTERTIMECODE: {
2550 uint64_t num;
2551 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
2552 break;
2553 cluster_time = num * matroska->time_scale;
2554 break;
2555 }
2556
2557 /* a group of blocks inside a cluster */
2558 case MATROSKA_ID_BLOCKGROUP:
2559 if ((res = ebml_read_master(matroska, &id)) < 0)
2560 break;
2561 res = matroska_parse_blockgroup(matroska, cluster_time);
2562 break;
2563
2564 default:
2565 av_log(matroska->ctx, AV_LOG_INFO,
2566 "Unknown entry 0x%x in cluster data\n", id);
2567 /* fall-through */
2568
2569 case EBML_ID_VOID:
2570 res = ebml_read_skip(matroska);
2571 break;
2572 }
2573
2574 if (matroska->level_up) {
2575 matroska->level_up--;
2576 break;
2577 }
2578 }
2579
2580 return res;
2581}
2582
2583static int
2584matroska_read_packet (AVFormatContext *s,
2585 AVPacket *pkt)
2586{
2587 MatroskaDemuxContext *matroska = s->priv_data;
2588 int res = 0;
2589 uint32_t id;
2590
2591 /* Do we still have a packet queued? */
2592 if (matroska_deliver_packet(matroska, pkt) == 0)
2593 return 0;
2594
2595 /* Have we already reached the end? */
2596 if (matroska->done)
2597 return AVERROR_IO;
2598
2599 while (res == 0) {
2600 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2601 res = AVERROR_IO;
2602 break;
2603 } else if (matroska->level_up) {
2604 matroska->level_up--;
2605 break;
2606 }
2607
2608 switch (id) {
2609 case MATROSKA_ID_CLUSTER:
2610 if ((res = ebml_read_master(matroska, &id)) < 0)
2611 break;
2612 if ((res = matroska_parse_cluster(matroska)) == 0)
2613 res = 1; /* Parsed one cluster, let's get out. */
2614 break;
2615
2616 default:
2617 case EBML_ID_VOID:
2618 res = ebml_read_skip(matroska);
2619 break;
2620 }
2621
2622 if (matroska->level_up) {
2623 matroska->level_up--;
2624 break;
2625 }
2626 }
2627
2628 if (res == -1)
2629 matroska->done = 1;
2630
2631 return matroska_deliver_packet(matroska, pkt);
2632}
2633
2634static int
2635matroska_read_close (AVFormatContext *s)
2636{
2637 MatroskaDemuxContext *matroska = s->priv_data;
2638 int n = 0;
2639
2640 if (matroska->writing_app)
2641 av_free(matroska->writing_app);
2642 if (matroska->muxing_app)
2643 av_free(matroska->muxing_app);
2644 if (matroska->index)
2645 av_free(matroska->index);
2646
2647 if (matroska->packets != NULL) {
2648 for (n = 0; n < matroska->num_packets; n++) {
2649 av_free_packet(matroska->packets[n]);
2650 av_free(matroska->packets[n]);
2651 }
2652 av_free(matroska->packets);
2653 }
2654
2655 for (n = 0; n < matroska->num_tracks; n++) {
2656 MatroskaTrack *track = matroska->tracks[n];
2657 if (track->codec_id)
2658 av_free(track->codec_id);
2659 if (track->codec_name)
2660 av_free(track->codec_name);
2661 if (track->codec_priv)
2662 av_free(track->codec_priv);
2663 if (track->name)
2664 av_free(track->name);
2665 if (track->language)
2666 av_free(track->language);
2667
2668 av_free(track);
2669 }
2670
2671 memset(matroska, 0, sizeof(MatroskaDemuxContext));
2672
2673 return 0;
2674}
2675
2676static AVInputFormat matroska_iformat = {
2677 "matroska",
2678 "Matroska file format",
2679 sizeof(MatroskaDemuxContext),
2680 matroska_probe,
2681 matroska_read_header,
2682 matroska_read_packet,
2683 matroska_read_close,
2684};
2685
2686int
2687matroska_init(void)
2688{
2689 av_register_input_format(&matroska_iformat);
2690 return 0;
2691}