xref: /freebsd/sys/netinet/ip_icmp.c (revision d2387d42b8da231a5b95cbc313825fb2aadf26f6)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
34  * $FreeBSD$
35  */
36 
37 #include "opt_ipsec.h"
38 #include "opt_mac.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/mac.h>
43 #include <sys/mbuf.h>
44 #include <sys/protosw.h>
45 #include <sys/socket.h>
46 #include <sys/time.h>
47 #include <sys/kernel.h>
48 #include <sys/sysctl.h>
49 
50 #include <net/if.h>
51 #include <net/if_types.h>
52 #include <net/route.h>
53 
54 #include <netinet/in.h>
55 #include <netinet/in_pcb.h>
56 #include <netinet/in_systm.h>
57 #include <netinet/in_var.h>
58 #include <netinet/ip.h>
59 #include <netinet/ip_icmp.h>
60 #include <netinet/ip_var.h>
61 #include <netinet/tcp.h>
62 #include <netinet/tcp_var.h>
63 #include <netinet/tcpip.h>
64 #include <netinet/icmp_var.h>
65 
66 #ifdef IPSEC
67 #include <netinet6/ipsec.h>
68 #include <netkey/key.h>
69 #endif
70 
71 #ifdef FAST_IPSEC
72 #include <netipsec/ipsec.h>
73 #include <netipsec/key.h>
74 #define	IPSEC
75 #endif
76 
77 #include <machine/in_cksum.h>
78 
79 /*
80  * ICMP routines: error generation, receive packet processing, and
81  * routines to turnaround packets back to the originator, and
82  * host table maintenance routines.
83  */
84 
85 struct	icmpstat icmpstat;
86 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
87 	&icmpstat, icmpstat, "");
88 
89 static int	icmpmaskrepl = 0;
90 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
91 	&icmpmaskrepl, 0, "Reply to ICMP Address Mask Request packets.");
92 
93 static u_int	icmpmaskfake = 0;
94 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
95 	&icmpmaskfake, 0, "Fake reply to ICMP Address Mask Request packets.");
96 
97 static int	drop_redirect = 0;
98 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
99 	&drop_redirect, 0, "");
100 
101 static int	log_redirect = 0;
102 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
103 	&log_redirect, 0, "");
104 
105 static int      icmplim = 200;
106 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
107 	&icmplim, 0, "");
108 
109 static int	icmplim_output = 1;
110 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
111 	&icmplim_output, 0, "");
112 
113 static char	reply_src[IFNAMSIZ+1];
114 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
115 	&reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
116 
117 /*
118  * ICMP broadcast echo sysctl
119  */
120 
121 static int	icmpbmcastecho = 0;
122 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
123 	&icmpbmcastecho, 0, "");
124 
125 
126 #ifdef ICMPPRINTFS
127 int	icmpprintfs = 0;
128 #endif
129 
130 static void	icmp_reflect(struct mbuf *);
131 static void	icmp_send(struct mbuf *, struct mbuf *);
132 static int	ip_next_mtu(int, int);
133 
134 extern	struct protosw inetsw[];
135 
136 /*
137  * Generate an error packet of type error
138  * in response to bad packet ip.
139  */
140 void
141 icmp_error(n, type, code, dest, destifp)
142 	struct mbuf *n;
143 	int type, code;
144 	n_long dest;
145 	struct ifnet *destifp;
146 {
147 	register struct ip *oip = mtod(n, struct ip *), *nip;
148 	register unsigned oiplen = oip->ip_hl << 2;
149 	register struct icmp *icp;
150 	register struct mbuf *m;
151 	register struct m_tag *mtag;
152 	unsigned icmplen;
153 
154 #ifdef ICMPPRINTFS
155 	if (icmpprintfs)
156 		printf("icmp_error(%p, %x, %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 &~ (IP_MF|IP_DF))
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 *)((caddr_t)oip + oiplen))->icmp_type)) {
170 		icmpstat.icps_oldicmp++;
171 		goto freeit;
172 	}
173 	/* Don't send error in response to a multicast or broadcast packet */
174 	if (n->m_flags & (M_BCAST|M_MCAST))
175 		goto freeit;
176 	/*
177 	 * First, formulate icmp message
178 	 */
179 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
180 	if (m == NULL)
181 		goto freeit;
182 #ifdef MAC
183 	mac_create_mbuf_netlayer(n, m);
184 #endif
185 	icmplen = min(oiplen + 8, oip->ip_len);
186 	if (icmplen < sizeof(struct ip))
187 		panic("icmp_error: bad length");
188 	m->m_len = icmplen + ICMP_MINLEN;
189 	MH_ALIGN(m, m->m_len);
190 	icp = mtod(m, struct icmp *);
191 	if ((u_int)type > ICMP_MAXTYPE)
192 		panic("icmp_error");
193 	icmpstat.icps_outhist[type]++;
194 	icp->icmp_type = type;
195 	if (type == ICMP_REDIRECT)
196 		icp->icmp_gwaddr.s_addr = dest;
197 	else {
198 		icp->icmp_void = 0;
199 		/*
200 		 * The following assignments assume an overlay with the
201 		 * zeroed icmp_void field.
202 		 */
203 		if (type == ICMP_PARAMPROB) {
204 			icp->icmp_pptr = code;
205 			code = 0;
206 		} else if (type == ICMP_UNREACH &&
207 			code == ICMP_UNREACH_NEEDFRAG && destifp) {
208 			icp->icmp_nextmtu = htons(destifp->if_mtu);
209 		}
210 	}
211 
212 	icp->icmp_code = code;
213 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
214 	nip = &icp->icmp_ip;
215 
216 	/*
217 	 * Convert fields to network representation.
218 	 */
219 	nip->ip_len = htons(nip->ip_len);
220 	nip->ip_off = htons(nip->ip_off);
221 
222 	/*
223 	 * Now, copy old ip header (without options)
224 	 * in front of icmp message.
225 	 */
226 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
227 		panic("icmp len");
228 	m->m_data -= sizeof(struct ip);
229 	m->m_len += sizeof(struct ip);
230 	m->m_pkthdr.len = m->m_len;
231 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
232 	nip = mtod(m, struct ip *);
233 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
234 	nip->ip_len = m->m_len;
235 	nip->ip_v = IPVERSION;
236 	nip->ip_hl = 5;
237 	nip->ip_p = IPPROTO_ICMP;
238 	nip->ip_tos = 0;
239 	/*
240 	 * XXX: Move PF_GENERATED m_tag to new packet, if it exists.
241 	 *	This should be replaced by unified flags/tags for
242 	 *	pf/ipfw/ipf and future pfil_hook applications.
243 	 */
244 	mtag = m_tag_find(n, PACKET_TAG_PF_GENERATED, NULL);
245 	if (mtag != NULL) {
246 		m_tag_unlink(n, mtag);
247 		m_tag_prepend(m, mtag);
248 	}
249 	icmp_reflect(m);
250 
251 freeit:
252 	m_freem(n);
253 }
254 
255 /*
256  * Process a received ICMP message.
257  */
258 void
259 icmp_input(m, off)
260 	struct mbuf *m;
261 	int off;
262 {
263 	struct icmp *icp;
264 	struct in_ifaddr *ia;
265 	struct ip *ip = mtod(m, struct ip *);
266 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
267 	int hlen = off;
268 	int icmplen = ip->ip_len;
269 	int i, code;
270 	void (*ctlfunc)(int, struct sockaddr *, void *);
271 
272 	/*
273 	 * Locate icmp structure in mbuf, and check
274 	 * that not corrupted and of at least minimum length.
275 	 */
276 #ifdef ICMPPRINTFS
277 	if (icmpprintfs) {
278 		char buf[4 * sizeof "123"];
279 		strcpy(buf, inet_ntoa(ip->ip_src));
280 		printf("icmp_input from %s to %s, len %d\n",
281 		       buf, inet_ntoa(ip->ip_dst), icmplen);
282 	}
283 #endif
284 	if (icmplen < ICMP_MINLEN) {
285 		icmpstat.icps_tooshort++;
286 		goto freeit;
287 	}
288 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
289 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
290 		icmpstat.icps_tooshort++;
291 		return;
292 	}
293 	ip = mtod(m, struct ip *);
294 	m->m_len -= hlen;
295 	m->m_data += hlen;
296 	icp = mtod(m, struct icmp *);
297 	if (in_cksum(m, icmplen)) {
298 		icmpstat.icps_checksum++;
299 		goto freeit;
300 	}
301 	m->m_len += hlen;
302 	m->m_data -= hlen;
303 
304 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
305 		/*
306 		 * Deliver very specific ICMP type only.
307 		 */
308 		switch (icp->icmp_type) {
309 		case ICMP_UNREACH:
310 		case ICMP_TIMXCEED:
311 			break;
312 		default:
313 			goto freeit;
314 		}
315 	}
316 
317 #ifdef ICMPPRINTFS
318 	if (icmpprintfs)
319 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
320 		    icp->icmp_code);
321 #endif
322 
323 	/*
324 	 * Message type specific processing.
325 	 */
326 	if (icp->icmp_type > ICMP_MAXTYPE)
327 		goto raw;
328 
329 	/* Initialize */
330 	bzero(&icmpsrc, sizeof(icmpsrc));
331 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
332 	icmpsrc.sin_family = AF_INET;
333 	bzero(&icmpdst, sizeof(icmpdst));
334 	icmpdst.sin_len = sizeof(struct sockaddr_in);
335 	icmpdst.sin_family = AF_INET;
336 	bzero(&icmpgw, sizeof(icmpgw));
337 	icmpgw.sin_len = sizeof(struct sockaddr_in);
338 	icmpgw.sin_family = AF_INET;
339 
340 	icmpstat.icps_inhist[icp->icmp_type]++;
341 	code = icp->icmp_code;
342 	switch (icp->icmp_type) {
343 
344 	case ICMP_UNREACH:
345 		switch (code) {
346 			case ICMP_UNREACH_NET:
347 			case ICMP_UNREACH_HOST:
348 			case ICMP_UNREACH_SRCFAIL:
349 			case ICMP_UNREACH_NET_UNKNOWN:
350 			case ICMP_UNREACH_HOST_UNKNOWN:
351 			case ICMP_UNREACH_ISOLATED:
352 			case ICMP_UNREACH_TOSNET:
353 			case ICMP_UNREACH_TOSHOST:
354 			case ICMP_UNREACH_HOST_PRECEDENCE:
355 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
356 				code = PRC_UNREACH_NET;
357 				break;
358 
359 			case ICMP_UNREACH_NEEDFRAG:
360 				code = PRC_MSGSIZE;
361 				break;
362 
363 			/*
364 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
365 			 * Treat subcodes 2,3 as immediate RST
366 			 */
367 			case ICMP_UNREACH_PROTOCOL:
368 			case ICMP_UNREACH_PORT:
369 				code = PRC_UNREACH_PORT;
370 				break;
371 
372 			case ICMP_UNREACH_NET_PROHIB:
373 			case ICMP_UNREACH_HOST_PROHIB:
374 			case ICMP_UNREACH_FILTER_PROHIB:
375 				code = PRC_UNREACH_ADMIN_PROHIB;
376 				break;
377 
378 			default:
379 				goto badcode;
380 		}
381 		goto deliver;
382 
383 	case ICMP_TIMXCEED:
384 		if (code > 1)
385 			goto badcode;
386 		code += PRC_TIMXCEED_INTRANS;
387 		goto deliver;
388 
389 	case ICMP_PARAMPROB:
390 		if (code > 1)
391 			goto badcode;
392 		code = PRC_PARAMPROB;
393 		goto deliver;
394 
395 	case ICMP_SOURCEQUENCH:
396 		if (code)
397 			goto badcode;
398 		code = PRC_QUENCH;
399 	deliver:
400 		/*
401 		 * Problem with datagram; advise higher level routines.
402 		 */
403 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
404 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
405 			icmpstat.icps_badlen++;
406 			goto freeit;
407 		}
408 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
409 		/* Discard ICMP's in response to multicast packets */
410 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
411 			goto badcode;
412 #ifdef ICMPPRINTFS
413 		if (icmpprintfs)
414 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
415 #endif
416 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
417 
418 		/*
419 		 * MTU discovery:
420 		 * If we got a needfrag and there is a host route to the
421 		 * original destination, and the MTU is not locked, then
422 		 * set the MTU in the route to the suggested new value
423 		 * (if given) and then notify as usual.  The ULPs will
424 		 * notice that the MTU has changed and adapt accordingly.
425 		 * If no new MTU was suggested, then we guess a new one
426 		 * less than the current value.  If the new MTU is
427 		 * unreasonably small (defined by sysctl tcp_minmss), then
428 		 * we don't update the MTU value.
429 		 *
430 		 * XXX: All this should be done in tcp_mtudisc() because
431 		 * the way we do it now, everyone can send us bogus ICMP
432 		 * MSGSIZE packets for any destination. By doing this far
433 		 * higher in the chain we have a matching tcp connection.
434 		 * Thus spoofing is much harder. However there is no easy
435 		 * non-hackish way to pass the new MTU up to tcp_mtudisc().
436 		 * Also see next XXX regarding IPv4 AH TCP.
437 		 */
438 		if (code == PRC_MSGSIZE) {
439 			int mtu;
440 			struct in_conninfo inc;
441 
442 			bzero(&inc, sizeof(inc));
443 			inc.inc_flags = 0; /* IPv4 */
444 			inc.inc_faddr = icmpsrc.sin_addr;
445 
446 			mtu = ntohs(icp->icmp_nextmtu);
447 			if (!mtu)
448 				mtu = ip_next_mtu(mtu, 1);
449 
450 			if (mtu >= max(296, (tcp_minmss +
451 					sizeof(struct tcpiphdr))))
452 				tcp_hc_updatemtu(&inc, mtu);
453 
454 #ifdef DEBUG_MTUDISC
455 			printf("MTU for %s reduced to %d\n",
456 				inet_ntoa(icmpsrc.sin_addr), mtu);
457 #endif
458 		}
459 
460 		/*
461 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
462 		 * notification to TCP layer.
463 		 */
464 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
465 		if (ctlfunc)
466 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
467 				   (void *)&icp->icmp_ip);
468 		break;
469 
470 	badcode:
471 		icmpstat.icps_badcode++;
472 		break;
473 
474 	case ICMP_ECHO:
475 		if (!icmpbmcastecho
476 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
477 			icmpstat.icps_bmcastecho++;
478 			break;
479 		}
480 		icp->icmp_type = ICMP_ECHOREPLY;
481 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
482 			goto freeit;
483 		else
484 			goto reflect;
485 
486 	case ICMP_TSTAMP:
487 		if (!icmpbmcastecho
488 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
489 			icmpstat.icps_bmcasttstamp++;
490 			break;
491 		}
492 		if (icmplen < ICMP_TSLEN) {
493 			icmpstat.icps_badlen++;
494 			break;
495 		}
496 		icp->icmp_type = ICMP_TSTAMPREPLY;
497 		icp->icmp_rtime = iptime();
498 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
499 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
500 			goto freeit;
501 		else
502 			goto reflect;
503 
504 	case ICMP_MASKREQ:
505 		if (icmpmaskrepl == 0)
506 			break;
507 		/*
508 		 * We are not able to respond with all ones broadcast
509 		 * unless we receive it over a point-to-point interface.
510 		 */
511 		if (icmplen < ICMP_MASKLEN)
512 			break;
513 		switch (ip->ip_dst.s_addr) {
514 
515 		case INADDR_BROADCAST:
516 		case INADDR_ANY:
517 			icmpdst.sin_addr = ip->ip_src;
518 			break;
519 
520 		default:
521 			icmpdst.sin_addr = ip->ip_dst;
522 		}
523 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
524 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
525 		if (ia == 0)
526 			break;
527 		if (ia->ia_ifp == 0)
528 			break;
529 		icp->icmp_type = ICMP_MASKREPLY;
530 		if (icmpmaskfake == 0)
531 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
532 		else
533 			icp->icmp_mask = icmpmaskfake;
534 		if (ip->ip_src.s_addr == 0) {
535 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
536 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
537 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
538 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
539 		}
540 reflect:
541 		ip->ip_len += hlen;	/* since ip_input deducts this */
542 		icmpstat.icps_reflect++;
543 		icmpstat.icps_outhist[icp->icmp_type]++;
544 		icmp_reflect(m);
545 		return;
546 
547 	case ICMP_REDIRECT:
548 		if (log_redirect) {
549 			u_long src, dst, gw;
550 
551 			src = ntohl(ip->ip_src.s_addr);
552 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
553 			gw = ntohl(icp->icmp_gwaddr.s_addr);
554 			printf("icmp redirect from %d.%d.%d.%d: "
555 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
556 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
557 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
558 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
559 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
560 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
561 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
562 		}
563 		/*
564 		 * RFC1812 says we must ignore ICMP redirects if we
565 		 * are acting as router.
566 		 */
567 		if (drop_redirect || ipforwarding)
568 			break;
569 		if (code > 3)
570 			goto badcode;
571 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
572 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
573 			icmpstat.icps_badlen++;
574 			break;
575 		}
576 		/*
577 		 * Short circuit routing redirects to force
578 		 * immediate change in the kernel's routing
579 		 * tables.  The message is also handed to anyone
580 		 * listening on a raw socket (e.g. the routing
581 		 * daemon for use in updating its tables).
582 		 */
583 		icmpgw.sin_addr = ip->ip_src;
584 		icmpdst.sin_addr = icp->icmp_gwaddr;
585 #ifdef	ICMPPRINTFS
586 		if (icmpprintfs) {
587 			char buf[4 * sizeof "123"];
588 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
589 
590 			printf("redirect dst %s to %s\n",
591 			       buf, inet_ntoa(icp->icmp_gwaddr));
592 		}
593 #endif
594 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
595 		rtredirect((struct sockaddr *)&icmpsrc,
596 		  (struct sockaddr *)&icmpdst,
597 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
598 		  (struct sockaddr *)&icmpgw);
599 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
600 #ifdef IPSEC
601 		key_sa_routechange((struct sockaddr *)&icmpsrc);
602 #endif
603 		break;
604 
605 	/*
606 	 * No kernel processing for the following;
607 	 * just fall through to send to raw listener.
608 	 */
609 	case ICMP_ECHOREPLY:
610 	case ICMP_ROUTERADVERT:
611 	case ICMP_ROUTERSOLICIT:
612 	case ICMP_TSTAMPREPLY:
613 	case ICMP_IREQREPLY:
614 	case ICMP_MASKREPLY:
615 	default:
616 		break;
617 	}
618 
619 raw:
620 	rip_input(m, off);
621 	return;
622 
623 freeit:
624 	m_freem(m);
625 }
626 
627 /*
628  * Reflect the ip packet back to the source
629  */
630 static void
631 icmp_reflect(m)
632 	struct mbuf *m;
633 {
634 	struct ip *ip = mtod(m, struct ip *);
635 	struct ifaddr *ifa;
636 	struct ifnet *ifn;
637 	struct in_ifaddr *ia;
638 	struct in_addr t;
639 	struct mbuf *opts = 0;
640 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
641 
642 	if (!in_canforward(ip->ip_src) &&
643 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
644 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
645 		m_freem(m);	/* Bad return address */
646 		icmpstat.icps_badaddr++;
647 		goto done;	/* Ip_output() will check for broadcast */
648 	}
649 	t = ip->ip_dst;
650 	ip->ip_dst = ip->ip_src;
651 
652 	/*
653 	 * Source selection for ICMP replies:
654 	 *
655 	 * If the incoming packet was addressed directly to one of our
656 	 * own addresses, use dst as the src for the reply.
657 	 */
658 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
659 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
660 			goto match;
661 	/*
662 	 * If the incoming packet was addressed to one of our broadcast
663 	 * addresses, use the first non-broadcast address which corresponds
664 	 * to the incoming interface.
665 	 */
666 	if (m->m_pkthdr.rcvif != NULL &&
667 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
668 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
669 			if (ifa->ifa_addr->sa_family != AF_INET)
670 				continue;
671 			ia = ifatoia(ifa);
672 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
673 			    t.s_addr)
674 				goto match;
675 		}
676 	}
677 	/*
678 	 * If the incoming packet was not addressed directly to us, use
679 	 * designated interface for icmp replies specified by sysctl
680 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
681 	 * with normal source selection.
682 	 */
683 	if (reply_src[0] != '\0' && (ifn = ifunit(reply_src))) {
684 		TAILQ_FOREACH(ifa, &ifn->if_addrhead, ifa_link) {
685 			if (ifa->ifa_addr->sa_family != AF_INET)
686 				continue;
687 			ia = ifatoia(ifa);
688 			goto match;
689 		}
690 	}
691 	/*
692 	 * If the packet was transiting through us, use the address of
693 	 * the interface that is the closest to the packet source.
694 	 * When we don't have a route back to the packet source, stop here
695 	 * and drop the packet.
696 	 */
697 	ia = ip_rtaddr(ip->ip_dst);
698 	if (ia == NULL) {
699 		m_freem(m);
700 		icmpstat.icps_noroute++;
701 		goto done;
702 	}
703 match:
704 #ifdef MAC
705 	mac_reflect_mbuf_icmp(m);
706 #endif
707 	t = IA_SIN(ia)->sin_addr;
708 	ip->ip_src = t;
709 	ip->ip_ttl = ip_defttl;
710 
711 	if (optlen > 0) {
712 		register u_char *cp;
713 		int opt, cnt;
714 		u_int len;
715 
716 		/*
717 		 * Retrieve any source routing from the incoming packet;
718 		 * add on any record-route or timestamp options.
719 		 */
720 		cp = (u_char *) (ip + 1);
721 		if ((opts = ip_srcroute()) == 0 &&
722 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
723 			opts->m_len = sizeof(struct in_addr);
724 			mtod(opts, struct in_addr *)->s_addr = 0;
725 		}
726 		if (opts) {
727 #ifdef ICMPPRINTFS
728 		    if (icmpprintfs)
729 			    printf("icmp_reflect optlen %d rt %d => ",
730 				optlen, opts->m_len);
731 #endif
732 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
733 			    opt = cp[IPOPT_OPTVAL];
734 			    if (opt == IPOPT_EOL)
735 				    break;
736 			    if (opt == IPOPT_NOP)
737 				    len = 1;
738 			    else {
739 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
740 					    break;
741 				    len = cp[IPOPT_OLEN];
742 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
743 				        len > cnt)
744 					    break;
745 			    }
746 			    /*
747 			     * Should check for overflow, but it "can't happen"
748 			     */
749 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
750 				opt == IPOPT_SECURITY) {
751 				    bcopy((caddr_t)cp,
752 					mtod(opts, caddr_t) + opts->m_len, len);
753 				    opts->m_len += len;
754 			    }
755 		    }
756 		    /* Terminate & pad, if necessary */
757 		    cnt = opts->m_len % 4;
758 		    if (cnt) {
759 			    for (; cnt < 4; cnt++) {
760 				    *(mtod(opts, caddr_t) + opts->m_len) =
761 					IPOPT_EOL;
762 				    opts->m_len++;
763 			    }
764 		    }
765 #ifdef ICMPPRINTFS
766 		    if (icmpprintfs)
767 			    printf("%d\n", opts->m_len);
768 #endif
769 		}
770 		/*
771 		 * Now strip out original options by copying rest of first
772 		 * mbuf's data back, and adjust the IP length.
773 		 */
774 		ip->ip_len -= optlen;
775 		ip->ip_v = IPVERSION;
776 		ip->ip_hl = 5;
777 		m->m_len -= optlen;
778 		if (m->m_flags & M_PKTHDR)
779 			m->m_pkthdr.len -= optlen;
780 		optlen += sizeof(struct ip);
781 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
782 			 (unsigned)(m->m_len - sizeof(struct ip)));
783 	}
784 	m_tag_delete_nonpersistent(m);
785 	m->m_flags &= ~(M_BCAST|M_MCAST);
786 	icmp_send(m, opts);
787 done:
788 	if (opts)
789 		(void)m_free(opts);
790 }
791 
792 /*
793  * Send an icmp packet back to the ip level,
794  * after supplying a checksum.
795  */
796 static void
797 icmp_send(m, opts)
798 	register struct mbuf *m;
799 	struct mbuf *opts;
800 {
801 	register struct ip *ip = mtod(m, struct ip *);
802 	register int hlen;
803 	register struct icmp *icp;
804 
805 	hlen = ip->ip_hl << 2;
806 	m->m_data += hlen;
807 	m->m_len -= hlen;
808 	icp = mtod(m, struct icmp *);
809 	icp->icmp_cksum = 0;
810 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
811 	m->m_data -= hlen;
812 	m->m_len += hlen;
813 	m->m_pkthdr.rcvif = (struct ifnet *)0;
814 #ifdef ICMPPRINTFS
815 	if (icmpprintfs) {
816 		char buf[4 * sizeof "123"];
817 		strcpy(buf, inet_ntoa(ip->ip_dst));
818 		printf("icmp_send dst %s src %s\n",
819 		       buf, inet_ntoa(ip->ip_src));
820 	}
821 #endif
822 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
823 }
824 
825 n_time
826 iptime()
827 {
828 	struct timeval atv;
829 	u_long t;
830 
831 	getmicrotime(&atv);
832 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
833 	return (htonl(t));
834 }
835 
836 /*
837  * Return the next larger or smaller MTU plateau (table from RFC 1191)
838  * given current value MTU.  If DIR is less than zero, a larger plateau
839  * is returned; otherwise, a smaller value is returned.
840  */
841 static int
842 ip_next_mtu(mtu, dir)
843 	int mtu;
844 	int dir;
845 {
846 	static int mtutab[] = {
847 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
848 		68, 0
849 	};
850 	int i;
851 
852 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
853 		if (mtu >= mtutab[i])
854 			break;
855 	}
856 
857 	if (dir < 0) {
858 		if (i == 0) {
859 			return 0;
860 		} else {
861 			return mtutab[i - 1];
862 		}
863 	} else {
864 		if (mtutab[i] == 0) {
865 			return 0;
866 		} else if(mtu > mtutab[i]) {
867 			return mtutab[i];
868 		} else {
869 			return mtutab[i + 1];
870 		}
871 	}
872 }
873 
874 
875 /*
876  * badport_bandlim() - check for ICMP bandwidth limit
877  *
878  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
879  *	hit our bandwidth limit and it is not ok.
880  *
881  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
882  *
883  *	For now we separate the TCP and UDP subsystems w/ different 'which'
884  *	values.  We may eventually remove this separation (and simplify the
885  *	code further).
886  *
887  *	Note that the printing of the error message is delayed so we can
888  *	properly print the icmp error rate that the system was trying to do
889  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
890  *	the 'final' error, but it doesn't make sense to solve the printing
891  *	delay with more complex code.
892  */
893 
894 int
895 badport_bandlim(int which)
896 {
897 #define	N(a)	(sizeof (a) / sizeof (a[0]))
898 	static struct rate {
899 		const char	*type;
900 		struct timeval	lasttime;
901 		int		curpps;;
902 	} rates[BANDLIM_MAX+1] = {
903 		{ "icmp unreach response" },
904 		{ "icmp ping response" },
905 		{ "icmp tstamp response" },
906 		{ "closed port RST response" },
907 		{ "open port RST response" }
908 	};
909 
910 	/*
911 	 * Return ok status if feature disabled or argument out of range.
912 	 */
913 	if (icmplim > 0 && (u_int) which < N(rates)) {
914 		struct rate *r = &rates[which];
915 		int opps = r->curpps;
916 
917 		if (!ppsratecheck(&r->lasttime, &r->curpps, icmplim))
918 			return -1;	/* discard packet */
919 		/*
920 		 * If we've dropped below the threshold after having
921 		 * rate-limited traffic print the message.  This preserves
922 		 * the previous behaviour at the expense of added complexity.
923 		 */
924 		if (icmplim_output && opps > icmplim)
925 			printf("Limiting %s from %d to %d packets/sec\n",
926 				r->type, opps, icmplim);
927 	}
928 	return 0;			/* okay to send packet */
929 #undef N
930 }
931