ip_input.c revision 1.59
1/*	$NetBSD: ip_input.c,v 1.59 1998/03/19 15:46:43 mrg Exp $	*/
2
3/*-
4 * Copyright (c) 1998 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Public Access Networks Corporation ("Panix").  It was developed under
9 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 *    notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 *    notice, this list of conditions and the following disclaimer in the
18 *    documentation and/or other materials provided with the distribution.
19 * 3. All advertising materials mentioning features or use of this software
20 *    must display the following acknowledgement:
21 *	This product includes software developed by the NetBSD
22 *	Foundation, Inc. and its contributors.
23 * 4. Neither the name of The NetBSD Foundation nor the names of its
24 *    contributors may be used to endorse or promote products derived
25 *    from this software without specific prior written permission.
26 *
27 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37 * POSSIBILITY OF SUCH DAMAGE.
38 */
39
40/*
41 * Copyright (c) 1982, 1986, 1988, 1993
42 *	The Regents of the University of California.  All rights reserved.
43 *
44 * Redistribution and use in source and binary forms, with or without
45 * modification, are permitted provided that the following conditions
46 * are met:
47 * 1. Redistributions of source code must retain the above copyright
48 *    notice, this list of conditions and the following disclaimer.
49 * 2. Redistributions in binary form must reproduce the above copyright
50 *    notice, this list of conditions and the following disclaimer in the
51 *    documentation and/or other materials provided with the distribution.
52 * 3. All advertising materials mentioning features or use of this software
53 *    must display the following acknowledgement:
54 *	This product includes software developed by the University of
55 *	California, Berkeley and its contributors.
56 * 4. Neither the name of the University nor the names of its contributors
57 *    may be used to endorse or promote products derived from this software
58 *    without specific prior written permission.
59 *
60 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
61 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
62 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
63 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
64 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
65 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
66 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
67 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
68 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
69 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
70 * SUCH DAMAGE.
71 *
72 *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
73 */
74
75#include "opt_mrouting.h"
76
77#include <sys/param.h>
78#include <sys/systm.h>
79#include <sys/malloc.h>
80#include <sys/mbuf.h>
81#include <sys/domain.h>
82#include <sys/protosw.h>
83#include <sys/socket.h>
84#include <sys/socketvar.h>
85#include <sys/errno.h>
86#include <sys/time.h>
87#include <sys/kernel.h>
88#include <sys/proc.h>
89
90#include <vm/vm.h>
91#include <sys/sysctl.h>
92
93#include <net/if.h>
94#include <net/if_dl.h>
95#include <net/route.h>
96#include <net/pfil.h>
97
98#include <netinet/in.h>
99#include <netinet/in_systm.h>
100#include <netinet/ip.h>
101#include <netinet/in_pcb.h>
102#include <netinet/in_var.h>
103#include <netinet/ip_var.h>
104#include <netinet/ip_icmp.h>
105
106#ifndef	IPFORWARDING
107#ifdef GATEWAY
108#define	IPFORWARDING	1	/* forward IP packets not for us */
109#else /* GATEWAY */
110#define	IPFORWARDING	0	/* don't forward IP packets not for us */
111#endif /* GATEWAY */
112#endif /* IPFORWARDING */
113#ifndef	IPSENDREDIRECTS
114#define	IPSENDREDIRECTS	1
115#endif
116#ifndef IPFORWSRCRT
117#define	IPFORWSRCRT	1	/* forward source-routed packets */
118#endif
119#ifndef IPALLOWSRCRT
120#define	IPALLOWSRCRT	1	/* allow source-routed packets */
121#endif
122#ifndef IPMTUDISC
123#define IPMTUDISC	0
124#endif
125
126/*
127 * Note: DIRECTED_BROADCAST is handled this way so that previous
128 * configuration using this option will Just Work.
129 */
130#ifndef IPDIRECTEDBCAST
131#ifdef DIRECTED_BROADCAST
132#define IPDIRECTEDBCAST	1
133#else
134#define	IPDIRECTEDBCAST	0
135#endif /* DIRECTED_BROADCAST */
136#endif /* IPDIRECTEDBCAST */
137int	ipforwarding = IPFORWARDING;
138int	ipsendredirects = IPSENDREDIRECTS;
139int	ip_defttl = IPDEFTTL;
140int	ip_forwsrcrt = IPFORWSRCRT;
141int	ip_directedbcast = IPDIRECTEDBCAST;
142int	ip_allowsrcrt = IPALLOWSRCRT;
143int	ip_mtudisc = IPMTUDISC;
144#ifdef DIAGNOSTIC
145int	ipprintfs = 0;
146#endif
147
148extern	struct domain inetdomain;
149extern	struct protosw inetsw[];
150u_char	ip_protox[IPPROTO_MAX];
151int	ipqmaxlen = IFQ_MAXLEN;
152struct	in_ifaddrhead in_ifaddr;
153struct	in_ifaddrhashhead *in_ifaddrhashtbl;
154struct	ifqueue ipintrq;
155
156/*
157 * We need to save the IP options in case a protocol wants to respond
158 * to an incoming packet over the same route if the packet got here
159 * using IP source routing.  This allows connection establishment and
160 * maintenance when the remote end is on a network that is not known
161 * to us.
162 */
163int	ip_nhops = 0;
164static	struct ip_srcrt {
165	struct	in_addr dst;			/* final destination */
166	char	nop;				/* one NOP to align */
167	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
168	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
169} ip_srcrt;
170
171static void save_rte __P((u_char *, struct in_addr));
172
173/*
174 * IP initialization: fill in IP protocol switch table.
175 * All protocols not implemented in kernel go to raw IP protocol handler.
176 */
177void
178ip_init()
179{
180	register struct protosw *pr;
181	register int i;
182
183	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
184	if (pr == 0)
185		panic("ip_init");
186	for (i = 0; i < IPPROTO_MAX; i++)
187		ip_protox[i] = pr - inetsw;
188	for (pr = inetdomain.dom_protosw;
189	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
190		if (pr->pr_domain->dom_family == PF_INET &&
191		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
192			ip_protox[pr->pr_protocol] = pr - inetsw;
193	LIST_INIT(&ipq);
194	ip_id = time.tv_sec & 0xffff;
195	ipintrq.ifq_maxlen = ipqmaxlen;
196	TAILQ_INIT(&in_ifaddr);
197	in_ifaddrhashtbl =
198	    hashinit(IN_IFADDR_HASH_SIZE, M_IFADDR, M_WAITOK, &in_ifaddrhash);
199}
200
201struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
202struct	route ipforward_rt;
203
204/*
205 * Ip input routine.  Checksum and byte swap header.  If fragmented
206 * try to reassemble.  Process options.  Pass to next level.
207 */
208void
209ipintr()
210{
211	register struct ip *ip = NULL;
212	register struct mbuf *m;
213	register struct ipq *fp;
214	register struct in_ifaddr *ia;
215	register struct ifaddr *ifa;
216	struct ipqent *ipqe;
217	int hlen = 0, mff, len, s;
218#ifdef PFIL_HOOKS
219	struct packet_filter_hook *pfh;
220	struct mbuf *m0;
221	int rv;
222#endif /* PFIL_HOOKS */
223
224next:
225	/*
226	 * Get next datagram off input queue and get IP header
227	 * in first mbuf.
228	 */
229	s = splimp();
230	IF_DEQUEUE(&ipintrq, m);
231	splx(s);
232	if (m == 0)
233		return;
234#ifdef	DIAGNOSTIC
235	if ((m->m_flags & M_PKTHDR) == 0)
236		panic("ipintr no HDR");
237#endif
238	/*
239	 * If no IP addresses have been set yet but the interfaces
240	 * are receiving, can't do anything with incoming packets yet.
241	 */
242	if (in_ifaddr.tqh_first == 0)
243		goto bad;
244	ipstat.ips_total++;
245	if (m->m_len < sizeof (struct ip) &&
246	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
247		ipstat.ips_toosmall++;
248		goto next;
249	}
250	ip = mtod(m, struct ip *);
251	if (ip->ip_v != IPVERSION) {
252		ipstat.ips_badvers++;
253		goto bad;
254	}
255	hlen = ip->ip_hl << 2;
256	if (hlen < sizeof(struct ip)) {	/* minimum header length */
257		ipstat.ips_badhlen++;
258		goto bad;
259	}
260	if (hlen > m->m_len) {
261		if ((m = m_pullup(m, hlen)) == 0) {
262			ipstat.ips_badhlen++;
263			goto next;
264		}
265		ip = mtod(m, struct ip *);
266	}
267	if ((ip->ip_sum = in_cksum(m, hlen)) != 0) {
268		ipstat.ips_badsum++;
269		goto bad;
270	}
271
272	/*
273	 * Convert fields to host representation.
274	 */
275	NTOHS(ip->ip_len);
276	NTOHS(ip->ip_id);
277	NTOHS(ip->ip_off);
278	len = ip->ip_len;
279
280	/*
281	 * Check that the amount of data in the buffers
282	 * is as at least much as the IP header would have us expect.
283	 * Trim mbufs if longer than we expect.
284	 * Drop packet if shorter than we expect.
285	 */
286	if (m->m_pkthdr.len < len) {
287		ipstat.ips_tooshort++;
288		goto bad;
289	}
290	if (m->m_pkthdr.len > len) {
291		if (m->m_len == m->m_pkthdr.len) {
292			m->m_len = len;
293			m->m_pkthdr.len = len;
294		} else
295			m_adj(m, len - m->m_pkthdr.len);
296	}
297
298#ifdef PFIL_HOOKS
299	/*
300	 * Run through list of hooks for input packets.
301	 */
302	m0 = m;
303	for (pfh = pfil_hook_get(PFIL_IN); pfh; pfh = pfh->pfil_link.tqe_next)
304		if (pfh->pfil_func) {
305			rv = pfh->pfil_func(ip, hlen, m->m_pkthdr.rcvif, 0, &m0);
306			if (rv)
307				goto next;
308			ip = mtod(m = m0, struct ip *);
309		}
310#endif /* PFIL_HOOKS */
311
312	/*
313	 * Process options and, if not destined for us,
314	 * ship it on.  ip_dooptions returns 1 when an
315	 * error was detected (causing an icmp message
316	 * to be sent and the original packet to be freed).
317	 */
318	ip_nhops = 0;		/* for source routed packets */
319	if (hlen > sizeof (struct ip) && ip_dooptions(m))
320		goto next;
321
322	/*
323	 * Check our list of addresses, to see if the packet is for us.
324	 */
325	INADDR_TO_IA(ip->ip_dst, ia);
326	if (ia != NULL) goto ours;
327	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
328		for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first;
329		    ifa != NULL; ifa = ifa->ifa_list.tqe_next) {
330			if (ifa->ifa_addr->sa_family != AF_INET) continue;
331			ia = ifatoia(ifa);
332			if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
333			    in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
334			    /*
335			     * Look for all-0's host part (old broadcast addr),
336			     * either for subnet or net.
337			     */
338			    ip->ip_dst.s_addr == ia->ia_subnet ||
339			    ip->ip_dst.s_addr == ia->ia_net)
340				goto ours;
341			/*
342			 * An interface with IP address zero accepts
343			 * all packets that arrive on that interface.
344			 */
345			if (in_nullhost(ia->ia_addr.sin_addr))
346				goto ours;
347		}
348	}
349	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
350		struct in_multi *inm;
351#ifdef MROUTING
352		extern struct socket *ip_mrouter;
353
354		if (m->m_flags & M_EXT) {
355			if ((m = m_pullup(m, hlen)) == 0) {
356				ipstat.ips_toosmall++;
357				goto next;
358			}
359			ip = mtod(m, struct ip *);
360		}
361
362		if (ip_mrouter) {
363			/*
364			 * If we are acting as a multicast router, all
365			 * incoming multicast packets are passed to the
366			 * kernel-level multicast forwarding function.
367			 * The packet is returned (relatively) intact; if
368			 * ip_mforward() returns a non-zero value, the packet
369			 * must be discarded, else it may be accepted below.
370			 *
371			 * (The IP ident field is put in the same byte order
372			 * as expected when ip_mforward() is called from
373			 * ip_output().)
374			 */
375			ip->ip_id = htons(ip->ip_id);
376			if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
377				ipstat.ips_cantforward++;
378				m_freem(m);
379				goto next;
380			}
381			ip->ip_id = ntohs(ip->ip_id);
382
383			/*
384			 * The process-level routing demon needs to receive
385			 * all multicast IGMP packets, whether or not this
386			 * host belongs to their destination groups.
387			 */
388			if (ip->ip_p == IPPROTO_IGMP)
389				goto ours;
390			ipstat.ips_forward++;
391		}
392#endif
393		/*
394		 * See if we belong to the destination multicast group on the
395		 * arrival interface.
396		 */
397		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
398		if (inm == NULL) {
399			ipstat.ips_cantforward++;
400			m_freem(m);
401			goto next;
402		}
403		goto ours;
404	}
405	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
406	    in_nullhost(ip->ip_dst))
407		goto ours;
408
409	/*
410	 * Not for us; forward if possible and desirable.
411	 */
412	if (ipforwarding == 0) {
413		ipstat.ips_cantforward++;
414		m_freem(m);
415	} else
416		ip_forward(m, 0);
417	goto next;
418
419ours:
420	/*
421	 * If offset or IP_MF are set, must reassemble.
422	 * Otherwise, nothing need be done.
423	 * (We could look in the reassembly queue to see
424	 * if the packet was previously fragmented,
425	 * but it's not worth the time; just let them time out.)
426	 */
427	if (ip->ip_off & ~(IP_DF|IP_RF)) {
428		/*
429		 * Look for queue of fragments
430		 * of this datagram.
431		 */
432		for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next)
433			if (ip->ip_id == fp->ipq_id &&
434			    in_hosteq(ip->ip_src, fp->ipq_src) &&
435			    in_hosteq(ip->ip_dst, fp->ipq_dst) &&
436			    ip->ip_p == fp->ipq_p)
437				goto found;
438		fp = 0;
439found:
440
441		/*
442		 * Adjust ip_len to not reflect header,
443		 * set ipqe_mff if more fragments are expected,
444		 * convert offset of this to bytes.
445		 */
446		ip->ip_len -= hlen;
447		mff = (ip->ip_off & IP_MF) != 0;
448		if (mff) {
449		        /*
450		         * Make sure that fragments have a data length
451			 * that's a non-zero multiple of 8 bytes.
452		         */
453			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
454				ipstat.ips_badfrags++;
455				goto bad;
456			}
457		}
458		ip->ip_off <<= 3;
459
460		/*
461		 * If datagram marked as having more fragments
462		 * or if this is not the first fragment,
463		 * attempt reassembly; if it succeeds, proceed.
464		 */
465		if (mff || ip->ip_off) {
466			ipstat.ips_fragments++;
467			MALLOC(ipqe, struct ipqent *, sizeof (struct ipqent),
468			    M_IPQ, M_NOWAIT);
469			if (ipqe == NULL) {
470				ipstat.ips_rcvmemdrop++;
471				goto bad;
472			}
473			ipqe->ipqe_mff = mff;
474			ipqe->ipqe_m = m;
475			ipqe->ipqe_ip = ip;
476			m = ip_reass(ipqe, fp);
477			if (m == 0)
478				goto next;
479			ipstat.ips_reassembled++;
480			ip = mtod(m, struct ip *);
481		} else
482			if (fp)
483				ip_freef(fp);
484	} else
485		ip->ip_len -= hlen;
486
487	/*
488	 * Switch out to protocol's input routine.
489	 */
490	ipstat.ips_delivered++;
491	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
492	goto next;
493bad:
494	m_freem(m);
495	goto next;
496}
497
498/*
499 * Take incoming datagram fragment and try to
500 * reassemble it into whole datagram.  If a chain for
501 * reassembly of this datagram already exists, then it
502 * is given as fp; otherwise have to make a chain.
503 */
504struct mbuf *
505ip_reass(ipqe, fp)
506	register struct ipqent *ipqe;
507	register struct ipq *fp;
508{
509	register struct mbuf *m = ipqe->ipqe_m;
510	register struct ipqent *nq, *p, *q;
511	struct ip *ip;
512	struct mbuf *t;
513	int hlen = ipqe->ipqe_ip->ip_hl << 2;
514	int i, next;
515
516	/*
517	 * Presence of header sizes in mbufs
518	 * would confuse code below.
519	 */
520	m->m_data += hlen;
521	m->m_len -= hlen;
522
523	/*
524	 * If first fragment to arrive, create a reassembly queue.
525	 */
526	if (fp == 0) {
527		MALLOC(fp, struct ipq *, sizeof (struct ipq),
528		    M_FTABLE, M_NOWAIT);
529		if (fp == NULL)
530			goto dropfrag;
531		LIST_INSERT_HEAD(&ipq, fp, ipq_q);
532		fp->ipq_ttl = IPFRAGTTL;
533		fp->ipq_p = ipqe->ipqe_ip->ip_p;
534		fp->ipq_id = ipqe->ipqe_ip->ip_id;
535		LIST_INIT(&fp->ipq_fragq);
536		fp->ipq_src = ipqe->ipqe_ip->ip_src;
537		fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
538		p = NULL;
539		goto insert;
540	}
541
542	/*
543	 * Find a segment which begins after this one does.
544	 */
545	for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL;
546	    p = q, q = q->ipqe_q.le_next)
547		if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off)
548			break;
549
550	/*
551	 * If there is a preceding segment, it may provide some of
552	 * our data already.  If so, drop the data from the incoming
553	 * segment.  If it provides all of our data, drop us.
554	 */
555	if (p != NULL) {
556		i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len -
557		    ipqe->ipqe_ip->ip_off;
558		if (i > 0) {
559			if (i >= ipqe->ipqe_ip->ip_len)
560				goto dropfrag;
561			m_adj(ipqe->ipqe_m, i);
562			ipqe->ipqe_ip->ip_off += i;
563			ipqe->ipqe_ip->ip_len -= i;
564		}
565	}
566
567	/*
568	 * While we overlap succeeding segments trim them or,
569	 * if they are completely covered, dequeue them.
570	 */
571	for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len >
572	    q->ipqe_ip->ip_off; q = nq) {
573		i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) -
574		    q->ipqe_ip->ip_off;
575		if (i < q->ipqe_ip->ip_len) {
576			q->ipqe_ip->ip_len -= i;
577			q->ipqe_ip->ip_off += i;
578			m_adj(q->ipqe_m, i);
579			break;
580		}
581		nq = q->ipqe_q.le_next;
582		m_freem(q->ipqe_m);
583		LIST_REMOVE(q, ipqe_q);
584		FREE(q, M_IPQ);
585	}
586
587insert:
588	/*
589	 * Stick new segment in its place;
590	 * check for complete reassembly.
591	 */
592	if (p == NULL) {
593		LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
594	} else {
595		LIST_INSERT_AFTER(p, ipqe, ipqe_q);
596	}
597	next = 0;
598	for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL;
599	    p = q, q = q->ipqe_q.le_next) {
600		if (q->ipqe_ip->ip_off != next)
601			return (0);
602		next += q->ipqe_ip->ip_len;
603	}
604	if (p->ipqe_mff)
605		return (0);
606
607	/*
608	 * Reassembly is complete.  Check for a bogus message size and
609	 * concatenate fragments.
610	 */
611	q = fp->ipq_fragq.lh_first;
612	ip = q->ipqe_ip;
613	if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
614		ipstat.ips_toolong++;
615		ip_freef(fp);
616		return (0);
617	}
618	m = q->ipqe_m;
619	t = m->m_next;
620	m->m_next = 0;
621	m_cat(m, t);
622	nq = q->ipqe_q.le_next;
623	FREE(q, M_IPQ);
624	for (q = nq; q != NULL; q = nq) {
625		t = q->ipqe_m;
626		nq = q->ipqe_q.le_next;
627		FREE(q, M_IPQ);
628		m_cat(m, t);
629	}
630
631	/*
632	 * Create header for new ip packet by
633	 * modifying header of first packet;
634	 * dequeue and discard fragment reassembly header.
635	 * Make header visible.
636	 */
637	ip->ip_len = next;
638	ip->ip_src = fp->ipq_src;
639	ip->ip_dst = fp->ipq_dst;
640	LIST_REMOVE(fp, ipq_q);
641	FREE(fp, M_FTABLE);
642	m->m_len += (ip->ip_hl << 2);
643	m->m_data -= (ip->ip_hl << 2);
644	/* some debugging cruft by sklower, below, will go away soon */
645	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
646		register int plen = 0;
647		for (t = m; t; t = t->m_next)
648			plen += t->m_len;
649		m->m_pkthdr.len = plen;
650	}
651	return (m);
652
653dropfrag:
654	ipstat.ips_fragdropped++;
655	m_freem(m);
656	FREE(ipqe, M_IPQ);
657	return (0);
658}
659
660/*
661 * Free a fragment reassembly header and all
662 * associated datagrams.
663 */
664void
665ip_freef(fp)
666	struct ipq *fp;
667{
668	register struct ipqent *q, *p;
669
670	for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) {
671		p = q->ipqe_q.le_next;
672		m_freem(q->ipqe_m);
673		LIST_REMOVE(q, ipqe_q);
674		FREE(q, M_IPQ);
675	}
676	LIST_REMOVE(fp, ipq_q);
677	FREE(fp, M_FTABLE);
678}
679
680/*
681 * IP timer processing;
682 * if a timer expires on a reassembly
683 * queue, discard it.
684 */
685void
686ip_slowtimo()
687{
688	register struct ipq *fp, *nfp;
689	int s = splsoftnet();
690
691	for (fp = ipq.lh_first; fp != NULL; fp = nfp) {
692		nfp = fp->ipq_q.le_next;
693		if (--fp->ipq_ttl == 0) {
694			ipstat.ips_fragtimeout++;
695			ip_freef(fp);
696		}
697	}
698	splx(s);
699}
700
701/*
702 * Drain off all datagram fragments.
703 */
704void
705ip_drain()
706{
707
708	while (ipq.lh_first != NULL) {
709		ipstat.ips_fragdropped++;
710		ip_freef(ipq.lh_first);
711	}
712}
713
714/*
715 * Do option processing on a datagram,
716 * possibly discarding it if bad options are encountered,
717 * or forwarding it if source-routed.
718 * Returns 1 if packet has been forwarded/freed,
719 * 0 if the packet should be processed further.
720 */
721int
722ip_dooptions(m)
723	struct mbuf *m;
724{
725	register struct ip *ip = mtod(m, struct ip *);
726	register u_char *cp;
727	register struct ip_timestamp *ipt;
728	register struct in_ifaddr *ia;
729	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
730	struct in_addr *sin, dst;
731	n_time ntime;
732
733	dst = ip->ip_dst;
734	cp = (u_char *)(ip + 1);
735	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
736	for (; cnt > 0; cnt -= optlen, cp += optlen) {
737		opt = cp[IPOPT_OPTVAL];
738		if (opt == IPOPT_EOL)
739			break;
740		if (opt == IPOPT_NOP)
741			optlen = 1;
742		else {
743			optlen = cp[IPOPT_OLEN];
744			if (optlen <= 0 || optlen > cnt) {
745				code = &cp[IPOPT_OLEN] - (u_char *)ip;
746				goto bad;
747			}
748		}
749		switch (opt) {
750
751		default:
752			break;
753
754		/*
755		 * Source routing with record.
756		 * Find interface with current destination address.
757		 * If none on this machine then drop if strictly routed,
758		 * or do nothing if loosely routed.
759		 * Record interface address and bring up next address
760		 * component.  If strictly routed make sure next
761		 * address is on directly accessible net.
762		 */
763		case IPOPT_LSRR:
764		case IPOPT_SSRR:
765			if (ip_allowsrcrt == 0) {
766				type = ICMP_UNREACH;
767				code = ICMP_UNREACH_NET_PROHIB;
768				goto bad;
769			}
770			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
771				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
772				goto bad;
773			}
774			ipaddr.sin_addr = ip->ip_dst;
775			ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
776			if (ia == 0) {
777				if (opt == IPOPT_SSRR) {
778					type = ICMP_UNREACH;
779					code = ICMP_UNREACH_SRCFAIL;
780					goto bad;
781				}
782				/*
783				 * Loose routing, and not at next destination
784				 * yet; nothing to do except forward.
785				 */
786				break;
787			}
788			off--;			/* 0 origin */
789			if (off > optlen - sizeof(struct in_addr)) {
790				/*
791				 * End of source route.  Should be for us.
792				 */
793				save_rte(cp, ip->ip_src);
794				break;
795			}
796			/*
797			 * locate outgoing interface
798			 */
799			bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr,
800			    sizeof(ipaddr.sin_addr));
801			if (opt == IPOPT_SSRR) {
802#define	INA	struct in_ifaddr *
803#define	SA	struct sockaddr *
804			    ia = (INA)ifa_ifwithladdr((SA)&ipaddr);
805			} else
806				ia = ip_rtaddr(ipaddr.sin_addr);
807			if (ia == 0) {
808				type = ICMP_UNREACH;
809				code = ICMP_UNREACH_SRCFAIL;
810				goto bad;
811			}
812			ip->ip_dst = ipaddr.sin_addr;
813			bcopy((caddr_t)&ia->ia_addr.sin_addr,
814			    (caddr_t)(cp + off), sizeof(struct in_addr));
815			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
816			/*
817			 * Let ip_intr's mcast routing check handle mcast pkts
818			 */
819			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
820			break;
821
822		case IPOPT_RR:
823			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
824				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
825				goto bad;
826			}
827			/*
828			 * If no space remains, ignore.
829			 */
830			off--;			/* 0 origin */
831			if (off > optlen - sizeof(struct in_addr))
832				break;
833			bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr,
834			    sizeof(ipaddr.sin_addr));
835			/*
836			 * locate outgoing interface; if we're the destination,
837			 * use the incoming interface (should be same).
838			 */
839			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
840			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
841				type = ICMP_UNREACH;
842				code = ICMP_UNREACH_HOST;
843				goto bad;
844			}
845			bcopy((caddr_t)&ia->ia_addr.sin_addr,
846			    (caddr_t)(cp + off), sizeof(struct in_addr));
847			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
848			break;
849
850		case IPOPT_TS:
851			code = cp - (u_char *)ip;
852			ipt = (struct ip_timestamp *)cp;
853			if (ipt->ipt_len < 5)
854				goto bad;
855			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
856				if (++ipt->ipt_oflw == 0)
857					goto bad;
858				break;
859			}
860			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
861			switch (ipt->ipt_flg) {
862
863			case IPOPT_TS_TSONLY:
864				break;
865
866			case IPOPT_TS_TSANDADDR:
867				if (ipt->ipt_ptr + sizeof(n_time) +
868				    sizeof(struct in_addr) > ipt->ipt_len)
869					goto bad;
870				ipaddr.sin_addr = dst;
871				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
872							    m->m_pkthdr.rcvif);
873				if (ia == 0)
874					continue;
875				bcopy((caddr_t)&ia->ia_addr.sin_addr,
876				    (caddr_t)sin, sizeof(struct in_addr));
877				ipt->ipt_ptr += sizeof(struct in_addr);
878				break;
879
880			case IPOPT_TS_PRESPEC:
881				if (ipt->ipt_ptr + sizeof(n_time) +
882				    sizeof(struct in_addr) > ipt->ipt_len)
883					goto bad;
884				bcopy((caddr_t)sin, (caddr_t)&ipaddr.sin_addr,
885				    sizeof(struct in_addr));
886				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
887					continue;
888				ipt->ipt_ptr += sizeof(struct in_addr);
889				break;
890
891			default:
892				goto bad;
893			}
894			ntime = iptime();
895			bcopy((caddr_t)&ntime, (caddr_t)cp + ipt->ipt_ptr - 1,
896			    sizeof(n_time));
897			ipt->ipt_ptr += sizeof(n_time);
898		}
899	}
900	if (forward) {
901		if (ip_forwsrcrt == 0) {
902			type = ICMP_UNREACH;
903			code = ICMP_UNREACH_SRCFAIL;
904			goto bad;
905		}
906		ip_forward(m, 1);
907		return (1);
908	}
909	return (0);
910bad:
911	ip->ip_len -= ip->ip_hl << 2;   /* XXX icmp_error adds in hdr length */
912	icmp_error(m, type, code, 0, 0);
913	ipstat.ips_badoptions++;
914	return (1);
915}
916
917/*
918 * Given address of next destination (final or next hop),
919 * return internet address info of interface to be used to get there.
920 */
921struct in_ifaddr *
922ip_rtaddr(dst)
923	 struct in_addr dst;
924{
925	register struct sockaddr_in *sin;
926
927	sin = satosin(&ipforward_rt.ro_dst);
928
929	if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) {
930		if (ipforward_rt.ro_rt) {
931			RTFREE(ipforward_rt.ro_rt);
932			ipforward_rt.ro_rt = 0;
933		}
934		sin->sin_family = AF_INET;
935		sin->sin_len = sizeof(*sin);
936		sin->sin_addr = dst;
937
938		rtalloc(&ipforward_rt);
939	}
940	if (ipforward_rt.ro_rt == 0)
941		return ((struct in_ifaddr *)0);
942	return (ifatoia(ipforward_rt.ro_rt->rt_ifa));
943}
944
945/*
946 * Save incoming source route for use in replies,
947 * to be picked up later by ip_srcroute if the receiver is interested.
948 */
949void
950save_rte(option, dst)
951	u_char *option;
952	struct in_addr dst;
953{
954	unsigned olen;
955
956	olen = option[IPOPT_OLEN];
957#ifdef DIAGNOSTIC
958	if (ipprintfs)
959		printf("save_rte: olen %d\n", olen);
960#endif
961	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
962		return;
963	bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen);
964	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
965	ip_srcrt.dst = dst;
966}
967
968/*
969 * Retrieve incoming source route for use in replies,
970 * in the same form used by setsockopt.
971 * The first hop is placed before the options, will be removed later.
972 */
973struct mbuf *
974ip_srcroute()
975{
976	register struct in_addr *p, *q;
977	register struct mbuf *m;
978
979	if (ip_nhops == 0)
980		return ((struct mbuf *)0);
981	m = m_get(M_DONTWAIT, MT_SOOPTS);
982	if (m == 0)
983		return ((struct mbuf *)0);
984
985#define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
986
987	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
988	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
989	    OPTSIZ;
990#ifdef DIAGNOSTIC
991	if (ipprintfs)
992		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
993#endif
994
995	/*
996	 * First save first hop for return route
997	 */
998	p = &ip_srcrt.route[ip_nhops - 1];
999	*(mtod(m, struct in_addr *)) = *p--;
1000#ifdef DIAGNOSTIC
1001	if (ipprintfs)
1002		printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1003#endif
1004
1005	/*
1006	 * Copy option fields and padding (nop) to mbuf.
1007	 */
1008	ip_srcrt.nop = IPOPT_NOP;
1009	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1010	bcopy((caddr_t)&ip_srcrt.nop,
1011	    mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ);
1012	q = (struct in_addr *)(mtod(m, caddr_t) +
1013	    sizeof(struct in_addr) + OPTSIZ);
1014#undef OPTSIZ
1015	/*
1016	 * Record return path as an IP source route,
1017	 * reversing the path (pointers are now aligned).
1018	 */
1019	while (p >= ip_srcrt.route) {
1020#ifdef DIAGNOSTIC
1021		if (ipprintfs)
1022			printf(" %x", ntohl(q->s_addr));
1023#endif
1024		*q++ = *p--;
1025	}
1026	/*
1027	 * Last hop goes to final destination.
1028	 */
1029	*q = ip_srcrt.dst;
1030#ifdef DIAGNOSTIC
1031	if (ipprintfs)
1032		printf(" %x\n", ntohl(q->s_addr));
1033#endif
1034	return (m);
1035}
1036
1037/*
1038 * Strip out IP options, at higher
1039 * level protocol in the kernel.
1040 * Second argument is buffer to which options
1041 * will be moved, and return value is their length.
1042 * XXX should be deleted; last arg currently ignored.
1043 */
1044void
1045ip_stripoptions(m, mopt)
1046	register struct mbuf *m;
1047	struct mbuf *mopt;
1048{
1049	register int i;
1050	struct ip *ip = mtod(m, struct ip *);
1051	register caddr_t opts;
1052	int olen;
1053
1054	olen = (ip->ip_hl<<2) - sizeof (struct ip);
1055	opts = (caddr_t)(ip + 1);
1056	i = m->m_len - (sizeof (struct ip) + olen);
1057	bcopy(opts  + olen, opts, (unsigned)i);
1058	m->m_len -= olen;
1059	if (m->m_flags & M_PKTHDR)
1060		m->m_pkthdr.len -= olen;
1061	ip->ip_hl = sizeof(struct ip) >> 2;
1062}
1063
1064int inetctlerrmap[PRC_NCMDS] = {
1065	0,		0,		0,		0,
1066	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1067	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1068	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1069	0,		0,		0,		0,
1070	ENOPROTOOPT
1071};
1072
1073/*
1074 * Forward a packet.  If some error occurs return the sender
1075 * an icmp packet.  Note we can't always generate a meaningful
1076 * icmp message because icmp doesn't have a large enough repertoire
1077 * of codes and types.
1078 *
1079 * If not forwarding, just drop the packet.  This could be confusing
1080 * if ipforwarding was zero but some routing protocol was advancing
1081 * us as a gateway to somewhere.  However, we must let the routing
1082 * protocol deal with that.
1083 *
1084 * The srcrt parameter indicates whether the packet is being forwarded
1085 * via a source route.
1086 */
1087void
1088ip_forward(m, srcrt)
1089	struct mbuf *m;
1090	int srcrt;
1091{
1092	register struct ip *ip = mtod(m, struct ip *);
1093	register struct sockaddr_in *sin;
1094	register struct rtentry *rt;
1095	int error, type = 0, code = 0;
1096	struct mbuf *mcopy;
1097	n_long dest;
1098	struct ifnet *destifp;
1099
1100	dest = 0;
1101#ifdef DIAGNOSTIC
1102	if (ipprintfs)
1103		printf("forward: src %x dst %x ttl %x\n",
1104		    ip->ip_src.s_addr, ip->ip_dst.s_addr, ip->ip_ttl);
1105#endif
1106	if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) {
1107		ipstat.ips_cantforward++;
1108		m_freem(m);
1109		return;
1110	}
1111	HTONS(ip->ip_id);
1112	if (ip->ip_ttl <= IPTTLDEC) {
1113		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1114		return;
1115	}
1116	ip->ip_ttl -= IPTTLDEC;
1117
1118	sin = satosin(&ipforward_rt.ro_dst);
1119	if ((rt = ipforward_rt.ro_rt) == 0 ||
1120	    !in_hosteq(ip->ip_dst, sin->sin_addr)) {
1121		if (ipforward_rt.ro_rt) {
1122			RTFREE(ipforward_rt.ro_rt);
1123			ipforward_rt.ro_rt = 0;
1124		}
1125		sin->sin_family = AF_INET;
1126		sin->sin_len = sizeof(struct sockaddr_in);
1127		sin->sin_addr = ip->ip_dst;
1128
1129		rtalloc(&ipforward_rt);
1130		if (ipforward_rt.ro_rt == 0) {
1131			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1132			return;
1133		}
1134		rt = ipforward_rt.ro_rt;
1135	}
1136
1137	/*
1138	 * Save at most 68 bytes of the packet in case
1139	 * we need to generate an ICMP message to the src.
1140	 */
1141	mcopy = m_copy(m, 0, imin((int)ip->ip_len, 68));
1142
1143	/*
1144	 * If forwarding packet using same interface that it came in on,
1145	 * perhaps should send a redirect to sender to shortcut a hop.
1146	 * Only send redirect if source is sending directly to us,
1147	 * and if packet was not source routed (or has any options).
1148	 * Also, don't send redirect if forwarding using a default route
1149	 * or a route modified by a redirect.
1150	 */
1151	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1152	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1153	    !in_nullhost(satosin(rt_key(rt))->sin_addr) &&
1154	    ipsendredirects && !srcrt) {
1155		if (rt->rt_ifa &&
1156		    (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1157		    ifatoia(rt->rt_ifa)->ia_subnet) {
1158		    if (rt->rt_flags & RTF_GATEWAY)
1159			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1160		    else
1161			dest = ip->ip_dst.s_addr;
1162		    /* Router requirements says to only send host redirects */
1163		    type = ICMP_REDIRECT;
1164		    code = ICMP_REDIRECT_HOST;
1165#ifdef DIAGNOSTIC
1166		    if (ipprintfs)
1167		    	printf("redirect (%d) to %x\n", code, (u_int32_t)dest);
1168#endif
1169		}
1170	}
1171
1172	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1173	    (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0);
1174	if (error)
1175		ipstat.ips_cantforward++;
1176	else {
1177		ipstat.ips_forward++;
1178		if (type)
1179			ipstat.ips_redirectsent++;
1180		else {
1181			if (mcopy)
1182				m_freem(mcopy);
1183			return;
1184		}
1185	}
1186	if (mcopy == NULL)
1187		return;
1188	destifp = NULL;
1189
1190	switch (error) {
1191
1192	case 0:				/* forwarded, but need redirect */
1193		/* type, code set above */
1194		break;
1195
1196	case ENETUNREACH:		/* shouldn't happen, checked above */
1197	case EHOSTUNREACH:
1198	case ENETDOWN:
1199	case EHOSTDOWN:
1200	default:
1201		type = ICMP_UNREACH;
1202		code = ICMP_UNREACH_HOST;
1203		break;
1204
1205	case EMSGSIZE:
1206		type = ICMP_UNREACH;
1207		code = ICMP_UNREACH_NEEDFRAG;
1208		if (ipforward_rt.ro_rt)
1209			destifp = ipforward_rt.ro_rt->rt_ifp;
1210		ipstat.ips_cantfrag++;
1211		break;
1212
1213	case ENOBUFS:
1214		type = ICMP_SOURCEQUENCH;
1215		code = 0;
1216		break;
1217	}
1218	icmp_error(mcopy, type, code, dest, destifp);
1219}
1220
1221void
1222ip_savecontrol(inp, mp, ip, m)
1223	register struct inpcb *inp;
1224	register struct mbuf **mp;
1225	register struct ip *ip;
1226	register struct mbuf *m;
1227{
1228
1229	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1230		struct timeval tv;
1231
1232		microtime(&tv);
1233		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1234		    SCM_TIMESTAMP, SOL_SOCKET);
1235		if (*mp)
1236			mp = &(*mp)->m_next;
1237	}
1238	if (inp->inp_flags & INP_RECVDSTADDR) {
1239		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1240		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1241		if (*mp)
1242			mp = &(*mp)->m_next;
1243	}
1244#ifdef notyet
1245	/*
1246	 * XXX
1247	 * Moving these out of udp_input() made them even more broken
1248	 * than they already were.
1249	 *	- fenner@parc.xerox.com
1250	 */
1251	/* options were tossed already */
1252	if (inp->inp_flags & INP_RECVOPTS) {
1253		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1254		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1255		if (*mp)
1256			mp = &(*mp)->m_next;
1257	}
1258	/* ip_srcroute doesn't do what we want here, need to fix */
1259	if (inp->inp_flags & INP_RECVRETOPTS) {
1260		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1261		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1262		if (*mp)
1263			mp = &(*mp)->m_next;
1264	}
1265#endif
1266	if (inp->inp_flags & INP_RECVIF) {
1267		struct sockaddr_dl sdl;
1268
1269		sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
1270		sdl.sdl_family = AF_LINK;
1271		sdl.sdl_index = m->m_pkthdr.rcvif ?
1272		    m->m_pkthdr.rcvif->if_index : 0;
1273		sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
1274		*mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
1275		    IP_RECVIF, IPPROTO_IP);
1276		if (*mp)
1277			mp = &(*mp)->m_next;
1278	}
1279}
1280
1281int
1282ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1283	int *name;
1284	u_int namelen;
1285	void *oldp;
1286	size_t *oldlenp;
1287	void *newp;
1288	size_t newlen;
1289{
1290	extern int subnetsarelocal;
1291
1292	int error, old;
1293
1294	/* All sysctl names at this level are terminal. */
1295	if (namelen != 1)
1296		return (ENOTDIR);
1297
1298	switch (name[0]) {
1299	case IPCTL_FORWARDING:
1300		return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding));
1301	case IPCTL_SENDREDIRECTS:
1302		return (sysctl_int(oldp, oldlenp, newp, newlen,
1303			&ipsendredirects));
1304	case IPCTL_DEFTTL:
1305		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl));
1306#ifdef notyet
1307	case IPCTL_DEFMTU:
1308		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu));
1309#endif
1310	case IPCTL_FORWSRCRT:
1311		/* Don't allow this to change in a secure environment.  */
1312		if (securelevel > 0)
1313			return (sysctl_rdint(oldp, oldlenp, newp,
1314			    ip_forwsrcrt));
1315		else
1316			return (sysctl_int(oldp, oldlenp, newp, newlen,
1317			    &ip_forwsrcrt));
1318	case IPCTL_DIRECTEDBCAST:
1319		return (sysctl_int(oldp, oldlenp, newp, newlen,
1320		    &ip_directedbcast));
1321	case IPCTL_ALLOWSRCRT:
1322		return (sysctl_int(oldp, oldlenp, newp, newlen,
1323		    &ip_allowsrcrt));
1324	case IPCTL_SUBNETSARELOCAL:
1325		return (sysctl_int(oldp, oldlenp, newp, newlen,
1326		    &subnetsarelocal));
1327	case IPCTL_MTUDISC:
1328		return (sysctl_int(oldp, oldlenp, newp, newlen,
1329		    &ip_mtudisc));
1330	case IPCTL_ANONPORTMIN:
1331		old = anonportmin;
1332		error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin);
1333		if (anonportmin >= anonportmax || anonportmin > 65535
1334#ifndef IPNOPRIVPORTS
1335		    || anonportmin < IPPORT_RESERVED
1336#endif
1337		    ) {
1338			anonportmin = old;
1339			return (EINVAL);
1340		}
1341		return (error);
1342	case IPCTL_ANONPORTMAX:
1343		old = anonportmax;
1344		error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax);
1345		if (anonportmin >= anonportmax || anonportmax > 65535
1346#ifndef IPNOPRIVPORTS
1347		    || anonportmax < IPPORT_RESERVED
1348#endif
1349		    ) {
1350			anonportmax = old;
1351			return (EINVAL);
1352		}
1353		return (error);
1354	default:
1355		return (EOPNOTSUPP);
1356	}
1357	/* NOTREACHED */
1358}
1359