xref: /freebsd/sys/netinet/ip_icmp.c (revision 77a0943ded95b9e6438f7db70c4a28e4d93946d4)
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/route.h>
50 
51 #define _IP_VHL
52 #include <netinet/in.h>
53 #include <netinet/in_systm.h>
54 #include <netinet/in_var.h>
55 #include <netinet/ip.h>
56 #include <netinet/ip_icmp.h>
57 #include <netinet/ip_var.h>
58 #include <netinet/icmp_var.h>
59 
60 #ifdef IPSEC
61 #include <netinet6/ipsec.h>
62 #include <netkey/key.h>
63 #endif
64 
65 #include "faith.h"
66 #if defined(NFAITH) && NFAITH > 0
67 #include <net/if_types.h>
68 #endif
69 
70 #include <machine/in_cksum.h>
71 
72 /*
73  * ICMP routines: error generation, receive packet processing, and
74  * routines to turnaround packets back to the originator, and
75  * host table maintenance routines.
76  */
77 
78 static struct	icmpstat icmpstat;
79 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RD,
80 	&icmpstat, icmpstat, "");
81 
82 static int	icmpmaskrepl = 0;
83 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
84 	&icmpmaskrepl, 0, "");
85 
86 static int	drop_redirect = 0;
87 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
88 	&drop_redirect, 0, "");
89 
90 static int	log_redirect = 0;
91 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
92 	&log_redirect, 0, "");
93 
94 static int      icmplim = 200;
95 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
96 	&icmplim, 0, "");
97 
98 static int	icmplim_output = 1;
99 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
100 	&icmplim_output, 0, "");
101 
102 /*
103  * ICMP broadcast echo sysctl
104  */
105 
106 static int	icmpbmcastecho = 0;
107 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
108 	&icmpbmcastecho, 0, "");
109 
110 
111 #ifdef ICMPPRINTFS
112 int	icmpprintfs = 0;
113 #endif
114 
115 static void	icmp_reflect __P((struct mbuf *));
116 static void	icmp_send __P((struct mbuf *, struct mbuf *));
117 static int	ip_next_mtu __P((int, int));
118 
119 extern	struct protosw inetsw[];
120 
121 /*
122  * Generate an error packet of type error
123  * in response to bad packet ip.
124  */
125 void
126 icmp_error(n, type, code, dest, destifp)
127 	struct mbuf *n;
128 	int type, code;
129 	n_long dest;
130 	struct ifnet *destifp;
131 {
132 	register struct ip *oip = mtod(n, struct ip *), *nip;
133 	register unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
134 	register struct icmp *icp;
135 	register struct mbuf *m;
136 	unsigned icmplen;
137 
138 #ifdef ICMPPRINTFS
139 	if (icmpprintfs)
140 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
141 #endif
142 	if (type != ICMP_REDIRECT)
143 		icmpstat.icps_error++;
144 	/*
145 	 * Don't send error if not the first fragment of message.
146 	 * Don't error if the old packet protocol was ICMP
147 	 * error message, only known informational types.
148 	 */
149 	if (oip->ip_off &~ (IP_MF|IP_DF))
150 		goto freeit;
151 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
152 	  n->m_len >= oiplen + ICMP_MINLEN &&
153 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
154 		icmpstat.icps_oldicmp++;
155 		goto freeit;
156 	}
157 	/* Don't send error in response to a multicast or broadcast packet */
158 	if (n->m_flags & (M_BCAST|M_MCAST))
159 		goto freeit;
160 	/*
161 	 * First, formulate icmp message
162 	 */
163 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
164 	if (m == NULL)
165 		goto freeit;
166 	icmplen = min(oiplen + 8, oip->ip_len);
167 	m->m_len = icmplen + ICMP_MINLEN;
168 	MH_ALIGN(m, m->m_len);
169 	icp = mtod(m, struct icmp *);
170 	if ((u_int)type > ICMP_MAXTYPE)
171 		panic("icmp_error");
172 	icmpstat.icps_outhist[type]++;
173 	icp->icmp_type = type;
174 	if (type == ICMP_REDIRECT)
175 		icp->icmp_gwaddr.s_addr = dest;
176 	else {
177 		icp->icmp_void = 0;
178 		/*
179 		 * The following assignments assume an overlay with the
180 		 * zeroed icmp_void field.
181 		 */
182 		if (type == ICMP_PARAMPROB) {
183 			icp->icmp_pptr = code;
184 			code = 0;
185 		} else if (type == ICMP_UNREACH &&
186 			code == ICMP_UNREACH_NEEDFRAG && destifp) {
187 			icp->icmp_nextmtu = htons(destifp->if_mtu);
188 		}
189 	}
190 
191 	icp->icmp_code = code;
192 	bcopy((caddr_t)oip, (caddr_t)&icp->icmp_ip, icmplen);
193 	nip = &icp->icmp_ip;
194 
195 	/*
196 	 * Convert fields to network representation.
197 	 */
198 	HTONS(nip->ip_len);
199 	HTONS(nip->ip_off);
200 
201 	/*
202 	 * Now, copy old ip header (without options)
203 	 * in front of icmp message.
204 	 */
205 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
206 		panic("icmp len");
207 	m->m_data -= sizeof(struct ip);
208 	m->m_len += sizeof(struct ip);
209 	m->m_pkthdr.len = m->m_len;
210 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
211 	nip = mtod(m, struct ip *);
212 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
213 	nip->ip_len = m->m_len;
214 	nip->ip_vhl = IP_VHL_BORING;
215 	nip->ip_p = IPPROTO_ICMP;
216 	nip->ip_tos = 0;
217 	icmp_reflect(m);
218 
219 freeit:
220 	m_freem(n);
221 }
222 
223 static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET };
224 static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET };
225 static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET };
226 
227 /*
228  * Process a received ICMP message.
229  */
230 void
231 icmp_input(m, off, proto)
232 	register struct mbuf *m;
233 	int off, proto;
234 {
235 	int hlen = off;
236 	register struct icmp *icp;
237 	register struct ip *ip = mtod(m, struct ip *);
238 	int icmplen = ip->ip_len;
239 	register int i;
240 	struct in_ifaddr *ia;
241 	void (*ctlfunc) __P((int, struct sockaddr *, void *));
242 	int code;
243 
244 	/*
245 	 * Locate icmp structure in mbuf, and check
246 	 * that not corrupted and of at least minimum length.
247 	 */
248 #ifdef ICMPPRINTFS
249 	if (icmpprintfs) {
250 		char buf[4 * sizeof "123"];
251 		strcpy(buf, inet_ntoa(ip->ip_src));
252 		printf("icmp_input from %s to %s, len %d\n",
253 		       buf, inet_ntoa(ip->ip_dst), icmplen);
254 	}
255 #endif
256 	if (icmplen < ICMP_MINLEN) {
257 		icmpstat.icps_tooshort++;
258 		goto freeit;
259 	}
260 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
261 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
262 		icmpstat.icps_tooshort++;
263 		return;
264 	}
265 	ip = mtod(m, struct ip *);
266 	m->m_len -= hlen;
267 	m->m_data += hlen;
268 	icp = mtod(m, struct icmp *);
269 	if (in_cksum(m, icmplen)) {
270 		icmpstat.icps_checksum++;
271 		goto freeit;
272 	}
273 	m->m_len += hlen;
274 	m->m_data -= hlen;
275 
276 #if defined(NFAITH) && 0 < NFAITH
277 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
278 		/*
279 		 * Deliver very specific ICMP type only.
280 		 */
281 		switch (icp->icmp_type) {
282 		case ICMP_UNREACH:
283 		case ICMP_TIMXCEED:
284 			break;
285 		default:
286 			goto freeit;
287 		}
288 	}
289 #endif
290 
291 #ifdef ICMPPRINTFS
292 	if (icmpprintfs)
293 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
294 		    icp->icmp_code);
295 #endif
296 
297 #ifdef IPSEC
298 	/* drop it if it does not match the policy */
299 	/* XXX Is there meaning of check in here ? */
300 	if (ipsec4_in_reject(m, NULL)) {
301 		ipsecstat.in_polvio++;
302 		goto freeit;
303 	}
304 #endif
305 
306 	/*
307 	 * Message type specific processing.
308 	 */
309 	if (icp->icmp_type > ICMP_MAXTYPE)
310 		goto raw;
311 	icmpstat.icps_inhist[icp->icmp_type]++;
312 	code = icp->icmp_code;
313 	switch (icp->icmp_type) {
314 
315 	case ICMP_UNREACH:
316 		switch (code) {
317 			case ICMP_UNREACH_NET:
318 			case ICMP_UNREACH_HOST:
319 			case ICMP_UNREACH_PROTOCOL:
320 			case ICMP_UNREACH_PORT:
321 			case ICMP_UNREACH_SRCFAIL:
322 				code += PRC_UNREACH_NET;
323 				break;
324 
325 			case ICMP_UNREACH_NEEDFRAG:
326 				code = PRC_MSGSIZE;
327 				break;
328 
329 			case ICMP_UNREACH_NET_UNKNOWN:
330 			case ICMP_UNREACH_NET_PROHIB:
331 			case ICMP_UNREACH_TOSNET:
332 				code = PRC_UNREACH_NET;
333 				break;
334 
335 			case ICMP_UNREACH_HOST_UNKNOWN:
336 			case ICMP_UNREACH_ISOLATED:
337 			case ICMP_UNREACH_HOST_PROHIB:
338 			case ICMP_UNREACH_TOSHOST:
339 				code = PRC_UNREACH_HOST;
340 				break;
341 
342 			case ICMP_UNREACH_FILTER_PROHIB:
343 			case ICMP_UNREACH_HOST_PRECEDENCE:
344 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
345 				code = PRC_UNREACH_PORT;
346 				break;
347 
348 			default:
349 				goto badcode;
350 		}
351 		goto deliver;
352 
353 	case ICMP_TIMXCEED:
354 		if (code > 1)
355 			goto badcode;
356 		code += PRC_TIMXCEED_INTRANS;
357 		goto deliver;
358 
359 	case ICMP_PARAMPROB:
360 		if (code > 1)
361 			goto badcode;
362 		code = PRC_PARAMPROB;
363 		goto deliver;
364 
365 	case ICMP_SOURCEQUENCH:
366 		if (code)
367 			goto badcode;
368 		code = PRC_QUENCH;
369 	deliver:
370 		/*
371 		 * Problem with datagram; advise higher level routines.
372 		 */
373 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
374 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
375 			icmpstat.icps_badlen++;
376 			goto freeit;
377 		}
378 		NTOHS(icp->icmp_ip.ip_len);
379 		/* Discard ICMP's in response to multicast packets */
380 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
381 			goto badcode;
382 #ifdef ICMPPRINTFS
383 		if (icmpprintfs)
384 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
385 #endif
386 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
387 #if 1
388 		/*
389 		 * MTU discovery:
390 		 * If we got a needfrag and there is a host route to the
391 		 * original destination, and the MTU is not locked, then
392 		 * set the MTU in the route to the suggested new value
393 		 * (if given) and then notify as usual.  The ULPs will
394 		 * notice that the MTU has changed and adapt accordingly.
395 		 * If no new MTU was suggested, then we guess a new one
396 		 * less than the current value.  If the new MTU is
397 		 * unreasonably small (arbitrarily set at 296), then
398 		 * we reset the MTU to the interface value and enable the
399 		 * lock bit, indicating that we are no longer doing MTU
400 		 * discovery.
401 		 */
402 		if (code == PRC_MSGSIZE) {
403 			struct rtentry *rt;
404 			int mtu;
405 
406 			rt = rtalloc1((struct sockaddr *)&icmpsrc, 0,
407 				      RTF_CLONING | RTF_PRCLONING);
408 			if (rt && (rt->rt_flags & RTF_HOST)
409 			    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
410 				mtu = ntohs(icp->icmp_nextmtu);
411 				if (!mtu)
412 					mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
413 							  1);
414 #ifdef DEBUG_MTUDISC
415 				printf("MTU for %s reduced to %d\n",
416 					inet_ntoa(icmpsrc.sin_addr), mtu);
417 #endif
418 				if (mtu < 296) {
419 					/* rt->rt_rmx.rmx_mtu =
420 						rt->rt_ifp->if_mtu; */
421 					rt->rt_rmx.rmx_locks |= RTV_MTU;
422 				} else if (rt->rt_rmx.rmx_mtu > mtu) {
423 					rt->rt_rmx.rmx_mtu = mtu;
424 				}
425 			}
426 			if (rt)
427 				RTFREE(rt);
428 		}
429 
430 #endif
431 		/*
432 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
433 		 * notification to TCP layer.
434 		 */
435 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
436 		if (ctlfunc)
437 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
438 				   (void *)&icp->icmp_ip);
439 		break;
440 
441 	badcode:
442 		icmpstat.icps_badcode++;
443 		break;
444 
445 	case ICMP_ECHO:
446 		if (!icmpbmcastecho
447 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
448 			icmpstat.icps_bmcastecho++;
449 			break;
450 		}
451 		icp->icmp_type = ICMP_ECHOREPLY;
452 		goto reflect;
453 
454 	case ICMP_TSTAMP:
455 		if (!icmpbmcastecho
456 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
457 			icmpstat.icps_bmcasttstamp++;
458 			break;
459 		}
460 		if (icmplen < ICMP_TSLEN) {
461 			icmpstat.icps_badlen++;
462 			break;
463 		}
464 		icp->icmp_type = ICMP_TSTAMPREPLY;
465 		icp->icmp_rtime = iptime();
466 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
467 		goto reflect;
468 
469 	case ICMP_MASKREQ:
470 #define	satosin(sa)	((struct sockaddr_in *)(sa))
471 		if (icmpmaskrepl == 0)
472 			break;
473 		/*
474 		 * We are not able to respond with all ones broadcast
475 		 * unless we receive it over a point-to-point interface.
476 		 */
477 		if (icmplen < ICMP_MASKLEN)
478 			break;
479 		switch (ip->ip_dst.s_addr) {
480 
481 		case INADDR_BROADCAST:
482 		case INADDR_ANY:
483 			icmpdst.sin_addr = ip->ip_src;
484 			break;
485 
486 		default:
487 			icmpdst.sin_addr = ip->ip_dst;
488 		}
489 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
490 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
491 		if (ia == 0)
492 			break;
493 		if (ia->ia_ifp == 0)
494 			break;
495 		icp->icmp_type = ICMP_MASKREPLY;
496 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
497 		if (ip->ip_src.s_addr == 0) {
498 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
499 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
500 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
501 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
502 		}
503 reflect:
504 		ip->ip_len += hlen;	/* since ip_input deducts this */
505 		icmpstat.icps_reflect++;
506 		icmpstat.icps_outhist[icp->icmp_type]++;
507 		icmp_reflect(m);
508 		return;
509 
510 	case ICMP_REDIRECT:
511 		if (log_redirect) {
512 			u_long src, dst, gw;
513 
514 			src = ntohl(ip->ip_src.s_addr);
515 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
516 			gw = ntohl(icp->icmp_gwaddr.s_addr);
517 			printf("icmp redirect from %d.%d.%d.%d: "
518 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
519 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
520 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
521 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
522 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
523 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
524 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
525 		}
526 		if (drop_redirect)
527 			break;
528 		if (code > 3)
529 			goto badcode;
530 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
531 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
532 			icmpstat.icps_badlen++;
533 			break;
534 		}
535 		/*
536 		 * Short circuit routing redirects to force
537 		 * immediate change in the kernel's routing
538 		 * tables.  The message is also handed to anyone
539 		 * listening on a raw socket (e.g. the routing
540 		 * daemon for use in updating its tables).
541 		 */
542 		icmpgw.sin_addr = ip->ip_src;
543 		icmpdst.sin_addr = icp->icmp_gwaddr;
544 #ifdef	ICMPPRINTFS
545 		if (icmpprintfs) {
546 			char buf[4 * sizeof "123"];
547 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
548 
549 			printf("redirect dst %s to %s\n",
550 			       buf, inet_ntoa(icp->icmp_gwaddr));
551 		}
552 #endif
553 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
554 		rtredirect((struct sockaddr *)&icmpsrc,
555 		  (struct sockaddr *)&icmpdst,
556 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
557 		  (struct sockaddr *)&icmpgw, (struct rtentry **)0);
558 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
559 #ifdef IPSEC
560 		key_sa_routechange((struct sockaddr *)&icmpsrc);
561 #endif
562 		break;
563 
564 	/*
565 	 * No kernel processing for the following;
566 	 * just fall through to send to raw listener.
567 	 */
568 	case ICMP_ECHOREPLY:
569 	case ICMP_ROUTERADVERT:
570 	case ICMP_ROUTERSOLICIT:
571 	case ICMP_TSTAMPREPLY:
572 	case ICMP_IREQREPLY:
573 	case ICMP_MASKREPLY:
574 	default:
575 		break;
576 	}
577 
578 raw:
579 	rip_input(m, off, proto);
580 	return;
581 
582 freeit:
583 	m_freem(m);
584 }
585 
586 /*
587  * Reflect the ip packet back to the source
588  */
589 static void
590 icmp_reflect(m)
591 	struct mbuf *m;
592 {
593 	register struct ip *ip = mtod(m, struct ip *);
594 	register struct in_ifaddr *ia;
595 	struct in_addr t;
596 	struct mbuf *opts = 0;
597 	int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
598 
599 	if (!in_canforward(ip->ip_src) &&
600 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
601 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
602 		m_freem(m);	/* Bad return address */
603 		goto done;	/* Ip_output() will check for broadcast */
604 	}
605 	t = ip->ip_dst;
606 	ip->ip_dst = ip->ip_src;
607 	/*
608 	 * If the incoming packet was addressed directly to us,
609 	 * use dst as the src for the reply.  Otherwise (broadcast
610 	 * or anonymous), use the address which corresponds
611 	 * to the incoming interface.
612 	 */
613 	for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) {
614 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
615 			break;
616 		if (ia->ia_ifp && (ia->ia_ifp->if_flags & IFF_BROADCAST) &&
617 		    t.s_addr == satosin(&ia->ia_broadaddr)->sin_addr.s_addr)
618 			break;
619 	}
620 	icmpdst.sin_addr = t;
621 	if ((ia == (struct in_ifaddr *)0) && m->m_pkthdr.rcvif)
622 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
623 			(struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
624 	/*
625 	 * The following happens if the packet was not addressed to us,
626 	 * and was received on an interface with no IP address.
627 	 */
628 	if (ia == (struct in_ifaddr *)0)
629 		ia = in_ifaddrhead.tqh_first;
630 	t = IA_SIN(ia)->sin_addr;
631 	ip->ip_src = t;
632 	ip->ip_ttl = MAXTTL;
633 
634 	if (optlen > 0) {
635 		register u_char *cp;
636 		int opt, cnt;
637 		u_int len;
638 
639 		/*
640 		 * Retrieve any source routing from the incoming packet;
641 		 * add on any record-route or timestamp options.
642 		 */
643 		cp = (u_char *) (ip + 1);
644 		if ((opts = ip_srcroute()) == 0 &&
645 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
646 			opts->m_len = sizeof(struct in_addr);
647 			mtod(opts, struct in_addr *)->s_addr = 0;
648 		}
649 		if (opts) {
650 #ifdef ICMPPRINTFS
651 		    if (icmpprintfs)
652 			    printf("icmp_reflect optlen %d rt %d => ",
653 				optlen, opts->m_len);
654 #endif
655 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
656 			    opt = cp[IPOPT_OPTVAL];
657 			    if (opt == IPOPT_EOL)
658 				    break;
659 			    if (opt == IPOPT_NOP)
660 				    len = 1;
661 			    else {
662 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
663 					    break;
664 				    len = cp[IPOPT_OLEN];
665 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
666 				        len > cnt)
667 					    break;
668 			    }
669 			    /*
670 			     * Should check for overflow, but it "can't happen"
671 			     */
672 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
673 				opt == IPOPT_SECURITY) {
674 				    bcopy((caddr_t)cp,
675 					mtod(opts, caddr_t) + opts->m_len, len);
676 				    opts->m_len += len;
677 			    }
678 		    }
679 		    /* Terminate & pad, if necessary */
680 		    cnt = opts->m_len % 4;
681 		    if (cnt) {
682 			    for (; cnt < 4; cnt++) {
683 				    *(mtod(opts, caddr_t) + opts->m_len) =
684 					IPOPT_EOL;
685 				    opts->m_len++;
686 			    }
687 		    }
688 #ifdef ICMPPRINTFS
689 		    if (icmpprintfs)
690 			    printf("%d\n", opts->m_len);
691 #endif
692 		}
693 		/*
694 		 * Now strip out original options by copying rest of first
695 		 * mbuf's data back, and adjust the IP length.
696 		 */
697 		ip->ip_len -= optlen;
698 		ip->ip_vhl = IP_VHL_BORING;
699 		m->m_len -= optlen;
700 		if (m->m_flags & M_PKTHDR)
701 			m->m_pkthdr.len -= optlen;
702 		optlen += sizeof(struct ip);
703 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
704 			 (unsigned)(m->m_len - sizeof(struct ip)));
705 	}
706 	m->m_flags &= ~(M_BCAST|M_MCAST);
707 	icmp_send(m, opts);
708 done:
709 	if (opts)
710 		(void)m_free(opts);
711 }
712 
713 /*
714  * Send an icmp packet back to the ip level,
715  * after supplying a checksum.
716  */
717 static void
718 icmp_send(m, opts)
719 	register struct mbuf *m;
720 	struct mbuf *opts;
721 {
722 	register struct ip *ip = mtod(m, struct ip *);
723 	register int hlen;
724 	register struct icmp *icp;
725 	struct route ro;
726 
727 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
728 	m->m_data += hlen;
729 	m->m_len -= hlen;
730 	icp = mtod(m, struct icmp *);
731 	icp->icmp_cksum = 0;
732 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
733 	m->m_data -= hlen;
734 	m->m_len += hlen;
735 	m->m_pkthdr.rcvif = (struct ifnet *)0;
736 #ifdef ICMPPRINTFS
737 	if (icmpprintfs) {
738 		char buf[4 * sizeof "123"];
739 		strcpy(buf, inet_ntoa(ip->ip_dst));
740 		printf("icmp_send dst %s src %s\n",
741 		       buf, inet_ntoa(ip->ip_src));
742 	}
743 #endif
744 	bzero(&ro, sizeof ro);
745 	(void) ip_output(m, opts, &ro, 0, NULL);
746 	if (ro.ro_rt)
747 		RTFREE(ro.ro_rt);
748 }
749 
750 n_time
751 iptime()
752 {
753 	struct timeval atv;
754 	u_long t;
755 
756 	microtime(&atv);
757 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
758 	return (htonl(t));
759 }
760 
761 #if 1
762 /*
763  * Return the next larger or smaller MTU plateau (table from RFC 1191)
764  * given current value MTU.  If DIR is less than zero, a larger plateau
765  * is returned; otherwise, a smaller value is returned.
766  */
767 static int
768 ip_next_mtu(mtu, dir)
769 	int mtu;
770 	int dir;
771 {
772 	static int mtutab[] = {
773 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
774 		68, 0
775 	};
776 	int i;
777 
778 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
779 		if (mtu >= mtutab[i])
780 			break;
781 	}
782 
783 	if (dir < 0) {
784 		if (i == 0) {
785 			return 0;
786 		} else {
787 			return mtutab[i - 1];
788 		}
789 	} else {
790 		if (mtutab[i] == 0) {
791 			return 0;
792 		} else if(mtu > mtutab[i]) {
793 			return mtutab[i];
794 		} else {
795 			return mtutab[i + 1];
796 		}
797 	}
798 }
799 #endif
800 
801 
802 /*
803  * badport_bandlim() - check for ICMP bandwidth limit
804  *
805  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
806  *	hit our bandwidth limit and it is not ok.
807  *
808  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
809  *
810  *	For now we separate the TCP and UDP subsystems w/ different 'which'
811  *	values.  We may eventually remove this separation (and simplify the
812  *	code further).
813  *
814  *	Note that the printing of the error message is delayed so we can
815  *	properly print the icmp error rate that the system was trying to do
816  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
817  *	the 'final' error, but it doesn't make sense to solve the printing
818  *	delay with more complex code.
819  */
820 
821 int
822 badport_bandlim(int which)
823 {
824 	static int lticks[2];
825 	static int lpackets[2];
826 	int dticks;
827 
828 	/*
829 	 * Return ok status if feature disabled or argument out of
830 	 * ranage.
831 	 */
832 
833 	if (icmplim <= 0 || which >= 2 || which < 0)
834 		return(0);
835 	dticks = ticks - lticks[which];
836 
837 	/*
838 	 * reset stats when cumulative dt exceeds one second.
839 	 */
840 
841 	if ((unsigned int)dticks > hz) {
842 		if (lpackets[which] > icmplim && icmplim_output) {
843 			printf("icmp-response bandwidth limit %d/%d pps\n",
844 				lpackets[which],
845 				icmplim
846 			);
847 		}
848 		lticks[which] = ticks;
849 		lpackets[which] = 0;
850 	}
851 
852 	/*
853 	 * bump packet count
854 	 */
855 
856 	if (++lpackets[which] > icmplim) {
857 		return(-1);
858 	}
859 	return(0);
860 }
861 
862