Subversion Repositories Kolibri OS

Rev

Blame | Last modification | View Log | RSS feed

  1. /*
  2.  * Matroska file demuxer
  3.  * Copyright (c) 2003-2008 The FFmpeg Project
  4.  *
  5.  * This file is part of FFmpeg.
  6.  *
  7.  * FFmpeg is free software; you can redistribute it and/or
  8.  * modify it under the terms of the GNU Lesser General Public
  9.  * License as published by the Free Software Foundation; either
  10.  * version 2.1 of the License, or (at your option) any later version.
  11.  *
  12.  * FFmpeg is distributed in the hope that it will be useful,
  13.  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15.  * Lesser General Public License for more details.
  16.  *
  17.  * You should have received a copy of the GNU Lesser General Public
  18.  * License along with FFmpeg; if not, write to the Free Software
  19.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20.  */
  21.  
  22. /**
  23.  * @file
  24.  * Matroska file demuxer
  25.  * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
  26.  * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
  27.  * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
  28.  * @see specs available on the Matroska project page: http://www.matroska.org/
  29.  */
  30.  
  31. #include "config.h"
  32.  
  33. #include <inttypes.h>
  34. #include <stdio.h>
  35.  
  36. #include "libavutil/avstring.h"
  37. #include "libavutil/base64.h"
  38. #include "libavutil/dict.h"
  39. #include "libavutil/intfloat.h"
  40. #include "libavutil/intreadwrite.h"
  41. #include "libavutil/lzo.h"
  42. #include "libavutil/mathematics.h"
  43. #include "libavutil/opt.h"
  44. #include "libavutil/time_internal.h"
  45.  
  46. #include "libavcodec/bytestream.h"
  47. #include "libavcodec/flac.h"
  48. #include "libavcodec/mpeg4audio.h"
  49.  
  50. #include "avformat.h"
  51. #include "avio_internal.h"
  52. #include "internal.h"
  53. #include "isom.h"
  54. #include "matroska.h"
  55. #include "oggdec.h"
  56. /* For ff_codec_get_id(). */
  57. #include "riff.h"
  58. #include "rmsipr.h"
  59.  
  60. #if CONFIG_BZLIB
  61. #include <bzlib.h>
  62. #endif
  63. #if CONFIG_ZLIB
  64. #include <zlib.h>
  65. #endif
  66.  
  67. typedef enum {
  68.     EBML_NONE,
  69.     EBML_UINT,
  70.     EBML_FLOAT,
  71.     EBML_STR,
  72.     EBML_UTF8,
  73.     EBML_BIN,
  74.     EBML_NEST,
  75.     EBML_LEVEL1,
  76.     EBML_PASS,
  77.     EBML_STOP,
  78.     EBML_SINT,
  79.     EBML_TYPE_COUNT
  80. } EbmlType;
  81.  
  82. typedef const struct EbmlSyntax {
  83.     uint32_t id;
  84.     EbmlType type;
  85.     int list_elem_size;
  86.     int data_offset;
  87.     union {
  88.         uint64_t    u;
  89.         double      f;
  90.         const char *s;
  91.         const struct EbmlSyntax *n;
  92.     } def;
  93. } EbmlSyntax;
  94.  
  95. typedef struct EbmlList {
  96.     int nb_elem;
  97.     void *elem;
  98. } EbmlList;
  99.  
  100. typedef struct EbmlBin {
  101.     int      size;
  102.     uint8_t *data;
  103.     int64_t  pos;
  104. } EbmlBin;
  105.  
  106. typedef struct Ebml {
  107.     uint64_t version;
  108.     uint64_t max_size;
  109.     uint64_t id_length;
  110.     char    *doctype;
  111.     uint64_t doctype_version;
  112. } Ebml;
  113.  
  114. typedef struct MatroskaTrackCompression {
  115.     uint64_t algo;
  116.     EbmlBin  settings;
  117. } MatroskaTrackCompression;
  118.  
  119. typedef struct MatroskaTrackEncryption {
  120.     uint64_t algo;
  121.     EbmlBin  key_id;
  122. } MatroskaTrackEncryption;
  123.  
  124. typedef struct MatroskaTrackEncoding {
  125.     uint64_t scope;
  126.     uint64_t type;
  127.     MatroskaTrackCompression compression;
  128.     MatroskaTrackEncryption encryption;
  129. } MatroskaTrackEncoding;
  130.  
  131. typedef struct MatroskaTrackVideo {
  132.     double   frame_rate;
  133.     uint64_t display_width;
  134.     uint64_t display_height;
  135.     uint64_t pixel_width;
  136.     uint64_t pixel_height;
  137.     EbmlBin color_space;
  138.     uint64_t stereo_mode;
  139.     uint64_t alpha_mode;
  140. } MatroskaTrackVideo;
  141.  
  142. typedef struct MatroskaTrackAudio {
  143.     double   samplerate;
  144.     double   out_samplerate;
  145.     uint64_t bitdepth;
  146.     uint64_t channels;
  147.  
  148.     /* real audio header (extracted from extradata) */
  149.     int      coded_framesize;
  150.     int      sub_packet_h;
  151.     int      frame_size;
  152.     int      sub_packet_size;
  153.     int      sub_packet_cnt;
  154.     int      pkt_cnt;
  155.     uint64_t buf_timecode;
  156.     uint8_t *buf;
  157. } MatroskaTrackAudio;
  158.  
  159. typedef struct MatroskaTrackPlane {
  160.     uint64_t uid;
  161.     uint64_t type;
  162. } MatroskaTrackPlane;
  163.  
  164. typedef struct MatroskaTrackOperation {
  165.     EbmlList combine_planes;
  166. } MatroskaTrackOperation;
  167.  
  168. typedef struct MatroskaTrack {
  169.     uint64_t num;
  170.     uint64_t uid;
  171.     uint64_t type;
  172.     char    *name;
  173.     char    *codec_id;
  174.     EbmlBin  codec_priv;
  175.     char    *language;
  176.     double time_scale;
  177.     uint64_t default_duration;
  178.     uint64_t flag_default;
  179.     uint64_t flag_forced;
  180.     uint64_t seek_preroll;
  181.     MatroskaTrackVideo video;
  182.     MatroskaTrackAudio audio;
  183.     MatroskaTrackOperation operation;
  184.     EbmlList encodings;
  185.     uint64_t codec_delay;
  186.  
  187.     AVStream *stream;
  188.     int64_t end_timecode;
  189.     int ms_compat;
  190.     uint64_t max_block_additional_id;
  191. } MatroskaTrack;
  192.  
  193. typedef struct MatroskaAttachment {
  194.     uint64_t uid;
  195.     char *filename;
  196.     char *mime;
  197.     EbmlBin bin;
  198.  
  199.     AVStream *stream;
  200. } MatroskaAttachment;
  201.  
  202. typedef struct MatroskaChapter {
  203.     uint64_t start;
  204.     uint64_t end;
  205.     uint64_t uid;
  206.     char    *title;
  207.  
  208.     AVChapter *chapter;
  209. } MatroskaChapter;
  210.  
  211. typedef struct MatroskaIndexPos {
  212.     uint64_t track;
  213.     uint64_t pos;
  214. } MatroskaIndexPos;
  215.  
  216. typedef struct MatroskaIndex {
  217.     uint64_t time;
  218.     EbmlList pos;
  219. } MatroskaIndex;
  220.  
  221. typedef struct MatroskaTag {
  222.     char *name;
  223.     char *string;
  224.     char *lang;
  225.     uint64_t def;
  226.     EbmlList sub;
  227. } MatroskaTag;
  228.  
  229. typedef struct MatroskaTagTarget {
  230.     char    *type;
  231.     uint64_t typevalue;
  232.     uint64_t trackuid;
  233.     uint64_t chapteruid;
  234.     uint64_t attachuid;
  235. } MatroskaTagTarget;
  236.  
  237. typedef struct MatroskaTags {
  238.     MatroskaTagTarget target;
  239.     EbmlList tag;
  240. } MatroskaTags;
  241.  
  242. typedef struct MatroskaSeekhead {
  243.     uint64_t id;
  244.     uint64_t pos;
  245. } MatroskaSeekhead;
  246.  
  247. typedef struct MatroskaLevel {
  248.     uint64_t start;
  249.     uint64_t length;
  250. } MatroskaLevel;
  251.  
  252. typedef struct MatroskaCluster {
  253.     uint64_t timecode;
  254.     EbmlList blocks;
  255. } MatroskaCluster;
  256.  
  257. typedef struct MatroskaLevel1Element {
  258.     uint64_t id;
  259.     uint64_t pos;
  260.     int parsed;
  261. } MatroskaLevel1Element;
  262.  
  263. typedef struct MatroskaDemuxContext {
  264.     const AVClass *class;
  265.     AVFormatContext *ctx;
  266.  
  267.     /* EBML stuff */
  268.     int num_levels;
  269.     MatroskaLevel levels[EBML_MAX_DEPTH];
  270.     int level_up;
  271.     uint32_t current_id;
  272.  
  273.     uint64_t time_scale;
  274.     double   duration;
  275.     char    *title;
  276.     char    *muxingapp;
  277.     EbmlBin date_utc;
  278.     EbmlList tracks;
  279.     EbmlList attachments;
  280.     EbmlList chapters;
  281.     EbmlList index;
  282.     EbmlList tags;
  283.     EbmlList seekhead;
  284.  
  285.     /* byte position of the segment inside the stream */
  286.     int64_t segment_start;
  287.  
  288.     /* the packet queue */
  289.     AVPacket **packets;
  290.     int num_packets;
  291.     AVPacket *prev_pkt;
  292.  
  293.     int done;
  294.  
  295.     /* What to skip before effectively reading a packet. */
  296.     int skip_to_keyframe;
  297.     uint64_t skip_to_timecode;
  298.  
  299.     /* File has a CUES element, but we defer parsing until it is needed. */
  300.     int cues_parsing_deferred;
  301.  
  302.     /* Level1 elements and whether they were read yet */
  303.     MatroskaLevel1Element level1_elems[64];
  304.     int num_level1_elems;
  305.  
  306.     int current_cluster_num_blocks;
  307.     int64_t current_cluster_pos;
  308.     MatroskaCluster current_cluster;
  309.  
  310.     /* File has SSA subtitles which prevent incremental cluster parsing. */
  311.     int contains_ssa;
  312.  
  313.     /* WebM DASH Manifest live flag/ */
  314.     int is_live;
  315. } MatroskaDemuxContext;
  316.  
  317. typedef struct MatroskaBlock {
  318.     uint64_t duration;
  319.     int64_t  reference;
  320.     uint64_t non_simple;
  321.     EbmlBin  bin;
  322.     uint64_t additional_id;
  323.     EbmlBin  additional;
  324.     int64_t discard_padding;
  325. } MatroskaBlock;
  326.  
  327. static const EbmlSyntax ebml_header[] = {
  328.     { EBML_ID_EBMLREADVERSION,    EBML_UINT, 0, offsetof(Ebml, version),         { .u = EBML_VERSION } },
  329.     { EBML_ID_EBMLMAXSIZELENGTH,  EBML_UINT, 0, offsetof(Ebml, max_size),        { .u = 8 } },
  330.     { EBML_ID_EBMLMAXIDLENGTH,    EBML_UINT, 0, offsetof(Ebml, id_length),       { .u = 4 } },
  331.     { EBML_ID_DOCTYPE,            EBML_STR,  0, offsetof(Ebml, doctype),         { .s = "(none)" } },
  332.     { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
  333.     { EBML_ID_EBMLVERSION,        EBML_NONE },
  334.     { EBML_ID_DOCTYPEVERSION,     EBML_NONE },
  335.     { 0 }
  336. };
  337.  
  338. static const EbmlSyntax ebml_syntax[] = {
  339.     { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
  340.     { 0 }
  341. };
  342.  
  343. static const EbmlSyntax matroska_info[] = {
  344.     { MATROSKA_ID_TIMECODESCALE, EBML_UINT,  0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
  345.     { MATROSKA_ID_DURATION,      EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
  346.     { MATROSKA_ID_TITLE,         EBML_UTF8,  0, offsetof(MatroskaDemuxContext, title) },
  347.     { MATROSKA_ID_WRITINGAPP,    EBML_NONE },
  348.     { MATROSKA_ID_MUXINGAPP,     EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
  349.     { MATROSKA_ID_DATEUTC,       EBML_BIN,  0, offsetof(MatroskaDemuxContext, date_utc) },
  350.     { MATROSKA_ID_SEGMENTUID,    EBML_NONE },
  351.     { 0 }
  352. };
  353.  
  354. static const EbmlSyntax matroska_track_video[] = {
  355.     { MATROSKA_ID_VIDEOFRAMERATE,      EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
  356.     { MATROSKA_ID_VIDEODISPLAYWIDTH,   EBML_UINT,  0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
  357.     { MATROSKA_ID_VIDEODISPLAYHEIGHT,  EBML_UINT,  0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
  358.     { MATROSKA_ID_VIDEOPIXELWIDTH,     EBML_UINT,  0, offsetof(MatroskaTrackVideo, pixel_width) },
  359.     { MATROSKA_ID_VIDEOPIXELHEIGHT,    EBML_UINT,  0, offsetof(MatroskaTrackVideo, pixel_height) },
  360.     { MATROSKA_ID_VIDEOCOLORSPACE,     EBML_BIN,   0, offsetof(MatroskaTrackVideo, color_space) },
  361.     { MATROSKA_ID_VIDEOALPHAMODE,      EBML_UINT,  0, offsetof(MatroskaTrackVideo, alpha_mode) },
  362.     { MATROSKA_ID_VIDEOPIXELCROPB,     EBML_NONE },
  363.     { MATROSKA_ID_VIDEOPIXELCROPT,     EBML_NONE },
  364.     { MATROSKA_ID_VIDEOPIXELCROPL,     EBML_NONE },
  365.     { MATROSKA_ID_VIDEOPIXELCROPR,     EBML_NONE },
  366.     { MATROSKA_ID_VIDEODISPLAYUNIT,    EBML_NONE },
  367.     { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_NONE },
  368.     { MATROSKA_ID_VIDEOSTEREOMODE,     EBML_UINT,  0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
  369.     { MATROSKA_ID_VIDEOASPECTRATIO,    EBML_NONE },
  370.     { 0 }
  371. };
  372.  
  373. static const EbmlSyntax matroska_track_audio[] = {
  374.     { MATROSKA_ID_AUDIOSAMPLINGFREQ,    EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
  375.     { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
  376.     { MATROSKA_ID_AUDIOBITDEPTH,        EBML_UINT,  0, offsetof(MatroskaTrackAudio, bitdepth) },
  377.     { MATROSKA_ID_AUDIOCHANNELS,        EBML_UINT,  0, offsetof(MatroskaTrackAudio, channels),   { .u = 1 } },
  378.     { 0 }
  379. };
  380.  
  381. static const EbmlSyntax matroska_track_encoding_compression[] = {
  382.     { MATROSKA_ID_ENCODINGCOMPALGO,     EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
  383.     { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN,  0, offsetof(MatroskaTrackCompression, settings) },
  384.     { 0 }
  385. };
  386.  
  387. static const EbmlSyntax matroska_track_encoding_encryption[] = {
  388.     { MATROSKA_ID_ENCODINGENCALGO,        EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
  389.     { MATROSKA_ID_ENCODINGENCKEYID,       EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
  390.     { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
  391.     { MATROSKA_ID_ENCODINGSIGALGO,        EBML_NONE },
  392.     { MATROSKA_ID_ENCODINGSIGHASHALGO,    EBML_NONE },
  393.     { MATROSKA_ID_ENCODINGSIGKEYID,       EBML_NONE },
  394.     { MATROSKA_ID_ENCODINGSIGNATURE,      EBML_NONE },
  395.     { 0 }
  396. };
  397. static const EbmlSyntax matroska_track_encoding[] = {
  398.     { MATROSKA_ID_ENCODINGSCOPE,       EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope),       { .u = 1 } },
  399.     { MATROSKA_ID_ENCODINGTYPE,        EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type),        { .u = 0 } },
  400.     { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
  401.     { MATROSKA_ID_ENCODINGENCRYPTION,  EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption),  { .n = matroska_track_encoding_encryption } },
  402.     { MATROSKA_ID_ENCODINGORDER,       EBML_NONE },
  403.     { 0 }
  404. };
  405.  
  406. static const EbmlSyntax matroska_track_encodings[] = {
  407.     { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
  408.     { 0 }
  409. };
  410.  
  411. static const EbmlSyntax matroska_track_plane[] = {
  412.     { MATROSKA_ID_TRACKPLANEUID,  EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
  413.     { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
  414.     { 0 }
  415. };
  416.  
  417. static const EbmlSyntax matroska_track_combine_planes[] = {
  418.     { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
  419.     { 0 }
  420. };
  421.  
  422. static const EbmlSyntax matroska_track_operation[] = {
  423.     { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
  424.     { 0 }
  425. };
  426.  
  427. static const EbmlSyntax matroska_track[] = {
  428.     { MATROSKA_ID_TRACKNUMBER,           EBML_UINT,  0, offsetof(MatroskaTrack, num) },
  429.     { MATROSKA_ID_TRACKNAME,             EBML_UTF8,  0, offsetof(MatroskaTrack, name) },
  430.     { MATROSKA_ID_TRACKUID,              EBML_UINT,  0, offsetof(MatroskaTrack, uid) },
  431.     { MATROSKA_ID_TRACKTYPE,             EBML_UINT,  0, offsetof(MatroskaTrack, type) },
  432.     { MATROSKA_ID_CODECID,               EBML_STR,   0, offsetof(MatroskaTrack, codec_id) },
  433.     { MATROSKA_ID_CODECPRIVATE,          EBML_BIN,   0, offsetof(MatroskaTrack, codec_priv) },
  434.     { MATROSKA_ID_CODECDELAY,            EBML_UINT,  0, offsetof(MatroskaTrack, codec_delay) },
  435.     { MATROSKA_ID_TRACKLANGUAGE,         EBML_UTF8,  0, offsetof(MatroskaTrack, language),     { .s = "eng" } },
  436.     { MATROSKA_ID_TRACKDEFAULTDURATION,  EBML_UINT,  0, offsetof(MatroskaTrack, default_duration) },
  437.     { MATROSKA_ID_TRACKTIMECODESCALE,    EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale),   { .f = 1.0 } },
  438.     { MATROSKA_ID_TRACKFLAGDEFAULT,      EBML_UINT,  0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
  439.     { MATROSKA_ID_TRACKFLAGFORCED,       EBML_UINT,  0, offsetof(MatroskaTrack, flag_forced),  { .u = 0 } },
  440.     { MATROSKA_ID_TRACKVIDEO,            EBML_NEST,  0, offsetof(MatroskaTrack, video),        { .n = matroska_track_video } },
  441.     { MATROSKA_ID_TRACKAUDIO,            EBML_NEST,  0, offsetof(MatroskaTrack, audio),        { .n = matroska_track_audio } },
  442.     { MATROSKA_ID_TRACKOPERATION,        EBML_NEST,  0, offsetof(MatroskaTrack, operation),    { .n = matroska_track_operation } },
  443.     { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST,  0, 0,                                     { .n = matroska_track_encodings } },
  444.     { MATROSKA_ID_TRACKMAXBLKADDID,      EBML_UINT,  0, offsetof(MatroskaTrack, max_block_additional_id) },
  445.     { MATROSKA_ID_SEEKPREROLL,           EBML_UINT,  0, offsetof(MatroskaTrack, seek_preroll) },
  446.     { MATROSKA_ID_TRACKFLAGENABLED,      EBML_NONE },
  447.     { MATROSKA_ID_TRACKFLAGLACING,       EBML_NONE },
  448.     { MATROSKA_ID_CODECNAME,             EBML_NONE },
  449.     { MATROSKA_ID_CODECDECODEALL,        EBML_NONE },
  450.     { MATROSKA_ID_CODECINFOURL,          EBML_NONE },
  451.     { MATROSKA_ID_CODECDOWNLOADURL,      EBML_NONE },
  452.     { MATROSKA_ID_TRACKMINCACHE,         EBML_NONE },
  453.     { MATROSKA_ID_TRACKMAXCACHE,         EBML_NONE },
  454.     { 0 }
  455. };
  456.  
  457. static const EbmlSyntax matroska_tracks[] = {
  458.     { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
  459.     { 0 }
  460. };
  461.  
  462. static const EbmlSyntax matroska_attachment[] = {
  463.     { MATROSKA_ID_FILEUID,      EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
  464.     { MATROSKA_ID_FILENAME,     EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
  465.     { MATROSKA_ID_FILEMIMETYPE, EBML_STR,  0, offsetof(MatroskaAttachment, mime) },
  466.     { MATROSKA_ID_FILEDATA,     EBML_BIN,  0, offsetof(MatroskaAttachment, bin) },
  467.     { MATROSKA_ID_FILEDESC,     EBML_NONE },
  468.     { 0 }
  469. };
  470.  
  471. static const EbmlSyntax matroska_attachments[] = {
  472.     { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
  473.     { 0 }
  474. };
  475.  
  476. static const EbmlSyntax matroska_chapter_display[] = {
  477.     { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
  478.     { MATROSKA_ID_CHAPLANG,   EBML_NONE },
  479.     { 0 }
  480. };
  481.  
  482. static const EbmlSyntax matroska_chapter_entry[] = {
  483.     { MATROSKA_ID_CHAPTERTIMESTART,   EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
  484.     { MATROSKA_ID_CHAPTERTIMEEND,     EBML_UINT, 0, offsetof(MatroskaChapter, end),   { .u = AV_NOPTS_VALUE } },
  485.     { MATROSKA_ID_CHAPTERUID,         EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
  486.     { MATROSKA_ID_CHAPTERDISPLAY,     EBML_NEST, 0,                        0,         { .n = matroska_chapter_display } },
  487.     { MATROSKA_ID_CHAPTERFLAGHIDDEN,  EBML_NONE },
  488.     { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
  489.     { MATROSKA_ID_CHAPTERPHYSEQUIV,   EBML_NONE },
  490.     { MATROSKA_ID_CHAPTERATOM,        EBML_NONE },
  491.     { 0 }
  492. };
  493.  
  494. static const EbmlSyntax matroska_chapter[] = {
  495.     { MATROSKA_ID_CHAPTERATOM,        EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
  496.     { MATROSKA_ID_EDITIONUID,         EBML_NONE },
  497.     { MATROSKA_ID_EDITIONFLAGHIDDEN,  EBML_NONE },
  498.     { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
  499.     { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
  500.     { 0 }
  501. };
  502.  
  503. static const EbmlSyntax matroska_chapters[] = {
  504.     { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
  505.     { 0 }
  506. };
  507.  
  508. static const EbmlSyntax matroska_index_pos[] = {
  509.     { MATROSKA_ID_CUETRACK,           EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
  510.     { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
  511.     { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
  512.     { MATROSKA_ID_CUEDURATION,        EBML_NONE },
  513.     { MATROSKA_ID_CUEBLOCKNUMBER,     EBML_NONE },
  514.     { 0 }
  515. };
  516.  
  517. static const EbmlSyntax matroska_index_entry[] = {
  518.     { MATROSKA_ID_CUETIME,          EBML_UINT, 0,                        offsetof(MatroskaIndex, time) },
  519.     { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
  520.     { 0 }
  521. };
  522.  
  523. static const EbmlSyntax matroska_index[] = {
  524.     { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
  525.     { 0 }
  526. };
  527.  
  528. static const EbmlSyntax matroska_simpletag[] = {
  529.     { MATROSKA_ID_TAGNAME,        EBML_UTF8, 0,                   offsetof(MatroskaTag, name) },
  530.     { MATROSKA_ID_TAGSTRING,      EBML_UTF8, 0,                   offsetof(MatroskaTag, string) },
  531.     { MATROSKA_ID_TAGLANG,        EBML_STR,  0,                   offsetof(MatroskaTag, lang), { .s = "und" } },
  532.     { MATROSKA_ID_TAGDEFAULT,     EBML_UINT, 0,                   offsetof(MatroskaTag, def) },
  533.     { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0,                   offsetof(MatroskaTag, def) },
  534.     { MATROSKA_ID_SIMPLETAG,      EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub),  { .n = matroska_simpletag } },
  535.     { 0 }
  536. };
  537.  
  538. static const EbmlSyntax matroska_tagtargets[] = {
  539.     { MATROSKA_ID_TAGTARGETS_TYPE,       EBML_STR,  0, offsetof(MatroskaTagTarget, type) },
  540.     { MATROSKA_ID_TAGTARGETS_TYPEVALUE,  EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
  541.     { MATROSKA_ID_TAGTARGETS_TRACKUID,   EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
  542.     { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
  543.     { MATROSKA_ID_TAGTARGETS_ATTACHUID,  EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
  544.     { 0 }
  545. };
  546.  
  547. static const EbmlSyntax matroska_tag[] = {
  548.     { MATROSKA_ID_SIMPLETAG,  EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag),    { .n = matroska_simpletag } },
  549.     { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0,                   offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
  550.     { 0 }
  551. };
  552.  
  553. static const EbmlSyntax matroska_tags[] = {
  554.     { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
  555.     { 0 }
  556. };
  557.  
  558. static const EbmlSyntax matroska_seekhead_entry[] = {
  559.     { MATROSKA_ID_SEEKID,       EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
  560.     { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
  561.     { 0 }
  562. };
  563.  
  564. static const EbmlSyntax matroska_seekhead[] = {
  565.     { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
  566.     { 0 }
  567. };
  568.  
  569. static const EbmlSyntax matroska_segment[] = {
  570.     { MATROSKA_ID_INFO,        EBML_LEVEL1, 0, 0, { .n = matroska_info } },
  571.     { MATROSKA_ID_TRACKS,      EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
  572.     { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
  573.     { MATROSKA_ID_CHAPTERS,    EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
  574.     { MATROSKA_ID_CUES,        EBML_LEVEL1, 0, 0, { .n = matroska_index } },
  575.     { MATROSKA_ID_TAGS,        EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
  576.     { MATROSKA_ID_SEEKHEAD,    EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
  577.     { MATROSKA_ID_CLUSTER,     EBML_STOP },
  578.     { 0 }
  579. };
  580.  
  581. static const EbmlSyntax matroska_segments[] = {
  582.     { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
  583.     { 0 }
  584. };
  585.  
  586. static const EbmlSyntax matroska_blockmore[] = {
  587.     { MATROSKA_ID_BLOCKADDID,      EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
  588.     { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN,  0, offsetof(MatroskaBlock,additional) },
  589.     { 0 }
  590. };
  591.  
  592. static const EbmlSyntax matroska_blockadditions[] = {
  593.     { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
  594.     { 0 }
  595. };
  596.  
  597. static const EbmlSyntax matroska_blockgroup[] = {
  598.     { MATROSKA_ID_BLOCK,          EBML_BIN,  0, offsetof(MatroskaBlock, bin) },
  599.     { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
  600.     { MATROSKA_ID_SIMPLEBLOCK,    EBML_BIN,  0, offsetof(MatroskaBlock, bin) },
  601.     { MATROSKA_ID_BLOCKDURATION,  EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
  602.     { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
  603.     { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference) },
  604.     { MATROSKA_ID_CODECSTATE,     EBML_NONE },
  605.     {                          1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
  606.     { 0 }
  607. };
  608.  
  609. static const EbmlSyntax matroska_cluster[] = {
  610.     { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0,                     offsetof(MatroskaCluster, timecode) },
  611.     { MATROSKA_ID_BLOCKGROUP,      EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
  612.     { MATROSKA_ID_SIMPLEBLOCK,     EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
  613.     { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
  614.     { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
  615.     { 0 }
  616. };
  617.  
  618. static const EbmlSyntax matroska_clusters[] = {
  619.     { MATROSKA_ID_CLUSTER,  EBML_NEST, 0, 0, { .n = matroska_cluster } },
  620.     { MATROSKA_ID_INFO,     EBML_NONE },
  621.     { MATROSKA_ID_CUES,     EBML_NONE },
  622.     { MATROSKA_ID_TAGS,     EBML_NONE },
  623.     { MATROSKA_ID_SEEKHEAD, EBML_NONE },
  624.     { 0 }
  625. };
  626.  
  627. static const EbmlSyntax matroska_cluster_incremental_parsing[] = {
  628.     { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0,                     offsetof(MatroskaCluster, timecode) },
  629.     { MATROSKA_ID_BLOCKGROUP,      EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
  630.     { MATROSKA_ID_SIMPLEBLOCK,     EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
  631.     { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
  632.     { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
  633.     { MATROSKA_ID_INFO,            EBML_NONE },
  634.     { MATROSKA_ID_CUES,            EBML_NONE },
  635.     { MATROSKA_ID_TAGS,            EBML_NONE },
  636.     { MATROSKA_ID_SEEKHEAD,        EBML_NONE },
  637.     { MATROSKA_ID_CLUSTER,         EBML_STOP },
  638.     { 0 }
  639. };
  640.  
  641. static const EbmlSyntax matroska_cluster_incremental[] = {
  642.     { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
  643.     { MATROSKA_ID_BLOCKGROUP,      EBML_STOP },
  644.     { MATROSKA_ID_SIMPLEBLOCK,     EBML_STOP },
  645.     { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
  646.     { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
  647.     { 0 }
  648. };
  649.  
  650. static const EbmlSyntax matroska_clusters_incremental[] = {
  651.     { MATROSKA_ID_CLUSTER,  EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
  652.     { MATROSKA_ID_INFO,     EBML_NONE },
  653.     { MATROSKA_ID_CUES,     EBML_NONE },
  654.     { MATROSKA_ID_TAGS,     EBML_NONE },
  655.     { MATROSKA_ID_SEEKHEAD, EBML_NONE },
  656.     { 0 }
  657. };
  658.  
  659. static const char *const matroska_doctypes[] = { "matroska", "webm" };
  660.  
  661. static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
  662. {
  663.     AVIOContext *pb = matroska->ctx->pb;
  664.     uint32_t id;
  665.     matroska->current_id = 0;
  666.     matroska->num_levels = 0;
  667.  
  668.     /* seek to next position to resync from */
  669.     if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
  670.         goto eof;
  671.  
  672.     id = avio_rb32(pb);
  673.  
  674.     // try to find a toplevel element
  675.     while (!avio_feof(pb)) {
  676.         if (id == MATROSKA_ID_INFO     || id == MATROSKA_ID_TRACKS      ||
  677.             id == MATROSKA_ID_CUES     || id == MATROSKA_ID_TAGS        ||
  678.             id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
  679.             id == MATROSKA_ID_CLUSTER  || id == MATROSKA_ID_CHAPTERS) {
  680.             matroska->current_id = id;
  681.             return 0;
  682.         }
  683.         id = (id << 8) | avio_r8(pb);
  684.     }
  685.  
  686. eof:
  687.     matroska->done = 1;
  688.     return AVERROR_EOF;
  689. }
  690.  
  691. /*
  692.  * Return: Whether we reached the end of a level in the hierarchy or not.
  693.  */
  694. static int ebml_level_end(MatroskaDemuxContext *matroska)
  695. {
  696.     AVIOContext *pb = matroska->ctx->pb;
  697.     int64_t pos = avio_tell(pb);
  698.  
  699.     if (matroska->num_levels > 0) {
  700.         MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
  701.         if (pos - level->start >= level->length || matroska->current_id) {
  702.             matroska->num_levels--;
  703.             return 1;
  704.         }
  705.     }
  706.     return (matroska->is_live && matroska->ctx->pb->eof_reached) ? 1 : 0;
  707. }
  708.  
  709. /*
  710.  * Read: an "EBML number", which is defined as a variable-length
  711.  * array of bytes. The first byte indicates the length by giving a
  712.  * number of 0-bits followed by a one. The position of the first
  713.  * "one" bit inside the first byte indicates the length of this
  714.  * number.
  715.  * Returns: number of bytes read, < 0 on error
  716.  */
  717. static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
  718.                          int max_size, uint64_t *number)
  719. {
  720.     int read = 1, n = 1;
  721.     uint64_t total = 0;
  722.  
  723.     /* The first byte tells us the length in bytes - avio_r8() can normally
  724.      * return 0, but since that's not a valid first ebmlID byte, we can
  725.      * use it safely here to catch EOS. */
  726.     if (!(total = avio_r8(pb))) {
  727.         /* we might encounter EOS here */
  728.         if (!avio_feof(pb)) {
  729.             int64_t pos = avio_tell(pb);
  730.             av_log(matroska->ctx, AV_LOG_ERROR,
  731.                    "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
  732.                    pos, pos);
  733.             return pb->error ? pb->error : AVERROR(EIO);
  734.         }
  735.         return AVERROR_EOF;
  736.     }
  737.  
  738.     /* get the length of the EBML number */
  739.     read = 8 - ff_log2_tab[total];
  740.     if (read > max_size) {
  741.         int64_t pos = avio_tell(pb) - 1;
  742.         av_log(matroska->ctx, AV_LOG_ERROR,
  743.                "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
  744.                (uint8_t) total, pos, pos);
  745.         return AVERROR_INVALIDDATA;
  746.     }
  747.  
  748.     /* read out length */
  749.     total ^= 1 << ff_log2_tab[total];
  750.     while (n++ < read)
  751.         total = (total << 8) | avio_r8(pb);
  752.  
  753.     *number = total;
  754.  
  755.     return read;
  756. }
  757.  
  758. /**
  759.  * Read a EBML length value.
  760.  * This needs special handling for the "unknown length" case which has multiple
  761.  * encodings.
  762.  */
  763. static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
  764.                             uint64_t *number)
  765. {
  766.     int res = ebml_read_num(matroska, pb, 8, number);
  767.     if (res > 0 && *number + 1 == 1ULL << (7 * res))
  768.         *number = 0xffffffffffffffULL;
  769.     return res;
  770. }
  771.  
  772. /*
  773.  * Read the next element as an unsigned int.
  774.  * 0 is success, < 0 is failure.
  775.  */
  776. static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
  777. {
  778.     int n = 0;
  779.  
  780.     if (size > 8)
  781.         return AVERROR_INVALIDDATA;
  782.  
  783.     /* big-endian ordering; build up number */
  784.     *num = 0;
  785.     while (n++ < size)
  786.         *num = (*num << 8) | avio_r8(pb);
  787.  
  788.     return 0;
  789. }
  790.  
  791. /*
  792.  * Read the next element as a signed int.
  793.  * 0 is success, < 0 is failure.
  794.  */
  795. static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
  796. {
  797.     int n = 1;
  798.  
  799.     if (size > 8)
  800.         return AVERROR_INVALIDDATA;
  801.  
  802.     if (size == 0) {
  803.         *num = 0;
  804.     } else {
  805.         *num = sign_extend(avio_r8(pb), 8);
  806.  
  807.         /* big-endian ordering; build up number */
  808.         while (n++ < size)
  809.             *num = ((uint64_t)*num << 8) | avio_r8(pb);
  810.     }
  811.  
  812.     return 0;
  813. }
  814.  
  815. /*
  816.  * Read the next element as a float.
  817.  * 0 is success, < 0 is failure.
  818.  */
  819. static int ebml_read_float(AVIOContext *pb, int size, double *num)
  820. {
  821.     if (size == 0)
  822.         *num = 0;
  823.     else if (size == 4)
  824.         *num = av_int2float(avio_rb32(pb));
  825.     else if (size == 8)
  826.         *num = av_int2double(avio_rb64(pb));
  827.     else
  828.         return AVERROR_INVALIDDATA;
  829.  
  830.     return 0;
  831. }
  832.  
  833. /*
  834.  * Read the next element as an ASCII string.
  835.  * 0 is success, < 0 is failure.
  836.  */
  837. static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
  838. {
  839.     char *res;
  840.  
  841.     /* EBML strings are usually not 0-terminated, so we allocate one
  842.      * byte more, read the string and NULL-terminate it ourselves. */
  843.     if (!(res = av_malloc(size + 1)))
  844.         return AVERROR(ENOMEM);
  845.     if (avio_read(pb, (uint8_t *) res, size) != size) {
  846.         av_free(res);
  847.         return AVERROR(EIO);
  848.     }
  849.     (res)[size] = '\0';
  850.     av_free(*str);
  851.     *str = res;
  852.  
  853.     return 0;
  854. }
  855.  
  856. /*
  857.  * Read the next element as binary data.
  858.  * 0 is success, < 0 is failure.
  859.  */
  860. static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
  861. {
  862.     av_fast_padded_malloc(&bin->data, &bin->size, length);
  863.     if (!bin->data)
  864.         return AVERROR(ENOMEM);
  865.  
  866.     bin->size = length;
  867.     bin->pos  = avio_tell(pb);
  868.     if (avio_read(pb, bin->data, length) != length) {
  869.         av_freep(&bin->data);
  870.         bin->size = 0;
  871.         return AVERROR(EIO);
  872.     }
  873.  
  874.     return 0;
  875. }
  876.  
  877. /*
  878.  * Read the next element, but only the header. The contents
  879.  * are supposed to be sub-elements which can be read separately.
  880.  * 0 is success, < 0 is failure.
  881.  */
  882. static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
  883. {
  884.     AVIOContext *pb = matroska->ctx->pb;
  885.     MatroskaLevel *level;
  886.  
  887.     if (matroska->num_levels >= EBML_MAX_DEPTH) {
  888.         av_log(matroska->ctx, AV_LOG_ERROR,
  889.                "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
  890.         return AVERROR(ENOSYS);
  891.     }
  892.  
  893.     level         = &matroska->levels[matroska->num_levels++];
  894.     level->start  = avio_tell(pb);
  895.     level->length = length;
  896.  
  897.     return 0;
  898. }
  899.  
  900. /*
  901.  * Read signed/unsigned "EBML" numbers.
  902.  * Return: number of bytes processed, < 0 on error
  903.  */
  904. static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
  905.                                  uint8_t *data, uint32_t size, uint64_t *num)
  906. {
  907.     AVIOContext pb;
  908.     ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
  909.     return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
  910. }
  911.  
  912. /*
  913.  * Same as above, but signed.
  914.  */
  915. static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
  916.                                  uint8_t *data, uint32_t size, int64_t *num)
  917. {
  918.     uint64_t unum;
  919.     int res;
  920.  
  921.     /* read as unsigned number first */
  922.     if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
  923.         return res;
  924.  
  925.     /* make signed (weird way) */
  926.     *num = unum - ((1LL << (7 * res - 1)) - 1);
  927.  
  928.     return res;
  929. }
  930.  
  931. static int ebml_parse_elem(MatroskaDemuxContext *matroska,
  932.                            EbmlSyntax *syntax, void *data);
  933.  
  934. static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
  935.                          uint32_t id, void *data)
  936. {
  937.     int i;
  938.     for (i = 0; syntax[i].id; i++)
  939.         if (id == syntax[i].id)
  940.             break;
  941.     if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
  942.         matroska->num_levels > 0                   &&
  943.         matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
  944.         return 0;  // we reached the end of an unknown size cluster
  945.     if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
  946.         av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32"\n", id);
  947.     }
  948.     return ebml_parse_elem(matroska, &syntax[i], data);
  949. }
  950.  
  951. static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
  952.                       void *data)
  953. {
  954.     if (!matroska->current_id) {
  955.         uint64_t id;
  956.         int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
  957.         if (res < 0) {
  958.             // in live mode, finish parsing if EOF is reached.
  959.             return (matroska->is_live && matroska->ctx->pb->eof_reached &&
  960.                     res == AVERROR_EOF) ? 1 : res;
  961.         }
  962.         matroska->current_id = id | 1 << 7 * res;
  963.     }
  964.     return ebml_parse_id(matroska, syntax, matroska->current_id, data);
  965. }
  966.  
  967. static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
  968.                            void *data)
  969. {
  970.     int i, res = 0;
  971.  
  972.     for (i = 0; syntax[i].id; i++)
  973.         switch (syntax[i].type) {
  974.         case EBML_UINT:
  975.             *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
  976.             break;
  977.         case EBML_FLOAT:
  978.             *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
  979.             break;
  980.         case EBML_STR:
  981.         case EBML_UTF8:
  982.             // the default may be NULL
  983.             if (syntax[i].def.s) {
  984.                 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
  985.                 *dst = av_strdup(syntax[i].def.s);
  986.                 if (!*dst)
  987.                     return AVERROR(ENOMEM);
  988.             }
  989.             break;
  990.         }
  991.  
  992.     while (!res && !ebml_level_end(matroska))
  993.         res = ebml_parse(matroska, syntax, data);
  994.  
  995.     return res;
  996. }
  997.  
  998. static int is_ebml_id_valid(uint32_t id)
  999. {
  1000.     // Due to endian nonsense in Matroska, the highest byte with any bits set
  1001.     // will contain the leading length bit. This bit in turn identifies the
  1002.     // total byte length of the element by its position within the byte.
  1003.     unsigned int bits = av_log2(id);
  1004.     return id && (bits + 7) / 8 ==  (8 - bits % 8);
  1005. }
  1006.  
  1007. /*
  1008.  * Allocate and return the entry for the level1 element with the given ID. If
  1009.  * an entry already exists, return the existing entry.
  1010.  */
  1011. static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
  1012.                                                         uint32_t id)
  1013. {
  1014.     int i;
  1015.     MatroskaLevel1Element *elem;
  1016.  
  1017.     if (!is_ebml_id_valid(id))
  1018.         return NULL;
  1019.  
  1020.     // Some files link to all clusters; useless.
  1021.     if (id == MATROSKA_ID_CLUSTER)
  1022.         return NULL;
  1023.  
  1024.     // There can be multiple seekheads.
  1025.     if (id != MATROSKA_ID_SEEKHEAD) {
  1026.         for (i = 0; i < matroska->num_level1_elems; i++) {
  1027.             if (matroska->level1_elems[i].id == id)
  1028.                 return &matroska->level1_elems[i];
  1029.         }
  1030.     }
  1031.  
  1032.     // Only a completely broken file would have more elements.
  1033.     // It also provides a low-effort way to escape from circular seekheads
  1034.     // (every iteration will add a level1 entry).
  1035.     if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
  1036.         av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
  1037.         return NULL;
  1038.     }
  1039.  
  1040.     elem = &matroska->level1_elems[matroska->num_level1_elems++];
  1041.     *elem = (MatroskaLevel1Element){.id = id};
  1042.  
  1043.     return elem;
  1044. }
  1045.  
  1046. static int ebml_parse_elem(MatroskaDemuxContext *matroska,
  1047.                            EbmlSyntax *syntax, void *data)
  1048. {
  1049.     static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
  1050.         [EBML_UINT]  = 8,
  1051.         [EBML_FLOAT] = 8,
  1052.         // max. 16 MB for strings
  1053.         [EBML_STR]   = 0x1000000,
  1054.         [EBML_UTF8]  = 0x1000000,
  1055.         // max. 256 MB for binary data
  1056.         [EBML_BIN]   = 0x10000000,
  1057.         // no limits for anything else
  1058.     };
  1059.     AVIOContext *pb = matroska->ctx->pb;
  1060.     uint32_t id = syntax->id;
  1061.     uint64_t length;
  1062.     int res;
  1063.     void *newelem;
  1064.     MatroskaLevel1Element *level1_elem;
  1065.  
  1066.     data = (char *) data + syntax->data_offset;
  1067.     if (syntax->list_elem_size) {
  1068.         EbmlList *list = data;
  1069.         newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
  1070.         if (!newelem)
  1071.             return AVERROR(ENOMEM);
  1072.         list->elem = newelem;
  1073.         data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
  1074.         memset(data, 0, syntax->list_elem_size);
  1075.         list->nb_elem++;
  1076.     }
  1077.  
  1078.     if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
  1079.         matroska->current_id = 0;
  1080.         if ((res = ebml_read_length(matroska, pb, &length)) < 0)
  1081.             return res;
  1082.         if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
  1083.             av_log(matroska->ctx, AV_LOG_ERROR,
  1084.                    "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
  1085.                    length, max_lengths[syntax->type], syntax->type);
  1086.             return AVERROR_INVALIDDATA;
  1087.         }
  1088.     }
  1089.  
  1090.     switch (syntax->type) {
  1091.     case EBML_UINT:
  1092.         res = ebml_read_uint(pb, length, data);
  1093.         break;
  1094.     case EBML_SINT:
  1095.         res = ebml_read_sint(pb, length, data);
  1096.         break;
  1097.     case EBML_FLOAT:
  1098.         res = ebml_read_float(pb, length, data);
  1099.         break;
  1100.     case EBML_STR:
  1101.     case EBML_UTF8:
  1102.         res = ebml_read_ascii(pb, length, data);
  1103.         break;
  1104.     case EBML_BIN:
  1105.         res = ebml_read_binary(pb, length, data);
  1106.         break;
  1107.     case EBML_LEVEL1:
  1108.     case EBML_NEST:
  1109.         if ((res = ebml_read_master(matroska, length)) < 0)
  1110.             return res;
  1111.         if (id == MATROSKA_ID_SEGMENT)
  1112.             matroska->segment_start = avio_tell(matroska->ctx->pb);
  1113.         if (id == MATROSKA_ID_CUES)
  1114.             matroska->cues_parsing_deferred = 0;
  1115.         if (syntax->type == EBML_LEVEL1 &&
  1116.             (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
  1117.             if (level1_elem->parsed)
  1118.                 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
  1119.             level1_elem->parsed = 1;
  1120.         }
  1121.         return ebml_parse_nest(matroska, syntax->def.n, data);
  1122.     case EBML_PASS:
  1123.         return ebml_parse_id(matroska, syntax->def.n, id, data);
  1124.     case EBML_STOP:
  1125.         return 1;
  1126.     default:
  1127.         if (ffio_limit(pb, length) != length)
  1128.             return AVERROR(EIO);
  1129.         return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
  1130.     }
  1131.     if (res == AVERROR_INVALIDDATA)
  1132.         av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
  1133.     else if (res == AVERROR(EIO))
  1134.         av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
  1135.     return res;
  1136. }
  1137.  
  1138. static void ebml_free(EbmlSyntax *syntax, void *data)
  1139. {
  1140.     int i, j;
  1141.     for (i = 0; syntax[i].id; i++) {
  1142.         void *data_off = (char *) data + syntax[i].data_offset;
  1143.         switch (syntax[i].type) {
  1144.         case EBML_STR:
  1145.         case EBML_UTF8:
  1146.             av_freep(data_off);
  1147.             break;
  1148.         case EBML_BIN:
  1149.             av_freep(&((EbmlBin *) data_off)->data);
  1150.             break;
  1151.         case EBML_LEVEL1:
  1152.         case EBML_NEST:
  1153.             if (syntax[i].list_elem_size) {
  1154.                 EbmlList *list = data_off;
  1155.                 char *ptr = list->elem;
  1156.                 for (j = 0; j < list->nb_elem;
  1157.                      j++, ptr += syntax[i].list_elem_size)
  1158.                     ebml_free(syntax[i].def.n, ptr);
  1159.                 av_freep(&list->elem);
  1160.             } else
  1161.                 ebml_free(syntax[i].def.n, data_off);
  1162.         default:
  1163.             break;
  1164.         }
  1165.     }
  1166. }
  1167.  
  1168. /*
  1169.  * Autodetecting...
  1170.  */
  1171. static int matroska_probe(AVProbeData *p)
  1172. {
  1173.     uint64_t total = 0;
  1174.     int len_mask = 0x80, size = 1, n = 1, i;
  1175.  
  1176.     /* EBML header? */
  1177.     if (AV_RB32(p->buf) != EBML_ID_HEADER)
  1178.         return 0;
  1179.  
  1180.     /* length of header */
  1181.     total = p->buf[4];
  1182.     while (size <= 8 && !(total & len_mask)) {
  1183.         size++;
  1184.         len_mask >>= 1;
  1185.     }
  1186.     if (size > 8)
  1187.         return 0;
  1188.     total &= (len_mask - 1);
  1189.     while (n < size)
  1190.         total = (total << 8) | p->buf[4 + n++];
  1191.  
  1192.     /* Does the probe data contain the whole header? */
  1193.     if (p->buf_size < 4 + size + total)
  1194.         return 0;
  1195.  
  1196.     /* The header should contain a known document type. For now,
  1197.      * we don't parse the whole header but simply check for the
  1198.      * availability of that array of characters inside the header.
  1199.      * Not fully fool-proof, but good enough. */
  1200.     for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
  1201.         size_t probelen = strlen(matroska_doctypes[i]);
  1202.         if (total < probelen)
  1203.             continue;
  1204.         for (n = 4 + size; n <= 4 + size + total - probelen; n++)
  1205.             if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
  1206.                 return AVPROBE_SCORE_MAX;
  1207.     }
  1208.  
  1209.     // probably valid EBML header but no recognized doctype
  1210.     return AVPROBE_SCORE_EXTENSION;
  1211. }
  1212.  
  1213. static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
  1214.                                                  int num)
  1215. {
  1216.     MatroskaTrack *tracks = matroska->tracks.elem;
  1217.     int i;
  1218.  
  1219.     for (i = 0; i < matroska->tracks.nb_elem; i++)
  1220.         if (tracks[i].num == num)
  1221.             return &tracks[i];
  1222.  
  1223.     av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
  1224.     return NULL;
  1225. }
  1226.  
  1227. static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
  1228.                                   MatroskaTrack *track)
  1229. {
  1230.     MatroskaTrackEncoding *encodings = track->encodings.elem;
  1231.     uint8_t *data = *buf;
  1232.     int isize = *buf_size;
  1233.     uint8_t *pkt_data = NULL;
  1234.     uint8_t av_unused *newpktdata;
  1235.     int pkt_size = isize;
  1236.     int result = 0;
  1237.     int olen;
  1238.  
  1239.     if (pkt_size >= 10000000U)
  1240.         return AVERROR_INVALIDDATA;
  1241.  
  1242.     switch (encodings[0].compression.algo) {
  1243.     case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
  1244.     {
  1245.         int header_size = encodings[0].compression.settings.size;
  1246.         uint8_t *header = encodings[0].compression.settings.data;
  1247.  
  1248.         if (header_size && !header) {
  1249.             av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
  1250.             return -1;
  1251.         }
  1252.  
  1253.         if (!header_size)
  1254.             return 0;
  1255.  
  1256.         pkt_size = isize + header_size;
  1257.         pkt_data = av_malloc(pkt_size);
  1258.         if (!pkt_data)
  1259.             return AVERROR(ENOMEM);
  1260.  
  1261.         memcpy(pkt_data, header, header_size);
  1262.         memcpy(pkt_data + header_size, data, isize);
  1263.         break;
  1264.     }
  1265. #if CONFIG_LZO
  1266.     case MATROSKA_TRACK_ENCODING_COMP_LZO:
  1267.         do {
  1268.             olen       = pkt_size *= 3;
  1269.             newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
  1270.             if (!newpktdata) {
  1271.                 result = AVERROR(ENOMEM);
  1272.                 goto failed;
  1273.             }
  1274.             pkt_data = newpktdata;
  1275.             result   = av_lzo1x_decode(pkt_data, &olen, data, &isize);
  1276.         } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
  1277.         if (result) {
  1278.             result = AVERROR_INVALIDDATA;
  1279.             goto failed;
  1280.         }
  1281.         pkt_size -= olen;
  1282.         break;
  1283. #endif
  1284. #if CONFIG_ZLIB
  1285.     case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
  1286.     {
  1287.         z_stream zstream = { 0 };
  1288.         if (inflateInit(&zstream) != Z_OK)
  1289.             return -1;
  1290.         zstream.next_in  = data;
  1291.         zstream.avail_in = isize;
  1292.         do {
  1293.             pkt_size  *= 3;
  1294.             newpktdata = av_realloc(pkt_data, pkt_size);
  1295.             if (!newpktdata) {
  1296.                 inflateEnd(&zstream);
  1297.                 result = AVERROR(ENOMEM);
  1298.                 goto failed;
  1299.             }
  1300.             pkt_data          = newpktdata;
  1301.             zstream.avail_out = pkt_size - zstream.total_out;
  1302.             zstream.next_out  = pkt_data + zstream.total_out;
  1303.             result = inflate(&zstream, Z_NO_FLUSH);
  1304.         } while (result == Z_OK && pkt_size < 10000000);
  1305.         pkt_size = zstream.total_out;
  1306.         inflateEnd(&zstream);
  1307.         if (result != Z_STREAM_END) {
  1308.             if (result == Z_MEM_ERROR)
  1309.                 result = AVERROR(ENOMEM);
  1310.             else
  1311.                 result = AVERROR_INVALIDDATA;
  1312.             goto failed;
  1313.         }
  1314.         break;
  1315.     }
  1316. #endif
  1317. #if CONFIG_BZLIB
  1318.     case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
  1319.     {
  1320.         bz_stream bzstream = { 0 };
  1321.         if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
  1322.             return -1;
  1323.         bzstream.next_in  = data;
  1324.         bzstream.avail_in = isize;
  1325.         do {
  1326.             pkt_size  *= 3;
  1327.             newpktdata = av_realloc(pkt_data, pkt_size);
  1328.             if (!newpktdata) {
  1329.                 BZ2_bzDecompressEnd(&bzstream);
  1330.                 result = AVERROR(ENOMEM);
  1331.                 goto failed;
  1332.             }
  1333.             pkt_data           = newpktdata;
  1334.             bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
  1335.             bzstream.next_out  = pkt_data + bzstream.total_out_lo32;
  1336.             result = BZ2_bzDecompress(&bzstream);
  1337.         } while (result == BZ_OK && pkt_size < 10000000);
  1338.         pkt_size = bzstream.total_out_lo32;
  1339.         BZ2_bzDecompressEnd(&bzstream);
  1340.         if (result != BZ_STREAM_END) {
  1341.             if (result == BZ_MEM_ERROR)
  1342.                 result = AVERROR(ENOMEM);
  1343.             else
  1344.                 result = AVERROR_INVALIDDATA;
  1345.             goto failed;
  1346.         }
  1347.         break;
  1348.     }
  1349. #endif
  1350.     default:
  1351.         return AVERROR_INVALIDDATA;
  1352.     }
  1353.  
  1354.     *buf      = pkt_data;
  1355.     *buf_size = pkt_size;
  1356.     return 0;
  1357.  
  1358. failed:
  1359.     av_free(pkt_data);
  1360.     return result;
  1361. }
  1362.  
  1363. static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
  1364.                                  AVDictionary **metadata, char *prefix)
  1365. {
  1366.     MatroskaTag *tags = list->elem;
  1367.     char key[1024];
  1368.     int i;
  1369.  
  1370.     for (i = 0; i < list->nb_elem; i++) {
  1371.         const char *lang = tags[i].lang &&
  1372.                            strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
  1373.  
  1374.         if (!tags[i].name) {
  1375.             av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
  1376.             continue;
  1377.         }
  1378.         if (prefix)
  1379.             snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
  1380.         else
  1381.             av_strlcpy(key, tags[i].name, sizeof(key));
  1382.         if (tags[i].def || !lang) {
  1383.             av_dict_set(metadata, key, tags[i].string, 0);
  1384.             if (tags[i].sub.nb_elem)
  1385.                 matroska_convert_tag(s, &tags[i].sub, metadata, key);
  1386.         }
  1387.         if (lang) {
  1388.             av_strlcat(key, "-", sizeof(key));
  1389.             av_strlcat(key, lang, sizeof(key));
  1390.             av_dict_set(metadata, key, tags[i].string, 0);
  1391.             if (tags[i].sub.nb_elem)
  1392.                 matroska_convert_tag(s, &tags[i].sub, metadata, key);
  1393.         }
  1394.     }
  1395.     ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
  1396. }
  1397.  
  1398. static void matroska_convert_tags(AVFormatContext *s)
  1399. {
  1400.     MatroskaDemuxContext *matroska = s->priv_data;
  1401.     MatroskaTags *tags = matroska->tags.elem;
  1402.     int i, j;
  1403.  
  1404.     for (i = 0; i < matroska->tags.nb_elem; i++) {
  1405.         if (tags[i].target.attachuid) {
  1406.             MatroskaAttachment *attachment = matroska->attachments.elem;
  1407.             for (j = 0; j < matroska->attachments.nb_elem; j++)
  1408.                 if (attachment[j].uid == tags[i].target.attachuid &&
  1409.                     attachment[j].stream)
  1410.                     matroska_convert_tag(s, &tags[i].tag,
  1411.                                          &attachment[j].stream->metadata, NULL);
  1412.         } else if (tags[i].target.chapteruid) {
  1413.             MatroskaChapter *chapter = matroska->chapters.elem;
  1414.             for (j = 0; j < matroska->chapters.nb_elem; j++)
  1415.                 if (chapter[j].uid == tags[i].target.chapteruid &&
  1416.                     chapter[j].chapter)
  1417.                     matroska_convert_tag(s, &tags[i].tag,
  1418.                                          &chapter[j].chapter->metadata, NULL);
  1419.         } else if (tags[i].target.trackuid) {
  1420.             MatroskaTrack *track = matroska->tracks.elem;
  1421.             for (j = 0; j < matroska->tracks.nb_elem; j++)
  1422.                 if (track[j].uid == tags[i].target.trackuid && track[j].stream)
  1423.                     matroska_convert_tag(s, &tags[i].tag,
  1424.                                          &track[j].stream->metadata, NULL);
  1425.         } else {
  1426.             matroska_convert_tag(s, &tags[i].tag, &s->metadata,
  1427.                                  tags[i].target.type);
  1428.         }
  1429.     }
  1430. }
  1431.  
  1432. static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
  1433.                                          uint64_t pos)
  1434. {
  1435.     uint32_t level_up       = matroska->level_up;
  1436.     uint32_t saved_id       = matroska->current_id;
  1437.     int64_t before_pos = avio_tell(matroska->ctx->pb);
  1438.     MatroskaLevel level;
  1439.     int64_t offset;
  1440.     int ret = 0;
  1441.  
  1442.     /* seek */
  1443.     offset = pos + matroska->segment_start;
  1444.     if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
  1445.         /* We don't want to lose our seekhead level, so we add
  1446.          * a dummy. This is a crude hack. */
  1447.         if (matroska->num_levels == EBML_MAX_DEPTH) {
  1448.             av_log(matroska->ctx, AV_LOG_INFO,
  1449.                    "Max EBML element depth (%d) reached, "
  1450.                    "cannot parse further.\n", EBML_MAX_DEPTH);
  1451.             ret = AVERROR_INVALIDDATA;
  1452.         } else {
  1453.             level.start  = 0;
  1454.             level.length = (uint64_t) -1;
  1455.             matroska->levels[matroska->num_levels] = level;
  1456.             matroska->num_levels++;
  1457.             matroska->current_id                   = 0;
  1458.  
  1459.             ret = ebml_parse(matroska, matroska_segment, matroska);
  1460.  
  1461.             /* remove dummy level */
  1462.             while (matroska->num_levels) {
  1463.                 uint64_t length = matroska->levels[--matroska->num_levels].length;
  1464.                 if (length == (uint64_t) -1)
  1465.                     break;
  1466.             }
  1467.         }
  1468.     }
  1469.     /* seek back */
  1470.     avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
  1471.     matroska->level_up   = level_up;
  1472.     matroska->current_id = saved_id;
  1473.  
  1474.     return ret;
  1475. }
  1476.  
  1477. static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
  1478. {
  1479.     EbmlList *seekhead_list = &matroska->seekhead;
  1480.     int i;
  1481.  
  1482.     // we should not do any seeking in the streaming case
  1483.     if (!matroska->ctx->pb->seekable)
  1484.         return;
  1485.  
  1486.     for (i = 0; i < seekhead_list->nb_elem; i++) {
  1487.         MatroskaSeekhead *seekheads = seekhead_list->elem;
  1488.         uint32_t id  = seekheads[i].id;
  1489.         uint64_t pos = seekheads[i].pos;
  1490.  
  1491.         MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
  1492.         if (!elem || elem->parsed)
  1493.             continue;
  1494.  
  1495.         elem->pos = pos;
  1496.  
  1497.         // defer cues parsing until we actually need cue data.
  1498.         if (id == MATROSKA_ID_CUES)
  1499.             continue;
  1500.  
  1501.         if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
  1502.             // mark index as broken
  1503.             matroska->cues_parsing_deferred = -1;
  1504.             break;
  1505.         }
  1506.  
  1507.         elem->parsed = 1;
  1508.     }
  1509. }
  1510.  
  1511. static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
  1512. {
  1513.     EbmlList *index_list;
  1514.     MatroskaIndex *index;
  1515.     uint64_t index_scale = 1;
  1516.     int i, j;
  1517.  
  1518.     if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
  1519.         return;
  1520.  
  1521.     index_list = &matroska->index;
  1522.     index      = index_list->elem;
  1523.     if (index_list->nb_elem &&
  1524.         index[0].time > 1E14 / matroska->time_scale) {
  1525.         av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
  1526.         index_scale = matroska->time_scale;
  1527.     }
  1528.     for (i = 0; i < index_list->nb_elem; i++) {
  1529.         EbmlList *pos_list    = &index[i].pos;
  1530.         MatroskaIndexPos *pos = pos_list->elem;
  1531.         for (j = 0; j < pos_list->nb_elem; j++) {
  1532.             MatroskaTrack *track = matroska_find_track_by_num(matroska,
  1533.                                                               pos[j].track);
  1534.             if (track && track->stream)
  1535.                 av_add_index_entry(track->stream,
  1536.                                    pos[j].pos + matroska->segment_start,
  1537.                                    index[i].time / index_scale, 0, 0,
  1538.                                    AVINDEX_KEYFRAME);
  1539.         }
  1540.     }
  1541. }
  1542.  
  1543. static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
  1544.     int i;
  1545.  
  1546.     if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
  1547.         return;
  1548.  
  1549.     for (i = 0; i < matroska->num_level1_elems; i++) {
  1550.         MatroskaLevel1Element *elem = &matroska->level1_elems[i];
  1551.         if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
  1552.             if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
  1553.                 matroska->cues_parsing_deferred = -1;
  1554.             elem->parsed = 1;
  1555.             break;
  1556.         }
  1557.     }
  1558.  
  1559.     matroska_add_index_entries(matroska);
  1560. }
  1561.  
  1562. static int matroska_aac_profile(char *codec_id)
  1563. {
  1564.     static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
  1565.     int profile;
  1566.  
  1567.     for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
  1568.         if (strstr(codec_id, aac_profiles[profile]))
  1569.             break;
  1570.     return profile + 1;
  1571. }
  1572.  
  1573. static int matroska_aac_sri(int samplerate)
  1574. {
  1575.     int sri;
  1576.  
  1577.     for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
  1578.         if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
  1579.             break;
  1580.     return sri;
  1581. }
  1582.  
  1583. static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
  1584. {
  1585.     char buffer[32];
  1586.     /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
  1587.     time_t creation_time = date_utc / 1000000000 + 978307200;
  1588.     struct tm tmpbuf, *ptm = gmtime_r(&creation_time, &tmpbuf);
  1589.     if (!ptm) return;
  1590.     if (strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm))
  1591.         av_dict_set(metadata, "creation_time", buffer, 0);
  1592. }
  1593.  
  1594. static int matroska_parse_flac(AVFormatContext *s,
  1595.                                MatroskaTrack *track,
  1596.                                int *offset)
  1597. {
  1598.     AVStream *st = track->stream;
  1599.     uint8_t *p = track->codec_priv.data;
  1600.     int size   = track->codec_priv.size;
  1601.  
  1602.     if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
  1603.         av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
  1604.         track->codec_priv.size = 0;
  1605.         return 0;
  1606.     }
  1607.     *offset = 8;
  1608.     track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
  1609.  
  1610.     p    += track->codec_priv.size;
  1611.     size -= track->codec_priv.size;
  1612.  
  1613.     /* parse the remaining metadata blocks if present */
  1614.     while (size >= 4) {
  1615.         int block_last, block_type, block_size;
  1616.  
  1617.         flac_parse_block_header(p, &block_last, &block_type, &block_size);
  1618.  
  1619.         p    += 4;
  1620.         size -= 4;
  1621.         if (block_size > size)
  1622.             return 0;
  1623.  
  1624.         /* check for the channel mask */
  1625.         if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
  1626.             AVDictionary *dict = NULL;
  1627.             AVDictionaryEntry *chmask;
  1628.  
  1629.             ff_vorbis_comment(s, &dict, p, block_size, 0);
  1630.             chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
  1631.             if (chmask) {
  1632.                 uint64_t mask = strtol(chmask->value, NULL, 0);
  1633.                 if (!mask || mask & ~0x3ffffULL) {
  1634.                     av_log(s, AV_LOG_WARNING,
  1635.                            "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
  1636.                 } else
  1637.                     st->codec->channel_layout = mask;
  1638.             }
  1639.             av_dict_free(&dict);
  1640.         }
  1641.  
  1642.         p    += block_size;
  1643.         size -= block_size;
  1644.     }
  1645.  
  1646.     return 0;
  1647. }
  1648.  
  1649. static int matroska_parse_tracks(AVFormatContext *s)
  1650. {
  1651.     MatroskaDemuxContext *matroska = s->priv_data;
  1652.     MatroskaTrack *tracks = matroska->tracks.elem;
  1653.     AVStream *st;
  1654.     int i, j, ret;
  1655.     int k;
  1656.  
  1657.     for (i = 0; i < matroska->tracks.nb_elem; i++) {
  1658.         MatroskaTrack *track = &tracks[i];
  1659.         enum AVCodecID codec_id = AV_CODEC_ID_NONE;
  1660.         EbmlList *encodings_list = &track->encodings;
  1661.         MatroskaTrackEncoding *encodings = encodings_list->elem;
  1662.         uint8_t *extradata = NULL;
  1663.         int extradata_size = 0;
  1664.         int extradata_offset = 0;
  1665.         uint32_t fourcc = 0;
  1666.         AVIOContext b;
  1667.         char* key_id_base64 = NULL;
  1668.         int bit_depth = -1;
  1669.  
  1670.         /* Apply some sanity checks. */
  1671.         if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
  1672.             track->type != MATROSKA_TRACK_TYPE_AUDIO &&
  1673.             track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
  1674.             track->type != MATROSKA_TRACK_TYPE_METADATA) {
  1675.             av_log(matroska->ctx, AV_LOG_INFO,
  1676.                    "Unknown or unsupported track type %"PRIu64"\n",
  1677.                    track->type);
  1678.             continue;
  1679.         }
  1680.         if (!track->codec_id)
  1681.             continue;
  1682.  
  1683.         if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
  1684.             isnan(track->audio.samplerate)) {
  1685.             av_log(matroska->ctx, AV_LOG_WARNING,
  1686.                    "Invalid sample rate %f, defaulting to 8000 instead.\n",
  1687.                    track->audio.samplerate);
  1688.             track->audio.samplerate = 8000;
  1689.         }
  1690.  
  1691.         if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
  1692.             if (!track->default_duration && track->video.frame_rate > 0)
  1693.                 track->default_duration = 1000000000 / track->video.frame_rate;
  1694.             if (track->video.display_width == -1)
  1695.                 track->video.display_width = track->video.pixel_width;
  1696.             if (track->video.display_height == -1)
  1697.                 track->video.display_height = track->video.pixel_height;
  1698.             if (track->video.color_space.size == 4)
  1699.                 fourcc = AV_RL32(track->video.color_space.data);
  1700.         } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
  1701.             if (!track->audio.out_samplerate)
  1702.                 track->audio.out_samplerate = track->audio.samplerate;
  1703.         }
  1704.         if (encodings_list->nb_elem > 1) {
  1705.             av_log(matroska->ctx, AV_LOG_ERROR,
  1706.                    "Multiple combined encodings not supported");
  1707.         } else if (encodings_list->nb_elem == 1) {
  1708.             if (encodings[0].type) {
  1709.                 if (encodings[0].encryption.key_id.size > 0) {
  1710.                     /* Save the encryption key id to be stored later as a
  1711.                        metadata tag. */
  1712.                     const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
  1713.                     key_id_base64 = av_malloc(b64_size);
  1714.                     if (key_id_base64 == NULL)
  1715.                         return AVERROR(ENOMEM);
  1716.  
  1717.                     av_base64_encode(key_id_base64, b64_size,
  1718.                                      encodings[0].encryption.key_id.data,
  1719.                                      encodings[0].encryption.key_id.size);
  1720.                 } else {
  1721.                     encodings[0].scope = 0;
  1722.                     av_log(matroska->ctx, AV_LOG_ERROR,
  1723.                            "Unsupported encoding type");
  1724.                 }
  1725.             } else if (
  1726. #if CONFIG_ZLIB
  1727.                  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB  &&
  1728. #endif
  1729. #if CONFIG_BZLIB
  1730.                  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
  1731. #endif
  1732. #if CONFIG_LZO
  1733.                  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO   &&
  1734. #endif
  1735.                  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
  1736.                 encodings[0].scope = 0;
  1737.                 av_log(matroska->ctx, AV_LOG_ERROR,
  1738.                        "Unsupported encoding type");
  1739.             } else if (track->codec_priv.size && encodings[0].scope & 2) {
  1740.                 uint8_t *codec_priv = track->codec_priv.data;
  1741.                 int ret = matroska_decode_buffer(&track->codec_priv.data,
  1742.                                                  &track->codec_priv.size,
  1743.                                                  track);
  1744.                 if (ret < 0) {
  1745.                     track->codec_priv.data = NULL;
  1746.                     track->codec_priv.size = 0;
  1747.                     av_log(matroska->ctx, AV_LOG_ERROR,
  1748.                            "Failed to decode codec private data\n");
  1749.                 }
  1750.  
  1751.                 if (codec_priv != track->codec_priv.data)
  1752.                     av_free(codec_priv);
  1753.             }
  1754.         }
  1755.  
  1756.         for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
  1757.             if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
  1758.                          strlen(ff_mkv_codec_tags[j].str))) {
  1759.                 codec_id = ff_mkv_codec_tags[j].id;
  1760.                 break;
  1761.             }
  1762.         }
  1763.  
  1764.         st = track->stream = avformat_new_stream(s, NULL);
  1765.         if (!st) {
  1766.             av_free(key_id_base64);
  1767.             return AVERROR(ENOMEM);
  1768.         }
  1769.  
  1770.         if (key_id_base64) {
  1771.             /* export encryption key id as base64 metadata tag */
  1772.             av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
  1773.             av_freep(&key_id_base64);
  1774.         }
  1775.  
  1776.         if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
  1777.              track->codec_priv.size >= 40               &&
  1778.             track->codec_priv.data) {
  1779.             track->ms_compat    = 1;
  1780.             bit_depth           = AV_RL16(track->codec_priv.data + 14);
  1781.             fourcc              = AV_RL32(track->codec_priv.data + 16);
  1782.             codec_id            = ff_codec_get_id(ff_codec_bmp_tags,
  1783.                                                   fourcc);
  1784.             if (!codec_id)
  1785.                 codec_id        = ff_codec_get_id(ff_codec_movvideo_tags,
  1786.                                                   fourcc);
  1787.             extradata_offset    = 40;
  1788.         } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
  1789.                    track->codec_priv.size >= 14         &&
  1790.                    track->codec_priv.data) {
  1791.             int ret;
  1792.             ffio_init_context(&b, track->codec_priv.data,
  1793.                               track->codec_priv.size,
  1794.                               0, NULL, NULL, NULL, NULL);
  1795.             ret = ff_get_wav_header(s, &b, st->codec, track->codec_priv.size, 0);
  1796.             if (ret < 0)
  1797.                 return ret;
  1798.             codec_id         = st->codec->codec_id;
  1799.             extradata_offset = FFMIN(track->codec_priv.size, 18);
  1800.         } else if (!strcmp(track->codec_id, "A_QUICKTIME")
  1801.                    && (track->codec_priv.size >= 86)
  1802.                    && (track->codec_priv.data)) {
  1803.             fourcc = AV_RL32(track->codec_priv.data + 4);
  1804.             codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
  1805.             if (ff_codec_get_id(ff_codec_movaudio_tags, AV_RL32(track->codec_priv.data))) {
  1806.                 fourcc = AV_RL32(track->codec_priv.data);
  1807.                 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
  1808.             }
  1809.         } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
  1810.                    (track->codec_priv.size >= 21)          &&
  1811.                    (track->codec_priv.data)) {
  1812.             fourcc   = AV_RL32(track->codec_priv.data + 4);
  1813.             codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
  1814.             if (ff_codec_get_id(ff_codec_movvideo_tags, AV_RL32(track->codec_priv.data))) {
  1815.                 fourcc   = AV_RL32(track->codec_priv.data);
  1816.                 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
  1817.             }
  1818.             if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI "))
  1819.                 codec_id = AV_CODEC_ID_SVQ3;
  1820.             if (codec_id == AV_CODEC_ID_NONE) {
  1821.                 char buf[32];
  1822.                 av_get_codec_tag_string(buf, sizeof(buf), fourcc);
  1823.                 av_log(matroska->ctx, AV_LOG_ERROR,
  1824.                        "mov FourCC not found %s.\n", buf);
  1825.             }
  1826.         } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
  1827.             switch (track->audio.bitdepth) {
  1828.             case  8:
  1829.                 codec_id = AV_CODEC_ID_PCM_U8;
  1830.                 break;
  1831.             case 24:
  1832.                 codec_id = AV_CODEC_ID_PCM_S24BE;
  1833.                 break;
  1834.             case 32:
  1835.                 codec_id = AV_CODEC_ID_PCM_S32BE;
  1836.                 break;
  1837.             }
  1838.         } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
  1839.             switch (track->audio.bitdepth) {
  1840.             case  8:
  1841.                 codec_id = AV_CODEC_ID_PCM_U8;
  1842.                 break;
  1843.             case 24:
  1844.                 codec_id = AV_CODEC_ID_PCM_S24LE;
  1845.                 break;
  1846.             case 32:
  1847.                 codec_id = AV_CODEC_ID_PCM_S32LE;
  1848.                 break;
  1849.             }
  1850.         } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
  1851.                    track->audio.bitdepth == 64) {
  1852.             codec_id = AV_CODEC_ID_PCM_F64LE;
  1853.         } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
  1854.             int profile = matroska_aac_profile(track->codec_id);
  1855.             int sri     = matroska_aac_sri(track->audio.samplerate);
  1856.             extradata   = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
  1857.             if (!extradata)
  1858.                 return AVERROR(ENOMEM);
  1859.             extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
  1860.             extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
  1861.             if (strstr(track->codec_id, "SBR")) {
  1862.                 sri            = matroska_aac_sri(track->audio.out_samplerate);
  1863.                 extradata[2]   = 0x56;
  1864.                 extradata[3]   = 0xE5;
  1865.                 extradata[4]   = 0x80 | (sri << 3);
  1866.                 extradata_size = 5;
  1867.             } else
  1868.                 extradata_size = 2;
  1869.         } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
  1870.             /* Only ALAC's magic cookie is stored in Matroska's track headers.
  1871.              * Create the "atom size", "tag", and "tag version" fields the
  1872.              * decoder expects manually. */
  1873.             extradata_size = 12 + track->codec_priv.size;
  1874.             extradata      = av_mallocz(extradata_size +
  1875.                                         AV_INPUT_BUFFER_PADDING_SIZE);
  1876.             if (!extradata)
  1877.                 return AVERROR(ENOMEM);
  1878.             AV_WB32(extradata, extradata_size);
  1879.             memcpy(&extradata[4], "alac", 4);
  1880.             AV_WB32(&extradata[8], 0);
  1881.             memcpy(&extradata[12], track->codec_priv.data,
  1882.                    track->codec_priv.size);
  1883.         } else if (codec_id == AV_CODEC_ID_TTA) {
  1884.             extradata_size = 30;
  1885.             extradata      = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
  1886.             if (!extradata)
  1887.                 return AVERROR(ENOMEM);
  1888.             ffio_init_context(&b, extradata, extradata_size, 1,
  1889.                               NULL, NULL, NULL, NULL);
  1890.             avio_write(&b, "TTA1", 4);
  1891.             avio_wl16(&b, 1);
  1892.             if (track->audio.channels > UINT16_MAX ||
  1893.                 track->audio.bitdepth > UINT16_MAX) {
  1894.                 av_log(matroska->ctx, AV_LOG_WARNING,
  1895.                        "Too large audio channel number %"PRIu64
  1896.                        " or bitdepth %"PRIu64". Skipping track.\n",
  1897.                        track->audio.channels, track->audio.bitdepth);
  1898.                 av_freep(&extradata);
  1899.                 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
  1900.                     return AVERROR_INVALIDDATA;
  1901.                 else
  1902.                     continue;
  1903.             }
  1904.             avio_wl16(&b, track->audio.channels);
  1905.             avio_wl16(&b, track->audio.bitdepth);
  1906.             if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
  1907.                 return AVERROR_INVALIDDATA;
  1908.             avio_wl32(&b, track->audio.out_samplerate);
  1909.             avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
  1910.                                      track->audio.out_samplerate,
  1911.                                      AV_TIME_BASE * 1000));
  1912.         } else if (codec_id == AV_CODEC_ID_RV10 ||
  1913.                    codec_id == AV_CODEC_ID_RV20 ||
  1914.                    codec_id == AV_CODEC_ID_RV30 ||
  1915.                    codec_id == AV_CODEC_ID_RV40) {
  1916.             extradata_offset = 26;
  1917.         } else if (codec_id == AV_CODEC_ID_RA_144) {
  1918.             track->audio.out_samplerate = 8000;
  1919.             track->audio.channels       = 1;
  1920.         } else if ((codec_id == AV_CODEC_ID_RA_288 ||
  1921.                     codec_id == AV_CODEC_ID_COOK   ||
  1922.                     codec_id == AV_CODEC_ID_ATRAC3 ||
  1923.                     codec_id == AV_CODEC_ID_SIPR)
  1924.                       && track->codec_priv.data) {
  1925.             int flavor;
  1926.  
  1927.             ffio_init_context(&b, track->codec_priv.data,
  1928.                               track->codec_priv.size,
  1929.                               0, NULL, NULL, NULL, NULL);
  1930.             avio_skip(&b, 22);
  1931.             flavor                       = avio_rb16(&b);
  1932.             track->audio.coded_framesize = avio_rb32(&b);
  1933.             avio_skip(&b, 12);
  1934.             track->audio.sub_packet_h    = avio_rb16(&b);
  1935.             track->audio.frame_size      = avio_rb16(&b);
  1936.             track->audio.sub_packet_size = avio_rb16(&b);
  1937.             if (flavor                        < 0 ||
  1938.                 track->audio.coded_framesize <= 0 ||
  1939.                 track->audio.sub_packet_h    <= 0 ||
  1940.                 track->audio.frame_size      <= 0 ||
  1941.                 track->audio.sub_packet_size <= 0)
  1942.                 return AVERROR_INVALIDDATA;
  1943.             track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
  1944.                                                track->audio.frame_size);
  1945.             if (!track->audio.buf)
  1946.                 return AVERROR(ENOMEM);
  1947.             if (codec_id == AV_CODEC_ID_RA_288) {
  1948.                 st->codec->block_align = track->audio.coded_framesize;
  1949.                 track->codec_priv.size = 0;
  1950.             } else {
  1951.                 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
  1952.                     static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
  1953.                     track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
  1954.                     st->codec->bit_rate          = sipr_bit_rate[flavor];
  1955.                 }
  1956.                 st->codec->block_align = track->audio.sub_packet_size;
  1957.                 extradata_offset       = 78;
  1958.             }
  1959.         } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
  1960.             ret = matroska_parse_flac(s, track, &extradata_offset);
  1961.             if (ret < 0)
  1962.                 return ret;
  1963.         } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
  1964.             fourcc = AV_RL32(track->codec_priv.data);
  1965.         }
  1966.         track->codec_priv.size -= extradata_offset;
  1967.  
  1968.         if (codec_id == AV_CODEC_ID_NONE)
  1969.             av_log(matroska->ctx, AV_LOG_INFO,
  1970.                    "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
  1971.  
  1972.         if (track->time_scale < 0.01)
  1973.             track->time_scale = 1.0;
  1974.         avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
  1975.                             1000 * 1000 * 1000);    /* 64 bit pts in ns */
  1976.  
  1977.         /* convert the delay from ns to the track timebase */
  1978.         track->codec_delay = av_rescale_q(track->codec_delay,
  1979.                                           (AVRational){ 1, 1000000000 },
  1980.                                           st->time_base);
  1981.  
  1982.         st->codec->codec_id = codec_id;
  1983.  
  1984.         if (strcmp(track->language, "und"))
  1985.             av_dict_set(&st->metadata, "language", track->language, 0);
  1986.         av_dict_set(&st->metadata, "title", track->name, 0);
  1987.  
  1988.         if (track->flag_default)
  1989.             st->disposition |= AV_DISPOSITION_DEFAULT;
  1990.         if (track->flag_forced)
  1991.             st->disposition |= AV_DISPOSITION_FORCED;
  1992.  
  1993.         if (!st->codec->extradata) {
  1994.             if (extradata) {
  1995.                 st->codec->extradata      = extradata;
  1996.                 st->codec->extradata_size = extradata_size;
  1997.             } else if (track->codec_priv.data && track->codec_priv.size > 0) {
  1998.                 if (ff_alloc_extradata(st->codec, track->codec_priv.size))
  1999.                     return AVERROR(ENOMEM);
  2000.                 memcpy(st->codec->extradata,
  2001.                        track->codec_priv.data + extradata_offset,
  2002.                        track->codec_priv.size);
  2003.             }
  2004.         }
  2005.  
  2006.         if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
  2007.             MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
  2008.  
  2009.             st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
  2010.             st->codec->codec_tag  = fourcc;
  2011.             if (bit_depth >= 0)
  2012.                 st->codec->bits_per_coded_sample = bit_depth;
  2013.             st->codec->width      = track->video.pixel_width;
  2014.             st->codec->height     = track->video.pixel_height;
  2015.             av_reduce(&st->sample_aspect_ratio.num,
  2016.                       &st->sample_aspect_ratio.den,
  2017.                       st->codec->height * track->video.display_width,
  2018.                       st->codec->width  * track->video.display_height,
  2019.                       255);
  2020.             if (st->codec->codec_id != AV_CODEC_ID_HEVC)
  2021.                 st->need_parsing = AVSTREAM_PARSE_HEADERS;
  2022.  
  2023.             if (track->default_duration) {
  2024.                 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
  2025.                           1000000000, track->default_duration, 30000);
  2026. #if FF_API_R_FRAME_RATE
  2027.                 if (   st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
  2028.                     && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
  2029.                     st->r_frame_rate = st->avg_frame_rate;
  2030. #endif
  2031.             }
  2032.  
  2033.             /* export stereo mode flag as metadata tag */
  2034.             if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
  2035.                 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
  2036.  
  2037.             /* export alpha mode flag as metadata tag  */
  2038.             if (track->video.alpha_mode)
  2039.                 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
  2040.  
  2041.             /* if we have virtual track, mark the real tracks */
  2042.             for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
  2043.                 char buf[32];
  2044.                 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
  2045.                     continue;
  2046.                 snprintf(buf, sizeof(buf), "%s_%d",
  2047.                          ff_matroska_video_stereo_plane[planes[j].type], i);
  2048.                 for (k=0; k < matroska->tracks.nb_elem; k++)
  2049.                     if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
  2050.                         av_dict_set(&tracks[k].stream->metadata,
  2051.                                     "stereo_mode", buf, 0);
  2052.                         break;
  2053.                     }
  2054.             }
  2055.             // add stream level stereo3d side data if it is a supported format
  2056.             if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
  2057.                 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
  2058.                 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
  2059.                 if (ret < 0)
  2060.                     return ret;
  2061.             }
  2062.         } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
  2063.             st->codec->codec_type  = AVMEDIA_TYPE_AUDIO;
  2064.             st->codec->sample_rate = track->audio.out_samplerate;
  2065.             st->codec->channels    = track->audio.channels;
  2066.             if (!st->codec->bits_per_coded_sample)
  2067.                 st->codec->bits_per_coded_sample = track->audio.bitdepth;
  2068.             if (st->codec->codec_id == AV_CODEC_ID_MP3)
  2069.                 st->need_parsing = AVSTREAM_PARSE_FULL;
  2070.             else if (st->codec->codec_id != AV_CODEC_ID_AAC)
  2071.                 st->need_parsing = AVSTREAM_PARSE_HEADERS;
  2072.             if (track->codec_delay > 0) {
  2073.                 st->codec->delay = av_rescale_q(track->codec_delay,
  2074.                                                 st->time_base,
  2075.                                                 (AVRational){1, st->codec->sample_rate});
  2076.             }
  2077.             if (track->seek_preroll > 0) {
  2078.                 av_codec_set_seek_preroll(st->codec,
  2079.                                           av_rescale_q(track->seek_preroll,
  2080.                                                        (AVRational){1, 1000000000},
  2081.                                                        (AVRational){1, st->codec->sample_rate}));
  2082.             }
  2083.         } else if (codec_id == AV_CODEC_ID_WEBVTT) {
  2084.             st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
  2085.  
  2086.             if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
  2087.                 st->disposition |= AV_DISPOSITION_CAPTIONS;
  2088.             } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
  2089.                 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
  2090.             } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
  2091.                 st->disposition |= AV_DISPOSITION_METADATA;
  2092.             }
  2093.         } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
  2094.             st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
  2095.             if (st->codec->codec_id == AV_CODEC_ID_ASS)
  2096.                 matroska->contains_ssa = 1;
  2097.         }
  2098.     }
  2099.  
  2100.     return 0;
  2101. }
  2102.  
  2103. static int matroska_read_header(AVFormatContext *s)
  2104. {
  2105.     MatroskaDemuxContext *matroska = s->priv_data;
  2106.     EbmlList *attachments_list = &matroska->attachments;
  2107.     EbmlList *chapters_list    = &matroska->chapters;
  2108.     MatroskaAttachment *attachments;
  2109.     MatroskaChapter *chapters;
  2110.     uint64_t max_start = 0;
  2111.     int64_t pos;
  2112.     Ebml ebml = { 0 };
  2113.     int i, j, res;
  2114.  
  2115.     matroska->ctx = s;
  2116.     matroska->cues_parsing_deferred = 1;
  2117.  
  2118.     /* First read the EBML header. */
  2119.     if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
  2120.         av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
  2121.         ebml_free(ebml_syntax, &ebml);
  2122.         return AVERROR_INVALIDDATA;
  2123.     }
  2124.     if (ebml.version         > EBML_VERSION      ||
  2125.         ebml.max_size        > sizeof(uint64_t)  ||
  2126.         ebml.id_length       > sizeof(uint32_t)  ||
  2127.         ebml.doctype_version > 3) {
  2128.         av_log(matroska->ctx, AV_LOG_ERROR,
  2129.                "EBML header using unsupported features\n"
  2130.                "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
  2131.                ebml.version, ebml.doctype, ebml.doctype_version);
  2132.         ebml_free(ebml_syntax, &ebml);
  2133.         return AVERROR_PATCHWELCOME;
  2134.     } else if (ebml.doctype_version == 3) {
  2135.         av_log(matroska->ctx, AV_LOG_WARNING,
  2136.                "EBML header using unsupported features\n"
  2137.                "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
  2138.                ebml.version, ebml.doctype, ebml.doctype_version);
  2139.     }
  2140.     for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
  2141.         if (!strcmp(ebml.doctype, matroska_doctypes[i]))
  2142.             break;
  2143.     if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
  2144.         av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
  2145.         if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
  2146.             ebml_free(ebml_syntax, &ebml);
  2147.             return AVERROR_INVALIDDATA;
  2148.         }
  2149.     }
  2150.     ebml_free(ebml_syntax, &ebml);
  2151.  
  2152.     /* The next thing is a segment. */
  2153.     pos = avio_tell(matroska->ctx->pb);
  2154.     res = ebml_parse(matroska, matroska_segments, matroska);
  2155.     // try resyncing until we find a EBML_STOP type element.
  2156.     while (res != 1) {
  2157.         res = matroska_resync(matroska, pos);
  2158.         if (res < 0)
  2159.             return res;
  2160.         pos = avio_tell(matroska->ctx->pb);
  2161.         res = ebml_parse(matroska, matroska_segment, matroska);
  2162.     }
  2163.     matroska_execute_seekhead(matroska);
  2164.  
  2165.     if (!matroska->time_scale)
  2166.         matroska->time_scale = 1000000;
  2167.     if (matroska->duration)
  2168.         matroska->ctx->duration = matroska->duration * matroska->time_scale *
  2169.                                   1000 / AV_TIME_BASE;
  2170.     av_dict_set(&s->metadata, "title", matroska->title, 0);
  2171.     av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
  2172.  
  2173.     if (matroska->date_utc.size == 8)
  2174.         matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
  2175.  
  2176.     res = matroska_parse_tracks(s);
  2177.     if (res < 0)
  2178.         return res;
  2179.  
  2180.     attachments = attachments_list->elem;
  2181.     for (j = 0; j < attachments_list->nb_elem; j++) {
  2182.         if (!(attachments[j].filename && attachments[j].mime &&
  2183.               attachments[j].bin.data && attachments[j].bin.size > 0)) {
  2184.             av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
  2185.         } else {
  2186.             AVStream *st = avformat_new_stream(s, NULL);
  2187.             if (!st)
  2188.                 break;
  2189.             av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
  2190.             av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
  2191.             st->codec->codec_id   = AV_CODEC_ID_NONE;
  2192.  
  2193.             for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
  2194.                 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
  2195.                              strlen(ff_mkv_image_mime_tags[i].str))) {
  2196.                     st->codec->codec_id = ff_mkv_image_mime_tags[i].id;
  2197.                     break;
  2198.                 }
  2199.             }
  2200.  
  2201.             attachments[j].stream = st;
  2202.  
  2203.             if (st->codec->codec_id != AV_CODEC_ID_NONE) {
  2204.                 st->disposition      |= AV_DISPOSITION_ATTACHED_PIC;
  2205.                 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
  2206.  
  2207.                 av_init_packet(&st->attached_pic);
  2208.                 if ((res = av_new_packet(&st->attached_pic, attachments[j].bin.size)) < 0)
  2209.                     return res;
  2210.                 memcpy(st->attached_pic.data, attachments[j].bin.data, attachments[j].bin.size);
  2211.                 st->attached_pic.stream_index = st->index;
  2212.                 st->attached_pic.flags       |= AV_PKT_FLAG_KEY;
  2213.             } else {
  2214.                 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
  2215.                 if (ff_alloc_extradata(st->codec, attachments[j].bin.size))
  2216.                     break;
  2217.                 memcpy(st->codec->extradata, attachments[j].bin.data,
  2218.                        attachments[j].bin.size);
  2219.  
  2220.                 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
  2221.                     if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
  2222.                                 strlen(ff_mkv_mime_tags[i].str))) {
  2223.                         st->codec->codec_id = ff_mkv_mime_tags[i].id;
  2224.                         break;
  2225.                     }
  2226.                 }
  2227.             }
  2228.         }
  2229.     }
  2230.  
  2231.     chapters = chapters_list->elem;
  2232.     for (i = 0; i < chapters_list->nb_elem; i++)
  2233.         if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
  2234.             (max_start == 0 || chapters[i].start > max_start)) {
  2235.             chapters[i].chapter =
  2236.                 avpriv_new_chapter(s, chapters[i].uid,
  2237.                                    (AVRational) { 1, 1000000000 },
  2238.                                    chapters[i].start, chapters[i].end,
  2239.                                    chapters[i].title);
  2240.             if (chapters[i].chapter) {
  2241.                 av_dict_set(&chapters[i].chapter->metadata,
  2242.                             "title", chapters[i].title, 0);
  2243.             }
  2244.             max_start = chapters[i].start;
  2245.         }
  2246.  
  2247.     matroska_add_index_entries(matroska);
  2248.  
  2249.     matroska_convert_tags(s);
  2250.  
  2251.     return 0;
  2252. }
  2253.  
  2254. /*
  2255.  * Put one packet in an application-supplied AVPacket struct.
  2256.  * Returns 0 on success or -1 on failure.
  2257.  */
  2258. static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
  2259.                                    AVPacket *pkt)
  2260. {
  2261.     if (matroska->num_packets > 0) {
  2262.         memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
  2263.         av_freep(&matroska->packets[0]);
  2264.         if (matroska->num_packets > 1) {
  2265.             void *newpackets;
  2266.             memmove(&matroska->packets[0], &matroska->packets[1],
  2267.                     (matroska->num_packets - 1) * sizeof(AVPacket *));
  2268.             newpackets = av_realloc(matroska->packets,
  2269.                                     (matroska->num_packets - 1) *
  2270.                                     sizeof(AVPacket *));
  2271.             if (newpackets)
  2272.                 matroska->packets = newpackets;
  2273.         } else {
  2274.             av_freep(&matroska->packets);
  2275.             matroska->prev_pkt = NULL;
  2276.         }
  2277.         matroska->num_packets--;
  2278.         return 0;
  2279.     }
  2280.  
  2281.     return -1;
  2282. }
  2283.  
  2284. /*
  2285.  * Free all packets in our internal queue.
  2286.  */
  2287. static void matroska_clear_queue(MatroskaDemuxContext *matroska)
  2288. {
  2289.     matroska->prev_pkt = NULL;
  2290.     if (matroska->packets) {
  2291.         int n;
  2292.         for (n = 0; n < matroska->num_packets; n++) {
  2293.             av_free_packet(matroska->packets[n]);
  2294.             av_freep(&matroska->packets[n]);
  2295.         }
  2296.         av_freep(&matroska->packets);
  2297.         matroska->num_packets = 0;
  2298.     }
  2299. }
  2300.  
  2301. static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
  2302.                                 int *buf_size, int type,
  2303.                                 uint32_t **lace_buf, int *laces)
  2304. {
  2305.     int res = 0, n, size = *buf_size;
  2306.     uint8_t *data = *buf;
  2307.     uint32_t *lace_size;
  2308.  
  2309.     if (!type) {
  2310.         *laces    = 1;
  2311.         *lace_buf = av_mallocz(sizeof(int));
  2312.         if (!*lace_buf)
  2313.             return AVERROR(ENOMEM);
  2314.  
  2315.         *lace_buf[0] = size;
  2316.         return 0;
  2317.     }
  2318.  
  2319.     av_assert0(size > 0);
  2320.     *laces    = *data + 1;
  2321.     data     += 1;
  2322.     size     -= 1;
  2323.     lace_size = av_mallocz(*laces * sizeof(int));
  2324.     if (!lace_size)
  2325.         return AVERROR(ENOMEM);
  2326.  
  2327.     switch (type) {
  2328.     case 0x1: /* Xiph lacing */
  2329.     {
  2330.         uint8_t temp;
  2331.         uint32_t total = 0;
  2332.         for (n = 0; res == 0 && n < *laces - 1; n++) {
  2333.             while (1) {
  2334.                 if (size <= total) {
  2335.                     res = AVERROR_INVALIDDATA;
  2336.                     break;
  2337.                 }
  2338.                 temp          = *data;
  2339.                 total        += temp;
  2340.                 lace_size[n] += temp;
  2341.                 data         += 1;
  2342.                 size         -= 1;
  2343.                 if (temp != 0xff)
  2344.                     break;
  2345.             }
  2346.         }
  2347.         if (size <= total) {
  2348.             res = AVERROR_INVALIDDATA;
  2349.             break;
  2350.         }
  2351.  
  2352.         lace_size[n] = size - total;
  2353.         break;
  2354.     }
  2355.  
  2356.     case 0x2: /* fixed-size lacing */
  2357.         if (size % (*laces)) {
  2358.             res = AVERROR_INVALIDDATA;
  2359.             break;
  2360.         }
  2361.         for (n = 0; n < *laces; n++)
  2362.             lace_size[n] = size / *laces;
  2363.         break;
  2364.  
  2365.     case 0x3: /* EBML lacing */
  2366.     {
  2367.         uint64_t num;
  2368.         uint64_t total;
  2369.         n = matroska_ebmlnum_uint(matroska, data, size, &num);
  2370.         if (n < 0 || num > INT_MAX) {
  2371.             av_log(matroska->ctx, AV_LOG_INFO,
  2372.                    "EBML block data error\n");
  2373.             res = n<0 ? n : AVERROR_INVALIDDATA;
  2374.             break;
  2375.         }
  2376.         data += n;
  2377.         size -= n;
  2378.         total = lace_size[0] = num;
  2379.         for (n = 1; res == 0 && n < *laces - 1; n++) {
  2380.             int64_t snum;
  2381.             int r;
  2382.             r = matroska_ebmlnum_sint(matroska, data, size, &snum);
  2383.             if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
  2384.                 av_log(matroska->ctx, AV_LOG_INFO,
  2385.                        "EBML block data error\n");
  2386.                 res = r<0 ? r : AVERROR_INVALIDDATA;
  2387.                 break;
  2388.             }
  2389.             data        += r;
  2390.             size        -= r;
  2391.             lace_size[n] = lace_size[n - 1] + snum;
  2392.             total       += lace_size[n];
  2393.         }
  2394.         if (size <= total) {
  2395.             res = AVERROR_INVALIDDATA;
  2396.             break;
  2397.         }
  2398.         lace_size[*laces - 1] = size - total;
  2399.         break;
  2400.     }
  2401.     }
  2402.  
  2403.     *buf      = data;
  2404.     *lace_buf = lace_size;
  2405.     *buf_size = size;
  2406.  
  2407.     return res;
  2408. }
  2409.  
  2410. static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
  2411.                                    MatroskaTrack *track, AVStream *st,
  2412.                                    uint8_t *data, int size, uint64_t timecode,
  2413.                                    int64_t pos)
  2414. {
  2415.     int a = st->codec->block_align;
  2416.     int sps = track->audio.sub_packet_size;
  2417.     int cfs = track->audio.coded_framesize;
  2418.     int h   = track->audio.sub_packet_h;
  2419.     int y   = track->audio.sub_packet_cnt;
  2420.     int w   = track->audio.frame_size;
  2421.     int x;
  2422.  
  2423.     if (!track->audio.pkt_cnt) {
  2424.         if (track->audio.sub_packet_cnt == 0)
  2425.             track->audio.buf_timecode = timecode;
  2426.         if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
  2427.             if (size < cfs * h / 2) {
  2428.                 av_log(matroska->ctx, AV_LOG_ERROR,
  2429.                        "Corrupt int4 RM-style audio packet size\n");
  2430.                 return AVERROR_INVALIDDATA;
  2431.             }
  2432.             for (x = 0; x < h / 2; x++)
  2433.                 memcpy(track->audio.buf + x * 2 * w + y * cfs,
  2434.                        data + x * cfs, cfs);
  2435.         } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
  2436.             if (size < w) {
  2437.                 av_log(matroska->ctx, AV_LOG_ERROR,
  2438.                        "Corrupt sipr RM-style audio packet size\n");
  2439.                 return AVERROR_INVALIDDATA;
  2440.             }
  2441.             memcpy(track->audio.buf + y * w, data, w);
  2442.         } else {
  2443.             if (size < sps * w / sps || h<=0 || w%sps) {
  2444.                 av_log(matroska->ctx, AV_LOG_ERROR,
  2445.                        "Corrupt generic RM-style audio packet size\n");
  2446.                 return AVERROR_INVALIDDATA;
  2447.             }
  2448.             for (x = 0; x < w / sps; x++)
  2449.                 memcpy(track->audio.buf +
  2450.                        sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
  2451.                        data + x * sps, sps);
  2452.         }
  2453.  
  2454.         if (++track->audio.sub_packet_cnt >= h) {
  2455.             if (st->codec->codec_id == AV_CODEC_ID_SIPR)
  2456.                 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
  2457.             track->audio.sub_packet_cnt = 0;
  2458.             track->audio.pkt_cnt        = h * w / a;
  2459.         }
  2460.     }
  2461.  
  2462.     while (track->audio.pkt_cnt) {
  2463.         int ret;
  2464.         AVPacket *pkt = av_mallocz(sizeof(AVPacket));
  2465.         if (!pkt)
  2466.             return AVERROR(ENOMEM);
  2467.  
  2468.         ret = av_new_packet(pkt, a);
  2469.         if (ret < 0) {
  2470.             av_free(pkt);
  2471.             return ret;
  2472.         }
  2473.         memcpy(pkt->data,
  2474.                track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
  2475.                a);
  2476.         pkt->pts                  = track->audio.buf_timecode;
  2477.         track->audio.buf_timecode = AV_NOPTS_VALUE;
  2478.         pkt->pos                  = pos;
  2479.         pkt->stream_index         = st->index;
  2480.         dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
  2481.     }
  2482.  
  2483.     return 0;
  2484. }
  2485.  
  2486. /* reconstruct full wavpack blocks from mangled matroska ones */
  2487. static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
  2488.                                   uint8_t **pdst, int *size)
  2489. {
  2490.     uint8_t *dst = NULL;
  2491.     int dstlen   = 0;
  2492.     int srclen   = *size;
  2493.     uint32_t samples;
  2494.     uint16_t ver;
  2495.     int ret, offset = 0;
  2496.  
  2497.     if (srclen < 12 || track->stream->codec->extradata_size < 2)
  2498.         return AVERROR_INVALIDDATA;
  2499.  
  2500.     ver = AV_RL16(track->stream->codec->extradata);
  2501.  
  2502.     samples = AV_RL32(src);
  2503.     src    += 4;
  2504.     srclen -= 4;
  2505.  
  2506.     while (srclen >= 8) {
  2507.         int multiblock;
  2508.         uint32_t blocksize;
  2509.         uint8_t *tmp;
  2510.  
  2511.         uint32_t flags = AV_RL32(src);
  2512.         uint32_t crc   = AV_RL32(src + 4);
  2513.         src    += 8;
  2514.         srclen -= 8;
  2515.  
  2516.         multiblock = (flags & 0x1800) != 0x1800;
  2517.         if (multiblock) {
  2518.             if (srclen < 4) {
  2519.                 ret = AVERROR_INVALIDDATA;
  2520.                 goto fail;
  2521.             }
  2522.             blocksize = AV_RL32(src);
  2523.             src      += 4;
  2524.             srclen   -= 4;
  2525.         } else
  2526.             blocksize = srclen;
  2527.  
  2528.         if (blocksize > srclen) {
  2529.             ret = AVERROR_INVALIDDATA;
  2530.             goto fail;
  2531.         }
  2532.  
  2533.         tmp = av_realloc(dst, dstlen + blocksize + 32);
  2534.         if (!tmp) {
  2535.             ret = AVERROR(ENOMEM);
  2536.             goto fail;
  2537.         }
  2538.         dst     = tmp;
  2539.         dstlen += blocksize + 32;
  2540.  
  2541.         AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k'));   // tag
  2542.         AV_WL32(dst + offset +  4, blocksize + 24);         // blocksize - 8
  2543.         AV_WL16(dst + offset +  8, ver);                    // version
  2544.         AV_WL16(dst + offset + 10, 0);                      // track/index_no
  2545.         AV_WL32(dst + offset + 12, 0);                      // total samples
  2546.         AV_WL32(dst + offset + 16, 0);                      // block index
  2547.         AV_WL32(dst + offset + 20, samples);                // number of samples
  2548.         AV_WL32(dst + offset + 24, flags);                  // flags
  2549.         AV_WL32(dst + offset + 28, crc);                    // crc
  2550.         memcpy(dst + offset + 32, src, blocksize);          // block data
  2551.  
  2552.         src    += blocksize;
  2553.         srclen -= blocksize;
  2554.         offset += blocksize + 32;
  2555.     }
  2556.  
  2557.     *pdst = dst;
  2558.     *size = dstlen;
  2559.  
  2560.     return 0;
  2561.  
  2562. fail:
  2563.     av_freep(&dst);
  2564.     return ret;
  2565. }
  2566.  
  2567. static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
  2568.                                  MatroskaTrack *track,
  2569.                                  AVStream *st,
  2570.                                  uint8_t *data, int data_len,
  2571.                                  uint64_t timecode,
  2572.                                  uint64_t duration,
  2573.                                  int64_t pos)
  2574. {
  2575.     AVPacket *pkt;
  2576.     uint8_t *id, *settings, *text, *buf;
  2577.     int id_len, settings_len, text_len;
  2578.     uint8_t *p, *q;
  2579.     int err;
  2580.  
  2581.     if (data_len <= 0)
  2582.         return AVERROR_INVALIDDATA;
  2583.  
  2584.     p = data;
  2585.     q = data + data_len;
  2586.  
  2587.     id = p;
  2588.     id_len = -1;
  2589.     while (p < q) {
  2590.         if (*p == '\r' || *p == '\n') {
  2591.             id_len = p - id;
  2592.             if (*p == '\r')
  2593.                 p++;
  2594.             break;
  2595.         }
  2596.         p++;
  2597.     }
  2598.  
  2599.     if (p >= q || *p != '\n')
  2600.         return AVERROR_INVALIDDATA;
  2601.     p++;
  2602.  
  2603.     settings = p;
  2604.     settings_len = -1;
  2605.     while (p < q) {
  2606.         if (*p == '\r' || *p == '\n') {
  2607.             settings_len = p - settings;
  2608.             if (*p == '\r')
  2609.                 p++;
  2610.             break;
  2611.         }
  2612.         p++;
  2613.     }
  2614.  
  2615.     if (p >= q || *p != '\n')
  2616.         return AVERROR_INVALIDDATA;
  2617.     p++;
  2618.  
  2619.     text = p;
  2620.     text_len = q - p;
  2621.     while (text_len > 0) {
  2622.         const int len = text_len - 1;
  2623.         const uint8_t c = p[len];
  2624.         if (c != '\r' && c != '\n')
  2625.             break;
  2626.         text_len = len;
  2627.     }
  2628.  
  2629.     if (text_len <= 0)
  2630.         return AVERROR_INVALIDDATA;
  2631.  
  2632.     pkt = av_mallocz(sizeof(*pkt));
  2633.     if (!pkt)
  2634.         return AVERROR(ENOMEM);
  2635.     err = av_new_packet(pkt, text_len);
  2636.     if (err < 0) {
  2637.         av_free(pkt);
  2638.         return AVERROR(err);
  2639.     }
  2640.  
  2641.     memcpy(pkt->data, text, text_len);
  2642.  
  2643.     if (id_len > 0) {
  2644.         buf = av_packet_new_side_data(pkt,
  2645.                                       AV_PKT_DATA_WEBVTT_IDENTIFIER,
  2646.                                       id_len);
  2647.         if (!buf) {
  2648.             av_free(pkt);
  2649.             return AVERROR(ENOMEM);
  2650.         }
  2651.         memcpy(buf, id, id_len);
  2652.     }
  2653.  
  2654.     if (settings_len > 0) {
  2655.         buf = av_packet_new_side_data(pkt,
  2656.                                       AV_PKT_DATA_WEBVTT_SETTINGS,
  2657.                                       settings_len);
  2658.         if (!buf) {
  2659.             av_free(pkt);
  2660.             return AVERROR(ENOMEM);
  2661.         }
  2662.         memcpy(buf, settings, settings_len);
  2663.     }
  2664.  
  2665.     // Do we need this for subtitles?
  2666.     // pkt->flags = AV_PKT_FLAG_KEY;
  2667.  
  2668.     pkt->stream_index = st->index;
  2669.     pkt->pts = timecode;
  2670.  
  2671.     // Do we need this for subtitles?
  2672.     // pkt->dts = timecode;
  2673.  
  2674.     pkt->duration = duration;
  2675.     pkt->pos = pos;
  2676.  
  2677.     dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
  2678.     matroska->prev_pkt = pkt;
  2679.  
  2680.     return 0;
  2681. }
  2682.  
  2683. static int matroska_parse_frame(MatroskaDemuxContext *matroska,
  2684.                                 MatroskaTrack *track, AVStream *st,
  2685.                                 uint8_t *data, int pkt_size,
  2686.                                 uint64_t timecode, uint64_t lace_duration,
  2687.                                 int64_t pos, int is_keyframe,
  2688.                                 uint8_t *additional, uint64_t additional_id, int additional_size,
  2689.                                 int64_t discard_padding)
  2690. {
  2691.     MatroskaTrackEncoding *encodings = track->encodings.elem;
  2692.     uint8_t *pkt_data = data;
  2693.     int offset = 0, res;
  2694.     AVPacket *pkt;
  2695.  
  2696.     if (encodings && !encodings->type && encodings->scope & 1) {
  2697.         res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
  2698.         if (res < 0)
  2699.             return res;
  2700.     }
  2701.  
  2702.     if (st->codec->codec_id == AV_CODEC_ID_WAVPACK) {
  2703.         uint8_t *wv_data;
  2704.         res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
  2705.         if (res < 0) {
  2706.             av_log(matroska->ctx, AV_LOG_ERROR,
  2707.                    "Error parsing a wavpack block.\n");
  2708.             goto fail;
  2709.         }
  2710.         if (pkt_data != data)
  2711.             av_freep(&pkt_data);
  2712.         pkt_data = wv_data;
  2713.     }
  2714.  
  2715.     if (st->codec->codec_id == AV_CODEC_ID_PRORES &&
  2716.         AV_RB32(&data[4]) != MKBETAG('i', 'c', 'p', 'f'))
  2717.         offset = 8;
  2718.  
  2719.     pkt = av_mallocz(sizeof(AVPacket));
  2720.     if (!pkt) {
  2721.         if (pkt_data != data)
  2722.             av_freep(&pkt_data);
  2723.         return AVERROR(ENOMEM);
  2724.     }
  2725.     /* XXX: prevent data copy... */
  2726.     if (av_new_packet(pkt, pkt_size + offset) < 0) {
  2727.         av_free(pkt);
  2728.         res = AVERROR(ENOMEM);
  2729.         goto fail;
  2730.     }
  2731.  
  2732.     if (st->codec->codec_id == AV_CODEC_ID_PRORES && offset == 8) {
  2733.         uint8_t *buf = pkt->data;
  2734.         bytestream_put_be32(&buf, pkt_size);
  2735.         bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
  2736.     }
  2737.  
  2738.     memcpy(pkt->data + offset, pkt_data, pkt_size);
  2739.  
  2740.     if (pkt_data != data)
  2741.         av_freep(&pkt_data);
  2742.  
  2743.     pkt->flags        = is_keyframe;
  2744.     pkt->stream_index = st->index;
  2745.  
  2746.     if (additional_size > 0) {
  2747.         uint8_t *side_data = av_packet_new_side_data(pkt,
  2748.                                                      AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
  2749.                                                      additional_size + 8);
  2750.         if (!side_data) {
  2751.             av_free_packet(pkt);
  2752.             av_free(pkt);
  2753.             return AVERROR(ENOMEM);
  2754.         }
  2755.         AV_WB64(side_data, additional_id);
  2756.         memcpy(side_data + 8, additional, additional_size);
  2757.     }
  2758.  
  2759.     if (discard_padding) {
  2760.         uint8_t *side_data = av_packet_new_side_data(pkt,
  2761.                                                      AV_PKT_DATA_SKIP_SAMPLES,
  2762.                                                      10);
  2763.         if (!side_data) {
  2764.             av_free_packet(pkt);
  2765.             av_free(pkt);
  2766.             return AVERROR(ENOMEM);
  2767.         }
  2768.         AV_WL32(side_data, 0);
  2769.         AV_WL32(side_data + 4, av_rescale_q(discard_padding,
  2770.                                             (AVRational){1, 1000000000},
  2771.                                             (AVRational){1, st->codec->sample_rate}));
  2772.     }
  2773.  
  2774.     if (track->ms_compat)
  2775.         pkt->dts = timecode;
  2776.     else
  2777.         pkt->pts = timecode;
  2778.     pkt->pos = pos;
  2779.     if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
  2780.         /*
  2781.          * For backward compatibility.
  2782.          * Historically, we have put subtitle duration
  2783.          * in convergence_duration, on the off chance
  2784.          * that the time_scale is less than 1us, which
  2785.          * could result in a 32bit overflow on the
  2786.          * normal duration field.
  2787.          */
  2788.         pkt->convergence_duration = lace_duration;
  2789.     }
  2790.  
  2791.     if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE ||
  2792.         lace_duration <= INT_MAX) {
  2793.         /*
  2794.          * For non subtitle tracks, just store the duration
  2795.          * as normal.
  2796.          *
  2797.          * If it's a subtitle track and duration value does
  2798.          * not overflow a uint32, then also store it normally.
  2799.          */
  2800.         pkt->duration = lace_duration;
  2801.     }
  2802.  
  2803.     dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
  2804.     matroska->prev_pkt = pkt;
  2805.  
  2806.     return 0;
  2807.  
  2808. fail:
  2809.     if (pkt_data != data)
  2810.         av_freep(&pkt_data);
  2811.     return res;
  2812. }
  2813.  
  2814. static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
  2815.                                 int size, int64_t pos, uint64_t cluster_time,
  2816.                                 uint64_t block_duration, int is_keyframe,
  2817.                                 uint8_t *additional, uint64_t additional_id, int additional_size,
  2818.                                 int64_t cluster_pos, int64_t discard_padding)
  2819. {
  2820.     uint64_t timecode = AV_NOPTS_VALUE;
  2821.     MatroskaTrack *track;
  2822.     int res = 0;
  2823.     AVStream *st;
  2824.     int16_t block_time;
  2825.     uint32_t *lace_size = NULL;
  2826.     int n, flags, laces = 0;
  2827.     uint64_t num;
  2828.     int trust_default_duration = 1;
  2829.  
  2830.     if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
  2831.         av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
  2832.         return n;
  2833.     }
  2834.     data += n;
  2835.     size -= n;
  2836.  
  2837.     track = matroska_find_track_by_num(matroska, num);
  2838.     if (!track || !track->stream) {
  2839.         av_log(matroska->ctx, AV_LOG_INFO,
  2840.                "Invalid stream %"PRIu64" or size %u\n", num, size);
  2841.         return AVERROR_INVALIDDATA;
  2842.     } else if (size <= 3)
  2843.         return 0;
  2844.     st = track->stream;
  2845.     if (st->discard >= AVDISCARD_ALL)
  2846.         return res;
  2847.     av_assert1(block_duration != AV_NOPTS_VALUE);
  2848.  
  2849.     block_time = sign_extend(AV_RB16(data), 16);
  2850.     data      += 2;
  2851.     flags      = *data++;
  2852.     size      -= 3;
  2853.     if (is_keyframe == -1)
  2854.         is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
  2855.  
  2856.     if (cluster_time != (uint64_t) -1 &&
  2857.         (block_time >= 0 || cluster_time >= -block_time)) {
  2858.         timecode = cluster_time + block_time - track->codec_delay;
  2859.         if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
  2860.             timecode < track->end_timecode)
  2861.             is_keyframe = 0;  /* overlapping subtitles are not key frame */
  2862.         if (is_keyframe)
  2863.             av_add_index_entry(st, cluster_pos, timecode, 0, 0,
  2864.                                AVINDEX_KEYFRAME);
  2865.     }
  2866.  
  2867.     if (matroska->skip_to_keyframe &&
  2868.         track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
  2869.         if (timecode < matroska->skip_to_timecode)
  2870.             return res;
  2871.         if (is_keyframe)
  2872.             matroska->skip_to_keyframe = 0;
  2873.         else if (!st->skip_to_keyframe) {
  2874.             av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
  2875.             matroska->skip_to_keyframe = 0;
  2876.         }
  2877.     }
  2878.  
  2879.     res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
  2880.                                &lace_size, &laces);
  2881.  
  2882.     if (res)
  2883.         goto end;
  2884.  
  2885.     if (track->audio.samplerate == 8000) {
  2886.         // If this is needed for more codecs, then add them here
  2887.         if (st->codec->codec_id == AV_CODEC_ID_AC3) {
  2888.             if (track->audio.samplerate != st->codec->sample_rate || !st->codec->frame_size)
  2889.                 trust_default_duration = 0;
  2890.         }
  2891.     }
  2892.  
  2893.     if (!block_duration && trust_default_duration)
  2894.         block_duration = track->default_duration * laces / matroska->time_scale;
  2895.  
  2896.     if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
  2897.         track->end_timecode =
  2898.             FFMAX(track->end_timecode, timecode + block_duration);
  2899.  
  2900.     for (n = 0; n < laces; n++) {
  2901.         int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
  2902.  
  2903.         if (lace_size[n] > size) {
  2904.             av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
  2905.             break;
  2906.         }
  2907.  
  2908.         if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
  2909.              st->codec->codec_id == AV_CODEC_ID_COOK   ||
  2910.              st->codec->codec_id == AV_CODEC_ID_SIPR   ||
  2911.              st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
  2912.             st->codec->block_align && track->audio.sub_packet_size) {
  2913.             res = matroska_parse_rm_audio(matroska, track, st, data,
  2914.                                           lace_size[n],
  2915.                                           timecode, pos);
  2916.             if (res)
  2917.                 goto end;
  2918.  
  2919.         } else if (st->codec->codec_id == AV_CODEC_ID_WEBVTT) {
  2920.             res = matroska_parse_webvtt(matroska, track, st,
  2921.                                         data, lace_size[n],
  2922.                                         timecode, lace_duration,
  2923.                                         pos);
  2924.             if (res)
  2925.                 goto end;
  2926.         } else {
  2927.             res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
  2928.                                        timecode, lace_duration, pos,
  2929.                                        !n ? is_keyframe : 0,
  2930.                                        additional, additional_id, additional_size,
  2931.                                        discard_padding);
  2932.             if (res)
  2933.                 goto end;
  2934.         }
  2935.  
  2936.         if (timecode != AV_NOPTS_VALUE)
  2937.             timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
  2938.         data += lace_size[n];
  2939.         size -= lace_size[n];
  2940.     }
  2941.  
  2942. end:
  2943.     av_free(lace_size);
  2944.     return res;
  2945. }
  2946.  
  2947. static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
  2948. {
  2949.     EbmlList *blocks_list;
  2950.     MatroskaBlock *blocks;
  2951.     int i, res;
  2952.     res = ebml_parse(matroska,
  2953.                      matroska_cluster_incremental_parsing,
  2954.                      &matroska->current_cluster);
  2955.     if (res == 1) {
  2956.         /* New Cluster */
  2957.         if (matroska->current_cluster_pos)
  2958.             ebml_level_end(matroska);
  2959.         ebml_free(matroska_cluster, &matroska->current_cluster);
  2960.         memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
  2961.         matroska->current_cluster_num_blocks = 0;
  2962.         matroska->current_cluster_pos        = avio_tell(matroska->ctx->pb);
  2963.         matroska->prev_pkt                   = NULL;
  2964.         /* sizeof the ID which was already read */
  2965.         if (matroska->current_id)
  2966.             matroska->current_cluster_pos -= 4;
  2967.         res = ebml_parse(matroska,
  2968.                          matroska_clusters_incremental,
  2969.                          &matroska->current_cluster);
  2970.         /* Try parsing the block again. */
  2971.         if (res == 1)
  2972.             res = ebml_parse(matroska,
  2973.                              matroska_cluster_incremental_parsing,
  2974.                              &matroska->current_cluster);
  2975.     }
  2976.  
  2977.     if (!res &&
  2978.         matroska->current_cluster_num_blocks <
  2979.         matroska->current_cluster.blocks.nb_elem) {
  2980.         blocks_list = &matroska->current_cluster.blocks;
  2981.         blocks      = blocks_list->elem;
  2982.  
  2983.         matroska->current_cluster_num_blocks = blocks_list->nb_elem;
  2984.         i                                    = blocks_list->nb_elem - 1;
  2985.         if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
  2986.             int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
  2987.             uint8_t* additional = blocks[i].additional.size > 0 ?
  2988.                                     blocks[i].additional.data : NULL;
  2989.             if (!blocks[i].non_simple)
  2990.                 blocks[i].duration = 0;
  2991.             res = matroska_parse_block(matroska, blocks[i].bin.data,
  2992.                                        blocks[i].bin.size, blocks[i].bin.pos,
  2993.                                        matroska->current_cluster.timecode,
  2994.                                        blocks[i].duration, is_keyframe,
  2995.                                        additional, blocks[i].additional_id,
  2996.                                        blocks[i].additional.size,
  2997.                                        matroska->current_cluster_pos,
  2998.                                        blocks[i].discard_padding);
  2999.         }
  3000.     }
  3001.  
  3002.     return res;
  3003. }
  3004.  
  3005. static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
  3006. {
  3007.     MatroskaCluster cluster = { 0 };
  3008.     EbmlList *blocks_list;
  3009.     MatroskaBlock *blocks;
  3010.     int i, res;
  3011.     int64_t pos;
  3012.  
  3013.     if (!matroska->contains_ssa)
  3014.         return matroska_parse_cluster_incremental(matroska);
  3015.     pos = avio_tell(matroska->ctx->pb);
  3016.     matroska->prev_pkt = NULL;
  3017.     if (matroska->current_id)
  3018.         pos -= 4;  /* sizeof the ID which was already read */
  3019.     res         = ebml_parse(matroska, matroska_clusters, &cluster);
  3020.     blocks_list = &cluster.blocks;
  3021.     blocks      = blocks_list->elem;
  3022.     for (i = 0; i < blocks_list->nb_elem; i++)
  3023.         if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
  3024.             int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
  3025.             res = matroska_parse_block(matroska, blocks[i].bin.data,
  3026.                                        blocks[i].bin.size, blocks[i].bin.pos,
  3027.                                        cluster.timecode, blocks[i].duration,
  3028.                                        is_keyframe, NULL, 0, 0, pos,
  3029.                                        blocks[i].discard_padding);
  3030.         }
  3031.     ebml_free(matroska_cluster, &cluster);
  3032.     return res;
  3033. }
  3034.  
  3035. static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
  3036. {
  3037.     MatroskaDemuxContext *matroska = s->priv_data;
  3038.  
  3039.     while (matroska_deliver_packet(matroska, pkt)) {
  3040.         int64_t pos = avio_tell(matroska->ctx->pb);
  3041.         if (matroska->done)
  3042.             return AVERROR_EOF;
  3043.         if (matroska_parse_cluster(matroska) < 0)
  3044.             matroska_resync(matroska, pos);
  3045.     }
  3046.  
  3047.     return 0;
  3048. }
  3049.  
  3050. static int matroska_read_seek(AVFormatContext *s, int stream_index,
  3051.                               int64_t timestamp, int flags)
  3052. {
  3053.     MatroskaDemuxContext *matroska = s->priv_data;
  3054.     MatroskaTrack *tracks = NULL;
  3055.     AVStream *st = s->streams[stream_index];
  3056.     int i, index, index_sub, index_min;
  3057.  
  3058.     /* Parse the CUES now since we need the index data to seek. */
  3059.     if (matroska->cues_parsing_deferred > 0) {
  3060.         matroska->cues_parsing_deferred = 0;
  3061.         matroska_parse_cues(matroska);
  3062.     }
  3063.  
  3064.     if (!st->nb_index_entries)
  3065.         goto err;
  3066.     timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
  3067.  
  3068.     if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
  3069.         avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
  3070.                   SEEK_SET);
  3071.         matroska->current_id = 0;
  3072.         while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
  3073.             matroska_clear_queue(matroska);
  3074.             if (matroska_parse_cluster(matroska) < 0)
  3075.                 break;
  3076.         }
  3077.     }
  3078.  
  3079.     matroska_clear_queue(matroska);
  3080.     if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
  3081.         goto err;
  3082.  
  3083.     index_min = index;
  3084.     tracks = matroska->tracks.elem;
  3085.     for (i = 0; i < matroska->tracks.nb_elem; i++) {
  3086.         tracks[i].audio.pkt_cnt        = 0;
  3087.         tracks[i].audio.sub_packet_cnt = 0;
  3088.         tracks[i].audio.buf_timecode   = AV_NOPTS_VALUE;
  3089.         tracks[i].end_timecode         = 0;
  3090.         if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
  3091.             tracks[i].stream &&
  3092.             tracks[i].stream->discard != AVDISCARD_ALL) {
  3093.             index_sub = av_index_search_timestamp(
  3094.                 tracks[i].stream, st->index_entries[index].timestamp,
  3095.                 AVSEEK_FLAG_BACKWARD);
  3096.             while (index_sub >= 0 &&
  3097.                   index_min > 0 &&
  3098.                   tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
  3099.                   st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
  3100.                 index_min--;
  3101.         }
  3102.     }
  3103.  
  3104.     avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
  3105.     matroska->current_id       = 0;
  3106.     if (flags & AVSEEK_FLAG_ANY) {
  3107.         st->skip_to_keyframe = 0;
  3108.         matroska->skip_to_timecode = timestamp;
  3109.     } else {
  3110.         st->skip_to_keyframe = 1;
  3111.         matroska->skip_to_timecode = st->index_entries[index].timestamp;
  3112.     }
  3113.     matroska->skip_to_keyframe = 1;
  3114.     matroska->done             = 0;
  3115.     matroska->num_levels       = 0;
  3116.     ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
  3117.     return 0;
  3118. err:
  3119.     // slightly hackish but allows proper fallback to
  3120.     // the generic seeking code.
  3121.     matroska_clear_queue(matroska);
  3122.     matroska->current_id = 0;
  3123.     st->skip_to_keyframe =
  3124.     matroska->skip_to_keyframe = 0;
  3125.     matroska->done = 0;
  3126.     matroska->num_levels = 0;
  3127.     return -1;
  3128. }
  3129.  
  3130. static int matroska_read_close(AVFormatContext *s)
  3131. {
  3132.     MatroskaDemuxContext *matroska = s->priv_data;
  3133.     MatroskaTrack *tracks = matroska->tracks.elem;
  3134.     int n;
  3135.  
  3136.     matroska_clear_queue(matroska);
  3137.  
  3138.     for (n = 0; n < matroska->tracks.nb_elem; n++)
  3139.         if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
  3140.             av_freep(&tracks[n].audio.buf);
  3141.     ebml_free(matroska_cluster, &matroska->current_cluster);
  3142.     ebml_free(matroska_segment, matroska);
  3143.  
  3144.     return 0;
  3145. }
  3146.  
  3147. typedef struct {
  3148.     int64_t start_time_ns;
  3149.     int64_t end_time_ns;
  3150.     int64_t start_offset;
  3151.     int64_t end_offset;
  3152. } CueDesc;
  3153.  
  3154. /* This function searches all the Cues and returns the CueDesc corresponding the
  3155.  * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
  3156.  * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
  3157.  */
  3158. static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
  3159.     MatroskaDemuxContext *matroska = s->priv_data;
  3160.     CueDesc cue_desc;
  3161.     int i;
  3162.     int nb_index_entries = s->streams[0]->nb_index_entries;
  3163.     AVIndexEntry *index_entries = s->streams[0]->index_entries;
  3164.     if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
  3165.     for (i = 1; i < nb_index_entries; i++) {
  3166.         if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
  3167.             index_entries[i].timestamp * matroska->time_scale > ts) {
  3168.             break;
  3169.         }
  3170.     }
  3171.     --i;
  3172.     cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
  3173.     cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
  3174.     if (i != nb_index_entries - 1) {
  3175.         cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
  3176.         cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
  3177.     } else {
  3178.         cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
  3179.         // FIXME: this needs special handling for files where Cues appear
  3180.         // before Clusters. the current logic assumes Cues appear after
  3181.         // Clusters.
  3182.         cue_desc.end_offset = cues_start - matroska->segment_start;
  3183.     }
  3184.     return cue_desc;
  3185. }
  3186.  
  3187. static int webm_clusters_start_with_keyframe(AVFormatContext *s)
  3188. {
  3189.     MatroskaDemuxContext *matroska = s->priv_data;
  3190.     int64_t cluster_pos, before_pos;
  3191.     int index, rv = 1;
  3192.     if (s->streams[0]->nb_index_entries <= 0) return 0;
  3193.     // seek to the first cluster using cues.
  3194.     index = av_index_search_timestamp(s->streams[0], 0, 0);
  3195.     if (index < 0)  return 0;
  3196.     cluster_pos = s->streams[0]->index_entries[index].pos;
  3197.     before_pos = avio_tell(s->pb);
  3198.     while (1) {
  3199.         int64_t cluster_id = 0, cluster_length = 0;
  3200.         AVPacket *pkt;
  3201.         avio_seek(s->pb, cluster_pos, SEEK_SET);
  3202.         // read cluster id and length
  3203.         ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id);
  3204.         ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
  3205.         if (cluster_id != 0xF43B675) { // done with all clusters
  3206.             break;
  3207.         }
  3208.         avio_seek(s->pb, cluster_pos, SEEK_SET);
  3209.         matroska->current_id = 0;
  3210.         matroska_clear_queue(matroska);
  3211.         if (matroska_parse_cluster(matroska) < 0 ||
  3212.             matroska->num_packets <= 0) {
  3213.             break;
  3214.         }
  3215.         pkt = matroska->packets[0];
  3216.         cluster_pos += cluster_length + 12; // 12 is the offset of the cluster id and length.
  3217.         if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
  3218.             rv = 0;
  3219.             break;
  3220.         }
  3221.     }
  3222.     avio_seek(s->pb, before_pos, SEEK_SET);
  3223.     return rv;
  3224. }
  3225.  
  3226. static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
  3227.                                              double min_buffer, double* buffer,
  3228.                                              double* sec_to_download, AVFormatContext *s,
  3229.                                              int64_t cues_start)
  3230. {
  3231.     double nano_seconds_per_second = 1000000000.0;
  3232.     double time_sec = time_ns / nano_seconds_per_second;
  3233.     int rv = 0;
  3234.     int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
  3235.     int64_t end_time_ns = time_ns + time_to_search_ns;
  3236.     double sec_downloaded = 0.0;
  3237.     CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
  3238.     if (desc_curr.start_time_ns == -1)
  3239.       return -1;
  3240.     *sec_to_download = 0.0;
  3241.  
  3242.     // Check for non cue start time.
  3243.     if (time_ns > desc_curr.start_time_ns) {
  3244.       int64_t cue_nano = desc_curr.end_time_ns - time_ns;
  3245.       double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
  3246.       double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
  3247.       double timeToDownload = (cueBytes * 8.0) / bps;
  3248.  
  3249.       sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
  3250.       *sec_to_download += timeToDownload;
  3251.  
  3252.       // Check if the search ends within the first cue.
  3253.       if (desc_curr.end_time_ns >= end_time_ns) {
  3254.           double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
  3255.           double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
  3256.           sec_downloaded = percent_to_sub * sec_downloaded;
  3257.           *sec_to_download = percent_to_sub * *sec_to_download;
  3258.       }
  3259.  
  3260.       if ((sec_downloaded + *buffer) <= min_buffer) {
  3261.           return 1;
  3262.       }
  3263.  
  3264.       // Get the next Cue.
  3265.       desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
  3266.     }
  3267.  
  3268.     while (desc_curr.start_time_ns != -1) {
  3269.         int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
  3270.         int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
  3271.         double desc_sec = desc_ns / nano_seconds_per_second;
  3272.         double bits = (desc_bytes * 8.0);
  3273.         double time_to_download = bits / bps;
  3274.  
  3275.         sec_downloaded += desc_sec - time_to_download;
  3276.         *sec_to_download += time_to_download;
  3277.  
  3278.         if (desc_curr.end_time_ns >= end_time_ns) {
  3279.             double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
  3280.             double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
  3281.             sec_downloaded = percent_to_sub * sec_downloaded;
  3282.             *sec_to_download = percent_to_sub * *sec_to_download;
  3283.  
  3284.             if ((sec_downloaded + *buffer) <= min_buffer)
  3285.                 rv = 1;
  3286.             break;
  3287.         }
  3288.  
  3289.         if ((sec_downloaded + *buffer) <= min_buffer) {
  3290.             rv = 1;
  3291.             break;
  3292.         }
  3293.  
  3294.         desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
  3295.     }
  3296.     *buffer = *buffer + sec_downloaded;
  3297.     return rv;
  3298. }
  3299.  
  3300. /* This function computes the bandwidth of the WebM file with the help of
  3301.  * buffer_size_after_time_downloaded() function. Both of these functions are
  3302.  * adapted from WebM Tools project and are adapted to work with FFmpeg's
  3303.  * Matroska parsing mechanism.
  3304.  *
  3305.  * Returns the bandwidth of the file on success; -1 on error.
  3306.  * */
  3307. static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
  3308. {
  3309.     MatroskaDemuxContext *matroska = s->priv_data;
  3310.     AVStream *st = s->streams[0];
  3311.     double bandwidth = 0.0;
  3312.     int i;
  3313.  
  3314.     for (i = 0; i < st->nb_index_entries; i++) {
  3315.         int64_t prebuffer_ns = 1000000000;
  3316.         int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
  3317.         double nano_seconds_per_second = 1000000000.0;
  3318.         int64_t prebuffered_ns = time_ns + prebuffer_ns;
  3319.         double prebuffer_bytes = 0.0;
  3320.         int64_t temp_prebuffer_ns = prebuffer_ns;
  3321.         int64_t pre_bytes, pre_ns;
  3322.         double pre_sec, prebuffer, bits_per_second;
  3323.         CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
  3324.  
  3325.         // Start with the first Cue.
  3326.         CueDesc desc_end = desc_beg;
  3327.  
  3328.         // Figure out how much data we have downloaded for the prebuffer. This will
  3329.         // be used later to adjust the bits per sample to try.
  3330.         while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
  3331.             // Prebuffered the entire Cue.
  3332.             prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
  3333.             temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
  3334.             desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
  3335.         }
  3336.         if (desc_end.start_time_ns == -1) {
  3337.             // The prebuffer is larger than the duration.
  3338.             if (matroska->duration * matroska->time_scale >= prebuffered_ns)
  3339.               return -1;
  3340.             bits_per_second = 0.0;
  3341.         } else {
  3342.             // The prebuffer ends in the last Cue. Estimate how much data was
  3343.             // prebuffered.
  3344.             pre_bytes = desc_end.end_offset - desc_end.start_offset;
  3345.             pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
  3346.             pre_sec = pre_ns / nano_seconds_per_second;
  3347.             prebuffer_bytes +=
  3348.                 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
  3349.  
  3350.             prebuffer = prebuffer_ns / nano_seconds_per_second;
  3351.  
  3352.             // Set this to 0.0 in case our prebuffer buffers the entire video.
  3353.             bits_per_second = 0.0;
  3354.             do {
  3355.                 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
  3356.                 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
  3357.                 double desc_sec = desc_ns / nano_seconds_per_second;
  3358.                 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
  3359.  
  3360.                 // Drop the bps by the percentage of bytes buffered.
  3361.                 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
  3362.                 double mod_bits_per_second = calc_bits_per_second * percent;
  3363.  
  3364.                 if (prebuffer < desc_sec) {
  3365.                     double search_sec =
  3366.                         (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
  3367.  
  3368.                     // Add 1 so the bits per second should be a little bit greater than file
  3369.                     // datarate.
  3370.                     int64_t bps = (int64_t)(mod_bits_per_second) + 1;
  3371.                     const double min_buffer = 0.0;
  3372.                     double buffer = prebuffer;
  3373.                     double sec_to_download = 0.0;
  3374.  
  3375.                     int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
  3376.                                                                min_buffer, &buffer, &sec_to_download,
  3377.                                                                s, cues_start);
  3378.                     if (rv < 0) {
  3379.                         return -1;
  3380.                     } else if (rv == 0) {
  3381.                         bits_per_second = (double)(bps);
  3382.                         break;
  3383.                     }
  3384.                 }
  3385.  
  3386.                 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
  3387.             } while (desc_end.start_time_ns != -1);
  3388.         }
  3389.         if (bandwidth < bits_per_second) bandwidth = bits_per_second;
  3390.     }
  3391.     return (int64_t)bandwidth;
  3392. }
  3393.  
  3394. static int webm_dash_manifest_cues(AVFormatContext *s)
  3395. {
  3396.     MatroskaDemuxContext *matroska = s->priv_data;
  3397.     EbmlList *seekhead_list = &matroska->seekhead;
  3398.     MatroskaSeekhead *seekhead = seekhead_list->elem;
  3399.     char *buf;
  3400.     int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
  3401.     int i;
  3402.  
  3403.     // determine cues start and end positions
  3404.     for (i = 0; i < seekhead_list->nb_elem; i++)
  3405.         if (seekhead[i].id == MATROSKA_ID_CUES)
  3406.             break;
  3407.  
  3408.     if (i >= seekhead_list->nb_elem) return -1;
  3409.  
  3410.     before_pos = avio_tell(matroska->ctx->pb);
  3411.     cues_start = seekhead[i].pos + matroska->segment_start;
  3412.     if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
  3413.         // cues_end is computed as cues_start + cues_length + length of the
  3414.         // Cues element ID + EBML length of the Cues element. cues_end is
  3415.         // inclusive and the above sum is reduced by 1.
  3416.         uint64_t cues_length = 0, cues_id = 0, bytes_read = 0;
  3417.         bytes_read += ebml_read_num(matroska, matroska->ctx->pb, 4, &cues_id);
  3418.         bytes_read += ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
  3419.         cues_end = cues_start + cues_length + bytes_read - 1;
  3420.     }
  3421.     avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
  3422.     if (cues_start == -1 || cues_end == -1) return -1;
  3423.  
  3424.     // parse the cues
  3425.     matroska_parse_cues(matroska);
  3426.  
  3427.     // cues start
  3428.     av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
  3429.  
  3430.     // cues end
  3431.     av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
  3432.  
  3433.     // bandwidth
  3434.     bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
  3435.     if (bandwidth < 0) return -1;
  3436.     av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
  3437.  
  3438.     // check if all clusters start with key frames
  3439.     av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
  3440.  
  3441.     // store cue point timestamps as a comma separated list for checking subsegment alignment in
  3442.     // the muxer. assumes that each timestamp cannot be more than 20 characters long.
  3443.     buf = av_malloc_array(s->streams[0]->nb_index_entries, 20 * sizeof(char));
  3444.     if (!buf) return -1;
  3445.     strcpy(buf, "");
  3446.     for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
  3447.         snprintf(buf, (i + 1) * 20 * sizeof(char),
  3448.                  "%s%" PRId64, buf, s->streams[0]->index_entries[i].timestamp);
  3449.         if (i != s->streams[0]->nb_index_entries - 1)
  3450.             strncat(buf, ",", sizeof(char));
  3451.     }
  3452.     av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
  3453.     av_free(buf);
  3454.  
  3455.     return 0;
  3456. }
  3457.  
  3458. static int webm_dash_manifest_read_header(AVFormatContext *s)
  3459. {
  3460.     char *buf;
  3461.     int ret = matroska_read_header(s);
  3462.     MatroskaTrack *tracks;
  3463.     MatroskaDemuxContext *matroska = s->priv_data;
  3464.     if (ret) {
  3465.         av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
  3466.         return -1;
  3467.     }
  3468.  
  3469.     if (!matroska->is_live) {
  3470.         buf = av_asprintf("%g", matroska->duration);
  3471.         if (!buf) return AVERROR(ENOMEM);
  3472.         av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
  3473.         av_free(buf);
  3474.  
  3475.         // initialization range
  3476.         // 5 is the offset of Cluster ID.
  3477.         av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, avio_tell(s->pb) - 5, 0);
  3478.     }
  3479.  
  3480.     // basename of the file
  3481.     buf = strrchr(s->filename, '/');
  3482.     av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->filename, 0);
  3483.  
  3484.     // track number
  3485.     tracks = matroska->tracks.elem;
  3486.     av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
  3487.  
  3488.     // parse the cues and populate Cue related fields
  3489.     return matroska->is_live ? 0 : webm_dash_manifest_cues(s);
  3490. }
  3491.  
  3492. static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
  3493. {
  3494.     return AVERROR_EOF;
  3495. }
  3496.  
  3497. #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
  3498. static const AVOption options[] = {
  3499.     { "live", "flag indicating that the input is a live file that only has the headers.", OFFSET(is_live), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AV_OPT_FLAG_DECODING_PARAM },
  3500.     { NULL },
  3501. };
  3502.  
  3503. static const AVClass webm_dash_class = {
  3504.     .class_name = "WebM DASH Manifest demuxer",
  3505.     .item_name  = av_default_item_name,
  3506.     .option     = options,
  3507.     .version    = LIBAVUTIL_VERSION_INT,
  3508. };
  3509.  
  3510. AVInputFormat ff_matroska_demuxer = {
  3511.     .name           = "matroska,webm",
  3512.     .long_name      = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
  3513.     .extensions     = "mkv,mk3d,mka,mks",
  3514.     .priv_data_size = sizeof(MatroskaDemuxContext),
  3515.     .read_probe     = matroska_probe,
  3516.     .read_header    = matroska_read_header,
  3517.     .read_packet    = matroska_read_packet,
  3518.     .read_close     = matroska_read_close,
  3519.     .read_seek      = matroska_read_seek,
  3520.     .mime_type      = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
  3521. };
  3522.  
  3523. AVInputFormat ff_webm_dash_manifest_demuxer = {
  3524.     .name           = "webm_dash_manifest",
  3525.     .long_name      = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
  3526.     .priv_data_size = sizeof(MatroskaDemuxContext),
  3527.     .read_header    = webm_dash_manifest_read_header,
  3528.     .read_packet    = webm_dash_manifest_read_packet,
  3529.     .read_close     = matroska_read_close,
  3530.     .priv_class     = &webm_dash_class,
  3531. };
  3532.