xref: /freebsd/sys/netinet/ip_input.c (revision 729362425c09cf6b362366aabc6fb547eee8035a)
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 #include "opt_bootp.h"
38 #include "opt_ipfw.h"
39 #include "opt_ipdn.h"
40 #include "opt_ipdivert.h"
41 #include "opt_ipfilter.h"
42 #include "opt_ipstealth.h"
43 #include "opt_ipsec.h"
44 #include "opt_mac.h"
45 #include "opt_pfil_hooks.h"
46 #include "opt_random_ip_id.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/mac.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/pfil.h>
62 #include <net/if.h>
63 #include <net/if_types.h>
64 #include <net/if_var.h>
65 #include <net/if_dl.h>
66 #include <net/route.h>
67 #include <net/netisr.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 <sys/socketvar.h>
79 
80 #include <netinet/ip_fw.h>
81 #include <netinet/ip_dummynet.h>
82 
83 #ifdef IPSEC
84 #include <netinet6/ipsec.h>
85 #include <netkey/key.h>
86 #endif
87 
88 #ifdef FAST_IPSEC
89 #include <netipsec/ipsec.h>
90 #include <netipsec/key.h>
91 #endif
92 
93 int rsvp_on = 0;
94 
95 int	ipforwarding = 0;
96 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
97     &ipforwarding, 0, "Enable IP forwarding between interfaces");
98 
99 static int	ipsendredirects = 1; /* XXX */
100 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
101     &ipsendredirects, 0, "Enable sending IP redirects");
102 
103 int	ip_defttl = IPDEFTTL;
104 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
105     &ip_defttl, 0, "Maximum TTL on IP packets");
106 
107 static int	ip_dosourceroute = 0;
108 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
109     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
110 
111 static int	ip_acceptsourceroute = 0;
112 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
113     CTLFLAG_RW, &ip_acceptsourceroute, 0,
114     "Enable accepting source routed IP packets");
115 
116 static int	ip_keepfaith = 0;
117 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
118 	&ip_keepfaith,	0,
119 	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
120 
121 static int    nipq = 0;         /* total # of reass queues */
122 static int    maxnipq;
123 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
124 	&maxnipq, 0,
125 	"Maximum number of IPv4 fragment reassembly queue entries");
126 
127 static int    maxfragsperpacket;
128 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
129 	&maxfragsperpacket, 0,
130 	"Maximum number of IPv4 fragments allowed per packet");
131 
132 static int	ip_sendsourcequench = 0;
133 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
134 	&ip_sendsourcequench, 0,
135 	"Enable the transmission of source quench packets");
136 
137 /*
138  * XXX - Setting ip_checkinterface mostly implements the receive side of
139  * the Strong ES model described in RFC 1122, but since the routing table
140  * and transmit implementation do not implement the Strong ES model,
141  * setting this to 1 results in an odd hybrid.
142  *
143  * XXX - ip_checkinterface currently must be disabled if you use ipnat
144  * to translate the destination address to another local interface.
145  *
146  * XXX - ip_checkinterface must be disabled if you add IP aliases
147  * to the loopback interface instead of the interface where the
148  * packets for those addresses are received.
149  */
150 static int	ip_checkinterface = 1;
151 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
152     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
153 
154 #ifdef DIAGNOSTIC
155 static int	ipprintfs = 0;
156 #endif
157 
158 static struct	ifqueue ipintrq;
159 static int	ipqmaxlen = IFQ_MAXLEN;
160 
161 extern	struct domain inetdomain;
162 extern	struct protosw inetsw[];
163 u_char	ip_protox[IPPROTO_MAX];
164 struct	in_ifaddrhead in_ifaddrhead; 		/* first inet address */
165 struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table  */
166 u_long 	in_ifaddrhmask;				/* mask for hash table */
167 
168 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
169     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
170 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
171     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
172 
173 struct ipstat ipstat;
174 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
175     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
176 
177 /* Packet reassembly stuff */
178 #define IPREASS_NHASH_LOG2      6
179 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
180 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
181 #define IPREASS_HASH(x,y) \
182 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
183 
184 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
185 
186 #ifdef IPCTL_DEFMTU
187 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
188     &ip_mtu, 0, "Default MTU");
189 #endif
190 
191 #ifdef IPSTEALTH
192 static int	ipstealth = 0;
193 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
194     &ipstealth, 0, "");
195 #endif
196 
197 
198 /* Firewall hooks */
199 ip_fw_chk_t *ip_fw_chk_ptr;
200 int fw_enable = 1 ;
201 int fw_one_pass = 1;
202 
203 /* Dummynet hooks */
204 ip_dn_io_t *ip_dn_io_ptr;
205 
206 
207 /*
208  * XXX this is ugly -- the following two global variables are
209  * used to store packet state while it travels through the stack.
210  * Note that the code even makes assumptions on the size and
211  * alignment of fields inside struct ip_srcrt so e.g. adding some
212  * fields will break the code. This needs to be fixed.
213  *
214  * We need to save the IP options in case a protocol wants to respond
215  * to an incoming packet over the same route if the packet got here
216  * using IP source routing.  This allows connection establishment and
217  * maintenance when the remote end is on a network that is not known
218  * to us.
219  */
220 static int	ip_nhops = 0;
221 static	struct ip_srcrt {
222 	struct	in_addr dst;			/* final destination */
223 	char	nop;				/* one NOP to align */
224 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
225 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
226 } ip_srcrt;
227 
228 static void	save_rte(u_char *, struct in_addr);
229 static int	ip_dooptions(struct mbuf *m, int,
230 			struct sockaddr_in *next_hop);
231 static void	ip_forward(struct mbuf *m, int srcrt,
232 			struct sockaddr_in *next_hop);
233 static void	ip_freef(struct ipqhead *, struct ipq *);
234 static struct	mbuf *ip_reass(struct mbuf *, struct ipqhead *,
235 		struct ipq *, u_int32_t *, u_int16_t *);
236 
237 /*
238  * IP initialization: fill in IP protocol switch table.
239  * All protocols not implemented in kernel go to raw IP protocol handler.
240  */
241 void
242 ip_init()
243 {
244 	register struct protosw *pr;
245 	register int i;
246 
247 	TAILQ_INIT(&in_ifaddrhead);
248 	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
249 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
250 	if (pr == 0)
251 		panic("ip_init");
252 	for (i = 0; i < IPPROTO_MAX; i++)
253 		ip_protox[i] = pr - inetsw;
254 	for (pr = inetdomain.dom_protosw;
255 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
256 		if (pr->pr_domain->dom_family == PF_INET &&
257 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
258 			ip_protox[pr->pr_protocol] = pr - inetsw;
259 
260 	for (i = 0; i < IPREASS_NHASH; i++)
261 	    TAILQ_INIT(&ipq[i]);
262 
263 	maxnipq = nmbclusters / 32;
264 	maxfragsperpacket = 16;
265 
266 #ifndef RANDOM_IP_ID
267 	ip_id = time_second & 0xffff;
268 #endif
269 	ipintrq.ifq_maxlen = ipqmaxlen;
270 	mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
271 	netisr_register(NETISR_IP, ip_input, &ipintrq);
272 }
273 
274 /*
275  * XXX watch out this one. It is perhaps used as a cache for
276  * the most recently used route ? it is cleared in in_addroute()
277  * when a new route is successfully created.
278  */
279 struct	route ipforward_rt;
280 
281 /*
282  * Ip input routine.  Checksum and byte swap header.  If fragmented
283  * try to reassemble.  Process options.  Pass to next level.
284  */
285 void
286 ip_input(struct mbuf *m)
287 {
288 	struct ip *ip;
289 	struct ipq *fp;
290 	struct in_ifaddr *ia = NULL;
291 	struct ifaddr *ifa;
292 	int    i, hlen, checkif;
293 	u_short sum;
294 	struct in_addr pkt_dst;
295 	u_int32_t divert_info = 0;		/* packet divert/tee info */
296 	struct ip_fw_args args;
297 #ifdef PFIL_HOOKS
298 	struct packet_filter_hook *pfh;
299 	struct mbuf *m0;
300 	int rv;
301 #endif /* PFIL_HOOKS */
302 #ifdef FAST_IPSEC
303 	struct m_tag *mtag;
304 	struct tdb_ident *tdbi;
305 	struct secpolicy *sp;
306 	int s, error;
307 #endif /* FAST_IPSEC */
308 
309 	args.eh = NULL;
310 	args.oif = NULL;
311 	args.rule = NULL;
312 	args.divert_rule = 0;			/* divert cookie */
313 	args.next_hop = NULL;
314 
315 	/* Grab info from MT_TAG mbufs prepended to the chain.	*/
316 	for (; m && m->m_type == MT_TAG; m = m->m_next) {
317 		switch(m->_m_tag_id) {
318 		default:
319 			printf("ip_input: unrecognised MT_TAG tag %d\n",
320 			    m->_m_tag_id);
321 			break;
322 
323 		case PACKET_TAG_DUMMYNET:
324 			args.rule = ((struct dn_pkt *)m)->rule;
325 			break;
326 
327 		case PACKET_TAG_DIVERT:
328 			args.divert_rule = (intptr_t)m->m_hdr.mh_data & 0xffff;
329 			break;
330 
331 		case PACKET_TAG_IPFORWARD:
332 			args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
333 			break;
334 		}
335 	}
336 
337 	KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0,
338 	    ("ip_input: no HDR"));
339 
340 	if (args.rule) {	/* dummynet already filtered us */
341 		ip = mtod(m, struct ip *);
342 		hlen = ip->ip_hl << 2;
343 		goto iphack ;
344 	}
345 
346 	ipstat.ips_total++;
347 
348 	if (m->m_pkthdr.len < sizeof(struct ip))
349 		goto tooshort;
350 
351 	if (m->m_len < sizeof (struct ip) &&
352 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
353 		ipstat.ips_toosmall++;
354 		return;
355 	}
356 	ip = mtod(m, struct ip *);
357 
358 	if (ip->ip_v != IPVERSION) {
359 		ipstat.ips_badvers++;
360 		goto bad;
361 	}
362 
363 	hlen = ip->ip_hl << 2;
364 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
365 		ipstat.ips_badhlen++;
366 		goto bad;
367 	}
368 	if (hlen > m->m_len) {
369 		if ((m = m_pullup(m, hlen)) == 0) {
370 			ipstat.ips_badhlen++;
371 			return;
372 		}
373 		ip = mtod(m, struct ip *);
374 	}
375 
376 	/* 127/8 must not appear on wire - RFC1122 */
377 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
378 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
379 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
380 			ipstat.ips_badaddr++;
381 			goto bad;
382 		}
383 	}
384 
385 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
386 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
387 	} else {
388 		if (hlen == sizeof(struct ip)) {
389 			sum = in_cksum_hdr(ip);
390 		} else {
391 			sum = in_cksum(m, hlen);
392 		}
393 	}
394 	if (sum) {
395 		ipstat.ips_badsum++;
396 		goto bad;
397 	}
398 
399 	/*
400 	 * Convert fields to host representation.
401 	 */
402 	ip->ip_len = ntohs(ip->ip_len);
403 	if (ip->ip_len < hlen) {
404 		ipstat.ips_badlen++;
405 		goto bad;
406 	}
407 	ip->ip_off = ntohs(ip->ip_off);
408 
409 	/*
410 	 * Check that the amount of data in the buffers
411 	 * is as at least much as the IP header would have us expect.
412 	 * Trim mbufs if longer than we expect.
413 	 * Drop packet if shorter than we expect.
414 	 */
415 	if (m->m_pkthdr.len < ip->ip_len) {
416 tooshort:
417 		ipstat.ips_tooshort++;
418 		goto bad;
419 	}
420 	if (m->m_pkthdr.len > ip->ip_len) {
421 		if (m->m_len == m->m_pkthdr.len) {
422 			m->m_len = ip->ip_len;
423 			m->m_pkthdr.len = ip->ip_len;
424 		} else
425 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
426 	}
427 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
428 	/*
429 	 * Bypass packet filtering for packets from a tunnel (gif).
430 	 */
431 	if (ipsec_gethist(m, NULL))
432 		goto pass;
433 #endif
434 
435 	/*
436 	 * IpHack's section.
437 	 * Right now when no processing on packet has done
438 	 * and it is still fresh out of network we do our black
439 	 * deals with it.
440 	 * - Firewall: deny/allow/divert
441 	 * - Xlate: translate packet's addr/port (NAT).
442 	 * - Pipe: pass pkt through dummynet.
443 	 * - Wrap: fake packet's addr/port <unimpl.>
444 	 * - Encapsulate: put it in another IP and send out. <unimp.>
445  	 */
446 
447 iphack:
448 
449 #ifdef PFIL_HOOKS
450 	/*
451 	 * Run through list of hooks for input packets.  If there are any
452 	 * filters which require that additional packets in the flow are
453 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
454 	 * Note that filters must _never_ set this flag, as another filter
455 	 * in the list may have previously cleared it.
456 	 */
457 	m0 = m;
458 	pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
459 	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
460 		if (pfh->pfil_func) {
461 			rv = pfh->pfil_func(ip, hlen,
462 					    m->m_pkthdr.rcvif, 0, &m0);
463 			if (rv)
464 				return;
465 			m = m0;
466 			if (m == NULL)
467 				return;
468 			ip = mtod(m, struct ip *);
469 		}
470 #endif /* PFIL_HOOKS */
471 
472 	if (fw_enable && IPFW_LOADED) {
473 		/*
474 		 * If we've been forwarded from the output side, then
475 		 * skip the firewall a second time
476 		 */
477 		if (args.next_hop)
478 			goto ours;
479 
480 		args.m = m;
481 		i = ip_fw_chk_ptr(&args);
482 		m = args.m;
483 
484 		if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
485 			if (m)
486 				m_freem(m);
487 			return;
488 		}
489 		ip = mtod(m, struct ip *); /* just in case m changed */
490 		if (i == 0 && args.next_hop == NULL)	/* common case */
491 			goto pass;
492                 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
493 			/* Send packet to the appropriate pipe */
494 			ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
495 			return;
496 		}
497 #ifdef IPDIVERT
498 		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
499 			/* Divert or tee packet */
500 			divert_info = i;
501 			goto ours;
502 		}
503 #endif
504 		if (i == 0 && args.next_hop != NULL)
505 			goto pass;
506 		/*
507 		 * if we get here, the packet must be dropped
508 		 */
509 		m_freem(m);
510 		return;
511 	}
512 pass:
513 
514 	/*
515 	 * Process options and, if not destined for us,
516 	 * ship it on.  ip_dooptions returns 1 when an
517 	 * error was detected (causing an icmp message
518 	 * to be sent and the original packet to be freed).
519 	 */
520 	ip_nhops = 0;		/* for source routed packets */
521 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
522 		return;
523 
524         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
525          * matter if it is destined to another node, or whether it is
526          * a multicast one, RSVP wants it! and prevents it from being forwarded
527          * anywhere else. Also checks if the rsvp daemon is running before
528 	 * grabbing the packet.
529          */
530 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
531 		goto ours;
532 
533 	/*
534 	 * Check our list of addresses, to see if the packet is for us.
535 	 * If we don't have any addresses, assume any unicast packet
536 	 * we receive might be for us (and let the upper layers deal
537 	 * with it).
538 	 */
539 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
540 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
541 		goto ours;
542 
543 	/*
544 	 * Cache the destination address of the packet; this may be
545 	 * changed by use of 'ipfw fwd'.
546 	 */
547 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
548 
549 	/*
550 	 * Enable a consistency check between the destination address
551 	 * and the arrival interface for a unicast packet (the RFC 1122
552 	 * strong ES model) if IP forwarding is disabled and the packet
553 	 * is not locally generated and the packet is not subject to
554 	 * 'ipfw fwd'.
555 	 *
556 	 * XXX - Checking also should be disabled if the destination
557 	 * address is ipnat'ed to a different interface.
558 	 *
559 	 * XXX - Checking is incompatible with IP aliases added
560 	 * to the loopback interface instead of the interface where
561 	 * the packets are received.
562 	 */
563 	checkif = ip_checkinterface && (ipforwarding == 0) &&
564 	    m->m_pkthdr.rcvif != NULL &&
565 	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
566 	    (args.next_hop == NULL);
567 
568 	/*
569 	 * Check for exact addresses in the hash bucket.
570 	 */
571 	LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
572 		/*
573 		 * If the address matches, verify that the packet
574 		 * arrived via the correct interface if checking is
575 		 * enabled.
576 		 */
577 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
578 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
579 			goto ours;
580 	}
581 	/*
582 	 * Check for broadcast addresses.
583 	 *
584 	 * Only accept broadcast packets that arrive via the matching
585 	 * interface.  Reception of forwarded directed broadcasts would
586 	 * be handled via ip_forward() and ether_output() with the loopback
587 	 * into the stack for SIMPLEX interfaces handled by ether_output().
588 	 */
589 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
590 	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
591 			if (ifa->ifa_addr->sa_family != AF_INET)
592 				continue;
593 			ia = ifatoia(ifa);
594 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
595 			    pkt_dst.s_addr)
596 				goto ours;
597 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
598 				goto ours;
599 #ifdef BOOTP_COMPAT
600 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
601 				goto ours;
602 #endif
603 		}
604 	}
605 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
606 		struct in_multi *inm;
607 		if (ip_mrouter) {
608 			/*
609 			 * If we are acting as a multicast router, all
610 			 * incoming multicast packets are passed to the
611 			 * kernel-level multicast forwarding function.
612 			 * The packet is returned (relatively) intact; if
613 			 * ip_mforward() returns a non-zero value, the packet
614 			 * must be discarded, else it may be accepted below.
615 			 */
616 			if (ip_mforward &&
617 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
618 				ipstat.ips_cantforward++;
619 				m_freem(m);
620 				return;
621 			}
622 
623 			/*
624 			 * The process-level routing daemon needs to receive
625 			 * all multicast IGMP packets, whether or not this
626 			 * host belongs to their destination groups.
627 			 */
628 			if (ip->ip_p == IPPROTO_IGMP)
629 				goto ours;
630 			ipstat.ips_forward++;
631 		}
632 		/*
633 		 * See if we belong to the destination multicast group on the
634 		 * arrival interface.
635 		 */
636 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
637 		if (inm == NULL) {
638 			ipstat.ips_notmember++;
639 			m_freem(m);
640 			return;
641 		}
642 		goto ours;
643 	}
644 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
645 		goto ours;
646 	if (ip->ip_dst.s_addr == INADDR_ANY)
647 		goto ours;
648 
649 	/*
650 	 * FAITH(Firewall Aided Internet Translator)
651 	 */
652 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
653 		if (ip_keepfaith) {
654 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
655 				goto ours;
656 		}
657 		m_freem(m);
658 		return;
659 	}
660 
661 	/*
662 	 * Not for us; forward if possible and desirable.
663 	 */
664 	if (ipforwarding == 0) {
665 		ipstat.ips_cantforward++;
666 		m_freem(m);
667 	} else {
668 #ifdef IPSEC
669 		/*
670 		 * Enforce inbound IPsec SPD.
671 		 */
672 		if (ipsec4_in_reject(m, NULL)) {
673 			ipsecstat.in_polvio++;
674 			goto bad;
675 		}
676 #endif /* IPSEC */
677 #ifdef FAST_IPSEC
678 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
679 		s = splnet();
680 		if (mtag != NULL) {
681 			tdbi = (struct tdb_ident *)(mtag + 1);
682 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
683 		} else {
684 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
685 						   IP_FORWARDING, &error);
686 		}
687 		if (sp == NULL) {	/* NB: can happen if error */
688 			splx(s);
689 			/*XXX error stat???*/
690 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
691 			goto bad;
692 		}
693 
694 		/*
695 		 * Check security policy against packet attributes.
696 		 */
697 		error = ipsec_in_reject(sp, m);
698 		KEY_FREESP(&sp);
699 		splx(s);
700 		if (error) {
701 			ipstat.ips_cantforward++;
702 			goto bad;
703 		}
704 #endif /* FAST_IPSEC */
705 		ip_forward(m, 0, args.next_hop);
706 	}
707 	return;
708 
709 ours:
710 #ifdef IPSTEALTH
711 	/*
712 	 * IPSTEALTH: Process non-routing options only
713 	 * if the packet is destined for us.
714 	 */
715 	if (ipstealth && hlen > sizeof (struct ip) &&
716 	    ip_dooptions(m, 1, args.next_hop))
717 		return;
718 #endif /* IPSTEALTH */
719 
720 	/* Count the packet in the ip address stats */
721 	if (ia != NULL) {
722 		ia->ia_ifa.if_ipackets++;
723 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
724 	}
725 
726 	/*
727 	 * If offset or IP_MF are set, must reassemble.
728 	 * Otherwise, nothing need be done.
729 	 * (We could look in the reassembly queue to see
730 	 * if the packet was previously fragmented,
731 	 * but it's not worth the time; just let them time out.)
732 	 */
733 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
734 
735 		/* If maxnipq is 0, never accept fragments. */
736 		if (maxnipq == 0) {
737                 	ipstat.ips_fragments++;
738 			ipstat.ips_fragdropped++;
739 			goto bad;
740 		}
741 
742 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
743 		/*
744 		 * Look for queue of fragments
745 		 * of this datagram.
746 		 */
747 		TAILQ_FOREACH(fp, &ipq[sum], ipq_list)
748 			if (ip->ip_id == fp->ipq_id &&
749 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
750 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
751 #ifdef MAC
752 			    mac_fragment_match(m, fp) &&
753 #endif
754 			    ip->ip_p == fp->ipq_p)
755 				goto found;
756 
757 		fp = 0;
758 
759 		/*
760 		 * Enforce upper bound on number of fragmented packets
761 		 * for which we attempt reassembly;
762 		 * If maxnipq is -1, accept all fragments without limitation.
763 		 */
764 		if ((nipq > maxnipq) && (maxnipq > 0)) {
765 		    /*
766 		     * drop something from the tail of the current queue
767 		     * before proceeding further
768 		     */
769 		    struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead);
770 		    if (q == NULL) {   /* gak */
771 			for (i = 0; i < IPREASS_NHASH; i++) {
772 			    struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
773 			    if (r) {
774 				ipstat.ips_fragtimeout += r->ipq_nfrags;
775 				ip_freef(&ipq[i], r);
776 				break;
777 			    }
778 			}
779 		    } else {
780 			ipstat.ips_fragtimeout += q->ipq_nfrags;
781 			ip_freef(&ipq[sum], q);
782 		    }
783 		}
784 found:
785 		/*
786 		 * Adjust ip_len to not reflect header,
787 		 * convert offset of this to bytes.
788 		 */
789 		ip->ip_len -= hlen;
790 		if (ip->ip_off & IP_MF) {
791 		        /*
792 		         * Make sure that fragments have a data length
793 			 * that's a non-zero multiple of 8 bytes.
794 		         */
795 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
796 				ipstat.ips_toosmall++; /* XXX */
797 				goto bad;
798 			}
799 			m->m_flags |= M_FRAG;
800 		} else
801 			m->m_flags &= ~M_FRAG;
802 		ip->ip_off <<= 3;
803 
804 		/*
805 		 * Attempt reassembly; if it succeeds, proceed.
806 		 * ip_reass() will return a different mbuf, and update
807 		 * the divert info in divert_info and args.divert_rule.
808 		 */
809 		ipstat.ips_fragments++;
810 		m->m_pkthdr.header = ip;
811 		m = ip_reass(m,
812 		    &ipq[sum], fp, &divert_info, &args.divert_rule);
813 		if (m == 0)
814 			return;
815 		ipstat.ips_reassembled++;
816 		ip = mtod(m, struct ip *);
817 		/* Get the header length of the reassembled packet */
818 		hlen = ip->ip_hl << 2;
819 #ifdef IPDIVERT
820 		/* Restore original checksum before diverting packet */
821 		if (divert_info != 0) {
822 			ip->ip_len += hlen;
823 			ip->ip_len = htons(ip->ip_len);
824 			ip->ip_off = htons(ip->ip_off);
825 			ip->ip_sum = 0;
826 			if (hlen == sizeof(struct ip))
827 				ip->ip_sum = in_cksum_hdr(ip);
828 			else
829 				ip->ip_sum = in_cksum(m, hlen);
830 			ip->ip_off = ntohs(ip->ip_off);
831 			ip->ip_len = ntohs(ip->ip_len);
832 			ip->ip_len -= hlen;
833 		}
834 #endif
835 	} else
836 		ip->ip_len -= hlen;
837 
838 #ifdef IPDIVERT
839 	/*
840 	 * Divert or tee packet to the divert protocol if required.
841 	 */
842 	if (divert_info != 0) {
843 		struct mbuf *clone = NULL;
844 
845 		/* Clone packet if we're doing a 'tee' */
846 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
847 			clone = m_dup(m, M_DONTWAIT);
848 
849 		/* Restore packet header fields to original values */
850 		ip->ip_len += hlen;
851 		ip->ip_len = htons(ip->ip_len);
852 		ip->ip_off = htons(ip->ip_off);
853 
854 		/* Deliver packet to divert input routine */
855 		divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
856 		ipstat.ips_delivered++;
857 
858 		/* If 'tee', continue with original packet */
859 		if (clone == NULL)
860 			return;
861 		m = clone;
862 		ip = mtod(m, struct ip *);
863 		ip->ip_len += hlen;
864 		/*
865 		 * Jump backwards to complete processing of the
866 		 * packet. But first clear divert_info to avoid
867 		 * entering this block again.
868 		 * We do not need to clear args.divert_rule
869 		 * or args.next_hop as they will not be used.
870 		 */
871 		divert_info = 0;
872 		goto pass;
873 	}
874 #endif
875 
876 #ifdef IPSEC
877 	/*
878 	 * enforce IPsec policy checking if we are seeing last header.
879 	 * note that we do not visit this with protocols with pcb layer
880 	 * code - like udp/tcp/raw ip.
881 	 */
882 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
883 	    ipsec4_in_reject(m, NULL)) {
884 		ipsecstat.in_polvio++;
885 		goto bad;
886 	}
887 #endif
888 #if FAST_IPSEC
889 	/*
890 	 * enforce IPsec policy checking if we are seeing last header.
891 	 * note that we do not visit this with protocols with pcb layer
892 	 * code - like udp/tcp/raw ip.
893 	 */
894 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
895 		/*
896 		 * Check if the packet has already had IPsec processing
897 		 * done.  If so, then just pass it along.  This tag gets
898 		 * set during AH, ESP, etc. input handling, before the
899 		 * packet is returned to the ip input queue for delivery.
900 		 */
901 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
902 		s = splnet();
903 		if (mtag != NULL) {
904 			tdbi = (struct tdb_ident *)(mtag + 1);
905 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
906 		} else {
907 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
908 						   IP_FORWARDING, &error);
909 		}
910 		if (sp != NULL) {
911 			/*
912 			 * Check security policy against packet attributes.
913 			 */
914 			error = ipsec_in_reject(sp, m);
915 			KEY_FREESP(&sp);
916 		} else {
917 			/* XXX error stat??? */
918 			error = EINVAL;
919 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
920 			goto bad;
921 		}
922 		splx(s);
923 		if (error)
924 			goto bad;
925 	}
926 #endif /* FAST_IPSEC */
927 
928 	/*
929 	 * Switch out to protocol's input routine.
930 	 */
931 	ipstat.ips_delivered++;
932 	if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
933 		/* TCP needs IPFORWARD info if available */
934 		struct m_hdr tag;
935 
936 		tag.mh_type = MT_TAG;
937 		tag.mh_flags = PACKET_TAG_IPFORWARD;
938 		tag.mh_data = (caddr_t)args.next_hop;
939 		tag.mh_next = m;
940 
941 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(
942 			(struct mbuf *)&tag, hlen);
943 	} else
944 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
945 	return;
946 bad:
947 	m_freem(m);
948 }
949 
950 /*
951  * Take incoming datagram fragment and try to reassemble it into
952  * whole datagram.  If a chain for reassembly of this datagram already
953  * exists, then it is given as fp; otherwise have to make a chain.
954  *
955  * When IPDIVERT enabled, keep additional state with each packet that
956  * tells us if we need to divert or tee the packet we're building.
957  * In particular, *divinfo includes the port and TEE flag,
958  * *divert_rule is the number of the matching rule.
959  */
960 
961 static struct mbuf *
962 ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp,
963 	u_int32_t *divinfo, u_int16_t *divert_rule)
964 {
965 	struct ip *ip = mtod(m, struct ip *);
966 	register struct mbuf *p, *q, *nq;
967 	struct mbuf *t;
968 	int hlen = ip->ip_hl << 2;
969 	int i, next;
970 
971 	/*
972 	 * Presence of header sizes in mbufs
973 	 * would confuse code below.
974 	 */
975 	m->m_data += hlen;
976 	m->m_len -= hlen;
977 
978 	/*
979 	 * If first fragment to arrive, create a reassembly queue.
980 	 */
981 	if (fp == 0) {
982 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
983 			goto dropfrag;
984 		fp = mtod(t, struct ipq *);
985 #ifdef MAC
986 		if (mac_init_ipq(fp, M_NOWAIT) != 0) {
987 			m_free(t);
988 			goto dropfrag;
989 		}
990 		mac_create_ipq(m, fp);
991 #endif
992 		TAILQ_INSERT_HEAD(head, fp, ipq_list);
993 		nipq++;
994 		fp->ipq_nfrags = 1;
995 		fp->ipq_ttl = IPFRAGTTL;
996 		fp->ipq_p = ip->ip_p;
997 		fp->ipq_id = ip->ip_id;
998 		fp->ipq_src = ip->ip_src;
999 		fp->ipq_dst = ip->ip_dst;
1000 		fp->ipq_frags = m;
1001 		m->m_nextpkt = NULL;
1002 #ifdef IPDIVERT
1003 		fp->ipq_div_info = 0;
1004 		fp->ipq_div_cookie = 0;
1005 #endif
1006 		goto inserted;
1007 	} else {
1008 		fp->ipq_nfrags++;
1009 #ifdef MAC
1010 		mac_update_ipq(m, fp);
1011 #endif
1012 	}
1013 
1014 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1015 
1016 	/*
1017 	 * Find a segment which begins after this one does.
1018 	 */
1019 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1020 		if (GETIP(q)->ip_off > ip->ip_off)
1021 			break;
1022 
1023 	/*
1024 	 * If there is a preceding segment, it may provide some of
1025 	 * our data already.  If so, drop the data from the incoming
1026 	 * segment.  If it provides all of our data, drop us, otherwise
1027 	 * stick new segment in the proper place.
1028 	 *
1029 	 * If some of the data is dropped from the the preceding
1030 	 * segment, then it's checksum is invalidated.
1031 	 */
1032 	if (p) {
1033 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1034 		if (i > 0) {
1035 			if (i >= ip->ip_len)
1036 				goto dropfrag;
1037 			m_adj(m, i);
1038 			m->m_pkthdr.csum_flags = 0;
1039 			ip->ip_off += i;
1040 			ip->ip_len -= i;
1041 		}
1042 		m->m_nextpkt = p->m_nextpkt;
1043 		p->m_nextpkt = m;
1044 	} else {
1045 		m->m_nextpkt = fp->ipq_frags;
1046 		fp->ipq_frags = m;
1047 	}
1048 
1049 	/*
1050 	 * While we overlap succeeding segments trim them or,
1051 	 * if they are completely covered, dequeue them.
1052 	 */
1053 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1054 	     q = nq) {
1055 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1056 		if (i < GETIP(q)->ip_len) {
1057 			GETIP(q)->ip_len -= i;
1058 			GETIP(q)->ip_off += i;
1059 			m_adj(q, i);
1060 			q->m_pkthdr.csum_flags = 0;
1061 			break;
1062 		}
1063 		nq = q->m_nextpkt;
1064 		m->m_nextpkt = nq;
1065 		ipstat.ips_fragdropped++;
1066 		fp->ipq_nfrags--;
1067 		m_freem(q);
1068 	}
1069 
1070 inserted:
1071 
1072 #ifdef IPDIVERT
1073 	/*
1074 	 * Transfer firewall instructions to the fragment structure.
1075 	 * Only trust info in the fragment at offset 0.
1076 	 */
1077 	if (ip->ip_off == 0) {
1078 		fp->ipq_div_info = *divinfo;
1079 		fp->ipq_div_cookie = *divert_rule;
1080 	}
1081 	*divinfo = 0;
1082 	*divert_rule = 0;
1083 #endif
1084 
1085 	/*
1086 	 * Check for complete reassembly and perform frag per packet
1087 	 * limiting.
1088 	 *
1089 	 * Frag limiting is performed here so that the nth frag has
1090 	 * a chance to complete the packet before we drop the packet.
1091 	 * As a result, n+1 frags are actually allowed per packet, but
1092 	 * only n will ever be stored. (n = maxfragsperpacket.)
1093 	 *
1094 	 */
1095 	next = 0;
1096 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1097 		if (GETIP(q)->ip_off != next) {
1098 			if (fp->ipq_nfrags > maxfragsperpacket) {
1099 				ipstat.ips_fragdropped += fp->ipq_nfrags;
1100 				ip_freef(head, fp);
1101 			}
1102 			return (0);
1103 		}
1104 		next += GETIP(q)->ip_len;
1105 	}
1106 	/* Make sure the last packet didn't have the IP_MF flag */
1107 	if (p->m_flags & M_FRAG) {
1108 		if (fp->ipq_nfrags > maxfragsperpacket) {
1109 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1110 			ip_freef(head, fp);
1111 		}
1112 		return (0);
1113 	}
1114 
1115 	/*
1116 	 * Reassembly is complete.  Make sure the packet is a sane size.
1117 	 */
1118 	q = fp->ipq_frags;
1119 	ip = GETIP(q);
1120 	if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1121 		ipstat.ips_toolong++;
1122 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1123 		ip_freef(head, fp);
1124 		return (0);
1125 	}
1126 
1127 	/*
1128 	 * Concatenate fragments.
1129 	 */
1130 	m = q;
1131 	t = m->m_next;
1132 	m->m_next = 0;
1133 	m_cat(m, t);
1134 	nq = q->m_nextpkt;
1135 	q->m_nextpkt = 0;
1136 	for (q = nq; q != NULL; q = nq) {
1137 		nq = q->m_nextpkt;
1138 		q->m_nextpkt = NULL;
1139 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1140 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1141 		m_cat(m, q);
1142 	}
1143 #ifdef MAC
1144 	mac_create_datagram_from_ipq(fp, m);
1145 	mac_destroy_ipq(fp);
1146 #endif
1147 
1148 #ifdef IPDIVERT
1149 	/*
1150 	 * Extract firewall instructions from the fragment structure.
1151 	 */
1152 	*divinfo = fp->ipq_div_info;
1153 	*divert_rule = fp->ipq_div_cookie;
1154 #endif
1155 
1156 	/*
1157 	 * Create header for new ip packet by
1158 	 * modifying header of first packet;
1159 	 * dequeue and discard fragment reassembly header.
1160 	 * Make header visible.
1161 	 */
1162 	ip->ip_len = next;
1163 	ip->ip_src = fp->ipq_src;
1164 	ip->ip_dst = fp->ipq_dst;
1165 	TAILQ_REMOVE(head, fp, ipq_list);
1166 	nipq--;
1167 	(void) m_free(dtom(fp));
1168 	m->m_len += (ip->ip_hl << 2);
1169 	m->m_data -= (ip->ip_hl << 2);
1170 	/* some debugging cruft by sklower, below, will go away soon */
1171 	if (m->m_flags & M_PKTHDR)	/* XXX this should be done elsewhere */
1172 		m_fixhdr(m);
1173 	return (m);
1174 
1175 dropfrag:
1176 #ifdef IPDIVERT
1177 	*divinfo = 0;
1178 	*divert_rule = 0;
1179 #endif
1180 	ipstat.ips_fragdropped++;
1181 	if (fp != 0)
1182 		fp->ipq_nfrags--;
1183 	m_freem(m);
1184 	return (0);
1185 
1186 #undef GETIP
1187 }
1188 
1189 /*
1190  * Free a fragment reassembly header and all
1191  * associated datagrams.
1192  */
1193 static void
1194 ip_freef(fhp, fp)
1195 	struct ipqhead *fhp;
1196 	struct ipq *fp;
1197 {
1198 	register struct mbuf *q;
1199 
1200 	while (fp->ipq_frags) {
1201 		q = fp->ipq_frags;
1202 		fp->ipq_frags = q->m_nextpkt;
1203 		m_freem(q);
1204 	}
1205 	TAILQ_REMOVE(fhp, fp, ipq_list);
1206 	(void) m_free(dtom(fp));
1207 	nipq--;
1208 }
1209 
1210 /*
1211  * IP timer processing;
1212  * if a timer expires on a reassembly
1213  * queue, discard it.
1214  */
1215 void
1216 ip_slowtimo()
1217 {
1218 	register struct ipq *fp;
1219 	int s = splnet();
1220 	int i;
1221 
1222 	for (i = 0; i < IPREASS_NHASH; i++) {
1223 		for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1224 			struct ipq *fpp;
1225 
1226 			fpp = fp;
1227 			fp = TAILQ_NEXT(fp, ipq_list);
1228 			if(--fpp->ipq_ttl == 0) {
1229 				ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1230 				ip_freef(&ipq[i], fpp);
1231 			}
1232 		}
1233 	}
1234 	/*
1235 	 * If we are over the maximum number of fragments
1236 	 * (due to the limit being lowered), drain off
1237 	 * enough to get down to the new limit.
1238 	 */
1239 	if (maxnipq >= 0 && nipq > maxnipq) {
1240 		for (i = 0; i < IPREASS_NHASH; i++) {
1241 			while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1242 				ipstat.ips_fragdropped +=
1243 				    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1244 				ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1245 			}
1246 		}
1247 	}
1248 	ipflow_slowtimo();
1249 	splx(s);
1250 }
1251 
1252 /*
1253  * Drain off all datagram fragments.
1254  */
1255 void
1256 ip_drain()
1257 {
1258 	int     i;
1259 
1260 	for (i = 0; i < IPREASS_NHASH; i++) {
1261 		while(!TAILQ_EMPTY(&ipq[i])) {
1262 			ipstat.ips_fragdropped +=
1263 			    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1264 			ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1265 		}
1266 	}
1267 	in_rtqdrain();
1268 }
1269 
1270 /*
1271  * Do option processing on a datagram,
1272  * possibly discarding it if bad options are encountered,
1273  * or forwarding it if source-routed.
1274  * The pass argument is used when operating in the IPSTEALTH
1275  * mode to tell what options to process:
1276  * [LS]SRR (pass 0) or the others (pass 1).
1277  * The reason for as many as two passes is that when doing IPSTEALTH,
1278  * non-routing options should be processed only if the packet is for us.
1279  * Returns 1 if packet has been forwarded/freed,
1280  * 0 if the packet should be processed further.
1281  */
1282 static int
1283 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1284 {
1285 	struct ip *ip = mtod(m, struct ip *);
1286 	u_char *cp;
1287 	struct in_ifaddr *ia;
1288 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1289 	struct in_addr *sin, dst;
1290 	n_time ntime;
1291 	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1292 
1293 	dst = ip->ip_dst;
1294 	cp = (u_char *)(ip + 1);
1295 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1296 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1297 		opt = cp[IPOPT_OPTVAL];
1298 		if (opt == IPOPT_EOL)
1299 			break;
1300 		if (opt == IPOPT_NOP)
1301 			optlen = 1;
1302 		else {
1303 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1304 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1305 				goto bad;
1306 			}
1307 			optlen = cp[IPOPT_OLEN];
1308 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1309 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1310 				goto bad;
1311 			}
1312 		}
1313 		switch (opt) {
1314 
1315 		default:
1316 			break;
1317 
1318 		/*
1319 		 * Source routing with record.
1320 		 * Find interface with current destination address.
1321 		 * If none on this machine then drop if strictly routed,
1322 		 * or do nothing if loosely routed.
1323 		 * Record interface address and bring up next address
1324 		 * component.  If strictly routed make sure next
1325 		 * address is on directly accessible net.
1326 		 */
1327 		case IPOPT_LSRR:
1328 		case IPOPT_SSRR:
1329 #ifdef IPSTEALTH
1330 			if (ipstealth && pass > 0)
1331 				break;
1332 #endif
1333 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1334 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1335 				goto bad;
1336 			}
1337 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1338 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1339 				goto bad;
1340 			}
1341 			ipaddr.sin_addr = ip->ip_dst;
1342 			ia = (struct in_ifaddr *)
1343 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1344 			if (ia == 0) {
1345 				if (opt == IPOPT_SSRR) {
1346 					type = ICMP_UNREACH;
1347 					code = ICMP_UNREACH_SRCFAIL;
1348 					goto bad;
1349 				}
1350 				if (!ip_dosourceroute)
1351 					goto nosourcerouting;
1352 				/*
1353 				 * Loose routing, and not at next destination
1354 				 * yet; nothing to do except forward.
1355 				 */
1356 				break;
1357 			}
1358 			off--;			/* 0 origin */
1359 			if (off > optlen - (int)sizeof(struct in_addr)) {
1360 				/*
1361 				 * End of source route.  Should be for us.
1362 				 */
1363 				if (!ip_acceptsourceroute)
1364 					goto nosourcerouting;
1365 				save_rte(cp, ip->ip_src);
1366 				break;
1367 			}
1368 #ifdef IPSTEALTH
1369 			if (ipstealth)
1370 				goto dropit;
1371 #endif
1372 			if (!ip_dosourceroute) {
1373 				if (ipforwarding) {
1374 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1375 					/*
1376 					 * Acting as a router, so generate ICMP
1377 					 */
1378 nosourcerouting:
1379 					strcpy(buf, inet_ntoa(ip->ip_dst));
1380 					log(LOG_WARNING,
1381 					    "attempted source route from %s to %s\n",
1382 					    inet_ntoa(ip->ip_src), buf);
1383 					type = ICMP_UNREACH;
1384 					code = ICMP_UNREACH_SRCFAIL;
1385 					goto bad;
1386 				} else {
1387 					/*
1388 					 * Not acting as a router, so silently drop.
1389 					 */
1390 #ifdef IPSTEALTH
1391 dropit:
1392 #endif
1393 					ipstat.ips_cantforward++;
1394 					m_freem(m);
1395 					return (1);
1396 				}
1397 			}
1398 
1399 			/*
1400 			 * locate outgoing interface
1401 			 */
1402 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1403 			    sizeof(ipaddr.sin_addr));
1404 
1405 			if (opt == IPOPT_SSRR) {
1406 #define	INA	struct in_ifaddr *
1407 #define	SA	struct sockaddr *
1408 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1409 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1410 			} else
1411 				ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1412 			if (ia == 0) {
1413 				type = ICMP_UNREACH;
1414 				code = ICMP_UNREACH_SRCFAIL;
1415 				goto bad;
1416 			}
1417 			ip->ip_dst = ipaddr.sin_addr;
1418 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1419 			    sizeof(struct in_addr));
1420 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1421 			/*
1422 			 * Let ip_intr's mcast routing check handle mcast pkts
1423 			 */
1424 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1425 			break;
1426 
1427 		case IPOPT_RR:
1428 #ifdef IPSTEALTH
1429 			if (ipstealth && pass == 0)
1430 				break;
1431 #endif
1432 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1433 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1434 				goto bad;
1435 			}
1436 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1437 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1438 				goto bad;
1439 			}
1440 			/*
1441 			 * If no space remains, ignore.
1442 			 */
1443 			off--;			/* 0 origin */
1444 			if (off > optlen - (int)sizeof(struct in_addr))
1445 				break;
1446 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1447 			    sizeof(ipaddr.sin_addr));
1448 			/*
1449 			 * locate outgoing interface; if we're the destination,
1450 			 * use the incoming interface (should be same).
1451 			 */
1452 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1453 			    (ia = ip_rtaddr(ipaddr.sin_addr,
1454 			    &ipforward_rt)) == 0) {
1455 				type = ICMP_UNREACH;
1456 				code = ICMP_UNREACH_HOST;
1457 				goto bad;
1458 			}
1459 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1460 			    sizeof(struct in_addr));
1461 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1462 			break;
1463 
1464 		case IPOPT_TS:
1465 #ifdef IPSTEALTH
1466 			if (ipstealth && pass == 0)
1467 				break;
1468 #endif
1469 			code = cp - (u_char *)ip;
1470 			if (optlen < 4 || optlen > 40) {
1471 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1472 				goto bad;
1473 			}
1474 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1475 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1476 				goto bad;
1477 			}
1478 			if (off > optlen - (int)sizeof(int32_t)) {
1479 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1480 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1481 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1482 					goto bad;
1483 				}
1484 				break;
1485 			}
1486 			off--;				/* 0 origin */
1487 			sin = (struct in_addr *)(cp + off);
1488 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1489 
1490 			case IPOPT_TS_TSONLY:
1491 				break;
1492 
1493 			case IPOPT_TS_TSANDADDR:
1494 				if (off + sizeof(n_time) +
1495 				    sizeof(struct in_addr) > optlen) {
1496 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1497 					goto bad;
1498 				}
1499 				ipaddr.sin_addr = dst;
1500 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1501 							    m->m_pkthdr.rcvif);
1502 				if (ia == 0)
1503 					continue;
1504 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1505 				    sizeof(struct in_addr));
1506 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1507 				off += sizeof(struct in_addr);
1508 				break;
1509 
1510 			case IPOPT_TS_PRESPEC:
1511 				if (off + sizeof(n_time) +
1512 				    sizeof(struct in_addr) > optlen) {
1513 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1514 					goto bad;
1515 				}
1516 				(void)memcpy(&ipaddr.sin_addr, sin,
1517 				    sizeof(struct in_addr));
1518 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1519 					continue;
1520 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1521 				off += sizeof(struct in_addr);
1522 				break;
1523 
1524 			default:
1525 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1526 				goto bad;
1527 			}
1528 			ntime = iptime();
1529 			(void)memcpy(cp + off, &ntime, sizeof(n_time));
1530 			cp[IPOPT_OFFSET] += sizeof(n_time);
1531 		}
1532 	}
1533 	if (forward && ipforwarding) {
1534 		ip_forward(m, 1, next_hop);
1535 		return (1);
1536 	}
1537 	return (0);
1538 bad:
1539 	icmp_error(m, type, code, 0, 0);
1540 	ipstat.ips_badoptions++;
1541 	return (1);
1542 }
1543 
1544 /*
1545  * Given address of next destination (final or next hop),
1546  * return internet address info of interface to be used to get there.
1547  */
1548 struct in_ifaddr *
1549 ip_rtaddr(dst, rt)
1550 	struct in_addr dst;
1551 	struct route *rt;
1552 {
1553 	register struct sockaddr_in *sin;
1554 
1555 	sin = (struct sockaddr_in *)&rt->ro_dst;
1556 
1557 	if (rt->ro_rt == 0 ||
1558 	    !(rt->ro_rt->rt_flags & RTF_UP) ||
1559 	    dst.s_addr != sin->sin_addr.s_addr) {
1560 		if (rt->ro_rt) {
1561 			RTFREE(rt->ro_rt);
1562 			rt->ro_rt = 0;
1563 		}
1564 		sin->sin_family = AF_INET;
1565 		sin->sin_len = sizeof(*sin);
1566 		sin->sin_addr = dst;
1567 
1568 		rtalloc_ign(rt, RTF_PRCLONING);
1569 	}
1570 	if (rt->ro_rt == 0)
1571 		return ((struct in_ifaddr *)0);
1572 	return (ifatoia(rt->ro_rt->rt_ifa));
1573 }
1574 
1575 /*
1576  * Save incoming source route for use in replies,
1577  * to be picked up later by ip_srcroute if the receiver is interested.
1578  */
1579 static void
1580 save_rte(option, dst)
1581 	u_char *option;
1582 	struct in_addr dst;
1583 {
1584 	unsigned olen;
1585 
1586 	olen = option[IPOPT_OLEN];
1587 #ifdef DIAGNOSTIC
1588 	if (ipprintfs)
1589 		printf("save_rte: olen %d\n", olen);
1590 #endif
1591 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1592 		return;
1593 	bcopy(option, ip_srcrt.srcopt, olen);
1594 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1595 	ip_srcrt.dst = dst;
1596 }
1597 
1598 /*
1599  * Retrieve incoming source route for use in replies,
1600  * in the same form used by setsockopt.
1601  * The first hop is placed before the options, will be removed later.
1602  */
1603 struct mbuf *
1604 ip_srcroute()
1605 {
1606 	register struct in_addr *p, *q;
1607 	register struct mbuf *m;
1608 
1609 	if (ip_nhops == 0)
1610 		return ((struct mbuf *)0);
1611 	m = m_get(M_DONTWAIT, MT_HEADER);
1612 	if (m == 0)
1613 		return ((struct mbuf *)0);
1614 
1615 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1616 
1617 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1618 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1619 	    OPTSIZ;
1620 #ifdef DIAGNOSTIC
1621 	if (ipprintfs)
1622 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1623 #endif
1624 
1625 	/*
1626 	 * First save first hop for return route
1627 	 */
1628 	p = &ip_srcrt.route[ip_nhops - 1];
1629 	*(mtod(m, struct in_addr *)) = *p--;
1630 #ifdef DIAGNOSTIC
1631 	if (ipprintfs)
1632 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1633 #endif
1634 
1635 	/*
1636 	 * Copy option fields and padding (nop) to mbuf.
1637 	 */
1638 	ip_srcrt.nop = IPOPT_NOP;
1639 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1640 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1641 	    &ip_srcrt.nop, OPTSIZ);
1642 	q = (struct in_addr *)(mtod(m, caddr_t) +
1643 	    sizeof(struct in_addr) + OPTSIZ);
1644 #undef OPTSIZ
1645 	/*
1646 	 * Record return path as an IP source route,
1647 	 * reversing the path (pointers are now aligned).
1648 	 */
1649 	while (p >= ip_srcrt.route) {
1650 #ifdef DIAGNOSTIC
1651 		if (ipprintfs)
1652 			printf(" %lx", (u_long)ntohl(q->s_addr));
1653 #endif
1654 		*q++ = *p--;
1655 	}
1656 	/*
1657 	 * Last hop goes to final destination.
1658 	 */
1659 	*q = ip_srcrt.dst;
1660 #ifdef DIAGNOSTIC
1661 	if (ipprintfs)
1662 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1663 #endif
1664 	return (m);
1665 }
1666 
1667 /*
1668  * Strip out IP options, at higher
1669  * level protocol in the kernel.
1670  * Second argument is buffer to which options
1671  * will be moved, and return value is their length.
1672  * XXX should be deleted; last arg currently ignored.
1673  */
1674 void
1675 ip_stripoptions(m, mopt)
1676 	register struct mbuf *m;
1677 	struct mbuf *mopt;
1678 {
1679 	register int i;
1680 	struct ip *ip = mtod(m, struct ip *);
1681 	register caddr_t opts;
1682 	int olen;
1683 
1684 	olen = (ip->ip_hl << 2) - sizeof (struct ip);
1685 	opts = (caddr_t)(ip + 1);
1686 	i = m->m_len - (sizeof (struct ip) + olen);
1687 	bcopy(opts + olen, opts, (unsigned)i);
1688 	m->m_len -= olen;
1689 	if (m->m_flags & M_PKTHDR)
1690 		m->m_pkthdr.len -= olen;
1691 	ip->ip_v = IPVERSION;
1692 	ip->ip_hl = sizeof(struct ip) >> 2;
1693 }
1694 
1695 u_char inetctlerrmap[PRC_NCMDS] = {
1696 	0,		0,		0,		0,
1697 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1698 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1699 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1700 	0,		0,		0,		0,
1701 	ENOPROTOOPT,	ECONNREFUSED
1702 };
1703 
1704 /*
1705  * Forward a packet.  If some error occurs return the sender
1706  * an icmp packet.  Note we can't always generate a meaningful
1707  * icmp message because icmp doesn't have a large enough repertoire
1708  * of codes and types.
1709  *
1710  * If not forwarding, just drop the packet.  This could be confusing
1711  * if ipforwarding was zero but some routing protocol was advancing
1712  * us as a gateway to somewhere.  However, we must let the routing
1713  * protocol deal with that.
1714  *
1715  * The srcrt parameter indicates whether the packet is being forwarded
1716  * via a source route.
1717  */
1718 static void
1719 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1720 {
1721 	struct ip *ip = mtod(m, struct ip *);
1722 	struct rtentry *rt;
1723 	int error, type = 0, code = 0;
1724 	struct mbuf *mcopy;
1725 	n_long dest;
1726 	struct in_addr pkt_dst;
1727 	struct ifnet *destifp;
1728 #if defined(IPSEC) || defined(FAST_IPSEC)
1729 	struct ifnet dummyifp;
1730 #endif
1731 
1732 	dest = 0;
1733 	/*
1734 	 * Cache the destination address of the packet; this may be
1735 	 * changed by use of 'ipfw fwd'.
1736 	 */
1737 	pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1738 
1739 #ifdef DIAGNOSTIC
1740 	if (ipprintfs)
1741 		printf("forward: src %lx dst %lx ttl %x\n",
1742 		    (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1743 		    ip->ip_ttl);
1744 #endif
1745 
1746 
1747 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1748 		ipstat.ips_cantforward++;
1749 		m_freem(m);
1750 		return;
1751 	}
1752 #ifdef IPSTEALTH
1753 	if (!ipstealth) {
1754 #endif
1755 		if (ip->ip_ttl <= IPTTLDEC) {
1756 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1757 			    dest, 0);
1758 			return;
1759 		}
1760 #ifdef IPSTEALTH
1761 	}
1762 #endif
1763 
1764 	if (ip_rtaddr(pkt_dst, &ipforward_rt) == 0) {
1765 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1766 		return;
1767 	} else
1768 		rt = ipforward_rt.ro_rt;
1769 
1770 	/*
1771 	 * Save the IP header and at most 8 bytes of the payload,
1772 	 * in case we need to generate an ICMP message to the src.
1773 	 *
1774 	 * XXX this can be optimized a lot by saving the data in a local
1775 	 * buffer on the stack (72 bytes at most), and only allocating the
1776 	 * mbuf if really necessary. The vast majority of the packets
1777 	 * are forwarded without having to send an ICMP back (either
1778 	 * because unnecessary, or because rate limited), so we are
1779 	 * really we are wasting a lot of work here.
1780 	 *
1781 	 * We don't use m_copy() because it might return a reference
1782 	 * to a shared cluster. Both this function and ip_output()
1783 	 * assume exclusive access to the IP header in `m', so any
1784 	 * data in a cluster may change before we reach icmp_error().
1785 	 */
1786 	MGET(mcopy, M_DONTWAIT, m->m_type);
1787 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1788 		/*
1789 		 * It's probably ok if the pkthdr dup fails (because
1790 		 * the deep copy of the tag chain failed), but for now
1791 		 * be conservative and just discard the copy since
1792 		 * code below may some day want the tags.
1793 		 */
1794 		m_free(mcopy);
1795 		mcopy = NULL;
1796 	}
1797 	if (mcopy != NULL) {
1798 		mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1799 		    (int)ip->ip_len);
1800 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1801 #ifdef MAC
1802 		/*
1803 		 * XXXMAC: This will eventually become an explicit
1804 		 * labeling point.
1805 		 */
1806 		mac_create_mbuf_from_mbuf(m, mcopy);
1807 #endif
1808 	}
1809 
1810 #ifdef IPSTEALTH
1811 	if (!ipstealth) {
1812 #endif
1813 		ip->ip_ttl -= IPTTLDEC;
1814 #ifdef IPSTEALTH
1815 	}
1816 #endif
1817 
1818 	/*
1819 	 * If forwarding packet using same interface that it came in on,
1820 	 * perhaps should send a redirect to sender to shortcut a hop.
1821 	 * Only send redirect if source is sending directly to us,
1822 	 * and if packet was not source routed (or has any options).
1823 	 * Also, don't send redirect if forwarding using a default route
1824 	 * or a route modified by a redirect.
1825 	 */
1826 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1827 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1828 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1829 	    ipsendredirects && !srcrt && !next_hop) {
1830 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1831 		u_long src = ntohl(ip->ip_src.s_addr);
1832 
1833 		if (RTA(rt) &&
1834 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1835 		    if (rt->rt_flags & RTF_GATEWAY)
1836 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1837 		    else
1838 			dest = pkt_dst.s_addr;
1839 		    /* Router requirements says to only send host redirects */
1840 		    type = ICMP_REDIRECT;
1841 		    code = ICMP_REDIRECT_HOST;
1842 #ifdef DIAGNOSTIC
1843 		    if (ipprintfs)
1844 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1845 #endif
1846 		}
1847 	}
1848 
1849     {
1850 	struct m_hdr tag;
1851 
1852 	if (next_hop) {
1853 		/* Pass IPFORWARD info if available */
1854 
1855 		tag.mh_type = MT_TAG;
1856 		tag.mh_flags = PACKET_TAG_IPFORWARD;
1857 		tag.mh_data = (caddr_t)next_hop;
1858 		tag.mh_next = m;
1859 		m = (struct mbuf *)&tag;
1860 	}
1861 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1862 			  IP_FORWARDING, 0, NULL);
1863     }
1864 	if (error)
1865 		ipstat.ips_cantforward++;
1866 	else {
1867 		ipstat.ips_forward++;
1868 		if (type)
1869 			ipstat.ips_redirectsent++;
1870 		else {
1871 			if (mcopy) {
1872 				ipflow_create(&ipforward_rt, mcopy);
1873 				m_freem(mcopy);
1874 			}
1875 			return;
1876 		}
1877 	}
1878 	if (mcopy == NULL)
1879 		return;
1880 	destifp = NULL;
1881 
1882 	switch (error) {
1883 
1884 	case 0:				/* forwarded, but need redirect */
1885 		/* type, code set above */
1886 		break;
1887 
1888 	case ENETUNREACH:		/* shouldn't happen, checked above */
1889 	case EHOSTUNREACH:
1890 	case ENETDOWN:
1891 	case EHOSTDOWN:
1892 	default:
1893 		type = ICMP_UNREACH;
1894 		code = ICMP_UNREACH_HOST;
1895 		break;
1896 
1897 	case EMSGSIZE:
1898 		type = ICMP_UNREACH;
1899 		code = ICMP_UNREACH_NEEDFRAG;
1900 #ifdef IPSEC
1901 		/*
1902 		 * If the packet is routed over IPsec tunnel, tell the
1903 		 * originator the tunnel MTU.
1904 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1905 		 * XXX quickhack!!!
1906 		 */
1907 		if (ipforward_rt.ro_rt) {
1908 			struct secpolicy *sp = NULL;
1909 			int ipsecerror;
1910 			int ipsechdr;
1911 			struct route *ro;
1912 
1913 			sp = ipsec4_getpolicybyaddr(mcopy,
1914 						    IPSEC_DIR_OUTBOUND,
1915 			                            IP_FORWARDING,
1916 			                            &ipsecerror);
1917 
1918 			if (sp == NULL)
1919 				destifp = ipforward_rt.ro_rt->rt_ifp;
1920 			else {
1921 				/* count IPsec header size */
1922 				ipsechdr = ipsec4_hdrsiz(mcopy,
1923 							 IPSEC_DIR_OUTBOUND,
1924 							 NULL);
1925 
1926 				/*
1927 				 * find the correct route for outer IPv4
1928 				 * header, compute tunnel MTU.
1929 				 *
1930 				 * XXX BUG ALERT
1931 				 * The "dummyifp" code relies upon the fact
1932 				 * that icmp_error() touches only ifp->if_mtu.
1933 				 */
1934 				/*XXX*/
1935 				destifp = NULL;
1936 				if (sp->req != NULL
1937 				 && sp->req->sav != NULL
1938 				 && sp->req->sav->sah != NULL) {
1939 					ro = &sp->req->sav->sah->sa_route;
1940 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1941 						dummyifp.if_mtu =
1942 						    ro->ro_rt->rt_ifp->if_mtu;
1943 						dummyifp.if_mtu -= ipsechdr;
1944 						destifp = &dummyifp;
1945 					}
1946 				}
1947 
1948 				key_freesp(sp);
1949 			}
1950 		}
1951 #elif FAST_IPSEC
1952 		/*
1953 		 * If the packet is routed over IPsec tunnel, tell the
1954 		 * originator the tunnel MTU.
1955 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1956 		 * XXX quickhack!!!
1957 		 */
1958 		if (ipforward_rt.ro_rt) {
1959 			struct secpolicy *sp = NULL;
1960 			int ipsecerror;
1961 			int ipsechdr;
1962 			struct route *ro;
1963 
1964 			sp = ipsec_getpolicybyaddr(mcopy,
1965 						   IPSEC_DIR_OUTBOUND,
1966 			                           IP_FORWARDING,
1967 			                           &ipsecerror);
1968 
1969 			if (sp == NULL)
1970 				destifp = ipforward_rt.ro_rt->rt_ifp;
1971 			else {
1972 				/* count IPsec header size */
1973 				ipsechdr = ipsec4_hdrsiz(mcopy,
1974 							 IPSEC_DIR_OUTBOUND,
1975 							 NULL);
1976 
1977 				/*
1978 				 * find the correct route for outer IPv4
1979 				 * header, compute tunnel MTU.
1980 				 *
1981 				 * XXX BUG ALERT
1982 				 * The "dummyifp" code relies upon the fact
1983 				 * that icmp_error() touches only ifp->if_mtu.
1984 				 */
1985 				/*XXX*/
1986 				destifp = NULL;
1987 				if (sp->req != NULL
1988 				 && sp->req->sav != NULL
1989 				 && sp->req->sav->sah != NULL) {
1990 					ro = &sp->req->sav->sah->sa_route;
1991 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1992 						dummyifp.if_mtu =
1993 						    ro->ro_rt->rt_ifp->if_mtu;
1994 						dummyifp.if_mtu -= ipsechdr;
1995 						destifp = &dummyifp;
1996 					}
1997 				}
1998 
1999 				KEY_FREESP(&sp);
2000 			}
2001 		}
2002 #else /* !IPSEC && !FAST_IPSEC */
2003 		if (ipforward_rt.ro_rt)
2004 			destifp = ipforward_rt.ro_rt->rt_ifp;
2005 #endif /*IPSEC*/
2006 		ipstat.ips_cantfrag++;
2007 		break;
2008 
2009 	case ENOBUFS:
2010 		/*
2011 		 * A router should not generate ICMP_SOURCEQUENCH as
2012 		 * required in RFC1812 Requirements for IP Version 4 Routers.
2013 		 * Source quench could be a big problem under DoS attacks,
2014 		 * or if the underlying interface is rate-limited.
2015 		 * Those who need source quench packets may re-enable them
2016 		 * via the net.inet.ip.sendsourcequench sysctl.
2017 		 */
2018 		if (ip_sendsourcequench == 0) {
2019 			m_freem(mcopy);
2020 			return;
2021 		} else {
2022 			type = ICMP_SOURCEQUENCH;
2023 			code = 0;
2024 		}
2025 		break;
2026 
2027 	case EACCES:			/* ipfw denied packet */
2028 		m_freem(mcopy);
2029 		return;
2030 	}
2031 	icmp_error(mcopy, type, code, dest, destifp);
2032 }
2033 
2034 void
2035 ip_savecontrol(inp, mp, ip, m)
2036 	register struct inpcb *inp;
2037 	register struct mbuf **mp;
2038 	register struct ip *ip;
2039 	register struct mbuf *m;
2040 {
2041 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2042 		struct timeval tv;
2043 
2044 		microtime(&tv);
2045 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2046 			SCM_TIMESTAMP, SOL_SOCKET);
2047 		if (*mp)
2048 			mp = &(*mp)->m_next;
2049 	}
2050 	if (inp->inp_flags & INP_RECVDSTADDR) {
2051 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2052 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2053 		if (*mp)
2054 			mp = &(*mp)->m_next;
2055 	}
2056 #ifdef notyet
2057 	/* XXX
2058 	 * Moving these out of udp_input() made them even more broken
2059 	 * than they already were.
2060 	 */
2061 	/* options were tossed already */
2062 	if (inp->inp_flags & INP_RECVOPTS) {
2063 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2064 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2065 		if (*mp)
2066 			mp = &(*mp)->m_next;
2067 	}
2068 	/* ip_srcroute doesn't do what we want here, need to fix */
2069 	if (inp->inp_flags & INP_RECVRETOPTS) {
2070 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2071 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2072 		if (*mp)
2073 			mp = &(*mp)->m_next;
2074 	}
2075 #endif
2076 	if (inp->inp_flags & INP_RECVIF) {
2077 		struct ifnet *ifp;
2078 		struct sdlbuf {
2079 			struct sockaddr_dl sdl;
2080 			u_char	pad[32];
2081 		} sdlbuf;
2082 		struct sockaddr_dl *sdp;
2083 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2084 
2085 		if (((ifp = m->m_pkthdr.rcvif))
2086 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
2087 			sdp = (struct sockaddr_dl *)
2088 			    (ifaddr_byindex(ifp->if_index)->ifa_addr);
2089 			/*
2090 			 * Change our mind and don't try copy.
2091 			 */
2092 			if ((sdp->sdl_family != AF_LINK)
2093 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
2094 				goto makedummy;
2095 			}
2096 			bcopy(sdp, sdl2, sdp->sdl_len);
2097 		} else {
2098 makedummy:
2099 			sdl2->sdl_len
2100 				= offsetof(struct sockaddr_dl, sdl_data[0]);
2101 			sdl2->sdl_family = AF_LINK;
2102 			sdl2->sdl_index = 0;
2103 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2104 		}
2105 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2106 			IP_RECVIF, IPPROTO_IP);
2107 		if (*mp)
2108 			mp = &(*mp)->m_next;
2109 	}
2110 }
2111 
2112 /*
2113  * XXX these routines are called from the upper part of the kernel.
2114  * They need to be locked when we remove Giant.
2115  *
2116  * They could also be moved to ip_mroute.c, since all the RSVP
2117  *  handling is done there already.
2118  */
2119 static int ip_rsvp_on;
2120 struct socket *ip_rsvpd;
2121 int
2122 ip_rsvp_init(struct socket *so)
2123 {
2124 	if (so->so_type != SOCK_RAW ||
2125 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2126 		return EOPNOTSUPP;
2127 
2128 	if (ip_rsvpd != NULL)
2129 		return EADDRINUSE;
2130 
2131 	ip_rsvpd = so;
2132 	/*
2133 	 * This may seem silly, but we need to be sure we don't over-increment
2134 	 * the RSVP counter, in case something slips up.
2135 	 */
2136 	if (!ip_rsvp_on) {
2137 		ip_rsvp_on = 1;
2138 		rsvp_on++;
2139 	}
2140 
2141 	return 0;
2142 }
2143 
2144 int
2145 ip_rsvp_done(void)
2146 {
2147 	ip_rsvpd = NULL;
2148 	/*
2149 	 * This may seem silly, but we need to be sure we don't over-decrement
2150 	 * the RSVP counter, in case something slips up.
2151 	 */
2152 	if (ip_rsvp_on) {
2153 		ip_rsvp_on = 0;
2154 		rsvp_on--;
2155 	}
2156 	return 0;
2157 }
2158 
2159 void
2160 rsvp_input(struct mbuf *m, int off)	/* XXX must fixup manually */
2161 {
2162 	if (rsvp_input_p) { /* call the real one if loaded */
2163 		rsvp_input_p(m, off);
2164 		return;
2165 	}
2166 
2167 	/* Can still get packets with rsvp_on = 0 if there is a local member
2168 	 * of the group to which the RSVP packet is addressed.  But in this
2169 	 * case we want to throw the packet away.
2170 	 */
2171 
2172 	if (!rsvp_on) {
2173 		m_freem(m);
2174 		return;
2175 	}
2176 
2177 	if (ip_rsvpd != NULL) {
2178 		rip_input(m, off);
2179 		return;
2180 	}
2181 	/* Drop the packet */
2182 	m_freem(m);
2183 }
2184