rtpenc: Allow requesting H264 RTP packetization mode 0
[libav.git] / libavformat / sdp.c
1 /*
2 * copyright (c) 2007 Luca Abeni
3 *
4 * This file is part of Libav.
5 *
6 * Libav is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * Libav is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 #include <string.h>
22 #include "libavutil/avstring.h"
23 #include "libavutil/base64.h"
24 #include "libavutil/dict.h"
25 #include "libavutil/parseutils.h"
26 #include "libavutil/opt.h"
27 #include "libavcodec/xiph.h"
28 #include "libavcodec/mpeg4audio.h"
29 #include "avformat.h"
30 #include "internal.h"
31 #include "avc.h"
32 #include "rtp.h"
33 #if CONFIG_NETWORK
34 #include "network.h"
35 #endif
36
37 #if CONFIG_RTP_MUXER
38 #define MAX_EXTRADATA_SIZE ((INT_MAX - 10) / 2)
39
40 struct sdp_session_level {
41 int sdp_version; /**< protocol version (currently 0) */
42 int id; /**< session ID */
43 int version; /**< session version */
44 int start_time; /**< session start time (NTP time, in seconds),
45 or 0 in case of permanent session */
46 int end_time; /**< session end time (NTP time, in seconds),
47 or 0 if the session is not bounded */
48 int ttl; /**< TTL, in case of multicast stream */
49 const char *user; /**< username of the session's creator */
50 const char *src_addr; /**< IP address of the machine from which the session was created */
51 const char *src_type; /**< address type of src_addr */
52 const char *dst_addr; /**< destination IP address (can be multicast) */
53 const char *dst_type; /**< destination IP address type */
54 const char *name; /**< session name (can be an empty string) */
55 };
56
57 static void sdp_write_address(char *buff, int size, const char *dest_addr,
58 const char *dest_type, int ttl)
59 {
60 if (dest_addr) {
61 if (!dest_type)
62 dest_type = "IP4";
63 if (ttl > 0 && !strcmp(dest_type, "IP4")) {
64 /* The TTL should only be specified for IPv4 multicast addresses,
65 * not for IPv6. */
66 av_strlcatf(buff, size, "c=IN %s %s/%d\r\n", dest_type, dest_addr, ttl);
67 } else {
68 av_strlcatf(buff, size, "c=IN %s %s\r\n", dest_type, dest_addr);
69 }
70 }
71 }
72
73 static void sdp_write_header(char *buff, int size, struct sdp_session_level *s)
74 {
75 av_strlcatf(buff, size, "v=%d\r\n"
76 "o=- %d %d IN %s %s\r\n"
77 "s=%s\r\n",
78 s->sdp_version,
79 s->id, s->version, s->src_type, s->src_addr,
80 s->name);
81 sdp_write_address(buff, size, s->dst_addr, s->dst_type, s->ttl);
82 av_strlcatf(buff, size, "t=%d %d\r\n"
83 "a=tool:libavformat " AV_STRINGIFY(LIBAVFORMAT_VERSION) "\r\n",
84 s->start_time, s->end_time);
85 }
86
87 #if CONFIG_NETWORK
88 static int resolve_destination(char *dest_addr, int size, char *type,
89 int type_size)
90 {
91 struct addrinfo hints = { 0 }, *ai;
92 int is_multicast;
93
94 av_strlcpy(type, "IP4", type_size);
95 if (!dest_addr[0])
96 return 0;
97
98 /* Resolve the destination, since it must be written
99 * as a numeric IP address in the SDP. */
100
101 if (getaddrinfo(dest_addr, NULL, &hints, &ai))
102 return 0;
103 getnameinfo(ai->ai_addr, ai->ai_addrlen, dest_addr, size,
104 NULL, 0, NI_NUMERICHOST);
105 #ifdef AF_INET6
106 if (ai->ai_family == AF_INET6)
107 av_strlcpy(type, "IP6", type_size);
108 #endif
109 is_multicast = ff_is_multicast_address(ai->ai_addr);
110 freeaddrinfo(ai);
111 return is_multicast;
112 }
113 #else
114 static int resolve_destination(char *dest_addr, int size, char *type,
115 int type_size)
116 {
117 return 0;
118 }
119 #endif
120
121 static int sdp_get_address(char *dest_addr, int size, int *ttl, const char *url)
122 {
123 int port;
124 const char *p;
125 char proto[32];
126
127 av_url_split(proto, sizeof(proto), NULL, 0, dest_addr, size, &port, NULL, 0, url);
128
129 *ttl = 0;
130
131 if (strcmp(proto, "rtp")) {
132 /* The url isn't for the actual rtp sessions,
133 * don't parse out anything else than the destination.
134 */
135 return 0;
136 }
137
138 p = strchr(url, '?');
139 if (p) {
140 char buff[64];
141
142 if (av_find_info_tag(buff, sizeof(buff), "ttl", p)) {
143 *ttl = strtol(buff, NULL, 10);
144 } else {
145 *ttl = 5;
146 }
147 }
148
149 return port;
150 }
151
152 #define MAX_PSET_SIZE 1024
153 static char *extradata2psets(AVCodecContext *c)
154 {
155 char *psets, *p;
156 const uint8_t *r;
157 const char *pset_string = "; sprop-parameter-sets=";
158 uint8_t *orig_extradata = NULL;
159 int orig_extradata_size = 0;
160
161 if (c->extradata_size > MAX_EXTRADATA_SIZE) {
162 av_log(c, AV_LOG_ERROR, "Too much extradata!\n");
163
164 return NULL;
165 }
166 if (c->extradata[0] == 1) {
167 uint8_t *dummy_p;
168 int dummy_int;
169 AVBitStreamFilterContext *bsfc= av_bitstream_filter_init("h264_mp4toannexb");
170
171 if (!bsfc) {
172 av_log(c, AV_LOG_ERROR, "Cannot open the h264_mp4toannexb BSF!\n");
173
174 return NULL;
175 }
176
177 orig_extradata_size = c->extradata_size;
178 orig_extradata = av_mallocz(orig_extradata_size +
179 FF_INPUT_BUFFER_PADDING_SIZE);
180 if (!orig_extradata) {
181 av_bitstream_filter_close(bsfc);
182 return NULL;
183 }
184 memcpy(orig_extradata, c->extradata, orig_extradata_size);
185 av_bitstream_filter_filter(bsfc, c, NULL, &dummy_p, &dummy_int, NULL, 0, 0);
186 av_bitstream_filter_close(bsfc);
187 }
188
189 psets = av_mallocz(MAX_PSET_SIZE);
190 if (psets == NULL) {
191 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the parameter sets.\n");
192 av_free(orig_extradata);
193 return NULL;
194 }
195 memcpy(psets, pset_string, strlen(pset_string));
196 p = psets + strlen(pset_string);
197 r = ff_avc_find_startcode(c->extradata, c->extradata + c->extradata_size);
198 while (r < c->extradata + c->extradata_size) {
199 const uint8_t *r1;
200 uint8_t nal_type;
201
202 while (!*(r++));
203 nal_type = *r & 0x1f;
204 r1 = ff_avc_find_startcode(r, c->extradata + c->extradata_size);
205 if (nal_type != 7 && nal_type != 8) { /* Only output SPS and PPS */
206 r = r1;
207 continue;
208 }
209 if (p != (psets + strlen(pset_string))) {
210 *p = ',';
211 p++;
212 }
213 if (av_base64_encode(p, MAX_PSET_SIZE - (p - psets), r, r1 - r) == NULL) {
214 av_log(c, AV_LOG_ERROR, "Cannot Base64-encode %td %td!\n", MAX_PSET_SIZE - (p - psets), r1 - r);
215 av_free(psets);
216
217 return NULL;
218 }
219 p += strlen(p);
220 r = r1;
221 }
222 if (orig_extradata) {
223 av_free(c->extradata);
224 c->extradata = orig_extradata;
225 c->extradata_size = orig_extradata_size;
226 }
227
228 return psets;
229 }
230
231 static char *extradata2config(AVCodecContext *c)
232 {
233 char *config;
234
235 if (c->extradata_size > MAX_EXTRADATA_SIZE) {
236 av_log(c, AV_LOG_ERROR, "Too much extradata!\n");
237
238 return NULL;
239 }
240 config = av_malloc(10 + c->extradata_size * 2);
241 if (config == NULL) {
242 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the config info.\n");
243 return NULL;
244 }
245 memcpy(config, "; config=", 9);
246 ff_data_to_hex(config + 9, c->extradata, c->extradata_size, 0);
247 config[9 + c->extradata_size * 2] = 0;
248
249 return config;
250 }
251
252 static char *xiph_extradata2config(AVCodecContext *c)
253 {
254 char *config, *encoded_config;
255 uint8_t *header_start[3];
256 int headers_len, header_len[3], config_len;
257 int first_header_size;
258
259 switch (c->codec_id) {
260 case CODEC_ID_THEORA:
261 first_header_size = 42;
262 break;
263 case CODEC_ID_VORBIS:
264 first_header_size = 30;
265 break;
266 default:
267 av_log(c, AV_LOG_ERROR, "Unsupported Xiph codec ID\n");
268 return NULL;
269 }
270
271 if (avpriv_split_xiph_headers(c->extradata, c->extradata_size,
272 first_header_size, header_start,
273 header_len) < 0) {
274 av_log(c, AV_LOG_ERROR, "Extradata corrupt.\n");
275 return NULL;
276 }
277
278 headers_len = header_len[0] + header_len[2];
279 config_len = 4 + // count
280 3 + // ident
281 2 + // packet size
282 1 + // header count
283 2 + // header size
284 headers_len; // and the rest
285
286 config = av_malloc(config_len);
287 if (!config)
288 goto xiph_fail;
289
290 encoded_config = av_malloc(AV_BASE64_SIZE(config_len));
291 if (!encoded_config) {
292 av_free(config);
293 goto xiph_fail;
294 }
295
296 config[0] = config[1] = config[2] = 0;
297 config[3] = 1;
298 config[4] = (RTP_XIPH_IDENT >> 16) & 0xff;
299 config[5] = (RTP_XIPH_IDENT >> 8) & 0xff;
300 config[6] = (RTP_XIPH_IDENT ) & 0xff;
301 config[7] = (headers_len >> 8) & 0xff;
302 config[8] = headers_len & 0xff;
303 config[9] = 2;
304 config[10] = header_len[0];
305 config[11] = 0; // size of comment header; nonexistent
306 memcpy(config + 12, header_start[0], header_len[0]);
307 memcpy(config + 12 + header_len[0], header_start[2], header_len[2]);
308
309 av_base64_encode(encoded_config, AV_BASE64_SIZE(config_len),
310 config, config_len);
311 av_free(config);
312
313 return encoded_config;
314
315 xiph_fail:
316 av_log(c, AV_LOG_ERROR,
317 "Not enough memory for configuration string\n");
318 return NULL;
319 }
320
321 static int latm_context2profilelevel(AVCodecContext *c)
322 {
323 /* MP4A-LATM
324 * The RTP payload format specification is described in RFC 3016
325 * The encoding specifications are provided in ISO/IEC 14496-3 */
326
327 int profile_level = 0x2B;
328
329 /* TODO: AAC Profile only supports AAC LC Object Type.
330 * Different Object Types should implement different Profile Levels */
331
332 if (c->sample_rate <= 24000) {
333 if (c->channels <= 2)
334 profile_level = 0x28; // AAC Profile, Level 1
335 } else if (c->sample_rate <= 48000) {
336 if (c->channels <= 2) {
337 profile_level = 0x29; // AAC Profile, Level 2
338 } else if (c->channels <= 5) {
339 profile_level = 0x2A; // AAC Profile, Level 4
340 }
341 } else if (c->sample_rate <= 96000) {
342 if (c->channels <= 5) {
343 profile_level = 0x2B; // AAC Profile, Level 5
344 }
345 }
346
347 return profile_level;
348 }
349
350 static char *latm_context2config(AVCodecContext *c)
351 {
352 /* MP4A-LATM
353 * The RTP payload format specification is described in RFC 3016
354 * The encoding specifications are provided in ISO/IEC 14496-3 */
355
356 uint8_t config_byte[6];
357 int rate_index;
358 char *config;
359
360 for (rate_index = 0; rate_index < 16; rate_index++)
361 if (avpriv_mpeg4audio_sample_rates[rate_index] == c->sample_rate)
362 break;
363 if (rate_index == 16) {
364 av_log(c, AV_LOG_ERROR, "Unsupported sample rate\n");
365 return NULL;
366 }
367
368 config_byte[0] = 0x40;
369 config_byte[1] = 0;
370 config_byte[2] = 0x20 | rate_index;
371 config_byte[3] = c->channels << 4;
372 config_byte[4] = 0x3f;
373 config_byte[5] = 0xc0;
374
375 config = av_malloc(6*2+1);
376 if (!config) {
377 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the config info.\n");
378 return NULL;
379 }
380 ff_data_to_hex(config, config_byte, 6, 1);
381 config[12] = 0;
382
383 return config;
384 }
385
386 static char *sdp_write_media_attributes(char *buff, int size, AVCodecContext *c, int payload_type, AVFormatContext *fmt)
387 {
388 char *config = NULL;
389
390 switch (c->codec_id) {
391 case CODEC_ID_H264: {
392 int mode = 1;
393 if (fmt && fmt->oformat->priv_class &&
394 av_opt_flag_is_set(fmt->priv_data, "rtpflags", "h264_mode0"))
395 mode = 0;
396 if (c->extradata_size) {
397 config = extradata2psets(c);
398 }
399 av_strlcatf(buff, size, "a=rtpmap:%d H264/90000\r\n"
400 "a=fmtp:%d packetization-mode=%d%s\r\n",
401 payload_type,
402 payload_type, mode, config ? config : "");
403 break;
404 }
405 case CODEC_ID_H263:
406 case CODEC_ID_H263P:
407 /* a=framesize is required by 3GPP TS 26.234 (PSS). It
408 * actually specifies the maximum video size, but we only know
409 * the current size. This is required for playback on Android
410 * stagefright and on Samsung bada. */
411 if (!fmt || !fmt->oformat->priv_class ||
412 !av_opt_flag_is_set(fmt->priv_data, "rtpflags", "rfc2190") ||
413 c->codec_id == CODEC_ID_H263P)
414 av_strlcatf(buff, size, "a=rtpmap:%d H263-2000/90000\r\n"
415 "a=framesize:%d %d-%d\r\n",
416 payload_type,
417 payload_type, c->width, c->height);
418 break;
419 case CODEC_ID_MPEG4:
420 if (c->extradata_size) {
421 config = extradata2config(c);
422 }
423 av_strlcatf(buff, size, "a=rtpmap:%d MP4V-ES/90000\r\n"
424 "a=fmtp:%d profile-level-id=1%s\r\n",
425 payload_type,
426 payload_type, config ? config : "");
427 break;
428 case CODEC_ID_AAC:
429 if (fmt && fmt->oformat->priv_class &&
430 av_opt_flag_is_set(fmt->priv_data, "rtpflags", "latm")) {
431 config = latm_context2config(c);
432 if (!config)
433 return NULL;
434 av_strlcatf(buff, size, "a=rtpmap:%d MP4A-LATM/%d/%d\r\n"
435 "a=fmtp:%d profile-level-id=%d;cpresent=0;config=%s\r\n",
436 payload_type, c->sample_rate, c->channels,
437 payload_type, latm_context2profilelevel(c), config);
438 } else {
439 if (c->extradata_size) {
440 config = extradata2config(c);
441 } else {
442 /* FIXME: maybe we can forge config information based on the
443 * codec parameters...
444 */
445 av_log(c, AV_LOG_ERROR, "AAC with no global headers is currently not supported.\n");
446 return NULL;
447 }
448 if (config == NULL) {
449 return NULL;
450 }
451 av_strlcatf(buff, size, "a=rtpmap:%d MPEG4-GENERIC/%d/%d\r\n"
452 "a=fmtp:%d profile-level-id=1;"
453 "mode=AAC-hbr;sizelength=13;indexlength=3;"
454 "indexdeltalength=3%s\r\n",
455 payload_type, c->sample_rate, c->channels,
456 payload_type, config);
457 }
458 break;
459 case CODEC_ID_PCM_S16BE:
460 if (payload_type >= RTP_PT_PRIVATE)
461 av_strlcatf(buff, size, "a=rtpmap:%d L16/%d/%d\r\n",
462 payload_type,
463 c->sample_rate, c->channels);
464 break;
465 case CODEC_ID_PCM_MULAW:
466 if (payload_type >= RTP_PT_PRIVATE)
467 av_strlcatf(buff, size, "a=rtpmap:%d PCMU/%d/%d\r\n",
468 payload_type,
469 c->sample_rate, c->channels);
470 break;
471 case CODEC_ID_PCM_ALAW:
472 if (payload_type >= RTP_PT_PRIVATE)
473 av_strlcatf(buff, size, "a=rtpmap:%d PCMA/%d/%d\r\n",
474 payload_type,
475 c->sample_rate, c->channels);
476 break;
477 case CODEC_ID_AMR_NB:
478 av_strlcatf(buff, size, "a=rtpmap:%d AMR/%d/%d\r\n"
479 "a=fmtp:%d octet-align=1\r\n",
480 payload_type, c->sample_rate, c->channels,
481 payload_type);
482 break;
483 case CODEC_ID_AMR_WB:
484 av_strlcatf(buff, size, "a=rtpmap:%d AMR-WB/%d/%d\r\n"
485 "a=fmtp:%d octet-align=1\r\n",
486 payload_type, c->sample_rate, c->channels,
487 payload_type);
488 break;
489 case CODEC_ID_VORBIS:
490 if (c->extradata_size)
491 config = xiph_extradata2config(c);
492 else
493 av_log(c, AV_LOG_ERROR, "Vorbis configuration info missing\n");
494 if (!config)
495 return NULL;
496
497 av_strlcatf(buff, size, "a=rtpmap:%d vorbis/%d/%d\r\n"
498 "a=fmtp:%d configuration=%s\r\n",
499 payload_type, c->sample_rate, c->channels,
500 payload_type, config);
501 break;
502 case CODEC_ID_THEORA: {
503 const char *pix_fmt;
504 if (c->extradata_size)
505 config = xiph_extradata2config(c);
506 else
507 av_log(c, AV_LOG_ERROR, "Theora configuation info missing\n");
508 if (!config)
509 return NULL;
510
511 switch (c->pix_fmt) {
512 case PIX_FMT_YUV420P:
513 pix_fmt = "YCbCr-4:2:0";
514 break;
515 case PIX_FMT_YUV422P:
516 pix_fmt = "YCbCr-4:2:2";
517 break;
518 case PIX_FMT_YUV444P:
519 pix_fmt = "YCbCr-4:4:4";
520 break;
521 default:
522 av_log(c, AV_LOG_ERROR, "Unsupported pixel format.\n");
523 return NULL;
524 }
525
526 av_strlcatf(buff, size, "a=rtpmap:%d theora/90000\r\n"
527 "a=fmtp:%d delivery-method=inline; "
528 "width=%d; height=%d; sampling=%s; "
529 "configuration=%s\r\n",
530 payload_type, payload_type,
531 c->width, c->height, pix_fmt, config);
532 break;
533 }
534 case CODEC_ID_VP8:
535 av_strlcatf(buff, size, "a=rtpmap:%d VP8/90000\r\n",
536 payload_type);
537 break;
538 case CODEC_ID_ADPCM_G722:
539 if (payload_type >= RTP_PT_PRIVATE)
540 av_strlcatf(buff, size, "a=rtpmap:%d G722/%d/%d\r\n",
541 payload_type,
542 8000, c->channels);
543 break;
544 case CODEC_ID_ADPCM_G726: {
545 if (payload_type >= RTP_PT_PRIVATE)
546 av_strlcatf(buff, size, "a=rtpmap:%d G726-%d/%d\r\n",
547 payload_type,
548 c->bits_per_coded_sample*8,
549 c->sample_rate);
550 break;
551 }
552 default:
553 /* Nothing special to do here... */
554 break;
555 }
556
557 av_free(config);
558
559 return buff;
560 }
561
562 void ff_sdp_write_media(char *buff, int size, AVCodecContext *c, const char *dest_addr, const char *dest_type, int port, int ttl, AVFormatContext *fmt)
563 {
564 const char *type;
565 int payload_type;
566
567 payload_type = ff_rtp_get_payload_type(fmt, c);
568
569 switch (c->codec_type) {
570 case AVMEDIA_TYPE_VIDEO : type = "video" ; break;
571 case AVMEDIA_TYPE_AUDIO : type = "audio" ; break;
572 case AVMEDIA_TYPE_SUBTITLE: type = "text" ; break;
573 default : type = "application"; break;
574 }
575
576 av_strlcatf(buff, size, "m=%s %d RTP/AVP %d\r\n", type, port, payload_type);
577 sdp_write_address(buff, size, dest_addr, dest_type, ttl);
578 if (c->bit_rate) {
579 av_strlcatf(buff, size, "b=AS:%d\r\n", c->bit_rate / 1000);
580 }
581
582 sdp_write_media_attributes(buff, size, c, payload_type, fmt);
583 }
584
585 int av_sdp_create(AVFormatContext *ac[], int n_files, char *buf, int size)
586 {
587 AVDictionaryEntry *title = av_dict_get(ac[0]->metadata, "title", NULL, 0);
588 struct sdp_session_level s = { 0 };
589 int i, j, port, ttl, is_multicast;
590 char dst[32], dst_type[5];
591
592 memset(buf, 0, size);
593 s.user = "-";
594 s.src_addr = "127.0.0.1"; /* FIXME: Properly set this */
595 s.src_type = "IP4";
596 s.name = title ? title->value : "No Name";
597
598 port = 0;
599 ttl = 0;
600 if (n_files == 1) {
601 port = sdp_get_address(dst, sizeof(dst), &ttl, ac[0]->filename);
602 is_multicast = resolve_destination(dst, sizeof(dst), dst_type,
603 sizeof(dst_type));
604 if (!is_multicast)
605 ttl = 0;
606 if (dst[0]) {
607 s.dst_addr = dst;
608 s.dst_type = dst_type;
609 s.ttl = ttl;
610 if (!strcmp(dst_type, "IP6")) {
611 s.src_addr = "::1";
612 s.src_type = "IP6";
613 }
614 }
615 }
616 sdp_write_header(buf, size, &s);
617
618 dst[0] = 0;
619 for (i = 0; i < n_files; i++) {
620 if (n_files != 1) {
621 port = sdp_get_address(dst, sizeof(dst), &ttl, ac[i]->filename);
622 is_multicast = resolve_destination(dst, sizeof(dst), dst_type,
623 sizeof(dst_type));
624 if (!is_multicast)
625 ttl = 0;
626 }
627 for (j = 0; j < ac[i]->nb_streams; j++) {
628 ff_sdp_write_media(buf, size,
629 ac[i]->streams[j]->codec, dst[0] ? dst : NULL,
630 dst_type, (port > 0) ? port + j * 2 : 0, ttl,
631 ac[i]);
632 if (port <= 0) {
633 av_strlcatf(buf, size,
634 "a=control:streamid=%d\r\n", i + j);
635 }
636 }
637 }
638
639 return 0;
640 }
641 #else
642 int av_sdp_create(AVFormatContext *ac[], int n_files, char *buf, int size)
643 {
644 return AVERROR(ENOSYS);
645 }
646
647 void ff_sdp_write_media(char *buff, int size, AVCodecContext *c, const char *dest_addr, const char *dest_type, int port, int ttl, AVFormatContext *fmt)
648 {
649 }
650 #endif