nd6.c revision 78468
1/*	$FreeBSD: head/sys/netinet6/nd6.c 78468 2001-06-19 14:48:02Z sumikawa $	*/
2/*	$KAME: nd6.c,v 1.144 2001/05/24 07:44:00 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/*
34 * XXX
35 * KAME 970409 note:
36 * BSD/OS version heavily modifies this code, related to llinfo.
37 * Since we don't have BSD/OS version of net/route.c in our hand,
38 * I left the code mostly as it was in 970310.  -- itojun
39 */
40
41#include "opt_inet.h"
42#include "opt_inet6.h"
43
44#include <sys/param.h>
45#include <sys/systm.h>
46#include <sys/callout.h>
47#include <sys/malloc.h>
48#include <sys/mbuf.h>
49#include <sys/socket.h>
50#include <sys/sockio.h>
51#include <sys/time.h>
52#include <sys/kernel.h>
53#include <sys/protosw.h>
54#include <sys/errno.h>
55#include <sys/syslog.h>
56#include <sys/queue.h>
57#include <sys/sysctl.h>
58
59#include <net/if.h>
60#include <net/if_dl.h>
61#include <net/if_types.h>
62#include <net/if_atm.h>
63#include <net/route.h>
64
65#include <netinet/in.h>
66#include <netinet/if_ether.h>
67#include <netinet/if_fddi.h>
68#include <netinet6/in6_var.h>
69#include <netinet/ip6.h>
70#include <netinet6/ip6_var.h>
71#include <netinet6/nd6.h>
72#include <netinet6/in6_prefix.h>
73#include <netinet/icmp6.h>
74
75#include <net/net_osdep.h>
76
77#define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
78#define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
79
80#define SIN6(s) ((struct sockaddr_in6 *)s)
81#define SDL(s) ((struct sockaddr_dl *)s)
82
83/* timer values */
84int	nd6_prune	= 1;	/* walk list every 1 seconds */
85int	nd6_delay	= 5;	/* delay first probe time 5 second */
86int	nd6_umaxtries	= 3;	/* maximum unicast query */
87int	nd6_mmaxtries	= 3;	/* maximum multicast query */
88int	nd6_useloopback = 1;	/* use loopback interface for local traffic */
89int	nd6_gctimer	= (60 * 60 * 24); /* 1 day: garbage collection timer */
90
91/* preventing too many loops in ND option parsing */
92int nd6_maxndopt = 10;	/* max # of ND options allowed */
93
94int nd6_maxnudhint = 0;	/* max # of subsequent upper layer hints */
95
96#ifdef ND6_DEBUG
97int nd6_debug = 1;
98#else
99int nd6_debug = 0;
100#endif
101
102/* for debugging? */
103static int nd6_inuse, nd6_allocated;
104
105struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6};
106static size_t nd_ifinfo_indexlim = 8;
107struct nd_ifinfo *nd_ifinfo = NULL;
108struct nd_drhead nd_defrouter;
109struct nd_prhead nd_prefix = { 0 };
110
111int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
112static struct sockaddr_in6 all1_sa;
113
114static void nd6_slowtimo __P((void *));
115static int regen_tmpaddr __P((struct in6_ifaddr *));
116
117struct callout nd6_slowtimo_ch;
118struct callout nd6_timer_ch;
119extern struct callout in6_tmpaddrtimer_ch;
120
121void
122nd6_init()
123{
124	static int nd6_init_done = 0;
125	int i;
126
127	if (nd6_init_done) {
128		log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
129		return;
130	}
131
132	all1_sa.sin6_family = AF_INET6;
133	all1_sa.sin6_len = sizeof(struct sockaddr_in6);
134	for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
135		all1_sa.sin6_addr.s6_addr[i] = 0xff;
136
137	/* initialization of the default router list */
138	TAILQ_INIT(&nd_defrouter);
139
140	nd6_init_done = 1;
141
142	/* start timer */
143	callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
144	    nd6_slowtimo, NULL);
145}
146
147void
148nd6_ifattach(ifp)
149	struct ifnet *ifp;
150{
151
152	/*
153	 * We have some arrays that should be indexed by if_index.
154	 * since if_index will grow dynamically, they should grow too.
155	 */
156	if (nd_ifinfo == NULL || if_index >= nd_ifinfo_indexlim) {
157		size_t n;
158		caddr_t q;
159
160		while (if_index >= nd_ifinfo_indexlim)
161			nd_ifinfo_indexlim <<= 1;
162
163		/* grow nd_ifinfo */
164		n = nd_ifinfo_indexlim * sizeof(struct nd_ifinfo);
165		q = (caddr_t)malloc(n, M_IP6NDP, M_WAITOK);
166		bzero(q, n);
167		if (nd_ifinfo) {
168			bcopy((caddr_t)nd_ifinfo, q, n/2);
169			free((caddr_t)nd_ifinfo, M_IP6NDP);
170		}
171		nd_ifinfo = (struct nd_ifinfo *)q;
172	}
173
174#define ND nd_ifinfo[ifp->if_index]
175
176	/*
177	 * Don't initialize if called twice.
178	 * XXX: to detect this, we should choose a member that is never set
179	 * before initialization of the ND structure itself.  We formaly used
180	 * the linkmtu member, which was not suitable because it could be
181	 * initialized via "ifconfig mtu".
182	 */
183	if (ND.basereachable)
184		return;
185
186	ND.linkmtu = ifindex2ifnet[ifp->if_index]->if_mtu;
187	ND.chlim = IPV6_DEFHLIM;
188	ND.basereachable = REACHABLE_TIME;
189	ND.reachable = ND_COMPUTE_RTIME(ND.basereachable);
190	ND.retrans = RETRANS_TIMER;
191	ND.receivedra = 0;
192	ND.flags = ND6_IFF_PERFORMNUD;
193	nd6_setmtu(ifp);
194#undef ND
195}
196
197/*
198 * Reset ND level link MTU. This function is called when the physical MTU
199 * changes, which means we might have to adjust the ND level MTU.
200 */
201void
202nd6_setmtu(ifp)
203	struct ifnet *ifp;
204{
205#define MIN(a,b) ((a) < (b) ? (a) : (b))
206	struct nd_ifinfo *ndi = &nd_ifinfo[ifp->if_index];
207	u_long oldmaxmtu = ndi->maxmtu;
208	u_long oldlinkmtu = ndi->linkmtu;
209
210	switch (ifp->if_type) {
211	case IFT_ARCNET:	/* XXX MTU handling needs more work */
212		ndi->maxmtu = MIN(60480, ifp->if_mtu);
213		break;
214	case IFT_ETHER:
215		ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
216		break;
217	case IFT_FDDI:
218		ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu);
219		break;
220	case IFT_ATM:
221		ndi->maxmtu = MIN(ATMMTU, ifp->if_mtu);
222		break;
223	case IFT_IEEE1394:	/* XXX should be IEEE1394MTU(1500) */
224		ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
225		break;
226#ifdef IFT_IEEE80211
227	case IFT_IEEE80211:	/* XXX should be IEEE80211MTU(1500) */
228		ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
229		break;
230#endif
231	default:
232		ndi->maxmtu = ifp->if_mtu;
233		break;
234	}
235
236	if (oldmaxmtu != ndi->maxmtu) {
237		/*
238		 * If the ND level MTU is not set yet, or if the maxmtu
239		 * is reset to a smaller value than the ND level MTU,
240		 * also reset the ND level MTU.
241		 */
242		if (ndi->linkmtu == 0 ||
243		    ndi->maxmtu < ndi->linkmtu) {
244			ndi->linkmtu = ndi->maxmtu;
245			/* also adjust in6_maxmtu if necessary. */
246			if (oldlinkmtu == 0) {
247				/*
248				 * XXX: the case analysis is grotty, but
249				 * it is not efficient to call in6_setmaxmtu()
250				 * here when we are during the initialization
251				 * procedure.
252				 */
253				if (in6_maxmtu < ndi->linkmtu)
254					in6_maxmtu = ndi->linkmtu;
255			} else
256				in6_setmaxmtu();
257		}
258	}
259#undef MIN
260}
261
262void
263nd6_option_init(opt, icmp6len, ndopts)
264	void *opt;
265	int icmp6len;
266	union nd_opts *ndopts;
267{
268	bzero(ndopts, sizeof(*ndopts));
269	ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
270	ndopts->nd_opts_last
271		= (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
272
273	if (icmp6len == 0) {
274		ndopts->nd_opts_done = 1;
275		ndopts->nd_opts_search = NULL;
276	}
277}
278
279/*
280 * Take one ND option.
281 */
282struct nd_opt_hdr *
283nd6_option(ndopts)
284	union nd_opts *ndopts;
285{
286	struct nd_opt_hdr *nd_opt;
287	int olen;
288
289	if (!ndopts)
290		panic("ndopts == NULL in nd6_option\n");
291	if (!ndopts->nd_opts_last)
292		panic("uninitialized ndopts in nd6_option\n");
293	if (!ndopts->nd_opts_search)
294		return NULL;
295	if (ndopts->nd_opts_done)
296		return NULL;
297
298	nd_opt = ndopts->nd_opts_search;
299
300	/* make sure nd_opt_len is inside the buffer */
301	if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
302		bzero(ndopts, sizeof(*ndopts));
303		return NULL;
304	}
305
306	olen = nd_opt->nd_opt_len << 3;
307	if (olen == 0) {
308		/*
309		 * Message validation requires that all included
310		 * options have a length that is greater than zero.
311		 */
312		bzero(ndopts, sizeof(*ndopts));
313		return NULL;
314	}
315
316	ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
317	if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
318		/* option overruns the end of buffer, invalid */
319		bzero(ndopts, sizeof(*ndopts));
320		return NULL;
321	} else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
322		/* reached the end of options chain */
323		ndopts->nd_opts_done = 1;
324		ndopts->nd_opts_search = NULL;
325	}
326	return nd_opt;
327}
328
329/*
330 * Parse multiple ND options.
331 * This function is much easier to use, for ND routines that do not need
332 * multiple options of the same type.
333 */
334int
335nd6_options(ndopts)
336	union nd_opts *ndopts;
337{
338	struct nd_opt_hdr *nd_opt;
339	int i = 0;
340
341	if (!ndopts)
342		panic("ndopts == NULL in nd6_options\n");
343	if (!ndopts->nd_opts_last)
344		panic("uninitialized ndopts in nd6_options\n");
345	if (!ndopts->nd_opts_search)
346		return 0;
347
348	while (1) {
349		nd_opt = nd6_option(ndopts);
350		if (!nd_opt && !ndopts->nd_opts_last) {
351			/*
352			 * Message validation requires that all included
353			 * options have a length that is greater than zero.
354			 */
355			icmp6stat.icp6s_nd_badopt++;
356			bzero(ndopts, sizeof(*ndopts));
357			return -1;
358		}
359
360		if (!nd_opt)
361			goto skip1;
362
363		switch (nd_opt->nd_opt_type) {
364		case ND_OPT_SOURCE_LINKADDR:
365		case ND_OPT_TARGET_LINKADDR:
366		case ND_OPT_MTU:
367		case ND_OPT_REDIRECTED_HEADER:
368			if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
369				nd6log((LOG_INFO,
370				    "duplicated ND6 option found (type=%d)\n",
371				    nd_opt->nd_opt_type));
372				/* XXX bark? */
373			} else {
374				ndopts->nd_opt_array[nd_opt->nd_opt_type]
375					= nd_opt;
376			}
377			break;
378		case ND_OPT_PREFIX_INFORMATION:
379			if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
380				ndopts->nd_opt_array[nd_opt->nd_opt_type]
381					= nd_opt;
382			}
383			ndopts->nd_opts_pi_end =
384				(struct nd_opt_prefix_info *)nd_opt;
385			break;
386		default:
387			/*
388			 * Unknown options must be silently ignored,
389			 * to accomodate future extension to the protocol.
390			 */
391			nd6log((LOG_DEBUG,
392			    "nd6_options: unsupported option %d - "
393			    "option ignored\n", nd_opt->nd_opt_type));
394		}
395
396skip1:
397		i++;
398		if (i > nd6_maxndopt) {
399			icmp6stat.icp6s_nd_toomanyopt++;
400			nd6log((LOG_INFO, "too many loop in nd opt\n"));
401			break;
402		}
403
404		if (ndopts->nd_opts_done)
405			break;
406	}
407
408	return 0;
409}
410
411/*
412 * ND6 timer routine to expire default route list and prefix list
413 */
414void
415nd6_timer(ignored_arg)
416	void	*ignored_arg;
417{
418	int s;
419	struct llinfo_nd6 *ln;
420	struct nd_defrouter *dr;
421	struct nd_prefix *pr;
422	struct ifnet *ifp;
423	struct in6_ifaddr *ia6, *nia6;
424	struct in6_addrlifetime *lt6;
425
426	s = splnet();
427	callout_reset(&nd6_timer_ch, nd6_prune * hz,
428		      nd6_timer, NULL);
429
430	ln = llinfo_nd6.ln_next;
431	/* XXX BSD/OS separates this code -- itojun */
432	while (ln && ln != &llinfo_nd6) {
433		struct rtentry *rt;
434		struct sockaddr_in6 *dst;
435		struct llinfo_nd6 *next = ln->ln_next;
436		/* XXX: used for the DELAY case only: */
437		struct nd_ifinfo *ndi = NULL;
438
439		if ((rt = ln->ln_rt) == NULL) {
440			ln = next;
441			continue;
442		}
443		if ((ifp = rt->rt_ifp) == NULL) {
444			ln = next;
445			continue;
446		}
447		ndi = &nd_ifinfo[ifp->if_index];
448		dst = (struct sockaddr_in6 *)rt_key(rt);
449
450		if (ln->ln_expire > time_second) {
451			ln = next;
452			continue;
453		}
454
455		/* sanity check */
456		if (!rt)
457			panic("rt=0 in nd6_timer(ln=%p)\n", ln);
458		if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln)
459			panic("rt_llinfo(%p) is not equal to ln(%p)\n",
460			      rt->rt_llinfo, ln);
461		if (!dst)
462			panic("dst=0 in nd6_timer(ln=%p)\n", ln);
463
464		switch (ln->ln_state) {
465		case ND6_LLINFO_INCOMPLETE:
466			if (ln->ln_asked < nd6_mmaxtries) {
467				ln->ln_asked++;
468				ln->ln_expire = time_second +
469					nd_ifinfo[ifp->if_index].retrans / 1000;
470				nd6_ns_output(ifp, NULL, &dst->sin6_addr,
471					ln, 0);
472			} else {
473				struct mbuf *m = ln->ln_hold;
474				if (m) {
475					if (rt->rt_ifp) {
476						/*
477						 * Fake rcvif to make ICMP error
478						 * more helpful in diagnosing
479						 * for the receiver.
480						 * XXX: should we consider
481						 * older rcvif?
482						 */
483						m->m_pkthdr.rcvif = rt->rt_ifp;
484					}
485					icmp6_error(m, ICMP6_DST_UNREACH,
486						    ICMP6_DST_UNREACH_ADDR, 0);
487					ln->ln_hold = NULL;
488				}
489				next = nd6_free(rt);
490			}
491			break;
492		case ND6_LLINFO_REACHABLE:
493			if (ln->ln_expire) {
494				ln->ln_state = ND6_LLINFO_STALE;
495				ln->ln_expire = time_second + nd6_gctimer;
496			}
497			break;
498
499		case ND6_LLINFO_STALE:
500			/* Garbage Collection(RFC 2461 5.3) */
501			if (ln->ln_expire)
502				next = nd6_free(rt);
503			break;
504
505		case ND6_LLINFO_DELAY:
506			if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
507				/* We need NUD */
508				ln->ln_asked = 1;
509				ln->ln_state = ND6_LLINFO_PROBE;
510				ln->ln_expire = time_second +
511					ndi->retrans / 1000;
512				nd6_ns_output(ifp, &dst->sin6_addr,
513					      &dst->sin6_addr,
514					      ln, 0);
515			} else {
516				ln->ln_state = ND6_LLINFO_STALE; /* XXX */
517				ln->ln_expire = time_second + nd6_gctimer;
518			}
519			break;
520		case ND6_LLINFO_PROBE:
521			if (ln->ln_asked < nd6_umaxtries) {
522				ln->ln_asked++;
523				ln->ln_expire = time_second +
524					nd_ifinfo[ifp->if_index].retrans / 1000;
525				nd6_ns_output(ifp, &dst->sin6_addr,
526					       &dst->sin6_addr, ln, 0);
527			} else {
528				next = nd6_free(rt);
529			}
530			break;
531		}
532		ln = next;
533	}
534
535	/* expire default router list */
536	dr = TAILQ_FIRST(&nd_defrouter);
537	while (dr) {
538		if (dr->expire && dr->expire < time_second) {
539			struct nd_defrouter *t;
540			t = TAILQ_NEXT(dr, dr_entry);
541			defrtrlist_del(dr);
542			dr = t;
543		} else {
544			dr = TAILQ_NEXT(dr, dr_entry);
545		}
546	}
547
548	/*
549	 * expire interface addresses.
550	 * in the past the loop was inside prefix expiry processing.
551	 * However, from a stricter speci-confrmance standpoint, we should
552	 * rather separate address lifetimes and prefix lifetimes.
553	 */
554  addrloop:
555	for (ia6 = in6_ifaddr; ia6; ia6 = nia6) {
556		nia6 = ia6->ia_next;
557		/* check address lifetime */
558		lt6 = &ia6->ia6_lifetime;
559		if (IFA6_IS_INVALID(ia6)) {
560			int regen = 0;
561
562			/*
563			 * If the expiring address is temporary, try
564			 * regenerating a new one.  This would be useful when
565			 * we suspended a laptop PC, then turned on after a
566			 * period that could invalidate all temporary
567			 * addresses.  Although we may have to restart the
568			 * loop (see below), it must be after purging the
569			 * address.  Otherwise, we'd see an infinite loop of
570			 * regeneration.
571			 */
572			if (ip6_use_tempaddr &&
573			    (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
574				if (regen_tmpaddr(ia6) == 0)
575					regen = 1;
576			}
577
578			in6_purgeaddr(&ia6->ia_ifa);
579
580			if (regen)
581				goto addrloop; /* XXX: see below */
582		} else if (IFA6_IS_DEPRECATED(ia6)) {
583			int oldflags = ia6->ia6_flags;
584
585			ia6->ia6_flags |= IN6_IFF_DEPRECATED;
586
587			/*
588			 * If a temporary address has just become deprecated,
589			 * regenerate a new one if possible.
590			 */
591			if (ip6_use_tempaddr &&
592			    (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
593			    (oldflags & IN6_IFF_DEPRECATED) == 0) {
594
595				if (regen_tmpaddr(ia6) == 0) {
596					/*
597					 * A new temporary address is
598					 * generated.
599					 * XXX: this means the address chain
600					 * has changed while we are still in
601					 * the loop.  Although the change
602					 * would not cause disaster (because
603					 * it's not an addition, but a
604					 * deletion,) we'd rather restart the
605					 * loop just for safety.  Or does this
606					 * significantly reduce performance??
607					 */
608					goto addrloop;
609				}
610			}
611		} else if (IFA6_IS_DEPRECATED(ia6)) {
612			/*
613			 * A new RA might have made a deprecated address
614			 * preferred.
615			 */
616			ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
617		}
618	}
619
620	/* expire prefix list */
621	pr = nd_prefix.lh_first;
622	while (pr) {
623		/*
624		 * check prefix lifetime.
625		 * since pltime is just for autoconf, pltime processing for
626		 * prefix is not necessary.
627		 *
628		 * we offset expire time by NDPR_KEEP_EXPIRE, so that we
629		 * can use the old prefix information to validate the
630		 * next prefix information to come.  See prelist_update()
631		 * for actual validation.
632		 *
633		 * I don't think such an offset is necessary.
634		 * (jinmei@kame.net, 20010130).
635		 */
636		if (pr->ndpr_expire && pr->ndpr_expire < time_second) {
637			struct nd_prefix *t;
638			t = pr->ndpr_next;
639
640			/*
641			 * address expiration and prefix expiration are
642			 * separate.  NEVER perform in6_purgeaddr here.
643			 */
644
645			prelist_remove(pr);
646			pr = t;
647		} else
648			pr = pr->ndpr_next;
649	}
650	splx(s);
651}
652
653static int
654regen_tmpaddr(ia6)
655	struct in6_ifaddr *ia6; /* deprecated/invalidated temporary address */
656{
657	struct ifaddr *ifa;
658	struct ifnet *ifp;
659	struct in6_ifaddr *public_ifa6 = NULL;
660
661	ifp = ia6->ia_ifa.ifa_ifp;
662	for (ifa = ifp->if_addrlist.tqh_first; ifa;
663	     ifa = ifa->ifa_list.tqe_next)
664	{
665		struct in6_ifaddr *it6;
666
667		if (ifa->ifa_addr->sa_family != AF_INET6)
668			continue;
669
670		it6 = (struct in6_ifaddr *)ifa;
671
672		/* ignore no autoconf addresses. */
673		if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
674			continue;
675
676		/* ignore autoconf addresses with different prefixes. */
677		if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
678			continue;
679
680		/*
681		 * Now we are looking at an autoconf address with the same
682		 * prefix as ours.  If the address is temporary and is still
683		 * preferred, do not create another one.  It would be rare, but
684		 * could happen, for example, when we resume a laptop PC after
685		 * a long period.
686		 */
687		if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
688		    !IFA6_IS_DEPRECATED(it6)) {
689			public_ifa6 = NULL;
690			break;
691		}
692
693		/*
694		 * This is a public autoconf address that has the same prefix
695		 * as ours.  If it is preferred, keep it.  We can't break the
696		 * loop here, because there may be a still-preferred temporary
697		 * address with the prefix.
698		 */
699		if (!IFA6_IS_DEPRECATED(it6))
700		    public_ifa6 = it6;
701	}
702
703	if (public_ifa6 != NULL) {
704		int e;
705
706		if ((e = in6_tmpifadd(public_ifa6, 0)) != 0) {
707			log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
708			    " tmp addr,errno=%d\n", e);
709			return(-1);
710		}
711		return(0);
712	}
713
714	return(-1);
715}
716
717/*
718 * Nuke neighbor cache/prefix/default router management table, right before
719 * ifp goes away.
720 */
721void
722nd6_purge(ifp)
723	struct ifnet *ifp;
724{
725	struct llinfo_nd6 *ln, *nln;
726	struct nd_defrouter *dr, *ndr, drany;
727	struct nd_prefix *pr, *npr;
728
729	/* Nuke default router list entries toward ifp */
730	if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
731		/*
732		 * The first entry of the list may be stored in
733		 * the routing table, so we'll delete it later.
734		 */
735		for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = ndr) {
736			ndr = TAILQ_NEXT(dr, dr_entry);
737			if (dr->ifp == ifp)
738				defrtrlist_del(dr);
739		}
740		dr = TAILQ_FIRST(&nd_defrouter);
741		if (dr->ifp == ifp)
742			defrtrlist_del(dr);
743	}
744
745	/* Nuke prefix list entries toward ifp */
746	for (pr = nd_prefix.lh_first; pr; pr = npr) {
747		npr = pr->ndpr_next;
748		if (pr->ndpr_ifp == ifp) {
749			/*
750			 * Previously, pr->ndpr_addr is removed as well,
751			 * but I strongly believe we don't have to do it.
752			 * nd6_purge() is only called from in6_ifdetach(),
753			 * which removes all the associated interface addresses
754			 * by itself.
755			 * (jinmei@kame.net 20010129)
756			 */
757			prelist_remove(pr);
758		}
759	}
760
761	/* cancel default outgoing interface setting */
762	if (nd6_defifindex == ifp->if_index)
763		nd6_setdefaultiface(0);
764
765	/* refresh default router list */
766	bzero(&drany, sizeof(drany));
767	defrouter_delreq(&drany, 0);
768	defrouter_select();
769
770	/*
771	 * Nuke neighbor cache entries for the ifp.
772	 * Note that rt->rt_ifp may not be the same as ifp,
773	 * due to KAME goto ours hack.  See RTM_RESOLVE case in
774	 * nd6_rtrequest(), and ip6_input().
775	 */
776	ln = llinfo_nd6.ln_next;
777	while (ln && ln != &llinfo_nd6) {
778		struct rtentry *rt;
779		struct sockaddr_dl *sdl;
780
781		nln = ln->ln_next;
782		rt = ln->ln_rt;
783		if (rt && rt->rt_gateway &&
784		    rt->rt_gateway->sa_family == AF_LINK) {
785			sdl = (struct sockaddr_dl *)rt->rt_gateway;
786			if (sdl->sdl_index == ifp->if_index)
787				nln = nd6_free(rt);
788		}
789		ln = nln;
790	}
791}
792
793struct rtentry *
794nd6_lookup(addr6, create, ifp)
795	struct in6_addr *addr6;
796	int create;
797	struct ifnet *ifp;
798{
799	struct rtentry *rt;
800	struct sockaddr_in6 sin6;
801
802	bzero(&sin6, sizeof(sin6));
803	sin6.sin6_len = sizeof(struct sockaddr_in6);
804	sin6.sin6_family = AF_INET6;
805	sin6.sin6_addr = *addr6;
806#ifdef SCOPEDROUTING
807	sin6.sin6_scope_id = in6_addr2scopeid(ifp, addr6);
808#endif
809	rt = rtalloc1((struct sockaddr *)&sin6, create, 0UL);
810	if (rt && (rt->rt_flags & RTF_LLINFO) == 0) {
811		/*
812		 * This is the case for the default route.
813		 * If we want to create a neighbor cache for the address, we
814		 * should free the route for the destination and allocate an
815		 * interface route.
816		 */
817		if (create) {
818			RTFREE(rt);
819			rt = 0;
820		}
821	}
822	if (!rt) {
823		if (create && ifp) {
824			int e;
825
826			/*
827			 * If no route is available and create is set,
828			 * we allocate a host route for the destination
829			 * and treat it like an interface route.
830			 * This hack is necessary for a neighbor which can't
831			 * be covered by our own prefix.
832			 */
833			struct ifaddr *ifa =
834				ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
835			if (ifa == NULL)
836				return(NULL);
837
838			/*
839			 * Create a new route. RTF_LLINFO is necessary
840			 * to create a Neighbor Cache entry for the
841			 * destination in nd6_rtrequest which will be
842			 * called in rtequest via ifa->ifa_rtrequest.
843			 */
844			if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6,
845					   ifa->ifa_addr,
846					   (struct sockaddr *)&all1_sa,
847					   (ifa->ifa_flags |
848					    RTF_HOST | RTF_LLINFO) &
849					   ~RTF_CLONING,
850					   &rt)) != 0)
851				log(LOG_ERR,
852				    "nd6_lookup: failed to add route for a "
853				    "neighbor(%s), errno=%d\n",
854				    ip6_sprintf(addr6), e);
855			if (rt == NULL)
856				return(NULL);
857			if (rt->rt_llinfo) {
858				struct llinfo_nd6 *ln =
859					(struct llinfo_nd6 *)rt->rt_llinfo;
860				ln->ln_state = ND6_LLINFO_NOSTATE;
861			}
862		} else
863			return(NULL);
864	}
865	rt->rt_refcnt--;
866	/*
867	 * Validation for the entry.
868	 * XXX: we can't use rt->rt_ifp to check for the interface, since
869	 *      it might be the loopback interface if the entry is for our
870	 *      own address on a non-loopback interface. Instead, we should
871	 *      use rt->rt_ifa->ifa_ifp, which would specify the REAL interface.
872	 */
873	if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
874	    rt->rt_gateway->sa_family != AF_LINK ||
875	    (ifp && rt->rt_ifa->ifa_ifp != ifp)) {
876		if (create) {
877			log(LOG_DEBUG, "nd6_lookup: failed to lookup %s (if = %s)\n",
878			    ip6_sprintf(addr6), ifp ? if_name(ifp) : "unspec");
879			/* xxx more logs... kazu */
880		}
881		return(0);
882	}
883	return(rt);
884}
885
886/*
887 * Detect if a given IPv6 address identifies a neighbor on a given link.
888 * XXX: should take care of the destination of a p2p link?
889 */
890int
891nd6_is_addr_neighbor(addr, ifp)
892	struct sockaddr_in6 *addr;
893	struct ifnet *ifp;
894{
895	struct ifaddr *ifa;
896	int i;
897
898#define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr)
899#define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr)
900
901	/*
902	 * A link-local address is always a neighbor.
903	 * XXX: we should use the sin6_scope_id field rather than the embedded
904	 * interface index.
905	 */
906	if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) &&
907	    ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index)
908		return(1);
909
910	/*
911	 * If the address matches one of our addresses,
912	 * it should be a neighbor.
913	 */
914	for (ifa = ifp->if_addrlist.tqh_first;
915	     ifa;
916	     ifa = ifa->ifa_list.tqe_next)
917	{
918		if (ifa->ifa_addr->sa_family != AF_INET6)
919			next: continue;
920
921		for (i = 0; i < 4; i++) {
922			if ((IFADDR6(ifa).s6_addr32[i] ^
923			     addr->sin6_addr.s6_addr32[i]) &
924			    IFMASK6(ifa).s6_addr32[i])
925				goto next;
926		}
927		return(1);
928	}
929
930	/*
931	 * Even if the address matches none of our addresses, it might be
932	 * in the neighbor cache.
933	 */
934	if (nd6_lookup(&addr->sin6_addr, 0, ifp))
935		return(1);
936
937	return(0);
938#undef IFADDR6
939#undef IFMASK6
940}
941
942/*
943 * Free an nd6 llinfo entry.
944 */
945struct llinfo_nd6 *
946nd6_free(rt)
947	struct rtentry *rt;
948{
949	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
950	struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
951	struct nd_defrouter *dr;
952
953	/*
954	 * we used to have pfctlinput(PRC_HOSTDEAD) here.
955	 * even though it is not harmful, it was not really necessary.
956	 */
957
958	if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
959		int s;
960		s = splnet();
961		dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
962				      rt->rt_ifp);
963
964		if (ln->ln_router || dr) {
965			/*
966			 * rt6_flush must be called whether or not the neighbor
967			 * is in the Default Router List.
968			 * See a corresponding comment in nd6_na_input().
969			 */
970			rt6_flush(&in6, rt->rt_ifp);
971		}
972
973		if (dr) {
974			/*
975			 * Unreachablity of a router might affect the default
976			 * router selection and on-link detection of advertised
977			 * prefixes.
978			 */
979
980			/*
981			 * Temporarily fake the state to choose a new default
982			 * router and to perform on-link determination of
983			 * prefixes coreectly.
984			 * Below the state will be set correctly,
985			 * or the entry itself will be deleted.
986			 */
987			ln->ln_state = ND6_LLINFO_INCOMPLETE;
988
989			/*
990			 * Since defrouter_select() does not affect the
991			 * on-link determination and MIP6 needs the check
992			 * before the default router selection, we perform
993			 * the check now.
994			 */
995			pfxlist_onlink_check();
996
997			if (dr == TAILQ_FIRST(&nd_defrouter)) {
998				/*
999				 * It is used as the current default router,
1000				 * so we have to move it to the end of the
1001				 * list and choose a new one.
1002				 * XXX: it is not very efficient if this is
1003				 *      the only router.
1004				 */
1005				TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
1006				TAILQ_INSERT_TAIL(&nd_defrouter, dr, dr_entry);
1007
1008				defrouter_select();
1009			}
1010		}
1011		splx(s);
1012	}
1013
1014	/*
1015	 * Before deleting the entry, remember the next entry as the
1016	 * return value.  We need this because pfxlist_onlink_check() above
1017	 * might have freed other entries (particularly the old next entry) as
1018	 * a side effect (XXX).
1019	 */
1020	next = ln->ln_next;
1021
1022	/*
1023	 * Detach the route from the routing tree and the list of neighbor
1024	 * caches, and disable the route entry not to be used in already
1025	 * cached routes.
1026	 */
1027	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0,
1028		  rt_mask(rt), 0, (struct rtentry **)0);
1029
1030	return(next);
1031}
1032
1033/*
1034 * Upper-layer reachability hint for Neighbor Unreachability Detection.
1035 *
1036 * XXX cost-effective metods?
1037 */
1038void
1039nd6_nud_hint(rt, dst6, force)
1040	struct rtentry *rt;
1041	struct in6_addr *dst6;
1042	int force;
1043{
1044	struct llinfo_nd6 *ln;
1045
1046	/*
1047	 * If the caller specified "rt", use that.  Otherwise, resolve the
1048	 * routing table by supplied "dst6".
1049	 */
1050	if (!rt) {
1051		if (!dst6)
1052			return;
1053		if (!(rt = nd6_lookup(dst6, 0, NULL)))
1054			return;
1055	}
1056
1057	if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
1058	    (rt->rt_flags & RTF_LLINFO) == 0 ||
1059	    !rt->rt_llinfo || !rt->rt_gateway ||
1060	    rt->rt_gateway->sa_family != AF_LINK) {
1061		/* This is not a host route. */
1062		return;
1063	}
1064
1065	ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1066	if (ln->ln_state < ND6_LLINFO_REACHABLE)
1067		return;
1068
1069	/*
1070	 * if we get upper-layer reachability confirmation many times,
1071	 * it is possible we have false information.
1072	 */
1073	if (!force) {
1074		ln->ln_byhint++;
1075		if (ln->ln_byhint > nd6_maxnudhint)
1076			return;
1077	}
1078
1079	ln->ln_state = ND6_LLINFO_REACHABLE;
1080	if (ln->ln_expire)
1081		ln->ln_expire = time_second +
1082			nd_ifinfo[rt->rt_ifp->if_index].reachable;
1083}
1084
1085void
1086nd6_rtrequest(req, rt, sa)
1087	int	req;
1088	struct rtentry *rt;
1089	struct sockaddr *sa; /* xxx unused */
1090{
1091	struct sockaddr *gate = rt->rt_gateway;
1092	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1093	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1094	struct ifnet *ifp = rt->rt_ifp;
1095	struct ifaddr *ifa;
1096
1097	if (rt->rt_flags & RTF_GATEWAY)
1098		return;
1099
1100	if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
1101		/*
1102		 * This is probably an interface direct route for a link
1103		 * which does not need neighbor caches (e.g. fe80::%lo0/64).
1104		 * We do not need special treatment below for such a route.
1105		 * Moreover, the RTF_LLINFO flag which would be set below
1106		 * would annoy the ndp(8) command.
1107		 */
1108		return;
1109	}
1110
1111	switch (req) {
1112	case RTM_ADD:
1113		/*
1114		 * There is no backward compatibility :)
1115		 *
1116		 * if ((rt->rt_flags & RTF_HOST) == 0 &&
1117		 *     SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
1118		 *	   rt->rt_flags |= RTF_CLONING;
1119		 */
1120		if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) {
1121			/*
1122			 * Case 1: This route should come from
1123			 * a route to interface. RTF_LLINFO flag is set
1124			 * for a host route whose destination should be
1125			 * treated as on-link.
1126			 */
1127			rt_setgate(rt, rt_key(rt),
1128				   (struct sockaddr *)&null_sdl);
1129			gate = rt->rt_gateway;
1130			SDL(gate)->sdl_type = ifp->if_type;
1131			SDL(gate)->sdl_index = ifp->if_index;
1132			if (ln)
1133				ln->ln_expire = time_second;
1134#if 1
1135			if (ln && ln->ln_expire == 0) {
1136				/* kludge for desktops */
1137#if 0
1138				printf("nd6_request: time.tv_sec is zero; "
1139				       "treat it as 1\n");
1140#endif
1141				ln->ln_expire = 1;
1142			}
1143#endif
1144			if (rt->rt_flags & RTF_CLONING)
1145				break;
1146		}
1147		/*
1148		 * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
1149		 * We don't do that here since llinfo is not ready yet.
1150		 *
1151		 * There are also couple of other things to be discussed:
1152		 * - unsolicited NA code needs improvement beforehand
1153		 * - RFC2461 says we MAY send multicast unsolicited NA
1154		 *   (7.2.6 paragraph 4), however, it also says that we
1155		 *   SHOULD provide a mechanism to prevent multicast NA storm.
1156		 *   we don't have anything like it right now.
1157		 *   note that the mechanism needs a mutual agreement
1158		 *   between proxies, which means that we need to implement
1159		 *   a new protocol, or a new kludge.
1160		 * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
1161		 *   we need to check ip6forwarding before sending it.
1162		 *   (or should we allow proxy ND configuration only for
1163		 *   routers?  there's no mention about proxy ND from hosts)
1164		 */
1165#if 0
1166		/* XXX it does not work */
1167		if (rt->rt_flags & RTF_ANNOUNCE)
1168			nd6_na_output(ifp,
1169			      &SIN6(rt_key(rt))->sin6_addr,
1170			      &SIN6(rt_key(rt))->sin6_addr,
1171			      ip6_forwarding ? ND_NA_FLAG_ROUTER : 0,
1172			      1, NULL);
1173#endif
1174		/* FALLTHROUGH */
1175	case RTM_RESOLVE:
1176		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
1177			/*
1178			 * Address resolution isn't necessary for a point to
1179			 * point link, so we can skip this test for a p2p link.
1180			 */
1181			if (gate->sa_family != AF_LINK ||
1182			    gate->sa_len < sizeof(null_sdl)) {
1183				log(LOG_DEBUG,
1184				    "nd6_rtrequest: bad gateway value: %s\n",
1185				    if_name(ifp));
1186				break;
1187			}
1188			SDL(gate)->sdl_type = ifp->if_type;
1189			SDL(gate)->sdl_index = ifp->if_index;
1190		}
1191		if (ln != NULL)
1192			break;	/* This happens on a route change */
1193		/*
1194		 * Case 2: This route may come from cloning, or a manual route
1195		 * add with a LL address.
1196		 */
1197		R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
1198		rt->rt_llinfo = (caddr_t)ln;
1199		if (!ln) {
1200			log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
1201			break;
1202		}
1203		nd6_inuse++;
1204		nd6_allocated++;
1205		Bzero(ln, sizeof(*ln));
1206		ln->ln_rt = rt;
1207		/* this is required for "ndp" command. - shin */
1208		if (req == RTM_ADD) {
1209		        /*
1210			 * gate should have some valid AF_LINK entry,
1211			 * and ln->ln_expire should have some lifetime
1212			 * which is specified by ndp command.
1213			 */
1214			ln->ln_state = ND6_LLINFO_REACHABLE;
1215			ln->ln_byhint = 0;
1216		} else {
1217		        /*
1218			 * When req == RTM_RESOLVE, rt is created and
1219			 * initialized in rtrequest(), so rt_expire is 0.
1220			 */
1221			ln->ln_state = ND6_LLINFO_NOSTATE;
1222			ln->ln_expire = time_second;
1223		}
1224		rt->rt_flags |= RTF_LLINFO;
1225		ln->ln_next = llinfo_nd6.ln_next;
1226		llinfo_nd6.ln_next = ln;
1227		ln->ln_prev = &llinfo_nd6;
1228		ln->ln_next->ln_prev = ln;
1229
1230		/*
1231		 * check if rt_key(rt) is one of my address assigned
1232		 * to the interface.
1233		 */
1234		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
1235					  &SIN6(rt_key(rt))->sin6_addr);
1236		if (ifa) {
1237			caddr_t macp = nd6_ifptomac(ifp);
1238			ln->ln_expire = 0;
1239			ln->ln_state = ND6_LLINFO_REACHABLE;
1240			ln->ln_byhint = 0;
1241			if (macp) {
1242				Bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
1243				SDL(gate)->sdl_alen = ifp->if_addrlen;
1244			}
1245			if (nd6_useloopback) {
1246				rt->rt_ifp = &loif[0];	/*XXX*/
1247				/*
1248				 * Make sure rt_ifa be equal to the ifaddr
1249				 * corresponding to the address.
1250				 * We need this because when we refer
1251				 * rt_ifa->ia6_flags in ip6_input, we assume
1252				 * that the rt_ifa points to the address instead
1253				 * of the loopback address.
1254				 */
1255				if (ifa != rt->rt_ifa) {
1256					IFAFREE(rt->rt_ifa);
1257					IFAREF(ifa);
1258					rt->rt_ifa = ifa;
1259				}
1260			}
1261		} else if (rt->rt_flags & RTF_ANNOUNCE) {
1262			ln->ln_expire = 0;
1263			ln->ln_state = ND6_LLINFO_REACHABLE;
1264			ln->ln_byhint = 0;
1265
1266			/* join solicited node multicast for proxy ND */
1267			if (ifp->if_flags & IFF_MULTICAST) {
1268				struct in6_addr llsol;
1269				int error;
1270
1271				llsol = SIN6(rt_key(rt))->sin6_addr;
1272				llsol.s6_addr16[0] = htons(0xff02);
1273				llsol.s6_addr16[1] = htons(ifp->if_index);
1274				llsol.s6_addr32[1] = 0;
1275				llsol.s6_addr32[2] = htonl(1);
1276				llsol.s6_addr8[12] = 0xff;
1277
1278				if (!in6_addmulti(&llsol, ifp, &error)) {
1279					nd6log((LOG_ERR, "%s: failed to join "
1280					    "%s (errno=%d)\n", if_name(ifp),
1281					    ip6_sprintf(&llsol), error));
1282				}
1283			}
1284		}
1285		break;
1286
1287	case RTM_DELETE:
1288		if (!ln)
1289			break;
1290		/* leave from solicited node multicast for proxy ND */
1291		if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
1292		    (ifp->if_flags & IFF_MULTICAST) != 0) {
1293			struct in6_addr llsol;
1294			struct in6_multi *in6m;
1295
1296			llsol = SIN6(rt_key(rt))->sin6_addr;
1297			llsol.s6_addr16[0] = htons(0xff02);
1298			llsol.s6_addr16[1] = htons(ifp->if_index);
1299			llsol.s6_addr32[1] = 0;
1300			llsol.s6_addr32[2] = htonl(1);
1301			llsol.s6_addr8[12] = 0xff;
1302
1303			IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1304			if (in6m)
1305				in6_delmulti(in6m);
1306		}
1307		nd6_inuse--;
1308		ln->ln_next->ln_prev = ln->ln_prev;
1309		ln->ln_prev->ln_next = ln->ln_next;
1310		ln->ln_prev = NULL;
1311		rt->rt_llinfo = 0;
1312		rt->rt_flags &= ~RTF_LLINFO;
1313		if (ln->ln_hold)
1314			m_freem(ln->ln_hold);
1315		Free((caddr_t)ln);
1316	}
1317}
1318
1319int
1320nd6_ioctl(cmd, data, ifp)
1321	u_long cmd;
1322	caddr_t	data;
1323	struct ifnet *ifp;
1324{
1325	struct in6_drlist *drl = (struct in6_drlist *)data;
1326	struct in6_prlist *prl = (struct in6_prlist *)data;
1327	struct in6_ndireq *ndi = (struct in6_ndireq *)data;
1328	struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
1329	struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
1330	struct nd_defrouter *dr, any;
1331	struct nd_prefix *pr;
1332	struct rtentry *rt;
1333	int i = 0, error = 0;
1334	int s;
1335
1336	switch (cmd) {
1337	case SIOCGDRLST_IN6:
1338		/*
1339		 * obsolete API, use sysctl under net.inet6.icmp6
1340		 */
1341		bzero(drl, sizeof(*drl));
1342		s = splnet();
1343		dr = TAILQ_FIRST(&nd_defrouter);
1344		while (dr && i < DRLSTSIZ) {
1345			drl->defrouter[i].rtaddr = dr->rtaddr;
1346			if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) {
1347				/* XXX: need to this hack for KAME stack */
1348				drl->defrouter[i].rtaddr.s6_addr16[1] = 0;
1349			} else
1350				log(LOG_ERR,
1351				    "default router list contains a "
1352				    "non-linklocal address(%s)\n",
1353				    ip6_sprintf(&drl->defrouter[i].rtaddr));
1354
1355			drl->defrouter[i].flags = dr->flags;
1356			drl->defrouter[i].rtlifetime = dr->rtlifetime;
1357			drl->defrouter[i].expire = dr->expire;
1358			drl->defrouter[i].if_index = dr->ifp->if_index;
1359			i++;
1360			dr = TAILQ_NEXT(dr, dr_entry);
1361		}
1362		splx(s);
1363		break;
1364	case SIOCGPRLST_IN6:
1365		/*
1366		 * obsolete API, use sysctl under net.inet6.icmp6
1367		 */
1368		/*
1369		 * XXX meaning of fields, especialy "raflags", is very
1370		 * differnet between RA prefix list and RR/static prefix list.
1371		 * how about separating ioctls into two?
1372		 */
1373		bzero(prl, sizeof(*prl));
1374		s = splnet();
1375		pr = nd_prefix.lh_first;
1376		while (pr && i < PRLSTSIZ) {
1377			struct nd_pfxrouter *pfr;
1378			int j;
1379
1380			(void)in6_embedscope(&prl->prefix[i].prefix,
1381			    &pr->ndpr_prefix, NULL, NULL);
1382			prl->prefix[i].raflags = pr->ndpr_raf;
1383			prl->prefix[i].prefixlen = pr->ndpr_plen;
1384			prl->prefix[i].vltime = pr->ndpr_vltime;
1385			prl->prefix[i].pltime = pr->ndpr_pltime;
1386			prl->prefix[i].if_index = pr->ndpr_ifp->if_index;
1387			prl->prefix[i].expire = pr->ndpr_expire;
1388
1389			pfr = pr->ndpr_advrtrs.lh_first;
1390			j = 0;
1391			while (pfr) {
1392				if (j < DRLSTSIZ) {
1393#define RTRADDR prl->prefix[i].advrtr[j]
1394					RTRADDR = pfr->router->rtaddr;
1395					if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) {
1396						/* XXX: hack for KAME */
1397						RTRADDR.s6_addr16[1] = 0;
1398					} else
1399						log(LOG_ERR,
1400						    "a router(%s) advertises "
1401						    "a prefix with "
1402						    "non-link local address\n",
1403						    ip6_sprintf(&RTRADDR));
1404#undef RTRADDR
1405				}
1406				j++;
1407				pfr = pfr->pfr_next;
1408			}
1409			prl->prefix[i].advrtrs = j;
1410			prl->prefix[i].origin = PR_ORIG_RA;
1411
1412			i++;
1413			pr = pr->ndpr_next;
1414		}
1415	      {
1416		struct rr_prefix *rpp;
1417
1418		for (rpp = LIST_FIRST(&rr_prefix); rpp;
1419		     rpp = LIST_NEXT(rpp, rp_entry)) {
1420			if (i >= PRLSTSIZ)
1421				break;
1422			(void)in6_embedscope(&prl->prefix[i].prefix,
1423			    &pr->ndpr_prefix, NULL, NULL);
1424			prl->prefix[i].raflags = rpp->rp_raf;
1425			prl->prefix[i].prefixlen = rpp->rp_plen;
1426			prl->prefix[i].vltime = rpp->rp_vltime;
1427			prl->prefix[i].pltime = rpp->rp_pltime;
1428			prl->prefix[i].if_index = rpp->rp_ifp->if_index;
1429			prl->prefix[i].expire = rpp->rp_expire;
1430			prl->prefix[i].advrtrs = 0;
1431			prl->prefix[i].origin = rpp->rp_origin;
1432			i++;
1433		}
1434	      }
1435		splx(s);
1436
1437		break;
1438	case OSIOCGIFINFO_IN6:
1439		if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1440			error = EINVAL;
1441			break;
1442		}
1443		ndi->ndi.linkmtu = nd_ifinfo[ifp->if_index].linkmtu;
1444		ndi->ndi.maxmtu = nd_ifinfo[ifp->if_index].maxmtu;
1445		ndi->ndi.basereachable =
1446		    nd_ifinfo[ifp->if_index].basereachable;
1447		ndi->ndi.reachable = nd_ifinfo[ifp->if_index].reachable;
1448		ndi->ndi.retrans = nd_ifinfo[ifp->if_index].retrans;
1449		ndi->ndi.flags = nd_ifinfo[ifp->if_index].flags;
1450		ndi->ndi.recalctm = nd_ifinfo[ifp->if_index].recalctm;
1451		ndi->ndi.chlim = nd_ifinfo[ifp->if_index].chlim;
1452		ndi->ndi.receivedra = nd_ifinfo[ifp->if_index].receivedra;
1453		break;
1454	case SIOCGIFINFO_IN6:
1455		if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1456			error = EINVAL;
1457			break;
1458		}
1459		ndi->ndi = nd_ifinfo[ifp->if_index];
1460		break;
1461	case SIOCSIFINFO_FLAGS:
1462		/* XXX: almost all other fields of ndi->ndi is unused */
1463		if (!nd_ifinfo || i >= nd_ifinfo_indexlim) {
1464			error = EINVAL;
1465			break;
1466		}
1467		nd_ifinfo[ifp->if_index].flags = ndi->ndi.flags;
1468		break;
1469	case SIOCSNDFLUSH_IN6:	/* XXX: the ioctl name is confusing... */
1470		/* flush default router list */
1471		/*
1472		 * xxx sumikawa: should not delete route if default
1473		 * route equals to the top of default router list
1474		 */
1475		bzero(&any, sizeof(any));
1476		defrouter_delreq(&any, 0);
1477		defrouter_select();
1478		/* xxx sumikawa: flush prefix list */
1479		break;
1480	case SIOCSPFXFLUSH_IN6:
1481	    {
1482		/* flush all the prefix advertised by routers */
1483		struct nd_prefix *pr, *next;
1484
1485		s = splnet();
1486		for (pr = nd_prefix.lh_first; pr; pr = next) {
1487			struct in6_ifaddr *ia, *ia_next;
1488
1489			next = pr->ndpr_next;
1490
1491			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1492				continue; /* XXX */
1493
1494			/* do we really have to remove addresses as well? */
1495			for (ia = in6_ifaddr; ia; ia = ia_next) {
1496				/* ia might be removed. keep the next ptr. */
1497				ia_next = ia->ia_next;
1498
1499				if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1500					continue;
1501
1502				if (ia->ia6_ndpr == pr)
1503					in6_purgeaddr(&ia->ia_ifa);
1504			}
1505			prelist_remove(pr);
1506		}
1507		splx(s);
1508		break;
1509	    }
1510	case SIOCSRTRFLUSH_IN6:
1511	    {
1512		/* flush all the default routers */
1513		struct nd_defrouter *dr, *next;
1514
1515		s = splnet();
1516		if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
1517			/*
1518			 * The first entry of the list may be stored in
1519			 * the routing table, so we'll delete it later.
1520			 */
1521			for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = next) {
1522				next = TAILQ_NEXT(dr, dr_entry);
1523				defrtrlist_del(dr);
1524			}
1525			defrtrlist_del(TAILQ_FIRST(&nd_defrouter));
1526		}
1527		splx(s);
1528		break;
1529	    }
1530	case SIOCGNBRINFO_IN6:
1531	    {
1532		struct llinfo_nd6 *ln;
1533		struct in6_addr nb_addr = nbi->addr; /* make local for safety */
1534
1535		/*
1536		 * XXX: KAME specific hack for scoped addresses
1537		 *      XXXX: for other scopes than link-local?
1538		 */
1539		if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) ||
1540		    IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) {
1541			u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2];
1542
1543			if (*idp == 0)
1544				*idp = htons(ifp->if_index);
1545		}
1546
1547		s = splnet();
1548		if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) {
1549			error = EINVAL;
1550			splx(s);
1551			break;
1552		}
1553		ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1554		nbi->state = ln->ln_state;
1555		nbi->asked = ln->ln_asked;
1556		nbi->isrouter = ln->ln_router;
1557		nbi->expire = ln->ln_expire;
1558		splx(s);
1559
1560		break;
1561	    }
1562	case SIOCGDEFIFACE_IN6:	/* XXX: should be implemented as a sysctl? */
1563		ndif->ifindex = nd6_defifindex;
1564		break;
1565	case SIOCSDEFIFACE_IN6:	/* XXX: should be implemented as a sysctl? */
1566		return(nd6_setdefaultiface(ndif->ifindex));
1567		break;
1568	}
1569	return(error);
1570}
1571
1572/*
1573 * Create neighbor cache entry and cache link-layer address,
1574 * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
1575 */
1576struct rtentry *
1577nd6_cache_lladdr(ifp, from, lladdr, lladdrlen, type, code)
1578	struct ifnet *ifp;
1579	struct in6_addr *from;
1580	char *lladdr;
1581	int lladdrlen;
1582	int type;	/* ICMP6 type */
1583	int code;	/* type dependent information */
1584{
1585	struct rtentry *rt = NULL;
1586	struct llinfo_nd6 *ln = NULL;
1587	int is_newentry;
1588	struct sockaddr_dl *sdl = NULL;
1589	int do_update;
1590	int olladdr;
1591	int llchange;
1592	int newstate = 0;
1593
1594	if (!ifp)
1595		panic("ifp == NULL in nd6_cache_lladdr");
1596	if (!from)
1597		panic("from == NULL in nd6_cache_lladdr");
1598
1599	/* nothing must be updated for unspecified address */
1600	if (IN6_IS_ADDR_UNSPECIFIED(from))
1601		return NULL;
1602
1603	/*
1604	 * Validation about ifp->if_addrlen and lladdrlen must be done in
1605	 * the caller.
1606	 *
1607	 * XXX If the link does not have link-layer adderss, what should
1608	 * we do? (ifp->if_addrlen == 0)
1609	 * Spec says nothing in sections for RA, RS and NA.  There's small
1610	 * description on it in NS section (RFC 2461 7.2.3).
1611	 */
1612
1613	rt = nd6_lookup(from, 0, ifp);
1614	if (!rt) {
1615#if 0
1616		/* nothing must be done if there's no lladdr */
1617		if (!lladdr || !lladdrlen)
1618			return NULL;
1619#endif
1620
1621		rt = nd6_lookup(from, 1, ifp);
1622		is_newentry = 1;
1623	} else {
1624		/* do nothing if static ndp is set */
1625		if (rt->rt_flags & RTF_STATIC)
1626			return NULL;
1627		is_newentry = 0;
1628	}
1629
1630	if (!rt)
1631		return NULL;
1632	if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
1633fail:
1634		(void)nd6_free(rt);
1635		return NULL;
1636	}
1637	ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1638	if (!ln)
1639		goto fail;
1640	if (!rt->rt_gateway)
1641		goto fail;
1642	if (rt->rt_gateway->sa_family != AF_LINK)
1643		goto fail;
1644	sdl = SDL(rt->rt_gateway);
1645
1646	olladdr = (sdl->sdl_alen) ? 1 : 0;
1647	if (olladdr && lladdr) {
1648		if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
1649			llchange = 1;
1650		else
1651			llchange = 0;
1652	} else
1653		llchange = 0;
1654
1655	/*
1656	 * newentry olladdr  lladdr  llchange	(*=record)
1657	 *	0	n	n	--	(1)
1658	 *	0	y	n	--	(2)
1659	 *	0	n	y	--	(3) * STALE
1660	 *	0	y	y	n	(4) *
1661	 *	0	y	y	y	(5) * STALE
1662	 *	1	--	n	--	(6)   NOSTATE(= PASSIVE)
1663	 *	1	--	y	--	(7) * STALE
1664	 */
1665
1666	if (lladdr) {		/*(3-5) and (7)*/
1667		/*
1668		 * Record source link-layer address
1669		 * XXX is it dependent to ifp->if_type?
1670		 */
1671		sdl->sdl_alen = ifp->if_addrlen;
1672		bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1673	}
1674
1675	if (!is_newentry) {
1676		if ((!olladdr && lladdr)		/*(3)*/
1677		 || (olladdr && lladdr && llchange)) {	/*(5)*/
1678			do_update = 1;
1679			newstate = ND6_LLINFO_STALE;
1680		} else					/*(1-2,4)*/
1681			do_update = 0;
1682	} else {
1683		do_update = 1;
1684		if (!lladdr)				/*(6)*/
1685			newstate = ND6_LLINFO_NOSTATE;
1686		else					/*(7)*/
1687			newstate = ND6_LLINFO_STALE;
1688	}
1689
1690	if (do_update) {
1691		/*
1692		 * Update the state of the neighbor cache.
1693		 */
1694		ln->ln_state = newstate;
1695
1696		if (ln->ln_state == ND6_LLINFO_STALE) {
1697			/*
1698			 * XXX: since nd6_output() below will cause
1699			 * state tansition to DELAY and reset the timer,
1700			 * we must set the timer now, although it is actually
1701			 * meaningless.
1702			 */
1703			ln->ln_expire = time_second + nd6_gctimer;
1704
1705			if (ln->ln_hold) {
1706				/*
1707				 * we assume ifp is not a p2p here, so just
1708				 * set the 2nd argument as the 1st one.
1709				 */
1710				nd6_output(ifp, ifp, ln->ln_hold,
1711					   (struct sockaddr_in6 *)rt_key(rt),
1712					   rt);
1713				ln->ln_hold = NULL;
1714			}
1715		} else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1716			/* probe right away */
1717			ln->ln_expire = time_second;
1718		}
1719	}
1720
1721	/*
1722	 * ICMP6 type dependent behavior.
1723	 *
1724	 * NS: clear IsRouter if new entry
1725	 * RS: clear IsRouter
1726	 * RA: set IsRouter if there's lladdr
1727	 * redir: clear IsRouter if new entry
1728	 *
1729	 * RA case, (1):
1730	 * The spec says that we must set IsRouter in the following cases:
1731	 * - If lladdr exist, set IsRouter.  This means (1-5).
1732	 * - If it is old entry (!newentry), set IsRouter.  This means (7).
1733	 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
1734	 * A quetion arises for (1) case.  (1) case has no lladdr in the
1735	 * neighbor cache, this is similar to (6).
1736	 * This case is rare but we figured that we MUST NOT set IsRouter.
1737	 *
1738	 * newentry olladdr  lladdr  llchange	    NS  RS  RA	redir
1739	 *							D R
1740	 *	0	n	n	--	(1)	c   ?     s
1741	 *	0	y	n	--	(2)	c   s     s
1742	 *	0	n	y	--	(3)	c   s     s
1743	 *	0	y	y	n	(4)	c   s     s
1744	 *	0	y	y	y	(5)	c   s     s
1745	 *	1	--	n	--	(6) c	c 	c s
1746	 *	1	--	y	--	(7) c	c   s	c s
1747	 *
1748	 *					(c=clear s=set)
1749	 */
1750	switch (type & 0xff) {
1751	case ND_NEIGHBOR_SOLICIT:
1752		/*
1753		 * New entry must have is_router flag cleared.
1754		 */
1755		if (is_newentry)	/*(6-7)*/
1756			ln->ln_router = 0;
1757		break;
1758	case ND_REDIRECT:
1759		/*
1760		 * If the icmp is a redirect to a better router, always set the
1761		 * is_router flag. Otherwise, if the entry is newly created,
1762		 * clear the flag. [RFC 2461, sec 8.3]
1763		 */
1764		if (code == ND_REDIRECT_ROUTER)
1765			ln->ln_router = 1;
1766		else if (is_newentry) /*(6-7)*/
1767			ln->ln_router = 0;
1768		break;
1769	case ND_ROUTER_SOLICIT:
1770		/*
1771		 * is_router flag must always be cleared.
1772		 */
1773		ln->ln_router = 0;
1774		break;
1775	case ND_ROUTER_ADVERT:
1776		/*
1777		 * Mark an entry with lladdr as a router.
1778		 */
1779		if ((!is_newentry && (olladdr || lladdr))	/*(2-5)*/
1780		 || (is_newentry && lladdr)) {			/*(7)*/
1781			ln->ln_router = 1;
1782		}
1783		break;
1784	}
1785
1786	return rt;
1787}
1788
1789static void
1790nd6_slowtimo(ignored_arg)
1791    void *ignored_arg;
1792{
1793	int s = splnet();
1794	int i;
1795	struct nd_ifinfo *nd6if;
1796
1797	callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
1798	    nd6_slowtimo, NULL);
1799	for (i = 1; i < if_index + 1; i++) {
1800		if (!nd_ifinfo || i >= nd_ifinfo_indexlim)
1801			continue;
1802		nd6if = &nd_ifinfo[i];
1803		if (nd6if->basereachable && /* already initialized */
1804		    (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
1805			/*
1806			 * Since reachable time rarely changes by router
1807			 * advertisements, we SHOULD insure that a new random
1808			 * value gets recomputed at least once every few hours.
1809			 * (RFC 2461, 6.3.4)
1810			 */
1811			nd6if->recalctm = nd6_recalc_reachtm_interval;
1812			nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
1813		}
1814	}
1815	splx(s);
1816}
1817
1818#define senderr(e) { error = (e); goto bad;}
1819int
1820nd6_output(ifp, origifp, m0, dst, rt0)
1821	struct ifnet *ifp;
1822	struct ifnet *origifp;
1823	struct mbuf *m0;
1824	struct sockaddr_in6 *dst;
1825	struct rtentry *rt0;
1826{
1827	struct mbuf *m = m0;
1828	struct rtentry *rt = rt0;
1829	struct sockaddr_in6 *gw6 = NULL;
1830	struct llinfo_nd6 *ln = NULL;
1831	int error = 0;
1832
1833	if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
1834		goto sendpkt;
1835
1836	if (nd6_need_cache(ifp) == 0)
1837		goto sendpkt;
1838
1839	/*
1840	 * next hop determination. This routine is derived from ether_outpout.
1841	 */
1842	if (rt) {
1843		if ((rt->rt_flags & RTF_UP) == 0) {
1844			if ((rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL)) !=
1845				NULL)
1846			{
1847				rt->rt_refcnt--;
1848				if (rt->rt_ifp != ifp) {
1849					/* XXX: loop care? */
1850					return nd6_output(ifp, origifp, m0,
1851							  dst, rt);
1852				}
1853			} else
1854				senderr(EHOSTUNREACH);
1855		}
1856
1857		if (rt->rt_flags & RTF_GATEWAY) {
1858			gw6 = (struct sockaddr_in6 *)rt->rt_gateway;
1859
1860			/*
1861			 * We skip link-layer address resolution and NUD
1862			 * if the gateway is not a neighbor from ND point
1863			 * of view, regardless the value of the
1864			 * nd_ifinfo.flags.
1865			 * The second condition is a bit tricky: we skip
1866			 * if the gateway is our own address, which is
1867			 * sometimes used to install a route to a p2p link.
1868			 */
1869			if (!nd6_is_addr_neighbor(gw6, ifp) ||
1870			    in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) {
1871				/*
1872				 * We allow this kind of tricky route only
1873				 * when the outgoing interface is p2p.
1874				 * XXX: we may need a more generic rule here.
1875				 */
1876				if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1877					senderr(EHOSTUNREACH);
1878
1879				goto sendpkt;
1880			}
1881
1882			if (rt->rt_gwroute == 0)
1883				goto lookup;
1884			if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) {
1885				rtfree(rt); rt = rt0;
1886			lookup: rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1, 0UL);
1887				if ((rt = rt->rt_gwroute) == 0)
1888					senderr(EHOSTUNREACH);
1889			}
1890		}
1891	}
1892
1893	/*
1894	 * Address resolution or Neighbor Unreachability Detection
1895	 * for the next hop.
1896	 * At this point, the destination of the packet must be a unicast
1897	 * or an anycast address(i.e. not a multicast).
1898	 */
1899
1900	/* Look up the neighbor cache for the nexthop */
1901	if (rt && (rt->rt_flags & RTF_LLINFO) != 0)
1902		ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1903	else {
1904		/*
1905		 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
1906		 * the condition below is not very efficient. But we believe
1907		 * it is tolerable, because this should be a rare case.
1908		 */
1909		if (nd6_is_addr_neighbor(dst, ifp) &&
1910		    (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL)
1911			ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1912	}
1913	if (!ln || !rt) {
1914		if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
1915		    !(nd_ifinfo[ifp->if_index].flags & ND6_IFF_PERFORMNUD)) {
1916			log(LOG_DEBUG,
1917			    "nd6_output: can't allocate llinfo for %s "
1918			    "(ln=%p, rt=%p)\n",
1919			    ip6_sprintf(&dst->sin6_addr), ln, rt);
1920			senderr(EIO);	/* XXX: good error? */
1921		}
1922
1923		goto sendpkt;	/* send anyway */
1924	}
1925
1926	/* We don't have to do link-layer address resolution on a p2p link. */
1927	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1928	    ln->ln_state < ND6_LLINFO_REACHABLE) {
1929		ln->ln_state = ND6_LLINFO_STALE;
1930		ln->ln_expire = time_second + nd6_gctimer;
1931	}
1932
1933	/*
1934	 * The first time we send a packet to a neighbor whose entry is
1935	 * STALE, we have to change the state to DELAY and a sets a timer to
1936	 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
1937	 * neighbor unreachability detection on expiration.
1938	 * (RFC 2461 7.3.3)
1939	 */
1940	if (ln->ln_state == ND6_LLINFO_STALE) {
1941		ln->ln_asked = 0;
1942		ln->ln_state = ND6_LLINFO_DELAY;
1943		ln->ln_expire = time_second + nd6_delay;
1944	}
1945
1946	/*
1947	 * If the neighbor cache entry has a state other than INCOMPLETE
1948	 * (i.e. its link-layer address is already reloved), just
1949	 * send the packet.
1950	 */
1951	if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
1952		goto sendpkt;
1953
1954	/*
1955	 * There is a neighbor cache entry, but no ethernet address
1956	 * response yet. Replace the held mbuf (if any) with this
1957	 * latest one.
1958	 *
1959	 * XXX Does the code conform to rate-limiting rule?
1960	 * (RFC 2461 7.2.2)
1961	 */
1962	if (ln->ln_state == ND6_LLINFO_NOSTATE)
1963		ln->ln_state = ND6_LLINFO_INCOMPLETE;
1964	if (ln->ln_hold)
1965		m_freem(ln->ln_hold);
1966	ln->ln_hold = m;
1967	if (ln->ln_expire) {
1968		if (ln->ln_asked < nd6_mmaxtries &&
1969		    ln->ln_expire < time_second) {
1970			ln->ln_asked++;
1971			ln->ln_expire = time_second +
1972				nd_ifinfo[ifp->if_index].retrans / 1000;
1973			nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
1974		}
1975	}
1976	return(0);
1977
1978  sendpkt:
1979
1980	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
1981		return((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
1982					 rt));
1983	}
1984	return((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt));
1985
1986  bad:
1987	if (m)
1988		m_freem(m);
1989	return (error);
1990}
1991#undef senderr
1992
1993int
1994nd6_need_cache(ifp)
1995	struct ifnet *ifp;
1996{
1997	/*
1998	 * XXX: we currently do not make neighbor cache on any interface
1999	 * other than ARCnet, Ethernet, FDDI and GIF.
2000	 *
2001	 * RFC2893 says:
2002	 * - unidirectional tunnels needs no ND
2003	 */
2004	switch (ifp->if_type) {
2005	case IFT_ARCNET:
2006	case IFT_ETHER:
2007	case IFT_FDDI:
2008	case IFT_IEEE1394:
2009#ifdef IFT_L2VLAN
2010	case IFT_L2VLAN:
2011#endif
2012#ifdef IFT_IEEE80211
2013	case IFT_IEEE80211:
2014#endif
2015	case IFT_GIF:		/* XXX need more cases? */
2016		return(1);
2017	default:
2018		return(0);
2019	}
2020}
2021
2022int
2023nd6_storelladdr(ifp, rt, m, dst, desten)
2024	struct ifnet *ifp;
2025	struct rtentry *rt;
2026	struct mbuf *m;
2027	struct sockaddr *dst;
2028	u_char *desten;
2029{
2030	int i;
2031	struct sockaddr_dl *sdl;
2032
2033	if (m->m_flags & M_MCAST) {
2034		switch (ifp->if_type) {
2035		case IFT_ETHER:
2036		case IFT_FDDI:
2037#ifdef IFT_L2VLAN
2038	case IFT_L2VLAN:
2039#endif
2040#ifdef IFT_IEEE80211
2041		case IFT_IEEE80211:
2042#endif
2043			ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
2044						 desten);
2045			return(1);
2046		case IFT_IEEE1394:
2047			for (i = 0; i < ifp->if_addrlen; i++)
2048				desten[i] = ~0;
2049			return(1);
2050		case IFT_ARCNET:
2051			*desten = 0;
2052			return(1);
2053		default:
2054			m_freem(m);
2055			return(0);
2056		}
2057	}
2058
2059	if (rt == NULL) {
2060		/* this could happen, if we could not allocate memory */
2061		m_freem(m);
2062		return(0);
2063	}
2064	if (rt->rt_gateway->sa_family != AF_LINK) {
2065		printf("nd6_storelladdr: something odd happens\n");
2066		m_freem(m);
2067		return(0);
2068	}
2069	sdl = SDL(rt->rt_gateway);
2070	if (sdl->sdl_alen == 0) {
2071		/* this should be impossible, but we bark here for debugging */
2072		printf("nd6_storelladdr: sdl_alen == 0\n");
2073		m_freem(m);
2074		return(0);
2075	}
2076
2077	bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
2078	return(1);
2079}
2080
2081static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
2082static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
2083#ifdef SYSCTL_DECL
2084SYSCTL_DECL(_net_inet6_icmp6);
2085#endif
2086SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2087	CTLFLAG_RD, nd6_sysctl_drlist, "");
2088SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
2089	CTLFLAG_RD, nd6_sysctl_prlist, "");
2090
2091static int
2092nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2093{
2094	int error;
2095	char buf[1024];
2096	struct in6_defrouter *d, *de;
2097	struct nd_defrouter *dr;
2098
2099	if (req->newptr)
2100		return EPERM;
2101	error = 0;
2102
2103	for (dr = TAILQ_FIRST(&nd_defrouter);
2104	     dr;
2105	     dr = TAILQ_NEXT(dr, dr_entry)) {
2106		d = (struct in6_defrouter *)buf;
2107		de = (struct in6_defrouter *)(buf + sizeof(buf));
2108
2109		if (d + 1 <= de) {
2110			bzero(d, sizeof(*d));
2111			d->rtaddr.sin6_family = AF_INET6;
2112			d->rtaddr.sin6_len = sizeof(d->rtaddr);
2113			if (in6_recoverscope(&d->rtaddr, &dr->rtaddr,
2114			    dr->ifp) != 0)
2115				log(LOG_ERR,
2116				    "scope error in "
2117				    "default router list (%s)\n",
2118				    ip6_sprintf(&dr->rtaddr));
2119			d->flags = dr->flags;
2120			d->rtlifetime = dr->rtlifetime;
2121			d->expire = dr->expire;
2122			d->if_index = dr->ifp->if_index;
2123		} else
2124			panic("buffer too short");
2125
2126		error = SYSCTL_OUT(req, buf, sizeof(*d));
2127		if (error)
2128			break;
2129	}
2130	return error;
2131}
2132
2133static int
2134nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
2135{
2136	int error;
2137	char buf[1024];
2138	struct in6_prefix *p, *pe;
2139	struct nd_prefix *pr;
2140
2141	if (req->newptr)
2142		return EPERM;
2143	error = 0;
2144
2145	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
2146		u_short advrtrs;
2147		size_t advance;
2148		struct sockaddr_in6 *sin6, *s6;
2149		struct nd_pfxrouter *pfr;
2150
2151		p = (struct in6_prefix *)buf;
2152		pe = (struct in6_prefix *)(buf + sizeof(buf));
2153
2154		if (p + 1 <= pe) {
2155			bzero(p, sizeof(*p));
2156			sin6 = (struct sockaddr_in6 *)(p + 1);
2157
2158			p->prefix = pr->ndpr_prefix;
2159			if (in6_recoverscope(&p->prefix,
2160			    &p->prefix.sin6_addr, pr->ndpr_ifp) != 0)
2161				log(LOG_ERR,
2162				    "scope error in prefix list (%s)\n",
2163				    ip6_sprintf(&p->prefix.sin6_addr));
2164			p->raflags = pr->ndpr_raf;
2165			p->prefixlen = pr->ndpr_plen;
2166			p->vltime = pr->ndpr_vltime;
2167			p->pltime = pr->ndpr_pltime;
2168			p->if_index = pr->ndpr_ifp->if_index;
2169			p->expire = pr->ndpr_expire;
2170			p->refcnt = pr->ndpr_refcnt;
2171			p->flags = pr->ndpr_stateflags;
2172			p->origin = PR_ORIG_RA;
2173			advrtrs = 0;
2174			for (pfr = pr->ndpr_advrtrs.lh_first;
2175			     pfr;
2176			     pfr = pfr->pfr_next) {
2177				if ((void *)&sin6[advrtrs + 1] >
2178				    (void *)pe) {
2179					advrtrs++;
2180					continue;
2181				}
2182				s6 = &sin6[advrtrs];
2183				bzero(s6, sizeof(*s6));
2184				s6->sin6_family = AF_INET6;
2185				s6->sin6_len = sizeof(*sin6);
2186				if (in6_recoverscope(s6,
2187				    &pfr->router->rtaddr,
2188				    pfr->router->ifp) != 0)
2189					log(LOG_ERR,
2190					    "scope error in "
2191					    "prefix list (%s)\n",
2192					    ip6_sprintf(&pfr->router->rtaddr));
2193				advrtrs++;
2194			}
2195			p->advrtrs = advrtrs;
2196		} else
2197			panic("buffer too short");
2198
2199		advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
2200		error = SYSCTL_OUT(req, buf, advance);
2201		if (error)
2202			break;
2203	}
2204	return error;
2205}
2206