xref: /netbsd-src/sys/netinet/ip_icmp.c (revision ca453df649ce9db45b64d73678ba06cbccf9aa11)
1 /*	$NetBSD: ip_icmp.c,v 1.125 2011/07/17 20:54:53 joerg Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1998, 2000 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Public Access Networks Corporation ("Panix").  It was developed under
38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39  *
40  * This code is derived from software contributed to The NetBSD Foundation
41  * by Jason R. Thorpe of Zembu Labs, Inc.
42  *
43  * Redistribution and use in source and binary forms, with or without
44  * modification, are permitted provided that the following conditions
45  * are met:
46  * 1. Redistributions of source code must retain the above copyright
47  *    notice, this list of conditions and the following disclaimer.
48  * 2. Redistributions in binary form must reproduce the above copyright
49  *    notice, this list of conditions and the following disclaimer in the
50  *    documentation and/or other materials provided with the distribution.
51  *
52  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
53  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
54  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
55  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
56  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
57  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
58  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
59  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
60  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
61  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
62  * POSSIBILITY OF SUCH DAMAGE.
63  */
64 
65 /*
66  * Copyright (c) 1982, 1986, 1988, 1993
67  *	The Regents of the University of California.  All rights reserved.
68  *
69  * Redistribution and use in source and binary forms, with or without
70  * modification, are permitted provided that the following conditions
71  * are met:
72  * 1. Redistributions of source code must retain the above copyright
73  *    notice, this list of conditions and the following disclaimer.
74  * 2. Redistributions in binary form must reproduce the above copyright
75  *    notice, this list of conditions and the following disclaimer in the
76  *    documentation and/or other materials provided with the distribution.
77  * 3. Neither the name of the University nor the names of its contributors
78  *    may be used to endorse or promote products derived from this software
79  *    without specific prior written permission.
80  *
81  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
82  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
83  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
84  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
85  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
86  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
87  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
88  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
89  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
90  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
91  * SUCH DAMAGE.
92  *
93  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
94  */
95 
96 #include <sys/cdefs.h>
97 __KERNEL_RCSID(0, "$NetBSD: ip_icmp.c,v 1.125 2011/07/17 20:54:53 joerg Exp $");
98 
99 #include "opt_ipsec.h"
100 
101 #include <sys/param.h>
102 #include <sys/systm.h>
103 #include <sys/malloc.h>
104 #include <sys/mbuf.h>
105 #include <sys/protosw.h>
106 #include <sys/socket.h>
107 #include <sys/time.h>
108 #include <sys/kernel.h>
109 #include <sys/syslog.h>
110 #include <sys/sysctl.h>
111 
112 #include <net/if.h>
113 #include <net/route.h>
114 
115 #include <netinet/in.h>
116 #include <netinet/in_systm.h>
117 #include <netinet/in_var.h>
118 #include <netinet/ip.h>
119 #include <netinet/ip_icmp.h>
120 #include <netinet/ip_var.h>
121 #include <netinet/in_pcb.h>
122 #include <netinet/in_proto.h>
123 #include <netinet/icmp_var.h>
124 #include <netinet/icmp_private.h>
125 
126 #ifdef IPSEC
127 #include <netinet6/ipsec.h>
128 #include <netkey/key.h>
129 #endif
130 
131 #ifdef FAST_IPSEC
132 #include <netipsec/ipsec.h>
133 #include <netipsec/key.h>
134 #endif	/* FAST_IPSEC*/
135 
136 /*
137  * ICMP routines: error generation, receive packet processing, and
138  * routines to turnaround packets back to the originator, and
139  * host table maintenance routines.
140  */
141 
142 int	icmpmaskrepl = 0;
143 int	icmpbmcastecho = 0;
144 #ifdef ICMPPRINTFS
145 int	icmpprintfs = 0;
146 #endif
147 int	icmpreturndatabytes = 8;
148 
149 percpu_t *icmpstat_percpu;
150 
151 /*
152  * List of callbacks to notify when Path MTU changes are made.
153  */
154 struct icmp_mtudisc_callback {
155 	LIST_ENTRY(icmp_mtudisc_callback) mc_list;
156 	void (*mc_func)(struct in_addr);
157 };
158 
159 LIST_HEAD(, icmp_mtudisc_callback) icmp_mtudisc_callbacks =
160     LIST_HEAD_INITIALIZER(&icmp_mtudisc_callbacks);
161 
162 #if 0
163 static u_int	ip_next_mtu(u_int, int);
164 #else
165 /*static*/ u_int	ip_next_mtu(u_int, int);
166 #endif
167 
168 extern int icmperrppslim;
169 static int icmperrpps_count = 0;
170 static struct timeval icmperrppslim_last;
171 static int icmp_rediraccept = 1;
172 static int icmp_redirtimeout = 600;
173 static struct rttimer_queue *icmp_redirect_timeout_q = NULL;
174 
175 static void icmp_mtudisc_timeout(struct rtentry *, struct rttimer *);
176 static void icmp_redirect_timeout(struct rtentry *, struct rttimer *);
177 
178 static void sysctl_netinet_icmp_setup(struct sysctllog **);
179 
180 void
181 icmp_init(void)
182 {
183 
184 	sysctl_netinet_icmp_setup(NULL);
185 
186 	/*
187 	 * This is only useful if the user initializes redirtimeout to
188 	 * something other than zero.
189 	 */
190 	if (icmp_redirtimeout != 0) {
191 		icmp_redirect_timeout_q =
192 			rt_timer_queue_create(icmp_redirtimeout);
193 	}
194 
195 	icmpstat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP_NSTATS);
196 }
197 
198 /*
199  * Register a Path MTU Discovery callback.
200  */
201 void
202 icmp_mtudisc_callback_register(void (*func)(struct in_addr))
203 {
204 	struct icmp_mtudisc_callback *mc;
205 
206 	for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL;
207 	     mc = LIST_NEXT(mc, mc_list)) {
208 		if (mc->mc_func == func)
209 			return;
210 	}
211 
212 	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
213 	if (mc == NULL)
214 		panic("icmp_mtudisc_callback_register");
215 
216 	mc->mc_func = func;
217 	LIST_INSERT_HEAD(&icmp_mtudisc_callbacks, mc, mc_list);
218 }
219 
220 /*
221  * Generate an error packet of type error
222  * in response to bad packet ip.
223  */
224 void
225 icmp_error(struct mbuf *n, int type, int code, n_long dest,
226     int destmtu)
227 {
228 	struct ip *oip = mtod(n, struct ip *), *nip;
229 	unsigned oiplen = oip->ip_hl << 2;
230 	struct icmp *icp;
231 	struct mbuf *m;
232 	struct m_tag *mtag;
233 	unsigned icmplen, mblen;
234 
235 #ifdef ICMPPRINTFS
236 	if (icmpprintfs)
237 		printf("icmp_error(%p, type:%d, code:%d)\n", oip, type, code);
238 #endif
239 	if (type != ICMP_REDIRECT)
240 		ICMP_STATINC(ICMP_STAT_ERROR);
241 	/*
242 	 * Don't send error if the original packet was encrypted.
243 	 * Don't send error if not the first fragment of message.
244 	 * Don't error if the old packet protocol was ICMP
245 	 * error message, only known informational types.
246 	 */
247 	if (n->m_flags & M_DECRYPTED)
248 		goto freeit;
249 	if (oip->ip_off &~ htons(IP_MF|IP_DF))
250 		goto freeit;
251 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
252 	  n->m_len >= oiplen + ICMP_MINLEN &&
253 	  !ICMP_INFOTYPE(((struct icmp *)((char *)oip + oiplen))->icmp_type)) {
254 		ICMP_STATINC(ICMP_STAT_OLDICMP);
255 		goto freeit;
256 	}
257 	/* Don't send error in response to a multicast or broadcast packet */
258 	if (n->m_flags & (M_BCAST|M_MCAST))
259 		goto freeit;
260 
261 	/*
262 	 * First, do a rate limitation check.
263 	 */
264 	if (icmp_ratelimit(&oip->ip_src, type, code)) {
265 		/* XXX stat */
266 		goto freeit;
267 	}
268 
269 	/*
270 	 * Now, formulate icmp message
271 	 */
272 	icmplen = oiplen + min(icmpreturndatabytes,
273 	    ntohs(oip->ip_len) - oiplen);
274 	/*
275 	 * Defend against mbuf chains shorter than oip->ip_len - oiplen:
276 	 */
277 	mblen = 0;
278 	for (m = n; m && (mblen < icmplen); m = m->m_next)
279 		mblen += m->m_len;
280 	icmplen = min(mblen, icmplen);
281 
282 	/*
283 	 * As we are not required to return everything we have,
284 	 * we return whatever we can return at ease.
285 	 *
286 	 * Note that ICMP datagrams longer than 576 octets are out of spec
287 	 * according to RFC1812; the limit on icmpreturndatabytes below in
288 	 * icmp_sysctl will keep things below that limit.
289 	 */
290 
291 	KASSERT(ICMP_MINLEN <= MCLBYTES);
292 
293 	if (icmplen + ICMP_MINLEN > MCLBYTES)
294 		icmplen = MCLBYTES - ICMP_MINLEN;
295 
296 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
297 	if (m && (icmplen + ICMP_MINLEN > MHLEN)) {
298 		MCLGET(m, M_DONTWAIT);
299 		if ((m->m_flags & M_EXT) == 0) {
300 			m_freem(m);
301 			m = NULL;
302 		}
303 	}
304 	if (m == NULL)
305 		goto freeit;
306 	MCLAIM(m, n->m_owner);
307 	m->m_len = icmplen + ICMP_MINLEN;
308 	if ((m->m_flags & M_EXT) == 0)
309 		MH_ALIGN(m, m->m_len);
310 	else {
311 		m->m_data += sizeof(struct ip);
312 		m->m_len -= sizeof(struct ip);
313 	}
314 	icp = mtod(m, struct icmp *);
315 	if ((u_int)type > ICMP_MAXTYPE)
316 		panic("icmp_error");
317 	ICMP_STATINC(ICMP_STAT_OUTHIST + type);
318 	icp->icmp_type = type;
319 	if (type == ICMP_REDIRECT)
320 		icp->icmp_gwaddr.s_addr = dest;
321 	else {
322 		icp->icmp_void = 0;
323 		/*
324 		 * The following assignments assume an overlay with the
325 		 * zeroed icmp_void field.
326 		 */
327 		if (type == ICMP_PARAMPROB) {
328 			icp->icmp_pptr = code;
329 			code = 0;
330 		} else if (type == ICMP_UNREACH &&
331 		    code == ICMP_UNREACH_NEEDFRAG && destmtu)
332 			icp->icmp_nextmtu = htons(destmtu);
333 	}
334 
335 	icp->icmp_code = code;
336 	m_copydata(n, 0, icmplen, (void *)&icp->icmp_ip);
337 
338 	/*
339 	 * Now, copy old ip header (without options)
340 	 * in front of icmp message.
341 	 */
342 	if ((m->m_flags & M_EXT) == 0 &&
343 	    m->m_data - sizeof(struct ip) < m->m_pktdat)
344 		panic("icmp len");
345 	m->m_data -= sizeof(struct ip);
346 	m->m_len += sizeof(struct ip);
347 	m->m_pkthdr.len = m->m_len;
348 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
349 	nip = mtod(m, struct ip *);
350 	/* ip_v set in ip_output */
351 	nip->ip_hl = sizeof(struct ip) >> 2;
352 	nip->ip_tos = 0;
353 	nip->ip_len = htons(m->m_len);
354 	/* ip_id set in ip_output */
355 	nip->ip_off = htons(0);
356 	/* ip_ttl set in icmp_reflect */
357 	nip->ip_p = IPPROTO_ICMP;
358 	nip->ip_src = oip->ip_src;
359 	nip->ip_dst = oip->ip_dst;
360 	/* move PF m_tag to new packet, if it exists */
361 	mtag = m_tag_find(n, PACKET_TAG_PF, NULL);
362 	if (mtag != NULL) {
363 		m_tag_unlink(n, mtag);
364 		m_tag_prepend(m, mtag);
365 	}
366 	icmp_reflect(m);
367 
368 freeit:
369 	m_freem(n);
370 }
371 
372 struct sockaddr_in icmpsrc = {
373 	.sin_len = sizeof (struct sockaddr_in),
374 	.sin_family = AF_INET,
375 };
376 static struct sockaddr_in icmpdst = {
377 	.sin_len = sizeof (struct sockaddr_in),
378 	.sin_family = AF_INET,
379 };
380 static struct sockaddr_in icmpgw = {
381 	.sin_len = sizeof (struct sockaddr_in),
382 	.sin_family = AF_INET,
383 };
384 struct sockaddr_in icmpmask = {
385 	.sin_len = 8,
386 	.sin_family = 0,
387 };
388 
389 /*
390  * Process a received ICMP message.
391  */
392 void
393 icmp_input(struct mbuf *m, ...)
394 {
395 	int proto;
396 	struct icmp *icp;
397 	struct ip *ip = mtod(m, struct ip *);
398 	int icmplen;
399 	int i;
400 	struct in_ifaddr *ia;
401 	void *(*ctlfunc)(int, const struct sockaddr *, void *);
402 	int code;
403 	int hlen;
404 	va_list ap;
405 	struct rtentry *rt;
406 
407 	va_start(ap, m);
408 	hlen = va_arg(ap, int);
409 	proto = va_arg(ap, int);
410 	va_end(ap);
411 
412 	/*
413 	 * Locate icmp structure in mbuf, and check
414 	 * that not corrupted and of at least minimum length.
415 	 */
416 	icmplen = ntohs(ip->ip_len) - hlen;
417 #ifdef ICMPPRINTFS
418 	if (icmpprintfs) {
419 		printf("icmp_input from `%s' to ", inet_ntoa(ip->ip_src));
420 		printf("`%s', len %d\n", inet_ntoa(ip->ip_dst), icmplen);
421 	}
422 #endif
423 	if (icmplen < ICMP_MINLEN) {
424 		ICMP_STATINC(ICMP_STAT_TOOSHORT);
425 		goto freeit;
426 	}
427 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
428 	if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) {
429 		ICMP_STATINC(ICMP_STAT_TOOSHORT);
430 		return;
431 	}
432 	ip = mtod(m, struct ip *);
433 	m->m_len -= hlen;
434 	m->m_data += hlen;
435 	icp = mtod(m, struct icmp *);
436 	/* Don't need to assert alignment, here. */
437 	if (in_cksum(m, icmplen)) {
438 		ICMP_STATINC(ICMP_STAT_CHECKSUM);
439 		goto freeit;
440 	}
441 	m->m_len += hlen;
442 	m->m_data -= hlen;
443 
444 #ifdef ICMPPRINTFS
445 	/*
446 	 * Message type specific processing.
447 	 */
448 	if (icmpprintfs)
449 		printf("icmp_input(type:%d, code:%d)\n", icp->icmp_type,
450 		    icp->icmp_code);
451 #endif
452 	if (icp->icmp_type > ICMP_MAXTYPE)
453 		goto raw;
454 	ICMP_STATINC(ICMP_STAT_INHIST + icp->icmp_type);
455 	code = icp->icmp_code;
456 	switch (icp->icmp_type) {
457 
458 	case ICMP_UNREACH:
459 		switch (code) {
460 			case ICMP_UNREACH_NET:
461 				code = PRC_UNREACH_NET;
462 				break;
463 
464 			case ICMP_UNREACH_HOST:
465 				code = PRC_UNREACH_HOST;
466 				break;
467 
468 			case ICMP_UNREACH_PROTOCOL:
469 				code = PRC_UNREACH_PROTOCOL;
470 				break;
471 
472 			case ICMP_UNREACH_PORT:
473 				code = PRC_UNREACH_PORT;
474 				break;
475 
476 			case ICMP_UNREACH_SRCFAIL:
477 				code = PRC_UNREACH_SRCFAIL;
478 				break;
479 
480 			case ICMP_UNREACH_NEEDFRAG:
481 				code = PRC_MSGSIZE;
482 				break;
483 
484 			case ICMP_UNREACH_NET_UNKNOWN:
485 			case ICMP_UNREACH_NET_PROHIB:
486 			case ICMP_UNREACH_TOSNET:
487 				code = PRC_UNREACH_NET;
488 				break;
489 
490 			case ICMP_UNREACH_HOST_UNKNOWN:
491 			case ICMP_UNREACH_ISOLATED:
492 			case ICMP_UNREACH_HOST_PROHIB:
493 			case ICMP_UNREACH_TOSHOST:
494 				code = PRC_UNREACH_HOST;
495 				break;
496 
497 			default:
498 				goto badcode;
499 		}
500 		goto deliver;
501 
502 	case ICMP_TIMXCEED:
503 		if (code > 1)
504 			goto badcode;
505 		code += PRC_TIMXCEED_INTRANS;
506 		goto deliver;
507 
508 	case ICMP_PARAMPROB:
509 		if (code > 1)
510 			goto badcode;
511 		code = PRC_PARAMPROB;
512 		goto deliver;
513 
514 	case ICMP_SOURCEQUENCH:
515 		if (code)
516 			goto badcode;
517 		code = PRC_QUENCH;
518 		goto deliver;
519 
520 	deliver:
521 		/*
522 		 * Problem with datagram; advise higher level routines.
523 		 */
524 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
525 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
526 			ICMP_STATINC(ICMP_STAT_BADLEN);
527 			goto freeit;
528 		}
529 		if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr))
530 			goto badcode;
531 #ifdef ICMPPRINTFS
532 		if (icmpprintfs)
533 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
534 #endif
535 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
536 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
537 		if (ctlfunc)
538 			(void) (*ctlfunc)(code, sintosa(&icmpsrc),
539 			    &icp->icmp_ip);
540 		break;
541 
542 	badcode:
543 		ICMP_STATINC(ICMP_STAT_BADCODE);
544 		break;
545 
546 	case ICMP_ECHO:
547 		if (!icmpbmcastecho &&
548 		    (m->m_flags & (M_MCAST | M_BCAST)) != 0)  {
549 			ICMP_STATINC(ICMP_STAT_BMCASTECHO);
550 			break;
551 		}
552 		icp->icmp_type = ICMP_ECHOREPLY;
553 		goto reflect;
554 
555 	case ICMP_TSTAMP:
556 		if (icmplen < ICMP_TSLEN) {
557 			ICMP_STATINC(ICMP_STAT_BADLEN);
558 			break;
559 		}
560 		if (!icmpbmcastecho &&
561 		    (m->m_flags & (M_MCAST | M_BCAST)) != 0)  {
562 			ICMP_STATINC(ICMP_STAT_BMCASTTSTAMP);
563 			break;
564 		}
565 		icp->icmp_type = ICMP_TSTAMPREPLY;
566 		icp->icmp_rtime = iptime();
567 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
568 		goto reflect;
569 
570 	case ICMP_MASKREQ:
571 		if (icmpmaskrepl == 0)
572 			break;
573 		/*
574 		 * We are not able to respond with all ones broadcast
575 		 * unless we receive it over a point-to-point interface.
576 		 */
577 		if (icmplen < ICMP_MASKLEN) {
578 			ICMP_STATINC(ICMP_STAT_BADLEN);
579 			break;
580 		}
581 		if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
582 		    in_nullhost(ip->ip_dst))
583 			icmpdst.sin_addr = ip->ip_src;
584 		else
585 			icmpdst.sin_addr = ip->ip_dst;
586 		ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst),
587 		    m->m_pkthdr.rcvif));
588 		if (ia == 0)
589 			break;
590 		icp->icmp_type = ICMP_MASKREPLY;
591 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
592 		if (in_nullhost(ip->ip_src)) {
593 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
594 				ip->ip_src = ia->ia_broadaddr.sin_addr;
595 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
596 				ip->ip_src = ia->ia_dstaddr.sin_addr;
597 		}
598 reflect:
599 		{
600 			uint64_t *icps = percpu_getref(icmpstat_percpu);
601 			icps[ICMP_STAT_REFLECT]++;
602 			icps[ICMP_STAT_OUTHIST + icp->icmp_type]++;
603 			percpu_putref(icmpstat_percpu);
604 		}
605 		icmp_reflect(m);
606 		return;
607 
608 	case ICMP_REDIRECT:
609 		if (code > 3)
610 			goto badcode;
611 		if (icmp_rediraccept == 0)
612 			goto freeit;
613 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
614 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
615 			ICMP_STATINC(ICMP_STAT_BADLEN);
616 			break;
617 		}
618 		/*
619 		 * Short circuit routing redirects to force
620 		 * immediate change in the kernel's routing
621 		 * tables.  The message is also handed to anyone
622 		 * listening on a raw socket (e.g. the routing
623 		 * daemon for use in updating its tables).
624 		 */
625 		icmpgw.sin_addr = ip->ip_src;
626 		icmpdst.sin_addr = icp->icmp_gwaddr;
627 #ifdef	ICMPPRINTFS
628 		if (icmpprintfs) {
629 			printf("redirect dst `%s' to `%s'\n",
630 			    inet_ntoa(icp->icmp_ip.ip_dst),
631 			    inet_ntoa(icp->icmp_gwaddr));
632 		}
633 #endif
634 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
635 		rt = NULL;
636 		rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst),
637 		    NULL, RTF_GATEWAY | RTF_HOST, sintosa(&icmpgw), &rt);
638 		if (rt != NULL && icmp_redirtimeout != 0) {
639 			i = rt_timer_add(rt, icmp_redirect_timeout,
640 					 icmp_redirect_timeout_q);
641 			if (i)
642 				log(LOG_ERR, "ICMP:  redirect failed to "
643 				    "register timeout for route to %x, "
644 				    "code %d\n",
645 				    icp->icmp_ip.ip_dst.s_addr, i);
646 		}
647 		if (rt != NULL)
648 			rtfree(rt);
649 
650 		pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc));
651 #if defined(IPSEC) || defined(FAST_IPSEC)
652 		key_sa_routechange((struct sockaddr *)&icmpsrc);
653 #endif
654 		break;
655 
656 	/*
657 	 * No kernel processing for the following;
658 	 * just fall through to send to raw listener.
659 	 */
660 	case ICMP_ECHOREPLY:
661 	case ICMP_ROUTERADVERT:
662 	case ICMP_ROUTERSOLICIT:
663 	case ICMP_TSTAMPREPLY:
664 	case ICMP_IREQREPLY:
665 	case ICMP_MASKREPLY:
666 	default:
667 		break;
668 	}
669 
670 raw:
671 	rip_input(m, hlen, proto);
672 	return;
673 
674 freeit:
675 	m_freem(m);
676 	return;
677 }
678 
679 /*
680  * Reflect the ip packet back to the source
681  */
682 void
683 icmp_reflect(struct mbuf *m)
684 {
685 	struct ip *ip = mtod(m, struct ip *);
686 	struct in_ifaddr *ia;
687 	struct ifaddr *ifa;
688 	struct sockaddr_in *sin = 0;
689 	struct in_addr t;
690 	struct mbuf *opts = 0;
691 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
692 
693 	if (!in_canforward(ip->ip_src) &&
694 	    ((ip->ip_src.s_addr & IN_CLASSA_NET) !=
695 	     htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
696 		m_freem(m);	/* Bad return address */
697 		goto done;	/* ip_output() will check for broadcast */
698 	}
699 	t = ip->ip_dst;
700 	ip->ip_dst = ip->ip_src;
701 	/*
702 	 * If the incoming packet was addressed directly to us, use
703 	 * dst as the src for the reply.  Otherwise (broadcast or
704 	 * anonymous), use an address which corresponds to the
705 	 * incoming interface, with a preference for the address which
706 	 * corresponds to the route to the destination of the ICMP.
707 	 */
708 
709 	/* Look for packet addressed to us */
710 	INADDR_TO_IA(t, ia);
711 
712 	/* look for packet sent to broadcast address */
713 	if (ia == NULL && m->m_pkthdr.rcvif &&
714 	    (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST)) {
715 		IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) {
716 			if (ifa->ifa_addr->sa_family != AF_INET)
717 				continue;
718 			if (in_hosteq(t,ifatoia(ifa)->ia_broadaddr.sin_addr)) {
719 				ia = ifatoia(ifa);
720 				break;
721 			}
722 		}
723 	}
724 
725 	if (ia)
726 		sin = &ia->ia_addr;
727 
728 	icmpdst.sin_addr = t;
729 
730 	/*
731 	 * if the packet is addressed somewhere else, compute the
732 	 * source address for packets routed back to the source, and
733 	 * use that, if it's an address on the interface which
734 	 * received the packet
735 	 */
736 	if (sin == NULL && m->m_pkthdr.rcvif) {
737 		struct sockaddr_in sin_dst;
738 		struct route icmproute;
739 		int errornum;
740 
741 		sockaddr_in_init(&sin_dst, &ip->ip_dst, 0);
742 		memset(&icmproute, 0, sizeof(icmproute));
743 		errornum = 0;
744 		sin = in_selectsrc(&sin_dst, &icmproute, 0, NULL, &errornum);
745 		/* errornum is never used */
746 		rtcache_free(&icmproute);
747 		/* check to make sure sin is a source address on rcvif */
748 		if (sin) {
749 			t = sin->sin_addr;
750 			sin = NULL;
751 			INADDR_TO_IA(t, ia);
752 			while (ia) {
753 				if (ia->ia_ifp == m->m_pkthdr.rcvif) {
754 					sin = &ia->ia_addr;
755 					break;
756 				}
757 				NEXT_IA_WITH_SAME_ADDR(ia);
758 			}
759 		}
760 	}
761 
762 	/*
763 	 * if it was not addressed to us, but the route doesn't go out
764 	 * the source interface, pick an address on the source
765 	 * interface.  This can happen when routing is asymmetric, or
766 	 * when the incoming packet was encapsulated
767 	 */
768 	if (sin == NULL && m->m_pkthdr.rcvif) {
769 		IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) {
770 			if (ifa->ifa_addr->sa_family != AF_INET)
771 				continue;
772 			sin = &(ifatoia(ifa)->ia_addr);
773 			break;
774 		}
775 	}
776 
777 	/*
778 	 * The following happens if the packet was not addressed to us,
779 	 * and was received on an interface with no IP address:
780 	 * We find the first AF_INET address on the first non-loopback
781 	 * interface.
782 	 */
783 	if (sin == NULL)
784 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
785 			if (ia->ia_ifp->if_flags & IFF_LOOPBACK)
786 				continue;
787 			sin = &ia->ia_addr;
788 			break;
789 		}
790 
791 	/*
792 	 * If we still didn't find an address, punt.  We could have an
793 	 * interface up (and receiving packets) with no address.
794 	 */
795 	if (sin == NULL) {
796 		m_freem(m);
797 		goto done;
798 	}
799 
800 	ip->ip_src = sin->sin_addr;
801 	ip->ip_ttl = MAXTTL;
802 
803 	if (optlen > 0) {
804 		u_char *cp;
805 		int opt, cnt;
806 		u_int len;
807 
808 		/*
809 		 * Retrieve any source routing from the incoming packet;
810 		 * add on any record-route or timestamp options.
811 		 */
812 		cp = (u_char *) (ip + 1);
813 		if ((opts = ip_srcroute()) == 0 &&
814 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
815 			MCLAIM(opts, m->m_owner);
816 			opts->m_len = sizeof(struct in_addr);
817 			*mtod(opts, struct in_addr *) = zeroin_addr;
818 		}
819 		if (opts) {
820 #ifdef ICMPPRINTFS
821 		    if (icmpprintfs)
822 			    printf("icmp_reflect optlen %d rt %d => ",
823 				optlen, opts->m_len);
824 #endif
825 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
826 			    opt = cp[IPOPT_OPTVAL];
827 			    if (opt == IPOPT_EOL)
828 				    break;
829 			    if (opt == IPOPT_NOP)
830 				    len = 1;
831 			    else {
832 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
833 					    break;
834 				    len = cp[IPOPT_OLEN];
835 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
836 				        len > cnt)
837 					    break;
838 			    }
839 			    /*
840 			     * Should check for overflow, but it "can't happen"
841 			     */
842 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
843 				opt == IPOPT_SECURITY) {
844 				    memmove(mtod(opts, char *) + opts->m_len,
845 					cp, len);
846 				    opts->m_len += len;
847 			    }
848 		    }
849 		    /* Terminate & pad, if necessary */
850 		    if ((cnt = opts->m_len % 4) != 0) {
851 			    for (; cnt < 4; cnt++) {
852 				    *(mtod(opts, char *) + opts->m_len) =
853 					IPOPT_EOL;
854 				    opts->m_len++;
855 			    }
856 		    }
857 #ifdef ICMPPRINTFS
858 		    if (icmpprintfs)
859 			    printf("%d\n", opts->m_len);
860 #endif
861 		}
862 		/*
863 		 * Now strip out original options by copying rest of first
864 		 * mbuf's data back, and adjust the IP length.
865 		 */
866 		ip->ip_len = htons(ntohs(ip->ip_len) - optlen);
867 		ip->ip_hl = sizeof(struct ip) >> 2;
868 		m->m_len -= optlen;
869 		if (m->m_flags & M_PKTHDR)
870 			m->m_pkthdr.len -= optlen;
871 		optlen += sizeof(struct ip);
872 		memmove(ip + 1, (char *)ip + optlen,
873 		    (unsigned)(m->m_len - sizeof(struct ip)));
874 	}
875 	m_tag_delete_nonpersistent(m);
876 	m->m_flags &= ~(M_BCAST|M_MCAST);
877 
878 	/*
879 	 * Clear any in-bound checksum flags for this packet.
880 	 */
881 	if (m->m_flags & M_PKTHDR)
882 		m->m_pkthdr.csum_flags = 0;
883 
884 	icmp_send(m, opts);
885 done:
886 	if (opts)
887 		(void)m_free(opts);
888 }
889 
890 /*
891  * Send an icmp packet back to the ip level,
892  * after supplying a checksum.
893  */
894 void
895 icmp_send(struct mbuf *m, struct mbuf *opts)
896 {
897 	struct ip *ip = mtod(m, struct ip *);
898 	int hlen;
899 	struct icmp *icp;
900 
901 	hlen = ip->ip_hl << 2;
902 	m->m_data += hlen;
903 	m->m_len -= hlen;
904 	icp = mtod(m, struct icmp *);
905 	icp->icmp_cksum = 0;
906 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
907 	m->m_data -= hlen;
908 	m->m_len += hlen;
909 #ifdef ICMPPRINTFS
910 	if (icmpprintfs) {
911 		printf("icmp_send to destination `%s' from `%s'\n",
912 		    inet_ntoa(ip->ip_dst), inet_ntoa(ip->ip_src));
913 	}
914 #endif
915 	(void)ip_output(m, opts, NULL, 0, NULL, NULL);
916 }
917 
918 n_time
919 iptime(void)
920 {
921 	struct timeval atv;
922 	u_long t;
923 
924 	microtime(&atv);
925 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
926 	return (htonl(t));
927 }
928 
929 /*
930  * sysctl helper routine for net.inet.icmp.returndatabytes.  ensures
931  * that the new value is in the correct range.
932  */
933 static int
934 sysctl_net_inet_icmp_returndatabytes(SYSCTLFN_ARGS)
935 {
936 	int error, t;
937 	struct sysctlnode node;
938 
939 	node = *rnode;
940 	node.sysctl_data = &t;
941 	t = icmpreturndatabytes;
942 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
943 	if (error || newp == NULL)
944 		return (error);
945 
946 	if (t < 8 || t > 512)
947 		return (EINVAL);
948 	icmpreturndatabytes = t;
949 
950 	return (0);
951 }
952 
953 /*
954  * sysctl helper routine for net.inet.icmp.redirtimeout.  ensures that
955  * the given value is not less than zero and then resets the timeout
956  * queue.
957  */
958 static int
959 sysctl_net_inet_icmp_redirtimeout(SYSCTLFN_ARGS)
960 {
961 	int error, tmp;
962 	struct sysctlnode node;
963 
964 	node = *rnode;
965 	node.sysctl_data = &tmp;
966 	tmp = icmp_redirtimeout;
967 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
968 	if (error || newp == NULL)
969 		return (error);
970 	if (tmp < 0)
971 		return (EINVAL);
972 	icmp_redirtimeout = tmp;
973 
974 	/*
975 	 * was it a *defined* side-effect that anyone even *reading*
976 	 * this value causes these things to happen?
977 	 */
978 	if (icmp_redirect_timeout_q != NULL) {
979 		if (icmp_redirtimeout == 0) {
980 			rt_timer_queue_destroy(icmp_redirect_timeout_q,
981 			    true);
982 			icmp_redirect_timeout_q = NULL;
983 		} else {
984 			rt_timer_queue_change(icmp_redirect_timeout_q,
985 			    icmp_redirtimeout);
986 		}
987 	} else if (icmp_redirtimeout > 0) {
988 		icmp_redirect_timeout_q =
989 		    rt_timer_queue_create(icmp_redirtimeout);
990 	}
991 
992 	return (0);
993 }
994 
995 static int
996 sysctl_net_inet_icmp_stats(SYSCTLFN_ARGS)
997 {
998 
999 	return (NETSTAT_SYSCTL(icmpstat_percpu, ICMP_NSTATS));
1000 }
1001 
1002 static void
1003 sysctl_netinet_icmp_setup(struct sysctllog **clog)
1004 {
1005 
1006 	sysctl_createv(clog, 0, NULL, NULL,
1007 		       CTLFLAG_PERMANENT,
1008 		       CTLTYPE_NODE, "net", NULL,
1009 		       NULL, 0, NULL, 0,
1010 		       CTL_NET, CTL_EOL);
1011 	sysctl_createv(clog, 0, NULL, NULL,
1012 		       CTLFLAG_PERMANENT,
1013 		       CTLTYPE_NODE, "inet", NULL,
1014 		       NULL, 0, NULL, 0,
1015 		       CTL_NET, PF_INET, CTL_EOL);
1016 	sysctl_createv(clog, 0, NULL, NULL,
1017 		       CTLFLAG_PERMANENT,
1018 		       CTLTYPE_NODE, "icmp",
1019 		       SYSCTL_DESCR("ICMPv4 related settings"),
1020 		       NULL, 0, NULL, 0,
1021 		       CTL_NET, PF_INET, IPPROTO_ICMP, CTL_EOL);
1022 
1023 	sysctl_createv(clog, 0, NULL, NULL,
1024 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1025 		       CTLTYPE_INT, "maskrepl",
1026 		       SYSCTL_DESCR("Respond to ICMP_MASKREQ messages"),
1027 		       NULL, 0, &icmpmaskrepl, 0,
1028 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1029 		       ICMPCTL_MASKREPL, CTL_EOL);
1030 	sysctl_createv(clog, 0, NULL, NULL,
1031 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1032 		       CTLTYPE_INT, "returndatabytes",
1033 		       SYSCTL_DESCR("Number of bytes to return in an ICMP "
1034 				    "error message"),
1035 		       sysctl_net_inet_icmp_returndatabytes, 0,
1036 		       &icmpreturndatabytes, 0,
1037 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1038 		       ICMPCTL_RETURNDATABYTES, CTL_EOL);
1039 	sysctl_createv(clog, 0, NULL, NULL,
1040 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1041 		       CTLTYPE_INT, "errppslimit",
1042 		       SYSCTL_DESCR("Maximum number of outgoing ICMP error "
1043 				    "messages per second"),
1044 		       NULL, 0, &icmperrppslim, 0,
1045 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1046 		       ICMPCTL_ERRPPSLIMIT, CTL_EOL);
1047 	sysctl_createv(clog, 0, NULL, NULL,
1048 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1049 		       CTLTYPE_INT, "rediraccept",
1050 		       SYSCTL_DESCR("Accept ICMP_REDIRECT messages"),
1051 		       NULL, 0, &icmp_rediraccept, 0,
1052 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1053 		       ICMPCTL_REDIRACCEPT, CTL_EOL);
1054 	sysctl_createv(clog, 0, NULL, NULL,
1055 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1056 		       CTLTYPE_INT, "redirtimeout",
1057 		       SYSCTL_DESCR("Lifetime of ICMP_REDIRECT generated "
1058 				    "routes"),
1059 		       sysctl_net_inet_icmp_redirtimeout, 0,
1060 		       &icmp_redirtimeout, 0,
1061 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1062 		       ICMPCTL_REDIRTIMEOUT, CTL_EOL);
1063 	sysctl_createv(clog, 0, NULL, NULL,
1064 		       CTLFLAG_PERMANENT,
1065 		       CTLTYPE_STRUCT, "stats",
1066 		       SYSCTL_DESCR("ICMP statistics"),
1067 		       sysctl_net_inet_icmp_stats, 0, NULL, 0,
1068 		       CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_STATS,
1069 		       CTL_EOL);
1070 	sysctl_createv(clog, 0, NULL, NULL,
1071 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1072 		       CTLTYPE_INT, "bmcastecho",
1073 		       SYSCTL_DESCR("Respond to ICMP_ECHO or ICMP_TIMESTAMP "
1074 				    "message to the broadcast or multicast"),
1075 		       NULL, 0, &icmpbmcastecho, 0,
1076 		       CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_BMCASTECHO,
1077 		       CTL_EOL);
1078 }
1079 
1080 void
1081 icmp_statinc(u_int stat)
1082 {
1083 
1084 	KASSERT(stat < ICMP_NSTATS);
1085 	ICMP_STATINC(stat);
1086 }
1087 
1088 /* Table of common MTUs: */
1089 
1090 static const u_int mtu_table[] = {
1091 	65535, 65280, 32000, 17914, 9180, 8166,
1092 	4352, 2002, 1492, 1006, 508, 296, 68, 0
1093 };
1094 
1095 void
1096 icmp_mtudisc(struct icmp *icp, struct in_addr faddr)
1097 {
1098 	struct icmp_mtudisc_callback *mc;
1099 	struct sockaddr *dst = sintosa(&icmpsrc);
1100 	struct rtentry *rt;
1101 	u_long mtu = ntohs(icp->icmp_nextmtu);  /* Why a long?  IPv6 */
1102 	int    error;
1103 
1104 	rt = rtalloc1(dst, 1);
1105 	if (rt == 0)
1106 		return;
1107 
1108 	/* If we didn't get a host route, allocate one */
1109 
1110 	if ((rt->rt_flags & RTF_HOST) == 0) {
1111 		struct rtentry *nrt;
1112 
1113 		error = rtrequest((int) RTM_ADD, dst,
1114 		    (struct sockaddr *) rt->rt_gateway,
1115 		    (struct sockaddr *) 0,
1116 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
1117 		if (error) {
1118 			rtfree(rt);
1119 			return;
1120 		}
1121 		nrt->rt_rmx = rt->rt_rmx;
1122 		rtfree(rt);
1123 		rt = nrt;
1124 	}
1125 	error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q);
1126 	if (error) {
1127 		rtfree(rt);
1128 		return;
1129 	}
1130 
1131 	if (mtu == 0) {
1132 		int i = 0;
1133 
1134 		mtu = ntohs(icp->icmp_ip.ip_len);
1135 		/* Some 4.2BSD-based routers incorrectly adjust the ip_len */
1136 		if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0)
1137 			mtu -= (icp->icmp_ip.ip_hl << 2);
1138 
1139 		/* If we still can't guess a value, try the route */
1140 
1141 		if (mtu == 0) {
1142 			mtu = rt->rt_rmx.rmx_mtu;
1143 
1144 			/* If no route mtu, default to the interface mtu */
1145 
1146 			if (mtu == 0)
1147 				mtu = rt->rt_ifp->if_mtu;
1148 		}
1149 
1150 		for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++)
1151 			if (mtu > mtu_table[i]) {
1152 				mtu = mtu_table[i];
1153 				break;
1154 			}
1155 	}
1156 
1157 	/*
1158 	 * XXX:   RTV_MTU is overloaded, since the admin can set it
1159 	 *	  to turn off PMTU for a route, and the kernel can
1160 	 *	  set it to indicate a serious problem with PMTU
1161 	 *	  on a route.  We should be using a separate flag
1162 	 *	  for the kernel to indicate this.
1163 	 */
1164 
1165 	if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
1166 		if (mtu < 296 || mtu > rt->rt_ifp->if_mtu)
1167 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1168 		else if (rt->rt_rmx.rmx_mtu > mtu ||
1169 			 rt->rt_rmx.rmx_mtu == 0) {
1170 			ICMP_STATINC(ICMP_STAT_PMTUCHG);
1171 			rt->rt_rmx.rmx_mtu = mtu;
1172 		}
1173 	}
1174 
1175 	if (rt)
1176 		rtfree(rt);
1177 
1178 	/*
1179 	 * Notify protocols that the MTU for this destination
1180 	 * has changed.
1181 	 */
1182 	for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL;
1183 	     mc = LIST_NEXT(mc, mc_list))
1184 		(*mc->mc_func)(faddr);
1185 }
1186 
1187 /*
1188  * Return the next larger or smaller MTU plateau (table from RFC 1191)
1189  * given current value MTU.  If DIR is less than zero, a larger plateau
1190  * is returned; otherwise, a smaller value is returned.
1191  */
1192 u_int
1193 ip_next_mtu(u_int mtu, int dir)	/* XXX */
1194 {
1195 	int i;
1196 
1197 	for (i = 0; i < (sizeof mtu_table) / (sizeof mtu_table[0]); i++) {
1198 		if (mtu >= mtu_table[i])
1199 			break;
1200 	}
1201 
1202 	if (dir < 0) {
1203 		if (i == 0) {
1204 			return 0;
1205 		} else {
1206 			return mtu_table[i - 1];
1207 		}
1208 	} else {
1209 		if (mtu_table[i] == 0) {
1210 			return 0;
1211 		} else if (mtu > mtu_table[i]) {
1212 			return mtu_table[i];
1213 		} else {
1214 			return mtu_table[i + 1];
1215 		}
1216 	}
1217 }
1218 
1219 static void
1220 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
1221 {
1222 	if (rt == NULL)
1223 		panic("icmp_mtudisc_timeout:  bad route to timeout");
1224 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1225 	    (RTF_DYNAMIC | RTF_HOST)) {
1226 		rtrequest((int) RTM_DELETE, rt_getkey(rt),
1227 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1228 	} else {
1229 		if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
1230 			rt->rt_rmx.rmx_mtu = 0;
1231 		}
1232 	}
1233 }
1234 
1235 static void
1236 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r)
1237 {
1238 	if (rt == NULL)
1239 		panic("icmp_redirect_timeout:  bad route to timeout");
1240 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1241 	    (RTF_DYNAMIC | RTF_HOST)) {
1242 		rtrequest((int) RTM_DELETE, rt_getkey(rt),
1243 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1244 	}
1245 }
1246 
1247 /*
1248  * Perform rate limit check.
1249  * Returns 0 if it is okay to send the icmp packet.
1250  * Returns 1 if the router SHOULD NOT send this icmp packet due to rate
1251  * limitation.
1252  *
1253  * XXX per-destination/type check necessary?
1254  */
1255 int
1256 icmp_ratelimit(const struct in_addr *dst, const int type,
1257     const int code)
1258 {
1259 
1260 	/* PPS limit */
1261 	if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count,
1262 	    icmperrppslim)) {
1263 		/* The packet is subject to rate limit */
1264 		return 1;
1265 	}
1266 
1267 	/* okay to send */
1268 	return 0;
1269 }
1270