xref: /freebsd/sys/netinet/ip_input.c (revision 2eb8169a1ad3bd8c7960767a5bc471c8d8bc677e)
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  * $FreeBSD$
35  */
36 
37 #define	_IP_VHL
38 
39 #include "opt_bootp.h"
40 #include "opt_ipfw.h"
41 #include "opt_ipdn.h"
42 #include "opt_ipdivert.h"
43 #include "opt_ipfilter.h"
44 #include "opt_ipstealth.h"
45 #include "opt_ipsec.h"
46 #include "opt_pfil_hooks.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/mbuf.h>
51 #include <sys/malloc.h>
52 #include <sys/domain.h>
53 #include <sys/protosw.h>
54 #include <sys/socket.h>
55 #include <sys/time.h>
56 #include <sys/kernel.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59 
60 #include <net/pfil.h>
61 #include <net/if.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
64 #include <net/route.h>
65 #include <net/netisr.h>
66 #include <net/intrq.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/in_var.h>
71 #include <netinet/ip.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/ip_icmp.h>
75 #include <machine/in_cksum.h>
76 
77 #include <netinet/ipprotosw.h>
78 
79 #include <sys/socketvar.h>
80 
81 #include <netinet/ip_fw.h>
82 
83 #ifdef IPSEC
84 #include <netinet6/ipsec.h>
85 #include <netkey/key.h>
86 #endif
87 
88 #include "faith.h"
89 #if defined(NFAITH) && NFAITH > 0
90 #include <net/if_types.h>
91 #endif
92 
93 #ifdef DUMMYNET
94 #include <netinet/ip_dummynet.h>
95 #endif
96 
97 int rsvp_on = 0;
98 static int ip_rsvp_on;
99 struct socket *ip_rsvpd;
100 
101 int	ipforwarding = 0;
102 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
103     &ipforwarding, 0, "Enable IP forwarding between interfaces");
104 
105 static int	ipsendredirects = 1; /* XXX */
106 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
107     &ipsendredirects, 0, "Enable sending IP redirects");
108 
109 int	ip_defttl = IPDEFTTL;
110 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
111     &ip_defttl, 0, "Maximum TTL on IP packets");
112 
113 static int	ip_dosourceroute = 0;
114 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
115     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
116 
117 static int	ip_acceptsourceroute = 0;
118 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
119     CTLFLAG_RW, &ip_acceptsourceroute, 0,
120     "Enable accepting source routed IP packets");
121 
122 static int	ip_keepfaith = 0;
123 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
124 	&ip_keepfaith,	0,
125 	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
126 
127 int	ip_send_unreach = 1;
128 SYSCTL_INT(_net_inet_ip, OID_AUTO, send_unreach, CTLFLAG_RW, &ip_send_unreach,
129 	0, "Send ICMP unreach when packet not for us rx, and forwarding disabled");
130 
131 #ifdef DIAGNOSTIC
132 static int	ipprintfs = 0;
133 #endif
134 
135 extern	struct domain inetdomain;
136 extern	struct ipprotosw inetsw[];
137 u_char	ip_protox[IPPROTO_MAX];
138 static int	ipqmaxlen = IFQ_MAXLEN;
139 struct	in_ifaddrhead in_ifaddrhead; /* first inet address */
140 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
141     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
142 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
143     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
144 
145 struct ipstat ipstat;
146 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD,
147     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
148 
149 /* Packet reassembly stuff */
150 #define IPREASS_NHASH_LOG2      6
151 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
152 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
153 #define IPREASS_HASH(x,y) \
154 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
155 
156 static struct ipq ipq[IPREASS_NHASH];
157 static int    nipq = 0;         /* total # of reass queues */
158 static int    maxnipq;
159 const  int    ipintrq_present = 1;
160 
161 #ifdef IPCTL_DEFMTU
162 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
163     &ip_mtu, 0, "Default MTU");
164 #endif
165 
166 #ifdef IPSTEALTH
167 static int	ipstealth = 0;
168 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
169     &ipstealth, 0, "");
170 #endif
171 
172 
173 /* Firewall hooks */
174 ip_fw_chk_t *ip_fw_chk_ptr;
175 ip_fw_ctl_t *ip_fw_ctl_ptr;
176 int fw_enable = 1 ;
177 
178 #ifdef DUMMYNET
179 ip_dn_ctl_t *ip_dn_ctl_ptr;
180 #endif
181 
182 
183 /*
184  * We need to save the IP options in case a protocol wants to respond
185  * to an incoming packet over the same route if the packet got here
186  * using IP source routing.  This allows connection establishment and
187  * maintenance when the remote end is on a network that is not known
188  * to us.
189  */
190 static int	ip_nhops = 0;
191 static	struct ip_srcrt {
192 	struct	in_addr dst;			/* final destination */
193 	char	nop;				/* one NOP to align */
194 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
195 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
196 } ip_srcrt;
197 
198 struct sockaddr_in *ip_fw_fwd_addr;
199 
200 static void	save_rte __P((u_char *, struct in_addr));
201 static int	ip_dooptions __P((struct mbuf *));
202 static void	ip_forward __P((struct mbuf *, int));
203 static void	ip_freef __P((struct ipq *));
204 #ifdef IPDIVERT
205 static struct	mbuf *ip_reass __P((struct mbuf *,
206 			struct ipq *, struct ipq *, u_int32_t *, u_int16_t *));
207 #else
208 static struct	mbuf *ip_reass __P((struct mbuf *, struct ipq *, struct ipq *));
209 #endif
210 static struct	in_ifaddr *ip_rtaddr __P((struct in_addr));
211 static void	ipintr __P((void));
212 
213 /*
214  * IP initialization: fill in IP protocol switch table.
215  * All protocols not implemented in kernel go to raw IP protocol handler.
216  */
217 void
218 ip_init()
219 {
220 	register struct ipprotosw *pr;
221 	register int i;
222 
223 	TAILQ_INIT(&in_ifaddrhead);
224 	pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
225 	if (pr == 0)
226 		panic("ip_init");
227 	for (i = 0; i < IPPROTO_MAX; i++)
228 		ip_protox[i] = pr - inetsw;
229 	for (pr = (struct ipprotosw *)inetdomain.dom_protosw;
230 	    pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++)
231 		if (pr->pr_domain->dom_family == PF_INET &&
232 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
233 			ip_protox[pr->pr_protocol] = pr - inetsw;
234 
235 	for (i = 0; i < IPREASS_NHASH; i++)
236 	    ipq[i].next = ipq[i].prev = &ipq[i];
237 
238 	maxnipq = nmbclusters/4;
239 
240 	ip_id = time_second & 0xffff;
241 	ipintrq.ifq_maxlen = ipqmaxlen;
242 	mtx_init(&ipintrq.ifq_mtx, "ip_inq", MTX_DEF);
243 
244 	register_netisr(NETISR_IP, ipintr);
245 }
246 
247 static struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
248 static struct	route ipforward_rt;
249 
250 /*
251  * Ip input routine.  Checksum and byte swap header.  If fragmented
252  * try to reassemble.  Process options.  Pass to next level.
253  */
254 void
255 ip_input(struct mbuf *m)
256 {
257 	struct ip *ip;
258 	struct ipq *fp;
259 	struct in_ifaddr *ia = NULL;
260 	int    i, hlen;
261 	u_short sum;
262 	u_int16_t divert_cookie;		/* firewall cookie */
263 #ifdef IPDIVERT
264 	u_int32_t divert_info = 0;		/* packet divert/tee info */
265 #endif
266 	struct ip_fw_chain *rule = NULL;
267 #ifdef PFIL_HOOKS
268 	struct packet_filter_hook *pfh;
269 	struct mbuf *m0;
270 	int rv;
271 #endif /* PFIL_HOOKS */
272 
273 #ifdef IPDIVERT
274 	/* Get and reset firewall cookie */
275 	divert_cookie = ip_divert_cookie;
276 	ip_divert_cookie = 0;
277 #else
278 	divert_cookie = 0;
279 #endif
280 
281 #if defined(IPFIREWALL) && defined(DUMMYNET)
282         /*
283          * dummynet packet are prepended a vestigial mbuf with
284          * m_type = MT_DUMMYNET and m_data pointing to the matching
285          * rule.
286          */
287         if (m->m_type == MT_DUMMYNET) {
288             rule = (struct ip_fw_chain *)(m->m_data) ;
289             m = m->m_next ;
290             ip = mtod(m, struct ip *);
291             hlen = IP_VHL_HL(ip->ip_vhl) << 2;
292             goto iphack ;
293         } else
294             rule = NULL ;
295 #endif
296 
297 #ifdef	DIAGNOSTIC
298 	if (m == NULL || (m->m_flags & M_PKTHDR) == 0)
299 		panic("ip_input no HDR");
300 #endif
301 	ipstat.ips_total++;
302 
303 	if (m->m_pkthdr.len < sizeof(struct ip))
304 		goto tooshort;
305 
306 	if (m->m_len < sizeof (struct ip) &&
307 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
308 		ipstat.ips_toosmall++;
309 		return;
310 	}
311 	ip = mtod(m, struct ip *);
312 
313 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
314 		ipstat.ips_badvers++;
315 		goto bad;
316 	}
317 
318 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
319 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
320 		ipstat.ips_badhlen++;
321 		goto bad;
322 	}
323 	if (hlen > m->m_len) {
324 		if ((m = m_pullup(m, hlen)) == 0) {
325 			ipstat.ips_badhlen++;
326 			return;
327 		}
328 		ip = mtod(m, struct ip *);
329 	}
330 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
331 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
332 	} else {
333 		if (hlen == sizeof(struct ip)) {
334 			sum = in_cksum_hdr(ip);
335 		} else {
336 			sum = in_cksum(m, hlen);
337 		}
338 	}
339 	if (sum) {
340 		ipstat.ips_badsum++;
341 		goto bad;
342 	}
343 
344 	/*
345 	 * Convert fields to host representation.
346 	 */
347 	NTOHS(ip->ip_len);
348 	if (ip->ip_len < hlen) {
349 		ipstat.ips_badlen++;
350 		goto bad;
351 	}
352 	NTOHS(ip->ip_off);
353 
354 	/*
355 	 * Check that the amount of data in the buffers
356 	 * is as at least much as the IP header would have us expect.
357 	 * Trim mbufs if longer than we expect.
358 	 * Drop packet if shorter than we expect.
359 	 */
360 	if (m->m_pkthdr.len < ip->ip_len) {
361 tooshort:
362 		ipstat.ips_tooshort++;
363 		goto bad;
364 	}
365 	if (m->m_pkthdr.len > ip->ip_len) {
366 		if (m->m_len == m->m_pkthdr.len) {
367 			m->m_len = ip->ip_len;
368 			m->m_pkthdr.len = ip->ip_len;
369 		} else
370 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
371 	}
372 	/*
373 	 * IpHack's section.
374 	 * Right now when no processing on packet has done
375 	 * and it is still fresh out of network we do our black
376 	 * deals with it.
377 	 * - Firewall: deny/allow/divert
378 	 * - Xlate: translate packet's addr/port (NAT).
379 	 * - Pipe: pass pkt through dummynet.
380 	 * - Wrap: fake packet's addr/port <unimpl.>
381 	 * - Encapsulate: put it in another IP and send out. <unimp.>
382  	 */
383 
384 #if defined(IPFIREWALL) && defined(DUMMYNET)
385 iphack:
386 #endif
387 
388 #ifdef PFIL_HOOKS
389 	/*
390 	 * Run through list of hooks for input packets.  If there are any
391 	 * filters which require that additional packets in the flow are
392 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
393 	 * Note that filters must _never_ set this flag, as another filter
394 	 * in the list may have previously cleared it.
395 	 */
396 	m0 = m;
397 	pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
398 	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
399 		if (pfh->pfil_func) {
400 			rv = pfh->pfil_func(ip, hlen,
401 					    m->m_pkthdr.rcvif, 0, &m0);
402 			if (rv)
403 				return;
404 			m = m0;
405 			if (m == NULL)
406 				return;
407 			ip = mtod(m, struct ip *);
408 		}
409 #endif /* PFIL_HOOKS */
410 
411 	if (fw_enable && ip_fw_chk_ptr) {
412 #ifdef IPFIREWALL_FORWARD
413 		/*
414 		 * If we've been forwarded from the output side, then
415 		 * skip the firewall a second time
416 		 */
417 		if (ip_fw_fwd_addr)
418 			goto ours;
419 #endif	/* IPFIREWALL_FORWARD */
420 		/*
421 		 * See the comment in ip_output for the return values
422 		 * produced by the firewall.
423 		 */
424 		i = (*ip_fw_chk_ptr)(&ip,
425 		    hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr);
426 		if (i & IP_FW_PORT_DENY_FLAG) { /* XXX new interface-denied */
427 		    if (m)
428 			m_freem(m);
429 		    return ;
430 		}
431 		if (m == NULL) {	/* Packet discarded by firewall */
432 		    static int __debug=10;
433 		    if (__debug >0) {
434 			printf("firewall returns NULL, please update!\n");
435 			__debug-- ;
436 		    }
437 		    return;
438 		}
439 		if (i == 0 && ip_fw_fwd_addr == NULL)	/* common case */
440 			goto pass;
441 #ifdef DUMMYNET
442                 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) {
443                         /* Send packet to the appropriate pipe */
444                         dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule,
445 				    0);
446 			return;
447 		}
448 #endif
449 #ifdef IPDIVERT
450 		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
451 			/* Divert or tee packet */
452 			divert_info = i;
453 			goto ours;
454 		}
455 #endif
456 #ifdef IPFIREWALL_FORWARD
457 		if (i == 0 && ip_fw_fwd_addr != NULL)
458 			goto pass;
459 #endif
460 		/*
461 		 * if we get here, the packet must be dropped
462 		 */
463 		m_freem(m);
464 		return;
465 	}
466 pass:
467 
468 	/*
469 	 * Process options and, if not destined for us,
470 	 * ship it on.  ip_dooptions returns 1 when an
471 	 * error was detected (causing an icmp message
472 	 * to be sent and the original packet to be freed).
473 	 */
474 	ip_nhops = 0;		/* for source routed packets */
475 	if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
476 #ifdef IPFIREWALL_FORWARD
477 		ip_fw_fwd_addr = NULL;
478 #endif
479 		return;
480 	}
481 
482         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
483          * matter if it is destined to another node, or whether it is
484          * a multicast one, RSVP wants it! and prevents it from being forwarded
485          * anywhere else. Also checks if the rsvp daemon is running before
486 	 * grabbing the packet.
487          */
488 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
489 		goto ours;
490 
491 	/*
492 	 * Check our list of addresses, to see if the packet is for us.
493 	 * If we don't have any addresses, assume any unicast packet
494 	 * we receive might be for us (and let the upper layers deal
495 	 * with it).
496 	 */
497 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
498 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
499 		goto ours;
500 
501 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
502 #define	satosin(sa)	((struct sockaddr_in *)(sa))
503 
504 #ifdef BOOTP_COMPAT
505 		if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
506 			goto ours;
507 #endif
508 #ifdef IPFIREWALL_FORWARD
509 		/*
510 		 * If the addr to forward to is one of ours, we pretend to
511 		 * be the destination for this packet.
512 		 */
513 		if (ip_fw_fwd_addr == NULL) {
514 			if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
515 				goto ours;
516 		} else if (IA_SIN(ia)->sin_addr.s_addr ==
517 					 ip_fw_fwd_addr->sin_addr.s_addr)
518 			goto ours;
519 #else
520 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
521 			goto ours;
522 #endif
523 		if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) {
524 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
525 			    ip->ip_dst.s_addr)
526 				goto ours;
527 			if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr)
528 				goto ours;
529 		}
530 	}
531 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
532 		struct in_multi *inm;
533 		if (ip_mrouter) {
534 			/*
535 			 * If we are acting as a multicast router, all
536 			 * incoming multicast packets are passed to the
537 			 * kernel-level multicast forwarding function.
538 			 * The packet is returned (relatively) intact; if
539 			 * ip_mforward() returns a non-zero value, the packet
540 			 * must be discarded, else it may be accepted below.
541 			 */
542 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
543 				ipstat.ips_cantforward++;
544 				m_freem(m);
545 				return;
546 			}
547 
548 			/*
549 			 * The process-level routing demon needs to receive
550 			 * all multicast IGMP packets, whether or not this
551 			 * host belongs to their destination groups.
552 			 */
553 			if (ip->ip_p == IPPROTO_IGMP)
554 				goto ours;
555 			ipstat.ips_forward++;
556 		}
557 		/*
558 		 * See if we belong to the destination multicast group on the
559 		 * arrival interface.
560 		 */
561 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
562 		if (inm == NULL) {
563 			ipstat.ips_notmember++;
564 			m_freem(m);
565 			return;
566 		}
567 		goto ours;
568 	}
569 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
570 		goto ours;
571 	if (ip->ip_dst.s_addr == INADDR_ANY)
572 		goto ours;
573 
574 #if defined(NFAITH) && 0 < NFAITH
575 	/*
576 	 * FAITH(Firewall Aided Internet Translator)
577 	 */
578 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
579 		if (ip_keepfaith) {
580 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
581 				goto ours;
582 		}
583 		m_freem(m);
584 		return;
585 	}
586 #endif
587 	/*
588 	 * Not for us; forward if possible and desirable.
589 	 */
590 	if (ipforwarding == 0) {
591 		ipstat.ips_cantforward++;
592 		/*
593 		 * If we receive a packet not for us, and forwarding disabled
594 		 * send a ICMP host unreachable back to the source.
595 		 */
596 		if (ip_send_unreach)
597 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
598 		else
599 			m_freem(m);
600 	} else
601 		ip_forward(m, 0);
602 #ifdef IPFIREWALL_FORWARD
603 	ip_fw_fwd_addr = NULL;
604 #endif
605 	return;
606 
607 ours:
608 	/* Count the packet in the ip address stats */
609 	if (ia != NULL) {
610 		ia->ia_ifa.if_ipackets++;
611 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
612 	}
613 
614 	/*
615 	 * If offset or IP_MF are set, must reassemble.
616 	 * Otherwise, nothing need be done.
617 	 * (We could look in the reassembly queue to see
618 	 * if the packet was previously fragmented,
619 	 * but it's not worth the time; just let them time out.)
620 	 */
621 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
622 
623 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
624 		/*
625 		 * Look for queue of fragments
626 		 * of this datagram.
627 		 */
628 		for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
629 			if (ip->ip_id == fp->ipq_id &&
630 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
631 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
632 			    ip->ip_p == fp->ipq_p)
633 				goto found;
634 
635 		fp = 0;
636 
637 		/* check if there's a place for the new queue */
638 		if (nipq > maxnipq) {
639 		    /*
640 		     * drop something from the tail of the current queue
641 		     * before proceeding further
642 		     */
643 		    if (ipq[sum].prev == &ipq[sum]) {   /* gak */
644 			for (i = 0; i < IPREASS_NHASH; i++) {
645 			    if (ipq[i].prev != &ipq[i]) {
646 				ip_freef(ipq[i].prev);
647 				break;
648 			    }
649 			}
650 		    } else
651 			ip_freef(ipq[sum].prev);
652 		}
653 found:
654 		/*
655 		 * Adjust ip_len to not reflect header,
656 		 * convert offset of this to bytes.
657 		 */
658 		ip->ip_len -= hlen;
659 		if (ip->ip_off & IP_MF) {
660 		        /*
661 		         * Make sure that fragments have a data length
662 			 * that's a non-zero multiple of 8 bytes.
663 		         */
664 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
665 				ipstat.ips_toosmall++; /* XXX */
666 				goto bad;
667 			}
668 			m->m_flags |= M_FRAG;
669 		}
670 		ip->ip_off <<= 3;
671 
672 		/*
673 		 * Attempt reassembly; if it succeeds, proceed.
674 		 */
675 		ipstat.ips_fragments++;
676 		m->m_pkthdr.header = ip;
677 #ifdef IPDIVERT
678 		m = ip_reass(m,
679 		    fp, &ipq[sum], &divert_info, &divert_cookie);
680 #else
681 		m = ip_reass(m, fp, &ipq[sum]);
682 #endif
683 		if (m == 0) {
684 #ifdef IPFIREWALL_FORWARD
685 			ip_fw_fwd_addr = NULL;
686 #endif
687 			return;
688 		}
689 		ipstat.ips_reassembled++;
690 		ip = mtod(m, struct ip *);
691 		/* Get the header length of the reassembled packet */
692 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
693 #ifdef IPDIVERT
694 		/* Restore original checksum before diverting packet */
695 		if (divert_info != 0) {
696 			ip->ip_len += hlen;
697 			HTONS(ip->ip_len);
698 			HTONS(ip->ip_off);
699 			ip->ip_sum = 0;
700 			if (hlen == sizeof(struct ip))
701 				ip->ip_sum = in_cksum_hdr(ip);
702 			else
703 				ip->ip_sum = in_cksum(m, hlen);
704 			NTOHS(ip->ip_off);
705 			NTOHS(ip->ip_len);
706 			ip->ip_len -= hlen;
707 		}
708 #endif
709 	} else
710 		ip->ip_len -= hlen;
711 
712 #ifdef IPDIVERT
713 	/*
714 	 * Divert or tee packet to the divert protocol if required.
715 	 *
716 	 * If divert_info is zero then cookie should be too, so we shouldn't
717 	 * need to clear them here.  Assume divert_packet() does so also.
718 	 */
719 	if (divert_info != 0) {
720 		struct mbuf *clone = NULL;
721 
722 		/* Clone packet if we're doing a 'tee' */
723 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
724 			clone = m_dup(m, M_DONTWAIT);
725 
726 		/* Restore packet header fields to original values */
727 		ip->ip_len += hlen;
728 		HTONS(ip->ip_len);
729 		HTONS(ip->ip_off);
730 
731 		/* Deliver packet to divert input routine */
732 		ip_divert_cookie = divert_cookie;
733 		divert_packet(m, 1, divert_info & 0xffff);
734 		ipstat.ips_delivered++;
735 
736 		/* If 'tee', continue with original packet */
737 		if (clone == NULL)
738 			return;
739 		m = clone;
740 		ip = mtod(m, struct ip *);
741 	}
742 #endif
743 
744 	/*
745 	 * Switch out to protocol's input routine.
746 	 */
747 	ipstat.ips_delivered++;
748     {
749 	int off = hlen, nh = ip->ip_p;
750 
751 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh);
752 #ifdef	IPFIREWALL_FORWARD
753 	ip_fw_fwd_addr = NULL;	/* tcp needed it */
754 #endif
755 	return;
756     }
757 bad:
758 #ifdef	IPFIREWALL_FORWARD
759 	ip_fw_fwd_addr = NULL;
760 #endif
761 	m_freem(m);
762 }
763 
764 /*
765  * IP software interrupt routine - to go away sometime soon
766  */
767 static void
768 ipintr(void)
769 {
770 	struct mbuf *m;
771 
772 	while (1) {
773 		IF_DEQUEUE(&ipintrq, m);
774 		if (m == 0)
775 			return;
776 		ip_input(m);
777 	}
778 }
779 
780 /*
781  * Take incoming datagram fragment and try to reassemble it into
782  * whole datagram.  If a chain for reassembly of this datagram already
783  * exists, then it is given as fp; otherwise have to make a chain.
784  *
785  * When IPDIVERT enabled, keep additional state with each packet that
786  * tells us if we need to divert or tee the packet we're building.
787  */
788 
789 static struct mbuf *
790 #ifdef IPDIVERT
791 ip_reass(m, fp, where, divinfo, divcookie)
792 #else
793 ip_reass(m, fp, where)
794 #endif
795 	register struct mbuf *m;
796 	register struct ipq *fp;
797 	struct   ipq    *where;
798 #ifdef IPDIVERT
799 	u_int32_t *divinfo;
800 	u_int16_t *divcookie;
801 #endif
802 {
803 	struct ip *ip = mtod(m, struct ip *);
804 	register struct mbuf *p, *q, *nq;
805 	struct mbuf *t;
806 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
807 	int i, next;
808 
809 	/*
810 	 * Presence of header sizes in mbufs
811 	 * would confuse code below.
812 	 */
813 	m->m_data += hlen;
814 	m->m_len -= hlen;
815 
816 	/*
817 	 * If first fragment to arrive, create a reassembly queue.
818 	 */
819 	if (fp == 0) {
820 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
821 			goto dropfrag;
822 		fp = mtod(t, struct ipq *);
823 		insque(fp, where);
824 		nipq++;
825 		fp->ipq_ttl = IPFRAGTTL;
826 		fp->ipq_p = ip->ip_p;
827 		fp->ipq_id = ip->ip_id;
828 		fp->ipq_src = ip->ip_src;
829 		fp->ipq_dst = ip->ip_dst;
830 		fp->ipq_frags = m;
831 		m->m_nextpkt = NULL;
832 #ifdef IPDIVERT
833 		fp->ipq_div_info = 0;
834 		fp->ipq_div_cookie = 0;
835 #endif
836 		goto inserted;
837 	}
838 
839 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
840 
841 	/*
842 	 * Find a segment which begins after this one does.
843 	 */
844 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
845 		if (GETIP(q)->ip_off > ip->ip_off)
846 			break;
847 
848 	/*
849 	 * If there is a preceding segment, it may provide some of
850 	 * our data already.  If so, drop the data from the incoming
851 	 * segment.  If it provides all of our data, drop us, otherwise
852 	 * stick new segment in the proper place.
853 	 *
854 	 * If some of the data is dropped from the the preceding
855 	 * segment, then it's checksum is invalidated.
856 	 */
857 	if (p) {
858 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
859 		if (i > 0) {
860 			if (i >= ip->ip_len)
861 				goto dropfrag;
862 			m_adj(m, i);
863 			m->m_pkthdr.csum_flags = 0;
864 			ip->ip_off += i;
865 			ip->ip_len -= i;
866 		}
867 		m->m_nextpkt = p->m_nextpkt;
868 		p->m_nextpkt = m;
869 	} else {
870 		m->m_nextpkt = fp->ipq_frags;
871 		fp->ipq_frags = m;
872 	}
873 
874 	/*
875 	 * While we overlap succeeding segments trim them or,
876 	 * if they are completely covered, dequeue them.
877 	 */
878 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
879 	     q = nq) {
880 		i = (ip->ip_off + ip->ip_len) -
881 		    GETIP(q)->ip_off;
882 		if (i < GETIP(q)->ip_len) {
883 			GETIP(q)->ip_len -= i;
884 			GETIP(q)->ip_off += i;
885 			m_adj(q, i);
886 			q->m_pkthdr.csum_flags = 0;
887 			break;
888 		}
889 		nq = q->m_nextpkt;
890 		m->m_nextpkt = nq;
891 		m_freem(q);
892 	}
893 
894 inserted:
895 
896 #ifdef IPDIVERT
897 	/*
898 	 * Transfer firewall instructions to the fragment structure.
899 	 * Any fragment diverting causes the whole packet to divert.
900 	 */
901 	fp->ipq_div_info = *divinfo;
902 	fp->ipq_div_cookie = *divcookie;
903 	*divinfo = 0;
904 	*divcookie = 0;
905 #endif
906 
907 	/*
908 	 * Check for complete reassembly.
909 	 */
910 	next = 0;
911 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
912 		if (GETIP(q)->ip_off != next)
913 			return (0);
914 		next += GETIP(q)->ip_len;
915 	}
916 	/* Make sure the last packet didn't have the IP_MF flag */
917 	if (p->m_flags & M_FRAG)
918 		return (0);
919 
920 	/*
921 	 * Reassembly is complete.  Make sure the packet is a sane size.
922 	 */
923 	q = fp->ipq_frags;
924 	ip = GETIP(q);
925 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
926 		ipstat.ips_toolong++;
927 		ip_freef(fp);
928 		return (0);
929 	}
930 
931 	/*
932 	 * Concatenate fragments.
933 	 */
934 	m = q;
935 	t = m->m_next;
936 	m->m_next = 0;
937 	m_cat(m, t);
938 	nq = q->m_nextpkt;
939 	q->m_nextpkt = 0;
940 	for (q = nq; q != NULL; q = nq) {
941 		nq = q->m_nextpkt;
942 		q->m_nextpkt = NULL;
943 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
944 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
945 		m_cat(m, q);
946 	}
947 
948 #ifdef IPDIVERT
949 	/*
950 	 * Extract firewall instructions from the fragment structure.
951 	 */
952 	*divinfo = fp->ipq_div_info;
953 	*divcookie = fp->ipq_div_cookie;
954 #endif
955 
956 	/*
957 	 * Create header for new ip packet by
958 	 * modifying header of first packet;
959 	 * dequeue and discard fragment reassembly header.
960 	 * Make header visible.
961 	 */
962 	ip->ip_len = next;
963 	ip->ip_src = fp->ipq_src;
964 	ip->ip_dst = fp->ipq_dst;
965 	remque(fp);
966 	nipq--;
967 	(void) m_free(dtom(fp));
968 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
969 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
970 	/* some debugging cruft by sklower, below, will go away soon */
971 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
972 		register int plen = 0;
973 		for (t = m; t; t = t->m_next)
974 			plen += t->m_len;
975 		m->m_pkthdr.len = plen;
976 	}
977 	return (m);
978 
979 dropfrag:
980 #ifdef IPDIVERT
981 	*divinfo = 0;
982 	*divcookie = 0;
983 #endif
984 	ipstat.ips_fragdropped++;
985 	m_freem(m);
986 	return (0);
987 
988 #undef GETIP
989 }
990 
991 /*
992  * Free a fragment reassembly header and all
993  * associated datagrams.
994  */
995 static void
996 ip_freef(fp)
997 	struct ipq *fp;
998 {
999 	register struct mbuf *q;
1000 
1001 	while (fp->ipq_frags) {
1002 		q = fp->ipq_frags;
1003 		fp->ipq_frags = q->m_nextpkt;
1004 		m_freem(q);
1005 	}
1006 	remque(fp);
1007 	(void) m_free(dtom(fp));
1008 	nipq--;
1009 }
1010 
1011 /*
1012  * IP timer processing;
1013  * if a timer expires on a reassembly
1014  * queue, discard it.
1015  */
1016 void
1017 ip_slowtimo()
1018 {
1019 	register struct ipq *fp;
1020 	int s = splnet();
1021 	int i;
1022 
1023 	for (i = 0; i < IPREASS_NHASH; i++) {
1024 		fp = ipq[i].next;
1025 		if (fp == 0)
1026 			continue;
1027 		while (fp != &ipq[i]) {
1028 			--fp->ipq_ttl;
1029 			fp = fp->next;
1030 			if (fp->prev->ipq_ttl == 0) {
1031 				ipstat.ips_fragtimeout++;
1032 				ip_freef(fp->prev);
1033 			}
1034 		}
1035 	}
1036 	ipflow_slowtimo();
1037 	splx(s);
1038 }
1039 
1040 /*
1041  * Drain off all datagram fragments.
1042  */
1043 void
1044 ip_drain()
1045 {
1046 	int     i;
1047 
1048 	for (i = 0; i < IPREASS_NHASH; i++) {
1049 		while (ipq[i].next != &ipq[i]) {
1050 			ipstat.ips_fragdropped++;
1051 			ip_freef(ipq[i].next);
1052 		}
1053 	}
1054 	in_rtqdrain();
1055 }
1056 
1057 /*
1058  * Do option processing on a datagram,
1059  * possibly discarding it if bad options are encountered,
1060  * or forwarding it if source-routed.
1061  * Returns 1 if packet has been forwarded/freed,
1062  * 0 if the packet should be processed further.
1063  */
1064 static int
1065 ip_dooptions(m)
1066 	struct mbuf *m;
1067 {
1068 	register struct ip *ip = mtod(m, struct ip *);
1069 	register u_char *cp;
1070 	register struct ip_timestamp *ipt;
1071 	register struct in_ifaddr *ia;
1072 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1073 	struct in_addr *sin, dst;
1074 	n_time ntime;
1075 
1076 	dst = ip->ip_dst;
1077 	cp = (u_char *)(ip + 1);
1078 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1079 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1080 		opt = cp[IPOPT_OPTVAL];
1081 		if (opt == IPOPT_EOL)
1082 			break;
1083 		if (opt == IPOPT_NOP)
1084 			optlen = 1;
1085 		else {
1086 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1087 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1088 				goto bad;
1089 			}
1090 			optlen = cp[IPOPT_OLEN];
1091 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1092 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1093 				goto bad;
1094 			}
1095 		}
1096 		switch (opt) {
1097 
1098 		default:
1099 			break;
1100 
1101 		/*
1102 		 * Source routing with record.
1103 		 * Find interface with current destination address.
1104 		 * If none on this machine then drop if strictly routed,
1105 		 * or do nothing if loosely routed.
1106 		 * Record interface address and bring up next address
1107 		 * component.  If strictly routed make sure next
1108 		 * address is on directly accessible net.
1109 		 */
1110 		case IPOPT_LSRR:
1111 		case IPOPT_SSRR:
1112 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1113 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1114 				goto bad;
1115 			}
1116 			ipaddr.sin_addr = ip->ip_dst;
1117 			ia = (struct in_ifaddr *)
1118 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1119 			if (ia == 0) {
1120 				if (opt == IPOPT_SSRR) {
1121 					type = ICMP_UNREACH;
1122 					code = ICMP_UNREACH_SRCFAIL;
1123 					goto bad;
1124 				}
1125 				if (!ip_dosourceroute)
1126 					goto nosourcerouting;
1127 				/*
1128 				 * Loose routing, and not at next destination
1129 				 * yet; nothing to do except forward.
1130 				 */
1131 				break;
1132 			}
1133 			off--;			/* 0 origin */
1134 			if (off > optlen - (int)sizeof(struct in_addr)) {
1135 				/*
1136 				 * End of source route.  Should be for us.
1137 				 */
1138 				if (!ip_acceptsourceroute)
1139 					goto nosourcerouting;
1140 				save_rte(cp, ip->ip_src);
1141 				break;
1142 			}
1143 
1144 			if (!ip_dosourceroute) {
1145 				if (ipforwarding) {
1146 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1147 					/*
1148 					 * Acting as a router, so generate ICMP
1149 					 */
1150 nosourcerouting:
1151 					strcpy(buf, inet_ntoa(ip->ip_dst));
1152 					log(LOG_WARNING,
1153 					    "attempted source route from %s to %s\n",
1154 					    inet_ntoa(ip->ip_src), buf);
1155 					type = ICMP_UNREACH;
1156 					code = ICMP_UNREACH_SRCFAIL;
1157 					goto bad;
1158 				} else {
1159 					/*
1160 					 * Not acting as a router, so silently drop.
1161 					 */
1162 					ipstat.ips_cantforward++;
1163 					m_freem(m);
1164 					return (1);
1165 				}
1166 			}
1167 
1168 			/*
1169 			 * locate outgoing interface
1170 			 */
1171 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1172 			    sizeof(ipaddr.sin_addr));
1173 
1174 			if (opt == IPOPT_SSRR) {
1175 #define	INA	struct in_ifaddr *
1176 #define	SA	struct sockaddr *
1177 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1178 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1179 			} else
1180 				ia = ip_rtaddr(ipaddr.sin_addr);
1181 			if (ia == 0) {
1182 				type = ICMP_UNREACH;
1183 				code = ICMP_UNREACH_SRCFAIL;
1184 				goto bad;
1185 			}
1186 			ip->ip_dst = ipaddr.sin_addr;
1187 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1188 			    sizeof(struct in_addr));
1189 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1190 			/*
1191 			 * Let ip_intr's mcast routing check handle mcast pkts
1192 			 */
1193 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1194 			break;
1195 
1196 		case IPOPT_RR:
1197 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1198 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1199 				goto bad;
1200 			}
1201 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1202 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1203 				goto bad;
1204 			}
1205 			/*
1206 			 * If no space remains, ignore.
1207 			 */
1208 			off--;			/* 0 origin */
1209 			if (off > optlen - (int)sizeof(struct in_addr))
1210 				break;
1211 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1212 			    sizeof(ipaddr.sin_addr));
1213 			/*
1214 			 * locate outgoing interface; if we're the destination,
1215 			 * use the incoming interface (should be same).
1216 			 */
1217 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1218 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1219 				type = ICMP_UNREACH;
1220 				code = ICMP_UNREACH_HOST;
1221 				goto bad;
1222 			}
1223 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1224 			    sizeof(struct in_addr));
1225 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1226 			break;
1227 
1228 		case IPOPT_TS:
1229 			code = cp - (u_char *)ip;
1230 			ipt = (struct ip_timestamp *)cp;
1231 			if (ipt->ipt_len < 5)
1232 				goto bad;
1233 			if (ipt->ipt_ptr >
1234 			    ipt->ipt_len - (int)sizeof(int32_t)) {
1235 				if (++ipt->ipt_oflw == 0)
1236 					goto bad;
1237 				break;
1238 			}
1239 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
1240 			switch (ipt->ipt_flg) {
1241 
1242 			case IPOPT_TS_TSONLY:
1243 				break;
1244 
1245 			case IPOPT_TS_TSANDADDR:
1246 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1247 				    sizeof(struct in_addr) > ipt->ipt_len)
1248 					goto bad;
1249 				ipaddr.sin_addr = dst;
1250 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1251 							    m->m_pkthdr.rcvif);
1252 				if (ia == 0)
1253 					continue;
1254 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1255 				    sizeof(struct in_addr));
1256 				ipt->ipt_ptr += sizeof(struct in_addr);
1257 				break;
1258 
1259 			case IPOPT_TS_PRESPEC:
1260 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1261 				    sizeof(struct in_addr) > ipt->ipt_len)
1262 					goto bad;
1263 				(void)memcpy(&ipaddr.sin_addr, sin,
1264 				    sizeof(struct in_addr));
1265 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1266 					continue;
1267 				ipt->ipt_ptr += sizeof(struct in_addr);
1268 				break;
1269 
1270 			default:
1271 				goto bad;
1272 			}
1273 			ntime = iptime();
1274 			(void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
1275 			    sizeof(n_time));
1276 			ipt->ipt_ptr += sizeof(n_time);
1277 		}
1278 	}
1279 	if (forward && ipforwarding) {
1280 		ip_forward(m, 1);
1281 		return (1);
1282 	}
1283 	return (0);
1284 bad:
1285 	icmp_error(m, type, code, 0, 0);
1286 	ipstat.ips_badoptions++;
1287 	return (1);
1288 }
1289 
1290 /*
1291  * Given address of next destination (final or next hop),
1292  * return internet address info of interface to be used to get there.
1293  */
1294 static struct in_ifaddr *
1295 ip_rtaddr(dst)
1296 	 struct in_addr dst;
1297 {
1298 	register struct sockaddr_in *sin;
1299 
1300 	sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1301 
1302 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1303 		if (ipforward_rt.ro_rt) {
1304 			RTFREE(ipforward_rt.ro_rt);
1305 			ipforward_rt.ro_rt = 0;
1306 		}
1307 		sin->sin_family = AF_INET;
1308 		sin->sin_len = sizeof(*sin);
1309 		sin->sin_addr = dst;
1310 
1311 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1312 	}
1313 	if (ipforward_rt.ro_rt == 0)
1314 		return ((struct in_ifaddr *)0);
1315 	return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
1316 }
1317 
1318 /*
1319  * Save incoming source route for use in replies,
1320  * to be picked up later by ip_srcroute if the receiver is interested.
1321  */
1322 void
1323 save_rte(option, dst)
1324 	u_char *option;
1325 	struct in_addr dst;
1326 {
1327 	unsigned olen;
1328 
1329 	olen = option[IPOPT_OLEN];
1330 #ifdef DIAGNOSTIC
1331 	if (ipprintfs)
1332 		printf("save_rte: olen %d\n", olen);
1333 #endif
1334 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1335 		return;
1336 	bcopy(option, ip_srcrt.srcopt, olen);
1337 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1338 	ip_srcrt.dst = dst;
1339 }
1340 
1341 /*
1342  * Retrieve incoming source route for use in replies,
1343  * in the same form used by setsockopt.
1344  * The first hop is placed before the options, will be removed later.
1345  */
1346 struct mbuf *
1347 ip_srcroute()
1348 {
1349 	register struct in_addr *p, *q;
1350 	register struct mbuf *m;
1351 
1352 	if (ip_nhops == 0)
1353 		return ((struct mbuf *)0);
1354 	m = m_get(M_DONTWAIT, MT_HEADER);
1355 	if (m == 0)
1356 		return ((struct mbuf *)0);
1357 
1358 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1359 
1360 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1361 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1362 	    OPTSIZ;
1363 #ifdef DIAGNOSTIC
1364 	if (ipprintfs)
1365 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1366 #endif
1367 
1368 	/*
1369 	 * First save first hop for return route
1370 	 */
1371 	p = &ip_srcrt.route[ip_nhops - 1];
1372 	*(mtod(m, struct in_addr *)) = *p--;
1373 #ifdef DIAGNOSTIC
1374 	if (ipprintfs)
1375 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1376 #endif
1377 
1378 	/*
1379 	 * Copy option fields and padding (nop) to mbuf.
1380 	 */
1381 	ip_srcrt.nop = IPOPT_NOP;
1382 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1383 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1384 	    &ip_srcrt.nop, OPTSIZ);
1385 	q = (struct in_addr *)(mtod(m, caddr_t) +
1386 	    sizeof(struct in_addr) + OPTSIZ);
1387 #undef OPTSIZ
1388 	/*
1389 	 * Record return path as an IP source route,
1390 	 * reversing the path (pointers are now aligned).
1391 	 */
1392 	while (p >= ip_srcrt.route) {
1393 #ifdef DIAGNOSTIC
1394 		if (ipprintfs)
1395 			printf(" %lx", (u_long)ntohl(q->s_addr));
1396 #endif
1397 		*q++ = *p--;
1398 	}
1399 	/*
1400 	 * Last hop goes to final destination.
1401 	 */
1402 	*q = ip_srcrt.dst;
1403 #ifdef DIAGNOSTIC
1404 	if (ipprintfs)
1405 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1406 #endif
1407 	return (m);
1408 }
1409 
1410 /*
1411  * Strip out IP options, at higher
1412  * level protocol in the kernel.
1413  * Second argument is buffer to which options
1414  * will be moved, and return value is their length.
1415  * XXX should be deleted; last arg currently ignored.
1416  */
1417 void
1418 ip_stripoptions(m, mopt)
1419 	register struct mbuf *m;
1420 	struct mbuf *mopt;
1421 {
1422 	register int i;
1423 	struct ip *ip = mtod(m, struct ip *);
1424 	register caddr_t opts;
1425 	int olen;
1426 
1427 	olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1428 	opts = (caddr_t)(ip + 1);
1429 	i = m->m_len - (sizeof (struct ip) + olen);
1430 	bcopy(opts + olen, opts, (unsigned)i);
1431 	m->m_len -= olen;
1432 	if (m->m_flags & M_PKTHDR)
1433 		m->m_pkthdr.len -= olen;
1434 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1435 }
1436 
1437 u_char inetctlerrmap[PRC_NCMDS] = {
1438 	0,		0,		0,		0,
1439 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1440 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1441 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1442 	0,		0,		0,		0,
1443 	ENOPROTOOPT
1444 };
1445 
1446 /*
1447  * Forward a packet.  If some error occurs return the sender
1448  * an icmp packet.  Note we can't always generate a meaningful
1449  * icmp message because icmp doesn't have a large enough repertoire
1450  * of codes and types.
1451  *
1452  * If not forwarding, just drop the packet.  This could be confusing
1453  * if ipforwarding was zero but some routing protocol was advancing
1454  * us as a gateway to somewhere.  However, we must let the routing
1455  * protocol deal with that.
1456  *
1457  * The srcrt parameter indicates whether the packet is being forwarded
1458  * via a source route.
1459  */
1460 static void
1461 ip_forward(m, srcrt)
1462 	struct mbuf *m;
1463 	int srcrt;
1464 {
1465 	register struct ip *ip = mtod(m, struct ip *);
1466 	register struct sockaddr_in *sin;
1467 	register struct rtentry *rt;
1468 	int error, type = 0, code = 0;
1469 	struct mbuf *mcopy;
1470 	n_long dest;
1471 	struct ifnet *destifp;
1472 #ifdef IPSEC
1473 	struct ifnet dummyifp;
1474 #endif
1475 
1476 	dest = 0;
1477 #ifdef DIAGNOSTIC
1478 	if (ipprintfs)
1479 		printf("forward: src %lx dst %lx ttl %x\n",
1480 		    (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1481 		    ip->ip_ttl);
1482 #endif
1483 
1484 
1485 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1486 		ipstat.ips_cantforward++;
1487 		m_freem(m);
1488 		return;
1489 	}
1490 #ifdef IPSTEALTH
1491 	if (!ipstealth) {
1492 #endif
1493 		if (ip->ip_ttl <= IPTTLDEC) {
1494 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1495 			    dest, 0);
1496 			return;
1497 		}
1498 #ifdef IPSTEALTH
1499 	}
1500 #endif
1501 
1502 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1503 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1504 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1505 		if (ipforward_rt.ro_rt) {
1506 			RTFREE(ipforward_rt.ro_rt);
1507 			ipforward_rt.ro_rt = 0;
1508 		}
1509 		sin->sin_family = AF_INET;
1510 		sin->sin_len = sizeof(*sin);
1511 		sin->sin_addr = ip->ip_dst;
1512 
1513 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1514 		if (ipforward_rt.ro_rt == 0) {
1515 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1516 			return;
1517 		}
1518 		rt = ipforward_rt.ro_rt;
1519 	}
1520 
1521 	/*
1522 	 * Save at most 64 bytes of the packet in case
1523 	 * we need to generate an ICMP message to the src.
1524 	 */
1525 	mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64));
1526 	if (mcopy && (mcopy->m_flags & M_EXT))
1527 		m_copydata(mcopy, 0, sizeof(struct ip), mtod(mcopy, caddr_t));
1528 
1529 #ifdef IPSTEALTH
1530 	if (!ipstealth) {
1531 #endif
1532 		ip->ip_ttl -= IPTTLDEC;
1533 #ifdef IPSTEALTH
1534 	}
1535 #endif
1536 
1537 	/*
1538 	 * If forwarding packet using same interface that it came in on,
1539 	 * perhaps should send a redirect to sender to shortcut a hop.
1540 	 * Only send redirect if source is sending directly to us,
1541 	 * and if packet was not source routed (or has any options).
1542 	 * Also, don't send redirect if forwarding using a default route
1543 	 * or a route modified by a redirect.
1544 	 */
1545 #define	satosin(sa)	((struct sockaddr_in *)(sa))
1546 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1547 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1548 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1549 	    ipsendredirects && !srcrt) {
1550 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1551 		u_long src = ntohl(ip->ip_src.s_addr);
1552 
1553 		if (RTA(rt) &&
1554 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1555 		    if (rt->rt_flags & RTF_GATEWAY)
1556 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1557 		    else
1558 			dest = ip->ip_dst.s_addr;
1559 		    /* Router requirements says to only send host redirects */
1560 		    type = ICMP_REDIRECT;
1561 		    code = ICMP_REDIRECT_HOST;
1562 #ifdef DIAGNOSTIC
1563 		    if (ipprintfs)
1564 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1565 #endif
1566 		}
1567 	}
1568 
1569 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1570 			  IP_FORWARDING, 0);
1571 	if (error)
1572 		ipstat.ips_cantforward++;
1573 	else {
1574 		ipstat.ips_forward++;
1575 		if (type)
1576 			ipstat.ips_redirectsent++;
1577 		else {
1578 			if (mcopy) {
1579 				ipflow_create(&ipforward_rt, mcopy);
1580 				m_freem(mcopy);
1581 			}
1582 			return;
1583 		}
1584 	}
1585 	if (mcopy == NULL)
1586 		return;
1587 	destifp = NULL;
1588 
1589 	switch (error) {
1590 
1591 	case 0:				/* forwarded, but need redirect */
1592 		/* type, code set above */
1593 		break;
1594 
1595 	case ENETUNREACH:		/* shouldn't happen, checked above */
1596 	case EHOSTUNREACH:
1597 	case ENETDOWN:
1598 	case EHOSTDOWN:
1599 	default:
1600 		type = ICMP_UNREACH;
1601 		code = ICMP_UNREACH_HOST;
1602 		break;
1603 
1604 	case EMSGSIZE:
1605 		type = ICMP_UNREACH;
1606 		code = ICMP_UNREACH_NEEDFRAG;
1607 #ifndef IPSEC
1608 		if (ipforward_rt.ro_rt)
1609 			destifp = ipforward_rt.ro_rt->rt_ifp;
1610 #else
1611 		/*
1612 		 * If the packet is routed over IPsec tunnel, tell the
1613 		 * originator the tunnel MTU.
1614 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1615 		 * XXX quickhack!!!
1616 		 */
1617 		if (ipforward_rt.ro_rt) {
1618 			struct secpolicy *sp = NULL;
1619 			int ipsecerror;
1620 			int ipsechdr;
1621 			struct route *ro;
1622 
1623 			sp = ipsec4_getpolicybyaddr(mcopy,
1624 						    IPSEC_DIR_OUTBOUND,
1625 			                            IP_FORWARDING,
1626 			                            &ipsecerror);
1627 
1628 			if (sp == NULL)
1629 				destifp = ipforward_rt.ro_rt->rt_ifp;
1630 			else {
1631 				/* count IPsec header size */
1632 				ipsechdr = ipsec4_hdrsiz(mcopy,
1633 							 IPSEC_DIR_OUTBOUND,
1634 							 NULL);
1635 
1636 				/*
1637 				 * find the correct route for outer IPv4
1638 				 * header, compute tunnel MTU.
1639 				 *
1640 				 * XXX BUG ALERT
1641 				 * The "dummyifp" code relies upon the fact
1642 				 * that icmp_error() touches only ifp->if_mtu.
1643 				 */
1644 				/*XXX*/
1645 				destifp = NULL;
1646 				if (sp->req != NULL
1647 				 && sp->req->sav != NULL
1648 				 && sp->req->sav->sah != NULL) {
1649 					ro = &sp->req->sav->sah->sa_route;
1650 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1651 						dummyifp.if_mtu =
1652 						    ro->ro_rt->rt_ifp->if_mtu;
1653 						dummyifp.if_mtu -= ipsechdr;
1654 						destifp = &dummyifp;
1655 					}
1656 				}
1657 
1658 				key_freesp(sp);
1659 			}
1660 		}
1661 #endif /*IPSEC*/
1662 		ipstat.ips_cantfrag++;
1663 		break;
1664 
1665 	case ENOBUFS:
1666 		type = ICMP_SOURCEQUENCH;
1667 		code = 0;
1668 		break;
1669 
1670 	case EACCES:			/* ipfw denied packet */
1671 		m_freem(mcopy);
1672 		return;
1673 	}
1674 	if (mcopy->m_flags & M_EXT)
1675 		m_copyback(mcopy, 0, sizeof(struct ip), mtod(mcopy, caddr_t));
1676 	icmp_error(mcopy, type, code, dest, destifp);
1677 }
1678 
1679 void
1680 ip_savecontrol(inp, mp, ip, m)
1681 	register struct inpcb *inp;
1682 	register struct mbuf **mp;
1683 	register struct ip *ip;
1684 	register struct mbuf *m;
1685 {
1686 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1687 		struct timeval tv;
1688 
1689 		microtime(&tv);
1690 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1691 			SCM_TIMESTAMP, SOL_SOCKET);
1692 		if (*mp)
1693 			mp = &(*mp)->m_next;
1694 	}
1695 	if (inp->inp_flags & INP_RECVDSTADDR) {
1696 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1697 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1698 		if (*mp)
1699 			mp = &(*mp)->m_next;
1700 	}
1701 #ifdef notyet
1702 	/* XXX
1703 	 * Moving these out of udp_input() made them even more broken
1704 	 * than they already were.
1705 	 */
1706 	/* options were tossed already */
1707 	if (inp->inp_flags & INP_RECVOPTS) {
1708 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1709 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1710 		if (*mp)
1711 			mp = &(*mp)->m_next;
1712 	}
1713 	/* ip_srcroute doesn't do what we want here, need to fix */
1714 	if (inp->inp_flags & INP_RECVRETOPTS) {
1715 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1716 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1717 		if (*mp)
1718 			mp = &(*mp)->m_next;
1719 	}
1720 #endif
1721 	if (inp->inp_flags & INP_RECVIF) {
1722 		struct ifnet *ifp;
1723 		struct sdlbuf {
1724 			struct sockaddr_dl sdl;
1725 			u_char	pad[32];
1726 		} sdlbuf;
1727 		struct sockaddr_dl *sdp;
1728 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1729 
1730 		if (((ifp = m->m_pkthdr.rcvif))
1731 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
1732 			sdp = (struct sockaddr_dl *)(ifnet_addrs
1733 					[ifp->if_index - 1]->ifa_addr);
1734 			/*
1735 			 * Change our mind and don't try copy.
1736 			 */
1737 			if ((sdp->sdl_family != AF_LINK)
1738 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
1739 				goto makedummy;
1740 			}
1741 			bcopy(sdp, sdl2, sdp->sdl_len);
1742 		} else {
1743 makedummy:
1744 			sdl2->sdl_len
1745 				= offsetof(struct sockaddr_dl, sdl_data[0]);
1746 			sdl2->sdl_family = AF_LINK;
1747 			sdl2->sdl_index = 0;
1748 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1749 		}
1750 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1751 			IP_RECVIF, IPPROTO_IP);
1752 		if (*mp)
1753 			mp = &(*mp)->m_next;
1754 	}
1755 }
1756 
1757 int
1758 ip_rsvp_init(struct socket *so)
1759 {
1760 	if (so->so_type != SOCK_RAW ||
1761 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1762 	  return EOPNOTSUPP;
1763 
1764 	if (ip_rsvpd != NULL)
1765 	  return EADDRINUSE;
1766 
1767 	ip_rsvpd = so;
1768 	/*
1769 	 * This may seem silly, but we need to be sure we don't over-increment
1770 	 * the RSVP counter, in case something slips up.
1771 	 */
1772 	if (!ip_rsvp_on) {
1773 		ip_rsvp_on = 1;
1774 		rsvp_on++;
1775 	}
1776 
1777 	return 0;
1778 }
1779 
1780 int
1781 ip_rsvp_done(void)
1782 {
1783 	ip_rsvpd = NULL;
1784 	/*
1785 	 * This may seem silly, but we need to be sure we don't over-decrement
1786 	 * the RSVP counter, in case something slips up.
1787 	 */
1788 	if (ip_rsvp_on) {
1789 		ip_rsvp_on = 0;
1790 		rsvp_on--;
1791 	}
1792 	return 0;
1793 }
1794