xref: /openbsd-src/sys/netinet/udp_usrreq.c (revision 8500990981f885cbe5e6a4958549cacc238b5ae6)
1 /*	$OpenBSD: udp_usrreq.c,v 1.93 2003/12/02 23:16:29 markus 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 struct	inpcbtable udbtable;
113 struct	udpstat udpstat;
114 
115 static	void udp_detach(struct inpcb *);
116 static	void udp_notify(struct inpcb *, int);
117 static	struct mbuf *udp_saveopt(caddr_t, int, int);
118 
119 #ifndef UDBHASHSIZE
120 #define	UDBHASHSIZE	128
121 #endif
122 int	udbhashsize = UDBHASHSIZE;
123 
124 /* from in_pcb.c */
125 extern	struct baddynamicports baddynamicports;
126 
127 void
128 udp_init()
129 {
130 	in_pcbinit(&udbtable, udbhashsize);
131 }
132 
133 #ifdef INET6
134 int
135 udp6_input(mp, offp, proto)
136 	struct mbuf **mp;
137 	int *offp, proto;
138 {
139 	struct mbuf *m = *mp;
140 
141 #if defined(NFAITH) && 0 < NFAITH
142 	if (m->m_pkthdr.rcvif) {
143 		if (m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
144 			/* XXX send icmp6 host/port unreach? */
145 			m_freem(m);
146 			return IPPROTO_DONE;
147 		}
148 	}
149 #endif
150 
151 	udp_input(m, *offp, proto);
152 	return IPPROTO_DONE;
153 }
154 #endif
155 
156 void
157 udp_input(struct mbuf *m, ...)
158 {
159 	register struct ip *ip;
160 	register struct udphdr *uh;
161 	register struct inpcb *inp;
162 	struct mbuf *opts = 0;
163 	struct ip save_ip;
164 	int iphlen, len;
165 	va_list ap;
166 	u_int16_t savesum;
167 	union {
168 		struct sockaddr sa;
169 		struct sockaddr_in sin;
170 #ifdef INET6
171 		struct sockaddr_in6 sin6;
172 #endif /* INET6 */
173 	} srcsa, dstsa;
174 #ifdef INET6
175 	struct ip6_hdr *ip6;
176 #endif /* INET6 */
177 #ifdef IPSEC
178 	struct m_tag *mtag;
179 	struct tdb_ident *tdbi;
180 	struct tdb *tdb;
181 	int error, s;
182 #endif /* IPSEC */
183 
184 	va_start(ap, m);
185 	iphlen = va_arg(ap, int);
186 	va_end(ap);
187 
188 	udpstat.udps_ipackets++;
189 
190 	switch (mtod(m, struct ip *)->ip_v) {
191 	case 4:
192 		ip = mtod(m, struct ip *);
193 #ifdef INET6
194 		ip6 = NULL;
195 #endif /* INET6 */
196 		srcsa.sa.sa_family = AF_INET;
197 		break;
198 #ifdef INET6
199 	case 6:
200 		ip = NULL;
201 		ip6 = mtod(m, struct ip6_hdr *);
202 		srcsa.sa.sa_family = AF_INET6;
203 		break;
204 #endif /* INET6 */
205 	default:
206 		goto bad;
207 	}
208 
209 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
210 	if (!uh) {
211 		udpstat.udps_hdrops++;
212 		return;
213 	}
214 
215 	/* Check for illegal destination port 0 */
216 	if (uh->uh_dport == 0) {
217 		udpstat.udps_noport++;
218 		goto bad;
219 	}
220 
221 	/*
222 	 * Make mbuf data length reflect UDP length.
223 	 * If not enough data to reflect UDP length, drop.
224 	 */
225 	len = ntohs((u_int16_t)uh->uh_ulen);
226 	if (ip) {
227 		if (m->m_pkthdr.len - iphlen != len) {
228 			if (len > (m->m_pkthdr.len - iphlen) ||
229 			    len < sizeof(struct udphdr)) {
230 				udpstat.udps_badlen++;
231 				goto bad;
232 			}
233 			m_adj(m, len - (m->m_pkthdr.len - iphlen));
234 		}
235 	}
236 #ifdef INET6
237 	else if (ip6) {
238 		/* jumbograms */
239 		if (len == 0 && m->m_pkthdr.len - iphlen > 0xffff)
240 			len = m->m_pkthdr.len - iphlen;
241 		if (len != m->m_pkthdr.len - iphlen) {
242 			udpstat.udps_badlen++;
243 			goto bad;
244 		}
245 	}
246 #endif
247 	else /* shouldn't happen */
248 		goto bad;
249 
250 	/*
251 	 * Save a copy of the IP header in case we want restore it
252 	 * for sending an ICMP error message in response.
253 	 */
254 	if (ip)
255 		save_ip = *ip;
256 
257 	/*
258 	 * Checksum extended UDP header and data.
259 	 * from W.R.Stevens: check incoming udp cksums even if
260 	 *	udpcksum is not set.
261 	 */
262 	savesum = uh->uh_sum;
263 #ifdef INET6
264 	if (ip6) {
265 		/* Be proactive about malicious use of IPv4 mapped address */
266 		if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
267 		    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
268 			/* XXX stat */
269 			goto bad;
270 		}
271 
272 		/*
273 		 * In IPv6, the UDP checksum is ALWAYS used.
274 		 */
275 		if ((uh->uh_sum = in6_cksum(m, IPPROTO_UDP, iphlen, len))) {
276 			udpstat.udps_badsum++;
277 			goto bad;
278 		}
279 	} else
280 #endif /* INET6 */
281 	if (uh->uh_sum) {
282 		if ((m->m_pkthdr.csum & M_UDP_CSUM_IN_OK) == 0) {
283 			if (m->m_pkthdr.csum & M_UDP_CSUM_IN_BAD) {
284 				udpstat.udps_badsum++;
285 				udpstat.udps_inhwcsum++;
286 				m_freem(m);
287 				return;
288 			}
289 
290 			bzero(((struct ipovly *)ip)->ih_x1,
291 			    sizeof ((struct ipovly *)ip)->ih_x1);
292 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
293 
294 			if ((uh->uh_sum = in_cksum(m, len +
295 			    sizeof (struct ip))) != 0) {
296 				udpstat.udps_badsum++;
297 				m_freem(m);
298 				return;
299 			}
300 		} else {
301 			m->m_pkthdr.csum &= ~M_UDP_CSUM_IN_OK;
302 			udpstat.udps_inhwcsum++;
303 		}
304 	} else
305 		udpstat.udps_nosum++;
306 
307 #ifdef IPSEC
308 	if (udpencap_enable && udpencap_port &&
309 	    uh->uh_dport == htons(udpencap_port)) {
310 		u_int32_t spi;
311 		int skip = iphlen + sizeof(struct udphdr);
312 
313 		if (m->m_pkthdr.len - skip < sizeof(u_int32_t)) {
314 			/* packet too short */
315 			m_freem(m);
316 			return;
317 		}
318 		m_copydata(m, skip, sizeof(u_int32_t), (caddr_t) &spi);
319 		/*
320 		 * decapsulate if the SPI is not zero, otherwise pass
321 		 * to userland
322 		 */
323 		if (spi != 0) {
324 			if ((m = m_pullup2(m, skip)) == NULL) {
325 				udpstat.udps_hdrops++;
326 				return;
327 			}
328 
329 			/* remove the UDP header */
330 			bcopy(mtod(m, u_char *),
331 			    mtod(m, u_char *) + sizeof(struct udphdr), iphlen);
332 			m_adj(m, sizeof(struct udphdr));
333 			skip -= sizeof(struct udphdr);
334 
335 			espstat.esps_udpencin++;
336 			ipsec_common_input(m, skip, offsetof(struct ip, ip_p),
337 			    srcsa.sa.sa_family, IPPROTO_ESP, 1);
338 			return;
339 		}
340 	}
341 #endif
342 
343 	switch (srcsa.sa.sa_family) {
344 	case AF_INET:
345 		bzero(&srcsa, sizeof(struct sockaddr_in));
346 		srcsa.sin.sin_len = sizeof(struct sockaddr_in);
347 		srcsa.sin.sin_family = AF_INET;
348 		srcsa.sin.sin_port = uh->uh_sport;
349 		srcsa.sin.sin_addr = ip->ip_src;
350 
351 		bzero(&dstsa, sizeof(struct sockaddr_in));
352 		dstsa.sin.sin_len = sizeof(struct sockaddr_in);
353 		dstsa.sin.sin_family = AF_INET;
354 		dstsa.sin.sin_port = uh->uh_dport;
355 		dstsa.sin.sin_addr = ip->ip_dst;
356 		break;
357 #ifdef INET6
358 	case AF_INET6:
359 		bzero(&srcsa, sizeof(struct sockaddr_in6));
360 		srcsa.sin6.sin6_len = sizeof(struct sockaddr_in6);
361 		srcsa.sin6.sin6_family = AF_INET6;
362 		srcsa.sin6.sin6_port = uh->uh_sport;
363 #if 0 /*XXX inbound flowinfo */
364 		srcsa.sin6.sin6_flowinfo = htonl(0x0fffffff) & ip6->ip6_flow;
365 #endif
366 		/* KAME hack: recover scopeid */
367 		(void)in6_recoverscope(&srcsa.sin6, &ip6->ip6_src,
368 		    m->m_pkthdr.rcvif);
369 
370 		bzero(&dstsa, sizeof(struct sockaddr_in6));
371 		dstsa.sin6.sin6_len = sizeof(struct sockaddr_in6);
372 		dstsa.sin6.sin6_family = AF_INET6;
373 		dstsa.sin6.sin6_port = uh->uh_dport;
374 		/* KAME hack: recover scopeid */
375 		(void)in6_recoverscope(&dstsa.sin6, &ip6->ip6_dst,
376 		    m->m_pkthdr.rcvif);
377 		break;
378 #endif /* INET6 */
379 	}
380 
381 #ifdef INET6
382 	if ((ip6 && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) ||
383 	    (ip && IN_MULTICAST(ip->ip_dst.s_addr)) ||
384 	    (ip && in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif))) {
385 #else /* INET6 */
386 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
387 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
388 #endif /* INET6 */
389 		struct socket *last;
390 		/*
391 		 * Deliver a multicast or broadcast datagram to *all* sockets
392 		 * for which the local and remote addresses and ports match
393 		 * those of the incoming datagram.  This allows more than
394 		 * one process to receive multi/broadcasts on the same port.
395 		 * (This really ought to be done for unicast datagrams as
396 		 * well, but that would cause problems with existing
397 		 * applications that open both address-specific sockets and
398 		 * a wildcard socket listening to the same port -- they would
399 		 * end up receiving duplicates of every unicast datagram.
400 		 * Those applications open the multiple sockets to overcome an
401 		 * inadequacy of the UDP socket interface, but for backwards
402 		 * compatibility we avoid the problem here rather than
403 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
404 		 */
405 
406 		iphlen += sizeof(struct udphdr);
407 
408 		/*
409 		 * Locate pcb(s) for datagram.
410 		 * (Algorithm copied from raw_intr().)
411 		 */
412 		last = NULL;
413 		for (inp = udbtable.inpt_queue.cqh_first;
414 		    inp != (struct inpcb *)&udbtable.inpt_queue;
415 		    inp = inp->inp_queue.cqe_next) {
416 #ifdef INET6
417 			/* don't accept it if AF does not match */
418 			if (ip6 && !(inp->inp_flags & INP_IPV6))
419 				continue;
420 			if (!ip6 && (inp->inp_flags & INP_IPV6))
421 				continue;
422 #endif
423 			if (inp->inp_lport != uh->uh_dport)
424 				continue;
425 #ifdef INET6
426 			if (ip6) {
427 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6))
428 					if (!IN6_ARE_ADDR_EQUAL(&inp->inp_laddr6,
429 					    &ip6->ip6_dst))
430 						continue;
431 			} else
432 #endif /* INET6 */
433 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
434 				if (inp->inp_laddr.s_addr !=
435 				    ip->ip_dst.s_addr)
436 					continue;
437 			}
438 #ifdef INET6
439 			if (ip6) {
440 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6))
441 					if (!IN6_ARE_ADDR_EQUAL(&inp->inp_faddr6,
442 					    &ip6->ip6_src) ||
443 					    inp->inp_fport != uh->uh_sport)
444 						continue;
445 			} else
446 #endif /* INET6 */
447 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
448 				if (inp->inp_faddr.s_addr !=
449 				    ip->ip_src.s_addr ||
450 				    inp->inp_fport != uh->uh_sport)
451 					continue;
452 			}
453 
454 			if (last != NULL) {
455 				struct mbuf *n;
456 
457 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
458 					opts = NULL;
459 #ifdef INET6
460 					if (ip6 && (inp->inp_flags & IN6P_CONTROLOPTS))
461 						ip6_savecontrol(inp, &opts, ip6, n);
462 #endif /* INET6 */
463 					m_adj(n, iphlen);
464 					if (sbappendaddr(&last->so_rcv,
465 					    &srcsa.sa, n, opts) == 0) {
466 						m_freem(n);
467 						if (opts)
468 							m_freem(opts);
469 						udpstat.udps_fullsock++;
470 					} else
471 						sorwakeup(last);
472 					opts = NULL;
473 				}
474 			}
475 			last = inp->inp_socket;
476 			/*
477 			 * Don't look for additional matches if this one does
478 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
479 			 * socket options set.  This heuristic avoids searching
480 			 * through all pcbs in the common case of a non-shared
481 			 * port.  It * assumes that an application will never
482 			 * clear these options after setting them.
483 			 */
484 			if ((last->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
485 				break;
486 		}
487 
488 		if (last == NULL) {
489 			/*
490 			 * No matching pcb found; discard datagram.
491 			 * (No need to send an ICMP Port Unreachable
492 			 * for a broadcast or multicast datgram.)
493 			 */
494 			udpstat.udps_noportbcast++;
495 			goto bad;
496 		}
497 
498 		opts = NULL;
499 #ifdef INET6
500 		if (ip6 && (inp->inp_flags & IN6P_CONTROLOPTS))
501 			ip6_savecontrol(inp, &opts, ip6, m);
502 #endif /* INET6 */
503 		m_adj(m, iphlen);
504 		if (sbappendaddr(&last->so_rcv,
505 		    &srcsa.sa, m, opts) == 0) {
506 			udpstat.udps_fullsock++;
507 			goto bad;
508 		}
509 		sorwakeup(last);
510 		return;
511 	}
512 	/*
513 	 * Locate pcb for datagram.
514 	 */
515 #ifdef INET6
516 	if (ip6)
517 		inp = in6_pcbhashlookup(&udbtable, &ip6->ip6_src, uh->uh_sport,
518 		    &ip6->ip6_dst, uh->uh_dport);
519 	else
520 #endif /* INET6 */
521 	inp = in_pcbhashlookup(&udbtable, ip->ip_src, uh->uh_sport,
522 	    ip->ip_dst, uh->uh_dport);
523 	if (inp == 0) {
524 		++udpstat.udps_pcbhashmiss;
525 #ifdef INET6
526 		if (ip6) {
527 			inp = in6_pcblookup_listen(&udbtable,
528 			    &ip6->ip6_dst, uh->uh_dport);
529 		} else
530 #endif /* INET6 */
531 		inp = in_pcblookup_listen(&udbtable,
532 		    ip->ip_dst, uh->uh_dport);
533 		if (inp == 0) {
534 			udpstat.udps_noport++;
535 			if (m->m_flags & (M_BCAST | M_MCAST)) {
536 				udpstat.udps_noportbcast++;
537 				goto bad;
538 			}
539 #ifdef INET6
540 			if (ip6) {
541 				icmp6_error(m, ICMP6_DST_UNREACH,
542 				    ICMP6_DST_UNREACH_NOPORT,0);
543 			} else
544 #endif /* INET6 */
545 			{
546 				*ip = save_ip;
547 				uh->uh_sum = savesum;
548 				icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT,
549 				    0, 0);
550 			}
551 			return;
552 		}
553 	}
554 
555 #ifdef IPSEC
556 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
557 	s = splnet();
558 	if (mtag != NULL) {
559 		tdbi = (struct tdb_ident *)(mtag + 1);
560 		tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
561 	} else
562 		tdb = NULL;
563 	ipsp_spd_lookup(m, srcsa.sa.sa_family, iphlen, &error,
564 	    IPSP_DIRECTION_IN, tdb, inp);
565 	if (error) {
566 		splx(s);
567 		goto bad;
568 	}
569 
570 	/* Latch SA only if the socket is connected */
571 	if (inp->inp_tdb_in != tdb &&
572 	    (inp->inp_socket->so_state & SS_ISCONNECTED)) {
573 		if (tdb) {
574 			tdb_add_inp(tdb, inp, 1);
575 			if (inp->inp_ipo == NULL) {
576 				inp->inp_ipo = ipsec_add_policy(inp,
577 				    srcsa.sa.sa_family, IPSP_DIRECTION_OUT);
578 				if (inp->inp_ipo == NULL) {
579 					splx(s);
580 					goto bad;
581 				}
582 			}
583 			if (inp->inp_ipo->ipo_dstid == NULL &&
584 			    tdb->tdb_srcid != NULL) {
585 				inp->inp_ipo->ipo_dstid = tdb->tdb_srcid;
586 				tdb->tdb_srcid->ref_count++;
587 			}
588 			if (inp->inp_ipsec_remotecred == NULL &&
589 			    tdb->tdb_remote_cred != NULL) {
590 				inp->inp_ipsec_remotecred =
591 				    tdb->tdb_remote_cred;
592 				tdb->tdb_remote_cred->ref_count++;
593 			}
594 			if (inp->inp_ipsec_remoteauth == NULL &&
595 			    tdb->tdb_remote_auth != NULL) {
596 				inp->inp_ipsec_remoteauth =
597 				    tdb->tdb_remote_auth;
598 				tdb->tdb_remote_auth->ref_count++;
599 			}
600 		} else { /* Just reset */
601 			TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in, inp,
602 			    inp_tdb_in_next);
603 			inp->inp_tdb_in = NULL;
604 		}
605 	}
606 	splx(s);
607 #endif /*IPSEC */
608 
609 	opts = NULL;
610 #ifdef INET6
611 	if (ip6 && (inp->inp_flags & IN6P_CONTROLOPTS))
612 		ip6_savecontrol(inp, &opts, ip6, m);
613 #endif /* INET6 */
614 	if (ip && (inp->inp_flags & INP_CONTROLOPTS)) {
615 		struct mbuf **mp = &opts;
616 
617 		if (inp->inp_flags & INP_RECVDSTADDR) {
618 			*mp = udp_saveopt((caddr_t) &ip->ip_dst,
619 			    sizeof(struct in_addr), IP_RECVDSTADDR);
620 			if (*mp)
621 				mp = &(*mp)->m_next;
622 		}
623 #ifdef notyet
624 		/* options were tossed above */
625 		if (inp->inp_flags & INP_RECVOPTS) {
626 			*mp = udp_saveopt((caddr_t) opts_deleted_above,
627 			    sizeof(struct in_addr), IP_RECVOPTS);
628 			if (*mp)
629 				mp = &(*mp)->m_next;
630 		}
631 		/* ip_srcroute doesn't do what we want here, need to fix */
632 		if (inp->inp_flags & INP_RECVRETOPTS) {
633 			*mp = udp_saveopt((caddr_t) ip_srcroute(),
634 			    sizeof(struct in_addr), IP_RECVRETOPTS);
635 			if (*mp)
636 				mp = &(*mp)->m_next;
637 		}
638 #endif
639 	}
640 	iphlen += sizeof(struct udphdr);
641 	m_adj(m, iphlen);
642 	if (sbappendaddr(&inp->inp_socket->so_rcv, &srcsa.sa, m, opts) == 0) {
643 		udpstat.udps_fullsock++;
644 		goto bad;
645 	}
646 	sorwakeup(inp->inp_socket);
647 	return;
648 bad:
649 	m_freem(m);
650 	if (opts)
651 		m_freem(opts);
652 }
653 
654 /*
655  * Create a "control" mbuf containing the specified data
656  * with the specified type for presentation with a datagram.
657  */
658 struct mbuf *
659 udp_saveopt(p, size, type)
660 	caddr_t p;
661 	register int size;
662 	int type;
663 {
664 	register struct cmsghdr *cp;
665 	struct mbuf *m;
666 
667 	if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL)
668 		return ((struct mbuf *) NULL);
669 	cp = (struct cmsghdr *) mtod(m, struct cmsghdr *);
670 	bcopy(p, CMSG_DATA(cp), size);
671 	size = CMSG_LEN(size);
672 	m->m_len = size;
673 	cp->cmsg_len = size;
674 	cp->cmsg_level = IPPROTO_IP;
675 	cp->cmsg_type = type;
676 	return (m);
677 }
678 
679 /*
680  * Notify a udp user of an asynchronous error;
681  * just wake up so that he can collect error status.
682  */
683 static void
684 udp_notify(inp, errno)
685 	register struct inpcb *inp;
686 	int errno;
687 {
688 	inp->inp_socket->so_error = errno;
689 	sorwakeup(inp->inp_socket);
690 	sowwakeup(inp->inp_socket);
691 }
692 
693 #ifdef INET6
694 void
695 udp6_ctlinput(cmd, sa, d)
696 	int cmd;
697 	struct sockaddr *sa;
698 	void *d;
699 {
700 	struct udphdr uh;
701 	struct sockaddr_in6 sa6;
702 	register struct ip6_hdr *ip6;
703 	struct mbuf *m;
704 	int off;
705 	void *cmdarg;
706 	struct ip6ctlparam *ip6cp = NULL;
707 	struct udp_portonly {
708 		u_int16_t uh_sport;
709 		u_int16_t uh_dport;
710 	} *uhp;
711 	void (*notify)(struct inpcb *, int) = udp_notify;
712 
713 	if (sa == NULL)
714 		return;
715 	if (sa->sa_family != AF_INET6 ||
716 	    sa->sa_len != sizeof(struct sockaddr_in6))
717 		return;
718 
719 	if ((unsigned)cmd >= PRC_NCMDS)
720 		return;
721 	if (PRC_IS_REDIRECT(cmd))
722 		notify = in_rtchange, d = NULL;
723 	else if (cmd == PRC_HOSTDEAD)
724 		d = NULL;
725 	else if (cmd == PRC_MSGSIZE)
726 		; /* special code is present, see below */
727 	else if (inet6ctlerrmap[cmd] == 0)
728 		return;
729 
730 	/* if the parameter is from icmp6, decode it. */
731 	if (d != NULL) {
732 		ip6cp = (struct ip6ctlparam *)d;
733 		m = ip6cp->ip6c_m;
734 		ip6 = ip6cp->ip6c_ip6;
735 		off = ip6cp->ip6c_off;
736 		cmdarg = ip6cp->ip6c_cmdarg;
737 	} else {
738 		m = NULL;
739 		ip6 = NULL;
740 		cmdarg = NULL;
741 		/* XXX: translate addresses into internal form */
742 		sa6 = *(struct sockaddr_in6 *)sa;
743 #ifndef SCOPEDROUTING
744 		if (in6_embedscope(&sa6.sin6_addr, &sa6, NULL, NULL)) {
745 			/* should be impossible */
746 			return;
747 		}
748 #endif
749 	}
750 
751 	if (ip6cp && ip6cp->ip6c_finaldst) {
752 		bzero(&sa6, sizeof(sa6));
753 		sa6.sin6_family = AF_INET6;
754 		sa6.sin6_len = sizeof(sa6);
755 		sa6.sin6_addr = *ip6cp->ip6c_finaldst;
756 		/* XXX: assuming M is valid in this case */
757 		sa6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
758 		    ip6cp->ip6c_finaldst);
759 #ifndef SCOPEDROUTING
760 		if (in6_embedscope(ip6cp->ip6c_finaldst, &sa6, NULL, NULL)) {
761 			/* should be impossible */
762 			return;
763 		}
764 #endif
765 	} else {
766 		/* XXX: translate addresses into internal form */
767 		sa6 = *(struct sockaddr_in6 *)sa;
768 #ifndef SCOPEDROUTING
769 		if (in6_embedscope(&sa6.sin6_addr, &sa6, NULL, NULL)) {
770 			/* should be impossible */
771 			return;
772 		}
773 #endif
774 	}
775 
776 	if (ip6) {
777 		/*
778 		 * XXX: We assume that when IPV6 is non NULL,
779 		 * M and OFF are valid.
780 		 */
781 		struct sockaddr_in6 sa6_src;
782 
783 		/* check if we can safely examine src and dst ports */
784 		if (m->m_pkthdr.len < off + sizeof(*uhp))
785 			return;
786 
787 		bzero(&uh, sizeof(uh));
788 		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
789 
790 		bzero(&sa6_src, sizeof(sa6_src));
791 		sa6_src.sin6_family = AF_INET6;
792 		sa6_src.sin6_len = sizeof(sa6_src);
793 		sa6_src.sin6_addr = ip6->ip6_src;
794 		sa6_src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
795 		    &ip6->ip6_src);
796 #ifndef SCOPEDROUTING
797 		if (in6_embedscope(&sa6_src.sin6_addr, &sa6_src, NULL, NULL)) {
798 			/* should be impossible */
799 			return;
800 		}
801 #endif
802 
803 		if (cmd == PRC_MSGSIZE) {
804 			int valid = 0;
805 
806 			/*
807 			 * Check to see if we have a valid UDP socket
808 			 * corresponding to the address in the ICMPv6 message
809 			 * payload.
810 			 */
811 			if (in6_pcbhashlookup(&udbtable, &sa6.sin6_addr,
812 			    uh.uh_dport, &sa6_src.sin6_addr, uh.uh_sport))
813 				valid = 1;
814 #if 0
815 			/*
816 			 * As the use of sendto(2) is fairly popular,
817 			 * we may want to allow non-connected pcb too.
818 			 * But it could be too weak against attacks...
819 			 * We should at least check if the local address (= s)
820 			 * is really ours.
821 			 */
822 			else if (in6_pcblookup_listen(&udbtable,
823 			    &sa6_src.sin6_addr, uh.uh_sport))
824 				valid = 1;
825 #endif
826 
827 			/*
828 			 * Depending on the value of "valid" and routing table
829 			 * size (mtudisc_{hi,lo}wat), we will:
830 			 * - recalcurate the new MTU and create the
831 			 *   corresponding routing entry, or
832 			 * - ignore the MTU change notification.
833 			 */
834 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
835 
836 			/*
837 			 * regardless of if we called icmp6_mtudisc_update(),
838 			 * we need to call in6_pcbnotify(), to notify path
839 			 * MTU change to the userland (2292bis-02), because
840 			 * some unconnected sockets may share the same
841 			 * destination and want to know the path MTU.
842 			 */
843 		}
844 
845 		(void) in6_pcbnotify(&udbtable, (struct sockaddr *)&sa6,
846 		    uh.uh_dport, (struct sockaddr *)&sa6_src,
847 		    uh.uh_sport, cmd, cmdarg, notify);
848 	} else {
849 		(void) in6_pcbnotify(&udbtable, (struct sockaddr *)&sa6, 0,
850 		    (struct sockaddr *)&sa6_any, 0, cmd, cmdarg, notify);
851 	}
852 }
853 #endif
854 
855 void *
856 udp_ctlinput(cmd, sa, v)
857 	int cmd;
858 	struct sockaddr *sa;
859 	void *v;
860 {
861 	register struct ip *ip = v;
862 	register struct udphdr *uhp;
863 	extern int inetctlerrmap[];
864 	void (*notify)(struct inpcb *, int) = udp_notify;
865 	int errno;
866 
867 	if (sa == NULL)
868 		return NULL;
869 	if (sa->sa_family != AF_INET ||
870 	    sa->sa_len != sizeof(struct sockaddr_in))
871 		return NULL;
872 
873 	if ((unsigned)cmd >= PRC_NCMDS)
874 		return NULL;
875 	errno = inetctlerrmap[cmd];
876 	if (PRC_IS_REDIRECT(cmd))
877 		notify = in_rtchange, ip = 0;
878 	else if (cmd == PRC_HOSTDEAD)
879 		ip = 0;
880 	else if (errno == 0)
881 		return NULL;
882 	if (ip) {
883 		uhp = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
884 		in_pcbnotify(&udbtable, sa, uhp->uh_dport, ip->ip_src,
885 		    uhp->uh_sport, errno, notify);
886 	} else
887 		in_pcbnotifyall(&udbtable, sa, errno, notify);
888 	return NULL;
889 }
890 
891 int
892 udp_output(struct mbuf *m, ...)
893 {
894 	register struct inpcb *inp;
895 	struct mbuf *addr, *control;
896 	register struct udpiphdr *ui;
897 	register int len = m->m_pkthdr.len;
898 	struct in_addr laddr;
899 	int s = 0, error = 0;
900 	va_list ap;
901 	int pcbflags = 0;
902 
903 	va_start(ap, m);
904 	inp = va_arg(ap, struct inpcb *);
905 	addr = va_arg(ap, struct mbuf *);
906 	control = va_arg(ap, struct mbuf *);
907 	va_end(ap);
908 
909 #ifdef DIAGNOSTIC
910 	if ((inp->inp_flags & INP_IPV6) != 0)
911 		panic("IPv6 inpcb to udp_output");
912 #endif
913 
914 	/*
915 	 * Compute the packet length of the IP header, and
916 	 * punt if the length looks bogus.
917 	 */
918 	if ((len + sizeof(struct udpiphdr)) > IP_MAXPACKET) {
919 		error = EMSGSIZE;
920 		goto release;
921 	}
922 
923 	if (addr) {
924 		/*
925 		 * Save current PCB flags because they may change during
926 		 * temporary connection.
927 		 */
928 		pcbflags = inp->inp_flags;
929 
930 		laddr = inp->inp_laddr;
931 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
932 			error = EISCONN;
933 			goto release;
934 		}
935 		/*
936 		 * Must block input while temporarily connected.
937 		 */
938 		s = splsoftnet();
939 		error = in_pcbconnect(inp, addr);
940 		if (error) {
941 			splx(s);
942 			goto release;
943 		}
944 	} else {
945 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
946 			error = ENOTCONN;
947 			goto release;
948 		}
949 	}
950 	/*
951 	 * Calculate data length and get a mbuf
952 	 * for UDP and IP headers.
953 	 */
954 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
955 	if (m == 0) {
956 		error = ENOBUFS;
957 		goto bail;
958 	}
959 
960 	/*
961 	 * Fill in mbuf with extended UDP header
962 	 * and addresses and length put into network format.
963 	 */
964 	ui = mtod(m, struct udpiphdr *);
965 	bzero(ui->ui_x1, sizeof ui->ui_x1);
966 	ui->ui_pr = IPPROTO_UDP;
967 	ui->ui_len = htons((u_int16_t)len + sizeof (struct udphdr));
968 	ui->ui_src = inp->inp_laddr;
969 	ui->ui_dst = inp->inp_faddr;
970 	ui->ui_sport = inp->inp_lport;
971 	ui->ui_dport = inp->inp_fport;
972 	ui->ui_ulen = ui->ui_len;
973 
974 	/*
975 	 * Compute the pseudo-header checksum; defer further checksumming
976 	 * until ip_output() or hardware (if it exists).
977 	 */
978 	if (udpcksum) {
979 		m->m_pkthdr.csum |= M_UDPV4_CSUM_OUT;
980 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
981 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
982 		    sizeof (struct udphdr) + IPPROTO_UDP));
983 	} else
984 		ui->ui_sum = 0;
985 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
986 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;
987 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;
988 
989 	udpstat.udps_opackets++;
990 	error = ip_output(m, inp->inp_options, &inp->inp_route,
991 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
992 	    inp->inp_moptions, inp, (void *)NULL);
993 
994 bail:
995 	if (addr) {
996 		in_pcbdisconnect(inp);
997 		inp->inp_flags = pcbflags;
998 		inp->inp_laddr = laddr;
999 		splx(s);
1000 	}
1001 	if (control)
1002 		m_freem(control);
1003 	return (error);
1004 
1005 release:
1006 	m_freem(m);
1007 	if (control)
1008 		m_freem(control);
1009 	return (error);
1010 }
1011 
1012 u_int	udp_sendspace = 9216;		/* really max datagram size */
1013 u_int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1014 					/* 40 1K datagrams */
1015 
1016 #ifdef INET6
1017 /*ARGSUSED*/
1018 int
1019 udp6_usrreq(so, req, m, addr, control, p)
1020 	struct socket *so;
1021 	int req;
1022 	struct mbuf *m, *addr, *control;
1023 	struct proc *p;
1024 {
1025 
1026 	return udp_usrreq(so, req, m, addr, control);
1027 }
1028 #endif
1029 
1030 /*ARGSUSED*/
1031 int
1032 udp_usrreq(so, req, m, addr, control)
1033 	struct socket *so;
1034 	int req;
1035 	struct mbuf *m, *addr, *control;
1036 {
1037 	struct inpcb *inp = sotoinpcb(so);
1038 	int error = 0;
1039 	int s;
1040 
1041 	if (req == PRU_CONTROL) {
1042 #ifdef INET6
1043 		if (inp->inp_flags & INP_IPV6)
1044 			return (in6_control(so, (u_long)m, (caddr_t)addr,
1045 			    (struct ifnet *)control, 0));
1046 		else
1047 #endif /* INET6 */
1048 			return (in_control(so, (u_long)m, (caddr_t)addr,
1049 			    (struct ifnet *)control));
1050 	}
1051 	if (inp == NULL && req != PRU_ATTACH) {
1052 		error = EINVAL;
1053 		goto release;
1054 	}
1055 	/*
1056 	 * Note: need to block udp_input while changing
1057 	 * the udp pcb queue and/or pcb addresses.
1058 	 */
1059 	switch (req) {
1060 
1061 	case PRU_ATTACH:
1062 		if (inp != NULL) {
1063 			error = EINVAL;
1064 			break;
1065 		}
1066 		s = splsoftnet();
1067 		error = in_pcballoc(so, &udbtable);
1068 		splx(s);
1069 		if (error)
1070 			break;
1071 		error = soreserve(so, udp_sendspace, udp_recvspace);
1072 		if (error)
1073 			break;
1074 #ifdef INET6
1075 		if (((struct inpcb *)so->so_pcb)->inp_flags & INP_IPV6)
1076 			((struct inpcb *) so->so_pcb)->inp_ipv6.ip6_hlim =
1077 			    ip6_defhlim;
1078 		else
1079 #endif /* INET6 */
1080 			((struct inpcb *) so->so_pcb)->inp_ip.ip_ttl = ip_defttl;
1081 		break;
1082 
1083 	case PRU_DETACH:
1084 		udp_detach(inp);
1085 		break;
1086 
1087 	case PRU_BIND:
1088 		s = splsoftnet();
1089 #ifdef INET6
1090 		if (inp->inp_flags & INP_IPV6)
1091 			error = in6_pcbbind(inp, addr);
1092 		else
1093 #endif
1094 			error = in_pcbbind(inp, addr);
1095 		splx(s);
1096 		break;
1097 
1098 	case PRU_LISTEN:
1099 		error = EOPNOTSUPP;
1100 		break;
1101 
1102 	case PRU_CONNECT:
1103 #ifdef INET6
1104 		if (inp->inp_flags & INP_IPV6) {
1105 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6)) {
1106 				error = EISCONN;
1107 				break;
1108 			}
1109 			s = splsoftnet();
1110 			error = in6_pcbconnect(inp, addr);
1111 			splx(s);
1112 		} else
1113 #endif /* INET6 */
1114 		{
1115 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
1116 				error = EISCONN;
1117 				break;
1118 			}
1119 			s = splsoftnet();
1120 			error = in_pcbconnect(inp, addr);
1121 			splx(s);
1122 		}
1123 
1124 		if (error == 0)
1125 			soisconnected(so);
1126 		break;
1127 
1128 	case PRU_CONNECT2:
1129 		error = EOPNOTSUPP;
1130 		break;
1131 
1132 	case PRU_ACCEPT:
1133 		error = EOPNOTSUPP;
1134 		break;
1135 
1136 	case PRU_DISCONNECT:
1137 #ifdef INET6
1138 		if (inp->inp_flags & INP_IPV6) {
1139 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6)) {
1140 				error = ENOTCONN;
1141 				break;
1142 			}
1143 		} else
1144 #endif /* INET6 */
1145 		{
1146 			if (inp->inp_faddr.s_addr == INADDR_ANY) {
1147 				error = ENOTCONN;
1148 				break;
1149 			}
1150 		}
1151 
1152 		s = splsoftnet();
1153 		in_pcbdisconnect(inp);
1154 #ifdef INET6
1155 		if (inp->inp_flags & INP_IPV6)
1156 			inp->inp_laddr6 = in6addr_any;
1157 		else
1158 #endif /* INET6 */
1159 			inp->inp_laddr.s_addr = INADDR_ANY;
1160 
1161 		splx(s);
1162 		so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1163 		break;
1164 
1165 	case PRU_SHUTDOWN:
1166 		socantsendmore(so);
1167 		break;
1168 
1169 	case PRU_SEND:
1170 #ifdef INET6
1171 		if (inp->inp_flags & INP_IPV6)
1172 			return (udp6_output(inp, m, addr, control));
1173 		else
1174 			return (udp_output(m, inp, addr, control));
1175 #else
1176 		return (udp_output(m, inp, addr, control));
1177 #endif
1178 
1179 	case PRU_ABORT:
1180 		soisdisconnected(so);
1181 		udp_detach(inp);
1182 		break;
1183 
1184 	case PRU_SOCKADDR:
1185 #ifdef INET6
1186 		if (inp->inp_flags & INP_IPV6)
1187 			in6_setsockaddr(inp, addr);
1188 		else
1189 #endif /* INET6 */
1190 			in_setsockaddr(inp, addr);
1191 		break;
1192 
1193 	case PRU_PEERADDR:
1194 #ifdef INET6
1195 		if (inp->inp_flags & INP_IPV6)
1196 			in6_setpeeraddr(inp, addr);
1197 		else
1198 #endif /* INET6 */
1199 			in_setpeeraddr(inp, addr);
1200 		break;
1201 
1202 	case PRU_SENSE:
1203 		/*
1204 		 * stat: don't bother with a blocksize.
1205 		 */
1206 		/*
1207 		 * Perhaps Path MTU might be returned for a connected
1208 		 * UDP socket in this case.
1209 		 */
1210 		return (0);
1211 
1212 	case PRU_SENDOOB:
1213 	case PRU_FASTTIMO:
1214 	case PRU_SLOWTIMO:
1215 	case PRU_PROTORCV:
1216 	case PRU_PROTOSEND:
1217 		error =  EOPNOTSUPP;
1218 		break;
1219 
1220 	case PRU_RCVD:
1221 	case PRU_RCVOOB:
1222 		return (EOPNOTSUPP);	/* do not free mbuf's */
1223 
1224 	default:
1225 		panic("udp_usrreq");
1226 	}
1227 
1228 release:
1229 	if (control) {
1230 		m_freem(control);
1231 	}
1232 	if (m)
1233 		m_freem(m);
1234 	return (error);
1235 }
1236 
1237 static void
1238 udp_detach(inp)
1239 	struct inpcb *inp;
1240 {
1241 	int s = splsoftnet();
1242 
1243 	in_pcbdetach(inp);
1244 	splx(s);
1245 }
1246 
1247 /*
1248  * Sysctl for udp variables.
1249  */
1250 int
1251 udp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1252 	int *name;
1253 	u_int namelen;
1254 	void *oldp;
1255 	size_t *oldlenp;
1256 	void *newp;
1257 	size_t newlen;
1258 {
1259 	/* All sysctl names at this level are terminal. */
1260 	if (namelen != 1)
1261 		return (ENOTDIR);
1262 
1263 	switch (name[0]) {
1264 	case UDPCTL_CHECKSUM:
1265 		return (sysctl_int(oldp, oldlenp, newp, newlen, &udpcksum));
1266 	case UDPCTL_BADDYNAMIC:
1267 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1268 		    baddynamicports.udp, sizeof(baddynamicports.udp)));
1269 	case UDPCTL_RECVSPACE:
1270 		return (sysctl_int(oldp, oldlenp, newp, newlen,&udp_recvspace));
1271 	case UDPCTL_SENDSPACE:
1272 		return (sysctl_int(oldp, oldlenp, newp, newlen,&udp_sendspace));
1273 	default:
1274 		return (ENOPROTOOPT);
1275 	}
1276 	/* NOTREACHED */
1277 }
1278