xref: /freebsd/sys/netinet/ip_icmp.c (revision 3fc36ee018bb836bd1796067cf4ef8683f166ebc)
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/sctp.h>
65 #include <netinet/tcp.h>
66 #include <netinet/tcp_var.h>
67 #include <netinet/tcpip.h>
68 #include <netinet/icmp_var.h>
69 
70 #ifdef INET
71 
72 #include <machine/in_cksum.h>
73 
74 #include <security/mac/mac_framework.h>
75 #endif /* INET */
76 
77 /*
78  * ICMP routines: error generation, receive packet processing, and
79  * routines to turnaround packets back to the originator, and
80  * host table maintenance routines.
81  */
82 static VNET_DEFINE(int, icmplim) = 200;
83 #define	V_icmplim			VNET(icmplim)
84 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
85 	&VNET_NAME(icmplim), 0,
86 	"Maximum number of ICMP responses per second");
87 
88 static VNET_DEFINE(int, icmplim_output) = 1;
89 #define	V_icmplim_output		VNET(icmplim_output)
90 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
91 	&VNET_NAME(icmplim_output), 0,
92 	"Enable logging of ICMP response rate limiting");
93 
94 #ifdef INET
95 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
96 VNET_PCPUSTAT_SYSINIT(icmpstat);
97 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
98     icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
99 
100 #ifdef VIMAGE
101 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
102 #endif /* VIMAGE */
103 
104 static VNET_DEFINE(int, icmpmaskrepl) = 0;
105 #define	V_icmpmaskrepl			VNET(icmpmaskrepl)
106 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
107 	&VNET_NAME(icmpmaskrepl), 0,
108 	"Reply to ICMP Address Mask Request packets");
109 
110 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
111 #define	V_icmpmaskfake			VNET(icmpmaskfake)
112 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
113 	&VNET_NAME(icmpmaskfake), 0,
114 	"Fake reply to ICMP Address Mask Request packets");
115 
116 VNET_DEFINE(int, drop_redirect) = 0;
117 #define	V_drop_redirect			VNET(drop_redirect)
118 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
119 	&VNET_NAME(drop_redirect), 0,
120 	"Ignore ICMP redirects");
121 
122 static VNET_DEFINE(int, log_redirect) = 0;
123 #define	V_log_redirect			VNET(log_redirect)
124 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
125 	&VNET_NAME(log_redirect), 0,
126 	"Log ICMP redirects to the console");
127 
128 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
129 #define	V_reply_src			VNET(reply_src)
130 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
131 	&VNET_NAME(reply_src), IFNAMSIZ,
132 	"ICMP reply source for non-local packets");
133 
134 static VNET_DEFINE(int, icmp_rfi) = 0;
135 #define	V_icmp_rfi			VNET(icmp_rfi)
136 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
137 	&VNET_NAME(icmp_rfi), 0,
138 	"ICMP reply from incoming interface for non-local packets");
139 
140 static VNET_DEFINE(int, icmp_quotelen) = 8;
141 #define	V_icmp_quotelen			VNET(icmp_quotelen)
142 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
143 	&VNET_NAME(icmp_quotelen), 0,
144 	"Number of bytes from original packet to quote in ICMP reply");
145 
146 static VNET_DEFINE(int, icmpbmcastecho) = 0;
147 #define	V_icmpbmcastecho		VNET(icmpbmcastecho)
148 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
149 	&VNET_NAME(icmpbmcastecho), 0,
150 	"Reply to multicast ICMP Echo Request and Timestamp packets");
151 
152 static VNET_DEFINE(int, icmptstamprepl) = 1;
153 #define	V_icmptstamprepl		VNET(icmptstamprepl)
154 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_RW,
155 	&VNET_NAME(icmptstamprepl), 0,
156 	"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 	 * Unfortunately 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 if (oip->ip_p == IPPROTO_SCTP) {
254 		struct sctphdr *sh;
255 		struct sctp_chunkhdr *ch;
256 
257 		if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
258 			goto stdreply;
259 		if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
260 		    n->m_next == NULL)
261 			goto stdreply;
262 		if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
263 		    (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
264 			goto freeit;
265 		icmpelen = max(sizeof(struct sctphdr),
266 		    min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
267 		sh = (struct sctphdr *)((caddr_t)oip + oiphlen);
268 		if (ntohl(sh->v_tag) == 0 &&
269 		    ntohs(oip->ip_len) >= oiphlen + sizeof(struct sctphdr) + 8 &&
270 		    (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
271 		     n->m_next != NULL)) {
272 			if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
273 			    (n = m_pullup(n, oiphlen + sizeof(struct sctphdr) + 8)) == NULL)
274 				goto freeit;
275 			ch = (struct sctp_chunkhdr *)(sh + 1);
276 			if (ch->chunk_type == SCTP_INITIATION) {
277 				icmpelen = max(sizeof(struct sctphdr) + 8,
278 				    min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
279 			}
280 		}
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 int
362 icmp_input(struct mbuf **mp, int *offp, int proto)
363 {
364 	struct icmp *icp;
365 	struct in_ifaddr *ia;
366 	struct mbuf *m = *mp;
367 	struct ip *ip = mtod(m, struct ip *);
368 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
369 	int hlen = *offp;
370 	int icmplen = ntohs(ip->ip_len) - *offp;
371 	int i, code;
372 	void (*ctlfunc)(int, struct sockaddr *, void *);
373 	int fibnum;
374 
375 	*mp = NULL;
376 
377 	/*
378 	 * Locate icmp structure in mbuf, and check
379 	 * that not corrupted and of at least minimum length.
380 	 */
381 #ifdef ICMPPRINTFS
382 	if (icmpprintfs) {
383 		char buf[4 * sizeof "123"];
384 		strcpy(buf, inet_ntoa(ip->ip_src));
385 		printf("icmp_input from %s to %s, len %d\n",
386 		       buf, inet_ntoa(ip->ip_dst), icmplen);
387 	}
388 #endif
389 	if (icmplen < ICMP_MINLEN) {
390 		ICMPSTAT_INC(icps_tooshort);
391 		goto freeit;
392 	}
393 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
394 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
395 		ICMPSTAT_INC(icps_tooshort);
396 		return (IPPROTO_DONE);
397 	}
398 	ip = mtod(m, struct ip *);
399 	m->m_len -= hlen;
400 	m->m_data += hlen;
401 	icp = mtod(m, struct icmp *);
402 	if (in_cksum(m, icmplen)) {
403 		ICMPSTAT_INC(icps_checksum);
404 		goto freeit;
405 	}
406 	m->m_len += hlen;
407 	m->m_data -= hlen;
408 
409 #ifdef ICMPPRINTFS
410 	if (icmpprintfs)
411 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
412 		    icp->icmp_code);
413 #endif
414 
415 	/*
416 	 * Message type specific processing.
417 	 */
418 	if (icp->icmp_type > ICMP_MAXTYPE)
419 		goto raw;
420 
421 	/* Initialize */
422 	bzero(&icmpsrc, sizeof(icmpsrc));
423 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
424 	icmpsrc.sin_family = AF_INET;
425 	bzero(&icmpdst, sizeof(icmpdst));
426 	icmpdst.sin_len = sizeof(struct sockaddr_in);
427 	icmpdst.sin_family = AF_INET;
428 	bzero(&icmpgw, sizeof(icmpgw));
429 	icmpgw.sin_len = sizeof(struct sockaddr_in);
430 	icmpgw.sin_family = AF_INET;
431 
432 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
433 	code = icp->icmp_code;
434 	switch (icp->icmp_type) {
435 
436 	case ICMP_UNREACH:
437 		switch (code) {
438 			case ICMP_UNREACH_NET:
439 			case ICMP_UNREACH_HOST:
440 			case ICMP_UNREACH_SRCFAIL:
441 			case ICMP_UNREACH_NET_UNKNOWN:
442 			case ICMP_UNREACH_HOST_UNKNOWN:
443 			case ICMP_UNREACH_ISOLATED:
444 			case ICMP_UNREACH_TOSNET:
445 			case ICMP_UNREACH_TOSHOST:
446 			case ICMP_UNREACH_HOST_PRECEDENCE:
447 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
448 				code = PRC_UNREACH_NET;
449 				break;
450 
451 			case ICMP_UNREACH_NEEDFRAG:
452 				code = PRC_MSGSIZE;
453 				break;
454 
455 			/*
456 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
457 			 * Treat subcodes 2,3 as immediate RST
458 			 */
459 			case ICMP_UNREACH_PROTOCOL:
460 			case ICMP_UNREACH_PORT:
461 				code = PRC_UNREACH_PORT;
462 				break;
463 
464 			case ICMP_UNREACH_NET_PROHIB:
465 			case ICMP_UNREACH_HOST_PROHIB:
466 			case ICMP_UNREACH_FILTER_PROHIB:
467 				code = PRC_UNREACH_ADMIN_PROHIB;
468 				break;
469 
470 			default:
471 				goto badcode;
472 		}
473 		goto deliver;
474 
475 	case ICMP_TIMXCEED:
476 		if (code > 1)
477 			goto badcode;
478 		code += PRC_TIMXCEED_INTRANS;
479 		goto deliver;
480 
481 	case ICMP_PARAMPROB:
482 		if (code > 1)
483 			goto badcode;
484 		code = PRC_PARAMPROB;
485 	deliver:
486 		/*
487 		 * Problem with datagram; advise higher level routines.
488 		 */
489 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
490 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
491 			ICMPSTAT_INC(icps_badlen);
492 			goto freeit;
493 		}
494 		/* Discard ICMP's in response to multicast packets */
495 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
496 			goto badcode;
497 #ifdef ICMPPRINTFS
498 		if (icmpprintfs)
499 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
500 #endif
501 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
502 		/*
503 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
504 		 * notification to TCP layer.
505 		 */
506 		i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
507 		ip_stripoptions(m);
508 		if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
509 			/* This should actually not happen */
510 			ICMPSTAT_INC(icps_tooshort);
511 			return (IPPROTO_DONE);
512 		}
513 		ip = mtod(m, struct ip *);
514 		icp = (struct icmp *)(ip + 1);
515 		/*
516 		 * The upper layer handler can rely on:
517 		 * - The outer IP header has no options.
518 		 * - The outer IP header, the ICMP header, the inner IP header,
519 		 *   and the first n bytes of the inner payload are contiguous.
520 		 *   n is at least 8, but might be larger based on
521 		 *   ICMP_ADVLENPREF. See its definition in ip_icmp.h.
522 		 */
523 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
524 		if (ctlfunc)
525 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
526 				   (void *)&icp->icmp_ip);
527 		break;
528 
529 	badcode:
530 		ICMPSTAT_INC(icps_badcode);
531 		break;
532 
533 	case ICMP_ECHO:
534 		if (!V_icmpbmcastecho
535 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
536 			ICMPSTAT_INC(icps_bmcastecho);
537 			break;
538 		}
539 		icp->icmp_type = ICMP_ECHOREPLY;
540 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
541 			goto freeit;
542 		else
543 			goto reflect;
544 
545 	case ICMP_TSTAMP:
546 		if (V_icmptstamprepl == 0)
547 			break;
548 		if (!V_icmpbmcastecho
549 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
550 			ICMPSTAT_INC(icps_bmcasttstamp);
551 			break;
552 		}
553 		if (icmplen < ICMP_TSLEN) {
554 			ICMPSTAT_INC(icps_badlen);
555 			break;
556 		}
557 		icp->icmp_type = ICMP_TSTAMPREPLY;
558 		icp->icmp_rtime = iptime();
559 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
560 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
561 			goto freeit;
562 		else
563 			goto reflect;
564 
565 	case ICMP_MASKREQ:
566 		if (V_icmpmaskrepl == 0)
567 			break;
568 		/*
569 		 * We are not able to respond with all ones broadcast
570 		 * unless we receive it over a point-to-point interface.
571 		 */
572 		if (icmplen < ICMP_MASKLEN)
573 			break;
574 		switch (ip->ip_dst.s_addr) {
575 
576 		case INADDR_BROADCAST:
577 		case INADDR_ANY:
578 			icmpdst.sin_addr = ip->ip_src;
579 			break;
580 
581 		default:
582 			icmpdst.sin_addr = ip->ip_dst;
583 		}
584 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
585 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
586 		if (ia == NULL)
587 			break;
588 		if (ia->ia_ifp == NULL) {
589 			ifa_free(&ia->ia_ifa);
590 			break;
591 		}
592 		icp->icmp_type = ICMP_MASKREPLY;
593 		if (V_icmpmaskfake == 0)
594 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
595 		else
596 			icp->icmp_mask = V_icmpmaskfake;
597 		if (ip->ip_src.s_addr == 0) {
598 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
599 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
600 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
601 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
602 		}
603 		ifa_free(&ia->ia_ifa);
604 reflect:
605 		ICMPSTAT_INC(icps_reflect);
606 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
607 		icmp_reflect(m);
608 		return (IPPROTO_DONE);
609 
610 	case ICMP_REDIRECT:
611 		if (V_log_redirect) {
612 			u_long src, dst, gw;
613 
614 			src = ntohl(ip->ip_src.s_addr);
615 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
616 			gw = ntohl(icp->icmp_gwaddr.s_addr);
617 			printf("icmp redirect from %d.%d.%d.%d: "
618 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
619 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
620 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
621 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
622 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
623 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
624 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
625 		}
626 		/*
627 		 * RFC1812 says we must ignore ICMP redirects if we
628 		 * are acting as router.
629 		 */
630 		if (V_drop_redirect || V_ipforwarding)
631 			break;
632 		if (code > 3)
633 			goto badcode;
634 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
635 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
636 			ICMPSTAT_INC(icps_badlen);
637 			break;
638 		}
639 		/*
640 		 * Short circuit routing redirects to force
641 		 * immediate change in the kernel's routing
642 		 * tables.  The message is also handed to anyone
643 		 * listening on a raw socket (e.g. the routing
644 		 * daemon for use in updating its tables).
645 		 */
646 		icmpgw.sin_addr = ip->ip_src;
647 		icmpdst.sin_addr = icp->icmp_gwaddr;
648 #ifdef	ICMPPRINTFS
649 		if (icmpprintfs) {
650 			char buf[4 * sizeof "123"];
651 			strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
652 
653 			printf("redirect dst %s to %s\n",
654 			       buf, inet_ntoa(icp->icmp_gwaddr));
655 		}
656 #endif
657 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
658 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
659 			in_rtredirect((struct sockaddr *)&icmpsrc,
660 			  (struct sockaddr *)&icmpdst,
661 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
662 			  (struct sockaddr *)&icmpgw, fibnum);
663 		}
664 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
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 	case ICMP_SOURCEQUENCH:
678 	default:
679 		break;
680 	}
681 
682 raw:
683 	*mp = m;
684 	rip_input(mp, offp, proto);
685 	return (IPPROTO_DONE);
686 
687 freeit:
688 	m_freem(m);
689 	return (IPPROTO_DONE);
690 }
691 
692 /*
693  * Reflect the ip packet back to the source
694  */
695 static void
696 icmp_reflect(struct mbuf *m)
697 {
698 	struct rm_priotracker in_ifa_tracker;
699 	struct ip *ip = mtod(m, struct ip *);
700 	struct ifaddr *ifa;
701 	struct ifnet *ifp;
702 	struct in_ifaddr *ia;
703 	struct in_addr t;
704 	struct nhop4_extended nh_ext;
705 	struct mbuf *opts = NULL;
706 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
707 
708 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
709 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
710 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
711 		m_freem(m);	/* Bad return address */
712 		ICMPSTAT_INC(icps_badaddr);
713 		goto done;	/* Ip_output() will check for broadcast */
714 	}
715 
716 	t = ip->ip_dst;
717 	ip->ip_dst = ip->ip_src;
718 
719 	/*
720 	 * Source selection for ICMP replies:
721 	 *
722 	 * If the incoming packet was addressed directly to one of our
723 	 * own addresses, use dst as the src for the reply.
724 	 */
725 	IN_IFADDR_RLOCK(&in_ifa_tracker);
726 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
727 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
728 			t = IA_SIN(ia)->sin_addr;
729 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
730 			goto match;
731 		}
732 	}
733 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
734 
735 	/*
736 	 * If the incoming packet was addressed to one of our broadcast
737 	 * addresses, use the first non-broadcast address which corresponds
738 	 * to the incoming interface.
739 	 */
740 	ifp = m->m_pkthdr.rcvif;
741 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
742 		IF_ADDR_RLOCK(ifp);
743 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
744 			if (ifa->ifa_addr->sa_family != AF_INET)
745 				continue;
746 			ia = ifatoia(ifa);
747 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
748 			    t.s_addr) {
749 				t = IA_SIN(ia)->sin_addr;
750 				IF_ADDR_RUNLOCK(ifp);
751 				goto match;
752 			}
753 		}
754 		IF_ADDR_RUNLOCK(ifp);
755 	}
756 	/*
757 	 * If the packet was transiting through us, use the address of
758 	 * the interface the packet came through in.  If that interface
759 	 * doesn't have a suitable IP address, the normal selection
760 	 * criteria apply.
761 	 */
762 	if (V_icmp_rfi && ifp != NULL) {
763 		IF_ADDR_RLOCK(ifp);
764 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
765 			if (ifa->ifa_addr->sa_family != AF_INET)
766 				continue;
767 			ia = ifatoia(ifa);
768 			t = IA_SIN(ia)->sin_addr;
769 			IF_ADDR_RUNLOCK(ifp);
770 			goto match;
771 		}
772 		IF_ADDR_RUNLOCK(ifp);
773 	}
774 	/*
775 	 * If the incoming packet was not addressed directly to us, use
776 	 * designated interface for icmp replies specified by sysctl
777 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
778 	 * with normal source selection.
779 	 */
780 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
781 		IF_ADDR_RLOCK(ifp);
782 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
783 			if (ifa->ifa_addr->sa_family != AF_INET)
784 				continue;
785 			ia = ifatoia(ifa);
786 			t = IA_SIN(ia)->sin_addr;
787 			IF_ADDR_RUNLOCK(ifp);
788 			goto match;
789 		}
790 		IF_ADDR_RUNLOCK(ifp);
791 	}
792 	/*
793 	 * If the packet was transiting through us, use the address of
794 	 * the interface that is the closest to the packet source.
795 	 * When we don't have a route back to the packet source, stop here
796 	 * and drop the packet.
797 	 */
798 	if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) {
799 		m_freem(m);
800 		ICMPSTAT_INC(icps_noroute);
801 		goto done;
802 	}
803 	t = nh_ext.nh_src;
804 match:
805 #ifdef MAC
806 	mac_netinet_icmp_replyinplace(m);
807 #endif
808 	ip->ip_src = t;
809 	ip->ip_ttl = V_ip_defttl;
810 
811 	if (optlen > 0) {
812 		register u_char *cp;
813 		int opt, cnt;
814 		u_int len;
815 
816 		/*
817 		 * Retrieve any source routing from the incoming packet;
818 		 * add on any record-route or timestamp options.
819 		 */
820 		cp = (u_char *) (ip + 1);
821 		if ((opts = ip_srcroute(m)) == NULL &&
822 		    (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
823 			opts->m_len = sizeof(struct in_addr);
824 			mtod(opts, struct in_addr *)->s_addr = 0;
825 		}
826 		if (opts) {
827 #ifdef ICMPPRINTFS
828 		    if (icmpprintfs)
829 			    printf("icmp_reflect optlen %d rt %d => ",
830 				optlen, opts->m_len);
831 #endif
832 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
833 			    opt = cp[IPOPT_OPTVAL];
834 			    if (opt == IPOPT_EOL)
835 				    break;
836 			    if (opt == IPOPT_NOP)
837 				    len = 1;
838 			    else {
839 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
840 					    break;
841 				    len = cp[IPOPT_OLEN];
842 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
843 				        len > cnt)
844 					    break;
845 			    }
846 			    /*
847 			     * Should check for overflow, but it "can't happen"
848 			     */
849 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
850 				opt == IPOPT_SECURITY) {
851 				    bcopy((caddr_t)cp,
852 					mtod(opts, caddr_t) + opts->m_len, len);
853 				    opts->m_len += len;
854 			    }
855 		    }
856 		    /* Terminate & pad, if necessary */
857 		    cnt = opts->m_len % 4;
858 		    if (cnt) {
859 			    for (; cnt < 4; cnt++) {
860 				    *(mtod(opts, caddr_t) + opts->m_len) =
861 					IPOPT_EOL;
862 				    opts->m_len++;
863 			    }
864 		    }
865 #ifdef ICMPPRINTFS
866 		    if (icmpprintfs)
867 			    printf("%d\n", opts->m_len);
868 #endif
869 		}
870 		ip_stripoptions(m);
871 	}
872 	m_tag_delete_nonpersistent(m);
873 	m->m_flags &= ~(M_BCAST|M_MCAST);
874 	icmp_send(m, opts);
875 done:
876 	if (opts)
877 		(void)m_free(opts);
878 }
879 
880 /*
881  * Send an icmp packet back to the ip level,
882  * after supplying a checksum.
883  */
884 static void
885 icmp_send(struct mbuf *m, struct mbuf *opts)
886 {
887 	register struct ip *ip = mtod(m, struct ip *);
888 	register int hlen;
889 	register struct icmp *icp;
890 
891 	hlen = ip->ip_hl << 2;
892 	m->m_data += hlen;
893 	m->m_len -= hlen;
894 	icp = mtod(m, struct icmp *);
895 	icp->icmp_cksum = 0;
896 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
897 	m->m_data -= hlen;
898 	m->m_len += hlen;
899 	m->m_pkthdr.rcvif = (struct ifnet *)0;
900 #ifdef ICMPPRINTFS
901 	if (icmpprintfs) {
902 		char buf[4 * sizeof "123"];
903 		strcpy(buf, inet_ntoa(ip->ip_dst));
904 		printf("icmp_send dst %s src %s\n",
905 		       buf, inet_ntoa(ip->ip_src));
906 	}
907 #endif
908 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
909 }
910 
911 /*
912  * Return milliseconds since 00:00 UTC in network format.
913  */
914 uint32_t
915 iptime(void)
916 {
917 	struct timeval atv;
918 	u_long t;
919 
920 	getmicrotime(&atv);
921 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
922 	return (htonl(t));
923 }
924 
925 /*
926  * Return the next larger or smaller MTU plateau (table from RFC 1191)
927  * given current value MTU.  If DIR is less than zero, a larger plateau
928  * is returned; otherwise, a smaller value is returned.
929  */
930 int
931 ip_next_mtu(int mtu, int dir)
932 {
933 	static int mtutab[] = {
934 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
935 		296, 68, 0
936 	};
937 	int i, size;
938 
939 	size = (sizeof mtutab) / (sizeof mtutab[0]);
940 	if (dir >= 0) {
941 		for (i = 0; i < size; i++)
942 			if (mtu > mtutab[i])
943 				return mtutab[i];
944 	} else {
945 		for (i = size - 1; i >= 0; i--)
946 			if (mtu < mtutab[i])
947 				return mtutab[i];
948 		if (mtu == mtutab[0])
949 			return mtutab[0];
950 	}
951 	return 0;
952 }
953 #endif /* INET */
954 
955 
956 /*
957  * badport_bandlim() - check for ICMP bandwidth limit
958  *
959  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
960  *	hit our bandwidth limit and it is not ok.
961  *
962  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
963  *
964  *	For now we separate the TCP and UDP subsystems w/ different 'which'
965  *	values.  We may eventually remove this separation (and simplify the
966  *	code further).
967  *
968  *	Note that the printing of the error message is delayed so we can
969  *	properly print the icmp error rate that the system was trying to do
970  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
971  *	the 'final' error, but it doesn't make sense to solve the printing
972  *	delay with more complex code.
973  */
974 
975 int
976 badport_bandlim(int which)
977 {
978 
979 #define	N(a)	(sizeof (a) / sizeof (a[0]))
980 	static struct rate {
981 		const char	*type;
982 		struct timeval	lasttime;
983 		int		curpps;
984 	} rates[BANDLIM_MAX+1] = {
985 		{ "icmp unreach response" },
986 		{ "icmp ping response" },
987 		{ "icmp tstamp response" },
988 		{ "closed port RST response" },
989 		{ "open port RST response" },
990 		{ "icmp6 unreach response" },
991 		{ "sctp ootb response" }
992 	};
993 
994 	/*
995 	 * Return ok status if feature disabled or argument out of range.
996 	 */
997 	if (V_icmplim > 0 && (u_int) which < N(rates)) {
998 		struct rate *r = &rates[which];
999 		int opps = r->curpps;
1000 
1001 		if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
1002 			return -1;	/* discard packet */
1003 		/*
1004 		 * If we've dropped below the threshold after having
1005 		 * rate-limited traffic print the message.  This preserves
1006 		 * the previous behaviour at the expense of added complexity.
1007 		 */
1008 		if (V_icmplim_output && opps > V_icmplim)
1009 			log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
1010 				r->type, opps, V_icmplim);
1011 	}
1012 	return 0;			/* okay to send packet */
1013 #undef N
1014 }
1015