xref: /freebsd/sys/netinet/ip_icmp.c (revision 49dae58b287906be26f56ba3e3dc693c3ba8cf37)
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 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mbuf.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/lock.h>
45 #include <sys/rmlock.h>
46 #include <sys/sysctl.h>
47 #include <sys/syslog.h>
48 
49 #include <net/if.h>
50 #include <net/if_var.h>
51 #include <net/if_types.h>
52 #include <net/route.h>
53 #include <net/vnet.h>
54 
55 #include <netinet/in.h>
56 #include <netinet/in_fib.h>
57 #include <netinet/in_pcb.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip.h>
61 #include <netinet/ip_icmp.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/ip_options.h>
64 #include <netinet/tcp.h>
65 #include <netinet/tcp_var.h>
66 #include <netinet/tcpip.h>
67 #include <netinet/icmp_var.h>
68 
69 #ifdef INET
70 
71 #include <machine/in_cksum.h>
72 
73 #include <security/mac/mac_framework.h>
74 #endif /* INET */
75 
76 /*
77  * ICMP routines: error generation, receive packet processing, and
78  * routines to turnaround packets back to the originator, and
79  * host table maintenance routines.
80  */
81 static VNET_DEFINE(int, icmplim) = 200;
82 #define	V_icmplim			VNET(icmplim)
83 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
84 	&VNET_NAME(icmplim), 0,
85 	"Maximum number of ICMP responses per second");
86 
87 static VNET_DEFINE(int, icmplim_output) = 1;
88 #define	V_icmplim_output		VNET(icmplim_output)
89 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
90 	&VNET_NAME(icmplim_output), 0,
91 	"Enable logging of ICMP response rate limiting");
92 
93 #ifdef INET
94 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
95 VNET_PCPUSTAT_SYSINIT(icmpstat);
96 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
97     icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
98 
99 #ifdef VIMAGE
100 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
101 #endif /* VIMAGE */
102 
103 static VNET_DEFINE(int, icmpmaskrepl) = 0;
104 #define	V_icmpmaskrepl			VNET(icmpmaskrepl)
105 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
106 	&VNET_NAME(icmpmaskrepl), 0,
107 	"Reply to ICMP Address Mask Request packets.");
108 
109 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
110 #define	V_icmpmaskfake			VNET(icmpmaskfake)
111 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
112 	&VNET_NAME(icmpmaskfake), 0,
113 	"Fake reply to ICMP Address Mask Request packets.");
114 
115 VNET_DEFINE(int, drop_redirect) = 0;
116 #define	V_drop_redirect			VNET(drop_redirect)
117 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
118 	&VNET_NAME(drop_redirect), 0, "Ignore ICMP redirects");
119 
120 static VNET_DEFINE(int, log_redirect) = 0;
121 #define	V_log_redirect			VNET(log_redirect)
122 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
123 	&VNET_NAME(log_redirect), 0,
124 	"Log ICMP redirects to the console");
125 
126 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
127 #define	V_reply_src			VNET(reply_src)
128 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
129 	&VNET_NAME(reply_src), IFNAMSIZ,
130 	"icmp reply source for non-local packets.");
131 
132 static VNET_DEFINE(int, icmp_rfi) = 0;
133 #define	V_icmp_rfi			VNET(icmp_rfi)
134 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
135 	&VNET_NAME(icmp_rfi), 0,
136 	"ICMP reply from incoming interface for non-local packets");
137 
138 static VNET_DEFINE(int, icmp_quotelen) = 8;
139 #define	V_icmp_quotelen			VNET(icmp_quotelen)
140 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
141 	&VNET_NAME(icmp_quotelen), 0,
142 	"Number of bytes from original packet to quote in ICMP reply");
143 
144 /*
145  * ICMP broadcast echo sysctl
146  */
147 static VNET_DEFINE(int, icmpbmcastecho) = 0;
148 #define	V_icmpbmcastecho		VNET(icmpbmcastecho)
149 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
150 	&VNET_NAME(icmpbmcastecho), 0,
151 	"");
152 
153 static VNET_DEFINE(int, icmptstamprepl) = 1;
154 #define	V_icmptstamprepl		VNET(icmptstamprepl)
155 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_RW,
156 	&VNET_NAME(icmptstamprepl), 0, "Respond to ICMP Timestamp packets");
157 
158 #ifdef ICMPPRINTFS
159 int	icmpprintfs = 0;
160 #endif
161 
162 static void	icmp_reflect(struct mbuf *);
163 static void	icmp_send(struct mbuf *, struct mbuf *);
164 
165 extern	struct protosw inetsw[];
166 
167 /*
168  * Kernel module interface for updating icmpstat.  The argument is an index
169  * into icmpstat treated as an array of u_long.  While this encodes the
170  * general layout of icmpstat into the caller, it doesn't encode its
171  * location, so that future changes to add, for example, per-CPU stats
172  * support won't cause binary compatibility problems for kernel modules.
173  */
174 void
175 kmod_icmpstat_inc(int statnum)
176 {
177 
178 	counter_u64_add(VNET(icmpstat)[statnum], 1);
179 }
180 
181 /*
182  * Generate an error packet of type error
183  * in response to bad packet ip.
184  */
185 void
186 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
187 {
188 	register struct ip *oip = mtod(n, struct ip *), *nip;
189 	register unsigned oiphlen = oip->ip_hl << 2;
190 	register struct icmp *icp;
191 	register struct mbuf *m;
192 	unsigned icmplen, icmpelen, nlen;
193 
194 	KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
195 #ifdef ICMPPRINTFS
196 	if (icmpprintfs)
197 		printf("icmp_error(%p, %x, %d)\n", oip, type, code);
198 #endif
199 	if (type != ICMP_REDIRECT)
200 		ICMPSTAT_INC(icps_error);
201 	/*
202 	 * Don't send error:
203 	 *  if the original packet was encrypted.
204 	 *  if not the first fragment of message.
205 	 *  in response to a multicast or broadcast packet.
206 	 *  if the old packet protocol was an ICMP error message.
207 	 */
208 	if (n->m_flags & M_DECRYPTED)
209 		goto freeit;
210 	if (oip->ip_off & htons(~(IP_MF|IP_DF)))
211 		goto freeit;
212 	if (n->m_flags & (M_BCAST|M_MCAST))
213 		goto freeit;
214 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
215 	  n->m_len >= oiphlen + ICMP_MINLEN &&
216 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
217 		ICMPSTAT_INC(icps_oldicmp);
218 		goto freeit;
219 	}
220 	/* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
221 	if (oiphlen + 8 > n->m_len)
222 		goto freeit;
223 	/*
224 	 * Calculate length to quote from original packet and
225 	 * prevent the ICMP mbuf from overflowing.
226 	 * Unfortunatly this is non-trivial since ip_forward()
227 	 * sends us truncated packets.
228 	 */
229 	nlen = m_length(n, NULL);
230 	if (oip->ip_p == IPPROTO_TCP) {
231 		struct tcphdr *th;
232 		int tcphlen;
233 
234 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
235 		    n->m_next == NULL)
236 			goto stdreply;
237 		if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
238 		    ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
239 			goto freeit;
240 		th = (struct tcphdr *)((caddr_t)oip + oiphlen);
241 		tcphlen = th->th_off << 2;
242 		if (tcphlen < sizeof(struct tcphdr))
243 			goto freeit;
244 		if (ntohs(oip->ip_len) < oiphlen + tcphlen)
245 			goto freeit;
246 		if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
247 			goto stdreply;
248 		if (n->m_len < oiphlen + tcphlen &&
249 		    ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
250 			goto freeit;
251 		icmpelen = max(tcphlen, min(V_icmp_quotelen,
252 		    ntohs(oip->ip_len) - oiphlen));
253 	} else
254 stdreply:	icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
255 
256 	icmplen = min(oiphlen + icmpelen, nlen);
257 	if (icmplen < sizeof(struct ip))
258 		goto freeit;
259 
260 	if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
261 		m = m_gethdr(M_NOWAIT, MT_DATA);
262 	else
263 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
264 	if (m == NULL)
265 		goto freeit;
266 #ifdef MAC
267 	mac_netinet_icmp_reply(n, m);
268 #endif
269 	icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
270 	m_align(m, ICMP_MINLEN + icmplen);
271 	m->m_len = ICMP_MINLEN + icmplen;
272 
273 	/* XXX MRT  make the outgoing packet use the same FIB
274 	 * that was associated with the incoming packet
275 	 */
276 	M_SETFIB(m, M_GETFIB(n));
277 	icp = mtod(m, struct icmp *);
278 	ICMPSTAT_INC(icps_outhist[type]);
279 	icp->icmp_type = type;
280 	if (type == ICMP_REDIRECT)
281 		icp->icmp_gwaddr.s_addr = dest;
282 	else {
283 		icp->icmp_void = 0;
284 		/*
285 		 * The following assignments assume an overlay with the
286 		 * just zeroed icmp_void field.
287 		 */
288 		if (type == ICMP_PARAMPROB) {
289 			icp->icmp_pptr = code;
290 			code = 0;
291 		} else if (type == ICMP_UNREACH &&
292 			code == ICMP_UNREACH_NEEDFRAG && mtu) {
293 			icp->icmp_nextmtu = htons(mtu);
294 		}
295 	}
296 	icp->icmp_code = code;
297 
298 	/*
299 	 * Copy the quotation into ICMP message and
300 	 * convert quoted IP header back to network representation.
301 	 */
302 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
303 	nip = &icp->icmp_ip;
304 
305 	/*
306 	 * Set up ICMP message mbuf and copy old IP header (without options
307 	 * in front of ICMP message.
308 	 * If the original mbuf was meant to bypass the firewall, the error
309 	 * reply should bypass as well.
310 	 */
311 	m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
312 	m->m_data -= sizeof(struct ip);
313 	m->m_len += sizeof(struct ip);
314 	m->m_pkthdr.len = m->m_len;
315 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
316 	nip = mtod(m, struct ip *);
317 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
318 	nip->ip_len = htons(m->m_len);
319 	nip->ip_v = IPVERSION;
320 	nip->ip_hl = 5;
321 	nip->ip_p = IPPROTO_ICMP;
322 	nip->ip_tos = 0;
323 	nip->ip_off = 0;
324 	icmp_reflect(m);
325 
326 freeit:
327 	m_freem(n);
328 }
329 
330 /*
331  * Process a received ICMP message.
332  */
333 int
334 icmp_input(struct mbuf **mp, int *offp, int proto)
335 {
336 	struct icmp *icp;
337 	struct in_ifaddr *ia;
338 	struct mbuf *m = *mp;
339 	struct ip *ip = mtod(m, struct ip *);
340 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
341 	int hlen = *offp;
342 	int icmplen = ntohs(ip->ip_len) - *offp;
343 	int i, code;
344 	void (*ctlfunc)(int, struct sockaddr *, void *);
345 	int fibnum;
346 
347 	*mp = NULL;
348 
349 	/*
350 	 * Locate icmp structure in mbuf, and check
351 	 * that not corrupted and of at least minimum length.
352 	 */
353 #ifdef ICMPPRINTFS
354 	if (icmpprintfs) {
355 		char buf[4 * sizeof "123"];
356 		strcpy(buf, inet_ntoa(ip->ip_src));
357 		printf("icmp_input from %s to %s, len %d\n",
358 		       buf, inet_ntoa(ip->ip_dst), icmplen);
359 	}
360 #endif
361 	if (icmplen < ICMP_MINLEN) {
362 		ICMPSTAT_INC(icps_tooshort);
363 		goto freeit;
364 	}
365 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
366 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
367 		ICMPSTAT_INC(icps_tooshort);
368 		return (IPPROTO_DONE);
369 	}
370 	ip = mtod(m, struct ip *);
371 	m->m_len -= hlen;
372 	m->m_data += hlen;
373 	icp = mtod(m, struct icmp *);
374 	if (in_cksum(m, icmplen)) {
375 		ICMPSTAT_INC(icps_checksum);
376 		goto freeit;
377 	}
378 	m->m_len += hlen;
379 	m->m_data -= hlen;
380 
381 #ifdef ICMPPRINTFS
382 	if (icmpprintfs)
383 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
384 		    icp->icmp_code);
385 #endif
386 
387 	/*
388 	 * Message type specific processing.
389 	 */
390 	if (icp->icmp_type > ICMP_MAXTYPE)
391 		goto raw;
392 
393 	/* Initialize */
394 	bzero(&icmpsrc, sizeof(icmpsrc));
395 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
396 	icmpsrc.sin_family = AF_INET;
397 	bzero(&icmpdst, sizeof(icmpdst));
398 	icmpdst.sin_len = sizeof(struct sockaddr_in);
399 	icmpdst.sin_family = AF_INET;
400 	bzero(&icmpgw, sizeof(icmpgw));
401 	icmpgw.sin_len = sizeof(struct sockaddr_in);
402 	icmpgw.sin_family = AF_INET;
403 
404 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
405 	code = icp->icmp_code;
406 	switch (icp->icmp_type) {
407 
408 	case ICMP_UNREACH:
409 		switch (code) {
410 			case ICMP_UNREACH_NET:
411 			case ICMP_UNREACH_HOST:
412 			case ICMP_UNREACH_SRCFAIL:
413 			case ICMP_UNREACH_NET_UNKNOWN:
414 			case ICMP_UNREACH_HOST_UNKNOWN:
415 			case ICMP_UNREACH_ISOLATED:
416 			case ICMP_UNREACH_TOSNET:
417 			case ICMP_UNREACH_TOSHOST:
418 			case ICMP_UNREACH_HOST_PRECEDENCE:
419 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
420 				code = PRC_UNREACH_NET;
421 				break;
422 
423 			case ICMP_UNREACH_NEEDFRAG:
424 				code = PRC_MSGSIZE;
425 				break;
426 
427 			/*
428 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
429 			 * Treat subcodes 2,3 as immediate RST
430 			 */
431 			case ICMP_UNREACH_PROTOCOL:
432 			case ICMP_UNREACH_PORT:
433 				code = PRC_UNREACH_PORT;
434 				break;
435 
436 			case ICMP_UNREACH_NET_PROHIB:
437 			case ICMP_UNREACH_HOST_PROHIB:
438 			case ICMP_UNREACH_FILTER_PROHIB:
439 				code = PRC_UNREACH_ADMIN_PROHIB;
440 				break;
441 
442 			default:
443 				goto badcode;
444 		}
445 		goto deliver;
446 
447 	case ICMP_TIMXCEED:
448 		if (code > 1)
449 			goto badcode;
450 		code += PRC_TIMXCEED_INTRANS;
451 		goto deliver;
452 
453 	case ICMP_PARAMPROB:
454 		if (code > 1)
455 			goto badcode;
456 		code = PRC_PARAMPROB;
457 	deliver:
458 		/*
459 		 * Problem with datagram; advise higher level routines.
460 		 */
461 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
462 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
463 			ICMPSTAT_INC(icps_badlen);
464 			goto freeit;
465 		}
466 		/* Discard ICMP's in response to multicast packets */
467 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
468 			goto badcode;
469 #ifdef ICMPPRINTFS
470 		if (icmpprintfs)
471 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
472 #endif
473 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
474 		/*
475 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
476 		 * notification to TCP layer.
477 		 */
478 		i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
479 		ip_stripoptions(m);
480 		if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
481 			/* This should actually not happen */
482 			ICMPSTAT_INC(icps_tooshort);
483 			return (IPPROTO_DONE);
484 		}
485 		ip = mtod(m, struct ip *);
486 		icp = (struct icmp *)(ip + 1);
487 		/*
488 		 * The upper layer handler can rely on:
489 		 * - The outer IP header has no options.
490 		 * - The outer IP header, the ICMP header, the inner IP header,
491 		 *   and the first n bytes of the inner payload are contiguous.
492 		 *   n is at least 8, but might be larger based on
493 		 *   ICMP_ADVLENPREF. See its definition in ip_icmp.h.
494 		 */
495 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
496 		if (ctlfunc)
497 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
498 				   (void *)&icp->icmp_ip);
499 		break;
500 
501 	badcode:
502 		ICMPSTAT_INC(icps_badcode);
503 		break;
504 
505 	case ICMP_ECHO:
506 		if (!V_icmpbmcastecho
507 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
508 			ICMPSTAT_INC(icps_bmcastecho);
509 			break;
510 		}
511 		icp->icmp_type = ICMP_ECHOREPLY;
512 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
513 			goto freeit;
514 		else
515 			goto reflect;
516 
517 	case ICMP_TSTAMP:
518 		if (V_icmptstamprepl == 0)
519 			break;
520 		if (!V_icmpbmcastecho
521 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
522 			ICMPSTAT_INC(icps_bmcasttstamp);
523 			break;
524 		}
525 		if (icmplen < ICMP_TSLEN) {
526 			ICMPSTAT_INC(icps_badlen);
527 			break;
528 		}
529 		icp->icmp_type = ICMP_TSTAMPREPLY;
530 		icp->icmp_rtime = iptime();
531 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
532 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
533 			goto freeit;
534 		else
535 			goto reflect;
536 
537 	case ICMP_MASKREQ:
538 		if (V_icmpmaskrepl == 0)
539 			break;
540 		/*
541 		 * We are not able to respond with all ones broadcast
542 		 * unless we receive it over a point-to-point interface.
543 		 */
544 		if (icmplen < ICMP_MASKLEN)
545 			break;
546 		switch (ip->ip_dst.s_addr) {
547 
548 		case INADDR_BROADCAST:
549 		case INADDR_ANY:
550 			icmpdst.sin_addr = ip->ip_src;
551 			break;
552 
553 		default:
554 			icmpdst.sin_addr = ip->ip_dst;
555 		}
556 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
557 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
558 		if (ia == NULL)
559 			break;
560 		if (ia->ia_ifp == NULL) {
561 			ifa_free(&ia->ia_ifa);
562 			break;
563 		}
564 		icp->icmp_type = ICMP_MASKREPLY;
565 		if (V_icmpmaskfake == 0)
566 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
567 		else
568 			icp->icmp_mask = V_icmpmaskfake;
569 		if (ip->ip_src.s_addr == 0) {
570 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
571 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
572 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
573 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
574 		}
575 		ifa_free(&ia->ia_ifa);
576 reflect:
577 		ICMPSTAT_INC(icps_reflect);
578 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
579 		icmp_reflect(m);
580 		return (IPPROTO_DONE);
581 
582 	case ICMP_REDIRECT:
583 		if (V_log_redirect) {
584 			u_long src, dst, gw;
585 
586 			src = ntohl(ip->ip_src.s_addr);
587 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
588 			gw = ntohl(icp->icmp_gwaddr.s_addr);
589 			printf("icmp redirect from %d.%d.%d.%d: "
590 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
591 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
592 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
593 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
594 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
595 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
596 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
597 		}
598 		/*
599 		 * RFC1812 says we must ignore ICMP redirects if we
600 		 * are acting as router.
601 		 */
602 		if (V_drop_redirect || V_ipforwarding)
603 			break;
604 		if (code > 3)
605 			goto badcode;
606 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
607 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
608 			ICMPSTAT_INC(icps_badlen);
609 			break;
610 		}
611 		/*
612 		 * Short circuit routing redirects to force
613 		 * immediate change in the kernel's routing
614 		 * tables.  The message is also handed to anyone
615 		 * listening on a raw socket (e.g. the routing
616 		 * daemon for use in updating its tables).
617 		 */
618 		icmpgw.sin_addr = ip->ip_src;
619 		icmpdst.sin_addr = icp->icmp_gwaddr;
620 #ifdef	ICMPPRINTFS
621 		if (icmpprintfs) {
622 			char buf[4 * sizeof "123"];
623 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
624 
625 			printf("redirect dst %s to %s\n",
626 			       buf, inet_ntoa(icp->icmp_gwaddr));
627 		}
628 #endif
629 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
630 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
631 			in_rtredirect((struct sockaddr *)&icmpsrc,
632 			  (struct sockaddr *)&icmpdst,
633 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
634 			  (struct sockaddr *)&icmpgw, fibnum);
635 		}
636 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
637 		break;
638 
639 	/*
640 	 * No kernel processing for the following;
641 	 * just fall through to send to raw listener.
642 	 */
643 	case ICMP_ECHOREPLY:
644 	case ICMP_ROUTERADVERT:
645 	case ICMP_ROUTERSOLICIT:
646 	case ICMP_TSTAMPREPLY:
647 	case ICMP_IREQREPLY:
648 	case ICMP_MASKREPLY:
649 	case ICMP_SOURCEQUENCH:
650 	default:
651 		break;
652 	}
653 
654 raw:
655 	*mp = m;
656 	rip_input(mp, offp, proto);
657 	return (IPPROTO_DONE);
658 
659 freeit:
660 	m_freem(m);
661 	return (IPPROTO_DONE);
662 }
663 
664 /*
665  * Reflect the ip packet back to the source
666  */
667 static void
668 icmp_reflect(struct mbuf *m)
669 {
670 	struct rm_priotracker in_ifa_tracker;
671 	struct ip *ip = mtod(m, struct ip *);
672 	struct ifaddr *ifa;
673 	struct ifnet *ifp;
674 	struct in_ifaddr *ia;
675 	struct in_addr t;
676 	struct nhop4_extended nh_ext;
677 	struct mbuf *opts = NULL;
678 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
679 
680 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
681 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
682 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
683 		m_freem(m);	/* Bad return address */
684 		ICMPSTAT_INC(icps_badaddr);
685 		goto done;	/* Ip_output() will check for broadcast */
686 	}
687 
688 	t = ip->ip_dst;
689 	ip->ip_dst = ip->ip_src;
690 
691 	/*
692 	 * Source selection for ICMP replies:
693 	 *
694 	 * If the incoming packet was addressed directly to one of our
695 	 * own addresses, use dst as the src for the reply.
696 	 */
697 	IN_IFADDR_RLOCK(&in_ifa_tracker);
698 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
699 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
700 			t = IA_SIN(ia)->sin_addr;
701 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
702 			goto match;
703 		}
704 	}
705 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
706 
707 	/*
708 	 * If the incoming packet was addressed to one of our broadcast
709 	 * addresses, use the first non-broadcast address which corresponds
710 	 * to the incoming interface.
711 	 */
712 	ifp = m->m_pkthdr.rcvif;
713 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
714 		IF_ADDR_RLOCK(ifp);
715 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
716 			if (ifa->ifa_addr->sa_family != AF_INET)
717 				continue;
718 			ia = ifatoia(ifa);
719 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
720 			    t.s_addr) {
721 				t = IA_SIN(ia)->sin_addr;
722 				IF_ADDR_RUNLOCK(ifp);
723 				goto match;
724 			}
725 		}
726 		IF_ADDR_RUNLOCK(ifp);
727 	}
728 	/*
729 	 * If the packet was transiting through us, use the address of
730 	 * the interface the packet came through in.  If that interface
731 	 * doesn't have a suitable IP address, the normal selection
732 	 * criteria apply.
733 	 */
734 	if (V_icmp_rfi && ifp != NULL) {
735 		IF_ADDR_RLOCK(ifp);
736 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
737 			if (ifa->ifa_addr->sa_family != AF_INET)
738 				continue;
739 			ia = ifatoia(ifa);
740 			t = IA_SIN(ia)->sin_addr;
741 			IF_ADDR_RUNLOCK(ifp);
742 			goto match;
743 		}
744 		IF_ADDR_RUNLOCK(ifp);
745 	}
746 	/*
747 	 * If the incoming packet was not addressed directly to us, use
748 	 * designated interface for icmp replies specified by sysctl
749 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
750 	 * with normal source selection.
751 	 */
752 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
753 		IF_ADDR_RLOCK(ifp);
754 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
755 			if (ifa->ifa_addr->sa_family != AF_INET)
756 				continue;
757 			ia = ifatoia(ifa);
758 			t = IA_SIN(ia)->sin_addr;
759 			IF_ADDR_RUNLOCK(ifp);
760 			goto match;
761 		}
762 		IF_ADDR_RUNLOCK(ifp);
763 	}
764 	/*
765 	 * If the packet was transiting through us, use the address of
766 	 * the interface that is the closest to the packet source.
767 	 * When we don't have a route back to the packet source, stop here
768 	 * and drop the packet.
769 	 */
770 	if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) {
771 		m_freem(m);
772 		ICMPSTAT_INC(icps_noroute);
773 		goto done;
774 	}
775 	t = nh_ext.nh_src;
776 match:
777 #ifdef MAC
778 	mac_netinet_icmp_replyinplace(m);
779 #endif
780 	ip->ip_src = t;
781 	ip->ip_ttl = V_ip_defttl;
782 
783 	if (optlen > 0) {
784 		register u_char *cp;
785 		int opt, cnt;
786 		u_int len;
787 
788 		/*
789 		 * Retrieve any source routing from the incoming packet;
790 		 * add on any record-route or timestamp options.
791 		 */
792 		cp = (u_char *) (ip + 1);
793 		if ((opts = ip_srcroute(m)) == NULL &&
794 		    (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
795 			opts->m_len = sizeof(struct in_addr);
796 			mtod(opts, struct in_addr *)->s_addr = 0;
797 		}
798 		if (opts) {
799 #ifdef ICMPPRINTFS
800 		    if (icmpprintfs)
801 			    printf("icmp_reflect optlen %d rt %d => ",
802 				optlen, opts->m_len);
803 #endif
804 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
805 			    opt = cp[IPOPT_OPTVAL];
806 			    if (opt == IPOPT_EOL)
807 				    break;
808 			    if (opt == IPOPT_NOP)
809 				    len = 1;
810 			    else {
811 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
812 					    break;
813 				    len = cp[IPOPT_OLEN];
814 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
815 				        len > cnt)
816 					    break;
817 			    }
818 			    /*
819 			     * Should check for overflow, but it "can't happen"
820 			     */
821 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
822 				opt == IPOPT_SECURITY) {
823 				    bcopy((caddr_t)cp,
824 					mtod(opts, caddr_t) + opts->m_len, len);
825 				    opts->m_len += len;
826 			    }
827 		    }
828 		    /* Terminate & pad, if necessary */
829 		    cnt = opts->m_len % 4;
830 		    if (cnt) {
831 			    for (; cnt < 4; cnt++) {
832 				    *(mtod(opts, caddr_t) + opts->m_len) =
833 					IPOPT_EOL;
834 				    opts->m_len++;
835 			    }
836 		    }
837 #ifdef ICMPPRINTFS
838 		    if (icmpprintfs)
839 			    printf("%d\n", opts->m_len);
840 #endif
841 		}
842 		ip_stripoptions(m);
843 	}
844 	m_tag_delete_nonpersistent(m);
845 	m->m_flags &= ~(M_BCAST|M_MCAST);
846 	icmp_send(m, opts);
847 done:
848 	if (opts)
849 		(void)m_free(opts);
850 }
851 
852 /*
853  * Send an icmp packet back to the ip level,
854  * after supplying a checksum.
855  */
856 static void
857 icmp_send(struct mbuf *m, struct mbuf *opts)
858 {
859 	register struct ip *ip = mtod(m, struct ip *);
860 	register int hlen;
861 	register struct icmp *icp;
862 
863 	hlen = ip->ip_hl << 2;
864 	m->m_data += hlen;
865 	m->m_len -= hlen;
866 	icp = mtod(m, struct icmp *);
867 	icp->icmp_cksum = 0;
868 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
869 	m->m_data -= hlen;
870 	m->m_len += hlen;
871 	m->m_pkthdr.rcvif = (struct ifnet *)0;
872 #ifdef ICMPPRINTFS
873 	if (icmpprintfs) {
874 		char buf[4 * sizeof "123"];
875 		strcpy(buf, inet_ntoa(ip->ip_dst));
876 		printf("icmp_send dst %s src %s\n",
877 		       buf, inet_ntoa(ip->ip_src));
878 	}
879 #endif
880 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
881 }
882 
883 /*
884  * Return milliseconds since 00:00 UTC in network format.
885  */
886 uint32_t
887 iptime(void)
888 {
889 	struct timeval atv;
890 	u_long t;
891 
892 	getmicrotime(&atv);
893 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
894 	return (htonl(t));
895 }
896 
897 /*
898  * Return the next larger or smaller MTU plateau (table from RFC 1191)
899  * given current value MTU.  If DIR is less than zero, a larger plateau
900  * is returned; otherwise, a smaller value is returned.
901  */
902 int
903 ip_next_mtu(int mtu, int dir)
904 {
905 	static int mtutab[] = {
906 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
907 		296, 68, 0
908 	};
909 	int i, size;
910 
911 	size = (sizeof mtutab) / (sizeof mtutab[0]);
912 	if (dir >= 0) {
913 		for (i = 0; i < size; i++)
914 			if (mtu > mtutab[i])
915 				return mtutab[i];
916 	} else {
917 		for (i = size - 1; i >= 0; i--)
918 			if (mtu < mtutab[i])
919 				return mtutab[i];
920 		if (mtu == mtutab[0])
921 			return mtutab[0];
922 	}
923 	return 0;
924 }
925 #endif /* INET */
926 
927 
928 /*
929  * badport_bandlim() - check for ICMP bandwidth limit
930  *
931  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
932  *	hit our bandwidth limit and it is not ok.
933  *
934  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
935  *
936  *	For now we separate the TCP and UDP subsystems w/ different 'which'
937  *	values.  We may eventually remove this separation (and simplify the
938  *	code further).
939  *
940  *	Note that the printing of the error message is delayed so we can
941  *	properly print the icmp error rate that the system was trying to do
942  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
943  *	the 'final' error, but it doesn't make sense to solve the printing
944  *	delay with more complex code.
945  */
946 
947 int
948 badport_bandlim(int which)
949 {
950 
951 #define	N(a)	(sizeof (a) / sizeof (a[0]))
952 	static struct rate {
953 		const char	*type;
954 		struct timeval	lasttime;
955 		int		curpps;
956 	} rates[BANDLIM_MAX+1] = {
957 		{ "icmp unreach response" },
958 		{ "icmp ping response" },
959 		{ "icmp tstamp response" },
960 		{ "closed port RST response" },
961 		{ "open port RST response" },
962 		{ "icmp6 unreach response" },
963 		{ "sctp ootb response" }
964 	};
965 
966 	/*
967 	 * Return ok status if feature disabled or argument out of range.
968 	 */
969 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
970 		struct rate *r = &rates[which];
971 		int opps = r->curpps;
972 
973 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
974 			return -1;	/* discard packet */
975 		/*
976 		 * If we've dropped below the threshold after having
977 		 * rate-limited traffic print the message.  This preserves
978 		 * the previous behaviour at the expense of added complexity.
979 		 */
980 		if (V_icmplim_output && opps > V_icmplim)
981 			log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
982 				r->type, opps, V_icmplim);
983 	}
984 	return 0;			/* okay to send packet */
985 #undef N
986 }
987