summaryrefslogtreecommitdiff
path: root/libavformat/dashenc.c (plain)
blob: 6232c70da243d0aca165779a448b28f9fc15551f
1/*
2 * MPEG-DASH ISO BMFF segmenter
3 * Copyright (c) 2014 Martin Storsjo
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg 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 * FFmpeg 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 FFmpeg; 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/avassert.h"
28#include "libavutil/avstring.h"
29#include "libavutil/intreadwrite.h"
30#include "libavutil/mathematics.h"
31#include "libavutil/opt.h"
32#include "libavutil/rational.h"
33#include "libavutil/time_internal.h"
34
35#include "avc.h"
36#include "avformat.h"
37#include "avio_internal.h"
38#include "internal.h"
39#include "isom.h"
40#include "os_support.h"
41#include "url.h"
42
43// See ISO/IEC 23009-1:2014 5.3.9.4.4
44typedef enum {
45 DASH_TMPL_ID_UNDEFINED = -1,
46 DASH_TMPL_ID_ESCAPE,
47 DASH_TMPL_ID_REP_ID,
48 DASH_TMPL_ID_NUMBER,
49 DASH_TMPL_ID_BANDWIDTH,
50 DASH_TMPL_ID_TIME,
51} DASHTmplId;
52
53typedef struct Segment {
54 char file[1024];
55 int64_t start_pos;
56 int range_length, index_length;
57 int64_t time;
58 int duration;
59 int n;
60} Segment;
61
62typedef struct OutputStream {
63 AVFormatContext *ctx;
64 int ctx_inited;
65 uint8_t iobuf[32768];
66 AVIOContext *out;
67 int packets_written;
68 char initfile[1024];
69 int64_t init_start_pos;
70 int init_range_length;
71 int nb_segments, segments_size, segment_index;
72 Segment **segments;
73 int64_t first_pts, start_pts, max_pts;
74 int64_t last_dts;
75 int bit_rate;
76 char bandwidth_str[64];
77
78 char codec_str[100];
79} OutputStream;
80
81typedef struct DASHContext {
82 const AVClass *class; /* Class for private options. */
83 int window_size;
84 int extra_window_size;
85 int min_seg_duration;
86 int remove_at_exit;
87 int use_template;
88 int use_timeline;
89 int single_file;
90 OutputStream *streams;
91 int has_video, has_audio;
92 int64_t last_duration;
93 int64_t total_duration;
94 char availability_start_time[100];
95 char dirname[1024];
96 const char *single_file_name;
97 const char *init_seg_name;
98 const char *media_seg_name;
99 AVRational min_frame_rate, max_frame_rate;
100 int ambiguous_frame_rate;
101} DASHContext;
102
103static int dash_write(void *opaque, uint8_t *buf, int buf_size)
104{
105 OutputStream *os = opaque;
106 if (os->out)
107 avio_write(os->out, buf, buf_size);
108 return buf_size;
109}
110
111// RFC 6381
112static void set_codec_str(AVFormatContext *s, AVCodecParameters *par,
113 char *str, int size)
114{
115 const AVCodecTag *tags[2] = { NULL, NULL };
116 uint32_t tag;
117 if (par->codec_type == AVMEDIA_TYPE_VIDEO)
118 tags[0] = ff_codec_movvideo_tags;
119 else if (par->codec_type == AVMEDIA_TYPE_AUDIO)
120 tags[0] = ff_codec_movaudio_tags;
121 else
122 return;
123
124 tag = av_codec_get_tag(tags, par->codec_id);
125 if (!tag)
126 return;
127 if (size < 5)
128 return;
129
130 AV_WL32(str, tag);
131 str[4] = '\0';
132 if (!strcmp(str, "mp4a") || !strcmp(str, "mp4v")) {
133 uint32_t oti;
134 tags[0] = ff_mp4_obj_type;
135 oti = av_codec_get_tag(tags, par->codec_id);
136 if (oti)
137 av_strlcatf(str, size, ".%02"PRIx32, oti);
138 else
139 return;
140
141 if (tag == MKTAG('m', 'p', '4', 'a')) {
142 if (par->extradata_size >= 2) {
143 int aot = par->extradata[0] >> 3;
144 if (aot == 31)
145 aot = ((AV_RB16(par->extradata) >> 5) & 0x3f) + 32;
146 av_strlcatf(str, size, ".%d", aot);
147 }
148 } else if (tag == MKTAG('m', 'p', '4', 'v')) {
149 // Unimplemented, should output ProfileLevelIndication as a decimal number
150 av_log(s, AV_LOG_WARNING, "Incomplete RFC 6381 codec string for mp4v\n");
151 }
152 } else if (!strcmp(str, "avc1")) {
153 uint8_t *tmpbuf = NULL;
154 uint8_t *extradata = par->extradata;
155 int extradata_size = par->extradata_size;
156 if (!extradata_size)
157 return;
158 if (extradata[0] != 1) {
159 AVIOContext *pb;
160 if (avio_open_dyn_buf(&pb) < 0)
161 return;
162 if (ff_isom_write_avcc(pb, extradata, extradata_size) < 0) {
163 ffio_free_dyn_buf(&pb);
164 return;
165 }
166 extradata_size = avio_close_dyn_buf(pb, &extradata);
167 tmpbuf = extradata;
168 }
169
170 if (extradata_size >= 4)
171 av_strlcatf(str, size, ".%02x%02x%02x",
172 extradata[1], extradata[2], extradata[3]);
173 av_free(tmpbuf);
174 }
175}
176
177static void dash_free(AVFormatContext *s)
178{
179 DASHContext *c = s->priv_data;
180 int i, j;
181 if (!c->streams)
182 return;
183 for (i = 0; i < s->nb_streams; i++) {
184 OutputStream *os = &c->streams[i];
185 if (os->ctx && os->ctx_inited)
186 av_write_trailer(os->ctx);
187 if (os->ctx && os->ctx->pb)
188 av_free(os->ctx->pb);
189 ff_format_io_close(s, &os->out);
190 if (os->ctx)
191 avformat_free_context(os->ctx);
192 for (j = 0; j < os->nb_segments; j++)
193 av_free(os->segments[j]);
194 av_free(os->segments);
195 }
196 av_freep(&c->streams);
197}
198
199static void output_segment_list(OutputStream *os, AVIOContext *out, DASHContext *c)
200{
201 int i, start_index = 0, start_number = 1;
202 if (c->window_size) {
203 start_index = FFMAX(os->nb_segments - c->window_size, 0);
204 start_number = FFMAX(os->segment_index - c->window_size, 1);
205 }
206
207 if (c->use_template) {
208 int timescale = c->use_timeline ? os->ctx->streams[0]->time_base.den : AV_TIME_BASE;
209 avio_printf(out, "\t\t\t\t<SegmentTemplate timescale=\"%d\" ", timescale);
210 if (!c->use_timeline)
211 avio_printf(out, "duration=\"%"PRId64"\" ", c->last_duration);
212 avio_printf(out, "initialization=\"%s\" media=\"%s\" startNumber=\"%d\">\n", c->init_seg_name, c->media_seg_name, c->use_timeline ? start_number : 1);
213 if (c->use_timeline) {
214 int64_t cur_time = 0;
215 avio_printf(out, "\t\t\t\t\t<SegmentTimeline>\n");
216 for (i = start_index; i < os->nb_segments; ) {
217 Segment *seg = os->segments[i];
218 int repeat = 0;
219 avio_printf(out, "\t\t\t\t\t\t<S ");
220 if (i == start_index || seg->time != cur_time) {
221 cur_time = seg->time;
222 avio_printf(out, "t=\"%"PRId64"\" ", seg->time);
223 }
224 avio_printf(out, "d=\"%d\" ", seg->duration);
225 while (i + repeat + 1 < os->nb_segments &&
226 os->segments[i + repeat + 1]->duration == seg->duration &&
227 os->segments[i + repeat + 1]->time == os->segments[i + repeat]->time + os->segments[i + repeat]->duration)
228 repeat++;
229 if (repeat > 0)
230 avio_printf(out, "r=\"%d\" ", repeat);
231 avio_printf(out, "/>\n");
232 i += 1 + repeat;
233 cur_time += (1 + repeat) * seg->duration;
234 }
235 avio_printf(out, "\t\t\t\t\t</SegmentTimeline>\n");
236 }
237 avio_printf(out, "\t\t\t\t</SegmentTemplate>\n");
238 } else if (c->single_file) {
239 avio_printf(out, "\t\t\t\t<BaseURL>%s</BaseURL>\n", os->initfile);
240 avio_printf(out, "\t\t\t\t<SegmentList timescale=\"%d\" duration=\"%"PRId64"\" startNumber=\"%d\">\n", AV_TIME_BASE, c->last_duration, start_number);
241 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);
242 for (i = start_index; i < os->nb_segments; i++) {
243 Segment *seg = os->segments[i];
244 avio_printf(out, "\t\t\t\t\t<SegmentURL mediaRange=\"%"PRId64"-%"PRId64"\" ", seg->start_pos, seg->start_pos + seg->range_length - 1);
245 if (seg->index_length)
246 avio_printf(out, "indexRange=\"%"PRId64"-%"PRId64"\" ", seg->start_pos, seg->start_pos + seg->index_length - 1);
247 avio_printf(out, "/>\n");
248 }
249 avio_printf(out, "\t\t\t\t</SegmentList>\n");
250 } else {
251 avio_printf(out, "\t\t\t\t<SegmentList timescale=\"%d\" duration=\"%"PRId64"\" startNumber=\"%d\">\n", AV_TIME_BASE, c->last_duration, start_number);
252 avio_printf(out, "\t\t\t\t\t<Initialization sourceURL=\"%s\" />\n", os->initfile);
253 for (i = start_index; i < os->nb_segments; i++) {
254 Segment *seg = os->segments[i];
255 avio_printf(out, "\t\t\t\t\t<SegmentURL media=\"%s\" />\n", seg->file);
256 }
257 avio_printf(out, "\t\t\t\t</SegmentList>\n");
258 }
259}
260
261static DASHTmplId dash_read_tmpl_id(const char *identifier, char *format_tag,
262 size_t format_tag_size, const char **ptr) {
263 const char *next_ptr;
264 DASHTmplId id_type = DASH_TMPL_ID_UNDEFINED;
265
266 if (av_strstart(identifier, "$$", &next_ptr)) {
267 id_type = DASH_TMPL_ID_ESCAPE;
268 *ptr = next_ptr;
269 } else if (av_strstart(identifier, "$RepresentationID$", &next_ptr)) {
270 id_type = DASH_TMPL_ID_REP_ID;
271 // default to basic format, as $RepresentationID$ identifiers
272 // are not allowed to have custom format-tags.
273 av_strlcpy(format_tag, "%d", format_tag_size);
274 *ptr = next_ptr;
275 } else { // the following identifiers may have an explicit format_tag
276 if (av_strstart(identifier, "$Number", &next_ptr))
277 id_type = DASH_TMPL_ID_NUMBER;
278 else if (av_strstart(identifier, "$Bandwidth", &next_ptr))
279 id_type = DASH_TMPL_ID_BANDWIDTH;
280 else if (av_strstart(identifier, "$Time", &next_ptr))
281 id_type = DASH_TMPL_ID_TIME;
282 else
283 id_type = DASH_TMPL_ID_UNDEFINED;
284
285 // next parse the dash format-tag and generate a c-string format tag
286 // (next_ptr now points at the first '%' at the beginning of the format-tag)
287 if (id_type != DASH_TMPL_ID_UNDEFINED) {
288 const char *number_format = (id_type == DASH_TMPL_ID_TIME) ? PRId64 : "d";
289 if (next_ptr[0] == '$') { // no dash format-tag
290 snprintf(format_tag, format_tag_size, "%%%s", number_format);
291 *ptr = &next_ptr[1];
292 } else {
293 const char *width_ptr;
294 // only tolerate single-digit width-field (i.e. up to 9-digit width)
295 if (av_strstart(next_ptr, "%0", &width_ptr) &&
296 av_isdigit(width_ptr[0]) &&
297 av_strstart(&width_ptr[1], "d$", &next_ptr)) {
298 // yes, we're using a format tag to build format_tag.
299 snprintf(format_tag, format_tag_size, "%s%c%s", "%0", width_ptr[0], number_format);
300 *ptr = next_ptr;
301 } else {
302 av_log(NULL, AV_LOG_WARNING, "Failed to parse format-tag beginning with %s. Expected either a "
303 "closing '$' character or a format-string like '%%0[width]d', "
304 "where width must be a single digit\n", next_ptr);
305 id_type = DASH_TMPL_ID_UNDEFINED;
306 }
307 }
308 }
309 }
310 return id_type;
311}
312
313static void dash_fill_tmpl_params(char *dst, size_t buffer_size,
314 const char *template, int rep_id,
315 int number, int bit_rate,
316 int64_t time) {
317 int dst_pos = 0;
318 const char *t_cur = template;
319 while (dst_pos < buffer_size - 1 && *t_cur) {
320 char format_tag[7]; // May be "%d", "%0Xd", or "%0Xlld" (for $Time$), where X is in [0-9]
321 int n = 0;
322 DASHTmplId id_type;
323 const char *t_next = strchr(t_cur, '$'); // copy over everything up to the first '$' character
324 if (t_next) {
325 int num_copy_bytes = FFMIN(t_next - t_cur, buffer_size - dst_pos - 1);
326 av_strlcpy(&dst[dst_pos], t_cur, num_copy_bytes + 1);
327 // advance
328 dst_pos += num_copy_bytes;
329 t_cur = t_next;
330 } else { // no more DASH identifiers to substitute - just copy the rest over and break
331 av_strlcpy(&dst[dst_pos], t_cur, buffer_size - dst_pos);
332 break;
333 }
334
335 if (dst_pos >= buffer_size - 1 || !*t_cur)
336 break;
337
338 // t_cur is now pointing to a '$' character
339 id_type = dash_read_tmpl_id(t_cur, format_tag, sizeof(format_tag), &t_next);
340 switch (id_type) {
341 case DASH_TMPL_ID_ESCAPE:
342 av_strlcpy(&dst[dst_pos], "$", 2);
343 n = 1;
344 break;
345 case DASH_TMPL_ID_REP_ID:
346 n = snprintf(&dst[dst_pos], buffer_size - dst_pos, format_tag, rep_id);
347 break;
348 case DASH_TMPL_ID_NUMBER:
349 n = snprintf(&dst[dst_pos], buffer_size - dst_pos, format_tag, number);
350 break;
351 case DASH_TMPL_ID_BANDWIDTH:
352 n = snprintf(&dst[dst_pos], buffer_size - dst_pos, format_tag, bit_rate);
353 break;
354 case DASH_TMPL_ID_TIME:
355 n = snprintf(&dst[dst_pos], buffer_size - dst_pos, format_tag, time);
356 break;
357 case DASH_TMPL_ID_UNDEFINED:
358 // copy over one byte and advance
359 av_strlcpy(&dst[dst_pos], t_cur, 2);
360 n = 1;
361 t_next = &t_cur[1];
362 break;
363 }
364 // t_next points just past the processed identifier
365 // n is the number of bytes that were attempted to be written to dst
366 // (may have failed to write all because buffer_size).
367
368 // advance
369 dst_pos += FFMIN(n, buffer_size - dst_pos - 1);
370 t_cur = t_next;
371 }
372}
373
374static char *xmlescape(const char *str) {
375 int outlen = strlen(str)*3/2 + 6;
376 char *out = av_realloc(NULL, outlen + 1);
377 int pos = 0;
378 if (!out)
379 return NULL;
380 for (; *str; str++) {
381 if (pos + 6 > outlen) {
382 char *tmp;
383 outlen = 2 * outlen + 6;
384 tmp = av_realloc(out, outlen + 1);
385 if (!tmp) {
386 av_free(out);
387 return NULL;
388 }
389 out = tmp;
390 }
391 if (*str == '&') {
392 memcpy(&out[pos], "&amp;", 5);
393 pos += 5;
394 } else if (*str == '<') {
395 memcpy(&out[pos], "&lt;", 4);
396 pos += 4;
397 } else if (*str == '>') {
398 memcpy(&out[pos], "&gt;", 4);
399 pos += 4;
400 } else if (*str == '\'') {
401 memcpy(&out[pos], "&apos;", 6);
402 pos += 6;
403 } else if (*str == '\"') {
404 memcpy(&out[pos], "&quot;", 6);
405 pos += 6;
406 } else {
407 out[pos++] = *str;
408 }
409 }
410 out[pos] = '\0';
411 return out;
412}
413
414static void write_time(AVIOContext *out, int64_t time)
415{
416 int seconds = time / AV_TIME_BASE;
417 int fractions = time % AV_TIME_BASE;
418 int minutes = seconds / 60;
419 int hours = minutes / 60;
420 seconds %= 60;
421 minutes %= 60;
422 avio_printf(out, "PT");
423 if (hours)
424 avio_printf(out, "%dH", hours);
425 if (hours || minutes)
426 avio_printf(out, "%dM", minutes);
427 avio_printf(out, "%d.%dS", seconds, fractions / (AV_TIME_BASE / 10));
428}
429
430static void format_date_now(char *buf, int size)
431{
432 time_t t = time(NULL);
433 struct tm *ptm, tmbuf;
434 ptm = gmtime_r(&t, &tmbuf);
435 if (ptm) {
436 if (!strftime(buf, size, "%Y-%m-%dT%H:%M:%S", ptm))
437 buf[0] = '\0';
438 }
439}
440
441static int write_manifest(AVFormatContext *s, int final)
442{
443 DASHContext *c = s->priv_data;
444 AVIOContext *out;
445 char temp_filename[1024];
446 int ret, i;
447 const char *proto = avio_find_protocol_name(s->filename);
448 int use_rename = proto && !strcmp(proto, "file");
449 static unsigned int warned_non_file = 0;
450 AVDictionaryEntry *title = av_dict_get(s->metadata, "title", NULL, 0);
451
452 if (!use_rename && !warned_non_file++)
453 av_log(s, AV_LOG_ERROR, "Cannot use rename on non file protocol, this may lead to races and temporary partial files\n");
454
455 snprintf(temp_filename, sizeof(temp_filename), use_rename ? "%s.tmp" : "%s", s->filename);
456 ret = s->io_open(s, &out, temp_filename, AVIO_FLAG_WRITE, NULL);
457 if (ret < 0) {
458 av_log(s, AV_LOG_ERROR, "Unable to open %s for writing\n", temp_filename);
459 return ret;
460 }
461 avio_printf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
462 avio_printf(out, "<MPD xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\"\n"
463 "\txmlns=\"urn:mpeg:dash:schema:mpd:2011\"\n"
464 "\txmlns:xlink=\"http://www.w3.org/1999/xlink\"\n"
465 "\txsi:schemaLocation=\"urn:mpeg:DASH:schema:MPD:2011 http://standards.iso.org/ittf/PubliclyAvailableStandards/MPEG-DASH_schema_files/DASH-MPD.xsd\"\n"
466 "\tprofiles=\"urn:mpeg:dash:profile:isoff-live:2011\"\n"
467 "\ttype=\"%s\"\n", final ? "static" : "dynamic");
468 if (final) {
469 avio_printf(out, "\tmediaPresentationDuration=\"");
470 write_time(out, c->total_duration);
471 avio_printf(out, "\"\n");
472 } else {
473 int64_t update_period = c->last_duration / AV_TIME_BASE;
474 char now_str[100];
475 if (c->use_template && !c->use_timeline)
476 update_period = 500;
477 avio_printf(out, "\tminimumUpdatePeriod=\"PT%"PRId64"S\"\n", update_period);
478 avio_printf(out, "\tsuggestedPresentationDelay=\"PT%"PRId64"S\"\n", c->last_duration / AV_TIME_BASE);
479 if (!c->availability_start_time[0] && s->nb_streams > 0 && c->streams[0].nb_segments > 0) {
480 format_date_now(c->availability_start_time, sizeof(c->availability_start_time));
481 }
482 if (c->availability_start_time[0])
483 avio_printf(out, "\tavailabilityStartTime=\"%s\"\n", c->availability_start_time);
484 format_date_now(now_str, sizeof(now_str));
485 if (now_str[0])
486 avio_printf(out, "\tpublishTime=\"%s\"\n", now_str);
487 if (c->window_size && c->use_template) {
488 avio_printf(out, "\ttimeShiftBufferDepth=\"");
489 write_time(out, c->last_duration * c->window_size);
490 avio_printf(out, "\"\n");
491 }
492 }
493 avio_printf(out, "\tminBufferTime=\"");
494 write_time(out, c->last_duration);
495 avio_printf(out, "\">\n");
496 avio_printf(out, "\t<ProgramInformation>\n");
497 if (title) {
498 char *escaped = xmlescape(title->value);
499 avio_printf(out, "\t\t<Title>%s</Title>\n", escaped);
500 av_free(escaped);
501 }
502 avio_printf(out, "\t</ProgramInformation>\n");
503 if (c->window_size && s->nb_streams > 0 && c->streams[0].nb_segments > 0 && !c->use_template) {
504 OutputStream *os = &c->streams[0];
505 int start_index = FFMAX(os->nb_segments - c->window_size, 0);
506 int64_t start_time = av_rescale_q(os->segments[start_index]->time, s->streams[0]->time_base, AV_TIME_BASE_Q);
507 avio_printf(out, "\t<Period start=\"");
508 write_time(out, start_time);
509 avio_printf(out, "\">\n");
510 } else {
511 avio_printf(out, "\t<Period start=\"PT0.0S\">\n");
512 }
513
514 if (c->has_video) {
515 avio_printf(out, "\t\t<AdaptationSet contentType=\"video\" segmentAlignment=\"true\" bitstreamSwitching=\"true\"");
516 if (c->max_frame_rate.num && !c->ambiguous_frame_rate)
517 avio_printf(out, " %s=\"%d/%d\"", (av_cmp_q(c->min_frame_rate, c->max_frame_rate) < 0) ? "maxFrameRate" : "frameRate", c->max_frame_rate.num, c->max_frame_rate.den);
518 avio_printf(out, ">\n");
519
520 for (i = 0; i < s->nb_streams; i++) {
521 AVStream *st = s->streams[i];
522 OutputStream *os = &c->streams[i];
523
524 if (st->codecpar->codec_type != AVMEDIA_TYPE_VIDEO)
525 continue;
526
527 avio_printf(out, "\t\t\t<Representation id=\"%d\" mimeType=\"video/mp4\" codecs=\"%s\"%s width=\"%d\" height=\"%d\"", i, os->codec_str, os->bandwidth_str, st->codecpar->width, st->codecpar->height);
528 if (st->avg_frame_rate.num)
529 avio_printf(out, " frameRate=\"%d/%d\"", st->avg_frame_rate.num, st->avg_frame_rate.den);
530 avio_printf(out, ">\n");
531
532 output_segment_list(&c->streams[i], out, c);
533 avio_printf(out, "\t\t\t</Representation>\n");
534 }
535 avio_printf(out, "\t\t</AdaptationSet>\n");
536 }
537 if (c->has_audio) {
538 avio_printf(out, "\t\t<AdaptationSet contentType=\"audio\" segmentAlignment=\"true\" bitstreamSwitching=\"true\">\n");
539 for (i = 0; i < s->nb_streams; i++) {
540 AVStream *st = s->streams[i];
541 OutputStream *os = &c->streams[i];
542
543 if (st->codecpar->codec_type != AVMEDIA_TYPE_AUDIO)
544 continue;
545
546 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->codecpar->sample_rate);
547 avio_printf(out, "\t\t\t\t<AudioChannelConfiguration schemeIdUri=\"urn:mpeg:dash:23003:3:audio_channel_configuration:2011\" value=\"%d\" />\n", st->codecpar->channels);
548 output_segment_list(&c->streams[i], out, c);
549 avio_printf(out, "\t\t\t</Representation>\n");
550 }
551 avio_printf(out, "\t\t</AdaptationSet>\n");
552 }
553 avio_printf(out, "\t</Period>\n");
554 avio_printf(out, "</MPD>\n");
555 avio_flush(out);
556 ff_format_io_close(s, &out);
557
558 if (use_rename)
559 return avpriv_io_move(temp_filename, s->filename);
560
561 return 0;
562}
563
564static int set_bitrate(AVFormatContext *s)
565{
566 DASHContext *c = s->priv_data;
567 int i;
568
569 for (i = 0; i < s->nb_streams; i++) {
570 OutputStream *os = &c->streams[i];
571
572 os->bit_rate = s->streams[i]->codecpar->bit_rate;
573 if (os->bit_rate) {
574 snprintf(os->bandwidth_str, sizeof(os->bandwidth_str),
575 " bandwidth=\"%d\"", os->bit_rate);
576 } else {
577 int level = s->strict_std_compliance >= FF_COMPLIANCE_STRICT ?
578 AV_LOG_ERROR : AV_LOG_WARNING;
579 av_log(s, level, "No bit rate set for stream %d\n", i);
580 if (s->strict_std_compliance >= FF_COMPLIANCE_STRICT)
581 return AVERROR(EINVAL);
582 }
583 }
584
585 return 0;
586}
587
588static int dash_init(AVFormatContext *s)
589{
590 DASHContext *c = s->priv_data;
591 int ret = 0, i;
592 AVOutputFormat *oformat;
593 char *ptr;
594 char basename[1024];
595
596 if (c->single_file_name)
597 c->single_file = 1;
598 if (c->single_file)
599 c->use_template = 0;
600 c->ambiguous_frame_rate = 0;
601
602 av_strlcpy(c->dirname, s->filename, sizeof(c->dirname));
603 ptr = strrchr(c->dirname, '/');
604 if (ptr) {
605 av_strlcpy(basename, &ptr[1], sizeof(basename));
606 ptr[1] = '\0';
607 } else {
608 c->dirname[0] = '\0';
609 av_strlcpy(basename, s->filename, sizeof(basename));
610 }
611
612 ptr = strrchr(basename, '.');
613 if (ptr)
614 *ptr = '\0';
615
616 oformat = av_guess_format("mp4", NULL, NULL);
617 if (!oformat)
618 return AVERROR_MUXER_NOT_FOUND;
619
620 c->streams = av_mallocz(sizeof(*c->streams) * s->nb_streams);
621 if (!c->streams)
622 return AVERROR(ENOMEM);
623
624 ret = set_bitrate(s);
625 if (ret < 0)
626 return ret;
627
628 for (i = 0; i < s->nb_streams; i++) {
629 OutputStream *os = &c->streams[i];
630 AVFormatContext *ctx;
631 AVStream *st;
632 AVDictionary *opts = NULL;
633 char filename[1024];
634
635 ctx = avformat_alloc_context();
636 if (!ctx)
637 return AVERROR(ENOMEM);
638 os->ctx = ctx;
639 ctx->oformat = oformat;
640 ctx->interrupt_callback = s->interrupt_callback;
641 ctx->opaque = s->opaque;
642 ctx->io_close = s->io_close;
643 ctx->io_open = s->io_open;
644
645 if (!(st = avformat_new_stream(ctx, NULL)))
646 return AVERROR(ENOMEM);
647 avcodec_parameters_copy(st->codecpar, s->streams[i]->codecpar);
648 st->sample_aspect_ratio = s->streams[i]->sample_aspect_ratio;
649 st->time_base = s->streams[i]->time_base;
650 ctx->avoid_negative_ts = s->avoid_negative_ts;
651 ctx->flags = s->flags;
652
653 ctx->pb = avio_alloc_context(os->iobuf, sizeof(os->iobuf), AVIO_FLAG_WRITE, os, NULL, dash_write, NULL);
654 if (!ctx->pb)
655 return AVERROR(ENOMEM);
656
657 if (c->single_file) {
658 if (c->single_file_name)
659 dash_fill_tmpl_params(os->initfile, sizeof(os->initfile), c->single_file_name, i, 0, os->bit_rate, 0);
660 else
661 snprintf(os->initfile, sizeof(os->initfile), "%s-stream%d.m4s", basename, i);
662 } else {
663 dash_fill_tmpl_params(os->initfile, sizeof(os->initfile), c->init_seg_name, i, 0, os->bit_rate, 0);
664 }
665 snprintf(filename, sizeof(filename), "%s%s", c->dirname, os->initfile);
666 ret = s->io_open(s, &os->out, filename, AVIO_FLAG_WRITE, NULL);
667 if (ret < 0)
668 return ret;
669 os->init_start_pos = 0;
670
671 av_dict_set(&opts, "movflags", "frag_custom+dash+delay_moov", 0);
672 if ((ret = avformat_init_output(ctx, &opts)) < 0)
673 return ret;
674 os->ctx_inited = 1;
675 avio_flush(ctx->pb);
676 av_dict_free(&opts);
677
678 av_log(s, AV_LOG_VERBOSE, "Representation %d init segment will be written to: %s\n", i, filename);
679
680 s->streams[i]->time_base = st->time_base;
681 // If the muxer wants to shift timestamps, request to have them shifted
682 // already before being handed to this muxer, so we don't have mismatches
683 // between the MPD and the actual segments.
684 s->avoid_negative_ts = ctx->avoid_negative_ts;
685 if (st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {
686 AVRational avg_frame_rate = s->streams[i]->avg_frame_rate;
687 if (avg_frame_rate.num > 0) {
688 if (av_cmp_q(avg_frame_rate, c->min_frame_rate) < 0)
689 c->min_frame_rate = avg_frame_rate;
690 if (av_cmp_q(c->max_frame_rate, avg_frame_rate) < 0)
691 c->max_frame_rate = avg_frame_rate;
692 } else {
693 c->ambiguous_frame_rate = 1;
694 }
695 c->has_video = 1;
696 } else if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
697 c->has_audio = 1;
698 }
699
700 set_codec_str(s, st->codecpar, os->codec_str, sizeof(os->codec_str));
701 os->first_pts = AV_NOPTS_VALUE;
702 os->max_pts = AV_NOPTS_VALUE;
703 os->last_dts = AV_NOPTS_VALUE;
704 os->segment_index = 1;
705 }
706
707 if (!c->has_video && c->min_seg_duration <= 0) {
708 av_log(s, AV_LOG_WARNING, "no video stream and no min seg duration set\n");
709 return AVERROR(EINVAL);
710 }
711 return 0;
712}
713
714static int dash_write_header(AVFormatContext *s)
715{
716 DASHContext *c = s->priv_data;
717 int i, ret;
718 for (i = 0; i < s->nb_streams; i++) {
719 OutputStream *os = &c->streams[i];
720 if ((ret = avformat_write_header(os->ctx, NULL)) < 0) {
721 dash_free(s);
722 return ret;
723 }
724 }
725 ret = write_manifest(s, 0);
726 if (!ret)
727 av_log(s, AV_LOG_VERBOSE, "Manifest written to: %s\n", s->filename);
728 return ret;
729}
730
731static int add_segment(OutputStream *os, const char *file,
732 int64_t time, int duration,
733 int64_t start_pos, int64_t range_length,
734 int64_t index_length)
735{
736 int err;
737 Segment *seg;
738 if (os->nb_segments >= os->segments_size) {
739 os->segments_size = (os->segments_size + 1) * 2;
740 if ((err = av_reallocp(&os->segments, sizeof(*os->segments) *
741 os->segments_size)) < 0) {
742 os->segments_size = 0;
743 os->nb_segments = 0;
744 return err;
745 }
746 }
747 seg = av_mallocz(sizeof(*seg));
748 if (!seg)
749 return AVERROR(ENOMEM);
750 av_strlcpy(seg->file, file, sizeof(seg->file));
751 seg->time = time;
752 seg->duration = duration;
753 if (seg->time < 0) { // If pts<0, it is expected to be cut away with an edit list
754 seg->duration += seg->time;
755 seg->time = 0;
756 }
757 seg->start_pos = start_pos;
758 seg->range_length = range_length;
759 seg->index_length = index_length;
760 os->segments[os->nb_segments++] = seg;
761 os->segment_index++;
762 return 0;
763}
764
765static void write_styp(AVIOContext *pb)
766{
767 avio_wb32(pb, 24);
768 ffio_wfourcc(pb, "styp");
769 ffio_wfourcc(pb, "msdh");
770 avio_wb32(pb, 0); /* minor */
771 ffio_wfourcc(pb, "msdh");
772 ffio_wfourcc(pb, "msix");
773}
774
775static void find_index_range(AVFormatContext *s, const char *full_path,
776 int64_t pos, int *index_length)
777{
778 uint8_t buf[8];
779 AVIOContext *pb;
780 int ret;
781
782 ret = s->io_open(s, &pb, full_path, AVIO_FLAG_READ, NULL);
783 if (ret < 0)
784 return;
785 if (avio_seek(pb, pos, SEEK_SET) != pos) {
786 ff_format_io_close(s, &pb);
787 return;
788 }
789 ret = avio_read(pb, buf, 8);
790 ff_format_io_close(s, &pb);
791 if (ret < 8)
792 return;
793 if (AV_RL32(&buf[4]) != MKTAG('s', 'i', 'd', 'x'))
794 return;
795 *index_length = AV_RB32(&buf[0]);
796}
797
798static int update_stream_extradata(AVFormatContext *s, OutputStream *os,
799 AVCodecParameters *par)
800{
801 uint8_t *extradata;
802
803 if (os->ctx->streams[0]->codecpar->extradata_size || !par->extradata_size)
804 return 0;
805
806 extradata = av_malloc(par->extradata_size);
807
808 if (!extradata)
809 return AVERROR(ENOMEM);
810
811 memcpy(extradata, par->extradata, par->extradata_size);
812
813 os->ctx->streams[0]->codecpar->extradata = extradata;
814 os->ctx->streams[0]->codecpar->extradata_size = par->extradata_size;
815
816 set_codec_str(s, par, os->codec_str, sizeof(os->codec_str));
817
818 return 0;
819}
820
821static int dash_flush(AVFormatContext *s, int final, int stream)
822{
823 DASHContext *c = s->priv_data;
824 int i, ret = 0;
825
826 const char *proto = avio_find_protocol_name(s->filename);
827 int use_rename = proto && !strcmp(proto, "file");
828
829 int cur_flush_segment_index = 0;
830 if (stream >= 0)
831 cur_flush_segment_index = c->streams[stream].segment_index;
832
833 for (i = 0; i < s->nb_streams; i++) {
834 OutputStream *os = &c->streams[i];
835 char filename[1024] = "", full_path[1024], temp_path[1024];
836 int64_t start_pos;
837 int range_length, index_length = 0;
838
839 if (!os->packets_written)
840 continue;
841
842 // Flush the single stream that got a keyframe right now.
843 // Flush all audio streams as well, in sync with video keyframes,
844 // but not the other video streams.
845 if (stream >= 0 && i != stream) {
846 if (s->streams[i]->codecpar->codec_type != AVMEDIA_TYPE_AUDIO)
847 continue;
848 // Make sure we don't flush audio streams multiple times, when
849 // all video streams are flushed one at a time.
850 if (c->has_video && os->segment_index > cur_flush_segment_index)
851 continue;
852 }
853
854 if (!os->init_range_length) {
855 av_write_frame(os->ctx, NULL);
856 os->init_range_length = avio_tell(os->ctx->pb);
857 if (!c->single_file)
858 ff_format_io_close(s, &os->out);
859 }
860
861 start_pos = avio_tell(os->ctx->pb);
862
863 if (!c->single_file) {
864 dash_fill_tmpl_params(filename, sizeof(filename), c->media_seg_name, i, os->segment_index, os->bit_rate, os->start_pts);
865 snprintf(full_path, sizeof(full_path), "%s%s", c->dirname, filename);
866 snprintf(temp_path, sizeof(temp_path), use_rename ? "%s.tmp" : "%s", full_path);
867 ret = s->io_open(s, &os->out, temp_path, AVIO_FLAG_WRITE, NULL);
868 if (ret < 0)
869 break;
870 write_styp(os->ctx->pb);
871 } else {
872 snprintf(full_path, sizeof(full_path), "%s%s", c->dirname, os->initfile);
873 }
874
875 av_write_frame(os->ctx, NULL);
876 avio_flush(os->ctx->pb);
877 os->packets_written = 0;
878
879 range_length = avio_tell(os->ctx->pb) - start_pos;
880 if (c->single_file) {
881 find_index_range(s, full_path, start_pos, &index_length);
882 } else {
883 ff_format_io_close(s, &os->out);
884
885 if (use_rename) {
886 ret = avpriv_io_move(temp_path, full_path);
887 if (ret < 0)
888 break;
889 }
890 }
891 add_segment(os, filename, os->start_pts, os->max_pts - os->start_pts, start_pos, range_length, index_length);
892 av_log(s, AV_LOG_VERBOSE, "Representation %d media segment %d written to: %s\n", i, os->segment_index, full_path);
893 }
894
895 if (c->window_size || (final && c->remove_at_exit)) {
896 for (i = 0; i < s->nb_streams; i++) {
897 OutputStream *os = &c->streams[i];
898 int j;
899 int remove = os->nb_segments - c->window_size - c->extra_window_size;
900 if (final && c->remove_at_exit)
901 remove = os->nb_segments;
902 if (remove > 0) {
903 for (j = 0; j < remove; j++) {
904 char filename[1024];
905 snprintf(filename, sizeof(filename), "%s%s", c->dirname, os->segments[j]->file);
906 unlink(filename);
907 av_free(os->segments[j]);
908 }
909 os->nb_segments -= remove;
910 memmove(os->segments, os->segments + remove, os->nb_segments * sizeof(*os->segments));
911 }
912 }
913 }
914
915 if (ret >= 0)
916 ret = write_manifest(s, final);
917 return ret;
918}
919
920static int dash_write_packet(AVFormatContext *s, AVPacket *pkt)
921{
922 DASHContext *c = s->priv_data;
923 AVStream *st = s->streams[pkt->stream_index];
924 OutputStream *os = &c->streams[pkt->stream_index];
925 int64_t seg_end_duration = (os->segment_index) * (int64_t) c->min_seg_duration;
926 int ret;
927
928 ret = update_stream_extradata(s, os, st->codecpar);
929 if (ret < 0)
930 return ret;
931
932 // Fill in a heuristic guess of the packet duration, if none is available.
933 // The mp4 muxer will do something similar (for the last packet in a fragment)
934 // if nothing is set (setting it for the other packets doesn't hurt).
935 // By setting a nonzero duration here, we can be sure that the mp4 muxer won't
936 // invoke its heuristic (this doesn't have to be identical to that algorithm),
937 // so that we know the exact timestamps of fragments.
938 if (!pkt->duration && os->last_dts != AV_NOPTS_VALUE)
939 pkt->duration = pkt->dts - os->last_dts;
940 os->last_dts = pkt->dts;
941
942 // If forcing the stream to start at 0, the mp4 muxer will set the start
943 // timestamps to 0. Do the same here, to avoid mismatches in duration/timestamps.
944 if (os->first_pts == AV_NOPTS_VALUE &&
945 s->avoid_negative_ts == AVFMT_AVOID_NEG_TS_MAKE_ZERO) {
946 pkt->pts -= pkt->dts;
947 pkt->dts = 0;
948 }
949
950 if (os->first_pts == AV_NOPTS_VALUE)
951 os->first_pts = pkt->pts;
952
953 if ((!c->has_video || st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) &&
954 pkt->flags & AV_PKT_FLAG_KEY && os->packets_written &&
955 av_compare_ts(pkt->pts - os->first_pts, st->time_base,
956 seg_end_duration, AV_TIME_BASE_Q) >= 0) {
957 int64_t prev_duration = c->last_duration;
958
959 c->last_duration = av_rescale_q(pkt->pts - os->start_pts,
960 st->time_base,
961 AV_TIME_BASE_Q);
962 c->total_duration = av_rescale_q(pkt->pts - os->first_pts,
963 st->time_base,
964 AV_TIME_BASE_Q);
965
966 if ((!c->use_timeline || !c->use_template) && prev_duration) {
967 if (c->last_duration < prev_duration*9/10 ||
968 c->last_duration > prev_duration*11/10) {
969 av_log(s, AV_LOG_WARNING,
970 "Segment durations differ too much, enable use_timeline "
971 "and use_template, or keep a stricter keyframe interval\n");
972 }
973 }
974
975 if ((ret = dash_flush(s, 0, pkt->stream_index)) < 0)
976 return ret;
977 }
978
979 if (!os->packets_written) {
980 // If we wrote a previous segment, adjust the start time of the segment
981 // to the end of the previous one (which is the same as the mp4 muxer
982 // does). This avoids gaps in the timeline.
983 if (os->max_pts != AV_NOPTS_VALUE)
984 os->start_pts = os->max_pts;
985 else
986 os->start_pts = pkt->pts;
987 }
988 if (os->max_pts == AV_NOPTS_VALUE)
989 os->max_pts = pkt->pts + pkt->duration;
990 else
991 os->max_pts = FFMAX(os->max_pts, pkt->pts + pkt->duration);
992 os->packets_written++;
993 return ff_write_chained(os->ctx, 0, pkt, s, 0);
994}
995
996static int dash_write_trailer(AVFormatContext *s)
997{
998 DASHContext *c = s->priv_data;
999
1000 set_bitrate(s);
1001
1002 if (s->nb_streams > 0) {
1003 OutputStream *os = &c->streams[0];
1004 // If no segments have been written so far, try to do a crude
1005 // guess of the segment duration
1006 if (!c->last_duration)
1007 c->last_duration = av_rescale_q(os->max_pts - os->start_pts,
1008 s->streams[0]->time_base,
1009 AV_TIME_BASE_Q);
1010 c->total_duration = av_rescale_q(os->max_pts - os->first_pts,
1011 s->streams[0]->time_base,
1012 AV_TIME_BASE_Q);
1013 }
1014 dash_flush(s, 1, -1);
1015
1016 if (c->remove_at_exit) {
1017 char filename[1024];
1018 int i;
1019 for (i = 0; i < s->nb_streams; i++) {
1020 OutputStream *os = &c->streams[i];
1021 snprintf(filename, sizeof(filename), "%s%s", c->dirname, os->initfile);
1022 unlink(filename);
1023 }
1024 unlink(s->filename);
1025 }
1026
1027 return 0;
1028}
1029
1030static int dash_check_bitstream(struct AVFormatContext *s, const AVPacket *avpkt)
1031{
1032 DASHContext *c = s->priv_data;
1033 OutputStream *os = &c->streams[avpkt->stream_index];
1034 AVFormatContext *oc = os->ctx;
1035 if (oc->oformat->check_bitstream) {
1036 int ret;
1037 AVPacket pkt = *avpkt;
1038 pkt.stream_index = 0;
1039 ret = oc->oformat->check_bitstream(oc, &pkt);
1040 if (ret == 1) {
1041 AVStream *st = s->streams[avpkt->stream_index];
1042 AVStream *ost = oc->streams[0];
1043 st->internal->bsfcs = ost->internal->bsfcs;
1044 st->internal->nb_bsfcs = ost->internal->nb_bsfcs;
1045 ost->internal->bsfcs = NULL;
1046 ost->internal->nb_bsfcs = 0;
1047 }
1048 return ret;
1049 }
1050 return 1;
1051}
1052
1053#define OFFSET(x) offsetof(DASHContext, x)
1054#define E AV_OPT_FLAG_ENCODING_PARAM
1055static const AVOption options[] = {
1056 { "window_size", "number of segments kept in the manifest", OFFSET(window_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, E },
1057 { "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 },
1058 { "min_seg_duration", "minimum segment duration (in microseconds)", OFFSET(min_seg_duration), AV_OPT_TYPE_INT64, { .i64 = 5000000 }, 0, INT_MAX, E },
1059 { "remove_at_exit", "remove all segments when finished", OFFSET(remove_at_exit), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, E },
1060 { "use_template", "Use SegmentTemplate instead of SegmentList", OFFSET(use_template), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, E },
1061 { "use_timeline", "Use SegmentTimeline in SegmentTemplate", OFFSET(use_timeline), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, E },
1062 { "single_file", "Store all segments in one file, accessed using byte ranges", OFFSET(single_file), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, E },
1063 { "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 },
1064 { "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 },
1065 { "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 },
1066 { NULL },
1067};
1068
1069static const AVClass dash_class = {
1070 .class_name = "dash muxer",
1071 .item_name = av_default_item_name,
1072 .option = options,
1073 .version = LIBAVUTIL_VERSION_INT,
1074};
1075
1076AVOutputFormat ff_dash_muxer = {
1077 .name = "dash",
1078 .long_name = NULL_IF_CONFIG_SMALL("DASH Muxer"),
1079 .priv_data_size = sizeof(DASHContext),
1080 .audio_codec = AV_CODEC_ID_AAC,
1081 .video_codec = AV_CODEC_ID_H264,
1082 .flags = AVFMT_GLOBALHEADER | AVFMT_NOFILE | AVFMT_TS_NEGATIVE,
1083 .init = dash_init,
1084 .write_header = dash_write_header,
1085 .write_packet = dash_write_packet,
1086 .write_trailer = dash_write_trailer,
1087 .deinit = dash_free,
1088 .codec_tag = (const AVCodecTag* const []){ ff_mp4_obj_type, 0 },
1089 .check_bitstream = dash_check_bitstream,
1090 .priv_class = &dash_class,
1091};
1092
generated by cgit at 2024-05-17 06:59:20 (GMT)