xref: /freebsd/sys/netinet/ip_icmp.c (revision b28624fde638caadd4a89f50c9b7e7da0f98c4d2)
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  * $FreeBSD$
31  */
32 
33 #include "opt_ipsec.h"
34 #include "opt_mac.h"
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/mbuf.h>
39 #include <sys/protosw.h>
40 #include <sys/socket.h>
41 #include <sys/time.h>
42 #include <sys/kernel.h>
43 #include <sys/sysctl.h>
44 
45 #include <net/if.h>
46 #include <net/if_types.h>
47 #include <net/route.h>
48 
49 #include <netinet/in.h>
50 #include <netinet/in_pcb.h>
51 #include <netinet/in_systm.h>
52 #include <netinet/in_var.h>
53 #include <netinet/ip.h>
54 #include <netinet/ip_icmp.h>
55 #include <netinet/ip_var.h>
56 #include <netinet/ip_options.h>
57 #include <netinet/tcp.h>
58 #include <netinet/tcp_var.h>
59 #include <netinet/tcpip.h>
60 #include <netinet/icmp_var.h>
61 
62 #ifdef IPSEC
63 #include <netipsec/ipsec.h>
64 #include <netipsec/key.h>
65 #endif
66 
67 #include <machine/in_cksum.h>
68 
69 #include <security/mac/mac_framework.h>
70 
71 /*
72  * ICMP routines: error generation, receive packet processing, and
73  * routines to turnaround packets back to the originator, and
74  * host table maintenance routines.
75  */
76 
77 struct	icmpstat icmpstat;
78 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
79 	&icmpstat, icmpstat, "");
80 
81 static int	icmpmaskrepl = 0;
82 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
83 	&icmpmaskrepl, 0, "Reply to ICMP Address Mask Request packets.");
84 
85 static u_int	icmpmaskfake = 0;
86 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
87 	&icmpmaskfake, 0, "Fake reply to ICMP Address Mask Request packets.");
88 
89 static int	drop_redirect = 0;
90 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
91 	&drop_redirect, 0, "Ignore ICMP redirects");
92 
93 static int	log_redirect = 0;
94 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
95 	&log_redirect, 0, "Log ICMP redirects to the console");
96 
97 static int      icmplim = 200;
98 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
99 	&icmplim, 0, "Maximum number of ICMP responses per second");
100 
101 static int	icmplim_output = 1;
102 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
103 	&icmplim_output, 0, "Enable rate limiting of ICMP responses");
104 
105 static char	reply_src[IFNAMSIZ];
106 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
107 	&reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
108 
109 static int	icmp_rfi = 0;
110 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
111 	&icmp_rfi, 0, "ICMP reply from incoming interface for "
112 	"non-local packets");
113 
114 static int	icmp_quotelen = 8;
115 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
116 	&icmp_quotelen, 0, "Number of bytes from original packet to "
117 	"quote in ICMP reply");
118 
119 /*
120  * ICMP broadcast echo sysctl
121  */
122 
123 static int	icmpbmcastecho = 0;
124 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
125 	&icmpbmcastecho, 0, "");
126 
127 
128 #ifdef ICMPPRINTFS
129 int	icmpprintfs = 0;
130 #endif
131 
132 static void	icmp_reflect(struct mbuf *);
133 static void	icmp_send(struct mbuf *, struct mbuf *);
134 
135 extern	struct protosw inetsw[];
136 
137 /*
138  * Generate an error packet of type error
139  * in response to bad packet ip.
140  */
141 void
142 icmp_error(struct mbuf *n, int type, int code, n_long dest, int mtu)
143 {
144 	register struct ip *oip = mtod(n, struct ip *), *nip;
145 	register unsigned oiphlen = oip->ip_hl << 2;
146 	register struct icmp *icp;
147 	register struct mbuf *m;
148 	unsigned icmplen, icmpelen, nlen;
149 
150 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
151 #ifdef ICMPPRINTFS
152 	if (icmpprintfs)
153 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
154 #endif
155 	if (type != ICMP_REDIRECT)
156 		icmpstat.icps_error++;
157 	/*
158 	 * Don't send error:
159 	 *  if the original packet was encrypted.
160 	 *  if not the first fragment of message.
161 	 *  in response to a multicast or broadcast packet.
162 	 *  if the old packet protocol was an ICMP error message.
163 	 */
164 	if (n->m_flags & M_DECRYPTED)
165 		goto freeit;
166 	if (oip->ip_off & ~(IP_MF|IP_DF))
167 		goto freeit;
168 	if (n->m_flags & (M_BCAST|M_MCAST))
169 		goto freeit;
170 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
171 	  n->m_len >= oiphlen + ICMP_MINLEN &&
172 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
173 		icmpstat.icps_oldicmp++;
174 		goto freeit;
175 	}
176 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
177 	if (oiphlen + 8 > n->m_len)
178 		goto freeit;
179 	/*
180 	 * Calculate length to quote from original packet and
181 	 * prevent the ICMP mbuf from overflowing.
182 	 * Unfortunatly this is non-trivial since ip_forward()
183 	 * sends us truncated packets.
184 	 */
185 	nlen = m_length(n, NULL);
186 	if (oip->ip_p == IPPROTO_TCP) {
187 		struct tcphdr *th;
188 		int tcphlen;
189 
190 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
191 		    n->m_next == NULL)
192 			goto stdreply;
193 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
194 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
195 			goto freeit;
196 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
197 		tcphlen = th->th_off << 2;
198 		if (tcphlen < sizeof(struct tcphdr))
199 			goto freeit;
200 		if (oip->ip_len < oiphlen + tcphlen)
201 			goto freeit;
202 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
203 			goto stdreply;
204 		if (n->m_len < oiphlen + tcphlen &&
205 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
206 			goto freeit;
207 		icmpelen = max(tcphlen, min(icmp_quotelen, oip->ip_len - oiphlen));
208 	} else
209 stdreply:	icmpelen = max(8, min(icmp_quotelen, oip->ip_len - oiphlen));
210 
211 	icmplen = min(oiphlen + icmpelen, nlen);
212 	if (icmplen < sizeof(struct ip))
213 		goto freeit;
214 
215 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
216 		m = m_gethdr(M_DONTWAIT, MT_DATA);
217 	else
218 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
219 	if (m == NULL)
220 		goto freeit;
221 #ifdef MAC
222 	mac_create_mbuf_netlayer(n, m);
223 #endif
224 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
225 	m_align(m, ICMP_MINLEN + icmplen);
226 	m->m_len = ICMP_MINLEN + icmplen;
227 
228 	icp = mtod(m, struct icmp *);
229 	icmpstat.icps_outhist[type]++;
230 	icp->icmp_type = type;
231 	if (type == ICMP_REDIRECT)
232 		icp->icmp_gwaddr.s_addr = dest;
233 	else {
234 		icp->icmp_void = 0;
235 		/*
236 		 * The following assignments assume an overlay with the
237 		 * just zeroed icmp_void field.
238 		 */
239 		if (type == ICMP_PARAMPROB) {
240 			icp->icmp_pptr = code;
241 			code = 0;
242 		} else if (type == ICMP_UNREACH &&
243 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
244 			icp->icmp_nextmtu = htons(mtu);
245 		}
246 	}
247 	icp->icmp_code = code;
248 
249 	/*
250 	 * Copy the quotation into ICMP message and
251 	 * convert quoted IP header back to network representation.
252 	 */
253 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
254 	nip = &icp->icmp_ip;
255 	nip->ip_len = htons(nip->ip_len);
256 	nip->ip_off = htons(nip->ip_off);
257 
258 	/*
259 	 * Set up ICMP message mbuf and copy old IP header (without options
260 	 * in front of ICMP message.
261 	 * If the original mbuf was meant to bypass the firewall, the error
262 	 * reply should bypass as well.
263 	 */
264 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
265 	m->m_data -= sizeof(struct ip);
266 	m->m_len += sizeof(struct ip);
267 	m->m_pkthdr.len = m->m_len;
268 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
269 	nip = mtod(m, struct ip *);
270 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
271 	nip->ip_len = m->m_len;
272 	nip->ip_v = IPVERSION;
273 	nip->ip_hl = 5;
274 	nip->ip_p = IPPROTO_ICMP;
275 	nip->ip_tos = 0;
276 	icmp_reflect(m);
277 
278 freeit:
279 	m_freem(n);
280 }
281 
282 /*
283  * Process a received ICMP message.
284  */
285 void
286 icmp_input(struct mbuf *m, int off)
287 {
288 	struct icmp *icp;
289 	struct in_ifaddr *ia;
290 	struct ip *ip = mtod(m, struct ip *);
291 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
292 	int hlen = off;
293 	int icmplen = ip->ip_len;
294 	int i, code;
295 	void (*ctlfunc)(int, struct sockaddr *, void *);
296 
297 	/*
298 	 * Locate icmp structure in mbuf, and check
299 	 * that not corrupted and of at least minimum length.
300 	 */
301 #ifdef ICMPPRINTFS
302 	if (icmpprintfs) {
303 		char buf[4 * sizeof "123"];
304 		strcpy(buf, inet_ntoa(ip->ip_src));
305 		printf("icmp_input from %s to %s, len %d\n",
306 		       buf, inet_ntoa(ip->ip_dst), icmplen);
307 	}
308 #endif
309 	if (icmplen < ICMP_MINLEN) {
310 		icmpstat.icps_tooshort++;
311 		goto freeit;
312 	}
313 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
314 	if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
315 		icmpstat.icps_tooshort++;
316 		return;
317 	}
318 	ip = mtod(m, struct ip *);
319 	m->m_len -= hlen;
320 	m->m_data += hlen;
321 	icp = mtod(m, struct icmp *);
322 	if (in_cksum(m, icmplen)) {
323 		icmpstat.icps_checksum++;
324 		goto freeit;
325 	}
326 	m->m_len += hlen;
327 	m->m_data -= hlen;
328 
329 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
330 		/*
331 		 * Deliver very specific ICMP type only.
332 		 */
333 		switch (icp->icmp_type) {
334 		case ICMP_UNREACH:
335 		case ICMP_TIMXCEED:
336 			break;
337 		default:
338 			goto freeit;
339 		}
340 	}
341 
342 #ifdef ICMPPRINTFS
343 	if (icmpprintfs)
344 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
345 		    icp->icmp_code);
346 #endif
347 
348 	/*
349 	 * Message type specific processing.
350 	 */
351 	if (icp->icmp_type > ICMP_MAXTYPE)
352 		goto raw;
353 
354 	/* Initialize */
355 	bzero(&icmpsrc, sizeof(icmpsrc));
356 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
357 	icmpsrc.sin_family = AF_INET;
358 	bzero(&icmpdst, sizeof(icmpdst));
359 	icmpdst.sin_len = sizeof(struct sockaddr_in);
360 	icmpdst.sin_family = AF_INET;
361 	bzero(&icmpgw, sizeof(icmpgw));
362 	icmpgw.sin_len = sizeof(struct sockaddr_in);
363 	icmpgw.sin_family = AF_INET;
364 
365 	icmpstat.icps_inhist[icp->icmp_type]++;
366 	code = icp->icmp_code;
367 	switch (icp->icmp_type) {
368 
369 	case ICMP_UNREACH:
370 		switch (code) {
371 			case ICMP_UNREACH_NET:
372 			case ICMP_UNREACH_HOST:
373 			case ICMP_UNREACH_SRCFAIL:
374 			case ICMP_UNREACH_NET_UNKNOWN:
375 			case ICMP_UNREACH_HOST_UNKNOWN:
376 			case ICMP_UNREACH_ISOLATED:
377 			case ICMP_UNREACH_TOSNET:
378 			case ICMP_UNREACH_TOSHOST:
379 			case ICMP_UNREACH_HOST_PRECEDENCE:
380 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
381 				code = PRC_UNREACH_NET;
382 				break;
383 
384 			case ICMP_UNREACH_NEEDFRAG:
385 				code = PRC_MSGSIZE;
386 				break;
387 
388 			/*
389 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
390 			 * Treat subcodes 2,3 as immediate RST
391 			 */
392 			case ICMP_UNREACH_PROTOCOL:
393 			case ICMP_UNREACH_PORT:
394 				code = PRC_UNREACH_PORT;
395 				break;
396 
397 			case ICMP_UNREACH_NET_PROHIB:
398 			case ICMP_UNREACH_HOST_PROHIB:
399 			case ICMP_UNREACH_FILTER_PROHIB:
400 				code = PRC_UNREACH_ADMIN_PROHIB;
401 				break;
402 
403 			default:
404 				goto badcode;
405 		}
406 		goto deliver;
407 
408 	case ICMP_TIMXCEED:
409 		if (code > 1)
410 			goto badcode;
411 		code += PRC_TIMXCEED_INTRANS;
412 		goto deliver;
413 
414 	case ICMP_PARAMPROB:
415 		if (code > 1)
416 			goto badcode;
417 		code = PRC_PARAMPROB;
418 		goto deliver;
419 
420 	case ICMP_SOURCEQUENCH:
421 		if (code)
422 			goto badcode;
423 		code = PRC_QUENCH;
424 	deliver:
425 		/*
426 		 * Problem with datagram; advise higher level routines.
427 		 */
428 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
429 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
430 			icmpstat.icps_badlen++;
431 			goto freeit;
432 		}
433 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
434 		/* Discard ICMP's in response to multicast packets */
435 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
436 			goto badcode;
437 #ifdef ICMPPRINTFS
438 		if (icmpprintfs)
439 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
440 #endif
441 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
442 		/*
443 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
444 		 * notification to TCP layer.
445 		 */
446 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
447 		if (ctlfunc)
448 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
449 				   (void *)&icp->icmp_ip);
450 		break;
451 
452 	badcode:
453 		icmpstat.icps_badcode++;
454 		break;
455 
456 	case ICMP_ECHO:
457 		if (!icmpbmcastecho
458 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
459 			icmpstat.icps_bmcastecho++;
460 			break;
461 		}
462 		icp->icmp_type = ICMP_ECHOREPLY;
463 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
464 			goto freeit;
465 		else
466 			goto reflect;
467 
468 	case ICMP_TSTAMP:
469 		if (!icmpbmcastecho
470 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
471 			icmpstat.icps_bmcasttstamp++;
472 			break;
473 		}
474 		if (icmplen < ICMP_TSLEN) {
475 			icmpstat.icps_badlen++;
476 			break;
477 		}
478 		icp->icmp_type = ICMP_TSTAMPREPLY;
479 		icp->icmp_rtime = iptime();
480 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
481 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
482 			goto freeit;
483 		else
484 			goto reflect;
485 
486 	case ICMP_MASKREQ:
487 		if (icmpmaskrepl == 0)
488 			break;
489 		/*
490 		 * We are not able to respond with all ones broadcast
491 		 * unless we receive it over a point-to-point interface.
492 		 */
493 		if (icmplen < ICMP_MASKLEN)
494 			break;
495 		switch (ip->ip_dst.s_addr) {
496 
497 		case INADDR_BROADCAST:
498 		case INADDR_ANY:
499 			icmpdst.sin_addr = ip->ip_src;
500 			break;
501 
502 		default:
503 			icmpdst.sin_addr = ip->ip_dst;
504 		}
505 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
506 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
507 		if (ia == 0)
508 			break;
509 		if (ia->ia_ifp == 0)
510 			break;
511 		icp->icmp_type = ICMP_MASKREPLY;
512 		if (icmpmaskfake == 0)
513 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
514 		else
515 			icp->icmp_mask = icmpmaskfake;
516 		if (ip->ip_src.s_addr == 0) {
517 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
518 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
519 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
520 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
521 		}
522 reflect:
523 		ip->ip_len += hlen;	/* since ip_input deducts this */
524 		icmpstat.icps_reflect++;
525 		icmpstat.icps_outhist[icp->icmp_type]++;
526 		icmp_reflect(m);
527 		return;
528 
529 	case ICMP_REDIRECT:
530 		if (log_redirect) {
531 			u_long src, dst, gw;
532 
533 			src = ntohl(ip->ip_src.s_addr);
534 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
535 			gw = ntohl(icp->icmp_gwaddr.s_addr);
536 			printf("icmp redirect from %d.%d.%d.%d: "
537 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
538 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
539 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
540 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
541 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
542 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
543 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
544 		}
545 		/*
546 		 * RFC1812 says we must ignore ICMP redirects if we
547 		 * are acting as router.
548 		 */
549 		if (drop_redirect || ipforwarding)
550 			break;
551 		if (code > 3)
552 			goto badcode;
553 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
554 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
555 			icmpstat.icps_badlen++;
556 			break;
557 		}
558 		/*
559 		 * Short circuit routing redirects to force
560 		 * immediate change in the kernel's routing
561 		 * tables.  The message is also handed to anyone
562 		 * listening on a raw socket (e.g. the routing
563 		 * daemon for use in updating its tables).
564 		 */
565 		icmpgw.sin_addr = ip->ip_src;
566 		icmpdst.sin_addr = icp->icmp_gwaddr;
567 #ifdef	ICMPPRINTFS
568 		if (icmpprintfs) {
569 			char buf[4 * sizeof "123"];
570 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
571 
572 			printf("redirect dst %s to %s\n",
573 			       buf, inet_ntoa(icp->icmp_gwaddr));
574 		}
575 #endif
576 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
577 		rtredirect((struct sockaddr *)&icmpsrc,
578 		  (struct sockaddr *)&icmpdst,
579 		  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
580 		  (struct sockaddr *)&icmpgw);
581 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
582 #ifdef IPSEC
583 		key_sa_routechange((struct sockaddr *)&icmpsrc);
584 #endif
585 		break;
586 
587 	/*
588 	 * No kernel processing for the following;
589 	 * just fall through to send to raw listener.
590 	 */
591 	case ICMP_ECHOREPLY:
592 	case ICMP_ROUTERADVERT:
593 	case ICMP_ROUTERSOLICIT:
594 	case ICMP_TSTAMPREPLY:
595 	case ICMP_IREQREPLY:
596 	case ICMP_MASKREPLY:
597 	default:
598 		break;
599 	}
600 
601 raw:
602 	rip_input(m, off);
603 	return;
604 
605 freeit:
606 	m_freem(m);
607 }
608 
609 /*
610  * Reflect the ip packet back to the source
611  */
612 static void
613 icmp_reflect(struct mbuf *m)
614 {
615 	struct ip *ip = mtod(m, struct ip *);
616 	struct ifaddr *ifa;
617 	struct ifnet *ifn;
618 	struct in_ifaddr *ia;
619 	struct in_addr t;
620 	struct mbuf *opts = 0;
621 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
622 
623 	if (!in_canforward(ip->ip_src) &&
624 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
625 	     (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
626 		m_freem(m);	/* Bad return address */
627 		icmpstat.icps_badaddr++;
628 		goto done;	/* Ip_output() will check for broadcast */
629 	}
630 	t = ip->ip_dst;
631 	ip->ip_dst = ip->ip_src;
632 
633 	/*
634 	 * Source selection for ICMP replies:
635 	 *
636 	 * If the incoming packet was addressed directly to one of our
637 	 * own addresses, use dst as the src for the reply.
638 	 */
639 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
640 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
641 			goto match;
642 	/*
643 	 * If the incoming packet was addressed to one of our broadcast
644 	 * addresses, use the first non-broadcast address which corresponds
645 	 * to the incoming interface.
646 	 */
647 	if (m->m_pkthdr.rcvif != NULL &&
648 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
649 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
650 			if (ifa->ifa_addr->sa_family != AF_INET)
651 				continue;
652 			ia = ifatoia(ifa);
653 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
654 			    t.s_addr)
655 				goto match;
656 		}
657 	}
658 	/*
659 	 * If the packet was transiting through us, use the address of
660 	 * the interface the packet came through in.  If that interface
661 	 * doesn't have a suitable IP address, the normal selection
662 	 * criteria apply.
663 	 */
664 	if (icmp_rfi && m->m_pkthdr.rcvif != NULL) {
665 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
666 			if (ifa->ifa_addr->sa_family != AF_INET)
667 				continue;
668 			ia = ifatoia(ifa);
669 			goto match;
670 		}
671 	}
672 	/*
673 	 * If the incoming packet was not addressed directly to us, use
674 	 * designated interface for icmp replies specified by sysctl
675 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
676 	 * with normal source selection.
677 	 */
678 	if (reply_src[0] != '\0' && (ifn = ifunit(reply_src))) {
679 		TAILQ_FOREACH(ifa, &ifn->if_addrhead, ifa_link) {
680 			if (ifa->ifa_addr->sa_family != AF_INET)
681 				continue;
682 			ia = ifatoia(ifa);
683 			goto match;
684 		}
685 	}
686 	/*
687 	 * If the packet was transiting through us, use the address of
688 	 * the interface that is the closest to the packet source.
689 	 * When we don't have a route back to the packet source, stop here
690 	 * and drop the packet.
691 	 */
692 	ia = ip_rtaddr(ip->ip_dst);
693 	if (ia == NULL) {
694 		m_freem(m);
695 		icmpstat.icps_noroute++;
696 		goto done;
697 	}
698 match:
699 #ifdef MAC
700 	mac_reflect_mbuf_icmp(m);
701 #endif
702 	t = IA_SIN(ia)->sin_addr;
703 	ip->ip_src = t;
704 	ip->ip_ttl = ip_defttl;
705 
706 	if (optlen > 0) {
707 		register u_char *cp;
708 		int opt, cnt;
709 		u_int len;
710 
711 		/*
712 		 * Retrieve any source routing from the incoming packet;
713 		 * add on any record-route or timestamp options.
714 		 */
715 		cp = (u_char *) (ip + 1);
716 		if ((opts = ip_srcroute(m)) == 0 &&
717 		    (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
718 			opts->m_len = sizeof(struct in_addr);
719 			mtod(opts, struct in_addr *)->s_addr = 0;
720 		}
721 		if (opts) {
722 #ifdef ICMPPRINTFS
723 		    if (icmpprintfs)
724 			    printf("icmp_reflect optlen %d rt %d => ",
725 				optlen, opts->m_len);
726 #endif
727 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
728 			    opt = cp[IPOPT_OPTVAL];
729 			    if (opt == IPOPT_EOL)
730 				    break;
731 			    if (opt == IPOPT_NOP)
732 				    len = 1;
733 			    else {
734 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
735 					    break;
736 				    len = cp[IPOPT_OLEN];
737 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
738 				        len > cnt)
739 					    break;
740 			    }
741 			    /*
742 			     * Should check for overflow, but it "can't happen"
743 			     */
744 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
745 				opt == IPOPT_SECURITY) {
746 				    bcopy((caddr_t)cp,
747 					mtod(opts, caddr_t) + opts->m_len, len);
748 				    opts->m_len += len;
749 			    }
750 		    }
751 		    /* Terminate & pad, if necessary */
752 		    cnt = opts->m_len % 4;
753 		    if (cnt) {
754 			    for (; cnt < 4; cnt++) {
755 				    *(mtod(opts, caddr_t) + opts->m_len) =
756 					IPOPT_EOL;
757 				    opts->m_len++;
758 			    }
759 		    }
760 #ifdef ICMPPRINTFS
761 		    if (icmpprintfs)
762 			    printf("%d\n", opts->m_len);
763 #endif
764 		}
765 		/*
766 		 * Now strip out original options by copying rest of first
767 		 * mbuf's data back, and adjust the IP length.
768 		 */
769 		ip->ip_len -= optlen;
770 		ip->ip_v = IPVERSION;
771 		ip->ip_hl = 5;
772 		m->m_len -= optlen;
773 		if (m->m_flags & M_PKTHDR)
774 			m->m_pkthdr.len -= optlen;
775 		optlen += sizeof(struct ip);
776 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
777 			 (unsigned)(m->m_len - sizeof(struct ip)));
778 	}
779 	m_tag_delete_nonpersistent(m);
780 	m->m_flags &= ~(M_BCAST|M_MCAST);
781 	icmp_send(m, opts);
782 done:
783 	if (opts)
784 		(void)m_free(opts);
785 }
786 
787 /*
788  * Send an icmp packet back to the ip level,
789  * after supplying a checksum.
790  */
791 static void
792 icmp_send(struct mbuf *m, struct mbuf *opts)
793 {
794 	register struct ip *ip = mtod(m, struct ip *);
795 	register int hlen;
796 	register struct icmp *icp;
797 
798 	hlen = ip->ip_hl << 2;
799 	m->m_data += hlen;
800 	m->m_len -= hlen;
801 	icp = mtod(m, struct icmp *);
802 	icp->icmp_cksum = 0;
803 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
804 	m->m_data -= hlen;
805 	m->m_len += hlen;
806 	m->m_pkthdr.rcvif = (struct ifnet *)0;
807 #ifdef ICMPPRINTFS
808 	if (icmpprintfs) {
809 		char buf[4 * sizeof "123"];
810 		strcpy(buf, inet_ntoa(ip->ip_dst));
811 		printf("icmp_send dst %s src %s\n",
812 		       buf, inet_ntoa(ip->ip_src));
813 	}
814 #endif
815 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
816 }
817 
818 n_time
819 iptime(void)
820 {
821 	struct timeval atv;
822 	u_long t;
823 
824 	getmicrotime(&atv);
825 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
826 	return (htonl(t));
827 }
828 
829 /*
830  * Return the next larger or smaller MTU plateau (table from RFC 1191)
831  * given current value MTU.  If DIR is less than zero, a larger plateau
832  * is returned; otherwise, a smaller value is returned.
833  */
834 int
835 ip_next_mtu(int mtu, int dir)
836 {
837 	static int mtutab[] = {
838 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
839 		296, 68, 0
840 	};
841 	int i, size;
842 
843 	size = (sizeof mtutab) / (sizeof mtutab[0]);
844 	if (dir >= 0) {
845 		for (i = 0; i < size; i++)
846 			if (mtu > mtutab[i])
847 				return mtutab[i];
848 	} else {
849 		for (i = size - 1; i >= 0; i--)
850 			if (mtu < mtutab[i])
851 				return mtutab[i];
852 		if (mtu == mtutab[0])
853 			return mtutab[0];
854 	}
855 	return 0;
856 }
857 
858 
859 /*
860  * badport_bandlim() - check for ICMP bandwidth limit
861  *
862  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
863  *	hit our bandwidth limit and it is not ok.
864  *
865  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
866  *
867  *	For now we separate the TCP and UDP subsystems w/ different 'which'
868  *	values.  We may eventually remove this separation (and simplify the
869  *	code further).
870  *
871  *	Note that the printing of the error message is delayed so we can
872  *	properly print the icmp error rate that the system was trying to do
873  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
874  *	the 'final' error, but it doesn't make sense to solve the printing
875  *	delay with more complex code.
876  */
877 
878 int
879 badport_bandlim(int which)
880 {
881 #define	N(a)	(sizeof (a) / sizeof (a[0]))
882 	static struct rate {
883 		const char	*type;
884 		struct timeval	lasttime;
885 		int		curpps;
886 	} rates[BANDLIM_MAX+1] = {
887 		{ "icmp unreach response" },
888 		{ "icmp ping response" },
889 		{ "icmp tstamp response" },
890 		{ "closed port RST response" },
891 		{ "open port RST response" },
892 		{ "icmp6 unreach response" }
893 	};
894 
895 	/*
896 	 * Return ok status if feature disabled or argument out of range.
897 	 */
898 	if (icmplim > 0 && (u_int) which < N(rates)) {
899 		struct rate *r = &rates[which];
900 		int opps = r->curpps;
901 
902 		if (!ppsratecheck(&r->lasttime, &r->curpps, icmplim))
903 			return -1;	/* discard packet */
904 		/*
905 		 * If we've dropped below the threshold after having
906 		 * rate-limited traffic print the message.  This preserves
907 		 * the previous behaviour at the expense of added complexity.
908 		 */
909 		if (icmplim_output && opps > icmplim)
910 			printf("Limiting %s from %d to %d packets/sec\n",
911 				r->type, opps, icmplim);
912 	}
913 	return 0;			/* okay to send packet */
914 #undef N
915 }
916