xref: /dragonfly/sys/netinet/ip_icmp.c (revision 25a2db75)
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  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
34  * $FreeBSD: src/sys/netinet/ip_icmp.c,v 1.39.2.19 2003/01/24 05:11:34 sam Exp $
35  */
36 
37 #include "opt_ipsec.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/mbuf.h>
42 #include <sys/protosw.h>
43 #include <sys/socket.h>
44 #include <sys/socketops.h>
45 #include <sys/time.h>
46 #include <sys/kernel.h>
47 #include <sys/sysctl.h>
48 #include <sys/in_cksum.h>
49 
50 #include <machine/stdarg.h>
51 
52 #include <net/if.h>
53 #include <net/if_types.h>
54 #include <net/route.h>
55 
56 #define _IP_VHL
57 #include <netinet/in.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip.h>
61 #include <netinet/ip_icmp.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/icmp_var.h>
64 
65 #ifdef IPSEC
66 #include <netinet6/ipsec.h>
67 #include <netproto/key/key.h>
68 #endif
69 
70 #ifdef FAST_IPSEC
71 #include <netproto/ipsec/ipsec.h>
72 #include <netproto/ipsec/key.h>
73 #define	IPSEC
74 #endif
75 
76 /*
77  * ICMP routines: error generation, receive packet processing, and
78  * routines to turnaround packets back to the originator, and
79  * host table maintenance routines.
80  */
81 
82 struct icmpstat icmpstat;
83 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
84 	&icmpstat, icmpstat, "ICMP statistics");
85 
86 static int	icmpmaskrepl = 0;
87 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
88 	&icmpmaskrepl, 0, "Allow replies to netmask requests");
89 
90 static int	drop_redirect = 0;
91 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
92 	&drop_redirect, 0, "Ignore ICMP redirects");
93 
94 static int	log_redirect = 0;
95 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
96 	&log_redirect, 0, "Enable output about ICMP redirects");
97 
98 static int	discard_sourcequench = 1;
99 SYSCTL_INT(_net_inet_icmp, OID_AUTO, discard_sourcequench, CTLFLAG_RW,
100 	&discard_sourcequench, 0, "Discard ICMP Source Quench");
101 
102 #ifdef ICMP_BANDLIM
103 
104 /*
105  * ICMP error-response bandwidth limiting sysctl.  If not enabled, sysctl
106  *      variable content is -1 and read-only.
107  */
108 
109 static int      icmplim = 200;
110 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
111 	&icmplim, 0, "ICMP bandwidth limit");
112 #else
113 
114 static int      icmplim = -1;
115 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
116 	&icmplim, 0, "ICMP bandwidth limit");
117 
118 #endif
119 
120 static int	icmplim_output = 0;
121 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
122 	&icmplim_output, 0, "Enable output about ICMP bandwidth limits");
123 
124 /*
125  * ICMP broadcast echo sysctl
126  */
127 
128 static int	icmpbmcastecho = 0;
129 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
130     &icmpbmcastecho, 0, "");
131 
132 static char	icmp_reply_src[IFNAMSIZ];
133 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
134 	icmp_reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
135 
136 static int	icmp_rfi;
137 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
138 	&icmp_rfi, 0, "ICMP reply from incoming interface for "
139 	"non-local packets");
140 
141 #ifdef ICMPPRINTFS
142 int	icmpprintfs = 0;
143 #endif
144 
145 static void	icmp_reflect (struct mbuf *);
146 static void	icmp_send (struct mbuf *, struct mbuf *, struct route *);
147 
148 extern	struct protosw inetsw[];
149 
150 /*
151  * Generate an error packet of type error
152  * in response to bad packet ip.
153  */
154 void
155 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu)
156 {
157 	struct ip *oip = mtod(n, struct ip *), *nip;
158 	unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
159 	struct icmp *icp;
160 	struct mbuf *m;
161 	unsigned icmplen;
162 
163 #ifdef ICMPPRINTFS
164 	if (icmpprintfs)
165 		kprintf("icmp_error(%p, %d, %d)\n", oip, type, code);
166 #endif
167 	if (type != ICMP_REDIRECT)
168 		icmpstat.icps_error++;
169 	/*
170 	 * Don't send error if the original packet was encrypted.
171 	 * Don't send error if not the first fragment of message.
172 	 * Don't error if the old packet protocol was ICMP
173 	 * error message, only known informational types.
174 	 */
175 	if (n->m_flags & M_DECRYPTED)
176 		goto freeit;
177 	if (oip->ip_off &~ (IP_MF|IP_DF))
178 		goto freeit;
179 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
180 	  n->m_len >= oiplen + ICMP_MINLEN &&
181 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
182 		icmpstat.icps_oldicmp++;
183 		goto freeit;
184 	}
185 	/* Don't send error in response to a multicast or broadcast packet */
186 	if (n->m_flags & (M_BCAST|M_MCAST))
187 		goto freeit;
188 	/*
189 	 * First, formulate icmp message
190 	 */
191 	m = m_gethdr(MB_DONTWAIT, MT_HEADER);
192 	if (m == NULL)
193 		goto freeit;
194 	icmplen = min(oiplen + 8, oip->ip_len);
195 	if (icmplen < sizeof(struct ip))
196 		panic("icmp_error: bad length");
197 	m->m_len = icmplen + ICMP_MINLEN;
198 	MH_ALIGN(m, m->m_len);
199 	icp = mtod(m, struct icmp *);
200 	if ((u_int)type > ICMP_MAXTYPE)
201 		panic("icmp_error");
202 	icmpstat.icps_outhist[type]++;
203 	icp->icmp_type = type;
204 	if (type == ICMP_REDIRECT)
205 		icp->icmp_gwaddr.s_addr = dest;
206 	else {
207 		icp->icmp_void = 0;
208 		/*
209 		 * The following assignments assume an overlay with the
210 		 * zeroed icmp_void field.
211 		 */
212 		if (type == ICMP_PARAMPROB) {
213 			icp->icmp_pptr = code;
214 			code = 0;
215 		} else if (type == ICMP_UNREACH &&
216 			code == ICMP_UNREACH_NEEDFRAG && destmtu) {
217 			icp->icmp_nextmtu = htons(destmtu);
218 		}
219 	}
220 
221 	icp->icmp_code = code;
222 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
223 	nip = &icp->icmp_ip;
224 
225 	/*
226 	 * Convert fields to network representation.
227 	 */
228 	nip->ip_len = htons(nip->ip_len);
229 	nip->ip_off = htons(nip->ip_off);
230 
231 	/*
232 	 * Now, copy old ip header (without options)
233 	 * in front of icmp message.
234 	 */
235 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
236 		panic("icmp len");
237 	m->m_data -= sizeof(struct ip);
238 	m->m_len += sizeof(struct ip);
239 	m->m_pkthdr.len = m->m_len;
240 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
241 	nip = mtod(m, struct ip *);
242 	bcopy(oip, nip, sizeof(struct ip));
243 	nip->ip_len = m->m_len;
244 	nip->ip_vhl = IP_VHL_BORING;
245 	nip->ip_p = IPPROTO_ICMP;
246 	nip->ip_tos = 0;
247 	m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & FW_MBUF_GENERATED;
248 	icmp_reflect(m);
249 
250 freeit:
251 	m_freem(n);
252 }
253 
254 /*
255  * Process a received ICMP message.
256  */
257 int
258 icmp_input(struct mbuf **mp, int *offp, int proto)
259 {
260 	struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
261 	struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
262 	struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
263 	struct icmp *icp;
264 	struct in_ifaddr *ia;
265 	struct mbuf *m = *mp;
266 	struct ip *ip = mtod(m, struct ip *);
267 	int icmplen = ip->ip_len;
268 	int i, hlen;
269 	int code;
270 
271 	*mp = NULL;
272 	hlen = *offp;
273 
274 	/*
275 	 * Locate icmp structure in mbuf, and check
276 	 * that not corrupted and of at least minimum length.
277 	 */
278 #ifdef ICMPPRINTFS
279 	if (icmpprintfs) {
280 		char buf[sizeof "aaa.bbb.ccc.ddd"];
281 
282 		strcpy(buf, inet_ntoa(ip->ip_src));
283 		kprintf("icmp_input from %s to %s, len %d\n",
284 		       buf, inet_ntoa(ip->ip_dst), icmplen);
285 	}
286 #endif
287 	if (icmplen < ICMP_MINLEN) {
288 		icmpstat.icps_tooshort++;
289 		goto freeit;
290 	}
291 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
292 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
293 		icmpstat.icps_tooshort++;
294 		return(IPPROTO_DONE);
295 	}
296 	ip = mtod(m, struct ip *);
297 	m->m_len -= hlen;
298 	m->m_data += hlen;
299 	icp = mtod(m, struct icmp *);
300 	if (in_cksum(m, icmplen)) {
301 		icmpstat.icps_checksum++;
302 		goto freeit;
303 	}
304 	m->m_len += hlen;
305 	m->m_data -= hlen;
306 
307 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
308 		/*
309 		 * Deliver very specific ICMP type only.
310 		 */
311 		switch (icp->icmp_type) {
312 		case ICMP_UNREACH:
313 		case ICMP_TIMXCEED:
314 			break;
315 		default:
316 			goto freeit;
317 		}
318 	}
319 
320 #ifdef ICMPPRINTFS
321 	if (icmpprintfs)
322 		kprintf("icmp_input, type %d code %d\n", icp->icmp_type,
323 		    icp->icmp_code);
324 #endif
325 
326 	/*
327 	 * Message type specific processing.
328 	 */
329 	if (icp->icmp_type > ICMP_MAXTYPE)
330 		goto raw;
331 	icmpstat.icps_inhist[icp->icmp_type]++;
332 	code = icp->icmp_code;
333 	switch (icp->icmp_type) {
334 
335 	case ICMP_UNREACH:
336 		switch (code) {
337 			case ICMP_UNREACH_NET:
338 			case ICMP_UNREACH_HOST:
339 			case ICMP_UNREACH_SRCFAIL:
340 			case ICMP_UNREACH_NET_UNKNOWN:
341 			case ICMP_UNREACH_HOST_UNKNOWN:
342 			case ICMP_UNREACH_ISOLATED:
343 			case ICMP_UNREACH_TOSNET:
344 			case ICMP_UNREACH_TOSHOST:
345 			case ICMP_UNREACH_HOST_PRECEDENCE:
346 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
347 				code = PRC_UNREACH_NET;
348 				break;
349 
350 			case ICMP_UNREACH_NEEDFRAG:
351 				code = PRC_MSGSIZE;
352 				break;
353 
354 			/*
355 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
356 			 * Treat subcodes 2,3 as immediate RST
357 			 */
358 			case ICMP_UNREACH_PROTOCOL:
359 			case ICMP_UNREACH_PORT:
360 				code = PRC_UNREACH_PORT;
361 				break;
362 
363 			case ICMP_UNREACH_NET_PROHIB:
364 			case ICMP_UNREACH_HOST_PROHIB:
365 			case ICMP_UNREACH_FILTER_PROHIB:
366 				code = PRC_UNREACH_ADMIN_PROHIB;
367 				break;
368 
369 			default:
370 				goto badcode;
371 		}
372 		goto deliver;
373 
374 	case ICMP_TIMXCEED:
375 		if (code > 1)
376 			goto badcode;
377 		code += PRC_TIMXCEED_INTRANS;
378 		goto deliver;
379 
380 	case ICMP_PARAMPROB:
381 		if (code > 1)
382 			goto badcode;
383 		code = PRC_PARAMPROB;
384 		goto deliver;
385 
386 	case ICMP_SOURCEQUENCH:
387 		if (code)
388 			goto badcode;
389 		if (discard_sourcequench)
390 			break;
391 		code = PRC_QUENCH;
392 deliver:
393 		/*
394 		 * Problem with datagram; advise higher level routines.
395 		 */
396 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
397 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
398 			icmpstat.icps_badlen++;
399 			goto freeit;
400 		}
401 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
402 		/* Discard ICMP's in response to multicast packets */
403 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
404 			goto badcode;
405 #ifdef ICMPPRINTFS
406 		if (icmpprintfs)
407 			kprintf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
408 #endif
409 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
410 #if 1
411 		/*
412 		 * MTU discovery:
413 		 * If we got a needfrag and there is a host route to the
414 		 * original destination, and the MTU is not locked, then
415 		 * set the MTU in the route to the suggested new value
416 		 * (if given) and then notify as usual.  The ULPs will
417 		 * notice that the MTU has changed and adapt accordingly.
418 		 * If no new MTU was suggested, then we guess a new one
419 		 * less than the current value.  If the new MTU is
420 		 * unreasonably small (arbitrarily set at 296), then
421 		 * we reset the MTU to the interface value and enable the
422 		 * lock bit, indicating that we are no longer doing MTU
423 		 * discovery.
424 		 */
425 		if (code == PRC_MSGSIZE) {
426 			struct rtentry *rt;
427 			int mtu;
428 
429 			rt = rtpurelookup((struct sockaddr *)&icmpsrc);
430 			if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
431 			    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
432 				mtu = ntohs(icp->icmp_nextmtu);
433 				if (!mtu)
434 					mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
435 							  1);
436 #ifdef DEBUG_MTUDISC
437 				kprintf("MTU for %s reduced to %d\n",
438 					inet_ntoa(icmpsrc.sin_addr), mtu);
439 #endif
440 				if (mtu < 296) {
441 					/* rt->rt_rmx.rmx_mtu =
442 						rt->rt_ifp->if_mtu; */
443 					rt->rt_rmx.rmx_locks |= RTV_MTU;
444 				} else if (rt->rt_rmx.rmx_mtu > mtu) {
445 					rt->rt_rmx.rmx_mtu = mtu;
446 				}
447 			}
448 			if (rt != NULL)
449 				--rt->rt_refcnt;
450 		}
451 #endif
452 		/*
453 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
454 		 * notification to TCP layer.
455 		 */
456 		so_pru_ctlinput(
457 			&inetsw[ip_protox[icp->icmp_ip.ip_p]],
458 			code, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
459 		break;
460 
461 badcode:
462 		icmpstat.icps_badcode++;
463 		break;
464 
465 	case ICMP_ECHO:
466 		if (!icmpbmcastecho
467 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
468 			icmpstat.icps_bmcastecho++;
469 			break;
470 		}
471 		icp->icmp_type = ICMP_ECHOREPLY;
472 #ifdef ICMP_BANDLIM
473 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
474 			goto freeit;
475 		else
476 #endif
477 			goto reflect;
478 
479 	case ICMP_TSTAMP:
480 		if (!icmpbmcastecho
481 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
482 			icmpstat.icps_bmcasttstamp++;
483 			break;
484 		}
485 		if (icmplen < ICMP_TSLEN) {
486 			icmpstat.icps_badlen++;
487 			break;
488 		}
489 		icp->icmp_type = ICMP_TSTAMPREPLY;
490 		icp->icmp_rtime = iptime();
491 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
492 #ifdef ICMP_BANDLIM
493 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
494 			goto freeit;
495 		else
496 #endif
497 			goto reflect;
498 
499 	case ICMP_MASKREQ:
500 		if (icmpmaskrepl == 0)
501 			break;
502 		/*
503 		 * We are not able to respond with all ones broadcast
504 		 * unless we receive it over a point-to-point interface.
505 		 */
506 		if (icmplen < ICMP_MASKLEN)
507 			break;
508 		switch (ip->ip_dst.s_addr) {
509 
510 		case INADDR_BROADCAST:
511 		case INADDR_ANY:
512 			icmpdst.sin_addr = ip->ip_src;
513 			break;
514 
515 		default:
516 			icmpdst.sin_addr = ip->ip_dst;
517 		}
518 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
519 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
520 		if (ia == NULL)
521 			break;
522 		if (ia->ia_ifp == 0)
523 			break;
524 		icp->icmp_type = ICMP_MASKREPLY;
525 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
526 		if (ip->ip_src.s_addr == 0) {
527 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
528 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
529 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
530 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
531 		}
532 reflect:
533 		ip->ip_len += hlen;	/* since ip_input deducts this */
534 		icmpstat.icps_reflect++;
535 		icmpstat.icps_outhist[icp->icmp_type]++;
536 		icmp_reflect(m);
537 		return(IPPROTO_DONE);
538 
539 	case ICMP_REDIRECT:
540 		if (log_redirect) {
541 			u_long src, dst, gw;
542 
543 			src = ntohl(ip->ip_src.s_addr);
544 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
545 			gw = ntohl(icp->icmp_gwaddr.s_addr);
546 			kprintf("icmp redirect from %d.%d.%d.%d: "
547 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
548 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
549 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
550 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
551 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
552 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
553 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
554 		}
555 		if (drop_redirect)
556 			break;
557 		if (code > 3)
558 			goto badcode;
559 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
560 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
561 			icmpstat.icps_badlen++;
562 			break;
563 		}
564 		/*
565 		 * Short circuit routing redirects to force
566 		 * immediate change in the kernel's routing
567 		 * tables.  The message is also handed to anyone
568 		 * listening on a raw socket (e.g. the routing
569 		 * daemon for use in updating its tables).
570 		 */
571 		icmpgw.sin_addr = ip->ip_src;
572 		icmpdst.sin_addr = icp->icmp_gwaddr;
573 #ifdef	ICMPPRINTFS
574 		if (icmpprintfs) {
575 			char buf[sizeof "aaa.bbb.ccc.ddd"];
576 
577 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
578 			kprintf("redirect dst %s to %s\n",
579 			       buf, inet_ntoa(icp->icmp_gwaddr));
580 		}
581 #endif
582 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
583 		rtredirect((struct sockaddr *)&icmpsrc,
584 		  (struct sockaddr *)&icmpdst,
585 		  NULL, RTF_GATEWAY | RTF_HOST,
586 		  (struct sockaddr *)&icmpgw);
587 		kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
588 #ifdef IPSEC
589 		key_sa_routechange((struct sockaddr *)&icmpsrc);
590 #endif
591 		break;
592 
593 	/*
594 	 * No kernel processing for the following;
595 	 * just fall through to send to raw listener.
596 	 */
597 	case ICMP_ECHOREPLY:
598 	case ICMP_ROUTERADVERT:
599 	case ICMP_ROUTERSOLICIT:
600 	case ICMP_TSTAMPREPLY:
601 	case ICMP_IREQREPLY:
602 	case ICMP_MASKREPLY:
603 	default:
604 		break;
605 	}
606 
607 raw:
608 	*mp = m;
609 	rip_input(mp, offp, proto);
610 	return(IPPROTO_DONE);
611 
612 freeit:
613 	m_freem(m);
614 	return(IPPROTO_DONE);
615 }
616 
617 /*
618  * Reflect the ip packet back to the source
619  */
620 static void
621 icmp_reflect(struct mbuf *m)
622 {
623 	struct ip *ip = mtod(m, struct ip *);
624 	struct in_ifaddr *ia;
625 	struct in_ifaddr_container *iac;
626 	struct ifaddr_container *ifac;
627 	struct ifnet *ifp;
628 	struct in_addr t;
629 	struct mbuf *opts = NULL;
630 	int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
631 	struct route *ro = NULL, rt;
632 
633 	if (!in_canforward(ip->ip_src) &&
634 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
635 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
636 		m_freem(m);	/* Bad return address */
637 		icmpstat.icps_badaddr++;
638 		goto done;	/* Ip_output() will check for broadcast */
639 	}
640 	t = ip->ip_dst;
641 	ip->ip_dst = ip->ip_src;
642 
643 	ro = &rt;
644 	bzero(ro, sizeof *ro);
645 
646 	/*
647 	 * If the incoming packet was addressed directly to us,
648 	 * use dst as the src for the reply.  Otherwise (broadcast
649 	 * or anonymous), use the address which corresponds
650 	 * to the incoming interface.
651 	 */
652 	ia = NULL;
653 	LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) {
654 		if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) {
655 			ia = iac->ia;
656 			goto match;
657 		}
658 	}
659 	ifp = m->m_pkthdr.rcvif;
660 	if (ifp != NULL && (ifp->if_flags & IFF_BROADCAST)) {
661 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
662 			struct ifaddr *ifa = ifac->ifa;
663 
664 			if (ifa->ifa_addr->sa_family != AF_INET)
665 				continue;
666 			ia = ifatoia(ifa);
667 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
668 			    t.s_addr)
669 				goto match;
670 		}
671 	}
672 	/*
673 	 * If the packet was transiting through us, use the address of
674 	 * the interface the packet came through in.  If that interface
675 	 * doesn't have a suitable IP address, the normal selection
676 	 * criteria apply.
677 	 */
678 	if (icmp_rfi && ifp != NULL) {
679 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
680 			struct ifaddr *ifa = ifac->ifa;
681 
682 			if (ifa->ifa_addr->sa_family != AF_INET)
683 				continue;
684 			ia = ifatoia(ifa);
685 			goto match;
686 		}
687 	}
688 	/*
689 	 * If the incoming packet was not addressed directly to us, use
690 	 * designated interface for icmp replies specified by sysctl
691 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
692 	 * with normal source selection.
693 	 */
694 	if (icmp_reply_src[0] != '\0' && (ifp = ifunit(icmp_reply_src))) {
695 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
696 			struct ifaddr *ifa = ifac->ifa;
697 
698 			if (ifa->ifa_addr->sa_family != AF_INET)
699 				continue;
700 			ia = ifatoia(ifa);
701 			goto match;
702 		}
703 	}
704 	/*
705 	 * If the packet was transiting through us, use the address of
706 	 * the interface that is the closest to the packet source.
707 	 * When we don't have a route back to the packet source, stop here
708 	 * and drop the packet.
709 	 */
710 	ia = ip_rtaddr(ip->ip_dst, ro);
711 	if (ia == NULL) {
712 		m_freem(m);
713 		icmpstat.icps_noroute++;
714 		goto done;
715 	}
716 match:
717 	t = IA_SIN(ia)->sin_addr;
718 	ip->ip_src = t;
719 	ip->ip_ttl = ip_defttl;
720 
721 	if (optlen > 0) {
722 		u_char *cp;
723 		int opt, cnt;
724 		u_int len;
725 
726 		/*
727 		 * Retrieve any source routing from the incoming packet;
728 		 * add on any record-route or timestamp options.
729 		 */
730 		cp = (u_char *) (ip + 1);
731 		if ((opts = ip_srcroute(m)) == NULL &&
732 		    (opts = m_gethdr(MB_DONTWAIT, MT_HEADER))) {
733 			opts->m_len = sizeof(struct in_addr);
734 			mtod(opts, struct in_addr *)->s_addr = 0;
735 		}
736 		if (opts) {
737 #ifdef ICMPPRINTFS
738 			if (icmpprintfs)
739 				kprintf("icmp_reflect optlen %d rt %d => ",
740 				       optlen, opts->m_len);
741 #endif
742 			for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
743 				opt = cp[IPOPT_OPTVAL];
744 				if (opt == IPOPT_EOL)
745 					break;
746 				if (opt == IPOPT_NOP)
747 					len = 1;
748 				else {
749 					if (cnt < IPOPT_OLEN + sizeof *cp)
750 						break;
751 					len = cp[IPOPT_OLEN];
752 					if (len < IPOPT_OLEN + sizeof *cp ||
753 					    len > cnt)
754 					break;
755 				}
756 				/*
757 				 * Should check for overflow, but it
758 				 * "can't happen".
759 				 */
760 				if (opt == IPOPT_RR || opt == IPOPT_TS ||
761 				    opt == IPOPT_SECURITY) {
762 					bcopy(cp,
763 					      mtod(opts, caddr_t) + opts->m_len,
764 					      len);
765 					opts->m_len += len;
766 				}
767 			}
768 			/* Terminate & pad, if necessary */
769 			cnt = opts->m_len % 4;
770 			if (cnt) {
771 				for (; cnt < 4; cnt++) {
772 					*(mtod(opts, caddr_t) + opts->m_len) =
773 					    IPOPT_EOL;
774 					opts->m_len++;
775 				}
776 			}
777 #ifdef ICMPPRINTFS
778 			if (icmpprintfs)
779 				kprintf("%d\n", opts->m_len);
780 #endif
781 		}
782 		/*
783 		 * Now strip out original options by copying rest of first
784 		 * mbuf's data back, and adjust the IP length.
785 		 */
786 		ip->ip_len -= optlen;
787 		ip->ip_vhl = IP_VHL_BORING;
788 		m->m_len -= optlen;
789 		if (m->m_flags & M_PKTHDR)
790 			m->m_pkthdr.len -= optlen;
791 		optlen += sizeof(struct ip);
792 		bcopy((caddr_t)ip + optlen, ip + 1,
793 		      m->m_len - sizeof(struct ip));
794 	}
795 	m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED;
796 	m->m_flags &= ~(M_BCAST|M_MCAST);
797 	icmp_send(m, opts, ro);
798 done:
799 	if (opts)
800 		m_free(opts);
801 	if (ro && ro->ro_rt)
802 		RTFREE(ro->ro_rt);
803 }
804 
805 /*
806  * Send an icmp packet back to the ip level,
807  * after supplying a checksum.
808  */
809 static void
810 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt)
811 {
812 	struct ip *ip = mtod(m, struct ip *);
813 	struct icmp *icp;
814 	int hlen;
815 
816 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
817 	m->m_data += hlen;
818 	m->m_len -= hlen;
819 	icp = mtod(m, struct icmp *);
820 	icp->icmp_cksum = 0;
821 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
822 	m->m_data -= hlen;
823 	m->m_len += hlen;
824 	m->m_pkthdr.rcvif = NULL;
825 #ifdef ICMPPRINTFS
826 	if (icmpprintfs) {
827 		char buf[sizeof "aaa.bbb.ccc.ddd"];
828 
829 		strcpy(buf, inet_ntoa(ip->ip_dst));
830 		kprintf("icmp_send dst %s src %s\n", buf, inet_ntoa(ip->ip_src));
831 	}
832 #endif
833 	ip_output(m, opts, rt, 0, NULL, NULL);
834 }
835 
836 n_time
837 iptime(void)
838 {
839 	struct timeval atv;
840 	u_long t;
841 
842 	getmicrotime(&atv);
843 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
844 	return (htonl(t));
845 }
846 
847 #if 1
848 /*
849  * Return the next larger or smaller MTU plateau (table from RFC 1191)
850  * given current value MTU.  If DIR is less than zero, a larger plateau
851  * is returned; otherwise, a smaller value is returned.
852  */
853 int
854 ip_next_mtu(int mtu, int dir)
855 {
856 	static int mtutab[] = {
857 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
858 		68, 0
859 	};
860 	int i;
861 
862 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
863 		if (mtu >= mtutab[i])
864 			break;
865 	}
866 
867 	if (dir < 0) {
868 		if (i == 0) {
869 			return 0;
870 		} else {
871 			return mtutab[i - 1];
872 		}
873 	} else {
874 		if (mtutab[i] == 0) {
875 			return 0;
876 		} else if(mtu > mtutab[i]) {
877 			return mtutab[i];
878 		} else {
879 			return mtutab[i + 1];
880 		}
881 	}
882 }
883 #endif
884 
885 #ifdef ICMP_BANDLIM
886 /*
887  * badport_bandlim() - check for ICMP bandwidth limit
888  *
889  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
890  *	hit our bandwidth limit and it is not ok.
891  *
892  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
893  *
894  *	For now we separate the TCP and UDP subsystems w/ different 'which'
895  *	values.  We may eventually remove this separation (and simplify the
896  *	code further).
897  *
898  *	Note that the printing of the error message is delayed so we can
899  *	properly print the icmp error rate that the system was trying to do
900  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
901  *	the 'final' error, but it doesn't make sense to solve the printing
902  *	delay with more complex code.
903  */
904 int
905 badport_bandlim(int which)
906 {
907 	static int lticks[BANDLIM_MAX + 1];
908 	static int lpackets[BANDLIM_MAX + 1];
909 	int dticks;
910 	const char *bandlimittype[] = {
911 		"Limiting icmp unreach response",
912 		"Limiting icmp ping response",
913 		"Limiting icmp tstamp response",
914 		"Limiting closed port RST response",
915 		"Limiting open port RST response"
916 		};
917 
918 	/*
919 	 * Return ok status if feature disabled or argument out of
920 	 * ranage.
921 	 */
922 
923 	if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
924 		return(0);
925 	dticks = ticks - lticks[which];
926 
927 	/*
928 	 * reset stats when cumulative dt exceeds one second.
929 	 */
930 
931 	if ((unsigned int)dticks > hz) {
932 		if (lpackets[which] > icmplim && icmplim_output) {
933 			kprintf("%s from %d to %d packets per second\n",
934 				bandlimittype[which],
935 				lpackets[which],
936 				icmplim
937 			);
938 		}
939 		lticks[which] = ticks;
940 		lpackets[which] = 0;
941 	}
942 
943 	/*
944 	 * bump packet count
945 	 */
946 
947 	if (++lpackets[which] > icmplim) {
948 		return(-1);
949 	}
950 	return(0);
951 }
952 #endif
953