matroska: Support V_QUICKTIME as written in the specification
[libav.git] / libavformat / matroskadec.c
1 /*
2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The Libav Project
4 *
5 * This file is part of Libav.
6 *
7 * Libav 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 * Libav 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 Libav; 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
24 * Matroska file demuxer
25 * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
26 * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
27 * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28 * @see specs available on the Matroska project page: http://www.matroska.org/
29 */
30
31 #include "config.h"
32
33 #include <inttypes.h>
34 #include <stdio.h>
35 #if CONFIG_BZLIB
36 #include <bzlib.h>
37 #endif
38 #if CONFIG_ZLIB
39 #include <zlib.h>
40 #endif
41
42 #include "libavutil/avstring.h"
43 #include "libavutil/dict.h"
44 #include "libavutil/intfloat.h"
45 #include "libavutil/intreadwrite.h"
46 #include "libavutil/lzo.h"
47 #include "libavutil/mathematics.h"
48
49 #include "libavcodec/bytestream.h"
50 #include "libavcodec/flac.h"
51 #include "libavcodec/mpeg4audio.h"
52
53 #include "avformat.h"
54 #include "avio_internal.h"
55 #include "internal.h"
56 #include "isom.h"
57 #include "matroska.h"
58 #include "oggdec.h"
59 /* For ff_codec_get_id(). */
60 #include "riff.h"
61 #include "rmsipr.h"
62
63 typedef enum {
64 EBML_NONE,
65 EBML_UINT,
66 EBML_FLOAT,
67 EBML_STR,
68 EBML_UTF8,
69 EBML_BIN,
70 EBML_NEST,
71 EBML_PASS,
72 EBML_STOP,
73 EBML_TYPE_COUNT
74 } EbmlType;
75
76 typedef const struct EbmlSyntax {
77 uint32_t id;
78 EbmlType type;
79 int list_elem_size;
80 int data_offset;
81 union {
82 uint64_t u;
83 double f;
84 const char *s;
85 const struct EbmlSyntax *n;
86 } def;
87 } EbmlSyntax;
88
89 typedef struct EbmlList {
90 int nb_elem;
91 void *elem;
92 } EbmlList;
93
94 typedef struct EbmlBin {
95 int size;
96 uint8_t *data;
97 int64_t pos;
98 } EbmlBin;
99
100 typedef struct Ebml {
101 uint64_t version;
102 uint64_t max_size;
103 uint64_t id_length;
104 char *doctype;
105 uint64_t doctype_version;
106 } Ebml;
107
108 typedef struct MatroskaTrackCompression {
109 uint64_t algo;
110 EbmlBin settings;
111 } MatroskaTrackCompression;
112
113 typedef struct MatroskaTrackEncoding {
114 uint64_t scope;
115 uint64_t type;
116 MatroskaTrackCompression compression;
117 } MatroskaTrackEncoding;
118
119 typedef struct MatroskaTrackVideo {
120 double frame_rate;
121 uint64_t display_width;
122 uint64_t display_height;
123 uint64_t pixel_width;
124 uint64_t pixel_height;
125 uint64_t fourcc;
126 uint64_t stereo_mode;
127 } MatroskaTrackVideo;
128
129 typedef struct MatroskaTrackAudio {
130 double samplerate;
131 double out_samplerate;
132 uint64_t bitdepth;
133 uint64_t channels;
134
135 /* real audio header (extracted from extradata) */
136 int coded_framesize;
137 int sub_packet_h;
138 int frame_size;
139 int sub_packet_size;
140 int sub_packet_cnt;
141 int pkt_cnt;
142 uint64_t buf_timecode;
143 uint8_t *buf;
144 } MatroskaTrackAudio;
145
146 typedef struct MatroskaTrack {
147 uint64_t num;
148 uint64_t uid;
149 uint64_t type;
150 char *name;
151 char *codec_id;
152 EbmlBin codec_priv;
153 char *language;
154 double time_scale;
155 uint64_t default_duration;
156 uint64_t flag_default;
157 uint64_t flag_forced;
158 MatroskaTrackVideo video;
159 MatroskaTrackAudio audio;
160 EbmlList encodings;
161 uint64_t codec_delay;
162
163 AVStream *stream;
164 int64_t end_timecode;
165 int ms_compat;
166 } MatroskaTrack;
167
168 typedef struct MatroskaAttachment {
169 uint64_t uid;
170 char *filename;
171 char *mime;
172 EbmlBin bin;
173
174 AVStream *stream;
175 } MatroskaAttachment;
176
177 typedef struct MatroskaChapter {
178 uint64_t start;
179 uint64_t end;
180 uint64_t uid;
181 char *title;
182
183 AVChapter *chapter;
184 } MatroskaChapter;
185
186 typedef struct MatroskaIndexPos {
187 uint64_t track;
188 uint64_t pos;
189 } MatroskaIndexPos;
190
191 typedef struct MatroskaIndex {
192 uint64_t time;
193 EbmlList pos;
194 } MatroskaIndex;
195
196 typedef struct MatroskaTag {
197 char *name;
198 char *string;
199 char *lang;
200 uint64_t def;
201 EbmlList sub;
202 } MatroskaTag;
203
204 typedef struct MatroskaTagTarget {
205 char *type;
206 uint64_t typevalue;
207 uint64_t trackuid;
208 uint64_t chapteruid;
209 uint64_t attachuid;
210 } MatroskaTagTarget;
211
212 typedef struct MatroskaTags {
213 MatroskaTagTarget target;
214 EbmlList tag;
215 } MatroskaTags;
216
217 typedef struct MatroskaSeekhead {
218 uint64_t id;
219 uint64_t pos;
220 } MatroskaSeekhead;
221
222 typedef struct MatroskaLevel {
223 uint64_t start;
224 uint64_t length;
225 } MatroskaLevel;
226
227 typedef struct MatroskaCluster {
228 uint64_t timecode;
229 EbmlList blocks;
230 } MatroskaCluster;
231
232 typedef struct MatroskaDemuxContext {
233 AVFormatContext *ctx;
234
235 /* EBML stuff */
236 int num_levels;
237 MatroskaLevel levels[EBML_MAX_DEPTH];
238 int level_up;
239 uint32_t current_id;
240
241 uint64_t time_scale;
242 double duration;
243 char *title;
244 EbmlList tracks;
245 EbmlList attachments;
246 EbmlList chapters;
247 EbmlList index;
248 EbmlList tags;
249 EbmlList seekhead;
250
251 /* byte position of the segment inside the stream */
252 int64_t segment_start;
253
254 /* the packet queue */
255 AVPacket **packets;
256 int num_packets;
257 AVPacket *prev_pkt;
258
259 int done;
260
261 /* What to skip before effectively reading a packet. */
262 int skip_to_keyframe;
263 uint64_t skip_to_timecode;
264
265 /* File has a CUES element, but we defer parsing until it is needed. */
266 int cues_parsing_deferred;
267
268 int current_cluster_num_blocks;
269 int64_t current_cluster_pos;
270 MatroskaCluster current_cluster;
271
272 /* File has SSA subtitles which prevent incremental cluster parsing. */
273 int contains_ssa;
274 } MatroskaDemuxContext;
275
276 typedef struct MatroskaBlock {
277 uint64_t duration;
278 int64_t reference;
279 uint64_t non_simple;
280 EbmlBin bin;
281 } MatroskaBlock;
282
283 static EbmlSyntax ebml_header[] = {
284 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
285 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
286 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
287 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
288 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
289 { EBML_ID_EBMLVERSION, EBML_NONE },
290 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
291 { 0 }
292 };
293
294 static EbmlSyntax ebml_syntax[] = {
295 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
296 { 0 }
297 };
298
299 static EbmlSyntax matroska_info[] = {
300 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
301 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
302 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
303 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
304 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
305 { MATROSKA_ID_DATEUTC, EBML_NONE },
306 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
307 { 0 }
308 };
309
310 static EbmlSyntax matroska_track_video[] = {
311 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
312 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width) },
313 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height) },
314 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
315 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
316 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, fourcc) },
317 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
318 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
319 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
320 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
321 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
322 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_NONE },
323 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
324 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
325 { 0 }
326 };
327
328 static EbmlSyntax matroska_track_audio[] = {
329 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
330 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
331 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
332 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
333 { 0 }
334 };
335
336 static EbmlSyntax matroska_track_encoding_compression[] = {
337 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
338 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
339 { 0 }
340 };
341
342 static EbmlSyntax matroska_track_encoding[] = {
343 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
344 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
345 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
346 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
347 { 0 }
348 };
349
350 static EbmlSyntax matroska_track_encodings[] = {
351 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
352 { 0 }
353 };
354
355 static EbmlSyntax matroska_track[] = {
356 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
357 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
358 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
359 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
360 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
361 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
362 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
363 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
364 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
365 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
366 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
367 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
368 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
369 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
370 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
371 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
372 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
373 { MATROSKA_ID_CODECNAME, EBML_NONE },
374 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
375 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
376 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
377 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
378 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
379 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
380 { 0 }
381 };
382
383 static EbmlSyntax matroska_tracks[] = {
384 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
385 { 0 }
386 };
387
388 static EbmlSyntax matroska_attachment[] = {
389 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
390 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
391 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
392 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
393 { MATROSKA_ID_FILEDESC, EBML_NONE },
394 { 0 }
395 };
396
397 static EbmlSyntax matroska_attachments[] = {
398 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
399 { 0 }
400 };
401
402 static EbmlSyntax matroska_chapter_display[] = {
403 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
404 { MATROSKA_ID_CHAPLANG, EBML_NONE },
405 { 0 }
406 };
407
408 static EbmlSyntax matroska_chapter_entry[] = {
409 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
410 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
411 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
412 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
413 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
414 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
415 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
416 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
417 { 0 }
418 };
419
420 static EbmlSyntax matroska_chapter[] = {
421 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
422 { MATROSKA_ID_EDITIONUID, EBML_NONE },
423 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
424 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
425 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
426 { 0 }
427 };
428
429 static EbmlSyntax matroska_chapters[] = {
430 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
431 { 0 }
432 };
433
434 static EbmlSyntax matroska_index_pos[] = {
435 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
436 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
437 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
438 { 0 }
439 };
440
441 static EbmlSyntax matroska_index_entry[] = {
442 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
443 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
444 { 0 }
445 };
446
447 static EbmlSyntax matroska_index[] = {
448 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
449 { 0 }
450 };
451
452 static EbmlSyntax matroska_simpletag[] = {
453 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
454 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
455 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
456 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
457 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
458 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
459 { 0 }
460 };
461
462 static EbmlSyntax matroska_tagtargets[] = {
463 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
464 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
465 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
466 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
467 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
468 { 0 }
469 };
470
471 static EbmlSyntax matroska_tag[] = {
472 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
473 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
474 { 0 }
475 };
476
477 static EbmlSyntax matroska_tags[] = {
478 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
479 { 0 }
480 };
481
482 static EbmlSyntax matroska_seekhead_entry[] = {
483 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
484 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
485 { 0 }
486 };
487
488 static EbmlSyntax matroska_seekhead[] = {
489 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
490 { 0 }
491 };
492
493 static EbmlSyntax matroska_segment[] = {
494 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, { .n = matroska_info } },
495 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, { .n = matroska_tracks } },
496 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, { .n = matroska_attachments } },
497 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, { .n = matroska_chapters } },
498 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, { .n = matroska_index } },
499 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, { .n = matroska_tags } },
500 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, { .n = matroska_seekhead } },
501 { MATROSKA_ID_CLUSTER, EBML_STOP },
502 { 0 }
503 };
504
505 static EbmlSyntax matroska_segments[] = {
506 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
507 { 0 }
508 };
509
510 static EbmlSyntax matroska_blockgroup[] = {
511 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
512 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
513 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration), { .u = AV_NOPTS_VALUE } },
514 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock, reference) },
515 { MATROSKA_ID_CODECSTATE, EBML_NONE },
516 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
517 { 0 }
518 };
519
520 static EbmlSyntax matroska_cluster[] = {
521 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
522 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
523 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
524 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
525 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
526 { 0 }
527 };
528
529 static EbmlSyntax matroska_clusters[] = {
530 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster } },
531 { MATROSKA_ID_INFO, EBML_NONE },
532 { MATROSKA_ID_CUES, EBML_NONE },
533 { MATROSKA_ID_TAGS, EBML_NONE },
534 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
535 { 0 }
536 };
537
538 static EbmlSyntax matroska_cluster_incremental_parsing[] = {
539 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
540 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
541 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
542 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
543 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
544 { MATROSKA_ID_INFO, EBML_NONE },
545 { MATROSKA_ID_CUES, EBML_NONE },
546 { MATROSKA_ID_TAGS, EBML_NONE },
547 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
548 { MATROSKA_ID_CLUSTER, EBML_STOP },
549 { 0 }
550 };
551
552 static EbmlSyntax matroska_cluster_incremental[] = {
553 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
554 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
555 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
556 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
557 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
558 { 0 }
559 };
560
561 static EbmlSyntax matroska_clusters_incremental[] = {
562 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
563 { MATROSKA_ID_INFO, EBML_NONE },
564 { MATROSKA_ID_CUES, EBML_NONE },
565 { MATROSKA_ID_TAGS, EBML_NONE },
566 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
567 { 0 }
568 };
569
570 static const char *const matroska_doctypes[] = { "matroska", "webm" };
571
572 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
573 {
574 AVIOContext *pb = matroska->ctx->pb;
575 uint32_t id;
576 matroska->current_id = 0;
577 matroska->num_levels = 0;
578
579 /* seek to next position to resync from */
580 if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
581 goto eof;
582
583 id = avio_rb32(pb);
584
585 // try to find a toplevel element
586 while (!pb->eof_reached) {
587 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
588 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
589 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
590 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
591 matroska->current_id = id;
592 return 0;
593 }
594 id = (id << 8) | avio_r8(pb);
595 }
596
597 eof:
598 matroska->done = 1;
599 return AVERROR_EOF;
600 }
601
602 /*
603 * Return: Whether we reached the end of a level in the hierarchy or not.
604 */
605 static int ebml_level_end(MatroskaDemuxContext *matroska)
606 {
607 AVIOContext *pb = matroska->ctx->pb;
608 int64_t pos = avio_tell(pb);
609
610 if (matroska->num_levels > 0) {
611 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
612 if (pos - level->start >= level->length || matroska->current_id) {
613 matroska->num_levels--;
614 return 1;
615 }
616 }
617 return 0;
618 }
619
620 /*
621 * Read: an "EBML number", which is defined as a variable-length
622 * array of bytes. The first byte indicates the length by giving a
623 * number of 0-bits followed by a one. The position of the first
624 * "one" bit inside the first byte indicates the length of this
625 * number.
626 * Returns: number of bytes read, < 0 on error
627 */
628 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
629 int max_size, uint64_t *number)
630 {
631 int read = 1, n = 1;
632 uint64_t total = 0;
633
634 /* The first byte tells us the length in bytes - avio_r8() can normally
635 * return 0, but since that's not a valid first ebmlID byte, we can
636 * use it safely here to catch EOS. */
637 if (!(total = avio_r8(pb))) {
638 /* we might encounter EOS here */
639 if (!pb->eof_reached) {
640 int64_t pos = avio_tell(pb);
641 av_log(matroska->ctx, AV_LOG_ERROR,
642 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
643 pos, pos);
644 return pb->error ? pb->error : AVERROR(EIO);
645 }
646 return AVERROR_EOF;
647 }
648
649 /* get the length of the EBML number */
650 read = 8 - ff_log2_tab[total];
651 if (read > max_size) {
652 int64_t pos = avio_tell(pb) - 1;
653 av_log(matroska->ctx, AV_LOG_ERROR,
654 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
655 (uint8_t) total, pos, pos);
656 return AVERROR_INVALIDDATA;
657 }
658
659 /* read out length */
660 total ^= 1 << ff_log2_tab[total];
661 while (n++ < read)
662 total = (total << 8) | avio_r8(pb);
663
664 *number = total;
665
666 return read;
667 }
668
669 /**
670 * Read a EBML length value.
671 * This needs special handling for the "unknown length" case which has multiple
672 * encodings.
673 */
674 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
675 uint64_t *number)
676 {
677 int res = ebml_read_num(matroska, pb, 8, number);
678 if (res > 0 && *number + 1 == 1ULL << (7 * res))
679 *number = 0xffffffffffffffULL;
680 return res;
681 }
682
683 /*
684 * Read the next element as an unsigned int.
685 * 0 is success, < 0 is failure.
686 */
687 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
688 {
689 int n = 0;
690
691 if (size > 8)
692 return AVERROR_INVALIDDATA;
693
694 /* big-endian ordering; build up number */
695 *num = 0;
696 while (n++ < size)
697 *num = (*num << 8) | avio_r8(pb);
698
699 return 0;
700 }
701
702 /*
703 * Read the next element as a float.
704 * 0 is success, < 0 is failure.
705 */
706 static int ebml_read_float(AVIOContext *pb, int size, double *num)
707 {
708 if (size == 0)
709 *num = 0;
710 else if (size == 4)
711 *num = av_int2float(avio_rb32(pb));
712 else if (size == 8)
713 *num = av_int2double(avio_rb64(pb));
714 else
715 return AVERROR_INVALIDDATA;
716
717 return 0;
718 }
719
720 /*
721 * Read the next element as an ASCII string.
722 * 0 is success, < 0 is failure.
723 */
724 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
725 {
726 char *res;
727
728 /* EBML strings are usually not 0-terminated, so we allocate one
729 * byte more, read the string and NULL-terminate it ourselves. */
730 if (!(res = av_malloc(size + 1)))
731 return AVERROR(ENOMEM);
732 if (avio_read(pb, (uint8_t *) res, size) != size) {
733 av_free(res);
734 return AVERROR(EIO);
735 }
736 (res)[size] = '\0';
737 av_free(*str);
738 *str = res;
739
740 return 0;
741 }
742
743 /*
744 * Read the next element as binary data.
745 * 0 is success, < 0 is failure.
746 */
747 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
748 {
749 av_free(bin->data);
750 if (!(bin->data = av_mallocz(length + AV_INPUT_BUFFER_PADDING_SIZE)))
751 return AVERROR(ENOMEM);
752
753 bin->size = length;
754 bin->pos = avio_tell(pb);
755 if (avio_read(pb, bin->data, length) != length) {
756 av_freep(&bin->data);
757 return AVERROR(EIO);
758 }
759
760 return 0;
761 }
762
763 /*
764 * Read the next element, but only the header. The contents
765 * are supposed to be sub-elements which can be read separately.
766 * 0 is success, < 0 is failure.
767 */
768 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
769 {
770 AVIOContext *pb = matroska->ctx->pb;
771 MatroskaLevel *level;
772
773 if (matroska->num_levels >= EBML_MAX_DEPTH) {
774 av_log(matroska->ctx, AV_LOG_ERROR,
775 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
776 return AVERROR(ENOSYS);
777 }
778
779 level = &matroska->levels[matroska->num_levels++];
780 level->start = avio_tell(pb);
781 level->length = length;
782
783 return 0;
784 }
785
786 /*
787 * Read signed/unsigned "EBML" numbers.
788 * Return: number of bytes processed, < 0 on error
789 */
790 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
791 uint8_t *data, uint32_t size, uint64_t *num)
792 {
793 AVIOContext pb;
794 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
795 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
796 }
797
798 /*
799 * Same as above, but signed.
800 */
801 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
802 uint8_t *data, uint32_t size, int64_t *num)
803 {
804 uint64_t unum;
805 int res;
806
807 /* read as unsigned number first */
808 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
809 return res;
810
811 /* make signed (weird way) */
812 *num = unum - ((1LL << (7 * res - 1)) - 1);
813
814 return res;
815 }
816
817 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
818 EbmlSyntax *syntax, void *data);
819
820 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
821 uint32_t id, void *data)
822 {
823 int i;
824 for (i = 0; syntax[i].id; i++)
825 if (id == syntax[i].id)
826 break;
827 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
828 matroska->num_levels > 0 &&
829 matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
830 return 0; // we reached the end of an unknown size cluster
831 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
832 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%"PRIX32"\n", id);
833 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
834 return AVERROR_INVALIDDATA;
835 }
836 return ebml_parse_elem(matroska, &syntax[i], data);
837 }
838
839 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
840 void *data)
841 {
842 if (!matroska->current_id) {
843 uint64_t id;
844 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
845 if (res < 0)
846 return res;
847 matroska->current_id = id | 1 << 7 * res;
848 }
849 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
850 }
851
852 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
853 void *data)
854 {
855 int i, res = 0;
856
857 for (i = 0; syntax[i].id; i++)
858 switch (syntax[i].type) {
859 case EBML_UINT:
860 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
861 break;
862 case EBML_FLOAT:
863 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
864 break;
865 case EBML_STR:
866 case EBML_UTF8:
867 // the default may be NULL
868 if (syntax[i].def.s) {
869 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
870 *dst = av_strdup(syntax[i].def.s);
871 if (!*dst)
872 return AVERROR(ENOMEM);
873 }
874 break;
875 }
876
877 while (!res && !ebml_level_end(matroska))
878 res = ebml_parse(matroska, syntax, data);
879
880 return res;
881 }
882
883 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
884 EbmlSyntax *syntax, void *data)
885 {
886 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
887 [EBML_UINT] = 8,
888 [EBML_FLOAT] = 8,
889 // max. 16 MB for strings
890 [EBML_STR] = 0x1000000,
891 [EBML_UTF8] = 0x1000000,
892 // max. 256 MB for binary data
893 [EBML_BIN] = 0x10000000,
894 // no limits for anything else
895 };
896 AVIOContext *pb = matroska->ctx->pb;
897 uint32_t id = syntax->id;
898 uint64_t length;
899 int res;
900
901 data = (char *) data + syntax->data_offset;
902 if (syntax->list_elem_size) {
903 EbmlList *list = data;
904 if ((res = av_reallocp_array(&list->elem,
905 list->nb_elem + 1,
906 syntax->list_elem_size)) < 0) {
907 list->nb_elem = 0;
908 return res;
909 }
910 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
911 memset(data, 0, syntax->list_elem_size);
912 list->nb_elem++;
913 }
914
915 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
916 matroska->current_id = 0;
917 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
918 return res;
919 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
920 av_log(matroska->ctx, AV_LOG_ERROR,
921 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
922 length, max_lengths[syntax->type], syntax->type);
923 return AVERROR_INVALIDDATA;
924 }
925 }
926
927 switch (syntax->type) {
928 case EBML_UINT:
929 res = ebml_read_uint(pb, length, data);
930 break;
931 case EBML_FLOAT:
932 res = ebml_read_float(pb, length, data);
933 break;
934 case EBML_STR:
935 case EBML_UTF8:
936 res = ebml_read_ascii(pb, length, data);
937 break;
938 case EBML_BIN:
939 res = ebml_read_binary(pb, length, data);
940 break;
941 case EBML_NEST:
942 if ((res = ebml_read_master(matroska, length)) < 0)
943 return res;
944 if (id == MATROSKA_ID_SEGMENT)
945 matroska->segment_start = avio_tell(matroska->ctx->pb);
946 return ebml_parse_nest(matroska, syntax->def.n, data);
947 case EBML_PASS:
948 return ebml_parse_id(matroska, syntax->def.n, id, data);
949 case EBML_STOP:
950 return 1;
951 default:
952 return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
953 }
954 if (res == AVERROR_INVALIDDATA)
955 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
956 else if (res == AVERROR(EIO))
957 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
958 return res;
959 }
960
961 static void ebml_free(EbmlSyntax *syntax, void *data)
962 {
963 int i, j;
964 for (i = 0; syntax[i].id; i++) {
965 void *data_off = (char *) data + syntax[i].data_offset;
966 switch (syntax[i].type) {
967 case EBML_STR:
968 case EBML_UTF8:
969 av_freep(data_off);
970 break;
971 case EBML_BIN:
972 av_freep(&((EbmlBin *) data_off)->data);
973 break;
974 case EBML_NEST:
975 if (syntax[i].list_elem_size) {
976 EbmlList *list = data_off;
977 char *ptr = list->elem;
978 for (j = 0; j < list->nb_elem;
979 j++, ptr += syntax[i].list_elem_size)
980 ebml_free(syntax[i].def.n, ptr);
981 av_free(list->elem);
982 } else
983 ebml_free(syntax[i].def.n, data_off);
984 default:
985 break;
986 }
987 }
988 }
989
990 /*
991 * Autodetecting...
992 */
993 static int matroska_probe(AVProbeData *p)
994 {
995 uint64_t total = 0;
996 int len_mask = 0x80, size = 1, n = 1, i;
997
998 /* EBML header? */
999 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1000 return 0;
1001
1002 /* length of header */
1003 total = p->buf[4];
1004 while (size <= 8 && !(total & len_mask)) {
1005 size++;
1006 len_mask >>= 1;
1007 }
1008 if (size > 8)
1009 return 0;
1010 total &= (len_mask - 1);
1011 while (n < size)
1012 total = (total << 8) | p->buf[4 + n++];
1013
1014 /* Does the probe data contain the whole header? */
1015 if (p->buf_size < 4 + size + total)
1016 return 0;
1017
1018 /* The header should contain a known document type. For now,
1019 * we don't parse the whole header but simply check for the
1020 * availability of that array of characters inside the header.
1021 * Not fully fool-proof, but good enough. */
1022 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1023 int probelen = strlen(matroska_doctypes[i]);
1024 if (total < probelen)
1025 continue;
1026 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1027 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1028 return AVPROBE_SCORE_MAX;
1029 }
1030
1031 // probably valid EBML header but no recognized doctype
1032 return AVPROBE_SCORE_EXTENSION;
1033 }
1034
1035 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1036 int num)
1037 {
1038 MatroskaTrack *tracks = matroska->tracks.elem;
1039 int i;
1040
1041 for (i = 0; i < matroska->tracks.nb_elem; i++)
1042 if (tracks[i].num == num)
1043 return &tracks[i];
1044
1045 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1046 return NULL;
1047 }
1048
1049 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1050 MatroskaTrack *track)
1051 {
1052 MatroskaTrackEncoding *encodings = track->encodings.elem;
1053 uint8_t *data = *buf;
1054 int isize = *buf_size;
1055 uint8_t *pkt_data = NULL;
1056 uint8_t av_unused *newpktdata;
1057 int pkt_size = isize;
1058 int result = 0;
1059 int olen;
1060
1061 if (pkt_size >= 10000000)
1062 return AVERROR_INVALIDDATA;
1063
1064 switch (encodings[0].compression.algo) {
1065 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1066 {
1067 int header_size = encodings[0].compression.settings.size;
1068 uint8_t *header = encodings[0].compression.settings.data;
1069
1070 if (!header_size)
1071 return 0;
1072
1073 pkt_size = isize + header_size;
1074 pkt_data = av_malloc(pkt_size);
1075 if (!pkt_data)
1076 return AVERROR(ENOMEM);
1077
1078 memcpy(pkt_data, header, header_size);
1079 memcpy(pkt_data + header_size, data, isize);
1080 break;
1081 }
1082 #if CONFIG_LZO
1083 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1084 do {
1085 olen = pkt_size *= 3;
1086 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1087 if (!newpktdata) {
1088 result = AVERROR(ENOMEM);
1089 goto failed;
1090 }
1091 pkt_data = newpktdata;
1092 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1093 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1094 if (result) {
1095 result = AVERROR_INVALIDDATA;
1096 goto failed;
1097 }
1098 pkt_size -= olen;
1099 break;
1100 #endif
1101 #if CONFIG_ZLIB
1102 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1103 {
1104 z_stream zstream = { 0 };
1105 if (inflateInit(&zstream) != Z_OK)
1106 return -1;
1107 zstream.next_in = data;
1108 zstream.avail_in = isize;
1109 do {
1110 pkt_size *= 3;
1111 newpktdata = av_realloc(pkt_data, pkt_size);
1112 if (!newpktdata) {
1113 inflateEnd(&zstream);
1114 goto failed;
1115 }
1116 pkt_data = newpktdata;
1117 zstream.avail_out = pkt_size - zstream.total_out;
1118 zstream.next_out = pkt_data + zstream.total_out;
1119 result = inflate(&zstream, Z_NO_FLUSH);
1120 } while (result == Z_OK && pkt_size < 10000000);
1121 pkt_size = zstream.total_out;
1122 inflateEnd(&zstream);
1123 if (result != Z_STREAM_END) {
1124 if (result == Z_MEM_ERROR)
1125 result = AVERROR(ENOMEM);
1126 else
1127 result = AVERROR_INVALIDDATA;
1128 goto failed;
1129 }
1130 break;
1131 }
1132 #endif
1133 #if CONFIG_BZLIB
1134 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1135 {
1136 bz_stream bzstream = { 0 };
1137 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1138 return -1;
1139 bzstream.next_in = data;
1140 bzstream.avail_in = isize;
1141 do {
1142 pkt_size *= 3;
1143 newpktdata = av_realloc(pkt_data, pkt_size);
1144 if (!newpktdata) {
1145 BZ2_bzDecompressEnd(&bzstream);
1146 goto failed;
1147 }
1148 pkt_data = newpktdata;
1149 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1150 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1151 result = BZ2_bzDecompress(&bzstream);
1152 } while (result == BZ_OK && pkt_size < 10000000);
1153 pkt_size = bzstream.total_out_lo32;
1154 BZ2_bzDecompressEnd(&bzstream);
1155 if (result != BZ_STREAM_END) {
1156 if (result == BZ_MEM_ERROR)
1157 result = AVERROR(ENOMEM);
1158 else
1159 result = AVERROR_INVALIDDATA;
1160 goto failed;
1161 }
1162 break;
1163 }
1164 #endif
1165 default:
1166 return AVERROR_INVALIDDATA;
1167 }
1168
1169 *buf = pkt_data;
1170 *buf_size = pkt_size;
1171 return 0;
1172
1173 failed:
1174 av_free(pkt_data);
1175 return result;
1176 }
1177
1178 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
1179 AVPacket *pkt, uint64_t display_duration)
1180 {
1181 AVBufferRef *line;
1182 char *layer, *ptr = pkt->data, *end = ptr + pkt->size;
1183
1184 for (; *ptr != ',' && ptr < end - 1; ptr++)
1185 ;
1186 if (*ptr == ',')
1187 layer = ++ptr;
1188 for (; *ptr != ',' && ptr < end - 1; ptr++)
1189 ;
1190 if (*ptr == ',') {
1191 int64_t end_pts = pkt->pts + display_duration;
1192 int sc = matroska->time_scale * pkt->pts / 10000000;
1193 int ec = matroska->time_scale * end_pts / 10000000;
1194 int sh, sm, ss, eh, em, es, len;
1195 sh = sc / 360000;
1196 sc -= 360000 * sh;
1197 sm = sc / 6000;
1198 sc -= 6000 * sm;
1199 ss = sc / 100;
1200 sc -= 100 * ss;
1201 eh = ec / 360000;
1202 ec -= 360000 * eh;
1203 em = ec / 6000;
1204 ec -= 6000 * em;
1205 es = ec / 100;
1206 ec -= 100 * es;
1207 *ptr++ = '\0';
1208 len = 50 + end - ptr + AV_INPUT_BUFFER_PADDING_SIZE;
1209 if (!(line = av_buffer_alloc(len)))
1210 return;
1211 snprintf(line->data, len,
1212 "Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
1213 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
1214 av_buffer_unref(&pkt->buf);
1215 pkt->buf = line;
1216 pkt->data = line->data;
1217 pkt->size = strlen(line->data);
1218 }
1219 }
1220
1221 static int matroska_merge_packets(AVPacket *out, AVPacket *in)
1222 {
1223 int old_size = out->size;
1224 int ret = av_grow_packet(out, in->size);
1225 if (ret < 0)
1226 return ret;
1227
1228 memcpy(out->data + old_size, in->data, in->size);
1229
1230 av_packet_unref(in);
1231 av_free(in);
1232 return 0;
1233 }
1234
1235 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1236 AVDictionary **metadata, char *prefix)
1237 {
1238 MatroskaTag *tags = list->elem;
1239 char key[1024];
1240 int i;
1241
1242 for (i = 0; i < list->nb_elem; i++) {
1243 const char *lang = tags[i].lang &&
1244 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1245
1246 if (!tags[i].name) {
1247 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1248 continue;
1249 }
1250 if (prefix)
1251 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1252 else
1253 av_strlcpy(key, tags[i].name, sizeof(key));
1254 if (tags[i].def || !lang) {
1255 av_dict_set(metadata, key, tags[i].string, 0);
1256 if (tags[i].sub.nb_elem)
1257 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1258 }
1259 if (lang) {
1260 av_strlcat(key, "-", sizeof(key));
1261 av_strlcat(key, lang, sizeof(key));
1262 av_dict_set(metadata, key, tags[i].string, 0);
1263 if (tags[i].sub.nb_elem)
1264 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1265 }
1266 }
1267 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1268 }
1269
1270 static void matroska_convert_tags(AVFormatContext *s)
1271 {
1272 MatroskaDemuxContext *matroska = s->priv_data;
1273 MatroskaTags *tags = matroska->tags.elem;
1274 int i, j;
1275
1276 for (i = 0; i < matroska->tags.nb_elem; i++) {
1277 if (tags[i].target.attachuid) {
1278 MatroskaAttachment *attachment = matroska->attachments.elem;
1279 int found = 0;
1280 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1281 if (attachment[j].uid == tags[i].target.attachuid &&
1282 attachment[j].stream) {
1283 matroska_convert_tag(s, &tags[i].tag,
1284 &attachment[j].stream->metadata, NULL);
1285 found = 1;
1286 }
1287 }
1288 if (!found) {
1289 av_log(NULL, AV_LOG_WARNING,
1290 "The tags at index %d refer to a "
1291 "non-existent attachment %"PRId64".\n",
1292 i, tags[i].target.attachuid);
1293 }
1294 } else if (tags[i].target.chapteruid) {
1295 MatroskaChapter *chapter = matroska->chapters.elem;
1296 int found = 0;
1297 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1298 if (chapter[j].uid == tags[i].target.chapteruid &&
1299 chapter[j].chapter) {
1300 matroska_convert_tag(s, &tags[i].tag,
1301 &chapter[j].chapter->metadata, NULL);
1302 found = 1;
1303 }
1304 }
1305 if (!found) {
1306 av_log(NULL, AV_LOG_WARNING,
1307 "The tags at index %d refer to a non-existent chapter "
1308 "%"PRId64".\n",
1309 i, tags[i].target.chapteruid);
1310 }
1311 } else if (tags[i].target.trackuid) {
1312 MatroskaTrack *track = matroska->tracks.elem;
1313 int found = 0;
1314 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1315 if (track[j].uid == tags[i].target.trackuid &&
1316 track[j].stream) {
1317 matroska_convert_tag(s, &tags[i].tag,
1318 &track[j].stream->metadata, NULL);
1319 found = 1;
1320 }
1321 }
1322 if (!found) {
1323 av_log(NULL, AV_LOG_WARNING,
1324 "The tags at index %d refer to a non-existent track "
1325 "%"PRId64".\n",
1326 i, tags[i].target.trackuid);
1327 }
1328 } else {
1329 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1330 tags[i].target.type);
1331 }
1332 }
1333 }
1334
1335 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1336 int idx)
1337 {
1338 EbmlList *seekhead_list = &matroska->seekhead;
1339 uint32_t level_up = matroska->level_up;
1340 uint32_t saved_id = matroska->current_id;
1341 MatroskaSeekhead *seekhead = seekhead_list->elem;
1342 int64_t before_pos = avio_tell(matroska->ctx->pb);
1343 MatroskaLevel level;
1344 int64_t offset;
1345 int ret = 0;
1346
1347 if (idx >= seekhead_list->nb_elem ||
1348 seekhead[idx].id == MATROSKA_ID_SEEKHEAD ||
1349 seekhead[idx].id == MATROSKA_ID_CLUSTER)
1350 return 0;
1351
1352 /* seek */
1353 offset = seekhead[idx].pos + matroska->segment_start;
1354 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1355 /* We don't want to lose our seekhead level, so we add
1356 * a dummy. This is a crude hack. */
1357 if (matroska->num_levels == EBML_MAX_DEPTH) {
1358 av_log(matroska->ctx, AV_LOG_INFO,
1359 "Max EBML element depth (%d) reached, "
1360 "cannot parse further.\n", EBML_MAX_DEPTH);
1361 ret = AVERROR_INVALIDDATA;
1362 } else {
1363 level.start = 0;
1364 level.length = (uint64_t) -1;
1365 matroska->levels[matroska->num_levels] = level;
1366 matroska->num_levels++;
1367 matroska->current_id = 0;
1368
1369 ret = ebml_parse(matroska, matroska_segment, matroska);
1370
1371 /* remove dummy level */
1372 while (matroska->num_levels) {
1373 uint64_t length = matroska->levels[--matroska->num_levels].length;
1374 if (length == (uint64_t) -1)
1375 break;
1376 }
1377 }
1378 }
1379 /* seek back */
1380 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1381 matroska->level_up = level_up;
1382 matroska->current_id = saved_id;
1383
1384 return ret;
1385 }
1386
1387 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1388 {
1389 EbmlList *seekhead_list = &matroska->seekhead;
1390 int64_t before_pos = avio_tell(matroska->ctx->pb);
1391 int i;
1392
1393 // we should not do any seeking in the streaming case
1394 if (!matroska->ctx->pb->seekable ||
1395 (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1396 return;
1397
1398 for (i = 0; i < seekhead_list->nb_elem; i++) {
1399 MatroskaSeekhead *seekhead = seekhead_list->elem;
1400 if (seekhead[i].pos <= before_pos)
1401 continue;
1402
1403 // defer cues parsing until we actually need cue data.
1404 if (seekhead[i].id == MATROSKA_ID_CUES) {
1405 matroska->cues_parsing_deferred = 1;
1406 continue;
1407 }
1408
1409 if (matroska_parse_seekhead_entry(matroska, i) < 0)
1410 break;
1411 }
1412 }
1413
1414 static void matroska_parse_cues(MatroskaDemuxContext *matroska)
1415 {
1416 EbmlList *seekhead_list = &matroska->seekhead;
1417 MatroskaSeekhead *seekhead = seekhead_list->elem;
1418 EbmlList *index_list;
1419 MatroskaIndex *index;
1420 int index_scale = 1;
1421 int i, j;
1422
1423 for (i = 0; i < seekhead_list->nb_elem; i++)
1424 if (seekhead[i].id == MATROSKA_ID_CUES)
1425 break;
1426 assert(i <= seekhead_list->nb_elem);
1427
1428 matroska_parse_seekhead_entry(matroska, i);
1429
1430 index_list = &matroska->index;
1431 index = index_list->elem;
1432 if (index_list->nb_elem &&
1433 index[0].time > 1E14 / matroska->time_scale) {
1434 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1435 index_scale = matroska->time_scale;
1436 }
1437 for (i = 0; i < index_list->nb_elem; i++) {
1438 EbmlList *pos_list = &index[i].pos;
1439 MatroskaIndexPos *pos = pos_list->elem;
1440 for (j = 0; j < pos_list->nb_elem; j++) {
1441 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1442 pos[j].track);
1443 if (track && track->stream)
1444 av_add_index_entry(track->stream,
1445 pos[j].pos + matroska->segment_start,
1446 index[i].time / index_scale, 0, 0,
1447 AVINDEX_KEYFRAME);
1448 }
1449 }
1450 }
1451
1452 static int matroska_aac_profile(char *codec_id)
1453 {
1454 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1455 int profile;
1456
1457 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1458 if (strstr(codec_id, aac_profiles[profile]))
1459 break;
1460 return profile + 1;
1461 }
1462
1463 static int matroska_aac_sri(int samplerate)
1464 {
1465 int sri;
1466
1467 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1468 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1469 break;
1470 return sri;
1471 }
1472
1473 static int matroska_parse_flac(AVFormatContext *s,
1474 MatroskaTrack *track,
1475 int *offset)
1476 {
1477 AVStream *st = track->stream;
1478 uint8_t *p = track->codec_priv.data;
1479 int size = track->codec_priv.size;
1480
1481 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1482 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1483 track->codec_priv.size = 0;
1484 return 0;
1485 }
1486 *offset = 8;
1487 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1488
1489 p += track->codec_priv.size;
1490 size -= track->codec_priv.size;
1491
1492 /* parse the remaining metadata blocks if present */
1493 while (size >= 4) {
1494 int block_last, block_type, block_size;
1495
1496 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1497
1498 p += 4;
1499 size -= 4;
1500 if (block_size > size)
1501 return 0;
1502
1503 /* check for the channel mask */
1504 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1505 AVDictionary *dict = NULL;
1506 AVDictionaryEntry *chmask;
1507
1508 ff_vorbis_comment(s, &dict, p, block_size, 0);
1509 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1510 if (chmask) {
1511 uint64_t mask = strtol(chmask->value, NULL, 0);
1512 if (!mask || mask & ~0x3ffffULL) {
1513 av_log(s, AV_LOG_WARNING,
1514 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1515 } else
1516 st->codecpar->channel_layout = mask;
1517 }
1518 av_dict_free(&dict);
1519 }
1520
1521 p += block_size;
1522 size -= block_size;
1523 }
1524
1525 return 0;
1526 }
1527
1528 static void mkv_stereo_mode_display_mul(int stereo_mode,
1529 int *h_width, int *h_height)
1530 {
1531 switch (stereo_mode) {
1532 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
1533 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
1534 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
1535 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
1536 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
1537 break;
1538 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
1539 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
1540 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
1541 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
1542 *h_width = 2;
1543 break;
1544 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
1545 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
1546 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
1547 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
1548 *h_height = 2;
1549 break;
1550 }
1551 }
1552
1553 static int matroska_parse_tracks(AVFormatContext *s)
1554 {
1555 MatroskaDemuxContext *matroska = s->priv_data;
1556 MatroskaTrack *tracks = matroska->tracks.elem;
1557 AVStream *st;
1558 int i, j, ret;
1559
1560 for (i = 0; i < matroska->tracks.nb_elem; i++) {
1561 MatroskaTrack *track = &tracks[i];
1562 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1563 EbmlList *encodings_list = &track->encodings;
1564 MatroskaTrackEncoding *encodings = encodings_list->elem;
1565 uint8_t *extradata = NULL;
1566 int extradata_size = 0;
1567 int extradata_offset = 0;
1568 AVIOContext b;
1569
1570 /* Apply some sanity checks. */
1571 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1572 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1573 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1574 av_log(matroska->ctx, AV_LOG_INFO,
1575 "Unknown or unsupported track type %"PRIu64"\n",
1576 track->type);
1577 continue;
1578 }
1579 if (!track->codec_id)
1580 continue;
1581
1582 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1583 if (!track->default_duration && track->video.frame_rate > 0)
1584 track->default_duration = 1000000000 / track->video.frame_rate;
1585 if (!track->video.display_width)
1586 track->video.display_width = track->video.pixel_width;
1587 if (!track->video.display_height)
1588 track->video.display_height = track->video.pixel_height;
1589 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1590 if (!track->audio.out_samplerate)
1591 track->audio.out_samplerate = track->audio.samplerate;
1592 }
1593 if (encodings_list->nb_elem > 1) {
1594 av_log(matroska->ctx, AV_LOG_ERROR,
1595 "Multiple combined encodings not supported");
1596 } else if (encodings_list->nb_elem == 1) {
1597 if (encodings[0].type ||
1598 (
1599 #if CONFIG_ZLIB
1600 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1601 #endif
1602 #if CONFIG_BZLIB
1603 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1604 #endif
1605 #if CONFIG_LZO
1606 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
1607 #endif
1608 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP)) {
1609 encodings[0].scope = 0;
1610 av_log(matroska->ctx, AV_LOG_ERROR,
1611 "Unsupported encoding type");
1612 } else if (track->codec_priv.size && encodings[0].scope & 2) {
1613 uint8_t *codec_priv = track->codec_priv.data;
1614 int ret = matroska_decode_buffer(&track->codec_priv.data,
1615 &track->codec_priv.size,
1616 track);
1617 if (ret < 0) {
1618 track->codec_priv.data = NULL;
1619 track->codec_priv.size = 0;
1620 av_log(matroska->ctx, AV_LOG_ERROR,
1621 "Failed to decode codec private data\n");
1622 }
1623
1624 if (codec_priv != track->codec_priv.data)
1625 av_free(codec_priv);
1626 }
1627 }
1628
1629 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
1630 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1631 strlen(ff_mkv_codec_tags[j].str))) {
1632 codec_id = ff_mkv_codec_tags[j].id;
1633 break;
1634 }
1635 }
1636
1637 st = track->stream = avformat_new_stream(s, NULL);
1638 if (!st)
1639 return AVERROR(ENOMEM);
1640
1641 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
1642 track->codec_priv.size >= 40 &&
1643 track->codec_priv.data) {
1644 track->ms_compat = 1;
1645 track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1646 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
1647 track->video.fourcc);
1648 extradata_offset = 40;
1649 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
1650 track->codec_priv.size >= 14 &&
1651 track->codec_priv.data) {
1652 int ret;
1653 ffio_init_context(&b, track->codec_priv.data,
1654 track->codec_priv.size,
1655 0, NULL, NULL, NULL, NULL);
1656 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size);
1657 if (ret < 0)
1658 return ret;
1659 codec_id = st->codecpar->codec_id;
1660 extradata_offset = FFMIN(track->codec_priv.size, 18);
1661 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
1662 (track->codec_priv.size >= 86) &&
1663 (track->codec_priv.data)) {
1664 if (track->codec_priv.size == AV_RB32(track->codec_priv.data)) {
1665 track->video.fourcc = AV_RL32(track->codec_priv.data + 4);
1666 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
1667 track->video.fourcc);
1668 }
1669 if (codec_id == AV_CODEC_ID_NONE) {
1670 track->video.fourcc = AV_RL32(track->codec_priv.data);
1671 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
1672 track->video.fourcc);
1673 }
1674 if (codec_id == AV_CODEC_ID_NONE) {
1675 char buf[32];
1676 av_get_codec_tag_string(buf, sizeof(buf), track->video.fourcc);
1677 av_log(matroska->ctx, AV_LOG_ERROR,
1678 "mov FourCC not found %s.\n", buf);
1679 }
1680 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1681 switch (track->audio.bitdepth) {
1682 case 8:
1683 codec_id = AV_CODEC_ID_PCM_U8;
1684 break;
1685 case 24:
1686 codec_id = AV_CODEC_ID_PCM_S24BE;
1687 break;
1688 case 32:
1689 codec_id = AV_CODEC_ID_PCM_S32BE;
1690 break;
1691 }
1692 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1693 switch (track->audio.bitdepth) {
1694 case 8:
1695 codec_id = AV_CODEC_ID_PCM_U8;
1696 break;
1697 case 24:
1698 codec_id = AV_CODEC_ID_PCM_S24LE;
1699 break;
1700 case 32:
1701 codec_id = AV_CODEC_ID_PCM_S32LE;
1702 break;
1703 }
1704 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
1705 track->audio.bitdepth == 64) {
1706 codec_id = AV_CODEC_ID_PCM_F64LE;
1707 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1708 int profile = matroska_aac_profile(track->codec_id);
1709 int sri = matroska_aac_sri(track->audio.samplerate);
1710 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
1711 if (!extradata)
1712 return AVERROR(ENOMEM);
1713 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
1714 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
1715 if (strstr(track->codec_id, "SBR")) {
1716 sri = matroska_aac_sri(track->audio.out_samplerate);
1717 extradata[2] = 0x56;
1718 extradata[3] = 0xE5;
1719 extradata[4] = 0x80 | (sri << 3);
1720 extradata_size = 5;
1721 } else
1722 extradata_size = 2;
1723 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size) {
1724 /* Only ALAC's magic cookie is stored in Matroska's track headers.
1725 * Create the "atom size", "tag", and "tag version" fields the
1726 * decoder expects manually. */
1727 extradata_size = 12 + track->codec_priv.size;
1728 extradata = av_mallocz(extradata_size +
1729 AV_INPUT_BUFFER_PADDING_SIZE);
1730 if (!extradata)
1731 return AVERROR(ENOMEM);
1732 AV_WB32(extradata, extradata_size);
1733 memcpy(&extradata[4], "alac", 4);
1734 AV_WB32(&extradata[8], 0);
1735 memcpy(&extradata[12], track->codec_priv.data,
1736 track->codec_priv.size);
1737 } else if (codec_id == AV_CODEC_ID_TTA) {
1738 extradata_size = 30;
1739 extradata = av_mallocz(extradata_size);
1740 if (!extradata)
1741 return AVERROR(ENOMEM);
1742 ffio_init_context(&b, extradata, extradata_size, 1,
1743 NULL, NULL, NULL, NULL);
1744 avio_write(&b, "TTA1", 4);
1745 avio_wl16(&b, 1);
1746 avio_wl16(&b, track->audio.channels);
1747 avio_wl16(&b, track->audio.bitdepth);
1748 avio_wl32(&b, track->audio.out_samplerate);
1749 avio_wl32(&b, matroska->ctx->duration *
1750 track->audio.out_samplerate);
1751 } else if (codec_id == AV_CODEC_ID_RV10 ||
1752 codec_id == AV_CODEC_ID_RV20 ||
1753 codec_id == AV_CODEC_ID_RV30 ||
1754 codec_id == AV_CODEC_ID_RV40) {
1755 extradata_offset = 26;
1756 } else if (codec_id == AV_CODEC_ID_RA_144) {
1757 track->audio.out_samplerate = 8000;
1758 track->audio.channels = 1;
1759 } else if (codec_id == AV_CODEC_ID_RA_288 ||
1760 codec_id == AV_CODEC_ID_COOK ||
1761 codec_id == AV_CODEC_ID_ATRAC3 ||
1762 codec_id == AV_CODEC_ID_SIPR) {
1763 int flavor;
1764 ffio_init_context(&b, track->codec_priv.data,
1765 track->codec_priv.size,
1766 0, NULL, NULL, NULL, NULL);
1767 avio_skip(&b, 22);
1768 flavor = avio_rb16(&b);
1769 track->audio.coded_framesize = avio_rb32(&b);
1770 avio_skip(&b, 12);
1771 track->audio.sub_packet_h = avio_rb16(&b);
1772 track->audio.frame_size = avio_rb16(&b);
1773 track->audio.sub_packet_size = avio_rb16(&b);
1774 if (flavor <= 0 ||
1775 track->audio.coded_framesize <= 0 ||
1776 track->audio.sub_packet_h <= 0 ||
1777 track->audio.frame_size <= 0 ||
1778 track->audio.sub_packet_size <= 0)
1779 return AVERROR_INVALIDDATA;
1780 track->audio.buf = av_malloc(track->audio.frame_size *
1781 track->audio.sub_packet_h);
1782 if (!track->audio.buf)
1783 return AVERROR(ENOMEM);
1784 if (codec_id == AV_CODEC_ID_RA_288) {
1785 st->codecpar->block_align = track->audio.coded_framesize;
1786 track->codec_priv.size = 0;
1787 } else {
1788 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1789 const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1790 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1791 st->codecpar->bit_rate = sipr_bit_rate[flavor];
1792 }
1793 st->codecpar->block_align = track->audio.sub_packet_size;
1794 extradata_offset = 78;
1795 }
1796 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
1797 ret = matroska_parse_flac(s, track, &extradata_offset);
1798 if (ret < 0)
1799 return ret;
1800 }
1801 track->codec_priv.size -= extradata_offset;
1802
1803 if (codec_id == AV_CODEC_ID_NONE)
1804 av_log(matroska->ctx, AV_LOG_INFO,
1805 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1806
1807 if (track->time_scale < 0.01)
1808 track->time_scale = 1.0;
1809 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
1810 1000 * 1000 * 1000); /* 64 bit pts in ns */
1811
1812 /* convert the delay from ns to the track timebase */
1813 track->codec_delay = av_rescale_q(track->codec_delay,
1814 (AVRational){ 1, 1000000000 },
1815 st->time_base);
1816
1817 st->codecpar->codec_id = codec_id;
1818 st->start_time = 0;
1819 if (strcmp(track->language, "und"))
1820 av_dict_set(&st->metadata, "language", track->language, 0);
1821 av_dict_set(&st->metadata, "title", track->name, 0);
1822
1823 if (track->flag_default)
1824 st->disposition |= AV_DISPOSITION_DEFAULT;
1825 if (track->flag_forced)
1826 st->disposition |= AV_DISPOSITION_FORCED;
1827
1828 if (!st->codecpar->extradata) {
1829 if (extradata) {
1830 st->codecpar->extradata = extradata;
1831 st->codecpar->extradata_size = extradata_size;
1832 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
1833 st->codecpar->extradata = av_mallocz(track->codec_priv.size +
1834 AV_INPUT_BUFFER_PADDING_SIZE);
1835 if (!st->codecpar->extradata)
1836 return AVERROR(ENOMEM);
1837 st->codecpar->extradata_size = track->codec_priv.size;
1838 memcpy(st->codecpar->extradata,
1839 track->codec_priv.data + extradata_offset,
1840 track->codec_priv.size);
1841 }
1842 }
1843
1844 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1845 int display_width_mul = 1;
1846 int display_height_mul = 1;
1847
1848 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
1849 st->codecpar->codec_tag = track->video.fourcc;
1850 st->codecpar->width = track->video.pixel_width;
1851 st->codecpar->height = track->video.pixel_height;
1852
1853 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
1854 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
1855
1856 av_reduce(&st->sample_aspect_ratio.num,
1857 &st->sample_aspect_ratio.den,
1858 st->codecpar->height * track->video.display_width * display_width_mul,
1859 st->codecpar->width * track->video.display_height * display_height_mul,
1860 255);
1861 if (st->codecpar->codec_id != AV_CODEC_ID_H264 &&
1862 st->codecpar->codec_id != AV_CODEC_ID_HEVC)
1863 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1864 if (track->default_duration) {
1865 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1866 1000000000, track->default_duration, 30000);
1867 }
1868 // add stream level stereo3d side data if it is a supported format
1869 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
1870 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
1871 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
1872 if (ret < 0)
1873 return ret;
1874 }
1875 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1876 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
1877 st->codecpar->sample_rate = track->audio.out_samplerate;
1878 st->codecpar->channels = track->audio.channels;
1879 if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
1880 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1881 if (st->codecpar->codec_id == AV_CODEC_ID_MP3)
1882 st->need_parsing = AVSTREAM_PARSE_FULL;
1883 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1884 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
1885 if (st->codecpar->codec_id == AV_CODEC_ID_SSA)
1886 matroska->contains_ssa = 1;
1887 }
1888 }
1889
1890 return 0;
1891 }
1892
1893 static int matroska_read_header(AVFormatContext *s)
1894 {
1895 MatroskaDemuxContext *matroska = s->priv_data;
1896 EbmlList *attachments_list = &matroska->attachments;
1897 EbmlList *chapters_list = &matroska->chapters;
1898 MatroskaAttachment *attachments;
1899 MatroskaChapter *chapters;
1900 uint64_t max_start = 0;
1901 int64_t pos;
1902 Ebml ebml = { 0 };
1903 int i, j, res;
1904
1905 matroska->ctx = s;
1906
1907 /* First read the EBML header. */
1908 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
1909 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
1910 ebml_free(ebml_syntax, &ebml);
1911 return AVERROR_INVALIDDATA;
1912 }
1913 if (ebml.version > EBML_VERSION ||
1914 ebml.max_size > sizeof(uint64_t) ||
1915 ebml.id_length > sizeof(uint32_t) ||
1916 ebml.doctype_version > 3) {
1917 av_log(matroska->ctx, AV_LOG_ERROR,
1918 "EBML header using unsupported features\n"
1919 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1920 ebml.version, ebml.doctype, ebml.doctype_version);
1921 ebml_free(ebml_syntax, &ebml);
1922 return AVERROR_PATCHWELCOME;
1923 }
1924 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
1925 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
1926 break;
1927 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
1928 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
1929 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
1930 ebml_free(ebml_syntax, &ebml);
1931 return AVERROR_INVALIDDATA;
1932 }
1933 }
1934 ebml_free(ebml_syntax, &ebml);
1935
1936 /* The next thing is a segment. */
1937 pos = avio_tell(matroska->ctx->pb);
1938 res = ebml_parse(matroska, matroska_segments, matroska);
1939 // try resyncing until we find a EBML_STOP type element.
1940 while (res != 1) {
1941 res = matroska_resync(matroska, pos);
1942 if (res < 0)
1943 return res;
1944 pos = avio_tell(matroska->ctx->pb);
1945 res = ebml_parse(matroska, matroska_segment, matroska);
1946 }
1947 matroska_execute_seekhead(matroska);
1948
1949 if (!matroska->time_scale)
1950 matroska->time_scale = 1000000;
1951 if (matroska->duration)
1952 matroska->ctx->duration = matroska->duration * matroska->time_scale *
1953 1000 / AV_TIME_BASE;
1954 av_dict_set(&s->metadata, "title", matroska->title, 0);
1955
1956 res = matroska_parse_tracks(s);
1957 if (res < 0)
1958 return res;
1959
1960 attachments = attachments_list->elem;
1961 for (j = 0; j < attachments_list->nb_elem; j++) {
1962 if (!(attachments[j].filename && attachments[j].mime &&
1963 attachments[j].bin.data && attachments[j].bin.size > 0)) {
1964 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1965 } else {
1966 AVStream *st = avformat_new_stream(s, NULL);
1967 if (!st)
1968 break;
1969 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
1970 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
1971 st->codecpar->codec_id = AV_CODEC_ID_NONE;
1972
1973 for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
1974 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
1975 strlen(ff_mkv_image_mime_tags[i].str))) {
1976 st->codecpar->codec_id = ff_mkv_image_mime_tags[i].id;
1977 break;
1978 }
1979 }
1980
1981 attachments[j].stream = st;
1982
1983 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
1984 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
1985 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
1986
1987 av_init_packet(&st->attached_pic);
1988 if ((res = av_new_packet(&st->attached_pic, attachments[j].bin.size)) < 0)
1989 return res;
1990 memcpy(st->attached_pic.data, attachments[j].bin.data, attachments[j].bin.size);
1991 st->attached_pic.stream_index = st->index;
1992 st->attached_pic.flags |= AV_PKT_FLAG_KEY;
1993 } else {
1994 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
1995 st->codecpar->extradata = av_malloc(attachments[j].bin.size);
1996 if (!st->codecpar->extradata)
1997 break;
1998
1999 st->codecpar->extradata_size = attachments[j].bin.size;
2000 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2001 attachments[j].bin.size);
2002
2003 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2004 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2005 strlen(ff_mkv_mime_tags[i].str))) {
2006 st->codecpar->codec_id = ff_mkv_mime_tags[i].id;
2007 break;
2008 }
2009 }
2010 }
2011 }
2012 }
2013
2014 chapters = chapters_list->elem;
2015 for (i = 0; i < chapters_list->nb_elem; i++)
2016 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2017 (max_start == 0 || chapters[i].start > max_start)) {
2018 chapters[i].chapter =
2019 avpriv_new_chapter(s, chapters[i].uid,
2020 (AVRational) { 1, 1000000000 },
2021 chapters[i].start, chapters[i].end,
2022 chapters[i].title);
2023 av_dict_set(&chapters[i].chapter->metadata,
2024 "title", chapters[i].title, 0);
2025 max_start = chapters[i].start;
2026 }
2027
2028 matroska_convert_tags(s);
2029
2030 return 0;
2031 }
2032
2033 /*
2034 * Put one packet in an application-supplied AVPacket struct.
2035 * Returns 0 on success or -1 on failure.
2036 */
2037 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2038 AVPacket *pkt)
2039 {
2040 if (matroska->num_packets > 0) {
2041 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2042 av_free(matroska->packets[0]);
2043 if (matroska->num_packets > 1) {
2044 void *newpackets;
2045 memmove(&matroska->packets[0], &matroska->packets[1],
2046 (matroska->num_packets - 1) * sizeof(AVPacket *));
2047 newpackets = av_realloc(matroska->packets,
2048 (matroska->num_packets - 1) *
2049 sizeof(AVPacket *));
2050 if (newpackets)
2051 matroska->packets = newpackets;
2052 } else {
2053 av_freep(&matroska->packets);
2054 matroska->prev_pkt = NULL;
2055 }
2056 matroska->num_packets--;
2057 return 0;
2058 }
2059
2060 return -1;
2061 }
2062
2063 /*
2064 * Free all packets in our internal queue.
2065 */
2066 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2067 {
2068 matroska->prev_pkt = NULL;
2069 if (matroska->packets) {
2070 int n;
2071 for (n = 0; n < matroska->num_packets; n++) {
2072 av_packet_unref(matroska->packets[n]);
2073 av_free(matroska->packets[n]);
2074 }
2075 av_freep(&matroska->packets);
2076 matroska->num_packets = 0;
2077 }
2078 }
2079
2080 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2081 int *buf_size, int type,
2082 uint32_t **lace_buf, int *laces)
2083 {
2084 int res = 0, n, size = *buf_size;
2085 uint8_t *data = *buf;
2086 uint32_t *lace_size;
2087
2088 if (!type) {
2089 *laces = 1;
2090 *lace_buf = av_mallocz(sizeof(int));
2091 if (!*lace_buf)
2092 return AVERROR(ENOMEM);
2093
2094 *lace_buf[0] = size;
2095 return 0;
2096 }
2097
2098 assert(size > 0);
2099 *laces = *data + 1;
2100 data += 1;
2101 size -= 1;
2102 lace_size = av_mallocz(*laces * sizeof(int));
2103 if (!lace_size)
2104 return AVERROR(ENOMEM);
2105
2106 switch (type) {
2107 case 0x1: /* Xiph lacing */
2108 {
2109 uint8_t temp;
2110 uint32_t total = 0;
2111 for (n = 0; res == 0 && n < *laces - 1; n++) {
2112 while (1) {
2113 if (size == 0) {
2114 res = AVERROR_EOF;
2115 break;
2116 }
2117 temp = *data;
2118 lace_size[n] += temp;
2119 data += 1;
2120 size -= 1;
2121 if (temp != 0xff)
2122 break;
2123 }
2124 total += lace_size[n];
2125 }
2126 if (size <= total) {
2127 res = AVERROR_INVALIDDATA;
2128 break;
2129 }
2130
2131 lace_size[n] = size - total;
2132 break;
2133 }
2134
2135 case 0x2: /* fixed-size lacing */
2136 if (size % (*laces)) {
2137 res = AVERROR_INVALIDDATA;
2138 break;
2139 }
2140 for (n = 0; n < *laces; n++)
2141 lace_size[n] = size / *laces;
2142 break;
2143
2144 case 0x3: /* EBML lacing */
2145 {
2146 uint64_t num;
2147 uint64_t total;
2148 n = matroska_ebmlnum_uint(matroska, data, size, &num);
2149 if (n < 0) {
2150 av_log(matroska->ctx, AV_LOG_INFO,
2151 "EBML block data error\n");
2152 res = n;
2153 break;
2154 }
2155 data += n;
2156 size -= n;
2157 total = lace_size[0] = num;
2158 for (n = 1; res == 0 && n < *laces - 1; n++) {
2159 int64_t snum;
2160 int r;
2161 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2162 if (r < 0) {
2163 av_log(matroska->ctx, AV_LOG_INFO,
2164 "EBML block data error\n");
2165 res = r;
2166 break;
2167 }
2168 data += r;
2169 size -= r;
2170 lace_size[n] = lace_size[n - 1] + snum;
2171 total += lace_size[n];
2172 }
2173 if (size <= total) {
2174 res = AVERROR_INVALIDDATA;
2175 break;
2176 }
2177 lace_size[*laces - 1] = size - total;
2178 break;
2179 }
2180 }
2181
2182 *buf = data;
2183 *lace_buf = lace_size;
2184 *buf_size = size;
2185
2186 return res;
2187 }
2188
2189 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2190 MatroskaTrack *track, AVStream *st,
2191 uint8_t *data, int size, uint64_t timecode,
2192 uint64_t duration, int64_t pos)
2193 {
2194 int a = st->codecpar->block_align;
2195 int sps = track->audio.sub_packet_size;
2196 int cfs = track->audio.coded_framesize;
2197 int h = track->audio.sub_packet_h;
2198 int y = track->audio.sub_packet_cnt;
2199 int w = track->audio.frame_size;
2200 int x;
2201
2202 if (!track->audio.pkt_cnt) {
2203 if (track->audio.sub_packet_cnt == 0)
2204 track->audio.buf_timecode = timecode;
2205 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
2206 if (size < cfs * h / 2) {
2207 av_log(matroska->ctx, AV_LOG_ERROR,
2208 "Corrupt int4 RM-style audio packet size\n");
2209 return AVERROR_INVALIDDATA;
2210 }
2211 for (x = 0; x < h / 2; x++)
2212 memcpy(track->audio.buf + x * 2 * w + y * cfs,
2213 data + x * cfs, cfs);
2214 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
2215 if (size < w) {
2216 av_log(matroska->ctx, AV_LOG_ERROR,
2217 "Corrupt sipr RM-style audio packet size\n");
2218 return AVERROR_INVALIDDATA;
2219 }
2220 memcpy(track->audio.buf + y * w, data, w);
2221 } else {
2222 if (size < sps * w / sps) {
2223 av_log(matroska->ctx, AV_LOG_ERROR,
2224 "Corrupt generic RM-style audio packet size\n");
2225 return AVERROR_INVALIDDATA;
2226 }
2227 for (x = 0; x < w / sps; x++)
2228 memcpy(track->audio.buf +
2229 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
2230 data + x * sps, sps);
2231 }
2232
2233 if (++track->audio.sub_packet_cnt >= h) {
2234 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
2235 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2236 track->audio.sub_packet_cnt = 0;
2237 track->audio.pkt_cnt = h * w / a;
2238 }
2239 }
2240
2241 while (track->audio.pkt_cnt) {
2242 int ret;
2243 AVPacket *pkt = av_mallocz(sizeof(AVPacket));
2244 if (!pkt)
2245 return AVERROR(ENOMEM);
2246
2247 ret = av_new_packet(pkt, a);
2248 if (ret < 0) {
2249 av_free(pkt);
2250 return ret;
2251 }
2252 memcpy(pkt->data,
2253 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
2254 a);
2255 pkt->pts = track->audio.buf_timecode;
2256 track->audio.buf_timecode = AV_NOPTS_VALUE;
2257 pkt->pos = pos;
2258 pkt->stream_index = st->index;
2259 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2260 }
2261
2262 return 0;
2263 }
2264
2265 /* reconstruct full wavpack blocks from mangled matroska ones */
2266 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
2267 uint8_t **pdst, int *size)
2268 {
2269 uint8_t *dst = NULL;
2270 int dstlen = 0;
2271 int srclen = *size;
2272 uint32_t samples;
2273 uint16_t ver;
2274 int ret, offset = 0;
2275
2276 if (srclen < 12 || track->stream->codecpar->extradata_size < 2)
2277 return AVERROR_INVALIDDATA;
2278
2279 ver = AV_RL16(track->stream->codecpar->extradata);
2280
2281 samples = AV_RL32(src);
2282 src += 4;
2283 srclen -= 4;
2284
2285 while (srclen >= 8) {
2286 int multiblock;
2287 uint32_t blocksize;
2288 uint8_t *tmp;
2289
2290 uint32_t flags = AV_RL32(src);
2291 uint32_t crc = AV_RL32(src + 4);
2292 src += 8;
2293 srclen -= 8;
2294
2295 multiblock = (flags & 0x1800) != 0x1800;
2296 if (multiblock) {
2297 if (srclen < 4) {
2298 ret = AVERROR_INVALIDDATA;
2299 goto fail;
2300 }
2301 blocksize = AV_RL32(src);
2302 src += 4;
2303 srclen -= 4;
2304 } else
2305 blocksize = srclen;
2306
2307 if (blocksize > srclen) {
2308 ret = AVERROR_INVALIDDATA;
2309 goto fail;
2310 }
2311
2312 tmp = av_realloc(dst, dstlen + blocksize + 32);
2313 if (!tmp) {
2314 ret = AVERROR(ENOMEM);
2315 goto fail;
2316 }
2317 dst = tmp;
2318 dstlen += blocksize + 32;
2319
2320 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2321 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2322 AV_WL16(dst + offset + 8, ver); // version
2323 AV_WL16(dst + offset + 10, 0); // track/index_no
2324 AV_WL32(dst + offset + 12, 0); // total samples
2325 AV_WL32(dst + offset + 16, 0); // block index
2326 AV_WL32(dst + offset + 20, samples); // number of samples
2327 AV_WL32(dst + offset + 24, flags); // flags
2328 AV_WL32(dst + offset + 28, crc); // crc
2329 memcpy(dst + offset + 32, src, blocksize); // block data
2330
2331 src += blocksize;
2332 srclen -= blocksize;
2333 offset += blocksize + 32;
2334 }
2335
2336 *pdst = dst;
2337 *size = dstlen;
2338
2339 return 0;
2340
2341 fail:
2342 av_freep(&dst);
2343 return ret;
2344 }
2345
2346 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2347 MatroskaTrack *track, AVStream *st,
2348 uint8_t *data, int pkt_size,
2349 uint64_t timecode, uint64_t duration,
2350 int64_t pos, int is_keyframe)
2351 {
2352 MatroskaTrackEncoding *encodings = track->encodings.elem;
2353 uint8_t *pkt_data = data;
2354 int offset = 0, res;
2355 AVPacket *pkt;
2356
2357 if (encodings && encodings->scope & 1) {
2358 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2359 if (res < 0)
2360 return res;
2361 }
2362
2363 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
2364 uint8_t *wv_data;
2365 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
2366 if (res < 0) {
2367 av_log(matroska->ctx, AV_LOG_ERROR,
2368 "Error parsing a wavpack block.\n");
2369 goto fail;
2370 }
2371 if (pkt_data != data)
2372 av_freep(&pkt_data);
2373 pkt_data = wv_data;
2374 }
2375
2376 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES)
2377 offset = 8;
2378
2379 pkt = av_mallocz(sizeof(AVPacket));
2380 if (!pkt) {
2381 av_freep(&pkt_data);
2382 return AVERROR(ENOMEM);
2383 }
2384 /* XXX: prevent data copy... */
2385 if (av_new_packet(pkt, pkt_size + offset) < 0) {
2386 av_free(pkt);
2387 av_freep(&pkt_data);
2388 return AVERROR(ENOMEM);
2389 }
2390
2391 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES) {
2392 uint8_t *buf = pkt->data;
2393 bytestream_put_be32(&buf, pkt_size);
2394 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2395 }
2396
2397 memcpy(pkt->data + offset, pkt_data, pkt_size);
2398
2399 if (pkt_data != data)
2400 av_free(pkt_data);
2401
2402 pkt->flags = is_keyframe;
2403 pkt->stream_index = st->index;
2404
2405 if (track->ms_compat)
2406 pkt->dts = timecode;
2407 else
2408 pkt->pts = timecode;
2409 pkt->pos = pos;
2410 if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE || st->codecpar->codec_id == AV_CODEC_ID_TEXT)
2411 pkt->duration = duration;
2412 #if FF_API_CONVERGENCE_DURATION
2413 FF_DISABLE_DEPRECATION_WARNINGS
2414 if (st->codecpar->codec_id == AV_CODEC_ID_TEXT)
2415 pkt->convergence_duration = duration;
2416 FF_ENABLE_DEPRECATION_WARNINGS
2417 #endif
2418
2419 if (st->codecpar->codec_id == AV_CODEC_ID_SSA)
2420 matroska_fix_ass_packet(matroska, pkt, duration);
2421
2422 if (matroska->prev_pkt &&
2423 timecode != AV_NOPTS_VALUE &&
2424 matroska->prev_pkt->pts == timecode &&
2425 matroska->prev_pkt->stream_index == st->index &&
2426 st->codecpar->codec_id == AV_CODEC_ID_SSA)
2427 matroska_merge_packets(matroska->prev_pkt, pkt);
2428 else {
2429 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2430 matroska->prev_pkt = pkt;
2431 }
2432
2433 return 0;
2434
2435 fail:
2436 if (pkt_data != data)
2437 av_freep(&pkt_data);
2438 return res;
2439 }
2440
2441 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
2442 int size, int64_t pos, uint64_t cluster_time,
2443 uint64_t block_duration, int is_keyframe,
2444 int64_t cluster_pos)
2445 {
2446 uint64_t timecode = AV_NOPTS_VALUE;
2447 MatroskaTrack *track;
2448 int res = 0;
2449 AVStream *st;
2450 int16_t block_time;
2451 uint32_t *lace_size = NULL;
2452 int n, flags, laces = 0;
2453 uint64_t num, duration;
2454
2455 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2456 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2457 return n;
2458 }
2459 data += n;
2460 size -= n;
2461
2462 track = matroska_find_track_by_num(matroska, num);
2463 if (!track || !track->stream) {
2464 av_log(matroska->ctx, AV_LOG_INFO,
2465 "Invalid stream %"PRIu64" or size %u\n", num, size);
2466 return AVERROR_INVALIDDATA;
2467 } else if (size <= 3)
2468 return 0;
2469 st = track->stream;
2470 if (st->discard >= AVDISCARD_ALL)
2471 return res;
2472
2473 block_time = AV_RB16(data);
2474 data += 2;
2475 flags = *data++;
2476 size -= 3;
2477 if (is_keyframe == -1)
2478 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2479
2480 if (cluster_time != (uint64_t) -1 &&
2481 (block_time >= 0 || cluster_time >= -block_time)) {
2482 timecode = cluster_time + block_time - track->codec_delay;
2483 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2484 timecode < track->end_timecode)
2485 is_keyframe = 0; /* overlapping subtitles are not key frame */
2486 if (is_keyframe)
2487 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
2488 AVINDEX_KEYFRAME);
2489 }
2490
2491 if (matroska->skip_to_keyframe &&
2492 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2493 if (!is_keyframe || timecode < matroska->skip_to_timecode)
2494 return res;
2495 matroska->skip_to_keyframe = 0;
2496 }
2497
2498 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
2499 &lace_size, &laces);
2500
2501 if (res)
2502 goto end;
2503
2504 if (block_duration != AV_NOPTS_VALUE) {
2505 duration = block_duration / laces;
2506 if (block_duration != duration * laces) {
2507 av_log(matroska->ctx, AV_LOG_WARNING,
2508 "Incorrect block_duration, possibly corrupted container");
2509 }
2510 } else {
2511 duration = track->default_duration / matroska->time_scale;
2512 block_duration = duration * laces;
2513 }
2514
2515 if (timecode != AV_NOPTS_VALUE)
2516 track->end_timecode =
2517 FFMAX(track->end_timecode, timecode + block_duration);
2518
2519 for (n = 0; n < laces; n++) {
2520 if ((st->codecpar->codec_id == AV_CODEC_ID_RA_288 ||
2521 st->codecpar->codec_id == AV_CODEC_ID_COOK ||
2522 st->codecpar->codec_id == AV_CODEC_ID_SIPR ||
2523 st->codecpar->codec_id == AV_CODEC_ID_ATRAC3) &&
2524 st->codecpar->block_align && track->audio.sub_packet_size) {
2525 res = matroska_parse_rm_audio(matroska, track, st, data,
2526 lace_size[n],
2527 timecode, duration, pos);
2528 if (res)
2529 goto end;
2530 } else {
2531 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2532 timecode, duration, pos,
2533 !n ? is_keyframe : 0);
2534 if (res)
2535 goto end;
2536 }
2537
2538 if (timecode != AV_NOPTS_VALUE)
2539 timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
2540 data += lace_size[n];
2541 }
2542
2543 end:
2544 av_free(lace_size);
2545 return res;
2546 }
2547
2548 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
2549 {
2550 EbmlList *blocks_list;
2551 MatroskaBlock *blocks;
2552 int i, res;
2553 res = ebml_parse(matroska,
2554 matroska_cluster_incremental_parsing,
2555 &matroska->current_cluster);
2556 if (res == 1) {
2557 /* New Cluster */
2558 if (matroska->current_cluster_pos)
2559 ebml_level_end(matroska);
2560 ebml_free(matroska_cluster, &matroska->current_cluster);
2561 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2562 matroska->current_cluster_num_blocks = 0;
2563 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2564 matroska->prev_pkt = NULL;
2565 /* sizeof the ID which was already read */
2566 if (matroska->current_id)
2567 matroska->current_cluster_pos -= 4;
2568 res = ebml_parse(matroska,
2569 matroska_clusters_incremental,
2570 &matroska->current_cluster);
2571 /* Try parsing the block again. */
2572 if (res == 1)
2573 res = ebml_parse(matroska,
2574 matroska_cluster_incremental_parsing,
2575 &matroska->current_cluster);
2576 }
2577
2578 if (!res &&
2579 matroska->current_cluster_num_blocks <
2580 matroska->current_cluster.blocks.nb_elem) {
2581 blocks_list = &matroska->current_cluster.blocks;
2582 blocks = blocks_list->elem;
2583
2584 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2585 i = blocks_list->nb_elem - 1;
2586 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2587 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2588 if (!blocks[i].non_simple)
2589 blocks[i].duration = AV_NOPTS_VALUE;
2590 res = matroska_parse_block(matroska, blocks[i].bin.data,
2591 blocks[i].bin.size, blocks[i].bin.pos,
2592 matroska->current_cluster.timecode,
2593 blocks[i].duration, is_keyframe,
2594 matroska->current_cluster_pos);
2595 }
2596 }
2597
2598 if (res < 0)
2599 matroska->done = 1;
2600 return res;
2601 }
2602
2603 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
2604 {
2605 MatroskaCluster cluster = { 0 };
2606 EbmlList *blocks_list;
2607 MatroskaBlock *blocks;
2608 int i, res;
2609 int64_t pos;
2610
2611 if (!matroska->contains_ssa)
2612 return matroska_parse_cluster_incremental(matroska);
2613 pos = avio_tell(matroska->ctx->pb);
2614 matroska->prev_pkt = NULL;
2615 if (matroska->current_id)
2616 pos -= 4; /* sizeof the ID which was already read */
2617 res = ebml_parse(matroska, matroska_clusters, &cluster);
2618 blocks_list = &cluster.blocks;
2619 blocks = blocks_list->elem;
2620 for (i = 0; i < blocks_list->nb_elem && !res; i++)
2621 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2622 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2623 if (!blocks[i].non_simple)
2624 blocks[i].duration = AV_NOPTS_VALUE;
2625 res = matroska_parse_block(matroska, blocks[i].bin.data,
2626 blocks[i].bin.size, blocks[i].bin.pos,
2627 cluster.timecode, blocks[i].duration,
2628 is_keyframe, pos);
2629 }
2630 ebml_free(matroska_cluster, &cluster);
2631 return res;
2632 }
2633
2634 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
2635 {
2636 MatroskaDemuxContext *matroska = s->priv_data;
2637 int ret = 0;
2638
2639 while (!ret && matroska_deliver_packet(matroska, pkt)) {
2640 int64_t pos = avio_tell(matroska->ctx->pb);
2641 if (matroska->done)
2642 return AVERROR_EOF;
2643 if (matroska_parse_cluster(matroska) < 0)
2644 ret = matroska_resync(matroska, pos);
2645 }
2646
2647 if (ret == AVERROR_INVALIDDATA && pkt->data) {
2648 pkt->flags |= AV_PKT_FLAG_CORRUPT;
2649 return 0;
2650 }
2651
2652 return ret;
2653 }
2654
2655 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2656 int64_t timestamp, int flags)
2657 {
2658 MatroskaDemuxContext *matroska = s->priv_data;
2659 MatroskaTrack *tracks = NULL;
2660 AVStream *st = s->streams[stream_index];
2661 int i, index, index_sub, index_min;
2662
2663 /* Parse the CUES now since we need the index data to seek. */
2664 if (matroska->cues_parsing_deferred) {
2665 matroska_parse_cues(matroska);
2666 matroska->cues_parsing_deferred = 0;
2667 }
2668
2669 if (!st->nb_index_entries)
2670 return 0;
2671 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2672
2673 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2674 avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
2675 SEEK_SET);
2676 matroska->current_id = 0;
2677 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2678 matroska_clear_queue(matroska);
2679 if (matroska_parse_cluster(matroska) < 0)
2680 break;
2681 }
2682 }
2683
2684 matroska_clear_queue(matroska);
2685 if (index < 0)
2686 return 0;
2687
2688 index_min = index;
2689 tracks = matroska->tracks.elem;
2690 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2691 tracks[i].audio.pkt_cnt = 0;
2692 tracks[i].audio.sub_packet_cnt = 0;
2693 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2694 tracks[i].end_timecode = 0;
2695 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2696 tracks[i].stream->discard != AVDISCARD_ALL) {
2697 index_sub = av_index_search_timestamp(
2698 tracks[i].stream, st->index_entries[index].timestamp,
2699 AVSEEK_FLAG_BACKWARD);
2700 if (index_sub >= 0 &&
2701 st->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
2702 st->index_entries[index].timestamp -
2703 st->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
2704 index_min = index_sub;
2705 }
2706 }
2707
2708 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2709 matroska->current_id = 0;
2710 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
2711 matroska->skip_to_timecode = st->index_entries[index].timestamp;
2712 matroska->done = 0;
2713 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2714 return 0;
2715 }
2716
2717 static int matroska_read_close(AVFormatContext *s)
2718 {
2719 MatroskaDemuxContext *matroska = s->priv_data;
2720 MatroskaTrack *tracks = matroska->tracks.elem;
2721 int n;
2722
2723 matroska_clear_queue(matroska);
2724
2725 for (n = 0; n < matroska->tracks.nb_elem; n++)
2726 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2727 av_free(tracks[n].audio.buf);
2728 ebml_free(matroska_cluster, &matroska->current_cluster);
2729 ebml_free(matroska_segment, matroska);
2730
2731 return 0;
2732 }
2733
2734 AVInputFormat ff_matroska_demuxer = {
2735 .name = "matroska,webm",
2736 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
2737 .extensions = "mkv,mk3d,mka,mks",
2738 .priv_data_size = sizeof(MatroskaDemuxContext),
2739 .read_probe = matroska_probe,
2740 .read_header = matroska_read_header,
2741 .read_packet = matroska_read_packet,
2742 .read_close = matroska_read_close,
2743 .read_seek = matroska_read_seek,
2744 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
2745 };