xref: /freebsd/sys/netinet/ip_input.c (revision 0de89efe5c443f213c7ea28773ef2dc6cf3af2ed)
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 1.65 1997/09/15 23:07:01 ache 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 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD,
114 	&ipstat, ipstat, "");
115 
116 /* Packet reassembly stuff */
117 #define IPREASS_NHASH_LOG2      6
118 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
119 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
120 #define IPREASS_HASH(x,y) \
121 	((((x) & 0xF | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
122 
123 static struct ipq ipq[IPREASS_NHASH];
124 static int    nipq = 0;         /* total # of reass queues */
125 static int    maxnipq;
126 
127 #ifdef IPCTL_DEFMTU
128 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
129 	&ip_mtu, 0, "");
130 #endif
131 
132 #if !defined(COMPAT_IPFW) || COMPAT_IPFW == 1
133 #undef COMPAT_IPFW
134 #define COMPAT_IPFW 1
135 #else
136 #undef COMPAT_IPFW
137 #endif
138 
139 #ifdef COMPAT_IPFW
140 /* Firewall hooks */
141 ip_fw_chk_t *ip_fw_chk_ptr;
142 ip_fw_ctl_t *ip_fw_ctl_ptr;
143 
144 /* IP Network Address Translation (NAT) hooks */
145 ip_nat_t *ip_nat_ptr;
146 ip_nat_ctl_t *ip_nat_ctl_ptr;
147 #endif
148 
149 #if defined(IPFILTER_LKM) || defined(IPFILTER)
150 int fr_check __P((struct ip *, int, struct ifnet *, int, struct mbuf **));
151 int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL;
152 #endif
153 
154 
155 /*
156  * We need to save the IP options in case a protocol wants to respond
157  * to an incoming packet over the same route if the packet got here
158  * using IP source routing.  This allows connection establishment and
159  * maintenance when the remote end is on a network that is not known
160  * to us.
161  */
162 static int	ip_nhops = 0;
163 static	struct ip_srcrt {
164 	struct	in_addr dst;			/* final destination */
165 	char	nop;				/* one NOP to align */
166 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
167 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
168 } ip_srcrt;
169 
170 #ifdef IPDIVERT
171 /*
172  * Shared variable between ip_input() and ip_reass() to communicate
173  * about which packets, once assembled from fragments, get diverted,
174  * and to which port.
175  */
176 static u_short	frag_divert_port;
177 #endif
178 
179 static void save_rte __P((u_char *, struct in_addr));
180 static void	 ip_deq __P((struct ipasfrag *));
181 static int	 ip_dooptions __P((struct mbuf *));
182 static void	 ip_enq __P((struct ipasfrag *, struct ipasfrag *));
183 static void	 ip_forward __P((struct mbuf *, int));
184 static void	 ip_freef __P((struct ipq *));
185 static struct ip *
186 	 ip_reass __P((struct ipasfrag *, struct ipq *, struct ipq *));
187 static struct in_ifaddr *
188 	 ip_rtaddr __P((struct in_addr));
189 static void	ipintr __P((void));
190 /*
191  * IP initialization: fill in IP protocol switch table.
192  * All protocols not implemented in kernel go to raw IP protocol handler.
193  */
194 void
195 ip_init()
196 {
197 	register struct protosw *pr;
198 	register int i;
199 
200 	TAILQ_INIT(&in_ifaddrhead);
201 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
202 	if (pr == 0)
203 		panic("ip_init");
204 	for (i = 0; i < IPPROTO_MAX; i++)
205 		ip_protox[i] = pr - inetsw;
206 	for (pr = inetdomain.dom_protosw;
207 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
208 		if (pr->pr_domain->dom_family == PF_INET &&
209 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
210 			ip_protox[pr->pr_protocol] = pr - inetsw;
211 
212 	for (i = 0; i < IPREASS_NHASH; i++)
213 	    ipq[i].next = ipq[i].prev = &ipq[i];
214 
215 	maxnipq = nmbclusters/4;
216 
217 	ip_id = time.tv_sec & 0xffff;
218 	ipintrq.ifq_maxlen = ipqmaxlen;
219 #ifdef IPFIREWALL
220 	ip_fw_init();
221 #endif
222 #ifdef IPNAT
223         ip_nat_init();
224 #endif
225 
226 }
227 
228 static struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
229 static struct	route ipforward_rt;
230 
231 /*
232  * Ip input routine.  Checksum and byte swap header.  If fragmented
233  * try to reassemble.  Process options.  Pass to next level.
234  */
235 void
236 ip_input(struct mbuf *m)
237 {
238 	struct ip *ip;
239 	struct ipq *fp;
240 	struct in_ifaddr *ia;
241 	int    i, hlen;
242 	u_short sum;
243 
244 #ifdef	DIAGNOSTIC
245 	if ((m->m_flags & M_PKTHDR) == 0)
246 		panic("ip_input no HDR");
247 #endif
248 	/*
249 	 * If no IP addresses have been set yet but the interfaces
250 	 * are receiving, can't do anything with incoming packets yet.
251 	 * XXX This is broken! We should be able to receive broadcasts
252 	 * and multicasts even without any local addresses configured.
253 	 */
254 	if (TAILQ_EMPTY(&in_ifaddrhead))
255 		goto bad;
256 	ipstat.ips_total++;
257 
258 	if (m->m_pkthdr.len < sizeof(struct ip))
259 		goto tooshort;
260 
261 #ifdef	DIAGNOSTIC
262 	if (m->m_len < sizeof(struct ip))
263 		panic("ipintr mbuf too short");
264 #endif
265 
266 	if (m->m_len < sizeof (struct ip) &&
267 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
268 		ipstat.ips_toosmall++;
269 		return;
270 	}
271 	ip = mtod(m, struct ip *);
272 
273 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
274 		ipstat.ips_badvers++;
275 		goto bad;
276 	}
277 
278 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
279 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
280 		ipstat.ips_badhlen++;
281 		goto bad;
282 	}
283 	if (hlen > m->m_len) {
284 		if ((m = m_pullup(m, hlen)) == 0) {
285 			ipstat.ips_badhlen++;
286 			return;
287 		}
288 		ip = mtod(m, struct ip *);
289 	}
290 	if (hlen == sizeof(struct ip)) {
291 		sum = in_cksum_hdr(ip);
292 	} else {
293 		sum = in_cksum(m, hlen);
294 	}
295 	if (sum) {
296 		ipstat.ips_badsum++;
297 		goto bad;
298 	}
299 
300 	/*
301 	 * Convert fields to host representation.
302 	 */
303 	NTOHS(ip->ip_len);
304 	if (ip->ip_len < hlen) {
305 		ipstat.ips_badlen++;
306 		goto bad;
307 	}
308 	NTOHS(ip->ip_id);
309 	NTOHS(ip->ip_off);
310 
311 	/*
312 	 * Check that the amount of data in the buffers
313 	 * is as at least much as the IP header would have us expect.
314 	 * Trim mbufs if longer than we expect.
315 	 * Drop packet if shorter than we expect.
316 	 */
317 	if (m->m_pkthdr.len < ip->ip_len) {
318 tooshort:
319 		ipstat.ips_tooshort++;
320 		goto bad;
321 	}
322 	if (m->m_pkthdr.len > ip->ip_len) {
323 		if (m->m_len == m->m_pkthdr.len) {
324 			m->m_len = ip->ip_len;
325 			m->m_pkthdr.len = ip->ip_len;
326 		} else
327 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
328 	}
329 	/*
330 	 * IpHack's section.
331 	 * Right now when no processing on packet has done
332 	 * and it is still fresh out of network we do our black
333 	 * deals with it.
334 	 * - Firewall: deny/allow/divert
335 	 * - Xlate: translate packet's addr/port (NAT).
336 	 * - Wrap: fake packet's addr/port <unimpl.>
337 	 * - Encapsulate: put it in another IP and send out. <unimp.>
338  	 */
339 #if defined(IPFILTER) || defined(IPFILTER_LKM)
340 	/*
341 	 * Check if we want to allow this packet to be processed.
342 	 * Consider it to be bad if not.
343 	 */
344 	if (fr_check) {
345 		struct	mbuf	*m1 = m;
346 
347 		if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1)
348 			return;
349 		ip = mtod(m = m1, struct ip *);
350 	}
351 #endif
352 #ifdef COMPAT_IPFW
353 	if (ip_fw_chk_ptr) {
354 #ifdef IPDIVERT
355 		u_short port;
356 
357 		port = (*ip_fw_chk_ptr)(&ip, hlen, NULL, ip_divert_ignore, &m);
358 		ip_divert_ignore = 0;
359 		if (port) {			/* Divert packet */
360 			frag_divert_port = port;
361 			goto ours;
362 		}
363 #else
364 		/* If ipfw says divert, we have to just drop packet */
365 		if ((*ip_fw_chk_ptr)(&ip, hlen, NULL, 0, &m)) {
366 			m_freem(m);
367 			m = NULL;
368 		}
369 #endif
370 		if (!m)
371 			return;
372 	}
373 
374         if (ip_nat_ptr && !(*ip_nat_ptr)(&ip, &m, m->m_pkthdr.rcvif, IP_NAT_IN))
375 		return;
376 #endif
377 
378 	/*
379 	 * Process options and, if not destined for us,
380 	 * ship it on.  ip_dooptions returns 1 when an
381 	 * error was detected (causing an icmp message
382 	 * to be sent and the original packet to be freed).
383 	 */
384 	ip_nhops = 0;		/* for source routed packets */
385 	if (hlen > sizeof (struct ip) && ip_dooptions(m))
386 		return;
387 
388         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
389          * matter if it is destined to another node, or whether it is
390          * a multicast one, RSVP wants it! and prevents it from being forwarded
391          * anywhere else. Also checks if the rsvp daemon is running before
392 	 * grabbing the packet.
393          */
394 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
395 		goto ours;
396 
397 	/*
398 	 * Check our list of addresses, to see if the packet is for us.
399 	 */
400 	for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) {
401 #define	satosin(sa)	((struct sockaddr_in *)(sa))
402 
403 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
404 			goto ours;
405 #ifdef BOOTP_COMPAT
406 		if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
407 			goto ours;
408 #endif
409 		if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) {
410 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
411 			    ip->ip_dst.s_addr)
412 				goto ours;
413 			if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr)
414 				goto ours;
415 		}
416 	}
417 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
418 		struct in_multi *inm;
419 		if (ip_mrouter) {
420 			/*
421 			 * If we are acting as a multicast router, all
422 			 * incoming multicast packets are passed to the
423 			 * kernel-level multicast forwarding function.
424 			 * The packet is returned (relatively) intact; if
425 			 * ip_mforward() returns a non-zero value, the packet
426 			 * must be discarded, else it may be accepted below.
427 			 *
428 			 * (The IP ident field is put in the same byte order
429 			 * as expected when ip_mforward() is called from
430 			 * ip_output().)
431 			 */
432 			ip->ip_id = htons(ip->ip_id);
433 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
434 				ipstat.ips_cantforward++;
435 				m_freem(m);
436 				return;
437 			}
438 			ip->ip_id = ntohs(ip->ip_id);
439 
440 			/*
441 			 * The process-level routing demon needs to receive
442 			 * all multicast IGMP packets, whether or not this
443 			 * host belongs to their destination groups.
444 			 */
445 			if (ip->ip_p == IPPROTO_IGMP)
446 				goto ours;
447 			ipstat.ips_forward++;
448 		}
449 		/*
450 		 * See if we belong to the destination multicast group on the
451 		 * arrival interface.
452 		 */
453 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
454 		if (inm == NULL) {
455 			ipstat.ips_notmember++;
456 			m_freem(m);
457 			return;
458 		}
459 		goto ours;
460 	}
461 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
462 		goto ours;
463 	if (ip->ip_dst.s_addr == INADDR_ANY)
464 		goto ours;
465 
466 	/*
467 	 * Not for us; forward if possible and desirable.
468 	 */
469 	if (ipforwarding == 0) {
470 		ipstat.ips_cantforward++;
471 		m_freem(m);
472 	} else
473 		ip_forward(m, 0);
474 	return;
475 
476 ours:
477 
478 	/*
479 	 * If offset or IP_MF are set, must reassemble.
480 	 * Otherwise, nothing need be done.
481 	 * (We could look in the reassembly queue to see
482 	 * if the packet was previously fragmented,
483 	 * but it's not worth the time; just let them time out.)
484 	 */
485 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
486 		if (m->m_flags & M_EXT) {		/* XXX */
487 			if ((m = m_pullup(m, sizeof (struct ip))) == 0) {
488 				ipstat.ips_toosmall++;
489 #ifdef IPDIVERT
490 				frag_divert_port = 0;
491 #endif
492 				return;
493 			}
494 			ip = mtod(m, struct ip *);
495 		}
496 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
497 		/*
498 		 * Look for queue of fragments
499 		 * of this datagram.
500 		 */
501 		for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
502 			if (ip->ip_id == fp->ipq_id &&
503 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
504 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
505 			    ip->ip_p == fp->ipq_p)
506 				goto found;
507 
508 		fp = 0;
509 
510 		/* check if there's a place for the new queue */
511 		if (nipq > maxnipq) {
512 		    /*
513 		     * drop something from the tail of the current queue
514 		     * before proceeding further
515 		     */
516 		    if (ipq[sum].prev == &ipq[sum]) {   /* gak */
517 			for (i = 0; i < IPREASS_NHASH; i++) {
518 			    if (ipq[i].prev != &ipq[i]) {
519 				ip_freef(ipq[i].prev);
520 				break;
521 			    }
522 			}
523 		    } else
524 			ip_freef(ipq[sum].prev);
525 		}
526 found:
527 		/*
528 		 * Adjust ip_len to not reflect header,
529 		 * set ip_mff if more fragments are expected,
530 		 * convert offset of this to bytes.
531 		 */
532 		ip->ip_len -= hlen;
533 		((struct ipasfrag *)ip)->ipf_mff &= ~1;
534 		if (ip->ip_off & IP_MF)
535 			((struct ipasfrag *)ip)->ipf_mff |= 1;
536 		ip->ip_off <<= 3;
537 
538 		/*
539 		 * If datagram marked as having more fragments
540 		 * or if this is not the first fragment,
541 		 * attempt reassembly; if it succeeds, proceed.
542 		 */
543 		if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) {
544 			ipstat.ips_fragments++;
545 			ip = ip_reass((struct ipasfrag *)ip, fp, &ipq[sum]);
546 			if (ip == 0)
547 				return;
548 			ipstat.ips_reassembled++;
549 			m = dtom(ip);
550 #ifdef IPDIVERT
551 			if (frag_divert_port) {
552 				ip->ip_len += hlen;
553 				HTONS(ip->ip_len);
554 				HTONS(ip->ip_off);
555 				HTONS(ip->ip_id);
556 				ip->ip_sum = 0;
557 				ip->ip_sum = in_cksum_hdr(ip);
558 				NTOHS(ip->ip_id);
559 				NTOHS(ip->ip_off);
560 				NTOHS(ip->ip_len);
561 				ip->ip_len -= hlen;
562 			}
563 #endif
564 		} else
565 			if (fp)
566 				ip_freef(fp);
567 	} else
568 		ip->ip_len -= hlen;
569 
570 #ifdef IPDIVERT
571 	/*
572 	 * Divert reassembled packets to the divert protocol if required
573 	 */
574 	if (frag_divert_port) {
575 		ipstat.ips_delivered++;
576 		ip_divert_port = frag_divert_port;
577 		frag_divert_port = 0;
578 		(*inetsw[ip_protox[IPPROTO_DIVERT]].pr_input)(m, hlen);
579 		return;
580 	}
581 #endif
582 
583 	/*
584 	 * Switch out to protocol's input routine.
585 	 */
586 	ipstat.ips_delivered++;
587 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
588 	return;
589 bad:
590 	m_freem(m);
591 }
592 
593 /*
594  * IP software interrupt routine - to go away sometime soon
595  */
596 static void
597 ipintr(void)
598 {
599 	int s;
600 	struct mbuf *m;
601 
602 	while(1) {
603 		s = splimp();
604 		IF_DEQUEUE(&ipintrq, m);
605 		splx(s);
606 		if (m == 0)
607 			return;
608 		ip_input(m);
609 	}
610 }
611 
612 NETISR_SET(NETISR_IP, ipintr);
613 
614 /*
615  * Take incoming datagram fragment and try to
616  * reassemble it into whole datagram.  If a chain for
617  * reassembly of this datagram already exists, then it
618  * is given as fp; otherwise have to make a chain.
619  */
620 static struct ip *
621 ip_reass(ip, fp, where)
622 	register struct ipasfrag *ip;
623 	register struct ipq *fp;
624 	struct   ipq    *where;
625 {
626 	register struct mbuf *m = dtom(ip);
627 	register struct ipasfrag *q;
628 	struct mbuf *t;
629 	int hlen = ip->ip_hl << 2;
630 	int i, next;
631 
632 	/*
633 	 * Presence of header sizes in mbufs
634 	 * would confuse code below.
635 	 */
636 	m->m_data += hlen;
637 	m->m_len -= hlen;
638 
639 	/*
640 	 * If first fragment to arrive, create a reassembly queue.
641 	 */
642 	if (fp == 0) {
643 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
644 			goto dropfrag;
645 		fp = mtod(t, struct ipq *);
646 		insque(fp, where);
647 		nipq++;
648 		fp->ipq_ttl = IPFRAGTTL;
649 		fp->ipq_p = ip->ip_p;
650 		fp->ipq_id = ip->ip_id;
651 		fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp;
652 		fp->ipq_src = ((struct ip *)ip)->ip_src;
653 		fp->ipq_dst = ((struct ip *)ip)->ip_dst;
654 #ifdef IPDIVERT
655 		fp->ipq_divert = 0;
656 #endif
657 		q = (struct ipasfrag *)fp;
658 		goto insert;
659 	}
660 
661 	/*
662 	 * Find a segment which begins after this one does.
663 	 */
664 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next)
665 		if (q->ip_off > ip->ip_off)
666 			break;
667 
668 	/*
669 	 * If there is a preceding segment, it may provide some of
670 	 * our data already.  If so, drop the data from the incoming
671 	 * segment.  If it provides all of our data, drop us.
672 	 */
673 	if (q->ipf_prev != (struct ipasfrag *)fp) {
674 		i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off;
675 		if (i > 0) {
676 			if (i >= ip->ip_len)
677 				goto dropfrag;
678 			m_adj(dtom(ip), i);
679 			ip->ip_off += i;
680 			ip->ip_len -= i;
681 		}
682 	}
683 
684 	/*
685 	 * While we overlap succeeding segments trim them or,
686 	 * if they are completely covered, dequeue them.
687 	 */
688 	while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) {
689 		struct mbuf *m0;
690 
691 		i = (ip->ip_off + ip->ip_len) - q->ip_off;
692 		if (i < q->ip_len) {
693 			q->ip_len -= i;
694 			q->ip_off += i;
695 			m_adj(dtom(q), i);
696 			break;
697 		}
698 		m0 = dtom(q);
699 		q = q->ipf_next;
700 		ip_deq(q->ipf_prev);
701 		m_freem(m0);
702 	}
703 
704 insert:
705 
706 #ifdef IPDIVERT
707 	/*
708 	 * Any fragment diverting causes the whole packet to divert
709 	 */
710 	if (frag_divert_port != 0)
711 		fp->ipq_divert = frag_divert_port;
712 	frag_divert_port = 0;
713 #endif
714 
715 	/*
716 	 * Stick new segment in its place;
717 	 * check for complete reassembly.
718 	 */
719 	ip_enq(ip, q->ipf_prev);
720 	next = 0;
721 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) {
722 		if (q->ip_off != next)
723 			return (0);
724 		next += q->ip_len;
725 	}
726 	if (q->ipf_prev->ipf_mff & 1)
727 		return (0);
728 
729 	/*
730 	 * Reassembly is complete.  Make sure the packet is a sane size.
731 	 */
732 	if (next + (IP_VHL_HL(((struct ip *)fp->ipq_next)->ip_vhl) << 2)
733 							> IP_MAXPACKET) {
734 		ipstat.ips_toolong++;
735 		ip_freef(fp);
736 		return (0);
737 	}
738 
739 	/*
740 	 * Concatenate fragments.
741 	 */
742 	q = fp->ipq_next;
743 	m = dtom(q);
744 	t = m->m_next;
745 	m->m_next = 0;
746 	m_cat(m, t);
747 	q = q->ipf_next;
748 	while (q != (struct ipasfrag *)fp) {
749 		t = dtom(q);
750 		q = q->ipf_next;
751 		m_cat(m, t);
752 	}
753 
754 #ifdef IPDIVERT
755 	/*
756 	 * Record divert port for packet, if any
757 	 */
758 	frag_divert_port = fp->ipq_divert;
759 #endif
760 
761 	/*
762 	 * Create header for new ip packet by
763 	 * modifying header of first packet;
764 	 * dequeue and discard fragment reassembly header.
765 	 * Make header visible.
766 	 */
767 	ip = fp->ipq_next;
768 	ip->ip_len = next;
769 	ip->ipf_mff &= ~1;
770 	((struct ip *)ip)->ip_src = fp->ipq_src;
771 	((struct ip *)ip)->ip_dst = fp->ipq_dst;
772 	remque(fp);
773 	nipq--;
774 	(void) m_free(dtom(fp));
775 	m = dtom(ip);
776 	m->m_len += (ip->ip_hl << 2);
777 	m->m_data -= (ip->ip_hl << 2);
778 	/* some debugging cruft by sklower, below, will go away soon */
779 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
780 		register int plen = 0;
781 		for (t = m; m; m = m->m_next)
782 			plen += m->m_len;
783 		t->m_pkthdr.len = plen;
784 	}
785 	return ((struct ip *)ip);
786 
787 dropfrag:
788 	ipstat.ips_fragdropped++;
789 	m_freem(m);
790 	return (0);
791 }
792 
793 /*
794  * Free a fragment reassembly header and all
795  * associated datagrams.
796  */
797 static void
798 ip_freef(fp)
799 	struct ipq *fp;
800 {
801 	register struct ipasfrag *q, *p;
802 
803 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) {
804 		p = q->ipf_next;
805 		ip_deq(q);
806 		m_freem(dtom(q));
807 	}
808 	remque(fp);
809 	(void) m_free(dtom(fp));
810 	nipq--;
811 }
812 
813 /*
814  * Put an ip fragment on a reassembly chain.
815  * Like insque, but pointers in middle of structure.
816  */
817 static void
818 ip_enq(p, prev)
819 	register struct ipasfrag *p, *prev;
820 {
821 
822 	p->ipf_prev = prev;
823 	p->ipf_next = prev->ipf_next;
824 	prev->ipf_next->ipf_prev = p;
825 	prev->ipf_next = p;
826 }
827 
828 /*
829  * To ip_enq as remque is to insque.
830  */
831 static void
832 ip_deq(p)
833 	register struct ipasfrag *p;
834 {
835 
836 	p->ipf_prev->ipf_next = p->ipf_next;
837 	p->ipf_next->ipf_prev = p->ipf_prev;
838 }
839 
840 /*
841  * IP timer processing;
842  * if a timer expires on a reassembly
843  * queue, discard it.
844  */
845 void
846 ip_slowtimo()
847 {
848 	register struct ipq *fp;
849 	int s = splnet();
850 	int i;
851 
852 	for (i = 0; i < IPREASS_NHASH; i++) {
853 		fp = ipq[i].next;
854 		if (fp == 0)
855 			continue;
856 		while (fp != &ipq[i]) {
857 			--fp->ipq_ttl;
858 			fp = fp->next;
859 			if (fp->prev->ipq_ttl == 0) {
860 				ipstat.ips_fragtimeout++;
861 				ip_freef(fp->prev);
862 			}
863 		}
864 	}
865 	splx(s);
866 }
867 
868 /*
869  * Drain off all datagram fragments.
870  */
871 void
872 ip_drain()
873 {
874 	int     i;
875 
876 	for (i = 0; i < IPREASS_NHASH; i++) {
877 		while (ipq[i].next != &ipq[i]) {
878 			ipstat.ips_fragdropped++;
879 			ip_freef(ipq[i].next);
880 		}
881 	}
882 	in_rtqdrain();
883 }
884 
885 /*
886  * Do option processing on a datagram,
887  * possibly discarding it if bad options are encountered,
888  * or forwarding it if source-routed.
889  * Returns 1 if packet has been forwarded/freed,
890  * 0 if the packet should be processed further.
891  */
892 static int
893 ip_dooptions(m)
894 	struct mbuf *m;
895 {
896 	register struct ip *ip = mtod(m, struct ip *);
897 	register u_char *cp;
898 	register struct ip_timestamp *ipt;
899 	register struct in_ifaddr *ia;
900 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
901 	struct in_addr *sin, dst;
902 	n_time ntime;
903 
904 	dst = ip->ip_dst;
905 	cp = (u_char *)(ip + 1);
906 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
907 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
908 		opt = cp[IPOPT_OPTVAL];
909 		if (opt == IPOPT_EOL)
910 			break;
911 		if (opt == IPOPT_NOP)
912 			optlen = 1;
913 		else {
914 			optlen = cp[IPOPT_OLEN];
915 			if (optlen <= 0 || optlen > cnt) {
916 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
917 				goto bad;
918 			}
919 		}
920 		switch (opt) {
921 
922 		default:
923 			break;
924 
925 		/*
926 		 * Source routing with record.
927 		 * Find interface with current destination address.
928 		 * If none on this machine then drop if strictly routed,
929 		 * or do nothing if loosely routed.
930 		 * Record interface address and bring up next address
931 		 * component.  If strictly routed make sure next
932 		 * address is on directly accessible net.
933 		 */
934 		case IPOPT_LSRR:
935 		case IPOPT_SSRR:
936 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
937 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
938 				goto bad;
939 			}
940 			ipaddr.sin_addr = ip->ip_dst;
941 			ia = (struct in_ifaddr *)
942 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
943 			if (ia == 0) {
944 				if (opt == IPOPT_SSRR) {
945 					type = ICMP_UNREACH;
946 					code = ICMP_UNREACH_SRCFAIL;
947 					goto bad;
948 				}
949 				/*
950 				 * Loose routing, and not at next destination
951 				 * yet; nothing to do except forward.
952 				 */
953 				break;
954 			}
955 			off--;			/* 0 origin */
956 			if (off > optlen - sizeof(struct in_addr)) {
957 				/*
958 				 * End of source route.  Should be for us.
959 				 */
960 				save_rte(cp, ip->ip_src);
961 				break;
962 			}
963 
964 			if (!ip_dosourceroute) {
965 				char buf[4*sizeof "123"];
966 				strcpy(buf, inet_ntoa(ip->ip_dst));
967 
968 				log(LOG_WARNING,
969 				    "attempted source route from %s to %s\n",
970 				    inet_ntoa(ip->ip_src), buf);
971 				type = ICMP_UNREACH;
972 				code = ICMP_UNREACH_SRCFAIL;
973 				goto bad;
974 			}
975 
976 			/*
977 			 * locate outgoing interface
978 			 */
979 			(void)memcpy(&ipaddr.sin_addr, cp + off,
980 			    sizeof(ipaddr.sin_addr));
981 
982 			if (opt == IPOPT_SSRR) {
983 #define	INA	struct in_ifaddr *
984 #define	SA	struct sockaddr *
985 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
986 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
987 			} else
988 				ia = ip_rtaddr(ipaddr.sin_addr);
989 			if (ia == 0) {
990 				type = ICMP_UNREACH;
991 				code = ICMP_UNREACH_SRCFAIL;
992 				goto bad;
993 			}
994 			ip->ip_dst = ipaddr.sin_addr;
995 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
996 			    sizeof(struct in_addr));
997 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
998 			/*
999 			 * Let ip_intr's mcast routing check handle mcast pkts
1000 			 */
1001 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1002 			break;
1003 
1004 		case IPOPT_RR:
1005 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1006 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1007 				goto bad;
1008 			}
1009 			/*
1010 			 * If no space remains, ignore.
1011 			 */
1012 			off--;			/* 0 origin */
1013 			if (off > optlen - sizeof(struct in_addr))
1014 				break;
1015 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1016 			    sizeof(ipaddr.sin_addr));
1017 			/*
1018 			 * locate outgoing interface; if we're the destination,
1019 			 * use the incoming interface (should be same).
1020 			 */
1021 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1022 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1023 				type = ICMP_UNREACH;
1024 				code = ICMP_UNREACH_HOST;
1025 				goto bad;
1026 			}
1027 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1028 			    sizeof(struct in_addr));
1029 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1030 			break;
1031 
1032 		case IPOPT_TS:
1033 			code = cp - (u_char *)ip;
1034 			ipt = (struct ip_timestamp *)cp;
1035 			if (ipt->ipt_len < 5)
1036 				goto bad;
1037 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) {
1038 				if (++ipt->ipt_oflw == 0)
1039 					goto bad;
1040 				break;
1041 			}
1042 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
1043 			switch (ipt->ipt_flg) {
1044 
1045 			case IPOPT_TS_TSONLY:
1046 				break;
1047 
1048 			case IPOPT_TS_TSANDADDR:
1049 				if (ipt->ipt_ptr + sizeof(n_time) +
1050 				    sizeof(struct in_addr) > ipt->ipt_len)
1051 					goto bad;
1052 				ipaddr.sin_addr = dst;
1053 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1054 							    m->m_pkthdr.rcvif);
1055 				if (ia == 0)
1056 					continue;
1057 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1058 				    sizeof(struct in_addr));
1059 				ipt->ipt_ptr += sizeof(struct in_addr);
1060 				break;
1061 
1062 			case IPOPT_TS_PRESPEC:
1063 				if (ipt->ipt_ptr + sizeof(n_time) +
1064 				    sizeof(struct in_addr) > ipt->ipt_len)
1065 					goto bad;
1066 				(void)memcpy(&ipaddr.sin_addr, sin,
1067 				    sizeof(struct in_addr));
1068 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1069 					continue;
1070 				ipt->ipt_ptr += sizeof(struct in_addr);
1071 				break;
1072 
1073 			default:
1074 				goto bad;
1075 			}
1076 			ntime = iptime();
1077 			(void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
1078 			    sizeof(n_time));
1079 			ipt->ipt_ptr += sizeof(n_time);
1080 		}
1081 	}
1082 	if (forward) {
1083 		ip_forward(m, 1);
1084 		return (1);
1085 	}
1086 	return (0);
1087 bad:
1088 	ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2;   /* XXX icmp_error adds in hdr length */
1089 	icmp_error(m, type, code, 0, 0);
1090 	ipstat.ips_badoptions++;
1091 	return (1);
1092 }
1093 
1094 /*
1095  * Given address of next destination (final or next hop),
1096  * return internet address info of interface to be used to get there.
1097  */
1098 static struct in_ifaddr *
1099 ip_rtaddr(dst)
1100 	 struct in_addr dst;
1101 {
1102 	register struct sockaddr_in *sin;
1103 
1104 	sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1105 
1106 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1107 		if (ipforward_rt.ro_rt) {
1108 			RTFREE(ipforward_rt.ro_rt);
1109 			ipforward_rt.ro_rt = 0;
1110 		}
1111 		sin->sin_family = AF_INET;
1112 		sin->sin_len = sizeof(*sin);
1113 		sin->sin_addr = dst;
1114 
1115 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1116 	}
1117 	if (ipforward_rt.ro_rt == 0)
1118 		return ((struct in_ifaddr *)0);
1119 	return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
1120 }
1121 
1122 /*
1123  * Save incoming source route for use in replies,
1124  * to be picked up later by ip_srcroute if the receiver is interested.
1125  */
1126 void
1127 save_rte(option, dst)
1128 	u_char *option;
1129 	struct in_addr dst;
1130 {
1131 	unsigned olen;
1132 
1133 	olen = option[IPOPT_OLEN];
1134 #ifdef DIAGNOSTIC
1135 	if (ipprintfs)
1136 		printf("save_rte: olen %d\n", olen);
1137 #endif
1138 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1139 		return;
1140 	bcopy(option, ip_srcrt.srcopt, olen);
1141 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1142 	ip_srcrt.dst = dst;
1143 }
1144 
1145 /*
1146  * Retrieve incoming source route for use in replies,
1147  * in the same form used by setsockopt.
1148  * The first hop is placed before the options, will be removed later.
1149  */
1150 struct mbuf *
1151 ip_srcroute()
1152 {
1153 	register struct in_addr *p, *q;
1154 	register struct mbuf *m;
1155 
1156 	if (ip_nhops == 0)
1157 		return ((struct mbuf *)0);
1158 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1159 	if (m == 0)
1160 		return ((struct mbuf *)0);
1161 
1162 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1163 
1164 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1165 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1166 	    OPTSIZ;
1167 #ifdef DIAGNOSTIC
1168 	if (ipprintfs)
1169 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1170 #endif
1171 
1172 	/*
1173 	 * First save first hop for return route
1174 	 */
1175 	p = &ip_srcrt.route[ip_nhops - 1];
1176 	*(mtod(m, struct in_addr *)) = *p--;
1177 #ifdef DIAGNOSTIC
1178 	if (ipprintfs)
1179 		printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr));
1180 #endif
1181 
1182 	/*
1183 	 * Copy option fields and padding (nop) to mbuf.
1184 	 */
1185 	ip_srcrt.nop = IPOPT_NOP;
1186 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1187 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1188 	    &ip_srcrt.nop, OPTSIZ);
1189 	q = (struct in_addr *)(mtod(m, caddr_t) +
1190 	    sizeof(struct in_addr) + OPTSIZ);
1191 #undef OPTSIZ
1192 	/*
1193 	 * Record return path as an IP source route,
1194 	 * reversing the path (pointers are now aligned).
1195 	 */
1196 	while (p >= ip_srcrt.route) {
1197 #ifdef DIAGNOSTIC
1198 		if (ipprintfs)
1199 			printf(" %lx", ntohl(q->s_addr));
1200 #endif
1201 		*q++ = *p--;
1202 	}
1203 	/*
1204 	 * Last hop goes to final destination.
1205 	 */
1206 	*q = ip_srcrt.dst;
1207 #ifdef DIAGNOSTIC
1208 	if (ipprintfs)
1209 		printf(" %lx\n", ntohl(q->s_addr));
1210 #endif
1211 	return (m);
1212 }
1213 
1214 /*
1215  * Strip out IP options, at higher
1216  * level protocol in the kernel.
1217  * Second argument is buffer to which options
1218  * will be moved, and return value is their length.
1219  * XXX should be deleted; last arg currently ignored.
1220  */
1221 void
1222 ip_stripoptions(m, mopt)
1223 	register struct mbuf *m;
1224 	struct mbuf *mopt;
1225 {
1226 	register int i;
1227 	struct ip *ip = mtod(m, struct ip *);
1228 	register caddr_t opts;
1229 	int olen;
1230 
1231 	olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1232 	opts = (caddr_t)(ip + 1);
1233 	i = m->m_len - (sizeof (struct ip) + olen);
1234 	bcopy(opts + olen, opts, (unsigned)i);
1235 	m->m_len -= olen;
1236 	if (m->m_flags & M_PKTHDR)
1237 		m->m_pkthdr.len -= olen;
1238 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1239 }
1240 
1241 u_char inetctlerrmap[PRC_NCMDS] = {
1242 	0,		0,		0,		0,
1243 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1244 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1245 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1246 	0,		0,		0,		0,
1247 	ENOPROTOOPT
1248 };
1249 
1250 /*
1251  * Forward a packet.  If some error occurs return the sender
1252  * an icmp packet.  Note we can't always generate a meaningful
1253  * icmp message because icmp doesn't have a large enough repertoire
1254  * of codes and types.
1255  *
1256  * If not forwarding, just drop the packet.  This could be confusing
1257  * if ipforwarding was zero but some routing protocol was advancing
1258  * us as a gateway to somewhere.  However, we must let the routing
1259  * protocol deal with that.
1260  *
1261  * The srcrt parameter indicates whether the packet is being forwarded
1262  * via a source route.
1263  */
1264 static void
1265 ip_forward(m, srcrt)
1266 	struct mbuf *m;
1267 	int srcrt;
1268 {
1269 	register struct ip *ip = mtod(m, struct ip *);
1270 	register struct sockaddr_in *sin;
1271 	register struct rtentry *rt;
1272 	int error, type = 0, code = 0;
1273 	struct mbuf *mcopy;
1274 	n_long dest;
1275 	struct ifnet *destifp;
1276 
1277 	dest = 0;
1278 #ifdef DIAGNOSTIC
1279 	if (ipprintfs)
1280 		printf("forward: src %lx dst %lx ttl %x\n",
1281 			ip->ip_src.s_addr, ip->ip_dst.s_addr, ip->ip_ttl);
1282 #endif
1283 
1284 
1285 	if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) {
1286 		ipstat.ips_cantforward++;
1287 		m_freem(m);
1288 		return;
1289 	}
1290 	HTONS(ip->ip_id);
1291 	if (ip->ip_ttl <= IPTTLDEC) {
1292 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1293 		return;
1294 	}
1295 	ip->ip_ttl -= IPTTLDEC;
1296 
1297 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1298 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1299 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1300 		if (ipforward_rt.ro_rt) {
1301 			RTFREE(ipforward_rt.ro_rt);
1302 			ipforward_rt.ro_rt = 0;
1303 		}
1304 		sin->sin_family = AF_INET;
1305 		sin->sin_len = sizeof(*sin);
1306 		sin->sin_addr = ip->ip_dst;
1307 
1308 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1309 		if (ipforward_rt.ro_rt == 0) {
1310 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1311 			return;
1312 		}
1313 		rt = ipforward_rt.ro_rt;
1314 	}
1315 
1316 	/*
1317 	 * Save at most 64 bytes of the packet in case
1318 	 * we need to generate an ICMP message to the src.
1319 	 */
1320 	mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64));
1321 
1322 	/*
1323 	 * If forwarding packet using same interface that it came in on,
1324 	 * perhaps should send a redirect to sender to shortcut a hop.
1325 	 * Only send redirect if source is sending directly to us,
1326 	 * and if packet was not source routed (or has any options).
1327 	 * Also, don't send redirect if forwarding using a default route
1328 	 * or a route modified by a redirect.
1329 	 */
1330 #define	satosin(sa)	((struct sockaddr_in *)(sa))
1331 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1332 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1333 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1334 	    ipsendredirects && !srcrt) {
1335 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1336 		u_long src = ntohl(ip->ip_src.s_addr);
1337 
1338 		if (RTA(rt) &&
1339 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1340 		    if (rt->rt_flags & RTF_GATEWAY)
1341 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1342 		    else
1343 			dest = ip->ip_dst.s_addr;
1344 		    /* Router requirements says to only send host redirects */
1345 		    type = ICMP_REDIRECT;
1346 		    code = ICMP_REDIRECT_HOST;
1347 #ifdef DIAGNOSTIC
1348 		    if (ipprintfs)
1349 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1350 #endif
1351 		}
1352 	}
1353 
1354 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1355 			  IP_FORWARDING, 0);
1356 	if (error)
1357 		ipstat.ips_cantforward++;
1358 	else {
1359 		ipstat.ips_forward++;
1360 		if (type)
1361 			ipstat.ips_redirectsent++;
1362 		else {
1363 			if (mcopy)
1364 				m_freem(mcopy);
1365 			return;
1366 		}
1367 	}
1368 	if (mcopy == NULL)
1369 		return;
1370 	destifp = NULL;
1371 
1372 	switch (error) {
1373 
1374 	case 0:				/* forwarded, but need redirect */
1375 		/* type, code set above */
1376 		break;
1377 
1378 	case ENETUNREACH:		/* shouldn't happen, checked above */
1379 	case EHOSTUNREACH:
1380 	case ENETDOWN:
1381 	case EHOSTDOWN:
1382 	default:
1383 		type = ICMP_UNREACH;
1384 		code = ICMP_UNREACH_HOST;
1385 		break;
1386 
1387 	case EMSGSIZE:
1388 		type = ICMP_UNREACH;
1389 		code = ICMP_UNREACH_NEEDFRAG;
1390 		if (ipforward_rt.ro_rt)
1391 			destifp = ipforward_rt.ro_rt->rt_ifp;
1392 		ipstat.ips_cantfrag++;
1393 		break;
1394 
1395 	case ENOBUFS:
1396 		type = ICMP_SOURCEQUENCH;
1397 		code = 0;
1398 		break;
1399 	}
1400 	icmp_error(mcopy, type, code, dest, destifp);
1401 }
1402 
1403 void
1404 ip_savecontrol(inp, mp, ip, m)
1405 	register struct inpcb *inp;
1406 	register struct mbuf **mp;
1407 	register struct ip *ip;
1408 	register struct mbuf *m;
1409 {
1410 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1411 		struct timeval tv;
1412 
1413 		microtime(&tv);
1414 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1415 			SCM_TIMESTAMP, SOL_SOCKET);
1416 		if (*mp)
1417 			mp = &(*mp)->m_next;
1418 	}
1419 	if (inp->inp_flags & INP_RECVDSTADDR) {
1420 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1421 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1422 		if (*mp)
1423 			mp = &(*mp)->m_next;
1424 	}
1425 #ifdef notyet
1426 	/* XXX
1427 	 * Moving these out of udp_input() made them even more broken
1428 	 * than they already were.
1429 	 */
1430 	/* options were tossed already */
1431 	if (inp->inp_flags & INP_RECVOPTS) {
1432 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1433 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1434 		if (*mp)
1435 			mp = &(*mp)->m_next;
1436 	}
1437 	/* ip_srcroute doesn't do what we want here, need to fix */
1438 	if (inp->inp_flags & INP_RECVRETOPTS) {
1439 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1440 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1441 		if (*mp)
1442 			mp = &(*mp)->m_next;
1443 	}
1444 #endif
1445 	if (inp->inp_flags & INP_RECVIF) {
1446 		struct sockaddr_dl sdl;
1447 
1448 		sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
1449 		sdl.sdl_family = AF_LINK;
1450 		sdl.sdl_index = m->m_pkthdr.rcvif ?
1451 			m->m_pkthdr.rcvif->if_index : 0;
1452 		sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
1453 		*mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
1454 			IP_RECVIF, IPPROTO_IP);
1455 		if (*mp)
1456 			mp = &(*mp)->m_next;
1457 	}
1458 }
1459 
1460 int
1461 ip_rsvp_init(struct socket *so)
1462 {
1463 	if (so->so_type != SOCK_RAW ||
1464 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1465 	  return EOPNOTSUPP;
1466 
1467 	if (ip_rsvpd != NULL)
1468 	  return EADDRINUSE;
1469 
1470 	ip_rsvpd = so;
1471 	/*
1472 	 * This may seem silly, but we need to be sure we don't over-increment
1473 	 * the RSVP counter, in case something slips up.
1474 	 */
1475 	if (!ip_rsvp_on) {
1476 		ip_rsvp_on = 1;
1477 		rsvp_on++;
1478 	}
1479 
1480 	return 0;
1481 }
1482 
1483 int
1484 ip_rsvp_done(void)
1485 {
1486 	ip_rsvpd = NULL;
1487 	/*
1488 	 * This may seem silly, but we need to be sure we don't over-decrement
1489 	 * the RSVP counter, in case something slips up.
1490 	 */
1491 	if (ip_rsvp_on) {
1492 		ip_rsvp_on = 0;
1493 		rsvp_on--;
1494 	}
1495 	return 0;
1496 }
1497