xref: /freebsd/sys/netinet/ip_icmp.c (revision 33644623554bb0fc57ed3c7d874193a498679b22)
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  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_ipsec.h"
36 #include "opt_mac.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/mbuf.h>
41 #include <sys/protosw.h>
42 #include <sys/socket.h>
43 #include <sys/time.h>
44 #include <sys/kernel.h>
45 #include <sys/sysctl.h>
46 #include <sys/vimage.h>
47 
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/route.h>
51 
52 #include <netinet/in.h>
53 #include <netinet/in_pcb.h>
54 #include <netinet/in_systm.h>
55 #include <netinet/in_var.h>
56 #include <netinet/ip.h>
57 #include <netinet/ip_icmp.h>
58 #include <netinet/ip_var.h>
59 #include <netinet/ip_options.h>
60 #include <netinet/tcp.h>
61 #include <netinet/tcp_var.h>
62 #include <netinet/tcpip.h>
63 #include <netinet/icmp_var.h>
64 
65 #ifdef IPSEC
66 #include <netipsec/ipsec.h>
67 #include <netipsec/key.h>
68 #endif
69 
70 #include <machine/in_cksum.h>
71 
72 #include <security/mac/mac_framework.h>
73 
74 /*
75  * ICMP routines: error generation, receive packet processing, and
76  * routines to turnaround packets back to the originator, and
77  * host table maintenance routines.
78  */
79 
80 #ifdef VIMAGE_GLOBALS
81 struct icmpstat	icmpstat;
82 static int	icmpmaskrepl;
83 static u_int	icmpmaskfake;
84 static int	drop_redirect;
85 static int	log_redirect;
86 static int	icmplim;
87 static int	icmplim_output;
88 static char	reply_src[IFNAMSIZ];
89 static int	icmp_rfi;
90 static int	icmp_quotelen;
91 static int	icmpbmcastecho;
92 #endif
93 
94 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_STATS, stats,
95 	CTLFLAG_RW, icmpstat, icmpstat, "");
96 
97 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_MASKREPL, maskrepl,
98 	CTLFLAG_RW, icmpmaskrepl, 0,
99 	"Reply to ICMP Address Mask Request packets.");
100 
101 SYSCTL_V_UINT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
102 	icmpmaskfake, 0, "Fake reply to ICMP Address Mask Request packets.");
103 
104 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, drop_redirect,
105 	CTLFLAG_RW, drop_redirect, 0, "Ignore ICMP redirects");
106 
107 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, log_redirect,
108 	CTLFLAG_RW, log_redirect, 0, "Log ICMP redirects to the console");
109 
110 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_ICMPLIM, icmplim,
111 	CTLFLAG_RW, icmplim, 0, "Maximum number of ICMP responses per second");
112 
113 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, icmplim_output,
114 	CTLFLAG_RW, icmplim_output, 0,
115 	"Enable rate limiting of ICMP responses");
116 
117 SYSCTL_V_STRING(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, reply_src,
118 	CTLFLAG_RW, reply_src, IFNAMSIZ,
119 	"icmp reply source for non-local packets.");
120 
121 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, reply_from_interface,
122 	CTLFLAG_RW, icmp_rfi, 0, "ICMP reply from incoming interface for "
123 	"non-local packets");
124 
125 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
126 	icmp_quotelen, 0, "Number of bytes from original packet to "
127 	"quote in ICMP reply");
128 
129 /*
130  * ICMP broadcast echo sysctl
131  */
132 
133 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, bmcastecho,
134 	CTLFLAG_RW, icmpbmcastecho, 0, "");
135 
136 
137 #ifdef ICMPPRINTFS
138 int	icmpprintfs = 0;
139 #endif
140 
141 static void	icmp_reflect(struct mbuf *);
142 static void	icmp_send(struct mbuf *, struct mbuf *);
143 
144 extern	struct protosw inetsw[];
145 
146 void
147 icmp_init(void)
148 {
149 	INIT_VNET_INET(curvnet);
150 
151 	V_icmpmaskrepl = 0;
152 	V_icmpmaskfake = 0;
153 	V_drop_redirect = 0;
154 	V_log_redirect = 0;
155 	V_icmplim = 200;
156 	V_icmplim_output = 1;
157 	V_icmp_rfi = 0;
158 	V_icmp_quotelen = 8;
159 	V_icmpbmcastecho = 0;
160 }
161 
162 /*
163  * Generate an error packet of type error
164  * in response to bad packet ip.
165  */
166 void
167 icmp_error(struct mbuf *n, int type, int code, n_long dest, int mtu)
168 {
169 	INIT_VNET_INET(curvnet);
170 	register struct ip *oip = mtod(n, struct ip *), *nip;
171 	register unsigned oiphlen = oip->ip_hl << 2;
172 	register struct icmp *icp;
173 	register struct mbuf *m;
174 	unsigned icmplen, icmpelen, nlen;
175 
176 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
177 #ifdef ICMPPRINTFS
178 	if (icmpprintfs)
179 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
180 #endif
181 	if (type != ICMP_REDIRECT)
182 		V_icmpstat.icps_error++;
183 	/*
184 	 * Don't send error:
185 	 *  if the original packet was encrypted.
186 	 *  if not the first fragment of message.
187 	 *  in response to a multicast or broadcast packet.
188 	 *  if the old packet protocol was an ICMP error message.
189 	 */
190 	if (n->m_flags & M_DECRYPTED)
191 		goto freeit;
192 	if (oip->ip_off & ~(IP_MF|IP_DF))
193 		goto freeit;
194 	if (n->m_flags & (M_BCAST|M_MCAST))
195 		goto freeit;
196 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
197 	  n->m_len >= oiphlen + ICMP_MINLEN &&
198 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
199 		V_icmpstat.icps_oldicmp++;
200 		goto freeit;
201 	}
202 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
203 	if (oiphlen + 8 > n->m_len)
204 		goto freeit;
205 	/*
206 	 * Calculate length to quote from original packet and
207 	 * prevent the ICMP mbuf from overflowing.
208 	 * Unfortunatly this is non-trivial since ip_forward()
209 	 * sends us truncated packets.
210 	 */
211 	nlen = m_length(n, NULL);
212 	if (oip->ip_p == IPPROTO_TCP) {
213 		struct tcphdr *th;
214 		int tcphlen;
215 
216 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
217 		    n->m_next == NULL)
218 			goto stdreply;
219 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
220 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
221 			goto freeit;
222 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
223 		tcphlen = th->th_off << 2;
224 		if (tcphlen < sizeof(struct tcphdr))
225 			goto freeit;
226 		if (oip->ip_len < oiphlen + tcphlen)
227 			goto freeit;
228 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
229 			goto stdreply;
230 		if (n->m_len < oiphlen + tcphlen &&
231 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
232 			goto freeit;
233 		icmpelen = max(tcphlen, min(V_icmp_quotelen, oip->ip_len - oiphlen));
234 	} else
235 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, oip->ip_len - oiphlen));
236 
237 	icmplen = min(oiphlen + icmpelen, nlen);
238 	if (icmplen < sizeof(struct ip))
239 		goto freeit;
240 
241 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
242 		m = m_gethdr(M_DONTWAIT, MT_DATA);
243 	else
244 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
245 	if (m == NULL)
246 		goto freeit;
247 #ifdef MAC
248 	mac_netinet_icmp_reply(n, m);
249 #endif
250 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
251 	m_align(m, ICMP_MINLEN + icmplen);
252 	m->m_len = ICMP_MINLEN + icmplen;
253 
254 	/* XXX MRT  make the outgoing packet use the same FIB
255 	 * that was associated with the incoming packet
256 	 */
257 	M_SETFIB(m, M_GETFIB(n));
258 	icp = mtod(m, struct icmp *);
259 	V_icmpstat.icps_outhist[type]++;
260 	icp->icmp_type = type;
261 	if (type == ICMP_REDIRECT)
262 		icp->icmp_gwaddr.s_addr = dest;
263 	else {
264 		icp->icmp_void = 0;
265 		/*
266 		 * The following assignments assume an overlay with the
267 		 * just zeroed icmp_void field.
268 		 */
269 		if (type == ICMP_PARAMPROB) {
270 			icp->icmp_pptr = code;
271 			code = 0;
272 		} else if (type == ICMP_UNREACH &&
273 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
274 			icp->icmp_nextmtu = htons(mtu);
275 		}
276 	}
277 	icp->icmp_code = code;
278 
279 	/*
280 	 * Copy the quotation into ICMP message and
281 	 * convert quoted IP header back to network representation.
282 	 */
283 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
284 	nip = &icp->icmp_ip;
285 	nip->ip_len = htons(nip->ip_len);
286 	nip->ip_off = htons(nip->ip_off);
287 
288 	/*
289 	 * Set up ICMP message mbuf and copy old IP header (without options
290 	 * in front of ICMP message.
291 	 * If the original mbuf was meant to bypass the firewall, the error
292 	 * reply should bypass as well.
293 	 */
294 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
295 	m->m_data -= sizeof(struct ip);
296 	m->m_len += sizeof(struct ip);
297 	m->m_pkthdr.len = m->m_len;
298 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
299 	nip = mtod(m, struct ip *);
300 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
301 	nip->ip_len = m->m_len;
302 	nip->ip_v = IPVERSION;
303 	nip->ip_hl = 5;
304 	nip->ip_p = IPPROTO_ICMP;
305 	nip->ip_tos = 0;
306 	icmp_reflect(m);
307 
308 freeit:
309 	m_freem(n);
310 }
311 
312 /*
313  * Process a received ICMP message.
314  */
315 void
316 icmp_input(struct mbuf *m, int off)
317 {
318 	INIT_VNET_INET(curvnet);
319 	struct icmp *icp;
320 	struct in_ifaddr *ia;
321 	struct ip *ip = mtod(m, struct ip *);
322 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
323 	int hlen = off;
324 	int icmplen = ip->ip_len;
325 	int i, code;
326 	void (*ctlfunc)(int, struct sockaddr *, void *);
327 	int fibnum;
328 
329 	/*
330 	 * Locate icmp structure in mbuf, and check
331 	 * that not corrupted and of at least minimum length.
332 	 */
333 #ifdef ICMPPRINTFS
334 	if (icmpprintfs) {
335 		char buf[4 * sizeof "123"];
336 		strcpy(buf, inet_ntoa(ip->ip_src));
337 		printf("icmp_input from %s to %s, len %d\n",
338 		       buf, inet_ntoa(ip->ip_dst), icmplen);
339 	}
340 #endif
341 	if (icmplen < ICMP_MINLEN) {
342 		V_icmpstat.icps_tooshort++;
343 		goto freeit;
344 	}
345 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
346 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
347 		V_icmpstat.icps_tooshort++;
348 		return;
349 	}
350 	ip = mtod(m, struct ip *);
351 	m->m_len -= hlen;
352 	m->m_data += hlen;
353 	icp = mtod(m, struct icmp *);
354 	if (in_cksum(m, icmplen)) {
355 		V_icmpstat.icps_checksum++;
356 		goto freeit;
357 	}
358 	m->m_len += hlen;
359 	m->m_data -= hlen;
360 
361 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
362 		/*
363 		 * Deliver very specific ICMP type only.
364 		 */
365 		switch (icp->icmp_type) {
366 		case ICMP_UNREACH:
367 		case ICMP_TIMXCEED:
368 			break;
369 		default:
370 			goto freeit;
371 		}
372 	}
373 
374 #ifdef ICMPPRINTFS
375 	if (icmpprintfs)
376 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
377 		    icp->icmp_code);
378 #endif
379 
380 	/*
381 	 * Message type specific processing.
382 	 */
383 	if (icp->icmp_type > ICMP_MAXTYPE)
384 		goto raw;
385 
386 	/* Initialize */
387 	bzero(&icmpsrc, sizeof(icmpsrc));
388 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
389 	icmpsrc.sin_family = AF_INET;
390 	bzero(&icmpdst, sizeof(icmpdst));
391 	icmpdst.sin_len = sizeof(struct sockaddr_in);
392 	icmpdst.sin_family = AF_INET;
393 	bzero(&icmpgw, sizeof(icmpgw));
394 	icmpgw.sin_len = sizeof(struct sockaddr_in);
395 	icmpgw.sin_family = AF_INET;
396 
397 	V_icmpstat.icps_inhist[icp->icmp_type]++;
398 	code = icp->icmp_code;
399 	switch (icp->icmp_type) {
400 
401 	case ICMP_UNREACH:
402 		switch (code) {
403 			case ICMP_UNREACH_NET:
404 			case ICMP_UNREACH_HOST:
405 			case ICMP_UNREACH_SRCFAIL:
406 			case ICMP_UNREACH_NET_UNKNOWN:
407 			case ICMP_UNREACH_HOST_UNKNOWN:
408 			case ICMP_UNREACH_ISOLATED:
409 			case ICMP_UNREACH_TOSNET:
410 			case ICMP_UNREACH_TOSHOST:
411 			case ICMP_UNREACH_HOST_PRECEDENCE:
412 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
413 				code = PRC_UNREACH_NET;
414 				break;
415 
416 			case ICMP_UNREACH_NEEDFRAG:
417 				code = PRC_MSGSIZE;
418 				break;
419 
420 			/*
421 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
422 			 * Treat subcodes 2,3 as immediate RST
423 			 */
424 			case ICMP_UNREACH_PROTOCOL:
425 			case ICMP_UNREACH_PORT:
426 				code = PRC_UNREACH_PORT;
427 				break;
428 
429 			case ICMP_UNREACH_NET_PROHIB:
430 			case ICMP_UNREACH_HOST_PROHIB:
431 			case ICMP_UNREACH_FILTER_PROHIB:
432 				code = PRC_UNREACH_ADMIN_PROHIB;
433 				break;
434 
435 			default:
436 				goto badcode;
437 		}
438 		goto deliver;
439 
440 	case ICMP_TIMXCEED:
441 		if (code > 1)
442 			goto badcode;
443 		code += PRC_TIMXCEED_INTRANS;
444 		goto deliver;
445 
446 	case ICMP_PARAMPROB:
447 		if (code > 1)
448 			goto badcode;
449 		code = PRC_PARAMPROB;
450 		goto deliver;
451 
452 	case ICMP_SOURCEQUENCH:
453 		if (code)
454 			goto badcode;
455 		code = PRC_QUENCH;
456 	deliver:
457 		/*
458 		 * Problem with datagram; advise higher level routines.
459 		 */
460 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
461 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
462 			V_icmpstat.icps_badlen++;
463 			goto freeit;
464 		}
465 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
466 		/* Discard ICMP's in response to multicast packets */
467 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
468 			goto badcode;
469 #ifdef ICMPPRINTFS
470 		if (icmpprintfs)
471 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
472 #endif
473 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
474 		/*
475 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
476 		 * notification to TCP layer.
477 		 */
478 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
479 		if (ctlfunc)
480 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
481 				   (void *)&icp->icmp_ip);
482 		break;
483 
484 	badcode:
485 		V_icmpstat.icps_badcode++;
486 		break;
487 
488 	case ICMP_ECHO:
489 		if (!V_icmpbmcastecho
490 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
491 			V_icmpstat.icps_bmcastecho++;
492 			break;
493 		}
494 		icp->icmp_type = ICMP_ECHOREPLY;
495 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
496 			goto freeit;
497 		else
498 			goto reflect;
499 
500 	case ICMP_TSTAMP:
501 		if (!V_icmpbmcastecho
502 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
503 			V_icmpstat.icps_bmcasttstamp++;
504 			break;
505 		}
506 		if (icmplen < ICMP_TSLEN) {
507 			V_icmpstat.icps_badlen++;
508 			break;
509 		}
510 		icp->icmp_type = ICMP_TSTAMPREPLY;
511 		icp->icmp_rtime = iptime();
512 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
513 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
514 			goto freeit;
515 		else
516 			goto reflect;
517 
518 	case ICMP_MASKREQ:
519 		if (V_icmpmaskrepl == 0)
520 			break;
521 		/*
522 		 * We are not able to respond with all ones broadcast
523 		 * unless we receive it over a point-to-point interface.
524 		 */
525 		if (icmplen < ICMP_MASKLEN)
526 			break;
527 		switch (ip->ip_dst.s_addr) {
528 
529 		case INADDR_BROADCAST:
530 		case INADDR_ANY:
531 			icmpdst.sin_addr = ip->ip_src;
532 			break;
533 
534 		default:
535 			icmpdst.sin_addr = ip->ip_dst;
536 		}
537 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
538 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
539 		if (ia == 0)
540 			break;
541 		if (ia->ia_ifp == 0)
542 			break;
543 		icp->icmp_type = ICMP_MASKREPLY;
544 		if (V_icmpmaskfake == 0)
545 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
546 		else
547 			icp->icmp_mask = V_icmpmaskfake;
548 		if (ip->ip_src.s_addr == 0) {
549 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
550 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
551 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
552 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
553 		}
554 reflect:
555 		ip->ip_len += hlen;	/* since ip_input deducts this */
556 		V_icmpstat.icps_reflect++;
557 		V_icmpstat.icps_outhist[icp->icmp_type]++;
558 		icmp_reflect(m);
559 		return;
560 
561 	case ICMP_REDIRECT:
562 		if (V_log_redirect) {
563 			u_long src, dst, gw;
564 
565 			src = ntohl(ip->ip_src.s_addr);
566 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
567 			gw = ntohl(icp->icmp_gwaddr.s_addr);
568 			printf("icmp redirect from %d.%d.%d.%d: "
569 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
570 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
571 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
572 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
573 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
574 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
575 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
576 		}
577 		/*
578 		 * RFC1812 says we must ignore ICMP redirects if we
579 		 * are acting as router.
580 		 */
581 		if (V_drop_redirect || V_ipforwarding)
582 			break;
583 		if (code > 3)
584 			goto badcode;
585 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
586 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
587 			V_icmpstat.icps_badlen++;
588 			break;
589 		}
590 		/*
591 		 * Short circuit routing redirects to force
592 		 * immediate change in the kernel's routing
593 		 * tables.  The message is also handed to anyone
594 		 * listening on a raw socket (e.g. the routing
595 		 * daemon for use in updating its tables).
596 		 */
597 		icmpgw.sin_addr = ip->ip_src;
598 		icmpdst.sin_addr = icp->icmp_gwaddr;
599 #ifdef	ICMPPRINTFS
600 		if (icmpprintfs) {
601 			char buf[4 * sizeof "123"];
602 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
603 
604 			printf("redirect dst %s to %s\n",
605 			       buf, inet_ntoa(icp->icmp_gwaddr));
606 		}
607 #endif
608 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
609 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
610 			in_rtredirect((struct sockaddr *)&icmpsrc,
611 			  (struct sockaddr *)&icmpdst,
612 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
613 			  (struct sockaddr *)&icmpgw, fibnum);
614 		}
615 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
616 #ifdef IPSEC
617 		key_sa_routechange((struct sockaddr *)&icmpsrc);
618 #endif
619 		break;
620 
621 	/*
622 	 * No kernel processing for the following;
623 	 * just fall through to send to raw listener.
624 	 */
625 	case ICMP_ECHOREPLY:
626 	case ICMP_ROUTERADVERT:
627 	case ICMP_ROUTERSOLICIT:
628 	case ICMP_TSTAMPREPLY:
629 	case ICMP_IREQREPLY:
630 	case ICMP_MASKREPLY:
631 	default:
632 		break;
633 	}
634 
635 raw:
636 	rip_input(m, off);
637 	return;
638 
639 freeit:
640 	m_freem(m);
641 }
642 
643 /*
644  * Reflect the ip packet back to the source
645  */
646 static void
647 icmp_reflect(struct mbuf *m)
648 {
649 	INIT_VNET_INET(curvnet);
650 	struct ip *ip = mtod(m, struct ip *);
651 	struct ifaddr *ifa;
652 	struct ifnet *ifn;
653 	struct in_ifaddr *ia;
654 	struct in_addr t;
655 	struct mbuf *opts = 0;
656 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
657 
658 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
659 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
660 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
661 		m_freem(m);	/* Bad return address */
662 		V_icmpstat.icps_badaddr++;
663 		goto done;	/* Ip_output() will check for broadcast */
664 	}
665 
666 	t = ip->ip_dst;
667 	ip->ip_dst = ip->ip_src;
668 
669 	/*
670 	 * Source selection for ICMP replies:
671 	 *
672 	 * If the incoming packet was addressed directly to one of our
673 	 * own addresses, use dst as the src for the reply.
674 	 */
675 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
676 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
677 			goto match;
678 	/*
679 	 * If the incoming packet was addressed to one of our broadcast
680 	 * addresses, use the first non-broadcast address which corresponds
681 	 * to the incoming interface.
682 	 */
683 	if (m->m_pkthdr.rcvif != NULL &&
684 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
685 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
686 			if (ifa->ifa_addr->sa_family != AF_INET)
687 				continue;
688 			ia = ifatoia(ifa);
689 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
690 			    t.s_addr)
691 				goto match;
692 		}
693 	}
694 	/*
695 	 * If the packet was transiting through us, use the address of
696 	 * the interface the packet came through in.  If that interface
697 	 * doesn't have a suitable IP address, the normal selection
698 	 * criteria apply.
699 	 */
700 	if (V_icmp_rfi && m->m_pkthdr.rcvif != NULL) {
701 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
702 			if (ifa->ifa_addr->sa_family != AF_INET)
703 				continue;
704 			ia = ifatoia(ifa);
705 			goto match;
706 		}
707 	}
708 	/*
709 	 * If the incoming packet was not addressed directly to us, use
710 	 * designated interface for icmp replies specified by sysctl
711 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
712 	 * with normal source selection.
713 	 */
714 	if (V_reply_src[0] != '\0' && (ifn = ifunit(V_reply_src))) {
715 		TAILQ_FOREACH(ifa, &ifn->if_addrhead, ifa_link) {
716 			if (ifa->ifa_addr->sa_family != AF_INET)
717 				continue;
718 			ia = ifatoia(ifa);
719 			goto match;
720 		}
721 	}
722 	/*
723 	 * If the packet was transiting through us, use the address of
724 	 * the interface that is the closest to the packet source.
725 	 * When we don't have a route back to the packet source, stop here
726 	 * and drop the packet.
727 	 */
728 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
729 	if (ia == NULL) {
730 		m_freem(m);
731 		V_icmpstat.icps_noroute++;
732 		goto done;
733 	}
734 match:
735 #ifdef MAC
736 	mac_netinet_icmp_replyinplace(m);
737 #endif
738 	t = IA_SIN(ia)->sin_addr;
739 	ip->ip_src = t;
740 	ip->ip_ttl = V_ip_defttl;
741 
742 	if (optlen > 0) {
743 		register u_char *cp;
744 		int opt, cnt;
745 		u_int len;
746 
747 		/*
748 		 * Retrieve any source routing from the incoming packet;
749 		 * add on any record-route or timestamp options.
750 		 */
751 		cp = (u_char *) (ip + 1);
752 		if ((opts = ip_srcroute(m)) == 0 &&
753 		    (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
754 			opts->m_len = sizeof(struct in_addr);
755 			mtod(opts, struct in_addr *)->s_addr = 0;
756 		}
757 		if (opts) {
758 #ifdef ICMPPRINTFS
759 		    if (icmpprintfs)
760 			    printf("icmp_reflect optlen %d rt %d => ",
761 				optlen, opts->m_len);
762 #endif
763 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
764 			    opt = cp[IPOPT_OPTVAL];
765 			    if (opt == IPOPT_EOL)
766 				    break;
767 			    if (opt == IPOPT_NOP)
768 				    len = 1;
769 			    else {
770 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
771 					    break;
772 				    len = cp[IPOPT_OLEN];
773 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
774 				        len > cnt)
775 					    break;
776 			    }
777 			    /*
778 			     * Should check for overflow, but it "can't happen"
779 			     */
780 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
781 				opt == IPOPT_SECURITY) {
782 				    bcopy((caddr_t)cp,
783 					mtod(opts, caddr_t) + opts->m_len, len);
784 				    opts->m_len += len;
785 			    }
786 		    }
787 		    /* Terminate & pad, if necessary */
788 		    cnt = opts->m_len % 4;
789 		    if (cnt) {
790 			    for (; cnt < 4; cnt++) {
791 				    *(mtod(opts, caddr_t) + opts->m_len) =
792 					IPOPT_EOL;
793 				    opts->m_len++;
794 			    }
795 		    }
796 #ifdef ICMPPRINTFS
797 		    if (icmpprintfs)
798 			    printf("%d\n", opts->m_len);
799 #endif
800 		}
801 		/*
802 		 * Now strip out original options by copying rest of first
803 		 * mbuf's data back, and adjust the IP length.
804 		 */
805 		ip->ip_len -= optlen;
806 		ip->ip_v = IPVERSION;
807 		ip->ip_hl = 5;
808 		m->m_len -= optlen;
809 		if (m->m_flags & M_PKTHDR)
810 			m->m_pkthdr.len -= optlen;
811 		optlen += sizeof(struct ip);
812 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
813 			 (unsigned)(m->m_len - sizeof(struct ip)));
814 	}
815 	m_tag_delete_nonpersistent(m);
816 	m->m_flags &= ~(M_BCAST|M_MCAST);
817 	icmp_send(m, opts);
818 done:
819 	if (opts)
820 		(void)m_free(opts);
821 }
822 
823 /*
824  * Send an icmp packet back to the ip level,
825  * after supplying a checksum.
826  */
827 static void
828 icmp_send(struct mbuf *m, struct mbuf *opts)
829 {
830 	register struct ip *ip = mtod(m, struct ip *);
831 	register int hlen;
832 	register struct icmp *icp;
833 
834 	hlen = ip->ip_hl << 2;
835 	m->m_data += hlen;
836 	m->m_len -= hlen;
837 	icp = mtod(m, struct icmp *);
838 	icp->icmp_cksum = 0;
839 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
840 	m->m_data -= hlen;
841 	m->m_len += hlen;
842 	m->m_pkthdr.rcvif = (struct ifnet *)0;
843 #ifdef ICMPPRINTFS
844 	if (icmpprintfs) {
845 		char buf[4 * sizeof "123"];
846 		strcpy(buf, inet_ntoa(ip->ip_dst));
847 		printf("icmp_send dst %s src %s\n",
848 		       buf, inet_ntoa(ip->ip_src));
849 	}
850 #endif
851 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
852 }
853 
854 n_time
855 iptime(void)
856 {
857 	struct timeval atv;
858 	u_long t;
859 
860 	getmicrotime(&atv);
861 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
862 	return (htonl(t));
863 }
864 
865 /*
866  * Return the next larger or smaller MTU plateau (table from RFC 1191)
867  * given current value MTU.  If DIR is less than zero, a larger plateau
868  * is returned; otherwise, a smaller value is returned.
869  */
870 int
871 ip_next_mtu(int mtu, int dir)
872 {
873 	static int mtutab[] = {
874 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
875 		296, 68, 0
876 	};
877 	int i, size;
878 
879 	size = (sizeof mtutab) / (sizeof mtutab[0]);
880 	if (dir >= 0) {
881 		for (i = 0; i < size; i++)
882 			if (mtu > mtutab[i])
883 				return mtutab[i];
884 	} else {
885 		for (i = size - 1; i >= 0; i--)
886 			if (mtu < mtutab[i])
887 				return mtutab[i];
888 		if (mtu == mtutab[0])
889 			return mtutab[0];
890 	}
891 	return 0;
892 }
893 
894 
895 /*
896  * badport_bandlim() - check for ICMP bandwidth limit
897  *
898  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
899  *	hit our bandwidth limit and it is not ok.
900  *
901  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
902  *
903  *	For now we separate the TCP and UDP subsystems w/ different 'which'
904  *	values.  We may eventually remove this separation (and simplify the
905  *	code further).
906  *
907  *	Note that the printing of the error message is delayed so we can
908  *	properly print the icmp error rate that the system was trying to do
909  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
910  *	the 'final' error, but it doesn't make sense to solve the printing
911  *	delay with more complex code.
912  */
913 
914 int
915 badport_bandlim(int which)
916 {
917 	INIT_VNET_INET(curvnet);
918 
919 #define	N(a)	(sizeof (a) / sizeof (a[0]))
920 	static struct rate {
921 		const char	*type;
922 		struct timeval	lasttime;
923 		int		curpps;
924 	} rates[BANDLIM_MAX+1] = {
925 		{ "icmp unreach response" },
926 		{ "icmp ping response" },
927 		{ "icmp tstamp response" },
928 		{ "closed port RST response" },
929 		{ "open port RST response" },
930 		{ "icmp6 unreach response" }
931 	};
932 
933 	/*
934 	 * Return ok status if feature disabled or argument out of range.
935 	 */
936 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
937 		struct rate *r = &rates[which];
938 		int opps = r->curpps;
939 
940 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
941 			return -1;	/* discard packet */
942 		/*
943 		 * If we've dropped below the threshold after having
944 		 * rate-limited traffic print the message.  This preserves
945 		 * the previous behaviour at the expense of added complexity.
946 		 */
947 		if (V_icmplim_output && opps > V_icmplim)
948 			printf("Limiting %s from %d to %d packets/sec\n",
949 				r->type, opps, V_icmplim);
950 	}
951 	return 0;			/* okay to send packet */
952 #undef N
953 }
954