xref: /freebsd/sys/netinet/ip_icmp.c (revision f9218d3d4fd34f082473b3a021c6d4d109fb47cf)
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, "");
88 
89 static int	drop_redirect = 0;
90 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
91 	&drop_redirect, 0, "");
92 
93 static int	log_redirect = 0;
94 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
95 	&log_redirect, 0, "");
96 
97 static int      icmplim = 200;
98 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
99 	&icmplim, 0, "");
100 
101 static int	icmplim_output = 1;
102 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
103 	&icmplim_output, 0, "");
104 
105 /*
106  * ICMP broadcast echo sysctl
107  */
108 
109 static int	icmpbmcastecho = 0;
110 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
111 	&icmpbmcastecho, 0, "");
112 
113 
114 #ifdef ICMPPRINTFS
115 int	icmpprintfs = 0;
116 #endif
117 
118 static void	icmp_reflect(struct mbuf *);
119 static void	icmp_send(struct mbuf *, struct mbuf *, struct route *);
120 static int	ip_next_mtu(int, int);
121 
122 extern	struct protosw inetsw[];
123 
124 /*
125  * Generate an error packet of type error
126  * in response to bad packet ip.
127  */
128 void
129 icmp_error(n, type, code, dest, destifp)
130 	struct mbuf *n;
131 	int type, code;
132 	n_long dest;
133 	struct ifnet *destifp;
134 {
135 	register struct ip *oip = mtod(n, struct ip *), *nip;
136 	register unsigned oiplen = oip->ip_hl << 2;
137 	register struct icmp *icp;
138 	register struct mbuf *m;
139 	unsigned icmplen;
140 
141 #ifdef ICMPPRINTFS
142 	if (icmpprintfs)
143 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
144 #endif
145 	if (type != ICMP_REDIRECT)
146 		icmpstat.icps_error++;
147 	/*
148 	 * Don't send error if not the first fragment of message.
149 	 * Don't error if the old packet protocol was ICMP
150 	 * error message, only known informational types.
151 	 */
152 	if (oip->ip_off &~ (IP_MF|IP_DF))
153 		goto freeit;
154 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
155 	  n->m_len >= oiplen + ICMP_MINLEN &&
156 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
157 		icmpstat.icps_oldicmp++;
158 		goto freeit;
159 	}
160 	/* Don't send error in response to a multicast or broadcast packet */
161 	if (n->m_flags & (M_BCAST|M_MCAST))
162 		goto freeit;
163 	/*
164 	 * First, formulate icmp message
165 	 */
166 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
167 	if (m == NULL)
168 		goto freeit;
169 #ifdef MAC
170 	mac_create_mbuf_netlayer(n, m);
171 #endif
172 	icmplen = min(oiplen + 8, oip->ip_len);
173 	if (icmplen < sizeof(struct ip))
174 		panic("icmp_error: bad length");
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 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
201 	nip = &icp->icmp_ip;
202 
203 	/*
204 	 * Convert fields to network representation.
205 	 */
206 	nip->ip_len = htons(nip->ip_len);
207 	nip->ip_off = htons(nip->ip_off);
208 
209 	/*
210 	 * Now, copy old ip header (without options)
211 	 * in front of icmp message.
212 	 */
213 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
214 		panic("icmp len");
215 	m->m_data -= sizeof(struct ip);
216 	m->m_len += sizeof(struct ip);
217 	m->m_pkthdr.len = m->m_len;
218 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
219 	nip = mtod(m, struct ip *);
220 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
221 	nip->ip_len = m->m_len;
222 	nip->ip_v = IPVERSION;
223 	nip->ip_hl = 5;
224 	nip->ip_p = IPPROTO_ICMP;
225 	nip->ip_tos = 0;
226 	icmp_reflect(m);
227 
228 freeit:
229 	m_freem(n);
230 }
231 
232 static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET };
233 static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET };
234 static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET };
235 
236 /*
237  * Process a received ICMP message.
238  */
239 void
240 icmp_input(m, off)
241 	register struct mbuf *m;
242 	int off;
243 {
244 	int hlen = off;
245 	register struct icmp *icp;
246 	register struct ip *ip = mtod(m, struct ip *);
247 	int icmplen = ip->ip_len;
248 	register int i;
249 	struct in_ifaddr *ia;
250 	void (*ctlfunc)(int, struct sockaddr *, void *);
251 	int code;
252 
253 	/*
254 	 * Locate icmp structure in mbuf, and check
255 	 * that not corrupted and of at least minimum length.
256 	 */
257 #ifdef ICMPPRINTFS
258 	if (icmpprintfs) {
259 		char buf[4 * sizeof "123"];
260 		strcpy(buf, inet_ntoa(ip->ip_src));
261 		printf("icmp_input from %s to %s, len %d\n",
262 		       buf, inet_ntoa(ip->ip_dst), icmplen);
263 	}
264 #endif
265 	if (icmplen < ICMP_MINLEN) {
266 		icmpstat.icps_tooshort++;
267 		goto freeit;
268 	}
269 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
270 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
271 		icmpstat.icps_tooshort++;
272 		return;
273 	}
274 	ip = mtod(m, struct ip *);
275 	m->m_len -= hlen;
276 	m->m_data += hlen;
277 	icp = mtod(m, struct icmp *);
278 	if (in_cksum(m, icmplen)) {
279 		icmpstat.icps_checksum++;
280 		goto freeit;
281 	}
282 	m->m_len += hlen;
283 	m->m_data -= hlen;
284 
285 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
286 		/*
287 		 * Deliver very specific ICMP type only.
288 		 */
289 		switch (icp->icmp_type) {
290 		case ICMP_UNREACH:
291 		case ICMP_TIMXCEED:
292 			break;
293 		default:
294 			goto freeit;
295 		}
296 	}
297 
298 #ifdef ICMPPRINTFS
299 	if (icmpprintfs)
300 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
301 		    icp->icmp_code);
302 #endif
303 
304 	/*
305 	 * Message type specific processing.
306 	 */
307 	if (icp->icmp_type > ICMP_MAXTYPE)
308 		goto raw;
309 	icmpstat.icps_inhist[icp->icmp_type]++;
310 	code = icp->icmp_code;
311 	switch (icp->icmp_type) {
312 
313 	case ICMP_UNREACH:
314 		switch (code) {
315 			case ICMP_UNREACH_NET:
316 			case ICMP_UNREACH_HOST:
317 			case ICMP_UNREACH_SRCFAIL:
318 			case ICMP_UNREACH_NET_UNKNOWN:
319 			case ICMP_UNREACH_HOST_UNKNOWN:
320 			case ICMP_UNREACH_ISOLATED:
321 			case ICMP_UNREACH_TOSNET:
322 			case ICMP_UNREACH_TOSHOST:
323 			case ICMP_UNREACH_HOST_PRECEDENCE:
324 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
325 				code = PRC_UNREACH_NET;
326 				break;
327 
328 			case ICMP_UNREACH_NEEDFRAG:
329 				code = PRC_MSGSIZE;
330 				break;
331 
332 			/*
333 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
334 			 * Treat subcodes 2,3 as immediate RST
335 			 */
336 			case ICMP_UNREACH_PROTOCOL:
337 			case ICMP_UNREACH_PORT:
338 				code = PRC_UNREACH_PORT;
339 				break;
340 
341 			case ICMP_UNREACH_NET_PROHIB:
342 			case ICMP_UNREACH_HOST_PROHIB:
343 			case ICMP_UNREACH_FILTER_PROHIB:
344 				code = PRC_UNREACH_ADMIN_PROHIB;
345 				break;
346 
347 			default:
348 				goto badcode;
349 		}
350 		goto deliver;
351 
352 	case ICMP_TIMXCEED:
353 		if (code > 1)
354 			goto badcode;
355 		code += PRC_TIMXCEED_INTRANS;
356 		goto deliver;
357 
358 	case ICMP_PARAMPROB:
359 		if (code > 1)
360 			goto badcode;
361 		code = PRC_PARAMPROB;
362 		goto deliver;
363 
364 	case ICMP_SOURCEQUENCH:
365 		if (code)
366 			goto badcode;
367 		code = PRC_QUENCH;
368 	deliver:
369 		/*
370 		 * Problem with datagram; advise higher level routines.
371 		 */
372 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
373 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
374 			icmpstat.icps_badlen++;
375 			goto freeit;
376 		}
377 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
378 		/* Discard ICMP's in response to multicast packets */
379 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
380 			goto badcode;
381 #ifdef ICMPPRINTFS
382 		if (icmpprintfs)
383 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
384 #endif
385 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
386 #if 1
387 		/*
388 		 * MTU discovery:
389 		 * If we got a needfrag and there is a host route to the
390 		 * original destination, and the MTU is not locked, then
391 		 * set the MTU in the route to the suggested new value
392 		 * (if given) and then notify as usual.  The ULPs will
393 		 * notice that the MTU has changed and adapt accordingly.
394 		 * If no new MTU was suggested, then we guess a new one
395 		 * less than the current value.  If the new MTU is
396 		 * unreasonably small (arbitrarily set at 296), then
397 		 * we reset the MTU to the interface value and enable the
398 		 * lock bit, indicating that we are no longer doing MTU
399 		 * discovery.
400 		 */
401 		if (code == PRC_MSGSIZE) {
402 			struct rtentry *rt;
403 			int mtu;
404 
405 			rt = rtalloc1((struct sockaddr *)&icmpsrc, 0,
406 				      RTF_CLONING | RTF_PRCLONING);
407 			if (rt && (rt->rt_flags & RTF_HOST)
408 			    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
409 				mtu = ntohs(icp->icmp_nextmtu);
410 				if (!mtu)
411 					mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
412 							  1);
413 #ifdef DEBUG_MTUDISC
414 				printf("MTU for %s reduced to %d\n",
415 					inet_ntoa(icmpsrc.sin_addr), mtu);
416 #endif
417 				if (mtu < 296) {
418 					/* rt->rt_rmx.rmx_mtu =
419 						rt->rt_ifp->if_mtu; */
420 					rt->rt_rmx.rmx_locks |= RTV_MTU;
421 				} else if (rt->rt_rmx.rmx_mtu > mtu) {
422 					rt->rt_rmx.rmx_mtu = mtu;
423 				}
424 			}
425 			if (rt)
426 				RTFREE(rt);
427 		}
428 
429 #endif
430 		/*
431 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
432 		 * notification to TCP layer.
433 		 */
434 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
435 		if (ctlfunc)
436 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
437 				   (void *)&icp->icmp_ip);
438 		break;
439 
440 	badcode:
441 		icmpstat.icps_badcode++;
442 		break;
443 
444 	case ICMP_ECHO:
445 		if (!icmpbmcastecho
446 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
447 			icmpstat.icps_bmcastecho++;
448 			break;
449 		}
450 		icp->icmp_type = ICMP_ECHOREPLY;
451 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
452 			goto freeit;
453 		else
454 			goto reflect;
455 
456 	case ICMP_TSTAMP:
457 		if (!icmpbmcastecho
458 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
459 			icmpstat.icps_bmcasttstamp++;
460 			break;
461 		}
462 		if (icmplen < ICMP_TSLEN) {
463 			icmpstat.icps_badlen++;
464 			break;
465 		}
466 		icp->icmp_type = ICMP_TSTAMPREPLY;
467 		icp->icmp_rtime = iptime();
468 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
469 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
470 			goto freeit;
471 		else
472 			goto reflect;
473 
474 	case ICMP_MASKREQ:
475 		if (icmpmaskrepl == 0)
476 			break;
477 		/*
478 		 * We are not able to respond with all ones broadcast
479 		 * unless we receive it over a point-to-point interface.
480 		 */
481 		if (icmplen < ICMP_MASKLEN)
482 			break;
483 		switch (ip->ip_dst.s_addr) {
484 
485 		case INADDR_BROADCAST:
486 		case INADDR_ANY:
487 			icmpdst.sin_addr = ip->ip_src;
488 			break;
489 
490 		default:
491 			icmpdst.sin_addr = ip->ip_dst;
492 		}
493 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
494 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
495 		if (ia == 0)
496 			break;
497 		if (ia->ia_ifp == 0)
498 			break;
499 		icp->icmp_type = ICMP_MASKREPLY;
500 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
501 		if (ip->ip_src.s_addr == 0) {
502 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
503 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
504 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
505 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
506 		}
507 reflect:
508 		ip->ip_len += hlen;	/* since ip_input deducts this */
509 		icmpstat.icps_reflect++;
510 		icmpstat.icps_outhist[icp->icmp_type]++;
511 		icmp_reflect(m);
512 		return;
513 
514 	case ICMP_REDIRECT:
515 		if (log_redirect) {
516 			u_long src, dst, gw;
517 
518 			src = ntohl(ip->ip_src.s_addr);
519 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
520 			gw = ntohl(icp->icmp_gwaddr.s_addr);
521 			printf("icmp redirect from %d.%d.%d.%d: "
522 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
523 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
524 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
525 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
526 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
527 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
528 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
529 		}
530 		if (drop_redirect)
531 			break;
532 		if (code > 3)
533 			goto badcode;
534 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
535 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
536 			icmpstat.icps_badlen++;
537 			break;
538 		}
539 		/*
540 		 * Short circuit routing redirects to force
541 		 * immediate change in the kernel's routing
542 		 * tables.  The message is also handed to anyone
543 		 * listening on a raw socket (e.g. the routing
544 		 * daemon for use in updating its tables).
545 		 */
546 		icmpgw.sin_addr = ip->ip_src;
547 		icmpdst.sin_addr = icp->icmp_gwaddr;
548 #ifdef	ICMPPRINTFS
549 		if (icmpprintfs) {
550 			char buf[4 * sizeof "123"];
551 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
552 
553 			printf("redirect dst %s to %s\n",
554 			       buf, inet_ntoa(icp->icmp_gwaddr));
555 		}
556 #endif
557 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
558 		rtredirect((struct sockaddr *)&icmpsrc,
559 		  (struct sockaddr *)&icmpdst,
560 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
561 		  (struct sockaddr *)&icmpgw, (struct rtentry **)0);
562 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
563 #ifdef IPSEC
564 		key_sa_routechange((struct sockaddr *)&icmpsrc);
565 #endif
566 		break;
567 
568 	/*
569 	 * No kernel processing for the following;
570 	 * just fall through to send to raw listener.
571 	 */
572 	case ICMP_ECHOREPLY:
573 	case ICMP_ROUTERADVERT:
574 	case ICMP_ROUTERSOLICIT:
575 	case ICMP_TSTAMPREPLY:
576 	case ICMP_IREQREPLY:
577 	case ICMP_MASKREPLY:
578 	default:
579 		break;
580 	}
581 
582 raw:
583 	rip_input(m, off);
584 	return;
585 
586 freeit:
587 	m_freem(m);
588 }
589 
590 /*
591  * Reflect the ip packet back to the source
592  */
593 static void
594 icmp_reflect(m)
595 	struct mbuf *m;
596 {
597 	struct ip *ip = mtod(m, struct ip *);
598 	struct ifaddr *ifa;
599 	struct in_ifaddr *ia;
600 	struct in_addr t;
601 	struct mbuf *opts = 0;
602 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
603 	struct route *ro = NULL, rt;
604 
605 	if (!in_canforward(ip->ip_src) &&
606 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
607 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
608 		m_freem(m);	/* Bad return address */
609 		icmpstat.icps_badaddr++;
610 		goto done;	/* Ip_output() will check for broadcast */
611 	}
612 	t = ip->ip_dst;
613 	ip->ip_dst = ip->ip_src;
614 	ro = &rt;
615 	bzero(ro, sizeof(*ro));
616 	/*
617 	 * If the incoming packet was addressed directly to us,
618 	 * use dst as the src for the reply.  Otherwise (broadcast
619 	 * or anonymous), use the address which corresponds
620 	 * to the incoming interface.
621 	 */
622 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
623 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
624 			goto match;
625 	if (m->m_pkthdr.rcvif != NULL &&
626 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
627 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
628 			if (ifa->ifa_addr->sa_family != AF_INET)
629 				continue;
630 			ia = ifatoia(ifa);
631 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
632 			    t.s_addr)
633 				goto match;
634 		}
635 	}
636 	ia = ip_rtaddr(ip->ip_dst, ro);
637 	/* We need a route to do anything useful. */
638 	if (ia == NULL) {
639 		m_freem(m);
640 		icmpstat.icps_noroute++;
641 		goto done;
642 	}
643 match:
644 	t = IA_SIN(ia)->sin_addr;
645 	ip->ip_src = t;
646 	ip->ip_ttl = ip_defttl;
647 
648 	if (optlen > 0) {
649 		register u_char *cp;
650 		int opt, cnt;
651 		u_int len;
652 
653 		/*
654 		 * Retrieve any source routing from the incoming packet;
655 		 * add on any record-route or timestamp options.
656 		 */
657 		cp = (u_char *) (ip + 1);
658 		if ((opts = ip_srcroute()) == 0 &&
659 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
660 			opts->m_len = sizeof(struct in_addr);
661 			mtod(opts, struct in_addr *)->s_addr = 0;
662 		}
663 		if (opts) {
664 #ifdef ICMPPRINTFS
665 		    if (icmpprintfs)
666 			    printf("icmp_reflect optlen %d rt %d => ",
667 				optlen, opts->m_len);
668 #endif
669 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
670 			    opt = cp[IPOPT_OPTVAL];
671 			    if (opt == IPOPT_EOL)
672 				    break;
673 			    if (opt == IPOPT_NOP)
674 				    len = 1;
675 			    else {
676 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
677 					    break;
678 				    len = cp[IPOPT_OLEN];
679 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
680 				        len > cnt)
681 					    break;
682 			    }
683 			    /*
684 			     * Should check for overflow, but it "can't happen"
685 			     */
686 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
687 				opt == IPOPT_SECURITY) {
688 				    bcopy((caddr_t)cp,
689 					mtod(opts, caddr_t) + opts->m_len, len);
690 				    opts->m_len += len;
691 			    }
692 		    }
693 		    /* Terminate & pad, if necessary */
694 		    cnt = opts->m_len % 4;
695 		    if (cnt) {
696 			    for (; cnt < 4; cnt++) {
697 				    *(mtod(opts, caddr_t) + opts->m_len) =
698 					IPOPT_EOL;
699 				    opts->m_len++;
700 			    }
701 		    }
702 #ifdef ICMPPRINTFS
703 		    if (icmpprintfs)
704 			    printf("%d\n", opts->m_len);
705 #endif
706 		}
707 		/*
708 		 * Now strip out original options by copying rest of first
709 		 * mbuf's data back, and adjust the IP length.
710 		 */
711 		ip->ip_len -= optlen;
712 		ip->ip_v = IPVERSION;
713 		ip->ip_hl = 5;
714 		m->m_len -= optlen;
715 		if (m->m_flags & M_PKTHDR)
716 			m->m_pkthdr.len -= optlen;
717 		optlen += sizeof(struct ip);
718 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
719 			 (unsigned)(m->m_len - sizeof(struct ip)));
720 	}
721 	m->m_flags &= ~(M_BCAST|M_MCAST);
722 	icmp_send(m, opts, ro);
723 done:
724 	if (opts)
725 		(void)m_free(opts);
726 	if (ro && ro->ro_rt)
727 		RTFREE(ro->ro_rt);
728 }
729 
730 /*
731  * Send an icmp packet back to the ip level,
732  * after supplying a checksum.
733  */
734 static void
735 icmp_send(m, opts, rt)
736 	register struct mbuf *m;
737 	struct mbuf *opts;
738 	struct route *rt;
739 {
740 	register struct ip *ip = mtod(m, struct ip *);
741 	register int hlen;
742 	register struct icmp *icp;
743 
744 	hlen = ip->ip_hl << 2;
745 	m->m_data += hlen;
746 	m->m_len -= hlen;
747 	icp = mtod(m, struct icmp *);
748 	icp->icmp_cksum = 0;
749 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
750 	m->m_data -= hlen;
751 	m->m_len += hlen;
752 	m->m_pkthdr.rcvif = (struct ifnet *)0;
753 #ifdef ICMPPRINTFS
754 	if (icmpprintfs) {
755 		char buf[4 * sizeof "123"];
756 		strcpy(buf, inet_ntoa(ip->ip_dst));
757 		printf("icmp_send dst %s src %s\n",
758 		       buf, inet_ntoa(ip->ip_src));
759 	}
760 #endif
761 	(void) ip_output(m, opts, rt, 0, NULL, NULL);
762 }
763 
764 n_time
765 iptime()
766 {
767 	struct timeval atv;
768 	u_long t;
769 
770 	getmicrotime(&atv);
771 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
772 	return (htonl(t));
773 }
774 
775 #if 1
776 /*
777  * Return the next larger or smaller MTU plateau (table from RFC 1191)
778  * given current value MTU.  If DIR is less than zero, a larger plateau
779  * is returned; otherwise, a smaller value is returned.
780  */
781 static int
782 ip_next_mtu(mtu, dir)
783 	int mtu;
784 	int dir;
785 {
786 	static int mtutab[] = {
787 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
788 		68, 0
789 	};
790 	int i;
791 
792 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
793 		if (mtu >= mtutab[i])
794 			break;
795 	}
796 
797 	if (dir < 0) {
798 		if (i == 0) {
799 			return 0;
800 		} else {
801 			return mtutab[i - 1];
802 		}
803 	} else {
804 		if (mtutab[i] == 0) {
805 			return 0;
806 		} else if(mtu > mtutab[i]) {
807 			return mtutab[i];
808 		} else {
809 			return mtutab[i + 1];
810 		}
811 	}
812 }
813 #endif
814 
815 
816 /*
817  * badport_bandlim() - check for ICMP bandwidth limit
818  *
819  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
820  *	hit our bandwidth limit and it is not ok.
821  *
822  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
823  *
824  *	For now we separate the TCP and UDP subsystems w/ different 'which'
825  *	values.  We may eventually remove this separation (and simplify the
826  *	code further).
827  *
828  *	Note that the printing of the error message is delayed so we can
829  *	properly print the icmp error rate that the system was trying to do
830  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
831  *	the 'final' error, but it doesn't make sense to solve the printing
832  *	delay with more complex code.
833  */
834 
835 int
836 badport_bandlim(int which)
837 {
838 #define	N(a)	(sizeof (a) / sizeof (a[0]))
839 	static struct rate {
840 		const char	*type;
841 		struct timeval	lasttime;
842 		int		curpps;;
843 	} rates[BANDLIM_MAX+1] = {
844 		{ "icmp unreach response" },
845 		{ "icmp ping response" },
846 		{ "icmp tstamp response" },
847 		{ "closed port RST response" },
848 		{ "open port RST response" }
849 	};
850 
851 	/*
852 	 * Return ok status if feature disabled or argument out of range.
853 	 */
854 	if (icmplim > 0 && (u_int) which < N(rates)) {
855 		struct rate *r = &rates[which];
856 		int opps = r->curpps;
857 
858 		if (!ppsratecheck(&r->lasttime, &r->curpps, icmplim))
859 			return -1;	/* discard packet */
860 		/*
861 		 * If we've dropped below the threshold after having
862 		 * rate-limited traffic print the message.  This preserves
863 		 * the previous behaviour at the expense of added complexity.
864 		 */
865 		if (icmplim_output && opps > icmplim)
866 			printf("Limiting %s from %d to %d packets/sec\n",
867 				r->type, opps, icmplim);
868 	}
869 	return 0;			/* okay to send packet */
870 #undef N
871 }
872