xref: /freebsd/sys/netinet/ip_icmp.c (revision c17d43407fe04133a94055b0dbc7ea8965654a9f)
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 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/mbuf.h>
42 #include <sys/protosw.h>
43 #include <sys/socket.h>
44 #include <sys/time.h>
45 #include <sys/kernel.h>
46 #include <sys/sysctl.h>
47 
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/route.h>
51 
52 #define _IP_VHL
53 #include <netinet/in.h>
54 #include <netinet/in_systm.h>
55 #include <netinet/in_var.h>
56 #include <netinet/ip.h>
57 #include <netinet/ip_icmp.h>
58 #include <netinet/ip_var.h>
59 #include <netinet/icmp_var.h>
60 
61 #ifdef IPSEC
62 #include <netinet6/ipsec.h>
63 #include <netkey/key.h>
64 #endif
65 
66 #include <machine/in_cksum.h>
67 
68 /*
69  * ICMP routines: error generation, receive packet processing, and
70  * routines to turnaround packets back to the originator, and
71  * host table maintenance routines.
72  */
73 
74 static struct	icmpstat icmpstat;
75 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
76 	&icmpstat, icmpstat, "");
77 
78 static int	icmpmaskrepl = 0;
79 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
80 	&icmpmaskrepl, 0, "");
81 
82 static int	drop_redirect = 0;
83 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
84 	&drop_redirect, 0, "");
85 
86 static int	log_redirect = 0;
87 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
88 	&log_redirect, 0, "");
89 
90 static int      icmplim = 200;
91 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
92 	&icmplim, 0, "");
93 
94 static int	icmplim_output = 1;
95 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
96 	&icmplim_output, 0, "");
97 
98 /*
99  * ICMP broadcast echo sysctl
100  */
101 
102 static int	icmpbmcastecho = 0;
103 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
104 	&icmpbmcastecho, 0, "");
105 
106 
107 #ifdef ICMPPRINTFS
108 int	icmpprintfs = 0;
109 #endif
110 
111 static void	icmp_reflect(struct mbuf *);
112 static void	icmp_send(struct mbuf *, struct mbuf *, struct route *);
113 static int	ip_next_mtu(int, int);
114 
115 extern	struct protosw inetsw[];
116 
117 /*
118  * Generate an error packet of type error
119  * in response to bad packet ip.
120  */
121 void
122 icmp_error(n, type, code, dest, destifp)
123 	struct mbuf *n;
124 	int type, code;
125 	n_long dest;
126 	struct ifnet *destifp;
127 {
128 	register struct ip *oip = mtod(n, struct ip *), *nip;
129 	register unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
130 	register struct icmp *icp;
131 	register struct mbuf *m;
132 	unsigned icmplen;
133 
134 #ifdef ICMPPRINTFS
135 	if (icmpprintfs)
136 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
137 #endif
138 	if (type != ICMP_REDIRECT)
139 		icmpstat.icps_error++;
140 	/*
141 	 * Don't send error if not the first fragment of message.
142 	 * Don't error if the old packet protocol was ICMP
143 	 * error message, only known informational types.
144 	 */
145 	if (oip->ip_off &~ (IP_MF|IP_DF))
146 		goto freeit;
147 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
148 	  n->m_len >= oiplen + ICMP_MINLEN &&
149 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
150 		icmpstat.icps_oldicmp++;
151 		goto freeit;
152 	}
153 	/* Don't send error in response to a multicast or broadcast packet */
154 	if (n->m_flags & (M_BCAST|M_MCAST))
155 		goto freeit;
156 	/*
157 	 * First, formulate icmp message
158 	 */
159 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
160 	if (m == NULL)
161 		goto freeit;
162 	icmplen = min(oiplen + 8, oip->ip_len);
163 	if (icmplen < sizeof(struct ip))
164 		panic("icmp_error: bad length");
165 	m->m_len = icmplen + ICMP_MINLEN;
166 	MH_ALIGN(m, m->m_len);
167 	icp = mtod(m, struct icmp *);
168 	if ((u_int)type > ICMP_MAXTYPE)
169 		panic("icmp_error");
170 	icmpstat.icps_outhist[type]++;
171 	icp->icmp_type = type;
172 	if (type == ICMP_REDIRECT)
173 		icp->icmp_gwaddr.s_addr = dest;
174 	else {
175 		icp->icmp_void = 0;
176 		/*
177 		 * The following assignments assume an overlay with the
178 		 * zeroed icmp_void field.
179 		 */
180 		if (type == ICMP_PARAMPROB) {
181 			icp->icmp_pptr = code;
182 			code = 0;
183 		} else if (type == ICMP_UNREACH &&
184 			code == ICMP_UNREACH_NEEDFRAG && destifp) {
185 			icp->icmp_nextmtu = htons(destifp->if_mtu);
186 		}
187 	}
188 
189 	icp->icmp_code = code;
190 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
191 	nip = &icp->icmp_ip;
192 
193 	/*
194 	 * Convert fields to network representation.
195 	 */
196 	nip->ip_len = htons(nip->ip_len);
197 	nip->ip_off = htons(nip->ip_off);
198 
199 	/*
200 	 * Now, copy old ip header (without options)
201 	 * in front of icmp message.
202 	 */
203 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
204 		panic("icmp len");
205 	m->m_data -= sizeof(struct ip);
206 	m->m_len += sizeof(struct ip);
207 	m->m_pkthdr.len = m->m_len;
208 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
209 	nip = mtod(m, struct ip *);
210 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
211 	nip->ip_len = m->m_len;
212 	nip->ip_vhl = IP_VHL_BORING;
213 	nip->ip_p = IPPROTO_ICMP;
214 	nip->ip_tos = 0;
215 	icmp_reflect(m);
216 
217 freeit:
218 	m_freem(n);
219 }
220 
221 static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET };
222 static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET };
223 static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET };
224 
225 /*
226  * Process a received ICMP message.
227  */
228 void
229 icmp_input(m, off)
230 	register struct mbuf *m;
231 	int off;
232 {
233 	int hlen = off;
234 	register struct icmp *icp;
235 	register struct ip *ip = mtod(m, struct ip *);
236 	int icmplen = ip->ip_len;
237 	register int i;
238 	struct in_ifaddr *ia;
239 	void (*ctlfunc)(int, struct sockaddr *, void *);
240 	int code;
241 
242 	/*
243 	 * Locate icmp structure in mbuf, and check
244 	 * that not corrupted and of at least minimum length.
245 	 */
246 #ifdef ICMPPRINTFS
247 	if (icmpprintfs) {
248 		char buf[4 * sizeof "123"];
249 		strcpy(buf, inet_ntoa(ip->ip_src));
250 		printf("icmp_input from %s to %s, len %d\n",
251 		       buf, inet_ntoa(ip->ip_dst), icmplen);
252 	}
253 #endif
254 	if (icmplen < ICMP_MINLEN) {
255 		icmpstat.icps_tooshort++;
256 		goto freeit;
257 	}
258 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
259 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
260 		icmpstat.icps_tooshort++;
261 		return;
262 	}
263 	ip = mtod(m, struct ip *);
264 	m->m_len -= hlen;
265 	m->m_data += hlen;
266 	icp = mtod(m, struct icmp *);
267 	if (in_cksum(m, icmplen)) {
268 		icmpstat.icps_checksum++;
269 		goto freeit;
270 	}
271 	m->m_len += hlen;
272 	m->m_data -= hlen;
273 
274 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
275 		/*
276 		 * Deliver very specific ICMP type only.
277 		 */
278 		switch (icp->icmp_type) {
279 		case ICMP_UNREACH:
280 		case ICMP_TIMXCEED:
281 			break;
282 		default:
283 			goto freeit;
284 		}
285 	}
286 
287 #ifdef ICMPPRINTFS
288 	if (icmpprintfs)
289 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
290 		    icp->icmp_code);
291 #endif
292 
293 	/*
294 	 * Message type specific processing.
295 	 */
296 	if (icp->icmp_type > ICMP_MAXTYPE)
297 		goto raw;
298 	icmpstat.icps_inhist[icp->icmp_type]++;
299 	code = icp->icmp_code;
300 	switch (icp->icmp_type) {
301 
302 	case ICMP_UNREACH:
303 		switch (code) {
304 			case ICMP_UNREACH_NET:
305 			case ICMP_UNREACH_HOST:
306 			case ICMP_UNREACH_SRCFAIL:
307 			case ICMP_UNREACH_NET_UNKNOWN:
308 			case ICMP_UNREACH_HOST_UNKNOWN:
309 			case ICMP_UNREACH_ISOLATED:
310 			case ICMP_UNREACH_TOSNET:
311 			case ICMP_UNREACH_TOSHOST:
312 			case ICMP_UNREACH_HOST_PRECEDENCE:
313 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
314 				code = PRC_UNREACH_NET;
315 				break;
316 
317 			case ICMP_UNREACH_NEEDFRAG:
318 				code = PRC_MSGSIZE;
319 				break;
320 
321 			/*
322 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
323 			 * Treat subcodes 2,3 as immediate RST
324 			 */
325 			case ICMP_UNREACH_PROTOCOL:
326 			case ICMP_UNREACH_PORT:
327 				code = PRC_UNREACH_PORT;
328 				break;
329 
330 			case ICMP_UNREACH_NET_PROHIB:
331 			case ICMP_UNREACH_HOST_PROHIB:
332 			case ICMP_UNREACH_FILTER_PROHIB:
333 				code = PRC_UNREACH_ADMIN_PROHIB;
334 				break;
335 
336 			default:
337 				goto badcode;
338 		}
339 		goto deliver;
340 
341 	case ICMP_TIMXCEED:
342 		if (code > 1)
343 			goto badcode;
344 		code += PRC_TIMXCEED_INTRANS;
345 		goto deliver;
346 
347 	case ICMP_PARAMPROB:
348 		if (code > 1)
349 			goto badcode;
350 		code = PRC_PARAMPROB;
351 		goto deliver;
352 
353 	case ICMP_SOURCEQUENCH:
354 		if (code)
355 			goto badcode;
356 		code = PRC_QUENCH;
357 	deliver:
358 		/*
359 		 * Problem with datagram; advise higher level routines.
360 		 */
361 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
362 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
363 			icmpstat.icps_badlen++;
364 			goto freeit;
365 		}
366 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
367 		/* Discard ICMP's in response to multicast packets */
368 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
369 			goto badcode;
370 #ifdef ICMPPRINTFS
371 		if (icmpprintfs)
372 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
373 #endif
374 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
375 #if 1
376 		/*
377 		 * MTU discovery:
378 		 * If we got a needfrag and there is a host route to the
379 		 * original destination, and the MTU is not locked, then
380 		 * set the MTU in the route to the suggested new value
381 		 * (if given) and then notify as usual.  The ULPs will
382 		 * notice that the MTU has changed and adapt accordingly.
383 		 * If no new MTU was suggested, then we guess a new one
384 		 * less than the current value.  If the new MTU is
385 		 * unreasonably small (arbitrarily set at 296), then
386 		 * we reset the MTU to the interface value and enable the
387 		 * lock bit, indicating that we are no longer doing MTU
388 		 * discovery.
389 		 */
390 		if (code == PRC_MSGSIZE) {
391 			struct rtentry *rt;
392 			int mtu;
393 
394 			rt = rtalloc1((struct sockaddr *)&icmpsrc, 0,
395 				      RTF_CLONING | RTF_PRCLONING);
396 			if (rt && (rt->rt_flags & RTF_HOST)
397 			    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
398 				mtu = ntohs(icp->icmp_nextmtu);
399 				if (!mtu)
400 					mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
401 							  1);
402 #ifdef DEBUG_MTUDISC
403 				printf("MTU for %s reduced to %d\n",
404 					inet_ntoa(icmpsrc.sin_addr), mtu);
405 #endif
406 				if (mtu < 296) {
407 					/* rt->rt_rmx.rmx_mtu =
408 						rt->rt_ifp->if_mtu; */
409 					rt->rt_rmx.rmx_locks |= RTV_MTU;
410 				} else if (rt->rt_rmx.rmx_mtu > mtu) {
411 					rt->rt_rmx.rmx_mtu = mtu;
412 				}
413 			}
414 			if (rt)
415 				RTFREE(rt);
416 		}
417 
418 #endif
419 		/*
420 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
421 		 * notification to TCP layer.
422 		 */
423 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
424 		if (ctlfunc)
425 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
426 				   (void *)&icp->icmp_ip);
427 		break;
428 
429 	badcode:
430 		icmpstat.icps_badcode++;
431 		break;
432 
433 	case ICMP_ECHO:
434 		if (!icmpbmcastecho
435 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
436 			icmpstat.icps_bmcastecho++;
437 			break;
438 		}
439 		icp->icmp_type = ICMP_ECHOREPLY;
440 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
441 			goto freeit;
442 		else
443 			goto reflect;
444 
445 	case ICMP_TSTAMP:
446 		if (!icmpbmcastecho
447 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
448 			icmpstat.icps_bmcasttstamp++;
449 			break;
450 		}
451 		if (icmplen < ICMP_TSLEN) {
452 			icmpstat.icps_badlen++;
453 			break;
454 		}
455 		icp->icmp_type = ICMP_TSTAMPREPLY;
456 		icp->icmp_rtime = iptime();
457 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
458 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
459 			goto freeit;
460 		else
461 			goto reflect;
462 
463 	case ICMP_MASKREQ:
464 		if (icmpmaskrepl == 0)
465 			break;
466 		/*
467 		 * We are not able to respond with all ones broadcast
468 		 * unless we receive it over a point-to-point interface.
469 		 */
470 		if (icmplen < ICMP_MASKLEN)
471 			break;
472 		switch (ip->ip_dst.s_addr) {
473 
474 		case INADDR_BROADCAST:
475 		case INADDR_ANY:
476 			icmpdst.sin_addr = ip->ip_src;
477 			break;
478 
479 		default:
480 			icmpdst.sin_addr = ip->ip_dst;
481 		}
482 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
483 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
484 		if (ia == 0)
485 			break;
486 		if (ia->ia_ifp == 0)
487 			break;
488 		icp->icmp_type = ICMP_MASKREPLY;
489 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
490 		if (ip->ip_src.s_addr == 0) {
491 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
492 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
493 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
494 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
495 		}
496 reflect:
497 		ip->ip_len += hlen;	/* since ip_input deducts this */
498 		icmpstat.icps_reflect++;
499 		icmpstat.icps_outhist[icp->icmp_type]++;
500 		icmp_reflect(m);
501 		return;
502 
503 	case ICMP_REDIRECT:
504 		if (log_redirect) {
505 			u_long src, dst, gw;
506 
507 			src = ntohl(ip->ip_src.s_addr);
508 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
509 			gw = ntohl(icp->icmp_gwaddr.s_addr);
510 			printf("icmp redirect from %d.%d.%d.%d: "
511 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
512 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
513 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
514 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
515 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
516 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
517 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
518 		}
519 		if (drop_redirect)
520 			break;
521 		if (code > 3)
522 			goto badcode;
523 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
524 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
525 			icmpstat.icps_badlen++;
526 			break;
527 		}
528 		/*
529 		 * Short circuit routing redirects to force
530 		 * immediate change in the kernel's routing
531 		 * tables.  The message is also handed to anyone
532 		 * listening on a raw socket (e.g. the routing
533 		 * daemon for use in updating its tables).
534 		 */
535 		icmpgw.sin_addr = ip->ip_src;
536 		icmpdst.sin_addr = icp->icmp_gwaddr;
537 #ifdef	ICMPPRINTFS
538 		if (icmpprintfs) {
539 			char buf[4 * sizeof "123"];
540 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
541 
542 			printf("redirect dst %s to %s\n",
543 			       buf, inet_ntoa(icp->icmp_gwaddr));
544 		}
545 #endif
546 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
547 		rtredirect((struct sockaddr *)&icmpsrc,
548 		  (struct sockaddr *)&icmpdst,
549 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
550 		  (struct sockaddr *)&icmpgw, (struct rtentry **)0);
551 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
552 #ifdef IPSEC
553 		key_sa_routechange((struct sockaddr *)&icmpsrc);
554 #endif
555 		break;
556 
557 	/*
558 	 * No kernel processing for the following;
559 	 * just fall through to send to raw listener.
560 	 */
561 	case ICMP_ECHOREPLY:
562 	case ICMP_ROUTERADVERT:
563 	case ICMP_ROUTERSOLICIT:
564 	case ICMP_TSTAMPREPLY:
565 	case ICMP_IREQREPLY:
566 	case ICMP_MASKREPLY:
567 	default:
568 		break;
569 	}
570 
571 raw:
572 	rip_input(m, off);
573 	return;
574 
575 freeit:
576 	m_freem(m);
577 }
578 
579 /*
580  * Reflect the ip packet back to the source
581  */
582 static void
583 icmp_reflect(m)
584 	struct mbuf *m;
585 {
586 	struct ip *ip = mtod(m, struct ip *);
587 	struct ifaddr *ifa;
588 	struct in_ifaddr *ia;
589 	struct in_addr t;
590 	struct mbuf *opts = 0;
591 	int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
592 	struct route *ro = NULL, rt;
593 
594 	if (!in_canforward(ip->ip_src) &&
595 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
596 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
597 		m_freem(m);	/* Bad return address */
598 		icmpstat.icps_badaddr++;
599 		goto done;	/* Ip_output() will check for broadcast */
600 	}
601 	t = ip->ip_dst;
602 	ip->ip_dst = ip->ip_src;
603 	ro = &rt;
604 	bzero(ro, sizeof(*ro));
605 	/*
606 	 * If the incoming packet was addressed directly to us,
607 	 * use dst as the src for the reply.  Otherwise (broadcast
608 	 * or anonymous), use the address which corresponds
609 	 * to the incoming interface.
610 	 */
611 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
612 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
613 			goto match;
614 	if (m->m_pkthdr.rcvif != NULL &&
615 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
616 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
617 			if (ifa->ifa_addr->sa_family != AF_INET)
618 				continue;
619 			ia = ifatoia(ifa);
620 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
621 			    t.s_addr)
622 				goto match;
623 		}
624 	}
625 	ia = ip_rtaddr(ip->ip_dst, ro);
626 	/* We need a route to do anything useful. */
627 	if (ia == NULL) {
628 		m_freem(m);
629 		icmpstat.icps_noroute++;
630 		goto done;
631 	}
632 match:
633 	t = IA_SIN(ia)->sin_addr;
634 	ip->ip_src = t;
635 	ip->ip_ttl = ip_defttl;
636 
637 	if (optlen > 0) {
638 		register u_char *cp;
639 		int opt, cnt;
640 		u_int len;
641 
642 		/*
643 		 * Retrieve any source routing from the incoming packet;
644 		 * add on any record-route or timestamp options.
645 		 */
646 		cp = (u_char *) (ip + 1);
647 		if ((opts = ip_srcroute()) == 0 &&
648 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
649 			opts->m_len = sizeof(struct in_addr);
650 			mtod(opts, struct in_addr *)->s_addr = 0;
651 		}
652 		if (opts) {
653 #ifdef ICMPPRINTFS
654 		    if (icmpprintfs)
655 			    printf("icmp_reflect optlen %d rt %d => ",
656 				optlen, opts->m_len);
657 #endif
658 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
659 			    opt = cp[IPOPT_OPTVAL];
660 			    if (opt == IPOPT_EOL)
661 				    break;
662 			    if (opt == IPOPT_NOP)
663 				    len = 1;
664 			    else {
665 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
666 					    break;
667 				    len = cp[IPOPT_OLEN];
668 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
669 				        len > cnt)
670 					    break;
671 			    }
672 			    /*
673 			     * Should check for overflow, but it "can't happen"
674 			     */
675 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
676 				opt == IPOPT_SECURITY) {
677 				    bcopy((caddr_t)cp,
678 					mtod(opts, caddr_t) + opts->m_len, len);
679 				    opts->m_len += len;
680 			    }
681 		    }
682 		    /* Terminate & pad, if necessary */
683 		    cnt = opts->m_len % 4;
684 		    if (cnt) {
685 			    for (; cnt < 4; cnt++) {
686 				    *(mtod(opts, caddr_t) + opts->m_len) =
687 					IPOPT_EOL;
688 				    opts->m_len++;
689 			    }
690 		    }
691 #ifdef ICMPPRINTFS
692 		    if (icmpprintfs)
693 			    printf("%d\n", opts->m_len);
694 #endif
695 		}
696 		/*
697 		 * Now strip out original options by copying rest of first
698 		 * mbuf's data back, and adjust the IP length.
699 		 */
700 		ip->ip_len -= optlen;
701 		ip->ip_vhl = IP_VHL_BORING;
702 		m->m_len -= optlen;
703 		if (m->m_flags & M_PKTHDR)
704 			m->m_pkthdr.len -= optlen;
705 		optlen += sizeof(struct ip);
706 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
707 			 (unsigned)(m->m_len - sizeof(struct ip)));
708 	}
709 	m->m_flags &= ~(M_BCAST|M_MCAST);
710 	icmp_send(m, opts, ro);
711 done:
712 	if (opts)
713 		(void)m_free(opts);
714 	if (ro && ro->ro_rt)
715 		RTFREE(ro->ro_rt);
716 }
717 
718 /*
719  * Send an icmp packet back to the ip level,
720  * after supplying a checksum.
721  */
722 static void
723 icmp_send(m, opts, rt)
724 	register struct mbuf *m;
725 	struct mbuf *opts;
726 	struct route *rt;
727 {
728 	register struct ip *ip = mtod(m, struct ip *);
729 	register int hlen;
730 	register struct icmp *icp;
731 
732 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
733 	m->m_data += hlen;
734 	m->m_len -= hlen;
735 	icp = mtod(m, struct icmp *);
736 	icp->icmp_cksum = 0;
737 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
738 	m->m_data -= hlen;
739 	m->m_len += hlen;
740 	m->m_pkthdr.rcvif = (struct ifnet *)0;
741 #ifdef ICMPPRINTFS
742 	if (icmpprintfs) {
743 		char buf[4 * sizeof "123"];
744 		strcpy(buf, inet_ntoa(ip->ip_dst));
745 		printf("icmp_send dst %s src %s\n",
746 		       buf, inet_ntoa(ip->ip_src));
747 	}
748 #endif
749 	(void) ip_output(m, opts, rt, 0, NULL);
750 }
751 
752 n_time
753 iptime()
754 {
755 	struct timeval atv;
756 	u_long t;
757 
758 	getmicrotime(&atv);
759 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
760 	return (htonl(t));
761 }
762 
763 #if 1
764 /*
765  * Return the next larger or smaller MTU plateau (table from RFC 1191)
766  * given current value MTU.  If DIR is less than zero, a larger plateau
767  * is returned; otherwise, a smaller value is returned.
768  */
769 static int
770 ip_next_mtu(mtu, dir)
771 	int mtu;
772 	int dir;
773 {
774 	static int mtutab[] = {
775 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
776 		68, 0
777 	};
778 	int i;
779 
780 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
781 		if (mtu >= mtutab[i])
782 			break;
783 	}
784 
785 	if (dir < 0) {
786 		if (i == 0) {
787 			return 0;
788 		} else {
789 			return mtutab[i - 1];
790 		}
791 	} else {
792 		if (mtutab[i] == 0) {
793 			return 0;
794 		} else if(mtu > mtutab[i]) {
795 			return mtutab[i];
796 		} else {
797 			return mtutab[i + 1];
798 		}
799 	}
800 }
801 #endif
802 
803 
804 /*
805  * badport_bandlim() - check for ICMP bandwidth limit
806  *
807  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
808  *	hit our bandwidth limit and it is not ok.
809  *
810  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
811  *
812  *	For now we separate the TCP and UDP subsystems w/ different 'which'
813  *	values.  We may eventually remove this separation (and simplify the
814  *	code further).
815  *
816  *	Note that the printing of the error message is delayed so we can
817  *	properly print the icmp error rate that the system was trying to do
818  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
819  *	the 'final' error, but it doesn't make sense to solve the printing
820  *	delay with more complex code.
821  */
822 
823 int
824 badport_bandlim(int which)
825 {
826 	static int lticks[BANDLIM_MAX + 1];
827 	static int lpackets[BANDLIM_MAX + 1];
828 	int dticks;
829 	const char *bandlimittype[] = {
830 		"Limiting icmp unreach response",
831 		"Limiting icmp ping response",
832 		"Limiting icmp tstamp response",
833 		"Limiting closed port RST response",
834 		"Limiting open port RST response"
835 		};
836 
837 	/*
838 	 * Return ok status if feature disabled or argument out of
839 	 * ranage.
840 	 */
841 
842 	if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
843 		return(0);
844 	dticks = ticks - lticks[which];
845 
846 	/*
847 	 * reset stats when cumulative dt exceeds one second.
848 	 */
849 
850 	if ((unsigned int)dticks > hz) {
851 		if (lpackets[which] > icmplim && icmplim_output) {
852 			printf("%s from %d to %d packets per second\n",
853 				bandlimittype[which],
854 				lpackets[which],
855 				icmplim
856 			);
857 		}
858 		lticks[which] = ticks;
859 		lpackets[which] = 0;
860 	}
861 
862 	/*
863 	 * bump packet count
864 	 */
865 
866 	if (++lpackets[which] > icmplim) {
867 		return(-1);
868 	}
869 	return(0);
870 }
871 
872