xref: /freebsd/sys/netinet/ip_icmp.c (revision f0a75d274af375d15b97b830966b99a02b7db911)
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(n, type, code, dest, mtu)
149 	struct mbuf *n;
150 	int type, code;
151 	n_long dest;
152 	int mtu;
153 {
154 	register struct ip *oip = mtod(n, struct ip *), *nip;
155 	register unsigned oiphlen = oip->ip_hl << 2;
156 	register struct icmp *icp;
157 	register struct mbuf *m;
158 	unsigned icmplen, icmpelen, nlen;
159 
160 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
161 #ifdef ICMPPRINTFS
162 	if (icmpprintfs)
163 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
164 #endif
165 	if (type != ICMP_REDIRECT)
166 		icmpstat.icps_error++;
167 	/*
168 	 * Don't send error:
169 	 *  if the original packet was encrypted.
170 	 *  if not the first fragment of message.
171 	 *  in response to a multicast or broadcast packet.
172 	 *  if the old packet protocol was an ICMP error message.
173 	 */
174 	if (n->m_flags & M_DECRYPTED)
175 		goto freeit;
176 	if (oip->ip_off & ~(IP_MF|IP_DF))
177 		goto freeit;
178 	if (n->m_flags & (M_BCAST|M_MCAST))
179 		goto freeit;
180 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
181 	  n->m_len >= oiphlen + ICMP_MINLEN &&
182 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
183 		icmpstat.icps_oldicmp++;
184 		goto freeit;
185 	}
186 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
187 	if (oiphlen + 8 > n->m_len)
188 		goto freeit;
189 	/*
190 	 * Calculate length to quote from original packet and
191 	 * prevent the ICMP mbuf from overflowing.
192 	 * Unfortunatly this is non-trivial since ip_forward()
193 	 * sends us truncated packets.
194 	 */
195 	nlen = m_length(n, NULL);
196 	if (oip->ip_p == IPPROTO_TCP) {
197 		struct tcphdr *th;
198 		int tcphlen;
199 
200 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
201 		    n->m_next == NULL)
202 			goto stdreply;
203 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
204 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
205 			goto freeit;
206 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
207 		tcphlen = th->th_off << 2;
208 		if (tcphlen < sizeof(struct tcphdr))
209 			goto freeit;
210 		if (oip->ip_len < oiphlen + tcphlen)
211 			goto freeit;
212 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
213 			goto stdreply;
214 		if (n->m_len < oiphlen + tcphlen &&
215 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
216 			goto freeit;
217 		icmpelen = max(tcphlen, min(icmp_quotelen, oip->ip_len - oiphlen));
218 	} else
219 stdreply:	icmpelen = max(8, min(icmp_quotelen, oip->ip_len - oiphlen));
220 
221 	icmplen = min(oiphlen + icmpelen, nlen);
222 	if (icmplen < sizeof(struct ip))
223 		goto freeit;
224 
225 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
226 		m = m_gethdr(M_DONTWAIT, MT_DATA);
227 	else
228 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
229 	if (m == NULL)
230 		goto freeit;
231 #ifdef MAC
232 	mac_create_mbuf_netlayer(n, m);
233 #endif
234 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
235 	m_align(m, ICMP_MINLEN + icmplen);
236 	m->m_len = ICMP_MINLEN + icmplen;
237 
238 	icp = mtod(m, struct icmp *);
239 	icmpstat.icps_outhist[type]++;
240 	icp->icmp_type = type;
241 	if (type == ICMP_REDIRECT)
242 		icp->icmp_gwaddr.s_addr = dest;
243 	else {
244 		icp->icmp_void = 0;
245 		/*
246 		 * The following assignments assume an overlay with the
247 		 * just zeroed icmp_void field.
248 		 */
249 		if (type == ICMP_PARAMPROB) {
250 			icp->icmp_pptr = code;
251 			code = 0;
252 		} else if (type == ICMP_UNREACH &&
253 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
254 			icp->icmp_nextmtu = htons(mtu);
255 		}
256 	}
257 	icp->icmp_code = code;
258 
259 	/*
260 	 * Copy the quotation into ICMP message and
261 	 * convert quoted IP header back to network representation.
262 	 */
263 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
264 	nip = &icp->icmp_ip;
265 	nip->ip_len = htons(nip->ip_len);
266 	nip->ip_off = htons(nip->ip_off);
267 
268 	/*
269 	 * Set up ICMP message mbuf and copy old IP header (without options
270 	 * in front of ICMP message.
271 	 * If the original mbuf was meant to bypass the firewall, the error
272 	 * reply should bypass as well.
273 	 */
274 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
275 	m->m_data -= sizeof(struct ip);
276 	m->m_len += sizeof(struct ip);
277 	m->m_pkthdr.len = m->m_len;
278 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
279 	nip = mtod(m, struct ip *);
280 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
281 	nip->ip_len = m->m_len;
282 	nip->ip_v = IPVERSION;
283 	nip->ip_hl = 5;
284 	nip->ip_p = IPPROTO_ICMP;
285 	nip->ip_tos = 0;
286 	icmp_reflect(m);
287 
288 freeit:
289 	m_freem(n);
290 }
291 
292 /*
293  * Process a received ICMP message.
294  */
295 void
296 icmp_input(m, off)
297 	struct mbuf *m;
298 	int off;
299 {
300 	struct icmp *icp;
301 	struct in_ifaddr *ia;
302 	struct ip *ip = mtod(m, struct ip *);
303 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
304 	int hlen = off;
305 	int icmplen = ip->ip_len;
306 	int i, code;
307 	void (*ctlfunc)(int, struct sockaddr *, void *);
308 
309 	/*
310 	 * Locate icmp structure in mbuf, and check
311 	 * that not corrupted and of at least minimum length.
312 	 */
313 #ifdef ICMPPRINTFS
314 	if (icmpprintfs) {
315 		char buf[4 * sizeof "123"];
316 		strcpy(buf, inet_ntoa(ip->ip_src));
317 		printf("icmp_input from %s to %s, len %d\n",
318 		       buf, inet_ntoa(ip->ip_dst), icmplen);
319 	}
320 #endif
321 	if (icmplen < ICMP_MINLEN) {
322 		icmpstat.icps_tooshort++;
323 		goto freeit;
324 	}
325 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
326 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
327 		icmpstat.icps_tooshort++;
328 		return;
329 	}
330 	ip = mtod(m, struct ip *);
331 	m->m_len -= hlen;
332 	m->m_data += hlen;
333 	icp = mtod(m, struct icmp *);
334 	if (in_cksum(m, icmplen)) {
335 		icmpstat.icps_checksum++;
336 		goto freeit;
337 	}
338 	m->m_len += hlen;
339 	m->m_data -= hlen;
340 
341 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
342 		/*
343 		 * Deliver very specific ICMP type only.
344 		 */
345 		switch (icp->icmp_type) {
346 		case ICMP_UNREACH:
347 		case ICMP_TIMXCEED:
348 			break;
349 		default:
350 			goto freeit;
351 		}
352 	}
353 
354 #ifdef ICMPPRINTFS
355 	if (icmpprintfs)
356 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
357 		    icp->icmp_code);
358 #endif
359 
360 	/*
361 	 * Message type specific processing.
362 	 */
363 	if (icp->icmp_type > ICMP_MAXTYPE)
364 		goto raw;
365 
366 	/* Initialize */
367 	bzero(&icmpsrc, sizeof(icmpsrc));
368 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
369 	icmpsrc.sin_family = AF_INET;
370 	bzero(&icmpdst, sizeof(icmpdst));
371 	icmpdst.sin_len = sizeof(struct sockaddr_in);
372 	icmpdst.sin_family = AF_INET;
373 	bzero(&icmpgw, sizeof(icmpgw));
374 	icmpgw.sin_len = sizeof(struct sockaddr_in);
375 	icmpgw.sin_family = AF_INET;
376 
377 	icmpstat.icps_inhist[icp->icmp_type]++;
378 	code = icp->icmp_code;
379 	switch (icp->icmp_type) {
380 
381 	case ICMP_UNREACH:
382 		switch (code) {
383 			case ICMP_UNREACH_NET:
384 			case ICMP_UNREACH_HOST:
385 			case ICMP_UNREACH_SRCFAIL:
386 			case ICMP_UNREACH_NET_UNKNOWN:
387 			case ICMP_UNREACH_HOST_UNKNOWN:
388 			case ICMP_UNREACH_ISOLATED:
389 			case ICMP_UNREACH_TOSNET:
390 			case ICMP_UNREACH_TOSHOST:
391 			case ICMP_UNREACH_HOST_PRECEDENCE:
392 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
393 				code = PRC_UNREACH_NET;
394 				break;
395 
396 			case ICMP_UNREACH_NEEDFRAG:
397 				code = PRC_MSGSIZE;
398 				break;
399 
400 			/*
401 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
402 			 * Treat subcodes 2,3 as immediate RST
403 			 */
404 			case ICMP_UNREACH_PROTOCOL:
405 			case ICMP_UNREACH_PORT:
406 				code = PRC_UNREACH_PORT;
407 				break;
408 
409 			case ICMP_UNREACH_NET_PROHIB:
410 			case ICMP_UNREACH_HOST_PROHIB:
411 			case ICMP_UNREACH_FILTER_PROHIB:
412 				code = PRC_UNREACH_ADMIN_PROHIB;
413 				break;
414 
415 			default:
416 				goto badcode;
417 		}
418 		goto deliver;
419 
420 	case ICMP_TIMXCEED:
421 		if (code > 1)
422 			goto badcode;
423 		code += PRC_TIMXCEED_INTRANS;
424 		goto deliver;
425 
426 	case ICMP_PARAMPROB:
427 		if (code > 1)
428 			goto badcode;
429 		code = PRC_PARAMPROB;
430 		goto deliver;
431 
432 	case ICMP_SOURCEQUENCH:
433 		if (code)
434 			goto badcode;
435 		code = PRC_QUENCH;
436 	deliver:
437 		/*
438 		 * Problem with datagram; advise higher level routines.
439 		 */
440 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
441 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
442 			icmpstat.icps_badlen++;
443 			goto freeit;
444 		}
445 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
446 		/* Discard ICMP's in response to multicast packets */
447 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
448 			goto badcode;
449 #ifdef ICMPPRINTFS
450 		if (icmpprintfs)
451 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
452 #endif
453 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
454 		/*
455 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
456 		 * notification to TCP layer.
457 		 */
458 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
459 		if (ctlfunc)
460 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
461 				   (void *)&icp->icmp_ip);
462 		break;
463 
464 	badcode:
465 		icmpstat.icps_badcode++;
466 		break;
467 
468 	case ICMP_ECHO:
469 		if (!icmpbmcastecho
470 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
471 			icmpstat.icps_bmcastecho++;
472 			break;
473 		}
474 		icp->icmp_type = ICMP_ECHOREPLY;
475 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
476 			goto freeit;
477 		else
478 			goto reflect;
479 
480 	case ICMP_TSTAMP:
481 		if (!icmpbmcastecho
482 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
483 			icmpstat.icps_bmcasttstamp++;
484 			break;
485 		}
486 		if (icmplen < ICMP_TSLEN) {
487 			icmpstat.icps_badlen++;
488 			break;
489 		}
490 		icp->icmp_type = ICMP_TSTAMPREPLY;
491 		icp->icmp_rtime = iptime();
492 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
493 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
494 			goto freeit;
495 		else
496 			goto reflect;
497 
498 	case ICMP_MASKREQ:
499 		if (icmpmaskrepl == 0)
500 			break;
501 		/*
502 		 * We are not able to respond with all ones broadcast
503 		 * unless we receive it over a point-to-point interface.
504 		 */
505 		if (icmplen < ICMP_MASKLEN)
506 			break;
507 		switch (ip->ip_dst.s_addr) {
508 
509 		case INADDR_BROADCAST:
510 		case INADDR_ANY:
511 			icmpdst.sin_addr = ip->ip_src;
512 			break;
513 
514 		default:
515 			icmpdst.sin_addr = ip->ip_dst;
516 		}
517 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
518 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
519 		if (ia == 0)
520 			break;
521 		if (ia->ia_ifp == 0)
522 			break;
523 		icp->icmp_type = ICMP_MASKREPLY;
524 		if (icmpmaskfake == 0)
525 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
526 		else
527 			icp->icmp_mask = icmpmaskfake;
528 		if (ip->ip_src.s_addr == 0) {
529 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
530 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
531 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
532 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
533 		}
534 reflect:
535 		ip->ip_len += hlen;	/* since ip_input deducts this */
536 		icmpstat.icps_reflect++;
537 		icmpstat.icps_outhist[icp->icmp_type]++;
538 		icmp_reflect(m);
539 		return;
540 
541 	case ICMP_REDIRECT:
542 		if (log_redirect) {
543 			u_long src, dst, gw;
544 
545 			src = ntohl(ip->ip_src.s_addr);
546 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
547 			gw = ntohl(icp->icmp_gwaddr.s_addr);
548 			printf("icmp redirect from %d.%d.%d.%d: "
549 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
550 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
551 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
552 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
553 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
554 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
555 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
556 		}
557 		/*
558 		 * RFC1812 says we must ignore ICMP redirects if we
559 		 * are acting as router.
560 		 */
561 		if (drop_redirect || ipforwarding)
562 			break;
563 		if (code > 3)
564 			goto badcode;
565 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
566 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
567 			icmpstat.icps_badlen++;
568 			break;
569 		}
570 		/*
571 		 * Short circuit routing redirects to force
572 		 * immediate change in the kernel's routing
573 		 * tables.  The message is also handed to anyone
574 		 * listening on a raw socket (e.g. the routing
575 		 * daemon for use in updating its tables).
576 		 */
577 		icmpgw.sin_addr = ip->ip_src;
578 		icmpdst.sin_addr = icp->icmp_gwaddr;
579 #ifdef	ICMPPRINTFS
580 		if (icmpprintfs) {
581 			char buf[4 * sizeof "123"];
582 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
583 
584 			printf("redirect dst %s to %s\n",
585 			       buf, inet_ntoa(icp->icmp_gwaddr));
586 		}
587 #endif
588 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
589 		rtredirect((struct sockaddr *)&icmpsrc,
590 		  (struct sockaddr *)&icmpdst,
591 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
592 		  (struct sockaddr *)&icmpgw);
593 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
594 #ifdef IPSEC
595 		key_sa_routechange((struct sockaddr *)&icmpsrc);
596 #endif
597 		break;
598 
599 	/*
600 	 * No kernel processing for the following;
601 	 * just fall through to send to raw listener.
602 	 */
603 	case ICMP_ECHOREPLY:
604 	case ICMP_ROUTERADVERT:
605 	case ICMP_ROUTERSOLICIT:
606 	case ICMP_TSTAMPREPLY:
607 	case ICMP_IREQREPLY:
608 	case ICMP_MASKREPLY:
609 	default:
610 		break;
611 	}
612 
613 raw:
614 	rip_input(m, off);
615 	return;
616 
617 freeit:
618 	m_freem(m);
619 }
620 
621 /*
622  * Reflect the ip packet back to the source
623  */
624 static void
625 icmp_reflect(m)
626 	struct mbuf *m;
627 {
628 	struct ip *ip = mtod(m, struct ip *);
629 	struct ifaddr *ifa;
630 	struct ifnet *ifn;
631 	struct in_ifaddr *ia;
632 	struct in_addr t;
633 	struct mbuf *opts = 0;
634 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
635 
636 	if (!in_canforward(ip->ip_src) &&
637 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
638 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
639 		m_freem(m);	/* Bad return address */
640 		icmpstat.icps_badaddr++;
641 		goto done;	/* Ip_output() will check for broadcast */
642 	}
643 	t = ip->ip_dst;
644 	ip->ip_dst = ip->ip_src;
645 
646 	/*
647 	 * Source selection for ICMP replies:
648 	 *
649 	 * If the incoming packet was addressed directly to one of our
650 	 * own addresses, use dst as the src for the reply.
651 	 */
652 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
653 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
654 			goto match;
655 	/*
656 	 * If the incoming packet was addressed to one of our broadcast
657 	 * addresses, use the first non-broadcast address which corresponds
658 	 * to the incoming interface.
659 	 */
660 	if (m->m_pkthdr.rcvif != NULL &&
661 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
662 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
663 			if (ifa->ifa_addr->sa_family != AF_INET)
664 				continue;
665 			ia = ifatoia(ifa);
666 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
667 			    t.s_addr)
668 				goto match;
669 		}
670 	}
671 	/*
672 	 * If the packet was transiting through us, use the address of
673 	 * the interface the packet came through in.  If that interface
674 	 * doesn't have a suitable IP address, the normal selection
675 	 * criteria apply.
676 	 */
677 	if (icmp_rfi && m->m_pkthdr.rcvif != NULL) {
678 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
679 			if (ifa->ifa_addr->sa_family != AF_INET)
680 				continue;
681 			ia = ifatoia(ifa);
682 			goto match;
683 		}
684 	}
685 	/*
686 	 * If the incoming packet was not addressed directly to us, use
687 	 * designated interface for icmp replies specified by sysctl
688 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
689 	 * with normal source selection.
690 	 */
691 	if (reply_src[0] != '\0' && (ifn = ifunit(reply_src))) {
692 		TAILQ_FOREACH(ifa, &ifn->if_addrhead, ifa_link) {
693 			if (ifa->ifa_addr->sa_family != AF_INET)
694 				continue;
695 			ia = ifatoia(ifa);
696 			goto match;
697 		}
698 	}
699 	/*
700 	 * If the packet was transiting through us, use the address of
701 	 * the interface that is the closest to the packet source.
702 	 * When we don't have a route back to the packet source, stop here
703 	 * and drop the packet.
704 	 */
705 	ia = ip_rtaddr(ip->ip_dst);
706 	if (ia == NULL) {
707 		m_freem(m);
708 		icmpstat.icps_noroute++;
709 		goto done;
710 	}
711 match:
712 #ifdef MAC
713 	mac_reflect_mbuf_icmp(m);
714 #endif
715 	t = IA_SIN(ia)->sin_addr;
716 	ip->ip_src = t;
717 	ip->ip_ttl = ip_defttl;
718 
719 	if (optlen > 0) {
720 		register u_char *cp;
721 		int opt, cnt;
722 		u_int len;
723 
724 		/*
725 		 * Retrieve any source routing from the incoming packet;
726 		 * add on any record-route or timestamp options.
727 		 */
728 		cp = (u_char *) (ip + 1);
729 		if ((opts = ip_srcroute(m)) == 0 &&
730 		    (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
731 			opts->m_len = sizeof(struct in_addr);
732 			mtod(opts, struct in_addr *)->s_addr = 0;
733 		}
734 		if (opts) {
735 #ifdef ICMPPRINTFS
736 		    if (icmpprintfs)
737 			    printf("icmp_reflect optlen %d rt %d => ",
738 				optlen, opts->m_len);
739 #endif
740 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
741 			    opt = cp[IPOPT_OPTVAL];
742 			    if (opt == IPOPT_EOL)
743 				    break;
744 			    if (opt == IPOPT_NOP)
745 				    len = 1;
746 			    else {
747 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
748 					    break;
749 				    len = cp[IPOPT_OLEN];
750 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
751 				        len > cnt)
752 					    break;
753 			    }
754 			    /*
755 			     * Should check for overflow, but it "can't happen"
756 			     */
757 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
758 				opt == IPOPT_SECURITY) {
759 				    bcopy((caddr_t)cp,
760 					mtod(opts, caddr_t) + opts->m_len, len);
761 				    opts->m_len += len;
762 			    }
763 		    }
764 		    /* Terminate & pad, if necessary */
765 		    cnt = opts->m_len % 4;
766 		    if (cnt) {
767 			    for (; cnt < 4; cnt++) {
768 				    *(mtod(opts, caddr_t) + opts->m_len) =
769 					IPOPT_EOL;
770 				    opts->m_len++;
771 			    }
772 		    }
773 #ifdef ICMPPRINTFS
774 		    if (icmpprintfs)
775 			    printf("%d\n", opts->m_len);
776 #endif
777 		}
778 		/*
779 		 * Now strip out original options by copying rest of first
780 		 * mbuf's data back, and adjust the IP length.
781 		 */
782 		ip->ip_len -= optlen;
783 		ip->ip_v = IPVERSION;
784 		ip->ip_hl = 5;
785 		m->m_len -= optlen;
786 		if (m->m_flags & M_PKTHDR)
787 			m->m_pkthdr.len -= optlen;
788 		optlen += sizeof(struct ip);
789 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
790 			 (unsigned)(m->m_len - sizeof(struct ip)));
791 	}
792 	m_tag_delete_nonpersistent(m);
793 	m->m_flags &= ~(M_BCAST|M_MCAST);
794 	icmp_send(m, opts);
795 done:
796 	if (opts)
797 		(void)m_free(opts);
798 }
799 
800 /*
801  * Send an icmp packet back to the ip level,
802  * after supplying a checksum.
803  */
804 static void
805 icmp_send(m, opts)
806 	register struct mbuf *m;
807 	struct mbuf *opts;
808 {
809 	register struct ip *ip = mtod(m, struct ip *);
810 	register int hlen;
811 	register struct icmp *icp;
812 
813 	hlen = ip->ip_hl << 2;
814 	m->m_data += hlen;
815 	m->m_len -= hlen;
816 	icp = mtod(m, struct icmp *);
817 	icp->icmp_cksum = 0;
818 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
819 	m->m_data -= hlen;
820 	m->m_len += hlen;
821 	m->m_pkthdr.rcvif = (struct ifnet *)0;
822 #ifdef ICMPPRINTFS
823 	if (icmpprintfs) {
824 		char buf[4 * sizeof "123"];
825 		strcpy(buf, inet_ntoa(ip->ip_dst));
826 		printf("icmp_send dst %s src %s\n",
827 		       buf, inet_ntoa(ip->ip_src));
828 	}
829 #endif
830 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
831 }
832 
833 n_time
834 iptime()
835 {
836 	struct timeval atv;
837 	u_long t;
838 
839 	getmicrotime(&atv);
840 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
841 	return (htonl(t));
842 }
843 
844 /*
845  * Return the next larger or smaller MTU plateau (table from RFC 1191)
846  * given current value MTU.  If DIR is less than zero, a larger plateau
847  * is returned; otherwise, a smaller value is returned.
848  */
849 int
850 ip_next_mtu(mtu, dir)
851 	int mtu;
852 	int dir;
853 {
854 	static int mtutab[] = {
855 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
856 		296, 68, 0
857 	};
858 	int i, size;
859 
860 	size = (sizeof mtutab) / (sizeof mtutab[0]);
861 	if (dir >= 0) {
862 		for (i = 0; i < size; i++)
863 			if (mtu > mtutab[i])
864 				return mtutab[i];
865 	} else {
866 		for (i = size - 1; i >= 0; i--)
867 			if (mtu < mtutab[i])
868 				return mtutab[i];
869 		if (mtu == mtutab[0])
870 			return mtutab[0];
871 	}
872 	return 0;
873 }
874 
875 
876 /*
877  * badport_bandlim() - check for ICMP bandwidth limit
878  *
879  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
880  *	hit our bandwidth limit and it is not ok.
881  *
882  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
883  *
884  *	For now we separate the TCP and UDP subsystems w/ different 'which'
885  *	values.  We may eventually remove this separation (and simplify the
886  *	code further).
887  *
888  *	Note that the printing of the error message is delayed so we can
889  *	properly print the icmp error rate that the system was trying to do
890  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
891  *	the 'final' error, but it doesn't make sense to solve the printing
892  *	delay with more complex code.
893  */
894 
895 int
896 badport_bandlim(int which)
897 {
898 #define	N(a)	(sizeof (a) / sizeof (a[0]))
899 	static struct rate {
900 		const char	*type;
901 		struct timeval	lasttime;
902 		int		curpps;
903 	} rates[BANDLIM_MAX+1] = {
904 		{ "icmp unreach response" },
905 		{ "icmp ping response" },
906 		{ "icmp tstamp response" },
907 		{ "closed port RST response" },
908 		{ "open port RST response" }
909 	};
910 
911 	/*
912 	 * Return ok status if feature disabled or argument out of range.
913 	 */
914 	if (icmplim > 0 && (u_int) which < N(rates)) {
915 		struct rate *r = &rates[which];
916 		int opps = r->curpps;
917 
918 		if (!ppsratecheck(&r->lasttime, &r->curpps, icmplim))
919 			return -1;	/* discard packet */
920 		/*
921 		 * If we've dropped below the threshold after having
922 		 * rate-limited traffic print the message.  This preserves
923 		 * the previous behaviour at the expense of added complexity.
924 		 */
925 		if (icmplim_output && opps > icmplim)
926 			printf("Limiting %s from %d to %d packets/sec\n",
927 				r->type, opps, icmplim);
928 	}
929 	return 0;			/* okay to send packet */
930 #undef N
931 }
932