xref: /freebsd/sys/netinet/ip_icmp.c (revision 9336e0699bda8a301cd2bfa37106b6ec5e32012e)
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 #include "opt_mac.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 
47 #include <net/if.h>
48 #include <net/if_types.h>
49 #include <net/route.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 
79 struct	icmpstat icmpstat;
80 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
81 	&icmpstat, icmpstat, "");
82 
83 static int	icmpmaskrepl = 0;
84 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
85 	&icmpmaskrepl, 0, "Reply to ICMP Address Mask Request packets.");
86 
87 static u_int	icmpmaskfake = 0;
88 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
89 	&icmpmaskfake, 0, "Fake reply to ICMP Address Mask Request packets.");
90 
91 static int	drop_redirect = 0;
92 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
93 	&drop_redirect, 0, "Ignore ICMP redirects");
94 
95 static int	log_redirect = 0;
96 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
97 	&log_redirect, 0, "Log ICMP redirects to the console");
98 
99 static int      icmplim = 200;
100 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
101 	&icmplim, 0, "Maximum number of ICMP responses per second");
102 
103 static int	icmplim_output = 1;
104 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
105 	&icmplim_output, 0, "Enable rate limiting of ICMP responses");
106 
107 static char	reply_src[IFNAMSIZ];
108 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
109 	&reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
110 
111 static int	icmp_rfi = 0;
112 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
113 	&icmp_rfi, 0, "ICMP reply from incoming interface for "
114 	"non-local packets");
115 
116 static int	icmp_quotelen = 8;
117 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
118 	&icmp_quotelen, 0, "Number of bytes from original packet to "
119 	"quote in ICMP reply");
120 
121 /*
122  * ICMP broadcast echo sysctl
123  */
124 
125 static int	icmpbmcastecho = 0;
126 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
127 	&icmpbmcastecho, 0, "");
128 
129 
130 #ifdef ICMPPRINTFS
131 int	icmpprintfs = 0;
132 #endif
133 
134 static void	icmp_reflect(struct mbuf *);
135 static void	icmp_send(struct mbuf *, struct mbuf *);
136 
137 extern	struct protosw inetsw[];
138 
139 /*
140  * Generate an error packet of type error
141  * in response to bad packet ip.
142  */
143 void
144 icmp_error(struct mbuf *n, int type, int code, n_long dest, int mtu)
145 {
146 	register struct ip *oip = mtod(n, struct ip *), *nip;
147 	register unsigned oiphlen = oip->ip_hl << 2;
148 	register struct icmp *icp;
149 	register struct mbuf *m;
150 	unsigned icmplen, icmpelen, nlen;
151 
152 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
153 #ifdef ICMPPRINTFS
154 	if (icmpprintfs)
155 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
156 #endif
157 	if (type != ICMP_REDIRECT)
158 		icmpstat.icps_error++;
159 	/*
160 	 * Don't send error:
161 	 *  if the original packet was encrypted.
162 	 *  if not the first fragment of message.
163 	 *  in response to a multicast or broadcast packet.
164 	 *  if the old packet protocol was an ICMP error message.
165 	 */
166 	if (n->m_flags & M_DECRYPTED)
167 		goto freeit;
168 	if (oip->ip_off & ~(IP_MF|IP_DF))
169 		goto freeit;
170 	if (n->m_flags & (M_BCAST|M_MCAST))
171 		goto freeit;
172 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
173 	  n->m_len >= oiphlen + ICMP_MINLEN &&
174 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
175 		icmpstat.icps_oldicmp++;
176 		goto freeit;
177 	}
178 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
179 	if (oiphlen + 8 > n->m_len)
180 		goto freeit;
181 	/*
182 	 * Calculate length to quote from original packet and
183 	 * prevent the ICMP mbuf from overflowing.
184 	 * Unfortunatly this is non-trivial since ip_forward()
185 	 * sends us truncated packets.
186 	 */
187 	nlen = m_length(n, NULL);
188 	if (oip->ip_p == IPPROTO_TCP) {
189 		struct tcphdr *th;
190 		int tcphlen;
191 
192 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
193 		    n->m_next == NULL)
194 			goto stdreply;
195 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
196 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
197 			goto freeit;
198 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
199 		tcphlen = th->th_off << 2;
200 		if (tcphlen < sizeof(struct tcphdr))
201 			goto freeit;
202 		if (oip->ip_len < oiphlen + tcphlen)
203 			goto freeit;
204 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
205 			goto stdreply;
206 		if (n->m_len < oiphlen + tcphlen &&
207 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
208 			goto freeit;
209 		icmpelen = max(tcphlen, min(icmp_quotelen, oip->ip_len - oiphlen));
210 	} else
211 stdreply:	icmpelen = max(8, min(icmp_quotelen, oip->ip_len - oiphlen));
212 
213 	icmplen = min(oiphlen + icmpelen, nlen);
214 	if (icmplen < sizeof(struct ip))
215 		goto freeit;
216 
217 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
218 		m = m_gethdr(M_DONTWAIT, MT_DATA);
219 	else
220 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
221 	if (m == NULL)
222 		goto freeit;
223 #ifdef MAC
224 	mac_netinet_icmp_reply(n, m);
225 #endif
226 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
227 	m_align(m, ICMP_MINLEN + icmplen);
228 	m->m_len = ICMP_MINLEN + icmplen;
229 
230 	icp = mtod(m, struct icmp *);
231 	icmpstat.icps_outhist[type]++;
232 	icp->icmp_type = type;
233 	if (type == ICMP_REDIRECT)
234 		icp->icmp_gwaddr.s_addr = dest;
235 	else {
236 		icp->icmp_void = 0;
237 		/*
238 		 * The following assignments assume an overlay with the
239 		 * just zeroed icmp_void field.
240 		 */
241 		if (type == ICMP_PARAMPROB) {
242 			icp->icmp_pptr = code;
243 			code = 0;
244 		} else if (type == ICMP_UNREACH &&
245 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
246 			icp->icmp_nextmtu = htons(mtu);
247 		}
248 	}
249 	icp->icmp_code = code;
250 
251 	/*
252 	 * Copy the quotation into ICMP message and
253 	 * convert quoted IP header back to network representation.
254 	 */
255 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
256 	nip = &icp->icmp_ip;
257 	nip->ip_len = htons(nip->ip_len);
258 	nip->ip_off = htons(nip->ip_off);
259 
260 	/*
261 	 * Set up ICMP message mbuf and copy old IP header (without options
262 	 * in front of ICMP message.
263 	 * If the original mbuf was meant to bypass the firewall, the error
264 	 * reply should bypass as well.
265 	 */
266 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
267 	m->m_data -= sizeof(struct ip);
268 	m->m_len += sizeof(struct ip);
269 	m->m_pkthdr.len = m->m_len;
270 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
271 	nip = mtod(m, struct ip *);
272 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
273 	nip->ip_len = m->m_len;
274 	nip->ip_v = IPVERSION;
275 	nip->ip_hl = 5;
276 	nip->ip_p = IPPROTO_ICMP;
277 	nip->ip_tos = 0;
278 	icmp_reflect(m);
279 
280 freeit:
281 	m_freem(n);
282 }
283 
284 /*
285  * Process a received ICMP message.
286  */
287 void
288 icmp_input(struct mbuf *m, int off)
289 {
290 	struct icmp *icp;
291 	struct in_ifaddr *ia;
292 	struct ip *ip = mtod(m, struct ip *);
293 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
294 	int hlen = off;
295 	int icmplen = ip->ip_len;
296 	int i, code;
297 	void (*ctlfunc)(int, struct sockaddr *, void *);
298 
299 	/*
300 	 * Locate icmp structure in mbuf, and check
301 	 * that not corrupted and of at least minimum length.
302 	 */
303 #ifdef ICMPPRINTFS
304 	if (icmpprintfs) {
305 		char buf[4 * sizeof "123"];
306 		strcpy(buf, inet_ntoa(ip->ip_src));
307 		printf("icmp_input from %s to %s, len %d\n",
308 		       buf, inet_ntoa(ip->ip_dst), icmplen);
309 	}
310 #endif
311 	if (icmplen < ICMP_MINLEN) {
312 		icmpstat.icps_tooshort++;
313 		goto freeit;
314 	}
315 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
316 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
317 		icmpstat.icps_tooshort++;
318 		return;
319 	}
320 	ip = mtod(m, struct ip *);
321 	m->m_len -= hlen;
322 	m->m_data += hlen;
323 	icp = mtod(m, struct icmp *);
324 	if (in_cksum(m, icmplen)) {
325 		icmpstat.icps_checksum++;
326 		goto freeit;
327 	}
328 	m->m_len += hlen;
329 	m->m_data -= hlen;
330 
331 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
332 		/*
333 		 * Deliver very specific ICMP type only.
334 		 */
335 		switch (icp->icmp_type) {
336 		case ICMP_UNREACH:
337 		case ICMP_TIMXCEED:
338 			break;
339 		default:
340 			goto freeit;
341 		}
342 	}
343 
344 #ifdef ICMPPRINTFS
345 	if (icmpprintfs)
346 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
347 		    icp->icmp_code);
348 #endif
349 
350 	/*
351 	 * Message type specific processing.
352 	 */
353 	if (icp->icmp_type > ICMP_MAXTYPE)
354 		goto raw;
355 
356 	/* Initialize */
357 	bzero(&icmpsrc, sizeof(icmpsrc));
358 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
359 	icmpsrc.sin_family = AF_INET;
360 	bzero(&icmpdst, sizeof(icmpdst));
361 	icmpdst.sin_len = sizeof(struct sockaddr_in);
362 	icmpdst.sin_family = AF_INET;
363 	bzero(&icmpgw, sizeof(icmpgw));
364 	icmpgw.sin_len = sizeof(struct sockaddr_in);
365 	icmpgw.sin_family = AF_INET;
366 
367 	icmpstat.icps_inhist[icp->icmp_type]++;
368 	code = icp->icmp_code;
369 	switch (icp->icmp_type) {
370 
371 	case ICMP_UNREACH:
372 		switch (code) {
373 			case ICMP_UNREACH_NET:
374 			case ICMP_UNREACH_HOST:
375 			case ICMP_UNREACH_SRCFAIL:
376 			case ICMP_UNREACH_NET_UNKNOWN:
377 			case ICMP_UNREACH_HOST_UNKNOWN:
378 			case ICMP_UNREACH_ISOLATED:
379 			case ICMP_UNREACH_TOSNET:
380 			case ICMP_UNREACH_TOSHOST:
381 			case ICMP_UNREACH_HOST_PRECEDENCE:
382 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
383 				code = PRC_UNREACH_NET;
384 				break;
385 
386 			case ICMP_UNREACH_NEEDFRAG:
387 				code = PRC_MSGSIZE;
388 				break;
389 
390 			/*
391 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
392 			 * Treat subcodes 2,3 as immediate RST
393 			 */
394 			case ICMP_UNREACH_PROTOCOL:
395 			case ICMP_UNREACH_PORT:
396 				code = PRC_UNREACH_PORT;
397 				break;
398 
399 			case ICMP_UNREACH_NET_PROHIB:
400 			case ICMP_UNREACH_HOST_PROHIB:
401 			case ICMP_UNREACH_FILTER_PROHIB:
402 				code = PRC_UNREACH_ADMIN_PROHIB;
403 				break;
404 
405 			default:
406 				goto badcode;
407 		}
408 		goto deliver;
409 
410 	case ICMP_TIMXCEED:
411 		if (code > 1)
412 			goto badcode;
413 		code += PRC_TIMXCEED_INTRANS;
414 		goto deliver;
415 
416 	case ICMP_PARAMPROB:
417 		if (code > 1)
418 			goto badcode;
419 		code = PRC_PARAMPROB;
420 		goto deliver;
421 
422 	case ICMP_SOURCEQUENCH:
423 		if (code)
424 			goto badcode;
425 		code = PRC_QUENCH;
426 	deliver:
427 		/*
428 		 * Problem with datagram; advise higher level routines.
429 		 */
430 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
431 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
432 			icmpstat.icps_badlen++;
433 			goto freeit;
434 		}
435 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
436 		/* Discard ICMP's in response to multicast packets */
437 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
438 			goto badcode;
439 #ifdef ICMPPRINTFS
440 		if (icmpprintfs)
441 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
442 #endif
443 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
444 		/*
445 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
446 		 * notification to TCP layer.
447 		 */
448 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
449 		if (ctlfunc)
450 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
451 				   (void *)&icp->icmp_ip);
452 		break;
453 
454 	badcode:
455 		icmpstat.icps_badcode++;
456 		break;
457 
458 	case ICMP_ECHO:
459 		if (!icmpbmcastecho
460 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
461 			icmpstat.icps_bmcastecho++;
462 			break;
463 		}
464 		icp->icmp_type = ICMP_ECHOREPLY;
465 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
466 			goto freeit;
467 		else
468 			goto reflect;
469 
470 	case ICMP_TSTAMP:
471 		if (!icmpbmcastecho
472 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
473 			icmpstat.icps_bmcasttstamp++;
474 			break;
475 		}
476 		if (icmplen < ICMP_TSLEN) {
477 			icmpstat.icps_badlen++;
478 			break;
479 		}
480 		icp->icmp_type = ICMP_TSTAMPREPLY;
481 		icp->icmp_rtime = iptime();
482 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
483 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
484 			goto freeit;
485 		else
486 			goto reflect;
487 
488 	case ICMP_MASKREQ:
489 		if (icmpmaskrepl == 0)
490 			break;
491 		/*
492 		 * We are not able to respond with all ones broadcast
493 		 * unless we receive it over a point-to-point interface.
494 		 */
495 		if (icmplen < ICMP_MASKLEN)
496 			break;
497 		switch (ip->ip_dst.s_addr) {
498 
499 		case INADDR_BROADCAST:
500 		case INADDR_ANY:
501 			icmpdst.sin_addr = ip->ip_src;
502 			break;
503 
504 		default:
505 			icmpdst.sin_addr = ip->ip_dst;
506 		}
507 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
508 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
509 		if (ia == 0)
510 			break;
511 		if (ia->ia_ifp == 0)
512 			break;
513 		icp->icmp_type = ICMP_MASKREPLY;
514 		if (icmpmaskfake == 0)
515 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
516 		else
517 			icp->icmp_mask = icmpmaskfake;
518 		if (ip->ip_src.s_addr == 0) {
519 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
520 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
521 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
522 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
523 		}
524 reflect:
525 		ip->ip_len += hlen;	/* since ip_input deducts this */
526 		icmpstat.icps_reflect++;
527 		icmpstat.icps_outhist[icp->icmp_type]++;
528 		icmp_reflect(m);
529 		return;
530 
531 	case ICMP_REDIRECT:
532 		if (log_redirect) {
533 			u_long src, dst, gw;
534 
535 			src = ntohl(ip->ip_src.s_addr);
536 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
537 			gw = ntohl(icp->icmp_gwaddr.s_addr);
538 			printf("icmp redirect from %d.%d.%d.%d: "
539 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
540 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
541 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
542 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
543 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
544 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
545 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
546 		}
547 		/*
548 		 * RFC1812 says we must ignore ICMP redirects if we
549 		 * are acting as router.
550 		 */
551 		if (drop_redirect || ipforwarding)
552 			break;
553 		if (code > 3)
554 			goto badcode;
555 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
556 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
557 			icmpstat.icps_badlen++;
558 			break;
559 		}
560 		/*
561 		 * Short circuit routing redirects to force
562 		 * immediate change in the kernel's routing
563 		 * tables.  The message is also handed to anyone
564 		 * listening on a raw socket (e.g. the routing
565 		 * daemon for use in updating its tables).
566 		 */
567 		icmpgw.sin_addr = ip->ip_src;
568 		icmpdst.sin_addr = icp->icmp_gwaddr;
569 #ifdef	ICMPPRINTFS
570 		if (icmpprintfs) {
571 			char buf[4 * sizeof "123"];
572 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
573 
574 			printf("redirect dst %s to %s\n",
575 			       buf, inet_ntoa(icp->icmp_gwaddr));
576 		}
577 #endif
578 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
579 		rtredirect((struct sockaddr *)&icmpsrc,
580 		  (struct sockaddr *)&icmpdst,
581 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
582 		  (struct sockaddr *)&icmpgw);
583 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
584 #ifdef IPSEC
585 		key_sa_routechange((struct sockaddr *)&icmpsrc);
586 #endif
587 		break;
588 
589 	/*
590 	 * No kernel processing for the following;
591 	 * just fall through to send to raw listener.
592 	 */
593 	case ICMP_ECHOREPLY:
594 	case ICMP_ROUTERADVERT:
595 	case ICMP_ROUTERSOLICIT:
596 	case ICMP_TSTAMPREPLY:
597 	case ICMP_IREQREPLY:
598 	case ICMP_MASKREPLY:
599 	default:
600 		break;
601 	}
602 
603 raw:
604 	rip_input(m, off);
605 	return;
606 
607 freeit:
608 	m_freem(m);
609 }
610 
611 /*
612  * Reflect the ip packet back to the source
613  */
614 static void
615 icmp_reflect(struct mbuf *m)
616 {
617 	struct ip *ip = mtod(m, struct ip *);
618 	struct ifaddr *ifa;
619 	struct ifnet *ifn;
620 	struct in_ifaddr *ia;
621 	struct in_addr t;
622 	struct mbuf *opts = 0;
623 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
624 
625 	if (!in_canforward(ip->ip_src) &&
626 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
627 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
628 		m_freem(m);	/* Bad return address */
629 		icmpstat.icps_badaddr++;
630 		goto done;	/* Ip_output() will check for broadcast */
631 	}
632 	t = ip->ip_dst;
633 	ip->ip_dst = ip->ip_src;
634 
635 	/*
636 	 * Source selection for ICMP replies:
637 	 *
638 	 * If the incoming packet was addressed directly to one of our
639 	 * own addresses, use dst as the src for the reply.
640 	 */
641 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
642 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
643 			goto match;
644 	/*
645 	 * If the incoming packet was addressed to one of our broadcast
646 	 * addresses, use the first non-broadcast address which corresponds
647 	 * to the incoming interface.
648 	 */
649 	if (m->m_pkthdr.rcvif != NULL &&
650 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
651 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
652 			if (ifa->ifa_addr->sa_family != AF_INET)
653 				continue;
654 			ia = ifatoia(ifa);
655 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
656 			    t.s_addr)
657 				goto match;
658 		}
659 	}
660 	/*
661 	 * If the packet was transiting through us, use the address of
662 	 * the interface the packet came through in.  If that interface
663 	 * doesn't have a suitable IP address, the normal selection
664 	 * criteria apply.
665 	 */
666 	if (icmp_rfi && m->m_pkthdr.rcvif != NULL) {
667 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
668 			if (ifa->ifa_addr->sa_family != AF_INET)
669 				continue;
670 			ia = ifatoia(ifa);
671 			goto match;
672 		}
673 	}
674 	/*
675 	 * If the incoming packet was not addressed directly to us, use
676 	 * designated interface for icmp replies specified by sysctl
677 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
678 	 * with normal source selection.
679 	 */
680 	if (reply_src[0] != '\0' && (ifn = ifunit(reply_src))) {
681 		TAILQ_FOREACH(ifa, &ifn->if_addrhead, ifa_link) {
682 			if (ifa->ifa_addr->sa_family != AF_INET)
683 				continue;
684 			ia = ifatoia(ifa);
685 			goto match;
686 		}
687 	}
688 	/*
689 	 * If the packet was transiting through us, use the address of
690 	 * the interface that is the closest to the packet source.
691 	 * When we don't have a route back to the packet source, stop here
692 	 * and drop the packet.
693 	 */
694 	ia = ip_rtaddr(ip->ip_dst);
695 	if (ia == NULL) {
696 		m_freem(m);
697 		icmpstat.icps_noroute++;
698 		goto done;
699 	}
700 match:
701 #ifdef MAC
702 	mac_netinet_icmp_replyinplace(m);
703 #endif
704 	t = IA_SIN(ia)->sin_addr;
705 	ip->ip_src = t;
706 	ip->ip_ttl = ip_defttl;
707 
708 	if (optlen > 0) {
709 		register u_char *cp;
710 		int opt, cnt;
711 		u_int len;
712 
713 		/*
714 		 * Retrieve any source routing from the incoming packet;
715 		 * add on any record-route or timestamp options.
716 		 */
717 		cp = (u_char *) (ip + 1);
718 		if ((opts = ip_srcroute(m)) == 0 &&
719 		    (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
720 			opts->m_len = sizeof(struct in_addr);
721 			mtod(opts, struct in_addr *)->s_addr = 0;
722 		}
723 		if (opts) {
724 #ifdef ICMPPRINTFS
725 		    if (icmpprintfs)
726 			    printf("icmp_reflect optlen %d rt %d => ",
727 				optlen, opts->m_len);
728 #endif
729 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
730 			    opt = cp[IPOPT_OPTVAL];
731 			    if (opt == IPOPT_EOL)
732 				    break;
733 			    if (opt == IPOPT_NOP)
734 				    len = 1;
735 			    else {
736 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
737 					    break;
738 				    len = cp[IPOPT_OLEN];
739 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
740 				        len > cnt)
741 					    break;
742 			    }
743 			    /*
744 			     * Should check for overflow, but it "can't happen"
745 			     */
746 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
747 				opt == IPOPT_SECURITY) {
748 				    bcopy((caddr_t)cp,
749 					mtod(opts, caddr_t) + opts->m_len, len);
750 				    opts->m_len += len;
751 			    }
752 		    }
753 		    /* Terminate & pad, if necessary */
754 		    cnt = opts->m_len % 4;
755 		    if (cnt) {
756 			    for (; cnt < 4; cnt++) {
757 				    *(mtod(opts, caddr_t) + opts->m_len) =
758 					IPOPT_EOL;
759 				    opts->m_len++;
760 			    }
761 		    }
762 #ifdef ICMPPRINTFS
763 		    if (icmpprintfs)
764 			    printf("%d\n", opts->m_len);
765 #endif
766 		}
767 		/*
768 		 * Now strip out original options by copying rest of first
769 		 * mbuf's data back, and adjust the IP length.
770 		 */
771 		ip->ip_len -= optlen;
772 		ip->ip_v = IPVERSION;
773 		ip->ip_hl = 5;
774 		m->m_len -= optlen;
775 		if (m->m_flags & M_PKTHDR)
776 			m->m_pkthdr.len -= optlen;
777 		optlen += sizeof(struct ip);
778 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
779 			 (unsigned)(m->m_len - sizeof(struct ip)));
780 	}
781 	m_tag_delete_nonpersistent(m);
782 	m->m_flags &= ~(M_BCAST|M_MCAST);
783 	icmp_send(m, opts);
784 done:
785 	if (opts)
786 		(void)m_free(opts);
787 }
788 
789 /*
790  * Send an icmp packet back to the ip level,
791  * after supplying a checksum.
792  */
793 static void
794 icmp_send(struct mbuf *m, struct mbuf *opts)
795 {
796 	register struct ip *ip = mtod(m, struct ip *);
797 	register int hlen;
798 	register struct icmp *icp;
799 
800 	hlen = ip->ip_hl << 2;
801 	m->m_data += hlen;
802 	m->m_len -= hlen;
803 	icp = mtod(m, struct icmp *);
804 	icp->icmp_cksum = 0;
805 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
806 	m->m_data -= hlen;
807 	m->m_len += hlen;
808 	m->m_pkthdr.rcvif = (struct ifnet *)0;
809 #ifdef ICMPPRINTFS
810 	if (icmpprintfs) {
811 		char buf[4 * sizeof "123"];
812 		strcpy(buf, inet_ntoa(ip->ip_dst));
813 		printf("icmp_send dst %s src %s\n",
814 		       buf, inet_ntoa(ip->ip_src));
815 	}
816 #endif
817 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
818 }
819 
820 n_time
821 iptime(void)
822 {
823 	struct timeval atv;
824 	u_long t;
825 
826 	getmicrotime(&atv);
827 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
828 	return (htonl(t));
829 }
830 
831 /*
832  * Return the next larger or smaller MTU plateau (table from RFC 1191)
833  * given current value MTU.  If DIR is less than zero, a larger plateau
834  * is returned; otherwise, a smaller value is returned.
835  */
836 int
837 ip_next_mtu(int mtu, int dir)
838 {
839 	static int mtutab[] = {
840 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
841 		296, 68, 0
842 	};
843 	int i, size;
844 
845 	size = (sizeof mtutab) / (sizeof mtutab[0]);
846 	if (dir >= 0) {
847 		for (i = 0; i < size; i++)
848 			if (mtu > mtutab[i])
849 				return mtutab[i];
850 	} else {
851 		for (i = size - 1; i >= 0; i--)
852 			if (mtu < mtutab[i])
853 				return mtutab[i];
854 		if (mtu == mtutab[0])
855 			return mtutab[0];
856 	}
857 	return 0;
858 }
859 
860 
861 /*
862  * badport_bandlim() - check for ICMP bandwidth limit
863  *
864  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
865  *	hit our bandwidth limit and it is not ok.
866  *
867  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
868  *
869  *	For now we separate the TCP and UDP subsystems w/ different 'which'
870  *	values.  We may eventually remove this separation (and simplify the
871  *	code further).
872  *
873  *	Note that the printing of the error message is delayed so we can
874  *	properly print the icmp error rate that the system was trying to do
875  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
876  *	the 'final' error, but it doesn't make sense to solve the printing
877  *	delay with more complex code.
878  */
879 
880 int
881 badport_bandlim(int which)
882 {
883 #define	N(a)	(sizeof (a) / sizeof (a[0]))
884 	static struct rate {
885 		const char	*type;
886 		struct timeval	lasttime;
887 		int		curpps;
888 	} rates[BANDLIM_MAX+1] = {
889 		{ "icmp unreach response" },
890 		{ "icmp ping response" },
891 		{ "icmp tstamp response" },
892 		{ "closed port RST response" },
893 		{ "open port RST response" },
894 		{ "icmp6 unreach response" }
895 	};
896 
897 	/*
898 	 * Return ok status if feature disabled or argument out of range.
899 	 */
900 	if (icmplim > 0 && (u_int) which < N(rates)) {
901 		struct rate *r = &rates[which];
902 		int opps = r->curpps;
903 
904 		if (!ppsratecheck(&r->lasttime, &r->curpps, icmplim))
905 			return -1;	/* discard packet */
906 		/*
907 		 * If we've dropped below the threshold after having
908 		 * rate-limited traffic print the message.  This preserves
909 		 * the previous behaviour at the expense of added complexity.
910 		 */
911 		if (icmplim_output && opps > icmplim)
912 			printf("Limiting %s from %d to %d packets/sec\n",
913 				r->type, opps, icmplim);
914 	}
915 	return 0;			/* okay to send packet */
916 #undef N
917 }
918