xref: /freebsd/sys/netinet6/nd6_rtr.c (revision 99f3b482da15db6f2ae4702d592b53dd36d4cf55)
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 #define SDL(s)	((struct sockaddr_dl *)s)
64 
65 static int rtpref __P((struct nd_defrouter *));
66 static struct nd_defrouter *defrtrlist_update __P((struct nd_defrouter *));
67 static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
68     struct mbuf *, int));
69 static struct in6_ifaddr *in6_ifadd __P((struct nd_prefixctl *,	int));
70 static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
71 	struct nd_defrouter *));
72 static void pfxrtr_add __P((struct nd_prefix *, struct nd_defrouter *));
73 static void pfxrtr_del __P((struct nd_pfxrouter *));
74 static struct nd_pfxrouter *find_pfxlist_reachable_router
75 	__P((struct nd_prefix *));
76 static void defrouter_delreq __P((struct nd_defrouter *));
77 static void nd6_rtmsg __P((int, struct rtentry *));
78 
79 static int in6_init_prefix_ltimes __P((struct nd_prefix *));
80 static void in6_init_address_ltimes __P((struct nd_prefix *,
81 	struct in6_addrlifetime *));
82 
83 static int rt6_deleteroute __P((struct radix_node *, 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 /*
96  * shorter lifetimes for debugging purposes.
97 int ip6_temp_preferred_lifetime = 800;
98 static int ip6_temp_valid_lifetime = 1800;
99 */
100 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
101 
102 /* RTPREF_MEDIUM has to be 0! */
103 #define RTPREF_HIGH	1
104 #define RTPREF_MEDIUM	0
105 #define RTPREF_LOW	(-1)
106 #define RTPREF_RESERVED	(-2)
107 #define RTPREF_INVALID	(-3)	/* internal */
108 
109 /*
110  * Receive Router Solicitation Message - just for routers.
111  * Router solicitation/advertisement is mostly managed by userland program
112  * (rtadvd) so here we have no function like nd6_ra_output().
113  *
114  * Based on RFC 2461
115  */
116 void
117 nd6_rs_input(m, off, icmp6len)
118 	struct	mbuf *m;
119 	int off, icmp6len;
120 {
121 	struct ifnet *ifp = m->m_pkthdr.rcvif;
122 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
123 	struct nd_router_solicit *nd_rs;
124 	struct in6_addr saddr6 = ip6->ip6_src;
125 	char *lladdr = NULL;
126 	int lladdrlen = 0;
127 	union nd_opts ndopts;
128 
129 	/* If I'm not a router, ignore it. */
130 	if (ip6_accept_rtadv != 0 || ip6_forwarding != 1)
131 		goto freeit;
132 
133 	/* Sanity checks */
134 	if (ip6->ip6_hlim != 255) {
135 		nd6log((LOG_ERR,
136 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
137 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
138 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
139 		goto bad;
140 	}
141 
142 	/*
143 	 * Don't update the neighbor cache, if src = ::.
144 	 * This indicates that the src has no IP address assigned yet.
145 	 */
146 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
147 		goto freeit;
148 
149 #ifndef PULLDOWN_TEST
150 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
151 	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
152 #else
153 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
154 	if (nd_rs == NULL) {
155 		icmp6stat.icp6s_tooshort++;
156 		return;
157 	}
158 #endif
159 
160 	icmp6len -= sizeof(*nd_rs);
161 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
162 	if (nd6_options(&ndopts) < 0) {
163 		nd6log((LOG_INFO,
164 		    "nd6_rs_input: invalid ND option, ignored\n"));
165 		/* nd6_options have incremented stats */
166 		goto freeit;
167 	}
168 
169 	if (ndopts.nd_opts_src_lladdr) {
170 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
171 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
172 	}
173 
174 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
175 		nd6log((LOG_INFO,
176 		    "nd6_rs_input: lladdrlen mismatch for %s "
177 		    "(if %d, RS packet %d)\n",
178 		    ip6_sprintf(&saddr6),
179 		    ifp->if_addrlen, lladdrlen - 2));
180 		goto bad;
181 	}
182 
183 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
184 
185  freeit:
186 	m_freem(m);
187 	return;
188 
189  bad:
190 	icmp6stat.icp6s_badrs++;
191 	m_freem(m);
192 }
193 
194 /*
195  * Receive Router Advertisement Message.
196  *
197  * Based on RFC 2461
198  * TODO: on-link bit on prefix information
199  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
200  */
201 void
202 nd6_ra_input(m, off, icmp6len)
203 	struct	mbuf *m;
204 	int off, icmp6len;
205 {
206 	struct ifnet *ifp = m->m_pkthdr.rcvif;
207 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
208 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
209 	struct nd_router_advert *nd_ra;
210 	struct in6_addr saddr6 = ip6->ip6_src;
211 	int mcast = 0;
212 	union nd_opts ndopts;
213 	struct nd_defrouter *dr;
214 
215 	/*
216 	 * We only accept RAs only when
217 	 * the system-wide variable allows the acceptance, and
218 	 * per-interface variable allows RAs on the receiving interface.
219 	 */
220 	if (ip6_accept_rtadv == 0)
221 		goto freeit;
222 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
223 		goto freeit;
224 
225 	if (ip6->ip6_hlim != 255) {
226 		nd6log((LOG_ERR,
227 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
228 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
229 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
230 		goto bad;
231 	}
232 
233 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
234 		nd6log((LOG_ERR,
235 		    "nd6_ra_input: src %s is not link-local\n",
236 		    ip6_sprintf(&saddr6)));
237 		goto bad;
238 	}
239 
240 #ifndef PULLDOWN_TEST
241 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
242 	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
243 #else
244 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
245 	if (nd_ra == NULL) {
246 		icmp6stat.icp6s_tooshort++;
247 		return;
248 	}
249 #endif
250 
251 	icmp6len -= sizeof(*nd_ra);
252 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
253 	if (nd6_options(&ndopts) < 0) {
254 		nd6log((LOG_INFO,
255 		    "nd6_ra_input: invalid ND option, ignored\n"));
256 		/* nd6_options have incremented stats */
257 		goto freeit;
258 	}
259 
260     {
261 	struct nd_defrouter dr0;
262 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
263 
264 	/* remember if this is a multicasted advertisement */
265 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
266 		mcast = 1;
267 
268 	bzero(&dr0, sizeof(dr0));
269 	dr0.rtaddr = saddr6;
270 	dr0.flags  = nd_ra->nd_ra_flags_reserved;
271 	dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
272 	dr0.expire = time_second + dr0.rtlifetime;
273 	dr0.ifp = ifp;
274 	/* unspecified or not? (RFC 2461 6.3.4) */
275 	if (advreachable) {
276 		advreachable = ntohl(advreachable);
277 		if (advreachable <= MAX_REACHABLE_TIME &&
278 		    ndi->basereachable != advreachable) {
279 			ndi->basereachable = advreachable;
280 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
281 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
282 		}
283 	}
284 	if (nd_ra->nd_ra_retransmit)
285 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
286 	if (nd_ra->nd_ra_curhoplimit)
287 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
288 	dr = defrtrlist_update(&dr0);
289     }
290 
291 	/*
292 	 * prefix
293 	 */
294 	if (ndopts.nd_opts_pi) {
295 		struct nd_opt_hdr *pt;
296 		struct nd_opt_prefix_info *pi = NULL;
297 		struct nd_prefixctl pr;
298 
299 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
300 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
301 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
302 						(pt->nd_opt_len << 3))) {
303 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
304 				continue;
305 			pi = (struct nd_opt_prefix_info *)pt;
306 
307 			if (pi->nd_opt_pi_len != 4) {
308 				nd6log((LOG_INFO,
309 				    "nd6_ra_input: invalid option "
310 				    "len %d for prefix information option, "
311 				    "ignored\n", pi->nd_opt_pi_len));
312 				continue;
313 			}
314 
315 			if (128 < pi->nd_opt_pi_prefix_len) {
316 				nd6log((LOG_INFO,
317 				    "nd6_ra_input: invalid prefix "
318 				    "len %d for prefix information option, "
319 				    "ignored\n", pi->nd_opt_pi_prefix_len));
320 				continue;
321 			}
322 
323 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
324 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
325 				nd6log((LOG_INFO,
326 				    "nd6_ra_input: invalid prefix "
327 				    "%s, ignored\n",
328 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
329 				continue;
330 			}
331 
332 			bzero(&pr, sizeof(pr));
333 			pr.ndpr_prefix.sin6_family = AF_INET6;
334 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
335 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
336 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
337 
338 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
339 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
340 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
341 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
342 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
343 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
344 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
345 			(void)prelist_update(&pr, dr, m, mcast);
346 		}
347 	}
348 
349 	/*
350 	 * MTU
351 	 */
352 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
353 		u_long mtu;
354 		u_long maxmtu;
355 
356 		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
357 
358 		/* lower bound */
359 		if (mtu < IPV6_MMTU) {
360 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
361 			    "mtu=%lu sent from %s, ignoring\n",
362 			    mtu, ip6_sprintf(&ip6->ip6_src)));
363 			goto skip;
364 		}
365 
366 		/* upper bound */
367 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
368 		    ? ndi->maxmtu : ifp->if_mtu;
369 		if (mtu <= maxmtu) {
370 			int change = (ndi->linkmtu != mtu);
371 
372 			ndi->linkmtu = mtu;
373 			if (change) /* in6_maxmtu may change */
374 				in6_setmaxmtu();
375 		} else {
376 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
377 			    "mtu=%lu sent from %s; "
378 			    "exceeds maxmtu %lu, ignoring\n",
379 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
380 		}
381 	}
382 
383  skip:
384 
385 	/*
386 	 * Source link layer address
387 	 */
388     {
389 	char *lladdr = NULL;
390 	int lladdrlen = 0;
391 
392 	if (ndopts.nd_opts_src_lladdr) {
393 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
394 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
395 	}
396 
397 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
398 		nd6log((LOG_INFO,
399 		    "nd6_ra_input: lladdrlen mismatch for %s "
400 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
401 		    ifp->if_addrlen, lladdrlen - 2));
402 		goto bad;
403 	}
404 
405 	nd6_cache_lladdr(ifp, &saddr6, lladdr,
406 	    lladdrlen, ND_ROUTER_ADVERT, 0);
407 
408 	/*
409 	 * Installing a link-layer address might change the state of the
410 	 * router's neighbor cache, which might also affect our on-link
411 	 * detection of adveritsed prefixes.
412 	 */
413 	pfxlist_onlink_check();
414     }
415 
416  freeit:
417 	m_freem(m);
418 	return;
419 
420  bad:
421 	icmp6stat.icp6s_badra++;
422 	m_freem(m);
423 }
424 
425 /*
426  * default router list proccessing sub routines
427  */
428 
429 /* tell the change to user processes watching the routing socket. */
430 static void
431 nd6_rtmsg(cmd, rt)
432 	int cmd;
433 	struct rtentry *rt;
434 {
435 	struct rt_addrinfo info;
436 
437 	bzero((caddr_t)&info, sizeof(info));
438 	info.rti_info[RTAX_DST] = rt_key(rt);
439 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
440 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
441 	if (rt->rt_ifp) {
442 		info.rti_info[RTAX_IFP] =
443 		    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
444 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
445 	}
446 
447 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
448 }
449 
450 void
451 defrouter_addreq(new)
452 	struct nd_defrouter *new;
453 {
454 	struct sockaddr_in6 def, mask, gate;
455 	struct rtentry *newrt = NULL;
456 	int s;
457 	int error;
458 
459 	bzero(&def, sizeof(def));
460 	bzero(&mask, sizeof(mask));
461 	bzero(&gate, sizeof(gate));
462 
463 	def.sin6_len = mask.sin6_len = gate.sin6_len =
464 	    sizeof(struct sockaddr_in6);
465 	def.sin6_family = gate.sin6_family = AF_INET6;
466 	gate.sin6_addr = new->rtaddr;
467 
468 	s = splnet();
469 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
470 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
471 	    RTF_GATEWAY, &newrt);
472 	if (newrt) {
473 		RT_LOCK(newrt);
474 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
475 		RT_REMREF(newrt);
476 		RT_UNLOCK(newrt);
477 	}
478 	if (error == 0)
479 		new->installed = 1;
480 	splx(s);
481 	return;
482 }
483 
484 struct nd_defrouter *
485 defrouter_lookup(addr, ifp)
486 	struct in6_addr *addr;
487 	struct ifnet *ifp;
488 {
489 	struct nd_defrouter *dr;
490 
491 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
492 	     dr = TAILQ_NEXT(dr, dr_entry)) {
493 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
494 			return (dr);
495 	}
496 
497 	return (NULL);		/* search failed */
498 }
499 
500 /*
501  * Remove the default route for a given router.
502  * This is just a subroutine function for defrouter_select(), and should
503  * not be called from anywhere else.
504  */
505 static void
506 defrouter_delreq(dr)
507 	struct nd_defrouter *dr;
508 {
509 	struct sockaddr_in6 def, mask, gate;
510 	struct rtentry *oldrt = NULL;
511 
512 	bzero(&def, sizeof(def));
513 	bzero(&mask, sizeof(mask));
514 	bzero(&gate, sizeof(gate));
515 
516 	def.sin6_len = mask.sin6_len = gate.sin6_len =
517 	    sizeof(struct sockaddr_in6);
518 	def.sin6_family = gate.sin6_family = AF_INET6;
519 	gate.sin6_addr = dr->rtaddr;
520 
521 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
522 	    (struct sockaddr *)&gate,
523 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
524 	if (oldrt) {
525 		nd6_rtmsg(RTM_DELETE, oldrt);
526 		RTFREE(oldrt);
527 	}
528 
529 	dr->installed = 0;
530 }
531 
532 /*
533  * remove all default routes from default router list
534  */
535 void
536 defrouter_reset()
537 {
538 	struct nd_defrouter *dr;
539 
540 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
541 	     dr = TAILQ_NEXT(dr, dr_entry))
542 		defrouter_delreq(dr);
543 
544 	/*
545 	 * XXX should we also nuke any default routers in the kernel, by
546 	 * going through them by rtalloc1()?
547 	 */
548 }
549 
550 void
551 defrtrlist_del(dr)
552 	struct nd_defrouter *dr;
553 {
554 	struct nd_defrouter *deldr = NULL;
555 	struct nd_prefix *pr;
556 
557 	/*
558 	 * Flush all the routing table entries that use the router
559 	 * as a next hop.
560 	 */
561 	if (!ip6_forwarding && ip6_accept_rtadv) /* XXX: better condition? */
562 		rt6_flush(&dr->rtaddr, dr->ifp);
563 
564 	if (dr->installed) {
565 		deldr = dr;
566 		defrouter_delreq(dr);
567 	}
568 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
569 
570 	/*
571 	 * Also delete all the pointers to the router in each prefix lists.
572 	 */
573 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
574 		struct nd_pfxrouter *pfxrtr;
575 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
576 			pfxrtr_del(pfxrtr);
577 	}
578 	pfxlist_onlink_check();
579 
580 	/*
581 	 * If the router is the primary one, choose a new one.
582 	 * Note that defrouter_select() will remove the current gateway
583 	 * from the routing table.
584 	 */
585 	if (deldr)
586 		defrouter_select();
587 
588 	free(dr, M_IP6NDP);
589 }
590 
591 /*
592  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
593  * draft-ietf-ipngwg-router-selection:
594  * 1) Routers that are reachable or probably reachable should be preferred.
595  *    If we have more than one (probably) reachable router, prefer ones
596  *    with the highest router preference.
597  * 2) When no routers on the list are known to be reachable or
598  *    probably reachable, routers SHOULD be selected in a round-robin
599  *    fashion, regardless of router preference values.
600  * 3) If the Default Router List is empty, assume that all
601  *    destinations are on-link.
602  *
603  * We assume nd_defrouter is sorted by router preference value.
604  * Since the code below covers both with and without router preference cases,
605  * we do not need to classify the cases by ifdef.
606  *
607  * At this moment, we do not try to install more than one default router,
608  * even when the multipath routing is available, because we're not sure about
609  * the benefits for stub hosts comparing to the risk of making the code
610  * complicated and the possibility of introducing bugs.
611  */
612 void
613 defrouter_select()
614 {
615 	int s = splnet();
616 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
617 	struct rtentry *rt = NULL;
618 	struct llinfo_nd6 *ln = NULL;
619 
620 	/*
621 	 * This function should be called only when acting as an autoconfigured
622 	 * host.  Although the remaining part of this function is not effective
623 	 * if the node is not an autoconfigured host, we explicitly exclude
624 	 * such cases here for safety.
625 	 */
626 	if (ip6_forwarding || !ip6_accept_rtadv) {
627 		nd6log((LOG_WARNING,
628 		    "defrouter_select: called unexpectedly (forwarding=%d, "
629 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
630 		splx(s);
631 		return;
632 	}
633 
634 	/*
635 	 * Let's handle easy case (3) first:
636 	 * If default router list is empty, there's nothing to be done.
637 	 */
638 	if (!TAILQ_FIRST(&nd_defrouter)) {
639 		splx(s);
640 		return;
641 	}
642 
643 	/*
644 	 * Search for a (probably) reachable router from the list.
645 	 * We just pick up the first reachable one (if any), assuming that
646 	 * the ordering rule of the list described in defrtrlist_update().
647 	 */
648 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
649 	     dr = TAILQ_NEXT(dr, dr_entry)) {
650 		if (selected_dr == NULL &&
651 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
652 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
653 		    ND6_IS_LLINFO_PROBREACH(ln)) {
654 			selected_dr = dr;
655 		}
656 
657 		if (dr->installed && installed_dr == NULL)
658 			installed_dr = dr;
659 		else if (dr->installed && installed_dr) {
660 			/* this should not happen.  warn for diagnosis. */
661 			log(LOG_ERR, "defrouter_select: more than one router"
662 			    " is installed\n");
663 		}
664 	}
665 	/*
666 	 * If none of the default routers was found to be reachable,
667 	 * round-robin the list regardless of preference.
668 	 * Otherwise, if we have an installed router, check if the selected
669 	 * (reachable) router should really be preferred to the installed one.
670 	 * We only prefer the new router when the old one is not reachable
671 	 * or when the new one has a really higher preference value.
672 	 */
673 	if (selected_dr == NULL) {
674 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
675 			selected_dr = TAILQ_FIRST(&nd_defrouter);
676 		else
677 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
678 	} else if (installed_dr &&
679 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
680 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
681 	    ND6_IS_LLINFO_PROBREACH(ln) &&
682 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
683 		selected_dr = installed_dr;
684 	}
685 
686 	/*
687 	 * If the selected router is different than the installed one,
688 	 * remove the installed router and install the selected one.
689 	 * Note that the selected router is never NULL here.
690 	 */
691 	if (installed_dr != selected_dr) {
692 		if (installed_dr)
693 			defrouter_delreq(installed_dr);
694 		defrouter_addreq(selected_dr);
695 	}
696 
697 	splx(s);
698 	return;
699 }
700 
701 /*
702  * for default router selection
703  * regards router-preference field as a 2-bit signed integer
704  */
705 static int
706 rtpref(struct nd_defrouter *dr)
707 {
708 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
709 	case ND_RA_FLAG_RTPREF_HIGH:
710 		return (RTPREF_HIGH);
711 	case ND_RA_FLAG_RTPREF_MEDIUM:
712 	case ND_RA_FLAG_RTPREF_RSV:
713 		return (RTPREF_MEDIUM);
714 	case ND_RA_FLAG_RTPREF_LOW:
715 		return (RTPREF_LOW);
716 	default:
717 		/*
718 		 * This case should never happen.  If it did, it would mean a
719 		 * serious bug of kernel internal.  We thus always bark here.
720 		 * Or, can we even panic?
721 		 */
722 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
723 		return (RTPREF_INVALID);
724 	}
725 	/* NOTREACHED */
726 }
727 
728 static struct nd_defrouter *
729 defrtrlist_update(new)
730 	struct nd_defrouter *new;
731 {
732 	struct nd_defrouter *dr, *n;
733 	int s = splnet();
734 
735 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
736 		/* entry exists */
737 		if (new->rtlifetime == 0) {
738 			defrtrlist_del(dr);
739 			dr = NULL;
740 		} else {
741 			int oldpref = rtpref(dr);
742 
743 			/* override */
744 			dr->flags = new->flags; /* xxx flag check */
745 			dr->rtlifetime = new->rtlifetime;
746 			dr->expire = new->expire;
747 
748 			/*
749 			 * If the preference does not change, there's no need
750 			 * to sort the entries.
751 			 */
752 			if (rtpref(new) == oldpref) {
753 				splx(s);
754 				return (dr);
755 			}
756 
757 			/*
758 			 * preferred router may be changed, so relocate
759 			 * this router.
760 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
761 			 * However, since defrtrlist_del() has many side
762 			 * effects, we intentionally do so here.
763 			 * defrouter_select() below will handle routing
764 			 * changes later.
765 			 */
766 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
767 			n = dr;
768 			goto insert;
769 		}
770 		splx(s);
771 		return (dr);
772 	}
773 
774 	/* entry does not exist */
775 	if (new->rtlifetime == 0) {
776 		splx(s);
777 		return (NULL);
778 	}
779 
780 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
781 	if (n == NULL) {
782 		splx(s);
783 		return (NULL);
784 	}
785 	bzero(n, sizeof(*n));
786 	*n = *new;
787 
788 insert:
789 	/*
790 	 * Insert the new router in the Default Router List;
791 	 * The Default Router List should be in the descending order
792 	 * of router-preferece.  Routers with the same preference are
793 	 * sorted in the arriving time order.
794 	 */
795 
796 	/* insert at the end of the group */
797 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
798 	     dr = TAILQ_NEXT(dr, dr_entry)) {
799 		if (rtpref(n) > rtpref(dr))
800 			break;
801 	}
802 	if (dr)
803 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
804 	else
805 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
806 
807 	defrouter_select();
808 
809 	splx(s);
810 
811 	return (n);
812 }
813 
814 static struct nd_pfxrouter *
815 pfxrtr_lookup(pr, dr)
816 	struct nd_prefix *pr;
817 	struct nd_defrouter *dr;
818 {
819 	struct nd_pfxrouter *search;
820 
821 	for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
822 		if (search->router == dr)
823 			break;
824 	}
825 
826 	return (search);
827 }
828 
829 static void
830 pfxrtr_add(pr, dr)
831 	struct nd_prefix *pr;
832 	struct nd_defrouter *dr;
833 {
834 	struct nd_pfxrouter *new;
835 
836 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
837 	if (new == NULL)
838 		return;
839 	bzero(new, sizeof(*new));
840 	new->router = dr;
841 
842 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
843 
844 	pfxlist_onlink_check();
845 }
846 
847 static void
848 pfxrtr_del(pfr)
849 	struct nd_pfxrouter *pfr;
850 {
851 	LIST_REMOVE(pfr, pfr_entry);
852 	free(pfr, M_IP6NDP);
853 }
854 
855 struct nd_prefix *
856 nd6_prefix_lookup(key)
857 	struct nd_prefixctl *key;
858 {
859 	struct nd_prefix *search;
860 
861 	for (search = nd_prefix.lh_first; search; search = search->ndpr_next) {
862 		if (key->ndpr_ifp == search->ndpr_ifp &&
863 		    key->ndpr_plen == search->ndpr_plen &&
864 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
865 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
866 			break;
867 		}
868 	}
869 
870 	return (search);
871 }
872 
873 int
874 nd6_prelist_add(pr, dr, newp)
875 	struct nd_prefixctl *pr;
876 	struct nd_prefix **newp;
877 	struct nd_defrouter *dr;
878 {
879 	struct nd_prefix *new = NULL;
880 	int error = 0;
881 	int i, s;
882 
883 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
884 	if (new == NULL)
885 		return(ENOMEM);
886 	bzero(new, sizeof(*new));
887 	new->ndpr_ifp = pr->ndpr_ifp;
888 	new->ndpr_prefix = pr->ndpr_prefix;
889 	new->ndpr_plen = pr->ndpr_plen;
890 	new->ndpr_vltime = pr->ndpr_vltime;
891 	new->ndpr_pltime = pr->ndpr_pltime;
892 	new->ndpr_flags = pr->ndpr_flags;
893 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
894 		free(new, M_IP6NDP);
895 		return(error);
896 	}
897 	new->ndpr_lastupdate = time_second;
898 	if (newp != NULL)
899 		*newp = new;
900 
901 	/* initialization */
902 	LIST_INIT(&new->ndpr_advrtrs);
903 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
904 	/* make prefix in the canonical form */
905 	for (i = 0; i < 4; i++)
906 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
907 		    new->ndpr_mask.s6_addr32[i];
908 
909 	s = splnet();
910 	/* link ndpr_entry to nd_prefix list */
911 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
912 	splx(s);
913 
914 	/* ND_OPT_PI_FLAG_ONLINK processing */
915 	if (new->ndpr_raf_onlink) {
916 		int e;
917 
918 		if ((e = nd6_prefix_onlink(new)) != 0) {
919 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
920 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
921 			    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
922 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
923 			/* proceed anyway. XXX: is it correct? */
924 		}
925 	}
926 
927 	if (dr)
928 		pfxrtr_add(new, dr);
929 
930 	return 0;
931 }
932 
933 void
934 prelist_remove(pr)
935 	struct nd_prefix *pr;
936 {
937 	struct nd_pfxrouter *pfr, *next;
938 	int e, s;
939 
940 	/* make sure to invalidate the prefix until it is really freed. */
941 	pr->ndpr_vltime = 0;
942 	pr->ndpr_pltime = 0;
943 
944 	/*
945 	 * Though these flags are now meaningless, we'd rather keep the value
946 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
947 	 * when executing "ndp -p".
948 	 */
949 
950 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
951 	    (e = nd6_prefix_offlink(pr)) != 0) {
952 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
953 		    "on %s, errno=%d\n",
954 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
955 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
956 		/* what should we do? */
957 	}
958 
959 	if (pr->ndpr_refcnt > 0)
960 		return;		/* notice here? */
961 
962 	s = splnet();
963 
964 	/* unlink ndpr_entry from nd_prefix list */
965 	LIST_REMOVE(pr, ndpr_entry);
966 
967 	/* free list of routers that adversed the prefix */
968 	for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
969 		next = pfr->pfr_next;
970 
971 		free(pfr, M_IP6NDP);
972 	}
973 	splx(s);
974 
975 	free(pr, M_IP6NDP);
976 
977 	pfxlist_onlink_check();
978 }
979 
980 static int
981 prelist_update(new, dr, m, mcast)
982 	struct nd_prefixctl *new;
983 	struct nd_defrouter *dr; /* may be NULL */
984 	struct mbuf *m;
985 	int mcast;
986 {
987 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
988 	struct ifaddr *ifa;
989 	struct ifnet *ifp = new->ndpr_ifp;
990 	struct nd_prefix *pr;
991 	int s = splnet();
992 	int error = 0;
993 	int newprefix = 0;
994 	int auth;
995 	struct in6_addrlifetime lt6_tmp;
996 
997 	auth = 0;
998 	if (m) {
999 		/*
1000 		 * Authenticity for NA consists authentication for
1001 		 * both IP header and IP datagrams, doesn't it ?
1002 		 */
1003 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1004 		auth = ((m->m_flags & M_AUTHIPHDR) &&
1005 		    (m->m_flags & M_AUTHIPDGM));
1006 #endif
1007 	}
1008 
1009 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1010 		/*
1011 		 * nd6_prefix_lookup() ensures that pr and new have the same
1012 		 * prefix on a same interface.
1013 		 */
1014 
1015 		/*
1016 		 * Update prefix information.  Note that the on-link (L) bit
1017 		 * and the autonomous (A) bit should NOT be changed from 1
1018 		 * to 0.
1019 		 */
1020 		if (new->ndpr_raf_onlink == 1)
1021 			pr->ndpr_raf_onlink = 1;
1022 		if (new->ndpr_raf_auto == 1)
1023 			pr->ndpr_raf_auto = 1;
1024 		if (new->ndpr_raf_onlink) {
1025 			pr->ndpr_vltime = new->ndpr_vltime;
1026 			pr->ndpr_pltime = new->ndpr_pltime;
1027 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1028 			pr->ndpr_lastupdate = time_second;
1029 		}
1030 
1031 		if (new->ndpr_raf_onlink &&
1032 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1033 			int e;
1034 
1035 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1036 				nd6log((LOG_ERR,
1037 				    "prelist_update: failed to make "
1038 				    "the prefix %s/%d on-link on %s "
1039 				    "(errno=%d)\n",
1040 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1041 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1042 				/* proceed anyway. XXX: is it correct? */
1043 			}
1044 		}
1045 
1046 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1047 			pfxrtr_add(pr, dr);
1048 	} else {
1049 		struct nd_prefix *newpr = NULL;
1050 
1051 		newprefix = 1;
1052 
1053 		if (new->ndpr_vltime == 0)
1054 			goto end;
1055 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1056 			goto end;
1057 
1058 		error = nd6_prelist_add(new, dr, &newpr);
1059 		if (error != 0 || newpr == NULL) {
1060 			nd6log((LOG_NOTICE, "prelist_update: "
1061 			    "nd6_prelist_add failed for %s/%d on %s "
1062 			    "errno=%d, returnpr=%p\n",
1063 			    ip6_sprintf(&new->ndpr_prefix.sin6_addr),
1064 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1065 			    error, newpr));
1066 			goto end; /* we should just give up in this case. */
1067 		}
1068 
1069 		/*
1070 		 * XXX: from the ND point of view, we can ignore a prefix
1071 		 * with the on-link bit being zero.  However, we need a
1072 		 * prefix structure for references from autoconfigured
1073 		 * addresses.  Thus, we explicitly make sure that the prefix
1074 		 * itself expires now.
1075 		 */
1076 		if (newpr->ndpr_raf_onlink == 0) {
1077 			newpr->ndpr_vltime = 0;
1078 			newpr->ndpr_pltime = 0;
1079 			in6_init_prefix_ltimes(newpr);
1080 		}
1081 
1082 		pr = newpr;
1083 	}
1084 
1085 	/*
1086 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1087 	 * Note that pr must be non NULL at this point.
1088 	 */
1089 
1090 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1091 	if (!new->ndpr_raf_auto)
1092 		goto end;
1093 
1094 	/*
1095 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1096 	 * nd6_ra_input.
1097 	 */
1098 
1099 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1100 	if (new->ndpr_pltime > new->ndpr_vltime) {
1101 		error = EINVAL;	/* XXX: won't be used */
1102 		goto end;
1103 	}
1104 
1105  	/*
1106 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1107 	 * an address configured by stateless autoconfiguration already in the
1108 	 * list of addresses associated with the interface, and the Valid
1109 	 * Lifetime is not 0, form an address.  We first check if we have
1110 	 * a matching prefix.
1111 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1112 	 * consider autoconfigured addresses while RFC2462 simply said
1113 	 * "address".
1114 	 */
1115 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1116 		struct in6_ifaddr *ifa6;
1117 		u_int32_t remaininglifetime;
1118 
1119 		if (ifa->ifa_addr->sa_family != AF_INET6)
1120 			continue;
1121 
1122 		ifa6 = (struct in6_ifaddr *)ifa;
1123 
1124 		/*
1125 		 * We only consider autoconfigured addresses as per rfc2462bis.
1126 		 */
1127 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1128 			continue;
1129 
1130 		/*
1131 		 * Spec is not clear here, but I believe we should concentrate
1132 		 * on unicast (i.e. not anycast) addresses.
1133 		 * XXX: other ia6_flags? detached or duplicated?
1134 		 */
1135 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1136 			continue;
1137 
1138 		/*
1139 		 * Ignore the address if it is not associated with a prefix
1140 		 * or is associated with a prefix that is different from this
1141 		 * one.  (pr is never NULL here)
1142 		 */
1143 		if (ifa6->ia6_ndpr != pr)
1144 			continue;
1145 
1146 		if (ia6_match == NULL) /* remember the first one */
1147 			ia6_match = ifa6;
1148 
1149 		/*
1150 		 * An already autoconfigured address matched.  Now that we
1151 		 * are sure there is at least one matched address, we can
1152 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1153 		 * "two hours" rule and the privacy extension.
1154 		 * We apply some clarifications in rfc2462bis:
1155 		 * - use remaininglifetime instead of storedlifetime as a
1156 		 *   variable name
1157 		 * - remove the dead code in the "two-hour" rule
1158 		 */
1159 #define TWOHOUR		(120*60)
1160 		lt6_tmp = ifa6->ia6_lifetime;
1161 
1162 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1163 			remaininglifetime = ND6_INFINITE_LIFETIME;
1164 		else if (time_second - ifa6->ia6_updatetime >
1165 			 lt6_tmp.ia6t_vltime) {
1166 			/*
1167 			 * The case of "invalid" address.  We should usually
1168 			 * not see this case.
1169 			 */
1170 			remaininglifetime = 0;
1171 		} else
1172 			remaininglifetime = lt6_tmp.ia6t_vltime -
1173 			    (time_second - ifa6->ia6_updatetime);
1174 
1175 		/* when not updating, keep the current stored lifetime. */
1176 		lt6_tmp.ia6t_vltime = remaininglifetime;
1177 
1178 		if (TWOHOUR < new->ndpr_vltime ||
1179 		    remaininglifetime < new->ndpr_vltime) {
1180 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1181 		} else if (remaininglifetime <= TWOHOUR) {
1182 			if (auth) {
1183 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1184 			}
1185 		} else {
1186 			/*
1187 			 * new->ndpr_vltime <= TWOHOUR &&
1188 			 * TWOHOUR < remaininglifetime
1189 			 */
1190 			lt6_tmp.ia6t_vltime = TWOHOUR;
1191 		}
1192 
1193 		/* The 2 hour rule is not imposed for preferred lifetime. */
1194 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1195 
1196 		in6_init_address_ltimes(pr, &lt6_tmp);
1197 
1198   		/*
1199 		 * We need to treat lifetimes for temporary addresses
1200 		 * differently, according to
1201 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1202 		 * we only update the lifetimes when they are in the maximum
1203 		 * intervals.
1204   		 */
1205   		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1206 			u_int32_t maxvltime, maxpltime;
1207 
1208 			if (ip6_temp_valid_lifetime >
1209 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1210 			    ip6_desync_factor)) {
1211 				maxvltime = ip6_temp_valid_lifetime -
1212 				    (time_second - ifa6->ia6_createtime) -
1213 				    ip6_desync_factor;
1214 			} else
1215 				maxvltime = 0;
1216 			if (ip6_temp_preferred_lifetime >
1217 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1218 			    ip6_desync_factor)) {
1219 				maxpltime = ip6_temp_preferred_lifetime -
1220 				    (time_second - ifa6->ia6_createtime) -
1221 				    ip6_desync_factor;
1222 			} else
1223 				maxpltime = 0;
1224 
1225 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1226 			    lt6_tmp.ia6t_vltime > maxvltime) {
1227 				lt6_tmp.ia6t_vltime = maxvltime;
1228 			}
1229 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1230 			    lt6_tmp.ia6t_pltime > maxpltime) {
1231 				lt6_tmp.ia6t_pltime = maxpltime;
1232 			}
1233 		}
1234 		ifa6->ia6_lifetime = lt6_tmp;
1235 		ifa6->ia6_updatetime = time_second;
1236 	}
1237 	if (ia6_match == NULL && new->ndpr_vltime) {
1238 		int ifidlen;
1239 
1240 		/*
1241 		 * 5.5.3 (d) (continued)
1242 		 * No address matched and the valid lifetime is non-zero.
1243 		 * Create a new address.
1244 		 */
1245 
1246 		/*
1247 		 * Prefix Length check:
1248 		 * If the sum of the prefix length and interface identifier
1249 		 * length does not equal 128 bits, the Prefix Information
1250 		 * option MUST be ignored.  The length of the interface
1251 		 * identifier is defined in a separate link-type specific
1252 		 * document.
1253 		 */
1254 		ifidlen = in6_if2idlen(ifp);
1255 		if (ifidlen < 0) {
1256 			/* this should not happen, so we always log it. */
1257 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1258 			    if_name(ifp));
1259 			goto end;
1260 		}
1261 		if (ifidlen + pr->ndpr_plen != 128) {
1262 			nd6log((LOG_INFO,
1263 			    "prelist_update: invalid prefixlen "
1264 			    "%d for %s, ignored\n",
1265 			    pr->ndpr_plen, if_name(ifp)));
1266 			goto end;
1267 		}
1268 
1269 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1270 			/*
1271 			 * note that we should use pr (not new) for reference.
1272 			 */
1273 			pr->ndpr_refcnt++;
1274 			ia6->ia6_ndpr = pr;
1275 
1276 			/*
1277 			 * RFC 3041 3.3 (2).
1278 			 * When a new public address is created as described
1279 			 * in RFC2462, also create a new temporary address.
1280 			 *
1281 			 * RFC 3041 3.5.
1282 			 * When an interface connects to a new link, a new
1283 			 * randomized interface identifier should be generated
1284 			 * immediately together with a new set of temporary
1285 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1286 			 * in6_tmpifadd().
1287 			 */
1288 			if (ip6_use_tempaddr) {
1289 				int e;
1290 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1291 					nd6log((LOG_NOTICE, "prelist_update: "
1292 					    "failed to create a temporary "
1293 					    "address, errno=%d\n",
1294 					    e));
1295 				}
1296 			}
1297 
1298 			/*
1299 			 * A newly added address might affect the status
1300 			 * of other addresses, so we check and update it.
1301 			 * XXX: what if address duplication happens?
1302 			 */
1303 			pfxlist_onlink_check();
1304 		} else {
1305 			/* just set an error. do not bark here. */
1306 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1307 		}
1308 	}
1309 
1310  end:
1311 	splx(s);
1312 	return error;
1313 }
1314 
1315 /*
1316  * A supplement function used in the on-link detection below;
1317  * detect if a given prefix has a (probably) reachable advertising router.
1318  * XXX: lengthy function name...
1319  */
1320 static struct nd_pfxrouter *
1321 find_pfxlist_reachable_router(pr)
1322 	struct nd_prefix *pr;
1323 {
1324 	struct nd_pfxrouter *pfxrtr;
1325 	struct rtentry *rt;
1326 	struct llinfo_nd6 *ln;
1327 
1328 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1329 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1330 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1331 		    pfxrtr->router->ifp)) &&
1332 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1333 		    ND6_IS_LLINFO_PROBREACH(ln))
1334 			break;	/* found */
1335 	}
1336 
1337 	return (pfxrtr);
1338 }
1339 
1340 /*
1341  * Check if each prefix in the prefix list has at least one available router
1342  * that advertised the prefix (a router is "available" if its neighbor cache
1343  * entry is reachable or probably reachable).
1344  * If the check fails, the prefix may be off-link, because, for example,
1345  * we have moved from the network but the lifetime of the prefix has not
1346  * expired yet.  So we should not use the prefix if there is another prefix
1347  * that has an available router.
1348  * But, if there is no prefix that has an available router, we still regards
1349  * all the prefixes as on-link.  This is because we can't tell if all the
1350  * routers are simply dead or if we really moved from the network and there
1351  * is no router around us.
1352  */
1353 void
1354 pfxlist_onlink_check()
1355 {
1356 	struct nd_prefix *pr;
1357 	struct in6_ifaddr *ifa;
1358 	struct nd_defrouter *dr;
1359 	struct nd_pfxrouter *pfxrtr = NULL;
1360 
1361 	/*
1362 	 * Check if there is a prefix that has a reachable advertising
1363 	 * router.
1364 	 */
1365 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1366 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1367 			break;
1368 	}
1369 
1370 	/*
1371 	 * If we have no such prefix, check whether we still have a router
1372 	 * that does not advertise any prefixes.
1373 	 */
1374 	if (pr == NULL) {
1375 		for (dr = TAILQ_FIRST(&nd_defrouter); dr;
1376 		    dr = TAILQ_NEXT(dr, dr_entry)) {
1377 			struct nd_prefix *pr0;
1378 
1379 			for (pr0 = nd_prefix.lh_first; pr0;
1380 			    pr0 = pr0->ndpr_next) {
1381 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1382 					break;
1383 			}
1384 			if (pfxrtr != NULL)
1385 				break;
1386 		}
1387 	}
1388 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && pfxrtr == NULL)) {
1389   		/*
1390 		 * There is at least one prefix that has a reachable router,
1391 		 * or at least a router which probably does not advertise
1392 		 * any prefixes.  The latter would be the case when we move
1393 		 * to a new link where we have a router that does not provide
1394 		 * prefixes and we configure an address by hand.
1395   		 * Detach prefixes which have no reachable advertising
1396   		 * router, and attach other prefixes.
1397   		 */
1398 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1399 			/* XXX: a link-local prefix should never be detached */
1400 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1401 				continue;
1402 
1403 			/*
1404 			 * we aren't interested in prefixes without the L bit
1405 			 * set.
1406 			 */
1407 			if (pr->ndpr_raf_onlink == 0)
1408 				continue;
1409 
1410 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1411 			    find_pfxlist_reachable_router(pr) == NULL)
1412 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1413 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1414 			    find_pfxlist_reachable_router(pr) != 0)
1415 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1416 		}
1417 	} else {
1418 		/* there is no prefix that has a reachable router */
1419 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1420 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1421 				continue;
1422 
1423 			if (pr->ndpr_raf_onlink == 0)
1424 				continue;
1425 
1426 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1427 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1428 		}
1429 	}
1430 
1431 	/*
1432 	 * Remove each interface route associated with a (just) detached
1433 	 * prefix, and reinstall the interface route for a (just) attached
1434 	 * prefix.  Note that all attempt of reinstallation does not
1435 	 * necessarily success, when a same prefix is shared among multiple
1436 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1437 	 * so we don't have to care about them.
1438 	 */
1439 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1440 		int e;
1441 
1442 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1443 			continue;
1444 
1445 		if (pr->ndpr_raf_onlink == 0)
1446 			continue;
1447 
1448 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1449 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1450 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1451 				nd6log((LOG_ERR,
1452 				    "pfxlist_onlink_check: failed to "
1453 				    "make %s/%d offlink, errno=%d\n",
1454 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1455 				    pr->ndpr_plen, e));
1456 			}
1457 		}
1458 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1459 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1460 		    pr->ndpr_raf_onlink) {
1461 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1462 				nd6log((LOG_ERR,
1463 				    "pfxlist_onlink_check: failed to "
1464 				    "make %s/%d onlink, errno=%d\n",
1465 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1466 				    pr->ndpr_plen, e));
1467 			}
1468 		}
1469 	}
1470 
1471 	/*
1472 	 * Changes on the prefix status might affect address status as well.
1473 	 * Make sure that all addresses derived from an attached prefix are
1474 	 * attached, and that all addresses derived from a detached prefix are
1475 	 * detached.  Note, however, that a manually configured address should
1476 	 * always be attached.
1477 	 * The precise detection logic is same as the one for prefixes.
1478 	 */
1479 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1480 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1481 			continue;
1482 
1483 		if (ifa->ia6_ndpr == NULL) {
1484 			/*
1485 			 * This can happen when we first configure the address
1486 			 * (i.e. the address exists, but the prefix does not).
1487 			 * XXX: complicated relationships...
1488 			 */
1489 			continue;
1490 		}
1491 
1492 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1493 			break;
1494 	}
1495 	if (ifa) {
1496 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1497 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1498 				continue;
1499 
1500 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1501 				continue;
1502 
1503 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1504 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1505 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1506 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1507 					nd6_dad_start((struct ifaddr *)ifa, 0);
1508 				}
1509 			} else {
1510 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1511 			}
1512 		}
1513 	}
1514 	else {
1515 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1516 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1517 				continue;
1518 
1519 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1520 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1521 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1522 				/* Do we need a delay in this case? */
1523 				nd6_dad_start((struct ifaddr *)ifa, 0);
1524 			}
1525 		}
1526 	}
1527 }
1528 
1529 int
1530 nd6_prefix_onlink(pr)
1531 	struct nd_prefix *pr;
1532 {
1533 	struct ifaddr *ifa;
1534 	struct ifnet *ifp = pr->ndpr_ifp;
1535 	struct sockaddr_in6 mask6;
1536 	struct nd_prefix *opr;
1537 	u_long rtflags;
1538 	int error = 0;
1539 	struct rtentry *rt = NULL;
1540 
1541 	/* sanity check */
1542 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1543 		nd6log((LOG_ERR,
1544 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1545 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1546 		return (EEXIST);
1547 	}
1548 
1549 	/*
1550 	 * Add the interface route associated with the prefix.  Before
1551 	 * installing the route, check if there's the same prefix on another
1552 	 * interface, and the prefix has already installed the interface route.
1553 	 * Although such a configuration is expected to be rare, we explicitly
1554 	 * allow it.
1555 	 */
1556 	for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1557 		if (opr == pr)
1558 			continue;
1559 
1560 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1561 			continue;
1562 
1563 		if (opr->ndpr_plen == pr->ndpr_plen &&
1564 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1565 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1566 			return (0);
1567 	}
1568 
1569 	/*
1570 	 * We prefer link-local addresses as the associated interface address.
1571 	 */
1572 	/* search for a link-local addr */
1573 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1574 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1575 	if (ifa == NULL) {
1576 		/* XXX: freebsd does not have ifa_ifwithaf */
1577 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1578 			if (ifa->ifa_addr->sa_family == AF_INET6)
1579 				break;
1580 		}
1581 		/* should we care about ia6_flags? */
1582 	}
1583 	if (ifa == NULL) {
1584 		/*
1585 		 * This can still happen, when, for example, we receive an RA
1586 		 * containing a prefix with the L bit set and the A bit clear,
1587 		 * after removing all IPv6 addresses on the receiving
1588 		 * interface.  This should, of course, be rare though.
1589 		 */
1590 		nd6log((LOG_NOTICE,
1591 		    "nd6_prefix_onlink: failed to find any ifaddr"
1592 		    " to add route for a prefix(%s/%d) on %s\n",
1593 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1594 		    pr->ndpr_plen, if_name(ifp)));
1595 		return (0);
1596 	}
1597 
1598 	/*
1599 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1600 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1601 	 */
1602 	bzero(&mask6, sizeof(mask6));
1603 	mask6.sin6_len = sizeof(mask6);
1604 	mask6.sin6_addr = pr->ndpr_mask;
1605 	rtflags = ifa->ifa_flags | RTF_CLONING | RTF_UP;
1606 	if (nd6_need_cache(ifp)) {
1607 		/* explicitly set in case ifa_flags does not set the flag. */
1608 		rtflags |= RTF_CLONING;
1609 	} else {
1610 		/*
1611 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1612 		 */
1613 		rtflags &= ~RTF_CLONING;
1614 	}
1615 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1616 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1617 	if (error == 0) {
1618 		if (rt != NULL) /* this should be non NULL, though */
1619 			nd6_rtmsg(RTM_ADD, rt);
1620 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1621 	} else {
1622 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1623 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1624 		    "errno = %d\n",
1625 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1626 		    pr->ndpr_plen, if_name(ifp),
1627 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1628 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
1629 	}
1630 
1631 	if (rt != NULL) {
1632 		RT_LOCK(rt);
1633 		RT_REMREF(rt);
1634 		RT_UNLOCK(rt);
1635 	}
1636 
1637 	return (error);
1638 }
1639 
1640 int
1641 nd6_prefix_offlink(pr)
1642 	struct nd_prefix *pr;
1643 {
1644 	int error = 0;
1645 	struct ifnet *ifp = pr->ndpr_ifp;
1646 	struct nd_prefix *opr;
1647 	struct sockaddr_in6 sa6, mask6;
1648 	struct rtentry *rt = NULL;
1649 
1650 	/* sanity check */
1651 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1652 		nd6log((LOG_ERR,
1653 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1654 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1655 		return (EEXIST);
1656 	}
1657 
1658 	bzero(&sa6, sizeof(sa6));
1659 	sa6.sin6_family = AF_INET6;
1660 	sa6.sin6_len = sizeof(sa6);
1661 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1662 	    sizeof(struct in6_addr));
1663 	bzero(&mask6, sizeof(mask6));
1664 	mask6.sin6_family = AF_INET6;
1665 	mask6.sin6_len = sizeof(sa6);
1666 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1667 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1668 	    (struct sockaddr *)&mask6, 0, &rt);
1669 	if (error == 0) {
1670 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1671 
1672 		/* report the route deletion to the routing socket. */
1673 		if (rt != NULL)
1674 			nd6_rtmsg(RTM_DELETE, rt);
1675 
1676 		/*
1677 		 * There might be the same prefix on another interface,
1678 		 * the prefix which could not be on-link just because we have
1679 		 * the interface route (see comments in nd6_prefix_onlink).
1680 		 * If there's one, try to make the prefix on-link on the
1681 		 * interface.
1682 		 */
1683 		for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1684 			if (opr == pr)
1685 				continue;
1686 
1687 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1688 				continue;
1689 
1690 			/*
1691 			 * KAME specific: detached prefixes should not be
1692 			 * on-link.
1693 			 */
1694 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1695 				continue;
1696 
1697 			if (opr->ndpr_plen == pr->ndpr_plen &&
1698 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1699 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1700 				int e;
1701 
1702 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1703 					nd6log((LOG_ERR,
1704 					    "nd6_prefix_offlink: failed to "
1705 					    "recover a prefix %s/%d from %s "
1706 					    "to %s (errno = %d)\n",
1707 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1708 					    opr->ndpr_plen, if_name(ifp),
1709 					    if_name(opr->ndpr_ifp), e));
1710 				}
1711 			}
1712 		}
1713 	} else {
1714 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1715 		nd6log((LOG_ERR,
1716 		    "nd6_prefix_offlink: failed to delete route: "
1717 		    "%s/%d on %s (errno = %d)\n",
1718 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1719 		    error));
1720 	}
1721 
1722 	if (rt != NULL) {
1723 		RTFREE(rt);
1724 	}
1725 
1726 	return (error);
1727 }
1728 
1729 static struct in6_ifaddr *
1730 in6_ifadd(pr, mcast)
1731 	struct nd_prefixctl *pr;
1732 	int mcast;
1733 {
1734 	struct ifnet *ifp = pr->ndpr_ifp;
1735 	struct ifaddr *ifa;
1736 	struct in6_aliasreq ifra;
1737 	struct in6_ifaddr *ia, *ib;
1738 	int error, plen0;
1739 	struct in6_addr mask;
1740 	int prefixlen = pr->ndpr_plen;
1741 	int updateflags;
1742 
1743 	in6_prefixlen2mask(&mask, prefixlen);
1744 
1745 	/*
1746 	 * find a link-local address (will be interface ID).
1747 	 * Is it really mandatory? Theoretically, a global or a site-local
1748 	 * address can be configured without a link-local address, if we
1749 	 * have a unique interface identifier...
1750 	 *
1751 	 * it is not mandatory to have a link-local address, we can generate
1752 	 * interface identifier on the fly.  we do this because:
1753 	 * (1) it should be the easiest way to find interface identifier.
1754 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1755 	 * for multiple addresses on a single interface, and possible shortcut
1756 	 * of DAD.  we omitted DAD for this reason in the past.
1757 	 * (3) a user can prevent autoconfiguration of global address
1758 	 * by removing link-local address by hand (this is partly because we
1759 	 * don't have other way to control the use of IPv6 on an interface.
1760 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1761 	 * (4) it is easier to manage when an interface has addresses
1762 	 * with the same interface identifier, than to have multiple addresses
1763 	 * with different interface identifiers.
1764 	 */
1765 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1766 	if (ifa)
1767 		ib = (struct in6_ifaddr *)ifa;
1768 	else
1769 		return NULL;
1770 
1771 	/* prefixlen + ifidlen must be equal to 128 */
1772 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1773 	if (prefixlen != plen0) {
1774 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1775 		    "(prefix=%d ifid=%d)\n",
1776 		    if_name(ifp), prefixlen, 128 - plen0));
1777 		return NULL;
1778 	}
1779 
1780 	/* make ifaddr */
1781 
1782 	bzero(&ifra, sizeof(ifra));
1783 	/*
1784 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1785 	 * for safety.
1786 	 */
1787 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1788 	ifra.ifra_addr.sin6_family = AF_INET6;
1789 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1790 	/* prefix */
1791 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1792 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1793 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1794 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1795 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1796 
1797 	/* interface ID */
1798 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1799 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1800 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1801 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1802 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1803 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1804 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1805 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1806 
1807 	/* new prefix mask. */
1808 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1809 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1810 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1811 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1812 
1813 	/* lifetimes. */
1814 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1815 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1816 
1817 	/* XXX: scope zone ID? */
1818 
1819 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1820 
1821   	/*
1822 	 * Make sure that we do not have this address already.  This should
1823 	 * usually not happen, but we can still see this case, e.g., if we
1824 	 * have manually configured the exact address to be configured.
1825 	 */
1826 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1827 		/* this should be rare enough to make an explicit log */
1828 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1829 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1830 		return (NULL);
1831 	}
1832 
1833 	/*
1834 	 * Allocate ifaddr structure, link into chain, etc.
1835 	 * If we are going to create a new address upon receiving a multicasted
1836 	 * RA, we need to impose a random delay before starting DAD.
1837 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1838 	 */
1839 	updateflags = 0;
1840 	if (mcast)
1841 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1842 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1843 		nd6log((LOG_ERR,
1844 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1845 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1846 		    error));
1847 		return (NULL);	/* ifaddr must not have been allocated. */
1848 	}
1849 
1850 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1851 
1852 	return (ia);		/* this is always non-NULL */
1853 }
1854 
1855 int
1856 in6_tmpifadd(ia0, forcegen, delay)
1857 	const struct in6_ifaddr *ia0; /* corresponding public address */
1858 	int forcegen, delay;
1859 {
1860 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1861 	struct in6_ifaddr *newia, *ia;
1862 	struct in6_aliasreq ifra;
1863 	int i, error;
1864 	int trylimit = 3;	/* XXX: adhoc value */
1865 	int updateflags;
1866 	u_int32_t randid[2];
1867 	time_t vltime0, pltime0;
1868 
1869 	bzero(&ifra, sizeof(ifra));
1870 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1871 	ifra.ifra_addr = ia0->ia_addr;
1872 	/* copy prefix mask */
1873 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1874 	/* clear the old IFID */
1875 	for (i = 0; i < 4; i++) {
1876 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1877 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1878 	}
1879 
1880   again:
1881 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1882 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1883 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1884 		    "random IFID\n"));
1885 		return (EINVAL);
1886 	}
1887 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1888 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1889 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1890 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1891 
1892   	/*
1893 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1894 	 * However, we may still have a chance to see collision, because
1895 	 * there may be a time lag between generation of the ID and generation
1896 	 * of the address.  So, we'll do one more sanity check.
1897 	 */
1898 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1899 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1900 		    &ifra.ifra_addr.sin6_addr)) {
1901   			if (trylimit-- == 0) {
1902 				/*
1903 				 * Give up.  Something strange should have
1904 				 * happened.
1905 				 */
1906 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1907 				    "find a unique random IFID\n"));
1908 				return (EEXIST);
1909 			}
1910 			forcegen = 1;
1911 			goto again;
1912 		}
1913 	}
1914 
1915 	/*
1916 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1917          * public address or TEMP_VALID_LIFETIME.
1918 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1919          * of the public address or TEMP_PREFERRED_LIFETIME -
1920          * DESYNC_FACTOR.
1921 	 */
1922 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1923 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1924 		    (ia0->ia6_lifetime.ia6t_vltime -
1925 		    (time_second - ia0->ia6_updatetime));
1926 		if (vltime0 > ip6_temp_valid_lifetime)
1927 			vltime0 = ip6_temp_valid_lifetime;
1928 	} else
1929 		vltime0 = ip6_temp_valid_lifetime;
1930 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1931 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1932 		    (ia0->ia6_lifetime.ia6t_pltime -
1933 		    (time_second - ia0->ia6_updatetime));
1934 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
1935 			pltime0 = ip6_temp_preferred_lifetime -
1936 			    ip6_desync_factor;
1937 		}
1938 	} else
1939 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
1940 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1941 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1942 
1943 	/*
1944 	 * A temporary address is created only if this calculated Preferred
1945 	 * Lifetime is greater than REGEN_ADVANCE time units.
1946 	 */
1947 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
1948 		return (0);
1949 
1950 	/* XXX: scope zone ID? */
1951 
1952 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1953 
1954 	/* allocate ifaddr structure, link into chain, etc. */
1955 	updateflags = 0;
1956 	if (delay)
1957 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1958 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1959 		return (error);
1960 
1961 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1962 	if (newia == NULL) {	/* XXX: can it happen? */
1963 		nd6log((LOG_ERR,
1964 		    "in6_tmpifadd: ifa update succeeded, but we got "
1965 		    "no ifaddr\n"));
1966 		return (EINVAL); /* XXX */
1967 	}
1968 	newia->ia6_ndpr = ia0->ia6_ndpr;
1969 	newia->ia6_ndpr->ndpr_refcnt++;
1970 
1971 	/*
1972 	 * A newly added address might affect the status of other addresses.
1973 	 * XXX: when the temporary address is generated with a new public
1974 	 * address, the onlink check is redundant.  However, it would be safe
1975 	 * to do the check explicitly everywhere a new address is generated,
1976 	 * and, in fact, we surely need the check when we create a new
1977 	 * temporary address due to deprecation of an old temporary address.
1978 	 */
1979 	pfxlist_onlink_check();
1980 
1981 	return (0);
1982 }
1983 
1984 static int
1985 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1986 {
1987 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1988 		ndpr->ndpr_preferred = 0;
1989 	else
1990 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
1991 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1992 		ndpr->ndpr_expire = 0;
1993 	else
1994 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
1995 
1996 	return 0;
1997 }
1998 
1999 static void
2000 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2001 {
2002 	/* init ia6t_expire */
2003 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2004 		lt6->ia6t_expire = 0;
2005 	else {
2006 		lt6->ia6t_expire = time_second;
2007 		lt6->ia6t_expire += lt6->ia6t_vltime;
2008 	}
2009 
2010 	/* init ia6t_preferred */
2011 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2012 		lt6->ia6t_preferred = 0;
2013 	else {
2014 		lt6->ia6t_preferred = time_second;
2015 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2016 	}
2017 }
2018 
2019 /*
2020  * Delete all the routing table entries that use the specified gateway.
2021  * XXX: this function causes search through all entries of routing table, so
2022  * it shouldn't be called when acting as a router.
2023  */
2024 void
2025 rt6_flush(gateway, ifp)
2026 	struct in6_addr *gateway;
2027 	struct ifnet *ifp;
2028 {
2029 	struct radix_node_head *rnh = rt_tables[AF_INET6];
2030 	int s = splnet();
2031 
2032 	/* We'll care only link-local addresses */
2033 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2034 		splx(s);
2035 		return;
2036 	}
2037 
2038 	RADIX_NODE_HEAD_LOCK(rnh);
2039 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2040 	RADIX_NODE_HEAD_UNLOCK(rnh);
2041 	splx(s);
2042 }
2043 
2044 static int
2045 rt6_deleteroute(rn, arg)
2046 	struct radix_node *rn;
2047 	void *arg;
2048 {
2049 #define SIN6(s)	((struct sockaddr_in6 *)s)
2050 	struct rtentry *rt = (struct rtentry *)rn;
2051 	struct in6_addr *gate = (struct in6_addr *)arg;
2052 
2053 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2054 		return (0);
2055 
2056 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2057 		return (0);
2058 	}
2059 
2060 	/*
2061 	 * Do not delete a static route.
2062 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2063 	 * 'cloned' bit instead?
2064 	 */
2065 	if ((rt->rt_flags & RTF_STATIC) != 0)
2066 		return (0);
2067 
2068 	/*
2069 	 * We delete only host route. This means, in particular, we don't
2070 	 * delete default route.
2071 	 */
2072 	if ((rt->rt_flags & RTF_HOST) == 0)
2073 		return (0);
2074 
2075 	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2076 	    rt_mask(rt), rt->rt_flags, 0));
2077 #undef SIN6
2078 }
2079 
2080 int
2081 nd6_setdefaultiface(ifindex)
2082 	int ifindex;
2083 {
2084 	int error = 0;
2085 
2086 	if (ifindex < 0 || if_index < ifindex)
2087 		return (EINVAL);
2088 	if (ifindex != 0 && !ifnet_byindex(ifindex))
2089 		return (EINVAL);
2090 
2091 	if (nd6_defifindex != ifindex) {
2092 		nd6_defifindex = ifindex;
2093 		if (nd6_defifindex > 0)
2094 			nd6_defifp = ifnet_byindex(nd6_defifindex);
2095 		else
2096 			nd6_defifp = NULL;
2097 
2098 		/*
2099 		 * Our current implementation assumes one-to-one maping between
2100 		 * interfaces and links, so it would be natural to use the
2101 		 * default interface as the default link.
2102 		 */
2103 		scope6_setdefault(nd6_defifp);
2104 	}
2105 
2106 	return (error);
2107 }
2108