xref: /freebsd/sys/netinet/ip_input.c (revision ce834215a70ff69e7e222827437116eee2f9ac6f)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
34  * $Id: ip_input.c,v 1.62 1997/05/11 18:05:37 tegge Exp $
35  *	$ANA: ip_input.c,v 1.5 1996/09/18 14:34:59 wollman Exp $
36  */
37 
38 #define	_IP_VHL
39 
40 #include "opt_ipfw.h"
41 
42 #include <stddef.h>
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/domain.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/errno.h>
52 #include <sys/time.h>
53 #include <sys/kernel.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56 
57 #include <net/if.h>
58 #include <net/if_dl.h>
59 #include <net/route.h>
60 #include <net/netisr.h>
61 
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/in_var.h>
65 #include <netinet/ip.h>
66 #include <netinet/in_pcb.h>
67 #include <netinet/in_var.h>
68 #include <netinet/ip_var.h>
69 #include <netinet/ip_icmp.h>
70 #include <machine/in_cksum.h>
71 
72 #include <sys/socketvar.h>
73 
74 #ifdef IPFIREWALL
75 #include <netinet/ip_fw.h>
76 #endif
77 
78 int rsvp_on = 0;
79 static int ip_rsvp_on;
80 struct socket *ip_rsvpd;
81 
82 static int	ipforwarding = 0;
83 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
84 	&ipforwarding, 0, "");
85 
86 static int	ipsendredirects = 1; /* XXX */
87 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
88 	&ipsendredirects, 0, "");
89 
90 int	ip_defttl = IPDEFTTL;
91 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
92 	&ip_defttl, 0, "");
93 
94 static int	ip_dosourceroute = 0;
95 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
96 	&ip_dosourceroute, 0, "");
97 #ifdef DIAGNOSTIC
98 static int	ipprintfs = 0;
99 #endif
100 
101 extern	struct domain inetdomain;
102 extern	struct protosw inetsw[];
103 u_char	ip_protox[IPPROTO_MAX];
104 static int	ipqmaxlen = IFQ_MAXLEN;
105 struct	in_ifaddrhead in_ifaddrhead; /* first inet address */
106 struct	ifqueue ipintrq;
107 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RD,
108 	&ipintrq.ifq_maxlen, 0, "");
109 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
110 	&ipintrq.ifq_drops, 0, "");
111 
112 struct ipstat ipstat;
113 static struct ipq ipq;
114 
115 #ifdef IPCTL_DEFMTU
116 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
117 	&ip_mtu, 0, "");
118 #endif
119 
120 #if !defined(COMPAT_IPFW) || COMPAT_IPFW == 1
121 #undef COMPAT_IPFW
122 #define COMPAT_IPFW 1
123 #else
124 #undef COMPAT_IPFW
125 #endif
126 
127 #ifdef COMPAT_IPFW
128 /* Firewall hooks */
129 ip_fw_chk_t *ip_fw_chk_ptr;
130 ip_fw_ctl_t *ip_fw_ctl_ptr;
131 
132 /* IP Network Address Translation (NAT) hooks */
133 ip_nat_t *ip_nat_ptr;
134 ip_nat_ctl_t *ip_nat_ctl_ptr;
135 #endif
136 
137 #if defined(IPFILTER_LKM) || defined(IPFILTER)
138 int fr_check __P((struct ip *, int, struct ifnet *, int, struct mbuf **));
139 int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL;
140 #endif
141 
142 
143 /*
144  * We need to save the IP options in case a protocol wants to respond
145  * to an incoming packet over the same route if the packet got here
146  * using IP source routing.  This allows connection establishment and
147  * maintenance when the remote end is on a network that is not known
148  * to us.
149  */
150 static int	ip_nhops = 0;
151 static	struct ip_srcrt {
152 	struct	in_addr dst;			/* final destination */
153 	char	nop;				/* one NOP to align */
154 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
155 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
156 } ip_srcrt;
157 
158 #ifdef IPDIVERT
159 /*
160  * Shared variable between ip_input() and ip_reass() to communicate
161  * about which packets, once assembled from fragments, get diverted,
162  * and to which port.
163  */
164 static u_short	frag_divert_port;
165 #endif
166 
167 static void save_rte __P((u_char *, struct in_addr));
168 static void	 ip_deq __P((struct ipasfrag *));
169 static int	 ip_dooptions __P((struct mbuf *));
170 static void	 ip_enq __P((struct ipasfrag *, struct ipasfrag *));
171 static void	 ip_forward __P((struct mbuf *, int));
172 static void	 ip_freef __P((struct ipq *));
173 static struct ip *
174 	 ip_reass __P((struct ipasfrag *, struct ipq *));
175 static struct in_ifaddr *
176 	 ip_rtaddr __P((struct in_addr));
177 static void	ipintr __P((void));
178 /*
179  * IP initialization: fill in IP protocol switch table.
180  * All protocols not implemented in kernel go to raw IP protocol handler.
181  */
182 void
183 ip_init()
184 {
185 	register struct protosw *pr;
186 	register int i;
187 
188 	TAILQ_INIT(&in_ifaddrhead);
189 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
190 	if (pr == 0)
191 		panic("ip_init");
192 	for (i = 0; i < IPPROTO_MAX; i++)
193 		ip_protox[i] = pr - inetsw;
194 	for (pr = inetdomain.dom_protosw;
195 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
196 		if (pr->pr_domain->dom_family == PF_INET &&
197 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
198 			ip_protox[pr->pr_protocol] = pr - inetsw;
199 	ipq.next = ipq.prev = &ipq;
200 	ip_id = time.tv_sec & 0xffff;
201 	ipintrq.ifq_maxlen = ipqmaxlen;
202 #ifdef IPFIREWALL
203 	ip_fw_init();
204 #endif
205 #ifdef IPNAT
206         ip_nat_init();
207 #endif
208 
209 }
210 
211 static struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
212 static struct	route ipforward_rt;
213 
214 /*
215  * Ip input routine.  Checksum and byte swap header.  If fragmented
216  * try to reassemble.  Process options.  Pass to next level.
217  */
218 void
219 ip_input(struct mbuf *m)
220 {
221 	struct ip *ip;
222 	struct ipq *fp;
223 	struct in_ifaddr *ia;
224 	int hlen;
225 	u_short sum;
226 
227 #ifdef	DIAGNOSTIC
228 	if ((m->m_flags & M_PKTHDR) == 0)
229 		panic("ip_input no HDR");
230 #endif
231 	/*
232 	 * If no IP addresses have been set yet but the interfaces
233 	 * are receiving, can't do anything with incoming packets yet.
234 	 * XXX This is broken! We should be able to receive broadcasts
235 	 * and multicasts even without any local addresses configured.
236 	 */
237 	if (TAILQ_EMPTY(&in_ifaddrhead))
238 		goto bad;
239 	ipstat.ips_total++;
240 
241 	if (m->m_pkthdr.len < sizeof(struct ip))
242 		goto tooshort;
243 
244 #ifdef	DIAGNOSTIC
245 	if (m->m_len < sizeof(struct ip))
246 		panic("ipintr mbuf too short");
247 #endif
248 
249 	if (m->m_len < sizeof (struct ip) &&
250 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
251 		ipstat.ips_toosmall++;
252 		return;
253 	}
254 	ip = mtod(m, struct ip *);
255 
256 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
257 		ipstat.ips_badvers++;
258 		goto bad;
259 	}
260 
261 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
262 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
263 		ipstat.ips_badhlen++;
264 		goto bad;
265 	}
266 	if (hlen > m->m_len) {
267 		if ((m = m_pullup(m, hlen)) == 0) {
268 			ipstat.ips_badhlen++;
269 			return;
270 		}
271 		ip = mtod(m, struct ip *);
272 	}
273 	if (hlen == sizeof(struct ip)) {
274 		sum = in_cksum_hdr(ip);
275 	} else {
276 		sum = in_cksum(m, hlen);
277 	}
278 	if (sum) {
279 		ipstat.ips_badsum++;
280 		goto bad;
281 	}
282 
283 	/*
284 	 * Convert fields to host representation.
285 	 */
286 	NTOHS(ip->ip_len);
287 	if (ip->ip_len < hlen) {
288 		ipstat.ips_badlen++;
289 		goto bad;
290 	}
291 	NTOHS(ip->ip_id);
292 	NTOHS(ip->ip_off);
293 
294 	/*
295 	 * Check that the amount of data in the buffers
296 	 * is as at least much as the IP header would have us expect.
297 	 * Trim mbufs if longer than we expect.
298 	 * Drop packet if shorter than we expect.
299 	 */
300 	if (m->m_pkthdr.len < ip->ip_len) {
301 tooshort:
302 		ipstat.ips_tooshort++;
303 		goto bad;
304 	}
305 	if (m->m_pkthdr.len > ip->ip_len) {
306 		if (m->m_len == m->m_pkthdr.len) {
307 			m->m_len = ip->ip_len;
308 			m->m_pkthdr.len = ip->ip_len;
309 		} else
310 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
311 	}
312 	/*
313 	 * IpHack's section.
314 	 * Right now when no processing on packet has done
315 	 * and it is still fresh out of network we do our black
316 	 * deals with it.
317 	 * - Firewall: deny/allow/divert
318 	 * - Xlate: translate packet's addr/port (NAT).
319 	 * - Wrap: fake packet's addr/port <unimpl.>
320 	 * - Encapsulate: put it in another IP and send out. <unimp.>
321  	 */
322 #if defined(IPFILTER) || defined(IPFILTER_LKM)
323 	/*
324 	 * Check if we want to allow this packet to be processed.
325 	 * Consider it to be bad if not.
326 	 */
327 	if (fr_check) {
328 		struct	mbuf	*m1 = m;
329 
330 		if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1)
331 			return;
332 		ip = mtod(m = m1, struct ip *);
333 	}
334 #endif
335 #ifdef COMPAT_IPFW
336 	if (ip_fw_chk_ptr) {
337 #ifdef IPDIVERT
338 		u_short port;
339 
340 		port = (*ip_fw_chk_ptr)(&ip, hlen, NULL, ip_divert_ignore, &m);
341 		ip_divert_ignore = 0;
342 		if (port) {			/* Divert packet */
343 			frag_divert_port = port;
344 			goto ours;
345 		}
346 #else
347 		/* If ipfw says divert, we have to just drop packet */
348 		if ((*ip_fw_chk_ptr)(&ip, hlen, NULL, 0, &m)) {
349 			m_freem(m);
350 			m = NULL;
351 		}
352 #endif
353 		if (!m)
354 			return;
355 	}
356 
357         if (ip_nat_ptr && !(*ip_nat_ptr)(&ip, &m, m->m_pkthdr.rcvif, IP_NAT_IN))
358 		return;
359 #endif
360 
361 	/*
362 	 * Process options and, if not destined for us,
363 	 * ship it on.  ip_dooptions returns 1 when an
364 	 * error was detected (causing an icmp message
365 	 * to be sent and the original packet to be freed).
366 	 */
367 	ip_nhops = 0;		/* for source routed packets */
368 	if (hlen > sizeof (struct ip) && ip_dooptions(m))
369 		return;
370 
371         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
372          * matter if it is destined to another node, or whether it is
373          * a multicast one, RSVP wants it! and prevents it from being forwarded
374          * anywhere else. Also checks if the rsvp daemon is running before
375 	 * grabbing the packet.
376          */
377 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
378 		goto ours;
379 
380 	/*
381 	 * Check our list of addresses, to see if the packet is for us.
382 	 */
383 	for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) {
384 #define	satosin(sa)	((struct sockaddr_in *)(sa))
385 
386 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
387 			goto ours;
388 #ifdef BOOTP_COMPAT
389 		if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
390 			goto ours;
391 #endif
392 		if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) {
393 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
394 			    ip->ip_dst.s_addr)
395 				goto ours;
396 			if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr)
397 				goto ours;
398 		}
399 	}
400 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
401 		struct in_multi *inm;
402 		if (ip_mrouter) {
403 			/*
404 			 * If we are acting as a multicast router, all
405 			 * incoming multicast packets are passed to the
406 			 * kernel-level multicast forwarding function.
407 			 * The packet is returned (relatively) intact; if
408 			 * ip_mforward() returns a non-zero value, the packet
409 			 * must be discarded, else it may be accepted below.
410 			 *
411 			 * (The IP ident field is put in the same byte order
412 			 * as expected when ip_mforward() is called from
413 			 * ip_output().)
414 			 */
415 			ip->ip_id = htons(ip->ip_id);
416 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
417 				ipstat.ips_cantforward++;
418 				m_freem(m);
419 				return;
420 			}
421 			ip->ip_id = ntohs(ip->ip_id);
422 
423 			/*
424 			 * The process-level routing demon needs to receive
425 			 * all multicast IGMP packets, whether or not this
426 			 * host belongs to their destination groups.
427 			 */
428 			if (ip->ip_p == IPPROTO_IGMP)
429 				goto ours;
430 			ipstat.ips_forward++;
431 		}
432 		/*
433 		 * See if we belong to the destination multicast group on the
434 		 * arrival interface.
435 		 */
436 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
437 		if (inm == NULL) {
438 			ipstat.ips_notmember++;
439 			m_freem(m);
440 			return;
441 		}
442 		goto ours;
443 	}
444 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
445 		goto ours;
446 	if (ip->ip_dst.s_addr == INADDR_ANY)
447 		goto ours;
448 
449 	/*
450 	 * Not for us; forward if possible and desirable.
451 	 */
452 	if (ipforwarding == 0) {
453 		ipstat.ips_cantforward++;
454 		m_freem(m);
455 	} else
456 		ip_forward(m, 0);
457 	return;
458 
459 ours:
460 
461 	/*
462 	 * If offset or IP_MF are set, must reassemble.
463 	 * Otherwise, nothing need be done.
464 	 * (We could look in the reassembly queue to see
465 	 * if the packet was previously fragmented,
466 	 * but it's not worth the time; just let them time out.)
467 	 */
468 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
469 		if (m->m_flags & M_EXT) {		/* XXX */
470 			if ((m = m_pullup(m, sizeof (struct ip))) == 0) {
471 				ipstat.ips_toosmall++;
472 #ifdef IPDIVERT
473 				frag_divert_port = 0;
474 #endif
475 				return;
476 			}
477 			ip = mtod(m, struct ip *);
478 		}
479 		/*
480 		 * Look for queue of fragments
481 		 * of this datagram.
482 		 */
483 		for (fp = ipq.next; fp != &ipq; fp = fp->next)
484 			if (ip->ip_id == fp->ipq_id &&
485 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
486 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
487 			    ip->ip_p == fp->ipq_p)
488 				goto found;
489 		fp = 0;
490 found:
491 
492 		/*
493 		 * Adjust ip_len to not reflect header,
494 		 * set ip_mff if more fragments are expected,
495 		 * convert offset of this to bytes.
496 		 */
497 		ip->ip_len -= hlen;
498 		((struct ipasfrag *)ip)->ipf_mff &= ~1;
499 		if (ip->ip_off & IP_MF)
500 			((struct ipasfrag *)ip)->ipf_mff |= 1;
501 		ip->ip_off <<= 3;
502 
503 		/*
504 		 * If datagram marked as having more fragments
505 		 * or if this is not the first fragment,
506 		 * attempt reassembly; if it succeeds, proceed.
507 		 */
508 		if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) {
509 			ipstat.ips_fragments++;
510 			ip = ip_reass((struct ipasfrag *)ip, fp);
511 			if (ip == 0)
512 				return;
513 			ipstat.ips_reassembled++;
514 			m = dtom(ip);
515 		} else
516 			if (fp)
517 				ip_freef(fp);
518 	} else
519 		ip->ip_len -= hlen;
520 
521 #ifdef IPDIVERT
522 	/*
523 	 * Divert reassembled packets to the divert protocol if required
524 	 */
525 	if (frag_divert_port) {
526 		ipstat.ips_delivered++;
527 		ip_divert_port = frag_divert_port;
528 		frag_divert_port = 0;
529 		(*inetsw[ip_protox[IPPROTO_DIVERT]].pr_input)(m, hlen);
530 		return;
531 	}
532 #endif
533 
534 	/*
535 	 * Switch out to protocol's input routine.
536 	 */
537 	ipstat.ips_delivered++;
538 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
539 	return;
540 bad:
541 	m_freem(m);
542 }
543 
544 /*
545  * IP software interrupt routine - to go away sometime soon
546  */
547 static void
548 ipintr(void)
549 {
550 	int s;
551 	struct mbuf *m;
552 
553 	while(1) {
554 		s = splimp();
555 		IF_DEQUEUE(&ipintrq, m);
556 		splx(s);
557 		if (m == 0)
558 			return;
559 		ip_input(m);
560 	}
561 }
562 
563 NETISR_SET(NETISR_IP, ipintr);
564 
565 /*
566  * Take incoming datagram fragment and try to
567  * reassemble it into whole datagram.  If a chain for
568  * reassembly of this datagram already exists, then it
569  * is given as fp; otherwise have to make a chain.
570  */
571 static struct ip *
572 ip_reass(ip, fp)
573 	register struct ipasfrag *ip;
574 	register struct ipq *fp;
575 {
576 	register struct mbuf *m = dtom(ip);
577 	register struct ipasfrag *q;
578 	struct mbuf *t;
579 	int hlen = ip->ip_hl << 2;
580 	int i, next;
581 
582 	/*
583 	 * Presence of header sizes in mbufs
584 	 * would confuse code below.
585 	 */
586 	m->m_data += hlen;
587 	m->m_len -= hlen;
588 
589 	/*
590 	 * If first fragment to arrive, create a reassembly queue.
591 	 */
592 	if (fp == 0) {
593 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
594 			goto dropfrag;
595 		fp = mtod(t, struct ipq *);
596 		insque(fp, &ipq);
597 		fp->ipq_ttl = IPFRAGTTL;
598 		fp->ipq_p = ip->ip_p;
599 		fp->ipq_id = ip->ip_id;
600 		fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp;
601 		fp->ipq_src = ((struct ip *)ip)->ip_src;
602 		fp->ipq_dst = ((struct ip *)ip)->ip_dst;
603 #ifdef IPDIVERT
604 		fp->ipq_divert = 0;
605 #endif
606 		q = (struct ipasfrag *)fp;
607 		goto insert;
608 	}
609 
610 	/*
611 	 * Find a segment which begins after this one does.
612 	 */
613 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next)
614 		if (q->ip_off > ip->ip_off)
615 			break;
616 
617 	/*
618 	 * If there is a preceding segment, it may provide some of
619 	 * our data already.  If so, drop the data from the incoming
620 	 * segment.  If it provides all of our data, drop us.
621 	 */
622 	if (q->ipf_prev != (struct ipasfrag *)fp) {
623 		i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off;
624 		if (i > 0) {
625 			if (i >= ip->ip_len)
626 				goto dropfrag;
627 			m_adj(dtom(ip), i);
628 			ip->ip_off += i;
629 			ip->ip_len -= i;
630 		}
631 	}
632 
633 	/*
634 	 * While we overlap succeeding segments trim them or,
635 	 * if they are completely covered, dequeue them.
636 	 */
637 	while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) {
638 		struct mbuf *m0;
639 
640 		i = (ip->ip_off + ip->ip_len) - q->ip_off;
641 		if (i < q->ip_len) {
642 			q->ip_len -= i;
643 			q->ip_off += i;
644 			m_adj(dtom(q), i);
645 			break;
646 		}
647 		m0 = dtom(q);
648 		q = q->ipf_next;
649 		ip_deq(q->ipf_prev);
650 		m_freem(m0);
651 	}
652 
653 insert:
654 
655 #ifdef IPDIVERT
656 	/*
657 	 * Any fragment diverting causes the whole packet to divert
658 	 */
659 	if (frag_divert_port != 0)
660 		fp->ipq_divert = frag_divert_port;
661 	frag_divert_port = 0;
662 #endif
663 
664 	/*
665 	 * Stick new segment in its place;
666 	 * check for complete reassembly.
667 	 */
668 	ip_enq(ip, q->ipf_prev);
669 	next = 0;
670 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) {
671 		if (q->ip_off != next)
672 			return (0);
673 		next += q->ip_len;
674 	}
675 	if (q->ipf_prev->ipf_mff & 1)
676 		return (0);
677 
678 	/*
679 	 * Reassembly is complete.  Make sure the packet is a sane size.
680 	 */
681 	if (next + (IP_VHL_HL(((struct ip *)fp->ipq_next)->ip_vhl) << 2)
682 							> IP_MAXPACKET) {
683 		ipstat.ips_toolong++;
684 		ip_freef(fp);
685 		return (0);
686 	}
687 
688 	/*
689 	 * Concatenate fragments.
690 	 */
691 	q = fp->ipq_next;
692 	m = dtom(q);
693 	t = m->m_next;
694 	m->m_next = 0;
695 	m_cat(m, t);
696 	q = q->ipf_next;
697 	while (q != (struct ipasfrag *)fp) {
698 		t = dtom(q);
699 		q = q->ipf_next;
700 		m_cat(m, t);
701 	}
702 
703 #ifdef IPDIVERT
704 	/*
705 	 * Record divert port for packet, if any
706 	 */
707 	frag_divert_port = fp->ipq_divert;
708 #endif
709 
710 	/*
711 	 * Create header for new ip packet by
712 	 * modifying header of first packet;
713 	 * dequeue and discard fragment reassembly header.
714 	 * Make header visible.
715 	 */
716 	ip = fp->ipq_next;
717 	ip->ip_len = next;
718 	ip->ipf_mff &= ~1;
719 	((struct ip *)ip)->ip_src = fp->ipq_src;
720 	((struct ip *)ip)->ip_dst = fp->ipq_dst;
721 	remque(fp);
722 	(void) m_free(dtom(fp));
723 	m = dtom(ip);
724 	m->m_len += (ip->ip_hl << 2);
725 	m->m_data -= (ip->ip_hl << 2);
726 	/* some debugging cruft by sklower, below, will go away soon */
727 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
728 		register int plen = 0;
729 		for (t = m; m; m = m->m_next)
730 			plen += m->m_len;
731 		t->m_pkthdr.len = plen;
732 	}
733 	return ((struct ip *)ip);
734 
735 dropfrag:
736 	ipstat.ips_fragdropped++;
737 	m_freem(m);
738 	return (0);
739 }
740 
741 /*
742  * Free a fragment reassembly header and all
743  * associated datagrams.
744  */
745 static void
746 ip_freef(fp)
747 	struct ipq *fp;
748 {
749 	register struct ipasfrag *q, *p;
750 
751 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) {
752 		p = q->ipf_next;
753 		ip_deq(q);
754 		m_freem(dtom(q));
755 	}
756 	remque(fp);
757 	(void) m_free(dtom(fp));
758 }
759 
760 /*
761  * Put an ip fragment on a reassembly chain.
762  * Like insque, but pointers in middle of structure.
763  */
764 static void
765 ip_enq(p, prev)
766 	register struct ipasfrag *p, *prev;
767 {
768 
769 	p->ipf_prev = prev;
770 	p->ipf_next = prev->ipf_next;
771 	prev->ipf_next->ipf_prev = p;
772 	prev->ipf_next = p;
773 }
774 
775 /*
776  * To ip_enq as remque is to insque.
777  */
778 static void
779 ip_deq(p)
780 	register struct ipasfrag *p;
781 {
782 
783 	p->ipf_prev->ipf_next = p->ipf_next;
784 	p->ipf_next->ipf_prev = p->ipf_prev;
785 }
786 
787 /*
788  * IP timer processing;
789  * if a timer expires on a reassembly
790  * queue, discard it.
791  */
792 void
793 ip_slowtimo()
794 {
795 	register struct ipq *fp;
796 	int s = splnet();
797 
798 	fp = ipq.next;
799 	if (fp == 0) {
800 		splx(s);
801 		return;
802 	}
803 	while (fp != &ipq) {
804 		--fp->ipq_ttl;
805 		fp = fp->next;
806 		if (fp->prev->ipq_ttl == 0) {
807 			ipstat.ips_fragtimeout++;
808 			ip_freef(fp->prev);
809 		}
810 	}
811 	splx(s);
812 }
813 
814 /*
815  * Drain off all datagram fragments.
816  */
817 void
818 ip_drain()
819 {
820 	while (ipq.next != &ipq) {
821 		ipstat.ips_fragdropped++;
822 		ip_freef(ipq.next);
823 	}
824 
825 	in_rtqdrain();
826 }
827 
828 /*
829  * Do option processing on a datagram,
830  * possibly discarding it if bad options are encountered,
831  * or forwarding it if source-routed.
832  * Returns 1 if packet has been forwarded/freed,
833  * 0 if the packet should be processed further.
834  */
835 static int
836 ip_dooptions(m)
837 	struct mbuf *m;
838 {
839 	register struct ip *ip = mtod(m, struct ip *);
840 	register u_char *cp;
841 	register struct ip_timestamp *ipt;
842 	register struct in_ifaddr *ia;
843 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
844 	struct in_addr *sin, dst;
845 	n_time ntime;
846 
847 	dst = ip->ip_dst;
848 	cp = (u_char *)(ip + 1);
849 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
850 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
851 		opt = cp[IPOPT_OPTVAL];
852 		if (opt == IPOPT_EOL)
853 			break;
854 		if (opt == IPOPT_NOP)
855 			optlen = 1;
856 		else {
857 			optlen = cp[IPOPT_OLEN];
858 			if (optlen <= 0 || optlen > cnt) {
859 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
860 				goto bad;
861 			}
862 		}
863 		switch (opt) {
864 
865 		default:
866 			break;
867 
868 		/*
869 		 * Source routing with record.
870 		 * Find interface with current destination address.
871 		 * If none on this machine then drop if strictly routed,
872 		 * or do nothing if loosely routed.
873 		 * Record interface address and bring up next address
874 		 * component.  If strictly routed make sure next
875 		 * address is on directly accessible net.
876 		 */
877 		case IPOPT_LSRR:
878 		case IPOPT_SSRR:
879 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
880 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
881 				goto bad;
882 			}
883 			ipaddr.sin_addr = ip->ip_dst;
884 			ia = (struct in_ifaddr *)
885 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
886 			if (ia == 0) {
887 				if (opt == IPOPT_SSRR) {
888 					type = ICMP_UNREACH;
889 					code = ICMP_UNREACH_SRCFAIL;
890 					goto bad;
891 				}
892 				/*
893 				 * Loose routing, and not at next destination
894 				 * yet; nothing to do except forward.
895 				 */
896 				break;
897 			}
898 			off--;			/* 0 origin */
899 			if (off > optlen - sizeof(struct in_addr)) {
900 				/*
901 				 * End of source route.  Should be for us.
902 				 */
903 				save_rte(cp, ip->ip_src);
904 				break;
905 			}
906 
907 			if (!ip_dosourceroute) {
908 				char buf[4*sizeof "123"];
909 				strcpy(buf, inet_ntoa(ip->ip_dst));
910 
911 				log(LOG_WARNING,
912 				    "attempted source route from %s to %s\n",
913 				    inet_ntoa(ip->ip_src), buf);
914 				type = ICMP_UNREACH;
915 				code = ICMP_UNREACH_SRCFAIL;
916 				goto bad;
917 			}
918 
919 			/*
920 			 * locate outgoing interface
921 			 */
922 			(void)memcpy(&ipaddr.sin_addr, cp + off,
923 			    sizeof(ipaddr.sin_addr));
924 
925 			if (opt == IPOPT_SSRR) {
926 #define	INA	struct in_ifaddr *
927 #define	SA	struct sockaddr *
928 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
929 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
930 			} else
931 				ia = ip_rtaddr(ipaddr.sin_addr);
932 			if (ia == 0) {
933 				type = ICMP_UNREACH;
934 				code = ICMP_UNREACH_SRCFAIL;
935 				goto bad;
936 			}
937 			ip->ip_dst = ipaddr.sin_addr;
938 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
939 			    sizeof(struct in_addr));
940 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
941 			/*
942 			 * Let ip_intr's mcast routing check handle mcast pkts
943 			 */
944 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
945 			break;
946 
947 		case IPOPT_RR:
948 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
949 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
950 				goto bad;
951 			}
952 			/*
953 			 * If no space remains, ignore.
954 			 */
955 			off--;			/* 0 origin */
956 			if (off > optlen - sizeof(struct in_addr))
957 				break;
958 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
959 			    sizeof(ipaddr.sin_addr));
960 			/*
961 			 * locate outgoing interface; if we're the destination,
962 			 * use the incoming interface (should be same).
963 			 */
964 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
965 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
966 				type = ICMP_UNREACH;
967 				code = ICMP_UNREACH_HOST;
968 				goto bad;
969 			}
970 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
971 			    sizeof(struct in_addr));
972 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
973 			break;
974 
975 		case IPOPT_TS:
976 			code = cp - (u_char *)ip;
977 			ipt = (struct ip_timestamp *)cp;
978 			if (ipt->ipt_len < 5)
979 				goto bad;
980 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) {
981 				if (++ipt->ipt_oflw == 0)
982 					goto bad;
983 				break;
984 			}
985 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
986 			switch (ipt->ipt_flg) {
987 
988 			case IPOPT_TS_TSONLY:
989 				break;
990 
991 			case IPOPT_TS_TSANDADDR:
992 				if (ipt->ipt_ptr + sizeof(n_time) +
993 				    sizeof(struct in_addr) > ipt->ipt_len)
994 					goto bad;
995 				ipaddr.sin_addr = dst;
996 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
997 							    m->m_pkthdr.rcvif);
998 				if (ia == 0)
999 					continue;
1000 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1001 				    sizeof(struct in_addr));
1002 				ipt->ipt_ptr += sizeof(struct in_addr);
1003 				break;
1004 
1005 			case IPOPT_TS_PRESPEC:
1006 				if (ipt->ipt_ptr + sizeof(n_time) +
1007 				    sizeof(struct in_addr) > ipt->ipt_len)
1008 					goto bad;
1009 				(void)memcpy(&ipaddr.sin_addr, sin,
1010 				    sizeof(struct in_addr));
1011 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1012 					continue;
1013 				ipt->ipt_ptr += sizeof(struct in_addr);
1014 				break;
1015 
1016 			default:
1017 				goto bad;
1018 			}
1019 			ntime = iptime();
1020 			(void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
1021 			    sizeof(n_time));
1022 			ipt->ipt_ptr += sizeof(n_time);
1023 		}
1024 	}
1025 	if (forward) {
1026 		ip_forward(m, 1);
1027 		return (1);
1028 	}
1029 	return (0);
1030 bad:
1031 	ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2;   /* XXX icmp_error adds in hdr length */
1032 	icmp_error(m, type, code, 0, 0);
1033 	ipstat.ips_badoptions++;
1034 	return (1);
1035 }
1036 
1037 /*
1038  * Given address of next destination (final or next hop),
1039  * return internet address info of interface to be used to get there.
1040  */
1041 static struct in_ifaddr *
1042 ip_rtaddr(dst)
1043 	 struct in_addr dst;
1044 {
1045 	register struct sockaddr_in *sin;
1046 
1047 	sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1048 
1049 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1050 		if (ipforward_rt.ro_rt) {
1051 			RTFREE(ipforward_rt.ro_rt);
1052 			ipforward_rt.ro_rt = 0;
1053 		}
1054 		sin->sin_family = AF_INET;
1055 		sin->sin_len = sizeof(*sin);
1056 		sin->sin_addr = dst;
1057 
1058 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1059 	}
1060 	if (ipforward_rt.ro_rt == 0)
1061 		return ((struct in_ifaddr *)0);
1062 	return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
1063 }
1064 
1065 /*
1066  * Save incoming source route for use in replies,
1067  * to be picked up later by ip_srcroute if the receiver is interested.
1068  */
1069 void
1070 save_rte(option, dst)
1071 	u_char *option;
1072 	struct in_addr dst;
1073 {
1074 	unsigned olen;
1075 
1076 	olen = option[IPOPT_OLEN];
1077 #ifdef DIAGNOSTIC
1078 	if (ipprintfs)
1079 		printf("save_rte: olen %d\n", olen);
1080 #endif
1081 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1082 		return;
1083 	bcopy(option, ip_srcrt.srcopt, olen);
1084 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1085 	ip_srcrt.dst = dst;
1086 }
1087 
1088 /*
1089  * Retrieve incoming source route for use in replies,
1090  * in the same form used by setsockopt.
1091  * The first hop is placed before the options, will be removed later.
1092  */
1093 struct mbuf *
1094 ip_srcroute()
1095 {
1096 	register struct in_addr *p, *q;
1097 	register struct mbuf *m;
1098 
1099 	if (ip_nhops == 0)
1100 		return ((struct mbuf *)0);
1101 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1102 	if (m == 0)
1103 		return ((struct mbuf *)0);
1104 
1105 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1106 
1107 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1108 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1109 	    OPTSIZ;
1110 #ifdef DIAGNOSTIC
1111 	if (ipprintfs)
1112 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1113 #endif
1114 
1115 	/*
1116 	 * First save first hop for return route
1117 	 */
1118 	p = &ip_srcrt.route[ip_nhops - 1];
1119 	*(mtod(m, struct in_addr *)) = *p--;
1120 #ifdef DIAGNOSTIC
1121 	if (ipprintfs)
1122 		printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr));
1123 #endif
1124 
1125 	/*
1126 	 * Copy option fields and padding (nop) to mbuf.
1127 	 */
1128 	ip_srcrt.nop = IPOPT_NOP;
1129 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1130 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1131 	    &ip_srcrt.nop, OPTSIZ);
1132 	q = (struct in_addr *)(mtod(m, caddr_t) +
1133 	    sizeof(struct in_addr) + OPTSIZ);
1134 #undef OPTSIZ
1135 	/*
1136 	 * Record return path as an IP source route,
1137 	 * reversing the path (pointers are now aligned).
1138 	 */
1139 	while (p >= ip_srcrt.route) {
1140 #ifdef DIAGNOSTIC
1141 		if (ipprintfs)
1142 			printf(" %lx", ntohl(q->s_addr));
1143 #endif
1144 		*q++ = *p--;
1145 	}
1146 	/*
1147 	 * Last hop goes to final destination.
1148 	 */
1149 	*q = ip_srcrt.dst;
1150 #ifdef DIAGNOSTIC
1151 	if (ipprintfs)
1152 		printf(" %lx\n", ntohl(q->s_addr));
1153 #endif
1154 	return (m);
1155 }
1156 
1157 /*
1158  * Strip out IP options, at higher
1159  * level protocol in the kernel.
1160  * Second argument is buffer to which options
1161  * will be moved, and return value is their length.
1162  * XXX should be deleted; last arg currently ignored.
1163  */
1164 void
1165 ip_stripoptions(m, mopt)
1166 	register struct mbuf *m;
1167 	struct mbuf *mopt;
1168 {
1169 	register int i;
1170 	struct ip *ip = mtod(m, struct ip *);
1171 	register caddr_t opts;
1172 	int olen;
1173 
1174 	olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1175 	opts = (caddr_t)(ip + 1);
1176 	i = m->m_len - (sizeof (struct ip) + olen);
1177 	bcopy(opts + olen, opts, (unsigned)i);
1178 	m->m_len -= olen;
1179 	if (m->m_flags & M_PKTHDR)
1180 		m->m_pkthdr.len -= olen;
1181 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1182 }
1183 
1184 u_char inetctlerrmap[PRC_NCMDS] = {
1185 	0,		0,		0,		0,
1186 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1187 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1188 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1189 	0,		0,		0,		0,
1190 	ENOPROTOOPT
1191 };
1192 
1193 /*
1194  * Forward a packet.  If some error occurs return the sender
1195  * an icmp packet.  Note we can't always generate a meaningful
1196  * icmp message because icmp doesn't have a large enough repertoire
1197  * of codes and types.
1198  *
1199  * If not forwarding, just drop the packet.  This could be confusing
1200  * if ipforwarding was zero but some routing protocol was advancing
1201  * us as a gateway to somewhere.  However, we must let the routing
1202  * protocol deal with that.
1203  *
1204  * The srcrt parameter indicates whether the packet is being forwarded
1205  * via a source route.
1206  */
1207 static void
1208 ip_forward(m, srcrt)
1209 	struct mbuf *m;
1210 	int srcrt;
1211 {
1212 	register struct ip *ip = mtod(m, struct ip *);
1213 	register struct sockaddr_in *sin;
1214 	register struct rtentry *rt;
1215 	int error, type = 0, code = 0;
1216 	struct mbuf *mcopy;
1217 	n_long dest;
1218 	struct ifnet *destifp;
1219 
1220 	dest = 0;
1221 #ifdef DIAGNOSTIC
1222 	if (ipprintfs)
1223 		printf("forward: src %lx dst %lx ttl %x\n",
1224 			ip->ip_src.s_addr, ip->ip_dst.s_addr, ip->ip_ttl);
1225 #endif
1226 
1227 
1228 	if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) {
1229 		ipstat.ips_cantforward++;
1230 		m_freem(m);
1231 		return;
1232 	}
1233 	HTONS(ip->ip_id);
1234 	if (ip->ip_ttl <= IPTTLDEC) {
1235 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1236 		return;
1237 	}
1238 	ip->ip_ttl -= IPTTLDEC;
1239 
1240 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1241 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1242 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1243 		if (ipforward_rt.ro_rt) {
1244 			RTFREE(ipforward_rt.ro_rt);
1245 			ipforward_rt.ro_rt = 0;
1246 		}
1247 		sin->sin_family = AF_INET;
1248 		sin->sin_len = sizeof(*sin);
1249 		sin->sin_addr = ip->ip_dst;
1250 
1251 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1252 		if (ipforward_rt.ro_rt == 0) {
1253 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1254 			return;
1255 		}
1256 		rt = ipforward_rt.ro_rt;
1257 	}
1258 
1259 	/*
1260 	 * Save at most 64 bytes of the packet in case
1261 	 * we need to generate an ICMP message to the src.
1262 	 */
1263 	mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64));
1264 
1265 	/*
1266 	 * If forwarding packet using same interface that it came in on,
1267 	 * perhaps should send a redirect to sender to shortcut a hop.
1268 	 * Only send redirect if source is sending directly to us,
1269 	 * and if packet was not source routed (or has any options).
1270 	 * Also, don't send redirect if forwarding using a default route
1271 	 * or a route modified by a redirect.
1272 	 */
1273 #define	satosin(sa)	((struct sockaddr_in *)(sa))
1274 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1275 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1276 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1277 	    ipsendredirects && !srcrt) {
1278 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1279 		u_long src = ntohl(ip->ip_src.s_addr);
1280 
1281 		if (RTA(rt) &&
1282 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1283 		    if (rt->rt_flags & RTF_GATEWAY)
1284 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1285 		    else
1286 			dest = ip->ip_dst.s_addr;
1287 		    /* Router requirements says to only send host redirects */
1288 		    type = ICMP_REDIRECT;
1289 		    code = ICMP_REDIRECT_HOST;
1290 #ifdef DIAGNOSTIC
1291 		    if (ipprintfs)
1292 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1293 #endif
1294 		}
1295 	}
1296 
1297 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1298 			  IP_FORWARDING, 0);
1299 	if (error)
1300 		ipstat.ips_cantforward++;
1301 	else {
1302 		ipstat.ips_forward++;
1303 		if (type)
1304 			ipstat.ips_redirectsent++;
1305 		else {
1306 			if (mcopy)
1307 				m_freem(mcopy);
1308 			return;
1309 		}
1310 	}
1311 	if (mcopy == NULL)
1312 		return;
1313 	destifp = NULL;
1314 
1315 	switch (error) {
1316 
1317 	case 0:				/* forwarded, but need redirect */
1318 		/* type, code set above */
1319 		break;
1320 
1321 	case ENETUNREACH:		/* shouldn't happen, checked above */
1322 	case EHOSTUNREACH:
1323 	case ENETDOWN:
1324 	case EHOSTDOWN:
1325 	default:
1326 		type = ICMP_UNREACH;
1327 		code = ICMP_UNREACH_HOST;
1328 		break;
1329 
1330 	case EMSGSIZE:
1331 		type = ICMP_UNREACH;
1332 		code = ICMP_UNREACH_NEEDFRAG;
1333 		if (ipforward_rt.ro_rt)
1334 			destifp = ipforward_rt.ro_rt->rt_ifp;
1335 		ipstat.ips_cantfrag++;
1336 		break;
1337 
1338 	case ENOBUFS:
1339 		type = ICMP_SOURCEQUENCH;
1340 		code = 0;
1341 		break;
1342 	}
1343 	icmp_error(mcopy, type, code, dest, destifp);
1344 }
1345 
1346 void
1347 ip_savecontrol(inp, mp, ip, m)
1348 	register struct inpcb *inp;
1349 	register struct mbuf **mp;
1350 	register struct ip *ip;
1351 	register struct mbuf *m;
1352 {
1353 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1354 		struct timeval tv;
1355 
1356 		microtime(&tv);
1357 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1358 			SCM_TIMESTAMP, SOL_SOCKET);
1359 		if (*mp)
1360 			mp = &(*mp)->m_next;
1361 	}
1362 	if (inp->inp_flags & INP_RECVDSTADDR) {
1363 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1364 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1365 		if (*mp)
1366 			mp = &(*mp)->m_next;
1367 	}
1368 #ifdef notyet
1369 	/* XXX
1370 	 * Moving these out of udp_input() made them even more broken
1371 	 * than they already were.
1372 	 */
1373 	/* options were tossed already */
1374 	if (inp->inp_flags & INP_RECVOPTS) {
1375 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1376 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1377 		if (*mp)
1378 			mp = &(*mp)->m_next;
1379 	}
1380 	/* ip_srcroute doesn't do what we want here, need to fix */
1381 	if (inp->inp_flags & INP_RECVRETOPTS) {
1382 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1383 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1384 		if (*mp)
1385 			mp = &(*mp)->m_next;
1386 	}
1387 #endif
1388 	if (inp->inp_flags & INP_RECVIF) {
1389 		struct sockaddr_dl sdl;
1390 
1391 		sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
1392 		sdl.sdl_family = AF_LINK;
1393 		sdl.sdl_index = m->m_pkthdr.rcvif ?
1394 			m->m_pkthdr.rcvif->if_index : 0;
1395 		sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
1396 		*mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
1397 			IP_RECVIF, IPPROTO_IP);
1398 		if (*mp)
1399 			mp = &(*mp)->m_next;
1400 	}
1401 }
1402 
1403 int
1404 ip_rsvp_init(struct socket *so)
1405 {
1406 	if (so->so_type != SOCK_RAW ||
1407 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1408 	  return EOPNOTSUPP;
1409 
1410 	if (ip_rsvpd != NULL)
1411 	  return EADDRINUSE;
1412 
1413 	ip_rsvpd = so;
1414 	/*
1415 	 * This may seem silly, but we need to be sure we don't over-increment
1416 	 * the RSVP counter, in case something slips up.
1417 	 */
1418 	if (!ip_rsvp_on) {
1419 		ip_rsvp_on = 1;
1420 		rsvp_on++;
1421 	}
1422 
1423 	return 0;
1424 }
1425 
1426 int
1427 ip_rsvp_done(void)
1428 {
1429 	ip_rsvpd = NULL;
1430 	/*
1431 	 * This may seem silly, but we need to be sure we don't over-decrement
1432 	 * the RSVP counter, in case something slips up.
1433 	 */
1434 	if (ip_rsvp_on) {
1435 		ip_rsvp_on = 0;
1436 		rsvp_on--;
1437 	}
1438 	return 0;
1439 }
1440