xref: /freebsd/sys/netinet/ip_icmp.c (revision 2b15cb3d0922bd70ea592f0da9b4a5b167f4d53f)
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_inet.h"
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mbuf.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/sysctl.h>
45 #include <sys/syslog.h>
46 
47 #include <net/if.h>
48 #include <net/if_var.h>
49 #include <net/if_types.h>
50 #include <net/route.h>
51 #include <net/vnet.h>
52 
53 #include <netinet/in.h>
54 #include <netinet/in_pcb.h>
55 #include <netinet/in_systm.h>
56 #include <netinet/in_var.h>
57 #include <netinet/ip.h>
58 #include <netinet/ip_icmp.h>
59 #include <netinet/ip_var.h>
60 #include <netinet/ip_options.h>
61 #include <netinet/tcp.h>
62 #include <netinet/tcp_var.h>
63 #include <netinet/tcpip.h>
64 #include <netinet/icmp_var.h>
65 
66 #ifdef INET
67 
68 #include <machine/in_cksum.h>
69 
70 #include <security/mac/mac_framework.h>
71 #endif /* INET */
72 
73 /*
74  * ICMP routines: error generation, receive packet processing, and
75  * routines to turnaround packets back to the originator, and
76  * host table maintenance routines.
77  */
78 static VNET_DEFINE(int, icmplim) = 200;
79 #define	V_icmplim			VNET(icmplim)
80 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
81 	&VNET_NAME(icmplim), 0,
82 	"Maximum number of ICMP responses per second");
83 
84 static VNET_DEFINE(int, icmplim_output) = 1;
85 #define	V_icmplim_output		VNET(icmplim_output)
86 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
87 	&VNET_NAME(icmplim_output), 0,
88 	"Enable logging of ICMP response rate limiting");
89 
90 #ifdef INET
91 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
92 VNET_PCPUSTAT_SYSINIT(icmpstat);
93 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
94     icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
95 
96 #ifdef VIMAGE
97 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
98 #endif /* VIMAGE */
99 
100 static VNET_DEFINE(int, icmpmaskrepl) = 0;
101 #define	V_icmpmaskrepl			VNET(icmpmaskrepl)
102 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
103 	&VNET_NAME(icmpmaskrepl), 0,
104 	"Reply to ICMP Address Mask Request packets.");
105 
106 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
107 #define	V_icmpmaskfake			VNET(icmpmaskfake)
108 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
109 	&VNET_NAME(icmpmaskfake), 0,
110 	"Fake reply to ICMP Address Mask Request packets.");
111 
112 VNET_DEFINE(int, drop_redirect) = 0;
113 #define	V_drop_redirect			VNET(drop_redirect)
114 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
115 	&VNET_NAME(drop_redirect), 0, "Ignore ICMP redirects");
116 
117 static VNET_DEFINE(int, log_redirect) = 0;
118 #define	V_log_redirect			VNET(log_redirect)
119 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
120 	&VNET_NAME(log_redirect), 0,
121 	"Log ICMP redirects to the console");
122 
123 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
124 #define	V_reply_src			VNET(reply_src)
125 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
126 	&VNET_NAME(reply_src), IFNAMSIZ,
127 	"icmp reply source for non-local packets.");
128 
129 static VNET_DEFINE(int, icmp_rfi) = 0;
130 #define	V_icmp_rfi			VNET(icmp_rfi)
131 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
132 	&VNET_NAME(icmp_rfi), 0,
133 	"ICMP reply from incoming interface for non-local packets");
134 
135 static VNET_DEFINE(int, icmp_quotelen) = 8;
136 #define	V_icmp_quotelen			VNET(icmp_quotelen)
137 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
138 	&VNET_NAME(icmp_quotelen), 0,
139 	"Number of bytes from original packet to quote in ICMP reply");
140 
141 /*
142  * ICMP broadcast echo sysctl
143  */
144 static VNET_DEFINE(int, icmpbmcastecho) = 0;
145 #define	V_icmpbmcastecho		VNET(icmpbmcastecho)
146 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
147 	&VNET_NAME(icmpbmcastecho), 0,
148 	"");
149 
150 static VNET_DEFINE(int, icmptstamprepl) = 1;
151 #define	V_icmptstamprepl		VNET(icmptstamprepl)
152 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_RW,
153 	&VNET_NAME(icmptstamprepl), 0, "Respond to ICMP Timestamp packets");
154 
155 #ifdef ICMPPRINTFS
156 int	icmpprintfs = 0;
157 #endif
158 
159 static void	icmp_reflect(struct mbuf *);
160 static void	icmp_send(struct mbuf *, struct mbuf *);
161 
162 extern	struct protosw inetsw[];
163 
164 /*
165  * Kernel module interface for updating icmpstat.  The argument is an index
166  * into icmpstat treated as an array of u_long.  While this encodes the
167  * general layout of icmpstat into the caller, it doesn't encode its
168  * location, so that future changes to add, for example, per-CPU stats
169  * support won't cause binary compatibility problems for kernel modules.
170  */
171 void
172 kmod_icmpstat_inc(int statnum)
173 {
174 
175 	counter_u64_add(VNET(icmpstat)[statnum], 1);
176 }
177 
178 /*
179  * Generate an error packet of type error
180  * in response to bad packet ip.
181  */
182 void
183 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
184 {
185 	register struct ip *oip = mtod(n, struct ip *), *nip;
186 	register unsigned oiphlen = oip->ip_hl << 2;
187 	register struct icmp *icp;
188 	register struct mbuf *m;
189 	unsigned icmplen, icmpelen, nlen;
190 
191 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
192 #ifdef ICMPPRINTFS
193 	if (icmpprintfs)
194 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
195 #endif
196 	if (type != ICMP_REDIRECT)
197 		ICMPSTAT_INC(icps_error);
198 	/*
199 	 * Don't send error:
200 	 *  if the original packet was encrypted.
201 	 *  if not the first fragment of message.
202 	 *  in response to a multicast or broadcast packet.
203 	 *  if the old packet protocol was an ICMP error message.
204 	 */
205 	if (n->m_flags & M_DECRYPTED)
206 		goto freeit;
207 	if (oip->ip_off & htons(~(IP_MF|IP_DF)))
208 		goto freeit;
209 	if (n->m_flags & (M_BCAST|M_MCAST))
210 		goto freeit;
211 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
212 	  n->m_len >= oiphlen + ICMP_MINLEN &&
213 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
214 		ICMPSTAT_INC(icps_oldicmp);
215 		goto freeit;
216 	}
217 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
218 	if (oiphlen + 8 > n->m_len)
219 		goto freeit;
220 	/*
221 	 * Calculate length to quote from original packet and
222 	 * prevent the ICMP mbuf from overflowing.
223 	 * Unfortunatly this is non-trivial since ip_forward()
224 	 * sends us truncated packets.
225 	 */
226 	nlen = m_length(n, NULL);
227 	if (oip->ip_p == IPPROTO_TCP) {
228 		struct tcphdr *th;
229 		int tcphlen;
230 
231 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
232 		    n->m_next == NULL)
233 			goto stdreply;
234 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
235 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
236 			goto freeit;
237 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
238 		tcphlen = th->th_off << 2;
239 		if (tcphlen < sizeof(struct tcphdr))
240 			goto freeit;
241 		if (ntohs(oip->ip_len) < oiphlen + tcphlen)
242 			goto freeit;
243 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
244 			goto stdreply;
245 		if (n->m_len < oiphlen + tcphlen &&
246 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
247 			goto freeit;
248 		icmpelen = max(tcphlen, min(V_icmp_quotelen,
249 		    ntohs(oip->ip_len) - oiphlen));
250 	} else
251 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
252 
253 	icmplen = min(oiphlen + icmpelen, nlen);
254 	if (icmplen < sizeof(struct ip))
255 		goto freeit;
256 
257 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
258 		m = m_gethdr(M_NOWAIT, MT_DATA);
259 	else
260 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
261 	if (m == NULL)
262 		goto freeit;
263 #ifdef MAC
264 	mac_netinet_icmp_reply(n, m);
265 #endif
266 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
267 	m_align(m, ICMP_MINLEN + icmplen);
268 	m->m_len = ICMP_MINLEN + icmplen;
269 
270 	/* XXX MRT  make the outgoing packet use the same FIB
271 	 * that was associated with the incoming packet
272 	 */
273 	M_SETFIB(m, M_GETFIB(n));
274 	icp = mtod(m, struct icmp *);
275 	ICMPSTAT_INC(icps_outhist[type]);
276 	icp->icmp_type = type;
277 	if (type == ICMP_REDIRECT)
278 		icp->icmp_gwaddr.s_addr = dest;
279 	else {
280 		icp->icmp_void = 0;
281 		/*
282 		 * The following assignments assume an overlay with the
283 		 * just zeroed icmp_void field.
284 		 */
285 		if (type == ICMP_PARAMPROB) {
286 			icp->icmp_pptr = code;
287 			code = 0;
288 		} else if (type == ICMP_UNREACH &&
289 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
290 			icp->icmp_nextmtu = htons(mtu);
291 		}
292 	}
293 	icp->icmp_code = code;
294 
295 	/*
296 	 * Copy the quotation into ICMP message and
297 	 * convert quoted IP header back to network representation.
298 	 */
299 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
300 	nip = &icp->icmp_ip;
301 
302 	/*
303 	 * Set up ICMP message mbuf and copy old IP header (without options
304 	 * in front of ICMP message.
305 	 * If the original mbuf was meant to bypass the firewall, the error
306 	 * reply should bypass as well.
307 	 */
308 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
309 	m->m_data -= sizeof(struct ip);
310 	m->m_len += sizeof(struct ip);
311 	m->m_pkthdr.len = m->m_len;
312 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
313 	nip = mtod(m, struct ip *);
314 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
315 	nip->ip_len = htons(m->m_len);
316 	nip->ip_v = IPVERSION;
317 	nip->ip_hl = 5;
318 	nip->ip_p = IPPROTO_ICMP;
319 	nip->ip_tos = 0;
320 	nip->ip_off = 0;
321 	icmp_reflect(m);
322 
323 freeit:
324 	m_freem(n);
325 }
326 
327 /*
328  * Process a received ICMP message.
329  */
330 int
331 icmp_input(struct mbuf **mp, int *offp, int proto)
332 {
333 	struct icmp *icp;
334 	struct in_ifaddr *ia;
335 	struct mbuf *m = *mp;
336 	struct ip *ip = mtod(m, struct ip *);
337 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
338 	int hlen = *offp;
339 	int icmplen = ntohs(ip->ip_len) - *offp;
340 	int i, code;
341 	void (*ctlfunc)(int, struct sockaddr *, void *);
342 	int fibnum;
343 
344 	*mp = NULL;
345 
346 	/*
347 	 * Locate icmp structure in mbuf, and check
348 	 * that not corrupted and of at least minimum length.
349 	 */
350 #ifdef ICMPPRINTFS
351 	if (icmpprintfs) {
352 		char buf[4 * sizeof "123"];
353 		strcpy(buf, inet_ntoa(ip->ip_src));
354 		printf("icmp_input from %s to %s, len %d\n",
355 		       buf, inet_ntoa(ip->ip_dst), icmplen);
356 	}
357 #endif
358 	if (icmplen < ICMP_MINLEN) {
359 		ICMPSTAT_INC(icps_tooshort);
360 		goto freeit;
361 	}
362 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
363 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
364 		ICMPSTAT_INC(icps_tooshort);
365 		return (IPPROTO_DONE);
366 	}
367 	ip = mtod(m, struct ip *);
368 	m->m_len -= hlen;
369 	m->m_data += hlen;
370 	icp = mtod(m, struct icmp *);
371 	if (in_cksum(m, icmplen)) {
372 		ICMPSTAT_INC(icps_checksum);
373 		goto freeit;
374 	}
375 	m->m_len += hlen;
376 	m->m_data -= hlen;
377 
378 #ifdef ICMPPRINTFS
379 	if (icmpprintfs)
380 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
381 		    icp->icmp_code);
382 #endif
383 
384 	/*
385 	 * Message type specific processing.
386 	 */
387 	if (icp->icmp_type > ICMP_MAXTYPE)
388 		goto raw;
389 
390 	/* Initialize */
391 	bzero(&icmpsrc, sizeof(icmpsrc));
392 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
393 	icmpsrc.sin_family = AF_INET;
394 	bzero(&icmpdst, sizeof(icmpdst));
395 	icmpdst.sin_len = sizeof(struct sockaddr_in);
396 	icmpdst.sin_family = AF_INET;
397 	bzero(&icmpgw, sizeof(icmpgw));
398 	icmpgw.sin_len = sizeof(struct sockaddr_in);
399 	icmpgw.sin_family = AF_INET;
400 
401 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
402 	code = icp->icmp_code;
403 	switch (icp->icmp_type) {
404 
405 	case ICMP_UNREACH:
406 		switch (code) {
407 			case ICMP_UNREACH_NET:
408 			case ICMP_UNREACH_HOST:
409 			case ICMP_UNREACH_SRCFAIL:
410 			case ICMP_UNREACH_NET_UNKNOWN:
411 			case ICMP_UNREACH_HOST_UNKNOWN:
412 			case ICMP_UNREACH_ISOLATED:
413 			case ICMP_UNREACH_TOSNET:
414 			case ICMP_UNREACH_TOSHOST:
415 			case ICMP_UNREACH_HOST_PRECEDENCE:
416 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
417 				code = PRC_UNREACH_NET;
418 				break;
419 
420 			case ICMP_UNREACH_NEEDFRAG:
421 				code = PRC_MSGSIZE;
422 				break;
423 
424 			/*
425 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
426 			 * Treat subcodes 2,3 as immediate RST
427 			 */
428 			case ICMP_UNREACH_PROTOCOL:
429 			case ICMP_UNREACH_PORT:
430 				code = PRC_UNREACH_PORT;
431 				break;
432 
433 			case ICMP_UNREACH_NET_PROHIB:
434 			case ICMP_UNREACH_HOST_PROHIB:
435 			case ICMP_UNREACH_FILTER_PROHIB:
436 				code = PRC_UNREACH_ADMIN_PROHIB;
437 				break;
438 
439 			default:
440 				goto badcode;
441 		}
442 		goto deliver;
443 
444 	case ICMP_TIMXCEED:
445 		if (code > 1)
446 			goto badcode;
447 		code += PRC_TIMXCEED_INTRANS;
448 		goto deliver;
449 
450 	case ICMP_PARAMPROB:
451 		if (code > 1)
452 			goto badcode;
453 		code = PRC_PARAMPROB;
454 	deliver:
455 		/*
456 		 * Problem with datagram; advise higher level routines.
457 		 */
458 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
459 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
460 			ICMPSTAT_INC(icps_badlen);
461 			goto freeit;
462 		}
463 		/* Discard ICMP's in response to multicast packets */
464 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
465 			goto badcode;
466 #ifdef ICMPPRINTFS
467 		if (icmpprintfs)
468 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
469 #endif
470 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
471 		/*
472 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
473 		 * notification to TCP layer.
474 		 */
475 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
476 		if (ctlfunc)
477 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
478 				   (void *)&icp->icmp_ip);
479 		break;
480 
481 	badcode:
482 		ICMPSTAT_INC(icps_badcode);
483 		break;
484 
485 	case ICMP_ECHO:
486 		if (!V_icmpbmcastecho
487 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
488 			ICMPSTAT_INC(icps_bmcastecho);
489 			break;
490 		}
491 		icp->icmp_type = ICMP_ECHOREPLY;
492 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
493 			goto freeit;
494 		else
495 			goto reflect;
496 
497 	case ICMP_TSTAMP:
498 		if (V_icmptstamprepl == 0)
499 			break;
500 		if (!V_icmpbmcastecho
501 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
502 			ICMPSTAT_INC(icps_bmcasttstamp);
503 			break;
504 		}
505 		if (icmplen < ICMP_TSLEN) {
506 			ICMPSTAT_INC(icps_badlen);
507 			break;
508 		}
509 		icp->icmp_type = ICMP_TSTAMPREPLY;
510 		icp->icmp_rtime = iptime();
511 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
512 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
513 			goto freeit;
514 		else
515 			goto reflect;
516 
517 	case ICMP_MASKREQ:
518 		if (V_icmpmaskrepl == 0)
519 			break;
520 		/*
521 		 * We are not able to respond with all ones broadcast
522 		 * unless we receive it over a point-to-point interface.
523 		 */
524 		if (icmplen < ICMP_MASKLEN)
525 			break;
526 		switch (ip->ip_dst.s_addr) {
527 
528 		case INADDR_BROADCAST:
529 		case INADDR_ANY:
530 			icmpdst.sin_addr = ip->ip_src;
531 			break;
532 
533 		default:
534 			icmpdst.sin_addr = ip->ip_dst;
535 		}
536 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
537 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
538 		if (ia == NULL)
539 			break;
540 		if (ia->ia_ifp == NULL) {
541 			ifa_free(&ia->ia_ifa);
542 			break;
543 		}
544 		icp->icmp_type = ICMP_MASKREPLY;
545 		if (V_icmpmaskfake == 0)
546 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
547 		else
548 			icp->icmp_mask = V_icmpmaskfake;
549 		if (ip->ip_src.s_addr == 0) {
550 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
551 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
552 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
553 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
554 		}
555 		ifa_free(&ia->ia_ifa);
556 reflect:
557 		ICMPSTAT_INC(icps_reflect);
558 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
559 		icmp_reflect(m);
560 		return (IPPROTO_DONE);
561 
562 	case ICMP_REDIRECT:
563 		if (V_log_redirect) {
564 			u_long src, dst, gw;
565 
566 			src = ntohl(ip->ip_src.s_addr);
567 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
568 			gw = ntohl(icp->icmp_gwaddr.s_addr);
569 			printf("icmp redirect from %d.%d.%d.%d: "
570 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
571 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
572 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
573 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
574 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
575 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
576 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
577 		}
578 		/*
579 		 * RFC1812 says we must ignore ICMP redirects if we
580 		 * are acting as router.
581 		 */
582 		if (V_drop_redirect || V_ipforwarding)
583 			break;
584 		if (code > 3)
585 			goto badcode;
586 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
587 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
588 			ICMPSTAT_INC(icps_badlen);
589 			break;
590 		}
591 		/*
592 		 * Short circuit routing redirects to force
593 		 * immediate change in the kernel's routing
594 		 * tables.  The message is also handed to anyone
595 		 * listening on a raw socket (e.g. the routing
596 		 * daemon for use in updating its tables).
597 		 */
598 		icmpgw.sin_addr = ip->ip_src;
599 		icmpdst.sin_addr = icp->icmp_gwaddr;
600 #ifdef	ICMPPRINTFS
601 		if (icmpprintfs) {
602 			char buf[4 * sizeof "123"];
603 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
604 
605 			printf("redirect dst %s to %s\n",
606 			       buf, inet_ntoa(icp->icmp_gwaddr));
607 		}
608 #endif
609 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
610 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
611 			in_rtredirect((struct sockaddr *)&icmpsrc,
612 			  (struct sockaddr *)&icmpdst,
613 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
614 			  (struct sockaddr *)&icmpgw, fibnum);
615 		}
616 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
617 		break;
618 
619 	/*
620 	 * No kernel processing for the following;
621 	 * just fall through to send to raw listener.
622 	 */
623 	case ICMP_ECHOREPLY:
624 	case ICMP_ROUTERADVERT:
625 	case ICMP_ROUTERSOLICIT:
626 	case ICMP_TSTAMPREPLY:
627 	case ICMP_IREQREPLY:
628 	case ICMP_MASKREPLY:
629 	case ICMP_SOURCEQUENCH:
630 	default:
631 		break;
632 	}
633 
634 raw:
635 	*mp = m;
636 	rip_input(mp, offp, proto);
637 	return (IPPROTO_DONE);
638 
639 freeit:
640 	m_freem(m);
641 	return (IPPROTO_DONE);
642 }
643 
644 /*
645  * Reflect the ip packet back to the source
646  */
647 static void
648 icmp_reflect(struct mbuf *m)
649 {
650 	struct ip *ip = mtod(m, struct ip *);
651 	struct ifaddr *ifa;
652 	struct ifnet *ifp;
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 		ICMPSTAT_INC(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 	IN_IFADDR_RLOCK();
676 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
677 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
678 			t = IA_SIN(ia)->sin_addr;
679 			IN_IFADDR_RUNLOCK();
680 			goto match;
681 		}
682 	}
683 	IN_IFADDR_RUNLOCK();
684 
685 	/*
686 	 * If the incoming packet was addressed to one of our broadcast
687 	 * addresses, use the first non-broadcast address which corresponds
688 	 * to the incoming interface.
689 	 */
690 	ifp = m->m_pkthdr.rcvif;
691 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
692 		IF_ADDR_RLOCK(ifp);
693 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
694 			if (ifa->ifa_addr->sa_family != AF_INET)
695 				continue;
696 			ia = ifatoia(ifa);
697 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
698 			    t.s_addr) {
699 				t = IA_SIN(ia)->sin_addr;
700 				IF_ADDR_RUNLOCK(ifp);
701 				goto match;
702 			}
703 		}
704 		IF_ADDR_RUNLOCK(ifp);
705 	}
706 	/*
707 	 * If the packet was transiting through us, use the address of
708 	 * the interface the packet came through in.  If that interface
709 	 * doesn't have a suitable IP address, the normal selection
710 	 * criteria apply.
711 	 */
712 	if (V_icmp_rfi && ifp != NULL) {
713 		IF_ADDR_RLOCK(ifp);
714 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
715 			if (ifa->ifa_addr->sa_family != AF_INET)
716 				continue;
717 			ia = ifatoia(ifa);
718 			t = IA_SIN(ia)->sin_addr;
719 			IF_ADDR_RUNLOCK(ifp);
720 			goto match;
721 		}
722 		IF_ADDR_RUNLOCK(ifp);
723 	}
724 	/*
725 	 * If the incoming packet was not addressed directly to us, use
726 	 * designated interface for icmp replies specified by sysctl
727 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
728 	 * with normal source selection.
729 	 */
730 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
731 		IF_ADDR_RLOCK(ifp);
732 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
733 			if (ifa->ifa_addr->sa_family != AF_INET)
734 				continue;
735 			ia = ifatoia(ifa);
736 			t = IA_SIN(ia)->sin_addr;
737 			IF_ADDR_RUNLOCK(ifp);
738 			goto match;
739 		}
740 		IF_ADDR_RUNLOCK(ifp);
741 	}
742 	/*
743 	 * If the packet was transiting through us, use the address of
744 	 * the interface that is the closest to the packet source.
745 	 * When we don't have a route back to the packet source, stop here
746 	 * and drop the packet.
747 	 */
748 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
749 	if (ia == NULL) {
750 		m_freem(m);
751 		ICMPSTAT_INC(icps_noroute);
752 		goto done;
753 	}
754 	t = IA_SIN(ia)->sin_addr;
755 	ifa_free(&ia->ia_ifa);
756 match:
757 #ifdef MAC
758 	mac_netinet_icmp_replyinplace(m);
759 #endif
760 	ip->ip_src = t;
761 	ip->ip_ttl = V_ip_defttl;
762 
763 	if (optlen > 0) {
764 		register u_char *cp;
765 		int opt, cnt;
766 		u_int len;
767 
768 		/*
769 		 * Retrieve any source routing from the incoming packet;
770 		 * add on any record-route or timestamp options.
771 		 */
772 		cp = (u_char *) (ip + 1);
773 		if ((opts = ip_srcroute(m)) == 0 &&
774 		    (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
775 			opts->m_len = sizeof(struct in_addr);
776 			mtod(opts, struct in_addr *)->s_addr = 0;
777 		}
778 		if (opts) {
779 #ifdef ICMPPRINTFS
780 		    if (icmpprintfs)
781 			    printf("icmp_reflect optlen %d rt %d => ",
782 				optlen, opts->m_len);
783 #endif
784 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
785 			    opt = cp[IPOPT_OPTVAL];
786 			    if (opt == IPOPT_EOL)
787 				    break;
788 			    if (opt == IPOPT_NOP)
789 				    len = 1;
790 			    else {
791 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
792 					    break;
793 				    len = cp[IPOPT_OLEN];
794 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
795 				        len > cnt)
796 					    break;
797 			    }
798 			    /*
799 			     * Should check for overflow, but it "can't happen"
800 			     */
801 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
802 				opt == IPOPT_SECURITY) {
803 				    bcopy((caddr_t)cp,
804 					mtod(opts, caddr_t) + opts->m_len, len);
805 				    opts->m_len += len;
806 			    }
807 		    }
808 		    /* Terminate & pad, if necessary */
809 		    cnt = opts->m_len % 4;
810 		    if (cnt) {
811 			    for (; cnt < 4; cnt++) {
812 				    *(mtod(opts, caddr_t) + opts->m_len) =
813 					IPOPT_EOL;
814 				    opts->m_len++;
815 			    }
816 		    }
817 #ifdef ICMPPRINTFS
818 		    if (icmpprintfs)
819 			    printf("%d\n", opts->m_len);
820 #endif
821 		}
822 		ip_stripoptions(m);
823 	}
824 	m_tag_delete_nonpersistent(m);
825 	m->m_flags &= ~(M_BCAST|M_MCAST);
826 	icmp_send(m, opts);
827 done:
828 	if (opts)
829 		(void)m_free(opts);
830 }
831 
832 /*
833  * Send an icmp packet back to the ip level,
834  * after supplying a checksum.
835  */
836 static void
837 icmp_send(struct mbuf *m, struct mbuf *opts)
838 {
839 	register struct ip *ip = mtod(m, struct ip *);
840 	register int hlen;
841 	register struct icmp *icp;
842 
843 	hlen = ip->ip_hl << 2;
844 	m->m_data += hlen;
845 	m->m_len -= hlen;
846 	icp = mtod(m, struct icmp *);
847 	icp->icmp_cksum = 0;
848 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
849 	m->m_data -= hlen;
850 	m->m_len += hlen;
851 	m->m_pkthdr.rcvif = (struct ifnet *)0;
852 #ifdef ICMPPRINTFS
853 	if (icmpprintfs) {
854 		char buf[4 * sizeof "123"];
855 		strcpy(buf, inet_ntoa(ip->ip_dst));
856 		printf("icmp_send dst %s src %s\n",
857 		       buf, inet_ntoa(ip->ip_src));
858 	}
859 #endif
860 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
861 }
862 
863 /*
864  * Return milliseconds since 00:00 UTC in network format.
865  */
866 uint32_t
867 iptime(void)
868 {
869 	struct timeval atv;
870 	u_long t;
871 
872 	getmicrotime(&atv);
873 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
874 	return (htonl(t));
875 }
876 
877 /*
878  * Return the next larger or smaller MTU plateau (table from RFC 1191)
879  * given current value MTU.  If DIR is less than zero, a larger plateau
880  * is returned; otherwise, a smaller value is returned.
881  */
882 int
883 ip_next_mtu(int mtu, int dir)
884 {
885 	static int mtutab[] = {
886 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
887 		296, 68, 0
888 	};
889 	int i, size;
890 
891 	size = (sizeof mtutab) / (sizeof mtutab[0]);
892 	if (dir >= 0) {
893 		for (i = 0; i < size; i++)
894 			if (mtu > mtutab[i])
895 				return mtutab[i];
896 	} else {
897 		for (i = size - 1; i >= 0; i--)
898 			if (mtu < mtutab[i])
899 				return mtutab[i];
900 		if (mtu == mtutab[0])
901 			return mtutab[0];
902 	}
903 	return 0;
904 }
905 #endif /* INET */
906 
907 
908 /*
909  * badport_bandlim() - check for ICMP bandwidth limit
910  *
911  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
912  *	hit our bandwidth limit and it is not ok.
913  *
914  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
915  *
916  *	For now we separate the TCP and UDP subsystems w/ different 'which'
917  *	values.  We may eventually remove this separation (and simplify the
918  *	code further).
919  *
920  *	Note that the printing of the error message is delayed so we can
921  *	properly print the icmp error rate that the system was trying to do
922  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
923  *	the 'final' error, but it doesn't make sense to solve the printing
924  *	delay with more complex code.
925  */
926 
927 int
928 badport_bandlim(int which)
929 {
930 
931 #define	N(a)	(sizeof (a) / sizeof (a[0]))
932 	static struct rate {
933 		const char	*type;
934 		struct timeval	lasttime;
935 		int		curpps;
936 	} rates[BANDLIM_MAX+1] = {
937 		{ "icmp unreach response" },
938 		{ "icmp ping response" },
939 		{ "icmp tstamp response" },
940 		{ "closed port RST response" },
941 		{ "open port RST response" },
942 		{ "icmp6 unreach response" },
943 		{ "sctp ootb response" }
944 	};
945 
946 	/*
947 	 * Return ok status if feature disabled or argument out of range.
948 	 */
949 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
950 		struct rate *r = &rates[which];
951 		int opps = r->curpps;
952 
953 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
954 			return -1;	/* discard packet */
955 		/*
956 		 * If we've dropped below the threshold after having
957 		 * rate-limited traffic print the message.  This preserves
958 		 * the previous behaviour at the expense of added complexity.
959 		 */
960 		if (V_icmplim_output && opps > V_icmplim)
961 			log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
962 				r->type, opps, V_icmplim);
963 	}
964 	return 0;			/* okay to send packet */
965 #undef N
966 }
967