git.sesse.net Git - ffmpeg/blob - tools/ismindex.c

   1 /*
   2  * Copyright (c) 2012 Martin Storsjo
   3  *
   4  * This file is part of Libav.
   5  *
   6  * Libav is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * Libav is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with Libav; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 /*
  22  * To create a simple file for smooth streaming:
  23  * ffmpeg <normal input/transcoding options> -movflags frag_keyframe foo.ismv
  24  * ismindex -n foo foo.ismv
  25  * This step creates foo.ism and foo.ismc that is required by IIS for
  26  * serving it.
  27  *
  28  * To pre-split files for serving as static files by a web server without
  29  * any extra server support, create the ismv file as above, and split it:
  30  * ismindex -split foo.ismv
  31  * This step creates a file Manifest and directories QualityLevel(...),
  32  * that can be read directly by a smooth streaming player.
  33  */
  34
  35 #include <stdio.h>
  36 #include <string.h>
  37 #include <sys/stat.h>
  38 #ifdef _WIN32
  39 #include <direct.h>
  40 #define mkdir(a, b) _mkdir(a)
  41 #endif
  42
  43 #include "cmdutils.h"
  44
  45 #include "libavformat/avformat.h"
  46 #include "libavutil/intreadwrite.h"
  47 #include "libavutil/mathematics.h"
  48
  49 static int usage(const char *argv0, int ret)
  50 {
  51     fprintf(stderr, "%s [-split] [-n basename] file1 [file2] ...\n", argv0);
  52     return ret;
  53 }
  54
  55 struct MoofOffset {
  56     int64_t time;
  57     int64_t offset;
  58     int duration;
  59 };
  60
  61 struct Track {
  62     const char *name;
  63     int64_t duration;
  64     int bitrate;
  65     int track_id;
  66     int is_audio, is_video;
  67     int width, height;
  68     int chunks;
  69     int sample_rate, channels;
  70     uint8_t *codec_private;
  71     int codec_private_size;
  72     struct MoofOffset *offsets;
  73     int timescale;
  74     const char *fourcc;
  75     int blocksize;
  76     int tag;
  77 };
  78
  79 struct Tracks {
  80     int nb_tracks;
  81     int64_t duration;
  82     struct Track **tracks;
  83     int video_track, audio_track;
  84     int nb_video_tracks, nb_audio_tracks;
  85 };
  86
  87 static int copy_tag(AVIOContext *in, AVIOContext *out, int32_t tag_name)
  88 {
  89     int32_t size, tag;
  90
  91     size = avio_rb32(in);
  92     tag  = avio_rb32(in);
  93     avio_wb32(out, size);
  94     avio_wb32(out, tag);
  95     if (tag != tag_name)
  96         return -1;
  97     size -= 8;
  98     while (size > 0) {
  99         char buf[1024];
 100         int len = FFMIN(sizeof(buf), size);
 101         if (avio_read(in, buf, len) != len)
 102             break;
 103         avio_write(out, buf, len);
 104         size -= len;
 105     }
 106     return 0;
 107 }
 108
 109 static int write_fragment(const char *filename, AVIOContext *in)
 110 {
 111     AVIOContext *out = NULL;
 112     int ret;
 113
 114     if ((ret = avio_open2(&out, filename, AVIO_FLAG_WRITE, NULL, NULL)) < 0)
 115         return ret;
 116     copy_tag(in, out, MKBETAG('m', 'o', 'o', 'f'));
 117     copy_tag(in, out, MKBETAG('m', 'd', 'a', 't'));
 118
 119     avio_flush(out);
 120     avio_close(out);
 121
 122     return ret;
 123 }
 124
 125 static int write_fragments(struct Tracks *tracks, int start_index,
 126                            AVIOContext *in)
 127 {
 128     char dirname[100], filename[500];
 129     int i, j;
 130
 131     for (i = start_index; i < tracks->nb_tracks; i++) {
 132         struct Track *track = tracks->tracks[i];
 133         const char *type    = track->is_video ? "video" : "audio";
 134         snprintf(dirname, sizeof(dirname), "QualityLevels(%d)", track->bitrate);
 135         mkdir(dirname, 0777);
 136         for (j = 0; j < track->chunks; j++) {
 137             snprintf(filename, sizeof(filename), "%s/Fragments(%s=%"PRId64")",
 138                      dirname, type, track->offsets[j].time);
 139             avio_seek(in, track->offsets[j].offset, SEEK_SET);
 140             write_fragment(filename, in);
 141         }
 142     }
 143     return 0;
 144 }
 145
 146 static int read_tfra(struct Tracks *tracks, int start_index, AVIOContext *f)
 147 {
 148     int ret = AVERROR_EOF, track_id;
 149     int version, fieldlength, i, j;
 150     int64_t pos   = avio_tell(f);
 151     uint32_t size = avio_rb32(f);
 152     struct Track *track = NULL;
 153
 154     if (avio_rb32(f) != MKBETAG('t', 'f', 'r', 'a'))
 155         goto fail;
 156     version = avio_r8(f);
 157     avio_rb24(f);
 158     track_id = avio_rb32(f); /* track id */
 159     for (i = start_index; i < tracks->nb_tracks && !track; i++)
 160         if (tracks->tracks[i]->track_id == track_id)
 161             track = tracks->tracks[i];
 162     if (!track) {
 163         /* Ok, continue parsing the next atom */
 164         ret = 0;
 165         goto fail;
 166     }
 167     fieldlength = avio_rb32(f);
 168     track->chunks  = avio_rb32(f);
 169     track->offsets = av_mallocz(sizeof(*track->offsets) * track->chunks);
 170     if (!track->offsets) {
 171         ret = AVERROR(ENOMEM);
 172         goto fail;
 173     }
 174     for (i = 0; i < track->chunks; i++) {
 175         if (version == 1) {
 176             track->offsets[i].time   = avio_rb64(f);
 177             track->offsets[i].offset = avio_rb64(f);
 178         } else {
 179             track->offsets[i].time   = avio_rb32(f);
 180             track->offsets[i].offset = avio_rb32(f);
 181         }
 182         for (j = 0; j < ((fieldlength >> 4) & 3) + 1; j++)
 183             avio_r8(f);
 184         for (j = 0; j < ((fieldlength >> 2) & 3) + 1; j++)
 185             avio_r8(f);
 186         for (j = 0; j < ((fieldlength >> 0) & 3) + 1; j++)
 187             avio_r8(f);
 188         if (i > 0)
 189             track->offsets[i - 1].duration = track->offsets[i].time -
 190                                              track->offsets[i - 1].time;
 191     }
 192     if (track->chunks > 0)
 193         track->offsets[track->chunks - 1].duration = track->duration -
 194                                                      track->offsets[track->chunks - 1].time;
 195     ret = 0;
 196
 197 fail:
 198     avio_seek(f, pos + size, SEEK_SET);
 199     return ret;
 200 }
 201
 202 static int read_mfra(struct Tracks *tracks, int start_index,
 203                      const char *file, int split)
 204 {
 205     int err = 0;
 206     AVIOContext *f = NULL;
 207     int32_t mfra_size;
 208
 209     if ((err = avio_open2(&f, file, AVIO_FLAG_READ, NULL, NULL)) < 0)
 210         goto fail;
 211     avio_seek(f, avio_size(f) - 4, SEEK_SET);
 212     mfra_size = avio_rb32(f);
 213     avio_seek(f, -mfra_size, SEEK_CUR);
 214     if (avio_rb32(f) != mfra_size) {
 215         err = AVERROR_INVALIDDATA;
 216         goto fail;
 217     }
 218     if (avio_rb32(f) != MKBETAG('m', 'f', 'r', 'a')) {
 219         err = AVERROR_INVALIDDATA;
 220         goto fail;
 221     }
 222     while (!read_tfra(tracks, start_index, f)) {
 223         /* Empty */
 224     }
 225
 226     if (split)
 227         write_fragments(tracks, start_index, f);
 228
 229 fail:
 230     if (f)
 231         avio_close(f);
 232     if (err)
 233         fprintf(stderr, "Unable to read the MFRA atom in %s\n", file);
 234     return err;
 235 }
 236
 237 static int get_private_data(struct Track *track, AVCodecContext *codec)
 238 {
 239     track->codec_private_size = codec->extradata_size;
 240     track->codec_private      = av_mallocz(codec->extradata_size);
 241     if (!track->codec_private)
 242         return AVERROR(ENOMEM);
 243     memcpy(track->codec_private, codec->extradata, codec->extradata_size);
 244     return 0;
 245 }
 246
 247 static int get_video_private_data(struct Track *track, AVCodecContext *codec)
 248 {
 249     AVIOContext *io = NULL;
 250     uint16_t sps_size, pps_size;
 251     int err = AVERROR(EINVAL);
 252
 253     if (codec->codec_id == AV_CODEC_ID_VC1)
 254         return get_private_data(track, codec);
 255
 256     if (avio_open_dyn_buf(&io) < 0)  {
 257         err = AVERROR(ENOMEM);
 258         goto fail;
 259     }
 260     if (codec->extradata_size < 11 || codec->extradata[0] != 1)
 261         goto fail;
 262     sps_size = AV_RB16(&codec->extradata[6]);
 263     if (11 + sps_size > codec->extradata_size)
 264         goto fail;
 265     avio_wb32(io, 0x00000001);
 266     avio_write(io, &codec->extradata[8], sps_size);
 267     pps_size = AV_RB16(&codec->extradata[9 + sps_size]);
 268     if (11 + sps_size + pps_size > codec->extradata_size)
 269         goto fail;
 270     avio_wb32(io, 0x00000001);
 271     avio_write(io, &codec->extradata[11 + sps_size], pps_size);
 272     err = 0;
 273
 274 fail:
 275     track->codec_private_size = avio_close_dyn_buf(io, &track->codec_private);
 276     return err;
 277 }
 278
 279 static int handle_file(struct Tracks *tracks, const char *file, int split)
 280 {
 281     AVFormatContext *ctx = NULL;
 282     int err = 0, i, orig_tracks = tracks->nb_tracks;
 283     char errbuf[50], *ptr;
 284     struct Track *track;
 285
 286     err = avformat_open_input(&ctx, file, NULL, NULL);
 287     if (err < 0) {
 288         av_strerror(err, errbuf, sizeof(errbuf));
 289         fprintf(stderr, "Unable to open %s: %s\n", file, errbuf);
 290         return 1;
 291     }
 292
 293     err = avformat_find_stream_info(ctx, NULL);
 294     if (err < 0) {
 295         av_strerror(err, errbuf, sizeof(errbuf));
 296         fprintf(stderr, "Unable to identify %s: %s\n", file, errbuf);
 297         goto fail;
 298     }
 299
 300     if (ctx->nb_streams < 1) {
 301         fprintf(stderr, "No streams found in %s\n", file);
 302         goto fail;
 303     }
 304     if (!tracks->duration)
 305         tracks->duration = ctx->duration;
 306
 307     for (i = 0; i < ctx->nb_streams; i++) {
 308         struct Track **temp;
 309         AVStream *st = ctx->streams[i];
 310         track = av_mallocz(sizeof(*track));
 311         if (!track) {
 312             err = AVERROR(ENOMEM);
 313             goto fail;
 314         }
 315         temp = av_realloc(tracks->tracks,
 316                           sizeof(*tracks->tracks) * (tracks->nb_tracks + 1));
 317         if (!temp) {
 318             av_free(track);
 319             err = AVERROR(ENOMEM);
 320             goto fail;
 321         }
 322         tracks->tracks = temp;
 323         tracks->tracks[tracks->nb_tracks] = track;
 324
 325         track->name = file;
 326         if ((ptr = strrchr(file, '/')) != NULL)
 327             track->name = ptr + 1;
 328
 329         track->bitrate   = st->codec->bit_rate;
 330         track->track_id  = st->id;
 331         track->timescale = st->time_base.den;
 332         track->duration  = av_rescale_rnd(ctx->duration, track->timescale,
 333                                           AV_TIME_BASE, AV_ROUND_UP);
 334         track->is_audio  = st->codec->codec_type == AVMEDIA_TYPE_AUDIO;
 335         track->is_video  = st->codec->codec_type == AVMEDIA_TYPE_VIDEO;
 336
 337         if (!track->is_audio && !track->is_video) {
 338             fprintf(stderr,
 339                     "Track %d in %s is neither video nor audio, skipping\n",
 340                     track->track_id, file);
 341             av_freep(&tracks->tracks[tracks->nb_tracks]);
 342             continue;
 343         }
 344
 345         if (track->is_audio) {
 346             if (tracks->audio_track < 0)
 347                 tracks->audio_track = tracks->nb_tracks;
 348             tracks->nb_audio_tracks++;
 349             track->channels    = st->codec->channels;
 350             track->sample_rate = st->codec->sample_rate;
 351             if (st->codec->codec_id == AV_CODEC_ID_AAC) {
 352                 track->fourcc    = "AACL";
 353                 track->tag       = 255;
 354                 track->blocksize = 4;
 355             } else if (st->codec->codec_id == AV_CODEC_ID_WMAPRO) {
 356                 track->fourcc    = "WMAP";
 357                 track->tag       = st->codec->codec_tag;
 358                 track->blocksize = st->codec->block_align;
 359             }
 360             get_private_data(track, st->codec);
 361         }
 362         if (track->is_video) {
 363             if (tracks->video_track < 0)
 364                 tracks->video_track = tracks->nb_tracks;
 365             tracks->nb_video_tracks++;
 366             track->width  = st->codec->width;
 367             track->height = st->codec->height;
 368             if (st->codec->codec_id == AV_CODEC_ID_H264)
 369                 track->fourcc = "H264";
 370             else if (st->codec->codec_id == AV_CODEC_ID_VC1)
 371                 track->fourcc = "WVC1";
 372             get_video_private_data(track, st->codec);
 373         }
 374
 375         tracks->nb_tracks++;
 376     }
 377
 378     avformat_close_input(&ctx);
 379
 380     err = read_mfra(tracks, orig_tracks, file, split);
 381
 382 fail:
 383     if (ctx)
 384         avformat_close_input(&ctx);
 385     return err;
 386 }
 387
 388 static void output_server_manifest(struct Tracks *tracks,
 389                                    const char *basename)
 390 {
 391     char filename[1000];
 392     FILE *out;
 393     int i;
 394
 395     snprintf(filename, sizeof(filename), "%s.ism", basename);
 396     out = fopen(filename, "w");
 397     if (!out) {
 398         perror(filename);
 399         return;
 400     }
 401     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
 402     fprintf(out, "<smil xmlns=\"http://www.w3.org/2001/SMIL20/Language\">\n");
 403     fprintf(out, "\t<head>\n");
 404     fprintf(out, "\t\t<meta name=\"clientManifestRelativePath\" "
 405                  "content=\"%s.ismc\" />\n", basename);
 406     fprintf(out, "\t</head>\n");
 407     fprintf(out, "\t<body>\n");
 408     fprintf(out, "\t\t<switch>\n");
 409     for (i = 0; i < tracks->nb_tracks; i++) {
 410         struct Track *track = tracks->tracks[i];
 411         const char *type    = track->is_video ? "video" : "audio";
 412         fprintf(out, "\t\t\t<%s src=\"%s\" systemBitrate=\"%d\">\n",
 413                 type, track->name, track->bitrate);
 414         fprintf(out, "\t\t\t\t<param name=\"trackID\" value=\"%d\" "
 415                      "valueType=\"data\" />\n", track->track_id);
 416         fprintf(out, "\t\t\t</%s>\n", type);
 417     }
 418     fprintf(out, "\t\t</switch>\n");
 419     fprintf(out, "\t</body>\n");
 420     fprintf(out, "</smil>\n");
 421     fclose(out);
 422 }
 423
 424 static void print_track_chunks(FILE *out, struct Tracks *tracks, int main,
 425                                const char *type)
 426 {
 427     int i, j;
 428     struct Track *track = tracks->tracks[main];
 429     for (i = 0; i < track->chunks; i++) {
 430         for (j = main + 1; j < tracks->nb_tracks; j++) {
 431             if (tracks->tracks[j]->is_audio == track->is_audio &&
 432                 track->offsets[i].duration != tracks->tracks[j]->offsets[i].duration)
 433                 fprintf(stderr, "Mismatched duration of %s chunk %d in %s and %s\n",
 434                         type, i, track->name, tracks->tracks[j]->name);
 435         }
 436         fprintf(out, "\t\t<c n=\"%d\" d=\"%d\" />\n",
 437                 i, track->offsets[i].duration);
 438     }
 439 }
 440
 441 static void output_client_manifest(struct Tracks *tracks,
 442                                    const char *basename, int split)
 443 {
 444     char filename[1000];
 445     FILE *out;
 446     int i, j;
 447
 448     if (split)
 449         snprintf(filename, sizeof(filename), "Manifest");
 450     else
 451         snprintf(filename, sizeof(filename), "%s.ismc", basename);
 452     out = fopen(filename, "w");
 453     if (!out) {
 454         perror(filename);
 455         return;
 456     }
 457     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
 458     fprintf(out, "<SmoothStreamingMedia MajorVersion=\"2\" MinorVersion=\"0\" "
 459                  "Duration=\"%"PRId64 "\">\n", tracks->duration * 10);
 460     if (tracks->video_track >= 0) {
 461         struct Track *track = tracks->tracks[tracks->video_track];
 462         struct Track *first_track = track;
 463         int index = 0;
 464         fprintf(out,
 465                 "\t<StreamIndex Type=\"video\" QualityLevels=\"%d\" "
 466                 "Chunks=\"%d\" "
 467                 "Url=\"QualityLevels({bitrate})/Fragments(video={start time})\">\n",
 468                 tracks->nb_video_tracks, track->chunks);
 469         for (i = 0; i < tracks->nb_tracks; i++) {
 470             track = tracks->tracks[i];
 471             if (!track->is_video)
 472                 continue;
 473             fprintf(out,
 474                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
 475                     "FourCC=\"%s\" MaxWidth=\"%d\" MaxHeight=\"%d\" "
 476                     "CodecPrivateData=\"",
 477                     index, track->bitrate, track->fourcc, track->width, track->height);
 478             for (j = 0; j < track->codec_private_size; j++)
 479                 fprintf(out, "%02X", track->codec_private[j]);
 480             fprintf(out, "\" />\n");
 481             index++;
 482             if (track->chunks != first_track->chunks)
 483                 fprintf(stderr, "Mismatched number of video chunks in %s and %s\n",
 484                         track->name, first_track->name);
 485         }
 486         print_track_chunks(out, tracks, tracks->video_track, "video");
 487         fprintf(out, "\t</StreamIndex>\n");
 488     }
 489     if (tracks->audio_track >= 0) {
 490         struct Track *track = tracks->tracks[tracks->audio_track];
 491         struct Track *first_track = track;
 492         int index = 0;
 493         fprintf(out,
 494                 "\t<StreamIndex Type=\"audio\" QualityLevels=\"%d\" "
 495                 "Chunks=\"%d\" "
 496                 "Url=\"QualityLevels({bitrate})/Fragments(audio={start time})\">\n",
 497                 tracks->nb_audio_tracks, track->chunks);
 498         for (i = 0; i < tracks->nb_tracks; i++) {
 499             track = tracks->tracks[i];
 500             if (!track->is_audio)
 501                 continue;
 502             fprintf(out,
 503                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
 504                     "FourCC=\"%s\" SamplingRate=\"%d\" Channels=\"%d\" "
 505                     "BitsPerSample=\"16\" PacketSize=\"%d\" "
 506                     "AudioTag=\"%d\" CodecPrivateData=\"",
 507                     index, track->bitrate, track->fourcc, track->sample_rate,
 508                     track->channels, track->blocksize, track->tag);
 509             for (j = 0; j < track->codec_private_size; j++)
 510                 fprintf(out, "%02X", track->codec_private[j]);
 511             fprintf(out, "\" />\n");
 512             index++;
 513             if (track->chunks != first_track->chunks)
 514                 fprintf(stderr, "Mismatched number of audio chunks in %s and %s\n",
 515                         track->name, first_track->name);
 516         }
 517         print_track_chunks(out, tracks, tracks->audio_track, "audio");
 518         fprintf(out, "\t</StreamIndex>\n");
 519     }
 520     fprintf(out, "</SmoothStreamingMedia>\n");
 521     fclose(out);
 522 }
 523
 524 static void clean_tracks(struct Tracks *tracks)
 525 {
 526     int i;
 527     for (i = 0; i < tracks->nb_tracks; i++) {
 528         av_freep(&tracks->tracks[i]->codec_private);
 529         av_freep(&tracks->tracks[i]->offsets);
 530         av_freep(&tracks->tracks[i]);
 531     }
 532     av_freep(&tracks->tracks);
 533     tracks->nb_tracks = 0;
 534 }
 535
 536 int main(int argc, char **argv)
 537 {
 538     const char *basename = NULL;
 539     int split = 0, i;
 540     struct Tracks tracks = { 0, .video_track = -1, .audio_track = -1 };
 541
 542     av_register_all();
 543
 544     for (i = 1; i < argc; i++) {
 545         if (!strcmp(argv[i], "-n")) {
 546             basename = argv[i + 1];
 547             i++;
 548         } else if (!strcmp(argv[i], "-split")) {
 549             split = 1;
 550         } else if (argv[i][0] == '-') {
 551             return usage(argv[0], 1);
 552         } else {
 553             if (handle_file(&tracks, argv[i], split))
 554                 return 1;
 555         }
 556     }
 557     if (!tracks.nb_tracks || (!basename && !split))
 558         return usage(argv[0], 1);
 559
 560     if (!split)
 561         output_server_manifest(&tracks, basename);
 562     output_client_manifest(&tracks, basename, split);
 563
 564     clean_tracks(&tracks);
 565
 566     return 0;
 567 }