xref: /openbsd-src/sys/netinet/ip_icmp.c (revision 50b7afb2c2c0993b0894d4e34bf857cb13ed9c80)
1 /*	$OpenBSD: ip_icmp.c,v 1.123 2014/07/13 13:57:56 mpi Exp $	*/
2 /*	$NetBSD: ip_icmp.c,v 1.19 1996/02/13 23:42:22 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)COPYRIGHT	1.1 (NRL) 17 January 1995
33  *
34  * NRL grants permission for redistribution and use in source and binary
35  * forms, with or without modification, of the software and documentation
36  * created at NRL provided that the following conditions are met:
37  *
38  * 1. Redistributions of source code must retain the above copyright
39  *    notice, this list of conditions and the following disclaimer.
40  * 2. Redistributions in binary form must reproduce the above copyright
41  *    notice, this list of conditions and the following disclaimer in the
42  *    documentation and/or other materials provided with the distribution.
43  * 3. All advertising materials mentioning features or use of this software
44  *    must display the following acknowledgements:
45  *	This product includes software developed by the University of
46  *	California, Berkeley and its contributors.
47  *	This product includes software developed at the Information
48  *	Technology Division, US Naval Research Laboratory.
49  * 4. Neither the name of the NRL nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
54  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
55  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
56  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
57  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
58  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
59  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
60  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
61  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
62  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
63  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
64  *
65  * The views and conclusions contained in the software and documentation
66  * are those of the authors and should not be interpreted as representing
67  * official policies, either expressed or implied, of the US Naval
68  * Research Laboratory (NRL).
69  */
70 
71 #include "carp.h"
72 #include "pf.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/mbuf.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/proc.h>
80 #include <sys/sysctl.h>
81 
82 #include <net/if.h>
83 #include <net/route.h>
84 
85 #include <netinet/in.h>
86 #include <netinet/in_systm.h>
87 #include <netinet/in_var.h>
88 #include <netinet/ip.h>
89 #include <netinet/ip_icmp.h>
90 #include <netinet/ip_var.h>
91 #include <netinet/icmp_var.h>
92 
93 #if NCARP > 0
94 #include <net/if_types.h>
95 #include <netinet/ip_carp.h>
96 #endif
97 
98 #if NPF > 0
99 #include <net/pfvar.h>
100 #endif
101 
102 /*
103  * ICMP routines: error generation, receive packet processing, and
104  * routines to turnaround packets back to the originator, and
105  * host table maintenance routines.
106  */
107 
108 #ifdef ICMPPRINTFS
109 int	icmpprintfs = 0;	/* Settable from ddb */
110 #endif
111 
112 /* values controllable via sysctl */
113 int	icmpmaskrepl = 0;
114 int	icmpbmcastecho = 0;
115 int	icmptstamprepl = 1;
116 int	icmperrppslim = 100;
117 int	icmp_rediraccept = 0;
118 int	icmp_redirtimeout = 10 * 60;
119 
120 static int icmperrpps_count = 0;
121 static struct timeval icmperrppslim_last;
122 
123 static struct rttimer_queue *icmp_redirect_timeout_q = NULL;
124 struct	icmpstat icmpstat;
125 
126 int *icmpctl_vars[ICMPCTL_MAXID] = ICMPCTL_VARS;
127 
128 void icmp_mtudisc_timeout(struct rtentry *, struct rttimer *);
129 int icmp_ratelimit(const struct in_addr *, const int, const int);
130 void icmp_redirect_timeout(struct rtentry *, struct rttimer *);
131 
132 void
133 icmp_init(void)
134 {
135 	/*
136 	 * This is only useful if the user initializes redirtimeout to
137 	 * something other than zero.
138 	 */
139 	if (icmp_redirtimeout != 0) {
140 		icmp_redirect_timeout_q =
141 		    rt_timer_queue_create(icmp_redirtimeout);
142 	}
143 }
144 
145 struct mbuf *
146 icmp_do_error(struct mbuf *n, int type, int code, u_int32_t dest, int destmtu)
147 {
148 	struct ip *oip = mtod(n, struct ip *), *nip;
149 	unsigned oiplen = oip->ip_hl << 2;
150 	struct icmp *icp;
151 	struct mbuf *m;
152 	unsigned icmplen, mblen;
153 
154 #ifdef ICMPPRINTFS
155 	if (icmpprintfs)
156 		printf("icmp_error(%x, %d, %d)\n", oip, type, code);
157 #endif
158 	if (type != ICMP_REDIRECT)
159 		icmpstat.icps_error++;
160 	/*
161 	 * Don't send error if not the first fragment of message.
162 	 * Don't error if the old packet protocol was ICMP
163 	 * error message, only known informational types.
164 	 */
165 	if (oip->ip_off & htons(IP_OFFMASK))
166 		goto freeit;
167 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
168 	    n->m_len >= oiplen + ICMP_MINLEN &&
169 	    !ICMP_INFOTYPE(((struct icmp *)
170 	    ((caddr_t)oip + oiplen))->icmp_type)) {
171 		icmpstat.icps_oldicmp++;
172 		goto freeit;
173 	}
174 	/* Don't send error in response to a multicast or broadcast packet */
175 	if (n->m_flags & (M_BCAST|M_MCAST))
176 		goto freeit;
177 
178 	/*
179 	 * First, do a rate limitation check.
180 	 */
181 	if (icmp_ratelimit(&oip->ip_src, type, code)) {
182 		icmpstat.icps_toofreq++;
183 		goto freeit;
184 	}
185 
186 	/*
187 	 * Now, formulate icmp message
188 	 */
189 	icmplen = oiplen + min(8, ntohs(oip->ip_len));
190 	/*
191 	 * Defend against mbuf chains shorter than oip->ip_len:
192 	 */
193 	mblen = 0;
194 	for (m = n; m && (mblen < icmplen); m = m->m_next)
195 		mblen += m->m_len;
196 	icmplen = min(mblen, icmplen);
197 
198 	/*
199 	 * As we are not required to return everything we have,
200 	 * we return whatever we can return at ease.
201 	 *
202 	 * Note that ICMP datagrams longer than 576 octets are out of spec
203 	 * according to RFC1812;
204 	 */
205 
206 	KASSERT(ICMP_MINLEN <= MCLBYTES);
207 
208 	if (icmplen + ICMP_MINLEN > MCLBYTES)
209 		icmplen = MCLBYTES - ICMP_MINLEN - sizeof (struct ip);
210 
211 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
212 	if (m && (sizeof (struct ip) + icmplen + ICMP_MINLEN > MHLEN)) {
213 		MCLGET(m, M_DONTWAIT);
214 		if ((m->m_flags & M_EXT) == 0) {
215 			m_freem(m);
216 			m = NULL;
217 		}
218 	}
219 	if (m == NULL)
220 		goto freeit;
221 	/* keep in same rtable */
222 	m->m_pkthdr.ph_rtableid = n->m_pkthdr.ph_rtableid;
223 	m->m_len = icmplen + ICMP_MINLEN;
224 	if ((m->m_flags & M_EXT) == 0)
225 		MH_ALIGN(m, m->m_len);
226 	icp = mtod(m, struct icmp *);
227 	if ((u_int)type > ICMP_MAXTYPE)
228 		panic("icmp_error");
229 	icmpstat.icps_outhist[type]++;
230 	icp->icmp_type = type;
231 	if (type == ICMP_REDIRECT)
232 		icp->icmp_gwaddr.s_addr = dest;
233 	else {
234 		icp->icmp_void = 0;
235 		/*
236 		 * The following assignments assume an overlay with the
237 		 * zeroed icmp_void field.
238 		 */
239 		if (type == ICMP_PARAMPROB) {
240 			icp->icmp_pptr = code;
241 			code = 0;
242 		} else if (type == ICMP_UNREACH &&
243 		    code == ICMP_UNREACH_NEEDFRAG && destmtu)
244 			icp->icmp_nextmtu = htons(destmtu);
245 	}
246 
247 	icp->icmp_code = code;
248 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
249 
250 	/*
251 	 * Now, copy old ip header (without options)
252 	 * in front of icmp message.
253 	 */
254 	if ((m->m_flags & M_EXT) == 0 &&
255 	    m->m_data - sizeof(struct ip) < m->m_pktdat)
256 		panic("icmp len");
257 	m->m_data -= sizeof(struct ip);
258 	m->m_len += sizeof(struct ip);
259 	m->m_pkthdr.len = m->m_len;
260 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
261 	nip = mtod(m, struct ip *);
262 	/* ip_v set in ip_output */
263 	nip->ip_hl = sizeof(struct ip) >> 2;
264 	nip->ip_tos = 0;
265 	nip->ip_len = htons(m->m_len);
266 	/* ip_id set in ip_output */
267 	nip->ip_off = 0;
268 	/* ip_ttl set in icmp_reflect */
269 	nip->ip_p = IPPROTO_ICMP;
270 	nip->ip_src = oip->ip_src;
271 	nip->ip_dst = oip->ip_dst;
272 
273 	/* move PF_GENERATED to new packet, if existent XXX preserve more? */
274 	if (n->m_pkthdr.pf.flags & PF_TAG_GENERATED)
275 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
276 
277 	m_freem(n);
278 	return (m);
279 
280 freeit:
281 	m_freem(n);
282 	return (NULL);
283 }
284 
285 /*
286  * Generate an error packet of type error
287  * in response to bad packet ip.
288  *
289  * The ip packet inside has ip_off and ip_len in host byte order.
290  */
291 void
292 icmp_error(struct mbuf *n, int type, int code, u_int32_t dest, int destmtu)
293 {
294 	struct mbuf *m;
295 
296 	m = icmp_do_error(n, type, code, dest, destmtu);
297 	if (m != NULL)
298 		if (!icmp_reflect(m, NULL, NULL))
299 			icmp_send(m, NULL);
300 }
301 
302 /*
303  * Process a received ICMP message.
304  */
305 void
306 icmp_input(struct mbuf *m, ...)
307 {
308 	struct ifnet *ifp;
309 	struct icmp *icp;
310 	struct ip *ip = mtod(m, struct ip *);
311 	struct sockaddr_in sin;
312 	int icmplen, i, code, hlen;
313 	struct in_ifaddr *ia;
314 	void *(*ctlfunc)(int, struct sockaddr *, u_int, void *);
315 	va_list ap;
316 	struct mbuf *opts;
317 
318 	va_start(ap, m);
319 	hlen = va_arg(ap, int);
320 	va_end(ap);
321 
322 	ifp = m->m_pkthdr.rcvif;
323 
324 	/*
325 	 * Locate icmp structure in mbuf, and check
326 	 * that not corrupted and of at least minimum length.
327 	 */
328 	icmplen = ntohs(ip->ip_len) - hlen;
329 #ifdef ICMPPRINTFS
330 	if (icmpprintfs) {
331 		char dst[INET_ADDRSTRLEN], src[INET_ADDRSTRLEN];
332 
333 		inet_ntop(AF_INET, &ip->ip_dst, dst, sizeof(dst));
334 		inet_ntop(AF_INET, &ip->ip_src, src, sizeof(src));
335 
336 		printf("icmp_input from %s to %s, len %d\n", src, dst, icmplen);
337 	}
338 #endif
339 	if (icmplen < ICMP_MINLEN) {
340 		icmpstat.icps_tooshort++;
341 		goto freeit;
342 	}
343 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
344 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
345 		icmpstat.icps_tooshort++;
346 		return;
347 	}
348 	ip = mtod(m, struct ip *);
349 	if (in4_cksum(m, 0, hlen, icmplen)) {
350 		icmpstat.icps_checksum++;
351 		goto freeit;
352 	}
353 
354 	icp = (struct icmp *)(mtod(m, caddr_t) + hlen);
355 #ifdef ICMPPRINTFS
356 	/*
357 	 * Message type specific processing.
358 	 */
359 	if (icmpprintfs)
360 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
361 		    icp->icmp_code);
362 #endif
363 	if (icp->icmp_type > ICMP_MAXTYPE)
364 		goto raw;
365 #if NPF > 0
366 	if (m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) {
367 		switch (icp->icmp_type) {
368 		/*
369 		 * These ICMP types map to other connections.  They must be
370 		 * delivered to pr_ctlinput() also for diverted connections.
371 		 */
372 		case ICMP_UNREACH:
373 		case ICMP_TIMXCEED:
374 		case ICMP_PARAMPROB:
375 		case ICMP_SOURCEQUENCH:
376 			break;
377 		 /*
378 		  * Although pf_icmp_mapping() considers redirects belonging
379 		  * to a diverted connection, we must process it here anyway.
380 		  */
381 		case ICMP_REDIRECT:
382 			break;
383 		default:
384 			goto raw;
385 		}
386 	}
387 #endif /* NPF */
388 	icmpstat.icps_inhist[icp->icmp_type]++;
389 	code = icp->icmp_code;
390 	switch (icp->icmp_type) {
391 
392 	case ICMP_UNREACH:
393 		switch (code) {
394 		case ICMP_UNREACH_NET:
395 		case ICMP_UNREACH_HOST:
396 		case ICMP_UNREACH_PROTOCOL:
397 		case ICMP_UNREACH_PORT:
398 		case ICMP_UNREACH_SRCFAIL:
399 			code += PRC_UNREACH_NET;
400 			break;
401 
402 		case ICMP_UNREACH_NEEDFRAG:
403 			code = PRC_MSGSIZE;
404 			break;
405 
406 		case ICMP_UNREACH_NET_UNKNOWN:
407 		case ICMP_UNREACH_NET_PROHIB:
408 		case ICMP_UNREACH_TOSNET:
409 			code = PRC_UNREACH_NET;
410 			break;
411 
412 		case ICMP_UNREACH_HOST_UNKNOWN:
413 		case ICMP_UNREACH_ISOLATED:
414 		case ICMP_UNREACH_HOST_PROHIB:
415 		case ICMP_UNREACH_TOSHOST:
416 		case ICMP_UNREACH_FILTER_PROHIB:
417 		case ICMP_UNREACH_HOST_PRECEDENCE:
418 		case ICMP_UNREACH_PRECEDENCE_CUTOFF:
419 			code = PRC_UNREACH_HOST;
420 			break;
421 
422 		default:
423 			goto badcode;
424 		}
425 		goto deliver;
426 
427 	case ICMP_TIMXCEED:
428 		if (code > 1)
429 			goto badcode;
430 		code += PRC_TIMXCEED_INTRANS;
431 		goto deliver;
432 
433 	case ICMP_PARAMPROB:
434 		if (code > 1)
435 			goto badcode;
436 		code = PRC_PARAMPROB;
437 		goto deliver;
438 
439 	case ICMP_SOURCEQUENCH:
440 		if (code)
441 			goto badcode;
442 		code = PRC_QUENCH;
443 	deliver:
444 		/* Free packet atttributes */
445 		if (m->m_flags & M_PKTHDR)
446 			m_tag_delete_chain(m);
447 
448 		/*
449 		 * Problem with datagram; advise higher level routines.
450 		 */
451 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
452 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
453 			icmpstat.icps_badlen++;
454 			goto freeit;
455 		}
456 		if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr))
457 			goto badcode;
458 #ifdef INET6
459 		/* Get more contiguous data for a v6 in v4 ICMP message. */
460 		if (icp->icmp_ip.ip_p == IPPROTO_IPV6) {
461 			if (icmplen < ICMP_V6ADVLENMIN ||
462 			    icmplen < ICMP_V6ADVLEN(icp)) {
463 				icmpstat.icps_badlen++;
464 				goto freeit;
465 			} else {
466 				if ((m = m_pullup(m, (ip->ip_hl << 2) +
467 				    ICMP_V6ADVLEN(icp))) == NULL) {
468 					icmpstat.icps_tooshort++;
469 					return;
470 				}
471 				ip = mtod(m, struct ip *);
472 				icp = (struct icmp *)
473 				    (m->m_data + (ip->ip_hl << 2));
474 			}
475 		}
476 #endif /* INET6 */
477 #ifdef ICMPPRINTFS
478 		if (icmpprintfs)
479 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
480 #endif
481 		memset(&sin, 0, sizeof(sin));
482 		sin.sin_family = AF_INET;
483 		sin.sin_len = sizeof(struct sockaddr_in);
484 		sin.sin_addr = icp->icmp_ip.ip_dst;
485 #if NCARP > 0
486 		if (ifp->if_type == IFT_CARP &&
487 		    carp_lsdrop(m, AF_INET, &sin.sin_addr.s_addr,
488 		    &ip->ip_dst.s_addr))
489 			goto freeit;
490 #endif
491 		/*
492 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
493 		 * notification to TCP layer.
494 		 */
495 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
496 		if (ctlfunc)
497 			(*ctlfunc)(code, sintosa(&sin), m->m_pkthdr.ph_rtableid,
498 			    &icp->icmp_ip);
499 		break;
500 
501 	badcode:
502 		icmpstat.icps_badcode++;
503 		break;
504 
505 	case ICMP_ECHO:
506 		if (!icmpbmcastecho &&
507 		    (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
508 			icmpstat.icps_bmcastecho++;
509 			break;
510 		}
511 		icp->icmp_type = ICMP_ECHOREPLY;
512 		goto reflect;
513 
514 	case ICMP_TSTAMP:
515 		if (icmptstamprepl == 0)
516 			break;
517 
518 		if (!icmpbmcastecho &&
519 		    (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
520 			icmpstat.icps_bmcastecho++;
521 			break;
522 		}
523 		if (icmplen < ICMP_TSLEN) {
524 			icmpstat.icps_badlen++;
525 			break;
526 		}
527 		icp->icmp_type = ICMP_TSTAMPREPLY;
528 		icp->icmp_rtime = iptime();
529 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
530 		goto reflect;
531 
532 	case ICMP_MASKREQ:
533 		if (icmpmaskrepl == 0)
534 			break;
535 		if (icmplen < ICMP_MASKLEN) {
536 			icmpstat.icps_badlen++;
537 			break;
538 		}
539 		/*
540 		 * We are not able to respond with all ones broadcast
541 		 * unless we receive it over a point-to-point interface.
542 		 */
543 		memset(&sin, 0, sizeof(sin));
544 		sin.sin_family = AF_INET;
545 		sin.sin_len = sizeof(struct sockaddr_in);
546 		if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
547 		    ip->ip_dst.s_addr == INADDR_ANY)
548 			sin.sin_addr = ip->ip_src;
549 		else
550 			sin.sin_addr = ip->ip_dst;
551 		if (ifp == NULL)
552 			break;
553 		ia = ifatoia(ifaof_ifpforaddr(sintosa(&sin), ifp));
554 		if (ia == 0)
555 			break;
556 		icp->icmp_type = ICMP_MASKREPLY;
557 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
558 		if (ip->ip_src.s_addr == 0) {
559 			if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
560 				if (ia->ia_broadaddr.sin_addr.s_addr)
561 					ip->ip_src = ia->ia_broadaddr.sin_addr;
562 				else
563 					ip->ip_src.s_addr = INADDR_BROADCAST;
564 			}
565 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
566 				ip->ip_src = ia->ia_dstaddr.sin_addr;
567 		}
568 reflect:
569 #if NCARP > 0
570 		if (ifp->if_type == IFT_CARP &&
571 		    carp_lsdrop(m, AF_INET, &ip->ip_src.s_addr,
572 		    &ip->ip_dst.s_addr))
573 			goto freeit;
574 #endif
575 		/* Free packet atttributes */
576 		if (m->m_flags & M_PKTHDR)
577 			m_tag_delete_chain(m);
578 
579 		icmpstat.icps_reflect++;
580 		icmpstat.icps_outhist[icp->icmp_type]++;
581 		if (!icmp_reflect(m, &opts, NULL))
582 			icmp_send(m, opts);
583 		return;
584 
585 	case ICMP_REDIRECT:
586 	{
587 		struct sockaddr_in sdst;
588 		struct sockaddr_in sgw;
589 		struct sockaddr_in ssrc;
590 		struct rtentry *newrt = NULL;
591 
592 		/* Free packet atttributes */
593 		if (m->m_flags & M_PKTHDR)
594 			m_tag_delete_chain(m);
595 		if (icmp_rediraccept == 0 || ipforwarding == 1)
596 			goto freeit;
597 		if (code > 3)
598 			goto badcode;
599 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
600 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
601 			icmpstat.icps_badlen++;
602 			break;
603 		}
604 		/*
605 		 * Short circuit routing redirects to force
606 		 * immediate change in the kernel's routing
607 		 * tables.  The message is also handed to anyone
608 		 * listening on a raw socket (e.g. the routing
609 		 * daemon for use in updating its tables).
610 		 */
611 		memset(&sdst, 0, sizeof(sdst));
612 		memset(&sgw, 0, sizeof(sgw));
613 		memset(&ssrc, 0, sizeof(ssrc));
614 		sdst.sin_family = sgw.sin_family = ssrc.sin_family = AF_INET;
615 		sdst.sin_len = sgw.sin_len = ssrc.sin_len = sizeof(sdst);
616 		memcpy(&sdst.sin_addr, &icp->icmp_ip.ip_dst,
617 		    sizeof(sdst.sin_addr));
618 		memcpy(&sgw.sin_addr, &icp->icmp_gwaddr,
619 		    sizeof(sgw.sin_addr));
620 		memcpy(&ssrc.sin_addr, &ip->ip_src,
621 		    sizeof(ssrc.sin_addr));
622 
623 #ifdef	ICMPPRINTFS
624 		if (icmpprintfs) {
625 			char gw[INET_ADDRSTRLEN], dst[INET_ADDRSTRLEN];
626 
627 			inet_ntop(AF_INET, &icp->icmp_gwaddr, gw, sizeof(gw));
628 			inet_ntop(AF_INET, &icp->icmp_ip.ip_dst,
629 			    dst, sizeof(dst));
630 
631 			printf("redirect dst %s to %s\n", dst, gw);
632 		}
633 #endif
634 
635 #if NCARP > 0
636 		if (ifp->if_type == IFT_CARP &&
637 		    carp_lsdrop(m, AF_INET, &sdst.sin_addr.s_addr,
638 		    &ip->ip_dst.s_addr))
639 			goto freeit;
640 #endif
641 		rtredirect(sintosa(&sdst), sintosa(&sgw), NULL,
642 		    RTF_GATEWAY | RTF_HOST, sintosa(&ssrc),
643 		    &newrt, m->m_pkthdr.ph_rtableid);
644 		if (newrt != NULL && icmp_redirtimeout != 0) {
645 			(void)rt_timer_add(newrt, icmp_redirect_timeout,
646 			    icmp_redirect_timeout_q, m->m_pkthdr.ph_rtableid);
647 		}
648 		if (newrt != NULL)
649 			rtfree(newrt);
650 		pfctlinput(PRC_REDIRECT_HOST, sintosa(&sdst));
651 		break;
652 	}
653 	/*
654 	 * No kernel processing for the following;
655 	 * just fall through to send to raw listener.
656 	 */
657 	case ICMP_ECHOREPLY:
658 	case ICMP_ROUTERADVERT:
659 	case ICMP_ROUTERSOLICIT:
660 	case ICMP_TSTAMPREPLY:
661 	case ICMP_IREQREPLY:
662 	case ICMP_MASKREPLY:
663 	case ICMP_TRACEROUTE:
664 	case ICMP_DATACONVERR:
665 	case ICMP_MOBILE_REDIRECT:
666 	case ICMP_IPV6_WHEREAREYOU:
667 	case ICMP_IPV6_IAMHERE:
668 	case ICMP_MOBILE_REGREQUEST:
669 	case ICMP_MOBILE_REGREPLY:
670 	case ICMP_PHOTURIS:
671 	default:
672 		break;
673 	}
674 
675 raw:
676 	rip_input(m);
677 	return;
678 
679 freeit:
680 	m_freem(m);
681 }
682 
683 /*
684  * Reflect the ip packet back to the source
685  */
686 int
687 icmp_reflect(struct mbuf *m, struct mbuf **op, struct in_ifaddr *ia)
688 {
689 	struct ip *ip = mtod(m, struct ip *);
690 	struct in_addr t;
691 	struct mbuf *opts = 0;
692 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
693 
694 	if (!in_canforward(ip->ip_src) &&
695 	    ((ip->ip_src.s_addr & IN_CLASSA_NET) !=
696 	    htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
697 		m_freem(m);		/* Bad return address */
698 		return (EHOSTUNREACH);
699 	}
700 
701 #if NPF > 0
702 	pf_pkt_addr_changed(m);
703 #endif
704 	t = ip->ip_dst;
705 	ip->ip_dst = ip->ip_src;
706 	/*
707 	 * If the incoming packet was addressed directly to us,
708 	 * use dst as the src for the reply.  For broadcast, use
709 	 * the address which corresponds to the incoming interface.
710 	 */
711 	if (ia == NULL) {
712 		TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
713 			if (ia->ia_ifp->if_rdomain !=
714 			    rtable_l2(m->m_pkthdr.ph_rtableid))
715 				continue;
716 			if (t.s_addr == ia->ia_addr.sin_addr.s_addr)
717 				break;
718 			if ((ia->ia_ifp->if_flags & IFF_BROADCAST) &&
719 			    ia->ia_broadaddr.sin_addr.s_addr != 0 &&
720 			    t.s_addr == ia->ia_broadaddr.sin_addr.s_addr)
721 				break;
722 		}
723 	}
724 	/*
725 	 * The following happens if the packet was not addressed to us.
726 	 * Use the new source address and do a route lookup. If it fails
727 	 * drop the packet as there is no path to the host.
728 	 */
729 	if (ia == NULL) {
730 		struct sockaddr_in *dst;
731 		struct route ro;
732 
733 		memset(&ro, 0, sizeof(ro));
734 		ro.ro_tableid = m->m_pkthdr.ph_rtableid;
735 		dst = satosin(&ro.ro_dst);
736 		dst->sin_family = AF_INET;
737 		dst->sin_len = sizeof(*dst);
738 		dst->sin_addr = ip->ip_src;
739 
740 		/* keep packet in the original virtual instance */
741 		ro.ro_rt = rtalloc1(&ro.ro_dst, RT_REPORT,
742 		     m->m_pkthdr.ph_rtableid);
743 		if (ro.ro_rt == 0) {
744 			ipstat.ips_noroute++;
745 			m_freem(m);
746 			return (EHOSTUNREACH);
747 		}
748 
749 		ia = ifatoia(ro.ro_rt->rt_ifa);
750 		ro.ro_rt->rt_use++;
751 		RTFREE(ro.ro_rt);
752 	}
753 
754 	t = ia->ia_addr.sin_addr;
755 	ip->ip_src = t;
756 	ip->ip_ttl = MAXTTL;
757 
758 	if (optlen > 0) {
759 		u_char *cp;
760 		int opt, cnt;
761 		u_int len;
762 
763 		/*
764 		 * Retrieve any source routing from the incoming packet;
765 		 * add on any record-route or timestamp options.
766 		 */
767 		cp = (u_char *) (ip + 1);
768 		if (op && (opts = ip_srcroute(m)) == NULL &&
769 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
770 			opts->m_len = sizeof(struct in_addr);
771 			mtod(opts, struct in_addr *)->s_addr = 0;
772 		}
773 		if (op && opts) {
774 #ifdef ICMPPRINTFS
775 			if (icmpprintfs)
776 				printf("icmp_reflect optlen %d rt %d => ",
777 				    optlen, opts->m_len);
778 #endif
779 			for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
780 				opt = cp[IPOPT_OPTVAL];
781 				if (opt == IPOPT_EOL)
782 					break;
783 				if (opt == IPOPT_NOP)
784 					len = 1;
785 				else {
786 					if (cnt < IPOPT_OLEN + sizeof(*cp))
787 						break;
788 					len = cp[IPOPT_OLEN];
789 					if (len < IPOPT_OLEN + sizeof(*cp) ||
790 					    len > cnt)
791 						break;
792 				}
793 				/*
794 				 * Should check for overflow, but it
795 				 * "can't happen"
796 				 */
797 				if (opt == IPOPT_RR || opt == IPOPT_TS ||
798 				    opt == IPOPT_SECURITY) {
799 					memcpy(mtod(opts, caddr_t) +
800 					    opts->m_len, cp, len);
801 					opts->m_len += len;
802 				}
803 			}
804 			/* Terminate & pad, if necessary */
805 			if ((cnt = opts->m_len % 4) != 0)
806 				for (; cnt < 4; cnt++) {
807 					*(mtod(opts, caddr_t) + opts->m_len) =
808 					    IPOPT_EOL;
809 					opts->m_len++;
810 				}
811 #ifdef ICMPPRINTFS
812 			if (icmpprintfs)
813 				printf("%d\n", opts->m_len);
814 #endif
815 		}
816 		ip_stripoptions(m);
817 	}
818 	m->m_flags &= ~(M_BCAST|M_MCAST);
819 	if (op)
820 		*op = opts;
821 
822 	return (0);
823 }
824 
825 /*
826  * Send an icmp packet back to the ip level
827  */
828 void
829 icmp_send(struct mbuf *m, struct mbuf *opts)
830 {
831 	struct ip *ip = mtod(m, struct ip *);
832 	int hlen;
833 	struct icmp *icp;
834 
835 	hlen = ip->ip_hl << 2;
836 	icp = (struct icmp *)(mtod(m, caddr_t) + hlen);
837 	icp->icmp_cksum = 0;
838 	m->m_pkthdr.csum_flags |= M_ICMP_CSUM_OUT;
839 #ifdef ICMPPRINTFS
840 	if (icmpprintfs) {
841 		char dst[INET_ADDRSTRLEN], src[INET_ADDRSTRLEN];
842 
843 		inet_ntop(AF_INET, &ip->ip_dst, dst, sizeof(dst));
844 		inet_ntop(AF_INET, &ip->ip_src, src, sizeof(src));
845 
846 		printf("icmp_send dst %s src %s\n", dst, src);
847 	}
848 #endif
849 	ip_output(m, opts, NULL, 0, NULL, NULL, 0);
850 }
851 
852 u_int32_t
853 iptime(void)
854 {
855 	struct timeval atv;
856 	u_long t;
857 
858 	microtime(&atv);
859 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
860 	return (htonl(t));
861 }
862 
863 int
864 icmp_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
865     size_t newlen)
866 {
867 	int s, error;
868 
869 	/* All sysctl names at this level are terminal. */
870 	if (namelen != 1)
871 		return (ENOTDIR);
872 
873 	s = splsoftnet();
874 	switch (name[0]) {
875 	case ICMPCTL_REDIRTIMEOUT:
876 
877 		error = sysctl_int(oldp, oldlenp, newp, newlen,
878 		    &icmp_redirtimeout);
879 		if (icmp_redirect_timeout_q != NULL) {
880 			if (icmp_redirtimeout == 0) {
881 				rt_timer_queue_destroy(icmp_redirect_timeout_q);
882 				icmp_redirect_timeout_q = NULL;
883 			} else
884 				rt_timer_queue_change(icmp_redirect_timeout_q,
885 				    icmp_redirtimeout);
886 		} else if (icmp_redirtimeout > 0) {
887 			icmp_redirect_timeout_q =
888 			    rt_timer_queue_create(icmp_redirtimeout);
889 		}
890 		break;
891 
892 	case ICMPCTL_STATS:
893 		if (newp != NULL) {
894 			error = EPERM;
895 			break;
896 		}
897 		error = sysctl_struct(oldp, oldlenp, newp, newlen,
898 		    &icmpstat, sizeof(icmpstat));
899 		break;
900 
901 	default:
902 		if (name[0] < ICMPCTL_MAXID) {
903 			error = sysctl_int_arr(icmpctl_vars, name, namelen,
904 			    oldp, oldlenp, newp, newlen);
905 			break;
906 		}
907 		error = ENOPROTOOPT;
908 		break;
909 	}
910 	splx(s);
911 
912 	return (error);
913 }
914 
915 
916 struct rtentry *
917 icmp_mtudisc_clone(struct in_addr dst, u_int rtableid)
918 {
919 	struct sockaddr_in *sin;
920 	struct route ro;
921 	struct rtentry *rt;
922 	int error;
923 
924 	memset(&ro, 0, sizeof(ro));
925 	ro.ro_tableid = rtableid;
926 	sin = satosin(&ro.ro_dst);
927 	sin->sin_family = AF_INET;
928 	sin->sin_len = sizeof(*sin);
929 	sin->sin_addr = dst;
930 
931 	rt = rtalloc1(&ro.ro_dst, RT_REPORT, rtableid);
932 	if (rt == NULL)
933 		return (NULL);
934 
935 	/* Check if the route is actually usable */
936 	if (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE) ||
937 	    (rt->rt_flags & RTF_UP) == 0)
938 		return (NULL);
939 
940 	/* If we didn't get a host route, allocate one */
941 
942 	if ((rt->rt_flags & RTF_HOST) == 0) {
943 		struct rtentry *nrt;
944 		struct rt_addrinfo info;
945 
946 		memset(&info, 0, sizeof(info));
947 		info.rti_info[RTAX_DST] = sintosa(sin);
948 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
949 		info.rti_flags = RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC;
950 
951 		error = rtrequest1(RTM_ADD, &info, RTP_DEFAULT, &nrt, rtableid);
952 		if (error) {
953 			rtfree(rt);
954 			return (NULL);
955 		}
956 		nrt->rt_rmx = rt->rt_rmx;
957 		rtfree(rt);
958 		rt = nrt;
959 	}
960 	error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q,
961 	    rtableid);
962 	if (error) {
963 		rtfree(rt);
964 		return (NULL);
965 	}
966 
967 	return (rt);
968 }
969 
970 /* Table of common MTUs: */
971 static const u_short mtu_table[] = {
972 	65535, 65280, 32000, 17914, 9180, 8166,
973 	4352, 2002, 1492, 1006, 508, 296, 68, 0
974 };
975 
976 void
977 icmp_mtudisc(struct icmp *icp, u_int rtableid)
978 {
979 	struct rtentry *rt;
980 	u_long mtu = ntohs(icp->icmp_nextmtu);  /* Why a long?  IPv6 */
981 
982 	rt = icmp_mtudisc_clone(icp->icmp_ip.ip_dst, rtableid);
983 	if (rt == 0)
984 		return;
985 
986 	if (mtu == 0) {
987 		int i = 0;
988 
989 		mtu = ntohs(icp->icmp_ip.ip_len);
990 		/* Some 4.2BSD-based routers incorrectly adjust the ip_len */
991 		if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0)
992 			mtu -= (icp->icmp_ip.ip_hl << 2);
993 
994 		/* If we still can't guess a value, try the route */
995 		if (mtu == 0) {
996 			mtu = rt->rt_rmx.rmx_mtu;
997 
998 			/* If no route mtu, default to the interface mtu */
999 
1000 			if (mtu == 0)
1001 				mtu = rt->rt_ifp->if_mtu;
1002 		}
1003 
1004 		for (i = 0; i < nitems(mtu_table); i++)
1005 			if (mtu > mtu_table[i]) {
1006 				mtu = mtu_table[i];
1007 				break;
1008 			}
1009 	}
1010 
1011 	/*
1012 	 * XXX:   RTV_MTU is overloaded, since the admin can set it
1013 	 *	  to turn off PMTU for a route, and the kernel can
1014 	 *	  set it to indicate a serious problem with PMTU
1015 	 *	  on a route.  We should be using a separate flag
1016 	 *	  for the kernel to indicate this.
1017 	 */
1018 
1019 	if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
1020 		if (mtu < 296 || mtu > rt->rt_ifp->if_mtu)
1021 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1022 		else if (rt->rt_rmx.rmx_mtu > mtu ||
1023 		    rt->rt_rmx.rmx_mtu == 0)
1024 			rt->rt_rmx.rmx_mtu = mtu;
1025 	}
1026 
1027 	rtfree(rt);
1028 }
1029 
1030 void
1031 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
1032 {
1033 	if (rt == NULL)
1034 		panic("icmp_mtudisc_timeout:  bad route to timeout");
1035 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1036 	    (RTF_DYNAMIC | RTF_HOST)) {
1037 		void *(*ctlfunc)(int, struct sockaddr *, u_int, void *);
1038 		struct sockaddr_in sa;
1039 		struct rt_addrinfo info;
1040 
1041 		memset(&info, 0, sizeof(info));
1042 		info.rti_info[RTAX_DST] = rt_key(rt);
1043 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1044 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1045 		info.rti_flags = rt->rt_flags;
1046 
1047 		sa = *(struct sockaddr_in *)rt_key(rt);
1048 		rtrequest1(RTM_DELETE, &info, rt->rt_priority, NULL,
1049 		    r->rtt_tableid);
1050 
1051 		/* Notify TCP layer of increased Path MTU estimate */
1052 		ctlfunc = inetsw[ip_protox[IPPROTO_TCP]].pr_ctlinput;
1053 		if (ctlfunc)
1054 			(*ctlfunc)(PRC_MTUINC,(struct sockaddr *)&sa,
1055 			    r->rtt_tableid, NULL);
1056 	} else
1057 		if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
1058 			rt->rt_rmx.rmx_mtu = 0;
1059 }
1060 
1061 /*
1062  * Perform rate limit check.
1063  * Returns 0 if it is okay to send the icmp packet.
1064  * Returns 1 if the router SHOULD NOT send this icmp packet due to rate
1065  * limitation.
1066  *
1067  * XXX per-destination/type check necessary?
1068  */
1069 int
1070 icmp_ratelimit(const struct in_addr *dst, const int type, const int code)
1071 {
1072 	/* PPS limit */
1073 	if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count,
1074 	    icmperrppslim))
1075 		return 1;	/* The packet is subject to rate limit */
1076 	return 0;	/* okay to send */
1077 }
1078 
1079 void
1080 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r)
1081 {
1082 	if (rt == NULL)
1083 		panic("icmp_redirect_timeout:  bad route to timeout");
1084 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1085 	    (RTF_DYNAMIC | RTF_HOST)) {
1086 		struct rt_addrinfo info;
1087 
1088 		memset(&info, 0, sizeof(info));
1089 		info.rti_info[RTAX_DST] = rt_key(rt);
1090 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1091 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1092 		info.rti_flags = rt->rt_flags;
1093 
1094 		rtrequest1(RTM_DELETE, &info, rt->rt_priority, NULL,
1095 		    r->rtt_tableid);
1096 	}
1097 }
1098 
1099 int
1100 icmp_do_exthdr(struct mbuf *m, u_int16_t class, u_int8_t ctype, void *buf,
1101     size_t len)
1102 {
1103 	struct ip *ip = mtod(m, struct ip *);
1104 	int hlen, off;
1105 	struct mbuf *n;
1106 	struct icmp *icp;
1107 	struct icmp_ext_hdr *ieh;
1108 	struct {
1109 		struct icmp_ext_hdr	ieh;
1110 		struct icmp_ext_obj_hdr	ieo;
1111 	} hdr;
1112 
1113 	hlen = ip->ip_hl << 2;
1114 	icp = (struct icmp *)(mtod(m, caddr_t) + hlen);
1115 	if (icp->icmp_type != ICMP_TIMXCEED && icp->icmp_type != ICMP_UNREACH &&
1116 	    icp->icmp_type != ICMP_PARAMPROB)
1117 		/* exthdr not supported */
1118 		return (0);
1119 
1120 	if (icp->icmp_length != 0)
1121 		/* exthdr already present, giving up */
1122 		return (0);
1123 
1124 	/* the actual offset starts after the common ICMP header */
1125 	hlen += ICMP_MINLEN;
1126 	/* exthdr must start on a word boundary */
1127 	off = roundup(ntohs(ip->ip_len) - hlen, sizeof(u_int32_t));
1128 	/* ... and at an offset of ICMP_EXT_OFFSET or bigger */
1129 	off = max(off, ICMP_EXT_OFFSET);
1130 	icp->icmp_length = off / sizeof(u_int32_t);
1131 
1132 	memset(&hdr, 0, sizeof(hdr));
1133 	hdr.ieh.ieh_version = ICMP_EXT_HDR_VERSION;
1134 	hdr.ieo.ieo_length = htons(sizeof(struct icmp_ext_obj_hdr) + len);
1135 	hdr.ieo.ieo_cnum = class;
1136 	hdr.ieo.ieo_ctype = ctype;
1137 
1138 	if (m_copyback(m, hlen + off, sizeof(hdr), &hdr, M_NOWAIT) ||
1139 	    m_copyback(m, hlen + off + sizeof(hdr), len, buf, M_NOWAIT)) {
1140 		m_freem(m);
1141 		return (ENOBUFS);
1142 	}
1143 
1144 	/* calculate checksum */
1145 	n = m_getptr(m, hlen + off, &off);
1146 	if (n == NULL)
1147 		panic("icmp_do_exthdr: m_getptr failure");
1148 	ieh = (struct icmp_ext_hdr *)(mtod(n, caddr_t) + off);
1149 	ieh->ieh_cksum = in4_cksum(n, 0, off, sizeof(hdr) + len);
1150 
1151 	ip->ip_len = htons(m->m_pkthdr.len);
1152 
1153 	return (0);
1154 }
1155