xref: /netbsd-src/sys/netinet/ip_icmp.c (revision 796c32c94f6e154afc9de0f63da35c91bb739b45)
1 /*	$NetBSD: ip_icmp.c,v 1.161 2017/03/31 06:49:44 ozaki-r 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 1.161 2017/03/31 06:49:44 ozaki-r 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	/* IPSEC*/
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 #ifdef ICMPPRINTFS
144 int	icmpprintfs = 0;
145 #endif
146 int	icmpreturndatabytes = 8;
147 
148 percpu_t *icmpstat_percpu;
149 
150 /*
151  * List of callbacks to notify when Path MTU changes are made.
152  */
153 struct icmp_mtudisc_callback {
154 	LIST_ENTRY(icmp_mtudisc_callback) mc_list;
155 	void (*mc_func)(struct in_addr);
156 };
157 
158 LIST_HEAD(, icmp_mtudisc_callback) icmp_mtudisc_callbacks =
159     LIST_HEAD_INITIALIZER(&icmp_mtudisc_callbacks);
160 
161 #if 0
162 static u_int	ip_next_mtu(u_int, int);
163 #else
164 /*static*/ u_int	ip_next_mtu(u_int, int);
165 #endif
166 
167 extern int icmperrppslim;
168 static int icmperrpps_count = 0;
169 static struct timeval icmperrppslim_last;
170 static int icmp_rediraccept = 1;
171 static int icmp_redirtimeout = 600;
172 static struct rttimer_queue *icmp_redirect_timeout_q = NULL;
173 
174 /* Protect mtudisc and redirect stuffs */
175 static kmutex_t icmp_mtx __cacheline_aligned;
176 
177 static void icmp_mtudisc_timeout(struct rtentry *, struct rttimer *);
178 static void icmp_redirect_timeout(struct rtentry *, struct rttimer *);
179 
180 static void sysctl_netinet_icmp_setup(struct sysctllog **);
181 
182 /* workqueue-based pr_input */
183 static struct wqinput *icmp_wqinput;
184 static void _icmp_input(struct mbuf *, int, int);
185 
186 void
187 icmp_init(void)
188 {
189 
190 	sysctl_netinet_icmp_setup(NULL);
191 
192 	mutex_init(&icmp_mtx, MUTEX_DEFAULT, IPL_NONE);
193 	/*
194 	 * This is only useful if the user initializes redirtimeout to
195 	 * something other than zero.
196 	 */
197 	mutex_enter(&icmp_mtx);
198 	icmp_redirect_timeout_q = rt_timer_queue_create(icmp_redirtimeout);
199 	mutex_exit(&icmp_mtx);
200 
201 	icmpstat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP_NSTATS);
202 	icmp_wqinput = wqinput_create("icmp", _icmp_input);
203 }
204 
205 void
206 icmp_mtudisc_lock(void)
207 {
208 
209 	mutex_enter(&icmp_mtx);
210 }
211 
212 void
213 icmp_mtudisc_unlock(void)
214 {
215 
216 	mutex_exit(&icmp_mtx);
217 }
218 
219 /*
220  * Register a Path MTU Discovery callback.
221  */
222 void
223 icmp_mtudisc_callback_register(void (*func)(struct in_addr))
224 {
225 	struct icmp_mtudisc_callback *mc, *new;
226 
227 	new = kmem_alloc(sizeof(*mc), KM_SLEEP);
228 
229 	mutex_enter(&icmp_mtx);
230 	for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL;
231 	     mc = LIST_NEXT(mc, mc_list)) {
232 		if (mc->mc_func == func) {
233 			mutex_exit(&icmp_mtx);
234 			kmem_free(new, sizeof(*mc));
235 			return;
236 		}
237 	}
238 
239 	new->mc_func = func;
240 	LIST_INSERT_HEAD(&icmp_mtudisc_callbacks, new, mc_list);
241 	mutex_exit(&icmp_mtx);
242 }
243 
244 /*
245  * Generate an error packet of type error
246  * in response to bad packet ip.
247  */
248 void
249 icmp_error(struct mbuf *n, int type, int code, n_long dest,
250     int destmtu)
251 {
252 	struct ip *oip = mtod(n, struct ip *), *nip;
253 	unsigned oiplen = oip->ip_hl << 2;
254 	struct icmp *icp;
255 	struct mbuf *m;
256 	struct m_tag *mtag;
257 	unsigned icmplen, mblen;
258 
259 #ifdef ICMPPRINTFS
260 	if (icmpprintfs)
261 		printf("icmp_error(%p, type:%d, code:%d)\n", oip, type, code);
262 #endif
263 	if (type != ICMP_REDIRECT)
264 		ICMP_STATINC(ICMP_STAT_ERROR);
265 	/*
266 	 * Don't send error if the original packet was encrypted.
267 	 * Don't send error if not the first fragment of message.
268 	 * Don't error if the old packet protocol was ICMP
269 	 * error message, only known informational types.
270 	 */
271 	if (n->m_flags & M_DECRYPTED)
272 		goto freeit;
273 	if (oip->ip_off &~ htons(IP_MF|IP_DF))
274 		goto freeit;
275 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
276 	  n->m_len >= oiplen + ICMP_MINLEN &&
277 	  !ICMP_INFOTYPE(((struct icmp *)((char *)oip + oiplen))->icmp_type)) {
278 		ICMP_STATINC(ICMP_STAT_OLDICMP);
279 		goto freeit;
280 	}
281 	/* Don't send error in response to a multicast or broadcast packet */
282 	if (n->m_flags & (M_BCAST|M_MCAST))
283 		goto freeit;
284 
285 	/*
286 	 * First, do a rate limitation check.
287 	 */
288 	if (icmp_ratelimit(&oip->ip_src, type, code)) {
289 		/* XXX stat */
290 		goto freeit;
291 	}
292 
293 	/*
294 	 * Now, formulate icmp message
295 	 */
296 	icmplen = oiplen + min(icmpreturndatabytes,
297 	    ntohs(oip->ip_len) - oiplen);
298 	/*
299 	 * Defend against mbuf chains shorter than oip->ip_len - oiplen:
300 	 */
301 	mblen = 0;
302 	for (m = n; m && (mblen < icmplen); m = m->m_next)
303 		mblen += m->m_len;
304 	icmplen = min(mblen, icmplen);
305 
306 	/*
307 	 * As we are not required to return everything we have,
308 	 * we return whatever we can return at ease.
309 	 *
310 	 * Note that ICMP datagrams longer than 576 octets are out of spec
311 	 * according to RFC1812; the limit on icmpreturndatabytes below in
312 	 * icmp_sysctl will keep things below that limit.
313 	 */
314 
315 	KASSERT(ICMP_MINLEN <= MCLBYTES);
316 
317 	if (icmplen + ICMP_MINLEN > MCLBYTES)
318 		icmplen = MCLBYTES - ICMP_MINLEN;
319 
320 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
321 	if (m && (icmplen + ICMP_MINLEN > MHLEN)) {
322 		MCLGET(m, M_DONTWAIT);
323 		if ((m->m_flags & M_EXT) == 0) {
324 			m_freem(m);
325 			m = NULL;
326 		}
327 	}
328 	if (m == NULL)
329 		goto freeit;
330 	MCLAIM(m, n->m_owner);
331 	m->m_len = icmplen + ICMP_MINLEN;
332 	if ((m->m_flags & M_EXT) == 0)
333 		MH_ALIGN(m, m->m_len);
334 	else {
335 		m->m_data += sizeof(struct ip);
336 		m->m_len -= sizeof(struct ip);
337 	}
338 	icp = mtod(m, struct icmp *);
339 	if ((u_int)type > ICMP_MAXTYPE)
340 		panic("icmp_error");
341 	ICMP_STATINC(ICMP_STAT_OUTHIST + type);
342 	icp->icmp_type = type;
343 	if (type == ICMP_REDIRECT)
344 		icp->icmp_gwaddr.s_addr = dest;
345 	else {
346 		icp->icmp_void = 0;
347 		/*
348 		 * The following assignments assume an overlay with the
349 		 * zeroed icmp_void field.
350 		 */
351 		if (type == ICMP_PARAMPROB) {
352 			icp->icmp_pptr = code;
353 			code = 0;
354 		} else if (type == ICMP_UNREACH &&
355 		    code == ICMP_UNREACH_NEEDFRAG && destmtu)
356 			icp->icmp_nextmtu = htons(destmtu);
357 	}
358 
359 	icp->icmp_code = code;
360 	m_copydata(n, 0, icmplen, (void *)&icp->icmp_ip);
361 
362 	/*
363 	 * Now, copy old ip header (without options)
364 	 * in front of icmp message.
365 	 */
366 	if ((m->m_flags & M_EXT) == 0 &&
367 	    m->m_data - sizeof(struct ip) < m->m_pktdat)
368 		panic("icmp len");
369 	m->m_data -= sizeof(struct ip);
370 	m->m_len += sizeof(struct ip);
371 	m->m_pkthdr.len = m->m_len;
372 	m_copy_rcvif(m, n);
373 	nip = mtod(m, struct ip *);
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, NULL);
386 	if (mtag != NULL) {
387 		m_tag_unlink(n, mtag);
388 		m_tag_prepend(m, mtag);
389 	}
390 	icmp_reflect(m);
391 
392 freeit:
393 	m_freem(n);
394 }
395 
396 struct sockaddr_in icmpsrc = {
397 	.sin_len = sizeof (struct sockaddr_in),
398 	.sin_family = AF_INET,
399 };
400 static struct sockaddr_in icmpdst = {
401 	.sin_len = sizeof (struct sockaddr_in),
402 	.sin_family = AF_INET,
403 };
404 static struct sockaddr_in icmpgw = {
405 	.sin_len = sizeof (struct sockaddr_in),
406 	.sin_family = AF_INET,
407 };
408 struct sockaddr_in icmpmask = {
409 	.sin_len = 8,
410 	.sin_family = 0,
411 };
412 
413 /*
414  * Process a received ICMP message.
415  */
416 static void
417 _icmp_input(struct mbuf *m, int hlen, int proto)
418 {
419 	struct icmp *icp;
420 	struct ip *ip = mtod(m, struct ip *);
421 	int icmplen;
422 	int i;
423 	struct in_ifaddr *ia;
424 	void *(*ctlfunc)(int, const struct sockaddr *, void *);
425 	int code;
426 	struct rtentry *rt;
427 
428 	/*
429 	 * Locate icmp structure in mbuf, and check
430 	 * that not corrupted and of at least minimum length.
431 	 */
432 	icmplen = ntohs(ip->ip_len) - hlen;
433 #ifdef ICMPPRINTFS
434 	if (icmpprintfs) {
435 		char sbuf[INET_ADDRSTRLEN], dbuf[INET_ADDRSTRLEN];
436 		printf("icmp_input from `%s' to `%s', len %d\n",
437 		    IN_PRINT(sbuf, &ip->ip_src), IN_PRINT(dbuf, &ip->ip_dst),
438 		    icmplen);
439 	}
440 #endif
441 	if (icmplen < ICMP_MINLEN) {
442 		ICMP_STATINC(ICMP_STAT_TOOSHORT);
443 		goto freeit;
444 	}
445 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
446 	if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == NULL) {
447 		ICMP_STATINC(ICMP_STAT_TOOSHORT);
448 		return;
449 	}
450 	ip = mtod(m, struct ip *);
451 	m->m_len -= hlen;
452 	m->m_data += hlen;
453 	icp = mtod(m, struct icmp *);
454 	/* Don't need to assert alignment, here. */
455 	if (in_cksum(m, icmplen)) {
456 		ICMP_STATINC(ICMP_STAT_CHECKSUM);
457 		goto freeit;
458 	}
459 	m->m_len += hlen;
460 	m->m_data -= hlen;
461 
462 #ifdef ICMPPRINTFS
463 	/*
464 	 * Message type specific processing.
465 	 */
466 	if (icmpprintfs)
467 		printf("icmp_input(type:%d, code:%d)\n", icp->icmp_type,
468 		    icp->icmp_code);
469 #endif
470 	if (icp->icmp_type > ICMP_MAXTYPE)
471 		goto raw;
472 	ICMP_STATINC(ICMP_STAT_INHIST + icp->icmp_type);
473 	code = icp->icmp_code;
474 	switch (icp->icmp_type) {
475 
476 	case ICMP_UNREACH:
477 		switch (code) {
478 		case ICMP_UNREACH_PROTOCOL:
479 			code = PRC_UNREACH_PROTOCOL;
480 			break;
481 
482 		case ICMP_UNREACH_PORT:
483 			code = PRC_UNREACH_PORT;
484 			break;
485 
486 		case ICMP_UNREACH_SRCFAIL:
487 			code = PRC_UNREACH_SRCFAIL;
488 			break;
489 
490 		case ICMP_UNREACH_NEEDFRAG:
491 			code = PRC_MSGSIZE;
492 			break;
493 
494 		case ICMP_UNREACH_NET:
495 		case ICMP_UNREACH_NET_UNKNOWN:
496 		case ICMP_UNREACH_NET_PROHIB:
497 		case ICMP_UNREACH_TOSNET:
498 			code = PRC_UNREACH_NET;
499 			break;
500 
501 		case ICMP_UNREACH_HOST:
502 		case ICMP_UNREACH_HOST_UNKNOWN:
503 		case ICMP_UNREACH_ISOLATED:
504 		case ICMP_UNREACH_HOST_PROHIB:
505 		case ICMP_UNREACH_TOSHOST:
506 		case ICMP_UNREACH_ADMIN_PROHIBIT:
507 		case ICMP_UNREACH_HOST_PREC:
508 		case ICMP_UNREACH_PREC_CUTOFF:
509 			code = PRC_UNREACH_HOST;
510 			break;
511 
512 		default:
513 			goto badcode;
514 		}
515 		goto deliver;
516 
517 	case ICMP_TIMXCEED:
518 		if (code > 1)
519 			goto badcode;
520 		code += PRC_TIMXCEED_INTRANS;
521 		goto deliver;
522 
523 	case ICMP_PARAMPROB:
524 		if (code > 1)
525 			goto badcode;
526 		code = PRC_PARAMPROB;
527 		goto deliver;
528 
529 	case ICMP_SOURCEQUENCH:
530 		if (code)
531 			goto badcode;
532 		code = PRC_QUENCH;
533 		goto deliver;
534 
535 	deliver:
536 		/*
537 		 * Problem with datagram; advise higher level routines.
538 		 */
539 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
540 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
541 			ICMP_STATINC(ICMP_STAT_BADLEN);
542 			goto freeit;
543 		}
544 		if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr))
545 			goto badcode;
546 #ifdef ICMPPRINTFS
547 		if (icmpprintfs)
548 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
549 #endif
550 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
551 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
552 		if (ctlfunc)
553 			(void) (*ctlfunc)(code, sintosa(&icmpsrc),
554 			    &icp->icmp_ip);
555 		break;
556 
557 	badcode:
558 		ICMP_STATINC(ICMP_STAT_BADCODE);
559 		break;
560 
561 	case ICMP_ECHO:
562 		if (!icmpbmcastecho &&
563 		    (m->m_flags & (M_MCAST | M_BCAST)) != 0)  {
564 			ICMP_STATINC(ICMP_STAT_BMCASTECHO);
565 			break;
566 		}
567 		icp->icmp_type = ICMP_ECHOREPLY;
568 		goto reflect;
569 
570 	case ICMP_TSTAMP:
571 		if (icmplen < ICMP_TSLEN) {
572 			ICMP_STATINC(ICMP_STAT_BADLEN);
573 			break;
574 		}
575 		if (!icmpbmcastecho &&
576 		    (m->m_flags & (M_MCAST | M_BCAST)) != 0)  {
577 			ICMP_STATINC(ICMP_STAT_BMCASTTSTAMP);
578 			break;
579 		}
580 		icp->icmp_type = ICMP_TSTAMPREPLY;
581 		icp->icmp_rtime = iptime();
582 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
583 		goto reflect;
584 
585 	case ICMP_MASKREQ: {
586 		struct ifnet *rcvif;
587 		int s, ss;
588 		struct ifaddr *ifa = NULL;
589 
590 		if (icmpmaskrepl == 0)
591 			break;
592 		/*
593 		 * We are not able to respond with all ones broadcast
594 		 * unless we receive it over a point-to-point interface.
595 		 */
596 		if (icmplen < ICMP_MASKLEN) {
597 			ICMP_STATINC(ICMP_STAT_BADLEN);
598 			break;
599 		}
600 		if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
601 		    in_nullhost(ip->ip_dst))
602 			icmpdst.sin_addr = ip->ip_src;
603 		else
604 			icmpdst.sin_addr = ip->ip_dst;
605 		ss = pserialize_read_enter();
606 		rcvif = m_get_rcvif(m, &s);
607 		if (__predict_true(rcvif != NULL))
608 			ifa = ifaof_ifpforaddr(sintosa(&icmpdst), rcvif);
609 		m_put_rcvif(rcvif, &s);
610 		if (ifa == NULL) {
611 			pserialize_read_exit(ss);
612 			break;
613 		}
614 		ia = ifatoia(ifa);
615 		icp->icmp_type = ICMP_MASKREPLY;
616 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
617 		if (in_nullhost(ip->ip_src)) {
618 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
619 				ip->ip_src = ia->ia_broadaddr.sin_addr;
620 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
621 				ip->ip_src = ia->ia_dstaddr.sin_addr;
622 		}
623 		pserialize_read_exit(ss);
624 reflect:
625 		{
626 			uint64_t *icps = percpu_getref(icmpstat_percpu);
627 			icps[ICMP_STAT_REFLECT]++;
628 			icps[ICMP_STAT_OUTHIST + icp->icmp_type]++;
629 			percpu_putref(icmpstat_percpu);
630 		}
631 		icmp_reflect(m);
632 		return;
633 	}
634 
635 	case ICMP_REDIRECT:
636 		if (code > 3)
637 			goto badcode;
638 		if (icmp_rediraccept == 0)
639 			goto freeit;
640 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
641 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
642 			ICMP_STATINC(ICMP_STAT_BADLEN);
643 			break;
644 		}
645 		/*
646 		 * Short circuit routing redirects to force
647 		 * immediate change in the kernel's routing
648 		 * tables.  The message is also handed to anyone
649 		 * listening on a raw socket (e.g. the routing
650 		 * daemon for use in updating its tables).
651 		 */
652 		icmpgw.sin_addr = ip->ip_src;
653 		icmpdst.sin_addr = icp->icmp_gwaddr;
654 #ifdef	ICMPPRINTFS
655 		if (icmpprintfs) {
656 			char gbuf[INET_ADDRSTRLEN], dbuf[INET_ADDRSTRLEN];
657 			printf("redirect dst `%s' to `%s'\n",
658 			    IN_PRINT(dbuf, &icp->icmp_ip.ip_dst),
659 			    IN_PRINT(gbuf, &icp->icmp_gwaddr));
660 		}
661 #endif
662 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
663 		rt = NULL;
664 		rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst),
665 		    NULL, RTF_GATEWAY | RTF_HOST, sintosa(&icmpgw), &rt);
666 		mutex_enter(&icmp_mtx);
667 		if (rt != NULL && icmp_redirtimeout != 0) {
668 			i = rt_timer_add(rt, icmp_redirect_timeout,
669 					 icmp_redirect_timeout_q);
670 			if (i) {
671 				char buf[INET_ADDRSTRLEN];
672 				log(LOG_ERR, "ICMP:  redirect failed to "
673 				    "register timeout for route to %s, "
674 				    "code %d\n",
675 				    IN_PRINT(buf, &icp->icmp_ip.ip_dst), i);
676 			}
677 		}
678 		mutex_exit(&icmp_mtx);
679 		if (rt != NULL)
680 			rt_unref(rt);
681 
682 		pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc));
683 #if defined(IPSEC)
684 		if (ipsec_used)
685 			key_sa_routechange((struct sockaddr *)&icmpsrc);
686 #endif
687 		break;
688 
689 	/*
690 	 * No kernel processing for the following;
691 	 * just fall through to send to raw listener.
692 	 */
693 	case ICMP_ECHOREPLY:
694 	case ICMP_ROUTERADVERT:
695 	case ICMP_ROUTERSOLICIT:
696 	case ICMP_TSTAMPREPLY:
697 	case ICMP_IREQREPLY:
698 	case ICMP_MASKREPLY:
699 	default:
700 		break;
701 	}
702 
703 raw:
704 	rip_input(m, hlen, proto);
705 	return;
706 
707 freeit:
708 	m_freem(m);
709 	return;
710 }
711 
712 void
713 icmp_input(struct mbuf *m, ...)
714 {
715 	int hlen, proto;
716 	va_list ap;
717 
718 	va_start(ap, m);
719 	hlen = va_arg(ap, int);
720 	proto = va_arg(ap, int);
721 	va_end(ap);
722 
723 	wqinput_input(icmp_wqinput, m, hlen, proto);
724 }
725 
726 /*
727  * Reflect the ip packet back to the source
728  */
729 void
730 icmp_reflect(struct mbuf *m)
731 {
732 	struct ip *ip = mtod(m, struct ip *);
733 	struct in_ifaddr *ia;
734 	struct ifaddr *ifa;
735 	struct sockaddr_in *sin;
736 	struct in_addr t;
737 	struct mbuf *opts = NULL;
738 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
739 	struct ifnet *rcvif;
740 	struct psref psref, psref_ia;
741 	int s;
742 	int bound;
743 
744 	bound = curlwp_bind();
745 
746 	if (!in_canforward(ip->ip_src) &&
747 	    ((ip->ip_src.s_addr & IN_CLASSA_NET) !=
748 	     htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
749 		m_freem(m);	/* Bad return address */
750 		goto done;	/* ip_output() will check for broadcast */
751 	}
752 	t = ip->ip_dst;
753 	ip->ip_dst = ip->ip_src;
754 	/*
755 	 * If the incoming packet was addressed directly to us, use
756 	 * dst as the src for the reply.  Otherwise (broadcast or
757 	 * anonymous), use an address which corresponds to the
758 	 * incoming interface, with a preference for the address which
759 	 * corresponds to the route to the destination of the ICMP.
760 	 */
761 
762 	/* Look for packet addressed to us */
763 	ia = in_get_ia_psref(t, &psref_ia);
764 	if (ia && (ia->ia4_flags & IN_IFF_NOTREADY)) {
765 		ia4_release(ia, &psref_ia);
766 		ia = NULL;
767 	}
768 
769 	rcvif = m_get_rcvif_psref(m, &psref);
770 
771 	/* look for packet sent to broadcast address */
772 	if (ia == NULL && rcvif &&
773 	    (rcvif->if_flags & IFF_BROADCAST)) {
774 		s = pserialize_read_enter();
775 		IFADDR_READER_FOREACH(ifa, rcvif) {
776 			if (ifa->ifa_addr->sa_family != AF_INET)
777 				continue;
778 			if (in_hosteq(t,ifatoia(ifa)->ia_broadaddr.sin_addr)) {
779 				ia = ifatoia(ifa);
780 				if ((ia->ia4_flags & IN_IFF_NOTREADY) == 0)
781 					break;
782 				ia = NULL;
783 			}
784 		}
785 		if (ia != NULL)
786 			ia4_acquire(ia, &psref_ia);
787 		pserialize_read_exit(s);
788 	}
789 
790 	sin = ia ? &ia->ia_addr : NULL;
791 
792 	icmpdst.sin_addr = t;
793 
794 	/*
795 	 * if the packet is addressed somewhere else, compute the
796 	 * source address for packets routed back to the source, and
797 	 * use that, if it's an address on the interface which
798 	 * received the packet
799 	 */
800 	if (sin == NULL && rcvif) {
801 		struct sockaddr_in sin_dst;
802 		struct route icmproute;
803 		int errornum;
804 
805 		sockaddr_in_init(&sin_dst, &ip->ip_dst, 0);
806 		memset(&icmproute, 0, sizeof(icmproute));
807 		errornum = 0;
808 		ia = in_selectsrc(&sin_dst, &icmproute, 0, NULL, &errornum,
809 		    &psref_ia);
810 		/* errornum is never used */
811 		rtcache_free(&icmproute);
812 		/* check to make sure sin is a source address on rcvif */
813 		if (ia != NULL) {
814 			sin = &ia->ia_addr;
815 			t = sin->sin_addr;
816 			sin = NULL;
817 			ia4_release(ia, &psref_ia);
818 			ia = in_get_ia_on_iface_psref(t, rcvif, &psref_ia);
819 			if (ia != NULL)
820 				sin = &ia->ia_addr;
821 		}
822 	}
823 
824 	/*
825 	 * if it was not addressed to us, but the route doesn't go out
826 	 * the source interface, pick an address on the source
827 	 * interface.  This can happen when routing is asymmetric, or
828 	 * when the incoming packet was encapsulated
829 	 */
830 	if (sin == NULL && rcvif) {
831 		KASSERT(ia == NULL);
832 		s = pserialize_read_enter();
833 		IFADDR_READER_FOREACH(ifa, rcvif) {
834 			if (ifa->ifa_addr->sa_family != AF_INET)
835 				continue;
836 			sin = &(ifatoia(ifa)->ia_addr);
837 			ia = ifatoia(ifa);
838 			ia4_acquire(ia, &psref_ia);
839 			break;
840 		}
841 		pserialize_read_exit(s);
842 	}
843 
844 	m_put_rcvif_psref(rcvif, &psref);
845 
846 	/*
847 	 * The following happens if the packet was not addressed to us,
848 	 * and was received on an interface with no IP address:
849 	 * We find the first AF_INET address on the first non-loopback
850 	 * interface.
851 	 */
852 	if (sin == NULL) {
853 		KASSERT(ia == NULL);
854 		s = pserialize_read_enter();
855 		IN_ADDRLIST_READER_FOREACH(ia) {
856 			if (ia->ia_ifp->if_flags & IFF_LOOPBACK)
857 				continue;
858 			sin = &ia->ia_addr;
859 			ia4_acquire(ia, &psref_ia);
860 			break;
861 		}
862 		pserialize_read_exit(s);
863 	}
864 
865 	/*
866 	 * If we still didn't find an address, punt.  We could have an
867 	 * interface up (and receiving packets) with no address.
868 	 */
869 	if (sin == NULL) {
870 		KASSERT(ia == NULL);
871 		m_freem(m);
872 		goto done;
873 	}
874 
875 	ip->ip_src = sin->sin_addr;
876 	ip->ip_ttl = MAXTTL;
877 
878 	if (ia != NULL)
879 		ia4_release(ia, &psref_ia);
880 
881 	if (optlen > 0) {
882 		u_char *cp;
883 		int opt, cnt;
884 		u_int len;
885 
886 		/*
887 		 * Retrieve any source routing from the incoming packet;
888 		 * add on any record-route or timestamp options.
889 		 */
890 		cp = (u_char *) (ip + 1);
891 		if ((opts = ip_srcroute(m)) == NULL &&
892 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
893 			MCLAIM(opts, m->m_owner);
894 			opts->m_len = sizeof(struct in_addr);
895 			*mtod(opts, struct in_addr *) = zeroin_addr;
896 		}
897 		if (opts) {
898 #ifdef ICMPPRINTFS
899 		    if (icmpprintfs)
900 			    printf("icmp_reflect optlen %d rt %d => ",
901 				optlen, opts->m_len);
902 #endif
903 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
904 			    opt = cp[IPOPT_OPTVAL];
905 			    if (opt == IPOPT_EOL)
906 				    break;
907 			    if (opt == IPOPT_NOP)
908 				    len = 1;
909 			    else {
910 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
911 					    break;
912 				    len = cp[IPOPT_OLEN];
913 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
914 				        len > cnt)
915 					    break;
916 			    }
917 			    /*
918 			     * Should check for overflow, but it "can't happen"
919 			     */
920 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
921 				opt == IPOPT_SECURITY) {
922 				    memmove(mtod(opts, char *) + opts->m_len,
923 					cp, len);
924 				    opts->m_len += len;
925 			    }
926 		    }
927 		    /* Terminate & pad, if necessary */
928 		    if ((cnt = opts->m_len % 4) != 0) {
929 			    for (; cnt < 4; cnt++) {
930 				    *(mtod(opts, char *) + opts->m_len) =
931 					IPOPT_EOL;
932 				    opts->m_len++;
933 			    }
934 		    }
935 #ifdef ICMPPRINTFS
936 		    if (icmpprintfs)
937 			    printf("%d\n", opts->m_len);
938 #endif
939 		}
940 		/*
941 		 * Now strip out original options by copying rest of first
942 		 * mbuf's data back, and adjust the IP length.
943 		 */
944 		ip->ip_len = htons(ntohs(ip->ip_len) - optlen);
945 		ip->ip_hl = sizeof(struct ip) >> 2;
946 		m->m_len -= optlen;
947 		if (m->m_flags & M_PKTHDR)
948 			m->m_pkthdr.len -= optlen;
949 		optlen += sizeof(struct ip);
950 		memmove(ip + 1, (char *)ip + optlen,
951 		    (unsigned)(m->m_len - sizeof(struct ip)));
952 	}
953 	m_tag_delete_nonpersistent(m);
954 	m->m_flags &= ~(M_BCAST|M_MCAST);
955 
956 	/*
957 	 * Clear any in-bound checksum flags for this packet.
958 	 */
959 	if (m->m_flags & M_PKTHDR)
960 		m->m_pkthdr.csum_flags = 0;
961 
962 	icmp_send(m, opts);
963 done:
964 	curlwp_bindx(bound);
965 	if (opts)
966 		(void)m_free(opts);
967 }
968 
969 /*
970  * Send an icmp packet back to the ip level,
971  * after supplying a checksum.
972  */
973 void
974 icmp_send(struct mbuf *m, struct mbuf *opts)
975 {
976 	struct ip *ip = mtod(m, struct ip *);
977 	int hlen;
978 	struct icmp *icp;
979 
980 	hlen = ip->ip_hl << 2;
981 	m->m_data += hlen;
982 	m->m_len -= hlen;
983 	icp = mtod(m, struct icmp *);
984 	icp->icmp_cksum = 0;
985 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
986 	m->m_data -= hlen;
987 	m->m_len += hlen;
988 #ifdef ICMPPRINTFS
989 	if (icmpprintfs) {
990 		char sbuf[INET_ADDRSTRLEN], dbuf[INET_ADDRSTRLEN];
991 		printf("icmp_send to destination `%s' from `%s'\n",
992 		    IN_PRINT(dbuf, &ip->ip_dst), IN_PRINT(sbuf, &ip->ip_src));
993 	}
994 #endif
995 	(void)ip_output(m, opts, NULL, 0, NULL, NULL);
996 }
997 
998 n_time
999 iptime(void)
1000 {
1001 	struct timeval atv;
1002 	u_long t;
1003 
1004 	microtime(&atv);
1005 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
1006 	return (htonl(t));
1007 }
1008 
1009 /*
1010  * sysctl helper routine for net.inet.icmp.returndatabytes.  ensures
1011  * that the new value is in the correct range.
1012  */
1013 static int
1014 sysctl_net_inet_icmp_returndatabytes(SYSCTLFN_ARGS)
1015 {
1016 	int error, t;
1017 	struct sysctlnode node;
1018 
1019 	node = *rnode;
1020 	node.sysctl_data = &t;
1021 	t = icmpreturndatabytes;
1022 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1023 	if (error || newp == NULL)
1024 		return (error);
1025 
1026 	if (t < 8 || t > 512)
1027 		return (EINVAL);
1028 	icmpreturndatabytes = t;
1029 
1030 	return (0);
1031 }
1032 
1033 /*
1034  * sysctl helper routine for net.inet.icmp.redirtimeout.  ensures that
1035  * the given value is not less than zero and then resets the timeout
1036  * queue.
1037  */
1038 static int
1039 sysctl_net_inet_icmp_redirtimeout(SYSCTLFN_ARGS)
1040 {
1041 	int error, tmp;
1042 	struct sysctlnode node;
1043 
1044 	mutex_enter(&icmp_mtx);
1045 
1046 	node = *rnode;
1047 	node.sysctl_data = &tmp;
1048 	tmp = icmp_redirtimeout;
1049 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1050 	if (error || newp == NULL)
1051 		goto out;
1052 	if (tmp < 0) {
1053 		error = EINVAL;
1054 		goto out;
1055 	}
1056 	icmp_redirtimeout = tmp;
1057 
1058 	/*
1059 	 * was it a *defined* side-effect that anyone even *reading*
1060 	 * this value causes these things to happen?
1061 	 */
1062 	if (icmp_redirect_timeout_q != NULL) {
1063 		if (icmp_redirtimeout == 0) {
1064 			rt_timer_queue_destroy(icmp_redirect_timeout_q);
1065 			icmp_redirect_timeout_q = NULL;
1066 		} else {
1067 			rt_timer_queue_change(icmp_redirect_timeout_q,
1068 			    icmp_redirtimeout);
1069 		}
1070 	} else if (icmp_redirtimeout > 0) {
1071 		icmp_redirect_timeout_q =
1072 		    rt_timer_queue_create(icmp_redirtimeout);
1073 	}
1074 	error = 0;
1075 out:
1076 	mutex_exit(&icmp_mtx);
1077 	return error;
1078 }
1079 
1080 static int
1081 sysctl_net_inet_icmp_stats(SYSCTLFN_ARGS)
1082 {
1083 
1084 	return (NETSTAT_SYSCTL(icmpstat_percpu, ICMP_NSTATS));
1085 }
1086 
1087 static void
1088 sysctl_netinet_icmp_setup(struct sysctllog **clog)
1089 {
1090 
1091 	sysctl_createv(clog, 0, NULL, NULL,
1092 		       CTLFLAG_PERMANENT,
1093 		       CTLTYPE_NODE, "inet", NULL,
1094 		       NULL, 0, NULL, 0,
1095 		       CTL_NET, PF_INET, CTL_EOL);
1096 	sysctl_createv(clog, 0, NULL, NULL,
1097 		       CTLFLAG_PERMANENT,
1098 		       CTLTYPE_NODE, "icmp",
1099 		       SYSCTL_DESCR("ICMPv4 related settings"),
1100 		       NULL, 0, NULL, 0,
1101 		       CTL_NET, PF_INET, IPPROTO_ICMP, CTL_EOL);
1102 
1103 	sysctl_createv(clog, 0, NULL, NULL,
1104 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1105 		       CTLTYPE_INT, "maskrepl",
1106 		       SYSCTL_DESCR("Respond to ICMP_MASKREQ messages"),
1107 		       NULL, 0, &icmpmaskrepl, 0,
1108 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1109 		       ICMPCTL_MASKREPL, CTL_EOL);
1110 	sysctl_createv(clog, 0, NULL, NULL,
1111 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1112 		       CTLTYPE_INT, "returndatabytes",
1113 		       SYSCTL_DESCR("Number of bytes to return in an ICMP "
1114 				    "error message"),
1115 		       sysctl_net_inet_icmp_returndatabytes, 0,
1116 		       &icmpreturndatabytes, 0,
1117 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1118 		       ICMPCTL_RETURNDATABYTES, CTL_EOL);
1119 	sysctl_createv(clog, 0, NULL, NULL,
1120 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1121 		       CTLTYPE_INT, "errppslimit",
1122 		       SYSCTL_DESCR("Maximum number of outgoing ICMP error "
1123 				    "messages per second"),
1124 		       NULL, 0, &icmperrppslim, 0,
1125 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1126 		       ICMPCTL_ERRPPSLIMIT, CTL_EOL);
1127 	sysctl_createv(clog, 0, NULL, NULL,
1128 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1129 		       CTLTYPE_INT, "rediraccept",
1130 		       SYSCTL_DESCR("Accept ICMP_REDIRECT messages"),
1131 		       NULL, 0, &icmp_rediraccept, 0,
1132 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1133 		       ICMPCTL_REDIRACCEPT, CTL_EOL);
1134 	sysctl_createv(clog, 0, NULL, NULL,
1135 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1136 		       CTLTYPE_INT, "redirtimeout",
1137 		       SYSCTL_DESCR("Lifetime of ICMP_REDIRECT generated "
1138 				    "routes"),
1139 		       sysctl_net_inet_icmp_redirtimeout, 0,
1140 		       &icmp_redirtimeout, 0,
1141 		       CTL_NET, PF_INET, IPPROTO_ICMP,
1142 		       ICMPCTL_REDIRTIMEOUT, CTL_EOL);
1143 	sysctl_createv(clog, 0, NULL, NULL,
1144 		       CTLFLAG_PERMANENT,
1145 		       CTLTYPE_STRUCT, "stats",
1146 		       SYSCTL_DESCR("ICMP statistics"),
1147 		       sysctl_net_inet_icmp_stats, 0, NULL, 0,
1148 		       CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_STATS,
1149 		       CTL_EOL);
1150 	sysctl_createv(clog, 0, NULL, NULL,
1151 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1152 		       CTLTYPE_INT, "bmcastecho",
1153 		       SYSCTL_DESCR("Respond to ICMP_ECHO or ICMP_TIMESTAMP "
1154 				    "message to the broadcast or multicast"),
1155 		       NULL, 0, &icmpbmcastecho, 0,
1156 		       CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_BMCASTECHO,
1157 		       CTL_EOL);
1158 }
1159 
1160 void
1161 icmp_statinc(u_int stat)
1162 {
1163 
1164 	KASSERT(stat < ICMP_NSTATS);
1165 	ICMP_STATINC(stat);
1166 }
1167 
1168 /* Table of common MTUs: */
1169 
1170 static const u_int mtu_table[] = {
1171 	65535, 65280, 32000, 17914, 9180, 8166,
1172 	4352, 2002, 1492, 1006, 508, 296, 68, 0
1173 };
1174 
1175 void
1176 icmp_mtudisc(struct icmp *icp, struct in_addr faddr)
1177 {
1178 	struct icmp_mtudisc_callback *mc;
1179 	struct sockaddr *dst = sintosa(&icmpsrc);
1180 	struct rtentry *rt;
1181 	u_long mtu = ntohs(icp->icmp_nextmtu);  /* Why a long?  IPv6 */
1182 	int    error;
1183 
1184 	rt = rtalloc1(dst, 1);
1185 	if (rt == NULL)
1186 		return;
1187 
1188 	/* If we didn't get a host route, allocate one */
1189 
1190 	if ((rt->rt_flags & RTF_HOST) == 0) {
1191 		struct rtentry *nrt;
1192 
1193 		error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL,
1194 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
1195 		if (error) {
1196 			rt_unref(rt);
1197 			return;
1198 		}
1199 		nrt->rt_rmx = rt->rt_rmx;
1200 		rt_unref(rt);
1201 		rt = nrt;
1202 	}
1203 
1204 	mutex_enter(&icmp_mtx);
1205 	error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q);
1206 	mutex_exit(&icmp_mtx);
1207 	if (error) {
1208 		rt_unref(rt);
1209 		return;
1210 	}
1211 
1212 	if (mtu == 0) {
1213 		int i = 0;
1214 
1215 		mtu = ntohs(icp->icmp_ip.ip_len);
1216 		/* Some 4.2BSD-based routers incorrectly adjust the ip_len */
1217 		if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0)
1218 			mtu -= (icp->icmp_ip.ip_hl << 2);
1219 
1220 		/* If we still can't guess a value, try the route */
1221 
1222 		if (mtu == 0) {
1223 			mtu = rt->rt_rmx.rmx_mtu;
1224 
1225 			/* If no route mtu, default to the interface mtu */
1226 
1227 			if (mtu == 0)
1228 				mtu = rt->rt_ifp->if_mtu;
1229 		}
1230 
1231 		for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++)
1232 			if (mtu > mtu_table[i]) {
1233 				mtu = mtu_table[i];
1234 				break;
1235 			}
1236 	}
1237 
1238 	/*
1239 	 * XXX:   RTV_MTU is overloaded, since the admin can set it
1240 	 *	  to turn off PMTU for a route, and the kernel can
1241 	 *	  set it to indicate a serious problem with PMTU
1242 	 *	  on a route.  We should be using a separate flag
1243 	 *	  for the kernel to indicate this.
1244 	 */
1245 
1246 	if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
1247 		if (mtu < 296 || mtu > rt->rt_ifp->if_mtu)
1248 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1249 		else if (rt->rt_rmx.rmx_mtu > mtu ||
1250 			 rt->rt_rmx.rmx_mtu == 0) {
1251 			ICMP_STATINC(ICMP_STAT_PMTUCHG);
1252 			rt->rt_rmx.rmx_mtu = mtu;
1253 		}
1254 	}
1255 
1256 	if (rt != NULL)
1257 		rt_unref(rt);
1258 
1259 	/*
1260 	 * Notify protocols that the MTU for this destination
1261 	 * has changed.
1262 	 */
1263 	mutex_enter(&icmp_mtx);
1264 	for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL;
1265 	     mc = LIST_NEXT(mc, mc_list))
1266 		(*mc->mc_func)(faddr);
1267 	mutex_exit(&icmp_mtx);
1268 }
1269 
1270 /*
1271  * Return the next larger or smaller MTU plateau (table from RFC 1191)
1272  * given current value MTU.  If DIR is less than zero, a larger plateau
1273  * is returned; otherwise, a smaller value is returned.
1274  */
1275 u_int
1276 ip_next_mtu(u_int mtu, int dir)	/* XXX */
1277 {
1278 	int i;
1279 
1280 	for (i = 0; i < (sizeof mtu_table) / (sizeof mtu_table[0]); i++) {
1281 		if (mtu >= mtu_table[i])
1282 			break;
1283 	}
1284 
1285 	if (dir < 0) {
1286 		if (i == 0) {
1287 			return 0;
1288 		} else {
1289 			return mtu_table[i - 1];
1290 		}
1291 	} else {
1292 		if (mtu_table[i] == 0) {
1293 			return 0;
1294 		} else if (mtu > mtu_table[i]) {
1295 			return mtu_table[i];
1296 		} else {
1297 			return mtu_table[i + 1];
1298 		}
1299 	}
1300 }
1301 
1302 static void
1303 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
1304 {
1305 
1306 	KASSERT(rt != NULL);
1307 	rt_assert_referenced(rt);
1308 
1309 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1310 	    (RTF_DYNAMIC | RTF_HOST)) {
1311 		rtrequest(RTM_DELETE, rt_getkey(rt),
1312 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
1313 	} else {
1314 		if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
1315 			rt->rt_rmx.rmx_mtu = 0;
1316 		}
1317 	}
1318 }
1319 
1320 static void
1321 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r)
1322 {
1323 
1324 	KASSERT(rt != NULL);
1325 	rt_assert_referenced(rt);
1326 
1327 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1328 	    (RTF_DYNAMIC | RTF_HOST)) {
1329 		rtrequest(RTM_DELETE, rt_getkey(rt),
1330 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
1331 	}
1332 }
1333 
1334 /*
1335  * Perform rate limit check.
1336  * Returns 0 if it is okay to send the icmp packet.
1337  * Returns 1 if the router SHOULD NOT send this icmp packet due to rate
1338  * limitation.
1339  *
1340  * XXX per-destination/type check necessary?
1341  */
1342 int
1343 icmp_ratelimit(const struct in_addr *dst, const int type,
1344     const int code)
1345 {
1346 
1347 	/* PPS limit */
1348 	if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count,
1349 	    icmperrppslim)) {
1350 		/* The packet is subject to rate limit */
1351 		return 1;
1352 	}
1353 
1354 	/* okay to send */
1355 	return 0;
1356 }
1357