xref: /openbsd-src/sys/netinet/ip_output.c (revision 4c1e55dc91edd6e69ccc60ce855900fbc12cf34f)
1 /*	$OpenBSD: ip_output.c,v 1.229 2012/04/13 09:38:32 deraadt 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 #if NPF > 0
125 	struct ifnet *encif;
126 #endif
127 #endif /* IPSEC */
128 
129 	va_start(ap, m0);
130 	opt = va_arg(ap, struct mbuf *);
131 	ro = va_arg(ap, struct route *);
132 	flags = va_arg(ap, int);
133 	imo = va_arg(ap, struct ip_moptions *);
134 #ifdef IPSEC
135 	inp = va_arg(ap, struct inpcb *);
136 	if (inp && (inp->inp_flags & INP_IPV6) != 0)
137 		panic("ip_output: IPv6 pcb is passed");
138 #endif /* IPSEC */
139 	va_end(ap);
140 
141 #ifdef	DIAGNOSTIC
142 	if ((m->m_flags & M_PKTHDR) == 0)
143 		panic("ip_output no HDR");
144 #endif
145 	if (opt) {
146 		m = ip_insertoptions(m, opt, &len);
147 		hlen = len;
148 	}
149 
150 	ip = mtod(m, struct ip *);
151 
152 	/*
153 	 * Fill in IP header.
154 	 */
155 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
156 		ip->ip_v = IPVERSION;
157 		ip->ip_off &= htons(IP_DF);
158 		ip->ip_id = htons(ip_randomid());
159 		ip->ip_hl = hlen >> 2;
160 		ipstat.ips_localout++;
161 	} else {
162 		hlen = ip->ip_hl << 2;
163 	}
164 
165 	/*
166 	 * We should not send traffic to 0/8 say both Stevens and RFCs
167 	 * 5735 section 3 and 1122 sections 3.2.1.3 and 3.3.6.
168 	 */
169 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == 0) {
170 		error = ENETUNREACH;
171 		goto bad;
172 	}
173 
174 	/*
175 	 * If we're missing the IP source address, do a route lookup. We'll
176 	 * remember this result, in case we don't need to do any IPsec
177 	 * processing on the packet. We need the source address so we can
178 	 * do an SPD lookup in IPsec; for most packets, the source address
179 	 * is set at a higher level protocol. ICMPs and other packets
180 	 * though (e.g., traceroute) have a source address of zeroes.
181 	 */
182 	if (ip->ip_src.s_addr == INADDR_ANY) {
183 		if (flags & IP_ROUTETOETHER) {
184 			error = EINVAL;
185 			goto bad;
186 		}
187 		donerouting = 1;
188 
189 		if (ro == 0) {
190 			ro = &iproute;
191 			bzero((caddr_t)ro, sizeof (*ro));
192 		}
193 
194 		dst = satosin(&ro->ro_dst);
195 
196 		/*
197 		 * If there is a cached route, check that it is to the same
198 		 * destination and is still up.  If not, free it and try again.
199 		 */
200 		if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
201 		    dst->sin_addr.s_addr != ip->ip_dst.s_addr ||
202 		    ro->ro_tableid != m->m_pkthdr.rdomain)) {
203 			RTFREE(ro->ro_rt);
204 			ro->ro_rt = (struct rtentry *)0;
205 		}
206 
207 		if (ro->ro_rt == 0) {
208 			dst->sin_family = AF_INET;
209 			dst->sin_len = sizeof(*dst);
210 			dst->sin_addr = ip->ip_dst;
211 			ro->ro_tableid = m->m_pkthdr.rdomain;
212 		}
213 
214 		/*
215 		 * If routing to interface only, short-circuit routing lookup.
216 		 */
217 		if (flags & IP_ROUTETOIF) {
218 			if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
219 			    m->m_pkthdr.rdomain))) == 0 &&
220 			    (ia = ifatoia(ifa_ifwithnet(sintosa(dst),
221 			    m->m_pkthdr.rdomain))) == 0) {
222 				ipstat.ips_noroute++;
223 				error = ENETUNREACH;
224 				goto bad;
225 			}
226 
227 			ifp = ia->ia_ifp;
228 			mtu = ifp->if_mtu;
229 			ip->ip_ttl = 1;
230 		} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
231 		    (ip->ip_dst.s_addr == INADDR_BROADCAST)) &&
232 		    imo != NULL && imo->imo_multicast_ifp != NULL) {
233 			ifp = imo->imo_multicast_ifp;
234 			mtu = ifp->if_mtu;
235 			IFP_TO_IA(ifp, ia);
236 		} else {
237 			if (ro->ro_rt == 0)
238 				rtalloc_mpath(ro, NULL);
239 
240 			if (ro->ro_rt == 0) {
241 				ipstat.ips_noroute++;
242 				error = EHOSTUNREACH;
243 				goto bad;
244 			}
245 
246 			ia = ifatoia(ro->ro_rt->rt_ifa);
247 			ifp = ro->ro_rt->rt_ifp;
248 			if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
249 				mtu = ifp->if_mtu;
250 			ro->ro_rt->rt_use++;
251 
252 			if (ro->ro_rt->rt_flags & RTF_GATEWAY)
253 				dst = satosin(ro->ro_rt->rt_gateway);
254 		}
255 
256 		/* Set the source IP address */
257 		if (!IN_MULTICAST(ip->ip_dst.s_addr))
258 			ip->ip_src = ia->ia_addr.sin_addr;
259 	}
260 
261 #if NPF > 0
262 reroute:
263 #endif
264 
265 #ifdef IPSEC
266 	if (!ipsec_in_use && inp == NULL)
267 		goto done_spd;
268 
269 	/*
270 	 * splnet is chosen over spltdb because we are not allowed to
271 	 * lower the level, and udp_output calls us in splnet().
272 	 */
273 	s = splnet();
274 
275 	/* Do we have any pending SAs to apply ? */
276 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
277 	if (mtag != NULL) {
278 #ifdef DIAGNOSTIC
279 		if (mtag->m_tag_len != sizeof (struct tdb_ident))
280 			panic("ip_output: tag of length %d (should be %d",
281 			    mtag->m_tag_len, sizeof (struct tdb_ident));
282 #endif
283 		tdbi = (struct tdb_ident *)(mtag + 1);
284 		tdb = gettdb(tdbi->rdomain,
285 		    tdbi->spi, &tdbi->dst, tdbi->proto);
286 		if (tdb == NULL)
287 			error = -EINVAL;
288 		m_tag_delete(m, mtag);
289 	}
290 	else
291 		tdb = ipsp_spd_lookup(m, AF_INET, hlen, &error,
292 		    IPSP_DIRECTION_OUT, NULL, inp);
293 
294 	if (tdb == NULL) {
295 		splx(s);
296 
297 		if (error == 0) {
298 			/*
299 			 * No IPsec processing required, we'll just send the
300 			 * packet out.
301 			 */
302 			sproto = 0;
303 
304 			/* Fall through to routing/multicast handling */
305 		} else {
306 			/*
307 			 * -EINVAL is used to indicate that the packet should
308 			 * be silently dropped, typically because we've asked
309 			 * key management for an SA.
310 			 */
311 			if (error == -EINVAL) /* Should silently drop packet */
312 			  error = 0;
313 
314 			m_freem(m);
315 			goto done;
316 		}
317 	} else {
318 		/* Loop detection */
319 		for (mtag = m_tag_first(m); mtag != NULL;
320 		    mtag = m_tag_next(m, mtag)) {
321 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
322 			    mtag->m_tag_id !=
323 			    PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
324 				continue;
325 			tdbi = (struct tdb_ident *)(mtag + 1);
326 			if (tdbi->spi == tdb->tdb_spi &&
327 			    tdbi->proto == tdb->tdb_sproto &&
328 			    tdbi->rdomain == tdb->tdb_rdomain &&
329 			    !bcmp(&tdbi->dst, &tdb->tdb_dst,
330 			    sizeof(union sockaddr_union))) {
331 				splx(s);
332 				sproto = 0; /* mark as no-IPsec-needed */
333 				goto done_spd;
334 			}
335 		}
336 
337 		/* We need to do IPsec */
338 		bcopy(&tdb->tdb_dst, &sdst, sizeof(sdst));
339 		sspi = tdb->tdb_spi;
340 		sproto = tdb->tdb_sproto;
341 		splx(s);
342 
343 		/*
344 		 * If it needs TCP/UDP hardware-checksumming, do the
345 		 * computation now.
346 		 */
347 		in_proto_cksum_out(m, NULL);
348 
349 		/* If it's not a multicast packet, try to fast-path */
350 		if (!IN_MULTICAST(ip->ip_dst.s_addr)) {
351 			goto sendit;
352 		}
353 	}
354 
355 	/* Fall through to the routing/multicast handling code */
356  done_spd:
357 #endif /* IPSEC */
358 
359 	if (flags & IP_ROUTETOETHER) {
360 		dst = satosin(&ro->ro_dst);
361 		ifp = ro->ro_rt->rt_ifp;
362 		mtu = ifp->if_mtu;
363 		ro->ro_rt = NULL;
364 	} else if (donerouting == 0) {
365 		if (ro == 0) {
366 			ro = &iproute;
367 			bzero((caddr_t)ro, sizeof (*ro));
368 		}
369 
370 		dst = satosin(&ro->ro_dst);
371 
372 		/*
373 		 * If there is a cached route, check that it is to the same
374 		 * destination and is still up.  If not, free it and try again.
375 		 */
376 		if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
377 		    dst->sin_addr.s_addr != ip->ip_dst.s_addr ||
378 		    ro->ro_tableid != m->m_pkthdr.rdomain)) {
379 			RTFREE(ro->ro_rt);
380 			ro->ro_rt = (struct rtentry *)0;
381 		}
382 
383 		if (ro->ro_rt == 0) {
384 			dst->sin_family = AF_INET;
385 			dst->sin_len = sizeof(*dst);
386 			dst->sin_addr = ip->ip_dst;
387 			ro->ro_tableid = m->m_pkthdr.rdomain;
388 		}
389 
390 		/*
391 		 * If routing to interface only, short-circuit routing lookup.
392 		 */
393 		if (flags & IP_ROUTETOIF) {
394 			if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
395 			    m->m_pkthdr.rdomain))) == 0 &&
396 			    (ia = ifatoia(ifa_ifwithnet(sintosa(dst),
397 			    m->m_pkthdr.rdomain))) == 0) {
398 				ipstat.ips_noroute++;
399 				error = ENETUNREACH;
400 				goto bad;
401 			}
402 
403 			ifp = ia->ia_ifp;
404 			mtu = ifp->if_mtu;
405 			ip->ip_ttl = 1;
406 		} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
407 		    (ip->ip_dst.s_addr == INADDR_BROADCAST)) &&
408 		    imo != NULL && imo->imo_multicast_ifp != NULL) {
409 			ifp = imo->imo_multicast_ifp;
410 			mtu = ifp->if_mtu;
411 			IFP_TO_IA(ifp, ia);
412 		} else {
413 			if (ro->ro_rt == 0)
414 				rtalloc_mpath(ro, &ip->ip_src.s_addr);
415 
416 			if (ro->ro_rt == 0) {
417 				ipstat.ips_noroute++;
418 				error = EHOSTUNREACH;
419 				goto bad;
420 			}
421 
422 			ia = ifatoia(ro->ro_rt->rt_ifa);
423 			ifp = ro->ro_rt->rt_ifp;
424 			if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
425 				mtu = ifp->if_mtu;
426 			ro->ro_rt->rt_use++;
427 
428 			if (ro->ro_rt->rt_flags & RTF_GATEWAY)
429 				dst = satosin(ro->ro_rt->rt_gateway);
430 		}
431 
432 		/* Set the source IP address */
433 		if (ip->ip_src.s_addr == INADDR_ANY)
434 			ip->ip_src = ia->ia_addr.sin_addr;
435 	}
436 
437 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
438 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
439 		struct in_multi *inm;
440 
441 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
442 			M_BCAST : M_MCAST;
443 
444 		/*
445 		 * IP destination address is multicast.  Make sure "dst"
446 		 * still points to the address in "ro".  (It may have been
447 		 * changed to point to a gateway address, above.)
448 		 */
449 		dst = satosin(&ro->ro_dst);
450 
451 		/*
452 		 * See if the caller provided any multicast options
453 		 */
454 		if (imo != NULL)
455 			ip->ip_ttl = imo->imo_multicast_ttl;
456 		else
457 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
458 
459 		/*
460 		 * if we don't know the outgoing ifp yet, we can't generate
461 		 * output
462 		 */
463 		if (!ifp) {
464 			ipstat.ips_noroute++;
465 			error = EHOSTUNREACH;
466 			goto bad;
467 		}
468 
469 		/*
470 		 * Confirm that the outgoing interface supports multicast,
471 		 * but only if the packet actually is going out on that
472 		 * interface (i.e., no IPsec is applied).
473 		 */
474 		if ((((m->m_flags & M_MCAST) &&
475 		      (ifp->if_flags & IFF_MULTICAST) == 0) ||
476 		     ((m->m_flags & M_BCAST) &&
477 		      (ifp->if_flags & IFF_BROADCAST) == 0)) && (sproto == 0)) {
478 			ipstat.ips_noroute++;
479 			error = ENETUNREACH;
480 			goto bad;
481 		}
482 
483 		/*
484 		 * If source address not specified yet, use address
485 		 * of outgoing interface.
486 		 */
487 		if (ip->ip_src.s_addr == INADDR_ANY) {
488 			struct in_ifaddr *ia;
489 
490 			TAILQ_FOREACH(ia, &in_ifaddr, ia_list)
491 				if (ia->ia_ifp == ifp) {
492 					ip->ip_src = ia->ia_addr.sin_addr;
493 					break;
494 				}
495 		}
496 
497 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
498 		if (inm != NULL &&
499 		   (imo == NULL || imo->imo_multicast_loop)) {
500 			/*
501 			 * If we belong to the destination multicast group
502 			 * on the outgoing interface, and the caller did not
503 			 * forbid loopback, loop back a copy.
504 			 * Can't defer TCP/UDP checksumming, do the
505 			 * computation now.
506 			 */
507 			in_proto_cksum_out(m, NULL);
508 			ip_mloopback(ifp, m, dst);
509 		}
510 #ifdef MROUTING
511 		else {
512 			/*
513 			 * If we are acting as a multicast router, perform
514 			 * multicast forwarding as if the packet had just
515 			 * arrived on the interface to which we are about
516 			 * to send.  The multicast forwarding function
517 			 * recursively calls this function, using the
518 			 * IP_FORWARDING flag to prevent infinite recursion.
519 			 *
520 			 * Multicasts that are looped back by ip_mloopback(),
521 			 * above, will be forwarded by the ip_input() routine,
522 			 * if necessary.
523 			 */
524 			extern struct socket *ip_mrouter;
525 
526 			if (ipmforwarding && ip_mrouter &&
527 			    (flags & IP_FORWARDING) == 0) {
528 				if (ip_mforward(m, ifp) != 0) {
529 					m_freem(m);
530 					goto done;
531 				}
532 			}
533 		}
534 #endif
535 		/*
536 		 * Multicasts with a time-to-live of zero may be looped-
537 		 * back, above, but must not be transmitted on a network.
538 		 * Also, multicasts addressed to the loopback interface
539 		 * are not sent -- the above call to ip_mloopback() will
540 		 * loop back a copy if this host actually belongs to the
541 		 * destination group on the loopback interface.
542 		 */
543 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
544 			m_freem(m);
545 			goto done;
546 		}
547 
548 		goto sendit;
549 	}
550 
551 	/*
552 	 * Look for broadcast address and verify user is allowed to send
553 	 * such a packet; if the packet is going in an IPsec tunnel, skip
554 	 * this check.
555 	 */
556 	if ((sproto == 0) && (in_broadcast(dst->sin_addr, ifp,
557 	    m->m_pkthdr.rdomain))) {
558 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
559 			error = EADDRNOTAVAIL;
560 			goto bad;
561 		}
562 		if ((flags & IP_ALLOWBROADCAST) == 0) {
563 			error = EACCES;
564 			goto bad;
565 		}
566 
567 		/* Don't allow broadcast messages to be fragmented */
568 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
569 			error = EMSGSIZE;
570 			goto bad;
571 		}
572 		m->m_flags |= M_BCAST;
573 	} else
574 		m->m_flags &= ~M_BCAST;
575 
576 sendit:
577 	/*
578 	 * If we're doing Path MTU discovery, we need to set DF unless
579 	 * the route's MTU is locked.
580 	 */
581 	if ((flags & IP_MTUDISC) && ro && ro->ro_rt &&
582 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
583 		ip->ip_off |= htons(IP_DF);
584 
585 #ifdef IPSEC
586 	/*
587 	 * Check if the packet needs encapsulation.
588 	 */
589 	if (sproto != 0) {
590 		s = splnet();
591 
592 		tdb = gettdb(rtable_l2(m->m_pkthdr.rdomain),
593 		    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 		/*
603 		 * Packet filter
604 		 */
605 #if NPF > 0
606 		if ((encif = enc_getif(tdb->tdb_rdomain,
607 		    tdb->tdb_tap)) == NULL ||
608 		    pf_test(AF_INET, PF_OUT, encif, &m, NULL) != PF_PASS) {
609 			error = EACCES;
610 			splx(s);
611 			m_freem(m);
612 			goto done;
613 		}
614 		if (m == NULL) {
615 			splx(s);
616 			goto done;
617 		}
618 		ip = mtod(m, struct ip *);
619 		hlen = ip->ip_hl << 2;
620 		/*
621 		 * PF_TAG_REROUTE handling or not...
622 		 * Packet is entering IPsec so the routing is
623 		 * already overruled by the IPsec policy.
624 		 * Until now the change was not reconsidered.
625 		 * What's the behaviour?
626 		 */
627 #endif
628 
629 		/* Check if we are allowed to fragment */
630 		if (ip_mtudisc && (ip->ip_off & htons(IP_DF)) && tdb->tdb_mtu &&
631 		    ntohs(ip->ip_len) > tdb->tdb_mtu &&
632 		    tdb->tdb_mtutimeout > time_second) {
633 			struct rtentry *rt = NULL;
634 			int rt_mtucloned = 0;
635 			int transportmode = 0;
636 
637 			transportmode = (tdb->tdb_dst.sa.sa_family == AF_INET) &&
638 			    (tdb->tdb_dst.sin.sin_addr.s_addr ==
639 			    ip->ip_dst.s_addr);
640 			icmp_mtu = tdb->tdb_mtu;
641 			splx(s);
642 
643 			/* Find a host route to store the mtu in */
644 			if (ro != NULL)
645 				rt = ro->ro_rt;
646 			/* but don't add a PMTU route for transport mode SAs */
647 			if (transportmode)
648 				rt = NULL;
649 			else if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0) {
650 				struct sockaddr_in dst = {
651 					sizeof(struct sockaddr_in), AF_INET};
652 				dst.sin_addr = ip->ip_dst;
653 				rt = icmp_mtudisc_clone((struct sockaddr *)&dst,
654 				    m->m_pkthdr.rdomain);
655 				rt_mtucloned = 1;
656 			}
657 			DPRINTF(("ip_output: spi %08x mtu %d rt %p cloned %d\n",
658 			    ntohl(tdb->tdb_spi), icmp_mtu, rt, rt_mtucloned));
659 			if (rt != NULL) {
660 				rt->rt_rmx.rmx_mtu = icmp_mtu;
661 				if (ro && ro->ro_rt != NULL) {
662 					RTFREE(ro->ro_rt);
663 					ro->ro_rt = rtalloc1(&ro->ro_dst, RT_REPORT,
664 					    m->m_pkthdr.rdomain);
665 				}
666 				if (rt_mtucloned)
667 					rtfree(rt);
668 			}
669 			error = EMSGSIZE;
670 			goto bad;
671 		}
672 
673 		/*
674 		 * Clear these -- they'll be set in the recursive invocation
675 		 * as needed.
676 		 */
677 		m->m_flags &= ~(M_MCAST | M_BCAST);
678 
679 		/* Callee frees mbuf */
680 		error = ipsp_process_packet(m, tdb, AF_INET, 0);
681 		splx(s);
682 		return error;  /* Nothing more to be done */
683 	}
684 
685 	/*
686 	 * If we got here and IPsec crypto processing didn't happen, drop it.
687 	 */
688 	if (ipsec_in_use && (mtag = m_tag_find(m,
689 	    PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL)) != NULL) {
690 		/* Notify IPsec to do its own crypto. */
691 		ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
692 		m_freem(m);
693 		error = EHOSTUNREACH;
694 		goto done;
695 	}
696 #endif /* IPSEC */
697 
698 	in_proto_cksum_out(m, ifp);
699 
700 	/*
701 	 * Packet filter
702 	 */
703 #if NPF > 0
704 	if (pf_test(AF_INET, PF_OUT, ifp, &m, NULL) != PF_PASS) {
705 		error = EHOSTUNREACH;
706 		m_freem(m);
707 		goto done;
708 	}
709 	if (m == NULL)
710 		goto done;
711 	ip = mtod(m, struct ip *);
712 	hlen = ip->ip_hl << 2;
713 	if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
714 	    (PF_TAG_REROUTE | PF_TAG_GENERATED))
715 		/* already rerun the route lookup, go on */
716 		m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
717 	else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
718 		/* tag as generated to skip over pf_test on rerun */
719 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
720 		ro = NULL;
721 		donerouting = 0;
722 		goto reroute;
723 	}
724 #endif
725 
726 #ifdef IPSEC
727 	if (ipsec_in_use && (flags & IP_FORWARDING) && (ipforwarding == 2) &&
728 	    (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) == NULL)) {
729 		error = EHOSTUNREACH;
730 		m_freem(m);
731 		goto done;
732 	}
733 #endif
734 
735 	/*
736 	 * If small enough for interface, can just send directly.
737 	 */
738 	if (ntohs(ip->ip_len) <= mtu) {
739 		ip->ip_sum = 0;
740 		if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
741 		    (ifp->if_bridge == NULL)) {
742 			m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
743 			ipstat.ips_outhwcsum++;
744 		} else
745 			ip->ip_sum = in_cksum(m, hlen);
746 		/* Update relevant hardware checksum stats for TCP/UDP */
747 		if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT)
748 			tcpstat.tcps_outhwcsum++;
749 		else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT)
750 			udpstat.udps_outhwcsum++;
751 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
752 		goto done;
753 	}
754 
755 	/*
756 	 * Too large for interface; fragment if possible.
757 	 * Must be able to put at least 8 bytes per fragment.
758 	 */
759 	if (ip->ip_off & htons(IP_DF)) {
760 #ifdef IPSEC
761 		icmp_mtu = ifp->if_mtu;
762 #endif
763 		error = EMSGSIZE;
764 		/*
765 		 * This case can happen if the user changed the MTU
766 		 * of an interface after enabling IP on it.  Because
767 		 * most netifs don't keep track of routes pointing to
768 		 * them, there is no way for one to update all its
769 		 * routes when the MTU is changed.
770 		 */
771 		if (ro->ro_rt != NULL &&
772 		    (ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
773 		    !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
774 		    (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
775 			ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
776 		}
777 		ipstat.ips_cantfrag++;
778 		goto bad;
779 	}
780 
781 	error = ip_fragment(m, ifp, mtu);
782 	if (error) {
783 		m = m0 = NULL;
784 		goto bad;
785 	}
786 
787 	for (; m; m = m0) {
788 		m0 = m->m_nextpkt;
789 		m->m_nextpkt = 0;
790 		if (error == 0)
791 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
792 			    ro->ro_rt);
793 		else
794 			m_freem(m);
795 	}
796 
797 	if (error == 0)
798 		ipstat.ips_fragmented++;
799 
800 done:
801 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt)
802 		RTFREE(ro->ro_rt);
803 	return (error);
804 bad:
805 #ifdef IPSEC
806 	if (error == EMSGSIZE && ip_mtudisc && icmp_mtu != 0 && m != NULL)
807 		ipsec_adjust_mtu(m, icmp_mtu);
808 #endif
809 	m_freem(m0);
810 	goto done;
811 }
812 
813 int
814 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
815 {
816 	struct ip *ip, *mhip;
817 	struct mbuf *m0;
818 	int len, hlen, off;
819 	int mhlen, firstlen;
820 	struct mbuf **mnext;
821 	int fragments = 0;
822 	int error = 0;
823 
824 	ip = mtod(m, struct ip *);
825 	hlen = ip->ip_hl << 2;
826 
827 	len = (mtu - hlen) &~ 7;
828 	if (len < 8) {
829 		m_freem(m);
830 		return (EMSGSIZE);
831 	}
832 
833 	/*
834 	 * If we are doing fragmentation, we can't defer TCP/UDP
835 	 * checksumming; compute the checksum and clear the flag.
836 	 */
837 	in_proto_cksum_out(m, NULL);
838 	firstlen = len;
839 	mnext = &m->m_nextpkt;
840 
841 	/*
842 	 * Loop through length of segment after first fragment,
843 	 * make new header and copy data of each part and link onto chain.
844 	 */
845 	m0 = m;
846 	mhlen = sizeof (struct ip);
847 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
848 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
849 		if (m == 0) {
850 			ipstat.ips_odropped++;
851 			error = ENOBUFS;
852 			goto sendorfree;
853 		}
854 		*mnext = m;
855 		mnext = &m->m_nextpkt;
856 		m->m_data += max_linkhdr;
857 		mhip = mtod(m, struct ip *);
858 		*mhip = *ip;
859 		/* we must inherit MCAST and BCAST flags and rdomain */
860 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
861 		m->m_pkthdr.rdomain = m0->m_pkthdr.rdomain;
862 		if (hlen > sizeof (struct ip)) {
863 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
864 			mhip->ip_hl = mhlen >> 2;
865 		}
866 		m->m_len = mhlen;
867 		mhip->ip_off = ((off - hlen) >> 3) +
868 		    (ntohs(ip->ip_off) & ~IP_MF);
869 		if (ip->ip_off & htons(IP_MF))
870 			mhip->ip_off |= IP_MF;
871 		if (off + len >= ntohs(ip->ip_len))
872 			len = ntohs(ip->ip_len) - off;
873 		else
874 			mhip->ip_off |= IP_MF;
875 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
876 		m->m_next = m_copy(m0, off, len);
877 		if (m->m_next == 0) {
878 			ipstat.ips_odropped++;
879 			error = ENOBUFS;
880 			goto sendorfree;
881 		}
882 		m->m_pkthdr.len = mhlen + len;
883 		m->m_pkthdr.rcvif = (struct ifnet *)0;
884 		mhip->ip_off = htons((u_int16_t)mhip->ip_off);
885 		mhip->ip_sum = 0;
886 		if ((ifp != NULL) &&
887 		    (ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
888 		    (ifp->if_bridge == NULL)) {
889 			m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
890 			ipstat.ips_outhwcsum++;
891 		} else
892 			mhip->ip_sum = in_cksum(m, mhlen);
893 		ipstat.ips_ofragments++;
894 		fragments++;
895 	}
896 	/*
897 	 * Update first fragment by trimming what's been copied out
898 	 * and updating header, then send each fragment (in order).
899 	 */
900 	m = m0;
901 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
902 	m->m_pkthdr.len = hlen + firstlen;
903 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
904 	ip->ip_off |= htons(IP_MF);
905 	ip->ip_sum = 0;
906 	if ((ifp != NULL) &&
907 	    (ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
908 	    (ifp->if_bridge == NULL)) {
909 		m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
910 		ipstat.ips_outhwcsum++;
911 	} else
912 		ip->ip_sum = in_cksum(m, hlen);
913 sendorfree:
914 	if (error) {
915 		for (m = m0; m; m = m0) {
916 			m0 = m->m_nextpkt;
917 			m->m_nextpkt = NULL;
918 			m_freem(m);
919 		}
920 	}
921 
922 	return (error);
923 }
924 
925 /*
926  * Insert IP options into preformed packet.
927  * Adjust IP destination as required for IP source routing,
928  * as indicated by a non-zero in_addr at the start of the options.
929  */
930 struct mbuf *
931 ip_insertoptions(m, opt, phlen)
932 	struct mbuf *m;
933 	struct mbuf *opt;
934 	int *phlen;
935 {
936 	struct ipoption *p = mtod(opt, struct ipoption *);
937 	struct mbuf *n;
938 	struct ip *ip = mtod(m, struct ip *);
939 	unsigned int optlen;
940 
941 	optlen = opt->m_len - sizeof(p->ipopt_dst);
942 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
943 		return (m);		/* XXX should fail */
944 	if (p->ipopt_dst.s_addr)
945 		ip->ip_dst = p->ipopt_dst;
946 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
947 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
948 		if (n == 0)
949 			return (m);
950 		M_MOVE_HDR(n, m);
951 		n->m_pkthdr.len += optlen;
952 		m->m_len -= sizeof(struct ip);
953 		m->m_data += sizeof(struct ip);
954 		n->m_next = m;
955 		m = n;
956 		m->m_len = optlen + sizeof(struct ip);
957 		m->m_data += max_linkhdr;
958 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
959 	} else {
960 		m->m_data -= optlen;
961 		m->m_len += optlen;
962 		m->m_pkthdr.len += optlen;
963 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
964 	}
965 	ip = mtod(m, struct ip *);
966 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), optlen);
967 	*phlen = sizeof(struct ip) + optlen;
968 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
969 	return (m);
970 }
971 
972 /*
973  * Copy options from ip to jp,
974  * omitting those not copied during fragmentation.
975  */
976 int
977 ip_optcopy(ip, jp)
978 	struct ip *ip, *jp;
979 {
980 	u_char *cp, *dp;
981 	int opt, optlen, cnt;
982 
983 	cp = (u_char *)(ip + 1);
984 	dp = (u_char *)(jp + 1);
985 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
986 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
987 		opt = cp[0];
988 		if (opt == IPOPT_EOL)
989 			break;
990 		if (opt == IPOPT_NOP) {
991 			/* Preserve for IP mcast tunnel's LSRR alignment. */
992 			*dp++ = IPOPT_NOP;
993 			optlen = 1;
994 			continue;
995 		}
996 #ifdef DIAGNOSTIC
997 		if (cnt < IPOPT_OLEN + sizeof(*cp))
998 			panic("malformed IPv4 option passed to ip_optcopy");
999 #endif
1000 		optlen = cp[IPOPT_OLEN];
1001 #ifdef DIAGNOSTIC
1002 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1003 			panic("malformed IPv4 option passed to ip_optcopy");
1004 #endif
1005 		/* bogus lengths should have been caught by ip_dooptions */
1006 		if (optlen > cnt)
1007 			optlen = cnt;
1008 		if (IPOPT_COPIED(opt)) {
1009 			bcopy((caddr_t)cp, (caddr_t)dp, optlen);
1010 			dp += optlen;
1011 		}
1012 	}
1013 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1014 		*dp++ = IPOPT_EOL;
1015 	return (optlen);
1016 }
1017 
1018 /*
1019  * IP socket option processing.
1020  */
1021 int
1022 ip_ctloutput(op, so, level, optname, mp)
1023 	int op;
1024 	struct socket *so;
1025 	int level, optname;
1026 	struct mbuf **mp;
1027 {
1028 	struct inpcb *inp = sotoinpcb(so);
1029 	struct mbuf *m = *mp;
1030 	int optval = 0;
1031 	struct proc *p = curproc; /* XXX */
1032 #ifdef IPSEC
1033 	struct ipsec_ref *ipr;
1034 	u_int16_t opt16val;
1035 #endif
1036 	int error = 0;
1037 	u_int rtid = 0;
1038 
1039 	if (level != IPPROTO_IP) {
1040 		error = EINVAL;
1041 		if (op == PRCO_SETOPT && *mp)
1042 			(void) m_free(*mp);
1043 	} else switch (op) {
1044 	case PRCO_SETOPT:
1045 		switch (optname) {
1046 		case IP_OPTIONS:
1047 #ifdef notyet
1048 		case IP_RETOPTS:
1049 			return (ip_pcbopts(optname, &inp->inp_options, m));
1050 #else
1051 			return (ip_pcbopts(&inp->inp_options, m));
1052 #endif
1053 
1054 		case IP_TOS:
1055 		case IP_TTL:
1056 		case IP_MINTTL:
1057 		case IP_RECVOPTS:
1058 		case IP_RECVRETOPTS:
1059 		case IP_RECVDSTADDR:
1060 		case IP_RECVIF:
1061 		case IP_RECVTTL:
1062 		case IP_RECVDSTPORT:
1063 		case IP_RECVRTABLE:
1064 			if (m == NULL || m->m_len != sizeof(int))
1065 				error = EINVAL;
1066 			else {
1067 				optval = *mtod(m, int *);
1068 				switch (optname) {
1069 
1070 				case IP_TOS:
1071 					inp->inp_ip.ip_tos = optval;
1072 					break;
1073 
1074 				case IP_TTL:
1075 					if (optval > 0 && optval <= MAXTTL)
1076 						inp->inp_ip.ip_ttl = optval;
1077 					else
1078 						error = EINVAL;
1079 					break;
1080 
1081 				case IP_MINTTL:
1082 					if (optval > 0 && optval <= MAXTTL)
1083 						inp->inp_ip_minttl = optval;
1084 					else
1085 						error = EINVAL;
1086 					break;
1087 #define	OPTSET(bit) \
1088 	if (optval) \
1089 		inp->inp_flags |= bit; \
1090 	else \
1091 		inp->inp_flags &= ~bit;
1092 
1093 				case IP_RECVOPTS:
1094 					OPTSET(INP_RECVOPTS);
1095 					break;
1096 
1097 				case IP_RECVRETOPTS:
1098 					OPTSET(INP_RECVRETOPTS);
1099 					break;
1100 
1101 				case IP_RECVDSTADDR:
1102 					OPTSET(INP_RECVDSTADDR);
1103 					break;
1104 				case IP_RECVIF:
1105 					OPTSET(INP_RECVIF);
1106 					break;
1107 				case IP_RECVTTL:
1108 					OPTSET(INP_RECVTTL);
1109 					break;
1110 				case IP_RECVDSTPORT:
1111 					OPTSET(INP_RECVDSTPORT);
1112 					break;
1113 				case IP_RECVRTABLE:
1114 					OPTSET(INP_RECVRTABLE);
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 			    inp->inp_rtableid);
1128 			break;
1129 
1130 		case IP_PORTRANGE:
1131 			if (m == 0 || m->m_len != sizeof(int))
1132 				error = EINVAL;
1133 			else {
1134 				optval = *mtod(m, int *);
1135 
1136 				switch (optval) {
1137 
1138 				case IP_PORTRANGE_DEFAULT:
1139 					inp->inp_flags &= ~(INP_LOWPORT);
1140 					inp->inp_flags &= ~(INP_HIGHPORT);
1141 					break;
1142 
1143 				case IP_PORTRANGE_HIGH:
1144 					inp->inp_flags &= ~(INP_LOWPORT);
1145 					inp->inp_flags |= INP_HIGHPORT;
1146 					break;
1147 
1148 				case IP_PORTRANGE_LOW:
1149 					inp->inp_flags &= ~(INP_HIGHPORT);
1150 					inp->inp_flags |= INP_LOWPORT;
1151 					break;
1152 
1153 				default:
1154 
1155 					error = EINVAL;
1156 					break;
1157 				}
1158 			}
1159 			break;
1160 		case IP_AUTH_LEVEL:
1161 		case IP_ESP_TRANS_LEVEL:
1162 		case IP_ESP_NETWORK_LEVEL:
1163 		case IP_IPCOMP_LEVEL:
1164 #ifndef IPSEC
1165 			error = EOPNOTSUPP;
1166 #else
1167 			if (m == 0 || m->m_len != sizeof(int)) {
1168 				error = EINVAL;
1169 				break;
1170 			}
1171 			optval = *mtod(m, int *);
1172 
1173 			if (optval < IPSEC_LEVEL_BYPASS ||
1174 			    optval > IPSEC_LEVEL_UNIQUE) {
1175 				error = EINVAL;
1176 				break;
1177 			}
1178 
1179 			/* Unlink cached output TDB to force a re-search */
1180 			if (inp->inp_tdb_out) {
1181 				int s = spltdb();
1182 				TAILQ_REMOVE(&inp->inp_tdb_out->tdb_inp_out,
1183 				    inp, inp_tdb_out_next);
1184 				splx(s);
1185 			}
1186 
1187 			if (inp->inp_tdb_in) {
1188 				int s = spltdb();
1189 				TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in,
1190 				    inp, inp_tdb_in_next);
1191 				splx(s);
1192 			}
1193 
1194 			switch (optname) {
1195 			case IP_AUTH_LEVEL:
1196 				if (optval < ipsec_auth_default_level &&
1197 				    suser(p, 0)) {
1198 					error = EACCES;
1199 					break;
1200 				}
1201 				inp->inp_seclevel[SL_AUTH] = optval;
1202 				break;
1203 
1204 			case IP_ESP_TRANS_LEVEL:
1205 				if (optval < ipsec_esp_trans_default_level &&
1206 				    suser(p, 0)) {
1207 					error = EACCES;
1208 					break;
1209 				}
1210 				inp->inp_seclevel[SL_ESP_TRANS] = optval;
1211 				break;
1212 
1213 			case IP_ESP_NETWORK_LEVEL:
1214 				if (optval < ipsec_esp_network_default_level &&
1215 				    suser(p, 0)) {
1216 					error = EACCES;
1217 					break;
1218 				}
1219 				inp->inp_seclevel[SL_ESP_NETWORK] = optval;
1220 				break;
1221 			case IP_IPCOMP_LEVEL:
1222 				if (optval < ipsec_ipcomp_default_level &&
1223 				    suser(p, 0)) {
1224 					error = EACCES;
1225 					break;
1226 				}
1227 				inp->inp_seclevel[SL_IPCOMP] = optval;
1228 				break;
1229 			}
1230 			if (!error)
1231 				inp->inp_secrequire = get_sa_require(inp);
1232 #endif
1233 			break;
1234 
1235 		case IP_IPSEC_REMOTE_CRED:
1236 		case IP_IPSEC_REMOTE_AUTH:
1237 			/* Can't set the remote credential or key */
1238 			error = EOPNOTSUPP;
1239 			break;
1240 
1241 		case IP_IPSEC_LOCAL_ID:
1242 		case IP_IPSEC_REMOTE_ID:
1243 		case IP_IPSEC_LOCAL_CRED:
1244 		case IP_IPSEC_LOCAL_AUTH:
1245 #ifndef IPSEC
1246 			error = EOPNOTSUPP;
1247 #else
1248 			if (m == NULL || m->m_len < 2) {
1249 				error = EINVAL;
1250 				break;
1251 			}
1252 
1253 			m_copydata(m, 0, 2, (caddr_t) &opt16val);
1254 
1255 			/* If the type is 0, then we cleanup and return */
1256 			if (opt16val == 0) {
1257 				switch (optname) {
1258 				case IP_IPSEC_LOCAL_ID:
1259 					if (inp->inp_ipo != NULL &&
1260 					    inp->inp_ipo->ipo_srcid != NULL) {
1261 						ipsp_reffree(inp->inp_ipo->ipo_srcid);
1262 						inp->inp_ipo->ipo_srcid = NULL;
1263 					}
1264 					break;
1265 
1266 				case IP_IPSEC_REMOTE_ID:
1267 					if (inp->inp_ipo != NULL &&
1268 					    inp->inp_ipo->ipo_dstid != NULL) {
1269 						ipsp_reffree(inp->inp_ipo->ipo_dstid);
1270 						inp->inp_ipo->ipo_dstid = NULL;
1271 					}
1272 					break;
1273 
1274 				case IP_IPSEC_LOCAL_CRED:
1275 					if (inp->inp_ipo != NULL &&
1276 					    inp->inp_ipo->ipo_local_cred != NULL) {
1277 						ipsp_reffree(inp->inp_ipo->ipo_local_cred);
1278 						inp->inp_ipo->ipo_local_cred = NULL;
1279 					}
1280 					break;
1281 
1282 				case IP_IPSEC_LOCAL_AUTH:
1283 					if (inp->inp_ipo != NULL &&
1284 					    inp->inp_ipo->ipo_local_auth != NULL) {
1285 						ipsp_reffree(inp->inp_ipo->ipo_local_auth);
1286 						inp->inp_ipo->ipo_local_auth = NULL;
1287 					}
1288 					break;
1289 				}
1290 
1291 				error = 0;
1292 				break;
1293 			}
1294 
1295 			/* Can't have an empty payload */
1296 			if (m->m_len == 2) {
1297 				error = EINVAL;
1298 				break;
1299 			}
1300 
1301 			/* Allocate if needed */
1302 			if (inp->inp_ipo == NULL) {
1303 				inp->inp_ipo = ipsec_add_policy(inp,
1304 				    AF_INET, IPSP_DIRECTION_OUT);
1305 				if (inp->inp_ipo == NULL) {
1306 					error = ENOBUFS;
1307 					break;
1308 				}
1309 			}
1310 
1311 			ipr = malloc(sizeof(struct ipsec_ref) + m->m_len - 2,
1312 			       M_CREDENTIALS, M_NOWAIT);
1313 			if (ipr == NULL) {
1314 				error = ENOBUFS;
1315 				break;
1316 			}
1317 
1318 			ipr->ref_count = 1;
1319 			ipr->ref_malloctype = M_CREDENTIALS;
1320 			ipr->ref_len = m->m_len - 2;
1321 			ipr->ref_type = opt16val;
1322 			m_copydata(m, 2, m->m_len - 2, (caddr_t)(ipr + 1));
1323 
1324 			switch (optname) {
1325 			case IP_IPSEC_LOCAL_ID:
1326 				/* Check valid types and NUL-termination */
1327 				if (ipr->ref_type < IPSP_IDENTITY_PREFIX ||
1328 				    ipr->ref_type > IPSP_IDENTITY_CONNECTION ||
1329 				    ((char *)(ipr + 1))[ipr->ref_len - 1]) {
1330 					free(ipr, M_CREDENTIALS);
1331 					error = EINVAL;
1332 				} else {
1333 					if (inp->inp_ipo->ipo_srcid != NULL)
1334 						ipsp_reffree(inp->inp_ipo->ipo_srcid);
1335 					inp->inp_ipo->ipo_srcid = ipr;
1336 				}
1337 				break;
1338 			case IP_IPSEC_REMOTE_ID:
1339 				/* Check valid types and NUL-termination */
1340 				if (ipr->ref_type < IPSP_IDENTITY_PREFIX ||
1341 				    ipr->ref_type > IPSP_IDENTITY_CONNECTION ||
1342 				    ((char *)(ipr + 1))[ipr->ref_len - 1]) {
1343 					free(ipr, M_CREDENTIALS);
1344 					error = EINVAL;
1345 				} else {
1346 					if (inp->inp_ipo->ipo_dstid != NULL)
1347 						ipsp_reffree(inp->inp_ipo->ipo_dstid);
1348 					inp->inp_ipo->ipo_dstid = ipr;
1349 				}
1350 				break;
1351 			case IP_IPSEC_LOCAL_CRED:
1352 				if (ipr->ref_type < IPSP_CRED_KEYNOTE ||
1353 				    ipr->ref_type > IPSP_CRED_X509) {
1354 					free(ipr, M_CREDENTIALS);
1355 					error = EINVAL;
1356 				} else {
1357 					if (inp->inp_ipo->ipo_local_cred != NULL)
1358 						ipsp_reffree(inp->inp_ipo->ipo_local_cred);
1359 					inp->inp_ipo->ipo_local_cred = ipr;
1360 				}
1361 				break;
1362 			case IP_IPSEC_LOCAL_AUTH:
1363 				if (ipr->ref_type < IPSP_AUTH_PASSPHRASE ||
1364 				    ipr->ref_type > IPSP_AUTH_RSA) {
1365 					free(ipr, M_CREDENTIALS);
1366 					error = EINVAL;
1367 				} else {
1368 					if (inp->inp_ipo->ipo_local_auth != NULL)
1369 						ipsp_reffree(inp->inp_ipo->ipo_local_auth);
1370 					inp->inp_ipo->ipo_local_auth = ipr;
1371 				}
1372 				break;
1373 			}
1374 
1375 			/* Unlink cached output TDB to force a re-search */
1376 			if (inp->inp_tdb_out) {
1377 				int s = spltdb();
1378 				TAILQ_REMOVE(&inp->inp_tdb_out->tdb_inp_out,
1379 				    inp, inp_tdb_out_next);
1380 				splx(s);
1381 			}
1382 
1383 			if (inp->inp_tdb_in) {
1384 				int s = spltdb();
1385 				TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in,
1386 				    inp, inp_tdb_in_next);
1387 				splx(s);
1388 			}
1389 #endif
1390 			break;
1391 		case SO_RTABLE:
1392 			if (m == NULL || m->m_len < sizeof(u_int)) {
1393 				error = EINVAL;
1394 				break;
1395 			}
1396 			rtid = *mtod(m, u_int *);
1397 			if (inp->inp_rtableid == rtid)
1398 				break;
1399 			/* needs priviledges to switch when already set */
1400 			if (p->p_p->ps_rtableid != rtid &&
1401 			    p->p_p->ps_rtableid != 0 &&
1402 			    (error = suser(p, 0)) != 0)
1403 				break;
1404 			/* table must exist */
1405 			if (!rtable_exists(rtid)) {
1406 				error = EINVAL;
1407 				break;
1408 			}
1409 			inp->inp_rtableid = rtid;
1410 			break;
1411 		case IP_PIPEX:
1412 			if (m != NULL && m->m_len == sizeof(int))
1413 				inp->inp_pipex = *mtod(m, int *);
1414 			else
1415 				error = EINVAL;
1416 			break;
1417 
1418 		default:
1419 			error = ENOPROTOOPT;
1420 			break;
1421 		}
1422 		if (m)
1423 			(void)m_free(m);
1424 		break;
1425 
1426 	case PRCO_GETOPT:
1427 		switch (optname) {
1428 		case IP_OPTIONS:
1429 		case IP_RETOPTS:
1430 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1431 			if (inp->inp_options) {
1432 				m->m_len = inp->inp_options->m_len;
1433 				bcopy(mtod(inp->inp_options, caddr_t),
1434 				    mtod(m, caddr_t), m->m_len);
1435 			} else
1436 				m->m_len = 0;
1437 			break;
1438 
1439 		case IP_TOS:
1440 		case IP_TTL:
1441 		case IP_MINTTL:
1442 		case IP_RECVOPTS:
1443 		case IP_RECVRETOPTS:
1444 		case IP_RECVDSTADDR:
1445 		case IP_RECVIF:
1446 		case IP_RECVTTL:
1447 		case IP_RECVDSTPORT:
1448 		case IP_RECVRTABLE:
1449 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1450 			m->m_len = sizeof(int);
1451 			switch (optname) {
1452 
1453 			case IP_TOS:
1454 				optval = inp->inp_ip.ip_tos;
1455 				break;
1456 
1457 			case IP_TTL:
1458 				optval = inp->inp_ip.ip_ttl;
1459 				break;
1460 
1461 			case IP_MINTTL:
1462 				optval = inp->inp_ip_minttl;
1463 				break;
1464 
1465 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1466 
1467 			case IP_RECVOPTS:
1468 				optval = OPTBIT(INP_RECVOPTS);
1469 				break;
1470 
1471 			case IP_RECVRETOPTS:
1472 				optval = OPTBIT(INP_RECVRETOPTS);
1473 				break;
1474 
1475 			case IP_RECVDSTADDR:
1476 				optval = OPTBIT(INP_RECVDSTADDR);
1477 				break;
1478 			case IP_RECVIF:
1479 				optval = OPTBIT(INP_RECVIF);
1480 				break;
1481 			case IP_RECVTTL:
1482 				optval = OPTBIT(INP_RECVTTL);
1483 				break;
1484 			case IP_RECVDSTPORT:
1485 				optval = OPTBIT(INP_RECVDSTPORT);
1486 				break;
1487 			case IP_RECVRTABLE:
1488 				optval = OPTBIT(INP_RECVRTABLE);
1489 				break;
1490 			}
1491 			*mtod(m, int *) = optval;
1492 			break;
1493 
1494 		case IP_MULTICAST_IF:
1495 		case IP_MULTICAST_TTL:
1496 		case IP_MULTICAST_LOOP:
1497 		case IP_ADD_MEMBERSHIP:
1498 		case IP_DROP_MEMBERSHIP:
1499 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1500 			break;
1501 
1502 		case IP_PORTRANGE:
1503 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1504 			m->m_len = sizeof(int);
1505 
1506 			if (inp->inp_flags & INP_HIGHPORT)
1507 				optval = IP_PORTRANGE_HIGH;
1508 			else if (inp->inp_flags & INP_LOWPORT)
1509 				optval = IP_PORTRANGE_LOW;
1510 			else
1511 				optval = 0;
1512 
1513 			*mtod(m, int *) = optval;
1514 			break;
1515 
1516 		case IP_AUTH_LEVEL:
1517 		case IP_ESP_TRANS_LEVEL:
1518 		case IP_ESP_NETWORK_LEVEL:
1519 		case IP_IPCOMP_LEVEL:
1520 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1521 #ifndef IPSEC
1522 			m->m_len = sizeof(int);
1523 			*mtod(m, int *) = IPSEC_LEVEL_NONE;
1524 #else
1525 			m->m_len = sizeof(int);
1526 			switch (optname) {
1527 			case IP_AUTH_LEVEL:
1528 				optval = inp->inp_seclevel[SL_AUTH];
1529 				break;
1530 
1531 			case IP_ESP_TRANS_LEVEL:
1532 				optval = inp->inp_seclevel[SL_ESP_TRANS];
1533 				break;
1534 
1535 			case IP_ESP_NETWORK_LEVEL:
1536 				optval = inp->inp_seclevel[SL_ESP_NETWORK];
1537 				break;
1538 			case IP_IPCOMP_LEVEL:
1539 				optval = inp->inp_seclevel[SL_IPCOMP];
1540 				break;
1541 			}
1542 			*mtod(m, int *) = optval;
1543 #endif
1544 			break;
1545 		case IP_IPSEC_LOCAL_ID:
1546 		case IP_IPSEC_REMOTE_ID:
1547 		case IP_IPSEC_LOCAL_CRED:
1548 		case IP_IPSEC_REMOTE_CRED:
1549 		case IP_IPSEC_LOCAL_AUTH:
1550 		case IP_IPSEC_REMOTE_AUTH:
1551 #ifndef IPSEC
1552 			error = EOPNOTSUPP;
1553 #else
1554 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1555 			m->m_len = sizeof(u_int16_t);
1556 			ipr = NULL;
1557 			switch (optname) {
1558 			case IP_IPSEC_LOCAL_ID:
1559 				if (inp->inp_ipo != NULL)
1560 					ipr = inp->inp_ipo->ipo_srcid;
1561 				opt16val = IPSP_IDENTITY_NONE;
1562 				break;
1563 			case IP_IPSEC_REMOTE_ID:
1564 				if (inp->inp_ipo != NULL)
1565 					ipr = inp->inp_ipo->ipo_dstid;
1566 				opt16val = IPSP_IDENTITY_NONE;
1567 				break;
1568 			case IP_IPSEC_LOCAL_CRED:
1569 				if (inp->inp_ipo != NULL)
1570 					ipr = inp->inp_ipo->ipo_local_cred;
1571 				opt16val = IPSP_CRED_NONE;
1572 				break;
1573 			case IP_IPSEC_REMOTE_CRED:
1574 				ipr = inp->inp_ipsec_remotecred;
1575 				opt16val = IPSP_CRED_NONE;
1576 				break;
1577 			case IP_IPSEC_LOCAL_AUTH:
1578 				if (inp->inp_ipo != NULL)
1579 					ipr = inp->inp_ipo->ipo_local_auth;
1580 				opt16val = IPSP_AUTH_NONE;
1581 				break;
1582 			case IP_IPSEC_REMOTE_AUTH:
1583 				ipr = inp->inp_ipsec_remoteauth;
1584 				opt16val = IPSP_AUTH_NONE;
1585 				break;
1586 			}
1587 			if (ipr == NULL)
1588 				*mtod(m, u_int16_t *) = opt16val;
1589 			else {
1590 				size_t len;
1591 
1592 				len = m->m_len + ipr->ref_len;
1593 				if (len > MCLBYTES) {
1594 					 m_free(m);
1595 					 error = EINVAL;
1596 					 break;
1597 				}
1598 				/* allocate mbuf cluster for larger option */
1599 				if (len > MLEN) {
1600 					 MCLGET(m, M_WAITOK);
1601 					 if ((m->m_flags & M_EXT) == 0) {
1602 						 m_free(m);
1603 						 error = ENOBUFS;
1604 						 break;
1605 					 }
1606 
1607 				}
1608 				m->m_len = len;
1609 				*mtod(m, u_int16_t *) = ipr->ref_type;
1610 				m_copyback(m, sizeof(u_int16_t), ipr->ref_len,
1611 				    ipr + 1, M_NOWAIT);
1612 			}
1613 #endif
1614 			break;
1615 		case SO_RTABLE:
1616 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1617 			m->m_len = sizeof(u_int);
1618 			*mtod(m, u_int *) = inp->inp_rtableid;
1619 			break;
1620 		case IP_PIPEX:
1621 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1622 			m->m_len = sizeof(int);
1623 			*mtod(m, int *) = inp->inp_pipex;
1624 			break;
1625 		default:
1626 			error = ENOPROTOOPT;
1627 			break;
1628 		}
1629 		break;
1630 	}
1631 	return (error);
1632 }
1633 
1634 /*
1635  * Set up IP options in pcb for insertion in output packets.
1636  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1637  * with destination address if source routed.
1638  */
1639 int
1640 #ifdef notyet
1641 ip_pcbopts(optname, pcbopt, m)
1642 	int optname;
1643 #else
1644 ip_pcbopts(pcbopt, m)
1645 #endif
1646 	struct mbuf **pcbopt;
1647 	struct mbuf *m;
1648 {
1649 	int cnt, optlen;
1650 	u_char *cp;
1651 	u_char opt;
1652 
1653 	/* turn off any old options */
1654 	if (*pcbopt)
1655 		(void)m_free(*pcbopt);
1656 	*pcbopt = 0;
1657 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1658 		/*
1659 		 * Only turning off any previous options.
1660 		 */
1661 		if (m)
1662 			(void)m_free(m);
1663 		return (0);
1664 	}
1665 
1666 	if (m->m_len % sizeof(int32_t))
1667 		goto bad;
1668 
1669 	/*
1670 	 * IP first-hop destination address will be stored before
1671 	 * actual options; move other options back
1672 	 * and clear it when none present.
1673 	 */
1674 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1675 		goto bad;
1676 	cnt = m->m_len;
1677 	m->m_len += sizeof(struct in_addr);
1678 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1679 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1680 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1681 
1682 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1683 		opt = cp[IPOPT_OPTVAL];
1684 		if (opt == IPOPT_EOL)
1685 			break;
1686 		if (opt == IPOPT_NOP)
1687 			optlen = 1;
1688 		else {
1689 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1690 				goto bad;
1691 			optlen = cp[IPOPT_OLEN];
1692 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1693 				goto bad;
1694 		}
1695 		switch (opt) {
1696 
1697 		default:
1698 			break;
1699 
1700 		case IPOPT_LSRR:
1701 		case IPOPT_SSRR:
1702 			/*
1703 			 * user process specifies route as:
1704 			 *	->A->B->C->D
1705 			 * D must be our final destination (but we can't
1706 			 * check that since we may not have connected yet).
1707 			 * A is first hop destination, which doesn't appear in
1708 			 * actual IP option, but is stored before the options.
1709 			 */
1710 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1711 				goto bad;
1712 			m->m_len -= sizeof(struct in_addr);
1713 			cnt -= sizeof(struct in_addr);
1714 			optlen -= sizeof(struct in_addr);
1715 			cp[IPOPT_OLEN] = optlen;
1716 			/*
1717 			 * Move first hop before start of options.
1718 			 */
1719 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1720 			    sizeof(struct in_addr));
1721 			/*
1722 			 * Then copy rest of options back
1723 			 * to close up the deleted entry.
1724 			 */
1725 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1726 			    sizeof(struct in_addr)),
1727 			    (caddr_t)&cp[IPOPT_OFFSET+1],
1728 			    (unsigned)cnt - (IPOPT_OFFSET+1));
1729 			break;
1730 		}
1731 	}
1732 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1733 		goto bad;
1734 	*pcbopt = m;
1735 	return (0);
1736 
1737 bad:
1738 	(void)m_free(m);
1739 	return (EINVAL);
1740 }
1741 
1742 /*
1743  * Set the IP multicast options in response to user setsockopt().
1744  */
1745 int
1746 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m,
1747     u_int rtableid)
1748 {
1749 	int error = 0;
1750 	u_char loop;
1751 	int i;
1752 	struct in_addr addr;
1753 	struct ip_mreq *mreq;
1754 	struct ifnet *ifp;
1755 	struct ip_moptions *imo = *imop;
1756 	struct in_multi **immp;
1757 	struct route ro;
1758 	struct sockaddr_in *dst;
1759 
1760 	if (imo == NULL) {
1761 		/*
1762 		 * No multicast option buffer attached to the pcb;
1763 		 * allocate one and initialize to default values.
1764 		 */
1765 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1766 		    M_WAITOK|M_ZERO);
1767 		immp = (struct in_multi **)malloc(
1768 		    (sizeof(*immp) * IP_MIN_MEMBERSHIPS), M_IPMOPTS,
1769 		    M_WAITOK|M_ZERO);
1770 		*imop = imo;
1771 		imo->imo_multicast_ifp = NULL;
1772 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1773 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1774 		imo->imo_num_memberships = 0;
1775 		imo->imo_max_memberships = IP_MIN_MEMBERSHIPS;
1776 		imo->imo_membership = immp;
1777 	}
1778 
1779 	switch (optname) {
1780 
1781 	case IP_MULTICAST_IF:
1782 		/*
1783 		 * Select the interface for outgoing multicast packets.
1784 		 */
1785 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1786 			error = EINVAL;
1787 			break;
1788 		}
1789 		addr = *(mtod(m, struct in_addr *));
1790 		/*
1791 		 * INADDR_ANY is used to remove a previous selection.
1792 		 * When no interface is selected, a default one is
1793 		 * chosen every time a multicast packet is sent.
1794 		 */
1795 		if (addr.s_addr == INADDR_ANY) {
1796 			imo->imo_multicast_ifp = NULL;
1797 			break;
1798 		}
1799 		/*
1800 		 * The selected interface is identified by its local
1801 		 * IP address.  Find the interface and confirm that
1802 		 * it supports multicasting.
1803 		 */
1804 		INADDR_TO_IFP(addr, ifp, rtableid);
1805 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1806 			error = EADDRNOTAVAIL;
1807 			break;
1808 		}
1809 		imo->imo_multicast_ifp = ifp;
1810 		break;
1811 
1812 	case IP_MULTICAST_TTL:
1813 		/*
1814 		 * Set the IP time-to-live for outgoing multicast packets.
1815 		 */
1816 		if (m == NULL || m->m_len != 1) {
1817 			error = EINVAL;
1818 			break;
1819 		}
1820 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1821 		break;
1822 
1823 	case IP_MULTICAST_LOOP:
1824 		/*
1825 		 * Set the loopback flag for outgoing multicast packets.
1826 		 * Must be zero or one.
1827 		 */
1828 		if (m == NULL || m->m_len != 1 ||
1829 		   (loop = *(mtod(m, u_char *))) > 1) {
1830 			error = EINVAL;
1831 			break;
1832 		}
1833 		imo->imo_multicast_loop = loop;
1834 		break;
1835 
1836 	case IP_ADD_MEMBERSHIP:
1837 		/*
1838 		 * Add a multicast group membership.
1839 		 * Group must be a valid IP multicast address.
1840 		 */
1841 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1842 			error = EINVAL;
1843 			break;
1844 		}
1845 		mreq = mtod(m, struct ip_mreq *);
1846 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1847 			error = EINVAL;
1848 			break;
1849 		}
1850 		/*
1851 		 * If no interface address was provided, use the interface of
1852 		 * the route to the given multicast address.
1853 		 */
1854 		if (mreq->imr_interface.s_addr == INADDR_ANY) {
1855 			ro.ro_rt = NULL;
1856 			dst = satosin(&ro.ro_dst);
1857 			dst->sin_len = sizeof(*dst);
1858 			dst->sin_family = AF_INET;
1859 			dst->sin_addr = mreq->imr_multiaddr;
1860 			if (!(ro.ro_rt && ro.ro_rt->rt_ifp &&
1861 			    (ro.ro_rt->rt_flags & RTF_UP)))
1862 				ro.ro_rt = rtalloc1(&ro.ro_dst, RT_REPORT,
1863 				    rtableid);
1864 			if (ro.ro_rt == NULL) {
1865 				error = EADDRNOTAVAIL;
1866 				break;
1867 			}
1868 			ifp = ro.ro_rt->rt_ifp;
1869 			rtfree(ro.ro_rt);
1870 		} else {
1871 			INADDR_TO_IFP(mreq->imr_interface, ifp, rtableid);
1872 		}
1873 		/*
1874 		 * See if we found an interface, and confirm that it
1875 		 * supports multicast.
1876 		 */
1877 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1878 			error = EADDRNOTAVAIL;
1879 			break;
1880 		}
1881 		/*
1882 		 * See if the membership already exists or if all the
1883 		 * membership slots are full.
1884 		 */
1885 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1886 			if (imo->imo_membership[i]->inm_ia->ia_ifp == ifp &&
1887 			    imo->imo_membership[i]->inm_addr.s_addr
1888 						== mreq->imr_multiaddr.s_addr)
1889 				break;
1890 		}
1891 		if (i < imo->imo_num_memberships) {
1892 			error = EADDRINUSE;
1893 			break;
1894 		}
1895 		if (imo->imo_num_memberships == imo->imo_max_memberships) {
1896 			struct in_multi **nmships, **omships;
1897 			size_t newmax;
1898 			/*
1899 			 * Resize the vector to next power-of-two minus 1. If the
1900 			 * size would exceed the maximum then we know we've really
1901 			 * run out of entries. Otherwise, we reallocate the vector.
1902 			 */
1903 			nmships = NULL;
1904 			omships = imo->imo_membership;
1905 			newmax = ((imo->imo_max_memberships + 1) * 2) - 1;
1906 			if (newmax <= IP_MAX_MEMBERSHIPS) {
1907 				nmships = (struct in_multi **)malloc(
1908 				    sizeof(*nmships) * newmax, M_IPMOPTS,
1909 				    M_NOWAIT|M_ZERO);
1910 				if (nmships != NULL) {
1911 					bcopy(omships, nmships,
1912 					    sizeof(*omships) *
1913 					    imo->imo_max_memberships);
1914 					free(omships, M_IPMOPTS);
1915 					imo->imo_membership = nmships;
1916 					imo->imo_max_memberships = newmax;
1917 				}
1918 			}
1919 			if (nmships == NULL) {
1920 				error = ETOOMANYREFS;
1921 				break;
1922 			}
1923 		}
1924 		/*
1925 		 * Everything looks good; add a new record to the multicast
1926 		 * address list for the given interface.
1927 		 */
1928 		if ((imo->imo_membership[i] =
1929 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1930 			error = ENOBUFS;
1931 			break;
1932 		}
1933 		++imo->imo_num_memberships;
1934 		break;
1935 
1936 	case IP_DROP_MEMBERSHIP:
1937 		/*
1938 		 * Drop a multicast group membership.
1939 		 * Group must be a valid IP multicast address.
1940 		 */
1941 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1942 			error = EINVAL;
1943 			break;
1944 		}
1945 		mreq = mtod(m, struct ip_mreq *);
1946 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1947 			error = EINVAL;
1948 			break;
1949 		}
1950 		/*
1951 		 * If an interface address was specified, get a pointer
1952 		 * to its ifnet structure.
1953 		 */
1954 		if (mreq->imr_interface.s_addr == INADDR_ANY)
1955 			ifp = NULL;
1956 		else {
1957 			INADDR_TO_IFP(mreq->imr_interface, ifp, rtableid);
1958 			if (ifp == NULL) {
1959 				error = EADDRNOTAVAIL;
1960 				break;
1961 			}
1962 		}
1963 		/*
1964 		 * Find the membership in the membership array.
1965 		 */
1966 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1967 			if ((ifp == NULL ||
1968 			     imo->imo_membership[i]->inm_ia->ia_ifp == ifp) &&
1969 			     imo->imo_membership[i]->inm_addr.s_addr ==
1970 			     mreq->imr_multiaddr.s_addr)
1971 				break;
1972 		}
1973 		if (i == imo->imo_num_memberships) {
1974 			error = EADDRNOTAVAIL;
1975 			break;
1976 		}
1977 		/*
1978 		 * Give up the multicast address record to which the
1979 		 * membership points.
1980 		 */
1981 		in_delmulti(imo->imo_membership[i]);
1982 		/*
1983 		 * Remove the gap in the membership array.
1984 		 */
1985 		for (++i; i < imo->imo_num_memberships; ++i)
1986 			imo->imo_membership[i-1] = imo->imo_membership[i];
1987 		--imo->imo_num_memberships;
1988 		break;
1989 
1990 	default:
1991 		error = EOPNOTSUPP;
1992 		break;
1993 	}
1994 
1995 	/*
1996 	 * If all options have default values, no need to keep the data.
1997 	 */
1998 	if (imo->imo_multicast_ifp == NULL &&
1999 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2000 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2001 	    imo->imo_num_memberships == 0) {
2002 		free(imo->imo_membership , M_IPMOPTS);
2003 		free(*imop, M_IPMOPTS);
2004 		*imop = NULL;
2005 	}
2006 
2007 	return (error);
2008 }
2009 
2010 /*
2011  * Return the IP multicast options in response to user getsockopt().
2012  */
2013 int
2014 ip_getmoptions(optname, imo, mp)
2015 	int optname;
2016 	struct ip_moptions *imo;
2017 	struct mbuf **mp;
2018 {
2019 	u_char *ttl;
2020 	u_char *loop;
2021 	struct in_addr *addr;
2022 	struct in_ifaddr *ia;
2023 
2024 	*mp = m_get(M_WAIT, MT_SOOPTS);
2025 
2026 	switch (optname) {
2027 
2028 	case IP_MULTICAST_IF:
2029 		addr = mtod(*mp, struct in_addr *);
2030 		(*mp)->m_len = sizeof(struct in_addr);
2031 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
2032 			addr->s_addr = INADDR_ANY;
2033 		else {
2034 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
2035 			addr->s_addr = (ia == NULL) ? INADDR_ANY
2036 					: ia->ia_addr.sin_addr.s_addr;
2037 		}
2038 		return (0);
2039 
2040 	case IP_MULTICAST_TTL:
2041 		ttl = mtod(*mp, u_char *);
2042 		(*mp)->m_len = 1;
2043 		*ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
2044 				     : imo->imo_multicast_ttl;
2045 		return (0);
2046 
2047 	case IP_MULTICAST_LOOP:
2048 		loop = mtod(*mp, u_char *);
2049 		(*mp)->m_len = 1;
2050 		*loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
2051 				      : imo->imo_multicast_loop;
2052 		return (0);
2053 
2054 	default:
2055 		return (EOPNOTSUPP);
2056 	}
2057 }
2058 
2059 /*
2060  * Discard the IP multicast options.
2061  */
2062 void
2063 ip_freemoptions(imo)
2064 	struct ip_moptions *imo;
2065 {
2066 	int i;
2067 
2068 	if (imo != NULL) {
2069 		for (i = 0; i < imo->imo_num_memberships; ++i)
2070 			in_delmulti(imo->imo_membership[i]);
2071 		free(imo->imo_membership, M_IPMOPTS);
2072 		free(imo, M_IPMOPTS);
2073 	}
2074 }
2075 
2076 /*
2077  * Routine called from ip_output() to loop back a copy of an IP multicast
2078  * packet to the input queue of a specified interface.  Note that this
2079  * calls the output routine of the loopback "driver", but with an interface
2080  * pointer that might NOT be &loif -- easier than replicating that code here.
2081  */
2082 void
2083 ip_mloopback(ifp, m, dst)
2084 	struct ifnet *ifp;
2085 	struct mbuf *m;
2086 	struct sockaddr_in *dst;
2087 {
2088 	struct ip *ip;
2089 	struct mbuf *copym;
2090 
2091 	copym = m_copym2(m, 0, M_COPYALL, M_DONTWAIT);
2092 	if (copym != NULL) {
2093 		/*
2094 		 * We don't bother to fragment if the IP length is greater
2095 		 * than the interface's MTU.  Can this possibly matter?
2096 		 */
2097 		ip = mtod(copym, struct ip *);
2098 		ip->ip_sum = 0;
2099 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
2100 		(void) looutput(ifp, copym, sintosa(dst), NULL);
2101 	}
2102 }
2103 
2104 /*
2105  * Process a delayed payload checksum calculation.
2106  */
2107 void
2108 in_delayed_cksum(struct mbuf *m)
2109 {
2110 	struct ip *ip;
2111 	u_int16_t csum, offset;
2112 
2113 	ip = mtod(m, struct ip *);
2114 	offset = ip->ip_hl << 2;
2115 	csum = in4_cksum(m, 0, offset, m->m_pkthdr.len - offset);
2116 	if (csum == 0 && ip->ip_p == IPPROTO_UDP)
2117 		csum = 0xffff;
2118 
2119 	switch (ip->ip_p) {
2120 	case IPPROTO_TCP:
2121 		offset += offsetof(struct tcphdr, th_sum);
2122 		break;
2123 
2124 	case IPPROTO_UDP:
2125 		offset += offsetof(struct udphdr, uh_sum);
2126 		break;
2127 
2128 	default:
2129 		return;
2130 	}
2131 
2132 	if ((offset + sizeof(u_int16_t)) > m->m_len)
2133 		m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
2134 	else
2135 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2136 }
2137 
2138 void
2139 in_proto_cksum_out(struct mbuf *m, struct ifnet *ifp)
2140 {
2141 	if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT) {
2142 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_TCPv4) ||
2143 		    ifp->if_bridge != NULL) {
2144 			in_delayed_cksum(m);
2145 			m->m_pkthdr.csum_flags &= ~M_TCP_CSUM_OUT; /* Clear */
2146 		}
2147 	} else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT) {
2148 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_UDPv4) ||
2149 		    ifp->if_bridge != NULL) {
2150 			in_delayed_cksum(m);
2151 			m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_OUT; /* Clear */
2152 		}
2153 	} else if (m->m_pkthdr.csum_flags & M_ICMP_CSUM_OUT) {
2154 		struct ip *ip = mtod(m, struct ip *);
2155 		int hlen;
2156 		struct icmp *icp;
2157 
2158 		hlen = ip->ip_hl << 2;
2159 		m->m_data += hlen;
2160 		m->m_len -= hlen;
2161 		icp = mtod(m, struct icmp *);
2162 		icp->icmp_cksum = 0;
2163 		icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
2164 		m->m_data -= hlen;
2165 		m->m_len += hlen;
2166 		m->m_pkthdr.csum_flags &= ~M_ICMP_CSUM_OUT; /* Clear */
2167 	}
2168 }
2169