xref: /netbsd-src/sys/netinet6/icmp6.c (revision fdd524d4ccd2bb0c6f67401e938dabf773eb0372)
1 /*	$NetBSD: icmp6.c,v 1.192 2016/07/07 09:32:03 ozaki-r 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 1.192 2016/07/07 09:32:03 ozaki-r Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_inet.h"
69 #include "opt_ipsec.h"
70 #endif
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/kmem.h>
75 #include <sys/mbuf.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/time.h>
80 #include <sys/kernel.h>
81 #include <sys/syslog.h>
82 #include <sys/domain.h>
83 #include <sys/sysctl.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 #include <net/if_dl.h>
88 #include <net/if_types.h>
89 
90 #include <netinet/in.h>
91 #include <netinet/in_var.h>
92 #include <netinet/ip6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet6/ip6_private.h>
95 #include <netinet/icmp6.h>
96 #include <netinet6/icmp6_private.h>
97 #include <netinet6/mld6_var.h>
98 #include <netinet6/in6_pcb.h>
99 #include <netinet6/nd6.h>
100 #include <netinet6/in6_ifattach.h>
101 #include <netinet6/ip6protosw.h>
102 #include <netinet6/scope6_var.h>
103 
104 #ifdef IPSEC
105 #include <netipsec/ipsec.h>
106 #include <netipsec/key.h>
107 #endif
108 
109 
110 #include "faith.h"
111 #if defined(NFAITH) && 0 < NFAITH
112 #include <net/if_faith.h>
113 #endif
114 
115 #include <net/net_osdep.h>
116 
117 extern struct domain inet6domain;
118 
119 percpu_t *icmp6stat_percpu;
120 
121 extern struct inpcbtable raw6cbtable;
122 extern int icmp6errppslim;
123 static int icmp6errpps_count = 0;
124 static struct timeval icmp6errppslim_last;
125 extern int icmp6_nodeinfo;
126 
127 /*
128  * List of callbacks to notify when Path MTU changes are made.
129  */
130 struct icmp6_mtudisc_callback {
131 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
132 	void (*mc_func)(struct in6_addr *);
133 };
134 
135 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
136     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
137 
138 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
139 extern int pmtu_expire;
140 
141 /* XXX do these values make any sense? */
142 static int icmp6_mtudisc_hiwat = 1280;
143 static int icmp6_mtudisc_lowat = 256;
144 
145 /*
146  * keep track of # of redirect routes.
147  */
148 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
149 
150 /* XXX experimental, turned off */
151 static int icmp6_redirect_hiwat = -1;
152 static int icmp6_redirect_lowat = -1;
153 
154 static void icmp6_errcount(u_int, int, int);
155 static int icmp6_rip6_input(struct mbuf **, int);
156 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
157 static const char *icmp6_redirect_diag(struct in6_addr *,
158 	struct in6_addr *, struct in6_addr *);
159 static struct mbuf *ni6_input(struct mbuf *, int);
160 static struct mbuf *ni6_nametodns(const char *, int, int);
161 static int ni6_dnsmatch(const char *, int, const char *, int);
162 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
163 			  struct ifnet **, char *);
164 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
165 				struct ifnet *, int);
166 static int icmp6_notify_error(struct mbuf *, int, int, int);
167 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
168 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
169 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
170 static void sysctl_net_inet6_icmp6_setup(struct sysctllog **);
171 
172 
173 void
174 icmp6_init(void)
175 {
176 
177 	sysctl_net_inet6_icmp6_setup(NULL);
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 = kmem_alloc(sizeof(*mc), KM_SLEEP);
259 	mc->mc_func = func;
260 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
261 }
262 
263 /*
264  * A wrapper function for icmp6_error() necessary when the erroneous packet
265  * may not contain enough scope zone information.
266  */
267 void
268 icmp6_error2(struct mbuf *m, int type, int code, int param,
269 	struct ifnet *ifp)
270 {
271 	struct ip6_hdr *ip6;
272 
273 	if (ifp == NULL)
274 		return;
275 
276 	if (m->m_len < sizeof(struct ip6_hdr)) {
277 		m = m_pullup(m, sizeof(struct ip6_hdr));
278 		if (m == NULL)
279 			return;
280 	}
281 
282 	ip6 = mtod(m, struct ip6_hdr *);
283 
284 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
285 		return;
286 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
287 		return;
288 
289 	icmp6_error(m, type, code, param);
290 }
291 
292 /*
293  * Generate an error packet of type error in response to bad IP6 packet.
294  */
295 void
296 icmp6_error(struct mbuf *m, int type, int code, int param)
297 {
298 	struct ip6_hdr *oip6, *nip6;
299 	struct icmp6_hdr *icmp6;
300 	u_int preplen;
301 	int off;
302 	int nxt;
303 
304 	ICMP6_STATINC(ICMP6_STAT_ERROR);
305 
306 	/* count per-type-code statistics */
307 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code);
308 
309 	if (m->m_flags & M_DECRYPTED) {
310 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
311 		goto freeit;
312 	}
313 
314 	if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) &&
315 	    (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL)
316 		return;
317 	oip6 = mtod(m, struct ip6_hdr *);
318 
319 	/*
320 	 * If the destination address of the erroneous packet is a multicast
321 	 * address, or the packet was sent using link-layer multicast,
322 	 * we should basically suppress sending an error (RFC 2463, Section
323 	 * 2.4).
324 	 * We have two exceptions (the item e.2 in that section):
325 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
326 	 * - the Parameter Problem Message that can be allowed an icmp6 error
327 	 *   in the option type field.  This check has been done in
328 	 *   ip6_unknown_opt(), so we can just check the type and code.
329 	 */
330 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
331 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
332 	    (type != ICMP6_PACKET_TOO_BIG &&
333 	     (type != ICMP6_PARAM_PROB ||
334 	      code != ICMP6_PARAMPROB_OPTION)))
335 		goto freeit;
336 
337 	/*
338 	 * RFC 2463, 2.4 (e.5): source address check.
339 	 * XXX: the case of anycast source?
340 	 */
341 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
342 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
343 		goto freeit;
344 
345 	/*
346 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
347 	 * don't do it.
348 	 */
349 	nxt = -1;
350 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
351 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
352 		struct icmp6_hdr *icp;
353 
354 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
355 			sizeof(*icp));
356 		if (icp == NULL) {
357 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
358 			return;
359 		}
360 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
361 		    icp->icmp6_type == ND_REDIRECT) {
362 			/*
363 			 * ICMPv6 error
364 			 * Special case: for redirect (which is
365 			 * informational) we must not send icmp6 error.
366 			 */
367 			ICMP6_STATINC(ICMP6_STAT_CANTERROR);
368 			goto freeit;
369 		} else {
370 			/* ICMPv6 informational - send the error */
371 		}
372 	}
373 #if 0 /* controversial */
374 	else if (off >= 0 && nxt == IPPROTO_ESP) {
375 		/*
376 		 * It could be ICMPv6 error inside ESP.  Take a safer side,
377 		 * don't respond.
378 		 */
379 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
380 		goto freeit;
381 	}
382 #endif
383 	else {
384 		/* non-ICMPv6 - send the error */
385 	}
386 
387 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
388 
389 	/* Finally, do rate limitation check. */
390 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
391 		ICMP6_STATINC(ICMP6_STAT_TOOFREQ);
392 		goto freeit;
393 	}
394 
395 	/*
396 	 * OK, ICMP6 can be generated.
397 	 */
398 
399 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
400 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
401 
402 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
403 	M_PREPEND(m, preplen, M_DONTWAIT);
404 	if (m && M_UNWRITABLE(m, preplen))
405 		m = m_pullup(m, preplen);
406 	if (m == NULL) {
407 		nd6log(LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__);
408 		return;
409 	}
410 
411 	nip6 = mtod(m, struct ip6_hdr *);
412 	nip6->ip6_src  = oip6->ip6_src;
413 	nip6->ip6_dst  = oip6->ip6_dst;
414 
415 	in6_clearscope(&oip6->ip6_src);
416 	in6_clearscope(&oip6->ip6_dst);
417 
418 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
419 	icmp6->icmp6_type = type;
420 	icmp6->icmp6_code = code;
421 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
422 
423 	/*
424 	 * icmp6_reflect() is designed to be in the input path.
425 	 * icmp6_error() can be called from both input and output path,
426 	 * and if we are in output path rcvif could contain bogus value.
427 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
428 	 * information in ip header (nip6).
429 	 */
430 	m_reset_rcvif(m);
431 
432 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
433 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
434 
435 	return;
436 
437   freeit:
438 	/*
439 	 * If we can't tell whether or not we can generate ICMP6, free it.
440 	 */
441 	m_freem(m);
442 }
443 
444 /*
445  * Process a received ICMP6 message.
446  */
447 int
448 icmp6_input(struct mbuf **mp, int *offp, int proto)
449 {
450 	struct mbuf *m = *mp, *n;
451 	struct ip6_hdr *ip6, *nip6;
452 	struct icmp6_hdr *icmp6, *nicmp6;
453 	int off = *offp;
454 	int icmp6len = m->m_pkthdr.len - *offp;
455 	int code, sum, noff, i;
456 	struct ifnet *rcvif;
457 	struct psref psref;
458 
459 	rcvif = m_get_rcvif_psref(m, &psref);
460 	if (__predict_false(rcvif == NULL))
461 		goto freeit;
462 
463 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
464 	KASSERT(ICMP6_MAXLEN < MCLBYTES);
465 	icmp6_ifstat_inc(rcvif, ifs6_in_msg);
466 
467 	/*
468 	 * Locate icmp6 structure in mbuf, and check
469 	 * that not corrupted and of at least minimum length
470 	 */
471 
472 	if (icmp6len < sizeof(struct icmp6_hdr)) {
473 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
474 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
475 		goto freeit;
476 	}
477 
478 	i = off + sizeof(*icmp6);
479 	if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) {
480 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
481 #if 0 /* m is 0 here */
482 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
483 #endif
484 		goto freeit;
485 	}
486 	ip6 = mtod(m, struct ip6_hdr *);
487 	/*
488 	 * calculate the checksum
489 	 */
490 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
491 	if (icmp6 == NULL) {
492 		m_put_rcvif_psref(rcvif, &psref);
493 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
494 		/* m is invalid */
495 		/*icmp6_ifstat_inc(rcvif, ifs6_in_error);*/
496 		return IPPROTO_DONE;
497 	}
498 	KASSERT(IP6_HDR_ALIGNED_P(icmp6));
499 	code = icmp6->icmp6_code;
500 
501 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
502 		nd6log(LOG_ERR, "ICMP6 checksum error(%d|%x) %s\n",
503 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src));
504 		ICMP6_STATINC(ICMP6_STAT_CHECKSUM);
505 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
506 		goto freeit;
507 	}
508 
509 #if defined(NFAITH) && 0 < NFAITH
510 	if (faithprefix(&ip6->ip6_dst)) {
511 		/*
512 		 * Deliver very specific ICMP6 type only.
513 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
514 		 * will not work.
515 		 */
516 		switch (icmp6->icmp6_type) {
517 		case ICMP6_DST_UNREACH:
518 		case ICMP6_PACKET_TOO_BIG:
519 		case ICMP6_TIME_EXCEEDED:
520 			break;
521 		default:
522 			goto freeit;
523 		}
524 	}
525 #endif
526 
527 	ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type);
528 
529 	switch (icmp6->icmp6_type) {
530 	case ICMP6_DST_UNREACH:
531 		icmp6_ifstat_inc(rcvif, ifs6_in_dstunreach);
532 		switch (code) {
533 		case ICMP6_DST_UNREACH_NOROUTE:
534 			code = PRC_UNREACH_NET;
535 			break;
536 		case ICMP6_DST_UNREACH_ADMIN:
537 			icmp6_ifstat_inc(rcvif, ifs6_in_adminprohib);
538 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
539 			break;
540 		case ICMP6_DST_UNREACH_ADDR:
541 			code = PRC_HOSTDEAD;
542 			break;
543 #ifdef COMPAT_RFC1885
544 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
545 			code = PRC_UNREACH_SRCFAIL;
546 			break;
547 #else
548 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
549 			/* I mean "source address was incorrect." */
550 			code = PRC_UNREACH_NET;
551 			break;
552 #endif
553 		case ICMP6_DST_UNREACH_NOPORT:
554 			code = PRC_UNREACH_PORT;
555 			break;
556 		default:
557 			goto badcode;
558 		}
559 		goto deliver;
560 
561 	case ICMP6_PACKET_TOO_BIG:
562 		icmp6_ifstat_inc(rcvif, ifs6_in_pkttoobig);
563 
564 		/*
565 		 * MTU is checked in icmp6_mtudisc.
566 		 */
567 		code = PRC_MSGSIZE;
568 
569 		/*
570 		 * Updating the path MTU will be done after examining
571 		 * intermediate extension headers.
572 		 */
573 		goto deliver;
574 
575 	case ICMP6_TIME_EXCEEDED:
576 		icmp6_ifstat_inc(rcvif, ifs6_in_timeexceed);
577 		switch (code) {
578 		case ICMP6_TIME_EXCEED_TRANSIT:
579 			code = PRC_TIMXCEED_INTRANS;
580 			break;
581 		case ICMP6_TIME_EXCEED_REASSEMBLY:
582 			code = PRC_TIMXCEED_REASS;
583 			break;
584 		default:
585 			goto badcode;
586 		}
587 		goto deliver;
588 
589 	case ICMP6_PARAM_PROB:
590 		icmp6_ifstat_inc(rcvif, ifs6_in_paramprob);
591 		switch (code) {
592 		case ICMP6_PARAMPROB_NEXTHEADER:
593 			code = PRC_UNREACH_PROTOCOL;
594 			break;
595 		case ICMP6_PARAMPROB_HEADER:
596 		case ICMP6_PARAMPROB_OPTION:
597 			code = PRC_PARAMPROB;
598 			break;
599 		default:
600 			goto badcode;
601 		}
602 		goto deliver;
603 
604 	case ICMP6_ECHO_REQUEST:
605 		icmp6_ifstat_inc(rcvif, ifs6_in_echo);
606 		if (code != 0)
607 			goto badcode;
608 		/*
609 		 * Copy mbuf to send to two data paths: userland socket(s),
610 		 * and to the querier (echo reply).
611 		 * m: a copy for socket, n: a copy for querier
612 		 *
613 		 * If the first mbuf is shared, or the first mbuf is too short,
614 		 * copy the first part of the data into a fresh mbuf.
615 		 * Otherwise, we will wrongly overwrite both copies.
616 		 */
617 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
618 			/* Give up local */
619 			n = m;
620 			m = NULL;
621 		} else if (M_READONLY(n) ||
622 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
623 			struct mbuf *n0 = n;
624 
625 			/*
626 			 * Prepare an internal mbuf.  m_pullup() doesn't
627 			 * always copy the length we specified.
628 			 */
629 			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
630 				/* Give up local */
631 				n = m;
632 				m = NULL;
633 			}
634 			m_freem(n0);
635 		}
636 		IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
637 		    sizeof(*nicmp6));
638 		if (nicmp6 == NULL)
639 			goto freeit;
640 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
641 		nicmp6->icmp6_code = 0;
642 		if (n) {
643 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
644 			icmp6s[ICMP6_STAT_REFLECT]++;
645 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++;
646 			ICMP6_STAT_PUTREF();
647 			icmp6_reflect(n, off);
648 		}
649 		if (!m)
650 			goto freeit;
651 		break;
652 
653 	case ICMP6_ECHO_REPLY:
654 		icmp6_ifstat_inc(rcvif, ifs6_in_echoreply);
655 		if (code != 0)
656 			goto badcode;
657 		break;
658 
659 	case MLD_LISTENER_QUERY:
660 	case MLD_LISTENER_REPORT:
661 		if (icmp6len < sizeof(struct mld_hdr))
662 			goto badlen;
663 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
664 			icmp6_ifstat_inc(rcvif, ifs6_in_mldquery);
665 		else
666 			icmp6_ifstat_inc(rcvif, ifs6_in_mldreport);
667 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
668 			/* give up local */
669 			mld_input(m, off);
670 			m = NULL;
671 			goto freeit;
672 		}
673 		mld_input(n, off);
674 		/* m stays. */
675 		break;
676 
677 	case MLD_LISTENER_DONE:
678 		icmp6_ifstat_inc(rcvif, ifs6_in_mlddone);
679 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
680 			goto badlen;
681 		break;		/* nothing to be done in kernel */
682 
683 	case MLD_MTRACE_RESP:
684 	case MLD_MTRACE:
685 		/* XXX: these two are experimental.  not officially defined. */
686 		/* XXX: per-interface statistics? */
687 		break;		/* just pass it to applications */
688 
689 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
690 	    {
691 		enum { WRU, FQDN } mode;
692 
693 		if (!icmp6_nodeinfo)
694 			break;
695 
696 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
697 			mode = WRU;
698 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
699 			mode = FQDN;
700 		else
701 			goto badlen;
702 
703 		if (mode == FQDN) {
704 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
705 			if (n)
706 				n = ni6_input(n, off);
707 			/* XXX meaningless if n == NULL */
708 			noff = sizeof(struct ip6_hdr);
709 		} else {
710 			u_char *p;
711 			int maxhlen;
712 
713 			if ((icmp6_nodeinfo & 5) != 5)
714 				break;
715 
716 			if (code != 0)
717 				goto badcode;
718 			MGETHDR(n, M_DONTWAIT, m->m_type);
719 			if (n && ICMP6_MAXLEN > MHLEN) {
720 				MCLGET(n, M_DONTWAIT);
721 				if ((n->m_flags & M_EXT) == 0) {
722 					m_free(n);
723 					n = NULL;
724 				}
725 			}
726 			if (n == NULL) {
727 				/* Give up remote */
728 				break;
729 			}
730 			m_reset_rcvif(n);
731 			n->m_len = 0;
732 			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
733 			if (maxhlen < 0)
734 				break;
735 			if (maxhlen > hostnamelen)
736 				maxhlen = hostnamelen;
737 			/*
738 			 * Copy IPv6 and ICMPv6 only.
739 			 */
740 			nip6 = mtod(n, struct ip6_hdr *);
741 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
742 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
743 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
744 			p = (u_char *)(nicmp6 + 1);
745 			memset(p, 0, 4);
746 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
747 			noff = sizeof(struct ip6_hdr);
748 			M_COPY_PKTHDR(n, m); /* just for rcvif */
749 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
750 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
751 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
752 			nicmp6->icmp6_code = 0;
753 		}
754 #undef hostnamelen
755 		if (n) {
756 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
757 			icmp6s[ICMP6_STAT_REFLECT]++;
758 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++;
759 			ICMP6_STAT_PUTREF();
760 			icmp6_reflect(n, noff);
761 		}
762 		break;
763 	    }
764 
765 	case ICMP6_WRUREPLY:
766 		if (code != 0)
767 			goto badcode;
768 		break;
769 
770 	case ND_ROUTER_SOLICIT:
771 		icmp6_ifstat_inc(rcvif, ifs6_in_routersolicit);
772 		if (code != 0)
773 			goto badcode;
774 		if (icmp6len < sizeof(struct nd_router_solicit))
775 			goto badlen;
776 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
777 			/* give up local */
778 			nd6_rs_input(m, off, icmp6len);
779 			m = NULL;
780 			goto freeit;
781 		}
782 		nd6_rs_input(n, off, icmp6len);
783 		/* m stays. */
784 		break;
785 
786 	case ND_ROUTER_ADVERT:
787 		icmp6_ifstat_inc(rcvif, ifs6_in_routeradvert);
788 		if (code != 0)
789 			goto badcode;
790 		if (icmp6len < sizeof(struct nd_router_advert))
791 			goto badlen;
792 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
793 			/* give up local */
794 			nd6_ra_input(m, off, icmp6len);
795 			m = NULL;
796 			goto freeit;
797 		}
798 		nd6_ra_input(n, off, icmp6len);
799 		/* m stays. */
800 		break;
801 
802 	case ND_NEIGHBOR_SOLICIT:
803 		icmp6_ifstat_inc(rcvif, ifs6_in_neighborsolicit);
804 		if (code != 0)
805 			goto badcode;
806 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
807 			goto badlen;
808 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
809 			/* give up local */
810 			nd6_ns_input(m, off, icmp6len);
811 			m = NULL;
812 			goto freeit;
813 		}
814 		nd6_ns_input(n, off, icmp6len);
815 		/* m stays. */
816 		break;
817 
818 	case ND_NEIGHBOR_ADVERT:
819 		icmp6_ifstat_inc(rcvif, ifs6_in_neighboradvert);
820 		if (code != 0)
821 			goto badcode;
822 		if (icmp6len < sizeof(struct nd_neighbor_advert))
823 			goto badlen;
824 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
825 			/* give up local */
826 			nd6_na_input(m, off, icmp6len);
827 			m = NULL;
828 			goto freeit;
829 		}
830 		nd6_na_input(n, off, icmp6len);
831 		/* m stays. */
832 		break;
833 
834 	case ND_REDIRECT:
835 		icmp6_ifstat_inc(rcvif, ifs6_in_redirect);
836 		if (code != 0)
837 			goto badcode;
838 		if (icmp6len < sizeof(struct nd_redirect))
839 			goto badlen;
840 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
841 			/* give up local */
842 			icmp6_redirect_input(m, off);
843 			m = NULL;
844 			goto freeit;
845 		}
846 		icmp6_redirect_input(n, off);
847 		/* m stays. */
848 		break;
849 
850 	case ICMP6_ROUTER_RENUMBERING:
851 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
852 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
853 			goto badcode;
854 		if (icmp6len < sizeof(struct icmp6_router_renum))
855 			goto badlen;
856 		break;
857 
858 	default:
859 		nd6log(LOG_DEBUG, "unknown type %d(src=%s, dst=%s, ifid=%d)\n",
860 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
861 		    ip6_sprintf(&ip6->ip6_dst),
862 		    rcvif ? rcvif->if_index : 0);
863 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
864 			/* ICMPv6 error: MUST deliver it by spec... */
865 			code = PRC_NCMDS;
866 			/* deliver */
867 		} else {
868 			/* ICMPv6 informational: MUST not deliver */
869 			break;
870 		}
871 	deliver:
872 		if (icmp6_notify_error(m, off, icmp6len, code)) {
873 			/* In this case, m should've been freed. */
874 			m_put_rcvif_psref(rcvif, &psref);
875 			return (IPPROTO_DONE);
876 		}
877 		break;
878 
879 	badcode:
880 		ICMP6_STATINC(ICMP6_STAT_BADCODE);
881 		break;
882 
883 	badlen:
884 		ICMP6_STATINC(ICMP6_STAT_BADLEN);
885 		break;
886 	}
887 	m_put_rcvif_psref(rcvif, &psref);
888 
889 	/* deliver the packet to appropriate sockets */
890 	icmp6_rip6_input(&m, *offp);
891 
892 	return IPPROTO_DONE;
893 
894  freeit:
895 	m_put_rcvif_psref(rcvif, &psref);
896 	m_freem(m);
897 	return IPPROTO_DONE;
898 }
899 
900 static int
901 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
902 {
903 	struct icmp6_hdr *icmp6;
904 	struct ip6_hdr *eip6;
905 	u_int32_t notifymtu;
906 	struct sockaddr_in6 icmp6src, icmp6dst;
907 
908 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
909 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
910 		goto freeit;
911 	}
912 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
913 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
914 	if (icmp6 == NULL) {
915 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
916 		return (-1);
917 	}
918 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
919 
920 	/* Detect the upper level protocol */
921 	{
922 		void (*ctlfunc)(int, struct sockaddr *, void *);
923 		u_int8_t nxt = eip6->ip6_nxt;
924 		int eoff = off + sizeof(struct icmp6_hdr) +
925 			sizeof(struct ip6_hdr);
926 		struct ip6ctlparam ip6cp;
927 		struct in6_addr *finaldst = NULL;
928 		int icmp6type = icmp6->icmp6_type;
929 		struct ip6_frag *fh;
930 		struct ip6_rthdr *rth;
931 		struct ip6_rthdr0 *rth0;
932 		int rthlen;
933 		struct ifnet *rcvif;
934 		int s;
935 
936 		while (1) { /* XXX: should avoid infinite loop explicitly? */
937 			struct ip6_ext *eh;
938 
939 			switch (nxt) {
940 			case IPPROTO_HOPOPTS:
941 			case IPPROTO_DSTOPTS:
942 			case IPPROTO_AH:
943 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
944 					       eoff, sizeof(*eh));
945 				if (eh == NULL) {
946 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
947 					return (-1);
948 				}
949 
950 				if (nxt == IPPROTO_AH)
951 					eoff += (eh->ip6e_len + 2) << 2;
952 				else
953 					eoff += (eh->ip6e_len + 1) << 3;
954 				nxt = eh->ip6e_nxt;
955 				break;
956 			case IPPROTO_ROUTING:
957 				/*
958 				 * When the erroneous packet contains a
959 				 * routing header, we should examine the
960 				 * header to determine the final destination.
961 				 * Otherwise, we can't properly update
962 				 * information that depends on the final
963 				 * destination (e.g. path MTU).
964 				 */
965 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
966 					       eoff, sizeof(*rth));
967 				if (rth == NULL) {
968 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
969 					return (-1);
970 				}
971 				rthlen = (rth->ip6r_len + 1) << 3;
972 				/*
973 				 * XXX: currently there is no
974 				 * officially defined type other
975 				 * than type-0.
976 				 * Note that if the segment left field
977 				 * is 0, all intermediate hops must
978 				 * have been passed.
979 				 */
980 				if (rth->ip6r_segleft &&
981 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
982 					int hops;
983 
984 					IP6_EXTHDR_GET(rth0,
985 						       struct ip6_rthdr0 *, m,
986 						       eoff, rthlen);
987 					if (rth0 == NULL) {
988 						ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
989 						return (-1);
990 					}
991 					/* just ignore a bogus header */
992 					if ((rth0->ip6r0_len % 2) == 0 &&
993 					    (hops = rth0->ip6r0_len/2))
994 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
995 				}
996 				eoff += rthlen;
997 				nxt = rth->ip6r_nxt;
998 				break;
999 			case IPPROTO_FRAGMENT:
1000 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1001 					       eoff, sizeof(*fh));
1002 				if (fh == NULL) {
1003 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1004 					return (-1);
1005 				}
1006 				/*
1007 				 * Data after a fragment header is meaningless
1008 				 * unless it is the first fragment, but
1009 				 * we'll go to the notify label for path MTU
1010 				 * discovery.
1011 				 */
1012 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1013 					goto notify;
1014 
1015 				eoff += sizeof(struct ip6_frag);
1016 				nxt = fh->ip6f_nxt;
1017 				break;
1018 			default:
1019 				/*
1020 				 * This case includes ESP and the No Next
1021 				 * Header.  In such cases going to the notify
1022 				 * label does not have any meaning
1023 				 * (i.e. ctlfunc will be NULL), but we go
1024 				 * anyway since we might have to update
1025 				 * path MTU information.
1026 				 */
1027 				goto notify;
1028 			}
1029 		}
1030 	  notify:
1031 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1032 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1033 		if (icmp6 == NULL) {
1034 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1035 			return (-1);
1036 		}
1037 
1038 		/*
1039 		 * retrieve parameters from the inner IPv6 header, and convert
1040 		 * them into sockaddr structures.
1041 		 * XXX: there is no guarantee that the source or destination
1042 		 * addresses of the inner packet are in the same scope zone as
1043 		 * the addresses of the icmp packet.  But there is no other
1044 		 * way to determine the zone.
1045 		 */
1046 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1047 
1048 		rcvif = m_get_rcvif(m, &s);
1049 		sockaddr_in6_init(&icmp6dst,
1050 		    (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0);
1051 		if (in6_setscope(&icmp6dst.sin6_addr, rcvif, NULL)) {
1052 			m_put_rcvif(rcvif, &s);
1053 			goto freeit;
1054 		}
1055 		sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0);
1056 		if (in6_setscope(&icmp6src.sin6_addr, rcvif, NULL)) {
1057 			m_put_rcvif(rcvif, &s);
1058 			goto freeit;
1059 		}
1060 		m_put_rcvif(rcvif, &s);
1061 
1062 		icmp6src.sin6_flowinfo =
1063 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1064 
1065 		if (finaldst == NULL)
1066 			finaldst = &eip6->ip6_dst;
1067 		ip6cp.ip6c_m = m;
1068 		ip6cp.ip6c_icmp6 = icmp6;
1069 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1070 		ip6cp.ip6c_off = eoff;
1071 		ip6cp.ip6c_finaldst = finaldst;
1072 		ip6cp.ip6c_src = &icmp6src;
1073 		ip6cp.ip6c_nxt = nxt;
1074 
1075 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1076 			notifymtu = ntohl(icmp6->icmp6_mtu);
1077 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1078 		}
1079 
1080 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1081 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1082 		if (ctlfunc) {
1083 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1084 					  &ip6cp);
1085 		}
1086 	}
1087 	return (0);
1088 
1089   freeit:
1090 	m_freem(m);
1091 	return (-1);
1092 }
1093 
1094 void
1095 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1096 {
1097 	unsigned long rtcount;
1098 	struct icmp6_mtudisc_callback *mc;
1099 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1100 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1101 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1102 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1103 	struct rtentry *rt = NULL;
1104 	struct sockaddr_in6 sin6;
1105 	struct ifnet *rcvif;
1106 	int s;
1107 
1108 	/*
1109 	 * The MTU should not be less than the minimal IPv6 MTU except for the
1110 	 * hack in ip6_output/ip6_setpmtu where we always include a frag header.
1111 	 * In that one case, the MTU might be less than 1280.
1112 	 */
1113 	if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) {
1114 		/* is the mtu even sane? */
1115 		if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1116 			return;
1117 		if (!validated)
1118 			return;
1119 		mtu = IPV6_MMTU - sizeof(struct ip6_frag);
1120 	}
1121 
1122 	/*
1123 	 * allow non-validated cases if memory is plenty, to make traffic
1124 	 * from non-connected pcb happy.
1125 	 */
1126 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1127 	if (validated) {
1128 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1129 			return;
1130 		else if (0 <= icmp6_mtudisc_lowat &&
1131 		    rtcount > icmp6_mtudisc_lowat) {
1132 			/*
1133 			 * XXX nuke a victim, install the new one.
1134 			 */
1135 		}
1136 	} else {
1137 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1138 			return;
1139 	}
1140 
1141 	memset(&sin6, 0, sizeof(sin6));
1142 	sin6.sin6_family = PF_INET6;
1143 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1144 	sin6.sin6_addr = *dst;
1145 	s = pserialize_read_enter();
1146 	rcvif = m_get_rcvif(m, &s);
1147 	if (in6_setscope(&sin6.sin6_addr, rcvif, NULL)) {
1148 		m_put_rcvif(rcvif, &s);
1149 		return;
1150 	}
1151 	m_put_rcvif(rcvif, &s);
1152 
1153 	rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1154 
1155 	if (rt && (rt->rt_flags & RTF_HOST) &&
1156 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1157 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1158 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1159 			ICMP6_STATINC(ICMP6_STAT_PMTUCHG);
1160 			rt->rt_rmx.rmx_mtu = mtu;
1161 		}
1162 	}
1163 	if (rt) {
1164 		rtfree(rt);
1165 	}
1166 
1167 	/*
1168 	 * Notify protocols that the MTU for this destination
1169 	 * has changed.
1170 	 */
1171 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1172 	     mc = LIST_NEXT(mc, mc_list))
1173 		(*mc->mc_func)(&sin6.sin6_addr);
1174 }
1175 
1176 /*
1177  * Process a Node Information Query packet, based on
1178  * draft-ietf-ipngwg-icmp-name-lookups-07.
1179  *
1180  * Spec incompatibilities:
1181  * - IPv6 Subject address handling
1182  * - IPv4 Subject address handling support missing
1183  * - Proxy reply (answer even if it's not for me)
1184  * - joins NI group address at in6_ifattach() time only, does not cope
1185  *   with hostname changes by sethostname(3)
1186  */
1187 static struct mbuf *
1188 ni6_input(struct mbuf *m, int off)
1189 {
1190 	struct icmp6_nodeinfo *ni6, *nni6;
1191 	struct mbuf *n = NULL;
1192 	u_int16_t qtype;
1193 	int subjlen;
1194 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1195 	struct ni_reply_fqdn *fqdn;
1196 	int addrs;		/* for NI_QTYPE_NODEADDR */
1197 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1198 	struct sockaddr_in6 sin6; /* ip6_dst */
1199 	struct in6_addr in6_subj; /* subject address */
1200 	struct ip6_hdr *ip6;
1201 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1202 	char *subj = NULL;
1203 	int s;
1204 	struct ifnet *rcvif;
1205 
1206 	ip6 = mtod(m, struct ip6_hdr *);
1207 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1208 	if (ni6 == NULL) {
1209 		/* m is already reclaimed */
1210 		return NULL;
1211 	}
1212 
1213 	/*
1214 	 * Validate IPv6 destination address.
1215 	 *
1216 	 * The Responder must discard the Query without further processing
1217 	 * unless it is one of the Responder's unicast or anycast addresses, or
1218 	 * a link-local scope multicast address which the Responder has joined.
1219 	 * [icmp-name-lookups-07, Section 4.]
1220 	 */
1221 	sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
1222 	/* XXX scopeid */
1223 	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1224 		; /* unicast/anycast, fine */
1225 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1226 		; /* link-local multicast, fine */
1227 	else
1228 		goto bad;
1229 
1230 	/* validate query Subject field. */
1231 	qtype = ntohs(ni6->ni_qtype);
1232 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1233 	switch (qtype) {
1234 	case NI_QTYPE_NOOP:
1235 	case NI_QTYPE_SUPTYPES:
1236 		/* 07 draft */
1237 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1238 			break;
1239 		/* FALLTHROUGH */
1240 	case NI_QTYPE_FQDN:
1241 	case NI_QTYPE_NODEADDR:
1242 	case NI_QTYPE_IPV4ADDR:
1243 		switch (ni6->ni_code) {
1244 		case ICMP6_NI_SUBJ_IPV6:
1245 #if ICMP6_NI_SUBJ_IPV6 != 0
1246 		case 0:
1247 #endif
1248 			/*
1249 			 * backward compatibility - try to accept 03 draft
1250 			 * format, where no Subject is present.
1251 			 */
1252 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1253 			    subjlen == 0) {
1254 				oldfqdn++;
1255 				break;
1256 			}
1257 #if ICMP6_NI_SUBJ_IPV6 != 0
1258 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1259 				goto bad;
1260 #endif
1261 
1262 			if (subjlen != sizeof(sin6.sin6_addr))
1263 				goto bad;
1264 
1265 			/*
1266 			 * Validate Subject address.
1267 			 *
1268 			 * Not sure what exactly "address belongs to the node"
1269 			 * means in the spec, is it just unicast, or what?
1270 			 *
1271 			 * At this moment we consider Subject address as
1272 			 * "belong to the node" if the Subject address equals
1273 			 * to the IPv6 destination address; validation for
1274 			 * IPv6 destination address should have done enough
1275 			 * check for us.
1276 			 *
1277 			 * We do not do proxy at this moment.
1278 			 */
1279 			/* m_pulldown instead of copy? */
1280 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1281 			    subjlen, (void *)&in6_subj);
1282 			rcvif = m_get_rcvif(m, &s);
1283 			if (in6_setscope(&in6_subj, rcvif, NULL)) {
1284 				m_put_rcvif(rcvif, &s);
1285 				goto bad;
1286 			}
1287 			m_put_rcvif(rcvif, &s);
1288 
1289 			subj = (char *)&in6_subj;
1290 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1291 				break;
1292 
1293 			/*
1294 			 * XXX if we are to allow other cases, we should really
1295 			 * be careful about scope here.
1296 			 * basically, we should disallow queries toward IPv6
1297 			 * destination X with subject Y, if scope(X) > scope(Y).
1298 			 * if we allow scope(X) > scope(Y), it will result in
1299 			 * information leakage across scope boundary.
1300 			 */
1301 			goto bad;
1302 
1303 		case ICMP6_NI_SUBJ_FQDN:
1304 			/*
1305 			 * Validate Subject name with gethostname(3).
1306 			 *
1307 			 * The behavior may need some debate, since:
1308 			 * - we are not sure if the node has FQDN as
1309 			 *   hostname (returned by gethostname(3)).
1310 			 * - the code does wildcard match for truncated names.
1311 			 *   however, we are not sure if we want to perform
1312 			 *   wildcard match, if gethostname(3) side has
1313 			 *   truncated hostname.
1314 			 */
1315 			n = ni6_nametodns(hostname, hostnamelen, 0);
1316 			if (!n || n->m_next || n->m_len == 0)
1317 				goto bad;
1318 			IP6_EXTHDR_GET(subj, char *, m,
1319 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1320 			if (subj == NULL)
1321 				goto bad;
1322 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1323 					n->m_len)) {
1324 				goto bad;
1325 			}
1326 			m_freem(n);
1327 			n = NULL;
1328 			break;
1329 
1330 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1331 		default:
1332 			goto bad;
1333 		}
1334 		break;
1335 	}
1336 
1337 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1338 	switch (qtype) {
1339 	case NI_QTYPE_FQDN:
1340 		if ((icmp6_nodeinfo & 1) == 0)
1341 			goto bad;
1342 		break;
1343 	case NI_QTYPE_NODEADDR:
1344 	case NI_QTYPE_IPV4ADDR:
1345 		if ((icmp6_nodeinfo & 2) == 0)
1346 			goto bad;
1347 		break;
1348 	}
1349 
1350 	/* guess reply length */
1351 	switch (qtype) {
1352 	case NI_QTYPE_NOOP:
1353 		break;		/* no reply data */
1354 	case NI_QTYPE_SUPTYPES:
1355 		replylen += sizeof(u_int32_t);
1356 		break;
1357 	case NI_QTYPE_FQDN:
1358 		/* XXX will append an mbuf */
1359 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1360 		break;
1361 	case NI_QTYPE_NODEADDR:
1362 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1363 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1364 					  sizeof(u_int32_t))) > MCLBYTES)
1365 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1366 		break;
1367 	case NI_QTYPE_IPV4ADDR:
1368 		/* unsupported - should respond with unknown Qtype? */
1369 		goto bad;
1370 	default:
1371 		/*
1372 		 * XXX: We must return a reply with the ICMP6 code
1373 		 * `unknown Qtype' in this case.  However we regard the case
1374 		 * as an FQDN query for backward compatibility.
1375 		 * Older versions set a random value to this field,
1376 		 * so it rarely varies in the defined qtypes.
1377 		 * But the mechanism is not reliable...
1378 		 * maybe we should obsolete older versions.
1379 		 */
1380 		qtype = NI_QTYPE_FQDN;
1381 		/* XXX will append an mbuf */
1382 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1383 		oldfqdn++;
1384 		break;
1385 	}
1386 
1387 	/* allocate an mbuf to reply. */
1388 	MGETHDR(n, M_DONTWAIT, m->m_type);
1389 	if (n == NULL) {
1390 		m_freem(m);
1391 		return (NULL);
1392 	}
1393 	M_MOVE_PKTHDR(n, m); /* just for rcvif */
1394 	if (replylen > MHLEN) {
1395 		if (replylen > MCLBYTES) {
1396 			/*
1397 			 * XXX: should we try to allocate more? But MCLBYTES
1398 			 * is probably much larger than IPV6_MMTU...
1399 			 */
1400 			goto bad;
1401 		}
1402 		MCLGET(n, M_DONTWAIT);
1403 		if ((n->m_flags & M_EXT) == 0) {
1404 			goto bad;
1405 		}
1406 	}
1407 	n->m_pkthdr.len = n->m_len = replylen;
1408 
1409 	/* copy mbuf header and IPv6 + Node Information base headers */
1410 	bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr));
1411 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1412 	bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo));
1413 
1414 	/* qtype dependent procedure */
1415 	switch (qtype) {
1416 	case NI_QTYPE_NOOP:
1417 		nni6->ni_code = ICMP6_NI_SUCCESS;
1418 		nni6->ni_flags = 0;
1419 		break;
1420 	case NI_QTYPE_SUPTYPES:
1421 	{
1422 		u_int32_t v;
1423 		nni6->ni_code = ICMP6_NI_SUCCESS;
1424 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1425 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1426 		v = (u_int32_t)htonl(0x0000000f);
1427 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1428 		break;
1429 	}
1430 	case NI_QTYPE_FQDN:
1431 		nni6->ni_code = ICMP6_NI_SUCCESS;
1432 		fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) +
1433 						sizeof(struct ip6_hdr) +
1434 						sizeof(struct icmp6_nodeinfo));
1435 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1436 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1437 		/*
1438 		 * XXX do we really have FQDN in variable "hostname"?
1439 		 */
1440 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1441 		if (n->m_next == NULL)
1442 			goto bad;
1443 		/* XXX we assume that n->m_next is not a chain */
1444 		if (n->m_next->m_next != NULL)
1445 			goto bad;
1446 		n->m_pkthdr.len += n->m_next->m_len;
1447 		break;
1448 	case NI_QTYPE_NODEADDR:
1449 	{
1450 		int lenlim, copied;
1451 
1452 		nni6->ni_code = ICMP6_NI_SUCCESS;
1453 		n->m_pkthdr.len = n->m_len =
1454 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1455 		lenlim = M_TRAILINGSPACE(n);
1456 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1457 		/* XXX: reset mbuf length */
1458 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1459 			sizeof(struct icmp6_nodeinfo) + copied;
1460 		break;
1461 	}
1462 	default:
1463 		break;		/* XXX impossible! */
1464 	}
1465 
1466 	nni6->ni_type = ICMP6_NI_REPLY;
1467 	m_freem(m);
1468 	return (n);
1469 
1470   bad:
1471 	m_freem(m);
1472 	if (n)
1473 		m_freem(n);
1474 	return (NULL);
1475 }
1476 #undef hostnamelen
1477 
1478 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1479 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1480 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1481 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1482 
1483 /*
1484  * make a mbuf with DNS-encoded string.  no compression support.
1485  *
1486  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1487  * treated as truncated name (two \0 at the end).  this is a wild guess.
1488  *
1489  * old - return pascal string if non-zero
1490  */
1491 static struct mbuf *
1492 ni6_nametodns(const char *name, int namelen, int old)
1493 {
1494 	struct mbuf *m;
1495 	char *cp, *ep;
1496 	const char *p, *q;
1497 	int i, len, nterm;
1498 
1499 	if (old)
1500 		len = namelen + 1;
1501 	else
1502 		len = MCLBYTES;
1503 
1504 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1505 	MGET(m, M_DONTWAIT, MT_DATA);
1506 	if (m && len > MLEN) {
1507 		MCLGET(m, M_DONTWAIT);
1508 		if ((m->m_flags & M_EXT) == 0)
1509 			goto fail;
1510 	}
1511 	if (!m)
1512 		goto fail;
1513 	m->m_next = NULL;
1514 
1515 	if (old) {
1516 		m->m_len = len;
1517 		*mtod(m, char *) = namelen;
1518 		bcopy(name, mtod(m, char *) + 1, namelen);
1519 		return m;
1520 	} else {
1521 		m->m_len = 0;
1522 		cp = mtod(m, char *);
1523 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1524 
1525 		/* if not certain about my name, return empty buffer */
1526 		if (namelen == 0)
1527 			return m;
1528 
1529 		/*
1530 		 * guess if it looks like shortened hostname, or FQDN.
1531 		 * shortened hostname needs two trailing "\0".
1532 		 */
1533 		i = 0;
1534 		for (p = name; p < name + namelen; p++) {
1535 			if (*p && *p == '.')
1536 				i++;
1537 		}
1538 		if (i < 2)
1539 			nterm = 2;
1540 		else
1541 			nterm = 1;
1542 
1543 		p = name;
1544 		while (cp < ep && p < name + namelen) {
1545 			i = 0;
1546 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1547 				i++;
1548 			/* result does not fit into mbuf */
1549 			if (cp + i + 1 >= ep)
1550 				goto fail;
1551 			/*
1552 			 * DNS label length restriction, RFC1035 page 8.
1553 			 * "i == 0" case is included here to avoid returning
1554 			 * 0-length label on "foo..bar".
1555 			 */
1556 			if (i <= 0 || i >= 64)
1557 				goto fail;
1558 			*cp++ = i;
1559 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1560 				goto fail;
1561 			while (i > 0) {
1562 				if (!isalnum(*p) && *p != '-')
1563 					goto fail;
1564 				if (isupper(*p)) {
1565 					*cp++ = tolower(*p);
1566 					p++;
1567 				} else
1568 					*cp++ = *p++;
1569 				i--;
1570 			}
1571 			p = q;
1572 			if (p < name + namelen && *p == '.')
1573 				p++;
1574 		}
1575 		/* termination */
1576 		if (cp + nterm >= ep)
1577 			goto fail;
1578 		while (nterm-- > 0)
1579 			*cp++ = '\0';
1580 		m->m_len = cp - mtod(m, char *);
1581 		return m;
1582 	}
1583 
1584 	panic("should not reach here");
1585 	/* NOTREACHED */
1586 
1587  fail:
1588 	if (m)
1589 		m_freem(m);
1590 	return NULL;
1591 }
1592 
1593 /*
1594  * check if two DNS-encoded string matches.  takes care of truncated
1595  * form (with \0\0 at the end).  no compression support.
1596  * XXX upper/lowercase match (see RFC2065)
1597  */
1598 static int
1599 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1600 {
1601 	const char *a0, *b0;
1602 	int l;
1603 
1604 	/* simplest case - need validation? */
1605 	if (alen == blen && memcmp(a, b, alen) == 0)
1606 		return 1;
1607 
1608 	a0 = a;
1609 	b0 = b;
1610 
1611 	/* termination is mandatory */
1612 	if (alen < 2 || blen < 2)
1613 		return 0;
1614 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1615 		return 0;
1616 	alen--;
1617 	blen--;
1618 
1619 	while (a - a0 < alen && b - b0 < blen) {
1620 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1621 			return 0;
1622 
1623 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1624 			return 0;
1625 		/* we don't support compression yet */
1626 		if (a[0] >= 64 || b[0] >= 64)
1627 			return 0;
1628 
1629 		/* truncated case */
1630 		if (a[0] == 0 && a - a0 == alen - 1)
1631 			return 1;
1632 		if (b[0] == 0 && b - b0 == blen - 1)
1633 			return 1;
1634 		if (a[0] == 0 || b[0] == 0)
1635 			return 0;
1636 
1637 		if (a[0] != b[0])
1638 			return 0;
1639 		l = a[0];
1640 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1641 			return 0;
1642 		if (memcmp(a + 1, b + 1, l) != 0)
1643 			return 0;
1644 
1645 		a += 1 + l;
1646 		b += 1 + l;
1647 	}
1648 
1649 	if (a - a0 == alen && b - b0 == blen)
1650 		return 1;
1651 	else
1652 		return 0;
1653 }
1654 
1655 /*
1656  * calculate the number of addresses to be returned in the node info reply.
1657  */
1658 static int
1659 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m,
1660     struct ifnet **ifpp, char *subj)
1661 {
1662 	struct ifnet *ifp;
1663 	struct in6_ifaddr *ia6;
1664 	struct ifaddr *ifa;
1665 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1666 	int addrs = 0, addrsofif, iffound = 0;
1667 	int niflags = ni6->ni_flags;
1668 
1669 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1670 		switch (ni6->ni_code) {
1671 		case ICMP6_NI_SUBJ_IPV6:
1672 			if (subj == NULL) /* must be impossible... */
1673 				return (0);
1674 			subj_ip6 = (struct sockaddr_in6 *)subj;
1675 			break;
1676 		default:
1677 			/*
1678 			 * XXX: we only support IPv6 subject address for
1679 			 * this Qtype.
1680 			 */
1681 			return (0);
1682 		}
1683 	}
1684 
1685 	IFNET_READER_FOREACH(ifp) {
1686 		addrsofif = 0;
1687 		IFADDR_READER_FOREACH(ifa, ifp) {
1688 			if (ifa->ifa_addr->sa_family != AF_INET6)
1689 				continue;
1690 			ia6 = (struct in6_ifaddr *)ifa;
1691 
1692 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1693 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1694 					       &ia6->ia_addr.sin6_addr))
1695 				iffound = 1;
1696 
1697 			/*
1698 			 * IPv4-mapped addresses can only be returned by a
1699 			 * Node Information proxy, since they represent
1700 			 * addresses of IPv4-only nodes, which perforce do
1701 			 * not implement this protocol.
1702 			 * [icmp-name-lookups-07, Section 5.4]
1703 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1704 			 * this function at this moment.
1705 			 */
1706 
1707 			/* What do we have to do about ::1? */
1708 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1709 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1710 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1711 					continue;
1712 				break;
1713 			case IPV6_ADDR_SCOPE_SITELOCAL:
1714 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1715 					continue;
1716 				break;
1717 			case IPV6_ADDR_SCOPE_GLOBAL:
1718 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1719 					continue;
1720 				break;
1721 			default:
1722 				continue;
1723 			}
1724 
1725 			/*
1726 			 * check if anycast is okay.
1727 			 * XXX: just experimental.  not in the spec.
1728 			 */
1729 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1730 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1731 				continue; /* we need only unicast addresses */
1732 
1733 			addrsofif++; /* count the address */
1734 		}
1735 		if (iffound) {
1736 			*ifpp = ifp;
1737 			return (addrsofif);
1738 		}
1739 
1740 		addrs += addrsofif;
1741 	}
1742 
1743 	return (addrs);
1744 }
1745 
1746 static int
1747 ni6_store_addrs(struct icmp6_nodeinfo *ni6,
1748 	struct icmp6_nodeinfo *nni6, struct ifnet *ifp0,
1749 	int resid)
1750 {
1751 	struct ifnet *ifp = ifp0 ? ifp0 : IFNET_READER_FIRST();
1752 	struct in6_ifaddr *ia6;
1753 	struct ifaddr *ifa;
1754 	struct ifnet *ifp_dep = NULL;
1755 	int copied = 0, allow_deprecated = 0;
1756 	u_char *cp = (u_char *)(nni6 + 1);
1757 	int niflags = ni6->ni_flags;
1758 	u_int32_t ltime;
1759 
1760 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1761 		return (0);	/* needless to copy */
1762 
1763   again:
1764 
1765 	for (; ifp; ifp = IFNET_READER_NEXT(ifp))
1766 	{
1767 		IFADDR_READER_FOREACH(ifa, ifp) {
1768 			if (ifa->ifa_addr->sa_family != AF_INET6)
1769 				continue;
1770 			ia6 = (struct in6_ifaddr *)ifa;
1771 
1772 			if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1773 			    allow_deprecated == 0) {
1774 				/*
1775 				 * prefererred address should be put before
1776 				 * deprecated addresses.
1777 				 */
1778 
1779 				/* record the interface for later search */
1780 				if (ifp_dep == NULL)
1781 					ifp_dep = ifp;
1782 
1783 				continue;
1784 			}
1785 			else if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1786 				 allow_deprecated != 0)
1787 				continue; /* we now collect deprecated addrs */
1788 
1789 			/* What do we have to do about ::1? */
1790 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1791 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1792 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1793 					continue;
1794 				break;
1795 			case IPV6_ADDR_SCOPE_SITELOCAL:
1796 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1797 					continue;
1798 				break;
1799 			case IPV6_ADDR_SCOPE_GLOBAL:
1800 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1801 					continue;
1802 				break;
1803 			default:
1804 				continue;
1805 			}
1806 
1807 			/*
1808 			 * check if anycast is okay.
1809 			 * XXX: just experimental.  not in the spec.
1810 			 */
1811 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1812 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1813 				continue;
1814 
1815 			/* now we can copy the address */
1816 			if (resid < sizeof(struct in6_addr) +
1817 			    sizeof(u_int32_t)) {
1818 				/*
1819 				 * We give up much more copy.
1820 				 * Set the truncate flag and return.
1821 				 */
1822 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1823 				return (copied);
1824 			}
1825 
1826 			/*
1827 			 * Set the TTL of the address.
1828 			 * The TTL value should be one of the following
1829 			 * according to the specification:
1830 			 *
1831 			 * 1. The remaining lifetime of a DHCP lease on the
1832 			 *    address, or
1833 			 * 2. The remaining Valid Lifetime of a prefix from
1834 			 *    which the address was derived through Stateless
1835 			 *    Autoconfiguration.
1836 			 *
1837 			 * Note that we currently do not support stateful
1838 			 * address configuration by DHCPv6, so the former
1839 			 * case can't happen.
1840 			 *
1841 			 * TTL must be 2^31 > TTL >= 0.
1842 			 */
1843 			if (ia6->ia6_lifetime.ia6t_expire == 0)
1844 				ltime = ND6_INFINITE_LIFETIME;
1845 			else {
1846 				if (ia6->ia6_lifetime.ia6t_expire >
1847 				    time_uptime)
1848 					ltime = ia6->ia6_lifetime.ia6t_expire -
1849 					    time_uptime;
1850 				else
1851 					ltime = 0;
1852 			}
1853 			if (ltime > 0x7fffffff)
1854 				ltime = 0x7fffffff;
1855 			ltime = htonl(ltime);
1856 
1857 			bcopy(&ltime, cp, sizeof(u_int32_t));
1858 			cp += sizeof(u_int32_t);
1859 
1860 			/* copy the address itself */
1861 			bcopy(&ia6->ia_addr.sin6_addr, cp,
1862 			      sizeof(struct in6_addr));
1863 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1864 			cp += sizeof(struct in6_addr);
1865 
1866 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1867 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1868 		}
1869 		if (ifp0)	/* we need search only on the specified IF */
1870 			break;
1871 	}
1872 
1873 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1874 		ifp = ifp_dep;
1875 		allow_deprecated = 1;
1876 
1877 		goto again;
1878 	}
1879 
1880 	return (copied);
1881 }
1882 
1883 /*
1884  * XXX almost dup'ed code with rip6_input.
1885  */
1886 static int
1887 icmp6_rip6_input(struct mbuf **mp, int off)
1888 {
1889 	struct mbuf *m = *mp;
1890 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1891 	struct inpcb_hdr *inph;
1892 	struct in6pcb *in6p;
1893 	struct in6pcb *last = NULL;
1894 	struct sockaddr_in6 rip6src;
1895 	struct icmp6_hdr *icmp6;
1896 	struct mbuf *opts = NULL;
1897 
1898 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1899 	if (icmp6 == NULL) {
1900 		/* m is already reclaimed */
1901 		return IPPROTO_DONE;
1902 	}
1903 
1904 	/*
1905 	 * XXX: the address may have embedded scope zone ID, which should be
1906 	 * hidden from applications.
1907 	 */
1908 	sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0);
1909 	if (sa6_recoverscope(&rip6src)) {
1910 		m_freem(m);
1911 		return (IPPROTO_DONE);
1912 	}
1913 
1914 	TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1915 		in6p = (struct in6pcb *)inph;
1916 		if (in6p->in6p_af != AF_INET6)
1917 			continue;
1918 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1919 			continue;
1920 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1921 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1922 			continue;
1923 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1924 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1925 			continue;
1926 		if (in6p->in6p_icmp6filt
1927 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1928 				 in6p->in6p_icmp6filt))
1929 			continue;
1930 		if (last) {
1931 			struct	mbuf *n;
1932 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1933 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1934 					ip6_savecontrol(last, &opts, ip6, n);
1935 				/* strip intermediate headers */
1936 				m_adj(n, off);
1937 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1938 						 (struct sockaddr *)&rip6src,
1939 						 n, opts) == 0) {
1940 					/* should notify about lost packet */
1941 					m_freem(n);
1942 					if (opts)
1943 						m_freem(opts);
1944 				} else
1945 					sorwakeup(last->in6p_socket);
1946 				opts = NULL;
1947 			}
1948 		}
1949 		last = in6p;
1950 	}
1951 	if (last) {
1952 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1953 			ip6_savecontrol(last, &opts, ip6, m);
1954 		/* strip intermediate headers */
1955 		m_adj(m, off);
1956 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1957 				(struct sockaddr *)&rip6src, m, opts) == 0) {
1958 			m_freem(m);
1959 			if (opts)
1960 				m_freem(opts);
1961 		} else
1962 			sorwakeup(last->in6p_socket);
1963 	} else {
1964 		m_freem(m);
1965 		IP6_STATDEC(IP6_STAT_DELIVERED);
1966 	}
1967 	return IPPROTO_DONE;
1968 }
1969 
1970 /*
1971  * Reflect the ip6 packet back to the source.
1972  * OFF points to the icmp6 header, counted from the top of the mbuf.
1973  *
1974  * Note: RFC 1885 required that an echo reply should be truncated if it
1975  * did not fit in with (return) path MTU, and KAME code supported the
1976  * behavior.  However, as a clarification after the RFC, this limitation
1977  * was removed in a revised version of the spec, RFC 2463.  We had kept the
1978  * old behavior, with a (non-default) ifdef block, while the new version of
1979  * the spec was an internet-draft status, and even after the new RFC was
1980  * published.  But it would rather make sense to clean the obsoleted part
1981  * up, and to make the code simpler at this stage.
1982  */
1983 void
1984 icmp6_reflect(struct mbuf *m, size_t off)
1985 {
1986 	struct ip6_hdr *ip6;
1987 	struct icmp6_hdr *icmp6;
1988 	const struct in6_ifaddr *ia;
1989 	const struct ip6aux *ip6a;
1990 	int plen;
1991 	int type, code;
1992 	struct ifnet *outif = NULL;
1993 	struct in6_addr origdst;
1994 	const struct in6_addr *src = NULL;
1995 	struct ifnet *rcvif;
1996 	int s;
1997 
1998 	/* too short to reflect */
1999 	if (off < sizeof(struct ip6_hdr)) {
2000 		nd6log(LOG_DEBUG,
2001 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2002 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2003 		    __FILE__, __LINE__);
2004 		goto bad;
2005 	}
2006 
2007 	/*
2008 	 * If there are extra headers between IPv6 and ICMPv6, strip
2009 	 * off that header first.
2010 	 */
2011 #ifdef DIAGNOSTIC
2012 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2013 		panic("assumption failed in icmp6_reflect");
2014 #endif
2015 	if (off > sizeof(struct ip6_hdr)) {
2016 		size_t l;
2017 		struct ip6_hdr nip6;
2018 
2019 		l = off - sizeof(struct ip6_hdr);
2020 		m_copydata(m, 0, sizeof(nip6), (void *)&nip6);
2021 		m_adj(m, l);
2022 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2023 		if (m->m_len < l) {
2024 			if ((m = m_pullup(m, l)) == NULL)
2025 				return;
2026 		}
2027 		bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6));
2028 	} else /* off == sizeof(struct ip6_hdr) */ {
2029 		size_t l;
2030 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2031 		if (m->m_len < l) {
2032 			if ((m = m_pullup(m, l)) == NULL)
2033 				return;
2034 		}
2035 	}
2036 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2037 	ip6 = mtod(m, struct ip6_hdr *);
2038 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2039 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2040 	type = icmp6->icmp6_type; /* keep type for statistics */
2041 	code = icmp6->icmp6_code; /* ditto. */
2042 
2043 	origdst = ip6->ip6_dst;
2044 	/*
2045 	 * ip6_input() drops a packet if its src is multicast.
2046 	 * So, the src is never multicast.
2047 	 */
2048 	ip6->ip6_dst = ip6->ip6_src;
2049 
2050 	/*
2051 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2052 	 * use dst as the src for the reply.
2053 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2054 	 * (for example) when we encounter an error while forwarding procedure
2055 	 * destined to a duplicated address of ours.
2056 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2057 	 * procedure of an outgoing packet of our own, in which case we need
2058 	 * to search in the ifaddr list.
2059 	 */
2060 	if (IN6_IS_ADDR_MULTICAST(&origdst))
2061 		;
2062 	else if ((ip6a = ip6_getdstifaddr(m)) != NULL) {
2063 		if ((ip6a->ip6a_flags &
2064 		     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0)
2065 			src = &ip6a->ip6a_src;
2066 	} else {
2067 		union {
2068 			struct sockaddr_in6 sin6;
2069 			struct sockaddr sa;
2070 		} u;
2071 
2072 		sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0);
2073 
2074 		ia = (struct in6_ifaddr *)ifa_ifwithaddr(&u.sa);
2075 
2076 		if (ia == NULL)
2077 			;
2078 		else if ((ia->ia6_flags &
2079 			 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0)
2080 			src = &ia->ia_addr.sin6_addr;
2081 	}
2082 
2083 	if (src == NULL) {
2084 		int e;
2085 		struct sockaddr_in6 sin6;
2086 		struct route ro;
2087 
2088 		/*
2089 		 * This case matches to multicasts, our anycast, or unicasts
2090 		 * that we do not own.  Select a source address based on the
2091 		 * source address of the erroneous packet.
2092 		 */
2093 		/* zone ID should be embedded */
2094 		sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
2095 
2096 		memset(&ro, 0, sizeof(ro));
2097 		src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, NULL, NULL, &e);
2098 		rtcache_free(&ro);
2099 		if (src == NULL) {
2100 			nd6log(LOG_DEBUG,
2101 			    "source can't be determined: "
2102 			    "dst=%s, error=%d\n",
2103 			    ip6_sprintf(&sin6.sin6_addr), e);
2104 			goto bad;
2105 		}
2106 	}
2107 
2108 	ip6->ip6_src = *src;
2109 	ip6->ip6_flow = 0;
2110 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2111 	ip6->ip6_vfc |= IPV6_VERSION;
2112 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2113 	rcvif = m_get_rcvif(m, &s);
2114 	if (rcvif) {
2115 		/* XXX: This may not be the outgoing interface */
2116 		ip6->ip6_hlim = ND_IFINFO(rcvif)->chlim;
2117 	} else
2118 		ip6->ip6_hlim = ip6_defhlim;
2119 	m_put_rcvif(rcvif, &s);
2120 
2121 	m->m_pkthdr.csum_flags = 0;
2122 	icmp6->icmp6_cksum = 0;
2123 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2124 					sizeof(struct ip6_hdr), plen);
2125 
2126 	/*
2127 	 * XXX option handling
2128 	 */
2129 
2130 	m->m_flags &= ~(M_BCAST|M_MCAST);
2131 
2132 	/*
2133 	 * To avoid a "too big" situation at an intermediate router
2134 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2135 	 * Note that only echo and node information replies are affected,
2136 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2137 	 */
2138 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif)
2139 	    != 0 && outif)
2140 		icmp6_ifstat_inc(outif, ifs6_out_error);
2141 	if (outif)
2142 		icmp6_ifoutstat_inc(outif, type, code);
2143 
2144 	return;
2145 
2146  bad:
2147 	m_freem(m);
2148 	return;
2149 }
2150 
2151 static const char *
2152 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2153 	struct in6_addr *tgt6)
2154 {
2155 	static char buf[1024];
2156 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2157 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2158 	return buf;
2159 }
2160 
2161 void
2162 icmp6_redirect_input(struct mbuf *m, int off)
2163 {
2164 	struct ifnet *ifp;
2165 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2166 	struct nd_redirect *nd_rd;
2167 	int icmp6len = ntohs(ip6->ip6_plen);
2168 	char *lladdr = NULL;
2169 	int lladdrlen = 0;
2170 	struct rtentry *rt = NULL;
2171 	int is_router;
2172 	int is_onlink;
2173 	struct in6_addr src6 = ip6->ip6_src;
2174 	struct in6_addr redtgt6;
2175 	struct in6_addr reddst6;
2176 	union nd_opts ndopts;
2177 	struct psref psref;
2178 
2179 	ifp = m_get_rcvif_psref(m, &psref);
2180 	if (ifp == NULL)
2181 		goto freeit;
2182 
2183 	/* XXX if we are router, we don't update route by icmp6 redirect */
2184 	if (ip6_forwarding)
2185 		goto freeit;
2186 	if (!icmp6_rediraccept)
2187 		goto freeit;
2188 
2189 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2190 	if (nd_rd == NULL) {
2191 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
2192 		m_put_rcvif_psref(ifp, &psref);
2193 		return;
2194 	}
2195 	redtgt6 = nd_rd->nd_rd_target;
2196 	reddst6 = nd_rd->nd_rd_dst;
2197 
2198 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2199 	    in6_setscope(&reddst6, ifp, NULL)) {
2200 		goto freeit;
2201 	}
2202 
2203 	/* validation */
2204 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2205 		nd6log(LOG_ERR,
2206 		    "ICMP6 redirect sent from %s rejected; "
2207 		    "must be from linklocal\n", ip6_sprintf(&src6));
2208 		goto bad;
2209 	}
2210 	if (ip6->ip6_hlim != 255) {
2211 		nd6log(LOG_ERR,
2212 		    "ICMP6 redirect sent from %s rejected; "
2213 		    "hlim=%d (must be 255)\n",
2214 		    ip6_sprintf(&src6), ip6->ip6_hlim);
2215 		goto bad;
2216 	}
2217     {
2218 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2219 	struct sockaddr_in6 sin6;
2220 	struct in6_addr *gw6;
2221 
2222 	sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0);
2223 	rt = rtalloc1((struct sockaddr *)&sin6, 0);
2224 	if (rt) {
2225 		if (rt->rt_gateway == NULL ||
2226 		    rt->rt_gateway->sa_family != AF_INET6) {
2227 			nd6log(LOG_ERR,
2228 			    "ICMP6 redirect rejected; no route "
2229 			    "with inet6 gateway found for redirect dst: %s\n",
2230 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2231 			rtfree(rt);
2232 			goto bad;
2233 		}
2234 
2235 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2236 		if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2237 			nd6log(LOG_ERR,
2238 			    "ICMP6 redirect rejected; "
2239 			    "not equal to gw-for-src=%s (must be same): %s\n",
2240 			    ip6_sprintf(gw6),
2241 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2242 			rtfree(rt);
2243 			goto bad;
2244 		}
2245 	} else {
2246 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2247 		    "no route found for redirect dst: %s\n",
2248 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2249 		goto bad;
2250 	}
2251 	rtfree(rt);
2252 	rt = NULL;
2253     }
2254 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2255 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2256 		    "redirect dst must be unicast: %s\n",
2257 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2258 		goto bad;
2259 	}
2260 
2261 	is_router = is_onlink = 0;
2262 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2263 		is_router = 1;	/* router case */
2264 	if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2265 		is_onlink = 1;	/* on-link destination case */
2266 	if (!is_router && !is_onlink) {
2267 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2268 		    "neither router case nor onlink case: %s\n",
2269 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2270 		goto bad;
2271 	}
2272 	/* validation passed */
2273 
2274 	icmp6len -= sizeof(*nd_rd);
2275 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2276 	if (nd6_options(&ndopts) < 0) {
2277 		nd6log(LOG_INFO, "invalid ND option, rejected: %s\n",
2278 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2279 		/* nd6_options have incremented stats */
2280 		goto freeit;
2281 	}
2282 
2283 	if (ndopts.nd_opts_tgt_lladdr) {
2284 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2285 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2286 	}
2287 
2288 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2289 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
2290 		    "(if %d, icmp6 packet %d): %s\n",
2291 		    ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2292 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2293 		goto bad;
2294 	}
2295 
2296 	/* RFC 2461 8.3 */
2297 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2298 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2299 
2300 	m_put_rcvif_psref(ifp, &psref);
2301 	ifp = NULL;
2302 
2303 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2304 		/* perform rtredirect */
2305 		struct sockaddr_in6 sdst;
2306 		struct sockaddr_in6 sgw;
2307 		struct sockaddr_in6 ssrc;
2308 		unsigned long rtcount;
2309 		struct rtentry *newrt = NULL;
2310 
2311 		/*
2312 		 * do not install redirect route, if the number of entries
2313 		 * is too much (> hiwat).  note that, the node (= host) will
2314 		 * work just fine even if we do not install redirect route
2315 		 * (there will be additional hops, though).
2316 		 */
2317 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2318 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes)
2319 			goto freeit;
2320 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat)
2321 			goto freeit;
2322 		else if (0 <= icmp6_redirect_lowat &&
2323 		    rtcount > icmp6_redirect_lowat) {
2324 			/*
2325 			 * XXX nuke a victim, install the new one.
2326 			 */
2327 		}
2328 
2329 		memset(&sdst, 0, sizeof(sdst));
2330 		memset(&sgw, 0, sizeof(sgw));
2331 		memset(&ssrc, 0, sizeof(ssrc));
2332 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2333 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2334 			sizeof(struct sockaddr_in6);
2335 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2336 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2337 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2338 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2339 			   NULL, RTF_GATEWAY | RTF_HOST,
2340 			   (struct sockaddr *)&ssrc,
2341 			   &newrt);
2342 
2343 		if (newrt) {
2344 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2345 			    icmp6_redirect_timeout_q);
2346 			rtfree(newrt);
2347 		}
2348 	}
2349 	/* finally update cached route in each socket via pfctlinput */
2350 	{
2351 		struct sockaddr_in6 sdst;
2352 
2353 		sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0);
2354 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2355 #if defined(IPSEC)
2356 		if (ipsec_used)
2357 			key_sa_routechange((struct sockaddr *)&sdst);
2358 #endif
2359 	}
2360 
2361  freeit:
2362 	if (ifp != NULL)
2363 		m_put_rcvif_psref(ifp, &psref);
2364 	m_freem(m);
2365 	return;
2366 
2367  bad:
2368 	m_put_rcvif_psref(ifp, &psref);
2369 	ICMP6_STATINC(ICMP6_STAT_BADREDIRECT);
2370 	m_freem(m);
2371 }
2372 
2373 void
2374 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2375 {
2376 	struct ifnet *ifp;	/* my outgoing interface */
2377 	struct in6_addr *ifp_ll6;
2378 	struct in6_addr *nexthop;
2379 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2380 	struct mbuf *m = NULL;	/* newly allocated one */
2381 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2382 	struct nd_redirect *nd_rd;
2383 	size_t maxlen;
2384 	u_char *p;
2385 	struct sockaddr_in6 src_sa;
2386 
2387 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0);
2388 
2389 	/* if we are not router, we don't send icmp6 redirect */
2390 	if (!ip6_forwarding)
2391 		goto fail;
2392 
2393 	/* sanity check */
2394 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2395 		goto fail;
2396 
2397 	/*
2398 	 * Address check:
2399 	 *  the source address must identify a neighbor, and
2400 	 *  the destination address must not be a multicast address
2401 	 *  [RFC 2461, sec 8.2]
2402 	 */
2403 	sip6 = mtod(m0, struct ip6_hdr *);
2404 	sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0);
2405 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2406 		goto fail;
2407 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2408 		goto fail;	/* what should we do here? */
2409 
2410 	/* rate limit */
2411 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2412 		goto fail;
2413 
2414 	/*
2415 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2416 	 * we almost always ask for an mbuf cluster for simplicity.
2417 	 * (MHLEN < IPV6_MMTU is almost always true)
2418 	 */
2419 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2420 	if (m && IPV6_MMTU >= MHLEN) {
2421 #if IPV6_MMTU >= MCLBYTES
2422 		_MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT);
2423 #else
2424 		MCLGET(m, M_DONTWAIT);
2425 #endif
2426 	}
2427 
2428 	if (!m)
2429 		goto fail;
2430 	m_reset_rcvif(m);
2431 	m->m_len = 0;
2432 	maxlen = M_TRAILINGSPACE(m);
2433 	maxlen = min(IPV6_MMTU, maxlen);
2434 	/* just for safety */
2435 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2436 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2437 		goto fail;
2438 	}
2439 
2440 	{
2441 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2442 		struct in6_ifaddr *ia;
2443 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2444 						 IN6_IFF_NOTREADY|
2445 						 IN6_IFF_ANYCAST)) == NULL)
2446 			goto fail;
2447 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2448 	}
2449 
2450 	/* get ip6 linklocal address for the router. */
2451 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2452 		struct sockaddr_in6 *sin6;
2453 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2454 		nexthop = &sin6->sin6_addr;
2455 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2456 			nexthop = NULL;
2457 	} else
2458 		nexthop = NULL;
2459 
2460 	/* ip6 */
2461 	ip6 = mtod(m, struct ip6_hdr *);
2462 	ip6->ip6_flow = 0;
2463 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2464 	ip6->ip6_vfc |= IPV6_VERSION;
2465 	/* ip6->ip6_plen will be set later */
2466 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2467 	ip6->ip6_hlim = 255;
2468 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2469 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2470 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2471 
2472 	/* ND Redirect */
2473 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2474 	nd_rd->nd_rd_type = ND_REDIRECT;
2475 	nd_rd->nd_rd_code = 0;
2476 	nd_rd->nd_rd_reserved = 0;
2477 	if (rt->rt_flags & RTF_GATEWAY) {
2478 		/*
2479 		 * nd_rd->nd_rd_target must be a link-local address in
2480 		 * better router cases.
2481 		 */
2482 		if (!nexthop)
2483 			goto fail;
2484 		bcopy(nexthop, &nd_rd->nd_rd_target,
2485 		      sizeof(nd_rd->nd_rd_target));
2486 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2487 		      sizeof(nd_rd->nd_rd_dst));
2488 	} else {
2489 		/* make sure redtgt == reddst */
2490 		nexthop = &sip6->ip6_dst;
2491 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2492 		      sizeof(nd_rd->nd_rd_target));
2493 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2494 		      sizeof(nd_rd->nd_rd_dst));
2495 	}
2496 
2497 	p = (u_char *)(nd_rd + 1);
2498 
2499 	{
2500 		/* target lladdr option */
2501 		struct llentry *ln = NULL;
2502 		int len;
2503 		struct nd_opt_hdr *nd_opt;
2504 		char *lladdr;
2505 
2506 		ln = nd6_lookup(nexthop, ifp, false);
2507 		if (ln == NULL)
2508 			goto nolladdropt;
2509 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2510 		len = (len + 7) & ~7;	/* round by 8 */
2511 		/* safety check */
2512 		if (len + (p - (u_char *)ip6) > maxlen) {
2513 			LLE_RUNLOCK(ln);
2514 			goto nolladdropt;
2515 		}
2516 		if (ln->la_flags & LLE_VALID) {
2517 			nd_opt = (struct nd_opt_hdr *)p;
2518 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2519 			nd_opt->nd_opt_len = len >> 3;
2520 			lladdr = (char *)(nd_opt + 1);
2521 			memcpy(lladdr, &ln->ll_addr, ifp->if_addrlen);
2522 			p += len;
2523 		}
2524 		LLE_RUNLOCK(ln);
2525 	}
2526   nolladdropt:;
2527 
2528 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2529 
2530 	/* just to be safe */
2531 	if (m0->m_flags & M_DECRYPTED)
2532 		goto noredhdropt;
2533 	if (p - (u_char *)ip6 > maxlen)
2534 		goto noredhdropt;
2535 
2536 	{
2537 		/* redirected header option */
2538 		int len;
2539 		struct nd_opt_rd_hdr *nd_opt_rh;
2540 
2541 		/*
2542 		 * compute the maximum size for icmp6 redirect header option.
2543 		 * XXX room for auth header?
2544 		 */
2545 		len = maxlen - (p - (u_char *)ip6);
2546 		len &= ~7;
2547 
2548 		/*
2549 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2550 		 * about padding/truncate rule for the original IP packet.
2551 		 * From the discussion on IPv6imp in Feb 1999,
2552 		 * the consensus was:
2553 		 * - "attach as much as possible" is the goal
2554 		 * - pad if not aligned (original size can be guessed by
2555 		 *   original ip6 header)
2556 		 * Following code adds the padding if it is simple enough,
2557 		 * and truncates if not.
2558 		 */
2559 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2560 			/* not enough room, truncate */
2561 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2562 			    m0->m_pkthdr.len);
2563 		} else {
2564 			/*
2565 			 * enough room, truncate if not aligned.
2566 			 * we don't pad here for simplicity.
2567 			 */
2568 			size_t extra;
2569 
2570 			extra = m0->m_pkthdr.len % 8;
2571 			if (extra) {
2572 				/* truncate */
2573 				m_adj(m0, -extra);
2574 			}
2575 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2576 		}
2577 
2578 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2579 		memset(nd_opt_rh, 0, sizeof(*nd_opt_rh));
2580 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2581 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2582 		p += sizeof(*nd_opt_rh);
2583 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2584 
2585 		/* connect m0 to m */
2586 		m->m_pkthdr.len += m0->m_pkthdr.len;
2587 		m_cat(m, m0);
2588 		m0 = NULL;
2589 	}
2590 noredhdropt:
2591 	if (m0) {
2592 		m_freem(m0);
2593 		m0 = NULL;
2594 	}
2595 
2596 	/* XXX: clear embedded link IDs in the inner header */
2597 	in6_clearscope(&sip6->ip6_src);
2598 	in6_clearscope(&sip6->ip6_dst);
2599 	in6_clearscope(&nd_rd->nd_rd_target);
2600 	in6_clearscope(&nd_rd->nd_rd_dst);
2601 
2602 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2603 
2604 	nd_rd->nd_rd_cksum = 0;
2605 	nd_rd->nd_rd_cksum
2606 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2607 
2608 	/* send the packet to outside... */
2609 	if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0)
2610 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2611 
2612 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2613 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2614 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT);
2615 
2616 	return;
2617 
2618 fail:
2619 	if (m)
2620 		m_freem(m);
2621 	if (m0)
2622 		m_freem(m0);
2623 }
2624 
2625 /*
2626  * ICMPv6 socket option processing.
2627  */
2628 int
2629 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2630 {
2631 	int error = 0;
2632 	struct in6pcb *in6p = sotoin6pcb(so);
2633 
2634 	if (sopt->sopt_level != IPPROTO_ICMPV6)
2635 		return rip6_ctloutput(op, so, sopt);
2636 
2637 	switch (op) {
2638 	case PRCO_SETOPT:
2639 		switch (sopt->sopt_name) {
2640 		case ICMP6_FILTER:
2641 		    {
2642 			struct icmp6_filter fil;
2643 
2644 			error = sockopt_get(sopt, &fil, sizeof(fil));
2645 			if (error)
2646 				break;
2647 			memcpy(in6p->in6p_icmp6filt, &fil,
2648 			    sizeof(struct icmp6_filter));
2649 			error = 0;
2650 			break;
2651 		    }
2652 
2653 		default:
2654 			error = ENOPROTOOPT;
2655 			break;
2656 		}
2657 		break;
2658 
2659 	case PRCO_GETOPT:
2660 		switch (sopt->sopt_name) {
2661 		case ICMP6_FILTER:
2662 		    {
2663 			if (in6p->in6p_icmp6filt == NULL) {
2664 				error = EINVAL;
2665 				break;
2666 			}
2667 			error = sockopt_set(sopt, in6p->in6p_icmp6filt,
2668 			    sizeof(struct icmp6_filter));
2669 			break;
2670 		    }
2671 
2672 		default:
2673 			error = ENOPROTOOPT;
2674 			break;
2675 		}
2676 		break;
2677 	}
2678 
2679 	return (error);
2680 }
2681 
2682 /*
2683  * Perform rate limit check.
2684  * Returns 0 if it is okay to send the icmp6 packet.
2685  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2686  * limitation.
2687  *
2688  * XXX per-destination/type check necessary?
2689  */
2690 static int
2691 icmp6_ratelimit(
2692 	const struct in6_addr *dst,	/* not used at this moment */
2693 	const int type,		/* not used at this moment */
2694 	const int code)		/* not used at this moment */
2695 {
2696 	int ret;
2697 
2698 	ret = 0;	/* okay to send */
2699 
2700 	/* PPS limit */
2701 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2702 	    icmp6errppslim)) {
2703 		/* The packet is subject to rate limit */
2704 		ret++;
2705 	}
2706 
2707 	return ret;
2708 }
2709 
2710 static struct rtentry *
2711 icmp6_mtudisc_clone(struct sockaddr *dst)
2712 {
2713 	struct rtentry *rt;
2714 	int    error;
2715 
2716 	rt = rtalloc1(dst, 1);
2717 	if (rt == 0)
2718 		return NULL;
2719 
2720 	/* If we didn't get a host route, allocate one */
2721 	if ((rt->rt_flags & RTF_HOST) == 0) {
2722 		struct rtentry *nrt;
2723 
2724 		error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL,
2725 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2726 		if (error) {
2727 			rtfree(rt);
2728 			return NULL;
2729 		}
2730 		nrt->rt_rmx = rt->rt_rmx;
2731 		rtfree(rt);
2732 		rt = nrt;
2733 	}
2734 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2735 			icmp6_mtudisc_timeout_q);
2736 	if (error) {
2737 		rtfree(rt);
2738 		return NULL;
2739 	}
2740 
2741 	return rt;	/* caller need to call rtfree() */
2742 }
2743 
2744 static void
2745 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
2746 {
2747 
2748 	KASSERT(rt != NULL);
2749 	rt_assert_referenced(rt);
2750 
2751 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2752 	    (RTF_DYNAMIC | RTF_HOST)) {
2753 		rtrequest(RTM_DELETE, rt_getkey(rt),
2754 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
2755 	} else {
2756 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2757 			rt->rt_rmx.rmx_mtu = 0;
2758 	}
2759 }
2760 
2761 static void
2762 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
2763 {
2764 
2765 	KASSERT(rt != NULL);
2766 	rt_assert_referenced(rt);
2767 
2768 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2769 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2770 		rtrequest(RTM_DELETE, rt_getkey(rt),
2771 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
2772 	}
2773 }
2774 
2775 /*
2776  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2777  */
2778 static int
2779 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2780 {
2781 	(void)&name;
2782 	(void)&l;
2783 	(void)&oname;
2784 
2785 	if (namelen != 0)
2786 		return (EINVAL);
2787 
2788 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2789 	    /*XXXUNCONST*/
2790 	    __UNCONST(newp), newlen));
2791 }
2792 
2793 static int
2794 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS)
2795 {
2796 
2797 	return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS));
2798 }
2799 
2800 static int
2801 sysctl_net_inet6_icmp6_redirtimeout(SYSCTLFN_ARGS)
2802 {
2803 	int error, tmp;
2804 	struct sysctlnode node;
2805 
2806 	node = *rnode;
2807 	node.sysctl_data = &tmp;
2808 	tmp = icmp6_redirtimeout;
2809 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2810 	if (error || newp == NULL)
2811 		return error;
2812 	if (tmp < 0)
2813 		return EINVAL;
2814 	icmp6_redirtimeout = tmp;
2815 
2816 	if (icmp6_redirect_timeout_q != NULL) {
2817 		if (icmp6_redirtimeout == 0) {
2818 			rt_timer_queue_destroy(icmp6_redirect_timeout_q,
2819 			    true);
2820 		} else {
2821 			rt_timer_queue_change(icmp6_redirect_timeout_q,
2822 			    icmp6_redirtimeout);
2823 		}
2824 	} else if (icmp6_redirtimeout > 0) {
2825 		icmp6_redirect_timeout_q =
2826 		    rt_timer_queue_create(icmp6_redirtimeout);
2827 	}
2828 
2829 	return 0;
2830 }
2831 
2832 static void
2833 sysctl_net_inet6_icmp6_setup(struct sysctllog **clog)
2834 {
2835 	extern int nd6_maxqueuelen; /* defined in nd6.c */
2836 
2837 	sysctl_createv(clog, 0, NULL, NULL,
2838 		       CTLFLAG_PERMANENT,
2839 		       CTLTYPE_NODE, "inet6", NULL,
2840 		       NULL, 0, NULL, 0,
2841 		       CTL_NET, PF_INET6, CTL_EOL);
2842 	sysctl_createv(clog, 0, NULL, NULL,
2843 		       CTLFLAG_PERMANENT,
2844 		       CTLTYPE_NODE, "icmp6",
2845 		       SYSCTL_DESCR("ICMPv6 related settings"),
2846 		       NULL, 0, NULL, 0,
2847 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2848 
2849 	sysctl_createv(clog, 0, NULL, NULL,
2850 		       CTLFLAG_PERMANENT,
2851 		       CTLTYPE_STRUCT, "stats",
2852 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2853 		       sysctl_net_inet6_icmp6_stats, 0, NULL, 0,
2854 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2855 		       ICMPV6CTL_STATS, CTL_EOL);
2856 	sysctl_createv(clog, 0, NULL, NULL,
2857 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2858 		       CTLTYPE_INT, "rediraccept",
2859 		       SYSCTL_DESCR("Accept and process redirect messages"),
2860 		       NULL, 0, &icmp6_rediraccept, 0,
2861 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2862 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2863 	sysctl_createv(clog, 0, NULL, NULL,
2864 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2865 		       CTLTYPE_INT, "redirtimeout",
2866 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2867 		       sysctl_net_inet6_icmp6_redirtimeout, 0,
2868 		       &icmp6_redirtimeout, 0,
2869 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2870 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2871 #if 0 /* obsoleted */
2872 	sysctl_createv(clog, 0, NULL, NULL,
2873 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2874 		       CTLTYPE_INT, "errratelimit", NULL,
2875 		       NULL, 0, &icmp6_errratelimit, 0,
2876 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2877 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2878 #endif
2879 	sysctl_createv(clog, 0, NULL, NULL,
2880 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2881 		       CTLTYPE_INT, "nd6_prune",
2882 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
2883 		       NULL, 0, &nd6_prune, 0,
2884 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2885 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
2886 	sysctl_createv(clog, 0, NULL, NULL,
2887 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2888 		       CTLTYPE_INT, "nd6_delay",
2889 		       SYSCTL_DESCR("First probe delay time"),
2890 		       NULL, 0, &nd6_delay, 0,
2891 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2892 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
2893 	sysctl_createv(clog, 0, NULL, NULL,
2894 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2895 		       CTLTYPE_INT, "nd6_umaxtries",
2896 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2897 		       NULL, 0, &nd6_umaxtries, 0,
2898 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2899 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
2900 	sysctl_createv(clog, 0, NULL, NULL,
2901 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2902 		       CTLTYPE_INT, "nd6_mmaxtries",
2903 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
2904 		       NULL, 0, &nd6_mmaxtries, 0,
2905 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2906 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
2907 	sysctl_createv(clog, 0, NULL, NULL,
2908 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2909 		       CTLTYPE_INT, "nd6_useloopback",
2910 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
2911 		       NULL, 0, &nd6_useloopback, 0,
2912 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2913 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
2914 #if 0 /* obsoleted */
2915 	sysctl_createv(clog, 0, NULL, NULL,
2916 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2917 		       CTLTYPE_INT, "nd6_proxyall", NULL,
2918 		       NULL, 0, &nd6_proxyall, 0,
2919 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2920 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
2921 #endif
2922 	sysctl_createv(clog, 0, NULL, NULL,
2923 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2924 		       CTLTYPE_INT, "nodeinfo",
2925 		       SYSCTL_DESCR("Respond to node information requests"),
2926 		       NULL, 0, &icmp6_nodeinfo, 0,
2927 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2928 		       ICMPV6CTL_NODEINFO, CTL_EOL);
2929 	sysctl_createv(clog, 0, NULL, NULL,
2930 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2931 		       CTLTYPE_INT, "errppslimit",
2932 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
2933 		       NULL, 0, &icmp6errppslim, 0,
2934 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2935 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
2936 	sysctl_createv(clog, 0, NULL, NULL,
2937 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2938 		       CTLTYPE_INT, "nd6_maxnudhint",
2939 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2940 		       NULL, 0, &nd6_maxnudhint, 0,
2941 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2942 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
2943 	sysctl_createv(clog, 0, NULL, NULL,
2944 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2945 		       CTLTYPE_INT, "mtudisc_hiwat",
2946 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2947 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
2948 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2949 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
2950 	sysctl_createv(clog, 0, NULL, NULL,
2951 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2952 		       CTLTYPE_INT, "mtudisc_lowat",
2953 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2954 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
2955 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2956 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
2957 	sysctl_createv(clog, 0, NULL, NULL,
2958 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2959 		       CTLTYPE_INT, "nd6_debug",
2960 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
2961 		       NULL, 0, &nd6_debug, 0,
2962 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2963 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
2964 	sysctl_createv(clog, 0, NULL, NULL,
2965 		       CTLFLAG_PERMANENT,
2966 		       CTLTYPE_STRUCT, "nd6_drlist",
2967 		       SYSCTL_DESCR("Default router list"),
2968 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2969 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2970 		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
2971 	sysctl_createv(clog, 0, NULL, NULL,
2972 		       CTLFLAG_PERMANENT,
2973 		       CTLTYPE_STRUCT, "nd6_prlist",
2974 		       SYSCTL_DESCR("Prefix list"),
2975 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2976 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2977 		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
2978 	sysctl_createv(clog, 0, NULL, NULL,
2979 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2980 		       CTLTYPE_INT, "maxqueuelen",
2981 		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
2982 		       NULL, 1, &nd6_maxqueuelen, 0,
2983 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2984 		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
2985 }
2986 
2987 void
2988 icmp6_statinc(u_int stat)
2989 {
2990 
2991 	KASSERT(stat < ICMP6_NSTATS);
2992 	ICMP6_STATINC(stat);
2993 }
2994