f15b0b3e9f5b925f8c0b76c42cb11e8117fadf8f
[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 {
90 int nb_elem;
91 void *elem;
92 } EbmlList;
93
94 typedef struct {
95 int size;
96 uint8_t *data;
97 int64_t pos;
98 } EbmlBin;
99
100 typedef struct {
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 {
109 uint64_t algo;
110 EbmlBin settings;
111 } MatroskaTrackCompression;
112
113 typedef struct {
114 uint64_t scope;
115 uint64_t type;
116 MatroskaTrackCompression compression;
117 } MatroskaTrackEncoding;
118
119 typedef struct {
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 } MatroskaTrackVideo;
127
128 typedef struct {
129 double samplerate;
130 double out_samplerate;
131 uint64_t bitdepth;
132 uint64_t channels;
133
134 /* real audio header (extracted from extradata) */
135 int coded_framesize;
136 int sub_packet_h;
137 int frame_size;
138 int sub_packet_size;
139 int sub_packet_cnt;
140 int pkt_cnt;
141 uint64_t buf_timecode;
142 uint8_t *buf;
143 } MatroskaTrackAudio;
144
145 typedef struct {
146 uint64_t num;
147 uint64_t uid;
148 uint64_t type;
149 char *name;
150 char *codec_id;
151 EbmlBin codec_priv;
152 char *language;
153 double time_scale;
154 uint64_t default_duration;
155 uint64_t flag_default;
156 uint64_t flag_forced;
157 MatroskaTrackVideo video;
158 MatroskaTrackAudio audio;
159 EbmlList encodings;
160 uint64_t codec_delay;
161
162 AVStream *stream;
163 int64_t end_timecode;
164 int ms_compat;
165 } MatroskaTrack;
166
167 typedef struct {
168 uint64_t uid;
169 char *filename;
170 char *mime;
171 EbmlBin bin;
172
173 AVStream *stream;
174 } MatroskaAttachment;
175
176 typedef struct {
177 uint64_t start;
178 uint64_t end;
179 uint64_t uid;
180 char *title;
181
182 AVChapter *chapter;
183 } MatroskaChapter;
184
185 typedef struct {
186 uint64_t track;
187 uint64_t pos;
188 } MatroskaIndexPos;
189
190 typedef struct {
191 uint64_t time;
192 EbmlList pos;
193 } MatroskaIndex;
194
195 typedef struct {
196 char *name;
197 char *string;
198 char *lang;
199 uint64_t def;
200 EbmlList sub;
201 } MatroskaTag;
202
203 typedef struct {
204 char *type;
205 uint64_t typevalue;
206 uint64_t trackuid;
207 uint64_t chapteruid;
208 uint64_t attachuid;
209 } MatroskaTagTarget;
210
211 typedef struct {
212 MatroskaTagTarget target;
213 EbmlList tag;
214 } MatroskaTags;
215
216 typedef struct {
217 uint64_t id;
218 uint64_t pos;
219 } MatroskaSeekhead;
220
221 typedef struct {
222 uint64_t start;
223 uint64_t length;
224 } MatroskaLevel;
225
226 typedef struct {
227 uint64_t timecode;
228 EbmlList blocks;
229 } MatroskaCluster;
230
231 typedef struct {
232 AVFormatContext *ctx;
233
234 /* EBML stuff */
235 int num_levels;
236 MatroskaLevel levels[EBML_MAX_DEPTH];
237 int level_up;
238 uint32_t current_id;
239
240 uint64_t time_scale;
241 double duration;
242 char *title;
243 EbmlList tracks;
244 EbmlList attachments;
245 EbmlList chapters;
246 EbmlList index;
247 EbmlList tags;
248 EbmlList seekhead;
249
250 /* byte position of the segment inside the stream */
251 int64_t segment_start;
252
253 /* the packet queue */
254 AVPacket **packets;
255 int num_packets;
256 AVPacket *prev_pkt;
257
258 int done;
259
260 /* What to skip before effectively reading a packet. */
261 int skip_to_keyframe;
262 uint64_t skip_to_timecode;
263
264 /* File has a CUES element, but we defer parsing until it is needed. */
265 int cues_parsing_deferred;
266
267 int current_cluster_num_blocks;
268 int64_t current_cluster_pos;
269 MatroskaCluster current_cluster;
270
271 /* File has SSA subtitles which prevent incremental cluster parsing. */
272 int contains_ssa;
273 } MatroskaDemuxContext;
274
275 typedef struct {
276 uint64_t duration;
277 int64_t reference;
278 uint64_t non_simple;
279 EbmlBin bin;
280 } MatroskaBlock;
281
282 static EbmlSyntax ebml_header[] = {
283 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
284 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
285 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
286 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
287 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
288 { EBML_ID_EBMLVERSION, EBML_NONE },
289 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
290 { 0 }
291 };
292
293 static EbmlSyntax ebml_syntax[] = {
294 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
295 { 0 }
296 };
297
298 static EbmlSyntax matroska_info[] = {
299 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
300 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
301 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
302 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
303 { MATROSKA_ID_MUXINGAPP, EBML_NONE },
304 { MATROSKA_ID_DATEUTC, EBML_NONE },
305 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
306 { 0 }
307 };
308
309 static EbmlSyntax matroska_track_video[] = {
310 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
311 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width) },
312 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height) },
313 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
314 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
315 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, fourcc) },
316 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
317 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
318 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
319 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
320 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
321 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_NONE },
322 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_NONE },
323 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
324 { 0 }
325 };
326
327 static EbmlSyntax matroska_track_audio[] = {
328 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
329 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
330 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
331 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
332 { 0 }
333 };
334
335 static EbmlSyntax matroska_track_encoding_compression[] = {
336 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
337 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
338 { 0 }
339 };
340
341 static EbmlSyntax matroska_track_encoding[] = {
342 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
343 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
344 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
345 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
346 { 0 }
347 };
348
349 static EbmlSyntax matroska_track_encodings[] = {
350 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
351 { 0 }
352 };
353
354 static EbmlSyntax matroska_track[] = {
355 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
356 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
357 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
358 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
359 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
360 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
361 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
362 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
363 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
364 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
365 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
366 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
367 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
368 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
369 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
370 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
371 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
372 { MATROSKA_ID_CODECNAME, EBML_NONE },
373 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
374 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
375 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
376 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
377 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
378 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
379 { 0 }
380 };
381
382 static EbmlSyntax matroska_tracks[] = {
383 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
384 { 0 }
385 };
386
387 static EbmlSyntax matroska_attachment[] = {
388 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
389 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
390 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
391 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
392 { MATROSKA_ID_FILEDESC, EBML_NONE },
393 { 0 }
394 };
395
396 static EbmlSyntax matroska_attachments[] = {
397 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
398 { 0 }
399 };
400
401 static EbmlSyntax matroska_chapter_display[] = {
402 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
403 { MATROSKA_ID_CHAPLANG, EBML_NONE },
404 { 0 }
405 };
406
407 static EbmlSyntax matroska_chapter_entry[] = {
408 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
409 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
410 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
411 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
412 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
413 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
414 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
415 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
416 { 0 }
417 };
418
419 static EbmlSyntax matroska_chapter[] = {
420 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
421 { MATROSKA_ID_EDITIONUID, EBML_NONE },
422 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
423 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
424 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
425 { 0 }
426 };
427
428 static EbmlSyntax matroska_chapters[] = {
429 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
430 { 0 }
431 };
432
433 static EbmlSyntax matroska_index_pos[] = {
434 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
435 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
436 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
437 { 0 }
438 };
439
440 static EbmlSyntax matroska_index_entry[] = {
441 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
442 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
443 { 0 }
444 };
445
446 static EbmlSyntax matroska_index[] = {
447 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
448 { 0 }
449 };
450
451 static EbmlSyntax matroska_simpletag[] = {
452 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
453 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
454 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
455 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
456 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
457 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
458 { 0 }
459 };
460
461 static EbmlSyntax matroska_tagtargets[] = {
462 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
463 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
464 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
465 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
466 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
467 { 0 }
468 };
469
470 static EbmlSyntax matroska_tag[] = {
471 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
472 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
473 { 0 }
474 };
475
476 static EbmlSyntax matroska_tags[] = {
477 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
478 { 0 }
479 };
480
481 static EbmlSyntax matroska_seekhead_entry[] = {
482 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
483 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
484 { 0 }
485 };
486
487 static EbmlSyntax matroska_seekhead[] = {
488 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
489 { 0 }
490 };
491
492 static EbmlSyntax matroska_segment[] = {
493 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, { .n = matroska_info } },
494 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, { .n = matroska_tracks } },
495 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, { .n = matroska_attachments } },
496 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, { .n = matroska_chapters } },
497 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, { .n = matroska_index } },
498 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, { .n = matroska_tags } },
499 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, { .n = matroska_seekhead } },
500 { MATROSKA_ID_CLUSTER, EBML_STOP },
501 { 0 }
502 };
503
504 static EbmlSyntax matroska_segments[] = {
505 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
506 { 0 }
507 };
508
509 static EbmlSyntax matroska_blockgroup[] = {
510 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
511 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
512 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration), { .u = AV_NOPTS_VALUE } },
513 { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock, reference) },
514 { MATROSKA_ID_CODECSTATE, EBML_NONE },
515 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
516 { 0 }
517 };
518
519 static EbmlSyntax matroska_cluster[] = {
520 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
521 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
522 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
523 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
524 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
525 { 0 }
526 };
527
528 static EbmlSyntax matroska_clusters[] = {
529 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster } },
530 { MATROSKA_ID_INFO, EBML_NONE },
531 { MATROSKA_ID_CUES, EBML_NONE },
532 { MATROSKA_ID_TAGS, EBML_NONE },
533 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
534 { 0 }
535 };
536
537 static EbmlSyntax matroska_cluster_incremental_parsing[] = {
538 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
539 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
540 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
541 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
542 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
543 { MATROSKA_ID_INFO, EBML_NONE },
544 { MATROSKA_ID_CUES, EBML_NONE },
545 { MATROSKA_ID_TAGS, EBML_NONE },
546 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
547 { MATROSKA_ID_CLUSTER, EBML_STOP },
548 { 0 }
549 };
550
551 static EbmlSyntax matroska_cluster_incremental[] = {
552 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
553 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
554 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
555 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
556 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
557 { 0 }
558 };
559
560 static EbmlSyntax matroska_clusters_incremental[] = {
561 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
562 { MATROSKA_ID_INFO, EBML_NONE },
563 { MATROSKA_ID_CUES, EBML_NONE },
564 { MATROSKA_ID_TAGS, EBML_NONE },
565 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
566 { 0 }
567 };
568
569 static const char *const matroska_doctypes[] = { "matroska", "webm" };
570
571 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
572 {
573 AVIOContext *pb = matroska->ctx->pb;
574 uint32_t id;
575 matroska->current_id = 0;
576 matroska->num_levels = 0;
577
578 /* seek to next position to resync from */
579 if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
580 goto eof;
581
582 id = avio_rb32(pb);
583
584 // try to find a toplevel element
585 while (!pb->eof_reached) {
586 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
587 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
588 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
589 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
590 matroska->current_id = id;
591 return 0;
592 }
593 id = (id << 8) | avio_r8(pb);
594 }
595
596 eof:
597 matroska->done = 1;
598 return AVERROR_EOF;
599 }
600
601 /*
602 * Return: Whether we reached the end of a level in the hierarchy or not.
603 */
604 static int ebml_level_end(MatroskaDemuxContext *matroska)
605 {
606 AVIOContext *pb = matroska->ctx->pb;
607 int64_t pos = avio_tell(pb);
608
609 if (matroska->num_levels > 0) {
610 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
611 if (pos - level->start >= level->length || matroska->current_id) {
612 matroska->num_levels--;
613 return 1;
614 }
615 }
616 return 0;
617 }
618
619 /*
620 * Read: an "EBML number", which is defined as a variable-length
621 * array of bytes. The first byte indicates the length by giving a
622 * number of 0-bits followed by a one. The position of the first
623 * "one" bit inside the first byte indicates the length of this
624 * number.
625 * Returns: number of bytes read, < 0 on error
626 */
627 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
628 int max_size, uint64_t *number)
629 {
630 int read = 1, n = 1;
631 uint64_t total = 0;
632
633 /* The first byte tells us the length in bytes - avio_r8() can normally
634 * return 0, but since that's not a valid first ebmlID byte, we can
635 * use it safely here to catch EOS. */
636 if (!(total = avio_r8(pb))) {
637 /* we might encounter EOS here */
638 if (!pb->eof_reached) {
639 int64_t pos = avio_tell(pb);
640 av_log(matroska->ctx, AV_LOG_ERROR,
641 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
642 pos, pos);
643 return pb->error ? pb->error : AVERROR(EIO);
644 }
645 return AVERROR_EOF;
646 }
647
648 /* get the length of the EBML number */
649 read = 8 - ff_log2_tab[total];
650 if (read > max_size) {
651 int64_t pos = avio_tell(pb) - 1;
652 av_log(matroska->ctx, AV_LOG_ERROR,
653 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
654 (uint8_t) total, pos, pos);
655 return AVERROR_INVALIDDATA;
656 }
657
658 /* read out length */
659 total ^= 1 << ff_log2_tab[total];
660 while (n++ < read)
661 total = (total << 8) | avio_r8(pb);
662
663 *number = total;
664
665 return read;
666 }
667
668 /**
669 * Read a EBML length value.
670 * This needs special handling for the "unknown length" case which has multiple
671 * encodings.
672 */
673 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
674 uint64_t *number)
675 {
676 int res = ebml_read_num(matroska, pb, 8, number);
677 if (res > 0 && *number + 1 == 1ULL << (7 * res))
678 *number = 0xffffffffffffffULL;
679 return res;
680 }
681
682 /*
683 * Read the next element as an unsigned int.
684 * 0 is success, < 0 is failure.
685 */
686 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
687 {
688 int n = 0;
689
690 if (size > 8)
691 return AVERROR_INVALIDDATA;
692
693 /* big-endian ordering; build up number */
694 *num = 0;
695 while (n++ < size)
696 *num = (*num << 8) | avio_r8(pb);
697
698 return 0;
699 }
700
701 /*
702 * Read the next element as a float.
703 * 0 is success, < 0 is failure.
704 */
705 static int ebml_read_float(AVIOContext *pb, int size, double *num)
706 {
707 if (size == 0)
708 *num = 0;
709 else if (size == 4)
710 *num = av_int2float(avio_rb32(pb));
711 else if (size == 8)
712 *num = av_int2double(avio_rb64(pb));
713 else
714 return AVERROR_INVALIDDATA;
715
716 return 0;
717 }
718
719 /*
720 * Read the next element as an ASCII string.
721 * 0 is success, < 0 is failure.
722 */
723 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
724 {
725 char *res;
726
727 /* EBML strings are usually not 0-terminated, so we allocate one
728 * byte more, read the string and NULL-terminate it ourselves. */
729 if (!(res = av_malloc(size + 1)))
730 return AVERROR(ENOMEM);
731 if (avio_read(pb, (uint8_t *) res, size) != size) {
732 av_free(res);
733 return AVERROR(EIO);
734 }
735 (res)[size] = '\0';
736 av_free(*str);
737 *str = res;
738
739 return 0;
740 }
741
742 /*
743 * Read the next element as binary data.
744 * 0 is success, < 0 is failure.
745 */
746 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
747 {
748 av_free(bin->data);
749 if (!(bin->data = av_mallocz(length + FF_INPUT_BUFFER_PADDING_SIZE)))
750 return AVERROR(ENOMEM);
751
752 bin->size = length;
753 bin->pos = avio_tell(pb);
754 if (avio_read(pb, bin->data, length) != length) {
755 av_freep(&bin->data);
756 return AVERROR(EIO);
757 }
758
759 return 0;
760 }
761
762 /*
763 * Read the next element, but only the header. The contents
764 * are supposed to be sub-elements which can be read separately.
765 * 0 is success, < 0 is failure.
766 */
767 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
768 {
769 AVIOContext *pb = matroska->ctx->pb;
770 MatroskaLevel *level;
771
772 if (matroska->num_levels >= EBML_MAX_DEPTH) {
773 av_log(matroska->ctx, AV_LOG_ERROR,
774 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
775 return AVERROR(ENOSYS);
776 }
777
778 level = &matroska->levels[matroska->num_levels++];
779 level->start = avio_tell(pb);
780 level->length = length;
781
782 return 0;
783 }
784
785 /*
786 * Read signed/unsigned "EBML" numbers.
787 * Return: number of bytes processed, < 0 on error
788 */
789 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
790 uint8_t *data, uint32_t size, uint64_t *num)
791 {
792 AVIOContext pb;
793 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
794 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
795 }
796
797 /*
798 * Same as above, but signed.
799 */
800 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
801 uint8_t *data, uint32_t size, int64_t *num)
802 {
803 uint64_t unum;
804 int res;
805
806 /* read as unsigned number first */
807 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
808 return res;
809
810 /* make signed (weird way) */
811 *num = unum - ((1LL << (7 * res - 1)) - 1);
812
813 return res;
814 }
815
816 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
817 EbmlSyntax *syntax, void *data);
818
819 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
820 uint32_t id, void *data)
821 {
822 int i;
823 for (i = 0; syntax[i].id; i++)
824 if (id == syntax[i].id)
825 break;
826 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
827 matroska->num_levels > 0 &&
828 matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
829 return 0; // we reached the end of an unknown size cluster
830 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
831 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%"PRIX32"\n", id);
832 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
833 return AVERROR_INVALIDDATA;
834 }
835 return ebml_parse_elem(matroska, &syntax[i], data);
836 }
837
838 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
839 void *data)
840 {
841 if (!matroska->current_id) {
842 uint64_t id;
843 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
844 if (res < 0)
845 return res;
846 matroska->current_id = id | 1 << 7 * res;
847 }
848 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
849 }
850
851 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
852 void *data)
853 {
854 int i, res = 0;
855
856 for (i = 0; syntax[i].id; i++)
857 switch (syntax[i].type) {
858 case EBML_UINT:
859 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
860 break;
861 case EBML_FLOAT:
862 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
863 break;
864 case EBML_STR:
865 case EBML_UTF8:
866 // the default may be NULL
867 if (syntax[i].def.s) {
868 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
869 *dst = av_strdup(syntax[i].def.s);
870 if (!*dst)
871 return AVERROR(ENOMEM);
872 }
873 break;
874 }
875
876 while (!res && !ebml_level_end(matroska))
877 res = ebml_parse(matroska, syntax, data);
878
879 return res;
880 }
881
882 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
883 EbmlSyntax *syntax, void *data)
884 {
885 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
886 [EBML_UINT] = 8,
887 [EBML_FLOAT] = 8,
888 // max. 16 MB for strings
889 [EBML_STR] = 0x1000000,
890 [EBML_UTF8] = 0x1000000,
891 // max. 256 MB for binary data
892 [EBML_BIN] = 0x10000000,
893 // no limits for anything else
894 };
895 AVIOContext *pb = matroska->ctx->pb;
896 uint32_t id = syntax->id;
897 uint64_t length;
898 int res;
899
900 data = (char *) data + syntax->data_offset;
901 if (syntax->list_elem_size) {
902 EbmlList *list = data;
903 if ((res = av_reallocp_array(&list->elem,
904 list->nb_elem + 1,
905 syntax->list_elem_size)) < 0) {
906 list->nb_elem = 0;
907 return res;
908 }
909 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
910 memset(data, 0, syntax->list_elem_size);
911 list->nb_elem++;
912 }
913
914 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
915 matroska->current_id = 0;
916 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
917 return res;
918 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
919 av_log(matroska->ctx, AV_LOG_ERROR,
920 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
921 length, max_lengths[syntax->type], syntax->type);
922 return AVERROR_INVALIDDATA;
923 }
924 }
925
926 switch (syntax->type) {
927 case EBML_UINT:
928 res = ebml_read_uint(pb, length, data);
929 break;
930 case EBML_FLOAT:
931 res = ebml_read_float(pb, length, data);
932 break;
933 case EBML_STR:
934 case EBML_UTF8:
935 res = ebml_read_ascii(pb, length, data);
936 break;
937 case EBML_BIN:
938 res = ebml_read_binary(pb, length, data);
939 break;
940 case EBML_NEST:
941 if ((res = ebml_read_master(matroska, length)) < 0)
942 return res;
943 if (id == MATROSKA_ID_SEGMENT)
944 matroska->segment_start = avio_tell(matroska->ctx->pb);
945 return ebml_parse_nest(matroska, syntax->def.n, data);
946 case EBML_PASS:
947 return ebml_parse_id(matroska, syntax->def.n, id, data);
948 case EBML_STOP:
949 return 1;
950 default:
951 return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
952 }
953 if (res == AVERROR_INVALIDDATA)
954 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
955 else if (res == AVERROR(EIO))
956 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
957 return res;
958 }
959
960 static void ebml_free(EbmlSyntax *syntax, void *data)
961 {
962 int i, j;
963 for (i = 0; syntax[i].id; i++) {
964 void *data_off = (char *) data + syntax[i].data_offset;
965 switch (syntax[i].type) {
966 case EBML_STR:
967 case EBML_UTF8:
968 av_freep(data_off);
969 break;
970 case EBML_BIN:
971 av_freep(&((EbmlBin *) data_off)->data);
972 break;
973 case EBML_NEST:
974 if (syntax[i].list_elem_size) {
975 EbmlList *list = data_off;
976 char *ptr = list->elem;
977 for (j = 0; j < list->nb_elem;
978 j++, ptr += syntax[i].list_elem_size)
979 ebml_free(syntax[i].def.n, ptr);
980 av_free(list->elem);
981 } else
982 ebml_free(syntax[i].def.n, data_off);
983 default:
984 break;
985 }
986 }
987 }
988
989 /*
990 * Autodetecting...
991 */
992 static int matroska_probe(AVProbeData *p)
993 {
994 uint64_t total = 0;
995 int len_mask = 0x80, size = 1, n = 1, i;
996
997 /* EBML header? */
998 if (AV_RB32(p->buf) != EBML_ID_HEADER)
999 return 0;
1000
1001 /* length of header */
1002 total = p->buf[4];
1003 while (size <= 8 && !(total & len_mask)) {
1004 size++;
1005 len_mask >>= 1;
1006 }
1007 if (size > 8)
1008 return 0;
1009 total &= (len_mask - 1);
1010 while (n < size)
1011 total = (total << 8) | p->buf[4 + n++];
1012
1013 /* Does the probe data contain the whole header? */
1014 if (p->buf_size < 4 + size + total)
1015 return 0;
1016
1017 /* The header should contain a known document type. For now,
1018 * we don't parse the whole header but simply check for the
1019 * availability of that array of characters inside the header.
1020 * Not fully fool-proof, but good enough. */
1021 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1022 int probelen = strlen(matroska_doctypes[i]);
1023 if (total < probelen)
1024 continue;
1025 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1026 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1027 return AVPROBE_SCORE_MAX;
1028 }
1029
1030 // probably valid EBML header but no recognized doctype
1031 return AVPROBE_SCORE_EXTENSION;
1032 }
1033
1034 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1035 int num)
1036 {
1037 MatroskaTrack *tracks = matroska->tracks.elem;
1038 int i;
1039
1040 for (i = 0; i < matroska->tracks.nb_elem; i++)
1041 if (tracks[i].num == num)
1042 return &tracks[i];
1043
1044 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1045 return NULL;
1046 }
1047
1048 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1049 MatroskaTrack *track)
1050 {
1051 MatroskaTrackEncoding *encodings = track->encodings.elem;
1052 uint8_t *data = *buf;
1053 int isize = *buf_size;
1054 uint8_t *pkt_data = NULL;
1055 uint8_t av_unused *newpktdata;
1056 int pkt_size = isize;
1057 int result = 0;
1058 int olen;
1059
1060 if (pkt_size >= 10000000)
1061 return AVERROR_INVALIDDATA;
1062
1063 switch (encodings[0].compression.algo) {
1064 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1065 {
1066 int header_size = encodings[0].compression.settings.size;
1067 uint8_t *header = encodings[0].compression.settings.data;
1068
1069 if (!header_size)
1070 return 0;
1071
1072 pkt_size = isize + header_size;
1073 pkt_data = av_malloc(pkt_size);
1074 if (!pkt_data)
1075 return AVERROR(ENOMEM);
1076
1077 memcpy(pkt_data, header, header_size);
1078 memcpy(pkt_data + header_size, data, isize);
1079 break;
1080 }
1081 #if CONFIG_LZO
1082 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1083 do {
1084 olen = pkt_size *= 3;
1085 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1086 if (!newpktdata) {
1087 result = AVERROR(ENOMEM);
1088 goto failed;
1089 }
1090 pkt_data = newpktdata;
1091 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1092 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1093 if (result) {
1094 result = AVERROR_INVALIDDATA;
1095 goto failed;
1096 }
1097 pkt_size -= olen;
1098 break;
1099 #endif
1100 #if CONFIG_ZLIB
1101 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1102 {
1103 z_stream zstream = { 0 };
1104 if (inflateInit(&zstream) != Z_OK)
1105 return -1;
1106 zstream.next_in = data;
1107 zstream.avail_in = isize;
1108 do {
1109 pkt_size *= 3;
1110 newpktdata = av_realloc(pkt_data, pkt_size);
1111 if (!newpktdata) {
1112 inflateEnd(&zstream);
1113 goto failed;
1114 }
1115 pkt_data = newpktdata;
1116 zstream.avail_out = pkt_size - zstream.total_out;
1117 zstream.next_out = pkt_data + zstream.total_out;
1118 result = inflate(&zstream, Z_NO_FLUSH);
1119 } while (result == Z_OK && pkt_size < 10000000);
1120 pkt_size = zstream.total_out;
1121 inflateEnd(&zstream);
1122 if (result != Z_STREAM_END) {
1123 if (result == Z_MEM_ERROR)
1124 result = AVERROR(ENOMEM);
1125 else
1126 result = AVERROR_INVALIDDATA;
1127 goto failed;
1128 }
1129 break;
1130 }
1131 #endif
1132 #if CONFIG_BZLIB
1133 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1134 {
1135 bz_stream bzstream = { 0 };
1136 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1137 return -1;
1138 bzstream.next_in = data;
1139 bzstream.avail_in = isize;
1140 do {
1141 pkt_size *= 3;
1142 newpktdata = av_realloc(pkt_data, pkt_size);
1143 if (!newpktdata) {
1144 BZ2_bzDecompressEnd(&bzstream);
1145 goto failed;
1146 }
1147 pkt_data = newpktdata;
1148 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1149 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1150 result = BZ2_bzDecompress(&bzstream);
1151 } while (result == BZ_OK && pkt_size < 10000000);
1152 pkt_size = bzstream.total_out_lo32;
1153 BZ2_bzDecompressEnd(&bzstream);
1154 if (result != BZ_STREAM_END) {
1155 if (result == BZ_MEM_ERROR)
1156 result = AVERROR(ENOMEM);
1157 else
1158 result = AVERROR_INVALIDDATA;
1159 goto failed;
1160 }
1161 break;
1162 }
1163 #endif
1164 default:
1165 return AVERROR_INVALIDDATA;
1166 }
1167
1168 *buf = pkt_data;
1169 *buf_size = pkt_size;
1170 return 0;
1171
1172 failed:
1173 av_free(pkt_data);
1174 return result;
1175 }
1176
1177 static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
1178 AVPacket *pkt, uint64_t display_duration)
1179 {
1180 AVBufferRef *line;
1181 char *layer, *ptr = pkt->data, *end = ptr + pkt->size;
1182
1183 for (; *ptr != ',' && ptr < end - 1; ptr++)
1184 ;
1185 if (*ptr == ',')
1186 layer = ++ptr;
1187 for (; *ptr != ',' && ptr < end - 1; ptr++)
1188 ;
1189 if (*ptr == ',') {
1190 int64_t end_pts = pkt->pts + display_duration;
1191 int sc = matroska->time_scale * pkt->pts / 10000000;
1192 int ec = matroska->time_scale * end_pts / 10000000;
1193 int sh, sm, ss, eh, em, es, len;
1194 sh = sc / 360000;
1195 sc -= 360000 * sh;
1196 sm = sc / 6000;
1197 sc -= 6000 * sm;
1198 ss = sc / 100;
1199 sc -= 100 * ss;
1200 eh = ec / 360000;
1201 ec -= 360000 * eh;
1202 em = ec / 6000;
1203 ec -= 6000 * em;
1204 es = ec / 100;
1205 ec -= 100 * es;
1206 *ptr++ = '\0';
1207 len = 50 + end - ptr + FF_INPUT_BUFFER_PADDING_SIZE;
1208 if (!(line = av_buffer_alloc(len)))
1209 return;
1210 snprintf(line->data, len,
1211 "Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
1212 layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
1213 av_buffer_unref(&pkt->buf);
1214 pkt->buf = line;
1215 pkt->data = line->data;
1216 pkt->size = strlen(line->data);
1217 }
1218 }
1219
1220 static int matroska_merge_packets(AVPacket *out, AVPacket *in)
1221 {
1222 int old_size = out->size;
1223 int ret = av_grow_packet(out, in->size);
1224 if (ret < 0)
1225 return ret;
1226
1227 memcpy(out->data + old_size, in->data, in->size);
1228
1229 av_free_packet(in);
1230 av_free(in);
1231 return 0;
1232 }
1233
1234 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1235 AVDictionary **metadata, char *prefix)
1236 {
1237 MatroskaTag *tags = list->elem;
1238 char key[1024];
1239 int i;
1240
1241 for (i = 0; i < list->nb_elem; i++) {
1242 const char *lang = tags[i].lang &&
1243 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1244
1245 if (!tags[i].name) {
1246 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1247 continue;
1248 }
1249 if (prefix)
1250 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1251 else
1252 av_strlcpy(key, tags[i].name, sizeof(key));
1253 if (tags[i].def || !lang) {
1254 av_dict_set(metadata, key, tags[i].string, 0);
1255 if (tags[i].sub.nb_elem)
1256 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1257 }
1258 if (lang) {
1259 av_strlcat(key, "-", sizeof(key));
1260 av_strlcat(key, lang, sizeof(key));
1261 av_dict_set(metadata, key, tags[i].string, 0);
1262 if (tags[i].sub.nb_elem)
1263 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1264 }
1265 }
1266 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1267 }
1268
1269 static void matroska_convert_tags(AVFormatContext *s)
1270 {
1271 MatroskaDemuxContext *matroska = s->priv_data;
1272 MatroskaTags *tags = matroska->tags.elem;
1273 int i, j;
1274
1275 for (i = 0; i < matroska->tags.nb_elem; i++) {
1276 if (tags[i].target.attachuid) {
1277 MatroskaAttachment *attachment = matroska->attachments.elem;
1278 for (j = 0; j < matroska->attachments.nb_elem; j++)
1279 if (attachment[j].uid == tags[i].target.attachuid &&
1280 attachment[j].stream)
1281 matroska_convert_tag(s, &tags[i].tag,
1282 &attachment[j].stream->metadata, NULL);
1283 } else if (tags[i].target.chapteruid) {
1284 MatroskaChapter *chapter = matroska->chapters.elem;
1285 for (j = 0; j < matroska->chapters.nb_elem; j++)
1286 if (chapter[j].uid == tags[i].target.chapteruid &&
1287 chapter[j].chapter)
1288 matroska_convert_tag(s, &tags[i].tag,
1289 &chapter[j].chapter->metadata, NULL);
1290 } else if (tags[i].target.trackuid) {
1291 MatroskaTrack *track = matroska->tracks.elem;
1292 for (j = 0; j < matroska->tracks.nb_elem; j++)
1293 if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1294 matroska_convert_tag(s, &tags[i].tag,
1295 &track[j].stream->metadata, NULL);
1296 } else {
1297 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1298 tags[i].target.type);
1299 }
1300 }
1301 }
1302
1303 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1304 int idx)
1305 {
1306 EbmlList *seekhead_list = &matroska->seekhead;
1307 uint32_t level_up = matroska->level_up;
1308 uint32_t saved_id = matroska->current_id;
1309 MatroskaSeekhead *seekhead = seekhead_list->elem;
1310 int64_t before_pos = avio_tell(matroska->ctx->pb);
1311 MatroskaLevel level;
1312 int64_t offset;
1313 int ret = 0;
1314
1315 if (idx >= seekhead_list->nb_elem ||
1316 seekhead[idx].id == MATROSKA_ID_SEEKHEAD ||
1317 seekhead[idx].id == MATROSKA_ID_CLUSTER)
1318 return 0;
1319
1320 /* seek */
1321 offset = seekhead[idx].pos + matroska->segment_start;
1322 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1323 /* We don't want to lose our seekhead level, so we add
1324 * a dummy. This is a crude hack. */
1325 if (matroska->num_levels == EBML_MAX_DEPTH) {
1326 av_log(matroska->ctx, AV_LOG_INFO,
1327 "Max EBML element depth (%d) reached, "
1328 "cannot parse further.\n", EBML_MAX_DEPTH);
1329 ret = AVERROR_INVALIDDATA;
1330 } else {
1331 level.start = 0;
1332 level.length = (uint64_t) -1;
1333 matroska->levels[matroska->num_levels] = level;
1334 matroska->num_levels++;
1335 matroska->current_id = 0;
1336
1337 ret = ebml_parse(matroska, matroska_segment, matroska);
1338
1339 /* remove dummy level */
1340 while (matroska->num_levels) {
1341 uint64_t length = matroska->levels[--matroska->num_levels].length;
1342 if (length == (uint64_t) -1)
1343 break;
1344 }
1345 }
1346 }
1347 /* seek back */
1348 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1349 matroska->level_up = level_up;
1350 matroska->current_id = saved_id;
1351
1352 return ret;
1353 }
1354
1355 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1356 {
1357 EbmlList *seekhead_list = &matroska->seekhead;
1358 int64_t before_pos = avio_tell(matroska->ctx->pb);
1359 int i;
1360
1361 // we should not do any seeking in the streaming case
1362 if (!matroska->ctx->pb->seekable ||
1363 (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1364 return;
1365
1366 for (i = 0; i < seekhead_list->nb_elem; i++) {
1367 MatroskaSeekhead *seekhead = seekhead_list->elem;
1368 if (seekhead[i].pos <= before_pos)
1369 continue;
1370
1371 // defer cues parsing until we actually need cue data.
1372 if (seekhead[i].id == MATROSKA_ID_CUES) {
1373 matroska->cues_parsing_deferred = 1;
1374 continue;
1375 }
1376
1377 if (matroska_parse_seekhead_entry(matroska, i) < 0)
1378 break;
1379 }
1380 }
1381
1382 static void matroska_parse_cues(MatroskaDemuxContext *matroska)
1383 {
1384 EbmlList *seekhead_list = &matroska->seekhead;
1385 MatroskaSeekhead *seekhead = seekhead_list->elem;
1386 EbmlList *index_list;
1387 MatroskaIndex *index;
1388 int index_scale = 1;
1389 int i, j;
1390
1391 for (i = 0; i < seekhead_list->nb_elem; i++)
1392 if (seekhead[i].id == MATROSKA_ID_CUES)
1393 break;
1394 assert(i <= seekhead_list->nb_elem);
1395
1396 matroska_parse_seekhead_entry(matroska, i);
1397
1398 index_list = &matroska->index;
1399 index = index_list->elem;
1400 if (index_list->nb_elem &&
1401 index[0].time > 1E14 / matroska->time_scale) {
1402 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1403 index_scale = matroska->time_scale;
1404 }
1405 for (i = 0; i < index_list->nb_elem; i++) {
1406 EbmlList *pos_list = &index[i].pos;
1407 MatroskaIndexPos *pos = pos_list->elem;
1408 for (j = 0; j < pos_list->nb_elem; j++) {
1409 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1410 pos[j].track);
1411 if (track && track->stream)
1412 av_add_index_entry(track->stream,
1413 pos[j].pos + matroska->segment_start,
1414 index[i].time / index_scale, 0, 0,
1415 AVINDEX_KEYFRAME);
1416 }
1417 }
1418 }
1419
1420 static int matroska_aac_profile(char *codec_id)
1421 {
1422 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1423 int profile;
1424
1425 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1426 if (strstr(codec_id, aac_profiles[profile]))
1427 break;
1428 return profile + 1;
1429 }
1430
1431 static int matroska_aac_sri(int samplerate)
1432 {
1433 int sri;
1434
1435 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1436 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1437 break;
1438 return sri;
1439 }
1440
1441 static int matroska_parse_flac(AVFormatContext *s,
1442 MatroskaTrack *track,
1443 int *offset)
1444 {
1445 AVStream *st = track->stream;
1446 uint8_t *p = track->codec_priv.data;
1447 int size = track->codec_priv.size;
1448
1449 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1450 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1451 track->codec_priv.size = 0;
1452 return 0;
1453 }
1454 *offset = 8;
1455 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1456
1457 p += track->codec_priv.size;
1458 size -= track->codec_priv.size;
1459
1460 /* parse the remaining metadata blocks if present */
1461 while (size >= 4) {
1462 int block_last, block_type, block_size;
1463
1464 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1465
1466 p += 4;
1467 size -= 4;
1468 if (block_size > size)
1469 return 0;
1470
1471 /* check for the channel mask */
1472 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1473 AVDictionary *dict = NULL;
1474 AVDictionaryEntry *chmask;
1475
1476 ff_vorbis_comment(s, &dict, p, block_size, 0);
1477 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1478 if (chmask) {
1479 uint64_t mask = strtol(chmask->value, NULL, 0);
1480 if (!mask || mask & ~0x3ffffULL) {
1481 av_log(s, AV_LOG_WARNING,
1482 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1483 } else
1484 st->codec->channel_layout = mask;
1485 }
1486 av_dict_free(&dict);
1487 }
1488
1489 p += block_size;
1490 size -= block_size;
1491 }
1492
1493 return 0;
1494 }
1495
1496 static int matroska_parse_tracks(AVFormatContext *s)
1497 {
1498 MatroskaDemuxContext *matroska = s->priv_data;
1499 MatroskaTrack *tracks = matroska->tracks.elem;
1500 AVStream *st;
1501 int i, j, ret;
1502
1503 for (i = 0; i < matroska->tracks.nb_elem; i++) {
1504 MatroskaTrack *track = &tracks[i];
1505 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1506 EbmlList *encodings_list = &track->encodings;
1507 MatroskaTrackEncoding *encodings = encodings_list->elem;
1508 uint8_t *extradata = NULL;
1509 int extradata_size = 0;
1510 int extradata_offset = 0;
1511 AVIOContext b;
1512
1513 /* Apply some sanity checks. */
1514 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1515 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1516 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
1517 av_log(matroska->ctx, AV_LOG_INFO,
1518 "Unknown or unsupported track type %"PRIu64"\n",
1519 track->type);
1520 continue;
1521 }
1522 if (track->codec_id == NULL)
1523 continue;
1524
1525 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1526 if (!track->default_duration && track->video.frame_rate > 0)
1527 track->default_duration = 1000000000 / track->video.frame_rate;
1528 if (!track->video.display_width)
1529 track->video.display_width = track->video.pixel_width;
1530 if (!track->video.display_height)
1531 track->video.display_height = track->video.pixel_height;
1532 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1533 if (!track->audio.out_samplerate)
1534 track->audio.out_samplerate = track->audio.samplerate;
1535 }
1536 if (encodings_list->nb_elem > 1) {
1537 av_log(matroska->ctx, AV_LOG_ERROR,
1538 "Multiple combined encodings not supported");
1539 } else if (encodings_list->nb_elem == 1) {
1540 if (encodings[0].type ||
1541 (
1542 #if CONFIG_ZLIB
1543 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1544 #endif
1545 #if CONFIG_BZLIB
1546 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1547 #endif
1548 #if CONFIG_LZO
1549 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
1550 #endif
1551 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP)) {
1552 encodings[0].scope = 0;
1553 av_log(matroska->ctx, AV_LOG_ERROR,
1554 "Unsupported encoding type");
1555 } else if (track->codec_priv.size && encodings[0].scope & 2) {
1556 uint8_t *codec_priv = track->codec_priv.data;
1557 int ret = matroska_decode_buffer(&track->codec_priv.data,
1558 &track->codec_priv.size,
1559 track);
1560 if (ret < 0) {
1561 track->codec_priv.data = NULL;
1562 track->codec_priv.size = 0;
1563 av_log(matroska->ctx, AV_LOG_ERROR,
1564 "Failed to decode codec private data\n");
1565 }
1566
1567 if (codec_priv != track->codec_priv.data)
1568 av_free(codec_priv);
1569 }
1570 }
1571
1572 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
1573 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1574 strlen(ff_mkv_codec_tags[j].str))) {
1575 codec_id = ff_mkv_codec_tags[j].id;
1576 break;
1577 }
1578 }
1579
1580 st = track->stream = avformat_new_stream(s, NULL);
1581 if (st == NULL)
1582 return AVERROR(ENOMEM);
1583
1584 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
1585 track->codec_priv.size >= 40 &&
1586 track->codec_priv.data != NULL) {
1587 track->ms_compat = 1;
1588 track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
1589 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
1590 track->video.fourcc);
1591 extradata_offset = 40;
1592 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
1593 track->codec_priv.size >= 14 &&
1594 track->codec_priv.data != NULL) {
1595 int ret;
1596 ffio_init_context(&b, track->codec_priv.data,
1597 track->codec_priv.size,
1598 0, NULL, NULL, NULL, NULL);
1599 ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1600 if (ret < 0)
1601 return ret;
1602 codec_id = st->codec->codec_id;
1603 extradata_offset = FFMIN(track->codec_priv.size, 18);
1604 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
1605 (track->codec_priv.size >= 86) &&
1606 (track->codec_priv.data != NULL)) {
1607 track->video.fourcc = AV_RL32(track->codec_priv.data);
1608 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
1609 track->video.fourcc);
1610 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1611 switch (track->audio.bitdepth) {
1612 case 8:
1613 codec_id = AV_CODEC_ID_PCM_U8;
1614 break;
1615 case 24:
1616 codec_id = AV_CODEC_ID_PCM_S24BE;
1617 break;
1618 case 32:
1619 codec_id = AV_CODEC_ID_PCM_S32BE;
1620 break;
1621 }
1622 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1623 switch (track->audio.bitdepth) {
1624 case 8:
1625 codec_id = AV_CODEC_ID_PCM_U8;
1626 break;
1627 case 24:
1628 codec_id = AV_CODEC_ID_PCM_S24LE;
1629 break;
1630 case 32:
1631 codec_id = AV_CODEC_ID_PCM_S32LE;
1632 break;
1633 }
1634 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
1635 track->audio.bitdepth == 64) {
1636 codec_id = AV_CODEC_ID_PCM_F64LE;
1637 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1638 int profile = matroska_aac_profile(track->codec_id);
1639 int sri = matroska_aac_sri(track->audio.samplerate);
1640 extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1641 if (extradata == NULL)
1642 return AVERROR(ENOMEM);
1643 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
1644 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
1645 if (strstr(track->codec_id, "SBR")) {
1646 sri = matroska_aac_sri(track->audio.out_samplerate);
1647 extradata[2] = 0x56;
1648 extradata[3] = 0xE5;
1649 extradata[4] = 0x80 | (sri << 3);
1650 extradata_size = 5;
1651 } else
1652 extradata_size = 2;
1653 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size) {
1654 /* Only ALAC's magic cookie is stored in Matroska's track headers.
1655 * Create the "atom size", "tag", and "tag version" fields the
1656 * decoder expects manually. */
1657 extradata_size = 12 + track->codec_priv.size;
1658 extradata = av_mallocz(extradata_size +
1659 FF_INPUT_BUFFER_PADDING_SIZE);
1660 if (extradata == NULL)
1661 return AVERROR(ENOMEM);
1662 AV_WB32(extradata, extradata_size);
1663 memcpy(&extradata[4], "alac", 4);
1664 AV_WB32(&extradata[8], 0);
1665 memcpy(&extradata[12], track->codec_priv.data,
1666 track->codec_priv.size);
1667 } else if (codec_id == AV_CODEC_ID_TTA) {
1668 extradata_size = 30;
1669 extradata = av_mallocz(extradata_size);
1670 if (extradata == NULL)
1671 return AVERROR(ENOMEM);
1672 ffio_init_context(&b, extradata, extradata_size, 1,
1673 NULL, NULL, NULL, NULL);
1674 avio_write(&b, "TTA1", 4);
1675 avio_wl16(&b, 1);
1676 avio_wl16(&b, track->audio.channels);
1677 avio_wl16(&b, track->audio.bitdepth);
1678 avio_wl32(&b, track->audio.out_samplerate);
1679 avio_wl32(&b, matroska->ctx->duration *
1680 track->audio.out_samplerate);
1681 } else if (codec_id == AV_CODEC_ID_RV10 ||
1682 codec_id == AV_CODEC_ID_RV20 ||
1683 codec_id == AV_CODEC_ID_RV30 ||
1684 codec_id == AV_CODEC_ID_RV40) {
1685 extradata_offset = 26;
1686 } else if (codec_id == AV_CODEC_ID_RA_144) {
1687 track->audio.out_samplerate = 8000;
1688 track->audio.channels = 1;
1689 } else if (codec_id == AV_CODEC_ID_RA_288 ||
1690 codec_id == AV_CODEC_ID_COOK ||
1691 codec_id == AV_CODEC_ID_ATRAC3 ||
1692 codec_id == AV_CODEC_ID_SIPR) {
1693 int flavor;
1694 ffio_init_context(&b, track->codec_priv.data,
1695 track->codec_priv.size,
1696 0, NULL, NULL, NULL, NULL);
1697 avio_skip(&b, 22);
1698 flavor = avio_rb16(&b);
1699 track->audio.coded_framesize = avio_rb32(&b);
1700 avio_skip(&b, 12);
1701 track->audio.sub_packet_h = avio_rb16(&b);
1702 track->audio.frame_size = avio_rb16(&b);
1703 track->audio.sub_packet_size = avio_rb16(&b);
1704 if (flavor <= 0 ||
1705 track->audio.coded_framesize <= 0 ||
1706 track->audio.sub_packet_h <= 0 ||
1707 track->audio.frame_size <= 0 ||
1708 track->audio.sub_packet_size <= 0)
1709 return AVERROR_INVALIDDATA;
1710 track->audio.buf = av_malloc(track->audio.frame_size *
1711 track->audio.sub_packet_h);
1712 if (codec_id == AV_CODEC_ID_RA_288) {
1713 st->codec->block_align = track->audio.coded_framesize;
1714 track->codec_priv.size = 0;
1715 } else {
1716 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1717 const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1718 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1719 st->codec->bit_rate = sipr_bit_rate[flavor];
1720 }
1721 st->codec->block_align = track->audio.sub_packet_size;
1722 extradata_offset = 78;
1723 }
1724 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
1725 ret = matroska_parse_flac(s, track, &extradata_offset);
1726 if (ret < 0)
1727 return ret;
1728 }
1729 track->codec_priv.size -= extradata_offset;
1730
1731 if (codec_id == AV_CODEC_ID_NONE)
1732 av_log(matroska->ctx, AV_LOG_INFO,
1733 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1734
1735 if (track->time_scale < 0.01)
1736 track->time_scale = 1.0;
1737 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
1738 1000 * 1000 * 1000); /* 64 bit pts in ns */
1739
1740 /* convert the delay from ns to the track timebase */
1741 track->codec_delay = av_rescale_q(track->codec_delay,
1742 (AVRational){ 1, 1000000000 },
1743 st->time_base);
1744
1745 st->codec->codec_id = codec_id;
1746 st->start_time = 0;
1747 if (strcmp(track->language, "und"))
1748 av_dict_set(&st->metadata, "language", track->language, 0);
1749 av_dict_set(&st->metadata, "title", track->name, 0);
1750
1751 if (track->flag_default)
1752 st->disposition |= AV_DISPOSITION_DEFAULT;
1753 if (track->flag_forced)
1754 st->disposition |= AV_DISPOSITION_FORCED;
1755
1756 if (!st->codec->extradata) {
1757 if (extradata) {
1758 st->codec->extradata = extradata;
1759 st->codec->extradata_size = extradata_size;
1760 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
1761 st->codec->extradata = av_mallocz(track->codec_priv.size +
1762 FF_INPUT_BUFFER_PADDING_SIZE);
1763 if (st->codec->extradata == NULL)
1764 return AVERROR(ENOMEM);
1765 st->codec->extradata_size = track->codec_priv.size;
1766 memcpy(st->codec->extradata,
1767 track->codec_priv.data + extradata_offset,
1768 track->codec_priv.size);
1769 }
1770 }
1771
1772 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1773 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1774 st->codec->codec_tag = track->video.fourcc;
1775 st->codec->width = track->video.pixel_width;
1776 st->codec->height = track->video.pixel_height;
1777 av_reduce(&st->sample_aspect_ratio.num,
1778 &st->sample_aspect_ratio.den,
1779 st->codec->height * track->video.display_width,
1780 st->codec->width * track->video.display_height,
1781 255);
1782 if (st->codec->codec_id != AV_CODEC_ID_H264 &&
1783 st->codec->codec_id != AV_CODEC_ID_HEVC)
1784 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1785 if (track->default_duration) {
1786 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1787 1000000000, track->default_duration, 30000);
1788 }
1789 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1790 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1791 st->codec->sample_rate = track->audio.out_samplerate;
1792 st->codec->channels = track->audio.channels;
1793 if (st->codec->codec_id != AV_CODEC_ID_AAC)
1794 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1795 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1796 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1797 if (st->codec->codec_id == AV_CODEC_ID_SSA)
1798 matroska->contains_ssa = 1;
1799 }
1800 }
1801
1802 return 0;
1803 }
1804
1805 static int matroska_read_header(AVFormatContext *s)
1806 {
1807 MatroskaDemuxContext *matroska = s->priv_data;
1808 EbmlList *attachments_list = &matroska->attachments;
1809 EbmlList *chapters_list = &matroska->chapters;
1810 MatroskaAttachment *attachments;
1811 MatroskaChapter *chapters;
1812 uint64_t max_start = 0;
1813 int64_t pos;
1814 Ebml ebml = { 0 };
1815 int i, j, res;
1816
1817 matroska->ctx = s;
1818
1819 /* First read the EBML header. */
1820 if (ebml_parse(matroska, ebml_syntax, &ebml) ||
1821 ebml.version > EBML_VERSION ||
1822 ebml.max_size > sizeof(uint64_t) ||
1823 ebml.id_length > sizeof(uint32_t) ||
1824 ebml.doctype_version > 2) {
1825 av_log(matroska->ctx, AV_LOG_ERROR,
1826 "EBML header using unsupported features\n"
1827 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
1828 ebml.version, ebml.doctype, ebml.doctype_version);
1829 ebml_free(ebml_syntax, &ebml);
1830 return AVERROR_PATCHWELCOME;
1831 }
1832 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
1833 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
1834 break;
1835 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
1836 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
1837 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
1838 ebml_free(ebml_syntax, &ebml);
1839 return AVERROR_INVALIDDATA;
1840 }
1841 }
1842 ebml_free(ebml_syntax, &ebml);
1843
1844 /* The next thing is a segment. */
1845 pos = avio_tell(matroska->ctx->pb);
1846 res = ebml_parse(matroska, matroska_segments, matroska);
1847 // try resyncing until we find a EBML_STOP type element.
1848 while (res != 1) {
1849 res = matroska_resync(matroska, pos);
1850 if (res < 0)
1851 return res;
1852 pos = avio_tell(matroska->ctx->pb);
1853 res = ebml_parse(matroska, matroska_segment, matroska);
1854 }
1855 matroska_execute_seekhead(matroska);
1856
1857 if (!matroska->time_scale)
1858 matroska->time_scale = 1000000;
1859 if (matroska->duration)
1860 matroska->ctx->duration = matroska->duration * matroska->time_scale *
1861 1000 / AV_TIME_BASE;
1862 av_dict_set(&s->metadata, "title", matroska->title, 0);
1863
1864 res = matroska_parse_tracks(s);
1865 if (res < 0)
1866 return res;
1867
1868 attachments = attachments_list->elem;
1869 for (j = 0; j < attachments_list->nb_elem; j++) {
1870 if (!(attachments[j].filename && attachments[j].mime &&
1871 attachments[j].bin.data && attachments[j].bin.size > 0)) {
1872 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
1873 } else {
1874 AVStream *st = avformat_new_stream(s, NULL);
1875 if (st == NULL)
1876 break;
1877 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
1878 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
1879 st->codec->codec_id = AV_CODEC_ID_NONE;
1880 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
1881 st->codec->extradata = av_malloc(attachments[j].bin.size);
1882 if (st->codec->extradata == NULL)
1883 break;
1884 st->codec->extradata_size = attachments[j].bin.size;
1885 memcpy(st->codec->extradata, attachments[j].bin.data,
1886 attachments[j].bin.size);
1887
1888 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
1889 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
1890 strlen(ff_mkv_mime_tags[i].str))) {
1891 st->codec->codec_id = ff_mkv_mime_tags[i].id;
1892 break;
1893 }
1894 }
1895 attachments[j].stream = st;
1896 }
1897 }
1898
1899 chapters = chapters_list->elem;
1900 for (i = 0; i < chapters_list->nb_elem; i++)
1901 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
1902 (max_start == 0 || chapters[i].start > max_start)) {
1903 chapters[i].chapter =
1904 avpriv_new_chapter(s, chapters[i].uid,
1905 (AVRational) { 1, 1000000000 },
1906 chapters[i].start, chapters[i].end,
1907 chapters[i].title);
1908 av_dict_set(&chapters[i].chapter->metadata,
1909 "title", chapters[i].title, 0);
1910 max_start = chapters[i].start;
1911 }
1912
1913 matroska_convert_tags(s);
1914
1915 return 0;
1916 }
1917
1918 /*
1919 * Put one packet in an application-supplied AVPacket struct.
1920 * Returns 0 on success or -1 on failure.
1921 */
1922 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
1923 AVPacket *pkt)
1924 {
1925 if (matroska->num_packets > 0) {
1926 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
1927 av_free(matroska->packets[0]);
1928 if (matroska->num_packets > 1) {
1929 void *newpackets;
1930 memmove(&matroska->packets[0], &matroska->packets[1],
1931 (matroska->num_packets - 1) * sizeof(AVPacket *));
1932 newpackets = av_realloc(matroska->packets,
1933 (matroska->num_packets - 1) *
1934 sizeof(AVPacket *));
1935 if (newpackets)
1936 matroska->packets = newpackets;
1937 } else {
1938 av_freep(&matroska->packets);
1939 matroska->prev_pkt = NULL;
1940 }
1941 matroska->num_packets--;
1942 return 0;
1943 }
1944
1945 return -1;
1946 }
1947
1948 /*
1949 * Free all packets in our internal queue.
1950 */
1951 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
1952 {
1953 matroska->prev_pkt = NULL;
1954 if (matroska->packets) {
1955 int n;
1956 for (n = 0; n < matroska->num_packets; n++) {
1957 av_free_packet(matroska->packets[n]);
1958 av_free(matroska->packets[n]);
1959 }
1960 av_freep(&matroska->packets);
1961 matroska->num_packets = 0;
1962 }
1963 }
1964
1965 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
1966 int *buf_size, int type,
1967 uint32_t **lace_buf, int *laces)
1968 {
1969 int res = 0, n, size = *buf_size;
1970 uint8_t *data = *buf;
1971 uint32_t *lace_size;
1972
1973 if (!type) {
1974 *laces = 1;
1975 *lace_buf = av_mallocz(sizeof(int));
1976 if (!*lace_buf)
1977 return AVERROR(ENOMEM);
1978
1979 *lace_buf[0] = size;
1980 return 0;
1981 }
1982
1983 assert(size > 0);
1984 *laces = *data + 1;
1985 data += 1;
1986 size -= 1;
1987 lace_size = av_mallocz(*laces * sizeof(int));
1988 if (!lace_size)
1989 return AVERROR(ENOMEM);
1990
1991 switch (type) {
1992 case 0x1: /* Xiph lacing */
1993 {
1994 uint8_t temp;
1995 uint32_t total = 0;
1996 for (n = 0; res == 0 && n < *laces - 1; n++) {
1997 while (1) {
1998 if (size == 0) {
1999 res = AVERROR_EOF;
2000 break;
2001 }
2002 temp = *data;
2003 lace_size[n] += temp;
2004 data += 1;
2005 size -= 1;
2006 if (temp != 0xff)
2007 break;
2008 }
2009 total += lace_size[n];
2010 }
2011 if (size <= total) {
2012 res = AVERROR_INVALIDDATA;
2013 break;
2014 }
2015
2016 lace_size[n] = size - total;
2017 break;
2018 }
2019
2020 case 0x2: /* fixed-size lacing */
2021 if (size % (*laces)) {
2022 res = AVERROR_INVALIDDATA;
2023 break;
2024 }
2025 for (n = 0; n < *laces; n++)
2026 lace_size[n] = size / *laces;
2027 break;
2028
2029 case 0x3: /* EBML lacing */
2030 {
2031 uint64_t num;
2032 uint64_t total;
2033 n = matroska_ebmlnum_uint(matroska, data, size, &num);
2034 if (n < 0) {
2035 av_log(matroska->ctx, AV_LOG_INFO,
2036 "EBML block data error\n");
2037 res = n;
2038 break;
2039 }
2040 data += n;
2041 size -= n;
2042 total = lace_size[0] = num;
2043 for (n = 1; res == 0 && n < *laces - 1; n++) {
2044 int64_t snum;
2045 int r;
2046 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2047 if (r < 0) {
2048 av_log(matroska->ctx, AV_LOG_INFO,
2049 "EBML block data error\n");
2050 res = r;
2051 break;
2052 }
2053 data += r;
2054 size -= r;
2055 lace_size[n] = lace_size[n - 1] + snum;
2056 total += lace_size[n];
2057 }
2058 if (size <= total) {
2059 res = AVERROR_INVALIDDATA;
2060 break;
2061 }
2062 lace_size[*laces - 1] = size - total;
2063 break;
2064 }
2065 }
2066
2067 *buf = data;
2068 *lace_buf = lace_size;
2069 *buf_size = size;
2070
2071 return res;
2072 }
2073
2074 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2075 MatroskaTrack *track, AVStream *st,
2076 uint8_t *data, int size, uint64_t timecode,
2077 uint64_t duration, int64_t pos)
2078 {
2079 int a = st->codec->block_align;
2080 int sps = track->audio.sub_packet_size;
2081 int cfs = track->audio.coded_framesize;
2082 int h = track->audio.sub_packet_h;
2083 int y = track->audio.sub_packet_cnt;
2084 int w = track->audio.frame_size;
2085 int x;
2086
2087 if (!track->audio.pkt_cnt) {
2088 if (track->audio.sub_packet_cnt == 0)
2089 track->audio.buf_timecode = timecode;
2090 if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2091 if (size < cfs * h / 2) {
2092 av_log(matroska->ctx, AV_LOG_ERROR,
2093 "Corrupt int4 RM-style audio packet size\n");
2094 return AVERROR_INVALIDDATA;
2095 }
2096 for (x = 0; x < h / 2; x++)
2097 memcpy(track->audio.buf + x * 2 * w + y * cfs,
2098 data + x * cfs, cfs);
2099 } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2100 if (size < w) {
2101 av_log(matroska->ctx, AV_LOG_ERROR,
2102 "Corrupt sipr RM-style audio packet size\n");
2103 return AVERROR_INVALIDDATA;
2104 }
2105 memcpy(track->audio.buf + y * w, data, w);
2106 } else {
2107 if (size < sps * w / sps) {
2108 av_log(matroska->ctx, AV_LOG_ERROR,
2109 "Corrupt generic RM-style audio packet size\n");
2110 return AVERROR_INVALIDDATA;
2111 }
2112 for (x = 0; x < w / sps; x++)
2113 memcpy(track->audio.buf +
2114 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
2115 data + x * sps, sps);
2116 }
2117
2118 if (++track->audio.sub_packet_cnt >= h) {
2119 if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2120 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2121 track->audio.sub_packet_cnt = 0;
2122 track->audio.pkt_cnt = h * w / a;
2123 }
2124 }
2125
2126 while (track->audio.pkt_cnt) {
2127 AVPacket *pkt = av_mallocz(sizeof(AVPacket));
2128 av_new_packet(pkt, a);
2129 memcpy(pkt->data,
2130 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
2131 a);
2132 pkt->pts = track->audio.buf_timecode;
2133 track->audio.buf_timecode = AV_NOPTS_VALUE;
2134 pkt->pos = pos;
2135 pkt->stream_index = st->index;
2136 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2137 }
2138
2139 return 0;
2140 }
2141
2142 /* reconstruct full wavpack blocks from mangled matroska ones */
2143 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
2144 uint8_t **pdst, int *size)
2145 {
2146 uint8_t *dst = NULL;
2147 int dstlen = 0;
2148 int srclen = *size;
2149 uint32_t samples;
2150 uint16_t ver;
2151 int ret, offset = 0;
2152
2153 if (srclen < 12 || track->stream->codec->extradata_size < 2)
2154 return AVERROR_INVALIDDATA;
2155
2156 ver = AV_RL16(track->stream->codec->extradata);
2157
2158 samples = AV_RL32(src);
2159 src += 4;
2160 srclen -= 4;
2161
2162 while (srclen >= 8) {
2163 int multiblock;
2164 uint32_t blocksize;
2165 uint8_t *tmp;
2166
2167 uint32_t flags = AV_RL32(src);
2168 uint32_t crc = AV_RL32(src + 4);
2169 src += 8;
2170 srclen -= 8;
2171
2172 multiblock = (flags & 0x1800) != 0x1800;
2173 if (multiblock) {
2174 if (srclen < 4) {
2175 ret = AVERROR_INVALIDDATA;
2176 goto fail;
2177 }
2178 blocksize = AV_RL32(src);
2179 src += 4;
2180 srclen -= 4;
2181 } else
2182 blocksize = srclen;
2183
2184 if (blocksize > srclen) {
2185 ret = AVERROR_INVALIDDATA;
2186 goto fail;
2187 }
2188
2189 tmp = av_realloc(dst, dstlen + blocksize + 32);
2190 if (!tmp) {
2191 ret = AVERROR(ENOMEM);
2192 goto fail;
2193 }
2194 dst = tmp;
2195 dstlen += blocksize + 32;
2196
2197 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2198 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2199 AV_WL16(dst + offset + 8, ver); // version
2200 AV_WL16(dst + offset + 10, 0); // track/index_no
2201 AV_WL32(dst + offset + 12, 0); // total samples
2202 AV_WL32(dst + offset + 16, 0); // block index
2203 AV_WL32(dst + offset + 20, samples); // number of samples
2204 AV_WL32(dst + offset + 24, flags); // flags
2205 AV_WL32(dst + offset + 28, crc); // crc
2206 memcpy(dst + offset + 32, src, blocksize); // block data
2207
2208 src += blocksize;
2209 srclen -= blocksize;
2210 offset += blocksize + 32;
2211 }
2212
2213 *pdst = dst;
2214 *size = dstlen;
2215
2216 return 0;
2217
2218 fail:
2219 av_freep(&dst);
2220 return ret;
2221 }
2222
2223 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2224 MatroskaTrack *track, AVStream *st,
2225 uint8_t *data, int pkt_size,
2226 uint64_t timecode, uint64_t duration,
2227 int64_t pos, int is_keyframe)
2228 {
2229 MatroskaTrackEncoding *encodings = track->encodings.elem;
2230 uint8_t *pkt_data = data;
2231 int offset = 0, res;
2232 AVPacket *pkt;
2233
2234 if (encodings && encodings->scope & 1) {
2235 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2236 if (res < 0)
2237 return res;
2238 }
2239
2240 if (st->codec->codec_id == AV_CODEC_ID_WAVPACK) {
2241 uint8_t *wv_data;
2242 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
2243 if (res < 0) {
2244 av_log(matroska->ctx, AV_LOG_ERROR,
2245 "Error parsing a wavpack block.\n");
2246 goto fail;
2247 }
2248 if (pkt_data != data)
2249 av_freep(&pkt_data);
2250 pkt_data = wv_data;
2251 }
2252
2253 if (st->codec->codec_id == AV_CODEC_ID_PRORES)
2254 offset = 8;
2255
2256 pkt = av_mallocz(sizeof(AVPacket));
2257 /* XXX: prevent data copy... */
2258 if (av_new_packet(pkt, pkt_size + offset) < 0) {
2259 av_free(pkt);
2260 return AVERROR(ENOMEM);
2261 }
2262
2263 if (st->codec->codec_id == AV_CODEC_ID_PRORES) {
2264 uint8_t *buf = pkt->data;
2265 bytestream_put_be32(&buf, pkt_size);
2266 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2267 }
2268
2269 memcpy(pkt->data + offset, pkt_data, pkt_size);
2270
2271 if (pkt_data != data)
2272 av_free(pkt_data);
2273
2274 pkt->flags = is_keyframe;
2275 pkt->stream_index = st->index;
2276
2277 if (track->ms_compat)
2278 pkt->dts = timecode;
2279 else
2280 pkt->pts = timecode;
2281 pkt->pos = pos;
2282 if (st->codec->codec_id == AV_CODEC_ID_TEXT)
2283 pkt->convergence_duration = duration;
2284 else if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE)
2285 pkt->duration = duration;
2286
2287 if (st->codec->codec_id == AV_CODEC_ID_SSA)
2288 matroska_fix_ass_packet(matroska, pkt, duration);
2289
2290 if (matroska->prev_pkt &&
2291 timecode != AV_NOPTS_VALUE &&
2292 matroska->prev_pkt->pts == timecode &&
2293 matroska->prev_pkt->stream_index == st->index &&
2294 st->codec->codec_id == AV_CODEC_ID_SSA)
2295 matroska_merge_packets(matroska->prev_pkt, pkt);
2296 else {
2297 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2298 matroska->prev_pkt = pkt;
2299 }
2300
2301 return 0;
2302
2303 fail:
2304 if (pkt_data != data)
2305 av_freep(&pkt_data);
2306 return res;
2307 }
2308
2309 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
2310 int size, int64_t pos, uint64_t cluster_time,
2311 uint64_t block_duration, int is_keyframe,
2312 int64_t cluster_pos)
2313 {
2314 uint64_t timecode = AV_NOPTS_VALUE;
2315 MatroskaTrack *track;
2316 int res = 0;
2317 AVStream *st;
2318 int16_t block_time;
2319 uint32_t *lace_size = NULL;
2320 int n, flags, laces = 0;
2321 uint64_t num, duration;
2322
2323 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2324 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2325 return n;
2326 }
2327 data += n;
2328 size -= n;
2329
2330 track = matroska_find_track_by_num(matroska, num);
2331 if (!track || !track->stream) {
2332 av_log(matroska->ctx, AV_LOG_INFO,
2333 "Invalid stream %"PRIu64" or size %u\n", num, size);
2334 return AVERROR_INVALIDDATA;
2335 } else if (size <= 3)
2336 return 0;
2337 st = track->stream;
2338 if (st->discard >= AVDISCARD_ALL)
2339 return res;
2340
2341 block_time = AV_RB16(data);
2342 data += 2;
2343 flags = *data++;
2344 size -= 3;
2345 if (is_keyframe == -1)
2346 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2347
2348 if (cluster_time != (uint64_t) -1 &&
2349 (block_time >= 0 || cluster_time >= -block_time)) {
2350 timecode = cluster_time + block_time - track->codec_delay;
2351 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2352 timecode < track->end_timecode)
2353 is_keyframe = 0; /* overlapping subtitles are not key frame */
2354 if (is_keyframe)
2355 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
2356 AVINDEX_KEYFRAME);
2357 }
2358
2359 if (matroska->skip_to_keyframe &&
2360 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2361 if (!is_keyframe || timecode < matroska->skip_to_timecode)
2362 return res;
2363 matroska->skip_to_keyframe = 0;
2364 }
2365
2366 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
2367 &lace_size, &laces);
2368
2369 if (res)
2370 goto end;
2371
2372 if (block_duration != AV_NOPTS_VALUE) {
2373 duration = block_duration / laces;
2374 if (block_duration != duration * laces) {
2375 av_log(matroska->ctx, AV_LOG_WARNING,
2376 "Incorrect block_duration, possibly corrupted container");
2377 }
2378 } else {
2379 duration = track->default_duration / matroska->time_scale;
2380 block_duration = duration * laces;
2381 }
2382
2383 if (timecode != AV_NOPTS_VALUE)
2384 track->end_timecode =
2385 FFMAX(track->end_timecode, timecode + block_duration);
2386
2387 for (n = 0; n < laces; n++) {
2388 if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2389 st->codec->codec_id == AV_CODEC_ID_COOK ||
2390 st->codec->codec_id == AV_CODEC_ID_SIPR ||
2391 st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2392 st->codec->block_align && track->audio.sub_packet_size) {
2393 res = matroska_parse_rm_audio(matroska, track, st, data,
2394 lace_size[n],
2395 timecode, duration, pos);
2396 if (res)
2397 goto end;
2398 } else {
2399 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2400 timecode, duration, pos,
2401 !n ? is_keyframe : 0);
2402 if (res)
2403 goto end;
2404 }
2405
2406 if (timecode != AV_NOPTS_VALUE)
2407 timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
2408 data += lace_size[n];
2409 }
2410
2411 end:
2412 av_free(lace_size);
2413 return res;
2414 }
2415
2416 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
2417 {
2418 EbmlList *blocks_list;
2419 MatroskaBlock *blocks;
2420 int i, res;
2421 res = ebml_parse(matroska,
2422 matroska_cluster_incremental_parsing,
2423 &matroska->current_cluster);
2424 if (res == 1) {
2425 /* New Cluster */
2426 if (matroska->current_cluster_pos)
2427 ebml_level_end(matroska);
2428 ebml_free(matroska_cluster, &matroska->current_cluster);
2429 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2430 matroska->current_cluster_num_blocks = 0;
2431 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2432 matroska->prev_pkt = NULL;
2433 /* sizeof the ID which was already read */
2434 if (matroska->current_id)
2435 matroska->current_cluster_pos -= 4;
2436 res = ebml_parse(matroska,
2437 matroska_clusters_incremental,
2438 &matroska->current_cluster);
2439 /* Try parsing the block again. */
2440 if (res == 1)
2441 res = ebml_parse(matroska,
2442 matroska_cluster_incremental_parsing,
2443 &matroska->current_cluster);
2444 }
2445
2446 if (!res &&
2447 matroska->current_cluster_num_blocks <
2448 matroska->current_cluster.blocks.nb_elem) {
2449 blocks_list = &matroska->current_cluster.blocks;
2450 blocks = blocks_list->elem;
2451
2452 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2453 i = blocks_list->nb_elem - 1;
2454 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2455 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2456 if (!blocks[i].non_simple)
2457 blocks[i].duration = AV_NOPTS_VALUE;
2458 res = matroska_parse_block(matroska, blocks[i].bin.data,
2459 blocks[i].bin.size, blocks[i].bin.pos,
2460 matroska->current_cluster.timecode,
2461 blocks[i].duration, is_keyframe,
2462 matroska->current_cluster_pos);
2463 }
2464 }
2465
2466 if (res < 0)
2467 matroska->done = 1;
2468 return res;
2469 }
2470
2471 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
2472 {
2473 MatroskaCluster cluster = { 0 };
2474 EbmlList *blocks_list;
2475 MatroskaBlock *blocks;
2476 int i, res;
2477 int64_t pos;
2478
2479 if (!matroska->contains_ssa)
2480 return matroska_parse_cluster_incremental(matroska);
2481 pos = avio_tell(matroska->ctx->pb);
2482 matroska->prev_pkt = NULL;
2483 if (matroska->current_id)
2484 pos -= 4; /* sizeof the ID which was already read */
2485 res = ebml_parse(matroska, matroska_clusters, &cluster);
2486 blocks_list = &cluster.blocks;
2487 blocks = blocks_list->elem;
2488 for (i = 0; i < blocks_list->nb_elem && !res; i++)
2489 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2490 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2491 if (!blocks[i].non_simple)
2492 blocks[i].duration = AV_NOPTS_VALUE;
2493 res = matroska_parse_block(matroska, blocks[i].bin.data,
2494 blocks[i].bin.size, blocks[i].bin.pos,
2495 cluster.timecode, blocks[i].duration,
2496 is_keyframe, pos);
2497 }
2498 ebml_free(matroska_cluster, &cluster);
2499 return res;
2500 }
2501
2502 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
2503 {
2504 MatroskaDemuxContext *matroska = s->priv_data;
2505 int ret = 0;
2506
2507 while (!ret && matroska_deliver_packet(matroska, pkt)) {
2508 int64_t pos = avio_tell(matroska->ctx->pb);
2509 if (matroska->done)
2510 return AVERROR_EOF;
2511 if (matroska_parse_cluster(matroska) < 0)
2512 ret = matroska_resync(matroska, pos);
2513 }
2514
2515 if (ret == AVERROR_INVALIDDATA && pkt->data) {
2516 pkt->flags |= AV_PKT_FLAG_CORRUPT;
2517 return 0;
2518 }
2519
2520 return ret;
2521 }
2522
2523 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2524 int64_t timestamp, int flags)
2525 {
2526 MatroskaDemuxContext *matroska = s->priv_data;
2527 MatroskaTrack *tracks = matroska->tracks.elem;
2528 AVStream *st = s->streams[stream_index];
2529 int i, index, index_sub, index_min;
2530
2531 /* Parse the CUES now since we need the index data to seek. */
2532 if (matroska->cues_parsing_deferred) {
2533 matroska_parse_cues(matroska);
2534 matroska->cues_parsing_deferred = 0;
2535 }
2536
2537 if (!st->nb_index_entries)
2538 return 0;
2539 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2540
2541 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2542 avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
2543 SEEK_SET);
2544 matroska->current_id = 0;
2545 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2546 matroska_clear_queue(matroska);
2547 if (matroska_parse_cluster(matroska) < 0)
2548 break;
2549 }
2550 }
2551
2552 matroska_clear_queue(matroska);
2553 if (index < 0)
2554 return 0;
2555
2556 index_min = index;
2557 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2558 tracks[i].audio.pkt_cnt = 0;
2559 tracks[i].audio.sub_packet_cnt = 0;
2560 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2561 tracks[i].end_timecode = 0;
2562 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2563 tracks[i].stream->discard != AVDISCARD_ALL) {
2564 index_sub = av_index_search_timestamp(
2565 tracks[i].stream, st->index_entries[index].timestamp,
2566 AVSEEK_FLAG_BACKWARD);
2567 if (index_sub >= 0 &&
2568 st->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
2569 st->index_entries[index].timestamp -
2570 st->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
2571 index_min = index_sub;
2572 }
2573 }
2574
2575 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2576 matroska->current_id = 0;
2577 matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
2578 matroska->skip_to_timecode = st->index_entries[index].timestamp;
2579 matroska->done = 0;
2580 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2581 return 0;
2582 }
2583
2584 static int matroska_read_close(AVFormatContext *s)
2585 {
2586 MatroskaDemuxContext *matroska = s->priv_data;
2587 MatroskaTrack *tracks = matroska->tracks.elem;
2588 int n;
2589
2590 matroska_clear_queue(matroska);
2591
2592 for (n = 0; n < matroska->tracks.nb_elem; n++)
2593 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2594 av_free(tracks[n].audio.buf);
2595 ebml_free(matroska_cluster, &matroska->current_cluster);
2596 ebml_free(matroska_segment, matroska);
2597
2598 return 0;
2599 }
2600
2601 AVInputFormat ff_matroska_demuxer = {
2602 .name = "matroska,webm",
2603 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
2604 .priv_data_size = sizeof(MatroskaDemuxContext),
2605 .read_probe = matroska_probe,
2606 .read_header = matroska_read_header,
2607 .read_packet = matroska_read_packet,
2608 .read_close = matroska_read_close,
2609 .read_seek = matroska_read_seek,
2610 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
2611 };