c84dd82e7da5328436e6e9ebeee5fa225067edc1
[libav.git] / libavformat / hlsenc.c
1 /*
2 * Apple HTTP Live Streaming segmenter
3 * Copyright (c) 2012, Luca Barbato
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 <float.h>
23 #include <stdint.h>
24
25 #include <config.h>
26
27 #if CONFIG_OPENSSL
28 #include <openssl/rand.h>
29 #endif
30
31 #include "libavutil/mathematics.h"
32 #include "libavutil/parseutils.h"
33 #include "libavutil/avstring.h"
34 #include "libavutil/intreadwrite.h"
35 #include "libavutil/opt.h"
36 #include "libavutil/random_seed.h"
37 #include "libavutil/log.h"
38
39 #include "avformat.h"
40 #include "internal.h"
41
42 typedef struct ListEntry {
43 char name[1024];
44 int64_t duration; // segment duration in AV_TIME_BASE units
45 struct ListEntry *next;
46 } ListEntry;
47
48 typedef struct HLSContext {
49 const AVClass *class; // Class for private options.
50 unsigned number;
51 int64_t sequence;
52 int64_t start_sequence;
53 AVOutputFormat *oformat;
54 AVFormatContext *avf;
55 float time; // Set by a private option.
56 int size; // Set by a private option.
57 int wrap; // Set by a private option.
58 int version; // Set by a private option.
59 int allowcache;
60 int64_t recording_time;
61 int has_video;
62 // The following timestamps are in AV_TIME_BASE units.
63 int64_t start_pts;
64 int64_t end_pts;
65 int64_t duration; // last segment duration computed so far.
66 int nb_entries;
67 ListEntry *list;
68 ListEntry *end_list;
69 char *basename;
70 char *baseurl;
71
72 int encrypt; // Set by a private option.
73 char *key; // Set by a private option.
74 int key_len;
75 char *key_url; // Set by a private option.
76 char *iv; // Set by a private option.
77 int iv_len;
78
79 char *key_basename;
80
81 AVDictionary *enc_opts;
82 } HLSContext;
83
84
85 static int randomize(uint8_t *buf, int len)
86 {
87 #if CONFIG_OPENSSL
88 if (RAND_bytes(buf, len))
89 return 0;
90 return AVERROR(EIO);
91 #else
92 return AVERROR(ENOSYS);
93 #endif
94 }
95
96 static void free_encryption(AVFormatContext *s)
97 {
98 HLSContext *hls = s->priv_data;
99
100 av_dict_free(&hls->enc_opts);
101
102 av_freep(&hls->key_basename);
103 }
104
105 static int dict_set_bin(AVDictionary **dict, const char *key,
106 uint8_t *buf, size_t len)
107 {
108 char hex[33];
109
110 ff_data_to_hex(hex, buf, len, 0);
111 hex[32] = '\0';
112
113 return av_dict_set(dict, key, hex, 0);
114 }
115
116 static int setup_encryption(AVFormatContext *s)
117 {
118 HLSContext *hls = s->priv_data;
119 AVIOContext *out = NULL;
120 int len, ret;
121 uint8_t buf[16];
122 uint8_t *k = NULL;
123
124 len = strlen(hls->basename) + 4 + 1;
125 hls->key_basename = av_mallocz(len);
126 if (!hls->key_basename)
127 return AVERROR(ENOMEM);
128
129 av_strlcpy(hls->key_basename, hls->basename + 7, len);
130 av_strlcat(hls->key_basename, ".key", len);
131
132 if (hls->key) {
133 if (hls->key_len != 16) {
134 av_log(s, AV_LOG_ERROR,
135 "Invalid key size %d, expected 16-bytes hex-coded key\n",
136 hls->key_len);
137 return AVERROR(EINVAL);
138 }
139
140 if ((ret = dict_set_bin(&hls->enc_opts, "key", hls->key, hls->key_len)) < 0)
141 return ret;
142 k = hls->key;
143 } else {
144 if (hls->start_sequence < 0) {
145 ret = s->io_open(s, &out, hls->key_basename, AVIO_FLAG_READ, NULL);
146 if (ret < 0) {
147 av_log(s, AV_LOG_WARNING,
148 "Cannot recover the key, generating a new one.\n");
149 } else {
150 avio_read(out, buf, 16);
151 k = buf;
152 avio_close(out);
153 }
154 }
155 if (!k) {
156 if ((ret = randomize(buf, sizeof(buf))) < 0) {
157 av_log(s, AV_LOG_ERROR, "Cannot generate a strong random key\n");
158 return ret;
159 }
160 }
161
162 if ((ret = dict_set_bin(&hls->enc_opts, "key", buf, sizeof(buf))) < 0)
163 return ret;
164 k = buf;
165 }
166
167 if (hls->iv) {
168 if (hls->iv_len != 16) {
169 av_log(s, AV_LOG_ERROR,
170 "Invalid key size %d, expected 16-bytes hex-coded initialization vector\n",
171 hls->iv_len);
172 return AVERROR(EINVAL);
173 }
174
175 if ((ret = dict_set_bin(&hls->enc_opts, "iv", hls->iv, hls->iv_len)) < 0)
176 return ret;
177 }
178
179 if ((ret = s->io_open(s, &out, hls->key_basename, AVIO_FLAG_WRITE, NULL)) < 0)
180 return ret;
181
182 avio_write(out, k, 16);
183
184 avio_close(out);
185
186 return 0;
187 }
188
189 static int hls_mux_init(AVFormatContext *s)
190 {
191 HLSContext *hls = s->priv_data;
192 AVFormatContext *oc;
193 int i;
194
195 hls->avf = oc = avformat_alloc_context();
196 if (!oc)
197 return AVERROR(ENOMEM);
198
199 oc->oformat = hls->oformat;
200 oc->interrupt_callback = s->interrupt_callback;
201 oc->opaque = s->opaque;
202 oc->io_open = s->io_open;
203 oc->io_close = s->io_close;
204
205 for (i = 0; i < s->nb_streams; i++) {
206 AVStream *st;
207 if (!(st = avformat_new_stream(oc, NULL)))
208 return AVERROR(ENOMEM);
209 avcodec_parameters_copy(st->codecpar, s->streams[i]->codecpar);
210 st->sample_aspect_ratio = s->streams[i]->sample_aspect_ratio;
211 st->time_base = s->streams[i]->time_base;
212 }
213
214 return 0;
215 }
216
217 static int append_entry(HLSContext *hls, int64_t duration, const char *name)
218 {
219 ListEntry *en = av_malloc(sizeof(*en));
220
221 if (!en)
222 return AVERROR(ENOMEM);
223
224 av_strlcpy(en->name, name, sizeof(en->name));
225
226 en->duration = duration;
227 en->next = NULL;
228
229 if (!hls->list)
230 hls->list = en;
231 else
232 hls->end_list->next = en;
233
234 hls->end_list = en;
235
236 if (hls->nb_entries >= hls->size) {
237 en = hls->list;
238 hls->list = en->next;
239 av_free(en);
240 } else
241 hls->nb_entries++;
242
243 hls->sequence++;
244
245 return 0;
246 }
247
248 static void free_entries(HLSContext *hls)
249 {
250 ListEntry *p = hls->list, *en;
251
252 while(p) {
253 en = p;
254 p = p->next;
255 av_free(en);
256 }
257 }
258
259 static int hls_window(AVFormatContext *s, int last)
260 {
261 HLSContext *hls = s->priv_data;
262 ListEntry *en;
263 int64_t target_duration = 0;
264 int ret = 0;
265 AVIOContext *out = NULL;
266 char temp_filename[1024];
267 int64_t sequence = FFMAX(hls->start_sequence, hls->sequence - hls->size);
268
269 snprintf(temp_filename, sizeof(temp_filename), "%s.tmp", s->filename);
270 if ((ret = s->io_open(s, &out, temp_filename, AVIO_FLAG_WRITE, NULL)) < 0)
271 goto fail;
272
273 for (en = hls->list; en; en = en->next) {
274 if (target_duration < en->duration)
275 target_duration = en->duration;
276 }
277
278 avio_printf(out, "#EXTM3U\n");
279 avio_printf(out, "#EXT-X-VERSION:%d\n", hls->version);
280 if (hls->allowcache == 0 || hls->allowcache == 1) {
281 avio_printf(out, "#EXT-X-ALLOW-CACHE:%s\n", hls->allowcache == 0 ? "NO" : "YES");
282 }
283 avio_printf(out, "#EXT-X-TARGETDURATION:%"PRId64"\n",
284 av_rescale_rnd(target_duration, 1, AV_TIME_BASE,
285 AV_ROUND_UP));
286 avio_printf(out, "#EXT-X-MEDIA-SEQUENCE:%"PRId64"\n", sequence);
287
288 av_log(s, AV_LOG_VERBOSE, "EXT-X-MEDIA-SEQUENCE:%"PRId64"\n",
289 sequence);
290
291 for (en = hls->list; en; en = en->next) {
292 if (hls->encrypt) {
293 char *key_url;
294
295 if (hls->key_url)
296 key_url = hls->key_url;
297 else
298 key_url = hls->baseurl;
299
300 avio_printf(out, "#EXT-X-KEY:METHOD=AES-128");
301 avio_printf(out, ",URI=\"");
302 if (key_url)
303 avio_printf(out, "%s", key_url);
304 avio_printf(out, "%s\"", av_basename(hls->key_basename));
305 if (hls->iv)
306 avio_printf(out, ",IV=\"0x%s\"", hls->iv);
307 avio_printf(out, "\n");
308 }
309
310 if (hls->version > 2)
311 avio_printf(out, "#EXTINF:%f\n",
312 (double)en->duration / AV_TIME_BASE);
313 else
314 avio_printf(out, "#EXTINF:%"PRId64",\n",
315 av_rescale(en->duration, 1, AV_TIME_BASE));
316 if (hls->baseurl)
317 avio_printf(out, "%s", hls->baseurl);
318 avio_printf(out, "%s\n", en->name);
319 }
320
321 if (last)
322 avio_printf(out, "#EXT-X-ENDLIST\n");
323
324 fail:
325 ff_format_io_close(s, &out);
326 if (ret >= 0)
327 ff_rename(temp_filename, s->filename);
328 return ret;
329 }
330
331 static int hls_start(AVFormatContext *s)
332 {
333 HLSContext *c = s->priv_data;
334 AVFormatContext *oc = c->avf;
335 int err = 0;
336 AVDictionary *opts = NULL;
337
338
339 if (av_get_frame_filename(oc->filename, sizeof(oc->filename),
340 c->basename, c->wrap ? c->sequence % c->wrap : c->sequence) < 0)
341 return AVERROR(EINVAL);
342 c->number++;
343
344 if (c->encrypt) {
345 if ((err = av_dict_copy(&opts, c->enc_opts, 0)) < 0)
346 return err;
347 if (!c->iv) {
348 uint8_t iv[16] = { 0 };
349 char buf[33];
350
351 AV_WB64(iv + 8, c->sequence);
352 ff_data_to_hex(buf, iv, sizeof(iv), 0);
353 buf[32] = '\0';
354
355 if ((err = av_dict_set(&opts, "iv", buf, 0)) < 0)
356 goto fail;
357 }
358 }
359
360 if ((err = s->io_open(s, &oc->pb, oc->filename, AVIO_FLAG_WRITE, &opts)) < 0)
361 return err;
362
363 if (oc->oformat->priv_class && oc->priv_data)
364 av_opt_set(oc->priv_data, "mpegts_flags", "resend_headers", 0);
365
366 fail:
367 av_dict_free(&opts);
368
369 return err;
370 }
371
372 static int read_chomp_line(AVIOContext *s, char *buf, int maxlen)
373 {
374 int len = ff_get_line(s, buf, maxlen);
375 while (len > 0 && av_isspace(buf[len - 1]))
376 buf[--len] = '\0';
377 return len;
378 }
379
380 static int hls_recover(AVFormatContext *s)
381 {
382 HLSContext *hls = s->priv_data;
383 char line[1024];
384 AVIOContext *io;
385 const char *ptr;
386 int ret, is_segment = 0;
387 int64_t duration = 0;
388
389 ret = s->io_open(s, &io, s->filename, AVIO_FLAG_READ, NULL);
390 if (ret < 0) {
391 av_log(s, AV_LOG_WARNING,
392 "Cannot recover the playlist, generating a new one.\n");
393 hls->start_sequence = 0;
394 hls->sequence = 0;
395 return 0;
396 }
397
398 read_chomp_line(io, line, sizeof(line));
399 if (strcmp(line, "#EXTM3U")) {
400 av_log(s, AV_LOG_ERROR,
401 "The playlist file is present but unparsable."
402 " Please remove it.\n");
403 return AVERROR_INVALIDDATA;
404 }
405
406 while (!io->eof_reached) {
407 read_chomp_line(io, line, sizeof(line));
408 if (av_strstart(line, "#EXT-X-MEDIA-SEQUENCE:", &ptr)) {
409 hls->sequence = hls->start_sequence = atoi(ptr);
410 } else if (av_strstart(line, "#EXTINF:", &ptr)) {
411 is_segment = 1;
412 duration = atof(ptr) * AV_TIME_BASE;
413 } else if (av_strstart(line, "#", NULL)) {
414 continue;
415 } else if (line[0]) {
416 if (is_segment) {
417 append_entry(hls, duration, av_basename(line));
418 is_segment = 0;
419 }
420 }
421 }
422
423 return 0;
424 }
425
426 static int hls_setup(AVFormatContext *s)
427 {
428 HLSContext *hls = s->priv_data;
429 const char *pattern = "%d.ts";
430 int basename_size = strlen(s->filename) + strlen(pattern) + 1;
431 char *p;
432 int ret;
433
434 if (hls->encrypt)
435 basename_size += 7;
436
437 hls->basename = av_mallocz(basename_size);
438 if (!hls->basename)
439 return AVERROR(ENOMEM);
440
441 // TODO: support protocol nesting?
442 if (hls->encrypt)
443 strcpy(hls->basename, "crypto:");
444
445 av_strlcat(hls->basename, s->filename, basename_size);
446
447 p = strrchr(hls->basename, '.');
448
449 if (p)
450 *p = '\0';
451
452 if (hls->encrypt) {
453 ret = setup_encryption(s);
454 if (ret < 0)
455 return ret;
456 }
457
458 if (hls->start_sequence < 0) {
459 ret = hls_recover(s);
460 if (ret < 0)
461 return ret;
462 }
463
464 av_strlcat(hls->basename, pattern, basename_size);
465
466 return 0;
467 }
468
469 static int hls_write_header(AVFormatContext *s)
470 {
471 HLSContext *hls = s->priv_data;
472 int ret, i;
473
474 hls->sequence = hls->start_sequence;
475 hls->recording_time = hls->time * AV_TIME_BASE;
476 hls->start_pts = AV_NOPTS_VALUE;
477
478 for (i = 0; i < s->nb_streams; i++)
479 hls->has_video +=
480 s->streams[i]->codecpar->codec_type == AVMEDIA_TYPE_VIDEO;
481
482 if (hls->has_video > 1)
483 av_log(s, AV_LOG_WARNING,
484 "More than a single video stream present, "
485 "expect issues decoding it.\n");
486
487 hls->oformat = av_guess_format("mpegts", NULL, NULL);
488
489 if (!hls->oformat) {
490 ret = AVERROR_MUXER_NOT_FOUND;
491 goto fail;
492 }
493
494 if ((ret = hls_setup(s)) < 0)
495 goto fail;
496
497 if ((ret = hls_mux_init(s)) < 0)
498 goto fail;
499
500 if ((ret = hls_start(s)) < 0)
501 goto fail;
502
503 if ((ret = avformat_write_header(hls->avf, NULL)) < 0)
504 return ret;
505
506
507 fail:
508 if (ret) {
509 av_free(hls->basename);
510 if (hls->avf)
511 avformat_free_context(hls->avf);
512
513 free_encryption(s);
514 }
515 return ret;
516 }
517
518 static int hls_write_packet(AVFormatContext *s, AVPacket *pkt)
519 {
520 HLSContext *hls = s->priv_data;
521 AVFormatContext *oc = hls->avf;
522 AVStream *st = s->streams[pkt->stream_index];
523 int64_t end_pts = hls->recording_time * hls->number;
524 int64_t pts = av_rescale_q(pkt->pts, st->time_base, AV_TIME_BASE_Q);
525 int ret, can_split = 1;
526
527 if (hls->start_pts == AV_NOPTS_VALUE) {
528 hls->start_pts = pts;
529 hls->end_pts = pts;
530 }
531
532 if (hls->has_video) {
533 can_split = st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO &&
534 pkt->flags & AV_PKT_FLAG_KEY;
535 }
536 if (pkt->pts == AV_NOPTS_VALUE)
537 can_split = 0;
538 else
539 hls->duration = pts - hls->end_pts;
540
541 if (can_split && pts - hls->start_pts >= end_pts) {
542 ret = append_entry(hls, hls->duration, av_basename(hls->avf->filename));
543 if (ret)
544 return ret;
545
546 hls->end_pts = pts;
547 hls->duration = 0;
548
549 av_write_frame(oc, NULL); /* Flush any buffered data */
550 ff_format_io_close(s, &oc->pb);
551
552 ret = hls_start(s);
553
554 if (ret)
555 return ret;
556
557 oc = hls->avf;
558
559 if ((ret = hls_window(s, 0)) < 0)
560 return ret;
561 }
562
563 ret = ff_write_chained(oc, pkt->stream_index, pkt, s);
564
565 return ret;
566 }
567
568 static int hls_write_trailer(struct AVFormatContext *s)
569 {
570 HLSContext *hls = s->priv_data;
571 AVFormatContext *oc = hls->avf;
572
573 av_write_trailer(oc);
574 ff_format_io_close(s, &oc->pb);
575 avformat_free_context(oc);
576 av_free(hls->basename);
577 append_entry(hls, hls->duration, av_basename(hls->avf->filename));
578 hls_window(s, 1);
579
580 free_entries(hls);
581 free_encryption(s);
582 return 0;
583 }
584
585 #define OFFSET(x) offsetof(HLSContext, x)
586 #define E AV_OPT_FLAG_ENCODING_PARAM
587 static const AVOption options[] = {
588 {"start_number", "first number in the sequence", OFFSET(start_sequence),AV_OPT_TYPE_INT64, {.i64 = 0}, -1, INT64_MAX, E, "start_number"},
589 {"recover", "If there is already a m3u8 file in the path, populate the sequence from it", 0, AV_OPT_TYPE_CONST, {.i64 = -1}, 0, 0, E, "start_number"},
590 {"hls_time", "segment length in seconds", OFFSET(time), AV_OPT_TYPE_FLOAT, {.dbl = 2}, 0, FLT_MAX, E},
591 {"hls_list_size", "maximum number of playlist entries", OFFSET(size), AV_OPT_TYPE_INT, {.i64 = 5}, 0, INT_MAX, E},
592 {"hls_wrap", "number after which the index wraps", OFFSET(wrap), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, E},
593 {"hls_allow_cache", "explicitly set whether the client MAY (1) or MUST NOT (0) cache media segments", OFFSET(allowcache), AV_OPT_TYPE_INT, {.i64 = -1}, INT_MIN, INT_MAX, E},
594 {"hls_base_url", "url to prepend to each playlist entry", OFFSET(baseurl), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, E},
595 {"hls_version", "protocol version", OFFSET(version), AV_OPT_TYPE_INT, {.i64 = 3}, 2, 3, E},
596 {"hls_enc", "AES128 encryption support", OFFSET(encrypt), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, E},
597 {"hls_enc_key", "use the specified hex-coded 16byte key to encrypt the segments", OFFSET(key), AV_OPT_TYPE_BINARY, .flags = E},
598 {"hls_enc_key_url", "url to access the key to decrypt the segments", OFFSET(key_url), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, E},
599 {"hls_enc_iv", "use the specified hex-coded 16byte initialization vector", OFFSET(iv), AV_OPT_TYPE_BINARY, .flags = E},
600 { NULL },
601 };
602
603 static const AVClass hls_class = {
604 .class_name = "hls muxer",
605 .item_name = av_default_item_name,
606 .option = options,
607 .version = LIBAVUTIL_VERSION_INT,
608 };
609
610
611 AVOutputFormat ff_hls_muxer = {
612 .name = "hls",
613 .long_name = NULL_IF_CONFIG_SMALL("Apple HTTP Live Streaming"),
614 .extensions = "m3u8",
615 .priv_data_size = sizeof(HLSContext),
616 .audio_codec = AV_CODEC_ID_AAC,
617 .video_codec = AV_CODEC_ID_H264,
618 .flags = AVFMT_NOFILE | AVFMT_ALLOW_FLUSH,
619 .write_header = hls_write_header,
620 .write_packet = hls_write_packet,
621 .write_trailer = hls_write_trailer,
622 .priv_class = &hls_class,
623 };