FFmpeg
matroskadec.c
Go to the documentation of this file.
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 #include "config_components.h"
33 
34 #include <inttypes.h>
35 #include <stdio.h>
36 
37 #include "libavutil/avstring.h"
38 #include "libavutil/base64.h"
39 #include "libavutil/bprint.h"
40 #include "libavutil/dict.h"
42 #include "libavutil/display.h"
44 #include "libavutil/intfloat.h"
45 #include "libavutil/intreadwrite.h"
46 #include "libavutil/lzo.h"
48 #include "libavutil/mathematics.h"
49 #include "libavutil/mem.h"
50 #include "libavutil/opt.h"
51 #include "libavutil/pixdesc.h"
53 #include "libavutil/spherical.h"
54 
55 #include "libavcodec/bytestream.h"
56 #include "libavcodec/defs.h"
57 #include "libavcodec/flac.h"
58 #include "libavcodec/itut35.h"
59 #include "libavcodec/mpeg4audio.h"
61 
62 #include "avformat.h"
63 #include "avio_internal.h"
64 #include "demux.h"
65 #include "dovi_isom.h"
66 #include "internal.h"
67 #include "isom.h"
68 #include "matroska.h"
69 #include "oggdec.h"
70 /* For ff_codec_get_id(). */
71 #include "riff.h"
72 #include "rmsipr.h"
73 
74 #if CONFIG_BZLIB
75 #include <bzlib.h>
76 #endif
77 #if CONFIG_ZLIB
78 #include <zlib.h>
79 #endif
80 
81 #include "qtpalette.h"
82 
83 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
84 #define NEEDS_CHECKING 2 /* Indicates that some error checks
85  * still need to be performed */
86 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
87  * syntax level used for parsing ended. */
88 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
89  * of unkown, potentially damaged data is encountered,
90  * it is considered an error. */
91 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
92  * to this many bytes of unknown data for the
93  * SKIP_THRESHOLD check. */
94 
95 typedef enum {
101  EBML_UTF8,
102  EBML_BIN,
104  EBML_LEVEL1,
108 
109 typedef struct CountedElement {
110  union {
111  uint64_t u;
112  int64_t i;
113  double f;
114  char *s;
115  } el;
116  unsigned count;
118 
119 typedef const struct EbmlSyntax {
120  uint32_t id;
121  uint8_t type;
122  uint8_t is_counted;
124  size_t data_offset;
125  union {
126  int64_t i;
127  uint64_t u;
128  double f;
129  const char *s;
130  const struct EbmlSyntax *n;
131  } def;
132 } EbmlSyntax;
133 
134 typedef struct EbmlList {
135  int nb_elem;
136  unsigned int alloc_elem_size;
137  void *elem;
139 
140 typedef struct EbmlBin {
141  int size;
143  uint8_t *data;
146 
147 typedef struct Ebml {
148  uint64_t version;
149  uint64_t max_size;
150  uint64_t id_length;
151  char *doctype;
152  uint64_t doctype_version;
153 } Ebml;
154 
155 typedef struct MatroskaTrackCompression {
156  uint64_t algo;
159 
160 typedef struct MatroskaTrackEncryption {
161  uint64_t algo;
164 
165 typedef struct MatroskaTrackEncoding {
166  uint64_t scope;
167  uint64_t type;
171 
172 typedef struct MatroskaMasteringMeta {
173  double r_x;
174  double r_y;
175  double g_x;
176  double g_y;
177  double b_x;
178  double b_y;
179  double white_x;
180  double white_y;
184 
185 typedef struct MatroskaTrackVideoColor {
188  uint64_t chroma_sub_horz;
189  uint64_t chroma_sub_vert;
190  uint64_t cb_sub_horz;
191  uint64_t cb_sub_vert;
194  uint64_t range;
195  uint64_t transfer_characteristics;
196  uint64_t primaries;
197  uint64_t max_cll;
198  uint64_t max_fall;
201 
202 typedef struct MatroskaTrackVideoProjection {
203  uint64_t type;
204  EbmlBin private;
205  double yaw;
206  double pitch;
207  double roll;
209 
210 typedef struct MatroskaTrackVideo {
211  double frame_rate;
212  uint64_t display_width;
213  uint64_t display_height;
214  uint64_t pixel_width;
215  uint64_t pixel_height;
216  uint64_t cropped_width;
217  uint64_t cropped_height;
219  uint64_t pixel_cropt;
220  uint64_t pixel_cropl;
221  uint64_t pixel_cropb;
222  uint64_t pixel_cropr;
223  uint64_t display_unit;
224  uint64_t interlaced;
225  uint64_t field_order;
226  uint64_t stereo_mode;
227  uint64_t alpha_mode;
231 
232 typedef struct MatroskaTrackAudio {
233  double samplerate;
235  uint64_t bitdepth;
236  uint64_t channels;
237 
238  /* real audio header (extracted from extradata) */
241  int frame_size;
242  int sub_packet_size;
244  int pkt_cnt;
245  uint64_t buf_timecode;
246  uint8_t *buf;
248 
249 typedef struct MatroskaTrackPlane {
250  uint64_t uid;
251  uint64_t type;
253 
254 typedef struct MatroskaTrackOperation {
257 
258 typedef struct MatroskaBlockAdditionMapping {
259  uint64_t value;
260  char *name;
261  uint64_t type;
264 
265 typedef struct MatroskaTrack {
266  uint64_t num;
267  uint64_t uid;
268  uint64_t type;
269  char *name;
270  char *codec_id;
272  char *language;
273  double time_scale;
275  uint64_t flag_default;
276  uint64_t flag_forced;
277  uint64_t flag_comment;
282  uint64_t seek_preroll;
287  uint64_t codec_delay;
289 
290  AVStream *stream;
293  int needs_decoding;
294  uint64_t max_block_additional_id;
296 
300 
301 typedef struct MatroskaAttachment {
302  uint64_t uid;
303  char *filename;
304  char *description;
305  char *mime;
307 
310 
311 typedef struct MatroskaChapter {
312  uint64_t start;
313  uint64_t end;
314  uint64_t uid;
315  char *title;
316 
319 
320 typedef struct MatroskaIndexPos {
321  uint64_t track;
322  uint64_t pos;
324 
325 typedef struct MatroskaIndex {
326  uint64_t time;
329 
330 typedef struct MatroskaTag {
331  char *name;
332  char *string;
333  char *lang;
334  uint64_t def;
337 
338 typedef struct MatroskaTagTarget {
339  char *type;
340  uint64_t typevalue;
341  uint64_t trackuid;
342  uint64_t chapteruid;
343  uint64_t attachuid;
345 
346 typedef struct MatroskaTags {
348  EbmlList tag;
349 } MatroskaTags;
350 
351 typedef struct MatroskaSeekhead {
352  uint64_t id;
353  uint64_t pos;
355 
356 typedef struct MatroskaLevel {
357  uint64_t start;
358  uint64_t length;
359 } MatroskaLevel;
360 
361 typedef struct MatroskaBlockMore {
362  uint64_t additional_id;
365 
366 typedef struct MatroskaBlock {
367  uint64_t duration;
369  uint64_t non_simple;
373 } MatroskaBlock;
374 
375 typedef struct MatroskaCluster {
377  uint64_t timecode;
380 
381 typedef struct MatroskaLevel1Element {
383  uint32_t id;
384  int parsed;
386 
387 typedef struct MatroskaDemuxContext {
388  const AVClass *class;
390 
391  /* EBML stuff */
394  uint32_t current_id;
397 
398  uint64_t time_scale;
399  double duration;
400  char *title;
401  char *muxingapp;
406  EbmlList index;
407  EbmlList tags;
409 
410  /* byte position of the segment inside the stream */
412 
413  /* This packet coincides with FFFormatContext.parse_pkt
414  * and is not owned by us. */
415  AVPacket *pkt;
416 
417  /* the packet queue */
419 
420  int done;
421 
422  /* What to skip before effectively reading a packet. */
423  int skip_to_keyframe;
425 
426  /* File has a CUES element, but we defer parsing until it is needed. */
428 
429  /* Level1 elements and whether they were read yet */
431  int num_level1_elems;
432 
434 
435  int is_webm;
436 
437  /* WebM DASH Manifest live flag */
438  int is_live;
439 
440  /* Bandwidth value for WebM DASH Manifest */
441  int bandwidth;
443 
444 #define CHILD_OF(parent) { .def = { .n = parent } }
445 
446 // The following forward declarations need their size because
447 // a tentative definition with internal linkage must not be an
448 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
449 // Removing the sizes breaks MSVC.
456 
457 static EbmlSyntax ebml_header[] = {
458  { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
459  { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, 0, offsetof(Ebml, max_size), { .u = 8 } },
460  { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, 0, offsetof(Ebml, id_length), { .u = 4 } },
461  { EBML_ID_DOCTYPE, EBML_STR, 0, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
462  { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
466 };
467 
469  { EBML_ID_HEADER, EBML_NEST, 0, 0, 0, { .n = ebml_header } },
471  { 0 }
472 };
473 
474 static EbmlSyntax matroska_info[] = {
475  { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
477  { MATROSKA_ID_TITLE, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, title) },
479  { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, muxingapp) },
480  { MATROSKA_ID_DATEUTC, EBML_BIN, 0, 0, offsetof(MatroskaDemuxContext, date_utc) },
483 };
484 
492  { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_x) },
493  { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_y) },
494  { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 1, 0, offsetof(MatroskaMasteringMeta, min_luminance) },
495  { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, max_luminance) },
497 };
498 
500  { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
501  { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u = 0 } },
502  { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz) },
503  { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert) },
504  { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz) },
505  { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert) },
512  { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_fall) },
515 };
516 
520  { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f = 0.0 } },
521  { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f = 0.0 } },
522  { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f = 0.0 } },
524 };
525 
526 static EbmlSyntax matroska_track_video[] = {
527  { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideo, frame_rate) },
528  { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
529  { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
530  { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_width) },
531  { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_height) },
532  { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideo, color_space) },
533  { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, alpha_mode), { .u = 0 } },
536  { MATROSKA_ID_VIDEOPIXELCROPB, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_cropb), {.u = 0 } },
537  { MATROSKA_ID_VIDEOPIXELCROPT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_cropt), {.u = 0 } },
538  { MATROSKA_ID_VIDEOPIXELCROPL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_cropl), {.u = 0 } },
539  { MATROSKA_ID_VIDEOPIXELCROPR, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_cropr), {.u = 0 } },
546 };
547 
548 static EbmlSyntax matroska_track_audio[] = {
549  { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
550  { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
552  { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
554 };
555 
560 };
561 
563  { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
571 };
573  { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
574  { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
579 };
580 
584 };
585 
586 static EbmlSyntax matroska_track_plane[] = {
590 };
591 
593  { MATROSKA_ID_TRACKPLANE, EBML_NEST, 0, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
595 };
596 
600 };
601 
608 };
609 
610 static EbmlSyntax matroska_track[] = {
611  { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, 0, offsetof(MatroskaTrack, num) },
612  { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTrack, name) },
613  { MATROSKA_ID_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, uid) },
614  { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrack, type) },
615  { MATROSKA_ID_CODECID, EBML_STR, 0, 0, offsetof(MatroskaTrack, codec_id) },
616  { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrack, codec_priv) },
617  { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, codec_delay), { .u = 0 } },
618  { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
619  { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, 0, offsetof(MatroskaTrack, default_duration) },
620  { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
621  { MATROSKA_ID_TRACKFLAGCOMMENTARY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_comment), { .u = 0 } },
622  { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
623  { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
624  { MATROSKA_ID_TRACKFLAGHEARINGIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_hearingimpaired), { .u = 0 } },
625  { MATROSKA_ID_TRACKFLAGVISUALIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_visualimpaired), { .u = 0 } },
626  { MATROSKA_ID_TRACKFLAGTEXTDESCRIPTIONS, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_textdescriptions), { .u = 0 } },
627  { MATROSKA_ID_TRACKFLAGORIGINAL, EBML_UINT, 1, 0, offsetof(MatroskaTrack, flag_original), {.u = 0 } },
629  { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
630  { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
632  { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, max_block_additional_id), { .u = 0 } },
633  { MATROSKA_ID_TRACKBLKADDMAPPING, EBML_NEST, 0, sizeof(MatroskaBlockAdditionMapping), offsetof(MatroskaTrack, block_addition_mappings), { .n = matroska_block_addition_mapping } },
634  { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, 0, offsetof(MatroskaTrack, seek_preroll), { .u = 0 } },
644 };
645 
646 static EbmlSyntax matroska_tracks[] = {
647  { MATROSKA_ID_TRACKENTRY, EBML_NEST, 0, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
649 };
650 
651 static EbmlSyntax matroska_attachment[] = {
652  { MATROSKA_ID_FILEUID, EBML_UINT, 0, 0, offsetof(MatroskaAttachment, uid) },
653  { MATROSKA_ID_FILENAME, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, filename) },
654  { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, 0, offsetof(MatroskaAttachment, mime) },
655  { MATROSKA_ID_FILEDATA, EBML_BIN, 0, 0, offsetof(MatroskaAttachment, bin) },
658 };
659 
660 static EbmlSyntax matroska_attachments[] = {
661  { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, 0, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
663 };
664 
666  { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaChapter, title) },
670 };
671 
673  { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
674  { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
675  { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaChapter, uid) },
682 };
683 
684 static EbmlSyntax matroska_chapter[] = {
685  { MATROSKA_ID_CHAPTERATOM, EBML_NEST, 0, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
691 };
692 
693 static EbmlSyntax matroska_chapters[] = {
694  { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter } },
696 };
697 
698 static EbmlSyntax matroska_index_pos[] = {
699  { MATROSKA_ID_CUETRACK, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, track) },
705 };
706 
707 static EbmlSyntax matroska_index_entry[] = {
708  { MATROSKA_ID_CUETIME, EBML_UINT, 0, 0, offsetof(MatroskaIndex, time) },
711 };
712 
713 static EbmlSyntax matroska_index[] = {
716 };
717 
718 static EbmlSyntax matroska_simpletag[] = {
719  { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTag, name) },
720  { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaTag, string) },
721  { MATROSKA_ID_TAGLANG, EBML_STR, 0, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
722  { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
723  { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
724  { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
726 };
727 
728 static EbmlSyntax matroska_tagtargets[] = {
730  { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
731  { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, trackuid), { .u = 0 } },
732  { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, chapteruid), { .u = 0 } },
733  { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, attachuid), { .u = 0 } },
735 };
736 
737 static EbmlSyntax matroska_tag[] = {
738  { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
739  { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
741 };
742 
743 static EbmlSyntax matroska_tags[] = {
744  { MATROSKA_ID_TAG, EBML_NEST, 0, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
746 };
747 
749  { MATROSKA_ID_SEEKID, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, id) },
750  { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
752 };
753 
754 static EbmlSyntax matroska_seekhead[] = {
755  { MATROSKA_ID_SEEKENTRY, EBML_NEST, 0, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
757 };
758 
759 static EbmlSyntax matroska_segment[] = {
761  { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, 0, { .n = matroska_info } },
762  { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tracks } },
764  { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_chapters } },
765  { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, 0, { .n = matroska_index } },
766  { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tags } },
767  { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, 0, { .n = matroska_seekhead } },
768  { 0 } /* We don't want to go back to level 0, so don't add the parent. */
769 };
770 
771 static EbmlSyntax matroska_segments[] = {
772  { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, 0, { .n = matroska_segment } },
773  { 0 }
774 };
775 
776 static EbmlSyntax matroska_blockmore[] = {
777  { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, 0, offsetof(MatroskaBlockMore,additional_id), { .u = MATROSKA_BLOCK_ADD_ID_OPAQUE } },
778  { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, 0, offsetof(MatroskaBlockMore,additional) },
780 };
781 
783  { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, sizeof(MatroskaBlockMore), offsetof(MatroskaBlock, blockmore), { .n = matroska_blockmore } },
785 };
786 
787 static EbmlSyntax matroska_blockgroup[] = {
788  { MATROSKA_ID_BLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
791  { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, 0, offsetof(MatroskaBlock, discard_padding) },
792  { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 1, 0, offsetof(MatroskaBlock, reference) },
794  { 1, EBML_UINT, 0, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
796 };
797 
798 // The following array contains SimpleBlock and BlockGroup twice
799 // in order to reuse the other values for matroska_cluster_enter.
801  { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
802  { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, 0, { .n = matroska_blockgroup } },
803  { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, 0, offsetof(MatroskaCluster, timecode) },
809 };
810 
812  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, 0, { .n = &matroska_cluster_parsing[2] } },
813  { 0 }
814 };
815 #undef CHILD_OF
816 
817 static const CodecMime mkv_image_mime_tags[] = {
818  {"image/gif" , AV_CODEC_ID_GIF},
819  {"image/jpeg" , AV_CODEC_ID_MJPEG},
820  {"image/png" , AV_CODEC_ID_PNG},
821  {"image/tiff" , AV_CODEC_ID_TIFF},
822 
823  {"" , AV_CODEC_ID_NONE}
824 };
825 
826 static const CodecMime mkv_mime_tags[] = {
827  {"application/x-truetype-font", AV_CODEC_ID_TTF},
828  {"application/x-font" , AV_CODEC_ID_TTF},
829  {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
830  {"binary" , AV_CODEC_ID_BIN_DATA},
831 
832  {"" , AV_CODEC_ID_NONE}
833 };
834 
836  "left",
837  "right",
838  "background",
839 };
840 
841 static const char *const matroska_doctypes[] = { "matroska", "webm" };
842 
843 /*
844  * This function prepares the status for parsing of level 1 elements.
845  */
846 static int matroska_reset_status(MatroskaDemuxContext *matroska,
847  uint32_t id, int64_t position)
848 {
849  int64_t err = 0;
850  if (position >= 0) {
851  err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
852  if (err > 0)
853  err = 0;
854  } else
855  position = avio_tell(matroska->ctx->pb);
856 
857  matroska->current_id = id;
858  matroska->num_levels = 1;
859  matroska->unknown_count = 0;
860  matroska->resync_pos = position;
861  if (id)
862  matroska->resync_pos -= (av_log2(id) + 7) / 8;
863 
864  return err;
865 }
866 
867 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
868 {
869  AVIOContext *pb = matroska->ctx->pb;
870  uint32_t id;
871 
872  /* Try to seek to the last position to resync from. If this doesn't work,
873  * we resync from the earliest position available: The start of the buffer. */
874  if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
875  av_log(matroska->ctx, AV_LOG_WARNING,
876  "Seek to desired resync point failed. Seeking to "
877  "earliest point available instead.\n");
878  avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
879  last_pos + 1), SEEK_SET);
880  }
881 
882  id = avio_rb32(pb);
883 
884  // try to find a toplevel element
885  while (!avio_feof(pb)) {
886  if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
887  id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
889  id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
890  /* Prepare the context for parsing of a level 1 element. */
891  matroska_reset_status(matroska, id, -1);
892  /* Given that we are here means that an error has occurred,
893  * so treat the segment as unknown length in order not to
894  * discard valid data that happens to be beyond the designated
895  * end of the segment. */
896  matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
897  return 0;
898  }
899  id = (id << 8) | avio_r8(pb);
900  }
901 
902  matroska->done = 1;
903  return pb->error ? pb->error : AVERROR_EOF;
904 }
905 
906 /*
907  * Read: an "EBML number", which is defined as a variable-length
908  * array of bytes. The first byte indicates the length by giving a
909  * number of 0-bits followed by a one. The position of the first
910  * "one" bit inside the first byte indicates the length of this
911  * number.
912  * Returns: number of bytes read, < 0 on error
913  */
914 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
915  int max_size, uint64_t *number, int eof_forbidden)
916 {
917  int read, n = 1;
918  uint64_t total;
919  int64_t pos;
920 
921  /* The first byte tells us the length in bytes - except when it is zero. */
922  total = avio_r8(pb);
923  if (pb->eof_reached)
924  goto err;
925 
926  /* get the length of the EBML number */
927  read = 8 - ff_log2_tab[total];
928 
929  if (!total || read > max_size) {
930  pos = avio_tell(pb) - 1;
931  if (!total) {
932  av_log(matroska->ctx, AV_LOG_ERROR,
933  "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
934  "of an EBML number\n", pos, pos);
935  } else {
936  av_log(matroska->ctx, AV_LOG_ERROR,
937  "Length %d indicated by an EBML number's first byte 0x%02x "
938  "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
939  read, (uint8_t) total, pos, pos, max_size);
940  }
941  return AVERROR_INVALIDDATA;
942  }
943 
944  /* read out length */
945  total ^= 1 << ff_log2_tab[total];
946  while (n++ < read)
947  total = (total << 8) | avio_r8(pb);
948 
949  if (pb->eof_reached) {
950  eof_forbidden = 1;
951  goto err;
952  }
953 
954  *number = total;
955 
956  return read;
957 
958 err:
959  pos = avio_tell(pb);
960  if (pb->error) {
961  av_log(matroska->ctx, AV_LOG_ERROR,
962  "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
963  pos, pos);
964  return pb->error;
965  }
966  if (eof_forbidden) {
967  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
968  "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
969  return AVERROR(EIO);
970  }
971  return AVERROR_EOF;
972 }
973 
974 /**
975  * Read a EBML length value.
976  * This needs special handling for the "unknown length" case which has multiple
977  * encodings.
978  */
979 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
980  uint64_t *number)
981 {
982  int res = ebml_read_num(matroska, pb, 8, number, 1);
983  if (res > 0 && *number + 1 == 1ULL << (7 * res))
984  *number = EBML_UNKNOWN_LENGTH;
985  return res;
986 }
987 
988 /*
989  * Read the next element as an unsigned int.
990  * Returns NEEDS_CHECKING unless size == 0.
991  */
992 static int ebml_read_uint(AVIOContext *pb, int size,
993  uint64_t default_value, uint64_t *num)
994 {
995  int n = 0;
996 
997  if (size == 0) {
998  *num = default_value;
999  return 0;
1000  }
1001  /* big-endian ordering; build up number */
1002  *num = 0;
1003  while (n++ < size)
1004  *num = (*num << 8) | avio_r8(pb);
1005 
1006  return NEEDS_CHECKING;
1008 
1009 /*
1010  * Read the next element as a signed int.
1011  * Returns NEEDS_CHECKING unless size == 0.
1012  */
1013 static int ebml_read_sint(AVIOContext *pb, int size,
1014  int64_t default_value, int64_t *num)
1015 {
1016  int n = 1;
1017 
1018  if (size == 0) {
1019  *num = default_value;
1020  return 0;
1021  } else {
1022  *num = sign_extend(avio_r8(pb), 8);
1023 
1024  /* big-endian ordering; build up number */
1025  while (n++ < size)
1026  *num = ((uint64_t)*num << 8) | avio_r8(pb);
1027  }
1028 
1029  return NEEDS_CHECKING;
1031 
1032 /*
1033  * Read the next element as a float.
1034  * Returns 0 if size == 0, NEEDS_CHECKING or < 0 on obvious failure.
1035  */
1036 static int ebml_read_float(AVIOContext *pb, int size,
1037  double default_value, double *num)
1038 {
1039  if (size == 0) {
1040  *num = default_value;
1041  return 0;
1042  } else if (size == 4) {
1043  *num = av_int2float(avio_rb32(pb));
1044  } else if (size == 8) {
1045  *num = av_int2double(avio_rb64(pb));
1046  } else
1047  return AVERROR_INVALIDDATA;
1048 
1049  return NEEDS_CHECKING;
1051 
1052 /*
1053  * Read the next element as an ASCII string.
1054  * 0 is success, < 0 or NEEDS_CHECKING is failure.
1055  */
1056 static int ebml_read_ascii(AVIOContext *pb, int size,
1057  const char *default_value, char **str)
1058 {
1059  char *res;
1060  int ret;
1061 
1062  if (size == 0 && default_value) {
1063  res = av_strdup(default_value);
1064  if (!res)
1065  return AVERROR(ENOMEM);
1066  } else {
1067  /* EBML strings are usually not 0-terminated, so we allocate one
1068  * byte more, read the string and NUL-terminate it ourselves. */
1069  if (!(res = av_malloc(size + 1)))
1070  return AVERROR(ENOMEM);
1071  if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
1072  av_free(res);
1073  return ret < 0 ? ret : NEEDS_CHECKING;
1074  }
1075  (res)[size] = '\0';
1076  }
1077  av_free(*str);
1078  *str = res;
1079 
1080  return 0;
1082 
1083 /*
1084  * Read the next element as binary data.
1085  * 0 is success, < 0 or NEEDS_CHECKING is failure.
1086  */
1087 static int ebml_read_binary(AVIOContext *pb, int length,
1088  int64_t pos, EbmlBin *bin)
1089 {
1090  int ret;
1091 
1093  if (ret < 0)
1094  return ret;
1095  memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1096 
1097  bin->data = bin->buf->data;
1098  bin->size = length;
1099  bin->pos = pos;
1100  if ((ret = avio_read(pb, bin->data, length)) != length) {
1101  av_buffer_unref(&bin->buf);
1102  bin->data = NULL;
1103  bin->size = 0;
1104  return ret < 0 ? ret : NEEDS_CHECKING;
1105  }
1106 
1107  return 0;
1108 }
1110 /*
1111  * Read the next element, but only the header. The contents
1112  * are supposed to be sub-elements which can be read separately.
1113  * 0 is success, < 0 is failure.
1114  */
1115 static int ebml_read_master(MatroskaDemuxContext *matroska,
1116  uint64_t length, int64_t pos)
1117 {
1119 
1120  if (matroska->num_levels >= EBML_MAX_DEPTH) {
1121  av_log(matroska->ctx, AV_LOG_ERROR,
1122  "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1123  return AVERROR(ENOSYS);
1124  }
1125 
1126  level = &matroska->levels[matroska->num_levels++];
1127  level->start = pos;
1128  level->length = length;
1129 
1130  return 0;
1132 
1133 /*
1134  * Read a signed "EBML number"
1135  * Return: number of bytes processed, < 0 on error
1136  */
1137 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1138  AVIOContext *pb, int64_t *num)
1139 {
1140  uint64_t unum;
1141  int res;
1142 
1143  /* read as unsigned number first */
1144  if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1145  return res;
1146 
1147  /* make signed (weird way) */
1148  *num = unum - ((1LL << (7 * res - 1)) - 1);
1149 
1150  return res;
1151 }
1152 
1153 static int ebml_parse(MatroskaDemuxContext *matroska,
1154  EbmlSyntax *syntax, void *data);
1155 
1156 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1157 {
1158  int i;
1159 
1160  // Whoever touches this should be aware of the duplication
1161  // existing in matroska_cluster_parsing.
1162  for (i = 0; syntax[i].id; i++)
1163  if (id == syntax[i].id)
1164  break;
1165 
1166  return &syntax[i];
1167 }
1168 
1169 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1170  void *data)
1171 {
1172  int res;
1173 
1174  if (data) {
1175  for (int i = 0; syntax[i].id; i++) {
1176  void *dst = (char *)data + syntax[i].data_offset;
1177  switch (syntax[i].type) {
1178  case EBML_UINT:
1179  *(uint64_t *)dst = syntax[i].def.u;
1180  break;
1181  case EBML_SINT:
1182  *(int64_t *) dst = syntax[i].def.i;
1183  break;
1184  case EBML_FLOAT:
1185  *(double *) dst = syntax[i].def.f;
1186  break;
1187  case EBML_STR:
1188  case EBML_UTF8:
1189  // the default may be NULL
1190  if (syntax[i].def.s) {
1191  *(char**)dst = av_strdup(syntax[i].def.s);
1192  if (!*(char**)dst)
1193  return AVERROR(ENOMEM);
1194  }
1195  break;
1196  }
1197  }
1198 
1199  if (!matroska->levels[matroska->num_levels - 1].length) {
1200  matroska->num_levels--;
1201  return 0;
1202  }
1203  }
1204 
1205  do {
1206  res = ebml_parse(matroska, syntax, data);
1207  } while (!res);
1208 
1209  return res == LEVEL_ENDED ? 0 : res;
1210 }
1211 
1212 static int is_ebml_id_valid(uint32_t id)
1213 {
1214  // Due to endian nonsense in Matroska, the highest byte with any bits set
1215  // will contain the leading length bit. This bit in turn identifies the
1216  // total byte length of the element by its position within the byte.
1217  unsigned int bits = av_log2(id);
1218  return id && (bits + 7) / 8 == (8 - bits % 8);
1220 
1221 /*
1222  * Allocate and return the entry for the level1 element with the given ID. If
1223  * an entry already exists, return the existing entry.
1224  */
1226  uint32_t id, int64_t pos)
1227 {
1228  int i;
1229  MatroskaLevel1Element *elem;
1230 
1231  if (!is_ebml_id_valid(id))
1232  return NULL;
1233 
1234  // Some files link to all clusters; useless.
1235  if (id == MATROSKA_ID_CLUSTER)
1236  return NULL;
1237 
1238  // There can be multiple SeekHeads and Tags.
1239  for (i = 0; i < matroska->num_level1_elems; i++) {
1240  if (matroska->level1_elems[i].id == id) {
1241  if (matroska->level1_elems[i].pos == pos ||
1242  id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1243  return &matroska->level1_elems[i];
1244  }
1245  }
1246 
1247  // Only a completely broken file would have more elements.
1248  if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1249  av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1250  return NULL;
1251  }
1252 
1253  elem = &matroska->level1_elems[matroska->num_level1_elems++];
1254  *elem = (MatroskaLevel1Element){.id = id};
1255 
1256  return elem;
1257 }
1258 
1259 static int ebml_parse(MatroskaDemuxContext *matroska,
1260  EbmlSyntax *syntax, void *data)
1261 {
1262  static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1263  // Forbid unknown-length EBML_NONE elements.
1265  [EBML_UINT] = 8,
1266  [EBML_SINT] = 8,
1267  [EBML_FLOAT] = 8,
1268  // max. 16 MB for strings
1269  [EBML_STR] = 0x1000000,
1270  [EBML_UTF8] = 0x1000000,
1271  // max. 256 MB for binary data
1272  [EBML_BIN] = 0x10000000,
1273  // no limits for anything else
1274  };
1275  AVIOContext *pb = matroska->ctx->pb;
1276  uint32_t id;
1277  uint64_t length;
1278  int64_t pos = avio_tell(pb), pos_alt;
1279  int res, update_pos = 1, level_check;
1280  MatroskaLevel1Element *level1_elem;
1281  MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1282 
1283  if (!matroska->current_id) {
1284  uint64_t id;
1285  res = ebml_read_num(matroska, pb, 4, &id, 0);
1286  if (res < 0) {
1287  if (pb->eof_reached && res == AVERROR_EOF) {
1288  if (matroska->is_live)
1289  // in live mode, finish parsing if EOF is reached.
1290  return 1;
1291  if (level && pos == avio_tell(pb)) {
1292  if (level->length == EBML_UNKNOWN_LENGTH) {
1293  // Unknown-length levels automatically end at EOF.
1294  matroska->num_levels--;
1295  return LEVEL_ENDED;
1296  } else {
1297  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1298  "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1299  }
1300  }
1301  }
1302  return res;
1303  }
1304  matroska->current_id = id | 1 << 7 * res;
1305  pos_alt = pos + res;
1306  } else {
1307  pos_alt = pos;
1308  pos -= (av_log2(matroska->current_id) + 7) / 8;
1309  }
1310 
1311  id = matroska->current_id;
1312 
1313  syntax = ebml_parse_id(syntax, id);
1314  if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1315  if (level && level->length == EBML_UNKNOWN_LENGTH) {
1316  // Unknown-length levels end when an element from an upper level
1317  // in the hierarchy is encountered.
1318  while (syntax->def.n) {
1319  syntax = ebml_parse_id(syntax->def.n, id);
1320  if (syntax->id) {
1321  matroska->num_levels--;
1322  return LEVEL_ENDED;
1323  }
1324  // We have not encountered a known element; syntax is a sentinel.
1325  av_assert1(syntax->type == EBML_NONE);
1326  };
1327  }
1328 
1329  av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1330  "%"PRId64"\n", id, pos);
1331  update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1332  }
1333 
1334  if (data) {
1335  data = (char *) data + syntax->data_offset;
1336  if (syntax->list_elem_size) {
1337  EbmlList *list = data;
1338  void *newelem;
1339 
1340  if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1341  return AVERROR(ENOMEM);
1342  newelem = av_fast_realloc(list->elem,
1343  &list->alloc_elem_size,
1344  (list->nb_elem + 1) * syntax->list_elem_size);
1345  if (!newelem)
1346  return AVERROR(ENOMEM);
1347  list->elem = newelem;
1348  data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1349  memset(data, 0, syntax->list_elem_size);
1350  list->nb_elem++;
1351  }
1352  }
1353 
1354  if (syntax->type != EBML_STOP) {
1355  matroska->current_id = 0;
1356  if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1357  return res;
1358 
1359  pos_alt += res;
1360 
1361  if (matroska->num_levels > 0) {
1362  if (length != EBML_UNKNOWN_LENGTH &&
1363  level->length != EBML_UNKNOWN_LENGTH) {
1364  uint64_t elem_end = pos_alt + length,
1365  level_end = level->start + level->length;
1366 
1367  if (elem_end < level_end) {
1368  level_check = 0;
1369  } else if (elem_end == level_end) {
1370  level_check = LEVEL_ENDED;
1371  } else {
1372  av_log(matroska->ctx, AV_LOG_ERROR,
1373  "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1374  "containing master element ending at 0x%"PRIx64"\n",
1375  pos, elem_end, level_end);
1376  return AVERROR_INVALIDDATA;
1377  }
1378  } else if (length != EBML_UNKNOWN_LENGTH) {
1379  level_check = 0;
1380  } else if (level->length != EBML_UNKNOWN_LENGTH) {
1381  av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1382  "at 0x%"PRIx64" inside parent with finite size\n", pos);
1383  return AVERROR_INVALIDDATA;
1384  } else {
1385  level_check = 0;
1386  if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1387  || syntax->type == EBML_NEST)) {
1388  // According to the current specifications only clusters and
1389  // segments are allowed to be unknown-length. We also accept
1390  // other unknown-length master elements.
1391  av_log(matroska->ctx, AV_LOG_WARNING,
1392  "Found unknown-length element 0x%"PRIX32" other than "
1393  "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1394  "parsing will nevertheless be attempted.\n", id, pos);
1395  update_pos = -1;
1396  }
1397  }
1398  } else
1399  level_check = 0;
1400 
1401  if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1402  if (length != EBML_UNKNOWN_LENGTH) {
1403  av_log(matroska->ctx, AV_LOG_ERROR,
1404  "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1405  "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1406  length, max_lengths[syntax->type], id, pos);
1407  } else if (syntax->type != EBML_NONE) {
1408  av_log(matroska->ctx, AV_LOG_ERROR,
1409  "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1410  "unknown length, yet the length of an element of its "
1411  "type must be known.\n", id, pos);
1412  } else {
1413  av_log(matroska->ctx, AV_LOG_ERROR,
1414  "Found unknown-length element with ID 0x%"PRIX32" at "
1415  "pos. 0x%"PRIx64" for which no syntax for parsing is "
1416  "available.\n", id, pos);
1417  }
1418  return AVERROR_INVALIDDATA;
1419  }
1420 
1421  if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1422  // Loosing sync will likely manifest itself as encountering unknown
1423  // elements which are not reliably distinguishable from elements
1424  // belonging to future extensions of the format.
1425  // We use a heuristic to detect such situations: If the current
1426  // element is not expected at the current syntax level and there
1427  // were only a few unknown elements in a row, then the element is
1428  // skipped or considered defective based upon the length of the
1429  // current element (i.e. how much would be skipped); if there were
1430  // more than a few skipped elements in a row and skipping the current
1431  // element would lead us more than SKIP_THRESHOLD away from the last
1432  // known good position, then it is inferred that an error occurred.
1433  // The dependency on the number of unknown elements in a row exists
1434  // because the distance to the last known good position is
1435  // automatically big if the last parsed element was big.
1436  // In both cases, each unknown element is considered equivalent to
1437  // UNKNOWN_EQUIV of skipped bytes for the check.
1438  // The whole check is only done for non-seekable output, because
1439  // in this situation skipped data can't simply be rechecked later.
1440  // This is especially important when using unkown length elements
1441  // as the check for whether a child exceeds its containing master
1442  // element is not effective in this situation.
1443  if (update_pos) {
1444  matroska->unknown_count = 0;
1445  } else {
1446  int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1447 
1448  if (matroska->unknown_count > 3)
1449  dist += pos_alt - matroska->resync_pos;
1450 
1451  if (dist > SKIP_THRESHOLD) {
1452  av_log(matroska->ctx, AV_LOG_ERROR,
1453  "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1454  "length 0x%"PRIx64" considered as invalid data. Last "
1455  "known good position 0x%"PRIx64", %d unknown elements"
1456  " in a row\n", id, pos, length, matroska->resync_pos,
1457  matroska->unknown_count);
1458  return AVERROR_INVALIDDATA;
1459  }
1460  }
1461  }
1462 
1463  if (update_pos > 0) {
1464  // We have found an element that is allowed at this place
1465  // in the hierarchy and it passed all checks, so treat the beginning
1466  // of the element as the "last known good" position.
1467  matroska->resync_pos = pos;
1468  }
1469 
1470  if (!data && length != EBML_UNKNOWN_LENGTH)
1471  goto skip;
1472  }
1473 
1474  switch (syntax->type) {
1475  case EBML_UINT:
1476  res = ebml_read_uint(pb, length, syntax->def.u, data);
1477  break;
1478  case EBML_SINT:
1479  res = ebml_read_sint(pb, length, syntax->def.i, data);
1480  break;
1481  case EBML_FLOAT:
1482  res = ebml_read_float(pb, length, syntax->def.f, data);
1483  break;
1484  case EBML_STR:
1485  case EBML_UTF8:
1486  res = ebml_read_ascii(pb, length, syntax->def.s, data);
1487  break;
1488  case EBML_BIN:
1489  res = ebml_read_binary(pb, length, pos_alt, data);
1490  break;
1491  case EBML_LEVEL1:
1492  case EBML_NEST:
1493  if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1494  return res;
1495  if (id == MATROSKA_ID_SEGMENT)
1496  matroska->segment_start = pos_alt;
1497  if (id == MATROSKA_ID_CUES)
1498  matroska->cues_parsing_deferred = 0;
1499  if (syntax->type == EBML_LEVEL1 &&
1500  (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1501  if (!level1_elem->pos) {
1502  // Zero is not a valid position for a level 1 element.
1503  level1_elem->pos = pos;
1504  } else if (level1_elem->pos != pos)
1505  av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1506  level1_elem->parsed = 1;
1507  }
1508  if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1509  return res;
1510  break;
1511  case EBML_STOP:
1512  return 1;
1513  skip:
1514  default:
1515  if (length) {
1516  int64_t res2;
1517  if (ffio_limit(pb, length) != length) {
1518  // ffio_limit emits its own error message,
1519  // so we don't have to.
1520  return AVERROR(EIO);
1521  }
1522  if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1523  // avio_skip might take us past EOF. We check for this
1524  // by skipping only length - 1 bytes, reading a byte and
1525  // checking the error flags. This is done in order to check
1526  // that the element has been properly skipped even when
1527  // no filesize (that ffio_limit relies on) is available.
1528  avio_r8(pb);
1529  res = NEEDS_CHECKING;
1530  } else
1531  res = res2;
1532  } else
1533  res = 0;
1534  }
1535  if (res) {
1536  if (res == NEEDS_CHECKING) {
1537  if (pb->eof_reached) {
1538  if (pb->error)
1539  res = pb->error;
1540  else
1541  res = AVERROR_EOF;
1542  } else
1543  goto level_check;
1544  }
1545 
1546  if (res == AVERROR_INVALIDDATA)
1547  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1548  else if (res == AVERROR(EIO))
1549  av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1550  else if (res == AVERROR_EOF) {
1551  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1552  res = AVERROR(EIO);
1553  }
1554 
1555  return res;
1556  }
1557 
1558 level_check:
1559  if (syntax->is_counted && data) {
1560  CountedElement *elem = data;
1561  if (elem->count != UINT_MAX)
1562  elem->count++;
1563  }
1564 
1565  if (level_check == LEVEL_ENDED && matroska->num_levels) {
1566  level = &matroska->levels[matroska->num_levels - 1];
1567  pos = avio_tell(pb);
1568 
1569  // Given that pos >= level->start no check for
1570  // level->length != EBML_UNKNOWN_LENGTH is necessary.
1571  while (matroska->num_levels && pos == level->start + level->length) {
1572  matroska->num_levels--;
1573  level--;
1574  }
1575  }
1576 
1577  return level_check;
1578 }
1579 
1580 static void ebml_free(EbmlSyntax *syntax, void *data)
1581 {
1582  int i, j;
1583  for (i = 0; syntax[i].id; i++) {
1584  void *data_off = (char *) data + syntax[i].data_offset;
1585  switch (syntax[i].type) {
1586  case EBML_STR:
1587  case EBML_UTF8:
1588  av_freep(data_off);
1589  break;
1590  case EBML_BIN:
1591  av_buffer_unref(&((EbmlBin *) data_off)->buf);
1592  break;
1593  case EBML_LEVEL1:
1594  case EBML_NEST:
1595  if (syntax[i].list_elem_size) {
1596  EbmlList *list = data_off;
1597  char *ptr = list->elem;
1598  for (j = 0; j < list->nb_elem;
1599  j++, ptr += syntax[i].list_elem_size)
1600  ebml_free(syntax[i].def.n, ptr);
1601  av_freep(&list->elem);
1602  list->nb_elem = 0;
1603  list->alloc_elem_size = 0;
1604  } else
1605  ebml_free(syntax[i].def.n, data_off);
1606  default:
1607  break;
1608  }
1609  }
1610 }
1611 
1612 /*
1613  * Autodetecting...
1614  */
1615 static int matroska_probe(const AVProbeData *p)
1616 {
1617  uint64_t total = 0;
1618  int len_mask = 0x80, size = 1, n = 1, i;
1619 
1620  /* EBML header? */
1621  if (AV_RB32(p->buf) != EBML_ID_HEADER)
1622  return 0;
1623 
1624  /* length of header */
1625  total = p->buf[4];
1626  while (size <= 8 && !(total & len_mask)) {
1627  size++;
1628  len_mask >>= 1;
1629  }
1630  if (size > 8)
1631  return 0;
1632  total &= (len_mask - 1);
1633  while (n < size)
1634  total = (total << 8) | p->buf[4 + n++];
1635 
1636  if (total + 1 == 1ULL << (7 * size)){
1637  /* Unknown-length header - simply parse the whole buffer. */
1638  total = p->buf_size - 4 - size;
1639  } else {
1640  /* Does the probe data contain the whole header? */
1641  if (p->buf_size < 4 + size + total)
1642  return 0;
1643  }
1644 
1645  /* The header should contain a known document type. For now,
1646  * we don't parse the whole header but simply check for the
1647  * availability of that array of characters inside the header.
1648  * Not fully fool-proof, but good enough. */
1649  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1650  size_t probelen = strlen(matroska_doctypes[i]);
1651  if (total < probelen)
1652  continue;
1653  for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1654  if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1655  return AVPROBE_SCORE_MAX;
1656  }
1657 
1658  // probably valid EBML header but no recognized doctype
1659  return AVPROBE_SCORE_EXTENSION;
1660 }
1661 
1663  uint64_t num)
1664 {
1665  MatroskaTrack *tracks = matroska->tracks.elem;
1666  int i;
1667 
1668  for (i = 0; i < matroska->tracks.nb_elem; i++)
1669  if (tracks[i].num == num)
1670  return &tracks[i];
1671 
1672  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1673  return NULL;
1674 }
1675 
1676 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1677  MatroskaTrack *track)
1678 {
1679  MatroskaTrackEncoding *encodings = track->encodings.elem;
1680  uint8_t *data = *buf;
1681  int isize = *buf_size;
1682  uint8_t *pkt_data = NULL;
1683  uint8_t av_unused *newpktdata;
1684  int pkt_size = isize;
1685  int result = 0;
1686  int olen;
1687 
1688  if (pkt_size >= 10000000U)
1689  return AVERROR_INVALIDDATA;
1690 
1691  switch (encodings[0].compression.algo) {
1693  {
1694  int header_size = encodings[0].compression.settings.size;
1695  uint8_t *header = encodings[0].compression.settings.data;
1696 
1697  if (header_size && !header) {
1698  av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1699  return -1;
1700  }
1701 
1702  if (!header_size)
1703  return 0;
1704 
1705  pkt_size = isize + header_size;
1706  pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1707  if (!pkt_data)
1708  return AVERROR(ENOMEM);
1709 
1710  memcpy(pkt_data, header, header_size);
1711  memcpy(pkt_data + header_size, data, isize);
1712  break;
1713  }
1715  do {
1716  int insize = isize;
1717  olen = pkt_size *= 3;
1718  newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1720  if (!newpktdata) {
1721  result = AVERROR(ENOMEM);
1722  goto failed;
1723  }
1724  pkt_data = newpktdata;
1725  result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1726  } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1727  if (result) {
1729  goto failed;
1730  }
1731  pkt_size -= olen;
1732  break;
1733 #if CONFIG_ZLIB
1735  {
1736  z_stream zstream = { 0 };
1737  if (!pkt_size || inflateInit(&zstream) != Z_OK)
1738  return -1;
1739  zstream.next_in = data;
1740  zstream.avail_in = isize;
1741  do {
1742  pkt_size *= 3;
1743  newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1744  if (!newpktdata) {
1745  inflateEnd(&zstream);
1746  result = AVERROR(ENOMEM);
1747  goto failed;
1748  }
1749  pkt_data = newpktdata;
1750  zstream.avail_out = pkt_size - zstream.total_out;
1751  zstream.next_out = pkt_data + zstream.total_out;
1752  result = inflate(&zstream, Z_NO_FLUSH);
1753  } while (result == Z_OK && pkt_size < 10000000);
1754  pkt_size = zstream.total_out;
1755  inflateEnd(&zstream);
1756  if (result != Z_STREAM_END) {
1757  if (result == Z_MEM_ERROR)
1758  result = AVERROR(ENOMEM);
1759  else
1761  goto failed;
1762  }
1763  break;
1764  }
1765 #endif
1766 #if CONFIG_BZLIB
1768  {
1769  bz_stream bzstream = { 0 };
1770  if (!pkt_size || BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1771  return -1;
1772  bzstream.next_in = data;
1773  bzstream.avail_in = isize;
1774  do {
1775  pkt_size *= 3;
1776  newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1777  if (!newpktdata) {
1778  BZ2_bzDecompressEnd(&bzstream);
1779  result = AVERROR(ENOMEM);
1780  goto failed;
1781  }
1782  pkt_data = newpktdata;
1783  bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1784  bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1785  result = BZ2_bzDecompress(&bzstream);
1786  } while (result == BZ_OK && pkt_size < 10000000);
1787  pkt_size = bzstream.total_out_lo32;
1788  BZ2_bzDecompressEnd(&bzstream);
1789  if (result != BZ_STREAM_END) {
1790  if (result == BZ_MEM_ERROR)
1791  result = AVERROR(ENOMEM);
1792  else
1794  goto failed;
1795  }
1796  break;
1797  }
1798 #endif
1799  default:
1800  return AVERROR_INVALIDDATA;
1801  }
1802 
1803  memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1804 
1805  *buf = pkt_data;
1806  *buf_size = pkt_size;
1807  return 0;
1809 failed:
1810  av_free(pkt_data);
1811  return result;
1812 }
1813 
1815  AVDictionary **metadata, char *prefix)
1816 {
1817  MatroskaTag *tags = list->elem;
1818  char key[1024];
1819  int i;
1820 
1821  for (i = 0; i < list->nb_elem; i++) {
1822  const char *lang = tags[i].lang &&
1823  strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1824 
1825  if (!tags[i].name) {
1826  av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1827  continue;
1828  }
1829  if (prefix)
1830  snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1831  else
1832  av_strlcpy(key, tags[i].name, sizeof(key));
1833  if (tags[i].def || !lang) {
1834  av_dict_set(metadata, key, tags[i].string, 0);
1835  if (tags[i].sub.nb_elem)
1836  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1837  }
1838  if (lang) {
1839  av_strlcat(key, "-", sizeof(key));
1840  av_strlcat(key, lang, sizeof(key));
1841  av_dict_set(metadata, key, tags[i].string, 0);
1842  if (tags[i].sub.nb_elem)
1843  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1844  }
1845  }
1847 }
1848 
1850 {
1851  MatroskaDemuxContext *matroska = s->priv_data;
1852  MatroskaTags *tags = matroska->tags.elem;
1853  int i, j;
1854 
1855  for (i = 0; i < matroska->tags.nb_elem; i++) {
1856  if (tags[i].target.attachuid) {
1857  MatroskaAttachment *attachment = matroska->attachments.elem;
1858  int found = 0;
1859  for (j = 0; j < matroska->attachments.nb_elem; j++) {
1860  if (attachment[j].uid == tags[i].target.attachuid &&
1861  attachment[j].stream) {
1862  matroska_convert_tag(s, &tags[i].tag,
1863  &attachment[j].stream->metadata, NULL);
1864  found = 1;
1865  }
1866  }
1867  if (!found) {
1869  "The tags at index %d refer to a "
1870  "non-existent attachment %"PRId64".\n",
1871  i, tags[i].target.attachuid);
1872  }
1873  } else if (tags[i].target.chapteruid) {
1874  MatroskaChapter *chapter = matroska->chapters.elem;
1875  int found = 0;
1876  for (j = 0; j < matroska->chapters.nb_elem; j++) {
1877  if (chapter[j].uid == tags[i].target.chapteruid &&
1878  chapter[j].chapter) {
1879  matroska_convert_tag(s, &tags[i].tag,
1880  &chapter[j].chapter->metadata, NULL);
1881  found = 1;
1882  }
1883  }
1884  if (!found) {
1886  "The tags at index %d refer to a non-existent chapter "
1887  "%"PRId64".\n",
1888  i, tags[i].target.chapteruid);
1889  }
1890  } else if (tags[i].target.trackuid) {
1891  MatroskaTrack *track = matroska->tracks.elem;
1892  int found = 0;
1893  for (j = 0; j < matroska->tracks.nb_elem; j++) {
1894  if (track[j].uid == tags[i].target.trackuid &&
1895  track[j].stream) {
1896  matroska_convert_tag(s, &tags[i].tag,
1897  &track[j].stream->metadata, NULL);
1898  found = 1;
1899  }
1900  }
1901  if (!found) {
1903  "The tags at index %d refer to a non-existent track "
1904  "%"PRId64".\n",
1905  i, tags[i].target.trackuid);
1906  }
1907  } else {
1908  matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1909  tags[i].target.type);
1910  }
1911  }
1912 }
1913 
1915  int64_t pos)
1916 {
1917  uint32_t saved_id = matroska->current_id;
1918  int64_t before_pos = avio_tell(matroska->ctx->pb);
1919  int ret = 0;
1920  int ret2;
1921 
1922  /* seek */
1923  if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1924  /* We don't want to lose our seekhead level, so we add
1925  * a dummy. This is a crude hack. */
1926  if (matroska->num_levels == EBML_MAX_DEPTH) {
1927  av_log(matroska->ctx, AV_LOG_INFO,
1928  "Max EBML element depth (%d) reached, "
1929  "cannot parse further.\n", EBML_MAX_DEPTH);
1931  } else {
1932  matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1933  matroska->num_levels++;
1934  matroska->current_id = 0;
1935 
1936  ret = ebml_parse(matroska, matroska_segment, matroska);
1937  if (ret == LEVEL_ENDED) {
1938  /* This can only happen if the seek brought us beyond EOF. */
1939  ret = AVERROR_EOF;
1940  }
1941  }
1942  }
1943  /* Seek back - notice that in all instances where this is used
1944  * it is safe to set the level to 1. */
1945  ret2 = matroska_reset_status(matroska, saved_id, before_pos);
1946  if (ret >= 0)
1947  ret = ret2;
1948 
1949  return ret;
1950 }
1951 
1952 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1953 {
1954  EbmlList *seekhead_list = &matroska->seekhead;
1955  int i;
1956 
1957  // we should not do any seeking in the streaming case
1958  if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1959  return;
1960 
1961  for (i = 0; i < seekhead_list->nb_elem; i++) {
1962  MatroskaSeekhead *seekheads = seekhead_list->elem;
1963  uint32_t id = seekheads[i].id;
1964  int64_t pos = seekheads[i].pos + matroska->segment_start;
1965  MatroskaLevel1Element *elem;
1966 
1967  if (id != seekheads[i].id || pos < matroska->segment_start)
1968  continue;
1969 
1970  elem = matroska_find_level1_elem(matroska, id, pos);
1971  if (!elem || elem->parsed)
1972  continue;
1973 
1974  elem->pos = pos;
1975 
1976  // defer cues parsing until we actually need cue data.
1977  if (id == MATROSKA_ID_CUES)
1978  continue;
1979 
1980  if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1981  // mark index as broken
1982  matroska->cues_parsing_deferred = -1;
1983  break;
1984  }
1985 
1986  elem->parsed = 1;
1987  }
1988 }
1989 
1990 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1991 {
1992  EbmlList *index_list;
1994  uint64_t index_scale = 1;
1995  int i, j;
1996 
1997  if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1998  return;
1999 
2000  index_list = &matroska->index;
2001  index = index_list->elem;
2002  if (index_list->nb_elem < 2)
2003  return;
2004  if (index[1].time > 1E14 / matroska->time_scale) {
2005  av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
2006  return;
2007  }
2008  for (i = 0; i < index_list->nb_elem; i++) {
2009  EbmlList *pos_list = &index[i].pos;
2010  MatroskaIndexPos *pos = pos_list->elem;
2011  for (j = 0; j < pos_list->nb_elem; j++) {
2012  MatroskaTrack *track = matroska_find_track_by_num(matroska,
2013  pos[j].track);
2014  if (track && track->stream)
2015  av_add_index_entry(track->stream,
2016  pos[j].pos + matroska->segment_start,
2017  index[i].time / index_scale, 0, 0,
2019  }
2020  }
2021 }
2022 
2023 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
2024  int i;
2025 
2026  if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
2027  return;
2028 
2029  for (i = 0; i < matroska->num_level1_elems; i++) {
2030  MatroskaLevel1Element *elem = &matroska->level1_elems[i];
2031  if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
2032  if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
2033  matroska->cues_parsing_deferred = -1;
2034  elem->parsed = 1;
2035  break;
2036  }
2037  }
2038 
2039  matroska_add_index_entries(matroska);
2040 }
2041 
2043  unsigned nb_encodings,
2044  MatroskaTrack *track,
2045  char **key_id_base64, void *logctx)
2046 {
2047  if (nb_encodings > 1) {
2048  av_log(logctx, AV_LOG_ERROR,
2049  "Multiple combined encodings not supported\n");
2050  return 0;
2051  }
2052  if (!nb_encodings)
2053  return 0;
2054  if (encodings->type) {
2055  if (encodings->encryption.key_id.size > 0) {
2056  /* Save the encryption key id to be stored later
2057  * as a metadata tag. */
2058  const int b64_size = AV_BASE64_SIZE(encodings->encryption.key_id.size);
2059  *key_id_base64 = av_malloc(b64_size);
2060  if (!*key_id_base64)
2061  return AVERROR(ENOMEM);
2062 
2063  av_base64_encode(*key_id_base64, b64_size,
2064  encodings->encryption.key_id.data,
2065  encodings->encryption.key_id.size);
2066  } else {
2067  encodings->scope = 0;
2068  av_log(logctx, AV_LOG_ERROR, "Unsupported encoding type\n");
2069  }
2070  } else if (
2071 #if CONFIG_ZLIB
2073 #endif
2074 #if CONFIG_BZLIB
2076 #endif
2079  encodings->scope = 0;
2080  av_log(logctx, AV_LOG_ERROR, "Unsupported encoding type\n");
2081  } else if (track->codec_priv.size && encodings[0].scope & 2) {
2082  uint8_t *codec_priv = track->codec_priv.data;
2083  int ret = matroska_decode_buffer(&track->codec_priv.data,
2084  &track->codec_priv.size,
2085  track);
2086  if (ret < 0) {
2087  track->codec_priv.data = NULL;
2088  track->codec_priv.size = 0;
2089  av_log(logctx, AV_LOG_ERROR,
2090  "Failed to decode codec private data\n");
2091  }
2092 
2093  if (codec_priv != track->codec_priv.data) {
2094  av_buffer_unref(&track->codec_priv.buf);
2095  if (track->codec_priv.data) {
2096  track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2098  NULL, NULL, 0);
2099  if (!track->codec_priv.buf) {
2100  av_freep(&track->codec_priv.data);
2101  track->codec_priv.size = 0;
2102  return AVERROR(ENOMEM);
2103  }
2104  }
2105  }
2106  }
2107  track->needs_decoding = !encodings->type &&
2108  encodings->scope & 1 &&
2109  (encodings->compression.algo !=
2111  encodings->compression.settings.size);
2112 
2113  return 0;
2114 }
2115 
2116 static int matroska_aac_profile(char *codec_id)
2117 {
2118  static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
2119  int profile;
2120 
2122  if (strstr(codec_id, aac_profiles[profile]))
2123  break;
2124  return profile + 1;
2125 }
2126 
2127 static int matroska_aac_sri(int samplerate)
2128 {
2129  int sri;
2130 
2131  for (sri = 0; sri < FF_ARRAY_ELEMS(ff_mpeg4audio_sample_rates); sri++)
2132  if (ff_mpeg4audio_sample_rates[sri] == samplerate)
2133  break;
2134  return sri;
2135 }
2136 
2137 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
2138 {
2139  /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
2140  avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
2141 }
2142 
2144  MatroskaTrack *track,
2145  int *offset)
2146 {
2147  AVStream *st = track->stream;
2148  uint8_t *p = track->codec_priv.data;
2149  int size = track->codec_priv.size;
2150 
2151  if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
2152  av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
2153  track->codec_priv.size = 0;
2154  return 0;
2155  }
2156  *offset = 8;
2157  track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
2158 
2159  p += track->codec_priv.size;
2160  size -= track->codec_priv.size;
2161 
2162  /* parse the remaining metadata blocks if present */
2163  while (size >= 4) {
2164  int block_last, block_type, block_size;
2165 
2166  flac_parse_block_header(p, &block_last, &block_type, &block_size);
2167 
2168  p += 4;
2169  size -= 4;
2170  if (block_size > size)
2171  return 0;
2172 
2173  /* check for the channel mask */
2174  if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2175  AVDictionary *dict = NULL;
2176  AVDictionaryEntry *chmask;
2177 
2178  ff_vorbis_comment(s, &dict, p, block_size, 0);
2179  chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2180  if (chmask) {
2181  uint64_t mask = strtol(chmask->value, NULL, 0);
2182  if (!mask || mask & ~0x3ffffULL) {
2184  "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2185  } else
2187  }
2188  av_dict_free(&dict);
2189  }
2190 
2191  p += block_size;
2192  size -= block_size;
2193  }
2194 
2195  return 0;
2196 }
2197 
2198 static int mkv_field_order(const MatroskaDemuxContext *matroska, uint64_t field_order)
2199 {
2200  int minor, micro, bttb = 0;
2201 
2202  /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2203  * this function, and fixed in 57.52 */
2204  if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf57.%d.%d", &minor, &micro) == 2)
2205  bttb = (minor >= 36 && minor <= 51 && micro >= 100);
2206 
2207  switch (field_order) {
2209  return AV_FIELD_PROGRESSIVE;
2211  return AV_FIELD_UNKNOWN;
2213  return AV_FIELD_TT;
2215  return AV_FIELD_BB;
2217  return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2219  return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2220  default:
2221  return AV_FIELD_UNKNOWN;
2222  }
2223 }
2224 
2225 static void mkv_stereo_mode_display_mul(int stereo_mode,
2226  int *h_width, int *h_height)
2227 {
2228  switch (stereo_mode) {
2234  break;
2239  *h_width = 2;
2240  break;
2245  *h_height = 2;
2246  break;
2247  }
2248 }
2249 
2250 static int mkv_stereo3d_conv(AVStream *st, MatroskaVideoStereoModeType stereo_mode)
2251 {
2252  static const struct {
2253  char type;
2254  char flags;
2255  } stereo_mode_conv [] = {
2256 #define STEREO_MODE_CONV(STEREOMODETYPE, STEREO3DTYPE, FLAGS, WDIV, HDIV, WEBM) \
2257  [(STEREOMODETYPE)] = { .type = (STEREO3DTYPE), .flags = (FLAGS) },
2258 #define NOTHING(STEREOMODETYPE, WDIV, HDIV, WEBM)
2260  };
2261  AVStereo3D *stereo;
2262  size_t size;
2263 
2264  stereo = av_stereo3d_alloc_size(&size);
2265  if (!stereo)
2266  return AVERROR(ENOMEM);
2267 
2268  stereo->type = stereo_mode_conv[stereo_mode].type;
2269  stereo->flags = stereo_mode_conv[stereo_mode].flags;
2270 
2272  AV_PKT_DATA_STEREO3D, stereo, size, 0)) {
2273  av_freep(&stereo);
2274  return AVERROR(ENOMEM);
2275  }
2276 
2277  return 0;
2278 }
2279 
2280 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2281  const MatroskaTrackVideoColor *color = track->video.color.elem;
2282  const MatroskaMasteringMeta *mastering_meta;
2283  int has_mastering_primaries, has_mastering_luminance;
2284 
2285  if (!track->video.color.nb_elem)
2286  return 0;
2287 
2288  mastering_meta = &color->mastering_meta;
2289  // Mastering primaries are CIE 1931 coords, and must be > 0.
2290  has_mastering_primaries =
2291  mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2292  mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2293  mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2294  mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2295  has_mastering_luminance = mastering_meta->max_luminance >
2296  mastering_meta->min_luminance.el.f &&
2297  mastering_meta->min_luminance.el.f >= 0 &&
2298  mastering_meta->min_luminance.count;
2299 
2300  if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2301  st->codecpar->color_space = color->matrix_coefficients;
2302  if (color->primaries != AVCOL_PRI_RESERVED &&
2303  color->primaries != AVCOL_PRI_RESERVED0)
2304  st->codecpar->color_primaries = color->primaries;
2305  if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2306  color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2307  st->codecpar->color_trc = color->transfer_characteristics;
2308  if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2309  color->range <= AVCOL_RANGE_JPEG)
2310  st->codecpar->color_range = color->range;
2311  if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2312  color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2313  color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2314  color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2315  st->codecpar->chroma_location =
2316  av_chroma_location_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2317  (color->chroma_siting_vert - 1) << 7);
2318  }
2319  if (color->max_cll && color->max_fall) {
2320  size_t size = 0;
2322  if (!metadata)
2323  return AVERROR(ENOMEM);
2325  AV_PKT_DATA_CONTENT_LIGHT_LEVEL, metadata, size, 0)) {
2326  av_freep(&metadata);
2327  return AVERROR(ENOMEM);
2328  }
2329  metadata->MaxCLL = color->max_cll;
2330  metadata->MaxFALL = color->max_fall;
2331  }
2332 
2333  if (has_mastering_primaries || has_mastering_luminance) {
2334  size_t size = 0;
2336  if (!metadata)
2337  return AVERROR(ENOMEM);
2340  av_freep(&metadata);
2341  return AVERROR(ENOMEM);
2342  }
2343  if (has_mastering_primaries) {
2344  metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2345  metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2346  metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2347  metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2348  metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2349  metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2350  metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2351  metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2352  metadata->has_primaries = 1;
2353  }
2354  if (has_mastering_luminance) {
2355  metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2356  metadata->min_luminance = av_d2q(mastering_meta->min_luminance.el.f, INT_MAX);
2357  metadata->has_luminance = 1;
2358  }
2359  }
2360  return 0;
2361 }
2362 
2363 static int mkv_create_display_matrix(AVStream *st,
2364  const MatroskaTrackVideoProjection *proj,
2365  void *logctx)
2366 {
2367  AVPacketSideData *sd;
2368  double pitch = proj->pitch, yaw = proj->yaw, roll = proj->roll;
2369  int32_t *matrix;
2370  int hflip;
2371 
2372  if (pitch == 0.0 && yaw == 0.0 && roll == 0.0)
2373  return 0;
2374 
2375  /* Note: The following constants are exactly representable
2376  * as floating-point numbers. */
2377  if (pitch != 0.0 || (yaw != 0.0 && yaw != 180.0 && yaw != -180.0) ||
2378  isnan(roll)) {
2379  av_log(logctx, AV_LOG_WARNING, "Ignoring non-2D rectangular "
2380  "projection in stream %u (yaw %f, pitch %f, roll %f)\n",
2381  st->index, yaw, pitch, roll);
2382  return 0;
2383  }
2387  9 * sizeof(*matrix), 0);
2388  if (!sd)
2389  return AVERROR(ENOMEM);
2390  matrix = (int32_t*)sd->data;
2391 
2392  hflip = yaw != 0.0;
2393  /* ProjectionPoseRoll is in the counter-clockwise direction
2394  * whereas av_display_rotation_set() expects its argument
2395  * to be oriented clockwise, so we need to negate roll.
2396  * Furthermore, if hflip is set, we need to negate it again
2397  * to account for the fact that the Matroska specifications
2398  * require the yaw rotation to be applied first. */
2399  av_display_rotation_set(matrix, roll * (2 * hflip - 1));
2400  av_display_matrix_flip(matrix, hflip, 0);
2401 
2402  return 0;
2403 }
2404 
2405 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2406  void *logctx)
2407 {
2408  AVSphericalMapping *spherical;
2409  const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2410  const uint8_t *priv_data = mkv_projection->private.data;
2411  enum AVSphericalProjection projection;
2412  size_t spherical_size;
2413  uint32_t l = 0, t = 0, r = 0, b = 0;
2414  uint32_t padding = 0;
2415 
2416  if (mkv_projection->private.size && priv_data[0] != 0) {
2417  av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2418  return 0;
2419  }
2420 
2421  switch (track->video.projection.type) {
2423  return mkv_create_display_matrix(st, mkv_projection, logctx);
2425  if (track->video.projection.private.size == 20) {
2426  t = AV_RB32(priv_data + 4);
2427  b = AV_RB32(priv_data + 8);
2428  l = AV_RB32(priv_data + 12);
2429  r = AV_RB32(priv_data + 16);
2430 
2431  if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2432  av_log(logctx, AV_LOG_ERROR,
2433  "Invalid bounding rectangle coordinates "
2434  "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2435  l, t, r, b);
2436  return AVERROR_INVALIDDATA;
2437  }
2438  } else if (track->video.projection.private.size != 0) {
2439  av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2440  return AVERROR_INVALIDDATA;
2441  }
2442 
2443  if (l || t || r || b)
2444  projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2445  else
2446  projection = AV_SPHERICAL_EQUIRECTANGULAR;
2447  break;
2449  if (track->video.projection.private.size < 4) {
2450  av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2451  return AVERROR_INVALIDDATA;
2452  } else if (track->video.projection.private.size == 12) {
2453  uint32_t layout = AV_RB32(priv_data + 4);
2454  if (layout) {
2455  av_log(logctx, AV_LOG_WARNING,
2456  "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2457  return 0;
2458  }
2459  projection = AV_SPHERICAL_CUBEMAP;
2460  padding = AV_RB32(priv_data + 8);
2461  } else {
2462  av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2463  return AVERROR_INVALIDDATA;
2464  }
2465  break;
2466  default:
2467  av_log(logctx, AV_LOG_WARNING,
2468  "Unknown spherical metadata type %"PRIu64"\n",
2469  track->video.projection.type);
2470  return 0;
2471  }
2472 
2473  spherical = av_spherical_alloc(&spherical_size);
2474  if (!spherical)
2475  return AVERROR(ENOMEM);
2476 
2477  spherical->projection = projection;
2478 
2479  spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2480  spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2481  spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2482 
2483  spherical->padding = padding;
2484 
2485  spherical->bound_left = l;
2486  spherical->bound_top = t;
2487  spherical->bound_right = r;
2488  spherical->bound_bottom = b;
2489 
2491  AV_PKT_DATA_SPHERICAL, spherical, spherical_size, 0)) {
2492  av_freep(&spherical);
2493  return AVERROR(ENOMEM);
2494  }
2495 
2496  return 0;
2497 }
2498 
2500  EbmlBin *bin)
2501 {
2502  return ff_isom_parse_dvcc_dvvc(s, st, bin->data, bin->size);
2503 }
2504 
2506 {
2507  const EbmlList *mappings_list = &track->block_addition_mappings;
2508  MatroskaBlockAdditionMapping *mappings = mappings_list->elem;
2509  int ret;
2510 
2511  for (int i = 0; i < mappings_list->nb_elem; i++) {
2512  MatroskaBlockAdditionMapping *mapping = &mappings[i];
2513  uint64_t type = mapping->type;
2514 
2515  switch (mapping->type) {
2518  "Explicit block Addition Mapping type \"Use BlockAddIDValue\", value %"PRIu64","
2519  " name \"%s\" found.\n", mapping->value, mapping->name ? mapping->name : "");
2521  // fall-through
2524  if (mapping->value != type) {
2525  int strict = s->strict_std_compliance >= FF_COMPLIANCE_STRICT;
2526  av_log(s, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
2527  "Invalid Block Addition Value 0x%"PRIx64" for Block Addition Mapping Type "
2528  "0x%"PRIx64", name \"%s\"\n", mapping->value, mapping->type,
2529  mapping->name ? mapping->name : "");
2530  if (strict)
2531  return AVERROR_INVALIDDATA;
2532  }
2533  break;
2536  if ((ret = mkv_parse_dvcc_dvvc(s, st, track, &mapping->extradata)) < 0)
2537  return ret;
2538 
2539  break;
2540  default:
2542  "Unknown Block Addition Mapping type 0x%"PRIx64", value %"PRIu64", name \"%s\"\n",
2543  mapping->type, mapping->value, mapping->name ? mapping->name : "");
2544  if (mapping->value < 2) {
2545  int strict = s->strict_std_compliance >= FF_COMPLIANCE_STRICT;
2546  av_log(s, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
2547  "Invalid Block Addition value 0x%"PRIu64" for unknown Block Addition Mapping "
2548  "type %"PRIx64", name \"%s\"\n", mapping->value, mapping->type,
2549  mapping->name ? mapping->name : "");
2550  if (strict)
2551  return AVERROR_INVALIDDATA;
2552  }
2553  break;
2554  }
2555  }
2556 
2557  return 0;
2558 }
2559 
2560 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2561 {
2562  const AVCodecTag *codec_tags;
2563 
2564  codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2566 
2567  /* Normalize noncompliant private data that starts with the fourcc
2568  * by expanding/shifting the data by 4 bytes and storing the data
2569  * size at the start. */
2570  if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2571  int ret = av_buffer_realloc(&track->codec_priv.buf,
2573  if (ret < 0)
2574  return ret;
2575 
2576  track->codec_priv.data = track->codec_priv.buf->data;
2577  memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2578  track->codec_priv.size += 4;
2579  AV_WB32(track->codec_priv.data, track->codec_priv.size);
2580  }
2581 
2582  *fourcc = AV_RL32(track->codec_priv.data + 4);
2583  *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2584 
2585  return 0;
2587 
2588 /* An enum with potential return values of the functions for parsing a track.
2589  * Apart from that all these functions can also indicate ordinary errors via
2590  * negative return values. */
2591 enum {
2592  SKIP_TRACK = 1,
2593 };
2594 
2595 #define AAC_MAX_EXTRADATA_SIZE 5
2596 #define TTA_EXTRADATA_SIZE 22
2597 #define WAVPACK_EXTRADATA_SIZE 2
2598 /* Performs the codec-specific part of parsing an audio track. */
2599 static int mka_parse_audio_codec(MatroskaTrack *track, AVCodecParameters *par,
2600  const MatroskaDemuxContext *matroska,
2601  AVFormatContext *s, int *extradata_offset)
2602 {
2603  uint8_t extradata[FFMAX3(AAC_MAX_EXTRADATA_SIZE,
2606  int extradata_size = 0; // > 0 means that the extradata buffer is used
2607  int ret;
2608 
2609  if (!strcmp(track->codec_id, "A_MS/ACM") &&
2610  track->codec_priv.size >= 14) {
2611  FFIOContext b;
2613  track->codec_priv.size);
2614  ret = ff_get_wav_header(s, &b.pub, par,
2615  track->codec_priv.size, 0);
2616  if (ret < 0)
2617  return ret;
2618  *extradata_offset = FFMIN(track->codec_priv.size, 18);
2619  return 0;
2620  } else if (!strcmp(track->codec_id, "A_QUICKTIME") &&
2621  /* Normally 36, but allow noncompliant private data */
2622  track->codec_priv.size >= 32) {
2623  enum AVCodecID codec_id;
2624  uint32_t fourcc;
2625  uint16_t sample_size;
2626 
2627  ret = get_qt_codec(track, &fourcc, &codec_id);
2628  if (ret < 0)
2629  return ret;
2630  sample_size = AV_RB16(track->codec_priv.data + 26);
2631  if (fourcc == 0) {
2632  if (sample_size == 8) {
2633  fourcc = MKTAG('r','a','w',' ');
2635  } else if (sample_size == 16) {
2636  fourcc = MKTAG('t','w','o','s');
2638  }
2639  }
2640  if ((fourcc == MKTAG('t','w','o','s') ||
2641  fourcc == MKTAG('s','o','w','t')) && sample_size == 8)
2643  par->codec_id = codec_id;
2644  par->codec_tag = fourcc;
2645  return 0;
2646  }
2647 
2648  switch (par->codec_id) {
2649  case AV_CODEC_ID_PCM_S16BE:
2650  switch (track->audio.bitdepth) {
2651  case 8:
2653  break;
2654  case 24:
2656  break;
2657  case 32:
2659  break;
2660  }
2661  break;
2662  case AV_CODEC_ID_PCM_S16LE:
2663  switch (track->audio.bitdepth) {
2664  case 8:
2666  break;
2667  case 24:
2669  break;
2670  case 32:
2672  break;
2673  }
2674  break;
2675  case AV_CODEC_ID_PCM_F32LE:
2676  if (track->audio.bitdepth == 64)
2678  break;
2679  case AV_CODEC_ID_AAC:
2680  if (!track->codec_priv.size) {
2681  int profile = matroska_aac_profile(track->codec_id);
2682  int sri = matroska_aac_sri(track->audio.samplerate);
2683 
2684  extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2685  extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2686  if (strstr(track->codec_id, "SBR")) {
2687  sri = matroska_aac_sri(track->audio.out_samplerate);
2688  extradata[2] = 0x56;
2689  extradata[3] = 0xE5;
2690  extradata[4] = 0x80 | (sri << 3);
2691  extradata_size = 5;
2692  } else
2693  extradata_size = 2;
2694  }
2695  break;
2696  case AV_CODEC_ID_ALAC:
2697  if (track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2698  /* Only ALAC's magic cookie is stored in Matroska's track headers.
2699  * Create the "atom size", "tag", and "tag version" fields the
2700  * decoder expects manually. */
2701  ret = ff_alloc_extradata(par, 12 + track->codec_priv.size);
2702  if (ret < 0)
2703  return ret;
2704  AV_WB32(par->extradata, par->extradata_size);
2705  AV_WB32(&par->extradata[4], MKBETAG('a', 'l', 'a', 'c'));
2706  AV_WB32(&par->extradata[8], 0);
2707  memcpy(&par->extradata[12], track->codec_priv.data,
2708  track->codec_priv.size);
2709  }
2710  break;
2711  case AV_CODEC_ID_TTA:
2712  {
2713  uint8_t *ptr;
2714  if (track->audio.channels > UINT16_MAX ||
2715  track->audio.bitdepth > UINT16_MAX) {
2716  av_log(matroska->ctx, AV_LOG_WARNING,
2717  "Too large audio channel number %"PRIu64
2718  " or bitdepth %"PRIu64". Skipping track.\n",
2719  track->audio.channels, track->audio.bitdepth);
2720  if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2721  return AVERROR_INVALIDDATA;
2722  else
2723  return SKIP_TRACK;
2724  }
2725  if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2726  return AVERROR_INVALIDDATA;
2727  extradata_size = TTA_EXTRADATA_SIZE;
2728  ptr = extradata;
2729  bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2730  bytestream_put_le16(&ptr, 1);
2731  bytestream_put_le16(&ptr, track->audio.channels);
2732  bytestream_put_le16(&ptr, track->audio.bitdepth);
2733  bytestream_put_le32(&ptr, track->audio.out_samplerate);
2734  bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2735  track->audio.out_samplerate,
2736  AV_TIME_BASE * 1000));
2737  break;
2738  }
2739  case AV_CODEC_ID_RA_144:
2740  track->audio.out_samplerate = 8000;
2741  track->audio.channels = 1;
2742  break;
2743  case AV_CODEC_ID_RA_288:
2744  case AV_CODEC_ID_COOK:
2745  case AV_CODEC_ID_ATRAC3:
2746  case AV_CODEC_ID_SIPR:
2747  {
2748  const uint8_t *ptr = track->codec_priv.data;
2749  int flavor;
2750 
2751  if (!track->codec_priv.size)
2752  break;
2753 
2754  if (track->codec_priv.size < 46)
2755  return AVERROR_INVALIDDATA;
2756  ptr += 22;
2757  flavor = bytestream_get_be16(&ptr);
2758  track->audio.coded_framesize = bytestream_get_be32(&ptr);
2759  ptr += 12;
2760  track->audio.sub_packet_h = bytestream_get_be16(&ptr);
2761  track->audio.frame_size = bytestream_get_be16(&ptr);
2762  track->audio.sub_packet_size = bytestream_get_be16(&ptr);
2763  if (track->audio.coded_framesize <= 0 ||
2764  track->audio.sub_packet_h <= 0 ||
2765  track->audio.frame_size <= 0)
2766  return AVERROR_INVALIDDATA;
2767 
2768  if (par->codec_id == AV_CODEC_ID_RA_288) {
2769  if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2770  != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2771  return AVERROR_INVALIDDATA;
2772  par->block_align = track->audio.coded_framesize;
2773  track->codec_priv.size = 0;
2774  } else {
2775  if (par->codec_id == AV_CODEC_ID_SIPR) {
2776  static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2777  if (flavor > 3)
2778  return AVERROR_INVALIDDATA;
2779  track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2780  par->bit_rate = sipr_bit_rate[flavor];
2781  } else if (track->audio.sub_packet_size <= 0 ||
2782  track->audio.frame_size % track->audio.sub_packet_size)
2783  return AVERROR_INVALIDDATA;
2784  par->block_align = track->audio.sub_packet_size;
2785  *extradata_offset = 78;
2786  }
2787  track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2788  track->audio.frame_size);
2789  if (!track->audio.buf)
2790  return AVERROR(ENOMEM);
2791  break;
2792  }
2793  case AV_CODEC_ID_ATRAC1:
2794  /* ATRAC1 uses a constant frame size.
2795  * Typical ATRAC1 streams are either mono or stereo.
2796  * At most, ATRAC1 was used to store 8 channels of audio. */
2797  if (track->audio.channels > 8)
2798  return AVERROR_INVALIDDATA;
2799  par->block_align = track->audio.channels * 212;
2800  break;
2801  case AV_CODEC_ID_FLAC:
2802  if (track->codec_priv.size) {
2803  ret = matroska_parse_flac(s, track, extradata_offset);
2804  if (ret < 0)
2805  return ret;
2806  }
2807  break;
2808  case AV_CODEC_ID_WAVPACK:
2809  if (track->codec_priv.size < 2) {
2810  av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2811  "in absence of valid CodecPrivate.\n");
2812  extradata_size = WAVPACK_EXTRADATA_SIZE;
2813  AV_WL16(extradata, 0x410);
2814  }
2815  break;
2816  }
2817 
2818  if (extradata_size > 0) {
2819  ret = ff_alloc_extradata(par, extradata_size);
2820  if (ret < 0)
2821  return ret;
2822  memcpy(par->extradata, extradata, extradata_size);
2823  }
2824 
2825  return 0;
2826 }
2827 
2828 /* Performs the generic part of parsing an audio track. */
2829 static int mka_parse_audio(MatroskaTrack *track, AVStream *st,
2830  AVCodecParameters *par,
2831  const MatroskaDemuxContext *matroska,
2832  AVFormatContext *s, int *extradata_offset)
2833 {
2834  FFStream *const sti = ffstream(st);
2835  int ret;
2836 
2837  ret = mka_parse_audio_codec(track, par, matroska,
2838  s, extradata_offset);
2839  if (ret)
2840  return ret;
2841 
2843  par->sample_rate = track->audio.out_samplerate;
2844  // channel layout may be already set by codec private checks above
2845  if (!av_channel_layout_check(&par->ch_layout)) {
2847  par->ch_layout.nb_channels = track->audio.channels;
2848  }
2849  if (!par->bits_per_coded_sample)
2850  par->bits_per_coded_sample = track->audio.bitdepth;
2851  if (par->codec_id == AV_CODEC_ID_MP3 ||
2852  par->codec_id == AV_CODEC_ID_MLP ||
2853  par->codec_id == AV_CODEC_ID_TRUEHD)
2855  else if (par->codec_id != AV_CODEC_ID_AAC)
2857  if (track->codec_delay > 0) {
2859  (AVRational){1, 1000000000},
2860  (AVRational){1, par->codec_id == AV_CODEC_ID_OPUS ?
2861  48000 : par->sample_rate});
2862  }
2863  if (track->seek_preroll > 0) {
2864  par->seek_preroll = av_rescale_q(track->seek_preroll,
2865  (AVRational){1, 1000000000},
2866  (AVRational){1, par->sample_rate});
2867  }
2868 
2869  return 0;
2870 }
2871 
2872 /* Performs the codec-specific part of parsing a video track. */
2873 static int mkv_parse_video_codec(MatroskaTrack *track, AVCodecParameters *par,
2874  const MatroskaDemuxContext *matroska,
2875  int *extradata_offset)
2876 {
2877  if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2878  track->codec_priv.size >= 40) {
2879  track->ms_compat = 1;
2880  par->bits_per_coded_sample = AV_RL16(track->codec_priv.data + 14);
2881  par->codec_tag = AV_RL32(track->codec_priv.data + 16);
2883  par->codec_tag);
2884  if (!par->codec_id)
2886  par->codec_tag);
2887  *extradata_offset = 40;
2888  return 0;
2889  } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2890  track->codec_priv.size >= 21) {
2891  enum AVCodecID codec_id;
2892  uint32_t fourcc;
2893  int ret = get_qt_codec(track, &fourcc, &codec_id);
2894  if (ret < 0)
2895  return ret;
2896  if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2897  fourcc = MKTAG('S','V','Q','3');
2899  }
2900  par->codec_id = codec_id;
2901  if (codec_id == AV_CODEC_ID_NONE)
2902  av_log(matroska->ctx, AV_LOG_ERROR,
2903  "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2904  if (track->codec_priv.size >= 86) {
2905  FFIOContext b;
2906  unsigned bit_depth = AV_RB16(track->codec_priv.data + 82);
2908  track->codec_priv.size);
2909  if (ff_get_qtpalette(codec_id, &b.pub, track->palette)) {
2910  bit_depth &= 0x1F;
2911  track->has_palette = 1;
2912  }
2914  }
2915  par->codec_tag = fourcc;
2916  return 0;
2917  }
2918 
2919  switch (par->codec_id) {
2920  case AV_CODEC_ID_RV10:
2921  case AV_CODEC_ID_RV20:
2922  case AV_CODEC_ID_RV30:
2923  case AV_CODEC_ID_RV40:
2924  *extradata_offset = 26;
2925  break;
2926  case AV_CODEC_ID_PRORES:
2927  if (track->codec_priv.size == 4)
2928  par->codec_tag = AV_RL32(track->codec_priv.data);
2929  break;
2930  case AV_CODEC_ID_VP9:
2931  /* we don't need any value stored in CodecPrivate.
2932  * make sure that it's not exported as extradata. */
2933  track->codec_priv.size = 0;
2934  break;
2935  }
2936 
2937  return 0;
2938 }
2939 
2940 /* Performs the generic part of parsing a video track. */
2941 static int mkv_parse_video(MatroskaTrack *track, AVStream *st,
2942  AVCodecParameters *par,
2943  const MatroskaDemuxContext *matroska,
2944  int *extradata_offset)
2945 {
2946  FFStream *const sti = ffstream(st);
2948  int display_width_mul = 1;
2949  int display_height_mul = 1;
2950  int ret;
2951 
2952  if (track->video.color_space.size == 4)
2953  par->codec_tag = AV_RL32(track->video.color_space.data);
2954 
2955  ret = mkv_parse_video_codec(track, par, matroska,
2956  extradata_offset);
2957  if (ret < 0)
2958  return ret;
2959 
2961  par->width = track->video.pixel_width;
2962  par->height = track->video.pixel_height;
2963 
2965  par->field_order = mkv_field_order(matroska, track->video.field_order);
2968 
2971  &display_width_mul, &display_height_mul);
2972 
2974  if (track->video.display_width && track->video.display_height &&
2975  track->video.display_width != -1 && track->video.display_height != -1 &&
2976  track->video.cropped_height < INT64_MAX / track->video.display_width / display_width_mul &&
2977  track->video.cropped_width < INT64_MAX / track->video.display_height / display_height_mul)
2979  &st->sample_aspect_ratio.den,
2980  track->video.cropped_height * track->video.display_width * display_width_mul,
2981  track->video.cropped_width * track->video.display_height * display_height_mul,
2982  INT_MAX);
2983  }
2984  if (track->video.cropped_width != track->video.pixel_width ||
2985  track->video.cropped_height != track->video.pixel_height) {
2986  uint8_t *cropping;
2990  sizeof(uint32_t) * 4, 0);
2991  if (!sd)
2992  return AVERROR(ENOMEM);
2993 
2994  cropping = sd->data;
2995  bytestream_put_le32(&cropping, track->video.pixel_cropt);
2996  bytestream_put_le32(&cropping, track->video.pixel_cropb);
2997  bytestream_put_le32(&cropping, track->video.pixel_cropl);
2998  bytestream_put_le32(&cropping, track->video.pixel_cropr);
2999  }
3000  if (par->codec_id != AV_CODEC_ID_HEVC)
3002 
3003  if (track->default_duration) {
3004  int div = track->default_duration <= INT64_MAX ? 1 : 2;
3006  1000000000 / div, track->default_duration / div, 30000);
3007 #if FF_API_R_FRAME_RATE
3008  if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
3009  && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
3010  st->r_frame_rate = st->avg_frame_rate;
3011 #endif
3012  }
3013 
3014  /* export stereo mode flag as metadata tag */
3016  av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
3017 
3018  /* export alpha mode flag as metadata tag */
3019  if (track->video.alpha_mode)
3020  av_dict_set_int(&st->metadata, "alpha_mode", 1, 0);
3021 
3022  /* if we have virtual track, mark the real tracks */
3024  for (int j = 0; j < track->operation.combine_planes.nb_elem; j++) {
3025  MatroskaTrack *tracks = matroska->tracks.elem;
3026  char buf[32];
3028  continue;
3029  snprintf(buf, sizeof(buf), "%s_%d",
3031  for (int k = 0; k < matroska->tracks.nb_elem; k++)
3032  if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
3033  av_dict_set(&tracks[k].stream->metadata,
3034  "stereo_mode", buf, 0);
3035  break;
3036  }
3037  }
3038  // add stream level stereo3d side data if it is a supported format
3042  int ret = mkv_stereo3d_conv(st, track->video.stereo_mode);
3043  if (ret < 0)
3044  return ret;
3045  }
3046 
3047  ret = mkv_parse_video_color(st, track);
3048  if (ret < 0)
3049  return ret;
3050  ret = mkv_parse_video_projection(st, track, matroska->ctx);
3051  if (ret < 0)
3052  return ret;
3053 
3054  return 0;
3055 }
3056 
3057 /* Performs the codec-specific part of parsing a subtitle track. */
3058 static int mkv_parse_subtitle_codec(MatroskaTrack *track, AVStream *st,
3059  AVCodecParameters *par,
3060  const MatroskaDemuxContext *matroska)
3061 {
3062  switch (par->codec_id) {
3064  if (track->codec_priv.size == 3) {
3065  int component_tag = track->codec_priv.data[0];
3066  int data_component_id = AV_RB16(track->codec_priv.data + 1);
3067 
3068  switch (data_component_id) {
3069  case 0x0008:
3070  // [0x30..0x37] are component tags utilized for
3071  // non-mobile captioning service ("profile A").
3072  if (component_tag >= 0x30 && component_tag <= 0x37) {
3074  }
3075  break;
3076  case 0x0012:
3077  // component tag 0x87 signifies a mobile/partial reception
3078  // (1seg) captioning service ("profile C").
3079  if (component_tag == 0x87) {
3081  }
3082  break;
3083  default:
3084  break;
3085  }
3086 
3087  if (par->profile == AV_PROFILE_UNKNOWN)
3088  av_log(matroska->ctx, AV_LOG_WARNING,
3089  "Unknown ARIB caption profile utilized: %02x / %04x\n",
3090  component_tag, data_component_id);
3091 
3092  track->codec_priv.size = 0;
3093  }
3094  break;
3095  case AV_CODEC_ID_WEBVTT:
3096  if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
3098  } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
3100  } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
3102  }
3103  break;
3104  }
3105 
3106  return 0;
3107 }
3108 
3110 {
3111  MatroskaDemuxContext *matroska = s->priv_data;
3112  MatroskaTrack *tracks = matroska->tracks.elem;
3113  int i, j, ret;
3114 
3115  for (i = 0; i < matroska->tracks.nb_elem; i++) {
3116  MatroskaTrack *track = &tracks[i];
3118  AVCodecParameters *par;
3120  int extradata_offset = 0;
3121  AVStream *st;
3122  char* key_id_base64 = NULL;
3123 
3124  /* Apply some sanity checks. */
3125  if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
3126  track->type != MATROSKA_TRACK_TYPE_AUDIO &&
3127  track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
3128  track->type != MATROSKA_TRACK_TYPE_METADATA) {
3129  av_log(matroska->ctx, AV_LOG_INFO,
3130  "Unknown or unsupported track type %"PRIu64"\n",
3131  track->type);
3132  continue;
3133  }
3134  if (!track->codec_id)
3135  continue;
3136 
3137  if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
3138  || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
3139  || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
3140  || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
3141  ) {
3142  av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
3143  continue;
3144  }
3145 
3146  if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
3147  isnan(track->audio.samplerate)) {
3148  av_log(matroska->ctx, AV_LOG_WARNING,
3149  "Invalid sample rate %f, defaulting to 8000 instead.\n",
3150  track->audio.samplerate);
3151  track->audio.samplerate = 8000;
3152  }
3153 
3154  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
3155  if (!track->default_duration && track->video.frame_rate > 0) {
3156  double default_duration = 1000000000 / track->video.frame_rate;
3157  if (default_duration > UINT64_MAX || default_duration < 0) {
3158  av_log(matroska->ctx, AV_LOG_WARNING,
3159  "Invalid frame rate %e. Cannot calculate default duration.\n",
3160  track->video.frame_rate);
3161  } else {
3162  track->default_duration = default_duration;
3163  }
3164  }
3165  if (track->video.pixel_cropl >= INT_MAX - track->video.pixel_cropr ||
3166  track->video.pixel_cropt >= INT_MAX - track->video.pixel_cropb ||
3167  (track->video.pixel_cropl + track->video.pixel_cropr) >= track->video.pixel_width ||
3168  (track->video.pixel_cropt + track->video.pixel_cropb) >= track->video.pixel_height)
3169  return AVERROR_INVALIDDATA;
3170  track->video.cropped_width = track->video.pixel_width -
3171  track->video.pixel_cropl - track->video.pixel_cropr;
3172  track->video.cropped_height = track->video.pixel_height -
3173  track->video.pixel_cropt - track->video.pixel_cropb;
3175  if (track->video.display_width == -1)
3176  track->video.display_width = track->video.cropped_width;
3177  if (track->video.display_height == -1)
3178  track->video.display_height = track->video.cropped_height;
3179  }
3180  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
3181  if (!track->audio.out_samplerate)
3182  track->audio.out_samplerate = track->audio.samplerate;
3183  }
3185  track->encodings.nb_elem,
3186  track, &key_id_base64, matroska->ctx);
3187  if (ret < 0)
3188  return ret;
3189 
3190  for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
3191  if (av_strstart(track->codec_id, ff_mkv_codec_tags[j].str, NULL)) {
3193  break;
3194  }
3195  }
3196 
3197  st = track->stream = avformat_new_stream(s, NULL);
3198  if (!st) {
3199  av_free(key_id_base64);
3200  return AVERROR(ENOMEM);
3201  }
3202  par = st->codecpar;
3203 
3204  par->codec_id = codec_id;
3205 
3206  if (track->flag_default)
3208  if (track->flag_forced)
3210  if (track->flag_comment)
3212  if (track->flag_hearingimpaired)
3214  if (track->flag_visualimpaired)
3216  if (track->flag_original.count > 0)
3219 
3220  if (key_id_base64) {
3221  /* export encryption key id as base64 metadata tag */
3222  av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
3224  }
3225 
3226  if (strcmp(track->language, "und"))
3227  av_dict_set(&st->metadata, "language", track->language, 0);
3228  av_dict_set(&st->metadata, "title", track->name, 0);
3229 
3230  if (track->time_scale < 0.01) {
3231  av_log(matroska->ctx, AV_LOG_WARNING,
3232  "Track TimestampScale too small %f, assuming 1.0.\n",
3233  track->time_scale);
3234  track->time_scale = 1.0;
3235  }
3236 
3237  if (matroska->time_scale * track->time_scale > UINT_MAX)
3238  return AVERROR_INVALIDDATA;
3239 
3240  avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
3241  1000 * 1000 * 1000); /* 64 bit pts in ns */
3242 
3243  /* convert the delay from ns to the track timebase */
3245  (AVRational){ 1, 1000000000 },
3246  st->time_base);
3247 
3248  type = track->type;
3249  if (par->codec_id == AV_CODEC_ID_WEBVTT)
3251  switch (type) {
3253  ret = mka_parse_audio(track, st, par, matroska,
3254  s, &extradata_offset);
3255  if (ret < 0)
3256  return ret;
3257  if (ret == SKIP_TRACK)
3258  continue;
3259  break;
3261  ret = mkv_parse_video(track, st, par, matroska, &extradata_offset);
3262  if (ret < 0)
3263  return ret;
3264  break;
3266  ret = mkv_parse_subtitle_codec(track, st, par, matroska);
3267  if (ret < 0)
3268  return ret;
3270 
3271  if (track->flag_textdescriptions)
3273  break;
3274  }
3275 
3276  if (par->codec_id == AV_CODEC_ID_NONE)
3277  av_log(matroska->ctx, AV_LOG_INFO,
3278  "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
3279 
3280  if (!par->extradata && track->codec_priv.size > extradata_offset) {
3281  const uint8_t *src = track->codec_priv.data + extradata_offset;
3282  unsigned extra_size = track->codec_priv.size - extradata_offset;
3283  ret = ff_alloc_extradata(par, extra_size);
3284  if (ret < 0)
3285  return ret;
3286  memcpy(par->extradata, src, extra_size);
3287  }
3288 
3289  ret = mkv_parse_block_addition_mappings(s, st, track);
3290  if (ret < 0)
3291  return ret;
3292  }
3293 
3294  return 0;
3295 }
3296 
3298 {
3299  FFFormatContext *const si = ffformatcontext(s);
3300  MatroskaDemuxContext *matroska = s->priv_data;
3301  EbmlList *attachments_list = &matroska->attachments;
3302  EbmlList *chapters_list = &matroska->chapters;
3303  MatroskaAttachment *attachments;
3304  MatroskaChapter *chapters;
3305  uint64_t max_start = 0;
3306  int64_t pos;
3307  Ebml ebml = { 0 };
3308  int i, j, res;
3309 
3310  matroska->ctx = s;
3311  matroska->cues_parsing_deferred = 1;
3312 
3313  /* First read the EBML header. */
3314  if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
3315  av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
3316  ebml_free(ebml_syntax, &ebml);
3317  return AVERROR_INVALIDDATA;
3318  }
3319  if (ebml.version > EBML_VERSION ||
3320  ebml.max_size > sizeof(uint64_t) ||
3321  ebml.id_length > sizeof(uint32_t) ||
3322  ebml.doctype_version > 3) {
3324  "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
3325  ebml.version, ebml.doctype, ebml.doctype_version);
3326  ebml_free(ebml_syntax, &ebml);
3327  return AVERROR_PATCHWELCOME;
3328  } else if (ebml.doctype_version == 3) {
3329  av_log(matroska->ctx, AV_LOG_WARNING,
3330  "EBML header using unsupported features\n"
3331  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
3332  ebml.version, ebml.doctype, ebml.doctype_version);
3333  }
3334  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
3335  if (!strcmp(ebml.doctype, matroska_doctypes[i]))
3336  break;
3338  av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
3339  if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
3340  ebml_free(ebml_syntax, &ebml);
3341  return AVERROR_INVALIDDATA;
3342  }
3343  }
3344  matroska->is_webm = !strcmp(ebml.doctype, "webm");
3345 
3346  ebml_free(ebml_syntax, &ebml);
3347 
3348  matroska->pkt = si->parse_pkt;
3349 
3350  /* The next thing is a segment. */
3351  pos = avio_tell(matroska->ctx->pb);
3352  res = ebml_parse(matroska, matroska_segments, matroska);
3353  // Try resyncing until we find an EBML_STOP type element.
3354  while (res != 1) {
3355  res = matroska_resync(matroska, pos);
3356  if (res < 0)
3357  return res;
3358  pos = avio_tell(matroska->ctx->pb);
3359  res = ebml_parse(matroska, matroska_segment, matroska);
3360  if (res == AVERROR(EIO)) // EOF is translated to EIO, this exists the loop on EOF
3361  return res;
3362  }
3363  /* Set data_offset as it might be needed later by seek_frame_generic. */
3364  if (matroska->current_id == MATROSKA_ID_CLUSTER)
3365  si->data_offset = avio_tell(matroska->ctx->pb) - 4;
3366  matroska_execute_seekhead(matroska);
3367 
3368  if (!matroska->time_scale)
3369  matroska->time_scale = 1000000;
3370  if (isnan(matroska->duration))
3371  matroska->duration = 0;
3372  if (matroska->duration)
3373  matroska->ctx->duration = matroska->duration * matroska->time_scale *
3374  1000 / AV_TIME_BASE;
3375  av_dict_set(&s->metadata, "title", matroska->title, 0);
3376  av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
3377 
3378  if (matroska->date_utc.size == 8)
3379  matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
3380 
3381  res = matroska_parse_tracks(s);
3382  if (res < 0)
3383  return res;
3384 
3385  attachments = attachments_list->elem;
3386  for (j = 0; j < attachments_list->nb_elem; j++) {
3387  if (!(attachments[j].filename && attachments[j].mime &&
3388  attachments[j].bin.data && attachments[j].bin.size > 0)) {
3389  av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
3390  } else {
3392  if (!st)
3393  break;
3394  av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
3395  av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
3396  if (attachments[j].description)
3397  av_dict_set(&st->metadata, "title", attachments[j].description, 0);
3399 
3400  for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3401  if (av_strstart(attachments[j].mime, mkv_image_mime_tags[i].str, NULL)) {
3403  break;
3404  }
3405  }
3406 
3407  attachments[j].stream = st;
3408 
3409  if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
3410  res = ff_add_attached_pic(s, st, NULL, &attachments[j].bin.buf, 0);
3411  if (res < 0)
3412  return res;
3413  } else {
3415  if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
3416  break;
3417  memcpy(st->codecpar->extradata, attachments[j].bin.data,
3418  attachments[j].bin.size);
3419 
3420  for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3421  if (av_strstart(attachments[j].mime, mkv_mime_tags[i].str, NULL)) {
3423  break;
3424  }
3425  }
3426  }
3427  }
3428  }
3429 
3430  chapters = chapters_list->elem;
3431  for (i = 0; i < chapters_list->nb_elem; i++)
3432  if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
3433  (max_start == 0 || chapters[i].start > max_start)) {
3434  chapters[i].chapter =
3435  avpriv_new_chapter(s, chapters[i].uid,
3436  (AVRational) { 1, 1000000000 },
3437  chapters[i].start, chapters[i].end,
3438  chapters[i].title);
3439  max_start = chapters[i].start;
3440  }
3441 
3442  matroska_add_index_entries(matroska);
3443 
3445 
3446  return 0;
3448 
3449 /*
3450  * Put one packet in an application-supplied AVPacket struct.
3451  * Returns 0 on success or -1 on failure.
3452  */
3453 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3454  AVPacket *pkt)
3455 {
3456  if (matroska->queue.head) {
3457  MatroskaTrack *tracks = matroska->tracks.elem;
3458  MatroskaTrack *track;
3459 
3460  avpriv_packet_list_get(&matroska->queue, pkt);
3461  track = &tracks[pkt->stream_index];
3462  if (track->has_palette) {
3464  if (!pal) {
3465  av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3466  } else {
3467  memcpy(pal, track->palette, AVPALETTE_SIZE);
3468  }
3469  track->has_palette = 0;
3470  }
3471  return 0;
3472  }
3473 
3474  return -1;
3475 }
3476 
3477 /*
3478  * Free all packets in our internal queue.
3479  */
3480 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3481 {
3482  avpriv_packet_list_free(&matroska->queue);
3483 }
3484 
3485 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3486  int size, int type, AVIOContext *pb,
3487  uint32_t lace_size[256], int *laces)
3488 {
3489  int n;
3490  uint8_t *data = *buf;
3491 
3492  if (!type) {
3493  *laces = 1;
3494  lace_size[0] = size;
3495  return 0;
3496  }
3497 
3498  if (size <= 0)
3499  return AVERROR_INVALIDDATA;
3500 
3501  *laces = *data + 1;
3502  data += 1;
3503  size -= 1;
3504 
3505  switch (type) {
3506  case 0x1: /* Xiph lacing */
3507  {
3508  uint8_t temp;
3509  uint32_t total = 0;
3510  for (n = 0; n < *laces - 1; n++) {
3511  lace_size[n] = 0;
3512 
3513  do {
3514  if (size <= total)
3515  return AVERROR_INVALIDDATA;
3516  temp = *data;
3517  total += temp;
3518  lace_size[n] += temp;
3519  data += 1;
3520  size -= 1;
3521  } while (temp == 0xff);
3522  }
3523  if (size < total)
3524  return AVERROR_INVALIDDATA;
3525 
3526  lace_size[n] = size - total;
3527  break;
3528  }
3529 
3530  case 0x2: /* fixed-size lacing */
3531  if (size % (*laces))
3532  return AVERROR_INVALIDDATA;
3533  for (n = 0; n < *laces; n++)
3534  lace_size[n] = size / *laces;
3535  break;
3536 
3537  case 0x3: /* EBML lacing */
3538  {
3539  uint64_t num;
3540  uint64_t total;
3541  int offset;
3542 
3543  avio_skip(pb, 4);
3544 
3545  n = ebml_read_num(matroska, pb, 8, &num, 1);
3546  if (n < 0)
3547  return n;
3548  if (num > INT_MAX)
3549  return AVERROR_INVALIDDATA;
3550 
3551  total = lace_size[0] = num;
3552  offset = n;
3553  for (n = 1; n < *laces - 1; n++) {
3554  int64_t snum;
3555  int r;
3556  r = matroska_ebmlnum_sint(matroska, pb, &snum);
3557  if (r < 0)
3558  return r;
3559  if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3560  return AVERROR_INVALIDDATA;
3561 
3562  lace_size[n] = lace_size[n - 1] + snum;
3563  total += lace_size[n];
3564  offset += r;
3565  }
3566  data += offset;
3567  size -= offset;
3568  if (size < total)
3569  return AVERROR_INVALIDDATA;
3570 
3571  lace_size[*laces - 1] = size - total;
3572  break;
3573  }
3574  }
3576  *buf = data;
3577 
3578  return 0;
3579 }
3580 
3581 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3582  MatroskaTrack *track, AVStream *st,
3583  uint8_t *data, int size, uint64_t timecode,
3584  int64_t pos)
3585 {
3586  const int a = st->codecpar->block_align;
3587  const int sps = track->audio.sub_packet_size;
3588  const int cfs = track->audio.coded_framesize;
3589  const int h = track->audio.sub_packet_h;
3590  const int w = track->audio.frame_size;
3591  int y = track->audio.sub_packet_cnt;
3592  int x;
3593 
3594  if (!track->audio.pkt_cnt) {
3595  if (track->audio.sub_packet_cnt == 0)
3596  track->audio.buf_timecode = timecode;
3597  if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3598  if (size < cfs * h / 2) {
3599  av_log(matroska->ctx, AV_LOG_ERROR,
3600  "Corrupt int4 RM-style audio packet size\n");
3601  return AVERROR_INVALIDDATA;
3602  }
3603  for (x = 0; x < h / 2; x++)
3604  memcpy(track->audio.buf + x * 2 * w + y * cfs,
3605  data + x * cfs, cfs);
3606  } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3607  if (size < w) {
3608  av_log(matroska->ctx, AV_LOG_ERROR,
3609  "Corrupt sipr RM-style audio packet size\n");
3610  return AVERROR_INVALIDDATA;
3611  }
3612  memcpy(track->audio.buf + y * w, data, w);
3613  } else {
3614  if (size < w) {
3615  av_log(matroska->ctx, AV_LOG_ERROR,
3616  "Corrupt generic RM-style audio packet size\n");
3617  return AVERROR_INVALIDDATA;
3618  }
3619  for (x = 0; x < w / sps; x++)
3620  memcpy(track->audio.buf +
3621  sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3622  data + x * sps, sps);
3623  }
3624 
3625  if (++track->audio.sub_packet_cnt >= h) {
3626  if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3627  ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3628  track->audio.sub_packet_cnt = 0;
3629  track->audio.pkt_cnt = h * w / a;
3630  }
3631  }
3632 
3633  while (track->audio.pkt_cnt) {
3634  int ret;
3635  AVPacket *pkt = matroska->pkt;
3636 
3637  ret = av_new_packet(pkt, a);
3638  if (ret < 0) {
3639  return ret;
3640  }
3641  memcpy(pkt->data,
3642  track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3643  a);
3644  pkt->pts = track->audio.buf_timecode;
3646  pkt->pos = pos;
3647  pkt->stream_index = st->index;
3648  ret = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
3649  if (ret < 0) {
3651  return AVERROR(ENOMEM);
3652  }
3653  }
3654 
3655  return 0;
3656 }
3657 
3658 /* reconstruct full wavpack blocks from mangled matroska ones */
3659 static int matroska_parse_wavpack(MatroskaTrack *track,
3660  uint8_t **data, int *size)
3661 {
3662  uint8_t *dst = NULL;
3663  uint8_t *src = *data;
3664  int dstlen = 0;
3665  int srclen = *size;
3666  uint32_t samples;
3667  uint16_t ver;
3668  int ret, offset = 0;
3669 
3670  if (srclen < 12)
3671  return AVERROR_INVALIDDATA;
3672 
3673  av_assert1(track->stream->codecpar->extradata_size >= 2);
3674  ver = AV_RL16(track->stream->codecpar->extradata);
3675 
3676  samples = AV_RL32(src);
3677  src += 4;
3678  srclen -= 4;
3679 
3680  while (srclen >= 8) {
3681  int multiblock;
3682  uint32_t blocksize;
3683  uint8_t *tmp;
3684 
3685  uint32_t flags = AV_RL32(src);
3686  uint32_t crc = AV_RL32(src + 4);
3687  src += 8;
3688  srclen -= 8;
3689 
3690  multiblock = (flags & 0x1800) != 0x1800;
3691  if (multiblock) {
3692  if (srclen < 4) {
3694  goto fail;
3695  }
3696  blocksize = AV_RL32(src);
3697  src += 4;
3698  srclen -= 4;
3699  } else
3700  blocksize = srclen;
3701 
3702  if (blocksize > srclen) {
3704  goto fail;
3705  }
3706 
3707  tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3708  if (!tmp) {
3709  ret = AVERROR(ENOMEM);
3710  goto fail;
3711  }
3712  dst = tmp;
3713  dstlen += blocksize + 32;
3714 
3715  AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3716  AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3717  AV_WL16(dst + offset + 8, ver); // version
3718  AV_WL16(dst + offset + 10, 0); // track/index_no
3719  AV_WL32(dst + offset + 12, 0); // total samples
3720  AV_WL32(dst + offset + 16, 0); // block index
3721  AV_WL32(dst + offset + 20, samples); // number of samples
3722  AV_WL32(dst + offset + 24, flags); // flags
3723  AV_WL32(dst + offset + 28, crc); // crc
3724  memcpy(dst + offset + 32, src, blocksize); // block data
3725 
3726  src += blocksize;
3727  srclen -= blocksize;
3728  offset += blocksize + 32;
3729  }
3730 
3731  memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3732 
3733  *data = dst;
3734  *size = dstlen;
3735 
3736  return 0;
3738 fail:
3739  av_freep(&dst);
3740  return ret;
3741 }
3742 
3743 static int matroska_parse_prores(MatroskaTrack *track,
3744  uint8_t **data, int *size)
3745 {
3746  uint8_t *dst;
3747  int dstlen = *size + 8;
3748 
3750  if (!dst)
3751  return AVERROR(ENOMEM);
3752 
3753  AV_WB32(dst, dstlen);
3754  AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3755  memcpy(dst + 8, *data, dstlen - 8);
3756  memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3757 
3758  *data = dst;
3759  *size = dstlen;
3760 
3761  return 0;
3762 }
3763 
3764 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3765  MatroskaTrack *track,
3766  AVStream *st,
3767  uint8_t *data, int data_len,
3768  uint64_t timecode,
3769  uint64_t duration,
3770  int64_t pos)
3771 {
3772  AVPacket *pkt = matroska->pkt;
3773  uint8_t *id, *settings, *text, *buf;
3774  int id_len, settings_len, text_len;
3775  uint8_t *p, *q;
3776  int err;
3777 
3778  if (data_len <= 0)
3779  return AVERROR_INVALIDDATA;
3780 
3781  p = data;
3782  q = data + data_len;
3783 
3784  id = p;
3785  id_len = -1;
3786  while (p < q) {
3787  if (*p == '\r' || *p == '\n') {
3788  id_len = p - id;
3789  if (*p == '\r')
3790  p++;
3791  break;
3792  }
3793  p++;
3794  }
3795 
3796  if (p >= q || *p != '\n')
3797  return AVERROR_INVALIDDATA;
3798  p++;
3799 
3800  settings = p;
3801  settings_len = -1;
3802  while (p < q) {
3803  if (*p == '\r' || *p == '\n') {
3804  settings_len = p - settings;
3805  if (*p == '\r')
3806  p++;
3807  break;
3808  }
3809  p++;
3810  }
3811 
3812  if (p >= q || *p != '\n')
3813  return AVERROR_INVALIDDATA;
3814  p++;
3815 
3816  text = p;
3817  text_len = q - p;
3818  while (text_len > 0) {
3819  const int len = text_len - 1;
3820  const uint8_t c = p[len];
3821  if (c != '\r' && c != '\n')
3822  break;
3823  text_len = len;
3824  }
3825 
3826  if (text_len <= 0)
3827  return AVERROR_INVALIDDATA;
3828 
3829  err = av_new_packet(pkt, text_len);
3830  if (err < 0) {
3831  return err;
3832  }
3833 
3834  memcpy(pkt->data, text, text_len);
3835 
3836  if (id_len > 0) {
3839  id_len);
3840  if (!buf) {
3842  return AVERROR(ENOMEM);
3843  }
3844  memcpy(buf, id, id_len);
3845  }
3846 
3847  if (settings_len > 0) {
3850  settings_len);
3851  if (!buf) {
3853  return AVERROR(ENOMEM);
3854  }
3855  memcpy(buf, settings, settings_len);
3856  }
3857 
3858  // Do we need this for subtitles?
3859  // pkt->flags = AV_PKT_FLAG_KEY;
3860 
3861  pkt->stream_index = st->index;
3862  pkt->pts = timecode;
3863 
3864  // Do we need this for subtitles?
3865  // pkt->dts = timecode;
3866 
3867  pkt->duration = duration;
3868  pkt->pos = pos;
3869 
3870  err = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
3871  if (err < 0) {
3873  return AVERROR(ENOMEM);
3874  }
3875 
3876  return 0;
3877 }
3878 
3880  MatroskaTrack *track, AVPacket *pkt,
3881  const uint8_t *data, int size, uint64_t id)
3882 {
3883  const EbmlList *mappings_list = &track->block_addition_mappings;
3884  MatroskaBlockAdditionMapping *mappings = mappings_list->elem, *mapping = NULL;
3885  uint8_t *side_data;
3886  int res;
3887 
3888  if (!matroska->is_webm && track->max_block_additional_id && id > track->max_block_additional_id) {
3889  int strict = matroska->ctx->strict_std_compliance >= FF_COMPLIANCE_STRICT;
3890  av_log(matroska->ctx, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
3891  "BlockAddID %"PRIu64" is higher than the reported MaxBlockAdditionID %"PRIu64" "
3892  "for Track with TrackNumber %"PRIu64"\n", id, track->max_block_additional_id,
3893  track->num);
3894  if (strict)
3895  return AVERROR_INVALIDDATA;
3896  }
3897 
3898  for (int i = 0; i < mappings_list->nb_elem; i++) {
3899  if (id != mappings[i].value)
3900  continue;
3901  mapping = &mappings[i];
3902  break;
3903  }
3904 
3905  if (id != 1 && !matroska->is_webm && !mapping) {
3906  av_log(matroska->ctx, AV_LOG_WARNING, "BlockAddID %"PRIu64" has no mapping. Skipping\n", id);
3907  return 0;
3908  }
3909 
3910  if (mapping && mapping->type)
3911  id = mapping->type;
3912 
3913  switch (id) {
3915  GetByteContext bc;
3916  int country_code, provider_code;
3917  int provider_oriented_code, application_identifier;
3918  size_t hdrplus_size;
3919  AVDynamicHDRPlus *hdrplus;
3920 
3921  if (size < 6)
3922  break; //ignore
3923 
3924  bytestream2_init(&bc, data, size);
3925 
3926  /* ITU-T T.35 metadata */
3927  country_code = bytestream2_get_byteu(&bc);
3928  provider_code = bytestream2_get_be16u(&bc);
3929 
3930  if (country_code != ITU_T_T35_COUNTRY_CODE_US ||
3931  provider_code != ITU_T_T35_PROVIDER_CODE_SMTPE)
3932  break; // ignore
3933 
3934  provider_oriented_code = bytestream2_get_be16u(&bc);
3935  application_identifier = bytestream2_get_byteu(&bc);
3936 
3937  if (provider_oriented_code != 1 || application_identifier != 4)
3938  break; // ignore
3939 
3940  hdrplus = av_dynamic_hdr_plus_alloc(&hdrplus_size);
3941  if (!hdrplus)
3942  return AVERROR(ENOMEM);
3943 
3944  if ((res = av_dynamic_hdr_plus_from_t35(hdrplus, bc.buffer,
3945  bytestream2_get_bytes_left(&bc))) < 0 ||
3947  (uint8_t *)hdrplus, hdrplus_size)) < 0) {
3948  av_free(hdrplus);
3949  return res;
3950  }
3951 
3952  return 0;
3953  }
3954  default:
3955  break;
3956  }
3957 
3959  size + (size_t)8);
3960  if (!side_data)
3961  return AVERROR(ENOMEM);
3962 
3963  AV_WB64(side_data, id);
3964  memcpy(side_data + 8, data, size);
3965 
3966  return 0;
3967 }
3968 
3969 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3970  MatroskaTrack *track, AVStream *st,
3971  AVBufferRef *buf, uint8_t *data, int pkt_size,
3972  uint64_t timecode, uint64_t lace_duration,
3973  int64_t pos, int is_keyframe,
3974  MatroskaBlockMore *blockmore, int nb_blockmore,
3975  int64_t discard_padding)
3976 {
3977  uint8_t *pkt_data = data;
3978  int res = 0;
3979  AVPacket *pkt = matroska->pkt;
3980 
3981  if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3982  res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3983  if (res < 0) {
3984  av_log(matroska->ctx, AV_LOG_ERROR,
3985  "Error parsing a wavpack block.\n");
3986  goto fail;
3987  }
3988  if (!buf)
3989  av_freep(&data);
3990  buf = NULL;
3991  }
3992 
3993  if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3994  AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3995  res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3996  if (res < 0) {
3997  av_log(matroska->ctx, AV_LOG_ERROR,
3998  "Error parsing a prores block.\n");
3999  goto fail;
4000  }
4001  if (!buf)
4002  av_freep(&data);
4003  buf = NULL;
4004  }
4005 
4006  if (!pkt_size && !nb_blockmore)
4007  goto no_output;
4008 
4009  if (!matroska->is_webm && nb_blockmore && !track->max_block_additional_id) {
4010  int strict = matroska->ctx->strict_std_compliance >= FF_COMPLIANCE_STRICT;
4011  av_log(matroska->ctx, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
4012  "Unexpected BlockAdditions found in a Block from Track with TrackNumber %"PRIu64" "
4013  "where MaxBlockAdditionID is 0\n", track->num);
4014  if (strict) {
4015  res = AVERROR_INVALIDDATA;
4016  goto fail;
4017  }
4018  }
4019 
4020  if (!buf)
4021  pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
4022  NULL, NULL, 0);
4023  else
4024  pkt->buf = av_buffer_ref(buf);
4025 
4026  if (!pkt->buf) {
4027  res = AVERROR(ENOMEM);
4028  goto fail;
4029  }
4030 
4031  pkt->data = pkt_data;
4032  pkt->size = pkt_size;
4033  pkt->flags = is_keyframe;
4034  pkt->stream_index = st->index;
4035 
4036  for (int i = 0; i < nb_blockmore; i++) {
4037  MatroskaBlockMore *more = &blockmore[i];
4038 
4039  if (!more->additional.size)
4040  continue;
4041 
4042  res = matroska_parse_block_additional(matroska, track, pkt, more->additional.data,
4043  more->additional.size, more->additional_id);
4044  if (res < 0) {
4046  return res;
4047  }
4048  }
4049 
4050  if (discard_padding) {
4051  uint8_t *side_data = av_packet_new_side_data(pkt,
4053  10);
4054  if (!side_data) {
4056  return AVERROR(ENOMEM);
4057  }
4058  discard_padding = av_rescale_q(discard_padding,
4059  (AVRational){1, 1000000000},
4060  (AVRational){1, st->codecpar->sample_rate});
4061  if (discard_padding > 0) {
4062  AV_WL32A(side_data + 4, discard_padding);
4063  } else {
4064  AV_WL32A(side_data, -discard_padding);
4065  }
4066  }
4067 
4068  if (track->ms_compat)
4069  pkt->dts = timecode;
4070  else
4071  pkt->pts = timecode;
4072  pkt->pos = pos;
4073  pkt->duration = lace_duration;
4074 
4075  res = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
4076  if (res < 0) {
4078  return AVERROR(ENOMEM);
4079  }
4080 
4081  return 0;
4082 
4083 no_output:
4085  if (!buf)
4086  av_free(pkt_data);
4087  return res;
4088 }
4089 
4090 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
4091  int size, int64_t pos, uint64_t cluster_time,
4092  uint64_t block_duration, int is_keyframe,
4093  MatroskaBlockMore *blockmore, int nb_blockmore,
4094  int64_t cluster_pos, int64_t discard_padding)
4095 {
4096  uint64_t timecode = AV_NOPTS_VALUE;
4097  MatroskaTrack *track;
4098  FFIOContext pb;
4099  int res = 0;
4100  AVStream *st;
4101  int16_t block_time;
4102  uint32_t lace_size[256];
4103  int n, flags, laces = 0;
4104  uint64_t num;
4105  int trust_default_duration;
4106 
4107  av_assert1(buf);
4108 
4110 
4111  if ((n = ebml_read_num(matroska, &pb.pub, 8, &num, 1)) < 0)
4112  return n;
4113  data += n;
4114  size -= n;
4115 
4116  track = matroska_find_track_by_num(matroska, num);
4117  if (!track || size < 3)
4118  return AVERROR_INVALIDDATA;
4119 
4120  if (!(st = track->stream)) {
4121  av_log(matroska->ctx, AV_LOG_VERBOSE,
4122  "No stream associated to TrackNumber %"PRIu64". "
4123  "Ignoring Block with this TrackNumber.\n", num);
4124  return 0;
4125  }
4126 
4127  if (st->discard >= AVDISCARD_ALL)
4128  return res;
4129  if (block_duration > INT64_MAX)
4130  block_duration = INT64_MAX;
4131 
4132  block_time = sign_extend(AV_RB16(data), 16);
4133  data += 2;
4134  flags = *data++;
4135  size -= 3;
4136  if (is_keyframe == -1)
4137  is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
4138 
4139  if (cluster_time != (uint64_t) -1 &&
4140  (block_time >= 0 || cluster_time >= -block_time)) {
4141  uint64_t timecode_cluster_in_track_tb = (double) cluster_time / track->time_scale;
4142  timecode = timecode_cluster_in_track_tb + block_time - track->codec_delay_in_track_tb;
4143  if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
4144  timecode < track->end_timecode)
4145  is_keyframe = 0; /* overlapping subtitles are not key frame */
4146  if (is_keyframe) {
4147  ff_reduce_index(matroska->ctx, st->index);
4148  av_add_index_entry(st, cluster_pos, timecode, 0, 0,
4150  }
4151  }
4152 
4153  if (matroska->skip_to_keyframe &&
4154  track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
4155  // Compare signed timecodes. Timecode may be negative due to codec delay
4156  // offset. We don't support timestamps greater than int64_t anyway - see
4157  // AVPacket's pts.
4158  if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
4159  return res;
4160  if (is_keyframe)
4161  matroska->skip_to_keyframe = 0;
4162  else if (!ffstream(st)->skip_to_keyframe) {
4163  av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
4164  matroska->skip_to_keyframe = 0;
4165  }
4166  }
4167 
4168  res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
4169  &pb.pub, lace_size, &laces);
4170  if (res < 0) {
4171  av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
4172  return res;
4173  }
4174 
4175  trust_default_duration = track->default_duration != 0;
4176  if (track->audio.samplerate == 8000 && trust_default_duration) {
4177  // If this is needed for more codecs, then add them here
4178  if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
4179  if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
4180  trust_default_duration = 0;
4181  }
4182  }
4183 
4184  if (!block_duration && trust_default_duration)
4185  block_duration = track->default_duration * laces / matroska->time_scale;
4186 
4187  if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
4188  track->end_timecode =
4189  FFMAX(track->end_timecode, timecode + block_duration);
4190 
4191  for (n = 0; n < laces; n++) {
4192  int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
4193  uint8_t *out_data = data;
4194  int out_size = lace_size[n];
4195 
4196  if (track->needs_decoding) {
4197  res = matroska_decode_buffer(&out_data, &out_size, track);
4198  if (res < 0)
4199  return res;
4200  /* Given that we are here means that out_data is no longer
4201  * owned by buf, so set it to NULL. This depends upon
4202  * zero-length header removal compression being ignored. */
4203  av_assert1(out_data != data);
4204  buf = NULL;
4205  }
4206 
4207  if (track->audio.buf) {
4208  res = matroska_parse_rm_audio(matroska, track, st,
4209  out_data, out_size,
4210  timecode, pos);
4211  if (!buf)
4212  av_free(out_data);
4213  if (res)
4214  return res;
4215  } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
4216  res = matroska_parse_webvtt(matroska, track, st,
4217  out_data, out_size,
4218  timecode, lace_duration,
4219  pos);
4220  if (!buf)
4221  av_free(out_data);
4222  if (res)
4223  return res;
4224  } else {
4225  res = matroska_parse_frame(matroska, track, st, buf, out_data,
4226  out_size, timecode, lace_duration,
4227  pos, !n ? is_keyframe : 0,
4228  blockmore, nb_blockmore,
4229  discard_padding);
4230  if (res)
4231  return res;
4232  }
4233 
4234  if (timecode != AV_NOPTS_VALUE)
4235  timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
4236  data += lace_size[n];
4237  }
4238 
4239  return 0;
4240 }
4241 
4242 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
4243 {
4244  MatroskaCluster *cluster = &matroska->current_cluster;
4245  MatroskaBlock *block = &cluster->block;
4246  int res;
4247 
4248  av_assert0(matroska->num_levels <= 2U);
4249 
4250  if (matroska->num_levels == 1) {
4251  res = ebml_parse(matroska, matroska_segment, NULL);
4252 
4253  if (res == 1) {
4254  /* Found a cluster: subtract the size of the ID already read. */
4255  cluster->pos = avio_tell(matroska->ctx->pb) - 4;
4256 
4257  res = ebml_parse(matroska, matroska_cluster_enter, cluster);
4258  if (res < 0)
4259  return res;
4260  }
4261  }
4262 
4263  if (matroska->num_levels == 2) {
4264  /* We are inside a cluster. */
4265  res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
4266 
4267  if (res >= 0 && block->bin.size > 0) {
4268  int is_keyframe = block->non_simple ? block->reference.count == 0 : -1;
4269 
4270  res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
4271  block->bin.size, block->bin.pos,
4272  cluster->timecode, block->duration,
4273  is_keyframe, block->blockmore.elem,
4274  block->blockmore.nb_elem, cluster->pos,
4275  block->discard_padding);
4276  }
4277 
4279  memset(block, 0, sizeof(*block));
4280  } else if (!matroska->num_levels) {
4281  if (!avio_feof(matroska->ctx->pb)) {
4282  avio_r8(matroska->ctx->pb);
4283  if (!avio_feof(matroska->ctx->pb)) {
4284  av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
4285  "end of segment.\n");
4286  return AVERROR_INVALIDDATA;
4287  }
4288  }
4289  matroska->done = 1;
4290  return AVERROR_EOF;
4291  }
4292 
4293  return res;
4294 }
4295 
4297 {
4298  MatroskaDemuxContext *matroska = s->priv_data;
4299  int ret = 0;
4300 
4301  if (matroska->resync_pos == -1) {
4302  // This can only happen if generic seeking has been used.
4303  matroska->resync_pos = avio_tell(s->pb);
4304  }
4305 
4306  while (matroska_deliver_packet(matroska, pkt)) {
4307  if (matroska->done)
4308  return (ret < 0) ? ret : AVERROR_EOF;
4309  if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
4310  ret = matroska_resync(matroska, matroska->resync_pos);
4311  }
4312 
4313  return 0;
4314 }
4315 
4316 static int matroska_read_seek(AVFormatContext *s, int stream_index,
4317  int64_t timestamp, int flags)
4318 {
4319  MatroskaDemuxContext *matroska = s->priv_data;
4320  MatroskaTrack *tracks = NULL;
4321  AVStream *st = s->streams[stream_index];
4322  FFStream *const sti = ffstream(st);
4323  int i, index;
4324 
4325  /* Parse the CUES now since we need the index data to seek. */
4326  if (matroska->cues_parsing_deferred > 0) {
4327  matroska->cues_parsing_deferred = 0;
4328  matroska_parse_cues(matroska);
4329  }
4330 
4331  if (!sti->nb_index_entries)
4332  goto err;
4333  timestamp = FFMAX(timestamp, sti->index_entries[0].timestamp);
4334 
4335  if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 ||
4336  index == sti->nb_index_entries - 1) {
4337  matroska_reset_status(matroska, 0, sti->index_entries[sti->nb_index_entries - 1].pos);
4338  while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 ||
4339  index == sti->nb_index_entries - 1) {
4340  matroska_clear_queue(matroska);
4341  if (matroska_parse_cluster(matroska) < 0)
4342  break;
4343  }
4344  }
4345 
4346  matroska_clear_queue(matroska);
4347  if (index < 0 || (matroska->cues_parsing_deferred < 0 &&
4348  index == sti->nb_index_entries - 1))
4349  goto err;
4350 
4351  tracks = matroska->tracks.elem;