xref: /freebsd/sys/netinet/ip_icmp.c (revision 41059135ce931c0f1014a999ffabc6bc470ce856)
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  * 3. 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 	struct ip *oip = mtod(n, struct ip *), *nip;
189 	unsigned oiphlen = oip->ip_hl << 2;
190 	struct icmp *icp;
191 	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 srcbuf[INET_ADDRSTRLEN];
384 		char dstbuf[INET_ADDRSTRLEN];
385 
386 		printf("icmp_input from %s to %s, len %d\n",
387 		    inet_ntoa_r(ip->ip_src, srcbuf),
388 		    inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
389 	}
390 #endif
391 	if (icmplen < ICMP_MINLEN) {
392 		ICMPSTAT_INC(icps_tooshort);
393 		goto freeit;
394 	}
395 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
396 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
397 		ICMPSTAT_INC(icps_tooshort);
398 		return (IPPROTO_DONE);
399 	}
400 	ip = mtod(m, struct ip *);
401 	m->m_len -= hlen;
402 	m->m_data += hlen;
403 	icp = mtod(m, struct icmp *);
404 	if (in_cksum(m, icmplen)) {
405 		ICMPSTAT_INC(icps_checksum);
406 		goto freeit;
407 	}
408 	m->m_len += hlen;
409 	m->m_data -= hlen;
410 
411 #ifdef ICMPPRINTFS
412 	if (icmpprintfs)
413 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
414 		    icp->icmp_code);
415 #endif
416 
417 	/*
418 	 * Message type specific processing.
419 	 */
420 	if (icp->icmp_type > ICMP_MAXTYPE)
421 		goto raw;
422 
423 	/* Initialize */
424 	bzero(&icmpsrc, sizeof(icmpsrc));
425 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
426 	icmpsrc.sin_family = AF_INET;
427 	bzero(&icmpdst, sizeof(icmpdst));
428 	icmpdst.sin_len = sizeof(struct sockaddr_in);
429 	icmpdst.sin_family = AF_INET;
430 	bzero(&icmpgw, sizeof(icmpgw));
431 	icmpgw.sin_len = sizeof(struct sockaddr_in);
432 	icmpgw.sin_family = AF_INET;
433 
434 	ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
435 	code = icp->icmp_code;
436 	switch (icp->icmp_type) {
437 
438 	case ICMP_UNREACH:
439 		switch (code) {
440 			case ICMP_UNREACH_NET:
441 			case ICMP_UNREACH_HOST:
442 			case ICMP_UNREACH_SRCFAIL:
443 			case ICMP_UNREACH_NET_UNKNOWN:
444 			case ICMP_UNREACH_HOST_UNKNOWN:
445 			case ICMP_UNREACH_ISOLATED:
446 			case ICMP_UNREACH_TOSNET:
447 			case ICMP_UNREACH_TOSHOST:
448 			case ICMP_UNREACH_HOST_PRECEDENCE:
449 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
450 				code = PRC_UNREACH_NET;
451 				break;
452 
453 			case ICMP_UNREACH_NEEDFRAG:
454 				code = PRC_MSGSIZE;
455 				break;
456 
457 			/*
458 			 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
459 			 * Treat subcodes 2,3 as immediate RST
460 			 */
461 			case ICMP_UNREACH_PROTOCOL:
462 				code = PRC_UNREACH_PROTOCOL;
463 				break;
464 			case ICMP_UNREACH_PORT:
465 				code = PRC_UNREACH_PORT;
466 				break;
467 
468 			case ICMP_UNREACH_NET_PROHIB:
469 			case ICMP_UNREACH_HOST_PROHIB:
470 			case ICMP_UNREACH_FILTER_PROHIB:
471 				code = PRC_UNREACH_ADMIN_PROHIB;
472 				break;
473 
474 			default:
475 				goto badcode;
476 		}
477 		goto deliver;
478 
479 	case ICMP_TIMXCEED:
480 		if (code > 1)
481 			goto badcode;
482 		code += PRC_TIMXCEED_INTRANS;
483 		goto deliver;
484 
485 	case ICMP_PARAMPROB:
486 		if (code > 1)
487 			goto badcode;
488 		code = PRC_PARAMPROB;
489 	deliver:
490 		/*
491 		 * Problem with datagram; advise higher level routines.
492 		 */
493 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
494 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
495 			ICMPSTAT_INC(icps_badlen);
496 			goto freeit;
497 		}
498 		/* Discard ICMP's in response to multicast packets */
499 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
500 			goto badcode;
501 #ifdef ICMPPRINTFS
502 		if (icmpprintfs)
503 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
504 #endif
505 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
506 		/*
507 		 * XXX if the packet contains [IPv4 AH TCP], we can't make a
508 		 * notification to TCP layer.
509 		 */
510 		i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
511 		ip_stripoptions(m);
512 		if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
513 			/* This should actually not happen */
514 			ICMPSTAT_INC(icps_tooshort);
515 			return (IPPROTO_DONE);
516 		}
517 		ip = mtod(m, struct ip *);
518 		icp = (struct icmp *)(ip + 1);
519 		/*
520 		 * The upper layer handler can rely on:
521 		 * - The outer IP header has no options.
522 		 * - The outer IP header, the ICMP header, the inner IP header,
523 		 *   and the first n bytes of the inner payload are contiguous.
524 		 *   n is at least 8, but might be larger based on
525 		 *   ICMP_ADVLENPREF. See its definition in ip_icmp.h.
526 		 */
527 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
528 		if (ctlfunc)
529 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
530 				   (void *)&icp->icmp_ip);
531 		break;
532 
533 	badcode:
534 		ICMPSTAT_INC(icps_badcode);
535 		break;
536 
537 	case ICMP_ECHO:
538 		if (!V_icmpbmcastecho
539 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
540 			ICMPSTAT_INC(icps_bmcastecho);
541 			break;
542 		}
543 		icp->icmp_type = ICMP_ECHOREPLY;
544 		if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
545 			goto freeit;
546 		else
547 			goto reflect;
548 
549 	case ICMP_TSTAMP:
550 		if (V_icmptstamprepl == 0)
551 			break;
552 		if (!V_icmpbmcastecho
553 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
554 			ICMPSTAT_INC(icps_bmcasttstamp);
555 			break;
556 		}
557 		if (icmplen < ICMP_TSLEN) {
558 			ICMPSTAT_INC(icps_badlen);
559 			break;
560 		}
561 		icp->icmp_type = ICMP_TSTAMPREPLY;
562 		icp->icmp_rtime = iptime();
563 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
564 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
565 			goto freeit;
566 		else
567 			goto reflect;
568 
569 	case ICMP_MASKREQ:
570 		if (V_icmpmaskrepl == 0)
571 			break;
572 		/*
573 		 * We are not able to respond with all ones broadcast
574 		 * unless we receive it over a point-to-point interface.
575 		 */
576 		if (icmplen < ICMP_MASKLEN)
577 			break;
578 		switch (ip->ip_dst.s_addr) {
579 
580 		case INADDR_BROADCAST:
581 		case INADDR_ANY:
582 			icmpdst.sin_addr = ip->ip_src;
583 			break;
584 
585 		default:
586 			icmpdst.sin_addr = ip->ip_dst;
587 		}
588 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
589 			    (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
590 		if (ia == NULL)
591 			break;
592 		if (ia->ia_ifp == NULL) {
593 			ifa_free(&ia->ia_ifa);
594 			break;
595 		}
596 		icp->icmp_type = ICMP_MASKREPLY;
597 		if (V_icmpmaskfake == 0)
598 			icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
599 		else
600 			icp->icmp_mask = V_icmpmaskfake;
601 		if (ip->ip_src.s_addr == 0) {
602 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
603 			    ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
604 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
605 			    ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
606 		}
607 		ifa_free(&ia->ia_ifa);
608 reflect:
609 		ICMPSTAT_INC(icps_reflect);
610 		ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
611 		icmp_reflect(m);
612 		return (IPPROTO_DONE);
613 
614 	case ICMP_REDIRECT:
615 		if (V_log_redirect) {
616 			u_long src, dst, gw;
617 
618 			src = ntohl(ip->ip_src.s_addr);
619 			dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
620 			gw = ntohl(icp->icmp_gwaddr.s_addr);
621 			printf("icmp redirect from %d.%d.%d.%d: "
622 			       "%d.%d.%d.%d => %d.%d.%d.%d\n",
623 			       (int)(src >> 24), (int)((src >> 16) & 0xff),
624 			       (int)((src >> 8) & 0xff), (int)(src & 0xff),
625 			       (int)(dst >> 24), (int)((dst >> 16) & 0xff),
626 			       (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
627 			       (int)(gw >> 24), (int)((gw >> 16) & 0xff),
628 			       (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
629 		}
630 		/*
631 		 * RFC1812 says we must ignore ICMP redirects if we
632 		 * are acting as router.
633 		 */
634 		if (V_drop_redirect || V_ipforwarding)
635 			break;
636 		if (code > 3)
637 			goto badcode;
638 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
639 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
640 			ICMPSTAT_INC(icps_badlen);
641 			break;
642 		}
643 		/*
644 		 * Short circuit routing redirects to force
645 		 * immediate change in the kernel's routing
646 		 * tables.  The message is also handed to anyone
647 		 * listening on a raw socket (e.g. the routing
648 		 * daemon for use in updating its tables).
649 		 */
650 		icmpgw.sin_addr = ip->ip_src;
651 		icmpdst.sin_addr = icp->icmp_gwaddr;
652 #ifdef	ICMPPRINTFS
653 		if (icmpprintfs) {
654 			char dstbuf[INET_ADDRSTRLEN];
655 			char gwbuf[INET_ADDRSTRLEN];
656 
657 			printf("redirect dst %s to %s\n",
658 			       inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
659 			       inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
660 		}
661 #endif
662 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
663 		for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
664 			in_rtredirect((struct sockaddr *)&icmpsrc,
665 			  (struct sockaddr *)&icmpdst,
666 			  (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
667 			  (struct sockaddr *)&icmpgw, fibnum);
668 		}
669 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
670 		break;
671 
672 	/*
673 	 * No kernel processing for the following;
674 	 * just fall through to send to raw listener.
675 	 */
676 	case ICMP_ECHOREPLY:
677 	case ICMP_ROUTERADVERT:
678 	case ICMP_ROUTERSOLICIT:
679 	case ICMP_TSTAMPREPLY:
680 	case ICMP_IREQREPLY:
681 	case ICMP_MASKREPLY:
682 	case ICMP_SOURCEQUENCH:
683 	default:
684 		break;
685 	}
686 
687 raw:
688 	*mp = m;
689 	rip_input(mp, offp, proto);
690 	return (IPPROTO_DONE);
691 
692 freeit:
693 	m_freem(m);
694 	return (IPPROTO_DONE);
695 }
696 
697 /*
698  * Reflect the ip packet back to the source
699  */
700 static void
701 icmp_reflect(struct mbuf *m)
702 {
703 	struct rm_priotracker in_ifa_tracker;
704 	struct ip *ip = mtod(m, struct ip *);
705 	struct ifaddr *ifa;
706 	struct ifnet *ifp;
707 	struct in_ifaddr *ia;
708 	struct in_addr t;
709 	struct nhop4_extended nh_ext;
710 	struct mbuf *opts = NULL;
711 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
712 
713 	if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
714 	    IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
715 	    IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
716 		m_freem(m);	/* Bad return address */
717 		ICMPSTAT_INC(icps_badaddr);
718 		goto done;	/* Ip_output() will check for broadcast */
719 	}
720 
721 	t = ip->ip_dst;
722 	ip->ip_dst = ip->ip_src;
723 
724 	/*
725 	 * Source selection for ICMP replies:
726 	 *
727 	 * If the incoming packet was addressed directly to one of our
728 	 * own addresses, use dst as the src for the reply.
729 	 */
730 	IN_IFADDR_RLOCK(&in_ifa_tracker);
731 	LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
732 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
733 			t = IA_SIN(ia)->sin_addr;
734 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
735 			goto match;
736 		}
737 	}
738 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
739 
740 	/*
741 	 * If the incoming packet was addressed to one of our broadcast
742 	 * addresses, use the first non-broadcast address which corresponds
743 	 * to the incoming interface.
744 	 */
745 	ifp = m->m_pkthdr.rcvif;
746 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
747 		IF_ADDR_RLOCK(ifp);
748 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
749 			if (ifa->ifa_addr->sa_family != AF_INET)
750 				continue;
751 			ia = ifatoia(ifa);
752 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
753 			    t.s_addr) {
754 				t = IA_SIN(ia)->sin_addr;
755 				IF_ADDR_RUNLOCK(ifp);
756 				goto match;
757 			}
758 		}
759 		IF_ADDR_RUNLOCK(ifp);
760 	}
761 	/*
762 	 * If the packet was transiting through us, use the address of
763 	 * the interface the packet came through in.  If that interface
764 	 * doesn't have a suitable IP address, the normal selection
765 	 * criteria apply.
766 	 */
767 	if (V_icmp_rfi && ifp != NULL) {
768 		IF_ADDR_RLOCK(ifp);
769 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
770 			if (ifa->ifa_addr->sa_family != AF_INET)
771 				continue;
772 			ia = ifatoia(ifa);
773 			t = IA_SIN(ia)->sin_addr;
774 			IF_ADDR_RUNLOCK(ifp);
775 			goto match;
776 		}
777 		IF_ADDR_RUNLOCK(ifp);
778 	}
779 	/*
780 	 * If the incoming packet was not addressed directly to us, use
781 	 * designated interface for icmp replies specified by sysctl
782 	 * net.inet.icmp.reply_src (default not set). Otherwise continue
783 	 * with normal source selection.
784 	 */
785 	if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
786 		IF_ADDR_RLOCK(ifp);
787 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
788 			if (ifa->ifa_addr->sa_family != AF_INET)
789 				continue;
790 			ia = ifatoia(ifa);
791 			t = IA_SIN(ia)->sin_addr;
792 			IF_ADDR_RUNLOCK(ifp);
793 			goto match;
794 		}
795 		IF_ADDR_RUNLOCK(ifp);
796 	}
797 	/*
798 	 * If the packet was transiting through us, use the address of
799 	 * the interface that is the closest to the packet source.
800 	 * When we don't have a route back to the packet source, stop here
801 	 * and drop the packet.
802 	 */
803 	if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) {
804 		m_freem(m);
805 		ICMPSTAT_INC(icps_noroute);
806 		goto done;
807 	}
808 	t = nh_ext.nh_src;
809 match:
810 #ifdef MAC
811 	mac_netinet_icmp_replyinplace(m);
812 #endif
813 	ip->ip_src = t;
814 	ip->ip_ttl = V_ip_defttl;
815 
816 	if (optlen > 0) {
817 		u_char *cp;
818 		int opt, cnt;
819 		u_int len;
820 
821 		/*
822 		 * Retrieve any source routing from the incoming packet;
823 		 * add on any record-route or timestamp options.
824 		 */
825 		cp = (u_char *) (ip + 1);
826 		if ((opts = ip_srcroute(m)) == NULL &&
827 		    (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
828 			opts->m_len = sizeof(struct in_addr);
829 			mtod(opts, struct in_addr *)->s_addr = 0;
830 		}
831 		if (opts) {
832 #ifdef ICMPPRINTFS
833 		    if (icmpprintfs)
834 			    printf("icmp_reflect optlen %d rt %d => ",
835 				optlen, opts->m_len);
836 #endif
837 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
838 			    opt = cp[IPOPT_OPTVAL];
839 			    if (opt == IPOPT_EOL)
840 				    break;
841 			    if (opt == IPOPT_NOP)
842 				    len = 1;
843 			    else {
844 				    if (cnt < IPOPT_OLEN + sizeof(*cp))
845 					    break;
846 				    len = cp[IPOPT_OLEN];
847 				    if (len < IPOPT_OLEN + sizeof(*cp) ||
848 				        len > cnt)
849 					    break;
850 			    }
851 			    /*
852 			     * Should check for overflow, but it "can't happen"
853 			     */
854 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
855 				opt == IPOPT_SECURITY) {
856 				    bcopy((caddr_t)cp,
857 					mtod(opts, caddr_t) + opts->m_len, len);
858 				    opts->m_len += len;
859 			    }
860 		    }
861 		    /* Terminate & pad, if necessary */
862 		    cnt = opts->m_len % 4;
863 		    if (cnt) {
864 			    for (; cnt < 4; cnt++) {
865 				    *(mtod(opts, caddr_t) + opts->m_len) =
866 					IPOPT_EOL;
867 				    opts->m_len++;
868 			    }
869 		    }
870 #ifdef ICMPPRINTFS
871 		    if (icmpprintfs)
872 			    printf("%d\n", opts->m_len);
873 #endif
874 		}
875 		ip_stripoptions(m);
876 	}
877 	m_tag_delete_nonpersistent(m);
878 	m->m_flags &= ~(M_BCAST|M_MCAST);
879 	icmp_send(m, opts);
880 done:
881 	if (opts)
882 		(void)m_free(opts);
883 }
884 
885 /*
886  * Send an icmp packet back to the ip level,
887  * after supplying a checksum.
888  */
889 static void
890 icmp_send(struct mbuf *m, struct mbuf *opts)
891 {
892 	struct ip *ip = mtod(m, struct ip *);
893 	int hlen;
894 	struct icmp *icp;
895 
896 	hlen = ip->ip_hl << 2;
897 	m->m_data += hlen;
898 	m->m_len -= hlen;
899 	icp = mtod(m, struct icmp *);
900 	icp->icmp_cksum = 0;
901 	icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
902 	m->m_data -= hlen;
903 	m->m_len += hlen;
904 	m->m_pkthdr.rcvif = (struct ifnet *)0;
905 #ifdef ICMPPRINTFS
906 	if (icmpprintfs) {
907 		char dstbuf[INET_ADDRSTRLEN];
908 		char srcbuf[INET_ADDRSTRLEN];
909 
910 		printf("icmp_send dst %s src %s\n",
911 		    inet_ntoa_r(ip->ip_dst, dstbuf),
912 		    inet_ntoa_r(ip->ip_src, srcbuf));
913 	}
914 #endif
915 	(void) ip_output(m, opts, NULL, 0, NULL, NULL);
916 }
917 
918 /*
919  * Return milliseconds since 00:00 UTC in network format.
920  */
921 uint32_t
922 iptime(void)
923 {
924 	struct timeval atv;
925 	u_long t;
926 
927 	getmicrotime(&atv);
928 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
929 	return (htonl(t));
930 }
931 
932 /*
933  * Return the next larger or smaller MTU plateau (table from RFC 1191)
934  * given current value MTU.  If DIR is less than zero, a larger plateau
935  * is returned; otherwise, a smaller value is returned.
936  */
937 int
938 ip_next_mtu(int mtu, int dir)
939 {
940 	static int mtutab[] = {
941 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
942 		296, 68, 0
943 	};
944 	int i, size;
945 
946 	size = (sizeof mtutab) / (sizeof mtutab[0]);
947 	if (dir >= 0) {
948 		for (i = 0; i < size; i++)
949 			if (mtu > mtutab[i])
950 				return mtutab[i];
951 	} else {
952 		for (i = size - 1; i >= 0; i--)
953 			if (mtu < mtutab[i])
954 				return mtutab[i];
955 		if (mtu == mtutab[0])
956 			return mtutab[0];
957 	}
958 	return 0;
959 }
960 #endif /* INET */
961 
962 
963 /*
964  * badport_bandlim() - check for ICMP bandwidth limit
965  *
966  *	Return 0 if it is ok to send an ICMP error response, -1 if we have
967  *	hit our bandwidth limit and it is not ok.
968  *
969  *	If icmplim is <= 0, the feature is disabled and 0 is returned.
970  *
971  *	For now we separate the TCP and UDP subsystems w/ different 'which'
972  *	values.  We may eventually remove this separation (and simplify the
973  *	code further).
974  *
975  *	Note that the printing of the error message is delayed so we can
976  *	properly print the icmp error rate that the system was trying to do
977  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
978  *	the 'final' error, but it doesn't make sense to solve the printing
979  *	delay with more complex code.
980  */
981 struct icmp_rate {
982 	const char *descr;
983 	struct counter_rate cr;
984 };
985 static VNET_DEFINE(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = {
986 	{ "icmp unreach response" },
987 	{ "icmp ping response" },
988 	{ "icmp tstamp response" },
989 	{ "closed port RST response" },
990 	{ "open port RST response" },
991 	{ "icmp6 unreach response" },
992 	{ "sctp ootb response" }
993 };
994 #define	V_icmp_rates	VNET(icmp_rates)
995 
996 static void
997 icmp_bandlimit_init(void)
998 {
999 
1000 	for (int i = 0; i < BANDLIM_MAX; i++) {
1001 		V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK);
1002 		V_icmp_rates[i].cr.cr_ticks = ticks;
1003 	}
1004 }
1005 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
1006     icmp_bandlimit_init, NULL);
1007 
1008 static void
1009 icmp_bandlimit_uninit(void)
1010 {
1011 
1012 	for (int i = 0; i < BANDLIM_MAX; i++)
1013 		counter_u64_free(V_icmp_rates[i].cr.cr_rate);
1014 }
1015 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
1016     icmp_bandlimit_uninit, NULL);
1017 
1018 int
1019 badport_bandlim(int which)
1020 {
1021 	int64_t pps;
1022 
1023 	if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
1024 		return (0);
1025 
1026 	KASSERT(which >= 0 && which < BANDLIM_MAX,
1027 	    ("%s: which %d", __func__, which));
1028 
1029 	pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim);
1030 	if (pps == -1)
1031 		return (-1);
1032 	if (pps > 0 && V_icmplim_output)
1033 		log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n",
1034 			V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim);
1035 	return (0);
1036 }
1037