xref: /freebsd/sys/netinet6/nd6.c (revision 5ca34122ecdd5abc62bdae39663fec9ac8523d87)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.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/sockio.h>
45 #include <sys/time.h>
46 #include <sys/kernel.h>
47 #include <sys/protosw.h>
48 #include <sys/errno.h>
49 #include <sys/syslog.h>
50 #include <sys/lock.h>
51 #include <sys/rwlock.h>
52 #include <sys/queue.h>
53 #include <sys/sdt.h>
54 #include <sys/sysctl.h>
55 
56 #include <net/if.h>
57 #include <net/if_var.h>
58 #include <net/if_arc.h>
59 #include <net/if_dl.h>
60 #include <net/if_types.h>
61 #include <net/iso88025.h>
62 #include <net/fddi.h>
63 #include <net/route.h>
64 #include <net/vnet.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/in_kdtrace.h>
68 #include <net/if_llatbl.h>
69 #include <netinet/if_ether.h>
70 #include <netinet6/in6_var.h>
71 #include <netinet/ip6.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/scope6_var.h>
74 #include <netinet6/nd6.h>
75 #include <netinet6/in6_ifattach.h>
76 #include <netinet/icmp6.h>
77 #include <netinet6/send.h>
78 
79 #include <sys/limits.h>
80 
81 #include <security/mac/mac_framework.h>
82 
83 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
84 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
85 
86 #define SIN6(s) ((const struct sockaddr_in6 *)(s))
87 
88 /* timer values */
89 VNET_DEFINE(int, nd6_prune)	= 1;	/* walk list every 1 seconds */
90 VNET_DEFINE(int, nd6_delay)	= 5;	/* delay first probe time 5 second */
91 VNET_DEFINE(int, nd6_umaxtries)	= 3;	/* maximum unicast query */
92 VNET_DEFINE(int, nd6_mmaxtries)	= 3;	/* maximum multicast query */
93 VNET_DEFINE(int, nd6_useloopback) = 1;	/* use loopback interface for
94 					 * local traffic */
95 VNET_DEFINE(int, nd6_gctimer)	= (60 * 60 * 24); /* 1 day: garbage
96 					 * collection timer */
97 
98 /* preventing too many loops in ND option parsing */
99 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */
100 
101 VNET_DEFINE(int, nd6_maxnudhint) = 0;	/* max # of subsequent upper
102 					 * layer hints */
103 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved
104 					 * ND entries */
105 #define	V_nd6_maxndopt			VNET(nd6_maxndopt)
106 #define	V_nd6_maxqueuelen		VNET(nd6_maxqueuelen)
107 
108 #ifdef ND6_DEBUG
109 VNET_DEFINE(int, nd6_debug) = 1;
110 #else
111 VNET_DEFINE(int, nd6_debug) = 0;
112 #endif
113 
114 static eventhandler_tag lle_event_eh;
115 
116 /* for debugging? */
117 #if 0
118 static int nd6_inuse, nd6_allocated;
119 #endif
120 
121 VNET_DEFINE(struct nd_drhead, nd_defrouter);
122 VNET_DEFINE(struct nd_prhead, nd_prefix);
123 
124 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL;
125 #define	V_nd6_recalc_reachtm_interval	VNET(nd6_recalc_reachtm_interval)
126 
127 int	(*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int);
128 
129 static int nd6_is_new_addr_neighbor(const struct sockaddr_in6 *,
130 	struct ifnet *);
131 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
132 static void nd6_slowtimo(void *);
133 static int regen_tmpaddr(struct in6_ifaddr *);
134 static void nd6_free(struct llentry *, int);
135 static void nd6_free_redirect(const struct llentry *);
136 static void nd6_llinfo_timer(void *);
137 static void clear_llinfo_pqueue(struct llentry *);
138 static void nd6_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
139 static int nd6_resolve_slow(struct ifnet *, struct mbuf *,
140     const struct sockaddr_in6 *, u_char *, uint32_t *);
141 static int nd6_need_cache(struct ifnet *);
142 
143 
144 static VNET_DEFINE(struct callout, nd6_slowtimo_ch);
145 #define	V_nd6_slowtimo_ch		VNET(nd6_slowtimo_ch)
146 
147 VNET_DEFINE(struct callout, nd6_timer_ch);
148 
149 static void
150 nd6_lle_event(void *arg __unused, struct llentry *lle, int evt)
151 {
152 	struct rt_addrinfo rtinfo;
153 	struct sockaddr_in6 dst;
154 	struct sockaddr_dl gw;
155 	struct ifnet *ifp;
156 	int type;
157 
158 	LLE_WLOCK_ASSERT(lle);
159 
160 	if (lltable_get_af(lle->lle_tbl) != AF_INET6)
161 		return;
162 
163 	switch (evt) {
164 	case LLENTRY_RESOLVED:
165 		type = RTM_ADD;
166 		KASSERT(lle->la_flags & LLE_VALID,
167 		    ("%s: %p resolved but not valid?", __func__, lle));
168 		break;
169 	case LLENTRY_EXPIRED:
170 		type = RTM_DELETE;
171 		break;
172 	default:
173 		return;
174 	}
175 
176 	ifp = lltable_get_ifp(lle->lle_tbl);
177 
178 	bzero(&dst, sizeof(dst));
179 	bzero(&gw, sizeof(gw));
180 	bzero(&rtinfo, sizeof(rtinfo));
181 	lltable_fill_sa_entry(lle, (struct sockaddr *)&dst);
182 	dst.sin6_scope_id = in6_getscopezone(ifp,
183 	    in6_addrscope(&dst.sin6_addr));
184 	gw.sdl_len = sizeof(struct sockaddr_dl);
185 	gw.sdl_family = AF_LINK;
186 	gw.sdl_alen = ifp->if_addrlen;
187 	gw.sdl_index = ifp->if_index;
188 	gw.sdl_type = ifp->if_type;
189 	if (evt == LLENTRY_RESOLVED)
190 		bcopy(&lle->ll_addr, gw.sdl_data, ifp->if_addrlen);
191 	rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst;
192 	rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw;
193 	rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY;
194 	rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | (
195 	    type == RTM_ADD ? RTF_UP: 0), 0, RT_DEFAULT_FIB);
196 }
197 
198 void
199 nd6_init(void)
200 {
201 
202 	LIST_INIT(&V_nd_prefix);
203 
204 	/* initialization of the default router list */
205 	TAILQ_INIT(&V_nd_defrouter);
206 
207 	/* start timer */
208 	callout_init(&V_nd6_slowtimo_ch, 0);
209 	callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
210 	    nd6_slowtimo, curvnet);
211 
212 	nd6_dad_init();
213 	if (IS_DEFAULT_VNET(curvnet))
214 		lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event,
215 		    NULL, EVENTHANDLER_PRI_ANY);
216 }
217 
218 #ifdef VIMAGE
219 void
220 nd6_destroy()
221 {
222 
223 	callout_drain(&V_nd6_slowtimo_ch);
224 	callout_drain(&V_nd6_timer_ch);
225 	if (IS_DEFAULT_VNET(curvnet))
226 		EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh);
227 }
228 #endif
229 
230 struct nd_ifinfo *
231 nd6_ifattach(struct ifnet *ifp)
232 {
233 	struct nd_ifinfo *nd;
234 
235 	nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO);
236 	nd->initialized = 1;
237 
238 	nd->chlim = IPV6_DEFHLIM;
239 	nd->basereachable = REACHABLE_TIME;
240 	nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
241 	nd->retrans = RETRANS_TIMER;
242 
243 	nd->flags = ND6_IFF_PERFORMNUD;
244 
245 	/* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL.
246 	 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by
247 	 * default regardless of the V_ip6_auto_linklocal configuration to
248 	 * give a reasonable default behavior.
249 	 */
250 	if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) ||
251 	    (ifp->if_flags & IFF_LOOPBACK))
252 		nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
253 	/*
254 	 * A loopback interface does not need to accept RTADV.
255 	 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by
256 	 * default regardless of the V_ip6_accept_rtadv configuration to
257 	 * prevent the interface from accepting RA messages arrived
258 	 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV.
259 	 */
260 	if (V_ip6_accept_rtadv &&
261 	    !(ifp->if_flags & IFF_LOOPBACK) &&
262 	    (ifp->if_type != IFT_BRIDGE))
263 			nd->flags |= ND6_IFF_ACCEPT_RTADV;
264 	if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK))
265 		nd->flags |= ND6_IFF_NO_RADR;
266 
267 	/* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
268 	nd6_setmtu0(ifp, nd);
269 
270 	return nd;
271 }
272 
273 void
274 nd6_ifdetach(struct nd_ifinfo *nd)
275 {
276 
277 	free(nd, M_IP6NDP);
278 }
279 
280 /*
281  * Reset ND level link MTU. This function is called when the physical MTU
282  * changes, which means we might have to adjust the ND level MTU.
283  */
284 void
285 nd6_setmtu(struct ifnet *ifp)
286 {
287 
288 	nd6_setmtu0(ifp, ND_IFINFO(ifp));
289 }
290 
291 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
292 void
293 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
294 {
295 	u_int32_t omaxmtu;
296 
297 	omaxmtu = ndi->maxmtu;
298 
299 	switch (ifp->if_type) {
300 	case IFT_ARCNET:
301 		ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
302 		break;
303 	case IFT_FDDI:
304 		ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
305 		break;
306 	case IFT_ISO88025:
307 		 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
308 		 break;
309 	default:
310 		ndi->maxmtu = ifp->if_mtu;
311 		break;
312 	}
313 
314 	/*
315 	 * Decreasing the interface MTU under IPV6 minimum MTU may cause
316 	 * undesirable situation.  We thus notify the operator of the change
317 	 * explicitly.  The check for omaxmtu is necessary to restrict the
318 	 * log to the case of changing the MTU, not initializing it.
319 	 */
320 	if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
321 		log(LOG_NOTICE, "nd6_setmtu0: "
322 		    "new link MTU on %s (%lu) is too small for IPv6\n",
323 		    if_name(ifp), (unsigned long)ndi->maxmtu);
324 	}
325 
326 	if (ndi->maxmtu > V_in6_maxmtu)
327 		in6_setmaxmtu(); /* check all interfaces just in case */
328 
329 }
330 
331 void
332 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
333 {
334 
335 	bzero(ndopts, sizeof(*ndopts));
336 	ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
337 	ndopts->nd_opts_last
338 		= (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
339 
340 	if (icmp6len == 0) {
341 		ndopts->nd_opts_done = 1;
342 		ndopts->nd_opts_search = NULL;
343 	}
344 }
345 
346 /*
347  * Take one ND option.
348  */
349 struct nd_opt_hdr *
350 nd6_option(union nd_opts *ndopts)
351 {
352 	struct nd_opt_hdr *nd_opt;
353 	int olen;
354 
355 	KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
356 	KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
357 	    __func__));
358 	if (ndopts->nd_opts_search == NULL)
359 		return NULL;
360 	if (ndopts->nd_opts_done)
361 		return NULL;
362 
363 	nd_opt = ndopts->nd_opts_search;
364 
365 	/* make sure nd_opt_len is inside the buffer */
366 	if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
367 		bzero(ndopts, sizeof(*ndopts));
368 		return NULL;
369 	}
370 
371 	olen = nd_opt->nd_opt_len << 3;
372 	if (olen == 0) {
373 		/*
374 		 * Message validation requires that all included
375 		 * options have a length that is greater than zero.
376 		 */
377 		bzero(ndopts, sizeof(*ndopts));
378 		return NULL;
379 	}
380 
381 	ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
382 	if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
383 		/* option overruns the end of buffer, invalid */
384 		bzero(ndopts, sizeof(*ndopts));
385 		return NULL;
386 	} else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
387 		/* reached the end of options chain */
388 		ndopts->nd_opts_done = 1;
389 		ndopts->nd_opts_search = NULL;
390 	}
391 	return nd_opt;
392 }
393 
394 /*
395  * Parse multiple ND options.
396  * This function is much easier to use, for ND routines that do not need
397  * multiple options of the same type.
398  */
399 int
400 nd6_options(union nd_opts *ndopts)
401 {
402 	struct nd_opt_hdr *nd_opt;
403 	int i = 0;
404 
405 	KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
406 	KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
407 	    __func__));
408 	if (ndopts->nd_opts_search == NULL)
409 		return 0;
410 
411 	while (1) {
412 		nd_opt = nd6_option(ndopts);
413 		if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
414 			/*
415 			 * Message validation requires that all included
416 			 * options have a length that is greater than zero.
417 			 */
418 			ICMP6STAT_INC(icp6s_nd_badopt);
419 			bzero(ndopts, sizeof(*ndopts));
420 			return -1;
421 		}
422 
423 		if (nd_opt == NULL)
424 			goto skip1;
425 
426 		switch (nd_opt->nd_opt_type) {
427 		case ND_OPT_SOURCE_LINKADDR:
428 		case ND_OPT_TARGET_LINKADDR:
429 		case ND_OPT_MTU:
430 		case ND_OPT_REDIRECTED_HEADER:
431 		case ND_OPT_NONCE:
432 			if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
433 				nd6log((LOG_INFO,
434 				    "duplicated ND6 option found (type=%d)\n",
435 				    nd_opt->nd_opt_type));
436 				/* XXX bark? */
437 			} else {
438 				ndopts->nd_opt_array[nd_opt->nd_opt_type]
439 					= nd_opt;
440 			}
441 			break;
442 		case ND_OPT_PREFIX_INFORMATION:
443 			if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
444 				ndopts->nd_opt_array[nd_opt->nd_opt_type]
445 					= nd_opt;
446 			}
447 			ndopts->nd_opts_pi_end =
448 				(struct nd_opt_prefix_info *)nd_opt;
449 			break;
450 		/* What about ND_OPT_ROUTE_INFO? RFC 4191 */
451 		case ND_OPT_RDNSS:	/* RFC 6106 */
452 		case ND_OPT_DNSSL:	/* RFC 6106 */
453 			/*
454 			 * Silently ignore options we know and do not care about
455 			 * in the kernel.
456 			 */
457 			break;
458 		default:
459 			/*
460 			 * Unknown options must be silently ignored,
461 			 * to accomodate future extension to the protocol.
462 			 */
463 			nd6log((LOG_DEBUG,
464 			    "nd6_options: unsupported option %d - "
465 			    "option ignored\n", nd_opt->nd_opt_type));
466 		}
467 
468 skip1:
469 		i++;
470 		if (i > V_nd6_maxndopt) {
471 			ICMP6STAT_INC(icp6s_nd_toomanyopt);
472 			nd6log((LOG_INFO, "too many loop in nd opt\n"));
473 			break;
474 		}
475 
476 		if (ndopts->nd_opts_done)
477 			break;
478 	}
479 
480 	return 0;
481 }
482 
483 /*
484  * ND6 timer routine to handle ND6 entries
485  */
486 void
487 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
488 {
489 	int canceled;
490 
491 	LLE_WLOCK_ASSERT(ln);
492 
493 	if (tick < 0) {
494 		ln->la_expire = 0;
495 		ln->ln_ntick = 0;
496 		canceled = callout_stop(&ln->lle_timer);
497 	} else {
498 		ln->la_expire = time_uptime + tick / hz;
499 		LLE_ADDREF(ln);
500 		if (tick > INT_MAX) {
501 			ln->ln_ntick = tick - INT_MAX;
502 			canceled = callout_reset(&ln->lle_timer, INT_MAX,
503 			    nd6_llinfo_timer, ln);
504 		} else {
505 			ln->ln_ntick = 0;
506 			canceled = callout_reset(&ln->lle_timer, tick,
507 			    nd6_llinfo_timer, ln);
508 		}
509 	}
510 	if (canceled)
511 		LLE_REMREF(ln);
512 }
513 
514 /*
515 * Gets source address of the first packet in hold queue
516 * and stores it in @src.
517 * Returns pointer to @src (if hold queue is not empty) or NULL.
518 *
519 */
520 static struct in6_addr *
521 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src)
522 {
523 	struct ip6_hdr hdr;
524 	struct mbuf *m;
525 
526 	if (ln->la_hold == NULL)
527 		return (NULL);
528 
529 	/*
530 	 * assume every packet in la_hold has the same IP header
531 	 */
532 	m = ln->la_hold;
533 	if (sizeof(hdr) < m->m_len)
534 		return (NULL);
535 
536 	m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr);
537 	*src = hdr.ip6_src;
538 
539 	return (src);
540 }
541 
542 /*
543  * Switch @lle state to new state optionally arming timers.
544  */
545 void
546 nd6_llinfo_setstate(struct llentry *lle, int newstate)
547 {
548 	struct ifnet *ifp;
549 	long delay;
550 
551 	delay = 0;
552 
553 	switch (newstate) {
554 	case ND6_LLINFO_INCOMPLETE:
555 		ifp = lle->lle_tbl->llt_ifp;
556 		delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000;
557 		break;
558 	case ND6_LLINFO_REACHABLE:
559 		if (!ND6_LLINFO_PERMANENT(lle)) {
560 			ifp = lle->lle_tbl->llt_ifp;
561 			delay = (long)ND_IFINFO(ifp)->reachable * hz;
562 		}
563 		break;
564 	case ND6_LLINFO_STALE:
565 		delay = (long)V_nd6_gctimer * hz;
566 		break;
567 	case ND6_LLINFO_DELAY:
568 		lle->la_asked = 0;
569 		delay = (long)V_nd6_delay * hz;
570 		break;
571 	}
572 
573 	if (delay > 0)
574 		nd6_llinfo_settimer_locked(lle, delay);
575 
576 	lle->ln_state = newstate;
577 }
578 
579 
580 void
581 nd6_llinfo_settimer(struct llentry *ln, long tick)
582 {
583 
584 	LLE_WLOCK(ln);
585 	nd6_llinfo_settimer_locked(ln, tick);
586 	LLE_WUNLOCK(ln);
587 }
588 
589 static void
590 nd6_llinfo_timer(void *arg)
591 {
592 	struct llentry *ln;
593 	struct in6_addr *dst, *pdst, *psrc, src;
594 	struct ifnet *ifp;
595 	struct nd_ifinfo *ndi = NULL;
596 	int send_ns;
597 
598 	KASSERT(arg != NULL, ("%s: arg NULL", __func__));
599 	ln = (struct llentry *)arg;
600 	LLE_WLOCK(ln);
601 	if (callout_pending(&ln->lle_timer)) {
602 		/*
603 		 * Here we are a bit odd here in the treatment of
604 		 * active/pending. If the pending bit is set, it got
605 		 * rescheduled before I ran. The active
606 		 * bit we ignore, since if it was stopped
607 		 * in ll_tablefree() and was currently running
608 		 * it would have return 0 so the code would
609 		 * not have deleted it since the callout could
610 		 * not be stopped so we want to go through
611 		 * with the delete here now. If the callout
612 		 * was restarted, the pending bit will be back on and
613 		 * we just want to bail since the callout_reset would
614 		 * return 1 and our reference would have been removed
615 		 * by nd6_llinfo_settimer_locked above since canceled
616 		 * would have been 1.
617 		 */
618 		LLE_WUNLOCK(ln);
619 		return;
620 	}
621 	ifp = ln->lle_tbl->llt_ifp;
622 	CURVNET_SET(ifp->if_vnet);
623 	ndi = ND_IFINFO(ifp);
624 	send_ns = 0;
625 	dst = &ln->r_l3addr.addr6;
626 	pdst = dst;
627 
628 	if (ln->ln_ntick > 0) {
629 		if (ln->ln_ntick > INT_MAX) {
630 			ln->ln_ntick -= INT_MAX;
631 			nd6_llinfo_settimer_locked(ln, INT_MAX);
632 		} else {
633 			ln->ln_ntick = 0;
634 			nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
635 		}
636 		goto done;
637 	}
638 
639 	if (ln->la_flags & LLE_STATIC) {
640 		goto done;
641 	}
642 
643 	if (ln->la_flags & LLE_DELETED) {
644 		nd6_free(ln, 0);
645 		ln = NULL;
646 		goto done;
647 	}
648 
649 	switch (ln->ln_state) {
650 	case ND6_LLINFO_INCOMPLETE:
651 		if (ln->la_asked < V_nd6_mmaxtries) {
652 			ln->la_asked++;
653 			send_ns = 1;
654 			/* Send NS to multicast address */
655 			pdst = NULL;
656 		} else {
657 			struct mbuf *m = ln->la_hold;
658 			if (m) {
659 				struct mbuf *m0;
660 
661 				/*
662 				 * assuming every packet in la_hold has the
663 				 * same IP header.  Send error after unlock.
664 				 */
665 				m0 = m->m_nextpkt;
666 				m->m_nextpkt = NULL;
667 				ln->la_hold = m0;
668 				clear_llinfo_pqueue(ln);
669 			}
670 			EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_TIMEDOUT);
671 			nd6_free(ln, 0);
672 			ln = NULL;
673 			if (m != NULL)
674 				icmp6_error2(m, ICMP6_DST_UNREACH,
675 				    ICMP6_DST_UNREACH_ADDR, 0, ifp);
676 		}
677 		break;
678 	case ND6_LLINFO_REACHABLE:
679 		if (!ND6_LLINFO_PERMANENT(ln))
680 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
681 		break;
682 
683 	case ND6_LLINFO_STALE:
684 		/* Garbage Collection(RFC 2461 5.3) */
685 		if (!ND6_LLINFO_PERMANENT(ln)) {
686 			EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
687 			nd6_free(ln, 1);
688 			ln = NULL;
689 		}
690 		break;
691 
692 	case ND6_LLINFO_DELAY:
693 		if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
694 			/* We need NUD */
695 			ln->la_asked = 1;
696 			nd6_llinfo_setstate(ln, ND6_LLINFO_PROBE);
697 			send_ns = 1;
698 		} else
699 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); /* XXX */
700 		break;
701 	case ND6_LLINFO_PROBE:
702 		if (ln->la_asked < V_nd6_umaxtries) {
703 			ln->la_asked++;
704 			send_ns = 1;
705 		} else {
706 			EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
707 			nd6_free(ln, 0);
708 			ln = NULL;
709 		}
710 		break;
711 	default:
712 		panic("%s: paths in a dark night can be confusing: %d",
713 		    __func__, ln->ln_state);
714 	}
715 done:
716 	if (send_ns != 0) {
717 		nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
718 		psrc = nd6_llinfo_get_holdsrc(ln, &src);
719 		LLE_FREE_LOCKED(ln);
720 		ln = NULL;
721 		nd6_ns_output(ifp, psrc, pdst, dst, NULL);
722 	}
723 
724 	if (ln != NULL)
725 		LLE_FREE_LOCKED(ln);
726 	CURVNET_RESTORE();
727 }
728 
729 
730 /*
731  * ND6 timer routine to expire default route list and prefix list
732  */
733 void
734 nd6_timer(void *arg)
735 {
736 	CURVNET_SET((struct vnet *) arg);
737 	struct nd_defrouter *dr, *ndr;
738 	struct nd_prefix *pr, *npr;
739 	struct in6_ifaddr *ia6, *nia6;
740 
741 	callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
742 	    nd6_timer, curvnet);
743 
744 	/* expire default router list */
745 	TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
746 		if (dr->expire && dr->expire < time_uptime)
747 			defrtrlist_del(dr);
748 	}
749 
750 	/*
751 	 * expire interface addresses.
752 	 * in the past the loop was inside prefix expiry processing.
753 	 * However, from a stricter speci-confrmance standpoint, we should
754 	 * rather separate address lifetimes and prefix lifetimes.
755 	 *
756 	 * XXXRW: in6_ifaddrhead locking.
757 	 */
758   addrloop:
759 	TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
760 		/* check address lifetime */
761 		if (IFA6_IS_INVALID(ia6)) {
762 			int regen = 0;
763 
764 			/*
765 			 * If the expiring address is temporary, try
766 			 * regenerating a new one.  This would be useful when
767 			 * we suspended a laptop PC, then turned it on after a
768 			 * period that could invalidate all temporary
769 			 * addresses.  Although we may have to restart the
770 			 * loop (see below), it must be after purging the
771 			 * address.  Otherwise, we'd see an infinite loop of
772 			 * regeneration.
773 			 */
774 			if (V_ip6_use_tempaddr &&
775 			    (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
776 				if (regen_tmpaddr(ia6) == 0)
777 					regen = 1;
778 			}
779 
780 			in6_purgeaddr(&ia6->ia_ifa);
781 
782 			if (regen)
783 				goto addrloop; /* XXX: see below */
784 		} else if (IFA6_IS_DEPRECATED(ia6)) {
785 			int oldflags = ia6->ia6_flags;
786 
787 			ia6->ia6_flags |= IN6_IFF_DEPRECATED;
788 
789 			/*
790 			 * If a temporary address has just become deprecated,
791 			 * regenerate a new one if possible.
792 			 */
793 			if (V_ip6_use_tempaddr &&
794 			    (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
795 			    (oldflags & IN6_IFF_DEPRECATED) == 0) {
796 
797 				if (regen_tmpaddr(ia6) == 0) {
798 					/*
799 					 * A new temporary address is
800 					 * generated.
801 					 * XXX: this means the address chain
802 					 * has changed while we are still in
803 					 * the loop.  Although the change
804 					 * would not cause disaster (because
805 					 * it's not a deletion, but an
806 					 * addition,) we'd rather restart the
807 					 * loop just for safety.  Or does this
808 					 * significantly reduce performance??
809 					 */
810 					goto addrloop;
811 				}
812 			}
813 		} else {
814 			/*
815 			 * A new RA might have made a deprecated address
816 			 * preferred.
817 			 */
818 			ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
819 		}
820 	}
821 
822 	/* expire prefix list */
823 	LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
824 		/*
825 		 * check prefix lifetime.
826 		 * since pltime is just for autoconf, pltime processing for
827 		 * prefix is not necessary.
828 		 */
829 		if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
830 		    time_uptime - pr->ndpr_lastupdate > pr->ndpr_vltime) {
831 
832 			/*
833 			 * address expiration and prefix expiration are
834 			 * separate.  NEVER perform in6_purgeaddr here.
835 			 */
836 			prelist_remove(pr);
837 		}
838 	}
839 	CURVNET_RESTORE();
840 }
841 
842 /*
843  * ia6 - deprecated/invalidated temporary address
844  */
845 static int
846 regen_tmpaddr(struct in6_ifaddr *ia6)
847 {
848 	struct ifaddr *ifa;
849 	struct ifnet *ifp;
850 	struct in6_ifaddr *public_ifa6 = NULL;
851 
852 	ifp = ia6->ia_ifa.ifa_ifp;
853 	IF_ADDR_RLOCK(ifp);
854 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
855 		struct in6_ifaddr *it6;
856 
857 		if (ifa->ifa_addr->sa_family != AF_INET6)
858 			continue;
859 
860 		it6 = (struct in6_ifaddr *)ifa;
861 
862 		/* ignore no autoconf addresses. */
863 		if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
864 			continue;
865 
866 		/* ignore autoconf addresses with different prefixes. */
867 		if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
868 			continue;
869 
870 		/*
871 		 * Now we are looking at an autoconf address with the same
872 		 * prefix as ours.  If the address is temporary and is still
873 		 * preferred, do not create another one.  It would be rare, but
874 		 * could happen, for example, when we resume a laptop PC after
875 		 * a long period.
876 		 */
877 		if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
878 		    !IFA6_IS_DEPRECATED(it6)) {
879 			public_ifa6 = NULL;
880 			break;
881 		}
882 
883 		/*
884 		 * This is a public autoconf address that has the same prefix
885 		 * as ours.  If it is preferred, keep it.  We can't break the
886 		 * loop here, because there may be a still-preferred temporary
887 		 * address with the prefix.
888 		 */
889 		if (!IFA6_IS_DEPRECATED(it6))
890 			public_ifa6 = it6;
891 	}
892 	if (public_ifa6 != NULL)
893 		ifa_ref(&public_ifa6->ia_ifa);
894 	IF_ADDR_RUNLOCK(ifp);
895 
896 	if (public_ifa6 != NULL) {
897 		int e;
898 
899 		if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
900 			ifa_free(&public_ifa6->ia_ifa);
901 			log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
902 			    " tmp addr,errno=%d\n", e);
903 			return (-1);
904 		}
905 		ifa_free(&public_ifa6->ia_ifa);
906 		return (0);
907 	}
908 
909 	return (-1);
910 }
911 
912 /*
913  * Nuke neighbor cache/prefix/default router management table, right before
914  * ifp goes away.
915  */
916 void
917 nd6_purge(struct ifnet *ifp)
918 {
919 	struct nd_defrouter *dr, *ndr;
920 	struct nd_prefix *pr, *npr;
921 
922 	/*
923 	 * Nuke default router list entries toward ifp.
924 	 * We defer removal of default router list entries that is installed
925 	 * in the routing table, in order to keep additional side effects as
926 	 * small as possible.
927 	 */
928 	TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
929 		if (dr->installed)
930 			continue;
931 
932 		if (dr->ifp == ifp)
933 			defrtrlist_del(dr);
934 	}
935 
936 	TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
937 		if (!dr->installed)
938 			continue;
939 
940 		if (dr->ifp == ifp)
941 			defrtrlist_del(dr);
942 	}
943 
944 	/* Nuke prefix list entries toward ifp */
945 	LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
946 		if (pr->ndpr_ifp == ifp) {
947 			/*
948 			 * Because if_detach() does *not* release prefixes
949 			 * while purging addresses the reference count will
950 			 * still be above zero. We therefore reset it to
951 			 * make sure that the prefix really gets purged.
952 			 */
953 			pr->ndpr_refcnt = 0;
954 
955 			/*
956 			 * Previously, pr->ndpr_addr is removed as well,
957 			 * but I strongly believe we don't have to do it.
958 			 * nd6_purge() is only called from in6_ifdetach(),
959 			 * which removes all the associated interface addresses
960 			 * by itself.
961 			 * (jinmei@kame.net 20010129)
962 			 */
963 			prelist_remove(pr);
964 		}
965 	}
966 
967 	/* cancel default outgoing interface setting */
968 	if (V_nd6_defifindex == ifp->if_index)
969 		nd6_setdefaultiface(0);
970 
971 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
972 		/* Refresh default router list. */
973 		defrouter_select();
974 	}
975 
976 	/* XXXXX
977 	 * We do not nuke the neighbor cache entries here any more
978 	 * because the neighbor cache is kept in if_afdata[AF_INET6].
979 	 * nd6_purge() is invoked by in6_ifdetach() which is called
980 	 * from if_detach() where everything gets purged. So let
981 	 * in6_domifdetach() do the actual L2 table purging work.
982 	 */
983 }
984 
985 /*
986  * the caller acquires and releases the lock on the lltbls
987  * Returns the llentry locked
988  */
989 struct llentry *
990 nd6_lookup(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
991 {
992 	struct sockaddr_in6 sin6;
993 	struct llentry *ln;
994 	int llflags;
995 
996 	bzero(&sin6, sizeof(sin6));
997 	sin6.sin6_len = sizeof(struct sockaddr_in6);
998 	sin6.sin6_family = AF_INET6;
999 	sin6.sin6_addr = *addr6;
1000 
1001 	IF_AFDATA_LOCK_ASSERT(ifp);
1002 
1003 	llflags = (flags & ND6_EXCLUSIVE) ? LLE_EXCLUSIVE : 0;
1004 	ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6);
1005 
1006 	return (ln);
1007 }
1008 
1009 struct llentry *
1010 nd6_alloc(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
1011 {
1012 	struct sockaddr_in6 sin6;
1013 	struct llentry *ln;
1014 
1015 	bzero(&sin6, sizeof(sin6));
1016 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1017 	sin6.sin6_family = AF_INET6;
1018 	sin6.sin6_addr = *addr6;
1019 
1020 	ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6);
1021 	if (ln != NULL)
1022 		ln->ln_state = ND6_LLINFO_NOSTATE;
1023 
1024 	return (ln);
1025 }
1026 
1027 /*
1028  * Test whether a given IPv6 address is a neighbor or not, ignoring
1029  * the actual neighbor cache.  The neighbor cache is ignored in order
1030  * to not reenter the routing code from within itself.
1031  */
1032 static int
1033 nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
1034 {
1035 	struct nd_prefix *pr;
1036 	struct ifaddr *dstaddr;
1037 
1038 	/*
1039 	 * A link-local address is always a neighbor.
1040 	 * XXX: a link does not necessarily specify a single interface.
1041 	 */
1042 	if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
1043 		struct sockaddr_in6 sin6_copy;
1044 		u_int32_t zone;
1045 
1046 		/*
1047 		 * We need sin6_copy since sa6_recoverscope() may modify the
1048 		 * content (XXX).
1049 		 */
1050 		sin6_copy = *addr;
1051 		if (sa6_recoverscope(&sin6_copy))
1052 			return (0); /* XXX: should be impossible */
1053 		if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
1054 			return (0);
1055 		if (sin6_copy.sin6_scope_id == zone)
1056 			return (1);
1057 		else
1058 			return (0);
1059 	}
1060 
1061 	/*
1062 	 * If the address matches one of our addresses,
1063 	 * it should be a neighbor.
1064 	 * If the address matches one of our on-link prefixes, it should be a
1065 	 * neighbor.
1066 	 */
1067 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1068 		if (pr->ndpr_ifp != ifp)
1069 			continue;
1070 
1071 		if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
1072 			struct rtentry *rt;
1073 
1074 			/* Always use the default FIB here. */
1075 			rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix,
1076 			    0, 0, RT_DEFAULT_FIB);
1077 			if (rt == NULL)
1078 				continue;
1079 			/*
1080 			 * This is the case where multiple interfaces
1081 			 * have the same prefix, but only one is installed
1082 			 * into the routing table and that prefix entry
1083 			 * is not the one being examined here. In the case
1084 			 * where RADIX_MPATH is enabled, multiple route
1085 			 * entries (of the same rt_key value) will be
1086 			 * installed because the interface addresses all
1087 			 * differ.
1088 			 */
1089 			if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
1090 			       &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) {
1091 				RTFREE_LOCKED(rt);
1092 				continue;
1093 			}
1094 			RTFREE_LOCKED(rt);
1095 		}
1096 
1097 		if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
1098 		    &addr->sin6_addr, &pr->ndpr_mask))
1099 			return (1);
1100 	}
1101 
1102 	/*
1103 	 * If the address is assigned on the node of the other side of
1104 	 * a p2p interface, the address should be a neighbor.
1105 	 */
1106 	dstaddr = ifa_ifwithdstaddr((const struct sockaddr *)addr, RT_ALL_FIBS);
1107 	if (dstaddr != NULL) {
1108 		if (dstaddr->ifa_ifp == ifp) {
1109 			ifa_free(dstaddr);
1110 			return (1);
1111 		}
1112 		ifa_free(dstaddr);
1113 	}
1114 
1115 	/*
1116 	 * If the default router list is empty, all addresses are regarded
1117 	 * as on-link, and thus, as a neighbor.
1118 	 */
1119 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV &&
1120 	    TAILQ_EMPTY(&V_nd_defrouter) &&
1121 	    V_nd6_defifindex == ifp->if_index) {
1122 		return (1);
1123 	}
1124 
1125 	return (0);
1126 }
1127 
1128 
1129 /*
1130  * Detect if a given IPv6 address identifies a neighbor on a given link.
1131  * XXX: should take care of the destination of a p2p link?
1132  */
1133 int
1134 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
1135 {
1136 	struct llentry *lle;
1137 	int rc = 0;
1138 
1139 	IF_AFDATA_UNLOCK_ASSERT(ifp);
1140 	if (nd6_is_new_addr_neighbor(addr, ifp))
1141 		return (1);
1142 
1143 	/*
1144 	 * Even if the address matches none of our addresses, it might be
1145 	 * in the neighbor cache.
1146 	 */
1147 	IF_AFDATA_RLOCK(ifp);
1148 	if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
1149 		LLE_RUNLOCK(lle);
1150 		rc = 1;
1151 	}
1152 	IF_AFDATA_RUNLOCK(ifp);
1153 	return (rc);
1154 }
1155 
1156 /*
1157  * Free an nd6 llinfo entry.
1158  * Since the function would cause significant changes in the kernel, DO NOT
1159  * make it global, unless you have a strong reason for the change, and are sure
1160  * that the change is safe.
1161  */
1162 static void
1163 nd6_free(struct llentry *ln, int gc)
1164 {
1165 	struct nd_defrouter *dr;
1166 	struct ifnet *ifp;
1167 
1168 	LLE_WLOCK_ASSERT(ln);
1169 
1170 	/*
1171 	 * we used to have pfctlinput(PRC_HOSTDEAD) here.
1172 	 * even though it is not harmful, it was not really necessary.
1173 	 */
1174 
1175 	/* cancel timer */
1176 	nd6_llinfo_settimer_locked(ln, -1);
1177 
1178 	ifp = ln->lle_tbl->llt_ifp;
1179 
1180 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1181 		dr = defrouter_lookup(&ln->r_l3addr.addr6, ifp);
1182 
1183 		if (dr != NULL && dr->expire &&
1184 		    ln->ln_state == ND6_LLINFO_STALE && gc) {
1185 			/*
1186 			 * If the reason for the deletion is just garbage
1187 			 * collection, and the neighbor is an active default
1188 			 * router, do not delete it.  Instead, reset the GC
1189 			 * timer using the router's lifetime.
1190 			 * Simply deleting the entry would affect default
1191 			 * router selection, which is not necessarily a good
1192 			 * thing, especially when we're using router preference
1193 			 * values.
1194 			 * XXX: the check for ln_state would be redundant,
1195 			 *      but we intentionally keep it just in case.
1196 			 */
1197 			if (dr->expire > time_uptime)
1198 				nd6_llinfo_settimer_locked(ln,
1199 				    (dr->expire - time_uptime) * hz);
1200 			else
1201 				nd6_llinfo_settimer_locked(ln,
1202 				    (long)V_nd6_gctimer * hz);
1203 
1204 			LLE_REMREF(ln);
1205 			LLE_WUNLOCK(ln);
1206 			return;
1207 		}
1208 
1209 		if (dr) {
1210 			/*
1211 			 * Unreachablity of a router might affect the default
1212 			 * router selection and on-link detection of advertised
1213 			 * prefixes.
1214 			 */
1215 
1216 			/*
1217 			 * Temporarily fake the state to choose a new default
1218 			 * router and to perform on-link determination of
1219 			 * prefixes correctly.
1220 			 * Below the state will be set correctly,
1221 			 * or the entry itself will be deleted.
1222 			 */
1223 			ln->ln_state = ND6_LLINFO_INCOMPLETE;
1224 		}
1225 
1226 		if (ln->ln_router || dr) {
1227 
1228 			/*
1229 			 * We need to unlock to avoid a LOR with rt6_flush() with the
1230 			 * rnh and for the calls to pfxlist_onlink_check() and
1231 			 * defrouter_select() in the block further down for calls
1232 			 * into nd6_lookup().  We still hold a ref.
1233 			 */
1234 			LLE_WUNLOCK(ln);
1235 
1236 			/*
1237 			 * rt6_flush must be called whether or not the neighbor
1238 			 * is in the Default Router List.
1239 			 * See a corresponding comment in nd6_na_input().
1240 			 */
1241 			rt6_flush(&ln->r_l3addr.addr6, ifp);
1242 		}
1243 
1244 		if (dr) {
1245 			/*
1246 			 * Since defrouter_select() does not affect the
1247 			 * on-link determination and MIP6 needs the check
1248 			 * before the default router selection, we perform
1249 			 * the check now.
1250 			 */
1251 			pfxlist_onlink_check();
1252 
1253 			/*
1254 			 * Refresh default router list.
1255 			 */
1256 			defrouter_select();
1257 		}
1258 
1259 		/*
1260 		 * If this entry was added by an on-link redirect, remove the
1261 		 * corresponding host route.
1262 		 */
1263 		if (ln->la_flags & LLE_REDIRECT)
1264 			nd6_free_redirect(ln);
1265 
1266 		if (ln->ln_router || dr)
1267 			LLE_WLOCK(ln);
1268 	}
1269 
1270 	/*
1271 	 * Save to unlock. We still hold an extra reference and will not
1272 	 * free(9) in llentry_free() if someone else holds one as well.
1273 	 */
1274 	LLE_WUNLOCK(ln);
1275 	IF_AFDATA_LOCK(ifp);
1276 	LLE_WLOCK(ln);
1277 	/* Guard against race with other llentry_free(). */
1278 	if (ln->la_flags & LLE_LINKED) {
1279 		/* Remove callout reference */
1280 		LLE_REMREF(ln);
1281 		lltable_unlink_entry(ln->lle_tbl, ln);
1282 	}
1283 	IF_AFDATA_UNLOCK(ifp);
1284 
1285 	llentry_free(ln);
1286 }
1287 
1288 /*
1289  * Remove the rtentry for the given llentry,
1290  * both of which were installed by a redirect.
1291  */
1292 static void
1293 nd6_free_redirect(const struct llentry *ln)
1294 {
1295 	int fibnum;
1296 	struct rtentry *rt;
1297 	struct radix_node_head *rnh;
1298 	struct sockaddr_in6 sin6;
1299 
1300 	lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6);
1301 	for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1302 		rnh = rt_tables_get_rnh(fibnum, AF_INET6);
1303 		if (rnh == NULL)
1304 			continue;
1305 
1306 		RADIX_NODE_HEAD_LOCK(rnh);
1307 		rt = in6_rtalloc1((struct sockaddr *)&sin6, 0,
1308 		    RTF_RNH_LOCKED, fibnum);
1309 		if (rt) {
1310 			if (rt->rt_flags == (RTF_UP | RTF_HOST | RTF_DYNAMIC))
1311 				rt_expunge(rnh, rt);
1312 			RTFREE_LOCKED(rt);
1313 		}
1314 		RADIX_NODE_HEAD_UNLOCK(rnh);
1315 	}
1316 }
1317 
1318 /*
1319  * Upper-layer reachability hint for Neighbor Unreachability Detection.
1320  *
1321  * XXX cost-effective methods?
1322  */
1323 void
1324 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
1325 {
1326 	struct llentry *ln;
1327 	struct ifnet *ifp;
1328 
1329 	if ((dst6 == NULL) || (rt == NULL))
1330 		return;
1331 
1332 	ifp = rt->rt_ifp;
1333 	IF_AFDATA_RLOCK(ifp);
1334 	ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL);
1335 	IF_AFDATA_RUNLOCK(ifp);
1336 	if (ln == NULL)
1337 		return;
1338 
1339 	if (ln->ln_state < ND6_LLINFO_REACHABLE)
1340 		goto done;
1341 
1342 	/*
1343 	 * if we get upper-layer reachability confirmation many times,
1344 	 * it is possible we have false information.
1345 	 */
1346 	if (!force) {
1347 		ln->ln_byhint++;
1348 		if (ln->ln_byhint > V_nd6_maxnudhint) {
1349 			goto done;
1350 		}
1351 	}
1352 
1353  	nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
1354 done:
1355 	LLE_WUNLOCK(ln);
1356 }
1357 
1358 
1359 /*
1360  * Rejuvenate this function for routing operations related
1361  * processing.
1362  */
1363 void
1364 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
1365 {
1366 	struct sockaddr_in6 *gateway;
1367 	struct nd_defrouter *dr;
1368 	struct ifnet *ifp;
1369 
1370 	gateway = (struct sockaddr_in6 *)rt->rt_gateway;
1371 	ifp = rt->rt_ifp;
1372 
1373 	switch (req) {
1374 	case RTM_ADD:
1375 		break;
1376 
1377 	case RTM_DELETE:
1378 		if (!ifp)
1379 			return;
1380 		/*
1381 		 * Only indirect routes are interesting.
1382 		 */
1383 		if ((rt->rt_flags & RTF_GATEWAY) == 0)
1384 			return;
1385 		/*
1386 		 * check for default route
1387 		 */
1388 		if (IN6_ARE_ADDR_EQUAL(&in6addr_any,
1389 				       &SIN6(rt_key(rt))->sin6_addr)) {
1390 
1391 			dr = defrouter_lookup(&gateway->sin6_addr, ifp);
1392 			if (dr != NULL)
1393 				dr->installed = 0;
1394 		}
1395 		break;
1396 	}
1397 }
1398 
1399 
1400 int
1401 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
1402 {
1403 	struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1404 	struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1405 	struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1406 	int error = 0;
1407 
1408 	if (ifp->if_afdata[AF_INET6] == NULL)
1409 		return (EPFNOSUPPORT);
1410 	switch (cmd) {
1411 	case OSIOCGIFINFO_IN6:
1412 #define ND	ndi->ndi
1413 		/* XXX: old ndp(8) assumes a positive value for linkmtu. */
1414 		bzero(&ND, sizeof(ND));
1415 		ND.linkmtu = IN6_LINKMTU(ifp);
1416 		ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
1417 		ND.basereachable = ND_IFINFO(ifp)->basereachable;
1418 		ND.reachable = ND_IFINFO(ifp)->reachable;
1419 		ND.retrans = ND_IFINFO(ifp)->retrans;
1420 		ND.flags = ND_IFINFO(ifp)->flags;
1421 		ND.recalctm = ND_IFINFO(ifp)->recalctm;
1422 		ND.chlim = ND_IFINFO(ifp)->chlim;
1423 		break;
1424 	case SIOCGIFINFO_IN6:
1425 		ND = *ND_IFINFO(ifp);
1426 		break;
1427 	case SIOCSIFINFO_IN6:
1428 		/*
1429 		 * used to change host variables from userland.
1430 		 * intented for a use on router to reflect RA configurations.
1431 		 */
1432 		/* 0 means 'unspecified' */
1433 		if (ND.linkmtu != 0) {
1434 			if (ND.linkmtu < IPV6_MMTU ||
1435 			    ND.linkmtu > IN6_LINKMTU(ifp)) {
1436 				error = EINVAL;
1437 				break;
1438 			}
1439 			ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
1440 		}
1441 
1442 		if (ND.basereachable != 0) {
1443 			int obasereachable = ND_IFINFO(ifp)->basereachable;
1444 
1445 			ND_IFINFO(ifp)->basereachable = ND.basereachable;
1446 			if (ND.basereachable != obasereachable)
1447 				ND_IFINFO(ifp)->reachable =
1448 				    ND_COMPUTE_RTIME(ND.basereachable);
1449 		}
1450 		if (ND.retrans != 0)
1451 			ND_IFINFO(ifp)->retrans = ND.retrans;
1452 		if (ND.chlim != 0)
1453 			ND_IFINFO(ifp)->chlim = ND.chlim;
1454 		/* FALLTHROUGH */
1455 	case SIOCSIFINFO_FLAGS:
1456 	{
1457 		struct ifaddr *ifa;
1458 		struct in6_ifaddr *ia;
1459 
1460 		if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1461 		    !(ND.flags & ND6_IFF_IFDISABLED)) {
1462 			/* ifdisabled 1->0 transision */
1463 
1464 			/*
1465 			 * If the interface is marked as ND6_IFF_IFDISABLED and
1466 			 * has an link-local address with IN6_IFF_DUPLICATED,
1467 			 * do not clear ND6_IFF_IFDISABLED.
1468 			 * See RFC 4862, Section 5.4.5.
1469 			 */
1470 			IF_ADDR_RLOCK(ifp);
1471 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1472 				if (ifa->ifa_addr->sa_family != AF_INET6)
1473 					continue;
1474 				ia = (struct in6_ifaddr *)ifa;
1475 				if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
1476 				    IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1477 					break;
1478 			}
1479 			IF_ADDR_RUNLOCK(ifp);
1480 
1481 			if (ifa != NULL) {
1482 				/* LLA is duplicated. */
1483 				ND.flags |= ND6_IFF_IFDISABLED;
1484 				log(LOG_ERR, "Cannot enable an interface"
1485 				    " with a link-local address marked"
1486 				    " duplicate.\n");
1487 			} else {
1488 				ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
1489 				if (ifp->if_flags & IFF_UP)
1490 					in6_if_up(ifp);
1491 			}
1492 		} else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
1493 			    (ND.flags & ND6_IFF_IFDISABLED)) {
1494 			/* ifdisabled 0->1 transision */
1495 			/* Mark all IPv6 address as tentative. */
1496 
1497 			ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1498 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) {
1499 				IF_ADDR_RLOCK(ifp);
1500 				TAILQ_FOREACH(ifa, &ifp->if_addrhead,
1501 				    ifa_link) {
1502 					if (ifa->ifa_addr->sa_family !=
1503 					    AF_INET6)
1504 						continue;
1505 					ia = (struct in6_ifaddr *)ifa;
1506 					ia->ia6_flags |= IN6_IFF_TENTATIVE;
1507 				}
1508 				IF_ADDR_RUNLOCK(ifp);
1509 			}
1510 		}
1511 
1512 		if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
1513 			if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
1514 				/* auto_linklocal 0->1 transision */
1515 
1516 				/* If no link-local address on ifp, configure */
1517 				ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
1518 				in6_ifattach(ifp, NULL);
1519 			} else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
1520 			    ifp->if_flags & IFF_UP) {
1521 				/*
1522 				 * When the IF already has
1523 				 * ND6_IFF_AUTO_LINKLOCAL, no link-local
1524 				 * address is assigned, and IFF_UP, try to
1525 				 * assign one.
1526 				 */
1527 				IF_ADDR_RLOCK(ifp);
1528 				TAILQ_FOREACH(ifa, &ifp->if_addrhead,
1529 				    ifa_link) {
1530 					if (ifa->ifa_addr->sa_family !=
1531 					    AF_INET6)
1532 						continue;
1533 					ia = (struct in6_ifaddr *)ifa;
1534 					if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
1535 						break;
1536 				}
1537 				IF_ADDR_RUNLOCK(ifp);
1538 				if (ifa != NULL)
1539 					/* No LLA is configured. */
1540 					in6_ifattach(ifp, NULL);
1541 			}
1542 		}
1543 	}
1544 		ND_IFINFO(ifp)->flags = ND.flags;
1545 		break;
1546 #undef ND
1547 	case SIOCSNDFLUSH_IN6:	/* XXX: the ioctl name is confusing... */
1548 		/* sync kernel routing table with the default router list */
1549 		defrouter_reset();
1550 		defrouter_select();
1551 		break;
1552 	case SIOCSPFXFLUSH_IN6:
1553 	{
1554 		/* flush all the prefix advertised by routers */
1555 		struct nd_prefix *pr, *next;
1556 
1557 		LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
1558 			struct in6_ifaddr *ia, *ia_next;
1559 
1560 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1561 				continue; /* XXX */
1562 
1563 			/* do we really have to remove addresses as well? */
1564 			/* XXXRW: in6_ifaddrhead locking. */
1565 			TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
1566 			    ia_next) {
1567 				if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1568 					continue;
1569 
1570 				if (ia->ia6_ndpr == pr)
1571 					in6_purgeaddr(&ia->ia_ifa);
1572 			}
1573 			prelist_remove(pr);
1574 		}
1575 		break;
1576 	}
1577 	case SIOCSRTRFLUSH_IN6:
1578 	{
1579 		/* flush all the default routers */
1580 		struct nd_defrouter *dr, *next;
1581 
1582 		defrouter_reset();
1583 		TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) {
1584 			defrtrlist_del(dr);
1585 		}
1586 		defrouter_select();
1587 		break;
1588 	}
1589 	case SIOCGNBRINFO_IN6:
1590 	{
1591 		struct llentry *ln;
1592 		struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1593 
1594 		if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
1595 			return (error);
1596 
1597 		IF_AFDATA_RLOCK(ifp);
1598 		ln = nd6_lookup(&nb_addr, 0, ifp);
1599 		IF_AFDATA_RUNLOCK(ifp);
1600 
1601 		if (ln == NULL) {
1602 			error = EINVAL;
1603 			break;
1604 		}
1605 		nbi->state = ln->ln_state;
1606 		nbi->asked = ln->la_asked;
1607 		nbi->isrouter = ln->ln_router;
1608 		if (ln->la_expire == 0)
1609 			nbi->expire = 0;
1610 		else
1611 			nbi->expire = ln->la_expire +
1612 			    (time_second - time_uptime);
1613 		LLE_RUNLOCK(ln);
1614 		break;
1615 	}
1616 	case SIOCGDEFIFACE_IN6:	/* XXX: should be implemented as a sysctl? */
1617 		ndif->ifindex = V_nd6_defifindex;
1618 		break;
1619 	case SIOCSDEFIFACE_IN6:	/* XXX: should be implemented as a sysctl? */
1620 		return (nd6_setdefaultiface(ndif->ifindex));
1621 	}
1622 	return (error);
1623 }
1624 
1625 /*
1626  * Calculates new isRouter value based on provided parameters and
1627  * returns it.
1628  */
1629 static int
1630 nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr,
1631     int ln_router)
1632 {
1633 
1634 	/*
1635 	 * ICMP6 type dependent behavior.
1636 	 *
1637 	 * NS: clear IsRouter if new entry
1638 	 * RS: clear IsRouter
1639 	 * RA: set IsRouter if there's lladdr
1640 	 * redir: clear IsRouter if new entry
1641 	 *
1642 	 * RA case, (1):
1643 	 * The spec says that we must set IsRouter in the following cases:
1644 	 * - If lladdr exist, set IsRouter.  This means (1-5).
1645 	 * - If it is old entry (!newentry), set IsRouter.  This means (7).
1646 	 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1647 	 * A quetion arises for (1) case.  (1) case has no lladdr in the
1648 	 * neighbor cache, this is similar to (6).
1649 	 * This case is rare but we figured that we MUST NOT set IsRouter.
1650 	 *
1651 	 *   is_new  old_addr new_addr 	    NS  RS  RA	redir
1652 	 *							D R
1653 	 *	0	n	n	(1)	c   ?     s
1654 	 *	0	y	n	(2)	c   s     s
1655 	 *	0	n	y	(3)	c   s     s
1656 	 *	0	y	y	(4)	c   s     s
1657 	 *	0	y	y	(5)	c   s     s
1658 	 *	1	--	n	(6) c	c	c s
1659 	 *	1	--	y	(7) c	c   s	c s
1660 	 *
1661 	 *					(c=clear s=set)
1662 	 */
1663 	switch (type & 0xff) {
1664 	case ND_NEIGHBOR_SOLICIT:
1665 		/*
1666 		 * New entry must have is_router flag cleared.
1667 		 */
1668 		if (is_new)					/* (6-7) */
1669 			ln_router = 0;
1670 		break;
1671 	case ND_REDIRECT:
1672 		/*
1673 		 * If the icmp is a redirect to a better router, always set the
1674 		 * is_router flag.  Otherwise, if the entry is newly created,
1675 		 * clear the flag.  [RFC 2461, sec 8.3]
1676 		 */
1677 		if (code == ND_REDIRECT_ROUTER)
1678 			ln_router = 1;
1679 		else {
1680 			if (is_new)				/* (6-7) */
1681 				ln_router = 0;
1682 		}
1683 		break;
1684 	case ND_ROUTER_SOLICIT:
1685 		/*
1686 		 * is_router flag must always be cleared.
1687 		 */
1688 		ln_router = 0;
1689 		break;
1690 	case ND_ROUTER_ADVERT:
1691 		/*
1692 		 * Mark an entry with lladdr as a router.
1693 		 */
1694 		if ((!is_new && (old_addr || new_addr)) ||	/* (2-5) */
1695 		    (is_new && new_addr)) {			/* (7) */
1696 			ln_router = 1;
1697 		}
1698 		break;
1699 	}
1700 
1701 	return (ln_router);
1702 }
1703 
1704 /*
1705  * Create neighbor cache entry and cache link-layer address,
1706  * on reception of inbound ND6 packets.  (RS/RA/NS/redirect)
1707  *
1708  * type - ICMP6 type
1709  * code - type dependent information
1710  *
1711  */
1712 void
1713 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
1714     int lladdrlen, int type, int code)
1715 {
1716 	struct llentry *ln = NULL, *ln_tmp;
1717 	int is_newentry;
1718 	int do_update;
1719 	int olladdr;
1720 	int llchange;
1721 	int flags;
1722 	uint16_t router = 0;
1723 	struct sockaddr_in6 sin6;
1724 	struct mbuf *chain = NULL;
1725 
1726 	IF_AFDATA_UNLOCK_ASSERT(ifp);
1727 
1728 	KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__));
1729 	KASSERT(from != NULL, ("%s: from == NULL", __func__));
1730 
1731 	/* nothing must be updated for unspecified address */
1732 	if (IN6_IS_ADDR_UNSPECIFIED(from))
1733 		return;
1734 
1735 	/*
1736 	 * Validation about ifp->if_addrlen and lladdrlen must be done in
1737 	 * the caller.
1738 	 *
1739 	 * XXX If the link does not have link-layer adderss, what should
1740 	 * we do? (ifp->if_addrlen == 0)
1741 	 * Spec says nothing in sections for RA, RS and NA.  There's small
1742 	 * description on it in NS section (RFC 2461 7.2.3).
1743 	 */
1744 	flags = lladdr ? ND6_EXCLUSIVE : 0;
1745 	IF_AFDATA_RLOCK(ifp);
1746 	ln = nd6_lookup(from, flags, ifp);
1747 	IF_AFDATA_RUNLOCK(ifp);
1748 	is_newentry = 0;
1749 	if (ln == NULL) {
1750 		flags |= ND6_EXCLUSIVE;
1751 		ln = nd6_alloc(from, 0, ifp);
1752 		if (ln == NULL)
1753 			return;
1754 
1755 		/*
1756 		 * Since we already know all the data for the new entry,
1757 		 * fill it before insertion.
1758 		 */
1759 		if (lladdr != NULL) {
1760 			bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
1761 			ln->la_flags |= LLE_VALID;
1762 		}
1763 		IF_AFDATA_WLOCK(ifp);
1764 		LLE_WLOCK(ln);
1765 		/* Prefer any existing lle over newly-created one */
1766 		ln_tmp = nd6_lookup(from, ND6_EXCLUSIVE, ifp);
1767 		if (ln_tmp == NULL)
1768 			lltable_link_entry(LLTABLE6(ifp), ln);
1769 		IF_AFDATA_WUNLOCK(ifp);
1770 		if (ln_tmp == NULL) {
1771 			/* No existing lle, mark as new entry */
1772 			is_newentry = 1;
1773 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
1774 		} else {
1775 			lltable_free_entry(LLTABLE6(ifp), ln);
1776 			ln = ln_tmp;
1777 			ln_tmp = NULL;
1778 		}
1779 	}
1780 	/* do nothing if static ndp is set */
1781 	if ((ln->la_flags & LLE_STATIC)) {
1782 		if (flags & ND6_EXCLUSIVE)
1783 			LLE_WUNLOCK(ln);
1784 		else
1785 			LLE_RUNLOCK(ln);
1786 		return;
1787 	}
1788 
1789 	olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
1790 	if (olladdr && lladdr) {
1791 		llchange = bcmp(lladdr, &ln->ll_addr,
1792 		    ifp->if_addrlen);
1793 	} else if (!olladdr && lladdr)
1794 		llchange = 1;
1795 	else
1796 		llchange = 0;
1797 
1798 	/*
1799 	 * newentry olladdr  lladdr  llchange	(*=record)
1800 	 *	0	n	n	--	(1)
1801 	 *	0	y	n	--	(2)
1802 	 *	0	n	y	y	(3) * STALE
1803 	 *	0	y	y	n	(4) *
1804 	 *	0	y	y	y	(5) * STALE
1805 	 *	1	--	n	--	(6)   NOSTATE(= PASSIVE)
1806 	 *	1	--	y	--	(7) * STALE
1807 	 */
1808 
1809 	do_update = 0;
1810 	if (!is_newentry && llchange != 0)
1811 		do_update = 1;	/* (3,5) */
1812 
1813 	if (lladdr) {		/* (3-5) and (7) */
1814 		/*
1815 		 * Record source link-layer address
1816 		 * XXX is it dependent to ifp->if_type?
1817 		 */
1818 		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
1819 		ln->la_flags |= LLE_VALID;
1820 		nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
1821 
1822 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
1823 
1824 		if (do_update) {
1825 			if (ln->la_hold != NULL)
1826 				nd6_grab_holdchain(ln, &chain, &sin6);
1827 		}
1828 	}
1829 
1830 	/* Calculates new router status */
1831 	router = nd6_is_router(type, code, is_newentry, olladdr,
1832 	    lladdr != NULL ? 1 : 0, ln->ln_router);
1833 
1834 	ln->ln_router = router;
1835 	/* Mark non-router redirects with special flag */
1836 	if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER)
1837 		ln->la_flags |= LLE_REDIRECT;
1838 
1839 	if (flags & ND6_EXCLUSIVE)
1840 		LLE_WUNLOCK(ln);
1841 	else
1842 		LLE_RUNLOCK(ln);
1843 
1844 	if (chain != NULL)
1845 		nd6_flush_holdchain(ifp, ifp, chain, &sin6);
1846 
1847 	/*
1848 	 * When the link-layer address of a router changes, select the
1849 	 * best router again.  In particular, when the neighbor entry is newly
1850 	 * created, it might affect the selection policy.
1851 	 * Question: can we restrict the first condition to the "is_newentry"
1852 	 * case?
1853 	 * XXX: when we hear an RA from a new router with the link-layer
1854 	 * address option, defrouter_select() is called twice, since
1855 	 * defrtrlist_update called the function as well.  However, I believe
1856 	 * we can compromise the overhead, since it only happens the first
1857 	 * time.
1858 	 * XXX: although defrouter_select() should not have a bad effect
1859 	 * for those are not autoconfigured hosts, we explicitly avoid such
1860 	 * cases for safety.
1861 	 */
1862 	if ((do_update || is_newentry) && router &&
1863 	    ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
1864 		/*
1865 		 * guaranteed recursion
1866 		 */
1867 		defrouter_select();
1868 	}
1869 }
1870 
1871 static void
1872 nd6_slowtimo(void *arg)
1873 {
1874 	CURVNET_SET((struct vnet *) arg);
1875 	struct nd_ifinfo *nd6if;
1876 	struct ifnet *ifp;
1877 
1878 	callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1879 	    nd6_slowtimo, curvnet);
1880 	IFNET_RLOCK_NOSLEEP();
1881 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1882 		if (ifp->if_afdata[AF_INET6] == NULL)
1883 			continue;
1884 		nd6if = ND_IFINFO(ifp);
1885 		if (nd6if->basereachable && /* already initialized */
1886 		    (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1887 			/*
1888 			 * Since reachable time rarely changes by router
1889 			 * advertisements, we SHOULD insure that a new random
1890 			 * value gets recomputed at least once every few hours.
1891 			 * (RFC 2461, 6.3.4)
1892 			 */
1893 			nd6if->recalctm = V_nd6_recalc_reachtm_interval;
1894 			nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1895 		}
1896 	}
1897 	IFNET_RUNLOCK_NOSLEEP();
1898 	CURVNET_RESTORE();
1899 }
1900 
1901 void
1902 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain,
1903     struct sockaddr_in6 *sin6)
1904 {
1905 
1906 	LLE_WLOCK_ASSERT(ln);
1907 
1908 	*chain = ln->la_hold;
1909 	ln->la_hold = NULL;
1910 	lltable_fill_sa_entry(ln, (struct sockaddr *)sin6);
1911 
1912 	if (ln->ln_state == ND6_LLINFO_STALE) {
1913 
1914 		/*
1915 		 * The first time we send a packet to a
1916 		 * neighbor whose entry is STALE, we have
1917 		 * to change the state to DELAY and a sets
1918 		 * a timer to expire in DELAY_FIRST_PROBE_TIME
1919 		 * seconds to ensure do neighbor unreachability
1920 		 * detection on expiration.
1921 		 * (RFC 2461 7.3.3)
1922 		 */
1923 		nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY);
1924 	}
1925 }
1926 
1927 int
1928 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
1929     struct sockaddr_in6 *dst)
1930 {
1931 	int error;
1932 	int ip6len;
1933 	struct ip6_hdr *ip6;
1934 	struct m_tag *mtag;
1935 
1936 #ifdef MAC
1937 	mac_netinet6_nd6_send(ifp, m);
1938 #endif
1939 
1940 	/*
1941 	 * If called from nd6_ns_output() (NS), nd6_na_output() (NA),
1942 	 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA
1943 	 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND
1944 	 * to be diverted to user space.  When re-injected into the kernel,
1945 	 * send_output() will directly dispatch them to the outgoing interface.
1946 	 */
1947 	if (send_sendso_input_hook != NULL) {
1948 		mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL);
1949 		if (mtag != NULL) {
1950 			ip6 = mtod(m, struct ip6_hdr *);
1951 			ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
1952 			/* Use the SEND socket */
1953 			error = send_sendso_input_hook(m, ifp, SND_OUT,
1954 			    ip6len);
1955 			/* -1 == no app on SEND socket */
1956 			if (error == 0 || error != -1)
1957 			    return (error);
1958 		}
1959 	}
1960 
1961 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
1962 	IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL,
1963 	    mtod(m, struct ip6_hdr *));
1964 
1965 	if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1966 		origifp = ifp;
1967 
1968 	error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, NULL);
1969 	return (error);
1970 }
1971 
1972 /*
1973  * Do L2 address resolution for @sa_dst address. Stores found
1974  * address in @desten buffer. Copy of lle ln_flags can be also
1975  * saved in @pflags if @pflags is non-NULL.
1976  *
1977  * If destination LLE does not exists or lle state modification
1978  * is required, call "slow" version.
1979  *
1980  * Return values:
1981  * - 0 on success (address copied to buffer).
1982  * - EWOULDBLOCK (no local error, but address is still unresolved)
1983  * - other errors (alloc failure, etc)
1984  */
1985 int
1986 nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
1987     const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags)
1988 {
1989 	struct llentry *ln = NULL;
1990 	const struct sockaddr_in6 *dst6;
1991 
1992 	if (pflags != NULL)
1993 		*pflags = 0;
1994 
1995 	dst6 = (const struct sockaddr_in6 *)sa_dst;
1996 
1997 	/* discard the packet if IPv6 operation is disabled on the interface */
1998 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
1999 		m_freem(m);
2000 		return (ENETDOWN); /* better error? */
2001 	}
2002 
2003 	if (m != NULL && m->m_flags & M_MCAST) {
2004 		switch (ifp->if_type) {
2005 		case IFT_ETHER:
2006 		case IFT_FDDI:
2007 		case IFT_L2VLAN:
2008 		case IFT_IEEE80211:
2009 		case IFT_BRIDGE:
2010 		case IFT_ISO88025:
2011 			ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr,
2012 						 desten);
2013 			return (0);
2014 		default:
2015 			m_freem(m);
2016 			return (EAFNOSUPPORT);
2017 		}
2018 	}
2019 
2020 	IF_AFDATA_RLOCK(ifp);
2021 	ln = nd6_lookup(&dst6->sin6_addr, 0, ifp);
2022 	IF_AFDATA_RUNLOCK(ifp);
2023 
2024 	/*
2025 	 * Perform fast path for the following cases:
2026 	 * 1) lle state is REACHABLE
2027 	 * 2) lle state is DELAY (NS message sent)
2028 	 *
2029 	 * Every other case involves lle modification, so we handle
2030 	 * them separately.
2031 	 */
2032 	if (ln == NULL || (ln->ln_state != ND6_LLINFO_REACHABLE &&
2033 	    ln->ln_state != ND6_LLINFO_DELAY)) {
2034 		/* Fall back to slow processing path */
2035 		if (ln != NULL)
2036 			LLE_RUNLOCK(ln);
2037 		return (nd6_resolve_slow(ifp, m, dst6, desten, pflags));
2038 	}
2039 
2040 
2041 	bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
2042 	if (pflags != NULL)
2043 		*pflags = ln->la_flags;
2044 	LLE_RUNLOCK(ln);
2045 	return (0);
2046 }
2047 
2048 
2049 /*
2050  * Do L2 address resolution for @sa_dst address. Stores found
2051  * address in @desten buffer. Copy of lle ln_flags can be also
2052  * saved in @pflags if @pflags is non-NULL.
2053  *
2054  * Heavy version.
2055  * Function assume that destination LLE does not exist,
2056  * is invalid or stale, so ND6_EXCLUSIVE lock needs to be acquired.
2057  */
2058 static int
2059 nd6_resolve_slow(struct ifnet *ifp, struct mbuf *m,
2060     const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags)
2061 {
2062 	struct llentry *lle = NULL, *lle_tmp;
2063 	struct in6_addr *psrc, src;
2064 	int send_ns;
2065 
2066 	/*
2067 	 * Address resolution or Neighbor Unreachability Detection
2068 	 * for the next hop.
2069 	 * At this point, the destination of the packet must be a unicast
2070 	 * or an anycast address(i.e. not a multicast).
2071 	 */
2072 	if (lle == NULL) {
2073 		IF_AFDATA_RLOCK(ifp);
2074 		lle = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp);
2075 		IF_AFDATA_RUNLOCK(ifp);
2076 		if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp))  {
2077 			/*
2078 			 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
2079 			 * the condition below is not very efficient.  But we believe
2080 			 * it is tolerable, because this should be a rare case.
2081 			 */
2082 			lle = nd6_alloc(&dst->sin6_addr, 0, ifp);
2083 			if (lle == NULL) {
2084 				char ip6buf[INET6_ADDRSTRLEN];
2085 				log(LOG_DEBUG,
2086 				    "nd6_output: can't allocate llinfo for %s "
2087 				    "(ln=%p)\n",
2088 				    ip6_sprintf(ip6buf, &dst->sin6_addr), lle);
2089 				m_freem(m);
2090 				return (ENOBUFS);
2091 			}
2092 
2093 			IF_AFDATA_WLOCK(ifp);
2094 			LLE_WLOCK(lle);
2095 			/* Prefer any existing entry over newly-created one */
2096 			lle_tmp = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp);
2097 			if (lle_tmp == NULL)
2098 				lltable_link_entry(LLTABLE6(ifp), lle);
2099 			IF_AFDATA_WUNLOCK(ifp);
2100 			if (lle_tmp != NULL) {
2101 				lltable_free_entry(LLTABLE6(ifp), lle);
2102 				lle = lle_tmp;
2103 				lle_tmp = NULL;
2104 			}
2105 		}
2106 	}
2107 	if (lle == NULL) {
2108 		if (!(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
2109 			m_freem(m);
2110 			return (ENOBUFS);
2111 		}
2112 
2113 		if (m != NULL)
2114 			m_freem(m);
2115 		return (ENOBUFS);
2116 	}
2117 
2118 	LLE_WLOCK_ASSERT(lle);
2119 
2120 	/*
2121 	 * The first time we send a packet to a neighbor whose entry is
2122 	 * STALE, we have to change the state to DELAY and a sets a timer to
2123 	 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
2124 	 * neighbor unreachability detection on expiration.
2125 	 * (RFC 2461 7.3.3)
2126 	 */
2127 	if (lle->ln_state == ND6_LLINFO_STALE)
2128 		nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY);
2129 
2130 	/*
2131 	 * If the neighbor cache entry has a state other than INCOMPLETE
2132 	 * (i.e. its link-layer address is already resolved), just
2133 	 * send the packet.
2134 	 */
2135 	if (lle->ln_state > ND6_LLINFO_INCOMPLETE) {
2136 		bcopy(&lle->ll_addr, desten, ifp->if_addrlen);
2137 		if (pflags != NULL)
2138 			*pflags = lle->la_flags;
2139 		LLE_WUNLOCK(lle);
2140 		return (0);
2141 	}
2142 
2143 	/*
2144 	 * There is a neighbor cache entry, but no ethernet address
2145 	 * response yet.  Append this latest packet to the end of the
2146 	 * packet queue in the mbuf, unless the number of the packet
2147 	 * does not exceed nd6_maxqueuelen.  When it exceeds nd6_maxqueuelen,
2148 	 * the oldest packet in the queue will be removed.
2149 	 */
2150 
2151 	if (lle->la_hold != NULL) {
2152 		struct mbuf *m_hold;
2153 		int i;
2154 
2155 		i = 0;
2156 		for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){
2157 			i++;
2158 			if (m_hold->m_nextpkt == NULL) {
2159 				m_hold->m_nextpkt = m;
2160 				break;
2161 			}
2162 		}
2163 		while (i >= V_nd6_maxqueuelen) {
2164 			m_hold = lle->la_hold;
2165 			lle->la_hold = lle->la_hold->m_nextpkt;
2166 			m_freem(m_hold);
2167 			i--;
2168 		}
2169 	} else {
2170 		lle->la_hold = m;
2171 	}
2172 
2173 	/*
2174 	 * If there has been no NS for the neighbor after entering the
2175 	 * INCOMPLETE state, send the first solicitation.
2176 	 * Note that for newly-created lle la_asked will be 0,
2177 	 * so we will transition from ND6_LLINFO_NOSTATE to
2178 	 * ND6_LLINFO_INCOMPLETE state here.
2179 	 */
2180 	psrc = NULL;
2181 	send_ns = 0;
2182 	if (lle->la_asked == 0) {
2183 		lle->la_asked++;
2184 		send_ns = 1;
2185 		psrc = nd6_llinfo_get_holdsrc(lle, &src);
2186 
2187 		nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE);
2188 	}
2189 	LLE_WUNLOCK(lle);
2190 	if (send_ns != 0)
2191 		nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL);
2192 
2193 	return (EWOULDBLOCK);
2194 }
2195 
2196 
2197 int
2198 nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
2199     struct sockaddr_in6 *dst)
2200 {
2201 	struct mbuf *m, *m_head;
2202 	struct ifnet *outifp;
2203 	int error = 0;
2204 
2205 	m_head = chain;
2206 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2207 		outifp = origifp;
2208 	else
2209 		outifp = ifp;
2210 
2211 	while (m_head) {
2212 		m = m_head;
2213 		m_head = m_head->m_nextpkt;
2214 		error = nd6_output_ifp(ifp, origifp, m, dst);
2215 	}
2216 
2217 	/*
2218 	 * XXX
2219 	 * note that intermediate errors are blindly ignored
2220 	 */
2221 	return (error);
2222 }
2223 
2224 static int
2225 nd6_need_cache(struct ifnet *ifp)
2226 {
2227 	/*
2228 	 * XXX: we currently do not make neighbor cache on any interface
2229 	 * other than ARCnet, Ethernet, FDDI and GIF.
2230 	 *
2231 	 * RFC2893 says:
2232 	 * - unidirectional tunnels needs no ND
2233 	 */
2234 	switch (ifp->if_type) {
2235 	case IFT_ARCNET:
2236 	case IFT_ETHER:
2237 	case IFT_FDDI:
2238 	case IFT_IEEE1394:
2239 	case IFT_L2VLAN:
2240 	case IFT_IEEE80211:
2241 	case IFT_INFINIBAND:
2242 	case IFT_BRIDGE:
2243 	case IFT_PROPVIRTUAL:
2244 		return (1);
2245 	default:
2246 		return (0);
2247 	}
2248 }
2249 
2250 /*
2251  * Add pernament ND6 link-layer record for given
2252  * interface address.
2253  *
2254  * Very similar to IPv4 arp_ifinit(), but:
2255  * 1) IPv6 DAD is performed in different place
2256  * 2) It is called by IPv6 protocol stack in contrast to
2257  * arp_ifinit() which is typically called in SIOCSIFADDR
2258  * driver ioctl handler.
2259  *
2260  */
2261 int
2262 nd6_add_ifa_lle(struct in6_ifaddr *ia)
2263 {
2264 	struct ifnet *ifp;
2265 	struct llentry *ln, *ln_tmp;
2266 	struct sockaddr *dst;
2267 
2268 	ifp = ia->ia_ifa.ifa_ifp;
2269 	if (nd6_need_cache(ifp) == 0)
2270 		return (0);
2271 
2272 	ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
2273 	dst = (struct sockaddr *)&ia->ia_addr;
2274 	ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst);
2275 	if (ln == NULL)
2276 		return (ENOBUFS);
2277 
2278 	IF_AFDATA_WLOCK(ifp);
2279 	LLE_WLOCK(ln);
2280 	/* Unlink any entry if exists */
2281 	ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst);
2282 	if (ln_tmp != NULL)
2283 		lltable_unlink_entry(LLTABLE6(ifp), ln_tmp);
2284 	lltable_link_entry(LLTABLE6(ifp), ln);
2285 	IF_AFDATA_WUNLOCK(ifp);
2286 
2287 	if (ln_tmp != NULL)
2288 		EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED);
2289 	EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
2290 
2291 	LLE_WUNLOCK(ln);
2292 	if (ln_tmp != NULL)
2293 		llentry_free(ln_tmp);
2294 
2295 	return (0);
2296 }
2297 
2298 /*
2299  * Removes either all lle entries for given @ia, or lle
2300  * corresponding to @ia address.
2301  */
2302 void
2303 nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all)
2304 {
2305 	struct sockaddr_in6 mask, addr;
2306 	struct sockaddr *saddr, *smask;
2307 	struct ifnet *ifp;
2308 
2309 	ifp = ia->ia_ifa.ifa_ifp;
2310 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
2311 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
2312 	saddr = (struct sockaddr *)&addr;
2313 	smask = (struct sockaddr *)&mask;
2314 
2315 	if (all != 0)
2316 		lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC);
2317 	else
2318 		lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr);
2319 }
2320 
2321 static void
2322 clear_llinfo_pqueue(struct llentry *ln)
2323 {
2324 	struct mbuf *m_hold, *m_hold_next;
2325 
2326 	for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
2327 		m_hold_next = m_hold->m_nextpkt;
2328 		m_freem(m_hold);
2329 	}
2330 
2331 	ln->la_hold = NULL;
2332 	return;
2333 }
2334 
2335 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2336 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2337 #ifdef SYSCTL_DECL
2338 SYSCTL_DECL(_net_inet6_icmp6);
2339 #endif
2340 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2341 	CTLFLAG_RD, nd6_sysctl_drlist, "");
2342 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2343 	CTLFLAG_RD, nd6_sysctl_prlist, "");
2344 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
2345 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
2346 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer,
2347 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), "");
2348 
2349 static int
2350 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2351 {
2352 	struct in6_defrouter d;
2353 	struct nd_defrouter *dr;
2354 	int error;
2355 
2356 	if (req->newptr)
2357 		return (EPERM);
2358 
2359 	bzero(&d, sizeof(d));
2360 	d.rtaddr.sin6_family = AF_INET6;
2361 	d.rtaddr.sin6_len = sizeof(d.rtaddr);
2362 
2363 	/*
2364 	 * XXX locking
2365 	 */
2366 	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
2367 		d.rtaddr.sin6_addr = dr->rtaddr;
2368 		error = sa6_recoverscope(&d.rtaddr);
2369 		if (error != 0)
2370 			return (error);
2371 		d.flags = dr->flags;
2372 		d.rtlifetime = dr->rtlifetime;
2373 		d.expire = dr->expire + (time_second - time_uptime);
2374 		d.if_index = dr->ifp->if_index;
2375 		error = SYSCTL_OUT(req, &d, sizeof(d));
2376 		if (error != 0)
2377 			return (error);
2378 	}
2379 	return (0);
2380 }
2381 
2382 static int
2383 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2384 {
2385 	struct in6_prefix p;
2386 	struct sockaddr_in6 s6;
2387 	struct nd_prefix *pr;
2388 	struct nd_pfxrouter *pfr;
2389 	time_t maxexpire;
2390 	int error;
2391 	char ip6buf[INET6_ADDRSTRLEN];
2392 
2393 	if (req->newptr)
2394 		return (EPERM);
2395 
2396 	bzero(&p, sizeof(p));
2397 	p.origin = PR_ORIG_RA;
2398 	bzero(&s6, sizeof(s6));
2399 	s6.sin6_family = AF_INET6;
2400 	s6.sin6_len = sizeof(s6);
2401 
2402 	/*
2403 	 * XXX locking
2404 	 */
2405 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
2406 		p.prefix = pr->ndpr_prefix;
2407 		if (sa6_recoverscope(&p.prefix)) {
2408 			log(LOG_ERR, "scope error in prefix list (%s)\n",
2409 			    ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
2410 			/* XXX: press on... */
2411 		}
2412 		p.raflags = pr->ndpr_raf;
2413 		p.prefixlen = pr->ndpr_plen;
2414 		p.vltime = pr->ndpr_vltime;
2415 		p.pltime = pr->ndpr_pltime;
2416 		p.if_index = pr->ndpr_ifp->if_index;
2417 		if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2418 			p.expire = 0;
2419 		else {
2420 			/* XXX: we assume time_t is signed. */
2421 			maxexpire = (-1) &
2422 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
2423 			if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
2424 				p.expire = pr->ndpr_lastupdate +
2425 				    pr->ndpr_vltime +
2426 				    (time_second - time_uptime);
2427 			else
2428 				p.expire = maxexpire;
2429 		}
2430 		p.refcnt = pr->ndpr_refcnt;
2431 		p.flags = pr->ndpr_stateflags;
2432 		p.advrtrs = 0;
2433 		LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
2434 			p.advrtrs++;
2435 		error = SYSCTL_OUT(req, &p, sizeof(p));
2436 		if (error != 0)
2437 			return (error);
2438 		LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
2439 			s6.sin6_addr = pfr->router->rtaddr;
2440 			if (sa6_recoverscope(&s6))
2441 				log(LOG_ERR,
2442 				    "scope error in prefix list (%s)\n",
2443 				    ip6_sprintf(ip6buf, &pfr->router->rtaddr));
2444 			error = SYSCTL_OUT(req, &s6, sizeof(s6));
2445 			if (error != 0)
2446 				return (error);
2447 		}
2448 	}
2449 	return (0);
2450 }
2451