xref: /dragonfly/sys/netinet/ip_input.c (revision 9dbf638f)
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: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
35  * $DragonFly: src/sys/netinet/ip_input.c,v 1.7 2003/09/15 23:38:14 hsu Exp $
36  */
37 
38 #define	_IP_VHL
39 
40 #include "opt_bootp.h"
41 #include "opt_ipfw.h"
42 #include "opt_ipdn.h"
43 #include "opt_ipdivert.h"
44 #include "opt_ipfilter.h"
45 #include "opt_ipstealth.h"
46 #include "opt_ipsec.h"
47 #include "opt_random_ip_id.h"
48 
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/mbuf.h>
52 #include <sys/malloc.h>
53 #include <sys/domain.h>
54 #include <sys/protosw.h>
55 #include <sys/socket.h>
56 #include <sys/time.h>
57 #include <sys/kernel.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60 
61 #include <net/if.h>
62 #include <net/if_types.h>
63 #include <net/if_var.h>
64 #include <net/if_dl.h>
65 #include <net/route.h>
66 #include <net/netisr.h>
67 #include <net/intrq.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/in_var.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_icmp.h>
76 #include <machine/in_cksum.h>
77 
78 #include <netinet/ipprotosw.h>
79 
80 #include <sys/socketvar.h>
81 
82 #include <net/ipfw/ip_fw.h>
83 #include <net/dummynet/ip_dummynet.h>
84 
85 #ifdef IPSEC
86 #include <netinet6/ipsec.h>
87 #include <netproto/key/key.h>
88 #endif
89 
90 #ifdef FAST_IPSEC
91 #include <netipsec/ipsec.h>
92 #include <netipsec/key.h>
93 #endif
94 
95 int rsvp_on = 0;
96 static int ip_rsvp_on;
97 struct socket *ip_rsvpd;
98 
99 int	ipforwarding = 0;
100 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
101     &ipforwarding, 0, "Enable IP forwarding between interfaces");
102 
103 static int	ipsendredirects = 1; /* XXX */
104 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
105     &ipsendredirects, 0, "Enable sending IP redirects");
106 
107 int	ip_defttl = IPDEFTTL;
108 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
109     &ip_defttl, 0, "Maximum TTL on IP packets");
110 
111 static int	ip_dosourceroute = 0;
112 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
113     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
114 
115 static int	ip_acceptsourceroute = 0;
116 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
117     CTLFLAG_RW, &ip_acceptsourceroute, 0,
118     "Enable accepting source routed IP packets");
119 
120 static int	ip_keepfaith = 0;
121 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
122 	&ip_keepfaith,	0,
123 	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
124 
125 static int	nipq = 0;	/* total # of reass queues */
126 static int	maxnipq;
127 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
128 	&maxnipq, 0,
129 	"Maximum number of IPv4 fragment reassembly queue entries");
130 
131 static int    maxfragsperpacket;
132 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
133 	&maxfragsperpacket, 0,
134 	"Maximum number of IPv4 fragments allowed per packet");
135 
136 static int	ip_sendsourcequench = 0;
137 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
138 	&ip_sendsourcequench, 0,
139 	"Enable the transmission of source quench packets");
140 
141 /*
142  * XXX - Setting ip_checkinterface mostly implements the receive side of
143  * the Strong ES model described in RFC 1122, but since the routing table
144  * and transmit implementation do not implement the Strong ES model,
145  * setting this to 1 results in an odd hybrid.
146  *
147  * XXX - ip_checkinterface currently must be disabled if you use ipnat
148  * to translate the destination address to another local interface.
149  *
150  * XXX - ip_checkinterface must be disabled if you add IP aliases
151  * to the loopback interface instead of the interface where the
152  * packets for those addresses are received.
153  */
154 static int	ip_checkinterface = 0;
155 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
156     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
157 
158 #ifdef DIAGNOSTIC
159 static int	ipprintfs = 0;
160 #endif
161 
162 static struct ifqueue ipintrq;
163 static int	ipqmaxlen = IFQ_MAXLEN;
164 
165 extern	struct domain inetdomain;
166 extern	struct ipprotosw inetsw[];
167 u_char	ip_protox[IPPROTO_MAX];
168 struct	in_ifaddrhead in_ifaddrhead;		/* first inet address */
169 struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table */
170 u_long	in_ifaddrhmask;				/* mask for hash table */
171 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
172     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
173 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
174     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
175 
176 struct ipstat ipstat;
177 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
178     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
179 
180 /* Packet reassembly stuff */
181 #define IPREASS_NHASH_LOG2      6
182 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
183 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
184 #define IPREASS_HASH(x,y) \
185 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
186 
187 static struct ipq ipq[IPREASS_NHASH];
188 const  int    ipintrq_present = 1;
189 
190 #ifdef IPCTL_DEFMTU
191 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
192     &ip_mtu, 0, "Default MTU");
193 #endif
194 
195 #ifdef IPSTEALTH
196 static int	ipstealth = 0;
197 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
198     &ipstealth, 0, "");
199 #endif
200 
201 
202 /* Firewall hooks */
203 ip_fw_chk_t *ip_fw_chk_ptr;
204 int fw_enable = 1 ;
205 int fw_one_pass = 1;
206 
207 /* Dummynet hooks */
208 ip_dn_io_t *ip_dn_io_ptr;
209 
210 int (*fr_checkp) (struct ip *, int, struct ifnet *, int, struct mbuf **) = NULL;
211 
212 /*
213  * XXX this is ugly -- the following two global variables are
214  * used to store packet state while it travels through the stack.
215  * Note that the code even makes assumptions on the size and
216  * alignment of fields inside struct ip_srcrt so e.g. adding some
217  * fields will break the code. This needs to be fixed.
218  *
219  * We need to save the IP options in case a protocol wants to respond
220  * to an incoming packet over the same route if the packet got here
221  * using IP source routing.  This allows connection establishment and
222  * maintenance when the remote end is on a network that is not known
223  * to us.
224  */
225 static int	ip_nhops = 0;
226 static	struct ip_srcrt {
227 	struct	in_addr dst;			/* final destination */
228 	char	nop;				/* one NOP to align */
229 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
230 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
231 } ip_srcrt;
232 
233 static void	save_rte(u_char *, struct in_addr);
234 static int	ip_dooptions(struct mbuf *m, int,
235 			struct sockaddr_in *next_hop);
236 static void	ip_forward(struct mbuf *m, int srcrt,
237 			struct sockaddr_in *next_hop);
238 static void	ip_freef(struct ipq *);
239 static struct	mbuf *ip_reass(struct mbuf *, struct ipq *,
240 		struct ipq *, u_int32_t *, u_int16_t *);
241 
242 /*
243  * IP initialization: fill in IP protocol switch table.
244  * All protocols not implemented in kernel go to raw IP protocol handler.
245  */
246 void
247 ip_init()
248 {
249 	struct ipprotosw *pr;
250 	int i;
251 
252 	TAILQ_INIT(&in_ifaddrhead);
253 	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
254 	pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
255 	if (pr == 0)
256 		panic("ip_init");
257 	for (i = 0; i < IPPROTO_MAX; i++)
258 		ip_protox[i] = pr - inetsw;
259 	for (pr = (struct ipprotosw *)inetdomain.dom_protosw;
260 	    pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++)
261 		if (pr->pr_domain->dom_family == PF_INET &&
262 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
263 			ip_protox[pr->pr_protocol] = pr - inetsw;
264 
265 	for (i = 0; i < IPREASS_NHASH; i++)
266 	    ipq[i].next = ipq[i].prev = &ipq[i];
267 
268 	maxnipq = nmbclusters / 32;
269 	maxfragsperpacket = 16;
270 
271 #ifndef RANDOM_IP_ID
272 	ip_id = time_second & 0xffff;
273 #endif
274 	ipintrq.ifq_maxlen = ipqmaxlen;
275 
276 	netisr_register(NETISR_IP, ip_input, &ipintrq);
277 }
278 
279 /*
280  * XXX watch out this one. It is perhaps used as a cache for
281  * the most recently used route ? it is cleared in in_addroute()
282  * when a new route is successfully created.
283  */
284 struct	route ipforward_rt;
285 static struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
286 
287 /*
288  * Ip input routine.  Checksum and byte swap header.  If fragmented
289  * try to reassemble.  Process options.  Pass to next level.
290  */
291 void
292 ip_input(struct mbuf *m)
293 {
294 	struct ip *ip;
295 	struct ipq *fp;
296 	struct in_ifaddr *ia = NULL;
297 	struct ifaddr *ifa;
298 	int    i, hlen, checkif;
299 	u_short sum;
300 	struct in_addr pkt_dst;
301 	u_int32_t divert_info = 0;		/* packet divert/tee info */
302 	struct ip_fw_args args;
303 #ifdef FAST_IPSEC
304 	struct m_tag *mtag;
305 	struct tdb_ident *tdbi;
306 	struct secpolicy *sp;
307 	int s, error;
308 #endif /* FAST_IPSEC */
309 
310 	args.eh = NULL;
311 	args.oif = NULL;
312 	args.rule = NULL;
313 	args.divert_rule = 0;			/* divert cookie */
314 	args.next_hop = NULL;
315 
316 	/* Grab info from MT_TAG mbufs prepended to the chain.	*/
317 	for (; m && m->m_type == MT_TAG; m = m->m_next) {
318 		switch(m->_m_tag_id) {
319 		default:
320 			printf("ip_input: unrecognised MT_TAG tag %d\n",
321 			    m->_m_tag_id);
322 			break;
323 
324 		case PACKET_TAG_DUMMYNET:
325 			args.rule = ((struct dn_pkt *)m)->rule;
326 			break;
327 
328 		case PACKET_TAG_DIVERT:
329 			args.divert_rule = (int)m->m_hdr.mh_data & 0xffff;
330 			break;
331 
332 		case PACKET_TAG_IPFORWARD:
333 			args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
334 			break;
335 		}
336 	}
337 
338 	KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0,
339 	    ("ip_input: no HDR"));
340 
341 	if (args.rule) {	/* dummynet already filtered us */
342 		ip = mtod(m, struct ip *);
343 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
344 		goto iphack ;
345 	}
346 
347 	ipstat.ips_total++;
348 
349 	if (m->m_pkthdr.len < sizeof(struct ip))
350 		goto tooshort;
351 
352 	if (m->m_len < sizeof (struct ip) &&
353 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
354 		ipstat.ips_toosmall++;
355 		return;
356 	}
357 	ip = mtod(m, struct ip *);
358 
359 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
360 		ipstat.ips_badvers++;
361 		goto bad;
362 	}
363 
364 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
365 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
366 		ipstat.ips_badhlen++;
367 		goto bad;
368 	}
369 	if (hlen > m->m_len) {
370 		if ((m = m_pullup(m, hlen)) == 0) {
371 			ipstat.ips_badhlen++;
372 			return;
373 		}
374 		ip = mtod(m, struct ip *);
375 	}
376 
377 	/* 127/8 must not appear on wire - RFC1122 */
378 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
379 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
380 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
381 			ipstat.ips_badaddr++;
382 			goto bad;
383 		}
384 	}
385 
386 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
387 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
388 	} else {
389 		if (hlen == sizeof(struct ip)) {
390 			sum = in_cksum_hdr(ip);
391 		} else {
392 			sum = in_cksum(m, hlen);
393 		}
394 	}
395 	if (sum) {
396 		ipstat.ips_badsum++;
397 		goto bad;
398 	}
399 
400 	/*
401 	 * Convert fields to host representation.
402 	 */
403 	ip->ip_len = ntohs(ip->ip_len);
404 	if (ip->ip_len < hlen) {
405 		ipstat.ips_badlen++;
406 		goto bad;
407 	}
408 	ip->ip_off = ntohs(ip->ip_off);
409 
410 	/*
411 	 * Check that the amount of data in the buffers
412 	 * is as at least much as the IP header would have us expect.
413 	 * Trim mbufs if longer than we expect.
414 	 * Drop packet if shorter than we expect.
415 	 */
416 	if (m->m_pkthdr.len < ip->ip_len) {
417 tooshort:
418 		ipstat.ips_tooshort++;
419 		goto bad;
420 	}
421 	if (m->m_pkthdr.len > ip->ip_len) {
422 		if (m->m_len == m->m_pkthdr.len) {
423 			m->m_len = ip->ip_len;
424 			m->m_pkthdr.len = ip->ip_len;
425 		} else
426 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
427 	}
428 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
429 	/*
430 	 * Bypass packet filtering for packets from a tunnel (gif).
431 	 */
432 	if (ipsec_gethist(m, NULL))
433 		goto pass;
434 #endif
435 
436 	/*
437 	 * IpHack's section.
438 	 * Right now when no processing on packet has done
439 	 * and it is still fresh out of network we do our black
440 	 * deals with it.
441 	 * - Firewall: deny/allow/divert
442 	 * - Xlate: translate packet's addr/port (NAT).
443 	 * - Pipe: pass pkt through dummynet.
444 	 * - Wrap: fake packet's addr/port <unimpl.>
445 	 * - Encapsulate: put it in another IP and send out. <unimp.>
446  	 */
447 
448 iphack:
449 	/*
450 	 * Check if we want to allow this packet to be processed.
451 	 * Consider it to be bad if not.
452 	 */
453 	if (fr_checkp) {
454 		struct	mbuf	*m1 = m;
455 
456 		if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1)
457 			return;
458 		ip = mtod(m = m1, struct ip *);
459 	}
460 	if (fw_enable && IPFW_LOADED) {
461 		/*
462 		 * If we've been forwarded from the output side, then
463 		 * skip the firewall a second time
464 		 */
465 		if (args.next_hop)
466 			goto ours;
467 
468 		args.m = m;
469 		i = ip_fw_chk_ptr(&args);
470 		m = args.m;
471 
472 		if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
473 			if (m)
474 				m_freem(m);
475 			return;
476 		}
477 		ip = mtod(m, struct ip *); /* just in case m changed */
478 		if (i == 0 && args.next_hop == NULL)	/* common case */
479 			goto pass;
480                 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
481 			/* Send packet to the appropriate pipe */
482 			ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
483 			return;
484 		}
485 #ifdef IPDIVERT
486 		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
487 			/* Divert or tee packet */
488 			divert_info = i;
489 			goto ours;
490 		}
491 #endif
492 		if (i == 0 && args.next_hop != NULL)
493 			goto pass;
494 		/*
495 		 * if we get here, the packet must be dropped
496 		 */
497 		m_freem(m);
498 		return;
499 	}
500 pass:
501 
502 	/*
503 	 * Process options and, if not destined for us,
504 	 * ship it on.  ip_dooptions returns 1 when an
505 	 * error was detected (causing an icmp message
506 	 * to be sent and the original packet to be freed).
507 	 */
508 	ip_nhops = 0;		/* for source routed packets */
509 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
510 		return;
511 
512         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
513          * matter if it is destined to another node, or whether it is
514          * a multicast one, RSVP wants it! and prevents it from being forwarded
515          * anywhere else. Also checks if the rsvp daemon is running before
516 	 * grabbing the packet.
517          */
518 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
519 		goto ours;
520 
521 	/*
522 	 * Check our list of addresses, to see if the packet is for us.
523 	 * If we don't have any addresses, assume any unicast packet
524 	 * we receive might be for us (and let the upper layers deal
525 	 * with it).
526 	 */
527 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
528 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
529 		goto ours;
530 
531 	/*
532 	 * Cache the destination address of the packet; this may be
533 	 * changed by use of 'ipfw fwd'.
534 	 */
535 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
536 
537 	/*
538 	 * Enable a consistency check between the destination address
539 	 * and the arrival interface for a unicast packet (the RFC 1122
540 	 * strong ES model) if IP forwarding is disabled and the packet
541 	 * is not locally generated and the packet is not subject to
542 	 * 'ipfw fwd'.
543 	 *
544 	 * XXX - Checking also should be disabled if the destination
545 	 * address is ipnat'ed to a different interface.
546 	 *
547 	 * XXX - Checking is incompatible with IP aliases added
548 	 * to the loopback interface instead of the interface where
549 	 * the packets are received.
550 	 */
551 	checkif = ip_checkinterface && (ipforwarding == 0) &&
552 	    m->m_pkthdr.rcvif != NULL &&
553 	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
554 	    (args.next_hop == NULL);
555 
556 	/*
557 	 * Check for exact addresses in the hash bucket.
558 	 */
559 	LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
560 		/*
561 		 * If the address matches, verify that the packet
562 		 * arrived via the correct interface if checking is
563 		 * enabled.
564 		 */
565 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
566 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
567 			goto ours;
568 	}
569 	/*
570 	 * Check for broadcast addresses.
571 	 *
572 	 * Only accept broadcast packets that arrive via the matching
573 	 * interface.  Reception of forwarded directed broadcasts would
574 	 * be handled via ip_forward() and ether_output() with the loopback
575 	 * into the stack for SIMPLEX interfaces handled by ether_output().
576 	 */
577 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
578 	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
579 			if (ifa->ifa_addr->sa_family != AF_INET)
580 				continue;
581 			ia = ifatoia(ifa);
582 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
583 			    pkt_dst.s_addr)
584 				goto ours;
585 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
586 				goto ours;
587 #ifdef BOOTP_COMPAT
588 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
589 				goto ours;
590 #endif
591 		}
592 	}
593 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
594 		struct in_multi *inm;
595 		if (ip_mrouter) {
596 			/*
597 			 * If we are acting as a multicast router, all
598 			 * incoming multicast packets are passed to the
599 			 * kernel-level multicast forwarding function.
600 			 * The packet is returned (relatively) intact; if
601 			 * ip_mforward() returns a non-zero value, the packet
602 			 * must be discarded, else it may be accepted below.
603 			 */
604 			if (ip_mforward &&
605 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
606 				ipstat.ips_cantforward++;
607 				m_freem(m);
608 				return;
609 			}
610 
611 			/*
612 			 * The process-level routing daemon needs to receive
613 			 * all multicast IGMP packets, whether or not this
614 			 * host belongs to their destination groups.
615 			 */
616 			if (ip->ip_p == IPPROTO_IGMP)
617 				goto ours;
618 			ipstat.ips_forward++;
619 		}
620 		/*
621 		 * See if we belong to the destination multicast group on the
622 		 * arrival interface.
623 		 */
624 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
625 		if (inm == NULL) {
626 			ipstat.ips_notmember++;
627 			m_freem(m);
628 			return;
629 		}
630 		goto ours;
631 	}
632 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
633 		goto ours;
634 	if (ip->ip_dst.s_addr == INADDR_ANY)
635 		goto ours;
636 
637 	/*
638 	 * FAITH(Firewall Aided Internet Translator)
639 	 */
640 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
641 		if (ip_keepfaith) {
642 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
643 				goto ours;
644 		}
645 		m_freem(m);
646 		return;
647 	}
648 
649 	/*
650 	 * Not for us; forward if possible and desirable.
651 	 */
652 	if (ipforwarding == 0) {
653 		ipstat.ips_cantforward++;
654 		m_freem(m);
655 	} else {
656 #ifdef IPSEC
657 		/*
658 		 * Enforce inbound IPsec SPD.
659 		 */
660 		if (ipsec4_in_reject(m, NULL)) {
661 			ipsecstat.in_polvio++;
662 			goto bad;
663 		}
664 #endif /* IPSEC */
665 #ifdef FAST_IPSEC
666 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
667 		s = splnet();
668 		if (mtag != NULL) {
669 			tdbi = (struct tdb_ident *)(mtag + 1);
670 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
671 		} else {
672 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
673 						   IP_FORWARDING, &error);
674 		}
675 		if (sp == NULL) {	/* NB: can happen if error */
676 			splx(s);
677 			/*XXX error stat???*/
678 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
679 			goto bad;
680 		}
681 
682 		/*
683 		 * Check security policy against packet attributes.
684 		 */
685 		error = ipsec_in_reject(sp, m);
686 		KEY_FREESP(&sp);
687 		splx(s);
688 		if (error) {
689 			ipstat.ips_cantforward++;
690 			goto bad;
691 		}
692 #endif /* FAST_IPSEC */
693 		ip_forward(m, 0, args.next_hop);
694 	}
695 	return;
696 
697 ours:
698 #ifdef IPSTEALTH
699 	/*
700 	 * IPSTEALTH: Process non-routing options only
701 	 * if the packet is destined for us.
702 	 */
703 	if (ipstealth && hlen > sizeof (struct ip) &&
704 	    ip_dooptions(m, 1, args.next_hop))
705 		return;
706 #endif /* IPSTEALTH */
707 
708 	/* Count the packet in the ip address stats */
709 	if (ia != NULL) {
710 		ia->ia_ifa.if_ipackets++;
711 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
712 	}
713 
714 	/*
715 	 * If offset or IP_MF are set, must reassemble.
716 	 * Otherwise, nothing need be done.
717 	 * (We could look in the reassembly queue to see
718 	 * if the packet was previously fragmented,
719 	 * but it's not worth the time; just let them time out.)
720 	 */
721 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
722 
723 		/* If maxnipq is 0, never accept fragments. */
724 		if (maxnipq == 0) {
725                 	ipstat.ips_fragments++;
726 			ipstat.ips_fragdropped++;
727 			goto bad;
728 		}
729 
730 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
731 		/*
732 		 * Look for queue of fragments
733 		 * of this datagram.
734 		 */
735 		for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
736 			if (ip->ip_id == fp->ipq_id &&
737 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
738 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
739 			    ip->ip_p == fp->ipq_p)
740 				goto found;
741 
742 		fp = 0;
743 
744 		/*
745 		 * Enforce upper bound on number of fragmented packets
746 		 * for which we attempt reassembly;
747 		 * If maxnipq is -1, accept all fragments without limitation.
748 		 */
749 		if ((nipq > maxnipq) && (maxnipq > 0)) {
750 		    /*
751 		     * drop something from the tail of the current queue
752 		     * before proceeding further
753 		     */
754 		    if (ipq[sum].prev == &ipq[sum]) {   /* gak */
755 			for (i = 0; i < IPREASS_NHASH; i++) {
756 			    if (ipq[i].prev != &ipq[i]) {
757 				ipstat.ips_fragtimeout +=
758 				    ipq[i].prev->ipq_nfrags;
759 				ip_freef(ipq[i].prev);
760 				break;
761 			    }
762 			}
763 		    } else {
764 			ipstat.ips_fragtimeout += ipq[sum].prev->ipq_nfrags;
765 			ip_freef(ipq[sum].prev);
766 		    }
767 		}
768 found:
769 		/*
770 		 * Adjust ip_len to not reflect header,
771 		 * convert offset of this to bytes.
772 		 */
773 		ip->ip_len -= hlen;
774 		if (ip->ip_off & IP_MF) {
775 		        /*
776 		         * Make sure that fragments have a data length
777 			 * that's a non-zero multiple of 8 bytes.
778 		         */
779 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
780 				ipstat.ips_toosmall++; /* XXX */
781 				goto bad;
782 			}
783 			m->m_flags |= M_FRAG;
784 		} else
785 			m->m_flags &= ~M_FRAG;
786 		ip->ip_off <<= 3;
787 
788 		/*
789 		 * Attempt reassembly; if it succeeds, proceed.
790 		 * ip_reass() will return a different mbuf, and update
791 		 * the divert info in divert_info and args.divert_rule.
792 		 */
793 		ipstat.ips_fragments++;
794 		m->m_pkthdr.header = ip;
795 		m = ip_reass(m,
796 		    fp, &ipq[sum], &divert_info, &args.divert_rule);
797 		if (m == 0)
798 			return;
799 		ipstat.ips_reassembled++;
800 		ip = mtod(m, struct ip *);
801 		/* Get the header length of the reassembled packet */
802 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
803 #ifdef IPDIVERT
804 		/* Restore original checksum before diverting packet */
805 		if (divert_info != 0) {
806 			ip->ip_len += hlen;
807 			ip->ip_len = htons(ip->ip_len);
808 			ip->ip_off = htons(ip->ip_off);
809 			ip->ip_sum = 0;
810 			if (hlen == sizeof(struct ip))
811 				ip->ip_sum = in_cksum_hdr(ip);
812 			else
813 				ip->ip_sum = in_cksum(m, hlen);
814 			ip->ip_off = ntohs(ip->ip_off);
815 			ip->ip_len = ntohs(ip->ip_len);
816 			ip->ip_len -= hlen;
817 		}
818 #endif
819 	} else
820 		ip->ip_len -= hlen;
821 
822 #ifdef IPDIVERT
823 	/*
824 	 * Divert or tee packet to the divert protocol if required.
825 	 */
826 	if (divert_info != 0) {
827 		struct mbuf *clone = NULL;
828 
829 		/* Clone packet if we're doing a 'tee' */
830 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
831 			clone = m_dup(m, M_DONTWAIT);
832 
833 		/* Restore packet header fields to original values */
834 		ip->ip_len += hlen;
835 		ip->ip_len = htons(ip->ip_len);
836 		ip->ip_off = htons(ip->ip_off);
837 
838 		/* Deliver packet to divert input routine */
839 		divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
840 		ipstat.ips_delivered++;
841 
842 		/* If 'tee', continue with original packet */
843 		if (clone == NULL)
844 			return;
845 		m = clone;
846 		ip = mtod(m, struct ip *);
847 		ip->ip_len += hlen;
848 		/*
849 		 * Jump backwards to complete processing of the
850 		 * packet. But first clear divert_info to avoid
851 		 * entering this block again.
852 		 * We do not need to clear args.divert_rule
853 		 * or args.next_hop as they will not be used.
854 		 */
855 		divert_info = 0;
856 		goto pass;
857 	}
858 #endif
859 
860 #ifdef IPSEC
861 	/*
862 	 * enforce IPsec policy checking if we are seeing last header.
863 	 * note that we do not visit this with protocols with pcb layer
864 	 * code - like udp/tcp/raw ip.
865 	 */
866 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
867 	    ipsec4_in_reject(m, NULL)) {
868 		ipsecstat.in_polvio++;
869 		goto bad;
870 	}
871 #endif
872 #if FAST_IPSEC
873 	/*
874 	 * enforce IPsec policy checking if we are seeing last header.
875 	 * note that we do not visit this with protocols with pcb layer
876 	 * code - like udp/tcp/raw ip.
877 	 */
878 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
879 		/*
880 		 * Check if the packet has already had IPsec processing
881 		 * done.  If so, then just pass it along.  This tag gets
882 		 * set during AH, ESP, etc. input handling, before the
883 		 * packet is returned to the ip input queue for delivery.
884 		 */
885 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
886 		s = splnet();
887 		if (mtag != NULL) {
888 			tdbi = (struct tdb_ident *)(mtag + 1);
889 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
890 		} else {
891 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
892 						   IP_FORWARDING, &error);
893 		}
894 		if (sp != NULL) {
895 			/*
896 			 * Check security policy against packet attributes.
897 			 */
898 			error = ipsec_in_reject(sp, m);
899 			KEY_FREESP(&sp);
900 		} else {
901 			/* XXX error stat??? */
902 			error = EINVAL;
903 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
904 			goto bad;
905 		}
906 		splx(s);
907 		if (error)
908 			goto bad;
909 	}
910 #endif /* FAST_IPSEC */
911 
912 	/*
913 	 * Switch out to protocol's input routine.
914 	 */
915 	ipstat.ips_delivered++;
916 	if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
917 		/* TCP needs IPFORWARD info if available */
918 		struct m_hdr tag;
919 
920 		tag.mh_type = MT_TAG;
921 		tag.mh_flags = PACKET_TAG_IPFORWARD;
922 		tag.mh_data = (caddr_t)args.next_hop;
923 		tag.mh_next = m;
924 
925 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(
926 			(struct mbuf *)&tag, hlen, ip->ip_p);
927     	} else
928 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
929 	return;
930 bad:
931 	m_freem(m);
932 }
933 
934 /*
935  * Take incoming datagram fragment and try to reassemble it into
936  * whole datagram.  If a chain for reassembly of this datagram already
937  * exists, then it is given as fp; otherwise have to make a chain.
938  *
939  * When IPDIVERT enabled, keep additional state with each packet that
940  * tells us if we need to divert or tee the packet we're building.
941  * In particular, *divinfo includes the port and TEE flag,
942  * *divert_rule is the number of the matching rule.
943  */
944 
945 static struct mbuf *
946 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
947 	u_int32_t *divinfo, u_int16_t *divert_rule)
948 {
949 	struct ip *ip = mtod(m, struct ip *);
950 	struct mbuf *p = 0, *q, *nq;
951 	struct mbuf *t;
952 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
953 	int i, next;
954 
955 	/*
956 	 * Presence of header sizes in mbufs
957 	 * would confuse code below.
958 	 */
959 	m->m_data += hlen;
960 	m->m_len -= hlen;
961 
962 	/*
963 	 * If first fragment to arrive, create a reassembly queue.
964 	 */
965 	if (fp == 0) {
966 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
967 			goto dropfrag;
968 		fp = mtod(t, struct ipq *);
969 		insque(fp, where);
970 		nipq++;
971 		fp->ipq_nfrags = 1;
972 		fp->ipq_ttl = IPFRAGTTL;
973 		fp->ipq_p = ip->ip_p;
974 		fp->ipq_id = ip->ip_id;
975 		fp->ipq_src = ip->ip_src;
976 		fp->ipq_dst = ip->ip_dst;
977 		fp->ipq_frags = m;
978 		m->m_nextpkt = NULL;
979 #ifdef IPDIVERT
980 		fp->ipq_div_info = 0;
981 		fp->ipq_div_cookie = 0;
982 #endif
983 		goto inserted;
984 	} else {
985 		fp->ipq_nfrags++;
986 	}
987 
988 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
989 
990 	/*
991 	 * Find a segment which begins after this one does.
992 	 */
993 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
994 		if (GETIP(q)->ip_off > ip->ip_off)
995 			break;
996 
997 	/*
998 	 * If there is a preceding segment, it may provide some of
999 	 * our data already.  If so, drop the data from the incoming
1000 	 * segment.  If it provides all of our data, drop us, otherwise
1001 	 * stick new segment in the proper place.
1002 	 *
1003 	 * If some of the data is dropped from the the preceding
1004 	 * segment, then it's checksum is invalidated.
1005 	 */
1006 	if (p) {
1007 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1008 		if (i > 0) {
1009 			if (i >= ip->ip_len)
1010 				goto dropfrag;
1011 			m_adj(m, i);
1012 			m->m_pkthdr.csum_flags = 0;
1013 			ip->ip_off += i;
1014 			ip->ip_len -= i;
1015 		}
1016 		m->m_nextpkt = p->m_nextpkt;
1017 		p->m_nextpkt = m;
1018 	} else {
1019 		m->m_nextpkt = fp->ipq_frags;
1020 		fp->ipq_frags = m;
1021 	}
1022 
1023 	/*
1024 	 * While we overlap succeeding segments trim them or,
1025 	 * if they are completely covered, dequeue them.
1026 	 */
1027 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1028 	     q = nq) {
1029 		i = (ip->ip_off + ip->ip_len) -
1030 		    GETIP(q)->ip_off;
1031 		if (i < GETIP(q)->ip_len) {
1032 			GETIP(q)->ip_len -= i;
1033 			GETIP(q)->ip_off += i;
1034 			m_adj(q, i);
1035 			q->m_pkthdr.csum_flags = 0;
1036 			break;
1037 		}
1038 		nq = q->m_nextpkt;
1039 		m->m_nextpkt = nq;
1040 		ipstat.ips_fragdropped++;
1041 		fp->ipq_nfrags--;
1042 		m_freem(q);
1043 	}
1044 
1045 inserted:
1046 
1047 #ifdef IPDIVERT
1048 	/*
1049 	 * Transfer firewall instructions to the fragment structure.
1050 	 * Only trust info in the fragment at offset 0.
1051 	 */
1052 	if (ip->ip_off == 0) {
1053 		fp->ipq_div_info = *divinfo;
1054 		fp->ipq_div_cookie = *divert_rule;
1055 	}
1056 	*divinfo = 0;
1057 	*divert_rule = 0;
1058 #endif
1059 
1060 	/*
1061 	 * Check for complete reassembly and perform frag per packet
1062 	 * limiting.
1063 	 *
1064 	 * Frag limiting is performed here so that the nth frag has
1065 	 * a chance to complete the packet before we drop the packet.
1066 	 * As a result, n+1 frags are actually allowed per packet, but
1067 	 * only n will ever be stored. (n = maxfragsperpacket.)
1068 	 *
1069 	 */
1070 	next = 0;
1071 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1072 		if (GETIP(q)->ip_off != next) {
1073 			if (fp->ipq_nfrags > maxfragsperpacket) {
1074 				ipstat.ips_fragdropped += fp->ipq_nfrags;
1075 				ip_freef(fp);
1076 			}
1077 			return (0);
1078 		}
1079 		next += GETIP(q)->ip_len;
1080 	}
1081 	/* Make sure the last packet didn't have the IP_MF flag */
1082 	if (p->m_flags & M_FRAG) {
1083 		if (fp->ipq_nfrags > maxfragsperpacket) {
1084 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1085 			ip_freef(fp);
1086 		}
1087 		return (0);
1088 	}
1089 
1090 	/*
1091 	 * Reassembly is complete.  Make sure the packet is a sane size.
1092 	 */
1093 	q = fp->ipq_frags;
1094 	ip = GETIP(q);
1095 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1096 		ipstat.ips_toolong++;
1097 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1098 		ip_freef(fp);
1099 		return (0);
1100 	}
1101 
1102 	/*
1103 	 * Concatenate fragments.
1104 	 */
1105 	m = q;
1106 	t = m->m_next;
1107 	m->m_next = 0;
1108 	m_cat(m, t);
1109 	nq = q->m_nextpkt;
1110 	q->m_nextpkt = 0;
1111 	for (q = nq; q != NULL; q = nq) {
1112 		nq = q->m_nextpkt;
1113 		q->m_nextpkt = NULL;
1114 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1115 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1116 		m_cat(m, q);
1117 	}
1118 
1119 #ifdef IPDIVERT
1120 	/*
1121 	 * Extract firewall instructions from the fragment structure.
1122 	 */
1123 	*divinfo = fp->ipq_div_info;
1124 	*divert_rule = fp->ipq_div_cookie;
1125 #endif
1126 
1127 	/*
1128 	 * Create header for new ip packet by
1129 	 * modifying header of first packet;
1130 	 * dequeue and discard fragment reassembly header.
1131 	 * Make header visible.
1132 	 */
1133 	ip->ip_len = next;
1134 	ip->ip_src = fp->ipq_src;
1135 	ip->ip_dst = fp->ipq_dst;
1136 	remque(fp);
1137 	nipq--;
1138 	(void) m_free(dtom(fp));
1139 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1140 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1141 	/* some debugging cruft by sklower, below, will go away soon */
1142 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1143 		int plen = 0;
1144 		for (t = m; t; t = t->m_next)
1145 			plen += t->m_len;
1146 		m->m_pkthdr.len = plen;
1147 	}
1148 	return (m);
1149 
1150 dropfrag:
1151 #ifdef IPDIVERT
1152 	*divinfo = 0;
1153 	*divert_rule = 0;
1154 #endif
1155 	ipstat.ips_fragdropped++;
1156 	if (fp != 0)
1157 		fp->ipq_nfrags--;
1158 	m_freem(m);
1159 	return (0);
1160 
1161 #undef GETIP
1162 }
1163 
1164 /*
1165  * Free a fragment reassembly header and all
1166  * associated datagrams.
1167  */
1168 static void
1169 ip_freef(fp)
1170 	struct ipq *fp;
1171 {
1172 	struct mbuf *q;
1173 
1174 	while (fp->ipq_frags) {
1175 		q = fp->ipq_frags;
1176 		fp->ipq_frags = q->m_nextpkt;
1177 		m_freem(q);
1178 	}
1179 	remque(fp);
1180 	(void) m_free(dtom(fp));
1181 	nipq--;
1182 }
1183 
1184 /*
1185  * IP timer processing;
1186  * if a timer expires on a reassembly
1187  * queue, discard it.
1188  */
1189 void
1190 ip_slowtimo()
1191 {
1192 	struct ipq *fp;
1193 	int s = splnet();
1194 	int i;
1195 
1196 	for (i = 0; i < IPREASS_NHASH; i++) {
1197 		fp = ipq[i].next;
1198 		if (fp == 0)
1199 			continue;
1200 		while (fp != &ipq[i]) {
1201 			--fp->ipq_ttl;
1202 			fp = fp->next;
1203 			if (fp->prev->ipq_ttl == 0) {
1204 				ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1205 				ip_freef(fp->prev);
1206 			}
1207 		}
1208 	}
1209 	/*
1210 	 * If we are over the maximum number of fragments
1211 	 * (due to the limit being lowered), drain off
1212 	 * enough to get down to the new limit.
1213 	 */
1214 	if (maxnipq >= 0 && nipq > maxnipq) {
1215 		for (i = 0; i < IPREASS_NHASH; i++) {
1216 			while (nipq > maxnipq &&
1217 				(ipq[i].next != &ipq[i])) {
1218 				ipstat.ips_fragdropped +=
1219 				    ipq[i].next->ipq_nfrags;
1220 				ip_freef(ipq[i].next);
1221 			}
1222 		}
1223 	}
1224 	ipflow_slowtimo();
1225 	splx(s);
1226 }
1227 
1228 /*
1229  * Drain off all datagram fragments.
1230  */
1231 void
1232 ip_drain()
1233 {
1234 	int     i;
1235 
1236 	for (i = 0; i < IPREASS_NHASH; i++) {
1237 		while (ipq[i].next != &ipq[i]) {
1238 			ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1239 			ip_freef(ipq[i].next);
1240 		}
1241 	}
1242 	in_rtqdrain();
1243 }
1244 
1245 /*
1246  * Do option processing on a datagram,
1247  * possibly discarding it if bad options are encountered,
1248  * or forwarding it if source-routed.
1249  * The pass argument is used when operating in the IPSTEALTH
1250  * mode to tell what options to process:
1251  * [LS]SRR (pass 0) or the others (pass 1).
1252  * The reason for as many as two passes is that when doing IPSTEALTH,
1253  * non-routing options should be processed only if the packet is for us.
1254  * Returns 1 if packet has been forwarded/freed,
1255  * 0 if the packet should be processed further.
1256  */
1257 static int
1258 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1259 {
1260 	struct ip *ip = mtod(m, struct ip *);
1261 	u_char *cp;
1262 	struct in_ifaddr *ia;
1263 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1264 	struct in_addr *sin, dst;
1265 	n_time ntime;
1266 
1267 	dst = ip->ip_dst;
1268 	cp = (u_char *)(ip + 1);
1269 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1270 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1271 		opt = cp[IPOPT_OPTVAL];
1272 		if (opt == IPOPT_EOL)
1273 			break;
1274 		if (opt == IPOPT_NOP)
1275 			optlen = 1;
1276 		else {
1277 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1278 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1279 				goto bad;
1280 			}
1281 			optlen = cp[IPOPT_OLEN];
1282 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1283 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1284 				goto bad;
1285 			}
1286 		}
1287 		switch (opt) {
1288 
1289 		default:
1290 			break;
1291 
1292 		/*
1293 		 * Source routing with record.
1294 		 * Find interface with current destination address.
1295 		 * If none on this machine then drop if strictly routed,
1296 		 * or do nothing if loosely routed.
1297 		 * Record interface address and bring up next address
1298 		 * component.  If strictly routed make sure next
1299 		 * address is on directly accessible net.
1300 		 */
1301 		case IPOPT_LSRR:
1302 		case IPOPT_SSRR:
1303 #ifdef IPSTEALTH
1304 			if (ipstealth && pass > 0)
1305 				break;
1306 #endif
1307 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1308 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1309 				goto bad;
1310 			}
1311 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1312 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1313 				goto bad;
1314 			}
1315 			ipaddr.sin_addr = ip->ip_dst;
1316 			ia = (struct in_ifaddr *)
1317 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1318 			if (ia == 0) {
1319 				if (opt == IPOPT_SSRR) {
1320 					type = ICMP_UNREACH;
1321 					code = ICMP_UNREACH_SRCFAIL;
1322 					goto bad;
1323 				}
1324 				if (!ip_dosourceroute)
1325 					goto nosourcerouting;
1326 				/*
1327 				 * Loose routing, and not at next destination
1328 				 * yet; nothing to do except forward.
1329 				 */
1330 				break;
1331 			}
1332 			off--;			/* 0 origin */
1333 			if (off > optlen - (int)sizeof(struct in_addr)) {
1334 				/*
1335 				 * End of source route.  Should be for us.
1336 				 */
1337 				if (!ip_acceptsourceroute)
1338 					goto nosourcerouting;
1339 				save_rte(cp, ip->ip_src);
1340 				break;
1341 			}
1342 #ifdef IPSTEALTH
1343 			if (ipstealth)
1344 				goto dropit;
1345 #endif
1346 			if (!ip_dosourceroute) {
1347 				if (ipforwarding) {
1348 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1349 					/*
1350 					 * Acting as a router, so generate ICMP
1351 					 */
1352 nosourcerouting:
1353 					strcpy(buf, inet_ntoa(ip->ip_dst));
1354 					log(LOG_WARNING,
1355 					    "attempted source route from %s to %s\n",
1356 					    inet_ntoa(ip->ip_src), buf);
1357 					type = ICMP_UNREACH;
1358 					code = ICMP_UNREACH_SRCFAIL;
1359 					goto bad;
1360 				} else {
1361 					/*
1362 					 * Not acting as a router, so silently drop.
1363 					 */
1364 #ifdef IPSTEALTH
1365 dropit:
1366 #endif
1367 					ipstat.ips_cantforward++;
1368 					m_freem(m);
1369 					return (1);
1370 				}
1371 			}
1372 
1373 			/*
1374 			 * locate outgoing interface
1375 			 */
1376 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1377 			    sizeof(ipaddr.sin_addr));
1378 
1379 			if (opt == IPOPT_SSRR) {
1380 #define	INA	struct in_ifaddr *
1381 #define	SA	struct sockaddr *
1382 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1383 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1384 			} else
1385 				ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1386 			if (ia == 0) {
1387 				type = ICMP_UNREACH;
1388 				code = ICMP_UNREACH_SRCFAIL;
1389 				goto bad;
1390 			}
1391 			ip->ip_dst = ipaddr.sin_addr;
1392 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1393 			    sizeof(struct in_addr));
1394 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1395 			/*
1396 			 * Let ip_intr's mcast routing check handle mcast pkts
1397 			 */
1398 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1399 			break;
1400 
1401 		case IPOPT_RR:
1402 #ifdef IPSTEALTH
1403 			if (ipstealth && pass == 0)
1404 				break;
1405 #endif
1406 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1407 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1408 				goto bad;
1409 			}
1410 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1411 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1412 				goto bad;
1413 			}
1414 			/*
1415 			 * If no space remains, ignore.
1416 			 */
1417 			off--;			/* 0 origin */
1418 			if (off > optlen - (int)sizeof(struct in_addr))
1419 				break;
1420 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1421 			    sizeof(ipaddr.sin_addr));
1422 			/*
1423 			 * locate outgoing interface; if we're the destination,
1424 			 * use the incoming interface (should be same).
1425 			 */
1426 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1427 			    (ia = ip_rtaddr(ipaddr.sin_addr,
1428 			    &ipforward_rt)) == 0) {
1429 				type = ICMP_UNREACH;
1430 				code = ICMP_UNREACH_HOST;
1431 				goto bad;
1432 			}
1433 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1434 			    sizeof(struct in_addr));
1435 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1436 			break;
1437 
1438 		case IPOPT_TS:
1439 #ifdef IPSTEALTH
1440 			if (ipstealth && pass == 0)
1441 				break;
1442 #endif
1443 			code = cp - (u_char *)ip;
1444 			if (optlen < 4 || optlen > 40) {
1445 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1446 				goto bad;
1447 			}
1448 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1449 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1450 				goto bad;
1451 			}
1452 			if (off > optlen - (int)sizeof(int32_t)) {
1453 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1454 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1455 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1456 					goto bad;
1457 				}
1458 				break;
1459 			}
1460 			off--;				/* 0 origin */
1461 			sin = (struct in_addr *)(cp + off);
1462 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1463 
1464 			case IPOPT_TS_TSONLY:
1465 				break;
1466 
1467 			case IPOPT_TS_TSANDADDR:
1468 				if (off + sizeof(n_time) +
1469 				    sizeof(struct in_addr) > optlen) {
1470 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1471 					goto bad;
1472 				}
1473 				ipaddr.sin_addr = dst;
1474 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1475 							    m->m_pkthdr.rcvif);
1476 				if (ia == 0)
1477 					continue;
1478 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1479 				    sizeof(struct in_addr));
1480 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1481 				off += sizeof(struct in_addr);
1482 				break;
1483 
1484 			case IPOPT_TS_PRESPEC:
1485 				if (off + sizeof(n_time) +
1486 				    sizeof(struct in_addr) > optlen) {
1487 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1488 					goto bad;
1489 				}
1490 				(void)memcpy(&ipaddr.sin_addr, sin,
1491 				    sizeof(struct in_addr));
1492 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1493 					continue;
1494 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1495 				off += sizeof(struct in_addr);
1496 				break;
1497 
1498 			default:
1499 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1500 				goto bad;
1501 			}
1502 			ntime = iptime();
1503 			(void)memcpy(cp + off, &ntime, sizeof(n_time));
1504 			cp[IPOPT_OFFSET] += sizeof(n_time);
1505 		}
1506 	}
1507 	if (forward && ipforwarding) {
1508 		ip_forward(m, 1, next_hop);
1509 		return (1);
1510 	}
1511 	return (0);
1512 bad:
1513 	icmp_error(m, type, code, 0, 0);
1514 	ipstat.ips_badoptions++;
1515 	return (1);
1516 }
1517 
1518 /*
1519  * Given address of next destination (final or next hop),
1520  * return internet address info of interface to be used to get there.
1521  */
1522 struct in_ifaddr *
1523 ip_rtaddr(dst, rt)
1524 	struct in_addr dst;
1525 	struct route *rt;
1526 {
1527 	struct sockaddr_in *sin;
1528 
1529 	sin = (struct sockaddr_in *)&rt->ro_dst;
1530 
1531 	if (rt->ro_rt == 0 ||
1532 	    dst.s_addr != sin->sin_addr.s_addr) {
1533 		if (rt->ro_rt) {
1534 			RTFREE(rt->ro_rt);
1535 			rt->ro_rt = 0;
1536 		}
1537 		sin->sin_family = AF_INET;
1538 		sin->sin_len = sizeof(*sin);
1539 		sin->sin_addr = dst;
1540 
1541 		rtalloc_ign(rt, RTF_PRCLONING);
1542 	}
1543 	if (rt->ro_rt == 0)
1544 		return ((struct in_ifaddr *)0);
1545 	return (ifatoia(rt->ro_rt->rt_ifa));
1546 }
1547 
1548 /*
1549  * Save incoming source route for use in replies,
1550  * to be picked up later by ip_srcroute if the receiver is interested.
1551  */
1552 void
1553 save_rte(option, dst)
1554 	u_char *option;
1555 	struct in_addr dst;
1556 {
1557 	unsigned olen;
1558 
1559 	olen = option[IPOPT_OLEN];
1560 #ifdef DIAGNOSTIC
1561 	if (ipprintfs)
1562 		printf("save_rte: olen %d\n", olen);
1563 #endif
1564 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1565 		return;
1566 	bcopy(option, ip_srcrt.srcopt, olen);
1567 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1568 	ip_srcrt.dst = dst;
1569 }
1570 
1571 /*
1572  * Retrieve incoming source route for use in replies,
1573  * in the same form used by setsockopt.
1574  * The first hop is placed before the options, will be removed later.
1575  */
1576 struct mbuf *
1577 ip_srcroute()
1578 {
1579 	struct in_addr *p, *q;
1580 	struct mbuf *m;
1581 
1582 	if (ip_nhops == 0)
1583 		return ((struct mbuf *)0);
1584 	m = m_get(M_DONTWAIT, MT_HEADER);
1585 	if (m == 0)
1586 		return ((struct mbuf *)0);
1587 
1588 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1589 
1590 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1591 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1592 	    OPTSIZ;
1593 #ifdef DIAGNOSTIC
1594 	if (ipprintfs)
1595 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1596 #endif
1597 
1598 	/*
1599 	 * First save first hop for return route
1600 	 */
1601 	p = &ip_srcrt.route[ip_nhops - 1];
1602 	*(mtod(m, struct in_addr *)) = *p--;
1603 #ifdef DIAGNOSTIC
1604 	if (ipprintfs)
1605 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1606 #endif
1607 
1608 	/*
1609 	 * Copy option fields and padding (nop) to mbuf.
1610 	 */
1611 	ip_srcrt.nop = IPOPT_NOP;
1612 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1613 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1614 	    &ip_srcrt.nop, OPTSIZ);
1615 	q = (struct in_addr *)(mtod(m, caddr_t) +
1616 	    sizeof(struct in_addr) + OPTSIZ);
1617 #undef OPTSIZ
1618 	/*
1619 	 * Record return path as an IP source route,
1620 	 * reversing the path (pointers are now aligned).
1621 	 */
1622 	while (p >= ip_srcrt.route) {
1623 #ifdef DIAGNOSTIC
1624 		if (ipprintfs)
1625 			printf(" %lx", (u_long)ntohl(q->s_addr));
1626 #endif
1627 		*q++ = *p--;
1628 	}
1629 	/*
1630 	 * Last hop goes to final destination.
1631 	 */
1632 	*q = ip_srcrt.dst;
1633 #ifdef DIAGNOSTIC
1634 	if (ipprintfs)
1635 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1636 #endif
1637 	return (m);
1638 }
1639 
1640 /*
1641  * Strip out IP options, at higher
1642  * level protocol in the kernel.
1643  * Second argument is buffer to which options
1644  * will be moved, and return value is their length.
1645  * XXX should be deleted; last arg currently ignored.
1646  */
1647 void
1648 ip_stripoptions(m, mopt)
1649 	struct mbuf *m;
1650 	struct mbuf *mopt;
1651 {
1652 	int i;
1653 	struct ip *ip = mtod(m, struct ip *);
1654 	caddr_t opts;
1655 	int olen;
1656 
1657 	olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1658 	opts = (caddr_t)(ip + 1);
1659 	i = m->m_len - (sizeof (struct ip) + olen);
1660 	bcopy(opts + olen, opts, (unsigned)i);
1661 	m->m_len -= olen;
1662 	if (m->m_flags & M_PKTHDR)
1663 		m->m_pkthdr.len -= olen;
1664 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1665 }
1666 
1667 u_char inetctlerrmap[PRC_NCMDS] = {
1668 	0,		0,		0,		0,
1669 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1670 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1671 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1672 	0,		0,		0,		0,
1673 	ENOPROTOOPT,	ECONNREFUSED
1674 };
1675 
1676 /*
1677  * Forward a packet.  If some error occurs return the sender
1678  * an icmp packet.  Note we can't always generate a meaningful
1679  * icmp message because icmp doesn't have a large enough repertoire
1680  * of codes and types.
1681  *
1682  * If not forwarding, just drop the packet.  This could be confusing
1683  * if ipforwarding was zero but some routing protocol was advancing
1684  * us as a gateway to somewhere.  However, we must let the routing
1685  * protocol deal with that.
1686  *
1687  * The srcrt parameter indicates whether the packet is being forwarded
1688  * via a source route.
1689  */
1690 static void
1691 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1692 {
1693 	struct ip *ip = mtod(m, struct ip *);
1694 	struct sockaddr_in *sin;
1695 	struct rtentry *rt;
1696 	int error, type = 0, code = 0;
1697 	struct mbuf *mcopy;
1698 	n_long dest;
1699 	struct in_addr pkt_dst;
1700 	struct ifnet *destifp;
1701 #if defined(IPSEC) || defined(FAST_IPSEC)
1702 	struct ifnet dummyifp;
1703 #endif
1704 
1705 	dest = 0;
1706 	/*
1707 	 * Cache the destination address of the packet; this may be
1708 	 * changed by use of 'ipfw fwd'.
1709 	 */
1710 	pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1711 
1712 #ifdef DIAGNOSTIC
1713 	if (ipprintfs)
1714 		printf("forward: src %lx dst %lx ttl %x\n",
1715 		    (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1716 		    ip->ip_ttl);
1717 #endif
1718 
1719 
1720 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1721 		ipstat.ips_cantforward++;
1722 		m_freem(m);
1723 		return;
1724 	}
1725 #ifdef IPSTEALTH
1726 	if (!ipstealth) {
1727 #endif
1728 		if (ip->ip_ttl <= IPTTLDEC) {
1729 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1730 			    dest, 0);
1731 			return;
1732 		}
1733 #ifdef IPSTEALTH
1734 	}
1735 #endif
1736 
1737 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1738 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1739 	    pkt_dst.s_addr != sin->sin_addr.s_addr) {
1740 		if (ipforward_rt.ro_rt) {
1741 			RTFREE(ipforward_rt.ro_rt);
1742 			ipforward_rt.ro_rt = 0;
1743 		}
1744 		sin->sin_family = AF_INET;
1745 		sin->sin_len = sizeof(*sin);
1746 		sin->sin_addr = pkt_dst;
1747 
1748 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1749 		if (ipforward_rt.ro_rt == 0) {
1750 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1751 			return;
1752 		}
1753 		rt = ipforward_rt.ro_rt;
1754 	}
1755 
1756 	/*
1757 	 * Save the IP header and at most 8 bytes of the payload,
1758 	 * in case we need to generate an ICMP message to the src.
1759 	 *
1760 	 * XXX this can be optimized a lot by saving the data in a local
1761 	 * buffer on the stack (72 bytes at most), and only allocating the
1762 	 * mbuf if really necessary. The vast majority of the packets
1763 	 * are forwarded without having to send an ICMP back (either
1764 	 * because unnecessary, or because rate limited), so we are
1765 	 * really we are wasting a lot of work here.
1766 	 *
1767 	 * We don't use m_copy() because it might return a reference
1768 	 * to a shared cluster. Both this function and ip_output()
1769 	 * assume exclusive access to the IP header in `m', so any
1770 	 * data in a cluster may change before we reach icmp_error().
1771 	 */
1772 	MGET(mcopy, M_DONTWAIT, m->m_type);
1773 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1774 		/*
1775 		 * It's probably ok if the pkthdr dup fails (because
1776 		 * the deep copy of the tag chain failed), but for now
1777 		 * be conservative and just discard the copy since
1778 		 * code below may some day want the tags.
1779 		 */
1780 		m_free(mcopy);
1781 		mcopy = NULL;
1782 	}
1783 	if (mcopy != NULL) {
1784 		mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1785 		    (int)ip->ip_len);
1786 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1787 	}
1788 
1789 #ifdef IPSTEALTH
1790 	if (!ipstealth) {
1791 #endif
1792 		ip->ip_ttl -= IPTTLDEC;
1793 #ifdef IPSTEALTH
1794 	}
1795 #endif
1796 
1797 	/*
1798 	 * If forwarding packet using same interface that it came in on,
1799 	 * perhaps should send a redirect to sender to shortcut a hop.
1800 	 * Only send redirect if source is sending directly to us,
1801 	 * and if packet was not source routed (or has any options).
1802 	 * Also, don't send redirect if forwarding using a default route
1803 	 * or a route modified by a redirect.
1804 	 */
1805 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1806 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1807 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1808 	    ipsendredirects && !srcrt && !next_hop) {
1809 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1810 		u_long src = ntohl(ip->ip_src.s_addr);
1811 
1812 		if (RTA(rt) &&
1813 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1814 		    if (rt->rt_flags & RTF_GATEWAY)
1815 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1816 		    else
1817 			dest = pkt_dst.s_addr;
1818 		    /* Router requirements says to only send host redirects */
1819 		    type = ICMP_REDIRECT;
1820 		    code = ICMP_REDIRECT_HOST;
1821 #ifdef DIAGNOSTIC
1822 		    if (ipprintfs)
1823 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1824 #endif
1825 		}
1826 	}
1827 
1828     {
1829 	struct m_hdr tag;
1830 
1831 	if (next_hop) {
1832 		/* Pass IPFORWARD info if available */
1833 
1834 		tag.mh_type = MT_TAG;
1835 		tag.mh_flags = PACKET_TAG_IPFORWARD;
1836 		tag.mh_data = (caddr_t)next_hop;
1837 		tag.mh_next = m;
1838 		m = (struct mbuf *)&tag;
1839 	}
1840 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1841 			  IP_FORWARDING, 0, NULL);
1842     }
1843 	if (error)
1844 		ipstat.ips_cantforward++;
1845 	else {
1846 		ipstat.ips_forward++;
1847 		if (type)
1848 			ipstat.ips_redirectsent++;
1849 		else {
1850 			if (mcopy) {
1851 				ipflow_create(&ipforward_rt, mcopy);
1852 				m_freem(mcopy);
1853 			}
1854 			return;
1855 		}
1856 	}
1857 	if (mcopy == NULL)
1858 		return;
1859 	destifp = NULL;
1860 
1861 	switch (error) {
1862 
1863 	case 0:				/* forwarded, but need redirect */
1864 		/* type, code set above */
1865 		break;
1866 
1867 	case ENETUNREACH:		/* shouldn't happen, checked above */
1868 	case EHOSTUNREACH:
1869 	case ENETDOWN:
1870 	case EHOSTDOWN:
1871 	default:
1872 		type = ICMP_UNREACH;
1873 		code = ICMP_UNREACH_HOST;
1874 		break;
1875 
1876 	case EMSGSIZE:
1877 		type = ICMP_UNREACH;
1878 		code = ICMP_UNREACH_NEEDFRAG;
1879 #ifdef IPSEC
1880 		/*
1881 		 * If the packet is routed over IPsec tunnel, tell the
1882 		 * originator the tunnel MTU.
1883 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1884 		 * XXX quickhack!!!
1885 		 */
1886 		if (ipforward_rt.ro_rt) {
1887 			struct secpolicy *sp = NULL;
1888 			int ipsecerror;
1889 			int ipsechdr;
1890 			struct route *ro;
1891 
1892 			sp = ipsec4_getpolicybyaddr(mcopy,
1893 						    IPSEC_DIR_OUTBOUND,
1894 			                            IP_FORWARDING,
1895 			                            &ipsecerror);
1896 
1897 			if (sp == NULL)
1898 				destifp = ipforward_rt.ro_rt->rt_ifp;
1899 			else {
1900 				/* count IPsec header size */
1901 				ipsechdr = ipsec4_hdrsiz(mcopy,
1902 							 IPSEC_DIR_OUTBOUND,
1903 							 NULL);
1904 
1905 				/*
1906 				 * find the correct route for outer IPv4
1907 				 * header, compute tunnel MTU.
1908 				 *
1909 				 * XXX BUG ALERT
1910 				 * The "dummyifp" code relies upon the fact
1911 				 * that icmp_error() touches only ifp->if_mtu.
1912 				 */
1913 				/*XXX*/
1914 				destifp = NULL;
1915 				if (sp->req != NULL
1916 				 && sp->req->sav != NULL
1917 				 && sp->req->sav->sah != NULL) {
1918 					ro = &sp->req->sav->sah->sa_route;
1919 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1920 						dummyifp.if_mtu =
1921 						    ro->ro_rt->rt_ifp->if_mtu;
1922 						dummyifp.if_mtu -= ipsechdr;
1923 						destifp = &dummyifp;
1924 					}
1925 				}
1926 
1927 				key_freesp(sp);
1928 			}
1929 		}
1930 #elif FAST_IPSEC
1931 		/*
1932 		 * If the packet is routed over IPsec tunnel, tell the
1933 		 * originator the tunnel MTU.
1934 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1935 		 * XXX quickhack!!!
1936 		 */
1937 		if (ipforward_rt.ro_rt) {
1938 			struct secpolicy *sp = NULL;
1939 			int ipsecerror;
1940 			int ipsechdr;
1941 			struct route *ro;
1942 
1943 			sp = ipsec_getpolicybyaddr(mcopy,
1944 						   IPSEC_DIR_OUTBOUND,
1945 			                           IP_FORWARDING,
1946 			                           &ipsecerror);
1947 
1948 			if (sp == NULL)
1949 				destifp = ipforward_rt.ro_rt->rt_ifp;
1950 			else {
1951 				/* count IPsec header size */
1952 				ipsechdr = ipsec4_hdrsiz(mcopy,
1953 							 IPSEC_DIR_OUTBOUND,
1954 							 NULL);
1955 
1956 				/*
1957 				 * find the correct route for outer IPv4
1958 				 * header, compute tunnel MTU.
1959 				 *
1960 				 * XXX BUG ALERT
1961 				 * The "dummyifp" code relies upon the fact
1962 				 * that icmp_error() touches only ifp->if_mtu.
1963 				 */
1964 				/*XXX*/
1965 				destifp = NULL;
1966 				if (sp->req != NULL
1967 				 && sp->req->sav != NULL
1968 				 && sp->req->sav->sah != NULL) {
1969 					ro = &sp->req->sav->sah->sa_route;
1970 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1971 						dummyifp.if_mtu =
1972 						    ro->ro_rt->rt_ifp->if_mtu;
1973 						dummyifp.if_mtu -= ipsechdr;
1974 						destifp = &dummyifp;
1975 					}
1976 				}
1977 
1978 				KEY_FREESP(&sp);
1979 			}
1980 		}
1981 #else /* !IPSEC && !FAST_IPSEC */
1982 		if (ipforward_rt.ro_rt)
1983 			destifp = ipforward_rt.ro_rt->rt_ifp;
1984 #endif /*IPSEC*/
1985 		ipstat.ips_cantfrag++;
1986 		break;
1987 
1988 	case ENOBUFS:
1989 		/*
1990 		 * A router should not generate ICMP_SOURCEQUENCH as
1991 		 * required in RFC1812 Requirements for IP Version 4 Routers.
1992 		 * Source quench could be a big problem under DoS attacks,
1993 		 * or if the underlying interface is rate-limited.
1994 		 * Those who need source quench packets may re-enable them
1995 		 * via the net.inet.ip.sendsourcequench sysctl.
1996 		 */
1997 		if (ip_sendsourcequench == 0) {
1998 			m_freem(mcopy);
1999 			return;
2000 		} else {
2001 			type = ICMP_SOURCEQUENCH;
2002 			code = 0;
2003 		}
2004 		break;
2005 
2006 	case EACCES:			/* ipfw denied packet */
2007 		m_freem(mcopy);
2008 		return;
2009 	}
2010 	icmp_error(mcopy, type, code, dest, destifp);
2011 }
2012 
2013 void
2014 ip_savecontrol(inp, mp, ip, m)
2015 	struct inpcb *inp;
2016 	struct mbuf **mp;
2017 	struct ip *ip;
2018 	struct mbuf *m;
2019 {
2020 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2021 		struct timeval tv;
2022 
2023 		microtime(&tv);
2024 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2025 			SCM_TIMESTAMP, SOL_SOCKET);
2026 		if (*mp)
2027 			mp = &(*mp)->m_next;
2028 	}
2029 	if (inp->inp_flags & INP_RECVDSTADDR) {
2030 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2031 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2032 		if (*mp)
2033 			mp = &(*mp)->m_next;
2034 	}
2035 #ifdef notyet
2036 	/* XXX
2037 	 * Moving these out of udp_input() made them even more broken
2038 	 * than they already were.
2039 	 */
2040 	/* options were tossed already */
2041 	if (inp->inp_flags & INP_RECVOPTS) {
2042 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2043 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2044 		if (*mp)
2045 			mp = &(*mp)->m_next;
2046 	}
2047 	/* ip_srcroute doesn't do what we want here, need to fix */
2048 	if (inp->inp_flags & INP_RECVRETOPTS) {
2049 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2050 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2051 		if (*mp)
2052 			mp = &(*mp)->m_next;
2053 	}
2054 #endif
2055 	if (inp->inp_flags & INP_RECVIF) {
2056 		struct ifnet *ifp;
2057 		struct sdlbuf {
2058 			struct sockaddr_dl sdl;
2059 			u_char	pad[32];
2060 		} sdlbuf;
2061 		struct sockaddr_dl *sdp;
2062 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2063 
2064 		if (((ifp = m->m_pkthdr.rcvif))
2065 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
2066 			sdp = (struct sockaddr_dl *)(ifnet_addrs
2067 					[ifp->if_index - 1]->ifa_addr);
2068 			/*
2069 			 * Change our mind and don't try copy.
2070 			 */
2071 			if ((sdp->sdl_family != AF_LINK)
2072 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
2073 				goto makedummy;
2074 			}
2075 			bcopy(sdp, sdl2, sdp->sdl_len);
2076 		} else {
2077 makedummy:
2078 			sdl2->sdl_len
2079 				= offsetof(struct sockaddr_dl, sdl_data[0]);
2080 			sdl2->sdl_family = AF_LINK;
2081 			sdl2->sdl_index = 0;
2082 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2083 		}
2084 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2085 			IP_RECVIF, IPPROTO_IP);
2086 		if (*mp)
2087 			mp = &(*mp)->m_next;
2088 	}
2089 }
2090 
2091 /*
2092  * XXX these routines are called from the upper part of the kernel.
2093  *
2094  * They could also be moved to ip_mroute.c, since all the RSVP
2095  *  handling is done there already.
2096  */
2097 int
2098 ip_rsvp_init(struct socket *so)
2099 {
2100 	if (so->so_type != SOCK_RAW ||
2101 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2102 		return EOPNOTSUPP;
2103 
2104 	if (ip_rsvpd != NULL)
2105 		return EADDRINUSE;
2106 
2107 	ip_rsvpd = so;
2108 	/*
2109 	 * This may seem silly, but we need to be sure we don't over-increment
2110 	 * the RSVP counter, in case something slips up.
2111 	 */
2112 	if (!ip_rsvp_on) {
2113 		ip_rsvp_on = 1;
2114 		rsvp_on++;
2115 	}
2116 
2117 	return 0;
2118 }
2119 
2120 int
2121 ip_rsvp_done(void)
2122 {
2123 	ip_rsvpd = NULL;
2124 	/*
2125 	 * This may seem silly, but we need to be sure we don't over-decrement
2126 	 * the RSVP counter, in case something slips up.
2127 	 */
2128 	if (ip_rsvp_on) {
2129 		ip_rsvp_on = 0;
2130 		rsvp_on--;
2131 	}
2132 	return 0;
2133 }
2134 
2135 void
2136 rsvp_input(struct mbuf *m, int off, int proto)	/* XXX must fixup manually */
2137 {
2138 	if (rsvp_input_p) { /* call the real one if loaded */
2139 		rsvp_input_p(m, off, proto);
2140 		return;
2141 	}
2142 
2143 	/* Can still get packets with rsvp_on = 0 if there is a local member
2144 	 * of the group to which the RSVP packet is addressed.  But in this
2145 	 * case we want to throw the packet away.
2146 	 */
2147 
2148 	if (!rsvp_on) {
2149 		m_freem(m);
2150 		return;
2151 	}
2152 
2153 	if (ip_rsvpd != NULL) {
2154 		rip_input(m, off, proto);
2155 		return;
2156 	}
2157 	/* Drop the packet */
2158 	m_freem(m);
2159 }
2160 
2161