xref: /openbsd-src/sys/netinet/udp_usrreq.c (revision daf88648c0e349d5c02e1504293082072c981640)
1 /*	$OpenBSD: udp_usrreq.c,v 1.112 2006/12/09 01:12:28 itojun Exp $	*/
2 /*	$NetBSD: udp_usrreq.c,v 1.28 1996/03/16 23:54:03 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1990, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)COPYRIGHT	1.1 (NRL) 17 January 1995
33  *
34  * NRL grants permission for redistribution and use in source and binary
35  * forms, with or without modification, of the software and documentation
36  * created at NRL provided that the following conditions are met:
37  *
38  * 1. Redistributions of source code must retain the above copyright
39  *    notice, this list of conditions and the following disclaimer.
40  * 2. Redistributions in binary form must reproduce the above copyright
41  *    notice, this list of conditions and the following disclaimer in the
42  *    documentation and/or other materials provided with the distribution.
43  * 3. All advertising materials mentioning features or use of this software
44  *    must display the following acknowledgements:
45  * 	This product includes software developed by the University of
46  * 	California, Berkeley and its contributors.
47  * 	This product includes software developed at the Information
48  * 	Technology Division, US Naval Research Laboratory.
49  * 4. Neither the name of the NRL nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
54  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
55  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
56  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
57  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
58  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
59  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
60  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
61  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
62  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
63  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
64  *
65  * The views and conclusions contained in the software and documentation
66  * are those of the authors and should not be interpreted as representing
67  * official policies, either expressed or implied, of the US Naval
68  * Research Laboratory (NRL).
69  */
70 
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/mbuf.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/sysctl.h>
78 
79 #include <net/if.h>
80 #include <net/route.h>
81 
82 #include <netinet/in.h>
83 #include <netinet/in_systm.h>
84 #include <netinet/in_var.h>
85 #include <netinet/ip.h>
86 #include <netinet/in_pcb.h>
87 #include <netinet/ip_var.h>
88 #include <netinet/ip_icmp.h>
89 #include <netinet/udp.h>
90 #include <netinet/udp_var.h>
91 
92 #ifdef IPSEC
93 #include <netinet/ip_ipsp.h>
94 #include <netinet/ip_esp.h>
95 #endif
96 
97 #ifdef INET6
98 #ifndef INET
99 #include <netinet/in.h>
100 #endif
101 #include <netinet6/ip6protosw.h>
102 
103 extern int ip6_defhlim;
104 #endif /* INET6 */
105 
106 /*
107  * UDP protocol implementation.
108  * Per RFC 768, August, 1980.
109  */
110 int	udpcksum = 1;
111 
112 u_int	udp_sendspace = 9216;		/* really max datagram size */
113 u_int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
114 					/* 40 1K datagrams */
115 
116 int *udpctl_vars[UDPCTL_MAXID] = UDPCTL_VARS;
117 
118 struct	inpcbtable udbtable;
119 struct	udpstat udpstat;
120 
121 static	void udp_detach(struct inpcb *);
122 static	void udp_notify(struct inpcb *, int);
123 
124 #ifndef UDBHASHSIZE
125 #define	UDBHASHSIZE	128
126 #endif
127 int	udbhashsize = UDBHASHSIZE;
128 
129 /* from in_pcb.c */
130 extern	struct baddynamicports baddynamicports;
131 
132 void
133 udp_init()
134 {
135 	in_pcbinit(&udbtable, udbhashsize);
136 }
137 
138 #ifdef INET6
139 int
140 udp6_input(mp, offp, proto)
141 	struct mbuf **mp;
142 	int *offp, proto;
143 {
144 	struct mbuf *m = *mp;
145 
146 #if defined(NFAITH) && 0 < NFAITH
147 	if (m->m_pkthdr.rcvif) {
148 		if (m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
149 			/* XXX send icmp6 host/port unreach? */
150 			m_freem(m);
151 			return IPPROTO_DONE;
152 		}
153 	}
154 #endif
155 
156 	udp_input(m, *offp, proto);
157 	return IPPROTO_DONE;
158 }
159 #endif
160 
161 void
162 udp_input(struct mbuf *m, ...)
163 {
164 	struct ip *ip;
165 	struct udphdr *uh;
166 	struct inpcb *inp;
167 	struct mbuf *opts = NULL;
168 	struct ip save_ip;
169 	int iphlen, len;
170 	va_list ap;
171 	u_int16_t savesum;
172 	union {
173 		struct sockaddr sa;
174 		struct sockaddr_in sin;
175 #ifdef INET6
176 		struct sockaddr_in6 sin6;
177 #endif /* INET6 */
178 	} srcsa, dstsa;
179 #ifdef INET6
180 	struct ip6_hdr *ip6;
181 #endif /* INET6 */
182 #ifdef IPSEC
183 	struct m_tag *mtag;
184 	struct tdb_ident *tdbi;
185 	struct tdb *tdb;
186 	int error, s;
187 #endif /* IPSEC */
188 
189 	va_start(ap, m);
190 	iphlen = va_arg(ap, int);
191 	va_end(ap);
192 
193 	udpstat.udps_ipackets++;
194 
195 	switch (mtod(m, struct ip *)->ip_v) {
196 	case 4:
197 		ip = mtod(m, struct ip *);
198 #ifdef INET6
199 		ip6 = NULL;
200 #endif /* INET6 */
201 		srcsa.sa.sa_family = AF_INET;
202 		break;
203 #ifdef INET6
204 	case 6:
205 		ip = NULL;
206 		ip6 = mtod(m, struct ip6_hdr *);
207 		srcsa.sa.sa_family = AF_INET6;
208 		break;
209 #endif /* INET6 */
210 	default:
211 		goto bad;
212 	}
213 
214 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
215 	if (!uh) {
216 		udpstat.udps_hdrops++;
217 		return;
218 	}
219 
220 	/* Check for illegal destination port 0 */
221 	if (uh->uh_dport == 0) {
222 		udpstat.udps_noport++;
223 		goto bad;
224 	}
225 
226 	/*
227 	 * Make mbuf data length reflect UDP length.
228 	 * If not enough data to reflect UDP length, drop.
229 	 */
230 	len = ntohs((u_int16_t)uh->uh_ulen);
231 	if (ip) {
232 		if (m->m_pkthdr.len - iphlen != len) {
233 			if (len > (m->m_pkthdr.len - iphlen) ||
234 			    len < sizeof(struct udphdr)) {
235 				udpstat.udps_badlen++;
236 				goto bad;
237 			}
238 			m_adj(m, len - (m->m_pkthdr.len - iphlen));
239 		}
240 	}
241 #ifdef INET6
242 	else if (ip6) {
243 		/* jumbograms */
244 		if (len == 0 && m->m_pkthdr.len - iphlen > 0xffff)
245 			len = m->m_pkthdr.len - iphlen;
246 		if (len != m->m_pkthdr.len - iphlen) {
247 			udpstat.udps_badlen++;
248 			goto bad;
249 		}
250 	}
251 #endif
252 	else /* shouldn't happen */
253 		goto bad;
254 
255 	/*
256 	 * Save a copy of the IP header in case we want restore it
257 	 * for sending an ICMP error message in response.
258 	 */
259 	if (ip)
260 		save_ip = *ip;
261 
262 	/*
263 	 * Checksum extended UDP header and data.
264 	 * from W.R.Stevens: check incoming udp cksums even if
265 	 *	udpcksum is not set.
266 	 */
267 	savesum = uh->uh_sum;
268 #ifdef INET6
269 	if (ip6) {
270 		/* Be proactive about malicious use of IPv4 mapped address */
271 		if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
272 		    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
273 			/* XXX stat */
274 			goto bad;
275 		}
276 
277 		/*
278 		 * In IPv6, the UDP checksum is ALWAYS used.
279 		 */
280 		if (uh->uh_sum == 0) {
281 			udpstat.udps_nosum++;
282 			goto bad;
283 		}
284 		if ((uh->uh_sum = in6_cksum(m, IPPROTO_UDP, iphlen, len))) {
285 			udpstat.udps_badsum++;
286 			goto bad;
287 		}
288 	} else
289 #endif /* INET6 */
290 	if (uh->uh_sum) {
291 		if ((m->m_pkthdr.csum_flags & M_UDP_CSUM_IN_OK) == 0) {
292 			if (m->m_pkthdr.csum_flags & M_UDP_CSUM_IN_BAD) {
293 				udpstat.udps_badsum++;
294 				udpstat.udps_inhwcsum++;
295 				m_freem(m);
296 				return;
297 			}
298 
299 			if ((uh->uh_sum = in4_cksum(m, IPPROTO_UDP,
300 			    iphlen, len))) {
301 				udpstat.udps_badsum++;
302 				m_freem(m);
303 				return;
304 			}
305 		} else {
306 			m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_IN_OK;
307 			udpstat.udps_inhwcsum++;
308 		}
309 	} else
310 		udpstat.udps_nosum++;
311 
312 #ifdef IPSEC
313 	if (udpencap_enable && udpencap_port &&
314 	    uh->uh_dport == htons(udpencap_port)) {
315 		u_int32_t spi;
316 		int skip = iphlen + sizeof(struct udphdr);
317 
318 		if (m->m_pkthdr.len - skip < sizeof(u_int32_t)) {
319 			/* packet too short */
320 			m_freem(m);
321 			return;
322 		}
323 		m_copydata(m, skip, sizeof(u_int32_t), (caddr_t) &spi);
324 		/*
325 		 * decapsulate if the SPI is not zero, otherwise pass
326 		 * to userland
327 		 */
328 		if (spi != 0) {
329 			if ((m = m_pullup2(m, skip)) == NULL) {
330 				udpstat.udps_hdrops++;
331 				return;
332 			}
333 
334 			/* remove the UDP header */
335 			bcopy(mtod(m, u_char *),
336 			    mtod(m, u_char *) + sizeof(struct udphdr), iphlen);
337 			m_adj(m, sizeof(struct udphdr));
338 			skip -= sizeof(struct udphdr);
339 
340 			espstat.esps_udpencin++;
341 			ipsec_common_input(m, skip, offsetof(struct ip, ip_p),
342 			    srcsa.sa.sa_family, IPPROTO_ESP, 1);
343 			return;
344 		}
345 	}
346 #endif
347 
348 	switch (srcsa.sa.sa_family) {
349 	case AF_INET:
350 		bzero(&srcsa, sizeof(struct sockaddr_in));
351 		srcsa.sin.sin_len = sizeof(struct sockaddr_in);
352 		srcsa.sin.sin_family = AF_INET;
353 		srcsa.sin.sin_port = uh->uh_sport;
354 		srcsa.sin.sin_addr = ip->ip_src;
355 
356 		bzero(&dstsa, sizeof(struct sockaddr_in));
357 		dstsa.sin.sin_len = sizeof(struct sockaddr_in);
358 		dstsa.sin.sin_family = AF_INET;
359 		dstsa.sin.sin_port = uh->uh_dport;
360 		dstsa.sin.sin_addr = ip->ip_dst;
361 		break;
362 #ifdef INET6
363 	case AF_INET6:
364 		bzero(&srcsa, sizeof(struct sockaddr_in6));
365 		srcsa.sin6.sin6_len = sizeof(struct sockaddr_in6);
366 		srcsa.sin6.sin6_family = AF_INET6;
367 		srcsa.sin6.sin6_port = uh->uh_sport;
368 #if 0 /*XXX inbound flowinfo */
369 		srcsa.sin6.sin6_flowinfo = htonl(0x0fffffff) & ip6->ip6_flow;
370 #endif
371 		/* KAME hack: recover scopeid */
372 		(void)in6_recoverscope(&srcsa.sin6, &ip6->ip6_src,
373 		    m->m_pkthdr.rcvif);
374 
375 		bzero(&dstsa, sizeof(struct sockaddr_in6));
376 		dstsa.sin6.sin6_len = sizeof(struct sockaddr_in6);
377 		dstsa.sin6.sin6_family = AF_INET6;
378 		dstsa.sin6.sin6_port = uh->uh_dport;
379 		/* KAME hack: recover scopeid */
380 		(void)in6_recoverscope(&dstsa.sin6, &ip6->ip6_dst,
381 		    m->m_pkthdr.rcvif);
382 		break;
383 #endif /* INET6 */
384 	}
385 
386 #ifdef INET6
387 	if ((ip6 && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) ||
388 	    (ip && IN_MULTICAST(ip->ip_dst.s_addr)) ||
389 	    (ip && in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif))) {
390 #else /* INET6 */
391 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
392 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
393 #endif /* INET6 */
394 		struct inpcb *last;
395 		/*
396 		 * Deliver a multicast or broadcast datagram to *all* sockets
397 		 * for which the local and remote addresses and ports match
398 		 * those of the incoming datagram.  This allows more than
399 		 * one process to receive multi/broadcasts on the same port.
400 		 * (This really ought to be done for unicast datagrams as
401 		 * well, but that would cause problems with existing
402 		 * applications that open both address-specific sockets and
403 		 * a wildcard socket listening to the same port -- they would
404 		 * end up receiving duplicates of every unicast datagram.
405 		 * Those applications open the multiple sockets to overcome an
406 		 * inadequacy of the UDP socket interface, but for backwards
407 		 * compatibility we avoid the problem here rather than
408 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
409 		 */
410 
411 		iphlen += sizeof(struct udphdr);
412 
413 		/*
414 		 * Locate pcb(s) for datagram.
415 		 * (Algorithm copied from raw_intr().)
416 		 */
417 		last = NULL;
418 		CIRCLEQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue) {
419 #ifdef INET6
420 			/* don't accept it if AF does not match */
421 			if (ip6 && !(inp->inp_flags & INP_IPV6))
422 				continue;
423 			if (!ip6 && (inp->inp_flags & INP_IPV6))
424 				continue;
425 #endif
426 			if (inp->inp_lport != uh->uh_dport)
427 				continue;
428 #ifdef INET6
429 			if (ip6) {
430 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6))
431 					if (!IN6_ARE_ADDR_EQUAL(&inp->inp_laddr6,
432 					    &ip6->ip6_dst))
433 						continue;
434 			} else
435 #endif /* INET6 */
436 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
437 				if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
438 					continue;
439 			}
440 #ifdef INET6
441 			if (ip6) {
442 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6))
443 					if (!IN6_ARE_ADDR_EQUAL(&inp->inp_faddr6,
444 					    &ip6->ip6_src) ||
445 					    inp->inp_fport != uh->uh_sport)
446 						continue;
447 			} else
448 #endif /* INET6 */
449 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
450 				if (inp->inp_faddr.s_addr !=
451 				    ip->ip_src.s_addr ||
452 				    inp->inp_fport != uh->uh_sport)
453 					continue;
454 			}
455 
456 			if (last != NULL) {
457 				struct mbuf *n;
458 
459 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
460 #ifdef INET6
461 					if (ip6 && (last->inp_flags &
462 					    IN6P_CONTROLOPTS))
463 						ip6_savecontrol(last, n, &opts);
464 #endif /* INET6 */
465 					if (ip && (last->inp_flags &
466 					    INP_CONTROLOPTS))
467 						ip_savecontrol(last, &opts,
468 						    ip, n);
469 
470 					m_adj(n, iphlen);
471 					if (sbappendaddr(
472 					    &last->inp_socket->so_rcv,
473 					    &srcsa.sa, n, opts) == 0) {
474 						m_freem(n);
475 						if (opts)
476 							m_freem(opts);
477 						udpstat.udps_fullsock++;
478 					} else
479 						sorwakeup(last->inp_socket);
480 					opts = NULL;
481 				}
482 			}
483 			last = inp;
484 			/*
485 			 * Don't look for additional matches if this one does
486 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
487 			 * socket options set.  This heuristic avoids searching
488 			 * through all pcbs in the common case of a non-shared
489 			 * port.  It * assumes that an application will never
490 			 * clear these options after setting them.
491 			 */
492 			if ((last->inp_socket->so_options & (SO_REUSEPORT |
493 			    SO_REUSEADDR)) == 0)
494 				break;
495 		}
496 
497 		if (last == NULL) {
498 			/*
499 			 * No matching pcb found; discard datagram.
500 			 * (No need to send an ICMP Port Unreachable
501 			 * for a broadcast or multicast datgram.)
502 			 */
503 			udpstat.udps_noportbcast++;
504 			goto bad;
505 		}
506 
507 #ifdef INET6
508 		if (ip6 && (last->inp_flags & IN6P_CONTROLOPTS))
509 			ip6_savecontrol(last, m, &opts);
510 #endif /* INET6 */
511 		if (ip && (last->inp_flags & INP_CONTROLOPTS))
512 			ip_savecontrol(last, &opts, ip, m);
513 
514 		m_adj(m, iphlen);
515 		if (sbappendaddr(&last->inp_socket->so_rcv,
516 		    &srcsa.sa, m, opts) == 0) {
517 			udpstat.udps_fullsock++;
518 			goto bad;
519 		}
520 		sorwakeup(last->inp_socket);
521 		return;
522 	}
523 	/*
524 	 * Locate pcb for datagram.
525 	 */
526 #ifdef INET6
527 	if (ip6)
528 		inp = in6_pcbhashlookup(&udbtable, &ip6->ip6_src, uh->uh_sport,
529 		    &ip6->ip6_dst, uh->uh_dport);
530 	else
531 #endif /* INET6 */
532 	inp = in_pcbhashlookup(&udbtable, ip->ip_src, uh->uh_sport,
533 	    ip->ip_dst, uh->uh_dport);
534 	if (inp == 0) {
535 		int	inpl_reverse = 0;
536 #if NPF > 0
537 		struct pf_mtag *t;
538 
539 		if ((t = pf_find_mtag(m)) != NULL &&
540 		    t->flags & PF_TAG_TRANSLATE_LOCALHOST)
541 			inpl_reverse = 1;
542 #endif
543 		++udpstat.udps_pcbhashmiss;
544 #ifdef INET6
545 		if (ip6) {
546 			inp = in6_pcblookup_listen(&udbtable,
547 			    &ip6->ip6_dst, uh->uh_dport, inpl_reverse);
548 		} else
549 #endif /* INET6 */
550 		inp = in_pcblookup_listen(&udbtable,
551 		    ip->ip_dst, uh->uh_dport, inpl_reverse);
552 		if (inp == 0) {
553 			udpstat.udps_noport++;
554 			if (m->m_flags & (M_BCAST | M_MCAST)) {
555 				udpstat.udps_noportbcast++;
556 				goto bad;
557 			}
558 #ifdef INET6
559 			if (ip6) {
560 				uh->uh_sum = savesum;
561 				icmp6_error(m, ICMP6_DST_UNREACH,
562 				    ICMP6_DST_UNREACH_NOPORT,0);
563 			} else
564 #endif /* INET6 */
565 			{
566 				*ip = save_ip;
567 				uh->uh_sum = savesum;
568 				icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT,
569 				    0, 0);
570 			}
571 			return;
572 		}
573 	}
574 
575 #ifdef IPSEC
576 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
577 	s = splnet();
578 	if (mtag != NULL) {
579 		tdbi = (struct tdb_ident *)(mtag + 1);
580 		tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
581 	} else
582 		tdb = NULL;
583 	ipsp_spd_lookup(m, srcsa.sa.sa_family, iphlen, &error,
584 	    IPSP_DIRECTION_IN, tdb, inp);
585 	if (error) {
586 		splx(s);
587 		goto bad;
588 	}
589 
590 	/* Latch SA only if the socket is connected */
591 	if (inp->inp_tdb_in != tdb &&
592 	    (inp->inp_socket->so_state & SS_ISCONNECTED)) {
593 		if (tdb) {
594 			tdb_add_inp(tdb, inp, 1);
595 			if (inp->inp_ipo == NULL) {
596 				inp->inp_ipo = ipsec_add_policy(inp,
597 				    srcsa.sa.sa_family, IPSP_DIRECTION_OUT);
598 				if (inp->inp_ipo == NULL) {
599 					splx(s);
600 					goto bad;
601 				}
602 			}
603 			if (inp->inp_ipo->ipo_dstid == NULL &&
604 			    tdb->tdb_srcid != NULL) {
605 				inp->inp_ipo->ipo_dstid = tdb->tdb_srcid;
606 				tdb->tdb_srcid->ref_count++;
607 			}
608 			if (inp->inp_ipsec_remotecred == NULL &&
609 			    tdb->tdb_remote_cred != NULL) {
610 				inp->inp_ipsec_remotecred =
611 				    tdb->tdb_remote_cred;
612 				tdb->tdb_remote_cred->ref_count++;
613 			}
614 			if (inp->inp_ipsec_remoteauth == NULL &&
615 			    tdb->tdb_remote_auth != NULL) {
616 				inp->inp_ipsec_remoteauth =
617 				    tdb->tdb_remote_auth;
618 				tdb->tdb_remote_auth->ref_count++;
619 			}
620 		} else { /* Just reset */
621 			TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in, inp,
622 			    inp_tdb_in_next);
623 			inp->inp_tdb_in = NULL;
624 		}
625 	}
626 	splx(s);
627 #endif /*IPSEC */
628 
629 	opts = NULL;
630 #ifdef INET6
631 	if (ip6 && (inp->inp_flags & IN6P_CONTROLOPTS))
632 		ip6_savecontrol(inp, m, &opts);
633 #endif /* INET6 */
634 	if (ip && (inp->inp_flags & INP_CONTROLOPTS))
635 		ip_savecontrol(inp, &opts, ip, m);
636 
637 	iphlen += sizeof(struct udphdr);
638 	m_adj(m, iphlen);
639 	if (sbappendaddr(&inp->inp_socket->so_rcv, &srcsa.sa, m, opts) == 0) {
640 		udpstat.udps_fullsock++;
641 		goto bad;
642 	}
643 	sorwakeup(inp->inp_socket);
644 	return;
645 bad:
646 	m_freem(m);
647 	if (opts)
648 		m_freem(opts);
649 }
650 
651 /*
652  * Notify a udp user of an asynchronous error;
653  * just wake up so that he can collect error status.
654  */
655 static void
656 udp_notify(inp, errno)
657 	struct inpcb *inp;
658 	int errno;
659 {
660 	inp->inp_socket->so_error = errno;
661 	sorwakeup(inp->inp_socket);
662 	sowwakeup(inp->inp_socket);
663 }
664 
665 #ifdef INET6
666 void
667 udp6_ctlinput(cmd, sa, d)
668 	int cmd;
669 	struct sockaddr *sa;
670 	void *d;
671 {
672 	struct udphdr uh;
673 	struct sockaddr_in6 sa6;
674 	struct ip6_hdr *ip6;
675 	struct mbuf *m;
676 	int off;
677 	void *cmdarg;
678 	struct ip6ctlparam *ip6cp = NULL;
679 	struct udp_portonly {
680 		u_int16_t uh_sport;
681 		u_int16_t uh_dport;
682 	} *uhp;
683 	void (*notify)(struct inpcb *, int) = udp_notify;
684 
685 	if (sa == NULL)
686 		return;
687 	if (sa->sa_family != AF_INET6 ||
688 	    sa->sa_len != sizeof(struct sockaddr_in6))
689 		return;
690 
691 	if ((unsigned)cmd >= PRC_NCMDS)
692 		return;
693 	if (PRC_IS_REDIRECT(cmd))
694 		notify = in_rtchange, d = NULL;
695 	else if (cmd == PRC_HOSTDEAD)
696 		d = NULL;
697 	else if (cmd == PRC_MSGSIZE)
698 		; /* special code is present, see below */
699 	else if (inet6ctlerrmap[cmd] == 0)
700 		return;
701 
702 	/* if the parameter is from icmp6, decode it. */
703 	if (d != NULL) {
704 		ip6cp = (struct ip6ctlparam *)d;
705 		m = ip6cp->ip6c_m;
706 		ip6 = ip6cp->ip6c_ip6;
707 		off = ip6cp->ip6c_off;
708 		cmdarg = ip6cp->ip6c_cmdarg;
709 	} else {
710 		m = NULL;
711 		ip6 = NULL;
712 		cmdarg = NULL;
713 		/* XXX: translate addresses into internal form */
714 		sa6 = *(struct sockaddr_in6 *)sa;
715 #ifndef SCOPEDROUTING
716 		if (in6_embedscope(&sa6.sin6_addr, &sa6, NULL, NULL)) {
717 			/* should be impossible */
718 			return;
719 		}
720 #endif
721 	}
722 
723 	if (ip6cp && ip6cp->ip6c_finaldst) {
724 		bzero(&sa6, sizeof(sa6));
725 		sa6.sin6_family = AF_INET6;
726 		sa6.sin6_len = sizeof(sa6);
727 		sa6.sin6_addr = *ip6cp->ip6c_finaldst;
728 		/* XXX: assuming M is valid in this case */
729 		sa6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
730 		    ip6cp->ip6c_finaldst);
731 #ifndef SCOPEDROUTING
732 		if (in6_embedscope(ip6cp->ip6c_finaldst, &sa6, NULL, NULL)) {
733 			/* should be impossible */
734 			return;
735 		}
736 #endif
737 	} else {
738 		/* XXX: translate addresses into internal form */
739 		sa6 = *(struct sockaddr_in6 *)sa;
740 #ifndef SCOPEDROUTING
741 		if (in6_embedscope(&sa6.sin6_addr, &sa6, NULL, NULL)) {
742 			/* should be impossible */
743 			return;
744 		}
745 #endif
746 	}
747 
748 	if (ip6) {
749 		/*
750 		 * XXX: We assume that when IPV6 is non NULL,
751 		 * M and OFF are valid.
752 		 */
753 		struct sockaddr_in6 sa6_src;
754 
755 		/* check if we can safely examine src and dst ports */
756 		if (m->m_pkthdr.len < off + sizeof(*uhp))
757 			return;
758 
759 		bzero(&uh, sizeof(uh));
760 		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
761 
762 		bzero(&sa6_src, sizeof(sa6_src));
763 		sa6_src.sin6_family = AF_INET6;
764 		sa6_src.sin6_len = sizeof(sa6_src);
765 		sa6_src.sin6_addr = ip6->ip6_src;
766 		sa6_src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
767 		    &ip6->ip6_src);
768 #ifndef SCOPEDROUTING
769 		if (in6_embedscope(&sa6_src.sin6_addr, &sa6_src, NULL, NULL)) {
770 			/* should be impossible */
771 			return;
772 		}
773 #endif
774 
775 		if (cmd == PRC_MSGSIZE) {
776 			int valid = 0;
777 
778 			/*
779 			 * Check to see if we have a valid UDP socket
780 			 * corresponding to the address in the ICMPv6 message
781 			 * payload.
782 			 */
783 			if (in6_pcbhashlookup(&udbtable, &sa6.sin6_addr,
784 			    uh.uh_dport, &sa6_src.sin6_addr, uh.uh_sport))
785 				valid = 1;
786 #if 0
787 			/*
788 			 * As the use of sendto(2) is fairly popular,
789 			 * we may want to allow non-connected pcb too.
790 			 * But it could be too weak against attacks...
791 			 * We should at least check if the local address (= s)
792 			 * is really ours.
793 			 */
794 			else if (in6_pcblookup_listen(&udbtable,
795 			    &sa6_src.sin6_addr, uh.uh_sport, 0);
796 				valid = 1;
797 #endif
798 
799 			/*
800 			 * Depending on the value of "valid" and routing table
801 			 * size (mtudisc_{hi,lo}wat), we will:
802 			 * - recalculate the new MTU and create the
803 			 *   corresponding routing entry, or
804 			 * - ignore the MTU change notification.
805 			 */
806 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
807 
808 			/*
809 			 * regardless of if we called icmp6_mtudisc_update(),
810 			 * we need to call in6_pcbnotify(), to notify path
811 			 * MTU change to the userland (2292bis-02), because
812 			 * some unconnected sockets may share the same
813 			 * destination and want to know the path MTU.
814 			 */
815 		}
816 
817 		(void) in6_pcbnotify(&udbtable, (struct sockaddr *)&sa6,
818 		    uh.uh_dport, (struct sockaddr *)&sa6_src,
819 		    uh.uh_sport, cmd, cmdarg, notify);
820 	} else {
821 		(void) in6_pcbnotify(&udbtable, (struct sockaddr *)&sa6, 0,
822 		    (struct sockaddr *)&sa6_any, 0, cmd, cmdarg, notify);
823 	}
824 }
825 #endif
826 
827 void *
828 udp_ctlinput(cmd, sa, v)
829 	int cmd;
830 	struct sockaddr *sa;
831 	void *v;
832 {
833 	struct ip *ip = v;
834 	struct udphdr *uhp;
835 	struct in_addr faddr;
836 	struct inpcb *inp;
837 	extern int inetctlerrmap[];
838 	void (*notify)(struct inpcb *, int) = udp_notify;
839 	int errno;
840 
841 	if (sa == NULL)
842 		return NULL;
843 	if (sa->sa_family != AF_INET ||
844 	    sa->sa_len != sizeof(struct sockaddr_in))
845 		return NULL;
846 	faddr = satosin(sa)->sin_addr;
847 	if (faddr.s_addr == INADDR_ANY)
848 		return NULL;
849 
850 	if ((unsigned)cmd >= PRC_NCMDS)
851 		return NULL;
852 	errno = inetctlerrmap[cmd];
853 	if (PRC_IS_REDIRECT(cmd))
854 		notify = in_rtchange, ip = 0;
855 	else if (cmd == PRC_HOSTDEAD)
856 		ip = 0;
857 	else if (errno == 0)
858 		return NULL;
859 	if (ip) {
860 		uhp = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
861 
862 #ifdef IPSEC
863 		/* PMTU discovery for udpencap */
864 		if (cmd == PRC_MSGSIZE && ip_mtudisc && udpencap_enable &&
865 		    udpencap_port && uhp->uh_sport == htons(udpencap_port)) {
866 			udpencap_ctlinput(cmd, sa, v);
867 			return (NULL);
868 		}
869 #endif
870 		inp = in_pcbhashlookup(&udbtable,
871 		    ip->ip_dst, uhp->uh_dport, ip->ip_src, uhp->uh_sport);
872 		if (inp && inp->inp_socket != NULL)
873 			notify(inp, errno);
874 	} else
875 		in_pcbnotifyall(&udbtable, sa, errno, notify);
876 	return (NULL);
877 }
878 
879 int
880 udp_output(struct mbuf *m, ...)
881 {
882 	struct inpcb *inp;
883 	struct mbuf *addr, *control;
884 	struct udpiphdr *ui;
885 	int len = m->m_pkthdr.len;
886 	struct in_addr laddr;
887 	int s = 0, error = 0;
888 	va_list ap;
889 
890 	va_start(ap, m);
891 	inp = va_arg(ap, struct inpcb *);
892 	addr = va_arg(ap, struct mbuf *);
893 	control = va_arg(ap, struct mbuf *);
894 	va_end(ap);
895 
896 #ifdef DIAGNOSTIC
897 	if ((inp->inp_flags & INP_IPV6) != 0)
898 		panic("IPv6 inpcb to udp_output");
899 #endif
900 
901 	/*
902 	 * Compute the packet length of the IP header, and
903 	 * punt if the length looks bogus.
904 	 */
905 	if ((len + sizeof(struct udpiphdr)) > IP_MAXPACKET) {
906 		error = EMSGSIZE;
907 		goto release;
908 	}
909 
910 	if (addr) {
911 		laddr = inp->inp_laddr;
912 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
913 			error = EISCONN;
914 			goto release;
915 		}
916 		/*
917 		 * Must block input while temporarily connected.
918 		 */
919 		s = splsoftnet();
920 		error = in_pcbconnect(inp, addr);
921 		if (error) {
922 			splx(s);
923 			goto release;
924 		}
925 	} else {
926 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
927 			error = ENOTCONN;
928 			goto release;
929 		}
930 	}
931 	/*
932 	 * Calculate data length and get a mbuf
933 	 * for UDP and IP headers.
934 	 */
935 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
936 	if (m == 0) {
937 		error = ENOBUFS;
938 		goto bail;
939 	}
940 
941 	/*
942 	 * Fill in mbuf with extended UDP header
943 	 * and addresses and length put into network format.
944 	 */
945 	ui = mtod(m, struct udpiphdr *);
946 	bzero(ui->ui_x1, sizeof ui->ui_x1);
947 	ui->ui_pr = IPPROTO_UDP;
948 	ui->ui_len = htons((u_int16_t)len + sizeof (struct udphdr));
949 	ui->ui_src = inp->inp_laddr;
950 	ui->ui_dst = inp->inp_faddr;
951 	ui->ui_sport = inp->inp_lport;
952 	ui->ui_dport = inp->inp_fport;
953 	ui->ui_ulen = ui->ui_len;
954 
955 	/*
956 	 * Compute the pseudo-header checksum; defer further checksumming
957 	 * until ip_output() or hardware (if it exists).
958 	 */
959 	if (udpcksum) {
960 		m->m_pkthdr.csum_flags |= M_UDPV4_CSUM_OUT;
961 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
962 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
963 		    sizeof (struct udphdr) + IPPROTO_UDP));
964 	} else
965 		ui->ui_sum = 0;
966 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
967 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;
968 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;
969 
970 	udpstat.udps_opackets++;
971 	error = ip_output(m, inp->inp_options, &inp->inp_route,
972 	    inp->inp_socket->so_options &
973 	    (SO_DONTROUTE | SO_BROADCAST | SO_JUMBO),
974 	    inp->inp_moptions, inp);
975 
976 bail:
977 	if (addr) {
978 		inp->inp_laddr = laddr;
979 		in_pcbdisconnect(inp);
980 		splx(s);
981 	}
982 	if (control)
983 		m_freem(control);
984 	return (error);
985 
986 release:
987 	m_freem(m);
988 	if (control)
989 		m_freem(control);
990 	return (error);
991 }
992 
993 #ifdef INET6
994 /*ARGSUSED*/
995 int
996 udp6_usrreq(so, req, m, addr, control, p)
997 	struct socket *so;
998 	int req;
999 	struct mbuf *m, *addr, *control;
1000 	struct proc *p;
1001 {
1002 
1003 	return udp_usrreq(so, req, m, addr, control);
1004 }
1005 #endif
1006 
1007 /*ARGSUSED*/
1008 int
1009 udp_usrreq(so, req, m, addr, control)
1010 	struct socket *so;
1011 	int req;
1012 	struct mbuf *m, *addr, *control;
1013 {
1014 	struct inpcb *inp = sotoinpcb(so);
1015 	int error = 0;
1016 	int s;
1017 
1018 	if (req == PRU_CONTROL) {
1019 #ifdef INET6
1020 		if (inp->inp_flags & INP_IPV6)
1021 			return (in6_control(so, (u_long)m, (caddr_t)addr,
1022 			    (struct ifnet *)control, 0));
1023 		else
1024 #endif /* INET6 */
1025 			return (in_control(so, (u_long)m, (caddr_t)addr,
1026 			    (struct ifnet *)control));
1027 	}
1028 	if (inp == NULL && req != PRU_ATTACH) {
1029 		error = EINVAL;
1030 		goto release;
1031 	}
1032 	/*
1033 	 * Note: need to block udp_input while changing
1034 	 * the udp pcb queue and/or pcb addresses.
1035 	 */
1036 	switch (req) {
1037 
1038 	case PRU_ATTACH:
1039 		if (inp != NULL) {
1040 			error = EINVAL;
1041 			break;
1042 		}
1043 		s = splsoftnet();
1044 		error = in_pcballoc(so, &udbtable);
1045 		splx(s);
1046 		if (error)
1047 			break;
1048 		error = soreserve(so, udp_sendspace, udp_recvspace);
1049 		if (error)
1050 			break;
1051 #ifdef INET6
1052 		if (((struct inpcb *)so->so_pcb)->inp_flags & INP_IPV6)
1053 			((struct inpcb *) so->so_pcb)->inp_ipv6.ip6_hlim =
1054 			    ip6_defhlim;
1055 		else
1056 #endif /* INET6 */
1057 			((struct inpcb *) so->so_pcb)->inp_ip.ip_ttl = ip_defttl;
1058 		break;
1059 
1060 	case PRU_DETACH:
1061 		udp_detach(inp);
1062 		break;
1063 
1064 	case PRU_BIND:
1065 		s = splsoftnet();
1066 #ifdef INET6
1067 		if (inp->inp_flags & INP_IPV6)
1068 			error = in6_pcbbind(inp, addr);
1069 		else
1070 #endif
1071 			error = in_pcbbind(inp, addr);
1072 		splx(s);
1073 		break;
1074 
1075 	case PRU_LISTEN:
1076 		error = EOPNOTSUPP;
1077 		break;
1078 
1079 	case PRU_CONNECT:
1080 #ifdef INET6
1081 		if (inp->inp_flags & INP_IPV6) {
1082 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6)) {
1083 				error = EISCONN;
1084 				break;
1085 			}
1086 			s = splsoftnet();
1087 			error = in6_pcbconnect(inp, addr);
1088 			splx(s);
1089 		} else
1090 #endif /* INET6 */
1091 		{
1092 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
1093 				error = EISCONN;
1094 				break;
1095 			}
1096 			s = splsoftnet();
1097 			error = in_pcbconnect(inp, addr);
1098 			splx(s);
1099 		}
1100 
1101 		if (error == 0)
1102 			soisconnected(so);
1103 		break;
1104 
1105 	case PRU_CONNECT2:
1106 		error = EOPNOTSUPP;
1107 		break;
1108 
1109 	case PRU_ACCEPT:
1110 		error = EOPNOTSUPP;
1111 		break;
1112 
1113 	case PRU_DISCONNECT:
1114 #ifdef INET6
1115 		if (inp->inp_flags & INP_IPV6) {
1116 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6)) {
1117 				error = ENOTCONN;
1118 				break;
1119 			}
1120 		} else
1121 #endif /* INET6 */
1122 		{
1123 			if (inp->inp_faddr.s_addr == INADDR_ANY) {
1124 				error = ENOTCONN;
1125 				break;
1126 			}
1127 		}
1128 
1129 		s = splsoftnet();
1130 #ifdef INET6
1131 		if (inp->inp_flags & INP_IPV6)
1132 			inp->inp_laddr6 = in6addr_any;
1133 		else
1134 #endif /* INET6 */
1135 			inp->inp_laddr.s_addr = INADDR_ANY;
1136 		in_pcbdisconnect(inp);
1137 
1138 		splx(s);
1139 		so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1140 		break;
1141 
1142 	case PRU_SHUTDOWN:
1143 		socantsendmore(so);
1144 		break;
1145 
1146 	case PRU_SEND:
1147 #ifdef INET6
1148 		if (inp->inp_flags & INP_IPV6)
1149 			return (udp6_output(inp, m, addr, control));
1150 		else
1151 			return (udp_output(m, inp, addr, control));
1152 #else
1153 		return (udp_output(m, inp, addr, control));
1154 #endif
1155 
1156 	case PRU_ABORT:
1157 		soisdisconnected(so);
1158 		udp_detach(inp);
1159 		break;
1160 
1161 	case PRU_SOCKADDR:
1162 #ifdef INET6
1163 		if (inp->inp_flags & INP_IPV6)
1164 			in6_setsockaddr(inp, addr);
1165 		else
1166 #endif /* INET6 */
1167 			in_setsockaddr(inp, addr);
1168 		break;
1169 
1170 	case PRU_PEERADDR:
1171 #ifdef INET6
1172 		if (inp->inp_flags & INP_IPV6)
1173 			in6_setpeeraddr(inp, addr);
1174 		else
1175 #endif /* INET6 */
1176 			in_setpeeraddr(inp, addr);
1177 		break;
1178 
1179 	case PRU_SENSE:
1180 		/*
1181 		 * stat: don't bother with a blocksize.
1182 		 */
1183 		/*
1184 		 * Perhaps Path MTU might be returned for a connected
1185 		 * UDP socket in this case.
1186 		 */
1187 		return (0);
1188 
1189 	case PRU_SENDOOB:
1190 	case PRU_FASTTIMO:
1191 	case PRU_SLOWTIMO:
1192 	case PRU_PROTORCV:
1193 	case PRU_PROTOSEND:
1194 		error =  EOPNOTSUPP;
1195 		break;
1196 
1197 	case PRU_RCVD:
1198 	case PRU_RCVOOB:
1199 		return (EOPNOTSUPP);	/* do not free mbuf's */
1200 
1201 	default:
1202 		panic("udp_usrreq");
1203 	}
1204 
1205 release:
1206 	if (control) {
1207 		m_freem(control);
1208 	}
1209 	if (m)
1210 		m_freem(m);
1211 	return (error);
1212 }
1213 
1214 static void
1215 udp_detach(inp)
1216 	struct inpcb *inp;
1217 {
1218 	int s = splsoftnet();
1219 
1220 	in_pcbdetach(inp);
1221 	splx(s);
1222 }
1223 
1224 /*
1225  * Sysctl for udp variables.
1226  */
1227 int
1228 udp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1229 	int *name;
1230 	u_int namelen;
1231 	void *oldp;
1232 	size_t *oldlenp;
1233 	void *newp;
1234 	size_t newlen;
1235 {
1236 	/* All sysctl names at this level are terminal. */
1237 	if (namelen != 1)
1238 		return (ENOTDIR);
1239 
1240 	switch (name[0]) {
1241 	case UDPCTL_BADDYNAMIC:
1242 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1243 		    baddynamicports.udp, sizeof(baddynamicports.udp)));
1244 	default:
1245 		if (name[0] < UDPCTL_MAXID)
1246 			return (sysctl_int_arr(udpctl_vars, name, namelen,
1247 			    oldp, oldlenp, newp, newlen));
1248 		return (ENOPROTOOPT);
1249 	}
1250 	/* NOTREACHED */
1251 }
1252