xref: /netbsd-src/sys/netinet6/icmp6.c (revision 413d532bcc3f62d122e56d92e13ac64825a40baf)
1 /*	$NetBSD: icmp6.c,v 1.165 2014/02/25 18:30:12 pooka 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 <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: icmp6.c,v 1.165 2014/02/25 18:30:12 pooka Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_ipsec.h"
69 
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/time.h>
78 #include <sys/kernel.h>
79 #include <sys/syslog.h>
80 #include <sys/domain.h>
81 #include <sys/sysctl.h>
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 #include <net/if_dl.h>
86 #include <net/if_types.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/ip6.h>
91 #include <netinet6/ip6_var.h>
92 #include <netinet6/ip6_private.h>
93 #include <netinet/icmp6.h>
94 #include <netinet6/icmp6_private.h>
95 #include <netinet6/mld6_var.h>
96 #include <netinet6/in6_pcb.h>
97 #include <netinet6/nd6.h>
98 #include <netinet6/in6_ifattach.h>
99 #include <netinet6/ip6protosw.h>
100 #include <netinet6/scope6_var.h>
101 
102 #ifdef IPSEC
103 #include <netipsec/ipsec.h>
104 #include <netipsec/key.h>
105 #endif
106 
107 
108 #include "faith.h"
109 #if defined(NFAITH) && 0 < NFAITH
110 #include <net/if_faith.h>
111 #endif
112 
113 #include <net/net_osdep.h>
114 
115 extern struct domain inet6domain;
116 
117 percpu_t *icmp6stat_percpu;
118 
119 extern struct inpcbtable raw6cbtable;
120 extern int icmp6errppslim;
121 static int icmp6errpps_count = 0;
122 static struct timeval icmp6errppslim_last;
123 extern int icmp6_nodeinfo;
124 
125 /*
126  * List of callbacks to notify when Path MTU changes are made.
127  */
128 struct icmp6_mtudisc_callback {
129 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
130 	void (*mc_func)(struct in6_addr *);
131 };
132 
133 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
134     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
135 
136 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
137 extern int pmtu_expire;
138 
139 /* XXX do these values make any sense? */
140 static int icmp6_mtudisc_hiwat = 1280;
141 static int icmp6_mtudisc_lowat = 256;
142 
143 /*
144  * keep track of # of redirect routes.
145  */
146 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
147 
148 /* XXX experimental, turned off */
149 static int icmp6_redirect_hiwat = -1;
150 static int icmp6_redirect_lowat = -1;
151 
152 static void icmp6_errcount(u_int, int, int);
153 static int icmp6_rip6_input(struct mbuf **, int);
154 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
155 static const char *icmp6_redirect_diag(struct in6_addr *,
156 	struct in6_addr *, struct in6_addr *);
157 static struct mbuf *ni6_input(struct mbuf *, int);
158 static struct mbuf *ni6_nametodns(const char *, int, int);
159 static int ni6_dnsmatch(const char *, int, const char *, int);
160 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
161 			  struct ifnet **, char *);
162 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
163 				struct ifnet *, int);
164 static int icmp6_notify_error(struct mbuf *, int, int, int);
165 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
166 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
167 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
168 static void sysctl_net_inet6_icmp6_setup(struct sysctllog **);
169 
170 
171 void
172 icmp6_init(void)
173 {
174 
175 	sysctl_net_inet6_icmp6_setup(NULL);
176 	mld_init();
177 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
178 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
179 
180 	icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS);
181 }
182 
183 static void
184 icmp6_errcount(u_int base, int type, int code)
185 {
186 	switch (type) {
187 	case ICMP6_DST_UNREACH:
188 		switch (code) {
189 		case ICMP6_DST_UNREACH_NOROUTE:
190 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE);
191 			return;
192 		case ICMP6_DST_UNREACH_ADMIN:
193 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN);
194 			return;
195 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
196 			ICMP6_STATINC(base +
197 				      ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE);
198 			return;
199 		case ICMP6_DST_UNREACH_ADDR:
200 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR);
201 			return;
202 		case ICMP6_DST_UNREACH_NOPORT:
203 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT);
204 			return;
205 		}
206 		break;
207 	case ICMP6_PACKET_TOO_BIG:
208 		ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG);
209 		return;
210 	case ICMP6_TIME_EXCEEDED:
211 		switch (code) {
212 		case ICMP6_TIME_EXCEED_TRANSIT:
213 			ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT);
214 			return;
215 		case ICMP6_TIME_EXCEED_REASSEMBLY:
216 			ICMP6_STATINC(base +
217 				      ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY);
218 			return;
219 		}
220 		break;
221 	case ICMP6_PARAM_PROB:
222 		switch (code) {
223 		case ICMP6_PARAMPROB_HEADER:
224 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER);
225 			return;
226 		case ICMP6_PARAMPROB_NEXTHEADER:
227 			ICMP6_STATINC(base +
228 				      ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER);
229 			return;
230 		case ICMP6_PARAMPROB_OPTION:
231 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION);
232 			return;
233 		}
234 		break;
235 	case ND_REDIRECT:
236 		ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT);
237 		return;
238 	}
239 	ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN);
240 }
241 
242 /*
243  * Register a Path MTU Discovery callback.
244  */
245 void
246 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *))
247 {
248 	struct icmp6_mtudisc_callback *mc;
249 
250 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
251 	     mc = LIST_NEXT(mc, mc_list)) {
252 		if (mc->mc_func == func)
253 			return;
254 	}
255 
256 	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
257 	if (mc == NULL)
258 		panic("icmp6_mtudisc_callback_register");
259 
260 	mc->mc_func = func;
261 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
262 }
263 
264 /*
265  * A wrapper function for icmp6_error() necessary when the erroneous packet
266  * may not contain enough scope zone information.
267  */
268 void
269 icmp6_error2(struct mbuf *m, int type, int code, int param,
270 	struct ifnet *ifp)
271 {
272 	struct ip6_hdr *ip6;
273 
274 	if (ifp == NULL)
275 		return;
276 
277 	if (m->m_len < sizeof(struct ip6_hdr)) {
278 		m = m_pullup(m, sizeof(struct ip6_hdr));
279 		if (m == NULL)
280 			return;
281 	}
282 
283 	ip6 = mtod(m, struct ip6_hdr *);
284 
285 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
286 		return;
287 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
288 		return;
289 
290 	icmp6_error(m, type, code, param);
291 }
292 
293 /*
294  * Generate an error packet of type error in response to bad IP6 packet.
295  */
296 void
297 icmp6_error(struct mbuf *m, int type, int code, int param)
298 {
299 	struct ip6_hdr *oip6, *nip6;
300 	struct icmp6_hdr *icmp6;
301 	u_int preplen;
302 	int off;
303 	int nxt;
304 
305 	ICMP6_STATINC(ICMP6_STAT_ERROR);
306 
307 	/* count per-type-code statistics */
308 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code);
309 
310 	if (m->m_flags & M_DECRYPTED) {
311 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
312 		goto freeit;
313 	}
314 
315 	if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) &&
316 	    (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL)
317 		return;
318 	oip6 = mtod(m, struct ip6_hdr *);
319 
320 	/*
321 	 * If the destination address of the erroneous packet is a multicast
322 	 * address, or the packet was sent using link-layer multicast,
323 	 * we should basically suppress sending an error (RFC 2463, Section
324 	 * 2.4).
325 	 * We have two exceptions (the item e.2 in that section):
326 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
327 	 * - the Parameter Problem Message that can be allowed an icmp6 error
328 	 *   in the option type field.  This check has been done in
329 	 *   ip6_unknown_opt(), so we can just check the type and code.
330 	 */
331 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
332 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
333 	    (type != ICMP6_PACKET_TOO_BIG &&
334 	     (type != ICMP6_PARAM_PROB ||
335 	      code != ICMP6_PARAMPROB_OPTION)))
336 		goto freeit;
337 
338 	/*
339 	 * RFC 2463, 2.4 (e.5): source address check.
340 	 * XXX: the case of anycast source?
341 	 */
342 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
343 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
344 		goto freeit;
345 
346 	/*
347 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
348 	 * don't do it.
349 	 */
350 	nxt = -1;
351 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
352 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
353 		struct icmp6_hdr *icp;
354 
355 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
356 			sizeof(*icp));
357 		if (icp == NULL) {
358 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
359 			return;
360 		}
361 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
362 		    icp->icmp6_type == ND_REDIRECT) {
363 			/*
364 			 * ICMPv6 error
365 			 * Special case: for redirect (which is
366 			 * informational) we must not send icmp6 error.
367 			 */
368 			ICMP6_STATINC(ICMP6_STAT_CANTERROR);
369 			goto freeit;
370 		} else {
371 			/* ICMPv6 informational - send the error */
372 		}
373 	}
374 #if 0 /* controversial */
375 	else if (off >= 0 && nxt == IPPROTO_ESP) {
376 		/*
377 		 * It could be ICMPv6 error inside ESP.  Take a safer side,
378 		 * don't respond.
379 		 */
380 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
381 		goto freeit;
382 	}
383 #endif
384 	else {
385 		/* non-ICMPv6 - send the error */
386 	}
387 
388 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
389 
390 	/* Finally, do rate limitation check. */
391 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
392 		ICMP6_STATINC(ICMP6_STAT_TOOFREQ);
393 		goto freeit;
394 	}
395 
396 	/*
397 	 * OK, ICMP6 can be generated.
398 	 */
399 
400 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
401 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
402 
403 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
404 	M_PREPEND(m, preplen, M_DONTWAIT);
405 	if (m && M_UNWRITABLE(m, preplen))
406 		m = m_pullup(m, preplen);
407 	if (m == NULL) {
408 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
409 		return;
410 	}
411 
412 	nip6 = mtod(m, struct ip6_hdr *);
413 	nip6->ip6_src  = oip6->ip6_src;
414 	nip6->ip6_dst  = oip6->ip6_dst;
415 
416 	in6_clearscope(&oip6->ip6_src);
417 	in6_clearscope(&oip6->ip6_dst);
418 
419 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
420 	icmp6->icmp6_type = type;
421 	icmp6->icmp6_code = code;
422 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
423 
424 	/*
425 	 * icmp6_reflect() is designed to be in the input path.
426 	 * icmp6_error() can be called from both input and output path,
427 	 * and if we are in output path rcvif could contain bogus value.
428 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
429 	 * information in ip header (nip6).
430 	 */
431 	m->m_pkthdr.rcvif = NULL;
432 
433 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
434 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
435 
436 	return;
437 
438   freeit:
439 	/*
440 	 * If we can't tell whether or not we can generate ICMP6, free it.
441 	 */
442 	m_freem(m);
443 }
444 
445 /*
446  * Process a received ICMP6 message.
447  */
448 int
449 icmp6_input(struct mbuf **mp, int *offp, int proto)
450 {
451 	struct mbuf *m = *mp, *n;
452 	struct ip6_hdr *ip6, *nip6;
453 	struct icmp6_hdr *icmp6, *nicmp6;
454 	int off = *offp;
455 	int icmp6len = m->m_pkthdr.len - *offp;
456 	int code, sum, noff, i;
457 
458 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
459 	KASSERT(ICMP6_MAXLEN < MCLBYTES);
460 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
461 
462 	/*
463 	 * Locate icmp6 structure in mbuf, and check
464 	 * that not corrupted and of at least minimum length
465 	 */
466 
467 	if (icmp6len < sizeof(struct icmp6_hdr)) {
468 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
469 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
470 		goto freeit;
471 	}
472 
473 	i = off + sizeof(*icmp6);
474 	if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) {
475 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
476 #if 0 /* m is 0 here */
477 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
478 #endif
479 		goto freeit;
480 	}
481 	ip6 = mtod(m, struct ip6_hdr *);
482 	/*
483 	 * calculate the checksum
484 	 */
485 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
486 	if (icmp6 == NULL) {
487 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
488 		/* m is invalid */
489 		/*icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);*/
490 		return IPPROTO_DONE;
491 	}
492 	KASSERT(IP6_HDR_ALIGNED_P(icmp6));
493 	code = icmp6->icmp6_code;
494 
495 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
496 		nd6log((LOG_ERR,
497 		    "ICMP6 checksum error(%d|%x) %s\n",
498 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
499 		ICMP6_STATINC(ICMP6_STAT_CHECKSUM);
500 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
501 		goto freeit;
502 	}
503 
504 #if defined(NFAITH) && 0 < NFAITH
505 	if (faithprefix(&ip6->ip6_dst)) {
506 		/*
507 		 * Deliver very specific ICMP6 type only.
508 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
509 		 * will not work.
510 		 */
511 		switch (icmp6->icmp6_type) {
512 		case ICMP6_DST_UNREACH:
513 		case ICMP6_PACKET_TOO_BIG:
514 		case ICMP6_TIME_EXCEEDED:
515 			break;
516 		default:
517 			goto freeit;
518 		}
519 	}
520 #endif
521 
522 	ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type);
523 
524 	switch (icmp6->icmp6_type) {
525 	case ICMP6_DST_UNREACH:
526 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
527 		switch (code) {
528 		case ICMP6_DST_UNREACH_NOROUTE:
529 			code = PRC_UNREACH_NET;
530 			break;
531 		case ICMP6_DST_UNREACH_ADMIN:
532 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
533 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
534 			break;
535 		case ICMP6_DST_UNREACH_ADDR:
536 			code = PRC_HOSTDEAD;
537 			break;
538 #ifdef COMPAT_RFC1885
539 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
540 			code = PRC_UNREACH_SRCFAIL;
541 			break;
542 #else
543 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
544 			/* I mean "source address was incorrect." */
545 			code = PRC_UNREACH_NET;
546 			break;
547 #endif
548 		case ICMP6_DST_UNREACH_NOPORT:
549 			code = PRC_UNREACH_PORT;
550 			break;
551 		default:
552 			goto badcode;
553 		}
554 		goto deliver;
555 
556 	case ICMP6_PACKET_TOO_BIG:
557 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
558 
559 		/*
560 		 * MTU is checked in icmp6_mtudisc.
561 		 */
562 		code = PRC_MSGSIZE;
563 
564 		/*
565 		 * Updating the path MTU will be done after examining
566 		 * intermediate extension headers.
567 		 */
568 		goto deliver;
569 
570 	case ICMP6_TIME_EXCEEDED:
571 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
572 		switch (code) {
573 		case ICMP6_TIME_EXCEED_TRANSIT:
574 			code = PRC_TIMXCEED_INTRANS;
575 			break;
576 		case ICMP6_TIME_EXCEED_REASSEMBLY:
577 			code = PRC_TIMXCEED_REASS;
578 			break;
579 		default:
580 			goto badcode;
581 		}
582 		goto deliver;
583 
584 	case ICMP6_PARAM_PROB:
585 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
586 		switch (code) {
587 		case ICMP6_PARAMPROB_NEXTHEADER:
588 			code = PRC_UNREACH_PROTOCOL;
589 			break;
590 		case ICMP6_PARAMPROB_HEADER:
591 		case ICMP6_PARAMPROB_OPTION:
592 			code = PRC_PARAMPROB;
593 			break;
594 		default:
595 			goto badcode;
596 		}
597 		goto deliver;
598 
599 	case ICMP6_ECHO_REQUEST:
600 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
601 		if (code != 0)
602 			goto badcode;
603 		/*
604 		 * Copy mbuf to send to two data paths: userland socket(s),
605 		 * and to the querier (echo reply).
606 		 * m: a copy for socket, n: a copy for querier
607 		 *
608 		 * If the first mbuf is shared, or the first mbuf is too short,
609 		 * copy the first part of the data into a fresh mbuf.
610 		 * Otherwise, we will wrongly overwrite both copies.
611 		 */
612 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
613 			/* Give up local */
614 			n = m;
615 			m = NULL;
616 		} else if (M_READONLY(n) ||
617 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
618 			struct mbuf *n0 = n;
619 
620 			/*
621 			 * Prepare an internal mbuf.  m_pullup() doesn't
622 			 * always copy the length we specified.
623 			 */
624 			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
625 				/* Give up local */
626 				n = m;
627 				m = NULL;
628 			}
629 			m_freem(n0);
630 		}
631 		IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
632 		    sizeof(*nicmp6));
633 		if (nicmp6 == NULL)
634 			goto freeit;
635 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
636 		nicmp6->icmp6_code = 0;
637 		if (n) {
638 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
639 			icmp6s[ICMP6_STAT_REFLECT]++;
640 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++;
641 			ICMP6_STAT_PUTREF();
642 			icmp6_reflect(n, off);
643 		}
644 		if (!m)
645 			goto freeit;
646 		break;
647 
648 	case ICMP6_ECHO_REPLY:
649 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
650 		if (code != 0)
651 			goto badcode;
652 		break;
653 
654 	case MLD_LISTENER_QUERY:
655 	case MLD_LISTENER_REPORT:
656 		if (icmp6len < sizeof(struct mld_hdr))
657 			goto badlen;
658 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
659 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
660 		else
661 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
662 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
663 			/* give up local */
664 			mld_input(m, off);
665 			m = NULL;
666 			goto freeit;
667 		}
668 		mld_input(n, off);
669 		/* m stays. */
670 		break;
671 
672 	case MLD_LISTENER_DONE:
673 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
674 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
675 			goto badlen;
676 		break;		/* nothing to be done in kernel */
677 
678 	case MLD_MTRACE_RESP:
679 	case MLD_MTRACE:
680 		/* XXX: these two are experimental.  not officially defined. */
681 		/* XXX: per-interface statistics? */
682 		break;		/* just pass it to applications */
683 
684 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
685 	    {
686 		enum { WRU, FQDN } mode;
687 
688 		if (!icmp6_nodeinfo)
689 			break;
690 
691 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
692 			mode = WRU;
693 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
694 			mode = FQDN;
695 		else
696 			goto badlen;
697 
698 		if (mode == FQDN) {
699 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
700 			if (n)
701 				n = ni6_input(n, off);
702 			/* XXX meaningless if n == NULL */
703 			noff = sizeof(struct ip6_hdr);
704 		} else {
705 			u_char *p;
706 			int maxhlen;
707 
708 			if ((icmp6_nodeinfo & 5) != 5)
709 				break;
710 
711 			if (code != 0)
712 				goto badcode;
713 			MGETHDR(n, M_DONTWAIT, m->m_type);
714 			if (n && ICMP6_MAXLEN > MHLEN) {
715 				MCLGET(n, M_DONTWAIT);
716 				if ((n->m_flags & M_EXT) == 0) {
717 					m_free(n);
718 					n = NULL;
719 				}
720 			}
721 			if (n == NULL) {
722 				/* Give up remote */
723 				break;
724 			}
725 			n->m_pkthdr.rcvif = NULL;
726 			n->m_len = 0;
727 			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
728 			if (maxhlen > hostnamelen)
729 				maxhlen = hostnamelen;
730 			/*
731 			 * Copy IPv6 and ICMPv6 only.
732 			 */
733 			nip6 = mtod(n, struct ip6_hdr *);
734 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
735 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
736 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
737 			p = (u_char *)(nicmp6 + 1);
738 			memset(p, 0, 4);
739 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
740 			noff = sizeof(struct ip6_hdr);
741 			M_COPY_PKTHDR(n, m); /* just for rcvif */
742 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
743 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
744 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
745 			nicmp6->icmp6_code = 0;
746 		}
747 #undef hostnamelen
748 		if (n) {
749 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
750 			icmp6s[ICMP6_STAT_REFLECT]++;
751 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++;
752 			ICMP6_STAT_PUTREF();
753 			icmp6_reflect(n, noff);
754 		}
755 		break;
756 	    }
757 
758 	case ICMP6_WRUREPLY:
759 		if (code != 0)
760 			goto badcode;
761 		break;
762 
763 	case ND_ROUTER_SOLICIT:
764 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
765 		if (code != 0)
766 			goto badcode;
767 		if (icmp6len < sizeof(struct nd_router_solicit))
768 			goto badlen;
769 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
770 			/* give up local */
771 			nd6_rs_input(m, off, icmp6len);
772 			m = NULL;
773 			goto freeit;
774 		}
775 		nd6_rs_input(n, off, icmp6len);
776 		/* m stays. */
777 		break;
778 
779 	case ND_ROUTER_ADVERT:
780 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
781 		if (code != 0)
782 			goto badcode;
783 		if (icmp6len < sizeof(struct nd_router_advert))
784 			goto badlen;
785 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
786 			/* give up local */
787 			nd6_ra_input(m, off, icmp6len);
788 			m = NULL;
789 			goto freeit;
790 		}
791 		nd6_ra_input(n, off, icmp6len);
792 		/* m stays. */
793 		break;
794 
795 	case ND_NEIGHBOR_SOLICIT:
796 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
797 		if (code != 0)
798 			goto badcode;
799 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
800 			goto badlen;
801 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
802 			/* give up local */
803 			nd6_ns_input(m, off, icmp6len);
804 			m = NULL;
805 			goto freeit;
806 		}
807 		nd6_ns_input(n, off, icmp6len);
808 		/* m stays. */
809 		break;
810 
811 	case ND_NEIGHBOR_ADVERT:
812 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
813 		if (code != 0)
814 			goto badcode;
815 		if (icmp6len < sizeof(struct nd_neighbor_advert))
816 			goto badlen;
817 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
818 			/* give up local */
819 			nd6_na_input(m, off, icmp6len);
820 			m = NULL;
821 			goto freeit;
822 		}
823 		nd6_na_input(n, off, icmp6len);
824 		/* m stays. */
825 		break;
826 
827 	case ND_REDIRECT:
828 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
829 		if (code != 0)
830 			goto badcode;
831 		if (icmp6len < sizeof(struct nd_redirect))
832 			goto badlen;
833 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
834 			/* give up local */
835 			icmp6_redirect_input(m, off);
836 			m = NULL;
837 			goto freeit;
838 		}
839 		icmp6_redirect_input(n, off);
840 		/* m stays. */
841 		break;
842 
843 	case ICMP6_ROUTER_RENUMBERING:
844 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
845 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
846 			goto badcode;
847 		if (icmp6len < sizeof(struct icmp6_router_renum))
848 			goto badlen;
849 		break;
850 
851 	default:
852 		nd6log((LOG_DEBUG,
853 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
854 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
855 		    ip6_sprintf(&ip6->ip6_dst),
856 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
857 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
858 			/* ICMPv6 error: MUST deliver it by spec... */
859 			code = PRC_NCMDS;
860 			/* deliver */
861 		} else {
862 			/* ICMPv6 informational: MUST not deliver */
863 			break;
864 		}
865 	deliver:
866 		if (icmp6_notify_error(m, off, icmp6len, code)) {
867 			/* In this case, m should've been freed. */
868 			return (IPPROTO_DONE);
869 		}
870 		break;
871 
872 	badcode:
873 		ICMP6_STATINC(ICMP6_STAT_BADCODE);
874 		break;
875 
876 	badlen:
877 		ICMP6_STATINC(ICMP6_STAT_BADLEN);
878 		break;
879 	}
880 
881 	/* deliver the packet to appropriate sockets */
882 	icmp6_rip6_input(&m, *offp);
883 
884 	return IPPROTO_DONE;
885 
886  freeit:
887 	m_freem(m);
888 	return IPPROTO_DONE;
889 }
890 
891 static int
892 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
893 {
894 	struct icmp6_hdr *icmp6;
895 	struct ip6_hdr *eip6;
896 	u_int32_t notifymtu;
897 	struct sockaddr_in6 icmp6src, icmp6dst;
898 
899 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
900 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
901 		goto freeit;
902 	}
903 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
904 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
905 	if (icmp6 == NULL) {
906 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
907 		return (-1);
908 	}
909 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
910 
911 	/* Detect the upper level protocol */
912 	{
913 		void (*ctlfunc)(int, struct sockaddr *, void *);
914 		u_int8_t nxt = eip6->ip6_nxt;
915 		int eoff = off + sizeof(struct icmp6_hdr) +
916 			sizeof(struct ip6_hdr);
917 		struct ip6ctlparam ip6cp;
918 		struct in6_addr *finaldst = NULL;
919 		int icmp6type = icmp6->icmp6_type;
920 		struct ip6_frag *fh;
921 		struct ip6_rthdr *rth;
922 		struct ip6_rthdr0 *rth0;
923 		int rthlen;
924 
925 		while (1) { /* XXX: should avoid infinite loop explicitly? */
926 			struct ip6_ext *eh;
927 
928 			switch (nxt) {
929 			case IPPROTO_HOPOPTS:
930 			case IPPROTO_DSTOPTS:
931 			case IPPROTO_AH:
932 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
933 					       eoff, sizeof(*eh));
934 				if (eh == NULL) {
935 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
936 					return (-1);
937 				}
938 
939 				if (nxt == IPPROTO_AH)
940 					eoff += (eh->ip6e_len + 2) << 2;
941 				else
942 					eoff += (eh->ip6e_len + 1) << 3;
943 				nxt = eh->ip6e_nxt;
944 				break;
945 			case IPPROTO_ROUTING:
946 				/*
947 				 * When the erroneous packet contains a
948 				 * routing header, we should examine the
949 				 * header to determine the final destination.
950 				 * Otherwise, we can't properly update
951 				 * information that depends on the final
952 				 * destination (e.g. path MTU).
953 				 */
954 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
955 					       eoff, sizeof(*rth));
956 				if (rth == NULL) {
957 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
958 					return (-1);
959 				}
960 				rthlen = (rth->ip6r_len + 1) << 3;
961 				/*
962 				 * XXX: currently there is no
963 				 * officially defined type other
964 				 * than type-0.
965 				 * Note that if the segment left field
966 				 * is 0, all intermediate hops must
967 				 * have been passed.
968 				 */
969 				if (rth->ip6r_segleft &&
970 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
971 					int hops;
972 
973 					IP6_EXTHDR_GET(rth0,
974 						       struct ip6_rthdr0 *, m,
975 						       eoff, rthlen);
976 					if (rth0 == NULL) {
977 						ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
978 						return (-1);
979 					}
980 					/* just ignore a bogus header */
981 					if ((rth0->ip6r0_len % 2) == 0 &&
982 					    (hops = rth0->ip6r0_len/2))
983 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
984 				}
985 				eoff += rthlen;
986 				nxt = rth->ip6r_nxt;
987 				break;
988 			case IPPROTO_FRAGMENT:
989 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
990 					       eoff, sizeof(*fh));
991 				if (fh == NULL) {
992 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
993 					return (-1);
994 				}
995 				/*
996 				 * Data after a fragment header is meaningless
997 				 * unless it is the first fragment, but
998 				 * we'll go to the notify label for path MTU
999 				 * discovery.
1000 				 */
1001 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1002 					goto notify;
1003 
1004 				eoff += sizeof(struct ip6_frag);
1005 				nxt = fh->ip6f_nxt;
1006 				break;
1007 			default:
1008 				/*
1009 				 * This case includes ESP and the No Next
1010 				 * Header.  In such cases going to the notify
1011 				 * label does not have any meaning
1012 				 * (i.e. ctlfunc will be NULL), but we go
1013 				 * anyway since we might have to update
1014 				 * path MTU information.
1015 				 */
1016 				goto notify;
1017 			}
1018 		}
1019 	  notify:
1020 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1021 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1022 		if (icmp6 == NULL) {
1023 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1024 			return (-1);
1025 		}
1026 
1027 		/*
1028 		 * retrieve parameters from the inner IPv6 header, and convert
1029 		 * them into sockaddr structures.
1030 		 * XXX: there is no guarantee that the source or destination
1031 		 * addresses of the inner packet are in the same scope zone as
1032 		 * the addresses of the icmp packet.  But there is no other
1033 		 * way to determine the zone.
1034 		 */
1035 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1036 
1037 		sockaddr_in6_init(&icmp6dst,
1038 		    (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0);
1039 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1040 			goto freeit;
1041 		sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0);
1042 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1043 			goto freeit;
1044 		icmp6src.sin6_flowinfo =
1045 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1046 
1047 		if (finaldst == NULL)
1048 			finaldst = &eip6->ip6_dst;
1049 		ip6cp.ip6c_m = m;
1050 		ip6cp.ip6c_icmp6 = icmp6;
1051 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1052 		ip6cp.ip6c_off = eoff;
1053 		ip6cp.ip6c_finaldst = finaldst;
1054 		ip6cp.ip6c_src = &icmp6src;
1055 		ip6cp.ip6c_nxt = nxt;
1056 
1057 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1058 			notifymtu = ntohl(icmp6->icmp6_mtu);
1059 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1060 		}
1061 
1062 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1063 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1064 		if (ctlfunc) {
1065 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1066 					  &ip6cp);
1067 		}
1068 	}
1069 	return (0);
1070 
1071   freeit:
1072 	m_freem(m);
1073 	return (-1);
1074 }
1075 
1076 void
1077 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1078 {
1079 	unsigned long rtcount;
1080 	struct icmp6_mtudisc_callback *mc;
1081 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1082 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1083 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1084 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1085 	struct rtentry *rt = NULL;
1086 	struct sockaddr_in6 sin6;
1087 
1088 	/*
1089 	 * The MTU should not be less than the minimal IPv6 MTU except for the
1090 	 * hack in ip6_output/ip6_setpmtu where we always include a frag header.
1091 	 * In that one case, the MTU might be less than 1280.
1092 	 */
1093 	if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) {
1094 		/* is the mtu even sane? */
1095 		if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1096 			return;
1097 		if (!validated)
1098 			return;
1099 		mtu = IPV6_MMTU - sizeof(struct ip6_frag);
1100 	}
1101 
1102 	/*
1103 	 * allow non-validated cases if memory is plenty, to make traffic
1104 	 * from non-connected pcb happy.
1105 	 */
1106 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1107 	if (validated) {
1108 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1109 			return;
1110 		else if (0 <= icmp6_mtudisc_lowat &&
1111 		    rtcount > icmp6_mtudisc_lowat) {
1112 			/*
1113 			 * XXX nuke a victim, install the new one.
1114 			 */
1115 		}
1116 	} else {
1117 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1118 			return;
1119 	}
1120 
1121 	memset(&sin6, 0, sizeof(sin6));
1122 	sin6.sin6_family = PF_INET6;
1123 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1124 	sin6.sin6_addr = *dst;
1125 	if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL))
1126 		return;
1127 
1128 	rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1129 
1130 	if (rt && (rt->rt_flags & RTF_HOST) &&
1131 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1132 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1133 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1134 			ICMP6_STATINC(ICMP6_STAT_PMTUCHG);
1135 			rt->rt_rmx.rmx_mtu = mtu;
1136 		}
1137 	}
1138 	if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1139 		RTFREE(rt);
1140 	}
1141 
1142 	/*
1143 	 * Notify protocols that the MTU for this destination
1144 	 * has changed.
1145 	 */
1146 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1147 	     mc = LIST_NEXT(mc, mc_list))
1148 		(*mc->mc_func)(&sin6.sin6_addr);
1149 }
1150 
1151 /*
1152  * Process a Node Information Query packet, based on
1153  * draft-ietf-ipngwg-icmp-name-lookups-07.
1154  *
1155  * Spec incompatibilities:
1156  * - IPv6 Subject address handling
1157  * - IPv4 Subject address handling support missing
1158  * - Proxy reply (answer even if it's not for me)
1159  * - joins NI group address at in6_ifattach() time only, does not cope
1160  *   with hostname changes by sethostname(3)
1161  */
1162 static struct mbuf *
1163 ni6_input(struct mbuf *m, int off)
1164 {
1165 	struct icmp6_nodeinfo *ni6, *nni6;
1166 	struct mbuf *n = NULL;
1167 	u_int16_t qtype;
1168 	int subjlen;
1169 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1170 	struct ni_reply_fqdn *fqdn;
1171 	int addrs;		/* for NI_QTYPE_NODEADDR */
1172 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1173 	struct sockaddr_in6 sin6; /* ip6_dst */
1174 	struct in6_addr in6_subj; /* subject address */
1175 	struct ip6_hdr *ip6;
1176 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1177 	char *subj = NULL;
1178 
1179 	ip6 = mtod(m, struct ip6_hdr *);
1180 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1181 	if (ni6 == NULL) {
1182 		/* m is already reclaimed */
1183 		return NULL;
1184 	}
1185 
1186 	/*
1187 	 * Validate IPv6 destination address.
1188 	 *
1189 	 * The Responder must discard the Query without further processing
1190 	 * unless it is one of the Responder's unicast or anycast addresses, or
1191 	 * a link-local scope multicast address which the Responder has joined.
1192 	 * [icmp-name-lookups-07, Section 4.]
1193 	 */
1194 	sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
1195 	/* XXX scopeid */
1196 	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1197 		; /* unicast/anycast, fine */
1198 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1199 		; /* link-local multicast, fine */
1200 	else
1201 		goto bad;
1202 
1203 	/* validate query Subject field. */
1204 	qtype = ntohs(ni6->ni_qtype);
1205 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1206 	switch (qtype) {
1207 	case NI_QTYPE_NOOP:
1208 	case NI_QTYPE_SUPTYPES:
1209 		/* 07 draft */
1210 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1211 			break;
1212 		/* FALLTHROUGH */
1213 	case NI_QTYPE_FQDN:
1214 	case NI_QTYPE_NODEADDR:
1215 	case NI_QTYPE_IPV4ADDR:
1216 		switch (ni6->ni_code) {
1217 		case ICMP6_NI_SUBJ_IPV6:
1218 #if ICMP6_NI_SUBJ_IPV6 != 0
1219 		case 0:
1220 #endif
1221 			/*
1222 			 * backward compatibility - try to accept 03 draft
1223 			 * format, where no Subject is present.
1224 			 */
1225 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1226 			    subjlen == 0) {
1227 				oldfqdn++;
1228 				break;
1229 			}
1230 #if ICMP6_NI_SUBJ_IPV6 != 0
1231 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1232 				goto bad;
1233 #endif
1234 
1235 			if (subjlen != sizeof(sin6.sin6_addr))
1236 				goto bad;
1237 
1238 			/*
1239 			 * Validate Subject address.
1240 			 *
1241 			 * Not sure what exactly "address belongs to the node"
1242 			 * means in the spec, is it just unicast, or what?
1243 			 *
1244 			 * At this moment we consider Subject address as
1245 			 * "belong to the node" if the Subject address equals
1246 			 * to the IPv6 destination address; validation for
1247 			 * IPv6 destination address should have done enough
1248 			 * check for us.
1249 			 *
1250 			 * We do not do proxy at this moment.
1251 			 */
1252 			/* m_pulldown instead of copy? */
1253 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1254 			    subjlen, (void *)&in6_subj);
1255 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1256 				goto bad;
1257 
1258 			subj = (char *)&in6_subj;
1259 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1260 				break;
1261 
1262 			/*
1263 			 * XXX if we are to allow other cases, we should really
1264 			 * be careful about scope here.
1265 			 * basically, we should disallow queries toward IPv6
1266 			 * destination X with subject Y, if scope(X) > scope(Y).
1267 			 * if we allow scope(X) > scope(Y), it will result in
1268 			 * information leakage across scope boundary.
1269 			 */
1270 			goto bad;
1271 
1272 		case ICMP6_NI_SUBJ_FQDN:
1273 			/*
1274 			 * Validate Subject name with gethostname(3).
1275 			 *
1276 			 * The behavior may need some debate, since:
1277 			 * - we are not sure if the node has FQDN as
1278 			 *   hostname (returned by gethostname(3)).
1279 			 * - the code does wildcard match for truncated names.
1280 			 *   however, we are not sure if we want to perform
1281 			 *   wildcard match, if gethostname(3) side has
1282 			 *   truncated hostname.
1283 			 */
1284 			n = ni6_nametodns(hostname, hostnamelen, 0);
1285 			if (!n || n->m_next || n->m_len == 0)
1286 				goto bad;
1287 			IP6_EXTHDR_GET(subj, char *, m,
1288 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1289 			if (subj == NULL)
1290 				goto bad;
1291 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1292 					n->m_len)) {
1293 				goto bad;
1294 			}
1295 			m_freem(n);
1296 			n = NULL;
1297 			break;
1298 
1299 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1300 		default:
1301 			goto bad;
1302 		}
1303 		break;
1304 	}
1305 
1306 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1307 	switch (qtype) {
1308 	case NI_QTYPE_FQDN:
1309 		if ((icmp6_nodeinfo & 1) == 0)
1310 			goto bad;
1311 		break;
1312 	case NI_QTYPE_NODEADDR:
1313 	case NI_QTYPE_IPV4ADDR:
1314 		if ((icmp6_nodeinfo & 2) == 0)
1315 			goto bad;
1316 		break;
1317 	}
1318 
1319 	/* guess reply length */
1320 	switch (qtype) {
1321 	case NI_QTYPE_NOOP:
1322 		break;		/* no reply data */
1323 	case NI_QTYPE_SUPTYPES:
1324 		replylen += sizeof(u_int32_t);
1325 		break;
1326 	case NI_QTYPE_FQDN:
1327 		/* XXX will append an mbuf */
1328 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1329 		break;
1330 	case NI_QTYPE_NODEADDR:
1331 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1332 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1333 					  sizeof(u_int32_t))) > MCLBYTES)
1334 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1335 		break;
1336 	case NI_QTYPE_IPV4ADDR:
1337 		/* unsupported - should respond with unknown Qtype? */
1338 		goto bad;
1339 	default:
1340 		/*
1341 		 * XXX: We must return a reply with the ICMP6 code
1342 		 * `unknown Qtype' in this case.  However we regard the case
1343 		 * as an FQDN query for backward compatibility.
1344 		 * Older versions set a random value to this field,
1345 		 * so it rarely varies in the defined qtypes.
1346 		 * But the mechanism is not reliable...
1347 		 * maybe we should obsolete older versions.
1348 		 */
1349 		qtype = NI_QTYPE_FQDN;
1350 		/* XXX will append an mbuf */
1351 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1352 		oldfqdn++;
1353 		break;
1354 	}
1355 
1356 	/* allocate an mbuf to reply. */
1357 	MGETHDR(n, M_DONTWAIT, m->m_type);
1358 	if (n == NULL) {
1359 		m_freem(m);
1360 		return (NULL);
1361 	}
1362 	M_MOVE_PKTHDR(n, m); /* just for rcvif */
1363 	if (replylen > MHLEN) {
1364 		if (replylen > MCLBYTES) {
1365 			/*
1366 			 * XXX: should we try to allocate more? But MCLBYTES
1367 			 * is probably much larger than IPV6_MMTU...
1368 			 */
1369 			goto bad;
1370 		}
1371 		MCLGET(n, M_DONTWAIT);
1372 		if ((n->m_flags & M_EXT) == 0) {
1373 			goto bad;
1374 		}
1375 	}
1376 	n->m_pkthdr.len = n->m_len = replylen;
1377 
1378 	/* copy mbuf header and IPv6 + Node Information base headers */
1379 	bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr));
1380 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1381 	bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo));
1382 
1383 	/* qtype dependent procedure */
1384 	switch (qtype) {
1385 	case NI_QTYPE_NOOP:
1386 		nni6->ni_code = ICMP6_NI_SUCCESS;
1387 		nni6->ni_flags = 0;
1388 		break;
1389 	case NI_QTYPE_SUPTYPES:
1390 	{
1391 		u_int32_t v;
1392 		nni6->ni_code = ICMP6_NI_SUCCESS;
1393 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1394 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1395 		v = (u_int32_t)htonl(0x0000000f);
1396 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1397 		break;
1398 	}
1399 	case NI_QTYPE_FQDN:
1400 		nni6->ni_code = ICMP6_NI_SUCCESS;
1401 		fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) +
1402 						sizeof(struct ip6_hdr) +
1403 						sizeof(struct icmp6_nodeinfo));
1404 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1405 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1406 		/*
1407 		 * XXX do we really have FQDN in variable "hostname"?
1408 		 */
1409 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1410 		if (n->m_next == NULL)
1411 			goto bad;
1412 		/* XXX we assume that n->m_next is not a chain */
1413 		if (n->m_next->m_next != NULL)
1414 			goto bad;
1415 		n->m_pkthdr.len += n->m_next->m_len;
1416 		break;
1417 	case NI_QTYPE_NODEADDR:
1418 	{
1419 		int lenlim, copied;
1420 
1421 		nni6->ni_code = ICMP6_NI_SUCCESS;
1422 		n->m_pkthdr.len = n->m_len =
1423 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1424 		lenlim = M_TRAILINGSPACE(n);
1425 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1426 		/* XXX: reset mbuf length */
1427 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1428 			sizeof(struct icmp6_nodeinfo) + copied;
1429 		break;
1430 	}
1431 	default:
1432 		break;		/* XXX impossible! */
1433 	}
1434 
1435 	nni6->ni_type = ICMP6_NI_REPLY;
1436 	m_freem(m);
1437 	return (n);
1438 
1439   bad:
1440 	m_freem(m);
1441 	if (n)
1442 		m_freem(n);
1443 	return (NULL);
1444 }
1445 #undef hostnamelen
1446 
1447 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1448 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1449 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1450 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1451 
1452 /*
1453  * make a mbuf with DNS-encoded string.  no compression support.
1454  *
1455  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1456  * treated as truncated name (two \0 at the end).  this is a wild guess.
1457  *
1458  * old - return pascal string if non-zero
1459  */
1460 static struct mbuf *
1461 ni6_nametodns(const char *name, int namelen, int old)
1462 {
1463 	struct mbuf *m;
1464 	char *cp, *ep;
1465 	const char *p, *q;
1466 	int i, len, nterm;
1467 
1468 	if (old)
1469 		len = namelen + 1;
1470 	else
1471 		len = MCLBYTES;
1472 
1473 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1474 	MGET(m, M_DONTWAIT, MT_DATA);
1475 	if (m && len > MLEN) {
1476 		MCLGET(m, M_DONTWAIT);
1477 		if ((m->m_flags & M_EXT) == 0)
1478 			goto fail;
1479 	}
1480 	if (!m)
1481 		goto fail;
1482 	m->m_next = NULL;
1483 
1484 	if (old) {
1485 		m->m_len = len;
1486 		*mtod(m, char *) = namelen;
1487 		bcopy(name, mtod(m, char *) + 1, namelen);
1488 		return m;
1489 	} else {
1490 		m->m_len = 0;
1491 		cp = mtod(m, char *);
1492 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1493 
1494 		/* if not certain about my name, return empty buffer */
1495 		if (namelen == 0)
1496 			return m;
1497 
1498 		/*
1499 		 * guess if it looks like shortened hostname, or FQDN.
1500 		 * shortened hostname needs two trailing "\0".
1501 		 */
1502 		i = 0;
1503 		for (p = name; p < name + namelen; p++) {
1504 			if (*p && *p == '.')
1505 				i++;
1506 		}
1507 		if (i < 2)
1508 			nterm = 2;
1509 		else
1510 			nterm = 1;
1511 
1512 		p = name;
1513 		while (cp < ep && p < name + namelen) {
1514 			i = 0;
1515 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1516 				i++;
1517 			/* result does not fit into mbuf */
1518 			if (cp + i + 1 >= ep)
1519 				goto fail;
1520 			/*
1521 			 * DNS label length restriction, RFC1035 page 8.
1522 			 * "i == 0" case is included here to avoid returning
1523 			 * 0-length label on "foo..bar".
1524 			 */
1525 			if (i <= 0 || i >= 64)
1526 				goto fail;
1527 			*cp++ = i;
1528 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1529 				goto fail;
1530 			while (i > 0) {
1531 				if (!isalnum(*p) && *p != '-')
1532 					goto fail;
1533 				if (isupper(*p)) {
1534 					*cp++ = tolower(*p);
1535 					p++;
1536 				} else
1537 					*cp++ = *p++;
1538 				i--;
1539 			}
1540 			p = q;
1541 			if (p < name + namelen && *p == '.')
1542 				p++;
1543 		}
1544 		/* termination */
1545 		if (cp + nterm >= ep)
1546 			goto fail;
1547 		while (nterm-- > 0)
1548 			*cp++ = '\0';
1549 		m->m_len = cp - mtod(m, char *);
1550 		return m;
1551 	}
1552 
1553 	panic("should not reach here");
1554 	/* NOTREACHED */
1555 
1556  fail:
1557 	if (m)
1558 		m_freem(m);
1559 	return NULL;
1560 }
1561 
1562 /*
1563  * check if two DNS-encoded string matches.  takes care of truncated
1564  * form (with \0\0 at the end).  no compression support.
1565  * XXX upper/lowercase match (see RFC2065)
1566  */
1567 static int
1568 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1569 {
1570 	const char *a0, *b0;
1571 	int l;
1572 
1573 	/* simplest case - need validation? */
1574 	if (alen == blen && memcmp(a, b, alen) == 0)
1575 		return 1;
1576 
1577 	a0 = a;
1578 	b0 = b;
1579 
1580 	/* termination is mandatory */
1581 	if (alen < 2 || blen < 2)
1582 		return 0;
1583 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1584 		return 0;
1585 	alen--;
1586 	blen--;
1587 
1588 	while (a - a0 < alen && b - b0 < blen) {
1589 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1590 			return 0;
1591 
1592 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1593 			return 0;
1594 		/* we don't support compression yet */
1595 		if (a[0] >= 64 || b[0] >= 64)
1596 			return 0;
1597 
1598 		/* truncated case */
1599 		if (a[0] == 0 && a - a0 == alen - 1)
1600 			return 1;
1601 		if (b[0] == 0 && b - b0 == blen - 1)
1602 			return 1;
1603 		if (a[0] == 0 || b[0] == 0)
1604 			return 0;
1605 
1606 		if (a[0] != b[0])
1607 			return 0;
1608 		l = a[0];
1609 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1610 			return 0;
1611 		if (memcmp(a + 1, b + 1, l) != 0)
1612 			return 0;
1613 
1614 		a += 1 + l;
1615 		b += 1 + l;
1616 	}
1617 
1618 	if (a - a0 == alen && b - b0 == blen)
1619 		return 1;
1620 	else
1621 		return 0;
1622 }
1623 
1624 /*
1625  * calculate the number of addresses to be returned in the node info reply.
1626  */
1627 static int
1628 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m,
1629     struct ifnet **ifpp, char *subj)
1630 {
1631 	struct ifnet *ifp;
1632 	struct in6_ifaddr *ifa6;
1633 	struct ifaddr *ifa;
1634 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1635 	int addrs = 0, addrsofif, iffound = 0;
1636 	int niflags = ni6->ni_flags;
1637 
1638 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1639 		switch (ni6->ni_code) {
1640 		case ICMP6_NI_SUBJ_IPV6:
1641 			if (subj == NULL) /* must be impossible... */
1642 				return (0);
1643 			subj_ip6 = (struct sockaddr_in6 *)subj;
1644 			break;
1645 		default:
1646 			/*
1647 			 * XXX: we only support IPv6 subject address for
1648 			 * this Qtype.
1649 			 */
1650 			return (0);
1651 		}
1652 	}
1653 
1654 	IFNET_FOREACH(ifp) {
1655 		addrsofif = 0;
1656 		IFADDR_FOREACH(ifa, ifp) {
1657 			if (ifa->ifa_addr->sa_family != AF_INET6)
1658 				continue;
1659 			ifa6 = (struct in6_ifaddr *)ifa;
1660 
1661 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1662 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1663 					       &ifa6->ia_addr.sin6_addr))
1664 				iffound = 1;
1665 
1666 			/*
1667 			 * IPv4-mapped addresses can only be returned by a
1668 			 * Node Information proxy, since they represent
1669 			 * addresses of IPv4-only nodes, which perforce do
1670 			 * not implement this protocol.
1671 			 * [icmp-name-lookups-07, Section 5.4]
1672 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1673 			 * this function at this moment.
1674 			 */
1675 
1676 			/* What do we have to do about ::1? */
1677 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1678 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1679 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1680 					continue;
1681 				break;
1682 			case IPV6_ADDR_SCOPE_SITELOCAL:
1683 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1684 					continue;
1685 				break;
1686 			case IPV6_ADDR_SCOPE_GLOBAL:
1687 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1688 					continue;
1689 				break;
1690 			default:
1691 				continue;
1692 			}
1693 
1694 			/*
1695 			 * check if anycast is okay.
1696 			 * XXX: just experimental.  not in the spec.
1697 			 */
1698 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1699 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1700 				continue; /* we need only unicast addresses */
1701 
1702 			addrsofif++; /* count the address */
1703 		}
1704 		if (iffound) {
1705 			*ifpp = ifp;
1706 			return (addrsofif);
1707 		}
1708 
1709 		addrs += addrsofif;
1710 	}
1711 
1712 	return (addrs);
1713 }
1714 
1715 static int
1716 ni6_store_addrs(struct icmp6_nodeinfo *ni6,
1717 	struct icmp6_nodeinfo *nni6, struct ifnet *ifp0,
1718 	int resid)
1719 {
1720 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1721 	struct in6_ifaddr *ifa6;
1722 	struct ifaddr *ifa;
1723 	struct ifnet *ifp_dep = NULL;
1724 	int copied = 0, allow_deprecated = 0;
1725 	u_char *cp = (u_char *)(nni6 + 1);
1726 	int niflags = ni6->ni_flags;
1727 	u_int32_t ltime;
1728 
1729 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1730 		return (0);	/* needless to copy */
1731 
1732   again:
1733 
1734 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1735 	{
1736 		IFADDR_FOREACH(ifa, ifp) {
1737 			if (ifa->ifa_addr->sa_family != AF_INET6)
1738 				continue;
1739 			ifa6 = (struct in6_ifaddr *)ifa;
1740 
1741 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1742 			    allow_deprecated == 0) {
1743 				/*
1744 				 * prefererred address should be put before
1745 				 * deprecated addresses.
1746 				 */
1747 
1748 				/* record the interface for later search */
1749 				if (ifp_dep == NULL)
1750 					ifp_dep = ifp;
1751 
1752 				continue;
1753 			}
1754 			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1755 				 allow_deprecated != 0)
1756 				continue; /* we now collect deprecated addrs */
1757 
1758 			/* What do we have to do about ::1? */
1759 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1760 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1761 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1762 					continue;
1763 				break;
1764 			case IPV6_ADDR_SCOPE_SITELOCAL:
1765 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1766 					continue;
1767 				break;
1768 			case IPV6_ADDR_SCOPE_GLOBAL:
1769 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1770 					continue;
1771 				break;
1772 			default:
1773 				continue;
1774 			}
1775 
1776 			/*
1777 			 * check if anycast is okay.
1778 			 * XXX: just experimental.  not in the spec.
1779 			 */
1780 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1781 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1782 				continue;
1783 
1784 			/* now we can copy the address */
1785 			if (resid < sizeof(struct in6_addr) +
1786 			    sizeof(u_int32_t)) {
1787 				/*
1788 				 * We give up much more copy.
1789 				 * Set the truncate flag and return.
1790 				 */
1791 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1792 				return (copied);
1793 			}
1794 
1795 			/*
1796 			 * Set the TTL of the address.
1797 			 * The TTL value should be one of the following
1798 			 * according to the specification:
1799 			 *
1800 			 * 1. The remaining lifetime of a DHCP lease on the
1801 			 *    address, or
1802 			 * 2. The remaining Valid Lifetime of a prefix from
1803 			 *    which the address was derived through Stateless
1804 			 *    Autoconfiguration.
1805 			 *
1806 			 * Note that we currently do not support stateful
1807 			 * address configuration by DHCPv6, so the former
1808 			 * case can't happen.
1809 			 *
1810 			 * TTL must be 2^31 > TTL >= 0.
1811 			 */
1812 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1813 				ltime = ND6_INFINITE_LIFETIME;
1814 			else {
1815 				if (ifa6->ia6_lifetime.ia6t_expire >
1816 				    time_second)
1817 					ltime = ifa6->ia6_lifetime.ia6t_expire -
1818 					    time_second;
1819 				else
1820 					ltime = 0;
1821 			}
1822 			if (ltime > 0x7fffffff)
1823 				ltime = 0x7fffffff;
1824 			ltime = htonl(ltime);
1825 
1826 			bcopy(&ltime, cp, sizeof(u_int32_t));
1827 			cp += sizeof(u_int32_t);
1828 
1829 			/* copy the address itself */
1830 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1831 			      sizeof(struct in6_addr));
1832 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1833 			cp += sizeof(struct in6_addr);
1834 
1835 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1836 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1837 		}
1838 		if (ifp0)	/* we need search only on the specified IF */
1839 			break;
1840 	}
1841 
1842 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1843 		ifp = ifp_dep;
1844 		allow_deprecated = 1;
1845 
1846 		goto again;
1847 	}
1848 
1849 	return (copied);
1850 }
1851 
1852 /*
1853  * XXX almost dup'ed code with rip6_input.
1854  */
1855 static int
1856 icmp6_rip6_input(struct mbuf **mp, int off)
1857 {
1858 	struct mbuf *m = *mp;
1859 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1860 	struct inpcb_hdr *inph;
1861 	struct in6pcb *in6p;
1862 	struct in6pcb *last = NULL;
1863 	struct sockaddr_in6 rip6src;
1864 	struct icmp6_hdr *icmp6;
1865 	struct mbuf *opts = NULL;
1866 
1867 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1868 	if (icmp6 == NULL) {
1869 		/* m is already reclaimed */
1870 		return IPPROTO_DONE;
1871 	}
1872 
1873 	/*
1874 	 * XXX: the address may have embedded scope zone ID, which should be
1875 	 * hidden from applications.
1876 	 */
1877 	sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0);
1878 	if (sa6_recoverscope(&rip6src)) {
1879 		m_freem(m);
1880 		return (IPPROTO_DONE);
1881 	}
1882 
1883 	TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1884 		in6p = (struct in6pcb *)inph;
1885 		if (in6p->in6p_af != AF_INET6)
1886 			continue;
1887 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1888 			continue;
1889 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1890 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1891 			continue;
1892 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1893 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1894 			continue;
1895 		if (in6p->in6p_icmp6filt
1896 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1897 				 in6p->in6p_icmp6filt))
1898 			continue;
1899 		if (last) {
1900 			struct	mbuf *n;
1901 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1902 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1903 					ip6_savecontrol(last, &opts, ip6, n);
1904 				/* strip intermediate headers */
1905 				m_adj(n, off);
1906 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1907 						 (struct sockaddr *)&rip6src,
1908 						 n, opts) == 0) {
1909 					/* should notify about lost packet */
1910 					m_freem(n);
1911 					if (opts)
1912 						m_freem(opts);
1913 				} else
1914 					sorwakeup(last->in6p_socket);
1915 				opts = NULL;
1916 			}
1917 		}
1918 		last = in6p;
1919 	}
1920 	if (last) {
1921 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1922 			ip6_savecontrol(last, &opts, ip6, m);
1923 		/* strip intermediate headers */
1924 		m_adj(m, off);
1925 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1926 				(struct sockaddr *)&rip6src, m, opts) == 0) {
1927 			m_freem(m);
1928 			if (opts)
1929 				m_freem(opts);
1930 		} else
1931 			sorwakeup(last->in6p_socket);
1932 	} else {
1933 		m_freem(m);
1934 		IP6_STATDEC(IP6_STAT_DELIVERED);
1935 	}
1936 	return IPPROTO_DONE;
1937 }
1938 
1939 /*
1940  * Reflect the ip6 packet back to the source.
1941  * OFF points to the icmp6 header, counted from the top of the mbuf.
1942  *
1943  * Note: RFC 1885 required that an echo reply should be truncated if it
1944  * did not fit in with (return) path MTU, and KAME code supported the
1945  * behavior.  However, as a clarification after the RFC, this limitation
1946  * was removed in a revised version of the spec, RFC 2463.  We had kept the
1947  * old behavior, with a (non-default) ifdef block, while the new version of
1948  * the spec was an internet-draft status, and even after the new RFC was
1949  * published.  But it would rather make sense to clean the obsoleted part
1950  * up, and to make the code simpler at this stage.
1951  */
1952 void
1953 icmp6_reflect(struct mbuf *m, size_t off)
1954 {
1955 	struct ip6_hdr *ip6;
1956 	struct icmp6_hdr *icmp6;
1957 	const struct in6_ifaddr *ia;
1958 	const struct ip6aux *ip6a;
1959 	int plen;
1960 	int type, code;
1961 	struct ifnet *outif = NULL;
1962 	struct in6_addr origdst;
1963 	const struct in6_addr *src = NULL;
1964 
1965 	/* too short to reflect */
1966 	if (off < sizeof(struct ip6_hdr)) {
1967 		nd6log((LOG_DEBUG,
1968 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1969 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1970 		    __FILE__, __LINE__));
1971 		goto bad;
1972 	}
1973 
1974 	/*
1975 	 * If there are extra headers between IPv6 and ICMPv6, strip
1976 	 * off that header first.
1977 	 */
1978 #ifdef DIAGNOSTIC
1979 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1980 		panic("assumption failed in icmp6_reflect");
1981 #endif
1982 	if (off > sizeof(struct ip6_hdr)) {
1983 		size_t l;
1984 		struct ip6_hdr nip6;
1985 
1986 		l = off - sizeof(struct ip6_hdr);
1987 		m_copydata(m, 0, sizeof(nip6), (void *)&nip6);
1988 		m_adj(m, l);
1989 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1990 		if (m->m_len < l) {
1991 			if ((m = m_pullup(m, l)) == NULL)
1992 				return;
1993 		}
1994 		bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6));
1995 	} else /* off == sizeof(struct ip6_hdr) */ {
1996 		size_t l;
1997 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1998 		if (m->m_len < l) {
1999 			if ((m = m_pullup(m, l)) == NULL)
2000 				return;
2001 		}
2002 	}
2003 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2004 	ip6 = mtod(m, struct ip6_hdr *);
2005 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2006 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2007 	type = icmp6->icmp6_type; /* keep type for statistics */
2008 	code = icmp6->icmp6_code; /* ditto. */
2009 
2010 	origdst = ip6->ip6_dst;
2011 	/*
2012 	 * ip6_input() drops a packet if its src is multicast.
2013 	 * So, the src is never multicast.
2014 	 */
2015 	ip6->ip6_dst = ip6->ip6_src;
2016 
2017 	/*
2018 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2019 	 * use dst as the src for the reply.
2020 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2021 	 * (for example) when we encounter an error while forwarding procedure
2022 	 * destined to a duplicated address of ours.
2023 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2024 	 * procedure of an outgoing packet of our own, in which case we need
2025 	 * to search in the ifaddr list.
2026 	 */
2027 	if (IN6_IS_ADDR_MULTICAST(&origdst))
2028 		;
2029 	else if ((ip6a = ip6_getdstifaddr(m)) != NULL) {
2030 		if ((ip6a->ip6a_flags &
2031 		     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0)
2032 			src = &ip6a->ip6a_src;
2033 	} else {
2034 		union {
2035 			struct sockaddr_in6 sin6;
2036 			struct sockaddr sa;
2037 		} u;
2038 
2039 		sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0);
2040 
2041 		ia = (struct in6_ifaddr *)ifa_ifwithaddr(&u.sa);
2042 
2043 		if (ia == NULL)
2044 			;
2045 		else if ((ia->ia6_flags &
2046 			 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0)
2047 			src = &ia->ia_addr.sin6_addr;
2048 	}
2049 
2050 	if (src == NULL) {
2051 		int e;
2052 		struct sockaddr_in6 sin6;
2053 		struct route ro;
2054 
2055 		/*
2056 		 * This case matches to multicasts, our anycast, or unicasts
2057 		 * that we do not own.  Select a source address based on the
2058 		 * source address of the erroneous packet.
2059 		 */
2060 		/* zone ID should be embedded */
2061 		sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
2062 
2063 		memset(&ro, 0, sizeof(ro));
2064 		src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e);
2065 		rtcache_free(&ro);
2066 		if (src == NULL) {
2067 			nd6log((LOG_DEBUG,
2068 			    "icmp6_reflect: source can't be determined: "
2069 			    "dst=%s, error=%d\n",
2070 			    ip6_sprintf(&sin6.sin6_addr), e));
2071 			goto bad;
2072 		}
2073 	}
2074 
2075 	ip6->ip6_src = *src;
2076 	ip6->ip6_flow = 0;
2077 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2078 	ip6->ip6_vfc |= IPV6_VERSION;
2079 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2080 	if (m->m_pkthdr.rcvif) {
2081 		/* XXX: This may not be the outgoing interface */
2082 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2083 	} else
2084 		ip6->ip6_hlim = ip6_defhlim;
2085 
2086 	m->m_pkthdr.csum_flags = 0;
2087 	icmp6->icmp6_cksum = 0;
2088 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2089 					sizeof(struct ip6_hdr), plen);
2090 
2091 	/*
2092 	 * XXX option handling
2093 	 */
2094 
2095 	m->m_flags &= ~(M_BCAST|M_MCAST);
2096 
2097 	/*
2098 	 * To avoid a "too big" situation at an intermediate router
2099 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2100 	 * Note that only echo and node information replies are affected,
2101 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2102 	 */
2103 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 &&
2104 	    outif)
2105 		icmp6_ifstat_inc(outif, ifs6_out_error);
2106 
2107 	if (outif)
2108 		icmp6_ifoutstat_inc(outif, type, code);
2109 
2110 	return;
2111 
2112  bad:
2113 	m_freem(m);
2114 	return;
2115 }
2116 
2117 static const char *
2118 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2119 	struct in6_addr *tgt6)
2120 {
2121 	static char buf[1024];
2122 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2123 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2124 	return buf;
2125 }
2126 
2127 void
2128 icmp6_redirect_input(struct mbuf *m, int off)
2129 {
2130 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2131 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2132 	struct nd_redirect *nd_rd;
2133 	int icmp6len = ntohs(ip6->ip6_plen);
2134 	char *lladdr = NULL;
2135 	int lladdrlen = 0;
2136 	struct rtentry *rt = NULL;
2137 	int is_router;
2138 	int is_onlink;
2139 	struct in6_addr src6 = ip6->ip6_src;
2140 	struct in6_addr redtgt6;
2141 	struct in6_addr reddst6;
2142 	union nd_opts ndopts;
2143 
2144 	if (!ifp)
2145 		return;
2146 
2147 	/* XXX if we are router, we don't update route by icmp6 redirect */
2148 	if (ip6_forwarding)
2149 		goto freeit;
2150 	if (!icmp6_rediraccept)
2151 		goto freeit;
2152 
2153 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2154 	if (nd_rd == NULL) {
2155 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
2156 		return;
2157 	}
2158 	redtgt6 = nd_rd->nd_rd_target;
2159 	reddst6 = nd_rd->nd_rd_dst;
2160 
2161 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2162 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2163 		goto freeit;
2164 	}
2165 
2166 	/* validation */
2167 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2168 		nd6log((LOG_ERR,
2169 			"ICMP6 redirect sent from %s rejected; "
2170 			"must be from linklocal\n", ip6_sprintf(&src6)));
2171 		goto bad;
2172 	}
2173 	if (ip6->ip6_hlim != 255) {
2174 		nd6log((LOG_ERR,
2175 			"ICMP6 redirect sent from %s rejected; "
2176 			"hlim=%d (must be 255)\n",
2177 			ip6_sprintf(&src6), ip6->ip6_hlim));
2178 		goto bad;
2179 	}
2180     {
2181 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2182 	struct sockaddr_in6 sin6;
2183 	struct in6_addr *gw6;
2184 
2185 	sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0);
2186 	rt = rtalloc1((struct sockaddr *)&sin6, 0);
2187 	if (rt) {
2188 		if (rt->rt_gateway == NULL ||
2189 		    rt->rt_gateway->sa_family != AF_INET6) {
2190 			nd6log((LOG_ERR,
2191 			    "ICMP6 redirect rejected; no route "
2192 			    "with inet6 gateway found for redirect dst: %s\n",
2193 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2194 			RTFREE(rt);
2195 			goto bad;
2196 		}
2197 
2198 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2199 		if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2200 			nd6log((LOG_ERR,
2201 				"ICMP6 redirect rejected; "
2202 				"not equal to gw-for-src=%s (must be same): "
2203 				"%s\n",
2204 				ip6_sprintf(gw6),
2205 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2206 			RTFREE(rt);
2207 			goto bad;
2208 		}
2209 	} else {
2210 		nd6log((LOG_ERR,
2211 			"ICMP6 redirect rejected; "
2212 			"no route found for redirect dst: %s\n",
2213 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2214 		goto bad;
2215 	}
2216 	RTFREE(rt);
2217 	rt = NULL;
2218     }
2219 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2220 		nd6log((LOG_ERR,
2221 			"ICMP6 redirect rejected; "
2222 			"redirect dst must be unicast: %s\n",
2223 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2224 		goto bad;
2225 	}
2226 
2227 	is_router = is_onlink = 0;
2228 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2229 		is_router = 1;	/* router case */
2230 	if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2231 		is_onlink = 1;	/* on-link destination case */
2232 	if (!is_router && !is_onlink) {
2233 		nd6log((LOG_ERR,
2234 			"ICMP6 redirect rejected; "
2235 			"neither router case nor onlink case: %s\n",
2236 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2237 		goto bad;
2238 	}
2239 	/* validation passed */
2240 
2241 	icmp6len -= sizeof(*nd_rd);
2242 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2243 	if (nd6_options(&ndopts) < 0) {
2244 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2245 			"invalid ND option, rejected: %s\n",
2246 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2247 		/* nd6_options have incremented stats */
2248 		goto freeit;
2249 	}
2250 
2251 	if (ndopts.nd_opts_tgt_lladdr) {
2252 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2253 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2254 	}
2255 
2256 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2257 		nd6log((LOG_INFO,
2258 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2259 			"(if %d, icmp6 packet %d): %s\n",
2260 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2261 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2262 		goto bad;
2263 	}
2264 
2265 	/* RFC 2461 8.3 */
2266 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2267 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2268 
2269 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2270 		/* perform rtredirect */
2271 		struct sockaddr_in6 sdst;
2272 		struct sockaddr_in6 sgw;
2273 		struct sockaddr_in6 ssrc;
2274 		unsigned long rtcount;
2275 		struct rtentry *newrt = NULL;
2276 
2277 		/*
2278 		 * do not install redirect route, if the number of entries
2279 		 * is too much (> hiwat).  note that, the node (= host) will
2280 		 * work just fine even if we do not install redirect route
2281 		 * (there will be additional hops, though).
2282 		 */
2283 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2284 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes)
2285 			goto freeit;
2286 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat)
2287 			return;
2288 		else if (0 <= icmp6_redirect_lowat &&
2289 		    rtcount > icmp6_redirect_lowat) {
2290 			/*
2291 			 * XXX nuke a victim, install the new one.
2292 			 */
2293 		}
2294 
2295 		memset(&sdst, 0, sizeof(sdst));
2296 		memset(&sgw, 0, sizeof(sgw));
2297 		memset(&ssrc, 0, sizeof(ssrc));
2298 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2299 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2300 			sizeof(struct sockaddr_in6);
2301 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2302 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2303 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2304 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2305 			   NULL, RTF_GATEWAY | RTF_HOST,
2306 			   (struct sockaddr *)&ssrc,
2307 			   &newrt);
2308 
2309 		if (newrt) {
2310 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2311 			    icmp6_redirect_timeout_q);
2312 			rtfree(newrt);
2313 		}
2314 	}
2315 	/* finally update cached route in each socket via pfctlinput */
2316 	{
2317 		struct sockaddr_in6 sdst;
2318 
2319 		sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0);
2320 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2321 #if defined(IPSEC)
2322 		key_sa_routechange((struct sockaddr *)&sdst);
2323 #endif
2324 	}
2325 
2326  freeit:
2327 	m_freem(m);
2328 	return;
2329 
2330  bad:
2331 	ICMP6_STATINC(ICMP6_STAT_BADREDIRECT);
2332 	m_freem(m);
2333 }
2334 
2335 void
2336 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2337 {
2338 	struct ifnet *ifp;	/* my outgoing interface */
2339 	struct in6_addr *ifp_ll6;
2340 	struct in6_addr *nexthop;
2341 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2342 	struct mbuf *m = NULL;	/* newly allocated one */
2343 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2344 	struct nd_redirect *nd_rd;
2345 	size_t maxlen;
2346 	u_char *p;
2347 	struct sockaddr_in6 src_sa;
2348 
2349 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0);
2350 
2351 	/* if we are not router, we don't send icmp6 redirect */
2352 	if (!ip6_forwarding)
2353 		goto fail;
2354 
2355 	/* sanity check */
2356 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2357 		goto fail;
2358 
2359 	/*
2360 	 * Address check:
2361 	 *  the source address must identify a neighbor, and
2362 	 *  the destination address must not be a multicast address
2363 	 *  [RFC 2461, sec 8.2]
2364 	 */
2365 	sip6 = mtod(m0, struct ip6_hdr *);
2366 	sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0);
2367 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2368 		goto fail;
2369 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2370 		goto fail;	/* what should we do here? */
2371 
2372 	/* rate limit */
2373 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2374 		goto fail;
2375 
2376 	/*
2377 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2378 	 * we almost always ask for an mbuf cluster for simplicity.
2379 	 * (MHLEN < IPV6_MMTU is almost always true)
2380 	 */
2381 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2382 	if (m && IPV6_MMTU >= MHLEN) {
2383 #if IPV6_MMTU >= MCLBYTES
2384 		_MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT);
2385 #else
2386 		MCLGET(m, M_DONTWAIT);
2387 #endif
2388 	}
2389 
2390 	if (!m)
2391 		goto fail;
2392 	m->m_pkthdr.rcvif = NULL;
2393 	m->m_len = 0;
2394 	maxlen = M_TRAILINGSPACE(m);
2395 	maxlen = min(IPV6_MMTU, maxlen);
2396 	/* just for safety */
2397 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2398 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2399 		goto fail;
2400 	}
2401 
2402 	{
2403 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2404 		struct in6_ifaddr *ia;
2405 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2406 						 IN6_IFF_NOTREADY|
2407 						 IN6_IFF_ANYCAST)) == NULL)
2408 			goto fail;
2409 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2410 	}
2411 
2412 	/* get ip6 linklocal address for the router. */
2413 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2414 		struct sockaddr_in6 *sin6;
2415 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2416 		nexthop = &sin6->sin6_addr;
2417 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2418 			nexthop = NULL;
2419 	} else
2420 		nexthop = NULL;
2421 
2422 	/* ip6 */
2423 	ip6 = mtod(m, struct ip6_hdr *);
2424 	ip6->ip6_flow = 0;
2425 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2426 	ip6->ip6_vfc |= IPV6_VERSION;
2427 	/* ip6->ip6_plen will be set later */
2428 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2429 	ip6->ip6_hlim = 255;
2430 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2431 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2432 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2433 
2434 	/* ND Redirect */
2435 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2436 	nd_rd->nd_rd_type = ND_REDIRECT;
2437 	nd_rd->nd_rd_code = 0;
2438 	nd_rd->nd_rd_reserved = 0;
2439 	if (rt->rt_flags & RTF_GATEWAY) {
2440 		/*
2441 		 * nd_rd->nd_rd_target must be a link-local address in
2442 		 * better router cases.
2443 		 */
2444 		if (!nexthop)
2445 			goto fail;
2446 		bcopy(nexthop, &nd_rd->nd_rd_target,
2447 		      sizeof(nd_rd->nd_rd_target));
2448 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2449 		      sizeof(nd_rd->nd_rd_dst));
2450 	} else {
2451 		/* make sure redtgt == reddst */
2452 		nexthop = &sip6->ip6_dst;
2453 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2454 		      sizeof(nd_rd->nd_rd_target));
2455 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2456 		      sizeof(nd_rd->nd_rd_dst));
2457 	}
2458 
2459 	p = (u_char *)(nd_rd + 1);
2460 
2461 	{
2462 		/* target lladdr option */
2463 		struct rtentry *rt_nexthop = NULL;
2464 		int len;
2465 		const struct sockaddr_dl *sdl;
2466 		struct nd_opt_hdr *nd_opt;
2467 		char *lladdr;
2468 
2469 		rt_nexthop = nd6_lookup(nexthop, 0, ifp);
2470 		if (!rt_nexthop)
2471 			goto nolladdropt;
2472 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2473 		len = (len + 7) & ~7;	/* round by 8 */
2474 		/* safety check */
2475 		if (len + (p - (u_char *)ip6) > maxlen)
2476 			goto nolladdropt;
2477 		if (!(rt_nexthop->rt_flags & RTF_GATEWAY) &&
2478 		    (rt_nexthop->rt_flags & RTF_LLINFO) &&
2479 		    (rt_nexthop->rt_gateway->sa_family == AF_LINK) &&
2480 		    (sdl = satocsdl(rt_nexthop->rt_gateway)) &&
2481 		    sdl->sdl_alen) {
2482 			nd_opt = (struct nd_opt_hdr *)p;
2483 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2484 			nd_opt->nd_opt_len = len >> 3;
2485 			lladdr = (char *)(nd_opt + 1);
2486 			memcpy(lladdr, CLLADDR(sdl), ifp->if_addrlen);
2487 			p += len;
2488 		}
2489 	}
2490   nolladdropt:;
2491 
2492 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2493 
2494 	/* just to be safe */
2495 	if (m0->m_flags & M_DECRYPTED)
2496 		goto noredhdropt;
2497 	if (p - (u_char *)ip6 > maxlen)
2498 		goto noredhdropt;
2499 
2500 	{
2501 		/* redirected header option */
2502 		int len;
2503 		struct nd_opt_rd_hdr *nd_opt_rh;
2504 
2505 		/*
2506 		 * compute the maximum size for icmp6 redirect header option.
2507 		 * XXX room for auth header?
2508 		 */
2509 		len = maxlen - (p - (u_char *)ip6);
2510 		len &= ~7;
2511 
2512 		/*
2513 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2514 		 * about padding/truncate rule for the original IP packet.
2515 		 * From the discussion on IPv6imp in Feb 1999,
2516 		 * the consensus was:
2517 		 * - "attach as much as possible" is the goal
2518 		 * - pad if not aligned (original size can be guessed by
2519 		 *   original ip6 header)
2520 		 * Following code adds the padding if it is simple enough,
2521 		 * and truncates if not.
2522 		 */
2523 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2524 			/* not enough room, truncate */
2525 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2526 			    m0->m_pkthdr.len);
2527 		} else {
2528 			/*
2529 			 * enough room, truncate if not aligned.
2530 			 * we don't pad here for simplicity.
2531 			 */
2532 			size_t extra;
2533 
2534 			extra = m0->m_pkthdr.len % 8;
2535 			if (extra) {
2536 				/* truncate */
2537 				m_adj(m0, -extra);
2538 			}
2539 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2540 		}
2541 
2542 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2543 		memset(nd_opt_rh, 0, sizeof(*nd_opt_rh));
2544 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2545 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2546 		p += sizeof(*nd_opt_rh);
2547 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2548 
2549 		/* connect m0 to m */
2550 		m->m_pkthdr.len += m0->m_pkthdr.len;
2551 		m_cat(m, m0);
2552 		m0 = NULL;
2553 	}
2554 noredhdropt:
2555 	if (m0) {
2556 		m_freem(m0);
2557 		m0 = NULL;
2558 	}
2559 
2560 	/* XXX: clear embedded link IDs in the inner header */
2561 	in6_clearscope(&sip6->ip6_src);
2562 	in6_clearscope(&sip6->ip6_dst);
2563 	in6_clearscope(&nd_rd->nd_rd_target);
2564 	in6_clearscope(&nd_rd->nd_rd_dst);
2565 
2566 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2567 
2568 	nd_rd->nd_rd_cksum = 0;
2569 	nd_rd->nd_rd_cksum
2570 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2571 
2572 	/* send the packet to outside... */
2573 	if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0)
2574 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2575 
2576 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2577 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2578 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT);
2579 
2580 	return;
2581 
2582 fail:
2583 	if (m)
2584 		m_freem(m);
2585 	if (m0)
2586 		m_freem(m0);
2587 }
2588 
2589 /*
2590  * ICMPv6 socket option processing.
2591  */
2592 int
2593 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2594 {
2595 	int error = 0;
2596 	struct in6pcb *in6p = sotoin6pcb(so);
2597 
2598 	if (sopt->sopt_level != IPPROTO_ICMPV6)
2599 		return rip6_ctloutput(op, so, sopt);
2600 
2601 	switch (op) {
2602 	case PRCO_SETOPT:
2603 		switch (sopt->sopt_name) {
2604 		case ICMP6_FILTER:
2605 		    {
2606 			struct icmp6_filter fil;
2607 
2608 			error = sockopt_get(sopt, &fil, sizeof(fil));
2609 			if (error)
2610 				break;
2611 			memcpy(in6p->in6p_icmp6filt, &fil,
2612 			    sizeof(struct icmp6_filter));
2613 			error = 0;
2614 			break;
2615 		    }
2616 
2617 		default:
2618 			error = ENOPROTOOPT;
2619 			break;
2620 		}
2621 		break;
2622 
2623 	case PRCO_GETOPT:
2624 		switch (sopt->sopt_name) {
2625 		case ICMP6_FILTER:
2626 		    {
2627 			if (in6p->in6p_icmp6filt == NULL) {
2628 				error = EINVAL;
2629 				break;
2630 			}
2631 			error = sockopt_set(sopt, in6p->in6p_icmp6filt,
2632 			    sizeof(struct icmp6_filter));
2633 			break;
2634 		    }
2635 
2636 		default:
2637 			error = ENOPROTOOPT;
2638 			break;
2639 		}
2640 		break;
2641 	}
2642 
2643 	return (error);
2644 }
2645 
2646 /*
2647  * Perform rate limit check.
2648  * Returns 0 if it is okay to send the icmp6 packet.
2649  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2650  * limitation.
2651  *
2652  * XXX per-destination/type check necessary?
2653  */
2654 static int
2655 icmp6_ratelimit(
2656 	const struct in6_addr *dst,	/* not used at this moment */
2657 	const int type,		/* not used at this moment */
2658 	const int code)		/* not used at this moment */
2659 {
2660 	int ret;
2661 
2662 	ret = 0;	/* okay to send */
2663 
2664 	/* PPS limit */
2665 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2666 	    icmp6errppslim)) {
2667 		/* The packet is subject to rate limit */
2668 		ret++;
2669 	}
2670 
2671 	return ret;
2672 }
2673 
2674 static struct rtentry *
2675 icmp6_mtudisc_clone(struct sockaddr *dst)
2676 {
2677 	struct rtentry *rt;
2678 	int    error;
2679 
2680 	rt = rtalloc1(dst, 1);
2681 	if (rt == 0)
2682 		return NULL;
2683 
2684 	/* If we didn't get a host route, allocate one */
2685 	if ((rt->rt_flags & RTF_HOST) == 0) {
2686 		struct rtentry *nrt;
2687 
2688 		error = rtrequest((int) RTM_ADD, dst,
2689 		    (struct sockaddr *) rt->rt_gateway, NULL,
2690 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2691 		if (error) {
2692 			rtfree(rt);
2693 			return NULL;
2694 		}
2695 		nrt->rt_rmx = rt->rt_rmx;
2696 		rtfree(rt);
2697 		rt = nrt;
2698 	}
2699 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2700 			icmp6_mtudisc_timeout_q);
2701 	if (error) {
2702 		rtfree(rt);
2703 		return NULL;
2704 	}
2705 
2706 	return rt;	/* caller need to call rtfree() */
2707 }
2708 
2709 static void
2710 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
2711 {
2712 	if (rt == NULL)
2713 		panic("icmp6_mtudisc_timeout: bad route to timeout");
2714 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2715 	    (RTF_DYNAMIC | RTF_HOST)) {
2716 		rtrequest((int) RTM_DELETE, rt_getkey(rt),
2717 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2718 	} else {
2719 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2720 			rt->rt_rmx.rmx_mtu = 0;
2721 	}
2722 }
2723 
2724 static void
2725 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
2726 {
2727 	if (rt == NULL)
2728 		panic("icmp6_redirect_timeout: bad route to timeout");
2729 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2730 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2731 		rtrequest((int) RTM_DELETE, rt_getkey(rt),
2732 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2733 	}
2734 }
2735 
2736 /*
2737  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2738  */
2739 static int
2740 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2741 {
2742 	(void)&name;
2743 	(void)&l;
2744 	(void)&oname;
2745 
2746 	if (namelen != 0)
2747 		return (EINVAL);
2748 
2749 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2750 	    /*XXXUNCONST*/
2751 	    __UNCONST(newp), newlen));
2752 }
2753 
2754 static int
2755 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS)
2756 {
2757 
2758 	return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS));
2759 }
2760 
2761 static void
2762 sysctl_net_inet6_icmp6_setup(struct sysctllog **clog)
2763 {
2764 	extern int nd6_maxqueuelen; /* defined in nd6.c */
2765 
2766 	sysctl_createv(clog, 0, NULL, NULL,
2767 		       CTLFLAG_PERMANENT,
2768 		       CTLTYPE_NODE, "inet6", NULL,
2769 		       NULL, 0, NULL, 0,
2770 		       CTL_NET, PF_INET6, CTL_EOL);
2771 	sysctl_createv(clog, 0, NULL, NULL,
2772 		       CTLFLAG_PERMANENT,
2773 		       CTLTYPE_NODE, "icmp6",
2774 		       SYSCTL_DESCR("ICMPv6 related settings"),
2775 		       NULL, 0, NULL, 0,
2776 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2777 
2778 	sysctl_createv(clog, 0, NULL, NULL,
2779 		       CTLFLAG_PERMANENT,
2780 		       CTLTYPE_STRUCT, "stats",
2781 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2782 		       sysctl_net_inet6_icmp6_stats, 0, NULL, 0,
2783 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2784 		       ICMPV6CTL_STATS, CTL_EOL);
2785 	sysctl_createv(clog, 0, NULL, NULL,
2786 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2787 		       CTLTYPE_INT, "rediraccept",
2788 		       SYSCTL_DESCR("Accept and process redirect messages"),
2789 		       NULL, 0, &icmp6_rediraccept, 0,
2790 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2791 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2792 	sysctl_createv(clog, 0, NULL, NULL,
2793 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2794 		       CTLTYPE_INT, "redirtimeout",
2795 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2796 		       NULL, 0, &icmp6_redirtimeout, 0,
2797 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2798 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2799 #if 0 /* obsoleted */
2800 	sysctl_createv(clog, 0, NULL, NULL,
2801 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2802 		       CTLTYPE_INT, "errratelimit", NULL,
2803 		       NULL, 0, &icmp6_errratelimit, 0,
2804 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2805 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2806 #endif
2807 	sysctl_createv(clog, 0, NULL, NULL,
2808 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2809 		       CTLTYPE_INT, "nd6_prune",
2810 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
2811 		       NULL, 0, &nd6_prune, 0,
2812 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2813 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
2814 	sysctl_createv(clog, 0, NULL, NULL,
2815 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2816 		       CTLTYPE_INT, "nd6_delay",
2817 		       SYSCTL_DESCR("First probe delay time"),
2818 		       NULL, 0, &nd6_delay, 0,
2819 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2820 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
2821 	sysctl_createv(clog, 0, NULL, NULL,
2822 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2823 		       CTLTYPE_INT, "nd6_umaxtries",
2824 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2825 		       NULL, 0, &nd6_umaxtries, 0,
2826 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2827 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
2828 	sysctl_createv(clog, 0, NULL, NULL,
2829 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2830 		       CTLTYPE_INT, "nd6_mmaxtries",
2831 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
2832 		       NULL, 0, &nd6_mmaxtries, 0,
2833 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2834 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
2835 	sysctl_createv(clog, 0, NULL, NULL,
2836 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2837 		       CTLTYPE_INT, "nd6_useloopback",
2838 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
2839 		       NULL, 0, &nd6_useloopback, 0,
2840 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2841 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
2842 #if 0 /* obsoleted */
2843 	sysctl_createv(clog, 0, NULL, NULL,
2844 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2845 		       CTLTYPE_INT, "nd6_proxyall", NULL,
2846 		       NULL, 0, &nd6_proxyall, 0,
2847 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2848 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
2849 #endif
2850 	sysctl_createv(clog, 0, NULL, NULL,
2851 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2852 		       CTLTYPE_INT, "nodeinfo",
2853 		       SYSCTL_DESCR("Respond to node information requests"),
2854 		       NULL, 0, &icmp6_nodeinfo, 0,
2855 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2856 		       ICMPV6CTL_NODEINFO, CTL_EOL);
2857 	sysctl_createv(clog, 0, NULL, NULL,
2858 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2859 		       CTLTYPE_INT, "errppslimit",
2860 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
2861 		       NULL, 0, &icmp6errppslim, 0,
2862 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2863 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
2864 	sysctl_createv(clog, 0, NULL, NULL,
2865 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2866 		       CTLTYPE_INT, "nd6_maxnudhint",
2867 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2868 		       NULL, 0, &nd6_maxnudhint, 0,
2869 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2870 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
2871 	sysctl_createv(clog, 0, NULL, NULL,
2872 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2873 		       CTLTYPE_INT, "mtudisc_hiwat",
2874 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2875 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
2876 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2877 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
2878 	sysctl_createv(clog, 0, NULL, NULL,
2879 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2880 		       CTLTYPE_INT, "mtudisc_lowat",
2881 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2882 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
2883 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2884 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
2885 	sysctl_createv(clog, 0, NULL, NULL,
2886 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2887 		       CTLTYPE_INT, "nd6_debug",
2888 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
2889 		       NULL, 0, &nd6_debug, 0,
2890 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2891 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
2892 	sysctl_createv(clog, 0, NULL, NULL,
2893 		       CTLFLAG_PERMANENT,
2894 		       CTLTYPE_STRUCT, "nd6_drlist",
2895 		       SYSCTL_DESCR("Default router list"),
2896 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2897 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2898 		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
2899 	sysctl_createv(clog, 0, NULL, NULL,
2900 		       CTLFLAG_PERMANENT,
2901 		       CTLTYPE_STRUCT, "nd6_prlist",
2902 		       SYSCTL_DESCR("Prefix list"),
2903 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2904 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2905 		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
2906 	sysctl_createv(clog, 0, NULL, NULL,
2907 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2908 		       CTLTYPE_INT, "maxqueuelen",
2909 		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
2910 		       NULL, 1, &nd6_maxqueuelen, 0,
2911 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2912 		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
2913 }
2914 
2915 void
2916 icmp6_statinc(u_int stat)
2917 {
2918 
2919 	KASSERT(stat < ICMP6_NSTATS);
2920 	ICMP6_STATINC(stat);
2921 }
2922