repositories / deb_ffmpeg.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
Imported Debian version 2.5.2~trusty
[deb_ffmpeg.git] / ffmpeg / libavformat / matroskadec.c
... / ...
CommitLineData
1/*
2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The FFmpeg Project
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22/**
23 * @file
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
36#include "libavutil/avstring.h"
37#include "libavutil/base64.h"
38#include "libavutil/dict.h"
39#include "libavutil/intfloat.h"
40#include "libavutil/intreadwrite.h"
41#include "libavutil/lzo.h"
42#include "libavutil/mathematics.h"
43#include "libavutil/time_internal.h"
44
45#include "libavcodec/bytestream.h"
46#include "libavcodec/flac.h"
47#include "libavcodec/mpeg4audio.h"
48
49#include "avformat.h"
50#include "avio_internal.h"
51#include "internal.h"
52#include "isom.h"
53#include "matroska.h"
54#include "oggdec.h"
55/* For ff_codec_get_id(). */
56#include "riff.h"
57#include "rmsipr.h"
58
59#if CONFIG_BZLIB
60#include <bzlib.h>
61#endif
62#if CONFIG_ZLIB
63#include <zlib.h>
64#endif
65
66typedef enum {
67 EBML_NONE,
68 EBML_UINT,
69 EBML_FLOAT,
70 EBML_STR,
71 EBML_UTF8,
72 EBML_BIN,
73 EBML_NEST,
74 EBML_PASS,
75 EBML_STOP,
76 EBML_SINT,
77 EBML_TYPE_COUNT
78} EbmlType;
79
80typedef const struct EbmlSyntax {
81 uint32_t id;
82 EbmlType type;
83 int list_elem_size;
84 int data_offset;
85 union {
86 uint64_t u;
87 double f;
88 const char *s;
89 const struct EbmlSyntax *n;
90 } def;
91} EbmlSyntax;
92
93typedef struct {
94 int nb_elem;
95 void *elem;
96} EbmlList;
97
98typedef struct {
99 int size;
100 uint8_t *data;
101 int64_t pos;
102} EbmlBin;
103
104typedef struct {
105 uint64_t version;
106 uint64_t max_size;
107 uint64_t id_length;
108 char *doctype;
109 uint64_t doctype_version;
110} Ebml;
111
112typedef struct {
113 uint64_t algo;
114 EbmlBin settings;
115} MatroskaTrackCompression;
116
117typedef struct {
118 uint64_t algo;
119 EbmlBin key_id;
120} MatroskaTrackEncryption;
121
122typedef struct {
123 uint64_t scope;
124 uint64_t type;
125 MatroskaTrackCompression compression;
126 MatroskaTrackEncryption encryption;
127} MatroskaTrackEncoding;
128
129typedef struct {
130 double frame_rate;
131 uint64_t display_width;
132 uint64_t display_height;
133 uint64_t pixel_width;
134 uint64_t pixel_height;
135 EbmlBin color_space;
136 uint64_t stereo_mode;
137 uint64_t alpha_mode;
138} MatroskaTrackVideo;
139
140typedef struct {
141 double samplerate;
142 double out_samplerate;
143 uint64_t bitdepth;
144 uint64_t channels;
145
146 /* real audio header (extracted from extradata) */
147 int coded_framesize;
148 int sub_packet_h;
149 int frame_size;
150 int sub_packet_size;
151 int sub_packet_cnt;
152 int pkt_cnt;
153 uint64_t buf_timecode;
154 uint8_t *buf;
155} MatroskaTrackAudio;
156
157typedef struct {
158 uint64_t uid;
159 uint64_t type;
160} MatroskaTrackPlane;
161
162typedef struct {
163 EbmlList combine_planes;
164} MatroskaTrackOperation;
165
166typedef struct {
167 uint64_t num;
168 uint64_t uid;
169 uint64_t type;
170 char *name;
171 char *codec_id;
172 EbmlBin codec_priv;
173 char *language;
174 double time_scale;
175 uint64_t default_duration;
176 uint64_t flag_default;
177 uint64_t flag_forced;
178 uint64_t seek_preroll;
179 MatroskaTrackVideo video;
180 MatroskaTrackAudio audio;
181 MatroskaTrackOperation operation;
182 EbmlList encodings;
183 uint64_t codec_delay;
184
185 AVStream *stream;
186 int64_t end_timecode;
187 int ms_compat;
188 uint64_t max_block_additional_id;
189} MatroskaTrack;
190
191typedef struct {
192 uint64_t uid;
193 char *filename;
194 char *mime;
195 EbmlBin bin;
196
197 AVStream *stream;
198} MatroskaAttachment;
199
200typedef struct {
201 uint64_t start;
202 uint64_t end;
203 uint64_t uid;
204 char *title;
205
206 AVChapter *chapter;
207} MatroskaChapter;
208
209typedef struct {
210 uint64_t track;
211 uint64_t pos;
212} MatroskaIndexPos;
213
214typedef struct {
215 uint64_t time;
216 EbmlList pos;
217} MatroskaIndex;
218
219typedef struct {
220 char *name;
221 char *string;
222 char *lang;
223 uint64_t def;
224 EbmlList sub;
225} MatroskaTag;
226
227typedef struct {
228 char *type;
229 uint64_t typevalue;
230 uint64_t trackuid;
231 uint64_t chapteruid;
232 uint64_t attachuid;
233} MatroskaTagTarget;
234
235typedef struct {
236 MatroskaTagTarget target;
237 EbmlList tag;
238} MatroskaTags;
239
240typedef struct {
241 uint64_t id;
242 uint64_t pos;
243} MatroskaSeekhead;
244
245typedef struct {
246 uint64_t start;
247 uint64_t length;
248} MatroskaLevel;
249
250typedef struct {
251 uint64_t timecode;
252 EbmlList blocks;
253} MatroskaCluster;
254
255typedef struct {
256 AVFormatContext *ctx;
257
258 /* EBML stuff */
259 int num_levels;
260 MatroskaLevel levels[EBML_MAX_DEPTH];
261 int level_up;
262 uint32_t current_id;
263
264 uint64_t time_scale;
265 double duration;
266 char *title;
267 char *muxingapp;
268 EbmlBin date_utc;
269 EbmlList tracks;
270 EbmlList attachments;
271 EbmlList chapters;
272 EbmlList index;
273 EbmlList tags;
274 EbmlList seekhead;
275
276 /* byte position of the segment inside the stream */
277 int64_t segment_start;
278
279 /* the packet queue */
280 AVPacket **packets;
281 int num_packets;
282 AVPacket *prev_pkt;
283
284 int done;
285
286 /* What to skip before effectively reading a packet. */
287 int skip_to_keyframe;
288 uint64_t skip_to_timecode;
289
290 /* File has a CUES element, but we defer parsing until it is needed. */
291 int cues_parsing_deferred;
292
293 int current_cluster_num_blocks;
294 int64_t current_cluster_pos;
295 MatroskaCluster current_cluster;
296
297 /* File has SSA subtitles which prevent incremental cluster parsing. */
298 int contains_ssa;
299} MatroskaDemuxContext;
300
301typedef struct {
302 uint64_t duration;
303 int64_t reference;
304 uint64_t non_simple;
305 EbmlBin bin;
306 uint64_t additional_id;
307 EbmlBin additional;
308 int64_t discard_padding;
309} MatroskaBlock;
310
311static EbmlSyntax ebml_header[] = {
312 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
313 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
314 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
315 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
316 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
317 { EBML_ID_EBMLVERSION, EBML_NONE },
318 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
319 { 0 }
320};
321
322static EbmlSyntax ebml_syntax[] = {
323 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
324 { 0 }
325};
326
327static EbmlSyntax matroska_info[] = {
328 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
329 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
330 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
331 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
332 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
333 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
334 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
335 { 0 }
336};
337
338static EbmlSyntax matroska_track_video[] = {
339 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
340 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
341 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
342 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
343 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
344 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
345 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
346 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
347 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
348 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
349 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
350 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
351 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_NONE },
352 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
353 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
354 { 0 }
355};
356
357static EbmlSyntax matroska_track_audio[] = {
358 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
359 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
360 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
361 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
362 { 0 }
363};
364
365static EbmlSyntax matroska_track_encoding_compression[] = {
366 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
367 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
368 { 0 }
369};
370
371static EbmlSyntax matroska_track_encoding_encryption[] = {
372 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
373 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
374 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
375 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
376 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
377 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
378 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
379 { 0 }
380};
381static EbmlSyntax matroska_track_encoding[] = {
382 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
383 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
384 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
385 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
386 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
387 { 0 }
388};
389
390static EbmlSyntax matroska_track_encodings[] = {
391 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
392 { 0 }
393};
394
395static EbmlSyntax matroska_track_plane[] = {
396 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
397 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
398 { 0 }
399};
400
401static EbmlSyntax matroska_track_combine_planes[] = {
402 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
403 { 0 }
404};
405
406static EbmlSyntax matroska_track_operation[] = {
407 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
408 { 0 }
409};
410
411static EbmlSyntax matroska_track[] = {
412 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
413 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
414 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
415 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
416 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
417 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
418 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
419 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
420 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
421 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
422 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
423 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
424 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
425 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
426 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
427 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
428 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
429 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
430 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
431 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
432 { MATROSKA_ID_CODECNAME, EBML_NONE },
433 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
434 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
435 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
436 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
437 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
438 { 0 }
439};
440
441static EbmlSyntax matroska_tracks[] = {
442 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
443 { 0 }
444};
445
446static EbmlSyntax matroska_attachment[] = {
447 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
448 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
449 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
450 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
451 { MATROSKA_ID_FILEDESC, EBML_NONE },
452 { 0 }
453};
454
455static EbmlSyntax matroska_attachments[] = {
456 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
457 { 0 }
458};
459
460static EbmlSyntax matroska_chapter_display[] = {
461 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
462 { MATROSKA_ID_CHAPLANG, EBML_NONE },
463 { 0 }
464};
465
466static EbmlSyntax matroska_chapter_entry[] = {
467 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
468 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
469 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
470 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
471 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
472 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
473 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
474 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
475 { 0 }
476};
477
478static EbmlSyntax matroska_chapter[] = {
479 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
480 { MATROSKA_ID_EDITIONUID, EBML_NONE },
481 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
482 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
483 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
484 { 0 }
485};
486
487static EbmlSyntax matroska_chapters[] = {
488 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
489 { 0 }
490};
491
492static EbmlSyntax matroska_index_pos[] = {
493 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
494 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
495 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
496 { MATROSKA_ID_CUEDURATION, EBML_NONE },
497 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
498 { 0 }
499};
500
501static EbmlSyntax matroska_index_entry[] = {
502 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
503 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
504 { 0 }
505};
506
507static EbmlSyntax matroska_index[] = {
508 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
509 { 0 }
510};
511
512static EbmlSyntax matroska_simpletag[] = {
513 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
514 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
515 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
516 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
517 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
518 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
519 { 0 }
520};
521
522static EbmlSyntax matroska_tagtargets[] = {
523 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
524 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
525 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
526 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
527 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
528 { 0 }
529};
530
531static EbmlSyntax matroska_tag[] = {
532 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
533 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
534 { 0 }
535};
536
537static EbmlSyntax matroska_tags[] = {
538 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
539 { 0 }
540};
541
542static EbmlSyntax matroska_seekhead_entry[] = {
543 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
544 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
545 { 0 }
546};
547
548static EbmlSyntax matroska_seekhead[] = {
549 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
550 { 0 }
551};
552
553static EbmlSyntax matroska_segment[] = {
554 { MATROSKA_ID_INFO, EBML_NEST, 0, 0, { .n = matroska_info } },
555 { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, { .n = matroska_tracks } },
556 { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, { .n = matroska_attachments } },
557 { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, { .n = matroska_chapters } },
558 { MATROSKA_ID_CUES, EBML_NEST, 0, 0, { .n = matroska_index } },
559 { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, { .n = matroska_tags } },
560 { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, { .n = matroska_seekhead } },
561 { MATROSKA_ID_CLUSTER, EBML_STOP },
562 { 0 }
563};
564
565static EbmlSyntax matroska_segments[] = {
566 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
567 { 0 }
568};
569
570static EbmlSyntax matroska_blockmore[] = {
571 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
572 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
573 { 0 }
574};
575
576static EbmlSyntax matroska_blockadditions[] = {
577 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
578 { 0 }
579};
580
581static EbmlSyntax matroska_blockgroup[] = {
582 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
583 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
584 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
585 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
586 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
587 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference) },
588 { MATROSKA_ID_CODECSTATE, EBML_NONE },
589 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
590 { 0 }
591};
592
593static EbmlSyntax matroska_cluster[] = {
594 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
595 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
596 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
597 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
598 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
599 { 0 }
600};
601
602static EbmlSyntax matroska_clusters[] = {
603 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster } },
604 { MATROSKA_ID_INFO, EBML_NONE },
605 { MATROSKA_ID_CUES, EBML_NONE },
606 { MATROSKA_ID_TAGS, EBML_NONE },
607 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
608 { 0 }
609};
610
611static EbmlSyntax matroska_cluster_incremental_parsing[] = {
612 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
613 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
614 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
615 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
616 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
617 { MATROSKA_ID_INFO, EBML_NONE },
618 { MATROSKA_ID_CUES, EBML_NONE },
619 { MATROSKA_ID_TAGS, EBML_NONE },
620 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
621 { MATROSKA_ID_CLUSTER, EBML_STOP },
622 { 0 }
623};
624
625static EbmlSyntax matroska_cluster_incremental[] = {
626 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
627 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
628 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
629 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
630 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
631 { 0 }
632};
633
634static EbmlSyntax matroska_clusters_incremental[] = {
635 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
636 { MATROSKA_ID_INFO, EBML_NONE },
637 { MATROSKA_ID_CUES, EBML_NONE },
638 { MATROSKA_ID_TAGS, EBML_NONE },
639 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
640 { 0 }
641};
642
643static const char *const matroska_doctypes[] = { "matroska", "webm" };
644
645static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
646{
647 AVIOContext *pb = matroska->ctx->pb;
648 uint32_t id;
649 matroska->current_id = 0;
650 matroska->num_levels = 0;
651
652 /* seek to next position to resync from */
653 if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
654 goto eof;
655
656 id = avio_rb32(pb);
657
658 // try to find a toplevel element
659 while (!avio_feof(pb)) {
660 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
661 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
662 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
663 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
664 matroska->current_id = id;
665 return 0;
666 }
667 id = (id << 8) | avio_r8(pb);
668 }
669
670eof:
671 matroska->done = 1;
672 return AVERROR_EOF;
673}
674
675/*
676 * Return: Whether we reached the end of a level in the hierarchy or not.
677 */
678static int ebml_level_end(MatroskaDemuxContext *matroska)
679{
680 AVIOContext *pb = matroska->ctx->pb;
681 int64_t pos = avio_tell(pb);
682
683 if (matroska->num_levels > 0) {
684 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
685 if (pos - level->start >= level->length || matroska->current_id) {
686 matroska->num_levels--;
687 return 1;
688 }
689 }
690 return 0;
691}
692
693/*
694 * Read: an "EBML number", which is defined as a variable-length
695 * array of bytes. The first byte indicates the length by giving a
696 * number of 0-bits followed by a one. The position of the first
697 * "one" bit inside the first byte indicates the length of this
698 * number.
699 * Returns: number of bytes read, < 0 on error
700 */
701static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
702 int max_size, uint64_t *number)
703{
704 int read = 1, n = 1;
705 uint64_t total = 0;
706
707 /* The first byte tells us the length in bytes - avio_r8() can normally
708 * return 0, but since that's not a valid first ebmlID byte, we can
709 * use it safely here to catch EOS. */
710 if (!(total = avio_r8(pb))) {
711 /* we might encounter EOS here */
712 if (!avio_feof(pb)) {
713 int64_t pos = avio_tell(pb);
714 av_log(matroska->ctx, AV_LOG_ERROR,
715 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
716 pos, pos);
717 return pb->error ? pb->error : AVERROR(EIO);
718 }
719 return AVERROR_EOF;
720 }
721
722 /* get the length of the EBML number */
723 read = 8 - ff_log2_tab[total];
724 if (read > max_size) {
725 int64_t pos = avio_tell(pb) - 1;
726 av_log(matroska->ctx, AV_LOG_ERROR,
727 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
728 (uint8_t) total, pos, pos);
729 return AVERROR_INVALIDDATA;
730 }
731
732 /* read out length */
733 total ^= 1 << ff_log2_tab[total];
734 while (n++ < read)
735 total = (total << 8) | avio_r8(pb);
736
737 *number = total;
738
739 return read;
740}
741
742/**
743 * Read a EBML length value.
744 * This needs special handling for the "unknown length" case which has multiple
745 * encodings.
746 */
747static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
748 uint64_t *number)
749{
750 int res = ebml_read_num(matroska, pb, 8, number);
751 if (res > 0 && *number + 1 == 1ULL << (7 * res))
752 *number = 0xffffffffffffffULL;
753 return res;
754}
755
756/*
757 * Read the next element as an unsigned int.
758 * 0 is success, < 0 is failure.
759 */
760static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
761{
762 int n = 0;
763
764 if (size > 8)
765 return AVERROR_INVALIDDATA;
766
767 /* big-endian ordering; build up number */
768 *num = 0;
769 while (n++ < size)
770 *num = (*num << 8) | avio_r8(pb);
771
772 return 0;
773}
774
775/*
776 * Read the next element as a signed int.
777 * 0 is success, < 0 is failure.
778 */
779static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
780{
781 int n = 1;
782
783 if (size > 8)
784 return AVERROR_INVALIDDATA;
785
786 if (size == 0) {
787 *num = 0;
788 } else {
789 *num = sign_extend(avio_r8(pb), 8);
790
791 /* big-endian ordering; build up number */
792 while (n++ < size)
793 *num = (*num << 8) | avio_r8(pb);
794 }
795
796 return 0;
797}
798
799/*
800 * Read the next element as a float.
801 * 0 is success, < 0 is failure.
802 */
803static int ebml_read_float(AVIOContext *pb, int size, double *num)
804{
805 if (size == 0)
806 *num = 0;
807 else if (size == 4)
808 *num = av_int2float(avio_rb32(pb));
809 else if (size == 8)
810 *num = av_int2double(avio_rb64(pb));
811 else
812 return AVERROR_INVALIDDATA;
813
814 return 0;
815}
816
817/*
818 * Read the next element as an ASCII string.
819 * 0 is success, < 0 is failure.
820 */
821static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
822{
823 char *res;
824
825 /* EBML strings are usually not 0-terminated, so we allocate one
826 * byte more, read the string and NULL-terminate it ourselves. */
827 if (!(res = av_malloc(size + 1)))
828 return AVERROR(ENOMEM);
829 if (avio_read(pb, (uint8_t *) res, size) != size) {
830 av_free(res);
831 return AVERROR(EIO);
832 }
833 (res)[size] = '\0';
834 av_free(*str);
835 *str = res;
836
837 return 0;
838}
839
840/*
841 * Read the next element as binary data.
842 * 0 is success, < 0 is failure.
843 */
844static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
845{
846 av_fast_padded_malloc(&bin->data, &bin->size, length);
847 if (!bin->data)
848 return AVERROR(ENOMEM);
849
850 bin->size = length;
851 bin->pos = avio_tell(pb);
852 if (avio_read(pb, bin->data, length) != length) {
853 av_freep(&bin->data);
854 bin->size = 0;
855 return AVERROR(EIO);
856 }
857
858 return 0;
859}
860
861/*
862 * Read the next element, but only the header. The contents
863 * are supposed to be sub-elements which can be read separately.
864 * 0 is success, < 0 is failure.
865 */
866static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
867{
868 AVIOContext *pb = matroska->ctx->pb;
869 MatroskaLevel *level;
870
871 if (matroska->num_levels >= EBML_MAX_DEPTH) {
872 av_log(matroska->ctx, AV_LOG_ERROR,
873 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
874 return AVERROR(ENOSYS);
875 }
876
877 level = &matroska->levels[matroska->num_levels++];
878 level->start = avio_tell(pb);
879 level->length = length;
880
881 return 0;
882}
883
884/*
885 * Read signed/unsigned "EBML" numbers.
886 * Return: number of bytes processed, < 0 on error
887 */
888static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
889 uint8_t *data, uint32_t size, uint64_t *num)
890{
891 AVIOContext pb;
892 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
893 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
894}
895
896/*
897 * Same as above, but signed.
898 */
899static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
900 uint8_t *data, uint32_t size, int64_t *num)
901{
902 uint64_t unum;
903 int res;
904
905 /* read as unsigned number first */
906 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
907 return res;
908
909 /* make signed (weird way) */
910 *num = unum - ((1LL << (7 * res - 1)) - 1);
911
912 return res;
913}
914
915static int ebml_parse_elem(MatroskaDemuxContext *matroska,
916 EbmlSyntax *syntax, void *data);
917
918static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
919 uint32_t id, void *data)
920{
921 int i;
922 for (i = 0; syntax[i].id; i++)
923 if (id == syntax[i].id)
924 break;
925 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
926 matroska->num_levels > 0 &&
927 matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
928 return 0; // we reached the end of an unknown size cluster
929 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
930 av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%"PRIX32"\n", id);
931 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
932 return AVERROR_INVALIDDATA;
933 }
934 return ebml_parse_elem(matroska, &syntax[i], data);
935}
936
937static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
938 void *data)
939{
940 if (!matroska->current_id) {
941 uint64_t id;
942 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
943 if (res < 0)
944 return res;
945 matroska->current_id = id | 1 << 7 * res;
946 }
947 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
948}
949
950static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
951 void *data)
952{
953 int i, res = 0;
954
955 for (i = 0; syntax[i].id; i++)
956 switch (syntax[i].type) {
957 case EBML_UINT:
958 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
959 break;
960 case EBML_FLOAT:
961 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
962 break;
963 case EBML_STR:
964 case EBML_UTF8:
965 // the default may be NULL
966 if (syntax[i].def.s) {
967 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
968 *dst = av_strdup(syntax[i].def.s);
969 if (!*dst)
970 return AVERROR(ENOMEM);
971 }
972 break;
973 }
974
975 while (!res && !ebml_level_end(matroska))
976 res = ebml_parse(matroska, syntax, data);
977
978 return res;
979}
980
981static int ebml_parse_elem(MatroskaDemuxContext *matroska,
982 EbmlSyntax *syntax, void *data)
983{
984 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
985 [EBML_UINT] = 8,
986 [EBML_FLOAT] = 8,
987 // max. 16 MB for strings
988 [EBML_STR] = 0x1000000,
989 [EBML_UTF8] = 0x1000000,
990 // max. 256 MB for binary data
991 [EBML_BIN] = 0x10000000,
992 // no limits for anything else
993 };
994 AVIOContext *pb = matroska->ctx->pb;
995 uint32_t id = syntax->id;
996 uint64_t length;
997 int res;
998 void *newelem;
999
1000 data = (char *) data + syntax->data_offset;
1001 if (syntax->list_elem_size) {
1002 EbmlList *list = data;
1003 newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
1004 if (!newelem)
1005 return AVERROR(ENOMEM);
1006 list->elem = newelem;
1007 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1008 memset(data, 0, syntax->list_elem_size);
1009 list->nb_elem++;
1010 }
1011
1012 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
1013 matroska->current_id = 0;
1014 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1015 return res;
1016 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1017 av_log(matroska->ctx, AV_LOG_ERROR,
1018 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1019 length, max_lengths[syntax->type], syntax->type);
1020 return AVERROR_INVALIDDATA;
1021 }
1022 }
1023
1024 switch (syntax->type) {
1025 case EBML_UINT:
1026 res = ebml_read_uint(pb, length, data);
1027 break;
1028 case EBML_SINT:
1029 res = ebml_read_sint(pb, length, data);
1030 break;
1031 case EBML_FLOAT:
1032 res = ebml_read_float(pb, length, data);
1033 break;
1034 case EBML_STR:
1035 case EBML_UTF8:
1036 res = ebml_read_ascii(pb, length, data);
1037 break;
1038 case EBML_BIN:
1039 res = ebml_read_binary(pb, length, data);
1040 break;
1041 case EBML_NEST:
1042 if ((res = ebml_read_master(matroska, length)) < 0)
1043 return res;
1044 if (id == MATROSKA_ID_SEGMENT)
1045 matroska->segment_start = avio_tell(matroska->ctx->pb);
1046 return ebml_parse_nest(matroska, syntax->def.n, data);
1047 case EBML_PASS:
1048 return ebml_parse_id(matroska, syntax->def.n, id, data);
1049 case EBML_STOP:
1050 return 1;
1051 default:
1052 if (ffio_limit(pb, length) != length)
1053 return AVERROR(EIO);
1054 return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
1055 }
1056 if (res == AVERROR_INVALIDDATA)
1057 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1058 else if (res == AVERROR(EIO))
1059 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1060 return res;
1061}
1062
1063static void ebml_free(EbmlSyntax *syntax, void *data)
1064{
1065 int i, j;
1066 for (i = 0; syntax[i].id; i++) {
1067 void *data_off = (char *) data + syntax[i].data_offset;
1068 switch (syntax[i].type) {
1069 case EBML_STR:
1070 case EBML_UTF8:
1071 av_freep(data_off);
1072 break;
1073 case EBML_BIN:
1074 av_freep(&((EbmlBin *) data_off)->data);
1075 break;
1076 case EBML_NEST:
1077 if (syntax[i].list_elem_size) {
1078 EbmlList *list = data_off;
1079 char *ptr = list->elem;
1080 for (j = 0; j < list->nb_elem;
1081 j++, ptr += syntax[i].list_elem_size)
1082 ebml_free(syntax[i].def.n, ptr);
1083 av_free(list->elem);
1084 } else
1085 ebml_free(syntax[i].def.n, data_off);
1086 default:
1087 break;
1088 }
1089 }
1090}
1091
1092/*
1093 * Autodetecting...
1094 */
1095static int matroska_probe(AVProbeData *p)
1096{
1097 uint64_t total = 0;
1098 int len_mask = 0x80, size = 1, n = 1, i;
1099
1100 /* EBML header? */
1101 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1102 return 0;
1103
1104 /* length of header */
1105 total = p->buf[4];
1106 while (size <= 8 && !(total & len_mask)) {
1107 size++;
1108 len_mask >>= 1;
1109 }
1110 if (size > 8)
1111 return 0;
1112 total &= (len_mask - 1);
1113 while (n < size)
1114 total = (total << 8) | p->buf[4 + n++];
1115
1116 /* Does the probe data contain the whole header? */
1117 if (p->buf_size < 4 + size + total)
1118 return 0;
1119
1120 /* The header should contain a known document type. For now,
1121 * we don't parse the whole header but simply check for the
1122 * availability of that array of characters inside the header.
1123 * Not fully fool-proof, but good enough. */
1124 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1125 int probelen = strlen(matroska_doctypes[i]);
1126 if (total < probelen)
1127 continue;
1128 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1129 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1130 return AVPROBE_SCORE_MAX;
1131 }
1132
1133 // probably valid EBML header but no recognized doctype
1134 return AVPROBE_SCORE_EXTENSION;
1135}
1136
1137static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1138 int num)
1139{
1140 MatroskaTrack *tracks = matroska->tracks.elem;
1141 int i;
1142
1143 for (i = 0; i < matroska->tracks.nb_elem; i++)
1144 if (tracks[i].num == num)
1145 return &tracks[i];
1146
1147 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1148 return NULL;
1149}
1150
1151static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1152 MatroskaTrack *track)
1153{
1154 MatroskaTrackEncoding *encodings = track->encodings.elem;
1155 uint8_t *data = *buf;
1156 int isize = *buf_size;
1157 uint8_t *pkt_data = NULL;
1158 uint8_t av_unused *newpktdata;
1159 int pkt_size = isize;
1160 int result = 0;
1161 int olen;
1162
1163 if (pkt_size >= 10000000U)
1164 return AVERROR_INVALIDDATA;
1165
1166 switch (encodings[0].compression.algo) {
1167 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1168 {
1169 int header_size = encodings[0].compression.settings.size;
1170 uint8_t *header = encodings[0].compression.settings.data;
1171
1172 if (header_size && !header) {
1173 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1174 return -1;
1175 }
1176
1177 if (!header_size)
1178 return 0;
1179
1180 pkt_size = isize + header_size;
1181 pkt_data = av_malloc(pkt_size);
1182 if (!pkt_data)
1183 return AVERROR(ENOMEM);
1184
1185 memcpy(pkt_data, header, header_size);
1186 memcpy(pkt_data + header_size, data, isize);
1187 break;
1188 }
1189#if CONFIG_LZO
1190 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1191 do {
1192 olen = pkt_size *= 3;
1193 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1194 if (!newpktdata) {
1195 result = AVERROR(ENOMEM);
1196 goto failed;
1197 }
1198 pkt_data = newpktdata;
1199 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1200 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1201 if (result) {
1202 result = AVERROR_INVALIDDATA;
1203 goto failed;
1204 }
1205 pkt_size -= olen;
1206 break;
1207#endif
1208#if CONFIG_ZLIB
1209 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1210 {
1211 z_stream zstream = { 0 };
1212 if (inflateInit(&zstream) != Z_OK)
1213 return -1;
1214 zstream.next_in = data;
1215 zstream.avail_in = isize;
1216 do {
1217 pkt_size *= 3;
1218 newpktdata = av_realloc(pkt_data, pkt_size);
1219 if (!newpktdata) {
1220 inflateEnd(&zstream);
1221 goto failed;
1222 }
1223 pkt_data = newpktdata;
1224 zstream.avail_out = pkt_size - zstream.total_out;
1225 zstream.next_out = pkt_data + zstream.total_out;
1226 if (pkt_data) {
1227 result = inflate(&zstream, Z_NO_FLUSH);
1228 } else
1229 result = Z_MEM_ERROR;
1230 } while (result == Z_OK && pkt_size < 10000000);
1231 pkt_size = zstream.total_out;
1232 inflateEnd(&zstream);
1233 if (result != Z_STREAM_END) {
1234 if (result == Z_MEM_ERROR)
1235 result = AVERROR(ENOMEM);
1236 else
1237 result = AVERROR_INVALIDDATA;
1238 goto failed;
1239 }
1240 break;
1241 }
1242#endif
1243#if CONFIG_BZLIB
1244 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1245 {
1246 bz_stream bzstream = { 0 };
1247 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1248 return -1;
1249 bzstream.next_in = data;
1250 bzstream.avail_in = isize;
1251 do {
1252 pkt_size *= 3;
1253 newpktdata = av_realloc(pkt_data, pkt_size);
1254 if (!newpktdata) {
1255 BZ2_bzDecompressEnd(&bzstream);
1256 goto failed;
1257 }
1258 pkt_data = newpktdata;
1259 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1260 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1261 if (pkt_data) {
1262 result = BZ2_bzDecompress(&bzstream);
1263 } else
1264 result = BZ_MEM_ERROR;
1265 } while (result == BZ_OK && pkt_size < 10000000);
1266 pkt_size = bzstream.total_out_lo32;
1267 BZ2_bzDecompressEnd(&bzstream);
1268 if (result != BZ_STREAM_END) {
1269 if (result == BZ_MEM_ERROR)
1270 result = AVERROR(ENOMEM);
1271 else
1272 result = AVERROR_INVALIDDATA;
1273 goto failed;
1274 }
1275 break;
1276 }
1277#endif
1278 default:
1279 return AVERROR_INVALIDDATA;
1280 }
1281
1282 *buf = pkt_data;
1283 *buf_size = pkt_size;
1284 return 0;
1285
1286failed:
1287 av_free(pkt_data);
1288 return result;
1289}
1290
1291static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1292 AVDictionary **metadata, char *prefix)
1293{
1294 MatroskaTag *tags = list->elem;
1295 char key[1024];
1296 int i;
1297
1298 for (i = 0; i < list->nb_elem; i++) {
1299 const char *lang = tags[i].lang &&
1300 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1301
1302 if (!tags[i].name) {
1303 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1304 continue;
1305 }
1306 if (prefix)
1307 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1308 else
1309 av_strlcpy(key, tags[i].name, sizeof(key));
1310 if (tags[i].def || !lang) {
1311 av_dict_set(metadata, key, tags[i].string, 0);
1312 if (tags[i].sub.nb_elem)
1313 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1314 }
1315 if (lang) {
1316 av_strlcat(key, "-", sizeof(key));
1317 av_strlcat(key, lang, sizeof(key));
1318 av_dict_set(metadata, key, tags[i].string, 0);
1319 if (tags[i].sub.nb_elem)
1320 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1321 }
1322 }
1323 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1324}
1325
1326static void matroska_convert_tags(AVFormatContext *s)
1327{
1328 MatroskaDemuxContext *matroska = s->priv_data;
1329 MatroskaTags *tags = matroska->tags.elem;
1330 int i, j;
1331
1332 for (i = 0; i < matroska->tags.nb_elem; i++) {
1333 if (tags[i].target.attachuid) {
1334 MatroskaAttachment *attachment = matroska->attachments.elem;
1335 for (j = 0; j < matroska->attachments.nb_elem; j++)
1336 if (attachment[j].uid == tags[i].target.attachuid &&
1337 attachment[j].stream)
1338 matroska_convert_tag(s, &tags[i].tag,
1339 &attachment[j].stream->metadata, NULL);
1340 } else if (tags[i].target.chapteruid) {
1341 MatroskaChapter *chapter = matroska->chapters.elem;
1342 for (j = 0; j < matroska->chapters.nb_elem; j++)
1343 if (chapter[j].uid == tags[i].target.chapteruid &&
1344 chapter[j].chapter)
1345 matroska_convert_tag(s, &tags[i].tag,
1346 &chapter[j].chapter->metadata, NULL);
1347 } else if (tags[i].target.trackuid) {
1348 MatroskaTrack *track = matroska->tracks.elem;
1349 for (j = 0; j < matroska->tracks.nb_elem; j++)
1350 if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1351 matroska_convert_tag(s, &tags[i].tag,
1352 &track[j].stream->metadata, NULL);
1353 } else {
1354 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1355 tags[i].target.type);
1356 }
1357 }
1358}
1359
1360static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1361 int idx)
1362{
1363 EbmlList *seekhead_list = &matroska->seekhead;
1364 uint32_t level_up = matroska->level_up;
1365 uint32_t saved_id = matroska->current_id;
1366 MatroskaSeekhead *seekhead = seekhead_list->elem;
1367 int64_t before_pos = avio_tell(matroska->ctx->pb);
1368 MatroskaLevel level;
1369 int64_t offset;
1370 int ret = 0;
1371
1372 if (idx >= seekhead_list->nb_elem ||
1373 seekhead[idx].id == MATROSKA_ID_SEEKHEAD ||
1374 seekhead[idx].id == MATROSKA_ID_CLUSTER)
1375 return 0;
1376
1377 /* seek */
1378 offset = seekhead[idx].pos + matroska->segment_start;
1379 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1380 /* We don't want to lose our seekhead level, so we add
1381 * a dummy. This is a crude hack. */
1382 if (matroska->num_levels == EBML_MAX_DEPTH) {
1383 av_log(matroska->ctx, AV_LOG_INFO,
1384 "Max EBML element depth (%d) reached, "
1385 "cannot parse further.\n", EBML_MAX_DEPTH);
1386 ret = AVERROR_INVALIDDATA;
1387 } else {
1388 level.start = 0;
1389 level.length = (uint64_t) -1;
1390 matroska->levels[matroska->num_levels] = level;
1391 matroska->num_levels++;
1392 matroska->current_id = 0;
1393
1394 ret = ebml_parse(matroska, matroska_segment, matroska);
1395
1396 /* remove dummy level */
1397 while (matroska->num_levels) {
1398 uint64_t length = matroska->levels[--matroska->num_levels].length;
1399 if (length == (uint64_t) -1)
1400 break;
1401 }
1402 }
1403 }
1404 /* seek back */
1405 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1406 matroska->level_up = level_up;
1407 matroska->current_id = saved_id;
1408
1409 return ret;
1410}
1411
1412static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1413{
1414 EbmlList *seekhead_list = &matroska->seekhead;
1415 int64_t before_pos = avio_tell(matroska->ctx->pb);
1416 int i;
1417 int nb_elem;
1418
1419 // we should not do any seeking in the streaming case
1420 if (!matroska->ctx->pb->seekable ||
1421 (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1422 return;
1423
1424 // do not read entries that are added while parsing seekhead entries
1425 nb_elem = seekhead_list->nb_elem;
1426
1427 for (i = 0; i < nb_elem; i++) {
1428 MatroskaSeekhead *seekhead = seekhead_list->elem;
1429 if (seekhead[i].pos <= before_pos)
1430 continue;
1431
1432 // defer cues parsing until we actually need cue data.
1433 if (seekhead[i].id == MATROSKA_ID_CUES) {
1434 matroska->cues_parsing_deferred = 1;
1435 continue;
1436 }
1437
1438 if (matroska_parse_seekhead_entry(matroska, i) < 0) {
1439 // mark index as broken
1440 matroska->cues_parsing_deferred = -1;
1441 break;
1442 }
1443 }
1444}
1445
1446static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1447{
1448 EbmlList *index_list;
1449 MatroskaIndex *index;
1450 int index_scale = 1;
1451 int i, j;
1452
1453 index_list = &matroska->index;
1454 index = index_list->elem;
1455 if (index_list->nb_elem &&
1456 index[0].time > 1E14 / matroska->time_scale) {
1457 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1458 index_scale = matroska->time_scale;
1459 }
1460 for (i = 0; i < index_list->nb_elem; i++) {
1461 EbmlList *pos_list = &index[i].pos;
1462 MatroskaIndexPos *pos = pos_list->elem;
1463 for (j = 0; j < pos_list->nb_elem; j++) {
1464 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1465 pos[j].track);
1466 if (track && track->stream)
1467 av_add_index_entry(track->stream,
1468 pos[j].pos + matroska->segment_start,
1469 index[i].time / index_scale, 0, 0,
1470 AVINDEX_KEYFRAME);
1471 }
1472 }
1473}
1474
1475static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1476 EbmlList *seekhead_list = &matroska->seekhead;
1477 MatroskaSeekhead *seekhead = seekhead_list->elem;
1478 int i;
1479
1480 for (i = 0; i < seekhead_list->nb_elem; i++)
1481 if (seekhead[i].id == MATROSKA_ID_CUES)
1482 break;
1483 av_assert1(i <= seekhead_list->nb_elem);
1484
1485 if (matroska_parse_seekhead_entry(matroska, i) < 0)
1486 matroska->cues_parsing_deferred = -1;
1487 matroska_add_index_entries(matroska);
1488}
1489
1490static int matroska_aac_profile(char *codec_id)
1491{
1492 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1493 int profile;
1494
1495 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1496 if (strstr(codec_id, aac_profiles[profile]))
1497 break;
1498 return profile + 1;
1499}
1500
1501static int matroska_aac_sri(int samplerate)
1502{
1503 int sri;
1504
1505 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1506 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1507 break;
1508 return sri;
1509}
1510
1511static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1512{
1513 char buffer[32];
1514 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1515 time_t creation_time = date_utc / 1000000000 + 978307200;
1516 struct tm tmpbuf, *ptm = gmtime_r(&creation_time, &tmpbuf);
1517 if (!ptm) return;
1518 if (strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm))
1519 av_dict_set(metadata, "creation_time", buffer, 0);
1520}
1521
1522static int matroska_parse_flac(AVFormatContext *s,
1523 MatroskaTrack *track,
1524 int *offset)
1525{
1526 AVStream *st = track->stream;
1527 uint8_t *p = track->codec_priv.data;
1528 int size = track->codec_priv.size;
1529
1530 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1531 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1532 track->codec_priv.size = 0;
1533 return 0;
1534 }
1535 *offset = 8;
1536 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1537
1538 p += track->codec_priv.size;
1539 size -= track->codec_priv.size;
1540
1541 /* parse the remaining metadata blocks if present */
1542 while (size >= 4) {
1543 int block_last, block_type, block_size;
1544
1545 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1546
1547 p += 4;
1548 size -= 4;
1549 if (block_size > size)
1550 return 0;
1551
1552 /* check for the channel mask */
1553 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1554 AVDictionary *dict = NULL;
1555 AVDictionaryEntry *chmask;
1556
1557 ff_vorbis_comment(s, &dict, p, block_size, 0);
1558 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1559 if (chmask) {
1560 uint64_t mask = strtol(chmask->value, NULL, 0);
1561 if (!mask || mask & ~0x3ffffULL) {
1562 av_log(s, AV_LOG_WARNING,
1563 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1564 } else
1565 st->codec->channel_layout = mask;
1566 }
1567 av_dict_free(&dict);
1568 }
1569
1570 p += block_size;
1571 size -= block_size;
1572 }
1573
1574 return 0;
1575}
1576
1577static int matroska_parse_tracks(AVFormatContext *s)
1578{
1579 MatroskaDemuxContext *matroska = s->priv_data;
1580 MatroskaTrack *tracks = matroska->tracks.elem;
1581 AVStream *st;
1582 int i, j, ret;
1583 int k;
1584
1585 for (i = 0; i < matroska->tracks.nb_elem; i++) {
1586 MatroskaTrack *track = &tracks[i];
1587 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1588 EbmlList *encodings_list = &track->encodings;
1589 MatroskaTrackEncoding *encodings = encodings_list->elem;
1590 uint8_t *extradata = NULL;
1591 int extradata_size = 0;
1592 int extradata_offset = 0;
1593 uint32_t fourcc = 0;
1594 AVIOContext b;
1595 char* key_id_base64 = NULL;
1596 int bit_depth = -1;
1597
1598 /* Apply some sanity checks. */
1599 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1600 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1601 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
1602 track->type != MATROSKA_TRACK_TYPE_METADATA) {
1603 av_log(matroska->ctx, AV_LOG_INFO,
1604 "Unknown or unsupported track type %"PRIu64"\n",
1605 track->type);
1606 continue;
1607 }
1608 if (!track->codec_id)
1609 continue;
1610
1611 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1612 if (!track->default_duration && track->video.frame_rate > 0)
1613 track->default_duration = 1000000000 / track->video.frame_rate;
1614 if (track->video.display_width == -1)
1615 track->video.display_width = track->video.pixel_width;
1616 if (track->video.display_height == -1)
1617 track->video.display_height = track->video.pixel_height;
1618 if (track->video.color_space.size == 4)
1619 fourcc = AV_RL32(track->video.color_space.data);
1620 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1621 if (!track->audio.out_samplerate)
1622 track->audio.out_samplerate = track->audio.samplerate;
1623 }
1624 if (encodings_list->nb_elem > 1) {
1625 av_log(matroska->ctx, AV_LOG_ERROR,
1626 "Multiple combined encodings not supported");
1627 } else if (encodings_list->nb_elem == 1) {
1628 if (encodings[0].type) {
1629 if (encodings[0].encryption.key_id.size > 0) {
1630 /* Save the encryption key id to be stored later as a
1631 metadata tag. */
1632 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
1633 key_id_base64 = av_malloc(b64_size);
1634 if (key_id_base64 == NULL)
1635 return AVERROR(ENOMEM);
1636
1637 av_base64_encode(key_id_base64, b64_size,
1638 encodings[0].encryption.key_id.data,
1639 encodings[0].encryption.key_id.size);
1640 } else {
1641 encodings[0].scope = 0;
1642 av_log(matroska->ctx, AV_LOG_ERROR,
1643 "Unsupported encoding type");
1644 }
1645 } else if (
1646#if CONFIG_ZLIB
1647 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1648#endif
1649#if CONFIG_BZLIB
1650 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1651#endif
1652#if CONFIG_LZO
1653 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
1654#endif
1655 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
1656 encodings[0].scope = 0;
1657 av_log(matroska->ctx, AV_LOG_ERROR,
1658 "Unsupported encoding type");
1659 } else if (track->codec_priv.size && encodings[0].scope & 2) {
1660 uint8_t *codec_priv = track->codec_priv.data;
1661 int ret = matroska_decode_buffer(&track->codec_priv.data,
1662 &track->codec_priv.size,
1663 track);
1664 if (ret < 0) {
1665 track->codec_priv.data = NULL;
1666 track->codec_priv.size = 0;
1667 av_log(matroska->ctx, AV_LOG_ERROR,
1668 "Failed to decode codec private data\n");
1669 }
1670
1671 if (codec_priv != track->codec_priv.data)
1672 av_free(codec_priv);
1673 }
1674 }
1675
1676 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
1677 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1678 strlen(ff_mkv_codec_tags[j].str))) {
1679 codec_id = ff_mkv_codec_tags[j].id;
1680 break;
1681 }
1682 }
1683
1684 st = track->stream = avformat_new_stream(s, NULL);
1685 if (!st) {
1686 av_free(key_id_base64);
1687 return AVERROR(ENOMEM);
1688 }
1689
1690 if (key_id_base64) {
1691 /* export encryption key id as base64 metadata tag */
1692 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
1693 av_freep(&key_id_base64);
1694 }
1695
1696 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
1697 track->codec_priv.size >= 40 &&
1698 track->codec_priv.data) {
1699 track->ms_compat = 1;
1700 bit_depth = AV_RL16(track->codec_priv.data + 14);
1701 fourcc = AV_RL32(track->codec_priv.data + 16);
1702 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
1703 fourcc);
1704 if (!codec_id)
1705 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
1706 fourcc);
1707 extradata_offset = 40;
1708 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
1709 track->codec_priv.size >= 14 &&
1710 track->codec_priv.data) {
1711 int ret;
1712 ffio_init_context(&b, track->codec_priv.data,
1713 track->codec_priv.size,
1714 0, NULL, NULL, NULL, NULL);
1715 ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1716 if (ret < 0)
1717 return ret;
1718 codec_id = st->codec->codec_id;
1719 extradata_offset = FFMIN(track->codec_priv.size, 18);
1720 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
1721 && (track->codec_priv.size >= 86)
1722 && (track->codec_priv.data)) {
1723 fourcc = AV_RL32(track->codec_priv.data + 4);
1724 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1725 if (ff_codec_get_id(ff_codec_movaudio_tags, AV_RL32(track->codec_priv.data))) {
1726 fourcc = AV_RL32(track->codec_priv.data);
1727 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1728 }
1729 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
1730 (track->codec_priv.size >= 21) &&
1731 (track->codec_priv.data)) {
1732 fourcc = AV_RL32(track->codec_priv.data + 4);
1733 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1734 if (ff_codec_get_id(ff_codec_movvideo_tags, AV_RL32(track->codec_priv.data))) {
1735 fourcc = AV_RL32(track->codec_priv.data);
1736 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1737 }
1738 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI "))
1739 codec_id = AV_CODEC_ID_SVQ3;
1740 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1741 switch (track->audio.bitdepth) {
1742 case 8:
1743 codec_id = AV_CODEC_ID_PCM_U8;
1744 break;
1745 case 24:
1746 codec_id = AV_CODEC_ID_PCM_S24BE;
1747 break;
1748 case 32:
1749 codec_id = AV_CODEC_ID_PCM_S32BE;
1750 break;
1751 }
1752 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1753 switch (track->audio.bitdepth) {
1754 case 8:
1755 codec_id = AV_CODEC_ID_PCM_U8;
1756 break;
1757 case 24:
1758 codec_id = AV_CODEC_ID_PCM_S24LE;
1759 break;
1760 case 32:
1761 codec_id = AV_CODEC_ID_PCM_S32LE;
1762 break;
1763 }
1764 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
1765 track->audio.bitdepth == 64) {
1766 codec_id = AV_CODEC_ID_PCM_F64LE;
1767 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1768 int profile = matroska_aac_profile(track->codec_id);
1769 int sri = matroska_aac_sri(track->audio.samplerate);
1770 extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1771 if (!extradata)
1772 return AVERROR(ENOMEM);
1773 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
1774 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
1775 if (strstr(track->codec_id, "SBR")) {
1776 sri = matroska_aac_sri(track->audio.out_samplerate);
1777 extradata[2] = 0x56;
1778 extradata[3] = 0xE5;
1779 extradata[4] = 0x80 | (sri << 3);
1780 extradata_size = 5;
1781 } else
1782 extradata_size = 2;
1783 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - FF_INPUT_BUFFER_PADDING_SIZE) {
1784 /* Only ALAC's magic cookie is stored in Matroska's track headers.
1785 * Create the "atom size", "tag", and "tag version" fields the
1786 * decoder expects manually. */
1787 extradata_size = 12 + track->codec_priv.size;
1788 extradata = av_mallocz(extradata_size +
1789 FF_INPUT_BUFFER_PADDING_SIZE);
1790 if (!extradata)
1791 return AVERROR(ENOMEM);
1792 AV_WB32(extradata, extradata_size);
1793 memcpy(&extradata[4], "alac", 4);
1794 AV_WB32(&extradata[8], 0);
1795 memcpy(&extradata[12], track->codec_priv.data,
1796 track->codec_priv.size);
1797 } else if (codec_id == AV_CODEC_ID_TTA) {
1798 extradata_size = 30;
1799 extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1800 if (!extradata)
1801 return AVERROR(ENOMEM);
1802 ffio_init_context(&b, extradata, extradata_size, 1,
1803 NULL, NULL, NULL, NULL);
1804 avio_write(&b, "TTA1", 4);
1805 avio_wl16(&b, 1);
1806 avio_wl16(&b, track->audio.channels);
1807 avio_wl16(&b, track->audio.bitdepth);
1808 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
1809 return AVERROR_INVALIDDATA;
1810 avio_wl32(&b, track->audio.out_samplerate);
1811 avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
1812 track->audio.out_samplerate,
1813 AV_TIME_BASE * 1000));
1814 } else if (codec_id == AV_CODEC_ID_RV10 ||
1815 codec_id == AV_CODEC_ID_RV20 ||
1816 codec_id == AV_CODEC_ID_RV30 ||
1817 codec_id == AV_CODEC_ID_RV40) {
1818 extradata_offset = 26;
1819 } else if (codec_id == AV_CODEC_ID_RA_144) {
1820 track->audio.out_samplerate = 8000;
1821 track->audio.channels = 1;
1822 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
1823 codec_id == AV_CODEC_ID_COOK ||
1824 codec_id == AV_CODEC_ID_ATRAC3 ||
1825 codec_id == AV_CODEC_ID_SIPR)
1826 && track->codec_priv.data) {
1827 int flavor;
1828
1829 ffio_init_context(&b, track->codec_priv.data,
1830 track->codec_priv.size,
1831 0, NULL, NULL, NULL, NULL);
1832 avio_skip(&b, 22);
1833 flavor = avio_rb16(&b);
1834 track->audio.coded_framesize = avio_rb32(&b);
1835 avio_skip(&b, 12);
1836 track->audio.sub_packet_h = avio_rb16(&b);
1837 track->audio.frame_size = avio_rb16(&b);
1838 track->audio.sub_packet_size = avio_rb16(&b);
1839 if (flavor < 0 ||
1840 track->audio.coded_framesize <= 0 ||
1841 track->audio.sub_packet_h <= 0 ||
1842 track->audio.frame_size <= 0 ||
1843 track->audio.sub_packet_size <= 0)
1844 return AVERROR_INVALIDDATA;
1845 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
1846 track->audio.frame_size);
1847 if (!track->audio.buf)
1848 return AVERROR(ENOMEM);
1849 if (codec_id == AV_CODEC_ID_RA_288) {
1850 st->codec->block_align = track->audio.coded_framesize;
1851 track->codec_priv.size = 0;
1852 } else {
1853 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1854 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1855 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1856 st->codec->bit_rate = sipr_bit_rate[flavor];
1857 }
1858 st->codec->block_align = track->audio.sub_packet_size;
1859 extradata_offset = 78;
1860 }
1861 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
1862 ret = matroska_parse_flac(s, track, &extradata_offset);
1863 if (ret < 0)
1864 return ret;
1865 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
1866 fourcc = AV_RL32(track->codec_priv.data);
1867 }
1868 track->codec_priv.size -= extradata_offset;
1869
1870 if (codec_id == AV_CODEC_ID_NONE)
1871 av_log(matroska->ctx, AV_LOG_INFO,
1872 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1873
1874 if (track->time_scale < 0.01)
1875 track->time_scale = 1.0;
1876 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
1877 1000 * 1000 * 1000); /* 64 bit pts in ns */
1878
1879 /* convert the delay from ns to the track timebase */
1880 track->codec_delay = av_rescale_q(track->codec_delay,
1881 (AVRational){ 1, 1000000000 },
1882 st->time_base);
1883
1884 st->codec->codec_id = codec_id;
1885
1886 if (strcmp(track->language, "und"))
1887 av_dict_set(&st->metadata, "language", track->language, 0);
1888 av_dict_set(&st->metadata, "title", track->name, 0);
1889
1890 if (track->flag_default)
1891 st->disposition |= AV_DISPOSITION_DEFAULT;
1892 if (track->flag_forced)
1893 st->disposition |= AV_DISPOSITION_FORCED;
1894
1895 if (!st->codec->extradata) {
1896 if (extradata) {
1897 st->codec->extradata = extradata;
1898 st->codec->extradata_size = extradata_size;
1899 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
1900 if (ff_alloc_extradata(st->codec, track->codec_priv.size))
1901 return AVERROR(ENOMEM);
1902 memcpy(st->codec->extradata,
1903 track->codec_priv.data + extradata_offset,
1904 track->codec_priv.size);
1905 }
1906 }
1907
1908 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1909 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
1910
1911 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1912 st->codec->codec_tag = fourcc;
1913 if (bit_depth >= 0)
1914 st->codec->bits_per_coded_sample = bit_depth;
1915 st->codec->width = track->video.pixel_width;
1916 st->codec->height = track->video.pixel_height;
1917 av_reduce(&st->sample_aspect_ratio.num,
1918 &st->sample_aspect_ratio.den,
1919 st->codec->height * track->video.display_width,
1920 st->codec->width * track->video.display_height,
1921 255);
1922 if (st->codec->codec_id != AV_CODEC_ID_HEVC)
1923 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1924
1925 if (track->default_duration) {
1926 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1927 1000000000, track->default_duration, 30000);
1928#if FF_API_R_FRAME_RATE
1929 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000L
1930 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5L)
1931 st->r_frame_rate = st->avg_frame_rate;
1932#endif
1933 }
1934
1935 /* export stereo mode flag as metadata tag */
1936 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
1937 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
1938
1939 /* export alpha mode flag as metadata tag */
1940 if (track->video.alpha_mode)
1941 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
1942
1943 /* if we have virtual track, mark the real tracks */
1944 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
1945 char buf[32];
1946 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
1947 continue;
1948 snprintf(buf, sizeof(buf), "%s_%d",
1949 ff_matroska_video_stereo_plane[planes[j].type], i);
1950 for (k=0; k < matroska->tracks.nb_elem; k++)
1951 if (planes[j].uid == tracks[k].uid) {
1952 av_dict_set(&s->streams[k]->metadata,
1953 "stereo_mode", buf, 0);
1954 break;
1955 }
1956 }
1957 // add stream level stereo3d side data if it is a supported format
1958 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
1959 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
1960 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
1961 if (ret < 0)
1962 return ret;
1963 }
1964 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1965 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1966 st->codec->sample_rate = track->audio.out_samplerate;
1967 st->codec->channels = track->audio.channels;
1968 if (!st->codec->bits_per_coded_sample)
1969 st->codec->bits_per_coded_sample = track->audio.bitdepth;
1970 if (st->codec->codec_id != AV_CODEC_ID_AAC)
1971 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1972 if (track->codec_delay > 0) {
1973 st->codec->delay = av_rescale_q(track->codec_delay,
1974 st->time_base,
1975 (AVRational){1, st->codec->sample_rate});
1976 }
1977 if (track->seek_preroll > 0) {
1978 av_codec_set_seek_preroll(st->codec,
1979 av_rescale_q(track->seek_preroll,
1980 (AVRational){1, 1000000000},
1981 (AVRational){1, st->codec->sample_rate}));
1982 }
1983 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
1984 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1985
1986 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
1987 st->disposition |= AV_DISPOSITION_CAPTIONS;
1988 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
1989 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
1990 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
1991 st->disposition |= AV_DISPOSITION_METADATA;
1992 }
1993 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1994 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1995 if (st->codec->codec_id == AV_CODEC_ID_ASS)
1996 matroska->contains_ssa = 1;
1997 }
1998 }
1999
2000 return 0;
2001}
2002
2003static int matroska_read_header(AVFormatContext *s)
2004{
2005 MatroskaDemuxContext *matroska = s->priv_data;
2006 EbmlList *attachments_list = &matroska->attachments;
2007 EbmlList *chapters_list = &matroska->chapters;
2008 MatroskaAttachment *attachments;
2009 MatroskaChapter *chapters;
2010 uint64_t max_start = 0;
2011 int64_t pos;
2012 Ebml ebml = { 0 };
2013 int i, j, res;
2014
2015 matroska->ctx = s;
2016
2017 /* First read the EBML header. */
2018 if (ebml_parse(matroska, ebml_syntax, &ebml) ||
2019 ebml.version > EBML_VERSION ||
2020 ebml.max_size > sizeof(uint64_t) ||
2021 ebml.id_length > sizeof(uint32_t) ||
2022 ebml.doctype_version > 3 ||
2023 !ebml.doctype) {
2024 av_log(matroska->ctx, AV_LOG_ERROR,
2025 "EBML header using unsupported features\n"
2026 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2027 ebml.version, ebml.doctype, ebml.doctype_version);
2028 ebml_free(ebml_syntax, &ebml);
2029 return AVERROR_PATCHWELCOME;
2030 } else if (ebml.doctype_version == 3) {
2031 av_log(matroska->ctx, AV_LOG_WARNING,
2032 "EBML header using unsupported features\n"
2033 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2034 ebml.version, ebml.doctype, ebml.doctype_version);
2035 }
2036 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2037 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2038 break;
2039 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2040 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2041 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2042 ebml_free(ebml_syntax, &ebml);
2043 return AVERROR_INVALIDDATA;
2044 }
2045 }
2046 ebml_free(ebml_syntax, &ebml);
2047
2048 /* The next thing is a segment. */
2049 pos = avio_tell(matroska->ctx->pb);
2050 res = ebml_parse(matroska, matroska_segments, matroska);
2051 // try resyncing until we find a EBML_STOP type element.
2052 while (res != 1) {
2053 res = matroska_resync(matroska, pos);
2054 if (res < 0)
2055 return res;
2056 pos = avio_tell(matroska->ctx->pb);
2057 res = ebml_parse(matroska, matroska_segment, matroska);
2058 }
2059 matroska_execute_seekhead(matroska);
2060
2061 if (!matroska->time_scale)
2062 matroska->time_scale = 1000000;
2063 if (matroska->duration)
2064 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2065 1000 / AV_TIME_BASE;
2066 av_dict_set(&s->metadata, "title", matroska->title, 0);
2067 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2068
2069 if (matroska->date_utc.size == 8)
2070 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2071
2072 res = matroska_parse_tracks(s);
2073 if (res < 0)
2074 return res;
2075
2076 attachments = attachments_list->elem;
2077 for (j = 0; j < attachments_list->nb_elem; j++) {
2078 if (!(attachments[j].filename && attachments[j].mime &&
2079 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2080 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2081 } else {
2082 AVStream *st = avformat_new_stream(s, NULL);
2083 if (!st)
2084 break;
2085 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2086 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2087 st->codec->codec_id = AV_CODEC_ID_NONE;
2088 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2089 if (ff_alloc_extradata(st->codec, attachments[j].bin.size))
2090 break;
2091 memcpy(st->codec->extradata, attachments[j].bin.data,
2092 attachments[j].bin.size);
2093
2094 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2095 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2096 strlen(ff_mkv_mime_tags[i].str))) {
2097 st->codec->codec_id = ff_mkv_mime_tags[i].id;
2098 break;
2099 }
2100 }
2101 attachments[j].stream = st;
2102 }
2103 }
2104
2105 chapters = chapters_list->elem;
2106 for (i = 0; i < chapters_list->nb_elem; i++)
2107 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2108 (max_start == 0 || chapters[i].start > max_start)) {
2109 chapters[i].chapter =
2110 avpriv_new_chapter(s, chapters[i].uid,
2111 (AVRational) { 1, 1000000000 },
2112 chapters[i].start, chapters[i].end,
2113 chapters[i].title);
2114 if (chapters[i].chapter) {
2115 av_dict_set(&chapters[i].chapter->metadata,
2116 "title", chapters[i].title, 0);
2117 }
2118 max_start = chapters[i].start;
2119 }
2120
2121 matroska_add_index_entries(matroska);
2122
2123 matroska_convert_tags(s);
2124
2125 return 0;
2126}
2127
2128/*
2129 * Put one packet in an application-supplied AVPacket struct.
2130 * Returns 0 on success or -1 on failure.
2131 */
2132static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2133 AVPacket *pkt)
2134{
2135 if (matroska->num_packets > 0) {
2136 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2137 av_free(matroska->packets[0]);
2138 if (matroska->num_packets > 1) {
2139 void *newpackets;
2140 memmove(&matroska->packets[0], &matroska->packets[1],
2141 (matroska->num_packets - 1) * sizeof(AVPacket *));
2142 newpackets = av_realloc(matroska->packets,
2143 (matroska->num_packets - 1) *
2144 sizeof(AVPacket *));
2145 if (newpackets)
2146 matroska->packets = newpackets;
2147 } else {
2148 av_freep(&matroska->packets);
2149 matroska->prev_pkt = NULL;
2150 }
2151 matroska->num_packets--;
2152 return 0;
2153 }
2154
2155 return -1;
2156}
2157
2158/*
2159 * Free all packets in our internal queue.
2160 */
2161static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2162{
2163 matroska->prev_pkt = NULL;
2164 if (matroska->packets) {
2165 int n;
2166 for (n = 0; n < matroska->num_packets; n++) {
2167 av_free_packet(matroska->packets[n]);
2168 av_free(matroska->packets[n]);
2169 }
2170 av_freep(&matroska->packets);
2171 matroska->num_packets = 0;
2172 }
2173}
2174
2175static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2176 int *buf_size, int type,
2177 uint32_t **lace_buf, int *laces)
2178{
2179 int res = 0, n, size = *buf_size;
2180 uint8_t *data = *buf;
2181 uint32_t *lace_size;
2182
2183 if (!type) {
2184 *laces = 1;
2185 *lace_buf = av_mallocz(sizeof(int));
2186 if (!*lace_buf)
2187 return AVERROR(ENOMEM);
2188
2189 *lace_buf[0] = size;
2190 return 0;
2191 }
2192
2193 av_assert0(size > 0);
2194 *laces = *data + 1;
2195 data += 1;
2196 size -= 1;
2197 lace_size = av_mallocz(*laces * sizeof(int));
2198 if (!lace_size)
2199 return AVERROR(ENOMEM);
2200
2201 switch (type) {
2202 case 0x1: /* Xiph lacing */
2203 {
2204 uint8_t temp;
2205 uint32_t total = 0;
2206 for (n = 0; res == 0 && n < *laces - 1; n++) {
2207 while (1) {
2208 if (size <= total) {
2209 res = AVERROR_INVALIDDATA;
2210 break;
2211 }
2212 temp = *data;
2213 total += temp;
2214 lace_size[n] += temp;
2215 data += 1;
2216 size -= 1;
2217 if (temp != 0xff)
2218 break;
2219 }
2220 }
2221 if (size <= total) {
2222 res = AVERROR_INVALIDDATA;
2223 break;
2224 }
2225
2226 lace_size[n] = size - total;
2227 break;
2228 }
2229
2230 case 0x2: /* fixed-size lacing */
2231 if (size % (*laces)) {
2232 res = AVERROR_INVALIDDATA;
2233 break;
2234 }
2235 for (n = 0; n < *laces; n++)
2236 lace_size[n] = size / *laces;
2237 break;
2238
2239 case 0x3: /* EBML lacing */
2240 {
2241 uint64_t num;
2242 uint64_t total;
2243 n = matroska_ebmlnum_uint(matroska, data, size, &num);
2244 if (n < 0 || num > INT_MAX) {
2245 av_log(matroska->ctx, AV_LOG_INFO,
2246 "EBML block data error\n");
2247 res = n<0 ? n : AVERROR_INVALIDDATA;
2248 break;
2249 }
2250 data += n;
2251 size -= n;
2252 total = lace_size[0] = num;
2253 for (n = 1; res == 0 && n < *laces - 1; n++) {
2254 int64_t snum;
2255 int r;
2256 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2257 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
2258 av_log(matroska->ctx, AV_LOG_INFO,
2259 "EBML block data error\n");
2260 res = r<0 ? r : AVERROR_INVALIDDATA;
2261 break;
2262 }
2263 data += r;
2264 size -= r;
2265 lace_size[n] = lace_size[n - 1] + snum;
2266 total += lace_size[n];
2267 }
2268 if (size <= total) {
2269 res = AVERROR_INVALIDDATA;
2270 break;
2271 }
2272 lace_size[*laces - 1] = size - total;
2273 break;
2274 }
2275 }
2276
2277 *buf = data;
2278 *lace_buf = lace_size;
2279 *buf_size = size;
2280
2281 return res;
2282}
2283
2284static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2285 MatroskaTrack *track, AVStream *st,
2286 uint8_t *data, int size, uint64_t timecode,
2287 int64_t pos)
2288{
2289 int a = st->codec->block_align;
2290 int sps = track->audio.sub_packet_size;
2291 int cfs = track->audio.coded_framesize;
2292 int h = track->audio.sub_packet_h;
2293 int y = track->audio.sub_packet_cnt;
2294 int w = track->audio.frame_size;
2295 int x;
2296
2297 if (!track->audio.pkt_cnt) {
2298 if (track->audio.sub_packet_cnt == 0)
2299 track->audio.buf_timecode = timecode;
2300 if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2301 if (size < cfs * h / 2) {
2302 av_log(matroska->ctx, AV_LOG_ERROR,
2303 "Corrupt int4 RM-style audio packet size\n");
2304 return AVERROR_INVALIDDATA;
2305 }
2306 for (x = 0; x < h / 2; x++)
2307 memcpy(track->audio.buf + x * 2 * w + y * cfs,
2308 data + x * cfs, cfs);
2309 } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2310 if (size < w) {
2311 av_log(matroska->ctx, AV_LOG_ERROR,
2312 "Corrupt sipr RM-style audio packet size\n");
2313 return AVERROR_INVALIDDATA;
2314 }
2315 memcpy(track->audio.buf + y * w, data, w);
2316 } else {
2317 if (size < sps * w / sps || h<=0 || w%sps) {
2318 av_log(matroska->ctx, AV_LOG_ERROR,
2319 "Corrupt generic RM-style audio packet size\n");
2320 return AVERROR_INVALIDDATA;
2321 }
2322 for (x = 0; x < w / sps; x++)
2323 memcpy(track->audio.buf +
2324 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
2325 data + x * sps, sps);
2326 }
2327
2328 if (++track->audio.sub_packet_cnt >= h) {
2329 if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2330 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2331 track->audio.sub_packet_cnt = 0;
2332 track->audio.pkt_cnt = h * w / a;
2333 }
2334 }
2335
2336 while (track->audio.pkt_cnt) {
2337 int ret;
2338 AVPacket *pkt = av_mallocz(sizeof(AVPacket));
2339 if (!pkt)
2340 return AVERROR(ENOMEM);
2341
2342 ret = av_new_packet(pkt, a);
2343 if (ret < 0) {
2344 av_free(pkt);
2345 return ret;
2346 }
2347 memcpy(pkt->data,
2348 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
2349 a);
2350 pkt->pts = track->audio.buf_timecode;
2351 track->audio.buf_timecode = AV_NOPTS_VALUE;
2352 pkt->pos = pos;
2353 pkt->stream_index = st->index;
2354 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2355 }
2356
2357 return 0;
2358}
2359
2360/* reconstruct full wavpack blocks from mangled matroska ones */
2361static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
2362 uint8_t **pdst, int *size)
2363{
2364 uint8_t *dst = NULL;
2365 int dstlen = 0;
2366 int srclen = *size;
2367 uint32_t samples;
2368 uint16_t ver;
2369 int ret, offset = 0;
2370
2371 if (srclen < 12 || track->stream->codec->extradata_size < 2)
2372 return AVERROR_INVALIDDATA;
2373
2374 ver = AV_RL16(track->stream->codec->extradata);
2375
2376 samples = AV_RL32(src);
2377 src += 4;
2378 srclen -= 4;
2379
2380 while (srclen >= 8) {
2381 int multiblock;
2382 uint32_t blocksize;
2383 uint8_t *tmp;
2384
2385 uint32_t flags = AV_RL32(src);
2386 uint32_t crc = AV_RL32(src + 4);
2387 src += 8;
2388 srclen -= 8;
2389
2390 multiblock = (flags & 0x1800) != 0x1800;
2391 if (multiblock) {
2392 if (srclen < 4) {
2393 ret = AVERROR_INVALIDDATA;
2394 goto fail;
2395 }
2396 blocksize = AV_RL32(src);
2397 src += 4;
2398 srclen -= 4;
2399 } else
2400 blocksize = srclen;
2401
2402 if (blocksize > srclen) {
2403 ret = AVERROR_INVALIDDATA;
2404 goto fail;
2405 }
2406
2407 tmp = av_realloc(dst, dstlen + blocksize + 32);
2408 if (!tmp) {
2409 ret = AVERROR(ENOMEM);
2410 goto fail;
2411 }
2412 dst = tmp;
2413 dstlen += blocksize + 32;
2414
2415 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2416 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2417 AV_WL16(dst + offset + 8, ver); // version
2418 AV_WL16(dst + offset + 10, 0); // track/index_no
2419 AV_WL32(dst + offset + 12, 0); // total samples
2420 AV_WL32(dst + offset + 16, 0); // block index
2421 AV_WL32(dst + offset + 20, samples); // number of samples
2422 AV_WL32(dst + offset + 24, flags); // flags
2423 AV_WL32(dst + offset + 28, crc); // crc
2424 memcpy(dst + offset + 32, src, blocksize); // block data
2425
2426 src += blocksize;
2427 srclen -= blocksize;
2428 offset += blocksize + 32;
2429 }
2430
2431 *pdst = dst;
2432 *size = dstlen;
2433
2434 return 0;
2435
2436fail:
2437 av_freep(&dst);
2438 return ret;
2439}
2440
2441static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
2442 MatroskaTrack *track,
2443 AVStream *st,
2444 uint8_t *data, int data_len,
2445 uint64_t timecode,
2446 uint64_t duration,
2447 int64_t pos)
2448{
2449 AVPacket *pkt;
2450 uint8_t *id, *settings, *text, *buf;
2451 int id_len, settings_len, text_len;
2452 uint8_t *p, *q;
2453 int err;
2454
2455 if (data_len <= 0)
2456 return AVERROR_INVALIDDATA;
2457
2458 p = data;
2459 q = data + data_len;
2460
2461 id = p;
2462 id_len = -1;
2463 while (p < q) {
2464 if (*p == '\r' || *p == '\n') {
2465 id_len = p - id;
2466 if (*p == '\r')
2467 p++;
2468 break;
2469 }
2470 p++;
2471 }
2472
2473 if (p >= q || *p != '\n')
2474 return AVERROR_INVALIDDATA;
2475 p++;
2476
2477 settings = p;
2478 settings_len = -1;
2479 while (p < q) {
2480 if (*p == '\r' || *p == '\n') {
2481 settings_len = p - settings;
2482 if (*p == '\r')
2483 p++;
2484 break;
2485 }
2486 p++;
2487 }
2488
2489 if (p >= q || *p != '\n')
2490 return AVERROR_INVALIDDATA;
2491 p++;
2492
2493 text = p;
2494 text_len = q - p;
2495 while (text_len > 0) {
2496 const int len = text_len - 1;
2497 const uint8_t c = p[len];
2498 if (c != '\r' && c != '\n')
2499 break;
2500 text_len = len;
2501 }
2502
2503 if (text_len <= 0)
2504 return AVERROR_INVALIDDATA;
2505
2506 pkt = av_mallocz(sizeof(*pkt));
2507 err = av_new_packet(pkt, text_len);
2508 if (err < 0) {
2509 av_free(pkt);
2510 return AVERROR(err);
2511 }
2512
2513 memcpy(pkt->data, text, text_len);
2514
2515 if (id_len > 0) {
2516 buf = av_packet_new_side_data(pkt,
2517 AV_PKT_DATA_WEBVTT_IDENTIFIER,
2518 id_len);
2519 if (!buf) {
2520 av_free(pkt);
2521 return AVERROR(ENOMEM);
2522 }
2523 memcpy(buf, id, id_len);
2524 }
2525
2526 if (settings_len > 0) {
2527 buf = av_packet_new_side_data(pkt,
2528 AV_PKT_DATA_WEBVTT_SETTINGS,
2529 settings_len);
2530 if (!buf) {
2531 av_free(pkt);
2532 return AVERROR(ENOMEM);
2533 }
2534 memcpy(buf, settings, settings_len);
2535 }
2536
2537 // Do we need this for subtitles?
2538 // pkt->flags = AV_PKT_FLAG_KEY;
2539
2540 pkt->stream_index = st->index;
2541 pkt->pts = timecode;
2542
2543 // Do we need this for subtitles?
2544 // pkt->dts = timecode;
2545
2546 pkt->duration = duration;
2547 pkt->pos = pos;
2548
2549 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2550 matroska->prev_pkt = pkt;
2551
2552 return 0;
2553}
2554
2555static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2556 MatroskaTrack *track, AVStream *st,
2557 uint8_t *data, int pkt_size,
2558 uint64_t timecode, uint64_t lace_duration,
2559 int64_t pos, int is_keyframe,
2560 uint8_t *additional, uint64_t additional_id, int additional_size,
2561 int64_t discard_padding)
2562{
2563 MatroskaTrackEncoding *encodings = track->encodings.elem;
2564 uint8_t *pkt_data = data;
2565 int offset = 0, res;
2566 AVPacket *pkt;
2567
2568 if (encodings && !encodings->type && encodings->scope & 1) {
2569 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2570 if (res < 0)
2571 return res;
2572 }
2573
2574 if (st->codec->codec_id == AV_CODEC_ID_WAVPACK) {
2575 uint8_t *wv_data;
2576 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
2577 if (res < 0) {
2578 av_log(matroska->ctx, AV_LOG_ERROR,
2579 "Error parsing a wavpack block.\n");
2580 goto fail;
2581 }
2582 if (pkt_data != data)
2583 av_freep(&pkt_data);
2584 pkt_data = wv_data;
2585 }
2586
2587 if (st->codec->codec_id == AV_CODEC_ID_PRORES &&
2588 AV_RB32(&data[4]) != MKBETAG('i', 'c', 'p', 'f'))
2589 offset = 8;
2590
2591 pkt = av_mallocz(sizeof(AVPacket));
2592 /* XXX: prevent data copy... */
2593 if (av_new_packet(pkt, pkt_size + offset) < 0) {
2594 av_free(pkt);
2595 res = AVERROR(ENOMEM);
2596 goto fail;
2597 }
2598
2599 if (st->codec->codec_id == AV_CODEC_ID_PRORES && offset == 8) {
2600 uint8_t *buf = pkt->data;
2601 bytestream_put_be32(&buf, pkt_size);
2602 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2603 }
2604
2605 memcpy(pkt->data + offset, pkt_data, pkt_size);
2606
2607 if (pkt_data != data)
2608 av_freep(&pkt_data);
2609
2610 pkt->flags = is_keyframe;
2611 pkt->stream_index = st->index;
2612
2613 if (additional_size > 0) {
2614 uint8_t *side_data = av_packet_new_side_data(pkt,
2615 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
2616 additional_size + 8);
2617 if (!side_data) {
2618 av_free_packet(pkt);
2619 av_free(pkt);
2620 return AVERROR(ENOMEM);
2621 }
2622 AV_WB64(side_data, additional_id);
2623 memcpy(side_data + 8, additional, additional_size);
2624 }
2625
2626 if (discard_padding) {
2627 uint8_t *side_data = av_packet_new_side_data(pkt,
2628 AV_PKT_DATA_SKIP_SAMPLES,
2629 10);
2630 if (!side_data) {
2631 av_free_packet(pkt);
2632 av_free(pkt);
2633 return AVERROR(ENOMEM);
2634 }
2635 AV_WL32(side_data, 0);
2636 AV_WL32(side_data + 4, av_rescale_q(discard_padding,
2637 (AVRational){1, 1000000000},
2638 (AVRational){1, st->codec->sample_rate}));
2639 }
2640
2641 if (track->ms_compat)
2642 pkt->dts = timecode;
2643 else
2644 pkt->pts = timecode;
2645 pkt->pos = pos;
2646 if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2647 /*
2648 * For backward compatibility.
2649 * Historically, we have put subtitle duration
2650 * in convergence_duration, on the off chance
2651 * that the time_scale is less than 1us, which
2652 * could result in a 32bit overflow on the
2653 * normal duration field.
2654 */
2655 pkt->convergence_duration = lace_duration;
2656 }
2657
2658 if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE ||
2659 lace_duration <= INT_MAX) {
2660 /*
2661 * For non subtitle tracks, just store the duration
2662 * as normal.
2663 *
2664 * If it's a subtitle track and duration value does
2665 * not overflow a uint32, then also store it normally.
2666 */
2667 pkt->duration = lace_duration;
2668 }
2669
2670 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2671 matroska->prev_pkt = pkt;
2672
2673 return 0;
2674
2675fail:
2676 if (pkt_data != data)
2677 av_freep(&pkt_data);
2678 return res;
2679}
2680
2681static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
2682 int size, int64_t pos, uint64_t cluster_time,
2683 uint64_t block_duration, int is_keyframe,
2684 uint8_t *additional, uint64_t additional_id, int additional_size,
2685 int64_t cluster_pos, int64_t discard_padding)
2686{
2687 uint64_t timecode = AV_NOPTS_VALUE;
2688 MatroskaTrack *track;
2689 int res = 0;
2690 AVStream *st;
2691 int16_t block_time;
2692 uint32_t *lace_size = NULL;
2693 int n, flags, laces = 0;
2694 uint64_t num;
2695 int trust_default_duration = 1;
2696
2697 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2698 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2699 return n;
2700 }
2701 data += n;
2702 size -= n;
2703
2704 track = matroska_find_track_by_num(matroska, num);
2705 if (!track || !track->stream) {
2706 av_log(matroska->ctx, AV_LOG_INFO,
2707 "Invalid stream %"PRIu64" or size %u\n", num, size);
2708 return AVERROR_INVALIDDATA;
2709 } else if (size <= 3)
2710 return 0;
2711 st = track->stream;
2712 if (st->discard >= AVDISCARD_ALL)
2713 return res;
2714 av_assert1(block_duration != AV_NOPTS_VALUE);
2715
2716 block_time = sign_extend(AV_RB16(data), 16);
2717 data += 2;
2718 flags = *data++;
2719 size -= 3;
2720 if (is_keyframe == -1)
2721 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2722
2723 if (cluster_time != (uint64_t) -1 &&
2724 (block_time >= 0 || cluster_time >= -block_time)) {
2725 timecode = cluster_time + block_time - track->codec_delay;
2726 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2727 timecode < track->end_timecode)
2728 is_keyframe = 0; /* overlapping subtitles are not key frame */
2729 if (is_keyframe)
2730 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
2731 AVINDEX_KEYFRAME);
2732 }
2733
2734 if (matroska->skip_to_keyframe &&
2735 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2736 if (timecode < matroska->skip_to_timecode)
2737 return res;
2738 if (is_keyframe)
2739 matroska->skip_to_keyframe = 0;
2740 else if (!st->skip_to_keyframe) {
2741 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2742 matroska->skip_to_keyframe = 0;
2743 }
2744 }
2745
2746 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
2747 &lace_size, &laces);
2748
2749 if (res)
2750 goto end;
2751
2752 if (track->audio.samplerate == 8000) {
2753 // If this is needed for more codecs, then add them here
2754 if (st->codec->codec_id == AV_CODEC_ID_AC3) {
2755 if (track->audio.samplerate != st->codec->sample_rate || !st->codec->frame_size)
2756 trust_default_duration = 0;
2757 }
2758 }
2759
2760 if (!block_duration && trust_default_duration)
2761 block_duration = track->default_duration * laces / matroska->time_scale;
2762
2763 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
2764 track->end_timecode =
2765 FFMAX(track->end_timecode, timecode + block_duration);
2766
2767 for (n = 0; n < laces; n++) {
2768 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
2769
2770 if (lace_size[n] > size) {
2771 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2772 break;
2773 }
2774
2775 if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2776 st->codec->codec_id == AV_CODEC_ID_COOK ||
2777 st->codec->codec_id == AV_CODEC_ID_SIPR ||
2778 st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2779 st->codec->block_align && track->audio.sub_packet_size) {
2780 res = matroska_parse_rm_audio(matroska, track, st, data,
2781 lace_size[n],
2782 timecode, pos);
2783 if (res)
2784 goto end;
2785
2786 } else if (st->codec->codec_id == AV_CODEC_ID_WEBVTT) {
2787 res = matroska_parse_webvtt(matroska, track, st,
2788 data, lace_size[n],
2789 timecode, lace_duration,
2790 pos);
2791 if (res)
2792 goto end;
2793 } else {
2794 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2795 timecode, lace_duration, pos,
2796 !n ? is_keyframe : 0,
2797 additional, additional_id, additional_size,
2798 discard_padding);
2799 if (res)
2800 goto end;
2801 }
2802
2803 if (timecode != AV_NOPTS_VALUE)
2804 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2805 data += lace_size[n];
2806 size -= lace_size[n];
2807 }
2808
2809end:
2810 av_free(lace_size);
2811 return res;
2812}
2813
2814static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
2815{
2816 EbmlList *blocks_list;
2817 MatroskaBlock *blocks;
2818 int i, res;
2819 res = ebml_parse(matroska,
2820 matroska_cluster_incremental_parsing,
2821 &matroska->current_cluster);
2822 if (res == 1) {
2823 /* New Cluster */
2824 if (matroska->current_cluster_pos)
2825 ebml_level_end(matroska);
2826 ebml_free(matroska_cluster, &matroska->current_cluster);
2827 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2828 matroska->current_cluster_num_blocks = 0;
2829 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2830 matroska->prev_pkt = NULL;
2831 /* sizeof the ID which was already read */
2832 if (matroska->current_id)
2833 matroska->current_cluster_pos -= 4;
2834 res = ebml_parse(matroska,
2835 matroska_clusters_incremental,
2836 &matroska->current_cluster);
2837 /* Try parsing the block again. */
2838 if (res == 1)
2839 res = ebml_parse(matroska,
2840 matroska_cluster_incremental_parsing,
2841 &matroska->current_cluster);
2842 }
2843
2844 if (!res &&
2845 matroska->current_cluster_num_blocks <
2846 matroska->current_cluster.blocks.nb_elem) {
2847 blocks_list = &matroska->current_cluster.blocks;
2848 blocks = blocks_list->elem;
2849
2850 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2851 i = blocks_list->nb_elem - 1;
2852 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2853 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2854 uint8_t* additional = blocks[i].additional.size > 0 ?
2855 blocks[i].additional.data : NULL;
2856 if (!blocks[i].non_simple)
2857 blocks[i].duration = 0;
2858 res = matroska_parse_block(matroska, blocks[i].bin.data,
2859 blocks[i].bin.size, blocks[i].bin.pos,
2860 matroska->current_cluster.timecode,
2861 blocks[i].duration, is_keyframe,
2862 additional, blocks[i].additional_id,
2863 blocks[i].additional.size,
2864 matroska->current_cluster_pos,
2865 blocks[i].discard_padding);
2866 }
2867 }
2868
2869 return res;
2870}
2871
2872static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
2873{
2874 MatroskaCluster cluster = { 0 };
2875 EbmlList *blocks_list;
2876 MatroskaBlock *blocks;
2877 int i, res;
2878 int64_t pos;
2879
2880 if (!matroska->contains_ssa)
2881 return matroska_parse_cluster_incremental(matroska);
2882 pos = avio_tell(matroska->ctx->pb);
2883 matroska->prev_pkt = NULL;
2884 if (matroska->current_id)
2885 pos -= 4; /* sizeof the ID which was already read */
2886 res = ebml_parse(matroska, matroska_clusters, &cluster);
2887 blocks_list = &cluster.blocks;
2888 blocks = blocks_list->elem;
2889 for (i = 0; i < blocks_list->nb_elem; i++)
2890 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2891 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2892 res = matroska_parse_block(matroska, blocks[i].bin.data,
2893 blocks[i].bin.size, blocks[i].bin.pos,
2894 cluster.timecode, blocks[i].duration,
2895 is_keyframe, NULL, 0, 0, pos,
2896 blocks[i].discard_padding);
2897 }
2898 ebml_free(matroska_cluster, &cluster);
2899 return res;
2900}
2901
2902static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
2903{
2904 MatroskaDemuxContext *matroska = s->priv_data;
2905
2906 while (matroska_deliver_packet(matroska, pkt)) {
2907 int64_t pos = avio_tell(matroska->ctx->pb);
2908 if (matroska->done)
2909 return AVERROR_EOF;
2910 if (matroska_parse_cluster(matroska) < 0)
2911 matroska_resync(matroska, pos);
2912 }
2913
2914 return 0;
2915}
2916
2917static int matroska_read_seek(AVFormatContext *s, int stream_index,
2918 int64_t timestamp, int flags)
2919{
2920 MatroskaDemuxContext *matroska = s->priv_data;
2921 MatroskaTrack *tracks = NULL;
2922 AVStream *st = s->streams[stream_index];
2923 int i, index, index_sub, index_min;
2924
2925 /* Parse the CUES now since we need the index data to seek. */
2926 if (matroska->cues_parsing_deferred > 0) {
2927 matroska->cues_parsing_deferred = 0;
2928 matroska_parse_cues(matroska);
2929 }
2930
2931 if (!st->nb_index_entries)
2932 goto err;
2933 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2934
2935 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
2936 avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
2937 SEEK_SET);
2938 matroska->current_id = 0;
2939 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
2940 matroska_clear_queue(matroska);
2941 if (matroska_parse_cluster(matroska) < 0)
2942 break;
2943 }
2944 }
2945
2946 matroska_clear_queue(matroska);
2947 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
2948 goto err;
2949
2950 index_min = index;
2951 tracks = matroska->tracks.elem;
2952 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2953 tracks[i].audio.pkt_cnt = 0;
2954 tracks[i].audio.sub_packet_cnt = 0;
2955 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2956 tracks[i].end_timecode = 0;
2957 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2958 tracks[i].stream->discard != AVDISCARD_ALL) {
2959 index_sub = av_index_search_timestamp(
2960 tracks[i].stream, st->index_entries[index].timestamp,
2961 AVSEEK_FLAG_BACKWARD);
2962 while (index_sub >= 0 &&
2963 index_min > 0 &&
2964 tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
2965 st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
2966 index_min--;
2967 }
2968 }
2969
2970 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2971 matroska->current_id = 0;
2972 if (flags & AVSEEK_FLAG_ANY) {
2973 st->skip_to_keyframe = 0;
2974 matroska->skip_to_timecode = timestamp;
2975 } else {
2976 st->skip_to_keyframe = 1;
2977 matroska->skip_to_timecode = st->index_entries[index].timestamp;
2978 }
2979 matroska->skip_to_keyframe = 1;
2980 matroska->done = 0;
2981 matroska->num_levels = 0;
2982 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2983 return 0;
2984err:
2985 // slightly hackish but allows proper fallback to
2986 // the generic seeking code.
2987 matroska_clear_queue(matroska);
2988 matroska->current_id = 0;
2989 st->skip_to_keyframe =
2990 matroska->skip_to_keyframe = 0;
2991 matroska->done = 0;
2992 matroska->num_levels = 0;
2993 return -1;
2994}
2995
2996static int matroska_read_close(AVFormatContext *s)
2997{
2998 MatroskaDemuxContext *matroska = s->priv_data;
2999 MatroskaTrack *tracks = matroska->tracks.elem;
3000 int n;
3001
3002 matroska_clear_queue(matroska);
3003
3004 for (n = 0; n < matroska->tracks.nb_elem; n++)
3005 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3006 av_free(tracks[n].audio.buf);
3007 ebml_free(matroska_cluster, &matroska->current_cluster);
3008 ebml_free(matroska_segment, matroska);
3009
3010 return 0;
3011}
3012
3013typedef struct {
3014 int64_t start_time_ns;
3015 int64_t end_time_ns;
3016 int64_t start_offset;
3017 int64_t end_offset;
3018} CueDesc;
3019
3020/* This function searches all the Cues and returns the CueDesc corresponding the
3021 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3022 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3023 */
3024static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3025 MatroskaDemuxContext *matroska = s->priv_data;
3026 CueDesc cue_desc;
3027 int i;
3028 int nb_index_entries = s->streams[0]->nb_index_entries;
3029 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3030 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3031 for (i = 1; i < nb_index_entries; i++) {
3032 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3033 index_entries[i].timestamp * matroska->time_scale > ts) {
3034 break;
3035 }
3036 }
3037 --i;
3038 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3039 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3040 if (i != nb_index_entries - 1) {
3041 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3042 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3043 } else {
3044 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3045 // FIXME: this needs special handling for files where Cues appear
3046 // before Clusters. the current logic assumes Cues appear after
3047 // Clusters.
3048 cue_desc.end_offset = cues_start - matroska->segment_start;
3049 }
3050 return cue_desc;
3051}
3052
3053static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3054{
3055 MatroskaDemuxContext *matroska = s->priv_data;
3056 int64_t cluster_pos, before_pos;
3057 int index, rv = 1;
3058 if (s->streams[0]->nb_index_entries <= 0) return 0;
3059 // seek to the first cluster using cues.
3060 index = av_index_search_timestamp(s->streams[0], 0, 0);
3061 if (index < 0) return 0;
3062 cluster_pos = s->streams[0]->index_entries[index].pos;
3063 before_pos = avio_tell(s->pb);
3064 while (1) {
3065 int64_t cluster_id = 0, cluster_length = 0;
3066 AVPacket *pkt;
3067 avio_seek(s->pb, cluster_pos, SEEK_SET);
3068 // read cluster id and length
3069 ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id);
3070 ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3071 if (cluster_id != 0xF43B675) { // done with all clusters
3072 break;
3073 }
3074 avio_seek(s->pb, cluster_pos, SEEK_SET);
3075 matroska->current_id = 0;
3076 matroska_clear_queue(matroska);
3077 if (matroska_parse_cluster(matroska) < 0 ||
3078 matroska->num_packets <= 0) {
3079 break;
3080 }
3081 pkt = matroska->packets[0];
3082 cluster_pos += cluster_length + 12; // 12 is the offset of the cluster id and length.
3083 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3084 rv = 0;
3085 break;
3086 }
3087 }
3088 avio_seek(s->pb, before_pos, SEEK_SET);
3089 return rv;
3090}
3091
3092static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3093 double min_buffer, double* buffer,
3094 double* sec_to_download, AVFormatContext *s,
3095 int64_t cues_start)
3096{
3097 double nano_seconds_per_second = 1000000000.0;
3098 double time_sec = time_ns / nano_seconds_per_second;
3099 int rv = 0;
3100 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3101 int64_t end_time_ns = time_ns + time_to_search_ns;
3102 double sec_downloaded = 0.0;
3103 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3104 if (desc_curr.start_time_ns == -1)
3105 return -1;
3106 *sec_to_download = 0.0;
3107
3108 // Check for non cue start time.
3109 if (time_ns > desc_curr.start_time_ns) {
3110 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3111 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3112 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3113 double timeToDownload = (cueBytes * 8.0) / bps;
3114
3115 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3116 *sec_to_download += timeToDownload;
3117
3118 // Check if the search ends within the first cue.
3119 if (desc_curr.end_time_ns >= end_time_ns) {
3120 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3121 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3122 sec_downloaded = percent_to_sub * sec_downloaded;
3123 *sec_to_download = percent_to_sub * *sec_to_download;
3124 }
3125
3126 if ((sec_downloaded + *buffer) <= min_buffer) {
3127 return 1;
3128 }
3129
3130 // Get the next Cue.
3131 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3132 }
3133
3134 while (desc_curr.start_time_ns != -1) {
3135 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3136 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3137 double desc_sec = desc_ns / nano_seconds_per_second;
3138 double bits = (desc_bytes * 8.0);
3139 double time_to_download = bits / bps;
3140
3141 sec_downloaded += desc_sec - time_to_download;
3142 *sec_to_download += time_to_download;
3143
3144 if (desc_curr.end_time_ns >= end_time_ns) {
3145 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3146 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3147 sec_downloaded = percent_to_sub * sec_downloaded;
3148 *sec_to_download = percent_to_sub * *sec_to_download;
3149
3150 if ((sec_downloaded + *buffer) <= min_buffer)
3151 rv = 1;
3152 break;
3153 }
3154
3155 if ((sec_downloaded + *buffer) <= min_buffer) {
3156 rv = 1;
3157 break;
3158 }
3159
3160 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3161 }
3162 *buffer = *buffer + sec_downloaded;
3163 return rv;
3164}
3165
3166/* This function computes the bandwidth of the WebM file with the help of
3167 * buffer_size_after_time_downloaded() function. Both of these functions are
3168 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3169 * Matroska parsing mechanism.
3170 *
3171 * Returns the bandwidth of the file on success; -1 on error.
3172 * */
3173static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3174{
3175 MatroskaDemuxContext *matroska = s->priv_data;
3176 AVStream *st = s->streams[0];
3177 double bandwidth = 0.0;
3178 int i;
3179
3180 for (i = 0; i < st->nb_index_entries; i++) {
3181 int64_t prebuffer_ns = 1000000000;
3182 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3183 double nano_seconds_per_second = 1000000000.0;
3184 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3185 double prebuffer_bytes = 0.0;
3186 int64_t temp_prebuffer_ns = prebuffer_ns;
3187 int64_t pre_bytes, pre_ns;
3188 double pre_sec, prebuffer, bits_per_second;
3189 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3190
3191 // Start with the first Cue.
3192 CueDesc desc_end = desc_beg;
3193
3194 // Figure out how much data we have downloaded for the prebuffer. This will
3195 // be used later to adjust the bits per sample to try.
3196 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3197 // Prebuffered the entire Cue.
3198 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3199 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3200 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3201 }
3202 if (desc_end.start_time_ns == -1) {
3203 // The prebuffer is larger than the duration.
3204 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3205 return -1;
3206 bits_per_second = 0.0;
3207 } else {
3208 // The prebuffer ends in the last Cue. Estimate how much data was
3209 // prebuffered.
3210 pre_bytes = desc_end.end_offset - desc_end.start_offset;
3211 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3212 pre_sec = pre_ns / nano_seconds_per_second;
3213 prebuffer_bytes +=
3214 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3215
3216 prebuffer = prebuffer_ns / nano_seconds_per_second;
3217
3218 // Set this to 0.0 in case our prebuffer buffers the entire video.
3219 bits_per_second = 0.0;
3220 do {
3221 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
3222 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
3223 double desc_sec = desc_ns / nano_seconds_per_second;
3224 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
3225
3226 // Drop the bps by the percentage of bytes buffered.
3227 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
3228 double mod_bits_per_second = calc_bits_per_second * percent;
3229
3230 if (prebuffer < desc_sec) {
3231 double search_sec =
3232 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
3233
3234 // Add 1 so the bits per second should be a little bit greater than file
3235 // datarate.
3236 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
3237 const double min_buffer = 0.0;
3238 double buffer = prebuffer;
3239 double sec_to_download = 0.0;
3240
3241 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
3242 min_buffer, &buffer, &sec_to_download,
3243 s, cues_start);
3244 if (rv < 0) {
3245 return -1;
3246 } else if (rv == 0) {
3247 bits_per_second = (double)(bps);
3248 break;
3249 }
3250 }
3251
3252 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3253 } while (desc_end.start_time_ns != -1);
3254 }
3255 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
3256 }
3257 return (int64_t)bandwidth;
3258}
3259
3260static int webm_dash_manifest_cues(AVFormatContext *s)
3261{
3262 MatroskaDemuxContext *matroska = s->priv_data;
3263 EbmlList *seekhead_list = &matroska->seekhead;
3264 MatroskaSeekhead *seekhead = seekhead_list->elem;
3265 char *buf;
3266 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
3267 int i;
3268
3269 // determine cues start and end positions
3270 for (i = 0; i < seekhead_list->nb_elem; i++)
3271 if (seekhead[i].id == MATROSKA_ID_CUES)
3272 break;
3273
3274 if (i >= seekhead_list->nb_elem) return -1;
3275
3276 before_pos = avio_tell(matroska->ctx->pb);
3277 cues_start = seekhead[i].pos + matroska->segment_start;
3278 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
3279 // cues_end is computed as cues_start + cues_length + length of the
3280 // Cues element ID + EBML length of the Cues element. cues_end is
3281 // inclusive and the above sum is reduced by 1.
3282 uint64_t cues_length = 0, cues_id = 0, bytes_read = 0;
3283 bytes_read += ebml_read_num(matroska, matroska->ctx->pb, 4, &cues_id);
3284 bytes_read += ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
3285 cues_end = cues_start + cues_length + bytes_read - 1;
3286 }
3287 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
3288 if (cues_start == -1 || cues_end == -1) return -1;
3289
3290 // parse the cues
3291 matroska_parse_cues(matroska);
3292
3293 // cues start
3294 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
3295
3296 // cues end
3297 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
3298
3299 // bandwidth
3300 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
3301 if (bandwidth < 0) return -1;
3302 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
3303
3304 // check if all clusters start with key frames
3305 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
3306
3307 // store cue point timestamps as a comma separated list for checking subsegment alignment in
3308 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
3309 buf = av_malloc(s->streams[0]->nb_index_entries * 20 * sizeof(char));
3310 if (!buf) return -1;
3311 strcpy(buf, "");
3312 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
3313 snprintf(buf, (i + 1) * 20 * sizeof(char),
3314 "%s%" PRId64, buf, s->streams[0]->index_entries[i].timestamp);
3315 if (i != s->streams[0]->nb_index_entries - 1)
3316 strncat(buf, ",", sizeof(char));
3317 }
3318 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
3319 av_free(buf);
3320
3321 return 0;
3322}
3323
3324static int webm_dash_manifest_read_header(AVFormatContext *s)
3325{
3326 char *buf;
3327 int ret = matroska_read_header(s);
3328 MatroskaTrack *tracks;
3329 MatroskaDemuxContext *matroska = s->priv_data;
3330 if (ret) {
3331 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
3332 return -1;
3333 }
3334
3335 // initialization range
3336 // 5 is the offset of Cluster ID.
3337 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, avio_tell(s->pb) - 5, 0);
3338
3339 // basename of the file
3340 buf = strrchr(s->filename, '/');
3341 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->filename, 0);
3342
3343 // duration
3344 buf = av_asprintf("%g", matroska->duration);
3345 if (!buf) return AVERROR(ENOMEM);
3346 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
3347 av_free(buf);
3348
3349 // track number
3350 tracks = matroska->tracks.elem;
3351 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
3352
3353 // parse the cues and populate Cue related fields
3354 return webm_dash_manifest_cues(s);
3355}
3356
3357static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
3358{
3359 return AVERROR_EOF;
3360}
3361
3362AVInputFormat ff_matroska_demuxer = {
3363 .name = "matroska,webm",
3364 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
3365 .extensions = "mkv,mk3d,mka,mks",
3366 .priv_data_size = sizeof(MatroskaDemuxContext),
3367 .read_probe = matroska_probe,
3368 .read_header = matroska_read_header,
3369 .read_packet = matroska_read_packet,
3370 .read_close = matroska_read_close,
3371 .read_seek = matroska_read_seek,
3372 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
3373};
3374
3375AVInputFormat ff_webm_dash_manifest_demuxer = {
3376 .name = "webm_dash_manifest",
3377 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
3378 .priv_data_size = sizeof(MatroskaDemuxContext),
3379 .read_header = webm_dash_manifest_read_header,
3380 .read_packet = webm_dash_manifest_read_packet,
3381 .read_close = matroska_read_close,
3382};
ffmpeg Piment Noir Debian package repository