FFmpeg: tools/ismindex.c Source File

00001 /*
00002  * Copyright (c) 2012 Martin Storsjo
00003  *
00004  * This file is part of Libav.
00005  *
00006  * Libav is free software; you can redistribute it and/or
00007  * modify it under the terms of the GNU Lesser General Public
00008  * License as published by the Free Software Foundation; either
00009  * version 2.1 of the License, or (at your option) any later version.
00010  *
00011  * Libav is distributed in the hope that it will be useful,
00012  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00013  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00014  * Lesser General Public License for more details.
00015  *
00016  * You should have received a copy of the GNU Lesser General Public
00017  * License along with Libav; if not, write to the Free Software
00018  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00019  */
00020 
00021 /*
00022  * To create a simple file for smooth streaming:
00023  * avconv <normal input/transcoding options> -movflags frag_keyframe foo.ismv
00024  * ismindex -n foo foo.ismv
00025  * This step creates foo.ism and foo.ismc that is required by IIS for
00026  * serving it.
00027  *
00028  * To pre-split files for serving as static files by a web server without
00029  * any extra server support, create the ismv file as above, and split it:
00030  * ismindex -split foo.ismv
00031  * This step creates a file Manifest and directories QualityLevel(...),
00032  * that can be read directly by a smooth streaming player.
00033  */
00034 
00035 #include <stdio.h>
00036 #include <string.h>
00037 #include <sys/stat.h>
00038 #ifdef _WIN32
00039 #include <io.h>
00040 #define mkdir(a, b) mkdir(a)
00041 #endif
00042 
00043 #include "libavformat/avformat.h"
00044 #include "libavutil/intreadwrite.h"
00045 #include "libavutil/mathematics.h"
00046 
00047 static int usage(const char *argv0, int ret)
00048 {
00049     fprintf(stderr, "%s [-split] [-n basename] file1 [file2] ...\n", argv0);
00050     return ret;
00051 }
00052 
00053 struct MoofOffset {
00054     int64_t time;
00055     int64_t offset;
00056     int duration;
00057 };
00058 
00059 struct VideoFile {
00060     const char *name;
00061     int64_t duration;
00062     int bitrate;
00063     int track_id;
00064     int is_audio, is_video;
00065     int width, height;
00066     int chunks;
00067     int sample_rate, channels;
00068     uint8_t *codec_private;
00069     int codec_private_size;
00070     struct MoofOffset *offsets;
00071     int timescale;
00072     const char *fourcc;
00073     int blocksize;
00074     int tag;
00075 };
00076 
00077 struct VideoFiles {
00078     int nb_files;
00079     int64_t duration;
00080     struct VideoFile **files;
00081     int video_file, audio_file;
00082     int nb_video_files, nb_audio_files;
00083 };
00084 
00085 static int copy_tag(AVIOContext *in, AVIOContext *out, int32_t tag_name)
00086 {
00087     int32_t size, tag;
00088 
00089     size = avio_rb32(in);
00090     tag  = avio_rb32(in);
00091     avio_wb32(out, size);
00092     avio_wb32(out, tag);
00093     if (tag != tag_name)
00094         return -1;
00095     size -= 8;
00096     while (size > 0) {
00097         char buf[1024];
00098         int len = FFMIN(sizeof(buf), size);
00099         if (avio_read(in, buf, len) != len)
00100             break;
00101         avio_write(out, buf, len);
00102         size -= len;
00103     }
00104     return 0;
00105 }
00106 
00107 static int write_fragment(const char *filename, AVIOContext *in)
00108 {
00109     AVIOContext *out = NULL;
00110     int ret;
00111 
00112     if ((ret = avio_open2(&out, filename, AVIO_FLAG_WRITE, NULL, NULL)) < 0)
00113         return ret;
00114     copy_tag(in, out, MKBETAG('m', 'o', 'o', 'f'));
00115     copy_tag(in, out, MKBETAG('m', 'd', 'a', 't'));
00116 
00117     avio_flush(out);
00118     avio_close(out);
00119 
00120     return ret;
00121 }
00122 
00123 static int write_fragments(struct VideoFiles *files, int start_index,
00124                            AVIOContext *in)
00125 {
00126     char dirname[100], filename[500];
00127     int i, j;
00128 
00129     for (i = start_index; i < files->nb_files; i++) {
00130         struct VideoFile *vf = files->files[i];
00131         const char *type     = vf->is_video ? "video" : "audio";
00132         snprintf(dirname, sizeof(dirname), "QualityLevels(%d)", vf->bitrate);
00133         mkdir(dirname, 0777);
00134         for (j = 0; j < vf->chunks; j++) {
00135             snprintf(filename, sizeof(filename), "%s/Fragments(%s=%"PRId64")",
00136                      dirname, type, vf->offsets[j].time);
00137             avio_seek(in, vf->offsets[j].offset, SEEK_SET);
00138             write_fragment(filename, in);
00139         }
00140     }
00141     return 0;
00142 }
00143 
00144 static int read_tfra(struct VideoFiles *files, int start_index, AVIOContext *f)
00145 {
00146     int ret = AVERROR_EOF, track_id;
00147     int version, fieldlength, i, j;
00148     int64_t pos   = avio_tell(f);
00149     uint32_t size = avio_rb32(f);
00150     struct VideoFile *vf = NULL;
00151 
00152     if (avio_rb32(f) != MKBETAG('t', 'f', 'r', 'a'))
00153         goto fail;
00154     version = avio_r8(f);
00155     avio_rb24(f);
00156     track_id = avio_rb32(f); /* track id */
00157     for (i = start_index; i < files->nb_files && !vf; i++)
00158         if (files->files[i]->track_id == track_id)
00159             vf = files->files[i];
00160     if (!vf) {
00161         /* Ok, continue parsing the next atom */
00162         ret = 0;
00163         goto fail;
00164     }
00165     fieldlength = avio_rb32(f);
00166     vf->chunks  = avio_rb32(f);
00167     vf->offsets = av_mallocz(sizeof(*vf->offsets) * vf->chunks);
00168     if (!vf->offsets) {
00169         ret = AVERROR(ENOMEM);
00170         goto fail;
00171     }
00172     for (i = 0; i < vf->chunks; i++) {
00173         if (version == 1) {
00174             vf->offsets[i].time   = avio_rb64(f);
00175             vf->offsets[i].offset = avio_rb64(f);
00176         } else {
00177             vf->offsets[i].time   = avio_rb32(f);
00178             vf->offsets[i].offset = avio_rb32(f);
00179         }
00180         for (j = 0; j < ((fieldlength >> 4) & 3) + 1; j++)
00181             avio_r8(f);
00182         for (j = 0; j < ((fieldlength >> 2) & 3) + 1; j++)
00183             avio_r8(f);
00184         for (j = 0; j < ((fieldlength >> 0) & 3) + 1; j++)
00185             avio_r8(f);
00186         if (i > 0)
00187             vf->offsets[i - 1].duration = vf->offsets[i].time -
00188                                           vf->offsets[i - 1].time;
00189     }
00190     if (vf->chunks > 0)
00191         vf->offsets[vf->chunks - 1].duration = vf->duration -
00192                                                vf->offsets[vf->chunks - 1].time;
00193     ret = 0;
00194 
00195 fail:
00196     avio_seek(f, pos + size, SEEK_SET);
00197     return ret;
00198 }
00199 
00200 static int read_mfra(struct VideoFiles *files, int start_index,
00201                      const char *file, int split)
00202 {
00203     int err = 0;
00204     AVIOContext *f = NULL;
00205     int32_t mfra_size;
00206 
00207     if ((err = avio_open2(&f, file, AVIO_FLAG_READ, NULL, NULL)) < 0)
00208         goto fail;
00209     avio_seek(f, avio_size(f) - 4, SEEK_SET);
00210     mfra_size = avio_rb32(f);
00211     avio_seek(f, -mfra_size, SEEK_CUR);
00212     if (avio_rb32(f) != mfra_size)
00213         goto fail;
00214     if (avio_rb32(f) != MKBETAG('m', 'f', 'r', 'a'))
00215         goto fail;
00216     while (!read_tfra(files, start_index, f)) {
00217         /* Empty */
00218     }
00219 
00220     if (split)
00221         write_fragments(files, start_index, f);
00222 
00223 fail:
00224     if (f)
00225         avio_close(f);
00226     return err;
00227 }
00228 
00229 static int get_private_data(struct VideoFile *vf, AVCodecContext *codec)
00230 {
00231     vf->codec_private_size = codec->extradata_size;
00232     vf->codec_private      = av_mallocz(codec->extradata_size);
00233     if (!vf->codec_private)
00234         return AVERROR(ENOMEM);
00235     memcpy(vf->codec_private, codec->extradata, codec->extradata_size);
00236     return 0;
00237 }
00238 
00239 static int get_video_private_data(struct VideoFile *vf, AVCodecContext *codec)
00240 {
00241     AVIOContext *io = NULL;
00242     uint16_t sps_size, pps_size;
00243     int err = AVERROR(EINVAL);
00244 
00245     if (codec->codec_id == CODEC_ID_VC1)
00246         return get_private_data(vf, codec);
00247 
00248     avio_open_dyn_buf(&io);
00249     if (codec->extradata_size < 11 || codec->extradata[0] != 1)
00250         goto fail;
00251     sps_size = AV_RB16(&codec->extradata[6]);
00252     if (11 + sps_size > codec->extradata_size)
00253         goto fail;
00254     avio_wb32(io, 0x00000001);
00255     avio_write(io, &codec->extradata[8], sps_size);
00256     pps_size = AV_RB16(&codec->extradata[9 + sps_size]);
00257     if (11 + sps_size + pps_size > codec->extradata_size)
00258         goto fail;
00259     avio_wb32(io, 0x00000001);
00260     avio_write(io, &codec->extradata[11 + sps_size], pps_size);
00261     err = 0;
00262 
00263 fail:
00264     vf->codec_private_size = avio_close_dyn_buf(io, &vf->codec_private);
00265     return err;
00266 }
00267 
00268 static int handle_file(struct VideoFiles *files, const char *file, int split)
00269 {
00270     AVFormatContext *ctx = NULL;
00271     int err = 0, i, orig_files = files->nb_files;
00272     char errbuf[50], *ptr;
00273     struct VideoFile *vf;
00274 
00275     err = avformat_open_input(&ctx, file, NULL, NULL);
00276     if (err < 0) {
00277         av_strerror(err, errbuf, sizeof(errbuf));
00278         fprintf(stderr, "Unable to open %s: %s\n", file, errbuf);
00279         return 1;
00280     }
00281 
00282     err = avformat_find_stream_info(ctx, NULL);
00283     if (err < 0) {
00284         av_strerror(err, errbuf, sizeof(errbuf));
00285         fprintf(stderr, "Unable to identify %s: %s\n", file, errbuf);
00286         goto fail;
00287     }
00288 
00289     if (ctx->nb_streams < 1) {
00290         fprintf(stderr, "No streams found in %s\n", file);
00291         goto fail;
00292     }
00293     if (!files->duration)
00294         files->duration = ctx->duration;
00295 
00296     for (i = 0; i < ctx->nb_streams; i++) {
00297         AVStream *st = ctx->streams[i];
00298         vf = av_mallocz(sizeof(*vf));
00299         files->files = av_realloc(files->files,
00300                                   sizeof(*files->files) * (files->nb_files + 1));
00301         files->files[files->nb_files] = vf;
00302 
00303         vf->name = file;
00304         if ((ptr = strrchr(file, '/')) != NULL)
00305             vf->name = ptr + 1;
00306 
00307         vf->bitrate   = st->codec->bit_rate;
00308         vf->track_id  = st->id;
00309         vf->timescale = st->time_base.den;
00310         vf->duration  = av_rescale_rnd(ctx->duration, vf->timescale,
00311                                        AV_TIME_BASE, AV_ROUND_UP);
00312         vf->is_audio  = st->codec->codec_type == AVMEDIA_TYPE_AUDIO;
00313         vf->is_video  = st->codec->codec_type == AVMEDIA_TYPE_VIDEO;
00314 
00315         if (!vf->is_audio && !vf->is_video) {
00316             fprintf(stderr,
00317                     "Track %d in %s is neither video nor audio, skipping\n",
00318                     vf->track_id, file);
00319             av_freep(&files->files[files->nb_files]);
00320             continue;
00321         }
00322 
00323         if (vf->is_audio) {
00324             if (files->audio_file < 0)
00325                 files->audio_file = files->nb_files;
00326             files->nb_audio_files++;
00327             vf->channels    = st->codec->channels;
00328             vf->sample_rate = st->codec->sample_rate;
00329             if (st->codec->codec_id == CODEC_ID_AAC) {
00330                 vf->fourcc    = "AACL";
00331                 vf->tag       = 255;
00332                 vf->blocksize = 4;
00333             } else if (st->codec->codec_id == CODEC_ID_WMAPRO) {
00334                 vf->fourcc    = "WMAP";
00335                 vf->tag       = st->codec->codec_tag;
00336                 vf->blocksize = st->codec->block_align;
00337             }
00338             get_private_data(vf, st->codec);
00339         }
00340         if (vf->is_video) {
00341             if (files->video_file < 0)
00342                 files->video_file = files->nb_files;
00343             files->nb_video_files++;
00344             vf->width  = st->codec->width;
00345             vf->height = st->codec->height;
00346             if (st->codec->codec_id == CODEC_ID_H264)
00347                 vf->fourcc = "H264";
00348             else if (st->codec->codec_id == CODEC_ID_VC1)
00349                 vf->fourcc = "WVC1";
00350             get_video_private_data(vf, st->codec);
00351         }
00352 
00353         files->nb_files++;
00354     }
00355 
00356     avformat_close_input(&ctx);
00357 
00358     read_mfra(files, orig_files, file, split);
00359 
00360 fail:
00361     if (ctx)
00362         avformat_close_input(&ctx);
00363     return err;
00364 }
00365 
00366 static void output_server_manifest(struct VideoFiles *files,
00367                                    const char *basename)
00368 {
00369     char filename[1000];
00370     FILE *out;
00371     int i;
00372 
00373     snprintf(filename, sizeof(filename), "%s.ism", basename);
00374     out = fopen(filename, "w");
00375     if (!out) {
00376         perror(filename);
00377         return;
00378     }
00379     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
00380     fprintf(out, "<smil xmlns=\"http://www.w3.org/2001/SMIL20/Language\">\n");
00381     fprintf(out, "\t<head>\n");
00382     fprintf(out, "\t\t<meta name=\"clientManifestRelativePath\" "
00383                  "content=\"%s.ismc\" />\n", basename);
00384     fprintf(out, "\t</head>\n");
00385     fprintf(out, "\t<body>\n");
00386     fprintf(out, "\t\t<switch>\n");
00387     for (i = 0; i < files->nb_files; i++) {
00388         struct VideoFile *vf = files->files[i];
00389         const char *type     = vf->is_video ? "video" : "audio";
00390         fprintf(out, "\t\t\t<%s src=\"%s\" systemBitrate=\"%d\">\n",
00391                 type, vf->name, vf->bitrate);
00392         fprintf(out, "\t\t\t\t<param name=\"trackID\" value=\"%d\" "
00393                      "valueType=\"data\" />\n", vf->track_id);
00394         fprintf(out, "\t\t\t</%s>\n", type);
00395     }
00396     fprintf(out, "\t\t</switch>\n");
00397     fprintf(out, "\t</body>\n");
00398     fprintf(out, "</smil>\n");
00399     fclose(out);
00400 }
00401 
00402 static void output_client_manifest(struct VideoFiles *files,
00403                                    const char *basename, int split)
00404 {
00405     char filename[1000];
00406     FILE *out;
00407     int i, j;
00408 
00409     if (split)
00410         snprintf(filename, sizeof(filename), "Manifest");
00411     else
00412         snprintf(filename, sizeof(filename), "%s.ismc", basename);
00413     out = fopen(filename, "w");
00414     if (!out) {
00415         perror(filename);
00416         return;
00417     }
00418     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
00419     fprintf(out, "<SmoothStreamingMedia MajorVersion=\"2\" MinorVersion=\"0\" "
00420                  "Duration=\"%"PRId64 "\">\n", files->duration * 10);
00421     if (files->video_file >= 0) {
00422         struct VideoFile *vf = files->files[files->video_file];
00423         int index = 0;
00424         fprintf(out,
00425                 "\t<StreamIndex Type=\"video\" QualityLevels=\"%d\" "
00426                 "Chunks=\"%d\" "
00427                 "Url=\"QualityLevels({bitrate})/Fragments(video={start time})\">\n",
00428                 files->nb_video_files, vf->chunks);
00429         for (i = 0; i < files->nb_files; i++) {
00430             vf = files->files[i];
00431             if (!vf->is_video)
00432                 continue;
00433             fprintf(out,
00434                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
00435                     "FourCC=\"%s\" MaxWidth=\"%d\" MaxHeight=\"%d\" "
00436                     "CodecPrivateData=\"",
00437                     index, vf->bitrate, vf->fourcc, vf->width, vf->height);
00438             for (j = 0; j < vf->codec_private_size; j++)
00439                 fprintf(out, "%02X", vf->codec_private[j]);
00440             fprintf(out, "\" />\n");
00441             index++;
00442         }
00443         vf = files->files[files->video_file];
00444         for (i = 0; i < vf->chunks; i++)
00445             fprintf(out, "\t\t<c n=\"%d\" d=\"%d\" />\n", i,
00446                     vf->offsets[i].duration);
00447         fprintf(out, "\t</StreamIndex>\n");
00448     }
00449     if (files->audio_file >= 0) {
00450         struct VideoFile *vf = files->files[files->audio_file];
00451         int index = 0;
00452         fprintf(out,
00453                 "\t<StreamIndex Type=\"audio\" QualityLevels=\"%d\" "
00454                 "Chunks=\"%d\" "
00455                 "Url=\"QualityLevels({bitrate})/Fragments(audio={start time})\">\n",
00456                 files->nb_audio_files, vf->chunks);
00457         for (i = 0; i < files->nb_files; i++) {
00458             vf = files->files[i];
00459             if (!vf->is_audio)
00460                 continue;
00461             fprintf(out,
00462                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
00463                     "FourCC=\"%s\" SamplingRate=\"%d\" Channels=\"%d\" "
00464                     "BitsPerSample=\"16\" PacketSize=\"%d\" "
00465                     "AudioTag=\"%d\" CodecPrivateData=\"",
00466                     index, vf->bitrate, vf->fourcc, vf->sample_rate,
00467                     vf->channels, vf->blocksize, vf->tag);
00468             for (j = 0; j < vf->codec_private_size; j++)
00469                 fprintf(out, "%02X", vf->codec_private[j]);
00470             fprintf(out, "\" />\n");
00471             index++;
00472         }
00473         vf = files->files[files->audio_file];
00474         for (i = 0; i < vf->chunks; i++)
00475             fprintf(out, "\t\t<c n=\"%d\" d=\"%d\" />\n",
00476                     i, vf->offsets[i].duration);
00477         fprintf(out, "\t</StreamIndex>\n");
00478     }
00479     fprintf(out, "</SmoothStreamingMedia>\n");
00480     fclose(out);
00481 }
00482 
00483 static void clean_files(struct VideoFiles *files)
00484 {
00485     int i;
00486     for (i = 0; i < files->nb_files; i++) {
00487         av_freep(&files->files[i]->codec_private);
00488         av_freep(&files->files[i]->offsets);
00489         av_freep(&files->files[i]);
00490     }
00491     av_freep(&files->files);
00492     files->nb_files = 0;
00493 }
00494 
00495 int main(int argc, char **argv)
00496 {
00497     const char *basename = NULL;
00498     int split = 0, i;
00499     struct VideoFiles vf = { 0, .video_file = -1, .audio_file = -1 };
00500 
00501     av_register_all();
00502 
00503     for (i = 1; i < argc; i++) {
00504         if (!strcmp(argv[i], "-n")) {
00505             basename = argv[i + 1];
00506             i++;
00507         } else if (!strcmp(argv[i], "-split")) {
00508             split = 1;
00509         } else if (argv[i][0] == '-') {
00510             return usage(argv[0], 1);
00511         } else {
00512             handle_file(&vf, argv[i], split);
00513         }
00514     }
00515     if (!vf.nb_files || (!basename && !split))
00516         return usage(argv[0], 1);
00517 
00518     if (!split)
00519         output_server_manifest(&vf, basename);
00520     output_client_manifest(&vf, basename, split);
00521 
00522     clean_files(&vf);
00523 
00524     return 0;
00525 }