xref: /freebsd/sys/netinet/ip_icmp.c (revision 46c1105fbb6fbff6d6ccd0a18571342eb992d637)
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/lock.h>
45 #include <sys/rmlock.h>
46 #include <sys/sysctl.h>
47 #include <sys/syslog.h>
48 
49 #include <net/if.h>
50 #include <net/if_var.h>
51 #include <net/if_types.h>
52 #include <net/route.h>
53 #include <net/vnet.h>
54 
55 #include <netinet/in.h>
56 #include <netinet/in_fib.h>
57 #include <netinet/in_pcb.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip.h>
61 #include <netinet/ip_icmp.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/ip_options.h>
64 #include <netinet/tcp.h>
65 #include <netinet/tcp_var.h>
66 #include <netinet/tcpip.h>
67 #include <netinet/icmp_var.h>
68 
69 #ifdef INET
70 
71 #include <machine/in_cksum.h>
72 
73 #include <security/mac/mac_framework.h>
74 #endif /* INET */
75 
76 /*
77  * ICMP routines: error generation, receive packet processing, and
78  * routines to turnaround packets back to the originator, and
79  * host table maintenance routines.
80  */
81 static VNET_DEFINE(int, icmplim) = 200;
82 #define	V_icmplim			VNET(icmplim)
83 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
84 	&VNET_NAME(icmplim), 0,
85 	"Maximum number of ICMP responses per second");
86 
87 static VNET_DEFINE(int, icmplim_output) = 1;
88 #define	V_icmplim_output		VNET(icmplim_output)
89 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
90 	&VNET_NAME(icmplim_output), 0,
91 	"Enable logging of ICMP response rate limiting");
92 
93 #ifdef INET
94 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
95 VNET_PCPUSTAT_SYSINIT(icmpstat);
96 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
97     icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
98 
99 #ifdef VIMAGE
100 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
101 #endif /* VIMAGE */
102 
103 static VNET_DEFINE(int, icmpmaskrepl) = 0;
104 #define	V_icmpmaskrepl			VNET(icmpmaskrepl)
105 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
106 	&VNET_NAME(icmpmaskrepl), 0,
107 	"Reply to ICMP Address Mask Request packets");
108 
109 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
110 #define	V_icmpmaskfake			VNET(icmpmaskfake)
111 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
112 	&VNET_NAME(icmpmaskfake), 0,
113 	"Fake reply to ICMP Address Mask Request packets");
114 
115 VNET_DEFINE(int, drop_redirect) = 0;
116 #define	V_drop_redirect			VNET(drop_redirect)
117 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
118 	&VNET_NAME(drop_redirect), 0,
119 	"Ignore ICMP redirects");
120 
121 static VNET_DEFINE(int, log_redirect) = 0;
122 #define	V_log_redirect			VNET(log_redirect)
123 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
124 	&VNET_NAME(log_redirect), 0,
125 	"Log ICMP redirects to the console");
126 
127 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
128 #define	V_reply_src			VNET(reply_src)
129 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
130 	&VNET_NAME(reply_src), IFNAMSIZ,
131 	"ICMP reply source for non-local packets");
132 
133 static VNET_DEFINE(int, icmp_rfi) = 0;
134 #define	V_icmp_rfi			VNET(icmp_rfi)
135 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
136 	&VNET_NAME(icmp_rfi), 0,
137 	"ICMP reply from incoming interface for non-local packets");
138 
139 static VNET_DEFINE(int, icmp_quotelen) = 8;
140 #define	V_icmp_quotelen			VNET(icmp_quotelen)
141 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
142 	&VNET_NAME(icmp_quotelen), 0,
143 	"Number of bytes from original packet to quote in ICMP reply");
144 
145 static VNET_DEFINE(int, icmpbmcastecho) = 0;
146 #define	V_icmpbmcastecho		VNET(icmpbmcastecho)
147 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
148 	&VNET_NAME(icmpbmcastecho), 0,
149 	"Reply to multicast ICMP Echo Request and Timestamp packets");
150 
151 static VNET_DEFINE(int, icmptstamprepl) = 1;
152 #define	V_icmptstamprepl		VNET(icmptstamprepl)
153 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_RW,
154 	&VNET_NAME(icmptstamprepl), 0,
155 	"Respond to ICMP Timestamp packets");
156 
157 #ifdef ICMPPRINTFS
158 int	icmpprintfs = 0;
159 #endif
160 
161 static void	icmp_reflect(struct mbuf *);
162 static void	icmp_send(struct mbuf *, struct mbuf *);
163 
164 extern	struct protosw inetsw[];
165 
166 /*
167  * Kernel module interface for updating icmpstat.  The argument is an index
168  * into icmpstat treated as an array of u_long.  While this encodes the
169  * general layout of icmpstat into the caller, it doesn't encode its
170  * location, so that future changes to add, for example, per-CPU stats
171  * support won't cause binary compatibility problems for kernel modules.
172  */
173 void
174 kmod_icmpstat_inc(int statnum)
175 {
176 
177 	counter_u64_add(VNET(icmpstat)[statnum], 1);
178 }
179 
180 /*
181  * Generate an error packet of type error
182  * in response to bad packet ip.
183  */
184 void
185 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
186 {
187 	register struct ip *oip = mtod(n, struct ip *), *nip;
188 	register unsigned oiphlen = oip->ip_hl << 2;
189 	register struct icmp *icp;
190 	register struct mbuf *m;
191 	unsigned icmplen, icmpelen, nlen;
192 
193 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
194 #ifdef ICMPPRINTFS
195 	if (icmpprintfs)
196 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
197 #endif
198 	if (type != ICMP_REDIRECT)
199 		ICMPSTAT_INC(icps_error);
200 	/*
201 	 * Don't send error:
202 	 *  if the original packet was encrypted.
203 	 *  if not the first fragment of message.
204 	 *  in response to a multicast or broadcast packet.
205 	 *  if the old packet protocol was an ICMP error message.
206 	 */
207 	if (n->m_flags & M_DECRYPTED)
208 		goto freeit;
209 	if (oip->ip_off & htons(~(IP_MF|IP_DF)))
210 		goto freeit;
211 	if (n->m_flags & (M_BCAST|M_MCAST))
212 		goto freeit;
213 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
214 	  n->m_len >= oiphlen + ICMP_MINLEN &&
215 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
216 		ICMPSTAT_INC(icps_oldicmp);
217 		goto freeit;
218 	}
219 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
220 	if (oiphlen + 8 > n->m_len)
221 		goto freeit;
222 	/*
223 	 * Calculate length to quote from original packet and
224 	 * prevent the ICMP mbuf from overflowing.
225 	 * Unfortunately this is non-trivial since ip_forward()
226 	 * sends us truncated packets.
227 	 */
228 	nlen = m_length(n, NULL);
229 	if (oip->ip_p == IPPROTO_TCP) {
230 		struct tcphdr *th;
231 		int tcphlen;
232 
233 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
234 		    n->m_next == NULL)
235 			goto stdreply;
236 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
237 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
238 			goto freeit;
239 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
240 		tcphlen = th->th_off << 2;
241 		if (tcphlen < sizeof(struct tcphdr))
242 			goto freeit;
243 		if (ntohs(oip->ip_len) < oiphlen + tcphlen)
244 			goto freeit;
245 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
246 			goto stdreply;
247 		if (n->m_len < oiphlen + tcphlen &&
248 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
249 			goto freeit;
250 		icmpelen = max(tcphlen, min(V_icmp_quotelen,
251 		    ntohs(oip->ip_len) - oiphlen));
252 	} else
253 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
254 
255 	icmplen = min(oiphlen + icmpelen, nlen);
256 	if (icmplen < sizeof(struct ip))
257 		goto freeit;
258 
259 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
260 		m = m_gethdr(M_NOWAIT, MT_DATA);
261 	else
262 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
263 	if (m == NULL)
264 		goto freeit;
265 #ifdef MAC
266 	mac_netinet_icmp_reply(n, m);
267 #endif
268 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
269 	m_align(m, ICMP_MINLEN + icmplen);
270 	m->m_len = ICMP_MINLEN + icmplen;
271 
272 	/* XXX MRT  make the outgoing packet use the same FIB
273 	 * that was associated with the incoming packet
274 	 */
275 	M_SETFIB(m, M_GETFIB(n));
276 	icp = mtod(m, struct icmp *);
277 	ICMPSTAT_INC(icps_outhist[type]);
278 	icp->icmp_type = type;
279 	if (type == ICMP_REDIRECT)
280 		icp->icmp_gwaddr.s_addr = dest;
281 	else {
282 		icp->icmp_void = 0;
283 		/*
284 		 * The following assignments assume an overlay with the
285 		 * just zeroed icmp_void field.
286 		 */
287 		if (type == ICMP_PARAMPROB) {
288 			icp->icmp_pptr = code;
289 			code = 0;
290 		} else if (type == ICMP_UNREACH &&
291 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
292 			icp->icmp_nextmtu = htons(mtu);
293 		}
294 	}
295 	icp->icmp_code = code;
296 
297 	/*
298 	 * Copy the quotation into ICMP message and
299 	 * convert quoted IP header back to network representation.
300 	 */
301 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
302 	nip = &icp->icmp_ip;
303 
304 	/*
305 	 * Set up ICMP message mbuf and copy old IP header (without options
306 	 * in front of ICMP message.
307 	 * If the original mbuf was meant to bypass the firewall, the error
308 	 * reply should bypass as well.
309 	 */
310 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
311 	m->m_data -= sizeof(struct ip);
312 	m->m_len += sizeof(struct ip);
313 	m->m_pkthdr.len = m->m_len;
314 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
315 	nip = mtod(m, struct ip *);
316 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
317 	nip->ip_len = htons(m->m_len);
318 	nip->ip_v = IPVERSION;
319 	nip->ip_hl = 5;
320 	nip->ip_p = IPPROTO_ICMP;
321 	nip->ip_tos = 0;
322 	nip->ip_off = 0;
323 	icmp_reflect(m);
324 
325 freeit:
326 	m_freem(n);
327 }
328 
329 /*
330  * Process a received ICMP message.
331  */
332 int
333 icmp_input(struct mbuf **mp, int *offp, int proto)
334 {
335 	struct icmp *icp;
336 	struct in_ifaddr *ia;
337 	struct mbuf *m = *mp;
338 	struct ip *ip = mtod(m, struct ip *);
339 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
340 	int hlen = *offp;
341 	int icmplen = ntohs(ip->ip_len) - *offp;
342 	int i, code;
343 	void (*ctlfunc)(int, struct sockaddr *, void *);
344 	int fibnum;
345 
346 	*mp = NULL;
347 
348 	/*
349 	 * Locate icmp structure in mbuf, and check
350 	 * that not corrupted and of at least minimum length.
351 	 */
352 #ifdef ICMPPRINTFS
353 	if (icmpprintfs) {
354 		char buf[4 * sizeof "123"];
355 		strcpy(buf, inet_ntoa(ip->ip_src));
356 		printf("icmp_input from %s to %s, len %d\n",
357 		       buf, inet_ntoa(ip->ip_dst), icmplen);
358 	}
359 #endif
360 	if (icmplen < ICMP_MINLEN) {
361 		ICMPSTAT_INC(icps_tooshort);
362 		goto freeit;
363 	}
364 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
365 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
366 		ICMPSTAT_INC(icps_tooshort);
367 		return (IPPROTO_DONE);
368 	}
369 	ip = mtod(m, struct ip *);
370 	m->m_len -= hlen;
371 	m->m_data += hlen;
372 	icp = mtod(m, struct icmp *);
373 	if (in_cksum(m, icmplen)) {
374 		ICMPSTAT_INC(icps_checksum);
375 		goto freeit;
376 	}
377 	m->m_len += hlen;
378 	m->m_data -= hlen;
379 
380 #ifdef ICMPPRINTFS
381 	if (icmpprintfs)
382 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
383 		    icp->icmp_code);
384 #endif
385 
386 	/*
387 	 * Message type specific processing.
388 	 */
389 	if (icp->icmp_type > ICMP_MAXTYPE)
390 		goto raw;
391 
392 	/* Initialize */
393 	bzero(&icmpsrc, sizeof(icmpsrc));
394 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
395 	icmpsrc.sin_family = AF_INET;
396 	bzero(&icmpdst, sizeof(icmpdst));
397 	icmpdst.sin_len = sizeof(struct sockaddr_in);
398 	icmpdst.sin_family = AF_INET;
399 	bzero(&icmpgw, sizeof(icmpgw));
400 	icmpgw.sin_len = sizeof(struct sockaddr_in);
401 	icmpgw.sin_family = AF_INET;
402 
403 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
404 	code = icp->icmp_code;
405 	switch (icp->icmp_type) {
406 
407 	case ICMP_UNREACH:
408 		switch (code) {
409 			case ICMP_UNREACH_NET:
410 			case ICMP_UNREACH_HOST:
411 			case ICMP_UNREACH_SRCFAIL:
412 			case ICMP_UNREACH_NET_UNKNOWN:
413 			case ICMP_UNREACH_HOST_UNKNOWN:
414 			case ICMP_UNREACH_ISOLATED:
415 			case ICMP_UNREACH_TOSNET:
416 			case ICMP_UNREACH_TOSHOST:
417 			case ICMP_UNREACH_HOST_PRECEDENCE:
418 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
419 				code = PRC_UNREACH_NET;
420 				break;
421 
422 			case ICMP_UNREACH_NEEDFRAG:
423 				code = PRC_MSGSIZE;
424 				break;
425 
426 			/*
427 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
428 			 * Treat subcodes 2,3 as immediate RST
429 			 */
430 			case ICMP_UNREACH_PROTOCOL:
431 			case ICMP_UNREACH_PORT:
432 				code = PRC_UNREACH_PORT;
433 				break;
434 
435 			case ICMP_UNREACH_NET_PROHIB:
436 			case ICMP_UNREACH_HOST_PROHIB:
437 			case ICMP_UNREACH_FILTER_PROHIB:
438 				code = PRC_UNREACH_ADMIN_PROHIB;
439 				break;
440 
441 			default:
442 				goto badcode;
443 		}
444 		goto deliver;
445 
446 	case ICMP_TIMXCEED:
447 		if (code > 1)
448 			goto badcode;
449 		code += PRC_TIMXCEED_INTRANS;
450 		goto deliver;
451 
452 	case ICMP_PARAMPROB:
453 		if (code > 1)
454 			goto badcode;
455 		code = PRC_PARAMPROB;
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 			ICMPSTAT_INC(icps_badlen);
463 			goto freeit;
464 		}
465 		/* Discard ICMP's in response to multicast packets */
466 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
467 			goto badcode;
468 #ifdef ICMPPRINTFS
469 		if (icmpprintfs)
470 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
471 #endif
472 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
473 		/*
474 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
475 		 * notification to TCP layer.
476 		 */
477 		i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
478 		ip_stripoptions(m);
479 		if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
480 			/* This should actually not happen */
481 			ICMPSTAT_INC(icps_tooshort);
482 			return (IPPROTO_DONE);
483 		}
484 		ip = mtod(m, struct ip *);
485 		icp = (struct icmp *)(ip + 1);
486 		/*
487 		 * The upper layer handler can rely on:
488 		 * - The outer IP header has no options.
489 		 * - The outer IP header, the ICMP header, the inner IP header,
490 		 *   and the first n bytes of the inner payload are contiguous.
491 		 *   n is at least 8, but might be larger based on
492 		 *   ICMP_ADVLENPREF. See its definition in ip_icmp.h.
493 		 */
494 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
495 		if (ctlfunc)
496 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
497 				   (void *)&icp->icmp_ip);
498 		break;
499 
500 	badcode:
501 		ICMPSTAT_INC(icps_badcode);
502 		break;
503 
504 	case ICMP_ECHO:
505 		if (!V_icmpbmcastecho
506 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
507 			ICMPSTAT_INC(icps_bmcastecho);
508 			break;
509 		}
510 		icp->icmp_type = ICMP_ECHOREPLY;
511 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
512 			goto freeit;
513 		else
514 			goto reflect;
515 
516 	case ICMP_TSTAMP:
517 		if (V_icmptstamprepl == 0)
518 			break;
519 		if (!V_icmpbmcastecho
520 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
521 			ICMPSTAT_INC(icps_bmcasttstamp);
522 			break;
523 		}
524 		if (icmplen < ICMP_TSLEN) {
525 			ICMPSTAT_INC(icps_badlen);
526 			break;
527 		}
528 		icp->icmp_type = ICMP_TSTAMPREPLY;
529 		icp->icmp_rtime = iptime();
530 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
531 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
532 			goto freeit;
533 		else
534 			goto reflect;
535 
536 	case ICMP_MASKREQ:
537 		if (V_icmpmaskrepl == 0)
538 			break;
539 		/*
540 		 * We are not able to respond with all ones broadcast
541 		 * unless we receive it over a point-to-point interface.
542 		 */
543 		if (icmplen < ICMP_MASKLEN)
544 			break;
545 		switch (ip->ip_dst.s_addr) {
546 
547 		case INADDR_BROADCAST:
548 		case INADDR_ANY:
549 			icmpdst.sin_addr = ip->ip_src;
550 			break;
551 
552 		default:
553 			icmpdst.sin_addr = ip->ip_dst;
554 		}
555 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
556 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
557 		if (ia == NULL)
558 			break;
559 		if (ia->ia_ifp == NULL) {
560 			ifa_free(&ia->ia_ifa);
561 			break;
562 		}
563 		icp->icmp_type = ICMP_MASKREPLY;
564 		if (V_icmpmaskfake == 0)
565 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
566 		else
567 			icp->icmp_mask = V_icmpmaskfake;
568 		if (ip->ip_src.s_addr == 0) {
569 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
570 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
571 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
572 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
573 		}
574 		ifa_free(&ia->ia_ifa);
575 reflect:
576 		ICMPSTAT_INC(icps_reflect);
577 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
578 		icmp_reflect(m);
579 		return (IPPROTO_DONE);
580 
581 	case ICMP_REDIRECT:
582 		if (V_log_redirect) {
583 			u_long src, dst, gw;
584 
585 			src = ntohl(ip->ip_src.s_addr);
586 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
587 			gw = ntohl(icp->icmp_gwaddr.s_addr);
588 			printf("icmp redirect from %d.%d.%d.%d: "
589 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
590 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
591 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
592 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
593 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
594 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
595 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
596 		}
597 		/*
598 		 * RFC1812 says we must ignore ICMP redirects if we
599 		 * are acting as router.
600 		 */
601 		if (V_drop_redirect || V_ipforwarding)
602 			break;
603 		if (code > 3)
604 			goto badcode;
605 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
606 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
607 			ICMPSTAT_INC(icps_badlen);
608 			break;
609 		}
610 		/*
611 		 * Short circuit routing redirects to force
612 		 * immediate change in the kernel's routing
613 		 * tables.  The message is also handed to anyone
614 		 * listening on a raw socket (e.g. the routing
615 		 * daemon for use in updating its tables).
616 		 */
617 		icmpgw.sin_addr = ip->ip_src;
618 		icmpdst.sin_addr = icp->icmp_gwaddr;
619 #ifdef	ICMPPRINTFS
620 		if (icmpprintfs) {
621 			char buf[4 * sizeof "123"];
622 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
623 
624 			printf("redirect dst %s to %s\n",
625 			       buf, inet_ntoa(icp->icmp_gwaddr));
626 		}
627 #endif
628 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
629 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
630 			in_rtredirect((struct sockaddr *)&icmpsrc,
631 			  (struct sockaddr *)&icmpdst,
632 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
633 			  (struct sockaddr *)&icmpgw, fibnum);
634 		}
635 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
636 		break;
637 
638 	/*
639 	 * No kernel processing for the following;
640 	 * just fall through to send to raw listener.
641 	 */
642 	case ICMP_ECHOREPLY:
643 	case ICMP_ROUTERADVERT:
644 	case ICMP_ROUTERSOLICIT:
645 	case ICMP_TSTAMPREPLY:
646 	case ICMP_IREQREPLY:
647 	case ICMP_MASKREPLY:
648 	case ICMP_SOURCEQUENCH:
649 	default:
650 		break;
651 	}
652 
653 raw:
654 	*mp = m;
655 	rip_input(mp, offp, proto);
656 	return (IPPROTO_DONE);
657 
658 freeit:
659 	m_freem(m);
660 	return (IPPROTO_DONE);
661 }
662 
663 /*
664  * Reflect the ip packet back to the source
665  */
666 static void
667 icmp_reflect(struct mbuf *m)
668 {
669 	struct rm_priotracker in_ifa_tracker;
670 	struct ip *ip = mtod(m, struct ip *);
671 	struct ifaddr *ifa;
672 	struct ifnet *ifp;
673 	struct in_ifaddr *ia;
674 	struct in_addr t;
675 	struct nhop4_extended nh_ext;
676 	struct mbuf *opts = NULL;
677 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
678 
679 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
680 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
681 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
682 		m_freem(m);	/* Bad return address */
683 		ICMPSTAT_INC(icps_badaddr);
684 		goto done;	/* Ip_output() will check for broadcast */
685 	}
686 
687 	t = ip->ip_dst;
688 	ip->ip_dst = ip->ip_src;
689 
690 	/*
691 	 * Source selection for ICMP replies:
692 	 *
693 	 * If the incoming packet was addressed directly to one of our
694 	 * own addresses, use dst as the src for the reply.
695 	 */
696 	IN_IFADDR_RLOCK(&in_ifa_tracker);
697 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
698 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
699 			t = IA_SIN(ia)->sin_addr;
700 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
701 			goto match;
702 		}
703 	}
704 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
705 
706 	/*
707 	 * If the incoming packet was addressed to one of our broadcast
708 	 * addresses, use the first non-broadcast address which corresponds
709 	 * to the incoming interface.
710 	 */
711 	ifp = m->m_pkthdr.rcvif;
712 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
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 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
719 			    t.s_addr) {
720 				t = IA_SIN(ia)->sin_addr;
721 				IF_ADDR_RUNLOCK(ifp);
722 				goto match;
723 			}
724 		}
725 		IF_ADDR_RUNLOCK(ifp);
726 	}
727 	/*
728 	 * If the packet was transiting through us, use the address of
729 	 * the interface the packet came through in.  If that interface
730 	 * doesn't have a suitable IP address, the normal selection
731 	 * criteria apply.
732 	 */
733 	if (V_icmp_rfi && ifp != NULL) {
734 		IF_ADDR_RLOCK(ifp);
735 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
736 			if (ifa->ifa_addr->sa_family != AF_INET)
737 				continue;
738 			ia = ifatoia(ifa);
739 			t = IA_SIN(ia)->sin_addr;
740 			IF_ADDR_RUNLOCK(ifp);
741 			goto match;
742 		}
743 		IF_ADDR_RUNLOCK(ifp);
744 	}
745 	/*
746 	 * If the incoming packet was not addressed directly to us, use
747 	 * designated interface for icmp replies specified by sysctl
748 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
749 	 * with normal source selection.
750 	 */
751 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
752 		IF_ADDR_RLOCK(ifp);
753 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
754 			if (ifa->ifa_addr->sa_family != AF_INET)
755 				continue;
756 			ia = ifatoia(ifa);
757 			t = IA_SIN(ia)->sin_addr;
758 			IF_ADDR_RUNLOCK(ifp);
759 			goto match;
760 		}
761 		IF_ADDR_RUNLOCK(ifp);
762 	}
763 	/*
764 	 * If the packet was transiting through us, use the address of
765 	 * the interface that is the closest to the packet source.
766 	 * When we don't have a route back to the packet source, stop here
767 	 * and drop the packet.
768 	 */
769 	if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) {
770 		m_freem(m);
771 		ICMPSTAT_INC(icps_noroute);
772 		goto done;
773 	}
774 	t = nh_ext.nh_src;
775 match:
776 #ifdef MAC
777 	mac_netinet_icmp_replyinplace(m);
778 #endif
779 	ip->ip_src = t;
780 	ip->ip_ttl = V_ip_defttl;
781 
782 	if (optlen > 0) {
783 		register u_char *cp;
784 		int opt, cnt;
785 		u_int len;
786 
787 		/*
788 		 * Retrieve any source routing from the incoming packet;
789 		 * add on any record-route or timestamp options.
790 		 */
791 		cp = (u_char *) (ip + 1);
792 		if ((opts = ip_srcroute(m)) == NULL &&
793 		    (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
794 			opts->m_len = sizeof(struct in_addr);
795 			mtod(opts, struct in_addr *)->s_addr = 0;
796 		}
797 		if (opts) {
798 #ifdef ICMPPRINTFS
799 		    if (icmpprintfs)
800 			    printf("icmp_reflect optlen %d rt %d => ",
801 				optlen, opts->m_len);
802 #endif
803 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
804 			    opt = cp[IPOPT_OPTVAL];
805 			    if (opt == IPOPT_EOL)
806 				    break;
807 			    if (opt == IPOPT_NOP)
808 				    len = 1;
809 			    else {
810 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
811 					    break;
812 				    len = cp[IPOPT_OLEN];
813 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
814 				        len > cnt)
815 					    break;
816 			    }
817 			    /*
818 			     * Should check for overflow, but it "can't happen"
819 			     */
820 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
821 				opt == IPOPT_SECURITY) {
822 				    bcopy((caddr_t)cp,
823 					mtod(opts, caddr_t) + opts->m_len, len);
824 				    opts->m_len += len;
825 			    }
826 		    }
827 		    /* Terminate & pad, if necessary */
828 		    cnt = opts->m_len % 4;
829 		    if (cnt) {
830 			    for (; cnt < 4; cnt++) {
831 				    *(mtod(opts, caddr_t) + opts->m_len) =
832 					IPOPT_EOL;
833 				    opts->m_len++;
834 			    }
835 		    }
836 #ifdef ICMPPRINTFS
837 		    if (icmpprintfs)
838 			    printf("%d\n", opts->m_len);
839 #endif
840 		}
841 		ip_stripoptions(m);
842 	}
843 	m_tag_delete_nonpersistent(m);
844 	m->m_flags &= ~(M_BCAST|M_MCAST);
845 	icmp_send(m, opts);
846 done:
847 	if (opts)
848 		(void)m_free(opts);
849 }
850 
851 /*
852  * Send an icmp packet back to the ip level,
853  * after supplying a checksum.
854  */
855 static void
856 icmp_send(struct mbuf *m, struct mbuf *opts)
857 {
858 	register struct ip *ip = mtod(m, struct ip *);
859 	register int hlen;
860 	register struct icmp *icp;
861 
862 	hlen = ip->ip_hl << 2;
863 	m->m_data += hlen;
864 	m->m_len -= hlen;
865 	icp = mtod(m, struct icmp *);
866 	icp->icmp_cksum = 0;
867 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
868 	m->m_data -= hlen;
869 	m->m_len += hlen;
870 	m->m_pkthdr.rcvif = (struct ifnet *)0;
871 #ifdef ICMPPRINTFS
872 	if (icmpprintfs) {
873 		char buf[4 * sizeof "123"];
874 		strcpy(buf, inet_ntoa(ip->ip_dst));
875 		printf("icmp_send dst %s src %s\n",
876 		       buf, inet_ntoa(ip->ip_src));
877 	}
878 #endif
879 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
880 }
881 
882 /*
883  * Return milliseconds since 00:00 UTC in network format.
884  */
885 uint32_t
886 iptime(void)
887 {
888 	struct timeval atv;
889 	u_long t;
890 
891 	getmicrotime(&atv);
892 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
893 	return (htonl(t));
894 }
895 
896 /*
897  * Return the next larger or smaller MTU plateau (table from RFC 1191)
898  * given current value MTU.  If DIR is less than zero, a larger plateau
899  * is returned; otherwise, a smaller value is returned.
900  */
901 int
902 ip_next_mtu(int mtu, int dir)
903 {
904 	static int mtutab[] = {
905 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
906 		296, 68, 0
907 	};
908 	int i, size;
909 
910 	size = (sizeof mtutab) / (sizeof mtutab[0]);
911 	if (dir >= 0) {
912 		for (i = 0; i < size; i++)
913 			if (mtu > mtutab[i])
914 				return mtutab[i];
915 	} else {
916 		for (i = size - 1; i >= 0; i--)
917 			if (mtu < mtutab[i])
918 				return mtutab[i];
919 		if (mtu == mtutab[0])
920 			return mtutab[0];
921 	}
922 	return 0;
923 }
924 #endif /* INET */
925 
926 
927 /*
928  * badport_bandlim() - check for ICMP bandwidth limit
929  *
930  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
931  *	hit our bandwidth limit and it is not ok.
932  *
933  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
934  *
935  *	For now we separate the TCP and UDP subsystems w/ different 'which'
936  *	values.  We may eventually remove this separation (and simplify the
937  *	code further).
938  *
939  *	Note that the printing of the error message is delayed so we can
940  *	properly print the icmp error rate that the system was trying to do
941  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
942  *	the 'final' error, but it doesn't make sense to solve the printing
943  *	delay with more complex code.
944  */
945 
946 int
947 badport_bandlim(int which)
948 {
949 
950 #define	N(a)	(sizeof (a) / sizeof (a[0]))
951 	static struct rate {
952 		const char	*type;
953 		struct timeval	lasttime;
954 		int		curpps;
955 	} rates[BANDLIM_MAX+1] = {
956 		{ "icmp unreach response" },
957 		{ "icmp ping response" },
958 		{ "icmp tstamp response" },
959 		{ "closed port RST response" },
960 		{ "open port RST response" },
961 		{ "icmp6 unreach response" },
962 		{ "sctp ootb response" }
963 	};
964 
965 	/*
966 	 * Return ok status if feature disabled or argument out of range.
967 	 */
968 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
969 		struct rate *r = &rates[which];
970 		int opps = r->curpps;
971 
972 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
973 			return -1;	/* discard packet */
974 		/*
975 		 * If we've dropped below the threshold after having
976 		 * rate-limited traffic print the message.  This preserves
977 		 * the previous behaviour at the expense of added complexity.
978 		 */
979 		if (V_icmplim_output && opps > V_icmplim)
980 			log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
981 				r->type, opps, V_icmplim);
982 	}
983 	return 0;			/* okay to send packet */
984 #undef N
985 }
986