xref: /netbsd-src/sys/netinet/ip_input.c (revision 37b34d511dea595d3ba03a661cf3b775038ea5f8)
1 /*	$NetBSD: ip_input.c,v 1.158 2002/09/23 13:43:27 itojun Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Public Access Networks Corporation ("Panix").  It was developed under
38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 3. All advertising materials mentioning features or use of this software
49  *    must display the following acknowledgement:
50  *	This product includes software developed by the NetBSD
51  *	Foundation, Inc. and its contributors.
52  * 4. Neither the name of The NetBSD Foundation nor the names of its
53  *    contributors may be used to endorse or promote products derived
54  *    from this software without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
57  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
58  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
59  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
60  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
61  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
62  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
63  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
64  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
65  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66  * POSSIBILITY OF SUCH DAMAGE.
67  */
68 
69 /*
70  * Copyright (c) 1982, 1986, 1988, 1993
71  *	The Regents of the University of California.  All rights reserved.
72  *
73  * Redistribution and use in source and binary forms, with or without
74  * modification, are permitted provided that the following conditions
75  * are met:
76  * 1. Redistributions of source code must retain the above copyright
77  *    notice, this list of conditions and the following disclaimer.
78  * 2. Redistributions in binary form must reproduce the above copyright
79  *    notice, this list of conditions and the following disclaimer in the
80  *    documentation and/or other materials provided with the distribution.
81  * 3. All advertising materials mentioning features or use of this software
82  *    must display the following acknowledgement:
83  *	This product includes software developed by the University of
84  *	California, Berkeley and its contributors.
85  * 4. Neither the name of the University nor the names of its contributors
86  *    may be used to endorse or promote products derived from this software
87  *    without specific prior written permission.
88  *
89  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
90  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
91  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
92  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
93  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
94  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
95  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
96  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
97  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
98  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
99  * SUCH DAMAGE.
100  *
101  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
102  */
103 
104 #include <sys/cdefs.h>
105 __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.158 2002/09/23 13:43:27 itojun Exp $");
106 
107 #include "opt_gateway.h"
108 #include "opt_pfil_hooks.h"
109 #include "opt_ipsec.h"
110 #include "opt_mrouting.h"
111 #include "opt_inet_csum.h"
112 
113 #include <sys/param.h>
114 #include <sys/systm.h>
115 #include <sys/malloc.h>
116 #include <sys/mbuf.h>
117 #include <sys/domain.h>
118 #include <sys/protosw.h>
119 #include <sys/socket.h>
120 #include <sys/socketvar.h>
121 #include <sys/errno.h>
122 #include <sys/time.h>
123 #include <sys/kernel.h>
124 #include <sys/pool.h>
125 #include <sys/sysctl.h>
126 
127 #include <net/if.h>
128 #include <net/if_dl.h>
129 #include <net/route.h>
130 #include <net/pfil.h>
131 
132 #include <netinet/in.h>
133 #include <netinet/in_systm.h>
134 #include <netinet/ip.h>
135 #include <netinet/in_pcb.h>
136 #include <netinet/in_var.h>
137 #include <netinet/ip_var.h>
138 #include <netinet/ip_icmp.h>
139 /* just for gif_ttl */
140 #include <netinet/in_gif.h>
141 #include "gif.h"
142 #include <net/if_gre.h>
143 #include "gre.h"
144 
145 #ifdef MROUTING
146 #include <netinet/ip_mroute.h>
147 #endif
148 
149 #ifdef IPSEC
150 #include <netinet6/ipsec.h>
151 #include <netkey/key.h>
152 #endif
153 
154 #ifndef	IPFORWARDING
155 #ifdef GATEWAY
156 #define	IPFORWARDING	1	/* forward IP packets not for us */
157 #else /* GATEWAY */
158 #define	IPFORWARDING	0	/* don't forward IP packets not for us */
159 #endif /* GATEWAY */
160 #endif /* IPFORWARDING */
161 #ifndef	IPSENDREDIRECTS
162 #define	IPSENDREDIRECTS	1
163 #endif
164 #ifndef IPFORWSRCRT
165 #define	IPFORWSRCRT	1	/* forward source-routed packets */
166 #endif
167 #ifndef IPALLOWSRCRT
168 #define	IPALLOWSRCRT	1	/* allow source-routed packets */
169 #endif
170 #ifndef IPMTUDISC
171 #define IPMTUDISC	1
172 #endif
173 #ifndef IPMTUDISCTIMEOUT
174 #define IPMTUDISCTIMEOUT (10 * 60)	/* as per RFC 1191 */
175 #endif
176 
177 /*
178  * Note: DIRECTED_BROADCAST is handled this way so that previous
179  * configuration using this option will Just Work.
180  */
181 #ifndef IPDIRECTEDBCAST
182 #ifdef DIRECTED_BROADCAST
183 #define IPDIRECTEDBCAST	1
184 #else
185 #define	IPDIRECTEDBCAST	0
186 #endif /* DIRECTED_BROADCAST */
187 #endif /* IPDIRECTEDBCAST */
188 int	ipforwarding = IPFORWARDING;
189 int	ipsendredirects = IPSENDREDIRECTS;
190 int	ip_defttl = IPDEFTTL;
191 int	ip_forwsrcrt = IPFORWSRCRT;
192 int	ip_directedbcast = IPDIRECTEDBCAST;
193 int	ip_allowsrcrt = IPALLOWSRCRT;
194 int	ip_mtudisc = IPMTUDISC;
195 int	ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
196 #ifdef DIAGNOSTIC
197 int	ipprintfs = 0;
198 #endif
199 
200 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
201 
202 extern	struct domain inetdomain;
203 int	ipqmaxlen = IFQ_MAXLEN;
204 u_long	in_ifaddrhash;				/* size of hash table - 1 */
205 int	in_ifaddrentries;			/* total number of addrs */
206 struct	in_ifaddrhead in_ifaddr;
207 struct	in_ifaddrhashhead *in_ifaddrhashtbl;
208 struct	ifqueue ipintrq;
209 struct	ipstat	ipstat;
210 u_int16_t	ip_id;
211 
212 #ifdef PFIL_HOOKS
213 struct pfil_head inet_pfil_hook;
214 #endif
215 
216 struct ipqhead ipq;
217 int	ipq_locked;
218 int	ip_nfragpackets = 0;
219 int	ip_maxfragpackets = 200;
220 
221 static __inline int ipq_lock_try __P((void));
222 static __inline void ipq_unlock __P((void));
223 
224 static __inline int
225 ipq_lock_try()
226 {
227 	int s;
228 
229 	/*
230 	 * Use splvm() -- we're blocking things that would cause
231 	 * mbuf allocation.
232 	 */
233 	s = splvm();
234 	if (ipq_locked) {
235 		splx(s);
236 		return (0);
237 	}
238 	ipq_locked = 1;
239 	splx(s);
240 	return (1);
241 }
242 
243 static __inline void
244 ipq_unlock()
245 {
246 	int s;
247 
248 	s = splvm();
249 	ipq_locked = 0;
250 	splx(s);
251 }
252 
253 #ifdef DIAGNOSTIC
254 #define	IPQ_LOCK()							\
255 do {									\
256 	if (ipq_lock_try() == 0) {					\
257 		printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \
258 		panic("ipq_lock");					\
259 	}								\
260 } while (0)
261 #define	IPQ_LOCK_CHECK()						\
262 do {									\
263 	if (ipq_locked == 0) {						\
264 		printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \
265 		panic("ipq lock check");				\
266 	}								\
267 } while (0)
268 #else
269 #define	IPQ_LOCK()		(void) ipq_lock_try()
270 #define	IPQ_LOCK_CHECK()	/* nothing */
271 #endif
272 
273 #define	IPQ_UNLOCK()		ipq_unlock()
274 
275 struct pool ipqent_pool;
276 
277 #ifdef INET_CSUM_COUNTERS
278 #include <sys/device.h>
279 
280 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
281     NULL, "inet", "hwcsum bad");
282 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
283     NULL, "inet", "hwcsum ok");
284 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
285     NULL, "inet", "swcsum");
286 
287 #define	INET_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
288 
289 #else
290 
291 #define	INET_CSUM_COUNTER_INCR(ev)	/* nothing */
292 
293 #endif /* INET_CSUM_COUNTERS */
294 
295 /*
296  * We need to save the IP options in case a protocol wants to respond
297  * to an incoming packet over the same route if the packet got here
298  * using IP source routing.  This allows connection establishment and
299  * maintenance when the remote end is on a network that is not known
300  * to us.
301  */
302 int	ip_nhops = 0;
303 static	struct ip_srcrt {
304 	struct	in_addr dst;			/* final destination */
305 	char	nop;				/* one NOP to align */
306 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
307 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
308 } ip_srcrt;
309 
310 static void save_rte __P((u_char *, struct in_addr));
311 
312 /*
313  * IP initialization: fill in IP protocol switch table.
314  * All protocols not implemented in kernel go to raw IP protocol handler.
315  */
316 void
317 ip_init()
318 {
319 	struct protosw *pr;
320 	int i;
321 
322 	pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl",
323 	    NULL);
324 
325 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
326 	if (pr == 0)
327 		panic("ip_init");
328 	for (i = 0; i < IPPROTO_MAX; i++)
329 		ip_protox[i] = pr - inetsw;
330 	for (pr = inetdomain.dom_protosw;
331 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
332 		if (pr->pr_domain->dom_family == PF_INET &&
333 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
334 			ip_protox[pr->pr_protocol] = pr - inetsw;
335 	LIST_INIT(&ipq);
336 	ip_id = time.tv_sec & 0xffff;
337 	ipintrq.ifq_maxlen = ipqmaxlen;
338 	TAILQ_INIT(&in_ifaddr);
339 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR,
340 	    M_WAITOK, &in_ifaddrhash);
341 	if (ip_mtudisc != 0)
342 		ip_mtudisc_timeout_q =
343 		    rt_timer_queue_create(ip_mtudisc_timeout);
344 #ifdef GATEWAY
345 	ipflow_init();
346 #endif
347 
348 #ifdef PFIL_HOOKS
349 	/* Register our Packet Filter hook. */
350 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
351 	inet_pfil_hook.ph_af   = AF_INET;
352 	i = pfil_head_register(&inet_pfil_hook);
353 	if (i != 0)
354 		printf("ip_init: WARNING: unable to register pfil hook, "
355 		    "error %d\n", i);
356 #endif /* PFIL_HOOKS */
357 
358 #ifdef INET_CSUM_COUNTERS
359 	evcnt_attach_static(&ip_hwcsum_bad);
360 	evcnt_attach_static(&ip_hwcsum_ok);
361 	evcnt_attach_static(&ip_swcsum);
362 #endif /* INET_CSUM_COUNTERS */
363 }
364 
365 struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
366 struct	route ipforward_rt;
367 
368 /*
369  * IP software interrupt routine
370  */
371 void
372 ipintr()
373 {
374 	int s;
375 	struct mbuf *m;
376 
377 	while (1) {
378 		s = splnet();
379 		IF_DEQUEUE(&ipintrq, m);
380 		splx(s);
381 		if (m == 0)
382 			return;
383 		ip_input(m);
384 	}
385 }
386 
387 /*
388  * Ip input routine.  Checksum and byte swap header.  If fragmented
389  * try to reassemble.  Process options.  Pass to next level.
390  */
391 void
392 ip_input(struct mbuf *m)
393 {
394 	struct ip *ip = NULL;
395 	struct ipq *fp;
396 	struct in_ifaddr *ia;
397 	struct ifaddr *ifa;
398 	struct ipqent *ipqe;
399 	int hlen = 0, mff, len;
400 	int downmatch;
401 
402 #ifdef	DIAGNOSTIC
403 	if ((m->m_flags & M_PKTHDR) == 0)
404 		panic("ipintr no HDR");
405 #endif
406 #ifdef IPSEC
407 	/*
408 	 * should the inner packet be considered authentic?
409 	 * see comment in ah4_input().
410 	 */
411 	if (m) {
412 		m->m_flags &= ~M_AUTHIPHDR;
413 		m->m_flags &= ~M_AUTHIPDGM;
414 	}
415 #endif
416 	/*
417 	 * If no IP addresses have been set yet but the interfaces
418 	 * are receiving, can't do anything with incoming packets yet.
419 	 */
420 	if (TAILQ_FIRST(&in_ifaddr) == 0)
421 		goto bad;
422 	ipstat.ips_total++;
423 	/*
424 	 * If the IP header is not aligned, slurp it up into a new
425 	 * mbuf with space for link headers, in the event we forward
426 	 * it.  Otherwise, if it is aligned, make sure the entire
427 	 * base IP header is in the first mbuf of the chain.
428 	 */
429 	if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
430 		if ((m = m_copyup(m, sizeof(struct ip),
431 				  (max_linkhdr + 3) & ~3)) == NULL) {
432 			/* XXXJRT new stat, please */
433 			ipstat.ips_toosmall++;
434 			return;
435 		}
436 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
437 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
438 			ipstat.ips_toosmall++;
439 			return;
440 		}
441 	}
442 	ip = mtod(m, struct ip *);
443 	if (ip->ip_v != IPVERSION) {
444 		ipstat.ips_badvers++;
445 		goto bad;
446 	}
447 	hlen = ip->ip_hl << 2;
448 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
449 		ipstat.ips_badhlen++;
450 		goto bad;
451 	}
452 	if (hlen > m->m_len) {
453 		if ((m = m_pullup(m, hlen)) == 0) {
454 			ipstat.ips_badhlen++;
455 			return;
456 		}
457 		ip = mtod(m, struct ip *);
458 	}
459 
460 	/*
461 	 * RFC1122: packets with a multicast source address are
462 	 * not allowed.
463 	 */
464 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
465 		ipstat.ips_badaddr++;
466 		goto bad;
467 	}
468 
469 	/* 127/8 must not appear on wire - RFC1122 */
470 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
471 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
472 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
473 			ipstat.ips_badaddr++;
474 			goto bad;
475 		}
476 	}
477 
478 	switch (m->m_pkthdr.csum_flags &
479 		((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
480 		 M_CSUM_IPv4_BAD)) {
481 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
482 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
483 		goto badcsum;
484 
485 	case M_CSUM_IPv4:
486 		/* Checksum was okay. */
487 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
488 		break;
489 
490 	default:
491 		/* Must compute it ourselves. */
492 		INET_CSUM_COUNTER_INCR(&ip_swcsum);
493 		if (in_cksum(m, hlen) != 0)
494 			goto bad;
495 		break;
496 	}
497 
498 	/* Retrieve the packet length. */
499 	len = ntohs(ip->ip_len);
500 
501 	/*
502 	 * Check for additional length bogosity
503 	 */
504 	if (len < hlen) {
505 	 	ipstat.ips_badlen++;
506 		goto bad;
507 	}
508 
509 	/*
510 	 * Check that the amount of data in the buffers
511 	 * is as at least much as the IP header would have us expect.
512 	 * Trim mbufs if longer than we expect.
513 	 * Drop packet if shorter than we expect.
514 	 */
515 	if (m->m_pkthdr.len < len) {
516 		ipstat.ips_tooshort++;
517 		goto bad;
518 	}
519 	if (m->m_pkthdr.len > len) {
520 		if (m->m_len == m->m_pkthdr.len) {
521 			m->m_len = len;
522 			m->m_pkthdr.len = len;
523 		} else
524 			m_adj(m, len - m->m_pkthdr.len);
525 	}
526 
527 #ifdef IPSEC
528 	/* ipflow (IP fast forwarding) is not compatible with IPsec. */
529 	m->m_flags &= ~M_CANFASTFWD;
530 #else
531 	/*
532 	 * Assume that we can create a fast-forward IP flow entry
533 	 * based on this packet.
534 	 */
535 	m->m_flags |= M_CANFASTFWD;
536 #endif
537 
538 #ifdef PFIL_HOOKS
539 	/*
540 	 * Run through list of hooks for input packets.  If there are any
541 	 * filters which require that additional packets in the flow are
542 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
543 	 * Note that filters must _never_ set this flag, as another filter
544 	 * in the list may have previously cleared it.
545 	 */
546 	/*
547 	 * let ipfilter look at packet on the wire,
548 	 * not the decapsulated packet.
549 	 */
550 #ifdef IPSEC
551 	if (!ipsec_getnhist(m))
552 #else
553 	if (1)
554 #endif
555 	{
556 		if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
557 				   PFIL_IN) != 0)
558 		return;
559 		if (m == NULL)
560 			return;
561 		ip = mtod(m, struct ip *);
562 		hlen = ip->ip_hl << 2;
563 	}
564 #endif /* PFIL_HOOKS */
565 
566 #ifdef ALTQ
567 	/* XXX Temporary until ALTQ is changed to use a pfil hook */
568 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
569 		/* packet dropped by traffic conditioner */
570 		return;
571 	}
572 #endif
573 
574 	/*
575 	 * Process options and, if not destined for us,
576 	 * ship it on.  ip_dooptions returns 1 when an
577 	 * error was detected (causing an icmp message
578 	 * to be sent and the original packet to be freed).
579 	 */
580 	ip_nhops = 0;		/* for source routed packets */
581 	if (hlen > sizeof (struct ip) && ip_dooptions(m))
582 		return;
583 
584 	/*
585 	 * Check our list of addresses, to see if the packet is for us.
586 	 *
587 	 * Traditional 4.4BSD did not consult IFF_UP at all.
588 	 * The behavior here is to treat addresses on !IFF_UP interface
589 	 * as not mine.
590 	 */
591 	downmatch = 0;
592 	LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
593 		if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
594 			if ((ia->ia_ifp->if_flags & IFF_UP) != 0)
595 				break;
596 			else
597 				downmatch++;
598 		}
599 	}
600 	if (ia != NULL)
601 		goto ours;
602 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
603 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) {
604 			if (ifa->ifa_addr->sa_family != AF_INET)
605 				continue;
606 			ia = ifatoia(ifa);
607 			if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
608 			    in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
609 			    /*
610 			     * Look for all-0's host part (old broadcast addr),
611 			     * either for subnet or net.
612 			     */
613 			    ip->ip_dst.s_addr == ia->ia_subnet ||
614 			    ip->ip_dst.s_addr == ia->ia_net)
615 				goto ours;
616 			/*
617 			 * An interface with IP address zero accepts
618 			 * all packets that arrive on that interface.
619 			 */
620 			if (in_nullhost(ia->ia_addr.sin_addr))
621 				goto ours;
622 		}
623 	}
624 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
625 		struct in_multi *inm;
626 #ifdef MROUTING
627 		extern struct socket *ip_mrouter;
628 
629 		if (M_READONLY(m)) {
630 			if ((m = m_pullup(m, hlen)) == 0) {
631 				ipstat.ips_toosmall++;
632 				return;
633 			}
634 			ip = mtod(m, struct ip *);
635 		}
636 
637 		if (ip_mrouter) {
638 			/*
639 			 * If we are acting as a multicast router, all
640 			 * incoming multicast packets are passed to the
641 			 * kernel-level multicast forwarding function.
642 			 * The packet is returned (relatively) intact; if
643 			 * ip_mforward() returns a non-zero value, the packet
644 			 * must be discarded, else it may be accepted below.
645 			 *
646 			 * (The IP ident field is put in the same byte order
647 			 * as expected when ip_mforward() is called from
648 			 * ip_output().)
649 			 */
650 			if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
651 				ipstat.ips_cantforward++;
652 				m_freem(m);
653 				return;
654 			}
655 
656 			/*
657 			 * The process-level routing demon needs to receive
658 			 * all multicast IGMP packets, whether or not this
659 			 * host belongs to their destination groups.
660 			 */
661 			if (ip->ip_p == IPPROTO_IGMP)
662 				goto ours;
663 			ipstat.ips_forward++;
664 		}
665 #endif
666 		/*
667 		 * See if we belong to the destination multicast group on the
668 		 * arrival interface.
669 		 */
670 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
671 		if (inm == NULL) {
672 			ipstat.ips_cantforward++;
673 			m_freem(m);
674 			return;
675 		}
676 		goto ours;
677 	}
678 	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
679 	    in_nullhost(ip->ip_dst))
680 		goto ours;
681 
682 	/*
683 	 * Not for us; forward if possible and desirable.
684 	 */
685 	if (ipforwarding == 0) {
686 		ipstat.ips_cantforward++;
687 		m_freem(m);
688 	} else {
689 		/*
690 		 * If ip_dst matched any of my address on !IFF_UP interface,
691 		 * and there's no IFF_UP interface that matches ip_dst,
692 		 * send icmp unreach.  Forwarding it will result in in-kernel
693 		 * forwarding loop till TTL goes to 0.
694 		 */
695 		if (downmatch) {
696 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
697 			ipstat.ips_cantforward++;
698 			return;
699 		}
700 #ifdef IPSEC
701 		if (ipsec4_in_reject(m, NULL)) {
702 			ipsecstat.in_polvio++;
703 			goto bad;
704 		}
705 #endif
706 
707 		ip_forward(m, 0);
708 	}
709 	return;
710 
711 ours:
712 	/*
713 	 * If offset or IP_MF are set, must reassemble.
714 	 * Otherwise, nothing need be done.
715 	 * (We could look in the reassembly queue to see
716 	 * if the packet was previously fragmented,
717 	 * but it's not worth the time; just let them time out.)
718 	 */
719 	if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
720 		if (M_READONLY(m)) {
721 			if ((m = m_pullup(m, hlen)) == NULL) {
722 				ipstat.ips_toosmall++;
723 				goto bad;
724 			}
725 			ip = mtod(m, struct ip *);
726 		}
727 
728 		/*
729 		 * Look for queue of fragments
730 		 * of this datagram.
731 		 */
732 		IPQ_LOCK();
733 		LIST_FOREACH(fp, &ipq, ipq_q)
734 			if (ip->ip_id == fp->ipq_id &&
735 			    in_hosteq(ip->ip_src, fp->ipq_src) &&
736 			    in_hosteq(ip->ip_dst, fp->ipq_dst) &&
737 			    ip->ip_p == fp->ipq_p)
738 				goto found;
739 		fp = 0;
740 found:
741 
742 		/*
743 		 * Adjust ip_len to not reflect header,
744 		 * set ipqe_mff if more fragments are expected,
745 		 * convert offset of this to bytes.
746 		 */
747 		ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
748 		mff = (ip->ip_off & htons(IP_MF)) != 0;
749 		if (mff) {
750 		        /*
751 		         * Make sure that fragments have a data length
752 			 * that's a non-zero multiple of 8 bytes.
753 		         */
754 			if (ntohs(ip->ip_len) == 0 ||
755 			    (ntohs(ip->ip_len) & 0x7) != 0) {
756 				ipstat.ips_badfrags++;
757 				IPQ_UNLOCK();
758 				goto bad;
759 			}
760 		}
761 		ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3);
762 
763 		/*
764 		 * If datagram marked as having more fragments
765 		 * or if this is not the first fragment,
766 		 * attempt reassembly; if it succeeds, proceed.
767 		 */
768 		if (mff || ip->ip_off != htons(0)) {
769 			ipstat.ips_fragments++;
770 			ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
771 			if (ipqe == NULL) {
772 				ipstat.ips_rcvmemdrop++;
773 				IPQ_UNLOCK();
774 				goto bad;
775 			}
776 			ipqe->ipqe_mff = mff;
777 			ipqe->ipqe_m = m;
778 			ipqe->ipqe_ip = ip;
779 			m = ip_reass(ipqe, fp);
780 			if (m == 0) {
781 				IPQ_UNLOCK();
782 				return;
783 			}
784 			ipstat.ips_reassembled++;
785 			ip = mtod(m, struct ip *);
786 			hlen = ip->ip_hl << 2;
787 			ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
788 		} else
789 			if (fp)
790 				ip_freef(fp);
791 		IPQ_UNLOCK();
792 	}
793 
794 #ifdef IPSEC
795 	/*
796 	 * enforce IPsec policy checking if we are seeing last header.
797 	 * note that we do not visit this with protocols with pcb layer
798 	 * code - like udp/tcp/raw ip.
799 	 */
800 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
801 	    ipsec4_in_reject(m, NULL)) {
802 		ipsecstat.in_polvio++;
803 		goto bad;
804 	}
805 #endif
806 
807 	/*
808 	 * Switch out to protocol's input routine.
809 	 */
810 #if IFA_STATS
811 	if (ia && ip)
812 		ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
813 #endif
814 	ipstat.ips_delivered++;
815     {
816 	int off = hlen, nh = ip->ip_p;
817 
818 	(*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
819 	return;
820     }
821 bad:
822 	m_freem(m);
823 	return;
824 
825 badcsum:
826 	ipstat.ips_badsum++;
827 	m_freem(m);
828 }
829 
830 /*
831  * Take incoming datagram fragment and try to
832  * reassemble it into whole datagram.  If a chain for
833  * reassembly of this datagram already exists, then it
834  * is given as fp; otherwise have to make a chain.
835  */
836 struct mbuf *
837 ip_reass(ipqe, fp)
838 	struct ipqent *ipqe;
839 	struct ipq *fp;
840 {
841 	struct mbuf *m = ipqe->ipqe_m;
842 	struct ipqent *nq, *p, *q;
843 	struct ip *ip;
844 	struct mbuf *t;
845 	int hlen = ipqe->ipqe_ip->ip_hl << 2;
846 	int i, next;
847 
848 	IPQ_LOCK_CHECK();
849 
850 	/*
851 	 * Presence of header sizes in mbufs
852 	 * would confuse code below.
853 	 */
854 	m->m_data += hlen;
855 	m->m_len -= hlen;
856 
857 	/*
858 	 * If first fragment to arrive, create a reassembly queue.
859 	 */
860 	if (fp == 0) {
861 		/*
862 		 * Enforce upper bound on number of fragmented packets
863 		 * for which we attempt reassembly;
864 		 * If maxfrag is 0, never accept fragments.
865 		 * If maxfrag is -1, accept all fragments without limitation.
866 		 */
867 		if (ip_maxfragpackets < 0)
868 			;
869 		else if (ip_nfragpackets >= ip_maxfragpackets)
870 			goto dropfrag;
871 		ip_nfragpackets++;
872 		MALLOC(fp, struct ipq *, sizeof (struct ipq),
873 		    M_FTABLE, M_NOWAIT);
874 		if (fp == NULL)
875 			goto dropfrag;
876 		LIST_INSERT_HEAD(&ipq, fp, ipq_q);
877 		fp->ipq_ttl = IPFRAGTTL;
878 		fp->ipq_p = ipqe->ipqe_ip->ip_p;
879 		fp->ipq_id = ipqe->ipqe_ip->ip_id;
880 		TAILQ_INIT(&fp->ipq_fragq);
881 		fp->ipq_src = ipqe->ipqe_ip->ip_src;
882 		fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
883 		p = NULL;
884 		goto insert;
885 	}
886 
887 	/*
888 	 * Find a segment which begins after this one does.
889 	 */
890 	for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
891 	    p = q, q = TAILQ_NEXT(q, ipqe_q))
892 		if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
893 			break;
894 
895 	/*
896 	 * If there is a preceding segment, it may provide some of
897 	 * our data already.  If so, drop the data from the incoming
898 	 * segment.  If it provides all of our data, drop us.
899 	 */
900 	if (p != NULL) {
901 		i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
902 		    ntohs(ipqe->ipqe_ip->ip_off);
903 		if (i > 0) {
904 			if (i >= ntohs(ipqe->ipqe_ip->ip_len))
905 				goto dropfrag;
906 			m_adj(ipqe->ipqe_m, i);
907 			ipqe->ipqe_ip->ip_off =
908 			    htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
909 			ipqe->ipqe_ip->ip_len =
910 			    htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
911 		}
912 	}
913 
914 	/*
915 	 * While we overlap succeeding segments trim them or,
916 	 * if they are completely covered, dequeue them.
917 	 */
918 	for (; q != NULL &&
919 	    ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
920 	    ntohs(q->ipqe_ip->ip_off); q = nq) {
921 		i = (ntohs(ipqe->ipqe_ip->ip_off) +
922 		    ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
923 		if (i < ntohs(q->ipqe_ip->ip_len)) {
924 			q->ipqe_ip->ip_len =
925 			    htons(ntohs(q->ipqe_ip->ip_len) - i);
926 			q->ipqe_ip->ip_off =
927 			    htons(ntohs(q->ipqe_ip->ip_off) + i);
928 			m_adj(q->ipqe_m, i);
929 			break;
930 		}
931 		nq = TAILQ_NEXT(q, ipqe_q);
932 		m_freem(q->ipqe_m);
933 		TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
934 		pool_put(&ipqent_pool, q);
935 	}
936 
937 insert:
938 	/*
939 	 * Stick new segment in its place;
940 	 * check for complete reassembly.
941 	 */
942 	if (p == NULL) {
943 		TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
944 	} else {
945 		TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q);
946 	}
947 	next = 0;
948 	for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
949 	    p = q, q = TAILQ_NEXT(q, ipqe_q)) {
950 		if (ntohs(q->ipqe_ip->ip_off) != next)
951 			return (0);
952 		next += ntohs(q->ipqe_ip->ip_len);
953 	}
954 	if (p->ipqe_mff)
955 		return (0);
956 
957 	/*
958 	 * Reassembly is complete.  Check for a bogus message size and
959 	 * concatenate fragments.
960 	 */
961 	q = TAILQ_FIRST(&fp->ipq_fragq);
962 	ip = q->ipqe_ip;
963 	if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
964 		ipstat.ips_toolong++;
965 		ip_freef(fp);
966 		return (0);
967 	}
968 	m = q->ipqe_m;
969 	t = m->m_next;
970 	m->m_next = 0;
971 	m_cat(m, t);
972 	nq = TAILQ_NEXT(q, ipqe_q);
973 	pool_put(&ipqent_pool, q);
974 	for (q = nq; q != NULL; q = nq) {
975 		t = q->ipqe_m;
976 		nq = TAILQ_NEXT(q, ipqe_q);
977 		pool_put(&ipqent_pool, q);
978 		m_cat(m, t);
979 	}
980 
981 	/*
982 	 * Create header for new ip packet by
983 	 * modifying header of first packet;
984 	 * dequeue and discard fragment reassembly header.
985 	 * Make header visible.
986 	 */
987 	ip->ip_len = htons(next);
988 	ip->ip_src = fp->ipq_src;
989 	ip->ip_dst = fp->ipq_dst;
990 	LIST_REMOVE(fp, ipq_q);
991 	FREE(fp, M_FTABLE);
992 	ip_nfragpackets--;
993 	m->m_len += (ip->ip_hl << 2);
994 	m->m_data -= (ip->ip_hl << 2);
995 	/* some debugging cruft by sklower, below, will go away soon */
996 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
997 		int plen = 0;
998 		for (t = m; t; t = t->m_next)
999 			plen += t->m_len;
1000 		m->m_pkthdr.len = plen;
1001 	}
1002 	return (m);
1003 
1004 dropfrag:
1005 	ipstat.ips_fragdropped++;
1006 	m_freem(m);
1007 	pool_put(&ipqent_pool, ipqe);
1008 	return (0);
1009 }
1010 
1011 /*
1012  * Free a fragment reassembly header and all
1013  * associated datagrams.
1014  */
1015 void
1016 ip_freef(fp)
1017 	struct ipq *fp;
1018 {
1019 	struct ipqent *q, *p;
1020 
1021 	IPQ_LOCK_CHECK();
1022 
1023 	for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) {
1024 		p = TAILQ_NEXT(q, ipqe_q);
1025 		m_freem(q->ipqe_m);
1026 		TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
1027 		pool_put(&ipqent_pool, q);
1028 	}
1029 	LIST_REMOVE(fp, ipq_q);
1030 	FREE(fp, M_FTABLE);
1031 	ip_nfragpackets--;
1032 }
1033 
1034 /*
1035  * IP timer processing;
1036  * if a timer expires on a reassembly
1037  * queue, discard it.
1038  */
1039 void
1040 ip_slowtimo()
1041 {
1042 	struct ipq *fp, *nfp;
1043 	int s = splsoftnet();
1044 
1045 	IPQ_LOCK();
1046 	for (fp = LIST_FIRST(&ipq); fp != NULL; fp = nfp) {
1047 		nfp = LIST_NEXT(fp, ipq_q);
1048 		if (--fp->ipq_ttl == 0) {
1049 			ipstat.ips_fragtimeout++;
1050 			ip_freef(fp);
1051 		}
1052 	}
1053 	/*
1054 	 * If we are over the maximum number of fragments
1055 	 * (due to the limit being lowered), drain off
1056 	 * enough to get down to the new limit.
1057 	 */
1058 	if (ip_maxfragpackets < 0)
1059 		;
1060 	else {
1061 		while (ip_nfragpackets > ip_maxfragpackets && LIST_FIRST(&ipq))
1062 			ip_freef(LIST_FIRST(&ipq));
1063 	}
1064 	IPQ_UNLOCK();
1065 #ifdef GATEWAY
1066 	ipflow_slowtimo();
1067 #endif
1068 	splx(s);
1069 }
1070 
1071 /*
1072  * Drain off all datagram fragments.
1073  */
1074 void
1075 ip_drain()
1076 {
1077 
1078 	/*
1079 	 * We may be called from a device's interrupt context.  If
1080 	 * the ipq is already busy, just bail out now.
1081 	 */
1082 	if (ipq_lock_try() == 0)
1083 		return;
1084 
1085 	while (LIST_FIRST(&ipq) != NULL) {
1086 		ipstat.ips_fragdropped++;
1087 		ip_freef(LIST_FIRST(&ipq));
1088 	}
1089 
1090 	IPQ_UNLOCK();
1091 }
1092 
1093 /*
1094  * Do option processing on a datagram,
1095  * possibly discarding it if bad options are encountered,
1096  * or forwarding it if source-routed.
1097  * Returns 1 if packet has been forwarded/freed,
1098  * 0 if the packet should be processed further.
1099  */
1100 int
1101 ip_dooptions(m)
1102 	struct mbuf *m;
1103 {
1104 	struct ip *ip = mtod(m, struct ip *);
1105 	u_char *cp, *cp0;
1106 	struct ip_timestamp *ipt;
1107 	struct in_ifaddr *ia;
1108 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1109 	struct in_addr dst;
1110 	n_time ntime;
1111 
1112 	dst = ip->ip_dst;
1113 	cp = (u_char *)(ip + 1);
1114 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1115 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1116 		opt = cp[IPOPT_OPTVAL];
1117 		if (opt == IPOPT_EOL)
1118 			break;
1119 		if (opt == IPOPT_NOP)
1120 			optlen = 1;
1121 		else {
1122 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1123 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1124 				goto bad;
1125 			}
1126 			optlen = cp[IPOPT_OLEN];
1127 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1128 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1129 				goto bad;
1130 			}
1131 		}
1132 		switch (opt) {
1133 
1134 		default:
1135 			break;
1136 
1137 		/*
1138 		 * Source routing with record.
1139 		 * Find interface with current destination address.
1140 		 * If none on this machine then drop if strictly routed,
1141 		 * or do nothing if loosely routed.
1142 		 * Record interface address and bring up next address
1143 		 * component.  If strictly routed make sure next
1144 		 * address is on directly accessible net.
1145 		 */
1146 		case IPOPT_LSRR:
1147 		case IPOPT_SSRR:
1148 			if (ip_allowsrcrt == 0) {
1149 				type = ICMP_UNREACH;
1150 				code = ICMP_UNREACH_NET_PROHIB;
1151 				goto bad;
1152 			}
1153 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1154 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1155 				goto bad;
1156 			}
1157 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1158 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1159 				goto bad;
1160 			}
1161 			ipaddr.sin_addr = ip->ip_dst;
1162 			ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
1163 			if (ia == 0) {
1164 				if (opt == IPOPT_SSRR) {
1165 					type = ICMP_UNREACH;
1166 					code = ICMP_UNREACH_SRCFAIL;
1167 					goto bad;
1168 				}
1169 				/*
1170 				 * Loose routing, and not at next destination
1171 				 * yet; nothing to do except forward.
1172 				 */
1173 				break;
1174 			}
1175 			off--;			/* 0 origin */
1176 			if ((off + sizeof(struct in_addr)) > optlen) {
1177 				/*
1178 				 * End of source route.  Should be for us.
1179 				 */
1180 				save_rte(cp, ip->ip_src);
1181 				break;
1182 			}
1183 			/*
1184 			 * locate outgoing interface
1185 			 */
1186 			bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr,
1187 			    sizeof(ipaddr.sin_addr));
1188 			if (opt == IPOPT_SSRR)
1189 				ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
1190 			else
1191 				ia = ip_rtaddr(ipaddr.sin_addr);
1192 			if (ia == 0) {
1193 				type = ICMP_UNREACH;
1194 				code = ICMP_UNREACH_SRCFAIL;
1195 				goto bad;
1196 			}
1197 			ip->ip_dst = ipaddr.sin_addr;
1198 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
1199 			    (caddr_t)(cp + off), sizeof(struct in_addr));
1200 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1201 			/*
1202 			 * Let ip_intr's mcast routing check handle mcast pkts
1203 			 */
1204 			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1205 			break;
1206 
1207 		case IPOPT_RR:
1208 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1209 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1210 				goto bad;
1211 			}
1212 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1213 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1214 				goto bad;
1215 			}
1216 			/*
1217 			 * If no space remains, ignore.
1218 			 */
1219 			off--;			/* 0 origin */
1220 			if ((off + sizeof(struct in_addr)) > optlen)
1221 				break;
1222 			bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr,
1223 			    sizeof(ipaddr.sin_addr));
1224 			/*
1225 			 * locate outgoing interface; if we're the destination,
1226 			 * use the incoming interface (should be same).
1227 			 */
1228 			if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))))
1229 			    == NULL &&
1230 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
1231 				type = ICMP_UNREACH;
1232 				code = ICMP_UNREACH_HOST;
1233 				goto bad;
1234 			}
1235 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
1236 			    (caddr_t)(cp + off), sizeof(struct in_addr));
1237 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1238 			break;
1239 
1240 		case IPOPT_TS:
1241 			code = cp - (u_char *)ip;
1242 			ipt = (struct ip_timestamp *)cp;
1243 			if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1244 				code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1245 				goto bad;
1246 			}
1247 			if (ipt->ipt_ptr < 5) {
1248 				code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1249 				goto bad;
1250 			}
1251 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
1252 				if (++ipt->ipt_oflw == 0) {
1253 					code = (u_char *)&ipt->ipt_ptr -
1254 					    (u_char *)ip;
1255 					goto bad;
1256 				}
1257 				break;
1258 			}
1259 			cp0 = (cp + ipt->ipt_ptr - 1);
1260 			switch (ipt->ipt_flg) {
1261 
1262 			case IPOPT_TS_TSONLY:
1263 				break;
1264 
1265 			case IPOPT_TS_TSANDADDR:
1266 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1267 				    sizeof(struct in_addr) > ipt->ipt_len) {
1268 					code = (u_char *)&ipt->ipt_ptr -
1269 					    (u_char *)ip;
1270 					goto bad;
1271 				}
1272 				ipaddr.sin_addr = dst;
1273 				ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
1274 				    m->m_pkthdr.rcvif));
1275 				if (ia == 0)
1276 					continue;
1277 				bcopy(&ia->ia_addr.sin_addr,
1278 				    cp0, sizeof(struct in_addr));
1279 				ipt->ipt_ptr += sizeof(struct in_addr);
1280 				break;
1281 
1282 			case IPOPT_TS_PRESPEC:
1283 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1284 				    sizeof(struct in_addr) > ipt->ipt_len) {
1285 					code = (u_char *)&ipt->ipt_ptr -
1286 					    (u_char *)ip;
1287 					goto bad;
1288 				}
1289 				bcopy(cp0, &ipaddr.sin_addr,
1290 				    sizeof(struct in_addr));
1291 				if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))
1292 				    == NULL)
1293 					continue;
1294 				ipt->ipt_ptr += sizeof(struct in_addr);
1295 				break;
1296 
1297 			default:
1298 				/* XXX can't take &ipt->ipt_flg */
1299 				code = (u_char *)&ipt->ipt_ptr -
1300 				    (u_char *)ip + 1;
1301 				goto bad;
1302 			}
1303 			ntime = iptime();
1304 			cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1305 			bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1,
1306 			    sizeof(n_time));
1307 			ipt->ipt_ptr += sizeof(n_time);
1308 		}
1309 	}
1310 	if (forward) {
1311 		if (ip_forwsrcrt == 0) {
1312 			type = ICMP_UNREACH;
1313 			code = ICMP_UNREACH_SRCFAIL;
1314 			goto bad;
1315 		}
1316 		ip_forward(m, 1);
1317 		return (1);
1318 	}
1319 	return (0);
1320 bad:
1321 	icmp_error(m, type, code, 0, 0);
1322 	ipstat.ips_badoptions++;
1323 	return (1);
1324 }
1325 
1326 /*
1327  * Given address of next destination (final or next hop),
1328  * return internet address info of interface to be used to get there.
1329  */
1330 struct in_ifaddr *
1331 ip_rtaddr(dst)
1332 	 struct in_addr dst;
1333 {
1334 	struct sockaddr_in *sin;
1335 
1336 	sin = satosin(&ipforward_rt.ro_dst);
1337 
1338 	if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) {
1339 		if (ipforward_rt.ro_rt) {
1340 			RTFREE(ipforward_rt.ro_rt);
1341 			ipforward_rt.ro_rt = 0;
1342 		}
1343 		sin->sin_family = AF_INET;
1344 		sin->sin_len = sizeof(*sin);
1345 		sin->sin_addr = dst;
1346 
1347 		rtalloc(&ipforward_rt);
1348 	}
1349 	if (ipforward_rt.ro_rt == 0)
1350 		return ((struct in_ifaddr *)0);
1351 	return (ifatoia(ipforward_rt.ro_rt->rt_ifa));
1352 }
1353 
1354 /*
1355  * Save incoming source route for use in replies,
1356  * to be picked up later by ip_srcroute if the receiver is interested.
1357  */
1358 void
1359 save_rte(option, dst)
1360 	u_char *option;
1361 	struct in_addr dst;
1362 {
1363 	unsigned olen;
1364 
1365 	olen = option[IPOPT_OLEN];
1366 #ifdef DIAGNOSTIC
1367 	if (ipprintfs)
1368 		printf("save_rte: olen %d\n", olen);
1369 #endif /* 0 */
1370 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1371 		return;
1372 	bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen);
1373 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1374 	ip_srcrt.dst = dst;
1375 }
1376 
1377 /*
1378  * Retrieve incoming source route for use in replies,
1379  * in the same form used by setsockopt.
1380  * The first hop is placed before the options, will be removed later.
1381  */
1382 struct mbuf *
1383 ip_srcroute()
1384 {
1385 	struct in_addr *p, *q;
1386 	struct mbuf *m;
1387 
1388 	if (ip_nhops == 0)
1389 		return ((struct mbuf *)0);
1390 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1391 	if (m == 0)
1392 		return ((struct mbuf *)0);
1393 
1394 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1395 
1396 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1397 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1398 	    OPTSIZ;
1399 #ifdef DIAGNOSTIC
1400 	if (ipprintfs)
1401 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1402 #endif
1403 
1404 	/*
1405 	 * First save first hop for return route
1406 	 */
1407 	p = &ip_srcrt.route[ip_nhops - 1];
1408 	*(mtod(m, struct in_addr *)) = *p--;
1409 #ifdef DIAGNOSTIC
1410 	if (ipprintfs)
1411 		printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1412 #endif
1413 
1414 	/*
1415 	 * Copy option fields and padding (nop) to mbuf.
1416 	 */
1417 	ip_srcrt.nop = IPOPT_NOP;
1418 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1419 	bcopy((caddr_t)&ip_srcrt.nop,
1420 	    mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ);
1421 	q = (struct in_addr *)(mtod(m, caddr_t) +
1422 	    sizeof(struct in_addr) + OPTSIZ);
1423 #undef OPTSIZ
1424 	/*
1425 	 * Record return path as an IP source route,
1426 	 * reversing the path (pointers are now aligned).
1427 	 */
1428 	while (p >= ip_srcrt.route) {
1429 #ifdef DIAGNOSTIC
1430 		if (ipprintfs)
1431 			printf(" %x", ntohl(q->s_addr));
1432 #endif
1433 		*q++ = *p--;
1434 	}
1435 	/*
1436 	 * Last hop goes to final destination.
1437 	 */
1438 	*q = ip_srcrt.dst;
1439 #ifdef DIAGNOSTIC
1440 	if (ipprintfs)
1441 		printf(" %x\n", ntohl(q->s_addr));
1442 #endif
1443 	return (m);
1444 }
1445 
1446 /*
1447  * Strip out IP options, at higher
1448  * level protocol in the kernel.
1449  * Second argument is buffer to which options
1450  * will be moved, and return value is their length.
1451  * XXX should be deleted; last arg currently ignored.
1452  */
1453 void
1454 ip_stripoptions(m, mopt)
1455 	struct mbuf *m;
1456 	struct mbuf *mopt;
1457 {
1458 	int i;
1459 	struct ip *ip = mtod(m, struct ip *);
1460 	caddr_t opts;
1461 	int olen;
1462 
1463 	olen = (ip->ip_hl << 2) - sizeof (struct ip);
1464 	opts = (caddr_t)(ip + 1);
1465 	i = m->m_len - (sizeof (struct ip) + olen);
1466 	bcopy(opts  + olen, opts, (unsigned)i);
1467 	m->m_len -= olen;
1468 	if (m->m_flags & M_PKTHDR)
1469 		m->m_pkthdr.len -= olen;
1470 	ip->ip_len = htons(ntohs(ip->ip_len) - olen);
1471 	ip->ip_hl = sizeof (struct ip) >> 2;
1472 }
1473 
1474 const int inetctlerrmap[PRC_NCMDS] = {
1475 	0,		0,		0,		0,
1476 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1477 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1478 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1479 	0,		0,		0,		0,
1480 	ENOPROTOOPT
1481 };
1482 
1483 /*
1484  * Forward a packet.  If some error occurs return the sender
1485  * an icmp packet.  Note we can't always generate a meaningful
1486  * icmp message because icmp doesn't have a large enough repertoire
1487  * of codes and types.
1488  *
1489  * If not forwarding, just drop the packet.  This could be confusing
1490  * if ipforwarding was zero but some routing protocol was advancing
1491  * us as a gateway to somewhere.  However, we must let the routing
1492  * protocol deal with that.
1493  *
1494  * The srcrt parameter indicates whether the packet is being forwarded
1495  * via a source route.
1496  */
1497 void
1498 ip_forward(m, srcrt)
1499 	struct mbuf *m;
1500 	int srcrt;
1501 {
1502 	struct ip *ip = mtod(m, struct ip *);
1503 	struct sockaddr_in *sin;
1504 	struct rtentry *rt;
1505 	int error, type = 0, code = 0;
1506 	struct mbuf *mcopy;
1507 	n_long dest;
1508 	struct ifnet *destifp;
1509 #ifdef IPSEC
1510 	struct ifnet dummyifp;
1511 #endif
1512 
1513 	/*
1514 	 * Clear any in-bound checksum flags for this packet.
1515 	 */
1516 	m->m_pkthdr.csum_flags = 0;
1517 
1518 	dest = 0;
1519 #ifdef DIAGNOSTIC
1520 	if (ipprintfs)
1521 		printf("forward: src %2.2x dst %2.2x ttl %x\n",
1522 		    ntohl(ip->ip_src.s_addr),
1523 		    ntohl(ip->ip_dst.s_addr), ip->ip_ttl);
1524 #endif
1525 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1526 		ipstat.ips_cantforward++;
1527 		m_freem(m);
1528 		return;
1529 	}
1530 	if (ip->ip_ttl <= IPTTLDEC) {
1531 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1532 		return;
1533 	}
1534 	ip->ip_ttl -= IPTTLDEC;
1535 
1536 	sin = satosin(&ipforward_rt.ro_dst);
1537 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1538 	    !in_hosteq(ip->ip_dst, sin->sin_addr)) {
1539 		if (ipforward_rt.ro_rt) {
1540 			RTFREE(ipforward_rt.ro_rt);
1541 			ipforward_rt.ro_rt = 0;
1542 		}
1543 		sin->sin_family = AF_INET;
1544 		sin->sin_len = sizeof(struct sockaddr_in);
1545 		sin->sin_addr = ip->ip_dst;
1546 
1547 		rtalloc(&ipforward_rt);
1548 		if (ipforward_rt.ro_rt == 0) {
1549 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1550 			return;
1551 		}
1552 		rt = ipforward_rt.ro_rt;
1553 	}
1554 
1555 	/*
1556 	 * Save at most 68 bytes of the packet in case
1557 	 * we need to generate an ICMP message to the src.
1558 	 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1559 	 */
1560 	mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1561 	if (mcopy)
1562 		mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1563 
1564 	/*
1565 	 * If forwarding packet using same interface that it came in on,
1566 	 * perhaps should send a redirect to sender to shortcut a hop.
1567 	 * Only send redirect if source is sending directly to us,
1568 	 * and if packet was not source routed (or has any options).
1569 	 * Also, don't send redirect if forwarding using a default route
1570 	 * or a route modified by a redirect.
1571 	 */
1572 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1573 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1574 	    !in_nullhost(satosin(rt_key(rt))->sin_addr) &&
1575 	    ipsendredirects && !srcrt) {
1576 		if (rt->rt_ifa &&
1577 		    (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1578 		    ifatoia(rt->rt_ifa)->ia_subnet) {
1579 			if (rt->rt_flags & RTF_GATEWAY)
1580 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1581 			else
1582 				dest = ip->ip_dst.s_addr;
1583 			/*
1584 			 * Router requirements says to only send host
1585 			 * redirects.
1586 			 */
1587 			type = ICMP_REDIRECT;
1588 			code = ICMP_REDIRECT_HOST;
1589 #ifdef DIAGNOSTIC
1590 			if (ipprintfs)
1591 				printf("redirect (%d) to %x\n", code,
1592 				    (u_int32_t)dest);
1593 #endif
1594 		}
1595 	}
1596 
1597 #ifdef IPSEC
1598 	/* Don't lookup socket in forwarding case */
1599 	(void)ipsec_setsocket(m, NULL);
1600 #endif
1601 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1602 	    (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0);
1603 	if (error)
1604 		ipstat.ips_cantforward++;
1605 	else {
1606 		ipstat.ips_forward++;
1607 		if (type)
1608 			ipstat.ips_redirectsent++;
1609 		else {
1610 			if (mcopy) {
1611 #ifdef GATEWAY
1612 				if (mcopy->m_flags & M_CANFASTFWD)
1613 					ipflow_create(&ipforward_rt, mcopy);
1614 #endif
1615 				m_freem(mcopy);
1616 			}
1617 			return;
1618 		}
1619 	}
1620 	if (mcopy == NULL)
1621 		return;
1622 	destifp = NULL;
1623 
1624 	switch (error) {
1625 
1626 	case 0:				/* forwarded, but need redirect */
1627 		/* type, code set above */
1628 		break;
1629 
1630 	case ENETUNREACH:		/* shouldn't happen, checked above */
1631 	case EHOSTUNREACH:
1632 	case ENETDOWN:
1633 	case EHOSTDOWN:
1634 	default:
1635 		type = ICMP_UNREACH;
1636 		code = ICMP_UNREACH_HOST;
1637 		break;
1638 
1639 	case EMSGSIZE:
1640 		type = ICMP_UNREACH;
1641 		code = ICMP_UNREACH_NEEDFRAG;
1642 #ifndef IPSEC
1643 		if (ipforward_rt.ro_rt)
1644 			destifp = ipforward_rt.ro_rt->rt_ifp;
1645 #else
1646 		/*
1647 		 * If the packet is routed over IPsec tunnel, tell the
1648 		 * originator the tunnel MTU.
1649 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1650 		 * XXX quickhack!!!
1651 		 */
1652 		if (ipforward_rt.ro_rt) {
1653 			struct secpolicy *sp;
1654 			int ipsecerror;
1655 			size_t ipsechdr;
1656 			struct route *ro;
1657 
1658 			sp = ipsec4_getpolicybyaddr(mcopy,
1659 			                            IPSEC_DIR_OUTBOUND,
1660 			                            IP_FORWARDING,
1661 			                            &ipsecerror);
1662 
1663 			if (sp == NULL)
1664 				destifp = ipforward_rt.ro_rt->rt_ifp;
1665 			else {
1666 				/* count IPsec header size */
1667 				ipsechdr = ipsec4_hdrsiz(mcopy,
1668 				                         IPSEC_DIR_OUTBOUND,
1669 				                         NULL);
1670 
1671 				/*
1672 				 * find the correct route for outer IPv4
1673 				 * header, compute tunnel MTU.
1674 				 *
1675 				 * XXX BUG ALERT
1676 				 * The "dummyifp" code relies upon the fact
1677 				 * that icmp_error() touches only ifp->if_mtu.
1678 				 */
1679 				/*XXX*/
1680 				destifp = NULL;
1681 				if (sp->req != NULL
1682 				 && sp->req->sav != NULL
1683 				 && sp->req->sav->sah != NULL) {
1684 					ro = &sp->req->sav->sah->sa_route;
1685 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1686 						dummyifp.if_mtu =
1687 						    ro->ro_rt->rt_rmx.rmx_mtu ?
1688 						    ro->ro_rt->rt_rmx.rmx_mtu :
1689 						    ro->ro_rt->rt_ifp->if_mtu;
1690 						dummyifp.if_mtu -= ipsechdr;
1691 						destifp = &dummyifp;
1692 					}
1693 				}
1694 
1695 				key_freesp(sp);
1696 			}
1697 		}
1698 #endif /*IPSEC*/
1699 		ipstat.ips_cantfrag++;
1700 		break;
1701 
1702 	case ENOBUFS:
1703 #if 1
1704 		/*
1705 		 * a router should not generate ICMP_SOURCEQUENCH as
1706 		 * required in RFC1812 Requirements for IP Version 4 Routers.
1707 		 * source quench could be a big problem under DoS attacks,
1708 		 * or if the underlying interface is rate-limited.
1709 		 */
1710 		if (mcopy)
1711 			m_freem(mcopy);
1712 		return;
1713 #else
1714 		type = ICMP_SOURCEQUENCH;
1715 		code = 0;
1716 		break;
1717 #endif
1718 	}
1719 	icmp_error(mcopy, type, code, dest, destifp);
1720 }
1721 
1722 void
1723 ip_savecontrol(inp, mp, ip, m)
1724 	struct inpcb *inp;
1725 	struct mbuf **mp;
1726 	struct ip *ip;
1727 	struct mbuf *m;
1728 {
1729 
1730 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1731 		struct timeval tv;
1732 
1733 		microtime(&tv);
1734 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1735 		    SCM_TIMESTAMP, SOL_SOCKET);
1736 		if (*mp)
1737 			mp = &(*mp)->m_next;
1738 	}
1739 	if (inp->inp_flags & INP_RECVDSTADDR) {
1740 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1741 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1742 		if (*mp)
1743 			mp = &(*mp)->m_next;
1744 	}
1745 #ifdef notyet
1746 	/*
1747 	 * XXX
1748 	 * Moving these out of udp_input() made them even more broken
1749 	 * than they already were.
1750 	 *	- fenner@parc.xerox.com
1751 	 */
1752 	/* options were tossed already */
1753 	if (inp->inp_flags & INP_RECVOPTS) {
1754 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1755 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1756 		if (*mp)
1757 			mp = &(*mp)->m_next;
1758 	}
1759 	/* ip_srcroute doesn't do what we want here, need to fix */
1760 	if (inp->inp_flags & INP_RECVRETOPTS) {
1761 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1762 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1763 		if (*mp)
1764 			mp = &(*mp)->m_next;
1765 	}
1766 #endif
1767 	if (inp->inp_flags & INP_RECVIF) {
1768 		struct sockaddr_dl sdl;
1769 
1770 		sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
1771 		sdl.sdl_family = AF_LINK;
1772 		sdl.sdl_index = m->m_pkthdr.rcvif ?
1773 		    m->m_pkthdr.rcvif->if_index : 0;
1774 		sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
1775 		*mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
1776 		    IP_RECVIF, IPPROTO_IP);
1777 		if (*mp)
1778 			mp = &(*mp)->m_next;
1779 	}
1780 }
1781 
1782 int
1783 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1784 	int *name;
1785 	u_int namelen;
1786 	void *oldp;
1787 	size_t *oldlenp;
1788 	void *newp;
1789 	size_t newlen;
1790 {
1791 	extern int subnetsarelocal, hostzeroisbroadcast;
1792 
1793 	int error, old;
1794 
1795 	/* All sysctl names at this level are terminal. */
1796 	if (namelen != 1)
1797 		return (ENOTDIR);
1798 
1799 	switch (name[0]) {
1800 	case IPCTL_FORWARDING:
1801 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding));
1802 	case IPCTL_SENDREDIRECTS:
1803 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1804 			&ipsendredirects));
1805 	case IPCTL_DEFTTL:
1806 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl));
1807 #ifdef notyet
1808 	case IPCTL_DEFMTU:
1809 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu));
1810 #endif
1811 	case IPCTL_FORWSRCRT:
1812 		/* Don't allow this to change in a secure environment.  */
1813 		if (securelevel > 0)
1814 			return (sysctl_rdint(oldp, oldlenp, newp,
1815 			    ip_forwsrcrt));
1816 		else
1817 			return (sysctl_int(oldp, oldlenp, newp, newlen,
1818 			    &ip_forwsrcrt));
1819 	case IPCTL_DIRECTEDBCAST:
1820 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1821 		    &ip_directedbcast));
1822 	case IPCTL_ALLOWSRCRT:
1823 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1824 		    &ip_allowsrcrt));
1825 	case IPCTL_SUBNETSARELOCAL:
1826 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1827 		    &subnetsarelocal));
1828 	case IPCTL_MTUDISC:
1829 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1830 		    &ip_mtudisc);
1831 		if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) {
1832 			ip_mtudisc_timeout_q =
1833 			    rt_timer_queue_create(ip_mtudisc_timeout);
1834 		} else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) {
1835 			rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE);
1836 			ip_mtudisc_timeout_q = NULL;
1837 		}
1838 		return error;
1839 	case IPCTL_ANONPORTMIN:
1840 		old = anonportmin;
1841 		error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin);
1842 		if (anonportmin >= anonportmax || anonportmin < 0
1843 		    || anonportmin > 65535
1844 #ifndef IPNOPRIVPORTS
1845 		    || anonportmin < IPPORT_RESERVED
1846 #endif
1847 		    ) {
1848 			anonportmin = old;
1849 			return (EINVAL);
1850 		}
1851 		return (error);
1852 	case IPCTL_ANONPORTMAX:
1853 		old = anonportmax;
1854 		error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax);
1855 		if (anonportmin >= anonportmax || anonportmax < 0
1856 		    || anonportmax > 65535
1857 #ifndef IPNOPRIVPORTS
1858 		    || anonportmax < IPPORT_RESERVED
1859 #endif
1860 		    ) {
1861 			anonportmax = old;
1862 			return (EINVAL);
1863 		}
1864 		return (error);
1865 	case IPCTL_MTUDISCTIMEOUT:
1866 		old = ip_mtudisc_timeout;
1867 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1868 		   &ip_mtudisc_timeout);
1869 		if (ip_mtudisc_timeout < 0) {
1870 			ip_mtudisc_timeout = old;
1871 			return (EINVAL);
1872 		}
1873 		if (ip_mtudisc_timeout_q != NULL)
1874 			rt_timer_queue_change(ip_mtudisc_timeout_q,
1875 					      ip_mtudisc_timeout);
1876 		return (error);
1877 #ifdef GATEWAY
1878 	case IPCTL_MAXFLOWS:
1879 	    {
1880 		int s;
1881 
1882 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1883 		   &ip_maxflows);
1884 		s = splsoftnet();
1885 		ipflow_reap(0);
1886 		splx(s);
1887 		return (error);
1888 	    }
1889 #endif
1890 	case IPCTL_HOSTZEROBROADCAST:
1891 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1892 		    &hostzeroisbroadcast));
1893 #if NGIF > 0
1894 	case IPCTL_GIF_TTL:
1895 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1896 				  &ip_gif_ttl));
1897 #endif
1898 
1899 #if NGRE > 0
1900 	case IPCTL_GRE_TTL:
1901 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1902 				  &ip_gre_ttl));
1903 #endif
1904 
1905 #ifndef IPNOPRIVPORTS
1906 	case IPCTL_LOWPORTMIN:
1907 		old = lowportmin;
1908 		error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin);
1909 		if (lowportmin >= lowportmax
1910 		    || lowportmin > IPPORT_RESERVEDMAX
1911 		    || lowportmin < IPPORT_RESERVEDMIN
1912 		    ) {
1913 			lowportmin = old;
1914 			return (EINVAL);
1915 		}
1916 		return (error);
1917 	case IPCTL_LOWPORTMAX:
1918 		old = lowportmax;
1919 		error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax);
1920 		if (lowportmin >= lowportmax
1921 		    || lowportmax > IPPORT_RESERVEDMAX
1922 		    || lowportmax < IPPORT_RESERVEDMIN
1923 		    ) {
1924 			lowportmax = old;
1925 			return (EINVAL);
1926 		}
1927 		return (error);
1928 #endif
1929 
1930 	case IPCTL_MAXFRAGPACKETS:
1931 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1932 		    &ip_maxfragpackets));
1933 
1934 	default:
1935 		return (EOPNOTSUPP);
1936 	}
1937 	/* NOTREACHED */
1938 }
1939