xref: /netbsd-src/sys/netinet/ip_icmp.c (revision 82d56013d7b633d116a93943de88e08335357a7c)
1 /*	$NetBSD: ip_icmp.c,v 1.177 2018/12/22 14:28:57 maxv Exp $	*/
2 
3 /*
4  * Copyright (c) 1998, 2000 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Public Access Networks Corporation ("Panix").  It was developed under
9  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
10  *
11  * This code is derived from software contributed to The NetBSD Foundation
12  * by Jason R. Thorpe of Zembu Labs, Inc.
13  *
14  * Redistribution and use in source and binary forms, with or without
15  * modification, are permitted provided that the following conditions
16  * are met:
17  * 1. Redistributions of source code must retain the above copyright
18  *    notice, this list of conditions and the following disclaimer.
19  * 2. Redistributions in binary form must reproduce the above copyright
20  *    notice, this list of conditions and the following disclaimer in the
21  *    documentation and/or other materials provided with the distribution.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
24  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
25  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
26  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
27  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
30  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
31  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
32  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
33  * POSSIBILITY OF SUCH DAMAGE.
34  */
35 
36 /*
37  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
38  * All rights reserved.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 3. Neither the name of the project nor the names of its contributors
49  *    may be used to endorse or promote products derived from this software
50  *    without specific prior written permission.
51  *
52  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
53  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
56  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62  * 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 1.177 2018/12/22 14:28:57 maxv Exp $");
98 
99 #ifdef _KERNEL_OPT
100 #include "opt_ipsec.h"
101 #endif
102 
103 #include <sys/param.h>
104 #include <sys/systm.h>
105 #include <sys/mbuf.h>
106 #include <sys/protosw.h>
107 #include <sys/socket.h>
108 #include <sys/socketvar.h> /* For softnet_lock */
109 #include <sys/kmem.h>
110 #include <sys/time.h>
111 #include <sys/kernel.h>
112 #include <sys/syslog.h>
113 #include <sys/sysctl.h>
114 
115 #include <net/if.h>
116 #include <net/route.h>
117 
118 #include <netinet/in.h>
119 #include <netinet/in_systm.h>
120 #include <netinet/in_var.h>
121 #include <netinet/ip.h>
122 #include <netinet/ip_icmp.h>
123 #include <netinet/ip_var.h>
124 #include <netinet/in_pcb.h>
125 #include <netinet/in_proto.h>
126 #include <netinet/icmp_var.h>
127 #include <netinet/icmp_private.h>
128 #include <netinet/wqinput.h>
129 
130 #ifdef IPSEC
131 #include <netipsec/ipsec.h>
132 #include <netipsec/key.h>
133 #endif
134 
135 /*
136  * ICMP routines: error generation, receive packet processing, and
137  * routines to turnaround packets back to the originator, and
138  * host table maintenance routines.
139  */
140 
141 int icmpmaskrepl = 0;
142 int icmpbmcastecho = 0;
143 int icmpreturndatabytes = 8;
144 
145 percpu_t *icmpstat_percpu;
146 
147 /*
148  * List of callbacks to notify when Path MTU changes are made.
149  */
150 struct icmp_mtudisc_callback {
151 	LIST_ENTRY(icmp_mtudisc_callback) mc_list;
152 	void (*mc_func)(struct in_addr);
153 };
154 
155 LIST_HEAD(, icmp_mtudisc_callback) icmp_mtudisc_callbacks =
156     LIST_HEAD_INITIALIZER(&icmp_mtudisc_callbacks);
157 
158 /* unused... */
159 u_int ip_next_mtu(u_int, int);
160 
161 static int icmperrppslim = 100;			/* 100pps */
162 static int icmperrpps_count = 0;
163 static struct timeval icmperrppslim_last;
164 static int icmp_rediraccept = 1;
165 static int icmp_redirtimeout = 600;
166 static struct rttimer_queue *icmp_redirect_timeout_q = NULL;
167 
168 /* Protect mtudisc and redirect stuff */
169 static kmutex_t icmp_mtx __cacheline_aligned;
170 
171 static void icmp_send(struct mbuf *, struct mbuf *);
172 static void icmp_mtudisc_timeout(struct rtentry *, struct rttimer *);
173 static void icmp_redirect_timeout(struct rtentry *, struct rttimer *);
174 
175 static void sysctl_netinet_icmp_setup(struct sysctllog **);
176 
177 /* workqueue-based pr_input */
178 static struct wqinput *icmp_wqinput;
179 static void _icmp_input(struct mbuf *, int, int);
180 
181 void
182 icmp_init(void)
183 {
184 
185 	sysctl_netinet_icmp_setup(NULL);
186 
187 	mutex_init(&icmp_mtx, MUTEX_DEFAULT, IPL_NONE);
188 	/*
189 	 * This is only useful if the user initializes redirtimeout to
190 	 * something other than zero.
191 	 */
192 	mutex_enter(&icmp_mtx);
193 	icmp_redirect_timeout_q = rt_timer_queue_create(icmp_redirtimeout);
194 	mutex_exit(&icmp_mtx);
195 
196 	icmpstat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP_NSTATS);
197 	icmp_wqinput = wqinput_create("icmp", _icmp_input);
198 }
199 
200 void
201 icmp_mtudisc_lock(void)
202 {
203 
204 	mutex_enter(&icmp_mtx);
205 }
206 
207 void
208 icmp_mtudisc_unlock(void)
209 {
210 
211 	mutex_exit(&icmp_mtx);
212 }
213 
214 /*
215  * Register a Path MTU Discovery callback.
216  */
217 void
218 icmp_mtudisc_callback_register(void (*func)(struct in_addr))
219 {
220 	struct icmp_mtudisc_callback *mc, *new;
221 
222 	new = kmem_alloc(sizeof(*mc), KM_SLEEP);
223 
224 	mutex_enter(&icmp_mtx);
225 	for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL;
226 	     mc = LIST_NEXT(mc, mc_list)) {
227 		if (mc->mc_func == func) {
228 			mutex_exit(&icmp_mtx);
229 			kmem_free(new, sizeof(*mc));
230 			return;
231 		}
232 	}
233 
234 	new->mc_func = func;
235 	LIST_INSERT_HEAD(&icmp_mtudisc_callbacks, new, mc_list);
236 	mutex_exit(&icmp_mtx);
237 }
238 
239 /*
240  * Generate an error packet of type error in response to a bad IP packet. 'n'
241  * contains this packet. We create 'm' and send it.
242  *
243  * As we are not required to return everything we have, we return whatever
244  * we can return at ease.
245  *
246  * Note that ICMP datagrams longer than 576 octets are out of spec according
247  * to RFC1812; the limit on icmpreturndatabytes will keep things below that
248  * limit.
249  */
250 void
251 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu)
252 {
253 	struct ip *oip = mtod(n, struct ip *), *nip;
254 	const unsigned oiphlen = oip->ip_hl << 2;
255 	struct icmp *icp;
256 	struct mbuf *m;
257 	struct m_tag *mtag;
258 	unsigned datalen, mblen;
259 	int totlen;
260 
261 	if (type != ICMP_REDIRECT)
262 		ICMP_STATINC(ICMP_STAT_ERROR);
263 
264 	/*
265 	 * Don't send error if:
266 	 *  - The original packet was encrypted.
267 	 *  - The packet is multicast or broadcast.
268 	 *  - The packet is not the first fragment of the message.
269 	 *  - The packet is an ICMP message with an unknown type.
270 	 */
271 	if (n->m_flags & M_DECRYPTED)
272 		goto freeit;
273 	if (n->m_flags & (M_BCAST|M_MCAST))
274 		goto freeit;
275 	if (oip->ip_off &~ htons(IP_MF|IP_DF))
276 		goto freeit;
277 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
278 	    n->m_len >= oiphlen + ICMP_MINLEN) {
279 		struct icmp *oicp = (struct icmp *)((char *)oip + oiphlen);
280 		if (!ICMP_INFOTYPE(oicp->icmp_type)) {
281 			ICMP_STATINC(ICMP_STAT_OLDICMP);
282 			goto freeit;
283 		}
284 	}
285 
286 	/*
287 	 * First, do a rate limitation check.
288 	 */
289 	if (icmp_ratelimit(&oip->ip_src, type, code)) {
290 		/* XXX stat */
291 		goto freeit;
292 	}
293 
294 	/*
295 	 * Compute the number of bytes we will put in 'icmp_ip'. Truncate
296 	 * it to the size of the mbuf, if it's too big.
297 	 */
298 	datalen = oiphlen + uimin(icmpreturndatabytes,
299 	    ntohs(oip->ip_len) - oiphlen);
300 	mblen = 0;
301 	for (m = n; m && (mblen < datalen); m = m->m_next)
302 		mblen += m->m_len;
303 	datalen = uimin(mblen, datalen);
304 
305 	/*
306 	 * Compute the total length of the new packet. Truncate it if it's
307 	 * bigger than the size of a cluster.
308 	 */
309 	CTASSERT(ICMP_MINLEN + sizeof(struct ip) <= MCLBYTES);
310 	totlen = sizeof(struct ip) + ICMP_MINLEN + datalen;
311 	if (totlen > MCLBYTES) {
312 		datalen = MCLBYTES - ICMP_MINLEN - sizeof(struct ip);
313 		totlen = MCLBYTES;
314 	}
315 
316 	/*
317 	 * Allocate the mbuf for the new packet.
318 	 */
319 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
320 	if (m && (totlen > MHLEN)) {
321 		MCLGET(m, M_DONTWAIT);
322 		if ((m->m_flags & M_EXT) == 0) {
323 			m_freem(m);
324 			m = NULL;
325 		}
326 	}
327 	if (m == NULL)
328 		goto freeit;
329 	MCLAIM(m, n->m_owner);
330 	m->m_len = totlen;
331 	m->m_pkthdr.len = m->m_len;
332 	m_copy_rcvif(m, n);
333 
334 	if ((u_int)type > ICMP_MAXTYPE)
335 		panic("icmp_error");
336 	ICMP_STATINC(ICMP_STAT_OUTHIST + type);
337 
338 	if ((m->m_flags & M_EXT) == 0)
339 		m_align(m, m->m_len);
340 
341 	/*
342 	 * Get pointers on the IP header and the ICMP header.
343 	 */
344 	nip = mtod(m, struct ip *);
345 	icp = (struct icmp *)(nip + 1);
346 
347 	/*
348 	 * Fill in the fields of the ICMP header: icmp_type, icmp_code
349 	 * and icmp_ip. icmp_cksum gets filled later.
350 	 */
351 	icp->icmp_type = type;
352 	if (type == ICMP_REDIRECT) {
353 		icp->icmp_gwaddr.s_addr = dest;
354 	} else {
355 		icp->icmp_void = 0;
356 		/*
357 		 * The following assignments assume an overlay with the
358 		 * zeroed icmp_void field.
359 		 */
360 		if (type == ICMP_PARAMPROB) {
361 			icp->icmp_pptr = code;
362 			code = 0;
363 		} else if (type == ICMP_UNREACH &&
364 		    code == ICMP_UNREACH_NEEDFRAG && destmtu)
365 			icp->icmp_nextmtu = htons(destmtu);
366 	}
367 	icp->icmp_code = code;
368 	m_copydata(n, 0, datalen, (void *)&icp->icmp_ip);
369 
370 	/*
371 	 * Now, copy the old IP header (without options) in front of the
372 	 * ICMP message. The src/dst fields will be swapped in icmp_reflect.
373 	 */
374 	/* ip_v set in ip_output */
375 	nip->ip_hl = sizeof(struct ip) >> 2;
376 	nip->ip_tos = 0;
377 	nip->ip_len = htons(m->m_len);
378 	/* ip_id set in ip_output */
379 	nip->ip_off = htons(0);
380 	/* ip_ttl set in icmp_reflect */
381 	nip->ip_p = IPPROTO_ICMP;
382 	nip->ip_src = oip->ip_src;
383 	nip->ip_dst = oip->ip_dst;
384 	/* move PF m_tag to new packet, if it exists */
385 	mtag = m_tag_find(n, PACKET_TAG_PF);
386 	if (mtag != NULL) {
387 		m_tag_unlink(n, mtag);
388 		m_tag_prepend(m, mtag);
389 	}
390 
391 	icmp_reflect(m);
392 
393 freeit:
394 	m_freem(n);
395 }
396 
397 struct sockaddr_in icmpsrc = {
398 	.sin_len = sizeof(struct sockaddr_in),
399 	.sin_family = AF_INET,
400 };
401 
402 /*
403  * Process a received ICMP message.
404  */
405 static void
406 _icmp_input(struct mbuf *m, int hlen, int proto)
407 {
408 	struct icmp *icp;
409 	struct ip *ip = mtod(m, struct ip *);
410 	int icmplen;
411 	int i;
412 	struct in_ifaddr *ia;
413 	void *(*ctlfunc)(int, const struct sockaddr *, void *);
414 	int code;
415 	struct rtentry *rt;
416 	struct sockaddr_in icmpdst = {
417 		.sin_len = sizeof(struct sockaddr_in),
418 		.sin_family = AF_INET,
419 	};
420 	struct sockaddr_in icmpgw = {
421 		.sin_len = sizeof(struct sockaddr_in),
422 		.sin_family = AF_INET,
423 	};
424 
425 	/*
426 	 * Locate icmp structure in mbuf, and check
427 	 * that not corrupted and of at least minimum length.
428 	 */
429 	icmplen = ntohs(ip->ip_len) - hlen;
430 	if (icmplen < ICMP_MINLEN) {
431 		ICMP_STATINC(ICMP_STAT_TOOSHORT);
432 		goto freeit;
433 	}
434 	i = hlen + uimin(icmplen, ICMP_ADVLENMIN);
435 	if (M_UNWRITABLE(m, i) && (m = m_pullup(m, i)) == NULL) {
436 		ICMP_STATINC(ICMP_STAT_TOOSHORT);
437 		return;
438 	}
439 	ip = mtod(m, struct ip *);
440 	m->m_len -= hlen;
441 	m->m_data += hlen;
442 	icp = mtod(m, struct icmp *);
443 	/* Don't need to assert alignment, here. */
444 	if (in_cksum(m, icmplen)) {
445 		ICMP_STATINC(ICMP_STAT_CHECKSUM);
446 		goto freeit;
447 	}
448 	m->m_len += hlen;
449 	m->m_data -= hlen;
450 
451 	if (icp->icmp_type > ICMP_MAXTYPE)
452 		goto raw;
453 	ICMP_STATINC(ICMP_STAT_INHIST + icp->icmp_type);
454 	code = icp->icmp_code;
455 
456 	switch (icp->icmp_type) {
457 	case ICMP_UNREACH:
458 		switch (code) {
459 		case ICMP_UNREACH_PROTOCOL:
460 			code = PRC_UNREACH_PROTOCOL;
461 			break;
462 
463 		case ICMP_UNREACH_PORT:
464 			code = PRC_UNREACH_PORT;
465 			break;
466 
467 		case ICMP_UNREACH_SRCFAIL:
468 			code = PRC_UNREACH_SRCFAIL;
469 			break;
470 
471 		case ICMP_UNREACH_NEEDFRAG:
472 			code = PRC_MSGSIZE;
473 			break;
474 
475 		case ICMP_UNREACH_NET:
476 		case ICMP_UNREACH_NET_UNKNOWN:
477 		case ICMP_UNREACH_NET_PROHIB:
478 		case ICMP_UNREACH_TOSNET:
479 			code = PRC_UNREACH_NET;
480 			break;
481 
482 		case ICMP_UNREACH_HOST:
483 		case ICMP_UNREACH_HOST_UNKNOWN:
484 		case ICMP_UNREACH_ISOLATED:
485 		case ICMP_UNREACH_HOST_PROHIB:
486 		case ICMP_UNREACH_TOSHOST:
487 		case ICMP_UNREACH_ADMIN_PROHIBIT:
488 		case ICMP_UNREACH_HOST_PREC:
489 		case ICMP_UNREACH_PREC_CUTOFF:
490 			code = PRC_UNREACH_HOST;
491 			break;
492 
493 		default:
494 			goto badcode;
495 		}
496 		goto deliver;
497 
498 	case ICMP_TIMXCEED:
499 		if (code > 1)
500 			goto badcode;
501 		code += PRC_TIMXCEED_INTRANS;
502 		goto deliver;
503 
504 	case ICMP_PARAMPROB:
505 		if (code > 1)
506 			goto badcode;
507 		code = PRC_PARAMPROB;
508 		goto deliver;
509 
510 	case ICMP_SOURCEQUENCH:
511 		if (code)
512 			goto badcode;
513 		code = PRC_QUENCH;
514 		goto deliver;
515 
516 	deliver:
517 		/*
518 		 * Problem with datagram; advise higher level routines.
519 		 */
520 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
521 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
522 			ICMP_STATINC(ICMP_STAT_BADLEN);
523 			goto freeit;
524 		}
525 		if (m->m_len < hlen + ICMP_ADVLEN(icp)) {
526 			m = m_pullup(m, hlen + ICMP_ADVLEN(icp));
527 			if (m == NULL)
528 				goto freeit;
529 		}
530 		ip = mtod(m, struct ip *);
531 		icp = (struct icmp *)(mtod(m, uint8_t *) + hlen);
532 
533 		if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr))
534 			goto badcode;
535 
536 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
537 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
538 		if (ctlfunc)
539 			(void) (*ctlfunc)(code, sintosa(&icmpsrc),
540 			    &icp->icmp_ip);
541 		break;
542 
543 	badcode:
544 		ICMP_STATINC(ICMP_STAT_BADCODE);
545 		break;
546 
547 	case ICMP_ECHO:
548 		if (!icmpbmcastecho &&
549 		    (m->m_flags & (M_MCAST | M_BCAST)) != 0)  {
550 			ICMP_STATINC(ICMP_STAT_BMCASTECHO);
551 			break;
552 		}
553 		icp->icmp_type = ICMP_ECHOREPLY;
554 		goto reflect;
555 
556 	case ICMP_TSTAMP:
557 		if (icmplen < ICMP_TSLEN) {
558 			ICMP_STATINC(ICMP_STAT_BADLEN);
559 			break;
560 		}
561 		if (!icmpbmcastecho &&
562 		    (m->m_flags & (M_MCAST | M_BCAST)) != 0)  {
563 			ICMP_STATINC(ICMP_STAT_BMCASTTSTAMP);
564 			break;
565 		}
566 		icp->icmp_type = ICMP_TSTAMPREPLY;
567 		icp->icmp_rtime = iptime();
568 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
569 		goto reflect;
570 
571 	case ICMP_MASKREQ: {
572 		struct ifnet *rcvif;
573 		int s, ss;
574 		struct ifaddr *ifa = NULL;
575 
576 		if (icmpmaskrepl == 0)
577 			break;
578 		/*
579 		 * We are not able to respond with all ones broadcast
580 		 * unless we receive it over a point-to-point interface.
581 		 */
582 		if (icmplen < ICMP_MASKLEN) {
583 			ICMP_STATINC(ICMP_STAT_BADLEN);
584 			break;
585 		}
586 		if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
587 		    in_nullhost(ip->ip_dst))
588 			icmpdst.sin_addr = ip->ip_src;
589 		else
590 			icmpdst.sin_addr = ip->ip_dst;
591 		ss = pserialize_read_enter();
592 		rcvif = m_get_rcvif(m, &s);
593 		if (__predict_true(rcvif != NULL))
594 			ifa = ifaof_ifpforaddr(sintosa(&icmpdst), rcvif);
595 		m_put_rcvif(rcvif, &s);
596 		if (ifa == NULL) {
597 			pserialize_read_exit(ss);
598 			break;
599 		}
600 		ia = ifatoia(ifa);
601 		icp->icmp_type = ICMP_MASKREPLY;
602 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
603 		if (in_nullhost(ip->ip_src)) {
604 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
605 				ip->ip_src = ia->ia_broadaddr.sin_addr;
606 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
607 				ip->ip_src = ia->ia_dstaddr.sin_addr;
608 		}
609 		pserialize_read_exit(ss);
610 reflect:
611 		{
612 			uint64_t *icps = percpu_getref(icmpstat_percpu);
613 			icps[ICMP_STAT_REFLECT]++;
614 			icps[ICMP_STAT_OUTHIST + icp->icmp_type]++;
615 			percpu_putref(icmpstat_percpu);
616 		}
617 		icmp_reflect(m);
618 		return;
619 	}
620 
621 	case ICMP_REDIRECT:
622 		if (code > 3)
623 			goto badcode;
624 		if (icmp_rediraccept == 0)
625 			goto freeit;
626 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
627 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
628 			ICMP_STATINC(ICMP_STAT_BADLEN);
629 			break;
630 		}
631 		/*
632 		 * Short circuit routing redirects to force
633 		 * immediate change in the kernel's routing
634 		 * tables.  The message is also handed to anyone
635 		 * listening on a raw socket (e.g. the routing
636 		 * daemon for use in updating its tables).
637 		 */
638 		icmpgw.sin_addr = ip->ip_src;
639 		icmpdst.sin_addr = icp->icmp_gwaddr;
640 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
641 		rt = NULL;
642 		rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst),
643 		    NULL, RTF_GATEWAY | RTF_HOST, sintosa(&icmpgw), &rt);
644 		mutex_enter(&icmp_mtx);
645 		if (rt != NULL && icmp_redirtimeout != 0) {
646 			i = rt_timer_add(rt, icmp_redirect_timeout,
647 					 icmp_redirect_timeout_q);
648 			if (i) {
649 				char buf[INET_ADDRSTRLEN];
650 				log(LOG_ERR, "ICMP:  redirect failed to "
651 				    "register timeout for route to %s, "
652 				    "code %d\n",
653 				    IN_PRINT(buf, &icp->icmp_ip.ip_dst), i);
654 			}
655 		}
656 		mutex_exit(&icmp_mtx);
657 		if (rt != NULL)
658 			rt_unref(rt);
659 
660 		pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc));
661 #if defined(IPSEC)
662 		if (ipsec_used)
663 			key_sa_routechange((struct sockaddr *)&icmpsrc);
664 #endif
665 		break;
666 
667 	/*
668 	 * No kernel processing for the following;
669 	 * just fall through to send to raw listener.
670 	 */
671 	case ICMP_ECHOREPLY:
672 	case ICMP_ROUTERADVERT:
673 	case ICMP_ROUTERSOLICIT:
674 	case ICMP_TSTAMPREPLY:
675 	case ICMP_IREQREPLY:
676 	case ICMP_MASKREPLY:
677 	default:
678 		break;
679 	}
680 
681 raw:
682 	/*
683 	 * Currently, pim_input() is always called holding softnet_lock
684 	 * by ipintr()(!NET_MPSAFE) or PR_INPUT_WRAP()(NET_MPSAFE).
685 	 */
686 	KASSERT(mutex_owned(softnet_lock));
687 	rip_input(m, hlen, proto);
688 	return;
689 
690 freeit:
691 	m_freem(m);
692 	return;
693 }
694 
695 void
696 icmp_input(struct mbuf *m, int off, int proto)
697 {
698 	wqinput_input(icmp_wqinput, m, off, proto);
699 }
700 
701 /*
702  * Reflect the ip packet back to the source
703  */
704 void
705 icmp_reflect(struct mbuf *m)
706 {
707 	struct ip *ip = mtod(m, struct ip *);
708 	struct in_ifaddr *ia;
709 	struct ifaddr *ifa;
710 	struct sockaddr_in *sin;
711 	struct in_addr t;
712 	struct mbuf *opts = NULL;
713 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
714 	struct ifnet *rcvif;
715 	struct psref psref, psref_ia;
716 	int s;
717 	int bound;
718 
719 	bound = curlwp_bind();
720 
721 	if (!in_canforward(ip->ip_src) &&
722 	    ((ip->ip_src.s_addr & IN_CLASSA_NET) !=
723 	     htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
724 		m_freem(m);	/* Bad return address */
725 		goto done;	/* ip_output() will check for broadcast */
726 	}
727 	t = ip->ip_dst;
728 	ip->ip_dst = ip->ip_src;
729 
730 	/*
731 	 * If the incoming packet was addressed directly to us, use
732 	 * dst as the src for the reply.  Otherwise (broadcast or
733 	 * anonymous), use an address which corresponds to the
734 	 * incoming interface, with a preference for the address which
735 	 * corresponds to the route to the destination of the ICMP.
736 	 */
737 
738 	/* Look for packet addressed to us */
739 	ia = in_get_ia_psref(t, &psref_ia);
740 	if (ia && (ia->ia4_flags & IN_IFF_NOTREADY)) {
741 		ia4_release(ia, &psref_ia);
742 		ia = NULL;
743 	}
744 
745 	rcvif = m_get_rcvif_psref(m, &psref);
746 
747 	/* look for packet sent to broadcast address */
748 	if (ia == NULL && rcvif &&
749 	    (rcvif->if_flags & IFF_BROADCAST)) {
750 		s = pserialize_read_enter();
751 		IFADDR_READER_FOREACH(ifa, rcvif) {
752 			if (ifa->ifa_addr->sa_family != AF_INET)
753 				continue;
754 			if (in_hosteq(t,ifatoia(ifa)->ia_broadaddr.sin_addr)) {
755 				ia = ifatoia(ifa);
756 				if ((ia->ia4_flags & IN_IFF_NOTREADY) == 0)
757 					break;
758 				ia = NULL;
759 			}
760 		}
761 		if (ia != NULL)
762 			ia4_acquire(ia, &psref_ia);
763 		pserialize_read_exit(s);
764 	}
765 
766 	sin = ia ? &ia->ia_addr : NULL;
767 
768 	/*
769 	 * if the packet is addressed somewhere else, compute the
770 	 * source address for packets routed back to the source, and
771 	 * use that, if it's an address on the interface which
772 	 * received the packet
773 	 */
774 	if (sin == NULL && rcvif) {
775 		struct sockaddr_in sin_dst;
776 		struct route icmproute;
777 		int errornum;
778 
779 		sockaddr_in_init(&sin_dst, &ip->ip_dst, 0);
780 		memset(&icmproute, 0, sizeof(icmproute));
781 		errornum = 0;
782 		ia = in_selectsrc(&sin_dst, &icmproute, 0, NULL, &errornum,
783 		    &psref_ia);
784 		/* errornum is never used */
785 		rtcache_free(&icmproute);
786 		/* check to make sure sin is a source address on rcvif */
787 		if (ia != NULL) {
788 			sin = &ia->ia_addr;
789 			t = sin->sin_addr;
790 			sin = NULL;
791 			ia4_release(ia, &psref_ia);
792 			ia = in_get_ia_on_iface_psref(t, rcvif, &psref_ia);
793 			if (ia != NULL)
794 				sin = &ia->ia_addr;
795 		}
796 	}
797 
798 	/*
799 	 * if it was not addressed to us, but the route doesn't go out
800 	 * the source interface, pick an address on the source
801 	 * interface.  This can happen when routing is asymmetric, or
802 	 * when the incoming packet was encapsulated
803 	 */
804 	if (sin == NULL && rcvif) {
805 		KASSERT(ia == NULL);
806 		s = pserialize_read_enter();
807 		IFADDR_READER_FOREACH(ifa, rcvif) {
808 			if (ifa->ifa_addr->sa_family != AF_INET)
809 				continue;
810 			sin = &(ifatoia(ifa)->ia_addr);
811 			ia = ifatoia(ifa);
812 			ia4_acquire(ia, &psref_ia);
813 			break;
814 		}
815 		pserialize_read_exit(s);
816 	}
817 
818 	m_put_rcvif_psref(rcvif, &psref);
819 
820 	/*
821 	 * The following happens if the packet was not addressed to us,
822 	 * and was received on an interface with no IP address:
823 	 * We find the first AF_INET address on the first non-loopback
824 	 * interface.
825 	 */
826 	if (sin == NULL) {
827 		KASSERT(ia == NULL);
828 		s = pserialize_read_enter();
829 		IN_ADDRLIST_READER_FOREACH(ia) {
830 			if (ia->ia_ifp->if_flags & IFF_LOOPBACK)
831 				continue;
832 			sin = &ia->ia_addr;
833 			ia4_acquire(ia, &psref_ia);
834 			break;
835 		}
836 		pserialize_read_exit(s);
837 	}
838 
839 	/*
840 	 * If we still didn't find an address, punt.  We could have an
841 	 * interface up (and receiving packets) with no address.
842 	 */
843 	if (sin == NULL) {
844 		KASSERT(ia == NULL);
845 		m_freem(m);
846 		goto done;
847 	}
848 
849 	ip->ip_src = sin->sin_addr;
850 	ip->ip_ttl = MAXTTL;
851 
852 	if (ia != NULL)
853 		ia4_release(ia, &psref_ia);
854 
855 	if (optlen > 0) {
856 		u_char *cp;
857 		int opt, cnt;
858 		u_int len;
859 
860 		/*
861 		 * Retrieve any source routing from the incoming packet;
862 		 * add on any record-route or timestamp options.
863 		 */
864 		cp = (u_char *)(ip + 1);
865 		if ((opts = ip_srcroute(m)) == NULL &&
866 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
867 			MCLAIM(opts, m->m_owner);
868 			opts->m_len = sizeof(struct in_addr);
869 			*mtod(opts, struct in_addr *) = zeroin_addr;
870 		}
871 
872 		if (opts) {
873 			for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
874 				opt = cp[IPOPT_OPTVAL];
875 				if (opt == IPOPT_EOL)
876 					break;
877 				if (opt == IPOPT_NOP)
878 					len = 1;
879 				else {
880 					if (cnt < IPOPT_OLEN + sizeof(*cp))
881 						break;
882 					len = cp[IPOPT_OLEN];
883 					if (len < IPOPT_OLEN + sizeof(*cp) ||
884 					    len > cnt)
885 						break;
886 				}
887 
888 				/* Overflows can't happen */
889 				KASSERT(opts->m_len + len <= MHLEN);
890 
891 				if (opt == IPOPT_RR || opt == IPOPT_TS ||
892 				    opt == IPOPT_SECURITY) {
893 					memmove(mtod(opts, char *) +
894 					    opts->m_len, cp, len);
895 					opts->m_len += len;
896 				}
897 			}
898 
899 			/* Terminate & pad, if necessary */
900 			if ((cnt = opts->m_len % 4) != 0) {
901 				for (; cnt < 4; cnt++) {
902 					*(mtod(opts, char *) + opts->m_len) =
903 					    IPOPT_EOL;
904 					opts->m_len++;
905 				}
906 			}
907 		}
908 
909 		/*
910 		 * Now strip out original options by copying rest of first
911 		 * mbuf's data back, and adjust the IP length.
912 		 */
913 		ip->ip_len = htons(ntohs(ip->ip_len) - optlen);
914 		ip->ip_hl = sizeof(struct ip) >> 2;
915 		m->m_len -= optlen;
916 		if (m->m_flags & M_PKTHDR)
917 			m->m_pkthdr.len -= optlen;
918 		optlen += sizeof(struct ip);
919 		memmove(ip + 1, (char *)ip + optlen,
920 		    (unsigned)(m->m_len - sizeof(struct ip)));
921 	}
922 	m_tag_delete_chain(m);
923 	m->m_flags &= ~(M_BCAST|M_MCAST);
924 
925 	/*
926 	 * Clear any in-bound checksum flags for this packet.
927 	 */
928 	if (m->m_flags & M_PKTHDR)
929 		m->m_pkthdr.csum_flags = 0;
930 
931 	icmp_send(m, opts);
932 done:
933 	curlwp_bindx(bound);
934 	if (opts)
935 		(void)m_free(opts);
936 }
937 
938 /*
939  * Send an icmp packet back to the ip level,
940  * after supplying a checksum.
941  */
942 static void
943 icmp_send(struct mbuf *m, struct mbuf *opts)
944 {
945 	struct ip *ip = mtod(m, struct ip *);
946 	int hlen;
947 	struct icmp *icp;
948 
949 	hlen = ip->ip_hl << 2;
950 	m->m_data += hlen;
951 	m->m_len -= hlen;
952 	icp = mtod(m, struct icmp *);
953 	icp->icmp_cksum = 0;
954 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
955 	m->m_data -= hlen;
956 	m->m_len += hlen;
957 
958 	(void)ip_output(m, opts, NULL, 0, NULL, NULL);
959 }
960 
961 n_time
962 iptime(void)
963 {
964 	struct timeval atv;
965 	u_long t;
966 
967 	microtime(&atv);
968 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
969 	return (htonl(t));
970 }
971 
972 /*
973  * sysctl helper routine for net.inet.icmp.returndatabytes.  ensures
974  * that the new value is in the correct range.
975  */
976 static int
977 sysctl_net_inet_icmp_returndatabytes(SYSCTLFN_ARGS)
978 {
979 	int error, t;
980 	struct sysctlnode node;
981 
982 	node = *rnode;
983 	node.sysctl_data = &t;
984 	t = icmpreturndatabytes;
985 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
986 	if (error || newp == NULL)
987 		return error;
988 
989 	if (t < 8 || t > 512)
990 		return EINVAL;
991 	icmpreturndatabytes = t;
992 
993 	return 0;
994 }
995 
996 /*
997  * sysctl helper routine for net.inet.icmp.redirtimeout.  ensures that
998  * the given value is not less than zero and then resets the timeout
999  * queue.
1000  */
1001 static int
1002 sysctl_net_inet_icmp_redirtimeout(SYSCTLFN_ARGS)
1003 {
1004 	int error, tmp;
1005 	struct sysctlnode node;
1006 
1007 	mutex_enter(&icmp_mtx);
1008 
1009 	node = *rnode;
1010 	node.sysctl_data = &tmp;
1011 	tmp = icmp_redirtimeout;
1012 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1013 	if (error || newp == NULL)
1014 		goto out;
1015 	if (tmp < 0) {
1016 		error = EINVAL;
1017 		goto out;
1018 	}
1019 	icmp_redirtimeout = tmp;
1020 
1021 	/*
1022 	 * was it a *defined* side-effect that anyone even *reading*
1023 	 * this value causes these things to happen?
1024 	 */
1025 	if (icmp_redirect_timeout_q != NULL) {
1026 		if (icmp_redirtimeout == 0) {
1027 			rt_timer_queue_destroy(icmp_redirect_timeout_q);
1028 			icmp_redirect_timeout_q = NULL;
1029 		} else {
1030 			rt_timer_queue_change(icmp_redirect_timeout_q,
1031 			    icmp_redirtimeout);
1032 		}
1033 	} else if (icmp_redirtimeout > 0) {
1034 		icmp_redirect_timeout_q =
1035 		    rt_timer_queue_create(icmp_redirtimeout);
1036 	}
1037 	error = 0;
1038 out:
1039 	mutex_exit(&icmp_mtx);
1040 	return error;
1041 }
1042 
1043 static int
1044 sysctl_net_inet_icmp_stats(SYSCTLFN_ARGS)
1045 {
1046 
1047 	return (NETSTAT_SYSCTL(icmpstat_percpu, ICMP_NSTATS));
1048 }
1049 
1050 static void
1051 sysctl_netinet_icmp_setup(struct sysctllog **clog)
1052 {
1053 
1054 	sysctl_createv(clog, 0, NULL, NULL,
1055 		       CTLFLAG_PERMANENT,
1056 		       CTLTYPE_NODE, "inet", NULL,
1057 		       NULL, 0, NULL, 0,
1058 		       CTL_NET, PF_INET, CTL_EOL);
1059 	sysctl_createv(clog, 0, NULL, NULL,
1060 		       CTLFLAG_PERMANENT,
1061 		       CTLTYPE_NODE, "icmp",
1062 		       SYSCTL_DESCR("ICMPv4 related settings"),
1063 		       NULL, 0, NULL, 0,
1064 		       CTL_NET, PF_INET, IPPROTO_ICMP, CTL_EOL);
1065 
1066 	sysctl_createv(clog, 0, NULL, NULL,
1067 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1068 		       CTLTYPE_INT, "maskrepl",
1069 		       SYSCTL_DESCR("Respond to ICMP_MASKREQ messages"),
1070 		       NULL, 0, &icmpmaskrepl, 0,
1071 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1072 		       ICMPCTL_MASKREPL, CTL_EOL);
1073 	sysctl_createv(clog, 0, NULL, NULL,
1074 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1075 		       CTLTYPE_INT, "returndatabytes",
1076 		       SYSCTL_DESCR("Number of bytes to return in an ICMP "
1077 				    "error message"),
1078 		       sysctl_net_inet_icmp_returndatabytes, 0,
1079 		       &icmpreturndatabytes, 0,
1080 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1081 		       ICMPCTL_RETURNDATABYTES, CTL_EOL);
1082 	sysctl_createv(clog, 0, NULL, NULL,
1083 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1084 		       CTLTYPE_INT, "errppslimit",
1085 		       SYSCTL_DESCR("Maximum number of outgoing ICMP error "
1086 				    "messages per second"),
1087 		       NULL, 0, &icmperrppslim, 0,
1088 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1089 		       ICMPCTL_ERRPPSLIMIT, CTL_EOL);
1090 	sysctl_createv(clog, 0, NULL, NULL,
1091 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1092 		       CTLTYPE_INT, "rediraccept",
1093 		       SYSCTL_DESCR("Accept ICMP_REDIRECT messages"),
1094 		       NULL, 0, &icmp_rediraccept, 0,
1095 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1096 		       ICMPCTL_REDIRACCEPT, CTL_EOL);
1097 	sysctl_createv(clog, 0, NULL, NULL,
1098 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1099 		       CTLTYPE_INT, "redirtimeout",
1100 		       SYSCTL_DESCR("Lifetime of ICMP_REDIRECT generated "
1101 				    "routes"),
1102 		       sysctl_net_inet_icmp_redirtimeout, 0,
1103 		       &icmp_redirtimeout, 0,
1104 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1105 		       ICMPCTL_REDIRTIMEOUT, CTL_EOL);
1106 	sysctl_createv(clog, 0, NULL, NULL,
1107 		       CTLFLAG_PERMANENT,
1108 		       CTLTYPE_STRUCT, "stats",
1109 		       SYSCTL_DESCR("ICMP statistics"),
1110 		       sysctl_net_inet_icmp_stats, 0, NULL, 0,
1111 		       CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_STATS,
1112 		       CTL_EOL);
1113 	sysctl_createv(clog, 0, NULL, NULL,
1114 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1115 		       CTLTYPE_INT, "bmcastecho",
1116 		       SYSCTL_DESCR("Respond to ICMP_ECHO or ICMP_TIMESTAMP "
1117 				    "message to the broadcast or multicast"),
1118 		       NULL, 0, &icmpbmcastecho, 0,
1119 		       CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_BMCASTECHO,
1120 		       CTL_EOL);
1121 }
1122 
1123 void
1124 icmp_statinc(u_int stat)
1125 {
1126 
1127 	KASSERT(stat < ICMP_NSTATS);
1128 	ICMP_STATINC(stat);
1129 }
1130 
1131 /* Table of common MTUs */
1132 static const u_int mtu_table[] = {
1133 	65535, 65280, 32000, 17914, 9180, 8166,
1134 	4352, 2002, 1492, 1006, 508, 296, 68, 0
1135 };
1136 
1137 void
1138 icmp_mtudisc(struct icmp *icp, struct in_addr faddr)
1139 {
1140 	struct icmp_mtudisc_callback *mc;
1141 	struct sockaddr *dst = sintosa(&icmpsrc);
1142 	struct rtentry *rt;
1143 	u_long mtu = ntohs(icp->icmp_nextmtu);  /* Why a long?  IPv6 */
1144 	int error;
1145 
1146 	rt = rtalloc1(dst, 1);
1147 	if (rt == NULL)
1148 		return;
1149 
1150 	/* If we didn't get a host route, allocate one */
1151 	if ((rt->rt_flags & RTF_HOST) == 0) {
1152 		struct rtentry *nrt;
1153 
1154 		error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL,
1155 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
1156 		if (error) {
1157 			rt_unref(rt);
1158 			return;
1159 		}
1160 		nrt->rt_rmx = rt->rt_rmx;
1161 		rt_unref(rt);
1162 		rt = nrt;
1163 	}
1164 
1165 	mutex_enter(&icmp_mtx);
1166 	error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q);
1167 	mutex_exit(&icmp_mtx);
1168 	if (error) {
1169 		rt_unref(rt);
1170 		return;
1171 	}
1172 
1173 	if (mtu == 0) {
1174 		int i = 0;
1175 
1176 		mtu = ntohs(icp->icmp_ip.ip_len);
1177 		/* Some 4.2BSD-based routers incorrectly adjust the ip_len */
1178 		if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0)
1179 			mtu -= (icp->icmp_ip.ip_hl << 2);
1180 
1181 		/* If we still can't guess a value, try the route */
1182 		if (mtu == 0) {
1183 			mtu = rt->rt_rmx.rmx_mtu;
1184 
1185 			/* If no route mtu, default to the interface mtu */
1186 			if (mtu == 0)
1187 				mtu = rt->rt_ifp->if_mtu;
1188 		}
1189 
1190 		for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++) {
1191 			if (mtu > mtu_table[i]) {
1192 				mtu = mtu_table[i];
1193 				break;
1194 			}
1195 		}
1196 	}
1197 
1198 	/*
1199 	 * XXX:   RTV_MTU is overloaded, since the admin can set it
1200 	 *	  to turn off PMTU for a route, and the kernel can
1201 	 *	  set it to indicate a serious problem with PMTU
1202 	 *	  on a route.  We should be using a separate flag
1203 	 *	  for the kernel to indicate this.
1204 	 */
1205 
1206 	if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
1207 		if (mtu < 296 || mtu > rt->rt_ifp->if_mtu)
1208 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1209 		else if (rt->rt_rmx.rmx_mtu > mtu ||
1210 			 rt->rt_rmx.rmx_mtu == 0) {
1211 			ICMP_STATINC(ICMP_STAT_PMTUCHG);
1212 			rt->rt_rmx.rmx_mtu = mtu;
1213 		}
1214 	}
1215 
1216 	if (rt != NULL)
1217 		rt_unref(rt);
1218 
1219 	/*
1220 	 * Notify protocols that the MTU for this destination
1221 	 * has changed.
1222 	 */
1223 	mutex_enter(&icmp_mtx);
1224 	for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL;
1225 	     mc = LIST_NEXT(mc, mc_list))
1226 		(*mc->mc_func)(faddr);
1227 	mutex_exit(&icmp_mtx);
1228 }
1229 
1230 /*
1231  * Return the next larger or smaller MTU plateau (table from RFC 1191)
1232  * given current value MTU.  If DIR is less than zero, a larger plateau
1233  * is returned; otherwise, a smaller value is returned.
1234  */
1235 u_int
1236 ip_next_mtu(u_int mtu, int dir)	/* XXX unused */
1237 {
1238 	int i;
1239 
1240 	for (i = 0; i < (sizeof mtu_table) / (sizeof mtu_table[0]); i++) {
1241 		if (mtu >= mtu_table[i])
1242 			break;
1243 	}
1244 
1245 	if (dir < 0) {
1246 		if (i == 0) {
1247 			return 0;
1248 		} else {
1249 			return mtu_table[i - 1];
1250 		}
1251 	} else {
1252 		if (mtu_table[i] == 0) {
1253 			return 0;
1254 		} else if (mtu > mtu_table[i]) {
1255 			return mtu_table[i];
1256 		} else {
1257 			return mtu_table[i + 1];
1258 		}
1259 	}
1260 }
1261 
1262 static void
1263 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
1264 {
1265 	struct rtentry *retrt;
1266 
1267 	KASSERT(rt != NULL);
1268 	rt_assert_referenced(rt);
1269 
1270 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1271 	    (RTF_DYNAMIC | RTF_HOST)) {
1272 		rtrequest(RTM_DELETE, rt_getkey(rt),
1273 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, &retrt);
1274 		rt_unref(rt);
1275 		rt_free(retrt);
1276 	} else {
1277 		if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
1278 			rt->rt_rmx.rmx_mtu = 0;
1279 		}
1280 	}
1281 }
1282 
1283 static void
1284 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r)
1285 {
1286 	struct rtentry *retrt;
1287 
1288 	KASSERT(rt != NULL);
1289 	rt_assert_referenced(rt);
1290 
1291 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1292 	    (RTF_DYNAMIC | RTF_HOST)) {
1293 		rtrequest(RTM_DELETE, rt_getkey(rt),
1294 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, &retrt);
1295 		rt_unref(rt);
1296 		rt_free(retrt);
1297 	}
1298 }
1299 
1300 /*
1301  * Perform rate limit check.
1302  * Returns 0 if it is okay to send the icmp packet.
1303  * Returns 1 if the router SHOULD NOT send this icmp packet due to rate
1304  * limitation.
1305  *
1306  * XXX per-destination/type check necessary?
1307  */
1308 int
1309 icmp_ratelimit(const struct in_addr *dst, const int type,
1310     const int code)
1311 {
1312 
1313 	/* PPS limit */
1314 	if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count,
1315 	    icmperrppslim)) {
1316 		/* The packet is subject to rate limit */
1317 		return 1;
1318 	}
1319 
1320 	/* okay to send */
1321 	return 0;
1322 }
1323