xref: /freebsd/sys/netinet/ip_icmp.c (revision 0572ccaa4543b0abef8ef81e384c1d04de9f3da1)
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_var.h>
50 #include <net/if_types.h>
51 #include <net/route.h>
52 #include <net/vnet.h>
53 
54 #include <netinet/in.h>
55 #include <netinet/in_pcb.h>
56 #include <netinet/in_systm.h>
57 #include <netinet/in_var.h>
58 #include <netinet/ip.h>
59 #include <netinet/ip_icmp.h>
60 #include <netinet/ip_var.h>
61 #include <netinet/ip_options.h>
62 #include <netinet/tcp.h>
63 #include <netinet/tcp_var.h>
64 #include <netinet/tcpip.h>
65 #include <netinet/icmp_var.h>
66 
67 #ifdef INET
68 #ifdef IPSEC
69 #include <netipsec/ipsec.h>
70 #include <netipsec/key.h>
71 #endif
72 
73 #include <machine/in_cksum.h>
74 
75 #include <security/mac/mac_framework.h>
76 #endif /* INET */
77 
78 /*
79  * ICMP routines: error generation, receive packet processing, and
80  * routines to turnaround packets back to the originator, and
81  * host table maintenance routines.
82  */
83 static VNET_DEFINE(int, icmplim) = 200;
84 #define	V_icmplim			VNET(icmplim)
85 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
86 	&VNET_NAME(icmplim), 0,
87 	"Maximum number of ICMP responses per second");
88 
89 static VNET_DEFINE(int, icmplim_output) = 1;
90 #define	V_icmplim_output		VNET(icmplim_output)
91 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
92 	&VNET_NAME(icmplim_output), 0,
93 	"Enable logging of ICMP response rate limiting");
94 
95 #ifdef INET
96 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
97 VNET_PCPUSTAT_SYSINIT(icmpstat);
98 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
99     icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
100 
101 #ifdef VIMAGE
102 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
103 #endif /* VIMAGE */
104 
105 static VNET_DEFINE(int, icmpmaskrepl) = 0;
106 #define	V_icmpmaskrepl			VNET(icmpmaskrepl)
107 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
108 	&VNET_NAME(icmpmaskrepl), 0,
109 	"Reply to ICMP Address Mask Request packets.");
110 
111 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
112 #define	V_icmpmaskfake			VNET(icmpmaskfake)
113 SYSCTL_VNET_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
114 	&VNET_NAME(icmpmaskfake), 0,
115 	"Fake reply to ICMP Address Mask Request packets.");
116 
117 VNET_DEFINE(int, drop_redirect) = 0;
118 
119 static VNET_DEFINE(int, log_redirect) = 0;
120 #define	V_log_redirect			VNET(log_redirect)
121 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
122 	&VNET_NAME(log_redirect), 0,
123 	"Log ICMP redirects to the console");
124 
125 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
126 #define	V_reply_src			VNET(reply_src)
127 SYSCTL_VNET_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
128 	&VNET_NAME(reply_src), IFNAMSIZ,
129 	"icmp reply source for non-local packets.");
130 
131 static VNET_DEFINE(int, icmp_rfi) = 0;
132 #define	V_icmp_rfi			VNET(icmp_rfi)
133 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
134 	&VNET_NAME(icmp_rfi), 0,
135 	"ICMP reply from incoming interface for non-local packets");
136 
137 static VNET_DEFINE(int, icmp_quotelen) = 8;
138 #define	V_icmp_quotelen			VNET(icmp_quotelen)
139 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
140 	&VNET_NAME(icmp_quotelen), 0,
141 	"Number of bytes from original packet to quote in ICMP reply");
142 
143 /*
144  * ICMP broadcast echo sysctl
145  */
146 static VNET_DEFINE(int, icmpbmcastecho) = 0;
147 #define	V_icmpbmcastecho		VNET(icmpbmcastecho)
148 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
149 	&VNET_NAME(icmpbmcastecho), 0,
150 	"");
151 
152 
153 #ifdef ICMPPRINTFS
154 int	icmpprintfs = 0;
155 #endif
156 
157 static void	icmp_reflect(struct mbuf *);
158 static void	icmp_send(struct mbuf *, struct mbuf *);
159 
160 extern	struct protosw inetsw[];
161 
162 static int
163 sysctl_net_icmp_drop_redir(SYSCTL_HANDLER_ARGS)
164 {
165 	int error, new;
166 	int i;
167 	struct radix_node_head *rnh;
168 
169 	new = V_drop_redirect;
170 	error = sysctl_handle_int(oidp, &new, 0, req);
171 	if (error == 0 && req->newptr) {
172 		new = (new != 0) ? 1 : 0;
173 
174 		if (new == V_drop_redirect)
175 			return (0);
176 
177 		for (i = 0; i < rt_numfibs; i++) {
178 			if ((rnh = rt_tables_get_rnh(i, AF_INET)) == NULL)
179 				continue;
180 			RADIX_NODE_HEAD_LOCK(rnh);
181 			in_setmatchfunc(rnh, new);
182 			RADIX_NODE_HEAD_UNLOCK(rnh);
183 		}
184 
185 		V_drop_redirect = new;
186 	}
187 
188 	return (error);
189 }
190 
191 SYSCTL_VNET_PROC(_net_inet_icmp, OID_AUTO, drop_redirect,
192     CTLTYPE_INT|CTLFLAG_RW, 0, 0,
193     sysctl_net_icmp_drop_redir, "I", "Ignore ICMP redirects");
194 
195 /*
196  * Kernel module interface for updating icmpstat.  The argument is an index
197  * into icmpstat treated as an array of u_long.  While this encodes the
198  * general layout of icmpstat into the caller, it doesn't encode its
199  * location, so that future changes to add, for example, per-CPU stats
200  * support won't cause binary compatibility problems for kernel modules.
201  */
202 void
203 kmod_icmpstat_inc(int statnum)
204 {
205 
206 	counter_u64_add(VNET(icmpstat)[statnum], 1);
207 }
208 
209 /*
210  * Generate an error packet of type error
211  * in response to bad packet ip.
212  */
213 void
214 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
215 {
216 	register struct ip *oip = mtod(n, struct ip *), *nip;
217 	register unsigned oiphlen = oip->ip_hl << 2;
218 	register struct icmp *icp;
219 	register struct mbuf *m;
220 	unsigned icmplen, icmpelen, nlen;
221 
222 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
223 #ifdef ICMPPRINTFS
224 	if (icmpprintfs)
225 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
226 #endif
227 	if (type != ICMP_REDIRECT)
228 		ICMPSTAT_INC(icps_error);
229 	/*
230 	 * Don't send error:
231 	 *  if the original packet was encrypted.
232 	 *  if not the first fragment of message.
233 	 *  in response to a multicast or broadcast packet.
234 	 *  if the old packet protocol was an ICMP error message.
235 	 */
236 	if (n->m_flags & M_DECRYPTED)
237 		goto freeit;
238 	if (oip->ip_off & htons(~(IP_MF|IP_DF)))
239 		goto freeit;
240 	if (n->m_flags & (M_BCAST|M_MCAST))
241 		goto freeit;
242 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
243 	  n->m_len >= oiphlen + ICMP_MINLEN &&
244 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
245 		ICMPSTAT_INC(icps_oldicmp);
246 		goto freeit;
247 	}
248 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
249 	if (oiphlen + 8 > n->m_len)
250 		goto freeit;
251 	/*
252 	 * Calculate length to quote from original packet and
253 	 * prevent the ICMP mbuf from overflowing.
254 	 * Unfortunatly this is non-trivial since ip_forward()
255 	 * sends us truncated packets.
256 	 */
257 	nlen = m_length(n, NULL);
258 	if (oip->ip_p == IPPROTO_TCP) {
259 		struct tcphdr *th;
260 		int tcphlen;
261 
262 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
263 		    n->m_next == NULL)
264 			goto stdreply;
265 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
266 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
267 			goto freeit;
268 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
269 		tcphlen = th->th_off << 2;
270 		if (tcphlen < sizeof(struct tcphdr))
271 			goto freeit;
272 		if (ntohs(oip->ip_len) < oiphlen + tcphlen)
273 			goto freeit;
274 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
275 			goto stdreply;
276 		if (n->m_len < oiphlen + tcphlen &&
277 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
278 			goto freeit;
279 		icmpelen = max(tcphlen, min(V_icmp_quotelen,
280 		    ntohs(oip->ip_len) - oiphlen));
281 	} else
282 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
283 
284 	icmplen = min(oiphlen + icmpelen, nlen);
285 	if (icmplen < sizeof(struct ip))
286 		goto freeit;
287 
288 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
289 		m = m_gethdr(M_NOWAIT, MT_DATA);
290 	else
291 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
292 	if (m == NULL)
293 		goto freeit;
294 #ifdef MAC
295 	mac_netinet_icmp_reply(n, m);
296 #endif
297 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
298 	m_align(m, ICMP_MINLEN + icmplen);
299 	m->m_len = ICMP_MINLEN + icmplen;
300 
301 	/* XXX MRT  make the outgoing packet use the same FIB
302 	 * that was associated with the incoming packet
303 	 */
304 	M_SETFIB(m, M_GETFIB(n));
305 	icp = mtod(m, struct icmp *);
306 	ICMPSTAT_INC(icps_outhist[type]);
307 	icp->icmp_type = type;
308 	if (type == ICMP_REDIRECT)
309 		icp->icmp_gwaddr.s_addr = dest;
310 	else {
311 		icp->icmp_void = 0;
312 		/*
313 		 * The following assignments assume an overlay with the
314 		 * just zeroed icmp_void field.
315 		 */
316 		if (type == ICMP_PARAMPROB) {
317 			icp->icmp_pptr = code;
318 			code = 0;
319 		} else if (type == ICMP_UNREACH &&
320 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
321 			icp->icmp_nextmtu = htons(mtu);
322 		}
323 	}
324 	icp->icmp_code = code;
325 
326 	/*
327 	 * Copy the quotation into ICMP message and
328 	 * convert quoted IP header back to network representation.
329 	 */
330 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
331 	nip = &icp->icmp_ip;
332 
333 	/*
334 	 * Set up ICMP message mbuf and copy old IP header (without options
335 	 * in front of ICMP message.
336 	 * If the original mbuf was meant to bypass the firewall, the error
337 	 * reply should bypass as well.
338 	 */
339 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
340 	m->m_data -= sizeof(struct ip);
341 	m->m_len += sizeof(struct ip);
342 	m->m_pkthdr.len = m->m_len;
343 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
344 	nip = mtod(m, struct ip *);
345 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
346 	nip->ip_len = htons(m->m_len);
347 	nip->ip_v = IPVERSION;
348 	nip->ip_hl = 5;
349 	nip->ip_p = IPPROTO_ICMP;
350 	nip->ip_tos = 0;
351 	nip->ip_off = 0;
352 	icmp_reflect(m);
353 
354 freeit:
355 	m_freem(n);
356 }
357 
358 /*
359  * Process a received ICMP message.
360  */
361 void
362 icmp_input(struct mbuf *m, int off)
363 {
364 	struct icmp *icp;
365 	struct in_ifaddr *ia;
366 	struct ip *ip = mtod(m, struct ip *);
367 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
368 	int hlen = off;
369 	int icmplen = ntohs(ip->ip_len) - off;
370 	int i, code;
371 	void (*ctlfunc)(int, struct sockaddr *, void *);
372 	int fibnum;
373 
374 	/*
375 	 * Locate icmp structure in mbuf, and check
376 	 * that not corrupted and of at least minimum length.
377 	 */
378 #ifdef ICMPPRINTFS
379 	if (icmpprintfs) {
380 		char buf[4 * sizeof "123"];
381 		strcpy(buf, inet_ntoa(ip->ip_src));
382 		printf("icmp_input from %s to %s, len %d\n",
383 		       buf, inet_ntoa(ip->ip_dst), icmplen);
384 	}
385 #endif
386 	if (icmplen < ICMP_MINLEN) {
387 		ICMPSTAT_INC(icps_tooshort);
388 		goto freeit;
389 	}
390 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
391 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
392 		ICMPSTAT_INC(icps_tooshort);
393 		return;
394 	}
395 	ip = mtod(m, struct ip *);
396 	m->m_len -= hlen;
397 	m->m_data += hlen;
398 	icp = mtod(m, struct icmp *);
399 	if (in_cksum(m, icmplen)) {
400 		ICMPSTAT_INC(icps_checksum);
401 		goto freeit;
402 	}
403 	m->m_len += hlen;
404 	m->m_data -= hlen;
405 
406 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
407 		/*
408 		 * Deliver very specific ICMP type only.
409 		 */
410 		switch (icp->icmp_type) {
411 		case ICMP_UNREACH:
412 		case ICMP_TIMXCEED:
413 			break;
414 		default:
415 			goto freeit;
416 		}
417 	}
418 
419 #ifdef ICMPPRINTFS
420 	if (icmpprintfs)
421 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
422 		    icp->icmp_code);
423 #endif
424 
425 	/*
426 	 * Message type specific processing.
427 	 */
428 	if (icp->icmp_type > ICMP_MAXTYPE)
429 		goto raw;
430 
431 	/* Initialize */
432 	bzero(&icmpsrc, sizeof(icmpsrc));
433 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
434 	icmpsrc.sin_family = AF_INET;
435 	bzero(&icmpdst, sizeof(icmpdst));
436 	icmpdst.sin_len = sizeof(struct sockaddr_in);
437 	icmpdst.sin_family = AF_INET;
438 	bzero(&icmpgw, sizeof(icmpgw));
439 	icmpgw.sin_len = sizeof(struct sockaddr_in);
440 	icmpgw.sin_family = AF_INET;
441 
442 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
443 	code = icp->icmp_code;
444 	switch (icp->icmp_type) {
445 
446 	case ICMP_UNREACH:
447 		switch (code) {
448 			case ICMP_UNREACH_NET:
449 			case ICMP_UNREACH_HOST:
450 			case ICMP_UNREACH_SRCFAIL:
451 			case ICMP_UNREACH_NET_UNKNOWN:
452 			case ICMP_UNREACH_HOST_UNKNOWN:
453 			case ICMP_UNREACH_ISOLATED:
454 			case ICMP_UNREACH_TOSNET:
455 			case ICMP_UNREACH_TOSHOST:
456 			case ICMP_UNREACH_HOST_PRECEDENCE:
457 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
458 				code = PRC_UNREACH_NET;
459 				break;
460 
461 			case ICMP_UNREACH_NEEDFRAG:
462 				code = PRC_MSGSIZE;
463 				break;
464 
465 			/*
466 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
467 			 * Treat subcodes 2,3 as immediate RST
468 			 */
469 			case ICMP_UNREACH_PROTOCOL:
470 			case ICMP_UNREACH_PORT:
471 				code = PRC_UNREACH_PORT;
472 				break;
473 
474 			case ICMP_UNREACH_NET_PROHIB:
475 			case ICMP_UNREACH_HOST_PROHIB:
476 			case ICMP_UNREACH_FILTER_PROHIB:
477 				code = PRC_UNREACH_ADMIN_PROHIB;
478 				break;
479 
480 			default:
481 				goto badcode;
482 		}
483 		goto deliver;
484 
485 	case ICMP_TIMXCEED:
486 		if (code > 1)
487 			goto badcode;
488 		code += PRC_TIMXCEED_INTRANS;
489 		goto deliver;
490 
491 	case ICMP_PARAMPROB:
492 		if (code > 1)
493 			goto badcode;
494 		code = PRC_PARAMPROB;
495 		goto deliver;
496 
497 	case ICMP_SOURCEQUENCH:
498 		if (code)
499 			goto badcode;
500 		code = PRC_QUENCH;
501 	deliver:
502 		/*
503 		 * Problem with datagram; advise higher level routines.
504 		 */
505 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
506 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
507 			ICMPSTAT_INC(icps_badlen);
508 			goto freeit;
509 		}
510 		/* Discard ICMP's in response to multicast packets */
511 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
512 			goto badcode;
513 #ifdef ICMPPRINTFS
514 		if (icmpprintfs)
515 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
516 #endif
517 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
518 		/*
519 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
520 		 * notification to TCP layer.
521 		 */
522 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
523 		if (ctlfunc)
524 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
525 				   (void *)&icp->icmp_ip);
526 		break;
527 
528 	badcode:
529 		ICMPSTAT_INC(icps_badcode);
530 		break;
531 
532 	case ICMP_ECHO:
533 		if (!V_icmpbmcastecho
534 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
535 			ICMPSTAT_INC(icps_bmcastecho);
536 			break;
537 		}
538 		icp->icmp_type = ICMP_ECHOREPLY;
539 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
540 			goto freeit;
541 		else
542 			goto reflect;
543 
544 	case ICMP_TSTAMP:
545 		if (!V_icmpbmcastecho
546 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
547 			ICMPSTAT_INC(icps_bmcasttstamp);
548 			break;
549 		}
550 		if (icmplen < ICMP_TSLEN) {
551 			ICMPSTAT_INC(icps_badlen);
552 			break;
553 		}
554 		icp->icmp_type = ICMP_TSTAMPREPLY;
555 		icp->icmp_rtime = iptime();
556 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
557 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
558 			goto freeit;
559 		else
560 			goto reflect;
561 
562 	case ICMP_MASKREQ:
563 		if (V_icmpmaskrepl == 0)
564 			break;
565 		/*
566 		 * We are not able to respond with all ones broadcast
567 		 * unless we receive it over a point-to-point interface.
568 		 */
569 		if (icmplen < ICMP_MASKLEN)
570 			break;
571 		switch (ip->ip_dst.s_addr) {
572 
573 		case INADDR_BROADCAST:
574 		case INADDR_ANY:
575 			icmpdst.sin_addr = ip->ip_src;
576 			break;
577 
578 		default:
579 			icmpdst.sin_addr = ip->ip_dst;
580 		}
581 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
582 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
583 		if (ia == NULL)
584 			break;
585 		if (ia->ia_ifp == NULL) {
586 			ifa_free(&ia->ia_ifa);
587 			break;
588 		}
589 		icp->icmp_type = ICMP_MASKREPLY;
590 		if (V_icmpmaskfake == 0)
591 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
592 		else
593 			icp->icmp_mask = V_icmpmaskfake;
594 		if (ip->ip_src.s_addr == 0) {
595 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
596 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
597 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
598 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
599 		}
600 		ifa_free(&ia->ia_ifa);
601 reflect:
602 		ICMPSTAT_INC(icps_reflect);
603 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
604 		icmp_reflect(m);
605 		return;
606 
607 	case ICMP_REDIRECT:
608 		if (V_log_redirect) {
609 			u_long src, dst, gw;
610 
611 			src = ntohl(ip->ip_src.s_addr);
612 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
613 			gw = ntohl(icp->icmp_gwaddr.s_addr);
614 			printf("icmp redirect from %d.%d.%d.%d: "
615 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
616 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
617 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
618 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
619 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
620 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
621 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
622 		}
623 		/*
624 		 * RFC1812 says we must ignore ICMP redirects if we
625 		 * are acting as router.
626 		 */
627 		if (V_drop_redirect || V_ipforwarding)
628 			break;
629 		if (code > 3)
630 			goto badcode;
631 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
632 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
633 			ICMPSTAT_INC(icps_badlen);
634 			break;
635 		}
636 		/*
637 		 * Short circuit routing redirects to force
638 		 * immediate change in the kernel's routing
639 		 * tables.  The message is also handed to anyone
640 		 * listening on a raw socket (e.g. the routing
641 		 * daemon for use in updating its tables).
642 		 */
643 		icmpgw.sin_addr = ip->ip_src;
644 		icmpdst.sin_addr = icp->icmp_gwaddr;
645 #ifdef	ICMPPRINTFS
646 		if (icmpprintfs) {
647 			char buf[4 * sizeof "123"];
648 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
649 
650 			printf("redirect dst %s to %s\n",
651 			       buf, inet_ntoa(icp->icmp_gwaddr));
652 		}
653 #endif
654 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
655 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
656 			in_rtredirect((struct sockaddr *)&icmpsrc,
657 			  (struct sockaddr *)&icmpdst,
658 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
659 			  (struct sockaddr *)&icmpgw, fibnum);
660 		}
661 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
662 #ifdef IPSEC
663 		key_sa_routechange((struct sockaddr *)&icmpsrc);
664 #endif
665 		break;
666 
667 	/*
668 	 * No kernel processing for the following;
669 	 * just fall through to send to raw listener.
670 	 */
671 	case ICMP_ECHOREPLY:
672 	case ICMP_ROUTERADVERT:
673 	case ICMP_ROUTERSOLICIT:
674 	case ICMP_TSTAMPREPLY:
675 	case ICMP_IREQREPLY:
676 	case ICMP_MASKREPLY:
677 	default:
678 		break;
679 	}
680 
681 raw:
682 	rip_input(m, off);
683 	return;
684 
685 freeit:
686 	m_freem(m);
687 }
688 
689 /*
690  * Reflect the ip packet back to the source
691  */
692 static void
693 icmp_reflect(struct mbuf *m)
694 {
695 	struct ip *ip = mtod(m, struct ip *);
696 	struct ifaddr *ifa;
697 	struct ifnet *ifp;
698 	struct in_ifaddr *ia;
699 	struct in_addr t;
700 	struct mbuf *opts = 0;
701 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
702 
703 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
704 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
705 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
706 		m_freem(m);	/* Bad return address */
707 		ICMPSTAT_INC(icps_badaddr);
708 		goto done;	/* Ip_output() will check for broadcast */
709 	}
710 
711 	t = ip->ip_dst;
712 	ip->ip_dst = ip->ip_src;
713 
714 	/*
715 	 * Source selection for ICMP replies:
716 	 *
717 	 * If the incoming packet was addressed directly to one of our
718 	 * own addresses, use dst as the src for the reply.
719 	 */
720 	IN_IFADDR_RLOCK();
721 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
722 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
723 			t = IA_SIN(ia)->sin_addr;
724 			IN_IFADDR_RUNLOCK();
725 			goto match;
726 		}
727 	}
728 	IN_IFADDR_RUNLOCK();
729 
730 	/*
731 	 * If the incoming packet was addressed to one of our broadcast
732 	 * addresses, use the first non-broadcast address which corresponds
733 	 * to the incoming interface.
734 	 */
735 	ifp = m->m_pkthdr.rcvif;
736 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
737 		IF_ADDR_RLOCK(ifp);
738 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
739 			if (ifa->ifa_addr->sa_family != AF_INET)
740 				continue;
741 			ia = ifatoia(ifa);
742 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
743 			    t.s_addr) {
744 				t = IA_SIN(ia)->sin_addr;
745 				IF_ADDR_RUNLOCK(ifp);
746 				goto match;
747 			}
748 		}
749 		IF_ADDR_RUNLOCK(ifp);
750 	}
751 	/*
752 	 * If the packet was transiting through us, use the address of
753 	 * the interface the packet came through in.  If that interface
754 	 * doesn't have a suitable IP address, the normal selection
755 	 * criteria apply.
756 	 */
757 	if (V_icmp_rfi && ifp != NULL) {
758 		IF_ADDR_RLOCK(ifp);
759 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
760 			if (ifa->ifa_addr->sa_family != AF_INET)
761 				continue;
762 			ia = ifatoia(ifa);
763 			t = IA_SIN(ia)->sin_addr;
764 			IF_ADDR_RUNLOCK(ifp);
765 			goto match;
766 		}
767 		IF_ADDR_RUNLOCK(ifp);
768 	}
769 	/*
770 	 * If the incoming packet was not addressed directly to us, use
771 	 * designated interface for icmp replies specified by sysctl
772 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
773 	 * with normal source selection.
774 	 */
775 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
776 		IF_ADDR_RLOCK(ifp);
777 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
778 			if (ifa->ifa_addr->sa_family != AF_INET)
779 				continue;
780 			ia = ifatoia(ifa);
781 			t = IA_SIN(ia)->sin_addr;
782 			IF_ADDR_RUNLOCK(ifp);
783 			goto match;
784 		}
785 		IF_ADDR_RUNLOCK(ifp);
786 	}
787 	/*
788 	 * If the packet was transiting through us, use the address of
789 	 * the interface that is the closest to the packet source.
790 	 * When we don't have a route back to the packet source, stop here
791 	 * and drop the packet.
792 	 */
793 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
794 	if (ia == NULL) {
795 		m_freem(m);
796 		ICMPSTAT_INC(icps_noroute);
797 		goto done;
798 	}
799 	t = IA_SIN(ia)->sin_addr;
800 	ifa_free(&ia->ia_ifa);
801 match:
802 #ifdef MAC
803 	mac_netinet_icmp_replyinplace(m);
804 #endif
805 	ip->ip_src = t;
806 	ip->ip_ttl = V_ip_defttl;
807 
808 	if (optlen > 0) {
809 		register u_char *cp;
810 		int opt, cnt;
811 		u_int len;
812 
813 		/*
814 		 * Retrieve any source routing from the incoming packet;
815 		 * add on any record-route or timestamp options.
816 		 */
817 		cp = (u_char *) (ip + 1);
818 		if ((opts = ip_srcroute(m)) == 0 &&
819 		    (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
820 			opts->m_len = sizeof(struct in_addr);
821 			mtod(opts, struct in_addr *)->s_addr = 0;
822 		}
823 		if (opts) {
824 #ifdef ICMPPRINTFS
825 		    if (icmpprintfs)
826 			    printf("icmp_reflect optlen %d rt %d => ",
827 				optlen, opts->m_len);
828 #endif
829 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
830 			    opt = cp[IPOPT_OPTVAL];
831 			    if (opt == IPOPT_EOL)
832 				    break;
833 			    if (opt == IPOPT_NOP)
834 				    len = 1;
835 			    else {
836 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
837 					    break;
838 				    len = cp[IPOPT_OLEN];
839 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
840 				        len > cnt)
841 					    break;
842 			    }
843 			    /*
844 			     * Should check for overflow, but it "can't happen"
845 			     */
846 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
847 				opt == IPOPT_SECURITY) {
848 				    bcopy((caddr_t)cp,
849 					mtod(opts, caddr_t) + opts->m_len, len);
850 				    opts->m_len += len;
851 			    }
852 		    }
853 		    /* Terminate & pad, if necessary */
854 		    cnt = opts->m_len % 4;
855 		    if (cnt) {
856 			    for (; cnt < 4; cnt++) {
857 				    *(mtod(opts, caddr_t) + opts->m_len) =
858 					IPOPT_EOL;
859 				    opts->m_len++;
860 			    }
861 		    }
862 #ifdef ICMPPRINTFS
863 		    if (icmpprintfs)
864 			    printf("%d\n", opts->m_len);
865 #endif
866 		}
867 		ip_stripoptions(m);
868 	}
869 	m_tag_delete_nonpersistent(m);
870 	m->m_flags &= ~(M_BCAST|M_MCAST);
871 	icmp_send(m, opts);
872 done:
873 	if (opts)
874 		(void)m_free(opts);
875 }
876 
877 /*
878  * Send an icmp packet back to the ip level,
879  * after supplying a checksum.
880  */
881 static void
882 icmp_send(struct mbuf *m, struct mbuf *opts)
883 {
884 	register struct ip *ip = mtod(m, struct ip *);
885 	register int hlen;
886 	register struct icmp *icp;
887 
888 	hlen = ip->ip_hl << 2;
889 	m->m_data += hlen;
890 	m->m_len -= hlen;
891 	icp = mtod(m, struct icmp *);
892 	icp->icmp_cksum = 0;
893 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
894 	m->m_data -= hlen;
895 	m->m_len += hlen;
896 	m->m_pkthdr.rcvif = (struct ifnet *)0;
897 #ifdef ICMPPRINTFS
898 	if (icmpprintfs) {
899 		char buf[4 * sizeof "123"];
900 		strcpy(buf, inet_ntoa(ip->ip_dst));
901 		printf("icmp_send dst %s src %s\n",
902 		       buf, inet_ntoa(ip->ip_src));
903 	}
904 #endif
905 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
906 }
907 
908 /*
909  * Return milliseconds since 00:00 GMT in network format.
910  */
911 uint32_t
912 iptime(void)
913 {
914 	struct timeval atv;
915 	u_long t;
916 
917 	getmicrotime(&atv);
918 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
919 	return (htonl(t));
920 }
921 
922 /*
923  * Return the next larger or smaller MTU plateau (table from RFC 1191)
924  * given current value MTU.  If DIR is less than zero, a larger plateau
925  * is returned; otherwise, a smaller value is returned.
926  */
927 int
928 ip_next_mtu(int mtu, int dir)
929 {
930 	static int mtutab[] = {
931 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
932 		296, 68, 0
933 	};
934 	int i, size;
935 
936 	size = (sizeof mtutab) / (sizeof mtutab[0]);
937 	if (dir >= 0) {
938 		for (i = 0; i < size; i++)
939 			if (mtu > mtutab[i])
940 				return mtutab[i];
941 	} else {
942 		for (i = size - 1; i >= 0; i--)
943 			if (mtu < mtutab[i])
944 				return mtutab[i];
945 		if (mtu == mtutab[0])
946 			return mtutab[0];
947 	}
948 	return 0;
949 }
950 #endif /* INET */
951 
952 
953 /*
954  * badport_bandlim() - check for ICMP bandwidth limit
955  *
956  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
957  *	hit our bandwidth limit and it is not ok.
958  *
959  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
960  *
961  *	For now we separate the TCP and UDP subsystems w/ different 'which'
962  *	values.  We may eventually remove this separation (and simplify the
963  *	code further).
964  *
965  *	Note that the printing of the error message is delayed so we can
966  *	properly print the icmp error rate that the system was trying to do
967  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
968  *	the 'final' error, but it doesn't make sense to solve the printing
969  *	delay with more complex code.
970  */
971 
972 int
973 badport_bandlim(int which)
974 {
975 
976 #define	N(a)	(sizeof (a) / sizeof (a[0]))
977 	static struct rate {
978 		const char	*type;
979 		struct timeval	lasttime;
980 		int		curpps;
981 	} rates[BANDLIM_MAX+1] = {
982 		{ "icmp unreach response" },
983 		{ "icmp ping response" },
984 		{ "icmp tstamp response" },
985 		{ "closed port RST response" },
986 		{ "open port RST response" },
987 		{ "icmp6 unreach response" },
988 		{ "sctp ootb response" }
989 	};
990 
991 	/*
992 	 * Return ok status if feature disabled or argument out of range.
993 	 */
994 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
995 		struct rate *r = &rates[which];
996 		int opps = r->curpps;
997 
998 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
999 			return -1;	/* discard packet */
1000 		/*
1001 		 * If we've dropped below the threshold after having
1002 		 * rate-limited traffic print the message.  This preserves
1003 		 * the previous behaviour at the expense of added complexity.
1004 		 */
1005 		if (V_icmplim_output && opps > V_icmplim)
1006 			log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
1007 				r->type, opps, V_icmplim);
1008 	}
1009 	return 0;			/* okay to send packet */
1010 #undef N
1011 }
1012