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 #if CONFIG_BZLIB
36 #include <bzlib.h>
37 #endif
38 #if CONFIG_ZLIB
39 #include <zlib.h>
40 #endif
41 
42 #include "libavutil/avstring.h"
43 #include "libavutil/base64.h"
44 #include "libavutil/dict.h"
45 #include "libavutil/intfloat.h"
46 #include "libavutil/intreadwrite.h"
47 #include "libavutil/lzo.h"
48 #include "libavutil/mathematics.h"
49 
50 #include "libavcodec/bytestream.h"
51 #include "libavcodec/flac.h"
52 #include "libavcodec/mpeg4audio.h"
53 
54 #include "avformat.h"
55 #include "avio_internal.h"
56 #include "internal.h"
57 #include "isom.h"
58 #include "matroska.h"
59 #include "oggdec.h"
60 /* For ff_codec_get_id(). */
61 #include "riff.h"
62 #include "rmsipr.h"
63 
64 typedef enum {
65     EBML_NONE,
66     EBML_UINT,
67     EBML_FLOAT,
68     EBML_STR,
69     EBML_UTF8,
70     EBML_BIN,
71     EBML_NEST,
72     EBML_PASS,
73     EBML_STOP,
74     EBML_SINT,
75     EBML_TYPE_COUNT
76 } EbmlType;
77 
78 typedef const struct EbmlSyntax {
79     uint32_t id;
80     EbmlType type;
81     int list_elem_size;
82     int data_offset;
83     union {
84         uint64_t    u;
85         double      f;
86         const char *s;
87         const struct EbmlSyntax *n;
88     } def;
89 } EbmlSyntax;
90 
91 typedef struct {
92     int nb_elem;
93     void *elem;
94 } EbmlList;
95 
96 typedef struct {
97     int      size;
98     uint8_t *data;
99     int64_t  pos;
100 } EbmlBin;
101 
102 typedef struct {
103     uint64_t version;
104     uint64_t max_size;
105     uint64_t id_length;
106     char    *doctype;
107     uint64_t doctype_version;
108 } Ebml;
109 
110 typedef struct {
111     uint64_t algo;
112     EbmlBin  settings;
113 } MatroskaTrackCompression;
114 
115 typedef struct {
116     uint64_t algo;
117     EbmlBin  key_id;
118 } MatroskaTrackEncryption;
119 
120 typedef struct {
121     uint64_t scope;
122     uint64_t type;
123     MatroskaTrackCompression compression;
124     MatroskaTrackEncryption encryption;
125 } MatroskaTrackEncoding;
126 
127 typedef struct {
128     double   frame_rate;
129     uint64_t display_width;
130     uint64_t display_height;
131     uint64_t pixel_width;
132     uint64_t pixel_height;
133     EbmlBin color_space;
134     uint64_t stereo_mode;
135     uint64_t alpha_mode;
136 } MatroskaTrackVideo;
137 
138 typedef struct {
139     double   samplerate;
140     double   out_samplerate;
141     uint64_t bitdepth;
142     uint64_t channels;
143 
144     /* real audio header (extracted from extradata) */
145     int      coded_framesize;
146     int      sub_packet_h;
147     int      frame_size;
148     int      sub_packet_size;
149     int      sub_packet_cnt;
150     int      pkt_cnt;
151     uint64_t buf_timecode;
152     uint8_t *buf;
153 } MatroskaTrackAudio;
154 
155 typedef struct {
156     uint64_t uid;
157     uint64_t type;
158 } MatroskaTrackPlane;
159 
160 typedef struct {
161     EbmlList combine_planes;
162 } MatroskaTrackOperation;
163 
164 typedef struct {
165     uint64_t num;
166     uint64_t uid;
167     uint64_t type;
168     char    *name;
169     char    *codec_id;
170     EbmlBin  codec_priv;
171     char    *language;
172     double time_scale;
173     uint64_t default_duration;
174     uint64_t flag_default;
175     uint64_t flag_forced;
176     uint64_t seek_preroll;
177     MatroskaTrackVideo video;
178     MatroskaTrackAudio audio;
179     MatroskaTrackOperation operation;
180     EbmlList encodings;
181     uint64_t codec_delay;
182 
183     AVStream *stream;
184     int64_t end_timecode;
185     int ms_compat;
186     uint64_t max_block_additional_id;
187 } MatroskaTrack;
188 
189 typedef struct {
190     uint64_t uid;
191     char *filename;
192     char *mime;
193     EbmlBin bin;
194 
195     AVStream *stream;
196 } MatroskaAttachment;
197 
198 typedef struct {
199     uint64_t start;
200     uint64_t end;
201     uint64_t uid;
202     char    *title;
203 
204     AVChapter *chapter;
205 } MatroskaChapter;
206 
207 typedef struct {
208     uint64_t track;
209     uint64_t pos;
210 } MatroskaIndexPos;
211 
212 typedef struct {
213     uint64_t time;
214     EbmlList pos;
215 } MatroskaIndex;
216 
217 typedef struct {
218     char *name;
219     char *string;
220     char *lang;
221     uint64_t def;
222     EbmlList sub;
223 } MatroskaTag;
224 
225 typedef struct {
226     char    *type;
227     uint64_t typevalue;
228     uint64_t trackuid;
229     uint64_t chapteruid;
230     uint64_t attachuid;
231 } MatroskaTagTarget;
232 
233 typedef struct {
234     MatroskaTagTarget target;
235     EbmlList tag;
236 } MatroskaTags;
237 
238 typedef struct {
239     uint64_t id;
240     uint64_t pos;
241 } MatroskaSeekhead;
242 
243 typedef struct {
244     uint64_t start;
245     uint64_t length;
246 } MatroskaLevel;
247 
248 typedef struct {
249     uint64_t timecode;
250     EbmlList blocks;
251 } MatroskaCluster;
252 
253 typedef struct {
254     AVFormatContext *ctx;
255 
256     /* EBML stuff */
257     int num_levels;
258     MatroskaLevel levels[EBML_MAX_DEPTH];
259     int level_up;
260     uint32_t current_id;
261 
262     uint64_t time_scale;
263     double   duration;
264     char    *title;
265     char    *muxingapp;
266     EbmlBin date_utc;
267     EbmlList tracks;
268     EbmlList attachments;
269     EbmlList chapters;
270     EbmlList index;
271     EbmlList tags;
272     EbmlList seekhead;
273 
274     /* byte position of the segment inside the stream */
275     int64_t segment_start;
276 
277     /* the packet queue */
278     AVPacket **packets;
279     int num_packets;
280     AVPacket *prev_pkt;
281 
282     int done;
283 
284     /* What to skip before effectively reading a packet. */
285     int skip_to_keyframe;
286     uint64_t skip_to_timecode;
287 
288     /* File has a CUES element, but we defer parsing until it is needed. */
289     int cues_parsing_deferred;
290 
291     int current_cluster_num_blocks;
292     int64_t current_cluster_pos;
293     MatroskaCluster current_cluster;
294 
295     /* File has SSA subtitles which prevent incremental cluster parsing. */
296     int contains_ssa;
297 } MatroskaDemuxContext;
298 
299 typedef struct {
300     uint64_t duration;
301     int64_t  reference;
302     uint64_t non_simple;
303     EbmlBin  bin;
304     uint64_t additional_id;
305     EbmlBin  additional;
306     int64_t discard_padding;
307 } MatroskaBlock;
308 
309 static EbmlSyntax ebml_header[] = {
310 	{ EBML_ID_EBMLREADVERSION,    EBML_UINT, 0, offsetof(Ebml, version),         { .u = EBML_VERSION } },
311     { EBML_ID_EBMLMAXSIZELENGTH,  EBML_UINT, 0, offsetof(Ebml, max_size),        { .u = 8 } },
312     { EBML_ID_EBMLMAXIDLENGTH,    EBML_UINT, 0, offsetof(Ebml, id_length),       { .u = 4 } },
313     { EBML_ID_DOCTYPE,            EBML_STR,  0, offsetof(Ebml, doctype),         { .s = "(none)" } },
314     { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
315     { EBML_ID_EBMLVERSION,        EBML_NONE },
316     { EBML_ID_DOCTYPEVERSION,     EBML_NONE },
317 	{ 0 }
318 };
319 
320 static EbmlSyntax ebml_syntax[] = {
321 	{ EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
322 	{ 0 }
323 };
324 
325 static EbmlSyntax matroska_info[] = {
326 	{ MATROSKA_ID_TIMECODESCALE, EBML_UINT,  0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
327     { MATROSKA_ID_DURATION,      EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
328     { MATROSKA_ID_TITLE,         EBML_UTF8,  0, offsetof(MatroskaDemuxContext, title) },
329     { MATROSKA_ID_WRITINGAPP,    EBML_NONE },
330     { MATROSKA_ID_MUXINGAPP,     EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
331     { MATROSKA_ID_DATEUTC,       EBML_BIN,  0, offsetof(MatroskaDemuxContext, date_utc) },
332     { MATROSKA_ID_SEGMENTUID,    EBML_NONE },
333 	{ 0 }
334 };
335 
336 static EbmlSyntax matroska_track_video[] = {
337 	{ MATROSKA_ID_VIDEOFRAMERATE,      EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
338     { MATROSKA_ID_VIDEODISPLAYWIDTH,   EBML_UINT,  0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
339     { MATROSKA_ID_VIDEODISPLAYHEIGHT,  EBML_UINT,  0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
340     { MATROSKA_ID_VIDEOPIXELWIDTH,     EBML_UINT,  0, offsetof(MatroskaTrackVideo, pixel_width) },
341     { MATROSKA_ID_VIDEOPIXELHEIGHT,    EBML_UINT,  0, offsetof(MatroskaTrackVideo, pixel_height) },
342     { MATROSKA_ID_VIDEOCOLORSPACE,     EBML_BIN,   0, offsetof(MatroskaTrackVideo, color_space) },
343     { MATROSKA_ID_VIDEOALPHAMODE,      EBML_UINT,  0, offsetof(MatroskaTrackVideo, alpha_mode) },
344     { MATROSKA_ID_VIDEOPIXELCROPB,     EBML_NONE },
345     { MATROSKA_ID_VIDEOPIXELCROPT,     EBML_NONE },
346     { MATROSKA_ID_VIDEOPIXELCROPL,     EBML_NONE },
347     { MATROSKA_ID_VIDEOPIXELCROPR,     EBML_NONE },
348     { MATROSKA_ID_VIDEODISPLAYUNIT,    EBML_NONE },
349     { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_NONE },
350     { MATROSKA_ID_VIDEOSTEREOMODE,     EBML_UINT,  0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
351     { MATROSKA_ID_VIDEOASPECTRATIO,    EBML_NONE },
352 	{ 0 }
353 };
354 
355 static EbmlSyntax matroska_track_audio[] = {
356 	{ MATROSKA_ID_AUDIOSAMPLINGFREQ,    EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
357     { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
358     { MATROSKA_ID_AUDIOBITDEPTH,        EBML_UINT,  0, offsetof(MatroskaTrackAudio, bitdepth) },
359     { MATROSKA_ID_AUDIOCHANNELS,        EBML_UINT,  0, offsetof(MatroskaTrackAudio, channels),   { .u = 1 } },
360 	{ 0 }
361 };
362 
363 static EbmlSyntax matroska_track_encoding_compression[] = {
364 	{ MATROSKA_ID_ENCODINGCOMPALGO,     EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
365     { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN,  0, offsetof(MatroskaTrackCompression, settings) },
366 	{ 0 }
367 };
368 
369 static EbmlSyntax matroska_track_encoding_encryption[] = {
370 	{ MATROSKA_ID_ENCODINGENCALGO,        EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
371     { MATROSKA_ID_ENCODINGENCKEYID,       EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
372     { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
373     { MATROSKA_ID_ENCODINGSIGALGO,        EBML_NONE },
374     { MATROSKA_ID_ENCODINGSIGHASHALGO,    EBML_NONE },
375     { MATROSKA_ID_ENCODINGSIGKEYID,       EBML_NONE },
376     { MATROSKA_ID_ENCODINGSIGNATURE,      EBML_NONE },
377 	{ 0 }
378 };
379 static EbmlSyntax matroska_track_encoding[] = {
380 	{ MATROSKA_ID_ENCODINGSCOPE,       EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope),       { .u = 1 } },
381     { MATROSKA_ID_ENCODINGTYPE,        EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type),        { .u = 0 } },
382     { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
383     { MATROSKA_ID_ENCODINGENCRYPTION,  EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption),  { .n = matroska_track_encoding_encryption } },
384     { MATROSKA_ID_ENCODINGORDER,       EBML_NONE },
385 	{ 0 }
386 };
387 
388 static EbmlSyntax matroska_track_encodings[] = {
389 	{ MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
390 	{ 0 }
391 };
392 
393 static EbmlSyntax matroska_track_plane[] = {
394     { MATROSKA_ID_TRACKPLANEUID,  EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
395     { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
396     { 0 }
397 };
398 
399 static EbmlSyntax matroska_track_combine_planes[] = {
400 	{ MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
401 	{ 0 }
402 };
403 
404 static EbmlSyntax matroska_track_operation[] = {
405 	{ MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
406 	{ 0 }
407 };
408 
409 static EbmlSyntax matroska_track[] = {
410 	{ MATROSKA_ID_TRACKNUMBER,           EBML_UINT,  0, offsetof(MatroskaTrack, num) },
411     { MATROSKA_ID_TRACKNAME,             EBML_UTF8,  0, offsetof(MatroskaTrack, name) },
412     { MATROSKA_ID_TRACKUID,              EBML_UINT,  0, offsetof(MatroskaTrack, uid) },
413     { MATROSKA_ID_TRACKTYPE,             EBML_UINT,  0, offsetof(MatroskaTrack, type) },
414     { MATROSKA_ID_CODECID,               EBML_STR,   0, offsetof(MatroskaTrack, codec_id) },
415     { MATROSKA_ID_CODECPRIVATE,          EBML_BIN,   0, offsetof(MatroskaTrack, codec_priv) },
416     { MATROSKA_ID_CODECDELAY,            EBML_UINT,  0, offsetof(MatroskaTrack, codec_delay) },
417     { MATROSKA_ID_TRACKLANGUAGE,         EBML_UTF8,  0, offsetof(MatroskaTrack, language),     { .s = "eng" } },
418     { MATROSKA_ID_TRACKDEFAULTDURATION,  EBML_UINT,  0, offsetof(MatroskaTrack, default_duration) },
419     { MATROSKA_ID_TRACKTIMECODESCALE,    EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale),   { .f = 1.0 } },
420     { MATROSKA_ID_TRACKFLAGDEFAULT,      EBML_UINT,  0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
421     { MATROSKA_ID_TRACKFLAGFORCED,       EBML_UINT,  0, offsetof(MatroskaTrack, flag_forced),  { .u = 0 } },
422     { MATROSKA_ID_TRACKVIDEO,            EBML_NEST,  0, offsetof(MatroskaTrack, video),        { .n = matroska_track_video } },
423     { MATROSKA_ID_TRACKAUDIO,            EBML_NEST,  0, offsetof(MatroskaTrack, audio),        { .n = matroska_track_audio } },
424     { MATROSKA_ID_TRACKOPERATION,        EBML_NEST,  0, offsetof(MatroskaTrack, operation),    { .n = matroska_track_operation } },
425     { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST,  0, 0,                                     { .n = matroska_track_encodings } },
426     { MATROSKA_ID_TRACKMAXBLKADDID,      EBML_UINT,  0, offsetof(MatroskaTrack, max_block_additional_id) },
427     { MATROSKA_ID_SEEKPREROLL,           EBML_UINT,  0, offsetof(MatroskaTrack, seek_preroll) },
428     { MATROSKA_ID_TRACKFLAGENABLED,      EBML_NONE },
429     { MATROSKA_ID_TRACKFLAGLACING,       EBML_NONE },
430     { MATROSKA_ID_CODECNAME,             EBML_NONE },
431     { MATROSKA_ID_CODECDECODEALL,        EBML_NONE },
432     { MATROSKA_ID_CODECINFOURL,          EBML_NONE },
433     { MATROSKA_ID_CODECDOWNLOADURL,      EBML_NONE },
434     { MATROSKA_ID_TRACKMINCACHE,         EBML_NONE },
435     { MATROSKA_ID_TRACKMAXCACHE,         EBML_NONE },
436 	{ 0 }
437 };
438 
439 static EbmlSyntax matroska_tracks[] = {
440 	{ MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
441 	{ 0 }
442 };
443 
444 static EbmlSyntax matroska_attachment[] = {
445     { MATROSKA_ID_FILEUID,      EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
446     { MATROSKA_ID_FILENAME,     EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
447     { MATROSKA_ID_FILEMIMETYPE, EBML_STR,  0, offsetof(MatroskaAttachment, mime) },
448     { MATROSKA_ID_FILEDATA,     EBML_BIN,  0, offsetof(MatroskaAttachment, bin) },
449     { MATROSKA_ID_FILEDESC,     EBML_NONE },
450     { 0 }
451 };
452 
453 static EbmlSyntax matroska_attachments[] = {
454 	{ MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
455 	{ 0 }
456 };
457 
458 static EbmlSyntax matroska_chapter_display[] = {
459     { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
460     { MATROSKA_ID_CHAPLANG,   EBML_NONE },
461     { 0 }
462 };
463 
464 static EbmlSyntax matroska_chapter_entry[] = {
465 	{ MATROSKA_ID_CHAPTERTIMESTART,   EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
466     { MATROSKA_ID_CHAPTERTIMEEND,     EBML_UINT, 0, offsetof(MatroskaChapter, end),   { .u = AV_NOPTS_VALUE } },
467     { MATROSKA_ID_CHAPTERUID,         EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
468     { MATROSKA_ID_CHAPTERDISPLAY,     EBML_NEST, 0,                        0,         { .n = matroska_chapter_display } },
469     { MATROSKA_ID_CHAPTERFLAGHIDDEN,  EBML_NONE },
470     { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
471     { MATROSKA_ID_CHAPTERPHYSEQUIV,   EBML_NONE },
472     { MATROSKA_ID_CHAPTERATOM,        EBML_NONE },
473 	{ 0 }
474 };
475 
476 static EbmlSyntax matroska_chapter[] = {
477 	{ MATROSKA_ID_CHAPTERATOM,        EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
478     { MATROSKA_ID_EDITIONUID,         EBML_NONE },
479     { MATROSKA_ID_EDITIONFLAGHIDDEN,  EBML_NONE },
480     { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
481     { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
482 	{ 0 }
483 };
484 
485 static EbmlSyntax matroska_chapters[] = {
486 	{ MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
487 	{ 0 }
488 };
489 
490 static EbmlSyntax matroska_index_pos[] = {
491     { MATROSKA_ID_CUETRACK,           EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
492     { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
493     { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
494     { MATROSKA_ID_CUEDURATION,        EBML_NONE },
495     { MATROSKA_ID_CUEBLOCKNUMBER,     EBML_NONE },
496     { 0 }
497 };
498 
499 static EbmlSyntax matroska_index_entry[] = {
500 	{ MATROSKA_ID_CUETIME,          EBML_UINT, 0,                        offsetof(MatroskaIndex, time) },
501     { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
502 	{ 0 }
503 };
504 
505 static EbmlSyntax matroska_index[] = {
506 	{ MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
507 	{ 0 }
508 };
509 
510 static EbmlSyntax matroska_simpletag[] = {
511 	{ MATROSKA_ID_TAGNAME,        EBML_UTF8, 0,                   offsetof(MatroskaTag, name) },
512     { MATROSKA_ID_TAGSTRING,      EBML_UTF8, 0,                   offsetof(MatroskaTag, string) },
513     { MATROSKA_ID_TAGLANG,        EBML_STR,  0,                   offsetof(MatroskaTag, lang), { .s = "und" } },
514     { MATROSKA_ID_TAGDEFAULT,     EBML_UINT, 0,                   offsetof(MatroskaTag, def) },
515     { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0,                   offsetof(MatroskaTag, def) },
516     { MATROSKA_ID_SIMPLETAG,      EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub),  { .n = matroska_simpletag } },
517 	{ 0 }
518 };
519 
520 static EbmlSyntax matroska_tagtargets[] = {
521 	{ MATROSKA_ID_TAGTARGETS_TYPE,       EBML_STR,  0, offsetof(MatroskaTagTarget, type) },
522     { MATROSKA_ID_TAGTARGETS_TYPEVALUE,  EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
523     { MATROSKA_ID_TAGTARGETS_TRACKUID,   EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
524     { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
525     { MATROSKA_ID_TAGTARGETS_ATTACHUID,  EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
526 	{ 0 }
527 };
528 
529 static EbmlSyntax matroska_tag[] = {
530 	{ MATROSKA_ID_SIMPLETAG,  EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag),    { .n = matroska_simpletag } },
531     { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0,                   offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
532 	{ 0 }
533 };
534 
535 static EbmlSyntax matroska_tags[] = {
536 	{ MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
537 	{ 0 }
538 };
539 
540 static EbmlSyntax matroska_seekhead_entry[] = {
541 	{ MATROSKA_ID_SEEKID,       EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
542     { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
543 	{ 0 }
544 };
545 
546 static EbmlSyntax matroska_seekhead[] = {
547 	{ MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
548 	{ 0 }
549 };
550 
551 static EbmlSyntax matroska_segment[] = {
552 	{ MATROSKA_ID_INFO,        EBML_NEST, 0, 0, { .n = matroska_info } },
553     { MATROSKA_ID_TRACKS,      EBML_NEST, 0, 0, { .n = matroska_tracks } },
554     { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, { .n = matroska_attachments } },
555     { MATROSKA_ID_CHAPTERS,    EBML_NEST, 0, 0, { .n = matroska_chapters } },
556     { MATROSKA_ID_CUES,        EBML_NEST, 0, 0, { .n = matroska_index } },
557     { MATROSKA_ID_TAGS,        EBML_NEST, 0, 0, { .n = matroska_tags } },
558     { MATROSKA_ID_SEEKHEAD,    EBML_NEST, 0, 0, { .n = matroska_seekhead } },
559     { MATROSKA_ID_CLUSTER,     EBML_STOP },
560 	{ 0 }
561 };
562 
563 static EbmlSyntax matroska_segments[] = {
564 	{ MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
565 	{ 0 }
566 };
567 
568 static EbmlSyntax matroska_blockmore[] = {
569     { MATROSKA_ID_BLOCKADDID,      EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
570     { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN,  0, offsetof(MatroskaBlock,additional) },
571     { 0 }
572 };
573 
574 static EbmlSyntax matroska_blockadditions[] = {
575 	{ MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
576 	{ 0 }
577 };
578 
579 static EbmlSyntax matroska_blockgroup[] = {
580 	{ MATROSKA_ID_BLOCK,          EBML_BIN,  0, offsetof(MatroskaBlock, bin) },
581     { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
582     { MATROSKA_ID_SIMPLEBLOCK,    EBML_BIN,  0, offsetof(MatroskaBlock, bin) },
583     { MATROSKA_ID_BLOCKDURATION,  EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
584     { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
585     { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference) },
586     { MATROSKA_ID_CODECSTATE,     EBML_NONE },
587     {                          1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
588 	{ 0 }
589 };
590 
591 static EbmlSyntax matroska_cluster[] = {
592 	{ MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0,                     offsetof(MatroskaCluster, timecode) },
593     { MATROSKA_ID_BLOCKGROUP,      EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
594     { MATROSKA_ID_SIMPLEBLOCK,     EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
595     { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
596     { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
597 	{ 0 }
598 };
599 
600 static EbmlSyntax matroska_clusters[] = {
601 	{ MATROSKA_ID_CLUSTER,  EBML_NEST, 0, 0, { .n = matroska_cluster } },
602     { MATROSKA_ID_INFO,     EBML_NONE },
603     { MATROSKA_ID_CUES,     EBML_NONE },
604     { MATROSKA_ID_TAGS,     EBML_NONE },
605     { MATROSKA_ID_SEEKHEAD, EBML_NONE },
606 	{ 0 }
607 };
608 
609 static EbmlSyntax matroska_cluster_incremental_parsing[] = {
610 	{ MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0,                     offsetof(MatroskaCluster, timecode) },
611     { MATROSKA_ID_BLOCKGROUP,      EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
612     { MATROSKA_ID_SIMPLEBLOCK,     EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
613     { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
614     { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
615     { MATROSKA_ID_INFO,            EBML_NONE },
616     { MATROSKA_ID_CUES,            EBML_NONE },
617     { MATROSKA_ID_TAGS,            EBML_NONE },
618     { MATROSKA_ID_SEEKHEAD,        EBML_NONE },
619     { MATROSKA_ID_CLUSTER,         EBML_STOP },
620 	{ 0 }
621 };
622 
623 static EbmlSyntax matroska_cluster_incremental[] = {
624     { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
625     { MATROSKA_ID_BLOCKGROUP,      EBML_STOP },
626     { MATROSKA_ID_SIMPLEBLOCK,     EBML_STOP },
627     { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
628     { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
629     { 0 }
630 };
631 
632 static EbmlSyntax matroska_clusters_incremental[] = {
633 	{ MATROSKA_ID_CLUSTER,  EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
634     { MATROSKA_ID_INFO,     EBML_NONE },
635     { MATROSKA_ID_CUES,     EBML_NONE },
636     { MATROSKA_ID_TAGS,     EBML_NONE },
637     { MATROSKA_ID_SEEKHEAD, EBML_NONE },
638 	{ 0 }
639 };
640 
641 static const char *const matroska_doctypes[] = { "matroska", "webm" };
642 
matroska_resync(MatroskaDemuxContext * matroska,int64_t last_pos)643 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
644 {
645     AVIOContext *pb = matroska->ctx->pb;
646     uint32_t id;
647     matroska->current_id = 0;
648     matroska->num_levels = 0;
649 
650     /* seek to next position to resync from */
651     if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
652         goto eof;
653 
654     id = avio_rb32(pb);
655 
656     // try to find a toplevel element
657     while (!avio_feof(pb)) {
658         if (id == MATROSKA_ID_INFO     || id == MATROSKA_ID_TRACKS      ||
659             id == MATROSKA_ID_CUES     || id == MATROSKA_ID_TAGS        ||
660             id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
661             id == MATROSKA_ID_CLUSTER  || id == MATROSKA_ID_CHAPTERS) {
662             matroska->current_id = id;
663             return 0;
664         }
665         id = (id << 8) | avio_r8(pb);
666     }
667 
668 eof:
669     matroska->done = 1;
670     return AVERROR_EOF;
671 }
672 
673 /*
674  * Return: Whether we reached the end of a level in the hierarchy or not.
675  */
ebml_level_end(MatroskaDemuxContext * matroska)676 static int ebml_level_end(MatroskaDemuxContext *matroska)
677 {
678     AVIOContext *pb = matroska->ctx->pb;
679     int64_t pos = avio_tell(pb);
680 
681     if (matroska->num_levels > 0) {
682         MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
683         if (pos - level->start >= level->length || matroska->current_id) {
684             matroska->num_levels--;
685             return 1;
686         }
687     }
688     return 0;
689 }
690 
691 /*
692  * Read: an "EBML number", which is defined as a variable-length
693  * array of bytes. The first byte indicates the length by giving a
694  * number of 0-bits followed by a one. The position of the first
695  * "one" bit inside the first byte indicates the length of this
696  * number.
697  * Returns: number of bytes read, < 0 on error
698  */
ebml_read_num(MatroskaDemuxContext * matroska,AVIOContext * pb,int max_size,uint64_t * number)699 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
700                          int max_size, uint64_t *number)
701 {
702     int read = 1, n = 1;
703     uint64_t total = 0;
704 
705     /* The first byte tells us the length in bytes - avio_r8() can normally
706      * return 0, but since that's not a valid first ebmlID byte, we can
707      * use it safely here to catch EOS. */
708     if (!(total = avio_r8(pb))) {
709         /* we might encounter EOS here */
710         if (!avio_feof(pb)) {
711             int64_t pos = avio_tell(pb);
712             av_log(matroska->ctx, AV_LOG_ERROR,
713                    "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
714                    pos, pos);
715             return pb->error ? pb->error : AVERROR(EIO);
716         }
717         return AVERROR_EOF;
718     }
719 
720     /* get the length of the EBML number */
721     read = 8 - ff_log2_tab[total];
722     if (read > max_size) {
723         int64_t pos = avio_tell(pb) - 1;
724         av_log(matroska->ctx, AV_LOG_ERROR,
725                "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
726                (uint8_t) total, pos, pos);
727         return AVERROR_INVALIDDATA;
728     }
729 
730     /* read out length */
731     total ^= 1 << ff_log2_tab[total];
732     while (n++ < read)
733         total = (total << 8) | avio_r8(pb);
734 
735     *number = total;
736 
737     return read;
738 }
739 
740 /**
741  * Read a EBML length value.
742  * This needs special handling for the "unknown length" case which has multiple
743  * encodings.
744  */
ebml_read_length(MatroskaDemuxContext * matroska,AVIOContext * pb,uint64_t * number)745 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
746                             uint64_t *number)
747 {
748     int res = ebml_read_num(matroska, pb, 8, number);
749     if (res > 0 && *number + 1 == ULLN(1) << (7 * res))
750         *number = ULLN(0xffffffffffffff);
751     return res;
752 }
753 
754 /*
755  * Read the next element as an unsigned int.
756  * 0 is success, < 0 is failure.
757  */
ebml_read_uint(AVIOContext * pb,int size,uint64_t * num)758 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
759 {
760     int n = 0;
761 
762     if (size > 8)
763         return AVERROR_INVALIDDATA;
764 
765     /* big-endian ordering; build up number */
766     *num = 0;
767     while (n++ < size)
768         *num = (*num << 8) | avio_r8(pb);
769 
770     return 0;
771 }
772 
773 /*
774  * Read the next element as a signed int.
775  * 0 is success, < 0 is failure.
776  */
ebml_read_sint(AVIOContext * pb,int size,int64_t * num)777 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
778 {
779     int n = 1;
780 
781     if (size > 8)
782         return AVERROR_INVALIDDATA;
783 
784     if (size == 0) {
785         *num = 0;
786     } else {
787         *num = sign_extend(avio_r8(pb), 8);
788 
789         /* big-endian ordering; build up number */
790         while (n++ < size)
791             *num = (*num << 8) | avio_r8(pb);
792     }
793 
794     return 0;
795 }
796 
797 /*
798  * Read the next element as a float.
799  * 0 is success, < 0 is failure.
800  */
ebml_read_float(AVIOContext * pb,int size,double * num)801 static int ebml_read_float(AVIOContext *pb, int size, double *num)
802 {
803     if (size == 0)
804         *num = 0;
805     else if (size == 4)
806         *num = av_int2float(avio_rb32(pb));
807     else if (size == 8)
808         *num = av_int2double(avio_rb64(pb));
809     else
810         return AVERROR_INVALIDDATA;
811 
812     return 0;
813 }
814 
815 /*
816  * Read the next element as an ASCII string.
817  * 0 is success, < 0 is failure.
818  */
ebml_read_ascii(AVIOContext * pb,int size,char ** str)819 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
820 {
821     char *res;
822 
823     /* EBML strings are usually not 0-terminated, so we allocate one
824      * byte more, read the string and NULL-terminate it ourselves. */
825     if (!(res = av_malloc(size + 1)))
826         return AVERROR(ENOMEM);
827     if (avio_read(pb, (uint8_t *) res, size) != size) {
828         av_free(res);
829         return AVERROR(EIO);
830     }
831     (res)[size] = '\0';
832     av_free(*str);
833     *str = res;
834 
835     return 0;
836 }
837 
838 /*
839  * Read the next element as binary data.
840  * 0 is success, < 0 is failure.
841  */
ebml_read_binary(AVIOContext * pb,int length,EbmlBin * bin)842 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
843 {
844     av_fast_padded_malloc(&bin->data, &bin->size, length);
845     if (!bin->data)
846         return AVERROR(ENOMEM);
847 
848     bin->size = length;
849     bin->pos  = avio_tell(pb);
850     if (avio_read(pb, bin->data, length) != length) {
851         av_freep(&bin->data);
852         bin->size = 0;
853         return AVERROR(EIO);
854     }
855 
856     return 0;
857 }
858 
859 /*
860  * Read the next element, but only the header. The contents
861  * are supposed to be sub-elements which can be read separately.
862  * 0 is success, < 0 is failure.
863  */
ebml_read_master(MatroskaDemuxContext * matroska,uint64_t length)864 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
865 {
866     AVIOContext *pb = matroska->ctx->pb;
867     MatroskaLevel *level;
868 
869     if (matroska->num_levels >= EBML_MAX_DEPTH) {
870         av_log(matroska->ctx, AV_LOG_ERROR,
871                "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
872         return AVERROR(ENOSYS);
873     }
874 
875     level         = &matroska->levels[matroska->num_levels++];
876     level->start  = avio_tell(pb);
877     level->length = length;
878 
879     return 0;
880 }
881 
882 /*
883  * Read signed/unsigned "EBML" numbers.
884  * Return: number of bytes processed, < 0 on error
885  */
matroska_ebmlnum_uint(MatroskaDemuxContext * matroska,uint8_t * data,uint32_t size,uint64_t * num)886 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
887                                  uint8_t *data, uint32_t size, uint64_t *num)
888 {
889     AVIOContext pb;
890     ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
891     return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
892 }
893 
894 /*
895  * Same as above, but signed.
896  */
matroska_ebmlnum_sint(MatroskaDemuxContext * matroska,uint8_t * data,uint32_t size,int64_t * num)897 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
898                                  uint8_t *data, uint32_t size, int64_t *num)
899 {
900     uint64_t unum;
901     int res;
902 
903     /* read as unsigned number first */
904     if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
905         return res;
906 
907     /* make signed (weird way) */
908     *num = unum - ((LLN(1) << (7 * res - 1)) - 1);
909 
910     return res;
911 }
912 
913 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
914                            EbmlSyntax *syntax, void *data);
915 
ebml_parse_id(MatroskaDemuxContext * matroska,EbmlSyntax * syntax,uint32_t id,void * data)916 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
917                          uint32_t id, void *data)
918 {
919     int i;
920     for (i = 0; syntax[i].id; i++)
921         if (id == syntax[i].id)
922             break;
923     if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
924         matroska->num_levels > 0                   &&
925         matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
926         return 0;  // we reached the end of an unknown size cluster
927     if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
928         av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%"PRIX32"\n", id);
929         if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
930             return AVERROR_INVALIDDATA;
931     }
932     return ebml_parse_elem(matroska, &syntax[i], data);
933 }
934 
ebml_parse(MatroskaDemuxContext * matroska,EbmlSyntax * syntax,void * data)935 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
936                       void *data)
937 {
938     if (!matroska->current_id) {
939         uint64_t id;
940         int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
941         if (res < 0)
942             return res;
943         matroska->current_id = id | 1 << 7 * res;
944     }
945     return ebml_parse_id(matroska, syntax, matroska->current_id, data);
946 }
947 
ebml_parse_nest(MatroskaDemuxContext * matroska,EbmlSyntax * syntax,void * data)948 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
949                            void *data)
950 {
951     int i, res = 0;
952 
953     for (i = 0; syntax[i].id; i++)
954         switch (syntax[i].type) {
955         case EBML_UINT:
956             *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
957             break;
958         case EBML_FLOAT:
959             *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
960             break;
961         case EBML_STR:
962         case EBML_UTF8:
963             // the default may be NULL
964             if (syntax[i].def.s) {
965                 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
966                 *dst = av_strdup(syntax[i].def.s);
967                 if (!*dst)
968                     return AVERROR(ENOMEM);
969             }
970             break;
971         }
972 
973     while (!res && !ebml_level_end(matroska))
974         res = ebml_parse(matroska, syntax, data);
975 
976     return res;
977 }
978 
ebml_parse_elem(MatroskaDemuxContext * matroska,EbmlSyntax * syntax,void * data)979 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
980                            EbmlSyntax *syntax, void *data)
981 {
982 	static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
983         [EBML_UINT]  = 8,
984         [EBML_FLOAT] = 8,
985         // max. 16 MB for strings
986         [EBML_STR]   = 0x1000000,
987         [EBML_UTF8]  = 0x1000000,
988         // max. 256 MB for binary data
989         [EBML_BIN]   = 0x10000000,
990         // no limits for anything else
991     };
992 	AVIOContext *pb = matroska->ctx->pb;
993     uint32_t id = syntax->id;
994     uint64_t length;
995     int res;
996     void *newelem;
997 
998     data = (char *) data + syntax->data_offset;
999     if (syntax->list_elem_size) {
1000         EbmlList *list = data;
1001         newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
1002         if (!newelem)
1003             return AVERROR(ENOMEM);
1004         list->elem = newelem;
1005         data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1006         memset(data, 0, syntax->list_elem_size);
1007         list->nb_elem++;
1008     }
1009 
1010     if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
1011         matroska->current_id = 0;
1012         if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1013             return res;
1014         if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1015             av_log(matroska->ctx, AV_LOG_ERROR,
1016                    "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1017                    length, max_lengths[syntax->type], syntax->type);
1018             return AVERROR_INVALIDDATA;
1019         }
1020     }
1021 
1022     switch (syntax->type) {
1023     case EBML_UINT:
1024         res = ebml_read_uint(pb, length, data);
1025         break;
1026     case EBML_SINT:
1027         res = ebml_read_sint(pb, length, data);
1028         break;
1029     case EBML_FLOAT:
1030         res = ebml_read_float(pb, length, data);
1031         break;
1032     case EBML_STR:
1033     case EBML_UTF8:
1034         res = ebml_read_ascii(pb, length, data);
1035         break;
1036     case EBML_BIN:
1037         res = ebml_read_binary(pb, length, data);
1038         break;
1039     case EBML_NEST:
1040         if ((res = ebml_read_master(matroska, length)) < 0)
1041             return res;
1042         if (id == MATROSKA_ID_SEGMENT)
1043             matroska->segment_start = avio_tell(matroska->ctx->pb);
1044         return ebml_parse_nest(matroska, syntax->def.n, data);
1045     case EBML_PASS:
1046         return ebml_parse_id(matroska, syntax->def.n, id, data);
1047     case EBML_STOP:
1048         return 1;
1049     default:
1050         if (ffio_limit(pb, length) != length)
1051             return AVERROR(EIO);
1052         return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
1053     }
1054     if (res == AVERROR_INVALIDDATA)
1055         av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1056     else if (res == AVERROR(EIO))
1057         av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1058     return res;
1059 }
1060 
ebml_free(EbmlSyntax * syntax,void * data)1061 static void ebml_free(EbmlSyntax *syntax, void *data)
1062 {
1063     int i, j;
1064     for (i = 0; syntax[i].id; i++) {
1065         void *data_off = (char *) data + syntax[i].data_offset;
1066         switch (syntax[i].type) {
1067         case EBML_STR:
1068         case EBML_UTF8:
1069             av_freep(data_off);
1070             break;
1071         case EBML_BIN:
1072             av_freep(&((EbmlBin *) data_off)->data);
1073             break;
1074         case EBML_NEST:
1075             if (syntax[i].list_elem_size) {
1076                 EbmlList *list = data_off;
1077                 char *ptr = list->elem;
1078                 for (j = 0; j < list->nb_elem;
1079                      j++, ptr += syntax[i].list_elem_size)
1080                     ebml_free(syntax[i].def.n, ptr);
1081                 av_free(list->elem);
1082             } else
1083                 ebml_free(syntax[i].def.n, data_off);
1084         default:
1085             break;
1086         }
1087     }
1088 }
1089 
1090 /*
1091  * Autodetecting...
1092  */
matroska_probe(AVProbeData * p)1093 static int matroska_probe(AVProbeData *p)
1094 {
1095     uint64_t total = 0;
1096     int len_mask = 0x80, size = 1, n = 1, i;
1097 
1098     /* EBML header? */
1099     if (AV_RB32(p->buf) != EBML_ID_HEADER)
1100         return 0;
1101 
1102     /* length of header */
1103     total = p->buf[4];
1104     while (size <= 8 && !(total & len_mask)) {
1105         size++;
1106         len_mask >>= 1;
1107     }
1108     if (size > 8)
1109         return 0;
1110     total &= (len_mask - 1);
1111     while (n < size)
1112         total = (total << 8) | p->buf[4 + n++];
1113 
1114     /* Does the probe data contain the whole header? */
1115     if (p->buf_size < 4 + size + total)
1116         return 0;
1117 
1118     /* The header should contain a known document type. For now,
1119      * we don't parse the whole header but simply check for the
1120      * availability of that array of characters inside the header.
1121      * Not fully fool-proof, but good enough. */
1122     for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1123         int probelen = strlen(matroska_doctypes[i]);
1124         if (total < probelen)
1125             continue;
1126         for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1127             if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1128                 return AVPROBE_SCORE_MAX;
1129     }
1130 
1131     // probably valid EBML header but no recognized doctype
1132     return AVPROBE_SCORE_EXTENSION;
1133 }
1134 
matroska_find_track_by_num(MatroskaDemuxContext * matroska,int num)1135 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1136                                                  int num)
1137 {
1138     MatroskaTrack *tracks = matroska->tracks.elem;
1139     int i;
1140 
1141     for (i = 0; i < matroska->tracks.nb_elem; i++)
1142         if (tracks[i].num == num)
1143             return &tracks[i];
1144 
1145     av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1146     return NULL;
1147 }
1148 
matroska_decode_buffer(uint8_t ** buf,int * buf_size,MatroskaTrack * track)1149 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1150                                   MatroskaTrack *track)
1151 {
1152     MatroskaTrackEncoding *encodings = track->encodings.elem;
1153     uint8_t *data = *buf;
1154     int isize = *buf_size;
1155     uint8_t *pkt_data = NULL;
1156     uint8_t av_unused *newpktdata;
1157     int pkt_size = isize;
1158     int result = 0;
1159     int olen;
1160 
1161     if (pkt_size >= 10000000U)
1162         return AVERROR_INVALIDDATA;
1163 
1164     switch (encodings[0].compression.algo) {
1165     case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1166     {
1167         int header_size = encodings[0].compression.settings.size;
1168         uint8_t *header = encodings[0].compression.settings.data;
1169 
1170         if (header_size && !header) {
1171             av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1172             return -1;
1173         }
1174 
1175         if (!header_size)
1176             return 0;
1177 
1178         pkt_size = isize + header_size;
1179         pkt_data = av_malloc(pkt_size);
1180         if (!pkt_data)
1181             return AVERROR(ENOMEM);
1182 
1183         memcpy(pkt_data, header, header_size);
1184         memcpy(pkt_data + header_size, data, isize);
1185         break;
1186     }
1187 #if CONFIG_LZO
1188     case MATROSKA_TRACK_ENCODING_COMP_LZO:
1189         do {
1190             olen       = pkt_size *= 3;
1191             newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1192             if (!newpktdata) {
1193                 result = AVERROR(ENOMEM);
1194                 goto failed;
1195             }
1196             pkt_data = newpktdata;
1197             result   = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1198         } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1199         if (result) {
1200             result = AVERROR_INVALIDDATA;
1201             goto failed;
1202         }
1203         pkt_size -= olen;
1204         break;
1205 #endif
1206 #if CONFIG_ZLIB
1207     case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1208     {
1209         z_stream zstream = { 0 };
1210         if (inflateInit(&zstream) != Z_OK)
1211             return -1;
1212         zstream.next_in  = data;
1213         zstream.avail_in = isize;
1214         do {
1215             pkt_size  *= 3;
1216             newpktdata = av_realloc(pkt_data, pkt_size);
1217             if (!newpktdata) {
1218                 inflateEnd(&zstream);
1219                 goto failed;
1220             }
1221             pkt_data          = newpktdata;
1222             zstream.avail_out = pkt_size - zstream.total_out;
1223             zstream.next_out  = pkt_data + zstream.total_out;
1224             if (pkt_data) {
1225                 result = inflate(&zstream, Z_NO_FLUSH);
1226             } else
1227                 result = Z_MEM_ERROR;
1228         } while (result == Z_OK && pkt_size < 10000000);
1229         pkt_size = zstream.total_out;
1230         inflateEnd(&zstream);
1231         if (result != Z_STREAM_END) {
1232             if (result == Z_MEM_ERROR)
1233                 result = AVERROR(ENOMEM);
1234             else
1235                 result = AVERROR_INVALIDDATA;
1236             goto failed;
1237         }
1238         break;
1239     }
1240 #endif
1241 #if CONFIG_BZLIB
1242     case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1243     {
1244         bz_stream bzstream = { 0 };
1245         if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1246             return -1;
1247         bzstream.next_in  = data;
1248         bzstream.avail_in = isize;
1249         do {
1250             pkt_size  *= 3;
1251             newpktdata = av_realloc(pkt_data, pkt_size);
1252             if (!newpktdata) {
1253                 BZ2_bzDecompressEnd(&bzstream);
1254                 goto failed;
1255             }
1256             pkt_data           = newpktdata;
1257             bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1258             bzstream.next_out  = pkt_data + bzstream.total_out_lo32;
1259             if (pkt_data) {
1260                 result = BZ2_bzDecompress(&bzstream);
1261             } else
1262                 result = BZ_MEM_ERROR;
1263         } while (result == BZ_OK && pkt_size < 10000000);
1264         pkt_size = bzstream.total_out_lo32;
1265         BZ2_bzDecompressEnd(&bzstream);
1266         if (result != BZ_STREAM_END) {
1267             if (result == BZ_MEM_ERROR)
1268                 result = AVERROR(ENOMEM);
1269             else
1270                 result = AVERROR_INVALIDDATA;
1271             goto failed;
1272         }
1273         break;
1274     }
1275 #endif
1276     default:
1277         return AVERROR_INVALIDDATA;
1278     }
1279 
1280     *buf      = pkt_data;
1281     *buf_size = pkt_size;
1282     return 0;
1283 
1284 failed:
1285     av_free(pkt_data);
1286     return result;
1287 }
1288 
matroska_convert_tag(AVFormatContext * s,EbmlList * list,AVDictionary ** metadata,char * prefix)1289 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1290                                  AVDictionary **metadata, char *prefix)
1291 {
1292     MatroskaTag *tags = list->elem;
1293     char key[1024];
1294     int i;
1295 
1296     for (i = 0; i < list->nb_elem; i++) {
1297         const char *lang = tags[i].lang &&
1298                            strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1299 
1300         if (!tags[i].name) {
1301             av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1302             continue;
1303         }
1304         if (prefix)
1305             snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1306         else
1307             av_strlcpy(key, tags[i].name, sizeof(key));
1308         if (tags[i].def || !lang) {
1309             av_dict_set(metadata, key, tags[i].string, 0);
1310             if (tags[i].sub.nb_elem)
1311                 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1312         }
1313         if (lang) {
1314             av_strlcat(key, "-", sizeof(key));
1315             av_strlcat(key, lang, sizeof(key));
1316             av_dict_set(metadata, key, tags[i].string, 0);
1317             if (tags[i].sub.nb_elem)
1318                 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1319         }
1320     }
1321     ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1322 }
1323 
matroska_convert_tags(AVFormatContext * s)1324 static void matroska_convert_tags(AVFormatContext *s)
1325 {
1326     MatroskaDemuxContext *matroska = s->priv_data;
1327     MatroskaTags *tags = matroska->tags.elem;
1328     int i, j;
1329 
1330     for (i = 0; i < matroska->tags.nb_elem; i++) {
1331         if (tags[i].target.attachuid) {
1332             MatroskaAttachment *attachment = matroska->attachments.elem;
1333             for (j = 0; j < matroska->attachments.nb_elem; j++)
1334                 if (attachment[j].uid == tags[i].target.attachuid &&
1335                     attachment[j].stream)
1336                     matroska_convert_tag(s, &tags[i].tag,
1337                                          &attachment[j].stream->metadata, NULL);
1338         } else if (tags[i].target.chapteruid) {
1339             MatroskaChapter *chapter = matroska->chapters.elem;
1340             for (j = 0; j < matroska->chapters.nb_elem; j++)
1341                 if (chapter[j].uid == tags[i].target.chapteruid &&
1342                     chapter[j].chapter)
1343                     matroska_convert_tag(s, &tags[i].tag,
1344                                          &chapter[j].chapter->metadata, NULL);
1345         } else if (tags[i].target.trackuid) {
1346             MatroskaTrack *track = matroska->tracks.elem;
1347             for (j = 0; j < matroska->tracks.nb_elem; j++)
1348                 if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1349                     matroska_convert_tag(s, &tags[i].tag,
1350                                          &track[j].stream->metadata, NULL);
1351         } else {
1352             matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1353                                  tags[i].target.type);
1354         }
1355     }
1356 }
1357 
matroska_parse_seekhead_entry(MatroskaDemuxContext * matroska,int idx)1358 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1359                                          int idx)
1360 {
1361     EbmlList *seekhead_list = &matroska->seekhead;
1362     uint32_t level_up       = matroska->level_up;
1363     uint32_t saved_id       = matroska->current_id;
1364     MatroskaSeekhead *seekhead = seekhead_list->elem;
1365     int64_t before_pos = avio_tell(matroska->ctx->pb);
1366     MatroskaLevel level;
1367     int64_t offset;
1368     int ret = 0;
1369 
1370     if (idx >= seekhead_list->nb_elem            ||
1371         seekhead[idx].id == MATROSKA_ID_SEEKHEAD ||
1372         seekhead[idx].id == MATROSKA_ID_CLUSTER)
1373         return 0;
1374 
1375     /* seek */
1376     offset = seekhead[idx].pos + matroska->segment_start;
1377     if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1378         /* We don't want to lose our seekhead level, so we add
1379          * a dummy. This is a crude hack. */
1380         if (matroska->num_levels == EBML_MAX_DEPTH) {
1381             av_log(matroska->ctx, AV_LOG_INFO,
1382                    "Max EBML element depth (%d) reached, "
1383                    "cannot parse further.\n", EBML_MAX_DEPTH);
1384             ret = AVERROR_INVALIDDATA;
1385         } else {
1386             level.start  = 0;
1387             level.length = (uint64_t) -1;
1388             matroska->levels[matroska->num_levels] = level;
1389             matroska->num_levels++;
1390             matroska->current_id                   = 0;
1391 
1392             ret = ebml_parse(matroska, matroska_segment, matroska);
1393 
1394             /* remove dummy level */
1395             while (matroska->num_levels) {
1396                 uint64_t length = matroska->levels[--matroska->num_levels].length;
1397                 if (length == (uint64_t) -1)
1398                     break;
1399             }
1400         }
1401     }
1402     /* seek back */
1403     avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1404     matroska->level_up   = level_up;
1405     matroska->current_id = saved_id;
1406 
1407     return ret;
1408 }
1409 
matroska_execute_seekhead(MatroskaDemuxContext * matroska)1410 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1411 {
1412     EbmlList *seekhead_list = &matroska->seekhead;
1413     int64_t before_pos = avio_tell(matroska->ctx->pb);
1414     int i;
1415 
1416     // we should not do any seeking in the streaming case
1417     if (!matroska->ctx->pb->seekable ||
1418         (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1419         return;
1420 
1421     for (i = 0; i < seekhead_list->nb_elem; i++) {
1422         MatroskaSeekhead *seekhead = seekhead_list->elem;
1423         if (seekhead[i].pos <= before_pos)
1424             continue;
1425 
1426         // defer cues parsing until we actually need cue data.
1427         if (seekhead[i].id == MATROSKA_ID_CUES) {
1428             matroska->cues_parsing_deferred = 1;
1429             continue;
1430         }
1431 
1432         if (matroska_parse_seekhead_entry(matroska, i) < 0) {
1433             // mark index as broken
1434             matroska->cues_parsing_deferred = -1;
1435             break;
1436         }
1437     }
1438 }
1439 
matroska_add_index_entries(MatroskaDemuxContext * matroska)1440 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1441 {
1442     EbmlList *index_list;
1443     MatroskaIndex *index;
1444     int index_scale = 1;
1445     int i, j;
1446 
1447     index_list = &matroska->index;
1448     index      = index_list->elem;
1449     if (index_list->nb_elem &&
1450         index[0].time > 1E14 / matroska->time_scale) {
1451         av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1452         index_scale = matroska->time_scale;
1453     }
1454     for (i = 0; i < index_list->nb_elem; i++) {
1455         EbmlList *pos_list    = &index[i].pos;
1456         MatroskaIndexPos *pos = pos_list->elem;
1457         for (j = 0; j < pos_list->nb_elem; j++) {
1458             MatroskaTrack *track = matroska_find_track_by_num(matroska,
1459                                                               pos[j].track);
1460             if (track && track->stream)
1461                 av_add_index_entry(track->stream,
1462                                    pos[j].pos + matroska->segment_start,
1463                                    index[i].time / index_scale, 0, 0,
1464                                    AVINDEX_KEYFRAME);
1465         }
1466     }
1467 }
1468 
matroska_parse_cues(MatroskaDemuxContext * matroska)1469 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1470     EbmlList *seekhead_list = &matroska->seekhead;
1471     MatroskaSeekhead *seekhead = seekhead_list->elem;
1472     int i;
1473 
1474     for (i = 0; i < seekhead_list->nb_elem; i++)
1475         if (seekhead[i].id == MATROSKA_ID_CUES)
1476             break;
1477     av_assert1(i <= seekhead_list->nb_elem);
1478 
1479     if (matroska_parse_seekhead_entry(matroska, i) < 0)
1480        matroska->cues_parsing_deferred = -1;
1481     matroska_add_index_entries(matroska);
1482 }
1483 
matroska_aac_profile(char * codec_id)1484 static int matroska_aac_profile(char *codec_id)
1485 {
1486     static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1487     int profile;
1488 
1489     for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1490         if (strstr(codec_id, aac_profiles[profile]))
1491             break;
1492     return profile + 1;
1493 }
1494 
matroska_aac_sri(int samplerate)1495 static int matroska_aac_sri(int samplerate)
1496 {
1497     int sri;
1498 
1499     for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1500         if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1501             break;
1502     return sri;
1503 }
1504 
matroska_metadata_creation_time(AVDictionary ** metadata,int64_t date_utc)1505 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1506 {
1507     char buffer[32];
1508     /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1509     time_t creation_time = date_utc / 1000000000 + 978307200;
1510     struct tm *ptm = gmtime(&creation_time);
1511     if (!ptm) return;
1512     strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm);
1513     av_dict_set(metadata, "creation_time", buffer, 0);
1514 }
1515 
matroska_parse_flac(AVFormatContext * s,MatroskaTrack * track,int * offset)1516 static int matroska_parse_flac(AVFormatContext *s,
1517                                MatroskaTrack *track,
1518                                int *offset)
1519 {
1520     AVStream *st = track->stream;
1521     uint8_t *p = track->codec_priv.data;
1522     int size   = track->codec_priv.size;
1523 
1524     if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1525         av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1526         track->codec_priv.size = 0;
1527         return 0;
1528     }
1529     *offset = 8;
1530     track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1531 
1532     p    += track->codec_priv.size;
1533     size -= track->codec_priv.size;
1534 
1535     /* parse the remaining metadata blocks if present */
1536     while (size >= 4) {
1537         int block_last, block_type, block_size;
1538 
1539         flac_parse_block_header(p, &block_last, &block_type, &block_size);
1540 
1541         p    += 4;
1542         size -= 4;
1543         if (block_size > size)
1544             return 0;
1545 
1546         /* check for the channel mask */
1547         if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1548             AVDictionary *dict = NULL;
1549             AVDictionaryEntry *chmask;
1550 
1551             ff_vorbis_comment(s, &dict, p, block_size, 0);
1552             chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1553             if (chmask) {
1554                 uint64_t mask = strtol(chmask->value, NULL, 0);
1555                 if (!mask || mask & ULLN(~0x3ffff)) {
1556                     av_log(s, AV_LOG_WARNING,
1557                            "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1558                 } else
1559                     st->codec->channel_layout = mask;
1560             }
1561             av_dict_free(&dict);
1562         }
1563 
1564         p    += block_size;
1565         size -= block_size;
1566     }
1567 
1568     return 0;
1569 }
1570 
matroska_parse_tracks(AVFormatContext * s)1571 static int matroska_parse_tracks(AVFormatContext *s)
1572 {
1573     MatroskaDemuxContext *matroska = s->priv_data;
1574     MatroskaTrack *tracks = matroska->tracks.elem;
1575     AVStream *st;
1576     int i, j, ret;
1577     int k;
1578 
1579     for (i = 0; i < matroska->tracks.nb_elem; i++) {
1580         MatroskaTrack *track = &tracks[i];
1581         enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1582         EbmlList *encodings_list = &track->encodings;
1583         MatroskaTrackEncoding *encodings = encodings_list->elem;
1584         uint8_t *extradata = NULL;
1585         int extradata_size = 0;
1586         int extradata_offset = 0;
1587         uint32_t fourcc = 0;
1588         AVIOContext b;
1589         char* key_id_base64 = NULL;
1590         int bit_depth = -1;
1591 
1592         /* Apply some sanity checks. */
1593         if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1594             track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1595             track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
1596             track->type != MATROSKA_TRACK_TYPE_METADATA) {
1597             av_log(matroska->ctx, AV_LOG_INFO,
1598                    "Unknown or unsupported track type %"PRIu64"\n",
1599                    track->type);
1600             continue;
1601         }
1602         if (!track->codec_id)
1603             continue;
1604 
1605         if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1606             if (!track->default_duration && track->video.frame_rate > 0)
1607                 track->default_duration = 1000000000 / track->video.frame_rate;
1608             if (track->video.display_width == -1)
1609                 track->video.display_width = track->video.pixel_width;
1610             if (track->video.display_height == -1)
1611                 track->video.display_height = track->video.pixel_height;
1612             if (track->video.color_space.size == 4)
1613                 fourcc = AV_RL32(track->video.color_space.data);
1614         } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1615             if (!track->audio.out_samplerate)
1616                 track->audio.out_samplerate = track->audio.samplerate;
1617         }
1618         if (encodings_list->nb_elem > 1) {
1619             av_log(matroska->ctx, AV_LOG_ERROR,
1620                    "Multiple combined encodings not supported");
1621         } else if (encodings_list->nb_elem == 1) {
1622             if (encodings[0].type) {
1623                 if (encodings[0].encryption.key_id.size > 0) {
1624                     /* Save the encryption key id to be stored later as a
1625                        metadata tag. */
1626                     const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
1627                     key_id_base64 = av_malloc(b64_size);
1628                     if (key_id_base64 == NULL)
1629                         return AVERROR(ENOMEM);
1630 
1631                     av_base64_encode(key_id_base64, b64_size,
1632                                      encodings[0].encryption.key_id.data,
1633                                      encodings[0].encryption.key_id.size);
1634                 } else {
1635                     encodings[0].scope = 0;
1636                     av_log(matroska->ctx, AV_LOG_ERROR,
1637                            "Unsupported encoding type");
1638                 }
1639             } else if (
1640 #if CONFIG_ZLIB
1641                  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB  &&
1642 #endif
1643 #if CONFIG_BZLIB
1644                  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1645 #endif
1646 #if CONFIG_LZO
1647                  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO   &&
1648 #endif
1649                  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
1650                 encodings[0].scope = 0;
1651                 av_log(matroska->ctx, AV_LOG_ERROR,
1652                        "Unsupported encoding type");
1653             } else if (track->codec_priv.size && encodings[0].scope & 2) {
1654                 uint8_t *codec_priv = track->codec_priv.data;
1655                 int ret = matroska_decode_buffer(&track->codec_priv.data,
1656                                                  &track->codec_priv.size,
1657                                                  track);
1658                 if (ret < 0) {
1659                     track->codec_priv.data = NULL;
1660                     track->codec_priv.size = 0;
1661                     av_log(matroska->ctx, AV_LOG_ERROR,
1662                            "Failed to decode codec private data\n");
1663                 }
1664 
1665                 if (codec_priv != track->codec_priv.data)
1666                     av_free(codec_priv);
1667             }
1668         }
1669 
1670         for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
1671             if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1672                          strlen(ff_mkv_codec_tags[j].str))) {
1673                 codec_id = ff_mkv_codec_tags[j].id;
1674                 break;
1675             }
1676         }
1677 
1678         st = track->stream = avformat_new_stream(s, NULL);
1679         if (!st) {
1680             av_free(key_id_base64);
1681             return AVERROR(ENOMEM);
1682         }
1683 
1684         if (key_id_base64) {
1685             /* export encryption key id as base64 metadata tag */
1686             av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
1687             av_freep(&key_id_base64);
1688         }
1689 
1690         if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
1691              track->codec_priv.size >= 40               &&
1692             track->codec_priv.data) {
1693             track->ms_compat    = 1;
1694             bit_depth           = AV_RL16(track->codec_priv.data + 14);
1695             fourcc              = AV_RL32(track->codec_priv.data + 16);
1696             codec_id            = ff_codec_get_id(ff_codec_bmp_tags,
1697                                                   fourcc);
1698             if (!codec_id)
1699                 codec_id        = ff_codec_get_id(ff_codec_movvideo_tags,
1700                                                   fourcc);
1701             extradata_offset    = 40;
1702         } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
1703                    track->codec_priv.size >= 14         &&
1704                    track->codec_priv.data) {
1705             int ret;
1706             ffio_init_context(&b, track->codec_priv.data,
1707                               track->codec_priv.size,
1708                               0, NULL, NULL, NULL, NULL);
1709             ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1710             if (ret < 0)
1711                 return ret;
1712             codec_id         = st->codec->codec_id;
1713             extradata_offset = FFMIN(track->codec_priv.size, 18);
1714         } else if (!strcmp(track->codec_id, "A_QUICKTIME")
1715                    && (track->codec_priv.size >= 86)
1716                    && (track->codec_priv.data)) {
1717             fourcc = AV_RL32(track->codec_priv.data + 4);
1718             codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1719             if (ff_codec_get_id(ff_codec_movaudio_tags, AV_RL32(track->codec_priv.data))) {
1720                 fourcc = AV_RL32(track->codec_priv.data);
1721                 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1722             }
1723         } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
1724                    (track->codec_priv.size >= 21)          &&
1725                    (track->codec_priv.data)) {
1726             fourcc   = AV_RL32(track->codec_priv.data + 4);
1727             codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1728             if (ff_codec_get_id(ff_codec_movvideo_tags, AV_RL32(track->codec_priv.data))) {
1729                 fourcc   = AV_RL32(track->codec_priv.data);
1730                 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1731             }
1732             if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI "))
1733                 codec_id = AV_CODEC_ID_SVQ3;
1734         } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1735             switch (track->audio.bitdepth) {
1736             case  8:
1737                 codec_id = AV_CODEC_ID_PCM_U8;
1738                 break;
1739             case 24:
1740                 codec_id = AV_CODEC_ID_PCM_S24BE;
1741                 break;
1742             case 32:
1743                 codec_id = AV_CODEC_ID_PCM_S32BE;
1744                 break;
1745             }
1746         } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1747             switch (track->audio.bitdepth) {
1748             case  8:
1749                 codec_id = AV_CODEC_ID_PCM_U8;
1750                 break;
1751             case 24:
1752                 codec_id = AV_CODEC_ID_PCM_S24LE;
1753                 break;
1754             case 32:
1755                 codec_id = AV_CODEC_ID_PCM_S32LE;
1756                 break;
1757             }
1758         } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
1759                    track->audio.bitdepth == 64) {
1760             codec_id = AV_CODEC_ID_PCM_F64LE;
1761         } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1762             int profile = matroska_aac_profile(track->codec_id);
1763             int sri     = matroska_aac_sri(track->audio.samplerate);
1764             extradata   = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1765             if (!extradata)
1766                 return AVERROR(ENOMEM);
1767             extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
1768             extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
1769             if (strstr(track->codec_id, "SBR")) {
1770                 sri            = matroska_aac_sri(track->audio.out_samplerate);
1771                 extradata[2]   = 0x56;
1772                 extradata[3]   = 0xE5;
1773                 extradata[4]   = 0x80 | (sri << 3);
1774                 extradata_size = 5;
1775             } else
1776                 extradata_size = 2;
1777         } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - FF_INPUT_BUFFER_PADDING_SIZE) {
1778             /* Only ALAC's magic cookie is stored in Matroska's track headers.
1779              * Create the "atom size", "tag", and "tag version" fields the
1780              * decoder expects manually. */
1781             extradata_size = 12 + track->codec_priv.size;
1782             extradata      = av_mallocz(extradata_size +
1783                                         FF_INPUT_BUFFER_PADDING_SIZE);
1784             if (!extradata)
1785                 return AVERROR(ENOMEM);
1786             AV_WB32(extradata, extradata_size);
1787             memcpy(&extradata[4], "alac", 4);
1788             AV_WB32(&extradata[8], 0);
1789             memcpy(&extradata[12], track->codec_priv.data,
1790                    track->codec_priv.size);
1791         } else if (codec_id == AV_CODEC_ID_TTA) {
1792             extradata_size = 30;
1793             extradata      = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1794             if (!extradata)
1795                 return AVERROR(ENOMEM);
1796             ffio_init_context(&b, extradata, extradata_size, 1,
1797                               NULL, NULL, NULL, NULL);
1798             avio_write(&b, "TTA1", 4);
1799             avio_wl16(&b, 1);
1800             avio_wl16(&b, track->audio.channels);
1801             avio_wl16(&b, track->audio.bitdepth);
1802             if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
1803                 return AVERROR_INVALIDDATA;
1804             avio_wl32(&b, track->audio.out_samplerate);
1805             avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
1806                                      track->audio.out_samplerate,
1807                                      AV_TIME_BASE * 1000));
1808         } else if (codec_id == AV_CODEC_ID_RV10 ||
1809                    codec_id == AV_CODEC_ID_RV20 ||
1810                    codec_id == AV_CODEC_ID_RV30 ||
1811                    codec_id == AV_CODEC_ID_RV40) {
1812             extradata_offset = 26;
1813         } else if (codec_id == AV_CODEC_ID_RA_144) {
1814             track->audio.out_samplerate = 8000;
1815             track->audio.channels       = 1;
1816         } else if ((codec_id == AV_CODEC_ID_RA_288 ||
1817                     codec_id == AV_CODEC_ID_COOK   ||
1818                     codec_id == AV_CODEC_ID_ATRAC3 ||
1819                     codec_id == AV_CODEC_ID_SIPR)
1820                       && track->codec_priv.data) {
1821             int flavor;
1822 
1823             ffio_init_context(&b, track->codec_priv.data,
1824                               track->codec_priv.size,
1825                               0, NULL, NULL, NULL, NULL);
1826             avio_skip(&b, 22);
1827             flavor                       = avio_rb16(&b);
1828             track->audio.coded_framesize = avio_rb32(&b);
1829             avio_skip(&b, 12);
1830             track->audio.sub_packet_h    = avio_rb16(&b);
1831             track->audio.frame_size      = avio_rb16(&b);
1832             track->audio.sub_packet_size = avio_rb16(&b);
1833             if (flavor                        < 0 ||
1834                 track->audio.coded_framesize <= 0 ||
1835                 track->audio.sub_packet_h    <= 0 ||
1836                 track->audio.frame_size      <= 0 ||
1837                 track->audio.sub_packet_size <= 0)
1838                 return AVERROR_INVALIDDATA;
1839             track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
1840                                                track->audio.frame_size);
1841             if (!track->audio.buf)
1842                 return AVERROR(ENOMEM);
1843             if (codec_id == AV_CODEC_ID_RA_288) {
1844                 st->codec->block_align = track->audio.coded_framesize;
1845                 track->codec_priv.size = 0;
1846             } else {
1847                 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1848                     static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1849                     track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1850                     st->codec->bit_rate          = sipr_bit_rate[flavor];
1851                 }
1852                 st->codec->block_align = track->audio.sub_packet_size;
1853                 extradata_offset       = 78;
1854             }
1855         } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
1856             ret = matroska_parse_flac(s, track, &extradata_offset);
1857             if (ret < 0)
1858                 return ret;
1859         } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
1860             fourcc = AV_RL32(track->codec_priv.data);
1861         }
1862         track->codec_priv.size -= extradata_offset;
1863 
1864         if (codec_id == AV_CODEC_ID_NONE)
1865             av_log(matroska->ctx, AV_LOG_INFO,
1866                    "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1867 
1868         if (track->time_scale < 0.01)
1869             track->time_scale = 1.0;
1870         avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
1871                             1000 * 1000 * 1000);    /* 64 bit pts in ns */
1872 
1873         /* convert the delay from ns to the track timebase */
1874 		track->codec_delay = av_rescale_q(track->codec_delay,
1875                                           (AVRational){ 1, 1000000000 },
1876                                           st->time_base);
1877 
1878         st->codec->codec_id = codec_id;
1879 
1880         if (strcmp(track->language, "und"))
1881             av_dict_set(&st->metadata, "language", track->language, 0);
1882         av_dict_set(&st->metadata, "title", track->name, 0);
1883 
1884         if (track->flag_default)
1885             st->disposition |= AV_DISPOSITION_DEFAULT;
1886         if (track->flag_forced)
1887             st->disposition |= AV_DISPOSITION_FORCED;
1888 
1889         if (!st->codec->extradata) {
1890             if (extradata) {
1891                 st->codec->extradata      = extradata;
1892                 st->codec->extradata_size = extradata_size;
1893             } else if (track->codec_priv.data && track->codec_priv.size > 0) {
1894                 if (ff_alloc_extradata(st->codec, track->codec_priv.size))
1895                     return AVERROR(ENOMEM);
1896                 memcpy(st->codec->extradata,
1897                        track->codec_priv.data + extradata_offset,
1898                        track->codec_priv.size);
1899             }
1900         }
1901 
1902         if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1903             MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
1904 
1905             st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1906             st->codec->codec_tag  = fourcc;
1907             if (bit_depth >= 0)
1908                 st->codec->bits_per_coded_sample = bit_depth;
1909             st->codec->width      = track->video.pixel_width;
1910             st->codec->height     = track->video.pixel_height;
1911             av_reduce(&st->sample_aspect_ratio.num,
1912                       &st->sample_aspect_ratio.den,
1913                       st->codec->height * track->video.display_width,
1914                       st->codec->width  * track->video.display_height,
1915                       255);
1916             if (st->codec->codec_id != AV_CODEC_ID_HEVC)
1917                 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1918 
1919             if (track->default_duration) {
1920                 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1921                           1000000000, track->default_duration, 30000);
1922 #if FF_API_R_FRAME_RATE
1923                 if (st->avg_frame_rate.num < st->avg_frame_rate.den * 1000L)
1924                     st->r_frame_rate = st->avg_frame_rate;
1925 #endif
1926             }
1927 
1928             /* export stereo mode flag as metadata tag */
1929             if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
1930                 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
1931 
1932             /* export alpha mode flag as metadata tag  */
1933             if (track->video.alpha_mode)
1934                 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
1935 
1936             /* if we have virtual track, mark the real tracks */
1937             for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
1938                 char buf[32];
1939                 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
1940                     continue;
1941                 snprintf(buf, sizeof(buf), "%s_%d",
1942                          ff_matroska_video_stereo_plane[planes[j].type], i);
1943                 for (k=0; k < matroska->tracks.nb_elem; k++)
1944                     if (planes[j].uid == tracks[k].uid) {
1945                         av_dict_set(&s->streams[k]->metadata,
1946                                     "stereo_mode", buf, 0);
1947                         break;
1948                     }
1949             }
1950             // add stream level stereo3d side data if it is a supported format
1951             if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
1952                 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
1953                 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
1954                 if (ret < 0)
1955                     return ret;
1956             }
1957         } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1958             st->codec->codec_type  = AVMEDIA_TYPE_AUDIO;
1959             st->codec->sample_rate = track->audio.out_samplerate;
1960             st->codec->channels    = track->audio.channels;
1961             if (!st->codec->bits_per_coded_sample)
1962                 st->codec->bits_per_coded_sample = track->audio.bitdepth;
1963             if (st->codec->codec_id != AV_CODEC_ID_AAC)
1964                 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1965             if (track->codec_delay > 0) {
1966 				st->codec->delay = av_rescale_q(track->codec_delay,
1967                                                 st->time_base,
1968                                                 (AVRational){1, st->codec->sample_rate});
1969 			}
1970             if (track->seek_preroll > 0) {
1971 				av_codec_set_seek_preroll(st->codec,
1972                                           av_rescale_q(track->seek_preroll,
1973                                                        (AVRational){1, 1000000000},
1974                                                        (AVRational){1, st->codec->sample_rate}));
1975 			}
1976         } else if (codec_id == AV_CODEC_ID_WEBVTT) {
1977             st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1978 
1979             if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
1980                 st->disposition |= AV_DISPOSITION_CAPTIONS;
1981             } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
1982                 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
1983             } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
1984                 st->disposition |= AV_DISPOSITION_METADATA;
1985             }
1986         } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1987             st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1988             if (st->codec->codec_id == AV_CODEC_ID_ASS)
1989                 matroska->contains_ssa = 1;
1990         }
1991     }
1992 
1993     return 0;
1994 }
1995 
matroska_read_header(AVFormatContext * s)1996 static int matroska_read_header(AVFormatContext *s)
1997 {
1998     MatroskaDemuxContext *matroska = s->priv_data;
1999     EbmlList *attachments_list = &matroska->attachments;
2000     EbmlList *chapters_list    = &matroska->chapters;
2001     MatroskaAttachment *attachments;
2002     MatroskaChapter *chapters;
2003     uint64_t max_start = 0;
2004     int64_t pos;
2005     Ebml ebml = { 0 };
2006     int i, j, res;
2007 
2008     matroska->ctx = s;
2009 
2010     /* First read the EBML header. */
2011     if (ebml_parse(matroska, ebml_syntax, &ebml) ||
2012         ebml.version         > EBML_VERSION      ||
2013         ebml.max_size        > sizeof(uint64_t)  ||
2014         ebml.id_length       > sizeof(uint32_t)  ||
2015         ebml.doctype_version > 3                 ||
2016         !ebml.doctype) {
2017         av_log(matroska->ctx, AV_LOG_ERROR,
2018                "EBML header using unsupported features\n"
2019                "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2020                ebml.version, ebml.doctype, ebml.doctype_version);
2021         ebml_free(ebml_syntax, &ebml);
2022         return AVERROR_PATCHWELCOME;
2023     } else if (ebml.doctype_version == 3) {
2024         av_log(matroska->ctx, AV_LOG_WARNING,
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     }
2029     for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2030         if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2031             break;
2032     if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2033         av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2034         if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2035             ebml_free(ebml_syntax, &ebml);
2036             return AVERROR_INVALIDDATA;
2037         }
2038     }
2039     ebml_free(ebml_syntax, &ebml);
2040 
2041     /* The next thing is a segment. */
2042     pos = avio_tell(matroska->ctx->pb);
2043     res = ebml_parse(matroska, matroska_segments, matroska);
2044     // try resyncing until we find a EBML_STOP type element.
2045     while (res != 1) {
2046         res = matroska_resync(matroska, pos);
2047         if (res < 0)
2048             return res;
2049         pos = avio_tell(matroska->ctx->pb);
2050         res = ebml_parse(matroska, matroska_segment, matroska);
2051     }
2052     matroska_execute_seekhead(matroska);
2053 
2054     if (!matroska->time_scale)
2055         matroska->time_scale = 1000000;
2056     if (matroska->duration)
2057         matroska->ctx->duration = matroska->duration * matroska->time_scale *
2058                                   1000 / AV_TIME_BASE;
2059     av_dict_set(&s->metadata, "title", matroska->title, 0);
2060     av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2061 
2062     if (matroska->date_utc.size == 8)
2063         matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2064 
2065     res = matroska_parse_tracks(s);
2066     if (res < 0)
2067         return res;
2068 
2069     attachments = attachments_list->elem;
2070     for (j = 0; j < attachments_list->nb_elem; j++) {
2071         if (!(attachments[j].filename && attachments[j].mime &&
2072               attachments[j].bin.data && attachments[j].bin.size > 0)) {
2073             av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2074         } else {
2075             AVStream *st = avformat_new_stream(s, NULL);
2076             if (!st)
2077                 break;
2078             av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2079             av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2080             st->codec->codec_id   = AV_CODEC_ID_NONE;
2081             st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2082             if (ff_alloc_extradata(st->codec, attachments[j].bin.size))
2083                 break;
2084             memcpy(st->codec->extradata, attachments[j].bin.data,
2085                    attachments[j].bin.size);
2086 
2087             for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2088                 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2089                              strlen(ff_mkv_mime_tags[i].str))) {
2090                     st->codec->codec_id = ff_mkv_mime_tags[i].id;
2091                     break;
2092                 }
2093             }
2094             attachments[j].stream = st;
2095         }
2096     }
2097 
2098     chapters = chapters_list->elem;
2099     for (i = 0; i < chapters_list->nb_elem; i++)
2100         if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2101             (max_start == 0 || chapters[i].start > max_start)) {
2102 			chapters[i].chapter =
2103                 avpriv_new_chapter(s, chapters[i].uid,
2104                                    (AVRational) { 1, 1000000000 },
2105                                    chapters[i].start, chapters[i].end,
2106                                    chapters[i].title);
2107 			if (chapters[i].chapter) {
2108                 av_dict_set(&chapters[i].chapter->metadata,
2109                             "title", chapters[i].title, 0);
2110             }
2111             max_start = chapters[i].start;
2112         }
2113 
2114     matroska_add_index_entries(matroska);
2115 
2116     matroska_convert_tags(s);
2117 
2118     return 0;
2119 }
2120 
2121 /*
2122  * Put one packet in an application-supplied AVPacket struct.
2123  * Returns 0 on success or -1 on failure.
2124  */
matroska_deliver_packet(MatroskaDemuxContext * matroska,AVPacket * pkt)2125 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2126                                    AVPacket *pkt)
2127 {
2128     if (matroska->num_packets > 0) {
2129         memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2130         av_free(matroska->packets[0]);
2131         if (matroska->num_packets > 1) {
2132             void *newpackets;
2133             memmove(&matroska->packets[0], &matroska->packets[1],
2134                     (matroska->num_packets - 1) * sizeof(AVPacket *));
2135             newpackets = av_realloc(matroska->packets,
2136                                     (matroska->num_packets - 1) *
2137                                     sizeof(AVPacket *));
2138             if (newpackets)
2139                 matroska->packets = newpackets;
2140         } else {
2141             av_freep(&matroska->packets);
2142             matroska->prev_pkt = NULL;
2143         }
2144         matroska->num_packets--;
2145         return 0;
2146     }
2147 
2148     return -1;
2149 }
2150 
2151 /*
2152  * Free all packets in our internal queue.
2153  */
matroska_clear_queue(MatroskaDemuxContext * matroska)2154 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2155 {
2156     matroska->prev_pkt = NULL;
2157     if (matroska->packets) {
2158         int n;
2159         for (n = 0; n < matroska->num_packets; n++) {
2160             av_free_packet(matroska->packets[n]);
2161             av_free(matroska->packets[n]);
2162         }
2163         av_freep(&matroska->packets);
2164         matroska->num_packets = 0;
2165     }
2166 }
2167 
matroska_parse_laces(MatroskaDemuxContext * matroska,uint8_t ** buf,int * buf_size,int type,uint32_t ** lace_buf,int * laces)2168 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2169                                 int *buf_size, int type,
2170                                 uint32_t **lace_buf, int *laces)
2171 {
2172     int res = 0, n, size = *buf_size;
2173     uint8_t *data = *buf;
2174     uint32_t *lace_size;
2175 
2176     if (!type) {
2177         *laces    = 1;
2178         *lace_buf = av_mallocz(sizeof(int));
2179         if (!*lace_buf)
2180             return AVERROR(ENOMEM);
2181 
2182         *lace_buf[0] = size;
2183         return 0;
2184     }
2185 
2186     av_assert0(size > 0);
2187     *laces    = *data + 1;
2188     data     += 1;
2189     size     -= 1;
2190     lace_size = av_mallocz(*laces * sizeof(int));
2191     if (!lace_size)
2192         return AVERROR(ENOMEM);
2193 
2194     switch (type) {
2195     case 0x1: /* Xiph lacing */
2196     {
2197         uint8_t temp;
2198         uint32_t total = 0;
2199         for (n = 0; res == 0 && n < *laces - 1; n++) {
2200             while (1) {
2201                 if (size <= total) {
2202                     res = AVERROR_INVALIDDATA;
2203                     break;
2204                 }
2205                 temp          = *data;
2206                 total        += temp;
2207                 lace_size[n] += temp;
2208                 data         += 1;
2209                 size         -= 1;
2210                 if (temp != 0xff)
2211                     break;
2212             }
2213         }
2214         if (size <= total) {
2215             res = AVERROR_INVALIDDATA;
2216             break;
2217         }
2218 
2219         lace_size[n] = size - total;
2220         break;
2221     }
2222 
2223     case 0x2: /* fixed-size lacing */
2224         if (size % (*laces)) {
2225             res = AVERROR_INVALIDDATA;
2226             break;
2227         }
2228         for (n = 0; n < *laces; n++)
2229             lace_size[n] = size / *laces;
2230         break;
2231 
2232     case 0x3: /* EBML lacing */
2233     {
2234         uint64_t num;
2235         uint64_t total;
2236         n = matroska_ebmlnum_uint(matroska, data, size, &num);
2237         if (n < 0 || num > INT_MAX) {
2238             av_log(matroska->ctx, AV_LOG_INFO,
2239                    "EBML block data error\n");
2240             res = n<0 ? n : AVERROR_INVALIDDATA;
2241             break;
2242         }
2243         data += n;
2244         size -= n;
2245         total = lace_size[0] = num;
2246         for (n = 1; res == 0 && n < *laces - 1; n++) {
2247             int64_t snum;
2248             int r;
2249             r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2250             if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
2251                 av_log(matroska->ctx, AV_LOG_INFO,
2252                        "EBML block data error\n");
2253                 res = r<0 ? r : AVERROR_INVALIDDATA;
2254                 break;
2255             }
2256             data        += r;
2257             size        -= r;
2258             lace_size[n] = lace_size[n - 1] + snum;
2259             total       += lace_size[n];
2260         }
2261         if (size <= total) {
2262             res = AVERROR_INVALIDDATA;
2263             break;
2264         }
2265         lace_size[*laces - 1] = size - total;
2266         break;
2267     }
2268     }
2269 
2270     *buf      = data;
2271     *lace_buf = lace_size;
2272     *buf_size = size;
2273 
2274     return res;
2275 }
2276 
matroska_parse_rm_audio(MatroskaDemuxContext * matroska,MatroskaTrack * track,AVStream * st,uint8_t * data,int size,uint64_t timecode,int64_t pos)2277 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2278                                    MatroskaTrack *track, AVStream *st,
2279                                    uint8_t *data, int size, uint64_t timecode,
2280                                    int64_t pos)
2281 {
2282     int a = st->codec->block_align;
2283     int sps = track->audio.sub_packet_size;
2284     int cfs = track->audio.coded_framesize;
2285     int h   = track->audio.sub_packet_h;
2286     int y   = track->audio.sub_packet_cnt;
2287     int w   = track->audio.frame_size;
2288     int x;
2289 
2290     if (!track->audio.pkt_cnt) {
2291         if (track->audio.sub_packet_cnt == 0)
2292             track->audio.buf_timecode = timecode;
2293         if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2294             if (size < cfs * h / 2) {
2295                 av_log(matroska->ctx, AV_LOG_ERROR,
2296                        "Corrupt int4 RM-style audio packet size\n");
2297                 return AVERROR_INVALIDDATA;
2298             }
2299             for (x = 0; x < h / 2; x++)
2300                 memcpy(track->audio.buf + x * 2 * w + y * cfs,
2301                        data + x * cfs, cfs);
2302         } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2303             if (size < w) {
2304                 av_log(matroska->ctx, AV_LOG_ERROR,
2305                        "Corrupt sipr RM-style audio packet size\n");
2306                 return AVERROR_INVALIDDATA;
2307             }
2308             memcpy(track->audio.buf + y * w, data, w);
2309         } else {
2310             if (size < sps * w / sps || h<=0 || w%sps) {
2311                 av_log(matroska->ctx, AV_LOG_ERROR,
2312                        "Corrupt generic RM-style audio packet size\n");
2313                 return AVERROR_INVALIDDATA;
2314             }
2315             for (x = 0; x < w / sps; x++)
2316                 memcpy(track->audio.buf +
2317                        sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
2318                        data + x * sps, sps);
2319         }
2320 
2321         if (++track->audio.sub_packet_cnt >= h) {
2322             if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2323                 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2324             track->audio.sub_packet_cnt = 0;
2325             track->audio.pkt_cnt        = h * w / a;
2326         }
2327     }
2328 
2329     while (track->audio.pkt_cnt) {
2330         AVPacket *pkt = NULL;
2331         if (!(pkt = av_mallocz(sizeof(AVPacket))) || av_new_packet(pkt, a) < 0) {
2332             av_free(pkt);
2333             return AVERROR(ENOMEM);
2334         }
2335         memcpy(pkt->data,
2336                track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
2337                a);
2338         pkt->pts                  = track->audio.buf_timecode;
2339         track->audio.buf_timecode = AV_NOPTS_VALUE;
2340         pkt->pos                  = pos;
2341         pkt->stream_index         = st->index;
2342         dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2343     }
2344 
2345     return 0;
2346 }
2347 
2348 /* reconstruct full wavpack blocks from mangled matroska ones */
matroska_parse_wavpack(MatroskaTrack * track,uint8_t * src,uint8_t ** pdst,int * size)2349 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
2350                                   uint8_t **pdst, int *size)
2351 {
2352     uint8_t *dst = NULL;
2353     int dstlen   = 0;
2354     int srclen   = *size;
2355     uint32_t samples;
2356     uint16_t ver;
2357     int ret, offset = 0;
2358 
2359     if (srclen < 12 || track->stream->codec->extradata_size < 2)
2360         return AVERROR_INVALIDDATA;
2361 
2362     ver = AV_RL16(track->stream->codec->extradata);
2363 
2364     samples = AV_RL32(src);
2365     src    += 4;
2366     srclen -= 4;
2367 
2368     while (srclen >= 8) {
2369         int multiblock;
2370         uint32_t blocksize;
2371         uint8_t *tmp;
2372 
2373         uint32_t flags = AV_RL32(src);
2374         uint32_t crc   = AV_RL32(src + 4);
2375         src    += 8;
2376         srclen -= 8;
2377 
2378         multiblock = (flags & 0x1800) != 0x1800;
2379         if (multiblock) {
2380             if (srclen < 4) {
2381                 ret = AVERROR_INVALIDDATA;
2382                 goto fail;
2383             }
2384             blocksize = AV_RL32(src);
2385             src      += 4;
2386             srclen   -= 4;
2387         } else
2388             blocksize = srclen;
2389 
2390         if (blocksize > srclen) {
2391             ret = AVERROR_INVALIDDATA;
2392             goto fail;
2393         }
2394 
2395         tmp = av_realloc(dst, dstlen + blocksize + 32);
2396         if (!tmp) {
2397             ret = AVERROR(ENOMEM);
2398             goto fail;
2399         }
2400         dst     = tmp;
2401         dstlen += blocksize + 32;
2402 
2403         AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k'));   // tag
2404         AV_WL32(dst + offset +  4, blocksize + 24);         // blocksize - 8
2405         AV_WL16(dst + offset +  8, ver);                    // version
2406         AV_WL16(dst + offset + 10, 0);                      // track/index_no
2407         AV_WL32(dst + offset + 12, 0);                      // total samples
2408         AV_WL32(dst + offset + 16, 0);                      // block index
2409         AV_WL32(dst + offset + 20, samples);                // number of samples
2410         AV_WL32(dst + offset + 24, flags);                  // flags
2411         AV_WL32(dst + offset + 28, crc);                    // crc
2412         memcpy(dst + offset + 32, src, blocksize);          // block data
2413 
2414         src    += blocksize;
2415         srclen -= blocksize;
2416         offset += blocksize + 32;
2417     }
2418 
2419     *pdst = dst;
2420     *size = dstlen;
2421 
2422     return 0;
2423 
2424 fail:
2425     av_freep(&dst);
2426     return ret;
2427 }
2428 
matroska_parse_webvtt(MatroskaDemuxContext * matroska,MatroskaTrack * track,AVStream * st,uint8_t * data,int data_len,uint64_t timecode,uint64_t duration,int64_t pos)2429 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
2430                                  MatroskaTrack *track,
2431                                  AVStream *st,
2432                                  uint8_t *data, int data_len,
2433                                  uint64_t timecode,
2434                                  uint64_t duration,
2435                                  int64_t pos)
2436 {
2437     AVPacket *pkt;
2438     uint8_t *id, *settings, *text, *buf;
2439     int id_len, settings_len, text_len;
2440     uint8_t *p, *q;
2441     int err;
2442 
2443     if (data_len <= 0)
2444         return AVERROR_INVALIDDATA;
2445 
2446     p = data;
2447     q = data + data_len;
2448 
2449     id = p;
2450     id_len = -1;
2451     while (p < q) {
2452         if (*p == '\r' || *p == '\n') {
2453             id_len = p - id;
2454             if (*p == '\r')
2455                 p++;
2456             break;
2457         }
2458         p++;
2459     }
2460 
2461     if (p >= q || *p != '\n')
2462         return AVERROR_INVALIDDATA;
2463     p++;
2464 
2465     settings = p;
2466     settings_len = -1;
2467     while (p < q) {
2468         if (*p == '\r' || *p == '\n') {
2469             settings_len = p - settings;
2470             if (*p == '\r')
2471                 p++;
2472             break;
2473         }
2474         p++;
2475     }
2476 
2477     if (p >= q || *p != '\n')
2478         return AVERROR_INVALIDDATA;
2479     p++;
2480 
2481     text = p;
2482     text_len = q - p;
2483     while (text_len > 0) {
2484         const int len = text_len - 1;
2485         const uint8_t c = p[len];
2486         if (c != '\r' && c != '\n')
2487             break;
2488         text_len = len;
2489     }
2490 
2491     if (text_len <= 0)
2492         return AVERROR_INVALIDDATA;
2493 
2494     pkt = av_mallocz(sizeof(*pkt));
2495     err = av_new_packet(pkt, text_len);
2496     if (err < 0) {
2497         av_free(pkt);
2498         return AVERROR(err);
2499     }
2500 
2501     memcpy(pkt->data, text, text_len);
2502 
2503     if (id_len > 0) {
2504         buf = av_packet_new_side_data(pkt,
2505                                       AV_PKT_DATA_WEBVTT_IDENTIFIER,
2506                                       id_len);
2507         if (!buf) {
2508             av_free(pkt);
2509             return AVERROR(ENOMEM);
2510         }
2511         memcpy(buf, id, id_len);
2512     }
2513 
2514     if (settings_len > 0) {
2515         buf = av_packet_new_side_data(pkt,
2516                                       AV_PKT_DATA_WEBVTT_SETTINGS,
2517                                       settings_len);
2518         if (!buf) {
2519             av_free(pkt);
2520             return AVERROR(ENOMEM);
2521         }
2522         memcpy(buf, settings, settings_len);
2523     }
2524 
2525     // Do we need this for subtitles?
2526     // pkt->flags = AV_PKT_FLAG_KEY;
2527 
2528     pkt->stream_index = st->index;
2529     pkt->pts = timecode;
2530 
2531     // Do we need this for subtitles?
2532     // pkt->dts = timecode;
2533 
2534     pkt->duration = duration;
2535     pkt->pos = pos;
2536 
2537     dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2538     matroska->prev_pkt = pkt;
2539 
2540     return 0;
2541 }
2542 
matroska_parse_frame(MatroskaDemuxContext * matroska,MatroskaTrack * track,AVStream * st,uint8_t * data,int pkt_size,uint64_t timecode,uint64_t lace_duration,int64_t pos,int is_keyframe,uint8_t * additional,uint64_t additional_id,int additional_size,int64_t discard_padding)2543 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2544                                 MatroskaTrack *track, AVStream *st,
2545                                 uint8_t *data, int pkt_size,
2546                                 uint64_t timecode, uint64_t lace_duration,
2547                                 int64_t pos, int is_keyframe,
2548                                 uint8_t *additional, uint64_t additional_id, int additional_size,
2549                                 int64_t discard_padding)
2550 {
2551     MatroskaTrackEncoding *encodings = track->encodings.elem;
2552     uint8_t *pkt_data = data;
2553     int offset = 0, res;
2554     AVPacket *pkt;
2555 
2556     if (encodings && !encodings->type && encodings->scope & 1) {
2557         res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2558         if (res < 0)
2559             return res;
2560     }
2561 
2562     if (st->codec->codec_id == AV_CODEC_ID_WAVPACK) {
2563         uint8_t *wv_data;
2564         res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
2565         if (res < 0) {
2566             av_log(matroska->ctx, AV_LOG_ERROR,
2567                    "Error parsing a wavpack block.\n");
2568             goto fail;
2569         }
2570         if (pkt_data != data)
2571             av_freep(&pkt_data);
2572         pkt_data = wv_data;
2573     }
2574 
2575     if (st->codec->codec_id == AV_CODEC_ID_PRORES &&
2576         AV_RB32(&data[4]) != MKBETAG('i', 'c', 'p', 'f'))
2577         offset = 8;
2578 
2579     pkt = av_mallocz(sizeof(AVPacket));
2580     /* XXX: prevent data copy... */
2581     if (av_new_packet(pkt, pkt_size + offset) < 0) {
2582         av_free(pkt);
2583         res = AVERROR(ENOMEM);
2584         goto fail;
2585     }
2586 
2587     if (st->codec->codec_id == AV_CODEC_ID_PRORES && offset == 8) {
2588         uint8_t *buf = pkt->data;
2589         bytestream_put_be32(&buf, pkt_size);
2590         bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2591     }
2592 
2593     memcpy(pkt->data + offset, pkt_data, pkt_size);
2594 
2595     if (pkt_data != data)
2596         av_freep(&pkt_data);
2597 
2598     pkt->flags        = is_keyframe;
2599     pkt->stream_index = st->index;
2600 
2601     if (additional_size > 0) {
2602         uint8_t *side_data = av_packet_new_side_data(pkt,
2603                                                      AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
2604                                                      additional_size + 8);
2605         if (!side_data) {
2606             av_free_packet(pkt);
2607             av_free(pkt);
2608             return AVERROR(ENOMEM);
2609         }
2610         AV_WB64(side_data, additional_id);
2611         memcpy(side_data + 8, additional, additional_size);
2612     }
2613 
2614     if (discard_padding) {
2615         uint8_t *side_data = av_packet_new_side_data(pkt,
2616                                                      AV_PKT_DATA_SKIP_SAMPLES,
2617                                                      10);
2618 
2619         if (!side_data) {
2620             av_free_packet(pkt);
2621             av_free(pkt);
2622             return AVERROR(ENOMEM);
2623         }
2624         AV_WL32(side_data, 0);
2625 		AV_WL32(side_data + 4, av_rescale_q(discard_padding,
2626                                             (AVRational){1, 1000000000},
2627                                             (AVRational){1, st->codec->sample_rate}));
2628 	}
2629 
2630     if (track->ms_compat)
2631         pkt->dts = timecode;
2632     else
2633         pkt->pts = timecode;
2634     pkt->pos = pos;
2635     if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2636         /*
2637          * For backward compatibility.
2638          * Historically, we have put subtitle duration
2639          * in convergence_duration, on the off chance
2640          * that the time_scale is less than 1us, which
2641          * could result in a 32bit overflow on the
2642          * normal duration field.
2643          */
2644         pkt->convergence_duration = lace_duration;
2645     }
2646 
2647     if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE ||
2648         lace_duration <= INT_MAX) {
2649         /*
2650          * For non subtitle tracks, just store the duration
2651          * as normal.
2652          *
2653          * If it's a subtitle track and duration value does
2654          * not overflow a uint32, then also store it normally.
2655          */
2656         pkt->duration = lace_duration;
2657     }
2658 
2659     dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2660     matroska->prev_pkt = pkt;
2661 
2662     return 0;
2663 
2664 fail:
2665     if (pkt_data != data)
2666         av_freep(&pkt_data);
2667     return res;
2668 }
2669 
matroska_parse_block(MatroskaDemuxContext * matroska,uint8_t * data,int size,int64_t pos,uint64_t cluster_time,uint64_t block_duration,int is_keyframe,uint8_t * additional,uint64_t additional_id,int additional_size,int64_t cluster_pos,int64_t discard_padding)2670 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
2671                                 int size, int64_t pos, uint64_t cluster_time,
2672                                 uint64_t block_duration, int is_keyframe,
2673                                 uint8_t *additional, uint64_t additional_id, int additional_size,
2674                                 int64_t cluster_pos, int64_t discard_padding)
2675 {
2676     uint64_t timecode = AV_NOPTS_VALUE;
2677     MatroskaTrack *track;
2678     int res = 0;
2679     AVStream *st;
2680     int16_t block_time;
2681     uint32_t *lace_size = NULL;
2682     int n, flags, laces = 0;
2683     uint64_t num;
2684     int trust_default_duration = 1;
2685 
2686     if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2687         av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2688         return n;
2689     }
2690     data += n;
2691     size -= n;
2692 
2693     track = matroska_find_track_by_num(matroska, num);
2694     if (!track || !track->stream) {
2695         av_log(matroska->ctx, AV_LOG_INFO,
2696                "Invalid stream %"PRIu64" or size %u\n", num, size);
2697         return AVERROR_INVALIDDATA;
2698     } else if (size <= 3)
2699         return 0;
2700     st = track->stream;
2701     if (st->discard >= AVDISCARD_ALL)
2702         return res;
2703     av_assert1(block_duration != AV_NOPTS_VALUE);
2704 
2705     block_time = sign_extend(AV_RB16(data), 16);
2706     data      += 2;
2707     flags      = *data++;
2708     size      -= 3;
2709     if (is_keyframe == -1)
2710         is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2711 
2712     if (cluster_time != (uint64_t) -1 &&
2713         (block_time >= 0 || cluster_time >= -block_time)) {
2714         timecode = cluster_time + block_time - track->codec_delay;
2715         if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2716             timecode < track->end_timecode)
2717             is_keyframe = 0;  /* overlapping subtitles are not key frame */
2718         if (is_keyframe)
2719             av_add_index_entry(st, cluster_pos, timecode, 0, 0,
2720                                AVINDEX_KEYFRAME);
2721     }
2722 
2723     if (matroska->skip_to_keyframe &&
2724         track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2725         if (timecode < matroska->skip_to_timecode)
2726             return res;
2727         if (is_keyframe)
2728             matroska->skip_to_keyframe = 0;
2729         else if (!st->skip_to_keyframe) {
2730             av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2731             matroska->skip_to_keyframe = 0;
2732         }
2733     }
2734 
2735     res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
2736                                &lace_size, &laces);
2737 
2738     if (res)
2739         goto end;
2740 
2741     if (track->audio.samplerate == 8000) {
2742         // If this is needed for more codecs, then add them here
2743         if (st->codec->codec_id == AV_CODEC_ID_AC3) {
2744             if (track->audio.samplerate != st->codec->sample_rate || !st->codec->frame_size)
2745                 trust_default_duration = 0;
2746         }
2747     }
2748 
2749     if (!block_duration && trust_default_duration)
2750         block_duration = track->default_duration * laces / matroska->time_scale;
2751 
2752     if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
2753         track->end_timecode =
2754             FFMAX(track->end_timecode, timecode + block_duration);
2755 
2756     for (n = 0; n < laces; n++) {
2757         int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
2758 
2759         if (lace_size[n] > size) {
2760             av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2761             break;
2762         }
2763 
2764         if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2765              st->codec->codec_id == AV_CODEC_ID_COOK   ||
2766              st->codec->codec_id == AV_CODEC_ID_SIPR   ||
2767              st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2768             st->codec->block_align && track->audio.sub_packet_size) {
2769             res = matroska_parse_rm_audio(matroska, track, st, data,
2770                                           lace_size[n],
2771                                           timecode, pos);
2772             if (res)
2773                 goto end;
2774 
2775         } else if (st->codec->codec_id == AV_CODEC_ID_WEBVTT) {
2776             res = matroska_parse_webvtt(matroska, track, st,
2777                                         data, lace_size[n],
2778                                         timecode, lace_duration,
2779                                         pos);
2780             if (res)
2781                 goto end;
2782         } else {
2783             res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2784                                        timecode, lace_duration, pos,
2785                                        !n ? is_keyframe : 0,
2786                                        additional, additional_id, additional_size,
2787                                        discard_padding);
2788             if (res)
2789                 goto end;
2790         }
2791 
2792         if (timecode != AV_NOPTS_VALUE)
2793             timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2794         data += lace_size[n];
2795         size -= lace_size[n];
2796     }
2797 
2798 end:
2799     av_free(lace_size);
2800     return res;
2801 }
2802 
matroska_parse_cluster_incremental(MatroskaDemuxContext * matroska)2803 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
2804 {
2805     EbmlList *blocks_list;
2806     MatroskaBlock *blocks;
2807     int i, res;
2808     res = ebml_parse(matroska,
2809                      matroska_cluster_incremental_parsing,
2810                      &matroska->current_cluster);
2811     if (res == 1) {
2812         /* New Cluster */
2813         if (matroska->current_cluster_pos)
2814             ebml_level_end(matroska);
2815         ebml_free(matroska_cluster, &matroska->current_cluster);
2816         memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2817         matroska->current_cluster_num_blocks = 0;
2818         matroska->current_cluster_pos        = avio_tell(matroska->ctx->pb);
2819         matroska->prev_pkt                   = NULL;
2820         /* sizeof the ID which was already read */
2821         if (matroska->current_id)
2822             matroska->current_cluster_pos -= 4;
2823         res = ebml_parse(matroska,
2824                          matroska_clusters_incremental,
2825                          &matroska->current_cluster);
2826         /* Try parsing the block again. */
2827         if (res == 1)
2828             res = ebml_parse(matroska,
2829                              matroska_cluster_incremental_parsing,
2830                              &matroska->current_cluster);
2831     }
2832 
2833     if (!res &&
2834         matroska->current_cluster_num_blocks <
2835         matroska->current_cluster.blocks.nb_elem) {
2836         blocks_list = &matroska->current_cluster.blocks;
2837         blocks      = blocks_list->elem;
2838 
2839         matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2840         i                                    = blocks_list->nb_elem - 1;
2841         if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2842             int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2843             uint8_t* additional = blocks[i].additional.size > 0 ?
2844                                     blocks[i].additional.data : NULL;
2845             if (!blocks[i].non_simple)
2846                 blocks[i].duration = 0;
2847             res = matroska_parse_block(matroska, blocks[i].bin.data,
2848                                        blocks[i].bin.size, blocks[i].bin.pos,
2849                                        matroska->current_cluster.timecode,
2850                                        blocks[i].duration, is_keyframe,
2851                                        additional, blocks[i].additional_id,
2852                                        blocks[i].additional.size,
2853                                        matroska->current_cluster_pos,
2854                                        blocks[i].discard_padding);
2855         }
2856     }
2857 
2858     return res;
2859 }
2860 
matroska_parse_cluster(MatroskaDemuxContext * matroska)2861 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
2862 {
2863     MatroskaCluster cluster = { 0 };
2864     EbmlList *blocks_list;
2865     MatroskaBlock *blocks;
2866     int i, res;
2867     int64_t pos;
2868 
2869     if (!matroska->contains_ssa)
2870         return matroska_parse_cluster_incremental(matroska);
2871     pos = avio_tell(matroska->ctx->pb);
2872     matroska->prev_pkt = NULL;
2873     if (matroska->current_id)
2874         pos -= 4;  /* sizeof the ID which was already read */
2875     res         = ebml_parse(matroska, matroska_clusters, &cluster);
2876     blocks_list = &cluster.blocks;
2877     blocks      = blocks_list->elem;
2878     for (i = 0; i < blocks_list->nb_elem; i++)
2879         if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2880             int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2881             res = matroska_parse_block(matroska, blocks[i].bin.data,
2882                                        blocks[i].bin.size, blocks[i].bin.pos,
2883                                        cluster.timecode, blocks[i].duration,
2884                                        is_keyframe, NULL, 0, 0, pos,
2885                                        blocks[i].discard_padding);
2886         }
2887     ebml_free(matroska_cluster, &cluster);
2888     return res;
2889 }
2890 
matroska_read_packet(AVFormatContext * s,AVPacket * pkt)2891 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
2892 {
2893     MatroskaDemuxContext *matroska = s->priv_data;
2894 
2895     while (matroska_deliver_packet(matroska, pkt)) {
2896         int64_t pos = avio_tell(matroska->ctx->pb);
2897         if (matroska->done)
2898             return AVERROR_EOF;
2899         if (matroska_parse_cluster(matroska) < 0)
2900             matroska_resync(matroska, pos);
2901     }
2902 
2903     return 0;
2904 }
2905 
matroska_read_seek(AVFormatContext * s,int stream_index,int64_t timestamp,int flags)2906 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2907                               int64_t timestamp, int flags)
2908 {
2909     MatroskaDemuxContext *matroska = s->priv_data;
2910     MatroskaTrack *tracks = matroska->tracks.elem;
2911     AVStream *st = s->streams[stream_index];
2912     int i, index, index_sub, index_min;
2913 
2914     /* Parse the CUES now since we need the index data to seek. */
2915     if (matroska->cues_parsing_deferred > 0) {
2916         matroska->cues_parsing_deferred = 0;
2917         matroska_parse_cues(matroska);
2918     }
2919 
2920     if (!st->nb_index_entries)
2921         goto err;
2922     timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2923 
2924     if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2925         avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
2926                   SEEK_SET);
2927         matroska->current_id = 0;
2928         while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
2929             matroska_clear_queue(matroska);
2930             if (matroska_parse_cluster(matroska) < 0)
2931                 break;
2932         }
2933     }
2934 
2935     matroska_clear_queue(matroska);
2936     if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
2937         goto err;
2938 
2939     index_min = index;
2940     for (i = 0; i < matroska->tracks.nb_elem; i++) {
2941         tracks[i].audio.pkt_cnt        = 0;
2942         tracks[i].audio.sub_packet_cnt = 0;
2943         tracks[i].audio.buf_timecode   = AV_NOPTS_VALUE;
2944         tracks[i].end_timecode         = 0;
2945         if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2946             tracks[i].stream->discard != AVDISCARD_ALL) {
2947             index_sub = av_index_search_timestamp(
2948                 tracks[i].stream, st->index_entries[index].timestamp,
2949                 AVSEEK_FLAG_BACKWARD);
2950             while (index_sub >= 0 &&
2951                   index_min > 0 &&
2952                   tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
2953                   st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
2954                 index_min--;
2955         }
2956     }
2957 
2958     avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2959     matroska->current_id       = 0;
2960     if (flags & AVSEEK_FLAG_ANY) {
2961         st->skip_to_keyframe = 0;
2962         matroska->skip_to_timecode = timestamp;
2963     } else {
2964         st->skip_to_keyframe = 1;
2965         matroska->skip_to_timecode = st->index_entries[index].timestamp;
2966     }
2967     matroska->skip_to_keyframe = 1;
2968     matroska->done             = 0;
2969     matroska->num_levels       = 0;
2970     ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2971     return 0;
2972 err:
2973     // slightly hackish but allows proper fallback to
2974     // the generic seeking code.
2975     matroska_clear_queue(matroska);
2976     matroska->current_id = 0;
2977     st->skip_to_keyframe =
2978     matroska->skip_to_keyframe = 0;
2979     matroska->done = 0;
2980     matroska->num_levels = 0;
2981     return -1;
2982 }
2983 
matroska_read_close(AVFormatContext * s)2984 static int matroska_read_close(AVFormatContext *s)
2985 {
2986     MatroskaDemuxContext *matroska = s->priv_data;
2987     MatroskaTrack *tracks = matroska->tracks.elem;
2988     int n;
2989 
2990     matroska_clear_queue(matroska);
2991 
2992     for (n = 0; n < matroska->tracks.nb_elem; n++)
2993         if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2994             av_free(tracks[n].audio.buf);
2995     ebml_free(matroska_cluster, &matroska->current_cluster);
2996     ebml_free(matroska_segment, matroska);
2997 
2998     return 0;
2999 }
3000 
3001 typedef struct {
3002     int64_t start_time_ns;
3003     int64_t end_time_ns;
3004     int64_t start_offset;
3005     int64_t end_offset;
3006 } CueDesc;
3007 
3008 /* This function searches all the Cues and returns the CueDesc corresponding the
3009  * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3010  * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3011  */
get_cue_desc(AVFormatContext * s,int64_t ts,int64_t cues_start)3012 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3013     MatroskaDemuxContext *matroska = s->priv_data;
3014     CueDesc cue_desc;
3015     int i;
3016     int nb_index_entries = s->streams[0]->nb_index_entries;
3017     AVIndexEntry *index_entries = s->streams[0]->index_entries;
3018 	if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3019 	for (i = 1; i < nb_index_entries; i++) {
3020         if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3021             index_entries[i].timestamp * matroska->time_scale > ts) {
3022             break;
3023         }
3024     }
3025     --i;
3026     cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3027     cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3028     if (i != nb_index_entries - 1) {
3029         cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3030         cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3031     } else {
3032         cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3033         // FIXME: this needs special handling for files where Cues appear
3034         // before Clusters. the current logic assumes Cues appear after
3035         // Clusters.
3036         cue_desc.end_offset = cues_start - matroska->segment_start;
3037     }
3038     return cue_desc;
3039 }
3040 
webm_clusters_start_with_keyframe(AVFormatContext * s)3041 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3042 {
3043     MatroskaDemuxContext *matroska = s->priv_data;
3044     int64_t cluster_pos, before_pos;
3045     int index, rv = 1;
3046     if (s->streams[0]->nb_index_entries <= 0) return 0;
3047     // seek to the first cluster using cues.
3048     index = av_index_search_timestamp(s->streams[0], 0, 0);
3049     if (index < 0)  return 0;
3050     cluster_pos = s->streams[0]->index_entries[index].pos;
3051     before_pos = avio_tell(s->pb);
3052     while (1) {
3053         int64_t cluster_id = 0, cluster_length = 0;
3054         AVPacket *pkt;
3055         avio_seek(s->pb, cluster_pos, SEEK_SET);
3056         // read cluster id and length
3057         ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id);
3058         ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3059         if (cluster_id != 0xF43B675) { // done with all clusters
3060             break;
3061         }
3062         avio_seek(s->pb, cluster_pos, SEEK_SET);
3063         matroska->current_id = 0;
3064         matroska_clear_queue(matroska);
3065         if (matroska_parse_cluster(matroska) < 0 ||
3066             matroska->num_packets <= 0) {
3067             break;
3068         }
3069         pkt = matroska->packets[0];
3070         cluster_pos += cluster_length + 12; // 12 is the offset of the cluster id and length.
3071         if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3072             rv = 0;
3073             break;
3074         }
3075     }
3076     avio_seek(s->pb, before_pos, SEEK_SET);
3077     return rv;
3078 }
3079 
buffer_size_after_time_downloaded(int64_t time_ns,double search_sec,int64_t bps,double min_buffer,double * buffer,double * sec_to_download,AVFormatContext * s,int64_t cues_start)3080 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3081                                              double min_buffer, double* buffer,
3082                                              double* sec_to_download, AVFormatContext *s,
3083                                              int64_t cues_start)
3084 {
3085     double nano_seconds_per_second = 1000000000.0;
3086     double time_sec = time_ns / nano_seconds_per_second;
3087     int rv = 0;
3088     int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3089     int64_t end_time_ns = time_ns + time_to_search_ns;
3090     double sec_downloaded = 0.0;
3091     CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3092     if (desc_curr.start_time_ns == -1)
3093       return -1;
3094     *sec_to_download = 0.0;
3095 
3096     // Check for non cue start time.
3097     if (time_ns > desc_curr.start_time_ns) {
3098       int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3099       double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3100       double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3101       double timeToDownload = (cueBytes * 8.0) / bps;
3102 
3103       sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3104       *sec_to_download += timeToDownload;
3105 
3106       // Check if the search ends within the first cue.
3107       if (desc_curr.end_time_ns >= end_time_ns) {
3108           double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3109           double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3110           sec_downloaded = percent_to_sub * sec_downloaded;
3111           *sec_to_download = percent_to_sub * *sec_to_download;
3112       }
3113 
3114       if ((sec_downloaded + *buffer) <= min_buffer) {
3115           return 1;
3116       }
3117 
3118       // Get the next Cue.
3119       desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3120     }
3121 
3122     while (desc_curr.start_time_ns != -1) {
3123         int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3124         int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3125         double desc_sec = desc_ns / nano_seconds_per_second;
3126         double bits = (desc_bytes * 8.0);
3127         double time_to_download = bits / bps;
3128 
3129         sec_downloaded += desc_sec - time_to_download;
3130         *sec_to_download += time_to_download;
3131 
3132         if (desc_curr.end_time_ns >= end_time_ns) {
3133             double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3134             double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3135             sec_downloaded = percent_to_sub * sec_downloaded;
3136             *sec_to_download = percent_to_sub * *sec_to_download;
3137 
3138             if ((sec_downloaded + *buffer) <= min_buffer)
3139                 rv = 1;
3140             break;
3141         }
3142 
3143         if ((sec_downloaded + *buffer) <= min_buffer) {
3144             rv = 1;
3145             break;
3146         }
3147 
3148         desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3149     }
3150     *buffer = *buffer + sec_downloaded;
3151     return rv;
3152 }
3153 
3154 /* This function computes the bandwidth of the WebM file with the help of
3155  * buffer_size_after_time_downloaded() function. Both of these functions are
3156  * adapted from WebM Tools project and are adapted to work with FFmpeg's
3157  * Matroska parsing mechanism.
3158  *
3159  * Returns the bandwidth of the file on success; -1 on error.
3160  * */
webm_dash_manifest_compute_bandwidth(AVFormatContext * s,int64_t cues_start)3161 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3162 {
3163     MatroskaDemuxContext *matroska = s->priv_data;
3164     AVStream *st = s->streams[0];
3165     double bandwidth = 0.0;
3166     int i;
3167 
3168     for (i = 0; i < st->nb_index_entries; i++) {
3169         int64_t prebuffer_ns = 1000000000;
3170         int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3171         double nano_seconds_per_second = 1000000000.0;
3172         int64_t prebuffered_ns = time_ns + prebuffer_ns;
3173         double prebuffer_bytes = 0.0;
3174         int64_t temp_prebuffer_ns = prebuffer_ns;
3175         int64_t pre_bytes, pre_ns;
3176         double pre_sec, prebuffer, bits_per_second;
3177         CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3178 
3179         // Start with the first Cue.
3180         CueDesc desc_end = desc_beg;
3181 
3182         // Figure out how much data we have downloaded for the prebuffer. This will
3183         // be used later to adjust the bits per sample to try.
3184         while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3185             // Prebuffered the entire Cue.
3186             prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3187             temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3188             desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3189         }
3190         if (desc_end.start_time_ns == -1) {
3191             // The prebuffer is larger than the duration.
3192             return (matroska->duration * matroska->time_scale >= prebuffered_ns) ? -1 : 0;
3193         }
3194 
3195         // The prebuffer ends in the last Cue. Estimate how much data was
3196         // prebuffered.
3197         pre_bytes = desc_end.end_offset - desc_end.start_offset;
3198         pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3199         pre_sec = pre_ns / nano_seconds_per_second;
3200         prebuffer_bytes +=
3201             pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3202 
3203         prebuffer = prebuffer_ns / nano_seconds_per_second;
3204 
3205         // Set this to 0.0 in case our prebuffer buffers the entire video.
3206         bits_per_second = 0.0;
3207         do {
3208             int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
3209             int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
3210             double desc_sec = desc_ns / nano_seconds_per_second;
3211             double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
3212 
3213             // Drop the bps by the percentage of bytes buffered.
3214             double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
3215             double mod_bits_per_second = calc_bits_per_second * percent;
3216 
3217             if (prebuffer < desc_sec) {
3218                 double search_sec =
3219                     (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
3220 
3221                 // Add 1 so the bits per second should be a little bit greater than file
3222                 // datarate.
3223                 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
3224                 const double min_buffer = 0.0;
3225                 double buffer = prebuffer;
3226                 double sec_to_download = 0.0;
3227 
3228                 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
3229                                                            min_buffer, &buffer, &sec_to_download,
3230                                                            s, cues_start);
3231                 if (rv < 0) {
3232                     return -1;
3233                 } else if (rv == 0) {
3234                     bits_per_second = (double)(bps);
3235                     break;
3236                 }
3237             }
3238 
3239             desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3240         } while (desc_end.start_time_ns != -1);
3241         if (bandwidth < bits_per_second) bandwidth = bits_per_second;
3242     }
3243     return (int64_t)bandwidth;
3244 }
3245 
webm_dash_manifest_cues(AVFormatContext * s)3246 static int webm_dash_manifest_cues(AVFormatContext *s)
3247 {
3248     MatroskaDemuxContext *matroska = s->priv_data;
3249     EbmlList *seekhead_list = &matroska->seekhead;
3250     MatroskaSeekhead *seekhead = seekhead_list->elem;
3251     char *buf;
3252     int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
3253     int i;
3254 
3255     // determine cues start and end positions
3256     for (i = 0; i < seekhead_list->nb_elem; i++)
3257         if (seekhead[i].id == MATROSKA_ID_CUES)
3258             break;
3259 
3260     if (i >= seekhead_list->nb_elem) return -1;
3261 
3262     before_pos = avio_tell(matroska->ctx->pb);
3263     cues_start = seekhead[i].pos + matroska->segment_start;
3264     if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
3265         uint64_t cues_length = 0, cues_id = 0;
3266         ebml_read_num(matroska, matroska->ctx->pb, 4, &cues_id);
3267         ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
3268         cues_end = cues_start + cues_length + 11; // 11 is the offset of Cues ID.
3269     }
3270     avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
3271     if (cues_start == -1 || cues_end == -1) return -1;
3272 
3273     // parse the cues
3274     matroska_parse_cues(matroska);
3275 
3276     // cues start
3277     av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
3278 
3279     // cues end
3280     av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
3281 
3282     // bandwidth
3283     bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
3284     if (bandwidth < 0) return -1;
3285     av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
3286 
3287     // check if all clusters start with key frames
3288     av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
3289 
3290     // store cue point timestamps as a comma separated list for checking subsegment alignment in
3291     // the muxer. assumes that each timestamp cannot be more than 20 characters long.
3292     buf = av_malloc(s->streams[0]->nb_index_entries * 20 * sizeof(char));
3293     if (!buf) return -1;
3294     strcpy(buf, "");
3295     for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
3296         snprintf(buf, (i + 1) * 20 * sizeof(char),
3297                  "%s%" PRId64, buf, s->streams[0]->index_entries[i].timestamp);
3298         if (i != s->streams[0]->nb_index_entries - 1)
3299             strncat(buf, ",", sizeof(char));
3300     }
3301     av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
3302     av_free(buf);
3303 
3304     return 0;
3305 }
3306 
webm_dash_manifest_read_header(AVFormatContext * s)3307 static int webm_dash_manifest_read_header(AVFormatContext *s)
3308 {
3309     char *buf;
3310     int ret = matroska_read_header(s);
3311     MatroskaTrack *tracks;
3312     MatroskaDemuxContext *matroska = s->priv_data;
3313     if (ret) {
3314         av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
3315         return -1;
3316     }
3317 
3318     // initialization range
3319     // 5 is the offset of Cluster ID.
3320     av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, avio_tell(s->pb) - 5, 0);
3321 
3322     // basename of the file
3323     buf = strrchr(s->filename, '/');
3324     if (!buf) return -1;
3325     av_dict_set(&s->streams[0]->metadata, FILENAME, ++buf, 0);
3326 
3327     // duration
3328     buf = av_asprintf("%g", matroska->duration);
3329     if (!buf) return AVERROR(ENOMEM);
3330     av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
3331     av_free(buf);
3332 
3333     // track number
3334     tracks = matroska->tracks.elem;
3335     av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
3336 
3337     // parse the cues and populate Cue related fields
3338     return webm_dash_manifest_cues(s);
3339 }
3340 
webm_dash_manifest_read_packet(AVFormatContext * s,AVPacket * pkt)3341 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
3342 {
3343     return AVERROR_EOF;
3344 }
3345 
3346 AVInputFormat ff_matroska_demuxer = {
3347 	.name           = "matroska,webm",
3348     .long_name      = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
3349     .extensions     = "mkv,mk3d,mka,mks",
3350     .priv_data_size = sizeof(MatroskaDemuxContext),
3351     .read_probe     = matroska_probe,
3352     .read_header    = matroska_read_header,
3353     .read_packet    = matroska_read_packet,
3354     .read_close     = matroska_read_close,
3355     .read_seek      = matroska_read_seek,
3356     .mime_type      = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
3357 };
3358 
3359 AVInputFormat ff_webm_dash_manifest_demuxer = {
3360 	.name           = "webm_dash_manifest",
3361     .long_name      = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
3362     .priv_data_size = sizeof(MatroskaDemuxContext),
3363     .read_header    = webm_dash_manifest_read_header,
3364     .read_packet    = webm_dash_manifest_read_packet,
3365     .read_close     = matroska_read_close,
3366 };
3367