xref: /openbsd/sys/netinet/ip_icmp.c (revision 610f49f8)
1 /*	$OpenBSD: ip_icmp.c,v 1.45 2002/01/21 05:33:14 itojun Exp $	*/
2 /*	$NetBSD: ip_icmp.c,v 1.19 1996/02/13 23:42:22 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)COPYRIGHT	1.1 (NRL) 17 January 1995
37  *
38  * NRL grants permission for redistribution and use in source and binary
39  * forms, with or without modification, of the software and documentation
40  * created at NRL provided that the following conditions are met:
41  *
42  * 1. Redistributions of source code must retain the above copyright
43  *    notice, this list of conditions and the following disclaimer.
44  * 2. Redistributions in binary form must reproduce the above copyright
45  *    notice, this list of conditions and the following disclaimer in the
46  *    documentation and/or other materials provided with the distribution.
47  * 3. All advertising materials mentioning features or use of this software
48  *    must display the following acknowledgements:
49  * 	This product includes software developed by the University of
50  * 	California, Berkeley and its contributors.
51  * 	This product includes software developed at the Information
52  * 	Technology Division, US Naval Research Laboratory.
53  * 4. Neither the name of the NRL nor the names of its contributors
54  *    may be used to endorse or promote products derived from this software
55  *    without specific prior written permission.
56  *
57  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
58  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
59  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
60  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
61  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
62  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
63  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
64  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
65  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
66  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
67  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
68  *
69  * The views and conclusions contained in the software and documentation
70  * are those of the authors and should not be interpreted as representing
71  * official policies, either expressed or implied, of the US Naval
72  * Research Laboratory (NRL).
73  */
74 
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/mbuf.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/sysctl.h>
81 
82 #include <net/if.h>
83 #include <net/route.h>
84 
85 #include <netinet/in.h>
86 #include <netinet/in_systm.h>
87 #include <netinet/in_var.h>
88 #include <netinet/ip.h>
89 #include <netinet/ip_icmp.h>
90 #include <netinet/ip_var.h>
91 #include <netinet/icmp_var.h>
92 
93 /*
94  * ICMP routines: error generation, receive packet processing, and
95  * routines to turnaround packets back to the originator, and
96  * host table maintenance routines.
97  */
98 
99 int	icmpmaskrepl = 0;
100 int	icmpbmcastecho = 0;
101 #ifdef ICMPPRINTFS
102 int	icmpprintfs = 0;
103 #endif
104 int	icmperrppslim = 100;
105 int	icmperrpps_count = 0;
106 struct timeval icmperrppslim_last;
107 int	icmp_rediraccept = 1;
108 int	icmp_redirtimeout = 10 * 60;
109 static struct rttimer_queue *icmp_redirect_timeout_q = NULL;
110 
111 void icmp_mtudisc_timeout __P((struct rtentry *, struct rttimer *));
112 int icmp_ratelimit __P((const struct in_addr *, const int, const int));
113 static void icmp_redirect_timeout __P((struct rtentry *, struct rttimer *));
114 
115 extern	struct protosw inetsw[];
116 
117 void
118 icmp_init()
119 {
120 	/*
121 	 * This is only useful if the user initializes redirtimeout to
122 	 * something other than zero.
123 	 */
124 	if (icmp_redirtimeout != 0) {
125 		icmp_redirect_timeout_q =
126 			rt_timer_queue_create(icmp_redirtimeout);
127 	}
128 }
129 
130 /*
131  * Generate an error packet of type error
132  * in response to bad packet ip.
133  *
134  * The ip packet inside has ip_off and ip_len in host byte order.
135  */
136 void
137 icmp_error(n, type, code, dest, destifp)
138 	struct mbuf *n;
139 	int type, code;
140 	n_long dest;
141 	struct ifnet *destifp;
142 {
143 	register struct ip *oip = mtod(n, struct ip *), *nip;
144 	register unsigned oiplen = oip->ip_hl << 2;
145 	register struct icmp *icp;
146 	struct mbuf *m;
147 	struct m_tag *mtag;
148 	unsigned icmplen, mblen;
149 
150 #ifdef ICMPPRINTFS
151 	if (icmpprintfs)
152 		printf("icmp_error(%x, %d, %d)\n", oip, type, code);
153 #endif
154 	if (type != ICMP_REDIRECT)
155 		icmpstat.icps_error++;
156 	/*
157 	 * Don't send error if not the first fragment of message.
158 	 * Don't error if the old packet protocol was ICMP
159 	 * error message, only known informational types.
160 	 */
161 	if (oip->ip_off & IP_OFFMASK)
162 		goto freeit;
163 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
164 	  n->m_len >= oiplen + ICMP_MINLEN &&
165 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
166 		icmpstat.icps_oldicmp++;
167 		goto freeit;
168 	}
169 	/* Don't send error in response to a multicast or broadcast packet */
170 	if (n->m_flags & (M_BCAST|M_MCAST))
171 		goto freeit;
172 
173 	/*
174 	 * First, do a rate limitation check.
175  	 */
176 	if (icmp_ratelimit(&oip->ip_src, type, code)) {
177 		/* XXX stat */
178 		goto freeit;
179 	}
180 
181 	/*
182 	 * Now, formulate icmp message
183 	 */
184 	icmplen = oiplen + min(8, oip->ip_len);
185  	/*
186 	 * Defend against mbuf chains shorter than oip->ip_len:
187 	 */
188 	mblen = 0;
189 	for (m = n; m && (mblen < icmplen); m = m->m_next)
190 		mblen += m->m_len;
191 	icmplen = min(mblen, icmplen);
192 
193 	/*
194 	 * As we are not required to return everything we have,
195 	 * we return whatever we can return at ease.
196 	 *
197 	 * Note that ICMP datagrams longer than 576 octets are out of spec
198 	 * according to RFC1812;
199 	 */
200 
201 	KASSERT(ICMP_MINLEN <= MCLBYTES);
202 
203 	if (icmplen + ICMP_MINLEN > MCLBYTES)
204 		icmplen = MCLBYTES - ICMP_MINLEN - sizeof (struct ip);
205 
206 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
207 	if (m && (sizeof (struct ip) + icmplen + ICMP_MINLEN > MHLEN)) {
208 		MCLGET(m, M_DONTWAIT);
209 		if ((m->m_flags & M_EXT) == 0) {
210 			m_freem(m);
211 			m = NULL;
212 		}
213 	}
214 	if (m == NULL)
215 		goto freeit;
216 	m->m_len = icmplen + ICMP_MINLEN;
217 	if ((m->m_flags & M_EXT) == 0)
218 		MH_ALIGN(m, m->m_len);
219 	icp = mtod(m, struct icmp *);
220 	if ((u_int)type > ICMP_MAXTYPE)
221 		panic("icmp_error");
222 	icmpstat.icps_outhist[type]++;
223 	icp->icmp_type = type;
224 	if (type == ICMP_REDIRECT)
225 		icp->icmp_gwaddr.s_addr = dest;
226 	else {
227 		icp->icmp_void = 0;
228 		/*
229 		 * The following assignments assume an overlay with the
230 		 * zeroed icmp_void field.
231 		 */
232 		if (type == ICMP_PARAMPROB) {
233 			icp->icmp_pptr = code;
234 			code = 0;
235 		} else if (type == ICMP_UNREACH &&
236 		    code == ICMP_UNREACH_NEEDFRAG && destifp)
237 			icp->icmp_nextmtu = htons(destifp->if_mtu);
238 	}
239 
240 	HTONS(oip->ip_off);
241 	HTONS(oip->ip_len);
242 	icp->icmp_code = code;
243 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
244 	nip = &icp->icmp_ip;
245 
246 	/*
247 	 * Now, copy old ip header (without options)
248 	 * in front of icmp message.
249 	 */
250 	if ((m->m_flags & M_EXT) == 0 &&
251 	    m->m_data - sizeof(struct ip) < m->m_pktdat)
252 		panic("icmp len");
253 	m->m_data -= sizeof(struct ip);
254 	m->m_len += sizeof(struct ip);
255 	m->m_pkthdr.len = m->m_len;
256 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
257 	nip = mtod(m, struct ip *);
258 	/* ip_v set in ip_output */
259 	nip->ip_hl = sizeof(struct ip) >> 2;
260 	nip->ip_tos = 0;
261 	nip->ip_len = m->m_len;
262 	/* ip_id set in ip_output */
263 	nip->ip_off = 0;
264 	/* ip_ttl set in icmp_reflect */
265 	nip->ip_p = IPPROTO_ICMP;
266 	nip->ip_src = oip->ip_src;
267 	nip->ip_dst = oip->ip_dst;
268 	/* move PF_GENERATED m_tag to new packet, if it exists */
269 	mtag = m_tag_find(n, PACKET_TAG_PF_GENERATED, NULL);
270 	if (mtag != NULL) {
271 		m_tag_unlink(n, mtag);
272 		m_tag_prepend(m, mtag);
273 	}
274 	icmp_reflect(m);
275 
276 freeit:
277 	m_freem(n);
278 }
279 
280 static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET };
281 static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET };
282 static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET };
283 struct sockaddr_in icmpmask = { 8, 0 };
284 
285 /*
286  * Process a received ICMP message.
287  */
288 void
289 #if __STDC__
290 icmp_input(struct mbuf *m, ...)
291 #else
292 icmp_input(m, va_alist)
293 	struct mbuf *m;
294 	va_dcl
295 #endif
296 {
297 	register struct icmp *icp;
298 	register struct ip *ip = mtod(m, struct ip *);
299 	int icmplen = ip->ip_len;
300 	register int i;
301 	struct in_ifaddr *ia;
302 	void *(*ctlfunc) __P((int, struct sockaddr *, void *));
303 	int code;
304 	extern u_char ip_protox[];
305 	int hlen;
306 	va_list ap;
307 	struct rtentry *rt;
308 
309 	va_start(ap, m);
310 	hlen = va_arg(ap, int);
311 	va_end(ap);
312 
313 	/*
314 	 * Locate icmp structure in mbuf, and check
315 	 * that not corrupted and of at least minimum length.
316 	 */
317 #ifdef ICMPPRINTFS
318 	if (icmpprintfs) {
319 		char buf[4*sizeof "123"];
320 
321 		strcpy(buf, inet_ntoa(ip->ip_dst));
322 		printf("icmp_input from %s to %s, len %d\n",
323 			inet_ntoa(ip->ip_src), buf, icmplen);
324 	}
325 #endif
326 	if (icmplen < ICMP_MINLEN) {
327 		icmpstat.icps_tooshort++;
328 		goto freeit;
329 	}
330 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
331 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
332 		icmpstat.icps_tooshort++;
333 		return;
334 	}
335 	ip = mtod(m, struct ip *);
336 	m->m_len -= hlen;
337 	m->m_data += hlen;
338 	icp = mtod(m, struct icmp *);
339 	if (in_cksum(m, icmplen)) {
340 		icmpstat.icps_checksum++;
341 		goto freeit;
342 	}
343 	m->m_len += hlen;
344 	m->m_data -= hlen;
345 
346 #ifdef ICMPPRINTFS
347 	/*
348 	 * Message type specific processing.
349 	 */
350 	if (icmpprintfs)
351 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
352 		    icp->icmp_code);
353 #endif
354 	if (icp->icmp_type > ICMP_MAXTYPE)
355 		goto raw;
356 	icmpstat.icps_inhist[icp->icmp_type]++;
357 	code = icp->icmp_code;
358 	switch (icp->icmp_type) {
359 
360 	case ICMP_UNREACH:
361 		switch (code) {
362 		case ICMP_UNREACH_NET:
363 		case ICMP_UNREACH_HOST:
364 		case ICMP_UNREACH_PROTOCOL:
365 		case ICMP_UNREACH_PORT:
366 		case ICMP_UNREACH_SRCFAIL:
367 			code += PRC_UNREACH_NET;
368 			break;
369 
370 		case ICMP_UNREACH_NEEDFRAG:
371 			code = PRC_MSGSIZE;
372 			break;
373 
374 		case ICMP_UNREACH_NET_UNKNOWN:
375 		case ICMP_UNREACH_NET_PROHIB:
376 		case ICMP_UNREACH_TOSNET:
377 			code = PRC_UNREACH_NET;
378 			break;
379 
380 		case ICMP_UNREACH_HOST_UNKNOWN:
381 		case ICMP_UNREACH_ISOLATED:
382 		case ICMP_UNREACH_HOST_PROHIB:
383 		case ICMP_UNREACH_TOSHOST:
384 		case ICMP_UNREACH_FILTER_PROHIB:
385 		case ICMP_UNREACH_HOST_PRECEDENCE:
386 		case ICMP_UNREACH_PRECEDENCE_CUTOFF:
387 			code = PRC_UNREACH_HOST;
388 			break;
389 
390 		default:
391 			goto badcode;
392 		}
393 		goto deliver;
394 
395 	case ICMP_TIMXCEED:
396 		if (code > 1)
397 			goto badcode;
398 		code += PRC_TIMXCEED_INTRANS;
399 		goto deliver;
400 
401 	case ICMP_PARAMPROB:
402 		if (code > 1)
403 			goto badcode;
404 		code = PRC_PARAMPROB;
405 		goto deliver;
406 
407 	case ICMP_SOURCEQUENCH:
408 		if (code)
409 			goto badcode;
410 		code = PRC_QUENCH;
411 	deliver:
412 		/* Free packet atttributes */
413 		if (m->m_flags & M_PKTHDR)
414 			m_tag_delete_chain(m, NULL);
415 
416 		/*
417 		 * Problem with datagram; advise higher level routines.
418 		 */
419 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
420 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
421 			icmpstat.icps_badlen++;
422 			goto freeit;
423 		}
424 		if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr))
425 			goto badcode;
426 		NTOHS(icp->icmp_ip.ip_len);
427 #ifdef INET6
428 		/* Get more contiguous data for a v6 in v4 ICMP message. */
429 		if (icp->icmp_ip.ip_p == IPPROTO_IPV6) {
430 			if (icmplen < ICMP_V6ADVLENMIN ||
431 			    icmplen < ICMP_V6ADVLEN(icp)) {
432 				icmpstat.icps_badlen++;
433 				goto freeit;
434 			} else {
435 				if ((m = m_pullup(m, (ip->ip_hl << 2) +
436 				    ICMP_V6ADVLEN(icp))) == NULL) {
437 					icmpstat.icps_tooshort++;
438 					return;
439 				}
440 				ip = mtod(m, struct ip *);
441 				icp = (struct icmp *)(m->m_data + (ip->ip_hl << 2));
442 			}
443 		}
444 #endif /* INET6 */
445 #ifdef ICMPPRINTFS
446 		if (icmpprintfs)
447 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
448 #endif
449 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
450 		/*
451 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
452 		 * notification to TCP layer.
453 		 */
454 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
455 		if (ctlfunc)
456 			(*ctlfunc)(code, sintosa(&icmpsrc), &icp->icmp_ip);
457 		break;
458 
459 	badcode:
460 		icmpstat.icps_badcode++;
461 		break;
462 
463 	case ICMP_ECHO:
464 		if (!icmpbmcastecho &&
465 		    (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
466 			icmpstat.icps_bmcastecho++;
467 			break;
468 		}
469 		icp->icmp_type = ICMP_ECHOREPLY;
470 		goto reflect;
471 
472 	case ICMP_TSTAMP:
473 		if (!icmpbmcastecho &&
474 		    (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
475 			icmpstat.icps_bmcastecho++;
476 			break;
477 		}
478 		if (icmplen < ICMP_TSLEN) {
479 			icmpstat.icps_badlen++;
480 			break;
481 		}
482 		icp->icmp_type = ICMP_TSTAMPREPLY;
483 		icp->icmp_rtime = iptime();
484 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
485 		goto reflect;
486 
487 	case ICMP_MASKREQ:
488 		if (icmpmaskrepl == 0)
489 			break;
490 		/*
491 		 * We are not able to respond with all ones broadcast
492 		 * unless we receive it over a point-to-point interface.
493 		 */
494 		if (icmplen < ICMP_MASKLEN) {
495 			icmpstat.icps_badlen++;
496 			break;
497 		}
498 		if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
499 		    ip->ip_dst.s_addr == INADDR_ANY)
500 			icmpdst.sin_addr = ip->ip_src;
501 		else
502 			icmpdst.sin_addr = ip->ip_dst;
503 		if (m->m_pkthdr.rcvif == NULL)
504 			break;
505 		ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst),
506 		    m->m_pkthdr.rcvif));
507 		if (ia == 0)
508 			break;
509 		icp->icmp_type = ICMP_MASKREPLY;
510 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
511 		if (ip->ip_src.s_addr == 0) {
512 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
513 				ip->ip_src = ia->ia_broadaddr.sin_addr;
514 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
515 				ip->ip_src = ia->ia_dstaddr.sin_addr;
516 		}
517 reflect:
518 		/* Free packet atttributes */
519 		if (m->m_flags & M_PKTHDR)
520 			m_tag_delete_chain(m, NULL);
521 
522 		ip->ip_len += hlen;	/* since ip_input deducts this */
523 		icmpstat.icps_reflect++;
524 		icmpstat.icps_outhist[icp->icmp_type]++;
525 		icmp_reflect(m);
526 		return;
527 
528 	case ICMP_REDIRECT:
529 		/* Free packet atttributes */
530 		if (m->m_flags & M_PKTHDR)
531 			m_tag_delete_chain(m, NULL);
532 		if (icmp_rediraccept == 0)
533 			goto freeit;
534 		if (code > 3)
535 			goto badcode;
536 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
537 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
538 			icmpstat.icps_badlen++;
539 			break;
540 		}
541 		/*
542 		 * Short circuit routing redirects to force
543 		 * immediate change in the kernel's routing
544 		 * tables.  The message is also handed to anyone
545 		 * listening on a raw socket (e.g. the routing
546 		 * daemon for use in updating its tables).
547 		 */
548 		icmpgw.sin_addr = ip->ip_src;
549 		icmpdst.sin_addr = icp->icmp_gwaddr;
550 #ifdef	ICMPPRINTFS
551 		if (icmpprintfs) {
552 			char buf[4 * sizeof "123"];
553 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
554 
555 			printf("redirect dst %s to %s\n",
556 			    buf, inet_ntoa(icp->icmp_gwaddr));
557 		}
558 #endif
559 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
560 		rt = NULL;
561 		rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst),
562 		    (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
563 		    sintosa(&icmpgw), (struct rtentry **)&rt);
564 		if (rt != NULL && icmp_redirtimeout != 0) {
565 			(void)rt_timer_add(rt, icmp_redirect_timeout,
566 			 		 icmp_redirect_timeout_q);
567 		}
568 		if (rt != NULL)
569 			rtfree(rt);
570 		pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc));
571 		break;
572 
573 	/*
574 	 * No kernel processing for the following;
575 	 * just fall through to send to raw listener.
576 	 */
577 	case ICMP_ECHOREPLY:
578 	case ICMP_ROUTERADVERT:
579 	case ICMP_ROUTERSOLICIT:
580 	case ICMP_TSTAMPREPLY:
581 	case ICMP_IREQREPLY:
582 	case ICMP_MASKREPLY:
583 	case ICMP_TRACEROUTE:
584 	case ICMP_DATACONVERR:
585 	case ICMP_MOBILE_REDIRECT:
586 	case ICMP_IPV6_WHEREAREYOU:
587 	case ICMP_IPV6_IAMHERE:
588 	case ICMP_MOBILE_REGREQUEST:
589 	case ICMP_MOBILE_REGREPLY:
590 	case ICMP_PHOTURIS:
591 	default:
592 		break;
593 	}
594 
595 raw:
596 	rip_input(m);
597 	return;
598 
599 freeit:
600 	m_freem(m);
601 }
602 
603 /*
604  * Reflect the ip packet back to the source
605  */
606 void
607 icmp_reflect(m)
608 	struct mbuf *m;
609 {
610 	register struct ip *ip = mtod(m, struct ip *);
611 	register struct in_ifaddr *ia;
612 	struct in_addr t;
613 	struct mbuf *opts = 0;
614 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
615 
616 	if (!in_canforward(ip->ip_src) &&
617 	    ((ip->ip_src.s_addr & IN_CLASSA_NET) !=
618 	     htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
619 		m_freem(m);	/* Bad return address */
620 		goto done;	/* ip_output() will check for broadcast */
621 	}
622 	t = ip->ip_dst;
623 	ip->ip_dst = ip->ip_src;
624 	/*
625 	 * If the incoming packet was addressed directly to us,
626 	 * use dst as the src for the reply.  Otherwise (broadcast
627 	 * or anonymous), use the address which corresponds
628 	 * to the incoming interface.
629 	 */
630 	for (ia = in_ifaddr.tqh_first; ia; ia = ia->ia_list.tqe_next) {
631 		if (t.s_addr == ia->ia_addr.sin_addr.s_addr)
632 			break;
633 		if ((ia->ia_ifp->if_flags & IFF_BROADCAST) &&
634 		    t.s_addr == ia->ia_broadaddr.sin_addr.s_addr)
635 			break;
636 	}
637 	icmpdst.sin_addr = t;
638 	if ((ia == (struct in_ifaddr *)0) && (m->m_pkthdr.rcvif != NULL))
639 		ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst),
640 					      m->m_pkthdr.rcvif));
641 	/*
642 	 * The following happens if the packet was not addressed to us,
643 	 * and was received on an interface with no IP address.
644 	 */
645 	if (ia == (struct in_ifaddr *)0) {
646 	        struct sockaddr_in *dst;
647 		struct route ro;
648 
649 		bzero((caddr_t) &ro, sizeof(ro));
650 		dst = satosin(&ro.ro_dst);
651 		dst->sin_family = AF_INET;
652 		dst->sin_len = sizeof(*dst);
653 		dst->sin_addr = t;
654 
655 		rtalloc(&ro);
656 		if (ro.ro_rt == 0)
657 		{
658 		    ipstat.ips_noroute++;
659 		    goto done;
660 		}
661 
662 		ia = ifatoia(ro.ro_rt->rt_ifa);
663 		ro.ro_rt->rt_use++;
664                 RTFREE(ro.ro_rt);
665         }
666 
667 	t = ia->ia_addr.sin_addr;
668 	ip->ip_src = t;
669 	ip->ip_ttl = MAXTTL;
670 
671 	if (optlen > 0) {
672 		register u_char *cp;
673 		int opt, cnt;
674 		u_int len;
675 
676 		/*
677 		 * Retrieve any source routing from the incoming packet;
678 		 * add on any record-route or timestamp options.
679 		 */
680 		cp = (u_char *) (ip + 1);
681 		if ((opts = ip_srcroute()) == 0 &&
682 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
683 			opts->m_len = sizeof(struct in_addr);
684 			mtod(opts, struct in_addr *)->s_addr = 0;
685 		}
686 		if (opts) {
687 #ifdef ICMPPRINTFS
688 		    if (icmpprintfs)
689 			    printf("icmp_reflect optlen %d rt %d => ",
690 				optlen, opts->m_len);
691 #endif
692 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
693 			    opt = cp[IPOPT_OPTVAL];
694 			    if (opt == IPOPT_EOL)
695 				    break;
696 			    if (opt == IPOPT_NOP)
697 				    len = 1;
698 			    else {
699 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
700 					    break;
701 				    len = cp[IPOPT_OLEN];
702 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
703 				        len > cnt)
704 					    break;
705 			    }
706 			    /*
707 			     * Should check for overflow, but it "can't happen"
708 			     */
709 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
710 				opt == IPOPT_SECURITY) {
711 				    bcopy((caddr_t)cp,
712 					mtod(opts, caddr_t) + opts->m_len, len);
713 				    opts->m_len += len;
714 			    }
715 		    }
716 		    /* Terminate & pad, if necessary */
717 		    if ((cnt = opts->m_len % 4) != 0) {
718 			    for (; cnt < 4; cnt++) {
719 				    *(mtod(opts, caddr_t) + opts->m_len) =
720 					IPOPT_EOL;
721 				    opts->m_len++;
722 			    }
723 		    }
724 #ifdef ICMPPRINTFS
725 		    if (icmpprintfs)
726 			    printf("%d\n", opts->m_len);
727 #endif
728 		}
729 		/*
730 		 * Now strip out original options by copying rest of first
731 		 * mbuf's data back, and adjust the IP length.
732 		 */
733 		ip->ip_len -= optlen;
734 		ip->ip_hl = sizeof(struct ip) >> 2;
735 		m->m_len -= optlen;
736 		if (m->m_flags & M_PKTHDR)
737 			m->m_pkthdr.len -= optlen;
738 		optlen += sizeof(struct ip);
739 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
740 			 (unsigned)(m->m_len - sizeof(struct ip)));
741 	}
742 	m->m_flags &= ~(M_BCAST|M_MCAST);
743 	icmp_send(m, opts);
744 done:
745 	if (opts)
746 		(void)m_free(opts);
747 }
748 
749 /*
750  * Send an icmp packet back to the ip level,
751  * after supplying a checksum.
752  */
753 void
754 icmp_send(m, opts)
755 	register struct mbuf *m;
756 	struct mbuf *opts;
757 {
758 	register struct ip *ip = mtod(m, struct ip *);
759 	register int hlen;
760 	register struct icmp *icp;
761 
762 	hlen = ip->ip_hl << 2;
763 	m->m_data += hlen;
764 	m->m_len -= hlen;
765 	icp = mtod(m, struct icmp *);
766 	icp->icmp_cksum = 0;
767 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
768 	m->m_data -= hlen;
769 	m->m_len += hlen;
770 #ifdef ICMPPRINTFS
771 	if (icmpprintfs) {
772 		char buf[4 * sizeof "123"];
773 
774 		strcpy(buf, inet_ntoa(ip->ip_dst));
775 		printf("icmp_send dst %s src %s\n",
776 		    buf, inet_ntoa(ip->ip_src));
777 	}
778 #endif
779 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
780 }
781 
782 n_time
783 iptime()
784 {
785 	struct timeval atv;
786 	u_long t;
787 
788 	microtime(&atv);
789 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
790 	return (htonl(t));
791 }
792 
793 int
794 icmp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
795 	int *name;
796 	u_int namelen;
797 	void *oldp;
798 	size_t *oldlenp;
799 	void *newp;
800 	size_t newlen;
801 {
802 
803 	/* All sysctl names at this level are terminal. */
804 	if (namelen != 1)
805 		return (ENOTDIR);
806 
807 	switch (name[0]) {
808 	case ICMPCTL_MASKREPL:
809 		return (sysctl_int(oldp, oldlenp, newp, newlen, &icmpmaskrepl));
810 	case ICMPCTL_BMCASTECHO:
811 		return (sysctl_int(oldp, oldlenp, newp, newlen, &icmpbmcastecho));
812 	case ICMPCTL_ERRPPSLIMIT:
813 		return (sysctl_int(oldp, oldlenp, newp, newlen,
814 		    &icmperrppslim));
815 		break;
816 	case ICMPCTL_REDIRACCEPT:
817 		return (sysctl_int(oldp, oldlenp, newp, newlen,
818 		    &icmp_rediraccept));
819 		break;
820 	case ICMPCTL_REDIRTIMEOUT: {
821 		int error;
822 
823 		error = sysctl_int(oldp, oldlenp, newp, newlen,
824 			    &icmp_redirtimeout);
825 		if (icmp_redirect_timeout_q != NULL) {
826 			if (icmp_redirtimeout == 0) {
827 				rt_timer_queue_destroy(icmp_redirect_timeout_q,
828 				    TRUE);
829 				icmp_redirect_timeout_q = NULL;
830 			} else {
831 				rt_timer_queue_change(icmp_redirect_timeout_q,
832 				    icmp_redirtimeout);
833 			}
834 		} else if (icmp_redirtimeout > 0) {
835 			icmp_redirect_timeout_q =
836 			    rt_timer_queue_create(icmp_redirtimeout);
837 		}
838 		return (error);
839 
840 		break;
841 	}
842 	default:
843 		return (ENOPROTOOPT);
844 	}
845 	/* NOTREACHED */
846 }
847 
848 struct rtentry *
849 icmp_mtudisc_clone(struct sockaddr *dst)
850 {
851 	struct rtentry *rt;
852 	int error;
853 
854 	rt = rtalloc1(dst, 1);
855 	if (rt == 0)
856 		return (NULL);
857 
858 	/* If we didn't get a host route, allocate one */
859 
860 	if ((rt->rt_flags & RTF_HOST) == 0) {
861 		struct rtentry *nrt;
862 
863 		error = rtrequest((int) RTM_ADD, dst,
864 		    (struct sockaddr *) rt->rt_gateway,
865 		    (struct sockaddr *) 0,
866 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
867 		if (error) {
868 			rtfree(rt);
869 			return (NULL);
870 		}
871 		nrt->rt_rmx = rt->rt_rmx;
872 		rtfree(rt);
873 		rt = nrt;
874 	}
875 	error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q);
876 	if (error) {
877 		rtfree(rt);
878 		return (NULL);
879 	}
880 
881 	return (rt);
882 }
883 
884 void
885 icmp_mtudisc(icp)
886 	struct icmp *icp;
887 {
888 	struct rtentry *rt;
889 	struct sockaddr *dst = sintosa(&icmpsrc);
890 	u_long mtu = ntohs(icp->icmp_nextmtu);  /* Why a long?  IPv6 */
891 
892 	/* Table of common MTUs: */
893 
894 	static u_short mtu_table[] = {65535, 65280, 32000, 17914, 9180, 8166,
895 				      4352, 2002, 1492, 1006, 508, 296, 68, 0};
896 
897 	rt = icmp_mtudisc_clone(dst);
898 	if (rt == 0)
899 		return;
900 
901 	if (mtu == 0) {
902 		int i = 0;
903 
904 		mtu = icp->icmp_ip.ip_len; /* NTOHS happened in deliver: */
905 		/* Some 4.2BSD-based routers incorrectly adjust the ip_len */
906 		if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0)
907 			mtu -= (icp->icmp_ip.ip_hl << 2);
908 
909 		/* If we still can't guess a value, try the route */
910 
911 		if (mtu == 0) {
912 			mtu = rt->rt_rmx.rmx_mtu;
913 
914 			/* If no route mtu, default to the interface mtu */
915 
916 			if (mtu == 0)
917 				mtu = rt->rt_ifp->if_mtu;
918 		}
919 
920 		for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++)
921 			if (mtu > mtu_table[i]) {
922 				mtu = mtu_table[i];
923 				break;
924 			}
925 	}
926 
927 	/*
928 	 * XXX:   RTV_MTU is overloaded, since the admin can set it
929 	 *	  to turn off PMTU for a route, and the kernel can
930 	 *	  set it to indicate a serious problem with PMTU
931 	 *	  on a route.  We should be using a separate flag
932 	 *	  for the kernel to indicate this.
933 	 */
934 
935 	if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
936 		if (mtu < 296 || mtu > rt->rt_ifp->if_mtu)
937 			rt->rt_rmx.rmx_locks |= RTV_MTU;
938 		else if (rt->rt_rmx.rmx_mtu > mtu ||
939 			 rt->rt_rmx.rmx_mtu == 0)
940 			rt->rt_rmx.rmx_mtu = mtu;
941 	}
942 
943 	rtfree(rt);
944 }
945 
946 void
947 icmp_mtudisc_timeout(rt, r)
948 	struct rtentry *rt;
949 	struct rttimer *r;
950 {
951 	if (rt == NULL)
952 		panic("icmp_mtudisc_timeout:  bad route to timeout");
953 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
954 	    (RTF_DYNAMIC | RTF_HOST)) {
955 		void *(*ctlfunc) __P((int, struct sockaddr *, void *));
956 		extern u_char ip_protox[];
957 		struct sockaddr_in sa;
958 
959 		sa = *(struct sockaddr_in *)rt_key(rt);
960 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
961 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
962 
963 		/* Notify TCP layer of increased Path MTU estimate */
964 		ctlfunc = inetsw[ip_protox[IPPROTO_TCP]].pr_ctlinput;
965 		if (ctlfunc)
966 			(*ctlfunc)(PRC_MTUINC,(struct sockaddr *)&sa, NULL);
967 	} else {
968 		if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
969 			rt->rt_rmx.rmx_mtu = 0;
970 		}
971 	}
972 }
973 
974 /*
975  * Perform rate limit check.
976  * Returns 0 if it is okay to send the icmp packet.
977  * Returns 1 if the router SHOULD NOT send this icmp packet due to rate
978  * limitation.
979  *
980  * XXX per-destination/type check necessary?
981  */
982 int
983 icmp_ratelimit(dst, type, code)
984 	const struct in_addr *dst;
985 	const int type;			/* not used at this moment */
986 	const int code;			/* not used at this moment */
987 {
988 
989 	/* PPS limit */
990 	if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count,
991 	    icmperrppslim)) {
992 		/* The packet is subject to rate limit */
993 		return 1;
994 	}
995 
996 	/*okay to send*/
997 	return 0;
998 }
999 
1000 static void
1001 icmp_redirect_timeout(rt, r)
1002 	struct rtentry *rt;
1003 	struct rttimer *r;
1004 {
1005 	if (rt == NULL)
1006 		panic("icmp_redirect_timeout:  bad route to timeout");
1007 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1008 	    (RTF_DYNAMIC | RTF_HOST)) {
1009 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
1010 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1011 	}
1012 }
1013