xref: /freebsd/sys/netinet/ip_icmp.c (revision 23f282aa31e9b6fceacd449020e936e98d6f2298)
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 /*
71  * ICMP routines: error generation, receive packet processing, and
72  * routines to turnaround packets back to the originator, and
73  * host table maintenance routines.
74  */
75 
76 static struct	icmpstat icmpstat;
77 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RD,
78 	&icmpstat, icmpstat, "");
79 
80 static int	icmpmaskrepl = 0;
81 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
82 	&icmpmaskrepl, 0, "");
83 
84 static int	drop_redirect = 0;
85 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
86 	&drop_redirect, 0, "");
87 
88 static int	log_redirect = 0;
89 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
90 	&log_redirect, 0, "");
91 
92 #ifdef ICMP_BANDLIM
93 
94 /*
95  * ICMP error-response bandwidth limiting sysctl.  If not enabled, sysctl
96  *      variable content is -1 and read-only.
97  */
98 
99 static int      icmplim = 200;
100 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
101 	&icmplim, 0, "");
102 #else
103 
104 static int      icmplim = -1;
105 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
106 	&icmplim, 0, "");
107 
108 #endif
109 
110 /*
111  * ICMP broadcast echo sysctl
112  */
113 
114 static int	icmpbmcastecho = 0;
115 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
116 	&icmpbmcastecho, 0, "");
117 
118 
119 #ifdef ICMPPRINTFS
120 int	icmpprintfs = 0;
121 #endif
122 
123 static void	icmp_reflect __P((struct mbuf *));
124 static void	icmp_send __P((struct mbuf *, struct mbuf *));
125 static int	ip_next_mtu __P((int, int));
126 
127 extern	struct protosw inetsw[];
128 
129 /*
130  * Generate an error packet of type error
131  * in response to bad packet ip.
132  */
133 void
134 icmp_error(n, type, code, dest, destifp)
135 	struct mbuf *n;
136 	int type, code;
137 	n_long dest;
138 	struct ifnet *destifp;
139 {
140 	register struct ip *oip = mtod(n, struct ip *), *nip;
141 	register unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
142 	register struct icmp *icp;
143 	register struct mbuf *m;
144 	unsigned icmplen;
145 
146 #ifdef ICMPPRINTFS
147 	if (icmpprintfs)
148 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
149 #endif
150 	if (type != ICMP_REDIRECT)
151 		icmpstat.icps_error++;
152 	/*
153 	 * Don't send error if not the first fragment of message.
154 	 * Don't error if the old packet protocol was ICMP
155 	 * error message, only known informational types.
156 	 */
157 	if (oip->ip_off &~ (IP_MF|IP_DF))
158 		goto freeit;
159 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
160 	  n->m_len >= oiplen + ICMP_MINLEN &&
161 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
162 		icmpstat.icps_oldicmp++;
163 		goto freeit;
164 	}
165 	/* Don't send error in response to a multicast or broadcast packet */
166 	if (n->m_flags & (M_BCAST|M_MCAST))
167 		goto freeit;
168 	/*
169 	 * First, formulate icmp message
170 	 */
171 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
172 	if (m == NULL)
173 		goto freeit;
174 	icmplen = oiplen + min(8, oip->ip_len);
175 	m->m_len = icmplen + ICMP_MINLEN;
176 	MH_ALIGN(m, m->m_len);
177 	icp = mtod(m, struct icmp *);
178 	if ((u_int)type > ICMP_MAXTYPE)
179 		panic("icmp_error");
180 	icmpstat.icps_outhist[type]++;
181 	icp->icmp_type = type;
182 	if (type == ICMP_REDIRECT)
183 		icp->icmp_gwaddr.s_addr = dest;
184 	else {
185 		icp->icmp_void = 0;
186 		/*
187 		 * The following assignments assume an overlay with the
188 		 * zeroed icmp_void field.
189 		 */
190 		if (type == ICMP_PARAMPROB) {
191 			icp->icmp_pptr = code;
192 			code = 0;
193 		} else if (type == ICMP_UNREACH &&
194 			code == ICMP_UNREACH_NEEDFRAG && destifp) {
195 			icp->icmp_nextmtu = htons(destifp->if_mtu);
196 		}
197 	}
198 
199 	icp->icmp_code = code;
200 	bcopy((caddr_t)oip, (caddr_t)&icp->icmp_ip, icmplen);
201 	nip = &icp->icmp_ip;
202 	nip->ip_len = htons((u_short)(nip->ip_len + oiplen));
203 
204 	/*
205 	 * Now, copy old ip header (without options)
206 	 * in front of icmp message.
207 	 */
208 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
209 		panic("icmp len");
210 	m->m_data -= sizeof(struct ip);
211 	m->m_len += sizeof(struct ip);
212 	m->m_pkthdr.len = m->m_len;
213 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
214 	nip = mtod(m, struct ip *);
215 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
216 	nip->ip_len = m->m_len;
217 	nip->ip_vhl = IP_VHL_BORING;
218 	nip->ip_p = IPPROTO_ICMP;
219 	nip->ip_tos = 0;
220 	icmp_reflect(m);
221 
222 freeit:
223 	m_freem(n);
224 }
225 
226 static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET };
227 static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET };
228 static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET };
229 
230 /*
231  * Process a received ICMP message.
232  */
233 void
234 icmp_input(m, off, proto)
235 	register struct mbuf *m;
236 	int off, proto;
237 {
238 	int hlen = off;
239 	register struct icmp *icp;
240 	register struct ip *ip = mtod(m, struct ip *);
241 	int icmplen = ip->ip_len;
242 	register int i;
243 	struct in_ifaddr *ia;
244 	void (*ctlfunc) __P((int, struct sockaddr *, void *));
245 	int code;
246 
247 	/*
248 	 * Locate icmp structure in mbuf, and check
249 	 * that not corrupted and of at least minimum length.
250 	 */
251 #ifdef ICMPPRINTFS
252 	if (icmpprintfs) {
253 		char buf[4 * sizeof "123"];
254 		strcpy(buf, inet_ntoa(ip->ip_src));
255 		printf("icmp_input from %s to %s, len %d\n",
256 		       buf, inet_ntoa(ip->ip_dst), icmplen);
257 	}
258 #endif
259 	if (icmplen < ICMP_MINLEN) {
260 		icmpstat.icps_tooshort++;
261 		goto freeit;
262 	}
263 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
264 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
265 		icmpstat.icps_tooshort++;
266 		return;
267 	}
268 	ip = mtod(m, struct ip *);
269 	m->m_len -= hlen;
270 	m->m_data += hlen;
271 	icp = mtod(m, struct icmp *);
272 	if (in_cksum(m, icmplen)) {
273 		icmpstat.icps_checksum++;
274 		goto freeit;
275 	}
276 	m->m_len += hlen;
277 	m->m_data -= hlen;
278 
279 #if defined(NFAITH) && 0 < NFAITH
280 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
281 		/*
282 		 * Deliver very specific ICMP type only.
283 		 */
284 		switch (icp->icmp_type) {
285 		case ICMP_UNREACH:
286 		case ICMP_TIMXCEED:
287 			break;
288 		default:
289 			goto freeit;
290 		}
291 	}
292 #endif
293 
294 #ifdef ICMPPRINTFS
295 	if (icmpprintfs)
296 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
297 		    icp->icmp_code);
298 #endif
299 
300 #ifdef IPSEC
301 	/* drop it if it does not match the policy */
302 	/* XXX Is there meaning of check in here ? */
303 	if (ipsec4_in_reject(m, NULL)) {
304 		ipsecstat.in_polvio++;
305 		goto freeit;
306 	}
307 #endif
308 
309 	/*
310 	 * Message type specific processing.
311 	 */
312 	if (icp->icmp_type > ICMP_MAXTYPE)
313 		goto raw;
314 	icmpstat.icps_inhist[icp->icmp_type]++;
315 	code = icp->icmp_code;
316 	switch (icp->icmp_type) {
317 
318 	case ICMP_UNREACH:
319 		switch (code) {
320 			case ICMP_UNREACH_NET:
321 			case ICMP_UNREACH_HOST:
322 			case ICMP_UNREACH_PROTOCOL:
323 			case ICMP_UNREACH_PORT:
324 			case ICMP_UNREACH_SRCFAIL:
325 				code += PRC_UNREACH_NET;
326 				break;
327 
328 			case ICMP_UNREACH_NEEDFRAG:
329 				code = PRC_MSGSIZE;
330 				break;
331 
332 			case ICMP_UNREACH_NET_UNKNOWN:
333 			case ICMP_UNREACH_NET_PROHIB:
334 			case ICMP_UNREACH_TOSNET:
335 				code = PRC_UNREACH_NET;
336 				break;
337 
338 			case ICMP_UNREACH_HOST_UNKNOWN:
339 			case ICMP_UNREACH_ISOLATED:
340 			case ICMP_UNREACH_HOST_PROHIB:
341 			case ICMP_UNREACH_TOSHOST:
342 				code = PRC_UNREACH_HOST;
343 				break;
344 
345 			case ICMP_UNREACH_FILTER_PROHIB:
346 			case ICMP_UNREACH_HOST_PRECEDENCE:
347 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
348 				code = PRC_UNREACH_PORT;
349 				break;
350 
351 			default:
352 				goto badcode;
353 		}
354 		goto deliver;
355 
356 	case ICMP_TIMXCEED:
357 		if (code > 1)
358 			goto badcode;
359 		code += PRC_TIMXCEED_INTRANS;
360 		goto deliver;
361 
362 	case ICMP_PARAMPROB:
363 		if (code > 1)
364 			goto badcode;
365 		code = PRC_PARAMPROB;
366 		goto deliver;
367 
368 	case ICMP_SOURCEQUENCH:
369 		if (code)
370 			goto badcode;
371 		code = PRC_QUENCH;
372 	deliver:
373 		/*
374 		 * Problem with datagram; advise higher level routines.
375 		 */
376 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
377 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
378 			icmpstat.icps_badlen++;
379 			goto freeit;
380 		}
381 		NTOHS(icp->icmp_ip.ip_len);
382 		/* Discard ICMP's in response to multicast packets */
383 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
384 			goto badcode;
385 #ifdef ICMPPRINTFS
386 		if (icmpprintfs)
387 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
388 #endif
389 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
390 #if 1
391 		/*
392 		 * MTU discovery:
393 		 * If we got a needfrag and there is a host route to the
394 		 * original destination, and the MTU is not locked, then
395 		 * set the MTU in the route to the suggested new value
396 		 * (if given) and then notify as usual.  The ULPs will
397 		 * notice that the MTU has changed and adapt accordingly.
398 		 * If no new MTU was suggested, then we guess a new one
399 		 * less than the current value.  If the new MTU is
400 		 * unreasonably small (arbitrarily set at 296), then
401 		 * we reset the MTU to the interface value and enable the
402 		 * lock bit, indicating that we are no longer doing MTU
403 		 * discovery.
404 		 */
405 		if (code == PRC_MSGSIZE) {
406 			struct rtentry *rt;
407 			int mtu;
408 
409 			rt = rtalloc1((struct sockaddr *)&icmpsrc, 0,
410 				      RTF_CLONING | RTF_PRCLONING);
411 			if (rt && (rt->rt_flags & RTF_HOST)
412 			    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
413 				mtu = ntohs(icp->icmp_nextmtu);
414 				if (!mtu)
415 					mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
416 							  1);
417 #ifdef DEBUG_MTUDISC
418 				printf("MTU for %s reduced to %d\n",
419 					inet_ntoa(icmpsrc.sin_addr), mtu);
420 #endif
421 				if (mtu < 296) {
422 					/* rt->rt_rmx.rmx_mtu =
423 						rt->rt_ifp->if_mtu; */
424 					rt->rt_rmx.rmx_locks |= RTV_MTU;
425 				} else if (rt->rt_rmx.rmx_mtu > mtu) {
426 					rt->rt_rmx.rmx_mtu = mtu;
427 				}
428 			}
429 			if (rt)
430 				RTFREE(rt);
431 		}
432 
433 #endif
434 		/*
435 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
436 		 * notification to TCP layer.
437 		 */
438 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
439 		if (ctlfunc)
440 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
441 				   (void *)&icp->icmp_ip);
442 		break;
443 
444 	badcode:
445 		icmpstat.icps_badcode++;
446 		break;
447 
448 	case ICMP_ECHO:
449 		if (!icmpbmcastecho
450 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
451 			icmpstat.icps_bmcastecho++;
452 			break;
453 		}
454 		icp->icmp_type = ICMP_ECHOREPLY;
455 		goto reflect;
456 
457 	case ICMP_TSTAMP:
458 		if (!icmpbmcastecho
459 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
460 			icmpstat.icps_bmcasttstamp++;
461 			break;
462 		}
463 		if (icmplen < ICMP_TSLEN) {
464 			icmpstat.icps_badlen++;
465 			break;
466 		}
467 		icp->icmp_type = ICMP_TSTAMPREPLY;
468 		icp->icmp_rtime = iptime();
469 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
470 		goto reflect;
471 
472 	case ICMP_MASKREQ:
473 #define	satosin(sa)	((struct sockaddr_in *)(sa))
474 		if (icmpmaskrepl == 0)
475 			break;
476 		/*
477 		 * We are not able to respond with all ones broadcast
478 		 * unless we receive it over a point-to-point interface.
479 		 */
480 		if (icmplen < ICMP_MASKLEN)
481 			break;
482 		switch (ip->ip_dst.s_addr) {
483 
484 		case INADDR_BROADCAST:
485 		case INADDR_ANY:
486 			icmpdst.sin_addr = ip->ip_src;
487 			break;
488 
489 		default:
490 			icmpdst.sin_addr = ip->ip_dst;
491 		}
492 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
493 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
494 		if (ia == 0)
495 			break;
496 		if (ia->ia_ifp == 0)
497 			break;
498 		icp->icmp_type = ICMP_MASKREPLY;
499 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
500 		if (ip->ip_src.s_addr == 0) {
501 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
502 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
503 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
504 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
505 		}
506 reflect:
507 		ip->ip_len += hlen;	/* since ip_input deducts this */
508 		icmpstat.icps_reflect++;
509 		icmpstat.icps_outhist[icp->icmp_type]++;
510 		icmp_reflect(m);
511 		return;
512 
513 	case ICMP_REDIRECT:
514 		if (log_redirect) {
515 			u_long src, dst, gw;
516 
517 			src = ntohl(ip->ip_src.s_addr);
518 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
519 			gw = ntohl(icp->icmp_gwaddr.s_addr);
520 			printf("icmp redirect from %d.%d.%d.%d: "
521 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
522 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
523 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
524 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
525 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
526 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
527 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
528 		}
529 		if (drop_redirect)
530 			break;
531 		if (code > 3)
532 			goto badcode;
533 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
534 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
535 			icmpstat.icps_badlen++;
536 			break;
537 		}
538 		/*
539 		 * Short circuit routing redirects to force
540 		 * immediate change in the kernel's routing
541 		 * tables.  The message is also handed to anyone
542 		 * listening on a raw socket (e.g. the routing
543 		 * daemon for use in updating its tables).
544 		 */
545 		icmpgw.sin_addr = ip->ip_src;
546 		icmpdst.sin_addr = icp->icmp_gwaddr;
547 #ifdef	ICMPPRINTFS
548 		if (icmpprintfs) {
549 			char buf[4 * sizeof "123"];
550 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
551 
552 			printf("redirect dst %s to %s\n",
553 			       buf, inet_ntoa(icp->icmp_gwaddr));
554 		}
555 #endif
556 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
557 		rtredirect((struct sockaddr *)&icmpsrc,
558 		  (struct sockaddr *)&icmpdst,
559 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
560 		  (struct sockaddr *)&icmpgw, (struct rtentry **)0);
561 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
562 #ifdef IPSEC
563 		key_sa_routechange((struct sockaddr *)&icmpsrc);
564 #endif
565 		break;
566 
567 	/*
568 	 * No kernel processing for the following;
569 	 * just fall through to send to raw listener.
570 	 */
571 	case ICMP_ECHOREPLY:
572 	case ICMP_ROUTERADVERT:
573 	case ICMP_ROUTERSOLICIT:
574 	case ICMP_TSTAMPREPLY:
575 	case ICMP_IREQREPLY:
576 	case ICMP_MASKREPLY:
577 	default:
578 		break;
579 	}
580 
581 raw:
582 	rip_input(m, off, proto);
583 	return;
584 
585 freeit:
586 	m_freem(m);
587 }
588 
589 /*
590  * Reflect the ip packet back to the source
591  */
592 static void
593 icmp_reflect(m)
594 	struct mbuf *m;
595 {
596 	register struct ip *ip = mtod(m, struct ip *);
597 	register struct in_ifaddr *ia;
598 	struct in_addr t;
599 	struct mbuf *opts = 0;
600 	int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
601 
602 	if (!in_canforward(ip->ip_src) &&
603 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
604 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
605 		m_freem(m);	/* Bad return address */
606 		goto done;	/* Ip_output() will check for broadcast */
607 	}
608 	t = ip->ip_dst;
609 	ip->ip_dst = ip->ip_src;
610 	/*
611 	 * If the incoming packet was addressed directly to us,
612 	 * use dst as the src for the reply.  Otherwise (broadcast
613 	 * or anonymous), use the address which corresponds
614 	 * to the incoming interface.
615 	 */
616 	for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) {
617 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
618 			break;
619 		if (ia->ia_ifp && (ia->ia_ifp->if_flags & IFF_BROADCAST) &&
620 		    t.s_addr == satosin(&ia->ia_broadaddr)->sin_addr.s_addr)
621 			break;
622 	}
623 	icmpdst.sin_addr = t;
624 	if ((ia == (struct in_ifaddr *)0) && m->m_pkthdr.rcvif)
625 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
626 			(struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
627 	/*
628 	 * The following happens if the packet was not addressed to us,
629 	 * and was received on an interface with no IP address.
630 	 */
631 	if (ia == (struct in_ifaddr *)0)
632 		ia = in_ifaddrhead.tqh_first;
633 	t = IA_SIN(ia)->sin_addr;
634 	ip->ip_src = t;
635 	ip->ip_ttl = MAXTTL;
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 				    len = cp[IPOPT_OLEN];
666 				    if (len <= 0 || 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 #ifdef ICMP_BANDLIM
802 
803 /*
804  * badport_bandlim() - check for ICMP bandwidth limit
805  *
806  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
807  *	hit our bandwidth limit and it is not ok.
808  *
809  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
810  *
811  *	For now we separate the TCP and UDP subsystems w/ different 'which'
812  *	values.  We may eventually remove this separation (and simplify the
813  *	code further).
814  *
815  *	Note that the printing of the error message is delayed so we can
816  *	properly print the icmp error rate that the system was trying to do
817  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
818  *	the 'final' error, but it doesn't make sense to solve the printing
819  *	delay with more complex code.
820  */
821 
822 int
823 badport_bandlim(int which)
824 {
825 	static int lticks[2];
826 	static int lpackets[2];
827 	int dticks;
828 
829 	/*
830 	 * Return ok status if feature disabled or argument out of
831 	 * ranage.
832 	 */
833 
834 	if (icmplim <= 0 || which >= 2 || which < 0)
835 		return(0);
836 	dticks = ticks - lticks[which];
837 
838 	/*
839 	 * reset stats when cumulative dt exceeds one second.
840 	 */
841 
842 	if ((unsigned int)dticks > hz) {
843 		if (lpackets[which] > icmplim) {
844 			printf("icmp-response bandwidth limit %d/%d pps\n",
845 				lpackets[which],
846 				icmplim
847 			);
848 		}
849 		lticks[which] = ticks;
850 		lpackets[which] = 0;
851 	}
852 
853 	/*
854 	 * bump packet count
855 	 */
856 
857 	if (++lpackets[which] > icmplim) {
858 		return(-1);
859 	}
860 	return(0);
861 }
862 
863 #endif
864 
865 
866