xref: /freebsd/sys/netinet/ip_icmp.c (revision 4cf49a43559ed9fdad601bdcccd2c55963008675)
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 <sys/param.h>
38 #include <sys/systm.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/route.h>
48 
49 #define _IP_VHL
50 #include <netinet/in.h>
51 #include <netinet/in_systm.h>
52 #include <netinet/in_var.h>
53 #include <netinet/ip.h>
54 #include <netinet/ip_icmp.h>
55 #include <netinet/ip_var.h>
56 #include <netinet/icmp_var.h>
57 
58 /*
59  * ICMP routines: error generation, receive packet processing, and
60  * routines to turnaround packets back to the originator, and
61  * host table maintenance routines.
62  */
63 
64 static struct	icmpstat icmpstat;
65 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RD,
66 	&icmpstat, icmpstat, "");
67 
68 static int	icmpmaskrepl = 0;
69 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
70 	&icmpmaskrepl, 0, "");
71 
72 static int	drop_redirect = 0;
73 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
74 	&drop_redirect, 0, "");
75 
76 static int	log_redirect = 0;
77 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
78 	&log_redirect, 0, "");
79 
80 #ifdef ICMP_BANDLIM
81 
82 /*
83  * ICMP error-response bandwidth limiting sysctl.  If not enabled, sysctl
84  *      variable content is -1 and read-only.
85  */
86 
87 static int      icmplim = 100;
88 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
89 	&icmplim, 0, "");
90 #else
91 
92 static int      icmplim = -1;
93 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
94 	&icmplim, 0, "");
95 
96 #endif
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 __P((struct mbuf *));
112 static void	icmp_send __P((struct mbuf *, struct mbuf *));
113 static int	ip_next_mtu __P((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 = oiplen + min(8, oip->ip_len);
163 	m->m_len = icmplen + ICMP_MINLEN;
164 	MH_ALIGN(m, m->m_len);
165 	icp = mtod(m, struct icmp *);
166 	if ((u_int)type > ICMP_MAXTYPE)
167 		panic("icmp_error");
168 	icmpstat.icps_outhist[type]++;
169 	icp->icmp_type = type;
170 	if (type == ICMP_REDIRECT)
171 		icp->icmp_gwaddr.s_addr = dest;
172 	else {
173 		icp->icmp_void = 0;
174 		/*
175 		 * The following assignments assume an overlay with the
176 		 * zeroed icmp_void field.
177 		 */
178 		if (type == ICMP_PARAMPROB) {
179 			icp->icmp_pptr = code;
180 			code = 0;
181 		} else if (type == ICMP_UNREACH &&
182 			code == ICMP_UNREACH_NEEDFRAG && destifp) {
183 			icp->icmp_nextmtu = htons(destifp->if_mtu);
184 		}
185 	}
186 
187 	icp->icmp_code = code;
188 	bcopy((caddr_t)oip, (caddr_t)&icp->icmp_ip, icmplen);
189 	nip = &icp->icmp_ip;
190 	nip->ip_len = htons((u_short)(nip->ip_len + oiplen));
191 
192 	/*
193 	 * Now, copy old ip header (without options)
194 	 * in front of icmp message.
195 	 */
196 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
197 		panic("icmp len");
198 	m->m_data -= sizeof(struct ip);
199 	m->m_len += sizeof(struct ip);
200 	m->m_pkthdr.len = m->m_len;
201 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
202 	nip = mtod(m, struct ip *);
203 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
204 	nip->ip_len = m->m_len;
205 	nip->ip_vhl = IP_VHL_BORING;
206 	nip->ip_p = IPPROTO_ICMP;
207 	nip->ip_tos = 0;
208 	icmp_reflect(m);
209 
210 freeit:
211 	m_freem(n);
212 }
213 
214 static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET };
215 static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET };
216 static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET };
217 
218 /*
219  * Process a received ICMP message.
220  */
221 void
222 icmp_input(m, hlen)
223 	register struct mbuf *m;
224 	int hlen;
225 {
226 	register struct icmp *icp;
227 	register struct ip *ip = mtod(m, struct ip *);
228 	int icmplen = ip->ip_len;
229 	register int i;
230 	struct in_ifaddr *ia;
231 	void (*ctlfunc) __P((int, struct sockaddr *, void *));
232 	int code;
233 
234 	/*
235 	 * Locate icmp structure in mbuf, and check
236 	 * that not corrupted and of at least minimum length.
237 	 */
238 #ifdef ICMPPRINTFS
239 	if (icmpprintfs) {
240 		char buf[4 * sizeof "123"];
241 		strcpy(buf, inet_ntoa(ip->ip_src));
242 		printf("icmp_input from %s to %s, len %d\n",
243 		       buf, inet_ntoa(ip->ip_dst), icmplen);
244 	}
245 #endif
246 	if (icmplen < ICMP_MINLEN) {
247 		icmpstat.icps_tooshort++;
248 		goto freeit;
249 	}
250 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
251 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
252 		icmpstat.icps_tooshort++;
253 		return;
254 	}
255 	ip = mtod(m, struct ip *);
256 	m->m_len -= hlen;
257 	m->m_data += hlen;
258 	icp = mtod(m, struct icmp *);
259 	if (in_cksum(m, icmplen)) {
260 		icmpstat.icps_checksum++;
261 		goto freeit;
262 	}
263 	m->m_len += hlen;
264 	m->m_data -= hlen;
265 
266 #ifdef ICMPPRINTFS
267 	if (icmpprintfs)
268 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
269 		    icp->icmp_code);
270 #endif
271 
272 	/*
273 	 * Message type specific processing.
274 	 */
275 	if (icp->icmp_type > ICMP_MAXTYPE)
276 		goto raw;
277 	icmpstat.icps_inhist[icp->icmp_type]++;
278 	code = icp->icmp_code;
279 	switch (icp->icmp_type) {
280 
281 	case ICMP_UNREACH:
282 		switch (code) {
283 			case ICMP_UNREACH_NET:
284 			case ICMP_UNREACH_HOST:
285 			case ICMP_UNREACH_PROTOCOL:
286 			case ICMP_UNREACH_PORT:
287 			case ICMP_UNREACH_SRCFAIL:
288 				code += PRC_UNREACH_NET;
289 				break;
290 
291 			case ICMP_UNREACH_NEEDFRAG:
292 				code = PRC_MSGSIZE;
293 				break;
294 
295 			case ICMP_UNREACH_NET_UNKNOWN:
296 			case ICMP_UNREACH_NET_PROHIB:
297 			case ICMP_UNREACH_TOSNET:
298 				code = PRC_UNREACH_NET;
299 				break;
300 
301 			case ICMP_UNREACH_HOST_UNKNOWN:
302 			case ICMP_UNREACH_ISOLATED:
303 			case ICMP_UNREACH_HOST_PROHIB:
304 			case ICMP_UNREACH_TOSHOST:
305 				code = PRC_UNREACH_HOST;
306 				break;
307 
308 			case ICMP_UNREACH_FILTER_PROHIB:
309 			case ICMP_UNREACH_HOST_PRECEDENCE:
310 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
311 				code = PRC_UNREACH_PORT;
312 				break;
313 
314 			default:
315 				goto badcode;
316 		}
317 		goto deliver;
318 
319 	case ICMP_TIMXCEED:
320 		if (code > 1)
321 			goto badcode;
322 		code += PRC_TIMXCEED_INTRANS;
323 		goto deliver;
324 
325 	case ICMP_PARAMPROB:
326 		if (code > 1)
327 			goto badcode;
328 		code = PRC_PARAMPROB;
329 		goto deliver;
330 
331 	case ICMP_SOURCEQUENCH:
332 		if (code)
333 			goto badcode;
334 		code = PRC_QUENCH;
335 	deliver:
336 		/*
337 		 * Problem with datagram; advise higher level routines.
338 		 */
339 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
340 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
341 			icmpstat.icps_badlen++;
342 			goto freeit;
343 		}
344 		NTOHS(icp->icmp_ip.ip_len);
345 		/* Discard ICMP's in response to multicast packets */
346 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
347 			goto badcode;
348 #ifdef ICMPPRINTFS
349 		if (icmpprintfs)
350 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
351 #endif
352 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
353 #if 1
354 		/*
355 		 * MTU discovery:
356 		 * If we got a needfrag and there is a host route to the
357 		 * original destination, and the MTU is not locked, then
358 		 * set the MTU in the route to the suggested new value
359 		 * (if given) and then notify as usual.  The ULPs will
360 		 * notice that the MTU has changed and adapt accordingly.
361 		 * If no new MTU was suggested, then we guess a new one
362 		 * less than the current value.  If the new MTU is
363 		 * unreasonably small (arbitrarily set at 296), then
364 		 * we reset the MTU to the interface value and enable the
365 		 * lock bit, indicating that we are no longer doing MTU
366 		 * discovery.
367 		 */
368 		if (code == PRC_MSGSIZE) {
369 			struct rtentry *rt;
370 			int mtu;
371 
372 			rt = rtalloc1((struct sockaddr *)&icmpsrc, 0,
373 				      RTF_CLONING | RTF_PRCLONING);
374 			if (rt && (rt->rt_flags & RTF_HOST)
375 			    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
376 				mtu = ntohs(icp->icmp_nextmtu);
377 				if (!mtu)
378 					mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
379 							  1);
380 #ifdef DEBUG_MTUDISC
381 				printf("MTU for %s reduced to %d\n",
382 					inet_ntoa(icmpsrc.sin_addr), mtu);
383 #endif
384 				if (mtu < 296) {
385 					/* rt->rt_rmx.rmx_mtu =
386 						rt->rt_ifp->if_mtu; */
387 					rt->rt_rmx.rmx_locks |= RTV_MTU;
388 				} else if (rt->rt_rmx.rmx_mtu > mtu) {
389 					rt->rt_rmx.rmx_mtu = mtu;
390 				}
391 			}
392 			if (rt)
393 				RTFREE(rt);
394 		}
395 
396 #endif
397 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
398 		if (ctlfunc)
399 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
400 				   (void *)&icp->icmp_ip);
401 		break;
402 
403 	badcode:
404 		icmpstat.icps_badcode++;
405 		break;
406 
407 	case ICMP_ECHO:
408 		if (!icmpbmcastecho
409 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
410 			icmpstat.icps_bmcastecho++;
411 			break;
412 		}
413 		icp->icmp_type = ICMP_ECHOREPLY;
414 		goto reflect;
415 
416 	case ICMP_TSTAMP:
417 		if (!icmpbmcastecho
418 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
419 			icmpstat.icps_bmcasttstamp++;
420 			break;
421 		}
422 		if (icmplen < ICMP_TSLEN) {
423 			icmpstat.icps_badlen++;
424 			break;
425 		}
426 		icp->icmp_type = ICMP_TSTAMPREPLY;
427 		icp->icmp_rtime = iptime();
428 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
429 		goto reflect;
430 
431 	case ICMP_MASKREQ:
432 #define	satosin(sa)	((struct sockaddr_in *)(sa))
433 		if (icmpmaskrepl == 0)
434 			break;
435 		/*
436 		 * We are not able to respond with all ones broadcast
437 		 * unless we receive it over a point-to-point interface.
438 		 */
439 		if (icmplen < ICMP_MASKLEN)
440 			break;
441 		switch (ip->ip_dst.s_addr) {
442 
443 		case INADDR_BROADCAST:
444 		case INADDR_ANY:
445 			icmpdst.sin_addr = ip->ip_src;
446 			break;
447 
448 		default:
449 			icmpdst.sin_addr = ip->ip_dst;
450 		}
451 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
452 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
453 		if (ia == 0)
454 			break;
455 		if (ia->ia_ifp == 0)
456 			break;
457 		icp->icmp_type = ICMP_MASKREPLY;
458 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
459 		if (ip->ip_src.s_addr == 0) {
460 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
461 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
462 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
463 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
464 		}
465 reflect:
466 		ip->ip_len += hlen;	/* since ip_input deducts this */
467 		icmpstat.icps_reflect++;
468 		icmpstat.icps_outhist[icp->icmp_type]++;
469 		icmp_reflect(m);
470 		return;
471 
472 	case ICMP_REDIRECT:
473 		if (log_redirect) {
474 			u_long src, dst, gw;
475 
476 			src = ntohl(ip->ip_src.s_addr);
477 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
478 			gw = ntohl(icp->icmp_gwaddr.s_addr);
479 			printf("icmp redirect from %d.%d.%d.%d: "
480 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
481 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
482 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
483 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
484 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
485 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
486 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
487 		}
488 		if (drop_redirect)
489 			break;
490 		if (code > 3)
491 			goto badcode;
492 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
493 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
494 			icmpstat.icps_badlen++;
495 			break;
496 		}
497 		/*
498 		 * Short circuit routing redirects to force
499 		 * immediate change in the kernel's routing
500 		 * tables.  The message is also handed to anyone
501 		 * listening on a raw socket (e.g. the routing
502 		 * daemon for use in updating its tables).
503 		 */
504 		icmpgw.sin_addr = ip->ip_src;
505 		icmpdst.sin_addr = icp->icmp_gwaddr;
506 #ifdef	ICMPPRINTFS
507 		if (icmpprintfs) {
508 			char buf[4 * sizeof "123"];
509 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
510 
511 			printf("redirect dst %s to %s\n",
512 			       buf, inet_ntoa(icp->icmp_gwaddr));
513 		}
514 #endif
515 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
516 		rtredirect((struct sockaddr *)&icmpsrc,
517 		  (struct sockaddr *)&icmpdst,
518 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
519 		  (struct sockaddr *)&icmpgw, (struct rtentry **)0);
520 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
521 		break;
522 
523 	/*
524 	 * No kernel processing for the following;
525 	 * just fall through to send to raw listener.
526 	 */
527 	case ICMP_ECHOREPLY:
528 	case ICMP_ROUTERADVERT:
529 	case ICMP_ROUTERSOLICIT:
530 	case ICMP_TSTAMPREPLY:
531 	case ICMP_IREQREPLY:
532 	case ICMP_MASKREPLY:
533 	default:
534 		break;
535 	}
536 
537 raw:
538 	rip_input(m, hlen);
539 	return;
540 
541 freeit:
542 	m_freem(m);
543 }
544 
545 /*
546  * Reflect the ip packet back to the source
547  */
548 static void
549 icmp_reflect(m)
550 	struct mbuf *m;
551 {
552 	register struct ip *ip = mtod(m, struct ip *);
553 	register struct in_ifaddr *ia;
554 	struct in_addr t;
555 	struct mbuf *opts = 0;
556 	int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
557 
558 	if (!in_canforward(ip->ip_src) &&
559 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
560 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
561 		m_freem(m);	/* Bad return address */
562 		goto done;	/* Ip_output() will check for broadcast */
563 	}
564 	t = ip->ip_dst;
565 	ip->ip_dst = ip->ip_src;
566 	/*
567 	 * If the incoming packet was addressed directly to us,
568 	 * use dst as the src for the reply.  Otherwise (broadcast
569 	 * or anonymous), use the address which corresponds
570 	 * to the incoming interface.
571 	 */
572 	for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) {
573 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
574 			break;
575 		if (ia->ia_ifp && (ia->ia_ifp->if_flags & IFF_BROADCAST) &&
576 		    t.s_addr == satosin(&ia->ia_broadaddr)->sin_addr.s_addr)
577 			break;
578 	}
579 	icmpdst.sin_addr = t;
580 	if ((ia == (struct in_ifaddr *)0) && m->m_pkthdr.rcvif)
581 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
582 			(struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
583 	/*
584 	 * The following happens if the packet was not addressed to us,
585 	 * and was received on an interface with no IP address.
586 	 */
587 	if (ia == (struct in_ifaddr *)0)
588 		ia = in_ifaddrhead.tqh_first;
589 	t = IA_SIN(ia)->sin_addr;
590 	ip->ip_src = t;
591 	ip->ip_ttl = MAXTTL;
592 
593 	if (optlen > 0) {
594 		register u_char *cp;
595 		int opt, cnt;
596 		u_int len;
597 
598 		/*
599 		 * Retrieve any source routing from the incoming packet;
600 		 * add on any record-route or timestamp options.
601 		 */
602 		cp = (u_char *) (ip + 1);
603 		if ((opts = ip_srcroute()) == 0 &&
604 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
605 			opts->m_len = sizeof(struct in_addr);
606 			mtod(opts, struct in_addr *)->s_addr = 0;
607 		}
608 		if (opts) {
609 #ifdef ICMPPRINTFS
610 		    if (icmpprintfs)
611 			    printf("icmp_reflect optlen %d rt %d => ",
612 				optlen, opts->m_len);
613 #endif
614 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
615 			    opt = cp[IPOPT_OPTVAL];
616 			    if (opt == IPOPT_EOL)
617 				    break;
618 			    if (opt == IPOPT_NOP)
619 				    len = 1;
620 			    else {
621 				    len = cp[IPOPT_OLEN];
622 				    if (len <= 0 || len > cnt)
623 					    break;
624 			    }
625 			    /*
626 			     * Should check for overflow, but it "can't happen"
627 			     */
628 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
629 				opt == IPOPT_SECURITY) {
630 				    bcopy((caddr_t)cp,
631 					mtod(opts, caddr_t) + opts->m_len, len);
632 				    opts->m_len += len;
633 			    }
634 		    }
635 		    /* Terminate & pad, if necessary */
636 		    cnt = opts->m_len % 4;
637 		    if (cnt) {
638 			    for (; cnt < 4; cnt++) {
639 				    *(mtod(opts, caddr_t) + opts->m_len) =
640 					IPOPT_EOL;
641 				    opts->m_len++;
642 			    }
643 		    }
644 #ifdef ICMPPRINTFS
645 		    if (icmpprintfs)
646 			    printf("%d\n", opts->m_len);
647 #endif
648 		}
649 		/*
650 		 * Now strip out original options by copying rest of first
651 		 * mbuf's data back, and adjust the IP length.
652 		 */
653 		ip->ip_len -= optlen;
654 		ip->ip_vhl = IP_VHL_BORING;
655 		m->m_len -= optlen;
656 		if (m->m_flags & M_PKTHDR)
657 			m->m_pkthdr.len -= optlen;
658 		optlen += sizeof(struct ip);
659 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
660 			 (unsigned)(m->m_len - sizeof(struct ip)));
661 	}
662 	m->m_flags &= ~(M_BCAST|M_MCAST);
663 	icmp_send(m, opts);
664 done:
665 	if (opts)
666 		(void)m_free(opts);
667 }
668 
669 /*
670  * Send an icmp packet back to the ip level,
671  * after supplying a checksum.
672  */
673 static void
674 icmp_send(m, opts)
675 	register struct mbuf *m;
676 	struct mbuf *opts;
677 {
678 	register struct ip *ip = mtod(m, struct ip *);
679 	register int hlen;
680 	register struct icmp *icp;
681 	struct route ro;
682 
683 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
684 	m->m_data += hlen;
685 	m->m_len -= hlen;
686 	icp = mtod(m, struct icmp *);
687 	icp->icmp_cksum = 0;
688 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
689 	m->m_data -= hlen;
690 	m->m_len += hlen;
691 	m->m_pkthdr.rcvif = (struct ifnet *)0;
692 #ifdef ICMPPRINTFS
693 	if (icmpprintfs) {
694 		char buf[4 * sizeof "123"];
695 		strcpy(buf, inet_ntoa(ip->ip_dst));
696 		printf("icmp_send dst %s src %s\n",
697 		       buf, inet_ntoa(ip->ip_src));
698 	}
699 #endif
700 	bzero(&ro, sizeof ro);
701 	(void) ip_output(m, opts, &ro, 0, NULL);
702 	if (ro.ro_rt)
703 		RTFREE(ro.ro_rt);
704 }
705 
706 n_time
707 iptime()
708 {
709 	struct timeval atv;
710 	u_long t;
711 
712 	microtime(&atv);
713 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
714 	return (htonl(t));
715 }
716 
717 #if 1
718 /*
719  * Return the next larger or smaller MTU plateau (table from RFC 1191)
720  * given current value MTU.  If DIR is less than zero, a larger plateau
721  * is returned; otherwise, a smaller value is returned.
722  */
723 static int
724 ip_next_mtu(mtu, dir)
725 	int mtu;
726 	int dir;
727 {
728 	static int mtutab[] = {
729 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
730 		68, 0
731 	};
732 	int i;
733 
734 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
735 		if (mtu >= mtutab[i])
736 			break;
737 	}
738 
739 	if (dir < 0) {
740 		if (i == 0) {
741 			return 0;
742 		} else {
743 			return mtutab[i - 1];
744 		}
745 	} else {
746 		if (mtutab[i] == 0) {
747 			return 0;
748 		} else if(mtu > mtutab[i]) {
749 			return mtutab[i];
750 		} else {
751 			return mtutab[i + 1];
752 		}
753 	}
754 }
755 #endif
756 
757 #ifdef ICMP_BANDLIM
758 
759 /*
760  * badport_bandlim() - check for ICMP bandwidth limit
761  *
762  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
763  *	hit our bandwidth limit and it is not ok.
764  *
765  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
766  *
767  *	For now we separate the TCP and UDP subsystems w/ different 'which'
768  *	values.  We may eventually remove this separation (and simplify the
769  *	code further).
770  *
771  *	Note that the printing of the error message is delayed so we can
772  *	properly print the icmp error rate that the system was trying to do
773  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
774  *	the 'final' error, but it doesn't make sense to solve the printing
775  *	delay with more complex code.
776  */
777 
778 int
779 badport_bandlim(int which)
780 {
781 	static int lticks[2];
782 	static int lpackets[2];
783 	int dticks;
784 
785 	/*
786 	 * Return ok status if feature disabled or argument out of
787 	 * ranage.
788 	 */
789 
790 	if (icmplim <= 0 || which >= 2 || which < 0)
791 		return(0);
792 	dticks = ticks - lticks[which];
793 
794 	/*
795 	 * reset stats when cumulative dt exceeds one second.
796 	 */
797 
798 	if ((unsigned int)dticks > hz) {
799 		if (lpackets[which] > icmplim) {
800 			printf("icmp-response bandwidth limit %d/%d pps\n",
801 				lpackets[which],
802 				icmplim
803 			);
804 		}
805 		lticks[which] = ticks;
806 		lpackets[which] = 0;
807 	}
808 
809 	/*
810 	 * bump packet count
811 	 */
812 
813 	if (++lpackets[which] > icmplim) {
814 		return(-1);
815 	}
816 	return(0);
817 }
818 
819 #endif
820 
821 
822