xref: /freebsd/sys/netinet/ip_icmp.c (revision 595e514d0df2bac5b813d35f83e32875dbf16a83)
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 #include "opt_ipsec.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/syslog.h>
47 
48 #include <net/if.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 #ifdef IPSEC
68 #include <netipsec/ipsec.h>
69 #include <netipsec/key.h>
70 #endif
71 
72 #include <machine/in_cksum.h>
73 
74 #include <security/mac/mac_framework.h>
75 #endif /* INET */
76 
77 /*
78  * ICMP routines: error generation, receive packet processing, and
79  * routines to turnaround packets back to the originator, and
80  * host table maintenance routines.
81  */
82 static VNET_DEFINE(int, icmplim) = 200;
83 #define	V_icmplim			VNET(icmplim)
84 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
85 	&VNET_NAME(icmplim), 0,
86 	"Maximum number of ICMP responses per second");
87 
88 static VNET_DEFINE(int, icmplim_output) = 1;
89 #define	V_icmplim_output		VNET(icmplim_output)
90 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
91 	&VNET_NAME(icmplim_output), 0,
92 	"Enable logging of ICMP response rate limiting");
93 
94 #ifdef INET
95 VNET_DEFINE(struct icmpstat, icmpstat);
96 SYSCTL_VNET_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
97 	&VNET_NAME(icmpstat), icmpstat, "");
98 
99 static VNET_DEFINE(int, icmpmaskrepl) = 0;
100 #define	V_icmpmaskrepl			VNET(icmpmaskrepl)
101 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
102 	&VNET_NAME(icmpmaskrepl), 0,
103 	"Reply to ICMP Address Mask Request packets.");
104 
105 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
106 #define	V_icmpmaskfake			VNET(icmpmaskfake)
107 SYSCTL_VNET_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
108 	&VNET_NAME(icmpmaskfake), 0,
109 	"Fake reply to ICMP Address Mask Request packets.");
110 
111 VNET_DEFINE(int, drop_redirect) = 0;
112 
113 static VNET_DEFINE(int, log_redirect) = 0;
114 #define	V_log_redirect			VNET(log_redirect)
115 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
116 	&VNET_NAME(log_redirect), 0,
117 	"Log ICMP redirects to the console");
118 
119 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
120 #define	V_reply_src			VNET(reply_src)
121 SYSCTL_VNET_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
122 	&VNET_NAME(reply_src), IFNAMSIZ,
123 	"icmp reply source for non-local packets.");
124 
125 static VNET_DEFINE(int, icmp_rfi) = 0;
126 #define	V_icmp_rfi			VNET(icmp_rfi)
127 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
128 	&VNET_NAME(icmp_rfi), 0,
129 	"ICMP reply from incoming interface for non-local packets");
130 
131 static VNET_DEFINE(int, icmp_quotelen) = 8;
132 #define	V_icmp_quotelen			VNET(icmp_quotelen)
133 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
134 	&VNET_NAME(icmp_quotelen), 0,
135 	"Number of bytes from original packet to quote in ICMP reply");
136 
137 /*
138  * ICMP broadcast echo sysctl
139  */
140 static VNET_DEFINE(int, icmpbmcastecho) = 0;
141 #define	V_icmpbmcastecho		VNET(icmpbmcastecho)
142 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
143 	&VNET_NAME(icmpbmcastecho), 0,
144 	"");
145 
146 
147 #ifdef ICMPPRINTFS
148 int	icmpprintfs = 0;
149 #endif
150 
151 static void	icmp_reflect(struct mbuf *);
152 static void	icmp_send(struct mbuf *, struct mbuf *);
153 
154 extern	struct protosw inetsw[];
155 
156 static int
157 sysctl_net_icmp_drop_redir(SYSCTL_HANDLER_ARGS)
158 {
159 	int error, new;
160 	int i;
161 	struct radix_node_head *rnh;
162 
163 	new = V_drop_redirect;
164 	error = sysctl_handle_int(oidp, &new, 0, req);
165 	if (error == 0 && req->newptr) {
166 		new = (new != 0) ? 1 : 0;
167 
168 		if (new == V_drop_redirect)
169 			return (0);
170 
171 		for (i = 0; i < rt_numfibs; i++) {
172 			if ((rnh = rt_tables_get_rnh(i, AF_INET)) == NULL)
173 				continue;
174 			RADIX_NODE_HEAD_LOCK(rnh);
175 			in_setmatchfunc(rnh, new);
176 			RADIX_NODE_HEAD_UNLOCK(rnh);
177 		}
178 
179 		V_drop_redirect = new;
180 	}
181 
182 	return (error);
183 }
184 
185 SYSCTL_VNET_PROC(_net_inet_icmp, OID_AUTO, drop_redirect,
186     CTLTYPE_INT|CTLFLAG_RW, 0, 0,
187     sysctl_net_icmp_drop_redir, "I", "Ignore ICMP redirects");
188 
189 /*
190  * Kernel module interface for updating icmpstat.  The argument is an index
191  * into icmpstat treated as an array of u_long.  While this encodes the
192  * general layout of icmpstat into the caller, it doesn't encode its
193  * location, so that future changes to add, for example, per-CPU stats
194  * support won't cause binary compatibility problems for kernel modules.
195  */
196 void
197 kmod_icmpstat_inc(int statnum)
198 {
199 
200 	(*((u_long *)&V_icmpstat + statnum))++;
201 }
202 
203 /*
204  * Generate an error packet of type error
205  * in response to bad packet ip.
206  */
207 void
208 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
209 {
210 	register struct ip *oip = mtod(n, struct ip *), *nip;
211 	register unsigned oiphlen = oip->ip_hl << 2;
212 	register struct icmp *icp;
213 	register struct mbuf *m;
214 	unsigned icmplen, icmpelen, nlen;
215 
216 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
217 #ifdef ICMPPRINTFS
218 	if (icmpprintfs)
219 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
220 #endif
221 	if (type != ICMP_REDIRECT)
222 		ICMPSTAT_INC(icps_error);
223 	/*
224 	 * Don't send error:
225 	 *  if the original packet was encrypted.
226 	 *  if not the first fragment of message.
227 	 *  in response to a multicast or broadcast packet.
228 	 *  if the old packet protocol was an ICMP error message.
229 	 */
230 	if (n->m_flags & M_DECRYPTED)
231 		goto freeit;
232 	if (oip->ip_off & htons(~(IP_MF|IP_DF)))
233 		goto freeit;
234 	if (n->m_flags & (M_BCAST|M_MCAST))
235 		goto freeit;
236 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
237 	  n->m_len >= oiphlen + ICMP_MINLEN &&
238 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
239 		ICMPSTAT_INC(icps_oldicmp);
240 		goto freeit;
241 	}
242 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
243 	if (oiphlen + 8 > n->m_len)
244 		goto freeit;
245 	/*
246 	 * Calculate length to quote from original packet and
247 	 * prevent the ICMP mbuf from overflowing.
248 	 * Unfortunatly this is non-trivial since ip_forward()
249 	 * sends us truncated packets.
250 	 */
251 	nlen = m_length(n, NULL);
252 	if (oip->ip_p == IPPROTO_TCP) {
253 		struct tcphdr *th;
254 		int tcphlen;
255 
256 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
257 		    n->m_next == NULL)
258 			goto stdreply;
259 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
260 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
261 			goto freeit;
262 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
263 		tcphlen = th->th_off << 2;
264 		if (tcphlen < sizeof(struct tcphdr))
265 			goto freeit;
266 		if (ntohs(oip->ip_len) < oiphlen + tcphlen)
267 			goto freeit;
268 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
269 			goto stdreply;
270 		if (n->m_len < oiphlen + tcphlen &&
271 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
272 			goto freeit;
273 		icmpelen = max(tcphlen, min(V_icmp_quotelen,
274 		    ntohs(oip->ip_len) - oiphlen));
275 	} else
276 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
277 
278 	icmplen = min(oiphlen + icmpelen, nlen);
279 	if (icmplen < sizeof(struct ip))
280 		goto freeit;
281 
282 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
283 		m = m_gethdr(M_NOWAIT, MT_DATA);
284 	else
285 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
286 	if (m == NULL)
287 		goto freeit;
288 #ifdef MAC
289 	mac_netinet_icmp_reply(n, m);
290 #endif
291 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
292 	m_align(m, ICMP_MINLEN + icmplen);
293 	m->m_len = ICMP_MINLEN + icmplen;
294 
295 	/* XXX MRT  make the outgoing packet use the same FIB
296 	 * that was associated with the incoming packet
297 	 */
298 	M_SETFIB(m, M_GETFIB(n));
299 	icp = mtod(m, struct icmp *);
300 	ICMPSTAT_INC(icps_outhist[type]);
301 	icp->icmp_type = type;
302 	if (type == ICMP_REDIRECT)
303 		icp->icmp_gwaddr.s_addr = dest;
304 	else {
305 		icp->icmp_void = 0;
306 		/*
307 		 * The following assignments assume an overlay with the
308 		 * just zeroed icmp_void field.
309 		 */
310 		if (type == ICMP_PARAMPROB) {
311 			icp->icmp_pptr = code;
312 			code = 0;
313 		} else if (type == ICMP_UNREACH &&
314 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
315 			icp->icmp_nextmtu = htons(mtu);
316 		}
317 	}
318 	icp->icmp_code = code;
319 
320 	/*
321 	 * Copy the quotation into ICMP message and
322 	 * convert quoted IP header back to network representation.
323 	 */
324 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
325 	nip = &icp->icmp_ip;
326 
327 	/*
328 	 * Set up ICMP message mbuf and copy old IP header (without options
329 	 * in front of ICMP message.
330 	 * If the original mbuf was meant to bypass the firewall, the error
331 	 * reply should bypass as well.
332 	 */
333 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
334 	m->m_data -= sizeof(struct ip);
335 	m->m_len += sizeof(struct ip);
336 	m->m_pkthdr.len = m->m_len;
337 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
338 	nip = mtod(m, struct ip *);
339 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
340 	nip->ip_len = htons(m->m_len);
341 	nip->ip_v = IPVERSION;
342 	nip->ip_hl = 5;
343 	nip->ip_p = IPPROTO_ICMP;
344 	nip->ip_tos = 0;
345 	icmp_reflect(m);
346 
347 freeit:
348 	m_freem(n);
349 }
350 
351 /*
352  * Process a received ICMP message.
353  */
354 void
355 icmp_input(struct mbuf *m, int off)
356 {
357 	struct icmp *icp;
358 	struct in_ifaddr *ia;
359 	struct ip *ip = mtod(m, struct ip *);
360 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
361 	int hlen = off;
362 	int icmplen = ntohs(ip->ip_len) - off;
363 	int i, code;
364 	void (*ctlfunc)(int, struct sockaddr *, void *);
365 	int fibnum;
366 
367 	/*
368 	 * Locate icmp structure in mbuf, and check
369 	 * that not corrupted and of at least minimum length.
370 	 */
371 #ifdef ICMPPRINTFS
372 	if (icmpprintfs) {
373 		char buf[4 * sizeof "123"];
374 		strcpy(buf, inet_ntoa(ip->ip_src));
375 		printf("icmp_input from %s to %s, len %d\n",
376 		       buf, inet_ntoa(ip->ip_dst), icmplen);
377 	}
378 #endif
379 	if (icmplen < ICMP_MINLEN) {
380 		ICMPSTAT_INC(icps_tooshort);
381 		goto freeit;
382 	}
383 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
384 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
385 		ICMPSTAT_INC(icps_tooshort);
386 		return;
387 	}
388 	ip = mtod(m, struct ip *);
389 	m->m_len -= hlen;
390 	m->m_data += hlen;
391 	icp = mtod(m, struct icmp *);
392 	if (in_cksum(m, icmplen)) {
393 		ICMPSTAT_INC(icps_checksum);
394 		goto freeit;
395 	}
396 	m->m_len += hlen;
397 	m->m_data -= hlen;
398 
399 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
400 		/*
401 		 * Deliver very specific ICMP type only.
402 		 */
403 		switch (icp->icmp_type) {
404 		case ICMP_UNREACH:
405 		case ICMP_TIMXCEED:
406 			break;
407 		default:
408 			goto freeit;
409 		}
410 	}
411 
412 #ifdef ICMPPRINTFS
413 	if (icmpprintfs)
414 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
415 		    icp->icmp_code);
416 #endif
417 
418 	/*
419 	 * Message type specific processing.
420 	 */
421 	if (icp->icmp_type > ICMP_MAXTYPE)
422 		goto raw;
423 
424 	/* Initialize */
425 	bzero(&icmpsrc, sizeof(icmpsrc));
426 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
427 	icmpsrc.sin_family = AF_INET;
428 	bzero(&icmpdst, sizeof(icmpdst));
429 	icmpdst.sin_len = sizeof(struct sockaddr_in);
430 	icmpdst.sin_family = AF_INET;
431 	bzero(&icmpgw, sizeof(icmpgw));
432 	icmpgw.sin_len = sizeof(struct sockaddr_in);
433 	icmpgw.sin_family = AF_INET;
434 
435 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
436 	code = icp->icmp_code;
437 	switch (icp->icmp_type) {
438 
439 	case ICMP_UNREACH:
440 		switch (code) {
441 			case ICMP_UNREACH_NET:
442 			case ICMP_UNREACH_HOST:
443 			case ICMP_UNREACH_SRCFAIL:
444 			case ICMP_UNREACH_NET_UNKNOWN:
445 			case ICMP_UNREACH_HOST_UNKNOWN:
446 			case ICMP_UNREACH_ISOLATED:
447 			case ICMP_UNREACH_TOSNET:
448 			case ICMP_UNREACH_TOSHOST:
449 			case ICMP_UNREACH_HOST_PRECEDENCE:
450 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
451 				code = PRC_UNREACH_NET;
452 				break;
453 
454 			case ICMP_UNREACH_NEEDFRAG:
455 				code = PRC_MSGSIZE;
456 				break;
457 
458 			/*
459 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
460 			 * Treat subcodes 2,3 as immediate RST
461 			 */
462 			case ICMP_UNREACH_PROTOCOL:
463 			case ICMP_UNREACH_PORT:
464 				code = PRC_UNREACH_PORT;
465 				break;
466 
467 			case ICMP_UNREACH_NET_PROHIB:
468 			case ICMP_UNREACH_HOST_PROHIB:
469 			case ICMP_UNREACH_FILTER_PROHIB:
470 				code = PRC_UNREACH_ADMIN_PROHIB;
471 				break;
472 
473 			default:
474 				goto badcode;
475 		}
476 		goto deliver;
477 
478 	case ICMP_TIMXCEED:
479 		if (code > 1)
480 			goto badcode;
481 		code += PRC_TIMXCEED_INTRANS;
482 		goto deliver;
483 
484 	case ICMP_PARAMPROB:
485 		if (code > 1)
486 			goto badcode;
487 		code = PRC_PARAMPROB;
488 		goto deliver;
489 
490 	case ICMP_SOURCEQUENCH:
491 		if (code)
492 			goto badcode;
493 		code = PRC_QUENCH;
494 	deliver:
495 		/*
496 		 * Problem with datagram; advise higher level routines.
497 		 */
498 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
499 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
500 			ICMPSTAT_INC(icps_badlen);
501 			goto freeit;
502 		}
503 		/* Discard ICMP's in response to multicast packets */
504 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
505 			goto badcode;
506 #ifdef ICMPPRINTFS
507 		if (icmpprintfs)
508 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
509 #endif
510 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
511 		/*
512 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
513 		 * notification to TCP layer.
514 		 */
515 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
516 		if (ctlfunc)
517 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
518 				   (void *)&icp->icmp_ip);
519 		break;
520 
521 	badcode:
522 		ICMPSTAT_INC(icps_badcode);
523 		break;
524 
525 	case ICMP_ECHO:
526 		if (!V_icmpbmcastecho
527 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
528 			ICMPSTAT_INC(icps_bmcastecho);
529 			break;
530 		}
531 		icp->icmp_type = ICMP_ECHOREPLY;
532 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
533 			goto freeit;
534 		else
535 			goto reflect;
536 
537 	case ICMP_TSTAMP:
538 		if (!V_icmpbmcastecho
539 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
540 			ICMPSTAT_INC(icps_bmcasttstamp);
541 			break;
542 		}
543 		if (icmplen < ICMP_TSLEN) {
544 			ICMPSTAT_INC(icps_badlen);
545 			break;
546 		}
547 		icp->icmp_type = ICMP_TSTAMPREPLY;
548 		icp->icmp_rtime = iptime();
549 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
550 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
551 			goto freeit;
552 		else
553 			goto reflect;
554 
555 	case ICMP_MASKREQ:
556 		if (V_icmpmaskrepl == 0)
557 			break;
558 		/*
559 		 * We are not able to respond with all ones broadcast
560 		 * unless we receive it over a point-to-point interface.
561 		 */
562 		if (icmplen < ICMP_MASKLEN)
563 			break;
564 		switch (ip->ip_dst.s_addr) {
565 
566 		case INADDR_BROADCAST:
567 		case INADDR_ANY:
568 			icmpdst.sin_addr = ip->ip_src;
569 			break;
570 
571 		default:
572 			icmpdst.sin_addr = ip->ip_dst;
573 		}
574 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
575 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
576 		if (ia == NULL)
577 			break;
578 		if (ia->ia_ifp == NULL) {
579 			ifa_free(&ia->ia_ifa);
580 			break;
581 		}
582 		icp->icmp_type = ICMP_MASKREPLY;
583 		if (V_icmpmaskfake == 0)
584 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
585 		else
586 			icp->icmp_mask = V_icmpmaskfake;
587 		if (ip->ip_src.s_addr == 0) {
588 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
589 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
590 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
591 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
592 		}
593 		ifa_free(&ia->ia_ifa);
594 reflect:
595 		ICMPSTAT_INC(icps_reflect);
596 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
597 		icmp_reflect(m);
598 		return;
599 
600 	case ICMP_REDIRECT:
601 		if (V_log_redirect) {
602 			u_long src, dst, gw;
603 
604 			src = ntohl(ip->ip_src.s_addr);
605 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
606 			gw = ntohl(icp->icmp_gwaddr.s_addr);
607 			printf("icmp redirect from %d.%d.%d.%d: "
608 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
609 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
610 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
611 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
612 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
613 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
614 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
615 		}
616 		/*
617 		 * RFC1812 says we must ignore ICMP redirects if we
618 		 * are acting as router.
619 		 */
620 		if (V_drop_redirect || V_ipforwarding)
621 			break;
622 		if (code > 3)
623 			goto badcode;
624 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
625 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
626 			ICMPSTAT_INC(icps_badlen);
627 			break;
628 		}
629 		/*
630 		 * Short circuit routing redirects to force
631 		 * immediate change in the kernel's routing
632 		 * tables.  The message is also handed to anyone
633 		 * listening on a raw socket (e.g. the routing
634 		 * daemon for use in updating its tables).
635 		 */
636 		icmpgw.sin_addr = ip->ip_src;
637 		icmpdst.sin_addr = icp->icmp_gwaddr;
638 #ifdef	ICMPPRINTFS
639 		if (icmpprintfs) {
640 			char buf[4 * sizeof "123"];
641 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
642 
643 			printf("redirect dst %s to %s\n",
644 			       buf, inet_ntoa(icp->icmp_gwaddr));
645 		}
646 #endif
647 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
648 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
649 			in_rtredirect((struct sockaddr *)&icmpsrc,
650 			  (struct sockaddr *)&icmpdst,
651 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
652 			  (struct sockaddr *)&icmpgw, fibnum);
653 		}
654 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
655 #ifdef IPSEC
656 		key_sa_routechange((struct sockaddr *)&icmpsrc);
657 #endif
658 		break;
659 
660 	/*
661 	 * No kernel processing for the following;
662 	 * just fall through to send to raw listener.
663 	 */
664 	case ICMP_ECHOREPLY:
665 	case ICMP_ROUTERADVERT:
666 	case ICMP_ROUTERSOLICIT:
667 	case ICMP_TSTAMPREPLY:
668 	case ICMP_IREQREPLY:
669 	case ICMP_MASKREPLY:
670 	default:
671 		break;
672 	}
673 
674 raw:
675 	rip_input(m, off);
676 	return;
677 
678 freeit:
679 	m_freem(m);
680 }
681 
682 /*
683  * Reflect the ip packet back to the source
684  */
685 static void
686 icmp_reflect(struct mbuf *m)
687 {
688 	struct ip *ip = mtod(m, struct ip *);
689 	struct ifaddr *ifa;
690 	struct ifnet *ifp;
691 	struct in_ifaddr *ia;
692 	struct in_addr t;
693 	struct mbuf *opts = 0;
694 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
695 
696 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
697 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
698 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
699 		m_freem(m);	/* Bad return address */
700 		ICMPSTAT_INC(icps_badaddr);
701 		goto done;	/* Ip_output() will check for broadcast */
702 	}
703 
704 	t = ip->ip_dst;
705 	ip->ip_dst = ip->ip_src;
706 
707 	/*
708 	 * Source selection for ICMP replies:
709 	 *
710 	 * If the incoming packet was addressed directly to one of our
711 	 * own addresses, use dst as the src for the reply.
712 	 */
713 	IN_IFADDR_RLOCK();
714 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
715 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
716 			t = IA_SIN(ia)->sin_addr;
717 			IN_IFADDR_RUNLOCK();
718 			goto match;
719 		}
720 	}
721 	IN_IFADDR_RUNLOCK();
722 
723 	/*
724 	 * If the incoming packet was addressed to one of our broadcast
725 	 * addresses, use the first non-broadcast address which corresponds
726 	 * to the incoming interface.
727 	 */
728 	ifp = m->m_pkthdr.rcvif;
729 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
730 		IF_ADDR_RLOCK(ifp);
731 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
732 			if (ifa->ifa_addr->sa_family != AF_INET)
733 				continue;
734 			ia = ifatoia(ifa);
735 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
736 			    t.s_addr) {
737 				t = IA_SIN(ia)->sin_addr;
738 				IF_ADDR_RUNLOCK(ifp);
739 				goto match;
740 			}
741 		}
742 		IF_ADDR_RUNLOCK(ifp);
743 	}
744 	/*
745 	 * If the packet was transiting through us, use the address of
746 	 * the interface the packet came through in.  If that interface
747 	 * doesn't have a suitable IP address, the normal selection
748 	 * criteria apply.
749 	 */
750 	if (V_icmp_rfi && ifp != NULL) {
751 		IF_ADDR_RLOCK(ifp);
752 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
753 			if (ifa->ifa_addr->sa_family != AF_INET)
754 				continue;
755 			ia = ifatoia(ifa);
756 			t = IA_SIN(ia)->sin_addr;
757 			IF_ADDR_RUNLOCK(ifp);
758 			goto match;
759 		}
760 		IF_ADDR_RUNLOCK(ifp);
761 	}
762 	/*
763 	 * If the incoming packet was not addressed directly to us, use
764 	 * designated interface for icmp replies specified by sysctl
765 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
766 	 * with normal source selection.
767 	 */
768 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
769 		IF_ADDR_RLOCK(ifp);
770 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
771 			if (ifa->ifa_addr->sa_family != AF_INET)
772 				continue;
773 			ia = ifatoia(ifa);
774 			t = IA_SIN(ia)->sin_addr;
775 			IF_ADDR_RUNLOCK(ifp);
776 			goto match;
777 		}
778 		IF_ADDR_RUNLOCK(ifp);
779 	}
780 	/*
781 	 * If the packet was transiting through us, use the address of
782 	 * the interface that is the closest to the packet source.
783 	 * When we don't have a route back to the packet source, stop here
784 	 * and drop the packet.
785 	 */
786 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
787 	if (ia == NULL) {
788 		m_freem(m);
789 		ICMPSTAT_INC(icps_noroute);
790 		goto done;
791 	}
792 	t = IA_SIN(ia)->sin_addr;
793 	ifa_free(&ia->ia_ifa);
794 match:
795 #ifdef MAC
796 	mac_netinet_icmp_replyinplace(m);
797 #endif
798 	ip->ip_src = t;
799 	ip->ip_ttl = V_ip_defttl;
800 
801 	if (optlen > 0) {
802 		register u_char *cp;
803 		int opt, cnt;
804 		u_int len;
805 
806 		/*
807 		 * Retrieve any source routing from the incoming packet;
808 		 * add on any record-route or timestamp options.
809 		 */
810 		cp = (u_char *) (ip + 1);
811 		if ((opts = ip_srcroute(m)) == 0 &&
812 		    (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
813 			opts->m_len = sizeof(struct in_addr);
814 			mtod(opts, struct in_addr *)->s_addr = 0;
815 		}
816 		if (opts) {
817 #ifdef ICMPPRINTFS
818 		    if (icmpprintfs)
819 			    printf("icmp_reflect optlen %d rt %d => ",
820 				optlen, opts->m_len);
821 #endif
822 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
823 			    opt = cp[IPOPT_OPTVAL];
824 			    if (opt == IPOPT_EOL)
825 				    break;
826 			    if (opt == IPOPT_NOP)
827 				    len = 1;
828 			    else {
829 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
830 					    break;
831 				    len = cp[IPOPT_OLEN];
832 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
833 				        len > cnt)
834 					    break;
835 			    }
836 			    /*
837 			     * Should check for overflow, but it "can't happen"
838 			     */
839 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
840 				opt == IPOPT_SECURITY) {
841 				    bcopy((caddr_t)cp,
842 					mtod(opts, caddr_t) + opts->m_len, len);
843 				    opts->m_len += len;
844 			    }
845 		    }
846 		    /* Terminate & pad, if necessary */
847 		    cnt = opts->m_len % 4;
848 		    if (cnt) {
849 			    for (; cnt < 4; cnt++) {
850 				    *(mtod(opts, caddr_t) + opts->m_len) =
851 					IPOPT_EOL;
852 				    opts->m_len++;
853 			    }
854 		    }
855 #ifdef ICMPPRINTFS
856 		    if (icmpprintfs)
857 			    printf("%d\n", opts->m_len);
858 #endif
859 		}
860 		ip_stripoptions(m);
861 	}
862 	m_tag_delete_nonpersistent(m);
863 	m->m_flags &= ~(M_BCAST|M_MCAST);
864 	icmp_send(m, opts);
865 done:
866 	if (opts)
867 		(void)m_free(opts);
868 }
869 
870 /*
871  * Send an icmp packet back to the ip level,
872  * after supplying a checksum.
873  */
874 static void
875 icmp_send(struct mbuf *m, struct mbuf *opts)
876 {
877 	register struct ip *ip = mtod(m, struct ip *);
878 	register int hlen;
879 	register struct icmp *icp;
880 
881 	hlen = ip->ip_hl << 2;
882 	m->m_data += hlen;
883 	m->m_len -= hlen;
884 	icp = mtod(m, struct icmp *);
885 	icp->icmp_cksum = 0;
886 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
887 	m->m_data -= hlen;
888 	m->m_len += hlen;
889 	m->m_pkthdr.rcvif = (struct ifnet *)0;
890 #ifdef ICMPPRINTFS
891 	if (icmpprintfs) {
892 		char buf[4 * sizeof "123"];
893 		strcpy(buf, inet_ntoa(ip->ip_dst));
894 		printf("icmp_send dst %s src %s\n",
895 		       buf, inet_ntoa(ip->ip_src));
896 	}
897 #endif
898 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
899 }
900 
901 /*
902  * Return milliseconds since 00:00 GMT in network format.
903  */
904 uint32_t
905 iptime(void)
906 {
907 	struct timeval atv;
908 	u_long t;
909 
910 	getmicrotime(&atv);
911 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
912 	return (htonl(t));
913 }
914 
915 /*
916  * Return the next larger or smaller MTU plateau (table from RFC 1191)
917  * given current value MTU.  If DIR is less than zero, a larger plateau
918  * is returned; otherwise, a smaller value is returned.
919  */
920 int
921 ip_next_mtu(int mtu, int dir)
922 {
923 	static int mtutab[] = {
924 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
925 		296, 68, 0
926 	};
927 	int i, size;
928 
929 	size = (sizeof mtutab) / (sizeof mtutab[0]);
930 	if (dir >= 0) {
931 		for (i = 0; i < size; i++)
932 			if (mtu > mtutab[i])
933 				return mtutab[i];
934 	} else {
935 		for (i = size - 1; i >= 0; i--)
936 			if (mtu < mtutab[i])
937 				return mtutab[i];
938 		if (mtu == mtutab[0])
939 			return mtutab[0];
940 	}
941 	return 0;
942 }
943 #endif /* INET */
944 
945 
946 /*
947  * badport_bandlim() - check for ICMP bandwidth limit
948  *
949  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
950  *	hit our bandwidth limit and it is not ok.
951  *
952  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
953  *
954  *	For now we separate the TCP and UDP subsystems w/ different 'which'
955  *	values.  We may eventually remove this separation (and simplify the
956  *	code further).
957  *
958  *	Note that the printing of the error message is delayed so we can
959  *	properly print the icmp error rate that the system was trying to do
960  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
961  *	the 'final' error, but it doesn't make sense to solve the printing
962  *	delay with more complex code.
963  */
964 
965 int
966 badport_bandlim(int which)
967 {
968 
969 #define	N(a)	(sizeof (a) / sizeof (a[0]))
970 	static struct rate {
971 		const char	*type;
972 		struct timeval	lasttime;
973 		int		curpps;
974 	} rates[BANDLIM_MAX+1] = {
975 		{ "icmp unreach response" },
976 		{ "icmp ping response" },
977 		{ "icmp tstamp response" },
978 		{ "closed port RST response" },
979 		{ "open port RST response" },
980 		{ "icmp6 unreach response" },
981 		{ "sctp ootb response" }
982 	};
983 
984 	/*
985 	 * Return ok status if feature disabled or argument out of range.
986 	 */
987 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
988 		struct rate *r = &rates[which];
989 		int opps = r->curpps;
990 
991 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
992 			return -1;	/* discard packet */
993 		/*
994 		 * If we've dropped below the threshold after having
995 		 * rate-limited traffic print the message.  This preserves
996 		 * the previous behaviour at the expense of added complexity.
997 		 */
998 		if (V_icmplim_output && opps > V_icmplim)
999 			log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
1000 				r->type, opps, V_icmplim);
1001 	}
1002 	return 0;			/* okay to send packet */
1003 #undef N
1004 }
1005