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