xref: /dragonfly/sys/netinet/ip_icmp.c (revision 5de36205)
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: src/sys/netinet/ip_icmp.c,v 1.39.2.19 2003/01/24 05:11:34 sam Exp $
35  * $DragonFly: src/sys/netinet/ip_icmp.c,v 1.22 2005/03/06 05:09:25 hsu Exp $
36  */
37 
38 #include "opt_ipsec.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/mbuf.h>
43 #include <sys/protosw.h>
44 #include <sys/socket.h>
45 #include <sys/time.h>
46 #include <sys/kernel.h>
47 #include <sys/sysctl.h>
48 #include <sys/in_cksum.h>
49 
50 #include <machine/stdarg.h>
51 
52 #include <net/if.h>
53 #include <net/if_types.h>
54 #include <net/route.h>
55 
56 #define _IP_VHL
57 #include <netinet/in.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip.h>
61 #include <netinet/ip_icmp.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/icmp_var.h>
64 
65 #ifdef IPSEC
66 #include <netinet6/ipsec.h>
67 #include <netproto/key/key.h>
68 #endif
69 
70 #ifdef FAST_IPSEC
71 #include <netproto/ipsec/ipsec.h>
72 #include <netproto/ipsec/key.h>
73 #define	IPSEC
74 #endif
75 
76 /*
77  * ICMP routines: error generation, receive packet processing, and
78  * routines to turnaround packets back to the originator, and
79  * host table maintenance routines.
80  */
81 
82 struct icmpstat icmpstat;
83 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
84 	&icmpstat, icmpstat, "");
85 
86 static int	icmpmaskrepl = 0;
87 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
88 	&icmpmaskrepl, 0, "");
89 
90 static int	drop_redirect = 0;
91 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
92 	&drop_redirect, 0, "");
93 
94 static int	log_redirect = 0;
95 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
96 	&log_redirect, 0, "");
97 
98 #ifdef ICMP_BANDLIM
99 
100 /*
101  * ICMP error-response bandwidth limiting sysctl.  If not enabled, sysctl
102  *      variable content is -1 and read-only.
103  */
104 
105 static int      icmplim = 200;
106 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
107 	&icmplim, 0, "");
108 #else
109 
110 static int      icmplim = -1;
111 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
112 	&icmplim, 0, "");
113 
114 #endif
115 
116 static int	icmplim_output = 1;
117 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
118 	&icmplim_output, 0, "");
119 
120 /*
121  * ICMP broadcast echo sysctl
122  */
123 
124 static int	icmpbmcastecho = 0;
125 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
126 	&icmpbmcastecho, 0, "");
127 
128 
129 #ifdef ICMPPRINTFS
130 int	icmpprintfs = 0;
131 #endif
132 
133 static void	icmp_reflect (struct mbuf *);
134 static void	icmp_send (struct mbuf *, struct mbuf *, struct route *);
135 
136 extern	struct protosw inetsw[];
137 
138 /*
139  * Generate an error packet of type error
140  * in response to bad packet ip.
141  */
142 void
143 icmp_error(n, type, code, dest, destifp)
144 	struct mbuf *n;
145 	int type, code;
146 	n_long dest;
147 	struct ifnet *destifp;
148 {
149 	struct ip *oip = mtod(n, struct ip *), *nip;
150 	unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
151 	struct icmp *icp;
152 	struct mbuf *m;
153 	unsigned icmplen;
154 
155 #ifdef ICMPPRINTFS
156 	if (icmpprintfs)
157 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
158 #endif
159 	if (type != ICMP_REDIRECT)
160 		icmpstat.icps_error++;
161 	/*
162 	 * Don't send error if the original packet was encrypted.
163 	 * Don't send error if not the first fragment of message.
164 	 * Don't error if the old packet protocol was ICMP
165 	 * error message, only known informational types.
166 	 */
167 	if (n->m_flags & M_DECRYPTED)
168 		goto freeit;
169 	if (oip->ip_off &~ (IP_MF|IP_DF))
170 		goto freeit;
171 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
172 	  n->m_len >= oiplen + ICMP_MINLEN &&
173 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
174 		icmpstat.icps_oldicmp++;
175 		goto freeit;
176 	}
177 	/* Don't send error in response to a multicast or broadcast packet */
178 	if (n->m_flags & (M_BCAST|M_MCAST))
179 		goto freeit;
180 	/*
181 	 * First, formulate icmp message
182 	 */
183 	m = m_gethdr(MB_DONTWAIT, MT_HEADER);
184 	if (m == NULL)
185 		goto freeit;
186 	icmplen = min(oiplen + 8, oip->ip_len);
187 	if (icmplen < sizeof(struct ip))
188 		panic("icmp_error: bad length");
189 	m->m_len = icmplen + ICMP_MINLEN;
190 	MH_ALIGN(m, m->m_len);
191 	icp = mtod(m, struct icmp *);
192 	if ((u_int)type > ICMP_MAXTYPE)
193 		panic("icmp_error");
194 	icmpstat.icps_outhist[type]++;
195 	icp->icmp_type = type;
196 	if (type == ICMP_REDIRECT)
197 		icp->icmp_gwaddr.s_addr = dest;
198 	else {
199 		icp->icmp_void = 0;
200 		/*
201 		 * The following assignments assume an overlay with the
202 		 * zeroed icmp_void field.
203 		 */
204 		if (type == ICMP_PARAMPROB) {
205 			icp->icmp_pptr = code;
206 			code = 0;
207 		} else if (type == ICMP_UNREACH &&
208 			code == ICMP_UNREACH_NEEDFRAG && destifp) {
209 			icp->icmp_nextmtu = htons(destifp->if_mtu);
210 		}
211 	}
212 
213 	icp->icmp_code = code;
214 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
215 	nip = &icp->icmp_ip;
216 
217 	/*
218 	 * Convert fields to network representation.
219 	 */
220 	nip->ip_len = htons(nip->ip_len);
221 	nip->ip_off = htons(nip->ip_off);
222 
223 	/*
224 	 * Now, copy old ip header (without options)
225 	 * in front of icmp message.
226 	 */
227 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
228 		panic("icmp len");
229 	m->m_data -= sizeof(struct ip);
230 	m->m_len += sizeof(struct ip);
231 	m->m_pkthdr.len = m->m_len;
232 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
233 	nip = mtod(m, struct ip *);
234 	bcopy(oip, nip, sizeof(struct ip));
235 	nip->ip_len = m->m_len;
236 	nip->ip_vhl = IP_VHL_BORING;
237 	nip->ip_p = IPPROTO_ICMP;
238 	nip->ip_tos = 0;
239 	m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & PF_MBUF_GENERATED;
240 	icmp_reflect(m);
241 
242 freeit:
243 	m_freem(n);
244 }
245 
246 /*
247  * Process a received ICMP message.
248  */
249 void
250 icmp_input(struct mbuf *m, ...)
251 {
252 	struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
253 	struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
254 	struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
255 	struct icmp *icp;
256 	struct in_ifaddr *ia;
257 	struct ip *ip = mtod(m, struct ip *);
258 	void (*ctlfunc) (int, struct sockaddr *, void *);
259 	int icmplen = ip->ip_len;
260 	int i, hlen;
261 	int code, off, proto;
262 	__va_list ap;
263 
264 	__va_start(ap, m);
265 	off = __va_arg(ap, int);
266 	proto = __va_arg(ap, int);
267 	__va_end(ap);
268 
269 	hlen = off;
270 
271 	/*
272 	 * Locate icmp structure in mbuf, and check
273 	 * that not corrupted and of at least minimum length.
274 	 */
275 #ifdef ICMPPRINTFS
276 	if (icmpprintfs) {
277 		char buf[sizeof "aaa.bbb.ccc.ddd"];
278 
279 		strcpy(buf, inet_ntoa(ip->ip_src));
280 		printf("icmp_input from %s to %s, len %d\n",
281 		       buf, inet_ntoa(ip->ip_dst), icmplen);
282 	}
283 #endif
284 	if (icmplen < ICMP_MINLEN) {
285 		icmpstat.icps_tooshort++;
286 		goto freeit;
287 	}
288 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
289 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
290 		icmpstat.icps_tooshort++;
291 		return;
292 	}
293 	ip = mtod(m, struct ip *);
294 	m->m_len -= hlen;
295 	m->m_data += hlen;
296 	icp = mtod(m, struct icmp *);
297 	if (in_cksum(m, icmplen)) {
298 		icmpstat.icps_checksum++;
299 		goto freeit;
300 	}
301 	m->m_len += hlen;
302 	m->m_data -= hlen;
303 
304 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
305 		/*
306 		 * Deliver very specific ICMP type only.
307 		 */
308 		switch (icp->icmp_type) {
309 		case ICMP_UNREACH:
310 		case ICMP_TIMXCEED:
311 			break;
312 		default:
313 			goto freeit;
314 		}
315 	}
316 
317 #ifdef ICMPPRINTFS
318 	if (icmpprintfs)
319 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
320 		    icp->icmp_code);
321 #endif
322 
323 	/*
324 	 * Message type specific processing.
325 	 */
326 	if (icp->icmp_type > ICMP_MAXTYPE)
327 		goto raw;
328 	icmpstat.icps_inhist[icp->icmp_type]++;
329 	code = icp->icmp_code;
330 	switch (icp->icmp_type) {
331 
332 	case ICMP_UNREACH:
333 		switch (code) {
334 			case ICMP_UNREACH_NET:
335 			case ICMP_UNREACH_HOST:
336 			case ICMP_UNREACH_SRCFAIL:
337 			case ICMP_UNREACH_NET_UNKNOWN:
338 			case ICMP_UNREACH_HOST_UNKNOWN:
339 			case ICMP_UNREACH_ISOLATED:
340 			case ICMP_UNREACH_TOSNET:
341 			case ICMP_UNREACH_TOSHOST:
342 			case ICMP_UNREACH_HOST_PRECEDENCE:
343 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
344 				code = PRC_UNREACH_NET;
345 				break;
346 
347 			case ICMP_UNREACH_NEEDFRAG:
348 				code = PRC_MSGSIZE;
349 				break;
350 
351 			/*
352 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
353 			 * Treat subcodes 2,3 as immediate RST
354 			 */
355 			case ICMP_UNREACH_PROTOCOL:
356 			case ICMP_UNREACH_PORT:
357 				code = PRC_UNREACH_PORT;
358 				break;
359 
360 			case ICMP_UNREACH_NET_PROHIB:
361 			case ICMP_UNREACH_HOST_PROHIB:
362 			case ICMP_UNREACH_FILTER_PROHIB:
363 				code = PRC_UNREACH_ADMIN_PROHIB;
364 				break;
365 
366 			default:
367 				goto badcode;
368 		}
369 		goto deliver;
370 
371 	case ICMP_TIMXCEED:
372 		if (code > 1)
373 			goto badcode;
374 		code += PRC_TIMXCEED_INTRANS;
375 		goto deliver;
376 
377 	case ICMP_PARAMPROB:
378 		if (code > 1)
379 			goto badcode;
380 		code = PRC_PARAMPROB;
381 		goto deliver;
382 
383 	case ICMP_SOURCEQUENCH:
384 		if (code)
385 			goto badcode;
386 		code = PRC_QUENCH;
387 deliver:
388 		/*
389 		 * Problem with datagram; advise higher level routines.
390 		 */
391 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
392 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
393 			icmpstat.icps_badlen++;
394 			goto freeit;
395 		}
396 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
397 		/* Discard ICMP's in response to multicast packets */
398 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
399 			goto badcode;
400 #ifdef ICMPPRINTFS
401 		if (icmpprintfs)
402 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
403 #endif
404 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
405 #if 1
406 		/*
407 		 * MTU discovery:
408 		 * If we got a needfrag and there is a host route to the
409 		 * original destination, and the MTU is not locked, then
410 		 * set the MTU in the route to the suggested new value
411 		 * (if given) and then notify as usual.  The ULPs will
412 		 * notice that the MTU has changed and adapt accordingly.
413 		 * If no new MTU was suggested, then we guess a new one
414 		 * less than the current value.  If the new MTU is
415 		 * unreasonably small (arbitrarily set at 296), then
416 		 * we reset the MTU to the interface value and enable the
417 		 * lock bit, indicating that we are no longer doing MTU
418 		 * discovery.
419 		 */
420 		if (code == PRC_MSGSIZE) {
421 			struct rtentry *rt;
422 			int mtu;
423 
424 			rt = rtpurelookup((struct sockaddr *)&icmpsrc);
425 			if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
426 			    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
427 				mtu = ntohs(icp->icmp_nextmtu);
428 				if (!mtu)
429 					mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
430 							  1);
431 #ifdef DEBUG_MTUDISC
432 				printf("MTU for %s reduced to %d\n",
433 					inet_ntoa(icmpsrc.sin_addr), mtu);
434 #endif
435 				if (mtu < 296) {
436 					/* rt->rt_rmx.rmx_mtu =
437 						rt->rt_ifp->if_mtu; */
438 					rt->rt_rmx.rmx_locks |= RTV_MTU;
439 				} else if (rt->rt_rmx.rmx_mtu > mtu) {
440 					rt->rt_rmx.rmx_mtu = mtu;
441 				}
442 			}
443 			if (rt != NULL)
444 				--rt->rt_refcnt;
445 		}
446 #endif
447 		/*
448 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
449 		 * notification to TCP layer.
450 		 */
451 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
452 		if (ctlfunc)
453 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
454 				   &icp->icmp_ip);
455 		break;
456 
457 badcode:
458 		icmpstat.icps_badcode++;
459 		break;
460 
461 	case ICMP_ECHO:
462 		if (!icmpbmcastecho
463 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
464 			icmpstat.icps_bmcastecho++;
465 			break;
466 		}
467 		icp->icmp_type = ICMP_ECHOREPLY;
468 #ifdef ICMP_BANDLIM
469 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
470 			goto freeit;
471 		else
472 #endif
473 			goto reflect;
474 
475 	case ICMP_TSTAMP:
476 		if (!icmpbmcastecho
477 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
478 			icmpstat.icps_bmcasttstamp++;
479 			break;
480 		}
481 		if (icmplen < ICMP_TSLEN) {
482 			icmpstat.icps_badlen++;
483 			break;
484 		}
485 		icp->icmp_type = ICMP_TSTAMPREPLY;
486 		icp->icmp_rtime = iptime();
487 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
488 #ifdef ICMP_BANDLIM
489 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
490 			goto freeit;
491 		else
492 #endif
493 			goto reflect;
494 
495 	case ICMP_MASKREQ:
496 		if (icmpmaskrepl == 0)
497 			break;
498 		/*
499 		 * We are not able to respond with all ones broadcast
500 		 * unless we receive it over a point-to-point interface.
501 		 */
502 		if (icmplen < ICMP_MASKLEN)
503 			break;
504 		switch (ip->ip_dst.s_addr) {
505 
506 		case INADDR_BROADCAST:
507 		case INADDR_ANY:
508 			icmpdst.sin_addr = ip->ip_src;
509 			break;
510 
511 		default:
512 			icmpdst.sin_addr = ip->ip_dst;
513 		}
514 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
515 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
516 		if (ia == 0)
517 			break;
518 		if (ia->ia_ifp == 0)
519 			break;
520 		icp->icmp_type = ICMP_MASKREPLY;
521 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
522 		if (ip->ip_src.s_addr == 0) {
523 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
524 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
525 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
526 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
527 		}
528 reflect:
529 		ip->ip_len += hlen;	/* since ip_input deducts this */
530 		icmpstat.icps_reflect++;
531 		icmpstat.icps_outhist[icp->icmp_type]++;
532 		icmp_reflect(m);
533 		return;
534 
535 	case ICMP_REDIRECT:
536 		if (log_redirect) {
537 			u_long src, dst, gw;
538 
539 			src = ntohl(ip->ip_src.s_addr);
540 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
541 			gw = ntohl(icp->icmp_gwaddr.s_addr);
542 			printf("icmp redirect from %d.%d.%d.%d: "
543 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
544 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
545 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
546 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
547 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
548 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
549 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
550 		}
551 		if (drop_redirect)
552 			break;
553 		if (code > 3)
554 			goto badcode;
555 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
556 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
557 			icmpstat.icps_badlen++;
558 			break;
559 		}
560 		/*
561 		 * Short circuit routing redirects to force
562 		 * immediate change in the kernel's routing
563 		 * tables.  The message is also handed to anyone
564 		 * listening on a raw socket (e.g. the routing
565 		 * daemon for use in updating its tables).
566 		 */
567 		icmpgw.sin_addr = ip->ip_src;
568 		icmpdst.sin_addr = icp->icmp_gwaddr;
569 #ifdef	ICMPPRINTFS
570 		if (icmpprintfs) {
571 			char buf[sizeof "aaa.bbb.ccc.ddd"];
572 
573 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
574 			printf("redirect dst %s to %s\n",
575 			       buf, inet_ntoa(icp->icmp_gwaddr));
576 		}
577 #endif
578 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
579 		rtredirect((struct sockaddr *)&icmpsrc,
580 		  (struct sockaddr *)&icmpdst,
581 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
582 		  (struct sockaddr *)&icmpgw);
583 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
584 #ifdef IPSEC
585 		key_sa_routechange((struct sockaddr *)&icmpsrc);
586 #endif
587 		break;
588 
589 	/*
590 	 * No kernel processing for the following;
591 	 * just fall through to send to raw listener.
592 	 */
593 	case ICMP_ECHOREPLY:
594 	case ICMP_ROUTERADVERT:
595 	case ICMP_ROUTERSOLICIT:
596 	case ICMP_TSTAMPREPLY:
597 	case ICMP_IREQREPLY:
598 	case ICMP_MASKREPLY:
599 	default:
600 		break;
601 	}
602 
603 raw:
604 	rip_input(m, off, proto);
605 	return;
606 
607 freeit:
608 	m_freem(m);
609 }
610 
611 /*
612  * Reflect the ip packet back to the source
613  */
614 static void
615 icmp_reflect(m)
616 	struct mbuf *m;
617 {
618 	struct ip *ip = mtod(m, struct ip *);
619 	struct ifaddr *ifa;
620 	struct in_ifaddr *ia;
621 	struct in_addr t;
622 	struct mbuf *opts = 0;
623 	int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
624 	struct route *ro = NULL, rt;
625 
626 	if (!in_canforward(ip->ip_src) &&
627 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
628 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
629 		m_freem(m);	/* Bad return address */
630 		icmpstat.icps_badaddr++;
631 		goto done;	/* Ip_output() will check for broadcast */
632 	}
633 	t = ip->ip_dst;
634 	ip->ip_dst = ip->ip_src;
635 	ro = &rt;
636 	bzero(ro, sizeof *ro);
637 	/*
638 	 * If the incoming packet was addressed directly to us,
639 	 * use dst as the src for the reply.  Otherwise (broadcast
640 	 * or anonymous), use the address which corresponds
641 	 * to the incoming interface.
642 	 */
643 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
644 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
645 			goto match;
646 	if (m->m_pkthdr.rcvif != NULL &&
647 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
648 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
649 			if (ifa->ifa_addr->sa_family != AF_INET)
650 				continue;
651 			ia = ifatoia(ifa);
652 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
653 			    t.s_addr)
654 				goto match;
655 		}
656 	}
657 	ia = ip_rtaddr(ip->ip_dst, ro);
658 	/* We need a route to do anything useful. */
659 	if (ia == NULL) {
660 		m_freem(m);
661 		icmpstat.icps_noroute++;
662 		goto done;
663 	}
664 match:
665 	t = IA_SIN(ia)->sin_addr;
666 	ip->ip_src = t;
667 	ip->ip_ttl = ip_defttl;
668 
669 	if (optlen > 0) {
670 		u_char *cp;
671 		int opt, cnt;
672 		u_int len;
673 
674 		/*
675 		 * Retrieve any source routing from the incoming packet;
676 		 * add on any record-route or timestamp options.
677 		 */
678 		cp = (u_char *) (ip + 1);
679 		if ((opts = ip_srcroute()) == 0 &&
680 		    (opts = m_gethdr(MB_DONTWAIT, MT_HEADER))) {
681 			opts->m_len = sizeof(struct in_addr);
682 			mtod(opts, struct in_addr *)->s_addr = 0;
683 		}
684 		if (opts) {
685 #ifdef ICMPPRINTFS
686 			if (icmpprintfs)
687 				printf("icmp_reflect optlen %d rt %d => ",
688 				       optlen, opts->m_len);
689 #endif
690 			for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
691 				opt = cp[IPOPT_OPTVAL];
692 				if (opt == IPOPT_EOL)
693 					break;
694 				if (opt == IPOPT_NOP)
695 					len = 1;
696 				else {
697 					if (cnt < IPOPT_OLEN + sizeof *cp)
698 						break;
699 					len = cp[IPOPT_OLEN];
700 					if (len < IPOPT_OLEN + sizeof *cp ||
701 					    len > cnt)
702 					break;
703 				}
704 				/*
705 				 * Should check for overflow, but it
706 				 * "can't happen".
707 				 */
708 				if (opt == IPOPT_RR || opt == IPOPT_TS ||
709 				    opt == IPOPT_SECURITY) {
710 					bcopy(cp,
711 					      mtod(opts, caddr_t) + opts->m_len,
712 					      len);
713 					opts->m_len += len;
714 				}
715 			}
716 			/* Terminate & pad, if necessary */
717 			cnt = opts->m_len % 4;
718 			if (cnt) {
719 				for (; cnt < 4; cnt++) {
720 					*(mtod(opts, caddr_t) + opts->m_len) =
721 					    IPOPT_EOL;
722 					opts->m_len++;
723 				}
724 			}
725 #ifdef ICMPPRINTFS
726 			if (icmpprintfs)
727 				printf("%d\n", opts->m_len);
728 #endif
729 		}
730 		/*
731 		 * Now strip out original options by copying rest of first
732 		 * mbuf's data back, and adjust the IP length.
733 		 */
734 		ip->ip_len -= optlen;
735 		ip->ip_vhl = IP_VHL_BORING;
736 		m->m_len -= optlen;
737 		if (m->m_flags & M_PKTHDR)
738 			m->m_pkthdr.len -= optlen;
739 		optlen += sizeof(struct ip);
740 		bcopy((caddr_t)ip + optlen, ip + 1,
741 		      m->m_len - sizeof(struct ip));
742 	}
743 	m->m_pkthdr.fw_flags &= PF_MBUF_GENERATED;
744 	m->m_flags &= ~(M_BCAST|M_MCAST);
745 	icmp_send(m, opts, ro);
746 done:
747 	if (opts)
748 		m_free(opts);
749 	if (ro && ro->ro_rt)
750 		RTFREE(ro->ro_rt);
751 }
752 
753 /*
754  * Send an icmp packet back to the ip level,
755  * after supplying a checksum.
756  */
757 static void
758 icmp_send(m, opts, rt)
759 	struct mbuf *m;
760 	struct mbuf *opts;
761 	struct route *rt;
762 {
763 	struct ip *ip = mtod(m, struct ip *);
764 	int hlen;
765 	struct icmp *icp;
766 
767 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
768 	m->m_data += hlen;
769 	m->m_len -= hlen;
770 	icp = mtod(m, struct icmp *);
771 	icp->icmp_cksum = 0;
772 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
773 	m->m_data -= hlen;
774 	m->m_len += hlen;
775 	m->m_pkthdr.rcvif = (struct ifnet *)0;
776 #ifdef ICMPPRINTFS
777 	if (icmpprintfs) {
778 		char buf[sizeof "aaa.bbb.ccc.ddd"];
779 
780 		strcpy(buf, inet_ntoa(ip->ip_dst));
781 		printf("icmp_send dst %s src %s\n", buf, inet_ntoa(ip->ip_src));
782 	}
783 #endif
784 	ip_output(m, opts, rt, 0, NULL, NULL);
785 }
786 
787 n_time
788 iptime()
789 {
790 	struct timeval atv;
791 	u_long t;
792 
793 	getmicrotime(&atv);
794 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
795 	return (htonl(t));
796 }
797 
798 #if 1
799 /*
800  * Return the next larger or smaller MTU plateau (table from RFC 1191)
801  * given current value MTU.  If DIR is less than zero, a larger plateau
802  * is returned; otherwise, a smaller value is returned.
803  */
804 int
805 ip_next_mtu(int mtu, int dir)
806 {
807 	static int mtutab[] = {
808 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
809 		68, 0
810 	};
811 	int i;
812 
813 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
814 		if (mtu >= mtutab[i])
815 			break;
816 	}
817 
818 	if (dir < 0) {
819 		if (i == 0) {
820 			return 0;
821 		} else {
822 			return mtutab[i - 1];
823 		}
824 	} else {
825 		if (mtutab[i] == 0) {
826 			return 0;
827 		} else if(mtu > mtutab[i]) {
828 			return mtutab[i];
829 		} else {
830 			return mtutab[i + 1];
831 		}
832 	}
833 }
834 #endif
835 
836 #ifdef ICMP_BANDLIM
837 /*
838  * badport_bandlim() - check for ICMP bandwidth limit
839  *
840  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
841  *	hit our bandwidth limit and it is not ok.
842  *
843  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
844  *
845  *	For now we separate the TCP and UDP subsystems w/ different 'which'
846  *	values.  We may eventually remove this separation (and simplify the
847  *	code further).
848  *
849  *	Note that the printing of the error message is delayed so we can
850  *	properly print the icmp error rate that the system was trying to do
851  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
852  *	the 'final' error, but it doesn't make sense to solve the printing
853  *	delay with more complex code.
854  */
855 int
856 badport_bandlim(int which)
857 {
858 	static int lticks[BANDLIM_MAX + 1];
859 	static int lpackets[BANDLIM_MAX + 1];
860 	int dticks;
861 	const char *bandlimittype[] = {
862 		"Limiting icmp unreach response",
863 		"Limiting icmp ping response",
864 		"Limiting icmp tstamp response",
865 		"Limiting closed port RST response",
866 		"Limiting open port RST response"
867 		};
868 
869 	/*
870 	 * Return ok status if feature disabled or argument out of
871 	 * ranage.
872 	 */
873 
874 	if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
875 		return(0);
876 	dticks = ticks - lticks[which];
877 
878 	/*
879 	 * reset stats when cumulative dt exceeds one second.
880 	 */
881 
882 	if ((unsigned int)dticks > hz) {
883 		if (lpackets[which] > icmplim && icmplim_output) {
884 			printf("%s from %d to %d packets per second\n",
885 				bandlimittype[which],
886 				lpackets[which],
887 				icmplim
888 			);
889 		}
890 		lticks[which] = ticks;
891 		lpackets[which] = 0;
892 	}
893 
894 	/*
895 	 * bump packet count
896 	 */
897 
898 	if (++lpackets[which] > icmplim) {
899 		return(-1);
900 	}
901 	return(0);
902 }
903 #endif
904