xref: /netbsd-src/sys/netinet6/nd6_rtr.c (revision 267197ec1eebfcb9810ea27a89625b6ddf68e3e7)
1 /*	$NetBSD: nd6_rtr.c,v 1.72 2007/12/20 19:53:34 dyoung Exp $	*/
2 /*	$KAME: nd6_rtr.c,v 1.95 2001/02/07 08:09:47 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: nd6_rtr.c,v 1.72 2007/12/20 19:53:34 dyoung Exp $");
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/ioctl.h>
46 #include <sys/syslog.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 static int rtpref(struct nd_defrouter *);
66 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
67 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
68     struct mbuf *, int);
69 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
70 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
71 	struct nd_defrouter *);
72 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
73 static void pfxrtr_del(struct nd_pfxrouter *);
74 static struct nd_pfxrouter *find_pfxlist_reachable_router
75 	(struct nd_prefix *);
76 static void defrouter_delreq(struct nd_defrouter *);
77 static void nd6_rtmsg(int, struct rtentry *);
78 
79 static int in6_init_prefix_ltimes(struct nd_prefix *);
80 static void in6_init_address_ltimes(struct nd_prefix *ndpr,
81 	struct in6_addrlifetime *lt6);
82 
83 static int rt6_deleteroute(struct rtentry *, void *);
84 
85 extern int nd6_recalc_reachtm_interval;
86 
87 static struct ifnet *nd6_defifp;
88 int nd6_defifindex;
89 
90 int ip6_use_tempaddr = 0;
91 
92 int ip6_desync_factor;
93 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
94 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
95 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
96 
97 /* RTPREF_MEDIUM has to be 0! */
98 #define RTPREF_HIGH	1
99 #define RTPREF_MEDIUM	0
100 #define RTPREF_LOW	(-1)
101 #define RTPREF_RESERVED	(-2)
102 #define RTPREF_INVALID	(-3)	/* internal */
103 
104 /*
105  * Receive Router Solicitation Message - just for routers.
106  * Router solicitation/advertisement is mostly managed by userland program
107  * (rtadvd) so here we have no function like nd6_ra_output().
108  *
109  * Based on RFC 2461
110  */
111 void
112 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
113 {
114 	struct ifnet *ifp = m->m_pkthdr.rcvif;
115 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
116 	struct nd_router_solicit *nd_rs;
117 	struct in6_addr saddr6 = ip6->ip6_src;
118 	char *lladdr = NULL;
119 	int lladdrlen = 0;
120 	union nd_opts ndopts;
121 
122 	/* If I'm not a router, ignore it. */
123 	if (ip6_accept_rtadv != 0 || !ip6_forwarding)
124 		goto freeit;
125 
126 	/* Sanity checks */
127 	if (ip6->ip6_hlim != 255) {
128 		nd6log((LOG_ERR,
129 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
130 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
131 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
132 		goto bad;
133 	}
134 
135 	/*
136 	 * Don't update the neighbor cache, if src = ::.
137 	 * This indicates that the src has no IP address assigned yet.
138 	 */
139 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
140 		goto freeit;
141 
142 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
143 	if (nd_rs == NULL) {
144 		icmp6stat.icp6s_tooshort++;
145 		return;
146 	}
147 
148 	icmp6len -= sizeof(*nd_rs);
149 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
150 	if (nd6_options(&ndopts) < 0) {
151 		nd6log((LOG_INFO,
152 		    "nd6_rs_input: invalid ND option, ignored\n"));
153 		/* nd6_options have incremented stats */
154 		goto freeit;
155 	}
156 
157 	if (ndopts.nd_opts_src_lladdr) {
158 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
159 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
160 	}
161 
162 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
163 		nd6log((LOG_INFO,
164 		    "nd6_rs_input: lladdrlen mismatch for %s "
165 		    "(if %d, RS packet %d)\n",
166 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
167 		goto bad;
168 	}
169 
170 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
171 
172  freeit:
173 	m_freem(m);
174 	return;
175 
176  bad:
177 	icmp6stat.icp6s_badrs++;
178 	m_freem(m);
179 }
180 
181 /*
182  * Receive Router Advertisement Message.
183  *
184  * Based on RFC 2461
185  * TODO: on-link bit on prefix information
186  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
187  */
188 void
189 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
190 {
191 	struct ifnet *ifp = m->m_pkthdr.rcvif;
192 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
193 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
194 	struct nd_router_advert *nd_ra;
195 	struct in6_addr saddr6 = ip6->ip6_src;
196 #if 0
197 	struct in6_addr daddr6 = ip6->ip6_dst;
198 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
199 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
200 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
201 #endif
202 	int mcast = 0;
203 	union nd_opts ndopts;
204 	struct nd_defrouter *dr;
205 
206 	/*
207 	 * We only accept RAs only when
208 	 * the system-wide variable allows the acceptance, and
209 	 * per-interface variable allows RAs on the receiving interface.
210 	 */
211 	if (ip6_accept_rtadv == 0)
212 		goto freeit;
213 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
214 		goto freeit;
215 
216 	if (ip6->ip6_hlim != 255) {
217 		nd6log((LOG_ERR,
218 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
219 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
220 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
221 		goto bad;
222 	}
223 
224 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
225 		nd6log((LOG_ERR,
226 		    "nd6_ra_input: src %s is not link-local\n",
227 		    ip6_sprintf(&saddr6)));
228 		goto bad;
229 	}
230 
231 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
232 	if (nd_ra == NULL) {
233 		icmp6stat.icp6s_tooshort++;
234 		return;
235 	}
236 
237 	icmp6len -= sizeof(*nd_ra);
238 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
239 	if (nd6_options(&ndopts) < 0) {
240 		nd6log((LOG_INFO,
241 		    "nd6_ra_input: invalid ND option, ignored\n"));
242 		/* nd6_options have incremented stats */
243 		goto freeit;
244 	}
245 
246     {
247 	struct nd_defrouter drtr;
248 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
249 
250 	/* remember if this is a multicasted advertisement */
251 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
252 		mcast = 1;
253 
254 	memset(&drtr, 0, sizeof(drtr));
255 	drtr.rtaddr = saddr6;
256 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
257 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
258 	drtr.expire = time_second + drtr.rtlifetime;
259 	drtr.ifp = ifp;
260 	/* unspecified or not? (RFC 2461 6.3.4) */
261 	if (advreachable) {
262 		NTOHL(advreachable);
263 		if (advreachable <= MAX_REACHABLE_TIME &&
264 		    ndi->basereachable != advreachable) {
265 			ndi->basereachable = advreachable;
266 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
267 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
268 		}
269 	}
270 	if (nd_ra->nd_ra_retransmit)
271 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
272 	if (nd_ra->nd_ra_curhoplimit)
273 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
274 	dr = defrtrlist_update(&drtr);
275     }
276 
277 	/*
278 	 * prefix
279 	 */
280 	if (ndopts.nd_opts_pi) {
281 		struct nd_opt_hdr *pt;
282 		struct nd_opt_prefix_info *pi = NULL;
283 		struct nd_prefixctl pr;
284 
285 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
286 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
287 		     pt = (struct nd_opt_hdr *)((char *)pt +
288 						(pt->nd_opt_len << 3))) {
289 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
290 				continue;
291 			pi = (struct nd_opt_prefix_info *)pt;
292 
293 			if (pi->nd_opt_pi_len != 4) {
294 				nd6log((LOG_INFO,
295 				    "nd6_ra_input: invalid option "
296 				    "len %d for prefix information option, "
297 				    "ignored\n", pi->nd_opt_pi_len));
298 				continue;
299 			}
300 
301 			if (128 < pi->nd_opt_pi_prefix_len) {
302 				nd6log((LOG_INFO,
303 				    "nd6_ra_input: invalid prefix "
304 				    "len %d for prefix information option, "
305 				    "ignored\n", pi->nd_opt_pi_prefix_len));
306 				continue;
307 			}
308 
309 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
310 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
311 				nd6log((LOG_INFO,
312 				    "nd6_ra_input: invalid prefix "
313 				    "%s, ignored\n",
314 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
315 				continue;
316 			}
317 
318 			bzero(&pr, sizeof(pr));
319 			sockaddr_in6_init(&pr.ndpr_prefix,
320 			    &pi->nd_opt_pi_prefix, 0, 0, 0);
321 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
322 
323 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
324 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
325 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
326 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
327 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
328 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
329 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
330 
331 			(void)prelist_update(&pr, dr, m, mcast);
332 		}
333 	}
334 
335 	/*
336 	 * MTU
337 	 */
338 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
339 		u_long mtu;
340 		u_long maxmtu;
341 
342 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
343 
344 		/* lower bound */
345 		if (mtu < IPV6_MMTU) {
346 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
347 			    "mtu=%lu sent from %s, ignoring\n",
348 			    mtu, ip6_sprintf(&ip6->ip6_src)));
349 			goto skip;
350 		}
351 
352 		/* upper bound */
353 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
354 		    ? ndi->maxmtu : ifp->if_mtu;
355 		if (mtu <= maxmtu) {
356 			int change = (ndi->linkmtu != mtu);
357 
358 			ndi->linkmtu = mtu;
359 			if (change) /* in6_maxmtu may change */
360 				in6_setmaxmtu();
361 		} else {
362 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
363 			    "mtu=%lu sent from %s; "
364 			    "exceeds maxmtu %lu, ignoring\n",
365 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
366 		}
367 	}
368 
369  skip:
370 
371 	/*
372 	 * Source link layer address
373 	 */
374     {
375 	char *lladdr = NULL;
376 	int lladdrlen = 0;
377 
378 	if (ndopts.nd_opts_src_lladdr) {
379 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
380 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
381 	}
382 
383 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
384 		nd6log((LOG_INFO,
385 		    "nd6_ra_input: lladdrlen mismatch for %s "
386 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
387 		    ifp->if_addrlen, lladdrlen - 2));
388 		goto bad;
389 	}
390 
391 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
392 
393 	/*
394 	 * Installing a link-layer address might change the state of the
395 	 * router's neighbor cache, which might also affect our on-link
396 	 * detection of adveritsed prefixes.
397 	 */
398 	pfxlist_onlink_check();
399     }
400 
401  freeit:
402 	m_freem(m);
403 	return;
404 
405  bad:
406 	icmp6stat.icp6s_badra++;
407 	m_freem(m);
408 }
409 
410 /*
411  * default router list processing sub routines
412  */
413 
414 /* tell the change to user processes watching the routing socket. */
415 static void
416 nd6_rtmsg(int cmd, struct rtentry *rt)
417 {
418 	struct rt_addrinfo info;
419 
420 	bzero((void *)&info, sizeof(info));
421 	info.rti_info[RTAX_DST] = rt_getkey(rt);
422 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
423 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
424 	if (rt->rt_ifp) {
425 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
426 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
427 	}
428 
429 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
430 }
431 
432 void
433 defrouter_addreq(struct nd_defrouter *new)
434 {
435 	struct sockaddr_in6 def, mask, gate;
436 	struct rtentry *newrt = NULL;
437 	int s;
438 	int error;
439 
440 	memset(&def, 0, sizeof(def));
441 	memset(&mask, 0, sizeof(mask));
442 	memset(&gate, 0,sizeof(gate)); /* for safety */
443 
444 	def.sin6_len = mask.sin6_len = gate.sin6_len =
445 	    sizeof(struct sockaddr_in6);
446 	def.sin6_family = mask.sin6_family = gate.sin6_family = AF_INET6;
447 	gate.sin6_addr = new->rtaddr;
448 #ifndef SCOPEDROUTING
449 	gate.sin6_scope_id = 0;	/* XXX */
450 #endif
451 
452 	s = splsoftnet();
453 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
454 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
455 	    RTF_GATEWAY, &newrt);
456 	if (newrt) {
457 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
458 		newrt->rt_refcnt--;
459 	}
460 	if (error == 0)
461 		new->installed = 1;
462 	splx(s);
463 	return;
464 }
465 
466 struct nd_defrouter *
467 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
468 {
469 	struct nd_defrouter *dr;
470 
471 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
472 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
473 			break;
474 	}
475 
476 	return dr;		/* search failed */
477 }
478 
479 void
480 defrtrlist_del(struct nd_defrouter *dr)
481 {
482 	struct nd_defrouter *deldr = NULL;
483 	struct nd_prefix *pr;
484 
485 	/*
486 	 * Flush all the routing table entries that use the router
487 	 * as a next hop.
488 	 */
489 	if (!ip6_forwarding && ip6_accept_rtadv) /* XXX: better condition? */
490 		rt6_flush(&dr->rtaddr, dr->ifp);
491 
492 	if (dr->installed) {
493 		deldr = dr;
494 		defrouter_delreq(dr);
495 	}
496 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
497 
498 	/*
499 	 * Also delete all the pointers to the router in each prefix lists.
500 	 */
501 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
502 		struct nd_pfxrouter *pfxrtr;
503 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
504 			pfxrtr_del(pfxrtr);
505 	}
506 	pfxlist_onlink_check();
507 
508 	/*
509 	 * If the router is the primary one, choose a new one.
510 	 * Note that defrouter_select() will remove the current gateway
511 	 * from the routing table.
512 	 */
513 	if (deldr)
514 		defrouter_select();
515 
516 	free(dr, M_IP6NDP);
517 }
518 
519 /*
520  * Remove the default route for a given router.
521  * This is just a subroutine function for defrouter_select(), and should
522  * not be called from anywhere else.
523  */
524 static void
525 defrouter_delreq(struct nd_defrouter *dr)
526 {
527 	struct sockaddr_in6 def, mask, gw;
528 	struct rtentry *oldrt = NULL;
529 
530 #ifdef DIAGNOSTIC
531 	if (dr == NULL)
532 		panic("dr == NULL in defrouter_delreq");
533 #endif
534 
535 	bzero(&def, sizeof(def));
536 	bzero(&mask, sizeof(mask));
537 	bzero(&gw, sizeof(gw));	/* for safety */
538 
539 	def.sin6_len = mask.sin6_len = gw.sin6_len =
540 	    sizeof(struct sockaddr_in6);
541 	def.sin6_family = mask.sin6_family = gw.sin6_family = AF_INET6;
542 	gw.sin6_addr = dr->rtaddr;
543 #ifndef SCOPEDROUTING
544 	gw.sin6_scope_id = 0;	/* XXX */
545 #endif
546 
547 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
548 	    (struct sockaddr *)&gw,
549 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
550 	if (oldrt) {
551 		nd6_rtmsg(RTM_DELETE, oldrt);
552 		if (oldrt->rt_refcnt <= 0) {
553 			/*
554 			 * XXX: borrowed from the RTM_DELETE case of
555 			 * rtrequest().
556 			 */
557 			oldrt->rt_refcnt++;
558 			rtfree(oldrt);
559 		}
560 	}
561 
562 	dr->installed = 0;
563 }
564 
565 /*
566  * remove all default routes from default router list
567  */
568 void
569 defrouter_reset()
570 {
571 	struct nd_defrouter *dr;
572 
573 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
574 	     dr = TAILQ_NEXT(dr, dr_entry))
575 		defrouter_delreq(dr);
576 
577 	/*
578 	 * XXX should we also nuke any default routers in the kernel, by
579 	 * going through them by rtalloc1()?
580 	 */
581 }
582 
583 /*
584  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
585  * draft-ietf-ipngwg-router-selection:
586  * 1) Routers that are reachable or probably reachable should be preferred.
587  *    If we have more than one (probably) reachable router, prefer ones
588  *    with the highest router preference.
589  * 2) When no routers on the list are known to be reachable or
590  *    probably reachable, routers SHOULD be selected in a round-robin
591  *    fashion, regardless of router preference values.
592  * 3) If the Default Router List is empty, assume that all
593  *    destinations are on-link.
594  *
595  * We assume nd_defrouter is sorted by router preference value.
596  * Since the code below covers both with and without router preference cases,
597  * we do not need to classify the cases by ifdef.
598  *
599  * At this moment, we do not try to install more than one default router,
600  * even when the multipath routing is available, because we're not sure about
601  * the benefits for stub hosts comparing to the risk of making the code
602  * complicated and the possibility of introducing bugs.
603  */
604 void
605 defrouter_select()
606 {
607 	int s = splsoftnet();
608 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
609 	struct rtentry *rt = NULL;
610 	struct llinfo_nd6 *ln = NULL;
611 
612 	/*
613 	 * This function should be called only when acting as an autoconfigured
614 	 * host.  Although the remaining part of this function is not effective
615 	 * if the node is not an autoconfigured host, we explicitly exclude
616 	 * such cases here for safety.
617 	 */
618 	if (ip6_forwarding || !ip6_accept_rtadv) {
619 		nd6log((LOG_WARNING,
620 		    "defrouter_select: called unexpectedly (forwarding=%d, "
621 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
622 		splx(s);
623 		return;
624 	}
625 
626 	/*
627 	 * Let's handle easy case (3) first:
628 	 * If default router list is empty, there's nothing to be done.
629 	 */
630 	if (!TAILQ_FIRST(&nd_defrouter)) {
631 		splx(s);
632 		return;
633 	}
634 
635 	/*
636 	 * Search for a (probably) reachable router from the list.
637 	 * We just pick up the first reachable one (if any), assuming that
638 	 * the ordering rule of the list described in defrtrlist_update().
639 	 */
640 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
641 	     dr = TAILQ_NEXT(dr, dr_entry)) {
642 		if (selected_dr == NULL &&
643 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) != NULL &&
644 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) != NULL &&
645 		    ND6_IS_LLINFO_PROBREACH(ln)) {
646 			selected_dr = dr;
647 		}
648 
649 		if (dr->installed && !installed_dr)
650 			installed_dr = dr;
651 		else if (dr->installed && installed_dr) {
652 			/* this should not happen.  warn for diagnosis. */
653 			log(LOG_ERR, "defrouter_select: more than one router"
654 			    " is installed\n");
655 		}
656 	}
657 	/*
658 	 * If none of the default routers was found to be reachable,
659 	 * round-robin the list regardless of preference.
660 	 * Otherwise, if we have an installed router, check if the selected
661 	 * (reachable) router should really be preferred to the installed one.
662 	 * We only prefer the new router when the old one is not reachable
663 	 * or when the new one has a really higher preference value.
664 	 */
665 	if (selected_dr == NULL) {
666 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
667 			selected_dr = TAILQ_FIRST(&nd_defrouter);
668 		else
669 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
670 	} else if (installed_dr &&
671 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
672 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
673 	    ND6_IS_LLINFO_PROBREACH(ln) &&
674 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
675 		selected_dr = installed_dr;
676 	}
677 
678 	/*
679 	 * If the selected router is different than the installed one,
680 	 * remove the installed router and install the selected one.
681 	 * Note that the selected router is never NULL here.
682 	 */
683 	if (installed_dr != selected_dr) {
684 		if (installed_dr)
685 			defrouter_delreq(installed_dr);
686 		defrouter_addreq(selected_dr);
687 	}
688 
689 	splx(s);
690 	return;
691 }
692 
693 /*
694  * for default router selection
695  * regards router-preference field as a 2-bit signed integer
696  */
697 static int
698 rtpref(struct nd_defrouter *dr)
699 {
700 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
701 	case ND_RA_FLAG_RTPREF_HIGH:
702 		return (RTPREF_HIGH);
703 	case ND_RA_FLAG_RTPREF_MEDIUM:
704 	case ND_RA_FLAG_RTPREF_RSV:
705 		return (RTPREF_MEDIUM);
706 	case ND_RA_FLAG_RTPREF_LOW:
707 		return (RTPREF_LOW);
708 	default:
709 		/*
710 		 * This case should never happen.  If it did, it would mean a
711 		 * serious bug of kernel internal.  We thus always bark here.
712 		 * Or, can we even panic?
713 		 */
714 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
715 		return (RTPREF_INVALID);
716 	}
717 	/* NOTREACHED */
718 }
719 
720 static struct nd_defrouter *
721 defrtrlist_update(struct nd_defrouter *new)
722 {
723 	struct nd_defrouter *dr, *n;
724 	int s = splsoftnet();
725 
726 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
727 		/* entry exists */
728 		if (new->rtlifetime == 0) {
729 			defrtrlist_del(dr);
730 			dr = NULL;
731 		} else {
732 			int oldpref = rtpref(dr);
733 
734 			/* override */
735 			dr->flags = new->flags; /* xxx flag check */
736 			dr->rtlifetime = new->rtlifetime;
737 			dr->expire = new->expire;
738 
739 			/*
740 			 * If the preference does not change, there's no need
741 			 * to sort the entries.
742 			 */
743 			if (rtpref(new) == oldpref) {
744 				splx(s);
745 				return (dr);
746 			}
747 
748 			/*
749 			 * preferred router may be changed, so relocate
750 			 * this router.
751 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
752 			 * However, since defrtrlist_del() has many side
753 			 * effects, we intentionally do so here.
754 			 * defrouter_select() below will handle routing
755 			 * changes later.
756 			 */
757 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
758 			n = dr;
759 			goto insert;
760 		}
761 		splx(s);
762 		return (dr);
763 	}
764 
765 	/* entry does not exist */
766 	if (new->rtlifetime == 0) {
767 		splx(s);
768 		return (NULL);
769 	}
770 
771 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
772 	if (n == NULL) {
773 		splx(s);
774 		return (NULL);
775 	}
776 	bzero(n, sizeof(*n));
777 	*n = *new;
778 
779 insert:
780 	/*
781 	 * Insert the new router in the Default Router List;
782 	 * The Default Router List should be in the descending order
783 	 * of router-preferece.  Routers with the same preference are
784 	 * sorted in the arriving time order.
785 	 */
786 
787 	/* insert at the end of the group */
788 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
789 	     dr = TAILQ_NEXT(dr, dr_entry)) {
790 		if (rtpref(n) > rtpref(dr))
791 			break;
792 	}
793 	if (dr)
794 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
795 	else
796 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
797 
798 	defrouter_select();
799 
800 	splx(s);
801 
802 	return (n);
803 }
804 
805 static struct nd_pfxrouter *
806 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
807 {
808 	struct nd_pfxrouter *search;
809 
810 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
811 		if (search->router == dr)
812 			break;
813 	}
814 
815 	return (search);
816 }
817 
818 static void
819 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
820 {
821 	struct nd_pfxrouter *new;
822 
823 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
824 	if (new == NULL)
825 		return;
826 	bzero(new, sizeof(*new));
827 	new->router = dr;
828 
829 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
830 
831 	pfxlist_onlink_check();
832 }
833 
834 static void
835 pfxrtr_del(struct nd_pfxrouter *pfr)
836 {
837 	LIST_REMOVE(pfr, pfr_entry);
838 	free(pfr, M_IP6NDP);
839 }
840 
841 struct nd_prefix *
842 nd6_prefix_lookup(struct nd_prefixctl *key)
843 {
844 	struct nd_prefix *search;
845 
846 	LIST_FOREACH(search, &nd_prefix, ndpr_entry) {
847 		if (key->ndpr_ifp == search->ndpr_ifp &&
848 		    key->ndpr_plen == search->ndpr_plen &&
849 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
850 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
851 			break;
852 		}
853 	}
854 
855 	return (search);
856 }
857 
858 int
859 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
860 	struct nd_prefix **newp)
861 {
862 	struct nd_prefix *new = NULL;
863 	int i, s;
864 	int error;
865 
866 	error = 0;
867 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
868 	if (new == NULL)
869 		return ENOMEM;
870 	bzero(new, sizeof(*new));
871 	new->ndpr_ifp = pr->ndpr_ifp;
872 	new->ndpr_prefix = pr->ndpr_prefix;
873 	new->ndpr_plen = pr->ndpr_plen;
874 	new->ndpr_vltime = pr->ndpr_vltime;
875 	new->ndpr_pltime = pr->ndpr_pltime;
876 	new->ndpr_flags = pr->ndpr_flags;
877 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
878 		free(new, M_IP6NDP);
879 		return(error);
880 	}
881 	new->ndpr_lastupdate = time_second;
882 	if (newp != NULL)
883 		*newp = new;
884 
885 	/* initialization */
886 	LIST_INIT(&new->ndpr_advrtrs);
887 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
888 	/* make prefix in the canonical form */
889 	for (i = 0; i < 4; i++)
890 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
891 		    new->ndpr_mask.s6_addr32[i];
892 
893 	s = splsoftnet();
894 	/* link ndpr_entry to nd_prefix list */
895 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
896 	splx(s);
897 
898 	/* ND_OPT_PI_FLAG_ONLINK processing */
899 	if (new->ndpr_raf_onlink) {
900 		int e;
901 
902 		if ((e = nd6_prefix_onlink(new)) != 0) {
903 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
904 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
905 			    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
906 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
907 			/* proceed anyway. XXX: is it correct? */
908 		}
909 	}
910 
911 	if (dr)
912 		pfxrtr_add(new, dr);
913 
914 	return 0;
915 }
916 
917 void
918 prelist_remove(struct nd_prefix *pr)
919 {
920 	struct nd_pfxrouter *pfr, *next;
921 	int e, s;
922 
923 	/* make sure to invalidate the prefix until it is really freed. */
924 	pr->ndpr_vltime = 0;
925 	pr->ndpr_pltime = 0;
926 #if 0
927 	/*
928 	 * Though these flags are now meaningless, we'd rather keep the value
929 	 * not to confuse users when executing "ndp -p".
930 	 */
931 	pr->ndpr_raf_onlink = 0;
932 	pr->ndpr_raf_auto = 0;
933 #endif
934 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
935 	    (e = nd6_prefix_offlink(pr)) != 0) {
936 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
937 		    "on %s, errno=%d\n",
938 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
939 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
940 		/* what should we do? */
941 	}
942 
943 	if (pr->ndpr_refcnt > 0)
944 		return;		/* notice here? */
945 
946 	s = splsoftnet();
947 	/* unlink ndpr_entry from nd_prefix list */
948 	LIST_REMOVE(pr, ndpr_entry);
949 
950 	/* free list of routers that adversed the prefix */
951 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
952 		next = LIST_NEXT(pfr, pfr_entry);
953 
954 		free(pfr, M_IP6NDP);
955 	}
956 	splx(s);
957 
958 	free(pr, M_IP6NDP);
959 
960 	pfxlist_onlink_check();
961 }
962 
963 static int
964 prelist_update(struct nd_prefixctl *new,
965 	struct nd_defrouter *dr, /* may be NULL */
966 	struct mbuf *m,
967 	int mcast)
968 {
969 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
970 	struct ifaddr *ifa;
971 	struct ifnet *ifp = new->ndpr_ifp;
972 	struct nd_prefix *pr;
973 	int s = splsoftnet();
974 	int error = 0;
975 	int newprefix = 0;
976 	int auth;
977 	struct in6_addrlifetime lt6_tmp;
978 
979 	auth = 0;
980 	if (m) {
981 		/*
982 		 * Authenticity for NA consists authentication for
983 		 * both IP header and IP datagrams, doesn't it ?
984 		 */
985 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
986 		auth = (m->m_flags & M_AUTHIPHDR
987 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
988 #endif
989 	}
990 
991 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
992 		/*
993 		 * nd6_prefix_lookup() ensures that pr and new have the same
994 		 * prefix on a same interface.
995 		 */
996 
997 		/*
998 		 * Update prefix information.  Note that the on-link (L) bit
999 		 * and the autonomous (A) bit should NOT be changed from 1
1000 		 * to 0.
1001 		 */
1002 		if (new->ndpr_raf_onlink == 1)
1003 			pr->ndpr_raf_onlink = 1;
1004 		if (new->ndpr_raf_auto == 1)
1005 			pr->ndpr_raf_auto = 1;
1006 		if (new->ndpr_raf_onlink) {
1007 			pr->ndpr_vltime = new->ndpr_vltime;
1008 			pr->ndpr_pltime = new->ndpr_pltime;
1009 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1010 			pr->ndpr_lastupdate = time_second;
1011 		}
1012 
1013 		if (new->ndpr_raf_onlink &&
1014 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1015 			int e;
1016 
1017 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1018 				nd6log((LOG_ERR,
1019 				    "prelist_update: failed to make "
1020 				    "the prefix %s/%d on-link on %s "
1021 				    "(errno=%d)\n",
1022 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1023 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1024 				/* proceed anyway. XXX: is it correct? */
1025 			}
1026 		}
1027 
1028 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1029 			pfxrtr_add(pr, dr);
1030 	} else {
1031 		struct nd_prefix *newpr = NULL;
1032 
1033 		newprefix = 1;
1034 
1035 		if (new->ndpr_vltime == 0)
1036 			goto end;
1037 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1038 			goto end;
1039 
1040 		error = nd6_prelist_add(new, dr, &newpr);
1041 		if (error != 0 || newpr == NULL) {
1042 			nd6log((LOG_NOTICE, "prelist_update: "
1043 			    "nd6_prelist_add failed for %s/%d on %s "
1044 			    "errno=%d, returnpr=%p\n",
1045 			    ip6_sprintf(&new->ndpr_prefix.sin6_addr),
1046 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1047 			    error, newpr));
1048 			goto end; /* we should just give up in this case. */
1049 		}
1050 
1051 		/*
1052 		 * XXX: from the ND point of view, we can ignore a prefix
1053 		 * with the on-link bit being zero.  However, we need a
1054 		 * prefix structure for references from autoconfigured
1055 		 * addresses.  Thus, we explicitly make sure that the prefix
1056 		 * itself expires now.
1057 		 */
1058 		if (newpr->ndpr_raf_onlink == 0) {
1059 			newpr->ndpr_vltime = 0;
1060 			newpr->ndpr_pltime = 0;
1061 			in6_init_prefix_ltimes(newpr);
1062 		}
1063 
1064 		pr = newpr;
1065 	}
1066 
1067 	/*
1068 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1069 	 * Note that pr must be non NULL at this point.
1070 	 */
1071 
1072 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1073 	if (!new->ndpr_raf_auto)
1074 		goto end;
1075 
1076 	/*
1077 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1078 	 * nd6_ra_input.
1079 	 */
1080 
1081 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1082 	if (new->ndpr_pltime > new->ndpr_vltime) {
1083 		error = EINVAL;	/* XXX: won't be used */
1084 		goto end;
1085 	}
1086 
1087 	/*
1088 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1089 	 * an address configured by stateless autoconfiguration already in the
1090 	 * list of addresses associated with the interface, and the Valid
1091 	 * Lifetime is not 0, form an address.  We first check if we have
1092 	 * a matching prefix.
1093 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1094 	 * consider autoconfigured addresses while RFC2462 simply said
1095 	 * "address".
1096 	 */
1097 	IFADDR_FOREACH(ifa, ifp) {
1098 		struct in6_ifaddr *ifa6;
1099 		u_int32_t remaininglifetime;
1100 
1101 		if (ifa->ifa_addr->sa_family != AF_INET6)
1102 			continue;
1103 
1104 		ifa6 = (struct in6_ifaddr *)ifa;
1105 
1106 		/*
1107 		 * We only consider autoconfigured addresses as per rfc2462bis.
1108 		 */
1109 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1110 			continue;
1111 
1112 		/*
1113 		 * Spec is not clear here, but I believe we should concentrate
1114 		 * on unicast (i.e. not anycast) addresses.
1115 		 * XXX: other ia6_flags? detached or duplicated?
1116 		 */
1117 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1118 			continue;
1119 
1120 		/*
1121 		 * Ignore the address if it is not associated with a prefix
1122 		 * or is associated with a prefix that is different from this
1123 		 * one.  (pr is never NULL here)
1124 		 */
1125 		if (ifa6->ia6_ndpr != pr)
1126 			continue;
1127 
1128 		if (ia6_match == NULL) /* remember the first one */
1129 			ia6_match = ifa6;
1130 
1131 		/*
1132 		 * An already autoconfigured address matched.  Now that we
1133 		 * are sure there is at least one matched address, we can
1134 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1135 		 * "two hours" rule and the privacy extension.
1136 		 * We apply some clarifications in rfc2462bis:
1137 		 * - use remaininglifetime instead of storedlifetime as a
1138 		 *   variable name
1139 		 * - remove the dead code in the "two-hour" rule
1140 		 */
1141 #define TWOHOUR		(120*60)
1142 		lt6_tmp = ifa6->ia6_lifetime;
1143 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1144 			remaininglifetime = ND6_INFINITE_LIFETIME;
1145 		else if (time_second - ifa6->ia6_updatetime >
1146 			 lt6_tmp.ia6t_vltime) {
1147 			/*
1148 			 * The case of "invalid" address.  We should usually
1149 			 * not see this case.
1150 			 */
1151 			remaininglifetime = 0;
1152 		} else
1153 			remaininglifetime = lt6_tmp.ia6t_vltime -
1154 			    (time_second - ifa6->ia6_updatetime);
1155 
1156 		/* when not updating, keep the current stored lifetime. */
1157 		lt6_tmp.ia6t_vltime = remaininglifetime;
1158 
1159 		if (TWOHOUR < new->ndpr_vltime ||
1160 		    remaininglifetime < new->ndpr_vltime) {
1161 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1162 		} else if (remaininglifetime <= TWOHOUR) {
1163 			if (auth)
1164 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1165 		} else {
1166 			/*
1167 			 * new->ndpr_vltime <= TWOHOUR &&
1168 			 * TWOHOUR < remaininglifetime
1169 			 */
1170 			lt6_tmp.ia6t_vltime = TWOHOUR;
1171 		}
1172 
1173 		/* The 2 hour rule is not imposed for preferred lifetime. */
1174 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1175 
1176 		in6_init_address_ltimes(pr, &lt6_tmp);
1177 
1178 		/*
1179 		 * We need to treat lifetimes for temporary addresses
1180 		 * differently, according to
1181 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1182 		 * we only update the lifetimes when they are in the maximum
1183 		 * intervals.
1184 		 */
1185 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1186 			u_int32_t maxvltime, maxpltime;
1187 
1188 			if (ip6_temp_valid_lifetime >
1189 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1190 			    ip6_desync_factor)) {
1191 				maxvltime = ip6_temp_valid_lifetime -
1192 				    (time_second - ifa6->ia6_createtime) -
1193 				    ip6_desync_factor;
1194 			} else
1195 				maxvltime = 0;
1196 			if (ip6_temp_preferred_lifetime >
1197 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1198 			    ip6_desync_factor)) {
1199 				maxpltime = ip6_temp_preferred_lifetime -
1200 				    (time_second - ifa6->ia6_createtime) -
1201 				    ip6_desync_factor;
1202 			} else
1203 				maxpltime = 0;
1204 
1205 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1206 			    lt6_tmp.ia6t_vltime > maxvltime) {
1207 				lt6_tmp.ia6t_vltime = maxvltime;
1208 			}
1209 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1210 			    lt6_tmp.ia6t_pltime > maxpltime) {
1211 				lt6_tmp.ia6t_pltime = maxpltime;
1212 			}
1213 		}
1214 
1215 		ifa6->ia6_lifetime = lt6_tmp;
1216 		ifa6->ia6_updatetime = time_second;
1217 	}
1218 	if (ia6_match == NULL && new->ndpr_vltime) {
1219 		int ifidlen;
1220 
1221 		/*
1222 		 * 5.5.3 (d) (continued)
1223 		 * No address matched and the valid lifetime is non-zero.
1224 		 * Create a new address.
1225 		 */
1226 
1227 		/*
1228 		 * Prefix Length check:
1229 		 * If the sum of the prefix length and interface identifier
1230 		 * length does not equal 128 bits, the Prefix Information
1231 		 * option MUST be ignored.  The length of the interface
1232 		 * identifier is defined in a separate link-type specific
1233 		 * document.
1234 		 */
1235 		ifidlen = in6_if2idlen(ifp);
1236 		if (ifidlen < 0) {
1237 			/* this should not happen, so we always log it. */
1238 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1239 			    if_name(ifp));
1240 			goto end;
1241 		}
1242 		if (ifidlen + pr->ndpr_plen != 128) {
1243 			nd6log((LOG_INFO,
1244 			    "prelist_update: invalid prefixlen "
1245 			    "%d for %s, ignored\n",
1246 			    pr->ndpr_plen, if_name(ifp)));
1247 			goto end;
1248 		}
1249 
1250 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1251 			/*
1252 			 * note that we should use pr (not new) for reference.
1253 			 */
1254 			pr->ndpr_refcnt++;
1255 			ia6->ia6_ndpr = pr;
1256 
1257 			/*
1258 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
1259 			 * When a new public address is created as described
1260 			 * in RFC2462, also create a new temporary address.
1261 			 *
1262 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
1263 			 * When an interface connects to a new link, a new
1264 			 * randomized interface identifier should be generated
1265 			 * immediately together with a new set of temporary
1266 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1267 			 * in6_tmpifadd().
1268 			 */
1269 			if (ip6_use_tempaddr) {
1270 				int e;
1271 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1272 					nd6log((LOG_NOTICE, "prelist_update: "
1273 					    "failed to create a temporary "
1274 					    "address, errno=%d\n",
1275 					    e));
1276 				}
1277 			}
1278 
1279 			/*
1280 			 * A newly added address might affect the status
1281 			 * of other addresses, so we check and update it.
1282 			 * XXX: what if address duplication happens?
1283 			 */
1284 			pfxlist_onlink_check();
1285 		} else {
1286 			/* just set an error. do not bark here. */
1287 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1288 		}
1289 	}
1290 
1291  end:
1292 	splx(s);
1293 	return error;
1294 }
1295 
1296 /*
1297  * A supplement function used in the on-link detection below;
1298  * detect if a given prefix has a (probably) reachable advertising router.
1299  * XXX: lengthy function name...
1300  */
1301 static struct nd_pfxrouter *
1302 find_pfxlist_reachable_router(struct nd_prefix *pr)
1303 {
1304 	struct nd_pfxrouter *pfxrtr;
1305 	struct rtentry *rt;
1306 	struct llinfo_nd6 *ln;
1307 
1308 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1309 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1310 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1311 		    pfxrtr->router->ifp)) &&
1312 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1313 		    ND6_IS_LLINFO_PROBREACH(ln))
1314 			break;	/* found */
1315 	}
1316 
1317 	return (pfxrtr);
1318 }
1319 
1320 /*
1321  * Check if each prefix in the prefix list has at least one available router
1322  * that advertised the prefix (a router is "available" if its neighbor cache
1323  * entry is reachable or probably reachable).
1324  * If the check fails, the prefix may be off-link, because, for example,
1325  * we have moved from the network but the lifetime of the prefix has not
1326  * expired yet.  So we should not use the prefix if there is another prefix
1327  * that has an available router.
1328  * But, if there is no prefix that has an available router, we still regards
1329  * all the prefixes as on-link.  This is because we can't tell if all the
1330  * routers are simply dead or if we really moved from the network and there
1331  * is no router around us.
1332  */
1333 void
1334 pfxlist_onlink_check()
1335 {
1336 	struct nd_prefix *pr;
1337 	struct in6_ifaddr *ifa;
1338 	struct nd_defrouter *dr;
1339 	struct nd_pfxrouter *pfxrtr = NULL;
1340 
1341 	/*
1342 	 * Check if there is a prefix that has a reachable advertising
1343 	 * router.
1344 	 */
1345 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1346 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1347 			break;
1348 	}
1349 	/*
1350 	 * If we have no such prefix, check whether we still have a router
1351 	 * that does not advertise any prefixes.
1352 	 */
1353 	if (pr == NULL) {
1354 		TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1355 			struct nd_prefix *pr0;
1356 
1357 			LIST_FOREACH(pr0, &nd_prefix, ndpr_entry) {
1358 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1359 					break;
1360 			}
1361 			if (pfxrtr)
1362 				break;
1363 		}
1364 	}
1365 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
1366 		/*
1367 		 * There is at least one prefix that has a reachable router,
1368 		 * or at least a router which probably does not advertise
1369 		 * any prefixes.  The latter would be the case when we move
1370 		 * to a new link where we have a router that does not provide
1371 		 * prefixes and we configure an address by hand.
1372 		 * Detach prefixes which have no reachable advertising
1373 		 * router, and attach other prefixes.
1374 		 */
1375 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1376 			/* XXX: a link-local prefix should never be detached */
1377 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1378 				continue;
1379 
1380 			/*
1381 			 * we aren't interested in prefixes without the L bit
1382 			 * set.
1383 			 */
1384 			if (pr->ndpr_raf_onlink == 0)
1385 				continue;
1386 
1387 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1388 			    find_pfxlist_reachable_router(pr) == NULL)
1389 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1390 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1391 			    find_pfxlist_reachable_router(pr) != 0)
1392 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1393 		}
1394 	} else {
1395 		/* there is no prefix that has a reachable router */
1396 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1397 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1398 				continue;
1399 
1400 			if (pr->ndpr_raf_onlink == 0)
1401 				continue;
1402 
1403 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1404 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1405 		}
1406 	}
1407 
1408 	/*
1409 	 * Remove each interface route associated with a (just) detached
1410 	 * prefix, and reinstall the interface route for a (just) attached
1411 	 * prefix.  Note that all attempt of reinstallation does not
1412 	 * necessarily success, when a same prefix is shared among multiple
1413 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1414 	 * so we don't have to care about them.
1415 	 */
1416 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1417 		int e;
1418 
1419 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1420 			continue;
1421 
1422 		if (pr->ndpr_raf_onlink == 0)
1423 			continue;
1424 
1425 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1426 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1427 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1428 				nd6log((LOG_ERR,
1429 				    "pfxlist_onlink_check: failed to "
1430 				    "make %s/%d offlink, errno=%d\n",
1431 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1432 				    pr->ndpr_plen, e));
1433 			}
1434 		}
1435 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1436 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1437 		    pr->ndpr_raf_onlink) {
1438 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1439 				nd6log((LOG_ERR,
1440 				    "pfxlist_onlink_check: failed to "
1441 				    "make %s/%d onlink, errno=%d\n",
1442 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1443 				    pr->ndpr_plen, e));
1444 			}
1445 		}
1446 	}
1447 
1448 	/*
1449 	 * Changes on the prefix status might affect address status as well.
1450 	 * Make sure that all addresses derived from an attached prefix are
1451 	 * attached, and that all addresses derived from a detached prefix are
1452 	 * detached.  Note, however, that a manually configured address should
1453 	 * always be attached.
1454 	 * The precise detection logic is same as the one for prefixes.
1455 	 */
1456 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1457 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1458 			continue;
1459 
1460 		if (ifa->ia6_ndpr == NULL) {
1461 			/*
1462 			 * This can happen when we first configure the address
1463 			 * (i.e. the address exists, but the prefix does not).
1464 			 * XXX: complicated relationships...
1465 			 */
1466 			continue;
1467 		}
1468 
1469 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1470 			break;
1471 	}
1472 	if (ifa) {
1473 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1474 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1475 				continue;
1476 
1477 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1478 				continue;
1479 
1480 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1481 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1482 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1483 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1484 					nd6_dad_start((struct ifaddr *)ifa,
1485 					    0);
1486 				}
1487 			} else {
1488 				if ((ifa->ia6_flags & IN6_IFF_DETACHED) == 0) {
1489 					ifa->ia6_flags |= IN6_IFF_DETACHED;
1490 				}
1491 			}
1492 		}
1493 	}
1494 	else {
1495 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1496 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1497 				continue;
1498 
1499 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1500 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1501 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1502 				/* Do we need a delay in this case? */
1503 				nd6_dad_start((struct ifaddr *)ifa, 0);
1504 			}
1505 		}
1506 	}
1507 }
1508 
1509 int
1510 nd6_prefix_onlink(struct nd_prefix *pr)
1511 {
1512 	struct ifaddr *ifa;
1513 	struct ifnet *ifp = pr->ndpr_ifp;
1514 	struct sockaddr_in6 mask6;
1515 	struct nd_prefix *opr;
1516 	u_long rtflags;
1517 	int error = 0;
1518 	struct rtentry *rt = NULL;
1519 
1520 	/* sanity check */
1521 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1522 		nd6log((LOG_ERR,
1523 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1524 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1525 		return (EEXIST);
1526 	}
1527 
1528 	/*
1529 	 * Add the interface route associated with the prefix.  Before
1530 	 * installing the route, check if there's the same prefix on another
1531 	 * interface, and the prefix has already installed the interface route.
1532 	 * Although such a configuration is expected to be rare, we explicitly
1533 	 * allow it.
1534 	 */
1535 	LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1536 		if (opr == pr)
1537 			continue;
1538 
1539 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1540 			continue;
1541 
1542 		if (opr->ndpr_plen == pr->ndpr_plen &&
1543 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1544 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1545 			return (0);
1546 	}
1547 
1548 	/*
1549 	 * We prefer link-local addresses as the associated interface address.
1550 	 */
1551 	/* search for a link-local addr */
1552 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1553 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1554 	if (ifa == NULL) {
1555 		/* XXX: freebsd does not have ifa_ifwithaf */
1556 		IFADDR_FOREACH(ifa, ifp) {
1557 			if (ifa->ifa_addr->sa_family == AF_INET6)
1558 				break;
1559 		}
1560 		/* should we care about ia6_flags? */
1561 	}
1562 	if (ifa == NULL) {
1563 		/*
1564 		 * This can still happen, when, for example, we receive an RA
1565 		 * containing a prefix with the L bit set and the A bit clear,
1566 		 * after removing all IPv6 addresses on the receiving
1567 		 * interface.  This should, of course, be rare though.
1568 		 */
1569 		nd6log((LOG_NOTICE,
1570 		    "nd6_prefix_onlink: failed to find any ifaddr"
1571 		    " to add route for a prefix(%s/%d) on %s\n",
1572 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1573 		    pr->ndpr_plen, if_name(ifp)));
1574 		return (0);
1575 	}
1576 
1577 	/*
1578 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1579 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1580 	 */
1581 	bzero(&mask6, sizeof(mask6));
1582 	mask6.sin6_len = sizeof(mask6);
1583 	mask6.sin6_addr = pr->ndpr_mask;
1584 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
1585 	rtflags = ifa->ifa_flags | RTF_UP;
1586 	if (nd6_need_cache(ifp)) {
1587 		/* explicitly set in case ifa_flags does not set the flag. */
1588 		rtflags |= RTF_CLONING;
1589 	} else {
1590 		/*
1591 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1592 		 */
1593 		rtflags &= ~RTF_CLONING;
1594 	}
1595 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1596 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1597 	if (error == 0) {
1598 		if (rt != NULL) /* this should be non NULL, though */
1599 			nd6_rtmsg(RTM_ADD, rt);
1600 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1601 	} else {
1602 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1603 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1604 		    "errno = %d\n",
1605 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1606 		    pr->ndpr_plen, if_name(ifp),
1607 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1608 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
1609 	}
1610 
1611 	if (rt != NULL)
1612 		rt->rt_refcnt--;
1613 
1614 	return (error);
1615 }
1616 
1617 int
1618 nd6_prefix_offlink(struct nd_prefix *pr)
1619 {
1620 	int error = 0;
1621 	struct ifnet *ifp = pr->ndpr_ifp;
1622 	struct nd_prefix *opr;
1623 	struct sockaddr_in6 sa6, mask6;
1624 	struct rtentry *rt = NULL;
1625 
1626 	/* sanity check */
1627 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1628 		nd6log((LOG_ERR,
1629 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1630 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1631 		return (EEXIST);
1632 	}
1633 
1634 	sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1635 	sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
1636 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1637 	    (struct sockaddr *)&mask6, 0, &rt);
1638 	if (error == 0) {
1639 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1640 
1641 		/* report the route deletion to the routing socket. */
1642 		if (rt != NULL)
1643 			nd6_rtmsg(RTM_DELETE, rt);
1644 
1645 		/*
1646 		 * There might be the same prefix on another interface,
1647 		 * the prefix which could not be on-link just because we have
1648 		 * the interface route (see comments in nd6_prefix_onlink).
1649 		 * If there's one, try to make the prefix on-link on the
1650 		 * interface.
1651 		 */
1652 		LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1653 			if (opr == pr)
1654 				continue;
1655 
1656 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1657 				continue;
1658 
1659 			/*
1660 			 * KAME specific: detached prefixes should not be
1661 			 * on-link.
1662 			 */
1663 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1664 				continue;
1665 
1666 			if (opr->ndpr_plen == pr->ndpr_plen &&
1667 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1668 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1669 				int e;
1670 
1671 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1672 					nd6log((LOG_ERR,
1673 					    "nd6_prefix_offlink: failed to "
1674 					    "recover a prefix %s/%d from %s "
1675 					    "to %s (errno = %d)\n",
1676 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1677 					    opr->ndpr_plen, if_name(ifp),
1678 					    if_name(opr->ndpr_ifp), e));
1679 				}
1680 			}
1681 		}
1682 	} else {
1683 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1684 		nd6log((LOG_ERR,
1685 		    "nd6_prefix_offlink: failed to delete route: "
1686 		    "%s/%d on %s (errno = %d)\n",
1687 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1688 		    error));
1689 	}
1690 
1691 	if (rt != NULL) {
1692 		if (rt->rt_refcnt <= 0) {
1693 			/* XXX: we should free the entry ourselves. */
1694 			rt->rt_refcnt++;
1695 			rtfree(rt);
1696 		}
1697 	}
1698 
1699 	return (error);
1700 }
1701 
1702 static struct in6_ifaddr *
1703 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1704 {
1705 	struct ifnet *ifp = pr->ndpr_ifp;
1706 	struct ifaddr *ifa;
1707 	struct in6_aliasreq ifra;
1708 	struct in6_ifaddr *ia, *ib;
1709 	int error, plen0;
1710 	struct in6_addr mask;
1711 	int prefixlen = pr->ndpr_plen;
1712 	int updateflags;
1713 
1714 	in6_prefixlen2mask(&mask, prefixlen);
1715 
1716 	/*
1717 	 * find a link-local address (will be interface ID).
1718 	 * Is it really mandatory? Theoretically, a global or a site-local
1719 	 * address can be configured without a link-local address, if we
1720 	 * have a unique interface identifier...
1721 	 *
1722 	 * it is not mandatory to have a link-local address, we can generate
1723 	 * interface identifier on the fly.  we do this because:
1724 	 * (1) it should be the easiest way to find interface identifier.
1725 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1726 	 * for multiple addresses on a single interface, and possible shortcut
1727 	 * of DAD.  we omitted DAD for this reason in the past.
1728 	 * (3) a user can prevent autoconfiguration of global address
1729 	 * by removing link-local address by hand (this is partly because we
1730 	 * don't have other way to control the use of IPv6 on an interface.
1731 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1732 	 * (4) it is easier to manage when an interface has addresses
1733 	 * with the same interface identifier, than to have multiple addresses
1734 	 * with different interface identifiers.
1735 	 */
1736 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1737 	if (ifa)
1738 		ib = (struct in6_ifaddr *)ifa;
1739 	else
1740 		return NULL;
1741 
1742 #if 0 /* don't care link local addr state, and always do DAD */
1743 	/* if link-local address is not eligible, do not autoconfigure. */
1744 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1745 		printf("in6_ifadd: link-local address not ready\n");
1746 		return NULL;
1747 	}
1748 #endif
1749 
1750 	/* prefixlen + ifidlen must be equal to 128 */
1751 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1752 	if (prefixlen != plen0) {
1753 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1754 		    "(prefix=%d ifid=%d)\n",
1755 		    if_name(ifp), prefixlen, 128 - plen0));
1756 		return NULL;
1757 	}
1758 
1759 	/* make ifaddr */
1760 
1761 	memset(&ifra, 0, sizeof(ifra));
1762 	/*
1763 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1764 	 * for safety.
1765 	 */
1766 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1767 	sockaddr_in6_init(&ifra.ifra_addr, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1768 	/* prefix */
1769 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1770 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1771 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1772 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1773 
1774 	/* interface ID */
1775 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1776 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1777 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1778 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1779 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1780 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1781 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1782 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1783 
1784 	/* new prefix mask. */
1785 	sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
1786 
1787 	/* lifetimes */
1788 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1789 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1790 
1791 	/* XXX: scope zone ID? */
1792 
1793 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1794 
1795 	/*
1796 	 * Make sure that we do not have this address already.  This should
1797 	 * usually not happen, but we can still see this case, e.g., if we
1798 	 * have manually configured the exact address to be configured.
1799 	 */
1800 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1801 		/* this should be rare enough to make an explicit log */
1802 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1803 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1804 		return (NULL);
1805 	}
1806 
1807 	/*
1808 	 * Allocate ifaddr structure, link into chain, etc.
1809 	 * If we are going to create a new address upon receiving a multicasted
1810 	 * RA, we need to impose a random delay before starting DAD.
1811 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1812 	 */
1813 	updateflags = 0;
1814 	if (mcast)
1815 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1816 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1817 		nd6log((LOG_ERR,
1818 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1819 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1820 		    error));
1821 		return (NULL);	/* ifaddr must not have been allocated. */
1822 	}
1823 
1824 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1825 
1826 	return (ia);		/* this is always non-NULL */
1827 }
1828 
1829 int
1830 in6_tmpifadd(
1831 	const struct in6_ifaddr *ia0, /* corresponding public address */
1832 	int forcegen,
1833 	int dad_delay)
1834 {
1835 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1836 	struct in6_ifaddr *newia, *ia;
1837 	struct in6_aliasreq ifra;
1838 	int i, error;
1839 	int trylimit = 3;	/* XXX: adhoc value */
1840 	int updateflags;
1841 	u_int32_t randid[2];
1842 	u_int32_t vltime0, pltime0;
1843 
1844 	memset(&ifra, 0, sizeof(ifra));
1845 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1846 	ifra.ifra_addr = ia0->ia_addr;
1847 	/* copy prefix mask */
1848 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1849 	/* clear the old IFID */
1850 	for (i = 0; i < 4; i++) {
1851 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1852 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1853 	}
1854 
1855   again:
1856 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1857 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1858 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1859 		    "random IFID\n"));
1860 		return (EINVAL);
1861 	}
1862 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1863 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1864 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1865 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1866 
1867 	/*
1868 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1869 	 * However, we may still have a chance to see collision, because
1870 	 * there may be a time lag between generation of the ID and generation
1871 	 * of the address.  So, we'll do one more sanity check.
1872 	 */
1873 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1874 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1875 		    &ifra.ifra_addr.sin6_addr)) {
1876 			if (trylimit-- == 0) {
1877 				/*
1878 				 * Give up.  Something strange should have
1879 				 * happened.
1880 				 */
1881 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1882 				    "find a unique random IFID\n"));
1883 				return (EEXIST);
1884 			}
1885 			forcegen = 1;
1886 			goto again;
1887 		}
1888 	}
1889 
1890 	/*
1891 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1892          * public address or TEMP_VALID_LIFETIME.
1893 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1894          * of the public address or TEMP_PREFERRED_LIFETIME -
1895          * DESYNC_FACTOR.
1896 	 */
1897 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1898 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1899 		    (ia0->ia6_lifetime.ia6t_vltime -
1900 		    (time_second - ia0->ia6_updatetime));
1901 		if (vltime0 > ip6_temp_valid_lifetime)
1902 			vltime0 = ip6_temp_valid_lifetime;
1903 	} else
1904 		vltime0 = ip6_temp_valid_lifetime;
1905 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1906 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1907 		    (ia0->ia6_lifetime.ia6t_pltime -
1908 		    (time_second - ia0->ia6_updatetime));
1909 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
1910 			pltime0 = ip6_temp_preferred_lifetime -
1911 			    ip6_desync_factor;
1912 		}
1913 	} else
1914 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
1915 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1916 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1917 
1918 	/*
1919 	 * A temporary address is created only if this calculated Preferred
1920 	 * Lifetime is greater than REGEN_ADVANCE time units.
1921 	 */
1922 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
1923 		return (0);
1924 
1925 	/* XXX: scope zone ID? */
1926 
1927 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1928 
1929 	/* allocate ifaddr structure, link into chain, etc. */
1930 	updateflags = 0;
1931 	if (dad_delay)
1932 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1933 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1934 		return (error);
1935 
1936 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1937 	if (newia == NULL) {	/* XXX: can it happen? */
1938 		nd6log((LOG_ERR,
1939 		    "in6_tmpifadd: ifa update succeeded, but we got "
1940 		    "no ifaddr\n"));
1941 		return (EINVAL); /* XXX */
1942 	}
1943 	newia->ia6_ndpr = ia0->ia6_ndpr;
1944 	newia->ia6_ndpr->ndpr_refcnt++;
1945 
1946 	/*
1947 	 * A newly added address might affect the status of other addresses.
1948 	 * XXX: when the temporary address is generated with a new public
1949 	 * address, the onlink check is redundant.  However, it would be safe
1950 	 * to do the check explicitly everywhere a new address is generated,
1951 	 * and, in fact, we surely need the check when we create a new
1952 	 * temporary address due to deprecation of an old temporary address.
1953 	 */
1954 	pfxlist_onlink_check();
1955 
1956 	return (0);
1957 }
1958 
1959 static int
1960 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1961 {
1962 
1963 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
1964 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
1965 		nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
1966 		    "(%d) is greater than valid lifetime(%d)\n",
1967 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
1968 		return (EINVAL);
1969 	}
1970 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1971 		ndpr->ndpr_preferred = 0;
1972 	else
1973 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
1974 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1975 		ndpr->ndpr_expire = 0;
1976 	else
1977 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
1978 
1979 	return 0;
1980 }
1981 
1982 static void
1983 in6_init_address_ltimes(struct nd_prefix *new,
1984     struct in6_addrlifetime *lt6)
1985 {
1986 
1987 	/* Valid lifetime must not be updated unless explicitly specified. */
1988 	/* init ia6t_expire */
1989 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
1990 		lt6->ia6t_expire = 0;
1991 	else {
1992 		lt6->ia6t_expire = time_second;
1993 		lt6->ia6t_expire += lt6->ia6t_vltime;
1994 	}
1995 
1996 	/* init ia6t_preferred */
1997 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
1998 		lt6->ia6t_preferred = 0;
1999 	else {
2000 		lt6->ia6t_preferred = time_second;
2001 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2002 	}
2003 }
2004 
2005 /*
2006  * Delete all the routing table entries that use the specified gateway.
2007  * XXX: this function causes search through all entries of routing table, so
2008  * it shouldn't be called when acting as a router.
2009  */
2010 void
2011 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2012 {
2013 	int s = splsoftnet();
2014 
2015 	/* We'll care only link-local addresses */
2016 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2017 		splx(s);
2018 		return;
2019 	}
2020 
2021 	rt_walktree(AF_INET6, rt6_deleteroute, (void *)gateway);
2022 	splx(s);
2023 }
2024 
2025 static int
2026 rt6_deleteroute(struct rtentry *rt, void *arg)
2027 {
2028 #define SIN6(s)	((struct sockaddr_in6 *)s)
2029 	struct in6_addr *gate = (struct in6_addr *)arg;
2030 
2031 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2032 		return (0);
2033 
2034 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr))
2035 		return (0);
2036 
2037 	/*
2038 	 * Do not delete a static route.
2039 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2040 	 * 'cloned' bit instead?
2041 	 */
2042 	if ((rt->rt_flags & RTF_STATIC) != 0)
2043 		return (0);
2044 
2045 	/*
2046 	 * We delete only host route. This means, in particular, we don't
2047 	 * delete default route.
2048 	 */
2049 	if ((rt->rt_flags & RTF_HOST) == 0)
2050 		return (0);
2051 
2052 	return (rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
2053 	    rt_mask(rt), rt->rt_flags, 0));
2054 #undef SIN6
2055 }
2056 
2057 int
2058 nd6_setdefaultiface(int ifindex)
2059 {
2060 	int error = 0;
2061 
2062 	if (ifindex < 0 || if_indexlim <= ifindex)
2063 		return (EINVAL);
2064 	if (ifindex != 0 && !ifindex2ifnet[ifindex])
2065 		return (EINVAL);
2066 
2067 	if (nd6_defifindex != ifindex) {
2068 		nd6_defifindex = ifindex;
2069 		if (nd6_defifindex > 0) {
2070 			nd6_defifp = ifindex2ifnet[nd6_defifindex];
2071 		} else
2072 			nd6_defifp = NULL;
2073 
2074 		/*
2075 		 * Our current implementation assumes one-to-one maping between
2076 		 * interfaces and links, so it would be natural to use the
2077 		 * default interface as the default link.
2078 		 */
2079 		scope6_setdefault(nd6_defifp);
2080 	}
2081 
2082 	return (error);
2083 }
2084