xref: /freebsd/sys/netinet/ip_icmp.c (revision 94942af266ac119ede0ca836f9aa5a5ac0582938)
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/mbuf.h>
39 #include <sys/protosw.h>
40 #include <sys/socket.h>
41 #include <sys/time.h>
42 #include <sys/kernel.h>
43 #include <sys/sysctl.h>
44 
45 #include <net/if.h>
46 #include <net/if_types.h>
47 #include <net/route.h>
48 
49 #include <netinet/in.h>
50 #include <netinet/in_pcb.h>
51 #include <netinet/in_systm.h>
52 #include <netinet/in_var.h>
53 #include <netinet/ip.h>
54 #include <netinet/ip_icmp.h>
55 #include <netinet/ip_var.h>
56 #include <netinet/ip_options.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 #include <security/mac/mac_framework.h>
76 
77 /*
78  * ICMP routines: error generation, receive packet processing, and
79  * routines to turnaround packets back to the originator, and
80  * host table maintenance routines.
81  */
82 
83 struct	icmpstat icmpstat;
84 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
85 	&icmpstat, icmpstat, "");
86 
87 static int	icmpmaskrepl = 0;
88 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
89 	&icmpmaskrepl, 0, "Reply to ICMP Address Mask Request packets.");
90 
91 static u_int	icmpmaskfake = 0;
92 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
93 	&icmpmaskfake, 0, "Fake reply to ICMP Address Mask Request packets.");
94 
95 static int	drop_redirect = 0;
96 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
97 	&drop_redirect, 0, "Ignore ICMP redirects");
98 
99 static int	log_redirect = 0;
100 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
101 	&log_redirect, 0, "Log ICMP redirects to the console");
102 
103 static int      icmplim = 200;
104 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
105 	&icmplim, 0, "Maximum number of ICMP responses per second");
106 
107 static int	icmplim_output = 1;
108 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
109 	&icmplim_output, 0, "Enable rate limiting of ICMP responses");
110 
111 static char	reply_src[IFNAMSIZ];
112 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
113 	&reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
114 
115 static int	icmp_rfi = 0;
116 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
117 	&icmp_rfi, 0, "ICMP reply from incoming interface for "
118 	"non-local packets");
119 
120 static int	icmp_quotelen = 8;
121 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
122 	&icmp_quotelen, 0, "Number of bytes from original packet to "
123 	"quote in ICMP reply");
124 
125 /*
126  * ICMP broadcast echo sysctl
127  */
128 
129 static int	icmpbmcastecho = 0;
130 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
131 	&icmpbmcastecho, 0, "");
132 
133 
134 #ifdef ICMPPRINTFS
135 int	icmpprintfs = 0;
136 #endif
137 
138 static void	icmp_reflect(struct mbuf *);
139 static void	icmp_send(struct mbuf *, struct mbuf *);
140 
141 extern	struct protosw inetsw[];
142 
143 /*
144  * Generate an error packet of type error
145  * in response to bad packet ip.
146  */
147 void
148 icmp_error(struct mbuf *n, int type, int code, n_long dest, int mtu)
149 {
150 	register struct ip *oip = mtod(n, struct ip *), *nip;
151 	register unsigned oiphlen = oip->ip_hl << 2;
152 	register struct icmp *icp;
153 	register struct mbuf *m;
154 	unsigned icmplen, icmpelen, nlen;
155 
156 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
157 #ifdef ICMPPRINTFS
158 	if (icmpprintfs)
159 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
160 #endif
161 	if (type != ICMP_REDIRECT)
162 		icmpstat.icps_error++;
163 	/*
164 	 * Don't send error:
165 	 *  if the original packet was encrypted.
166 	 *  if not the first fragment of message.
167 	 *  in response to a multicast or broadcast packet.
168 	 *  if the old packet protocol was an ICMP error message.
169 	 */
170 	if (n->m_flags & M_DECRYPTED)
171 		goto freeit;
172 	if (oip->ip_off & ~(IP_MF|IP_DF))
173 		goto freeit;
174 	if (n->m_flags & (M_BCAST|M_MCAST))
175 		goto freeit;
176 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
177 	  n->m_len >= oiphlen + ICMP_MINLEN &&
178 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
179 		icmpstat.icps_oldicmp++;
180 		goto freeit;
181 	}
182 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
183 	if (oiphlen + 8 > n->m_len)
184 		goto freeit;
185 	/*
186 	 * Calculate length to quote from original packet and
187 	 * prevent the ICMP mbuf from overflowing.
188 	 * Unfortunatly this is non-trivial since ip_forward()
189 	 * sends us truncated packets.
190 	 */
191 	nlen = m_length(n, NULL);
192 	if (oip->ip_p == IPPROTO_TCP) {
193 		struct tcphdr *th;
194 		int tcphlen;
195 
196 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
197 		    n->m_next == NULL)
198 			goto stdreply;
199 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
200 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
201 			goto freeit;
202 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
203 		tcphlen = th->th_off << 2;
204 		if (tcphlen < sizeof(struct tcphdr))
205 			goto freeit;
206 		if (oip->ip_len < oiphlen + tcphlen)
207 			goto freeit;
208 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
209 			goto stdreply;
210 		if (n->m_len < oiphlen + tcphlen &&
211 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
212 			goto freeit;
213 		icmpelen = max(tcphlen, min(icmp_quotelen, oip->ip_len - oiphlen));
214 	} else
215 stdreply:	icmpelen = max(8, min(icmp_quotelen, oip->ip_len - oiphlen));
216 
217 	icmplen = min(oiphlen + icmpelen, nlen);
218 	if (icmplen < sizeof(struct ip))
219 		goto freeit;
220 
221 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
222 		m = m_gethdr(M_DONTWAIT, MT_DATA);
223 	else
224 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
225 	if (m == NULL)
226 		goto freeit;
227 #ifdef MAC
228 	mac_create_mbuf_netlayer(n, m);
229 #endif
230 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
231 	m_align(m, ICMP_MINLEN + icmplen);
232 	m->m_len = ICMP_MINLEN + icmplen;
233 
234 	icp = mtod(m, struct icmp *);
235 	icmpstat.icps_outhist[type]++;
236 	icp->icmp_type = type;
237 	if (type == ICMP_REDIRECT)
238 		icp->icmp_gwaddr.s_addr = dest;
239 	else {
240 		icp->icmp_void = 0;
241 		/*
242 		 * The following assignments assume an overlay with the
243 		 * just zeroed icmp_void field.
244 		 */
245 		if (type == ICMP_PARAMPROB) {
246 			icp->icmp_pptr = code;
247 			code = 0;
248 		} else if (type == ICMP_UNREACH &&
249 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
250 			icp->icmp_nextmtu = htons(mtu);
251 		}
252 	}
253 	icp->icmp_code = code;
254 
255 	/*
256 	 * Copy the quotation into ICMP message and
257 	 * convert quoted IP header back to network representation.
258 	 */
259 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
260 	nip = &icp->icmp_ip;
261 	nip->ip_len = htons(nip->ip_len);
262 	nip->ip_off = htons(nip->ip_off);
263 
264 	/*
265 	 * Set up ICMP message mbuf and copy old IP header (without options
266 	 * in front of ICMP message.
267 	 * If the original mbuf was meant to bypass the firewall, the error
268 	 * reply should bypass as well.
269 	 */
270 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
271 	m->m_data -= sizeof(struct ip);
272 	m->m_len += sizeof(struct ip);
273 	m->m_pkthdr.len = m->m_len;
274 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
275 	nip = mtod(m, struct ip *);
276 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
277 	nip->ip_len = m->m_len;
278 	nip->ip_v = IPVERSION;
279 	nip->ip_hl = 5;
280 	nip->ip_p = IPPROTO_ICMP;
281 	nip->ip_tos = 0;
282 	icmp_reflect(m);
283 
284 freeit:
285 	m_freem(n);
286 }
287 
288 /*
289  * Process a received ICMP message.
290  */
291 void
292 icmp_input(struct mbuf *m, int off)
293 {
294 	struct icmp *icp;
295 	struct in_ifaddr *ia;
296 	struct ip *ip = mtod(m, struct ip *);
297 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
298 	int hlen = off;
299 	int icmplen = ip->ip_len;
300 	int i, code;
301 	void (*ctlfunc)(int, struct sockaddr *, void *);
302 
303 	/*
304 	 * Locate icmp structure in mbuf, and check
305 	 * that not corrupted and of at least minimum length.
306 	 */
307 #ifdef ICMPPRINTFS
308 	if (icmpprintfs) {
309 		char buf[4 * sizeof "123"];
310 		strcpy(buf, inet_ntoa(ip->ip_src));
311 		printf("icmp_input from %s to %s, len %d\n",
312 		       buf, inet_ntoa(ip->ip_dst), icmplen);
313 	}
314 #endif
315 	if (icmplen < ICMP_MINLEN) {
316 		icmpstat.icps_tooshort++;
317 		goto freeit;
318 	}
319 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
320 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
321 		icmpstat.icps_tooshort++;
322 		return;
323 	}
324 	ip = mtod(m, struct ip *);
325 	m->m_len -= hlen;
326 	m->m_data += hlen;
327 	icp = mtod(m, struct icmp *);
328 	if (in_cksum(m, icmplen)) {
329 		icmpstat.icps_checksum++;
330 		goto freeit;
331 	}
332 	m->m_len += hlen;
333 	m->m_data -= hlen;
334 
335 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
336 		/*
337 		 * Deliver very specific ICMP type only.
338 		 */
339 		switch (icp->icmp_type) {
340 		case ICMP_UNREACH:
341 		case ICMP_TIMXCEED:
342 			break;
343 		default:
344 			goto freeit;
345 		}
346 	}
347 
348 #ifdef ICMPPRINTFS
349 	if (icmpprintfs)
350 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
351 		    icp->icmp_code);
352 #endif
353 
354 	/*
355 	 * Message type specific processing.
356 	 */
357 	if (icp->icmp_type > ICMP_MAXTYPE)
358 		goto raw;
359 
360 	/* Initialize */
361 	bzero(&icmpsrc, sizeof(icmpsrc));
362 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
363 	icmpsrc.sin_family = AF_INET;
364 	bzero(&icmpdst, sizeof(icmpdst));
365 	icmpdst.sin_len = sizeof(struct sockaddr_in);
366 	icmpdst.sin_family = AF_INET;
367 	bzero(&icmpgw, sizeof(icmpgw));
368 	icmpgw.sin_len = sizeof(struct sockaddr_in);
369 	icmpgw.sin_family = AF_INET;
370 
371 	icmpstat.icps_inhist[icp->icmp_type]++;
372 	code = icp->icmp_code;
373 	switch (icp->icmp_type) {
374 
375 	case ICMP_UNREACH:
376 		switch (code) {
377 			case ICMP_UNREACH_NET:
378 			case ICMP_UNREACH_HOST:
379 			case ICMP_UNREACH_SRCFAIL:
380 			case ICMP_UNREACH_NET_UNKNOWN:
381 			case ICMP_UNREACH_HOST_UNKNOWN:
382 			case ICMP_UNREACH_ISOLATED:
383 			case ICMP_UNREACH_TOSNET:
384 			case ICMP_UNREACH_TOSHOST:
385 			case ICMP_UNREACH_HOST_PRECEDENCE:
386 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
387 				code = PRC_UNREACH_NET;
388 				break;
389 
390 			case ICMP_UNREACH_NEEDFRAG:
391 				code = PRC_MSGSIZE;
392 				break;
393 
394 			/*
395 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
396 			 * Treat subcodes 2,3 as immediate RST
397 			 */
398 			case ICMP_UNREACH_PROTOCOL:
399 			case ICMP_UNREACH_PORT:
400 				code = PRC_UNREACH_PORT;
401 				break;
402 
403 			case ICMP_UNREACH_NET_PROHIB:
404 			case ICMP_UNREACH_HOST_PROHIB:
405 			case ICMP_UNREACH_FILTER_PROHIB:
406 				code = PRC_UNREACH_ADMIN_PROHIB;
407 				break;
408 
409 			default:
410 				goto badcode;
411 		}
412 		goto deliver;
413 
414 	case ICMP_TIMXCEED:
415 		if (code > 1)
416 			goto badcode;
417 		code += PRC_TIMXCEED_INTRANS;
418 		goto deliver;
419 
420 	case ICMP_PARAMPROB:
421 		if (code > 1)
422 			goto badcode;
423 		code = PRC_PARAMPROB;
424 		goto deliver;
425 
426 	case ICMP_SOURCEQUENCH:
427 		if (code)
428 			goto badcode;
429 		code = PRC_QUENCH;
430 	deliver:
431 		/*
432 		 * Problem with datagram; advise higher level routines.
433 		 */
434 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
435 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
436 			icmpstat.icps_badlen++;
437 			goto freeit;
438 		}
439 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
440 		/* Discard ICMP's in response to multicast packets */
441 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
442 			goto badcode;
443 #ifdef ICMPPRINTFS
444 		if (icmpprintfs)
445 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
446 #endif
447 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
448 		/*
449 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
450 		 * notification to TCP layer.
451 		 */
452 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
453 		if (ctlfunc)
454 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
455 				   (void *)&icp->icmp_ip);
456 		break;
457 
458 	badcode:
459 		icmpstat.icps_badcode++;
460 		break;
461 
462 	case ICMP_ECHO:
463 		if (!icmpbmcastecho
464 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
465 			icmpstat.icps_bmcastecho++;
466 			break;
467 		}
468 		icp->icmp_type = ICMP_ECHOREPLY;
469 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
470 			goto freeit;
471 		else
472 			goto reflect;
473 
474 	case ICMP_TSTAMP:
475 		if (!icmpbmcastecho
476 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
477 			icmpstat.icps_bmcasttstamp++;
478 			break;
479 		}
480 		if (icmplen < ICMP_TSLEN) {
481 			icmpstat.icps_badlen++;
482 			break;
483 		}
484 		icp->icmp_type = ICMP_TSTAMPREPLY;
485 		icp->icmp_rtime = iptime();
486 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
487 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
488 			goto freeit;
489 		else
490 			goto reflect;
491 
492 	case ICMP_MASKREQ:
493 		if (icmpmaskrepl == 0)
494 			break;
495 		/*
496 		 * We are not able to respond with all ones broadcast
497 		 * unless we receive it over a point-to-point interface.
498 		 */
499 		if (icmplen < ICMP_MASKLEN)
500 			break;
501 		switch (ip->ip_dst.s_addr) {
502 
503 		case INADDR_BROADCAST:
504 		case INADDR_ANY:
505 			icmpdst.sin_addr = ip->ip_src;
506 			break;
507 
508 		default:
509 			icmpdst.sin_addr = ip->ip_dst;
510 		}
511 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
512 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
513 		if (ia == 0)
514 			break;
515 		if (ia->ia_ifp == 0)
516 			break;
517 		icp->icmp_type = ICMP_MASKREPLY;
518 		if (icmpmaskfake == 0)
519 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
520 		else
521 			icp->icmp_mask = icmpmaskfake;
522 		if (ip->ip_src.s_addr == 0) {
523 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
524 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
525 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
526 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
527 		}
528 reflect:
529 		ip->ip_len += hlen;	/* since ip_input deducts this */
530 		icmpstat.icps_reflect++;
531 		icmpstat.icps_outhist[icp->icmp_type]++;
532 		icmp_reflect(m);
533 		return;
534 
535 	case ICMP_REDIRECT:
536 		if (log_redirect) {
537 			u_long src, dst, gw;
538 
539 			src = ntohl(ip->ip_src.s_addr);
540 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
541 			gw = ntohl(icp->icmp_gwaddr.s_addr);
542 			printf("icmp redirect from %d.%d.%d.%d: "
543 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
544 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
545 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
546 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
547 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
548 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
549 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
550 		}
551 		/*
552 		 * RFC1812 says we must ignore ICMP redirects if we
553 		 * are acting as router.
554 		 */
555 		if (drop_redirect || ipforwarding)
556 			break;
557 		if (code > 3)
558 			goto badcode;
559 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
560 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
561 			icmpstat.icps_badlen++;
562 			break;
563 		}
564 		/*
565 		 * Short circuit routing redirects to force
566 		 * immediate change in the kernel's routing
567 		 * tables.  The message is also handed to anyone
568 		 * listening on a raw socket (e.g. the routing
569 		 * daemon for use in updating its tables).
570 		 */
571 		icmpgw.sin_addr = ip->ip_src;
572 		icmpdst.sin_addr = icp->icmp_gwaddr;
573 #ifdef	ICMPPRINTFS
574 		if (icmpprintfs) {
575 			char buf[4 * sizeof "123"];
576 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
577 
578 			printf("redirect dst %s to %s\n",
579 			       buf, inet_ntoa(icp->icmp_gwaddr));
580 		}
581 #endif
582 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
583 		rtredirect((struct sockaddr *)&icmpsrc,
584 		  (struct sockaddr *)&icmpdst,
585 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
586 		  (struct sockaddr *)&icmpgw);
587 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
588 #ifdef IPSEC
589 		key_sa_routechange((struct sockaddr *)&icmpsrc);
590 #endif
591 		break;
592 
593 	/*
594 	 * No kernel processing for the following;
595 	 * just fall through to send to raw listener.
596 	 */
597 	case ICMP_ECHOREPLY:
598 	case ICMP_ROUTERADVERT:
599 	case ICMP_ROUTERSOLICIT:
600 	case ICMP_TSTAMPREPLY:
601 	case ICMP_IREQREPLY:
602 	case ICMP_MASKREPLY:
603 	default:
604 		break;
605 	}
606 
607 raw:
608 	rip_input(m, off);
609 	return;
610 
611 freeit:
612 	m_freem(m);
613 }
614 
615 /*
616  * Reflect the ip packet back to the source
617  */
618 static void
619 icmp_reflect(struct mbuf *m)
620 {
621 	struct ip *ip = mtod(m, struct ip *);
622 	struct ifaddr *ifa;
623 	struct ifnet *ifn;
624 	struct in_ifaddr *ia;
625 	struct in_addr t;
626 	struct mbuf *opts = 0;
627 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
628 
629 	if (!in_canforward(ip->ip_src) &&
630 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
631 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
632 		m_freem(m);	/* Bad return address */
633 		icmpstat.icps_badaddr++;
634 		goto done;	/* Ip_output() will check for broadcast */
635 	}
636 	t = ip->ip_dst;
637 	ip->ip_dst = ip->ip_src;
638 
639 	/*
640 	 * Source selection for ICMP replies:
641 	 *
642 	 * If the incoming packet was addressed directly to one of our
643 	 * own addresses, use dst as the src for the reply.
644 	 */
645 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
646 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
647 			goto match;
648 	/*
649 	 * If the incoming packet was addressed to one of our broadcast
650 	 * addresses, use the first non-broadcast address which corresponds
651 	 * to the incoming interface.
652 	 */
653 	if (m->m_pkthdr.rcvif != NULL &&
654 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
655 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
656 			if (ifa->ifa_addr->sa_family != AF_INET)
657 				continue;
658 			ia = ifatoia(ifa);
659 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
660 			    t.s_addr)
661 				goto match;
662 		}
663 	}
664 	/*
665 	 * If the packet was transiting through us, use the address of
666 	 * the interface the packet came through in.  If that interface
667 	 * doesn't have a suitable IP address, the normal selection
668 	 * criteria apply.
669 	 */
670 	if (icmp_rfi && m->m_pkthdr.rcvif != NULL) {
671 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
672 			if (ifa->ifa_addr->sa_family != AF_INET)
673 				continue;
674 			ia = ifatoia(ifa);
675 			goto match;
676 		}
677 	}
678 	/*
679 	 * If the incoming packet was not addressed directly to us, use
680 	 * designated interface for icmp replies specified by sysctl
681 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
682 	 * with normal source selection.
683 	 */
684 	if (reply_src[0] != '\0' && (ifn = ifunit(reply_src))) {
685 		TAILQ_FOREACH(ifa, &ifn->if_addrhead, ifa_link) {
686 			if (ifa->ifa_addr->sa_family != AF_INET)
687 				continue;
688 			ia = ifatoia(ifa);
689 			goto match;
690 		}
691 	}
692 	/*
693 	 * If the packet was transiting through us, use the address of
694 	 * the interface that is the closest to the packet source.
695 	 * When we don't have a route back to the packet source, stop here
696 	 * and drop the packet.
697 	 */
698 	ia = ip_rtaddr(ip->ip_dst);
699 	if (ia == NULL) {
700 		m_freem(m);
701 		icmpstat.icps_noroute++;
702 		goto done;
703 	}
704 match:
705 #ifdef MAC
706 	mac_reflect_mbuf_icmp(m);
707 #endif
708 	t = IA_SIN(ia)->sin_addr;
709 	ip->ip_src = t;
710 	ip->ip_ttl = ip_defttl;
711 
712 	if (optlen > 0) {
713 		register u_char *cp;
714 		int opt, cnt;
715 		u_int len;
716 
717 		/*
718 		 * Retrieve any source routing from the incoming packet;
719 		 * add on any record-route or timestamp options.
720 		 */
721 		cp = (u_char *) (ip + 1);
722 		if ((opts = ip_srcroute(m)) == 0 &&
723 		    (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
724 			opts->m_len = sizeof(struct in_addr);
725 			mtod(opts, struct in_addr *)->s_addr = 0;
726 		}
727 		if (opts) {
728 #ifdef ICMPPRINTFS
729 		    if (icmpprintfs)
730 			    printf("icmp_reflect optlen %d rt %d => ",
731 				optlen, opts->m_len);
732 #endif
733 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
734 			    opt = cp[IPOPT_OPTVAL];
735 			    if (opt == IPOPT_EOL)
736 				    break;
737 			    if (opt == IPOPT_NOP)
738 				    len = 1;
739 			    else {
740 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
741 					    break;
742 				    len = cp[IPOPT_OLEN];
743 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
744 				        len > cnt)
745 					    break;
746 			    }
747 			    /*
748 			     * Should check for overflow, but it "can't happen"
749 			     */
750 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
751 				opt == IPOPT_SECURITY) {
752 				    bcopy((caddr_t)cp,
753 					mtod(opts, caddr_t) + opts->m_len, len);
754 				    opts->m_len += len;
755 			    }
756 		    }
757 		    /* Terminate & pad, if necessary */
758 		    cnt = opts->m_len % 4;
759 		    if (cnt) {
760 			    for (; cnt < 4; cnt++) {
761 				    *(mtod(opts, caddr_t) + opts->m_len) =
762 					IPOPT_EOL;
763 				    opts->m_len++;
764 			    }
765 		    }
766 #ifdef ICMPPRINTFS
767 		    if (icmpprintfs)
768 			    printf("%d\n", opts->m_len);
769 #endif
770 		}
771 		/*
772 		 * Now strip out original options by copying rest of first
773 		 * mbuf's data back, and adjust the IP length.
774 		 */
775 		ip->ip_len -= optlen;
776 		ip->ip_v = IPVERSION;
777 		ip->ip_hl = 5;
778 		m->m_len -= optlen;
779 		if (m->m_flags & M_PKTHDR)
780 			m->m_pkthdr.len -= optlen;
781 		optlen += sizeof(struct ip);
782 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
783 			 (unsigned)(m->m_len - sizeof(struct ip)));
784 	}
785 	m_tag_delete_nonpersistent(m);
786 	m->m_flags &= ~(M_BCAST|M_MCAST);
787 	icmp_send(m, opts);
788 done:
789 	if (opts)
790 		(void)m_free(opts);
791 }
792 
793 /*
794  * Send an icmp packet back to the ip level,
795  * after supplying a checksum.
796  */
797 static void
798 icmp_send(struct mbuf *m, struct mbuf *opts)
799 {
800 	register struct ip *ip = mtod(m, struct ip *);
801 	register int hlen;
802 	register struct icmp *icp;
803 
804 	hlen = ip->ip_hl << 2;
805 	m->m_data += hlen;
806 	m->m_len -= hlen;
807 	icp = mtod(m, struct icmp *);
808 	icp->icmp_cksum = 0;
809 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
810 	m->m_data -= hlen;
811 	m->m_len += hlen;
812 	m->m_pkthdr.rcvif = (struct ifnet *)0;
813 #ifdef ICMPPRINTFS
814 	if (icmpprintfs) {
815 		char buf[4 * sizeof "123"];
816 		strcpy(buf, inet_ntoa(ip->ip_dst));
817 		printf("icmp_send dst %s src %s\n",
818 		       buf, inet_ntoa(ip->ip_src));
819 	}
820 #endif
821 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
822 }
823 
824 n_time
825 iptime(void)
826 {
827 	struct timeval atv;
828 	u_long t;
829 
830 	getmicrotime(&atv);
831 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
832 	return (htonl(t));
833 }
834 
835 /*
836  * Return the next larger or smaller MTU plateau (table from RFC 1191)
837  * given current value MTU.  If DIR is less than zero, a larger plateau
838  * is returned; otherwise, a smaller value is returned.
839  */
840 int
841 ip_next_mtu(int mtu, int dir)
842 {
843 	static int mtutab[] = {
844 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
845 		296, 68, 0
846 	};
847 	int i, size;
848 
849 	size = (sizeof mtutab) / (sizeof mtutab[0]);
850 	if (dir >= 0) {
851 		for (i = 0; i < size; i++)
852 			if (mtu > mtutab[i])
853 				return mtutab[i];
854 	} else {
855 		for (i = size - 1; i >= 0; i--)
856 			if (mtu < mtutab[i])
857 				return mtutab[i];
858 		if (mtu == mtutab[0])
859 			return mtutab[0];
860 	}
861 	return 0;
862 }
863 
864 
865 /*
866  * badport_bandlim() - check for ICMP bandwidth limit
867  *
868  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
869  *	hit our bandwidth limit and it is not ok.
870  *
871  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
872  *
873  *	For now we separate the TCP and UDP subsystems w/ different 'which'
874  *	values.  We may eventually remove this separation (and simplify the
875  *	code further).
876  *
877  *	Note that the printing of the error message is delayed so we can
878  *	properly print the icmp error rate that the system was trying to do
879  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
880  *	the 'final' error, but it doesn't make sense to solve the printing
881  *	delay with more complex code.
882  */
883 
884 int
885 badport_bandlim(int which)
886 {
887 #define	N(a)	(sizeof (a) / sizeof (a[0]))
888 	static struct rate {
889 		const char	*type;
890 		struct timeval	lasttime;
891 		int		curpps;
892 	} rates[BANDLIM_MAX+1] = {
893 		{ "icmp unreach response" },
894 		{ "icmp ping response" },
895 		{ "icmp tstamp response" },
896 		{ "closed port RST response" },
897 		{ "open port RST response" }
898 	};
899 
900 	/*
901 	 * Return ok status if feature disabled or argument out of range.
902 	 */
903 	if (icmplim > 0 && (u_int) which < N(rates)) {
904 		struct rate *r = &rates[which];
905 		int opps = r->curpps;
906 
907 		if (!ppsratecheck(&r->lasttime, &r->curpps, icmplim))
908 			return -1;	/* discard packet */
909 		/*
910 		 * If we've dropped below the threshold after having
911 		 * rate-limited traffic print the message.  This preserves
912 		 * the previous behaviour at the expense of added complexity.
913 		 */
914 		if (icmplim_output && opps > icmplim)
915 			printf("Limiting %s from %d to %d packets/sec\n",
916 				r->type, opps, icmplim);
917 	}
918 	return 0;			/* okay to send packet */
919 #undef N
920 }
921