xref: /openbsd-src/sys/netinet/ip_output.c (revision a28daedfc357b214be5c701aa8ba8adb29a7f1c2)
1 /*	$OpenBSD: ip_output.c,v 1.193 2009/01/30 20:46:33 claudio Exp $	*/
2 /*	$NetBSD: ip_output.c,v 1.28 1996/02/13 23:43:07 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1990, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
33  */
34 
35 #include "pf.h"
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mbuf.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/socketvar.h>
43 #include <sys/proc.h>
44 #include <sys/kernel.h>
45 
46 #include <net/if.h>
47 #include <net/if_enc.h>
48 #include <net/route.h>
49 
50 #include <netinet/in.h>
51 #include <netinet/in_systm.h>
52 #include <netinet/ip.h>
53 #include <netinet/in_pcb.h>
54 #include <netinet/in_var.h>
55 #include <netinet/ip_var.h>
56 #include <netinet/ip_icmp.h>
57 #include <netinet/tcp.h>
58 #include <netinet/udp.h>
59 #include <netinet/tcp_timer.h>
60 #include <netinet/tcp_var.h>
61 #include <netinet/udp_var.h>
62 
63 #if NPF > 0
64 #include <net/pfvar.h>
65 #endif
66 
67 #ifdef IPSEC
68 #ifdef ENCDEBUG
69 #define DPRINTF(x)    do { if (encdebug) printf x ; } while (0)
70 #else
71 #define DPRINTF(x)
72 #endif
73 
74 extern u_int8_t get_sa_require(struct inpcb *);
75 
76 extern int ipsec_auth_default_level;
77 extern int ipsec_esp_trans_default_level;
78 extern int ipsec_esp_network_default_level;
79 extern int ipsec_ipcomp_default_level;
80 extern int ipforwarding;
81 #endif /* IPSEC */
82 
83 #ifdef MROUTING
84 extern int ipmforwarding;
85 #endif
86 
87 struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
88 void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *);
89 
90 /*
91  * IP output.  The packet in mbuf chain m contains a skeletal IP
92  * header (with len, off, ttl, proto, tos, src, dst).
93  * The mbuf chain containing the packet will be freed.
94  * The mbuf opt, if present, will not be freed.
95  */
96 int
97 ip_output(struct mbuf *m0, ...)
98 {
99 	struct ip *ip;
100 	struct ifnet *ifp;
101 	struct mbuf *m = m0;
102 	int hlen = sizeof (struct ip);
103 	int len, error = 0;
104 	struct route iproute;
105 	struct sockaddr_in *dst;
106 	struct in_ifaddr *ia;
107 	struct mbuf *opt;
108 	struct route *ro;
109 	int flags;
110 	struct ip_moptions *imo;
111 	va_list ap;
112 	u_int8_t sproto = 0, donerouting = 0;
113 	u_long mtu;
114 #ifdef IPSEC
115 	u_int32_t icmp_mtu = 0;
116 	union sockaddr_union sdst;
117 	u_int32_t sspi;
118 	struct m_tag *mtag;
119 	struct tdb_ident *tdbi;
120 
121 	struct inpcb *inp;
122 	struct tdb *tdb;
123 	int s;
124 #endif /* IPSEC */
125 
126 	va_start(ap, m0);
127 	opt = va_arg(ap, struct mbuf *);
128 	ro = va_arg(ap, struct route *);
129 	flags = va_arg(ap, int);
130 	imo = va_arg(ap, struct ip_moptions *);
131 #ifdef IPSEC
132 	inp = va_arg(ap, struct inpcb *);
133 	if (inp && (inp->inp_flags & INP_IPV6) != 0)
134 		panic("ip_output: IPv6 pcb is passed");
135 #endif /* IPSEC */
136 	va_end(ap);
137 
138 #ifdef	DIAGNOSTIC
139 	if ((m->m_flags & M_PKTHDR) == 0)
140 		panic("ip_output no HDR");
141 #endif
142 	if (opt) {
143 		m = ip_insertoptions(m, opt, &len);
144 		hlen = len;
145 	}
146 
147 	ip = mtod(m, struct ip *);
148 
149 	/*
150 	 * Fill in IP header.
151 	 */
152 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
153 		ip->ip_v = IPVERSION;
154 		ip->ip_off &= htons(IP_DF);
155 		ip->ip_id = htons(ip_randomid());
156 		ip->ip_hl = hlen >> 2;
157 		ipstat.ips_localout++;
158 	} else {
159 		hlen = ip->ip_hl << 2;
160 	}
161 
162 	/*
163 	 * If we're missing the IP source address, do a route lookup. We'll
164 	 * remember this result, in case we don't need to do any IPsec
165 	 * processing on the packet. We need the source address so we can
166 	 * do an SPD lookup in IPsec; for most packets, the source address
167 	 * is set at a higher level protocol. ICMPs and other packets
168 	 * though (e.g., traceroute) have a source address of zeroes.
169 	 */
170 	if (ip->ip_src.s_addr == INADDR_ANY) {
171 		if (flags & IP_ROUTETOETHER) {
172 			error = EINVAL;
173 			goto bad;
174 		}
175 		donerouting = 1;
176 
177 		if (ro == 0) {
178 			ro = &iproute;
179 			bzero((caddr_t)ro, sizeof (*ro));
180 		}
181 
182 		dst = satosin(&ro->ro_dst);
183 
184 		/*
185 		 * If there is a cached route, check that it is to the same
186 		 * destination and is still up.  If not, free it and try again.
187 		 */
188 		if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
189 				  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
190 			RTFREE(ro->ro_rt);
191 			ro->ro_rt = (struct rtentry *)0;
192 		}
193 
194 		if (ro->ro_rt == 0) {
195 			dst->sin_family = AF_INET;
196 			dst->sin_len = sizeof(*dst);
197 			dst->sin_addr = ip->ip_dst;
198 		}
199 
200 		/*
201 		 * If routing to interface only, short-circuit routing lookup.
202 		 */
203 		if (flags & IP_ROUTETOIF) {
204 			if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 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 		ip->ip_sum = 0;
725 		if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
726 		    ifp->if_bridge == NULL) {
727 			m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
728 			ipstat.ips_outhwcsum++;
729 		} else
730 			ip->ip_sum = in_cksum(m, hlen);
731 		/* Update relevant hardware checksum stats for TCP/UDP */
732 		if (m->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT)
733 			tcpstat.tcps_outhwcsum++;
734 		else if (m->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT)
735 			udpstat.udps_outhwcsum++;
736 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
737 		goto done;
738 	}
739 
740 	/*
741 	 * Too large for interface; fragment if possible.
742 	 * Must be able to put at least 8 bytes per fragment.
743 	 */
744 	if (ip->ip_off & htons(IP_DF)) {
745 #ifdef IPSEC
746 		icmp_mtu = ifp->if_mtu;
747 #endif
748 		error = EMSGSIZE;
749 		/*
750 		 * This case can happen if the user changed the MTU
751 		 * of an interface after enabling IP on it.  Because
752 		 * most netifs don't keep track of routes pointing to
753 		 * them, there is no way for one to update all its
754 		 * routes when the MTU is changed.
755 		 */
756 		if (ro->ro_rt != NULL &&
757 		    (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 		mhip->ip_sum = 0;
875 		if ((ifp != NULL) &&
876 		    (ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
877 		    ifp->if_bridge == NULL) {
878 			m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
879 			ipstat.ips_outhwcsum++;
880 		} else
881 			mhip->ip_sum = in_cksum(m, mhlen);
882 		ipstat.ips_ofragments++;
883 		fragments++;
884 	}
885 	/*
886 	 * Update first fragment by trimming what's been copied out
887 	 * and updating header, then send each fragment (in order).
888 	 */
889 	m = m0;
890 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
891 	m->m_pkthdr.len = hlen + firstlen;
892 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
893 	ip->ip_off |= htons(IP_MF);
894 	ip->ip_sum = 0;
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 = in_cksum(m, hlen);
902 sendorfree:
903 	/*
904 	 * If there is no room for all the fragments, don't queue
905 	 * any of them.
906 	 */
907 	if (ifp != NULL) {
908 		s = splnet();
909 		if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
910 		    error == 0) {
911 			error = ENOBUFS;
912 			ipstat.ips_odropped++;
913 			IFQ_INC_DROPS(&ifp->if_snd);
914 		}
915 		splx(s);
916 	}
917 	if (error) {
918 		for (m = m0; m; m = m0) {
919 			m0 = m->m_nextpkt;
920 			m->m_nextpkt = NULL;
921 			m_freem(m);
922 		}
923 	}
924 
925 	return (error);
926 }
927 
928 /*
929  * Insert IP options into preformed packet.
930  * Adjust IP destination as required for IP source routing,
931  * as indicated by a non-zero in_addr at the start of the options.
932  */
933 struct mbuf *
934 ip_insertoptions(m, opt, phlen)
935 	struct mbuf *m;
936 	struct mbuf *opt;
937 	int *phlen;
938 {
939 	struct ipoption *p = mtod(opt, struct ipoption *);
940 	struct mbuf *n;
941 	struct ip *ip = mtod(m, struct ip *);
942 	unsigned optlen;
943 
944 	optlen = opt->m_len - sizeof(p->ipopt_dst);
945 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
946 		return (m);		/* XXX should fail */
947 	if (p->ipopt_dst.s_addr)
948 		ip->ip_dst = p->ipopt_dst;
949 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
950 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
951 		if (n == 0)
952 			return (m);
953 		M_MOVE_HDR(n, m);
954 		n->m_pkthdr.len += optlen;
955 		m->m_len -= sizeof(struct ip);
956 		m->m_data += sizeof(struct ip);
957 		n->m_next = m;
958 		m = n;
959 		m->m_len = optlen + sizeof(struct ip);
960 		m->m_data += max_linkhdr;
961 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
962 	} else {
963 		m->m_data -= optlen;
964 		m->m_len += optlen;
965 		m->m_pkthdr.len += optlen;
966 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
967 	}
968 	ip = mtod(m, struct ip *);
969 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
970 	*phlen = sizeof(struct ip) + optlen;
971 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
972 	return (m);
973 }
974 
975 /*
976  * Copy options from ip to jp,
977  * omitting those not copied during fragmentation.
978  */
979 int
980 ip_optcopy(ip, jp)
981 	struct ip *ip, *jp;
982 {
983 	u_char *cp, *dp;
984 	int opt, optlen, cnt;
985 
986 	cp = (u_char *)(ip + 1);
987 	dp = (u_char *)(jp + 1);
988 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
989 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
990 		opt = cp[0];
991 		if (opt == IPOPT_EOL)
992 			break;
993 		if (opt == IPOPT_NOP) {
994 			/* Preserve for IP mcast tunnel's LSRR alignment. */
995 			*dp++ = IPOPT_NOP;
996 			optlen = 1;
997 			continue;
998 		}
999 #ifdef DIAGNOSTIC
1000 		if (cnt < IPOPT_OLEN + sizeof(*cp))
1001 			panic("malformed IPv4 option passed to ip_optcopy");
1002 #endif
1003 		optlen = cp[IPOPT_OLEN];
1004 #ifdef DIAGNOSTIC
1005 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1006 			panic("malformed IPv4 option passed to ip_optcopy");
1007 #endif
1008 		/* bogus lengths should have been caught by ip_dooptions */
1009 		if (optlen > cnt)
1010 			optlen = cnt;
1011 		if (IPOPT_COPIED(opt)) {
1012 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
1013 			dp += optlen;
1014 		}
1015 	}
1016 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1017 		*dp++ = IPOPT_EOL;
1018 	return (optlen);
1019 }
1020 
1021 /*
1022  * IP socket option processing.
1023  */
1024 int
1025 ip_ctloutput(op, so, level, optname, mp)
1026 	int op;
1027 	struct socket *so;
1028 	int level, optname;
1029 	struct mbuf **mp;
1030 {
1031 	struct inpcb *inp = sotoinpcb(so);
1032 	struct mbuf *m = *mp;
1033 	int optval = 0;
1034 #ifdef IPSEC
1035 	struct proc *p = curproc; /* XXX */
1036 	struct ipsec_ref *ipr;
1037 	u_int16_t opt16val;
1038 #endif
1039 	int error = 0;
1040 
1041 	if (level != IPPROTO_IP) {
1042 		error = EINVAL;
1043 		if (op == PRCO_SETOPT && *mp)
1044 			(void) m_free(*mp);
1045 	} else switch (op) {
1046 	case PRCO_SETOPT:
1047 		switch (optname) {
1048 		case IP_OPTIONS:
1049 #ifdef notyet
1050 		case IP_RETOPTS:
1051 			return (ip_pcbopts(optname, &inp->inp_options, m));
1052 #else
1053 			return (ip_pcbopts(&inp->inp_options, m));
1054 #endif
1055 
1056 		case IP_TOS:
1057 		case IP_TTL:
1058 		case IP_MINTTL:
1059 		case IP_RECVOPTS:
1060 		case IP_RECVRETOPTS:
1061 		case IP_RECVDSTADDR:
1062 		case IP_RECVIF:
1063 		case IP_RECVTTL:
1064 		case IP_RECVDSTPORT:
1065 			if (m == NULL || m->m_len != sizeof(int))
1066 				error = EINVAL;
1067 			else {
1068 				optval = *mtod(m, int *);
1069 				switch (optname) {
1070 
1071 				case IP_TOS:
1072 					inp->inp_ip.ip_tos = optval;
1073 					break;
1074 
1075 				case IP_TTL:
1076 					if (optval > 0 && optval <= MAXTTL)
1077 						inp->inp_ip.ip_ttl = optval;
1078 					else
1079 						error = EINVAL;
1080 					break;
1081 
1082 				case IP_MINTTL:
1083 					if (optval > 0 && optval <= MAXTTL)
1084 						inp->inp_ip_minttl = optval;
1085 					else
1086 						error = EINVAL;
1087 					break;
1088 #define	OPTSET(bit) \
1089 	if (optval) \
1090 		inp->inp_flags |= bit; \
1091 	else \
1092 		inp->inp_flags &= ~bit;
1093 
1094 				case IP_RECVOPTS:
1095 					OPTSET(INP_RECVOPTS);
1096 					break;
1097 
1098 				case IP_RECVRETOPTS:
1099 					OPTSET(INP_RECVRETOPTS);
1100 					break;
1101 
1102 				case IP_RECVDSTADDR:
1103 					OPTSET(INP_RECVDSTADDR);
1104 					break;
1105 				case IP_RECVIF:
1106 					OPTSET(INP_RECVIF);
1107 					break;
1108 				case IP_RECVTTL:
1109 					OPTSET(INP_RECVTTL);
1110 					break;
1111 				case IP_RECVDSTPORT:
1112 					OPTSET(INP_RECVDSTPORT);
1113 					break;
1114 				}
1115 			}
1116 			break;
1117 #undef OPTSET
1118 
1119 		case IP_MULTICAST_IF:
1120 		case IP_MULTICAST_TTL:
1121 		case IP_MULTICAST_LOOP:
1122 		case IP_ADD_MEMBERSHIP:
1123 		case IP_DROP_MEMBERSHIP:
1124 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
1125 			break;
1126 
1127 		case IP_PORTRANGE:
1128 			if (m == 0 || m->m_len != sizeof(int))
1129 				error = EINVAL;
1130 			else {
1131 				optval = *mtod(m, int *);
1132 
1133 				switch (optval) {
1134 
1135 				case IP_PORTRANGE_DEFAULT:
1136 					inp->inp_flags &= ~(INP_LOWPORT);
1137 					inp->inp_flags &= ~(INP_HIGHPORT);
1138 					break;
1139 
1140 				case IP_PORTRANGE_HIGH:
1141 					inp->inp_flags &= ~(INP_LOWPORT);
1142 					inp->inp_flags |= INP_HIGHPORT;
1143 					break;
1144 
1145 				case IP_PORTRANGE_LOW:
1146 					inp->inp_flags &= ~(INP_HIGHPORT);
1147 					inp->inp_flags |= INP_LOWPORT;
1148 					break;
1149 
1150 				default:
1151 
1152 					error = EINVAL;
1153 					break;
1154 				}
1155 			}
1156 			break;
1157 		case IP_AUTH_LEVEL:
1158 		case IP_ESP_TRANS_LEVEL:
1159 		case IP_ESP_NETWORK_LEVEL:
1160 		case IP_IPCOMP_LEVEL:
1161 #ifndef IPSEC
1162 			error = EOPNOTSUPP;
1163 #else
1164 			if (m == 0 || m->m_len != sizeof(int)) {
1165 				error = EINVAL;
1166 				break;
1167 			}
1168 			optval = *mtod(m, int *);
1169 
1170 			if (optval < IPSEC_LEVEL_BYPASS ||
1171 			    optval > IPSEC_LEVEL_UNIQUE) {
1172 				error = EINVAL;
1173 				break;
1174 			}
1175 
1176 			/* Unlink cached output TDB to force a re-search */
1177 			if (inp->inp_tdb_out) {
1178 				int s = spltdb();
1179 				TAILQ_REMOVE(&inp->inp_tdb_out->tdb_inp_out,
1180 				    inp, inp_tdb_out_next);
1181 				splx(s);
1182 			}
1183 
1184 			if (inp->inp_tdb_in) {
1185 				int s = spltdb();
1186 				TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in,
1187 				    inp, inp_tdb_in_next);
1188 				splx(s);
1189 			}
1190 
1191 			switch (optname) {
1192 			case IP_AUTH_LEVEL:
1193 				if (optval < ipsec_auth_default_level &&
1194 				    suser(p, 0)) {
1195 					error = EACCES;
1196 					break;
1197 				}
1198 				inp->inp_seclevel[SL_AUTH] = optval;
1199 				break;
1200 
1201 			case IP_ESP_TRANS_LEVEL:
1202 				if (optval < ipsec_esp_trans_default_level &&
1203 				    suser(p, 0)) {
1204 					error = EACCES;
1205 					break;
1206 				}
1207 				inp->inp_seclevel[SL_ESP_TRANS] = optval;
1208 				break;
1209 
1210 			case IP_ESP_NETWORK_LEVEL:
1211 				if (optval < ipsec_esp_network_default_level &&
1212 				    suser(p, 0)) {
1213 					error = EACCES;
1214 					break;
1215 				}
1216 				inp->inp_seclevel[SL_ESP_NETWORK] = optval;
1217 				break;
1218 			case IP_IPCOMP_LEVEL:
1219 				if (optval < ipsec_ipcomp_default_level &&
1220 				    suser(p, 0)) {
1221 					error = EACCES;
1222 					break;
1223 				}
1224 				inp->inp_seclevel[SL_IPCOMP] = optval;
1225 				break;
1226 			}
1227 			if (!error)
1228 				inp->inp_secrequire = get_sa_require(inp);
1229 #endif
1230 			break;
1231 
1232 		case IP_IPSEC_REMOTE_CRED:
1233 		case IP_IPSEC_REMOTE_AUTH:
1234 			/* Can't set the remote credential or key */
1235 			error = EOPNOTSUPP;
1236 			break;
1237 
1238 		case IP_IPSEC_LOCAL_ID:
1239 		case IP_IPSEC_REMOTE_ID:
1240 		case IP_IPSEC_LOCAL_CRED:
1241 		case IP_IPSEC_LOCAL_AUTH:
1242 #ifndef IPSEC
1243 			error = EOPNOTSUPP;
1244 #else
1245 			if (m->m_len < 2) {
1246 				error = EINVAL;
1247 				break;
1248 			}
1249 
1250 			m_copydata(m, 0, 2, (caddr_t) &opt16val);
1251 
1252 			/* If the type is 0, then we cleanup and return */
1253 			if (opt16val == 0) {
1254 				switch (optname) {
1255 				case IP_IPSEC_LOCAL_ID:
1256 					if (inp->inp_ipo != NULL &&
1257 					    inp->inp_ipo->ipo_srcid != NULL) {
1258 						ipsp_reffree(inp->inp_ipo->ipo_srcid);
1259 						inp->inp_ipo->ipo_srcid = NULL;
1260 					}
1261 					break;
1262 
1263 				case IP_IPSEC_REMOTE_ID:
1264 					if (inp->inp_ipo != NULL &&
1265 					    inp->inp_ipo->ipo_dstid != NULL) {
1266 						ipsp_reffree(inp->inp_ipo->ipo_dstid);
1267 						inp->inp_ipo->ipo_dstid = NULL;
1268 					}
1269 					break;
1270 
1271 				case IP_IPSEC_LOCAL_CRED:
1272 					if (inp->inp_ipo != NULL &&
1273 					    inp->inp_ipo->ipo_local_cred != NULL) {
1274 						ipsp_reffree(inp->inp_ipo->ipo_local_cred);
1275 						inp->inp_ipo->ipo_local_cred = NULL;
1276 					}
1277 					break;
1278 
1279 				case IP_IPSEC_LOCAL_AUTH:
1280 					if (inp->inp_ipo != NULL &&
1281 					    inp->inp_ipo->ipo_local_auth != NULL) {
1282 						ipsp_reffree(inp->inp_ipo->ipo_local_auth);
1283 						inp->inp_ipo->ipo_local_auth = NULL;
1284 					}
1285 					break;
1286 				}
1287 
1288 				error = 0;
1289 				break;
1290 			}
1291 
1292 			/* Can't have an empty payload */
1293 			if (m->m_len == 2) {
1294 				error = EINVAL;
1295 				break;
1296 			}
1297 
1298 			/* Allocate if needed */
1299 			if (inp->inp_ipo == NULL) {
1300 				inp->inp_ipo = ipsec_add_policy(inp,
1301 				    AF_INET, IPSP_DIRECTION_OUT);
1302 				if (inp->inp_ipo == NULL) {
1303 					error = ENOBUFS;
1304 					break;
1305 				}
1306 			}
1307 
1308 			ipr = malloc(sizeof(struct ipsec_ref) + m->m_len - 2,
1309 			       M_CREDENTIALS, M_NOWAIT);
1310 			if (ipr == NULL) {
1311 				error = ENOBUFS;
1312 				break;
1313 			}
1314 
1315 			ipr->ref_count = 1;
1316 			ipr->ref_malloctype = M_CREDENTIALS;
1317 			ipr->ref_len = m->m_len - 2;
1318 			ipr->ref_type = opt16val;
1319 			m_copydata(m, 2, m->m_len - 2, (caddr_t)(ipr + 1));
1320 
1321 			switch (optname) {
1322 			case IP_IPSEC_LOCAL_ID:
1323 				/* Check valid types and NUL-termination */
1324 				if (ipr->ref_type < IPSP_IDENTITY_PREFIX ||
1325 				    ipr->ref_type > IPSP_IDENTITY_CONNECTION ||
1326 				    ((char *)(ipr + 1))[ipr->ref_len - 1]) {
1327 					free(ipr, M_CREDENTIALS);
1328 					error = EINVAL;
1329 				} else {
1330 					if (inp->inp_ipo->ipo_srcid != NULL)
1331 						ipsp_reffree(inp->inp_ipo->ipo_srcid);
1332 					inp->inp_ipo->ipo_srcid = ipr;
1333 				}
1334 				break;
1335 			case IP_IPSEC_REMOTE_ID:
1336 				/* Check valid types and NUL-termination */
1337 				if (ipr->ref_type < IPSP_IDENTITY_PREFIX ||
1338 				    ipr->ref_type > IPSP_IDENTITY_CONNECTION ||
1339 				    ((char *)(ipr + 1))[ipr->ref_len - 1]) {
1340 					free(ipr, M_CREDENTIALS);
1341 					error = EINVAL;
1342 				} else {
1343 					if (inp->inp_ipo->ipo_dstid != NULL)
1344 						ipsp_reffree(inp->inp_ipo->ipo_dstid);
1345 					inp->inp_ipo->ipo_dstid = ipr;
1346 				}
1347 				break;
1348 			case IP_IPSEC_LOCAL_CRED:
1349 				if (ipr->ref_type < IPSP_CRED_KEYNOTE ||
1350 				    ipr->ref_type > IPSP_CRED_X509) {
1351 					free(ipr, M_CREDENTIALS);
1352 					error = EINVAL;
1353 				} else {
1354 					if (inp->inp_ipo->ipo_local_cred != NULL)
1355 						ipsp_reffree(inp->inp_ipo->ipo_local_cred);
1356 					inp->inp_ipo->ipo_local_cred = ipr;
1357 				}
1358 				break;
1359 			case IP_IPSEC_LOCAL_AUTH:
1360 				if (ipr->ref_type < IPSP_AUTH_PASSPHRASE ||
1361 				    ipr->ref_type > IPSP_AUTH_RSA) {
1362 					free(ipr, M_CREDENTIALS);
1363 					error = EINVAL;
1364 				} else {
1365 					if (inp->inp_ipo->ipo_local_auth != NULL)
1366 						ipsp_reffree(inp->inp_ipo->ipo_local_auth);
1367 					inp->inp_ipo->ipo_local_auth = ipr;
1368 				}
1369 				break;
1370 			}
1371 
1372 			/* Unlink cached output TDB to force a re-search */
1373 			if (inp->inp_tdb_out) {
1374 				int s = spltdb();
1375 				TAILQ_REMOVE(&inp->inp_tdb_out->tdb_inp_out,
1376 				    inp, inp_tdb_out_next);
1377 				splx(s);
1378 			}
1379 
1380 			if (inp->inp_tdb_in) {
1381 				int s = spltdb();
1382 				TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in,
1383 				    inp, inp_tdb_in_next);
1384 				splx(s);
1385 			}
1386 #endif
1387 			break;
1388 		default:
1389 			error = ENOPROTOOPT;
1390 			break;
1391 		}
1392 		if (m)
1393 			(void)m_free(m);
1394 		break;
1395 
1396 	case PRCO_GETOPT:
1397 		switch (optname) {
1398 		case IP_OPTIONS:
1399 		case IP_RETOPTS:
1400 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1401 			if (inp->inp_options) {
1402 				m->m_len = inp->inp_options->m_len;
1403 				bcopy(mtod(inp->inp_options, caddr_t),
1404 				    mtod(m, caddr_t), (unsigned)m->m_len);
1405 			} else
1406 				m->m_len = 0;
1407 			break;
1408 
1409 		case IP_TOS:
1410 		case IP_TTL:
1411 		case IP_MINTTL:
1412 		case IP_RECVOPTS:
1413 		case IP_RECVRETOPTS:
1414 		case IP_RECVDSTADDR:
1415 		case IP_RECVIF:
1416 		case IP_RECVTTL:
1417 		case IP_RECVDSTPORT:
1418 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1419 			m->m_len = sizeof(int);
1420 			switch (optname) {
1421 
1422 			case IP_TOS:
1423 				optval = inp->inp_ip.ip_tos;
1424 				break;
1425 
1426 			case IP_TTL:
1427 				optval = inp->inp_ip.ip_ttl;
1428 				break;
1429 
1430 			case IP_MINTTL:
1431 				optval = inp->inp_ip_minttl;
1432 				break;
1433 
1434 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1435 
1436 			case IP_RECVOPTS:
1437 				optval = OPTBIT(INP_RECVOPTS);
1438 				break;
1439 
1440 			case IP_RECVRETOPTS:
1441 				optval = OPTBIT(INP_RECVRETOPTS);
1442 				break;
1443 
1444 			case IP_RECVDSTADDR:
1445 				optval = OPTBIT(INP_RECVDSTADDR);
1446 				break;
1447 			case IP_RECVIF:
1448 				optval = OPTBIT(INP_RECVIF);
1449 				break;
1450 			case IP_RECVTTL:
1451 				optval = OPTBIT(INP_RECVTTL);
1452 				break;
1453 			case IP_RECVDSTPORT:
1454 				optval = OPTBIT(INP_RECVDSTPORT);
1455 				break;
1456 			}
1457 			*mtod(m, int *) = optval;
1458 			break;
1459 
1460 		case IP_MULTICAST_IF:
1461 		case IP_MULTICAST_TTL:
1462 		case IP_MULTICAST_LOOP:
1463 		case IP_ADD_MEMBERSHIP:
1464 		case IP_DROP_MEMBERSHIP:
1465 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1466 			break;
1467 
1468 		case IP_PORTRANGE:
1469 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1470 			m->m_len = sizeof(int);
1471 
1472 			if (inp->inp_flags & INP_HIGHPORT)
1473 				optval = IP_PORTRANGE_HIGH;
1474 			else if (inp->inp_flags & INP_LOWPORT)
1475 				optval = IP_PORTRANGE_LOW;
1476 			else
1477 				optval = 0;
1478 
1479 			*mtod(m, int *) = optval;
1480 			break;
1481 
1482 		case IP_AUTH_LEVEL:
1483 		case IP_ESP_TRANS_LEVEL:
1484 		case IP_ESP_NETWORK_LEVEL:
1485 		case IP_IPCOMP_LEVEL:
1486 #ifndef IPSEC
1487 			m->m_len = sizeof(int);
1488 			*mtod(m, int *) = IPSEC_LEVEL_NONE;
1489 #else
1490 			m->m_len = sizeof(int);
1491 			switch (optname) {
1492 			case IP_AUTH_LEVEL:
1493 				optval = inp->inp_seclevel[SL_AUTH];
1494 				break;
1495 
1496 			case IP_ESP_TRANS_LEVEL:
1497 				optval = inp->inp_seclevel[SL_ESP_TRANS];
1498 				break;
1499 
1500 			case IP_ESP_NETWORK_LEVEL:
1501 				optval = inp->inp_seclevel[SL_ESP_NETWORK];
1502 				break;
1503 			case IP_IPCOMP_LEVEL:
1504 				optval = inp->inp_seclevel[SL_IPCOMP];
1505 				break;
1506 			}
1507 			*mtod(m, int *) = optval;
1508 #endif
1509 			break;
1510 		case IP_IPSEC_LOCAL_ID:
1511 		case IP_IPSEC_REMOTE_ID:
1512 		case IP_IPSEC_LOCAL_CRED:
1513 		case IP_IPSEC_REMOTE_CRED:
1514 		case IP_IPSEC_LOCAL_AUTH:
1515 		case IP_IPSEC_REMOTE_AUTH:
1516 #ifndef IPSEC
1517 			error = EOPNOTSUPP;
1518 #else
1519 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1520 			m->m_len = sizeof(u_int16_t);
1521 			ipr = NULL;
1522 			switch (optname) {
1523 			case IP_IPSEC_LOCAL_ID:
1524 				if (inp->inp_ipo != NULL)
1525 					ipr = inp->inp_ipo->ipo_srcid;
1526 				opt16val = IPSP_IDENTITY_NONE;
1527 				break;
1528 			case IP_IPSEC_REMOTE_ID:
1529 				if (inp->inp_ipo != NULL)
1530 					ipr = inp->inp_ipo->ipo_dstid;
1531 				opt16val = IPSP_IDENTITY_NONE;
1532 				break;
1533 			case IP_IPSEC_LOCAL_CRED:
1534 				if (inp->inp_ipo != NULL)
1535 					ipr = inp->inp_ipo->ipo_local_cred;
1536 				opt16val = IPSP_CRED_NONE;
1537 				break;
1538 			case IP_IPSEC_REMOTE_CRED:
1539 				ipr = inp->inp_ipsec_remotecred;
1540 				opt16val = IPSP_CRED_NONE;
1541 				break;
1542 			case IP_IPSEC_LOCAL_AUTH:
1543 				if (inp->inp_ipo != NULL)
1544 					ipr = inp->inp_ipo->ipo_local_auth;
1545 				break;
1546 			case IP_IPSEC_REMOTE_AUTH:
1547 				ipr = inp->inp_ipsec_remoteauth;
1548 				break;
1549 			}
1550 			if (ipr == NULL)
1551 				*mtod(m, u_int16_t *) = opt16val;
1552 			else {
1553 				size_t len;
1554 
1555 				len = m->m_len + ipr->ref_len;
1556 				if (len > MCLBYTES) {
1557 					 m_free(m);
1558 					 error = EINVAL;
1559 					 break;
1560 				}
1561 				/* allocate mbuf cluster for larger option */
1562 				if (len > MLEN) {
1563 					 MCLGET(m, M_WAITOK);
1564 					 if ((m->m_flags & M_EXT) == 0) {
1565 						 m_free(m);
1566 						 error = ENOBUFS;
1567 						 break;
1568 					 }
1569 
1570 				}
1571 				m->m_len = len;
1572 				*mtod(m, u_int16_t *) = ipr->ref_type;
1573 				m_copyback(m, sizeof(u_int16_t), ipr->ref_len,
1574 				    ipr + 1);
1575 			}
1576 #endif
1577 			break;
1578 		default:
1579 			error = ENOPROTOOPT;
1580 			break;
1581 		}
1582 		break;
1583 	}
1584 	return (error);
1585 }
1586 
1587 /*
1588  * Set up IP options in pcb for insertion in output packets.
1589  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1590  * with destination address if source routed.
1591  */
1592 int
1593 #ifdef notyet
1594 ip_pcbopts(optname, pcbopt, m)
1595 	int optname;
1596 #else
1597 ip_pcbopts(pcbopt, m)
1598 #endif
1599 	struct mbuf **pcbopt;
1600 	struct mbuf *m;
1601 {
1602 	int cnt, optlen;
1603 	u_char *cp;
1604 	u_char opt;
1605 
1606 	/* turn off any old options */
1607 	if (*pcbopt)
1608 		(void)m_free(*pcbopt);
1609 	*pcbopt = 0;
1610 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1611 		/*
1612 		 * Only turning off any previous options.
1613 		 */
1614 		if (m)
1615 			(void)m_free(m);
1616 		return (0);
1617 	}
1618 
1619 	if (m->m_len % sizeof(int32_t))
1620 		goto bad;
1621 
1622 	/*
1623 	 * IP first-hop destination address will be stored before
1624 	 * actual options; move other options back
1625 	 * and clear it when none present.
1626 	 */
1627 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1628 		goto bad;
1629 	cnt = m->m_len;
1630 	m->m_len += sizeof(struct in_addr);
1631 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1632 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1633 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1634 
1635 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1636 		opt = cp[IPOPT_OPTVAL];
1637 		if (opt == IPOPT_EOL)
1638 			break;
1639 		if (opt == IPOPT_NOP)
1640 			optlen = 1;
1641 		else {
1642 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1643 				goto bad;
1644 			optlen = cp[IPOPT_OLEN];
1645 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1646 				goto bad;
1647 		}
1648 		switch (opt) {
1649 
1650 		default:
1651 			break;
1652 
1653 		case IPOPT_LSRR:
1654 		case IPOPT_SSRR:
1655 			/*
1656 			 * user process specifies route as:
1657 			 *	->A->B->C->D
1658 			 * D must be our final destination (but we can't
1659 			 * check that since we may not have connected yet).
1660 			 * A is first hop destination, which doesn't appear in
1661 			 * actual IP option, but is stored before the options.
1662 			 */
1663 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1664 				goto bad;
1665 			m->m_len -= sizeof(struct in_addr);
1666 			cnt -= sizeof(struct in_addr);
1667 			optlen -= sizeof(struct in_addr);
1668 			cp[IPOPT_OLEN] = optlen;
1669 			/*
1670 			 * Move first hop before start of options.
1671 			 */
1672 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1673 			    sizeof(struct in_addr));
1674 			/*
1675 			 * Then copy rest of options back
1676 			 * to close up the deleted entry.
1677 			 */
1678 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1679 			    sizeof(struct in_addr)),
1680 			    (caddr_t)&cp[IPOPT_OFFSET+1],
1681 			    (unsigned)cnt - (IPOPT_OFFSET+1));
1682 			break;
1683 		}
1684 	}
1685 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1686 		goto bad;
1687 	*pcbopt = m;
1688 	return (0);
1689 
1690 bad:
1691 	(void)m_free(m);
1692 	return (EINVAL);
1693 }
1694 
1695 /*
1696  * Set the IP multicast options in response to user setsockopt().
1697  */
1698 int
1699 ip_setmoptions(optname, imop, m)
1700 	int optname;
1701 	struct ip_moptions **imop;
1702 	struct mbuf *m;
1703 {
1704 	int error = 0;
1705 	u_char loop;
1706 	int i;
1707 	struct in_addr addr;
1708 	struct ip_mreq *mreq;
1709 	struct ifnet *ifp;
1710 	struct ip_moptions *imo = *imop;
1711 	struct in_multi **immp;
1712 	struct route ro;
1713 	struct sockaddr_in *dst;
1714 
1715 	if (imo == NULL) {
1716 		/*
1717 		 * No multicast option buffer attached to the pcb;
1718 		 * allocate one and initialize to default values.
1719 		 */
1720 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1721 		    M_WAITOK|M_ZERO);
1722 		immp = (struct in_multi **)malloc(
1723 		    (sizeof(*immp) * IP_MIN_MEMBERSHIPS), M_IPMOPTS,
1724 		    M_WAITOK|M_ZERO);
1725 		*imop = imo;
1726 		imo->imo_multicast_ifp = NULL;
1727 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1728 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1729 		imo->imo_num_memberships = 0;
1730 		imo->imo_max_memberships = IP_MIN_MEMBERSHIPS;
1731 		imo->imo_membership = immp;
1732 	}
1733 
1734 	switch (optname) {
1735 
1736 	case IP_MULTICAST_IF:
1737 		/*
1738 		 * Select the interface for outgoing multicast packets.
1739 		 */
1740 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1741 			error = EINVAL;
1742 			break;
1743 		}
1744 		addr = *(mtod(m, struct in_addr *));
1745 		/*
1746 		 * INADDR_ANY is used to remove a previous selection.
1747 		 * When no interface is selected, a default one is
1748 		 * chosen every time a multicast packet is sent.
1749 		 */
1750 		if (addr.s_addr == INADDR_ANY) {
1751 			imo->imo_multicast_ifp = NULL;
1752 			break;
1753 		}
1754 		/*
1755 		 * The selected interface is identified by its local
1756 		 * IP address.  Find the interface and confirm that
1757 		 * it supports multicasting.
1758 		 */
1759 		INADDR_TO_IFP(addr, ifp);
1760 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1761 			error = EADDRNOTAVAIL;
1762 			break;
1763 		}
1764 		imo->imo_multicast_ifp = ifp;
1765 		break;
1766 
1767 	case IP_MULTICAST_TTL:
1768 		/*
1769 		 * Set the IP time-to-live for outgoing multicast packets.
1770 		 */
1771 		if (m == NULL || m->m_len != 1) {
1772 			error = EINVAL;
1773 			break;
1774 		}
1775 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1776 		break;
1777 
1778 	case IP_MULTICAST_LOOP:
1779 		/*
1780 		 * Set the loopback flag for outgoing multicast packets.
1781 		 * Must be zero or one.
1782 		 */
1783 		if (m == NULL || m->m_len != 1 ||
1784 		   (loop = *(mtod(m, u_char *))) > 1) {
1785 			error = EINVAL;
1786 			break;
1787 		}
1788 		imo->imo_multicast_loop = loop;
1789 		break;
1790 
1791 	case IP_ADD_MEMBERSHIP:
1792 		/*
1793 		 * Add a multicast group membership.
1794 		 * Group must be a valid IP multicast address.
1795 		 */
1796 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1797 			error = EINVAL;
1798 			break;
1799 		}
1800 		mreq = mtod(m, struct ip_mreq *);
1801 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1802 			error = EINVAL;
1803 			break;
1804 		}
1805 		/*
1806 		 * If no interface address was provided, use the interface of
1807 		 * the route to the given multicast address.
1808 		 */
1809 		if (mreq->imr_interface.s_addr == INADDR_ANY) {
1810 			ro.ro_rt = NULL;
1811 			dst = satosin(&ro.ro_dst);
1812 			dst->sin_len = sizeof(*dst);
1813 			dst->sin_family = AF_INET;
1814 			dst->sin_addr = mreq->imr_multiaddr;
1815 			rtalloc(&ro);
1816 			if (ro.ro_rt == NULL) {
1817 				error = EADDRNOTAVAIL;
1818 				break;
1819 			}
1820 			ifp = ro.ro_rt->rt_ifp;
1821 			rtfree(ro.ro_rt);
1822 		} else {
1823 			INADDR_TO_IFP(mreq->imr_interface, ifp);
1824 		}
1825 		/*
1826 		 * See if we found an interface, and confirm that it
1827 		 * supports multicast.
1828 		 */
1829 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1830 			error = EADDRNOTAVAIL;
1831 			break;
1832 		}
1833 		/*
1834 		 * See if the membership already exists or if all the
1835 		 * membership slots are full.
1836 		 */
1837 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1838 			if (imo->imo_membership[i]->inm_ia->ia_ifp == ifp &&
1839 			    imo->imo_membership[i]->inm_addr.s_addr
1840 						== mreq->imr_multiaddr.s_addr)
1841 				break;
1842 		}
1843 		if (i < imo->imo_num_memberships) {
1844 			error = EADDRINUSE;
1845 			break;
1846 		}
1847 		if (imo->imo_num_memberships == imo->imo_max_memberships) {
1848 			struct in_multi **nmships, **omships;
1849 			size_t newmax;
1850 			/*
1851 			 * Resize the vector to next power-of-two minus 1. If the
1852 			 * size would exceed the maximum then we know we've really
1853 			 * run out of entries. Otherwise, we reallocate the vector.
1854 			 */
1855 			nmships = NULL;
1856 			omships = imo->imo_membership;
1857 			newmax = ((imo->imo_max_memberships + 1) * 2) - 1;
1858 			if (newmax <= IP_MAX_MEMBERSHIPS) {
1859 				nmships = (struct in_multi **)malloc(
1860 				    sizeof(*nmships) * newmax, M_IPMOPTS,
1861 				    M_NOWAIT|M_ZERO);
1862 				if (nmships != NULL) {
1863 					bcopy(omships, nmships,
1864 					    sizeof(*omships) *
1865 					    imo->imo_max_memberships);
1866 					free(omships, M_IPMOPTS);
1867 					imo->imo_membership = nmships;
1868 					imo->imo_max_memberships = newmax;
1869 				}
1870 			}
1871 			if (nmships == NULL) {
1872 				error = ETOOMANYREFS;
1873 				break;
1874 			}
1875 		}
1876 		/*
1877 		 * Everything looks good; add a new record to the multicast
1878 		 * address list for the given interface.
1879 		 */
1880 		if ((imo->imo_membership[i] =
1881 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1882 			error = ENOBUFS;
1883 			break;
1884 		}
1885 		++imo->imo_num_memberships;
1886 		break;
1887 
1888 	case IP_DROP_MEMBERSHIP:
1889 		/*
1890 		 * Drop a multicast group membership.
1891 		 * Group must be a valid IP multicast address.
1892 		 */
1893 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1894 			error = EINVAL;
1895 			break;
1896 		}
1897 		mreq = mtod(m, struct ip_mreq *);
1898 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1899 			error = EINVAL;
1900 			break;
1901 		}
1902 		/*
1903 		 * If an interface address was specified, get a pointer
1904 		 * to its ifnet structure.
1905 		 */
1906 		if (mreq->imr_interface.s_addr == INADDR_ANY)
1907 			ifp = NULL;
1908 		else {
1909 			INADDR_TO_IFP(mreq->imr_interface, ifp);
1910 			if (ifp == NULL) {
1911 				error = EADDRNOTAVAIL;
1912 				break;
1913 			}
1914 		}
1915 		/*
1916 		 * Find the membership in the membership array.
1917 		 */
1918 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1919 			if ((ifp == NULL ||
1920 			     imo->imo_membership[i]->inm_ia->ia_ifp == ifp) &&
1921 			     imo->imo_membership[i]->inm_addr.s_addr ==
1922 			     mreq->imr_multiaddr.s_addr)
1923 				break;
1924 		}
1925 		if (i == imo->imo_num_memberships) {
1926 			error = EADDRNOTAVAIL;
1927 			break;
1928 		}
1929 		/*
1930 		 * Give up the multicast address record to which the
1931 		 * membership points.
1932 		 */
1933 		in_delmulti(imo->imo_membership[i]);
1934 		/*
1935 		 * Remove the gap in the membership array.
1936 		 */
1937 		for (++i; i < imo->imo_num_memberships; ++i)
1938 			imo->imo_membership[i-1] = imo->imo_membership[i];
1939 		--imo->imo_num_memberships;
1940 		break;
1941 
1942 	default:
1943 		error = EOPNOTSUPP;
1944 		break;
1945 	}
1946 
1947 	/*
1948 	 * If all options have default values, no need to keep the data.
1949 	 */
1950 	if (imo->imo_multicast_ifp == NULL &&
1951 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1952 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1953 	    imo->imo_num_memberships == 0) {
1954 		free(imo->imo_membership , M_IPMOPTS);
1955 		free(*imop, M_IPMOPTS);
1956 		*imop = NULL;
1957 	}
1958 
1959 	return (error);
1960 }
1961 
1962 /*
1963  * Return the IP multicast options in response to user getsockopt().
1964  */
1965 int
1966 ip_getmoptions(optname, imo, mp)
1967 	int optname;
1968 	struct ip_moptions *imo;
1969 	struct mbuf **mp;
1970 {
1971 	u_char *ttl;
1972 	u_char *loop;
1973 	struct in_addr *addr;
1974 	struct in_ifaddr *ia;
1975 
1976 	*mp = m_get(M_WAIT, MT_SOOPTS);
1977 
1978 	switch (optname) {
1979 
1980 	case IP_MULTICAST_IF:
1981 		addr = mtod(*mp, struct in_addr *);
1982 		(*mp)->m_len = sizeof(struct in_addr);
1983 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1984 			addr->s_addr = INADDR_ANY;
1985 		else {
1986 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1987 			addr->s_addr = (ia == NULL) ? INADDR_ANY
1988 					: ia->ia_addr.sin_addr.s_addr;
1989 		}
1990 		return (0);
1991 
1992 	case IP_MULTICAST_TTL:
1993 		ttl = mtod(*mp, u_char *);
1994 		(*mp)->m_len = 1;
1995 		*ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
1996 				     : imo->imo_multicast_ttl;
1997 		return (0);
1998 
1999 	case IP_MULTICAST_LOOP:
2000 		loop = mtod(*mp, u_char *);
2001 		(*mp)->m_len = 1;
2002 		*loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
2003 				      : imo->imo_multicast_loop;
2004 		return (0);
2005 
2006 	default:
2007 		return (EOPNOTSUPP);
2008 	}
2009 }
2010 
2011 /*
2012  * Discard the IP multicast options.
2013  */
2014 void
2015 ip_freemoptions(imo)
2016 	struct ip_moptions *imo;
2017 {
2018 	int i;
2019 
2020 	if (imo != NULL) {
2021 		for (i = 0; i < imo->imo_num_memberships; ++i)
2022 			in_delmulti(imo->imo_membership[i]);
2023 		free(imo->imo_membership, M_IPMOPTS);
2024 		free(imo, M_IPMOPTS);
2025 	}
2026 }
2027 
2028 /*
2029  * Routine called from ip_output() to loop back a copy of an IP multicast
2030  * packet to the input queue of a specified interface.  Note that this
2031  * calls the output routine of the loopback "driver", but with an interface
2032  * pointer that might NOT be &loif -- easier than replicating that code here.
2033  */
2034 void
2035 ip_mloopback(ifp, m, dst)
2036 	struct ifnet *ifp;
2037 	struct mbuf *m;
2038 	struct sockaddr_in *dst;
2039 {
2040 	struct ip *ip;
2041 	struct mbuf *copym;
2042 
2043 	copym = m_copym2(m, 0, M_COPYALL, M_DONTWAIT);
2044 	if (copym != NULL) {
2045 		/*
2046 		 * We don't bother to fragment if the IP length is greater
2047 		 * than the interface's MTU.  Can this possibly matter?
2048 		 */
2049 		ip = mtod(copym, struct ip *);
2050 		ip->ip_sum = 0;
2051 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
2052 		(void) looutput(ifp, copym, sintosa(dst), NULL);
2053 	}
2054 }
2055 
2056 /*
2057  * Process a delayed payload checksum calculation.
2058  */
2059 void
2060 in_delayed_cksum(struct mbuf *m)
2061 {
2062 	struct ip *ip;
2063 	u_int16_t csum, offset;
2064 
2065 	ip = mtod(m, struct ip *);
2066 	offset = ip->ip_hl << 2;
2067 	csum = in4_cksum(m, 0, offset, m->m_pkthdr.len - offset);
2068 	if (csum == 0 && ip->ip_p == IPPROTO_UDP)
2069 		csum = 0xffff;
2070 
2071 	switch (ip->ip_p) {
2072 	case IPPROTO_TCP:
2073 		offset += offsetof(struct tcphdr, th_sum);
2074 		break;
2075 
2076 	case IPPROTO_UDP:
2077 		offset += offsetof(struct udphdr, uh_sum);
2078 		break;
2079 
2080 	default:
2081 		return;
2082 	}
2083 
2084 	if ((offset + sizeof(u_int16_t)) > m->m_len)
2085 		m_copyback(m, offset, sizeof(csum), &csum);
2086 	else
2087 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2088 }
2089