xref: /netbsd-src/sys/netinet6/nd6.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: nd6.c,v 1.152 2014/06/06 01:02:47 rmind Exp $	*/
2 /*	$KAME: nd6.c,v 1.279 2002/06/08 11:16:51 itojun 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 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: nd6.c,v 1.152 2014/06/06 01:02:47 rmind Exp $");
35 
36 #include "opt_ipsec.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/callout.h>
41 #include <sys/malloc.h>
42 #include <sys/mbuf.h>
43 #include <sys/socket.h>
44 #include <sys/socketvar.h>
45 #include <sys/sockio.h>
46 #include <sys/time.h>
47 #include <sys/kernel.h>
48 #include <sys/protosw.h>
49 #include <sys/errno.h>
50 #include <sys/ioctl.h>
51 #include <sys/syslog.h>
52 #include <sys/queue.h>
53 #include <sys/cprng.h>
54 
55 #include <net/if.h>
56 #include <net/if_dl.h>
57 #include <net/if_types.h>
58 #include <net/route.h>
59 #include <net/if_ether.h>
60 #include <net/if_fddi.h>
61 #include <net/if_arc.h>
62 
63 #include <netinet/in.h>
64 #include <netinet6/in6_var.h>
65 #include <netinet/ip6.h>
66 #include <netinet6/ip6_var.h>
67 #include <netinet6/scope6_var.h>
68 #include <netinet6/nd6.h>
69 #include <netinet6/in6_ifattach.h>
70 #include <netinet/icmp6.h>
71 #include <netinet6/icmp6_private.h>
72 
73 #include <net/net_osdep.h>
74 
75 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
76 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
77 
78 /* timer values */
79 int	nd6_prune	= 1;	/* walk list every 1 seconds */
80 int	nd6_delay	= 5;	/* delay first probe time 5 second */
81 int	nd6_umaxtries	= 3;	/* maximum unicast query */
82 int	nd6_mmaxtries	= 3;	/* maximum multicast query */
83 int	nd6_useloopback = 1;	/* use loopback interface for local traffic */
84 int	nd6_gctimer	= (60 * 60 * 24); /* 1 day: garbage collection timer */
85 
86 /* preventing too many loops in ND option parsing */
87 int nd6_maxndopt = 10;	/* max # of ND options allowed */
88 
89 int nd6_maxnudhint = 0;	/* max # of subsequent upper layer hints */
90 
91 int nd6_maxqueuelen = 1; /* max # of packets cached in unresolved ND entries */
92 
93 #ifdef ND6_DEBUG
94 int nd6_debug = 1;
95 #else
96 int nd6_debug = 0;
97 #endif
98 
99 /* for debugging? */
100 static int nd6_inuse, nd6_allocated;
101 
102 struct llinfo_nd6 llinfo_nd6 = {
103 	.ln_prev = &llinfo_nd6,
104 	.ln_next = &llinfo_nd6,
105 };
106 struct nd_drhead nd_defrouter;
107 struct nd_prhead nd_prefix = { 0 };
108 
109 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
110 static const struct sockaddr_in6 all1_sa = {
111 	  .sin6_family = AF_INET6
112 	, .sin6_len = sizeof(struct sockaddr_in6)
113 	, .sin6_addr = {.s6_addr = {0xff, 0xff, 0xff, 0xff,
114 				    0xff, 0xff, 0xff, 0xff,
115 				    0xff, 0xff, 0xff, 0xff,
116 				    0xff, 0xff, 0xff, 0xff}}
117 };
118 
119 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
120 static void nd6_slowtimo(void *);
121 static int regen_tmpaddr(struct in6_ifaddr *);
122 static struct llinfo_nd6 *nd6_free(struct rtentry *, int);
123 static void nd6_llinfo_timer(void *);
124 static void clear_llinfo_pqueue(struct llinfo_nd6 *);
125 
126 callout_t nd6_slowtimo_ch;
127 callout_t nd6_timer_ch;
128 extern callout_t in6_tmpaddrtimer_ch;
129 
130 static int fill_drlist(void *, size_t *, size_t);
131 static int fill_prlist(void *, size_t *, size_t);
132 
133 MALLOC_DEFINE(M_IP6NDP, "NDP", "IPv6 Neighbour Discovery");
134 
135 #define LN_DEQUEUE(ln) do { \
136 	(ln)->ln_next->ln_prev = (ln)->ln_prev; \
137 	(ln)->ln_prev->ln_next = (ln)->ln_next; \
138 	} while (/*CONSTCOND*/0)
139 #define LN_INSERTHEAD(ln) do { \
140 	(ln)->ln_next = llinfo_nd6.ln_next; \
141 	llinfo_nd6.ln_next = (ln); \
142 	(ln)->ln_prev = &llinfo_nd6; \
143 	(ln)->ln_next->ln_prev = (ln); \
144 	} while (/*CONSTCOND*/0)
145 void
146 nd6_init(void)
147 {
148 	static int nd6_init_done = 0;
149 
150 	if (nd6_init_done) {
151 		log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
152 		return;
153 	}
154 
155 	/* initialization of the default router list */
156 	TAILQ_INIT(&nd_defrouter);
157 
158 	nd6_init_done = 1;
159 
160 	callout_init(&nd6_slowtimo_ch, CALLOUT_MPSAFE);
161 	callout_init(&nd6_timer_ch, CALLOUT_MPSAFE);
162 
163 	/* start timer */
164 	callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
165 	    nd6_slowtimo, NULL);
166 }
167 
168 struct nd_ifinfo *
169 nd6_ifattach(struct ifnet *ifp)
170 {
171 	struct nd_ifinfo *nd;
172 
173 	nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO);
174 
175 	nd->initialized = 1;
176 
177 	nd->chlim = IPV6_DEFHLIM;
178 	nd->basereachable = REACHABLE_TIME;
179 	nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
180 	nd->retrans = RETRANS_TIMER;
181 
182 	nd->flags = ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV;
183 
184 	/* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL.
185 	 * A bridge interface should not have ND6_IFF_AUTO_LINKLOCAL
186 	 * because one of it's members should. */
187 	if ((ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) ||
188 	    (ifp->if_flags & IFF_LOOPBACK))
189 		nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
190 
191 	/* A loopback interface does not need to accept RTADV.
192 	 * A bridge interface should not accept RTADV
193 	 * because one of it's members should. */
194 	if (ip6_accept_rtadv &&
195 	    !(ifp->if_flags & IFF_LOOPBACK) &&
196 	    !(ifp->if_type != IFT_BRIDGE))
197 		nd->flags |= ND6_IFF_ACCEPT_RTADV;
198 
199 	/* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
200 	nd6_setmtu0(ifp, nd);
201 
202 	return nd;
203 }
204 
205 void
206 nd6_ifdetach(struct nd_ifinfo *nd)
207 {
208 
209 	free(nd, M_IP6NDP);
210 }
211 
212 void
213 nd6_setmtu(struct ifnet *ifp)
214 {
215 	nd6_setmtu0(ifp, ND_IFINFO(ifp));
216 }
217 
218 void
219 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
220 {
221 	u_int32_t omaxmtu;
222 
223 	omaxmtu = ndi->maxmtu;
224 
225 	switch (ifp->if_type) {
226 	case IFT_ARCNET:
227 		ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
228 		break;
229 	case IFT_FDDI:
230 		ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu);
231 		break;
232 	default:
233 		ndi->maxmtu = ifp->if_mtu;
234 		break;
235 	}
236 
237 	/*
238 	 * Decreasing the interface MTU under IPV6 minimum MTU may cause
239 	 * undesirable situation.  We thus notify the operator of the change
240 	 * explicitly.  The check for omaxmtu is necessary to restrict the
241 	 * log to the case of changing the MTU, not initializing it.
242 	 */
243 	if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
244 		log(LOG_NOTICE, "nd6_setmtu0: new link MTU on %s (%lu) is too"
245 		    " small for IPv6 which needs %lu\n",
246 		    if_name(ifp), (unsigned long)ndi->maxmtu, (unsigned long)
247 		    IPV6_MMTU);
248 	}
249 
250 	if (ndi->maxmtu > in6_maxmtu)
251 		in6_setmaxmtu(); /* check all interfaces just in case */
252 }
253 
254 void
255 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
256 {
257 
258 	memset(ndopts, 0, sizeof(*ndopts));
259 	ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
260 	ndopts->nd_opts_last
261 		= (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
262 
263 	if (icmp6len == 0) {
264 		ndopts->nd_opts_done = 1;
265 		ndopts->nd_opts_search = NULL;
266 	}
267 }
268 
269 /*
270  * Take one ND option.
271  */
272 struct nd_opt_hdr *
273 nd6_option(union nd_opts *ndopts)
274 {
275 	struct nd_opt_hdr *nd_opt;
276 	int olen;
277 
278 	if (ndopts == NULL)
279 		panic("ndopts == NULL in nd6_option");
280 	if (ndopts->nd_opts_last == NULL)
281 		panic("uninitialized ndopts in nd6_option");
282 	if (ndopts->nd_opts_search == NULL)
283 		return NULL;
284 	if (ndopts->nd_opts_done)
285 		return NULL;
286 
287 	nd_opt = ndopts->nd_opts_search;
288 
289 	/* make sure nd_opt_len is inside the buffer */
290 	if ((void *)&nd_opt->nd_opt_len >= (void *)ndopts->nd_opts_last) {
291 		memset(ndopts, 0, sizeof(*ndopts));
292 		return NULL;
293 	}
294 
295 	olen = nd_opt->nd_opt_len << 3;
296 	if (olen == 0) {
297 		/*
298 		 * Message validation requires that all included
299 		 * options have a length that is greater than zero.
300 		 */
301 		memset(ndopts, 0, sizeof(*ndopts));
302 		return NULL;
303 	}
304 
305 	ndopts->nd_opts_search = (struct nd_opt_hdr *)((char *)nd_opt + olen);
306 	if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
307 		/* option overruns the end of buffer, invalid */
308 		memset(ndopts, 0, sizeof(*ndopts));
309 		return NULL;
310 	} else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
311 		/* reached the end of options chain */
312 		ndopts->nd_opts_done = 1;
313 		ndopts->nd_opts_search = NULL;
314 	}
315 	return nd_opt;
316 }
317 
318 /*
319  * Parse multiple ND options.
320  * This function is much easier to use, for ND routines that do not need
321  * multiple options of the same type.
322  */
323 int
324 nd6_options(union nd_opts *ndopts)
325 {
326 	struct nd_opt_hdr *nd_opt;
327 	int i = 0;
328 
329 	if (ndopts == NULL)
330 		panic("ndopts == NULL in nd6_options");
331 	if (ndopts->nd_opts_last == NULL)
332 		panic("uninitialized ndopts in nd6_options");
333 	if (ndopts->nd_opts_search == NULL)
334 		return 0;
335 
336 	while (1) {
337 		nd_opt = nd6_option(ndopts);
338 		if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
339 			/*
340 			 * Message validation requires that all included
341 			 * options have a length that is greater than zero.
342 			 */
343 			ICMP6_STATINC(ICMP6_STAT_ND_BADOPT);
344 			memset(ndopts, 0, sizeof(*ndopts));
345 			return -1;
346 		}
347 
348 		if (nd_opt == NULL)
349 			goto skip1;
350 
351 		switch (nd_opt->nd_opt_type) {
352 		case ND_OPT_SOURCE_LINKADDR:
353 		case ND_OPT_TARGET_LINKADDR:
354 		case ND_OPT_MTU:
355 		case ND_OPT_REDIRECTED_HEADER:
356 			if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
357 				nd6log((LOG_INFO,
358 				    "duplicated ND6 option found (type=%d)\n",
359 				    nd_opt->nd_opt_type));
360 				/* XXX bark? */
361 			} else {
362 				ndopts->nd_opt_array[nd_opt->nd_opt_type]
363 					= nd_opt;
364 			}
365 			break;
366 		case ND_OPT_PREFIX_INFORMATION:
367 			if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
368 				ndopts->nd_opt_array[nd_opt->nd_opt_type]
369 					= nd_opt;
370 			}
371 			ndopts->nd_opts_pi_end =
372 				(struct nd_opt_prefix_info *)nd_opt;
373 			break;
374 		default:
375 			/*
376 			 * Unknown options must be silently ignored,
377 			 * to accommodate future extension to the protocol.
378 			 */
379 			nd6log((LOG_DEBUG,
380 			    "nd6_options: unsupported option %d - "
381 			    "option ignored\n", nd_opt->nd_opt_type));
382 		}
383 
384 skip1:
385 		i++;
386 		if (i > nd6_maxndopt) {
387 			ICMP6_STATINC(ICMP6_STAT_ND_TOOMANYOPT);
388 			nd6log((LOG_INFO, "too many loop in nd opt\n"));
389 			break;
390 		}
391 
392 		if (ndopts->nd_opts_done)
393 			break;
394 	}
395 
396 	return 0;
397 }
398 
399 /*
400  * ND6 timer routine to handle ND6 entries
401  */
402 void
403 nd6_llinfo_settimer(struct llinfo_nd6 *ln, long xtick)
404 {
405 	int s;
406 
407 	s = splsoftnet();
408 
409 	if (xtick < 0) {
410 		ln->ln_expire = 0;
411 		ln->ln_ntick = 0;
412 		callout_stop(&ln->ln_timer_ch);
413 	} else {
414 		ln->ln_expire = time_second + xtick / hz;
415 		if (xtick > INT_MAX) {
416 			ln->ln_ntick = xtick - INT_MAX;
417 			callout_reset(&ln->ln_timer_ch, INT_MAX,
418 			    nd6_llinfo_timer, ln);
419 		} else {
420 			ln->ln_ntick = 0;
421 			callout_reset(&ln->ln_timer_ch, xtick,
422 			    nd6_llinfo_timer, ln);
423 		}
424 	}
425 
426 	splx(s);
427 }
428 
429 static void
430 nd6_llinfo_timer(void *arg)
431 {
432 	struct llinfo_nd6 *ln;
433 	struct rtentry *rt;
434 	const struct sockaddr_in6 *dst;
435 	struct ifnet *ifp;
436 	struct nd_ifinfo *ndi = NULL;
437 
438 	mutex_enter(softnet_lock);
439 	KERNEL_LOCK(1, NULL);
440 
441 	ln = (struct llinfo_nd6 *)arg;
442 
443 	if (ln->ln_ntick > 0) {
444 		nd6_llinfo_settimer(ln, ln->ln_ntick);
445 		KERNEL_UNLOCK_ONE(NULL);
446 		mutex_exit(softnet_lock);
447 		return;
448 	}
449 
450 	if ((rt = ln->ln_rt) == NULL)
451 		panic("ln->ln_rt == NULL");
452 	if ((ifp = rt->rt_ifp) == NULL)
453 		panic("ln->ln_rt->rt_ifp == NULL");
454 	ndi = ND_IFINFO(ifp);
455 	dst = satocsin6(rt_getkey(rt));
456 
457 	/* sanity check */
458 	if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln)
459 		panic("rt_llinfo(%p) is not equal to ln(%p)",
460 		      rt->rt_llinfo, ln);
461 	if (!dst)
462 		panic("dst=0 in nd6_timer(ln=%p)", ln);
463 
464 	switch (ln->ln_state) {
465 	case ND6_LLINFO_INCOMPLETE:
466 		if (ln->ln_asked < nd6_mmaxtries) {
467 			ln->ln_asked++;
468 			nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
469 			nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
470 		} else {
471 			struct mbuf *m = ln->ln_hold;
472 			if (m) {
473 				struct mbuf *m0;
474 
475 				/*
476 				 * assuming every packet in ln_hold has
477 				 * the same IP header
478 				 */
479 				m0 = m->m_nextpkt;
480 				m->m_nextpkt = NULL;
481 				icmp6_error2(m, ICMP6_DST_UNREACH,
482 				    ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp);
483 
484 				ln->ln_hold = m0;
485 				clear_llinfo_pqueue(ln);
486  			}
487 			(void)nd6_free(rt, 0);
488 			ln = NULL;
489 		}
490 		break;
491 	case ND6_LLINFO_REACHABLE:
492 		if (!ND6_LLINFO_PERMANENT(ln)) {
493 			ln->ln_state = ND6_LLINFO_STALE;
494 			nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
495 		}
496 		break;
497 
498 	case ND6_LLINFO_PURGE:
499 	case ND6_LLINFO_STALE:
500 		/* Garbage Collection(RFC 2461 5.3) */
501 		if (!ND6_LLINFO_PERMANENT(ln)) {
502 			(void)nd6_free(rt, 1);
503 			ln = NULL;
504 		}
505 		break;
506 
507 	case ND6_LLINFO_DELAY:
508 		if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
509 			/* We need NUD */
510 			ln->ln_asked = 1;
511 			ln->ln_state = ND6_LLINFO_PROBE;
512 			nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
513 			nd6_ns_output(ifp, &dst->sin6_addr,
514 			    &dst->sin6_addr, ln, 0);
515 		} else {
516 			ln->ln_state = ND6_LLINFO_STALE; /* XXX */
517 			nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
518 		}
519 		break;
520 	case ND6_LLINFO_PROBE:
521 		if (ln->ln_asked < nd6_umaxtries) {
522 			ln->ln_asked++;
523 			nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000);
524 			nd6_ns_output(ifp, &dst->sin6_addr,
525 			    &dst->sin6_addr, ln, 0);
526 		} else {
527 			(void)nd6_free(rt, 0);
528 			ln = NULL;
529 		}
530 		break;
531 	}
532 
533 	KERNEL_UNLOCK_ONE(NULL);
534 	mutex_exit(softnet_lock);
535 }
536 
537 /*
538  * ND6 timer routine to expire default route list and prefix list
539  */
540 void
541 nd6_timer(void *ignored_arg)
542 {
543 	struct nd_defrouter *next_dr, *dr;
544 	struct nd_prefix *next_pr, *pr;
545 	struct in6_ifaddr *ia6, *nia6;
546 
547 	callout_reset(&nd6_timer_ch, nd6_prune * hz,
548 	    nd6_timer, NULL);
549 
550 	mutex_enter(softnet_lock);
551 	KERNEL_LOCK(1, NULL);
552 
553 	/* expire default router list */
554 
555 	TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, next_dr) {
556 		if (dr->expire && dr->expire < time_second) {
557 			defrtrlist_del(dr);
558 		}
559 	}
560 
561 	/*
562 	 * expire interface addresses.
563 	 * in the past the loop was inside prefix expiry processing.
564 	 * However, from a stricter speci-confrmance standpoint, we should
565 	 * rather separate address lifetimes and prefix lifetimes.
566 	 */
567   addrloop:
568 	for (ia6 = in6_ifaddr; ia6; ia6 = nia6) {
569 		nia6 = ia6->ia_next;
570 		/* check address lifetime */
571 		if (IFA6_IS_INVALID(ia6)) {
572 			int regen = 0;
573 
574 			/*
575 			 * If the expiring address is temporary, try
576 			 * regenerating a new one.  This would be useful when
577 			 * we suspended a laptop PC, then turned it on after a
578 			 * period that could invalidate all temporary
579 			 * addresses.  Although we may have to restart the
580 			 * loop (see below), it must be after purging the
581 			 * address.  Otherwise, we'd see an infinite loop of
582 			 * regeneration.
583 			 */
584 			if (ip6_use_tempaddr &&
585 			    (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
586 				if (regen_tmpaddr(ia6) == 0)
587 					regen = 1;
588 			}
589 
590  			in6_purgeaddr(&ia6->ia_ifa);
591 
592 			if (regen)
593 				goto addrloop; /* XXX: see below */
594 		} else if (IFA6_IS_DEPRECATED(ia6)) {
595 			int oldflags = ia6->ia6_flags;
596 
597 			if ((oldflags & IN6_IFF_DEPRECATED) == 0) {
598 				ia6->ia6_flags |= IN6_IFF_DEPRECATED;
599 				nd6_newaddrmsg((struct ifaddr *)ia6);
600 			}
601 
602 			/*
603 			 * If a temporary address has just become deprecated,
604 			 * regenerate a new one if possible.
605 			 */
606 			if (ip6_use_tempaddr &&
607 			    (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
608 			    (oldflags & IN6_IFF_DEPRECATED) == 0) {
609 
610 				if (regen_tmpaddr(ia6) == 0) {
611 					/*
612 					 * A new temporary address is
613 					 * generated.
614 					 * XXX: this means the address chain
615 					 * has changed while we are still in
616 					 * the loop.  Although the change
617 					 * would not cause disaster (because
618 					 * it's not a deletion, but an
619 					 * addition,) we'd rather restart the
620 					 * loop just for safety.  Or does this
621 					 * significantly reduce performance??
622 					 */
623 					goto addrloop;
624 				}
625 			}
626 		} else {
627 			/*
628 			 * A new RA might have made a deprecated address
629 			 * preferred.
630 			 */
631 			if (ia6->ia6_flags & IN6_IFF_DEPRECATED) {
632 				ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
633 				nd6_newaddrmsg((struct ifaddr *)ia6);
634 			}
635 		}
636 	}
637 
638 	/* expire prefix list */
639 	LIST_FOREACH_SAFE(pr, &nd_prefix, ndpr_entry, next_pr) {
640 		/*
641 		 * check prefix lifetime.
642 		 * since pltime is just for autoconf, pltime processing for
643 		 * prefix is not necessary.
644 		 */
645 		if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
646 		    time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) {
647 
648 			/*
649 			 * address expiration and prefix expiration are
650 			 * separate.  NEVER perform in6_purgeaddr here.
651 			 */
652 
653 			prelist_remove(pr);
654 		}
655 	}
656 
657 	KERNEL_UNLOCK_ONE(NULL);
658 	mutex_exit(softnet_lock);
659 }
660 
661 /* ia6: deprecated/invalidated temporary address */
662 static int
663 regen_tmpaddr(struct in6_ifaddr *ia6)
664 {
665 	struct ifaddr *ifa;
666 	struct ifnet *ifp;
667 	struct in6_ifaddr *public_ifa6 = NULL;
668 
669 	ifp = ia6->ia_ifa.ifa_ifp;
670 	IFADDR_FOREACH(ifa, ifp) {
671 		struct in6_ifaddr *it6;
672 
673 		if (ifa->ifa_addr->sa_family != AF_INET6)
674 			continue;
675 
676 		it6 = (struct in6_ifaddr *)ifa;
677 
678 		/* ignore no autoconf addresses. */
679 		if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
680 			continue;
681 
682 		/* ignore autoconf addresses with different prefixes. */
683 		if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
684 			continue;
685 
686 		/*
687 		 * Now we are looking at an autoconf address with the same
688 		 * prefix as ours.  If the address is temporary and is still
689 		 * preferred, do not create another one.  It would be rare, but
690 		 * could happen, for example, when we resume a laptop PC after
691 		 * a long period.
692 		 */
693 		if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
694 		    !IFA6_IS_DEPRECATED(it6)) {
695 			public_ifa6 = NULL;
696 			break;
697 		}
698 
699 		/*
700 		 * This is a public autoconf address that has the same prefix
701 		 * as ours.  If it is preferred, keep it.  We can't break the
702 		 * loop here, because there may be a still-preferred temporary
703 		 * address with the prefix.
704 		 */
705 		if (!IFA6_IS_DEPRECATED(it6))
706 		    public_ifa6 = it6;
707 	}
708 
709 	if (public_ifa6 != NULL) {
710 		int e;
711 
712 		/*
713 		 * Random factor is introduced in the preferred lifetime, so
714 		 * we do not need additional delay (3rd arg to in6_tmpifadd).
715 		 */
716 		if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
717 			log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
718 			    " tmp addr, errno=%d\n", e);
719 			return -1;
720 		}
721 		return 0;
722 	}
723 
724 	return -1;
725 }
726 
727 bool
728 nd6_accepts_rtadv(const struct nd_ifinfo *ndi)
729 {
730 	switch (ndi->flags & (ND6_IFF_ACCEPT_RTADV|ND6_IFF_OVERRIDE_RTADV)) {
731 	case ND6_IFF_OVERRIDE_RTADV|ND6_IFF_ACCEPT_RTADV:
732 		return true;
733 	case ND6_IFF_ACCEPT_RTADV:
734 		return ip6_accept_rtadv != 0;
735 	case ND6_IFF_OVERRIDE_RTADV:
736 	case 0:
737 	default:
738 		return false;
739 	}
740 }
741 
742 /*
743  * Nuke neighbor cache/prefix/default router management table, right before
744  * ifp goes away.
745  */
746 void
747 nd6_purge(struct ifnet *ifp)
748 {
749 	struct llinfo_nd6 *ln, *nln;
750 	struct nd_defrouter *dr, *ndr;
751 	struct nd_prefix *pr, *npr;
752 
753 	/*
754 	 * Nuke default router list entries toward ifp.
755 	 * We defer removal of default router list entries that is installed
756 	 * in the routing table, in order to keep additional side effects as
757 	 * small as possible.
758 	 */
759 	TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, ndr) {
760 		if (dr->installed)
761 			continue;
762 
763 		if (dr->ifp == ifp)
764 			defrtrlist_del(dr);
765 	}
766 
767 	TAILQ_FOREACH_SAFE(dr, &nd_defrouter, dr_entry, ndr) {
768 		if (!dr->installed)
769 			continue;
770 
771 		if (dr->ifp == ifp)
772 			defrtrlist_del(dr);
773 	}
774 
775 	/* Nuke prefix list entries toward ifp */
776 	LIST_FOREACH_SAFE(pr, &nd_prefix, ndpr_entry, npr) {
777 		if (pr->ndpr_ifp == ifp) {
778 			/*
779 			 * Because if_detach() does *not* release prefixes
780 			 * while purging addresses the reference count will
781 			 * still be above zero. We therefore reset it to
782 			 * make sure that the prefix really gets purged.
783 			 */
784 			pr->ndpr_refcnt = 0;
785 			/*
786 			 * Previously, pr->ndpr_addr is removed as well,
787 			 * but I strongly believe we don't have to do it.
788 			 * nd6_purge() is only called from in6_ifdetach(),
789 			 * which removes all the associated interface addresses
790 			 * by itself.
791 			 * (jinmei@kame.net 20010129)
792 			 */
793 			prelist_remove(pr);
794 		}
795 	}
796 
797 	/* cancel default outgoing interface setting */
798 	if (nd6_defifindex == ifp->if_index)
799 		nd6_setdefaultiface(0);
800 
801 	/* XXX: too restrictive? */
802 	if (!ip6_forwarding && ifp->if_afdata[AF_INET6]) {
803 		struct nd_ifinfo *ndi = ND_IFINFO(ifp);
804 		if (ndi && nd6_accepts_rtadv(ndi)) {
805 			/* refresh default router list */
806 			defrouter_select();
807 		}
808 	}
809 
810 	/*
811 	 * Nuke neighbor cache entries for the ifp.
812 	 * Note that rt->rt_ifp may not be the same as ifp,
813 	 * due to KAME goto ours hack.  See RTM_RESOLVE case in
814 	 * nd6_rtrequest(), and ip6_input().
815 	 */
816 	ln = llinfo_nd6.ln_next;
817 	while (ln != NULL && ln != &llinfo_nd6) {
818 		struct rtentry *rt;
819 		const struct sockaddr_dl *sdl;
820 
821 		nln = ln->ln_next;
822 		rt = ln->ln_rt;
823 		if (rt && rt->rt_gateway &&
824 		    rt->rt_gateway->sa_family == AF_LINK) {
825 			sdl = satocsdl(rt->rt_gateway);
826 			if (sdl->sdl_index == ifp->if_index)
827 				nln = nd6_free(rt, 0);
828 		}
829 		ln = nln;
830 	}
831 }
832 
833 static struct rtentry *
834 nd6_lookup1(const struct in6_addr *addr6, int create, struct ifnet *ifp,
835     int cloning)
836 {
837 	struct rtentry *rt;
838 	struct sockaddr_in6 sin6;
839 
840 	sockaddr_in6_init(&sin6, addr6, 0, 0, 0);
841 	rt = rtalloc1((struct sockaddr *)&sin6, create);
842 	if (rt != NULL && (rt->rt_flags & RTF_LLINFO) == 0) {
843 		/*
844 		 * This is the case for the default route.
845 		 * If we want to create a neighbor cache for the address, we
846 		 * should free the route for the destination and allocate an
847 		 * interface route.
848 		 */
849 		if (create) {
850 			rtfree(rt);
851 			rt = NULL;
852 		}
853 	}
854 	if (rt != NULL)
855 		;
856 	else if (create && ifp) {
857 		int e;
858 
859 		/*
860 		 * If no route is available and create is set,
861 		 * we allocate a host route for the destination
862 		 * and treat it like an interface route.
863 		 * This hack is necessary for a neighbor which can't
864 		 * be covered by our own prefix.
865 		 */
866 		struct ifaddr *ifa =
867 		    ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
868 		if (ifa == NULL)
869 			return NULL;
870 
871 		/*
872 		 * Create a new route.  RTF_LLINFO is necessary
873 		 * to create a Neighbor Cache entry for the
874 		 * destination in nd6_rtrequest which will be
875 		 * called in rtrequest via ifa->ifa_rtrequest.
876 		 */
877 		if ((e = rtrequest(RTM_ADD, (const struct sockaddr *)&sin6,
878 		    ifa->ifa_addr, (const struct sockaddr *)&all1_sa,
879 		    (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) &
880 		    ~RTF_CLONING, &rt)) != 0) {
881 #if 0
882 			log(LOG_ERR,
883 			    "nd6_lookup: failed to add route for a "
884 			    "neighbor(%s), errno=%d\n",
885 			    ip6_sprintf(addr6), e);
886 #endif
887 			return NULL;
888 		}
889 		if (rt == NULL)
890 			return NULL;
891 		if (rt->rt_llinfo) {
892 			struct llinfo_nd6 *ln =
893 			    (struct llinfo_nd6 *)rt->rt_llinfo;
894 			ln->ln_state = ND6_LLINFO_NOSTATE;
895 		}
896 	} else
897 		return NULL;
898 	rt->rt_refcnt--;
899 
900 	/*
901 	 * Check for a cloning route to match the address.
902 	 * This should only be set from in6_is_addr_neighbor so we avoid
903 	 * a potentially expensive second call to rtalloc1.
904 	 */
905 	if (cloning &&
906 	    rt->rt_flags & (RTF_CLONING | RTF_CLONED) &&
907 	    (rt->rt_ifp == ifp
908 #if NBRIDGE > 0
909 	    || SAME_BRIDGE(rt->rt_ifp->if_bridgeport, ifp->if_bridgeport)
910 #endif
911 #if NCARP > 0
912 	    || (ifp->if_type == IFT_CARP && rt->rt_ifp == ifp->if_carpdev) ||
913 	    (rt->rt_ifp->if_type == IFT_CARP && rt->rt_ifp->if_carpdev == ifp)||
914 	    (ifp->if_type == IFT_CARP && rt->rt_ifp->if_type == IFT_CARP &&
915 	    rt->rt_ifp->if_carpdev == ifp->if_carpdev)
916 #endif
917 	    ))
918 		return rt;
919 
920 	/*
921 	 * Validation for the entry.
922 	 * Note that the check for rt_llinfo is necessary because a cloned
923 	 * route from a parent route that has the L flag (e.g. the default
924 	 * route to a p2p interface) may have the flag, too, while the
925 	 * destination is not actually a neighbor.
926 	 * XXX: we can't use rt->rt_ifp to check for the interface, since
927 	 *      it might be the loopback interface if the entry is for our
928 	 *      own address on a non-loopback interface. Instead, we should
929 	 *      use rt->rt_ifa->ifa_ifp, which would specify the REAL
930 	 *	interface.
931 	 * Note also that ifa_ifp and ifp may differ when we connect two
932 	 * interfaces to a same link, install a link prefix to an interface,
933 	 * and try to install a neighbor cache on an interface that does not
934 	 * have a route to the prefix.
935 	 */
936 	if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
937 	    rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL ||
938 	    (ifp && rt->rt_ifa->ifa_ifp != ifp)) {
939 		if (create) {
940 			nd6log((LOG_DEBUG,
941 			    "nd6_lookup: failed to lookup %s (if = %s)\n",
942 			    ip6_sprintf(addr6),
943 			    ifp ? if_name(ifp) : "unspec"));
944 		}
945 		return NULL;
946 	}
947 	return rt;
948 }
949 
950 struct rtentry *
951 nd6_lookup(const struct in6_addr *addr6, int create, struct ifnet *ifp)
952 {
953 
954 	return nd6_lookup1(addr6, create, ifp, 0);
955 }
956 
957 /*
958  * Detect if a given IPv6 address identifies a neighbor on a given link.
959  * XXX: should take care of the destination of a p2p link?
960  */
961 int
962 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
963 {
964 	struct nd_prefix *pr;
965 
966 	/*
967 	 * A link-local address is always a neighbor.
968 	 * XXX: a link does not necessarily specify a single interface.
969 	 */
970 	if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
971 		struct sockaddr_in6 sin6_copy;
972 		u_int32_t zone;
973 
974 		/*
975 		 * We need sin6_copy since sa6_recoverscope() may modify the
976 		 * content (XXX).
977 		 */
978 		sin6_copy = *addr;
979 		if (sa6_recoverscope(&sin6_copy))
980 			return 0; /* XXX: should be impossible */
981 		if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
982 			return 0;
983 		if (sin6_copy.sin6_scope_id == zone)
984 			return 1;
985 		else
986 			return 0;
987 	}
988 
989 	/*
990 	 * If the address matches one of our on-link prefixes, it should be a
991 	 * neighbor.
992 	 */
993 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
994 		if (pr->ndpr_ifp != ifp)
995 			continue;
996 
997 		if (!(pr->ndpr_stateflags & NDPRF_ONLINK))
998 			continue;
999 
1000 		if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
1001 		    &addr->sin6_addr, &pr->ndpr_mask))
1002 			return 1;
1003 	}
1004 
1005 	/*
1006 	 * If the default router list is empty, all addresses are regarded
1007 	 * as on-link, and thus, as a neighbor.
1008 	 * XXX: we restrict the condition to hosts, because routers usually do
1009 	 * not have the "default router list".
1010 	 */
1011 	if (!ip6_forwarding && TAILQ_FIRST(&nd_defrouter) == NULL &&
1012 	    nd6_defifindex == ifp->if_index) {
1013 		return 1;
1014 	}
1015 
1016 	/*
1017 	 * Even if the address matches none of our addresses, it might match
1018 	 * a cloning route or be in the neighbor cache.
1019 	 */
1020 	if (nd6_lookup1(&addr->sin6_addr, 0, ifp, 1) != NULL)
1021 		return 1;
1022 
1023 	return 0;
1024 }
1025 
1026 /*
1027  * Free an nd6 llinfo entry.
1028  * Since the function would cause significant changes in the kernel, DO NOT
1029  * make it global, unless you have a strong reason for the change, and are sure
1030  * that the change is safe.
1031  */
1032 static struct llinfo_nd6 *
1033 nd6_free(struct rtentry *rt, int gc)
1034 {
1035 	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
1036 	struct in6_addr in6 = satocsin6(rt_getkey(rt))->sin6_addr;
1037 	struct nd_defrouter *dr;
1038 
1039 	/*
1040 	 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1041 	 * even though it is not harmful, it was not really necessary.
1042 	 */
1043 
1044 	/* cancel timer */
1045 	nd6_llinfo_settimer(ln, -1);
1046 
1047 	if (!ip6_forwarding) {
1048 		int s;
1049 		s = splsoftnet();
1050 		dr = defrouter_lookup(&satocsin6(rt_getkey(rt))->sin6_addr,
1051 		    rt->rt_ifp);
1052 
1053 		if (dr != NULL && dr->expire &&
1054 		    ln->ln_state == ND6_LLINFO_STALE && gc) {
1055 			/*
1056 			 * If the reason for the deletion is just garbage
1057 			 * collection, and the neighbor is an active default
1058 			 * router, do not delete it.  Instead, reset the GC
1059 			 * timer using the router's lifetime.
1060 			 * Simply deleting the entry would affect default
1061 			 * router selection, which is not necessarily a good
1062 			 * thing, especially when we're using router preference
1063 			 * values.
1064 			 * XXX: the check for ln_state would be redundant,
1065 			 *      but we intentionally keep it just in case.
1066 			 */
1067 			if (dr->expire > time_second)
1068 				nd6_llinfo_settimer(ln,
1069 				    (dr->expire - time_second) * hz);
1070 			else
1071 				nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
1072 			splx(s);
1073 			return ln->ln_next;
1074 		}
1075 
1076 		if (ln->ln_router || dr) {
1077 			/*
1078 			 * rt6_flush must be called whether or not the neighbor
1079 			 * is in the Default Router List.
1080 			 * See a corresponding comment in nd6_na_input().
1081 			 */
1082 			rt6_flush(&in6, rt->rt_ifp);
1083 		}
1084 
1085 		if (dr) {
1086 			/*
1087 			 * Unreachablity of a router might affect the default
1088 			 * router selection and on-link detection of advertised
1089 			 * prefixes.
1090 			 */
1091 
1092 			/*
1093 			 * Temporarily fake the state to choose a new default
1094 			 * router and to perform on-link determination of
1095 			 * prefixes correctly.
1096 			 * Below the state will be set correctly,
1097 			 * or the entry itself will be deleted.
1098 			 */
1099 			ln->ln_state = ND6_LLINFO_INCOMPLETE;
1100 
1101 			/*
1102 			 * Since defrouter_select() does not affect the
1103 			 * on-link determination and MIP6 needs the check
1104 			 * before the default router selection, we perform
1105 			 * the check now.
1106 			 */
1107 			pfxlist_onlink_check();
1108 
1109 			/*
1110 			 * refresh default router list
1111 			 */
1112 			defrouter_select();
1113 		}
1114 		splx(s);
1115 	}
1116 
1117 	/*
1118 	 * Before deleting the entry, remember the next entry as the
1119 	 * return value.  We need this because pfxlist_onlink_check() above
1120 	 * might have freed other entries (particularly the old next entry) as
1121 	 * a side effect (XXX).
1122 	 */
1123 	next = ln->ln_next;
1124 
1125 	/*
1126 	 * Detach the route from the routing tree and the list of neighbor
1127 	 * caches, and disable the route entry not to be used in already
1128 	 * cached routes.
1129 	 */
1130 	rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL);
1131 
1132 	return next;
1133 }
1134 
1135 /*
1136  * Upper-layer reachability hint for Neighbor Unreachability Detection.
1137  *
1138  * XXX cost-effective methods?
1139  */
1140 void
1141 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1142 {
1143 	struct llinfo_nd6 *ln;
1144 
1145 	/*
1146 	 * If the caller specified "rt", use that.  Otherwise, resolve the
1147 	 * routing table by supplied "dst6".
1148 	 */
1149 	if (rt == NULL) {
1150 		if (dst6 == NULL)
1151 			return;
1152 		if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL)
1153 			return;
1154 	}
1155 
1156 	if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
1157 	    (rt->rt_flags & RTF_LLINFO) == 0 ||
1158 	    !rt->rt_llinfo || !rt->rt_gateway ||
1159 	    rt->rt_gateway->sa_family != AF_LINK) {
1160 		/* This is not a host route. */
1161 		return;
1162 	}
1163 
1164 	ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1165 	if (ln->ln_state < ND6_LLINFO_REACHABLE)
1166 		return;
1167 
1168 	/*
1169 	 * if we get upper-layer reachability confirmation many times,
1170 	 * it is possible we have false information.
1171 	 */
1172 	if (!force) {
1173 		ln->ln_byhint++;
1174 		if (ln->ln_byhint > nd6_maxnudhint)
1175 			return;
1176 	}
1177 
1178 	ln->ln_state = ND6_LLINFO_REACHABLE;
1179 	if (!ND6_LLINFO_PERMANENT(ln)) {
1180 		nd6_llinfo_settimer(ln,
1181 		    (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
1182 	}
1183 }
1184 
1185 void
1186 nd6_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info)
1187 {
1188 	struct sockaddr *gate = rt->rt_gateway;
1189 	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1190 	struct ifnet *ifp = rt->rt_ifp;
1191 	uint8_t namelen = strlen(ifp->if_xname), addrlen = ifp->if_addrlen;
1192 	struct ifaddr *ifa;
1193 
1194 	RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1195 
1196 	if (req == RTM_LLINFO_UPD) {
1197 		int rc;
1198 		struct in6_addr *in6;
1199 		struct in6_addr in6_all;
1200 		int anycast;
1201 
1202 		if ((ifa = info->rti_ifa) == NULL)
1203 			return;
1204 
1205 		in6 = &ifatoia6(ifa)->ia_addr.sin6_addr;
1206 		anycast = ifatoia6(ifa)->ia6_flags & IN6_IFF_ANYCAST;
1207 
1208 		in6_all = in6addr_linklocal_allnodes;
1209 		if ((rc = in6_setscope(&in6_all, ifa->ifa_ifp, NULL)) != 0) {
1210 			log(LOG_ERR, "%s: failed to set scope %s "
1211 			    "(errno=%d)\n", __func__, if_name(ifp), rc);
1212 			return;
1213 		}
1214 
1215 		/* XXX don't set Override for proxy addresses */
1216 		nd6_na_output(ifa->ifa_ifp, &in6_all, in6,
1217 		    (anycast ? 0 : ND_NA_FLAG_OVERRIDE)
1218 #if 0
1219 		    | (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0)
1220 #endif
1221 		    , 1, NULL);
1222 		return;
1223 	}
1224 
1225 	if ((rt->rt_flags & RTF_GATEWAY) != 0)
1226 		return;
1227 
1228 	if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
1229 		RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1230 		/*
1231 		 * This is probably an interface direct route for a link
1232 		 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1233 		 * We do not need special treatment below for such a route.
1234 		 * Moreover, the RTF_LLINFO flag which would be set below
1235 		 * would annoy the ndp(8) command.
1236 		 */
1237 		return;
1238 	}
1239 
1240 	if (req == RTM_RESOLVE &&
1241 	    (nd6_need_cache(ifp) == 0 || /* stf case */
1242 	     !nd6_is_addr_neighbor(satocsin6(rt_getkey(rt)), ifp))) {
1243 		RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1244 		/*
1245 		 * FreeBSD and BSD/OS often make a cloned host route based
1246 		 * on a less-specific route (e.g. the default route).
1247 		 * If the less specific route does not have a "gateway"
1248 		 * (this is the case when the route just goes to a p2p or an
1249 		 * stf interface), we'll mistakenly make a neighbor cache for
1250 		 * the host route, and will see strange neighbor solicitation
1251 		 * for the corresponding destination.  In order to avoid the
1252 		 * confusion, we check if the destination of the route is
1253 		 * a neighbor in terms of neighbor discovery, and stop the
1254 		 * process if not.  Additionally, we remove the LLINFO flag
1255 		 * so that ndp(8) will not try to get the neighbor information
1256 		 * of the destination.
1257 		 */
1258 		rt->rt_flags &= ~RTF_LLINFO;
1259 		return;
1260 	}
1261 
1262 	switch (req) {
1263 	case RTM_ADD:
1264 		RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1265 		/*
1266 		 * There is no backward compatibility :)
1267 		 *
1268 		 * if ((rt->rt_flags & RTF_HOST) == 0 &&
1269 		 *     SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1270 		 *	   rt->rt_flags |= RTF_CLONING;
1271 		 */
1272 		if ((rt->rt_flags & RTF_CLONING) ||
1273 		    ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) {
1274 			union {
1275 				struct sockaddr sa;
1276 				struct sockaddr_dl sdl;
1277 				struct sockaddr_storage ss;
1278 			} u;
1279 			/*
1280 			 * Case 1: This route should come from a route to
1281 			 * interface (RTF_CLONING case) or the route should be
1282 			 * treated as on-link but is currently not
1283 			 * (RTF_LLINFO && ln == NULL case).
1284 			 */
1285 			if (sockaddr_dl_init(&u.sdl, sizeof(u.ss),
1286 			    ifp->if_index, ifp->if_type,
1287 			    NULL, namelen, NULL, addrlen) == NULL) {
1288 				printf("%s.%d: sockaddr_dl_init(, %zu, ) "
1289 				    "failed on %s\n", __func__, __LINE__,
1290 				    sizeof(u.ss), if_name(ifp));
1291 			}
1292 			rt_setgate(rt, &u.sa);
1293 			gate = rt->rt_gateway;
1294 			RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1295 			if (ln != NULL)
1296 				nd6_llinfo_settimer(ln, 0);
1297 			RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1298 			if ((rt->rt_flags & RTF_CLONING) != 0)
1299 				break;
1300 		}
1301 		RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1302 		/*
1303 		 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1304 		 * We don't do that here since llinfo is not ready yet.
1305 		 *
1306 		 * There are also couple of other things to be discussed:
1307 		 * - unsolicited NA code needs improvement beforehand
1308 		 * - RFC2461 says we MAY send multicast unsolicited NA
1309 		 *   (7.2.6 paragraph 4), however, it also says that we
1310 		 *   SHOULD provide a mechanism to prevent multicast NA storm.
1311 		 *   we don't have anything like it right now.
1312 		 *   note that the mechanism needs a mutual agreement
1313 		 *   between proxies, which means that we need to implement
1314 		 *   a new protocol, or a new kludge.
1315 		 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1316 		 *   we need to check ip6forwarding before sending it.
1317 		 *   (or should we allow proxy ND configuration only for
1318 		 *   routers?  there's no mention about proxy ND from hosts)
1319 		 */
1320 #if 0
1321 		/* XXX it does not work */
1322 		if (rt->rt_flags & RTF_ANNOUNCE)
1323 			nd6_na_output(ifp,
1324 			      &satocsin6(rt_getkey(rt))->sin6_addr,
1325 			      &satocsin6(rt_getkey(rt))->sin6_addr,
1326 			      ip6_forwarding ? ND_NA_FLAG_ROUTER : 0,
1327 			      1, NULL);
1328 #endif
1329 		/* FALLTHROUGH */
1330 	case RTM_RESOLVE:
1331 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
1332 			RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1333 			/*
1334 			 * Address resolution isn't necessary for a point to
1335 			 * point link, so we can skip this test for a p2p link.
1336 			 */
1337 			if (gate->sa_family != AF_LINK ||
1338 			    gate->sa_len <
1339 			    sockaddr_dl_measure(namelen, addrlen)) {
1340 				log(LOG_DEBUG,
1341 				    "nd6_rtrequest: bad gateway value: %s\n",
1342 				    if_name(ifp));
1343 				break;
1344 			}
1345 			satosdl(gate)->sdl_type = ifp->if_type;
1346 			satosdl(gate)->sdl_index = ifp->if_index;
1347 			RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1348 		}
1349 		if (ln != NULL)
1350 			break;	/* This happens on a route change */
1351 		RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1352 		/*
1353 		 * Case 2: This route may come from cloning, or a manual route
1354 		 * add with a LL address.
1355 		 */
1356 		R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
1357 		rt->rt_llinfo = ln;
1358 		RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1359 		if (ln == NULL) {
1360 			log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
1361 			break;
1362 		}
1363 		RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1364 		nd6_inuse++;
1365 		nd6_allocated++;
1366 		memset(ln, 0, sizeof(*ln));
1367 		ln->ln_rt = rt;
1368 		callout_init(&ln->ln_timer_ch, CALLOUT_MPSAFE);
1369 		/* this is required for "ndp" command. - shin */
1370 		if (req == RTM_ADD) {
1371 		        /*
1372 			 * gate should have some valid AF_LINK entry,
1373 			 * and ln->ln_expire should have some lifetime
1374 			 * which is specified by ndp command.
1375 			 */
1376 			ln->ln_state = ND6_LLINFO_REACHABLE;
1377 			ln->ln_byhint = 0;
1378 		} else {
1379 		        /*
1380 			 * When req == RTM_RESOLVE, rt is created and
1381 			 * initialized in rtrequest(), so rt_expire is 0.
1382 			 */
1383 			ln->ln_state = ND6_LLINFO_NOSTATE;
1384 			nd6_llinfo_settimer(ln, 0);
1385 		}
1386 		RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1387 		rt->rt_flags |= RTF_LLINFO;
1388 		ln->ln_next = llinfo_nd6.ln_next;
1389 		llinfo_nd6.ln_next = ln;
1390 		ln->ln_prev = &llinfo_nd6;
1391 		ln->ln_next->ln_prev = ln;
1392 
1393 		/*
1394 		 * If we have too many cache entries, initiate immediate
1395 		 * purging for some "less recently used" entries.  Note that
1396 		 * we cannot directly call nd6_free() here because it would
1397 		 * cause re-entering rtable related routines triggering an LOR
1398 		 * problem for FreeBSD.
1399 		 */
1400 		if (ip6_neighborgcthresh >= 0 &&
1401 		    nd6_inuse >= ip6_neighborgcthresh) {
1402 			int i;
1403 
1404 			for (i = 0; i < 10 && llinfo_nd6.ln_prev != ln; i++) {
1405 				struct llinfo_nd6 *ln_end = llinfo_nd6.ln_prev;
1406 
1407 				/* Move this entry to the head */
1408 				LN_DEQUEUE(ln_end);
1409 				LN_INSERTHEAD(ln_end);
1410 
1411 				if (ND6_LLINFO_PERMANENT(ln_end))
1412 					continue;
1413 
1414 				if (ln_end->ln_state > ND6_LLINFO_INCOMPLETE)
1415 					ln_end->ln_state = ND6_LLINFO_STALE;
1416 				else
1417 					ln_end->ln_state = ND6_LLINFO_PURGE;
1418 				nd6_llinfo_settimer(ln_end, 0);
1419 			}
1420 		}
1421 
1422 		RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1423 		/*
1424 		 * check if rt_getkey(rt) is an address assigned
1425 		 * to the interface.
1426 		 */
1427 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1428 		    &satocsin6(rt_getkey(rt))->sin6_addr);
1429 		RT_DPRINTF("rt_getkey(rt) = %p\n", rt_getkey(rt));
1430 		if (ifa != NULL) {
1431 			const void *mac;
1432 			nd6_llinfo_settimer(ln, -1);
1433 			ln->ln_state = ND6_LLINFO_REACHABLE;
1434 			ln->ln_byhint = 0;
1435 			if ((mac = nd6_ifptomac(ifp)) != NULL) {
1436 				/* XXX check for error */
1437 				if (sockaddr_dl_setaddr(satosdl(gate),
1438 				    gate->sa_len, mac,
1439 				    ifp->if_addrlen) == NULL) {
1440 					printf("%s.%d: "
1441 					    "sockaddr_dl_setaddr(, %d, ) "
1442 					    "failed on %s\n", __func__,
1443 					    __LINE__, gate->sa_len,
1444 					    if_name(ifp));
1445 				}
1446 			}
1447 			if (nd6_useloopback) {
1448 				ifp = rt->rt_ifp = lo0ifp;	/* XXX */
1449 				/*
1450 				 * Make sure rt_ifa be equal to the ifaddr
1451 				 * corresponding to the address.
1452 				 * We need this because when we refer
1453 				 * rt_ifa->ia6_flags in ip6_input, we assume
1454 				 * that the rt_ifa points to the address instead
1455 				 * of the loopback address.
1456 				 */
1457 				if (ifa != rt->rt_ifa)
1458 					rt_replace_ifa(rt, ifa);
1459 				rt->rt_flags &= ~RTF_CLONED;
1460 			}
1461 		} else if (rt->rt_flags & RTF_ANNOUNCE) {
1462 			nd6_llinfo_settimer(ln, -1);
1463 			ln->ln_state = ND6_LLINFO_REACHABLE;
1464 			ln->ln_byhint = 0;
1465 
1466 			/* join solicited node multicast for proxy ND */
1467 			if (ifp->if_flags & IFF_MULTICAST) {
1468 				struct in6_addr llsol;
1469 				int error;
1470 
1471 				llsol = satocsin6(rt_getkey(rt))->sin6_addr;
1472 				llsol.s6_addr32[0] = htonl(0xff020000);
1473 				llsol.s6_addr32[1] = 0;
1474 				llsol.s6_addr32[2] = htonl(1);
1475 				llsol.s6_addr8[12] = 0xff;
1476 				if (in6_setscope(&llsol, ifp, NULL))
1477 					break;
1478 				if (!in6_addmulti(&llsol, ifp, &error, 0)) {
1479 					nd6log((LOG_ERR, "%s: failed to join "
1480 					    "%s (errno=%d)\n", if_name(ifp),
1481 					    ip6_sprintf(&llsol), error));
1482 				}
1483 			}
1484 		}
1485 		break;
1486 
1487 	case RTM_DELETE:
1488 		if (ln == NULL)
1489 			break;
1490 		/* leave from solicited node multicast for proxy ND */
1491 		if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
1492 		    (ifp->if_flags & IFF_MULTICAST) != 0) {
1493 			struct in6_addr llsol;
1494 			struct in6_multi *in6m;
1495 
1496 			llsol = satocsin6(rt_getkey(rt))->sin6_addr;
1497 			llsol.s6_addr32[0] = htonl(0xff020000);
1498 			llsol.s6_addr32[1] = 0;
1499 			llsol.s6_addr32[2] = htonl(1);
1500 			llsol.s6_addr8[12] = 0xff;
1501 			if (in6_setscope(&llsol, ifp, NULL) == 0) {
1502 				IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1503 				if (in6m)
1504 					in6_delmulti(in6m);
1505 			}
1506 		}
1507 		nd6_inuse--;
1508 		ln->ln_next->ln_prev = ln->ln_prev;
1509 		ln->ln_prev->ln_next = ln->ln_next;
1510 		ln->ln_prev = NULL;
1511 		nd6_llinfo_settimer(ln, -1);
1512 		rt->rt_llinfo = 0;
1513 		rt->rt_flags &= ~RTF_LLINFO;
1514 		clear_llinfo_pqueue(ln);
1515 		Free(ln);
1516 	}
1517 }
1518 
1519 int
1520 nd6_ioctl(u_long cmd, void *data, struct ifnet *ifp)
1521 {
1522 	struct in6_drlist *drl = (struct in6_drlist *)data;
1523 	struct in6_oprlist *oprl = (struct in6_oprlist *)data;
1524 	struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1525 	struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1526 	struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1527 	struct nd_defrouter *dr;
1528 	struct nd_prefix *pr;
1529 	struct rtentry *rt;
1530 	int i = 0, error = 0;
1531 	int s;
1532 
1533 	switch (cmd) {
1534 	case SIOCGDRLST_IN6:
1535 		/*
1536 		 * obsolete API, use sysctl under net.inet6.icmp6
1537 		 */
1538 		memset(drl, 0, sizeof(*drl));
1539 		s = splsoftnet();
1540 		TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1541 			if (i >= DRLSTSIZ)
1542 				break;
1543 			drl->defrouter[i].rtaddr = dr->rtaddr;
1544 			in6_clearscope(&drl->defrouter[i].rtaddr);
1545 
1546 			drl->defrouter[i].flags = dr->flags;
1547 			drl->defrouter[i].rtlifetime = dr->rtlifetime;
1548 			drl->defrouter[i].expire = dr->expire;
1549 			drl->defrouter[i].if_index = dr->ifp->if_index;
1550 			i++;
1551 		}
1552 		splx(s);
1553 		break;
1554 	case SIOCGPRLST_IN6:
1555 		/*
1556 		 * obsolete API, use sysctl under net.inet6.icmp6
1557 		 *
1558 		 * XXX the structure in6_prlist was changed in backward-
1559 		 * incompatible manner.  in6_oprlist is used for SIOCGPRLST_IN6,
1560 		 * in6_prlist is used for nd6_sysctl() - fill_prlist().
1561 		 */
1562 		/*
1563 		 * XXX meaning of fields, especialy "raflags", is very
1564 		 * differnet between RA prefix list and RR/static prefix list.
1565 		 * how about separating ioctls into two?
1566 		 */
1567 		memset(oprl, 0, sizeof(*oprl));
1568 		s = splsoftnet();
1569 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1570 			struct nd_pfxrouter *pfr;
1571 			int j;
1572 
1573 			if (i >= PRLSTSIZ)
1574 				break;
1575 			oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
1576 			oprl->prefix[i].raflags = pr->ndpr_raf;
1577 			oprl->prefix[i].prefixlen = pr->ndpr_plen;
1578 			oprl->prefix[i].vltime = pr->ndpr_vltime;
1579 			oprl->prefix[i].pltime = pr->ndpr_pltime;
1580 			oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1581 			if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1582 				oprl->prefix[i].expire = 0;
1583 			else {
1584 				time_t maxexpire;
1585 
1586 				/* XXX: we assume time_t is signed. */
1587 				maxexpire = (-1) &
1588 				    ~((time_t)1 <<
1589 				    ((sizeof(maxexpire) * 8) - 1));
1590 				if (pr->ndpr_vltime <
1591 				    maxexpire - pr->ndpr_lastupdate) {
1592 					oprl->prefix[i].expire =
1593 						 pr->ndpr_lastupdate +
1594 						pr->ndpr_vltime;
1595 				} else
1596 					oprl->prefix[i].expire = maxexpire;
1597 			}
1598 
1599 			j = 0;
1600 			LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
1601 				if (j < DRLSTSIZ) {
1602 #define RTRADDR oprl->prefix[i].advrtr[j]
1603 					RTRADDR = pfr->router->rtaddr;
1604 					in6_clearscope(&RTRADDR);
1605 #undef RTRADDR
1606 				}
1607 				j++;
1608 			}
1609 			oprl->prefix[i].advrtrs = j;
1610 			oprl->prefix[i].origin = PR_ORIG_RA;
1611 
1612 			i++;
1613 		}
1614 		splx(s);
1615 
1616 		break;
1617 	case OSIOCGIFINFO_IN6:
1618 #define ND	ndi->ndi
1619 		/* XXX: old ndp(8) assumes a positive value for linkmtu. */
1620 		memset(&ND, 0, sizeof(ND));
1621 		ND.linkmtu = IN6_LINKMTU(ifp);
1622 		ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1623 		ND.basereachable = ND_IFINFO(ifp)->basereachable;
1624 		ND.reachable = ND_IFINFO(ifp)->reachable;
1625 		ND.retrans = ND_IFINFO(ifp)->retrans;
1626 		ND.flags = ND_IFINFO(ifp)->flags;
1627 		ND.recalctm = ND_IFINFO(ifp)->recalctm;
1628 		ND.chlim = ND_IFINFO(ifp)->chlim;
1629 		break;
1630 	case SIOCGIFINFO_IN6:
1631 		ND = *ND_IFINFO(ifp);
1632 		break;
1633 	case SIOCSIFINFO_IN6:
1634 		/*
1635 		 * used to change host variables from userland.
1636 		 * intented for a use on router to reflect RA configurations.
1637 		 */
1638 		/* 0 means 'unspecified' */
1639 		if (ND.linkmtu != 0) {
1640 			if (ND.linkmtu < IPV6_MMTU ||
1641 			    ND.linkmtu > IN6_LINKMTU(ifp)) {
1642 				error = EINVAL;
1643 				break;
1644 			}
1645 			ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1646 		}
1647 
1648 		if (ND.basereachable != 0) {
1649 			int obasereachable = ND_IFINFO(ifp)->basereachable;
1650 
1651 			ND_IFINFO(ifp)->basereachable = ND.basereachable;
1652 			if (ND.basereachable != obasereachable)
1653 				ND_IFINFO(ifp)->reachable =
1654 				    ND_COMPUTE_RTIME(ND.basereachable);
1655 		}
1656 		if (ND.retrans != 0)
1657 			ND_IFINFO(ifp)->retrans = ND.retrans;
1658 		if (ND.chlim != 0)
1659 			ND_IFINFO(ifp)->chlim = ND.chlim;
1660 		/* FALLTHROUGH */
1661 	case SIOCSIFINFO_FLAGS:
1662 	{
1663 		struct ifaddr *ifa;
1664 		struct in6_ifaddr *ia;
1665 
1666 		if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1667 		    !(ND.flags & ND6_IFF_IFDISABLED))
1668 		{
1669 			/*
1670 			 * If the interface is marked as ND6_IFF_IFDISABLED and
1671 			 * has a link-local address with IN6_IFF_DUPLICATED,
1672 			 * do not clear ND6_IFF_IFDISABLED.
1673 			 * See RFC 4862, section 5.4.5.
1674 			 */
1675 			int duplicated_linklocal = 0;
1676 
1677 			IFADDR_FOREACH(ifa, ifp) {
1678 				if (ifa->ifa_addr->sa_family != AF_INET6)
1679 					continue;
1680 				ia = (struct in6_ifaddr *)ifa;
1681 				if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1682 				    IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1683 				{
1684 					duplicated_linklocal = 1;
1685 					break;
1686 				}
1687 			}
1688 
1689 			if (duplicated_linklocal) {
1690 				ND.flags |= ND6_IFF_IFDISABLED;
1691 				log(LOG_ERR, "Cannot enable an interface"
1692 				    " with a link-local address marked"
1693 				    " duplicate.\n");
1694 			} else {
1695 				ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1696 				if (ifp->if_flags & IFF_UP)
1697 					in6_if_up(ifp);
1698 			}
1699 		} else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1700 		    (ND.flags & ND6_IFF_IFDISABLED))
1701 		{
1702 			/* Mark all IPv6 addresses as tentative. */
1703 
1704 			ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1705 			IFADDR_FOREACH(ifa, ifp) {
1706 				if (ifa->ifa_addr->sa_family != AF_INET6)
1707 					continue;
1708 				nd6_dad_stop(ifa);
1709 				ia = (struct in6_ifaddr *)ifa;
1710 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
1711 			}
1712 		}
1713 
1714 		if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
1715 			if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
1716 				/* auto_linklocal 0->1 transition */
1717 
1718 				ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1719 				in6_ifattach(ifp, NULL);
1720 			} else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
1721 			    ifp->if_flags & IFF_UP)
1722 			{
1723 				/*
1724 				 * When the IF already has
1725 				 * ND6_IFF_AUTO_LINKLOCAL, no link-local
1726 				 * address is assigned, and IFF_UP, try to
1727 				 * assign one.
1728 				 */
1729 				 int haslinklocal = 0;
1730 
1731 				 IFADDR_FOREACH(ifa, ifp) {
1732 					if (ifa->ifa_addr->sa_family !=AF_INET6)
1733 						continue;
1734 					ia = (struct in6_ifaddr *)ifa;
1735 					if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))){
1736 						haslinklocal = 1;
1737 						break;
1738 					}
1739 				 }
1740 				 if (!haslinklocal)
1741 					in6_ifattach(ifp, NULL);
1742 			}
1743 		}
1744 	}
1745 		ND_IFINFO(ifp)->flags = ND.flags;
1746 		break;
1747 #undef ND
1748 	case SIOCSNDFLUSH_IN6:	/* XXX: the ioctl name is confusing... */
1749 		/* sync kernel routing table with the default router list */
1750 		defrouter_reset();
1751 		defrouter_select();
1752 		break;
1753 	case SIOCSPFXFLUSH_IN6:
1754 	{
1755 		/* flush all the prefix advertised by routers */
1756 		struct nd_prefix *pfx, *next;
1757 
1758 		s = splsoftnet();
1759 		LIST_FOREACH_SAFE(pfx, &nd_prefix, ndpr_entry, next) {
1760 			struct in6_ifaddr *ia, *ia_next;
1761 
1762 			if (IN6_IS_ADDR_LINKLOCAL(&pfx->ndpr_prefix.sin6_addr))
1763 				continue; /* XXX */
1764 
1765 			/* do we really have to remove addresses as well? */
1766 			for (ia = in6_ifaddr; ia; ia = ia_next) {
1767 				/* ia might be removed.  keep the next ptr. */
1768 				ia_next = ia->ia_next;
1769 
1770 				if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1771 					continue;
1772 
1773 				if (ia->ia6_ndpr == pfx)
1774 					in6_purgeaddr(&ia->ia_ifa);
1775 			}
1776 			prelist_remove(pfx);
1777 		}
1778 		splx(s);
1779 		break;
1780 	}
1781 	case SIOCSRTRFLUSH_IN6:
1782 	{
1783 		/* flush all the default routers */
1784 		struct nd_defrouter *drtr, *next;
1785 
1786 		s = splsoftnet();
1787 		defrouter_reset();
1788 		TAILQ_FOREACH_SAFE(drtr, &nd_defrouter, dr_entry, next) {
1789 			defrtrlist_del(drtr);
1790 		}
1791 		defrouter_select();
1792 		splx(s);
1793 		break;
1794 	}
1795 	case SIOCGNBRINFO_IN6:
1796 	{
1797 		struct llinfo_nd6 *ln;
1798 		struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1799 
1800 		if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1801 			return error;
1802 
1803 		s = splsoftnet();
1804 		if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL ||
1805 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) {
1806 			error = EINVAL;
1807 			splx(s);
1808 			break;
1809 		}
1810 		nbi->state = ln->ln_state;
1811 		nbi->asked = ln->ln_asked;
1812 		nbi->isrouter = ln->ln_router;
1813 		nbi->expire = ln->ln_expire;
1814 		splx(s);
1815 
1816 		break;
1817 	}
1818 	case SIOCGDEFIFACE_IN6:	/* XXX: should be implemented as a sysctl? */
1819 		ndif->ifindex = nd6_defifindex;
1820 		break;
1821 	case SIOCSDEFIFACE_IN6:	/* XXX: should be implemented as a sysctl? */
1822 		return nd6_setdefaultiface(ndif->ifindex);
1823 	}
1824 	return error;
1825 }
1826 
1827 void
1828 nd6_llinfo_release_pkts(struct llinfo_nd6 *ln, struct ifnet *ifp,
1829     struct rtentry *rt)
1830 {
1831 	struct mbuf *m_hold, *m_hold_next;
1832 
1833 	for (m_hold = ln->ln_hold, ln->ln_hold = NULL;
1834 	     m_hold != NULL;
1835 	     m_hold = m_hold_next) {
1836 		m_hold_next = m_hold->m_nextpkt;
1837 		m_hold->m_nextpkt = NULL;
1838 
1839 		/*
1840 		 * we assume ifp is not a p2p here, so
1841 		 * just set the 2nd argument as the
1842 		 * 1st one.
1843 		 */
1844 		nd6_output(ifp, ifp, m_hold, satocsin6(rt_getkey(rt)), rt);
1845 	}
1846 }
1847 
1848 /*
1849  * Create neighbor cache entry and cache link-layer address,
1850  * on reception of inbound ND6 packets.  (RS/RA/NS/redirect)
1851  */
1852 struct rtentry *
1853 nd6_cache_lladdr(
1854     struct ifnet *ifp,
1855     struct in6_addr *from,
1856     char *lladdr,
1857     int lladdrlen,
1858     int type,	/* ICMP6 type */
1859     int code	/* type dependent information */
1860 )
1861 {
1862 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
1863 	struct rtentry *rt = NULL;
1864 	struct llinfo_nd6 *ln = NULL;
1865 	int is_newentry;
1866 	struct sockaddr_dl *sdl = NULL;
1867 	int do_update;
1868 	int olladdr;
1869 	int llchange;
1870 	int newstate = 0;
1871 
1872 	if (ifp == NULL)
1873 		panic("ifp == NULL in nd6_cache_lladdr");
1874 	if (from == NULL)
1875 		panic("from == NULL in nd6_cache_lladdr");
1876 
1877 	/* nothing must be updated for unspecified address */
1878 	if (IN6_IS_ADDR_UNSPECIFIED(from))
1879 		return NULL;
1880 
1881 	/*
1882 	 * Validation about ifp->if_addrlen and lladdrlen must be done in
1883 	 * the caller.
1884 	 *
1885 	 * XXX If the link does not have link-layer adderss, what should
1886 	 * we do? (ifp->if_addrlen == 0)
1887 	 * Spec says nothing in sections for RA, RS and NA.  There's small
1888 	 * description on it in NS section (RFC 2461 7.2.3).
1889 	 */
1890 
1891 	rt = nd6_lookup(from, 0, ifp);
1892 	if (rt == NULL) {
1893 #if 0
1894 		/* nothing must be done if there's no lladdr */
1895 		if (!lladdr || !lladdrlen)
1896 			return NULL;
1897 #endif
1898 
1899 		rt = nd6_lookup(from, 1, ifp);
1900 		is_newentry = 1;
1901 	} else {
1902 		/* do nothing if static ndp is set */
1903 		if (rt->rt_flags & RTF_STATIC)
1904 			return NULL;
1905 		is_newentry = 0;
1906 	}
1907 
1908 	if (rt == NULL)
1909 		return NULL;
1910 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1911 fail:
1912 		(void)nd6_free(rt, 0);
1913 		return NULL;
1914 	}
1915 	ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1916 	if (ln == NULL)
1917 		goto fail;
1918 	if (rt->rt_gateway == NULL)
1919 		goto fail;
1920 	if (rt->rt_gateway->sa_family != AF_LINK)
1921 		goto fail;
1922 	sdl = satosdl(rt->rt_gateway);
1923 
1924 	olladdr = (sdl->sdl_alen) ? 1 : 0;
1925 	if (olladdr && lladdr) {
1926 		if (memcmp(lladdr, CLLADDR(sdl), ifp->if_addrlen))
1927 			llchange = 1;
1928 		else
1929 			llchange = 0;
1930 	} else
1931 		llchange = 0;
1932 
1933 	/*
1934 	 * newentry olladdr  lladdr  llchange	(*=record)
1935 	 *	0	n	n	--	(1)
1936 	 *	0	y	n	--	(2)
1937 	 *	0	n	y	--	(3) * STALE
1938 	 *	0	y	y	n	(4) *
1939 	 *	0	y	y	y	(5) * STALE
1940 	 *	1	--	n	--	(6)   NOSTATE(= PASSIVE)
1941 	 *	1	--	y	--	(7) * STALE
1942 	 */
1943 
1944 	if (lladdr) {		/* (3-5) and (7) */
1945 		/*
1946 		 * Record source link-layer address
1947 		 * XXX is it dependent to ifp->if_type?
1948 		 */
1949 		/* XXX check for error */
1950 		if (sockaddr_dl_setaddr(sdl, sdl->sdl_len, lladdr,
1951 		    ifp->if_addrlen) == NULL) {
1952 			printf("%s.%d: sockaddr_dl_setaddr(, %d, ) "
1953 			    "failed on %s\n", __func__, __LINE__,
1954 			    sdl->sdl_len, if_name(ifp));
1955 		}
1956 	}
1957 
1958 	if (!is_newentry) {
1959 		if ((!olladdr && lladdr) ||		/* (3) */
1960 		    (olladdr && lladdr && llchange)) {	/* (5) */
1961 			do_update = 1;
1962 			newstate = ND6_LLINFO_STALE;
1963 		} else					/* (1-2,4) */
1964 			do_update = 0;
1965 	} else {
1966 		do_update = 1;
1967 		if (lladdr == NULL)			/* (6) */
1968 			newstate = ND6_LLINFO_NOSTATE;
1969 		else					/* (7) */
1970 			newstate = ND6_LLINFO_STALE;
1971 	}
1972 
1973 	if (do_update) {
1974 		/*
1975 		 * Update the state of the neighbor cache.
1976 		 */
1977 		ln->ln_state = newstate;
1978 
1979 		if (ln->ln_state == ND6_LLINFO_STALE) {
1980 			/*
1981 			 * XXX: since nd6_output() below will cause
1982 			 * state tansition to DELAY and reset the timer,
1983 			 * we must set the timer now, although it is actually
1984 			 * meaningless.
1985 			 */
1986 			nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
1987 
1988 			nd6_llinfo_release_pkts(ln, ifp, rt);
1989 		} else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1990 			/* probe right away */
1991 			nd6_llinfo_settimer((void *)ln, 0);
1992 		}
1993 	}
1994 
1995 	/*
1996 	 * ICMP6 type dependent behavior.
1997 	 *
1998 	 * NS: clear IsRouter if new entry
1999 	 * RS: clear IsRouter
2000 	 * RA: set IsRouter if there's lladdr
2001 	 * redir: clear IsRouter if new entry
2002 	 *
2003 	 * RA case, (1):
2004 	 * The spec says that we must set IsRouter in the following cases:
2005 	 * - If lladdr exist, set IsRouter.  This means (1-5).
2006 	 * - If it is old entry (!newentry), set IsRouter.  This means (7).
2007 	 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
2008 	 * A quetion arises for (1) case.  (1) case has no lladdr in the
2009 	 * neighbor cache, this is similar to (6).
2010 	 * This case is rare but we figured that we MUST NOT set IsRouter.
2011 	 *
2012 	 * newentry olladdr  lladdr  llchange	    NS  RS  RA	redir
2013 	 *							D R
2014 	 *	0	n	n	--	(1)	c   ?     s
2015 	 *	0	y	n	--	(2)	c   s     s
2016 	 *	0	n	y	--	(3)	c   s     s
2017 	 *	0	y	y	n	(4)	c   s     s
2018 	 *	0	y	y	y	(5)	c   s     s
2019 	 *	1	--	n	--	(6) c	c 	c s
2020 	 *	1	--	y	--	(7) c	c   s	c s
2021 	 *
2022 	 *					(c=clear s=set)
2023 	 */
2024 	switch (type & 0xff) {
2025 	case ND_NEIGHBOR_SOLICIT:
2026 		/*
2027 		 * New entry must have is_router flag cleared.
2028 		 */
2029 		if (is_newentry)	/* (6-7) */
2030 			ln->ln_router = 0;
2031 		break;
2032 	case ND_REDIRECT:
2033 		/*
2034 		 * If the icmp is a redirect to a better router, always set the
2035 		 * is_router flag.  Otherwise, if the entry is newly created,
2036 		 * clear the flag.  [RFC 2461, sec 8.3]
2037 		 */
2038 		if (code == ND_REDIRECT_ROUTER)
2039 			ln->ln_router = 1;
2040 		else if (is_newentry) /* (6-7) */
2041 			ln->ln_router = 0;
2042 		break;
2043 	case ND_ROUTER_SOLICIT:
2044 		/*
2045 		 * is_router flag must always be cleared.
2046 		 */
2047 		ln->ln_router = 0;
2048 		break;
2049 	case ND_ROUTER_ADVERT:
2050 		/*
2051 		 * Mark an entry with lladdr as a router.
2052 		 */
2053 		if ((!is_newentry && (olladdr || lladdr)) ||	/* (2-5) */
2054 		    (is_newentry && lladdr)) {			/* (7) */
2055 			ln->ln_router = 1;
2056 		}
2057 		break;
2058 	}
2059 
2060 	/*
2061 	 * When the link-layer address of a router changes, select the
2062 	 * best router again.  In particular, when the neighbor entry is newly
2063 	 * created, it might affect the selection policy.
2064 	 * Question: can we restrict the first condition to the "is_newentry"
2065 	 * case?
2066 	 * XXX: when we hear an RA from a new router with the link-layer
2067 	 * address option, defrouter_select() is called twice, since
2068 	 * defrtrlist_update called the function as well.  However, I believe
2069 	 * we can compromise the overhead, since it only happens the first
2070 	 * time.
2071 	 * XXX: although defrouter_select() should not have a bad effect
2072 	 * for those are not autoconfigured hosts, we explicitly avoid such
2073 	 * cases for safety.
2074 	 */
2075 	if (do_update && ln->ln_router && !ip6_forwarding &&
2076 	    nd6_accepts_rtadv(ndi))
2077 		defrouter_select();
2078 
2079 	return rt;
2080 }
2081 
2082 static void
2083 nd6_slowtimo(void *ignored_arg)
2084 {
2085 	struct nd_ifinfo *nd6if;
2086 	struct ifnet *ifp;
2087 
2088 	mutex_enter(softnet_lock);
2089 	KERNEL_LOCK(1, NULL);
2090       	callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
2091 	    nd6_slowtimo, NULL);
2092 	IFNET_FOREACH(ifp) {
2093 		nd6if = ND_IFINFO(ifp);
2094 		if (nd6if->basereachable && /* already initialized */
2095 		    (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
2096 			/*
2097 			 * Since reachable time rarely changes by router
2098 			 * advertisements, we SHOULD insure that a new random
2099 			 * value gets recomputed at least once every few hours.
2100 			 * (RFC 2461, 6.3.4)
2101 			 */
2102 			nd6if->recalctm = nd6_recalc_reachtm_interval;
2103 			nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
2104 		}
2105 	}
2106 	KERNEL_UNLOCK_ONE(NULL);
2107 	mutex_exit(softnet_lock);
2108 }
2109 
2110 #define senderr(e) { error = (e); goto bad;}
2111 int
2112 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
2113     const struct sockaddr_in6 *dst, struct rtentry *rt0)
2114 {
2115 	struct mbuf *m = m0;
2116 	struct rtentry *rt = rt0;
2117 	struct sockaddr_in6 *gw6 = NULL;
2118 	struct llinfo_nd6 *ln = NULL;
2119 	int error = 0;
2120 
2121 	if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
2122 		goto sendpkt;
2123 
2124 	if (nd6_need_cache(ifp) == 0)
2125 		goto sendpkt;
2126 
2127 	/*
2128 	 * next hop determination.  This routine is derived from ether_output.
2129 	 */
2130 	if (rt) {
2131 		if ((rt->rt_flags & RTF_UP) == 0) {
2132 			if ((rt0 = rt = rtalloc1(sin6tocsa(dst), 1)) != NULL) {
2133 				rt->rt_refcnt--;
2134 				if (rt->rt_ifp != ifp)
2135 					senderr(EHOSTUNREACH);
2136 			} else
2137 				senderr(EHOSTUNREACH);
2138 		}
2139 
2140 		if (rt->rt_flags & RTF_GATEWAY) {
2141 			gw6 = (struct sockaddr_in6 *)rt->rt_gateway;
2142 
2143 			/*
2144 			 * We skip link-layer address resolution and NUD
2145 			 * if the gateway is not a neighbor from ND point
2146 			 * of view, regardless of the value of nd_ifinfo.flags.
2147 			 * The second condition is a bit tricky; we skip
2148 			 * if the gateway is our own address, which is
2149 			 * sometimes used to install a route to a p2p link.
2150 			 */
2151 			if (!nd6_is_addr_neighbor(gw6, ifp) ||
2152 			    in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) {
2153 				/*
2154 				 * We allow this kind of tricky route only
2155 				 * when the outgoing interface is p2p.
2156 				 * XXX: we may need a more generic rule here.
2157 				 */
2158 				if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
2159 					senderr(EHOSTUNREACH);
2160 
2161 				goto sendpkt;
2162 			}
2163 
2164 			if (rt->rt_gwroute == NULL)
2165 				goto lookup;
2166 			if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) {
2167 				rtfree(rt); rt = rt0;
2168 			lookup:
2169 				rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1);
2170 				if ((rt = rt->rt_gwroute) == NULL)
2171 					senderr(EHOSTUNREACH);
2172 				/* the "G" test below also prevents rt == rt0 */
2173 				if ((rt->rt_flags & RTF_GATEWAY) ||
2174 				    (rt->rt_ifp != ifp)) {
2175 					rt->rt_refcnt--;
2176 					rt0->rt_gwroute = NULL;
2177 					senderr(EHOSTUNREACH);
2178 				}
2179 			}
2180 		}
2181 	}
2182 
2183 	/*
2184 	 * Address resolution or Neighbor Unreachability Detection
2185 	 * for the next hop.
2186 	 * At this point, the destination of the packet must be a unicast
2187 	 * or an anycast address(i.e. not a multicast).
2188 	 */
2189 
2190 	/* Look up the neighbor cache for the nexthop */
2191 	if (rt != NULL && (rt->rt_flags & RTF_LLINFO) != 0)
2192 		ln = (struct llinfo_nd6 *)rt->rt_llinfo;
2193 	else {
2194 		/*
2195 		 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
2196 		 * the condition below is not very efficient.  But we believe
2197 		 * it is tolerable, because this should be a rare case.
2198 		 */
2199 		if (nd6_is_addr_neighbor(dst, ifp) &&
2200 		    (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL)
2201 			ln = (struct llinfo_nd6 *)rt->rt_llinfo;
2202 	}
2203 	if (ln == NULL || rt == NULL) {
2204 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
2205 		    !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
2206 			log(LOG_DEBUG,
2207 			    "nd6_output: can't allocate llinfo for %s "
2208 			    "(ln=%p, rt=%p)\n",
2209 			    ip6_sprintf(&dst->sin6_addr), ln, rt);
2210 			senderr(EIO);	/* XXX: good error? */
2211 		}
2212 
2213 		goto sendpkt;	/* send anyway */
2214 	}
2215 
2216 	/*
2217 	 * Move this entry to the head of the queue so that it is less likely
2218 	 * for this entry to be a target of forced garbage collection (see
2219 	 * nd6_rtrequest()).
2220 	 */
2221 	LN_DEQUEUE(ln);
2222 	LN_INSERTHEAD(ln);
2223 
2224 	/* We don't have to do link-layer address resolution on a p2p link. */
2225 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
2226 	    ln->ln_state < ND6_LLINFO_REACHABLE) {
2227 		ln->ln_state = ND6_LLINFO_STALE;
2228 		nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
2229 	}
2230 
2231 	/*
2232 	 * The first time we send a packet to a neighbor whose entry is
2233 	 * STALE, we have to change the state to DELAY and a sets a timer to
2234 	 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2235 	 * neighbor unreachability detection on expiration.
2236 	 * (RFC 2461 7.3.3)
2237 	 */
2238 	if (ln->ln_state == ND6_LLINFO_STALE) {
2239 		ln->ln_asked = 0;
2240 		ln->ln_state = ND6_LLINFO_DELAY;
2241 		nd6_llinfo_settimer(ln, (long)nd6_delay * hz);
2242 	}
2243 
2244 	/*
2245 	 * If the neighbor cache entry has a state other than INCOMPLETE
2246 	 * (i.e. its link-layer address is already resolved), just
2247 	 * send the packet.
2248 	 */
2249 	if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
2250 		goto sendpkt;
2251 
2252 	/*
2253 	 * There is a neighbor cache entry, but no ethernet address
2254 	 * response yet.  Append this latest packet to the end of the
2255 	 * packet queue in the mbuf, unless the number of the packet
2256 	 * does not exceed nd6_maxqueuelen.  When it exceeds nd6_maxqueuelen,
2257 	 * the oldest packet in the queue will be removed.
2258 	 */
2259 	if (ln->ln_state == ND6_LLINFO_NOSTATE)
2260 		ln->ln_state = ND6_LLINFO_INCOMPLETE;
2261 	if (ln->ln_hold) {
2262 		struct mbuf *m_hold;
2263 		int i;
2264 
2265 		i = 0;
2266 		for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) {
2267 			i++;
2268 			if (m_hold->m_nextpkt == NULL) {
2269 				m_hold->m_nextpkt = m;
2270 				break;
2271 			}
2272 		}
2273 		while (i >= nd6_maxqueuelen) {
2274 			m_hold = ln->ln_hold;
2275 			ln->ln_hold = ln->ln_hold->m_nextpkt;
2276 			m_freem(m_hold);
2277 			i--;
2278 		}
2279 	} else {
2280 		ln->ln_hold = m;
2281 	}
2282 
2283 	/*
2284 	 * If there has been no NS for the neighbor after entering the
2285 	 * INCOMPLETE state, send the first solicitation.
2286 	 */
2287 	if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) {
2288 		ln->ln_asked++;
2289 		nd6_llinfo_settimer(ln,
2290 		    (long)ND_IFINFO(ifp)->retrans * hz / 1000);
2291 		nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
2292 	}
2293 	return 0;
2294 
2295   sendpkt:
2296 	/* discard the packet if IPv6 operation is disabled on the interface */
2297 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
2298 		error = ENETDOWN; /* better error? */
2299 		goto bad;
2300 	}
2301 
2302 	KERNEL_LOCK(1, NULL);
2303 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2304 		error = (*ifp->if_output)(origifp, m, sin6tocsa(dst), rt);
2305 	else
2306 		error = (*ifp->if_output)(ifp, m, sin6tocsa(dst), rt);
2307 	KERNEL_UNLOCK_ONE(NULL);
2308 	return error;
2309 
2310   bad:
2311 	if (m != NULL)
2312 		m_freem(m);
2313 	return error;
2314 }
2315 #undef senderr
2316 
2317 int
2318 nd6_need_cache(struct ifnet *ifp)
2319 {
2320 	/*
2321 	 * XXX: we currently do not make neighbor cache on any interface
2322 	 * other than ARCnet, Ethernet, FDDI and GIF.
2323 	 *
2324 	 * RFC2893 says:
2325 	 * - unidirectional tunnels needs no ND
2326 	 */
2327 	switch (ifp->if_type) {
2328 	case IFT_ARCNET:
2329 	case IFT_ETHER:
2330 	case IFT_FDDI:
2331 	case IFT_IEEE1394:
2332 	case IFT_CARP:
2333 	case IFT_GIF:		/* XXX need more cases? */
2334 	case IFT_PPP:
2335 	case IFT_TUNNEL:
2336 		return 1;
2337 	default:
2338 		return 0;
2339 	}
2340 }
2341 
2342 int
2343 nd6_storelladdr(const struct ifnet *ifp, const struct rtentry *rt,
2344     struct mbuf *m, const struct sockaddr *dst, uint8_t *lldst,
2345     size_t dstsize)
2346 {
2347 	const struct sockaddr_dl *sdl;
2348 
2349 	if (m->m_flags & M_MCAST) {
2350 		switch (ifp->if_type) {
2351 		case IFT_ETHER:
2352 		case IFT_FDDI:
2353 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
2354 			    lldst);
2355 			return 1;
2356 		case IFT_IEEE1394:
2357 			memcpy(lldst, ifp->if_broadcastaddr,
2358 			    MIN(dstsize, ifp->if_addrlen));
2359 			return 1;
2360 		case IFT_ARCNET:
2361 			*lldst = 0;
2362 			return 1;
2363 		default:
2364 			m_freem(m);
2365 			return 0;
2366 		}
2367 	}
2368 
2369 	if (rt == NULL) {
2370 		/* this could happen, if we could not allocate memory */
2371 		m_freem(m);
2372 		return 0;
2373 	}
2374 	if (rt->rt_gateway->sa_family != AF_LINK) {
2375 		printf("%s: something odd happens\n", __func__);
2376 		m_freem(m);
2377 		return 0;
2378 	}
2379 	sdl = satocsdl(rt->rt_gateway);
2380 	if (sdl->sdl_alen == 0 || sdl->sdl_alen > dstsize) {
2381 		/* this should be impossible, but we bark here for debugging */
2382 		printf("%s: sdl_alen == %" PRIu8 ", dst=%s, if=%s\n", __func__,
2383 		    sdl->sdl_alen, ip6_sprintf(&satocsin6(dst)->sin6_addr),
2384 		    if_name(ifp));
2385 		m_freem(m);
2386 		return 0;
2387 	}
2388 
2389 	memcpy(lldst, CLLADDR(sdl), MIN(dstsize, sdl->sdl_alen));
2390 	return 1;
2391 }
2392 
2393 static void
2394 clear_llinfo_pqueue(struct llinfo_nd6 *ln)
2395 {
2396 	struct mbuf *m_hold, *m_hold_next;
2397 
2398 	for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) {
2399 		m_hold_next = m_hold->m_nextpkt;
2400 		m_hold->m_nextpkt = NULL;
2401 		m_freem(m_hold);
2402 	}
2403 
2404 	ln->ln_hold = NULL;
2405 	return;
2406 }
2407 
2408 int
2409 nd6_sysctl(
2410     int name,
2411     void *oldp,	/* syscall arg, need copyout */
2412     size_t *oldlenp,
2413     void *newp,	/* syscall arg, need copyin */
2414     size_t newlen
2415 )
2416 {
2417 	void *p;
2418 	size_t ol;
2419 	int error;
2420 
2421 	error = 0;
2422 
2423 	if (newp)
2424 		return EPERM;
2425 	if (oldp && !oldlenp)
2426 		return EINVAL;
2427 	ol = oldlenp ? *oldlenp : 0;
2428 
2429 	if (oldp) {
2430 		p = malloc(*oldlenp, M_TEMP, M_WAITOK);
2431 		if (p == NULL)
2432 			return ENOMEM;
2433 	} else
2434 		p = NULL;
2435 	switch (name) {
2436 	case ICMPV6CTL_ND6_DRLIST:
2437 		error = fill_drlist(p, oldlenp, ol);
2438 		if (!error && p != NULL && oldp != NULL)
2439 			error = copyout(p, oldp, *oldlenp);
2440 		break;
2441 
2442 	case ICMPV6CTL_ND6_PRLIST:
2443 		error = fill_prlist(p, oldlenp, ol);
2444 		if (!error && p != NULL && oldp != NULL)
2445 			error = copyout(p, oldp, *oldlenp);
2446 		break;
2447 
2448 	case ICMPV6CTL_ND6_MAXQLEN:
2449 		break;
2450 
2451 	default:
2452 		error = ENOPROTOOPT;
2453 		break;
2454 	}
2455 	if (p)
2456 		free(p, M_TEMP);
2457 
2458 	return error;
2459 }
2460 
2461 static int
2462 fill_drlist(void *oldp, size_t *oldlenp, size_t ol)
2463 {
2464 	int error = 0, s;
2465 	struct in6_defrouter *d = NULL, *de = NULL;
2466 	struct nd_defrouter *dr;
2467 	size_t l;
2468 
2469 	s = splsoftnet();
2470 
2471 	if (oldp) {
2472 		d = (struct in6_defrouter *)oldp;
2473 		de = (struct in6_defrouter *)((char *)oldp + *oldlenp);
2474 	}
2475 	l = 0;
2476 
2477 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
2478 
2479 		if (oldp && d + 1 <= de) {
2480 			memset(d, 0, sizeof(*d));
2481 			sockaddr_in6_init(&d->rtaddr, &dr->rtaddr, 0, 0, 0);
2482 			if (sa6_recoverscope(&d->rtaddr)) {
2483 				log(LOG_ERR,
2484 				    "scope error in router list (%s)\n",
2485 				    ip6_sprintf(&d->rtaddr.sin6_addr));
2486 				/* XXX: press on... */
2487 			}
2488 			d->flags = dr->flags;
2489 			d->rtlifetime = dr->rtlifetime;
2490 			d->expire = dr->expire;
2491 			d->if_index = dr->ifp->if_index;
2492 		}
2493 
2494 		l += sizeof(*d);
2495 		if (d)
2496 			d++;
2497 	}
2498 
2499 	if (oldp) {
2500 		if (l > ol)
2501 			error = ENOMEM;
2502 	}
2503 	if (oldlenp)
2504 		*oldlenp = l;	/* (void *)d - (void *)oldp */
2505 
2506 	splx(s);
2507 
2508 	return error;
2509 }
2510 
2511 static int
2512 fill_prlist(void *oldp, size_t *oldlenp, size_t ol)
2513 {
2514 	int error = 0, s;
2515 	struct nd_prefix *pr;
2516 	uint8_t *p = NULL, *ps = NULL;
2517 	uint8_t *pe = NULL;
2518 	size_t l;
2519 
2520 	s = splsoftnet();
2521 
2522 	if (oldp) {
2523 		ps = p = (uint8_t*)oldp;
2524 		pe = (uint8_t*)oldp + *oldlenp;
2525 	}
2526 	l = 0;
2527 
2528 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
2529 		u_short advrtrs;
2530 		struct sockaddr_in6 sin6;
2531 		struct nd_pfxrouter *pfr;
2532 		struct in6_prefix pfx;
2533 
2534 		if (oldp && p + sizeof(struct in6_prefix) <= pe)
2535 		{
2536 			memset(&pfx, 0, sizeof(pfx));
2537 			ps = p;
2538 			pfx.prefix = pr->ndpr_prefix;
2539 
2540 			if (sa6_recoverscope(&pfx.prefix)) {
2541 				log(LOG_ERR,
2542 				    "scope error in prefix list (%s)\n",
2543 				    ip6_sprintf(&pfx.prefix.sin6_addr));
2544 				/* XXX: press on... */
2545 			}
2546 			pfx.raflags = pr->ndpr_raf;
2547 			pfx.prefixlen = pr->ndpr_plen;
2548 			pfx.vltime = pr->ndpr_vltime;
2549 			pfx.pltime = pr->ndpr_pltime;
2550 			pfx.if_index = pr->ndpr_ifp->if_index;
2551 			if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2552 				pfx.expire = 0;
2553 			else {
2554 				time_t maxexpire;
2555 
2556 				/* XXX: we assume time_t is signed. */
2557 				maxexpire = (-1) &
2558 				    ~((time_t)1 <<
2559 				    ((sizeof(maxexpire) * 8) - 1));
2560 				if (pr->ndpr_vltime <
2561 				    maxexpire - pr->ndpr_lastupdate) {
2562 					pfx.expire = pr->ndpr_lastupdate +
2563 						pr->ndpr_vltime;
2564 				} else
2565 					pfx.expire = maxexpire;
2566 			}
2567 			pfx.refcnt = pr->ndpr_refcnt;
2568 			pfx.flags = pr->ndpr_stateflags;
2569 			pfx.origin = PR_ORIG_RA;
2570 
2571 			p += sizeof(pfx); l += sizeof(pfx);
2572 
2573 			advrtrs = 0;
2574 			LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2575 				if (p + sizeof(sin6) > pe) {
2576 					advrtrs++;
2577 					continue;
2578 				}
2579 
2580 				sockaddr_in6_init(&sin6, &pfr->router->rtaddr,
2581 				    0, 0, 0);
2582 				if (sa6_recoverscope(&sin6)) {
2583 					log(LOG_ERR,
2584 					    "scope error in "
2585 					    "prefix list (%s)\n",
2586 					    ip6_sprintf(&pfr->router->rtaddr));
2587 				}
2588 				advrtrs++;
2589 				memcpy(p, &sin6, sizeof(sin6));
2590 				p += sizeof(sin6);
2591 				l += sizeof(sin6);
2592 			}
2593 			pfx.advrtrs = advrtrs;
2594 			memcpy(ps, &pfx, sizeof(pfx));
2595 		}
2596 		else {
2597 			l += sizeof(pfx);
2598 			advrtrs = 0;
2599 			LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2600 				advrtrs++;
2601 				l += sizeof(sin6);
2602 			}
2603 		}
2604 	}
2605 
2606 	if (oldp) {
2607 		*oldlenp = l;	/* (void *)d - (void *)oldp */
2608 		if (l > ol)
2609 			error = ENOMEM;
2610 	} else
2611 		*oldlenp = l;
2612 
2613 	splx(s);
2614 
2615 	return error;
2616 }
2617