libavformat/dashenc.c

   1 /*
   2  * MPEG-DASH ISO BMFF segmenter
   3  * Copyright (c) 2014 Martin Storsjo
   4  *
   5  * This file is part of Libav.
   6  *
   7  * Libav is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * Libav is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with Libav; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 #include "config.h"
  23 #if HAVE_UNISTD_H
  24 #include <unistd.h>
  25 #endif
  26
  27 #include "libavutil/avstring.h"
  28 #include "libavutil/intreadwrite.h"
  29 #include "libavutil/mathematics.h"
  30 #include "libavutil/opt.h"
  31 #include "libavutil/time_internal.h"
  32
  33 #include "avc.h"
  34 #include "avformat.h"
  35 #include "avio_internal.h"
  36 #include "internal.h"
  37 #include "isom.h"
  38 #include "os_support.h"
  39 #include "url.h"
  40
  41 // See ISO/IEC 23009-1:2014 5.3.9.4.4
  42 typedef enum {
  43     DASH_TMPL_ID_UNDEFINED = -1,
  44     DASH_TMPL_ID_ESCAPE,
  45     DASH_TMPL_ID_REP_ID,
  46     DASH_TMPL_ID_NUMBER,
  47     DASH_TMPL_ID_BANDWIDTH,
  48     DASH_TMPL_ID_TIME,
  49 } DASHTmplId;
  50
  51 typedef struct Segment {
  52     char file[1024];
  53     int64_t start_pos;
  54     int range_length, index_length;
  55     int64_t time;
  56     int duration;
  57     int n;
  58 } Segment;
  59
  60 typedef struct OutputStream {
  61     AVFormatContext *ctx;
  62     int ctx_inited;
  63     uint8_t iobuf[32768];
  64     URLContext *out;
  65     int packets_written;
  66     char initfile[1024];
  67     int64_t init_start_pos;
  68     int init_range_length;
  69     int nb_segments, segments_size, segment_index;
  70     Segment **segments;
  71     int64_t first_pts, start_pts, max_pts;
  72     int64_t last_dts;
  73     int bit_rate;
  74     char bandwidth_str[64];
  75
  76     char codec_str[100];
  77 } OutputStream;
  78
  79 typedef struct DASHContext {
  80     const AVClass *class;  /* Class for private options. */
  81     int window_size;
  82     int extra_window_size;
  83     int min_seg_duration;
  84     int remove_at_exit;
  85     int use_template;
  86     int use_timeline;
  87     int single_file;
  88     OutputStream *streams;
  89     int has_video, has_audio;
  90     int64_t last_duration;
  91     int64_t total_duration;
  92     char availability_start_time[100];
  93     char dirname[1024];
  94     const char *single_file_name;
  95     const char *init_seg_name;
  96     const char *media_seg_name;
  97 } DASHContext;
  98
  99 static int dash_write(void *opaque, uint8_t *buf, int buf_size)
 100 {
 101     OutputStream *os = opaque;
 102     if (os->out)
 103         ffurl_write(os->out, buf, buf_size);
 104     return buf_size;
 105 }
 106
 107 // RFC 6381
 108 static void set_codec_str(AVFormatContext *s, AVCodecContext *codec,
 109                           char *str, int size)
 110 {
 111     const AVCodecTag *tags[2] = { NULL, NULL };
 112     uint32_t tag;
 113     if (codec->codec_type == AVMEDIA_TYPE_VIDEO)
 114         tags[0] = ff_codec_movvideo_tags;
 115     else if (codec->codec_type == AVMEDIA_TYPE_AUDIO)
 116         tags[0] = ff_codec_movaudio_tags;
 117     else
 118         return;
 119
 120     tag = av_codec_get_tag(tags, codec->codec_id);
 121     if (!tag)
 122         return;
 123     if (size < 5)
 124         return;
 125
 126     AV_WL32(str, tag);
 127     str[4] = '\0';
 128     if (!strcmp(str, "mp4a") || !strcmp(str, "mp4v")) {
 129         uint32_t oti;
 130         tags[0] = ff_mp4_obj_type;
 131         oti = av_codec_get_tag(tags, codec->codec_id);
 132         if (oti)
 133             av_strlcatf(str, size, ".%02x", oti);
 134         else
 135             return;
 136
 137         if (tag == MKTAG('m', 'p', '4', 'a')) {
 138             if (codec->extradata_size >= 2) {
 139                 int aot = codec->extradata[0] >> 3;
 140                 if (aot == 31)
 141                     aot = ((AV_RB16(codec->extradata) >> 5) & 0x3f) + 32;
 142                 av_strlcatf(str, size, ".%d", aot);
 143             }
 144         } else if (tag == MKTAG('m', 'p', '4', 'v')) {
 145             // Unimplemented, should output ProfileLevelIndication as a decimal number
 146             av_log(s, AV_LOG_WARNING, "Incomplete RFC 6381 codec string for mp4v\n");
 147         }
 148     } else if (!strcmp(str, "avc1")) {
 149         uint8_t *tmpbuf = NULL;
 150         uint8_t *extradata = codec->extradata;
 151         int extradata_size = codec->extradata_size;
 152         if (!extradata_size)
 153             return;
 154         if (extradata[0] != 1) {
 155             AVIOContext *pb;
 156             if (avio_open_dyn_buf(&pb) < 0)
 157                 return;
 158             if (ff_isom_write_avcc(pb, extradata, extradata_size) < 0) {
 159                 ffio_free_dyn_buf(&pb);
 160                 return;
 161             }
 162             extradata_size = avio_close_dyn_buf(pb, &extradata);
 163             tmpbuf = extradata;
 164         }
 165
 166         if (extradata_size >= 4)
 167             av_strlcatf(str, size, ".%02x%02x%02x",
 168                         extradata[1], extradata[2], extradata[3]);
 169         av_free(tmpbuf);
 170     }
 171 }
 172
 173 static void dash_free(AVFormatContext *s)
 174 {
 175     DASHContext *c = s->priv_data;
 176     int i, j;
 177     if (!c->streams)
 178         return;
 179     for (i = 0; i < s->nb_streams; i++) {
 180         OutputStream *os = &c->streams[i];
 181         if (os->ctx && os->ctx_inited)
 182             av_write_trailer(os->ctx);
 183         if (os->ctx && os->ctx->pb)
 184             av_free(os->ctx->pb);
 185         ffurl_close(os->out);
 186         os->out =  NULL;
 187         if (os->ctx)
 188             avformat_free_context(os->ctx);
 189         for (j = 0; j < os->nb_segments; j++)
 190             av_free(os->segments[j]);
 191         av_free(os->segments);
 192     }
 193     av_freep(&c->streams);
 194 }
 195
 196 static void output_segment_list(OutputStream *os, AVIOContext *out, DASHContext *c)
 197 {
 198     int i, start_index = 0, start_number = 1;
 199     if (c->window_size) {
 200         start_index  = FFMAX(os->nb_segments   - c->window_size, 0);
 201         start_number = FFMAX(os->segment_index - c->window_size, 1);
 202     }
 203
 204     if (c->use_template) {
 205         int timescale = c->use_timeline ? os->ctx->streams[0]->time_base.den : AV_TIME_BASE;
 206         avio_printf(out, "\t\t\t\t<SegmentTemplate timescale=\"%d\" ", timescale);
 207         if (!c->use_timeline)
 208             avio_printf(out, "duration=\"%"PRId64"\" ", c->last_duration);
 209         avio_printf(out, "initialization=\"%s\" media=\"%s\" startNumber=\"%d\">\n", c->init_seg_name, c->media_seg_name, c->use_timeline ? start_number : 1);
 210         if (c->use_timeline) {
 211             int64_t cur_time = 0;
 212             avio_printf(out, "\t\t\t\t\t<SegmentTimeline>\n");
 213             for (i = start_index; i < os->nb_segments; ) {
 214                 Segment *seg = os->segments[i];
 215                 int repeat = 0;
 216                 avio_printf(out, "\t\t\t\t\t\t<S ");
 217                 if (i == start_index || seg->time != cur_time) {
 218                     cur_time = seg->time;
 219                     avio_printf(out, "t=\"%"PRId64"\" ", seg->time);
 220                 }
 221                 avio_printf(out, "d=\"%d\" ", seg->duration);
 222                 while (i + repeat + 1 < os->nb_segments &&
 223                        os->segments[i + repeat + 1]->duration == seg->duration &&
 224                        os->segments[i + repeat + 1]->time == os->segments[i + repeat]->time + os->segments[i + repeat]->duration)
 225                     repeat++;
 226                 if (repeat > 0)
 227                     avio_printf(out, "r=\"%d\" ", repeat);
 228                 avio_printf(out, "/>\n");
 229                 i += 1 + repeat;
 230                 cur_time += (1 + repeat) * seg->duration;
 231             }
 232             avio_printf(out, "\t\t\t\t\t</SegmentTimeline>\n");
 233         }
 234         avio_printf(out, "\t\t\t\t</SegmentTemplate>\n");
 235     } else if (c->single_file) {
 236         avio_printf(out, "\t\t\t\t<BaseURL>%s</BaseURL>\n", os->initfile);
 237         avio_printf(out, "\t\t\t\t<SegmentList timescale=\"%d\" duration=\"%"PRId64"\" startNumber=\"%d\">\n", AV_TIME_BASE, c->last_duration, start_number);
 238         avio_printf(out, "\t\t\t\t\t<Initialization range=\"%"PRId64"-%"PRId64"\" />\n", os->init_start_pos, os->init_start_pos + os->init_range_length - 1);
 239         for (i = start_index; i < os->nb_segments; i++) {
 240             Segment *seg = os->segments[i];
 241             avio_printf(out, "\t\t\t\t\t<SegmentURL mediaRange=\"%"PRId64"-%"PRId64"\" ", seg->start_pos, seg->start_pos + seg->range_length - 1);
 242             if (seg->index_length)
 243                 avio_printf(out, "indexRange=\"%"PRId64"-%"PRId64"\" ", seg->start_pos, seg->start_pos + seg->index_length - 1);
 244             avio_printf(out, "/>\n");
 245         }
 246         avio_printf(out, "\t\t\t\t</SegmentList>\n");
 247     } else {
 248         avio_printf(out, "\t\t\t\t<SegmentList timescale=\"%d\" duration=\"%"PRId64"\" startNumber=\"%d\">\n", AV_TIME_BASE, c->last_duration, start_number);
 249         avio_printf(out, "\t\t\t\t\t<Initialization sourceURL=\"%s\" />\n", os->initfile);
 250         for (i = start_index; i < os->nb_segments; i++) {
 251             Segment *seg = os->segments[i];
 252             avio_printf(out, "\t\t\t\t\t<SegmentURL media=\"%s\" />\n", seg->file);
 253         }
 254         avio_printf(out, "\t\t\t\t</SegmentList>\n");
 255     }
 256 }
 257
 258 static DASHTmplId dash_read_tmpl_id(const char *identifier, char *format_tag,
 259                                     size_t format_tag_size, const char **ptr) {
 260     const char *next_ptr;
 261     DASHTmplId id_type = DASH_TMPL_ID_UNDEFINED;
 262
 263     if (av_strstart(identifier, "$$", &next_ptr)) {
 264         id_type = DASH_TMPL_ID_ESCAPE;
 265         *ptr = next_ptr;
 266     } else if (av_strstart(identifier, "$RepresentationID$", &next_ptr)) {
 267         id_type = DASH_TMPL_ID_REP_ID;
 268         // default to basic format, as $RepresentationID$ identifiers
 269         // are not allowed to have custom format-tags.
 270         av_strlcpy(format_tag, "%d", format_tag_size);
 271         *ptr = next_ptr;
 272     } else { // the following identifiers may have an explicit format_tag
 273         if (av_strstart(identifier, "$Number", &next_ptr))
 274             id_type = DASH_TMPL_ID_NUMBER;
 275         else if (av_strstart(identifier, "$Bandwidth", &next_ptr))
 276             id_type = DASH_TMPL_ID_BANDWIDTH;
 277         else if (av_strstart(identifier, "$Time", &next_ptr))
 278             id_type = DASH_TMPL_ID_TIME;
 279         else
 280             id_type = DASH_TMPL_ID_UNDEFINED;
 281
 282         // next parse the dash format-tag and generate a c-string format tag
 283         // (next_ptr now points at the first '%' at the beginning of the format-tag)
 284         if (id_type != DASH_TMPL_ID_UNDEFINED) {
 285             const char *number_format = (id_type == DASH_TMPL_ID_TIME) ? PRId64 : "d";
 286             if (next_ptr[0] == '$') { // no dash format-tag
 287                 snprintf(format_tag, format_tag_size, "%%%s", number_format);
 288                 *ptr = &next_ptr[1];
 289             } else {
 290                 const char *width_ptr;
 291                 // only tolerate single-digit width-field (i.e. up to 9-digit width)
 292                 if (av_strstart(next_ptr, "%0", &width_ptr) &&
 293                     av_isdigit(width_ptr[0]) &&
 294                     av_strstart(&width_ptr[1], "d$", &next_ptr)) {
 295                     // yes, we're using a format tag to build format_tag.
 296                     snprintf(format_tag, format_tag_size, "%s%c%s", "%0", width_ptr[0], number_format);
 297                     *ptr = next_ptr;
 298                 } else {
 299                     av_log(NULL, AV_LOG_WARNING, "Failed to parse format-tag beginning with %s. Expected either a "
 300                                                  "closing '$' character or a format-string like '%%0[width]d', "
 301                                                  "where width must be a single digit\n", next_ptr);
 302                     id_type = DASH_TMPL_ID_UNDEFINED;
 303                 }
 304             }
 305         }
 306     }
 307     return id_type;
 308 }
 309
 310 static void dash_fill_tmpl_params(char *dst, size_t buffer_size,
 311                                   const char *template, int rep_id,
 312                                   int number, int bit_rate,
 313                                   int64_t time) {
 314     int dst_pos = 0;
 315     const char *t_cur = template;
 316     while (dst_pos < buffer_size - 1 && *t_cur) {
 317         char format_tag[7]; // May be "%d", "%0Xd", or "%0Xlld" (for $Time$), where X is in [0-9]
 318         int n = 0;
 319         DASHTmplId id_type;
 320         const char *t_next = strchr(t_cur, '$'); // copy over everything up to the first '$' character
 321         if (t_next) {
 322             int num_copy_bytes = FFMIN(t_next - t_cur, buffer_size - dst_pos - 1);
 323             av_strlcpy(&dst[dst_pos], t_cur, num_copy_bytes + 1);
 324             // advance
 325             dst_pos += num_copy_bytes;
 326             t_cur = t_next;
 327         } else { // no more DASH identifiers to substitute - just copy the rest over and break
 328             av_strlcpy(&dst[dst_pos], t_cur, buffer_size - dst_pos);
 329             break;
 330         }
 331
 332         if (dst_pos >= buffer_size - 1 || !*t_cur)
 333             break;
 334
 335         // t_cur is now pointing to a '$' character
 336         id_type = dash_read_tmpl_id(t_cur, format_tag, sizeof(format_tag), &t_next);
 337         switch (id_type) {
 338         case DASH_TMPL_ID_ESCAPE:
 339             av_strlcpy(&dst[dst_pos], "$", 2);
 340             n = 1;
 341             break;
 342         case DASH_TMPL_ID_REP_ID:
 343             n = snprintf(&dst[dst_pos], buffer_size - dst_pos, format_tag, rep_id);
 344             break;
 345         case DASH_TMPL_ID_NUMBER:
 346             n = snprintf(&dst[dst_pos], buffer_size - dst_pos, format_tag, number);
 347             break;
 348         case DASH_TMPL_ID_BANDWIDTH:
 349             n = snprintf(&dst[dst_pos], buffer_size - dst_pos, format_tag, bit_rate);
 350             break;
 351         case DASH_TMPL_ID_TIME:
 352             n = snprintf(&dst[dst_pos], buffer_size - dst_pos, format_tag, time);
 353             break;
 354         case DASH_TMPL_ID_UNDEFINED:
 355             // copy over one byte and advance
 356             av_strlcpy(&dst[dst_pos], t_cur, 2);
 357             n = 1;
 358             t_next = &t_cur[1];
 359             break;
 360         }
 361         // t_next points just past the processed identifier
 362         // n is the number of bytes that were attempted to be written to dst
 363         // (may have failed to write all because buffer_size).
 364
 365         // advance
 366         dst_pos += FFMIN(n, buffer_size - dst_pos - 1);
 367         t_cur = t_next;
 368     }
 369 }
 370
 371 static char *xmlescape(const char *str) {
 372     int outlen = strlen(str)*3/2 + 6;
 373     char *out = av_realloc(NULL, outlen + 1);
 374     int pos = 0;
 375     if (!out)
 376         return NULL;
 377     for (; *str; str++) {
 378         if (pos + 6 > outlen) {
 379             char *tmp;
 380             outlen = 2 * outlen + 6;
 381             tmp = av_realloc(out, outlen + 1);
 382             if (!tmp) {
 383                 av_free(out);
 384                 return NULL;
 385             }
 386             out = tmp;
 387         }
 388         if (*str == '&') {
 389             memcpy(&out[pos], "&amp;", 5);
 390             pos += 5;
 391         } else if (*str == '<') {
 392             memcpy(&out[pos], "&lt;", 4);
 393             pos += 4;
 394         } else if (*str == '>') {
 395             memcpy(&out[pos], "&gt;", 4);
 396             pos += 4;
 397         } else if (*str == '\'') {
 398             memcpy(&out[pos], "&apos;", 6);
 399             pos += 6;
 400         } else if (*str == '\"') {
 401             memcpy(&out[pos], "&quot;", 6);
 402             pos += 6;
 403         } else {
 404             out[pos++] = *str;
 405         }
 406     }
 407     out[pos] = '\0';
 408     return out;
 409 }
 410
 411 static void write_time(AVIOContext *out, int64_t time)
 412 {
 413     int seconds = time / AV_TIME_BASE;
 414     int fractions = time % AV_TIME_BASE;
 415     int minutes = seconds / 60;
 416     int hours = minutes / 60;
 417     seconds %= 60;
 418     minutes %= 60;
 419     avio_printf(out, "PT");
 420     if (hours)
 421         avio_printf(out, "%dH", hours);
 422     if (hours || minutes)
 423         avio_printf(out, "%dM", minutes);
 424     avio_printf(out, "%d.%dS", seconds, fractions / (AV_TIME_BASE / 10));
 425 }
 426
 427 static void format_date_now(char *buf, int size)
 428 {
 429     time_t t = time(NULL);
 430     struct tm *ptm, tmbuf;
 431     ptm = gmtime_r(&t, &tmbuf);
 432     if (ptm) {
 433         if (!strftime(buf, size, "%Y-%m-%dT%H:%M:%S", ptm))
 434             buf[0] = '\0';
 435     }
 436 }
 437
 438 static int write_manifest(AVFormatContext *s, int final)
 439 {
 440     DASHContext *c = s->priv_data;
 441     AVIOContext *out;
 442     char temp_filename[1024];
 443     int ret, i;
 444     AVDictionaryEntry *title = av_dict_get(s->metadata, "title", NULL, 0);
 445
 446     snprintf(temp_filename, sizeof(temp_filename), "%s.tmp", s->filename);
 447     ret = avio_open2(&out, temp_filename, AVIO_FLAG_WRITE, &s->interrupt_callback, NULL);
 448     if (ret < 0) {
 449         av_log(s, AV_LOG_ERROR, "Unable to open %s for writing\n", temp_filename);
 450         return ret;
 451     }
 452     avio_printf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
 453     avio_printf(out, "<MPD xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\"\n"
 454                 "\txmlns=\"urn:mpeg:dash:schema:mpd:2011\"\n"
 455                 "\txmlns:xlink=\"http://www.w3.org/1999/xlink\"\n"
 456                 "\txsi:schemaLocation=\"urn:mpeg:DASH:schema:MPD:2011 http://standards.iso.org/ittf/PubliclyAvailableStandards/MPEG-DASH_schema_files/DASH-MPD.xsd\"\n"
 457                 "\tprofiles=\"urn:mpeg:dash:profile:isoff-live:2011\"\n"
 458                 "\ttype=\"%s\"\n", final ? "static" : "dynamic");
 459     if (final) {
 460         avio_printf(out, "\tmediaPresentationDuration=\"");
 461         write_time(out, c->total_duration);
 462         avio_printf(out, "\"\n");
 463     } else {
 464         int64_t update_period = c->last_duration / AV_TIME_BASE;
 465         char now_str[100];
 466         if (c->use_template && !c->use_timeline)
 467             update_period = 500;
 468         avio_printf(out, "\tminimumUpdatePeriod=\"PT%"PRId64"S\"\n", update_period);
 469         avio_printf(out, "\tsuggestedPresentationDelay=\"PT%"PRId64"S\"\n", c->last_duration / AV_TIME_BASE);
 470         if (!c->availability_start_time[0] && s->nb_streams > 0 && c->streams[0].nb_segments > 0) {
 471             format_date_now(c->availability_start_time, sizeof(c->availability_start_time));
 472         }
 473         if (c->availability_start_time[0])
 474             avio_printf(out, "\tavailabilityStartTime=\"%s\"\n", c->availability_start_time);
 475         format_date_now(now_str, sizeof(now_str));
 476         if (now_str[0])
 477             avio_printf(out, "\tpublishTime=\"%s\"\n", now_str);
 478         if (c->window_size && c->use_template) {
 479             avio_printf(out, "\ttimeShiftBufferDepth=\"");
 480             write_time(out, c->last_duration * c->window_size);
 481             avio_printf(out, "\"\n");
 482         }
 483     }
 484     avio_printf(out, "\tminBufferTime=\"");
 485     write_time(out, c->last_duration);
 486     avio_printf(out, "\">\n");
 487     avio_printf(out, "\t<ProgramInformation>\n");
 488     if (title) {
 489         char *escaped = xmlescape(title->value);
 490         avio_printf(out, "\t\t<Title>%s</Title>\n", escaped);
 491         av_free(escaped);
 492     }
 493     avio_printf(out, "\t</ProgramInformation>\n");
 494     if (c->window_size && s->nb_streams > 0 && c->streams[0].nb_segments > 0 && !c->use_template) {
 495         OutputStream *os = &c->streams[0];
 496         int start_index = FFMAX(os->nb_segments - c->window_size, 0);
 497         int64_t start_time = av_rescale_q(os->segments[start_index]->time, s->streams[0]->time_base, AV_TIME_BASE_Q);
 498         avio_printf(out, "\t<Period start=\"");
 499         write_time(out, start_time);
 500         avio_printf(out, "\">\n");
 501     } else {
 502         avio_printf(out, "\t<Period start=\"PT0.0S\">\n");
 503     }
 504
 505     if (c->has_video) {
 506         avio_printf(out, "\t\t<AdaptationSet id=\"video\" segmentAlignment=\"true\" bitstreamSwitching=\"true\">\n");
 507         for (i = 0; i < s->nb_streams; i++) {
 508             AVStream *st = s->streams[i];
 509             OutputStream *os = &c->streams[i];
 510             if (st->codec->codec_type != AVMEDIA_TYPE_VIDEO)
 511                 continue;
 512             avio_printf(out, "\t\t\t<Representation id=\"%d\" mimeType=\"video/mp4\" codecs=\"%s\"%s width=\"%d\" height=\"%d\">\n", i, os->codec_str, os->bandwidth_str, st->codec->width, st->codec->height);
 513             output_segment_list(&c->streams[i], out, c);
 514             avio_printf(out, "\t\t\t</Representation>\n");
 515         }
 516         avio_printf(out, "\t\t</AdaptationSet>\n");
 517     }
 518     if (c->has_audio) {
 519         avio_printf(out, "\t\t<AdaptationSet id=\"audio\" segmentAlignment=\"true\" bitstreamSwitching=\"true\">\n");
 520         for (i = 0; i < s->nb_streams; i++) {
 521             AVStream *st = s->streams[i];
 522             OutputStream *os = &c->streams[i];
 523             if (st->codec->codec_type != AVMEDIA_TYPE_AUDIO)
 524                 continue;
 525             avio_printf(out, "\t\t\t<Representation id=\"%d\" mimeType=\"audio/mp4\" codecs=\"%s\"%s audioSamplingRate=\"%d\">\n", i, os->codec_str, os->bandwidth_str, st->codec->sample_rate);
 526             avio_printf(out, "\t\t\t\t<AudioChannelConfiguration schemeIdUri=\"urn:mpeg:dash:23003:3:audio_channel_configuration:2011\" value=\"%d\" />\n", st->codec->channels);
 527             output_segment_list(&c->streams[i], out, c);
 528             avio_printf(out, "\t\t\t</Representation>\n");
 529         }
 530         avio_printf(out, "\t\t</AdaptationSet>\n");
 531     }
 532     avio_printf(out, "\t</Period>\n");
 533     avio_printf(out, "</MPD>\n");
 534     avio_flush(out);
 535     avio_close(out);
 536     return ff_rename(temp_filename, s->filename);
 537 }
 538
 539 static int dash_write_header(AVFormatContext *s)
 540 {
 541     DASHContext *c = s->priv_data;
 542     int ret = 0, i;
 543     AVOutputFormat *oformat;
 544     char *ptr;
 545     char basename[1024];
 546
 547     if (c->single_file_name)
 548         c->single_file = 1;
 549     if (c->single_file)
 550         c->use_template = 0;
 551
 552     av_strlcpy(c->dirname, s->filename, sizeof(c->dirname));
 553     ptr = strrchr(c->dirname, '/');
 554     if (ptr) {
 555         av_strlcpy(basename, &ptr[1], sizeof(basename));
 556         ptr[1] = '\0';
 557     } else {
 558         c->dirname[0] = '\0';
 559         av_strlcpy(basename, s->filename, sizeof(basename));
 560     }
 561
 562     ptr = strrchr(basename, '.');
 563     if (ptr)
 564         *ptr = '\0';
 565
 566     oformat = av_guess_format("mp4", NULL, NULL);
 567     if (!oformat) {
 568         ret = AVERROR_MUXER_NOT_FOUND;
 569         goto fail;
 570     }
 571
 572     c->streams = av_mallocz(sizeof(*c->streams) * s->nb_streams);
 573     if (!c->streams) {
 574         ret = AVERROR(ENOMEM);
 575         goto fail;
 576     }
 577
 578     for (i = 0; i < s->nb_streams; i++) {
 579         OutputStream *os = &c->streams[i];
 580         AVFormatContext *ctx;
 581         AVStream *st;
 582         AVDictionary *opts = NULL;
 583         char filename[1024];
 584
 585         os->bit_rate = s->streams[i]->codec->bit_rate;
 586         if (os->bit_rate) {
 587             snprintf(os->bandwidth_str, sizeof(os->bandwidth_str),
 588                      " bandwidth=\"%d\"", os->bit_rate);
 589         } else {
 590             int level = s->strict_std_compliance >= FF_COMPLIANCE_STRICT ?
 591                         AV_LOG_ERROR : AV_LOG_WARNING;
 592             av_log(s, level, "No bit rate set for stream %d\n", i);
 593             if (s->strict_std_compliance >= FF_COMPLIANCE_STRICT) {
 594                 ret = AVERROR(EINVAL);
 595                 goto fail;
 596             }
 597         }
 598
 599         ctx = avformat_alloc_context();
 600         if (!ctx) {
 601             ret = AVERROR(ENOMEM);
 602             goto fail;
 603         }
 604         os->ctx = ctx;
 605         ctx->oformat = oformat;
 606         ctx->interrupt_callback = s->interrupt_callback;
 607
 608         if (!(st = avformat_new_stream(ctx, NULL))) {
 609             ret = AVERROR(ENOMEM);
 610             goto fail;
 611         }
 612         avcodec_copy_context(st->codec, s->streams[i]->codec);
 613         st->sample_aspect_ratio = s->streams[i]->sample_aspect_ratio;
 614         st->time_base = s->streams[i]->time_base;
 615         ctx->avoid_negative_ts = s->avoid_negative_ts;
 616
 617         ctx->pb = avio_alloc_context(os->iobuf, sizeof(os->iobuf), AVIO_FLAG_WRITE, os, NULL, dash_write, NULL);
 618         if (!ctx->pb) {
 619             ret = AVERROR(ENOMEM);
 620             goto fail;
 621         }
 622
 623         if (c->single_file) {
 624             if (c->single_file_name)
 625                 dash_fill_tmpl_params(os->initfile, sizeof(os->initfile), c->single_file_name, i, 0, os->bit_rate, 0);
 626             else
 627                 snprintf(os->initfile, sizeof(os->initfile), "%s-stream%d.m4s", basename, i);
 628         } else {
 629             dash_fill_tmpl_params(os->initfile, sizeof(os->initfile), c->init_seg_name, i, 0, os->bit_rate, 0);
 630         }
 631         snprintf(filename, sizeof(filename), "%s%s", c->dirname, os->initfile);
 632         ret = ffurl_open(&os->out, filename, AVIO_FLAG_WRITE, &s->interrupt_callback, NULL);
 633         if (ret < 0)
 634             goto fail;
 635         os->init_start_pos = 0;
 636
 637         av_dict_set(&opts, "movflags", "frag_custom+dash+delay_moov", 0);
 638         if ((ret = avformat_write_header(ctx, &opts)) < 0) {
 639              goto fail;
 640         }
 641         os->ctx_inited = 1;
 642         avio_flush(ctx->pb);
 643         av_dict_free(&opts);
 644
 645         av_log(s, AV_LOG_VERBOSE, "Representation %d init segment will be written to: %s\n", i, filename);
 646
 647         s->streams[i]->time_base = st->time_base;
 648         // If the muxer wants to shift timestamps, request to have them shifted
 649         // already before being handed to this muxer, so we don't have mismatches
 650         // between the MPD and the actual segments.
 651         s->avoid_negative_ts = ctx->avoid_negative_ts;
 652         if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
 653             c->has_video = 1;
 654         else if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
 655             c->has_audio = 1;
 656
 657         set_codec_str(s, st->codec, os->codec_str, sizeof(os->codec_str));
 658         os->first_pts = AV_NOPTS_VALUE;
 659         os->max_pts = AV_NOPTS_VALUE;
 660         os->last_dts = AV_NOPTS_VALUE;
 661         os->segment_index = 1;
 662     }
 663
 664     if (!c->has_video && c->min_seg_duration <= 0) {
 665         av_log(s, AV_LOG_WARNING, "no video stream and no min seg duration set\n");
 666         ret = AVERROR(EINVAL);
 667     }
 668     ret = write_manifest(s, 0);
 669     if (!ret)
 670         av_log(s, AV_LOG_VERBOSE, "Manifest written to: %s\n", s->filename);
 671
 672 fail:
 673     if (ret)
 674         dash_free(s);
 675     return ret;
 676 }
 677
 678 static int add_segment(OutputStream *os, const char *file,
 679                        int64_t time, int duration,
 680                        int64_t start_pos, int64_t range_length,
 681                        int64_t index_length)
 682 {
 683     int err;
 684     Segment *seg;
 685     if (os->nb_segments >= os->segments_size) {
 686         os->segments_size = (os->segments_size + 1) * 2;
 687         if ((err = av_reallocp(&os->segments, sizeof(*os->segments) *
 688                                os->segments_size)) < 0) {
 689             os->segments_size = 0;
 690             os->nb_segments = 0;
 691             return err;
 692         }
 693     }
 694     seg = av_mallocz(sizeof(*seg));
 695     if (!seg)
 696         return AVERROR(ENOMEM);
 697     av_strlcpy(seg->file, file, sizeof(seg->file));
 698     seg->time = time;
 699     if (seg->time < 0) // If pts<0, it is expected to be cut away with an edit list
 700         seg->time = 0;
 701     seg->duration = duration;
 702     seg->start_pos = start_pos;
 703     seg->range_length = range_length;
 704     seg->index_length = index_length;
 705     os->segments[os->nb_segments++] = seg;
 706     os->segment_index++;
 707     return 0;
 708 }
 709
 710 static void write_styp(AVIOContext *pb)
 711 {
 712     avio_wb32(pb, 24);
 713     ffio_wfourcc(pb, "styp");
 714     ffio_wfourcc(pb, "msdh");
 715     avio_wb32(pb, 0); /* minor */
 716     ffio_wfourcc(pb, "msdh");
 717     ffio_wfourcc(pb, "msix");
 718 }
 719
 720 static void find_index_range(AVFormatContext *s, const char *full_path,
 721                              int64_t pos, int *index_length)
 722 {
 723     uint8_t buf[8];
 724     URLContext *fd;
 725     int ret;
 726
 727     ret = ffurl_open(&fd, full_path, AVIO_FLAG_READ, &s->interrupt_callback, NULL);
 728     if (ret < 0)
 729         return;
 730     if (ffurl_seek(fd, pos, SEEK_SET) != pos) {
 731         ffurl_close(fd);
 732         return;
 733     }
 734     ret = ffurl_read(fd, buf, 8);
 735     ffurl_close(fd);
 736     if (ret < 8)
 737         return;
 738     if (AV_RL32(&buf[4]) != MKTAG('s', 'i', 'd', 'x'))
 739         return;
 740     *index_length = AV_RB32(&buf[0]);
 741 }
 742
 743 static int update_stream_extradata(AVFormatContext *s, OutputStream *os,
 744                                    AVCodecContext *codec)
 745 {
 746     uint8_t *extradata;
 747
 748     if (os->ctx->streams[0]->codec->extradata_size || !codec->extradata_size)
 749         return 0;
 750
 751     extradata = av_malloc(codec->extradata_size);
 752
 753     if (!extradata)
 754         return AVERROR(ENOMEM);
 755
 756     memcpy(extradata, codec->extradata, codec->extradata_size);
 757
 758     os->ctx->streams[0]->codec->extradata = extradata;
 759     os->ctx->streams[0]->codec->extradata_size = codec->extradata_size;
 760
 761     set_codec_str(s, codec, os->codec_str, sizeof(os->codec_str));
 762
 763     return 0;
 764 }
 765
 766 static int dash_flush(AVFormatContext *s, int final, int stream)
 767 {
 768     DASHContext *c = s->priv_data;
 769     int i, ret = 0;
 770     int cur_flush_segment_index = 0;
 771     if (stream >= 0)
 772         cur_flush_segment_index = c->streams[stream].segment_index;
 773
 774     for (i = 0; i < s->nb_streams; i++) {
 775         OutputStream *os = &c->streams[i];
 776         char filename[1024] = "", full_path[1024], temp_path[1024];
 777         int64_t start_pos;
 778         int range_length, index_length = 0;
 779
 780         if (!os->packets_written)
 781             continue;
 782
 783         // Flush the single stream that got a keyframe right now.
 784         // Flush all audio streams as well, in sync with video keyframes,
 785         // but not the other video streams.
 786         if (stream >= 0 && i != stream) {
 787             if (s->streams[i]->codec->codec_type != AVMEDIA_TYPE_AUDIO)
 788                 continue;
 789             // Make sure we don't flush audio streams multiple times, when
 790             // all video streams are flushed one at a time.
 791             if (c->has_video && os->segment_index > cur_flush_segment_index)
 792                 continue;
 793         }
 794
 795         if (!os->init_range_length) {
 796             av_write_frame(os->ctx, NULL);
 797             os->init_range_length = avio_tell(os->ctx->pb);
 798             if (!c->single_file) {
 799                 ffurl_close(os->out);
 800                 os->out = NULL;
 801             }
 802         }
 803
 804         start_pos = avio_tell(os->ctx->pb);
 805
 806         if (!c->single_file) {
 807             dash_fill_tmpl_params(filename, sizeof(filename), c->media_seg_name, i, os->segment_index, os->bit_rate, os->start_pts);
 808             snprintf(full_path, sizeof(full_path), "%s%s", c->dirname, filename);
 809             snprintf(temp_path, sizeof(temp_path), "%s.tmp", full_path);
 810             ret = ffurl_open(&os->out, temp_path, AVIO_FLAG_WRITE, &s->interrupt_callback, NULL);
 811             if (ret < 0)
 812                 break;
 813             write_styp(os->ctx->pb);
 814         } else {
 815             snprintf(full_path, sizeof(full_path), "%s%s", c->dirname, os->initfile);
 816         }
 817
 818         av_write_frame(os->ctx, NULL);
 819         avio_flush(os->ctx->pb);
 820         os->packets_written = 0;
 821
 822         range_length = avio_tell(os->ctx->pb) - start_pos;
 823         if (c->single_file) {
 824             find_index_range(s, full_path, start_pos, &index_length);
 825         } else {
 826             ffurl_close(os->out);
 827             os->out = NULL;
 828             ret = ff_rename(temp_path, full_path);
 829             if (ret < 0)
 830                 break;
 831         }
 832         add_segment(os, filename, os->start_pts, os->max_pts - os->start_pts, start_pos, range_length, index_length);
 833         av_log(s, AV_LOG_VERBOSE, "Representation %d media segment %d written to: %s\n", i, os->segment_index, full_path);
 834     }
 835
 836     if (c->window_size || (final && c->remove_at_exit)) {
 837         for (i = 0; i < s->nb_streams; i++) {
 838             OutputStream *os = &c->streams[i];
 839             int j;
 840             int remove = os->nb_segments - c->window_size - c->extra_window_size;
 841             if (final && c->remove_at_exit)
 842                 remove = os->nb_segments;
 843             if (remove > 0) {
 844                 for (j = 0; j < remove; j++) {
 845                     char filename[1024];
 846                     snprintf(filename, sizeof(filename), "%s%s", c->dirname, os->segments[j]->file);
 847                     unlink(filename);
 848                     av_free(os->segments[j]);
 849                 }
 850                 os->nb_segments -= remove;
 851                 memmove(os->segments, os->segments + remove, os->nb_segments * sizeof(*os->segments));
 852             }
 853         }
 854     }
 855
 856     if (ret >= 0)
 857         ret = write_manifest(s, final);
 858     return ret;
 859 }
 860
 861 static int dash_write_packet(AVFormatContext *s, AVPacket *pkt)
 862 {
 863     DASHContext *c = s->priv_data;
 864     AVStream *st = s->streams[pkt->stream_index];
 865     OutputStream *os = &c->streams[pkt->stream_index];
 866     int64_t seg_end_duration = (os->segment_index) * (int64_t) c->min_seg_duration;
 867     int ret;
 868
 869     ret = update_stream_extradata(s, os, st->codec);
 870     if (ret < 0)
 871         return ret;
 872
 873     // Fill in a heuristic guess of the packet duration, if none is available.
 874     // The mp4 muxer will do something similar (for the last packet in a fragment)
 875     // if nothing is set (setting it for the other packets doesn't hurt).
 876     // By setting a nonzero duration here, we can be sure that the mp4 muxer won't
 877     // invoke its heuristic (this doesn't have to be identical to that algorithm),
 878     // so that we know the exact timestamps of fragments.
 879     if (!pkt->duration && os->last_dts != AV_NOPTS_VALUE)
 880         pkt->duration = pkt->dts - os->last_dts;
 881     os->last_dts = pkt->dts;
 882
 883     // If forcing the stream to start at 0, the mp4 muxer will set the start
 884     // timestamps to 0. Do the same here, to avoid mismatches in duration/timestamps.
 885     if (os->first_pts == AV_NOPTS_VALUE &&
 886         s->avoid_negative_ts == AVFMT_AVOID_NEG_TS_MAKE_ZERO) {
 887         pkt->pts -= pkt->dts;
 888         pkt->dts  = 0;
 889     }
 890
 891     if (os->first_pts == AV_NOPTS_VALUE)
 892         os->first_pts = pkt->pts;
 893
 894     if ((!c->has_video || st->codec->codec_type == AVMEDIA_TYPE_VIDEO) &&
 895         pkt->flags & AV_PKT_FLAG_KEY && os->packets_written &&
 896         av_compare_ts(pkt->pts - os->first_pts, st->time_base,
 897                       seg_end_duration, AV_TIME_BASE_Q) >= 0) {
 898         int64_t prev_duration = c->last_duration;
 899
 900         c->last_duration = av_rescale_q(pkt->pts - os->start_pts,
 901                                         st->time_base,
 902                                         AV_TIME_BASE_Q);
 903         c->total_duration = av_rescale_q(pkt->pts - os->first_pts,
 904                                          st->time_base,
 905                                          AV_TIME_BASE_Q);
 906
 907         if ((!c->use_timeline || !c->use_template) && prev_duration) {
 908             if (c->last_duration < prev_duration*9/10 ||
 909                 c->last_duration > prev_duration*11/10) {
 910                 av_log(s, AV_LOG_WARNING,
 911                        "Segment durations differ too much, enable use_timeline "
 912                        "and use_template, or keep a stricter keyframe interval\n");
 913             }
 914         }
 915
 916         if ((ret = dash_flush(s, 0, pkt->stream_index)) < 0)
 917             return ret;
 918     }
 919
 920     if (!os->packets_written) {
 921         // If we wrote a previous segment, adjust the start time of the segment
 922         // to the end of the previous one (which is the same as the mp4 muxer
 923         // does). This avoids gaps in the timeline.
 924         if (os->max_pts != AV_NOPTS_VALUE)
 925             os->start_pts = os->max_pts;
 926         else
 927             os->start_pts = pkt->pts;
 928     }
 929     if (os->max_pts == AV_NOPTS_VALUE)
 930         os->max_pts = pkt->pts + pkt->duration;
 931     else
 932         os->max_pts = FFMAX(os->max_pts, pkt->pts + pkt->duration);
 933     os->packets_written++;
 934     return ff_write_chained(os->ctx, 0, pkt, s);
 935 }
 936
 937 static int dash_write_trailer(AVFormatContext *s)
 938 {
 939     DASHContext *c = s->priv_data;
 940
 941     if (s->nb_streams > 0) {
 942         OutputStream *os = &c->streams[0];
 943         // If no segments have been written so far, try to do a crude
 944         // guess of the segment duration
 945         if (!c->last_duration)
 946             c->last_duration = av_rescale_q(os->max_pts - os->start_pts,
 947                                             s->streams[0]->time_base,
 948                                             AV_TIME_BASE_Q);
 949         c->total_duration = av_rescale_q(os->max_pts - os->first_pts,
 950                                          s->streams[0]->time_base,
 951                                          AV_TIME_BASE_Q);
 952     }
 953     dash_flush(s, 1, -1);
 954
 955     if (c->remove_at_exit) {
 956         char filename[1024];
 957         int i;
 958         for (i = 0; i < s->nb_streams; i++) {
 959             OutputStream *os = &c->streams[i];
 960             snprintf(filename, sizeof(filename), "%s%s", c->dirname, os->initfile);
 961             unlink(filename);
 962         }
 963         unlink(s->filename);
 964     }
 965
 966     dash_free(s);
 967     return 0;
 968 }
 969
 970 #define OFFSET(x) offsetof(DASHContext, x)
 971 #define E AV_OPT_FLAG_ENCODING_PARAM
 972 static const AVOption options[] = {
 973     { "window_size", "number of segments kept in the manifest", OFFSET(window_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, E },
 974     { "extra_window_size", "number of segments kept outside of the manifest before removing from disk", OFFSET(extra_window_size), AV_OPT_TYPE_INT, { .i64 = 5 }, 0, INT_MAX, E },
 975     { "min_seg_duration", "minimum segment duration (in microseconds)", OFFSET(min_seg_duration), AV_OPT_TYPE_INT64, { .i64 = 5000000 }, 0, INT_MAX, E },
 976     { "remove_at_exit", "remove all segments when finished", OFFSET(remove_at_exit), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, E },
 977     { "use_template", "Use SegmentTemplate instead of SegmentList", OFFSET(use_template), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, E },
 978     { "use_timeline", "Use SegmentTimeline in SegmentTemplate", OFFSET(use_timeline), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, E },
 979     { "single_file", "Store all segments in one file, accessed using byte ranges", OFFSET(single_file), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, E },
 980     { "single_file_name", "DASH-templated name to be used for baseURL. Implies storing all segments in one file, accessed using byte ranges", OFFSET(single_file_name), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, E },
 981     { "init_seg_name", "DASH-templated name to used for the initialization segment", OFFSET(init_seg_name), AV_OPT_TYPE_STRING, {.str = "init-stream$RepresentationID$.m4s"}, 0, 0, E },
 982     { "media_seg_name", "DASH-templated name to used for the media segments", OFFSET(media_seg_name), AV_OPT_TYPE_STRING, {.str = "chunk-stream$RepresentationID$-$Number%05d$.m4s"}, 0, 0, E },
 983     { NULL },
 984 };
 985
 986 static const AVClass dash_class = {
 987     .class_name = "dash muxer",
 988     .item_name  = av_default_item_name,
 989     .option     = options,
 990     .version    = LIBAVUTIL_VERSION_INT,
 991 };
 992
 993 AVOutputFormat ff_dash_muxer = {
 994     .name           = "dash",
 995     .long_name      = NULL_IF_CONFIG_SMALL("DASH Muxer"),
 996     .priv_data_size = sizeof(DASHContext),
 997     .audio_codec    = AV_CODEC_ID_AAC,
 998     .video_codec    = AV_CODEC_ID_H264,
 999     .flags          = AVFMT_GLOBALHEADER | AVFMT_NOFILE | AVFMT_TS_NEGATIVE,
1000     .write_header   = dash_write_header,
1001     .write_packet   = dash_write_packet,
1002     .write_trailer  = dash_write_trailer,
1003     .codec_tag      = (const AVCodecTag* const []){ ff_mp4_obj_type, 0 },
1004     .priv_class     = &dash_class,
1005 };