matroskadec: stop parsing when skipping en element crossing over the end of file
[libav.git] / libavformat / matroskadec.c
1 /*
2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The FFmpeg Project
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 /**
23 * @file matroskadec.c
24 * Matroska file demuxer
25 * by Ronald Bultje <rbultje@ronald.bitfreak.net>
26 * with a little help from Moritz Bunkus <moritz@bunkus.org>
27 * totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28 * Specs available on the Matroska project page: http://www.matroska.org/.
29 */
30
31 #include "avformat.h"
32 /* For codec_get_id(). */
33 #include "riff.h"
34 #include "isom.h"
35 #include "matroska.h"
36 #include "libavcodec/mpeg4audio.h"
37 #include "libavutil/intfloat_readwrite.h"
38 #include "libavutil/avstring.h"
39 #include "libavutil/lzo.h"
40 #ifdef CONFIG_ZLIB
41 #include <zlib.h>
42 #endif
43 #ifdef CONFIG_BZLIB
44 #include <bzlib.h>
45 #endif
46
47 typedef enum {
48 EBML_NONE,
49 EBML_UINT,
50 EBML_FLOAT,
51 EBML_STR,
52 EBML_UTF8,
53 EBML_BIN,
54 EBML_NEST,
55 EBML_PASS,
56 EBML_STOP,
57 } EbmlType;
58
59 typedef const struct EbmlSyntax {
60 uint32_t id;
61 EbmlType type;
62 int list_elem_size;
63 int data_offset;
64 union {
65 uint64_t u;
66 double f;
67 const char *s;
68 const struct EbmlSyntax *n;
69 } def;
70 } EbmlSyntax;
71
72 typedef struct {
73 int nb_elem;
74 void *elem;
75 } EbmlList;
76
77 typedef struct {
78 int size;
79 uint8_t *data;
80 int64_t pos;
81 } EbmlBin;
82
83 typedef struct {
84 uint64_t version;
85 uint64_t max_size;
86 uint64_t id_length;
87 char *doctype;
88 uint64_t doctype_version;
89 } Ebml;
90
91 typedef struct {
92 uint64_t algo;
93 EbmlBin settings;
94 } MatroskaTrackCompression;
95
96 typedef struct {
97 uint64_t scope;
98 uint64_t type;
99 MatroskaTrackCompression compression;
100 } MatroskaTrackEncoding;
101
102 typedef struct {
103 double frame_rate;
104 uint64_t display_width;
105 uint64_t display_height;
106 uint64_t pixel_width;
107 uint64_t pixel_height;
108 uint64_t fourcc;
109 } MatroskaTrackVideo;
110
111 typedef struct {
112 double samplerate;
113 double out_samplerate;
114 uint64_t bitdepth;
115 uint64_t channels;
116
117 /* real audio header (extracted from extradata) */
118 int coded_framesize;
119 int sub_packet_h;
120 int frame_size;
121 int sub_packet_size;
122 int sub_packet_cnt;
123 int pkt_cnt;
124 uint8_t *buf;
125 } MatroskaTrackAudio;
126
127 typedef struct {
128 uint64_t num;
129 uint64_t type;
130 char *codec_id;
131 EbmlBin codec_priv;
132 char *language;
133 double time_scale;
134 uint64_t default_duration;
135 uint64_t flag_default;
136 MatroskaTrackVideo video;
137 MatroskaTrackAudio audio;
138 EbmlList encodings;
139
140 AVStream *stream;
141 } MatroskaTrack;
142
143 typedef struct {
144 char *filename;
145 char *mime;
146 EbmlBin bin;
147 } MatroskaAttachement;
148
149 typedef struct {
150 uint64_t start;
151 uint64_t end;
152 uint64_t uid;
153 char *title;
154 } MatroskaChapter;
155
156 typedef struct {
157 uint64_t track;
158 uint64_t pos;
159 } MatroskaIndexPos;
160
161 typedef struct {
162 uint64_t time;
163 EbmlList pos;
164 } MatroskaIndex;
165
166 typedef struct {
167 char *name;
168 char *string;
169 EbmlList sub;
170 } MatroskaTag;
171
172 typedef struct {
173 uint64_t id;
174 uint64_t pos;
175 } MatroskaSeekhead;
176
177 typedef struct {
178 uint64_t start;
179 uint64_t length;
180 } MatroskaLevel;
181
182 typedef struct {
183 AVFormatContext *ctx;
184
185 /* EBML stuff */
186 int num_levels;
187 MatroskaLevel levels[EBML_MAX_DEPTH];
188 int level_up;
189
190 uint64_t time_scale;
191 double duration;
192 char *title;
193 EbmlList tracks;
194 EbmlList attachments;
195 EbmlList chapters;
196 EbmlList index;
197 EbmlList tags;
198 EbmlList seekhead;
199
200 /* byte position of the segment inside the stream */
201 offset_t segment_start;
202
203 /* the packet queue */
204 AVPacket **packets;
205 int num_packets;
206
207 int done;
208 int has_cluster_id;
209
210 /* What to skip before effectively reading a packet. */
211 int skip_to_keyframe;
212 AVStream *skip_to_stream;
213 } MatroskaDemuxContext;
214
215 typedef struct {
216 uint64_t duration;
217 int64_t reference;
218 EbmlBin bin;
219 } MatroskaBlock;
220
221 typedef struct {
222 uint64_t timecode;
223 EbmlList blocks;
224 } MatroskaCluster;
225
226 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(*x))
227
228 static EbmlSyntax ebml_header[] = {
229 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
230 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
231 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
232 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
233 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
234 { EBML_ID_EBMLVERSION, EBML_NONE },
235 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
236 { 0 }
237 };
238
239 static EbmlSyntax ebml_syntax[] = {
240 { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
241 { 0 }
242 };
243
244 static EbmlSyntax matroska_info[] = {
245 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
246 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
247 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
248 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
249 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
250 { MATROSKA_ID_DATEUTC, EBML_NONE },
251 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
252 { 0 }
253 };
254
255 static EbmlSyntax matroska_track_video[] = {
256 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
257 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
258 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
259 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
260 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
261 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
262 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
263 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
264 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
265 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
266 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
267 { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
268 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_NONE },
269 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
270 { 0 }
271 };
272
273 static EbmlSyntax matroska_track_audio[] = {
274 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
275 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
276 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
277 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
278 { 0 }
279 };
280
281 static EbmlSyntax matroska_track_encoding_compression[] = {
282 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
283 { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
284 { 0 }
285 };
286
287 static EbmlSyntax matroska_track_encoding[] = {
288 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
289 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
290 { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
291 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
292 { 0 }
293 };
294
295 static EbmlSyntax matroska_track_encodings[] = {
296 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
297 { 0 }
298 };
299
300 static EbmlSyntax matroska_track[] = {
301 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
302 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
303 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
304 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
305 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
306 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
307 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
308 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
309 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
310 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
311 { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
312 { MATROSKA_ID_TRACKUID, EBML_NONE },
313 { MATROSKA_ID_TRACKNAME, EBML_NONE },
314 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
315 { MATROSKA_ID_TRACKFLAGFORCED, EBML_NONE },
316 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
317 { MATROSKA_ID_CODECNAME, EBML_NONE },
318 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
319 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
320 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
321 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
322 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
323 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
324 { 0 }
325 };
326
327 static EbmlSyntax matroska_tracks[] = {
328 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
329 { 0 }
330 };
331
332 static EbmlSyntax matroska_attachment[] = {
333 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
334 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
335 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
336 { MATROSKA_ID_FILEDESC, EBML_NONE },
337 { MATROSKA_ID_FILEUID, EBML_NONE },
338 { 0 }
339 };
340
341 static EbmlSyntax matroska_attachments[] = {
342 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
343 { 0 }
344 };
345
346 static EbmlSyntax matroska_chapter_display[] = {
347 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
348 { MATROSKA_ID_CHAPLANG, EBML_NONE },
349 { 0 }
350 };
351
352 static EbmlSyntax matroska_chapter_entry[] = {
353 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
354 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
355 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
356 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
357 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
358 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
359 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
360 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
361 { 0 }
362 };
363
364 static EbmlSyntax matroska_chapter[] = {
365 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
366 { MATROSKA_ID_EDITIONUID, EBML_NONE },
367 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
368 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
369 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
370 { 0 }
371 };
372
373 static EbmlSyntax matroska_chapters[] = {
374 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
375 { 0 }
376 };
377
378 static EbmlSyntax matroska_index_pos[] = {
379 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
380 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
381 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
382 { 0 }
383 };
384
385 static EbmlSyntax matroska_index_entry[] = {
386 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
387 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
388 { 0 }
389 };
390
391 static EbmlSyntax matroska_index[] = {
392 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
393 { 0 }
394 };
395
396 static EbmlSyntax matroska_simpletag[] = {
397 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
398 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
399 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
400 { MATROSKA_ID_TAGLANG, EBML_NONE },
401 { MATROSKA_ID_TAGDEFAULT, EBML_NONE },
402 { 0 }
403 };
404
405 static EbmlSyntax matroska_tag[] = {
406 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), 0, {.n=matroska_simpletag} },
407 { MATROSKA_ID_TAGTARGETS, EBML_NONE },
408 { 0 }
409 };
410
411 static EbmlSyntax matroska_tags[] = {
412 { MATROSKA_ID_TAG, EBML_NEST, 0, offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
413 { 0 }
414 };
415
416 static EbmlSyntax matroska_seekhead_entry[] = {
417 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
418 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
419 { 0 }
420 };
421
422 static EbmlSyntax matroska_seekhead[] = {
423 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
424 { 0 }
425 };
426
427 static EbmlSyntax matroska_segment[] = {
428 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
429 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
430 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
431 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
432 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
433 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
434 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
435 { MATROSKA_ID_CLUSTER, EBML_STOP, 0, offsetof(MatroskaDemuxContext,has_cluster_id) },
436 { 0 }
437 };
438
439 static EbmlSyntax matroska_segments[] = {
440 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
441 { 0 }
442 };
443
444 static EbmlSyntax matroska_blockgroup[] = {
445 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
446 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
447 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration), {.u=AV_NOPTS_VALUE} },
448 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
449 { 0 }
450 };
451
452 static EbmlSyntax matroska_cluster[] = {
453 { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
454 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
455 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
456 { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
457 { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
458 { 0 }
459 };
460
461 static EbmlSyntax matroska_clusters[] = {
462 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
463 { MATROSKA_ID_INFO, EBML_NONE },
464 { MATROSKA_ID_CUES, EBML_NONE },
465 { MATROSKA_ID_TAGS, EBML_NONE },
466 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
467 { 0 }
468 };
469
470 #define SIZE_OFF(x) sizeof(((AVFormatContext*)0)->x),offsetof(AVFormatContext,x)
471 const struct {
472 const char name[16];
473 int size;
474 int offset;
475 } metadata[] = {
476 { "TITLE", SIZE_OFF(title) },
477 { "ARTIST", SIZE_OFF(author) },
478 { "WRITTEN_BY", SIZE_OFF(author) },
479 { "LEAD_PERFORMER", SIZE_OFF(author) },
480 { "COPYRIGHT", SIZE_OFF(copyright) },
481 { "COMMENT", SIZE_OFF(comment) },
482 { "ALBUM", SIZE_OFF(album) },
483 { "DATE_WRITTEN", SIZE_OFF(year) },
484 { "DATE_RELEASED", SIZE_OFF(year) },
485 { "PART_NUMBER", SIZE_OFF(track) },
486 { "GENRE", SIZE_OFF(genre) },
487 };
488
489 /*
490 * Return: Whether we reached the end of a level in the hierarchy or not.
491 */
492 static int ebml_level_end(MatroskaDemuxContext *matroska)
493 {
494 ByteIOContext *pb = matroska->ctx->pb;
495 offset_t pos = url_ftell(pb);
496
497 if (matroska->num_levels > 0) {
498 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
499 if (pos - level->start >= level->length) {
500 matroska->num_levels--;
501 return 1;
502 }
503 }
504 return 0;
505 }
506
507 /*
508 * Read: an "EBML number", which is defined as a variable-length
509 * array of bytes. The first byte indicates the length by giving a
510 * number of 0-bits followed by a one. The position of the first
511 * "one" bit inside the first byte indicates the length of this
512 * number.
513 * Returns: number of bytes read, < 0 on error
514 */
515 static int ebml_read_num(MatroskaDemuxContext *matroska, ByteIOContext *pb,
516 int max_size, uint64_t *number)
517 {
518 int len_mask = 0x80, read = 1, n = 1;
519 int64_t total = 0;
520
521 /* The first byte tells us the length in bytes - get_byte() can normally
522 * return 0, but since that's not a valid first ebmlID byte, we can
523 * use it safely here to catch EOS. */
524 if (!(total = get_byte(pb))) {
525 /* we might encounter EOS here */
526 if (!url_feof(pb)) {
527 offset_t pos = url_ftell(pb);
528 av_log(matroska->ctx, AV_LOG_ERROR,
529 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
530 pos, pos);
531 }
532 return AVERROR(EIO); /* EOS or actual I/O error */
533 }
534
535 /* get the length of the EBML number */
536 while (read <= max_size && !(total & len_mask)) {
537 read++;
538 len_mask >>= 1;
539 }
540 if (read > max_size) {
541 offset_t pos = url_ftell(pb) - 1;
542 av_log(matroska->ctx, AV_LOG_ERROR,
543 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
544 (uint8_t) total, pos, pos);
545 return AVERROR_INVALIDDATA;
546 }
547
548 /* read out length */
549 total &= ~len_mask;
550 while (n++ < read)
551 total = (total << 8) | get_byte(pb);
552
553 *number = total;
554
555 return read;
556 }
557
558 /*
559 * Read the next element as an unsigned int.
560 * 0 is success, < 0 is failure.
561 */
562 static int ebml_read_uint(ByteIOContext *pb, int size, uint64_t *num)
563 {
564 int n = 0;
565
566 if (size < 1 || size > 8)
567 return AVERROR_INVALIDDATA;
568
569 /* big-endian ordering; build up number */
570 *num = 0;
571 while (n++ < size)
572 *num = (*num << 8) | get_byte(pb);
573
574 return 0;
575 }
576
577 /*
578 * Read the next element as a float.
579 * 0 is success, < 0 is failure.
580 */
581 static int ebml_read_float(ByteIOContext *pb, int size, double *num)
582 {
583 if (size == 4) {
584 *num= av_int2flt(get_be32(pb));
585 } else if(size==8){
586 *num= av_int2dbl(get_be64(pb));
587 } else
588 return AVERROR_INVALIDDATA;
589
590 return 0;
591 }
592
593 /*
594 * Read the next element as an ASCII string.
595 * 0 is success, < 0 is failure.
596 */
597 static int ebml_read_ascii(ByteIOContext *pb, int size, char **str)
598 {
599 av_free(*str);
600 /* EBML strings are usually not 0-terminated, so we allocate one
601 * byte more, read the string and NULL-terminate it ourselves. */
602 if (!(*str = av_malloc(size + 1)))
603 return AVERROR(ENOMEM);
604 if (get_buffer(pb, (uint8_t *) *str, size) != size) {
605 av_free(*str);
606 return AVERROR(EIO);
607 }
608 (*str)[size] = '\0';
609
610 return 0;
611 }
612
613 /*
614 * Read the next element as binary data.
615 * 0 is success, < 0 is failure.
616 */
617 static int ebml_read_binary(ByteIOContext *pb, int length, EbmlBin *bin)
618 {
619 av_free(bin->data);
620 if (!(bin->data = av_malloc(length)))
621 return AVERROR(ENOMEM);
622
623 bin->size = length;
624 bin->pos = url_ftell(pb);
625 if (get_buffer(pb, bin->data, length) != length)
626 return AVERROR(EIO);
627
628 return 0;
629 }
630
631 /*
632 * Read the next element, but only the header. The contents
633 * are supposed to be sub-elements which can be read separately.
634 * 0 is success, < 0 is failure.
635 */
636 static int ebml_read_master(MatroskaDemuxContext *matroska, int length)
637 {
638 ByteIOContext *pb = matroska->ctx->pb;
639 MatroskaLevel *level;
640
641 if (matroska->num_levels >= EBML_MAX_DEPTH) {
642 av_log(matroska->ctx, AV_LOG_ERROR,
643 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
644 return AVERROR(ENOSYS);
645 }
646
647 level = &matroska->levels[matroska->num_levels++];
648 level->start = url_ftell(pb);
649 level->length = length;
650
651 return 0;
652 }
653
654 /*
655 * Read signed/unsigned "EBML" numbers.
656 * Return: number of bytes processed, < 0 on error
657 */
658 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
659 uint8_t *data, uint32_t size, uint64_t *num)
660 {
661 ByteIOContext pb;
662 init_put_byte(&pb, data, size, 0, NULL, NULL, NULL, NULL);
663 return ebml_read_num(matroska, &pb, 8, num);
664 }
665
666 /*
667 * Same as above, but signed.
668 */
669 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
670 uint8_t *data, uint32_t size, int64_t *num)
671 {
672 uint64_t unum;
673 int res;
674
675 /* read as unsigned number first */
676 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
677 return res;
678
679 /* make signed (weird way) */
680 *num = unum - ((1LL << (7*res - 1)) - 1);
681
682 return res;
683 }
684
685 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
686 EbmlSyntax *syntax, void *data);
687
688 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
689 uint32_t id, void *data)
690 {
691 int i;
692 for (i=0; syntax[i].id; i++)
693 if (id == syntax[i].id)
694 break;
695 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32)
696 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
697 return ebml_parse_elem(matroska, &syntax[i], data);
698 }
699
700 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
701 void *data)
702 {
703 uint64_t id;
704 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
705 id |= 1 << 7*res;
706 return res < 0 ? res : ebml_parse_id(matroska, syntax, id, data);
707 }
708
709 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
710 void *data)
711 {
712 int i, res = 0;
713
714 for (i=0; syntax[i].id; i++)
715 switch (syntax[i].type) {
716 case EBML_UINT:
717 *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
718 break;
719 case EBML_FLOAT:
720 *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
721 break;
722 case EBML_STR:
723 case EBML_UTF8:
724 *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
725 break;
726 }
727
728 while (!res && !ebml_level_end(matroska))
729 res = ebml_parse(matroska, syntax, data);
730
731 return res;
732 }
733
734 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
735 EbmlSyntax *syntax, void *data)
736 {
737 ByteIOContext *pb = matroska->ctx->pb;
738 uint32_t id = syntax->id;
739 uint64_t length;
740 int res;
741
742 data = (char *)data + syntax->data_offset;
743 if (syntax->list_elem_size) {
744 EbmlList *list = data;
745 list->elem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
746 data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
747 memset(data, 0, syntax->list_elem_size);
748 list->nb_elem++;
749 }
750
751 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP)
752 if ((res = ebml_read_num(matroska, pb, 8, &length)) < 0)
753 return res;
754
755 switch (syntax->type) {
756 case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
757 case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
758 case EBML_STR:
759 case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
760 case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
761 case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
762 return res;
763 if (id == MATROSKA_ID_SEGMENT)
764 matroska->segment_start = url_ftell(matroska->ctx->pb);
765 return ebml_parse_nest(matroska, syntax->def.n, data);
766 case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
767 case EBML_STOP: *(int *)data = 1; return 1;
768 default: return url_fseek(pb,length,SEEK_CUR)<0 ? AVERROR(EIO) : 0;
769 }
770 if (res == AVERROR_INVALIDDATA)
771 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
772 else if (res == AVERROR(EIO))
773 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
774 return res;
775 }
776
777 static void ebml_free(EbmlSyntax *syntax, void *data)
778 {
779 int i, j;
780 for (i=0; syntax[i].id; i++) {
781 void *data_off = (char *)data + syntax[i].data_offset;
782 switch (syntax[i].type) {
783 case EBML_STR:
784 case EBML_UTF8: av_freep(data_off); break;
785 case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
786 case EBML_NEST:
787 if (syntax[i].list_elem_size) {
788 EbmlList *list = data_off;
789 char *ptr = list->elem;
790 for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
791 ebml_free(syntax[i].def.n, ptr);
792 av_free(list->elem);
793 } else
794 ebml_free(syntax[i].def.n, data_off);
795 default: break;
796 }
797 }
798 }
799
800
801 /*
802 * Autodetecting...
803 */
804 static int matroska_probe(AVProbeData *p)
805 {
806 uint64_t total = 0;
807 int len_mask = 0x80, size = 1, n = 1;
808 char probe_data[] = "matroska";
809
810 /* EBML header? */
811 if (AV_RB32(p->buf) != EBML_ID_HEADER)
812 return 0;
813
814 /* length of header */
815 total = p->buf[4];
816 while (size <= 8 && !(total & len_mask)) {
817 size++;
818 len_mask >>= 1;
819 }
820 if (size > 8)
821 return 0;
822 total &= (len_mask - 1);
823 while (n < size)
824 total = (total << 8) | p->buf[4 + n++];
825
826 /* Does the probe data contain the whole header? */
827 if (p->buf_size < 4 + size + total)
828 return 0;
829
830 /* The header must contain the document type 'matroska'. For now,
831 * we don't parse the whole header but simply check for the
832 * availability of that array of characters inside the header.
833 * Not fully fool-proof, but good enough. */
834 for (n = 4+size; n <= 4+size+total-(sizeof(probe_data)-1); n++)
835 if (!memcmp(p->buf+n, probe_data, sizeof(probe_data)-1))
836 return AVPROBE_SCORE_MAX;
837
838 return 0;
839 }
840
841 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
842 int num)
843 {
844 MatroskaTrack *tracks = matroska->tracks.elem;
845 int i;
846
847 for (i=0; i < matroska->tracks.nb_elem; i++)
848 if (tracks[i].num == num)
849 return &tracks[i];
850
851 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
852 return NULL;
853 }
854
855 static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
856 MatroskaTrack *track)
857 {
858 MatroskaTrackEncoding *encodings = track->encodings.elem;
859 uint8_t* data = *buf;
860 int isize = *buf_size;
861 uint8_t* pkt_data = NULL;
862 int pkt_size = isize;
863 int result = 0;
864 int olen;
865
866 switch (encodings[0].compression.algo) {
867 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
868 return encodings[0].compression.settings.size;
869 case MATROSKA_TRACK_ENCODING_COMP_LZO:
870 do {
871 olen = pkt_size *= 3;
872 pkt_data = av_realloc(pkt_data,
873 pkt_size+LZO_OUTPUT_PADDING);
874 result = lzo1x_decode(pkt_data, &olen, data, &isize);
875 } while (result==LZO_OUTPUT_FULL && pkt_size<10000000);
876 if (result)
877 goto failed;
878 pkt_size -= olen;
879 break;
880 #ifdef CONFIG_ZLIB
881 case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
882 z_stream zstream = {0};
883 if (inflateInit(&zstream) != Z_OK)
884 return -1;
885 zstream.next_in = data;
886 zstream.avail_in = isize;
887 do {
888 pkt_size *= 3;
889 pkt_data = av_realloc(pkt_data, pkt_size);
890 zstream.avail_out = pkt_size - zstream.total_out;
891 zstream.next_out = pkt_data + zstream.total_out;
892 result = inflate(&zstream, Z_NO_FLUSH);
893 } while (result==Z_OK && pkt_size<10000000);
894 pkt_size = zstream.total_out;
895 inflateEnd(&zstream);
896 if (result != Z_STREAM_END)
897 goto failed;
898 break;
899 }
900 #endif
901 #ifdef CONFIG_BZLIB
902 case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
903 bz_stream bzstream = {0};
904 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
905 return -1;
906 bzstream.next_in = data;
907 bzstream.avail_in = isize;
908 do {
909 pkt_size *= 3;
910 pkt_data = av_realloc(pkt_data, pkt_size);
911 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
912 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
913 result = BZ2_bzDecompress(&bzstream);
914 } while (result==BZ_OK && pkt_size<10000000);
915 pkt_size = bzstream.total_out_lo32;
916 BZ2_bzDecompressEnd(&bzstream);
917 if (result != BZ_STREAM_END)
918 goto failed;
919 break;
920 }
921 #endif
922 }
923
924 *buf = pkt_data;
925 *buf_size = pkt_size;
926 return 0;
927 failed:
928 av_free(pkt_data);
929 return -1;
930 }
931
932 static void matroska_convert_tags(AVFormatContext *s, EbmlList *list)
933 {
934 MatroskaTag *tags = list->elem;
935 int i, j;
936
937 for (i=0; i < list->nb_elem; i++) {
938 for (j=0; j < ARRAY_SIZE(metadata); j++){
939 if (!strcmp(tags[i].name, metadata[j].name)) {
940 int *ptr = (int *)((char *)s + metadata[j].offset);
941 if (*ptr) continue;
942 if (metadata[j].size > sizeof(int))
943 av_strlcpy((char *)ptr, tags[i].string, metadata[j].size);
944 else
945 *ptr = atoi(tags[i].string);
946 }
947 }
948 if (tags[i].sub.nb_elem)
949 matroska_convert_tags(s, &tags[i].sub);
950 }
951 }
952
953 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
954 {
955 EbmlList *seekhead_list = &matroska->seekhead;
956 MatroskaSeekhead *seekhead = seekhead_list->elem;
957 uint32_t level_up = matroska->level_up;
958 offset_t before_pos = url_ftell(matroska->ctx->pb);
959 MatroskaLevel level;
960 int i;
961
962 for (i=0; i<seekhead_list->nb_elem; i++) {
963 offset_t offset = seekhead[i].pos + matroska->segment_start;
964
965 if (seekhead[i].pos <= before_pos
966 || seekhead[i].id == MATROSKA_ID_SEEKHEAD
967 || seekhead[i].id == MATROSKA_ID_CLUSTER)
968 continue;
969
970 /* seek */
971 if (url_fseek(matroska->ctx->pb, offset, SEEK_SET) != offset)
972 continue;
973
974 /* We don't want to lose our seekhead level, so we add
975 * a dummy. This is a crude hack. */
976 if (matroska->num_levels == EBML_MAX_DEPTH) {
977 av_log(matroska->ctx, AV_LOG_INFO,
978 "Max EBML element depth (%d) reached, "
979 "cannot parse further.\n", EBML_MAX_DEPTH);
980 break;
981 }
982
983 level.start = 0;
984 level.length = (uint64_t)-1;
985 matroska->levels[matroska->num_levels] = level;
986 matroska->num_levels++;
987
988 ebml_parse(matroska, matroska_segment, matroska);
989
990 /* remove dummy level */
991 while (matroska->num_levels) {
992 uint64_t length = matroska->levels[--matroska->num_levels].length;
993 if (length == (uint64_t)-1)
994 break;
995 }
996 }
997
998 /* seek back */
999 url_fseek(matroska->ctx->pb, before_pos, SEEK_SET);
1000 matroska->level_up = level_up;
1001 }
1002
1003 static int matroska_aac_profile(char *codec_id)
1004 {
1005 static const char *aac_profiles[] = { "MAIN", "LC", "SSR" };
1006 int profile;
1007
1008 for (profile=0; profile<ARRAY_SIZE(aac_profiles); profile++)
1009 if (strstr(codec_id, aac_profiles[profile]))
1010 break;
1011 return profile + 1;
1012 }
1013
1014 static int matroska_aac_sri(int samplerate)
1015 {
1016 int sri;
1017
1018 for (sri=0; sri<ARRAY_SIZE(ff_mpeg4audio_sample_rates); sri++)
1019 if (ff_mpeg4audio_sample_rates[sri] == samplerate)
1020 break;
1021 return sri;
1022 }
1023
1024 static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
1025 {
1026 MatroskaDemuxContext *matroska = s->priv_data;
1027 EbmlList *attachements_list = &matroska->attachments;
1028 MatroskaAttachement *attachements;
1029 EbmlList *chapters_list = &matroska->chapters;
1030 MatroskaChapter *chapters;
1031 MatroskaTrack *tracks;
1032 EbmlList *index_list;
1033 MatroskaIndex *index;
1034 Ebml ebml = { 0 };
1035 AVStream *st;
1036 int i, j;
1037
1038 matroska->ctx = s;
1039
1040 /* First read the EBML header. */
1041 if (ebml_parse(matroska, ebml_syntax, &ebml)
1042 || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
1043 || ebml.id_length > sizeof(uint32_t) || strcmp(ebml.doctype, "matroska")
1044 || ebml.doctype_version > 2) {
1045 av_log(matroska->ctx, AV_LOG_ERROR,
1046 "EBML header using unsupported features\n"
1047 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1048 ebml.version, ebml.doctype, ebml.doctype_version);
1049 return AVERROR_NOFMT;
1050 }
1051 ebml_free(ebml_syntax, &ebml);
1052
1053 /* The next thing is a segment. */
1054 if (ebml_parse(matroska, matroska_segments, matroska) < 0)
1055 return -1;
1056 matroska_execute_seekhead(matroska);
1057
1058 if (matroska->duration)
1059 matroska->ctx->duration = matroska->duration * matroska->time_scale
1060 * 1000 / AV_TIME_BASE;
1061 if (matroska->title)
1062 strncpy(matroska->ctx->title, matroska->title,
1063 sizeof(matroska->ctx->title)-1);
1064 matroska_convert_tags(s, &matroska->tags);
1065
1066 tracks = matroska->tracks.elem;
1067 for (i=0; i < matroska->tracks.nb_elem; i++) {
1068 MatroskaTrack *track = &tracks[i];
1069 enum CodecID codec_id = CODEC_ID_NONE;
1070 EbmlList *encodings_list = &tracks->encodings;
1071 MatroskaTrackEncoding *encodings = encodings_list->elem;
1072 uint8_t *extradata = NULL;
1073 int extradata_size = 0;
1074 int extradata_offset = 0;
1075
1076 /* Apply some sanity checks. */
1077 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1078 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1079 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1080 av_log(matroska->ctx, AV_LOG_INFO,
1081 "Unknown or unsupported track type %"PRIu64"\n",
1082 track->type);
1083 continue;
1084 }
1085 if (track->codec_id == NULL)
1086 continue;
1087
1088 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1089 if (!track->default_duration)
1090 track->default_duration = 1000000000/track->video.frame_rate;
1091 if (!track->video.display_width)
1092 track->video.display_width = track->video.pixel_width;
1093 if (!track->video.display_height)
1094 track->video.display_height = track->video.pixel_height;
1095 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1096 if (!track->audio.out_samplerate)
1097 track->audio.out_samplerate = track->audio.samplerate;
1098 }
1099 if (encodings_list->nb_elem > 1) {
1100 av_log(matroska->ctx, AV_LOG_ERROR,
1101 "Multiple combined encodings no supported");
1102 } else if (encodings_list->nb_elem == 1) {
1103 if (encodings[0].type ||
1104 (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
1105 #ifdef CONFIG_ZLIB
1106 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1107 #endif
1108 #ifdef CONFIG_BZLIB
1109 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1110 #endif
1111 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
1112 encodings[0].scope = 0;
1113 av_log(matroska->ctx, AV_LOG_ERROR,
1114 "Unsupported encoding type");
1115 } else if (track->codec_priv.size && encodings[0].scope&2) {
1116 uint8_t *codec_priv = track->codec_priv.data;
1117 int offset = matroska_decode_buffer(&track->codec_priv.data,
1118 &track->codec_priv.size,
1119 track);
1120 if (offset < 0) {
1121 track->codec_priv.data = NULL;
1122 track->codec_priv.size = 0;
1123 av_log(matroska->ctx, AV_LOG_ERROR,
1124 "Failed to decode codec private data\n");
1125 } else if (offset > 0) {
1126 track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
1127 memcpy(track->codec_priv.data,
1128 encodings[0].compression.settings.data, offset);
1129 memcpy(track->codec_priv.data+offset, codec_priv,
1130 track->codec_priv.size);
1131 track->codec_priv.size += offset;
1132 }
1133 if (codec_priv != track->codec_priv.data)
1134 av_free(codec_priv);
1135 }
1136 }
1137
1138 for(j=0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++){
1139 if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1140 strlen(ff_mkv_codec_tags[j].str))){
1141 codec_id= ff_mkv_codec_tags[j].id;
1142 break;
1143 }
1144 }
1145
1146 st = track->stream = av_new_stream(s, 0);
1147 if (st == NULL)
1148 return AVERROR(ENOMEM);
1149
1150 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
1151 && track->codec_priv.size >= 40
1152 && track->codec_priv.data != NULL) {
1153 track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1154 codec_id = codec_get_id(codec_bmp_tags, track->video.fourcc);
1155 } else if (!strcmp(track->codec_id, "A_MS/ACM")
1156 && track->codec_priv.size >= 18
1157 && track->codec_priv.data != NULL) {
1158 uint16_t tag = AV_RL16(track->codec_priv.data);
1159 codec_id = codec_get_id(codec_wav_tags, tag);
1160 } else if (!strcmp(track->codec_id, "V_QUICKTIME")
1161 && (track->codec_priv.size >= 86)
1162 && (track->codec_priv.data != NULL)) {
1163 track->video.fourcc = AV_RL32(track->codec_priv.data);
1164 codec_id=codec_get_id(codec_movvideo_tags, track->video.fourcc);
1165 } else if (codec_id == CODEC_ID_PCM_S16BE) {
1166 switch (track->audio.bitdepth) {
1167 case 8: codec_id = CODEC_ID_PCM_U8; break;
1168 case 24: codec_id = CODEC_ID_PCM_S24BE; break;
1169 case 32: codec_id = CODEC_ID_PCM_S32BE; break;
1170 }
1171 } else if (codec_id == CODEC_ID_PCM_S16LE) {
1172 switch (track->audio.bitdepth) {
1173 case 8: codec_id = CODEC_ID_PCM_U8; break;
1174 case 24: codec_id = CODEC_ID_PCM_S24LE; break;
1175 case 32: codec_id = CODEC_ID_PCM_S32LE; break;
1176 }
1177 } else if (codec_id==CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
1178 codec_id = CODEC_ID_PCM_F64LE;
1179 } else if (codec_id == CODEC_ID_AAC && !track->codec_priv.size) {
1180 int profile = matroska_aac_profile(track->codec_id);
1181 int sri = matroska_aac_sri(track->audio.samplerate);
1182 extradata = av_malloc(5);
1183 if (extradata == NULL)
1184 return AVERROR(ENOMEM);
1185 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
1186 extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
1187 if (strstr(track->codec_id, "SBR")) {
1188 sri = matroska_aac_sri(track->audio.out_samplerate);
1189 extradata[2] = 0x56;
1190 extradata[3] = 0xE5;
1191 extradata[4] = 0x80 | (sri<<3);
1192 extradata_size = 5;
1193 } else
1194 extradata_size = 2;
1195 } else if (codec_id == CODEC_ID_TTA) {
1196 ByteIOContext b;
1197 extradata_size = 30;
1198 extradata = av_mallocz(extradata_size);
1199 if (extradata == NULL)
1200 return AVERROR(ENOMEM);
1201 init_put_byte(&b, extradata, extradata_size, 1,
1202 NULL, NULL, NULL, NULL);
1203 put_buffer(&b, "TTA1", 4);
1204 put_le16(&b, 1);
1205 put_le16(&b, track->audio.channels);
1206 put_le16(&b, track->audio.bitdepth);
1207 put_le32(&b, track->audio.out_samplerate);
1208 put_le32(&b, matroska->ctx->duration * track->audio.out_samplerate);
1209 } else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
1210 codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
1211 extradata_offset = 26;
1212 track->codec_priv.size -= extradata_offset;
1213 } else if (codec_id == CODEC_ID_RA_144) {
1214 track->audio.out_samplerate = 8000;
1215 track->audio.channels = 1;
1216 } else if (codec_id == CODEC_ID_RA_288 || codec_id == CODEC_ID_COOK ||
1217 codec_id == CODEC_ID_ATRAC3) {
1218 ByteIOContext b;
1219
1220 init_put_byte(&b, track->codec_priv.data,track->codec_priv.size,
1221 0, NULL, NULL, NULL, NULL);
1222 url_fskip(&b, 24);
1223 track->audio.coded_framesize = get_be32(&b);
1224 url_fskip(&b, 12);
1225 track->audio.sub_packet_h = get_be16(&b);
1226 track->audio.frame_size = get_be16(&b);
1227 track->audio.sub_packet_size = get_be16(&b);
1228 track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
1229 if (codec_id == CODEC_ID_RA_288) {
1230 st->codec->block_align = track->audio.coded_framesize;
1231 track->codec_priv.size = 0;
1232 } else {
1233 st->codec->block_align = track->audio.sub_packet_size;
1234 extradata_offset = 78;
1235 track->codec_priv.size -= extradata_offset;
1236 }
1237 }
1238
1239 if (codec_id == CODEC_ID_NONE)
1240 av_log(matroska->ctx, AV_LOG_INFO,
1241 "Unknown/unsupported CodecID %s.\n", track->codec_id);
1242
1243 av_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
1244
1245 st->codec->codec_id = codec_id;
1246 st->start_time = 0;
1247 if (strcmp(track->language, "und"))
1248 av_strlcpy(st->language, track->language, 4);
1249
1250 if (track->flag_default)
1251 st->disposition |= AV_DISPOSITION_DEFAULT;
1252
1253 if (track->default_duration)
1254 av_reduce(&st->codec->time_base.num, &st->codec->time_base.den,
1255 track->default_duration, 1000000000, 30000);
1256
1257 if(extradata){
1258 st->codec->extradata = extradata;
1259 st->codec->extradata_size = extradata_size;
1260 } else if(track->codec_priv.data && track->codec_priv.size > 0){
1261 st->codec->extradata = av_malloc(track->codec_priv.size);
1262 if(st->codec->extradata == NULL)
1263 return AVERROR(ENOMEM);
1264 st->codec->extradata_size = track->codec_priv.size;
1265 memcpy(st->codec->extradata,
1266 track->codec_priv.data + extradata_offset,
1267 track->codec_priv.size);
1268 }
1269
1270 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1271 st->codec->codec_type = CODEC_TYPE_VIDEO;
1272 st->codec->codec_tag = track->video.fourcc;
1273 st->codec->width = track->video.pixel_width;
1274 st->codec->height = track->video.pixel_height;
1275 av_reduce(&st->codec->sample_aspect_ratio.num,
1276 &st->codec->sample_aspect_ratio.den,
1277 st->codec->height * track->video.display_width,
1278 st->codec-> width * track->video.display_height,
1279 255);
1280 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1281 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1282 st->codec->codec_type = CODEC_TYPE_AUDIO;
1283 st->codec->sample_rate = track->audio.out_samplerate;
1284 st->codec->channels = track->audio.channels;
1285 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1286 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
1287 }
1288 }
1289
1290 attachements = attachements_list->elem;
1291 for (j=0; j<attachements_list->nb_elem; j++) {
1292 if (!(attachements[j].filename && attachements[j].mime &&
1293 attachements[j].bin.data && attachements[j].bin.size > 0)) {
1294 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1295 } else {
1296 AVStream *st = av_new_stream(s, 0);
1297 if (st == NULL)
1298 break;
1299 st->filename = av_strdup(attachements[j].filename);
1300 st->codec->codec_id = CODEC_ID_NONE;
1301 st->codec->codec_type = CODEC_TYPE_ATTACHMENT;
1302 st->codec->extradata = av_malloc(attachements[j].bin.size);
1303 if(st->codec->extradata == NULL)
1304 break;
1305 st->codec->extradata_size = attachements[j].bin.size;
1306 memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
1307
1308 for (i=0; ff_mkv_mime_tags[i].id != CODEC_ID_NONE; i++) {
1309 if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
1310 strlen(ff_mkv_mime_tags[i].str))) {
1311 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1312 break;
1313 }
1314 }
1315 }
1316 }
1317
1318 chapters = chapters_list->elem;
1319 for (i=0; i<chapters_list->nb_elem; i++)
1320 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid)
1321 ff_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
1322 chapters[i].start, chapters[i].end,
1323 chapters[i].title);
1324
1325 index_list = &matroska->index;
1326 index = index_list->elem;
1327 for (i=0; i<index_list->nb_elem; i++) {
1328 EbmlList *pos_list = &index[i].pos;
1329 MatroskaIndexPos *pos = pos_list->elem;
1330 for (j=0; j<pos_list->nb_elem; j++) {
1331 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1332 pos[j].track);
1333 if (track && track->stream)
1334 av_add_index_entry(track->stream,
1335 pos[j].pos + matroska->segment_start,
1336 index[i].time*matroska->time_scale/AV_TIME_BASE,
1337 0, 0, AVINDEX_KEYFRAME);
1338 }
1339 }
1340
1341 return 0;
1342 }
1343
1344 /*
1345 * Put one packet in an application-supplied AVPacket struct.
1346 * Returns 0 on success or -1 on failure.
1347 */
1348 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1349 AVPacket *pkt)
1350 {
1351 if (matroska->num_packets > 0) {
1352 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1353 av_free(matroska->packets[0]);
1354 if (matroska->num_packets > 1) {
1355 memmove(&matroska->packets[0], &matroska->packets[1],
1356 (matroska->num_packets - 1) * sizeof(AVPacket *));
1357 matroska->packets =
1358 av_realloc(matroska->packets, (matroska->num_packets - 1) *
1359 sizeof(AVPacket *));
1360 } else {
1361 av_freep(&matroska->packets);
1362 }
1363 matroska->num_packets--;
1364 return 0;
1365 }
1366
1367 return -1;
1368 }
1369
1370 /*
1371 * Free all packets in our internal queue.
1372 */
1373 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1374 {
1375 if (matroska->packets) {
1376 int n;
1377 for (n = 0; n < matroska->num_packets; n++) {
1378 av_free_packet(matroska->packets[n]);
1379 av_free(matroska->packets[n]);
1380 }
1381 av_freep(&matroska->packets);
1382 matroska->num_packets = 0;
1383 }
1384 }
1385
1386 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
1387 int size, int64_t pos, uint64_t cluster_time,
1388 uint64_t duration, int is_keyframe)
1389 {
1390 MatroskaTrack *track;
1391 int res = 0;
1392 AVStream *st;
1393 AVPacket *pkt;
1394 int16_t block_time;
1395 uint32_t *lace_size = NULL;
1396 int n, flags, laces = 0;
1397 uint64_t num;
1398
1399 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
1400 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
1401 return res;
1402 }
1403 data += n;
1404 size -= n;
1405
1406 track = matroska_find_track_by_num(matroska, num);
1407 if (size <= 3 || !track || !track->stream) {
1408 av_log(matroska->ctx, AV_LOG_INFO,
1409 "Invalid stream %"PRIu64" or size %u\n", num, size);
1410 return res;
1411 }
1412 st = track->stream;
1413 if (st->discard >= AVDISCARD_ALL)
1414 return res;
1415 if (duration == AV_NOPTS_VALUE)
1416 duration = track->default_duration / matroska->time_scale;
1417
1418 block_time = AV_RB16(data);
1419 data += 2;
1420 flags = *data++;
1421 size -= 3;
1422 if (is_keyframe == -1)
1423 is_keyframe = flags & 0x80 ? PKT_FLAG_KEY : 0;
1424
1425 if (matroska->skip_to_keyframe) {
1426 if (!is_keyframe || st != matroska->skip_to_stream)
1427 return res;
1428 matroska->skip_to_keyframe = 0;
1429 }
1430
1431 switch ((flags & 0x06) >> 1) {
1432 case 0x0: /* no lacing */
1433 laces = 1;
1434 lace_size = av_mallocz(sizeof(int));
1435 lace_size[0] = size;
1436 break;
1437
1438 case 0x1: /* Xiph lacing */
1439 case 0x2: /* fixed-size lacing */
1440 case 0x3: /* EBML lacing */
1441 assert(size>0); // size <=3 is checked before size-=3 above
1442 laces = (*data) + 1;
1443 data += 1;
1444 size -= 1;
1445 lace_size = av_mallocz(laces * sizeof(int));
1446
1447 switch ((flags & 0x06) >> 1) {
1448 case 0x1: /* Xiph lacing */ {
1449 uint8_t temp;
1450 uint32_t total = 0;
1451 for (n = 0; res == 0 && n < laces - 1; n++) {
1452 while (1) {
1453 if (size == 0) {
1454 res = -1;
1455 break;
1456 }
1457 temp = *data;
1458 lace_size[n] += temp;
1459 data += 1;
1460 size -= 1;
1461 if (temp != 0xff)
1462 break;
1463 }
1464 total += lace_size[n];
1465 }
1466 lace_size[n] = size - total;
1467 break;
1468 }
1469
1470 case 0x2: /* fixed-size lacing */
1471 for (n = 0; n < laces; n++)
1472 lace_size[n] = size / laces;
1473 break;
1474
1475 case 0x3: /* EBML lacing */ {
1476 uint32_t total;
1477 n = matroska_ebmlnum_uint(matroska, data, size, &num);
1478 if (n < 0) {
1479 av_log(matroska->ctx, AV_LOG_INFO,
1480 "EBML block data error\n");
1481 break;
1482 }
1483 data += n;
1484 size -= n;
1485 total = lace_size[0] = num;
1486 for (n = 1; res == 0 && n < laces - 1; n++) {
1487 int64_t snum;
1488 int r;
1489 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
1490 if (r < 0) {
1491 av_log(matroska->ctx, AV_LOG_INFO,
1492 "EBML block data error\n");
1493 break;
1494 }
1495 data += r;
1496 size -= r;
1497 lace_size[n] = lace_size[n - 1] + snum;
1498 total += lace_size[n];
1499 }
1500 lace_size[n] = size - total;
1501 break;
1502 }
1503 }
1504 break;
1505 }
1506
1507 if (res == 0) {
1508 uint64_t timecode = AV_NOPTS_VALUE;
1509
1510 if (cluster_time != (uint64_t)-1
1511 && (block_time >= 0 || cluster_time >= -block_time))
1512 timecode = cluster_time + block_time;
1513
1514 for (n = 0; n < laces; n++) {
1515 if (st->codec->codec_id == CODEC_ID_RA_288 ||
1516 st->codec->codec_id == CODEC_ID_COOK ||
1517 st->codec->codec_id == CODEC_ID_ATRAC3) {
1518 int a = st->codec->block_align;
1519 int sps = track->audio.sub_packet_size;
1520 int cfs = track->audio.coded_framesize;
1521 int h = track->audio.sub_packet_h;
1522 int y = track->audio.sub_packet_cnt;
1523 int w = track->audio.frame_size;
1524 int x;
1525
1526 if (!track->audio.pkt_cnt) {
1527 if (st->codec->codec_id == CODEC_ID_RA_288)
1528 for (x=0; x<h/2; x++)
1529 memcpy(track->audio.buf+x*2*w+y*cfs,
1530 data+x*cfs, cfs);
1531 else
1532 for (x=0; x<w/sps; x++)
1533 memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
1534
1535 if (++track->audio.sub_packet_cnt >= h) {
1536 track->audio.sub_packet_cnt = 0;
1537 track->audio.pkt_cnt = h*w / a;
1538 }
1539 }
1540 while (track->audio.pkt_cnt) {
1541 pkt = av_mallocz(sizeof(AVPacket));
1542 av_new_packet(pkt, a);
1543 memcpy(pkt->data, track->audio.buf
1544 + a * (h*w / a - track->audio.pkt_cnt--), a);
1545 pkt->pos = pos;
1546 pkt->stream_index = st->index;
1547 dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
1548 }
1549 } else {
1550 MatroskaTrackEncoding *encodings = track->encodings.elem;
1551 int offset = 0, pkt_size = lace_size[n];
1552 uint8_t *pkt_data = data;
1553
1554 if (encodings && encodings->scope & 1) {
1555 offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
1556 if (offset < 0)
1557 continue;
1558 }
1559
1560 pkt = av_mallocz(sizeof(AVPacket));
1561 /* XXX: prevent data copy... */
1562 if (av_new_packet(pkt, pkt_size+offset) < 0) {
1563 av_free(pkt);
1564 res = AVERROR(ENOMEM);
1565 n = laces-1;
1566 break;
1567 }
1568 if (offset)
1569 memcpy (pkt->data, encodings->compression.settings.data, offset);
1570 memcpy (pkt->data+offset, pkt_data, pkt_size);
1571
1572 if (pkt_data != data)
1573 av_free(pkt_data);
1574
1575 if (n == 0)
1576 pkt->flags = is_keyframe;
1577 pkt->stream_index = st->index;
1578
1579 pkt->pts = timecode;
1580 pkt->pos = pos;
1581 pkt->duration = duration;
1582
1583 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
1584 }
1585
1586 if (timecode != AV_NOPTS_VALUE)
1587 timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
1588 data += lace_size[n];
1589 }
1590 }
1591
1592 av_free(lace_size);
1593 return res;
1594 }
1595
1596 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
1597 {
1598 MatroskaCluster cluster = { 0 };
1599 EbmlList *blocks_list;
1600 MatroskaBlock *blocks;
1601 int i, res;
1602 if (matroska->has_cluster_id){
1603 /* For the first cluster we parse, its ID was already read as
1604 part of matroska_read_header(), so don't read it again */
1605 res = ebml_parse_id(matroska, matroska_clusters,
1606 MATROSKA_ID_CLUSTER, &cluster);
1607 matroska->has_cluster_id = 0;
1608 } else
1609 res = ebml_parse(matroska, matroska_clusters, &cluster);
1610 blocks_list = &cluster.blocks;
1611 blocks = blocks_list->elem;
1612 for (i=0; i<blocks_list->nb_elem; i++)
1613 if (blocks[i].bin.size > 0)
1614 res=matroska_parse_block(matroska,
1615 blocks[i].bin.data, blocks[i].bin.size,
1616 blocks[i].bin.pos, cluster.timecode,
1617 blocks[i].duration, !blocks[i].reference);
1618 ebml_free(matroska_cluster, &cluster);
1619 return res;
1620 }
1621
1622 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
1623 {
1624 MatroskaDemuxContext *matroska = s->priv_data;
1625
1626 while (matroska_deliver_packet(matroska, pkt)) {
1627 if (matroska->done)
1628 return AVERROR(EIO);
1629 if (matroska_parse_cluster(matroska) < 0)
1630 matroska->done = 1;
1631 }
1632
1633 return 0;
1634 }
1635
1636 static int matroska_read_seek(AVFormatContext *s, int stream_index,
1637 int64_t timestamp, int flags)
1638 {
1639 MatroskaDemuxContext *matroska = s->priv_data;
1640 AVStream *st = s->streams[stream_index];
1641 int index;
1642
1643 index = av_index_search_timestamp(st, timestamp, flags);
1644 if (index < 0)
1645 return 0;
1646
1647 matroska_clear_queue(matroska);
1648
1649 url_fseek(s->pb, st->index_entries[index].pos, SEEK_SET);
1650 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
1651 matroska->skip_to_stream = st;
1652 av_update_cur_dts(s, st, st->index_entries[index].timestamp);
1653 return 0;
1654 }
1655
1656 static int matroska_read_close(AVFormatContext *s)
1657 {
1658 MatroskaDemuxContext *matroska = s->priv_data;
1659 MatroskaTrack *tracks = matroska->tracks.elem;
1660 int n;
1661
1662 matroska_clear_queue(matroska);
1663
1664 for (n=0; n < matroska->tracks.nb_elem; n++)
1665 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
1666 av_free(tracks[n].audio.buf);
1667 ebml_free(matroska_segment, matroska);
1668
1669 return 0;
1670 }
1671
1672 AVInputFormat matroska_demuxer = {
1673 "matroska",
1674 NULL_IF_CONFIG_SMALL("Matroska file format"),
1675 sizeof(MatroskaDemuxContext),
1676 matroska_probe,
1677 matroska_read_header,
1678 matroska_read_packet,
1679 matroska_read_close,
1680 matroska_read_seek,
1681 };