xref: /freebsd/sys/netinet/ip_icmp.c (revision d429ea332342fcb98d27a350d0c4944bf9aec3f9)
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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
30  * $FreeBSD$
31  */
32 
33 #include "opt_ipsec.h"
34 #include "opt_mac.h"
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/mac.h>
39 #include <sys/mbuf.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/sysctl.h>
45 
46 #include <net/if.h>
47 #include <net/if_types.h>
48 #include <net/route.h>
49 
50 #include <netinet/in.h>
51 #include <netinet/in_pcb.h>
52 #include <netinet/in_systm.h>
53 #include <netinet/in_var.h>
54 #include <netinet/ip.h>
55 #include <netinet/ip_icmp.h>
56 #include <netinet/ip_var.h>
57 #include <netinet/tcp.h>
58 #include <netinet/tcp_var.h>
59 #include <netinet/tcpip.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 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 static char	reply_src[IFNAMSIZ];
110 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
111 	&reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
112 
113 /*
114  * ICMP broadcast echo sysctl
115  */
116 
117 static int	icmpbmcastecho = 0;
118 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
119 	&icmpbmcastecho, 0, "");
120 
121 
122 #ifdef ICMPPRINTFS
123 int	icmpprintfs = 0;
124 #endif
125 
126 static void	icmp_reflect(struct mbuf *);
127 static void	icmp_send(struct mbuf *, struct mbuf *);
128 
129 extern	struct protosw inetsw[];
130 
131 /*
132  * Generate an error packet of type error
133  * in response to bad packet ip.
134  */
135 void
136 icmp_error(n, type, code, dest, mtu)
137 	struct mbuf *n;
138 	int type, code;
139 	n_long dest;
140 	int mtu;
141 {
142 	register struct ip *oip = mtod(n, struct ip *), *nip;
143 	register unsigned oiplen = oip->ip_hl << 2;
144 	register struct icmp *icp;
145 	register struct mbuf *m;
146 	unsigned icmplen;
147 
148 #ifdef ICMPPRINTFS
149 	if (icmpprintfs)
150 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
151 #endif
152 	if (type != ICMP_REDIRECT)
153 		icmpstat.icps_error++;
154 	/*
155 	 * Don't send error if the original packet was encrypted.
156 	 * Don't send error if not the first fragment of message.
157 	 * Don't error if the old packet protocol was ICMP
158 	 * error message, only known informational types.
159 	 */
160 	if (n->m_flags & M_DECRYPTED)
161 		goto freeit;
162 	if (oip->ip_off &~ (IP_MF|IP_DF))
163 		goto freeit;
164 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
165 	  n->m_len >= oiplen + ICMP_MINLEN &&
166 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
167 		icmpstat.icps_oldicmp++;
168 		goto freeit;
169 	}
170 	/* Don't send error in response to a multicast or broadcast packet */
171 	if (n->m_flags & (M_BCAST|M_MCAST))
172 		goto freeit;
173 	/*
174 	 * First, formulate icmp message
175 	 */
176 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
177 	if (m == NULL)
178 		goto freeit;
179 #ifdef MAC
180 	mac_create_mbuf_netlayer(n, m);
181 #endif
182 	icmplen = min(oiplen + 8, oip->ip_len);
183 	if (icmplen < sizeof(struct ip))
184 		panic("icmp_error: bad length");
185 	m->m_len = icmplen + ICMP_MINLEN;
186 	MH_ALIGN(m, m->m_len);
187 	icp = mtod(m, struct icmp *);
188 	if ((u_int)type > ICMP_MAXTYPE)
189 		panic("icmp_error");
190 	icmpstat.icps_outhist[type]++;
191 	icp->icmp_type = type;
192 	if (type == ICMP_REDIRECT)
193 		icp->icmp_gwaddr.s_addr = dest;
194 	else {
195 		icp->icmp_void = 0;
196 		/*
197 		 * The following assignments assume an overlay with the
198 		 * zeroed icmp_void field.
199 		 */
200 		if (type == ICMP_PARAMPROB) {
201 			icp->icmp_pptr = code;
202 			code = 0;
203 		} else if (type == ICMP_UNREACH &&
204 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
205 			icp->icmp_nextmtu = htons(mtu);
206 		}
207 	}
208 
209 	icp->icmp_code = code;
210 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
211 	nip = &icp->icmp_ip;
212 
213 	/*
214 	 * Convert fields to network representation.
215 	 */
216 	nip->ip_len = htons(nip->ip_len);
217 	nip->ip_off = htons(nip->ip_off);
218 
219 	/*
220 	 * Now, copy old ip header (without options)
221 	 * in front of icmp message.
222 	 */
223 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
224 		panic("icmp len");
225 	/*
226 	 * If the original mbuf was meant to bypass the firewall, the error
227 	 * reply should bypass as well.
228 	 */
229 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
230 	m->m_data -= sizeof(struct ip);
231 	m->m_len += sizeof(struct ip);
232 	m->m_pkthdr.len = m->m_len;
233 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
234 	nip = mtod(m, struct ip *);
235 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
236 	nip->ip_len = m->m_len;
237 	nip->ip_v = IPVERSION;
238 	nip->ip_hl = 5;
239 	nip->ip_p = IPPROTO_ICMP;
240 	nip->ip_tos = 0;
241 	icmp_reflect(m);
242 
243 freeit:
244 	m_freem(n);
245 }
246 
247 /*
248  * Process a received ICMP message.
249  */
250 void
251 icmp_input(m, off)
252 	struct mbuf *m;
253 	int off;
254 {
255 	struct icmp *icp;
256 	struct in_ifaddr *ia;
257 	struct ip *ip = mtod(m, struct ip *);
258 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
259 	int hlen = off;
260 	int icmplen = ip->ip_len;
261 	int i, code;
262 	void (*ctlfunc)(int, struct sockaddr *, void *);
263 
264 	/*
265 	 * Locate icmp structure in mbuf, and check
266 	 * that not corrupted and of at least minimum length.
267 	 */
268 #ifdef ICMPPRINTFS
269 	if (icmpprintfs) {
270 		char buf[4 * sizeof "123"];
271 		strcpy(buf, inet_ntoa(ip->ip_src));
272 		printf("icmp_input from %s to %s, len %d\n",
273 		       buf, inet_ntoa(ip->ip_dst), icmplen);
274 	}
275 #endif
276 	if (icmplen < ICMP_MINLEN) {
277 		icmpstat.icps_tooshort++;
278 		goto freeit;
279 	}
280 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
281 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
282 		icmpstat.icps_tooshort++;
283 		return;
284 	}
285 	ip = mtod(m, struct ip *);
286 	m->m_len -= hlen;
287 	m->m_data += hlen;
288 	icp = mtod(m, struct icmp *);
289 	if (in_cksum(m, icmplen)) {
290 		icmpstat.icps_checksum++;
291 		goto freeit;
292 	}
293 	m->m_len += hlen;
294 	m->m_data -= hlen;
295 
296 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
297 		/*
298 		 * Deliver very specific ICMP type only.
299 		 */
300 		switch (icp->icmp_type) {
301 		case ICMP_UNREACH:
302 		case ICMP_TIMXCEED:
303 			break;
304 		default:
305 			goto freeit;
306 		}
307 	}
308 
309 #ifdef ICMPPRINTFS
310 	if (icmpprintfs)
311 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
312 		    icp->icmp_code);
313 #endif
314 
315 	/*
316 	 * Message type specific processing.
317 	 */
318 	if (icp->icmp_type > ICMP_MAXTYPE)
319 		goto raw;
320 
321 	/* Initialize */
322 	bzero(&icmpsrc, sizeof(icmpsrc));
323 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
324 	icmpsrc.sin_family = AF_INET;
325 	bzero(&icmpdst, sizeof(icmpdst));
326 	icmpdst.sin_len = sizeof(struct sockaddr_in);
327 	icmpdst.sin_family = AF_INET;
328 	bzero(&icmpgw, sizeof(icmpgw));
329 	icmpgw.sin_len = sizeof(struct sockaddr_in);
330 	icmpgw.sin_family = AF_INET;
331 
332 	icmpstat.icps_inhist[icp->icmp_type]++;
333 	code = icp->icmp_code;
334 	switch (icp->icmp_type) {
335 
336 	case ICMP_UNREACH:
337 		switch (code) {
338 			case ICMP_UNREACH_NET:
339 			case ICMP_UNREACH_HOST:
340 			case ICMP_UNREACH_SRCFAIL:
341 			case ICMP_UNREACH_NET_UNKNOWN:
342 			case ICMP_UNREACH_HOST_UNKNOWN:
343 			case ICMP_UNREACH_ISOLATED:
344 			case ICMP_UNREACH_TOSNET:
345 			case ICMP_UNREACH_TOSHOST:
346 			case ICMP_UNREACH_HOST_PRECEDENCE:
347 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
348 				code = PRC_UNREACH_NET;
349 				break;
350 
351 			case ICMP_UNREACH_NEEDFRAG:
352 				code = PRC_MSGSIZE;
353 				break;
354 
355 			/*
356 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
357 			 * Treat subcodes 2,3 as immediate RST
358 			 */
359 			case ICMP_UNREACH_PROTOCOL:
360 			case ICMP_UNREACH_PORT:
361 				code = PRC_UNREACH_PORT;
362 				break;
363 
364 			case ICMP_UNREACH_NET_PROHIB:
365 			case ICMP_UNREACH_HOST_PROHIB:
366 			case ICMP_UNREACH_FILTER_PROHIB:
367 				code = PRC_UNREACH_ADMIN_PROHIB;
368 				break;
369 
370 			default:
371 				goto badcode;
372 		}
373 		goto deliver;
374 
375 	case ICMP_TIMXCEED:
376 		if (code > 1)
377 			goto badcode;
378 		code += PRC_TIMXCEED_INTRANS;
379 		goto deliver;
380 
381 	case ICMP_PARAMPROB:
382 		if (code > 1)
383 			goto badcode;
384 		code = PRC_PARAMPROB;
385 		goto deliver;
386 
387 	case ICMP_SOURCEQUENCH:
388 		if (code)
389 			goto badcode;
390 		code = PRC_QUENCH;
391 	deliver:
392 		/*
393 		 * Problem with datagram; advise higher level routines.
394 		 */
395 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
396 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
397 			icmpstat.icps_badlen++;
398 			goto freeit;
399 		}
400 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
401 		/* Discard ICMP's in response to multicast packets */
402 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
403 			goto badcode;
404 #ifdef ICMPPRINTFS
405 		if (icmpprintfs)
406 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
407 #endif
408 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
409 		/*
410 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
411 		 * notification to TCP layer.
412 		 */
413 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
414 		if (ctlfunc)
415 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
416 				   (void *)&icp->icmp_ip);
417 		break;
418 
419 	badcode:
420 		icmpstat.icps_badcode++;
421 		break;
422 
423 	case ICMP_ECHO:
424 		if (!icmpbmcastecho
425 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
426 			icmpstat.icps_bmcastecho++;
427 			break;
428 		}
429 		icp->icmp_type = ICMP_ECHOREPLY;
430 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
431 			goto freeit;
432 		else
433 			goto reflect;
434 
435 	case ICMP_TSTAMP:
436 		if (!icmpbmcastecho
437 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
438 			icmpstat.icps_bmcasttstamp++;
439 			break;
440 		}
441 		if (icmplen < ICMP_TSLEN) {
442 			icmpstat.icps_badlen++;
443 			break;
444 		}
445 		icp->icmp_type = ICMP_TSTAMPREPLY;
446 		icp->icmp_rtime = iptime();
447 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
448 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
449 			goto freeit;
450 		else
451 			goto reflect;
452 
453 	case ICMP_MASKREQ:
454 		if (icmpmaskrepl == 0)
455 			break;
456 		/*
457 		 * We are not able to respond with all ones broadcast
458 		 * unless we receive it over a point-to-point interface.
459 		 */
460 		if (icmplen < ICMP_MASKLEN)
461 			break;
462 		switch (ip->ip_dst.s_addr) {
463 
464 		case INADDR_BROADCAST:
465 		case INADDR_ANY:
466 			icmpdst.sin_addr = ip->ip_src;
467 			break;
468 
469 		default:
470 			icmpdst.sin_addr = ip->ip_dst;
471 		}
472 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
473 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
474 		if (ia == 0)
475 			break;
476 		if (ia->ia_ifp == 0)
477 			break;
478 		icp->icmp_type = ICMP_MASKREPLY;
479 		if (icmpmaskfake == 0)
480 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
481 		else
482 			icp->icmp_mask = icmpmaskfake;
483 		if (ip->ip_src.s_addr == 0) {
484 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
485 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
486 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
487 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
488 		}
489 reflect:
490 		ip->ip_len += hlen;	/* since ip_input deducts this */
491 		icmpstat.icps_reflect++;
492 		icmpstat.icps_outhist[icp->icmp_type]++;
493 		icmp_reflect(m);
494 		return;
495 
496 	case ICMP_REDIRECT:
497 		if (log_redirect) {
498 			u_long src, dst, gw;
499 
500 			src = ntohl(ip->ip_src.s_addr);
501 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
502 			gw = ntohl(icp->icmp_gwaddr.s_addr);
503 			printf("icmp redirect from %d.%d.%d.%d: "
504 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
505 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
506 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
507 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
508 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
509 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
510 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
511 		}
512 		/*
513 		 * RFC1812 says we must ignore ICMP redirects if we
514 		 * are acting as router.
515 		 */
516 		if (drop_redirect || ipforwarding)
517 			break;
518 		if (code > 3)
519 			goto badcode;
520 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
521 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
522 			icmpstat.icps_badlen++;
523 			break;
524 		}
525 		/*
526 		 * Short circuit routing redirects to force
527 		 * immediate change in the kernel's routing
528 		 * tables.  The message is also handed to anyone
529 		 * listening on a raw socket (e.g. the routing
530 		 * daemon for use in updating its tables).
531 		 */
532 		icmpgw.sin_addr = ip->ip_src;
533 		icmpdst.sin_addr = icp->icmp_gwaddr;
534 #ifdef	ICMPPRINTFS
535 		if (icmpprintfs) {
536 			char buf[4 * sizeof "123"];
537 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
538 
539 			printf("redirect dst %s to %s\n",
540 			       buf, inet_ntoa(icp->icmp_gwaddr));
541 		}
542 #endif
543 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
544 		rtredirect((struct sockaddr *)&icmpsrc,
545 		  (struct sockaddr *)&icmpdst,
546 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
547 		  (struct sockaddr *)&icmpgw);
548 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
549 #ifdef IPSEC
550 		key_sa_routechange((struct sockaddr *)&icmpsrc);
551 #endif
552 		break;
553 
554 	/*
555 	 * No kernel processing for the following;
556 	 * just fall through to send to raw listener.
557 	 */
558 	case ICMP_ECHOREPLY:
559 	case ICMP_ROUTERADVERT:
560 	case ICMP_ROUTERSOLICIT:
561 	case ICMP_TSTAMPREPLY:
562 	case ICMP_IREQREPLY:
563 	case ICMP_MASKREPLY:
564 	default:
565 		break;
566 	}
567 
568 raw:
569 	rip_input(m, off);
570 	return;
571 
572 freeit:
573 	m_freem(m);
574 }
575 
576 /*
577  * Reflect the ip packet back to the source
578  */
579 static void
580 icmp_reflect(m)
581 	struct mbuf *m;
582 {
583 	struct ip *ip = mtod(m, struct ip *);
584 	struct ifaddr *ifa;
585 	struct ifnet *ifn;
586 	struct in_ifaddr *ia;
587 	struct in_addr t;
588 	struct mbuf *opts = 0;
589 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
590 
591 	if (!in_canforward(ip->ip_src) &&
592 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
593 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
594 		m_freem(m);	/* Bad return address */
595 		icmpstat.icps_badaddr++;
596 		goto done;	/* Ip_output() will check for broadcast */
597 	}
598 	t = ip->ip_dst;
599 	ip->ip_dst = ip->ip_src;
600 
601 	/*
602 	 * Source selection for ICMP replies:
603 	 *
604 	 * If the incoming packet was addressed directly to one of our
605 	 * own addresses, use dst as the src for the reply.
606 	 */
607 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
608 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
609 			goto match;
610 	/*
611 	 * If the incoming packet was addressed to one of our broadcast
612 	 * addresses, use the first non-broadcast address which corresponds
613 	 * to the incoming interface.
614 	 */
615 	if (m->m_pkthdr.rcvif != NULL &&
616 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
617 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
618 			if (ifa->ifa_addr->sa_family != AF_INET)
619 				continue;
620 			ia = ifatoia(ifa);
621 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
622 			    t.s_addr)
623 				goto match;
624 		}
625 	}
626 	/*
627 	 * If the incoming packet was not addressed directly to us, use
628 	 * designated interface for icmp replies specified by sysctl
629 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
630 	 * with normal source selection.
631 	 */
632 	if (reply_src[0] != '\0' && (ifn = ifunit(reply_src))) {
633 		TAILQ_FOREACH(ifa, &ifn->if_addrhead, ifa_link) {
634 			if (ifa->ifa_addr->sa_family != AF_INET)
635 				continue;
636 			ia = ifatoia(ifa);
637 			goto match;
638 		}
639 	}
640 	/*
641 	 * If the packet was transiting through us, use the address of
642 	 * the interface that is the closest to the packet source.
643 	 * When we don't have a route back to the packet source, stop here
644 	 * and drop the packet.
645 	 */
646 	ia = ip_rtaddr(ip->ip_dst);
647 	if (ia == NULL) {
648 		m_freem(m);
649 		icmpstat.icps_noroute++;
650 		goto done;
651 	}
652 match:
653 #ifdef MAC
654 	mac_reflect_mbuf_icmp(m);
655 #endif
656 	t = IA_SIN(ia)->sin_addr;
657 	ip->ip_src = t;
658 	ip->ip_ttl = ip_defttl;
659 
660 	if (optlen > 0) {
661 		register u_char *cp;
662 		int opt, cnt;
663 		u_int len;
664 
665 		/*
666 		 * Retrieve any source routing from the incoming packet;
667 		 * add on any record-route or timestamp options.
668 		 */
669 		cp = (u_char *) (ip + 1);
670 		if ((opts = ip_srcroute(m)) == 0 &&
671 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
672 			opts->m_len = sizeof(struct in_addr);
673 			mtod(opts, struct in_addr *)->s_addr = 0;
674 		}
675 		if (opts) {
676 #ifdef ICMPPRINTFS
677 		    if (icmpprintfs)
678 			    printf("icmp_reflect optlen %d rt %d => ",
679 				optlen, opts->m_len);
680 #endif
681 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
682 			    opt = cp[IPOPT_OPTVAL];
683 			    if (opt == IPOPT_EOL)
684 				    break;
685 			    if (opt == IPOPT_NOP)
686 				    len = 1;
687 			    else {
688 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
689 					    break;
690 				    len = cp[IPOPT_OLEN];
691 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
692 				        len > cnt)
693 					    break;
694 			    }
695 			    /*
696 			     * Should check for overflow, but it "can't happen"
697 			     */
698 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
699 				opt == IPOPT_SECURITY) {
700 				    bcopy((caddr_t)cp,
701 					mtod(opts, caddr_t) + opts->m_len, len);
702 				    opts->m_len += len;
703 			    }
704 		    }
705 		    /* Terminate & pad, if necessary */
706 		    cnt = opts->m_len % 4;
707 		    if (cnt) {
708 			    for (; cnt < 4; cnt++) {
709 				    *(mtod(opts, caddr_t) + opts->m_len) =
710 					IPOPT_EOL;
711 				    opts->m_len++;
712 			    }
713 		    }
714 #ifdef ICMPPRINTFS
715 		    if (icmpprintfs)
716 			    printf("%d\n", opts->m_len);
717 #endif
718 		}
719 		/*
720 		 * Now strip out original options by copying rest of first
721 		 * mbuf's data back, and adjust the IP length.
722 		 */
723 		ip->ip_len -= optlen;
724 		ip->ip_v = IPVERSION;
725 		ip->ip_hl = 5;
726 		m->m_len -= optlen;
727 		if (m->m_flags & M_PKTHDR)
728 			m->m_pkthdr.len -= optlen;
729 		optlen += sizeof(struct ip);
730 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
731 			 (unsigned)(m->m_len - sizeof(struct ip)));
732 	}
733 	m_tag_delete_nonpersistent(m);
734 	m->m_flags &= ~(M_BCAST|M_MCAST);
735 	icmp_send(m, opts);
736 done:
737 	if (opts)
738 		(void)m_free(opts);
739 }
740 
741 /*
742  * Send an icmp packet back to the ip level,
743  * after supplying a checksum.
744  */
745 static void
746 icmp_send(m, opts)
747 	register struct mbuf *m;
748 	struct mbuf *opts;
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, NULL, 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 /*
786  * Return the next larger or smaller MTU plateau (table from RFC 1191)
787  * given current value MTU.  If DIR is less than zero, a larger plateau
788  * is returned; otherwise, a smaller value is returned.
789  */
790 int
791 ip_next_mtu(mtu, dir)
792 	int mtu;
793 	int dir;
794 {
795 	static int mtutab[] = {
796 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
797 		296, 68, 0
798 	};
799 	int i;
800 
801 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
802 		if (mtu >= mtutab[i])
803 			break;
804 	}
805 
806 	if (dir < 0) {
807 		if (i == 0) {
808 			return 0;
809 		} else {
810 			return mtutab[i - 1];
811 		}
812 	} else {
813 		if (mtutab[i] == 0) {
814 			return 0;
815 		} else if(mtu > mtutab[i]) {
816 			return mtutab[i];
817 		} else {
818 			return mtutab[i + 1];
819 		}
820 	}
821 }
822 
823 
824 /*
825  * badport_bandlim() - check for ICMP bandwidth limit
826  *
827  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
828  *	hit our bandwidth limit and it is not ok.
829  *
830  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
831  *
832  *	For now we separate the TCP and UDP subsystems w/ different 'which'
833  *	values.  We may eventually remove this separation (and simplify the
834  *	code further).
835  *
836  *	Note that the printing of the error message is delayed so we can
837  *	properly print the icmp error rate that the system was trying to do
838  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
839  *	the 'final' error, but it doesn't make sense to solve the printing
840  *	delay with more complex code.
841  */
842 
843 int
844 badport_bandlim(int which)
845 {
846 #define	N(a)	(sizeof (a) / sizeof (a[0]))
847 	static struct rate {
848 		const char	*type;
849 		struct timeval	lasttime;
850 		int		curpps;
851 	} rates[BANDLIM_MAX+1] = {
852 		{ "icmp unreach response" },
853 		{ "icmp ping response" },
854 		{ "icmp tstamp response" },
855 		{ "closed port RST response" },
856 		{ "open port RST response" }
857 	};
858 
859 	/*
860 	 * Return ok status if feature disabled or argument out of range.
861 	 */
862 	if (icmplim > 0 && (u_int) which < N(rates)) {
863 		struct rate *r = &rates[which];
864 		int opps = r->curpps;
865 
866 		if (!ppsratecheck(&r->lasttime, &r->curpps, icmplim))
867 			return -1;	/* discard packet */
868 		/*
869 		 * If we've dropped below the threshold after having
870 		 * rate-limited traffic print the message.  This preserves
871 		 * the previous behaviour at the expense of added complexity.
872 		 */
873 		if (icmplim_output && opps > icmplim)
874 			printf("Limiting %s from %d to %d packets/sec\n",
875 				r->type, opps, icmplim);
876 	}
877 	return 0;			/* okay to send packet */
878 #undef N
879 }
880