xref: /freebsd/sys/netinet/ip_icmp.c (revision 2f02600abfddfc4e9f20dd384a2e729b451e16bd)
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 	icmp_reflect(m);
352 
353 freeit:
354 	m_freem(n);
355 }
356 
357 /*
358  * Process a received ICMP message.
359  */
360 void
361 icmp_input(struct mbuf *m, int off)
362 {
363 	struct icmp *icp;
364 	struct in_ifaddr *ia;
365 	struct ip *ip = mtod(m, struct ip *);
366 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
367 	int hlen = off;
368 	int icmplen = ntohs(ip->ip_len) - off;
369 	int i, code;
370 	void (*ctlfunc)(int, struct sockaddr *, void *);
371 	int fibnum;
372 
373 	/*
374 	 * Locate icmp structure in mbuf, and check
375 	 * that not corrupted and of at least minimum length.
376 	 */
377 #ifdef ICMPPRINTFS
378 	if (icmpprintfs) {
379 		char buf[4 * sizeof "123"];
380 		strcpy(buf, inet_ntoa(ip->ip_src));
381 		printf("icmp_input from %s to %s, len %d\n",
382 		       buf, inet_ntoa(ip->ip_dst), icmplen);
383 	}
384 #endif
385 	if (icmplen < ICMP_MINLEN) {
386 		ICMPSTAT_INC(icps_tooshort);
387 		goto freeit;
388 	}
389 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
390 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
391 		ICMPSTAT_INC(icps_tooshort);
392 		return;
393 	}
394 	ip = mtod(m, struct ip *);
395 	m->m_len -= hlen;
396 	m->m_data += hlen;
397 	icp = mtod(m, struct icmp *);
398 	if (in_cksum(m, icmplen)) {
399 		ICMPSTAT_INC(icps_checksum);
400 		goto freeit;
401 	}
402 	m->m_len += hlen;
403 	m->m_data -= hlen;
404 
405 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
406 		/*
407 		 * Deliver very specific ICMP type only.
408 		 */
409 		switch (icp->icmp_type) {
410 		case ICMP_UNREACH:
411 		case ICMP_TIMXCEED:
412 			break;
413 		default:
414 			goto freeit;
415 		}
416 	}
417 
418 #ifdef ICMPPRINTFS
419 	if (icmpprintfs)
420 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
421 		    icp->icmp_code);
422 #endif
423 
424 	/*
425 	 * Message type specific processing.
426 	 */
427 	if (icp->icmp_type > ICMP_MAXTYPE)
428 		goto raw;
429 
430 	/* Initialize */
431 	bzero(&icmpsrc, sizeof(icmpsrc));
432 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
433 	icmpsrc.sin_family = AF_INET;
434 	bzero(&icmpdst, sizeof(icmpdst));
435 	icmpdst.sin_len = sizeof(struct sockaddr_in);
436 	icmpdst.sin_family = AF_INET;
437 	bzero(&icmpgw, sizeof(icmpgw));
438 	icmpgw.sin_len = sizeof(struct sockaddr_in);
439 	icmpgw.sin_family = AF_INET;
440 
441 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
442 	code = icp->icmp_code;
443 	switch (icp->icmp_type) {
444 
445 	case ICMP_UNREACH:
446 		switch (code) {
447 			case ICMP_UNREACH_NET:
448 			case ICMP_UNREACH_HOST:
449 			case ICMP_UNREACH_SRCFAIL:
450 			case ICMP_UNREACH_NET_UNKNOWN:
451 			case ICMP_UNREACH_HOST_UNKNOWN:
452 			case ICMP_UNREACH_ISOLATED:
453 			case ICMP_UNREACH_TOSNET:
454 			case ICMP_UNREACH_TOSHOST:
455 			case ICMP_UNREACH_HOST_PRECEDENCE:
456 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
457 				code = PRC_UNREACH_NET;
458 				break;
459 
460 			case ICMP_UNREACH_NEEDFRAG:
461 				code = PRC_MSGSIZE;
462 				break;
463 
464 			/*
465 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
466 			 * Treat subcodes 2,3 as immediate RST
467 			 */
468 			case ICMP_UNREACH_PROTOCOL:
469 			case ICMP_UNREACH_PORT:
470 				code = PRC_UNREACH_PORT;
471 				break;
472 
473 			case ICMP_UNREACH_NET_PROHIB:
474 			case ICMP_UNREACH_HOST_PROHIB:
475 			case ICMP_UNREACH_FILTER_PROHIB:
476 				code = PRC_UNREACH_ADMIN_PROHIB;
477 				break;
478 
479 			default:
480 				goto badcode;
481 		}
482 		goto deliver;
483 
484 	case ICMP_TIMXCEED:
485 		if (code > 1)
486 			goto badcode;
487 		code += PRC_TIMXCEED_INTRANS;
488 		goto deliver;
489 
490 	case ICMP_PARAMPROB:
491 		if (code > 1)
492 			goto badcode;
493 		code = PRC_PARAMPROB;
494 		goto deliver;
495 
496 	case ICMP_SOURCEQUENCH:
497 		if (code)
498 			goto badcode;
499 		code = PRC_QUENCH;
500 	deliver:
501 		/*
502 		 * Problem with datagram; advise higher level routines.
503 		 */
504 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
505 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
506 			ICMPSTAT_INC(icps_badlen);
507 			goto freeit;
508 		}
509 		/* Discard ICMP's in response to multicast packets */
510 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
511 			goto badcode;
512 #ifdef ICMPPRINTFS
513 		if (icmpprintfs)
514 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
515 #endif
516 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
517 		/*
518 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
519 		 * notification to TCP layer.
520 		 */
521 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
522 		if (ctlfunc)
523 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
524 				   (void *)&icp->icmp_ip);
525 		break;
526 
527 	badcode:
528 		ICMPSTAT_INC(icps_badcode);
529 		break;
530 
531 	case ICMP_ECHO:
532 		if (!V_icmpbmcastecho
533 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
534 			ICMPSTAT_INC(icps_bmcastecho);
535 			break;
536 		}
537 		icp->icmp_type = ICMP_ECHOREPLY;
538 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
539 			goto freeit;
540 		else
541 			goto reflect;
542 
543 	case ICMP_TSTAMP:
544 		if (!V_icmpbmcastecho
545 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
546 			ICMPSTAT_INC(icps_bmcasttstamp);
547 			break;
548 		}
549 		if (icmplen < ICMP_TSLEN) {
550 			ICMPSTAT_INC(icps_badlen);
551 			break;
552 		}
553 		icp->icmp_type = ICMP_TSTAMPREPLY;
554 		icp->icmp_rtime = iptime();
555 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
556 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
557 			goto freeit;
558 		else
559 			goto reflect;
560 
561 	case ICMP_MASKREQ:
562 		if (V_icmpmaskrepl == 0)
563 			break;
564 		/*
565 		 * We are not able to respond with all ones broadcast
566 		 * unless we receive it over a point-to-point interface.
567 		 */
568 		if (icmplen < ICMP_MASKLEN)
569 			break;
570 		switch (ip->ip_dst.s_addr) {
571 
572 		case INADDR_BROADCAST:
573 		case INADDR_ANY:
574 			icmpdst.sin_addr = ip->ip_src;
575 			break;
576 
577 		default:
578 			icmpdst.sin_addr = ip->ip_dst;
579 		}
580 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
581 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
582 		if (ia == NULL)
583 			break;
584 		if (ia->ia_ifp == NULL) {
585 			ifa_free(&ia->ia_ifa);
586 			break;
587 		}
588 		icp->icmp_type = ICMP_MASKREPLY;
589 		if (V_icmpmaskfake == 0)
590 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
591 		else
592 			icp->icmp_mask = V_icmpmaskfake;
593 		if (ip->ip_src.s_addr == 0) {
594 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
595 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
596 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
597 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
598 		}
599 		ifa_free(&ia->ia_ifa);
600 reflect:
601 		ICMPSTAT_INC(icps_reflect);
602 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
603 		icmp_reflect(m);
604 		return;
605 
606 	case ICMP_REDIRECT:
607 		if (V_log_redirect) {
608 			u_long src, dst, gw;
609 
610 			src = ntohl(ip->ip_src.s_addr);
611 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
612 			gw = ntohl(icp->icmp_gwaddr.s_addr);
613 			printf("icmp redirect from %d.%d.%d.%d: "
614 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
615 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
616 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
617 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
618 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
619 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
620 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
621 		}
622 		/*
623 		 * RFC1812 says we must ignore ICMP redirects if we
624 		 * are acting as router.
625 		 */
626 		if (V_drop_redirect || V_ipforwarding)
627 			break;
628 		if (code > 3)
629 			goto badcode;
630 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
631 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
632 			ICMPSTAT_INC(icps_badlen);
633 			break;
634 		}
635 		/*
636 		 * Short circuit routing redirects to force
637 		 * immediate change in the kernel's routing
638 		 * tables.  The message is also handed to anyone
639 		 * listening on a raw socket (e.g. the routing
640 		 * daemon for use in updating its tables).
641 		 */
642 		icmpgw.sin_addr = ip->ip_src;
643 		icmpdst.sin_addr = icp->icmp_gwaddr;
644 #ifdef	ICMPPRINTFS
645 		if (icmpprintfs) {
646 			char buf[4 * sizeof "123"];
647 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
648 
649 			printf("redirect dst %s to %s\n",
650 			       buf, inet_ntoa(icp->icmp_gwaddr));
651 		}
652 #endif
653 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
654 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
655 			in_rtredirect((struct sockaddr *)&icmpsrc,
656 			  (struct sockaddr *)&icmpdst,
657 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
658 			  (struct sockaddr *)&icmpgw, fibnum);
659 		}
660 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
661 #ifdef IPSEC
662 		key_sa_routechange((struct sockaddr *)&icmpsrc);
663 #endif
664 		break;
665 
666 	/*
667 	 * No kernel processing for the following;
668 	 * just fall through to send to raw listener.
669 	 */
670 	case ICMP_ECHOREPLY:
671 	case ICMP_ROUTERADVERT:
672 	case ICMP_ROUTERSOLICIT:
673 	case ICMP_TSTAMPREPLY:
674 	case ICMP_IREQREPLY:
675 	case ICMP_MASKREPLY:
676 	default:
677 		break;
678 	}
679 
680 raw:
681 	rip_input(m, off);
682 	return;
683 
684 freeit:
685 	m_freem(m);
686 }
687 
688 /*
689  * Reflect the ip packet back to the source
690  */
691 static void
692 icmp_reflect(struct mbuf *m)
693 {
694 	struct ip *ip = mtod(m, struct ip *);
695 	struct ifaddr *ifa;
696 	struct ifnet *ifp;
697 	struct in_ifaddr *ia;
698 	struct in_addr t;
699 	struct mbuf *opts = 0;
700 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
701 
702 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
703 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
704 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
705 		m_freem(m);	/* Bad return address */
706 		ICMPSTAT_INC(icps_badaddr);
707 		goto done;	/* Ip_output() will check for broadcast */
708 	}
709 
710 	t = ip->ip_dst;
711 	ip->ip_dst = ip->ip_src;
712 
713 	/*
714 	 * Source selection for ICMP replies:
715 	 *
716 	 * If the incoming packet was addressed directly to one of our
717 	 * own addresses, use dst as the src for the reply.
718 	 */
719 	IN_IFADDR_RLOCK();
720 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
721 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
722 			t = IA_SIN(ia)->sin_addr;
723 			IN_IFADDR_RUNLOCK();
724 			goto match;
725 		}
726 	}
727 	IN_IFADDR_RUNLOCK();
728 
729 	/*
730 	 * If the incoming packet was addressed to one of our broadcast
731 	 * addresses, use the first non-broadcast address which corresponds
732 	 * to the incoming interface.
733 	 */
734 	ifp = m->m_pkthdr.rcvif;
735 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
736 		IF_ADDR_RLOCK(ifp);
737 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
738 			if (ifa->ifa_addr->sa_family != AF_INET)
739 				continue;
740 			ia = ifatoia(ifa);
741 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
742 			    t.s_addr) {
743 				t = IA_SIN(ia)->sin_addr;
744 				IF_ADDR_RUNLOCK(ifp);
745 				goto match;
746 			}
747 		}
748 		IF_ADDR_RUNLOCK(ifp);
749 	}
750 	/*
751 	 * If the packet was transiting through us, use the address of
752 	 * the interface the packet came through in.  If that interface
753 	 * doesn't have a suitable IP address, the normal selection
754 	 * criteria apply.
755 	 */
756 	if (V_icmp_rfi && ifp != NULL) {
757 		IF_ADDR_RLOCK(ifp);
758 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
759 			if (ifa->ifa_addr->sa_family != AF_INET)
760 				continue;
761 			ia = ifatoia(ifa);
762 			t = IA_SIN(ia)->sin_addr;
763 			IF_ADDR_RUNLOCK(ifp);
764 			goto match;
765 		}
766 		IF_ADDR_RUNLOCK(ifp);
767 	}
768 	/*
769 	 * If the incoming packet was not addressed directly to us, use
770 	 * designated interface for icmp replies specified by sysctl
771 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
772 	 * with normal source selection.
773 	 */
774 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
775 		IF_ADDR_RLOCK(ifp);
776 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
777 			if (ifa->ifa_addr->sa_family != AF_INET)
778 				continue;
779 			ia = ifatoia(ifa);
780 			t = IA_SIN(ia)->sin_addr;
781 			IF_ADDR_RUNLOCK(ifp);
782 			goto match;
783 		}
784 		IF_ADDR_RUNLOCK(ifp);
785 	}
786 	/*
787 	 * If the packet was transiting through us, use the address of
788 	 * the interface that is the closest to the packet source.
789 	 * When we don't have a route back to the packet source, stop here
790 	 * and drop the packet.
791 	 */
792 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
793 	if (ia == NULL) {
794 		m_freem(m);
795 		ICMPSTAT_INC(icps_noroute);
796 		goto done;
797 	}
798 	t = IA_SIN(ia)->sin_addr;
799 	ifa_free(&ia->ia_ifa);
800 match:
801 #ifdef MAC
802 	mac_netinet_icmp_replyinplace(m);
803 #endif
804 	ip->ip_src = t;
805 	ip->ip_ttl = V_ip_defttl;
806 
807 	if (optlen > 0) {
808 		register u_char *cp;
809 		int opt, cnt;
810 		u_int len;
811 
812 		/*
813 		 * Retrieve any source routing from the incoming packet;
814 		 * add on any record-route or timestamp options.
815 		 */
816 		cp = (u_char *) (ip + 1);
817 		if ((opts = ip_srcroute(m)) == 0 &&
818 		    (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
819 			opts->m_len = sizeof(struct in_addr);
820 			mtod(opts, struct in_addr *)->s_addr = 0;
821 		}
822 		if (opts) {
823 #ifdef ICMPPRINTFS
824 		    if (icmpprintfs)
825 			    printf("icmp_reflect optlen %d rt %d => ",
826 				optlen, opts->m_len);
827 #endif
828 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
829 			    opt = cp[IPOPT_OPTVAL];
830 			    if (opt == IPOPT_EOL)
831 				    break;
832 			    if (opt == IPOPT_NOP)
833 				    len = 1;
834 			    else {
835 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
836 					    break;
837 				    len = cp[IPOPT_OLEN];
838 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
839 				        len > cnt)
840 					    break;
841 			    }
842 			    /*
843 			     * Should check for overflow, but it "can't happen"
844 			     */
845 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
846 				opt == IPOPT_SECURITY) {
847 				    bcopy((caddr_t)cp,
848 					mtod(opts, caddr_t) + opts->m_len, len);
849 				    opts->m_len += len;
850 			    }
851 		    }
852 		    /* Terminate & pad, if necessary */
853 		    cnt = opts->m_len % 4;
854 		    if (cnt) {
855 			    for (; cnt < 4; cnt++) {
856 				    *(mtod(opts, caddr_t) + opts->m_len) =
857 					IPOPT_EOL;
858 				    opts->m_len++;
859 			    }
860 		    }
861 #ifdef ICMPPRINTFS
862 		    if (icmpprintfs)
863 			    printf("%d\n", opts->m_len);
864 #endif
865 		}
866 		ip_stripoptions(m);
867 	}
868 	m_tag_delete_nonpersistent(m);
869 	m->m_flags &= ~(M_BCAST|M_MCAST);
870 	icmp_send(m, opts);
871 done:
872 	if (opts)
873 		(void)m_free(opts);
874 }
875 
876 /*
877  * Send an icmp packet back to the ip level,
878  * after supplying a checksum.
879  */
880 static void
881 icmp_send(struct mbuf *m, struct mbuf *opts)
882 {
883 	register struct ip *ip = mtod(m, struct ip *);
884 	register int hlen;
885 	register struct icmp *icp;
886 
887 	hlen = ip->ip_hl << 2;
888 	m->m_data += hlen;
889 	m->m_len -= hlen;
890 	icp = mtod(m, struct icmp *);
891 	icp->icmp_cksum = 0;
892 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
893 	m->m_data -= hlen;
894 	m->m_len += hlen;
895 	m->m_pkthdr.rcvif = (struct ifnet *)0;
896 #ifdef ICMPPRINTFS
897 	if (icmpprintfs) {
898 		char buf[4 * sizeof "123"];
899 		strcpy(buf, inet_ntoa(ip->ip_dst));
900 		printf("icmp_send dst %s src %s\n",
901 		       buf, inet_ntoa(ip->ip_src));
902 	}
903 #endif
904 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
905 }
906 
907 /*
908  * Return milliseconds since 00:00 GMT in network format.
909  */
910 uint32_t
911 iptime(void)
912 {
913 	struct timeval atv;
914 	u_long t;
915 
916 	getmicrotime(&atv);
917 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
918 	return (htonl(t));
919 }
920 
921 /*
922  * Return the next larger or smaller MTU plateau (table from RFC 1191)
923  * given current value MTU.  If DIR is less than zero, a larger plateau
924  * is returned; otherwise, a smaller value is returned.
925  */
926 int
927 ip_next_mtu(int mtu, int dir)
928 {
929 	static int mtutab[] = {
930 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
931 		296, 68, 0
932 	};
933 	int i, size;
934 
935 	size = (sizeof mtutab) / (sizeof mtutab[0]);
936 	if (dir >= 0) {
937 		for (i = 0; i < size; i++)
938 			if (mtu > mtutab[i])
939 				return mtutab[i];
940 	} else {
941 		for (i = size - 1; i >= 0; i--)
942 			if (mtu < mtutab[i])
943 				return mtutab[i];
944 		if (mtu == mtutab[0])
945 			return mtutab[0];
946 	}
947 	return 0;
948 }
949 #endif /* INET */
950 
951 
952 /*
953  * badport_bandlim() - check for ICMP bandwidth limit
954  *
955  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
956  *	hit our bandwidth limit and it is not ok.
957  *
958  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
959  *
960  *	For now we separate the TCP and UDP subsystems w/ different 'which'
961  *	values.  We may eventually remove this separation (and simplify the
962  *	code further).
963  *
964  *	Note that the printing of the error message is delayed so we can
965  *	properly print the icmp error rate that the system was trying to do
966  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
967  *	the 'final' error, but it doesn't make sense to solve the printing
968  *	delay with more complex code.
969  */
970 
971 int
972 badport_bandlim(int which)
973 {
974 
975 #define	N(a)	(sizeof (a) / sizeof (a[0]))
976 	static struct rate {
977 		const char	*type;
978 		struct timeval	lasttime;
979 		int		curpps;
980 	} rates[BANDLIM_MAX+1] = {
981 		{ "icmp unreach response" },
982 		{ "icmp ping response" },
983 		{ "icmp tstamp response" },
984 		{ "closed port RST response" },
985 		{ "open port RST response" },
986 		{ "icmp6 unreach response" },
987 		{ "sctp ootb response" }
988 	};
989 
990 	/*
991 	 * Return ok status if feature disabled or argument out of range.
992 	 */
993 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
994 		struct rate *r = &rates[which];
995 		int opps = r->curpps;
996 
997 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
998 			return -1;	/* discard packet */
999 		/*
1000 		 * If we've dropped below the threshold after having
1001 		 * rate-limited traffic print the message.  This preserves
1002 		 * the previous behaviour at the expense of added complexity.
1003 		 */
1004 		if (V_icmplim_output && opps > V_icmplim)
1005 			log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
1006 				r->type, opps, V_icmplim);
1007 	}
1008 	return 0;			/* okay to send packet */
1009 #undef N
1010 }
1011