rtpenc: Add support for mp4 format h264
[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, *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 memset(&hints, 0, sizeof(hints));
102 if (getaddrinfo(dest_addr, NULL, &hints, &ai))
103 return 0;
104 getnameinfo(ai->ai_addr, ai->ai_addrlen, dest_addr, size,
105 NULL, 0, NI_NUMERICHOST);
106 #ifdef AF_INET6
107 if (ai->ai_family == AF_INET6)
108 av_strlcpy(type, "IP6", type_size);
109 #endif
110 is_multicast = ff_is_multicast_address(ai->ai_addr);
111 freeaddrinfo(ai);
112 return is_multicast;
113 }
114 #else
115 static int resolve_destination(char *dest_addr, int size, char *type,
116 int type_size)
117 {
118 return 0;
119 }
120 #endif
121
122 static int sdp_get_address(char *dest_addr, int size, int *ttl, const char *url)
123 {
124 int port;
125 const char *p;
126 char proto[32];
127
128 av_url_split(proto, sizeof(proto), NULL, 0, dest_addr, size, &port, NULL, 0, url);
129
130 *ttl = 0;
131
132 if (strcmp(proto, "rtp")) {
133 /* The url isn't for the actual rtp sessions,
134 * don't parse out anything else than the destination.
135 */
136 return 0;
137 }
138
139 p = strchr(url, '?');
140 if (p) {
141 char buff[64];
142
143 if (av_find_info_tag(buff, sizeof(buff), "ttl", p)) {
144 *ttl = strtol(buff, NULL, 10);
145 } else {
146 *ttl = 5;
147 }
148 }
149
150 return port;
151 }
152
153 #define MAX_PSET_SIZE 1024
154 static char *extradata2psets(AVCodecContext *c)
155 {
156 char *psets, *p;
157 const uint8_t *r;
158 const char *pset_string = "; sprop-parameter-sets=";
159
160 if (c->extradata_size > MAX_EXTRADATA_SIZE) {
161 av_log(c, AV_LOG_ERROR, "Too much extradata!\n");
162
163 return NULL;
164 }
165 if (c->extradata[0] == 1) {
166 uint8_t *dummy_p;
167 int dummy_int;
168 AVBitStreamFilterContext *bsfc= av_bitstream_filter_init("h264_mp4toannexb");
169
170 if (!bsfc) {
171 av_log(c, AV_LOG_ERROR, "Cannot open the h264_mp4toannexb BSF!\n");
172
173 return NULL;
174 }
175 av_bitstream_filter_filter(bsfc, c, NULL, &dummy_p, &dummy_int, NULL, 0, 0);
176 av_bitstream_filter_close(bsfc);
177 }
178
179 psets = av_mallocz(MAX_PSET_SIZE);
180 if (psets == NULL) {
181 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the parameter sets.\n");
182 return NULL;
183 }
184 memcpy(psets, pset_string, strlen(pset_string));
185 p = psets + strlen(pset_string);
186 r = ff_avc_find_startcode(c->extradata, c->extradata + c->extradata_size);
187 while (r < c->extradata + c->extradata_size) {
188 const uint8_t *r1;
189 uint8_t nal_type;
190
191 while (!*(r++));
192 nal_type = *r & 0x1f;
193 r1 = ff_avc_find_startcode(r, c->extradata + c->extradata_size);
194 if (nal_type != 7 && nal_type != 8) { /* Only output SPS and PPS */
195 r = r1;
196 continue;
197 }
198 if (p != (psets + strlen(pset_string))) {
199 *p = ',';
200 p++;
201 }
202 if (av_base64_encode(p, MAX_PSET_SIZE - (p - psets), r, r1 - r) == NULL) {
203 av_log(c, AV_LOG_ERROR, "Cannot Base64-encode %td %td!\n", MAX_PSET_SIZE - (p - psets), r1 - r);
204 av_free(psets);
205
206 return NULL;
207 }
208 p += strlen(p);
209 r = r1;
210 }
211
212 return psets;
213 }
214
215 static char *extradata2config(AVCodecContext *c)
216 {
217 char *config;
218
219 if (c->extradata_size > MAX_EXTRADATA_SIZE) {
220 av_log(c, AV_LOG_ERROR, "Too much extradata!\n");
221
222 return NULL;
223 }
224 config = av_malloc(10 + c->extradata_size * 2);
225 if (config == NULL) {
226 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the config info.\n");
227 return NULL;
228 }
229 memcpy(config, "; config=", 9);
230 ff_data_to_hex(config + 9, c->extradata, c->extradata_size, 0);
231 config[9 + c->extradata_size * 2] = 0;
232
233 return config;
234 }
235
236 static char *xiph_extradata2config(AVCodecContext *c)
237 {
238 char *config, *encoded_config;
239 uint8_t *header_start[3];
240 int headers_len, header_len[3], config_len;
241 int first_header_size;
242
243 switch (c->codec_id) {
244 case CODEC_ID_THEORA:
245 first_header_size = 42;
246 break;
247 case CODEC_ID_VORBIS:
248 first_header_size = 30;
249 break;
250 default:
251 av_log(c, AV_LOG_ERROR, "Unsupported Xiph codec ID\n");
252 return NULL;
253 }
254
255 if (avpriv_split_xiph_headers(c->extradata, c->extradata_size,
256 first_header_size, header_start,
257 header_len) < 0) {
258 av_log(c, AV_LOG_ERROR, "Extradata corrupt.\n");
259 return NULL;
260 }
261
262 headers_len = header_len[0] + header_len[2];
263 config_len = 4 + // count
264 3 + // ident
265 2 + // packet size
266 1 + // header count
267 2 + // header size
268 headers_len; // and the rest
269
270 config = av_malloc(config_len);
271 if (!config)
272 goto xiph_fail;
273
274 encoded_config = av_malloc(AV_BASE64_SIZE(config_len));
275 if (!encoded_config) {
276 av_free(config);
277 goto xiph_fail;
278 }
279
280 config[0] = config[1] = config[2] = 0;
281 config[3] = 1;
282 config[4] = (RTP_XIPH_IDENT >> 16) & 0xff;
283 config[5] = (RTP_XIPH_IDENT >> 8) & 0xff;
284 config[6] = (RTP_XIPH_IDENT ) & 0xff;
285 config[7] = (headers_len >> 8) & 0xff;
286 config[8] = headers_len & 0xff;
287 config[9] = 2;
288 config[10] = header_len[0];
289 config[11] = 0; // size of comment header; nonexistent
290 memcpy(config + 12, header_start[0], header_len[0]);
291 memcpy(config + 12 + header_len[0], header_start[2], header_len[2]);
292
293 av_base64_encode(encoded_config, AV_BASE64_SIZE(config_len),
294 config, config_len);
295 av_free(config);
296
297 return encoded_config;
298
299 xiph_fail:
300 av_log(c, AV_LOG_ERROR,
301 "Not enough memory for configuration string\n");
302 return NULL;
303 }
304
305 static int latm_context2profilelevel(AVCodecContext *c)
306 {
307 /* MP4A-LATM
308 * The RTP payload format specification is described in RFC 3016
309 * The encoding specifications are provided in ISO/IEC 14496-3 */
310
311 int profile_level = 0x2B;
312
313 /* TODO: AAC Profile only supports AAC LC Object Type.
314 * Different Object Types should implement different Profile Levels */
315
316 if (c->sample_rate <= 24000) {
317 if (c->channels <= 2)
318 profile_level = 0x28; // AAC Profile, Level 1
319 } else if (c->sample_rate <= 48000) {
320 if (c->channels <= 2) {
321 profile_level = 0x29; // AAC Profile, Level 2
322 } else if (c->channels <= 5) {
323 profile_level = 0x2A; // AAC Profile, Level 4
324 }
325 } else if (c->sample_rate <= 96000) {
326 if (c->channels <= 5) {
327 profile_level = 0x2B; // AAC Profile, Level 5
328 }
329 }
330
331 return profile_level;
332 }
333
334 static char *latm_context2config(AVCodecContext *c)
335 {
336 /* MP4A-LATM
337 * The RTP payload format specification is described in RFC 3016
338 * The encoding specifications are provided in ISO/IEC 14496-3 */
339
340 uint8_t config_byte[6];
341 int rate_index;
342 char *config;
343
344 for (rate_index = 0; rate_index < 16; rate_index++)
345 if (avpriv_mpeg4audio_sample_rates[rate_index] == c->sample_rate)
346 break;
347 if (rate_index == 16) {
348 av_log(c, AV_LOG_ERROR, "Unsupported sample rate\n");
349 return NULL;
350 }
351
352 config_byte[0] = 0x40;
353 config_byte[1] = 0;
354 config_byte[2] = 0x20 | rate_index;
355 config_byte[3] = c->channels << 4;
356 config_byte[4] = 0x3f;
357 config_byte[5] = 0xc0;
358
359 config = av_malloc(6*2+1);
360 if (!config) {
361 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the config info.\n");
362 return NULL;
363 }
364 ff_data_to_hex(config, config_byte, 6, 1);
365 config[12] = 0;
366
367 return config;
368 }
369
370 static char *sdp_write_media_attributes(char *buff, int size, AVCodecContext *c, int payload_type, AVFormatContext *fmt)
371 {
372 char *config = NULL;
373
374 switch (c->codec_id) {
375 case CODEC_ID_H264:
376 if (c->extradata_size) {
377 config = extradata2psets(c);
378 }
379 av_strlcatf(buff, size, "a=rtpmap:%d H264/90000\r\n"
380 "a=fmtp:%d packetization-mode=1%s\r\n",
381 payload_type,
382 payload_type, config ? config : "");
383 break;
384 case CODEC_ID_H263:
385 case CODEC_ID_H263P:
386 /* a=framesize is required by 3GPP TS 26.234 (PSS). It
387 * actually specifies the maximum video size, but we only know
388 * the current size. This is required for playback on Android
389 * stagefright and on Samsung bada. */
390 av_strlcatf(buff, size, "a=rtpmap:%d H263-2000/90000\r\n"
391 "a=framesize:%d %d-%d\r\n",
392 payload_type,
393 payload_type, c->width, c->height);
394 break;
395 case CODEC_ID_MPEG4:
396 if (c->extradata_size) {
397 config = extradata2config(c);
398 }
399 av_strlcatf(buff, size, "a=rtpmap:%d MP4V-ES/90000\r\n"
400 "a=fmtp:%d profile-level-id=1%s\r\n",
401 payload_type,
402 payload_type, config ? config : "");
403 break;
404 case CODEC_ID_AAC:
405 if (fmt && fmt->oformat->priv_class &&
406 av_opt_flag_is_set(fmt->priv_data, "rtpflags", "latm")) {
407 config = latm_context2config(c);
408 if (!config)
409 return NULL;
410 av_strlcatf(buff, size, "a=rtpmap:%d MP4A-LATM/%d/%d\r\n"
411 "a=fmtp:%d profile-level-id=%d;cpresent=0;config=%s\r\n",
412 payload_type, c->sample_rate, c->channels,
413 payload_type, latm_context2profilelevel(c), config);
414 } else {
415 if (c->extradata_size) {
416 config = extradata2config(c);
417 } else {
418 /* FIXME: maybe we can forge config information based on the
419 * codec parameters...
420 */
421 av_log(c, AV_LOG_ERROR, "AAC with no global headers is currently not supported.\n");
422 return NULL;
423 }
424 if (config == NULL) {
425 return NULL;
426 }
427 av_strlcatf(buff, size, "a=rtpmap:%d MPEG4-GENERIC/%d/%d\r\n"
428 "a=fmtp:%d profile-level-id=1;"
429 "mode=AAC-hbr;sizelength=13;indexlength=3;"
430 "indexdeltalength=3%s\r\n",
431 payload_type, c->sample_rate, c->channels,
432 payload_type, config);
433 }
434 break;
435 case CODEC_ID_PCM_S16BE:
436 if (payload_type >= RTP_PT_PRIVATE)
437 av_strlcatf(buff, size, "a=rtpmap:%d L16/%d/%d\r\n",
438 payload_type,
439 c->sample_rate, c->channels);
440 break;
441 case CODEC_ID_PCM_MULAW:
442 if (payload_type >= RTP_PT_PRIVATE)
443 av_strlcatf(buff, size, "a=rtpmap:%d PCMU/%d/%d\r\n",
444 payload_type,
445 c->sample_rate, c->channels);
446 break;
447 case CODEC_ID_PCM_ALAW:
448 if (payload_type >= RTP_PT_PRIVATE)
449 av_strlcatf(buff, size, "a=rtpmap:%d PCMA/%d/%d\r\n",
450 payload_type,
451 c->sample_rate, c->channels);
452 break;
453 case CODEC_ID_AMR_NB:
454 av_strlcatf(buff, size, "a=rtpmap:%d AMR/%d/%d\r\n"
455 "a=fmtp:%d octet-align=1\r\n",
456 payload_type, c->sample_rate, c->channels,
457 payload_type);
458 break;
459 case CODEC_ID_AMR_WB:
460 av_strlcatf(buff, size, "a=rtpmap:%d AMR-WB/%d/%d\r\n"
461 "a=fmtp:%d octet-align=1\r\n",
462 payload_type, c->sample_rate, c->channels,
463 payload_type);
464 break;
465 case CODEC_ID_VORBIS:
466 if (c->extradata_size)
467 config = xiph_extradata2config(c);
468 else
469 av_log(c, AV_LOG_ERROR, "Vorbis configuration info missing\n");
470 if (!config)
471 return NULL;
472
473 av_strlcatf(buff, size, "a=rtpmap:%d vorbis/%d/%d\r\n"
474 "a=fmtp:%d configuration=%s\r\n",
475 payload_type, c->sample_rate, c->channels,
476 payload_type, config);
477 break;
478 case CODEC_ID_THEORA: {
479 const char *pix_fmt;
480 if (c->extradata_size)
481 config = xiph_extradata2config(c);
482 else
483 av_log(c, AV_LOG_ERROR, "Theora configuation info missing\n");
484 if (!config)
485 return NULL;
486
487 switch (c->pix_fmt) {
488 case PIX_FMT_YUV420P:
489 pix_fmt = "YCbCr-4:2:0";
490 break;
491 case PIX_FMT_YUV422P:
492 pix_fmt = "YCbCr-4:2:2";
493 break;
494 case PIX_FMT_YUV444P:
495 pix_fmt = "YCbCr-4:4:4";
496 break;
497 default:
498 av_log(c, AV_LOG_ERROR, "Unsupported pixel format.\n");
499 return NULL;
500 }
501
502 av_strlcatf(buff, size, "a=rtpmap:%d theora/90000\r\n"
503 "a=fmtp:%d delivery-method=inline; "
504 "width=%d; height=%d; sampling=%s; "
505 "configuration=%s\r\n",
506 payload_type, payload_type,
507 c->width, c->height, pix_fmt, config);
508 break;
509 }
510 case CODEC_ID_VP8:
511 av_strlcatf(buff, size, "a=rtpmap:%d VP8/90000\r\n",
512 payload_type);
513 break;
514 case CODEC_ID_ADPCM_G722:
515 if (payload_type >= RTP_PT_PRIVATE)
516 av_strlcatf(buff, size, "a=rtpmap:%d G722/%d/%d\r\n",
517 payload_type,
518 8000, c->channels);
519 break;
520 case CODEC_ID_ADPCM_G726: {
521 if (payload_type >= RTP_PT_PRIVATE)
522 av_strlcatf(buff, size, "a=rtpmap:%d G726-%d/%d\r\n",
523 payload_type,
524 c->bits_per_coded_sample*8,
525 c->sample_rate);
526 break;
527 }
528 default:
529 /* Nothing special to do here... */
530 break;
531 }
532
533 av_free(config);
534
535 return buff;
536 }
537
538 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)
539 {
540 const char *type;
541 int payload_type;
542
543 payload_type = ff_rtp_get_payload_type(fmt, c);
544
545 switch (c->codec_type) {
546 case AVMEDIA_TYPE_VIDEO : type = "video" ; break;
547 case AVMEDIA_TYPE_AUDIO : type = "audio" ; break;
548 case AVMEDIA_TYPE_SUBTITLE: type = "text" ; break;
549 default : type = "application"; break;
550 }
551
552 av_strlcatf(buff, size, "m=%s %d RTP/AVP %d\r\n", type, port, payload_type);
553 sdp_write_address(buff, size, dest_addr, dest_type, ttl);
554 if (c->bit_rate) {
555 av_strlcatf(buff, size, "b=AS:%d\r\n", c->bit_rate / 1000);
556 }
557
558 sdp_write_media_attributes(buff, size, c, payload_type, fmt);
559 }
560
561 int av_sdp_create(AVFormatContext *ac[], int n_files, char *buf, int size)
562 {
563 AVDictionaryEntry *title = av_dict_get(ac[0]->metadata, "title", NULL, 0);
564 struct sdp_session_level s;
565 int i, j, port, ttl, is_multicast;
566 char dst[32], dst_type[5];
567
568 memset(buf, 0, size);
569 memset(&s, 0, sizeof(struct sdp_session_level));
570 s.user = "-";
571 s.src_addr = "127.0.0.1"; /* FIXME: Properly set this */
572 s.src_type = "IP4";
573 s.name = title ? title->value : "No Name";
574
575 port = 0;
576 ttl = 0;
577 if (n_files == 1) {
578 port = sdp_get_address(dst, sizeof(dst), &ttl, ac[0]->filename);
579 is_multicast = resolve_destination(dst, sizeof(dst), dst_type,
580 sizeof(dst_type));
581 if (!is_multicast)
582 ttl = 0;
583 if (dst[0]) {
584 s.dst_addr = dst;
585 s.dst_type = dst_type;
586 s.ttl = ttl;
587 if (!strcmp(dst_type, "IP6")) {
588 s.src_addr = "::1";
589 s.src_type = "IP6";
590 }
591 }
592 }
593 sdp_write_header(buf, size, &s);
594
595 dst[0] = 0;
596 for (i = 0; i < n_files; i++) {
597 if (n_files != 1) {
598 port = sdp_get_address(dst, sizeof(dst), &ttl, ac[i]->filename);
599 is_multicast = resolve_destination(dst, sizeof(dst), dst_type,
600 sizeof(dst_type));
601 if (!is_multicast)
602 ttl = 0;
603 }
604 for (j = 0; j < ac[i]->nb_streams; j++) {
605 ff_sdp_write_media(buf, size,
606 ac[i]->streams[j]->codec, dst[0] ? dst : NULL,
607 dst_type, (port > 0) ? port + j * 2 : 0, ttl,
608 ac[i]);
609 if (port <= 0) {
610 av_strlcatf(buf, size,
611 "a=control:streamid=%d\r\n", i + j);
612 }
613 }
614 }
615
616 return 0;
617 }
618 #else
619 int av_sdp_create(AVFormatContext *ac[], int n_files, char *buf, int size)
620 {
621 return AVERROR(ENOSYS);
622 }
623
624 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)
625 {
626 }
627 #endif
628
629 #if FF_API_SDP_CREATE
630 int avf_sdp_create(AVFormatContext *ac[], int n_files, char *buff, int size)
631 {
632 return av_sdp_create(ac, n_files, buff, size);
633 }
634 #endif