xref: /netbsd-src/sys/netinet6/icmp6.c (revision cac8e449158efc7261bebc8657cbb0125a2cfdde)
1 /*	$NetBSD: icmp6.c,v 1.148 2008/05/07 11:28:37 bouyer 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 1.148 2008/05/07 11:28:37 bouyer 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 <netinet6/ipsec.h>
104 #include <netkey/key.h>
105 #endif
106 
107 #ifdef FAST_IPSEC
108 #include <netipsec/ipsec.h>
109 #include <netipsec/key.h>
110 #endif
111 
112 
113 #include "faith.h"
114 #if defined(NFAITH) && 0 < NFAITH
115 #include <net/if_faith.h>
116 #endif
117 
118 #include <net/net_osdep.h>
119 
120 extern struct domain inet6domain;
121 
122 percpu_t *icmp6stat_percpu;
123 
124 extern struct inpcbtable raw6cbtable;
125 extern int icmp6errppslim;
126 static int icmp6errpps_count = 0;
127 static struct timeval icmp6errppslim_last;
128 extern int icmp6_nodeinfo;
129 
130 /*
131  * List of callbacks to notify when Path MTU changes are made.
132  */
133 struct icmp6_mtudisc_callback {
134 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
135 	void (*mc_func)(struct in6_addr *);
136 };
137 
138 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
139     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
140 
141 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
142 extern int pmtu_expire;
143 
144 /* XXX do these values make any sense? */
145 static int icmp6_mtudisc_hiwat = 1280;
146 static int icmp6_mtudisc_lowat = 256;
147 
148 /*
149  * keep track of # of redirect routes.
150  */
151 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
152 
153 /* XXX experimental, turned off */
154 static int icmp6_redirect_hiwat = -1;
155 static int icmp6_redirect_lowat = -1;
156 
157 static void icmp6_errcount(u_int, int, int);
158 static int icmp6_rip6_input(struct mbuf **, int);
159 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
160 static const char *icmp6_redirect_diag(struct in6_addr *,
161 	struct in6_addr *, struct in6_addr *);
162 static struct mbuf *ni6_input(struct mbuf *, int);
163 static struct mbuf *ni6_nametodns(const char *, int, int);
164 static int ni6_dnsmatch(const char *, int, const char *, int);
165 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
166 			  struct ifnet **, char *);
167 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
168 				struct ifnet *, int);
169 static int icmp6_notify_error(struct mbuf *, int, int, int);
170 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
171 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
172 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
173 
174 
175 void
176 icmp6_init(void)
177 {
178 	mld_init();
179 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
180 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
181 
182 	icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS);
183 }
184 
185 static void
186 icmp6_errcount(u_int base, int type, int code)
187 {
188 	switch (type) {
189 	case ICMP6_DST_UNREACH:
190 		switch (code) {
191 		case ICMP6_DST_UNREACH_NOROUTE:
192 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE);
193 			return;
194 		case ICMP6_DST_UNREACH_ADMIN:
195 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN);
196 			return;
197 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
198 			ICMP6_STATINC(base +
199 				      ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE);
200 			return;
201 		case ICMP6_DST_UNREACH_ADDR:
202 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR);
203 			return;
204 		case ICMP6_DST_UNREACH_NOPORT:
205 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT);
206 			return;
207 		}
208 		break;
209 	case ICMP6_PACKET_TOO_BIG:
210 		ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG);
211 		return;
212 	case ICMP6_TIME_EXCEEDED:
213 		switch (code) {
214 		case ICMP6_TIME_EXCEED_TRANSIT:
215 			ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT);
216 			return;
217 		case ICMP6_TIME_EXCEED_REASSEMBLY:
218 			ICMP6_STATINC(base +
219 				      ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY);
220 			return;
221 		}
222 		break;
223 	case ICMP6_PARAM_PROB:
224 		switch (code) {
225 		case ICMP6_PARAMPROB_HEADER:
226 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER);
227 			return;
228 		case ICMP6_PARAMPROB_NEXTHEADER:
229 			ICMP6_STATINC(base +
230 				      ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER);
231 			return;
232 		case ICMP6_PARAMPROB_OPTION:
233 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION);
234 			return;
235 		}
236 		break;
237 	case ND_REDIRECT:
238 		ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT);
239 		return;
240 	}
241 	ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN);
242 }
243 
244 /*
245  * Register a Path MTU Discovery callback.
246  */
247 void
248 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *))
249 {
250 	struct icmp6_mtudisc_callback *mc;
251 
252 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
253 	     mc = LIST_NEXT(mc, mc_list)) {
254 		if (mc->mc_func == func)
255 			return;
256 	}
257 
258 	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
259 	if (mc == NULL)
260 		panic("icmp6_mtudisc_callback_register");
261 
262 	mc->mc_func = func;
263 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
264 }
265 
266 /*
267  * A wrapper function for icmp6_error() necessary when the erroneous packet
268  * may not contain enough scope zone information.
269  */
270 void
271 icmp6_error2(struct mbuf *m, int type, int code, int param,
272 	struct ifnet *ifp)
273 {
274 	struct ip6_hdr *ip6;
275 
276 	if (ifp == NULL)
277 		return;
278 
279 	if (m->m_len < sizeof(struct ip6_hdr)) {
280 		m = m_pullup(m, sizeof(struct ip6_hdr));
281 		if (m == NULL)
282 			return;
283 	}
284 
285 	ip6 = mtod(m, struct ip6_hdr *);
286 
287 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
288 		return;
289 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
290 		return;
291 
292 	icmp6_error(m, type, code, param);
293 }
294 
295 /*
296  * Generate an error packet of type error in response to bad IP6 packet.
297  */
298 void
299 icmp6_error(struct mbuf *m, int type, int code, int param)
300 {
301 	struct ip6_hdr *oip6, *nip6;
302 	struct icmp6_hdr *icmp6;
303 	u_int preplen;
304 	int off;
305 	int nxt;
306 
307 	ICMP6_STATINC(ICMP6_STAT_ERROR);
308 
309 	/* count per-type-code statistics */
310 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code);
311 
312 	if (m->m_flags & M_DECRYPTED) {
313 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
314 		goto freeit;
315 	}
316 
317 	if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) &&
318 	    (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL)
319 		return;
320 	oip6 = mtod(m, struct ip6_hdr *);
321 
322 	/*
323 	 * If the destination address of the erroneous packet is a multicast
324 	 * address, or the packet was sent using link-layer multicast,
325 	 * we should basically suppress sending an error (RFC 2463, Section
326 	 * 2.4).
327 	 * We have two exceptions (the item e.2 in that section):
328 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
329 	 * - the Parameter Problem Message that can be allowed an icmp6 error
330 	 *   in the option type field.  This check has been done in
331 	 *   ip6_unknown_opt(), so we can just check the type and code.
332 	 */
333 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
334 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
335 	    (type != ICMP6_PACKET_TOO_BIG &&
336 	     (type != ICMP6_PARAM_PROB ||
337 	      code != ICMP6_PARAMPROB_OPTION)))
338 		goto freeit;
339 
340 	/*
341 	 * RFC 2463, 2.4 (e.5): source address check.
342 	 * XXX: the case of anycast source?
343 	 */
344 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
345 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
346 		goto freeit;
347 
348 	/*
349 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
350 	 * don't do it.
351 	 */
352 	nxt = -1;
353 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
354 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
355 		struct icmp6_hdr *icp;
356 
357 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
358 			sizeof(*icp));
359 		if (icp == NULL) {
360 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
361 			return;
362 		}
363 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
364 		    icp->icmp6_type == ND_REDIRECT) {
365 			/*
366 			 * ICMPv6 error
367 			 * Special case: for redirect (which is
368 			 * informational) we must not send icmp6 error.
369 			 */
370 			ICMP6_STATINC(ICMP6_STAT_CANTERROR);
371 			goto freeit;
372 		} else {
373 			/* ICMPv6 informational - send the error */
374 		}
375 	}
376 #if 0 /* controversial */
377 	else if (off >= 0 && nxt == IPPROTO_ESP) {
378 		/*
379 		 * It could be ICMPv6 error inside ESP.  Take a safer side,
380 		 * don't respond.
381 		 */
382 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
383 		goto freeit;
384 	}
385 #endif
386 	else {
387 		/* non-ICMPv6 - send the error */
388 	}
389 
390 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
391 
392 	/* Finally, do rate limitation check. */
393 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
394 		ICMP6_STATINC(ICMP6_STAT_TOOFREQ);
395 		goto freeit;
396 	}
397 
398 	/*
399 	 * OK, ICMP6 can be generated.
400 	 */
401 
402 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
403 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
404 
405 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
406 	M_PREPEND(m, preplen, M_DONTWAIT);
407 	if (m && M_UNWRITABLE(m, preplen))
408 		m = m_pullup(m, preplen);
409 	if (m == NULL) {
410 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
411 		return;
412 	}
413 
414 	nip6 = mtod(m, struct ip6_hdr *);
415 	nip6->ip6_src  = oip6->ip6_src;
416 	nip6->ip6_dst  = oip6->ip6_dst;
417 
418 	in6_clearscope(&oip6->ip6_src);
419 	in6_clearscope(&oip6->ip6_dst);
420 
421 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
422 	icmp6->icmp6_type = type;
423 	icmp6->icmp6_code = code;
424 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
425 
426 	/*
427 	 * icmp6_reflect() is designed to be in the input path.
428 	 * icmp6_error() can be called from both input and output path,
429 	 * and if we are in output path rcvif could contain bogus value.
430 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
431 	 * information in ip header (nip6).
432 	 */
433 	m->m_pkthdr.rcvif = NULL;
434 
435 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
436 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
437 
438 	return;
439 
440   freeit:
441 	/*
442 	 * If we can't tell whether or not we can generate ICMP6, free it.
443 	 */
444 	m_freem(m);
445 }
446 
447 /*
448  * Process a received ICMP6 message.
449  */
450 int
451 icmp6_input(struct mbuf **mp, int *offp, int proto)
452 {
453 	struct mbuf *m = *mp, *n;
454 	struct ip6_hdr *ip6, *nip6;
455 	struct icmp6_hdr *icmp6, *nicmp6;
456 	int off = *offp;
457 	int icmp6len = m->m_pkthdr.len - *offp;
458 	int code, sum, noff, i;
459 
460 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
461 	KASSERT(ICMP6_MAXLEN < MCLBYTES);
462 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
463 
464 	/*
465 	 * Locate icmp6 structure in mbuf, and check
466 	 * that not corrupted and of at least minimum length
467 	 */
468 
469 	if (icmp6len < sizeof(struct icmp6_hdr)) {
470 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
471 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
472 		goto freeit;
473 	}
474 
475 	i = off + sizeof(*icmp6);
476 	if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) {
477 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
478 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
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 		code = PRC_MSGSIZE;
560 
561 		/*
562 		 * Updating the path MTU will be done after examining
563 		 * intermediate extension headers.
564 		 */
565 		goto deliver;
566 
567 	case ICMP6_TIME_EXCEEDED:
568 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
569 		switch (code) {
570 		case ICMP6_TIME_EXCEED_TRANSIT:
571 			code = PRC_TIMXCEED_INTRANS;
572 			break;
573 		case ICMP6_TIME_EXCEED_REASSEMBLY:
574 			code = PRC_TIMXCEED_REASS;
575 			break;
576 		default:
577 			goto badcode;
578 		}
579 		goto deliver;
580 
581 	case ICMP6_PARAM_PROB:
582 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
583 		switch (code) {
584 		case ICMP6_PARAMPROB_NEXTHEADER:
585 			code = PRC_UNREACH_PROTOCOL;
586 			break;
587 		case ICMP6_PARAMPROB_HEADER:
588 		case ICMP6_PARAMPROB_OPTION:
589 			code = PRC_PARAMPROB;
590 			break;
591 		default:
592 			goto badcode;
593 		}
594 		goto deliver;
595 
596 	case ICMP6_ECHO_REQUEST:
597 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
598 		if (code != 0)
599 			goto badcode;
600 		/*
601 		 * Copy mbuf to send to two data paths: userland socket(s),
602 		 * and to the querier (echo reply).
603 		 * m: a copy for socket, n: a copy for querier
604 		 *
605 		 * If the first mbuf is shared, or the first mbuf is too short,
606 		 * copy the first part of the data into a fresh mbuf.
607 		 * Otherwise, we will wrongly overwrite both copies.
608 		 */
609 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
610 			/* Give up local */
611 			n = m;
612 			m = NULL;
613 		} else if (M_READONLY(n) ||
614 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
615 			struct mbuf *n0 = n;
616 
617 			/*
618 			 * Prepare an internal mbuf.  m_pullup() doesn't
619 			 * always copy the length we specified.
620 			 */
621 			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
622 				/* Give up local */
623 				n = m;
624 				m = NULL;
625 			}
626 			m_freem(n0);
627 		}
628 		IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
629 		    sizeof(*nicmp6));
630 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
631 		nicmp6->icmp6_code = 0;
632 		if (n) {
633 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
634 			icmp6s[ICMP6_STAT_REFLECT]++;
635 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++;
636 			ICMP6_STAT_PUTREF();
637 			icmp6_reflect(n, off);
638 		}
639 		if (!m)
640 			goto freeit;
641 		break;
642 
643 	case ICMP6_ECHO_REPLY:
644 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
645 		if (code != 0)
646 			goto badcode;
647 		break;
648 
649 	case MLD_LISTENER_QUERY:
650 	case MLD_LISTENER_REPORT:
651 		if (icmp6len < sizeof(struct mld_hdr))
652 			goto badlen;
653 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
654 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
655 		else
656 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
657 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
658 			/* give up local */
659 			mld_input(m, off);
660 			m = NULL;
661 			goto freeit;
662 		}
663 		mld_input(n, off);
664 		/* m stays. */
665 		break;
666 
667 	case MLD_LISTENER_DONE:
668 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
669 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
670 			goto badlen;
671 		break;		/* nothing to be done in kernel */
672 
673 	case MLD_MTRACE_RESP:
674 	case MLD_MTRACE:
675 		/* XXX: these two are experimental.  not officially defined. */
676 		/* XXX: per-interface statistics? */
677 		break;		/* just pass it to applications */
678 
679 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
680 	    {
681 		enum { WRU, FQDN } mode;
682 
683 		if (!icmp6_nodeinfo)
684 			break;
685 
686 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
687 			mode = WRU;
688 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
689 			mode = FQDN;
690 		else
691 			goto badlen;
692 
693 		if (mode == FQDN) {
694 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
695 			if (n)
696 				n = ni6_input(n, off);
697 			/* XXX meaningless if n == NULL */
698 			noff = sizeof(struct ip6_hdr);
699 		} else {
700 			u_char *p;
701 			int maxhlen;
702 
703 			if ((icmp6_nodeinfo & 5) != 5)
704 				break;
705 
706 			if (code != 0)
707 				goto badcode;
708 			MGETHDR(n, M_DONTWAIT, m->m_type);
709 			if (n && ICMP6_MAXLEN > MHLEN) {
710 				MCLGET(n, M_DONTWAIT);
711 				if ((n->m_flags & M_EXT) == 0) {
712 					m_free(n);
713 					n = NULL;
714 				}
715 			}
716 			if (n == NULL) {
717 				/* Give up remote */
718 				break;
719 			}
720 			n->m_pkthdr.rcvif = NULL;
721 			n->m_len = 0;
722 			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
723 			if (maxhlen > hostnamelen)
724 				maxhlen = hostnamelen;
725 			/*
726 			 * Copy IPv6 and ICMPv6 only.
727 			 */
728 			nip6 = mtod(n, struct ip6_hdr *);
729 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
730 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
731 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
732 			p = (u_char *)(nicmp6 + 1);
733 			bzero(p, 4);
734 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
735 			noff = sizeof(struct ip6_hdr);
736 			M_COPY_PKTHDR(n, m); /* just for rcvif */
737 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
738 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
739 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
740 			nicmp6->icmp6_code = 0;
741 		}
742 #undef hostnamelen
743 		if (n) {
744 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
745 			icmp6s[ICMP6_STAT_REFLECT]++;
746 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++;
747 			ICMP6_STAT_PUTREF();
748 			icmp6_reflect(n, noff);
749 		}
750 		break;
751 	    }
752 
753 	case ICMP6_WRUREPLY:
754 		if (code != 0)
755 			goto badcode;
756 		break;
757 
758 	case ND_ROUTER_SOLICIT:
759 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
760 		if (code != 0)
761 			goto badcode;
762 		if (icmp6len < sizeof(struct nd_router_solicit))
763 			goto badlen;
764 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
765 			/* give up local */
766 			nd6_rs_input(m, off, icmp6len);
767 			m = NULL;
768 			goto freeit;
769 		}
770 		nd6_rs_input(n, off, icmp6len);
771 		/* m stays. */
772 		break;
773 
774 	case ND_ROUTER_ADVERT:
775 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
776 		if (code != 0)
777 			goto badcode;
778 		if (icmp6len < sizeof(struct nd_router_advert))
779 			goto badlen;
780 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
781 			/* give up local */
782 			nd6_ra_input(m, off, icmp6len);
783 			m = NULL;
784 			goto freeit;
785 		}
786 		nd6_ra_input(n, off, icmp6len);
787 		/* m stays. */
788 		break;
789 
790 	case ND_NEIGHBOR_SOLICIT:
791 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
792 		if (code != 0)
793 			goto badcode;
794 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
795 			goto badlen;
796 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
797 			/* give up local */
798 			nd6_ns_input(m, off, icmp6len);
799 			m = NULL;
800 			goto freeit;
801 		}
802 		nd6_ns_input(n, off, icmp6len);
803 		/* m stays. */
804 		break;
805 
806 	case ND_NEIGHBOR_ADVERT:
807 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
808 		if (code != 0)
809 			goto badcode;
810 		if (icmp6len < sizeof(struct nd_neighbor_advert))
811 			goto badlen;
812 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
813 			/* give up local */
814 			nd6_na_input(m, off, icmp6len);
815 			m = NULL;
816 			goto freeit;
817 		}
818 		nd6_na_input(n, off, icmp6len);
819 		/* m stays. */
820 		break;
821 
822 	case ND_REDIRECT:
823 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
824 		if (code != 0)
825 			goto badcode;
826 		if (icmp6len < sizeof(struct nd_redirect))
827 			goto badlen;
828 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
829 			/* give up local */
830 			icmp6_redirect_input(m, off);
831 			m = NULL;
832 			goto freeit;
833 		}
834 		icmp6_redirect_input(n, off);
835 		/* m stays. */
836 		break;
837 
838 	case ICMP6_ROUTER_RENUMBERING:
839 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
840 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
841 			goto badcode;
842 		if (icmp6len < sizeof(struct icmp6_router_renum))
843 			goto badlen;
844 		break;
845 
846 	default:
847 		nd6log((LOG_DEBUG,
848 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
849 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
850 		    ip6_sprintf(&ip6->ip6_dst),
851 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
852 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
853 			/* ICMPv6 error: MUST deliver it by spec... */
854 			code = PRC_NCMDS;
855 			/* deliver */
856 		} else {
857 			/* ICMPv6 informational: MUST not deliver */
858 			break;
859 		}
860 	deliver:
861 		if (icmp6_notify_error(m, off, icmp6len, code)) {
862 			/* In this case, m should've been freed. */
863 			return (IPPROTO_DONE);
864 		}
865 		break;
866 
867 	badcode:
868 		ICMP6_STATINC(ICMP6_STAT_BADCODE);
869 		break;
870 
871 	badlen:
872 		ICMP6_STATINC(ICMP6_STAT_BADLEN);
873 		break;
874 	}
875 
876 	/* deliver the packet to appropriate sockets */
877 	icmp6_rip6_input(&m, *offp);
878 
879 	return IPPROTO_DONE;
880 
881  freeit:
882 	m_freem(m);
883 	return IPPROTO_DONE;
884 }
885 
886 static int
887 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
888 {
889 	struct icmp6_hdr *icmp6;
890 	struct ip6_hdr *eip6;
891 	u_int32_t notifymtu;
892 	struct sockaddr_in6 icmp6src, icmp6dst;
893 
894 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
895 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
896 		goto freeit;
897 	}
898 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
899 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
900 	if (icmp6 == NULL) {
901 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
902 		return (-1);
903 	}
904 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
905 
906 	/* Detect the upper level protocol */
907 	{
908 		void (*ctlfunc)(int, struct sockaddr *, void *);
909 		u_int8_t nxt = eip6->ip6_nxt;
910 		int eoff = off + sizeof(struct icmp6_hdr) +
911 			sizeof(struct ip6_hdr);
912 		struct ip6ctlparam ip6cp;
913 		struct in6_addr *finaldst = NULL;
914 		int icmp6type = icmp6->icmp6_type;
915 		struct ip6_frag *fh;
916 		struct ip6_rthdr *rth;
917 		struct ip6_rthdr0 *rth0;
918 		int rthlen;
919 
920 		while (1) { /* XXX: should avoid infinite loop explicitly? */
921 			struct ip6_ext *eh;
922 
923 			switch (nxt) {
924 			case IPPROTO_HOPOPTS:
925 			case IPPROTO_DSTOPTS:
926 			case IPPROTO_AH:
927 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
928 					       eoff, sizeof(*eh));
929 				if (eh == NULL) {
930 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
931 					return (-1);
932 				}
933 
934 				if (nxt == IPPROTO_AH)
935 					eoff += (eh->ip6e_len + 2) << 2;
936 				else
937 					eoff += (eh->ip6e_len + 1) << 3;
938 				nxt = eh->ip6e_nxt;
939 				break;
940 			case IPPROTO_ROUTING:
941 				/*
942 				 * When the erroneous packet contains a
943 				 * routing header, we should examine the
944 				 * header to determine the final destination.
945 				 * Otherwise, we can't properly update
946 				 * information that depends on the final
947 				 * destination (e.g. path MTU).
948 				 */
949 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
950 					       eoff, sizeof(*rth));
951 				if (rth == NULL) {
952 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
953 					return (-1);
954 				}
955 				rthlen = (rth->ip6r_len + 1) << 3;
956 				/*
957 				 * XXX: currently there is no
958 				 * officially defined type other
959 				 * than type-0.
960 				 * Note that if the segment left field
961 				 * is 0, all intermediate hops must
962 				 * have been passed.
963 				 */
964 				if (rth->ip6r_segleft &&
965 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
966 					int hops;
967 
968 					IP6_EXTHDR_GET(rth0,
969 						       struct ip6_rthdr0 *, m,
970 						       eoff, rthlen);
971 					if (rth0 == NULL) {
972 						ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
973 						return (-1);
974 					}
975 					/* just ignore a bogus header */
976 					if ((rth0->ip6r0_len % 2) == 0 &&
977 					    (hops = rth0->ip6r0_len/2))
978 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
979 				}
980 				eoff += rthlen;
981 				nxt = rth->ip6r_nxt;
982 				break;
983 			case IPPROTO_FRAGMENT:
984 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
985 					       eoff, sizeof(*fh));
986 				if (fh == NULL) {
987 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
988 					return (-1);
989 				}
990 				/*
991 				 * Data after a fragment header is meaningless
992 				 * unless it is the first fragment, but
993 				 * we'll go to the notify label for path MTU
994 				 * discovery.
995 				 */
996 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
997 					goto notify;
998 
999 				eoff += sizeof(struct ip6_frag);
1000 				nxt = fh->ip6f_nxt;
1001 				break;
1002 			default:
1003 				/*
1004 				 * This case includes ESP and the No Next
1005 				 * Header.  In such cases going to the notify
1006 				 * label does not have any meaning
1007 				 * (i.e. ctlfunc will be NULL), but we go
1008 				 * anyway since we might have to update
1009 				 * path MTU information.
1010 				 */
1011 				goto notify;
1012 			}
1013 		}
1014 	  notify:
1015 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1016 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1017 		if (icmp6 == NULL) {
1018 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1019 			return (-1);
1020 		}
1021 
1022 		/*
1023 		 * retrieve parameters from the inner IPv6 header, and convert
1024 		 * them into sockaddr structures.
1025 		 * XXX: there is no guarantee that the source or destination
1026 		 * addresses of the inner packet are in the same scope zone as
1027 		 * the addresses of the icmp packet.  But there is no other
1028 		 * way to determine the zone.
1029 		 */
1030 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1031 
1032 		sockaddr_in6_init(&icmp6dst,
1033 		    (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0);
1034 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1035 			goto freeit;
1036 		sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0);
1037 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1038 			goto freeit;
1039 		icmp6src.sin6_flowinfo =
1040 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1041 
1042 		if (finaldst == NULL)
1043 			finaldst = &eip6->ip6_dst;
1044 		ip6cp.ip6c_m = m;
1045 		ip6cp.ip6c_icmp6 = icmp6;
1046 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1047 		ip6cp.ip6c_off = eoff;
1048 		ip6cp.ip6c_finaldst = finaldst;
1049 		ip6cp.ip6c_src = &icmp6src;
1050 		ip6cp.ip6c_nxt = nxt;
1051 
1052 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1053 			notifymtu = ntohl(icmp6->icmp6_mtu);
1054 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1055 		}
1056 
1057 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1058 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1059 		if (ctlfunc) {
1060 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1061 					  &ip6cp);
1062 		}
1063 	}
1064 	return (0);
1065 
1066   freeit:
1067 	m_freem(m);
1068 	return (-1);
1069 }
1070 
1071 void
1072 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1073 {
1074 	unsigned long rtcount;
1075 	struct icmp6_mtudisc_callback *mc;
1076 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1077 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1078 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1079 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1080 	struct rtentry *rt = NULL;
1081 	struct sockaddr_in6 sin6;
1082 
1083 	/*
1084 	 * allow non-validated cases if memory is plenty, to make traffic
1085 	 * from non-connected pcb happy.
1086 	 */
1087 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1088 	if (validated) {
1089 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1090 			return;
1091 		else if (0 <= icmp6_mtudisc_lowat &&
1092 		    rtcount > icmp6_mtudisc_lowat) {
1093 			/*
1094 			 * XXX nuke a victim, install the new one.
1095 			 */
1096 		}
1097 	} else {
1098 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1099 			return;
1100 	}
1101 
1102 	bzero(&sin6, sizeof(sin6));
1103 	sin6.sin6_family = PF_INET6;
1104 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1105 	sin6.sin6_addr = *dst;
1106 	if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL))
1107 		return;
1108 
1109 	rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1110 
1111 	if (rt && (rt->rt_flags & RTF_HOST) &&
1112 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1113 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1114 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1115 			ICMP6_STATINC(ICMP6_STAT_PMTUCHG);
1116 			rt->rt_rmx.rmx_mtu = mtu;
1117 		}
1118 	}
1119 	if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1120 		RTFREE(rt);
1121 	}
1122 
1123 	/*
1124 	 * Notify protocols that the MTU for this destination
1125 	 * has changed.
1126 	 */
1127 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1128 	     mc = LIST_NEXT(mc, mc_list))
1129 		(*mc->mc_func)(&sin6.sin6_addr);
1130 }
1131 
1132 /*
1133  * Process a Node Information Query packet, based on
1134  * draft-ietf-ipngwg-icmp-name-lookups-07.
1135  *
1136  * Spec incompatibilities:
1137  * - IPv6 Subject address handling
1138  * - IPv4 Subject address handling support missing
1139  * - Proxy reply (answer even if it's not for me)
1140  * - joins NI group address at in6_ifattach() time only, does not cope
1141  *   with hostname changes by sethostname(3)
1142  */
1143 #ifndef offsetof		/* XXX */
1144 #define	offsetof(type, member)	((size_t)(&((type *)0)->member))
1145 #endif
1146 static struct mbuf *
1147 ni6_input(struct mbuf *m, int off)
1148 {
1149 	struct icmp6_nodeinfo *ni6, *nni6;
1150 	struct mbuf *n = NULL;
1151 	u_int16_t qtype;
1152 	int subjlen;
1153 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1154 	struct ni_reply_fqdn *fqdn;
1155 	int addrs;		/* for NI_QTYPE_NODEADDR */
1156 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1157 	struct sockaddr_in6 sin6; /* ip6_dst */
1158 	struct in6_addr in6_subj; /* subject address */
1159 	struct ip6_hdr *ip6;
1160 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1161 	char *subj = NULL;
1162 
1163 	ip6 = mtod(m, struct ip6_hdr *);
1164 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1165 	if (ni6 == NULL) {
1166 		/* m is already reclaimed */
1167 		return NULL;
1168 	}
1169 
1170 	/*
1171 	 * Validate IPv6 destination address.
1172 	 *
1173 	 * The Responder must discard the Query without further processing
1174 	 * unless it is one of the Responder's unicast or anycast addresses, or
1175 	 * a link-local scope multicast address which the Responder has joined.
1176 	 * [icmp-name-lookups-07, Section 4.]
1177 	 */
1178 	sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
1179 	/* XXX scopeid */
1180 	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1181 		; /* unicast/anycast, fine */
1182 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1183 		; /* link-local multicast, fine */
1184 	else
1185 		goto bad;
1186 
1187 	/* validate query Subject field. */
1188 	qtype = ntohs(ni6->ni_qtype);
1189 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1190 	switch (qtype) {
1191 	case NI_QTYPE_NOOP:
1192 	case NI_QTYPE_SUPTYPES:
1193 		/* 07 draft */
1194 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1195 			break;
1196 		/* FALLTHROUGH */
1197 	case NI_QTYPE_FQDN:
1198 	case NI_QTYPE_NODEADDR:
1199 	case NI_QTYPE_IPV4ADDR:
1200 		switch (ni6->ni_code) {
1201 		case ICMP6_NI_SUBJ_IPV6:
1202 #if ICMP6_NI_SUBJ_IPV6 != 0
1203 		case 0:
1204 #endif
1205 			/*
1206 			 * backward compatibility - try to accept 03 draft
1207 			 * format, where no Subject is present.
1208 			 */
1209 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1210 			    subjlen == 0) {
1211 				oldfqdn++;
1212 				break;
1213 			}
1214 #if ICMP6_NI_SUBJ_IPV6 != 0
1215 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1216 				goto bad;
1217 #endif
1218 
1219 			if (subjlen != sizeof(sin6.sin6_addr))
1220 				goto bad;
1221 
1222 			/*
1223 			 * Validate Subject address.
1224 			 *
1225 			 * Not sure what exactly "address belongs to the node"
1226 			 * means in the spec, is it just unicast, or what?
1227 			 *
1228 			 * At this moment we consider Subject address as
1229 			 * "belong to the node" if the Subject address equals
1230 			 * to the IPv6 destination address; validation for
1231 			 * IPv6 destination address should have done enough
1232 			 * check for us.
1233 			 *
1234 			 * We do not do proxy at this moment.
1235 			 */
1236 			/* m_pulldown instead of copy? */
1237 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1238 			    subjlen, (void *)&in6_subj);
1239 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1240 				goto bad;
1241 
1242 			subj = (char *)&in6_subj;
1243 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1244 				break;
1245 
1246 			/*
1247 			 * XXX if we are to allow other cases, we should really
1248 			 * be careful about scope here.
1249 			 * basically, we should disallow queries toward IPv6
1250 			 * destination X with subject Y, if scope(X) > scope(Y).
1251 			 * if we allow scope(X) > scope(Y), it will result in
1252 			 * information leakage across scope boundary.
1253 			 */
1254 			goto bad;
1255 
1256 		case ICMP6_NI_SUBJ_FQDN:
1257 			/*
1258 			 * Validate Subject name with gethostname(3).
1259 			 *
1260 			 * The behavior may need some debate, since:
1261 			 * - we are not sure if the node has FQDN as
1262 			 *   hostname (returned by gethostname(3)).
1263 			 * - the code does wildcard match for truncated names.
1264 			 *   however, we are not sure if we want to perform
1265 			 *   wildcard match, if gethostname(3) side has
1266 			 *   truncated hostname.
1267 			 */
1268 			n = ni6_nametodns(hostname, hostnamelen, 0);
1269 			if (!n || n->m_next || n->m_len == 0)
1270 				goto bad;
1271 			IP6_EXTHDR_GET(subj, char *, m,
1272 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1273 			if (subj == NULL)
1274 				goto bad;
1275 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1276 					n->m_len)) {
1277 				goto bad;
1278 			}
1279 			m_freem(n);
1280 			n = NULL;
1281 			break;
1282 
1283 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1284 		default:
1285 			goto bad;
1286 		}
1287 		break;
1288 	}
1289 
1290 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1291 	switch (qtype) {
1292 	case NI_QTYPE_FQDN:
1293 		if ((icmp6_nodeinfo & 1) == 0)
1294 			goto bad;
1295 		break;
1296 	case NI_QTYPE_NODEADDR:
1297 	case NI_QTYPE_IPV4ADDR:
1298 		if ((icmp6_nodeinfo & 2) == 0)
1299 			goto bad;
1300 		break;
1301 	}
1302 
1303 	/* guess reply length */
1304 	switch (qtype) {
1305 	case NI_QTYPE_NOOP:
1306 		break;		/* no reply data */
1307 	case NI_QTYPE_SUPTYPES:
1308 		replylen += sizeof(u_int32_t);
1309 		break;
1310 	case NI_QTYPE_FQDN:
1311 		/* XXX will append an mbuf */
1312 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1313 		break;
1314 	case NI_QTYPE_NODEADDR:
1315 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1316 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1317 					  sizeof(u_int32_t))) > MCLBYTES)
1318 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1319 		break;
1320 	case NI_QTYPE_IPV4ADDR:
1321 		/* unsupported - should respond with unknown Qtype? */
1322 		goto bad;
1323 	default:
1324 		/*
1325 		 * XXX: We must return a reply with the ICMP6 code
1326 		 * `unknown Qtype' in this case.  However we regard the case
1327 		 * as an FQDN query for backward compatibility.
1328 		 * Older versions set a random value to this field,
1329 		 * so it rarely varies in the defined qtypes.
1330 		 * But the mechanism is not reliable...
1331 		 * maybe we should obsolete older versions.
1332 		 */
1333 		qtype = NI_QTYPE_FQDN;
1334 		/* XXX will append an mbuf */
1335 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1336 		oldfqdn++;
1337 		break;
1338 	}
1339 
1340 	/* allocate an mbuf to reply. */
1341 	MGETHDR(n, M_DONTWAIT, m->m_type);
1342 	if (n == NULL) {
1343 		m_freem(m);
1344 		return (NULL);
1345 	}
1346 	M_MOVE_PKTHDR(n, m); /* just for rcvif */
1347 	if (replylen > MHLEN) {
1348 		if (replylen > MCLBYTES) {
1349 			/*
1350 			 * XXX: should we try to allocate more? But MCLBYTES
1351 			 * is probably much larger than IPV6_MMTU...
1352 			 */
1353 			goto bad;
1354 		}
1355 		MCLGET(n, M_DONTWAIT);
1356 		if ((n->m_flags & M_EXT) == 0) {
1357 			goto bad;
1358 		}
1359 	}
1360 	n->m_pkthdr.len = n->m_len = replylen;
1361 
1362 	/* copy mbuf header and IPv6 + Node Information base headers */
1363 	bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr));
1364 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1365 	bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo));
1366 
1367 	/* qtype dependent procedure */
1368 	switch (qtype) {
1369 	case NI_QTYPE_NOOP:
1370 		nni6->ni_code = ICMP6_NI_SUCCESS;
1371 		nni6->ni_flags = 0;
1372 		break;
1373 	case NI_QTYPE_SUPTYPES:
1374 	{
1375 		u_int32_t v;
1376 		nni6->ni_code = ICMP6_NI_SUCCESS;
1377 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1378 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1379 		v = (u_int32_t)htonl(0x0000000f);
1380 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1381 		break;
1382 	}
1383 	case NI_QTYPE_FQDN:
1384 		nni6->ni_code = ICMP6_NI_SUCCESS;
1385 		fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) +
1386 						sizeof(struct ip6_hdr) +
1387 						sizeof(struct icmp6_nodeinfo));
1388 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1389 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1390 		/*
1391 		 * XXX do we really have FQDN in variable "hostname"?
1392 		 */
1393 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1394 		if (n->m_next == NULL)
1395 			goto bad;
1396 		/* XXX we assume that n->m_next is not a chain */
1397 		if (n->m_next->m_next != NULL)
1398 			goto bad;
1399 		n->m_pkthdr.len += n->m_next->m_len;
1400 		break;
1401 	case NI_QTYPE_NODEADDR:
1402 	{
1403 		int lenlim, copied;
1404 
1405 		nni6->ni_code = ICMP6_NI_SUCCESS;
1406 		n->m_pkthdr.len = n->m_len =
1407 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1408 		lenlim = M_TRAILINGSPACE(n);
1409 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1410 		/* XXX: reset mbuf length */
1411 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1412 			sizeof(struct icmp6_nodeinfo) + copied;
1413 		break;
1414 	}
1415 	default:
1416 		break;		/* XXX impossible! */
1417 	}
1418 
1419 	nni6->ni_type = ICMP6_NI_REPLY;
1420 	m_freem(m);
1421 	return (n);
1422 
1423   bad:
1424 	m_freem(m);
1425 	if (n)
1426 		m_freem(n);
1427 	return (NULL);
1428 }
1429 #undef hostnamelen
1430 
1431 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1432 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1433 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1434 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1435 
1436 /*
1437  * make a mbuf with DNS-encoded string.  no compression support.
1438  *
1439  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1440  * treated as truncated name (two \0 at the end).  this is a wild guess.
1441  *
1442  * old - return pascal string if non-zero
1443  */
1444 static struct mbuf *
1445 ni6_nametodns(const char *name, int namelen, int old)
1446 {
1447 	struct mbuf *m;
1448 	char *cp, *ep;
1449 	const char *p, *q;
1450 	int i, len, nterm;
1451 
1452 	if (old)
1453 		len = namelen + 1;
1454 	else
1455 		len = MCLBYTES;
1456 
1457 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1458 	MGET(m, M_DONTWAIT, MT_DATA);
1459 	if (m && len > MLEN) {
1460 		MCLGET(m, M_DONTWAIT);
1461 		if ((m->m_flags & M_EXT) == 0)
1462 			goto fail;
1463 	}
1464 	if (!m)
1465 		goto fail;
1466 	m->m_next = NULL;
1467 
1468 	if (old) {
1469 		m->m_len = len;
1470 		*mtod(m, char *) = namelen;
1471 		bcopy(name, mtod(m, char *) + 1, namelen);
1472 		return m;
1473 	} else {
1474 		m->m_len = 0;
1475 		cp = mtod(m, char *);
1476 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1477 
1478 		/* if not certain about my name, return empty buffer */
1479 		if (namelen == 0)
1480 			return m;
1481 
1482 		/*
1483 		 * guess if it looks like shortened hostname, or FQDN.
1484 		 * shortened hostname needs two trailing "\0".
1485 		 */
1486 		i = 0;
1487 		for (p = name; p < name + namelen; p++) {
1488 			if (*p && *p == '.')
1489 				i++;
1490 		}
1491 		if (i < 2)
1492 			nterm = 2;
1493 		else
1494 			nterm = 1;
1495 
1496 		p = name;
1497 		while (cp < ep && p < name + namelen) {
1498 			i = 0;
1499 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1500 				i++;
1501 			/* result does not fit into mbuf */
1502 			if (cp + i + 1 >= ep)
1503 				goto fail;
1504 			/*
1505 			 * DNS label length restriction, RFC1035 page 8.
1506 			 * "i == 0" case is included here to avoid returning
1507 			 * 0-length label on "foo..bar".
1508 			 */
1509 			if (i <= 0 || i >= 64)
1510 				goto fail;
1511 			*cp++ = i;
1512 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1513 				goto fail;
1514 			while (i > 0) {
1515 				if (!isalnum(*p) && *p != '-')
1516 					goto fail;
1517 				if (isupper(*p)) {
1518 					*cp++ = tolower(*p);
1519 					p++;
1520 				} else
1521 					*cp++ = *p++;
1522 				i--;
1523 			}
1524 			p = q;
1525 			if (p < name + namelen && *p == '.')
1526 				p++;
1527 		}
1528 		/* termination */
1529 		if (cp + nterm >= ep)
1530 			goto fail;
1531 		while (nterm-- > 0)
1532 			*cp++ = '\0';
1533 		m->m_len = cp - mtod(m, char *);
1534 		return m;
1535 	}
1536 
1537 	panic("should not reach here");
1538 	/* NOTREACHED */
1539 
1540  fail:
1541 	if (m)
1542 		m_freem(m);
1543 	return NULL;
1544 }
1545 
1546 /*
1547  * check if two DNS-encoded string matches.  takes care of truncated
1548  * form (with \0\0 at the end).  no compression support.
1549  * XXX upper/lowercase match (see RFC2065)
1550  */
1551 static int
1552 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1553 {
1554 	const char *a0, *b0;
1555 	int l;
1556 
1557 	/* simplest case - need validation? */
1558 	if (alen == blen && bcmp(a, b, alen) == 0)
1559 		return 1;
1560 
1561 	a0 = a;
1562 	b0 = b;
1563 
1564 	/* termination is mandatory */
1565 	if (alen < 2 || blen < 2)
1566 		return 0;
1567 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1568 		return 0;
1569 	alen--;
1570 	blen--;
1571 
1572 	while (a - a0 < alen && b - b0 < blen) {
1573 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1574 			return 0;
1575 
1576 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1577 			return 0;
1578 		/* we don't support compression yet */
1579 		if (a[0] >= 64 || b[0] >= 64)
1580 			return 0;
1581 
1582 		/* truncated case */
1583 		if (a[0] == 0 && a - a0 == alen - 1)
1584 			return 1;
1585 		if (b[0] == 0 && b - b0 == blen - 1)
1586 			return 1;
1587 		if (a[0] == 0 || b[0] == 0)
1588 			return 0;
1589 
1590 		if (a[0] != b[0])
1591 			return 0;
1592 		l = a[0];
1593 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1594 			return 0;
1595 		if (bcmp(a + 1, b + 1, l) != 0)
1596 			return 0;
1597 
1598 		a += 1 + l;
1599 		b += 1 + l;
1600 	}
1601 
1602 	if (a - a0 == alen && b - b0 == blen)
1603 		return 1;
1604 	else
1605 		return 0;
1606 }
1607 
1608 /*
1609  * calculate the number of addresses to be returned in the node info reply.
1610  */
1611 static int
1612 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m,
1613     struct ifnet **ifpp, char *subj)
1614 {
1615 	struct ifnet *ifp;
1616 	struct in6_ifaddr *ifa6;
1617 	struct ifaddr *ifa;
1618 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1619 	int addrs = 0, addrsofif, iffound = 0;
1620 	int niflags = ni6->ni_flags;
1621 
1622 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1623 		switch (ni6->ni_code) {
1624 		case ICMP6_NI_SUBJ_IPV6:
1625 			if (subj == NULL) /* must be impossible... */
1626 				return (0);
1627 			subj_ip6 = (struct sockaddr_in6 *)subj;
1628 			break;
1629 		default:
1630 			/*
1631 			 * XXX: we only support IPv6 subject address for
1632 			 * this Qtype.
1633 			 */
1634 			return (0);
1635 		}
1636 	}
1637 
1638 	IFNET_FOREACH(ifp) {
1639 		addrsofif = 0;
1640 		IFADDR_FOREACH(ifa, ifp) {
1641 			if (ifa->ifa_addr->sa_family != AF_INET6)
1642 				continue;
1643 			ifa6 = (struct in6_ifaddr *)ifa;
1644 
1645 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1646 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1647 					       &ifa6->ia_addr.sin6_addr))
1648 				iffound = 1;
1649 
1650 			/*
1651 			 * IPv4-mapped addresses can only be returned by a
1652 			 * Node Information proxy, since they represent
1653 			 * addresses of IPv4-only nodes, which perforce do
1654 			 * not implement this protocol.
1655 			 * [icmp-name-lookups-07, Section 5.4]
1656 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1657 			 * this function at this moment.
1658 			 */
1659 
1660 			/* What do we have to do about ::1? */
1661 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1662 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1663 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1664 					continue;
1665 				break;
1666 			case IPV6_ADDR_SCOPE_SITELOCAL:
1667 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1668 					continue;
1669 				break;
1670 			case IPV6_ADDR_SCOPE_GLOBAL:
1671 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1672 					continue;
1673 				break;
1674 			default:
1675 				continue;
1676 			}
1677 
1678 			/*
1679 			 * check if anycast is okay.
1680 			 * XXX: just experimental.  not in the spec.
1681 			 */
1682 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1683 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1684 				continue; /* we need only unicast addresses */
1685 
1686 			addrsofif++; /* count the address */
1687 		}
1688 		if (iffound) {
1689 			*ifpp = ifp;
1690 			return (addrsofif);
1691 		}
1692 
1693 		addrs += addrsofif;
1694 	}
1695 
1696 	return (addrs);
1697 }
1698 
1699 static int
1700 ni6_store_addrs(struct icmp6_nodeinfo *ni6,
1701 	struct icmp6_nodeinfo *nni6, struct ifnet *ifp0,
1702 	int resid)
1703 {
1704 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1705 	struct in6_ifaddr *ifa6;
1706 	struct ifaddr *ifa;
1707 	struct ifnet *ifp_dep = NULL;
1708 	int copied = 0, allow_deprecated = 0;
1709 	u_char *cp = (u_char *)(nni6 + 1);
1710 	int niflags = ni6->ni_flags;
1711 	u_int32_t ltime;
1712 
1713 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1714 		return (0);	/* needless to copy */
1715 
1716   again:
1717 
1718 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1719 	{
1720 		IFADDR_FOREACH(ifa, ifp) {
1721 			if (ifa->ifa_addr->sa_family != AF_INET6)
1722 				continue;
1723 			ifa6 = (struct in6_ifaddr *)ifa;
1724 
1725 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1726 			    allow_deprecated == 0) {
1727 				/*
1728 				 * prefererred address should be put before
1729 				 * deprecated addresses.
1730 				 */
1731 
1732 				/* record the interface for later search */
1733 				if (ifp_dep == NULL)
1734 					ifp_dep = ifp;
1735 
1736 				continue;
1737 			}
1738 			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1739 				 allow_deprecated != 0)
1740 				continue; /* we now collect deprecated addrs */
1741 
1742 			/* What do we have to do about ::1? */
1743 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1744 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1745 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1746 					continue;
1747 				break;
1748 			case IPV6_ADDR_SCOPE_SITELOCAL:
1749 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1750 					continue;
1751 				break;
1752 			case IPV6_ADDR_SCOPE_GLOBAL:
1753 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1754 					continue;
1755 				break;
1756 			default:
1757 				continue;
1758 			}
1759 
1760 			/*
1761 			 * check if anycast is okay.
1762 			 * XXX: just experimental.  not in the spec.
1763 			 */
1764 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1765 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1766 				continue;
1767 
1768 			/* now we can copy the address */
1769 			if (resid < sizeof(struct in6_addr) +
1770 			    sizeof(u_int32_t)) {
1771 				/*
1772 				 * We give up much more copy.
1773 				 * Set the truncate flag and return.
1774 				 */
1775 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1776 				return (copied);
1777 			}
1778 
1779 			/*
1780 			 * Set the TTL of the address.
1781 			 * The TTL value should be one of the following
1782 			 * according to the specification:
1783 			 *
1784 			 * 1. The remaining lifetime of a DHCP lease on the
1785 			 *    address, or
1786 			 * 2. The remaining Valid Lifetime of a prefix from
1787 			 *    which the address was derived through Stateless
1788 			 *    Autoconfiguration.
1789 			 *
1790 			 * Note that we currently do not support stateful
1791 			 * address configuration by DHCPv6, so the former
1792 			 * case can't happen.
1793 			 *
1794 			 * TTL must be 2^31 > TTL >= 0.
1795 			 */
1796 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1797 				ltime = ND6_INFINITE_LIFETIME;
1798 			else {
1799 				if (ifa6->ia6_lifetime.ia6t_expire >
1800 				    time_second)
1801 					ltime = ifa6->ia6_lifetime.ia6t_expire -
1802 					    time_second;
1803 				else
1804 					ltime = 0;
1805 			}
1806 			if (ltime > 0x7fffffff)
1807 				ltime = 0x7fffffff;
1808 			ltime = htonl(ltime);
1809 
1810 			bcopy(&ltime, cp, sizeof(u_int32_t));
1811 			cp += sizeof(u_int32_t);
1812 
1813 			/* copy the address itself */
1814 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1815 			      sizeof(struct in6_addr));
1816 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1817 			cp += sizeof(struct in6_addr);
1818 
1819 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1820 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1821 		}
1822 		if (ifp0)	/* we need search only on the specified IF */
1823 			break;
1824 	}
1825 
1826 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1827 		ifp = ifp_dep;
1828 		allow_deprecated = 1;
1829 
1830 		goto again;
1831 	}
1832 
1833 	return (copied);
1834 }
1835 
1836 /*
1837  * XXX almost dup'ed code with rip6_input.
1838  */
1839 static int
1840 icmp6_rip6_input(struct mbuf **mp, int off)
1841 {
1842 	struct mbuf *m = *mp;
1843 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1844 	struct inpcb_hdr *inph;
1845 	struct in6pcb *in6p;
1846 	struct in6pcb *last = NULL;
1847 	struct sockaddr_in6 rip6src;
1848 	struct icmp6_hdr *icmp6;
1849 	struct mbuf *opts = NULL;
1850 
1851 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1852 	if (icmp6 == NULL) {
1853 		/* m is already reclaimed */
1854 		return IPPROTO_DONE;
1855 	}
1856 
1857 	/*
1858 	 * XXX: the address may have embedded scope zone ID, which should be
1859 	 * hidden from applications.
1860 	 */
1861 	sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0);
1862 	if (sa6_recoverscope(&rip6src)) {
1863 		m_freem(m);
1864 		return (IPPROTO_DONE);
1865 	}
1866 
1867 	CIRCLEQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1868 		in6p = (struct in6pcb *)inph;
1869 		if (in6p->in6p_af != AF_INET6)
1870 			continue;
1871 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1872 			continue;
1873 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1874 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1875 			continue;
1876 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1877 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1878 			continue;
1879 		if (in6p->in6p_icmp6filt
1880 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1881 				 in6p->in6p_icmp6filt))
1882 			continue;
1883 		if (last) {
1884 			struct	mbuf *n;
1885 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1886 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1887 					ip6_savecontrol(last, &opts, ip6, n);
1888 				/* strip intermediate headers */
1889 				m_adj(n, off);
1890 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1891 						 (struct sockaddr *)&rip6src,
1892 						 n, opts) == 0) {
1893 					/* should notify about lost packet */
1894 					m_freem(n);
1895 					if (opts)
1896 						m_freem(opts);
1897 				} else
1898 					sorwakeup(last->in6p_socket);
1899 				opts = NULL;
1900 			}
1901 		}
1902 		last = in6p;
1903 	}
1904 	if (last) {
1905 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1906 			ip6_savecontrol(last, &opts, ip6, m);
1907 		/* strip intermediate headers */
1908 		m_adj(m, off);
1909 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1910 				(struct sockaddr *)&rip6src, m, opts) == 0) {
1911 			m_freem(m);
1912 			if (opts)
1913 				m_freem(opts);
1914 		} else
1915 			sorwakeup(last->in6p_socket);
1916 	} else {
1917 		m_freem(m);
1918 		IP6_STATDEC(IP6_STAT_DELIVERED);
1919 	}
1920 	return IPPROTO_DONE;
1921 }
1922 
1923 /*
1924  * Reflect the ip6 packet back to the source.
1925  * OFF points to the icmp6 header, counted from the top of the mbuf.
1926  *
1927  * Note: RFC 1885 required that an echo reply should be truncated if it
1928  * did not fit in with (return) path MTU, and KAME code supported the
1929  * behavior.  However, as a clarification after the RFC, this limitation
1930  * was removed in a revised version of the spec, RFC 2463.  We had kept the
1931  * old behavior, with a (non-default) ifdef block, while the new version of
1932  * the spec was an internet-draft status, and even after the new RFC was
1933  * published.  But it would rather make sense to clean the obsoleted part
1934  * up, and to make the code simpler at this stage.
1935  */
1936 void
1937 icmp6_reflect(struct mbuf *m, size_t off)
1938 {
1939 	struct ip6_hdr *ip6;
1940 	struct icmp6_hdr *icmp6;
1941 	const struct in6_ifaddr *ia;
1942 	const struct ip6aux *ip6a;
1943 	int plen;
1944 	int type, code;
1945 	struct ifnet *outif = NULL;
1946 	struct in6_addr origdst;
1947 	const struct in6_addr *src = NULL;
1948 
1949 	/* too short to reflect */
1950 	if (off < sizeof(struct ip6_hdr)) {
1951 		nd6log((LOG_DEBUG,
1952 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1953 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1954 		    __FILE__, __LINE__));
1955 		goto bad;
1956 	}
1957 
1958 	/*
1959 	 * If there are extra headers between IPv6 and ICMPv6, strip
1960 	 * off that header first.
1961 	 */
1962 #ifdef DIAGNOSTIC
1963 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1964 		panic("assumption failed in icmp6_reflect");
1965 #endif
1966 	if (off > sizeof(struct ip6_hdr)) {
1967 		size_t l;
1968 		struct ip6_hdr nip6;
1969 
1970 		l = off - sizeof(struct ip6_hdr);
1971 		m_copydata(m, 0, sizeof(nip6), (void *)&nip6);
1972 		m_adj(m, l);
1973 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1974 		if (m->m_len < l) {
1975 			if ((m = m_pullup(m, l)) == NULL)
1976 				return;
1977 		}
1978 		bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6));
1979 	} else /* off == sizeof(struct ip6_hdr) */ {
1980 		size_t l;
1981 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1982 		if (m->m_len < l) {
1983 			if ((m = m_pullup(m, l)) == NULL)
1984 				return;
1985 		}
1986 	}
1987 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1988 	ip6 = mtod(m, struct ip6_hdr *);
1989 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1990 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1991 	type = icmp6->icmp6_type; /* keep type for statistics */
1992 	code = icmp6->icmp6_code; /* ditto. */
1993 
1994 	origdst = ip6->ip6_dst;
1995 	/*
1996 	 * ip6_input() drops a packet if its src is multicast.
1997 	 * So, the src is never multicast.
1998 	 */
1999 	ip6->ip6_dst = ip6->ip6_src;
2000 
2001 	/*
2002 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2003 	 * use dst as the src for the reply.
2004 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2005 	 * (for example) when we encounter an error while forwarding procedure
2006 	 * destined to a duplicated address of ours.
2007 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2008 	 * procedure of an outgoing packet of our own, in which case we need
2009 	 * to search in the ifaddr list.
2010 	 */
2011 	if (IN6_IS_ADDR_MULTICAST(&origdst))
2012 		;
2013 	else if ((ip6a = ip6_getdstifaddr(m)) != NULL) {
2014 		if ((ip6a->ip6a_flags &
2015 		     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0)
2016 			src = &ip6a->ip6a_src;
2017 	} else {
2018 		union {
2019 			struct sockaddr_in6 sin6;
2020 			struct sockaddr sa;
2021 		} u;
2022 
2023 		sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0);
2024 
2025 		ia = (struct in6_ifaddr *)ifa_ifwithaddr(&u.sa);
2026 
2027 		if (ia == NULL)
2028 			;
2029 		else if ((ia->ia6_flags &
2030 			 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0)
2031 			src = &ia->ia_addr.sin6_addr;
2032 	}
2033 
2034 	if (src == NULL) {
2035 		int e;
2036 		struct sockaddr_in6 sin6;
2037 		struct route ro;
2038 
2039 		/*
2040 		 * This case matches to multicasts, our anycast, or unicasts
2041 		 * that we do not own.  Select a source address based on the
2042 		 * source address of the erroneous packet.
2043 		 */
2044 		/* zone ID should be embedded */
2045 		sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
2046 
2047 		memset(&ro, 0, sizeof(ro));
2048 		src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e);
2049 		rtcache_free(&ro);
2050 		if (src == NULL) {
2051 			nd6log((LOG_DEBUG,
2052 			    "icmp6_reflect: source can't be determined: "
2053 			    "dst=%s, error=%d\n",
2054 			    ip6_sprintf(&sin6.sin6_addr), e));
2055 			goto bad;
2056 		}
2057 	}
2058 
2059 	ip6->ip6_src = *src;
2060 	ip6->ip6_flow = 0;
2061 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2062 	ip6->ip6_vfc |= IPV6_VERSION;
2063 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2064 	if (m->m_pkthdr.rcvif) {
2065 		/* XXX: This may not be the outgoing interface */
2066 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2067 	} else
2068 		ip6->ip6_hlim = ip6_defhlim;
2069 
2070 	m->m_pkthdr.csum_flags = 0;
2071 	icmp6->icmp6_cksum = 0;
2072 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2073 					sizeof(struct ip6_hdr), plen);
2074 
2075 	/*
2076 	 * XXX option handling
2077 	 */
2078 
2079 	m->m_flags &= ~(M_BCAST|M_MCAST);
2080 
2081 	/*
2082 	 * To avoid a "too big" situation at an intermediate router
2083 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2084 	 * Note that only echo and node information replies are affected,
2085 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2086 	 */
2087 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 &&
2088 	    outif)
2089 		icmp6_ifstat_inc(outif, ifs6_out_error);
2090 
2091 	if (outif)
2092 		icmp6_ifoutstat_inc(outif, type, code);
2093 
2094 	return;
2095 
2096  bad:
2097 	m_freem(m);
2098 	return;
2099 }
2100 
2101 static const char *
2102 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2103 	struct in6_addr *tgt6)
2104 {
2105 	static char buf[1024];
2106 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2107 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2108 	return buf;
2109 }
2110 
2111 void
2112 icmp6_redirect_input(struct mbuf *m, int off)
2113 {
2114 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2115 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2116 	struct nd_redirect *nd_rd;
2117 	int icmp6len = ntohs(ip6->ip6_plen);
2118 	char *lladdr = NULL;
2119 	int lladdrlen = 0;
2120 	struct rtentry *rt = NULL;
2121 	int is_router;
2122 	int is_onlink;
2123 	struct in6_addr src6 = ip6->ip6_src;
2124 	struct in6_addr redtgt6;
2125 	struct in6_addr reddst6;
2126 	union nd_opts ndopts;
2127 
2128 	if (!ifp)
2129 		return;
2130 
2131 	/* XXX if we are router, we don't update route by icmp6 redirect */
2132 	if (ip6_forwarding)
2133 		goto freeit;
2134 	if (!icmp6_rediraccept)
2135 		goto freeit;
2136 
2137 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2138 	if (nd_rd == NULL) {
2139 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
2140 		return;
2141 	}
2142 	redtgt6 = nd_rd->nd_rd_target;
2143 	reddst6 = nd_rd->nd_rd_dst;
2144 
2145 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2146 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2147 		goto freeit;
2148 	}
2149 
2150 	/* validation */
2151 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2152 		nd6log((LOG_ERR,
2153 			"ICMP6 redirect sent from %s rejected; "
2154 			"must be from linklocal\n", ip6_sprintf(&src6)));
2155 		goto bad;
2156 	}
2157 	if (ip6->ip6_hlim != 255) {
2158 		nd6log((LOG_ERR,
2159 			"ICMP6 redirect sent from %s rejected; "
2160 			"hlim=%d (must be 255)\n",
2161 			ip6_sprintf(&src6), ip6->ip6_hlim));
2162 		goto bad;
2163 	}
2164     {
2165 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2166 	struct sockaddr_in6 sin6;
2167 	struct in6_addr *gw6;
2168 
2169 	sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0);
2170 	rt = rtalloc1((struct sockaddr *)&sin6, 0);
2171 	if (rt) {
2172 		if (rt->rt_gateway == NULL ||
2173 		    rt->rt_gateway->sa_family != AF_INET6) {
2174 			nd6log((LOG_ERR,
2175 			    "ICMP6 redirect rejected; no route "
2176 			    "with inet6 gateway found for redirect dst: %s\n",
2177 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2178 			RTFREE(rt);
2179 			goto bad;
2180 		}
2181 
2182 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2183 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2184 			nd6log((LOG_ERR,
2185 				"ICMP6 redirect rejected; "
2186 				"not equal to gw-for-src=%s (must be same): "
2187 				"%s\n",
2188 				ip6_sprintf(gw6),
2189 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2190 			RTFREE(rt);
2191 			goto bad;
2192 		}
2193 	} else {
2194 		nd6log((LOG_ERR,
2195 			"ICMP6 redirect rejected; "
2196 			"no route found for redirect dst: %s\n",
2197 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2198 		goto bad;
2199 	}
2200 	RTFREE(rt);
2201 	rt = NULL;
2202     }
2203 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2204 		nd6log((LOG_ERR,
2205 			"ICMP6 redirect rejected; "
2206 			"redirect dst must be unicast: %s\n",
2207 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2208 		goto bad;
2209 	}
2210 
2211 	is_router = is_onlink = 0;
2212 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2213 		is_router = 1;	/* router case */
2214 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2215 		is_onlink = 1;	/* on-link destination case */
2216 	if (!is_router && !is_onlink) {
2217 		nd6log((LOG_ERR,
2218 			"ICMP6 redirect rejected; "
2219 			"neither router case nor onlink case: %s\n",
2220 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2221 		goto bad;
2222 	}
2223 	/* validation passed */
2224 
2225 	icmp6len -= sizeof(*nd_rd);
2226 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2227 	if (nd6_options(&ndopts) < 0) {
2228 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2229 			"invalid ND option, rejected: %s\n",
2230 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2231 		/* nd6_options have incremented stats */
2232 		goto freeit;
2233 	}
2234 
2235 	if (ndopts.nd_opts_tgt_lladdr) {
2236 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2237 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2238 	}
2239 
2240 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2241 		nd6log((LOG_INFO,
2242 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2243 			"(if %d, icmp6 packet %d): %s\n",
2244 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2245 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2246 		goto bad;
2247 	}
2248 
2249 	/* RFC 2461 8.3 */
2250 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2251 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2252 
2253 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2254 		/* perform rtredirect */
2255 		struct sockaddr_in6 sdst;
2256 		struct sockaddr_in6 sgw;
2257 		struct sockaddr_in6 ssrc;
2258 		unsigned long rtcount;
2259 		struct rtentry *newrt = NULL;
2260 
2261 		/*
2262 		 * do not install redirect route, if the number of entries
2263 		 * is too much (> hiwat).  note that, the node (= host) will
2264 		 * work just fine even if we do not install redirect route
2265 		 * (there will be additional hops, though).
2266 		 */
2267 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2268 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat)
2269 			return;
2270 		else if (0 <= icmp6_redirect_lowat &&
2271 		    rtcount > icmp6_redirect_lowat) {
2272 			/*
2273 			 * XXX nuke a victim, install the new one.
2274 			 */
2275 		}
2276 
2277 		bzero(&sdst, sizeof(sdst));
2278 		bzero(&sgw, sizeof(sgw));
2279 		bzero(&ssrc, sizeof(ssrc));
2280 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2281 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2282 			sizeof(struct sockaddr_in6);
2283 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2284 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2285 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2286 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2287 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2288 			   (struct sockaddr *)&ssrc,
2289 			   &newrt);
2290 
2291 		if (newrt) {
2292 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2293 			    icmp6_redirect_timeout_q);
2294 			rtfree(newrt);
2295 		}
2296 	}
2297 	/* finally update cached route in each socket via pfctlinput */
2298 	{
2299 		struct sockaddr_in6 sdst;
2300 
2301 		sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0);
2302 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2303 #if defined(IPSEC) || defined(FAST_IPSEC)
2304 		key_sa_routechange((struct sockaddr *)&sdst);
2305 #endif
2306 	}
2307 
2308  freeit:
2309 	m_freem(m);
2310 	return;
2311 
2312  bad:
2313 	ICMP6_STATINC(ICMP6_STAT_BADREDIRECT);
2314 	m_freem(m);
2315 }
2316 
2317 void
2318 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2319 {
2320 	struct ifnet *ifp;	/* my outgoing interface */
2321 	struct in6_addr *ifp_ll6;
2322 	struct in6_addr *nexthop;
2323 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2324 	struct mbuf *m = NULL;	/* newly allocated one */
2325 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2326 	struct nd_redirect *nd_rd;
2327 	size_t maxlen;
2328 	u_char *p;
2329 	struct sockaddr_in6 src_sa;
2330 
2331 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0);
2332 
2333 	/* if we are not router, we don't send icmp6 redirect */
2334 	if (!ip6_forwarding)
2335 		goto fail;
2336 
2337 	/* sanity check */
2338 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2339 		goto fail;
2340 
2341 	/*
2342 	 * Address check:
2343 	 *  the source address must identify a neighbor, and
2344 	 *  the destination address must not be a multicast address
2345 	 *  [RFC 2461, sec 8.2]
2346 	 */
2347 	sip6 = mtod(m0, struct ip6_hdr *);
2348 	sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0);
2349 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2350 		goto fail;
2351 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2352 		goto fail;	/* what should we do here? */
2353 
2354 	/* rate limit */
2355 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2356 		goto fail;
2357 
2358 	/*
2359 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2360 	 * we almost always ask for an mbuf cluster for simplicity.
2361 	 * (MHLEN < IPV6_MMTU is almost always true)
2362 	 */
2363 #if IPV6_MMTU >= MCLBYTES
2364 # error assumption failed about IPV6_MMTU and MCLBYTES
2365 #endif
2366 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2367 	if (m && IPV6_MMTU >= MHLEN)
2368 		MCLGET(m, M_DONTWAIT);
2369 	if (!m)
2370 		goto fail;
2371 	m->m_pkthdr.rcvif = NULL;
2372 	m->m_len = 0;
2373 	maxlen = M_TRAILINGSPACE(m);
2374 	maxlen = min(IPV6_MMTU, maxlen);
2375 	/* just for safety */
2376 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2377 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2378 		goto fail;
2379 	}
2380 
2381 	{
2382 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2383 		struct in6_ifaddr *ia;
2384 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2385 						 IN6_IFF_NOTREADY|
2386 						 IN6_IFF_ANYCAST)) == NULL)
2387 			goto fail;
2388 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2389 	}
2390 
2391 	/* get ip6 linklocal address for the router. */
2392 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2393 		struct sockaddr_in6 *sin6;
2394 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2395 		nexthop = &sin6->sin6_addr;
2396 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2397 			nexthop = NULL;
2398 	} else
2399 		nexthop = NULL;
2400 
2401 	/* ip6 */
2402 	ip6 = mtod(m, struct ip6_hdr *);
2403 	ip6->ip6_flow = 0;
2404 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2405 	ip6->ip6_vfc |= IPV6_VERSION;
2406 	/* ip6->ip6_plen will be set later */
2407 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2408 	ip6->ip6_hlim = 255;
2409 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2410 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2411 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2412 
2413 	/* ND Redirect */
2414 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2415 	nd_rd->nd_rd_type = ND_REDIRECT;
2416 	nd_rd->nd_rd_code = 0;
2417 	nd_rd->nd_rd_reserved = 0;
2418 	if (rt->rt_flags & RTF_GATEWAY) {
2419 		/*
2420 		 * nd_rd->nd_rd_target must be a link-local address in
2421 		 * better router cases.
2422 		 */
2423 		if (!nexthop)
2424 			goto fail;
2425 		bcopy(nexthop, &nd_rd->nd_rd_target,
2426 		      sizeof(nd_rd->nd_rd_target));
2427 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2428 		      sizeof(nd_rd->nd_rd_dst));
2429 	} else {
2430 		/* make sure redtgt == reddst */
2431 		nexthop = &sip6->ip6_dst;
2432 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2433 		      sizeof(nd_rd->nd_rd_target));
2434 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2435 		      sizeof(nd_rd->nd_rd_dst));
2436 	}
2437 
2438 	p = (u_char *)(nd_rd + 1);
2439 
2440 	{
2441 		/* target lladdr option */
2442 		struct rtentry *rt_nexthop = NULL;
2443 		int len;
2444 		const struct sockaddr_dl *sdl;
2445 		struct nd_opt_hdr *nd_opt;
2446 		char *lladdr;
2447 
2448 		rt_nexthop = nd6_lookup(nexthop, 0, ifp);
2449 		if (!rt_nexthop)
2450 			goto nolladdropt;
2451 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2452 		len = (len + 7) & ~7;	/* round by 8 */
2453 		/* safety check */
2454 		if (len + (p - (u_char *)ip6) > maxlen)
2455 			goto nolladdropt;
2456 		if (!(rt_nexthop->rt_flags & RTF_GATEWAY) &&
2457 		    (rt_nexthop->rt_flags & RTF_LLINFO) &&
2458 		    (rt_nexthop->rt_gateway->sa_family == AF_LINK) &&
2459 		    (sdl = satocsdl(rt_nexthop->rt_gateway)) &&
2460 		    sdl->sdl_alen) {
2461 			nd_opt = (struct nd_opt_hdr *)p;
2462 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2463 			nd_opt->nd_opt_len = len >> 3;
2464 			lladdr = (char *)(nd_opt + 1);
2465 			memcpy(lladdr, CLLADDR(sdl), ifp->if_addrlen);
2466 			p += len;
2467 		}
2468 	}
2469   nolladdropt:;
2470 
2471 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2472 
2473 	/* just to be safe */
2474 	if (m0->m_flags & M_DECRYPTED)
2475 		goto noredhdropt;
2476 	if (p - (u_char *)ip6 > maxlen)
2477 		goto noredhdropt;
2478 
2479 	{
2480 		/* redirected header option */
2481 		int len;
2482 		struct nd_opt_rd_hdr *nd_opt_rh;
2483 
2484 		/*
2485 		 * compute the maximum size for icmp6 redirect header option.
2486 		 * XXX room for auth header?
2487 		 */
2488 		len = maxlen - (p - (u_char *)ip6);
2489 		len &= ~7;
2490 
2491 		/*
2492 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2493 		 * about padding/truncate rule for the original IP packet.
2494 		 * From the discussion on IPv6imp in Feb 1999,
2495 		 * the consensus was:
2496 		 * - "attach as much as possible" is the goal
2497 		 * - pad if not aligned (original size can be guessed by
2498 		 *   original ip6 header)
2499 		 * Following code adds the padding if it is simple enough,
2500 		 * and truncates if not.
2501 		 */
2502 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2503 			/* not enough room, truncate */
2504 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2505 			    m0->m_pkthdr.len);
2506 		} else {
2507 			/*
2508 			 * enough room, truncate if not aligned.
2509 			 * we don't pad here for simplicity.
2510 			 */
2511 			size_t extra;
2512 
2513 			extra = m0->m_pkthdr.len % 8;
2514 			if (extra) {
2515 				/* truncate */
2516 				m_adj(m0, -extra);
2517 			}
2518 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2519 		}
2520 
2521 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2522 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2523 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2524 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2525 		p += sizeof(*nd_opt_rh);
2526 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2527 
2528 		/* connect m0 to m */
2529 		m->m_pkthdr.len += m0->m_pkthdr.len;
2530 		m_cat(m, m0);
2531 		m0 = NULL;
2532 	}
2533 noredhdropt:
2534 	if (m0) {
2535 		m_freem(m0);
2536 		m0 = NULL;
2537 	}
2538 
2539 	/* XXX: clear embedded link IDs in the inner header */
2540 	in6_clearscope(&sip6->ip6_src);
2541 	in6_clearscope(&sip6->ip6_dst);
2542 	in6_clearscope(&nd_rd->nd_rd_target);
2543 	in6_clearscope(&nd_rd->nd_rd_dst);
2544 
2545 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2546 
2547 	nd_rd->nd_rd_cksum = 0;
2548 	nd_rd->nd_rd_cksum
2549 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2550 
2551 	/* send the packet to outside... */
2552 	if (ip6_output(m, NULL, NULL, 0,
2553 		(struct ip6_moptions *)NULL, (struct socket *)NULL, NULL) != 0)
2554 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2555 
2556 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2557 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2558 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT);
2559 
2560 	return;
2561 
2562 fail:
2563 	if (m)
2564 		m_freem(m);
2565 	if (m0)
2566 		m_freem(m0);
2567 }
2568 
2569 /*
2570  * ICMPv6 socket option processing.
2571  */
2572 int
2573 icmp6_ctloutput(int op, struct socket *so, int level,
2574 	int optname, struct mbuf **mp)
2575 {
2576 	int error = 0;
2577 	int optlen;
2578 	struct in6pcb *in6p = sotoin6pcb(so);
2579 	struct mbuf *m = *mp;
2580 
2581 	optlen = m ? m->m_len : 0;
2582 
2583 	if (level != IPPROTO_ICMPV6)
2584 		return rip6_ctloutput(op, so, level, optname, mp);
2585 
2586 	switch (op) {
2587 	case PRCO_SETOPT:
2588 		switch (optname) {
2589 		case ICMP6_FILTER:
2590 		    {
2591 			struct icmp6_filter *p;
2592 
2593 			if (optlen != sizeof(*p)) {
2594 				error = EMSGSIZE;
2595 				break;
2596 			}
2597 			p = mtod(m, struct icmp6_filter *);
2598 			if (!p || !in6p->in6p_icmp6filt) {
2599 				error = EINVAL;
2600 				break;
2601 			}
2602 			bcopy(p, in6p->in6p_icmp6filt,
2603 				sizeof(struct icmp6_filter));
2604 			error = 0;
2605 			break;
2606 		    }
2607 
2608 		default:
2609 			error = ENOPROTOOPT;
2610 			break;
2611 		}
2612 		if (m)
2613 			(void)m_freem(m);
2614 		break;
2615 
2616 	case PRCO_GETOPT:
2617 		switch (optname) {
2618 		case ICMP6_FILTER:
2619 		    {
2620 			struct icmp6_filter *p;
2621 
2622 			if (!in6p->in6p_icmp6filt) {
2623 				error = EINVAL;
2624 				break;
2625 			}
2626 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2627 			m->m_len = sizeof(struct icmp6_filter);
2628 			p = mtod(m, struct icmp6_filter *);
2629 			bcopy(in6p->in6p_icmp6filt, p,
2630 				sizeof(struct icmp6_filter));
2631 			error = 0;
2632 			break;
2633 		    }
2634 
2635 		default:
2636 			error = ENOPROTOOPT;
2637 			break;
2638 		}
2639 		break;
2640 	}
2641 
2642 	return (error);
2643 }
2644 
2645 /*
2646  * Perform rate limit check.
2647  * Returns 0 if it is okay to send the icmp6 packet.
2648  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2649  * limitation.
2650  *
2651  * XXX per-destination/type check necessary?
2652  */
2653 static int
2654 icmp6_ratelimit(
2655 	const struct in6_addr *dst,	/* not used at this moment */
2656 	const int type,		/* not used at this moment */
2657 	const int code)		/* not used at this moment */
2658 {
2659 	int ret;
2660 
2661 	ret = 0;	/* okay to send */
2662 
2663 	/* PPS limit */
2664 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2665 	    icmp6errppslim)) {
2666 		/* The packet is subject to rate limit */
2667 		ret++;
2668 	}
2669 
2670 	return ret;
2671 }
2672 
2673 static struct rtentry *
2674 icmp6_mtudisc_clone(struct sockaddr *dst)
2675 {
2676 	struct rtentry *rt;
2677 	int    error;
2678 
2679 	rt = rtalloc1(dst, 1);
2680 	if (rt == 0)
2681 		return NULL;
2682 
2683 	/* If we didn't get a host route, allocate one */
2684 	if ((rt->rt_flags & RTF_HOST) == 0) {
2685 		struct rtentry *nrt;
2686 
2687 		error = rtrequest((int) RTM_ADD, dst,
2688 		    (struct sockaddr *) rt->rt_gateway,
2689 		    (struct sockaddr *) 0,
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 SYSCTL_SETUP(sysctl_net_inet6_icmp6_setup,
2762 	     "sysctl net.inet6.icmp6 subtree setup")
2763 {
2764 	extern int nd6_maxqueuelen; /* defined in nd6.c */
2765 
2766 	sysctl_createv(clog, 0, NULL, NULL,
2767 		       CTLFLAG_PERMANENT,
2768 		       CTLTYPE_NODE, "net", NULL,
2769 		       NULL, 0, NULL, 0,
2770 		       CTL_NET, CTL_EOL);
2771 	sysctl_createv(clog, 0, NULL, NULL,
2772 		       CTLFLAG_PERMANENT,
2773 		       CTLTYPE_NODE, "inet6", NULL,
2774 		       NULL, 0, NULL, 0,
2775 		       CTL_NET, PF_INET6, CTL_EOL);
2776 	sysctl_createv(clog, 0, NULL, NULL,
2777 		       CTLFLAG_PERMANENT,
2778 		       CTLTYPE_NODE, "icmp6",
2779 		       SYSCTL_DESCR("ICMPv6 related settings"),
2780 		       NULL, 0, NULL, 0,
2781 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2782 
2783 	sysctl_createv(clog, 0, NULL, NULL,
2784 		       CTLFLAG_PERMANENT,
2785 		       CTLTYPE_STRUCT, "stats",
2786 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2787 		       sysctl_net_inet6_icmp6_stats, 0, NULL, 0,
2788 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2789 		       ICMPV6CTL_STATS, CTL_EOL);
2790 	sysctl_createv(clog, 0, NULL, NULL,
2791 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2792 		       CTLTYPE_INT, "rediraccept",
2793 		       SYSCTL_DESCR("Accept and process redirect messages"),
2794 		       NULL, 0, &icmp6_rediraccept, 0,
2795 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2796 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2797 	sysctl_createv(clog, 0, NULL, NULL,
2798 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2799 		       CTLTYPE_INT, "redirtimeout",
2800 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2801 		       NULL, 0, &icmp6_redirtimeout, 0,
2802 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2803 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2804 #if 0 /* obsoleted */
2805 	sysctl_createv(clog, 0, NULL, NULL,
2806 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2807 		       CTLTYPE_INT, "errratelimit", NULL,
2808 		       NULL, 0, &icmp6_errratelimit, 0,
2809 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2810 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2811 #endif
2812 	sysctl_createv(clog, 0, NULL, NULL,
2813 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2814 		       CTLTYPE_INT, "nd6_prune",
2815 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
2816 		       NULL, 0, &nd6_prune, 0,
2817 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2818 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
2819 	sysctl_createv(clog, 0, NULL, NULL,
2820 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2821 		       CTLTYPE_INT, "nd6_delay",
2822 		       SYSCTL_DESCR("First probe delay time"),
2823 		       NULL, 0, &nd6_delay, 0,
2824 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2825 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
2826 	sysctl_createv(clog, 0, NULL, NULL,
2827 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2828 		       CTLTYPE_INT, "nd6_umaxtries",
2829 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2830 		       NULL, 0, &nd6_umaxtries, 0,
2831 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2832 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
2833 	sysctl_createv(clog, 0, NULL, NULL,
2834 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2835 		       CTLTYPE_INT, "nd6_mmaxtries",
2836 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
2837 		       NULL, 0, &nd6_mmaxtries, 0,
2838 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2839 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
2840 	sysctl_createv(clog, 0, NULL, NULL,
2841 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2842 		       CTLTYPE_INT, "nd6_useloopback",
2843 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
2844 		       NULL, 0, &nd6_useloopback, 0,
2845 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2846 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
2847 #if 0 /* obsoleted */
2848 	sysctl_createv(clog, 0, NULL, NULL,
2849 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2850 		       CTLTYPE_INT, "nd6_proxyall", NULL,
2851 		       NULL, 0, &nd6_proxyall, 0,
2852 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2853 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
2854 #endif
2855 	sysctl_createv(clog, 0, NULL, NULL,
2856 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2857 		       CTLTYPE_INT, "nodeinfo",
2858 		       SYSCTL_DESCR("Respond to node information requests"),
2859 		       NULL, 0, &icmp6_nodeinfo, 0,
2860 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2861 		       ICMPV6CTL_NODEINFO, CTL_EOL);
2862 	sysctl_createv(clog, 0, NULL, NULL,
2863 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2864 		       CTLTYPE_INT, "errppslimit",
2865 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
2866 		       NULL, 0, &icmp6errppslim, 0,
2867 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2868 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
2869 	sysctl_createv(clog, 0, NULL, NULL,
2870 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2871 		       CTLTYPE_INT, "nd6_maxnudhint",
2872 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2873 		       NULL, 0, &nd6_maxnudhint, 0,
2874 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2875 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
2876 	sysctl_createv(clog, 0, NULL, NULL,
2877 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2878 		       CTLTYPE_INT, "mtudisc_hiwat",
2879 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2880 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
2881 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2882 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
2883 	sysctl_createv(clog, 0, NULL, NULL,
2884 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2885 		       CTLTYPE_INT, "mtudisc_lowat",
2886 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2887 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
2888 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2889 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
2890 	sysctl_createv(clog, 0, NULL, NULL,
2891 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2892 		       CTLTYPE_INT, "nd6_debug",
2893 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
2894 		       NULL, 0, &nd6_debug, 0,
2895 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2896 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
2897 	sysctl_createv(clog, 0, NULL, NULL,
2898 		       CTLFLAG_PERMANENT,
2899 		       CTLTYPE_STRUCT, "nd6_drlist",
2900 		       SYSCTL_DESCR("Default router list"),
2901 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2902 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2903 		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
2904 	sysctl_createv(clog, 0, NULL, NULL,
2905 		       CTLFLAG_PERMANENT,
2906 		       CTLTYPE_STRUCT, "nd6_prlist",
2907 		       SYSCTL_DESCR("Prefix list"),
2908 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2909 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2910 		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
2911 	sysctl_createv(clog, 0, NULL, NULL,
2912 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2913 		       CTLTYPE_INT, "maxqueuelen",
2914 		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
2915 		       NULL, 1, &nd6_maxqueuelen, 0,
2916 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2917 		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
2918 }
2919 
2920 void
2921 icmp6_statinc(u_int stat)
2922 {
2923 
2924 	KASSERT(stat < ICMP6_NSTATS);
2925 	ICMP6_STATINC(stat);
2926 }
2927