libavformat/matroskaenc.c

   1 /*
   2  * Matroska muxer
   3  * Copyright (c) 2007 David Conrad
   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 #include "avformat.h"
  23 #include "riff.h"
  24 #include "matroska.h"
  25 #include "avc.h"
  26 #include "libavutil/md5.h"
  27 #include "libavcodec/xiph.h"
  28 #include "libavcodec/mpeg4audio.h"
  29
  30 typedef struct ebml_master {
  31     offset_t        pos;                ///< absolute offset in the file where the master's elements start
  32     int             sizebytes;          ///< how many bytes were reserved for the size
  33 } ebml_master;
  34
  35 typedef struct mkv_seekhead_entry {
  36     unsigned int    elementid;
  37     uint64_t        segmentpos;
  38 } mkv_seekhead_entry;
  39
  40 typedef struct mkv_seekhead {
  41     offset_t                filepos;
  42     offset_t                segment_offset;     ///< the file offset to the beginning of the segment
  43     int                     reserved_size;      ///< -1 if appending to file
  44     int                     max_entries;
  45     mkv_seekhead_entry      *entries;
  46     int                     num_entries;
  47 } mkv_seekhead;
  48
  49 typedef struct {
  50     uint64_t        pts;
  51     int             tracknum;
  52     offset_t        cluster_pos;        ///< file offset of the cluster containing the block
  53 } mkv_cuepoint;
  54
  55 typedef struct {
  56     offset_t        segment_offset;
  57     mkv_cuepoint    *entries;
  58     int             num_entries;
  59 } mkv_cues;
  60
  61 typedef struct MatroskaMuxContext {
  62     ebml_master     segment;
  63     offset_t        segment_offset;
  64     offset_t        segment_uid;
  65     ebml_master     cluster;
  66     offset_t        cluster_pos;        ///< file offset of the current cluster
  67     uint64_t        cluster_pts;
  68     offset_t        duration_offset;
  69     uint64_t        duration;
  70     mkv_seekhead    *main_seekhead;
  71     mkv_seekhead    *cluster_seekhead;
  72     mkv_cues        *cues;
  73
  74     struct AVMD5    *md5_ctx;
  75 } MatroskaMuxContext;
  76
  77
  78 /** 2 bytes * 3 for EBML IDs, 3 1-byte EBML lengths, 8 bytes for 64 bit
  79  * offset, 4 bytes for target EBML ID */
  80 #define MAX_SEEKENTRY_SIZE 21
  81
  82 /** per-cuepoint-track - 3 1-byte EBML IDs, 3 1-byte EBML sizes, 2
  83  * 8-byte uint max */
  84 #define MAX_CUETRACKPOS_SIZE 22
  85
  86 /** per-cuepoint - 2 1-byte EBML IDs, 2 1-byte EBML sizes, 8-byte uint max */
  87 #define MAX_CUEPOINT_SIZE(num_tracks) 12 + MAX_CUETRACKPOS_SIZE*num_tracks
  88
  89
  90 static int ebml_id_size(unsigned int id)
  91 {
  92     return (av_log2(id+1)-1)/7+1;
  93 }
  94
  95 static void put_ebml_id(ByteIOContext *pb, unsigned int id)
  96 {
  97     int i = ebml_id_size(id);
  98     while (i--)
  99         put_byte(pb, id >> (i*8));
 100 }
 101
 102 /**
 103  * Write an EBML size meaning "unknown size".
 104  *
 105  * @param bytes The number of bytes the size should occupy (maximum: 8).
 106  */
 107 static void put_ebml_size_unknown(ByteIOContext *pb, int bytes)
 108 {
 109     assert(bytes <= 8);
 110     put_byte(pb, 0x1ff >> bytes);
 111     while (--bytes)
 112         put_byte(pb, 0xff);
 113 }
 114
 115 /**
 116  * Calculate how many bytes are needed to represent a given number in EBML.
 117  */
 118 static int ebml_num_size(uint64_t num)
 119 {
 120     int bytes = 1;
 121     while ((num+1) >> bytes*7) bytes++;
 122     return bytes;
 123 }
 124
 125 /**
 126  * Write a number in EBML variable length format.
 127  *
 128  * @param bytes The number of bytes that need to be used to write the number.
 129  *              If zero, any number of bytes can be used.
 130  */
 131 static void put_ebml_num(ByteIOContext *pb, uint64_t num, int bytes)
 132 {
 133     int i, needed_bytes = ebml_num_size(num);
 134
 135     // sizes larger than this are currently undefined in EBML
 136     assert(num < (1ULL<<56)-1);
 137
 138     if (bytes == 0)
 139         // don't care how many bytes are used, so use the min
 140         bytes = needed_bytes;
 141     // the bytes needed to write the given size would exceed the bytes
 142     // that we need to use, so write unknown size. This shouldn't happen.
 143     assert(bytes >= needed_bytes);
 144
 145     num |= 1ULL << bytes*7;
 146     for (i = bytes - 1; i >= 0; i--)
 147         put_byte(pb, num >> i*8);
 148 }
 149
 150 static void put_ebml_uint(ByteIOContext *pb, unsigned int elementid, uint64_t val)
 151 {
 152     int i, bytes = 1;
 153     uint64_t tmp = val;
 154     while (tmp>>=8) bytes++;
 155
 156     put_ebml_id(pb, elementid);
 157     put_ebml_num(pb, bytes, 0);
 158     for (i = bytes - 1; i >= 0; i--)
 159         put_byte(pb, val >> i*8);
 160 }
 161
 162 static void put_ebml_float(ByteIOContext *pb, unsigned int elementid, double val)
 163 {
 164     put_ebml_id(pb, elementid);
 165     put_ebml_num(pb, 8, 0);
 166     put_be64(pb, av_dbl2int(val));
 167 }
 168
 169 static void put_ebml_binary(ByteIOContext *pb, unsigned int elementid,
 170                             const uint8_t *buf, int size)
 171 {
 172     put_ebml_id(pb, elementid);
 173     put_ebml_num(pb, size, 0);
 174     put_buffer(pb, buf, size);
 175 }
 176
 177 static void put_ebml_string(ByteIOContext *pb, unsigned int elementid, const char *str)
 178 {
 179     put_ebml_binary(pb, elementid, str, strlen(str));
 180 }
 181
 182 /**
 183  * Writes a void element of a given size. Useful for reserving space in
 184  * the file to be written to later.
 185  *
 186  * @param size The number of bytes to reserve, which must be at least 2.
 187  */
 188 static void put_ebml_void(ByteIOContext *pb, uint64_t size)
 189 {
 190     offset_t currentpos = url_ftell(pb);
 191
 192     assert(size >= 2);
 193
 194     put_ebml_id(pb, EBML_ID_VOID);
 195     // we need to subtract the length needed to store the size from the
 196     // size we need to reserve so 2 cases, we use 8 bytes to store the
 197     // size if possible, 1 byte otherwise
 198     if (size < 10)
 199         put_ebml_num(pb, size-1, 0);
 200     else
 201         put_ebml_num(pb, size-9, 8);
 202     while(url_ftell(pb) < currentpos + size)
 203         put_byte(pb, 0);
 204 }
 205
 206 static ebml_master start_ebml_master(ByteIOContext *pb, unsigned int elementid, uint64_t expectedsize)
 207 {
 208     int bytes = expectedsize ? ebml_num_size(expectedsize) : 8;
 209     put_ebml_id(pb, elementid);
 210     put_ebml_size_unknown(pb, bytes);
 211     return (ebml_master){ url_ftell(pb), bytes };
 212 }
 213
 214 static void end_ebml_master(ByteIOContext *pb, ebml_master master)
 215 {
 216     offset_t pos = url_ftell(pb);
 217
 218     // leave the unknown size for masters when streaming
 219     if (url_is_streamed(pb))
 220         return;
 221
 222     url_fseek(pb, master.pos - master.sizebytes, SEEK_SET);
 223     put_ebml_num(pb, pos - master.pos, master.sizebytes);
 224     url_fseek(pb, pos, SEEK_SET);
 225 }
 226
 227 static void put_xiph_size(ByteIOContext *pb, int size)
 228 {
 229     int i;
 230     for (i = 0; i < size / 255; i++)
 231         put_byte(pb, 255);
 232     put_byte(pb, size % 255);
 233 }
 234
 235 /**
 236  * Initialize a mkv_seekhead element to be ready to index level 1 Matroska
 237  * elements. If a maximum number of elements is specified, enough space
 238  * will be reserved at the current file location to write a seek head of
 239  * that size.
 240  *
 241  * @param segment_offset The absolute offset to the position in the file
 242  *                       where the segment begins.
 243  * @param numelements The maximum number of elements that will be indexed
 244  *                    by this seek head, 0 if unlimited.
 245  */
 246 static mkv_seekhead * mkv_start_seekhead(ByteIOContext *pb, offset_t segment_offset, int numelements)
 247 {
 248     mkv_seekhead *new_seekhead = av_mallocz(sizeof(mkv_seekhead));
 249     if (new_seekhead == NULL)
 250         return NULL;
 251
 252     new_seekhead->segment_offset = segment_offset;
 253
 254     if (numelements > 0) {
 255         new_seekhead->filepos = url_ftell(pb);
 256         // 21 bytes max for a seek entry, 10 bytes max for the SeekHead ID
 257         // and size, and 3 bytes to guarantee that an EBML void element
 258         // will fit afterwards
 259         new_seekhead->reserved_size = numelements * MAX_SEEKENTRY_SIZE + 13;
 260         new_seekhead->max_entries = numelements;
 261         put_ebml_void(pb, new_seekhead->reserved_size);
 262     }
 263     return new_seekhead;
 264 }
 265
 266 static int mkv_add_seekhead_entry(mkv_seekhead *seekhead, unsigned int elementid, uint64_t filepos)
 267 {
 268     mkv_seekhead_entry *entries = seekhead->entries;
 269
 270     // don't store more elements than we reserved space for
 271     if (seekhead->max_entries > 0 && seekhead->max_entries <= seekhead->num_entries)
 272         return -1;
 273
 274     entries = av_realloc(entries, (seekhead->num_entries + 1) * sizeof(mkv_seekhead_entry));
 275     if (entries == NULL)
 276         return AVERROR(ENOMEM);
 277
 278     entries[seekhead->num_entries  ].elementid = elementid;
 279     entries[seekhead->num_entries++].segmentpos = filepos - seekhead->segment_offset;
 280
 281     seekhead->entries = entries;
 282     return 0;
 283 }
 284
 285 /**
 286  * Write the seek head to the file and free it. If a maximum number of
 287  * elements was specified to mkv_start_seekhead(), the seek head will
 288  * be written at the location reserved for it. Otherwise, it is written
 289  * at the current location in the file.
 290  *
 291  * @return The file offset where the seekhead was written.
 292  */
 293 static offset_t mkv_write_seekhead(ByteIOContext *pb, mkv_seekhead *seekhead)
 294 {
 295     ebml_master metaseek, seekentry;
 296     offset_t currentpos;
 297     int i;
 298
 299     currentpos = url_ftell(pb);
 300
 301     if (seekhead->reserved_size > 0)
 302         url_fseek(pb, seekhead->filepos, SEEK_SET);
 303
 304     metaseek = start_ebml_master(pb, MATROSKA_ID_SEEKHEAD, seekhead->reserved_size);
 305     for (i = 0; i < seekhead->num_entries; i++) {
 306         mkv_seekhead_entry *entry = &seekhead->entries[i];
 307
 308         seekentry = start_ebml_master(pb, MATROSKA_ID_SEEKENTRY, MAX_SEEKENTRY_SIZE);
 309
 310         put_ebml_id(pb, MATROSKA_ID_SEEKID);
 311         put_ebml_num(pb, ebml_id_size(entry->elementid), 0);
 312         put_ebml_id(pb, entry->elementid);
 313
 314         put_ebml_uint(pb, MATROSKA_ID_SEEKPOSITION, entry->segmentpos);
 315         end_ebml_master(pb, seekentry);
 316     }
 317     end_ebml_master(pb, metaseek);
 318
 319     if (seekhead->reserved_size > 0) {
 320         uint64_t remaining = seekhead->filepos + seekhead->reserved_size - url_ftell(pb);
 321         put_ebml_void(pb, remaining);
 322         url_fseek(pb, currentpos, SEEK_SET);
 323
 324         currentpos = seekhead->filepos;
 325     }
 326     av_free(seekhead->entries);
 327     av_free(seekhead);
 328
 329     return currentpos;
 330 }
 331
 332 static mkv_cues * mkv_start_cues(offset_t segment_offset)
 333 {
 334     mkv_cues *cues = av_mallocz(sizeof(mkv_cues));
 335     if (cues == NULL)
 336         return NULL;
 337
 338     cues->segment_offset = segment_offset;
 339     return cues;
 340 }
 341
 342 static int mkv_add_cuepoint(mkv_cues *cues, AVPacket *pkt, offset_t cluster_pos)
 343 {
 344     mkv_cuepoint *entries = cues->entries;
 345
 346     entries = av_realloc(entries, (cues->num_entries + 1) * sizeof(mkv_cuepoint));
 347     if (entries == NULL)
 348         return AVERROR(ENOMEM);
 349
 350     entries[cues->num_entries  ].pts = pkt->pts;
 351     entries[cues->num_entries  ].tracknum = pkt->stream_index + 1;
 352     entries[cues->num_entries++].cluster_pos = cluster_pos - cues->segment_offset;
 353
 354     cues->entries = entries;
 355     return 0;
 356 }
 357
 358 static offset_t mkv_write_cues(ByteIOContext *pb, mkv_cues *cues, int num_tracks)
 359 {
 360     ebml_master cues_element;
 361     offset_t currentpos;
 362     int i, j;
 363
 364     currentpos = url_ftell(pb);
 365     cues_element = start_ebml_master(pb, MATROSKA_ID_CUES, 0);
 366
 367     for (i = 0; i < cues->num_entries; i++) {
 368         ebml_master cuepoint, track_positions;
 369         mkv_cuepoint *entry = &cues->entries[i];
 370         uint64_t pts = entry->pts;
 371
 372         cuepoint = start_ebml_master(pb, MATROSKA_ID_POINTENTRY, MAX_CUEPOINT_SIZE(num_tracks));
 373         put_ebml_uint(pb, MATROSKA_ID_CUETIME, pts);
 374
 375         // put all the entries from different tracks that have the exact same
 376         // timestamp into the same CuePoint
 377         for (j = 0; j < cues->num_entries - i && entry[j].pts == pts; j++) {
 378             track_positions = start_ebml_master(pb, MATROSKA_ID_CUETRACKPOSITION, MAX_CUETRACKPOS_SIZE);
 379             put_ebml_uint(pb, MATROSKA_ID_CUETRACK          , entry[j].tracknum   );
 380             put_ebml_uint(pb, MATROSKA_ID_CUECLUSTERPOSITION, entry[j].cluster_pos);
 381             end_ebml_master(pb, track_positions);
 382         }
 383         i += j - 1;
 384         end_ebml_master(pb, cuepoint);
 385     }
 386     end_ebml_master(pb, cues_element);
 387
 388     av_free(cues->entries);
 389     av_free(cues);
 390     return currentpos;
 391 }
 392
 393 static int put_xiph_codecpriv(AVFormatContext *s, ByteIOContext *pb, AVCodecContext *codec)
 394 {
 395     uint8_t *header_start[3];
 396     int header_len[3];
 397     int first_header_size;
 398     int j;
 399
 400     if (codec->codec_id == CODEC_ID_VORBIS)
 401         first_header_size = 30;
 402     else
 403         first_header_size = 42;
 404
 405     if (ff_split_xiph_headers(codec->extradata, codec->extradata_size,
 406                               first_header_size, header_start, header_len) < 0) {
 407         av_log(s, AV_LOG_ERROR, "Extradata corrupt.\n");
 408         return -1;
 409     }
 410
 411     put_byte(pb, 2);                    // number packets - 1
 412     for (j = 0; j < 2; j++) {
 413         put_xiph_size(pb, header_len[j]);
 414     }
 415     for (j = 0; j < 3; j++)
 416         put_buffer(pb, header_start[j], header_len[j]);
 417
 418     return 0;
 419 }
 420
 421 #define FLAC_STREAMINFO_SIZE 34
 422
 423 static int put_flac_codecpriv(AVFormatContext *s, ByteIOContext *pb, AVCodecContext *codec)
 424 {
 425     // if the extradata_size is greater than FLAC_STREAMINFO_SIZE,
 426     // assume that it's in Matroska's format already
 427     if (codec->extradata_size < FLAC_STREAMINFO_SIZE) {
 428         av_log(s, AV_LOG_ERROR, "Invalid FLAC extradata\n");
 429         return -1;
 430     } else if (codec->extradata_size == FLAC_STREAMINFO_SIZE) {
 431         // only the streaminfo packet
 432         put_buffer(pb, "fLaC", 4);
 433         put_byte(pb, 0x80);
 434         put_be24(pb, FLAC_STREAMINFO_SIZE);
 435     } else if(memcmp("fLaC", codec->extradata, 4)) {
 436         av_log(s, AV_LOG_ERROR, "Invalid FLAC extradata\n");
 437         return -1;
 438     }
 439     put_buffer(pb, codec->extradata, codec->extradata_size);
 440     return 0;
 441 }
 442
 443 static void get_aac_sample_rates(AVFormatContext *s, AVCodecContext *codec, int *sample_rate, int *output_sample_rate)
 444 {
 445     int sri;
 446
 447     if (codec->extradata_size < 2) {
 448         av_log(s, AV_LOG_WARNING, "No AAC extradata, unable to determine samplerate.\n");
 449         return;
 450     }
 451
 452     sri = ((codec->extradata[0] << 1) & 0xE) | (codec->extradata[1] >> 7);
 453     if (sri > 12) {
 454         av_log(s, AV_LOG_WARNING, "AAC samplerate index out of bounds\n");
 455         return;
 456     }
 457     *sample_rate = ff_mpeg4audio_sample_rates[sri];
 458
 459     // if sbr, get output sample rate as well
 460     if (codec->extradata_size == 5) {
 461         sri = (codec->extradata[4] >> 3) & 0xF;
 462         if (sri > 12) {
 463             av_log(s, AV_LOG_WARNING, "AAC output samplerate index out of bounds\n");
 464             return;
 465         }
 466         *output_sample_rate = ff_mpeg4audio_sample_rates[sri];
 467     }
 468 }
 469
 470 static int mkv_write_codecprivate(AVFormatContext *s, ByteIOContext *pb, AVCodecContext *codec, int native_id)
 471 {
 472     ByteIOContext *dyn_cp;
 473     uint8_t *codecpriv;
 474     int ret, codecpriv_size;
 475
 476     ret = url_open_dyn_buf(&dyn_cp);
 477     if(ret < 0)
 478         return ret;
 479
 480     if (native_id) {
 481         if (codec->codec_id == CODEC_ID_VORBIS || codec->codec_id == CODEC_ID_THEORA)
 482             ret = put_xiph_codecpriv(s, dyn_cp, codec);
 483         else if (codec->codec_id == CODEC_ID_FLAC)
 484             ret = put_flac_codecpriv(s, dyn_cp, codec);
 485         else if (codec->codec_id == CODEC_ID_H264)
 486             ret = ff_isom_write_avcc(dyn_cp, codec->extradata, codec->extradata_size);
 487         else if (codec->extradata_size)
 488             put_buffer(dyn_cp, codec->extradata, codec->extradata_size);
 489     } else if (codec->codec_type == CODEC_TYPE_VIDEO) {
 490         if (!codec->codec_tag)
 491             codec->codec_tag = codec_get_tag(codec_bmp_tags, codec->codec_id);
 492         if (!codec->codec_tag) {
 493             av_log(s, AV_LOG_ERROR, "No bmp codec ID found.");
 494             ret = -1;
 495         }
 496
 497         put_bmp_header(dyn_cp, codec, codec_bmp_tags, 0);
 498
 499     } else if (codec->codec_type == CODEC_TYPE_AUDIO) {
 500         if (!codec->codec_tag)
 501             codec->codec_tag = codec_get_tag(codec_wav_tags, codec->codec_id);
 502         if (!codec->codec_tag) {
 503             av_log(s, AV_LOG_ERROR, "No wav codec ID found.");
 504             ret = -1;
 505         }
 506
 507         put_wav_header(dyn_cp, codec);
 508     }
 509
 510     codecpriv_size = url_close_dyn_buf(dyn_cp, &codecpriv);
 511     if (codecpriv_size)
 512         put_ebml_binary(pb, MATROSKA_ID_CODECPRIVATE, codecpriv, codecpriv_size);
 513     av_free(codecpriv);
 514     return ret;
 515 }
 516
 517 static int mkv_write_tracks(AVFormatContext *s)
 518 {
 519     MatroskaMuxContext *mkv = s->priv_data;
 520     ByteIOContext *pb = s->pb;
 521     ebml_master tracks;
 522     int i, j, ret;
 523
 524     ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_TRACKS, url_ftell(pb));
 525     if (ret < 0) return ret;
 526
 527     tracks = start_ebml_master(pb, MATROSKA_ID_TRACKS, 0);
 528     for (i = 0; i < s->nb_streams; i++) {
 529         AVStream *st = s->streams[i];
 530         AVCodecContext *codec = st->codec;
 531         ebml_master subinfo, track;
 532         int native_id = 0;
 533         int bit_depth = av_get_bits_per_sample(codec->codec_id);
 534         int sample_rate = codec->sample_rate;
 535         int output_sample_rate = 0;
 536
 537         if (!bit_depth)
 538             bit_depth = av_get_bits_per_sample_format(codec->sample_fmt);
 539
 540         if (codec->codec_id == CODEC_ID_AAC)
 541             get_aac_sample_rates(s, codec, &sample_rate, &output_sample_rate);
 542
 543         track = start_ebml_master(pb, MATROSKA_ID_TRACKENTRY, 0);
 544         put_ebml_uint (pb, MATROSKA_ID_TRACKNUMBER     , i + 1);
 545         put_ebml_uint (pb, MATROSKA_ID_TRACKUID        , i + 1);
 546         put_ebml_uint (pb, MATROSKA_ID_TRACKFLAGLACING , 0);    // no lacing (yet)
 547
 548         if (st->language[0])
 549             put_ebml_string(pb, MATROSKA_ID_TRACKLANGUAGE, st->language);
 550         else
 551             put_ebml_string(pb, MATROSKA_ID_TRACKLANGUAGE, "und");
 552
 553         put_ebml_uint(pb, MATROSKA_ID_TRACKFLAGDEFAULT, !!(st->disposition & AV_DISPOSITION_DEFAULT));
 554
 555         // look for a codec ID string specific to mkv to use,
 556         // if none are found, use AVI codes
 557         for (j = 0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++) {
 558             if (ff_mkv_codec_tags[j].id == codec->codec_id) {
 559                 put_ebml_string(pb, MATROSKA_ID_CODECID, ff_mkv_codec_tags[j].str);
 560                 native_id = 1;
 561                 break;
 562             }
 563         }
 564
 565         switch (codec->codec_type) {
 566             case CODEC_TYPE_VIDEO:
 567                 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_VIDEO);
 568
 569                 if (!native_id)
 570                     // if there is no mkv-specific codec ID, use VFW mode
 571                     put_ebml_string(pb, MATROSKA_ID_CODECID, MATROSKA_CODEC_ID_VIDEO_VFW_FOURCC);
 572
 573                 subinfo = start_ebml_master(pb, MATROSKA_ID_TRACKVIDEO, 0);
 574                 // XXX: interlace flag?
 575                 put_ebml_uint (pb, MATROSKA_ID_VIDEOPIXELWIDTH , codec->width);
 576                 put_ebml_uint (pb, MATROSKA_ID_VIDEOPIXELHEIGHT, codec->height);
 577                 if (codec->sample_aspect_ratio.num) {
 578                     AVRational dar = av_mul_q(codec->sample_aspect_ratio,
 579                                     (AVRational){codec->width, codec->height});
 580                     put_ebml_uint(pb, MATROSKA_ID_VIDEODISPLAYWIDTH , dar.num);
 581                     put_ebml_uint(pb, MATROSKA_ID_VIDEODISPLAYHEIGHT, dar.den);
 582                 }
 583                 end_ebml_master(pb, subinfo);
 584                 break;
 585
 586             case CODEC_TYPE_AUDIO:
 587                 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_AUDIO);
 588
 589                 if (!native_id)
 590                     // no mkv-specific ID, use ACM mode
 591                     put_ebml_string(pb, MATROSKA_ID_CODECID, MATROSKA_CODEC_ID_AUDIO_ACM);
 592
 593                 subinfo = start_ebml_master(pb, MATROSKA_ID_TRACKAUDIO, 0);
 594                 put_ebml_uint  (pb, MATROSKA_ID_AUDIOCHANNELS    , codec->channels);
 595                 put_ebml_float (pb, MATROSKA_ID_AUDIOSAMPLINGFREQ, sample_rate);
 596                 if (output_sample_rate)
 597                     put_ebml_float(pb, MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, output_sample_rate);
 598                 if (bit_depth)
 599                     put_ebml_uint(pb, MATROSKA_ID_AUDIOBITDEPTH, bit_depth);
 600                 end_ebml_master(pb, subinfo);
 601                 break;
 602
 603             case CODEC_TYPE_SUBTITLE:
 604                 put_ebml_uint(pb, MATROSKA_ID_TRACKTYPE, MATROSKA_TRACK_TYPE_SUBTITLE);
 605                 break;
 606             default:
 607                 av_log(s, AV_LOG_ERROR, "Only audio, video, and subtitles are supported for Matroska.");
 608                 break;
 609         }
 610         ret = mkv_write_codecprivate(s, pb, codec, native_id);
 611         if (ret < 0) return ret;
 612
 613         end_ebml_master(pb, track);
 614
 615         // ms precision is the de-facto standard timescale for mkv files
 616         av_set_pts_info(st, 64, 1, 1000);
 617     }
 618     end_ebml_master(pb, tracks);
 619     return 0;
 620 }
 621
 622 static int mkv_write_header(AVFormatContext *s)
 623 {
 624     MatroskaMuxContext *mkv = s->priv_data;
 625     ByteIOContext *pb = s->pb;
 626     ebml_master ebml_header, segment_info;
 627     int ret;
 628
 629     mkv->md5_ctx = av_mallocz(av_md5_size);
 630     av_md5_init(mkv->md5_ctx);
 631
 632     ebml_header = start_ebml_master(pb, EBML_ID_HEADER, 0);
 633     put_ebml_uint   (pb, EBML_ID_EBMLVERSION        ,           1);
 634     put_ebml_uint   (pb, EBML_ID_EBMLREADVERSION    ,           1);
 635     put_ebml_uint   (pb, EBML_ID_EBMLMAXIDLENGTH    ,           4);
 636     put_ebml_uint   (pb, EBML_ID_EBMLMAXSIZELENGTH  ,           8);
 637     put_ebml_string (pb, EBML_ID_DOCTYPE            ,  "matroska");
 638     put_ebml_uint   (pb, EBML_ID_DOCTYPEVERSION     ,           2);
 639     put_ebml_uint   (pb, EBML_ID_DOCTYPEREADVERSION ,           2);
 640     end_ebml_master(pb, ebml_header);
 641
 642     mkv->segment = start_ebml_master(pb, MATROSKA_ID_SEGMENT, 0);
 643     mkv->segment_offset = url_ftell(pb);
 644
 645     // we write 2 seek heads - one at the end of the file to point to each
 646     // cluster, and one at the beginning to point to all other level one
 647     // elements (including the seek head at the end of the file), which
 648     // isn't more than 10 elements if we only write one of each other
 649     // currently defined level 1 element
 650     mkv->main_seekhead    = mkv_start_seekhead(pb, mkv->segment_offset, 10);
 651     mkv->cluster_seekhead = mkv_start_seekhead(pb, mkv->segment_offset, 0);
 652     if (mkv->main_seekhead == NULL || mkv->cluster_seekhead == NULL)
 653         return AVERROR(ENOMEM);
 654
 655     ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_INFO, url_ftell(pb));
 656     if (ret < 0) return ret;
 657
 658     segment_info = start_ebml_master(pb, MATROSKA_ID_INFO, 0);
 659     put_ebml_uint(pb, MATROSKA_ID_TIMECODESCALE, 1000000);
 660     if (strlen(s->title))
 661         put_ebml_string(pb, MATROSKA_ID_TITLE, s->title);
 662     if (!(s->streams[0]->codec->flags & CODEC_FLAG_BITEXACT)) {
 663         put_ebml_string(pb, MATROSKA_ID_MUXINGAPP , LIBAVFORMAT_IDENT);
 664         put_ebml_string(pb, MATROSKA_ID_WRITINGAPP, LIBAVFORMAT_IDENT);
 665
 666         // reserve space to write the segment UID later
 667         mkv->segment_uid = url_ftell(pb);
 668         put_ebml_void(pb, 19);
 669     }
 670
 671     // reserve space for the duration
 672     mkv->duration = 0;
 673     mkv->duration_offset = url_ftell(pb);
 674     put_ebml_void(pb, 11);                  // assumes double-precision float to be written
 675     end_ebml_master(pb, segment_info);
 676
 677     ret = mkv_write_tracks(s);
 678     if (ret < 0) return ret;
 679
 680     ret = mkv_add_seekhead_entry(mkv->cluster_seekhead, MATROSKA_ID_CLUSTER, url_ftell(pb));
 681     if (ret < 0) return ret;
 682
 683     mkv->cluster_pos = url_ftell(pb);
 684     mkv->cluster = start_ebml_master(pb, MATROSKA_ID_CLUSTER, 0);
 685     put_ebml_uint(pb, MATROSKA_ID_CLUSTERTIMECODE, 0);
 686     mkv->cluster_pts = 0;
 687
 688     mkv->cues = mkv_start_cues(mkv->segment_offset);
 689     if (mkv->cues == NULL)
 690         return AVERROR(ENOMEM);
 691
 692     return 0;
 693 }
 694
 695 static int mkv_block_size(AVPacket *pkt)
 696 {
 697     int size = 4;           // track num + timecode + flags
 698     return size + pkt->size;
 699 }
 700
 701 static int mkv_blockgroup_size(AVPacket *pkt)
 702 {
 703     int size = mkv_block_size(pkt);
 704     size += ebml_num_size(size);
 705     size += 2;              // EBML ID for block and block duration
 706     size += 8;              // max size of block duration
 707     size += ebml_num_size(size);
 708     size += 1;              // blockgroup EBML ID
 709     return size;
 710 }
 711
 712 static void mkv_write_block(AVFormatContext *s, unsigned int blockid, AVPacket *pkt, int flags)
 713 {
 714     MatroskaMuxContext *mkv = s->priv_data;
 715     ByteIOContext *pb = s->pb;
 716
 717     av_log(s, AV_LOG_DEBUG, "Writing block at offset %" PRIu64 ", size %d, "
 718            "pts %" PRId64 ", dts %" PRId64 ", duration %d, flags %d\n",
 719            url_ftell(pb), pkt->size, pkt->pts, pkt->dts, pkt->duration, flags);
 720     put_ebml_id(pb, blockid);
 721     put_ebml_num(pb, mkv_block_size(pkt), 0);
 722     put_byte(pb, 0x80 | (pkt->stream_index + 1));     // this assumes stream_index is less than 126
 723     put_be16(pb, pkt->pts - mkv->cluster_pts);
 724     put_byte(pb, flags);
 725     put_buffer(pb, pkt->data, pkt->size);
 726 }
 727
 728 static int mkv_write_packet(AVFormatContext *s, AVPacket *pkt)
 729 {
 730     MatroskaMuxContext *mkv = s->priv_data;
 731     ByteIOContext *pb = s->pb;
 732     AVCodecContext *codec = s->streams[pkt->stream_index]->codec;
 733     int keyframe = !!(pkt->flags & PKT_FLAG_KEY);
 734     int ret;
 735
 736     // start a new cluster every 5 MB or 5 sec
 737     if (url_ftell(pb) > mkv->cluster_pos + 5*1024*1024 || pkt->pts > mkv->cluster_pts + 5000) {
 738         av_log(s, AV_LOG_DEBUG, "Starting new cluster at offset %" PRIu64
 739                " bytes, pts %" PRIu64 "\n", url_ftell(pb), pkt->pts);
 740         end_ebml_master(pb, mkv->cluster);
 741
 742         ret = mkv_add_seekhead_entry(mkv->cluster_seekhead, MATROSKA_ID_CLUSTER, url_ftell(pb));
 743         if (ret < 0) return ret;
 744
 745         mkv->cluster_pos = url_ftell(pb);
 746         mkv->cluster = start_ebml_master(pb, MATROSKA_ID_CLUSTER, 0);
 747         put_ebml_uint(pb, MATROSKA_ID_CLUSTERTIMECODE, pkt->pts);
 748         mkv->cluster_pts = pkt->pts;
 749         av_md5_update(mkv->md5_ctx, pkt->data, FFMIN(200, pkt->size));
 750     }
 751
 752     if (codec->codec_id == CODEC_ID_H264 &&
 753         codec->extradata_size > 0 && AV_RB32(codec->extradata) == 0x00000001) {
 754         /* from x264 or from bytestream h264 */
 755         /* nal reformating needed */
 756         int ret = ff_avc_parse_nal_units(pkt->data, &pkt->data, &pkt->size);
 757         if (ret < 0)
 758             return ret;
 759         assert(pkt->size);
 760     }
 761
 762     if (codec->codec_type != CODEC_TYPE_SUBTITLE) {
 763         mkv_write_block(s, MATROSKA_ID_SIMPLEBLOCK, pkt, keyframe << 7);
 764     } else {
 765         ebml_master blockgroup = start_ebml_master(pb, MATROSKA_ID_BLOCKGROUP, mkv_blockgroup_size(pkt));
 766         mkv_write_block(s, MATROSKA_ID_BLOCK, pkt, 0);
 767         put_ebml_uint(pb, MATROSKA_ID_DURATION, pkt->duration);
 768         end_ebml_master(pb, blockgroup);
 769     }
 770
 771     if (codec->codec_type == CODEC_TYPE_VIDEO && keyframe) {
 772         ret = mkv_add_cuepoint(mkv->cues, pkt, mkv->cluster_pos);
 773         if (ret < 0) return ret;
 774     }
 775
 776     mkv->duration = FFMAX(mkv->duration, pkt->pts + pkt->duration);
 777     return 0;
 778 }
 779
 780 static int mkv_write_trailer(AVFormatContext *s)
 781 {
 782     MatroskaMuxContext *mkv = s->priv_data;
 783     ByteIOContext *pb = s->pb;
 784     offset_t currentpos, second_seekhead, cuespos;
 785     int ret;
 786
 787     end_ebml_master(pb, mkv->cluster);
 788
 789     if (!url_is_streamed(pb)) {
 790         cuespos = mkv_write_cues(pb, mkv->cues, s->nb_streams);
 791         second_seekhead = mkv_write_seekhead(pb, mkv->cluster_seekhead);
 792
 793         ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_CUES    , cuespos);
 794         if (ret < 0) return ret;
 795         ret = mkv_add_seekhead_entry(mkv->main_seekhead, MATROSKA_ID_SEEKHEAD, second_seekhead);
 796         if (ret < 0) return ret;
 797         mkv_write_seekhead(pb, mkv->main_seekhead);
 798
 799         // update the duration
 800         av_log(s, AV_LOG_DEBUG, "end duration = %" PRIu64 "\n", mkv->duration);
 801         currentpos = url_ftell(pb);
 802         url_fseek(pb, mkv->duration_offset, SEEK_SET);
 803         put_ebml_float(pb, MATROSKA_ID_DURATION, mkv->duration);
 804
 805         // write the md5sum of some frames as the segment UID
 806         if (!(s->streams[0]->codec->flags & CODEC_FLAG_BITEXACT)) {
 807             uint8_t segment_uid[16];
 808             av_md5_final(mkv->md5_ctx, segment_uid);
 809             url_fseek(pb, mkv->segment_uid, SEEK_SET);
 810             put_ebml_binary(pb, MATROSKA_ID_SEGMENTUID, segment_uid, 16);
 811         }
 812         url_fseek(pb, currentpos, SEEK_SET);
 813     }
 814
 815     end_ebml_master(pb, mkv->segment);
 816     av_free(mkv->md5_ctx);
 817     return 0;
 818 }
 819
 820 AVOutputFormat matroska_muxer = {
 821     "matroska",
 822     "Matroska File Format",
 823     "video/x-matroska",
 824     "mkv",
 825     sizeof(MatroskaMuxContext),
 826     CODEC_ID_MP2,
 827     CODEC_ID_MPEG4,
 828     mkv_write_header,
 829     mkv_write_packet,
 830     mkv_write_trailer,
 831     .codec_tag = (const AVCodecTag*[]){codec_bmp_tags, codec_wav_tags, 0},
 832     .subtitle_codec = CODEC_ID_TEXT,
 833 };
 834
 835 AVOutputFormat matroska_audio_muxer = {
 836     "matroska",
 837     "Matroska File Format",
 838     "audio/x-matroska",
 839     "mka",
 840     sizeof(MatroskaMuxContext),
 841     CODEC_ID_MP2,
 842     CODEC_ID_NONE,
 843     mkv_write_header,
 844     mkv_write_packet,
 845     mkv_write_trailer,
 846     .codec_tag = (const AVCodecTag*[]){codec_wav_tags, 0},
 847 };