xref: /freebsd/sys/netinet/ip_icmp.c (revision 7660b554bc59a07be0431c17e0e33815818baa69)
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 #include "opt_mac.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/mac.h>
43 #include <sys/mbuf.h>
44 #include <sys/protosw.h>
45 #include <sys/socket.h>
46 #include <sys/time.h>
47 #include <sys/kernel.h>
48 #include <sys/sysctl.h>
49 
50 #include <net/if.h>
51 #include <net/if_types.h>
52 #include <net/route.h>
53 
54 #include <netinet/in.h>
55 #include <netinet/in_systm.h>
56 #include <netinet/in_var.h>
57 #include <netinet/ip.h>
58 #include <netinet/ip_icmp.h>
59 #include <netinet/ip_var.h>
60 #include <netinet/icmp_var.h>
61 
62 #ifdef IPSEC
63 #include <netinet6/ipsec.h>
64 #include <netkey/key.h>
65 #endif
66 
67 #ifdef FAST_IPSEC
68 #include <netipsec/ipsec.h>
69 #include <netipsec/key.h>
70 #define	IPSEC
71 #endif
72 
73 #include <machine/in_cksum.h>
74 
75 /*
76  * ICMP routines: error generation, receive packet processing, and
77  * routines to turnaround packets back to the originator, and
78  * host table maintenance routines.
79  */
80 
81 static struct	icmpstat icmpstat;
82 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
83 	&icmpstat, icmpstat, "");
84 
85 static int	icmpmaskrepl = 0;
86 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
87 	&icmpmaskrepl, 0, "Reply to ICMP Address Mask Request packets.");
88 
89 static u_int	icmpmaskfake = 0;
90 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
91 	&icmpmaskfake, 0, "Fake reply to ICMP Address Mask Request packets.");
92 
93 static int	drop_redirect = 0;
94 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
95 	&drop_redirect, 0, "");
96 
97 static int	log_redirect = 0;
98 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
99 	&log_redirect, 0, "");
100 
101 static int      icmplim = 200;
102 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
103 	&icmplim, 0, "");
104 
105 static int	icmplim_output = 1;
106 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
107 	&icmplim_output, 0, "");
108 
109 /*
110  * ICMP broadcast echo sysctl
111  */
112 
113 static int	icmpbmcastecho = 0;
114 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
115 	&icmpbmcastecho, 0, "");
116 
117 
118 #ifdef ICMPPRINTFS
119 int	icmpprintfs = 0;
120 #endif
121 
122 static void	icmp_reflect(struct mbuf *);
123 static void	icmp_send(struct mbuf *, struct mbuf *, struct route *);
124 static int	ip_next_mtu(int, int);
125 
126 extern	struct protosw inetsw[];
127 
128 /*
129  * Generate an error packet of type error
130  * in response to bad packet ip.
131  */
132 void
133 icmp_error(n, type, code, dest, destifp)
134 	struct mbuf *n;
135 	int type, code;
136 	n_long dest;
137 	struct ifnet *destifp;
138 {
139 	register struct ip *oip = mtod(n, struct ip *), *nip;
140 	register unsigned oiplen = oip->ip_hl << 2;
141 	register struct icmp *icp;
142 	register struct mbuf *m;
143 	unsigned icmplen;
144 
145 #ifdef ICMPPRINTFS
146 	if (icmpprintfs)
147 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
148 #endif
149 	if (type != ICMP_REDIRECT)
150 		icmpstat.icps_error++;
151 	/*
152 	 * Don't send error if not the first fragment of message.
153 	 * Don't error if the old packet protocol was ICMP
154 	 * error message, only known informational types.
155 	 */
156 	if (oip->ip_off &~ (IP_MF|IP_DF))
157 		goto freeit;
158 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
159 	  n->m_len >= oiplen + ICMP_MINLEN &&
160 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
161 		icmpstat.icps_oldicmp++;
162 		goto freeit;
163 	}
164 	/* Don't send error in response to a multicast or broadcast packet */
165 	if (n->m_flags & (M_BCAST|M_MCAST))
166 		goto freeit;
167 	/*
168 	 * First, formulate icmp message
169 	 */
170 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
171 	if (m == NULL)
172 		goto freeit;
173 #ifdef MAC
174 	mac_create_mbuf_netlayer(n, m);
175 #endif
176 	icmplen = min(oiplen + 8, oip->ip_len);
177 	if (icmplen < sizeof(struct ip))
178 		panic("icmp_error: bad length");
179 	m->m_len = icmplen + ICMP_MINLEN;
180 	MH_ALIGN(m, m->m_len);
181 	icp = mtod(m, struct icmp *);
182 	if ((u_int)type > ICMP_MAXTYPE)
183 		panic("icmp_error");
184 	icmpstat.icps_outhist[type]++;
185 	icp->icmp_type = type;
186 	if (type == ICMP_REDIRECT)
187 		icp->icmp_gwaddr.s_addr = dest;
188 	else {
189 		icp->icmp_void = 0;
190 		/*
191 		 * The following assignments assume an overlay with the
192 		 * zeroed icmp_void field.
193 		 */
194 		if (type == ICMP_PARAMPROB) {
195 			icp->icmp_pptr = code;
196 			code = 0;
197 		} else if (type == ICMP_UNREACH &&
198 			code == ICMP_UNREACH_NEEDFRAG && destifp) {
199 			icp->icmp_nextmtu = htons(destifp->if_mtu);
200 		}
201 	}
202 
203 	icp->icmp_code = code;
204 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
205 	nip = &icp->icmp_ip;
206 
207 	/*
208 	 * Convert fields to network representation.
209 	 */
210 	nip->ip_len = htons(nip->ip_len);
211 	nip->ip_off = htons(nip->ip_off);
212 
213 	/*
214 	 * Now, copy old ip header (without options)
215 	 * in front of icmp message.
216 	 */
217 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
218 		panic("icmp len");
219 	m->m_data -= sizeof(struct ip);
220 	m->m_len += sizeof(struct ip);
221 	m->m_pkthdr.len = m->m_len;
222 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
223 	nip = mtod(m, struct ip *);
224 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
225 	nip->ip_len = m->m_len;
226 	nip->ip_v = IPVERSION;
227 	nip->ip_hl = 5;
228 	nip->ip_p = IPPROTO_ICMP;
229 	nip->ip_tos = 0;
230 	icmp_reflect(m);
231 
232 freeit:
233 	m_freem(n);
234 }
235 
236 static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET };
237 static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET };
238 static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET };
239 
240 /*
241  * Process a received ICMP message.
242  */
243 void
244 icmp_input(m, off)
245 	register struct mbuf *m;
246 	int off;
247 {
248 	int hlen = off;
249 	register struct icmp *icp;
250 	register struct ip *ip = mtod(m, struct ip *);
251 	int icmplen = ip->ip_len;
252 	register int i;
253 	struct in_ifaddr *ia;
254 	void (*ctlfunc)(int, struct sockaddr *, void *);
255 	int code;
256 
257 	/*
258 	 * Locate icmp structure in mbuf, and check
259 	 * that not corrupted and of at least minimum length.
260 	 */
261 #ifdef ICMPPRINTFS
262 	if (icmpprintfs) {
263 		char buf[4 * sizeof "123"];
264 		strcpy(buf, inet_ntoa(ip->ip_src));
265 		printf("icmp_input from %s to %s, len %d\n",
266 		       buf, inet_ntoa(ip->ip_dst), icmplen);
267 	}
268 #endif
269 	if (icmplen < ICMP_MINLEN) {
270 		icmpstat.icps_tooshort++;
271 		goto freeit;
272 	}
273 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
274 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
275 		icmpstat.icps_tooshort++;
276 		return;
277 	}
278 	ip = mtod(m, struct ip *);
279 	m->m_len -= hlen;
280 	m->m_data += hlen;
281 	icp = mtod(m, struct icmp *);
282 	if (in_cksum(m, icmplen)) {
283 		icmpstat.icps_checksum++;
284 		goto freeit;
285 	}
286 	m->m_len += hlen;
287 	m->m_data -= hlen;
288 
289 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
290 		/*
291 		 * Deliver very specific ICMP type only.
292 		 */
293 		switch (icp->icmp_type) {
294 		case ICMP_UNREACH:
295 		case ICMP_TIMXCEED:
296 			break;
297 		default:
298 			goto freeit;
299 		}
300 	}
301 
302 #ifdef ICMPPRINTFS
303 	if (icmpprintfs)
304 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
305 		    icp->icmp_code);
306 #endif
307 
308 	/*
309 	 * Message type specific processing.
310 	 */
311 	if (icp->icmp_type > ICMP_MAXTYPE)
312 		goto raw;
313 	icmpstat.icps_inhist[icp->icmp_type]++;
314 	code = icp->icmp_code;
315 	switch (icp->icmp_type) {
316 
317 	case ICMP_UNREACH:
318 		switch (code) {
319 			case ICMP_UNREACH_NET:
320 			case ICMP_UNREACH_HOST:
321 			case ICMP_UNREACH_SRCFAIL:
322 			case ICMP_UNREACH_NET_UNKNOWN:
323 			case ICMP_UNREACH_HOST_UNKNOWN:
324 			case ICMP_UNREACH_ISOLATED:
325 			case ICMP_UNREACH_TOSNET:
326 			case ICMP_UNREACH_TOSHOST:
327 			case ICMP_UNREACH_HOST_PRECEDENCE:
328 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
329 				code = PRC_UNREACH_NET;
330 				break;
331 
332 			case ICMP_UNREACH_NEEDFRAG:
333 				code = PRC_MSGSIZE;
334 				break;
335 
336 			/*
337 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
338 			 * Treat subcodes 2,3 as immediate RST
339 			 */
340 			case ICMP_UNREACH_PROTOCOL:
341 			case ICMP_UNREACH_PORT:
342 				code = PRC_UNREACH_PORT;
343 				break;
344 
345 			case ICMP_UNREACH_NET_PROHIB:
346 			case ICMP_UNREACH_HOST_PROHIB:
347 			case ICMP_UNREACH_FILTER_PROHIB:
348 				code = PRC_UNREACH_ADMIN_PROHIB;
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 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
378 			icmpstat.icps_badlen++;
379 			goto freeit;
380 		}
381 		icp->icmp_ip.ip_len = 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 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
456 			goto freeit;
457 		else
458 			goto reflect;
459 
460 	case ICMP_TSTAMP:
461 		if (!icmpbmcastecho
462 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
463 			icmpstat.icps_bmcasttstamp++;
464 			break;
465 		}
466 		if (icmplen < ICMP_TSLEN) {
467 			icmpstat.icps_badlen++;
468 			break;
469 		}
470 		icp->icmp_type = ICMP_TSTAMPREPLY;
471 		icp->icmp_rtime = iptime();
472 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
473 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
474 			goto freeit;
475 		else
476 			goto reflect;
477 
478 	case ICMP_MASKREQ:
479 		if (icmpmaskrepl == 0)
480 			break;
481 		/*
482 		 * We are not able to respond with all ones broadcast
483 		 * unless we receive it over a point-to-point interface.
484 		 */
485 		if (icmplen < ICMP_MASKLEN)
486 			break;
487 		switch (ip->ip_dst.s_addr) {
488 
489 		case INADDR_BROADCAST:
490 		case INADDR_ANY:
491 			icmpdst.sin_addr = ip->ip_src;
492 			break;
493 
494 		default:
495 			icmpdst.sin_addr = ip->ip_dst;
496 		}
497 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
498 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
499 		if (ia == 0)
500 			break;
501 		if (ia->ia_ifp == 0)
502 			break;
503 		icp->icmp_type = ICMP_MASKREPLY;
504 		if (icmpmaskfake == 0)
505 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
506 		else
507 			icp->icmp_mask = icmpmaskfake;
508 		if (ip->ip_src.s_addr == 0) {
509 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
510 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
511 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
512 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
513 		}
514 reflect:
515 		ip->ip_len += hlen;	/* since ip_input deducts this */
516 		icmpstat.icps_reflect++;
517 		icmpstat.icps_outhist[icp->icmp_type]++;
518 		icmp_reflect(m);
519 		return;
520 
521 	case ICMP_REDIRECT:
522 		if (log_redirect) {
523 			u_long src, dst, gw;
524 
525 			src = ntohl(ip->ip_src.s_addr);
526 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
527 			gw = ntohl(icp->icmp_gwaddr.s_addr);
528 			printf("icmp redirect from %d.%d.%d.%d: "
529 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
530 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
531 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
532 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
533 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
534 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
535 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
536 		}
537 		if (drop_redirect)
538 			break;
539 		if (code > 3)
540 			goto badcode;
541 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
542 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
543 			icmpstat.icps_badlen++;
544 			break;
545 		}
546 		/*
547 		 * Short circuit routing redirects to force
548 		 * immediate change in the kernel's routing
549 		 * tables.  The message is also handed to anyone
550 		 * listening on a raw socket (e.g. the routing
551 		 * daemon for use in updating its tables).
552 		 */
553 		icmpgw.sin_addr = ip->ip_src;
554 		icmpdst.sin_addr = icp->icmp_gwaddr;
555 #ifdef	ICMPPRINTFS
556 		if (icmpprintfs) {
557 			char buf[4 * sizeof "123"];
558 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
559 
560 			printf("redirect dst %s to %s\n",
561 			       buf, inet_ntoa(icp->icmp_gwaddr));
562 		}
563 #endif
564 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
565 		rtredirect((struct sockaddr *)&icmpsrc,
566 		  (struct sockaddr *)&icmpdst,
567 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
568 		  (struct sockaddr *)&icmpgw, (struct rtentry **)0);
569 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
570 #ifdef IPSEC
571 		key_sa_routechange((struct sockaddr *)&icmpsrc);
572 #endif
573 		break;
574 
575 	/*
576 	 * No kernel processing for the following;
577 	 * just fall through to send to raw listener.
578 	 */
579 	case ICMP_ECHOREPLY:
580 	case ICMP_ROUTERADVERT:
581 	case ICMP_ROUTERSOLICIT:
582 	case ICMP_TSTAMPREPLY:
583 	case ICMP_IREQREPLY:
584 	case ICMP_MASKREPLY:
585 	default:
586 		break;
587 	}
588 
589 raw:
590 	rip_input(m, off);
591 	return;
592 
593 freeit:
594 	m_freem(m);
595 }
596 
597 /*
598  * Reflect the ip packet back to the source
599  */
600 static void
601 icmp_reflect(m)
602 	struct mbuf *m;
603 {
604 	struct ip *ip = mtod(m, struct ip *);
605 	struct ifaddr *ifa;
606 	struct in_ifaddr *ia;
607 	struct in_addr t;
608 	struct mbuf *opts = 0;
609 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
610 	struct route *ro = NULL, rt;
611 
612 	if (!in_canforward(ip->ip_src) &&
613 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
614 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
615 		m_freem(m);	/* Bad return address */
616 		icmpstat.icps_badaddr++;
617 		goto done;	/* Ip_output() will check for broadcast */
618 	}
619 	t = ip->ip_dst;
620 	ip->ip_dst = ip->ip_src;
621 	ro = &rt;
622 	bzero(ro, sizeof(*ro));
623 	/*
624 	 * If the incoming packet was addressed directly to us,
625 	 * use dst as the src for the reply.  Otherwise (broadcast
626 	 * or anonymous), use the address which corresponds
627 	 * to the incoming interface.
628 	 */
629 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
630 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
631 			goto match;
632 	if (m->m_pkthdr.rcvif != NULL &&
633 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
634 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
635 			if (ifa->ifa_addr->sa_family != AF_INET)
636 				continue;
637 			ia = ifatoia(ifa);
638 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
639 			    t.s_addr)
640 				goto match;
641 		}
642 	}
643 	ia = ip_rtaddr(ip->ip_dst, ro);
644 	/* We need a route to do anything useful. */
645 	if (ia == NULL) {
646 		m_freem(m);
647 		icmpstat.icps_noroute++;
648 		goto done;
649 	}
650 match:
651 #ifdef MAC
652 	mac_reflect_mbuf_icmp(m);
653 #endif
654 	t = IA_SIN(ia)->sin_addr;
655 	ip->ip_src = t;
656 	ip->ip_ttl = ip_defttl;
657 
658 	if (optlen > 0) {
659 		register u_char *cp;
660 		int opt, cnt;
661 		u_int len;
662 
663 		/*
664 		 * Retrieve any source routing from the incoming packet;
665 		 * add on any record-route or timestamp options.
666 		 */
667 		cp = (u_char *) (ip + 1);
668 		if ((opts = ip_srcroute()) == 0 &&
669 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
670 			opts->m_len = sizeof(struct in_addr);
671 			mtod(opts, struct in_addr *)->s_addr = 0;
672 		}
673 		if (opts) {
674 #ifdef ICMPPRINTFS
675 		    if (icmpprintfs)
676 			    printf("icmp_reflect optlen %d rt %d => ",
677 				optlen, opts->m_len);
678 #endif
679 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
680 			    opt = cp[IPOPT_OPTVAL];
681 			    if (opt == IPOPT_EOL)
682 				    break;
683 			    if (opt == IPOPT_NOP)
684 				    len = 1;
685 			    else {
686 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
687 					    break;
688 				    len = cp[IPOPT_OLEN];
689 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
690 				        len > cnt)
691 					    break;
692 			    }
693 			    /*
694 			     * Should check for overflow, but it "can't happen"
695 			     */
696 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
697 				opt == IPOPT_SECURITY) {
698 				    bcopy((caddr_t)cp,
699 					mtod(opts, caddr_t) + opts->m_len, len);
700 				    opts->m_len += len;
701 			    }
702 		    }
703 		    /* Terminate & pad, if necessary */
704 		    cnt = opts->m_len % 4;
705 		    if (cnt) {
706 			    for (; cnt < 4; cnt++) {
707 				    *(mtod(opts, caddr_t) + opts->m_len) =
708 					IPOPT_EOL;
709 				    opts->m_len++;
710 			    }
711 		    }
712 #ifdef ICMPPRINTFS
713 		    if (icmpprintfs)
714 			    printf("%d\n", opts->m_len);
715 #endif
716 		}
717 		/*
718 		 * Now strip out original options by copying rest of first
719 		 * mbuf's data back, and adjust the IP length.
720 		 */
721 		ip->ip_len -= optlen;
722 		ip->ip_v = IPVERSION;
723 		ip->ip_hl = 5;
724 		m->m_len -= optlen;
725 		if (m->m_flags & M_PKTHDR)
726 			m->m_pkthdr.len -= optlen;
727 		optlen += sizeof(struct ip);
728 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
729 			 (unsigned)(m->m_len - sizeof(struct ip)));
730 	}
731 	m->m_flags &= ~(M_BCAST|M_MCAST);
732 	icmp_send(m, opts, ro);
733 done:
734 	if (opts)
735 		(void)m_free(opts);
736 	if (ro && ro->ro_rt)
737 		RTFREE(ro->ro_rt);
738 }
739 
740 /*
741  * Send an icmp packet back to the ip level,
742  * after supplying a checksum.
743  */
744 static void
745 icmp_send(m, opts, rt)
746 	register struct mbuf *m;
747 	struct mbuf *opts;
748 	struct route *rt;
749 {
750 	register struct ip *ip = mtod(m, struct ip *);
751 	register int hlen;
752 	register struct icmp *icp;
753 
754 	hlen = ip->ip_hl << 2;
755 	m->m_data += hlen;
756 	m->m_len -= hlen;
757 	icp = mtod(m, struct icmp *);
758 	icp->icmp_cksum = 0;
759 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
760 	m->m_data -= hlen;
761 	m->m_len += hlen;
762 	m->m_pkthdr.rcvif = (struct ifnet *)0;
763 #ifdef ICMPPRINTFS
764 	if (icmpprintfs) {
765 		char buf[4 * sizeof "123"];
766 		strcpy(buf, inet_ntoa(ip->ip_dst));
767 		printf("icmp_send dst %s src %s\n",
768 		       buf, inet_ntoa(ip->ip_src));
769 	}
770 #endif
771 	(void) ip_output(m, opts, rt, 0, NULL, NULL);
772 }
773 
774 n_time
775 iptime()
776 {
777 	struct timeval atv;
778 	u_long t;
779 
780 	getmicrotime(&atv);
781 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
782 	return (htonl(t));
783 }
784 
785 #if 1
786 /*
787  * Return the next larger or smaller MTU plateau (table from RFC 1191)
788  * given current value MTU.  If DIR is less than zero, a larger plateau
789  * is returned; otherwise, a smaller value is returned.
790  */
791 static int
792 ip_next_mtu(mtu, dir)
793 	int mtu;
794 	int dir;
795 {
796 	static int mtutab[] = {
797 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
798 		68, 0
799 	};
800 	int i;
801 
802 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
803 		if (mtu >= mtutab[i])
804 			break;
805 	}
806 
807 	if (dir < 0) {
808 		if (i == 0) {
809 			return 0;
810 		} else {
811 			return mtutab[i - 1];
812 		}
813 	} else {
814 		if (mtutab[i] == 0) {
815 			return 0;
816 		} else if(mtu > mtutab[i]) {
817 			return mtutab[i];
818 		} else {
819 			return mtutab[i + 1];
820 		}
821 	}
822 }
823 #endif
824 
825 
826 /*
827  * badport_bandlim() - check for ICMP bandwidth limit
828  *
829  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
830  *	hit our bandwidth limit and it is not ok.
831  *
832  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
833  *
834  *	For now we separate the TCP and UDP subsystems w/ different 'which'
835  *	values.  We may eventually remove this separation (and simplify the
836  *	code further).
837  *
838  *	Note that the printing of the error message is delayed so we can
839  *	properly print the icmp error rate that the system was trying to do
840  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
841  *	the 'final' error, but it doesn't make sense to solve the printing
842  *	delay with more complex code.
843  */
844 
845 int
846 badport_bandlim(int which)
847 {
848 #define	N(a)	(sizeof (a) / sizeof (a[0]))
849 	static struct rate {
850 		const char	*type;
851 		struct timeval	lasttime;
852 		int		curpps;;
853 	} rates[BANDLIM_MAX+1] = {
854 		{ "icmp unreach response" },
855 		{ "icmp ping response" },
856 		{ "icmp tstamp response" },
857 		{ "closed port RST response" },
858 		{ "open port RST response" }
859 	};
860 
861 	/*
862 	 * Return ok status if feature disabled or argument out of range.
863 	 */
864 	if (icmplim > 0 && (u_int) which < N(rates)) {
865 		struct rate *r = &rates[which];
866 		int opps = r->curpps;
867 
868 		if (!ppsratecheck(&r->lasttime, &r->curpps, icmplim))
869 			return -1;	/* discard packet */
870 		/*
871 		 * If we've dropped below the threshold after having
872 		 * rate-limited traffic print the message.  This preserves
873 		 * the previous behaviour at the expense of added complexity.
874 		 */
875 		if (icmplim_output && opps > icmplim)
876 			printf("Limiting %s from %d to %d packets/sec\n",
877 				r->type, opps, icmplim);
878 	}
879 	return 0;			/* okay to send packet */
880 #undef N
881 }
882