You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1901 lines
68KB

  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. * @file
  23. * Matroska file demuxer
  24. * by Ronald Bultje <rbultje@ronald.bitfreak.net>
  25. * with a little help from Moritz Bunkus <moritz@bunkus.org>
  26. * totally reworked by Aurelien Jacobs <aurel@gnuage.org>
  27. * Specs available on the Matroska project page: http://www.matroska.org/.
  28. */
  29. #include <stdio.h>
  30. #include "avformat.h"
  31. #include "internal.h"
  32. /* For ff_codec_get_id(). */
  33. #include "riff.h"
  34. #include "isom.h"
  35. #include "rm.h"
  36. #include "matroska.h"
  37. #include "libavcodec/mpeg4audio.h"
  38. #include "libavutil/intfloat_readwrite.h"
  39. #include "libavutil/intreadwrite.h"
  40. #include "libavutil/avstring.h"
  41. #include "libavutil/lzo.h"
  42. #if CONFIG_ZLIB
  43. #include <zlib.h>
  44. #endif
  45. #if CONFIG_BZLIB
  46. #include <bzlib.h>
  47. #endif
  48. typedef enum {
  49. EBML_NONE,
  50. EBML_UINT,
  51. EBML_FLOAT,
  52. EBML_STR,
  53. EBML_UTF8,
  54. EBML_BIN,
  55. EBML_NEST,
  56. EBML_PASS,
  57. EBML_STOP,
  58. } EbmlType;
  59. typedef const struct EbmlSyntax {
  60. uint32_t id;
  61. EbmlType type;
  62. int list_elem_size;
  63. int data_offset;
  64. union {
  65. uint64_t u;
  66. double f;
  67. const char *s;
  68. const struct EbmlSyntax *n;
  69. } def;
  70. } EbmlSyntax;
  71. typedef struct {
  72. int nb_elem;
  73. void *elem;
  74. } EbmlList;
  75. typedef struct {
  76. int size;
  77. uint8_t *data;
  78. int64_t pos;
  79. } EbmlBin;
  80. typedef struct {
  81. uint64_t version;
  82. uint64_t max_size;
  83. uint64_t id_length;
  84. char *doctype;
  85. uint64_t doctype_version;
  86. } Ebml;
  87. typedef struct {
  88. uint64_t algo;
  89. EbmlBin settings;
  90. } MatroskaTrackCompression;
  91. typedef struct {
  92. uint64_t scope;
  93. uint64_t type;
  94. MatroskaTrackCompression compression;
  95. } MatroskaTrackEncoding;
  96. typedef struct {
  97. double frame_rate;
  98. uint64_t display_width;
  99. uint64_t display_height;
  100. uint64_t pixel_width;
  101. uint64_t pixel_height;
  102. uint64_t fourcc;
  103. } MatroskaTrackVideo;
  104. typedef struct {
  105. double samplerate;
  106. double out_samplerate;
  107. uint64_t bitdepth;
  108. uint64_t channels;
  109. /* real audio header (extracted from extradata) */
  110. int coded_framesize;
  111. int sub_packet_h;
  112. int frame_size;
  113. int sub_packet_size;
  114. int sub_packet_cnt;
  115. int pkt_cnt;
  116. uint8_t *buf;
  117. } MatroskaTrackAudio;
  118. typedef struct {
  119. uint64_t num;
  120. uint64_t uid;
  121. uint64_t type;
  122. char *name;
  123. char *codec_id;
  124. EbmlBin codec_priv;
  125. char *language;
  126. double time_scale;
  127. uint64_t default_duration;
  128. uint64_t flag_default;
  129. MatroskaTrackVideo video;
  130. MatroskaTrackAudio audio;
  131. EbmlList encodings;
  132. AVStream *stream;
  133. int64_t end_timecode;
  134. int ms_compat;
  135. } MatroskaTrack;
  136. typedef struct {
  137. uint64_t uid;
  138. char *filename;
  139. char *mime;
  140. EbmlBin bin;
  141. AVStream *stream;
  142. } MatroskaAttachement;
  143. typedef struct {
  144. uint64_t start;
  145. uint64_t end;
  146. uint64_t uid;
  147. char *title;
  148. AVChapter *chapter;
  149. } MatroskaChapter;
  150. typedef struct {
  151. uint64_t track;
  152. uint64_t pos;
  153. } MatroskaIndexPos;
  154. typedef struct {
  155. uint64_t time;
  156. EbmlList pos;
  157. } MatroskaIndex;
  158. typedef struct {
  159. char *name;
  160. char *string;
  161. char *lang;
  162. uint64_t def;
  163. EbmlList sub;
  164. } MatroskaTag;
  165. typedef struct {
  166. char *type;
  167. uint64_t typevalue;
  168. uint64_t trackuid;
  169. uint64_t chapteruid;
  170. uint64_t attachuid;
  171. } MatroskaTagTarget;
  172. typedef struct {
  173. MatroskaTagTarget target;
  174. EbmlList tag;
  175. } MatroskaTags;
  176. typedef struct {
  177. uint64_t id;
  178. uint64_t pos;
  179. } MatroskaSeekhead;
  180. typedef struct {
  181. uint64_t start;
  182. uint64_t length;
  183. } MatroskaLevel;
  184. typedef struct {
  185. AVFormatContext *ctx;
  186. /* EBML stuff */
  187. int num_levels;
  188. MatroskaLevel levels[EBML_MAX_DEPTH];
  189. int level_up;
  190. uint64_t time_scale;
  191. double duration;
  192. char *title;
  193. EbmlList tracks;
  194. EbmlList attachments;
  195. EbmlList chapters;
  196. EbmlList index;
  197. EbmlList tags;
  198. EbmlList seekhead;
  199. /* byte position of the segment inside the stream */
  200. int64_t segment_start;
  201. /* the packet queue */
  202. AVPacket **packets;
  203. int num_packets;
  204. AVPacket *prev_pkt;
  205. int done;
  206. int has_cluster_id;
  207. /* What to skip before effectively reading a packet. */
  208. int skip_to_keyframe;
  209. uint64_t skip_to_timecode;
  210. } MatroskaDemuxContext;
  211. typedef struct {
  212. uint64_t duration;
  213. int64_t reference;
  214. uint64_t non_simple;
  215. EbmlBin bin;
  216. } MatroskaBlock;
  217. typedef struct {
  218. uint64_t timecode;
  219. EbmlList blocks;
  220. } MatroskaCluster;
  221. static EbmlSyntax ebml_header[] = {
  222. { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml,version), {.u=EBML_VERSION} },
  223. { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml,max_size), {.u=8} },
  224. { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml,id_length), {.u=4} },
  225. { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml,doctype), {.s="(none)"} },
  226. { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml,doctype_version), {.u=1} },
  227. { EBML_ID_EBMLVERSION, EBML_NONE },
  228. { EBML_ID_DOCTYPEVERSION, EBML_NONE },
  229. { 0 }
  230. };
  231. static EbmlSyntax ebml_syntax[] = {
  232. { EBML_ID_HEADER, EBML_NEST, 0, 0, {.n=ebml_header} },
  233. { 0 }
  234. };
  235. static EbmlSyntax matroska_info[] = {
  236. { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext,time_scale), {.u=1000000} },
  237. { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext,duration) },
  238. { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext,title) },
  239. { MATROSKA_ID_WRITINGAPP, EBML_NONE },
  240. { MATROSKA_ID_MUXINGAPP, EBML_NONE },
  241. { MATROSKA_ID_DATEUTC, EBML_NONE },
  242. { MATROSKA_ID_SEGMENTUID, EBML_NONE },
  243. { 0 }
  244. };
  245. static EbmlSyntax matroska_track_video[] = {
  246. { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT,0, offsetof(MatroskaTrackVideo,frame_rate) },
  247. { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_width) },
  248. { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,display_height) },
  249. { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_width) },
  250. { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo,pixel_height) },
  251. { MATROSKA_ID_VIDEOCOLORSPACE, EBML_UINT, 0, offsetof(MatroskaTrackVideo,fourcc) },
  252. { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
  253. { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
  254. { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
  255. { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
  256. { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
  257. { MATROSKA_ID_VIDEOFLAGINTERLACED,EBML_NONE },
  258. { MATROSKA_ID_VIDEOSTEREOMODE, EBML_NONE },
  259. { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
  260. { 0 }
  261. };
  262. static EbmlSyntax matroska_track_audio[] = {
  263. { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT,0, offsetof(MatroskaTrackAudio,samplerate), {.f=8000.0} },
  264. { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ,EBML_FLOAT,0,offsetof(MatroskaTrackAudio,out_samplerate) },
  265. { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio,bitdepth) },
  266. { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio,channels), {.u=1} },
  267. { 0 }
  268. };
  269. static EbmlSyntax matroska_track_encoding_compression[] = {
  270. { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression,algo), {.u=0} },
  271. { MATROSKA_ID_ENCODINGCOMPSETTINGS,EBML_BIN, 0, offsetof(MatroskaTrackCompression,settings) },
  272. { 0 }
  273. };
  274. static EbmlSyntax matroska_track_encoding[] = {
  275. { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,scope), {.u=1} },
  276. { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding,type), {.u=0} },
  277. { MATROSKA_ID_ENCODINGCOMPRESSION,EBML_NEST, 0, offsetof(MatroskaTrackEncoding,compression), {.n=matroska_track_encoding_compression} },
  278. { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
  279. { 0 }
  280. };
  281. static EbmlSyntax matroska_track_encodings[] = {
  282. { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack,encodings), {.n=matroska_track_encoding} },
  283. { 0 }
  284. };
  285. static EbmlSyntax matroska_track[] = {
  286. { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack,num) },
  287. { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack,name) },
  288. { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack,uid) },
  289. { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack,type) },
  290. { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack,codec_id) },
  291. { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack,codec_priv) },
  292. { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack,language), {.s="eng"} },
  293. { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack,default_duration) },
  294. { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT,0, offsetof(MatroskaTrack,time_scale), {.f=1.0} },
  295. { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack,flag_default), {.u=1} },
  296. { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack,video), {.n=matroska_track_video} },
  297. { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack,audio), {.n=matroska_track_audio} },
  298. { MATROSKA_ID_TRACKCONTENTENCODINGS,EBML_NEST, 0, 0, {.n=matroska_track_encodings} },
  299. { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
  300. { MATROSKA_ID_TRACKFLAGFORCED, EBML_NONE },
  301. { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
  302. { MATROSKA_ID_CODECNAME, EBML_NONE },
  303. { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
  304. { MATROSKA_ID_CODECINFOURL, EBML_NONE },
  305. { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
  306. { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
  307. { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
  308. { MATROSKA_ID_TRACKMAXBLKADDID, EBML_NONE },
  309. { 0 }
  310. };
  311. static EbmlSyntax matroska_tracks[] = {
  312. { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext,tracks), {.n=matroska_track} },
  313. { 0 }
  314. };
  315. static EbmlSyntax matroska_attachment[] = {
  316. { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachement,uid) },
  317. { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachement,filename) },
  318. { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachement,mime) },
  319. { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachement,bin) },
  320. { MATROSKA_ID_FILEDESC, EBML_NONE },
  321. { 0 }
  322. };
  323. static EbmlSyntax matroska_attachments[] = {
  324. { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachement), offsetof(MatroskaDemuxContext,attachments), {.n=matroska_attachment} },
  325. { 0 }
  326. };
  327. static EbmlSyntax matroska_chapter_display[] = {
  328. { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter,title) },
  329. { MATROSKA_ID_CHAPLANG, EBML_NONE },
  330. { 0 }
  331. };
  332. static EbmlSyntax matroska_chapter_entry[] = {
  333. { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter,start), {.u=AV_NOPTS_VALUE} },
  334. { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter,end), {.u=AV_NOPTS_VALUE} },
  335. { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter,uid) },
  336. { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, {.n=matroska_chapter_display} },
  337. { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
  338. { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
  339. { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
  340. { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
  341. { 0 }
  342. };
  343. static EbmlSyntax matroska_chapter[] = {
  344. { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext,chapters), {.n=matroska_chapter_entry} },
  345. { MATROSKA_ID_EDITIONUID, EBML_NONE },
  346. { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
  347. { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
  348. { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
  349. { 0 }
  350. };
  351. static EbmlSyntax matroska_chapters[] = {
  352. { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, {.n=matroska_chapter} },
  353. { 0 }
  354. };
  355. static EbmlSyntax matroska_index_pos[] = {
  356. { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos,track) },
  357. { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos,pos) },
  358. { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
  359. { 0 }
  360. };
  361. static EbmlSyntax matroska_index_entry[] = {
  362. { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex,time) },
  363. { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex,pos), {.n=matroska_index_pos} },
  364. { 0 }
  365. };
  366. static EbmlSyntax matroska_index[] = {
  367. { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext,index), {.n=matroska_index_entry} },
  368. { 0 }
  369. };
  370. static EbmlSyntax matroska_simpletag[] = {
  371. { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag,name) },
  372. { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag,string) },
  373. { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag,lang), {.s="und"} },
  374. { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag,def) },
  375. { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag,def) },
  376. { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag,sub), {.n=matroska_simpletag} },
  377. { 0 }
  378. };
  379. static EbmlSyntax matroska_tagtargets[] = {
  380. { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget,type) },
  381. { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget,typevalue), {.u=50} },
  382. { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,trackuid) },
  383. { MATROSKA_ID_TAGTARGETS_CHAPTERUID,EBML_UINT, 0, offsetof(MatroskaTagTarget,chapteruid) },
  384. { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget,attachuid) },
  385. { 0 }
  386. };
  387. static EbmlSyntax matroska_tag[] = {
  388. { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags,tag), {.n=matroska_simpletag} },
  389. { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags,target), {.n=matroska_tagtargets} },
  390. { 0 }
  391. };
  392. static EbmlSyntax matroska_tags[] = {
  393. { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext,tags), {.n=matroska_tag} },
  394. { 0 }
  395. };
  396. static EbmlSyntax matroska_seekhead_entry[] = {
  397. { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead,id) },
  398. { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead,pos), {.u=-1} },
  399. { 0 }
  400. };
  401. static EbmlSyntax matroska_seekhead[] = {
  402. { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext,seekhead), {.n=matroska_seekhead_entry} },
  403. { 0 }
  404. };
  405. static EbmlSyntax matroska_segment[] = {
  406. { MATROSKA_ID_INFO, EBML_NEST, 0, 0, {.n=matroska_info } },
  407. { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, {.n=matroska_tracks } },
  408. { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, {.n=matroska_attachments} },
  409. { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, {.n=matroska_chapters } },
  410. { MATROSKA_ID_CUES, EBML_NEST, 0, 0, {.n=matroska_index } },
  411. { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, {.n=matroska_tags } },
  412. { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, {.n=matroska_seekhead } },
  413. { MATROSKA_ID_CLUSTER, EBML_STOP, 0, offsetof(MatroskaDemuxContext,has_cluster_id) },
  414. { 0 }
  415. };
  416. static EbmlSyntax matroska_segments[] = {
  417. { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, {.n=matroska_segment } },
  418. { 0 }
  419. };
  420. static EbmlSyntax matroska_blockgroup[] = {
  421. { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
  422. { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock,bin) },
  423. { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock,duration), {.u=AV_NOPTS_VALUE} },
  424. { MATROSKA_ID_BLOCKREFERENCE, EBML_UINT, 0, offsetof(MatroskaBlock,reference) },
  425. { 1, EBML_UINT, 0, offsetof(MatroskaBlock,non_simple), {.u=1} },
  426. { 0 }
  427. };
  428. static EbmlSyntax matroska_cluster[] = {
  429. { MATROSKA_ID_CLUSTERTIMECODE,EBML_UINT,0, offsetof(MatroskaCluster,timecode) },
  430. { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
  431. { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster,blocks), {.n=matroska_blockgroup} },
  432. { MATROSKA_ID_CLUSTERPOSITION,EBML_NONE },
  433. { MATROSKA_ID_CLUSTERPREVSIZE,EBML_NONE },
  434. { 0 }
  435. };
  436. static EbmlSyntax matroska_clusters[] = {
  437. { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, {.n=matroska_cluster} },
  438. { MATROSKA_ID_INFO, EBML_NONE },
  439. { MATROSKA_ID_CUES, EBML_NONE },
  440. { MATROSKA_ID_TAGS, EBML_NONE },
  441. { MATROSKA_ID_SEEKHEAD, EBML_NONE },
  442. { 0 }
  443. };
  444. static const char *matroska_doctypes[] = { "matroska", "webm" };
  445. /*
  446. * Return: Whether we reached the end of a level in the hierarchy or not.
  447. */
  448. static int ebml_level_end(MatroskaDemuxContext *matroska)
  449. {
  450. ByteIOContext *pb = matroska->ctx->pb;
  451. int64_t pos = url_ftell(pb);
  452. if (matroska->num_levels > 0) {
  453. MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
  454. if (pos - level->start >= level->length) {
  455. matroska->num_levels--;
  456. return 1;
  457. }
  458. }
  459. return 0;
  460. }
  461. /*
  462. * Read: an "EBML number", which is defined as a variable-length
  463. * array of bytes. The first byte indicates the length by giving a
  464. * number of 0-bits followed by a one. The position of the first
  465. * "one" bit inside the first byte indicates the length of this
  466. * number.
  467. * Returns: number of bytes read, < 0 on error
  468. */
  469. static int ebml_read_num(MatroskaDemuxContext *matroska, ByteIOContext *pb,
  470. int max_size, uint64_t *number)
  471. {
  472. int len_mask = 0x80, read = 1, n = 1;
  473. int64_t total = 0;
  474. /* The first byte tells us the length in bytes - get_byte() can normally
  475. * return 0, but since that's not a valid first ebmlID byte, we can
  476. * use it safely here to catch EOS. */
  477. if (!(total = get_byte(pb))) {
  478. /* we might encounter EOS here */
  479. if (!url_feof(pb)) {
  480. int64_t pos = url_ftell(pb);
  481. av_log(matroska->ctx, AV_LOG_ERROR,
  482. "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
  483. pos, pos);
  484. }
  485. return AVERROR(EIO); /* EOS or actual I/O error */
  486. }
  487. /* get the length of the EBML number */
  488. while (read <= max_size && !(total & len_mask)) {
  489. read++;
  490. len_mask >>= 1;
  491. }
  492. if (read > max_size) {
  493. int64_t pos = url_ftell(pb) - 1;
  494. av_log(matroska->ctx, AV_LOG_ERROR,
  495. "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
  496. (uint8_t) total, pos, pos);
  497. return AVERROR_INVALIDDATA;
  498. }
  499. /* read out length */
  500. total &= ~len_mask;
  501. while (n++ < read)
  502. total = (total << 8) | get_byte(pb);
  503. *number = total;
  504. return read;
  505. }
  506. /*
  507. * Read the next element as an unsigned int.
  508. * 0 is success, < 0 is failure.
  509. */
  510. static int ebml_read_uint(ByteIOContext *pb, int size, uint64_t *num)
  511. {
  512. int n = 0;
  513. if (size < 1 || size > 8)
  514. return AVERROR_INVALIDDATA;
  515. /* big-endian ordering; build up number */
  516. *num = 0;
  517. while (n++ < size)
  518. *num = (*num << 8) | get_byte(pb);
  519. return 0;
  520. }
  521. /*
  522. * Read the next element as a float.
  523. * 0 is success, < 0 is failure.
  524. */
  525. static int ebml_read_float(ByteIOContext *pb, int size, double *num)
  526. {
  527. if (size == 4) {
  528. *num= av_int2flt(get_be32(pb));
  529. } else if(size==8){
  530. *num= av_int2dbl(get_be64(pb));
  531. } else
  532. return AVERROR_INVALIDDATA;
  533. return 0;
  534. }
  535. /*
  536. * Read the next element as an ASCII string.
  537. * 0 is success, < 0 is failure.
  538. */
  539. static int ebml_read_ascii(ByteIOContext *pb, int size, char **str)
  540. {
  541. av_free(*str);
  542. /* EBML strings are usually not 0-terminated, so we allocate one
  543. * byte more, read the string and NULL-terminate it ourselves. */
  544. if (!(*str = av_malloc(size + 1)))
  545. return AVERROR(ENOMEM);
  546. if (get_buffer(pb, (uint8_t *) *str, size) != size) {
  547. av_freep(str);
  548. return AVERROR(EIO);
  549. }
  550. (*str)[size] = '\0';
  551. return 0;
  552. }
  553. /*
  554. * Read the next element as binary data.
  555. * 0 is success, < 0 is failure.
  556. */
  557. static int ebml_read_binary(ByteIOContext *pb, int length, EbmlBin *bin)
  558. {
  559. av_free(bin->data);
  560. if (!(bin->data = av_malloc(length)))
  561. return AVERROR(ENOMEM);
  562. bin->size = length;
  563. bin->pos = url_ftell(pb);
  564. if (get_buffer(pb, bin->data, length) != length) {
  565. av_freep(&bin->data);
  566. return AVERROR(EIO);
  567. }
  568. return 0;
  569. }
  570. /*
  571. * Read the next element, but only the header. The contents
  572. * are supposed to be sub-elements which can be read separately.
  573. * 0 is success, < 0 is failure.
  574. */
  575. static int ebml_read_master(MatroskaDemuxContext *matroska, int length)
  576. {
  577. ByteIOContext *pb = matroska->ctx->pb;
  578. MatroskaLevel *level;
  579. if (matroska->num_levels >= EBML_MAX_DEPTH) {
  580. av_log(matroska->ctx, AV_LOG_ERROR,
  581. "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
  582. return AVERROR(ENOSYS);
  583. }
  584. level = &matroska->levels[matroska->num_levels++];
  585. level->start = url_ftell(pb);
  586. level->length = length;
  587. return 0;
  588. }
  589. /*
  590. * Read signed/unsigned "EBML" numbers.
  591. * Return: number of bytes processed, < 0 on error
  592. */
  593. static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
  594. uint8_t *data, uint32_t size, uint64_t *num)
  595. {
  596. ByteIOContext pb;
  597. init_put_byte(&pb, data, size, 0, NULL, NULL, NULL, NULL);
  598. return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
  599. }
  600. /*
  601. * Same as above, but signed.
  602. */
  603. static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
  604. uint8_t *data, uint32_t size, int64_t *num)
  605. {
  606. uint64_t unum;
  607. int res;
  608. /* read as unsigned number first */
  609. if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
  610. return res;
  611. /* make signed (weird way) */
  612. *num = unum - ((1LL << (7*res - 1)) - 1);
  613. return res;
  614. }
  615. static int ebml_parse_elem(MatroskaDemuxContext *matroska,
  616. EbmlSyntax *syntax, void *data);
  617. static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
  618. uint32_t id, void *data)
  619. {
  620. int i;
  621. for (i=0; syntax[i].id; i++)
  622. if (id == syntax[i].id)
  623. break;
  624. if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32)
  625. av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%X\n", id);
  626. return ebml_parse_elem(matroska, &syntax[i], data);
  627. }
  628. static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
  629. void *data)
  630. {
  631. uint64_t id;
  632. int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
  633. id |= 1 << 7*res;
  634. return res < 0 ? res : ebml_parse_id(matroska, syntax, id, data);
  635. }
  636. static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
  637. void *data)
  638. {
  639. int i, res = 0;
  640. for (i=0; syntax[i].id; i++)
  641. switch (syntax[i].type) {
  642. case EBML_UINT:
  643. *(uint64_t *)((char *)data+syntax[i].data_offset) = syntax[i].def.u;
  644. break;
  645. case EBML_FLOAT:
  646. *(double *)((char *)data+syntax[i].data_offset) = syntax[i].def.f;
  647. break;
  648. case EBML_STR:
  649. case EBML_UTF8:
  650. *(char **)((char *)data+syntax[i].data_offset) = av_strdup(syntax[i].def.s);
  651. break;
  652. }
  653. while (!res && !ebml_level_end(matroska))
  654. res = ebml_parse(matroska, syntax, data);
  655. return res;
  656. }
  657. static int ebml_parse_elem(MatroskaDemuxContext *matroska,
  658. EbmlSyntax *syntax, void *data)
  659. {
  660. ByteIOContext *pb = matroska->ctx->pb;
  661. uint32_t id = syntax->id;
  662. uint64_t length;
  663. int res;
  664. data = (char *)data + syntax->data_offset;
  665. if (syntax->list_elem_size) {
  666. EbmlList *list = data;
  667. list->elem = av_realloc(list->elem, (list->nb_elem+1)*syntax->list_elem_size);
  668. data = (char*)list->elem + list->nb_elem*syntax->list_elem_size;
  669. memset(data, 0, syntax->list_elem_size);
  670. list->nb_elem++;
  671. }
  672. if (syntax->type != EBML_PASS && syntax->type != EBML_STOP)
  673. if ((res = ebml_read_num(matroska, pb, 8, &length)) < 0)
  674. return res;
  675. switch (syntax->type) {
  676. case EBML_UINT: res = ebml_read_uint (pb, length, data); break;
  677. case EBML_FLOAT: res = ebml_read_float (pb, length, data); break;
  678. case EBML_STR:
  679. case EBML_UTF8: res = ebml_read_ascii (pb, length, data); break;
  680. case EBML_BIN: res = ebml_read_binary(pb, length, data); break;
  681. case EBML_NEST: if ((res=ebml_read_master(matroska, length)) < 0)
  682. return res;
  683. if (id == MATROSKA_ID_SEGMENT)
  684. matroska->segment_start = url_ftell(matroska->ctx->pb);
  685. return ebml_parse_nest(matroska, syntax->def.n, data);
  686. case EBML_PASS: return ebml_parse_id(matroska, syntax->def.n, id, data);
  687. case EBML_STOP: *(int *)data = 1; return 1;
  688. default: return url_fseek(pb,length,SEEK_CUR)<0 ? AVERROR(EIO) : 0;
  689. }
  690. if (res == AVERROR_INVALIDDATA)
  691. av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
  692. else if (res == AVERROR(EIO))
  693. av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
  694. return res;
  695. }
  696. static void ebml_free(EbmlSyntax *syntax, void *data)
  697. {
  698. int i, j;
  699. for (i=0; syntax[i].id; i++) {
  700. void *data_off = (char *)data + syntax[i].data_offset;
  701. switch (syntax[i].type) {
  702. case EBML_STR:
  703. case EBML_UTF8: av_freep(data_off); break;
  704. case EBML_BIN: av_freep(&((EbmlBin *)data_off)->data); break;
  705. case EBML_NEST:
  706. if (syntax[i].list_elem_size) {
  707. EbmlList *list = data_off;
  708. char *ptr = list->elem;
  709. for (j=0; j<list->nb_elem; j++, ptr+=syntax[i].list_elem_size)
  710. ebml_free(syntax[i].def.n, ptr);
  711. av_free(list->elem);
  712. } else
  713. ebml_free(syntax[i].def.n, data_off);
  714. default: break;
  715. }
  716. }
  717. }
  718. /*
  719. * Autodetecting...
  720. */
  721. static int matroska_probe(AVProbeData *p)
  722. {
  723. uint64_t total = 0;
  724. int len_mask = 0x80, size = 1, n = 1, i;
  725. /* EBML header? */
  726. if (AV_RB32(p->buf) != EBML_ID_HEADER)
  727. return 0;
  728. /* length of header */
  729. total = p->buf[4];
  730. while (size <= 8 && !(total & len_mask)) {
  731. size++;
  732. len_mask >>= 1;
  733. }
  734. if (size > 8)
  735. return 0;
  736. total &= (len_mask - 1);
  737. while (n < size)
  738. total = (total << 8) | p->buf[4 + n++];
  739. /* Does the probe data contain the whole header? */
  740. if (p->buf_size < 4 + size + total)
  741. return 0;
  742. /* The header should contain a known document type. For now,
  743. * we don't parse the whole header but simply check for the
  744. * availability of that array of characters inside the header.
  745. * Not fully fool-proof, but good enough. */
  746. for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
  747. int probelen = strlen(matroska_doctypes[i]);
  748. for (n = 4+size; n <= 4+size+total-probelen; n++)
  749. if (!memcmp(p->buf+n, matroska_doctypes[i], probelen))
  750. return AVPROBE_SCORE_MAX;
  751. }
  752. // probably valid EBML header but no recognized doctype
  753. return AVPROBE_SCORE_MAX/2;
  754. }
  755. static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
  756. int num)
  757. {
  758. MatroskaTrack *tracks = matroska->tracks.elem;
  759. int i;
  760. for (i=0; i < matroska->tracks.nb_elem; i++)
  761. if (tracks[i].num == num)
  762. return &tracks[i];
  763. av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
  764. return NULL;
  765. }
  766. static int matroska_decode_buffer(uint8_t** buf, int* buf_size,
  767. MatroskaTrack *track)
  768. {
  769. MatroskaTrackEncoding *encodings = track->encodings.elem;
  770. uint8_t* data = *buf;
  771. int isize = *buf_size;
  772. uint8_t* pkt_data = NULL;
  773. int pkt_size = isize;
  774. int result = 0;
  775. int olen;
  776. switch (encodings[0].compression.algo) {
  777. case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
  778. return encodings[0].compression.settings.size;
  779. case MATROSKA_TRACK_ENCODING_COMP_LZO:
  780. do {
  781. olen = pkt_size *= 3;
  782. pkt_data = av_realloc(pkt_data, pkt_size+AV_LZO_OUTPUT_PADDING);
  783. result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
  784. } while (result==AV_LZO_OUTPUT_FULL && pkt_size<10000000);
  785. if (result)
  786. goto failed;
  787. pkt_size -= olen;
  788. break;
  789. #if CONFIG_ZLIB
  790. case MATROSKA_TRACK_ENCODING_COMP_ZLIB: {
  791. z_stream zstream = {0};
  792. if (inflateInit(&zstream) != Z_OK)
  793. return -1;
  794. zstream.next_in = data;
  795. zstream.avail_in = isize;
  796. do {
  797. pkt_size *= 3;
  798. pkt_data = av_realloc(pkt_data, pkt_size);
  799. zstream.avail_out = pkt_size - zstream.total_out;
  800. zstream.next_out = pkt_data + zstream.total_out;
  801. result = inflate(&zstream, Z_NO_FLUSH);
  802. } while (result==Z_OK && pkt_size<10000000);
  803. pkt_size = zstream.total_out;
  804. inflateEnd(&zstream);
  805. if (result != Z_STREAM_END)
  806. goto failed;
  807. break;
  808. }
  809. #endif
  810. #if CONFIG_BZLIB
  811. case MATROSKA_TRACK_ENCODING_COMP_BZLIB: {
  812. bz_stream bzstream = {0};
  813. if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
  814. return -1;
  815. bzstream.next_in = data;
  816. bzstream.avail_in = isize;
  817. do {
  818. pkt_size *= 3;
  819. pkt_data = av_realloc(pkt_data, pkt_size);
  820. bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
  821. bzstream.next_out = pkt_data + bzstream.total_out_lo32;
  822. result = BZ2_bzDecompress(&bzstream);
  823. } while (result==BZ_OK && pkt_size<10000000);
  824. pkt_size = bzstream.total_out_lo32;
  825. BZ2_bzDecompressEnd(&bzstream);
  826. if (result != BZ_STREAM_END)
  827. goto failed;
  828. break;
  829. }
  830. #endif
  831. default:
  832. return -1;
  833. }
  834. *buf = pkt_data;
  835. *buf_size = pkt_size;
  836. return 0;
  837. failed:
  838. av_free(pkt_data);
  839. return -1;
  840. }
  841. static void matroska_fix_ass_packet(MatroskaDemuxContext *matroska,
  842. AVPacket *pkt, uint64_t display_duration)
  843. {
  844. char *line, *layer, *ptr = pkt->data, *end = ptr+pkt->size;
  845. for (; *ptr!=',' && ptr<end-1; ptr++);
  846. if (*ptr == ',')
  847. layer = ++ptr;
  848. for (; *ptr!=',' && ptr<end-1; ptr++);
  849. if (*ptr == ',') {
  850. int64_t end_pts = pkt->pts + display_duration;
  851. int sc = matroska->time_scale * pkt->pts / 10000000;
  852. int ec = matroska->time_scale * end_pts / 10000000;
  853. int sh, sm, ss, eh, em, es, len;
  854. sh = sc/360000; sc -= 360000*sh;
  855. sm = sc/ 6000; sc -= 6000*sm;
  856. ss = sc/ 100; sc -= 100*ss;
  857. eh = ec/360000; ec -= 360000*eh;
  858. em = ec/ 6000; ec -= 6000*em;
  859. es = ec/ 100; ec -= 100*es;
  860. *ptr++ = '\0';
  861. len = 50 + end-ptr + FF_INPUT_BUFFER_PADDING_SIZE;
  862. if (!(line = av_malloc(len)))
  863. return;
  864. snprintf(line,len,"Dialogue: %s,%d:%02d:%02d.%02d,%d:%02d:%02d.%02d,%s\r\n",
  865. layer, sh, sm, ss, sc, eh, em, es, ec, ptr);
  866. av_free(pkt->data);
  867. pkt->data = line;
  868. pkt->size = strlen(line);
  869. }
  870. }
  871. static void matroska_merge_packets(AVPacket *out, AVPacket *in)
  872. {
  873. out->data = av_realloc(out->data, out->size+in->size);
  874. memcpy(out->data+out->size, in->data, in->size);
  875. out->size += in->size;
  876. av_destruct_packet(in);
  877. av_free(in);
  878. }
  879. static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
  880. AVMetadata **metadata, char *prefix)
  881. {
  882. MatroskaTag *tags = list->elem;
  883. char key[1024];
  884. int i;
  885. for (i=0; i < list->nb_elem; i++) {
  886. const char *lang = strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
  887. if (prefix) snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
  888. else av_strlcpy(key, tags[i].name, sizeof(key));
  889. if (tags[i].def || !lang) {
  890. av_metadata_set2(metadata, key, tags[i].string, 0);
  891. if (tags[i].sub.nb_elem)
  892. matroska_convert_tag(s, &tags[i].sub, metadata, key);
  893. }
  894. if (lang) {
  895. av_strlcat(key, "-", sizeof(key));
  896. av_strlcat(key, lang, sizeof(key));
  897. av_metadata_set2(metadata, key, tags[i].string, 0);
  898. if (tags[i].sub.nb_elem)
  899. matroska_convert_tag(s, &tags[i].sub, metadata, key);
  900. }
  901. }
  902. }
  903. static void matroska_convert_tags(AVFormatContext *s)
  904. {
  905. MatroskaDemuxContext *matroska = s->priv_data;
  906. MatroskaTags *tags = matroska->tags.elem;
  907. int i, j;
  908. for (i=0; i < matroska->tags.nb_elem; i++) {
  909. if (tags[i].target.attachuid) {
  910. MatroskaAttachement *attachment = matroska->attachments.elem;
  911. for (j=0; j<matroska->attachments.nb_elem; j++)
  912. if (attachment[j].uid == tags[i].target.attachuid)
  913. matroska_convert_tag(s, &tags[i].tag,
  914. &attachment[j].stream->metadata, NULL);
  915. } else if (tags[i].target.chapteruid) {
  916. MatroskaChapter *chapter = matroska->chapters.elem;
  917. for (j=0; j<matroska->chapters.nb_elem; j++)
  918. if (chapter[j].uid == tags[i].target.chapteruid)
  919. matroska_convert_tag(s, &tags[i].tag,
  920. &chapter[j].chapter->metadata, NULL);
  921. } else if (tags[i].target.trackuid) {
  922. MatroskaTrack *track = matroska->tracks.elem;
  923. for (j=0; j<matroska->tracks.nb_elem; j++)
  924. if (track[j].uid == tags[i].target.trackuid)
  925. matroska_convert_tag(s, &tags[i].tag,
  926. &track[j].stream->metadata, NULL);
  927. } else {
  928. matroska_convert_tag(s, &tags[i].tag, &s->metadata,
  929. tags[i].target.type);
  930. }
  931. }
  932. }
  933. static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
  934. {
  935. EbmlList *seekhead_list = &matroska->seekhead;
  936. MatroskaSeekhead *seekhead = seekhead_list->elem;
  937. uint32_t level_up = matroska->level_up;
  938. int64_t before_pos = url_ftell(matroska->ctx->pb);
  939. MatroskaLevel level;
  940. int i;
  941. // we should not do any seeking in the streaming case
  942. if (url_is_streamed(matroska->ctx->pb) ||
  943. (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
  944. return;
  945. for (i=0; i<seekhead_list->nb_elem; i++) {
  946. int64_t offset = seekhead[i].pos + matroska->segment_start;
  947. if (seekhead[i].pos <= before_pos
  948. || seekhead[i].id == MATROSKA_ID_SEEKHEAD
  949. || seekhead[i].id == MATROSKA_ID_CLUSTER)
  950. continue;
  951. /* seek */
  952. if (url_fseek(matroska->ctx->pb, offset, SEEK_SET) != offset)
  953. continue;
  954. /* We don't want to lose our seekhead level, so we add
  955. * a dummy. This is a crude hack. */
  956. if (matroska->num_levels == EBML_MAX_DEPTH) {
  957. av_log(matroska->ctx, AV_LOG_INFO,
  958. "Max EBML element depth (%d) reached, "
  959. "cannot parse further.\n", EBML_MAX_DEPTH);
  960. break;
  961. }
  962. level.start = 0;
  963. level.length = (uint64_t)-1;
  964. matroska->levels[matroska->num_levels] = level;
  965. matroska->num_levels++;
  966. ebml_parse(matroska, matroska_segment, matroska);
  967. /* remove dummy level */
  968. while (matroska->num_levels) {
  969. uint64_t length = matroska->levels[--matroska->num_levels].length;
  970. if (length == (uint64_t)-1)
  971. break;
  972. }
  973. }
  974. /* seek back */
  975. url_fseek(matroska->ctx->pb, before_pos, SEEK_SET);
  976. matroska->level_up = level_up;
  977. }
  978. static int matroska_aac_profile(char *codec_id)
  979. {
  980. static const char * const aac_profiles[] = { "MAIN", "LC", "SSR" };
  981. int profile;
  982. for (profile=0; profile<FF_ARRAY_ELEMS(aac_profiles); profile++)
  983. if (strstr(codec_id, aac_profiles[profile]))
  984. break;
  985. return profile + 1;
  986. }
  987. static int matroska_aac_sri(int samplerate)
  988. {
  989. int sri;
  990. for (sri=0; sri<FF_ARRAY_ELEMS(ff_mpeg4audio_sample_rates); sri++)
  991. if (ff_mpeg4audio_sample_rates[sri] == samplerate)
  992. break;
  993. return sri;
  994. }
  995. static int matroska_read_header(AVFormatContext *s, AVFormatParameters *ap)
  996. {
  997. MatroskaDemuxContext *matroska = s->priv_data;
  998. EbmlList *attachements_list = &matroska->attachments;
  999. MatroskaAttachement *attachements;
  1000. EbmlList *chapters_list = &matroska->chapters;
  1001. MatroskaChapter *chapters;
  1002. MatroskaTrack *tracks;
  1003. EbmlList *index_list;
  1004. MatroskaIndex *index;
  1005. int index_scale = 1;
  1006. uint64_t max_start = 0;
  1007. Ebml ebml = { 0 };
  1008. AVStream *st;
  1009. int i, j;
  1010. matroska->ctx = s;
  1011. /* First read the EBML header. */
  1012. if (ebml_parse(matroska, ebml_syntax, &ebml)
  1013. || ebml.version > EBML_VERSION || ebml.max_size > sizeof(uint64_t)
  1014. || ebml.id_length > sizeof(uint32_t) || ebml.doctype_version > 2) {
  1015. av_log(matroska->ctx, AV_LOG_ERROR,
  1016. "EBML header using unsupported features\n"
  1017. "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
  1018. ebml.version, ebml.doctype, ebml.doctype_version);
  1019. ebml_free(ebml_syntax, &ebml);
  1020. return AVERROR_PATCHWELCOME;
  1021. }
  1022. for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
  1023. if (!strcmp(ebml.doctype, matroska_doctypes[i]))
  1024. break;
  1025. if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
  1026. av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
  1027. }
  1028. ebml_free(ebml_syntax, &ebml);
  1029. /* The next thing is a segment. */
  1030. if (ebml_parse(matroska, matroska_segments, matroska) < 0)
  1031. return -1;
  1032. matroska_execute_seekhead(matroska);
  1033. if (!matroska->time_scale)
  1034. matroska->time_scale = 1000000;
  1035. if (matroska->duration)
  1036. matroska->ctx->duration = matroska->duration * matroska->time_scale
  1037. * 1000 / AV_TIME_BASE;
  1038. av_metadata_set2(&s->metadata, "title", matroska->title, 0);
  1039. tracks = matroska->tracks.elem;
  1040. for (i=0; i < matroska->tracks.nb_elem; i++) {
  1041. MatroskaTrack *track = &tracks[i];
  1042. enum CodecID codec_id = CODEC_ID_NONE;
  1043. EbmlList *encodings_list = &tracks->encodings;
  1044. MatroskaTrackEncoding *encodings = encodings_list->elem;
  1045. uint8_t *extradata = NULL;
  1046. int extradata_size = 0;
  1047. int extradata_offset = 0;
  1048. ByteIOContext b;
  1049. /* Apply some sanity checks. */
  1050. if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
  1051. track->type != MATROSKA_TRACK_TYPE_AUDIO &&
  1052. track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
  1053. av_log(matroska->ctx, AV_LOG_INFO,
  1054. "Unknown or unsupported track type %"PRIu64"\n",
  1055. track->type);
  1056. continue;
  1057. }
  1058. if (track->codec_id == NULL)
  1059. continue;
  1060. if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
  1061. if (!track->default_duration)
  1062. track->default_duration = 1000000000/track->video.frame_rate;
  1063. if (!track->video.display_width)
  1064. track->video.display_width = track->video.pixel_width;
  1065. if (!track->video.display_height)
  1066. track->video.display_height = track->video.pixel_height;
  1067. } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
  1068. if (!track->audio.out_samplerate)
  1069. track->audio.out_samplerate = track->audio.samplerate;
  1070. }
  1071. if (encodings_list->nb_elem > 1) {
  1072. av_log(matroska->ctx, AV_LOG_ERROR,
  1073. "Multiple combined encodings no supported");
  1074. } else if (encodings_list->nb_elem == 1) {
  1075. if (encodings[0].type ||
  1076. (encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP &&
  1077. #if CONFIG_ZLIB
  1078. encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
  1079. #endif
  1080. #if CONFIG_BZLIB
  1081. encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
  1082. #endif
  1083. encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO)) {
  1084. encodings[0].scope = 0;
  1085. av_log(matroska->ctx, AV_LOG_ERROR,
  1086. "Unsupported encoding type");
  1087. } else if (track->codec_priv.size && encodings[0].scope&2) {
  1088. uint8_t *codec_priv = track->codec_priv.data;
  1089. int offset = matroska_decode_buffer(&track->codec_priv.data,
  1090. &track->codec_priv.size,
  1091. track);
  1092. if (offset < 0) {
  1093. track->codec_priv.data = NULL;
  1094. track->codec_priv.size = 0;
  1095. av_log(matroska->ctx, AV_LOG_ERROR,
  1096. "Failed to decode codec private data\n");
  1097. } else if (offset > 0) {
  1098. track->codec_priv.data = av_malloc(track->codec_priv.size + offset);
  1099. memcpy(track->codec_priv.data,
  1100. encodings[0].compression.settings.data, offset);
  1101. memcpy(track->codec_priv.data+offset, codec_priv,
  1102. track->codec_priv.size);
  1103. track->codec_priv.size += offset;
  1104. }
  1105. if (codec_priv != track->codec_priv.data)
  1106. av_free(codec_priv);
  1107. }
  1108. }
  1109. for(j=0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++){
  1110. if(!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
  1111. strlen(ff_mkv_codec_tags[j].str))){
  1112. codec_id= ff_mkv_codec_tags[j].id;
  1113. break;
  1114. }
  1115. }
  1116. st = track->stream = av_new_stream(s, 0);
  1117. if (st == NULL)
  1118. return AVERROR(ENOMEM);
  1119. if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC")
  1120. && track->codec_priv.size >= 40
  1121. && track->codec_priv.data != NULL) {
  1122. track->ms_compat = 1;
  1123. track->video.fourcc = AV_RL32(track->codec_priv.data + 16);
  1124. codec_id = ff_codec_get_id(ff_codec_bmp_tags, track->video.fourcc);
  1125. extradata_offset = 40;
  1126. } else if (!strcmp(track->codec_id, "A_MS/ACM")
  1127. && track->codec_priv.size >= 14
  1128. && track->codec_priv.data != NULL) {
  1129. init_put_byte(&b, track->codec_priv.data, track->codec_priv.size,
  1130. URL_RDONLY, NULL, NULL, NULL, NULL);
  1131. ff_get_wav_header(&b, st->codec, track->codec_priv.size);
  1132. codec_id = st->codec->codec_id;
  1133. extradata_offset = FFMIN(track->codec_priv.size, 18);
  1134. } else if (!strcmp(track->codec_id, "V_QUICKTIME")
  1135. && (track->codec_priv.size >= 86)
  1136. && (track->codec_priv.data != NULL)) {
  1137. track->video.fourcc = AV_RL32(track->codec_priv.data);
  1138. codec_id=ff_codec_get_id(codec_movvideo_tags, track->video.fourcc);
  1139. } else if (codec_id == CODEC_ID_PCM_S16BE) {
  1140. switch (track->audio.bitdepth) {
  1141. case 8: codec_id = CODEC_ID_PCM_U8; break;
  1142. case 24: codec_id = CODEC_ID_PCM_S24BE; break;
  1143. case 32: codec_id = CODEC_ID_PCM_S32BE; break;
  1144. }
  1145. } else if (codec_id == CODEC_ID_PCM_S16LE) {
  1146. switch (track->audio.bitdepth) {
  1147. case 8: codec_id = CODEC_ID_PCM_U8; break;
  1148. case 24: codec_id = CODEC_ID_PCM_S24LE; break;
  1149. case 32: codec_id = CODEC_ID_PCM_S32LE; break;
  1150. }
  1151. } else if (codec_id==CODEC_ID_PCM_F32LE && track->audio.bitdepth==64) {
  1152. codec_id = CODEC_ID_PCM_F64LE;
  1153. } else if (codec_id == CODEC_ID_AAC && !track->codec_priv.size) {
  1154. int profile = matroska_aac_profile(track->codec_id);
  1155. int sri = matroska_aac_sri(track->audio.samplerate);
  1156. extradata = av_malloc(5);
  1157. if (extradata == NULL)
  1158. return AVERROR(ENOMEM);
  1159. extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
  1160. extradata[1] = ((sri&0x01) << 7) | (track->audio.channels<<3);
  1161. if (strstr(track->codec_id, "SBR")) {
  1162. sri = matroska_aac_sri(track->audio.out_samplerate);
  1163. extradata[2] = 0x56;
  1164. extradata[3] = 0xE5;
  1165. extradata[4] = 0x80 | (sri<<3);
  1166. extradata_size = 5;
  1167. } else
  1168. extradata_size = 2;
  1169. } else if (codec_id == CODEC_ID_TTA) {
  1170. extradata_size = 30;
  1171. extradata = av_mallocz(extradata_size);
  1172. if (extradata == NULL)
  1173. return AVERROR(ENOMEM);
  1174. init_put_byte(&b, extradata, extradata_size, 1,
  1175. NULL, NULL, NULL, NULL);
  1176. put_buffer(&b, "TTA1", 4);
  1177. put_le16(&b, 1);
  1178. put_le16(&b, track->audio.channels);
  1179. put_le16(&b, track->audio.bitdepth);
  1180. put_le32(&b, track->audio.out_samplerate);
  1181. put_le32(&b, matroska->ctx->duration * track->audio.out_samplerate);
  1182. } else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
  1183. codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
  1184. extradata_offset = 26;
  1185. } else if (codec_id == CODEC_ID_RA_144) {
  1186. track->audio.out_samplerate = 8000;
  1187. track->audio.channels = 1;
  1188. } else if (codec_id == CODEC_ID_RA_288 || codec_id == CODEC_ID_COOK ||
  1189. codec_id == CODEC_ID_ATRAC3 || codec_id == CODEC_ID_SIPR) {
  1190. int flavor;
  1191. init_put_byte(&b, track->codec_priv.data,track->codec_priv.size,
  1192. 0, NULL, NULL, NULL, NULL);
  1193. url_fskip(&b, 22);
  1194. flavor = get_be16(&b);
  1195. track->audio.coded_framesize = get_be32(&b);
  1196. url_fskip(&b, 12);
  1197. track->audio.sub_packet_h = get_be16(&b);
  1198. track->audio.frame_size = get_be16(&b);
  1199. track->audio.sub_packet_size = get_be16(&b);
  1200. track->audio.buf = av_malloc(track->audio.frame_size * track->audio.sub_packet_h);
  1201. if (codec_id == CODEC_ID_RA_288) {
  1202. st->codec->block_align = track->audio.coded_framesize;
  1203. track->codec_priv.size = 0;
  1204. } else {
  1205. if (codec_id == CODEC_ID_SIPR && flavor < 4) {
  1206. const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
  1207. track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
  1208. st->codec->bit_rate = sipr_bit_rate[flavor];
  1209. }
  1210. st->codec->block_align = track->audio.sub_packet_size;
  1211. extradata_offset = 78;
  1212. }
  1213. }
  1214. track->codec_priv.size -= extradata_offset;
  1215. if (codec_id == CODEC_ID_NONE)
  1216. av_log(matroska->ctx, AV_LOG_INFO,
  1217. "Unknown/unsupported CodecID %s.\n", track->codec_id);
  1218. if (track->time_scale < 0.01)
  1219. track->time_scale = 1.0;
  1220. av_set_pts_info(st, 64, matroska->time_scale*track->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
  1221. st->codec->codec_id = codec_id;
  1222. st->start_time = 0;
  1223. if (strcmp(track->language, "und"))
  1224. av_metadata_set2(&st->metadata, "language", track->language, 0);
  1225. av_metadata_set2(&st->metadata, "title", track->name, 0);
  1226. if (track->flag_default)
  1227. st->disposition |= AV_DISPOSITION_DEFAULT;
  1228. if (track->default_duration)
  1229. av_reduce(&st->codec->time_base.num, &st->codec->time_base.den,
  1230. track->default_duration, 1000000000, 30000);
  1231. if (!st->codec->extradata) {
  1232. if(extradata){
  1233. st->codec->extradata = extradata;
  1234. st->codec->extradata_size = extradata_size;
  1235. } else if(track->codec_priv.data && track->codec_priv.size > 0){
  1236. st->codec->extradata = av_mallocz(track->codec_priv.size +
  1237. FF_INPUT_BUFFER_PADDING_SIZE);
  1238. if(st->codec->extradata == NULL)
  1239. return AVERROR(ENOMEM);
  1240. st->codec->extradata_size = track->codec_priv.size;
  1241. memcpy(st->codec->extradata,
  1242. track->codec_priv.data + extradata_offset,
  1243. track->codec_priv.size);
  1244. }
  1245. }
  1246. if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
  1247. st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
  1248. st->codec->codec_tag = track->video.fourcc;
  1249. st->codec->width = track->video.pixel_width;
  1250. st->codec->height = track->video.pixel_height;
  1251. av_reduce(&st->sample_aspect_ratio.num,
  1252. &st->sample_aspect_ratio.den,
  1253. st->codec->height * track->video.display_width,
  1254. st->codec-> width * track->video.display_height,
  1255. 255);
  1256. if (st->codec->codec_id != CODEC_ID_H264)
  1257. st->need_parsing = AVSTREAM_PARSE_HEADERS;
  1258. if (track->default_duration)
  1259. st->avg_frame_rate = av_d2q(1000000000.0/track->default_duration, INT_MAX);
  1260. } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
  1261. st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
  1262. st->codec->sample_rate = track->audio.out_samplerate;
  1263. st->codec->channels = track->audio.channels;
  1264. if (st->codec->codec_id != CODEC_ID_AAC)
  1265. st->need_parsing = AVSTREAM_PARSE_HEADERS;
  1266. } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
  1267. st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
  1268. }
  1269. }
  1270. attachements = attachements_list->elem;
  1271. for (j=0; j<attachements_list->nb_elem; j++) {
  1272. if (!(attachements[j].filename && attachements[j].mime &&
  1273. attachements[j].bin.data && attachements[j].bin.size > 0)) {
  1274. av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
  1275. } else {
  1276. AVStream *st = av_new_stream(s, 0);
  1277. if (st == NULL)
  1278. break;
  1279. av_metadata_set2(&st->metadata, "filename",attachements[j].filename, 0);
  1280. st->codec->codec_id = CODEC_ID_NONE;
  1281. st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
  1282. st->codec->extradata = av_malloc(attachements[j].bin.size);
  1283. if(st->codec->extradata == NULL)
  1284. break;
  1285. st->codec->extradata_size = attachements[j].bin.size;
  1286. memcpy(st->codec->extradata, attachements[j].bin.data, attachements[j].bin.size);
  1287. for (i=0; ff_mkv_mime_tags[i].id != CODEC_ID_NONE; i++) {
  1288. if (!strncmp(ff_mkv_mime_tags[i].str, attachements[j].mime,
  1289. strlen(ff_mkv_mime_tags[i].str))) {
  1290. st->codec->codec_id = ff_mkv_mime_tags[i].id;
  1291. break;
  1292. }
  1293. }
  1294. attachements[j].stream = st;
  1295. }
  1296. }
  1297. chapters = chapters_list->elem;
  1298. for (i=0; i<chapters_list->nb_elem; i++)
  1299. if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid
  1300. && (max_start==0 || chapters[i].start > max_start)) {
  1301. chapters[i].chapter =
  1302. ff_new_chapter(s, chapters[i].uid, (AVRational){1, 1000000000},
  1303. chapters[i].start, chapters[i].end,
  1304. chapters[i].title);
  1305. av_metadata_set2(&chapters[i].chapter->metadata,
  1306. "title", chapters[i].title, 0);
  1307. max_start = chapters[i].start;
  1308. }
  1309. index_list = &matroska->index;
  1310. index = index_list->elem;
  1311. if (index_list->nb_elem
  1312. && index[0].time > 100000000000000/matroska->time_scale) {
  1313. av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
  1314. index_scale = matroska->time_scale;
  1315. }
  1316. for (i=0; i<index_list->nb_elem; i++) {
  1317. EbmlList *pos_list = &index[i].pos;
  1318. MatroskaIndexPos *pos = pos_list->elem;
  1319. for (j=0; j<pos_list->nb_elem; j++) {
  1320. MatroskaTrack *track = matroska_find_track_by_num(matroska,
  1321. pos[j].track);
  1322. if (track && track->stream)
  1323. av_add_index_entry(track->stream,
  1324. pos[j].pos + matroska->segment_start,
  1325. index[i].time/index_scale, 0, 0,
  1326. AVINDEX_KEYFRAME);
  1327. }
  1328. }
  1329. matroska_convert_tags(s);
  1330. return 0;
  1331. }
  1332. /*
  1333. * Put one packet in an application-supplied AVPacket struct.
  1334. * Returns 0 on success or -1 on failure.
  1335. */
  1336. static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
  1337. AVPacket *pkt)
  1338. {
  1339. if (matroska->num_packets > 0) {
  1340. memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
  1341. av_free(matroska->packets[0]);
  1342. if (matroska->num_packets > 1) {
  1343. memmove(&matroska->packets[0], &matroska->packets[1],
  1344. (matroska->num_packets - 1) * sizeof(AVPacket *));
  1345. matroska->packets =
  1346. av_realloc(matroska->packets, (matroska->num_packets - 1) *
  1347. sizeof(AVPacket *));
  1348. } else {
  1349. av_freep(&matroska->packets);
  1350. }
  1351. matroska->num_packets--;
  1352. return 0;
  1353. }
  1354. return -1;
  1355. }
  1356. /*
  1357. * Free all packets in our internal queue.
  1358. */
  1359. static void matroska_clear_queue(MatroskaDemuxContext *matroska)
  1360. {
  1361. if (matroska->packets) {
  1362. int n;
  1363. for (n = 0; n < matroska->num_packets; n++) {
  1364. av_free_packet(matroska->packets[n]);
  1365. av_free(matroska->packets[n]);
  1366. }
  1367. av_freep(&matroska->packets);
  1368. matroska->num_packets = 0;
  1369. }
  1370. }
  1371. static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
  1372. int size, int64_t pos, uint64_t cluster_time,
  1373. uint64_t duration, int is_keyframe,
  1374. int64_t cluster_pos)
  1375. {
  1376. uint64_t timecode = AV_NOPTS_VALUE;
  1377. MatroskaTrack *track;
  1378. int res = 0;
  1379. AVStream *st;
  1380. AVPacket *pkt;
  1381. int16_t block_time;
  1382. uint32_t *lace_size = NULL;
  1383. int n, flags, laces = 0;
  1384. uint64_t num;
  1385. if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
  1386. av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
  1387. return res;
  1388. }
  1389. data += n;
  1390. size -= n;
  1391. track = matroska_find_track_by_num(matroska, num);
  1392. if (size <= 3 || !track || !track->stream) {
  1393. av_log(matroska->ctx, AV_LOG_INFO,
  1394. "Invalid stream %"PRIu64" or size %u\n", num, size);
  1395. return res;
  1396. }
  1397. st = track->stream;
  1398. if (st->discard >= AVDISCARD_ALL)
  1399. return res;
  1400. if (duration == AV_NOPTS_VALUE)
  1401. duration = track->default_duration / matroska->time_scale;
  1402. block_time = AV_RB16(data);
  1403. data += 2;
  1404. flags = *data++;
  1405. size -= 3;
  1406. if (is_keyframe == -1)
  1407. is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
  1408. if (cluster_time != (uint64_t)-1
  1409. && (block_time >= 0 || cluster_time >= -block_time)) {
  1410. timecode = cluster_time + block_time;
  1411. if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE
  1412. && timecode < track->end_timecode)
  1413. is_keyframe = 0; /* overlapping subtitles are not key frame */
  1414. if (is_keyframe)
  1415. av_add_index_entry(st, cluster_pos, timecode, 0,0,AVINDEX_KEYFRAME);
  1416. track->end_timecode = FFMAX(track->end_timecode, timecode+duration);
  1417. }
  1418. if (matroska->skip_to_keyframe && track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
  1419. if (!is_keyframe || timecode < matroska->skip_to_timecode)
  1420. return res;
  1421. matroska->skip_to_keyframe = 0;
  1422. }
  1423. switch ((flags & 0x06) >> 1) {
  1424. case 0x0: /* no lacing */
  1425. laces = 1;
  1426. lace_size = av_mallocz(sizeof(int));
  1427. lace_size[0] = size;
  1428. break;
  1429. case 0x1: /* Xiph lacing */
  1430. case 0x2: /* fixed-size lacing */
  1431. case 0x3: /* EBML lacing */
  1432. assert(size>0); // size <=3 is checked before size-=3 above
  1433. laces = (*data) + 1;
  1434. data += 1;
  1435. size -= 1;
  1436. lace_size = av_mallocz(laces * sizeof(int));
  1437. switch ((flags & 0x06) >> 1) {
  1438. case 0x1: /* Xiph lacing */ {
  1439. uint8_t temp;
  1440. uint32_t total = 0;
  1441. for (n = 0; res == 0 && n < laces - 1; n++) {
  1442. while (1) {
  1443. if (size == 0) {
  1444. res = -1;
  1445. break;
  1446. }
  1447. temp = *data;
  1448. lace_size[n] += temp;
  1449. data += 1;
  1450. size -= 1;
  1451. if (temp != 0xff)
  1452. break;
  1453. }
  1454. total += lace_size[n];
  1455. }
  1456. lace_size[n] = size - total;
  1457. break;
  1458. }
  1459. case 0x2: /* fixed-size lacing */
  1460. for (n = 0; n < laces; n++)
  1461. lace_size[n] = size / laces;
  1462. break;
  1463. case 0x3: /* EBML lacing */ {
  1464. uint32_t total;
  1465. n = matroska_ebmlnum_uint(matroska, data, size, &num);
  1466. if (n < 0) {
  1467. av_log(matroska->ctx, AV_LOG_INFO,
  1468. "EBML block data error\n");
  1469. break;
  1470. }
  1471. data += n;
  1472. size -= n;
  1473. total = lace_size[0] = num;
  1474. for (n = 1; res == 0 && n < laces - 1; n++) {
  1475. int64_t snum;
  1476. int r;
  1477. r = matroska_ebmlnum_sint(matroska, data, size, &snum);
  1478. if (r < 0) {
  1479. av_log(matroska->ctx, AV_LOG_INFO,
  1480. "EBML block data error\n");
  1481. break;
  1482. }
  1483. data += r;
  1484. size -= r;
  1485. lace_size[n] = lace_size[n - 1] + snum;
  1486. total += lace_size[n];
  1487. }
  1488. lace_size[n] = size - total;
  1489. break;
  1490. }
  1491. }
  1492. break;
  1493. }
  1494. if (res == 0) {
  1495. for (n = 0; n < laces; n++) {
  1496. if ((st->codec->codec_id == CODEC_ID_RA_288 ||
  1497. st->codec->codec_id == CODEC_ID_COOK ||
  1498. st->codec->codec_id == CODEC_ID_SIPR ||
  1499. st->codec->codec_id == CODEC_ID_ATRAC3) &&
  1500. st->codec->block_align && track->audio.sub_packet_size) {
  1501. int a = st->codec->block_align;
  1502. int sps = track->audio.sub_packet_size;
  1503. int cfs = track->audio.coded_framesize;
  1504. int h = track->audio.sub_packet_h;
  1505. int y = track->audio.sub_packet_cnt;
  1506. int w = track->audio.frame_size;
  1507. int x;
  1508. if (!track->audio.pkt_cnt) {
  1509. if (st->codec->codec_id == CODEC_ID_RA_288)
  1510. for (x=0; x<h/2; x++)
  1511. memcpy(track->audio.buf+x*2*w+y*cfs,
  1512. data+x*cfs, cfs);
  1513. else if (st->codec->codec_id == CODEC_ID_SIPR)
  1514. memcpy(track->audio.buf + y*w, data, w);
  1515. else
  1516. for (x=0; x<w/sps; x++)
  1517. memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
  1518. if (++track->audio.sub_packet_cnt >= h) {
  1519. if (st->codec->codec_id == CODEC_ID_SIPR)
  1520. ff_rm_reorder_sipr_data(track->audio.buf, h, w);
  1521. track->audio.sub_packet_cnt = 0;
  1522. track->audio.pkt_cnt = h*w / a;
  1523. }
  1524. }
  1525. while (track->audio.pkt_cnt) {
  1526. pkt = av_mallocz(sizeof(AVPacket));
  1527. av_new_packet(pkt, a);
  1528. memcpy(pkt->data, track->audio.buf
  1529. + a * (h*w / a - track->audio.pkt_cnt--), a);
  1530. pkt->pos = pos;
  1531. pkt->stream_index = st->index;
  1532. dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
  1533. }
  1534. } else {
  1535. MatroskaTrackEncoding *encodings = track->encodings.elem;
  1536. int offset = 0, pkt_size = lace_size[n];
  1537. uint8_t *pkt_data = data;
  1538. if (lace_size[n] > size) {
  1539. av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
  1540. break;
  1541. }
  1542. if (encodings && encodings->scope & 1) {
  1543. offset = matroska_decode_buffer(&pkt_data,&pkt_size, track);
  1544. if (offset < 0)
  1545. continue;
  1546. }
  1547. pkt = av_mallocz(sizeof(AVPacket));
  1548. /* XXX: prevent data copy... */
  1549. if (av_new_packet(pkt, pkt_size+offset) < 0) {
  1550. av_free(pkt);
  1551. res = AVERROR(ENOMEM);
  1552. break;
  1553. }
  1554. if (offset)
  1555. memcpy (pkt->data, encodings->compression.settings.data, offset);
  1556. memcpy (pkt->data+offset, pkt_data, pkt_size);
  1557. if (pkt_data != data)
  1558. av_free(pkt_data);
  1559. if (n == 0)
  1560. pkt->flags = is_keyframe;
  1561. pkt->stream_index = st->index;
  1562. if (track->ms_compat)
  1563. pkt->dts = timecode;
  1564. else
  1565. pkt->pts = timecode;
  1566. pkt->pos = pos;
  1567. if (st->codec->codec_id == CODEC_ID_TEXT)
  1568. pkt->convergence_duration = duration;
  1569. else if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE)
  1570. pkt->duration = duration;
  1571. if (st->codec->codec_id == CODEC_ID_SSA)
  1572. matroska_fix_ass_packet(matroska, pkt, duration);
  1573. if (matroska->prev_pkt &&
  1574. timecode != AV_NOPTS_VALUE &&
  1575. matroska->prev_pkt->pts == timecode &&
  1576. matroska->prev_pkt->stream_index == st->index)
  1577. matroska_merge_packets(matroska->prev_pkt, pkt);
  1578. else {
  1579. dynarray_add(&matroska->packets,&matroska->num_packets,pkt);
  1580. matroska->prev_pkt = pkt;
  1581. }
  1582. }
  1583. if (timecode != AV_NOPTS_VALUE)
  1584. timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
  1585. data += lace_size[n];
  1586. size -= lace_size[n];
  1587. }
  1588. }
  1589. av_free(lace_size);
  1590. return res;
  1591. }
  1592. static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
  1593. {
  1594. MatroskaCluster cluster = { 0 };
  1595. EbmlList *blocks_list;
  1596. MatroskaBlock *blocks;
  1597. int i, res;
  1598. int64_t pos = url_ftell(matroska->ctx->pb);
  1599. matroska->prev_pkt = NULL;
  1600. if (matroska->has_cluster_id){
  1601. /* For the first cluster we parse, its ID was already read as
  1602. part of matroska_read_header(), so don't read it again */
  1603. res = ebml_parse_id(matroska, matroska_clusters,
  1604. MATROSKA_ID_CLUSTER, &cluster);
  1605. pos -= 4; /* sizeof the ID which was already read */
  1606. matroska->has_cluster_id = 0;
  1607. } else
  1608. res = ebml_parse(matroska, matroska_clusters, &cluster);
  1609. blocks_list = &cluster.blocks;
  1610. blocks = blocks_list->elem;
  1611. for (i=0; i<blocks_list->nb_elem; i++)
  1612. if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
  1613. int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
  1614. res=matroska_parse_block(matroska,
  1615. blocks[i].bin.data, blocks[i].bin.size,
  1616. blocks[i].bin.pos, cluster.timecode,
  1617. blocks[i].duration, is_keyframe,
  1618. pos);
  1619. }
  1620. ebml_free(matroska_cluster, &cluster);
  1621. if (res < 0) matroska->done = 1;
  1622. return res;
  1623. }
  1624. static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
  1625. {
  1626. MatroskaDemuxContext *matroska = s->priv_data;
  1627. while (matroska_deliver_packet(matroska, pkt)) {
  1628. if (matroska->done)
  1629. return AVERROR_EOF;
  1630. matroska_parse_cluster(matroska);
  1631. }
  1632. return 0;
  1633. }
  1634. static int matroska_read_seek(AVFormatContext *s, int stream_index,
  1635. int64_t timestamp, int flags)
  1636. {
  1637. MatroskaDemuxContext *matroska = s->priv_data;
  1638. MatroskaTrack *tracks = matroska->tracks.elem;
  1639. AVStream *st = s->streams[stream_index];
  1640. int i, index, index_sub, index_min;
  1641. if (!st->nb_index_entries)
  1642. return 0;
  1643. timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
  1644. if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
  1645. url_fseek(s->pb, st->index_entries[st->nb_index_entries-1].pos, SEEK_SET);
  1646. while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0) {
  1647. matroska_clear_queue(matroska);
  1648. if (matroska_parse_cluster(matroska) < 0)
  1649. break;
  1650. }
  1651. }
  1652. matroska_clear_queue(matroska);
  1653. if (index < 0)
  1654. return 0;
  1655. index_min = index;
  1656. for (i=0; i < matroska->tracks.nb_elem; i++) {
  1657. tracks[i].end_timecode = 0;
  1658. if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE
  1659. && !tracks[i].stream->discard != AVDISCARD_ALL) {
  1660. index_sub = av_index_search_timestamp(tracks[i].stream, st->index_entries[index].timestamp, AVSEEK_FLAG_BACKWARD);
  1661. if (index_sub >= 0
  1662. && st->index_entries[index_sub].pos < st->index_entries[index_min].pos
  1663. && st->index_entries[index].timestamp - st->index_entries[index_sub].timestamp < 30000000000/matroska->time_scale)
  1664. index_min = index_sub;
  1665. }
  1666. }
  1667. url_fseek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
  1668. matroska->skip_to_keyframe = !(flags & AVSEEK_FLAG_ANY);
  1669. matroska->skip_to_timecode = st->index_entries[index].timestamp;
  1670. matroska->done = 0;
  1671. av_update_cur_dts(s, st, st->index_entries[index].timestamp);
  1672. return 0;
  1673. }
  1674. static int matroska_read_close(AVFormatContext *s)
  1675. {
  1676. MatroskaDemuxContext *matroska = s->priv_data;
  1677. MatroskaTrack *tracks = matroska->tracks.elem;
  1678. int n;
  1679. matroska_clear_queue(matroska);
  1680. for (n=0; n < matroska->tracks.nb_elem; n++)
  1681. if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
  1682. av_free(tracks[n].audio.buf);
  1683. ebml_free(matroska_segment, matroska);
  1684. return 0;
  1685. }
  1686. AVInputFormat matroska_demuxer = {
  1687. "matroska",
  1688. NULL_IF_CONFIG_SMALL("Matroska file format"),
  1689. sizeof(MatroskaDemuxContext),
  1690. matroska_probe,
  1691. matroska_read_header,
  1692. matroska_read_packet,
  1693. matroska_read_close,
  1694. matroska_read_seek,
  1695. .metadata_conv = ff_mkv_metadata_conv,
  1696. };