xref: /freebsd/sys/netinet/ip_input.c (revision 6af83ee0d2941d18880b6aaa2b4facd1d30c6106)
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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
30  * $FreeBSD$
31  */
32 
33 #include "opt_bootp.h"
34 #include "opt_ipfw.h"
35 #include "opt_ipstealth.h"
36 #include "opt_ipsec.h"
37 #include "opt_mac.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/callout.h>
42 #include <sys/mac.h>
43 #include <sys/mbuf.h>
44 #include <sys/malloc.h>
45 #include <sys/domain.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
48 #include <sys/time.h>
49 #include <sys/kernel.h>
50 #include <sys/syslog.h>
51 #include <sys/sysctl.h>
52 
53 #include <net/pfil.h>
54 #include <net/if.h>
55 #include <net/if_types.h>
56 #include <net/if_var.h>
57 #include <net/if_dl.h>
58 #include <net/route.h>
59 #include <net/netisr.h>
60 
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/in_var.h>
64 #include <netinet/ip.h>
65 #include <netinet/in_pcb.h>
66 #include <netinet/ip_var.h>
67 #include <netinet/ip_icmp.h>
68 #include <machine/in_cksum.h>
69 
70 #include <sys/socketvar.h>
71 
72 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
73 #include <netinet/ip_fw.h>
74 #include <netinet/ip_dummynet.h>
75 
76 #ifdef IPSEC
77 #include <netinet6/ipsec.h>
78 #include <netkey/key.h>
79 #endif
80 
81 #ifdef FAST_IPSEC
82 #include <netipsec/ipsec.h>
83 #include <netipsec/key.h>
84 #endif
85 
86 int rsvp_on = 0;
87 
88 int	ipforwarding = 0;
89 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
90     &ipforwarding, 0, "Enable IP forwarding between interfaces");
91 
92 static int	ipsendredirects = 1; /* XXX */
93 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
94     &ipsendredirects, 0, "Enable sending IP redirects");
95 
96 int	ip_defttl = IPDEFTTL;
97 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
98     &ip_defttl, 0, "Maximum TTL on IP packets");
99 
100 static int	ip_dosourceroute = 0;
101 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
102     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
103 
104 static int	ip_acceptsourceroute = 0;
105 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
106     CTLFLAG_RW, &ip_acceptsourceroute, 0,
107     "Enable accepting source routed IP packets");
108 
109 int		ip_doopts = 1;	/* 0 = ignore, 1 = process, 2 = reject */
110 SYSCTL_INT(_net_inet_ip, OID_AUTO, process_options, CTLFLAG_RW,
111     &ip_doopts, 0, "Enable IP options processing ([LS]SRR, RR, TS)");
112 
113 static int	ip_keepfaith = 0;
114 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
115 	&ip_keepfaith,	0,
116 	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
117 
118 static int    nipq = 0;         /* total # of reass queues */
119 static int    maxnipq;
120 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
121 	&maxnipq, 0,
122 	"Maximum number of IPv4 fragment reassembly queue entries");
123 
124 static int    maxfragsperpacket;
125 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
126 	&maxfragsperpacket, 0,
127 	"Maximum number of IPv4 fragments allowed per packet");
128 
129 static int	ip_sendsourcequench = 0;
130 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
131 	&ip_sendsourcequench, 0,
132 	"Enable the transmission of source quench packets");
133 
134 int	ip_do_randomid = 0;
135 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
136 	&ip_do_randomid, 0,
137 	"Assign random ip_id values");
138 
139 /*
140  * XXX - Setting ip_checkinterface mostly implements the receive side of
141  * the Strong ES model described in RFC 1122, but since the routing table
142  * and transmit implementation do not implement the Strong ES model,
143  * setting this to 1 results in an odd hybrid.
144  *
145  * XXX - ip_checkinterface currently must be disabled if you use ipnat
146  * to translate the destination address to another local interface.
147  *
148  * XXX - ip_checkinterface must be disabled if you add IP aliases
149  * to the loopback interface instead of the interface where the
150  * packets for those addresses are received.
151  */
152 static int	ip_checkinterface = 0;
153 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
154     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
155 
156 #ifdef DIAGNOSTIC
157 static int	ipprintfs = 0;
158 #endif
159 
160 struct pfil_head inet_pfil_hook;	/* Packet filter hooks */
161 
162 static struct	ifqueue ipintrq;
163 static int	ipqmaxlen = IFQ_MAXLEN;
164 
165 extern	struct domain inetdomain;
166 extern	struct protosw 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 
172 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
173     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
174 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
175     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
176 
177 struct ipstat ipstat;
178 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
179     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
180 
181 /* Packet reassembly stuff */
182 #define IPREASS_NHASH_LOG2      6
183 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
184 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
185 #define IPREASS_HASH(x,y) \
186 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
187 
188 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
189 struct mtx ipqlock;
190 struct callout ipport_tick_callout;
191 
192 #define	IPQ_LOCK()	mtx_lock(&ipqlock)
193 #define	IPQ_UNLOCK()	mtx_unlock(&ipqlock)
194 #define	IPQ_LOCK_INIT()	mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
195 #define	IPQ_LOCK_ASSERT()	mtx_assert(&ipqlock, MA_OWNED)
196 
197 #ifdef IPCTL_DEFMTU
198 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
199     &ip_mtu, 0, "Default MTU");
200 #endif
201 
202 #ifdef IPSTEALTH
203 int	ipstealth = 0;
204 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
205     &ipstealth, 0, "");
206 #endif
207 
208 /*
209  * ipfw_ether and ipfw_bridge hooks.
210  * XXX: Temporary until those are converted to pfil_hooks as well.
211  */
212 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
213 ip_dn_io_t *ip_dn_io_ptr = NULL;
214 int fw_enable = 1;
215 int fw_one_pass = 1;
216 
217 /*
218  * XXX this is ugly.  IP options source routing magic.
219  */
220 struct ipoptrt {
221 	struct	in_addr dst;			/* final destination */
222 	char	nop;				/* one NOP to align */
223 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
224 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
225 };
226 
227 struct ipopt_tag {
228 	struct	m_tag tag;
229 	int	ip_nhops;
230 	struct	ipoptrt ip_srcrt;
231 };
232 
233 static void	save_rte(struct mbuf *, u_char *, struct in_addr);
234 static int	ip_dooptions(struct mbuf *m, int);
235 static void	ip_forward(struct mbuf *m, int srcrt);
236 static void	ip_freef(struct ipqhead *, struct ipq *);
237 
238 /*
239  * IP initialization: fill in IP protocol switch table.
240  * All protocols not implemented in kernel go to raw IP protocol handler.
241  */
242 void
243 ip_init()
244 {
245 	register struct protosw *pr;
246 	register int i;
247 
248 	TAILQ_INIT(&in_ifaddrhead);
249 	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
250 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
251 	if (pr == NULL)
252 		panic("ip_init: PF_INET not found");
253 
254 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
255 	for (i = 0; i < IPPROTO_MAX; i++)
256 		ip_protox[i] = pr - inetsw;
257 	/*
258 	 * Cycle through IP protocols and put them into the appropriate place
259 	 * in ip_protox[].
260 	 */
261 	for (pr = inetdomain.dom_protosw;
262 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
263 		if (pr->pr_domain->dom_family == PF_INET &&
264 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
265 			/* Be careful to only index valid IP protocols. */
266 			if (pr->pr_protocol <= IPPROTO_MAX)
267 				ip_protox[pr->pr_protocol] = pr - inetsw;
268 		}
269 
270 	/* Initialize packet filter hooks. */
271 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
272 	inet_pfil_hook.ph_af = AF_INET;
273 	if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
274 		printf("%s: WARNING: unable to register pfil hook, "
275 			"error %d\n", __func__, i);
276 
277 	/* Initialize IP reassembly queue. */
278 	IPQ_LOCK_INIT();
279 	for (i = 0; i < IPREASS_NHASH; i++)
280 	    TAILQ_INIT(&ipq[i]);
281 	maxnipq = nmbclusters / 32;
282 	maxfragsperpacket = 16;
283 
284 	/* Start ipport_tick. */
285 	callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
286 	ipport_tick(NULL);
287 	EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
288 		SHUTDOWN_PRI_DEFAULT);
289 
290 	/* Initialize various other remaining things. */
291 	ip_id = time_second & 0xffff;
292 	ipintrq.ifq_maxlen = ipqmaxlen;
293 	mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
294 	netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
295 }
296 
297 void ip_fini(xtp)
298 	void *xtp;
299 {
300 	callout_stop(&ipport_tick_callout);
301 }
302 
303 /*
304  * Ip input routine.  Checksum and byte swap header.  If fragmented
305  * try to reassemble.  Process options.  Pass to next level.
306  */
307 void
308 ip_input(struct mbuf *m)
309 {
310 	struct ip *ip = NULL;
311 	struct in_ifaddr *ia = NULL;
312 	struct ifaddr *ifa;
313 	int    checkif, hlen = 0;
314 	u_short sum;
315 	int dchg = 0;				/* dest changed after fw */
316 	struct in_addr odst;			/* original dst address */
317 #ifdef FAST_IPSEC
318 	struct m_tag *mtag;
319 	struct tdb_ident *tdbi;
320 	struct secpolicy *sp;
321 	int s, error;
322 #endif /* FAST_IPSEC */
323 
324   	M_ASSERTPKTHDR(m);
325 
326 	if (m->m_flags & M_FASTFWD_OURS) {
327 		/*
328 		 * Firewall or NAT changed destination to local.
329 		 * We expect ip_len and ip_off to be in host byte order.
330 		 */
331 		m->m_flags &= ~M_FASTFWD_OURS;
332 		/* Set up some basics that will be used later. */
333 		ip = mtod(m, struct ip *);
334 		hlen = ip->ip_hl << 2;
335   		goto ours;
336   	}
337 
338 	ipstat.ips_total++;
339 
340 	if (m->m_pkthdr.len < sizeof(struct ip))
341 		goto tooshort;
342 
343 	if (m->m_len < sizeof (struct ip) &&
344 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
345 		ipstat.ips_toosmall++;
346 		return;
347 	}
348 	ip = mtod(m, struct ip *);
349 
350 	if (ip->ip_v != IPVERSION) {
351 		ipstat.ips_badvers++;
352 		goto bad;
353 	}
354 
355 	hlen = ip->ip_hl << 2;
356 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
357 		ipstat.ips_badhlen++;
358 		goto bad;
359 	}
360 	if (hlen > m->m_len) {
361 		if ((m = m_pullup(m, hlen)) == NULL) {
362 			ipstat.ips_badhlen++;
363 			return;
364 		}
365 		ip = mtod(m, struct ip *);
366 	}
367 
368 	/* 127/8 must not appear on wire - RFC1122 */
369 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
370 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
371 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
372 			ipstat.ips_badaddr++;
373 			goto bad;
374 		}
375 	}
376 
377 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
378 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
379 	} else {
380 		if (hlen == sizeof(struct ip)) {
381 			sum = in_cksum_hdr(ip);
382 		} else {
383 			sum = in_cksum(m, hlen);
384 		}
385 	}
386 	if (sum) {
387 		ipstat.ips_badsum++;
388 		goto bad;
389 	}
390 
391 #ifdef ALTQ
392 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
393 		/* packet is dropped by traffic conditioner */
394 		return;
395 #endif
396 
397 	/*
398 	 * Convert fields to host representation.
399 	 */
400 	ip->ip_len = ntohs(ip->ip_len);
401 	if (ip->ip_len < hlen) {
402 		ipstat.ips_badlen++;
403 		goto bad;
404 	}
405 	ip->ip_off = ntohs(ip->ip_off);
406 
407 	/*
408 	 * Check that the amount of data in the buffers
409 	 * is as at least much as the IP header would have us expect.
410 	 * Trim mbufs if longer than we expect.
411 	 * Drop packet if shorter than we expect.
412 	 */
413 	if (m->m_pkthdr.len < ip->ip_len) {
414 tooshort:
415 		ipstat.ips_tooshort++;
416 		goto bad;
417 	}
418 	if (m->m_pkthdr.len > ip->ip_len) {
419 		if (m->m_len == m->m_pkthdr.len) {
420 			m->m_len = ip->ip_len;
421 			m->m_pkthdr.len = ip->ip_len;
422 		} else
423 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
424 	}
425 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
426 	/*
427 	 * Bypass packet filtering for packets from a tunnel (gif).
428 	 */
429 	if (ipsec_getnhist(m))
430 		goto passin;
431 #endif
432 #if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF)
433 	/*
434 	 * Bypass packet filtering for packets from a tunnel (gif).
435 	 */
436 	if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
437 		goto passin;
438 #endif
439 
440 	/*
441 	 * Run through list of hooks for input packets.
442 	 *
443 	 * NB: Beware of the destination address changing (e.g.
444 	 *     by NAT rewriting).  When this happens, tell
445 	 *     ip_forward to do the right thing.
446 	 */
447 
448 	/* Jump over all PFIL processing if hooks are not active. */
449 	if (inet_pfil_hook.ph_busy_count == -1)
450 		goto passin;
451 
452 	odst = ip->ip_dst;
453 	if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
454 	    PFIL_IN, NULL) != 0)
455 		return;
456 	if (m == NULL)			/* consumed by filter */
457 		return;
458 
459 	ip = mtod(m, struct ip *);
460 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
461 
462 #ifdef IPFIREWALL_FORWARD
463 	if (m->m_flags & M_FASTFWD_OURS) {
464 		m->m_flags &= ~M_FASTFWD_OURS;
465 		goto ours;
466 	}
467 	dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
468 #endif /* IPFIREWALL_FORWARD */
469 
470 passin:
471 	/*
472 	 * Process options and, if not destined for us,
473 	 * ship it on.  ip_dooptions returns 1 when an
474 	 * error was detected (causing an icmp message
475 	 * to be sent and the original packet to be freed).
476 	 */
477 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
478 		return;
479 
480         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
481          * matter if it is destined to another node, or whether it is
482          * a multicast one, RSVP wants it! and prevents it from being forwarded
483          * anywhere else. Also checks if the rsvp daemon is running before
484 	 * grabbing the packet.
485          */
486 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
487 		goto ours;
488 
489 	/*
490 	 * Check our list of addresses, to see if the packet is for us.
491 	 * If we don't have any addresses, assume any unicast packet
492 	 * we receive might be for us (and let the upper layers deal
493 	 * with it).
494 	 */
495 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
496 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
497 		goto ours;
498 
499 	/*
500 	 * Enable a consistency check between the destination address
501 	 * and the arrival interface for a unicast packet (the RFC 1122
502 	 * strong ES model) if IP forwarding is disabled and the packet
503 	 * is not locally generated and the packet is not subject to
504 	 * 'ipfw fwd'.
505 	 *
506 	 * XXX - Checking also should be disabled if the destination
507 	 * address is ipnat'ed to a different interface.
508 	 *
509 	 * XXX - Checking is incompatible with IP aliases added
510 	 * to the loopback interface instead of the interface where
511 	 * the packets are received.
512 	 */
513 	checkif = ip_checkinterface && (ipforwarding == 0) &&
514 	    m->m_pkthdr.rcvif != NULL &&
515 	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
516 	    (dchg == 0);
517 
518 	/*
519 	 * Check for exact addresses in the hash bucket.
520 	 */
521 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
522 		/*
523 		 * If the address matches, verify that the packet
524 		 * arrived via the correct interface if checking is
525 		 * enabled.
526 		 */
527 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
528 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
529 			goto ours;
530 	}
531 	/*
532 	 * Check for broadcast addresses.
533 	 *
534 	 * Only accept broadcast packets that arrive via the matching
535 	 * interface.  Reception of forwarded directed broadcasts would
536 	 * be handled via ip_forward() and ether_output() with the loopback
537 	 * into the stack for SIMPLEX interfaces handled by ether_output().
538 	 */
539 	if (m->m_pkthdr.rcvif != NULL &&
540 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
541 	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
542 			if (ifa->ifa_addr->sa_family != AF_INET)
543 				continue;
544 			ia = ifatoia(ifa);
545 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
546 			    ip->ip_dst.s_addr)
547 				goto ours;
548 			if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
549 				goto ours;
550 #ifdef BOOTP_COMPAT
551 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
552 				goto ours;
553 #endif
554 		}
555 	}
556 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
557 		struct in_multi *inm;
558 		if (ip_mrouter) {
559 			/*
560 			 * If we are acting as a multicast router, all
561 			 * incoming multicast packets are passed to the
562 			 * kernel-level multicast forwarding function.
563 			 * The packet is returned (relatively) intact; if
564 			 * ip_mforward() returns a non-zero value, the packet
565 			 * must be discarded, else it may be accepted below.
566 			 */
567 			if (ip_mforward &&
568 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
569 				ipstat.ips_cantforward++;
570 				m_freem(m);
571 				return;
572 			}
573 
574 			/*
575 			 * The process-level routing daemon needs to receive
576 			 * all multicast IGMP packets, whether or not this
577 			 * host belongs to their destination groups.
578 			 */
579 			if (ip->ip_p == IPPROTO_IGMP)
580 				goto ours;
581 			ipstat.ips_forward++;
582 		}
583 		/*
584 		 * See if we belong to the destination multicast group on the
585 		 * arrival interface.
586 		 */
587 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
588 		if (inm == NULL) {
589 			ipstat.ips_notmember++;
590 			m_freem(m);
591 			return;
592 		}
593 		goto ours;
594 	}
595 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
596 		goto ours;
597 	if (ip->ip_dst.s_addr == INADDR_ANY)
598 		goto ours;
599 
600 	/*
601 	 * FAITH(Firewall Aided Internet Translator)
602 	 */
603 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
604 		if (ip_keepfaith) {
605 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
606 				goto ours;
607 		}
608 		m_freem(m);
609 		return;
610 	}
611 
612 	/*
613 	 * Not for us; forward if possible and desirable.
614 	 */
615 	if (ipforwarding == 0) {
616 		ipstat.ips_cantforward++;
617 		m_freem(m);
618 	} else {
619 #ifdef IPSEC
620 		/*
621 		 * Enforce inbound IPsec SPD.
622 		 */
623 		if (ipsec4_in_reject(m, NULL)) {
624 			ipsecstat.in_polvio++;
625 			goto bad;
626 		}
627 #endif /* IPSEC */
628 #ifdef FAST_IPSEC
629 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
630 		s = splnet();
631 		if (mtag != NULL) {
632 			tdbi = (struct tdb_ident *)(mtag + 1);
633 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
634 		} else {
635 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
636 						   IP_FORWARDING, &error);
637 		}
638 		if (sp == NULL) {	/* NB: can happen if error */
639 			splx(s);
640 			/*XXX error stat???*/
641 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
642 			goto bad;
643 		}
644 
645 		/*
646 		 * Check security policy against packet attributes.
647 		 */
648 		error = ipsec_in_reject(sp, m);
649 		KEY_FREESP(&sp);
650 		splx(s);
651 		if (error) {
652 			ipstat.ips_cantforward++;
653 			goto bad;
654 		}
655 #endif /* FAST_IPSEC */
656 		ip_forward(m, dchg);
657 	}
658 	return;
659 
660 ours:
661 #ifdef IPSTEALTH
662 	/*
663 	 * IPSTEALTH: Process non-routing options only
664 	 * if the packet is destined for us.
665 	 */
666 	if (ipstealth && hlen > sizeof (struct ip) &&
667 	    ip_dooptions(m, 1))
668 		return;
669 #endif /* IPSTEALTH */
670 
671 	/* Count the packet in the ip address stats */
672 	if (ia != NULL) {
673 		ia->ia_ifa.if_ipackets++;
674 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
675 	}
676 
677 	/*
678 	 * Attempt reassembly; if it succeeds, proceed.
679 	 * ip_reass() will return a different mbuf.
680 	 */
681 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
682 		m = ip_reass(m);
683 		if (m == NULL)
684 			return;
685 		ip = mtod(m, struct ip *);
686 		/* Get the header length of the reassembled packet */
687 		hlen = ip->ip_hl << 2;
688 	}
689 
690 	/*
691 	 * Further protocols expect the packet length to be w/o the
692 	 * IP header.
693 	 */
694 	ip->ip_len -= hlen;
695 
696 #ifdef IPSEC
697 	/*
698 	 * enforce IPsec policy checking if we are seeing last header.
699 	 * note that we do not visit this with protocols with pcb layer
700 	 * code - like udp/tcp/raw ip.
701 	 */
702 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
703 	    ipsec4_in_reject(m, NULL)) {
704 		ipsecstat.in_polvio++;
705 		goto bad;
706 	}
707 #endif
708 #if FAST_IPSEC
709 	/*
710 	 * enforce IPsec policy checking if we are seeing last header.
711 	 * note that we do not visit this with protocols with pcb layer
712 	 * code - like udp/tcp/raw ip.
713 	 */
714 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
715 		/*
716 		 * Check if the packet has already had IPsec processing
717 		 * done.  If so, then just pass it along.  This tag gets
718 		 * set during AH, ESP, etc. input handling, before the
719 		 * packet is returned to the ip input queue for delivery.
720 		 */
721 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
722 		s = splnet();
723 		if (mtag != NULL) {
724 			tdbi = (struct tdb_ident *)(mtag + 1);
725 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
726 		} else {
727 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
728 						   IP_FORWARDING, &error);
729 		}
730 		if (sp != NULL) {
731 			/*
732 			 * Check security policy against packet attributes.
733 			 */
734 			error = ipsec_in_reject(sp, m);
735 			KEY_FREESP(&sp);
736 		} else {
737 			/* XXX error stat??? */
738 			error = EINVAL;
739 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
740 			goto bad;
741 		}
742 		splx(s);
743 		if (error)
744 			goto bad;
745 	}
746 #endif /* FAST_IPSEC */
747 
748 	/*
749 	 * Switch out to protocol's input routine.
750 	 */
751 	ipstat.ips_delivered++;
752 
753 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
754 	return;
755 bad:
756 	m_freem(m);
757 }
758 
759 /*
760  * Take incoming datagram fragment and try to reassemble it into
761  * whole datagram.  If the argument is the first fragment or one
762  * in between the function will return NULL and store the mbuf
763  * in the fragment chain.  If the argument is the last fragment
764  * the packet will be reassembled and the pointer to the new
765  * mbuf returned for further processing.  Only m_tags attached
766  * to the first packet/fragment are preserved.
767  * The IP header is *NOT* adjusted out of iplen.
768  */
769 
770 struct mbuf *
771 ip_reass(struct mbuf *m)
772 {
773 	struct ip *ip;
774 	struct mbuf *p, *q, *nq, *t;
775 	struct ipq *fp = NULL;
776 	struct ipqhead *head;
777 	int i, hlen, next;
778 	u_int8_t ecn, ecn0;
779 	u_short hash;
780 
781 	/* If maxnipq is 0, never accept fragments. */
782 	if (maxnipq == 0) {
783 		ipstat.ips_fragments++;
784 		ipstat.ips_fragdropped++;
785 		m_freem(m);
786 		return (NULL);
787 	}
788 
789 	ip = mtod(m, struct ip *);
790 	hlen = ip->ip_hl << 2;
791 
792 	hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
793 	head = &ipq[hash];
794 	IPQ_LOCK();
795 
796 	/*
797 	 * Look for queue of fragments
798 	 * of this datagram.
799 	 */
800 	TAILQ_FOREACH(fp, head, ipq_list)
801 		if (ip->ip_id == fp->ipq_id &&
802 		    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
803 		    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
804 #ifdef MAC
805 		    mac_fragment_match(m, fp) &&
806 #endif
807 		    ip->ip_p == fp->ipq_p)
808 			goto found;
809 
810 	fp = NULL;
811 
812 	/*
813 	 * Enforce upper bound on number of fragmented packets
814 	 * for which we attempt reassembly;
815 	 * If maxnipq is -1, accept all fragments without limitation.
816 	 */
817 	if ((nipq > maxnipq) && (maxnipq > 0)) {
818 		/*
819 		 * drop something from the tail of the current queue
820 		 * before proceeding further
821 		 */
822 		struct ipq *q = TAILQ_LAST(head, ipqhead);
823 		if (q == NULL) {   /* gak */
824 			for (i = 0; i < IPREASS_NHASH; i++) {
825 				struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
826 				if (r) {
827 					ipstat.ips_fragtimeout += r->ipq_nfrags;
828 					ip_freef(&ipq[i], r);
829 					break;
830 				}
831 			}
832 		} else {
833 			ipstat.ips_fragtimeout += q->ipq_nfrags;
834 			ip_freef(head, q);
835 		}
836 	}
837 
838 found:
839 	/*
840 	 * Adjust ip_len to not reflect header,
841 	 * convert offset of this to bytes.
842 	 */
843 	ip->ip_len -= hlen;
844 	if (ip->ip_off & IP_MF) {
845 		/*
846 		 * Make sure that fragments have a data length
847 		 * that's a non-zero multiple of 8 bytes.
848 		 */
849 		if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
850 			ipstat.ips_toosmall++; /* XXX */
851 			goto dropfrag;
852 		}
853 		m->m_flags |= M_FRAG;
854 	} else
855 		m->m_flags &= ~M_FRAG;
856 	ip->ip_off <<= 3;
857 
858 
859 	/*
860 	 * Attempt reassembly; if it succeeds, proceed.
861 	 * ip_reass() will return a different mbuf.
862 	 */
863 	ipstat.ips_fragments++;
864 	m->m_pkthdr.header = ip;
865 
866 	/* Previous ip_reass() started here. */
867 	/*
868 	 * Presence of header sizes in mbufs
869 	 * would confuse code below.
870 	 */
871 	m->m_data += hlen;
872 	m->m_len -= hlen;
873 
874 	/*
875 	 * If first fragment to arrive, create a reassembly queue.
876 	 */
877 	if (fp == NULL) {
878 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
879 			goto dropfrag;
880 		fp = mtod(t, struct ipq *);
881 #ifdef MAC
882 		if (mac_init_ipq(fp, M_NOWAIT) != 0) {
883 			m_free(t);
884 			goto dropfrag;
885 		}
886 		mac_create_ipq(m, fp);
887 #endif
888 		TAILQ_INSERT_HEAD(head, fp, ipq_list);
889 		nipq++;
890 		fp->ipq_nfrags = 1;
891 		fp->ipq_ttl = IPFRAGTTL;
892 		fp->ipq_p = ip->ip_p;
893 		fp->ipq_id = ip->ip_id;
894 		fp->ipq_src = ip->ip_src;
895 		fp->ipq_dst = ip->ip_dst;
896 		fp->ipq_frags = m;
897 		m->m_nextpkt = NULL;
898 		goto inserted;
899 	} else {
900 		fp->ipq_nfrags++;
901 #ifdef MAC
902 		mac_update_ipq(m, fp);
903 #endif
904 	}
905 
906 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
907 
908 	/*
909 	 * Handle ECN by comparing this segment with the first one;
910 	 * if CE is set, do not lose CE.
911 	 * drop if CE and not-ECT are mixed for the same packet.
912 	 */
913 	ecn = ip->ip_tos & IPTOS_ECN_MASK;
914 	ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
915 	if (ecn == IPTOS_ECN_CE) {
916 		if (ecn0 == IPTOS_ECN_NOTECT)
917 			goto dropfrag;
918 		if (ecn0 != IPTOS_ECN_CE)
919 			GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
920 	}
921 	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
922 		goto dropfrag;
923 
924 	/*
925 	 * Find a segment which begins after this one does.
926 	 */
927 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
928 		if (GETIP(q)->ip_off > ip->ip_off)
929 			break;
930 
931 	/*
932 	 * If there is a preceding segment, it may provide some of
933 	 * our data already.  If so, drop the data from the incoming
934 	 * segment.  If it provides all of our data, drop us, otherwise
935 	 * stick new segment in the proper place.
936 	 *
937 	 * If some of the data is dropped from the the preceding
938 	 * segment, then it's checksum is invalidated.
939 	 */
940 	if (p) {
941 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
942 		if (i > 0) {
943 			if (i >= ip->ip_len)
944 				goto dropfrag;
945 			m_adj(m, i);
946 			m->m_pkthdr.csum_flags = 0;
947 			ip->ip_off += i;
948 			ip->ip_len -= i;
949 		}
950 		m->m_nextpkt = p->m_nextpkt;
951 		p->m_nextpkt = m;
952 	} else {
953 		m->m_nextpkt = fp->ipq_frags;
954 		fp->ipq_frags = m;
955 	}
956 
957 	/*
958 	 * While we overlap succeeding segments trim them or,
959 	 * if they are completely covered, dequeue them.
960 	 */
961 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
962 	     q = nq) {
963 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
964 		if (i < GETIP(q)->ip_len) {
965 			GETIP(q)->ip_len -= i;
966 			GETIP(q)->ip_off += i;
967 			m_adj(q, i);
968 			q->m_pkthdr.csum_flags = 0;
969 			break;
970 		}
971 		nq = q->m_nextpkt;
972 		m->m_nextpkt = nq;
973 		ipstat.ips_fragdropped++;
974 		fp->ipq_nfrags--;
975 		m_freem(q);
976 	}
977 
978 inserted:
979 
980 	/*
981 	 * Check for complete reassembly and perform frag per packet
982 	 * limiting.
983 	 *
984 	 * Frag limiting is performed here so that the nth frag has
985 	 * a chance to complete the packet before we drop the packet.
986 	 * As a result, n+1 frags are actually allowed per packet, but
987 	 * only n will ever be stored. (n = maxfragsperpacket.)
988 	 *
989 	 */
990 	next = 0;
991 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
992 		if (GETIP(q)->ip_off != next) {
993 			if (fp->ipq_nfrags > maxfragsperpacket) {
994 				ipstat.ips_fragdropped += fp->ipq_nfrags;
995 				ip_freef(head, fp);
996 			}
997 			goto done;
998 		}
999 		next += GETIP(q)->ip_len;
1000 	}
1001 	/* Make sure the last packet didn't have the IP_MF flag */
1002 	if (p->m_flags & M_FRAG) {
1003 		if (fp->ipq_nfrags > maxfragsperpacket) {
1004 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1005 			ip_freef(head, fp);
1006 		}
1007 		goto done;
1008 	}
1009 
1010 	/*
1011 	 * Reassembly is complete.  Make sure the packet is a sane size.
1012 	 */
1013 	q = fp->ipq_frags;
1014 	ip = GETIP(q);
1015 	if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1016 		ipstat.ips_toolong++;
1017 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1018 		ip_freef(head, fp);
1019 		goto done;
1020 	}
1021 
1022 	/*
1023 	 * Concatenate fragments.
1024 	 */
1025 	m = q;
1026 	t = m->m_next;
1027 	m->m_next = NULL;
1028 	m_cat(m, t);
1029 	nq = q->m_nextpkt;
1030 	q->m_nextpkt = NULL;
1031 	for (q = nq; q != NULL; q = nq) {
1032 		nq = q->m_nextpkt;
1033 		q->m_nextpkt = NULL;
1034 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1035 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1036 		m_cat(m, q);
1037 	}
1038 #ifdef MAC
1039 	mac_create_datagram_from_ipq(fp, m);
1040 	mac_destroy_ipq(fp);
1041 #endif
1042 
1043 	/*
1044 	 * Create header for new ip packet by modifying header of first
1045 	 * packet;  dequeue and discard fragment reassembly header.
1046 	 * Make header visible.
1047 	 */
1048 	ip->ip_len = (ip->ip_hl << 2) + next;
1049 	ip->ip_src = fp->ipq_src;
1050 	ip->ip_dst = fp->ipq_dst;
1051 	TAILQ_REMOVE(head, fp, ipq_list);
1052 	nipq--;
1053 	(void) m_free(dtom(fp));
1054 	m->m_len += (ip->ip_hl << 2);
1055 	m->m_data -= (ip->ip_hl << 2);
1056 	/* some debugging cruft by sklower, below, will go away soon */
1057 	if (m->m_flags & M_PKTHDR)	/* XXX this should be done elsewhere */
1058 		m_fixhdr(m);
1059 	ipstat.ips_reassembled++;
1060 	IPQ_UNLOCK();
1061 	return (m);
1062 
1063 dropfrag:
1064 	ipstat.ips_fragdropped++;
1065 	if (fp != NULL)
1066 		fp->ipq_nfrags--;
1067 	m_freem(m);
1068 done:
1069 	IPQ_UNLOCK();
1070 	return (NULL);
1071 
1072 #undef GETIP
1073 }
1074 
1075 /*
1076  * Free a fragment reassembly header and all
1077  * associated datagrams.
1078  */
1079 static void
1080 ip_freef(fhp, fp)
1081 	struct ipqhead *fhp;
1082 	struct ipq *fp;
1083 {
1084 	register struct mbuf *q;
1085 
1086 	IPQ_LOCK_ASSERT();
1087 
1088 	while (fp->ipq_frags) {
1089 		q = fp->ipq_frags;
1090 		fp->ipq_frags = q->m_nextpkt;
1091 		m_freem(q);
1092 	}
1093 	TAILQ_REMOVE(fhp, fp, ipq_list);
1094 	(void) m_free(dtom(fp));
1095 	nipq--;
1096 }
1097 
1098 /*
1099  * IP timer processing;
1100  * if a timer expires on a reassembly
1101  * queue, discard it.
1102  */
1103 void
1104 ip_slowtimo()
1105 {
1106 	register struct ipq *fp;
1107 	int s = splnet();
1108 	int i;
1109 
1110 	IPQ_LOCK();
1111 	for (i = 0; i < IPREASS_NHASH; i++) {
1112 		for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1113 			struct ipq *fpp;
1114 
1115 			fpp = fp;
1116 			fp = TAILQ_NEXT(fp, ipq_list);
1117 			if(--fpp->ipq_ttl == 0) {
1118 				ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1119 				ip_freef(&ipq[i], fpp);
1120 			}
1121 		}
1122 	}
1123 	/*
1124 	 * If we are over the maximum number of fragments
1125 	 * (due to the limit being lowered), drain off
1126 	 * enough to get down to the new limit.
1127 	 */
1128 	if (maxnipq >= 0 && nipq > maxnipq) {
1129 		for (i = 0; i < IPREASS_NHASH; i++) {
1130 			while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1131 				ipstat.ips_fragdropped +=
1132 				    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1133 				ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1134 			}
1135 		}
1136 	}
1137 	IPQ_UNLOCK();
1138 	splx(s);
1139 }
1140 
1141 /*
1142  * Drain off all datagram fragments.
1143  */
1144 void
1145 ip_drain()
1146 {
1147 	int     i;
1148 
1149 	IPQ_LOCK();
1150 	for (i = 0; i < IPREASS_NHASH; i++) {
1151 		while(!TAILQ_EMPTY(&ipq[i])) {
1152 			ipstat.ips_fragdropped +=
1153 			    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1154 			ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1155 		}
1156 	}
1157 	IPQ_UNLOCK();
1158 	in_rtqdrain();
1159 }
1160 
1161 /*
1162  * The protocol to be inserted into ip_protox[] must be already registered
1163  * in inetsw[], either statically or through pf_proto_register().
1164  */
1165 int
1166 ipproto_register(u_char ipproto)
1167 {
1168 	struct protosw *pr;
1169 
1170 	/* Sanity checks. */
1171 	if (ipproto == 0)
1172 		return (EPROTONOSUPPORT);
1173 
1174 	/*
1175 	 * The protocol slot must not be occupied by another protocol
1176 	 * already.  An index pointing to IPPROTO_RAW is unused.
1177 	 */
1178 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1179 	if (pr == NULL)
1180 		return (EPFNOSUPPORT);
1181 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
1182 		return (EEXIST);
1183 
1184 	/* Find the protocol position in inetsw[] and set the index. */
1185 	for (pr = inetdomain.dom_protosw;
1186 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1187 		if (pr->pr_domain->dom_family == PF_INET &&
1188 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
1189 			/* Be careful to only index valid IP protocols. */
1190 			if (pr->pr_protocol <= IPPROTO_MAX) {
1191 				ip_protox[pr->pr_protocol] = pr - inetsw;
1192 				return (0);
1193 			} else
1194 				return (EINVAL);
1195 		}
1196 	}
1197 	return (EPROTONOSUPPORT);
1198 }
1199 
1200 int
1201 ipproto_unregister(u_char ipproto)
1202 {
1203 	struct protosw *pr;
1204 
1205 	/* Sanity checks. */
1206 	if (ipproto == 0)
1207 		return (EPROTONOSUPPORT);
1208 
1209 	/* Check if the protocol was indeed registered. */
1210 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1211 	if (pr == NULL)
1212 		return (EPFNOSUPPORT);
1213 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
1214 		return (ENOENT);
1215 
1216 	/* Reset the protocol slot to IPPROTO_RAW. */
1217 	ip_protox[ipproto] = pr - inetsw;
1218 	return (0);
1219 }
1220 
1221 
1222 /*
1223  * Do option processing on a datagram,
1224  * possibly discarding it if bad options are encountered,
1225  * or forwarding it if source-routed.
1226  * The pass argument is used when operating in the IPSTEALTH
1227  * mode to tell what options to process:
1228  * [LS]SRR (pass 0) or the others (pass 1).
1229  * The reason for as many as two passes is that when doing IPSTEALTH,
1230  * non-routing options should be processed only if the packet is for us.
1231  * Returns 1 if packet has been forwarded/freed,
1232  * 0 if the packet should be processed further.
1233  */
1234 static int
1235 ip_dooptions(struct mbuf *m, int pass)
1236 {
1237 	struct ip *ip = mtod(m, struct ip *);
1238 	u_char *cp;
1239 	struct in_ifaddr *ia;
1240 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1241 	struct in_addr *sin, dst;
1242 	n_time ntime;
1243 	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1244 
1245 	/* ignore or reject packets with IP options */
1246 	if (ip_doopts == 0)
1247 		return 0;
1248 	else if (ip_doopts == 2) {
1249 		type = ICMP_UNREACH;
1250 		code = ICMP_UNREACH_FILTER_PROHIB;
1251 		goto bad;
1252 	}
1253 
1254 	dst = ip->ip_dst;
1255 	cp = (u_char *)(ip + 1);
1256 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1257 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1258 		opt = cp[IPOPT_OPTVAL];
1259 		if (opt == IPOPT_EOL)
1260 			break;
1261 		if (opt == IPOPT_NOP)
1262 			optlen = 1;
1263 		else {
1264 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1265 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1266 				goto bad;
1267 			}
1268 			optlen = cp[IPOPT_OLEN];
1269 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1270 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1271 				goto bad;
1272 			}
1273 		}
1274 		switch (opt) {
1275 
1276 		default:
1277 			break;
1278 
1279 		/*
1280 		 * Source routing with record.
1281 		 * Find interface with current destination address.
1282 		 * If none on this machine then drop if strictly routed,
1283 		 * or do nothing if loosely routed.
1284 		 * Record interface address and bring up next address
1285 		 * component.  If strictly routed make sure next
1286 		 * address is on directly accessible net.
1287 		 */
1288 		case IPOPT_LSRR:
1289 		case IPOPT_SSRR:
1290 #ifdef IPSTEALTH
1291 			if (ipstealth && pass > 0)
1292 				break;
1293 #endif
1294 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1295 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1296 				goto bad;
1297 			}
1298 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1299 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1300 				goto bad;
1301 			}
1302 			ipaddr.sin_addr = ip->ip_dst;
1303 			ia = (struct in_ifaddr *)
1304 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1305 			if (ia == NULL) {
1306 				if (opt == IPOPT_SSRR) {
1307 					type = ICMP_UNREACH;
1308 					code = ICMP_UNREACH_SRCFAIL;
1309 					goto bad;
1310 				}
1311 				if (!ip_dosourceroute)
1312 					goto nosourcerouting;
1313 				/*
1314 				 * Loose routing, and not at next destination
1315 				 * yet; nothing to do except forward.
1316 				 */
1317 				break;
1318 			}
1319 			off--;			/* 0 origin */
1320 			if (off > optlen - (int)sizeof(struct in_addr)) {
1321 				/*
1322 				 * End of source route.  Should be for us.
1323 				 */
1324 				if (!ip_acceptsourceroute)
1325 					goto nosourcerouting;
1326 				save_rte(m, cp, ip->ip_src);
1327 				break;
1328 			}
1329 #ifdef IPSTEALTH
1330 			if (ipstealth)
1331 				goto dropit;
1332 #endif
1333 			if (!ip_dosourceroute) {
1334 				if (ipforwarding) {
1335 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1336 					/*
1337 					 * Acting as a router, so generate ICMP
1338 					 */
1339 nosourcerouting:
1340 					strcpy(buf, inet_ntoa(ip->ip_dst));
1341 					log(LOG_WARNING,
1342 					    "attempted source route from %s to %s\n",
1343 					    inet_ntoa(ip->ip_src), buf);
1344 					type = ICMP_UNREACH;
1345 					code = ICMP_UNREACH_SRCFAIL;
1346 					goto bad;
1347 				} else {
1348 					/*
1349 					 * Not acting as a router, so silently drop.
1350 					 */
1351 #ifdef IPSTEALTH
1352 dropit:
1353 #endif
1354 					ipstat.ips_cantforward++;
1355 					m_freem(m);
1356 					return (1);
1357 				}
1358 			}
1359 
1360 			/*
1361 			 * locate outgoing interface
1362 			 */
1363 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1364 			    sizeof(ipaddr.sin_addr));
1365 
1366 			if (opt == IPOPT_SSRR) {
1367 #define	INA	struct in_ifaddr *
1368 #define	SA	struct sockaddr *
1369 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == NULL)
1370 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1371 			} else
1372 				ia = ip_rtaddr(ipaddr.sin_addr);
1373 			if (ia == NULL) {
1374 				type = ICMP_UNREACH;
1375 				code = ICMP_UNREACH_SRCFAIL;
1376 				goto bad;
1377 			}
1378 			ip->ip_dst = ipaddr.sin_addr;
1379 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1380 			    sizeof(struct in_addr));
1381 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1382 			/*
1383 			 * Let ip_intr's mcast routing check handle mcast pkts
1384 			 */
1385 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1386 			break;
1387 
1388 		case IPOPT_RR:
1389 #ifdef IPSTEALTH
1390 			if (ipstealth && pass == 0)
1391 				break;
1392 #endif
1393 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1394 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1395 				goto bad;
1396 			}
1397 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1398 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1399 				goto bad;
1400 			}
1401 			/*
1402 			 * If no space remains, ignore.
1403 			 */
1404 			off--;			/* 0 origin */
1405 			if (off > optlen - (int)sizeof(struct in_addr))
1406 				break;
1407 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1408 			    sizeof(ipaddr.sin_addr));
1409 			/*
1410 			 * locate outgoing interface; if we're the destination,
1411 			 * use the incoming interface (should be same).
1412 			 */
1413 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1414 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
1415 				type = ICMP_UNREACH;
1416 				code = ICMP_UNREACH_HOST;
1417 				goto bad;
1418 			}
1419 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1420 			    sizeof(struct in_addr));
1421 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1422 			break;
1423 
1424 		case IPOPT_TS:
1425 #ifdef IPSTEALTH
1426 			if (ipstealth && pass == 0)
1427 				break;
1428 #endif
1429 			code = cp - (u_char *)ip;
1430 			if (optlen < 4 || optlen > 40) {
1431 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1432 				goto bad;
1433 			}
1434 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1435 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1436 				goto bad;
1437 			}
1438 			if (off > optlen - (int)sizeof(int32_t)) {
1439 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1440 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1441 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1442 					goto bad;
1443 				}
1444 				break;
1445 			}
1446 			off--;				/* 0 origin */
1447 			sin = (struct in_addr *)(cp + off);
1448 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1449 
1450 			case IPOPT_TS_TSONLY:
1451 				break;
1452 
1453 			case IPOPT_TS_TSANDADDR:
1454 				if (off + sizeof(n_time) +
1455 				    sizeof(struct in_addr) > optlen) {
1456 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1457 					goto bad;
1458 				}
1459 				ipaddr.sin_addr = dst;
1460 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1461 							    m->m_pkthdr.rcvif);
1462 				if (ia == NULL)
1463 					continue;
1464 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1465 				    sizeof(struct in_addr));
1466 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1467 				off += sizeof(struct in_addr);
1468 				break;
1469 
1470 			case IPOPT_TS_PRESPEC:
1471 				if (off + sizeof(n_time) +
1472 				    sizeof(struct in_addr) > optlen) {
1473 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1474 					goto bad;
1475 				}
1476 				(void)memcpy(&ipaddr.sin_addr, sin,
1477 				    sizeof(struct in_addr));
1478 				if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1479 					continue;
1480 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1481 				off += sizeof(struct in_addr);
1482 				break;
1483 
1484 			default:
1485 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1486 				goto bad;
1487 			}
1488 			ntime = iptime();
1489 			(void)memcpy(cp + off, &ntime, sizeof(n_time));
1490 			cp[IPOPT_OFFSET] += sizeof(n_time);
1491 		}
1492 	}
1493 	if (forward && ipforwarding) {
1494 		ip_forward(m, 1);
1495 		return (1);
1496 	}
1497 	return (0);
1498 bad:
1499 	icmp_error(m, type, code, 0, 0);
1500 	ipstat.ips_badoptions++;
1501 	return (1);
1502 }
1503 
1504 /*
1505  * Given address of next destination (final or next hop),
1506  * return internet address info of interface to be used to get there.
1507  */
1508 struct in_ifaddr *
1509 ip_rtaddr(dst)
1510 	struct in_addr dst;
1511 {
1512 	struct route sro;
1513 	struct sockaddr_in *sin;
1514 	struct in_ifaddr *ifa;
1515 
1516 	bzero(&sro, sizeof(sro));
1517 	sin = (struct sockaddr_in *)&sro.ro_dst;
1518 	sin->sin_family = AF_INET;
1519 	sin->sin_len = sizeof(*sin);
1520 	sin->sin_addr = dst;
1521 	rtalloc_ign(&sro, RTF_CLONING);
1522 
1523 	if (sro.ro_rt == NULL)
1524 		return (NULL);
1525 
1526 	ifa = ifatoia(sro.ro_rt->rt_ifa);
1527 	RTFREE(sro.ro_rt);
1528 	return (ifa);
1529 }
1530 
1531 /*
1532  * Save incoming source route for use in replies,
1533  * to be picked up later by ip_srcroute if the receiver is interested.
1534  */
1535 static void
1536 save_rte(m, option, dst)
1537 	struct mbuf *m;
1538 	u_char *option;
1539 	struct in_addr dst;
1540 {
1541 	unsigned olen;
1542 	struct ipopt_tag *opts;
1543 
1544 	opts = (struct ipopt_tag *)m_tag_get(PACKET_TAG_IPOPTIONS,
1545 					sizeof(struct ipopt_tag), M_NOWAIT);
1546 	if (opts == NULL)
1547 		return;
1548 
1549 	olen = option[IPOPT_OLEN];
1550 #ifdef DIAGNOSTIC
1551 	if (ipprintfs)
1552 		printf("save_rte: olen %d\n", olen);
1553 #endif
1554 	if (olen > sizeof(opts->ip_srcrt) - (1 + sizeof(dst)))
1555 		return;
1556 	bcopy(option, opts->ip_srcrt.srcopt, olen);
1557 	opts->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1558 	opts->ip_srcrt.dst = dst;
1559 	m_tag_prepend(m, (struct m_tag *)opts);
1560 }
1561 
1562 /*
1563  * Retrieve incoming source route for use in replies,
1564  * in the same form used by setsockopt.
1565  * The first hop is placed before the options, will be removed later.
1566  */
1567 struct mbuf *
1568 ip_srcroute(m0)
1569 	struct mbuf *m0;
1570 {
1571 	register struct in_addr *p, *q;
1572 	register struct mbuf *m;
1573 	struct ipopt_tag *opts;
1574 
1575 	opts = (struct ipopt_tag *)m_tag_find(m0, PACKET_TAG_IPOPTIONS, NULL);
1576 	if (opts == NULL)
1577 		return (NULL);
1578 
1579 	if (opts->ip_nhops == 0)
1580 		return (NULL);
1581 	m = m_get(M_DONTWAIT, MT_HEADER);
1582 	if (m == NULL)
1583 		return (NULL);
1584 
1585 #define OPTSIZ	(sizeof(opts->ip_srcrt.nop) + sizeof(opts->ip_srcrt.srcopt))
1586 
1587 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1588 	m->m_len = opts->ip_nhops * sizeof(struct in_addr) +
1589 	    sizeof(struct in_addr) + OPTSIZ;
1590 #ifdef DIAGNOSTIC
1591 	if (ipprintfs)
1592 		printf("ip_srcroute: nhops %d mlen %d", opts->ip_nhops, m->m_len);
1593 #endif
1594 
1595 	/*
1596 	 * First save first hop for return route
1597 	 */
1598 	p = &(opts->ip_srcrt.route[opts->ip_nhops - 1]);
1599 	*(mtod(m, struct in_addr *)) = *p--;
1600 #ifdef DIAGNOSTIC
1601 	if (ipprintfs)
1602 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1603 #endif
1604 
1605 	/*
1606 	 * Copy option fields and padding (nop) to mbuf.
1607 	 */
1608 	opts->ip_srcrt.nop = IPOPT_NOP;
1609 	opts->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1610 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1611 	    &(opts->ip_srcrt.nop), OPTSIZ);
1612 	q = (struct in_addr *)(mtod(m, caddr_t) +
1613 	    sizeof(struct in_addr) + OPTSIZ);
1614 #undef OPTSIZ
1615 	/*
1616 	 * Record return path as an IP source route,
1617 	 * reversing the path (pointers are now aligned).
1618 	 */
1619 	while (p >= opts->ip_srcrt.route) {
1620 #ifdef DIAGNOSTIC
1621 		if (ipprintfs)
1622 			printf(" %lx", (u_long)ntohl(q->s_addr));
1623 #endif
1624 		*q++ = *p--;
1625 	}
1626 	/*
1627 	 * Last hop goes to final destination.
1628 	 */
1629 	*q = opts->ip_srcrt.dst;
1630 #ifdef DIAGNOSTIC
1631 	if (ipprintfs)
1632 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1633 #endif
1634 	m_tag_delete(m0, (struct m_tag *)opts);
1635 	return (m);
1636 }
1637 
1638 /*
1639  * Strip out IP options, at higher
1640  * level protocol in the kernel.
1641  * Second argument is buffer to which options
1642  * will be moved, and return value is their length.
1643  * XXX should be deleted; last arg currently ignored.
1644  */
1645 void
1646 ip_stripoptions(m, mopt)
1647 	register struct mbuf *m;
1648 	struct mbuf *mopt;
1649 {
1650 	register int i;
1651 	struct ip *ip = mtod(m, struct ip *);
1652 	register caddr_t opts;
1653 	int olen;
1654 
1655 	olen = (ip->ip_hl << 2) - sizeof (struct ip);
1656 	opts = (caddr_t)(ip + 1);
1657 	i = m->m_len - (sizeof (struct ip) + olen);
1658 	bcopy(opts + olen, opts, (unsigned)i);
1659 	m->m_len -= olen;
1660 	if (m->m_flags & M_PKTHDR)
1661 		m->m_pkthdr.len -= olen;
1662 	ip->ip_v = IPVERSION;
1663 	ip->ip_hl = sizeof(struct ip) >> 2;
1664 }
1665 
1666 u_char inetctlerrmap[PRC_NCMDS] = {
1667 	0,		0,		0,		0,
1668 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1669 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1670 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1671 	0,		0,		EHOSTUNREACH,	0,
1672 	ENOPROTOOPT,	ECONNREFUSED
1673 };
1674 
1675 /*
1676  * Forward a packet.  If some error occurs return the sender
1677  * an icmp packet.  Note we can't always generate a meaningful
1678  * icmp message because icmp doesn't have a large enough repertoire
1679  * of codes and types.
1680  *
1681  * If not forwarding, just drop the packet.  This could be confusing
1682  * if ipforwarding was zero but some routing protocol was advancing
1683  * us as a gateway to somewhere.  However, we must let the routing
1684  * protocol deal with that.
1685  *
1686  * The srcrt parameter indicates whether the packet is being forwarded
1687  * via a source route.
1688  */
1689 void
1690 ip_forward(struct mbuf *m, int srcrt)
1691 {
1692 	struct ip *ip = mtod(m, struct ip *);
1693 	struct in_ifaddr *ia = NULL;
1694 	int error, type = 0, code = 0;
1695 	struct mbuf *mcopy;
1696 	struct in_addr dest;
1697 	struct ifnet *destifp, dummyifp;
1698 
1699 #ifdef DIAGNOSTIC
1700 	if (ipprintfs)
1701 		printf("forward: src %lx dst %lx ttl %x\n",
1702 		    (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1703 		    ip->ip_ttl);
1704 #endif
1705 
1706 
1707 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1708 		ipstat.ips_cantforward++;
1709 		m_freem(m);
1710 		return;
1711 	}
1712 #ifdef IPSTEALTH
1713 	if (!ipstealth) {
1714 #endif
1715 		if (ip->ip_ttl <= IPTTLDEC) {
1716 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1717 			    0, 0);
1718 			return;
1719 		}
1720 #ifdef IPSTEALTH
1721 	}
1722 #endif
1723 
1724 	if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) {
1725 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1726 		return;
1727 	}
1728 
1729 	/*
1730 	 * Save the IP header and at most 8 bytes of the payload,
1731 	 * in case we need to generate an ICMP message to the src.
1732 	 *
1733 	 * XXX this can be optimized a lot by saving the data in a local
1734 	 * buffer on the stack (72 bytes at most), and only allocating the
1735 	 * mbuf if really necessary. The vast majority of the packets
1736 	 * are forwarded without having to send an ICMP back (either
1737 	 * because unnecessary, or because rate limited), so we are
1738 	 * really we are wasting a lot of work here.
1739 	 *
1740 	 * We don't use m_copy() because it might return a reference
1741 	 * to a shared cluster. Both this function and ip_output()
1742 	 * assume exclusive access to the IP header in `m', so any
1743 	 * data in a cluster may change before we reach icmp_error().
1744 	 */
1745 	MGET(mcopy, M_DONTWAIT, m->m_type);
1746 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1747 		/*
1748 		 * It's probably ok if the pkthdr dup fails (because
1749 		 * the deep copy of the tag chain failed), but for now
1750 		 * be conservative and just discard the copy since
1751 		 * code below may some day want the tags.
1752 		 */
1753 		m_free(mcopy);
1754 		mcopy = NULL;
1755 	}
1756 	if (mcopy != NULL) {
1757 		mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1758 		    (int)ip->ip_len);
1759 		mcopy->m_pkthdr.len = mcopy->m_len;
1760 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1761 	}
1762 
1763 #ifdef IPSTEALTH
1764 	if (!ipstealth) {
1765 #endif
1766 		ip->ip_ttl -= IPTTLDEC;
1767 #ifdef IPSTEALTH
1768 	}
1769 #endif
1770 
1771 	/*
1772 	 * If forwarding packet using same interface that it came in on,
1773 	 * perhaps should send a redirect to sender to shortcut a hop.
1774 	 * Only send redirect if source is sending directly to us,
1775 	 * and if packet was not source routed (or has any options).
1776 	 * Also, don't send redirect if forwarding using a default route
1777 	 * or a route modified by a redirect.
1778 	 */
1779 	dest.s_addr = 0;
1780 	if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1781 		struct sockaddr_in *sin;
1782 		struct route ro;
1783 		struct rtentry *rt;
1784 
1785 		bzero(&ro, sizeof(ro));
1786 		sin = (struct sockaddr_in *)&ro.ro_dst;
1787 		sin->sin_family = AF_INET;
1788 		sin->sin_len = sizeof(*sin);
1789 		sin->sin_addr = ip->ip_dst;
1790 		rtalloc_ign(&ro, RTF_CLONING);
1791 
1792 		rt = ro.ro_rt;
1793 
1794 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1795 		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1796 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1797 			u_long src = ntohl(ip->ip_src.s_addr);
1798 
1799 			if (RTA(rt) &&
1800 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1801 				if (rt->rt_flags & RTF_GATEWAY)
1802 					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1803 				else
1804 					dest.s_addr = ip->ip_dst.s_addr;
1805 				/* Router requirements says to only send host redirects */
1806 				type = ICMP_REDIRECT;
1807 				code = ICMP_REDIRECT_HOST;
1808 #ifdef DIAGNOSTIC
1809 				if (ipprintfs)
1810 					printf("redirect (%d) to %lx\n", code, (u_long)dest.s_addr);
1811 #endif
1812 			}
1813 		}
1814 		if (rt)
1815 			RTFREE(rt);
1816 	}
1817 
1818 	error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
1819 	if (error)
1820 		ipstat.ips_cantforward++;
1821 	else {
1822 		ipstat.ips_forward++;
1823 		if (type)
1824 			ipstat.ips_redirectsent++;
1825 		else {
1826 			if (mcopy)
1827 				m_freem(mcopy);
1828 			return;
1829 		}
1830 	}
1831 	if (mcopy == NULL)
1832 		return;
1833 	destifp = NULL;
1834 
1835 	switch (error) {
1836 
1837 	case 0:				/* forwarded, but need redirect */
1838 		/* type, code set above */
1839 		break;
1840 
1841 	case ENETUNREACH:		/* shouldn't happen, checked above */
1842 	case EHOSTUNREACH:
1843 	case ENETDOWN:
1844 	case EHOSTDOWN:
1845 	default:
1846 		type = ICMP_UNREACH;
1847 		code = ICMP_UNREACH_HOST;
1848 		break;
1849 
1850 	case EMSGSIZE:
1851 		type = ICMP_UNREACH;
1852 		code = ICMP_UNREACH_NEEDFRAG;
1853 #if defined(IPSEC) || defined(FAST_IPSEC)
1854 		/*
1855 		 * If the packet is routed over IPsec tunnel, tell the
1856 		 * originator the tunnel MTU.
1857 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1858 		 * XXX quickhack!!!
1859 		 */
1860 		{
1861 			struct secpolicy *sp = NULL;
1862 			int ipsecerror;
1863 			int ipsechdr;
1864 			struct route *ro;
1865 
1866 #ifdef IPSEC
1867 			sp = ipsec4_getpolicybyaddr(mcopy,
1868 						    IPSEC_DIR_OUTBOUND,
1869 						    IP_FORWARDING,
1870 						    &ipsecerror);
1871 #else /* FAST_IPSEC */
1872 			sp = ipsec_getpolicybyaddr(mcopy,
1873 						   IPSEC_DIR_OUTBOUND,
1874 						   IP_FORWARDING,
1875 						   &ipsecerror);
1876 #endif
1877 			if (sp != NULL) {
1878 				/* count IPsec header size */
1879 				ipsechdr = ipsec4_hdrsiz(mcopy,
1880 							 IPSEC_DIR_OUTBOUND,
1881 							 NULL);
1882 
1883 				/*
1884 				 * find the correct route for outer IPv4
1885 				 * header, compute tunnel MTU.
1886 				 *
1887 				 * XXX BUG ALERT
1888 				 * The "dummyifp" code relies upon the fact
1889 				 * that icmp_error() touches only ifp->if_mtu.
1890 				 */
1891 				/*XXX*/
1892 				destifp = NULL;
1893 				if (sp->req != NULL
1894 				 && sp->req->sav != NULL
1895 				 && sp->req->sav->sah != NULL) {
1896 					ro = &sp->req->sav->sah->sa_route;
1897 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1898 						dummyifp.if_mtu =
1899 						    ro->ro_rt->rt_rmx.rmx_mtu ?
1900 						    ro->ro_rt->rt_rmx.rmx_mtu :
1901 						    ro->ro_rt->rt_ifp->if_mtu;
1902 						dummyifp.if_mtu -= ipsechdr;
1903 						destifp = &dummyifp;
1904 					}
1905 				}
1906 
1907 #ifdef IPSEC
1908 				key_freesp(sp);
1909 #else /* FAST_IPSEC */
1910 				KEY_FREESP(&sp);
1911 #endif
1912 				ipstat.ips_cantfrag++;
1913 				break;
1914 			} else
1915 #endif /*IPSEC || FAST_IPSEC*/
1916 		/*
1917 		 * When doing source routing 'ia' can be NULL.  Fall back
1918 		 * to the minimum guaranteed routeable packet size and use
1919 		 * the same hack as IPSEC to setup a dummyifp for icmp.
1920 		 */
1921 		if (ia == NULL) {
1922 			dummyifp.if_mtu = IP_MSS;
1923 			destifp = &dummyifp;
1924 		} else
1925 			destifp = ia->ia_ifp;
1926 #if defined(IPSEC) || defined(FAST_IPSEC)
1927 		}
1928 #endif /*IPSEC || FAST_IPSEC*/
1929 		ipstat.ips_cantfrag++;
1930 		break;
1931 
1932 	case ENOBUFS:
1933 		/*
1934 		 * A router should not generate ICMP_SOURCEQUENCH as
1935 		 * required in RFC1812 Requirements for IP Version 4 Routers.
1936 		 * Source quench could be a big problem under DoS attacks,
1937 		 * or if the underlying interface is rate-limited.
1938 		 * Those who need source quench packets may re-enable them
1939 		 * via the net.inet.ip.sendsourcequench sysctl.
1940 		 */
1941 		if (ip_sendsourcequench == 0) {
1942 			m_freem(mcopy);
1943 			return;
1944 		} else {
1945 			type = ICMP_SOURCEQUENCH;
1946 			code = 0;
1947 		}
1948 		break;
1949 
1950 	case EACCES:			/* ipfw denied packet */
1951 		m_freem(mcopy);
1952 		return;
1953 	}
1954 	icmp_error(mcopy, type, code, dest.s_addr, destifp);
1955 }
1956 
1957 void
1958 ip_savecontrol(inp, mp, ip, m)
1959 	register struct inpcb *inp;
1960 	register struct mbuf **mp;
1961 	register struct ip *ip;
1962 	register struct mbuf *m;
1963 {
1964 	if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1965 		struct bintime bt;
1966 
1967 		bintime(&bt);
1968 		if (inp->inp_socket->so_options & SO_BINTIME) {
1969 			*mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1970 			SCM_BINTIME, SOL_SOCKET);
1971 			if (*mp)
1972 				mp = &(*mp)->m_next;
1973 		}
1974 		if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1975 			struct timeval tv;
1976 
1977 			bintime2timeval(&bt, &tv);
1978 			*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1979 				SCM_TIMESTAMP, SOL_SOCKET);
1980 			if (*mp)
1981 				mp = &(*mp)->m_next;
1982 		}
1983 	}
1984 	if (inp->inp_flags & INP_RECVDSTADDR) {
1985 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1986 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1987 		if (*mp)
1988 			mp = &(*mp)->m_next;
1989 	}
1990 	if (inp->inp_flags & INP_RECVTTL) {
1991 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1992 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1993 		if (*mp)
1994 			mp = &(*mp)->m_next;
1995 	}
1996 #ifdef notyet
1997 	/* XXX
1998 	 * Moving these out of udp_input() made them even more broken
1999 	 * than they already were.
2000 	 */
2001 	/* options were tossed already */
2002 	if (inp->inp_flags & INP_RECVOPTS) {
2003 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2004 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2005 		if (*mp)
2006 			mp = &(*mp)->m_next;
2007 	}
2008 	/* ip_srcroute doesn't do what we want here, need to fix */
2009 	if (inp->inp_flags & INP_RECVRETOPTS) {
2010 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2011 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2012 		if (*mp)
2013 			mp = &(*mp)->m_next;
2014 	}
2015 #endif
2016 	if (inp->inp_flags & INP_RECVIF) {
2017 		struct ifnet *ifp;
2018 		struct sdlbuf {
2019 			struct sockaddr_dl sdl;
2020 			u_char	pad[32];
2021 		} sdlbuf;
2022 		struct sockaddr_dl *sdp;
2023 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2024 
2025 		if (((ifp = m->m_pkthdr.rcvif))
2026 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
2027 			sdp = (struct sockaddr_dl *)
2028 			    (ifaddr_byindex(ifp->if_index)->ifa_addr);
2029 			/*
2030 			 * Change our mind and don't try copy.
2031 			 */
2032 			if ((sdp->sdl_family != AF_LINK)
2033 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
2034 				goto makedummy;
2035 			}
2036 			bcopy(sdp, sdl2, sdp->sdl_len);
2037 		} else {
2038 makedummy:
2039 			sdl2->sdl_len
2040 				= offsetof(struct sockaddr_dl, sdl_data[0]);
2041 			sdl2->sdl_family = AF_LINK;
2042 			sdl2->sdl_index = 0;
2043 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2044 		}
2045 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2046 			IP_RECVIF, IPPROTO_IP);
2047 		if (*mp)
2048 			mp = &(*mp)->m_next;
2049 	}
2050 }
2051 
2052 /*
2053  * XXX these routines are called from the upper part of the kernel.
2054  * They need to be locked when we remove Giant.
2055  *
2056  * They could also be moved to ip_mroute.c, since all the RSVP
2057  *  handling is done there already.
2058  */
2059 static int ip_rsvp_on;
2060 struct socket *ip_rsvpd;
2061 int
2062 ip_rsvp_init(struct socket *so)
2063 {
2064 	if (so->so_type != SOCK_RAW ||
2065 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2066 		return EOPNOTSUPP;
2067 
2068 	if (ip_rsvpd != NULL)
2069 		return EADDRINUSE;
2070 
2071 	ip_rsvpd = so;
2072 	/*
2073 	 * This may seem silly, but we need to be sure we don't over-increment
2074 	 * the RSVP counter, in case something slips up.
2075 	 */
2076 	if (!ip_rsvp_on) {
2077 		ip_rsvp_on = 1;
2078 		rsvp_on++;
2079 	}
2080 
2081 	return 0;
2082 }
2083 
2084 int
2085 ip_rsvp_done(void)
2086 {
2087 	ip_rsvpd = NULL;
2088 	/*
2089 	 * This may seem silly, but we need to be sure we don't over-decrement
2090 	 * the RSVP counter, in case something slips up.
2091 	 */
2092 	if (ip_rsvp_on) {
2093 		ip_rsvp_on = 0;
2094 		rsvp_on--;
2095 	}
2096 	return 0;
2097 }
2098 
2099 void
2100 rsvp_input(struct mbuf *m, int off)	/* XXX must fixup manually */
2101 {
2102 	if (rsvp_input_p) { /* call the real one if loaded */
2103 		rsvp_input_p(m, off);
2104 		return;
2105 	}
2106 
2107 	/* Can still get packets with rsvp_on = 0 if there is a local member
2108 	 * of the group to which the RSVP packet is addressed.  But in this
2109 	 * case we want to throw the packet away.
2110 	 */
2111 
2112 	if (!rsvp_on) {
2113 		m_freem(m);
2114 		return;
2115 	}
2116 
2117 	if (ip_rsvpd != NULL) {
2118 		rip_input(m, off);
2119 		return;
2120 	}
2121 	/* Drop the packet */
2122 	m_freem(m);
2123 }
2124