xref: /freebsd/sys/netinet/ip_icmp.c (revision 4a5216a6dc0c3ce4cf5f2d3ee8af0c3ff3402c4f)
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 struct	icmpstat icmpstat;
81 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_STATS, stats,
82 	CTLFLAG_RW, icmpstat, icmpstat, "");
83 
84 static int	icmpmaskrepl = 0;
85 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_MASKREPL, maskrepl,
86 	CTLFLAG_RW, icmpmaskrepl, 0,
87 	"Reply to ICMP Address Mask Request packets.");
88 
89 static u_int	icmpmaskfake = 0;
90 SYSCTL_V_UINT(V_NET, vnet_inet, _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_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, drop_redirect,
95 	CTLFLAG_RW, drop_redirect, 0, "Ignore ICMP redirects");
96 
97 static int	log_redirect = 0;
98 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, log_redirect,
99 	CTLFLAG_RW, log_redirect, 0, "Log ICMP redirects to the console");
100 
101 static int      icmplim = 200;
102 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_ICMPLIM, icmplim,
103 	CTLFLAG_RW, icmplim, 0, "Maximum number of ICMP responses per second");
104 
105 static int	icmplim_output = 1;
106 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, icmplim_output,
107 	CTLFLAG_RW, icmplim_output, 0,
108 	"Enable rate limiting of ICMP responses");
109 
110 static char	reply_src[IFNAMSIZ];
111 SYSCTL_V_STRING(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, reply_src,
112 	CTLFLAG_RW, reply_src, IFNAMSIZ,
113 	"icmp reply source for non-local packets.");
114 
115 static int	icmp_rfi = 0;
116 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, reply_from_interface,
117 	CTLFLAG_RW, icmp_rfi, 0, "ICMP reply from incoming interface for "
118 	"non-local packets");
119 
120 static int	icmp_quotelen = 8;
121 SYSCTL_V_INT(V_NET, vnet_inet, _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_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, bmcastecho,
131 	CTLFLAG_RW, 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 	INIT_VNET_INET(curvnet);
151 	register struct ip *oip = mtod(n, struct ip *), *nip;
152 	register unsigned oiphlen = oip->ip_hl << 2;
153 	register struct icmp *icp;
154 	register struct mbuf *m;
155 	unsigned icmplen, icmpelen, nlen;
156 
157 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
158 #ifdef ICMPPRINTFS
159 	if (icmpprintfs)
160 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
161 #endif
162 	if (type != ICMP_REDIRECT)
163 		V_icmpstat.icps_error++;
164 	/*
165 	 * Don't send error:
166 	 *  if the original packet was encrypted.
167 	 *  if not the first fragment of message.
168 	 *  in response to a multicast or broadcast packet.
169 	 *  if the old packet protocol was an ICMP error message.
170 	 */
171 	if (n->m_flags & M_DECRYPTED)
172 		goto freeit;
173 	if (oip->ip_off & ~(IP_MF|IP_DF))
174 		goto freeit;
175 	if (n->m_flags & (M_BCAST|M_MCAST))
176 		goto freeit;
177 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
178 	  n->m_len >= oiphlen + ICMP_MINLEN &&
179 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
180 		V_icmpstat.icps_oldicmp++;
181 		goto freeit;
182 	}
183 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
184 	if (oiphlen + 8 > n->m_len)
185 		goto freeit;
186 	/*
187 	 * Calculate length to quote from original packet and
188 	 * prevent the ICMP mbuf from overflowing.
189 	 * Unfortunatly this is non-trivial since ip_forward()
190 	 * sends us truncated packets.
191 	 */
192 	nlen = m_length(n, NULL);
193 	if (oip->ip_p == IPPROTO_TCP) {
194 		struct tcphdr *th;
195 		int tcphlen;
196 
197 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
198 		    n->m_next == NULL)
199 			goto stdreply;
200 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
201 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
202 			goto freeit;
203 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
204 		tcphlen = th->th_off << 2;
205 		if (tcphlen < sizeof(struct tcphdr))
206 			goto freeit;
207 		if (oip->ip_len < oiphlen + tcphlen)
208 			goto freeit;
209 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
210 			goto stdreply;
211 		if (n->m_len < oiphlen + tcphlen &&
212 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
213 			goto freeit;
214 		icmpelen = max(tcphlen, min(V_icmp_quotelen, oip->ip_len - oiphlen));
215 	} else
216 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, oip->ip_len - oiphlen));
217 
218 	icmplen = min(oiphlen + icmpelen, nlen);
219 	if (icmplen < sizeof(struct ip))
220 		goto freeit;
221 
222 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
223 		m = m_gethdr(M_DONTWAIT, MT_DATA);
224 	else
225 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
226 	if (m == NULL)
227 		goto freeit;
228 #ifdef MAC
229 	mac_netinet_icmp_reply(n, m);
230 #endif
231 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
232 	m_align(m, ICMP_MINLEN + icmplen);
233 	m->m_len = ICMP_MINLEN + icmplen;
234 
235 	/* XXX MRT  make the outgoing packet use the same FIB
236 	 * that was associated with the incoming packet
237 	 */
238 	M_SETFIB(m, M_GETFIB(n));
239 	icp = mtod(m, struct icmp *);
240 	V_icmpstat.icps_outhist[type]++;
241 	icp->icmp_type = type;
242 	if (type == ICMP_REDIRECT)
243 		icp->icmp_gwaddr.s_addr = dest;
244 	else {
245 		icp->icmp_void = 0;
246 		/*
247 		 * The following assignments assume an overlay with the
248 		 * just zeroed icmp_void field.
249 		 */
250 		if (type == ICMP_PARAMPROB) {
251 			icp->icmp_pptr = code;
252 			code = 0;
253 		} else if (type == ICMP_UNREACH &&
254 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
255 			icp->icmp_nextmtu = htons(mtu);
256 		}
257 	}
258 	icp->icmp_code = code;
259 
260 	/*
261 	 * Copy the quotation into ICMP message and
262 	 * convert quoted IP header back to network representation.
263 	 */
264 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
265 	nip = &icp->icmp_ip;
266 	nip->ip_len = htons(nip->ip_len);
267 	nip->ip_off = htons(nip->ip_off);
268 
269 	/*
270 	 * Set up ICMP message mbuf and copy old IP header (without options
271 	 * in front of ICMP message.
272 	 * If the original mbuf was meant to bypass the firewall, the error
273 	 * reply should bypass as well.
274 	 */
275 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
276 	m->m_data -= sizeof(struct ip);
277 	m->m_len += sizeof(struct ip);
278 	m->m_pkthdr.len = m->m_len;
279 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
280 	nip = mtod(m, struct ip *);
281 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
282 	nip->ip_len = m->m_len;
283 	nip->ip_v = IPVERSION;
284 	nip->ip_hl = 5;
285 	nip->ip_p = IPPROTO_ICMP;
286 	nip->ip_tos = 0;
287 	icmp_reflect(m);
288 
289 freeit:
290 	m_freem(n);
291 }
292 
293 /*
294  * Process a received ICMP message.
295  */
296 void
297 icmp_input(struct mbuf *m, int off)
298 {
299 	INIT_VNET_INET(curvnet);
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 	int fibnum;
309 
310 	/*
311 	 * Locate icmp structure in mbuf, and check
312 	 * that not corrupted and of at least minimum length.
313 	 */
314 #ifdef ICMPPRINTFS
315 	if (icmpprintfs) {
316 		char buf[4 * sizeof "123"];
317 		strcpy(buf, inet_ntoa(ip->ip_src));
318 		printf("icmp_input from %s to %s, len %d\n",
319 		       buf, inet_ntoa(ip->ip_dst), icmplen);
320 	}
321 #endif
322 	if (icmplen < ICMP_MINLEN) {
323 		V_icmpstat.icps_tooshort++;
324 		goto freeit;
325 	}
326 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
327 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
328 		V_icmpstat.icps_tooshort++;
329 		return;
330 	}
331 	ip = mtod(m, struct ip *);
332 	m->m_len -= hlen;
333 	m->m_data += hlen;
334 	icp = mtod(m, struct icmp *);
335 	if (in_cksum(m, icmplen)) {
336 		V_icmpstat.icps_checksum++;
337 		goto freeit;
338 	}
339 	m->m_len += hlen;
340 	m->m_data -= hlen;
341 
342 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
343 		/*
344 		 * Deliver very specific ICMP type only.
345 		 */
346 		switch (icp->icmp_type) {
347 		case ICMP_UNREACH:
348 		case ICMP_TIMXCEED:
349 			break;
350 		default:
351 			goto freeit;
352 		}
353 	}
354 
355 #ifdef ICMPPRINTFS
356 	if (icmpprintfs)
357 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
358 		    icp->icmp_code);
359 #endif
360 
361 	/*
362 	 * Message type specific processing.
363 	 */
364 	if (icp->icmp_type > ICMP_MAXTYPE)
365 		goto raw;
366 
367 	/* Initialize */
368 	bzero(&icmpsrc, sizeof(icmpsrc));
369 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
370 	icmpsrc.sin_family = AF_INET;
371 	bzero(&icmpdst, sizeof(icmpdst));
372 	icmpdst.sin_len = sizeof(struct sockaddr_in);
373 	icmpdst.sin_family = AF_INET;
374 	bzero(&icmpgw, sizeof(icmpgw));
375 	icmpgw.sin_len = sizeof(struct sockaddr_in);
376 	icmpgw.sin_family = AF_INET;
377 
378 	V_icmpstat.icps_inhist[icp->icmp_type]++;
379 	code = icp->icmp_code;
380 	switch (icp->icmp_type) {
381 
382 	case ICMP_UNREACH:
383 		switch (code) {
384 			case ICMP_UNREACH_NET:
385 			case ICMP_UNREACH_HOST:
386 			case ICMP_UNREACH_SRCFAIL:
387 			case ICMP_UNREACH_NET_UNKNOWN:
388 			case ICMP_UNREACH_HOST_UNKNOWN:
389 			case ICMP_UNREACH_ISOLATED:
390 			case ICMP_UNREACH_TOSNET:
391 			case ICMP_UNREACH_TOSHOST:
392 			case ICMP_UNREACH_HOST_PRECEDENCE:
393 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
394 				code = PRC_UNREACH_NET;
395 				break;
396 
397 			case ICMP_UNREACH_NEEDFRAG:
398 				code = PRC_MSGSIZE;
399 				break;
400 
401 			/*
402 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
403 			 * Treat subcodes 2,3 as immediate RST
404 			 */
405 			case ICMP_UNREACH_PROTOCOL:
406 			case ICMP_UNREACH_PORT:
407 				code = PRC_UNREACH_PORT;
408 				break;
409 
410 			case ICMP_UNREACH_NET_PROHIB:
411 			case ICMP_UNREACH_HOST_PROHIB:
412 			case ICMP_UNREACH_FILTER_PROHIB:
413 				code = PRC_UNREACH_ADMIN_PROHIB;
414 				break;
415 
416 			default:
417 				goto badcode;
418 		}
419 		goto deliver;
420 
421 	case ICMP_TIMXCEED:
422 		if (code > 1)
423 			goto badcode;
424 		code += PRC_TIMXCEED_INTRANS;
425 		goto deliver;
426 
427 	case ICMP_PARAMPROB:
428 		if (code > 1)
429 			goto badcode;
430 		code = PRC_PARAMPROB;
431 		goto deliver;
432 
433 	case ICMP_SOURCEQUENCH:
434 		if (code)
435 			goto badcode;
436 		code = PRC_QUENCH;
437 	deliver:
438 		/*
439 		 * Problem with datagram; advise higher level routines.
440 		 */
441 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
442 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
443 			V_icmpstat.icps_badlen++;
444 			goto freeit;
445 		}
446 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
447 		/* Discard ICMP's in response to multicast packets */
448 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
449 			goto badcode;
450 #ifdef ICMPPRINTFS
451 		if (icmpprintfs)
452 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
453 #endif
454 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
455 		/*
456 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
457 		 * notification to TCP layer.
458 		 */
459 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
460 		if (ctlfunc)
461 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
462 				   (void *)&icp->icmp_ip);
463 		break;
464 
465 	badcode:
466 		V_icmpstat.icps_badcode++;
467 		break;
468 
469 	case ICMP_ECHO:
470 		if (!V_icmpbmcastecho
471 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
472 			V_icmpstat.icps_bmcastecho++;
473 			break;
474 		}
475 		icp->icmp_type = ICMP_ECHOREPLY;
476 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
477 			goto freeit;
478 		else
479 			goto reflect;
480 
481 	case ICMP_TSTAMP:
482 		if (!V_icmpbmcastecho
483 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
484 			V_icmpstat.icps_bmcasttstamp++;
485 			break;
486 		}
487 		if (icmplen < ICMP_TSLEN) {
488 			V_icmpstat.icps_badlen++;
489 			break;
490 		}
491 		icp->icmp_type = ICMP_TSTAMPREPLY;
492 		icp->icmp_rtime = iptime();
493 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
494 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
495 			goto freeit;
496 		else
497 			goto reflect;
498 
499 	case ICMP_MASKREQ:
500 		if (V_icmpmaskrepl == 0)
501 			break;
502 		/*
503 		 * We are not able to respond with all ones broadcast
504 		 * unless we receive it over a point-to-point interface.
505 		 */
506 		if (icmplen < ICMP_MASKLEN)
507 			break;
508 		switch (ip->ip_dst.s_addr) {
509 
510 		case INADDR_BROADCAST:
511 		case INADDR_ANY:
512 			icmpdst.sin_addr = ip->ip_src;
513 			break;
514 
515 		default:
516 			icmpdst.sin_addr = ip->ip_dst;
517 		}
518 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
519 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
520 		if (ia == 0)
521 			break;
522 		if (ia->ia_ifp == 0)
523 			break;
524 		icp->icmp_type = ICMP_MASKREPLY;
525 		if (V_icmpmaskfake == 0)
526 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
527 		else
528 			icp->icmp_mask = V_icmpmaskfake;
529 		if (ip->ip_src.s_addr == 0) {
530 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
531 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
532 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
533 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
534 		}
535 reflect:
536 		ip->ip_len += hlen;	/* since ip_input deducts this */
537 		V_icmpstat.icps_reflect++;
538 		V_icmpstat.icps_outhist[icp->icmp_type]++;
539 		icmp_reflect(m);
540 		return;
541 
542 	case ICMP_REDIRECT:
543 		if (V_log_redirect) {
544 			u_long src, dst, gw;
545 
546 			src = ntohl(ip->ip_src.s_addr);
547 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
548 			gw = ntohl(icp->icmp_gwaddr.s_addr);
549 			printf("icmp redirect from %d.%d.%d.%d: "
550 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
551 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
552 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
553 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
554 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
555 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
556 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
557 		}
558 		/*
559 		 * RFC1812 says we must ignore ICMP redirects if we
560 		 * are acting as router.
561 		 */
562 		if (V_drop_redirect || V_ipforwarding)
563 			break;
564 		if (code > 3)
565 			goto badcode;
566 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
567 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
568 			V_icmpstat.icps_badlen++;
569 			break;
570 		}
571 		/*
572 		 * Short circuit routing redirects to force
573 		 * immediate change in the kernel's routing
574 		 * tables.  The message is also handed to anyone
575 		 * listening on a raw socket (e.g. the routing
576 		 * daemon for use in updating its tables).
577 		 */
578 		icmpgw.sin_addr = ip->ip_src;
579 		icmpdst.sin_addr = icp->icmp_gwaddr;
580 #ifdef	ICMPPRINTFS
581 		if (icmpprintfs) {
582 			char buf[4 * sizeof "123"];
583 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
584 
585 			printf("redirect dst %s to %s\n",
586 			       buf, inet_ntoa(icp->icmp_gwaddr));
587 		}
588 #endif
589 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
590 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
591 			in_rtredirect((struct sockaddr *)&icmpsrc,
592 			  (struct sockaddr *)&icmpdst,
593 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
594 			  (struct sockaddr *)&icmpgw, fibnum);
595 		}
596 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
597 #ifdef IPSEC
598 		key_sa_routechange((struct sockaddr *)&icmpsrc);
599 #endif
600 		break;
601 
602 	/*
603 	 * No kernel processing for the following;
604 	 * just fall through to send to raw listener.
605 	 */
606 	case ICMP_ECHOREPLY:
607 	case ICMP_ROUTERADVERT:
608 	case ICMP_ROUTERSOLICIT:
609 	case ICMP_TSTAMPREPLY:
610 	case ICMP_IREQREPLY:
611 	case ICMP_MASKREPLY:
612 	default:
613 		break;
614 	}
615 
616 raw:
617 	rip_input(m, off);
618 	return;
619 
620 freeit:
621 	m_freem(m);
622 }
623 
624 /*
625  * Reflect the ip packet back to the source
626  */
627 static void
628 icmp_reflect(struct mbuf *m)
629 {
630 	INIT_VNET_INET(curvnet);
631 	struct ip *ip = mtod(m, struct ip *);
632 	struct ifaddr *ifa;
633 	struct ifnet *ifn;
634 	struct in_ifaddr *ia;
635 	struct in_addr t;
636 	struct mbuf *opts = 0;
637 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
638 
639 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
640 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
641 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
642 		m_freem(m);	/* Bad return address */
643 		V_icmpstat.icps_badaddr++;
644 		goto done;	/* Ip_output() will check for broadcast */
645 	}
646 
647 	t = ip->ip_dst;
648 	ip->ip_dst = ip->ip_src;
649 
650 	/*
651 	 * Source selection for ICMP replies:
652 	 *
653 	 * If the incoming packet was addressed directly to one of our
654 	 * own addresses, use dst as the src for the reply.
655 	 */
656 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
657 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
658 			goto match;
659 	/*
660 	 * If the incoming packet was addressed to one of our broadcast
661 	 * addresses, use the first non-broadcast address which corresponds
662 	 * to the incoming interface.
663 	 */
664 	if (m->m_pkthdr.rcvif != NULL &&
665 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
666 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
667 			if (ifa->ifa_addr->sa_family != AF_INET)
668 				continue;
669 			ia = ifatoia(ifa);
670 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
671 			    t.s_addr)
672 				goto match;
673 		}
674 	}
675 	/*
676 	 * If the packet was transiting through us, use the address of
677 	 * the interface the packet came through in.  If that interface
678 	 * doesn't have a suitable IP address, the normal selection
679 	 * criteria apply.
680 	 */
681 	if (V_icmp_rfi && m->m_pkthdr.rcvif != NULL) {
682 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
683 			if (ifa->ifa_addr->sa_family != AF_INET)
684 				continue;
685 			ia = ifatoia(ifa);
686 			goto match;
687 		}
688 	}
689 	/*
690 	 * If the incoming packet was not addressed directly to us, use
691 	 * designated interface for icmp replies specified by sysctl
692 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
693 	 * with normal source selection.
694 	 */
695 	if (V_reply_src[0] != '\0' && (ifn = ifunit(V_reply_src))) {
696 		TAILQ_FOREACH(ifa, &ifn->if_addrhead, ifa_link) {
697 			if (ifa->ifa_addr->sa_family != AF_INET)
698 				continue;
699 			ia = ifatoia(ifa);
700 			goto match;
701 		}
702 	}
703 	/*
704 	 * If the packet was transiting through us, use the address of
705 	 * the interface that is the closest to the packet source.
706 	 * When we don't have a route back to the packet source, stop here
707 	 * and drop the packet.
708 	 */
709 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
710 	if (ia == NULL) {
711 		m_freem(m);
712 		V_icmpstat.icps_noroute++;
713 		goto done;
714 	}
715 match:
716 #ifdef MAC
717 	mac_netinet_icmp_replyinplace(m);
718 #endif
719 	t = IA_SIN(ia)->sin_addr;
720 	ip->ip_src = t;
721 	ip->ip_ttl = V_ip_defttl;
722 
723 	if (optlen > 0) {
724 		register u_char *cp;
725 		int opt, cnt;
726 		u_int len;
727 
728 		/*
729 		 * Retrieve any source routing from the incoming packet;
730 		 * add on any record-route or timestamp options.
731 		 */
732 		cp = (u_char *) (ip + 1);
733 		if ((opts = ip_srcroute(m)) == 0 &&
734 		    (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
735 			opts->m_len = sizeof(struct in_addr);
736 			mtod(opts, struct in_addr *)->s_addr = 0;
737 		}
738 		if (opts) {
739 #ifdef ICMPPRINTFS
740 		    if (icmpprintfs)
741 			    printf("icmp_reflect optlen %d rt %d => ",
742 				optlen, opts->m_len);
743 #endif
744 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
745 			    opt = cp[IPOPT_OPTVAL];
746 			    if (opt == IPOPT_EOL)
747 				    break;
748 			    if (opt == IPOPT_NOP)
749 				    len = 1;
750 			    else {
751 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
752 					    break;
753 				    len = cp[IPOPT_OLEN];
754 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
755 				        len > cnt)
756 					    break;
757 			    }
758 			    /*
759 			     * Should check for overflow, but it "can't happen"
760 			     */
761 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
762 				opt == IPOPT_SECURITY) {
763 				    bcopy((caddr_t)cp,
764 					mtod(opts, caddr_t) + opts->m_len, len);
765 				    opts->m_len += len;
766 			    }
767 		    }
768 		    /* Terminate & pad, if necessary */
769 		    cnt = opts->m_len % 4;
770 		    if (cnt) {
771 			    for (; cnt < 4; cnt++) {
772 				    *(mtod(opts, caddr_t) + opts->m_len) =
773 					IPOPT_EOL;
774 				    opts->m_len++;
775 			    }
776 		    }
777 #ifdef ICMPPRINTFS
778 		    if (icmpprintfs)
779 			    printf("%d\n", opts->m_len);
780 #endif
781 		}
782 		/*
783 		 * Now strip out original options by copying rest of first
784 		 * mbuf's data back, and adjust the IP length.
785 		 */
786 		ip->ip_len -= optlen;
787 		ip->ip_v = IPVERSION;
788 		ip->ip_hl = 5;
789 		m->m_len -= optlen;
790 		if (m->m_flags & M_PKTHDR)
791 			m->m_pkthdr.len -= optlen;
792 		optlen += sizeof(struct ip);
793 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
794 			 (unsigned)(m->m_len - sizeof(struct ip)));
795 	}
796 	m_tag_delete_nonpersistent(m);
797 	m->m_flags &= ~(M_BCAST|M_MCAST);
798 	icmp_send(m, opts);
799 done:
800 	if (opts)
801 		(void)m_free(opts);
802 }
803 
804 /*
805  * Send an icmp packet back to the ip level,
806  * after supplying a checksum.
807  */
808 static void
809 icmp_send(struct mbuf *m, struct mbuf *opts)
810 {
811 	register struct ip *ip = mtod(m, struct ip *);
812 	register int hlen;
813 	register struct icmp *icp;
814 
815 	hlen = ip->ip_hl << 2;
816 	m->m_data += hlen;
817 	m->m_len -= hlen;
818 	icp = mtod(m, struct icmp *);
819 	icp->icmp_cksum = 0;
820 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
821 	m->m_data -= hlen;
822 	m->m_len += hlen;
823 	m->m_pkthdr.rcvif = (struct ifnet *)0;
824 #ifdef ICMPPRINTFS
825 	if (icmpprintfs) {
826 		char buf[4 * sizeof "123"];
827 		strcpy(buf, inet_ntoa(ip->ip_dst));
828 		printf("icmp_send dst %s src %s\n",
829 		       buf, inet_ntoa(ip->ip_src));
830 	}
831 #endif
832 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
833 }
834 
835 n_time
836 iptime(void)
837 {
838 	struct timeval atv;
839 	u_long t;
840 
841 	getmicrotime(&atv);
842 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
843 	return (htonl(t));
844 }
845 
846 /*
847  * Return the next larger or smaller MTU plateau (table from RFC 1191)
848  * given current value MTU.  If DIR is less than zero, a larger plateau
849  * is returned; otherwise, a smaller value is returned.
850  */
851 int
852 ip_next_mtu(int mtu, 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 	INIT_VNET_INET(curvnet);
899 
900 #define	N(a)	(sizeof (a) / sizeof (a[0]))
901 	static struct rate {
902 		const char	*type;
903 		struct timeval	lasttime;
904 		int		curpps;
905 	} rates[BANDLIM_MAX+1] = {
906 		{ "icmp unreach response" },
907 		{ "icmp ping response" },
908 		{ "icmp tstamp response" },
909 		{ "closed port RST response" },
910 		{ "open port RST response" },
911 		{ "icmp6 unreach response" }
912 	};
913 
914 	/*
915 	 * Return ok status if feature disabled or argument out of range.
916 	 */
917 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
918 		struct rate *r = &rates[which];
919 		int opps = r->curpps;
920 
921 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
922 			return -1;	/* discard packet */
923 		/*
924 		 * If we've dropped below the threshold after having
925 		 * rate-limited traffic print the message.  This preserves
926 		 * the previous behaviour at the expense of added complexity.
927 		 */
928 		if (V_icmplim_output && opps > V_icmplim)
929 			printf("Limiting %s from %d to %d packets/sec\n",
930 				r->type, opps, V_icmplim);
931 	}
932 	return 0;			/* okay to send packet */
933 #undef N
934 }
935