xref: /freebsd/sys/netinet6/nd6_rtr.c (revision 4f29da19bd44f0e99f021510460a81bf754c21d2)
1 /*	$FreeBSD$	*/
2 /*	$KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $	*/
3 
4 /*-
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include "opt_inet.h"
34 #include "opt_inet6.h"
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/malloc.h>
39 #include <sys/mbuf.h>
40 #include <sys/socket.h>
41 #include <sys/sockio.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/errno.h>
45 #include <sys/syslog.h>
46 #include <sys/queue.h>
47 
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/if_dl.h>
51 #include <net/route.h>
52 #include <net/radix.h>
53 
54 #include <netinet/in.h>
55 #include <netinet6/in6_var.h>
56 #include <netinet6/in6_ifattach.h>
57 #include <netinet/ip6.h>
58 #include <netinet6/ip6_var.h>
59 #include <netinet6/nd6.h>
60 #include <netinet/icmp6.h>
61 #include <netinet6/scope6_var.h>
62 
63 #include <net/net_osdep.h>
64 
65 #define SDL(s)	((struct sockaddr_dl *)s)
66 
67 static int rtpref __P((struct nd_defrouter *));
68 static struct nd_defrouter *defrtrlist_update __P((struct nd_defrouter *));
69 static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
70     struct mbuf *, int));
71 static struct in6_ifaddr *in6_ifadd __P((struct nd_prefixctl *,	int));
72 static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
73 	struct nd_defrouter *));
74 static void pfxrtr_add __P((struct nd_prefix *, struct nd_defrouter *));
75 static void pfxrtr_del __P((struct nd_pfxrouter *));
76 static struct nd_pfxrouter *find_pfxlist_reachable_router
77 	__P((struct nd_prefix *));
78 static void defrouter_delreq __P((struct nd_defrouter *));
79 static void nd6_rtmsg __P((int, struct rtentry *));
80 
81 static int in6_init_prefix_ltimes __P((struct nd_prefix *));
82 static void in6_init_address_ltimes __P((struct nd_prefix *,
83 	struct in6_addrlifetime *));
84 
85 static int rt6_deleteroute __P((struct radix_node *, void *));
86 
87 extern int nd6_recalc_reachtm_interval;
88 
89 static struct ifnet *nd6_defifp;
90 int nd6_defifindex;
91 
92 int ip6_use_tempaddr = 0;
93 
94 int ip6_desync_factor;
95 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
96 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
97 /*
98  * shorter lifetimes for debugging purposes.
99 int ip6_temp_preferred_lifetime = 800;
100 static int ip6_temp_valid_lifetime = 1800;
101 */
102 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
103 
104 /* RTPREF_MEDIUM has to be 0! */
105 #define RTPREF_HIGH	1
106 #define RTPREF_MEDIUM	0
107 #define RTPREF_LOW	(-1)
108 #define RTPREF_RESERVED	(-2)
109 #define RTPREF_INVALID	(-3)	/* internal */
110 
111 /*
112  * Receive Router Solicitation Message - just for routers.
113  * Router solicitation/advertisement is mostly managed by userland program
114  * (rtadvd) so here we have no function like nd6_ra_output().
115  *
116  * Based on RFC 2461
117  */
118 void
119 nd6_rs_input(m, off, icmp6len)
120 	struct	mbuf *m;
121 	int off, icmp6len;
122 {
123 	struct ifnet *ifp = m->m_pkthdr.rcvif;
124 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
125 	struct nd_router_solicit *nd_rs;
126 	struct in6_addr saddr6 = ip6->ip6_src;
127 	char *lladdr = NULL;
128 	int lladdrlen = 0;
129 	union nd_opts ndopts;
130 
131 	/* If I'm not a router, ignore it. */
132 	if (ip6_accept_rtadv != 0 || ip6_forwarding != 1)
133 		goto freeit;
134 
135 	/* Sanity checks */
136 	if (ip6->ip6_hlim != 255) {
137 		nd6log((LOG_ERR,
138 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
139 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
140 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
141 		goto bad;
142 	}
143 
144 	/*
145 	 * Don't update the neighbor cache, if src = ::.
146 	 * This indicates that the src has no IP address assigned yet.
147 	 */
148 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
149 		goto freeit;
150 
151 #ifndef PULLDOWN_TEST
152 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
153 	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
154 #else
155 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
156 	if (nd_rs == NULL) {
157 		icmp6stat.icp6s_tooshort++;
158 		return;
159 	}
160 #endif
161 
162 	icmp6len -= sizeof(*nd_rs);
163 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
164 	if (nd6_options(&ndopts) < 0) {
165 		nd6log((LOG_INFO,
166 		    "nd6_rs_input: invalid ND option, ignored\n"));
167 		/* nd6_options have incremented stats */
168 		goto freeit;
169 	}
170 
171 	if (ndopts.nd_opts_src_lladdr) {
172 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
173 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
174 	}
175 
176 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
177 		nd6log((LOG_INFO,
178 		    "nd6_rs_input: lladdrlen mismatch for %s "
179 		    "(if %d, RS packet %d)\n",
180 		    ip6_sprintf(&saddr6),
181 		    ifp->if_addrlen, lladdrlen - 2));
182 		goto bad;
183 	}
184 
185 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
186 
187  freeit:
188 	m_freem(m);
189 	return;
190 
191  bad:
192 	icmp6stat.icp6s_badrs++;
193 	m_freem(m);
194 }
195 
196 /*
197  * Receive Router Advertisement Message.
198  *
199  * Based on RFC 2461
200  * TODO: on-link bit on prefix information
201  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
202  */
203 void
204 nd6_ra_input(m, off, icmp6len)
205 	struct	mbuf *m;
206 	int off, icmp6len;
207 {
208 	struct ifnet *ifp = m->m_pkthdr.rcvif;
209 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
210 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
211 	struct nd_router_advert *nd_ra;
212 	struct in6_addr saddr6 = ip6->ip6_src;
213 	int mcast = 0;
214 	union nd_opts ndopts;
215 	struct nd_defrouter *dr;
216 
217 	/*
218 	 * We only accept RAs only when
219 	 * the system-wide variable allows the acceptance, and
220 	 * per-interface variable allows RAs on the receiving interface.
221 	 */
222 	if (ip6_accept_rtadv == 0)
223 		goto freeit;
224 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
225 		goto freeit;
226 
227 	if (ip6->ip6_hlim != 255) {
228 		nd6log((LOG_ERR,
229 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
230 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
231 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
232 		goto bad;
233 	}
234 
235 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
236 		nd6log((LOG_ERR,
237 		    "nd6_ra_input: src %s is not link-local\n",
238 		    ip6_sprintf(&saddr6)));
239 		goto bad;
240 	}
241 
242 #ifndef PULLDOWN_TEST
243 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
244 	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
245 #else
246 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
247 	if (nd_ra == NULL) {
248 		icmp6stat.icp6s_tooshort++;
249 		return;
250 	}
251 #endif
252 
253 	icmp6len -= sizeof(*nd_ra);
254 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
255 	if (nd6_options(&ndopts) < 0) {
256 		nd6log((LOG_INFO,
257 		    "nd6_ra_input: invalid ND option, ignored\n"));
258 		/* nd6_options have incremented stats */
259 		goto freeit;
260 	}
261 
262     {
263 	struct nd_defrouter dr0;
264 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
265 
266 	/* remember if this is a multicasted advertisement */
267 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
268 		mcast = 1;
269 
270 	bzero(&dr0, sizeof(dr0));
271 	dr0.rtaddr = saddr6;
272 	dr0.flags  = nd_ra->nd_ra_flags_reserved;
273 	dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
274 	dr0.expire = time_second + dr0.rtlifetime;
275 	dr0.ifp = ifp;
276 	/* unspecified or not? (RFC 2461 6.3.4) */
277 	if (advreachable) {
278 		advreachable = ntohl(advreachable);
279 		if (advreachable <= MAX_REACHABLE_TIME &&
280 		    ndi->basereachable != advreachable) {
281 			ndi->basereachable = advreachable;
282 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
283 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
284 		}
285 	}
286 	if (nd_ra->nd_ra_retransmit)
287 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
288 	if (nd_ra->nd_ra_curhoplimit)
289 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
290 	dr = defrtrlist_update(&dr0);
291     }
292 
293 	/*
294 	 * prefix
295 	 */
296 	if (ndopts.nd_opts_pi) {
297 		struct nd_opt_hdr *pt;
298 		struct nd_opt_prefix_info *pi = NULL;
299 		struct nd_prefixctl pr;
300 
301 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
302 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
303 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
304 						(pt->nd_opt_len << 3))) {
305 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
306 				continue;
307 			pi = (struct nd_opt_prefix_info *)pt;
308 
309 			if (pi->nd_opt_pi_len != 4) {
310 				nd6log((LOG_INFO,
311 				    "nd6_ra_input: invalid option "
312 				    "len %d for prefix information option, "
313 				    "ignored\n", pi->nd_opt_pi_len));
314 				continue;
315 			}
316 
317 			if (128 < pi->nd_opt_pi_prefix_len) {
318 				nd6log((LOG_INFO,
319 				    "nd6_ra_input: invalid prefix "
320 				    "len %d for prefix information option, "
321 				    "ignored\n", pi->nd_opt_pi_prefix_len));
322 				continue;
323 			}
324 
325 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
326 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
327 				nd6log((LOG_INFO,
328 				    "nd6_ra_input: invalid prefix "
329 				    "%s, ignored\n",
330 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
331 				continue;
332 			}
333 
334 			bzero(&pr, sizeof(pr));
335 			pr.ndpr_prefix.sin6_family = AF_INET6;
336 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
337 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
338 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
339 
340 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
341 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
342 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
343 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
344 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
345 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
346 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
347 			(void)prelist_update(&pr, dr, m, mcast);
348 		}
349 	}
350 
351 	/*
352 	 * MTU
353 	 */
354 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
355 		u_long mtu;
356 		u_long maxmtu;
357 
358 		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
359 
360 		/* lower bound */
361 		if (mtu < IPV6_MMTU) {
362 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
363 			    "mtu=%lu sent from %s, ignoring\n",
364 			    mtu, ip6_sprintf(&ip6->ip6_src)));
365 			goto skip;
366 		}
367 
368 		/* upper bound */
369 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
370 		    ? ndi->maxmtu : ifp->if_mtu;
371 		if (mtu <= maxmtu) {
372 			int change = (ndi->linkmtu != mtu);
373 
374 			ndi->linkmtu = mtu;
375 			if (change) /* in6_maxmtu may change */
376 				in6_setmaxmtu();
377 		} else {
378 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
379 			    "mtu=%lu sent from %s; "
380 			    "exceeds maxmtu %lu, ignoring\n",
381 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
382 		}
383 	}
384 
385  skip:
386 
387 	/*
388 	 * Source link layer address
389 	 */
390     {
391 	char *lladdr = NULL;
392 	int lladdrlen = 0;
393 
394 	if (ndopts.nd_opts_src_lladdr) {
395 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
396 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
397 	}
398 
399 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
400 		nd6log((LOG_INFO,
401 		    "nd6_ra_input: lladdrlen mismatch for %s "
402 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
403 		    ifp->if_addrlen, lladdrlen - 2));
404 		goto bad;
405 	}
406 
407 	nd6_cache_lladdr(ifp, &saddr6, lladdr,
408 	    lladdrlen, ND_ROUTER_ADVERT, 0);
409 
410 	/*
411 	 * Installing a link-layer address might change the state of the
412 	 * router's neighbor cache, which might also affect our on-link
413 	 * detection of adveritsed prefixes.
414 	 */
415 	pfxlist_onlink_check();
416     }
417 
418  freeit:
419 	m_freem(m);
420 	return;
421 
422  bad:
423 	icmp6stat.icp6s_badra++;
424 	m_freem(m);
425 }
426 
427 /*
428  * default router list proccessing sub routines
429  */
430 
431 /* tell the change to user processes watching the routing socket. */
432 static void
433 nd6_rtmsg(cmd, rt)
434 	int cmd;
435 	struct rtentry *rt;
436 {
437 	struct rt_addrinfo info;
438 
439 	bzero((caddr_t)&info, sizeof(info));
440 	info.rti_info[RTAX_DST] = rt_key(rt);
441 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
442 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
443 	if (rt->rt_ifp) {
444 		info.rti_info[RTAX_IFP] =
445 		    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
446 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
447 	}
448 
449 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
450 }
451 
452 void
453 defrouter_addreq(new)
454 	struct nd_defrouter *new;
455 {
456 	struct sockaddr_in6 def, mask, gate;
457 	struct rtentry *newrt = NULL;
458 	int s;
459 	int error;
460 
461 	bzero(&def, sizeof(def));
462 	bzero(&mask, sizeof(mask));
463 	bzero(&gate, sizeof(gate));
464 
465 	def.sin6_len = mask.sin6_len = gate.sin6_len =
466 	    sizeof(struct sockaddr_in6);
467 	def.sin6_family = gate.sin6_family = AF_INET6;
468 	gate.sin6_addr = new->rtaddr;
469 
470 	s = splnet();
471 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
472 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
473 	    RTF_GATEWAY, &newrt);
474 	if (newrt) {
475 		RT_LOCK(newrt);
476 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
477 		RT_REMREF(newrt);
478 		RT_UNLOCK(newrt);
479 	}
480 	if (error == 0)
481 		new->installed = 1;
482 	splx(s);
483 	return;
484 }
485 
486 struct nd_defrouter *
487 defrouter_lookup(addr, ifp)
488 	struct in6_addr *addr;
489 	struct ifnet *ifp;
490 {
491 	struct nd_defrouter *dr;
492 
493 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
494 	     dr = TAILQ_NEXT(dr, dr_entry)) {
495 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
496 			return (dr);
497 	}
498 
499 	return (NULL);		/* search failed */
500 }
501 
502 /*
503  * Remove the default route for a given router.
504  * This is just a subroutine function for defrouter_select(), and should
505  * not be called from anywhere else.
506  */
507 static void
508 defrouter_delreq(dr)
509 	struct nd_defrouter *dr;
510 {
511 	struct sockaddr_in6 def, mask, gate;
512 	struct rtentry *oldrt = NULL;
513 
514 	bzero(&def, sizeof(def));
515 	bzero(&mask, sizeof(mask));
516 	bzero(&gate, sizeof(gate));
517 
518 	def.sin6_len = mask.sin6_len = gate.sin6_len =
519 	    sizeof(struct sockaddr_in6);
520 	def.sin6_family = gate.sin6_family = AF_INET6;
521 	gate.sin6_addr = dr->rtaddr;
522 
523 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
524 	    (struct sockaddr *)&gate,
525 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
526 	if (oldrt) {
527 		nd6_rtmsg(RTM_DELETE, oldrt);
528 		RTFREE(oldrt);
529 	}
530 
531 	dr->installed = 0;
532 }
533 
534 /*
535  * remove all default routes from default router list
536  */
537 void
538 defrouter_reset()
539 {
540 	struct nd_defrouter *dr;
541 
542 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
543 	     dr = TAILQ_NEXT(dr, dr_entry))
544 		defrouter_delreq(dr);
545 
546 	/*
547 	 * XXX should we also nuke any default routers in the kernel, by
548 	 * going through them by rtalloc1()?
549 	 */
550 }
551 
552 void
553 defrtrlist_del(dr)
554 	struct nd_defrouter *dr;
555 {
556 	struct nd_defrouter *deldr = NULL;
557 	struct nd_prefix *pr;
558 
559 	/*
560 	 * Flush all the routing table entries that use the router
561 	 * as a next hop.
562 	 */
563 	if (!ip6_forwarding && ip6_accept_rtadv) /* XXX: better condition? */
564 		rt6_flush(&dr->rtaddr, dr->ifp);
565 
566 	if (dr->installed) {
567 		deldr = dr;
568 		defrouter_delreq(dr);
569 	}
570 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
571 
572 	/*
573 	 * Also delete all the pointers to the router in each prefix lists.
574 	 */
575 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
576 		struct nd_pfxrouter *pfxrtr;
577 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
578 			pfxrtr_del(pfxrtr);
579 	}
580 	pfxlist_onlink_check();
581 
582 	/*
583 	 * If the router is the primary one, choose a new one.
584 	 * Note that defrouter_select() will remove the current gateway
585 	 * from the routing table.
586 	 */
587 	if (deldr)
588 		defrouter_select();
589 
590 	free(dr, M_IP6NDP);
591 }
592 
593 /*
594  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
595  * draft-ietf-ipngwg-router-selection:
596  * 1) Routers that are reachable or probably reachable should be preferred.
597  *    If we have more than one (probably) reachable router, prefer ones
598  *    with the highest router preference.
599  * 2) When no routers on the list are known to be reachable or
600  *    probably reachable, routers SHOULD be selected in a round-robin
601  *    fashion, regardless of router preference values.
602  * 3) If the Default Router List is empty, assume that all
603  *    destinations are on-link.
604  *
605  * We assume nd_defrouter is sorted by router preference value.
606  * Since the code below covers both with and without router preference cases,
607  * we do not need to classify the cases by ifdef.
608  *
609  * At this moment, we do not try to install more than one default router,
610  * even when the multipath routing is available, because we're not sure about
611  * the benefits for stub hosts comparing to the risk of making the code
612  * complicated and the possibility of introducing bugs.
613  */
614 void
615 defrouter_select()
616 {
617 	int s = splnet();
618 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
619 	struct rtentry *rt = NULL;
620 	struct llinfo_nd6 *ln = NULL;
621 
622 	/*
623 	 * This function should be called only when acting as an autoconfigured
624 	 * host.  Although the remaining part of this function is not effective
625 	 * if the node is not an autoconfigured host, we explicitly exclude
626 	 * such cases here for safety.
627 	 */
628 	if (ip6_forwarding || !ip6_accept_rtadv) {
629 		nd6log((LOG_WARNING,
630 		    "defrouter_select: called unexpectedly (forwarding=%d, "
631 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
632 		splx(s);
633 		return;
634 	}
635 
636 	/*
637 	 * Let's handle easy case (3) first:
638 	 * If default router list is empty, there's nothing to be done.
639 	 */
640 	if (!TAILQ_FIRST(&nd_defrouter)) {
641 		splx(s);
642 		return;
643 	}
644 
645 	/*
646 	 * Search for a (probably) reachable router from the list.
647 	 * We just pick up the first reachable one (if any), assuming that
648 	 * the ordering rule of the list described in defrtrlist_update().
649 	 */
650 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
651 	     dr = TAILQ_NEXT(dr, dr_entry)) {
652 		if (selected_dr == NULL &&
653 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
654 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
655 		    ND6_IS_LLINFO_PROBREACH(ln)) {
656 			selected_dr = dr;
657 		}
658 
659 		if (dr->installed && installed_dr == NULL)
660 			installed_dr = dr;
661 		else if (dr->installed && installed_dr) {
662 			/* this should not happen.  warn for diagnosis. */
663 			log(LOG_ERR, "defrouter_select: more than one router"
664 			    " is installed\n");
665 		}
666 	}
667 	/*
668 	 * If none of the default routers was found to be reachable,
669 	 * round-robin the list regardless of preference.
670 	 * Otherwise, if we have an installed router, check if the selected
671 	 * (reachable) router should really be preferred to the installed one.
672 	 * We only prefer the new router when the old one is not reachable
673 	 * or when the new one has a really higher preference value.
674 	 */
675 	if (selected_dr == NULL) {
676 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
677 			selected_dr = TAILQ_FIRST(&nd_defrouter);
678 		else
679 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
680 	} else if (installed_dr &&
681 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
682 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
683 	    ND6_IS_LLINFO_PROBREACH(ln) &&
684 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
685 		selected_dr = installed_dr;
686 	}
687 
688 	/*
689 	 * If the selected router is different than the installed one,
690 	 * remove the installed router and install the selected one.
691 	 * Note that the selected router is never NULL here.
692 	 */
693 	if (installed_dr != selected_dr) {
694 		if (installed_dr)
695 			defrouter_delreq(installed_dr);
696 		defrouter_addreq(selected_dr);
697 	}
698 
699 	splx(s);
700 	return;
701 }
702 
703 /*
704  * for default router selection
705  * regards router-preference field as a 2-bit signed integer
706  */
707 static int
708 rtpref(struct nd_defrouter *dr)
709 {
710 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
711 	case ND_RA_FLAG_RTPREF_HIGH:
712 		return (RTPREF_HIGH);
713 	case ND_RA_FLAG_RTPREF_MEDIUM:
714 	case ND_RA_FLAG_RTPREF_RSV:
715 		return (RTPREF_MEDIUM);
716 	case ND_RA_FLAG_RTPREF_LOW:
717 		return (RTPREF_LOW);
718 	default:
719 		/*
720 		 * This case should never happen.  If it did, it would mean a
721 		 * serious bug of kernel internal.  We thus always bark here.
722 		 * Or, can we even panic?
723 		 */
724 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
725 		return (RTPREF_INVALID);
726 	}
727 	/* NOTREACHED */
728 }
729 
730 static struct nd_defrouter *
731 defrtrlist_update(new)
732 	struct nd_defrouter *new;
733 {
734 	struct nd_defrouter *dr, *n;
735 	int s = splnet();
736 
737 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
738 		/* entry exists */
739 		if (new->rtlifetime == 0) {
740 			defrtrlist_del(dr);
741 			dr = NULL;
742 		} else {
743 			int oldpref = rtpref(dr);
744 
745 			/* override */
746 			dr->flags = new->flags; /* xxx flag check */
747 			dr->rtlifetime = new->rtlifetime;
748 			dr->expire = new->expire;
749 
750 			/*
751 			 * If the preference does not change, there's no need
752 			 * to sort the entries.
753 			 */
754 			if (rtpref(new) == oldpref) {
755 				splx(s);
756 				return (dr);
757 			}
758 
759 			/*
760 			 * preferred router may be changed, so relocate
761 			 * this router.
762 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
763 			 * However, since defrtrlist_del() has many side
764 			 * effects, we intentionally do so here.
765 			 * defrouter_select() below will handle routing
766 			 * changes later.
767 			 */
768 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
769 			n = dr;
770 			goto insert;
771 		}
772 		splx(s);
773 		return (dr);
774 	}
775 
776 	/* entry does not exist */
777 	if (new->rtlifetime == 0) {
778 		splx(s);
779 		return (NULL);
780 	}
781 
782 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
783 	if (n == NULL) {
784 		splx(s);
785 		return (NULL);
786 	}
787 	bzero(n, sizeof(*n));
788 	*n = *new;
789 
790 insert:
791 	/*
792 	 * Insert the new router in the Default Router List;
793 	 * The Default Router List should be in the descending order
794 	 * of router-preferece.  Routers with the same preference are
795 	 * sorted in the arriving time order.
796 	 */
797 
798 	/* insert at the end of the group */
799 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
800 	     dr = TAILQ_NEXT(dr, dr_entry)) {
801 		if (rtpref(n) > rtpref(dr))
802 			break;
803 	}
804 	if (dr)
805 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
806 	else
807 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
808 
809 	defrouter_select();
810 
811 	splx(s);
812 
813 	return (n);
814 }
815 
816 static struct nd_pfxrouter *
817 pfxrtr_lookup(pr, dr)
818 	struct nd_prefix *pr;
819 	struct nd_defrouter *dr;
820 {
821 	struct nd_pfxrouter *search;
822 
823 	for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
824 		if (search->router == dr)
825 			break;
826 	}
827 
828 	return (search);
829 }
830 
831 static void
832 pfxrtr_add(pr, dr)
833 	struct nd_prefix *pr;
834 	struct nd_defrouter *dr;
835 {
836 	struct nd_pfxrouter *new;
837 
838 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
839 	if (new == NULL)
840 		return;
841 	bzero(new, sizeof(*new));
842 	new->router = dr;
843 
844 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
845 
846 	pfxlist_onlink_check();
847 }
848 
849 static void
850 pfxrtr_del(pfr)
851 	struct nd_pfxrouter *pfr;
852 {
853 	LIST_REMOVE(pfr, pfr_entry);
854 	free(pfr, M_IP6NDP);
855 }
856 
857 struct nd_prefix *
858 nd6_prefix_lookup(key)
859 	struct nd_prefixctl *key;
860 {
861 	struct nd_prefix *search;
862 
863 	for (search = nd_prefix.lh_first; search; search = search->ndpr_next) {
864 		if (key->ndpr_ifp == search->ndpr_ifp &&
865 		    key->ndpr_plen == search->ndpr_plen &&
866 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
867 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
868 			break;
869 		}
870 	}
871 
872 	return (search);
873 }
874 
875 int
876 nd6_prelist_add(pr, dr, newp)
877 	struct nd_prefixctl *pr;
878 	struct nd_prefix **newp;
879 	struct nd_defrouter *dr;
880 {
881 	struct nd_prefix *new = NULL;
882 	int error = 0;
883 	int i, s;
884 
885 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
886 	if (new == NULL)
887 		return(ENOMEM);
888 	bzero(new, sizeof(*new));
889 	new->ndpr_ifp = pr->ndpr_ifp;
890 	new->ndpr_prefix = pr->ndpr_prefix;
891 	new->ndpr_plen = pr->ndpr_plen;
892 	new->ndpr_vltime = pr->ndpr_vltime;
893 	new->ndpr_pltime = pr->ndpr_pltime;
894 	new->ndpr_flags = pr->ndpr_flags;
895 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
896 		free(new, M_IP6NDP);
897 		return(error);
898 	}
899 	new->ndpr_lastupdate = time_second;
900 	if (newp != NULL)
901 		*newp = new;
902 
903 	/* initialization */
904 	LIST_INIT(&new->ndpr_advrtrs);
905 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
906 	/* make prefix in the canonical form */
907 	for (i = 0; i < 4; i++)
908 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
909 		    new->ndpr_mask.s6_addr32[i];
910 
911 	s = splnet();
912 	/* link ndpr_entry to nd_prefix list */
913 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
914 	splx(s);
915 
916 	/* ND_OPT_PI_FLAG_ONLINK processing */
917 	if (new->ndpr_raf_onlink) {
918 		int e;
919 
920 		if ((e = nd6_prefix_onlink(new)) != 0) {
921 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
922 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
923 			    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
924 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
925 			/* proceed anyway. XXX: is it correct? */
926 		}
927 	}
928 
929 	if (dr)
930 		pfxrtr_add(new, dr);
931 
932 	return 0;
933 }
934 
935 void
936 prelist_remove(pr)
937 	struct nd_prefix *pr;
938 {
939 	struct nd_pfxrouter *pfr, *next;
940 	int e, s;
941 
942 	/* make sure to invalidate the prefix until it is really freed. */
943 	pr->ndpr_vltime = 0;
944 	pr->ndpr_pltime = 0;
945 
946 	/*
947 	 * Though these flags are now meaningless, we'd rather keep the value
948 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
949 	 * when executing "ndp -p".
950 	 */
951 
952 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
953 	    (e = nd6_prefix_offlink(pr)) != 0) {
954 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
955 		    "on %s, errno=%d\n",
956 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
957 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
958 		/* what should we do? */
959 	}
960 
961 	if (pr->ndpr_refcnt > 0)
962 		return;		/* notice here? */
963 
964 	s = splnet();
965 
966 	/* unlink ndpr_entry from nd_prefix list */
967 	LIST_REMOVE(pr, ndpr_entry);
968 
969 	/* free list of routers that adversed the prefix */
970 	for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
971 		next = pfr->pfr_next;
972 
973 		free(pfr, M_IP6NDP);
974 	}
975 	splx(s);
976 
977 	free(pr, M_IP6NDP);
978 
979 	pfxlist_onlink_check();
980 }
981 
982 static int
983 prelist_update(new, dr, m, mcast)
984 	struct nd_prefixctl *new;
985 	struct nd_defrouter *dr; /* may be NULL */
986 	struct mbuf *m;
987 	int mcast;
988 {
989 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
990 	struct ifaddr *ifa;
991 	struct ifnet *ifp = new->ndpr_ifp;
992 	struct nd_prefix *pr;
993 	int s = splnet();
994 	int error = 0;
995 	int newprefix = 0;
996 	int auth;
997 	struct in6_addrlifetime lt6_tmp;
998 
999 	auth = 0;
1000 	if (m) {
1001 		/*
1002 		 * Authenticity for NA consists authentication for
1003 		 * both IP header and IP datagrams, doesn't it ?
1004 		 */
1005 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1006 		auth = ((m->m_flags & M_AUTHIPHDR) &&
1007 		    (m->m_flags & M_AUTHIPDGM));
1008 #endif
1009 	}
1010 
1011 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1012 		/*
1013 		 * nd6_prefix_lookup() ensures that pr and new have the same
1014 		 * prefix on a same interface.
1015 		 */
1016 
1017 		/*
1018 		 * Update prefix information.  Note that the on-link (L) bit
1019 		 * and the autonomous (A) bit should NOT be changed from 1
1020 		 * to 0.
1021 		 */
1022 		if (new->ndpr_raf_onlink == 1)
1023 			pr->ndpr_raf_onlink = 1;
1024 		if (new->ndpr_raf_auto == 1)
1025 			pr->ndpr_raf_auto = 1;
1026 		if (new->ndpr_raf_onlink) {
1027 			pr->ndpr_vltime = new->ndpr_vltime;
1028 			pr->ndpr_pltime = new->ndpr_pltime;
1029 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1030 			pr->ndpr_lastupdate = time_second;
1031 		}
1032 
1033 		if (new->ndpr_raf_onlink &&
1034 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1035 			int e;
1036 
1037 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1038 				nd6log((LOG_ERR,
1039 				    "prelist_update: failed to make "
1040 				    "the prefix %s/%d on-link on %s "
1041 				    "(errno=%d)\n",
1042 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1043 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1044 				/* proceed anyway. XXX: is it correct? */
1045 			}
1046 		}
1047 
1048 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1049 			pfxrtr_add(pr, dr);
1050 	} else {
1051 		struct nd_prefix *newpr = NULL;
1052 
1053 		newprefix = 1;
1054 
1055 		if (new->ndpr_vltime == 0)
1056 			goto end;
1057 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1058 			goto end;
1059 
1060 		error = nd6_prelist_add(new, dr, &newpr);
1061 		if (error != 0 || newpr == NULL) {
1062 			nd6log((LOG_NOTICE, "prelist_update: "
1063 			    "nd6_prelist_add failed for %s/%d on %s "
1064 			    "errno=%d, returnpr=%p\n",
1065 			    ip6_sprintf(&new->ndpr_prefix.sin6_addr),
1066 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1067 			    error, newpr));
1068 			goto end; /* we should just give up in this case. */
1069 		}
1070 
1071 		/*
1072 		 * XXX: from the ND point of view, we can ignore a prefix
1073 		 * with the on-link bit being zero.  However, we need a
1074 		 * prefix structure for references from autoconfigured
1075 		 * addresses.  Thus, we explicitly make sure that the prefix
1076 		 * itself expires now.
1077 		 */
1078 		if (newpr->ndpr_raf_onlink == 0) {
1079 			newpr->ndpr_vltime = 0;
1080 			newpr->ndpr_pltime = 0;
1081 			in6_init_prefix_ltimes(newpr);
1082 		}
1083 
1084 		pr = newpr;
1085 	}
1086 
1087 	/*
1088 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1089 	 * Note that pr must be non NULL at this point.
1090 	 */
1091 
1092 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1093 	if (!new->ndpr_raf_auto)
1094 		goto end;
1095 
1096 	/*
1097 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1098 	 * nd6_ra_input.
1099 	 */
1100 
1101 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1102 	if (new->ndpr_pltime > new->ndpr_vltime) {
1103 		error = EINVAL;	/* XXX: won't be used */
1104 		goto end;
1105 	}
1106 
1107  	/*
1108 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1109 	 * an address configured by stateless autoconfiguration already in the
1110 	 * list of addresses associated with the interface, and the Valid
1111 	 * Lifetime is not 0, form an address.  We first check if we have
1112 	 * a matching prefix.
1113 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1114 	 * consider autoconfigured addresses while RFC2462 simply said
1115 	 * "address".
1116 	 */
1117 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1118 		struct in6_ifaddr *ifa6;
1119 		u_int32_t remaininglifetime;
1120 
1121 		if (ifa->ifa_addr->sa_family != AF_INET6)
1122 			continue;
1123 
1124 		ifa6 = (struct in6_ifaddr *)ifa;
1125 
1126 		/*
1127 		 * We only consider autoconfigured addresses as per rfc2462bis.
1128 		 */
1129 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1130 			continue;
1131 
1132 		/*
1133 		 * Spec is not clear here, but I believe we should concentrate
1134 		 * on unicast (i.e. not anycast) addresses.
1135 		 * XXX: other ia6_flags? detached or duplicated?
1136 		 */
1137 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1138 			continue;
1139 
1140 		/*
1141 		 * Ignore the address if it is not associated with a prefix
1142 		 * or is associated with a prefix that is different from this
1143 		 * one.  (pr is never NULL here)
1144 		 */
1145 		if (ifa6->ia6_ndpr != pr)
1146 			continue;
1147 
1148 		if (ia6_match == NULL) /* remember the first one */
1149 			ia6_match = ifa6;
1150 
1151 		/*
1152 		 * An already autoconfigured address matched.  Now that we
1153 		 * are sure there is at least one matched address, we can
1154 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1155 		 * "two hours" rule and the privacy extension.
1156 		 * We apply some clarifications in rfc2462bis:
1157 		 * - use remaininglifetime instead of storedlifetime as a
1158 		 *   variable name
1159 		 * - remove the dead code in the "two-hour" rule
1160 		 */
1161 #define TWOHOUR		(120*60)
1162 		lt6_tmp = ifa6->ia6_lifetime;
1163 
1164 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1165 			remaininglifetime = ND6_INFINITE_LIFETIME;
1166 		else if (time_second - ifa6->ia6_updatetime >
1167 			 lt6_tmp.ia6t_vltime) {
1168 			/*
1169 			 * The case of "invalid" address.  We should usually
1170 			 * not see this case.
1171 			 */
1172 			remaininglifetime = 0;
1173 		} else
1174 			remaininglifetime = lt6_tmp.ia6t_vltime -
1175 			    (time_second - ifa6->ia6_updatetime);
1176 
1177 		/* when not updating, keep the current stored lifetime. */
1178 		lt6_tmp.ia6t_vltime = remaininglifetime;
1179 
1180 		if (TWOHOUR < new->ndpr_vltime ||
1181 		    remaininglifetime < new->ndpr_vltime) {
1182 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1183 		} else if (remaininglifetime <= TWOHOUR) {
1184 			if (auth) {
1185 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1186 			}
1187 		} else {
1188 			/*
1189 			 * new->ndpr_vltime <= TWOHOUR &&
1190 			 * TWOHOUR < remaininglifetime
1191 			 */
1192 			lt6_tmp.ia6t_vltime = TWOHOUR;
1193 		}
1194 
1195 		/* The 2 hour rule is not imposed for preferred lifetime. */
1196 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1197 
1198 		in6_init_address_ltimes(pr, &lt6_tmp);
1199 
1200   		/*
1201 		 * We need to treat lifetimes for temporary addresses
1202 		 * differently, according to
1203 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1204 		 * we only update the lifetimes when they are in the maximum
1205 		 * intervals.
1206   		 */
1207   		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1208 			u_int32_t maxvltime, maxpltime;
1209 
1210 			if (ip6_temp_valid_lifetime >
1211 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1212 			    ip6_desync_factor)) {
1213 				maxvltime = ip6_temp_valid_lifetime -
1214 				    (time_second - ifa6->ia6_createtime) -
1215 				    ip6_desync_factor;
1216 			} else
1217 				maxvltime = 0;
1218 			if (ip6_temp_preferred_lifetime >
1219 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1220 			    ip6_desync_factor)) {
1221 				maxpltime = ip6_temp_preferred_lifetime -
1222 				    (time_second - ifa6->ia6_createtime) -
1223 				    ip6_desync_factor;
1224 			} else
1225 				maxpltime = 0;
1226 
1227 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1228 			    lt6_tmp.ia6t_vltime > maxvltime) {
1229 				lt6_tmp.ia6t_vltime = maxvltime;
1230 			}
1231 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1232 			    lt6_tmp.ia6t_pltime > maxpltime) {
1233 				lt6_tmp.ia6t_pltime = maxpltime;
1234 			}
1235 		}
1236 		ifa6->ia6_lifetime = lt6_tmp;
1237 		ifa6->ia6_updatetime = time_second;
1238 	}
1239 	if (ia6_match == NULL && new->ndpr_vltime) {
1240 		int ifidlen;
1241 
1242 		/*
1243 		 * 5.5.3 (d) (continued)
1244 		 * No address matched and the valid lifetime is non-zero.
1245 		 * Create a new address.
1246 		 */
1247 
1248 		/*
1249 		 * Prefix Length check:
1250 		 * If the sum of the prefix length and interface identifier
1251 		 * length does not equal 128 bits, the Prefix Information
1252 		 * option MUST be ignored.  The length of the interface
1253 		 * identifier is defined in a separate link-type specific
1254 		 * document.
1255 		 */
1256 		ifidlen = in6_if2idlen(ifp);
1257 		if (ifidlen < 0) {
1258 			/* this should not happen, so we always log it. */
1259 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1260 			    if_name(ifp));
1261 			goto end;
1262 		}
1263 		if (ifidlen + pr->ndpr_plen != 128) {
1264 			nd6log((LOG_INFO,
1265 			    "prelist_update: invalid prefixlen "
1266 			    "%d for %s, ignored\n",
1267 			    pr->ndpr_plen, if_name(ifp)));
1268 			goto end;
1269 		}
1270 
1271 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1272 			/*
1273 			 * note that we should use pr (not new) for reference.
1274 			 */
1275 			pr->ndpr_refcnt++;
1276 			ia6->ia6_ndpr = pr;
1277 
1278 			/*
1279 			 * RFC 3041 3.3 (2).
1280 			 * When a new public address is created as described
1281 			 * in RFC2462, also create a new temporary address.
1282 			 *
1283 			 * RFC 3041 3.5.
1284 			 * When an interface connects to a new link, a new
1285 			 * randomized interface identifier should be generated
1286 			 * immediately together with a new set of temporary
1287 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1288 			 * in6_tmpifadd().
1289 			 */
1290 			if (ip6_use_tempaddr) {
1291 				int e;
1292 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1293 					nd6log((LOG_NOTICE, "prelist_update: "
1294 					    "failed to create a temporary "
1295 					    "address, errno=%d\n",
1296 					    e));
1297 				}
1298 			}
1299 
1300 			/*
1301 			 * A newly added address might affect the status
1302 			 * of other addresses, so we check and update it.
1303 			 * XXX: what if address duplication happens?
1304 			 */
1305 			pfxlist_onlink_check();
1306 		} else {
1307 			/* just set an error. do not bark here. */
1308 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1309 		}
1310 	}
1311 
1312  end:
1313 	splx(s);
1314 	return error;
1315 }
1316 
1317 /*
1318  * A supplement function used in the on-link detection below;
1319  * detect if a given prefix has a (probably) reachable advertising router.
1320  * XXX: lengthy function name...
1321  */
1322 static struct nd_pfxrouter *
1323 find_pfxlist_reachable_router(pr)
1324 	struct nd_prefix *pr;
1325 {
1326 	struct nd_pfxrouter *pfxrtr;
1327 	struct rtentry *rt;
1328 	struct llinfo_nd6 *ln;
1329 
1330 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1331 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1332 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1333 		    pfxrtr->router->ifp)) &&
1334 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1335 		    ND6_IS_LLINFO_PROBREACH(ln))
1336 			break;	/* found */
1337 	}
1338 
1339 	return (pfxrtr);
1340 }
1341 
1342 /*
1343  * Check if each prefix in the prefix list has at least one available router
1344  * that advertised the prefix (a router is "available" if its neighbor cache
1345  * entry is reachable or probably reachable).
1346  * If the check fails, the prefix may be off-link, because, for example,
1347  * we have moved from the network but the lifetime of the prefix has not
1348  * expired yet.  So we should not use the prefix if there is another prefix
1349  * that has an available router.
1350  * But, if there is no prefix that has an available router, we still regards
1351  * all the prefixes as on-link.  This is because we can't tell if all the
1352  * routers are simply dead or if we really moved from the network and there
1353  * is no router around us.
1354  */
1355 void
1356 pfxlist_onlink_check()
1357 {
1358 	struct nd_prefix *pr;
1359 	struct in6_ifaddr *ifa;
1360 	struct nd_defrouter *dr;
1361 	struct nd_pfxrouter *pfxrtr = NULL;
1362 
1363 	/*
1364 	 * Check if there is a prefix that has a reachable advertising
1365 	 * router.
1366 	 */
1367 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1368 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1369 			break;
1370 	}
1371 
1372 	/*
1373 	 * If we have no such prefix, check whether we still have a router
1374 	 * that does not advertise any prefixes.
1375 	 */
1376 	if (pr == NULL) {
1377 		for (dr = TAILQ_FIRST(&nd_defrouter); dr;
1378 		    dr = TAILQ_NEXT(dr, dr_entry)) {
1379 			struct nd_prefix *pr0;
1380 
1381 			for (pr0 = nd_prefix.lh_first; pr0;
1382 			    pr0 = pr0->ndpr_next) {
1383 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1384 					break;
1385 			}
1386 			if (pfxrtr != NULL)
1387 				break;
1388 		}
1389 	}
1390 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && pfxrtr == NULL)) {
1391   		/*
1392 		 * There is at least one prefix that has a reachable router,
1393 		 * or at least a router which probably does not advertise
1394 		 * any prefixes.  The latter would be the case when we move
1395 		 * to a new link where we have a router that does not provide
1396 		 * prefixes and we configure an address by hand.
1397   		 * Detach prefixes which have no reachable advertising
1398   		 * router, and attach other prefixes.
1399   		 */
1400 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1401 			/* XXX: a link-local prefix should never be detached */
1402 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1403 				continue;
1404 
1405 			/*
1406 			 * we aren't interested in prefixes without the L bit
1407 			 * set.
1408 			 */
1409 			if (pr->ndpr_raf_onlink == 0)
1410 				continue;
1411 
1412 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1413 			    find_pfxlist_reachable_router(pr) == NULL)
1414 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1415 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1416 			    find_pfxlist_reachable_router(pr) != 0)
1417 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1418 		}
1419 	} else {
1420 		/* there is no prefix that has a reachable router */
1421 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1422 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1423 				continue;
1424 
1425 			if (pr->ndpr_raf_onlink == 0)
1426 				continue;
1427 
1428 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1429 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1430 		}
1431 	}
1432 
1433 	/*
1434 	 * Remove each interface route associated with a (just) detached
1435 	 * prefix, and reinstall the interface route for a (just) attached
1436 	 * prefix.  Note that all attempt of reinstallation does not
1437 	 * necessarily success, when a same prefix is shared among multiple
1438 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1439 	 * so we don't have to care about them.
1440 	 */
1441 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1442 		int e;
1443 
1444 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1445 			continue;
1446 
1447 		if (pr->ndpr_raf_onlink == 0)
1448 			continue;
1449 
1450 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1451 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1452 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1453 				nd6log((LOG_ERR,
1454 				    "pfxlist_onlink_check: failed to "
1455 				    "make %s/%d offlink, errno=%d\n",
1456 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1457 				    pr->ndpr_plen, e));
1458 			}
1459 		}
1460 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1461 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1462 		    pr->ndpr_raf_onlink) {
1463 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1464 				nd6log((LOG_ERR,
1465 				    "pfxlist_onlink_check: failed to "
1466 				    "make %s/%d onlink, errno=%d\n",
1467 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1468 				    pr->ndpr_plen, e));
1469 			}
1470 		}
1471 	}
1472 
1473 	/*
1474 	 * Changes on the prefix status might affect address status as well.
1475 	 * Make sure that all addresses derived from an attached prefix are
1476 	 * attached, and that all addresses derived from a detached prefix are
1477 	 * detached.  Note, however, that a manually configured address should
1478 	 * always be attached.
1479 	 * The precise detection logic is same as the one for prefixes.
1480 	 */
1481 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1482 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1483 			continue;
1484 
1485 		if (ifa->ia6_ndpr == NULL) {
1486 			/*
1487 			 * This can happen when we first configure the address
1488 			 * (i.e. the address exists, but the prefix does not).
1489 			 * XXX: complicated relationships...
1490 			 */
1491 			continue;
1492 		}
1493 
1494 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1495 			break;
1496 	}
1497 	if (ifa) {
1498 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1499 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1500 				continue;
1501 
1502 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1503 				continue;
1504 
1505 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1506 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1507 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1508 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1509 					nd6_dad_start((struct ifaddr *)ifa, 0);
1510 				}
1511 			} else {
1512 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1513 			}
1514 		}
1515 	}
1516 	else {
1517 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1518 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1519 				continue;
1520 
1521 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1522 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1523 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1524 				/* Do we need a delay in this case? */
1525 				nd6_dad_start((struct ifaddr *)ifa, 0);
1526 			}
1527 		}
1528 	}
1529 }
1530 
1531 int
1532 nd6_prefix_onlink(pr)
1533 	struct nd_prefix *pr;
1534 {
1535 	struct ifaddr *ifa;
1536 	struct ifnet *ifp = pr->ndpr_ifp;
1537 	struct sockaddr_in6 mask6;
1538 	struct nd_prefix *opr;
1539 	u_long rtflags;
1540 	int error = 0;
1541 	struct rtentry *rt = NULL;
1542 
1543 	/* sanity check */
1544 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1545 		nd6log((LOG_ERR,
1546 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1547 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1548 		return (EEXIST);
1549 	}
1550 
1551 	/*
1552 	 * Add the interface route associated with the prefix.  Before
1553 	 * installing the route, check if there's the same prefix on another
1554 	 * interface, and the prefix has already installed the interface route.
1555 	 * Although such a configuration is expected to be rare, we explicitly
1556 	 * allow it.
1557 	 */
1558 	for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1559 		if (opr == pr)
1560 			continue;
1561 
1562 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1563 			continue;
1564 
1565 		if (opr->ndpr_plen == pr->ndpr_plen &&
1566 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1567 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1568 			return (0);
1569 	}
1570 
1571 	/*
1572 	 * We prefer link-local addresses as the associated interface address.
1573 	 */
1574 	/* search for a link-local addr */
1575 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1576 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1577 	if (ifa == NULL) {
1578 		/* XXX: freebsd does not have ifa_ifwithaf */
1579 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1580 			if (ifa->ifa_addr->sa_family == AF_INET6)
1581 				break;
1582 		}
1583 		/* should we care about ia6_flags? */
1584 	}
1585 	if (ifa == NULL) {
1586 		/*
1587 		 * This can still happen, when, for example, we receive an RA
1588 		 * containing a prefix with the L bit set and the A bit clear,
1589 		 * after removing all IPv6 addresses on the receiving
1590 		 * interface.  This should, of course, be rare though.
1591 		 */
1592 		nd6log((LOG_NOTICE,
1593 		    "nd6_prefix_onlink: failed to find any ifaddr"
1594 		    " to add route for a prefix(%s/%d) on %s\n",
1595 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1596 		    pr->ndpr_plen, if_name(ifp)));
1597 		return (0);
1598 	}
1599 
1600 	/*
1601 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1602 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1603 	 */
1604 	bzero(&mask6, sizeof(mask6));
1605 	mask6.sin6_len = sizeof(mask6);
1606 	mask6.sin6_addr = pr->ndpr_mask;
1607 	rtflags = ifa->ifa_flags | RTF_CLONING | RTF_UP;
1608 	if (nd6_need_cache(ifp)) {
1609 		/* explicitly set in case ifa_flags does not set the flag. */
1610 		rtflags |= RTF_CLONING;
1611 	} else {
1612 		/*
1613 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1614 		 */
1615 		rtflags &= ~RTF_CLONING;
1616 	}
1617 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1618 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1619 	if (error == 0) {
1620 		if (rt != NULL) /* this should be non NULL, though */
1621 			nd6_rtmsg(RTM_ADD, rt);
1622 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1623 	} else {
1624 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1625 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1626 		    "errno = %d\n",
1627 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1628 		    pr->ndpr_plen, if_name(ifp),
1629 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1630 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
1631 	}
1632 
1633 	if (rt != NULL) {
1634 		RT_LOCK(rt);
1635 		RT_REMREF(rt);
1636 		RT_UNLOCK(rt);
1637 	}
1638 
1639 	return (error);
1640 }
1641 
1642 int
1643 nd6_prefix_offlink(pr)
1644 	struct nd_prefix *pr;
1645 {
1646 	int error = 0;
1647 	struct ifnet *ifp = pr->ndpr_ifp;
1648 	struct nd_prefix *opr;
1649 	struct sockaddr_in6 sa6, mask6;
1650 	struct rtentry *rt = NULL;
1651 
1652 	/* sanity check */
1653 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1654 		nd6log((LOG_ERR,
1655 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1656 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1657 		return (EEXIST);
1658 	}
1659 
1660 	bzero(&sa6, sizeof(sa6));
1661 	sa6.sin6_family = AF_INET6;
1662 	sa6.sin6_len = sizeof(sa6);
1663 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1664 	    sizeof(struct in6_addr));
1665 	bzero(&mask6, sizeof(mask6));
1666 	mask6.sin6_family = AF_INET6;
1667 	mask6.sin6_len = sizeof(sa6);
1668 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1669 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1670 	    (struct sockaddr *)&mask6, 0, &rt);
1671 	if (error == 0) {
1672 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1673 
1674 		/* report the route deletion to the routing socket. */
1675 		if (rt != NULL)
1676 			nd6_rtmsg(RTM_DELETE, rt);
1677 
1678 		/*
1679 		 * There might be the same prefix on another interface,
1680 		 * the prefix which could not be on-link just because we have
1681 		 * the interface route (see comments in nd6_prefix_onlink).
1682 		 * If there's one, try to make the prefix on-link on the
1683 		 * interface.
1684 		 */
1685 		for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1686 			if (opr == pr)
1687 				continue;
1688 
1689 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1690 				continue;
1691 
1692 			/*
1693 			 * KAME specific: detached prefixes should not be
1694 			 * on-link.
1695 			 */
1696 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1697 				continue;
1698 
1699 			if (opr->ndpr_plen == pr->ndpr_plen &&
1700 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1701 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1702 				int e;
1703 
1704 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1705 					nd6log((LOG_ERR,
1706 					    "nd6_prefix_offlink: failed to "
1707 					    "recover a prefix %s/%d from %s "
1708 					    "to %s (errno = %d)\n",
1709 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1710 					    opr->ndpr_plen, if_name(ifp),
1711 					    if_name(opr->ndpr_ifp), e));
1712 				}
1713 			}
1714 		}
1715 	} else {
1716 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1717 		nd6log((LOG_ERR,
1718 		    "nd6_prefix_offlink: failed to delete route: "
1719 		    "%s/%d on %s (errno = %d)\n",
1720 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1721 		    error));
1722 	}
1723 
1724 	if (rt != NULL) {
1725 		RTFREE(rt);
1726 	}
1727 
1728 	return (error);
1729 }
1730 
1731 static struct in6_ifaddr *
1732 in6_ifadd(pr, mcast)
1733 	struct nd_prefixctl *pr;
1734 	int mcast;
1735 {
1736 	struct ifnet *ifp = pr->ndpr_ifp;
1737 	struct ifaddr *ifa;
1738 	struct in6_aliasreq ifra;
1739 	struct in6_ifaddr *ia, *ib;
1740 	int error, plen0;
1741 	struct in6_addr mask;
1742 	int prefixlen = pr->ndpr_plen;
1743 	int updateflags;
1744 
1745 	in6_prefixlen2mask(&mask, prefixlen);
1746 
1747 	/*
1748 	 * find a link-local address (will be interface ID).
1749 	 * Is it really mandatory? Theoretically, a global or a site-local
1750 	 * address can be configured without a link-local address, if we
1751 	 * have a unique interface identifier...
1752 	 *
1753 	 * it is not mandatory to have a link-local address, we can generate
1754 	 * interface identifier on the fly.  we do this because:
1755 	 * (1) it should be the easiest way to find interface identifier.
1756 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1757 	 * for multiple addresses on a single interface, and possible shortcut
1758 	 * of DAD.  we omitted DAD for this reason in the past.
1759 	 * (3) a user can prevent autoconfiguration of global address
1760 	 * by removing link-local address by hand (this is partly because we
1761 	 * don't have other way to control the use of IPv6 on an interface.
1762 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1763 	 * (4) it is easier to manage when an interface has addresses
1764 	 * with the same interface identifier, than to have multiple addresses
1765 	 * with different interface identifiers.
1766 	 */
1767 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1768 	if (ifa)
1769 		ib = (struct in6_ifaddr *)ifa;
1770 	else
1771 		return NULL;
1772 
1773 	/* prefixlen + ifidlen must be equal to 128 */
1774 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1775 	if (prefixlen != plen0) {
1776 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1777 		    "(prefix=%d ifid=%d)\n",
1778 		    if_name(ifp), prefixlen, 128 - plen0));
1779 		return NULL;
1780 	}
1781 
1782 	/* make ifaddr */
1783 
1784 	bzero(&ifra, sizeof(ifra));
1785 	/*
1786 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1787 	 * for safety.
1788 	 */
1789 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1790 	ifra.ifra_addr.sin6_family = AF_INET6;
1791 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1792 	/* prefix */
1793 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1794 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1795 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1796 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1797 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1798 
1799 	/* interface ID */
1800 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1801 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1802 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1803 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1804 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1805 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1806 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1807 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1808 
1809 	/* new prefix mask. */
1810 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1811 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1812 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1813 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1814 
1815 	/* lifetimes. */
1816 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1817 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1818 
1819 	/* XXX: scope zone ID? */
1820 
1821 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1822 
1823   	/*
1824 	 * Make sure that we do not have this address already.  This should
1825 	 * usually not happen, but we can still see this case, e.g., if we
1826 	 * have manually configured the exact address to be configured.
1827 	 */
1828 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1829 		/* this should be rare enough to make an explicit log */
1830 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1831 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1832 		return (NULL);
1833 	}
1834 
1835 	/*
1836 	 * Allocate ifaddr structure, link into chain, etc.
1837 	 * If we are going to create a new address upon receiving a multicasted
1838 	 * RA, we need to impose a random delay before starting DAD.
1839 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1840 	 */
1841 	updateflags = 0;
1842 	if (mcast)
1843 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1844 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1845 		nd6log((LOG_ERR,
1846 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1847 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1848 		    error));
1849 		return (NULL);	/* ifaddr must not have been allocated. */
1850 	}
1851 
1852 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1853 
1854 	return (ia);		/* this is always non-NULL */
1855 }
1856 
1857 int
1858 in6_tmpifadd(ia0, forcegen, delay)
1859 	const struct in6_ifaddr *ia0; /* corresponding public address */
1860 	int forcegen, delay;
1861 {
1862 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1863 	struct in6_ifaddr *newia, *ia;
1864 	struct in6_aliasreq ifra;
1865 	int i, error;
1866 	int trylimit = 3;	/* XXX: adhoc value */
1867 	int updateflags;
1868 	u_int32_t randid[2];
1869 	time_t vltime0, pltime0;
1870 
1871 	bzero(&ifra, sizeof(ifra));
1872 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1873 	ifra.ifra_addr = ia0->ia_addr;
1874 	/* copy prefix mask */
1875 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1876 	/* clear the old IFID */
1877 	for (i = 0; i < 4; i++) {
1878 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1879 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1880 	}
1881 
1882   again:
1883 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1884 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1885 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1886 		    "random IFID\n"));
1887 		return (EINVAL);
1888 	}
1889 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1890 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1891 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1892 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1893 
1894   	/*
1895 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1896 	 * However, we may still have a chance to see collision, because
1897 	 * there may be a time lag between generation of the ID and generation
1898 	 * of the address.  So, we'll do one more sanity check.
1899 	 */
1900 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1901 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1902 		    &ifra.ifra_addr.sin6_addr)) {
1903   			if (trylimit-- == 0) {
1904 				/*
1905 				 * Give up.  Something strange should have
1906 				 * happened.
1907 				 */
1908 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1909 				    "find a unique random IFID\n"));
1910 				return (EEXIST);
1911 			}
1912 			forcegen = 1;
1913 			goto again;
1914 		}
1915 	}
1916 
1917 	/*
1918 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1919          * public address or TEMP_VALID_LIFETIME.
1920 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1921          * of the public address or TEMP_PREFERRED_LIFETIME -
1922          * DESYNC_FACTOR.
1923 	 */
1924 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1925 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1926 		    (ia0->ia6_lifetime.ia6t_vltime -
1927 		    (time_second - ia0->ia6_updatetime));
1928 		if (vltime0 > ip6_temp_valid_lifetime)
1929 			vltime0 = ip6_temp_valid_lifetime;
1930 	} else
1931 		vltime0 = ip6_temp_valid_lifetime;
1932 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1933 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1934 		    (ia0->ia6_lifetime.ia6t_pltime -
1935 		    (time_second - ia0->ia6_updatetime));
1936 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
1937 			pltime0 = ip6_temp_preferred_lifetime -
1938 			    ip6_desync_factor;
1939 		}
1940 	} else
1941 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
1942 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1943 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1944 
1945 	/*
1946 	 * A temporary address is created only if this calculated Preferred
1947 	 * Lifetime is greater than REGEN_ADVANCE time units.
1948 	 */
1949 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
1950 		return (0);
1951 
1952 	/* XXX: scope zone ID? */
1953 
1954 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1955 
1956 	/* allocate ifaddr structure, link into chain, etc. */
1957 	updateflags = 0;
1958 	if (delay)
1959 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1960 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1961 		return (error);
1962 
1963 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1964 	if (newia == NULL) {	/* XXX: can it happen? */
1965 		nd6log((LOG_ERR,
1966 		    "in6_tmpifadd: ifa update succeeded, but we got "
1967 		    "no ifaddr\n"));
1968 		return (EINVAL); /* XXX */
1969 	}
1970 	newia->ia6_ndpr = ia0->ia6_ndpr;
1971 	newia->ia6_ndpr->ndpr_refcnt++;
1972 
1973 	/*
1974 	 * A newly added address might affect the status of other addresses.
1975 	 * XXX: when the temporary address is generated with a new public
1976 	 * address, the onlink check is redundant.  However, it would be safe
1977 	 * to do the check explicitly everywhere a new address is generated,
1978 	 * and, in fact, we surely need the check when we create a new
1979 	 * temporary address due to deprecation of an old temporary address.
1980 	 */
1981 	pfxlist_onlink_check();
1982 
1983 	return (0);
1984 }
1985 
1986 static int
1987 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1988 {
1989 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1990 		ndpr->ndpr_preferred = 0;
1991 	else
1992 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
1993 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1994 		ndpr->ndpr_expire = 0;
1995 	else
1996 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
1997 
1998 	return 0;
1999 }
2000 
2001 static void
2002 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2003 {
2004 	/* init ia6t_expire */
2005 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2006 		lt6->ia6t_expire = 0;
2007 	else {
2008 		lt6->ia6t_expire = time_second;
2009 		lt6->ia6t_expire += lt6->ia6t_vltime;
2010 	}
2011 
2012 	/* init ia6t_preferred */
2013 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2014 		lt6->ia6t_preferred = 0;
2015 	else {
2016 		lt6->ia6t_preferred = time_second;
2017 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2018 	}
2019 }
2020 
2021 /*
2022  * Delete all the routing table entries that use the specified gateway.
2023  * XXX: this function causes search through all entries of routing table, so
2024  * it shouldn't be called when acting as a router.
2025  */
2026 void
2027 rt6_flush(gateway, ifp)
2028 	struct in6_addr *gateway;
2029 	struct ifnet *ifp;
2030 {
2031 	struct radix_node_head *rnh = rt_tables[AF_INET6];
2032 	int s = splnet();
2033 
2034 	/* We'll care only link-local addresses */
2035 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2036 		splx(s);
2037 		return;
2038 	}
2039 
2040 	RADIX_NODE_HEAD_LOCK(rnh);
2041 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2042 	RADIX_NODE_HEAD_UNLOCK(rnh);
2043 	splx(s);
2044 }
2045 
2046 static int
2047 rt6_deleteroute(rn, arg)
2048 	struct radix_node *rn;
2049 	void *arg;
2050 {
2051 #define SIN6(s)	((struct sockaddr_in6 *)s)
2052 	struct rtentry *rt = (struct rtentry *)rn;
2053 	struct in6_addr *gate = (struct in6_addr *)arg;
2054 
2055 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2056 		return (0);
2057 
2058 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2059 		return (0);
2060 	}
2061 
2062 	/*
2063 	 * Do not delete a static route.
2064 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2065 	 * 'cloned' bit instead?
2066 	 */
2067 	if ((rt->rt_flags & RTF_STATIC) != 0)
2068 		return (0);
2069 
2070 	/*
2071 	 * We delete only host route. This means, in particular, we don't
2072 	 * delete default route.
2073 	 */
2074 	if ((rt->rt_flags & RTF_HOST) == 0)
2075 		return (0);
2076 
2077 	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2078 	    rt_mask(rt), rt->rt_flags, 0));
2079 #undef SIN6
2080 }
2081 
2082 int
2083 nd6_setdefaultiface(ifindex)
2084 	int ifindex;
2085 {
2086 	int error = 0;
2087 
2088 	if (ifindex < 0 || if_index < ifindex)
2089 		return (EINVAL);
2090 	if (ifindex != 0 && !ifnet_byindex(ifindex))
2091 		return (EINVAL);
2092 
2093 	if (nd6_defifindex != ifindex) {
2094 		nd6_defifindex = ifindex;
2095 		if (nd6_defifindex > 0)
2096 			nd6_defifp = ifnet_byindex(nd6_defifindex);
2097 		else
2098 			nd6_defifp = NULL;
2099 
2100 		/*
2101 		 * Our current implementation assumes one-to-one maping between
2102 		 * interfaces and links, so it would be natural to use the
2103 		 * default interface as the default link.
2104 		 */
2105 		scope6_setdefault(nd6_defifp);
2106 	}
2107 
2108 	return (error);
2109 }
2110