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 
33 #include <inttypes.h>
34 #include <stdio.h>
35 
36 #include "libavutil/avstring.h"
37 #include "libavutil/base64.h"
38 #include "libavutil/dict.h"
39 #include "libavutil/intfloat.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavutil/lzo.h"
43 #include "libavutil/mathematics.h"
44 #include "libavutil/opt.h"
46 #include "libavutil/spherical.h"
47 
48 #include "libavcodec/bytestream.h"
49 #include "libavcodec/flac.h"
50 #include "libavcodec/mpeg4audio.h"
51 
52 #include "avformat.h"
53 #include "avio_internal.h"
54 #include "internal.h"
55 #include "isom.h"
56 #include "matroska.h"
57 #include "oggdec.h"
58 /* For ff_codec_get_id(). */
59 #include "riff.h"
60 #include "rmsipr.h"
61 
62 #if CONFIG_BZLIB
63 #include <bzlib.h>
64 #endif
65 #if CONFIG_ZLIB
66 #include <zlib.h>
67 #endif
68 
69 #include "qtpalette.h"
70 
71 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
72 #define NEEDS_CHECKING 2 /* Indicates that some error checks
73  * still need to be performed */
74 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
75  * syntax level used for parsing ended. */
76 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
77  * of unkown, potentially damaged data is encountered,
78  * it is considered an error. */
79 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
80  * to this many bytes of unknown data for the
81  * SKIP_THRESHOLD check. */
82 
83 typedef enum {
89  EBML_UTF8,
90  EBML_BIN,
96 
97 typedef const struct EbmlSyntax {
98  uint32_t id;
101  size_t data_offset;
102  union {
103  int64_t i;
104  uint64_t u;
105  double f;
106  const char *s;
107  const struct EbmlSyntax *n;
108  } def;
109 } EbmlSyntax;
110 
111 typedef struct EbmlList {
112  int nb_elem;
113  unsigned int alloc_elem_size;
114  void *elem;
116 
117 typedef struct EbmlBin {
118  int size;
121  int64_t pos;
123 
124 typedef struct Ebml {
125  uint64_t version;
126  uint64_t max_size;
127  uint64_t id_length;
128  char *doctype;
129  uint64_t doctype_version;
130 } Ebml;
131 
132 typedef struct MatroskaTrackCompression {
133  uint64_t algo;
136 
137 typedef struct MatroskaTrackEncryption {
138  uint64_t algo;
141 
142 typedef struct MatroskaTrackEncoding {
143  uint64_t scope;
144  uint64_t type;
148 
149 typedef struct MatroskaMasteringMeta {
150  double r_x;
151  double r_y;
152  double g_x;
153  double g_y;
154  double b_x;
155  double b_y;
156  double white_x;
157  double white_y;
161 
162 typedef struct MatroskaTrackVideoColor {
165  uint64_t chroma_sub_horz;
166  uint64_t chroma_sub_vert;
167  uint64_t cb_sub_horz;
168  uint64_t cb_sub_vert;
171  uint64_t range;
172  uint64_t transfer_characteristics;
173  uint64_t primaries;
174  uint64_t max_cll;
175  uint64_t max_fall;
178 
179 typedef struct MatroskaTrackVideoProjection {
180  uint64_t type;
181  EbmlBin private;
182  double yaw;
183  double pitch;
184  double roll;
186 
187 typedef struct MatroskaTrackVideo {
188  double frame_rate;
189  uint64_t display_width;
190  uint64_t display_height;
191  uint64_t pixel_width;
192  uint64_t pixel_height;
194  uint64_t display_unit;
195  uint64_t interlaced;
196  uint64_t field_order;
197  uint64_t stereo_mode;
198  uint64_t alpha_mode;
202 
203 typedef struct MatroskaTrackAudio {
204  double samplerate;
206  uint64_t bitdepth;
207  uint64_t channels;
208 
209  /* real audio header (extracted from extradata) */
212  int frame_size;
213  int sub_packet_size;
215  int pkt_cnt;
216  uint64_t buf_timecode;
217  uint8_t *buf;
219 
220 typedef struct MatroskaTrackPlane {
221  uint64_t uid;
222  uint64_t type;
224 
225 typedef struct MatroskaTrackOperation {
228 
229 typedef struct MatroskaTrack {
230  uint64_t num;
231  uint64_t uid;
232  uint64_t type;
233  char *name;
234  char *codec_id;
236  char *language;
237  double time_scale;
239  uint64_t flag_default;
240  uint64_t flag_forced;
241  uint64_t seek_preroll;
246  uint64_t codec_delay;
248 
250  int64_t end_timecode;
251  int ms_compat;
252  int needs_decoding;
254 
258 
259 typedef struct MatroskaAttachment {
260  uint64_t uid;
261  char *filename;
262  char *description;
263  char *mime;
265 
268 
269 typedef struct MatroskaChapter {
270  uint64_t start;
271  uint64_t end;
272  uint64_t uid;
273  char *title;
274 
277 
278 typedef struct MatroskaIndexPos {
279  uint64_t track;
280  uint64_t pos;
282 
283 typedef struct MatroskaIndex {
284  uint64_t time;
287 
288 typedef struct MatroskaTag {
289  char *name;
290  char *string;
291  char *lang;
292  uint64_t def;
295 
296 typedef struct MatroskaTagTarget {
297  char *type;
298  uint64_t typevalue;
299  uint64_t trackuid;
300  uint64_t chapteruid;
301  uint64_t attachuid;
303 
304 typedef struct MatroskaTags {
306  EbmlList tag;
307 } MatroskaTags;
308 
309 typedef struct MatroskaSeekhead {
310  uint64_t id;
311  uint64_t pos;
313 
314 typedef struct MatroskaLevel {
315  uint64_t start;
316  uint64_t length;
318 
319 typedef struct MatroskaBlock {
320  uint64_t duration;
321  int64_t reference;
322  uint64_t non_simple;
324  uint64_t additional_id;
327 } MatroskaBlock;
328 
329 typedef struct MatroskaCluster {
331  uint64_t timecode;
332  int64_t pos;
334 
335 typedef struct MatroskaLevel1Element {
336  int64_t pos;
337  uint32_t id;
338  int parsed;
340 
341 typedef struct MatroskaDemuxContext {
342  const AVClass *class;
344 
345  /* EBML stuff */
348  uint32_t current_id;
349  int64_t resync_pos;
351 
352  uint64_t time_scale;
353  double duration;
354  char *title;
355  char *muxingapp;
360  EbmlList index;
361  EbmlList tags;
363 
364  /* byte position of the segment inside the stream */
365  int64_t segment_start;
366 
367  /* the packet queue */
370 
371  int done;
372 
373  /* What to skip before effectively reading a packet. */
374  int skip_to_keyframe;
376 
377  /* File has a CUES element, but we defer parsing until it is needed. */
379 
380  /* Level1 elements and whether they were read yet */
382  int num_level1_elems;
383 
385 
386  /* WebM DASH Manifest live flag */
387  int is_live;
388 
389  /* Bandwidth value for WebM DASH Manifest */
390  int bandwidth;
392 
393 #define CHILD_OF(parent) { .def = { .n = parent } }
394 
395 // The following forward declarations need their size because
396 // a tentative definition with internal linkage must not be an
397 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
398 // Removing the sizes breaks MSVC.
405 
406 static EbmlSyntax ebml_header[] = {
407  { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
408  { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
409  { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
410  { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
411  { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
415 };
416 
418  { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
420  { 0 }
421 };
422 
423 static EbmlSyntax matroska_info[] = {
424  { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
426  { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
428  { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
429  { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
432 };
433 
435  { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_x), { .f=-1 } },
436  { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_y), { .f=-1 } },
437  { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_x), { .f=-1 } },
438  { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_y), { .f=-1 } },
439  { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_x), { .f=-1 } },
440  { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_y), { .f=-1 } },
441  { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_x), { .f=-1 } },
442  { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_y), { .f=-1 } },
443  { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, min_luminance), { .f=-1 } },
444  { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, max_luminance), { .f=-1 } },
446 };
447 
449  { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
450  { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u=0 } },
451  { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz), { .u=0 } },
452  { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert), { .u=0 } },
453  { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz), { .u=0 } },
454  { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert), { .u=0 } },
460  { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_cll), { .u=0 } },
461  { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_fall), { .u=0 } },
464 };
465 
473 };
474 
475 static EbmlSyntax matroska_track_video[] = {
476  { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
477  { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
478  { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
479  { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
480  { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
481  { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
482  { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
495 };
496 
497 static EbmlSyntax matroska_track_audio[] = {
498  { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
500  { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
501  { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
503 };
504 
509 };
510 
520 };
522  { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
523  { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
528 };
529 
533 };
534 
535 static EbmlSyntax matroska_track_plane[] = {
539 };
540 
544 };
545 
549 };
550 
551 static EbmlSyntax matroska_track[] = {
552  { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
554  { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
557  { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
558  { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
559  { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
560  { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
561  { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
562  { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
563  { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
565  { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
566  { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
568  { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
569  { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
579 };
580 
581 static EbmlSyntax matroska_tracks[] = {
582  { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
584 };
585 
586 static EbmlSyntax matroska_attachment[] = {
588  { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
589  { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
590  { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
593 };
594 
595 static EbmlSyntax matroska_attachments[] = {
596  { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
598 };
599 
601  { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
605 };
606 
608  { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
617 };
618 
619 static EbmlSyntax matroska_chapter[] = {
626 };
627 
628 static EbmlSyntax matroska_chapters[] = {
631 };
632 
633 static EbmlSyntax matroska_index_pos[] = {
634  { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
640 };
641 
642 static EbmlSyntax matroska_index_entry[] = {
643  { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
646 };
647 
648 static EbmlSyntax matroska_index[] = {
651 };
652 
653 static EbmlSyntax matroska_simpletag[] = {
654  { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
655  { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
656  { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
658  { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
659  { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
661 };
662 
663 static EbmlSyntax matroska_tagtargets[] = {
665  { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
666  { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
667  { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
668  { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
670 };
671 
672 static EbmlSyntax matroska_tag[] = {
674  { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
676 };
677 
678 static EbmlSyntax matroska_tags[] = {
679  { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
681 };
682 
684  { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
685  { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
687 };
688 
689 static EbmlSyntax matroska_seekhead[] = {
692 };
693 
694 static EbmlSyntax matroska_segment[] = {
696  { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
697  { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
701  { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
703  { 0 } /* We don't want to go back to level 0, so don't add the parent. */
704 };
705 
706 static EbmlSyntax matroska_segments[] = {
707  { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
708  { 0 }
709 };
710 
711 static EbmlSyntax matroska_blockmore[] = {
712  { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
713  { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
715 };
716 
720 };
721 
722 static EbmlSyntax matroska_blockgroup[] = {
723  { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
726  { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
727  { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference), { .i = INT64_MIN } },
729  { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
731 };
732 
733 // The following array contains SimpleBlock and BlockGroup twice
734 // in order to reuse the other values for matroska_cluster_enter.
736  { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
738  { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
744 };
745 
747  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = &matroska_cluster_parsing[2] } },
748  { 0 }
749 };
750 #undef CHILD_OF
751 
752 static const CodecMime mkv_image_mime_tags[] = {
753  {"image/gif" , AV_CODEC_ID_GIF},
754  {"image/jpeg" , AV_CODEC_ID_MJPEG},
755  {"image/png" , AV_CODEC_ID_PNG},
756  {"image/tiff" , AV_CODEC_ID_TIFF},
757 
758  {"" , AV_CODEC_ID_NONE}
759 };
760 
761 static const CodecMime mkv_mime_tags[] = {
762  {"text/plain" , AV_CODEC_ID_TEXT},
763  {"application/x-truetype-font", AV_CODEC_ID_TTF},
764  {"application/x-font" , AV_CODEC_ID_TTF},
765  {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
766  {"binary" , AV_CODEC_ID_BIN_DATA},
767 
768  {"" , AV_CODEC_ID_NONE}
769 };
770 
771 static const char *const matroska_doctypes[] = { "matroska", "webm" };
772 
774 
775 /*
776  * This function prepares the status for parsing of level 1 elements.
777  */
778 static int matroska_reset_status(MatroskaDemuxContext *matroska,
779  uint32_t id, int64_t position)
780 {
781  int64_t err = 0;
782  if (position >= 0) {
783  err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
784  if (err > 0)
785  err = 0;
786  } else
787  position = avio_tell(matroska->ctx->pb);
788 
789  matroska->current_id = id;
790  matroska->num_levels = 1;
791  matroska->unknown_count = 0;
792  matroska->resync_pos = position;
793  if (id)
794  matroska->resync_pos -= (av_log2(id) + 7) / 8;
795 
796  return err;
797 }
798 
799 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
800 {
801  AVIOContext *pb = matroska->ctx->pb;
802  uint32_t id;
803 
804  /* Try to seek to the last position to resync from. If this doesn't work,
805  * we resync from the earliest position available: The start of the buffer. */
806  if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
807  av_log(matroska->ctx, AV_LOG_WARNING,
808  "Seek to desired resync point failed. Seeking to "
809  "earliest point available instead.\n");
810  avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
811  last_pos + 1), SEEK_SET);
812  }
813 
814  id = avio_rb32(pb);
815 
816  // try to find a toplevel element
817  while (!avio_feof(pb)) {
818  if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
819  id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
821  id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
822  /* Prepare the context for parsing of a level 1 element. */
823  matroska_reset_status(matroska, id, -1);
824  /* Given that we are here means that an error has occurred,
825  * so treat the segment as unknown length in order not to
826  * discard valid data that happens to be beyond the designated
827  * end of the segment. */
828  matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
829  return 0;
830  }
831  id = (id << 8) | avio_r8(pb);
832  }
833 
834  matroska->done = 1;
835  return pb->error ? pb->error : AVERROR_EOF;
836 }
837 
838 /*
839  * Read: an "EBML number", which is defined as a variable-length
840  * array of bytes. The first byte indicates the length by giving a
841  * number of 0-bits followed by a one. The position of the first
842  * "one" bit inside the first byte indicates the length of this
843  * number.
844  * Returns: number of bytes read, < 0 on error
845  */
846 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
847  int max_size, uint64_t *number, int eof_forbidden)
848 {
849  int read, n = 1;
850  uint64_t total;
851  int64_t pos;
852 
853  /* The first byte tells us the length in bytes - except when it is zero. */
854  total = avio_r8(pb);
855  if (pb->eof_reached)
856  goto err;
857 
858  /* get the length of the EBML number */
859  read = 8 - ff_log2_tab[total];
860 
861  if (!total || read > max_size) {
862  pos = avio_tell(pb) - 1;
863  if (!total) {
864  av_log(matroska->ctx, AV_LOG_ERROR,
865  "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
866  "of an EBML number\n", pos, pos);
867  } else {
868  av_log(matroska->ctx, AV_LOG_ERROR,
869  "Length %d indicated by an EBML number's first byte 0x%02x "
870  "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
871  read, (uint8_t) total, pos, pos, max_size);
872  }
873  return AVERROR_INVALIDDATA;
874  }
875 
876  /* read out length */
877  total ^= 1 << ff_log2_tab[total];
878  while (n++ < read)
879  total = (total << 8) | avio_r8(pb);
880 
881  if (pb->eof_reached) {
882  eof_forbidden = 1;
883  goto err;
884  }
885 
886  *number = total;
887 
888  return read;
889 
890 err:
891  pos = avio_tell(pb);
892  if (pb->error) {
893  av_log(matroska->ctx, AV_LOG_ERROR,
894  "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
895  pos, pos);
896  return pb->error;
897  }
898  if (eof_forbidden) {
899  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
900  "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
901  return AVERROR(EIO);
902  }
903  return AVERROR_EOF;
904 }
905 
906 /**
907  * Read a EBML length value.
908  * This needs special handling for the "unknown length" case which has multiple
909  * encodings.
910  */
911 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
912  uint64_t *number)
913 {
914  int res = ebml_read_num(matroska, pb, 8, number, 1);
915  if (res > 0 && *number + 1 == 1ULL << (7 * res))
916  *number = EBML_UNKNOWN_LENGTH;
917  return res;
918 }
919 
920 /*
921  * Read the next element as an unsigned int.
922  * Returns NEEDS_CHECKING.
923  */
924 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
925 {
926  int n = 0;
927 
928  /* big-endian ordering; build up number */
929  *num = 0;
930  while (n++ < size)
931  *num = (*num << 8) | avio_r8(pb);
932 
933  return NEEDS_CHECKING;
934 }
935 
936 /*
937  * Read the next element as a signed int.
938  * Returns NEEDS_CHECKING.
939  */
940 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
941 {
942  int n = 1;
943 
944  if (size == 0) {
945  *num = 0;
946  } else {
947  *num = sign_extend(avio_r8(pb), 8);
948 
949  /* big-endian ordering; build up number */
950  while (n++ < size)
951  *num = ((uint64_t)*num << 8) | avio_r8(pb);
952  }
953 
954  return NEEDS_CHECKING;
955 }
956 
957 /*
958  * Read the next element as a float.
959  * Returns NEEDS_CHECKING or < 0 on obvious failure.
960  */
961 static int ebml_read_float(AVIOContext *pb, int size, double *num)
962 {
963  if (size == 0)
964  *num = 0;
965  else if (size == 4)
966  *num = av_int2float(avio_rb32(pb));
967  else if (size == 8)
968  *num = av_int2double(avio_rb64(pb));
969  else
970  return AVERROR_INVALIDDATA;
971 
972  return NEEDS_CHECKING;
973 }
974 
975 /*
976  * Read the next element as an ASCII string.
977  * 0 is success, < 0 or NEEDS_CHECKING is failure.
978  */
979 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
980 {
981  char *res;
982  int ret;
983 
984  /* EBML strings are usually not 0-terminated, so we allocate one
985  * byte more, read the string and NULL-terminate it ourselves. */
986  if (!(res = av_malloc(size + 1)))
987  return AVERROR(ENOMEM);
988  if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
989  av_free(res);
990  return ret < 0 ? ret : NEEDS_CHECKING;
991  }
992  (res)[size] = '\0';
993  av_free(*str);
994  *str = res;
995 
996  return 0;
997 }
998 
999 /*
1000  * Read the next element as binary data.
1001  * 0 is success, < 0 or NEEDS_CHECKING is failure.
1002  */
1003 static int ebml_read_binary(AVIOContext *pb, int length,
1004  int64_t pos, EbmlBin *bin)
1005 {
1006  int ret;
1007 
1009  if (ret < 0)
1010  return ret;
1011  memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1012 
1013  bin->data = bin->buf->data;
1014  bin->size = length;
1015  bin->pos = pos;
1016  if ((ret = avio_read(pb, bin->data, length)) != length) {
1017  av_buffer_unref(&bin->buf);
1018  bin->data = NULL;
1019  bin->size = 0;
1020  return ret < 0 ? ret : NEEDS_CHECKING;
1021  }
1022 
1023  return 0;
1024 }
1026 /*
1027  * Read the next element, but only the header. The contents
1028  * are supposed to be sub-elements which can be read separately.
1029  * 0 is success, < 0 is failure.
1030  */
1031 static int ebml_read_master(MatroskaDemuxContext *matroska,
1032  uint64_t length, int64_t pos)
1033 {
1035 
1036  if (matroska->num_levels >= EBML_MAX_DEPTH) {
1037  av_log(matroska->ctx, AV_LOG_ERROR,
1038  "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1039  return AVERROR(ENOSYS);
1040  }
1041 
1042  level = &matroska->levels[matroska->num_levels++];
1043  level->start = pos;
1044  level->length = length;
1045 
1046  return 0;
1048 
1049 /*
1050  * Read a signed "EBML number"
1051  * Return: number of bytes processed, < 0 on error
1052  */
1053 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1054  AVIOContext *pb, int64_t *num)
1055 {
1056  uint64_t unum;
1057  int res;
1058 
1059  /* read as unsigned number first */
1060  if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1061  return res;
1062 
1063  /* make signed (weird way) */
1064  *num = unum - ((1LL << (7 * res - 1)) - 1);
1065 
1066  return res;
1067 }
1068 
1069 static int ebml_parse(MatroskaDemuxContext *matroska,
1070  EbmlSyntax *syntax, void *data);
1071 
1072 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1073 {
1074  int i;
1075 
1076  // Whoever touches this should be aware of the duplication
1077  // existing in matroska_cluster_parsing.
1078  for (i = 0; syntax[i].id; i++)
1079  if (id == syntax[i].id)
1080  break;
1081 
1082  return &syntax[i];
1083 }
1084 
1085 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1086  void *data)
1087 {
1088  int res;
1089 
1090  if (data) {
1091  for (int i = 0; syntax[i].id; i++)
1092  switch (syntax[i].type) {
1093  case EBML_UINT:
1094  *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1095  break;
1096  case EBML_SINT:
1097  *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1098  break;
1099  case EBML_FLOAT:
1100  *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1101  break;
1102  case EBML_STR:
1103  case EBML_UTF8:
1104  // the default may be NULL
1105  if (syntax[i].def.s) {
1106  uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1107  *dst = av_strdup(syntax[i].def.s);
1108  if (!*dst)
1109  return AVERROR(ENOMEM);
1110  }
1111  break;
1112  }
1113 
1114  if (!matroska->levels[matroska->num_levels - 1].length) {
1115  matroska->num_levels--;
1116  return 0;
1117  }
1118  }
1119 
1120  do {
1121  res = ebml_parse(matroska, syntax, data);
1122  } while (!res);
1123 
1124  return res == LEVEL_ENDED ? 0 : res;
1125 }
1126 
1127 static int is_ebml_id_valid(uint32_t id)
1128 {
1129  // Due to endian nonsense in Matroska, the highest byte with any bits set
1130  // will contain the leading length bit. This bit in turn identifies the
1131  // total byte length of the element by its position within the byte.
1132  unsigned int bits = av_log2(id);
1133  return id && (bits + 7) / 8 == (8 - bits % 8);
1135 
1136 /*
1137  * Allocate and return the entry for the level1 element with the given ID. If
1138  * an entry already exists, return the existing entry.
1139  */
1141  uint32_t id, int64_t pos)
1142 {
1143  int i;
1144  MatroskaLevel1Element *elem;
1145 
1146  if (!is_ebml_id_valid(id))
1147  return NULL;
1148 
1149  // Some files link to all clusters; useless.
1150  if (id == MATROSKA_ID_CLUSTER)
1151  return NULL;
1152 
1153  // There can be multiple SeekHeads and Tags.
1154  for (i = 0; i < matroska->num_level1_elems; i++) {
1155  if (matroska->level1_elems[i].id == id) {
1156  if (matroska->level1_elems[i].pos == pos ||
1157  id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1158  return &matroska->level1_elems[i];
1159  }
1160  }
1161 
1162  // Only a completely broken file would have more elements.
1163  if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1164  av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1165  return NULL;
1166  }
1167 
1168  elem = &matroska->level1_elems[matroska->num_level1_elems++];
1169  *elem = (MatroskaLevel1Element){.id = id};
1170 
1171  return elem;
1172 }
1173 
1174 static int ebml_parse(MatroskaDemuxContext *matroska,
1175  EbmlSyntax *syntax, void *data)
1176 {
1177  static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1178  // Forbid unknown-length EBML_NONE elements.
1180  [EBML_UINT] = 8,
1181  [EBML_SINT] = 8,
1182  [EBML_FLOAT] = 8,
1183  // max. 16 MB for strings
1184  [EBML_STR] = 0x1000000,
1185  [EBML_UTF8] = 0x1000000,
1186  // max. 256 MB for binary data
1187  [EBML_BIN] = 0x10000000,
1188  // no limits for anything else
1189  };
1190  AVIOContext *pb = matroska->ctx->pb;
1191  uint32_t id;
1192  uint64_t length;
1193  int64_t pos = avio_tell(pb), pos_alt;
1194  int res, update_pos = 1, level_check;
1195  MatroskaLevel1Element *level1_elem;
1196  MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1197 
1198  if (!matroska->current_id) {
1199  uint64_t id;
1200  res = ebml_read_num(matroska, pb, 4, &id, 0);
1201  if (res < 0) {
1202  if (pb->eof_reached && res == AVERROR_EOF) {
1203  if (matroska->is_live)
1204  // in live mode, finish parsing if EOF is reached.
1205  return 1;
1206  if (level && pos == avio_tell(pb)) {
1207  if (level->length == EBML_UNKNOWN_LENGTH) {
1208  // Unknown-length levels automatically end at EOF.
1209  matroska->num_levels--;
1210  return LEVEL_ENDED;
1211  } else {
1212  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1213  "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1214  }
1215  }
1216  }
1217  return res;
1218  }
1219  matroska->current_id = id | 1 << 7 * res;
1220  pos_alt = pos + res;
1221  } else {
1222  pos_alt = pos;
1223  pos -= (av_log2(matroska->current_id) + 7) / 8;
1224  }
1225 
1226  id = matroska->current_id;
1227 
1228  syntax = ebml_parse_id(syntax, id);
1229  if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1230  if (level && level->length == EBML_UNKNOWN_LENGTH) {
1231  // Unknown-length levels end when an element from an upper level
1232  // in the hierarchy is encountered.
1233  while (syntax->def.n) {
1234  syntax = ebml_parse_id(syntax->def.n, id);
1235  if (syntax->id) {
1236  matroska->num_levels--;
1237  return LEVEL_ENDED;
1238  }
1239  };
1240  }
1241 
1242  av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1243  "%"PRId64"\n", id, pos);
1244  update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1245  }
1246 
1247  if (data) {
1248  data = (char *) data + syntax->data_offset;
1249  if (syntax->list_elem_size) {
1250  EbmlList *list = data;
1251  void *newelem;
1252 
1253  if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1254  return AVERROR(ENOMEM);
1255  newelem = av_fast_realloc(list->elem,
1256  &list->alloc_elem_size,
1257  (list->nb_elem + 1) * syntax->list_elem_size);
1258  if (!newelem)
1259  return AVERROR(ENOMEM);
1260  list->elem = newelem;
1261  data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1262  memset(data, 0, syntax->list_elem_size);
1263  list->nb_elem++;
1264  }
1265  }
1266 
1267  if (syntax->type != EBML_STOP) {
1268  matroska->current_id = 0;
1269  if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1270  return res;
1271 
1272  pos_alt += res;
1273 
1274  if (matroska->num_levels > 0) {
1275  if (length != EBML_UNKNOWN_LENGTH &&
1276  level->length != EBML_UNKNOWN_LENGTH) {
1277  uint64_t elem_end = pos_alt + length,
1278  level_end = level->start + level->length;
1279 
1280  if (elem_end < level_end) {
1281  level_check = 0;
1282  } else if (elem_end == level_end) {
1283  level_check = LEVEL_ENDED;
1284  } else {
1285  av_log(matroska->ctx, AV_LOG_ERROR,
1286  "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1287  "containing master element ending at 0x%"PRIx64"\n",
1288  pos, elem_end, level_end);
1289  return AVERROR_INVALIDDATA;
1290  }
1291  } else if (length != EBML_UNKNOWN_LENGTH) {
1292  level_check = 0;
1293  } else if (level->length != EBML_UNKNOWN_LENGTH) {
1294  av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1295  "at 0x%"PRIx64" inside parent with finite size\n", pos);
1296  return AVERROR_INVALIDDATA;
1297  } else {
1298  level_check = 0;
1299  if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1300  || syntax->type == EBML_NEST)) {
1301  // According to the current specifications only clusters and
1302  // segments are allowed to be unknown-length. We also accept
1303  // other unknown-length master elements.
1304  av_log(matroska->ctx, AV_LOG_WARNING,
1305  "Found unknown-length element 0x%"PRIX32" other than "
1306  "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1307  "parsing will nevertheless be attempted.\n", id, pos);
1308  update_pos = -1;
1309  }
1310  }
1311  } else
1312  level_check = 0;
1313 
1314  if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1315  if (length != EBML_UNKNOWN_LENGTH) {
1316  av_log(matroska->ctx, AV_LOG_ERROR,
1317  "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1318  "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1319  length, max_lengths[syntax->type], id, pos);
1320  } else if (syntax->type != EBML_NONE) {
1321  av_log(matroska->ctx, AV_LOG_ERROR,
1322  "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1323  "unknown length, yet the length of an element of its "
1324  "type must be known.\n", id, pos);
1325  } else {
1326  av_log(matroska->ctx, AV_LOG_ERROR,
1327  "Found unknown-length element with ID 0x%"PRIX32" at "
1328  "pos. 0x%"PRIx64" for which no syntax for parsing is "
1329  "available.\n", id, pos);
1330  }
1331  return AVERROR_INVALIDDATA;
1332  }
1333 
1334  if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1335  // Loosing sync will likely manifest itself as encountering unknown
1336  // elements which are not reliably distinguishable from elements
1337  // belonging to future extensions of the format.
1338  // We use a heuristic to detect such situations: If the current
1339  // element is not expected at the current syntax level and there
1340  // were only a few unknown elements in a row, then the element is
1341  // skipped or considered defective based upon the length of the
1342  // current element (i.e. how much would be skipped); if there were
1343  // more than a few skipped elements in a row and skipping the current
1344  // element would lead us more than SKIP_THRESHOLD away from the last
1345  // known good position, then it is inferred that an error occurred.
1346  // The dependency on the number of unknown elements in a row exists
1347  // because the distance to the last known good position is
1348  // automatically big if the last parsed element was big.
1349  // In both cases, each unknown element is considered equivalent to
1350  // UNKNOWN_EQUIV of skipped bytes for the check.
1351  // The whole check is only done for non-seekable output, because
1352  // in this situation skipped data can't simply be rechecked later.
1353  // This is especially important when using unkown length elements
1354  // as the check for whether a child exceeds its containing master
1355  // element is not effective in this situation.
1356  if (update_pos) {
1357  matroska->unknown_count = 0;
1358  } else {
1359  int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1360 
1361  if (matroska->unknown_count > 3)
1362  dist += pos_alt - matroska->resync_pos;
1363 
1364  if (dist > SKIP_THRESHOLD) {
1365  av_log(matroska->ctx, AV_LOG_ERROR,
1366  "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1367  "length 0x%"PRIx64" considered as invalid data. Last "
1368  "known good position 0x%"PRIx64", %d unknown elements"
1369  " in a row\n", id, pos, length, matroska->resync_pos,
1370  matroska->unknown_count);
1371  return AVERROR_INVALIDDATA;
1372  }
1373  }
1374  }
1375 
1376  if (update_pos > 0) {
1377  // We have found an element that is allowed at this place
1378  // in the hierarchy and it passed all checks, so treat the beginning
1379  // of the element as the "last known good" position.
1380  matroska->resync_pos = pos;
1381  }
1382 
1383  if (!data && length != EBML_UNKNOWN_LENGTH)
1384  goto skip;
1385  }
1386 
1387  switch (syntax->type) {
1388  case EBML_UINT:
1389  res = ebml_read_uint(pb, length, data);
1390  break;
1391  case EBML_SINT:
1392  res = ebml_read_sint(pb, length, data);
1393  break;
1394  case EBML_FLOAT:
1395  res = ebml_read_float(pb, length, data);
1396  break;
1397  case EBML_STR:
1398  case EBML_UTF8:
1399  res = ebml_read_ascii(pb, length, data);
1400  break;
1401  case EBML_BIN:
1402  res = ebml_read_binary(pb, length, pos_alt, data);
1403  break;
1404  case EBML_LEVEL1:
1405  case EBML_NEST:
1406  if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1407  return res;
1408  if (id == MATROSKA_ID_SEGMENT)
1409  matroska->segment_start = pos_alt;
1410  if (id == MATROSKA_ID_CUES)
1411  matroska->cues_parsing_deferred = 0;
1412  if (syntax->type == EBML_LEVEL1 &&
1413  (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1414  if (!level1_elem->pos) {
1415  // Zero is not a valid position for a level 1 element.
1416  level1_elem->pos = pos;
1417  } else if (level1_elem->pos != pos)
1418  av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1419  level1_elem->parsed = 1;
1420  }
1421  if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1422  return res;
1423  break;
1424  case EBML_STOP:
1425  return 1;
1426  skip:
1427  default:
1428  if (length) {
1429  int64_t res2;
1430  if (ffio_limit(pb, length) != length) {
1431  // ffio_limit emits its own error message,
1432  // so we don't have to.
1433  return AVERROR(EIO);
1434  }
1435  if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1436  // avio_skip might take us past EOF. We check for this
1437  // by skipping only length - 1 bytes, reading a byte and
1438  // checking the error flags. This is done in order to check
1439  // that the element has been properly skipped even when
1440  // no filesize (that ffio_limit relies on) is available.
1441  avio_r8(pb);
1442  res = NEEDS_CHECKING;
1443  } else
1444  res = res2;
1445  } else
1446  res = 0;
1447  }
1448  if (res) {
1449  if (res == NEEDS_CHECKING) {
1450  if (pb->eof_reached) {
1451  if (pb->error)
1452  res = pb->error;
1453  else
1454  res = AVERROR_EOF;
1455  } else
1456  goto level_check;
1457  }
1458 
1459  if (res == AVERROR_INVALIDDATA)
1460  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1461  else if (res == AVERROR(EIO))
1462  av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1463  else if (res == AVERROR_EOF) {
1464  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1465  res = AVERROR(EIO);
1466  }
1467 
1468  return res;
1469  }
1470 
1471 level_check:
1472  if (level_check == LEVEL_ENDED && matroska->num_levels) {
1473  level = &matroska->levels[matroska->num_levels - 1];
1474  pos = avio_tell(pb);
1475 
1476  // Given that pos >= level->start no check for
1477  // level->length != EBML_UNKNOWN_LENGTH is necessary.
1478  while (matroska->num_levels && pos == level->start + level->length) {
1479  matroska->num_levels--;
1480  level--;
1481  }
1482  }
1483 
1484  return level_check;
1485 }
1486 
1487 static void ebml_free(EbmlSyntax *syntax, void *data)
1488 {
1489  int i, j;
1490  for (i = 0; syntax[i].id; i++) {
1491  void *data_off = (char *) data + syntax[i].data_offset;
1492  switch (syntax[i].type) {
1493  case EBML_STR:
1494  case EBML_UTF8:
1495  av_freep(data_off);
1496  break;
1497  case EBML_BIN:
1498  av_buffer_unref(&((EbmlBin *) data_off)->buf);
1499  break;
1500  case EBML_LEVEL1:
1501  case EBML_NEST:
1502  if (syntax[i].list_elem_size) {
1503  EbmlList *list = data_off;
1504  char *ptr = list->elem;
1505  for (j = 0; j < list->nb_elem;
1506  j++, ptr += syntax[i].list_elem_size)
1507  ebml_free(syntax[i].def.n, ptr);
1508  av_freep(&list->elem);
1509  list->nb_elem = 0;
1510  list->alloc_elem_size = 0;
1511  } else
1512  ebml_free(syntax[i].def.n, data_off);
1513  default:
1514  break;
1515  }
1516  }
1517 }
1518 
1519 /*
1520  * Autodetecting...
1521  */
1522 static int matroska_probe(const AVProbeData *p)
1523 {
1524  uint64_t total = 0;
1525  int len_mask = 0x80, size = 1, n = 1, i;
1526 
1527  /* EBML header? */
1528  if (AV_RB32(p->buf) != EBML_ID_HEADER)
1529  return 0;
1530 
1531  /* length of header */
1532  total = p->buf[4];
1533  while (size <= 8 && !(total & len_mask)) {
1534  size++;
1535  len_mask >>= 1;
1536  }
1537  if (size > 8)
1538  return 0;
1539  total &= (len_mask - 1);
1540  while (n < size)
1541  total = (total << 8) | p->buf[4 + n++];
1542 
1543  if (total + 1 == 1ULL << (7 * size)){
1544  /* Unknown-length header - simply parse the whole buffer. */
1545  total = p->buf_size - 4 - size;
1546  } else {
1547  /* Does the probe data contain the whole header? */
1548  if (p->buf_size < 4 + size + total)
1549  return 0;
1550  }
1551 
1552  /* The header should contain a known document type. For now,
1553  * we don't parse the whole header but simply check for the
1554  * availability of that array of characters inside the header.
1555  * Not fully fool-proof, but good enough. */
1556  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1557  size_t probelen = strlen(matroska_doctypes[i]);
1558  if (total < probelen)
1559  continue;
1560  for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1561  if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1562  return AVPROBE_SCORE_MAX;
1563  }
1564 
1565  // probably valid EBML header but no recognized doctype
1566  return AVPROBE_SCORE_EXTENSION;
1567 }
1568 
1570  uint64_t num)
1571 {
1572  MatroskaTrack *tracks = matroska->tracks.elem;
1573  int i;
1574 
1575  for (i = 0; i < matroska->tracks.nb_elem; i++)
1576  if (tracks[i].num == num)
1577  return &tracks[i];
1578 
1579  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1580  return NULL;
1581 }
1582 
1583 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1584  MatroskaTrack *track)
1585 {
1586  MatroskaTrackEncoding *encodings = track->encodings.elem;
1587  uint8_t *data = *buf;
1588  int isize = *buf_size;
1589  uint8_t *pkt_data = NULL;
1590  uint8_t av_unused *newpktdata;
1591  int pkt_size = isize;
1592  int result = 0;
1593  int olen;
1594 
1595  if (pkt_size >= 10000000U)
1596  return AVERROR_INVALIDDATA;
1597 
1598  switch (encodings[0].compression.algo) {
1600  {
1601  int header_size = encodings[0].compression.settings.size;
1602  uint8_t *header = encodings[0].compression.settings.data;
1603 
1604  if (header_size && !header) {
1605  av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1606  return -1;
1607  }
1608 
1609  if (!header_size)
1610  return 0;
1611 
1612  pkt_size = isize + header_size;
1613  pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1614  if (!pkt_data)
1615  return AVERROR(ENOMEM);
1616 
1617  memcpy(pkt_data, header, header_size);
1618  memcpy(pkt_data + header_size, data, isize);
1619  break;
1620  }
1621 #if CONFIG_LZO
1623  do {
1624  int insize = isize;
1625  olen = pkt_size *= 3;
1626  newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1628  if (!newpktdata) {
1629  result = AVERROR(ENOMEM);
1630  goto failed;
1631  }
1632  pkt_data = newpktdata;
1633  result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1634  } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1635  if (result) {
1637  goto failed;
1638  }
1639  pkt_size -= olen;
1640  break;
1641 #endif
1642 #if CONFIG_ZLIB
1644  {
1645  z_stream zstream = { 0 };
1646  if (!pkt_size || inflateInit(&zstream) != Z_OK)
1647  return -1;
1648  zstream.next_in = data;
1649  zstream.avail_in = isize;
1650  do {
1651  pkt_size *= 3;
1652  newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1653  if (!newpktdata) {
1654  inflateEnd(&zstream);
1655  result = AVERROR(ENOMEM);
1656  goto failed;
1657  }
1658  pkt_data = newpktdata;
1659  zstream.avail_out = pkt_size - zstream.total_out;
1660  zstream.next_out = pkt_data + zstream.total_out;
1661  result = inflate(&zstream, Z_NO_FLUSH);
1662  } while (result == Z_OK && pkt_size < 10000000);
1663  pkt_size = zstream.total_out;
1664  inflateEnd(&zstream);
1665  if (result != Z_STREAM_END) {
1666  if (result == Z_MEM_ERROR)
1667  result = AVERROR(ENOMEM);
1668  else
1670  goto failed;
1671  }
1672  break;
1673  }
1674 #endif
1675 #if CONFIG_BZLIB
1677  {
1678  bz_stream bzstream = { 0 };
1679  if (!pkt_size || BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1680  return -1;
1681  bzstream.next_in = data;
1682  bzstream.avail_in = isize;
1683  do {
1684  pkt_size *= 3;
1685  newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1686  if (!newpktdata) {
1687  BZ2_bzDecompressEnd(&bzstream);
1688  result = AVERROR(ENOMEM);
1689  goto failed;
1690  }
1691  pkt_data = newpktdata;
1692  bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1693  bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1694  result = BZ2_bzDecompress(&bzstream);
1695  } while (result == BZ_OK && pkt_size < 10000000);
1696  pkt_size = bzstream.total_out_lo32;
1697  BZ2_bzDecompressEnd(&bzstream);
1698  if (result != BZ_STREAM_END) {
1699  if (result == BZ_MEM_ERROR)
1700  result = AVERROR(ENOMEM);
1701  else
1703  goto failed;
1704  }
1705  break;
1706  }
1707 #endif
1708  default:
1709  return AVERROR_INVALIDDATA;
1710  }
1711 
1712  memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1713 
1714  *buf = pkt_data;
1715  *buf_size = pkt_size;
1716  return 0;
1718 failed:
1719  av_free(pkt_data);
1720  return result;
1721 }
1722 
1724  AVDictionary **metadata, char *prefix)
1725 {
1726  MatroskaTag *tags = list->elem;
1727  char key[1024];
1728  int i;
1729 
1730  for (i = 0; i < list->nb_elem; i++) {
1731  const char *lang = tags[i].lang &&
1732  strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1733 
1734  if (!tags[i].name) {
1735  av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1736  continue;
1737  }
1738  if (prefix)
1739  snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1740  else
1741  av_strlcpy(key, tags[i].name, sizeof(key));
1742  if (tags[i].def || !lang) {
1743  av_dict_set(metadata, key, tags[i].string, 0);
1744  if (tags[i].sub.nb_elem)
1745  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1746  }
1747  if (lang) {
1748  av_strlcat(key, "-", sizeof(key));
1749  av_strlcat(key, lang, sizeof(key));
1750  av_dict_set(metadata, key, tags[i].string, 0);
1751  if (tags[i].sub.nb_elem)
1752  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1753  }
1754  }
1756 }
1757 
1759 {
1760  MatroskaDemuxContext *matroska = s->priv_data;
1761  MatroskaTags *tags = matroska->tags.elem;
1762  int i, j;
1763 
1764  for (i = 0; i < matroska->tags.nb_elem; i++) {
1765  if (tags[i].target.attachuid) {
1766  MatroskaAttachment *attachment = matroska->attachments.elem;
1767  int found = 0;
1768  for (j = 0; j < matroska->attachments.nb_elem; j++) {
1769  if (attachment[j].uid == tags[i].target.attachuid &&
1770  attachment[j].stream) {
1771  matroska_convert_tag(s, &tags[i].tag,
1772  &attachment[j].stream->metadata, NULL);
1773  found = 1;
1774  }
1775  }
1776  if (!found) {
1778  "The tags at index %d refer to a "
1779  "non-existent attachment %"PRId64".\n",
1780  i, tags[i].target.attachuid);
1781  }
1782  } else if (tags[i].target.chapteruid) {
1783  MatroskaChapter *chapter = matroska->chapters.elem;
1784  int found = 0;
1785  for (j = 0; j < matroska->chapters.nb_elem; j++) {
1786  if (chapter[j].uid == tags[i].target.chapteruid &&
1787  chapter[j].chapter) {
1788  matroska_convert_tag(s, &tags[i].tag,
1789  &chapter[j].chapter->metadata, NULL);
1790  found = 1;
1791  }
1792  }
1793  if (!found) {
1795  "The tags at index %d refer to a non-existent chapter "
1796  "%"PRId64".\n",
1797  i, tags[i].target.chapteruid);
1798  }
1799  } else if (tags[i].target.trackuid) {
1800  MatroskaTrack *track = matroska->tracks.elem;
1801  int found = 0;
1802  for (j = 0; j < matroska->tracks.nb_elem; j++) {
1803  if (track[j].uid == tags[i].target.trackuid &&
1804  track[j].stream) {
1805  matroska_convert_tag(s, &tags[i].tag,
1806  &track[j].stream->metadata, NULL);
1807  found = 1;
1808  }
1809  }
1810  if (!found) {
1812  "The tags at index %d refer to a non-existent track "
1813  "%"PRId64".\n",
1814  i, tags[i].target.trackuid);
1815  }
1816  } else {
1817  matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1818  tags[i].target.type);
1819  }
1820  }
1821 }
1822 
1824  int64_t pos)
1825 {
1826  uint32_t saved_id = matroska->current_id;
1827  int64_t before_pos = avio_tell(matroska->ctx->pb);
1828  int ret = 0;
1829  int ret2;
1830 
1831  /* seek */
1832  if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1833  /* We don't want to lose our seekhead level, so we add
1834  * a dummy. This is a crude hack. */
1835  if (matroska->num_levels == EBML_MAX_DEPTH) {
1836  av_log(matroska->ctx, AV_LOG_INFO,
1837  "Max EBML element depth (%d) reached, "
1838  "cannot parse further.\n", EBML_MAX_DEPTH);
1840  } else {
1841  matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1842  matroska->num_levels++;
1843  matroska->current_id = 0;
1844 
1845  ret = ebml_parse(matroska, matroska_segment, matroska);
1846  if (ret == LEVEL_ENDED) {
1847  /* This can only happen if the seek brought us beyond EOF. */
1848  ret = AVERROR_EOF;
1849  }
1850  }
1851  }
1852  /* Seek back - notice that in all instances where this is used
1853  * it is safe to set the level to 1. */
1854  ret2 = matroska_reset_status(matroska, saved_id, before_pos);
1855  if (ret >= 0)
1856  ret = ret2;
1857 
1858  return ret;
1859 }
1860 
1861 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1862 {
1863  EbmlList *seekhead_list = &matroska->seekhead;
1864  int i;
1865 
1866  // we should not do any seeking in the streaming case
1867  if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1868  return;
1869 
1870  for (i = 0; i < seekhead_list->nb_elem; i++) {
1871  MatroskaSeekhead *seekheads = seekhead_list->elem;
1872  uint32_t id = seekheads[i].id;
1873  int64_t pos = seekheads[i].pos + matroska->segment_start;
1874  MatroskaLevel1Element *elem;
1875 
1876  if (id != seekheads[i].id || pos < matroska->segment_start)
1877  continue;
1878 
1879  elem = matroska_find_level1_elem(matroska, id, pos);
1880  if (!elem || elem->parsed)
1881  continue;
1882 
1883  elem->pos = pos;
1884 
1885  // defer cues parsing until we actually need cue data.
1886  if (id == MATROSKA_ID_CUES)
1887  continue;
1888 
1889  if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1890  // mark index as broken
1891  matroska->cues_parsing_deferred = -1;
1892  break;
1893  }
1894 
1895  elem->parsed = 1;
1896  }
1897 }
1898 
1899 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1900 {
1901  EbmlList *index_list;
1903  uint64_t index_scale = 1;
1904  int i, j;
1905 
1906  if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1907  return;
1908 
1909  index_list = &matroska->index;
1910  index = index_list->elem;
1911  if (index_list->nb_elem < 2)
1912  return;
1913  if (index[1].time > 1E14 / matroska->time_scale) {
1914  av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1915  return;
1916  }
1917  for (i = 0; i < index_list->nb_elem; i++) {
1918  EbmlList *pos_list = &index[i].pos;
1919  MatroskaIndexPos *pos = pos_list->elem;
1920  for (j = 0; j < pos_list->nb_elem; j++) {
1921  MatroskaTrack *track = matroska_find_track_by_num(matroska,
1922  pos[j].track);
1923  if (track && track->stream)
1924  av_add_index_entry(track->stream,
1925  pos[j].pos + matroska->segment_start,
1926  index[i].time / index_scale, 0, 0,
1928  }
1929  }
1930 }
1931 
1932 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1933  int i;
1934 
1935  if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1936  return;
1937 
1938  for (i = 0; i < matroska->num_level1_elems; i++) {
1939  MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1940  if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1941  if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1942  matroska->cues_parsing_deferred = -1;
1943  elem->parsed = 1;
1944  break;
1945  }
1946  }
1947 
1948  matroska_add_index_entries(matroska);
1949 }
1950 
1951 static int matroska_aac_profile(char *codec_id)
1952 {
1953  static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1954  int profile;
1955 
1957  if (strstr(codec_id, aac_profiles[profile]))
1958  break;
1959  return profile + 1;
1960 }
1961 
1962 static int matroska_aac_sri(int samplerate)
1963 {
1964  int sri;
1965 
1966  for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1967  if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1968  break;
1969  return sri;
1970 }
1971 
1972 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1973 {
1974  /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1975  avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1976 }
1977 
1979  MatroskaTrack *track,
1980  int *offset)
1981 {
1982  AVStream *st = track->stream;
1983  uint8_t *p = track->codec_priv.data;
1984  int size = track->codec_priv.size;
1985 
1986  if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1987  av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1988  track->codec_priv.size = 0;
1989  return 0;
1990  }
1991  *offset = 8;
1992  track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1993 
1994  p += track->codec_priv.size;
1995  size -= track->codec_priv.size;
1996 
1997  /* parse the remaining metadata blocks if present */
1998  while (size >= 4) {
1999  int block_last, block_type, block_size;
2000 
2001  flac_parse_block_header(p, &block_last, &block_type, &block_size);
2002 
2003  p += 4;
2004  size -= 4;
2005  if (block_size > size)
2006  return 0;
2007 
2008  /* check for the channel mask */
2009  if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2010  AVDictionary *dict = NULL;
2011  AVDictionaryEntry *chmask;
2012 
2013  ff_vorbis_comment(s, &dict, p, block_size, 0);
2014  chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2015  if (chmask) {
2016  uint64_t mask = strtol(chmask->value, NULL, 0);
2017  if (!mask || mask & ~0x3ffffULL) {
2019  "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2020  } else
2021  st->codecpar->channel_layout = mask;
2022  }
2023  av_dict_free(&dict);
2024  }
2025 
2026  p += block_size;
2027  size -= block_size;
2028  }
2029 
2030  return 0;
2031 }
2032 
2033 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2034 {
2035  int major, minor, micro, bttb = 0;
2036 
2037  /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2038  * this function, and fixed in 57.52 */
2039  if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf%d.%d.%d", &major, &minor, &micro) == 3)
2040  bttb = (major == 57 && minor >= 36 && minor <= 51 && micro >= 100);
2041 
2042  switch (field_order) {
2044  return AV_FIELD_PROGRESSIVE;
2046  return AV_FIELD_UNKNOWN;
2048  return AV_FIELD_TT;
2050  return AV_FIELD_BB;
2052  return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2054  return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2055  default:
2056  return AV_FIELD_UNKNOWN;
2057  }
2058 }
2059 
2060 static void mkv_stereo_mode_display_mul(int stereo_mode,
2061  int *h_width, int *h_height)
2062 {
2063  switch (stereo_mode) {
2069  break;
2074  *h_width = 2;
2075  break;
2080  *h_height = 2;
2081  break;
2082  }
2083 }
2084 
2085 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2086  const MatroskaTrackVideoColor *color = track->video.color.elem;
2087  const MatroskaMasteringMeta *mastering_meta;
2088  int has_mastering_primaries, has_mastering_luminance;
2089 
2090  if (!track->video.color.nb_elem)
2091  return 0;
2092 
2093  mastering_meta = &color->mastering_meta;
2094  // Mastering primaries are CIE 1931 coords, and must be > 0.
2095  has_mastering_primaries =
2096  mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2097  mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2098  mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2099  mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2100  has_mastering_luminance = mastering_meta->max_luminance > 0;
2101 
2102  if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2103  st->codecpar->color_space = color->matrix_coefficients;
2104  if (color->primaries != AVCOL_PRI_RESERVED &&
2105  color->primaries != AVCOL_PRI_RESERVED0)
2106  st->codecpar->color_primaries = color->primaries;
2107  if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2108  color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2109  st->codecpar->color_trc = color->transfer_characteristics;
2110  if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2111  color->range <= AVCOL_RANGE_JPEG)
2112  st->codecpar->color_range = color->range;
2113  if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2114  color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2115  color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2116  color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2117  st->codecpar->chroma_location =
2118  avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2119  (color->chroma_siting_vert - 1) << 7);
2120  }
2121  if (color->max_cll && color->max_fall) {
2122  size_t size = 0;
2123  int ret;
2125  if (!metadata)
2126  return AVERROR(ENOMEM);
2128  (uint8_t *)metadata, size);
2129  if (ret < 0) {
2130  av_freep(&metadata);
2131  return ret;
2132  }
2133  metadata->MaxCLL = color->max_cll;
2134  metadata->MaxFALL = color->max_fall;
2135  }
2136 
2137  if (has_mastering_primaries || has_mastering_luminance) {
2138  AVMasteringDisplayMetadata *metadata =
2141  sizeof(AVMasteringDisplayMetadata));
2142  if (!metadata) {
2143  return AVERROR(ENOMEM);
2144  }
2145  memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2146  if (has_mastering_primaries) {
2147  metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2148  metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2149  metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2150  metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2151  metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2152  metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2153  metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2154  metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2155  metadata->has_primaries = 1;
2156  }
2157  if (has_mastering_luminance) {
2158  metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2159  metadata->min_luminance = av_d2q(mastering_meta->min_luminance, INT_MAX);
2160  metadata->has_luminance = 1;
2161  }
2162  }
2163  return 0;
2164 }
2165 
2166 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2167  void *logctx)
2168 {
2169  AVSphericalMapping *spherical;
2170  const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2171  const uint8_t *priv_data = mkv_projection->private.data;
2172  enum AVSphericalProjection projection;
2173  size_t spherical_size;
2174  uint32_t l = 0, t = 0, r = 0, b = 0;
2175  uint32_t padding = 0;
2176  int ret;
2177 
2178  if (mkv_projection->private.size && priv_data[0] != 0) {
2179  av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2180  return 0;
2181  }
2182 
2183  switch (track->video.projection.type) {
2185  if (track->video.projection.private.size == 20) {
2186  t = AV_RB32(priv_data + 4);
2187  b = AV_RB32(priv_data + 8);
2188  l = AV_RB32(priv_data + 12);
2189  r = AV_RB32(priv_data + 16);
2190 
2191  if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2192  av_log(logctx, AV_LOG_ERROR,
2193  "Invalid bounding rectangle coordinates "
2194  "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2195  l, t, r, b);
2196  return AVERROR_INVALIDDATA;
2197  }
2198  } else if (track->video.projection.private.size != 0) {
2199  av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2200  return AVERROR_INVALIDDATA;
2201  }
2202 
2203  if (l || t || r || b)
2204  projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2205  else
2206  projection = AV_SPHERICAL_EQUIRECTANGULAR;
2207  break;
2209  if (track->video.projection.private.size < 4) {
2210  av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2211  return AVERROR_INVALIDDATA;
2212  } else if (track->video.projection.private.size == 12) {
2213  uint32_t layout = AV_RB32(priv_data + 4);
2214  if (layout) {
2215  av_log(logctx, AV_LOG_WARNING,
2216  "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2217  return 0;
2218  }
2219  projection = AV_SPHERICAL_CUBEMAP;
2220  padding = AV_RB32(priv_data + 8);
2221  } else {
2222  av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2223  return AVERROR_INVALIDDATA;
2224  }
2225  break;
2227  /* No Spherical metadata */
2228  return 0;
2229  default:
2230  av_log(logctx, AV_LOG_WARNING,
2231  "Unknown spherical metadata type %"PRIu64"\n",
2232  track->video.projection.type);
2233  return 0;
2234  }
2235 
2236  spherical = av_spherical_alloc(&spherical_size);
2237  if (!spherical)
2238  return AVERROR(ENOMEM);
2239 
2240  spherical->projection = projection;
2241 
2242  spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2243  spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2244  spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2245 
2246  spherical->padding = padding;
2247 
2248  spherical->bound_left = l;
2249  spherical->bound_top = t;
2250  spherical->bound_right = r;
2251  spherical->bound_bottom = b;
2252 
2254  spherical_size);
2255  if (ret < 0) {
2256  av_freep(&spherical);
2257  return ret;
2258  }
2259 
2260  return 0;
2261 }
2262 
2263 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2264 {
2265  const AVCodecTag *codec_tags;
2266 
2267  codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2269 
2270  /* Normalize noncompliant private data that starts with the fourcc
2271  * by expanding/shifting the data by 4 bytes and storing the data
2272  * size at the start. */
2273  if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2274  int ret = av_buffer_realloc(&track->codec_priv.buf,
2276  if (ret < 0)
2277  return ret;
2278 
2279  track->codec_priv.data = track->codec_priv.buf->data;
2280  memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2281  track->codec_priv.size += 4;
2282  AV_WB32(track->codec_priv.data, track->codec_priv.size);
2283  }
2284 
2285  *fourcc = AV_RL32(track->codec_priv.data + 4);
2286  *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2287 
2288  return 0;
2289 }
2290 
2292 {
2293  MatroskaDemuxContext *matroska = s->priv_data;
2294  MatroskaTrack *tracks = matroska->tracks.elem;
2295  AVStream *st;
2296  int i, j, ret;
2297  int k;
2298 
2299  for (i = 0; i < matroska->tracks.nb_elem; i++) {
2300  MatroskaTrack *track = &tracks[i];
2302  EbmlList *encodings_list = &track->encodings;
2303  MatroskaTrackEncoding *encodings = encodings_list->elem;
2304  uint8_t *extradata = NULL;
2305  int extradata_size = 0;
2306  int extradata_offset = 0;
2307  uint32_t fourcc = 0;
2308  AVIOContext b;
2309  char* key_id_base64 = NULL;
2310  int bit_depth = -1;
2311 
2312  /* Apply some sanity checks. */
2313  if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2314  track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2315  track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2316  track->type != MATROSKA_TRACK_TYPE_METADATA) {
2317  av_log(matroska->ctx, AV_LOG_INFO,
2318  "Unknown or unsupported track type %"PRIu64"\n",
2319  track->type);
2320  continue;
2321  }
2322  if (!track->codec_id)
2323  continue;
2324 
2325  if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
2326  || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
2327  || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2328  || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2329  ) {
2330  av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
2331  continue;
2332  }
2333 
2334  if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2335  isnan(track->audio.samplerate)) {
2336  av_log(matroska->ctx, AV_LOG_WARNING,
2337  "Invalid sample rate %f, defaulting to 8000 instead.\n",
2338  track->audio.samplerate);
2339  track->audio.samplerate = 8000;
2340  }
2341 
2342  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2343  if (!track->default_duration && track->video.frame_rate > 0) {
2344  double default_duration = 1000000000 / track->video.frame_rate;
2345  if (default_duration > UINT64_MAX || default_duration < 0) {
2346  av_log(matroska->ctx, AV_LOG_WARNING,
2347  "Invalid frame rate %e. Cannot calculate default duration.\n",
2348  track->video.frame_rate);
2349  } else {
2350  track->default_duration = default_duration;
2351  }
2352  }
2353  if (track->video.display_width == -1)
2354  track->video.display_width = track->video.pixel_width;
2355  if (track->video.display_height == -1)
2356  track->video.display_height = track->video.pixel_height;
2357  if (track->video.color_space.size == 4)
2358  fourcc = AV_RL32(track->video.color_space.data);
2359  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2360  if (!track->audio.out_samplerate)
2361  track->audio.out_samplerate = track->audio.samplerate;
2362  }
2363  if (encodings_list->nb_elem > 1) {
2364  av_log(matroska->ctx, AV_LOG_ERROR,
2365  "Multiple combined encodings not supported");
2366  } else if (encodings_list->nb_elem == 1) {
2367  if (encodings[0].type) {
2368  if (encodings[0].encryption.key_id.size > 0) {
2369  /* Save the encryption key id to be stored later as a
2370  metadata tag. */
2371  const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2372  key_id_base64 = av_malloc(b64_size);
2373  if (key_id_base64 == NULL)
2374  return AVERROR(ENOMEM);
2375 
2376  av_base64_encode(key_id_base64, b64_size,
2377  encodings[0].encryption.key_id.data,
2378  encodings[0].encryption.key_id.size);
2379  } else {
2380  encodings[0].scope = 0;
2381  av_log(matroska->ctx, AV_LOG_ERROR,
2382  "Unsupported encoding type");
2383  }
2384  } else if (
2385 #if CONFIG_ZLIB
2386  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2387 #endif
2388 #if CONFIG_BZLIB
2389  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2390 #endif
2391 #if CONFIG_LZO
2392  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2393 #endif
2394  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2395  encodings[0].scope = 0;
2396  av_log(matroska->ctx, AV_LOG_ERROR,
2397  "Unsupported encoding type");
2398  } else if (track->codec_priv.size && encodings[0].scope & 2) {
2399  uint8_t *codec_priv = track->codec_priv.data;
2400  int ret = matroska_decode_buffer(&track->codec_priv.data,
2401  &track->codec_priv.size,
2402  track);
2403  if (ret < 0) {
2404  track->codec_priv.data = NULL;
2405  track->codec_priv.size = 0;
2406  av_log(matroska->ctx, AV_LOG_ERROR,
2407  "Failed to decode codec private data\n");
2408  }
2409 
2410  if (codec_priv != track->codec_priv.data) {
2411  av_buffer_unref(&track->codec_priv.buf);
2412  if (track->codec_priv.data) {
2413  track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2415  NULL, NULL, 0);
2416  if (!track->codec_priv.buf) {
2417  av_freep(&track->codec_priv.data);
2418  track->codec_priv.size = 0;
2419  return AVERROR(ENOMEM);
2420  }
2421  }
2422  }
2423  }
2424  }
2425  track->needs_decoding = encodings && !encodings[0].type &&
2426  encodings[0].scope & 1 &&
2427  (encodings[0].compression.algo !=
2429  encodings[0].compression.settings.size);
2430 
2431  for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2432  if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2433  strlen(ff_mkv_codec_tags[j].str))) {
2435  break;
2436  }
2437  }
2438 
2439  st = track->stream = avformat_new_stream(s, NULL);
2440  if (!st) {
2441  av_free(key_id_base64);
2442  return AVERROR(ENOMEM);
2443  }
2444 
2445  if (key_id_base64) {
2446  /* export encryption key id as base64 metadata tag */
2447  av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
2449  }
2450 
2451  if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2452  track->codec_priv.size >= 40 &&
2453  track->codec_priv.data) {
2454  track->ms_compat = 1;
2455  bit_depth = AV_RL16(track->codec_priv.data + 14);
2456  fourcc = AV_RL32(track->codec_priv.data + 16);
2458  fourcc);
2459  if (!codec_id)
2461  fourcc);
2462  extradata_offset = 40;
2463  } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2464  track->codec_priv.size >= 14 &&
2465  track->codec_priv.data) {
2466  int ret;
2468  track->codec_priv.size,
2469  0, NULL, NULL, NULL, NULL);
2470  ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2471  if (ret < 0)
2472  return ret;
2473  codec_id = st->codecpar->codec_id;
2474  fourcc = st->codecpar->codec_tag;
2475  extradata_offset = FFMIN(track->codec_priv.size, 18);
2476  } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2477  /* Normally 36, but allow noncompliant private data */
2478  && (track->codec_priv.size >= 32)
2479  && (track->codec_priv.data)) {
2480  uint16_t sample_size;
2481  int ret = get_qt_codec(track, &fourcc, &codec_id);
2482  if (ret < 0)
2483  return ret;
2484  sample_size = AV_RB16(track->codec_priv.data + 26);
2485  if (fourcc == 0) {
2486  if (sample_size == 8) {
2487  fourcc = MKTAG('r','a','w',' ');
2489  } else if (sample_size == 16) {
2490  fourcc = MKTAG('t','w','o','s');
2492  }
2493  }
2494  if ((fourcc == MKTAG('t','w','o','s') ||
2495  fourcc == MKTAG('s','o','w','t')) &&
2496  sample_size == 8)
2498  } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2499  (track->codec_priv.size >= 21) &&
2500  (track->codec_priv.data)) {
2501  int ret = get_qt_codec(track, &fourcc, &codec_id);
2502  if (ret < 0)
2503  return ret;
2504  if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2505  fourcc = MKTAG('S','V','Q','3');
2507  }
2508  if (codec_id == AV_CODEC_ID_NONE)
2509  av_log(matroska->ctx, AV_LOG_ERROR,
2510  "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2511  if (track->codec_priv.size >= 86) {
2512  bit_depth = AV_RB16(track->codec_priv.data + 82);
2514  track->codec_priv.size,
2515  0, NULL, NULL, NULL, NULL);
2516  if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2517  bit_depth &= 0x1F;
2518  track->has_palette = 1;
2519  }
2520  }
2521  } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2522  switch (track->audio.bitdepth) {
2523  case 8:
2525  break;
2526  case 24:
2528  break;
2529  case 32:
2531  break;
2532  }
2533  } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2534  switch (track->audio.bitdepth) {
2535  case 8:
2537  break;
2538  case 24:
2540  break;
2541  case 32:
2543  break;
2544  }
2545  } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2546  track->audio.bitdepth == 64) {
2548  } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2549  int profile = matroska_aac_profile(track->codec_id);
2550  int sri = matroska_aac_sri(track->audio.samplerate);
2551  extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2552  if (!extradata)
2553  return AVERROR(ENOMEM);
2554  extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2555  extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2556  if (strstr(track->codec_id, "SBR")) {
2557  sri = matroska_aac_sri(track->audio.out_samplerate);
2558  extradata[2] = 0x56;
2559  extradata[3] = 0xE5;
2560  extradata[4] = 0x80 | (sri << 3);
2561  extradata_size = 5;
2562  } else
2563  extradata_size = 2;
2564  } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2565  /* Only ALAC's magic cookie is stored in Matroska's track headers.
2566  * Create the "atom size", "tag", and "tag version" fields the
2567  * decoder expects manually. */
2568  extradata_size = 12 + track->codec_priv.size;
2569  extradata = av_mallocz(extradata_size +
2571  if (!extradata)
2572  return AVERROR(ENOMEM);
2573  AV_WB32(extradata, extradata_size);
2574  memcpy(&extradata[4], "alac", 4);
2575  AV_WB32(&extradata[8], 0);
2576  memcpy(&extradata[12], track->codec_priv.data,
2577  track->codec_priv.size);
2578  } else if (codec_id == AV_CODEC_ID_TTA) {
2579  uint8_t *ptr;
2580  if (track->audio.channels > UINT16_MAX ||
2581  track->audio.bitdepth > UINT16_MAX) {
2582  av_log(matroska->ctx, AV_LOG_WARNING,
2583  "Too large audio channel number %"PRIu64
2584  " or bitdepth %"PRIu64". Skipping track.\n",
2585  track->audio.channels, track->audio.bitdepth);
2586  if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2587  return AVERROR_INVALIDDATA;
2588  else
2589  continue;
2590  }
2591  if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2592  return AVERROR_INVALIDDATA;
2593  extradata_size = 22;
2594  extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2595  if (!extradata)
2596  return AVERROR(ENOMEM);
2597  ptr = extradata;
2598  bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2599  bytestream_put_le16(&ptr, 1);
2600  bytestream_put_le16(&ptr, track->audio.channels);
2601  bytestream_put_le16(&ptr, track->audio.bitdepth);
2602  bytestream_put_le32(&ptr, track->audio.out_samplerate);
2603  bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2604  track->audio.out_samplerate,
2605  AV_TIME_BASE * 1000));
2606  } else if (codec_id == AV_CODEC_ID_RV10 ||
2610  extradata_offset = 26;
2611  } else if (codec_id == AV_CODEC_ID_RA_144) {
2612  track->audio.out_samplerate = 8000;
2613  track->audio.channels = 1;
2614  } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2618  && track->codec_priv.data) {
2619  int flavor;
2620 
2622  track->codec_priv.size,
2623  0, NULL, NULL, NULL, NULL);
2624  avio_skip(&b, 22);
2625  flavor = avio_rb16(&b);
2626  track->audio.coded_framesize = avio_rb32(&b);
2627  avio_skip(&b, 12);
2628  track->audio.sub_packet_h = avio_rb16(&b);
2629  track->audio.frame_size = avio_rb16(&b);
2630  track->audio.sub_packet_size = avio_rb16(&b);
2631  if (track->audio.coded_framesize <= 0 ||
2632  track->audio.sub_packet_h <= 0 ||
2633  track->audio.frame_size <= 0)
2634  return AVERROR_INVALIDDATA;
2635 
2636  if (codec_id == AV_CODEC_ID_RA_288) {
2637  if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2638  != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2639  return AVERROR_INVALIDDATA;
2640  st->codecpar->block_align = track->audio.coded_framesize;
2641  track->codec_priv.size = 0;
2642  } else {
2643  if (codec_id == AV_CODEC_ID_SIPR) {
2644  static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2645  if (flavor > 3)
2646  return AVERROR_INVALIDDATA;
2647  track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2648  st->codecpar->bit_rate = sipr_bit_rate[flavor];
2649  } else if (track->audio.sub_packet_size <= 0 ||
2650  track->audio.frame_size % track->audio.sub_packet_size)
2651  return AVERROR_INVALIDDATA;
2652  st->codecpar->block_align = track->audio.sub_packet_size;
2653  extradata_offset = 78;
2654  }
2655  track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2656  track->audio.frame_size);
2657  if (!track->audio.buf)
2658  return AVERROR(ENOMEM);
2659  } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2660  ret = matroska_parse_flac(s, track, &extradata_offset);
2661  if (ret < 0)
2662  return ret;
2663  } else if (codec_id == AV_CODEC_ID_WAVPACK && track->codec_priv.size < 2) {
2664  av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2665  "in absence of valid CodecPrivate.\n");
2666  extradata_size = 2;
2667  extradata = av_mallocz(2 + AV_INPUT_BUFFER_PADDING_SIZE);
2668  if (!extradata)
2669  return AVERROR(ENOMEM);
2670  AV_WL16(extradata, 0x410);
2671  } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2672  fourcc = AV_RL32(track->codec_priv.data);
2673  } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2674  /* we don't need any value stored in CodecPrivate.
2675  make sure that it's not exported as extradata. */
2676  track->codec_priv.size = 0;
2677  } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2678  /* For now, propagate only the OBUs, if any. Once libavcodec is
2679  updated to handle isobmff style extradata this can be removed. */
2680  extradata_offset = 4;
2681  }
2682  track->codec_priv.size -= extradata_offset;
2683 
2684  if (codec_id == AV_CODEC_ID_NONE)
2685  av_log(matroska->ctx, AV_LOG_INFO,
2686  "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2687 
2688  if (track->time_scale < 0.01)
2689  track->time_scale = 1.0;
2690  avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2691  1000 * 1000 * 1000); /* 64 bit pts in ns */
2692 
2693  /* convert the delay from ns to the track timebase */
2695  (AVRational){ 1, 1000000000 },
2696  st->time_base);
2697 
2698  st->codecpar->codec_id = codec_id;
2699 
2700  if (strcmp(track->language, "und"))
2701  av_dict_set(&st->metadata, "language", track->language, 0);
2702  av_dict_set(&st->metadata, "title", track->name, 0);
2703 
2704  if (track->flag_default)
2706  if (track->flag_forced)
2708 
2709  if (!st->codecpar->extradata) {
2710  if (extradata) {
2711  st->codecpar->extradata = extradata;
2712  st->codecpar->extradata_size = extradata_size;
2713  } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2714  if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2715  return AVERROR(ENOMEM);
2716  memcpy(st->codecpar->extradata,
2717  track->codec_priv.data + extradata_offset,
2718  track->codec_priv.size);
2719  }
2720  }
2721 
2722  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2724  int display_width_mul = 1;
2725  int display_height_mul = 1;
2726 
2728  st->codecpar->codec_tag = fourcc;
2729  if (bit_depth >= 0)
2731  st->codecpar->width = track->video.pixel_width;
2732  st->codecpar->height = track->video.pixel_height;
2733 
2735  st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2738 
2740  mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2741 
2744  &st->sample_aspect_ratio.den,
2745  st->codecpar->height * track->video.display_width * display_width_mul,
2746  st->codecpar->width * track->video.display_height * display_height_mul,
2747  255);
2748  }
2749  if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2751 
2752  if (track->default_duration) {
2753  int div = track->default_duration <= INT64_MAX ? 1 : 2;
2755  1000000000 / div, track->default_duration / div, 30000);
2756 #if FF_API_R_FRAME_RATE
2757  if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2758  && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2759  st->r_frame_rate = st->avg_frame_rate;
2760 #endif
2761  }
2762 
2763  /* export stereo mode flag as metadata tag */
2765  av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2766 
2767  /* export alpha mode flag as metadata tag */
2768  if (track->video.alpha_mode)
2769  av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2770 
2771  /* if we have virtual track, mark the real tracks */
2772  for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2773  char buf[32];
2775  continue;
2776  snprintf(buf, sizeof(buf), "%s_%d",
2778  for (k=0; k < matroska->tracks.nb_elem; k++)
2779  if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2780  av_dict_set(&tracks[k].stream->metadata,
2781  "stereo_mode", buf, 0);
2782  break;
2783  }
2784  }
2785  // add stream level stereo3d side data if it is a supported format
2787  track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2788  int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2789  if (ret < 0)
2790  return ret;
2791  }
2792 
2793  ret = mkv_parse_video_color(st, track);
2794  if (ret < 0)
2795  return ret;
2796  ret = mkv_parse_video_projection(st, track, matroska->ctx);
2797  if (ret < 0)
2798  return ret;
2799  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2801  st->codecpar->codec_tag = fourcc;
2802  st->codecpar->sample_rate = track->audio.out_samplerate;
2803  st->codecpar->channels = track->audio.channels;
2804  if (!st->codecpar->bits_per_coded_sample)
2806  if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2807  st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2810  else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2812  if (track->codec_delay > 0) {
2814  (AVRational){1, 1000000000},
2815  (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2816  48000 : st->codecpar->sample_rate});
2817  }
2818  if (track->seek_preroll > 0) {
2820  (AVRational){1, 1000000000},
2821  (AVRational){1, st->codecpar->sample_rate});
2822  }
2823  } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2825 
2826  if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2828  } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2830  } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2832  }
2833  } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2835  }
2836  }
2837 
2838  return 0;
2839 }
2840 
2842 {
2843  MatroskaDemuxContext *matroska = s->priv_data;
2844  EbmlList *attachments_list = &matroska->attachments;
2845  EbmlList *chapters_list = &matroska->chapters;
2846  MatroskaAttachment *attachments;
2847  MatroskaChapter *chapters;
2848  uint64_t max_start = 0;
2849  int64_t pos;
2850  Ebml ebml = { 0 };
2851  int i, j, res;
2852 
2853  matroska->ctx = s;
2854  matroska->cues_parsing_deferred = 1;
2855 
2856  /* First read the EBML header. */
2857  if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2858  av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2859  ebml_free(ebml_syntax, &ebml);
2860  return AVERROR_INVALIDDATA;
2861  }
2862  if (ebml.version > EBML_VERSION ||
2863  ebml.max_size > sizeof(uint64_t) ||
2864  ebml.id_length > sizeof(uint32_t) ||
2865  ebml.doctype_version > 3) {
2867  "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2868  ebml.version, ebml.doctype, ebml.doctype_version);
2869  ebml_free(ebml_syntax, &ebml);
2870  return AVERROR_PATCHWELCOME;
2871  } else if (ebml.doctype_version == 3) {
2872  av_log(matroska->ctx, AV_LOG_WARNING,
2873  "EBML header using unsupported features\n"
2874  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2875  ebml.version, ebml.doctype, ebml.doctype_version);
2876  }
2877  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2878  if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2879  break;
2881  av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2882  if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2883  ebml_free(ebml_syntax, &ebml);
2884  return AVERROR_INVALIDDATA;
2885  }
2886  }
2887  ebml_free(ebml_syntax, &ebml);
2888 
2889  /* The next thing is a segment. */
2890  pos = avio_tell(matroska->ctx->pb);
2891  res = ebml_parse(matroska, matroska_segments, matroska);
2892  // Try resyncing until we find an EBML_STOP type element.
2893  while (res != 1) {
2894  res = matroska_resync(matroska, pos);
2895  if (res < 0)
2896  goto fail;
2897  pos = avio_tell(matroska->ctx->pb);
2898  res = ebml_parse(matroska, matroska_segment, matroska);
2899  if (res == AVERROR(EIO)) // EOF is translated to EIO, this exists the loop on EOF
2900  goto fail;
2901  }
2902  /* Set data_offset as it might be needed later by seek_frame_generic. */
2903  if (matroska->current_id == MATROSKA_ID_CLUSTER)
2904  s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2905  matroska_execute_seekhead(matroska);
2906 
2907  if (!matroska->time_scale)
2908  matroska->time_scale = 1000000;
2909  if (isnan(matroska->duration))
2910  matroska->duration = 0;
2911  if (matroska->duration)
2912  matroska->ctx->duration = matroska->duration * matroska->time_scale *
2913  1000 / AV_TIME_BASE;
2914  av_dict_set(&s->metadata, "title", matroska->title, 0);
2915  av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2916 
2917  if (matroska->date_utc.size == 8)
2918  matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2919 
2920  res = matroska_parse_tracks(s);
2921  if (res < 0)
2922  goto fail;
2923 
2924  attachments = attachments_list->elem;
2925  for (j = 0; j < attachments_list->nb_elem; j++) {
2926  if (!(attachments[j].filename && attachments[j].mime &&
2927  attachments[j].bin.data && attachments[j].bin.size > 0)) {
2928  av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2929  } else {
2931  if (!st)
2932  break;
2933  av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2934  av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2935  if (attachments[j].description)
2936  av_dict_set(&st->metadata, "title", attachments[j].description, 0);
2938 
2939  for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2940  if (!strncmp(mkv_image_mime_tags[i].str, attachments[j].mime,
2941  strlen(mkv_image_mime_tags[i].str))) {
2943  break;
2944  }
2945  }
2946 
2947  attachments[j].stream = st;
2948 
2949  if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2950  AVPacket *pkt = &st->attached_pic;
2951 
2954 
2956  pkt->buf = attachments[j].bin.buf;
2957  attachments[j].bin.buf = NULL;
2958  pkt->data = attachments[j].bin.data;
2959  pkt->size = attachments[j].bin.size;
2960  pkt->stream_index = st->index;
2962  } else {
2964  if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2965  break;
2966  memcpy(st->codecpar->extradata, attachments[j].bin.data,
2967  attachments[j].bin.size);
2968 
2969  for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2970  if (!strncmp(mkv_mime_tags[i].str, attachments[j].mime,
2971  strlen(mkv_mime_tags[i].str))) {
2973  break;
2974  }
2975  }
2976  }
2977  }
2978  }
2979 
2980  chapters = chapters_list->elem;
2981  for (i = 0; i < chapters_list->nb_elem; i++)
2982  if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2983  (max_start == 0 || chapters[i].start > max_start)) {
2984  chapters[i].chapter =
2985  avpriv_new_chapter(s, chapters[i].uid,
2986  (AVRational) { 1, 1000000000 },
2987  chapters[i].start, chapters[i].end,
2988  chapters[i].title);
2989  max_start = chapters[i].start;
2990  }
2991 
2992  matroska_add_index_entries(matroska);
2993 
2995 
2996  return 0;
2997 fail:
2999  return res;
3001 
3002 /*
3003  * Put one packet in an application-supplied AVPacket struct.
3004  * Returns 0 on success or -1 on failure.
3005  */
3006 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3007  AVPacket *pkt)
3008 {
3009  if (matroska->queue) {
3010  MatroskaTrack *tracks = matroska->tracks.elem;
3011  MatroskaTrack *track;
3012 
3013  ff_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
3014  track = &tracks[pkt->stream_index];
3015  if (track->has_palette) {
3017  if (!pal) {
3018  av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3019  } else {
3020  memcpy(pal, track->palette, AVPALETTE_SIZE);
3021  }
3022  track->has_palette = 0;
3023  }
3024  return 0;
3025  }
3026 
3027  return -1;
3028 }
3029 
3030 /*
3031  * Free all packets in our internal queue.
3032  */
3033 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3034 {
3035  ff_packet_list_free(&matroska->queue, &matroska->queue_end);
3036 }
3037 
3038 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3039  int size, int type, AVIOContext *pb,
3040  uint32_t lace_size[256], int *laces)
3041 {
3042  int n;
3043  uint8_t *data = *buf;
3044 
3045  if (!type) {
3046  *laces = 1;
3047  lace_size[0] = size;
3048  return 0;
3049  }
3050 
3051  if (size <= 0)
3052  return AVERROR_INVALIDDATA;
3053 
3054  *laces = *data + 1;
3055  data += 1;
3056  size -= 1;
3057 
3058  switch (type) {
3059  case 0x1: /* Xiph lacing */
3060  {
3061  uint8_t temp;
3062  uint32_t total = 0;
3063  for (n = 0; n < *laces - 1; n++) {
3064  lace_size[n] = 0;
3065 
3066  do {
3067  if (size <= total)
3068  return AVERROR_INVALIDDATA;
3069  temp = *data;
3070  total += temp;
3071  lace_size[n] += temp;
3072  data += 1;
3073  size -= 1;
3074  } while (temp == 0xff);
3075  }
3076  if (size < total)
3077  return AVERROR_INVALIDDATA;
3078 
3079  lace_size[n] = size - total;
3080  break;
3081  }
3082 
3083  case 0x2: /* fixed-size lacing */
3084  if (size % (*laces))
3085  return AVERROR_INVALIDDATA;
3086  for (n = 0; n < *laces; n++)
3087  lace_size[n] = size / *laces;
3088  break;
3089 
3090  case 0x3: /* EBML lacing */
3091  {
3092  uint64_t num;
3093  uint64_t total;
3094  int offset;
3095 
3096  avio_skip(pb, 4);
3097 
3098  n = ebml_read_num(matroska, pb, 8, &num, 1);
3099  if (n < 0)
3100  return n;
3101  if (num > INT_MAX)
3102  return AVERROR_INVALIDDATA;
3103 
3104  total = lace_size[0] = num;
3105  offset = n;
3106  for (n = 1; n < *laces - 1; n++) {
3107  int64_t snum;
3108  int r;
3109  r = matroska_ebmlnum_sint(matroska, pb, &snum);
3110  if (r < 0)
3111  return r;
3112  if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3113  return AVERROR_INVALIDDATA;
3114 
3115  lace_size[n] = lace_size[n - 1] + snum;
3116  total += lace_size[n];
3117  offset += r;
3118  }
3119  data += offset;
3120  size -= offset;
3121  if (size < total)
3122  return AVERROR_INVALIDDATA;
3123 
3124  lace_size[*laces - 1] = size - total;
3125  break;
3126  }
3127  }
3129  *buf = data;
3130 
3131  return 0;
3132 }
3133 
3134 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3135  MatroskaTrack *track, AVStream *st,
3136  uint8_t *data, int size, uint64_t timecode,
3137  int64_t pos)
3138 {
3139  const int a = st->codecpar->block_align;
3140  const int sps = track->audio.sub_packet_size;
3141  const int cfs = track->audio.coded_framesize;
3142  const int h = track->audio.sub_packet_h;
3143  const int w = track->audio.frame_size;
3144  int y = track->audio.sub_packet_cnt;
3145  int x;
3146 
3147  if (!track->audio.pkt_cnt) {
3148  if (track->audio.sub_packet_cnt == 0)
3149  track->audio.buf_timecode = timecode;
3150  if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3151  if (size < cfs * h / 2) {
3152  av_log(matroska->ctx, AV_LOG_ERROR,
3153  "Corrupt int4 RM-style audio packet size\n");
3154  return AVERROR_INVALIDDATA;
3155  }
3156  for (x = 0; x < h / 2; x++)
3157  memcpy(track->audio.buf + x * 2 * w + y * cfs,
3158  data + x * cfs, cfs);
3159  } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3160  if (size < w) {
3161  av_log(matroska->ctx, AV_LOG_ERROR,
3162  "Corrupt sipr RM-style audio packet size\n");
3163  return AVERROR_INVALIDDATA;
3164  }
3165  memcpy(track->audio.buf + y * w, data, w);
3166  } else {
3167  if (size < w) {
3168  av_log(matroska->ctx, AV_LOG_ERROR,
3169  "Corrupt generic RM-style audio packet size\n");
3170  return AVERROR_INVALIDDATA;
3171  }
3172  for (x = 0; x < w / sps; x++)
3173  memcpy(track->audio.buf +
3174  sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3175  data + x * sps, sps);
3176  }
3177 
3178  if (++track->audio.sub_packet_cnt >= h) {
3179  if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3180  ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3181  track->audio.sub_packet_cnt = 0;
3182  track->audio.pkt_cnt = h * w / a;
3183  }
3184  }
3185 
3186  while (track->audio.pkt_cnt) {
3187  int ret;
3188  AVPacket pktl, *pkt = &pktl;
3189 
3190  ret = av_new_packet(pkt, a);
3191  if (ret < 0) {
3192  return ret;
3193  }
3194  memcpy(pkt->data,
3195  track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3196  a);
3197  pkt->pts = track->audio.buf_timecode;
3199  pkt->pos = pos;
3200  pkt->stream_index = st->index;
3201  ret = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3202  if (ret < 0) {
3204  return AVERROR(ENOMEM);
3205  }
3206  }
3207 
3208  return 0;
3209 }
3210 
3211 /* reconstruct full wavpack blocks from mangled matroska ones */
3212 static int matroska_parse_wavpack(MatroskaTrack *track,
3213  uint8_t **data, int *size)
3214 {
3215  uint8_t *dst = NULL;
3216  uint8_t *src = *data;
3217  int dstlen = 0;
3218  int srclen = *size;
3219  uint32_t samples;
3220  uint16_t ver;
3221  int ret, offset = 0;
3222 
3223  if (srclen < 12)
3224  return AVERROR_INVALIDDATA;
3225 
3226  av_assert1(track->stream->codecpar->extradata_size >= 2);
3227  ver = AV_RL16(track->stream->codecpar->extradata);
3228 
3229  samples = AV_RL32(src);
3230  src += 4;
3231  srclen -= 4;
3232 
3233  while (srclen >= 8) {
3234  int multiblock;
3235  uint32_t blocksize;
3236  uint8_t *tmp;
3237 
3238  uint32_t flags = AV_RL32(src);
3239  uint32_t crc = AV_RL32(src + 4);
3240  src += 8;
3241  srclen -= 8;
3242 
3243  multiblock = (flags & 0x1800) != 0x1800;
3244  if (multiblock) {
3245  if (srclen < 4) {
3247  goto fail;
3248  }
3249  blocksize = AV_RL32(src);
3250  src += 4;
3251  srclen -= 4;
3252  } else
3253  blocksize = srclen;
3254 
3255  if (blocksize > srclen) {
3257  goto fail;
3258  }
3259 
3260  tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3261  if (!tmp) {
3262  ret = AVERROR(ENOMEM);
3263  goto fail;
3264  }
3265  dst = tmp;
3266  dstlen += blocksize + 32;
3267 
3268  AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3269  AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3270  AV_WL16(dst + offset + 8, ver); // version
3271  AV_WL16(dst + offset + 10, 0); // track/index_no
3272  AV_WL32(dst + offset + 12, 0); // total samples
3273  AV_WL32(dst + offset + 16, 0); // block index
3274  AV_WL32(dst + offset + 20, samples); // number of samples
3275  AV_WL32(dst + offset + 24, flags); // flags
3276  AV_WL32(dst + offset + 28, crc); // crc
3277  memcpy(dst + offset + 32, src, blocksize); // block data
3278 
3279  src += blocksize;
3280  srclen -= blocksize;
3281  offset += blocksize + 32;
3282  }
3283 
3284  memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3285 
3286  *data = dst;
3287  *size = dstlen;
3288 
3289  return 0;
3291 fail:
3292  av_freep(&dst);
3293  return ret;
3294 }
3295 
3296 static int matroska_parse_prores(MatroskaTrack *track,
3297  uint8_t **data, int *size)
3298 {
3299  uint8_t *dst;
3300  int dstlen = *size + 8;
3301 
3302  dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3303  if (!dst)
3304  return AVERROR(ENOMEM);
3305 
3306  AV_WB32(dst, dstlen);
3307  AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3308  memcpy(dst + 8, *data, dstlen - 8);
3309  memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3310 
3311  *data = dst;
3312  *size = dstlen;
3313 
3314  return 0;
3315 }
3316 
3317 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3318  MatroskaTrack *track,
3319  AVStream *st,
3320  uint8_t *data, int data_len,
3321  uint64_t timecode,
3322  uint64_t duration,
3323  int64_t pos)
3324 {
3325  AVPacket pktl, *pkt = &pktl;
3326  uint8_t *id, *settings, *text, *buf;
3327  int id_len, settings_len, text_len;
3328  uint8_t *p, *q;
3329  int err;
3330 
3331  if (data_len <= 0)
3332  return AVERROR_INVALIDDATA;
3333 
3334  p = data;
3335  q = data + data_len;
3336 
3337  id = p;
3338  id_len = -1;
3339  while (p < q) {
3340  if (*p == '\r' || *p == '\n') {
3341  id_len = p - id;
3342  if (*p == '\r')
3343  p++;
3344  break;
3345  }
3346  p++;
3347  }
3348 
3349  if (p >= q || *p != '\n')
3350  return AVERROR_INVALIDDATA;
3351  p++;
3352 
3353  settings = p;
3354  settings_len = -1;
3355  while (p < q) {
3356  if (*p == '\r' || *p == '\n') {
3357  settings_len = p - settings;
3358  if (*p == '\r')
3359  p++;
3360  break;
3361  }
3362  p++;
3363  }
3364 
3365  if (p >= q || *p != '\n')
3366  return AVERROR_INVALIDDATA;
3367  p++;
3368 
3369  text = p;
3370  text_len = q - p;
3371  while (text_len > 0) {
3372  const int len = text_len - 1;
3373  const uint8_t c = p[len];
3374  if (c != '\r' && c != '\n')
3375  break;
3376  text_len = len;
3377  }
3378 
3379  if (text_len <= 0)
3380  return AVERROR_INVALIDDATA;
3381 
3382  err = av_new_packet(pkt, text_len);
3383  if (err < 0) {
3384  return err;
3385  }
3386 
3387  memcpy(pkt->data, text, text_len);
3388 
3389  if (id_len > 0) {
3392  id_len);
3393  if (!buf) {
3395  return AVERROR(ENOMEM);
3396  }
3397  memcpy(buf, id, id_len);
3398  }
3399 
3400  if (settings_len > 0) {
3403  settings_len);
3404  if (!buf) {
3406  return AVERROR(ENOMEM);
3407  }
3408  memcpy(buf, settings, settings_len);
3409  }
3410 
3411  // Do we need this for subtitles?
3412  // pkt->flags = AV_PKT_FLAG_KEY;
3413 
3414  pkt->stream_index = st->index;
3415  pkt->pts = timecode;
3416 
3417  // Do we need this for subtitles?
3418  // pkt->dts = timecode;
3419 
3420  pkt->duration = duration;
3421  pkt->pos = pos;
3422 
3423  err = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3424  if (err < 0) {
3426  return AVERROR(ENOMEM);
3427  }
3428 
3429  return 0;
3430 }
3431 
3432 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3433  MatroskaTrack *track, AVStream *st,
3434  AVBufferRef *buf, uint8_t *data, int pkt_size,
3435  uint64_t timecode, uint64_t lace_duration,
3436  int64_t pos, int is_keyframe,
3437  uint8_t *additional, uint64_t additional_id, int additional_size,
3438  int64_t discard_padding)
3439 {
3440  uint8_t *pkt_data = data;
3441  int res = 0;
3442  AVPacket pktl, *pkt = &pktl;
3443 
3444  if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3445  res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3446  if (res < 0) {
3447  av_log(matroska->ctx, AV_LOG_ERROR,
3448  "Error parsing a wavpack block.\n");
3449  goto fail;
3450  }
3451  if (!buf)
3452  av_freep(&data);
3453  buf = NULL;
3454  }
3455 
3456  if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3457  AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3458  res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3459  if (res < 0) {
3460  av_log(matroska->ctx, AV_LOG_ERROR,
3461  "Error parsing a prores block.\n");
3462  goto fail;
3463  }
3464  if (!buf)
3465  av_freep(&data);
3466  buf = NULL;
3467  }
3468 
3469  if (!pkt_size && !additional_size)
3470  goto no_output;
3471 
3473  if (!buf)
3474  pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3475  NULL, NULL, 0);
3476  else
3477  pkt->buf = av_buffer_ref(buf);
3478 
3479  if (!pkt->buf) {
3480  res = AVERROR(ENOMEM);
3481  goto fail;
3482  }
3483 
3484  pkt->data = pkt_data;
3485  pkt->size = pkt_size;
3486  pkt->flags = is_keyframe;
3487  pkt->stream_index = st->index;
3488 
3489  if (additional_size > 0) {
3490  uint8_t *side_data = av_packet_new_side_data(pkt,
3492  additional_size + 8);
3493  if (!side_data) {
3495  return AVERROR(ENOMEM);
3496  }
3497  AV_WB64(side_data, additional_id);
3498  memcpy(side_data + 8, additional, additional_size);
3499  }
3500 
3501  if (discard_padding) {
3502  uint8_t *side_data = av_packet_new_side_data(pkt,
3504  10);
3505  if (!side_data) {
3507  return AVERROR(ENOMEM);
3508  }
3509  discard_padding = av_rescale_q(discard_padding,
3510  (AVRational){1, 1000000000},
3511  (AVRational){1, st->codecpar->sample_rate});
3512  if (discard_padding > 0) {
3513  AV_WL32(side_data + 4, discard_padding);
3514  } else {
3515  AV_WL32(side_data, -discard_padding);
3516  }
3517  }
3518 
3519  if (track->ms_compat)
3520  pkt->dts = timecode;
3521  else
3522  pkt->pts = timecode;
3523  pkt->pos = pos;
3524  pkt->duration = lace_duration;
3525 
3526 #if FF_API_CONVERGENCE_DURATION
3528  if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3529  pkt->convergence_duration = lace_duration;
3530  }
3532 #endif
3533 
3534  res = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3535  if (res < 0) {
3537  return AVERROR(ENOMEM);
3538  }
3539 
3540  return 0;
3541 
3542 no_output:
3544  if (!buf)
3545  av_free(pkt_data);
3546  return res;
3547 }
3548 
3550  int size, int64_t pos, uint64_t cluster_time,
3551  uint64_t block_duration, int is_keyframe,
3552  uint8_t *additional, uint64_t additional_id, int additional_size,
3553  int64_t cluster_pos, int64_t discard_padding)
3554 {
3555  uint64_t timecode = AV_NOPTS_VALUE;
3556  MatroskaTrack *track;
3557  AVIOContext pb;
3558  int res = 0;
3559  AVStream *st;
3560  int16_t block_time;
3561  uint32_t lace_size[256];
3562  int n, flags, laces = 0;
3563  uint64_t num;
3564  int trust_default_duration = 1;
3565 
3566  ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3567 
3568  if ((n = ebml_read_num(matroska, &pb, 8, &num, 1)) < 0)
3569  return n;
3570  data += n;
3571  size -= n;
3572 
3573  track = matroska_find_track_by_num(matroska, num);
3574  if (!track || size < 3)
3575  return AVERROR_INVALIDDATA;
3576 
3577  if (!(st = track->stream)) {
3578  av_log(matroska->ctx, AV_LOG_VERBOSE,
3579  "No stream associated to TrackNumber %"PRIu64". "
3580  "Ignoring Block with this TrackNumber.\n", num);
3581  return 0;
3582  }
3583 
3584  if (st->discard >= AVDISCARD_ALL)
3585  return res;
3586  if (block_duration > INT64_MAX)
3587  block_duration = INT64_MAX;
3588 
3589  block_time = sign_extend(AV_RB16(data), 16);
3590  data += 2;
3591  flags = *data++;
3592  size -= 3;
3593  if (is_keyframe == -1)
3594  is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3595 
3596  if (cluster_time != (uint64_t) -1 &&
3597  (block_time >= 0 || cluster_time >= -block_time)) {
3598  timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3599  if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3600  timecode < track->end_timecode)
3601  is_keyframe = 0; /* overlapping subtitles are not key frame */
3602  if (is_keyframe) {
3603  ff_reduce_index(matroska->ctx, st->index);
3604  av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3606  }
3607  }
3608 
3609  if (matroska->skip_to_keyframe &&
3610  track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3611  // Compare signed timecodes. Timecode may be negative due to codec delay
3612  // offset. We don't support timestamps greater than int64_t anyway - see
3613  // AVPacket's pts.
3614  if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3615  return res;
3616  if (is_keyframe)
3617  matroska->skip_to_keyframe = 0;
3618  else if (!st->skip_to_keyframe) {
3619  av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3620  matroska->skip_to_keyframe = 0;
3621  }
3622  }
3623 
3624  res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3625  &pb, lace_size, &laces);
3626  if (res < 0) {
3627  av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3628  return res;
3629  }
3630 
3631  if (track->audio.samplerate == 8000) {
3632  // If this is needed for more codecs, then add them here
3633  if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3634  if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3635  trust_default_duration = 0;
3636  }
3637  }
3638 
3639  if (!block_duration && trust_default_duration)
3640  block_duration = track->default_duration * laces / matroska->time_scale;
3641 
3642  if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3643  track->end_timecode =
3644  FFMAX(track->end_timecode, timecode + block_duration);
3645 
3646  for (n = 0; n < laces; n++) {
3647  int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3648  uint8_t *out_data = data;
3649  int out_size = lace_size[n];
3650 
3651  if (track->needs_decoding) {
3652  res = matroska_decode_buffer(&out_data, &out_size, track);
3653  if (res < 0)
3654  return res;
3655  /* Given that we are here means that out_data is no longer
3656  * owned by buf, so set it to NULL. This depends upon
3657  * zero-length header removal compression being ignored. */
3658  av_assert1(out_data != data);
3659  buf = NULL;
3660  }
3661 
3662  if (track->audio.buf) {
3663  res = matroska_parse_rm_audio(matroska, track, st,
3664  out_data, out_size,
3665  timecode, pos);
3666  if (!buf)
3667  av_free(out_data);
3668  if (res)
3669  return res;
3670  } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3671  res = matroska_parse_webvtt(matroska, track, st,
3672  out_data, out_size,
3673  timecode, lace_duration,
3674  pos);
3675  if (!buf)
3676  av_free(out_data);
3677  if (res)
3678  return res;
3679  } else {
3680  res = matroska_parse_frame(matroska, track, st, buf, out_data,
3681  out_size, timecode, lace_duration,
3682  pos, !n ? is_keyframe : 0,
3683  additional, additional_id, additional_size,
3684  discard_padding);
3685  if (res)
3686  return res;
3687  }
3688 
3689  if (timecode != AV_NOPTS_VALUE)
3690  timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3691  data += lace_size[n];
3692  }
3693 
3694  return 0;
3695 }
3696 
3697 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3698 {
3699  MatroskaCluster *cluster = &matroska->current_cluster;
3700  MatroskaBlock *block = &cluster->block;
3701  int res;
3702 
3703  av_assert0(matroska->num_levels <= 2);
3704 
3705  if (matroska->num_levels == 1) {
3706  res = ebml_parse(matroska, matroska_segment, NULL);
3707 
3708  if (res == 1) {
3709  /* Found a cluster: subtract the size of the ID already read. */
3710  cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3711 
3712  res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3713  if (res < 0)
3714  return res;
3715  }
3716  }
3717 
3718  if (matroska->num_levels == 2) {
3719  /* We are inside a cluster. */
3720  res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3721 
3722  if (res >= 0 && block->bin.size > 0) {
3723  int is_keyframe = block->non_simple ? block->reference == INT64_MIN : -1;
3724  uint8_t* additional = block->additional.size > 0 ?
3725  block->additional.data : NULL;
3726 
3727  res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3728  block->bin.size, block->bin.pos,
3729  cluster->timecode, block->duration,
3730  is_keyframe, additional, block->additional_id,
3731  block->additional.size, cluster->pos,
3732  block->discard_padding);
3733  }
3734 
3736  memset(block, 0, sizeof(*block));
3737  } else if (!matroska->num_levels) {
3738  if (!avio_feof(matroska->ctx->pb)) {
3739  avio_r8(matroska->ctx->pb);
3740  if (!avio_feof(matroska->ctx->pb)) {
3741  av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3742  "end of segment.\n");
3743  return AVERROR_INVALIDDATA;
3744  }
3745  }
3746  matroska->done = 1;
3747  return AVERROR_EOF;
3748  }
3749 
3750  return res;
3751 }
3752 
3754 {
3755  MatroskaDemuxContext *matroska = s->priv_data;
3756  int ret = 0;
3757 
3758  if (matroska->resync_pos == -1) {
3759  // This can only happen if generic seeking has been used.
3760  matroska->resync_pos = avio_tell(s->pb);
3761  }
3762 
3763  while (matroska_deliver_packet(matroska, pkt)) {
3764  if (matroska->done)
3765  return (ret < 0) ? ret : AVERROR_EOF;
3766  if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3767  ret = matroska_resync(matroska, matroska->resync_pos);
3768  }
3769 
3770  return 0;
3771 }
3772 
3773 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3774  int64_t timestamp, int flags)
3775 {
3776  MatroskaDemuxContext *matroska = s->priv_data;
3777  MatroskaTrack *tracks = NULL;
3778  AVStream *st = s->streams[stream_index];
3779  int i, index;
3780 
3781  /* Parse the CUES now since we need the index data to seek. */
3782  if (matroska->cues_parsing_deferred > 0) {
3783  matroska->cues_parsing_deferred = 0;
3784  matroska_parse_cues(matroska);
3785  }
3786 
3787  if (!st->nb_index_entries)
3788  goto err;
3789  timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3790 
3791  if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3792  matroska_reset_status(matroska, 0, st->index_entries[st->nb_index_entries - 1].pos);
3793  while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3794  matroska_clear_queue(matroska);
3795  if (matroska_parse_cluster(matroska) < 0)
3796  break;
3797  }
3798  }
3799 
3800  matroska_clear_queue(matroska);
3801  if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3802  goto err;
3803 
3804  tracks = matroska->tracks.elem;
3805  for (i = 0; i < matroska->tracks.nb_elem; i++) {
3806  tracks[i].audio.pkt_cnt = 0;
3807  tracks[i].audio.sub_packet_cnt = 0;
3808  tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3809  tracks[i].end_timecode = 0;
3810  }
3811 
3812  /* We seek to a level 1 element, so set the appropriate status. */
3813  matroska_reset_status(matroska, 0, st->index_entries[index].pos);
3814  if (flags & AVSEEK_FLAG_ANY) {
3815  st->skip_to_keyframe = 0;
3816  matroska->skip_to_timecode = timestamp;
3817  } else {
3818  st->skip_to_keyframe = 1;
3819  matroska->skip_to_timecode = st->index_entries[index].timestamp;
3820  }
3821  matroska->skip_to_keyframe = 1;
3822  matroska->done = 0;
3824  return 0;
3825 err:
3826  // slightly hackish but allows proper fallback to
3827  // the generic seeking code.
3828  matroska_reset_status(matroska, 0, -1);
3829  matroska->resync_pos = -1;
3830  matroska_clear_queue(matroska);
3832  matroska->skip_to_keyframe = 0;
3833  matroska->done = 0;
3834  return -1;
3835 }
3836 
3838 {
3839  MatroskaDemuxContext *matroska = s->priv_data;
3840  MatroskaTrack *tracks = matroska->tracks.elem;
3841  int n;
3842 
3843  matroska_clear_queue(matroska);
3844 
3845  for (n = 0; n < matroska->tracks.nb_elem; n++)
3846  if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3847  av_freep(&tracks[n].audio.buf);
3850  return 0;
3852 
3853 typedef struct {
3854  int64_t start_time_ns;
3855  int64_t end_time_ns;
3856  int64_t start_offset;
3857  int64_t end_offset;
3859 
3860 /* This function searches all the Cues and returns the CueDesc corresponding to
3861  * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3862  * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3863  */
3864 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3865  MatroskaDemuxContext *matroska = s->priv_data;
3866  CueDesc cue_desc;
3867  int i;
3868  int nb_index_entries = s->streams[0]->nb_index_entries;
3869  AVIndexEntry *index_entries = s->streams[0]->index_entries;
3870 
3871  if (ts >= (int64_t)(matroska->duration * matroska->time_scale))
3872  return (CueDesc) {-1, -1, -1, -1};
3873  for (i = 1; i < nb_index_entries; i++) {
3874  if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3875  index_entries[i].timestamp * matroska->time_scale > ts) {
3876  break;
3877  }
3878  }
3879  --i;
3880  cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3881  cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3882  if (i != nb_index_entries - 1) {
3883  cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3884  cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3885  } else {
3886  cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3887  // FIXME: this needs special handling for files where Cues appear
3888  // before Clusters. the current logic assumes Cues appear after
3889  // Clusters.
3890  cue_desc.end_offset = cues_start - matroska->segment_start;
3891  }
3892  return cue_desc;
3893 }
3894 
3896 {
3897  MatroskaDemuxContext *matroska = s->priv_data;
3898  uint32_t id = matroska->current_id;
3899  int64_t cluster_pos, before_pos;
3900  int index, rv = 1;
3901  if (s->streams[0]->nb_index_entries <= 0) return 0;
3902  // seek to the first cluster using cues.
3903  index = av_index_search_timestamp(s->streams[0], 0, 0);
3904  if (index < 0) return 0;
3905  cluster_pos = s->streams[0]->index_entries[index].pos;
3906  before_pos = avio_tell(s->pb);
3907  while (1) {
3908  uint64_t cluster_id, cluster_length;
3909  int read;
3910  AVPacket *pkt;
3911  avio_seek(s->pb, cluster_pos, SEEK_SET);
3912  // read cluster id and length
3913  read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3914  if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3915  break;
3916  read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3917  if (read < 0)
3918  break;
3919 
3920  matroska_reset_status(matroska, 0, cluster_pos);
3921  matroska_clear_queue(matroska);
3922  if (matroska_parse_cluster(matroska) < 0 ||
3923  !matroska->queue) {
3924  break;
3925  }
3926  pkt = &matroska->queue->pkt;
3927  // 4 + read is the length of the cluster id and the cluster length field.
3928  cluster_pos += 4 + read + cluster_length;
3929  if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3930  rv = 0;
3931  break;
3932  }
3933  }
3934 
3935  /* Restore the status after matroska_read_header: */
3936  matroska_reset_status(matroska, id, before_pos);
3937 
3938  return rv;
3939 }
3940 
3941 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3942  double min_buffer, double* buffer,
3943  double* sec_to_download, AVFormatContext *s,
3944  int64_t cues_start)
3945 {
3946  double nano_seconds_per_second = 1000000000.0;
3947  double time_sec = time_ns / nano_seconds_per_second;
3948  int rv = 0;
3949  int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3950  int64_t end_time_ns = time_ns + time_to_search_ns;
3951  double sec_downloaded = 0.0;
3952  CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3953  if (desc_curr.start_time_ns == -1)
3954  return -1;
3955  *sec_to_download = 0.0;
3956 
3957  // Check for non cue start time.
3958  if (time_ns > desc_curr.start_time_ns) {
3959  int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3960  double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3961  double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3962  double timeToDownload = (cueBytes * 8.0) / bps;
3963 
3964  sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3965  *sec_to_download += timeToDownload;
3966 
3967  // Check if the search ends within the first cue.
3968  if (desc_curr.end_time_ns >= end_time_ns) {
3969  double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3970  double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3971  sec_downloaded = percent_to_sub * sec_downloaded;
3972  *sec_to_download = percent_to_sub * *sec_to_download;
3973  }
3974 
3975  if ((sec_downloaded + *buffer) <= min_buffer) {
3976  return 1;
3977  }
3978 
3979  // Get the next Cue.
3980  desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3981  }
3982 
3983  while (desc_curr.start_time_ns != -1) {
3984  int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3985  int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3986  double desc_sec = desc_ns / nano_seconds_per_second;
3987  double bits = (desc_bytes * 8.0);
3988  double time_to_download = bits / bps;
3989 
3990  sec_downloaded += desc_sec - time_to_download;
3991  *sec_to_download += time_to_download;
3992 
3993  if (desc_curr.end_time_ns >= end_time_ns) {
3994  double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3995  double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3996  sec_downloaded = percent_to_sub * sec_downloaded;
3997  *sec_to_download = percent_to_sub * *sec_to_download;
3998 
3999  if ((sec_downloaded + *buffer) <= min_buffer)
4000  rv = 1;
4001  break;
4002  }
4003 
4004  if ((sec_downloaded + *buffer) <= min_buffer) {
4005  rv = 1;
4006  break;
4007  }
4008 
4009  desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4010  }
4011  *buffer = *buffer + sec_downloaded;
4012  return rv;
4013 }
4014 
4015 /* This function computes the bandwidth of the WebM file with the help of
4016  * buffer_size_after_time_downloaded() function. Both of these functions are
4017  * adapted from WebM Tools project and are adapted to work with FFmpeg's
4018  * Matroska parsing mechanism.
4019  *
4020  * Returns the bandwidth of the file on success; -1 on error.
4021  * */
4022 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4023 {
4024  MatroskaDemuxContext *matroska = s->priv_data;
4025  AVStream *st = s->streams[0];
4026  double bandwidth = 0.0;
4027  int i;
4028 
4029  for (i = 0; i < st->nb_index_entries; i++) {
4030  int64_t prebuffer_ns = 1000000000;
4031  int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
4032  double nano_seconds_per_second = 1000000000.0;
4033  int64_t prebuffered_ns = time_ns + prebuffer_ns;
4034  double prebuffer_bytes = 0.0;
4035  int64_t temp_prebuffer_ns = prebuffer_ns;
4036  int64_t pre_bytes, pre_ns;
4037  double pre_sec, prebuffer, bits_per_second;
4038  CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4039 
4040  // Start with the first Cue.
4041  CueDesc desc_end = desc_beg;
4042 
4043  // Figure out how much data we have downloaded for the prebuffer. This will
4044  // be used later to adjust the bits per sample to try.
4045  while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4046  // Prebuffered the entire Cue.
4047  prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4048  temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4049  desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4050  }
4051  if (desc_end.start_time_ns == -1) {
4052  // The prebuffer is larger than the duration.
4053  if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4054  return -1;
4055  bits_per_second = 0.0;
4056  } else {
4057  // The prebuffer ends in the last Cue. Estimate how much data was
4058  // prebuffered.
4059  pre_bytes = desc_end.end_offset - desc_end.start_offset;
4060  pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4061  if (pre_ns <= 0)
4062  return -1;
4063  pre_sec = pre_ns / nano_seconds_per_second;
4064  prebuffer_bytes +=
4065  pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4066 
4067  prebuffer = prebuffer_ns / nano_seconds_per_second;
4068 
4069  // Set this to 0.0 in case our prebuffer buffers the entire video.
4070  bits_per_second = 0.0;
4071  do {
4072  int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4073  int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4074  double desc_sec, calc_bits_per_second, percent, mod_bits_per_second;
4075  if (desc_bytes <= 0)
4076  return -1;
4077 
4078  desc_sec = desc_ns / nano_seconds_per_second;
4079  calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4080 
4081  // Drop the bps by the percentage of bytes buffered.
4082  percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4083  mod_bits_per_second = calc_bits_per_second * percent;
4084 
4085  if (prebuffer < desc_sec) {
4086  double search_sec =
4087  (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4088 
4089  // Add 1 so the bits per second should be a little bit greater than file
4090  // datarate.
4091  int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4092  const double min_buffer = 0.0;
4093  double buffer = prebuffer;
4094  double sec_to_download = 0.0;
4095 
4096  int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4097  min_buffer, &buffer, &sec_to_download,
4098  s, cues_start);
4099  if (rv < 0) {
4100  return -1;
4101  } else if (rv == 0) {
4102  bits_per_second = (double)(bps);
4103  break;
4104  }
4105  }
4106 
4107  desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4108  } while (desc_end.start_time_ns != -1);
4109  }
4110  if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4111  }
4112  return (int64_t)bandwidth;
4113 }
4114 
4115 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4116 {
4117  MatroskaDemuxContext *matroska = s->priv_data;
4118  EbmlList *seekhead_list = &matroska->seekhead;
4119  MatroskaSeekhead *seekhead = seekhead_list->elem;
4120  char *buf;
4121  int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4122  int i;
4123  int end = 0;
4124 
4125  // determine cues start and end positions
4126  for (i = 0; i < seekhead_list->nb_elem; i++)
4127  if (seekhead[i].id == MATROSKA_ID_CUES)
4128  break;
4129 
4130  if (i >= seekhead_list->nb_elem) return -1;
4131 
4132  before_pos = avio_tell(matroska->ctx->pb);
4133  cues_start = seekhead[i].pos + matroska->segment_start;
4134  if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4135  // cues_end is computed as cues_start + cues_length + length of the
4136  // Cues element ID (i.e. 4) + EBML length of the Cues element.
4137  // cues_end is inclusive and the above sum is reduced by 1.
4138  uint64_t cues_length, cues_id;
4139  int bytes_read;
4140  bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4141  if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4142  return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4143  bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4144  if (bytes_read < 0)
4145  return bytes_read;
4146  cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4147  }
4148  avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4149  if (cues_start == -1 || cues_end == -1) return -1;
4150 
4151  // parse the cues
4152  matroska_parse_cues(matroska);
4153 
4154  // cues start
4155  av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4156 
4157  // cues end
4158  av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4159 
4160  // if the file has cues at the start, fix up the init range so that
4161  // it does not include it
4162  if (cues_start <= init_range)
4163  av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4164 
4165  // bandwidth
4166  bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4167  if (bandwidth < 0) return -1;
4168  av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4169 
4170  // check if all clusters start with key frames
4172 
4173  // store cue point timestamps as a comma separated list for checking subsegment alignment in
4174  // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4175  buf = av_malloc_array(s->streams[0]->nb_index_entries, 20);
4176  if (!buf) return -1;
4177  strcpy(buf, "");
4178  for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
4179  int ret = snprintf(buf + end, 20,
4180  "%" PRId64"%s", s->streams[0]->index_entries[i].timestamp,
4181  i != s->streams[0]->nb_index_entries - 1 ? "," : "");
4182  if (ret <= 0 || (ret == 20 && i == s->streams[0]->nb_index_entries - 1)) {
4183  av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4184  av_free(buf);
4185  return AVERROR_INVALIDDATA;
4186  }
4187  end += ret;
4188  }
4189  av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4191 
4192  return 0;
4193 }
4194 
4196 {
4197  char *buf;
4198  int ret = matroska_read_header(s);
4199  int64_t init_range;
4200  MatroskaTrack *tracks;
4201  MatroskaDemuxContext *matroska = s->priv_data;
4202  if (ret) {
4203  av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4204  return -1;
4205  }
4206  if (!matroska->tracks.nb_elem || !s->nb_streams) {
4207  av_log(s, AV_LOG_ERROR, "No track found\n");
4209  goto fail;
4210  }
4211 
4212  if (!matroska->is_live) {
4213  buf = av_asprintf("%g", matroska->duration);
4214  if (!buf) {
4215  ret = AVERROR(ENOMEM);
4216  goto fail;
4217  }
4218  av_dict_set(&s->streams[0]->metadata, DURATION,
4220 
4221  // initialization range
4222  // 5 is the offset of Cluster ID.
4223  init_range = avio_tell(s->pb) - 5;
4224  av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4225  }
4226 
4227  // basename of the file
4228  buf = strrchr(s->url, '/');
4229  av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4230 
4231  // track number
4232  tracks = matroska->tracks.elem;
4233  av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4234 
4235  // parse the cues and populate Cue related fields
4236  if (!matroska->is_live) {
4237  ret = webm_dash_manifest_cues(s, init_range);
4238  if (ret < 0) {
4239  av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4240  goto fail;
4241  }
4242  }
4243 
4244  // use the bandwidth from the command line if it was provided
4245  if (matroska->bandwidth > 0) {
4246  av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4247  matroska->bandwidth, 0);
4248  }
4249  return 0;
4250 fail:
4252  return ret;
4253 }
4256 {
4257  return AVERROR_EOF;
4258 }
4259 
4260 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4261 static const AVOption options[] = {
4262  { "live", "flag indicating that the input is a live file that only has the headers.", OFFSET(is_live), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_DECODING_PARAM },
4263  { "bandwidth", "bandwidth of this stream to be specified in the DASH manifest.", OFFSET(bandwidth), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, AV_OPT_FLAG_DECODING_PARAM },
4264  { NULL },
4265 };
4266 
4267 static const AVClass webm_dash_class = {
4268  .class_name = "WebM DASH Manifest demuxer",
4269  .item_name = av_default_item_name,
4270  .option = options,
4271  .version = LIBAVUTIL_VERSION_INT,
4272 };
4273 
4275  .name = "matroska,webm",
4276  .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4277  .extensions = "mkv,mk3d,mka,mks",
4278  .priv_data_size = sizeof(MatroskaDemuxContext),
4284  .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4285 };
4286 
4288  .name = "webm_dash_manifest",
4289  .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4290  .priv_data_size = sizeof(MatroskaDemuxContext),
4294  .priv_class = &webm_dash_class,
4295 };
AVCOL_PRI_RESERVED
@ AVCOL_PRI_RESERVED
Definition: pixfmt.h:459
MatroskaCluster::timecode
uint64_t timecode
Definition: matroskadec.c:325
AVStream::index_entries
AVIndexEntry * index_entries
Only used if the format does not support seeking natively.
Definition: avformat.h:1094
MATROSKA_ID_ENCODINGENCRYPTION
#define MATROSKA_ID_ENCODINGENCRYPTION
Definition: matroska.h:178
MatroskaDemuxContext::segment_start
int64_t segment_start
Definition: matroskadec.c:359
AV_CODEC_ID_PCM_S16LE
@ AV_CODEC_ID_PCM_S16LE
Definition: codec_id.h:301
MATROSKA_ID_TAGTARGETS_ATTACHUID
#define MATROSKA_ID_TAGTARGETS_ATTACHUID
Definition: matroska.h:214
MATROSKA_ID_CHAPCOUNTRY
#define MATROSKA_ID_CHAPCOUNTRY
Definition: matroska.h:257
AVMasteringDisplayMetadata::has_primaries
int has_primaries
Flag indicating whether the display primaries (and white point) are set.
Definition: mastering_display_metadata.h:62
ebml_read_master
static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length, int64_t pos)
Definition: matroskadec.c:1025
ebml_read_binary
static int ebml_read_binary(AVIOContext *pb, int length, int64_t pos, EbmlBin *bin)
Definition: matroskadec.c:997
av_packet_unref
void av_packet_unref(AVPacket *pkt)
Wipe the packet.
Definition: avpacket.c:605
MATROSKA_ID_VIDEODISPLAYUNIT
#define MATROSKA_ID_VIDEODISPLAYUNIT
Definition: matroska.h:120
mkv_image_mime_tags
static const CodecMime mkv_image_mime_tags[]
Definition: matroskadec.c:746
matroska_parse_laces
static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf, int size, int type, AVIOContext *pb, uint32_t lace_size[256], int *laces)
Definition: matroskadec.c:3032
FF_ENABLE_DEPRECATION_WARNINGS
#define FF_ENABLE_DEPRECATION_WARNINGS
Definition: internal.h:85
MATROSKA_ID_CODECPRIVATE
#define MATROSKA_ID_CODECPRIVATE
Definition: matroska.h:89
MATROSKA_ID_TRACKNUMBER
#define MATROSKA_ID_TRACKNUMBER
Definition: matroska.h:78
MatroskaTag
Definition: matroskadec.c:282
ff_packet_list_get
int ff_packet_list_get(AVPacketList **head, AVPacketList **tail, AVPacket *pkt)
Remove the oldest AVPacket in the list and return it.
Definition: utils.c:1560
AVMEDIA_TYPE_SUBTITLE
@ AVMEDIA_TYPE_SUBTITLE
Definition: avutil.h:204
bit_depth
static void bit_depth(AudioStatsContext *s, uint64_t mask, uint64_t imask, AVRational *depth)
Definition: af_astats.c:254
MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL
@ MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL
Definition: matroska.h:308
AV_LOG_WARNING
#define AV_LOG_WARNING
Something somehow does not look correct.
Definition: log.h:182
MATROSKA_ID_CHAPSTRING
#define MATROSKA_ID_CHAPSTRING
Definition: matroska.h:255
MATROSKA_ID_ENCODINGSIGHASHALGO
#define MATROSKA_ID_ENCODINGSIGHASHALGO
Definition: matroska.h:183
MatroskaTrack::codec_priv
EbmlBin codec_priv
Definition: matroskadec.c:229
MATROSKA_ID_TAGTARGETS
#define MATROSKA_ID_TAGTARGETS
Definition: matroska.h:209
AVMasteringDisplayMetadata::max_luminance
AVRational max_luminance
Max luminance of mastering display (cd/m^2).
Definition: mastering_display_metadata.h:57
AVSphericalProjection
AVSphericalProjection
Projection of the video surface(s) on a sphere.
Definition: spherical.h:51
AVCodecParameters::extradata
uint8_t * extradata
Extra binary data needed for initializing the decoder, codec-dependent.
Definition: codec_par.h:74
EbmlSyntax::id
uint32_t id
Definition: matroskadec.c:92
name
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf default minimum maximum flags name is the option name
Definition: writing_filters.txt:88
MATROSKA_ID_CLUSTERTIMECODE
#define MATROSKA_ID_CLUSTERTIMECODE
Definition: matroska.h:224
level
uint8_t level
Definition: svq3.c:210
matroska_attachment
static EbmlSyntax matroska_attachment[]
Definition: matroskadec.c:580
AV_CODEC_ID_AC3
@ AV_CODEC_ID_AC3
Definition: codec_id.h:413
ebml_header
static EbmlSyntax ebml_header[]
Definition: matroskadec.c:400
MatroskaDemuxContext::current_id
uint32_t current_id
Definition: matroskadec.c:342
MatroskaDemuxContext::bandwidth
int bandwidth
Definition: matroskadec.c:384</