2 * Matroska file demuxer (no muxer yet)
3 * Copyright (c) 2003-2004 The ffmpeg Project
5 * This file is part of FFmpeg.
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.
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.
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
24 * Matroska file demuxer
25 * by Ronald Bultje <rbultje@ronald.bitfreak.net>
26 * with a little help from Moritz Bunkus <moritz@bunkus.org>
27 * Specs available on the matroska project page:
28 * http://www.matroska.org/.
32 /* For codec_get_id(). */
34 #include "intfloat_readwrite.h"
36 /* EBML version supported */
37 #define EBML_VERSION 1
39 /* top-level master-IDs */
40 #define EBML_ID_HEADER 0x1A45DFA3
42 /* IDs in the HEADER master */
43 #define EBML_ID_EBMLVERSION 0x4286
44 #define EBML_ID_EBMLREADVERSION 0x42F7
45 #define EBML_ID_EBMLMAXIDLENGTH 0x42F2
46 #define EBML_ID_EBMLMAXSIZELENGTH 0x42F3
47 #define EBML_ID_DOCTYPE 0x4282
48 #define EBML_ID_DOCTYPEVERSION 0x4287
49 #define EBML_ID_DOCTYPEREADVERSION 0x4285
51 /* general EBML types */
52 #define EBML_ID_VOID 0xEC
55 * Matroska element IDs. max. 32-bit.
58 /* toplevel segment */
59 #define MATROSKA_ID_SEGMENT 0x18538067
61 /* matroska top-level master IDs */
62 #define MATROSKA_ID_INFO 0x1549A966
63 #define MATROSKA_ID_TRACKS 0x1654AE6B
64 #define MATROSKA_ID_CUES 0x1C53BB6B
65 #define MATROSKA_ID_TAGS 0x1254C367
66 #define MATROSKA_ID_SEEKHEAD 0x114D9B74
67 #define MATROSKA_ID_CLUSTER 0x1F43B675
69 /* IDs in the info master */
70 #define MATROSKA_ID_TIMECODESCALE 0x2AD7B1
71 #define MATROSKA_ID_DURATION 0x4489
72 #define MATROSKA_ID_TITLE 0x7BA9
73 #define MATROSKA_ID_WRITINGAPP 0x5741
74 #define MATROSKA_ID_MUXINGAPP 0x4D80
75 #define MATROSKA_ID_DATEUTC 0x4461
77 /* ID in the tracks master */
78 #define MATROSKA_ID_TRACKENTRY 0xAE
80 /* IDs in the trackentry master */
81 #define MATROSKA_ID_TRACKNUMBER 0xD7
82 #define MATROSKA_ID_TRACKUID 0x73C5
83 #define MATROSKA_ID_TRACKTYPE 0x83
84 #define MATROSKA_ID_TRACKAUDIO 0xE1
85 #define MATROSKA_ID_TRACKVIDEO 0xE0
86 #define MATROSKA_ID_CODECID 0x86
87 #define MATROSKA_ID_CODECPRIVATE 0x63A2
88 #define MATROSKA_ID_CODECNAME 0x258688
89 #define MATROSKA_ID_CODECINFOURL 0x3B4040
90 #define MATROSKA_ID_CODECDOWNLOADURL 0x26B240
91 #define MATROSKA_ID_TRACKNAME 0x536E
92 #define MATROSKA_ID_TRACKLANGUAGE 0x22B59C
93 #define MATROSKA_ID_TRACKFLAGENABLED 0xB9
94 #define MATROSKA_ID_TRACKFLAGDEFAULT 0x88
95 #define MATROSKA_ID_TRACKFLAGLACING 0x9C
96 #define MATROSKA_ID_TRACKMINCACHE 0x6DE7
97 #define MATROSKA_ID_TRACKMAXCACHE 0x6DF8
98 #define MATROSKA_ID_TRACKDEFAULTDURATION 0x23E383
100 /* IDs in the trackvideo master */
101 #define MATROSKA_ID_VIDEOFRAMERATE 0x2383E3
102 #define MATROSKA_ID_VIDEODISPLAYWIDTH 0x54B0
103 #define MATROSKA_ID_VIDEODISPLAYHEIGHT 0x54BA
104 #define MATROSKA_ID_VIDEOPIXELWIDTH 0xB0
105 #define MATROSKA_ID_VIDEOPIXELHEIGHT 0xBA
106 #define MATROSKA_ID_VIDEOFLAGINTERLACED 0x9A
107 #define MATROSKA_ID_VIDEOSTEREOMODE 0x53B9
108 #define MATROSKA_ID_VIDEOASPECTRATIO 0x54B3
109 #define MATROSKA_ID_VIDEOCOLOURSPACE 0x2EB524
111 /* IDs in the trackaudio master */
112 #define MATROSKA_ID_AUDIOSAMPLINGFREQ 0xB5
113 #define MATROSKA_ID_AUDIOOUTSAMPLINGFREQ 0x78B5
115 #define MATROSKA_ID_AUDIOBITDEPTH 0x6264
116 #define MATROSKA_ID_AUDIOCHANNELS 0x9F
118 /* ID in the cues master */
119 #define MATROSKA_ID_POINTENTRY 0xBB
121 /* IDs in the pointentry master */
122 #define MATROSKA_ID_CUETIME 0xB3
123 #define MATROSKA_ID_CUETRACKPOSITION 0xB7
125 /* IDs in the cuetrackposition master */
126 #define MATROSKA_ID_CUETRACK 0xF7
127 #define MATROSKA_ID_CUECLUSTERPOSITION 0xF1
129 /* IDs in the tags master */
132 /* IDs in the seekhead master */
133 #define MATROSKA_ID_SEEKENTRY 0x4DBB
135 /* IDs in the seekpoint master */
136 #define MATROSKA_ID_SEEKID 0x53AB
137 #define MATROSKA_ID_SEEKPOSITION 0x53AC
139 /* IDs in the cluster master */
140 #define MATROSKA_ID_CLUSTERTIMECODE 0xE7
141 #define MATROSKA_ID_BLOCKGROUP 0xA0
142 #define MATROSKA_ID_SIMPLEBLOCK 0xA3
144 /* IDs in the blockgroup master */
145 #define MATROSKA_ID_BLOCK 0xA1
146 #define MATROSKA_ID_BLOCKDURATION 0x9B
147 #define MATROSKA_ID_BLOCKREFERENCE 0xFB
150 MATROSKA_TRACK_TYPE_VIDEO = 0x1,
151 MATROSKA_TRACK_TYPE_AUDIO = 0x2,
152 MATROSKA_TRACK_TYPE_COMPLEX = 0x3,
153 MATROSKA_TRACK_TYPE_LOGO = 0x10,
154 MATROSKA_TRACK_TYPE_SUBTITLE = 0x11,
155 MATROSKA_TRACK_TYPE_CONTROL = 0x20,
159 MATROSKA_EYE_MODE_MONO = 0x0,
160 MATROSKA_EYE_MODE_RIGHT = 0x1,
161 MATROSKA_EYE_MODE_LEFT = 0x2,
162 MATROSKA_EYE_MODE_BOTH = 0x3,
166 MATROSKA_ASPECT_RATIO_MODE_FREE = 0x0,
167 MATROSKA_ASPECT_RATIO_MODE_KEEP = 0x1,
168 MATROSKA_ASPECT_RATIO_MODE_FIXED = 0x2,
169 } MatroskaAspectRatioMode;
172 * These aren't in any way "matroska-form" things,
173 * it's just something I use in the muxer/demuxer.
177 MATROSKA_TRACK_ENABLED = (1<<0),
178 MATROSKA_TRACK_DEFAULT = (1<<1),
179 MATROSKA_TRACK_LACING = (1<<2),
180 MATROSKA_TRACK_REAL_V = (1<<4),
181 MATROSKA_TRACK_REORDER = (1<<8),
182 MATROSKA_TRACK_SHIFT = (1<<16)
183 } MatroskaTrackFlags;
186 MATROSKA_VIDEOTRACK_INTERLACED = (MATROSKA_TRACK_SHIFT<<0)
187 } MatroskaVideoTrackFlags;
190 * Matroska Codec IDs. Strings.
193 typedef struct CodecTags{
198 #define MATROSKA_CODEC_ID_VIDEO_VFW_FOURCC "V_MS/VFW/FOURCC"
199 #define MATROSKA_CODEC_ID_AUDIO_ACM "A_MS/ACM"
201 static CodecTags codec_tags[]={
202 // {"V_MS/VFW/FOURCC" , CODEC_ID_NONE},
203 {"V_UNCOMPRESSED" , CODEC_ID_RAWVIDEO},
204 {"V_MPEG4/ISO/SP" , CODEC_ID_MPEG4},
205 {"V_MPEG4/ISO/ASP" , CODEC_ID_MPEG4},
206 {"V_MPEG4/ISO/AP" , CODEC_ID_MPEG4},
207 {"V_MPEG4/ISO/AVC" , CODEC_ID_H264},
208 {"V_MPEG4/MS/V3" , CODEC_ID_MSMPEG4V3},
209 {"V_MPEG1" , CODEC_ID_MPEG1VIDEO},
210 {"V_MPEG2" , CODEC_ID_MPEG2VIDEO},
211 {"V_MJPEG" , CODEC_ID_MJPEG},
212 {"V_REAL/RV10" , CODEC_ID_RV10},
213 {"V_REAL/RV20" , CODEC_ID_RV20},
214 {"V_REAL/RV30" , CODEC_ID_RV30},
215 {"V_REAL/RV40" , CODEC_ID_RV40},
216 {"V_THEORA" , CODEC_ID_THEORA},
217 /* TODO: Real/Quicktime */
219 // {"A_MS/ACM" , CODEC_ID_NONE},
220 {"A_MPEG/L1" , CODEC_ID_MP3},
221 {"A_MPEG/L2" , CODEC_ID_MP3},
222 {"A_MPEG/L3" , CODEC_ID_MP3},
223 {"A_PCM/INT/BIG" , CODEC_ID_PCM_U16BE},
224 {"A_PCM/INT/LIT" , CODEC_ID_PCM_U16LE},
225 // {"A_PCM/FLOAT/IEEE" , CODEC_ID_NONE},
226 {"A_AC3" , CODEC_ID_AC3},
227 {"A_DTS" , CODEC_ID_DTS},
228 {"A_VORBIS" , CODEC_ID_VORBIS},
229 {"A_AAC" , CODEC_ID_AAC},
230 {"A_FLAC" , CODEC_ID_FLAC},
231 {"A_WAVPACK4" , CODEC_ID_WAVPACK},
232 {"A_TTA1" , CODEC_ID_TTA},
233 {NULL , CODEC_ID_NONE}
234 /* TODO: AC3-9/10 (?), Real, Musepack, Quicktime */
237 /* max. depth in the EBML tree structure */
238 #define EBML_MAX_DEPTH 16
240 typedef struct Track {
241 MatroskaTrackType type;
243 /* Unique track number and track ID. stream_index is the index that
244 * the calling app uses for this track. */
255 unsigned char *codec_priv;
258 uint64_t default_duration;
259 MatroskaTrackFlags flags;
262 typedef struct MatroskaVideoTrack {
272 MatroskaAspectRatioMode ar_mode;
273 MatroskaEyeMode eye_mode;
276 } MatroskaVideoTrack;
278 typedef struct MatroskaAudioTrack {
286 } MatroskaAudioTrack;
288 typedef struct MatroskaSubtitleTrack {
292 } MatroskaSubtitleTrack;
294 #define MAX_TRACK_SIZE (FFMAX(FFMAX(sizeof(MatroskaVideoTrack), \
295 sizeof(MatroskaAudioTrack)), \
296 sizeof(MatroskaSubtitleTrack)))
298 typedef struct MatroskaLevel {
299 uint64_t start, length;
302 typedef struct MatroskaDemuxIndex {
303 uint64_t pos; /* of the corresponding *cluster*! */
304 uint16_t track; /* reference to 'num' */
305 uint64_t time; /* in nanoseconds */
306 } MatroskaDemuxIndex;
308 typedef struct MatroskaDemuxContext {
309 AVFormatContext *ctx;
313 MatroskaLevel levels[EBML_MAX_DEPTH];
321 /* timescale in the file */
324 /* position (time, ns) */
327 /* num_streams is the number of streams that av_new_stream() was called
328 * for ( = that are available to the calling program). */
329 int num_tracks, num_streams;
330 MatroskaTrack *tracks[MAX_STREAMS];
332 /* cache for ID peeking */
335 /* byte position of the segment inside the stream */
336 offset_t segment_start;
338 /* The packet queue. */
341 /* Second packet queue used to reorder pts of some video track. */
342 AVPacket **packets_reorder;
343 int num_packets_reorder;
344 uint64_t reorder_max_pts;
346 /* have we already parse metadata/cues/clusters? */
351 /* The index for seeking. */
353 MatroskaDemuxIndex *index;
354 } MatroskaDemuxContext;
357 * The first few functions handle EBML file parsing. The rest
358 * is the document interpretation. Matroska really just is a
363 * Return: the amount of levels in the hierarchy that the
364 * current element lies higher than the previous one.
365 * The opposite isn't done - that's auto-done using master
370 ebml_read_element_level_up (MatroskaDemuxContext *matroska)
372 ByteIOContext *pb = &matroska->ctx->pb;
373 offset_t pos = url_ftell(pb);
376 while (matroska->num_levels > 0) {
377 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
379 if (pos >= level->start + level->length) {
380 matroska->num_levels--;
391 * Read: an "EBML number", which is defined as a variable-length
392 * array of bytes. The first byte indicates the length by giving a
393 * number of 0-bits followed by a one. The position of the first
394 * "one" bit inside the first byte indicates the length of this
396 * Returns: num. of bytes read. < 0 on error.
400 ebml_read_num (MatroskaDemuxContext *matroska,
404 ByteIOContext *pb = &matroska->ctx->pb;
405 int len_mask = 0x80, read = 1, n = 1;
408 /* the first byte tells us the length in bytes - get_byte() can normally
409 * return 0, but since that's not a valid first ebmlID byte, we can
410 * use it safely here to catch EOS. */
411 if (!(total = get_byte(pb))) {
412 /* we might encounter EOS here */
414 offset_t pos = url_ftell(pb);
415 av_log(matroska->ctx, AV_LOG_ERROR,
416 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
419 return AVERROR_IO; /* EOS or actual I/O error */
422 /* get the length of the EBML number */
423 while (read <= max_size && !(total & len_mask)) {
427 if (read > max_size) {
428 offset_t pos = url_ftell(pb) - 1;
429 av_log(matroska->ctx, AV_LOG_ERROR,
430 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
431 (uint8_t) total, pos, pos);
432 return AVERROR_INVALIDDATA;
435 /* read out length */
438 total = (total << 8) | get_byte(pb);
446 * Read: the element content data ID.
447 * Return: the number of bytes read or < 0 on error.
451 ebml_read_element_id (MatroskaDemuxContext *matroska,
458 /* if we re-call this, use our cached ID */
459 if (matroska->peek_id != 0) {
462 *id = matroska->peek_id;
466 /* read out the "EBML number", include tag in ID */
467 if ((read = ebml_read_num(matroska, 4, &total)) < 0)
469 *id = matroska->peek_id = total | (1 << (read * 7));
473 *level_up = ebml_read_element_level_up(matroska);
479 * Read: element content length.
480 * Return: the number of bytes read or < 0 on error.
484 ebml_read_element_length (MatroskaDemuxContext *matroska,
487 /* clear cache since we're now beyond that data point */
488 matroska->peek_id = 0;
490 /* read out the "EBML number", include tag in ID */
491 return ebml_read_num(matroska, 8, length);
495 * Return: the ID of the next element, or 0 on error.
496 * Level_up contains the amount of levels that this
497 * next element lies higher than the previous one.
501 ebml_peek_id (MatroskaDemuxContext *matroska,
506 assert(level_up != NULL);
508 if (ebml_read_element_id(matroska, &id, level_up) < 0)
515 * Seek to a given offset.
516 * 0 is success, -1 is failure.
520 ebml_read_seek (MatroskaDemuxContext *matroska,
523 ByteIOContext *pb = &matroska->ctx->pb;
525 /* clear ID cache, if any */
526 matroska->peek_id = 0;
528 return (url_fseek(pb, offset, SEEK_SET) == offset) ? 0 : -1;
532 * Skip the next element.
533 * 0 is success, -1 is failure.
537 ebml_read_skip (MatroskaDemuxContext *matroska)
539 ByteIOContext *pb = &matroska->ctx->pb;
544 if ((res = ebml_read_element_id(matroska, &id, NULL)) < 0 ||
545 (res = ebml_read_element_length(matroska, &length)) < 0)
548 url_fskip(pb, length);
554 * Read the next element as an unsigned int.
555 * 0 is success, < 0 is failure.
559 ebml_read_uint (MatroskaDemuxContext *matroska,
563 ByteIOContext *pb = &matroska->ctx->pb;
564 int n = 0, size, res;
567 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
568 (res = ebml_read_element_length(matroska, &rlength)) < 0)
571 if (size < 1 || size > 8) {
572 offset_t pos = url_ftell(pb);
573 av_log(matroska->ctx, AV_LOG_ERROR,
574 "Invalid uint element size %d at position %"PRId64" (0x%"PRIx64")\n",
576 return AVERROR_INVALIDDATA;
579 /* big-endian ordening; build up number */
582 *num = (*num << 8) | get_byte(pb);
588 * Read the next element as a signed int.
589 * 0 is success, < 0 is failure.
593 ebml_read_sint (MatroskaDemuxContext *matroska,
597 ByteIOContext *pb = &matroska->ctx->pb;
598 int size, n = 1, negative = 0, res;
601 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
602 (res = ebml_read_element_length(matroska, &rlength)) < 0)
605 if (size < 1 || size > 8) {
606 offset_t pos = url_ftell(pb);
607 av_log(matroska->ctx, AV_LOG_ERROR,
608 "Invalid sint element size %d at position %"PRId64" (0x%"PRIx64")\n",
610 return AVERROR_INVALIDDATA;
612 if ((*num = get_byte(pb)) & 0x80) {
617 *num = (*num << 8) | get_byte(pb);
621 *num = *num - (1LL << ((8 * size) - 1));
627 * Read the next element as a float.
628 * 0 is success, < 0 is failure.
632 ebml_read_float (MatroskaDemuxContext *matroska,
636 ByteIOContext *pb = &matroska->ctx->pb;
640 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
641 (res = ebml_read_element_length(matroska, &rlength)) < 0)
646 *num= av_int2flt(get_be32(pb));
648 *num= av_int2dbl(get_be64(pb));
650 offset_t pos = url_ftell(pb);
651 av_log(matroska->ctx, AV_LOG_ERROR,
652 "Invalid float element size %d at position %"PRIu64" (0x%"PRIx64")\n",
654 return AVERROR_INVALIDDATA;
661 * Read the next element as an ASCII string.
662 * 0 is success, < 0 is failure.
666 ebml_read_ascii (MatroskaDemuxContext *matroska,
670 ByteIOContext *pb = &matroska->ctx->pb;
674 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
675 (res = ebml_read_element_length(matroska, &rlength)) < 0)
679 /* ebml strings are usually not 0-terminated, so we allocate one
680 * byte more, read the string and NULL-terminate it ourselves. */
681 if (size < 0 || !(*str = av_malloc(size + 1))) {
682 av_log(matroska->ctx, AV_LOG_ERROR, "Memory allocation failed\n");
683 return AVERROR_NOMEM;
685 if (get_buffer(pb, (uint8_t *) *str, size) != size) {
686 offset_t pos = url_ftell(pb);
687 av_log(matroska->ctx, AV_LOG_ERROR,
688 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
697 * Read the next element as a UTF-8 string.
698 * 0 is success, < 0 is failure.
702 ebml_read_utf8 (MatroskaDemuxContext *matroska,
706 return ebml_read_ascii(matroska, id, str);
710 * Read the next element as a date (nanoseconds since 1/1/2000).
711 * 0 is success, < 0 is failure.
715 ebml_read_date (MatroskaDemuxContext *matroska,
719 return ebml_read_sint(matroska, id, date);
723 * Read the next element, but only the header. The contents
724 * are supposed to be sub-elements which can be read separately.
725 * 0 is success, < 0 is failure.
729 ebml_read_master (MatroskaDemuxContext *matroska,
732 ByteIOContext *pb = &matroska->ctx->pb;
734 MatroskaLevel *level;
737 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
738 (res = ebml_read_element_length(matroska, &length)) < 0)
741 /* protect... (Heaven forbids that the '>' is true) */
742 if (matroska->num_levels >= EBML_MAX_DEPTH) {
743 av_log(matroska->ctx, AV_LOG_ERROR,
744 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
745 return AVERROR_NOTSUPP;
749 level = &matroska->levels[matroska->num_levels++];
750 level->start = url_ftell(pb);
751 level->length = length;
757 * Read the next element as binary data.
758 * 0 is success, < 0 is failure.
762 ebml_read_binary (MatroskaDemuxContext *matroska,
767 ByteIOContext *pb = &matroska->ctx->pb;
771 if ((res = ebml_read_element_id(matroska, id, NULL)) < 0 ||
772 (res = ebml_read_element_length(matroska, &rlength)) < 0)
776 if (!(*binary = av_malloc(*size))) {
777 av_log(matroska->ctx, AV_LOG_ERROR,
778 "Memory allocation error\n");
779 return AVERROR_NOMEM;
782 if (get_buffer(pb, *binary, *size) != *size) {
783 offset_t pos = url_ftell(pb);
784 av_log(matroska->ctx, AV_LOG_ERROR,
785 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
793 * Read signed/unsigned "EBML" numbers.
794 * Return: number of bytes processed, < 0 on error.
795 * XXX: use ebml_read_num().
799 matroska_ebmlnum_uint (uint8_t *data,
803 int len_mask = 0x80, read = 1, n = 1, num_ffs = 0;
807 return AVERROR_INVALIDDATA;
810 while (read <= 8 && !(total & len_mask)) {
815 return AVERROR_INVALIDDATA;
817 if ((total &= (len_mask - 1)) == len_mask - 1)
820 return AVERROR_INVALIDDATA;
824 total = (total << 8) | data[n];
837 * Same as above, but signed.
841 matroska_ebmlnum_sint (uint8_t *data,
848 /* read as unsigned number first */
849 if ((res = matroska_ebmlnum_uint(data, size, &unum)) < 0)
852 /* make signed (weird way) */
853 if (unum == (uint64_t)-1)
856 *num = unum - ((1LL << ((7 * res) - 1)) - 1);
862 * Read an EBML header.
863 * 0 is success, < 0 is failure.
867 ebml_read_header (MatroskaDemuxContext *matroska,
872 int level_up, res = 0;
880 if (!(id = ebml_peek_id(matroska, &level_up)) ||
881 level_up != 0 || id != EBML_ID_HEADER) {
882 av_log(matroska->ctx, AV_LOG_ERROR,
883 "This is not an EBML file (id=0x%x/0x%x)\n", id, EBML_ID_HEADER);
884 return AVERROR_INVALIDDATA;
886 if ((res = ebml_read_master(matroska, &id)) < 0)
890 if (!(id = ebml_peek_id(matroska, &level_up)))
898 /* is our read version uptodate? */
899 case EBML_ID_EBMLREADVERSION: {
902 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
904 if (num > EBML_VERSION) {
905 av_log(matroska->ctx, AV_LOG_ERROR,
906 "EBML version %"PRIu64" (> %d) is not supported\n",
908 return AVERROR_INVALIDDATA;
913 /* we only handle 8 byte lengths at max */
914 case EBML_ID_EBMLMAXSIZELENGTH: {
917 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
919 if (num > sizeof(uint64_t)) {
920 av_log(matroska->ctx, AV_LOG_ERROR,
921 "Integers of size %"PRIu64" (> %zd) not supported\n",
922 num, sizeof(uint64_t));
923 return AVERROR_INVALIDDATA;
928 /* we handle 4 byte IDs at max */
929 case EBML_ID_EBMLMAXIDLENGTH: {
932 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
934 if (num > sizeof(uint32_t)) {
935 av_log(matroska->ctx, AV_LOG_ERROR,
936 "IDs of size %"PRIu64" (> %zu) not supported\n",
937 num, sizeof(uint32_t));
938 return AVERROR_INVALIDDATA;
943 case EBML_ID_DOCTYPE: {
946 if ((res = ebml_read_ascii(matroska, &id, &text)) < 0)
957 case EBML_ID_DOCTYPEREADVERSION: {
960 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
968 av_log(matroska->ctx, AV_LOG_INFO,
969 "Unknown data type 0x%x in EBML header", id);
973 /* we ignore these two, as they don't tell us anything we
975 case EBML_ID_EBMLVERSION:
976 case EBML_ID_DOCTYPEVERSION:
977 res = ebml_read_skip (matroska);
986 * Put one packet in an application-supplied AVPacket struct.
987 * Returns 0 on success or -1 on failure.
991 matroska_deliver_packet (MatroskaDemuxContext *matroska,
994 if (matroska->num_packets > 0) {
995 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
996 av_free(matroska->packets[0]);
997 if (matroska->num_packets > 1) {
998 memmove(&matroska->packets[0], &matroska->packets[1],
999 (matroska->num_packets - 1) * sizeof(AVPacket *));
1001 av_realloc(matroska->packets, (matroska->num_packets - 1) *
1002 sizeof(AVPacket *));
1004 av_freep(&matroska->packets);
1006 matroska->num_packets--;
1014 * Put a packet into our internal queue. Will be delivered to the
1015 * user/application during the next get_packet() call.
1019 matroska_queue_packet (MatroskaDemuxContext *matroska,
1023 av_realloc(matroska->packets, (matroska->num_packets + 1) *
1024 sizeof(AVPacket *));
1025 matroska->packets[matroska->num_packets] = pkt;
1026 matroska->num_packets++;
1030 * Put a packet into our internal reordering queue. Will be moved to the
1031 * main packet queue when enough packets are available to reorder pts.
1035 matroska_queue_packet_reordered (MatroskaDemuxContext *matroska,
1039 if (matroska->num_packets_reorder && !is_bframe
1040 && pkt->pts > matroska->reorder_max_pts) {
1043 for (j=matroska->num_packets_reorder-1; j && k; j--) {
1045 for (i=0; i<j; i++) {
1046 if (matroska->packets_reorder[i]->pts > matroska->packets_reorder[i+1]->pts) {
1047 FFSWAP(uint64_t, matroska->packets_reorder[i]->pts, matroska->packets_reorder[i+1]->pts);
1052 /* then really queue the packets */
1053 for (i=0; i<matroska->num_packets_reorder; i++)
1054 matroska_queue_packet (matroska, matroska->packets_reorder[i]);
1055 matroska->num_packets_reorder = 0;
1057 matroska->packets_reorder =
1058 av_realloc(matroska->packets_reorder,
1059 (matroska->num_packets_reorder + 1) * sizeof(AVPacket *));
1060 matroska->packets_reorder[matroska->num_packets_reorder++] = pkt;
1061 if (pkt->pts > matroska->reorder_max_pts)
1062 matroska->reorder_max_pts = pkt->pts;
1071 matroska_probe (AVProbeData *p)
1074 int len_mask = 0x80, size = 1, n = 1;
1075 uint8_t probe_data[] = { 'm', 'a', 't', 'r', 'o', 's', 'k', 'a' };
1077 if (p->buf_size < 5)
1081 if ((p->buf[0] << 24 | p->buf[1] << 16 |
1082 p->buf[2] << 8 | p->buf[3]) != EBML_ID_HEADER)
1085 /* length of header */
1087 while (size <= 8 && !(total & len_mask)) {
1093 total &= (len_mask - 1);
1095 total = (total << 8) | p->buf[4 + n++];
1097 /* does the probe data contain the whole header? */
1098 if (p->buf_size < 4 + size + total)
1101 /* the header must contain the document type 'matroska'. For now,
1102 * we don't parse the whole header but simply check for the
1103 * availability of that array of characters inside the header.
1104 * Not fully fool-proof, but good enough. */
1105 for (n = 4 + size; n <= 4 + size + total - sizeof(probe_data); n++)
1106 if (!memcmp (&p->buf[n], probe_data, sizeof(probe_data)))
1107 return AVPROBE_SCORE_MAX;
1113 * From here on, it's all XML-style DTD stuff... Needs no comments.
1117 matroska_parse_info (MatroskaDemuxContext *matroska)
1122 av_log(matroska->ctx, AV_LOG_DEBUG, "Parsing info...\n");
1125 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1128 } else if (matroska->level_up) {
1129 matroska->level_up--;
1134 /* cluster timecode */
1135 case MATROSKA_ID_TIMECODESCALE: {
1137 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1139 matroska->time_scale = num;
1143 case MATROSKA_ID_DURATION: {
1145 if ((res = ebml_read_float(matroska, &id, &num)) < 0)
1147 matroska->ctx->duration = num * matroska->time_scale * 1000 / AV_TIME_BASE;
1151 case MATROSKA_ID_TITLE: {
1153 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1155 strncpy(matroska->ctx->title, text,
1156 sizeof(matroska->ctx->title)-1);
1161 case MATROSKA_ID_WRITINGAPP: {
1163 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1165 matroska->writing_app = text;
1169 case MATROSKA_ID_MUXINGAPP: {
1171 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1173 matroska->muxing_app = text;
1177 case MATROSKA_ID_DATEUTC: {
1179 if ((res = ebml_read_date(matroska, &id, &time)) < 0)
1181 matroska->created = time;
1186 av_log(matroska->ctx, AV_LOG_INFO,
1187 "Unknown entry 0x%x in info header\n", id);
1191 res = ebml_read_skip(matroska);
1195 if (matroska->level_up) {
1196 matroska->level_up--;
1205 matroska_add_stream (MatroskaDemuxContext *matroska)
1209 MatroskaTrack *track;
1211 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing track, adding stream..,\n");
1213 /* Allocate a generic track. As soon as we know its type we'll realloc. */
1214 track = av_mallocz(MAX_TRACK_SIZE);
1215 matroska->num_tracks++;
1217 /* start with the master */
1218 if ((res = ebml_read_master(matroska, &id)) < 0)
1221 /* try reading the trackentry headers */
1223 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1226 } else if (matroska->level_up > 0) {
1227 matroska->level_up--;
1232 /* track number (unique stream ID) */
1233 case MATROSKA_ID_TRACKNUMBER: {
1235 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1241 /* track UID (unique identifier) */
1242 case MATROSKA_ID_TRACKUID: {
1244 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1250 /* track type (video, audio, combined, subtitle, etc.) */
1251 case MATROSKA_ID_TRACKTYPE: {
1253 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1255 if (track->type && track->type != num) {
1256 av_log(matroska->ctx, AV_LOG_INFO,
1257 "More than one tracktype in an entry - skip\n");
1262 switch (track->type) {
1263 case MATROSKA_TRACK_TYPE_VIDEO:
1264 case MATROSKA_TRACK_TYPE_AUDIO:
1265 case MATROSKA_TRACK_TYPE_SUBTITLE:
1267 case MATROSKA_TRACK_TYPE_COMPLEX:
1268 case MATROSKA_TRACK_TYPE_LOGO:
1269 case MATROSKA_TRACK_TYPE_CONTROL:
1271 av_log(matroska->ctx, AV_LOG_INFO,
1272 "Unknown or unsupported track type 0x%x\n",
1277 matroska->tracks[matroska->num_tracks - 1] = track;
1281 /* tracktype specific stuff for video */
1282 case MATROSKA_ID_TRACKVIDEO: {
1283 MatroskaVideoTrack *videotrack;
1285 track->type = MATROSKA_TRACK_TYPE_VIDEO;
1286 if (track->type != MATROSKA_TRACK_TYPE_VIDEO) {
1287 av_log(matroska->ctx, AV_LOG_INFO,
1288 "video data in non-video track - ignoring\n");
1289 res = AVERROR_INVALIDDATA;
1291 } else if ((res = ebml_read_master(matroska, &id)) < 0)
1293 videotrack = (MatroskaVideoTrack *)track;
1296 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1299 } else if (matroska->level_up > 0) {
1300 matroska->level_up--;
1305 /* fixme, this should be one-up, but I get it here */
1306 case MATROSKA_ID_TRACKDEFAULTDURATION: {
1308 if ((res = ebml_read_uint (matroska, &id,
1311 track->default_duration = num/matroska->time_scale;
1315 /* video framerate */
1316 case MATROSKA_ID_VIDEOFRAMERATE: {
1318 if ((res = ebml_read_float(matroska, &id,
1321 track->default_duration = 1000000000/(matroska->time_scale*num);
1325 /* width of the size to display the video at */
1326 case MATROSKA_ID_VIDEODISPLAYWIDTH: {
1328 if ((res = ebml_read_uint(matroska, &id,
1331 videotrack->display_width = num;
1335 /* height of the size to display the video at */
1336 case MATROSKA_ID_VIDEODISPLAYHEIGHT: {
1338 if ((res = ebml_read_uint(matroska, &id,
1341 videotrack->display_height = num;
1345 /* width of the video in the file */
1346 case MATROSKA_ID_VIDEOPIXELWIDTH: {
1348 if ((res = ebml_read_uint(matroska, &id,
1351 videotrack->pixel_width = num;
1355 /* height of the video in the file */
1356 case MATROSKA_ID_VIDEOPIXELHEIGHT: {
1358 if ((res = ebml_read_uint(matroska, &id,
1361 videotrack->pixel_height = num;
1365 /* whether the video is interlaced */
1366 case MATROSKA_ID_VIDEOFLAGINTERLACED: {
1368 if ((res = ebml_read_uint(matroska, &id,
1373 MATROSKA_VIDEOTRACK_INTERLACED;
1376 ~MATROSKA_VIDEOTRACK_INTERLACED;
1380 /* stereo mode (whether the video has two streams,
1381 * where one is for the left eye and the other for
1382 * the right eye, which creates a 3D-like
1384 case MATROSKA_ID_VIDEOSTEREOMODE: {
1386 if ((res = ebml_read_uint(matroska, &id,
1389 if (num != MATROSKA_EYE_MODE_MONO &&
1390 num != MATROSKA_EYE_MODE_LEFT &&
1391 num != MATROSKA_EYE_MODE_RIGHT &&
1392 num != MATROSKA_EYE_MODE_BOTH) {
1393 av_log(matroska->ctx, AV_LOG_INFO,
1394 "Ignoring unknown eye mode 0x%x\n",
1398 videotrack->eye_mode = num;
1402 /* aspect ratio behaviour */
1403 case MATROSKA_ID_VIDEOASPECTRATIO: {
1405 if ((res = ebml_read_uint(matroska, &id,
1408 if (num != MATROSKA_ASPECT_RATIO_MODE_FREE &&
1409 num != MATROSKA_ASPECT_RATIO_MODE_KEEP &&
1410 num != MATROSKA_ASPECT_RATIO_MODE_FIXED) {
1411 av_log(matroska->ctx, AV_LOG_INFO,
1412 "Ignoring unknown aspect ratio 0x%x\n",
1416 videotrack->ar_mode = num;
1420 /* colourspace (only matters for raw video)
1422 case MATROSKA_ID_VIDEOCOLOURSPACE: {
1424 if ((res = ebml_read_uint(matroska, &id,
1427 videotrack->fourcc = num;
1432 av_log(matroska->ctx, AV_LOG_INFO,
1433 "Unknown video track header entry "
1434 "0x%x - ignoring\n", id);
1438 res = ebml_read_skip(matroska);
1442 if (matroska->level_up) {
1443 matroska->level_up--;
1450 /* tracktype specific stuff for audio */
1451 case MATROSKA_ID_TRACKAUDIO: {
1452 MatroskaAudioTrack *audiotrack;
1454 track->type = MATROSKA_TRACK_TYPE_AUDIO;
1455 if (track->type != MATROSKA_TRACK_TYPE_AUDIO) {
1456 av_log(matroska->ctx, AV_LOG_INFO,
1457 "audio data in non-audio track - ignoring\n");
1458 res = AVERROR_INVALIDDATA;
1460 } else if ((res = ebml_read_master(matroska, &id)) < 0)
1462 audiotrack = (MatroskaAudioTrack *)track;
1463 audiotrack->channels = 1;
1464 audiotrack->samplerate = 8000;
1467 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1470 } else if (matroska->level_up > 0) {
1471 matroska->level_up--;
1477 case MATROSKA_ID_AUDIOSAMPLINGFREQ: {
1479 if ((res = ebml_read_float(matroska, &id,
1482 audiotrack->internal_samplerate =
1483 audiotrack->samplerate = num;
1487 case MATROSKA_ID_AUDIOOUTSAMPLINGFREQ: {
1489 if ((res = ebml_read_float(matroska, &id,
1492 audiotrack->samplerate = num;
1497 case MATROSKA_ID_AUDIOBITDEPTH: {
1499 if ((res = ebml_read_uint(matroska, &id,
1502 audiotrack->bitdepth = num;
1507 case MATROSKA_ID_AUDIOCHANNELS: {
1509 if ((res = ebml_read_uint(matroska, &id,
1512 audiotrack->channels = num;
1517 av_log(matroska->ctx, AV_LOG_INFO,
1518 "Unknown audio track header entry "
1519 "0x%x - ignoring\n", id);
1523 res = ebml_read_skip(matroska);
1527 if (matroska->level_up) {
1528 matroska->level_up--;
1535 /* codec identifier */
1536 case MATROSKA_ID_CODECID: {
1538 if ((res = ebml_read_ascii(matroska, &id, &text)) < 0)
1540 track->codec_id = text;
1544 /* codec private data */
1545 case MATROSKA_ID_CODECPRIVATE: {
1548 if ((res = ebml_read_binary(matroska, &id, &data, &size) < 0))
1550 track->codec_priv = data;
1551 track->codec_priv_size = size;
1555 /* name of the codec */
1556 case MATROSKA_ID_CODECNAME: {
1558 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1560 track->codec_name = text;
1564 /* name of this track */
1565 case MATROSKA_ID_TRACKNAME: {
1567 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1573 /* language (matters for audio/subtitles, mostly) */
1574 case MATROSKA_ID_TRACKLANGUAGE: {
1576 if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
1578 track->language = text;
1582 /* whether this is actually used */
1583 case MATROSKA_ID_TRACKFLAGENABLED: {
1585 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1588 track->flags |= MATROSKA_TRACK_ENABLED;
1590 track->flags &= ~MATROSKA_TRACK_ENABLED;
1594 /* whether it's the default for this track type */
1595 case MATROSKA_ID_TRACKFLAGDEFAULT: {
1597 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1600 track->flags |= MATROSKA_TRACK_DEFAULT;
1602 track->flags &= ~MATROSKA_TRACK_DEFAULT;
1606 /* lacing (like MPEG, where blocks don't end/start on frame
1608 case MATROSKA_ID_TRACKFLAGLACING: {
1610 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1613 track->flags |= MATROSKA_TRACK_LACING;
1615 track->flags &= ~MATROSKA_TRACK_LACING;
1619 /* default length (in time) of one data block in this track */
1620 case MATROSKA_ID_TRACKDEFAULTDURATION: {
1622 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
1624 track->default_duration = num / matroska->time_scale;
1629 av_log(matroska->ctx, AV_LOG_INFO,
1630 "Unknown track header entry 0x%x - ignoring\n", id);
1634 /* we ignore these because they're nothing useful. */
1635 case MATROSKA_ID_CODECINFOURL:
1636 case MATROSKA_ID_CODECDOWNLOADURL:
1637 case MATROSKA_ID_TRACKMINCACHE:
1638 case MATROSKA_ID_TRACKMAXCACHE:
1639 res = ebml_read_skip(matroska);
1643 if (matroska->level_up) {
1644 matroska->level_up--;
1653 matroska_parse_tracks (MatroskaDemuxContext *matroska)
1658 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing tracks...\n");
1661 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1664 } else if (matroska->level_up) {
1665 matroska->level_up--;
1670 /* one track within the "all-tracks" header */
1671 case MATROSKA_ID_TRACKENTRY:
1672 res = matroska_add_stream(matroska);
1676 av_log(matroska->ctx, AV_LOG_INFO,
1677 "Unknown entry 0x%x in track header\n", id);
1681 res = ebml_read_skip(matroska);
1685 if (matroska->level_up) {
1686 matroska->level_up--;
1695 matroska_parse_index (MatroskaDemuxContext *matroska)
1699 MatroskaDemuxIndex idx;
1701 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing index...\n");
1704 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1707 } else if (matroska->level_up) {
1708 matroska->level_up--;
1713 /* one single index entry ('point') */
1714 case MATROSKA_ID_POINTENTRY:
1715 if ((res = ebml_read_master(matroska, &id)) < 0)
1718 /* in the end, we hope to fill one entry with a
1719 * timestamp, a file position and a tracknum */
1720 idx.pos = (uint64_t) -1;
1721 idx.time = (uint64_t) -1;
1722 idx.track = (uint16_t) -1;
1725 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1728 } else if (matroska->level_up) {
1729 matroska->level_up--;
1734 /* one single index entry ('point') */
1735 case MATROSKA_ID_CUETIME: {
1737 if ((res = ebml_read_uint(matroska, &id,
1740 idx.time = time * matroska->time_scale;
1744 /* position in the file + track to which it
1746 case MATROSKA_ID_CUETRACKPOSITION:
1747 if ((res = ebml_read_master(matroska, &id)) < 0)
1751 if (!(id = ebml_peek_id (matroska,
1752 &matroska->level_up))) {
1755 } else if (matroska->level_up) {
1756 matroska->level_up--;
1762 case MATROSKA_ID_CUETRACK: {
1764 if ((res = ebml_read_uint(matroska,
1771 /* position in file */
1772 case MATROSKA_ID_CUECLUSTERPOSITION: {
1774 if ((res = ebml_read_uint(matroska,
1782 av_log(matroska->ctx, AV_LOG_INFO,
1783 "Unknown entry 0x%x in "
1784 "CuesTrackPositions\n", id);
1788 res = ebml_read_skip(matroska);
1792 if (matroska->level_up) {
1793 matroska->level_up--;
1801 av_log(matroska->ctx, AV_LOG_INFO,
1802 "Unknown entry 0x%x in cuespoint "
1807 res = ebml_read_skip(matroska);
1811 if (matroska->level_up) {
1812 matroska->level_up--;
1817 /* so let's see if we got what we wanted */
1818 if (idx.pos != (uint64_t) -1 &&
1819 idx.time != (uint64_t) -1 &&
1820 idx.track != (uint16_t) -1) {
1821 if (matroska->num_indexes % 32 == 0) {
1822 /* re-allocate bigger index */
1824 av_realloc(matroska->index,
1825 (matroska->num_indexes + 32) *
1826 sizeof(MatroskaDemuxIndex));
1828 matroska->index[matroska->num_indexes] = idx;
1829 matroska->num_indexes++;
1834 av_log(matroska->ctx, AV_LOG_INFO,
1835 "Unknown entry 0x%x in cues header\n", id);
1839 res = ebml_read_skip(matroska);
1843 if (matroska->level_up) {
1844 matroska->level_up--;
1853 matroska_parse_metadata (MatroskaDemuxContext *matroska)
1859 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1862 } else if (matroska->level_up) {
1863 matroska->level_up--;
1868 /* Hm, this is unsupported... */
1870 av_log(matroska->ctx, AV_LOG_INFO,
1871 "Unknown entry 0x%x in metadata header\n", id);
1875 res = ebml_read_skip(matroska);
1879 if (matroska->level_up) {
1880 matroska->level_up--;
1889 matroska_parse_seekhead (MatroskaDemuxContext *matroska)
1894 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing seekhead...\n");
1897 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1900 } else if (matroska->level_up) {
1901 matroska->level_up--;
1906 case MATROSKA_ID_SEEKENTRY: {
1907 uint32_t seek_id = 0, peek_id_cache = 0;
1908 uint64_t seek_pos = (uint64_t) -1, t;
1910 if ((res = ebml_read_master(matroska, &id)) < 0)
1914 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
1917 } else if (matroska->level_up) {
1918 matroska->level_up--;
1923 case MATROSKA_ID_SEEKID:
1924 res = ebml_read_uint(matroska, &id, &t);
1928 case MATROSKA_ID_SEEKPOSITION:
1929 res = ebml_read_uint(matroska, &id, &seek_pos);
1933 av_log(matroska->ctx, AV_LOG_INFO,
1934 "Unknown seekhead ID 0x%x\n", id);
1938 res = ebml_read_skip(matroska);
1942 if (matroska->level_up) {
1943 matroska->level_up--;
1948 if (!seek_id || seek_pos == (uint64_t) -1) {
1949 av_log(matroska->ctx, AV_LOG_INFO,
1950 "Incomplete seekhead entry (0x%x/%"PRIu64")\n",
1956 case MATROSKA_ID_CUES:
1957 case MATROSKA_ID_TAGS: {
1958 uint32_t level_up = matroska->level_up;
1959 offset_t before_pos;
1961 MatroskaLevel level;
1963 /* remember the peeked ID and the current position */
1964 peek_id_cache = matroska->peek_id;
1965 before_pos = url_ftell(&matroska->ctx->pb);
1968 if ((res = ebml_read_seek(matroska, seek_pos +
1969 matroska->segment_start)) < 0)
1972 /* we don't want to lose our seekhead level, so we add
1973 * a dummy. This is a crude hack. */
1974 if (matroska->num_levels == EBML_MAX_DEPTH) {
1975 av_log(matroska->ctx, AV_LOG_INFO,
1976 "Max EBML element depth (%d) reached, "
1977 "cannot parse further.\n", EBML_MAX_DEPTH);
1978 return AVERROR_UNKNOWN;
1982 level.length = (uint64_t)-1;
1983 matroska->levels[matroska->num_levels] = level;
1984 matroska->num_levels++;
1987 if (!(id = ebml_peek_id (matroska,
1988 &matroska->level_up)))
1990 if (id != seek_id) {
1991 av_log(matroska->ctx, AV_LOG_INFO,
1992 "We looked for ID=0x%x but got "
1993 "ID=0x%x (pos=%"PRIu64")",
1994 seek_id, id, seek_pos +
1995 matroska->segment_start);
1999 /* read master + parse */
2000 if ((res = ebml_read_master(matroska, &id)) < 0)
2003 case MATROSKA_ID_CUES:
2004 if (!(res = matroska_parse_index(matroska)) ||
2005 url_feof(&matroska->ctx->pb)) {
2006 matroska->index_parsed = 1;
2010 case MATROSKA_ID_TAGS:
2011 if (!(res = matroska_parse_metadata(matroska)) ||
2012 url_feof(&matroska->ctx->pb)) {
2013 matroska->metadata_parsed = 1;
2020 /* remove dummy level */
2021 while (matroska->num_levels) {
2022 matroska->num_levels--;
2024 matroska->levels[matroska->num_levels].length;
2025 if (length == (uint64_t)-1)
2030 if ((res = ebml_read_seek(matroska, before_pos)) < 0)
2032 matroska->peek_id = peek_id_cache;
2033 matroska->level_up = level_up;
2038 av_log(matroska->ctx, AV_LOG_INFO,
2039 "Ignoring seekhead entry for ID=0x%x\n",
2048 av_log(matroska->ctx, AV_LOG_INFO,
2049 "Unknown seekhead ID 0x%x\n", id);
2053 res = ebml_read_skip(matroska);
2057 if (matroska->level_up) {
2058 matroska->level_up--;
2066 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(*x))
2069 matroska_aac_profile (char *codec_id)
2071 static const char *aac_profiles[] = {
2076 for (profile=0; profile<ARRAY_SIZE(aac_profiles); profile++)
2077 if (strstr(codec_id, aac_profiles[profile]))
2083 matroska_aac_sri (int samplerate)
2085 static const int aac_sample_rates[] = {
2086 96000, 88200, 64000, 48000, 44100, 32000,
2087 24000, 22050, 16000, 12000, 11025, 8000,
2091 for (sri=0; sri<ARRAY_SIZE(aac_sample_rates); sri++)
2092 if (aac_sample_rates[sri] == samplerate)
2098 matroska_read_header (AVFormatContext *s,
2099 AVFormatParameters *ap)
2101 MatroskaDemuxContext *matroska = s->priv_data;
2103 int version, last_level, res = 0;
2108 /* First read the EBML header. */
2110 if ((res = ebml_read_header(matroska, &doctype, &version)) < 0)
2112 if ((doctype == NULL) || strcmp(doctype, "matroska")) {
2113 av_log(matroska->ctx, AV_LOG_ERROR,
2114 "Wrong EBML doctype ('%s' != 'matroska').\n",
2115 doctype ? doctype : "(none)");
2118 return AVERROR_NOFMT;
2122 av_log(matroska->ctx, AV_LOG_ERROR,
2123 "Matroska demuxer version 2 too old for file version %d\n",
2125 return AVERROR_NOFMT;
2128 /* The next thing is a segment. */
2130 if (!(id = ebml_peek_id(matroska, &last_level)))
2132 if (id == MATROSKA_ID_SEGMENT)
2136 av_log(matroska->ctx, AV_LOG_INFO,
2137 "Expected a Segment ID (0x%x), but received 0x%x!\n",
2138 MATROSKA_ID_SEGMENT, id);
2139 if ((res = ebml_read_skip(matroska)) < 0)
2143 /* We now have a Matroska segment.
2144 * Seeks are from the beginning of the segment,
2145 * after the segment ID/length. */
2146 if ((res = ebml_read_master(matroska, &id)) < 0)
2148 matroska->segment_start = url_ftell(&s->pb);
2150 matroska->time_scale = 1000000;
2151 /* we've found our segment, start reading the different contents in here */
2153 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2156 } else if (matroska->level_up) {
2157 matroska->level_up--;
2163 case MATROSKA_ID_INFO: {
2164 if ((res = ebml_read_master(matroska, &id)) < 0)
2166 res = matroska_parse_info(matroska);
2170 /* track info headers */
2171 case MATROSKA_ID_TRACKS: {
2172 if ((res = ebml_read_master(matroska, &id)) < 0)
2174 res = matroska_parse_tracks(matroska);
2179 case MATROSKA_ID_CUES: {
2180 if (!matroska->index_parsed) {
2181 if ((res = ebml_read_master(matroska, &id)) < 0)
2183 res = matroska_parse_index(matroska);
2185 res = ebml_read_skip(matroska);
2190 case MATROSKA_ID_TAGS: {
2191 if (!matroska->metadata_parsed) {
2192 if ((res = ebml_read_master(matroska, &id)) < 0)
2194 res = matroska_parse_metadata(matroska);
2196 res = ebml_read_skip(matroska);
2200 /* file index (if seekable, seek to Cues/Tags to parse it) */
2201 case MATROSKA_ID_SEEKHEAD: {
2202 if ((res = ebml_read_master(matroska, &id)) < 0)
2204 res = matroska_parse_seekhead(matroska);
2208 case MATROSKA_ID_CLUSTER: {
2209 /* Do not read the master - this will be done in the next
2210 * call to matroska_read_packet. */
2216 av_log(matroska->ctx, AV_LOG_INFO,
2217 "Unknown matroska file header ID 0x%x\n", id);
2221 res = ebml_read_skip(matroska);
2225 if (matroska->level_up) {
2226 matroska->level_up--;
2231 /* Have we found a cluster? */
2232 if (ebml_peek_id(matroska, NULL) == MATROSKA_ID_CLUSTER) {
2234 MatroskaTrack *track;
2237 for (i = 0; i < matroska->num_tracks; i++) {
2238 enum CodecID codec_id = CODEC_ID_NONE;
2239 uint8_t *extradata = NULL;
2240 int extradata_size = 0;
2241 int extradata_offset = 0;
2242 track = matroska->tracks[i];
2244 /* libavformat does not really support subtitles.
2245 * Also apply some sanity checks. */
2246 if ((track->type == MATROSKA_TRACK_TYPE_SUBTITLE) ||
2247 (track->codec_id == NULL))
2250 for(j=0; codec_tags[j].str; j++){
2251 if(!strncmp(codec_tags[j].str, track->codec_id,
2252 strlen(codec_tags[j].str))){
2253 codec_id= codec_tags[j].id;
2258 /* Set the FourCC from the CodecID. */
2259 /* This is the MS compatibility mode which stores a
2260 * BITMAPINFOHEADER in the CodecPrivate. */
2261 if (!strcmp(track->codec_id,
2262 MATROSKA_CODEC_ID_VIDEO_VFW_FOURCC) &&
2263 (track->codec_priv_size >= 40) &&
2264 (track->codec_priv != NULL)) {
2267 /* Offset of biCompression. Stored in LE. */
2268 p = (unsigned char *)track->codec_priv + 16;
2269 ((MatroskaVideoTrack *)track)->fourcc = (p[3] << 24) |
2270 (p[2] << 16) | (p[1] << 8) | p[0];
2271 codec_id = codec_get_id(codec_bmp_tags, ((MatroskaVideoTrack *)track)->fourcc);
2275 /* This is the MS compatibility mode which stores a
2276 * WAVEFORMATEX in the CodecPrivate. */
2277 else if (!strcmp(track->codec_id,
2278 MATROSKA_CODEC_ID_AUDIO_ACM) &&
2279 (track->codec_priv_size >= 18) &&
2280 (track->codec_priv != NULL)) {
2284 /* Offset of wFormatTag. Stored in LE. */
2285 p = (unsigned char *)track->codec_priv;
2286 tag = (p[1] << 8) | p[0];
2287 codec_id = codec_get_id(codec_wav_tags, tag);
2291 else if (codec_id == CODEC_ID_AAC && !track->codec_priv_size) {
2292 MatroskaAudioTrack *audiotrack = (MatroskaAudioTrack *) track;
2293 int profile = matroska_aac_profile(track->codec_id);
2294 int sri = matroska_aac_sri(audiotrack->internal_samplerate);
2295 extradata = av_malloc(5);
2296 if (extradata == NULL)
2297 return AVERROR_NOMEM;
2298 extradata[0] = (profile << 3) | ((sri&0x0E) >> 1);
2299 extradata[1] = ((sri&0x01) << 7) | (audiotrack->channels<<3);
2300 if (strstr(track->codec_id, "SBR")) {
2301 sri = matroska_aac_sri(audiotrack->samplerate);
2302 extradata[2] = 0x56;
2303 extradata[3] = 0xE5;
2304 extradata[4] = 0x80 | (sri<<3);
2309 track->default_duration = 1024*1000 / audiotrack->internal_samplerate;
2312 else if (codec_id == CODEC_ID_TTA) {
2313 MatroskaAudioTrack *audiotrack = (MatroskaAudioTrack *) track;
2315 extradata_size = 30;
2316 extradata = av_mallocz(extradata_size);
2317 if (extradata == NULL)
2318 return AVERROR_NOMEM;
2319 init_put_byte(&b, extradata, extradata_size, 1,
2320 NULL, NULL, NULL, NULL);
2321 put_buffer(&b, (uint8_t *) "TTA1", 4);
2323 put_le16(&b, audiotrack->channels);
2324 put_le16(&b, audiotrack->bitdepth);
2325 put_le32(&b, audiotrack->samplerate);
2326 put_le32(&b, matroska->ctx->duration * audiotrack->samplerate);
2329 else if (codec_id == CODEC_ID_RV10 || codec_id == CODEC_ID_RV20 ||
2330 codec_id == CODEC_ID_RV30 || codec_id == CODEC_ID_RV40) {
2331 extradata_offset = 26;
2332 track->codec_priv_size -= extradata_offset;
2333 track->flags |= MATROSKA_TRACK_REAL_V;
2336 if (codec_id == CODEC_ID_NONE) {
2337 av_log(matroska->ctx, AV_LOG_INFO,
2338 "Unknown/unsupported CodecID %s.\n",
2342 track->stream_index = matroska->num_streams;
2344 matroska->num_streams++;
2345 st = av_new_stream(s, track->stream_index);
2347 return AVERROR_NOMEM;
2348 av_set_pts_info(st, 64, matroska->time_scale, 1000*1000*1000); /* 64 bit pts in ns */
2350 st->codec->codec_id = codec_id;
2352 if (track->default_duration)
2353 av_reduce(&st->codec->time_base.num, &st->codec->time_base.den,
2354 track->default_duration, 1000, 30000);
2357 st->codec->extradata = extradata;
2358 st->codec->extradata_size = extradata_size;
2359 } else if(track->codec_priv && track->codec_priv_size > 0){
2360 st->codec->extradata = av_malloc(track->codec_priv_size);
2361 if(st->codec->extradata == NULL)
2362 return AVERROR_NOMEM;
2363 st->codec->extradata_size = track->codec_priv_size;
2364 memcpy(st->codec->extradata,track->codec_priv+extradata_offset,
2365 track->codec_priv_size);
2368 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2369 MatroskaVideoTrack *videotrack = (MatroskaVideoTrack *)track;
2371 st->codec->codec_type = CODEC_TYPE_VIDEO;
2372 st->codec->codec_tag = videotrack->fourcc;
2373 st->codec->width = videotrack->pixel_width;
2374 st->codec->height = videotrack->pixel_height;
2375 if (videotrack->display_width == 0)
2376 videotrack->display_width= videotrack->pixel_width;
2377 if (videotrack->display_height == 0)
2378 videotrack->display_height= videotrack->pixel_height;
2379 av_reduce(&st->codec->sample_aspect_ratio.num,
2380 &st->codec->sample_aspect_ratio.den,
2381 st->codec->height * videotrack->display_width,
2382 st->codec-> width * videotrack->display_height,
2384 st->need_parsing = 2;
2385 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2386 MatroskaAudioTrack *audiotrack = (MatroskaAudioTrack *)track;
2388 st->codec->codec_type = CODEC_TYPE_AUDIO;
2389 st->codec->sample_rate = audiotrack->samplerate;
2390 st->codec->channels = audiotrack->channels;
2391 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2392 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
2395 /* What do we do with private data? E.g. for Vorbis. */
2404 matroska_find_track_by_num (MatroskaDemuxContext *matroska,
2409 for (i = 0; i < matroska->num_tracks; i++)
2410 if (matroska->tracks[i]->num == num)
2417 rv_offset(uint8_t *data, int slice, int slices)
2419 return AV_RL32(data+8*slice+4) + 8*slices;
2423 matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data, int size,
2424 int64_t pos, uint64_t cluster_time, uint64_t duration,
2425 int is_keyframe, int is_bframe)
2430 uint8_t *origdata = data;
2432 uint32_t *lace_size = NULL;
2433 int n, flags, laces = 0;
2436 /* first byte(s): tracknum */
2437 if ((n = matroska_ebmlnum_uint(data, size, &num)) < 0) {
2438 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2445 /* fetch track from num */
2446 track = matroska_find_track_by_num(matroska, num);
2447 if (size <= 3 || track < 0 || track >= matroska->num_tracks) {
2448 av_log(matroska->ctx, AV_LOG_INFO,
2449 "Invalid stream %d or size %u\n", track, size);
2453 if(matroska->ctx->streams[ matroska->tracks[track]->stream_index ]->discard >= AVDISCARD_ALL){
2457 if (duration == AV_NOPTS_VALUE)
2458 duration = matroska->tracks[track]->default_duration;
2460 /* block_time (relative to cluster time) */
2461 block_time = (data[0] << 8) | data[1];
2467 if (is_keyframe == -1)
2468 is_keyframe = flags & 1 ? PKT_FLAG_KEY : 0;
2469 switch ((flags & 0x06) >> 1) {
2470 case 0x0: /* no lacing */
2472 lace_size = av_mallocz(sizeof(int));
2473 lace_size[0] = size;
2476 case 0x1: /* xiph lacing */
2477 case 0x2: /* fixed-size lacing */
2478 case 0x3: /* EBML lacing */
2483 laces = (*data) + 1;
2486 lace_size = av_mallocz(laces * sizeof(int));
2488 switch ((flags & 0x06) >> 1) {
2489 case 0x1: /* xiph lacing */ {
2492 for (n = 0; res == 0 && n < laces - 1; n++) {
2499 lace_size[n] += temp;
2505 total += lace_size[n];
2507 lace_size[n] = size - total;
2511 case 0x2: /* fixed-size lacing */
2512 for (n = 0; n < laces; n++)
2513 lace_size[n] = size / laces;
2516 case 0x3: /* EBML lacing */ {
2518 n = matroska_ebmlnum_uint(data, size, &num);
2520 av_log(matroska->ctx, AV_LOG_INFO,
2521 "EBML block data error\n");
2526 total = lace_size[0] = num;
2527 for (n = 1; res == 0 && n < laces - 1; n++) {
2530 r = matroska_ebmlnum_sint (data, size, &snum);
2532 av_log(matroska->ctx, AV_LOG_INFO,
2533 "EBML block data error\n");
2538 lace_size[n] = lace_size[n - 1] + snum;
2539 total += lace_size[n];
2541 lace_size[n] = size - total;
2549 int real_v = matroska->tracks[track]->flags & MATROSKA_TRACK_REAL_V;
2550 uint64_t timecode = AV_NOPTS_VALUE;
2552 if (cluster_time != (uint64_t)-1 && cluster_time + block_time >= 0)
2553 timecode = cluster_time + block_time;
2555 for (n = 0; n < laces; n++) {
2556 int slice, slices = 1;
2559 slices = *data++ + 1;
2563 for (slice=0; slice<slices; slice++) {
2564 int slice_size, slice_offset = 0;
2566 slice_offset = rv_offset(data, slice, slices);
2567 if (slice+1 == slices)
2568 slice_size = lace_size[n] - slice_offset;
2570 slice_size = rv_offset(data, slice+1, slices) - slice_offset;
2571 pkt = av_mallocz(sizeof(AVPacket));
2572 /* XXX: prevent data copy... */
2573 if (av_new_packet(pkt, slice_size) < 0) {
2574 res = AVERROR_NOMEM;
2578 memcpy (pkt->data, data+slice_offset, slice_size);
2581 pkt->flags = is_keyframe;
2582 pkt->stream_index = matroska->tracks[track]->stream_index;
2584 pkt->pts = timecode;
2586 pkt->duration = duration;
2588 if (matroska->tracks[track]->flags & MATROSKA_TRACK_REORDER)
2589 matroska_queue_packet_reordered(matroska, pkt, is_bframe);
2591 matroska_queue_packet(matroska, pkt);
2593 if (timecode != AV_NOPTS_VALUE)
2594 timecode = duration ? timecode + duration : AV_NOPTS_VALUE;
2596 data += lace_size[n];
2606 matroska_parse_blockgroup (MatroskaDemuxContext *matroska,
2607 uint64_t cluster_time)
2612 int is_keyframe = PKT_FLAG_KEY, last_num_packets = matroska->num_packets;
2613 uint64_t duration = AV_NOPTS_VALUE;
2618 av_log(matroska->ctx, AV_LOG_DEBUG, "parsing blockgroup...\n");
2621 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2624 } else if (matroska->level_up) {
2625 matroska->level_up--;
2630 /* one block inside the group. Note, block parsing is one
2631 * of the harder things, so this code is a bit complicated.
2632 * See http://www.matroska.org/ for documentation. */
2633 case MATROSKA_ID_BLOCK: {
2634 pos = url_ftell(&matroska->ctx->pb);
2635 res = ebml_read_binary(matroska, &id, &data, &size);
2639 case MATROSKA_ID_BLOCKDURATION: {
2640 if ((res = ebml_read_uint(matroska, &id, &duration)) < 0)
2642 duration /= matroska->time_scale;
2646 case MATROSKA_ID_BLOCKREFERENCE: {
2648 /* We've found a reference, so not even the first frame in
2649 * the lace is a key frame. */
2651 if (last_num_packets != matroska->num_packets)
2652 matroska->packets[last_num_packets]->flags = 0;
2653 if ((res = ebml_read_sint(matroska, &id, &num)) < 0)
2661 av_log(matroska->ctx, AV_LOG_INFO,
2662 "Unknown entry 0x%x in blockgroup data\n", id);
2666 res = ebml_read_skip(matroska);
2670 if (matroska->level_up) {
2671 matroska->level_up--;
2680 res = matroska_parse_block(matroska, data, size, pos, cluster_time,
2681 duration, is_keyframe, is_bframe);
2687 matroska_parse_cluster (MatroskaDemuxContext *matroska)
2691 uint64_t cluster_time = 0;
2696 av_log(matroska->ctx, AV_LOG_DEBUG,
2697 "parsing cluster at %"PRId64"\n", url_ftell(&matroska->ctx->pb));
2700 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2703 } else if (matroska->level_up) {
2704 matroska->level_up--;
2709 /* cluster timecode */
2710 case MATROSKA_ID_CLUSTERTIMECODE: {
2712 if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
2718 /* a group of blocks inside a cluster */
2719 case MATROSKA_ID_BLOCKGROUP:
2720 if ((res = ebml_read_master(matroska, &id)) < 0)
2722 res = matroska_parse_blockgroup(matroska, cluster_time);
2725 case MATROSKA_ID_SIMPLEBLOCK:
2726 pos = url_ftell(&matroska->ctx->pb);
2727 res = ebml_read_binary(matroska, &id, &data, &size);
2729 res = matroska_parse_block(matroska, data, size, pos,
2730 cluster_time, AV_NOPTS_VALUE,
2735 av_log(matroska->ctx, AV_LOG_INFO,
2736 "Unknown entry 0x%x in cluster data\n", id);
2740 res = ebml_read_skip(matroska);
2744 if (matroska->level_up) {
2745 matroska->level_up--;
2754 matroska_read_packet (AVFormatContext *s,
2757 MatroskaDemuxContext *matroska = s->priv_data;
2761 /* Read stream until we have a packet queued. */
2762 while (matroska_deliver_packet(matroska, pkt)) {
2764 /* Have we already reached the end? */
2769 if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
2771 } else if (matroska->level_up) {
2772 matroska->level_up--;
2777 case MATROSKA_ID_CLUSTER:
2778 if ((res = ebml_read_master(matroska, &id)) < 0)
2780 if ((res = matroska_parse_cluster(matroska)) == 0)
2781 res = 1; /* Parsed one cluster, let's get out. */
2786 res = ebml_read_skip(matroska);
2790 if (matroska->level_up) {
2791 matroska->level_up--;
2804 matroska_read_close (AVFormatContext *s)
2806 MatroskaDemuxContext *matroska = s->priv_data;
2809 av_free(matroska->writing_app);
2810 av_free(matroska->muxing_app);
2811 av_free(matroska->index);
2813 if (matroska->packets != NULL) {
2814 for (n = 0; n < matroska->num_packets; n++) {
2815 av_free_packet(matroska->packets[n]);
2816 av_free(matroska->packets[n]);
2818 av_free(matroska->packets);
2820 if (matroska->packets_reorder) {
2821 for (n = 0; n < matroska->num_packets_reorder; n++) {
2822 av_free_packet(matroska->packets_reorder[n]);
2823 av_free(matroska->packets_reorder[n]);
2825 av_free(matroska->packets_reorder);
2828 for (n = 0; n < matroska->num_tracks; n++) {
2829 MatroskaTrack *track = matroska->tracks[n];
2830 av_free(track->codec_id);
2831 av_free(track->codec_name);
2832 av_free(track->codec_priv);
2833 av_free(track->name);
2834 av_free(track->language);
2842 AVInputFormat matroska_demuxer = {
2844 "Matroska file format",
2845 sizeof(MatroskaDemuxContext),
2847 matroska_read_header,
2848 matroska_read_packet,
2849 matroska_read_close,