xref: /freebsd/sys/netinet/ip_icmp.c (revision 5861f9665471e98e544f6fa3ce73c4912229ff82)
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 #include <sys/vimage.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 #include <netinet/vinet.h>
64 
65 #ifdef IPSEC
66 #include <netipsec/ipsec.h>
67 #include <netipsec/key.h>
68 #endif
69 
70 #include <machine/in_cksum.h>
71 
72 #include <security/mac/mac_framework.h>
73 
74 /*
75  * ICMP routines: error generation, receive packet processing, and
76  * routines to turnaround packets back to the originator, and
77  * host table maintenance routines.
78  */
79 
80 #ifdef VIMAGE_GLOBALS
81 struct icmpstat	icmpstat;
82 static int	icmpmaskrepl;
83 static u_int	icmpmaskfake;
84 static int	drop_redirect;
85 static int	log_redirect;
86 static int	icmplim;
87 static int	icmplim_output;
88 static char	reply_src[IFNAMSIZ];
89 static int	icmp_rfi;
90 static int	icmp_quotelen;
91 static int	icmpbmcastecho;
92 #endif
93 
94 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_STATS, stats,
95 	CTLFLAG_RW, icmpstat, icmpstat, "");
96 
97 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_MASKREPL, maskrepl,
98 	CTLFLAG_RW, icmpmaskrepl, 0,
99 	"Reply to ICMP Address Mask Request packets.");
100 
101 SYSCTL_V_UINT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
102 	icmpmaskfake, 0, "Fake reply to ICMP Address Mask Request packets.");
103 
104 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, drop_redirect,
105 	CTLFLAG_RW, drop_redirect, 0, "Ignore ICMP redirects");
106 
107 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, log_redirect,
108 	CTLFLAG_RW, log_redirect, 0, "Log ICMP redirects to the console");
109 
110 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_ICMPLIM, icmplim,
111 	CTLFLAG_RW, icmplim, 0, "Maximum number of ICMP responses per second");
112 
113 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, icmplim_output,
114 	CTLFLAG_RW, icmplim_output, 0,
115 	"Enable rate limiting of ICMP responses");
116 
117 SYSCTL_V_STRING(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, reply_src,
118 	CTLFLAG_RW, reply_src, IFNAMSIZ,
119 	"icmp reply source for non-local packets.");
120 
121 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, reply_from_interface,
122 	CTLFLAG_RW, icmp_rfi, 0, "ICMP reply from incoming interface for "
123 	"non-local packets");
124 
125 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
126 	icmp_quotelen, 0, "Number of bytes from original packet to "
127 	"quote in ICMP reply");
128 
129 /*
130  * ICMP broadcast echo sysctl
131  */
132 
133 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, bmcastecho,
134 	CTLFLAG_RW, icmpbmcastecho, 0, "");
135 
136 
137 #ifdef ICMPPRINTFS
138 int	icmpprintfs = 0;
139 #endif
140 
141 static void	icmp_reflect(struct mbuf *);
142 static void	icmp_send(struct mbuf *, struct mbuf *);
143 
144 extern	struct protosw inetsw[];
145 
146 void
147 icmp_init(void)
148 {
149 	INIT_VNET_INET(curvnet);
150 
151 	V_icmpmaskrepl = 0;
152 	V_icmpmaskfake = 0;
153 	V_drop_redirect = 0;
154 	V_log_redirect = 0;
155 	V_icmplim = 200;
156 	V_icmplim_output = 1;
157 	V_icmp_rfi = 0;
158 	V_icmp_quotelen = 8;
159 	V_icmpbmcastecho = 0;
160 }
161 
162 /*
163  * Generate an error packet of type error
164  * in response to bad packet ip.
165  */
166 void
167 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
168 {
169 	INIT_VNET_INET(curvnet);
170 	register struct ip *oip = mtod(n, struct ip *), *nip;
171 	register unsigned oiphlen = oip->ip_hl << 2;
172 	register struct icmp *icp;
173 	register struct mbuf *m;
174 	unsigned icmplen, icmpelen, nlen;
175 
176 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
177 #ifdef ICMPPRINTFS
178 	if (icmpprintfs)
179 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
180 #endif
181 	if (type != ICMP_REDIRECT)
182 		ICMPSTAT_INC(icps_error);
183 	/*
184 	 * Don't send error:
185 	 *  if the original packet was encrypted.
186 	 *  if not the first fragment of message.
187 	 *  in response to a multicast or broadcast packet.
188 	 *  if the old packet protocol was an ICMP error message.
189 	 */
190 	if (n->m_flags & M_DECRYPTED)
191 		goto freeit;
192 	if (oip->ip_off & ~(IP_MF|IP_DF))
193 		goto freeit;
194 	if (n->m_flags & (M_BCAST|M_MCAST))
195 		goto freeit;
196 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
197 	  n->m_len >= oiphlen + ICMP_MINLEN &&
198 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
199 		ICMPSTAT_INC(icps_oldicmp);
200 		goto freeit;
201 	}
202 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
203 	if (oiphlen + 8 > n->m_len)
204 		goto freeit;
205 	/*
206 	 * Calculate length to quote from original packet and
207 	 * prevent the ICMP mbuf from overflowing.
208 	 * Unfortunatly this is non-trivial since ip_forward()
209 	 * sends us truncated packets.
210 	 */
211 	nlen = m_length(n, NULL);
212 	if (oip->ip_p == IPPROTO_TCP) {
213 		struct tcphdr *th;
214 		int tcphlen;
215 
216 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
217 		    n->m_next == NULL)
218 			goto stdreply;
219 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
220 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
221 			goto freeit;
222 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
223 		tcphlen = th->th_off << 2;
224 		if (tcphlen < sizeof(struct tcphdr))
225 			goto freeit;
226 		if (oip->ip_len < oiphlen + tcphlen)
227 			goto freeit;
228 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
229 			goto stdreply;
230 		if (n->m_len < oiphlen + tcphlen &&
231 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
232 			goto freeit;
233 		icmpelen = max(tcphlen, min(V_icmp_quotelen, oip->ip_len - oiphlen));
234 	} else
235 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, oip->ip_len - oiphlen));
236 
237 	icmplen = min(oiphlen + icmpelen, nlen);
238 	if (icmplen < sizeof(struct ip))
239 		goto freeit;
240 
241 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
242 		m = m_gethdr(M_DONTWAIT, MT_DATA);
243 	else
244 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
245 	if (m == NULL)
246 		goto freeit;
247 #ifdef MAC
248 	mac_netinet_icmp_reply(n, m);
249 #endif
250 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
251 	m_align(m, ICMP_MINLEN + icmplen);
252 	m->m_len = ICMP_MINLEN + icmplen;
253 
254 	/* XXX MRT  make the outgoing packet use the same FIB
255 	 * that was associated with the incoming packet
256 	 */
257 	M_SETFIB(m, M_GETFIB(n));
258 	icp = mtod(m, struct icmp *);
259 	ICMPSTAT_INC(icps_outhist[type]);
260 	icp->icmp_type = type;
261 	if (type == ICMP_REDIRECT)
262 		icp->icmp_gwaddr.s_addr = dest;
263 	else {
264 		icp->icmp_void = 0;
265 		/*
266 		 * The following assignments assume an overlay with the
267 		 * just zeroed icmp_void field.
268 		 */
269 		if (type == ICMP_PARAMPROB) {
270 			icp->icmp_pptr = code;
271 			code = 0;
272 		} else if (type == ICMP_UNREACH &&
273 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
274 			icp->icmp_nextmtu = htons(mtu);
275 		}
276 	}
277 	icp->icmp_code = code;
278 
279 	/*
280 	 * Copy the quotation into ICMP message and
281 	 * convert quoted IP header back to network representation.
282 	 */
283 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
284 	nip = &icp->icmp_ip;
285 	nip->ip_len = htons(nip->ip_len);
286 	nip->ip_off = htons(nip->ip_off);
287 
288 	/*
289 	 * Set up ICMP message mbuf and copy old IP header (without options
290 	 * in front of ICMP message.
291 	 * If the original mbuf was meant to bypass the firewall, the error
292 	 * reply should bypass as well.
293 	 */
294 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
295 	m->m_data -= sizeof(struct ip);
296 	m->m_len += sizeof(struct ip);
297 	m->m_pkthdr.len = m->m_len;
298 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
299 	nip = mtod(m, struct ip *);
300 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
301 	nip->ip_len = m->m_len;
302 	nip->ip_v = IPVERSION;
303 	nip->ip_hl = 5;
304 	nip->ip_p = IPPROTO_ICMP;
305 	nip->ip_tos = 0;
306 	icmp_reflect(m);
307 
308 freeit:
309 	m_freem(n);
310 }
311 
312 /*
313  * Process a received ICMP message.
314  */
315 void
316 icmp_input(struct mbuf *m, int off)
317 {
318 	INIT_VNET_INET(curvnet);
319 	struct icmp *icp;
320 	struct in_ifaddr *ia;
321 	struct ip *ip = mtod(m, struct ip *);
322 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
323 	int hlen = off;
324 	int icmplen = ip->ip_len;
325 	int i, code;
326 	void (*ctlfunc)(int, struct sockaddr *, void *);
327 	int fibnum;
328 
329 	/*
330 	 * Locate icmp structure in mbuf, and check
331 	 * that not corrupted and of at least minimum length.
332 	 */
333 #ifdef ICMPPRINTFS
334 	if (icmpprintfs) {
335 		char buf[4 * sizeof "123"];
336 		strcpy(buf, inet_ntoa(ip->ip_src));
337 		printf("icmp_input from %s to %s, len %d\n",
338 		       buf, inet_ntoa(ip->ip_dst), icmplen);
339 	}
340 #endif
341 	if (icmplen < ICMP_MINLEN) {
342 		ICMPSTAT_INC(icps_tooshort);
343 		goto freeit;
344 	}
345 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
346 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
347 		ICMPSTAT_INC(icps_tooshort);
348 		return;
349 	}
350 	ip = mtod(m, struct ip *);
351 	m->m_len -= hlen;
352 	m->m_data += hlen;
353 	icp = mtod(m, struct icmp *);
354 	if (in_cksum(m, icmplen)) {
355 		ICMPSTAT_INC(icps_checksum);
356 		goto freeit;
357 	}
358 	m->m_len += hlen;
359 	m->m_data -= hlen;
360 
361 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
362 		/*
363 		 * Deliver very specific ICMP type only.
364 		 */
365 		switch (icp->icmp_type) {
366 		case ICMP_UNREACH:
367 		case ICMP_TIMXCEED:
368 			break;
369 		default:
370 			goto freeit;
371 		}
372 	}
373 
374 #ifdef ICMPPRINTFS
375 	if (icmpprintfs)
376 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
377 		    icp->icmp_code);
378 #endif
379 
380 	/*
381 	 * Message type specific processing.
382 	 */
383 	if (icp->icmp_type > ICMP_MAXTYPE)
384 		goto raw;
385 
386 	/* Initialize */
387 	bzero(&icmpsrc, sizeof(icmpsrc));
388 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
389 	icmpsrc.sin_family = AF_INET;
390 	bzero(&icmpdst, sizeof(icmpdst));
391 	icmpdst.sin_len = sizeof(struct sockaddr_in);
392 	icmpdst.sin_family = AF_INET;
393 	bzero(&icmpgw, sizeof(icmpgw));
394 	icmpgw.sin_len = sizeof(struct sockaddr_in);
395 	icmpgw.sin_family = AF_INET;
396 
397 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
398 	code = icp->icmp_code;
399 	switch (icp->icmp_type) {
400 
401 	case ICMP_UNREACH:
402 		switch (code) {
403 			case ICMP_UNREACH_NET:
404 			case ICMP_UNREACH_HOST:
405 			case ICMP_UNREACH_SRCFAIL:
406 			case ICMP_UNREACH_NET_UNKNOWN:
407 			case ICMP_UNREACH_HOST_UNKNOWN:
408 			case ICMP_UNREACH_ISOLATED:
409 			case ICMP_UNREACH_TOSNET:
410 			case ICMP_UNREACH_TOSHOST:
411 			case ICMP_UNREACH_HOST_PRECEDENCE:
412 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
413 				code = PRC_UNREACH_NET;
414 				break;
415 
416 			case ICMP_UNREACH_NEEDFRAG:
417 				code = PRC_MSGSIZE;
418 				break;
419 
420 			/*
421 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
422 			 * Treat subcodes 2,3 as immediate RST
423 			 */
424 			case ICMP_UNREACH_PROTOCOL:
425 			case ICMP_UNREACH_PORT:
426 				code = PRC_UNREACH_PORT;
427 				break;
428 
429 			case ICMP_UNREACH_NET_PROHIB:
430 			case ICMP_UNREACH_HOST_PROHIB:
431 			case ICMP_UNREACH_FILTER_PROHIB:
432 				code = PRC_UNREACH_ADMIN_PROHIB;
433 				break;
434 
435 			default:
436 				goto badcode;
437 		}
438 		goto deliver;
439 
440 	case ICMP_TIMXCEED:
441 		if (code > 1)
442 			goto badcode;
443 		code += PRC_TIMXCEED_INTRANS;
444 		goto deliver;
445 
446 	case ICMP_PARAMPROB:
447 		if (code > 1)
448 			goto badcode;
449 		code = PRC_PARAMPROB;
450 		goto deliver;
451 
452 	case ICMP_SOURCEQUENCH:
453 		if (code)
454 			goto badcode;
455 		code = PRC_QUENCH;
456 	deliver:
457 		/*
458 		 * Problem with datagram; advise higher level routines.
459 		 */
460 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
461 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
462 			ICMPSTAT_INC(icps_badlen);
463 			goto freeit;
464 		}
465 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
466 		/* Discard ICMP's in response to multicast packets */
467 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
468 			goto badcode;
469 #ifdef ICMPPRINTFS
470 		if (icmpprintfs)
471 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
472 #endif
473 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
474 		/*
475 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
476 		 * notification to TCP layer.
477 		 */
478 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
479 		if (ctlfunc)
480 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
481 				   (void *)&icp->icmp_ip);
482 		break;
483 
484 	badcode:
485 		ICMPSTAT_INC(icps_badcode);
486 		break;
487 
488 	case ICMP_ECHO:
489 		if (!V_icmpbmcastecho
490 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
491 			ICMPSTAT_INC(icps_bmcastecho);
492 			break;
493 		}
494 		icp->icmp_type = ICMP_ECHOREPLY;
495 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
496 			goto freeit;
497 		else
498 			goto reflect;
499 
500 	case ICMP_TSTAMP:
501 		if (!V_icmpbmcastecho
502 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
503 			ICMPSTAT_INC(icps_bmcasttstamp);
504 			break;
505 		}
506 		if (icmplen < ICMP_TSLEN) {
507 			ICMPSTAT_INC(icps_badlen);
508 			break;
509 		}
510 		icp->icmp_type = ICMP_TSTAMPREPLY;
511 		icp->icmp_rtime = iptime();
512 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
513 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
514 			goto freeit;
515 		else
516 			goto reflect;
517 
518 	case ICMP_MASKREQ:
519 		if (V_icmpmaskrepl == 0)
520 			break;
521 		/*
522 		 * We are not able to respond with all ones broadcast
523 		 * unless we receive it over a point-to-point interface.
524 		 */
525 		if (icmplen < ICMP_MASKLEN)
526 			break;
527 		switch (ip->ip_dst.s_addr) {
528 
529 		case INADDR_BROADCAST:
530 		case INADDR_ANY:
531 			icmpdst.sin_addr = ip->ip_src;
532 			break;
533 
534 		default:
535 			icmpdst.sin_addr = ip->ip_dst;
536 		}
537 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
538 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
539 		if (ia == NULL)
540 			break;
541 		if (ia->ia_ifp == NULL) {
542 			ifa_free(&ia->ia_ifa);
543 			break;
544 		}
545 		icp->icmp_type = ICMP_MASKREPLY;
546 		if (V_icmpmaskfake == 0)
547 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
548 		else
549 			icp->icmp_mask = V_icmpmaskfake;
550 		if (ip->ip_src.s_addr == 0) {
551 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
552 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
553 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
554 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
555 		}
556 		ifa_free(&ia->ia_ifa);
557 reflect:
558 		ip->ip_len += hlen;	/* since ip_input deducts this */
559 		ICMPSTAT_INC(icps_reflect);
560 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
561 		icmp_reflect(m);
562 		return;
563 
564 	case ICMP_REDIRECT:
565 		if (V_log_redirect) {
566 			u_long src, dst, gw;
567 
568 			src = ntohl(ip->ip_src.s_addr);
569 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
570 			gw = ntohl(icp->icmp_gwaddr.s_addr);
571 			printf("icmp redirect from %d.%d.%d.%d: "
572 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
573 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
574 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
575 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
576 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
577 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
578 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
579 		}
580 		/*
581 		 * RFC1812 says we must ignore ICMP redirects if we
582 		 * are acting as router.
583 		 */
584 		if (V_drop_redirect || V_ipforwarding)
585 			break;
586 		if (code > 3)
587 			goto badcode;
588 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
589 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
590 			ICMPSTAT_INC(icps_badlen);
591 			break;
592 		}
593 		/*
594 		 * Short circuit routing redirects to force
595 		 * immediate change in the kernel's routing
596 		 * tables.  The message is also handed to anyone
597 		 * listening on a raw socket (e.g. the routing
598 		 * daemon for use in updating its tables).
599 		 */
600 		icmpgw.sin_addr = ip->ip_src;
601 		icmpdst.sin_addr = icp->icmp_gwaddr;
602 #ifdef	ICMPPRINTFS
603 		if (icmpprintfs) {
604 			char buf[4 * sizeof "123"];
605 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
606 
607 			printf("redirect dst %s to %s\n",
608 			       buf, inet_ntoa(icp->icmp_gwaddr));
609 		}
610 #endif
611 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
612 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
613 			in_rtredirect((struct sockaddr *)&icmpsrc,
614 			  (struct sockaddr *)&icmpdst,
615 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
616 			  (struct sockaddr *)&icmpgw, fibnum);
617 		}
618 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
619 #ifdef IPSEC
620 		key_sa_routechange((struct sockaddr *)&icmpsrc);
621 #endif
622 		break;
623 
624 	/*
625 	 * No kernel processing for the following;
626 	 * just fall through to send to raw listener.
627 	 */
628 	case ICMP_ECHOREPLY:
629 	case ICMP_ROUTERADVERT:
630 	case ICMP_ROUTERSOLICIT:
631 	case ICMP_TSTAMPREPLY:
632 	case ICMP_IREQREPLY:
633 	case ICMP_MASKREPLY:
634 	default:
635 		break;
636 	}
637 
638 raw:
639 	rip_input(m, off);
640 	return;
641 
642 freeit:
643 	m_freem(m);
644 }
645 
646 /*
647  * Reflect the ip packet back to the source
648  */
649 static void
650 icmp_reflect(struct mbuf *m)
651 {
652 	INIT_VNET_INET(curvnet);
653 	struct ip *ip = mtod(m, struct ip *);
654 	struct ifaddr *ifa;
655 	struct ifnet *ifp;
656 	struct in_ifaddr *ia;
657 	struct in_addr t;
658 	struct mbuf *opts = 0;
659 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
660 
661 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
662 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
663 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
664 		m_freem(m);	/* Bad return address */
665 		ICMPSTAT_INC(icps_badaddr);
666 		goto done;	/* Ip_output() will check for broadcast */
667 	}
668 
669 	t = ip->ip_dst;
670 	ip->ip_dst = ip->ip_src;
671 
672 	/*
673 	 * Source selection for ICMP replies:
674 	 *
675 	 * If the incoming packet was addressed directly to one of our
676 	 * own addresses, use dst as the src for the reply.
677 	 */
678 	IN_IFADDR_RLOCK();
679 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
680 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
681 			t = IA_SIN(ia)->sin_addr;
682 			IN_IFADDR_RUNLOCK();
683 			goto match;
684 		}
685 	}
686 	IN_IFADDR_RUNLOCK();
687 
688 	/*
689 	 * If the incoming packet was addressed to one of our broadcast
690 	 * addresses, use the first non-broadcast address which corresponds
691 	 * to the incoming interface.
692 	 */
693 	ifp = m->m_pkthdr.rcvif;
694 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
695 		IF_ADDR_LOCK(ifp);
696 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
697 			if (ifa->ifa_addr->sa_family != AF_INET)
698 				continue;
699 			ia = ifatoia(ifa);
700 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
701 			    t.s_addr) {
702 				t = IA_SIN(ia)->sin_addr;
703 				IF_ADDR_UNLOCK(ifp);
704 				goto match;
705 			}
706 		}
707 		IF_ADDR_UNLOCK(ifp);
708 	}
709 	/*
710 	 * If the packet was transiting through us, use the address of
711 	 * the interface the packet came through in.  If that interface
712 	 * doesn't have a suitable IP address, the normal selection
713 	 * criteria apply.
714 	 */
715 	if (V_icmp_rfi && ifp != NULL) {
716 		IF_ADDR_LOCK(ifp);
717 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
718 			if (ifa->ifa_addr->sa_family != AF_INET)
719 				continue;
720 			ia = ifatoia(ifa);
721 			t = IA_SIN(ia)->sin_addr;
722 			IF_ADDR_UNLOCK(ifp);
723 			goto match;
724 		}
725 		IF_ADDR_UNLOCK(ifp);
726 	}
727 	/*
728 	 * If the incoming packet was not addressed directly to us, use
729 	 * designated interface for icmp replies specified by sysctl
730 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
731 	 * with normal source selection.
732 	 */
733 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
734 		IF_ADDR_LOCK(ifp);
735 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
736 			if (ifa->ifa_addr->sa_family != AF_INET)
737 				continue;
738 			ia = ifatoia(ifa);
739 			t = IA_SIN(ia)->sin_addr;
740 			IF_ADDR_UNLOCK(ifp);
741 			goto match;
742 		}
743 		IF_ADDR_UNLOCK(ifp);
744 	}
745 	/*
746 	 * If the packet was transiting through us, use the address of
747 	 * the interface that is the closest to the packet source.
748 	 * When we don't have a route back to the packet source, stop here
749 	 * and drop the packet.
750 	 */
751 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
752 	if (ia == NULL) {
753 		m_freem(m);
754 		ICMPSTAT_INC(icps_noroute);
755 		goto done;
756 	}
757 	t = IA_SIN(ia)->sin_addr;
758 	ifa_free(&ia->ia_ifa);
759 match:
760 #ifdef MAC
761 	mac_netinet_icmp_replyinplace(m);
762 #endif
763 	ip->ip_src = t;
764 	ip->ip_ttl = V_ip_defttl;
765 
766 	if (optlen > 0) {
767 		register u_char *cp;
768 		int opt, cnt;
769 		u_int len;
770 
771 		/*
772 		 * Retrieve any source routing from the incoming packet;
773 		 * add on any record-route or timestamp options.
774 		 */
775 		cp = (u_char *) (ip + 1);
776 		if ((opts = ip_srcroute(m)) == 0 &&
777 		    (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
778 			opts->m_len = sizeof(struct in_addr);
779 			mtod(opts, struct in_addr *)->s_addr = 0;
780 		}
781 		if (opts) {
782 #ifdef ICMPPRINTFS
783 		    if (icmpprintfs)
784 			    printf("icmp_reflect optlen %d rt %d => ",
785 				optlen, opts->m_len);
786 #endif
787 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
788 			    opt = cp[IPOPT_OPTVAL];
789 			    if (opt == IPOPT_EOL)
790 				    break;
791 			    if (opt == IPOPT_NOP)
792 				    len = 1;
793 			    else {
794 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
795 					    break;
796 				    len = cp[IPOPT_OLEN];
797 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
798 				        len > cnt)
799 					    break;
800 			    }
801 			    /*
802 			     * Should check for overflow, but it "can't happen"
803 			     */
804 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
805 				opt == IPOPT_SECURITY) {
806 				    bcopy((caddr_t)cp,
807 					mtod(opts, caddr_t) + opts->m_len, len);
808 				    opts->m_len += len;
809 			    }
810 		    }
811 		    /* Terminate & pad, if necessary */
812 		    cnt = opts->m_len % 4;
813 		    if (cnt) {
814 			    for (; cnt < 4; cnt++) {
815 				    *(mtod(opts, caddr_t) + opts->m_len) =
816 					IPOPT_EOL;
817 				    opts->m_len++;
818 			    }
819 		    }
820 #ifdef ICMPPRINTFS
821 		    if (icmpprintfs)
822 			    printf("%d\n", opts->m_len);
823 #endif
824 		}
825 		/*
826 		 * Now strip out original options by copying rest of first
827 		 * mbuf's data back, and adjust the IP length.
828 		 */
829 		ip->ip_len -= optlen;
830 		ip->ip_v = IPVERSION;
831 		ip->ip_hl = 5;
832 		m->m_len -= optlen;
833 		if (m->m_flags & M_PKTHDR)
834 			m->m_pkthdr.len -= optlen;
835 		optlen += sizeof(struct ip);
836 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
837 			 (unsigned)(m->m_len - sizeof(struct ip)));
838 	}
839 	m_tag_delete_nonpersistent(m);
840 	m->m_flags &= ~(M_BCAST|M_MCAST);
841 	icmp_send(m, opts);
842 done:
843 	if (opts)
844 		(void)m_free(opts);
845 }
846 
847 /*
848  * Send an icmp packet back to the ip level,
849  * after supplying a checksum.
850  */
851 static void
852 icmp_send(struct mbuf *m, struct mbuf *opts)
853 {
854 	register struct ip *ip = mtod(m, struct ip *);
855 	register int hlen;
856 	register struct icmp *icp;
857 
858 	hlen = ip->ip_hl << 2;
859 	m->m_data += hlen;
860 	m->m_len -= hlen;
861 	icp = mtod(m, struct icmp *);
862 	icp->icmp_cksum = 0;
863 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
864 	m->m_data -= hlen;
865 	m->m_len += hlen;
866 	m->m_pkthdr.rcvif = (struct ifnet *)0;
867 #ifdef ICMPPRINTFS
868 	if (icmpprintfs) {
869 		char buf[4 * sizeof "123"];
870 		strcpy(buf, inet_ntoa(ip->ip_dst));
871 		printf("icmp_send dst %s src %s\n",
872 		       buf, inet_ntoa(ip->ip_src));
873 	}
874 #endif
875 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
876 }
877 
878 /*
879  * Return milliseconds since 00:00 GMT in network format.
880  */
881 uint32_t
882 iptime(void)
883 {
884 	struct timeval atv;
885 	u_long t;
886 
887 	getmicrotime(&atv);
888 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
889 	return (htonl(t));
890 }
891 
892 /*
893  * Return the next larger or smaller MTU plateau (table from RFC 1191)
894  * given current value MTU.  If DIR is less than zero, a larger plateau
895  * is returned; otherwise, a smaller value is returned.
896  */
897 int
898 ip_next_mtu(int mtu, int dir)
899 {
900 	static int mtutab[] = {
901 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
902 		296, 68, 0
903 	};
904 	int i, size;
905 
906 	size = (sizeof mtutab) / (sizeof mtutab[0]);
907 	if (dir >= 0) {
908 		for (i = 0; i < size; i++)
909 			if (mtu > mtutab[i])
910 				return mtutab[i];
911 	} else {
912 		for (i = size - 1; i >= 0; i--)
913 			if (mtu < mtutab[i])
914 				return mtutab[i];
915 		if (mtu == mtutab[0])
916 			return mtutab[0];
917 	}
918 	return 0;
919 }
920 
921 
922 /*
923  * badport_bandlim() - check for ICMP bandwidth limit
924  *
925  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
926  *	hit our bandwidth limit and it is not ok.
927  *
928  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
929  *
930  *	For now we separate the TCP and UDP subsystems w/ different 'which'
931  *	values.  We may eventually remove this separation (and simplify the
932  *	code further).
933  *
934  *	Note that the printing of the error message is delayed so we can
935  *	properly print the icmp error rate that the system was trying to do
936  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
937  *	the 'final' error, but it doesn't make sense to solve the printing
938  *	delay with more complex code.
939  */
940 
941 int
942 badport_bandlim(int which)
943 {
944 	INIT_VNET_INET(curvnet);
945 
946 #define	N(a)	(sizeof (a) / sizeof (a[0]))
947 	static struct rate {
948 		const char	*type;
949 		struct timeval	lasttime;
950 		int		curpps;
951 	} rates[BANDLIM_MAX+1] = {
952 		{ "icmp unreach response" },
953 		{ "icmp ping response" },
954 		{ "icmp tstamp response" },
955 		{ "closed port RST response" },
956 		{ "open port RST response" },
957 		{ "icmp6 unreach response" }
958 	};
959 
960 	/*
961 	 * Return ok status if feature disabled or argument out of range.
962 	 */
963 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
964 		struct rate *r = &rates[which];
965 		int opps = r->curpps;
966 
967 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
968 			return -1;	/* discard packet */
969 		/*
970 		 * If we've dropped below the threshold after having
971 		 * rate-limited traffic print the message.  This preserves
972 		 * the previous behaviour at the expense of added complexity.
973 		 */
974 		if (V_icmplim_output && opps > V_icmplim)
975 			printf("Limiting %s from %d to %d packets/sec\n",
976 				r->type, opps, V_icmplim);
977 	}
978 	return 0;			/* okay to send packet */
979 #undef N
980 }
981