xref: /freebsd/sys/netinet/ip_icmp.c (revision aa64588d28258aef88cc33b8043112e8856948d0)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_ipsec.h"
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mbuf.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/sysctl.h>
45 
46 #include <net/if.h>
47 #include <net/if_types.h>
48 #include <net/route.h>
49 #include <net/vnet.h>
50 
51 #include <netinet/in.h>
52 #include <netinet/in_pcb.h>
53 #include <netinet/in_systm.h>
54 #include <netinet/in_var.h>
55 #include <netinet/ip.h>
56 #include <netinet/ip_icmp.h>
57 #include <netinet/ip_var.h>
58 #include <netinet/ip_options.h>
59 #include <netinet/tcp.h>
60 #include <netinet/tcp_var.h>
61 #include <netinet/tcpip.h>
62 #include <netinet/icmp_var.h>
63 
64 #ifdef IPSEC
65 #include <netipsec/ipsec.h>
66 #include <netipsec/key.h>
67 #endif
68 
69 #include <machine/in_cksum.h>
70 
71 #include <security/mac/mac_framework.h>
72 
73 /*
74  * ICMP routines: error generation, receive packet processing, and
75  * routines to turnaround packets back to the originator, and
76  * host table maintenance routines.
77  */
78 VNET_DEFINE(struct icmpstat, icmpstat);
79 SYSCTL_VNET_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
80 	&VNET_NAME(icmpstat), icmpstat, "");
81 
82 static VNET_DEFINE(int, icmpmaskrepl) = 0;
83 #define	V_icmpmaskrepl			VNET(icmpmaskrepl)
84 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
85 	&VNET_NAME(icmpmaskrepl), 0,
86 	"Reply to ICMP Address Mask Request packets.");
87 
88 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
89 #define	V_icmpmaskfake			VNET(icmpmaskfake)
90 SYSCTL_VNET_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
91 	&VNET_NAME(icmpmaskfake), 0,
92 	"Fake reply to ICMP Address Mask Request packets.");
93 
94 static VNET_DEFINE(int, drop_redirect) = 0;
95 #define	V_drop_redirect			VNET(drop_redirect)
96 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
97 	&VNET_NAME(drop_redirect), 0,
98 	"Ignore ICMP redirects");
99 
100 static VNET_DEFINE(int, log_redirect) = 0;
101 #define	V_log_redirect			VNET(log_redirect)
102 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
103 	&VNET_NAME(log_redirect), 0,
104 	"Log ICMP redirects to the console");
105 
106 static VNET_DEFINE(int, icmplim) = 200;
107 #define	V_icmplim			VNET(icmplim)
108 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
109 	&VNET_NAME(icmplim), 0,
110 	"Maximum number of ICMP responses per second");
111 
112 static VNET_DEFINE(int, icmplim_output) = 1;
113 #define	V_icmplim_output		VNET(icmplim_output)
114 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
115 	&VNET_NAME(icmplim_output), 0,
116 	"Enable rate limiting of ICMP responses");
117 
118 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
119 #define	V_reply_src			VNET(reply_src)
120 SYSCTL_VNET_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
121 	&VNET_NAME(reply_src), IFNAMSIZ,
122 	"icmp reply source for non-local packets.");
123 
124 static VNET_DEFINE(int, icmp_rfi) = 0;
125 #define	V_icmp_rfi			VNET(icmp_rfi)
126 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
127 	&VNET_NAME(icmp_rfi), 0,
128 	"ICMP reply from incoming interface for non-local packets");
129 
130 static VNET_DEFINE(int, icmp_quotelen) = 8;
131 #define	V_icmp_quotelen			VNET(icmp_quotelen)
132 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
133 	&VNET_NAME(icmp_quotelen), 0,
134 	"Number of bytes from original packet to quote in ICMP reply");
135 
136 /*
137  * ICMP broadcast echo sysctl
138  */
139 static VNET_DEFINE(int, icmpbmcastecho) = 0;
140 #define	V_icmpbmcastecho		VNET(icmpbmcastecho)
141 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
142 	&VNET_NAME(icmpbmcastecho), 0,
143 	"");
144 
145 
146 #ifdef ICMPPRINTFS
147 int	icmpprintfs = 0;
148 #endif
149 
150 static void	icmp_reflect(struct mbuf *);
151 static void	icmp_send(struct mbuf *, struct mbuf *);
152 
153 extern	struct protosw inetsw[];
154 
155 /*
156  * Kernel module interface for updating icmpstat.  The argument is an index
157  * into icmpstat treated as an array of u_long.  While this encodes the
158  * general layout of icmpstat into the caller, it doesn't encode its
159  * location, so that future changes to add, for example, per-CPU stats
160  * support won't cause binary compatibility problems for kernel modules.
161  */
162 void
163 kmod_icmpstat_inc(int statnum)
164 {
165 
166 	(*((u_long *)&V_icmpstat + statnum))++;
167 }
168 
169 /*
170  * Generate an error packet of type error
171  * in response to bad packet ip.
172  */
173 void
174 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
175 {
176 	register struct ip *oip = mtod(n, struct ip *), *nip;
177 	register unsigned oiphlen = oip->ip_hl << 2;
178 	register struct icmp *icp;
179 	register struct mbuf *m;
180 	unsigned icmplen, icmpelen, nlen;
181 
182 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
183 #ifdef ICMPPRINTFS
184 	if (icmpprintfs)
185 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
186 #endif
187 	if (type != ICMP_REDIRECT)
188 		ICMPSTAT_INC(icps_error);
189 	/*
190 	 * Don't send error:
191 	 *  if the original packet was encrypted.
192 	 *  if not the first fragment of message.
193 	 *  in response to a multicast or broadcast packet.
194 	 *  if the old packet protocol was an ICMP error message.
195 	 */
196 	if (n->m_flags & M_DECRYPTED)
197 		goto freeit;
198 	if (oip->ip_off & ~(IP_MF|IP_DF))
199 		goto freeit;
200 	if (n->m_flags & (M_BCAST|M_MCAST))
201 		goto freeit;
202 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
203 	  n->m_len >= oiphlen + ICMP_MINLEN &&
204 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
205 		ICMPSTAT_INC(icps_oldicmp);
206 		goto freeit;
207 	}
208 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
209 	if (oiphlen + 8 > n->m_len)
210 		goto freeit;
211 	/*
212 	 * Calculate length to quote from original packet and
213 	 * prevent the ICMP mbuf from overflowing.
214 	 * Unfortunatly this is non-trivial since ip_forward()
215 	 * sends us truncated packets.
216 	 */
217 	nlen = m_length(n, NULL);
218 	if (oip->ip_p == IPPROTO_TCP) {
219 		struct tcphdr *th;
220 		int tcphlen;
221 
222 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
223 		    n->m_next == NULL)
224 			goto stdreply;
225 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
226 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
227 			goto freeit;
228 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
229 		tcphlen = th->th_off << 2;
230 		if (tcphlen < sizeof(struct tcphdr))
231 			goto freeit;
232 		if (oip->ip_len < oiphlen + tcphlen)
233 			goto freeit;
234 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
235 			goto stdreply;
236 		if (n->m_len < oiphlen + tcphlen &&
237 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
238 			goto freeit;
239 		icmpelen = max(tcphlen, min(V_icmp_quotelen, oip->ip_len - oiphlen));
240 	} else
241 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, oip->ip_len - oiphlen));
242 
243 	icmplen = min(oiphlen + icmpelen, nlen);
244 	if (icmplen < sizeof(struct ip))
245 		goto freeit;
246 
247 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
248 		m = m_gethdr(M_DONTWAIT, MT_DATA);
249 	else
250 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
251 	if (m == NULL)
252 		goto freeit;
253 #ifdef MAC
254 	mac_netinet_icmp_reply(n, m);
255 #endif
256 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
257 	m_align(m, ICMP_MINLEN + icmplen);
258 	m->m_len = ICMP_MINLEN + icmplen;
259 
260 	/* XXX MRT  make the outgoing packet use the same FIB
261 	 * that was associated with the incoming packet
262 	 */
263 	M_SETFIB(m, M_GETFIB(n));
264 	icp = mtod(m, struct icmp *);
265 	ICMPSTAT_INC(icps_outhist[type]);
266 	icp->icmp_type = type;
267 	if (type == ICMP_REDIRECT)
268 		icp->icmp_gwaddr.s_addr = dest;
269 	else {
270 		icp->icmp_void = 0;
271 		/*
272 		 * The following assignments assume an overlay with the
273 		 * just zeroed icmp_void field.
274 		 */
275 		if (type == ICMP_PARAMPROB) {
276 			icp->icmp_pptr = code;
277 			code = 0;
278 		} else if (type == ICMP_UNREACH &&
279 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
280 			icp->icmp_nextmtu = htons(mtu);
281 		}
282 	}
283 	icp->icmp_code = code;
284 
285 	/*
286 	 * Copy the quotation into ICMP message and
287 	 * convert quoted IP header back to network representation.
288 	 */
289 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
290 	nip = &icp->icmp_ip;
291 	nip->ip_len = htons(nip->ip_len);
292 	nip->ip_off = htons(nip->ip_off);
293 
294 	/*
295 	 * Set up ICMP message mbuf and copy old IP header (without options
296 	 * in front of ICMP message.
297 	 * If the original mbuf was meant to bypass the firewall, the error
298 	 * reply should bypass as well.
299 	 */
300 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
301 	m->m_data -= sizeof(struct ip);
302 	m->m_len += sizeof(struct ip);
303 	m->m_pkthdr.len = m->m_len;
304 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
305 	nip = mtod(m, struct ip *);
306 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
307 	nip->ip_len = m->m_len;
308 	nip->ip_v = IPVERSION;
309 	nip->ip_hl = 5;
310 	nip->ip_p = IPPROTO_ICMP;
311 	nip->ip_tos = 0;
312 	icmp_reflect(m);
313 
314 freeit:
315 	m_freem(n);
316 }
317 
318 /*
319  * Process a received ICMP message.
320  */
321 void
322 icmp_input(struct mbuf *m, int off)
323 {
324 	struct icmp *icp;
325 	struct in_ifaddr *ia;
326 	struct ip *ip = mtod(m, struct ip *);
327 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
328 	int hlen = off;
329 	int icmplen = ip->ip_len;
330 	int i, code;
331 	void (*ctlfunc)(int, struct sockaddr *, void *);
332 	int fibnum;
333 
334 	/*
335 	 * Locate icmp structure in mbuf, and check
336 	 * that not corrupted and of at least minimum length.
337 	 */
338 #ifdef ICMPPRINTFS
339 	if (icmpprintfs) {
340 		char buf[4 * sizeof "123"];
341 		strcpy(buf, inet_ntoa(ip->ip_src));
342 		printf("icmp_input from %s to %s, len %d\n",
343 		       buf, inet_ntoa(ip->ip_dst), icmplen);
344 	}
345 #endif
346 	if (icmplen < ICMP_MINLEN) {
347 		ICMPSTAT_INC(icps_tooshort);
348 		goto freeit;
349 	}
350 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
351 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
352 		ICMPSTAT_INC(icps_tooshort);
353 		return;
354 	}
355 	ip = mtod(m, struct ip *);
356 	m->m_len -= hlen;
357 	m->m_data += hlen;
358 	icp = mtod(m, struct icmp *);
359 	if (in_cksum(m, icmplen)) {
360 		ICMPSTAT_INC(icps_checksum);
361 		goto freeit;
362 	}
363 	m->m_len += hlen;
364 	m->m_data -= hlen;
365 
366 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
367 		/*
368 		 * Deliver very specific ICMP type only.
369 		 */
370 		switch (icp->icmp_type) {
371 		case ICMP_UNREACH:
372 		case ICMP_TIMXCEED:
373 			break;
374 		default:
375 			goto freeit;
376 		}
377 	}
378 
379 #ifdef ICMPPRINTFS
380 	if (icmpprintfs)
381 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
382 		    icp->icmp_code);
383 #endif
384 
385 	/*
386 	 * Message type specific processing.
387 	 */
388 	if (icp->icmp_type > ICMP_MAXTYPE)
389 		goto raw;
390 
391 	/* Initialize */
392 	bzero(&icmpsrc, sizeof(icmpsrc));
393 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
394 	icmpsrc.sin_family = AF_INET;
395 	bzero(&icmpdst, sizeof(icmpdst));
396 	icmpdst.sin_len = sizeof(struct sockaddr_in);
397 	icmpdst.sin_family = AF_INET;
398 	bzero(&icmpgw, sizeof(icmpgw));
399 	icmpgw.sin_len = sizeof(struct sockaddr_in);
400 	icmpgw.sin_family = AF_INET;
401 
402 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
403 	code = icp->icmp_code;
404 	switch (icp->icmp_type) {
405 
406 	case ICMP_UNREACH:
407 		switch (code) {
408 			case ICMP_UNREACH_NET:
409 			case ICMP_UNREACH_HOST:
410 			case ICMP_UNREACH_SRCFAIL:
411 			case ICMP_UNREACH_NET_UNKNOWN:
412 			case ICMP_UNREACH_HOST_UNKNOWN:
413 			case ICMP_UNREACH_ISOLATED:
414 			case ICMP_UNREACH_TOSNET:
415 			case ICMP_UNREACH_TOSHOST:
416 			case ICMP_UNREACH_HOST_PRECEDENCE:
417 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
418 				code = PRC_UNREACH_NET;
419 				break;
420 
421 			case ICMP_UNREACH_NEEDFRAG:
422 				code = PRC_MSGSIZE;
423 				break;
424 
425 			/*
426 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
427 			 * Treat subcodes 2,3 as immediate RST
428 			 */
429 			case ICMP_UNREACH_PROTOCOL:
430 			case ICMP_UNREACH_PORT:
431 				code = PRC_UNREACH_PORT;
432 				break;
433 
434 			case ICMP_UNREACH_NET_PROHIB:
435 			case ICMP_UNREACH_HOST_PROHIB:
436 			case ICMP_UNREACH_FILTER_PROHIB:
437 				code = PRC_UNREACH_ADMIN_PROHIB;
438 				break;
439 
440 			default:
441 				goto badcode;
442 		}
443 		goto deliver;
444 
445 	case ICMP_TIMXCEED:
446 		if (code > 1)
447 			goto badcode;
448 		code += PRC_TIMXCEED_INTRANS;
449 		goto deliver;
450 
451 	case ICMP_PARAMPROB:
452 		if (code > 1)
453 			goto badcode;
454 		code = PRC_PARAMPROB;
455 		goto deliver;
456 
457 	case ICMP_SOURCEQUENCH:
458 		if (code)
459 			goto badcode;
460 		code = PRC_QUENCH;
461 	deliver:
462 		/*
463 		 * Problem with datagram; advise higher level routines.
464 		 */
465 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
466 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
467 			ICMPSTAT_INC(icps_badlen);
468 			goto freeit;
469 		}
470 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
471 		/* Discard ICMP's in response to multicast packets */
472 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
473 			goto badcode;
474 #ifdef ICMPPRINTFS
475 		if (icmpprintfs)
476 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
477 #endif
478 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
479 		/*
480 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
481 		 * notification to TCP layer.
482 		 */
483 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
484 		if (ctlfunc)
485 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
486 				   (void *)&icp->icmp_ip);
487 		break;
488 
489 	badcode:
490 		ICMPSTAT_INC(icps_badcode);
491 		break;
492 
493 	case ICMP_ECHO:
494 		if (!V_icmpbmcastecho
495 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
496 			ICMPSTAT_INC(icps_bmcastecho);
497 			break;
498 		}
499 		icp->icmp_type = ICMP_ECHOREPLY;
500 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
501 			goto freeit;
502 		else
503 			goto reflect;
504 
505 	case ICMP_TSTAMP:
506 		if (!V_icmpbmcastecho
507 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
508 			ICMPSTAT_INC(icps_bmcasttstamp);
509 			break;
510 		}
511 		if (icmplen < ICMP_TSLEN) {
512 			ICMPSTAT_INC(icps_badlen);
513 			break;
514 		}
515 		icp->icmp_type = ICMP_TSTAMPREPLY;
516 		icp->icmp_rtime = iptime();
517 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
518 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
519 			goto freeit;
520 		else
521 			goto reflect;
522 
523 	case ICMP_MASKREQ:
524 		if (V_icmpmaskrepl == 0)
525 			break;
526 		/*
527 		 * We are not able to respond with all ones broadcast
528 		 * unless we receive it over a point-to-point interface.
529 		 */
530 		if (icmplen < ICMP_MASKLEN)
531 			break;
532 		switch (ip->ip_dst.s_addr) {
533 
534 		case INADDR_BROADCAST:
535 		case INADDR_ANY:
536 			icmpdst.sin_addr = ip->ip_src;
537 			break;
538 
539 		default:
540 			icmpdst.sin_addr = ip->ip_dst;
541 		}
542 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
543 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
544 		if (ia == NULL)
545 			break;
546 		if (ia->ia_ifp == NULL) {
547 			ifa_free(&ia->ia_ifa);
548 			break;
549 		}
550 		icp->icmp_type = ICMP_MASKREPLY;
551 		if (V_icmpmaskfake == 0)
552 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
553 		else
554 			icp->icmp_mask = V_icmpmaskfake;
555 		if (ip->ip_src.s_addr == 0) {
556 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
557 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
558 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
559 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
560 		}
561 		ifa_free(&ia->ia_ifa);
562 reflect:
563 		ip->ip_len += hlen;	/* since ip_input deducts this */
564 		ICMPSTAT_INC(icps_reflect);
565 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
566 		icmp_reflect(m);
567 		return;
568 
569 	case ICMP_REDIRECT:
570 		if (V_log_redirect) {
571 			u_long src, dst, gw;
572 
573 			src = ntohl(ip->ip_src.s_addr);
574 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
575 			gw = ntohl(icp->icmp_gwaddr.s_addr);
576 			printf("icmp redirect from %d.%d.%d.%d: "
577 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
578 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
579 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
580 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
581 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
582 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
583 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
584 		}
585 		/*
586 		 * RFC1812 says we must ignore ICMP redirects if we
587 		 * are acting as router.
588 		 */
589 		if (V_drop_redirect || V_ipforwarding)
590 			break;
591 		if (code > 3)
592 			goto badcode;
593 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
594 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
595 			ICMPSTAT_INC(icps_badlen);
596 			break;
597 		}
598 		/*
599 		 * Short circuit routing redirects to force
600 		 * immediate change in the kernel's routing
601 		 * tables.  The message is also handed to anyone
602 		 * listening on a raw socket (e.g. the routing
603 		 * daemon for use in updating its tables).
604 		 */
605 		icmpgw.sin_addr = ip->ip_src;
606 		icmpdst.sin_addr = icp->icmp_gwaddr;
607 #ifdef	ICMPPRINTFS
608 		if (icmpprintfs) {
609 			char buf[4 * sizeof "123"];
610 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
611 
612 			printf("redirect dst %s to %s\n",
613 			       buf, inet_ntoa(icp->icmp_gwaddr));
614 		}
615 #endif
616 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
617 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
618 			in_rtredirect((struct sockaddr *)&icmpsrc,
619 			  (struct sockaddr *)&icmpdst,
620 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
621 			  (struct sockaddr *)&icmpgw, fibnum);
622 		}
623 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
624 #ifdef IPSEC
625 		key_sa_routechange((struct sockaddr *)&icmpsrc);
626 #endif
627 		break;
628 
629 	/*
630 	 * No kernel processing for the following;
631 	 * just fall through to send to raw listener.
632 	 */
633 	case ICMP_ECHOREPLY:
634 	case ICMP_ROUTERADVERT:
635 	case ICMP_ROUTERSOLICIT:
636 	case ICMP_TSTAMPREPLY:
637 	case ICMP_IREQREPLY:
638 	case ICMP_MASKREPLY:
639 	default:
640 		break;
641 	}
642 
643 raw:
644 	rip_input(m, off);
645 	return;
646 
647 freeit:
648 	m_freem(m);
649 }
650 
651 /*
652  * Reflect the ip packet back to the source
653  */
654 static void
655 icmp_reflect(struct mbuf *m)
656 {
657 	struct ip *ip = mtod(m, struct ip *);
658 	struct ifaddr *ifa;
659 	struct ifnet *ifp;
660 	struct in_ifaddr *ia;
661 	struct in_addr t;
662 	struct mbuf *opts = 0;
663 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
664 
665 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
666 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
667 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
668 		m_freem(m);	/* Bad return address */
669 		ICMPSTAT_INC(icps_badaddr);
670 		goto done;	/* Ip_output() will check for broadcast */
671 	}
672 
673 	t = ip->ip_dst;
674 	ip->ip_dst = ip->ip_src;
675 
676 	/*
677 	 * Source selection for ICMP replies:
678 	 *
679 	 * If the incoming packet was addressed directly to one of our
680 	 * own addresses, use dst as the src for the reply.
681 	 */
682 	IN_IFADDR_RLOCK();
683 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
684 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
685 			t = IA_SIN(ia)->sin_addr;
686 			IN_IFADDR_RUNLOCK();
687 			goto match;
688 		}
689 	}
690 	IN_IFADDR_RUNLOCK();
691 
692 	/*
693 	 * If the incoming packet was addressed to one of our broadcast
694 	 * addresses, use the first non-broadcast address which corresponds
695 	 * to the incoming interface.
696 	 */
697 	ifp = m->m_pkthdr.rcvif;
698 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
699 		IF_ADDR_LOCK(ifp);
700 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
701 			if (ifa->ifa_addr->sa_family != AF_INET)
702 				continue;
703 			ia = ifatoia(ifa);
704 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
705 			    t.s_addr) {
706 				t = IA_SIN(ia)->sin_addr;
707 				IF_ADDR_UNLOCK(ifp);
708 				goto match;
709 			}
710 		}
711 		IF_ADDR_UNLOCK(ifp);
712 	}
713 	/*
714 	 * If the packet was transiting through us, use the address of
715 	 * the interface the packet came through in.  If that interface
716 	 * doesn't have a suitable IP address, the normal selection
717 	 * criteria apply.
718 	 */
719 	if (V_icmp_rfi && ifp != NULL) {
720 		IF_ADDR_LOCK(ifp);
721 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
722 			if (ifa->ifa_addr->sa_family != AF_INET)
723 				continue;
724 			ia = ifatoia(ifa);
725 			t = IA_SIN(ia)->sin_addr;
726 			IF_ADDR_UNLOCK(ifp);
727 			goto match;
728 		}
729 		IF_ADDR_UNLOCK(ifp);
730 	}
731 	/*
732 	 * If the incoming packet was not addressed directly to us, use
733 	 * designated interface for icmp replies specified by sysctl
734 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
735 	 * with normal source selection.
736 	 */
737 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
738 		IF_ADDR_LOCK(ifp);
739 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
740 			if (ifa->ifa_addr->sa_family != AF_INET)
741 				continue;
742 			ia = ifatoia(ifa);
743 			t = IA_SIN(ia)->sin_addr;
744 			IF_ADDR_UNLOCK(ifp);
745 			goto match;
746 		}
747 		IF_ADDR_UNLOCK(ifp);
748 	}
749 	/*
750 	 * If the packet was transiting through us, use the address of
751 	 * the interface that is the closest to the packet source.
752 	 * When we don't have a route back to the packet source, stop here
753 	 * and drop the packet.
754 	 */
755 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
756 	if (ia == NULL) {
757 		m_freem(m);
758 		ICMPSTAT_INC(icps_noroute);
759 		goto done;
760 	}
761 	t = IA_SIN(ia)->sin_addr;
762 	ifa_free(&ia->ia_ifa);
763 match:
764 #ifdef MAC
765 	mac_netinet_icmp_replyinplace(m);
766 #endif
767 	ip->ip_src = t;
768 	ip->ip_ttl = V_ip_defttl;
769 
770 	if (optlen > 0) {
771 		register u_char *cp;
772 		int opt, cnt;
773 		u_int len;
774 
775 		/*
776 		 * Retrieve any source routing from the incoming packet;
777 		 * add on any record-route or timestamp options.
778 		 */
779 		cp = (u_char *) (ip + 1);
780 		if ((opts = ip_srcroute(m)) == 0 &&
781 		    (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
782 			opts->m_len = sizeof(struct in_addr);
783 			mtod(opts, struct in_addr *)->s_addr = 0;
784 		}
785 		if (opts) {
786 #ifdef ICMPPRINTFS
787 		    if (icmpprintfs)
788 			    printf("icmp_reflect optlen %d rt %d => ",
789 				optlen, opts->m_len);
790 #endif
791 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
792 			    opt = cp[IPOPT_OPTVAL];
793 			    if (opt == IPOPT_EOL)
794 				    break;
795 			    if (opt == IPOPT_NOP)
796 				    len = 1;
797 			    else {
798 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
799 					    break;
800 				    len = cp[IPOPT_OLEN];
801 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
802 				        len > cnt)
803 					    break;
804 			    }
805 			    /*
806 			     * Should check for overflow, but it "can't happen"
807 			     */
808 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
809 				opt == IPOPT_SECURITY) {
810 				    bcopy((caddr_t)cp,
811 					mtod(opts, caddr_t) + opts->m_len, len);
812 				    opts->m_len += len;
813 			    }
814 		    }
815 		    /* Terminate & pad, if necessary */
816 		    cnt = opts->m_len % 4;
817 		    if (cnt) {
818 			    for (; cnt < 4; cnt++) {
819 				    *(mtod(opts, caddr_t) + opts->m_len) =
820 					IPOPT_EOL;
821 				    opts->m_len++;
822 			    }
823 		    }
824 #ifdef ICMPPRINTFS
825 		    if (icmpprintfs)
826 			    printf("%d\n", opts->m_len);
827 #endif
828 		}
829 		/*
830 		 * Now strip out original options by copying rest of first
831 		 * mbuf's data back, and adjust the IP length.
832 		 */
833 		ip->ip_len -= optlen;
834 		ip->ip_v = IPVERSION;
835 		ip->ip_hl = 5;
836 		m->m_len -= optlen;
837 		if (m->m_flags & M_PKTHDR)
838 			m->m_pkthdr.len -= optlen;
839 		optlen += sizeof(struct ip);
840 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
841 			 (unsigned)(m->m_len - sizeof(struct ip)));
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, 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 GMT 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 
925 
926 /*
927  * badport_bandlim() - check for ICMP bandwidth limit
928  *
929  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
930  *	hit our bandwidth limit and it is not ok.
931  *
932  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
933  *
934  *	For now we separate the TCP and UDP subsystems w/ different 'which'
935  *	values.  We may eventually remove this separation (and simplify the
936  *	code further).
937  *
938  *	Note that the printing of the error message is delayed so we can
939  *	properly print the icmp error rate that the system was trying to do
940  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
941  *	the 'final' error, but it doesn't make sense to solve the printing
942  *	delay with more complex code.
943  */
944 
945 int
946 badport_bandlim(int which)
947 {
948 
949 #define	N(a)	(sizeof (a) / sizeof (a[0]))
950 	static struct rate {
951 		const char	*type;
952 		struct timeval	lasttime;
953 		int		curpps;
954 	} rates[BANDLIM_MAX+1] = {
955 		{ "icmp unreach response" },
956 		{ "icmp ping response" },
957 		{ "icmp tstamp response" },
958 		{ "closed port RST response" },
959 		{ "open port RST response" },
960 		{ "icmp6 unreach response" }
961 	};
962 
963 	/*
964 	 * Return ok status if feature disabled or argument out of range.
965 	 */
966 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
967 		struct rate *r = &rates[which];
968 		int opps = r->curpps;
969 
970 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
971 			return -1;	/* discard packet */
972 		/*
973 		 * If we've dropped below the threshold after having
974 		 * rate-limited traffic print the message.  This preserves
975 		 * the previous behaviour at the expense of added complexity.
976 		 */
977 		if (V_icmplim_output && opps > V_icmplim)
978 			printf("Limiting %s from %d to %d packets/sec\n",
979 				r->type, opps, V_icmplim);
980 	}
981 	return 0;			/* okay to send packet */
982 #undef N
983 }
984