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