xref: /openbsd-src/sys/netinet6/icmp6.c (revision 46035553bfdd96e63c94e32da0210227ec2e3cf1)
1 /*	$OpenBSD: icmp6.c,v 1.233 2020/10/28 17:27:35 bluhm Exp $	*/
2 /*	$KAME: icmp6.c,v 1.217 2001/06/20 15:03:29 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "carp.h"
65 #include "pf.h"
66 
67 #include <sys/param.h>
68 #include <sys/systm.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/sysctl.h>
72 #include <sys/protosw.h>
73 #include <sys/socket.h>
74 #include <sys/socketvar.h>
75 #include <sys/time.h>
76 #include <sys/kernel.h>
77 #include <sys/syslog.h>
78 #include <sys/domain.h>
79 
80 #include <net/if.h>
81 #include <net/if_var.h>
82 #include <net/route.h>
83 #include <net/if_dl.h>
84 #include <net/if_types.h>
85 
86 #include <netinet/in.h>
87 #include <netinet/ip.h>
88 #include <netinet6/in6_var.h>
89 #include <netinet/ip6.h>
90 #include <netinet6/ip6_var.h>
91 #include <netinet/icmp6.h>
92 #include <netinet6/mld6_var.h>
93 #include <netinet/in_pcb.h>
94 #include <netinet6/nd6.h>
95 #include <netinet6/ip6protosw.h>
96 
97 #if NCARP > 0
98 #include <netinet/ip_carp.h>
99 #endif
100 
101 #if NPF > 0
102 #include <net/pfvar.h>
103 #endif
104 
105 struct cpumem *icmp6counters;
106 
107 extern int icmp6errppslim;
108 static int icmp6errpps_count = 0;
109 static struct timeval icmp6errppslim_last;
110 
111 /*
112  * List of callbacks to notify when Path MTU changes are made.
113  */
114 struct icmp6_mtudisc_callback {
115 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
116 	void (*mc_func)(struct sockaddr_in6 *, u_int);
117 };
118 
119 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
120     LIST_HEAD_INITIALIZER(icmp6_mtudisc_callbacks);
121 
122 struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
123 
124 /* XXX do these values make any sense? */
125 static int icmp6_mtudisc_hiwat = 1280;
126 static int icmp6_mtudisc_lowat = 256;
127 
128 /*
129  * keep track of # of redirect routes.
130  */
131 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
132 
133 /* XXX experimental, turned off */
134 static int icmp6_redirect_lowat = -1;
135 
136 void	icmp6_errcount(int, int);
137 int	icmp6_ratelimit(const struct in6_addr *, const int, const int);
138 const char *icmp6_redirect_diag(struct in6_addr *, struct in6_addr *,
139 	    struct in6_addr *);
140 int	icmp6_notify_error(struct mbuf *, int, int, int);
141 struct rtentry *icmp6_mtudisc_clone(struct sockaddr *, u_int);
142 void	icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
143 void	icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
144 
145 void
146 icmp6_init(void)
147 {
148 	mld6_init();
149 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(ip6_mtudisc_timeout);
150 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
151 	icmp6counters = counters_alloc(icp6s_ncounters);
152 }
153 
154 void
155 icmp6_errcount(int type, int code)
156 {
157 	enum icmp6stat_counters c = icp6s_ounknown;
158 
159 	switch (type) {
160 	case ICMP6_DST_UNREACH:
161 		switch (code) {
162 		case ICMP6_DST_UNREACH_NOROUTE:
163 			c = icp6s_odst_unreach_noroute;
164 			break;
165 		case ICMP6_DST_UNREACH_ADMIN:
166 			c = icp6s_odst_unreach_admin;
167 			break;
168 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
169 			c = icp6s_odst_unreach_beyondscope;
170 			break;
171 		case ICMP6_DST_UNREACH_ADDR:
172 			c = icp6s_odst_unreach_addr;
173 			break;
174 		case ICMP6_DST_UNREACH_NOPORT:
175 			c = icp6s_odst_unreach_noport;
176 			break;
177 		}
178 		break;
179 	case ICMP6_PACKET_TOO_BIG:
180 		c = icp6s_opacket_too_big;
181 		break;
182 	case ICMP6_TIME_EXCEEDED:
183 		switch (code) {
184 		case ICMP6_TIME_EXCEED_TRANSIT:
185 			c = icp6s_otime_exceed_transit;
186 			break;
187 		case ICMP6_TIME_EXCEED_REASSEMBLY:
188 			c = icp6s_otime_exceed_reassembly;
189 			break;
190 		}
191 		break;
192 	case ICMP6_PARAM_PROB:
193 		switch (code) {
194 		case ICMP6_PARAMPROB_HEADER:
195 			c = icp6s_oparamprob_header;
196 			break;
197 		case ICMP6_PARAMPROB_NEXTHEADER:
198 			c = icp6s_oparamprob_nextheader;
199 			break;
200 		case ICMP6_PARAMPROB_OPTION:
201 			c = icp6s_oparamprob_option;
202 			break;
203 		}
204 		break;
205 	case ND_REDIRECT:
206 		c = icp6s_oredirect;
207 		break;
208 	}
209 
210 	icmp6stat_inc(c);
211 }
212 
213 /*
214  * Register a Path MTU Discovery callback.
215  */
216 void
217 icmp6_mtudisc_callback_register(void (*func)(struct sockaddr_in6 *, u_int))
218 {
219 	struct icmp6_mtudisc_callback *mc;
220 
221 	LIST_FOREACH(mc, &icmp6_mtudisc_callbacks, mc_list) {
222 		if (mc->mc_func == func)
223 			return;
224 	}
225 
226 	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
227 	if (mc == NULL)
228 		panic("%s", __func__);
229 
230 	mc->mc_func = func;
231 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
232 }
233 
234 struct mbuf *
235 icmp6_do_error(struct mbuf *m, int type, int code, int param)
236 {
237 	struct ip6_hdr *oip6, *nip6;
238 	struct icmp6_hdr *icmp6;
239 	u_int preplen;
240 	int off;
241 	int nxt;
242 
243 	icmp6stat_inc(icp6s_error);
244 
245 	/* count per-type-code statistics */
246 	icmp6_errcount(type, code);
247 
248 	if (m->m_len < sizeof(struct ip6_hdr)) {
249 		m = m_pullup(m, sizeof(struct ip6_hdr));
250 		if (m == NULL)
251 			return (NULL);
252 	}
253 	oip6 = mtod(m, struct ip6_hdr *);
254 
255 	/*
256 	 * If the destination address of the erroneous packet is a multicast
257 	 * address, or the packet was sent using link-layer multicast,
258 	 * we should basically suppress sending an error (RFC 2463, Section
259 	 * 2.4).
260 	 * We have two exceptions (the item e.2 in that section):
261 	 * - the Packet Too Big message can be sent for path MTU discovery.
262 	 * - the Parameter Problem Message that can be allowed an icmp6 error
263 	 *   in the option type field.  This check has been done in
264 	 *   ip6_unknown_opt(), so we can just check the type and code.
265 	 */
266 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
267 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
268 	    (type != ICMP6_PACKET_TOO_BIG &&
269 	     (type != ICMP6_PARAM_PROB ||
270 	      code != ICMP6_PARAMPROB_OPTION)))
271 		goto freeit;
272 
273 	/*
274 	 * RFC 2463, 2.4 (e.5): source address check.
275 	 * XXX: the case of anycast source?
276 	 */
277 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
278 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
279 		goto freeit;
280 
281 	/*
282 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
283 	 * don't do it.
284 	 */
285 	nxt = -1;
286 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
287 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
288 		struct icmp6_hdr *icp;
289 
290 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
291 			sizeof(*icp));
292 		if (icp == NULL) {
293 			icmp6stat_inc(icp6s_tooshort);
294 			return (NULL);
295 		}
296 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
297 		    icp->icmp6_type == ND_REDIRECT) {
298 			/*
299 			 * ICMPv6 error
300 			 * Special case: for redirect (which is
301 			 * informational) we must not send icmp6 error.
302 			 */
303 			icmp6stat_inc(icp6s_canterror);
304 			goto freeit;
305 		} else {
306 			/* ICMPv6 informational - send the error */
307 		}
308 	}
309 	else {
310 		/* non-ICMPv6 - send the error */
311 	}
312 
313 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
314 
315 	/* Finally, do rate limitation check. */
316 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
317 		icmp6stat_inc(icp6s_toofreq);
318 		goto freeit;
319 	}
320 
321 	/*
322 	 * OK, ICMP6 can be generated.
323 	 */
324 
325 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
326 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
327 
328 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
329 	M_PREPEND(m, preplen, M_DONTWAIT);
330 	if (m && m->m_len < preplen)
331 		m = m_pullup(m, preplen);
332 	if (m == NULL) {
333 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
334 		return (NULL);
335 	}
336 
337 	nip6 = mtod(m, struct ip6_hdr *);
338 	nip6->ip6_src  = oip6->ip6_src;
339 	nip6->ip6_dst  = oip6->ip6_dst;
340 
341 	if (IN6_IS_SCOPE_EMBED(&oip6->ip6_src))
342 		oip6->ip6_src.s6_addr16[1] = 0;
343 	if (IN6_IS_SCOPE_EMBED(&oip6->ip6_dst))
344 		oip6->ip6_dst.s6_addr16[1] = 0;
345 
346 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
347 	icmp6->icmp6_type = type;
348 	icmp6->icmp6_code = code;
349 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
350 
351 	/*
352 	 * icmp6_reflect() is designed to be in the input path.
353 	 * icmp6_error() can be called from both input and outut path,
354 	 * and if we are in output path rcvif could contain bogus value.
355 	 * clear m->m_pkthdr.ph_ifidx for safety, we should have enough
356 	 * scope information in ip header (nip6).
357 	 */
358 	m->m_pkthdr.ph_ifidx = 0;
359 
360 	icmp6stat_inc(icp6s_outhist + type);
361 
362 	return (m);
363 
364   freeit:
365 	/*
366 	 * If we can't tell wheter or not we can generate ICMP6, free it.
367 	 */
368 	return (m_freem(m));
369 }
370 
371 /*
372  * Generate an error packet of type error in response to bad IP6 packet.
373  */
374 void
375 icmp6_error(struct mbuf *m, int type, int code, int param)
376 {
377 	struct mbuf	*n;
378 
379 	n = icmp6_do_error(m, type, code, param);
380 	if (n != NULL) {
381 		/* header order: IPv6 - ICMPv6 */
382 		if (!icmp6_reflect(&n, sizeof(struct ip6_hdr), NULL))
383 			ip6_send(n);
384 	}
385 }
386 
387 /*
388  * Process a received ICMP6 message.
389  */
390 int
391 icmp6_input(struct mbuf **mp, int *offp, int proto, int af)
392 {
393 #if NCARP > 0
394 	struct ifnet *ifp;
395 #endif
396 	struct mbuf *m = *mp, *n;
397 	struct ip6_hdr *ip6, *nip6;
398 	struct icmp6_hdr *icmp6, *nicmp6;
399 	int off = *offp;
400 	int icmp6len = m->m_pkthdr.len - *offp;
401 	int code, sum, noff;
402 	char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
403 
404 	/*
405 	 * Locate icmp6 structure in mbuf, and check
406 	 * that not corrupted and of at least minimum length
407 	 */
408 
409 	ip6 = mtod(m, struct ip6_hdr *);
410 	if (icmp6len < sizeof(struct icmp6_hdr)) {
411 		icmp6stat_inc(icp6s_tooshort);
412 		goto freeit;
413 	}
414 
415 	/*
416 	 * calculate the checksum
417 	 */
418 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
419 	if (icmp6 == NULL) {
420 		icmp6stat_inc(icp6s_tooshort);
421 		return IPPROTO_DONE;
422 	}
423 	code = icmp6->icmp6_code;
424 
425 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
426 		nd6log((LOG_ERR,
427 		    "ICMP6 checksum error(%d|%x) %s\n",
428 		    icmp6->icmp6_type, sum,
429 		    inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src))));
430 		icmp6stat_inc(icp6s_checksum);
431 		goto freeit;
432 	}
433 
434 #if NPF > 0
435 	if (m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) {
436 		switch (icmp6->icmp6_type) {
437 		/*
438 		 * These ICMP6 types map to other connections.  They must be
439 		 * delivered to pr_ctlinput() also for diverted connections.
440 		 */
441 		case ICMP6_DST_UNREACH:
442 		case ICMP6_PACKET_TOO_BIG:
443 		case ICMP6_TIME_EXCEEDED:
444 		case ICMP6_PARAM_PROB:
445 			/*
446 			 * Do not use the divert-to property of the TCP or UDP
447 			 * rule when doing the PCB lookup for the raw socket.
448 			 */
449 			m->m_pkthdr.pf.flags &=~ PF_TAG_DIVERTED;
450 			break;
451 		default:
452 			goto raw;
453 		}
454 	}
455 #endif /* NPF */
456 
457 #if NCARP > 0
458 	ifp = if_get(m->m_pkthdr.ph_ifidx);
459 	if (ifp == NULL)
460 		goto freeit;
461 
462 	if (icmp6->icmp6_type == ICMP6_ECHO_REQUEST &&
463 	    carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
464 	    ip6->ip6_dst.s6_addr32, 1)) {
465 		if_put(ifp);
466 		goto freeit;
467 	}
468 
469 	if_put(ifp);
470 #endif
471 	icmp6stat_inc(icp6s_inhist + icmp6->icmp6_type);
472 
473 	switch (icmp6->icmp6_type) {
474 	case ICMP6_DST_UNREACH:
475 		switch (code) {
476 		case ICMP6_DST_UNREACH_NOROUTE:
477 			code = PRC_UNREACH_NET;
478 			break;
479 		case ICMP6_DST_UNREACH_ADMIN:
480 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
481 			break;
482 		case ICMP6_DST_UNREACH_ADDR:
483 			code = PRC_HOSTDEAD;
484 			break;
485 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
486 			/* I mean "source address was incorrect." */
487 			code = PRC_PARAMPROB;
488 			break;
489 		case ICMP6_DST_UNREACH_NOPORT:
490 			code = PRC_UNREACH_PORT;
491 			break;
492 		default:
493 			goto badcode;
494 		}
495 		goto deliver;
496 
497 	case ICMP6_PACKET_TOO_BIG:
498 		/* MTU is checked in icmp6_mtudisc_update. */
499 		code = PRC_MSGSIZE;
500 
501 		/*
502 		 * Updating the path MTU will be done after examining
503 		 * intermediate extension headers.
504 		 */
505 		goto deliver;
506 
507 	case ICMP6_TIME_EXCEEDED:
508 		switch (code) {
509 		case ICMP6_TIME_EXCEED_TRANSIT:
510 			code = PRC_TIMXCEED_INTRANS;
511 			break;
512 		case ICMP6_TIME_EXCEED_REASSEMBLY:
513 			code = PRC_TIMXCEED_REASS;
514 			break;
515 		default:
516 			goto badcode;
517 		}
518 		goto deliver;
519 
520 	case ICMP6_PARAM_PROB:
521 		switch (code) {
522 		case ICMP6_PARAMPROB_NEXTHEADER:
523 			code = PRC_UNREACH_PROTOCOL;
524 			break;
525 		case ICMP6_PARAMPROB_HEADER:
526 		case ICMP6_PARAMPROB_OPTION:
527 			code = PRC_PARAMPROB;
528 			break;
529 		default:
530 			goto badcode;
531 		}
532 		goto deliver;
533 
534 	case ICMP6_ECHO_REQUEST:
535 		if (code != 0)
536 			goto badcode;
537 		/*
538 		 * Copy mbuf to send to two data paths: userland socket(s),
539 		 * and to the querier (echo reply).
540 		 * m: a copy for socket, n: a copy for querier
541 		 */
542 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
543 			/* Give up local */
544 			n = m;
545 			m = *mp = NULL;
546 			goto deliverecho;
547 		}
548 		/*
549 		 * If the first mbuf is shared, or the first mbuf is too short,
550 		 * copy the first part of the data into a fresh mbuf.
551 		 * Otherwise, we will wrongly overwrite both copies.
552 		 */
553 		if ((n->m_flags & M_EXT) != 0 ||
554 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
555 			struct mbuf *n0 = n;
556 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
557 
558 			/*
559 			 * Prepare an internal mbuf.  m_pullup() doesn't
560 			 * always copy the length we specified.
561 			 */
562 			if (maxlen >= MCLBYTES) {
563 				/* Give up remote */
564 				m_freem(n0);
565 				break;
566 			}
567 			MGETHDR(n, M_DONTWAIT, n0->m_type);
568 			if (n && maxlen >= MHLEN) {
569 				MCLGET(n, M_DONTWAIT);
570 				if ((n->m_flags & M_EXT) == 0) {
571 					m_free(n);
572 					n = NULL;
573 				}
574 			}
575 			if (n == NULL) {
576 				/* Give up local */
577 				m_freem(n0);
578 				n = m;
579 				m = *mp = NULL;
580 				goto deliverecho;
581 			}
582 			M_MOVE_PKTHDR(n, n0);
583 			/*
584 			 * Copy IPv6 and ICMPv6 only.
585 			 */
586 			nip6 = mtod(n, struct ip6_hdr *);
587 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
588 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
589 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
590 			noff = sizeof(struct ip6_hdr);
591 			n->m_len = noff + sizeof(struct icmp6_hdr);
592 			/*
593 			 * Adjust mbuf.  ip6_plen will be adjusted in
594 			 * ip6_output().
595 			 * n->m_pkthdr.len == n0->m_pkthdr.len at this point.
596 			 */
597 			n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr);
598 			n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr));
599 			m_adj(n0, off + sizeof(struct icmp6_hdr));
600 			n->m_next = n0;
601 		} else {
602 	 deliverecho:
603 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
604 			    sizeof(*nicmp6));
605 			noff = off;
606 		}
607 		if (n) {
608 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
609 			nicmp6->icmp6_code = 0;
610 			icmp6stat_inc(icp6s_reflect);
611 			icmp6stat_inc(icp6s_outhist + ICMP6_ECHO_REPLY);
612 			if (!icmp6_reflect(&n, noff, NULL))
613 				ip6_send(n);
614 		}
615 		if (!m)
616 			goto freeit;
617 		break;
618 
619 	case ICMP6_ECHO_REPLY:
620 		if (code != 0)
621 			goto badcode;
622 		break;
623 
624 	case MLD_LISTENER_QUERY:
625 	case MLD_LISTENER_REPORT:
626 		if (icmp6len < sizeof(struct mld_hdr))
627 			goto badlen;
628 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
629 			/* give up local */
630 			mld6_input(m, off);
631 			m = NULL;
632 			goto freeit;
633 		}
634 		mld6_input(n, off);
635 		/* m stays. */
636 		break;
637 
638 	case MLD_LISTENER_DONE:
639 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
640 			goto badlen;
641 		break;		/* nothing to be done in kernel */
642 
643 	case MLD_MTRACE_RESP:
644 	case MLD_MTRACE:
645 		/* XXX: these two are experimental.  not officially defined. */
646 		/* XXX: per-interface statistics? */
647 		break;		/* just pass it to applications */
648 
649 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
650 		/* IPv6 Node Information Queries are not supported */
651 		break;
652 	case ICMP6_WRUREPLY:
653 		break;
654 
655 	case ND_ROUTER_SOLICIT:
656 	case ND_ROUTER_ADVERT:
657 		if (code != 0)
658 			goto badcode;
659 		if ((icmp6->icmp6_type == ND_ROUTER_SOLICIT && icmp6len <
660 		    sizeof(struct nd_router_solicit)) ||
661 		    (icmp6->icmp6_type == ND_ROUTER_ADVERT && icmp6len <
662 		    sizeof(struct nd_router_advert)))
663 			goto badlen;
664 
665 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
666 			/* give up local */
667 			nd6_rtr_cache(m, off, icmp6len,
668 			    icmp6->icmp6_type);
669 			m = NULL;
670 			goto freeit;
671 		}
672 		nd6_rtr_cache(n, off, icmp6len, icmp6->icmp6_type);
673 		/* m stays. */
674 		break;
675 
676 	case ND_NEIGHBOR_SOLICIT:
677 		if (code != 0)
678 			goto badcode;
679 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
680 			goto badlen;
681 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
682 			/* give up local */
683 			nd6_ns_input(m, off, icmp6len);
684 			m = NULL;
685 			goto freeit;
686 		}
687 		nd6_ns_input(n, off, icmp6len);
688 		/* m stays. */
689 		break;
690 
691 	case ND_NEIGHBOR_ADVERT:
692 		if (code != 0)
693 			goto badcode;
694 		if (icmp6len < sizeof(struct nd_neighbor_advert))
695 			goto badlen;
696 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
697 			/* give up local */
698 			nd6_na_input(m, off, icmp6len);
699 			m = NULL;
700 			goto freeit;
701 		}
702 		nd6_na_input(n, off, icmp6len);
703 		/* m stays. */
704 		break;
705 
706 	case ND_REDIRECT:
707 		if (code != 0)
708 			goto badcode;
709 		if (icmp6len < sizeof(struct nd_redirect))
710 			goto badlen;
711 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
712 			/* give up local */
713 			icmp6_redirect_input(m, off);
714 			m = NULL;
715 			goto freeit;
716 		}
717 		icmp6_redirect_input(n, off);
718 		/* m stays. */
719 		break;
720 
721 	case ICMP6_ROUTER_RENUMBERING:
722 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
723 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
724 			goto badcode;
725 		if (icmp6len < sizeof(struct icmp6_router_renum))
726 			goto badlen;
727 		break;
728 
729 	default:
730 		nd6log((LOG_DEBUG,
731 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%u)\n",
732 		    icmp6->icmp6_type,
733 		    inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src)),
734 		    inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst)),
735 		    m->m_pkthdr.ph_ifidx));
736 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
737 			/* ICMPv6 error: MUST deliver it by spec... */
738 			code = PRC_NCMDS;
739 			/* deliver */
740 		} else {
741 			/* ICMPv6 informational: MUST not deliver */
742 			break;
743 		}
744 deliver:
745 		if (icmp6_notify_error(m, off, icmp6len, code)) {
746 			/* In this case, m should've been freed. */
747 			return (IPPROTO_DONE);
748 		}
749 		break;
750 
751 badcode:
752 		icmp6stat_inc(icp6s_badcode);
753 		break;
754 
755 badlen:
756 		icmp6stat_inc(icp6s_badlen);
757 		break;
758 	}
759 
760 #if NPF > 0
761 raw:
762 #endif
763 	/* deliver the packet to appropriate sockets */
764 	return rip6_input(mp, offp, proto, af);
765 
766  freeit:
767 	m_freem(m);
768 	return IPPROTO_DONE;
769 }
770 
771 int
772 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
773 {
774 	struct icmp6_hdr *icmp6;
775 	struct ip6_hdr *eip6;
776 	u_int32_t notifymtu;
777 	struct sockaddr_in6 icmp6src, icmp6dst;
778 
779 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
780 		icmp6stat_inc(icp6s_tooshort);
781 		goto freeit;
782 	}
783 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
784 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
785 	if (icmp6 == NULL) {
786 		icmp6stat_inc(icp6s_tooshort);
787 		return (-1);
788 	}
789 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
790 
791 	/* Detect the upper level protocol */
792 	{
793 		void (*ctlfunc)(int, struct sockaddr *, u_int, void *);
794 		u_int8_t nxt = eip6->ip6_nxt;
795 		int eoff = off + sizeof(struct icmp6_hdr) +
796 			sizeof(struct ip6_hdr);
797 		struct ip6ctlparam ip6cp;
798 		struct in6_addr *finaldst = NULL;
799 		int icmp6type = icmp6->icmp6_type;
800 		struct ip6_frag *fh;
801 		struct ip6_rthdr *rth;
802 		struct ip6_rthdr0 *rth0;
803 		int rthlen;
804 
805 		while (1) { /* XXX: should avoid infinite loop explicitly? */
806 			struct ip6_ext *eh;
807 
808 			switch (nxt) {
809 			case IPPROTO_HOPOPTS:
810 			case IPPROTO_DSTOPTS:
811 			case IPPROTO_AH:
812 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
813 					       eoff, sizeof(*eh));
814 				if (eh == NULL) {
815 					icmp6stat_inc(icp6s_tooshort);
816 					return (-1);
817 				}
818 
819 				if (nxt == IPPROTO_AH)
820 					eoff += (eh->ip6e_len + 2) << 2;
821 				else
822 					eoff += (eh->ip6e_len + 1) << 3;
823 				nxt = eh->ip6e_nxt;
824 				break;
825 			case IPPROTO_ROUTING:
826 				/*
827 				 * When the erroneous packet contains a
828 				 * routing header, we should examine the
829 				 * header to determine the final destination.
830 				 * Otherwise, we can't properly update
831 				 * information that depends on the final
832 				 * destination (e.g. path MTU).
833 				 */
834 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
835 					       eoff, sizeof(*rth));
836 				if (rth == NULL) {
837 					icmp6stat_inc(icp6s_tooshort);
838 					return (-1);
839 				}
840 				rthlen = (rth->ip6r_len + 1) << 3;
841 				/*
842 				 * XXX: currently there is no
843 				 * officially defined type other
844 				 * than type-0.
845 				 * Note that if the segment left field
846 				 * is 0, all intermediate hops must
847 				 * have been passed.
848 				 */
849 				if (rth->ip6r_segleft &&
850 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
851 					int hops;
852 
853 					IP6_EXTHDR_GET(rth0,
854 						       struct ip6_rthdr0 *, m,
855 						       eoff, rthlen);
856 					if (rth0 == NULL) {
857 						icmp6stat_inc(icp6s_tooshort);
858 						return (-1);
859 					}
860 					/* just ignore a bogus header */
861 					if ((rth0->ip6r0_len % 2) == 0 &&
862 					    (hops = rth0->ip6r0_len/2))
863 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
864 				}
865 				eoff += rthlen;
866 				nxt = rth->ip6r_nxt;
867 				break;
868 			case IPPROTO_FRAGMENT:
869 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
870 					       eoff, sizeof(*fh));
871 				if (fh == NULL) {
872 					icmp6stat_inc(icp6s_tooshort);
873 					return (-1);
874 				}
875 				/*
876 				 * Data after a fragment header is meaningless
877 				 * unless it is the first fragment, but
878 				 * we'll go to the notify label for path MTU
879 				 * discovery.
880 				 */
881 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
882 					goto notify;
883 
884 				eoff += sizeof(struct ip6_frag);
885 				nxt = fh->ip6f_nxt;
886 				break;
887 			default:
888 				/*
889 				 * This case includes ESP and the No Next
890 				 * Header.  In such cases going to the notify
891 				 * label does not have any meaning
892 				 * (i.e. ctlfunc will be NULL), but we go
893 				 * anyway since we might have to update
894 				 * path MTU information.
895 				 */
896 				goto notify;
897 			}
898 		}
899 	  notify:
900 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
901 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
902 		if (icmp6 == NULL) {
903 			icmp6stat_inc(icp6s_tooshort);
904 			return (-1);
905 		}
906 
907 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
908 		bzero(&icmp6dst, sizeof(icmp6dst));
909 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
910 		icmp6dst.sin6_family = AF_INET6;
911 		if (finaldst == NULL)
912 			icmp6dst.sin6_addr = eip6->ip6_dst;
913 		else
914 			icmp6dst.sin6_addr = *finaldst;
915 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.ph_ifidx,
916 		    &icmp6dst.sin6_addr);
917 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst, NULL)) {
918 			/* should be impossbile */
919 			nd6log((LOG_DEBUG,
920 			    "icmp6_notify_error: in6_embedscope failed\n"));
921 			goto freeit;
922 		}
923 
924 		/*
925 		 * retrieve parameters from the inner IPv6 header, and convert
926 		 * them into sockaddr structures.
927 		 */
928 		bzero(&icmp6src, sizeof(icmp6src));
929 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
930 		icmp6src.sin6_family = AF_INET6;
931 		icmp6src.sin6_addr = eip6->ip6_src;
932 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.ph_ifidx,
933 		    &icmp6src.sin6_addr);
934 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src, NULL)) {
935 			/* should be impossbile */
936 			nd6log((LOG_DEBUG,
937 			    "icmp6_notify_error: in6_embedscope failed\n"));
938 			goto freeit;
939 		}
940 		icmp6src.sin6_flowinfo =
941 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
942 
943 		if (finaldst == NULL)
944 			finaldst = &eip6->ip6_dst;
945 		ip6cp.ip6c_m = m;
946 		ip6cp.ip6c_icmp6 = icmp6;
947 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
948 		ip6cp.ip6c_off = eoff;
949 		ip6cp.ip6c_finaldst = finaldst;
950 		ip6cp.ip6c_src = &icmp6src;
951 		ip6cp.ip6c_nxt = nxt;
952 #if NPF > 0
953 		pf_pkt_addr_changed(m);
954 #endif
955 
956 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
957 			notifymtu = ntohl(icmp6->icmp6_mtu);
958 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
959 		}
960 
961 		ctlfunc = inet6sw[ip6_protox[nxt]].pr_ctlinput;
962 		if (ctlfunc)
963 			(*ctlfunc)(code, sin6tosa(&icmp6dst),
964 			    m->m_pkthdr.ph_rtableid, &ip6cp);
965 	}
966 	return (0);
967 
968   freeit:
969 	m_freem(m);
970 	return (-1);
971 }
972 
973 void
974 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
975 {
976 	unsigned long rtcount;
977 	struct icmp6_mtudisc_callback *mc;
978 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
979 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
980 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
981 	u_int mtu = ntohl(icmp6->icmp6_mtu);
982 	struct rtentry *rt = NULL;
983 	struct sockaddr_in6 sin6;
984 
985 	if (mtu < IPV6_MMTU)
986 		return;
987 
988 	/*
989 	 * allow non-validated cases if memory is plenty, to make traffic
990 	 * from non-connected pcb happy.
991 	 */
992 	rtcount = rt_timer_queue_count(icmp6_mtudisc_timeout_q);
993 	if (validated) {
994 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
995 			return;
996 		else if (0 <= icmp6_mtudisc_lowat &&
997 		    rtcount > icmp6_mtudisc_lowat) {
998 			/*
999 			 * XXX nuke a victim, install the new one.
1000 			 */
1001 		}
1002 	} else {
1003 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1004 			return;
1005 	}
1006 
1007 	bzero(&sin6, sizeof(sin6));
1008 	sin6.sin6_family = PF_INET6;
1009 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1010 	sin6.sin6_addr = *dst;
1011 	/* XXX normally, this won't happen */
1012 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1013 		sin6.sin6_addr.s6_addr16[1] = htons(m->m_pkthdr.ph_ifidx);
1014 	}
1015 	sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.ph_ifidx,
1016 	    &sin6.sin6_addr);
1017 
1018 	rt = icmp6_mtudisc_clone(sin6tosa(&sin6), m->m_pkthdr.ph_rtableid);
1019 
1020 	if (rt != NULL && ISSET(rt->rt_flags, RTF_HOST) &&
1021 	    !(rt->rt_locks & RTV_MTU) &&
1022 	    (rt->rt_mtu > mtu || rt->rt_mtu == 0)) {
1023 		struct ifnet *ifp;
1024 
1025 		ifp = if_get(rt->rt_ifidx);
1026 		if (ifp != NULL && mtu < ifp->if_mtu) {
1027 			icmp6stat_inc(icp6s_pmtuchg);
1028 			rt->rt_mtu = mtu;
1029 		}
1030 		if_put(ifp);
1031 	}
1032 	rtfree(rt);
1033 
1034 	/*
1035 	 * Notify protocols that the MTU for this destination
1036 	 * has changed.
1037 	 */
1038 	LIST_FOREACH(mc, &icmp6_mtudisc_callbacks, mc_list)
1039 		(*mc->mc_func)(&sin6, m->m_pkthdr.ph_rtableid);
1040 }
1041 
1042 /*
1043  * Reflect the ip6 packet back to the source.
1044  * OFF points to the icmp6 header, counted from the top of the mbuf.
1045  */
1046 int
1047 icmp6_reflect(struct mbuf **mp, size_t off, struct sockaddr *sa)
1048 {
1049 	struct mbuf *m = *mp;
1050 	struct rtentry *rt = NULL;
1051 	struct ip6_hdr *ip6;
1052 	struct icmp6_hdr *icmp6;
1053 	struct in6_addr t, *src = NULL;
1054 	struct sockaddr_in6 sa6_src, sa6_dst;
1055 	u_int rtableid;
1056 
1057 	CTASSERT(sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) <= MHLEN);
1058 
1059 	/* too short to reflect */
1060 	if (off < sizeof(struct ip6_hdr)) {
1061 		nd6log((LOG_DEBUG,
1062 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1063 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1064 		    __FILE__, __LINE__));
1065 		goto bad;
1066 	}
1067 
1068 	if (m->m_pkthdr.ph_loopcnt++ >= M_MAXLOOP) {
1069 		m_freemp(mp);
1070 		return (ELOOP);
1071 	}
1072 	rtableid = m->m_pkthdr.ph_rtableid;
1073 	m_resethdr(m);
1074 	m->m_pkthdr.ph_rtableid = rtableid;
1075 
1076 	/*
1077 	 * If there are extra headers between IPv6 and ICMPv6, strip
1078 	 * off that header first.
1079 	 */
1080 	if (off > sizeof(struct ip6_hdr)) {
1081 		size_t l;
1082 		struct ip6_hdr nip6;
1083 
1084 		l = off - sizeof(struct ip6_hdr);
1085 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1086 		m_adj(m, l);
1087 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1088 		if (m->m_len < l) {
1089 			if ((m = *mp = m_pullup(m, l)) == NULL)
1090 				return (EMSGSIZE);
1091 		}
1092 		memcpy(mtod(m, caddr_t), &nip6, sizeof(nip6));
1093 	} else /* off == sizeof(struct ip6_hdr) */ {
1094 		size_t l;
1095 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1096 		if (m->m_len < l) {
1097 			if ((m = *mp = m_pullup(m, l)) == NULL)
1098 				return (EMSGSIZE);
1099 		}
1100 	}
1101 	ip6 = mtod(m, struct ip6_hdr *);
1102 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1103 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1104 
1105 	t = ip6->ip6_dst;
1106 	/*
1107 	 * ip6_input() drops a packet if its src is multicast.
1108 	 * So, the src is never multicast.
1109 	 */
1110 	ip6->ip6_dst = ip6->ip6_src;
1111 
1112 	/*
1113 	 * XXX: make sure to embed scope zone information, using
1114 	 * already embedded IDs or the received interface (if any).
1115 	 * Note that rcvif may be NULL.
1116 	 * TODO: scoped routing case (XXX).
1117 	 */
1118 	bzero(&sa6_src, sizeof(sa6_src));
1119 	sa6_src.sin6_family = AF_INET6;
1120 	sa6_src.sin6_len = sizeof(sa6_src);
1121 	sa6_src.sin6_addr = ip6->ip6_dst;
1122 	bzero(&sa6_dst, sizeof(sa6_dst));
1123 	sa6_dst.sin6_family = AF_INET6;
1124 	sa6_dst.sin6_len = sizeof(sa6_dst);
1125 	sa6_dst.sin6_addr = t;
1126 
1127 	if (sa == NULL) {
1128 		/*
1129 		 * If the incoming packet was addressed directly to us (i.e.
1130 		 * unicast), use dst as the src for the reply. The
1131 		 * IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED case would be VERY rare,
1132 		 * but is possible (for example) when we encounter an error
1133 		 * while forwarding procedure destined to a duplicated address
1134 		 * of ours.
1135 		 */
1136 		rt = rtalloc(sin6tosa(&sa6_dst), 0, rtableid);
1137 		if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL) &&
1138 		    !ISSET(ifatoia6(rt->rt_ifa)->ia6_flags,
1139 		    IN6_IFF_ANYCAST|IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED)) {
1140 			src = &t;
1141 		}
1142 		rtfree(rt);
1143 		rt = NULL;
1144 		sa = sin6tosa(&sa6_src);
1145 	}
1146 
1147 	if (src == NULL) {
1148 		struct in6_ifaddr *ia6;
1149 
1150 		/*
1151 		 * This case matches to multicasts, our anycast, or unicasts
1152 		 * that we do not own.  Select a source address based on the
1153 		 * source address of the erroneous packet.
1154 		 */
1155 		rt = rtalloc(sa, RT_RESOLVE, rtableid);
1156 		if (!rtisvalid(rt)) {
1157 			char addr[INET6_ADDRSTRLEN];
1158 
1159 			nd6log((LOG_DEBUG,
1160 			    "%s: source can't be determined: dst=%s\n",
1161 			    __func__, inet_ntop(AF_INET6, &sa6_src.sin6_addr,
1162 			    addr, sizeof(addr))));
1163 			rtfree(rt);
1164 			goto bad;
1165 		}
1166 		ia6 = in6_ifawithscope(rt->rt_ifa->ifa_ifp, &t, rtableid);
1167 		if (ia6 != NULL)
1168 			src = &ia6->ia_addr.sin6_addr;
1169 		if (src == NULL)
1170 			src = &ifatoia6(rt->rt_ifa)->ia_addr.sin6_addr;
1171 	}
1172 
1173 	ip6->ip6_src = *src;
1174 	rtfree(rt);
1175 
1176 	ip6->ip6_flow = 0;
1177 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1178 	ip6->ip6_vfc |= IPV6_VERSION;
1179 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1180 	ip6->ip6_hlim = ip6_defhlim;
1181 
1182 	icmp6->icmp6_cksum = 0;
1183 	m->m_pkthdr.csum_flags = M_ICMP_CSUM_OUT;
1184 
1185 	/*
1186 	 * XXX option handling
1187 	 */
1188 
1189 	m->m_flags &= ~(M_BCAST|M_MCAST);
1190 	return (0);
1191 
1192  bad:
1193 	m_freemp(mp);
1194 	return (EHOSTUNREACH);
1195 }
1196 
1197 void
1198 icmp6_fasttimo(void)
1199 {
1200 
1201 	mld6_fasttimeo();
1202 }
1203 
1204 const char *
1205 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
1206     struct in6_addr *tgt6)
1207 {
1208 	static char buf[1024]; /* XXX */
1209 	char src[INET6_ADDRSTRLEN];
1210 	char dst[INET6_ADDRSTRLEN];
1211 	char tgt[INET6_ADDRSTRLEN];
1212 
1213 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
1214 		 inet_ntop(AF_INET6, src6, src, sizeof(src)),
1215 		 inet_ntop(AF_INET6, dst6, dst, sizeof(dst)),
1216 		 inet_ntop(AF_INET6, tgt6, tgt, sizeof(tgt)));
1217 	return buf;
1218 }
1219 
1220 void
1221 icmp6_redirect_input(struct mbuf *m, int off)
1222 {
1223 	struct ifnet *ifp;
1224 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1225 	struct nd_redirect *nd_rd;
1226 	int icmp6len = ntohs(ip6->ip6_plen);
1227 	char *lladdr = NULL;
1228 	int lladdrlen = 0;
1229 	struct rtentry *rt = NULL;
1230 	int is_router;
1231 	int is_onlink;
1232 	struct in6_addr src6 = ip6->ip6_src;
1233 	struct in6_addr redtgt6;
1234 	struct in6_addr reddst6;
1235 	union nd_opts ndopts;
1236 	char addr[INET6_ADDRSTRLEN];
1237 
1238 	ifp = if_get(m->m_pkthdr.ph_ifidx);
1239 	if (ifp == NULL)
1240 		return;
1241 
1242 	/* XXX if we are router, we don't update route by icmp6 redirect */
1243 	if (ip6_forwarding)
1244 		goto freeit;
1245 	if (!(ifp->if_xflags & IFXF_AUTOCONF6))
1246 		goto freeit;
1247 
1248 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
1249 	if (nd_rd == NULL) {
1250 		icmp6stat_inc(icp6s_tooshort);
1251 		if_put(ifp);
1252 		return;
1253 	}
1254 	redtgt6 = nd_rd->nd_rd_target;
1255 	reddst6 = nd_rd->nd_rd_dst;
1256 
1257 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
1258 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
1259 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
1260 		reddst6.s6_addr16[1] = htons(ifp->if_index);
1261 
1262 	/* validation */
1263 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
1264 		nd6log((LOG_ERR,
1265 			"ICMP6 redirect sent from %s rejected; "
1266 			"must be from linklocal\n",
1267 			inet_ntop(AF_INET6, &src6, addr, sizeof(addr))));
1268 		goto bad;
1269 	}
1270 	if (ip6->ip6_hlim != 255) {
1271 		nd6log((LOG_ERR,
1272 			"ICMP6 redirect sent from %s rejected; "
1273 			"hlim=%d (must be 255)\n",
1274 			inet_ntop(AF_INET6, &src6, addr, sizeof(addr)),
1275 			ip6->ip6_hlim));
1276 		goto bad;
1277 	}
1278 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
1279 		nd6log((LOG_ERR,
1280 			"ICMP6 redirect rejected; "
1281 			"redirect dst must be unicast: %s\n",
1282 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1283 		goto bad;
1284 	}
1285     {
1286 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
1287 	struct sockaddr_in6 sin6;
1288 	struct in6_addr *gw6;
1289 
1290 	bzero(&sin6, sizeof(sin6));
1291 	sin6.sin6_family = AF_INET6;
1292 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1293 	memcpy(&sin6.sin6_addr, &reddst6, sizeof(reddst6));
1294 	rt = rtalloc(sin6tosa(&sin6), 0, m->m_pkthdr.ph_rtableid);
1295 	if (rt) {
1296 		if (rt->rt_gateway == NULL ||
1297 		    rt->rt_gateway->sa_family != AF_INET6) {
1298 			nd6log((LOG_ERR,
1299 			    "ICMP6 redirect rejected; no route "
1300 			    "with inet6 gateway found for redirect dst: %s\n",
1301 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1302 			rtfree(rt);
1303 			goto bad;
1304 		}
1305 
1306 		gw6 = &(satosin6(rt->rt_gateway)->sin6_addr);
1307 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
1308 			nd6log((LOG_ERR,
1309 				"ICMP6 redirect rejected; "
1310 				"not equal to gw-for-src=%s (must be same): "
1311 				"%s\n",
1312 				inet_ntop(AF_INET6, gw6, addr, sizeof(addr)),
1313 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1314 			rtfree(rt);
1315 			goto bad;
1316 		}
1317 	} else {
1318 		nd6log((LOG_ERR,
1319 			"ICMP6 redirect rejected; "
1320 			"no route found for redirect dst: %s\n",
1321 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1322 		goto bad;
1323 	}
1324 	rtfree(rt);
1325 	rt = NULL;
1326     }
1327 
1328 	is_router = is_onlink = 0;
1329 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
1330 		is_router = 1;	/* router case */
1331 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
1332 		is_onlink = 1;	/* on-link destination case */
1333 	if (!is_router && !is_onlink) {
1334 		nd6log((LOG_ERR,
1335 			"ICMP6 redirect rejected; "
1336 			"neither router case nor onlink case: %s\n",
1337 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1338 		goto bad;
1339 	}
1340 	/* validation passed */
1341 
1342 	icmp6len -= sizeof(*nd_rd);
1343 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
1344 	if (nd6_options(&ndopts) < 0) {
1345 		nd6log((LOG_INFO, "icmp6_redirect_input: "
1346 			"invalid ND option, rejected: %s\n",
1347 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1348 		/* nd6_options have incremented stats */
1349 		goto freeit;
1350 	}
1351 
1352 	if (ndopts.nd_opts_tgt_lladdr) {
1353 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
1354 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
1355 	}
1356 
1357 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
1358 		nd6log((LOG_INFO,
1359 			"icmp6_redirect_input: lladdrlen mismatch for %s "
1360 			"(if %d, icmp6 packet %d): %s\n",
1361 			inet_ntop(AF_INET6, &redtgt6, addr, sizeof(addr)),
1362 			ifp->if_addrlen, lladdrlen - 2,
1363 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1364 		goto bad;
1365 	}
1366 
1367 	/* RFC 2461 8.3 */
1368 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
1369 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
1370 
1371 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
1372 		/* perform rtredirect */
1373 		struct sockaddr_in6 sdst;
1374 		struct sockaddr_in6 sgw;
1375 		struct sockaddr_in6 ssrc;
1376 		unsigned long rtcount;
1377 		struct rtentry *newrt = NULL;
1378 
1379 		/*
1380 		 * do not install redirect route, if the number of entries
1381 		 * is too much (> hiwat).  note that, the node (= host) will
1382 		 * work just fine even if we do not install redirect route
1383 		 * (there will be additional hops, though).
1384 		 */
1385 		rtcount = rt_timer_queue_count(icmp6_redirect_timeout_q);
1386 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes)
1387 			goto freeit;
1388 		else if (0 <= icmp6_redirect_lowat &&
1389 		    rtcount > icmp6_redirect_lowat) {
1390 			/*
1391 			 * XXX nuke a victim, install the new one.
1392 			 */
1393 		}
1394 
1395 		bzero(&sdst, sizeof(sdst));
1396 		bzero(&sgw, sizeof(sgw));
1397 		bzero(&ssrc, sizeof(ssrc));
1398 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
1399 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
1400 			sizeof(struct sockaddr_in6);
1401 		memcpy(&sgw.sin6_addr, &redtgt6, sizeof(struct in6_addr));
1402 		memcpy(&sdst.sin6_addr, &reddst6, sizeof(struct in6_addr));
1403 		memcpy(&ssrc.sin6_addr, &src6, sizeof(struct in6_addr));
1404 		rtredirect(sin6tosa(&sdst), sin6tosa(&sgw), sin6tosa(&ssrc),
1405 		    &newrt, m->m_pkthdr.ph_rtableid);
1406 
1407 		if (newrt) {
1408 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
1409 			    icmp6_redirect_timeout_q, m->m_pkthdr.ph_rtableid);
1410 			rtfree(newrt);
1411 		}
1412 	}
1413 	/* finally update cached route in each socket via pfctlinput */
1414 	{
1415 		struct sockaddr_in6 sdst;
1416 
1417 		bzero(&sdst, sizeof(sdst));
1418 		sdst.sin6_family = AF_INET6;
1419 		sdst.sin6_len = sizeof(struct sockaddr_in6);
1420 		memcpy(&sdst.sin6_addr, &reddst6, sizeof(struct in6_addr));
1421 		pfctlinput(PRC_REDIRECT_HOST, sin6tosa(&sdst));
1422 	}
1423 
1424  freeit:
1425 	if_put(ifp);
1426 	m_freem(m);
1427 	return;
1428 
1429  bad:
1430 	if_put(ifp);
1431 	icmp6stat_inc(icp6s_badredirect);
1432 	m_freem(m);
1433 }
1434 
1435 void
1436 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
1437 {
1438 	struct ifnet *ifp = NULL;
1439 	struct in6_addr *ifp_ll6;
1440 	struct in6_addr *nexthop;
1441 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
1442 	struct mbuf *m = NULL;	/* newly allocated one */
1443 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
1444 	struct nd_redirect *nd_rd;
1445 	size_t maxlen;
1446 	u_char *p;
1447 	struct sockaddr_in6 src_sa;
1448 
1449 	icmp6_errcount(ND_REDIRECT, 0);
1450 
1451 	/* if we are not router, we don't send icmp6 redirect */
1452 	if (!ip6_forwarding)
1453 		goto fail;
1454 
1455 	/* sanity check */
1456 	if (m0 == NULL || !rtisvalid(rt))
1457 		goto fail;
1458 
1459 	ifp = if_get(rt->rt_ifidx);
1460 	if (ifp == NULL)
1461 		goto fail;
1462 
1463 	/*
1464 	 * Address check:
1465 	 *  the source address must identify a neighbor, and
1466 	 *  the destination address must not be a multicast address
1467 	 *  [RFC 2461, sec 8.2]
1468 	 */
1469 	sip6 = mtod(m0, struct ip6_hdr *);
1470 	bzero(&src_sa, sizeof(src_sa));
1471 	src_sa.sin6_family = AF_INET6;
1472 	src_sa.sin6_len = sizeof(src_sa);
1473 	src_sa.sin6_addr = sip6->ip6_src;
1474 	/* we don't currently use sin6_scope_id, but eventually use it */
1475 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp->if_index, &sip6->ip6_src);
1476 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
1477 		goto fail;
1478 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
1479 		goto fail;	/* what should we do here? */
1480 
1481 	/* rate limit */
1482 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
1483 		goto fail;
1484 
1485 	/*
1486 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
1487 	 * we almost always ask for an mbuf cluster for simplicity.
1488 	 * (MHLEN < IPV6_MMTU is almost always true)
1489 	 */
1490 #if IPV6_MMTU >= MCLBYTES
1491 # error assumption failed about IPV6_MMTU and MCLBYTES
1492 #endif
1493 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
1494 	if (m && IPV6_MMTU >= MHLEN)
1495 		MCLGET(m, M_DONTWAIT);
1496 	if (!m)
1497 		goto fail;
1498 	m->m_pkthdr.ph_ifidx = 0;
1499 	m->m_len = 0;
1500 	maxlen = m_trailingspace(m);
1501 	maxlen = min(IPV6_MMTU, maxlen);
1502 	/* just for safety */
1503 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
1504 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
1505 		goto fail;
1506 	}
1507 
1508 	{
1509 		/* get ip6 linklocal address for ifp(my outgoing interface). */
1510 		struct in6_ifaddr *ia6;
1511 		if ((ia6 = in6ifa_ifpforlinklocal(ifp, IN6_IFF_TENTATIVE|
1512 		    IN6_IFF_DUPLICATED|IN6_IFF_ANYCAST)) == NULL)
1513 			goto fail;
1514 		ifp_ll6 = &ia6->ia_addr.sin6_addr;
1515 	}
1516 
1517 	/* get ip6 linklocal address for the router. */
1518 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
1519 		struct sockaddr_in6 *sin6;
1520 		sin6 = satosin6(rt->rt_gateway);
1521 		nexthop = &sin6->sin6_addr;
1522 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
1523 			nexthop = NULL;
1524 	} else
1525 		nexthop = NULL;
1526 
1527 	/* ip6 */
1528 	ip6 = mtod(m, struct ip6_hdr *);
1529 	ip6->ip6_flow = 0;
1530 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1531 	ip6->ip6_vfc |= IPV6_VERSION;
1532 	/* ip6->ip6_plen will be set later */
1533 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1534 	ip6->ip6_hlim = 255;
1535 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
1536 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
1537 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
1538 
1539 	/* ND Redirect */
1540 	nd_rd = (struct nd_redirect *)(ip6 + 1);
1541 	nd_rd->nd_rd_type = ND_REDIRECT;
1542 	nd_rd->nd_rd_code = 0;
1543 	nd_rd->nd_rd_reserved = 0;
1544 	if (rt->rt_flags & RTF_GATEWAY) {
1545 		/*
1546 		 * nd_rd->nd_rd_target must be a link-local address in
1547 		 * better router cases.
1548 		 */
1549 		if (!nexthop)
1550 			goto fail;
1551 		bcopy(nexthop, &nd_rd->nd_rd_target,
1552 		      sizeof(nd_rd->nd_rd_target));
1553 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
1554 		      sizeof(nd_rd->nd_rd_dst));
1555 	} else {
1556 		/* make sure redtgt == reddst */
1557 		nexthop = &sip6->ip6_dst;
1558 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
1559 		      sizeof(nd_rd->nd_rd_target));
1560 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
1561 		      sizeof(nd_rd->nd_rd_dst));
1562 	}
1563 
1564 	p = (u_char *)(nd_rd + 1);
1565 
1566 	{
1567 		/* target lladdr option */
1568 		struct rtentry *nrt;
1569 		int len;
1570 		struct sockaddr_dl *sdl;
1571 		struct nd_opt_hdr *nd_opt;
1572 		char *lladdr;
1573 
1574 		len = sizeof(*nd_opt) + ifp->if_addrlen;
1575 		len = (len + 7) & ~7;	/* round by 8 */
1576 		/* safety check */
1577 		if (len + (p - (u_char *)ip6) > maxlen)
1578 			goto nolladdropt;
1579 		nrt = nd6_lookup(nexthop, 0, ifp, ifp->if_rdomain);
1580 		if ((nrt != NULL) &&
1581 		    (nrt->rt_flags & (RTF_GATEWAY|RTF_LLINFO)) == RTF_LLINFO &&
1582 		    (nrt->rt_gateway->sa_family == AF_LINK) &&
1583 		    (sdl = satosdl(nrt->rt_gateway)) &&
1584 		    sdl->sdl_alen) {
1585 			nd_opt = (struct nd_opt_hdr *)p;
1586 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1587 			nd_opt->nd_opt_len = len >> 3;
1588 			lladdr = (char *)(nd_opt + 1);
1589 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
1590 			p += len;
1591 		}
1592 		rtfree(nrt);
1593 	}
1594   nolladdropt:;
1595 
1596 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
1597 
1598 	/* just to be safe */
1599 	if (p - (u_char *)ip6 > maxlen)
1600 		goto noredhdropt;
1601 
1602 	{
1603 		/* redirected header option */
1604 		int len;
1605 		struct nd_opt_rd_hdr *nd_opt_rh;
1606 
1607 		/*
1608 		 * compute the maximum size for icmp6 redirect header option.
1609 		 * XXX room for auth header?
1610 		 */
1611 		len = maxlen - (p - (u_char *)ip6);
1612 		len &= ~7;
1613 
1614 		/*
1615 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
1616 		 * about padding/truncate rule for the original IP packet.
1617 		 * From the discussion on IPv6imp in Feb 1999,
1618 		 * the consensus was:
1619 		 * - "attach as much as possible" is the goal
1620 		 * - pad if not aligned (original size can be guessed by
1621 		 *   original ip6 header)
1622 		 * Following code adds the padding if it is simple enough,
1623 		 * and truncates if not.
1624 		 */
1625 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
1626 			/* not enough room, truncate */
1627 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
1628 			    m0->m_pkthdr.len);
1629 		} else {
1630 			/*
1631 			 * enough room, truncate if not aligned.
1632 			 * we don't pad here for simplicity.
1633 			 */
1634 			size_t extra;
1635 
1636 			extra = m0->m_pkthdr.len % 8;
1637 			if (extra) {
1638 				/* truncate */
1639 				m_adj(m0, -extra);
1640 			}
1641 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
1642 		}
1643 
1644 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
1645 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
1646 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
1647 		nd_opt_rh->nd_opt_rh_len = len >> 3;
1648 		p += sizeof(*nd_opt_rh);
1649 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
1650 
1651 		/* connect m0 to m */
1652 		m->m_pkthdr.len += m0->m_pkthdr.len;
1653 		m_cat(m, m0);
1654 		m0 = NULL;
1655 	}
1656 noredhdropt:
1657 	m_freem(m0);
1658 	m0 = NULL;
1659 
1660 	sip6 = mtod(m, struct ip6_hdr *);
1661 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
1662 		sip6->ip6_src.s6_addr16[1] = 0;
1663 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
1664 		sip6->ip6_dst.s6_addr16[1] = 0;
1665 #if 0
1666 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1667 		ip6->ip6_src.s6_addr16[1] = 0;
1668 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
1669 		ip6->ip6_dst.s6_addr16[1] = 0;
1670 #endif
1671 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
1672 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
1673 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
1674 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
1675 
1676 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
1677 
1678 	nd_rd->nd_rd_cksum = 0;
1679 	m->m_pkthdr.csum_flags = M_ICMP_CSUM_OUT;
1680 
1681 	/* send the packet to outside... */
1682 	ip6_output(m, NULL, NULL, 0, NULL, NULL);
1683 
1684 	icmp6stat_inc(icp6s_outhist + ND_REDIRECT);
1685 
1686 	if_put(ifp);
1687 	return;
1688 
1689 fail:
1690 	if_put(ifp);
1691 	m_freem(m);
1692 	m_freem(m0);
1693 }
1694 
1695 /*
1696  * ICMPv6 socket option processing.
1697  */
1698 int
1699 icmp6_ctloutput(int op, struct socket *so, int level, int optname,
1700     struct mbuf *m)
1701 {
1702 	int error = 0;
1703 	struct inpcb *in6p = sotoinpcb(so);
1704 
1705 	if (level != IPPROTO_ICMPV6)
1706 		return EINVAL;
1707 
1708 	switch (op) {
1709 	case PRCO_SETOPT:
1710 		switch (optname) {
1711 		case ICMP6_FILTER:
1712 		    {
1713 			struct icmp6_filter *p;
1714 
1715 			if (m == NULL || m->m_len != sizeof(*p)) {
1716 				error = EMSGSIZE;
1717 				break;
1718 			}
1719 			p = mtod(m, struct icmp6_filter *);
1720 			if (!p || !in6p->inp_icmp6filt) {
1721 				error = EINVAL;
1722 				break;
1723 			}
1724 			bcopy(p, in6p->inp_icmp6filt,
1725 				sizeof(struct icmp6_filter));
1726 			error = 0;
1727 			break;
1728 		    }
1729 
1730 		default:
1731 			error = ENOPROTOOPT;
1732 			break;
1733 		}
1734 		break;
1735 
1736 	case PRCO_GETOPT:
1737 		switch (optname) {
1738 		case ICMP6_FILTER:
1739 		    {
1740 			struct icmp6_filter *p;
1741 
1742 			if (!in6p->inp_icmp6filt) {
1743 				error = EINVAL;
1744 				break;
1745 			}
1746 			m->m_len = sizeof(struct icmp6_filter);
1747 			p = mtod(m, struct icmp6_filter *);
1748 			bcopy(in6p->inp_icmp6filt, p,
1749 				sizeof(struct icmp6_filter));
1750 			error = 0;
1751 			break;
1752 		    }
1753 
1754 		default:
1755 			error = ENOPROTOOPT;
1756 			break;
1757 		}
1758 		break;
1759 	}
1760 
1761 	return (error);
1762 }
1763 
1764 /*
1765  * Perform rate limit check.
1766  * Returns 0 if it is okay to send the icmp6 packet.
1767  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
1768  * limitation.
1769  *
1770  * XXX per-destination/type check necessary?
1771  *
1772  * dst - not used at this moment
1773  * type - not used at this moment
1774  * code - not used at this moment
1775  */
1776 int
1777 icmp6_ratelimit(const struct in6_addr *dst, const int type, const int code)
1778 {
1779 	/* PPS limit */
1780 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
1781 	    icmp6errppslim))
1782 		return 1;	/* The packet is subject to rate limit */
1783 	return 0;		/* okay to send */
1784 }
1785 
1786 struct rtentry *
1787 icmp6_mtudisc_clone(struct sockaddr *dst, u_int rtableid)
1788 {
1789 	struct rtentry *rt;
1790 	int    error;
1791 
1792 	rt = rtalloc(dst, RT_RESOLVE, rtableid);
1793 
1794 	/* Check if the route is actually usable */
1795 	if (!rtisvalid(rt) || (rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)))
1796 		goto bad;
1797 
1798 	/*
1799 	 * No PMTU for local routes and permanent neighbors,
1800 	 * ARP and NDP use the same expire timer as the route.
1801 	 */
1802 	if (ISSET(rt->rt_flags, RTF_LOCAL) ||
1803 	    (ISSET(rt->rt_flags, RTF_LLINFO) && rt->rt_expire == 0))
1804 		goto bad;
1805 
1806 	/* If we didn't get a host route, allocate one */
1807 	if ((rt->rt_flags & RTF_HOST) == 0) {
1808 		struct rtentry *nrt;
1809 		struct rt_addrinfo info;
1810 		struct sockaddr_rtlabel sa_rl;
1811 
1812 		memset(&info, 0, sizeof(info));
1813 		info.rti_ifa = rt->rt_ifa;
1814 		info.rti_flags = RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC;
1815 		info.rti_info[RTAX_DST] = dst;
1816 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1817 		info.rti_info[RTAX_LABEL] =
1818 		    rtlabel_id2sa(rt->rt_labelid, &sa_rl);
1819 
1820 		error = rtrequest(RTM_ADD, &info, rt->rt_priority, &nrt,
1821 		    rtableid);
1822 		if (error)
1823 			goto bad;
1824 		nrt->rt_rmx = rt->rt_rmx;
1825 		rtfree(rt);
1826 		rt = nrt;
1827 		rtm_send(rt, RTM_ADD, 0, rtableid);
1828 	}
1829 	error = rt_timer_add(rt, icmp6_mtudisc_timeout, icmp6_mtudisc_timeout_q,
1830 	    rtableid);
1831 	if (error)
1832 		goto bad;
1833 
1834 	return (rt);
1835 bad:
1836 	rtfree(rt);
1837 	return (NULL);
1838 }
1839 
1840 void
1841 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
1842 {
1843 	struct ifnet *ifp;
1844 
1845 	NET_ASSERT_LOCKED();
1846 
1847 	ifp = if_get(rt->rt_ifidx);
1848 	if (ifp == NULL)
1849 		return;
1850 
1851 	if ((rt->rt_flags & (RTF_DYNAMIC|RTF_HOST)) == (RTF_DYNAMIC|RTF_HOST)) {
1852 		rtdeletemsg(rt, ifp, r->rtt_tableid);
1853 	} else {
1854 		if (!(rt->rt_locks & RTV_MTU))
1855 			rt->rt_mtu = 0;
1856 	}
1857 
1858 	if_put(ifp);
1859 }
1860 
1861 void
1862 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
1863 {
1864 	struct ifnet *ifp;
1865 
1866 	NET_ASSERT_LOCKED();
1867 
1868 	ifp = if_get(rt->rt_ifidx);
1869 	if (ifp == NULL)
1870 		return;
1871 
1872 	if ((rt->rt_flags & (RTF_DYNAMIC|RTF_HOST)) == (RTF_DYNAMIC|RTF_HOST)) {
1873 		rtdeletemsg(rt, ifp, r->rtt_tableid);
1874 	}
1875 
1876 	if_put(ifp);
1877 }
1878 
1879 const struct sysctl_bounded_args icmpv6ctl_vars[] = {
1880 	{ ICMPV6CTL_REDIRTIMEOUT, &icmp6_redirtimeout, 0, INT_MAX },
1881 	{ ICMPV6CTL_ND6_DELAY, &nd6_delay, 0, INT_MAX },
1882 	{ ICMPV6CTL_ND6_UMAXTRIES, &nd6_umaxtries, 0, INT_MAX },
1883 	{ ICMPV6CTL_ND6_MMAXTRIES, &nd6_mmaxtries, 0, INT_MAX },
1884 	{ ICMPV6CTL_ERRPPSLIMIT, &icmp6errppslim, -1, 1000 },
1885 	{ ICMPV6CTL_ND6_MAXNUDHINT, &nd6_maxnudhint, 0, INT_MAX },
1886 	{ ICMPV6CTL_MTUDISC_HIWAT, &icmp6_mtudisc_hiwat, -1, INT_MAX },
1887 	{ ICMPV6CTL_MTUDISC_LOWAT, &icmp6_mtudisc_lowat, -1, INT_MAX },
1888 	{ ICMPV6CTL_ND6_DEBUG, &nd6_debug, 0, 1 },
1889 };
1890 
1891 int
1892 icmp6_sysctl_icmp6stat(void *oldp, size_t *oldlenp, void *newp)
1893 {
1894 	struct icmp6stat *icmp6stat;
1895 	int ret;
1896 
1897 	CTASSERT(sizeof(*icmp6stat) == icp6s_ncounters * sizeof(uint64_t));
1898 	icmp6stat = malloc(sizeof(*icmp6stat), M_TEMP, M_WAITOK|M_ZERO);
1899 	counters_read(icmp6counters, (uint64_t *)icmp6stat, icp6s_ncounters);
1900 	ret = sysctl_rdstruct(oldp, oldlenp, newp,
1901 	    icmp6stat, sizeof(*icmp6stat));
1902 	free(icmp6stat, M_TEMP, sizeof(*icmp6stat));
1903 
1904 	return (ret);
1905 }
1906 
1907 int
1908 icmp6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1909     void *newp, size_t newlen)
1910 {
1911 	int error;
1912 
1913 	/* All sysctl names at this level are terminal. */
1914 	if (namelen != 1)
1915 		return ENOTDIR;
1916 
1917 	switch (name[0]) {
1918 
1919 	case ICMPV6CTL_STATS:
1920 		return icmp6_sysctl_icmp6stat(oldp, oldlenp, newp);
1921 	default:
1922 		NET_LOCK();
1923 		error = sysctl_bounded_arr(icmpv6ctl_vars,
1924 		    nitems(icmpv6ctl_vars), name, namelen, oldp, oldlenp, newp,
1925 		    newlen);
1926 		NET_UNLOCK();
1927 		return (error);
1928 	}
1929 	/* NOTREACHED */
1930 }
1931