xref: /freebsd/sys/netinet/ip_input.c (revision 952d112864d8008aa87278a30a539d888a8493cd)
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 1.1.1.2 1997/04/03 10:39:25 darrenr 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 		int action;
338 
339 #ifdef IPDIVERT
340 		action = (*ip_fw_chk_ptr)(&ip, hlen,
341 				m->m_pkthdr.rcvif, ip_divert_ignore, &m);
342 		ip_divert_ignore = 0;
343 #else
344 		action = (*ip_fw_chk_ptr)(&ip, hlen, m->m_pkthdr.rcvif, 0, &m);
345 #endif
346 		if (action == -1)
347 			return;
348 		if (action != 0) {
349 #ifdef IPDIVERT
350 			frag_divert_port = action;
351 			goto ours;
352 #else
353 			goto bad;	/* ipfw said divert but we can't */
354 #endif
355 		}
356 	}
357 
358         if (ip_nat_ptr && !(*ip_nat_ptr)(&ip, &m, m->m_pkthdr.rcvif, IP_NAT_IN))
359 		return;
360 #endif
361 
362 	/*
363 	 * Process options and, if not destined for us,
364 	 * ship it on.  ip_dooptions returns 1 when an
365 	 * error was detected (causing an icmp message
366 	 * to be sent and the original packet to be freed).
367 	 */
368 	ip_nhops = 0;		/* for source routed packets */
369 	if (hlen > sizeof (struct ip) && ip_dooptions(m))
370 		return;
371 
372         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
373          * matter if it is destined to another node, or whether it is
374          * a multicast one, RSVP wants it! and prevents it from being forwarded
375          * anywhere else. Also checks if the rsvp daemon is running before
376 	 * grabbing the packet.
377          */
378 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
379 		goto ours;
380 
381 	/*
382 	 * Check our list of addresses, to see if the packet is for us.
383 	 */
384 	for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) {
385 #define	satosin(sa)	((struct sockaddr_in *)(sa))
386 
387 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
388 			goto ours;
389 		if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) {
390 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
391 			    ip->ip_dst.s_addr)
392 				goto ours;
393 			if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr)
394 				goto ours;
395 		}
396 	}
397 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
398 		struct in_multi *inm;
399 		if (ip_mrouter) {
400 			/*
401 			 * If we are acting as a multicast router, all
402 			 * incoming multicast packets are passed to the
403 			 * kernel-level multicast forwarding function.
404 			 * The packet is returned (relatively) intact; if
405 			 * ip_mforward() returns a non-zero value, the packet
406 			 * must be discarded, else it may be accepted below.
407 			 *
408 			 * (The IP ident field is put in the same byte order
409 			 * as expected when ip_mforward() is called from
410 			 * ip_output().)
411 			 */
412 			ip->ip_id = htons(ip->ip_id);
413 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
414 				ipstat.ips_cantforward++;
415 				m_freem(m);
416 				return;
417 			}
418 			ip->ip_id = ntohs(ip->ip_id);
419 
420 			/*
421 			 * The process-level routing demon needs to receive
422 			 * all multicast IGMP packets, whether or not this
423 			 * host belongs to their destination groups.
424 			 */
425 			if (ip->ip_p == IPPROTO_IGMP)
426 				goto ours;
427 			ipstat.ips_forward++;
428 		}
429 		/*
430 		 * See if we belong to the destination multicast group on the
431 		 * arrival interface.
432 		 */
433 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
434 		if (inm == NULL) {
435 			ipstat.ips_notmember++;
436 			m_freem(m);
437 			return;
438 		}
439 		goto ours;
440 	}
441 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
442 		goto ours;
443 	if (ip->ip_dst.s_addr == INADDR_ANY)
444 		goto ours;
445 
446 	/*
447 	 * Not for us; forward if possible and desirable.
448 	 */
449 	if (ipforwarding == 0) {
450 		ipstat.ips_cantforward++;
451 		m_freem(m);
452 	} else
453 		ip_forward(m, 0);
454 	return;
455 
456 ours:
457 
458 	/*
459 	 * If offset or IP_MF are set, must reassemble.
460 	 * Otherwise, nothing need be done.
461 	 * (We could look in the reassembly queue to see
462 	 * if the packet was previously fragmented,
463 	 * but it's not worth the time; just let them time out.)
464 	 */
465 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
466 		if (m->m_flags & M_EXT) {		/* XXX */
467 			if ((m = m_pullup(m, sizeof (struct ip))) == 0) {
468 				ipstat.ips_toosmall++;
469 #ifdef IPDIVERT
470 				frag_divert_port = 0;
471 #endif
472 				return;
473 			}
474 			ip = mtod(m, struct ip *);
475 		}
476 		/*
477 		 * Look for queue of fragments
478 		 * of this datagram.
479 		 */
480 		for (fp = ipq.next; fp != &ipq; fp = fp->next)
481 			if (ip->ip_id == fp->ipq_id &&
482 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
483 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
484 			    ip->ip_p == fp->ipq_p)
485 				goto found;
486 		fp = 0;
487 found:
488 
489 		/*
490 		 * Adjust ip_len to not reflect header,
491 		 * set ip_mff if more fragments are expected,
492 		 * convert offset of this to bytes.
493 		 */
494 		ip->ip_len -= hlen;
495 		((struct ipasfrag *)ip)->ipf_mff &= ~1;
496 		if (ip->ip_off & IP_MF)
497 			((struct ipasfrag *)ip)->ipf_mff |= 1;
498 		ip->ip_off <<= 3;
499 
500 		/*
501 		 * If datagram marked as having more fragments
502 		 * or if this is not the first fragment,
503 		 * attempt reassembly; if it succeeds, proceed.
504 		 */
505 		if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) {
506 			ipstat.ips_fragments++;
507 			ip = ip_reass((struct ipasfrag *)ip, fp);
508 			if (ip == 0)
509 				return;
510 			ipstat.ips_reassembled++;
511 			m = dtom(ip);
512 		} else
513 			if (fp)
514 				ip_freef(fp);
515 	} else
516 		ip->ip_len -= hlen;
517 
518 #ifdef IPDIVERT
519 	/*
520 	 * Divert packets here to the divert protocol if required
521 	 */
522 	if (frag_divert_port) {
523 		ip_divert_port = frag_divert_port;
524 		frag_divert_port = 0;
525 		(*inetsw[ip_protox[IPPROTO_DIVERT]].pr_input)(m, hlen);
526 		return;
527 	}
528 #endif
529 
530 	/*
531 	 * Switch out to protocol's input routine.
532 	 */
533 	ipstat.ips_delivered++;
534 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
535 	return;
536 bad:
537 	m_freem(m);
538 }
539 
540 /*
541  * IP software interrupt routine - to go away sometime soon
542  */
543 static void
544 ipintr(void)
545 {
546 	int s;
547 	struct mbuf *m;
548 
549 	while(1) {
550 		s = splimp();
551 		IF_DEQUEUE(&ipintrq, m);
552 		splx(s);
553 		if (m == 0)
554 			return;
555 		ip_input(m);
556 	}
557 }
558 
559 NETISR_SET(NETISR_IP, ipintr);
560 
561 /*
562  * Take incoming datagram fragment and try to
563  * reassemble it into whole datagram.  If a chain for
564  * reassembly of this datagram already exists, then it
565  * is given as fp; otherwise have to make a chain.
566  */
567 static struct ip *
568 ip_reass(ip, fp)
569 	register struct ipasfrag *ip;
570 	register struct ipq *fp;
571 {
572 	register struct mbuf *m = dtom(ip);
573 	register struct ipasfrag *q;
574 	struct mbuf *t;
575 	int hlen = ip->ip_hl << 2;
576 	int i, next;
577 
578 	/*
579 	 * Presence of header sizes in mbufs
580 	 * would confuse code below.
581 	 */
582 	m->m_data += hlen;
583 	m->m_len -= hlen;
584 
585 	/*
586 	 * If first fragment to arrive, create a reassembly queue.
587 	 */
588 	if (fp == 0) {
589 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
590 			goto dropfrag;
591 		fp = mtod(t, struct ipq *);
592 		insque(fp, &ipq);
593 		fp->ipq_ttl = IPFRAGTTL;
594 		fp->ipq_p = ip->ip_p;
595 		fp->ipq_id = ip->ip_id;
596 		fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp;
597 		fp->ipq_src = ((struct ip *)ip)->ip_src;
598 		fp->ipq_dst = ((struct ip *)ip)->ip_dst;
599 #ifdef IPDIVERT
600 		fp->ipq_divert = 0;
601 #endif
602 		q = (struct ipasfrag *)fp;
603 		goto insert;
604 	}
605 
606 	/*
607 	 * Find a segment which begins after this one does.
608 	 */
609 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next)
610 		if (q->ip_off > ip->ip_off)
611 			break;
612 
613 	/*
614 	 * If there is a preceding segment, it may provide some of
615 	 * our data already.  If so, drop the data from the incoming
616 	 * segment.  If it provides all of our data, drop us.
617 	 */
618 	if (q->ipf_prev != (struct ipasfrag *)fp) {
619 		i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off;
620 		if (i > 0) {
621 			if (i >= ip->ip_len)
622 				goto dropfrag;
623 			m_adj(dtom(ip), i);
624 			ip->ip_off += i;
625 			ip->ip_len -= i;
626 		}
627 	}
628 
629 	/*
630 	 * While we overlap succeeding segments trim them or,
631 	 * if they are completely covered, dequeue them.
632 	 */
633 	while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) {
634 		struct mbuf *m0;
635 
636 		i = (ip->ip_off + ip->ip_len) - q->ip_off;
637 		if (i < q->ip_len) {
638 			q->ip_len -= i;
639 			q->ip_off += i;
640 			m_adj(dtom(q), i);
641 			break;
642 		}
643 		m0 = dtom(q);
644 		q = q->ipf_next;
645 		ip_deq(q->ipf_prev);
646 		m_freem(m0);
647 	}
648 
649 insert:
650 
651 #ifdef IPDIVERT
652 	/*
653 	 * Any fragment diverting causes the whole packet to divert
654 	 */
655 	if (frag_divert_port != 0)
656 		fp->ipq_divert = frag_divert_port;
657 	frag_divert_port = 0;
658 #endif
659 
660 	/*
661 	 * Stick new segment in its place;
662 	 * check for complete reassembly.
663 	 */
664 	ip_enq(ip, q->ipf_prev);
665 	next = 0;
666 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) {
667 		if (q->ip_off != next)
668 			return (0);
669 		next += q->ip_len;
670 	}
671 	if (q->ipf_prev->ipf_mff & 1)
672 		return (0);
673 
674 	/*
675 	 * Reassembly is complete.  Make sure the packet is a sane size.
676 	 */
677 	if (next + (IP_VHL_HL(((struct ip *)fp->ipq_next)->ip_vhl) << 2)
678 							> IP_MAXPACKET) {
679 		ipstat.ips_toolong++;
680 		ip_freef(fp);
681 		return (0);
682 	}
683 
684 	/*
685 	 * Concatenate fragments.
686 	 */
687 	q = fp->ipq_next;
688 	m = dtom(q);
689 	t = m->m_next;
690 	m->m_next = 0;
691 	m_cat(m, t);
692 	q = q->ipf_next;
693 	while (q != (struct ipasfrag *)fp) {
694 		t = dtom(q);
695 		q = q->ipf_next;
696 		m_cat(m, t);
697 	}
698 
699 #ifdef IPDIVERT
700 	/*
701 	 * Record divert port for packet, if any
702 	 */
703 	frag_divert_port = fp->ipq_divert;
704 #endif
705 
706 	/*
707 	 * Create header for new ip packet by
708 	 * modifying header of first packet;
709 	 * dequeue and discard fragment reassembly header.
710 	 * Make header visible.
711 	 */
712 	ip = fp->ipq_next;
713 	ip->ip_len = next;
714 	ip->ipf_mff &= ~1;
715 	((struct ip *)ip)->ip_src = fp->ipq_src;
716 	((struct ip *)ip)->ip_dst = fp->ipq_dst;
717 	remque(fp);
718 	(void) m_free(dtom(fp));
719 	m = dtom(ip);
720 	m->m_len += (ip->ip_hl << 2);
721 	m->m_data -= (ip->ip_hl << 2);
722 	/* some debugging cruft by sklower, below, will go away soon */
723 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
724 		register int plen = 0;
725 		for (t = m; m; m = m->m_next)
726 			plen += m->m_len;
727 		t->m_pkthdr.len = plen;
728 	}
729 	return ((struct ip *)ip);
730 
731 dropfrag:
732 	ipstat.ips_fragdropped++;
733 	m_freem(m);
734 	return (0);
735 }
736 
737 /*
738  * Free a fragment reassembly header and all
739  * associated datagrams.
740  */
741 static void
742 ip_freef(fp)
743 	struct ipq *fp;
744 {
745 	register struct ipasfrag *q, *p;
746 
747 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) {
748 		p = q->ipf_next;
749 		ip_deq(q);
750 		m_freem(dtom(q));
751 	}
752 	remque(fp);
753 	(void) m_free(dtom(fp));
754 }
755 
756 /*
757  * Put an ip fragment on a reassembly chain.
758  * Like insque, but pointers in middle of structure.
759  */
760 static void
761 ip_enq(p, prev)
762 	register struct ipasfrag *p, *prev;
763 {
764 
765 	p->ipf_prev = prev;
766 	p->ipf_next = prev->ipf_next;
767 	prev->ipf_next->ipf_prev = p;
768 	prev->ipf_next = p;
769 }
770 
771 /*
772  * To ip_enq as remque is to insque.
773  */
774 static void
775 ip_deq(p)
776 	register struct ipasfrag *p;
777 {
778 
779 	p->ipf_prev->ipf_next = p->ipf_next;
780 	p->ipf_next->ipf_prev = p->ipf_prev;
781 }
782 
783 /*
784  * IP timer processing;
785  * if a timer expires on a reassembly
786  * queue, discard it.
787  */
788 void
789 ip_slowtimo()
790 {
791 	register struct ipq *fp;
792 	int s = splnet();
793 
794 	fp = ipq.next;
795 	if (fp == 0) {
796 		splx(s);
797 		return;
798 	}
799 	while (fp != &ipq) {
800 		--fp->ipq_ttl;
801 		fp = fp->next;
802 		if (fp->prev->ipq_ttl == 0) {
803 			ipstat.ips_fragtimeout++;
804 			ip_freef(fp->prev);
805 		}
806 	}
807 	splx(s);
808 }
809 
810 /*
811  * Drain off all datagram fragments.
812  */
813 void
814 ip_drain()
815 {
816 	while (ipq.next != &ipq) {
817 		ipstat.ips_fragdropped++;
818 		ip_freef(ipq.next);
819 	}
820 
821 	in_rtqdrain();
822 }
823 
824 /*
825  * Do option processing on a datagram,
826  * possibly discarding it if bad options are encountered,
827  * or forwarding it if source-routed.
828  * Returns 1 if packet has been forwarded/freed,
829  * 0 if the packet should be processed further.
830  */
831 static int
832 ip_dooptions(m)
833 	struct mbuf *m;
834 {
835 	register struct ip *ip = mtod(m, struct ip *);
836 	register u_char *cp;
837 	register struct ip_timestamp *ipt;
838 	register struct in_ifaddr *ia;
839 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
840 	struct in_addr *sin, dst;
841 	n_time ntime;
842 
843 	dst = ip->ip_dst;
844 	cp = (u_char *)(ip + 1);
845 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
846 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
847 		opt = cp[IPOPT_OPTVAL];
848 		if (opt == IPOPT_EOL)
849 			break;
850 		if (opt == IPOPT_NOP)
851 			optlen = 1;
852 		else {
853 			optlen = cp[IPOPT_OLEN];
854 			if (optlen <= 0 || optlen > cnt) {
855 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
856 				goto bad;
857 			}
858 		}
859 		switch (opt) {
860 
861 		default:
862 			break;
863 
864 		/*
865 		 * Source routing with record.
866 		 * Find interface with current destination address.
867 		 * If none on this machine then drop if strictly routed,
868 		 * or do nothing if loosely routed.
869 		 * Record interface address and bring up next address
870 		 * component.  If strictly routed make sure next
871 		 * address is on directly accessible net.
872 		 */
873 		case IPOPT_LSRR:
874 		case IPOPT_SSRR:
875 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
876 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
877 				goto bad;
878 			}
879 			ipaddr.sin_addr = ip->ip_dst;
880 			ia = (struct in_ifaddr *)
881 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
882 			if (ia == 0) {
883 				if (opt == IPOPT_SSRR) {
884 					type = ICMP_UNREACH;
885 					code = ICMP_UNREACH_SRCFAIL;
886 					goto bad;
887 				}
888 				/*
889 				 * Loose routing, and not at next destination
890 				 * yet; nothing to do except forward.
891 				 */
892 				break;
893 			}
894 			off--;			/* 0 origin */
895 			if (off > optlen - sizeof(struct in_addr)) {
896 				/*
897 				 * End of source route.  Should be for us.
898 				 */
899 				save_rte(cp, ip->ip_src);
900 				break;
901 			}
902 
903 			if (!ip_dosourceroute) {
904 				char buf[4*sizeof "123"];
905 				strcpy(buf, inet_ntoa(ip->ip_dst));
906 
907 				log(LOG_WARNING,
908 				    "attempted source route from %s to %s\n",
909 				    inet_ntoa(ip->ip_src), buf);
910 				type = ICMP_UNREACH;
911 				code = ICMP_UNREACH_SRCFAIL;
912 				goto bad;
913 			}
914 
915 			/*
916 			 * locate outgoing interface
917 			 */
918 			(void)memcpy(&ipaddr.sin_addr, cp + off,
919 			    sizeof(ipaddr.sin_addr));
920 
921 			if (opt == IPOPT_SSRR) {
922 #define	INA	struct in_ifaddr *
923 #define	SA	struct sockaddr *
924 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
925 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
926 			} else
927 				ia = ip_rtaddr(ipaddr.sin_addr);
928 			if (ia == 0) {
929 				type = ICMP_UNREACH;
930 				code = ICMP_UNREACH_SRCFAIL;
931 				goto bad;
932 			}
933 			ip->ip_dst = ipaddr.sin_addr;
934 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
935 			    sizeof(struct in_addr));
936 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
937 			/*
938 			 * Let ip_intr's mcast routing check handle mcast pkts
939 			 */
940 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
941 			break;
942 
943 		case IPOPT_RR:
944 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
945 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
946 				goto bad;
947 			}
948 			/*
949 			 * If no space remains, ignore.
950 			 */
951 			off--;			/* 0 origin */
952 			if (off > optlen - sizeof(struct in_addr))
953 				break;
954 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
955 			    sizeof(ipaddr.sin_addr));
956 			/*
957 			 * locate outgoing interface; if we're the destination,
958 			 * use the incoming interface (should be same).
959 			 */
960 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
961 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
962 				type = ICMP_UNREACH;
963 				code = ICMP_UNREACH_HOST;
964 				goto bad;
965 			}
966 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
967 			    sizeof(struct in_addr));
968 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
969 			break;
970 
971 		case IPOPT_TS:
972 			code = cp - (u_char *)ip;
973 			ipt = (struct ip_timestamp *)cp;
974 			if (ipt->ipt_len < 5)
975 				goto bad;
976 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) {
977 				if (++ipt->ipt_oflw == 0)
978 					goto bad;
979 				break;
980 			}
981 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
982 			switch (ipt->ipt_flg) {
983 
984 			case IPOPT_TS_TSONLY:
985 				break;
986 
987 			case IPOPT_TS_TSANDADDR:
988 				if (ipt->ipt_ptr + sizeof(n_time) +
989 				    sizeof(struct in_addr) > ipt->ipt_len)
990 					goto bad;
991 				ipaddr.sin_addr = dst;
992 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
993 							    m->m_pkthdr.rcvif);
994 				if (ia == 0)
995 					continue;
996 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
997 				    sizeof(struct in_addr));
998 				ipt->ipt_ptr += sizeof(struct in_addr);
999 				break;
1000 
1001 			case IPOPT_TS_PRESPEC:
1002 				if (ipt->ipt_ptr + sizeof(n_time) +
1003 				    sizeof(struct in_addr) > ipt->ipt_len)
1004 					goto bad;
1005 				(void)memcpy(&ipaddr.sin_addr, sin,
1006 				    sizeof(struct in_addr));
1007 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1008 					continue;
1009 				ipt->ipt_ptr += sizeof(struct in_addr);
1010 				break;
1011 
1012 			default:
1013 				goto bad;
1014 			}
1015 			ntime = iptime();
1016 			(void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
1017 			    sizeof(n_time));
1018 			ipt->ipt_ptr += sizeof(n_time);
1019 		}
1020 	}
1021 	if (forward) {
1022 		ip_forward(m, 1);
1023 		return (1);
1024 	}
1025 	return (0);
1026 bad:
1027 	ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2;   /* XXX icmp_error adds in hdr length */
1028 	icmp_error(m, type, code, 0, 0);
1029 	ipstat.ips_badoptions++;
1030 	return (1);
1031 }
1032 
1033 /*
1034  * Given address of next destination (final or next hop),
1035  * return internet address info of interface to be used to get there.
1036  */
1037 static struct in_ifaddr *
1038 ip_rtaddr(dst)
1039 	 struct in_addr dst;
1040 {
1041 	register struct sockaddr_in *sin;
1042 
1043 	sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1044 
1045 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1046 		if (ipforward_rt.ro_rt) {
1047 			RTFREE(ipforward_rt.ro_rt);
1048 			ipforward_rt.ro_rt = 0;
1049 		}
1050 		sin->sin_family = AF_INET;
1051 		sin->sin_len = sizeof(*sin);
1052 		sin->sin_addr = dst;
1053 
1054 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1055 	}
1056 	if (ipforward_rt.ro_rt == 0)
1057 		return ((struct in_ifaddr *)0);
1058 	return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
1059 }
1060 
1061 /*
1062  * Save incoming source route for use in replies,
1063  * to be picked up later by ip_srcroute if the receiver is interested.
1064  */
1065 void
1066 save_rte(option, dst)
1067 	u_char *option;
1068 	struct in_addr dst;
1069 {
1070 	unsigned olen;
1071 
1072 	olen = option[IPOPT_OLEN];
1073 #ifdef DIAGNOSTIC
1074 	if (ipprintfs)
1075 		printf("save_rte: olen %d\n", olen);
1076 #endif
1077 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1078 		return;
1079 	bcopy(option, ip_srcrt.srcopt, olen);
1080 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1081 	ip_srcrt.dst = dst;
1082 }
1083 
1084 /*
1085  * Retrieve incoming source route for use in replies,
1086  * in the same form used by setsockopt.
1087  * The first hop is placed before the options, will be removed later.
1088  */
1089 struct mbuf *
1090 ip_srcroute()
1091 {
1092 	register struct in_addr *p, *q;
1093 	register struct mbuf *m;
1094 
1095 	if (ip_nhops == 0)
1096 		return ((struct mbuf *)0);
1097 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1098 	if (m == 0)
1099 		return ((struct mbuf *)0);
1100 
1101 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1102 
1103 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1104 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1105 	    OPTSIZ;
1106 #ifdef DIAGNOSTIC
1107 	if (ipprintfs)
1108 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1109 #endif
1110 
1111 	/*
1112 	 * First save first hop for return route
1113 	 */
1114 	p = &ip_srcrt.route[ip_nhops - 1];
1115 	*(mtod(m, struct in_addr *)) = *p--;
1116 #ifdef DIAGNOSTIC
1117 	if (ipprintfs)
1118 		printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr));
1119 #endif
1120 
1121 	/*
1122 	 * Copy option fields and padding (nop) to mbuf.
1123 	 */
1124 	ip_srcrt.nop = IPOPT_NOP;
1125 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1126 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1127 	    &ip_srcrt.nop, OPTSIZ);
1128 	q = (struct in_addr *)(mtod(m, caddr_t) +
1129 	    sizeof(struct in_addr) + OPTSIZ);
1130 #undef OPTSIZ
1131 	/*
1132 	 * Record return path as an IP source route,
1133 	 * reversing the path (pointers are now aligned).
1134 	 */
1135 	while (p >= ip_srcrt.route) {
1136 #ifdef DIAGNOSTIC
1137 		if (ipprintfs)
1138 			printf(" %lx", ntohl(q->s_addr));
1139 #endif
1140 		*q++ = *p--;
1141 	}
1142 	/*
1143 	 * Last hop goes to final destination.
1144 	 */
1145 	*q = ip_srcrt.dst;
1146 #ifdef DIAGNOSTIC
1147 	if (ipprintfs)
1148 		printf(" %lx\n", ntohl(q->s_addr));
1149 #endif
1150 	return (m);
1151 }
1152 
1153 /*
1154  * Strip out IP options, at higher
1155  * level protocol in the kernel.
1156  * Second argument is buffer to which options
1157  * will be moved, and return value is their length.
1158  * XXX should be deleted; last arg currently ignored.
1159  */
1160 void
1161 ip_stripoptions(m, mopt)
1162 	register struct mbuf *m;
1163 	struct mbuf *mopt;
1164 {
1165 	register int i;
1166 	struct ip *ip = mtod(m, struct ip *);
1167 	register caddr_t opts;
1168 	int olen;
1169 
1170 	olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1171 	opts = (caddr_t)(ip + 1);
1172 	i = m->m_len - (sizeof (struct ip) + olen);
1173 	bcopy(opts + olen, opts, (unsigned)i);
1174 	m->m_len -= olen;
1175 	if (m->m_flags & M_PKTHDR)
1176 		m->m_pkthdr.len -= olen;
1177 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1178 }
1179 
1180 u_char inetctlerrmap[PRC_NCMDS] = {
1181 	0,		0,		0,		0,
1182 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1183 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1184 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1185 	0,		0,		0,		0,
1186 	ENOPROTOOPT
1187 };
1188 
1189 /*
1190  * Forward a packet.  If some error occurs return the sender
1191  * an icmp packet.  Note we can't always generate a meaningful
1192  * icmp message because icmp doesn't have a large enough repertoire
1193  * of codes and types.
1194  *
1195  * If not forwarding, just drop the packet.  This could be confusing
1196  * if ipforwarding was zero but some routing protocol was advancing
1197  * us as a gateway to somewhere.  However, we must let the routing
1198  * protocol deal with that.
1199  *
1200  * The srcrt parameter indicates whether the packet is being forwarded
1201  * via a source route.
1202  */
1203 static void
1204 ip_forward(m, srcrt)
1205 	struct mbuf *m;
1206 	int srcrt;
1207 {
1208 	register struct ip *ip = mtod(m, struct ip *);
1209 	register struct sockaddr_in *sin;
1210 	register struct rtentry *rt;
1211 	int error, type = 0, code = 0;
1212 	struct mbuf *mcopy;
1213 	n_long dest;
1214 	struct ifnet *destifp;
1215 
1216 	dest = 0;
1217 #ifdef DIAGNOSTIC
1218 	if (ipprintfs)
1219 		printf("forward: src %lx dst %lx ttl %x\n",
1220 			ip->ip_src.s_addr, ip->ip_dst.s_addr, ip->ip_ttl);
1221 #endif
1222 
1223 
1224 	if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) {
1225 		ipstat.ips_cantforward++;
1226 		m_freem(m);
1227 		return;
1228 	}
1229 	HTONS(ip->ip_id);
1230 	if (ip->ip_ttl <= IPTTLDEC) {
1231 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1232 		return;
1233 	}
1234 	ip->ip_ttl -= IPTTLDEC;
1235 
1236 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1237 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1238 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1239 		if (ipforward_rt.ro_rt) {
1240 			RTFREE(ipforward_rt.ro_rt);
1241 			ipforward_rt.ro_rt = 0;
1242 		}
1243 		sin->sin_family = AF_INET;
1244 		sin->sin_len = sizeof(*sin);
1245 		sin->sin_addr = ip->ip_dst;
1246 
1247 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1248 		if (ipforward_rt.ro_rt == 0) {
1249 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1250 			return;
1251 		}
1252 		rt = ipforward_rt.ro_rt;
1253 	}
1254 
1255 	/*
1256 	 * Save at most 64 bytes of the packet in case
1257 	 * we need to generate an ICMP message to the src.
1258 	 */
1259 	mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64));
1260 
1261 	/*
1262 	 * If forwarding packet using same interface that it came in on,
1263 	 * perhaps should send a redirect to sender to shortcut a hop.
1264 	 * Only send redirect if source is sending directly to us,
1265 	 * and if packet was not source routed (or has any options).
1266 	 * Also, don't send redirect if forwarding using a default route
1267 	 * or a route modified by a redirect.
1268 	 */
1269 #define	satosin(sa)	((struct sockaddr_in *)(sa))
1270 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1271 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1272 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1273 	    ipsendredirects && !srcrt) {
1274 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1275 		u_long src = ntohl(ip->ip_src.s_addr);
1276 
1277 		if (RTA(rt) &&
1278 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1279 		    if (rt->rt_flags & RTF_GATEWAY)
1280 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1281 		    else
1282 			dest = ip->ip_dst.s_addr;
1283 		    /* Router requirements says to only send host redirects */
1284 		    type = ICMP_REDIRECT;
1285 		    code = ICMP_REDIRECT_HOST;
1286 #ifdef DIAGNOSTIC
1287 		    if (ipprintfs)
1288 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1289 #endif
1290 		}
1291 	}
1292 
1293 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1294 			  IP_FORWARDING, 0);
1295 	if (error)
1296 		ipstat.ips_cantforward++;
1297 	else {
1298 		ipstat.ips_forward++;
1299 		if (type)
1300 			ipstat.ips_redirectsent++;
1301 		else {
1302 			if (mcopy)
1303 				m_freem(mcopy);
1304 			return;
1305 		}
1306 	}
1307 	if (mcopy == NULL)
1308 		return;
1309 	destifp = NULL;
1310 
1311 	switch (error) {
1312 
1313 	case 0:				/* forwarded, but need redirect */
1314 		/* type, code set above */
1315 		break;
1316 
1317 	case ENETUNREACH:		/* shouldn't happen, checked above */
1318 	case EHOSTUNREACH:
1319 	case ENETDOWN:
1320 	case EHOSTDOWN:
1321 	default:
1322 		type = ICMP_UNREACH;
1323 		code = ICMP_UNREACH_HOST;
1324 		break;
1325 
1326 	case EMSGSIZE:
1327 		type = ICMP_UNREACH;
1328 		code = ICMP_UNREACH_NEEDFRAG;
1329 		if (ipforward_rt.ro_rt)
1330 			destifp = ipforward_rt.ro_rt->rt_ifp;
1331 		ipstat.ips_cantfrag++;
1332 		break;
1333 
1334 	case ENOBUFS:
1335 		type = ICMP_SOURCEQUENCH;
1336 		code = 0;
1337 		break;
1338 	}
1339 	icmp_error(mcopy, type, code, dest, destifp);
1340 }
1341 
1342 void
1343 ip_savecontrol(inp, mp, ip, m)
1344 	register struct inpcb *inp;
1345 	register struct mbuf **mp;
1346 	register struct ip *ip;
1347 	register struct mbuf *m;
1348 {
1349 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1350 		struct timeval tv;
1351 
1352 		microtime(&tv);
1353 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1354 			SCM_TIMESTAMP, SOL_SOCKET);
1355 		if (*mp)
1356 			mp = &(*mp)->m_next;
1357 	}
1358 	if (inp->inp_flags & INP_RECVDSTADDR) {
1359 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1360 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1361 		if (*mp)
1362 			mp = &(*mp)->m_next;
1363 	}
1364 #ifdef notyet
1365 	/* XXX
1366 	 * Moving these out of udp_input() made them even more broken
1367 	 * than they already were.
1368 	 */
1369 	/* options were tossed already */
1370 	if (inp->inp_flags & INP_RECVOPTS) {
1371 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1372 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1373 		if (*mp)
1374 			mp = &(*mp)->m_next;
1375 	}
1376 	/* ip_srcroute doesn't do what we want here, need to fix */
1377 	if (inp->inp_flags & INP_RECVRETOPTS) {
1378 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1379 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1380 		if (*mp)
1381 			mp = &(*mp)->m_next;
1382 	}
1383 #endif
1384 	if (inp->inp_flags & INP_RECVIF) {
1385 		struct sockaddr_dl sdl;
1386 
1387 		sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
1388 		sdl.sdl_family = AF_LINK;
1389 		sdl.sdl_index = m->m_pkthdr.rcvif ?
1390 			m->m_pkthdr.rcvif->if_index : 0;
1391 		sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
1392 		*mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
1393 			IP_RECVIF, IPPROTO_IP);
1394 		if (*mp)
1395 			mp = &(*mp)->m_next;
1396 	}
1397 }
1398 
1399 int
1400 ip_rsvp_init(struct socket *so)
1401 {
1402 	if (so->so_type != SOCK_RAW ||
1403 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1404 	  return EOPNOTSUPP;
1405 
1406 	if (ip_rsvpd != NULL)
1407 	  return EADDRINUSE;
1408 
1409 	ip_rsvpd = so;
1410 	/*
1411 	 * This may seem silly, but we need to be sure we don't over-increment
1412 	 * the RSVP counter, in case something slips up.
1413 	 */
1414 	if (!ip_rsvp_on) {
1415 		ip_rsvp_on = 1;
1416 		rsvp_on++;
1417 	}
1418 
1419 	return 0;
1420 }
1421 
1422 int
1423 ip_rsvp_done(void)
1424 {
1425 	ip_rsvpd = NULL;
1426 	/*
1427 	 * This may seem silly, but we need to be sure we don't over-decrement
1428 	 * the RSVP counter, in case something slips up.
1429 	 */
1430 	if (ip_rsvp_on) {
1431 		ip_rsvp_on = 0;
1432 		rsvp_on--;
1433 	}
1434 	return 0;
1435 }
1436