xref: /freebsd/sys/netinet/ip_icmp.c (revision 61af1d13936ec56808f62d13dd8698f73b440dc1)
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 rate limiting of ICMP responses");
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 static VNET_DEFINE(int, drop_redirect) = 0;
112 #define	V_drop_redirect			VNET(drop_redirect)
113 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
114 	&VNET_NAME(drop_redirect), 0,
115 	"Ignore ICMP redirects");
116 
117 static VNET_DEFINE(int, log_redirect) = 0;
118 #define	V_log_redirect			VNET(log_redirect)
119 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, log_redirect, 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_VNET_STRING(_net_inet_icmp, OID_AUTO, reply_src, 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_VNET_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, 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_VNET_INT(_net_inet_icmp, OID_AUTO, quotelen, 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_VNET_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
147 	&VNET_NAME(icmpbmcastecho), 0,
148 	"");
149 
150 
151 #ifdef ICMPPRINTFS
152 int	icmpprintfs = 0;
153 #endif
154 
155 static void	icmp_reflect(struct mbuf *);
156 static void	icmp_send(struct mbuf *, struct mbuf *);
157 
158 extern	struct protosw inetsw[];
159 
160 /*
161  * Kernel module interface for updating icmpstat.  The argument is an index
162  * into icmpstat treated as an array of u_long.  While this encodes the
163  * general layout of icmpstat into the caller, it doesn't encode its
164  * location, so that future changes to add, for example, per-CPU stats
165  * support won't cause binary compatibility problems for kernel modules.
166  */
167 void
168 kmod_icmpstat_inc(int statnum)
169 {
170 
171 	(*((u_long *)&V_icmpstat + statnum))++;
172 }
173 
174 /*
175  * Generate an error packet of type error
176  * in response to bad packet ip.
177  */
178 void
179 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
180 {
181 	register struct ip *oip = mtod(n, struct ip *), *nip;
182 	register unsigned oiphlen = oip->ip_hl << 2;
183 	register struct icmp *icp;
184 	register struct mbuf *m;
185 	unsigned icmplen, icmpelen, nlen;
186 
187 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
188 #ifdef ICMPPRINTFS
189 	if (icmpprintfs)
190 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
191 #endif
192 	if (type != ICMP_REDIRECT)
193 		ICMPSTAT_INC(icps_error);
194 	/*
195 	 * Don't send error:
196 	 *  if the original packet was encrypted.
197 	 *  if not the first fragment of message.
198 	 *  in response to a multicast or broadcast packet.
199 	 *  if the old packet protocol was an ICMP error message.
200 	 */
201 	if (n->m_flags & M_DECRYPTED)
202 		goto freeit;
203 	if (oip->ip_off & ~(IP_MF|IP_DF))
204 		goto freeit;
205 	if (n->m_flags & (M_BCAST|M_MCAST))
206 		goto freeit;
207 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
208 	  n->m_len >= oiphlen + ICMP_MINLEN &&
209 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
210 		ICMPSTAT_INC(icps_oldicmp);
211 		goto freeit;
212 	}
213 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
214 	if (oiphlen + 8 > n->m_len)
215 		goto freeit;
216 	/*
217 	 * Calculate length to quote from original packet and
218 	 * prevent the ICMP mbuf from overflowing.
219 	 * Unfortunatly this is non-trivial since ip_forward()
220 	 * sends us truncated packets.
221 	 */
222 	nlen = m_length(n, NULL);
223 	if (oip->ip_p == IPPROTO_TCP) {
224 		struct tcphdr *th;
225 		int tcphlen;
226 
227 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
228 		    n->m_next == NULL)
229 			goto stdreply;
230 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
231 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
232 			goto freeit;
233 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
234 		tcphlen = th->th_off << 2;
235 		if (tcphlen < sizeof(struct tcphdr))
236 			goto freeit;
237 		if (oip->ip_len < oiphlen + tcphlen)
238 			goto freeit;
239 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
240 			goto stdreply;
241 		if (n->m_len < oiphlen + tcphlen &&
242 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
243 			goto freeit;
244 		icmpelen = max(tcphlen, min(V_icmp_quotelen, oip->ip_len - oiphlen));
245 	} else
246 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, oip->ip_len - oiphlen));
247 
248 	icmplen = min(oiphlen + icmpelen, nlen);
249 	if (icmplen < sizeof(struct ip))
250 		goto freeit;
251 
252 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
253 		m = m_gethdr(M_DONTWAIT, MT_DATA);
254 	else
255 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
256 	if (m == NULL)
257 		goto freeit;
258 #ifdef MAC
259 	mac_netinet_icmp_reply(n, m);
260 #endif
261 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
262 	m_align(m, ICMP_MINLEN + icmplen);
263 	m->m_len = ICMP_MINLEN + icmplen;
264 
265 	/* XXX MRT  make the outgoing packet use the same FIB
266 	 * that was associated with the incoming packet
267 	 */
268 	M_SETFIB(m, M_GETFIB(n));
269 	icp = mtod(m, struct icmp *);
270 	ICMPSTAT_INC(icps_outhist[type]);
271 	icp->icmp_type = type;
272 	if (type == ICMP_REDIRECT)
273 		icp->icmp_gwaddr.s_addr = dest;
274 	else {
275 		icp->icmp_void = 0;
276 		/*
277 		 * The following assignments assume an overlay with the
278 		 * just zeroed icmp_void field.
279 		 */
280 		if (type == ICMP_PARAMPROB) {
281 			icp->icmp_pptr = code;
282 			code = 0;
283 		} else if (type == ICMP_UNREACH &&
284 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
285 			icp->icmp_nextmtu = htons(mtu);
286 		}
287 	}
288 	icp->icmp_code = code;
289 
290 	/*
291 	 * Copy the quotation into ICMP message and
292 	 * convert quoted IP header back to network representation.
293 	 */
294 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
295 	nip = &icp->icmp_ip;
296 	nip->ip_len = htons(nip->ip_len);
297 	nip->ip_off = htons(nip->ip_off);
298 
299 	/*
300 	 * Set up ICMP message mbuf and copy old IP header (without options
301 	 * in front of ICMP message.
302 	 * If the original mbuf was meant to bypass the firewall, the error
303 	 * reply should bypass as well.
304 	 */
305 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
306 	m->m_data -= sizeof(struct ip);
307 	m->m_len += sizeof(struct ip);
308 	m->m_pkthdr.len = m->m_len;
309 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
310 	nip = mtod(m, struct ip *);
311 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
312 	nip->ip_len = m->m_len;
313 	nip->ip_v = IPVERSION;
314 	nip->ip_hl = 5;
315 	nip->ip_p = IPPROTO_ICMP;
316 	nip->ip_tos = 0;
317 	icmp_reflect(m);
318 
319 freeit:
320 	m_freem(n);
321 }
322 
323 /*
324  * Process a received ICMP message.
325  */
326 void
327 icmp_input(struct mbuf *m, int off)
328 {
329 	struct icmp *icp;
330 	struct in_ifaddr *ia;
331 	struct ip *ip = mtod(m, struct ip *);
332 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
333 	int hlen = off;
334 	int icmplen = ip->ip_len;
335 	int i, code;
336 	void (*ctlfunc)(int, struct sockaddr *, void *);
337 	int fibnum;
338 
339 	/*
340 	 * Locate icmp structure in mbuf, and check
341 	 * that not corrupted and of at least minimum length.
342 	 */
343 #ifdef ICMPPRINTFS
344 	if (icmpprintfs) {
345 		char buf[4 * sizeof "123"];
346 		strcpy(buf, inet_ntoa(ip->ip_src));
347 		printf("icmp_input from %s to %s, len %d\n",
348 		       buf, inet_ntoa(ip->ip_dst), icmplen);
349 	}
350 #endif
351 	if (icmplen < ICMP_MINLEN) {
352 		ICMPSTAT_INC(icps_tooshort);
353 		goto freeit;
354 	}
355 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
356 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
357 		ICMPSTAT_INC(icps_tooshort);
358 		return;
359 	}
360 	ip = mtod(m, struct ip *);
361 	m->m_len -= hlen;
362 	m->m_data += hlen;
363 	icp = mtod(m, struct icmp *);
364 	if (in_cksum(m, icmplen)) {
365 		ICMPSTAT_INC(icps_checksum);
366 		goto freeit;
367 	}
368 	m->m_len += hlen;
369 	m->m_data -= hlen;
370 
371 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
372 		/*
373 		 * Deliver very specific ICMP type only.
374 		 */
375 		switch (icp->icmp_type) {
376 		case ICMP_UNREACH:
377 		case ICMP_TIMXCEED:
378 			break;
379 		default:
380 			goto freeit;
381 		}
382 	}
383 
384 #ifdef ICMPPRINTFS
385 	if (icmpprintfs)
386 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
387 		    icp->icmp_code);
388 #endif
389 
390 	/*
391 	 * Message type specific processing.
392 	 */
393 	if (icp->icmp_type > ICMP_MAXTYPE)
394 		goto raw;
395 
396 	/* Initialize */
397 	bzero(&icmpsrc, sizeof(icmpsrc));
398 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
399 	icmpsrc.sin_family = AF_INET;
400 	bzero(&icmpdst, sizeof(icmpdst));
401 	icmpdst.sin_len = sizeof(struct sockaddr_in);
402 	icmpdst.sin_family = AF_INET;
403 	bzero(&icmpgw, sizeof(icmpgw));
404 	icmpgw.sin_len = sizeof(struct sockaddr_in);
405 	icmpgw.sin_family = AF_INET;
406 
407 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
408 	code = icp->icmp_code;
409 	switch (icp->icmp_type) {
410 
411 	case ICMP_UNREACH:
412 		switch (code) {
413 			case ICMP_UNREACH_NET:
414 			case ICMP_UNREACH_HOST:
415 			case ICMP_UNREACH_SRCFAIL:
416 			case ICMP_UNREACH_NET_UNKNOWN:
417 			case ICMP_UNREACH_HOST_UNKNOWN:
418 			case ICMP_UNREACH_ISOLATED:
419 			case ICMP_UNREACH_TOSNET:
420 			case ICMP_UNREACH_TOSHOST:
421 			case ICMP_UNREACH_HOST_PRECEDENCE:
422 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
423 				code = PRC_UNREACH_NET;
424 				break;
425 
426 			case ICMP_UNREACH_NEEDFRAG:
427 				code = PRC_MSGSIZE;
428 				break;
429 
430 			/*
431 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
432 			 * Treat subcodes 2,3 as immediate RST
433 			 */
434 			case ICMP_UNREACH_PROTOCOL:
435 			case ICMP_UNREACH_PORT:
436 				code = PRC_UNREACH_PORT;
437 				break;
438 
439 			case ICMP_UNREACH_NET_PROHIB:
440 			case ICMP_UNREACH_HOST_PROHIB:
441 			case ICMP_UNREACH_FILTER_PROHIB:
442 				code = PRC_UNREACH_ADMIN_PROHIB;
443 				break;
444 
445 			default:
446 				goto badcode;
447 		}
448 		goto deliver;
449 
450 	case ICMP_TIMXCEED:
451 		if (code > 1)
452 			goto badcode;
453 		code += PRC_TIMXCEED_INTRANS;
454 		goto deliver;
455 
456 	case ICMP_PARAMPROB:
457 		if (code > 1)
458 			goto badcode;
459 		code = PRC_PARAMPROB;
460 		goto deliver;
461 
462 	case ICMP_SOURCEQUENCH:
463 		if (code)
464 			goto badcode;
465 		code = PRC_QUENCH;
466 	deliver:
467 		/*
468 		 * Problem with datagram; advise higher level routines.
469 		 */
470 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
471 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
472 			ICMPSTAT_INC(icps_badlen);
473 			goto freeit;
474 		}
475 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
476 		/* Discard ICMP's in response to multicast packets */
477 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
478 			goto badcode;
479 #ifdef ICMPPRINTFS
480 		if (icmpprintfs)
481 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
482 #endif
483 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
484 		/*
485 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
486 		 * notification to TCP layer.
487 		 */
488 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
489 		if (ctlfunc)
490 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
491 				   (void *)&icp->icmp_ip);
492 		break;
493 
494 	badcode:
495 		ICMPSTAT_INC(icps_badcode);
496 		break;
497 
498 	case ICMP_ECHO:
499 		if (!V_icmpbmcastecho
500 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
501 			ICMPSTAT_INC(icps_bmcastecho);
502 			break;
503 		}
504 		icp->icmp_type = ICMP_ECHOREPLY;
505 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
506 			goto freeit;
507 		else
508 			goto reflect;
509 
510 	case ICMP_TSTAMP:
511 		if (!V_icmpbmcastecho
512 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
513 			ICMPSTAT_INC(icps_bmcasttstamp);
514 			break;
515 		}
516 		if (icmplen < ICMP_TSLEN) {
517 			ICMPSTAT_INC(icps_badlen);
518 			break;
519 		}
520 		icp->icmp_type = ICMP_TSTAMPREPLY;
521 		icp->icmp_rtime = iptime();
522 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
523 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
524 			goto freeit;
525 		else
526 			goto reflect;
527 
528 	case ICMP_MASKREQ:
529 		if (V_icmpmaskrepl == 0)
530 			break;
531 		/*
532 		 * We are not able to respond with all ones broadcast
533 		 * unless we receive it over a point-to-point interface.
534 		 */
535 		if (icmplen < ICMP_MASKLEN)
536 			break;
537 		switch (ip->ip_dst.s_addr) {
538 
539 		case INADDR_BROADCAST:
540 		case INADDR_ANY:
541 			icmpdst.sin_addr = ip->ip_src;
542 			break;
543 
544 		default:
545 			icmpdst.sin_addr = ip->ip_dst;
546 		}
547 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
548 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
549 		if (ia == NULL)
550 			break;
551 		if (ia->ia_ifp == NULL) {
552 			ifa_free(&ia->ia_ifa);
553 			break;
554 		}
555 		icp->icmp_type = ICMP_MASKREPLY;
556 		if (V_icmpmaskfake == 0)
557 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
558 		else
559 			icp->icmp_mask = V_icmpmaskfake;
560 		if (ip->ip_src.s_addr == 0) {
561 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
562 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
563 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
564 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
565 		}
566 		ifa_free(&ia->ia_ifa);
567 reflect:
568 		ip->ip_len += hlen;	/* since ip_input deducts this */
569 		ICMPSTAT_INC(icps_reflect);
570 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
571 		icmp_reflect(m);
572 		return;
573 
574 	case ICMP_REDIRECT:
575 		if (V_log_redirect) {
576 			u_long src, dst, gw;
577 
578 			src = ntohl(ip->ip_src.s_addr);
579 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
580 			gw = ntohl(icp->icmp_gwaddr.s_addr);
581 			printf("icmp redirect from %d.%d.%d.%d: "
582 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
583 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
584 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
585 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
586 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
587 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
588 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
589 		}
590 		/*
591 		 * RFC1812 says we must ignore ICMP redirects if we
592 		 * are acting as router.
593 		 */
594 		if (V_drop_redirect || V_ipforwarding)
595 			break;
596 		if (code > 3)
597 			goto badcode;
598 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
599 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
600 			ICMPSTAT_INC(icps_badlen);
601 			break;
602 		}
603 		/*
604 		 * Short circuit routing redirects to force
605 		 * immediate change in the kernel's routing
606 		 * tables.  The message is also handed to anyone
607 		 * listening on a raw socket (e.g. the routing
608 		 * daemon for use in updating its tables).
609 		 */
610 		icmpgw.sin_addr = ip->ip_src;
611 		icmpdst.sin_addr = icp->icmp_gwaddr;
612 #ifdef	ICMPPRINTFS
613 		if (icmpprintfs) {
614 			char buf[4 * sizeof "123"];
615 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
616 
617 			printf("redirect dst %s to %s\n",
618 			       buf, inet_ntoa(icp->icmp_gwaddr));
619 		}
620 #endif
621 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
622 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
623 			in_rtredirect((struct sockaddr *)&icmpsrc,
624 			  (struct sockaddr *)&icmpdst,
625 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
626 			  (struct sockaddr *)&icmpgw, fibnum);
627 		}
628 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
629 #ifdef IPSEC
630 		key_sa_routechange((struct sockaddr *)&icmpsrc);
631 #endif
632 		break;
633 
634 	/*
635 	 * No kernel processing for the following;
636 	 * just fall through to send to raw listener.
637 	 */
638 	case ICMP_ECHOREPLY:
639 	case ICMP_ROUTERADVERT:
640 	case ICMP_ROUTERSOLICIT:
641 	case ICMP_TSTAMPREPLY:
642 	case ICMP_IREQREPLY:
643 	case ICMP_MASKREPLY:
644 	default:
645 		break;
646 	}
647 
648 raw:
649 	rip_input(m, off);
650 	return;
651 
652 freeit:
653 	m_freem(m);
654 }
655 
656 /*
657  * Reflect the ip packet back to the source
658  */
659 static void
660 icmp_reflect(struct mbuf *m)
661 {
662 	struct ip *ip = mtod(m, struct ip *);
663 	struct ifaddr *ifa;
664 	struct ifnet *ifp;
665 	struct in_ifaddr *ia;
666 	struct in_addr t;
667 	struct mbuf *opts = 0;
668 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
669 
670 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
671 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
672 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
673 		m_freem(m);	/* Bad return address */
674 		ICMPSTAT_INC(icps_badaddr);
675 		goto done;	/* Ip_output() will check for broadcast */
676 	}
677 
678 	m_addr_changed(m);
679 
680 	t = ip->ip_dst;
681 	ip->ip_dst = ip->ip_src;
682 
683 	/*
684 	 * Source selection for ICMP replies:
685 	 *
686 	 * If the incoming packet was addressed directly to one of our
687 	 * own addresses, use dst as the src for the reply.
688 	 */
689 	IN_IFADDR_RLOCK();
690 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
691 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
692 			t = IA_SIN(ia)->sin_addr;
693 			IN_IFADDR_RUNLOCK();
694 			goto match;
695 		}
696 	}
697 	IN_IFADDR_RUNLOCK();
698 
699 	/*
700 	 * If the incoming packet was addressed to one of our broadcast
701 	 * addresses, use the first non-broadcast address which corresponds
702 	 * to the incoming interface.
703 	 */
704 	ifp = m->m_pkthdr.rcvif;
705 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
706 		IF_ADDR_LOCK(ifp);
707 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
708 			if (ifa->ifa_addr->sa_family != AF_INET)
709 				continue;
710 			ia = ifatoia(ifa);
711 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
712 			    t.s_addr) {
713 				t = IA_SIN(ia)->sin_addr;
714 				IF_ADDR_UNLOCK(ifp);
715 				goto match;
716 			}
717 		}
718 		IF_ADDR_UNLOCK(ifp);
719 	}
720 	/*
721 	 * If the packet was transiting through us, use the address of
722 	 * the interface the packet came through in.  If that interface
723 	 * doesn't have a suitable IP address, the normal selection
724 	 * criteria apply.
725 	 */
726 	if (V_icmp_rfi && ifp != NULL) {
727 		IF_ADDR_LOCK(ifp);
728 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
729 			if (ifa->ifa_addr->sa_family != AF_INET)
730 				continue;
731 			ia = ifatoia(ifa);
732 			t = IA_SIN(ia)->sin_addr;
733 			IF_ADDR_UNLOCK(ifp);
734 			goto match;
735 		}
736 		IF_ADDR_UNLOCK(ifp);
737 	}
738 	/*
739 	 * If the incoming packet was not addressed directly to us, use
740 	 * designated interface for icmp replies specified by sysctl
741 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
742 	 * with normal source selection.
743 	 */
744 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
745 		IF_ADDR_LOCK(ifp);
746 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
747 			if (ifa->ifa_addr->sa_family != AF_INET)
748 				continue;
749 			ia = ifatoia(ifa);
750 			t = IA_SIN(ia)->sin_addr;
751 			IF_ADDR_UNLOCK(ifp);
752 			goto match;
753 		}
754 		IF_ADDR_UNLOCK(ifp);
755 	}
756 	/*
757 	 * If the packet was transiting through us, use the address of
758 	 * the interface that is the closest to the packet source.
759 	 * When we don't have a route back to the packet source, stop here
760 	 * and drop the packet.
761 	 */
762 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
763 	if (ia == NULL) {
764 		m_freem(m);
765 		ICMPSTAT_INC(icps_noroute);
766 		goto done;
767 	}
768 	t = IA_SIN(ia)->sin_addr;
769 	ifa_free(&ia->ia_ifa);
770 match:
771 #ifdef MAC
772 	mac_netinet_icmp_replyinplace(m);
773 #endif
774 	ip->ip_src = t;
775 	ip->ip_ttl = V_ip_defttl;
776 
777 	if (optlen > 0) {
778 		register u_char *cp;
779 		int opt, cnt;
780 		u_int len;
781 
782 		/*
783 		 * Retrieve any source routing from the incoming packet;
784 		 * add on any record-route or timestamp options.
785 		 */
786 		cp = (u_char *) (ip + 1);
787 		if ((opts = ip_srcroute(m)) == 0 &&
788 		    (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
789 			opts->m_len = sizeof(struct in_addr);
790 			mtod(opts, struct in_addr *)->s_addr = 0;
791 		}
792 		if (opts) {
793 #ifdef ICMPPRINTFS
794 		    if (icmpprintfs)
795 			    printf("icmp_reflect optlen %d rt %d => ",
796 				optlen, opts->m_len);
797 #endif
798 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
799 			    opt = cp[IPOPT_OPTVAL];
800 			    if (opt == IPOPT_EOL)
801 				    break;
802 			    if (opt == IPOPT_NOP)
803 				    len = 1;
804 			    else {
805 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
806 					    break;
807 				    len = cp[IPOPT_OLEN];
808 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
809 				        len > cnt)
810 					    break;
811 			    }
812 			    /*
813 			     * Should check for overflow, but it "can't happen"
814 			     */
815 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
816 				opt == IPOPT_SECURITY) {
817 				    bcopy((caddr_t)cp,
818 					mtod(opts, caddr_t) + opts->m_len, len);
819 				    opts->m_len += len;
820 			    }
821 		    }
822 		    /* Terminate & pad, if necessary */
823 		    cnt = opts->m_len % 4;
824 		    if (cnt) {
825 			    for (; cnt < 4; cnt++) {
826 				    *(mtod(opts, caddr_t) + opts->m_len) =
827 					IPOPT_EOL;
828 				    opts->m_len++;
829 			    }
830 		    }
831 #ifdef ICMPPRINTFS
832 		    if (icmpprintfs)
833 			    printf("%d\n", opts->m_len);
834 #endif
835 		}
836 		/*
837 		 * Now strip out original options by copying rest of first
838 		 * mbuf's data back, and adjust the IP length.
839 		 */
840 		ip->ip_len -= optlen;
841 		ip->ip_v = IPVERSION;
842 		ip->ip_hl = 5;
843 		m->m_len -= optlen;
844 		if (m->m_flags & M_PKTHDR)
845 			m->m_pkthdr.len -= optlen;
846 		optlen += sizeof(struct ip);
847 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
848 			 (unsigned)(m->m_len - sizeof(struct ip)));
849 	}
850 	m_tag_delete_nonpersistent(m);
851 	m->m_flags &= ~(M_BCAST|M_MCAST);
852 	icmp_send(m, opts);
853 done:
854 	if (opts)
855 		(void)m_free(opts);
856 }
857 
858 /*
859  * Send an icmp packet back to the ip level,
860  * after supplying a checksum.
861  */
862 static void
863 icmp_send(struct mbuf *m, struct mbuf *opts)
864 {
865 	register struct ip *ip = mtod(m, struct ip *);
866 	register int hlen;
867 	register struct icmp *icp;
868 
869 	hlen = ip->ip_hl << 2;
870 	m->m_data += hlen;
871 	m->m_len -= hlen;
872 	icp = mtod(m, struct icmp *);
873 	icp->icmp_cksum = 0;
874 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
875 	m->m_data -= hlen;
876 	m->m_len += hlen;
877 	m->m_pkthdr.rcvif = (struct ifnet *)0;
878 #ifdef ICMPPRINTFS
879 	if (icmpprintfs) {
880 		char buf[4 * sizeof "123"];
881 		strcpy(buf, inet_ntoa(ip->ip_dst));
882 		printf("icmp_send dst %s src %s\n",
883 		       buf, inet_ntoa(ip->ip_src));
884 	}
885 #endif
886 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
887 }
888 
889 /*
890  * Return milliseconds since 00:00 GMT in network format.
891  */
892 uint32_t
893 iptime(void)
894 {
895 	struct timeval atv;
896 	u_long t;
897 
898 	getmicrotime(&atv);
899 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
900 	return (htonl(t));
901 }
902 
903 /*
904  * Return the next larger or smaller MTU plateau (table from RFC 1191)
905  * given current value MTU.  If DIR is less than zero, a larger plateau
906  * is returned; otherwise, a smaller value is returned.
907  */
908 int
909 ip_next_mtu(int mtu, int dir)
910 {
911 	static int mtutab[] = {
912 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
913 		296, 68, 0
914 	};
915 	int i, size;
916 
917 	size = (sizeof mtutab) / (sizeof mtutab[0]);
918 	if (dir >= 0) {
919 		for (i = 0; i < size; i++)
920 			if (mtu > mtutab[i])
921 				return mtutab[i];
922 	} else {
923 		for (i = size - 1; i >= 0; i--)
924 			if (mtu < mtutab[i])
925 				return mtutab[i];
926 		if (mtu == mtutab[0])
927 			return mtutab[0];
928 	}
929 	return 0;
930 }
931 #endif /* INET */
932 
933 
934 /*
935  * badport_bandlim() - check for ICMP bandwidth limit
936  *
937  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
938  *	hit our bandwidth limit and it is not ok.
939  *
940  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
941  *
942  *	For now we separate the TCP and UDP subsystems w/ different 'which'
943  *	values.  We may eventually remove this separation (and simplify the
944  *	code further).
945  *
946  *	Note that the printing of the error message is delayed so we can
947  *	properly print the icmp error rate that the system was trying to do
948  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
949  *	the 'final' error, but it doesn't make sense to solve the printing
950  *	delay with more complex code.
951  */
952 
953 int
954 badport_bandlim(int which)
955 {
956 
957 #define	N(a)	(sizeof (a) / sizeof (a[0]))
958 	static struct rate {
959 		const char	*type;
960 		struct timeval	lasttime;
961 		int		curpps;
962 	} rates[BANDLIM_MAX+1] = {
963 		{ "icmp unreach response" },
964 		{ "icmp ping response" },
965 		{ "icmp tstamp response" },
966 		{ "closed port RST response" },
967 		{ "open port RST response" },
968 		{ "icmp6 unreach response" }
969 	};
970 
971 	/*
972 	 * Return ok status if feature disabled or argument out of range.
973 	 */
974 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
975 		struct rate *r = &rates[which];
976 		int opps = r->curpps;
977 
978 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
979 			return -1;	/* discard packet */
980 		/*
981 		 * If we've dropped below the threshold after having
982 		 * rate-limited traffic print the message.  This preserves
983 		 * the previous behaviour at the expense of added complexity.
984 		 */
985 		if (V_icmplim_output && opps > V_icmplim)
986 			log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
987 				r->type, opps, V_icmplim);
988 	}
989 	return 0;			/* okay to send packet */
990 #undef N
991 }
992