xref: /openbsd-src/sys/netinet/ip_output.c (revision 5054e3e78af0749a9bb00ba9a024b3ee2d90290f)
1 /*	$OpenBSD: ip_output.c,v 1.198 2009/11/13 14:14:56 claudio Exp $	*/
2 /*	$NetBSD: ip_output.c,v 1.28 1996/02/13 23:43:07 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  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
33  */
34 
35 #include "pf.h"
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mbuf.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/socketvar.h>
43 #include <sys/proc.h>
44 #include <sys/kernel.h>
45 
46 #include <net/if.h>
47 #include <net/if_enc.h>
48 #include <net/route.h>
49 
50 #include <netinet/in.h>
51 #include <netinet/in_systm.h>
52 #include <netinet/ip.h>
53 #include <netinet/in_pcb.h>
54 #include <netinet/in_var.h>
55 #include <netinet/ip_var.h>
56 #include <netinet/ip_icmp.h>
57 #include <netinet/tcp.h>
58 #include <netinet/udp.h>
59 #include <netinet/tcp_timer.h>
60 #include <netinet/tcp_var.h>
61 #include <netinet/udp_var.h>
62 
63 #if NPF > 0
64 #include <net/pfvar.h>
65 #endif
66 
67 #ifdef IPSEC
68 #ifdef ENCDEBUG
69 #define DPRINTF(x)    do { if (encdebug) printf x ; } while (0)
70 #else
71 #define DPRINTF(x)
72 #endif
73 
74 extern u_int8_t get_sa_require(struct inpcb *);
75 
76 extern int ipsec_auth_default_level;
77 extern int ipsec_esp_trans_default_level;
78 extern int ipsec_esp_network_default_level;
79 extern int ipsec_ipcomp_default_level;
80 extern int ipforwarding;
81 #endif /* IPSEC */
82 
83 #ifdef MROUTING
84 extern int ipmforwarding;
85 #endif
86 
87 struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
88 void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *);
89 
90 /*
91  * IP output.  The packet in mbuf chain m contains a skeletal IP
92  * header (with len, off, ttl, proto, tos, src, dst).
93  * The mbuf chain containing the packet will be freed.
94  * The mbuf opt, if present, will not be freed.
95  */
96 int
97 ip_output(struct mbuf *m0, ...)
98 {
99 	struct ip *ip;
100 	struct ifnet *ifp;
101 	struct mbuf *m = m0;
102 	int hlen = sizeof (struct ip);
103 	int len, error = 0;
104 	struct route iproute;
105 	struct sockaddr_in *dst;
106 	struct in_ifaddr *ia;
107 	struct mbuf *opt;
108 	struct route *ro;
109 	int flags;
110 	struct ip_moptions *imo;
111 	va_list ap;
112 	u_int8_t sproto = 0, donerouting = 0;
113 	u_long mtu;
114 #ifdef IPSEC
115 	u_int32_t icmp_mtu = 0;
116 	union sockaddr_union sdst;
117 	u_int32_t sspi;
118 	struct m_tag *mtag;
119 	struct tdb_ident *tdbi;
120 
121 	struct inpcb *inp;
122 	struct tdb *tdb;
123 	int s;
124 #endif /* IPSEC */
125 
126 	va_start(ap, m0);
127 	opt = va_arg(ap, struct mbuf *);
128 	ro = va_arg(ap, struct route *);
129 	flags = va_arg(ap, int);
130 	imo = va_arg(ap, struct ip_moptions *);
131 #ifdef IPSEC
132 	inp = va_arg(ap, struct inpcb *);
133 	if (inp && (inp->inp_flags & INP_IPV6) != 0)
134 		panic("ip_output: IPv6 pcb is passed");
135 #endif /* IPSEC */
136 	va_end(ap);
137 
138 #ifdef	DIAGNOSTIC
139 	if ((m->m_flags & M_PKTHDR) == 0)
140 		panic("ip_output no HDR");
141 #endif
142 	if (opt) {
143 		m = ip_insertoptions(m, opt, &len);
144 		hlen = len;
145 	}
146 
147 	ip = mtod(m, struct ip *);
148 
149 	/*
150 	 * Fill in IP header.
151 	 */
152 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
153 		ip->ip_v = IPVERSION;
154 		ip->ip_off &= htons(IP_DF);
155 		ip->ip_id = htons(ip_randomid());
156 		ip->ip_hl = hlen >> 2;
157 		ipstat.ips_localout++;
158 	} else {
159 		hlen = ip->ip_hl << 2;
160 	}
161 
162 	/*
163 	 * If we're missing the IP source address, do a route lookup. We'll
164 	 * remember this result, in case we don't need to do any IPsec
165 	 * processing on the packet. We need the source address so we can
166 	 * do an SPD lookup in IPsec; for most packets, the source address
167 	 * is set at a higher level protocol. ICMPs and other packets
168 	 * though (e.g., traceroute) have a source address of zeroes.
169 	 */
170 	if (ip->ip_src.s_addr == INADDR_ANY) {
171 		if (flags & IP_ROUTETOETHER) {
172 			error = EINVAL;
173 			goto bad;
174 		}
175 		donerouting = 1;
176 
177 		if (ro == 0) {
178 			ro = &iproute;
179 			bzero((caddr_t)ro, sizeof (*ro));
180 		}
181 
182 		dst = satosin(&ro->ro_dst);
183 
184 		/*
185 		 * If there is a cached route, check that it is to the same
186 		 * destination and is still up.  If not, free it and try again.
187 		 */
188 		if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
189 				  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
190 			RTFREE(ro->ro_rt);
191 			ro->ro_rt = (struct rtentry *)0;
192 		}
193 
194 		if (ro->ro_rt == 0) {
195 			dst->sin_family = AF_INET;
196 			dst->sin_len = sizeof(*dst);
197 			dst->sin_addr = ip->ip_dst;
198 		}
199 
200 		/*
201 		 * If routing to interface only, short-circuit routing lookup.
202 		 */
203 		if (flags & IP_ROUTETOIF) {
204 			if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
205 			    m->m_pkthdr.rdomain))) == 0 &&
206 			    (ia = ifatoia(ifa_ifwithnet(sintosa(dst),
207 			    m->m_pkthdr.rdomain))) == 0) {
208 				ipstat.ips_noroute++;
209 				error = ENETUNREACH;
210 				goto bad;
211 			}
212 
213 			ifp = ia->ia_ifp;
214 			mtu = ifp->if_mtu;
215 			ip->ip_ttl = 1;
216 		} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
217 		    (ip->ip_dst.s_addr == INADDR_BROADCAST)) &&
218 		    imo != NULL && imo->imo_multicast_ifp != NULL) {
219 			ifp = imo->imo_multicast_ifp;
220 			mtu = ifp->if_mtu;
221 			IFP_TO_IA(ifp, ia);
222 		} else {
223 			if (ro->ro_rt == 0)
224 				rtalloc_mpath(ro, NULL,
225 				    m->m_pkthdr.rdomain);
226 
227 			if (ro->ro_rt == 0) {
228 				ipstat.ips_noroute++;
229 				error = EHOSTUNREACH;
230 				goto bad;
231 			}
232 
233 			ia = ifatoia(ro->ro_rt->rt_ifa);
234 			ifp = ro->ro_rt->rt_ifp;
235 			if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
236 				mtu = ifp->if_mtu;
237 			ro->ro_rt->rt_use++;
238 
239 			if (ro->ro_rt->rt_flags & RTF_GATEWAY)
240 				dst = satosin(ro->ro_rt->rt_gateway);
241 		}
242 
243 		/* Set the source IP address */
244 		if (!IN_MULTICAST(ip->ip_dst.s_addr))
245 			ip->ip_src = ia->ia_addr.sin_addr;
246 	}
247 
248 #if NPF > 0
249 reroute:
250 #endif
251 
252 #ifdef IPSEC
253 	if (!ipsec_in_use && inp == NULL)
254 		goto done_spd;
255 
256 	/*
257 	 * splnet is chosen over spltdb because we are not allowed to
258 	 * lower the level, and udp_output calls us in splnet().
259 	 */
260 	s = splnet();
261 
262 	/* Do we have any pending SAs to apply ? */
263 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
264 	if (mtag != NULL) {
265 #ifdef DIAGNOSTIC
266 		if (mtag->m_tag_len != sizeof (struct tdb_ident))
267 			panic("ip_output: tag of length %d (should be %d",
268 			    mtag->m_tag_len, sizeof (struct tdb_ident));
269 #endif
270 		tdbi = (struct tdb_ident *)(mtag + 1);
271 		tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
272 		if (tdb == NULL)
273 			error = -EINVAL;
274 		m_tag_delete(m, mtag);
275 	}
276 	else
277 		tdb = ipsp_spd_lookup(m, AF_INET, hlen, &error,
278 		    IPSP_DIRECTION_OUT, NULL, inp);
279 
280 	if (tdb == NULL) {
281 		splx(s);
282 
283 		if (error == 0) {
284 			/*
285 			 * No IPsec processing required, we'll just send the
286 			 * packet out.
287 			 */
288 			sproto = 0;
289 
290 			/* Fall through to routing/multicast handling */
291 		} else {
292 			/*
293 			 * -EINVAL is used to indicate that the packet should
294 			 * be silently dropped, typically because we've asked
295 			 * key management for an SA.
296 			 */
297 			if (error == -EINVAL) /* Should silently drop packet */
298 			  error = 0;
299 
300 			m_freem(m);
301 			goto done;
302 		}
303 	} else {
304 		/* Loop detection */
305 		for (mtag = m_tag_first(m); mtag != NULL;
306 		    mtag = m_tag_next(m, mtag)) {
307 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
308 			    mtag->m_tag_id !=
309 			    PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
310 				continue;
311 			tdbi = (struct tdb_ident *)(mtag + 1);
312 			if (tdbi->spi == tdb->tdb_spi &&
313 			    tdbi->proto == tdb->tdb_sproto &&
314 			    !bcmp(&tdbi->dst, &tdb->tdb_dst,
315 			    sizeof(union sockaddr_union))) {
316 				splx(s);
317 				sproto = 0; /* mark as no-IPsec-needed */
318 				goto done_spd;
319 			}
320 		}
321 
322 		/* We need to do IPsec */
323 		bcopy(&tdb->tdb_dst, &sdst, sizeof(sdst));
324 		sspi = tdb->tdb_spi;
325 		sproto = tdb->tdb_sproto;
326 		splx(s);
327 
328 		/*
329 		 * If it needs TCP/UDP hardware-checksumming, do the
330 		 * computation now.
331 		 */
332 		if (m->m_pkthdr.csum_flags & (M_TCPV4_CSUM_OUT | M_UDPV4_CSUM_OUT)) {
333 			in_delayed_cksum(m);
334 			m->m_pkthdr.csum_flags &=
335 			    ~(M_UDPV4_CSUM_OUT | M_TCPV4_CSUM_OUT);
336 		}
337 
338 		/* If it's not a multicast packet, try to fast-path */
339 		if (!IN_MULTICAST(ip->ip_dst.s_addr)) {
340 			goto sendit;
341 		}
342 	}
343 
344 	/* Fall through to the routing/multicast handling code */
345  done_spd:
346 #endif /* IPSEC */
347 
348 	if (flags & IP_ROUTETOETHER) {
349 		dst = satosin(&ro->ro_dst);
350 		ifp = ro->ro_rt->rt_ifp;
351 		mtu = ifp->if_mtu;
352 		ro->ro_rt = NULL;
353 	} else if (donerouting == 0) {
354 		if (ro == 0) {
355 			ro = &iproute;
356 			bzero((caddr_t)ro, sizeof (*ro));
357 		}
358 
359 		dst = satosin(&ro->ro_dst);
360 
361 		/*
362 		 * If there is a cached route, check that it is to the same
363 		 * destination and is still up.  If not, free it and try again.
364 		 */
365 		if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
366 				  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
367 			RTFREE(ro->ro_rt);
368 			ro->ro_rt = (struct rtentry *)0;
369 		}
370 
371 		if (ro->ro_rt == 0) {
372 			dst->sin_family = AF_INET;
373 			dst->sin_len = sizeof(*dst);
374 			dst->sin_addr = ip->ip_dst;
375 		}
376 
377 		/*
378 		 * If routing to interface only, short-circuit routing lookup.
379 		 */
380 		if (flags & IP_ROUTETOIF) {
381 			if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
382 			    m->m_pkthdr.rdomain))) == 0 &&
383 			    (ia = ifatoia(ifa_ifwithnet(sintosa(dst),
384 			    m->m_pkthdr.rdomain))) == 0) {
385 				ipstat.ips_noroute++;
386 				error = ENETUNREACH;
387 				goto bad;
388 			}
389 
390 			ifp = ia->ia_ifp;
391 			mtu = ifp->if_mtu;
392 			ip->ip_ttl = 1;
393 		} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
394 		    (ip->ip_dst.s_addr == INADDR_BROADCAST)) &&
395 		    imo != NULL && imo->imo_multicast_ifp != NULL) {
396 			ifp = imo->imo_multicast_ifp;
397 			mtu = ifp->if_mtu;
398 			IFP_TO_IA(ifp, ia);
399 		} else {
400 			if (ro->ro_rt == 0)
401 				rtalloc_mpath(ro, &ip->ip_src.s_addr,
402 				    m->m_pkthdr.rdomain);
403 
404 			if (ro->ro_rt == 0) {
405 				ipstat.ips_noroute++;
406 				error = EHOSTUNREACH;
407 				goto bad;
408 			}
409 
410 			ia = ifatoia(ro->ro_rt->rt_ifa);
411 			ifp = ro->ro_rt->rt_ifp;
412 			if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
413 				mtu = ifp->if_mtu;
414 			ro->ro_rt->rt_use++;
415 
416 			if (ro->ro_rt->rt_flags & RTF_GATEWAY)
417 				dst = satosin(ro->ro_rt->rt_gateway);
418 		}
419 
420 		/* Set the source IP address */
421 		if (ip->ip_src.s_addr == INADDR_ANY)
422 			ip->ip_src = ia->ia_addr.sin_addr;
423 	}
424 
425 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
426 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
427 		struct in_multi *inm;
428 
429 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
430 			M_BCAST : M_MCAST;
431 
432 		/*
433 		 * IP destination address is multicast.  Make sure "dst"
434 		 * still points to the address in "ro".  (It may have been
435 		 * changed to point to a gateway address, above.)
436 		 */
437 		dst = satosin(&ro->ro_dst);
438 
439 		/*
440 		 * See if the caller provided any multicast options
441 		 */
442 		if (imo != NULL)
443 			ip->ip_ttl = imo->imo_multicast_ttl;
444 		else
445 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
446 
447 		/*
448 		 * if we don't know the outgoing ifp yet, we can't generate
449 		 * output
450 		 */
451 		if (!ifp) {
452 			ipstat.ips_noroute++;
453 			error = EHOSTUNREACH;
454 			goto bad;
455 		}
456 
457 		/*
458 		 * Confirm that the outgoing interface supports multicast,
459 		 * but only if the packet actually is going out on that
460 		 * interface (i.e., no IPsec is applied).
461 		 */
462 		if ((((m->m_flags & M_MCAST) &&
463 		      (ifp->if_flags & IFF_MULTICAST) == 0) ||
464 		     ((m->m_flags & M_BCAST) &&
465 		      (ifp->if_flags & IFF_BROADCAST) == 0)) && (sproto == 0)) {
466 			ipstat.ips_noroute++;
467 			error = ENETUNREACH;
468 			goto bad;
469 		}
470 
471 		/*
472 		 * If source address not specified yet, use address
473 		 * of outgoing interface.
474 		 */
475 		if (ip->ip_src.s_addr == INADDR_ANY) {
476 			struct in_ifaddr *ia;
477 
478 			TAILQ_FOREACH(ia, &in_ifaddr, ia_list)
479 				if (ia->ia_ifp == ifp) {
480 					ip->ip_src = ia->ia_addr.sin_addr;
481 					break;
482 				}
483 		}
484 
485 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
486 		if (inm != NULL &&
487 		   (imo == NULL || imo->imo_multicast_loop)) {
488 			/*
489 			 * If we belong to the destination multicast group
490 			 * on the outgoing interface, and the caller did not
491 			 * forbid loopback, loop back a copy.
492 			 * Can't defer TCP/UDP checksumming, do the
493 			 * computation now.
494 			 */
495 			if (m->m_pkthdr.csum_flags &
496 			    (M_TCPV4_CSUM_OUT | M_UDPV4_CSUM_OUT)) {
497 				in_delayed_cksum(m);
498 				m->m_pkthdr.csum_flags &=
499 				    ~(M_UDPV4_CSUM_OUT | M_TCPV4_CSUM_OUT);
500 			}
501 			ip_mloopback(ifp, m, dst);
502 		}
503 #ifdef MROUTING
504 		else {
505 			/*
506 			 * If we are acting as a multicast router, perform
507 			 * multicast forwarding as if the packet had just
508 			 * arrived on the interface to which we are about
509 			 * to send.  The multicast forwarding function
510 			 * recursively calls this function, using the
511 			 * IP_FORWARDING flag to prevent infinite recursion.
512 			 *
513 			 * Multicasts that are looped back by ip_mloopback(),
514 			 * above, will be forwarded by the ip_input() routine,
515 			 * if necessary.
516 			 */
517 			extern struct socket *ip_mrouter;
518 
519 			if (ipmforwarding && ip_mrouter &&
520 			    (flags & IP_FORWARDING) == 0) {
521 				if (ip_mforward(m, ifp) != 0) {
522 					m_freem(m);
523 					goto done;
524 				}
525 			}
526 		}
527 #endif
528 		/*
529 		 * Multicasts with a time-to-live of zero may be looped-
530 		 * back, above, but must not be transmitted on a network.
531 		 * Also, multicasts addressed to the loopback interface
532 		 * are not sent -- the above call to ip_mloopback() will
533 		 * loop back a copy if this host actually belongs to the
534 		 * destination group on the loopback interface.
535 		 */
536 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
537 			m_freem(m);
538 			goto done;
539 		}
540 
541 		goto sendit;
542 	}
543 
544 	/*
545 	 * Look for broadcast address and and verify user is allowed to send
546 	 * such a packet; if the packet is going in an IPsec tunnel, skip
547 	 * this check.
548 	 */
549 	if ((sproto == 0) && (in_broadcast(dst->sin_addr, ifp))) {
550 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
551 			error = EADDRNOTAVAIL;
552 			goto bad;
553 		}
554 		if ((flags & IP_ALLOWBROADCAST) == 0) {
555 			error = EACCES;
556 			goto bad;
557 		}
558 
559 		/* Don't allow broadcast messages to be fragmented */
560 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
561 			error = EMSGSIZE;
562 			goto bad;
563 		}
564 		m->m_flags |= M_BCAST;
565 	} else
566 		m->m_flags &= ~M_BCAST;
567 
568 sendit:
569 	/*
570 	 * If we're doing Path MTU discovery, we need to set DF unless
571 	 * the route's MTU is locked.
572 	 */
573 	if ((flags & IP_MTUDISC) && ro && ro->ro_rt &&
574 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
575 		ip->ip_off |= htons(IP_DF);
576 
577 #ifdef IPSEC
578 	/*
579 	 * Check if the packet needs encapsulation.
580 	 */
581 	if (sproto != 0) {
582 		s = splnet();
583 
584 		/*
585 		 * Packet filter
586 		 */
587 #if NPF > 0
588 
589 		if (pf_test(PF_OUT, &encif[0].sc_if, &m, NULL) != PF_PASS) {
590 			error = EHOSTUNREACH;
591 			splx(s);
592 			m_freem(m);
593 			goto done;
594 		}
595 		if (m == NULL) {
596 			splx(s);
597 			goto done;
598 		}
599 		ip = mtod(m, struct ip *);
600 		hlen = ip->ip_hl << 2;
601 		/*
602 		 * PF_TAG_REROUTE handling or not...
603 		 * Packet is entering IPsec so the routing is
604 		 * already overruled by the IPsec policy.
605 		 * Until now the change was not reconsidered.
606 		 * What's the behaviour?
607 		 */
608 #endif
609 
610 		tdb = gettdb(sspi, &sdst, sproto);
611 		if (tdb == NULL) {
612 			DPRINTF(("ip_output: unknown TDB"));
613 			error = EHOSTUNREACH;
614 			splx(s);
615 			m_freem(m);
616 			goto done;
617 		}
618 
619 		/* Check if we are allowed to fragment */
620 		if (ip_mtudisc && (ip->ip_off & htons(IP_DF)) && tdb->tdb_mtu &&
621 		    ntohs(ip->ip_len) > tdb->tdb_mtu &&
622 		    tdb->tdb_mtutimeout > time_second) {
623 			struct rtentry *rt = NULL;
624 			int rt_mtucloned = 0;
625 			int transportmode = 0;
626 
627 			transportmode = (tdb->tdb_dst.sa.sa_family == AF_INET) &&
628 			    (tdb->tdb_dst.sin.sin_addr.s_addr ==
629 			    ip->ip_dst.s_addr);
630 			icmp_mtu = tdb->tdb_mtu;
631 			splx(s);
632 
633 			/* Find a host route to store the mtu in */
634 			if (ro != NULL)
635 				rt = ro->ro_rt;
636 			/* but don't add a PMTU route for transport mode SAs */
637 			if (transportmode)
638 				rt = NULL;
639 			else if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0) {
640 				struct sockaddr_in dst = {
641 					sizeof(struct sockaddr_in), AF_INET};
642 				dst.sin_addr = ip->ip_dst;
643 				rt = icmp_mtudisc_clone((struct sockaddr *)&dst,
644 				    m->m_pkthdr.rdomain);
645 				rt_mtucloned = 1;
646 			}
647 			DPRINTF(("ip_output: spi %08x mtu %d rt %p cloned %d\n",
648 			    ntohl(tdb->tdb_spi), icmp_mtu, rt, rt_mtucloned));
649 			if (rt != NULL) {
650 				rt->rt_rmx.rmx_mtu = icmp_mtu;
651 				if (ro && ro->ro_rt != NULL) {
652 					RTFREE(ro->ro_rt);
653 					ro->ro_rt = NULL;
654 					rtalloc1(&ro->ro_dst, 1,
655 					    m->m_pkthdr.rdomain);
656 				}
657 				if (rt_mtucloned)
658 					rtfree(rt);
659 			}
660 			error = EMSGSIZE;
661 			goto bad;
662 		}
663 
664 		/*
665 		 * Clear these -- they'll be set in the recursive invocation
666 		 * as needed.
667 		 */
668 		m->m_flags &= ~(M_MCAST | M_BCAST);
669 
670 		/* Callee frees mbuf */
671 		error = ipsp_process_packet(m, tdb, AF_INET, 0);
672 		splx(s);
673 		return error;  /* Nothing more to be done */
674 	}
675 
676 	/*
677 	 * If deferred crypto processing is needed, check that the
678 	 * interface supports it.
679 	 */
680 	if (ipsec_in_use && (mtag = m_tag_find(m,
681 	    PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL)) != NULL &&
682 	    (ifp->if_capabilities & IFCAP_IPSEC) == 0) {
683 		/* Notify IPsec to do its own crypto. */
684 		ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
685 		m_freem(m);
686 		error = EHOSTUNREACH;
687 		goto done;
688 	}
689 #endif /* IPSEC */
690 
691 	/* Catch routing changes wrt. hardware checksumming for TCP or UDP. */
692 	if (m->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) {
693 		if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) ||
694 		    ifp->if_bridge != NULL) {
695 			in_delayed_cksum(m);
696 			m->m_pkthdr.csum_flags &= ~M_TCPV4_CSUM_OUT; /* Clear */
697 		}
698 	} else if (m->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) {
699 		if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) ||
700 		    ifp->if_bridge != NULL) {
701 			in_delayed_cksum(m);
702 			m->m_pkthdr.csum_flags &= ~M_UDPV4_CSUM_OUT; /* Clear */
703 		}
704 	}
705 
706 	/*
707 	 * Packet filter
708 	 */
709 #if NPF > 0
710 	if (pf_test(PF_OUT, ifp, &m, NULL) != PF_PASS) {
711 		error = EHOSTUNREACH;
712 		m_freem(m);
713 		goto done;
714 	}
715 	if (m == NULL)
716 		goto done;
717 	ip = mtod(m, struct ip *);
718 	hlen = ip->ip_hl << 2;
719 	if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
720 	    (PF_TAG_REROUTE | PF_TAG_GENERATED))
721 		/* already rerun the route lookup, go on */
722 		m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
723 	else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
724 		/* tag as generated to skip over pf_test on rerun */
725 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
726 		ro = NULL;
727 		donerouting = 0;
728 		goto reroute;
729 	}
730 #endif
731 
732 #ifdef IPSEC
733 	if (ipsec_in_use && (flags & IP_FORWARDING) && (ipforwarding == 2) &&
734 	    (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) == NULL)) {
735 		error = EHOSTUNREACH;
736 		m_freem(m);
737 		goto done;
738 	}
739 #endif
740 
741 	/* XXX
742 	 * Try to use jumbograms based on socket option, or the route
743 	 * or... for other reasons later on.
744 	 */
745 	if ((flags & IP_JUMBO) && ro->ro_rt && (ro->ro_rt->rt_flags & RTF_JUMBO) &&
746 	    ro->ro_rt->rt_ifp)
747 		mtu = ro->ro_rt->rt_ifp->if_hardmtu;
748 
749 	/*
750 	 * If small enough for interface, can just send directly.
751 	 */
752 	if (ntohs(ip->ip_len) <= mtu) {
753 		ip->ip_sum = 0;
754 		if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
755 		    ifp->if_bridge == NULL) {
756 			m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
757 			ipstat.ips_outhwcsum++;
758 		} else
759 			ip->ip_sum = in_cksum(m, hlen);
760 		/* Update relevant hardware checksum stats for TCP/UDP */
761 		if (m->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT)
762 			tcpstat.tcps_outhwcsum++;
763 		else if (m->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT)
764 			udpstat.udps_outhwcsum++;
765 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
766 		goto done;
767 	}
768 
769 	/*
770 	 * Too large for interface; fragment if possible.
771 	 * Must be able to put at least 8 bytes per fragment.
772 	 */
773 	if (ip->ip_off & htons(IP_DF)) {
774 #ifdef IPSEC
775 		icmp_mtu = ifp->if_mtu;
776 #endif
777 		error = EMSGSIZE;
778 		/*
779 		 * This case can happen if the user changed the MTU
780 		 * of an interface after enabling IP on it.  Because
781 		 * most netifs don't keep track of routes pointing to
782 		 * them, there is no way for one to update all its
783 		 * routes when the MTU is changed.
784 		 */
785 		if (ro->ro_rt != NULL &&
786 		    (ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
787 		    !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
788 		    (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
789 			ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
790 		}
791 		ipstat.ips_cantfrag++;
792 		goto bad;
793 	}
794 
795 	error = ip_fragment(m, ifp, mtu);
796 	if (error) {
797 		m = m0 = NULL;
798 		goto bad;
799 	}
800 
801 	for (; m; m = m0) {
802 		m0 = m->m_nextpkt;
803 		m->m_nextpkt = 0;
804 		if (error == 0)
805 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
806 			    ro->ro_rt);
807 		else
808 			m_freem(m);
809 	}
810 
811 	if (error == 0)
812 		ipstat.ips_fragmented++;
813 
814 done:
815 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt)
816 		RTFREE(ro->ro_rt);
817 	return (error);
818 bad:
819 #ifdef IPSEC
820 	if (error == EMSGSIZE && ip_mtudisc && icmp_mtu != 0 && m != NULL)
821 		ipsec_adjust_mtu(m, icmp_mtu);
822 #endif
823 	m_freem(m0);
824 	goto done;
825 }
826 
827 int
828 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
829 {
830 	struct ip *ip, *mhip;
831 	struct mbuf *m0;
832 	int len, hlen, off;
833 	int mhlen, firstlen;
834 	struct mbuf **mnext;
835 	int fragments = 0;
836 	int s;
837 	int error = 0;
838 
839 	ip = mtod(m, struct ip *);
840 	hlen = ip->ip_hl << 2;
841 
842 	len = (mtu - hlen) &~ 7;
843 	if (len < 8) {
844 		m_freem(m);
845 		return (EMSGSIZE);
846 	}
847 
848 	/*
849 	 * If we are doing fragmentation, we can't defer TCP/UDP
850 	 * checksumming; compute the checksum and clear the flag.
851 	 */
852 	if (m->m_pkthdr.csum_flags & (M_TCPV4_CSUM_OUT | M_UDPV4_CSUM_OUT)) {
853 		in_delayed_cksum(m);
854 		m->m_pkthdr.csum_flags &= ~(M_UDPV4_CSUM_OUT | M_TCPV4_CSUM_OUT);
855 	}
856 
857 	firstlen = len;
858 	mnext = &m->m_nextpkt;
859 
860 	/*
861 	 * Loop through length of segment after first fragment,
862 	 * make new header and copy data of each part and link onto chain.
863 	 */
864 	m0 = m;
865 	mhlen = sizeof (struct ip);
866 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
867 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
868 		if (m == 0) {
869 			ipstat.ips_odropped++;
870 			error = ENOBUFS;
871 			goto sendorfree;
872 		}
873 		*mnext = m;
874 		mnext = &m->m_nextpkt;
875 		m->m_data += max_linkhdr;
876 		mhip = mtod(m, struct ip *);
877 		*mhip = *ip;
878 		/* we must inherit MCAST and BCAST flags and rdomain */
879 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
880 		m->m_pkthdr.rdomain = m0->m_pkthdr.rdomain;
881 		if (hlen > sizeof (struct ip)) {
882 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
883 			mhip->ip_hl = mhlen >> 2;
884 		}
885 		m->m_len = mhlen;
886 		mhip->ip_off = ((off - hlen) >> 3) +
887 		    (ntohs(ip->ip_off) & ~IP_MF);
888 		if (ip->ip_off & htons(IP_MF))
889 			mhip->ip_off |= IP_MF;
890 		if (off + len >= ntohs(ip->ip_len))
891 			len = ntohs(ip->ip_len) - off;
892 		else
893 			mhip->ip_off |= IP_MF;
894 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
895 		m->m_next = m_copy(m0, off, len);
896 		if (m->m_next == 0) {
897 			ipstat.ips_odropped++;
898 			error = ENOBUFS;
899 			goto sendorfree;
900 		}
901 		m->m_pkthdr.len = mhlen + len;
902 		m->m_pkthdr.rcvif = (struct ifnet *)0;
903 		mhip->ip_off = htons((u_int16_t)mhip->ip_off);
904 		mhip->ip_sum = 0;
905 		if ((ifp != NULL) &&
906 		    (ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
907 		    ifp->if_bridge == NULL) {
908 			m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
909 			ipstat.ips_outhwcsum++;
910 		} else
911 			mhip->ip_sum = in_cksum(m, mhlen);
912 		ipstat.ips_ofragments++;
913 		fragments++;
914 	}
915 	/*
916 	 * Update first fragment by trimming what's been copied out
917 	 * and updating header, then send each fragment (in order).
918 	 */
919 	m = m0;
920 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
921 	m->m_pkthdr.len = hlen + firstlen;
922 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
923 	ip->ip_off |= htons(IP_MF);
924 	ip->ip_sum = 0;
925 	if ((ifp != NULL) &&
926 	    (ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
927 	    ifp->if_bridge == NULL) {
928 		m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
929 		ipstat.ips_outhwcsum++;
930 	} else
931 		ip->ip_sum = in_cksum(m, hlen);
932 sendorfree:
933 	/*
934 	 * If there is no room for all the fragments, don't queue
935 	 * any of them.
936 	 */
937 	if (ifp != NULL) {
938 		s = splnet();
939 		if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
940 		    error == 0) {
941 			error = ENOBUFS;
942 			ipstat.ips_odropped++;
943 			IFQ_INC_DROPS(&ifp->if_snd);
944 		}
945 		splx(s);
946 	}
947 	if (error) {
948 		for (m = m0; m; m = m0) {
949 			m0 = m->m_nextpkt;
950 			m->m_nextpkt = NULL;
951 			m_freem(m);
952 		}
953 	}
954 
955 	return (error);
956 }
957 
958 /*
959  * Insert IP options into preformed packet.
960  * Adjust IP destination as required for IP source routing,
961  * as indicated by a non-zero in_addr at the start of the options.
962  */
963 struct mbuf *
964 ip_insertoptions(m, opt, phlen)
965 	struct mbuf *m;
966 	struct mbuf *opt;
967 	int *phlen;
968 {
969 	struct ipoption *p = mtod(opt, struct ipoption *);
970 	struct mbuf *n;
971 	struct ip *ip = mtod(m, struct ip *);
972 	unsigned optlen;
973 
974 	optlen = opt->m_len - sizeof(p->ipopt_dst);
975 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
976 		return (m);		/* XXX should fail */
977 	if (p->ipopt_dst.s_addr)
978 		ip->ip_dst = p->ipopt_dst;
979 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
980 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
981 		if (n == 0)
982 			return (m);
983 		M_MOVE_HDR(n, m);
984 		n->m_pkthdr.len += optlen;
985 		m->m_len -= sizeof(struct ip);
986 		m->m_data += sizeof(struct ip);
987 		n->m_next = m;
988 		m = n;
989 		m->m_len = optlen + sizeof(struct ip);
990 		m->m_data += max_linkhdr;
991 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
992 	} else {
993 		m->m_data -= optlen;
994 		m->m_len += optlen;
995 		m->m_pkthdr.len += optlen;
996 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
997 	}
998 	ip = mtod(m, struct ip *);
999 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
1000 	*phlen = sizeof(struct ip) + optlen;
1001 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
1002 	return (m);
1003 }
1004 
1005 /*
1006  * Copy options from ip to jp,
1007  * omitting those not copied during fragmentation.
1008  */
1009 int
1010 ip_optcopy(ip, jp)
1011 	struct ip *ip, *jp;
1012 {
1013 	u_char *cp, *dp;
1014 	int opt, optlen, cnt;
1015 
1016 	cp = (u_char *)(ip + 1);
1017 	dp = (u_char *)(jp + 1);
1018 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1019 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1020 		opt = cp[0];
1021 		if (opt == IPOPT_EOL)
1022 			break;
1023 		if (opt == IPOPT_NOP) {
1024 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1025 			*dp++ = IPOPT_NOP;
1026 			optlen = 1;
1027 			continue;
1028 		}
1029 #ifdef DIAGNOSTIC
1030 		if (cnt < IPOPT_OLEN + sizeof(*cp))
1031 			panic("malformed IPv4 option passed to ip_optcopy");
1032 #endif
1033 		optlen = cp[IPOPT_OLEN];
1034 #ifdef DIAGNOSTIC
1035 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1036 			panic("malformed IPv4 option passed to ip_optcopy");
1037 #endif
1038 		/* bogus lengths should have been caught by ip_dooptions */
1039 		if (optlen > cnt)
1040 			optlen = cnt;
1041 		if (IPOPT_COPIED(opt)) {
1042 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
1043 			dp += optlen;
1044 		}
1045 	}
1046 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1047 		*dp++ = IPOPT_EOL;
1048 	return (optlen);
1049 }
1050 
1051 /*
1052  * IP socket option processing.
1053  */
1054 int
1055 ip_ctloutput(op, so, level, optname, mp)
1056 	int op;
1057 	struct socket *so;
1058 	int level, optname;
1059 	struct mbuf **mp;
1060 {
1061 	struct inpcb *inp = sotoinpcb(so);
1062 	struct mbuf *m = *mp;
1063 	int optval = 0;
1064 #ifdef IPSEC
1065 	struct proc *p = curproc; /* XXX */
1066 	struct ipsec_ref *ipr;
1067 	u_int16_t opt16val;
1068 #endif
1069 	int error = 0;
1070 	u_int rtid = 0;
1071 
1072 	if (level != IPPROTO_IP) {
1073 		error = EINVAL;
1074 		if (op == PRCO_SETOPT && *mp)
1075 			(void) m_free(*mp);
1076 	} else switch (op) {
1077 	case PRCO_SETOPT:
1078 		switch (optname) {
1079 		case IP_OPTIONS:
1080 #ifdef notyet
1081 		case IP_RETOPTS:
1082 			return (ip_pcbopts(optname, &inp->inp_options, m));
1083 #else
1084 			return (ip_pcbopts(&inp->inp_options, m));
1085 #endif
1086 
1087 		case IP_TOS:
1088 		case IP_TTL:
1089 		case IP_MINTTL:
1090 		case IP_RECVOPTS:
1091 		case IP_RECVRETOPTS:
1092 		case IP_RECVDSTADDR:
1093 		case IP_RECVIF:
1094 		case IP_RECVTTL:
1095 		case IP_RECVDSTPORT:
1096 			if (m == NULL || m->m_len != sizeof(int))
1097 				error = EINVAL;
1098 			else {
1099 				optval = *mtod(m, int *);
1100 				switch (optname) {
1101 
1102 				case IP_TOS:
1103 					inp->inp_ip.ip_tos = optval;
1104 					break;
1105 
1106 				case IP_TTL:
1107 					if (optval > 0 && optval <= MAXTTL)
1108 						inp->inp_ip.ip_ttl = optval;
1109 					else
1110 						error = EINVAL;
1111 					break;
1112 
1113 				case IP_MINTTL:
1114 					if (optval > 0 && optval <= MAXTTL)
1115 						inp->inp_ip_minttl = optval;
1116 					else
1117 						error = EINVAL;
1118 					break;
1119 #define	OPTSET(bit) \
1120 	if (optval) \
1121 		inp->inp_flags |= bit; \
1122 	else \
1123 		inp->inp_flags &= ~bit;
1124 
1125 				case IP_RECVOPTS:
1126 					OPTSET(INP_RECVOPTS);
1127 					break;
1128 
1129 				case IP_RECVRETOPTS:
1130 					OPTSET(INP_RECVRETOPTS);
1131 					break;
1132 
1133 				case IP_RECVDSTADDR:
1134 					OPTSET(INP_RECVDSTADDR);
1135 					break;
1136 				case IP_RECVIF:
1137 					OPTSET(INP_RECVIF);
1138 					break;
1139 				case IP_RECVTTL:
1140 					OPTSET(INP_RECVTTL);
1141 					break;
1142 				case IP_RECVDSTPORT:
1143 					OPTSET(INP_RECVDSTPORT);
1144 					break;
1145 				}
1146 			}
1147 			break;
1148 #undef OPTSET
1149 
1150 		case IP_MULTICAST_IF:
1151 		case IP_MULTICAST_TTL:
1152 		case IP_MULTICAST_LOOP:
1153 		case IP_ADD_MEMBERSHIP:
1154 		case IP_DROP_MEMBERSHIP:
1155 			error = ip_setmoptions(optname, &inp->inp_moptions, m,
1156 			    inp->inp_rdomain);
1157 			break;
1158 
1159 		case IP_PORTRANGE:
1160 			if (m == 0 || m->m_len != sizeof(int))
1161 				error = EINVAL;
1162 			else {
1163 				optval = *mtod(m, int *);
1164 
1165 				switch (optval) {
1166 
1167 				case IP_PORTRANGE_DEFAULT:
1168 					inp->inp_flags &= ~(INP_LOWPORT);
1169 					inp->inp_flags &= ~(INP_HIGHPORT);
1170 					break;
1171 
1172 				case IP_PORTRANGE_HIGH:
1173 					inp->inp_flags &= ~(INP_LOWPORT);
1174 					inp->inp_flags |= INP_HIGHPORT;
1175 					break;
1176 
1177 				case IP_PORTRANGE_LOW:
1178 					inp->inp_flags &= ~(INP_HIGHPORT);
1179 					inp->inp_flags |= INP_LOWPORT;
1180 					break;
1181 
1182 				default:
1183 
1184 					error = EINVAL;
1185 					break;
1186 				}
1187 			}
1188 			break;
1189 		case IP_AUTH_LEVEL:
1190 		case IP_ESP_TRANS_LEVEL:
1191 		case IP_ESP_NETWORK_LEVEL:
1192 		case IP_IPCOMP_LEVEL:
1193 #ifndef IPSEC
1194 			error = EOPNOTSUPP;
1195 #else
1196 			if (m == 0 || m->m_len != sizeof(int)) {
1197 				error = EINVAL;
1198 				break;
1199 			}
1200 			optval = *mtod(m, int *);
1201 
1202 			if (optval < IPSEC_LEVEL_BYPASS ||
1203 			    optval > IPSEC_LEVEL_UNIQUE) {
1204 				error = EINVAL;
1205 				break;
1206 			}
1207 
1208 			/* Unlink cached output TDB to force a re-search */
1209 			if (inp->inp_tdb_out) {
1210 				int s = spltdb();
1211 				TAILQ_REMOVE(&inp->inp_tdb_out->tdb_inp_out,
1212 				    inp, inp_tdb_out_next);
1213 				splx(s);
1214 			}
1215 
1216 			if (inp->inp_tdb_in) {
1217 				int s = spltdb();
1218 				TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in,
1219 				    inp, inp_tdb_in_next);
1220 				splx(s);
1221 			}
1222 
1223 			switch (optname) {
1224 			case IP_AUTH_LEVEL:
1225 				if (optval < ipsec_auth_default_level &&
1226 				    suser(p, 0)) {
1227 					error = EACCES;
1228 					break;
1229 				}
1230 				inp->inp_seclevel[SL_AUTH] = optval;
1231 				break;
1232 
1233 			case IP_ESP_TRANS_LEVEL:
1234 				if (optval < ipsec_esp_trans_default_level &&
1235 				    suser(p, 0)) {
1236 					error = EACCES;
1237 					break;
1238 				}
1239 				inp->inp_seclevel[SL_ESP_TRANS] = optval;
1240 				break;
1241 
1242 			case IP_ESP_NETWORK_LEVEL:
1243 				if (optval < ipsec_esp_network_default_level &&
1244 				    suser(p, 0)) {
1245 					error = EACCES;
1246 					break;
1247 				}
1248 				inp->inp_seclevel[SL_ESP_NETWORK] = optval;
1249 				break;
1250 			case IP_IPCOMP_LEVEL:
1251 				if (optval < ipsec_ipcomp_default_level &&
1252 				    suser(p, 0)) {
1253 					error = EACCES;
1254 					break;
1255 				}
1256 				inp->inp_seclevel[SL_IPCOMP] = optval;
1257 				break;
1258 			}
1259 			if (!error)
1260 				inp->inp_secrequire = get_sa_require(inp);
1261 #endif
1262 			break;
1263 
1264 		case IP_IPSEC_REMOTE_CRED:
1265 		case IP_IPSEC_REMOTE_AUTH:
1266 			/* Can't set the remote credential or key */
1267 			error = EOPNOTSUPP;
1268 			break;
1269 
1270 		case IP_IPSEC_LOCAL_ID:
1271 		case IP_IPSEC_REMOTE_ID:
1272 		case IP_IPSEC_LOCAL_CRED:
1273 		case IP_IPSEC_LOCAL_AUTH:
1274 #ifndef IPSEC
1275 			error = EOPNOTSUPP;
1276 #else
1277 			if (m->m_len < 2) {
1278 				error = EINVAL;
1279 				break;
1280 			}
1281 
1282 			m_copydata(m, 0, 2, (caddr_t) &opt16val);
1283 
1284 			/* If the type is 0, then we cleanup and return */
1285 			if (opt16val == 0) {
1286 				switch (optname) {
1287 				case IP_IPSEC_LOCAL_ID:
1288 					if (inp->inp_ipo != NULL &&
1289 					    inp->inp_ipo->ipo_srcid != NULL) {
1290 						ipsp_reffree(inp->inp_ipo->ipo_srcid);
1291 						inp->inp_ipo->ipo_srcid = NULL;
1292 					}
1293 					break;
1294 
1295 				case IP_IPSEC_REMOTE_ID:
1296 					if (inp->inp_ipo != NULL &&
1297 					    inp->inp_ipo->ipo_dstid != NULL) {
1298 						ipsp_reffree(inp->inp_ipo->ipo_dstid);
1299 						inp->inp_ipo->ipo_dstid = NULL;
1300 					}
1301 					break;
1302 
1303 				case IP_IPSEC_LOCAL_CRED:
1304 					if (inp->inp_ipo != NULL &&
1305 					    inp->inp_ipo->ipo_local_cred != NULL) {
1306 						ipsp_reffree(inp->inp_ipo->ipo_local_cred);
1307 						inp->inp_ipo->ipo_local_cred = NULL;
1308 					}
1309 					break;
1310 
1311 				case IP_IPSEC_LOCAL_AUTH:
1312 					if (inp->inp_ipo != NULL &&
1313 					    inp->inp_ipo->ipo_local_auth != NULL) {
1314 						ipsp_reffree(inp->inp_ipo->ipo_local_auth);
1315 						inp->inp_ipo->ipo_local_auth = NULL;
1316 					}
1317 					break;
1318 				}
1319 
1320 				error = 0;
1321 				break;
1322 			}
1323 
1324 			/* Can't have an empty payload */
1325 			if (m->m_len == 2) {
1326 				error = EINVAL;
1327 				break;
1328 			}
1329 
1330 			/* Allocate if needed */
1331 			if (inp->inp_ipo == NULL) {
1332 				inp->inp_ipo = ipsec_add_policy(inp,
1333 				    AF_INET, IPSP_DIRECTION_OUT);
1334 				if (inp->inp_ipo == NULL) {
1335 					error = ENOBUFS;
1336 					break;
1337 				}
1338 			}
1339 
1340 			ipr = malloc(sizeof(struct ipsec_ref) + m->m_len - 2,
1341 			       M_CREDENTIALS, M_NOWAIT);
1342 			if (ipr == NULL) {
1343 				error = ENOBUFS;
1344 				break;
1345 			}
1346 
1347 			ipr->ref_count = 1;
1348 			ipr->ref_malloctype = M_CREDENTIALS;
1349 			ipr->ref_len = m->m_len - 2;
1350 			ipr->ref_type = opt16val;
1351 			m_copydata(m, 2, m->m_len - 2, (caddr_t)(ipr + 1));
1352 
1353 			switch (optname) {
1354 			case IP_IPSEC_LOCAL_ID:
1355 				/* Check valid types and NUL-termination */
1356 				if (ipr->ref_type < IPSP_IDENTITY_PREFIX ||
1357 				    ipr->ref_type > IPSP_IDENTITY_CONNECTION ||
1358 				    ((char *)(ipr + 1))[ipr->ref_len - 1]) {
1359 					free(ipr, M_CREDENTIALS);
1360 					error = EINVAL;
1361 				} else {
1362 					if (inp->inp_ipo->ipo_srcid != NULL)
1363 						ipsp_reffree(inp->inp_ipo->ipo_srcid);
1364 					inp->inp_ipo->ipo_srcid = ipr;
1365 				}
1366 				break;
1367 			case IP_IPSEC_REMOTE_ID:
1368 				/* Check valid types and NUL-termination */
1369 				if (ipr->ref_type < IPSP_IDENTITY_PREFIX ||
1370 				    ipr->ref_type > IPSP_IDENTITY_CONNECTION ||
1371 				    ((char *)(ipr + 1))[ipr->ref_len - 1]) {
1372 					free(ipr, M_CREDENTIALS);
1373 					error = EINVAL;
1374 				} else {
1375 					if (inp->inp_ipo->ipo_dstid != NULL)
1376 						ipsp_reffree(inp->inp_ipo->ipo_dstid);
1377 					inp->inp_ipo->ipo_dstid = ipr;
1378 				}
1379 				break;
1380 			case IP_IPSEC_LOCAL_CRED:
1381 				if (ipr->ref_type < IPSP_CRED_KEYNOTE ||
1382 				    ipr->ref_type > IPSP_CRED_X509) {
1383 					free(ipr, M_CREDENTIALS);
1384 					error = EINVAL;
1385 				} else {
1386 					if (inp->inp_ipo->ipo_local_cred != NULL)
1387 						ipsp_reffree(inp->inp_ipo->ipo_local_cred);
1388 					inp->inp_ipo->ipo_local_cred = ipr;
1389 				}
1390 				break;
1391 			case IP_IPSEC_LOCAL_AUTH:
1392 				if (ipr->ref_type < IPSP_AUTH_PASSPHRASE ||
1393 				    ipr->ref_type > IPSP_AUTH_RSA) {
1394 					free(ipr, M_CREDENTIALS);
1395 					error = EINVAL;
1396 				} else {
1397 					if (inp->inp_ipo->ipo_local_auth != NULL)
1398 						ipsp_reffree(inp->inp_ipo->ipo_local_auth);
1399 					inp->inp_ipo->ipo_local_auth = ipr;
1400 				}
1401 				break;
1402 			}
1403 
1404 			/* Unlink cached output TDB to force a re-search */
1405 			if (inp->inp_tdb_out) {
1406 				int s = spltdb();
1407 				TAILQ_REMOVE(&inp->inp_tdb_out->tdb_inp_out,
1408 				    inp, inp_tdb_out_next);
1409 				splx(s);
1410 			}
1411 
1412 			if (inp->inp_tdb_in) {
1413 				int s = spltdb();
1414 				TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in,
1415 				    inp, inp_tdb_in_next);
1416 				splx(s);
1417 			}
1418 #endif
1419 			break;
1420 		case SO_RDOMAIN:
1421 			if (m == NULL || m->m_len < sizeof(u_int)) {
1422 				error = EINVAL;
1423 				break;
1424 			}
1425 			rtid = *mtod(m, u_int *);
1426 			/* table must exist and be a domain */
1427 			if (!rtable_exists(rtid) || rtid != rtable_l2(rtid)) {
1428 				error = EINVAL;
1429 				break;
1430 			}
1431 			inp->inp_rdomain = rtid;
1432 			break;
1433 		default:
1434 			error = ENOPROTOOPT;
1435 			break;
1436 		}
1437 		if (m)
1438 			(void)m_free(m);
1439 		break;
1440 
1441 	case PRCO_GETOPT:
1442 		switch (optname) {
1443 		case IP_OPTIONS:
1444 		case IP_RETOPTS:
1445 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1446 			if (inp->inp_options) {
1447 				m->m_len = inp->inp_options->m_len;
1448 				bcopy(mtod(inp->inp_options, caddr_t),
1449 				    mtod(m, caddr_t), (unsigned)m->m_len);
1450 			} else
1451 				m->m_len = 0;
1452 			break;
1453 
1454 		case IP_TOS:
1455 		case IP_TTL:
1456 		case IP_MINTTL:
1457 		case IP_RECVOPTS:
1458 		case IP_RECVRETOPTS:
1459 		case IP_RECVDSTADDR:
1460 		case IP_RECVIF:
1461 		case IP_RECVTTL:
1462 		case IP_RECVDSTPORT:
1463 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1464 			m->m_len = sizeof(int);
1465 			switch (optname) {
1466 
1467 			case IP_TOS:
1468 				optval = inp->inp_ip.ip_tos;
1469 				break;
1470 
1471 			case IP_TTL:
1472 				optval = inp->inp_ip.ip_ttl;
1473 				break;
1474 
1475 			case IP_MINTTL:
1476 				optval = inp->inp_ip_minttl;
1477 				break;
1478 
1479 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1480 
1481 			case IP_RECVOPTS:
1482 				optval = OPTBIT(INP_RECVOPTS);
1483 				break;
1484 
1485 			case IP_RECVRETOPTS:
1486 				optval = OPTBIT(INP_RECVRETOPTS);
1487 				break;
1488 
1489 			case IP_RECVDSTADDR:
1490 				optval = OPTBIT(INP_RECVDSTADDR);
1491 				break;
1492 			case IP_RECVIF:
1493 				optval = OPTBIT(INP_RECVIF);
1494 				break;
1495 			case IP_RECVTTL:
1496 				optval = OPTBIT(INP_RECVTTL);
1497 				break;
1498 			case IP_RECVDSTPORT:
1499 				optval = OPTBIT(INP_RECVDSTPORT);
1500 				break;
1501 			}
1502 			*mtod(m, int *) = optval;
1503 			break;
1504 
1505 		case IP_MULTICAST_IF:
1506 		case IP_MULTICAST_TTL:
1507 		case IP_MULTICAST_LOOP:
1508 		case IP_ADD_MEMBERSHIP:
1509 		case IP_DROP_MEMBERSHIP:
1510 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1511 			break;
1512 
1513 		case IP_PORTRANGE:
1514 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1515 			m->m_len = sizeof(int);
1516 
1517 			if (inp->inp_flags & INP_HIGHPORT)
1518 				optval = IP_PORTRANGE_HIGH;
1519 			else if (inp->inp_flags & INP_LOWPORT)
1520 				optval = IP_PORTRANGE_LOW;
1521 			else
1522 				optval = 0;
1523 
1524 			*mtod(m, int *) = optval;
1525 			break;
1526 
1527 		case IP_AUTH_LEVEL:
1528 		case IP_ESP_TRANS_LEVEL:
1529 		case IP_ESP_NETWORK_LEVEL:
1530 		case IP_IPCOMP_LEVEL:
1531 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1532 #ifndef IPSEC
1533 			m->m_len = sizeof(int);
1534 			*mtod(m, int *) = IPSEC_LEVEL_NONE;
1535 #else
1536 			m->m_len = sizeof(int);
1537 			switch (optname) {
1538 			case IP_AUTH_LEVEL:
1539 				optval = inp->inp_seclevel[SL_AUTH];
1540 				break;
1541 
1542 			case IP_ESP_TRANS_LEVEL:
1543 				optval = inp->inp_seclevel[SL_ESP_TRANS];
1544 				break;
1545 
1546 			case IP_ESP_NETWORK_LEVEL:
1547 				optval = inp->inp_seclevel[SL_ESP_NETWORK];
1548 				break;
1549 			case IP_IPCOMP_LEVEL:
1550 				optval = inp->inp_seclevel[SL_IPCOMP];
1551 				break;
1552 			}
1553 			*mtod(m, int *) = optval;
1554 #endif
1555 			break;
1556 		case IP_IPSEC_LOCAL_ID:
1557 		case IP_IPSEC_REMOTE_ID:
1558 		case IP_IPSEC_LOCAL_CRED:
1559 		case IP_IPSEC_REMOTE_CRED:
1560 		case IP_IPSEC_LOCAL_AUTH:
1561 		case IP_IPSEC_REMOTE_AUTH:
1562 #ifndef IPSEC
1563 			error = EOPNOTSUPP;
1564 #else
1565 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1566 			m->m_len = sizeof(u_int16_t);
1567 			ipr = NULL;
1568 			switch (optname) {
1569 			case IP_IPSEC_LOCAL_ID:
1570 				if (inp->inp_ipo != NULL)
1571 					ipr = inp->inp_ipo->ipo_srcid;
1572 				opt16val = IPSP_IDENTITY_NONE;
1573 				break;
1574 			case IP_IPSEC_REMOTE_ID:
1575 				if (inp->inp_ipo != NULL)
1576 					ipr = inp->inp_ipo->ipo_dstid;
1577 				opt16val = IPSP_IDENTITY_NONE;
1578 				break;
1579 			case IP_IPSEC_LOCAL_CRED:
1580 				if (inp->inp_ipo != NULL)
1581 					ipr = inp->inp_ipo->ipo_local_cred;
1582 				opt16val = IPSP_CRED_NONE;
1583 				break;
1584 			case IP_IPSEC_REMOTE_CRED:
1585 				ipr = inp->inp_ipsec_remotecred;
1586 				opt16val = IPSP_CRED_NONE;
1587 				break;
1588 			case IP_IPSEC_LOCAL_AUTH:
1589 				if (inp->inp_ipo != NULL)
1590 					ipr = inp->inp_ipo->ipo_local_auth;
1591 				break;
1592 			case IP_IPSEC_REMOTE_AUTH:
1593 				ipr = inp->inp_ipsec_remoteauth;
1594 				break;
1595 			}
1596 			if (ipr == NULL)
1597 				*mtod(m, u_int16_t *) = opt16val;
1598 			else {
1599 				size_t len;
1600 
1601 				len = m->m_len + ipr->ref_len;
1602 				if (len > MCLBYTES) {
1603 					 m_free(m);
1604 					 error = EINVAL;
1605 					 break;
1606 				}
1607 				/* allocate mbuf cluster for larger option */
1608 				if (len > MLEN) {
1609 					 MCLGET(m, M_WAITOK);
1610 					 if ((m->m_flags & M_EXT) == 0) {
1611 						 m_free(m);
1612 						 error = ENOBUFS;
1613 						 break;
1614 					 }
1615 
1616 				}
1617 				m->m_len = len;
1618 				*mtod(m, u_int16_t *) = ipr->ref_type;
1619 				m_copyback(m, sizeof(u_int16_t), ipr->ref_len,
1620 				    ipr + 1);
1621 			}
1622 #endif
1623 			break;
1624 		case SO_RDOMAIN:
1625 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1626 			m->m_len = sizeof(u_int);
1627 			*mtod(m, u_int *) = inp->inp_rdomain;
1628 			break;
1629 		default:
1630 			error = ENOPROTOOPT;
1631 			break;
1632 		}
1633 		break;
1634 	}
1635 	return (error);
1636 }
1637 
1638 /*
1639  * Set up IP options in pcb for insertion in output packets.
1640  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1641  * with destination address if source routed.
1642  */
1643 int
1644 #ifdef notyet
1645 ip_pcbopts(optname, pcbopt, m)
1646 	int optname;
1647 #else
1648 ip_pcbopts(pcbopt, m)
1649 #endif
1650 	struct mbuf **pcbopt;
1651 	struct mbuf *m;
1652 {
1653 	int cnt, optlen;
1654 	u_char *cp;
1655 	u_char opt;
1656 
1657 	/* turn off any old options */
1658 	if (*pcbopt)
1659 		(void)m_free(*pcbopt);
1660 	*pcbopt = 0;
1661 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1662 		/*
1663 		 * Only turning off any previous options.
1664 		 */
1665 		if (m)
1666 			(void)m_free(m);
1667 		return (0);
1668 	}
1669 
1670 	if (m->m_len % sizeof(int32_t))
1671 		goto bad;
1672 
1673 	/*
1674 	 * IP first-hop destination address will be stored before
1675 	 * actual options; move other options back
1676 	 * and clear it when none present.
1677 	 */
1678 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1679 		goto bad;
1680 	cnt = m->m_len;
1681 	m->m_len += sizeof(struct in_addr);
1682 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1683 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1684 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1685 
1686 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1687 		opt = cp[IPOPT_OPTVAL];
1688 		if (opt == IPOPT_EOL)
1689 			break;
1690 		if (opt == IPOPT_NOP)
1691 			optlen = 1;
1692 		else {
1693 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1694 				goto bad;
1695 			optlen = cp[IPOPT_OLEN];
1696 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1697 				goto bad;
1698 		}
1699 		switch (opt) {
1700 
1701 		default:
1702 			break;
1703 
1704 		case IPOPT_LSRR:
1705 		case IPOPT_SSRR:
1706 			/*
1707 			 * user process specifies route as:
1708 			 *	->A->B->C->D
1709 			 * D must be our final destination (but we can't
1710 			 * check that since we may not have connected yet).
1711 			 * A is first hop destination, which doesn't appear in
1712 			 * actual IP option, but is stored before the options.
1713 			 */
1714 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1715 				goto bad;
1716 			m->m_len -= sizeof(struct in_addr);
1717 			cnt -= sizeof(struct in_addr);
1718 			optlen -= sizeof(struct in_addr);
1719 			cp[IPOPT_OLEN] = optlen;
1720 			/*
1721 			 * Move first hop before start of options.
1722 			 */
1723 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1724 			    sizeof(struct in_addr));
1725 			/*
1726 			 * Then copy rest of options back
1727 			 * to close up the deleted entry.
1728 			 */
1729 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1730 			    sizeof(struct in_addr)),
1731 			    (caddr_t)&cp[IPOPT_OFFSET+1],
1732 			    (unsigned)cnt - (IPOPT_OFFSET+1));
1733 			break;
1734 		}
1735 	}
1736 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1737 		goto bad;
1738 	*pcbopt = m;
1739 	return (0);
1740 
1741 bad:
1742 	(void)m_free(m);
1743 	return (EINVAL);
1744 }
1745 
1746 /*
1747  * Set the IP multicast options in response to user setsockopt().
1748  */
1749 int
1750 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m,
1751     u_int rdomain)
1752 {
1753 	int error = 0;
1754 	u_char loop;
1755 	int i;
1756 	struct in_addr addr;
1757 	struct ip_mreq *mreq;
1758 	struct ifnet *ifp;
1759 	struct ip_moptions *imo = *imop;
1760 	struct in_multi **immp;
1761 	struct route ro;
1762 	struct sockaddr_in *dst;
1763 
1764 	if (imo == NULL) {
1765 		/*
1766 		 * No multicast option buffer attached to the pcb;
1767 		 * allocate one and initialize to default values.
1768 		 */
1769 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1770 		    M_WAITOK|M_ZERO);
1771 		immp = (struct in_multi **)malloc(
1772 		    (sizeof(*immp) * IP_MIN_MEMBERSHIPS), M_IPMOPTS,
1773 		    M_WAITOK|M_ZERO);
1774 		*imop = imo;
1775 		imo->imo_multicast_ifp = NULL;
1776 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1777 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1778 		imo->imo_num_memberships = 0;
1779 		imo->imo_max_memberships = IP_MIN_MEMBERSHIPS;
1780 		imo->imo_membership = immp;
1781 	}
1782 
1783 	switch (optname) {
1784 
1785 	case IP_MULTICAST_IF:
1786 		/*
1787 		 * Select the interface for outgoing multicast packets.
1788 		 */
1789 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1790 			error = EINVAL;
1791 			break;
1792 		}
1793 		addr = *(mtod(m, struct in_addr *));
1794 		/*
1795 		 * INADDR_ANY is used to remove a previous selection.
1796 		 * When no interface is selected, a default one is
1797 		 * chosen every time a multicast packet is sent.
1798 		 */
1799 		if (addr.s_addr == INADDR_ANY) {
1800 			imo->imo_multicast_ifp = NULL;
1801 			break;
1802 		}
1803 		/*
1804 		 * The selected interface is identified by its local
1805 		 * IP address.  Find the interface and confirm that
1806 		 * it supports multicasting.
1807 		 */
1808 		INADDR_TO_IFP(addr, ifp, rdomain);
1809 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1810 			error = EADDRNOTAVAIL;
1811 			break;
1812 		}
1813 		imo->imo_multicast_ifp = ifp;
1814 		break;
1815 
1816 	case IP_MULTICAST_TTL:
1817 		/*
1818 		 * Set the IP time-to-live for outgoing multicast packets.
1819 		 */
1820 		if (m == NULL || m->m_len != 1) {
1821 			error = EINVAL;
1822 			break;
1823 		}
1824 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1825 		break;
1826 
1827 	case IP_MULTICAST_LOOP:
1828 		/*
1829 		 * Set the loopback flag for outgoing multicast packets.
1830 		 * Must be zero or one.
1831 		 */
1832 		if (m == NULL || m->m_len != 1 ||
1833 		   (loop = *(mtod(m, u_char *))) > 1) {
1834 			error = EINVAL;
1835 			break;
1836 		}
1837 		imo->imo_multicast_loop = loop;
1838 		break;
1839 
1840 	case IP_ADD_MEMBERSHIP:
1841 		/*
1842 		 * Add a multicast group membership.
1843 		 * Group must be a valid IP multicast address.
1844 		 */
1845 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1846 			error = EINVAL;
1847 			break;
1848 		}
1849 		mreq = mtod(m, struct ip_mreq *);
1850 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1851 			error = EINVAL;
1852 			break;
1853 		}
1854 		/*
1855 		 * If no interface address was provided, use the interface of
1856 		 * the route to the given multicast address.
1857 		 */
1858 		if (mreq->imr_interface.s_addr == INADDR_ANY) {
1859 			ro.ro_rt = NULL;
1860 			dst = satosin(&ro.ro_dst);
1861 			dst->sin_len = sizeof(*dst);
1862 			dst->sin_family = AF_INET;
1863 			dst->sin_addr = mreq->imr_multiaddr;
1864 			if (!(ro.ro_rt && ro.ro_rt->rt_ifp &&
1865 			    (ro.ro_rt->rt_flags & RTF_UP)))
1866 				ro.ro_rt = rtalloc1(&ro.ro_dst, 1, rdomain);
1867 			if (ro.ro_rt == NULL) {
1868 				error = EADDRNOTAVAIL;
1869 				break;
1870 			}
1871 			ifp = ro.ro_rt->rt_ifp;
1872 			rtfree(ro.ro_rt);
1873 		} else {
1874 			INADDR_TO_IFP(mreq->imr_interface, ifp, rdomain);
1875 		}
1876 		/*
1877 		 * See if we found an interface, and confirm that it
1878 		 * supports multicast.
1879 		 */
1880 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1881 			error = EADDRNOTAVAIL;
1882 			break;
1883 		}
1884 		/*
1885 		 * See if the membership already exists or if all the
1886 		 * membership slots are full.
1887 		 */
1888 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1889 			if (imo->imo_membership[i]->inm_ia->ia_ifp == ifp &&
1890 			    imo->imo_membership[i]->inm_addr.s_addr
1891 						== mreq->imr_multiaddr.s_addr)
1892 				break;
1893 		}
1894 		if (i < imo->imo_num_memberships) {
1895 			error = EADDRINUSE;
1896 			break;
1897 		}
1898 		if (imo->imo_num_memberships == imo->imo_max_memberships) {
1899 			struct in_multi **nmships, **omships;
1900 			size_t newmax;
1901 			/*
1902 			 * Resize the vector to next power-of-two minus 1. If the
1903 			 * size would exceed the maximum then we know we've really
1904 			 * run out of entries. Otherwise, we reallocate the vector.
1905 			 */
1906 			nmships = NULL;
1907 			omships = imo->imo_membership;
1908 			newmax = ((imo->imo_max_memberships + 1) * 2) - 1;
1909 			if (newmax <= IP_MAX_MEMBERSHIPS) {
1910 				nmships = (struct in_multi **)malloc(
1911 				    sizeof(*nmships) * newmax, M_IPMOPTS,
1912 				    M_NOWAIT|M_ZERO);
1913 				if (nmships != NULL) {
1914 					bcopy(omships, nmships,
1915 					    sizeof(*omships) *
1916 					    imo->imo_max_memberships);
1917 					free(omships, M_IPMOPTS);
1918 					imo->imo_membership = nmships;
1919 					imo->imo_max_memberships = newmax;
1920 				}
1921 			}
1922 			if (nmships == NULL) {
1923 				error = ETOOMANYREFS;
1924 				break;
1925 			}
1926 		}
1927 		/*
1928 		 * Everything looks good; add a new record to the multicast
1929 		 * address list for the given interface.
1930 		 */
1931 		if ((imo->imo_membership[i] =
1932 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1933 			error = ENOBUFS;
1934 			break;
1935 		}
1936 		++imo->imo_num_memberships;
1937 		break;
1938 
1939 	case IP_DROP_MEMBERSHIP:
1940 		/*
1941 		 * Drop a multicast group membership.
1942 		 * Group must be a valid IP multicast address.
1943 		 */
1944 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1945 			error = EINVAL;
1946 			break;
1947 		}
1948 		mreq = mtod(m, struct ip_mreq *);
1949 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1950 			error = EINVAL;
1951 			break;
1952 		}
1953 		/*
1954 		 * If an interface address was specified, get a pointer
1955 		 * to its ifnet structure.
1956 		 */
1957 		if (mreq->imr_interface.s_addr == INADDR_ANY)
1958 			ifp = NULL;
1959 		else {
1960 			INADDR_TO_IFP(mreq->imr_interface, ifp, rdomain);
1961 			if (ifp == NULL) {
1962 				error = EADDRNOTAVAIL;
1963 				break;
1964 			}
1965 		}
1966 		/*
1967 		 * Find the membership in the membership array.
1968 		 */
1969 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1970 			if ((ifp == NULL ||
1971 			     imo->imo_membership[i]->inm_ia->ia_ifp == ifp) &&
1972 			     imo->imo_membership[i]->inm_addr.s_addr ==
1973 			     mreq->imr_multiaddr.s_addr)
1974 				break;
1975 		}
1976 		if (i == imo->imo_num_memberships) {
1977 			error = EADDRNOTAVAIL;
1978 			break;
1979 		}
1980 		/*
1981 		 * Give up the multicast address record to which the
1982 		 * membership points.
1983 		 */
1984 		in_delmulti(imo->imo_membership[i]);
1985 		/*
1986 		 * Remove the gap in the membership array.
1987 		 */
1988 		for (++i; i < imo->imo_num_memberships; ++i)
1989 			imo->imo_membership[i-1] = imo->imo_membership[i];
1990 		--imo->imo_num_memberships;
1991 		break;
1992 
1993 	default:
1994 		error = EOPNOTSUPP;
1995 		break;
1996 	}
1997 
1998 	/*
1999 	 * If all options have default values, no need to keep the data.
2000 	 */
2001 	if (imo->imo_multicast_ifp == NULL &&
2002 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2003 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2004 	    imo->imo_num_memberships == 0) {
2005 		free(imo->imo_membership , M_IPMOPTS);
2006 		free(*imop, M_IPMOPTS);
2007 		*imop = NULL;
2008 	}
2009 
2010 	return (error);
2011 }
2012 
2013 /*
2014  * Return the IP multicast options in response to user getsockopt().
2015  */
2016 int
2017 ip_getmoptions(optname, imo, mp)
2018 	int optname;
2019 	struct ip_moptions *imo;
2020 	struct mbuf **mp;
2021 {
2022 	u_char *ttl;
2023 	u_char *loop;
2024 	struct in_addr *addr;
2025 	struct in_ifaddr *ia;
2026 
2027 	*mp = m_get(M_WAIT, MT_SOOPTS);
2028 
2029 	switch (optname) {
2030 
2031 	case IP_MULTICAST_IF:
2032 		addr = mtod(*mp, struct in_addr *);
2033 		(*mp)->m_len = sizeof(struct in_addr);
2034 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
2035 			addr->s_addr = INADDR_ANY;
2036 		else {
2037 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
2038 			addr->s_addr = (ia == NULL) ? INADDR_ANY
2039 					: ia->ia_addr.sin_addr.s_addr;
2040 		}
2041 		return (0);
2042 
2043 	case IP_MULTICAST_TTL:
2044 		ttl = mtod(*mp, u_char *);
2045 		(*mp)->m_len = 1;
2046 		*ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
2047 				     : imo->imo_multicast_ttl;
2048 		return (0);
2049 
2050 	case IP_MULTICAST_LOOP:
2051 		loop = mtod(*mp, u_char *);
2052 		(*mp)->m_len = 1;
2053 		*loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
2054 				      : imo->imo_multicast_loop;
2055 		return (0);
2056 
2057 	default:
2058 		return (EOPNOTSUPP);
2059 	}
2060 }
2061 
2062 /*
2063  * Discard the IP multicast options.
2064  */
2065 void
2066 ip_freemoptions(imo)
2067 	struct ip_moptions *imo;
2068 {
2069 	int i;
2070 
2071 	if (imo != NULL) {
2072 		for (i = 0; i < imo->imo_num_memberships; ++i)
2073 			in_delmulti(imo->imo_membership[i]);
2074 		free(imo->imo_membership, M_IPMOPTS);
2075 		free(imo, M_IPMOPTS);
2076 	}
2077 }
2078 
2079 /*
2080  * Routine called from ip_output() to loop back a copy of an IP multicast
2081  * packet to the input queue of a specified interface.  Note that this
2082  * calls the output routine of the loopback "driver", but with an interface
2083  * pointer that might NOT be &loif -- easier than replicating that code here.
2084  */
2085 void
2086 ip_mloopback(ifp, m, dst)
2087 	struct ifnet *ifp;
2088 	struct mbuf *m;
2089 	struct sockaddr_in *dst;
2090 {
2091 	struct ip *ip;
2092 	struct mbuf *copym;
2093 
2094 	copym = m_copym2(m, 0, M_COPYALL, M_DONTWAIT);
2095 	if (copym != NULL) {
2096 		/*
2097 		 * We don't bother to fragment if the IP length is greater
2098 		 * than the interface's MTU.  Can this possibly matter?
2099 		 */
2100 		ip = mtod(copym, struct ip *);
2101 		ip->ip_sum = 0;
2102 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
2103 		(void) looutput(ifp, copym, sintosa(dst), NULL);
2104 	}
2105 }
2106 
2107 /*
2108  * Process a delayed payload checksum calculation.
2109  */
2110 void
2111 in_delayed_cksum(struct mbuf *m)
2112 {
2113 	struct ip *ip;
2114 	u_int16_t csum, offset;
2115 
2116 	ip = mtod(m, struct ip *);
2117 	offset = ip->ip_hl << 2;
2118 	csum = in4_cksum(m, 0, offset, m->m_pkthdr.len - offset);
2119 	if (csum == 0 && ip->ip_p == IPPROTO_UDP)
2120 		csum = 0xffff;
2121 
2122 	switch (ip->ip_p) {
2123 	case IPPROTO_TCP:
2124 		offset += offsetof(struct tcphdr, th_sum);
2125 		break;
2126 
2127 	case IPPROTO_UDP:
2128 		offset += offsetof(struct udphdr, uh_sum);
2129 		break;
2130 
2131 	default:
2132 		return;
2133 	}
2134 
2135 	if ((offset + sizeof(u_int16_t)) > m->m_len)
2136 		m_copyback(m, offset, sizeof(csum), &csum);
2137 	else
2138 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2139 }
2140