xref: /freebsd/sys/netinet/ip_icmp.c (revision 145992504973bd16cf3518af9ba5ce185fefa82a)
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_inet.h"
36 #include "opt_ipsec.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 #include <sys/syslog.h>
47 
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/route.h>
51 #include <net/vnet.h>
52 
53 #include <netinet/in.h>
54 #include <netinet/in_pcb.h>
55 #include <netinet/in_systm.h>
56 #include <netinet/in_var.h>
57 #include <netinet/ip.h>
58 #include <netinet/ip_icmp.h>
59 #include <netinet/ip_var.h>
60 #include <netinet/ip_options.h>
61 #include <netinet/tcp.h>
62 #include <netinet/tcp_var.h>
63 #include <netinet/tcpip.h>
64 #include <netinet/icmp_var.h>
65 
66 #ifdef INET
67 #ifdef IPSEC
68 #include <netipsec/ipsec.h>
69 #include <netipsec/key.h>
70 #endif
71 
72 #include <machine/in_cksum.h>
73 
74 #include <security/mac/mac_framework.h>
75 #endif /* INET */
76 
77 /*
78  * ICMP routines: error generation, receive packet processing, and
79  * routines to turnaround packets back to the originator, and
80  * host table maintenance routines.
81  */
82 static VNET_DEFINE(int, icmplim) = 200;
83 #define	V_icmplim			VNET(icmplim)
84 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
85 	&VNET_NAME(icmplim), 0,
86 	"Maximum number of ICMP responses per second");
87 
88 static VNET_DEFINE(int, icmplim_output) = 1;
89 #define	V_icmplim_output		VNET(icmplim_output)
90 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
91 	&VNET_NAME(icmplim_output), 0,
92 	"Enable logging of ICMP response rate limiting");
93 
94 #ifdef INET
95 VNET_DEFINE(struct icmpstat, icmpstat);
96 SYSCTL_VNET_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
97 	&VNET_NAME(icmpstat), icmpstat, "");
98 
99 static VNET_DEFINE(int, icmpmaskrepl) = 0;
100 #define	V_icmpmaskrepl			VNET(icmpmaskrepl)
101 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
102 	&VNET_NAME(icmpmaskrepl), 0,
103 	"Reply to ICMP Address Mask Request packets.");
104 
105 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
106 #define	V_icmpmaskfake			VNET(icmpmaskfake)
107 SYSCTL_VNET_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
108 	&VNET_NAME(icmpmaskfake), 0,
109 	"Fake reply to ICMP Address Mask Request packets.");
110 
111 VNET_DEFINE(int, drop_redirect) = 0;
112 
113 static VNET_DEFINE(int, log_redirect) = 0;
114 #define	V_log_redirect			VNET(log_redirect)
115 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
116 	&VNET_NAME(log_redirect), 0,
117 	"Log ICMP redirects to the console");
118 
119 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
120 #define	V_reply_src			VNET(reply_src)
121 SYSCTL_VNET_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
122 	&VNET_NAME(reply_src), IFNAMSIZ,
123 	"icmp reply source for non-local packets.");
124 
125 static VNET_DEFINE(int, icmp_rfi) = 0;
126 #define	V_icmp_rfi			VNET(icmp_rfi)
127 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
128 	&VNET_NAME(icmp_rfi), 0,
129 	"ICMP reply from incoming interface for non-local packets");
130 
131 static VNET_DEFINE(int, icmp_quotelen) = 8;
132 #define	V_icmp_quotelen			VNET(icmp_quotelen)
133 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
134 	&VNET_NAME(icmp_quotelen), 0,
135 	"Number of bytes from original packet to quote in ICMP reply");
136 
137 /*
138  * ICMP broadcast echo sysctl
139  */
140 static VNET_DEFINE(int, icmpbmcastecho) = 0;
141 #define	V_icmpbmcastecho		VNET(icmpbmcastecho)
142 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
143 	&VNET_NAME(icmpbmcastecho), 0,
144 	"");
145 
146 
147 #ifdef ICMPPRINTFS
148 int	icmpprintfs = 0;
149 #endif
150 
151 static void	icmp_reflect(struct mbuf *);
152 static void	icmp_send(struct mbuf *, struct mbuf *);
153 
154 extern	struct protosw inetsw[];
155 
156 static int
157 sysctl_net_icmp_drop_redir(SYSCTL_HANDLER_ARGS)
158 {
159 	int error, new;
160 	int i;
161 	struct radix_node_head *rnh;
162 
163 	new = V_drop_redirect;
164 	error = sysctl_handle_int(oidp, &new, 0, req);
165 	if (error == 0 && req->newptr) {
166 		new = (new != 0) ? 1 : 0;
167 
168 		if (new == V_drop_redirect)
169 			return (0);
170 
171 		for (i = 0; i < rt_numfibs; i++) {
172 			if ((rnh = rt_tables_get_rnh(i, AF_INET)) == NULL)
173 				continue;
174 			RADIX_NODE_HEAD_LOCK(rnh);
175 			in_setmatchfunc(rnh, new);
176 			RADIX_NODE_HEAD_UNLOCK(rnh);
177 		}
178 
179 		V_drop_redirect = new;
180 	}
181 
182 	return (error);
183 }
184 
185 SYSCTL_VNET_PROC(_net_inet_icmp, OID_AUTO, drop_redirect,
186     CTLTYPE_INT|CTLFLAG_RW, 0, 0,
187     sysctl_net_icmp_drop_redir, "I", "Ignore ICMP redirects");
188 
189 /*
190  * Kernel module interface for updating icmpstat.  The argument is an index
191  * into icmpstat treated as an array of u_long.  While this encodes the
192  * general layout of icmpstat into the caller, it doesn't encode its
193  * location, so that future changes to add, for example, per-CPU stats
194  * support won't cause binary compatibility problems for kernel modules.
195  */
196 void
197 kmod_icmpstat_inc(int statnum)
198 {
199 
200 	(*((u_long *)&V_icmpstat + statnum))++;
201 }
202 
203 /*
204  * Generate an error packet of type error
205  * in response to bad packet ip.
206  */
207 void
208 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
209 {
210 	register struct ip *oip = mtod(n, struct ip *), *nip;
211 	register unsigned oiphlen = oip->ip_hl << 2;
212 	register struct icmp *icp;
213 	register struct mbuf *m;
214 	unsigned icmplen, icmpelen, nlen;
215 
216 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
217 #ifdef ICMPPRINTFS
218 	if (icmpprintfs)
219 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
220 #endif
221 	if (type != ICMP_REDIRECT)
222 		ICMPSTAT_INC(icps_error);
223 	/*
224 	 * Don't send error:
225 	 *  if the original packet was encrypted.
226 	 *  if not the first fragment of message.
227 	 *  in response to a multicast or broadcast packet.
228 	 *  if the old packet protocol was an ICMP error message.
229 	 */
230 	if (n->m_flags & M_DECRYPTED)
231 		goto freeit;
232 	if (oip->ip_off & ~(IP_MF|IP_DF))
233 		goto freeit;
234 	if (n->m_flags & (M_BCAST|M_MCAST))
235 		goto freeit;
236 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
237 	  n->m_len >= oiphlen + ICMP_MINLEN &&
238 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
239 		ICMPSTAT_INC(icps_oldicmp);
240 		goto freeit;
241 	}
242 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
243 	if (oiphlen + 8 > n->m_len)
244 		goto freeit;
245 	/*
246 	 * Calculate length to quote from original packet and
247 	 * prevent the ICMP mbuf from overflowing.
248 	 * Unfortunatly this is non-trivial since ip_forward()
249 	 * sends us truncated packets.
250 	 */
251 	nlen = m_length(n, NULL);
252 	if (oip->ip_p == IPPROTO_TCP) {
253 		struct tcphdr *th;
254 		int tcphlen;
255 
256 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
257 		    n->m_next == NULL)
258 			goto stdreply;
259 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
260 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
261 			goto freeit;
262 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
263 		tcphlen = th->th_off << 2;
264 		if (tcphlen < sizeof(struct tcphdr))
265 			goto freeit;
266 		if (oip->ip_len < oiphlen + tcphlen)
267 			goto freeit;
268 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
269 			goto stdreply;
270 		if (n->m_len < oiphlen + tcphlen &&
271 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
272 			goto freeit;
273 		icmpelen = max(tcphlen, min(V_icmp_quotelen, oip->ip_len - oiphlen));
274 	} else
275 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, oip->ip_len - oiphlen));
276 
277 	icmplen = min(oiphlen + icmpelen, nlen);
278 	if (icmplen < sizeof(struct ip))
279 		goto freeit;
280 
281 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
282 		m = m_gethdr(M_DONTWAIT, MT_DATA);
283 	else
284 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
285 	if (m == NULL)
286 		goto freeit;
287 #ifdef MAC
288 	mac_netinet_icmp_reply(n, m);
289 #endif
290 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
291 	m_align(m, ICMP_MINLEN + icmplen);
292 	m->m_len = ICMP_MINLEN + icmplen;
293 
294 	/* XXX MRT  make the outgoing packet use the same FIB
295 	 * that was associated with the incoming packet
296 	 */
297 	M_SETFIB(m, M_GETFIB(n));
298 	icp = mtod(m, struct icmp *);
299 	ICMPSTAT_INC(icps_outhist[type]);
300 	icp->icmp_type = type;
301 	if (type == ICMP_REDIRECT)
302 		icp->icmp_gwaddr.s_addr = dest;
303 	else {
304 		icp->icmp_void = 0;
305 		/*
306 		 * The following assignments assume an overlay with the
307 		 * just zeroed icmp_void field.
308 		 */
309 		if (type == ICMP_PARAMPROB) {
310 			icp->icmp_pptr = code;
311 			code = 0;
312 		} else if (type == ICMP_UNREACH &&
313 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
314 			icp->icmp_nextmtu = htons(mtu);
315 		}
316 	}
317 	icp->icmp_code = code;
318 
319 	/*
320 	 * Copy the quotation into ICMP message and
321 	 * convert quoted IP header back to network representation.
322 	 */
323 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
324 	nip = &icp->icmp_ip;
325 	nip->ip_len = htons(nip->ip_len);
326 	nip->ip_off = htons(nip->ip_off);
327 
328 	/*
329 	 * Set up ICMP message mbuf and copy old IP header (without options
330 	 * in front of ICMP message.
331 	 * If the original mbuf was meant to bypass the firewall, the error
332 	 * reply should bypass as well.
333 	 */
334 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
335 	m->m_data -= sizeof(struct ip);
336 	m->m_len += sizeof(struct ip);
337 	m->m_pkthdr.len = m->m_len;
338 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
339 	nip = mtod(m, struct ip *);
340 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
341 	nip->ip_len = m->m_len;
342 	nip->ip_v = IPVERSION;
343 	nip->ip_hl = 5;
344 	nip->ip_p = IPPROTO_ICMP;
345 	nip->ip_tos = 0;
346 	icmp_reflect(m);
347 
348 freeit:
349 	m_freem(n);
350 }
351 
352 /*
353  * Process a received ICMP message.
354  */
355 void
356 icmp_input(struct mbuf *m, int off)
357 {
358 	struct icmp *icp;
359 	struct in_ifaddr *ia;
360 	struct ip *ip = mtod(m, struct ip *);
361 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
362 	int hlen = off;
363 	int icmplen = ip->ip_len;
364 	int i, code;
365 	void (*ctlfunc)(int, struct sockaddr *, void *);
366 	int fibnum;
367 
368 	/*
369 	 * Locate icmp structure in mbuf, and check
370 	 * that not corrupted and of at least minimum length.
371 	 */
372 #ifdef ICMPPRINTFS
373 	if (icmpprintfs) {
374 		char buf[4 * sizeof "123"];
375 		strcpy(buf, inet_ntoa(ip->ip_src));
376 		printf("icmp_input from %s to %s, len %d\n",
377 		       buf, inet_ntoa(ip->ip_dst), icmplen);
378 	}
379 #endif
380 	if (icmplen < ICMP_MINLEN) {
381 		ICMPSTAT_INC(icps_tooshort);
382 		goto freeit;
383 	}
384 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
385 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
386 		ICMPSTAT_INC(icps_tooshort);
387 		return;
388 	}
389 	ip = mtod(m, struct ip *);
390 	m->m_len -= hlen;
391 	m->m_data += hlen;
392 	icp = mtod(m, struct icmp *);
393 	if (in_cksum(m, icmplen)) {
394 		ICMPSTAT_INC(icps_checksum);
395 		goto freeit;
396 	}
397 	m->m_len += hlen;
398 	m->m_data -= hlen;
399 
400 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
401 		/*
402 		 * Deliver very specific ICMP type only.
403 		 */
404 		switch (icp->icmp_type) {
405 		case ICMP_UNREACH:
406 		case ICMP_TIMXCEED:
407 			break;
408 		default:
409 			goto freeit;
410 		}
411 	}
412 
413 #ifdef ICMPPRINTFS
414 	if (icmpprintfs)
415 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
416 		    icp->icmp_code);
417 #endif
418 
419 	/*
420 	 * Message type specific processing.
421 	 */
422 	if (icp->icmp_type > ICMP_MAXTYPE)
423 		goto raw;
424 
425 	/* Initialize */
426 	bzero(&icmpsrc, sizeof(icmpsrc));
427 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
428 	icmpsrc.sin_family = AF_INET;
429 	bzero(&icmpdst, sizeof(icmpdst));
430 	icmpdst.sin_len = sizeof(struct sockaddr_in);
431 	icmpdst.sin_family = AF_INET;
432 	bzero(&icmpgw, sizeof(icmpgw));
433 	icmpgw.sin_len = sizeof(struct sockaddr_in);
434 	icmpgw.sin_family = AF_INET;
435 
436 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
437 	code = icp->icmp_code;
438 	switch (icp->icmp_type) {
439 
440 	case ICMP_UNREACH:
441 		switch (code) {
442 			case ICMP_UNREACH_NET:
443 			case ICMP_UNREACH_HOST:
444 			case ICMP_UNREACH_SRCFAIL:
445 			case ICMP_UNREACH_NET_UNKNOWN:
446 			case ICMP_UNREACH_HOST_UNKNOWN:
447 			case ICMP_UNREACH_ISOLATED:
448 			case ICMP_UNREACH_TOSNET:
449 			case ICMP_UNREACH_TOSHOST:
450 			case ICMP_UNREACH_HOST_PRECEDENCE:
451 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
452 				code = PRC_UNREACH_NET;
453 				break;
454 
455 			case ICMP_UNREACH_NEEDFRAG:
456 				code = PRC_MSGSIZE;
457 				break;
458 
459 			/*
460 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
461 			 * Treat subcodes 2,3 as immediate RST
462 			 */
463 			case ICMP_UNREACH_PROTOCOL:
464 			case ICMP_UNREACH_PORT:
465 				code = PRC_UNREACH_PORT;
466 				break;
467 
468 			case ICMP_UNREACH_NET_PROHIB:
469 			case ICMP_UNREACH_HOST_PROHIB:
470 			case ICMP_UNREACH_FILTER_PROHIB:
471 				code = PRC_UNREACH_ADMIN_PROHIB;
472 				break;
473 
474 			default:
475 				goto badcode;
476 		}
477 		goto deliver;
478 
479 	case ICMP_TIMXCEED:
480 		if (code > 1)
481 			goto badcode;
482 		code += PRC_TIMXCEED_INTRANS;
483 		goto deliver;
484 
485 	case ICMP_PARAMPROB:
486 		if (code > 1)
487 			goto badcode;
488 		code = PRC_PARAMPROB;
489 		goto deliver;
490 
491 	case ICMP_SOURCEQUENCH:
492 		if (code)
493 			goto badcode;
494 		code = PRC_QUENCH;
495 	deliver:
496 		/*
497 		 * Problem with datagram; advise higher level routines.
498 		 */
499 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
500 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
501 			ICMPSTAT_INC(icps_badlen);
502 			goto freeit;
503 		}
504 		icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
505 		/* Discard ICMP's in response to multicast packets */
506 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
507 			goto badcode;
508 #ifdef ICMPPRINTFS
509 		if (icmpprintfs)
510 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
511 #endif
512 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
513 		/*
514 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
515 		 * notification to TCP layer.
516 		 */
517 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
518 		if (ctlfunc)
519 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
520 				   (void *)&icp->icmp_ip);
521 		break;
522 
523 	badcode:
524 		ICMPSTAT_INC(icps_badcode);
525 		break;
526 
527 	case ICMP_ECHO:
528 		if (!V_icmpbmcastecho
529 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
530 			ICMPSTAT_INC(icps_bmcastecho);
531 			break;
532 		}
533 		icp->icmp_type = ICMP_ECHOREPLY;
534 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
535 			goto freeit;
536 		else
537 			goto reflect;
538 
539 	case ICMP_TSTAMP:
540 		if (!V_icmpbmcastecho
541 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
542 			ICMPSTAT_INC(icps_bmcasttstamp);
543 			break;
544 		}
545 		if (icmplen < ICMP_TSLEN) {
546 			ICMPSTAT_INC(icps_badlen);
547 			break;
548 		}
549 		icp->icmp_type = ICMP_TSTAMPREPLY;
550 		icp->icmp_rtime = iptime();
551 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
552 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
553 			goto freeit;
554 		else
555 			goto reflect;
556 
557 	case ICMP_MASKREQ:
558 		if (V_icmpmaskrepl == 0)
559 			break;
560 		/*
561 		 * We are not able to respond with all ones broadcast
562 		 * unless we receive it over a point-to-point interface.
563 		 */
564 		if (icmplen < ICMP_MASKLEN)
565 			break;
566 		switch (ip->ip_dst.s_addr) {
567 
568 		case INADDR_BROADCAST:
569 		case INADDR_ANY:
570 			icmpdst.sin_addr = ip->ip_src;
571 			break;
572 
573 		default:
574 			icmpdst.sin_addr = ip->ip_dst;
575 		}
576 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
577 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
578 		if (ia == NULL)
579 			break;
580 		if (ia->ia_ifp == NULL) {
581 			ifa_free(&ia->ia_ifa);
582 			break;
583 		}
584 		icp->icmp_type = ICMP_MASKREPLY;
585 		if (V_icmpmaskfake == 0)
586 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
587 		else
588 			icp->icmp_mask = V_icmpmaskfake;
589 		if (ip->ip_src.s_addr == 0) {
590 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
591 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
592 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
593 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
594 		}
595 		ifa_free(&ia->ia_ifa);
596 reflect:
597 		ip->ip_len += hlen;	/* since ip_input deducts this */
598 		ICMPSTAT_INC(icps_reflect);
599 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
600 		icmp_reflect(m);
601 		return;
602 
603 	case ICMP_REDIRECT:
604 		if (V_log_redirect) {
605 			u_long src, dst, gw;
606 
607 			src = ntohl(ip->ip_src.s_addr);
608 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
609 			gw = ntohl(icp->icmp_gwaddr.s_addr);
610 			printf("icmp redirect from %d.%d.%d.%d: "
611 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
612 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
613 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
614 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
615 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
616 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
617 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
618 		}
619 		/*
620 		 * RFC1812 says we must ignore ICMP redirects if we
621 		 * are acting as router.
622 		 */
623 		if (V_drop_redirect || V_ipforwarding)
624 			break;
625 		if (code > 3)
626 			goto badcode;
627 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
628 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
629 			ICMPSTAT_INC(icps_badlen);
630 			break;
631 		}
632 		/*
633 		 * Short circuit routing redirects to force
634 		 * immediate change in the kernel's routing
635 		 * tables.  The message is also handed to anyone
636 		 * listening on a raw socket (e.g. the routing
637 		 * daemon for use in updating its tables).
638 		 */
639 		icmpgw.sin_addr = ip->ip_src;
640 		icmpdst.sin_addr = icp->icmp_gwaddr;
641 #ifdef	ICMPPRINTFS
642 		if (icmpprintfs) {
643 			char buf[4 * sizeof "123"];
644 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
645 
646 			printf("redirect dst %s to %s\n",
647 			       buf, inet_ntoa(icp->icmp_gwaddr));
648 		}
649 #endif
650 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
651 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
652 			in_rtredirect((struct sockaddr *)&icmpsrc,
653 			  (struct sockaddr *)&icmpdst,
654 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
655 			  (struct sockaddr *)&icmpgw, fibnum);
656 		}
657 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
658 #ifdef IPSEC
659 		key_sa_routechange((struct sockaddr *)&icmpsrc);
660 #endif
661 		break;
662 
663 	/*
664 	 * No kernel processing for the following;
665 	 * just fall through to send to raw listener.
666 	 */
667 	case ICMP_ECHOREPLY:
668 	case ICMP_ROUTERADVERT:
669 	case ICMP_ROUTERSOLICIT:
670 	case ICMP_TSTAMPREPLY:
671 	case ICMP_IREQREPLY:
672 	case ICMP_MASKREPLY:
673 	default:
674 		break;
675 	}
676 
677 raw:
678 	rip_input(m, off);
679 	return;
680 
681 freeit:
682 	m_freem(m);
683 }
684 
685 /*
686  * Reflect the ip packet back to the source
687  */
688 static void
689 icmp_reflect(struct mbuf *m)
690 {
691 	struct ip *ip = mtod(m, struct ip *);
692 	struct ifaddr *ifa;
693 	struct ifnet *ifp;
694 	struct in_ifaddr *ia;
695 	struct in_addr t;
696 	struct mbuf *opts = 0;
697 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
698 
699 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
700 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
701 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
702 		m_freem(m);	/* Bad return address */
703 		ICMPSTAT_INC(icps_badaddr);
704 		goto done;	/* Ip_output() will check for broadcast */
705 	}
706 
707 	t = ip->ip_dst;
708 	ip->ip_dst = ip->ip_src;
709 
710 	/*
711 	 * Source selection for ICMP replies:
712 	 *
713 	 * If the incoming packet was addressed directly to one of our
714 	 * own addresses, use dst as the src for the reply.
715 	 */
716 	IN_IFADDR_RLOCK();
717 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
718 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
719 			t = IA_SIN(ia)->sin_addr;
720 			IN_IFADDR_RUNLOCK();
721 			goto match;
722 		}
723 	}
724 	IN_IFADDR_RUNLOCK();
725 
726 	/*
727 	 * If the incoming packet was addressed to one of our broadcast
728 	 * addresses, use the first non-broadcast address which corresponds
729 	 * to the incoming interface.
730 	 */
731 	ifp = m->m_pkthdr.rcvif;
732 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
733 		IF_ADDR_RLOCK(ifp);
734 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
735 			if (ifa->ifa_addr->sa_family != AF_INET)
736 				continue;
737 			ia = ifatoia(ifa);
738 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
739 			    t.s_addr) {
740 				t = IA_SIN(ia)->sin_addr;
741 				IF_ADDR_RUNLOCK(ifp);
742 				goto match;
743 			}
744 		}
745 		IF_ADDR_RUNLOCK(ifp);
746 	}
747 	/*
748 	 * If the packet was transiting through us, use the address of
749 	 * the interface the packet came through in.  If that interface
750 	 * doesn't have a suitable IP address, the normal selection
751 	 * criteria apply.
752 	 */
753 	if (V_icmp_rfi && ifp != NULL) {
754 		IF_ADDR_RLOCK(ifp);
755 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
756 			if (ifa->ifa_addr->sa_family != AF_INET)
757 				continue;
758 			ia = ifatoia(ifa);
759 			t = IA_SIN(ia)->sin_addr;
760 			IF_ADDR_RUNLOCK(ifp);
761 			goto match;
762 		}
763 		IF_ADDR_RUNLOCK(ifp);
764 	}
765 	/*
766 	 * If the incoming packet was not addressed directly to us, use
767 	 * designated interface for icmp replies specified by sysctl
768 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
769 	 * with normal source selection.
770 	 */
771 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
772 		IF_ADDR_RLOCK(ifp);
773 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
774 			if (ifa->ifa_addr->sa_family != AF_INET)
775 				continue;
776 			ia = ifatoia(ifa);
777 			t = IA_SIN(ia)->sin_addr;
778 			IF_ADDR_RUNLOCK(ifp);
779 			goto match;
780 		}
781 		IF_ADDR_RUNLOCK(ifp);
782 	}
783 	/*
784 	 * If the packet was transiting through us, use the address of
785 	 * the interface that is the closest to the packet source.
786 	 * When we don't have a route back to the packet source, stop here
787 	 * and drop the packet.
788 	 */
789 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
790 	if (ia == NULL) {
791 		m_freem(m);
792 		ICMPSTAT_INC(icps_noroute);
793 		goto done;
794 	}
795 	t = IA_SIN(ia)->sin_addr;
796 	ifa_free(&ia->ia_ifa);
797 match:
798 #ifdef MAC
799 	mac_netinet_icmp_replyinplace(m);
800 #endif
801 	ip->ip_src = t;
802 	ip->ip_ttl = V_ip_defttl;
803 
804 	if (optlen > 0) {
805 		register u_char *cp;
806 		int opt, cnt;
807 		u_int len;
808 
809 		/*
810 		 * Retrieve any source routing from the incoming packet;
811 		 * add on any record-route or timestamp options.
812 		 */
813 		cp = (u_char *) (ip + 1);
814 		if ((opts = ip_srcroute(m)) == 0 &&
815 		    (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
816 			opts->m_len = sizeof(struct in_addr);
817 			mtod(opts, struct in_addr *)->s_addr = 0;
818 		}
819 		if (opts) {
820 #ifdef ICMPPRINTFS
821 		    if (icmpprintfs)
822 			    printf("icmp_reflect optlen %d rt %d => ",
823 				optlen, opts->m_len);
824 #endif
825 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
826 			    opt = cp[IPOPT_OPTVAL];
827 			    if (opt == IPOPT_EOL)
828 				    break;
829 			    if (opt == IPOPT_NOP)
830 				    len = 1;
831 			    else {
832 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
833 					    break;
834 				    len = cp[IPOPT_OLEN];
835 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
836 				        len > cnt)
837 					    break;
838 			    }
839 			    /*
840 			     * Should check for overflow, but it "can't happen"
841 			     */
842 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
843 				opt == IPOPT_SECURITY) {
844 				    bcopy((caddr_t)cp,
845 					mtod(opts, caddr_t) + opts->m_len, len);
846 				    opts->m_len += len;
847 			    }
848 		    }
849 		    /* Terminate & pad, if necessary */
850 		    cnt = opts->m_len % 4;
851 		    if (cnt) {
852 			    for (; cnt < 4; cnt++) {
853 				    *(mtod(opts, caddr_t) + opts->m_len) =
854 					IPOPT_EOL;
855 				    opts->m_len++;
856 			    }
857 		    }
858 #ifdef ICMPPRINTFS
859 		    if (icmpprintfs)
860 			    printf("%d\n", opts->m_len);
861 #endif
862 		}
863 		/*
864 		 * Now strip out original options by copying rest of first
865 		 * mbuf's data back, and adjust the IP length.
866 		 */
867 		ip->ip_len -= optlen;
868 		ip->ip_v = IPVERSION;
869 		ip->ip_hl = 5;
870 		m->m_len -= optlen;
871 		if (m->m_flags & M_PKTHDR)
872 			m->m_pkthdr.len -= optlen;
873 		optlen += sizeof(struct ip);
874 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
875 			 (unsigned)(m->m_len - sizeof(struct ip)));
876 	}
877 	m_tag_delete_nonpersistent(m);
878 	m->m_flags &= ~(M_BCAST|M_MCAST);
879 	icmp_send(m, opts);
880 done:
881 	if (opts)
882 		(void)m_free(opts);
883 }
884 
885 /*
886  * Send an icmp packet back to the ip level,
887  * after supplying a checksum.
888  */
889 static void
890 icmp_send(struct mbuf *m, struct mbuf *opts)
891 {
892 	register struct ip *ip = mtod(m, struct ip *);
893 	register int hlen;
894 	register struct icmp *icp;
895 
896 	hlen = ip->ip_hl << 2;
897 	m->m_data += hlen;
898 	m->m_len -= hlen;
899 	icp = mtod(m, struct icmp *);
900 	icp->icmp_cksum = 0;
901 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
902 	m->m_data -= hlen;
903 	m->m_len += hlen;
904 	m->m_pkthdr.rcvif = (struct ifnet *)0;
905 #ifdef ICMPPRINTFS
906 	if (icmpprintfs) {
907 		char buf[4 * sizeof "123"];
908 		strcpy(buf, inet_ntoa(ip->ip_dst));
909 		printf("icmp_send dst %s src %s\n",
910 		       buf, inet_ntoa(ip->ip_src));
911 	}
912 #endif
913 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
914 }
915 
916 /*
917  * Return milliseconds since 00:00 GMT in network format.
918  */
919 uint32_t
920 iptime(void)
921 {
922 	struct timeval atv;
923 	u_long t;
924 
925 	getmicrotime(&atv);
926 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
927 	return (htonl(t));
928 }
929 
930 /*
931  * Return the next larger or smaller MTU plateau (table from RFC 1191)
932  * given current value MTU.  If DIR is less than zero, a larger plateau
933  * is returned; otherwise, a smaller value is returned.
934  */
935 int
936 ip_next_mtu(int mtu, int dir)
937 {
938 	static int mtutab[] = {
939 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
940 		296, 68, 0
941 	};
942 	int i, size;
943 
944 	size = (sizeof mtutab) / (sizeof mtutab[0]);
945 	if (dir >= 0) {
946 		for (i = 0; i < size; i++)
947 			if (mtu > mtutab[i])
948 				return mtutab[i];
949 	} else {
950 		for (i = size - 1; i >= 0; i--)
951 			if (mtu < mtutab[i])
952 				return mtutab[i];
953 		if (mtu == mtutab[0])
954 			return mtutab[0];
955 	}
956 	return 0;
957 }
958 #endif /* INET */
959 
960 
961 /*
962  * badport_bandlim() - check for ICMP bandwidth limit
963  *
964  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
965  *	hit our bandwidth limit and it is not ok.
966  *
967  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
968  *
969  *	For now we separate the TCP and UDP subsystems w/ different 'which'
970  *	values.  We may eventually remove this separation (and simplify the
971  *	code further).
972  *
973  *	Note that the printing of the error message is delayed so we can
974  *	properly print the icmp error rate that the system was trying to do
975  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
976  *	the 'final' error, but it doesn't make sense to solve the printing
977  *	delay with more complex code.
978  */
979 
980 int
981 badport_bandlim(int which)
982 {
983 
984 #define	N(a)	(sizeof (a) / sizeof (a[0]))
985 	static struct rate {
986 		const char	*type;
987 		struct timeval	lasttime;
988 		int		curpps;
989 	} rates[BANDLIM_MAX+1] = {
990 		{ "icmp unreach response" },
991 		{ "icmp ping response" },
992 		{ "icmp tstamp response" },
993 		{ "closed port RST response" },
994 		{ "open port RST response" },
995 		{ "icmp6 unreach response" },
996 		{ "sctp ootb response" }
997 	};
998 
999 	/*
1000 	 * Return ok status if feature disabled or argument out of range.
1001 	 */
1002 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
1003 		struct rate *r = &rates[which];
1004 		int opps = r->curpps;
1005 
1006 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
1007 			return -1;	/* discard packet */
1008 		/*
1009 		 * If we've dropped below the threshold after having
1010 		 * rate-limited traffic print the message.  This preserves
1011 		 * the previous behaviour at the expense of added complexity.
1012 		 */
1013 		if (V_icmplim_output && opps > V_icmplim)
1014 			log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
1015 				r->type, opps, V_icmplim);
1016 	}
1017 	return 0;			/* okay to send packet */
1018 #undef N
1019 }
1020