xref: /freebsd/sys/netinet/ip_icmp.c (revision 822923447e454b30d310cb46903c9ddeca9f0a7a)
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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
30  * $FreeBSD$
31  */
32 
33 #include "opt_ipsec.h"
34 #include "opt_mac.h"
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/mac.h>
39 #include <sys/mbuf.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/sysctl.h>
45 
46 #include <net/if.h>
47 #include <net/if_types.h>
48 #include <net/route.h>
49 
50 #include <netinet/in.h>
51 #include <netinet/in_pcb.h>
52 #include <netinet/in_systm.h>
53 #include <netinet/in_var.h>
54 #include <netinet/ip.h>
55 #include <netinet/ip_icmp.h>
56 #include <netinet/ip_var.h>
57 #include <netinet/tcp.h>
58 #include <netinet/tcp_var.h>
59 #include <netinet/tcpip.h>
60 #include <netinet/icmp_var.h>
61 
62 #ifdef IPSEC
63 #include <netinet6/ipsec.h>
64 #include <netkey/key.h>
65 #endif
66 
67 #ifdef FAST_IPSEC
68 #include <netipsec/ipsec.h>
69 #include <netipsec/key.h>
70 #define	IPSEC
71 #endif
72 
73 #include <machine/in_cksum.h>
74 
75 /*
76  * ICMP routines: error generation, receive packet processing, and
77  * routines to turnaround packets back to the originator, and
78  * host table maintenance routines.
79  */
80 
81 struct	icmpstat icmpstat;
82 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
83 	&icmpstat, icmpstat, "");
84 
85 static int	icmpmaskrepl = 0;
86 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
87 	&icmpmaskrepl, 0, "Reply to ICMP Address Mask Request packets.");
88 
89 static u_int	icmpmaskfake = 0;
90 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
91 	&icmpmaskfake, 0, "Fake reply to ICMP Address Mask Request packets.");
92 
93 static int	drop_redirect = 0;
94 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
95 	&drop_redirect, 0, "");
96 
97 static int	log_redirect = 0;
98 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
99 	&log_redirect, 0, "");
100 
101 static int      icmplim = 200;
102 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
103 	&icmplim, 0, "");
104 
105 static int	icmplim_output = 1;
106 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
107 	&icmplim_output, 0, "");
108 
109 static char	reply_src[IFNAMSIZ];
110 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
111 	&reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
112 
113 static int	icmp_rfi = 0;
114 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
115 	&icmp_rfi, 0, "ICMP reply from incoming interface for "
116 	"non-local packets");
117 
118 static int	icmp_quotelen = 8;
119 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
120 	&icmp_quotelen, 0, "Number of bytes from original packet to "
121 	"quote in ICMP reply");
122 
123 /*
124  * ICMP broadcast echo sysctl
125  */
126 
127 static int	icmpbmcastecho = 0;
128 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
129 	&icmpbmcastecho, 0, "");
130 
131 
132 #ifdef ICMPPRINTFS
133 int	icmpprintfs = 0;
134 #endif
135 
136 static void	icmp_reflect(struct mbuf *);
137 static void	icmp_send(struct mbuf *, struct mbuf *);
138 
139 extern	struct protosw inetsw[];
140 
141 /*
142  * Generate an error packet of type error
143  * in response to bad packet ip.
144  */
145 void
146 icmp_error(n, type, code, dest, mtu)
147 	struct mbuf *n;
148 	int type, code;
149 	n_long dest;
150 	int mtu;
151 {
152 	register struct ip *oip = mtod(n, struct ip *), *nip;
153 	register unsigned oiplen = oip->ip_hl << 2;
154 	register struct icmp *icp;
155 	register struct mbuf *m;
156 	unsigned icmplen;
157 
158 #ifdef ICMPPRINTFS
159 	if (icmpprintfs)
160 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
161 #endif
162 	if (type != ICMP_REDIRECT)
163 		icmpstat.icps_error++;
164 	/*
165 	 * Don't send error if the original packet was encrypted.
166 	 * Don't send error if not the first fragment of message.
167 	 * Don't error if the old packet protocol was ICMP
168 	 * error message, only known informational types.
169 	 */
170 	if (n->m_flags & M_DECRYPTED)
171 		goto freeit;
172 	if (oip->ip_off &~ (IP_MF|IP_DF))
173 		goto freeit;
174 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
175 	  n->m_len >= oiplen + ICMP_MINLEN &&
176 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
177 		icmpstat.icps_oldicmp++;
178 		goto freeit;
179 	}
180 	/* Don't send error in response to a multicast or broadcast packet */
181 	if (n->m_flags & (M_BCAST|M_MCAST))
182 		goto freeit;
183 	/*
184 	 * First, formulate icmp message
185 	 */
186 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
187 	if (m == NULL)
188 		goto freeit;
189 #ifdef MAC
190 	mac_create_mbuf_netlayer(n, m);
191 #endif
192 	/*
193 	 * Calculate length to quote from original packet and
194 	 * prevent the ICMP mbuf from overflowing.
195 	 */
196 	if (oip->ip_p == IPPROTO_TCP) {
197 		struct tcphdr *th;
198 		int tcphlen;
199 
200 		if (n->m_len < oiplen + sizeof(struct tcphdr) &&
201 		    ((n = m_pullup(n, oiplen + sizeof(struct tcphdr))) == NULL))
202 			goto freeit;
203 		th = (struct tcphdr *)((caddr_t)oip + oiplen);
204 		tcphlen = th->th_off << 2;
205 		if (tcphlen < sizeof(struct tcphdr))
206 			goto freeit;
207 		if (oip->ip_len < oiplen + tcphlen)
208 			goto freeit;
209 		if (n->m_len < oiplen + tcphlen &&
210 		    ((n = m_pullup(n, oiplen + tcphlen)) == NULL))
211 			goto freeit;
212 		icmplen = max(oiplen + tcphlen, min(icmp_quotelen, oip->ip_len));
213 	} else
214 		icmplen = min(oiplen + max(8, icmp_quotelen), oip->ip_len);
215 	if (icmplen < sizeof(struct ip))
216 		goto freeit;
217 	if (icmplen + ICMP_MINLEN + sizeof(struct ip) > MHLEN) {
218 		MCLGET(m, M_DONTWAIT);
219 		if (!(m->m_flags & M_EXT))
220 			goto freeit;
221 	}
222 	m->m_len = icmplen + ICMP_MINLEN;
223 	MH_ALIGN(m, m->m_len);
224 	icp = mtod(m, struct icmp *);
225 	if ((u_int)type > ICMP_MAXTYPE)
226 		panic("icmp_error");
227 	icmpstat.icps_outhist[type]++;
228 	icp->icmp_type = type;
229 	if (type == ICMP_REDIRECT)
230 		icp->icmp_gwaddr.s_addr = dest;
231 	else {
232 		icp->icmp_void = 0;
233 		/*
234 		 * The following assignments assume an overlay with the
235 		 * zeroed icmp_void field.
236 		 */
237 		if (type == ICMP_PARAMPROB) {
238 			icp->icmp_pptr = code;
239 			code = 0;
240 		} else if (type == ICMP_UNREACH &&
241 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
242 			icp->icmp_nextmtu = htons(mtu);
243 		}
244 	}
245 
246 	icp->icmp_code = code;
247 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
248 	nip = &icp->icmp_ip;
249 
250 	/*
251 	 * Convert fields to network representation.
252 	 */
253 	nip->ip_len = htons(nip->ip_len);
254 	nip->ip_off = htons(nip->ip_off);
255 
256 	/*
257 	 * Now, copy old ip header (without options)
258 	 * in front of icmp message.
259 	 */
260 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
261 		panic("icmp len");
262 	/*
263 	 * If the original mbuf was meant to bypass the firewall, the error
264 	 * reply should bypass as well.
265 	 */
266 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
267 	m->m_data -= sizeof(struct ip);
268 	m->m_len += sizeof(struct ip);
269 	m->m_pkthdr.len = m->m_len;
270 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
271 	nip = mtod(m, struct ip *);
272 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
273 	nip->ip_len = m->m_len;
274 	nip->ip_v = IPVERSION;
275 	nip->ip_hl = 5;
276 	nip->ip_p = IPPROTO_ICMP;
277 	nip->ip_tos = 0;
278 	icmp_reflect(m);
279 
280 freeit:
281 	m_freem(n);
282 }
283 
284 /*
285  * Process a received ICMP message.
286  */
287 void
288 icmp_input(m, off)
289 	struct mbuf *m;
290 	int off;
291 {
292 	struct icmp *icp;
293 	struct in_ifaddr *ia;
294 	struct ip *ip = mtod(m, struct ip *);
295 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
296 	int hlen = off;
297 	int icmplen = ip->ip_len;
298 	int i, code;
299 	void (*ctlfunc)(int, struct sockaddr *, void *);
300 
301 	/*
302 	 * Locate icmp structure in mbuf, and check
303 	 * that not corrupted and of at least minimum length.
304 	 */
305 #ifdef ICMPPRINTFS
306 	if (icmpprintfs) {
307 		char buf[4 * sizeof "123"];
308 		strcpy(buf, inet_ntoa(ip->ip_src));
309 		printf("icmp_input from %s to %s, len %d\n",
310 		       buf, inet_ntoa(ip->ip_dst), icmplen);
311 	}
312 #endif
313 	if (icmplen < ICMP_MINLEN) {
314 		icmpstat.icps_tooshort++;
315 		goto freeit;
316 	}
317 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
318 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
319 		icmpstat.icps_tooshort++;
320 		return;
321 	}
322 	ip = mtod(m, struct ip *);
323 	m->m_len -= hlen;
324 	m->m_data += hlen;
325 	icp = mtod(m, struct icmp *);
326 	if (in_cksum(m, icmplen)) {
327 		icmpstat.icps_checksum++;
328 		goto freeit;
329 	}
330 	m->m_len += hlen;
331 	m->m_data -= hlen;
332 
333 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
334 		/*
335 		 * Deliver very specific ICMP type only.
336 		 */
337 		switch (icp->icmp_type) {
338 		case ICMP_UNREACH:
339 		case ICMP_TIMXCEED:
340 			break;
341 		default:
342 			goto freeit;
343 		}
344 	}
345 
346 #ifdef ICMPPRINTFS
347 	if (icmpprintfs)
348 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
349 		    icp->icmp_code);
350 #endif
351 
352 	/*
353 	 * Message type specific processing.
354 	 */
355 	if (icp->icmp_type > ICMP_MAXTYPE)
356 		goto raw;
357 
358 	/* Initialize */
359 	bzero(&icmpsrc, sizeof(icmpsrc));
360 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
361 	icmpsrc.sin_family = AF_INET;
362 	bzero(&icmpdst, sizeof(icmpdst));
363 	icmpdst.sin_len = sizeof(struct sockaddr_in);
364 	icmpdst.sin_family = AF_INET;
365 	bzero(&icmpgw, sizeof(icmpgw));
366 	icmpgw.sin_len = sizeof(struct sockaddr_in);
367 	icmpgw.sin_family = AF_INET;
368 
369 	icmpstat.icps_inhist[icp->icmp_type]++;
370 	code = icp->icmp_code;
371 	switch (icp->icmp_type) {
372 
373 	case ICMP_UNREACH:
374 		switch (code) {
375 			case ICMP_UNREACH_NET:
376 			case ICMP_UNREACH_HOST:
377 			case ICMP_UNREACH_SRCFAIL:
378 			case ICMP_UNREACH_NET_UNKNOWN:
379 			case ICMP_UNREACH_HOST_UNKNOWN:
380 			case ICMP_UNREACH_ISOLATED:
381 			case ICMP_UNREACH_TOSNET:
382 			case ICMP_UNREACH_TOSHOST:
383 			case ICMP_UNREACH_HOST_PRECEDENCE:
384 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
385 				code = PRC_UNREACH_NET;
386 				break;
387 
388 			case ICMP_UNREACH_NEEDFRAG:
389 				code = PRC_MSGSIZE;
390 				break;
391 
392 			/*
393 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
394 			 * Treat subcodes 2,3 as immediate RST
395 			 */
396 			case ICMP_UNREACH_PROTOCOL:
397 			case ICMP_UNREACH_PORT:
398 				code = PRC_UNREACH_PORT;
399 				break;
400 
401 			case ICMP_UNREACH_NET_PROHIB:
402 			case ICMP_UNREACH_HOST_PROHIB:
403 			case ICMP_UNREACH_FILTER_PROHIB:
404 				code = PRC_UNREACH_ADMIN_PROHIB;
405 				break;
406 
407 			default:
408 				goto badcode;
409 		}
410 		goto deliver;
411 
412 	case ICMP_TIMXCEED:
413 		if (code > 1)
414 			goto badcode;
415 		code += PRC_TIMXCEED_INTRANS;
416 		goto deliver;
417 
418 	case ICMP_PARAMPROB:
419 		if (code > 1)
420 			goto badcode;
421 		code = PRC_PARAMPROB;
422 		goto deliver;
423 
424 	case ICMP_SOURCEQUENCH:
425 		if (code)
426 			goto badcode;
427 		code = PRC_QUENCH;
428 	deliver:
429 		/*
430 		 * Problem with datagram; advise higher level routines.
431 		 */
432 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
433 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
434 			icmpstat.icps_badlen++;
435 			goto freeit;
436 		}
437 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
438 		/* Discard ICMP's in response to multicast packets */
439 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
440 			goto badcode;
441 #ifdef ICMPPRINTFS
442 		if (icmpprintfs)
443 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
444 #endif
445 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
446 		/*
447 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
448 		 * notification to TCP layer.
449 		 */
450 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
451 		if (ctlfunc)
452 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
453 				   (void *)&icp->icmp_ip);
454 		break;
455 
456 	badcode:
457 		icmpstat.icps_badcode++;
458 		break;
459 
460 	case ICMP_ECHO:
461 		if (!icmpbmcastecho
462 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
463 			icmpstat.icps_bmcastecho++;
464 			break;
465 		}
466 		icp->icmp_type = ICMP_ECHOREPLY;
467 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
468 			goto freeit;
469 		else
470 			goto reflect;
471 
472 	case ICMP_TSTAMP:
473 		if (!icmpbmcastecho
474 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
475 			icmpstat.icps_bmcasttstamp++;
476 			break;
477 		}
478 		if (icmplen < ICMP_TSLEN) {
479 			icmpstat.icps_badlen++;
480 			break;
481 		}
482 		icp->icmp_type = ICMP_TSTAMPREPLY;
483 		icp->icmp_rtime = iptime();
484 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
485 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
486 			goto freeit;
487 		else
488 			goto reflect;
489 
490 	case ICMP_MASKREQ:
491 		if (icmpmaskrepl == 0)
492 			break;
493 		/*
494 		 * We are not able to respond with all ones broadcast
495 		 * unless we receive it over a point-to-point interface.
496 		 */
497 		if (icmplen < ICMP_MASKLEN)
498 			break;
499 		switch (ip->ip_dst.s_addr) {
500 
501 		case INADDR_BROADCAST:
502 		case INADDR_ANY:
503 			icmpdst.sin_addr = ip->ip_src;
504 			break;
505 
506 		default:
507 			icmpdst.sin_addr = ip->ip_dst;
508 		}
509 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
510 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
511 		if (ia == 0)
512 			break;
513 		if (ia->ia_ifp == 0)
514 			break;
515 		icp->icmp_type = ICMP_MASKREPLY;
516 		if (icmpmaskfake == 0)
517 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
518 		else
519 			icp->icmp_mask = icmpmaskfake;
520 		if (ip->ip_src.s_addr == 0) {
521 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
522 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
523 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
524 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
525 		}
526 reflect:
527 		ip->ip_len += hlen;	/* since ip_input deducts this */
528 		icmpstat.icps_reflect++;
529 		icmpstat.icps_outhist[icp->icmp_type]++;
530 		icmp_reflect(m);
531 		return;
532 
533 	case ICMP_REDIRECT:
534 		if (log_redirect) {
535 			u_long src, dst, gw;
536 
537 			src = ntohl(ip->ip_src.s_addr);
538 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
539 			gw = ntohl(icp->icmp_gwaddr.s_addr);
540 			printf("icmp redirect from %d.%d.%d.%d: "
541 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
542 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
543 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
544 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
545 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
546 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
547 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
548 		}
549 		/*
550 		 * RFC1812 says we must ignore ICMP redirects if we
551 		 * are acting as router.
552 		 */
553 		if (drop_redirect || ipforwarding)
554 			break;
555 		if (code > 3)
556 			goto badcode;
557 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
558 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
559 			icmpstat.icps_badlen++;
560 			break;
561 		}
562 		/*
563 		 * Short circuit routing redirects to force
564 		 * immediate change in the kernel's routing
565 		 * tables.  The message is also handed to anyone
566 		 * listening on a raw socket (e.g. the routing
567 		 * daemon for use in updating its tables).
568 		 */
569 		icmpgw.sin_addr = ip->ip_src;
570 		icmpdst.sin_addr = icp->icmp_gwaddr;
571 #ifdef	ICMPPRINTFS
572 		if (icmpprintfs) {
573 			char buf[4 * sizeof "123"];
574 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
575 
576 			printf("redirect dst %s to %s\n",
577 			       buf, inet_ntoa(icp->icmp_gwaddr));
578 		}
579 #endif
580 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
581 		rtredirect((struct sockaddr *)&icmpsrc,
582 		  (struct sockaddr *)&icmpdst,
583 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
584 		  (struct sockaddr *)&icmpgw);
585 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
586 #ifdef IPSEC
587 		key_sa_routechange((struct sockaddr *)&icmpsrc);
588 #endif
589 		break;
590 
591 	/*
592 	 * No kernel processing for the following;
593 	 * just fall through to send to raw listener.
594 	 */
595 	case ICMP_ECHOREPLY:
596 	case ICMP_ROUTERADVERT:
597 	case ICMP_ROUTERSOLICIT:
598 	case ICMP_TSTAMPREPLY:
599 	case ICMP_IREQREPLY:
600 	case ICMP_MASKREPLY:
601 	default:
602 		break;
603 	}
604 
605 raw:
606 	rip_input(m, off);
607 	return;
608 
609 freeit:
610 	m_freem(m);
611 }
612 
613 /*
614  * Reflect the ip packet back to the source
615  */
616 static void
617 icmp_reflect(m)
618 	struct mbuf *m;
619 {
620 	struct ip *ip = mtod(m, struct ip *);
621 	struct ifaddr *ifa;
622 	struct ifnet *ifn;
623 	struct in_ifaddr *ia;
624 	struct in_addr t;
625 	struct mbuf *opts = 0;
626 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
627 
628 	if (!in_canforward(ip->ip_src) &&
629 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
630 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
631 		m_freem(m);	/* Bad return address */
632 		icmpstat.icps_badaddr++;
633 		goto done;	/* Ip_output() will check for broadcast */
634 	}
635 	t = ip->ip_dst;
636 	ip->ip_dst = ip->ip_src;
637 
638 	/*
639 	 * Source selection for ICMP replies:
640 	 *
641 	 * If the incoming packet was addressed directly to one of our
642 	 * own addresses, use dst as the src for the reply.
643 	 */
644 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
645 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
646 			goto match;
647 	/*
648 	 * If the incoming packet was addressed to one of our broadcast
649 	 * addresses, use the first non-broadcast address which corresponds
650 	 * to the incoming interface.
651 	 */
652 	if (m->m_pkthdr.rcvif != NULL &&
653 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
654 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
655 			if (ifa->ifa_addr->sa_family != AF_INET)
656 				continue;
657 			ia = ifatoia(ifa);
658 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
659 			    t.s_addr)
660 				goto match;
661 		}
662 	}
663 	/*
664 	 * If the packet was transiting through us, use the address of
665 	 * the interface the packet came through in.  If that interface
666 	 * doesn't have a suitable IP address, the normal selection
667 	 * criteria apply.
668 	 */
669 	if (icmp_rfi && m->m_pkthdr.rcvif != NULL) {
670 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
671 			if (ifa->ifa_addr->sa_family != AF_INET)
672 				continue;
673 			ia = ifatoia(ifa);
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(m)) == 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 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, 1280, 1006, 508,
848 		296, 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