xref: /openbsd-src/sys/netinet/ip_output.c (revision 850e275390052b330d93020bf619a739a3c277ac)
1 /*	$OpenBSD: ip_output.c,v 1.191 2008/05/09 02:56:36 markus 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)))) == 0 &&
205 			    (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
206 			    ipstat.ips_noroute++;
207 			    error = ENETUNREACH;
208 			    goto bad;
209 			}
210 
211 			ifp = ia->ia_ifp;
212 			mtu = ifp->if_mtu;
213 			ip->ip_ttl = 1;
214 		} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
215 		    (ip->ip_dst.s_addr == INADDR_BROADCAST)) &&
216 		    imo != NULL && imo->imo_multicast_ifp != NULL) {
217 			ifp = imo->imo_multicast_ifp;
218 			mtu = ifp->if_mtu;
219 			IFP_TO_IA(ifp, ia);
220 		} else {
221 			if (ro->ro_rt == 0)
222 				rtalloc_mpath(ro, NULL, 0);
223 
224 			if (ro->ro_rt == 0) {
225 				ipstat.ips_noroute++;
226 				error = EHOSTUNREACH;
227 				goto bad;
228 			}
229 
230 			ia = ifatoia(ro->ro_rt->rt_ifa);
231 			ifp = ro->ro_rt->rt_ifp;
232 			if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
233 				mtu = ifp->if_mtu;
234 			ro->ro_rt->rt_use++;
235 
236 			if (ro->ro_rt->rt_flags & RTF_GATEWAY)
237 				dst = satosin(ro->ro_rt->rt_gateway);
238 		}
239 
240 		/* Set the source IP address */
241 		if (!IN_MULTICAST(ip->ip_dst.s_addr))
242 			ip->ip_src = ia->ia_addr.sin_addr;
243 	}
244 
245 #ifdef IPSEC
246 	if (!ipsec_in_use && inp == NULL)
247 		goto done_spd;
248 
249 	/*
250 	 * splnet is chosen over spltdb because we are not allowed to
251 	 * lower the level, and udp_output calls us in splnet().
252 	 */
253 	s = splnet();
254 
255 	/* Do we have any pending SAs to apply ? */
256 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
257 	if (mtag != NULL) {
258 #ifdef DIAGNOSTIC
259 		if (mtag->m_tag_len != sizeof (struct tdb_ident))
260 			panic("ip_output: tag of length %d (should be %d",
261 			    mtag->m_tag_len, sizeof (struct tdb_ident));
262 #endif
263 		tdbi = (struct tdb_ident *)(mtag + 1);
264 		tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
265 		if (tdb == NULL)
266 			error = -EINVAL;
267 		m_tag_delete(m, mtag);
268 	}
269 	else
270 		tdb = ipsp_spd_lookup(m, AF_INET, hlen, &error,
271 		    IPSP_DIRECTION_OUT, NULL, inp);
272 
273 	if (tdb == NULL) {
274 		splx(s);
275 
276 		if (error == 0) {
277 			/*
278 			 * No IPsec processing required, we'll just send the
279 			 * packet out.
280 			 */
281 			sproto = 0;
282 
283 			/* Fall through to routing/multicast handling */
284 		} else {
285 			/*
286 			 * -EINVAL is used to indicate that the packet should
287 			 * be silently dropped, typically because we've asked
288 			 * key management for an SA.
289 			 */
290 			if (error == -EINVAL) /* Should silently drop packet */
291 			  error = 0;
292 
293 			m_freem(m);
294 			goto done;
295 		}
296 	} else {
297 		/* Loop detection */
298 		for (mtag = m_tag_first(m); mtag != NULL;
299 		    mtag = m_tag_next(m, mtag)) {
300 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
301 			    mtag->m_tag_id !=
302 			    PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
303 				continue;
304 			tdbi = (struct tdb_ident *)(mtag + 1);
305 			if (tdbi->spi == tdb->tdb_spi &&
306 			    tdbi->proto == tdb->tdb_sproto &&
307 			    !bcmp(&tdbi->dst, &tdb->tdb_dst,
308 			    sizeof(union sockaddr_union))) {
309 				splx(s);
310 				sproto = 0; /* mark as no-IPsec-needed */
311 				goto done_spd;
312 			}
313 		}
314 
315 		/* We need to do IPsec */
316 		bcopy(&tdb->tdb_dst, &sdst, sizeof(sdst));
317 		sspi = tdb->tdb_spi;
318 		sproto = tdb->tdb_sproto;
319 		splx(s);
320 
321 		/*
322 		 * If it needs TCP/UDP hardware-checksumming, do the
323 		 * computation now.
324 		 */
325 		if (m->m_pkthdr.csum_flags & (M_TCPV4_CSUM_OUT | M_UDPV4_CSUM_OUT)) {
326 			in_delayed_cksum(m);
327 			m->m_pkthdr.csum_flags &=
328 			    ~(M_UDPV4_CSUM_OUT | M_TCPV4_CSUM_OUT);
329 		}
330 
331 		/* If it's not a multicast packet, try to fast-path */
332 		if (!IN_MULTICAST(ip->ip_dst.s_addr)) {
333 			goto sendit;
334 		}
335 	}
336 
337 	/* Fall through to the routing/multicast handling code */
338  done_spd:
339 #endif /* IPSEC */
340 
341 	if (flags & IP_ROUTETOETHER) {
342 		dst = satosin(&ro->ro_dst);
343 		ifp = ro->ro_rt->rt_ifp;
344 		mtu = ifp->if_mtu;
345 		ro->ro_rt = NULL;
346 	} else if (donerouting == 0) {
347 		if (ro == 0) {
348 			ro = &iproute;
349 			bzero((caddr_t)ro, sizeof (*ro));
350 		}
351 
352 		dst = satosin(&ro->ro_dst);
353 
354 		/*
355 		 * If there is a cached route, check that it is to the same
356 		 * destination and is still up.  If not, free it and try again.
357 		 */
358 		if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
359 				  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
360 			RTFREE(ro->ro_rt);
361 			ro->ro_rt = (struct rtentry *)0;
362 		}
363 
364 		if (ro->ro_rt == 0) {
365 			dst->sin_family = AF_INET;
366 			dst->sin_len = sizeof(*dst);
367 			dst->sin_addr = ip->ip_dst;
368 		}
369 
370 		/*
371 		 * If routing to interface only, short-circuit routing lookup.
372 		 */
373 		if (flags & IP_ROUTETOIF) {
374 			if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
375 			    (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
376 			    ipstat.ips_noroute++;
377 			    error = ENETUNREACH;
378 			    goto bad;
379 			}
380 
381 			ifp = ia->ia_ifp;
382 			mtu = ifp->if_mtu;
383 			ip->ip_ttl = 1;
384 		} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
385 		    (ip->ip_dst.s_addr == INADDR_BROADCAST)) &&
386 		    imo != NULL && imo->imo_multicast_ifp != NULL) {
387 			ifp = imo->imo_multicast_ifp;
388 			mtu = ifp->if_mtu;
389 			IFP_TO_IA(ifp, ia);
390 		} else {
391 			if (ro->ro_rt == 0)
392 				rtalloc_mpath(ro, &ip->ip_src.s_addr, 0);
393 
394 			if (ro->ro_rt == 0) {
395 				ipstat.ips_noroute++;
396 				error = EHOSTUNREACH;
397 				goto bad;
398 			}
399 
400 			ia = ifatoia(ro->ro_rt->rt_ifa);
401 			ifp = ro->ro_rt->rt_ifp;
402 			if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
403 				mtu = ifp->if_mtu;
404 			ro->ro_rt->rt_use++;
405 
406 			if (ro->ro_rt->rt_flags & RTF_GATEWAY)
407 				dst = satosin(ro->ro_rt->rt_gateway);
408 		}
409 
410 		/* Set the source IP address */
411 		if (ip->ip_src.s_addr == INADDR_ANY)
412 			ip->ip_src = ia->ia_addr.sin_addr;
413 	}
414 
415 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
416 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
417 		struct in_multi *inm;
418 
419 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
420 			M_BCAST : M_MCAST;
421 
422 		/*
423 		 * IP destination address is multicast.  Make sure "dst"
424 		 * still points to the address in "ro".  (It may have been
425 		 * changed to point to a gateway address, above.)
426 		 */
427 		dst = satosin(&ro->ro_dst);
428 
429 		/*
430 		 * See if the caller provided any multicast options
431 		 */
432 		if (imo != NULL)
433 			ip->ip_ttl = imo->imo_multicast_ttl;
434 		else
435 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
436 
437 		/*
438 		 * if we don't know the outgoing ifp yet, we can't generate
439 		 * output
440 		 */
441 		if (!ifp) {
442 			ipstat.ips_noroute++;
443 			error = EHOSTUNREACH;
444 			goto bad;
445 		}
446 
447 		/*
448 		 * Confirm that the outgoing interface supports multicast,
449 		 * but only if the packet actually is going out on that
450 		 * interface (i.e., no IPsec is applied).
451 		 */
452 		if ((((m->m_flags & M_MCAST) &&
453 		      (ifp->if_flags & IFF_MULTICAST) == 0) ||
454 		     ((m->m_flags & M_BCAST) &&
455 		      (ifp->if_flags & IFF_BROADCAST) == 0)) && (sproto == 0)) {
456 			ipstat.ips_noroute++;
457 			error = ENETUNREACH;
458 			goto bad;
459 		}
460 
461 		/*
462 		 * If source address not specified yet, use address
463 		 * of outgoing interface.
464 		 */
465 		if (ip->ip_src.s_addr == INADDR_ANY) {
466 			struct in_ifaddr *ia;
467 
468 			TAILQ_FOREACH(ia, &in_ifaddr, ia_list)
469 				if (ia->ia_ifp == ifp) {
470 					ip->ip_src = ia->ia_addr.sin_addr;
471 					break;
472 				}
473 		}
474 
475 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
476 		if (inm != NULL &&
477 		   (imo == NULL || imo->imo_multicast_loop)) {
478 			/*
479 			 * If we belong to the destination multicast group
480 			 * on the outgoing interface, and the caller did not
481 			 * forbid loopback, loop back a copy.
482 			 * Can't defer TCP/UDP checksumming, do the
483 			 * computation now.
484 			 */
485 			if (m->m_pkthdr.csum_flags &
486 			    (M_TCPV4_CSUM_OUT | M_UDPV4_CSUM_OUT)) {
487 				in_delayed_cksum(m);
488 				m->m_pkthdr.csum_flags &=
489 				    ~(M_UDPV4_CSUM_OUT | M_TCPV4_CSUM_OUT);
490 			}
491 			ip_mloopback(ifp, m, dst);
492 		}
493 #ifdef MROUTING
494 		else {
495 			/*
496 			 * If we are acting as a multicast router, perform
497 			 * multicast forwarding as if the packet had just
498 			 * arrived on the interface to which we are about
499 			 * to send.  The multicast forwarding function
500 			 * recursively calls this function, using the
501 			 * IP_FORWARDING flag to prevent infinite recursion.
502 			 *
503 			 * Multicasts that are looped back by ip_mloopback(),
504 			 * above, will be forwarded by the ip_input() routine,
505 			 * if necessary.
506 			 */
507 			extern struct socket *ip_mrouter;
508 
509 			if (ipmforwarding && ip_mrouter &&
510 			    (flags & IP_FORWARDING) == 0) {
511 				if (ip_mforward(m, ifp) != 0) {
512 					m_freem(m);
513 					goto done;
514 				}
515 			}
516 		}
517 #endif
518 		/*
519 		 * Multicasts with a time-to-live of zero may be looped-
520 		 * back, above, but must not be transmitted on a network.
521 		 * Also, multicasts addressed to the loopback interface
522 		 * are not sent -- the above call to ip_mloopback() will
523 		 * loop back a copy if this host actually belongs to the
524 		 * destination group on the loopback interface.
525 		 */
526 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
527 			m_freem(m);
528 			goto done;
529 		}
530 
531 		goto sendit;
532 	}
533 
534 	/*
535 	 * Look for broadcast address and and verify user is allowed to send
536 	 * such a packet; if the packet is going in an IPsec tunnel, skip
537 	 * this check.
538 	 */
539 	if ((sproto == 0) && (in_broadcast(dst->sin_addr, ifp))) {
540 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
541 			error = EADDRNOTAVAIL;
542 			goto bad;
543 		}
544 		if ((flags & IP_ALLOWBROADCAST) == 0) {
545 			error = EACCES;
546 			goto bad;
547 		}
548 
549 		/* Don't allow broadcast messages to be fragmented */
550 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
551 			error = EMSGSIZE;
552 			goto bad;
553 		}
554 		m->m_flags |= M_BCAST;
555 	} else
556 		m->m_flags &= ~M_BCAST;
557 
558 sendit:
559 	/*
560 	 * If we're doing Path MTU discovery, we need to set DF unless
561 	 * the route's MTU is locked.
562 	 */
563 	if ((flags & IP_MTUDISC) && ro && ro->ro_rt &&
564 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
565 		ip->ip_off |= htons(IP_DF);
566 
567 #ifdef IPSEC
568 	/*
569 	 * Check if the packet needs encapsulation.
570 	 */
571 	if (sproto != 0) {
572 		s = splnet();
573 
574 		/*
575 		 * Packet filter
576 		 */
577 #if NPF > 0
578 
579 		if (pf_test(PF_OUT, &encif[0].sc_if, &m, NULL) != PF_PASS) {
580 			error = EHOSTUNREACH;
581 			splx(s);
582 			m_freem(m);
583 			goto done;
584 		}
585 		if (m == NULL) {
586 			splx(s);
587 			goto done;
588 		}
589 		ip = mtod(m, struct ip *);
590 		hlen = ip->ip_hl << 2;
591 #endif
592 
593 		tdb = gettdb(sspi, &sdst, sproto);
594 		if (tdb == NULL) {
595 			DPRINTF(("ip_output: unknown TDB"));
596 			error = EHOSTUNREACH;
597 			splx(s);
598 			m_freem(m);
599 			goto done;
600 		}
601 
602 		/* Check if we are allowed to fragment */
603 		if (ip_mtudisc && (ip->ip_off & htons(IP_DF)) && tdb->tdb_mtu &&
604 		    ntohs(ip->ip_len) > tdb->tdb_mtu &&
605 		    tdb->tdb_mtutimeout > time_second) {
606 			struct rtentry *rt = NULL;
607 			int rt_mtucloned = 0;
608 			int transportmode = 0;
609 
610 			transportmode = (tdb->tdb_dst.sa.sa_family == AF_INET) &&
611 			    (tdb->tdb_dst.sin.sin_addr.s_addr ==
612 			    ip->ip_dst.s_addr);
613 			icmp_mtu = tdb->tdb_mtu;
614 			splx(s);
615 
616 			/* Find a host route to store the mtu in */
617 			if (ro != NULL)
618 				rt = ro->ro_rt;
619 			/* but don't add a PMTU route for transport mode SAs */
620 			if (transportmode)
621 				rt = NULL;
622 			else if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0) {
623 				struct sockaddr_in dst = {
624 					sizeof(struct sockaddr_in), AF_INET};
625 				dst.sin_addr = ip->ip_dst;
626 				rt = icmp_mtudisc_clone((struct sockaddr *)&dst);
627 				rt_mtucloned = 1;
628 			}
629 			DPRINTF(("ip_output: spi %08x mtu %d rt %p cloned %d\n",
630 			    ntohl(tdb->tdb_spi), icmp_mtu, rt, rt_mtucloned));
631 			if (rt != NULL) {
632 				rt->rt_rmx.rmx_mtu = icmp_mtu;
633 				if (ro && ro->ro_rt != NULL) {
634 					RTFREE(ro->ro_rt);
635 					ro->ro_rt = (struct rtentry *) 0;
636 					rtalloc(ro);
637 				}
638 				if (rt_mtucloned)
639 					rtfree(rt);
640 			}
641 			error = EMSGSIZE;
642 			goto bad;
643 		}
644 
645 		/*
646 		 * Clear these -- they'll be set in the recursive invocation
647 		 * as needed.
648 		 */
649 		m->m_flags &= ~(M_MCAST | M_BCAST);
650 
651 		/* Callee frees mbuf */
652 		error = ipsp_process_packet(m, tdb, AF_INET, 0);
653 		splx(s);
654 		return error;  /* Nothing more to be done */
655 	}
656 
657 	/*
658 	 * If deferred crypto processing is needed, check that the
659 	 * interface supports it.
660 	 */
661 	if (ipsec_in_use && (mtag = m_tag_find(m,
662 	    PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL)) != NULL &&
663 	    (ifp->if_capabilities & IFCAP_IPSEC) == 0) {
664 		/* Notify IPsec to do its own crypto. */
665 		ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
666 		m_freem(m);
667 		error = EHOSTUNREACH;
668 		goto done;
669 	}
670 #endif /* IPSEC */
671 
672 	/* Catch routing changes wrt. hardware checksumming for TCP or UDP. */
673 	if (m->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) {
674 		if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) ||
675 		    ifp->if_bridge != NULL) {
676 			in_delayed_cksum(m);
677 			m->m_pkthdr.csum_flags &= ~M_TCPV4_CSUM_OUT; /* Clear */
678 		}
679 	} else if (m->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) {
680 		if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) ||
681 		    ifp->if_bridge != NULL) {
682 			in_delayed_cksum(m);
683 			m->m_pkthdr.csum_flags &= ~M_UDPV4_CSUM_OUT; /* Clear */
684 		}
685 	}
686 
687 	/*
688 	 * Packet filter
689 	 */
690 #if NPF > 0
691 	if (pf_test(PF_OUT, ifp, &m, NULL) != PF_PASS) {
692 		error = EHOSTUNREACH;
693 		m_freem(m);
694 		goto done;
695 	}
696 	if (m == NULL)
697 		goto done;
698 
699 	ip = mtod(m, struct ip *);
700 	hlen = ip->ip_hl << 2;
701 #endif
702 
703 #ifdef IPSEC
704 	if (ipsec_in_use && (flags & IP_FORWARDING) && (ipforwarding == 2) &&
705 	    (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) == NULL)) {
706 		error = EHOSTUNREACH;
707 		m_freem(m);
708 		goto done;
709 	}
710 #endif
711 
712 	/* XXX
713 	 * Try to use jumbograms based on socket option, or the route
714 	 * or... for other reasons later on.
715 	 */
716 	if ((flags & IP_JUMBO) && ro->ro_rt && (ro->ro_rt->rt_flags & RTF_JUMBO) &&
717 	    ro->ro_rt->rt_ifp)
718 		mtu = ro->ro_rt->rt_ifp->if_hardmtu;
719 
720 	/*
721 	 * If small enough for interface, can just send directly.
722 	 */
723 	if (ntohs(ip->ip_len) <= mtu) {
724 		if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
725 		    ifp->if_bridge == NULL) {
726 			m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
727 			ipstat.ips_outhwcsum++;
728 		} else {
729 			ip->ip_sum = 0;
730 			ip->ip_sum = in_cksum(m, hlen);
731 		}
732 		/* Update relevant hardware checksum stats for TCP/UDP */
733 		if (m->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT)
734 			tcpstat.tcps_outhwcsum++;
735 		else if (m->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT)
736 			udpstat.udps_outhwcsum++;
737 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
738 		goto done;
739 	}
740 
741 	/*
742 	 * Too large for interface; fragment if possible.
743 	 * Must be able to put at least 8 bytes per fragment.
744 	 */
745 	if (ip->ip_off & htons(IP_DF)) {
746 #ifdef IPSEC
747 		icmp_mtu = ifp->if_mtu;
748 #endif
749 		error = EMSGSIZE;
750 		/*
751 		 * This case can happen if the user changed the MTU
752 		 * of an interface after enabling IP on it.  Because
753 		 * most netifs don't keep track of routes pointing to
754 		 * them, there is no way for one to update all its
755 		 * routes when the MTU is changed.
756 		 */
757 		if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
758 		    !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
759 		    (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
760 			ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
761 		}
762 		ipstat.ips_cantfrag++;
763 		goto bad;
764 	}
765 
766 	error = ip_fragment(m, ifp, mtu);
767 	if (error) {
768 		m = m0 = NULL;
769 		goto bad;
770 	}
771 
772 	for (; m; m = m0) {
773 		m0 = m->m_nextpkt;
774 		m->m_nextpkt = 0;
775 		if (error == 0)
776 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
777 			    ro->ro_rt);
778 		else
779 			m_freem(m);
780 	}
781 
782 	if (error == 0)
783 		ipstat.ips_fragmented++;
784 
785 done:
786 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt)
787 		RTFREE(ro->ro_rt);
788 	return (error);
789 bad:
790 #ifdef IPSEC
791 	if (error == EMSGSIZE && ip_mtudisc && icmp_mtu != 0 && m != NULL)
792 		ipsec_adjust_mtu(m, icmp_mtu);
793 #endif
794 	m_freem(m0);
795 	goto done;
796 }
797 
798 int
799 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
800 {
801 	struct ip *ip, *mhip;
802 	struct mbuf *m0;
803 	int len, hlen, off;
804 	int mhlen, firstlen;
805 	struct mbuf **mnext;
806 	int fragments = 0;
807 	int s;
808 	int error = 0;
809 
810 	ip = mtod(m, struct ip *);
811 	hlen = ip->ip_hl << 2;
812 
813 	len = (mtu - hlen) &~ 7;
814 	if (len < 8) {
815 		m_freem(m);
816 		return (EMSGSIZE);
817 	}
818 
819 	/*
820 	 * If we are doing fragmentation, we can't defer TCP/UDP
821 	 * checksumming; compute the checksum and clear the flag.
822 	 */
823 	if (m->m_pkthdr.csum_flags & (M_TCPV4_CSUM_OUT | M_UDPV4_CSUM_OUT)) {
824 		in_delayed_cksum(m);
825 		m->m_pkthdr.csum_flags &= ~(M_UDPV4_CSUM_OUT | M_TCPV4_CSUM_OUT);
826 	}
827 
828 	firstlen = len;
829 	mnext = &m->m_nextpkt;
830 
831 	/*
832 	 * Loop through length of segment after first fragment,
833 	 * make new header and copy data of each part and link onto chain.
834 	 */
835 	m0 = m;
836 	mhlen = sizeof (struct ip);
837 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
838 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
839 		if (m == 0) {
840 			ipstat.ips_odropped++;
841 			error = ENOBUFS;
842 			goto sendorfree;
843 		}
844 		*mnext = m;
845 		mnext = &m->m_nextpkt;
846 		m->m_data += max_linkhdr;
847 		mhip = mtod(m, struct ip *);
848 		*mhip = *ip;
849 		/* we must inherit MCAST and BCAST flags */
850 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
851 		if (hlen > sizeof (struct ip)) {
852 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
853 			mhip->ip_hl = mhlen >> 2;
854 		}
855 		m->m_len = mhlen;
856 		mhip->ip_off = ((off - hlen) >> 3) +
857 		    (ntohs(ip->ip_off) & ~IP_MF);
858 		if (ip->ip_off & htons(IP_MF))
859 			mhip->ip_off |= IP_MF;
860 		if (off + len >= ntohs(ip->ip_len))
861 			len = ntohs(ip->ip_len) - off;
862 		else
863 			mhip->ip_off |= IP_MF;
864 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
865 		m->m_next = m_copy(m0, off, len);
866 		if (m->m_next == 0) {
867 			ipstat.ips_odropped++;
868 			error = ENOBUFS;
869 			goto sendorfree;
870 		}
871 		m->m_pkthdr.len = mhlen + len;
872 		m->m_pkthdr.rcvif = (struct ifnet *)0;
873 		mhip->ip_off = htons((u_int16_t)mhip->ip_off);
874 		if ((ifp != NULL) &&
875 		    (ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
876 		    ifp->if_bridge == NULL) {
877 			m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
878 			ipstat.ips_outhwcsum++;
879 		} else {
880 			mhip->ip_sum = 0;
881 			mhip->ip_sum = in_cksum(m, mhlen);
882 		}
883 		ipstat.ips_ofragments++;
884 		fragments++;
885 	}
886 	/*
887 	 * Update first fragment by trimming what's been copied out
888 	 * and updating header, then send each fragment (in order).
889 	 */
890 	m = m0;
891 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
892 	m->m_pkthdr.len = hlen + firstlen;
893 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
894 	ip->ip_off |= htons(IP_MF);
895 	if ((ifp != NULL) &&
896 	    (ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
897 	    ifp->if_bridge == NULL) {
898 		m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
899 		ipstat.ips_outhwcsum++;
900 	} else {
901 		ip->ip_sum = 0;
902 		ip->ip_sum = in_cksum(m, hlen);
903 	}
904 sendorfree:
905 	/*
906 	 * If there is no room for all the fragments, don't queue
907 	 * any of them.
908 	 */
909 	if (ifp != NULL) {
910 		s = splnet();
911 		if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
912 		    error == 0) {
913 			error = ENOBUFS;
914 			ipstat.ips_odropped++;
915 			IFQ_INC_DROPS(&ifp->if_snd);
916 		}
917 		splx(s);
918 	}
919 	if (error) {
920 		for (m = m0; m; m = m0) {
921 			m0 = m->m_nextpkt;
922 			m->m_nextpkt = NULL;
923 			m_freem(m);
924 		}
925 	}
926 
927 	return (error);
928 }
929 
930 /*
931  * Insert IP options into preformed packet.
932  * Adjust IP destination as required for IP source routing,
933  * as indicated by a non-zero in_addr at the start of the options.
934  */
935 struct mbuf *
936 ip_insertoptions(m, opt, phlen)
937 	struct mbuf *m;
938 	struct mbuf *opt;
939 	int *phlen;
940 {
941 	struct ipoption *p = mtod(opt, struct ipoption *);
942 	struct mbuf *n;
943 	struct ip *ip = mtod(m, struct ip *);
944 	unsigned optlen;
945 
946 	optlen = opt->m_len - sizeof(p->ipopt_dst);
947 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
948 		return (m);		/* XXX should fail */
949 	if (p->ipopt_dst.s_addr)
950 		ip->ip_dst = p->ipopt_dst;
951 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
952 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
953 		if (n == 0)
954 			return (m);
955 		M_MOVE_HDR(n, m);
956 		n->m_pkthdr.len += optlen;
957 		m->m_len -= sizeof(struct ip);
958 		m->m_data += sizeof(struct ip);
959 		n->m_next = m;
960 		m = n;
961 		m->m_len = optlen + sizeof(struct ip);
962 		m->m_data += max_linkhdr;
963 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
964 	} else {
965 		m->m_data -= optlen;
966 		m->m_len += optlen;
967 		m->m_pkthdr.len += optlen;
968 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
969 	}
970 	ip = mtod(m, struct ip *);
971 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
972 	*phlen = sizeof(struct ip) + optlen;
973 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
974 	return (m);
975 }
976 
977 /*
978  * Copy options from ip to jp,
979  * omitting those not copied during fragmentation.
980  */
981 int
982 ip_optcopy(ip, jp)
983 	struct ip *ip, *jp;
984 {
985 	u_char *cp, *dp;
986 	int opt, optlen, cnt;
987 
988 	cp = (u_char *)(ip + 1);
989 	dp = (u_char *)(jp + 1);
990 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
991 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
992 		opt = cp[0];
993 		if (opt == IPOPT_EOL)
994 			break;
995 		if (opt == IPOPT_NOP) {
996 			/* Preserve for IP mcast tunnel's LSRR alignment. */
997 			*dp++ = IPOPT_NOP;
998 			optlen = 1;
999 			continue;
1000 		}
1001 #ifdef DIAGNOSTIC
1002 		if (cnt < IPOPT_OLEN + sizeof(*cp))
1003 			panic("malformed IPv4 option passed to ip_optcopy");
1004 #endif
1005 		optlen = cp[IPOPT_OLEN];
1006 #ifdef DIAGNOSTIC
1007 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1008 			panic("malformed IPv4 option passed to ip_optcopy");
1009 #endif
1010 		/* bogus lengths should have been caught by ip_dooptions */
1011 		if (optlen > cnt)
1012 			optlen = cnt;
1013 		if (IPOPT_COPIED(opt)) {
1014 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
1015 			dp += optlen;
1016 		}
1017 	}
1018 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1019 		*dp++ = IPOPT_EOL;
1020 	return (optlen);
1021 }
1022 
1023 /*
1024  * IP socket option processing.
1025  */
1026 int
1027 ip_ctloutput(op, so, level, optname, mp)
1028 	int op;
1029 	struct socket *so;
1030 	int level, optname;
1031 	struct mbuf **mp;
1032 {
1033 	struct inpcb *inp = sotoinpcb(so);
1034 	struct mbuf *m = *mp;
1035 	int optval = 0;
1036 #ifdef IPSEC
1037 	struct proc *p = curproc; /* XXX */
1038 	struct ipsec_ref *ipr;
1039 	u_int16_t opt16val;
1040 #endif
1041 	int error = 0;
1042 
1043 	if (level != IPPROTO_IP) {
1044 		error = EINVAL;
1045 		if (op == PRCO_SETOPT && *mp)
1046 			(void) m_free(*mp);
1047 	} else switch (op) {
1048 	case PRCO_SETOPT:
1049 		switch (optname) {
1050 		case IP_OPTIONS:
1051 #ifdef notyet
1052 		case IP_RETOPTS:
1053 			return (ip_pcbopts(optname, &inp->inp_options, m));
1054 #else
1055 			return (ip_pcbopts(&inp->inp_options, m));
1056 #endif
1057 
1058 		case IP_TOS:
1059 		case IP_TTL:
1060 		case IP_MINTTL:
1061 		case IP_RECVOPTS:
1062 		case IP_RECVRETOPTS:
1063 		case IP_RECVDSTADDR:
1064 		case IP_RECVIF:
1065 		case IP_RECVTTL:
1066 		case IP_RECVDSTPORT:
1067 			if (m == NULL || m->m_len != sizeof(int))
1068 				error = EINVAL;
1069 			else {
1070 				optval = *mtod(m, int *);
1071 				switch (optname) {
1072 
1073 				case IP_TOS:
1074 					inp->inp_ip.ip_tos = optval;
1075 					break;
1076 
1077 				case IP_TTL:
1078 					if (optval > 0 && optval <= MAXTTL)
1079 						inp->inp_ip.ip_ttl = optval;
1080 					else
1081 						error = EINVAL;
1082 					break;
1083 
1084 				case IP_MINTTL:
1085 					if (optval > 0 && optval <= MAXTTL)
1086 						inp->inp_ip_minttl = optval;
1087 					else
1088 						error = EINVAL;
1089 					break;
1090 #define	OPTSET(bit) \
1091 	if (optval) \
1092 		inp->inp_flags |= bit; \
1093 	else \
1094 		inp->inp_flags &= ~bit;
1095 
1096 				case IP_RECVOPTS:
1097 					OPTSET(INP_RECVOPTS);
1098 					break;
1099 
1100 				case IP_RECVRETOPTS:
1101 					OPTSET(INP_RECVRETOPTS);
1102 					break;
1103 
1104 				case IP_RECVDSTADDR:
1105 					OPTSET(INP_RECVDSTADDR);
1106 					break;
1107 				case IP_RECVIF:
1108 					OPTSET(INP_RECVIF);
1109 					break;
1110 				case IP_RECVTTL:
1111 					OPTSET(INP_RECVTTL);
1112 					break;
1113 				case IP_RECVDSTPORT:
1114 					OPTSET(INP_RECVDSTPORT);
1115 					break;
1116 				}
1117 			}
1118 			break;
1119 #undef OPTSET
1120 
1121 		case IP_MULTICAST_IF:
1122 		case IP_MULTICAST_TTL:
1123 		case IP_MULTICAST_LOOP:
1124 		case IP_ADD_MEMBERSHIP:
1125 		case IP_DROP_MEMBERSHIP:
1126 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
1127 			break;
1128 
1129 		case IP_PORTRANGE:
1130 			if (m == 0 || m->m_len != sizeof(int))
1131 				error = EINVAL;
1132 			else {
1133 				optval = *mtod(m, int *);
1134 
1135 				switch (optval) {
1136 
1137 				case IP_PORTRANGE_DEFAULT:
1138 					inp->inp_flags &= ~(INP_LOWPORT);
1139 					inp->inp_flags &= ~(INP_HIGHPORT);
1140 					break;
1141 
1142 				case IP_PORTRANGE_HIGH:
1143 					inp->inp_flags &= ~(INP_LOWPORT);
1144 					inp->inp_flags |= INP_HIGHPORT;
1145 					break;
1146 
1147 				case IP_PORTRANGE_LOW:
1148 					inp->inp_flags &= ~(INP_HIGHPORT);
1149 					inp->inp_flags |= INP_LOWPORT;
1150 					break;
1151 
1152 				default:
1153 
1154 					error = EINVAL;
1155 					break;
1156 				}
1157 			}
1158 			break;
1159 		case IP_AUTH_LEVEL:
1160 		case IP_ESP_TRANS_LEVEL:
1161 		case IP_ESP_NETWORK_LEVEL:
1162 		case IP_IPCOMP_LEVEL:
1163 #ifndef IPSEC
1164 			error = EOPNOTSUPP;
1165 #else
1166 			if (m == 0 || m->m_len != sizeof(int)) {
1167 				error = EINVAL;
1168 				break;
1169 			}
1170 			optval = *mtod(m, int *);
1171 
1172 			if (optval < IPSEC_LEVEL_BYPASS ||
1173 			    optval > IPSEC_LEVEL_UNIQUE) {
1174 				error = EINVAL;
1175 				break;
1176 			}
1177 
1178 			/* Unlink cached output TDB to force a re-search */
1179 			if (inp->inp_tdb_out) {
1180 				int s = spltdb();
1181 				TAILQ_REMOVE(&inp->inp_tdb_out->tdb_inp_out,
1182 				    inp, inp_tdb_out_next);
1183 				splx(s);
1184 			}
1185 
1186 			if (inp->inp_tdb_in) {
1187 				int s = spltdb();
1188 				TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in,
1189 				    inp, inp_tdb_in_next);
1190 				splx(s);
1191 			}
1192 
1193 			switch (optname) {
1194 			case IP_AUTH_LEVEL:
1195 				if (optval < ipsec_auth_default_level &&
1196 				    suser(p, 0)) {
1197 					error = EACCES;
1198 					break;
1199 				}
1200 				inp->inp_seclevel[SL_AUTH] = optval;
1201 				break;
1202 
1203 			case IP_ESP_TRANS_LEVEL:
1204 				if (optval < ipsec_esp_trans_default_level &&
1205 				    suser(p, 0)) {
1206 					error = EACCES;
1207 					break;
1208 				}
1209 				inp->inp_seclevel[SL_ESP_TRANS] = optval;
1210 				break;
1211 
1212 			case IP_ESP_NETWORK_LEVEL:
1213 				if (optval < ipsec_esp_network_default_level &&
1214 				    suser(p, 0)) {
1215 					error = EACCES;
1216 					break;
1217 				}
1218 				inp->inp_seclevel[SL_ESP_NETWORK] = optval;
1219 				break;
1220 			case IP_IPCOMP_LEVEL:
1221 				if (optval < ipsec_ipcomp_default_level &&
1222 				    suser(p, 0)) {
1223 					error = EACCES;
1224 					break;
1225 				}
1226 				inp->inp_seclevel[SL_IPCOMP] = optval;
1227 				break;
1228 			}
1229 			if (!error)
1230 				inp->inp_secrequire = get_sa_require(inp);
1231 #endif
1232 			break;
1233 
1234 		case IP_IPSEC_REMOTE_CRED:
1235 		case IP_IPSEC_REMOTE_AUTH:
1236 			/* Can't set the remote credential or key */
1237 			error = EOPNOTSUPP;
1238 			break;
1239 
1240 		case IP_IPSEC_LOCAL_ID:
1241 		case IP_IPSEC_REMOTE_ID:
1242 		case IP_IPSEC_LOCAL_CRED:
1243 		case IP_IPSEC_LOCAL_AUTH:
1244 #ifndef IPSEC
1245 			error = EOPNOTSUPP;
1246 #else
1247 			if (m->m_len < 2) {
1248 				error = EINVAL;
1249 				break;
1250 			}
1251 
1252 			m_copydata(m, 0, 2, (caddr_t) &opt16val);
1253 
1254 			/* If the type is 0, then we cleanup and return */
1255 			if (opt16val == 0) {
1256 				switch (optname) {
1257 				case IP_IPSEC_LOCAL_ID:
1258 					if (inp->inp_ipo != NULL &&
1259 					    inp->inp_ipo->ipo_srcid != NULL) {
1260 						ipsp_reffree(inp->inp_ipo->ipo_srcid);
1261 						inp->inp_ipo->ipo_srcid = NULL;
1262 					}
1263 					break;
1264 
1265 				case IP_IPSEC_REMOTE_ID:
1266 					if (inp->inp_ipo != NULL &&
1267 					    inp->inp_ipo->ipo_dstid != NULL) {
1268 						ipsp_reffree(inp->inp_ipo->ipo_dstid);
1269 						inp->inp_ipo->ipo_dstid = NULL;
1270 					}
1271 					break;
1272 
1273 				case IP_IPSEC_LOCAL_CRED:
1274 					if (inp->inp_ipo != NULL &&
1275 					    inp->inp_ipo->ipo_local_cred != NULL) {
1276 						ipsp_reffree(inp->inp_ipo->ipo_local_cred);
1277 						inp->inp_ipo->ipo_local_cred = NULL;
1278 					}
1279 					break;
1280 
1281 				case IP_IPSEC_LOCAL_AUTH:
1282 					if (inp->inp_ipo != NULL &&
1283 					    inp->inp_ipo->ipo_local_auth != NULL) {
1284 						ipsp_reffree(inp->inp_ipo->ipo_local_auth);
1285 						inp->inp_ipo->ipo_local_auth = NULL;
1286 					}
1287 					break;
1288 				}
1289 
1290 				error = 0;
1291 				break;
1292 			}
1293 
1294 			/* Can't have an empty payload */
1295 			if (m->m_len == 2) {
1296 				error = EINVAL;
1297 				break;
1298 			}
1299 
1300 			/* Allocate if needed */
1301 			if (inp->inp_ipo == NULL) {
1302 				inp->inp_ipo = ipsec_add_policy(inp,
1303 				    AF_INET, IPSP_DIRECTION_OUT);
1304 				if (inp->inp_ipo == NULL) {
1305 					error = ENOBUFS;
1306 					break;
1307 				}
1308 			}
1309 
1310 			ipr = malloc(sizeof(struct ipsec_ref) + m->m_len - 2,
1311 			       M_CREDENTIALS, M_NOWAIT);
1312 			if (ipr == NULL) {
1313 				error = ENOBUFS;
1314 				break;
1315 			}
1316 
1317 			ipr->ref_count = 1;
1318 			ipr->ref_malloctype = M_CREDENTIALS;
1319 			ipr->ref_len = m->m_len - 2;
1320 			ipr->ref_type = opt16val;
1321 			m_copydata(m, 2, m->m_len - 2, (caddr_t)(ipr + 1));
1322 
1323 			switch (optname) {
1324 			case IP_IPSEC_LOCAL_ID:
1325 				/* Check valid types and NUL-termination */
1326 				if (ipr->ref_type < IPSP_IDENTITY_PREFIX ||
1327 				    ipr->ref_type > IPSP_IDENTITY_CONNECTION ||
1328 				    ((char *)(ipr + 1))[ipr->ref_len - 1]) {
1329 					free(ipr, M_CREDENTIALS);
1330 					error = EINVAL;
1331 				} else {
1332 					if (inp->inp_ipo->ipo_srcid != NULL)
1333 						ipsp_reffree(inp->inp_ipo->ipo_srcid);
1334 					inp->inp_ipo->ipo_srcid = ipr;
1335 				}
1336 				break;
1337 			case IP_IPSEC_REMOTE_ID:
1338 				/* Check valid types and NUL-termination */
1339 				if (ipr->ref_type < IPSP_IDENTITY_PREFIX ||
1340 				    ipr->ref_type > IPSP_IDENTITY_CONNECTION ||
1341 				    ((char *)(ipr + 1))[ipr->ref_len - 1]) {
1342 					free(ipr, M_CREDENTIALS);
1343 					error = EINVAL;
1344 				} else {
1345 					if (inp->inp_ipo->ipo_dstid != NULL)
1346 						ipsp_reffree(inp->inp_ipo->ipo_dstid);
1347 					inp->inp_ipo->ipo_dstid = ipr;
1348 				}
1349 				break;
1350 			case IP_IPSEC_LOCAL_CRED:
1351 				if (ipr->ref_type < IPSP_CRED_KEYNOTE ||
1352 				    ipr->ref_type > IPSP_CRED_X509) {
1353 					free(ipr, M_CREDENTIALS);
1354 					error = EINVAL;
1355 				} else {
1356 					if (inp->inp_ipo->ipo_local_cred != NULL)
1357 						ipsp_reffree(inp->inp_ipo->ipo_local_cred);
1358 					inp->inp_ipo->ipo_local_cred = ipr;
1359 				}
1360 				break;
1361 			case IP_IPSEC_LOCAL_AUTH:
1362 				if (ipr->ref_type < IPSP_AUTH_PASSPHRASE ||
1363 				    ipr->ref_type > IPSP_AUTH_RSA) {
1364 					free(ipr, M_CREDENTIALS);
1365 					error = EINVAL;
1366 				} else {
1367 					if (inp->inp_ipo->ipo_local_auth != NULL)
1368 						ipsp_reffree(inp->inp_ipo->ipo_local_auth);
1369 					inp->inp_ipo->ipo_local_auth = ipr;
1370 				}
1371 				break;
1372 			}
1373 
1374 			/* Unlink cached output TDB to force a re-search */
1375 			if (inp->inp_tdb_out) {
1376 				int s = spltdb();
1377 				TAILQ_REMOVE(&inp->inp_tdb_out->tdb_inp_out,
1378 				    inp, inp_tdb_out_next);
1379 				splx(s);
1380 			}
1381 
1382 			if (inp->inp_tdb_in) {
1383 				int s = spltdb();
1384 				TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in,
1385 				    inp, inp_tdb_in_next);
1386 				splx(s);
1387 			}
1388 #endif
1389 			break;
1390 		default:
1391 			error = ENOPROTOOPT;
1392 			break;
1393 		}
1394 		if (m)
1395 			(void)m_free(m);
1396 		break;
1397 
1398 	case PRCO_GETOPT:
1399 		switch (optname) {
1400 		case IP_OPTIONS:
1401 		case IP_RETOPTS:
1402 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1403 			if (inp->inp_options) {
1404 				m->m_len = inp->inp_options->m_len;
1405 				bcopy(mtod(inp->inp_options, caddr_t),
1406 				    mtod(m, caddr_t), (unsigned)m->m_len);
1407 			} else
1408 				m->m_len = 0;
1409 			break;
1410 
1411 		case IP_TOS:
1412 		case IP_TTL:
1413 		case IP_MINTTL:
1414 		case IP_RECVOPTS:
1415 		case IP_RECVRETOPTS:
1416 		case IP_RECVDSTADDR:
1417 		case IP_RECVIF:
1418 		case IP_RECVTTL:
1419 		case IP_RECVDSTPORT:
1420 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1421 			m->m_len = sizeof(int);
1422 			switch (optname) {
1423 
1424 			case IP_TOS:
1425 				optval = inp->inp_ip.ip_tos;
1426 				break;
1427 
1428 			case IP_TTL:
1429 				optval = inp->inp_ip.ip_ttl;
1430 				break;
1431 
1432 			case IP_MINTTL:
1433 				optval = inp->inp_ip_minttl;
1434 				break;
1435 
1436 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1437 
1438 			case IP_RECVOPTS:
1439 				optval = OPTBIT(INP_RECVOPTS);
1440 				break;
1441 
1442 			case IP_RECVRETOPTS:
1443 				optval = OPTBIT(INP_RECVRETOPTS);
1444 				break;
1445 
1446 			case IP_RECVDSTADDR:
1447 				optval = OPTBIT(INP_RECVDSTADDR);
1448 				break;
1449 			case IP_RECVIF:
1450 				optval = OPTBIT(INP_RECVIF);
1451 				break;
1452 			case IP_RECVTTL:
1453 				optval = OPTBIT(INP_RECVTTL);
1454 				break;
1455 			case IP_RECVDSTPORT:
1456 				optval = OPTBIT(INP_RECVDSTPORT);
1457 				break;
1458 			}
1459 			*mtod(m, int *) = optval;
1460 			break;
1461 
1462 		case IP_MULTICAST_IF:
1463 		case IP_MULTICAST_TTL:
1464 		case IP_MULTICAST_LOOP:
1465 		case IP_ADD_MEMBERSHIP:
1466 		case IP_DROP_MEMBERSHIP:
1467 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1468 			break;
1469 
1470 		case IP_PORTRANGE:
1471 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1472 			m->m_len = sizeof(int);
1473 
1474 			if (inp->inp_flags & INP_HIGHPORT)
1475 				optval = IP_PORTRANGE_HIGH;
1476 			else if (inp->inp_flags & INP_LOWPORT)
1477 				optval = IP_PORTRANGE_LOW;
1478 			else
1479 				optval = 0;
1480 
1481 			*mtod(m, int *) = optval;
1482 			break;
1483 
1484 		case IP_AUTH_LEVEL:
1485 		case IP_ESP_TRANS_LEVEL:
1486 		case IP_ESP_NETWORK_LEVEL:
1487 		case IP_IPCOMP_LEVEL:
1488 #ifndef IPSEC
1489 			m->m_len = sizeof(int);
1490 			*mtod(m, int *) = IPSEC_LEVEL_NONE;
1491 #else
1492 			m->m_len = sizeof(int);
1493 			switch (optname) {
1494 			case IP_AUTH_LEVEL:
1495 				optval = inp->inp_seclevel[SL_AUTH];
1496 				break;
1497 
1498 			case IP_ESP_TRANS_LEVEL:
1499 				optval = inp->inp_seclevel[SL_ESP_TRANS];
1500 				break;
1501 
1502 			case IP_ESP_NETWORK_LEVEL:
1503 				optval = inp->inp_seclevel[SL_ESP_NETWORK];
1504 				break;
1505 			case IP_IPCOMP_LEVEL:
1506 				optval = inp->inp_seclevel[SL_IPCOMP];
1507 				break;
1508 			}
1509 			*mtod(m, int *) = optval;
1510 #endif
1511 			break;
1512 		case IP_IPSEC_LOCAL_ID:
1513 		case IP_IPSEC_REMOTE_ID:
1514 		case IP_IPSEC_LOCAL_CRED:
1515 		case IP_IPSEC_REMOTE_CRED:
1516 		case IP_IPSEC_LOCAL_AUTH:
1517 		case IP_IPSEC_REMOTE_AUTH:
1518 #ifndef IPSEC
1519 			error = EOPNOTSUPP;
1520 #else
1521 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1522 			m->m_len = sizeof(u_int16_t);
1523 			ipr = NULL;
1524 			switch (optname) {
1525 			case IP_IPSEC_LOCAL_ID:
1526 				if (inp->inp_ipo != NULL)
1527 					ipr = inp->inp_ipo->ipo_srcid;
1528 				opt16val = IPSP_IDENTITY_NONE;
1529 				break;
1530 			case IP_IPSEC_REMOTE_ID:
1531 				if (inp->inp_ipo != NULL)
1532 					ipr = inp->inp_ipo->ipo_dstid;
1533 				opt16val = IPSP_IDENTITY_NONE;
1534 				break;
1535 			case IP_IPSEC_LOCAL_CRED:
1536 				if (inp->inp_ipo != NULL)
1537 					ipr = inp->inp_ipo->ipo_local_cred;
1538 				opt16val = IPSP_CRED_NONE;
1539 				break;
1540 			case IP_IPSEC_REMOTE_CRED:
1541 				ipr = inp->inp_ipsec_remotecred;
1542 				opt16val = IPSP_CRED_NONE;
1543 				break;
1544 			case IP_IPSEC_LOCAL_AUTH:
1545 				if (inp->inp_ipo != NULL)
1546 					ipr = inp->inp_ipo->ipo_local_auth;
1547 				break;
1548 			case IP_IPSEC_REMOTE_AUTH:
1549 				ipr = inp->inp_ipsec_remoteauth;
1550 				break;
1551 			}
1552 			if (ipr == NULL)
1553 				*mtod(m, u_int16_t *) = opt16val;
1554 			else {
1555 				size_t len;
1556 
1557 				len = m->m_len + ipr->ref_len;
1558 				if (len > MCLBYTES) {
1559 					 m_free(m);
1560 					 error = EINVAL;
1561 					 break;
1562 				}
1563 				/* allocate mbuf cluster for larger option */
1564 				if (len > MLEN) {
1565 					 MCLGET(m, M_WAITOK);
1566 					 if ((m->m_flags & M_EXT) == 0) {
1567 						 m_free(m);
1568 						 error = ENOBUFS;
1569 						 break;
1570 					 }
1571 
1572 				}
1573 				m->m_len = len;
1574 				*mtod(m, u_int16_t *) = ipr->ref_type;
1575 				m_copyback(m, sizeof(u_int16_t), ipr->ref_len,
1576 				    ipr + 1);
1577 			}
1578 #endif
1579 			break;
1580 		default:
1581 			error = ENOPROTOOPT;
1582 			break;
1583 		}
1584 		break;
1585 	}
1586 	return (error);
1587 }
1588 
1589 /*
1590  * Set up IP options in pcb for insertion in output packets.
1591  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1592  * with destination address if source routed.
1593  */
1594 int
1595 #ifdef notyet
1596 ip_pcbopts(optname, pcbopt, m)
1597 	int optname;
1598 #else
1599 ip_pcbopts(pcbopt, m)
1600 #endif
1601 	struct mbuf **pcbopt;
1602 	struct mbuf *m;
1603 {
1604 	int cnt, optlen;
1605 	u_char *cp;
1606 	u_char opt;
1607 
1608 	/* turn off any old options */
1609 	if (*pcbopt)
1610 		(void)m_free(*pcbopt);
1611 	*pcbopt = 0;
1612 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1613 		/*
1614 		 * Only turning off any previous options.
1615 		 */
1616 		if (m)
1617 			(void)m_free(m);
1618 		return (0);
1619 	}
1620 
1621 	if (m->m_len % sizeof(int32_t))
1622 		goto bad;
1623 
1624 	/*
1625 	 * IP first-hop destination address will be stored before
1626 	 * actual options; move other options back
1627 	 * and clear it when none present.
1628 	 */
1629 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1630 		goto bad;
1631 	cnt = m->m_len;
1632 	m->m_len += sizeof(struct in_addr);
1633 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1634 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1635 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1636 
1637 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1638 		opt = cp[IPOPT_OPTVAL];
1639 		if (opt == IPOPT_EOL)
1640 			break;
1641 		if (opt == IPOPT_NOP)
1642 			optlen = 1;
1643 		else {
1644 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1645 				goto bad;
1646 			optlen = cp[IPOPT_OLEN];
1647 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1648 				goto bad;
1649 		}
1650 		switch (opt) {
1651 
1652 		default:
1653 			break;
1654 
1655 		case IPOPT_LSRR:
1656 		case IPOPT_SSRR:
1657 			/*
1658 			 * user process specifies route as:
1659 			 *	->A->B->C->D
1660 			 * D must be our final destination (but we can't
1661 			 * check that since we may not have connected yet).
1662 			 * A is first hop destination, which doesn't appear in
1663 			 * actual IP option, but is stored before the options.
1664 			 */
1665 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1666 				goto bad;
1667 			m->m_len -= sizeof(struct in_addr);
1668 			cnt -= sizeof(struct in_addr);
1669 			optlen -= sizeof(struct in_addr);
1670 			cp[IPOPT_OLEN] = optlen;
1671 			/*
1672 			 * Move first hop before start of options.
1673 			 */
1674 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1675 			    sizeof(struct in_addr));
1676 			/*
1677 			 * Then copy rest of options back
1678 			 * to close up the deleted entry.
1679 			 */
1680 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1681 			    sizeof(struct in_addr)),
1682 			    (caddr_t)&cp[IPOPT_OFFSET+1],
1683 			    (unsigned)cnt - (IPOPT_OFFSET+1));
1684 			break;
1685 		}
1686 	}
1687 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1688 		goto bad;
1689 	*pcbopt = m;
1690 	return (0);
1691 
1692 bad:
1693 	(void)m_free(m);
1694 	return (EINVAL);
1695 }
1696 
1697 /*
1698  * Set the IP multicast options in response to user setsockopt().
1699  */
1700 int
1701 ip_setmoptions(optname, imop, m)
1702 	int optname;
1703 	struct ip_moptions **imop;
1704 	struct mbuf *m;
1705 {
1706 	int error = 0;
1707 	u_char loop;
1708 	int i;
1709 	struct in_addr addr;
1710 	struct ip_mreq *mreq;
1711 	struct ifnet *ifp;
1712 	struct ip_moptions *imo = *imop;
1713 	struct in_multi **immp;
1714 	struct route ro;
1715 	struct sockaddr_in *dst;
1716 
1717 	if (imo == NULL) {
1718 		/*
1719 		 * No multicast option buffer attached to the pcb;
1720 		 * allocate one and initialize to default values.
1721 		 */
1722 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1723 		    M_WAITOK|M_ZERO);
1724 		immp = (struct in_multi **)malloc(
1725 		    (sizeof(*immp) * IP_MIN_MEMBERSHIPS), M_IPMOPTS,
1726 		    M_WAITOK|M_ZERO);
1727 		*imop = imo;
1728 		imo->imo_multicast_ifp = NULL;
1729 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1730 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1731 		imo->imo_num_memberships = 0;
1732 		imo->imo_max_memberships = IP_MIN_MEMBERSHIPS;
1733 		imo->imo_membership = immp;
1734 	}
1735 
1736 	switch (optname) {
1737 
1738 	case IP_MULTICAST_IF:
1739 		/*
1740 		 * Select the interface for outgoing multicast packets.
1741 		 */
1742 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1743 			error = EINVAL;
1744 			break;
1745 		}
1746 		addr = *(mtod(m, struct in_addr *));
1747 		/*
1748 		 * INADDR_ANY is used to remove a previous selection.
1749 		 * When no interface is selected, a default one is
1750 		 * chosen every time a multicast packet is sent.
1751 		 */
1752 		if (addr.s_addr == INADDR_ANY) {
1753 			imo->imo_multicast_ifp = NULL;
1754 			break;
1755 		}
1756 		/*
1757 		 * The selected interface is identified by its local
1758 		 * IP address.  Find the interface and confirm that
1759 		 * it supports multicasting.
1760 		 */
1761 		INADDR_TO_IFP(addr, ifp);
1762 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1763 			error = EADDRNOTAVAIL;
1764 			break;
1765 		}
1766 		imo->imo_multicast_ifp = ifp;
1767 		break;
1768 
1769 	case IP_MULTICAST_TTL:
1770 		/*
1771 		 * Set the IP time-to-live for outgoing multicast packets.
1772 		 */
1773 		if (m == NULL || m->m_len != 1) {
1774 			error = EINVAL;
1775 			break;
1776 		}
1777 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1778 		break;
1779 
1780 	case IP_MULTICAST_LOOP:
1781 		/*
1782 		 * Set the loopback flag for outgoing multicast packets.
1783 		 * Must be zero or one.
1784 		 */
1785 		if (m == NULL || m->m_len != 1 ||
1786 		   (loop = *(mtod(m, u_char *))) > 1) {
1787 			error = EINVAL;
1788 			break;
1789 		}
1790 		imo->imo_multicast_loop = loop;
1791 		break;
1792 
1793 	case IP_ADD_MEMBERSHIP:
1794 		/*
1795 		 * Add a multicast group membership.
1796 		 * Group must be a valid IP multicast address.
1797 		 */
1798 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1799 			error = EINVAL;
1800 			break;
1801 		}
1802 		mreq = mtod(m, struct ip_mreq *);
1803 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1804 			error = EINVAL;
1805 			break;
1806 		}
1807 		/*
1808 		 * If no interface address was provided, use the interface of
1809 		 * the route to the given multicast address.
1810 		 */
1811 		if (mreq->imr_interface.s_addr == INADDR_ANY) {
1812 			ro.ro_rt = NULL;
1813 			dst = satosin(&ro.ro_dst);
1814 			dst->sin_len = sizeof(*dst);
1815 			dst->sin_family = AF_INET;
1816 			dst->sin_addr = mreq->imr_multiaddr;
1817 			rtalloc(&ro);
1818 			if (ro.ro_rt == NULL) {
1819 				error = EADDRNOTAVAIL;
1820 				break;
1821 			}
1822 			ifp = ro.ro_rt->rt_ifp;
1823 			rtfree(ro.ro_rt);
1824 		} else {
1825 			INADDR_TO_IFP(mreq->imr_interface, ifp);
1826 		}
1827 		/*
1828 		 * See if we found an interface, and confirm that it
1829 		 * supports multicast.
1830 		 */
1831 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1832 			error = EADDRNOTAVAIL;
1833 			break;
1834 		}
1835 		/*
1836 		 * See if the membership already exists or if all the
1837 		 * membership slots are full.
1838 		 */
1839 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1840 			if (imo->imo_membership[i]->inm_ia->ia_ifp == ifp &&
1841 			    imo->imo_membership[i]->inm_addr.s_addr
1842 						== mreq->imr_multiaddr.s_addr)
1843 				break;
1844 		}
1845 		if (i < imo->imo_num_memberships) {
1846 			error = EADDRINUSE;
1847 			break;
1848 		}
1849 		if (imo->imo_num_memberships == imo->imo_max_memberships) {
1850 			struct in_multi **nmships, **omships;
1851 			size_t newmax;
1852 			/*
1853 			 * Resize the vector to next power-of-two minus 1. If the
1854 			 * size would exceed the maximum then we know we've really
1855 			 * run out of entries. Otherwise, we reallocate the vector.
1856 			 */
1857 			nmships = NULL;
1858 			omships = imo->imo_membership;
1859 			newmax = ((imo->imo_max_memberships + 1) * 2) - 1;
1860 			if (newmax <= IP_MAX_MEMBERSHIPS) {
1861 				nmships = (struct in_multi **)malloc(
1862 				    sizeof(*nmships) * newmax, M_IPMOPTS,
1863 				    M_NOWAIT|M_ZERO);
1864 				if (nmships != NULL) {
1865 					bcopy(omships, nmships,
1866 					    sizeof(*omships) *
1867 					    imo->imo_max_memberships);
1868 					free(omships, M_IPMOPTS);
1869 					imo->imo_membership = nmships;
1870 					imo->imo_max_memberships = newmax;
1871 				}
1872 			}
1873 			if (nmships == NULL) {
1874 				error = ETOOMANYREFS;
1875 				break;
1876 			}
1877 		}
1878 		/*
1879 		 * Everything looks good; add a new record to the multicast
1880 		 * address list for the given interface.
1881 		 */
1882 		if ((imo->imo_membership[i] =
1883 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1884 			error = ENOBUFS;
1885 			break;
1886 		}
1887 		++imo->imo_num_memberships;
1888 		break;
1889 
1890 	case IP_DROP_MEMBERSHIP:
1891 		/*
1892 		 * Drop a multicast group membership.
1893 		 * Group must be a valid IP multicast address.
1894 		 */
1895 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1896 			error = EINVAL;
1897 			break;
1898 		}
1899 		mreq = mtod(m, struct ip_mreq *);
1900 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1901 			error = EINVAL;
1902 			break;
1903 		}
1904 		/*
1905 		 * If an interface address was specified, get a pointer
1906 		 * to its ifnet structure.
1907 		 */
1908 		if (mreq->imr_interface.s_addr == INADDR_ANY)
1909 			ifp = NULL;
1910 		else {
1911 			INADDR_TO_IFP(mreq->imr_interface, ifp);
1912 			if (ifp == NULL) {
1913 				error = EADDRNOTAVAIL;
1914 				break;
1915 			}
1916 		}
1917 		/*
1918 		 * Find the membership in the membership array.
1919 		 */
1920 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1921 			if ((ifp == NULL ||
1922 			     imo->imo_membership[i]->inm_ia->ia_ifp == ifp) &&
1923 			     imo->imo_membership[i]->inm_addr.s_addr ==
1924 			     mreq->imr_multiaddr.s_addr)
1925 				break;
1926 		}
1927 		if (i == imo->imo_num_memberships) {
1928 			error = EADDRNOTAVAIL;
1929 			break;
1930 		}
1931 		/*
1932 		 * Give up the multicast address record to which the
1933 		 * membership points.
1934 		 */
1935 		in_delmulti(imo->imo_membership[i]);
1936 		/*
1937 		 * Remove the gap in the membership array.
1938 		 */
1939 		for (++i; i < imo->imo_num_memberships; ++i)
1940 			imo->imo_membership[i-1] = imo->imo_membership[i];
1941 		--imo->imo_num_memberships;
1942 		break;
1943 
1944 	default:
1945 		error = EOPNOTSUPP;
1946 		break;
1947 	}
1948 
1949 	/*
1950 	 * If all options have default values, no need to keep the data.
1951 	 */
1952 	if (imo->imo_multicast_ifp == NULL &&
1953 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1954 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1955 	    imo->imo_num_memberships == 0) {
1956 		free(imo->imo_membership , M_IPMOPTS);
1957 		free(*imop, M_IPMOPTS);
1958 		*imop = NULL;
1959 	}
1960 
1961 	return (error);
1962 }
1963 
1964 /*
1965  * Return the IP multicast options in response to user getsockopt().
1966  */
1967 int
1968 ip_getmoptions(optname, imo, mp)
1969 	int optname;
1970 	struct ip_moptions *imo;
1971 	struct mbuf **mp;
1972 {
1973 	u_char *ttl;
1974 	u_char *loop;
1975 	struct in_addr *addr;
1976 	struct in_ifaddr *ia;
1977 
1978 	*mp = m_get(M_WAIT, MT_SOOPTS);
1979 
1980 	switch (optname) {
1981 
1982 	case IP_MULTICAST_IF:
1983 		addr = mtod(*mp, struct in_addr *);
1984 		(*mp)->m_len = sizeof(struct in_addr);
1985 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1986 			addr->s_addr = INADDR_ANY;
1987 		else {
1988 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1989 			addr->s_addr = (ia == NULL) ? INADDR_ANY
1990 					: ia->ia_addr.sin_addr.s_addr;
1991 		}
1992 		return (0);
1993 
1994 	case IP_MULTICAST_TTL:
1995 		ttl = mtod(*mp, u_char *);
1996 		(*mp)->m_len = 1;
1997 		*ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
1998 				     : imo->imo_multicast_ttl;
1999 		return (0);
2000 
2001 	case IP_MULTICAST_LOOP:
2002 		loop = mtod(*mp, u_char *);
2003 		(*mp)->m_len = 1;
2004 		*loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
2005 				      : imo->imo_multicast_loop;
2006 		return (0);
2007 
2008 	default:
2009 		return (EOPNOTSUPP);
2010 	}
2011 }
2012 
2013 /*
2014  * Discard the IP multicast options.
2015  */
2016 void
2017 ip_freemoptions(imo)
2018 	struct ip_moptions *imo;
2019 {
2020 	int i;
2021 
2022 	if (imo != NULL) {
2023 		for (i = 0; i < imo->imo_num_memberships; ++i)
2024 			in_delmulti(imo->imo_membership[i]);
2025 		free(imo->imo_membership, M_IPMOPTS);
2026 		free(imo, M_IPMOPTS);
2027 	}
2028 }
2029 
2030 /*
2031  * Routine called from ip_output() to loop back a copy of an IP multicast
2032  * packet to the input queue of a specified interface.  Note that this
2033  * calls the output routine of the loopback "driver", but with an interface
2034  * pointer that might NOT be &loif -- easier than replicating that code here.
2035  */
2036 void
2037 ip_mloopback(ifp, m, dst)
2038 	struct ifnet *ifp;
2039 	struct mbuf *m;
2040 	struct sockaddr_in *dst;
2041 {
2042 	struct ip *ip;
2043 	struct mbuf *copym;
2044 
2045 	copym = m_copym2(m, 0, M_COPYALL, M_DONTWAIT);
2046 	if (copym != NULL) {
2047 		/*
2048 		 * We don't bother to fragment if the IP length is greater
2049 		 * than the interface's MTU.  Can this possibly matter?
2050 		 */
2051 		ip = mtod(copym, struct ip *);
2052 		ip->ip_sum = 0;
2053 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
2054 		(void) looutput(ifp, copym, sintosa(dst), NULL);
2055 	}
2056 }
2057 
2058 /*
2059  * Process a delayed payload checksum calculation.
2060  */
2061 void
2062 in_delayed_cksum(struct mbuf *m)
2063 {
2064 	struct ip *ip;
2065 	u_int16_t csum, offset;
2066 
2067 	ip = mtod(m, struct ip *);
2068 	offset = ip->ip_hl << 2;
2069 	csum = in4_cksum(m, 0, offset, m->m_pkthdr.len - offset);
2070 	if (csum == 0 && ip->ip_p == IPPROTO_UDP)
2071 		csum = 0xffff;
2072 
2073 	switch (ip->ip_p) {
2074 	case IPPROTO_TCP:
2075 		offset += offsetof(struct tcphdr, th_sum);
2076 		break;
2077 
2078 	case IPPROTO_UDP:
2079 		offset += offsetof(struct udphdr, uh_sum);
2080 		break;
2081 
2082 	default:
2083 		return;
2084 	}
2085 
2086 	if ((offset + sizeof(u_int16_t)) > m->m_len)
2087 		m_copyback(m, offset, sizeof(csum), &csum);
2088 	else
2089 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2090 }
2091