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  * ffmpeg <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 <direct.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         err = AVERROR_INVALIDDATA;
00214         goto fail;
00215     }
00216     if (avio_rb32(f) != MKBETAG('m', 'f', 'r', 'a')) {
00217         err = AVERROR_INVALIDDATA;
00218         goto fail;
00219     }
00220     while (!read_tfra(files, start_index, f)) {
00221         /* Empty */
00222     }
00223 
00224     if (split)
00225         write_fragments(files, start_index, f);
00226 
00227 fail:
00228     if (f)
00229         avio_close(f);
00230     if (err)
00231         fprintf(stderr, "Unable to read the MFRA atom in %s\n", file);
00232     return err;
00233 }
00234 
00235 static int get_private_data(struct VideoFile *vf, AVCodecContext *codec)
00236 {
00237     vf->codec_private_size = codec->extradata_size;
00238     vf->codec_private      = av_mallocz(codec->extradata_size);
00239     if (!vf->codec_private)
00240         return AVERROR(ENOMEM);
00241     memcpy(vf->codec_private, codec->extradata, codec->extradata_size);
00242     return 0;
00243 }
00244 
00245 static int get_video_private_data(struct VideoFile *vf, AVCodecContext *codec)
00246 {
00247     AVIOContext *io = NULL;
00248     uint16_t sps_size, pps_size;
00249     int err = AVERROR(EINVAL);
00250 
00251     if (codec->codec_id == AV_CODEC_ID_VC1)
00252         return get_private_data(vf, codec);
00253 
00254     avio_open_dyn_buf(&io);
00255     if (codec->extradata_size < 11 || codec->extradata[0] != 1)
00256         goto fail;
00257     sps_size = AV_RB16(&codec->extradata[6]);
00258     if (11 + sps_size > codec->extradata_size)
00259         goto fail;
00260     avio_wb32(io, 0x00000001);
00261     avio_write(io, &codec->extradata[8], sps_size);
00262     pps_size = AV_RB16(&codec->extradata[9 + sps_size]);
00263     if (11 + sps_size + pps_size > codec->extradata_size)
00264         goto fail;
00265     avio_wb32(io, 0x00000001);
00266     avio_write(io, &codec->extradata[11 + sps_size], pps_size);
00267     err = 0;
00268 
00269 fail:
00270     vf->codec_private_size = avio_close_dyn_buf(io, &vf->codec_private);
00271     return err;
00272 }
00273 
00274 static int handle_file(struct VideoFiles *files, const char *file, int split)
00275 {
00276     AVFormatContext *ctx = NULL;
00277     int err = 0, i, orig_files = files->nb_files;
00278     char errbuf[50], *ptr;
00279     struct VideoFile *vf;
00280 
00281     err = avformat_open_input(&ctx, file, NULL, NULL);
00282     if (err < 0) {
00283         av_strerror(err, errbuf, sizeof(errbuf));
00284         fprintf(stderr, "Unable to open %s: %s\n", file, errbuf);
00285         return 1;
00286     }
00287 
00288     err = avformat_find_stream_info(ctx, NULL);
00289     if (err < 0) {
00290         av_strerror(err, errbuf, sizeof(errbuf));
00291         fprintf(stderr, "Unable to identify %s: %s\n", file, errbuf);
00292         goto fail;
00293     }
00294 
00295     if (ctx->nb_streams < 1) {
00296         fprintf(stderr, "No streams found in %s\n", file);
00297         goto fail;
00298     }
00299     if (!files->duration)
00300         files->duration = ctx->duration;
00301 
00302     for (i = 0; i < ctx->nb_streams; i++) {
00303         AVStream *st = ctx->streams[i];
00304         vf = av_mallocz(sizeof(*vf));
00305         files->files = av_realloc(files->files,
00306                                   sizeof(*files->files) * (files->nb_files + 1));
00307         files->files[files->nb_files] = vf;
00308 
00309         vf->name = file;
00310         if ((ptr = strrchr(file, '/')) != NULL)
00311             vf->name = ptr + 1;
00312 
00313         vf->bitrate   = st->codec->bit_rate;
00314         vf->track_id  = st->id;
00315         vf->timescale = st->time_base.den;
00316         vf->duration  = av_rescale_rnd(ctx->duration, vf->timescale,
00317                                        AV_TIME_BASE, AV_ROUND_UP);
00318         vf->is_audio  = st->codec->codec_type == AVMEDIA_TYPE_AUDIO;
00319         vf->is_video  = st->codec->codec_type == AVMEDIA_TYPE_VIDEO;
00320 
00321         if (!vf->is_audio && !vf->is_video) {
00322             fprintf(stderr,
00323                     "Track %d in %s is neither video nor audio, skipping\n",
00324                     vf->track_id, file);
00325             av_freep(&files->files[files->nb_files]);
00326             continue;
00327         }
00328 
00329         if (vf->is_audio) {
00330             if (files->audio_file < 0)
00331                 files->audio_file = files->nb_files;
00332             files->nb_audio_files++;
00333             vf->channels    = st->codec->channels;
00334             vf->sample_rate = st->codec->sample_rate;
00335             if (st->codec->codec_id == AV_CODEC_ID_AAC) {
00336                 vf->fourcc    = "AACL";
00337                 vf->tag       = 255;
00338                 vf->blocksize = 4;
00339             } else if (st->codec->codec_id == AV_CODEC_ID_WMAPRO) {
00340                 vf->fourcc    = "WMAP";
00341                 vf->tag       = st->codec->codec_tag;
00342                 vf->blocksize = st->codec->block_align;
00343             }
00344             get_private_data(vf, st->codec);
00345         }
00346         if (vf->is_video) {
00347             if (files->video_file < 0)
00348                 files->video_file = files->nb_files;
00349             files->nb_video_files++;
00350             vf->width  = st->codec->width;
00351             vf->height = st->codec->height;
00352             if (st->codec->codec_id == AV_CODEC_ID_H264)
00353                 vf->fourcc = "H264";
00354             else if (st->codec->codec_id == AV_CODEC_ID_VC1)
00355                 vf->fourcc = "WVC1";
00356             get_video_private_data(vf, st->codec);
00357         }
00358 
00359         files->nb_files++;
00360     }
00361 
00362     avformat_close_input(&ctx);
00363 
00364     err = read_mfra(files, orig_files, file, split);
00365 
00366 fail:
00367     if (ctx)
00368         avformat_close_input(&ctx);
00369     return err;
00370 }
00371 
00372 static void output_server_manifest(struct VideoFiles *files,
00373                                    const char *basename)
00374 {
00375     char filename[1000];
00376     FILE *out;
00377     int i;
00378 
00379     snprintf(filename, sizeof(filename), "%s.ism", basename);
00380     out = fopen(filename, "w");
00381     if (!out) {
00382         perror(filename);
00383         return;
00384     }
00385     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
00386     fprintf(out, "<smil xmlns=\"http://www.w3.org/2001/SMIL20/Language\">\n");
00387     fprintf(out, "\t<head>\n");
00388     fprintf(out, "\t\t<meta name=\"clientManifestRelativePath\" "
00389                  "content=\"%s.ismc\" />\n", basename);
00390     fprintf(out, "\t</head>\n");
00391     fprintf(out, "\t<body>\n");
00392     fprintf(out, "\t\t<switch>\n");
00393     for (i = 0; i < files->nb_files; i++) {
00394         struct VideoFile *vf = files->files[i];
00395         const char *type     = vf->is_video ? "video" : "audio";
00396         fprintf(out, "\t\t\t<%s src=\"%s\" systemBitrate=\"%d\">\n",
00397                 type, vf->name, vf->bitrate);
00398         fprintf(out, "\t\t\t\t<param name=\"trackID\" value=\"%d\" "
00399                      "valueType=\"data\" />\n", vf->track_id);
00400         fprintf(out, "\t\t\t</%s>\n", type);
00401     }
00402     fprintf(out, "\t\t</switch>\n");
00403     fprintf(out, "\t</body>\n");
00404     fprintf(out, "</smil>\n");
00405     fclose(out);
00406 }
00407 
00408 static void output_client_manifest(struct VideoFiles *files,
00409                                    const char *basename, int split)
00410 {
00411     char filename[1000];
00412     FILE *out;
00413     int i, j;
00414 
00415     if (split)
00416         snprintf(filename, sizeof(filename), "Manifest");
00417     else
00418         snprintf(filename, sizeof(filename), "%s.ismc", basename);
00419     out = fopen(filename, "w");
00420     if (!out) {
00421         perror(filename);
00422         return;
00423     }
00424     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
00425     fprintf(out, "<SmoothStreamingMedia MajorVersion=\"2\" MinorVersion=\"0\" "
00426                  "Duration=\"%"PRId64 "\">\n", files->duration * 10);
00427     if (files->video_file >= 0) {
00428         struct VideoFile *vf = files->files[files->video_file];
00429         struct VideoFile *first_vf = vf;
00430         int index = 0;
00431         fprintf(out,
00432                 "\t<StreamIndex Type=\"video\" QualityLevels=\"%d\" "
00433                 "Chunks=\"%d\" "
00434                 "Url=\"QualityLevels({bitrate})/Fragments(video={start time})\">\n",
00435                 files->nb_video_files, vf->chunks);
00436         for (i = 0; i < files->nb_files; i++) {
00437             vf = files->files[i];
00438             if (!vf->is_video)
00439                 continue;
00440             fprintf(out,
00441                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
00442                     "FourCC=\"%s\" MaxWidth=\"%d\" MaxHeight=\"%d\" "
00443                     "CodecPrivateData=\"",
00444                     index, vf->bitrate, vf->fourcc, vf->width, vf->height);
00445             for (j = 0; j < vf->codec_private_size; j++)
00446                 fprintf(out, "%02X", vf->codec_private[j]);
00447             fprintf(out, "\" />\n");
00448             index++;
00449             if (vf->chunks != first_vf->chunks)
00450                 fprintf(stderr, "Mismatched number of video chunks in %s and %s\n",
00451                         vf->name, first_vf->name);
00452         }
00453         vf = first_vf;
00454         for (i = 0; i < vf->chunks; i++) {
00455             for (j = files->video_file + 1; j < files->nb_files; j++) {
00456                 if (files->files[j]->is_video &&
00457                     vf->offsets[i].duration != files->files[j]->offsets[i].duration)
00458                     fprintf(stderr, "Mismatched duration of video chunk %d in %s and %s\n",
00459                             i, vf->name, files->files[j]->name);
00460             }
00461             fprintf(out, "\t\t<c n=\"%d\" d=\"%d\" />\n", i,
00462                     vf->offsets[i].duration);
00463         }
00464         fprintf(out, "\t</StreamIndex>\n");
00465     }
00466     if (files->audio_file >= 0) {
00467         struct VideoFile *vf = files->files[files->audio_file];
00468         struct VideoFile *first_vf = vf;
00469         int index = 0;
00470         fprintf(out,
00471                 "\t<StreamIndex Type=\"audio\" QualityLevels=\"%d\" "
00472                 "Chunks=\"%d\" "
00473                 "Url=\"QualityLevels({bitrate})/Fragments(audio={start time})\">\n",
00474                 files->nb_audio_files, vf->chunks);
00475         for (i = 0; i < files->nb_files; i++) {
00476             vf = files->files[i];
00477             if (!vf->is_audio)
00478                 continue;
00479             fprintf(out,
00480                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
00481                     "FourCC=\"%s\" SamplingRate=\"%d\" Channels=\"%d\" "
00482                     "BitsPerSample=\"16\" PacketSize=\"%d\" "
00483                     "AudioTag=\"%d\" CodecPrivateData=\"",
00484                     index, vf->bitrate, vf->fourcc, vf->sample_rate,
00485                     vf->channels, vf->blocksize, vf->tag);
00486             for (j = 0; j < vf->codec_private_size; j++)
00487                 fprintf(out, "%02X", vf->codec_private[j]);
00488             fprintf(out, "\" />\n");
00489             index++;
00490             if (vf->chunks != first_vf->chunks)
00491                 fprintf(stderr, "Mismatched number of audio chunks in %s and %s\n",
00492                         vf->name, first_vf->name);
00493         }
00494         vf = first_vf;
00495         for (i = 0; i < vf->chunks; i++) {
00496             for (j = files->audio_file + 1; j < files->nb_files; j++) {
00497                 if (files->files[j]->is_audio &&
00498                     vf->offsets[i].duration != files->files[j]->offsets[i].duration)
00499                     fprintf(stderr, "Mismatched duration of audio chunk %d in %s and %s\n",
00500                             i, vf->name, files->files[j]->name);
00501             }
00502             fprintf(out, "\t\t<c n=\"%d\" d=\"%d\" />\n",
00503                     i, vf->offsets[i].duration);
00504         }
00505         fprintf(out, "\t</StreamIndex>\n");
00506     }
00507     fprintf(out, "</SmoothStreamingMedia>\n");
00508     fclose(out);
00509 }
00510 
00511 static void clean_files(struct VideoFiles *files)
00512 {
00513     int i;
00514     for (i = 0; i < files->nb_files; i++) {
00515         av_freep(&files->files[i]->codec_private);
00516         av_freep(&files->files[i]->offsets);
00517         av_freep(&files->files[i]);
00518     }
00519     av_freep(&files->files);
00520     files->nb_files = 0;
00521 }
00522 
00523 int main(int argc, char **argv)
00524 {
00525     const char *basename = NULL;
00526     int split = 0, i;
00527     struct VideoFiles vf = { 0, .video_file = -1, .audio_file = -1 };
00528 
00529     av_register_all();
00530 
00531     for (i = 1; i < argc; i++) {
00532         if (!strcmp(argv[i], "-n")) {
00533             basename = argv[i + 1];
00534             i++;
00535         } else if (!strcmp(argv[i], "-split")) {
00536             split = 1;
00537         } else if (argv[i][0] == '-') {
00538             return usage(argv[0], 1);
00539         } else {
00540             if (handle_file(&vf, argv[i], split))
00541                 return 1;
00542         }
00543     }
00544     if (!vf.nb_files || (!basename && !split))
00545         return usage(argv[0], 1);
00546 
00547     if (!split)
00548         output_server_manifest(&vf, basename);
00549     output_client_manifest(&vf, basename, split);
00550 
00551     clean_files(&vf);
00552 
00553     return 0;
00554 }