xref: /freebsd/sys/netinet6/nd6_rtr.c (revision 10b59a9b4add0320d52c15ce057dd697261e7dfc)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/socket.h>
43 #include <sys/sockio.h>
44 #include <sys/time.h>
45 #include <sys/kernel.h>
46 #include <sys/lock.h>
47 #include <sys/errno.h>
48 #include <sys/rwlock.h>
49 #include <sys/syslog.h>
50 #include <sys/queue.h>
51 
52 #include <net/if.h>
53 #include <net/if_types.h>
54 #include <net/if_dl.h>
55 #include <net/route.h>
56 #include <net/radix.h>
57 #include <net/vnet.h>
58 
59 #include <netinet/in.h>
60 #include <net/if_llatbl.h>
61 #include <netinet6/in6_var.h>
62 #include <netinet6/in6_ifattach.h>
63 #include <netinet/ip6.h>
64 #include <netinet6/ip6_var.h>
65 #include <netinet6/nd6.h>
66 #include <netinet/icmp6.h>
67 #include <netinet6/scope6_var.h>
68 
69 static int rtpref(struct nd_defrouter *);
70 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
71 static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
72     struct mbuf *, int));
73 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *,	int);
74 static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
75 	struct nd_defrouter *));
76 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
77 static void pfxrtr_del(struct nd_pfxrouter *);
78 static struct nd_pfxrouter *find_pfxlist_reachable_router
79 (struct nd_prefix *);
80 static void defrouter_delreq(struct nd_defrouter *);
81 static void nd6_rtmsg(int, struct rtentry *);
82 
83 static int in6_init_prefix_ltimes(struct nd_prefix *);
84 static void in6_init_address_ltimes __P((struct nd_prefix *,
85 	struct in6_addrlifetime *));
86 
87 static int rt6_deleteroute(struct radix_node *, void *);
88 
89 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
90 #define	V_nd6_recalc_reachtm_interval	VNET(nd6_recalc_reachtm_interval)
91 
92 static VNET_DEFINE(struct ifnet *, nd6_defifp);
93 VNET_DEFINE(int, nd6_defifindex);
94 #define	V_nd6_defifp			VNET(nd6_defifp)
95 
96 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
97 
98 VNET_DEFINE(int, ip6_desync_factor);
99 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
100 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
101 
102 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
103 
104 /* RTPREF_MEDIUM has to be 0! */
105 #define RTPREF_HIGH	1
106 #define RTPREF_MEDIUM	0
107 #define RTPREF_LOW	(-1)
108 #define RTPREF_RESERVED	(-2)
109 #define RTPREF_INVALID	(-3)	/* internal */
110 
111 /*
112  * Receive Router Solicitation Message - just for routers.
113  * Router solicitation/advertisement is mostly managed by userland program
114  * (rtadvd) so here we have no function like nd6_ra_output().
115  *
116  * Based on RFC 2461
117  */
118 void
119 nd6_rs_input(struct mbuf *m, int off, int 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 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
129 
130 	/*
131 	 * Accept RS only when V_ip6_forwarding=1 and the interface has
132 	 * no ND6_IFF_ACCEPT_RTADV.
133 	 */
134 	if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
135 		goto freeit;
136 
137 	/* Sanity checks */
138 	if (ip6->ip6_hlim != 255) {
139 		nd6log((LOG_ERR,
140 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
141 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
142 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
143 		goto bad;
144 	}
145 
146 	/*
147 	 * Don't update the neighbor cache, if src = ::.
148 	 * This indicates that the src has no IP address assigned yet.
149 	 */
150 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
151 		goto freeit;
152 
153 #ifndef PULLDOWN_TEST
154 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
155 	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
156 #else
157 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
158 	if (nd_rs == NULL) {
159 		ICMP6STAT_INC(icp6s_tooshort);
160 		return;
161 	}
162 #endif
163 
164 	icmp6len -= sizeof(*nd_rs);
165 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
166 	if (nd6_options(&ndopts) < 0) {
167 		nd6log((LOG_INFO,
168 		    "nd6_rs_input: invalid ND option, ignored\n"));
169 		/* nd6_options have incremented stats */
170 		goto freeit;
171 	}
172 
173 	if (ndopts.nd_opts_src_lladdr) {
174 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
175 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
176 	}
177 
178 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
179 		nd6log((LOG_INFO,
180 		    "nd6_rs_input: lladdrlen mismatch for %s "
181 		    "(if %d, RS packet %d)\n",
182 		    ip6_sprintf(ip6bufs, &saddr6),
183 		    ifp->if_addrlen, lladdrlen - 2));
184 		goto bad;
185 	}
186 
187 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
188 
189  freeit:
190 	m_freem(m);
191 	return;
192 
193  bad:
194 	ICMP6STAT_INC(icp6s_badrs);
195 	m_freem(m);
196 }
197 
198 /*
199  * Receive Router Advertisement Message.
200  *
201  * Based on RFC 2461
202  * TODO: on-link bit on prefix information
203  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
204  */
205 void
206 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
207 {
208 	struct ifnet *ifp = m->m_pkthdr.rcvif;
209 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
210 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
211 	struct nd_router_advert *nd_ra;
212 	struct in6_addr saddr6 = ip6->ip6_src;
213 	int mcast = 0;
214 	union nd_opts ndopts;
215 	struct nd_defrouter *dr;
216 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
217 
218 	/*
219 	 * We only accept RAs only when the per-interface flag
220 	 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
221 	 */
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(ip6bufs, &ip6->ip6_src),
229 		    ip6_sprintf(ip6bufd, &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(ip6bufs, &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_INC(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 	/*
272 	 * Effectively-disable routes from RA messages when
273 	 * ND6_IFF_NO_RADR enabled on the receiving interface or
274 	 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
275 	 */
276 	if (ndi->flags & ND6_IFF_NO_RADR)
277 		dr0.rtlifetime = 0;
278 	else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
279 		dr0.rtlifetime = 0;
280 	else
281 		dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
282 	dr0.expire = time_second + dr0.rtlifetime;
283 	dr0.ifp = ifp;
284 	/* unspecified or not? (RFC 2461 6.3.4) */
285 	if (advreachable) {
286 		advreachable = ntohl(advreachable);
287 		if (advreachable <= MAX_REACHABLE_TIME &&
288 		    ndi->basereachable != advreachable) {
289 			ndi->basereachable = advreachable;
290 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
291 			ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
292 		}
293 	}
294 	if (nd_ra->nd_ra_retransmit)
295 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
296 	if (nd_ra->nd_ra_curhoplimit)
297 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
298 	dr = defrtrlist_update(&dr0);
299     }
300 
301 	/*
302 	 * prefix
303 	 */
304 	if (ndopts.nd_opts_pi) {
305 		struct nd_opt_hdr *pt;
306 		struct nd_opt_prefix_info *pi = NULL;
307 		struct nd_prefixctl pr;
308 
309 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
310 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
311 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
312 						(pt->nd_opt_len << 3))) {
313 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
314 				continue;
315 			pi = (struct nd_opt_prefix_info *)pt;
316 
317 			if (pi->nd_opt_pi_len != 4) {
318 				nd6log((LOG_INFO,
319 				    "nd6_ra_input: invalid option "
320 				    "len %d for prefix information option, "
321 				    "ignored\n", pi->nd_opt_pi_len));
322 				continue;
323 			}
324 
325 			if (128 < pi->nd_opt_pi_prefix_len) {
326 				nd6log((LOG_INFO,
327 				    "nd6_ra_input: invalid prefix "
328 				    "len %d for prefix information option, "
329 				    "ignored\n", pi->nd_opt_pi_prefix_len));
330 				continue;
331 			}
332 
333 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
334 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
335 				nd6log((LOG_INFO,
336 				    "nd6_ra_input: invalid prefix "
337 				    "%s, ignored\n",
338 				    ip6_sprintf(ip6bufs,
339 					&pi->nd_opt_pi_prefix)));
340 				continue;
341 			}
342 
343 			bzero(&pr, sizeof(pr));
344 			pr.ndpr_prefix.sin6_family = AF_INET6;
345 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
346 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
347 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
348 
349 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
350 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
351 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
352 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
353 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
354 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
355 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
356 			(void)prelist_update(&pr, dr, m, mcast);
357 		}
358 	}
359 
360 	/*
361 	 * MTU
362 	 */
363 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
364 		u_long mtu;
365 		u_long maxmtu;
366 
367 		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
368 
369 		/* lower bound */
370 		if (mtu < IPV6_MMTU) {
371 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
372 			    "mtu=%lu sent from %s, ignoring\n",
373 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
374 			goto skip;
375 		}
376 
377 		/* upper bound */
378 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
379 		    ? ndi->maxmtu : ifp->if_mtu;
380 		if (mtu <= maxmtu) {
381 			int change = (ndi->linkmtu != mtu);
382 
383 			ndi->linkmtu = mtu;
384 			if (change) /* in6_maxmtu may change */
385 				in6_setmaxmtu();
386 		} else {
387 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
388 			    "mtu=%lu sent from %s; "
389 			    "exceeds maxmtu %lu, ignoring\n",
390 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
391 		}
392 	}
393 
394  skip:
395 
396 	/*
397 	 * Source link layer address
398 	 */
399     {
400 	char *lladdr = NULL;
401 	int lladdrlen = 0;
402 
403 	if (ndopts.nd_opts_src_lladdr) {
404 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
405 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
406 	}
407 
408 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
409 		nd6log((LOG_INFO,
410 		    "nd6_ra_input: lladdrlen mismatch for %s "
411 		    "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
412 		    ifp->if_addrlen, lladdrlen - 2));
413 		goto bad;
414 	}
415 
416 	nd6_cache_lladdr(ifp, &saddr6, lladdr,
417 	    lladdrlen, ND_ROUTER_ADVERT, 0);
418 
419 	/*
420 	 * Installing a link-layer address might change the state of the
421 	 * router's neighbor cache, which might also affect our on-link
422 	 * detection of adveritsed prefixes.
423 	 */
424 	pfxlist_onlink_check();
425     }
426 
427  freeit:
428 	m_freem(m);
429 	return;
430 
431  bad:
432 	ICMP6STAT_INC(icp6s_badra);
433 	m_freem(m);
434 }
435 
436 /*
437  * default router list proccessing sub routines
438  */
439 
440 /* tell the change to user processes watching the routing socket. */
441 static void
442 nd6_rtmsg(int cmd, struct rtentry *rt)
443 {
444 	struct rt_addrinfo info;
445 	struct ifnet *ifp;
446 	struct ifaddr *ifa;
447 
448 	bzero((caddr_t)&info, sizeof(info));
449 	info.rti_info[RTAX_DST] = rt_key(rt);
450 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
451 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
452 	ifp = rt->rt_ifp;
453 	if (ifp != NULL) {
454 		IF_ADDR_LOCK(ifp);
455 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
456 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
457 		ifa_ref(ifa);
458 		IF_ADDR_UNLOCK(ifp);
459 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
460 	} else
461 		ifa = NULL;
462 
463 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
464 	if (ifa != NULL)
465 		ifa_free(ifa);
466 }
467 
468 void
469 defrouter_addreq(struct nd_defrouter *new)
470 {
471 	struct sockaddr_in6 def, mask, gate;
472 	struct rtentry *newrt = NULL;
473 	int s;
474 	int error;
475 
476 	bzero(&def, sizeof(def));
477 	bzero(&mask, sizeof(mask));
478 	bzero(&gate, sizeof(gate));
479 
480 	def.sin6_len = mask.sin6_len = gate.sin6_len =
481 	    sizeof(struct sockaddr_in6);
482 	def.sin6_family = gate.sin6_family = AF_INET6;
483 	gate.sin6_addr = new->rtaddr;
484 
485 	s = splnet();
486 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
487 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
488 	    RTF_GATEWAY, &newrt);
489 	if (newrt) {
490 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
491 		RTFREE(newrt);
492 	}
493 	if (error == 0)
494 		new->installed = 1;
495 	splx(s);
496 	return;
497 }
498 
499 struct nd_defrouter *
500 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
501 {
502 	struct nd_defrouter *dr;
503 
504 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
505 	     dr = TAILQ_NEXT(dr, dr_entry)) {
506 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
507 			return (dr);
508 	}
509 
510 	return (NULL);		/* search failed */
511 }
512 
513 /*
514  * Remove the default route for a given router.
515  * This is just a subroutine function for defrouter_select(), and should
516  * not be called from anywhere else.
517  */
518 static void
519 defrouter_delreq(struct nd_defrouter *dr)
520 {
521 	struct sockaddr_in6 def, mask, gate;
522 	struct rtentry *oldrt = NULL;
523 
524 	bzero(&def, sizeof(def));
525 	bzero(&mask, sizeof(mask));
526 	bzero(&gate, sizeof(gate));
527 
528 	def.sin6_len = mask.sin6_len = gate.sin6_len =
529 	    sizeof(struct sockaddr_in6);
530 	def.sin6_family = gate.sin6_family = AF_INET6;
531 	gate.sin6_addr = dr->rtaddr;
532 
533 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
534 	    (struct sockaddr *)&gate,
535 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
536 	if (oldrt) {
537 		nd6_rtmsg(RTM_DELETE, oldrt);
538 		RTFREE(oldrt);
539 	}
540 
541 	dr->installed = 0;
542 }
543 
544 /*
545  * remove all default routes from default router list
546  */
547 void
548 defrouter_reset(void)
549 {
550 	struct nd_defrouter *dr;
551 
552 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
553 	     dr = TAILQ_NEXT(dr, dr_entry))
554 		defrouter_delreq(dr);
555 
556 	/*
557 	 * XXX should we also nuke any default routers in the kernel, by
558 	 * going through them by rtalloc1()?
559 	 */
560 }
561 
562 void
563 defrtrlist_del(struct nd_defrouter *dr)
564 {
565 	struct nd_defrouter *deldr = NULL;
566 	struct nd_prefix *pr;
567 
568 	/*
569 	 * Flush all the routing table entries that use the router
570 	 * as a next hop.
571 	 */
572 	if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
573 		rt6_flush(&dr->rtaddr, dr->ifp);
574 
575 	if (dr->installed) {
576 		deldr = dr;
577 		defrouter_delreq(dr);
578 	}
579 	TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
580 
581 	/*
582 	 * Also delete all the pointers to the router in each prefix lists.
583 	 */
584 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
585 		struct nd_pfxrouter *pfxrtr;
586 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
587 			pfxrtr_del(pfxrtr);
588 	}
589 	pfxlist_onlink_check();
590 
591 	/*
592 	 * If the router is the primary one, choose a new one.
593 	 * Note that defrouter_select() will remove the current gateway
594 	 * from the routing table.
595 	 */
596 	if (deldr)
597 		defrouter_select();
598 
599 	free(dr, M_IP6NDP);
600 }
601 
602 /*
603  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
604  * draft-ietf-ipngwg-router-selection:
605  * 1) Routers that are reachable or probably reachable should be preferred.
606  *    If we have more than one (probably) reachable router, prefer ones
607  *    with the highest router preference.
608  * 2) When no routers on the list are known to be reachable or
609  *    probably reachable, routers SHOULD be selected in a round-robin
610  *    fashion, regardless of router preference values.
611  * 3) If the Default Router List is empty, assume that all
612  *    destinations are on-link.
613  *
614  * We assume nd_defrouter is sorted by router preference value.
615  * Since the code below covers both with and without router preference cases,
616  * we do not need to classify the cases by ifdef.
617  *
618  * At this moment, we do not try to install more than one default router,
619  * even when the multipath routing is available, because we're not sure about
620  * the benefits for stub hosts comparing to the risk of making the code
621  * complicated and the possibility of introducing bugs.
622  */
623 void
624 defrouter_select(void)
625 {
626 	int s = splnet();
627 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
628 	struct llentry *ln = NULL;
629 
630 	/*
631 	 * Let's handle easy case (3) first:
632 	 * If default router list is empty, there's nothing to be done.
633 	 */
634 	if (!TAILQ_FIRST(&V_nd_defrouter)) {
635 		splx(s);
636 		return;
637 	}
638 
639 	/*
640 	 * Search for a (probably) reachable router from the list.
641 	 * We just pick up the first reachable one (if any), assuming that
642 	 * the ordering rule of the list described in defrtrlist_update().
643 	 */
644 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
645 	     dr = TAILQ_NEXT(dr, dr_entry)) {
646 		IF_AFDATA_LOCK(dr->ifp);
647 		if (selected_dr == NULL &&
648 		    (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
649 		    ND6_IS_LLINFO_PROBREACH(ln)) {
650 			selected_dr = dr;
651 		}
652 		IF_AFDATA_UNLOCK(dr->ifp);
653 		if (ln != NULL) {
654 			LLE_RUNLOCK(ln);
655 			ln = NULL;
656 		}
657 
658 		if (dr->installed && installed_dr == NULL)
659 			installed_dr = dr;
660 		else if (dr->installed && installed_dr) {
661 			/* this should not happen.  warn for diagnosis. */
662 			log(LOG_ERR, "defrouter_select: more than one router"
663 			    " is installed\n");
664 		}
665 	}
666 	/*
667 	 * If none of the default routers was found to be reachable,
668 	 * round-robin the list regardless of preference.
669 	 * Otherwise, if we have an installed router, check if the selected
670 	 * (reachable) router should really be preferred to the installed one.
671 	 * We only prefer the new router when the old one is not reachable
672 	 * or when the new one has a really higher preference value.
673 	 */
674 	if (selected_dr == NULL) {
675 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
676 			selected_dr = TAILQ_FIRST(&V_nd_defrouter);
677 		else
678 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
679 	} else if (installed_dr) {
680 		IF_AFDATA_LOCK(installed_dr->ifp);
681 		if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
682 		    ND6_IS_LLINFO_PROBREACH(ln) &&
683 		    rtpref(selected_dr) <= rtpref(installed_dr)) {
684 			selected_dr = installed_dr;
685 		}
686 		IF_AFDATA_UNLOCK(installed_dr->ifp);
687 		if (ln != NULL)
688 			LLE_RUNLOCK(ln);
689 	}
690 
691 	/*
692 	 * If the selected router is different than the installed one,
693 	 * remove the installed router and install the selected one.
694 	 * Note that the selected router is never NULL here.
695 	 */
696 	if (installed_dr != selected_dr) {
697 		if (installed_dr)
698 			defrouter_delreq(installed_dr);
699 		defrouter_addreq(selected_dr);
700 	}
701 
702 	splx(s);
703 	return;
704 }
705 
706 /*
707  * for default router selection
708  * regards router-preference field as a 2-bit signed integer
709  */
710 static int
711 rtpref(struct nd_defrouter *dr)
712 {
713 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
714 	case ND_RA_FLAG_RTPREF_HIGH:
715 		return (RTPREF_HIGH);
716 	case ND_RA_FLAG_RTPREF_MEDIUM:
717 	case ND_RA_FLAG_RTPREF_RSV:
718 		return (RTPREF_MEDIUM);
719 	case ND_RA_FLAG_RTPREF_LOW:
720 		return (RTPREF_LOW);
721 	default:
722 		/*
723 		 * This case should never happen.  If it did, it would mean a
724 		 * serious bug of kernel internal.  We thus always bark here.
725 		 * Or, can we even panic?
726 		 */
727 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
728 		return (RTPREF_INVALID);
729 	}
730 	/* NOTREACHED */
731 }
732 
733 static struct nd_defrouter *
734 defrtrlist_update(struct nd_defrouter *new)
735 {
736 	struct nd_defrouter *dr, *n;
737 	int s = splnet();
738 
739 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
740 		/* entry exists */
741 		if (new->rtlifetime == 0) {
742 			defrtrlist_del(dr);
743 			dr = NULL;
744 		} else {
745 			int oldpref = rtpref(dr);
746 
747 			/* override */
748 			dr->flags = new->flags; /* xxx flag check */
749 			dr->rtlifetime = new->rtlifetime;
750 			dr->expire = new->expire;
751 
752 			/*
753 			 * If the preference does not change, there's no need
754 			 * to sort the entries. Also make sure the selected
755 			 * router is still installed in the kernel.
756 			 */
757 			if (dr->installed && rtpref(new) == oldpref) {
758 				splx(s);
759 				return (dr);
760 			}
761 
762 			/*
763 			 * preferred router may be changed, so relocate
764 			 * this router.
765 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
766 			 * However, since defrtrlist_del() has many side
767 			 * effects, we intentionally do so here.
768 			 * defrouter_select() below will handle routing
769 			 * changes later.
770 			 */
771 			TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
772 			n = dr;
773 			goto insert;
774 		}
775 		splx(s);
776 		return (dr);
777 	}
778 
779 	/* entry does not exist */
780 	if (new->rtlifetime == 0) {
781 		splx(s);
782 		return (NULL);
783 	}
784 
785 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
786 	if (n == NULL) {
787 		splx(s);
788 		return (NULL);
789 	}
790 	bzero(n, sizeof(*n));
791 	*n = *new;
792 
793 insert:
794 	/*
795 	 * Insert the new router in the Default Router List;
796 	 * The Default Router List should be in the descending order
797 	 * of router-preferece.  Routers with the same preference are
798 	 * sorted in the arriving time order.
799 	 */
800 
801 	/* insert at the end of the group */
802 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
803 	     dr = TAILQ_NEXT(dr, dr_entry)) {
804 		if (rtpref(n) > rtpref(dr))
805 			break;
806 	}
807 	if (dr)
808 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
809 	else
810 		TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
811 
812 	defrouter_select();
813 
814 	splx(s);
815 
816 	return (n);
817 }
818 
819 static struct nd_pfxrouter *
820 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
821 {
822 	struct nd_pfxrouter *search;
823 
824 	for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
825 		if (search->router == dr)
826 			break;
827 	}
828 
829 	return (search);
830 }
831 
832 static void
833 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
834 {
835 	struct nd_pfxrouter *new;
836 
837 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
838 	if (new == NULL)
839 		return;
840 	bzero(new, sizeof(*new));
841 	new->router = dr;
842 
843 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
844 
845 	pfxlist_onlink_check();
846 }
847 
848 static void
849 pfxrtr_del(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(struct nd_prefixctl *key)
857 {
858 	struct nd_prefix *search;
859 
860 	for (search = V_nd_prefix.lh_first;
861 	    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(struct nd_prefixctl *pr, struct nd_defrouter *dr,
875     struct nd_prefix **newp)
876 {
877 	struct nd_prefix *new = NULL;
878 	int error = 0;
879 	int i, s;
880 	char ip6buf[INET6_ADDRSTRLEN];
881 
882 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
883 	if (new == NULL)
884 		return(ENOMEM);
885 	bzero(new, sizeof(*new));
886 	new->ndpr_ifp = pr->ndpr_ifp;
887 	new->ndpr_prefix = pr->ndpr_prefix;
888 	new->ndpr_plen = pr->ndpr_plen;
889 	new->ndpr_vltime = pr->ndpr_vltime;
890 	new->ndpr_pltime = pr->ndpr_pltime;
891 	new->ndpr_flags = pr->ndpr_flags;
892 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
893 		free(new, M_IP6NDP);
894 		return(error);
895 	}
896 	new->ndpr_lastupdate = time_second;
897 	if (newp != NULL)
898 		*newp = new;
899 
900 	/* initialization */
901 	LIST_INIT(&new->ndpr_advrtrs);
902 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
903 	/* make prefix in the canonical form */
904 	for (i = 0; i < 4; i++)
905 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
906 		    new->ndpr_mask.s6_addr32[i];
907 
908 	s = splnet();
909 	/* link ndpr_entry to nd_prefix list */
910 	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
911 	splx(s);
912 
913 	/* ND_OPT_PI_FLAG_ONLINK processing */
914 	if (new->ndpr_raf_onlink) {
915 		int e;
916 
917 		if ((e = nd6_prefix_onlink(new)) != 0) {
918 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
919 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
920 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
921 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
922 			/* proceed anyway. XXX: is it correct? */
923 		}
924 	}
925 
926 	if (dr)
927 		pfxrtr_add(new, dr);
928 
929 	return 0;
930 }
931 
932 void
933 prelist_remove(struct nd_prefix *pr)
934 {
935 	struct nd_pfxrouter *pfr, *next;
936 	int e, s;
937 	char ip6buf[INET6_ADDRSTRLEN];
938 
939 	/* make sure to invalidate the prefix until it is really freed. */
940 	pr->ndpr_vltime = 0;
941 	pr->ndpr_pltime = 0;
942 
943 	/*
944 	 * Though these flags are now meaningless, we'd rather keep the value
945 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
946 	 * when executing "ndp -p".
947 	 */
948 
949 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
950 	    (e = nd6_prefix_offlink(pr)) != 0) {
951 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
952 		    "on %s, errno=%d\n",
953 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
954 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
955 		/* what should we do? */
956 	}
957 
958 	if (pr->ndpr_refcnt > 0)
959 		return;		/* notice here? */
960 
961 	s = splnet();
962 
963 	/* unlink ndpr_entry from nd_prefix list */
964 	LIST_REMOVE(pr, ndpr_entry);
965 
966 	/* free list of routers that adversed the prefix */
967 	for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
968 		next = pfr->pfr_next;
969 
970 		free(pfr, M_IP6NDP);
971 	}
972 	splx(s);
973 
974 	free(pr, M_IP6NDP);
975 
976 	pfxlist_onlink_check();
977 }
978 
979 /*
980  * dr - may be NULL
981  */
982 
983 static int
984 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
985     struct mbuf *m, 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 	char ip6buf[INET6_ADDRSTRLEN];
997 
998 	auth = 0;
999 	if (m) {
1000 		/*
1001 		 * Authenticity for NA consists authentication for
1002 		 * both IP header and IP datagrams, doesn't it ?
1003 		 */
1004 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1005 		auth = ((m->m_flags & M_AUTHIPHDR) &&
1006 		    (m->m_flags & M_AUTHIPDGM));
1007 #endif
1008 	}
1009 
1010 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1011 		/*
1012 		 * nd6_prefix_lookup() ensures that pr and new have the same
1013 		 * prefix on a same interface.
1014 		 */
1015 
1016 		/*
1017 		 * Update prefix information.  Note that the on-link (L) bit
1018 		 * and the autonomous (A) bit should NOT be changed from 1
1019 		 * to 0.
1020 		 */
1021 		if (new->ndpr_raf_onlink == 1)
1022 			pr->ndpr_raf_onlink = 1;
1023 		if (new->ndpr_raf_auto == 1)
1024 			pr->ndpr_raf_auto = 1;
1025 		if (new->ndpr_raf_onlink) {
1026 			pr->ndpr_vltime = new->ndpr_vltime;
1027 			pr->ndpr_pltime = new->ndpr_pltime;
1028 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1029 			pr->ndpr_lastupdate = time_second;
1030 		}
1031 
1032 		if (new->ndpr_raf_onlink &&
1033 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1034 			int e;
1035 
1036 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1037 				nd6log((LOG_ERR,
1038 				    "prelist_update: failed to make "
1039 				    "the prefix %s/%d on-link on %s "
1040 				    "(errno=%d)\n",
1041 				    ip6_sprintf(ip6buf,
1042 					    &pr->ndpr_prefix.sin6_addr),
1043 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1044 				/* proceed anyway. XXX: is it correct? */
1045 			}
1046 		}
1047 
1048 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1049 			pfxrtr_add(pr, dr);
1050 	} else {
1051 		struct nd_prefix *newpr = NULL;
1052 
1053 		newprefix = 1;
1054 
1055 		if (new->ndpr_vltime == 0)
1056 			goto end;
1057 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1058 			goto end;
1059 
1060 		error = nd6_prelist_add(new, dr, &newpr);
1061 		if (error != 0 || newpr == NULL) {
1062 			nd6log((LOG_NOTICE, "prelist_update: "
1063 			    "nd6_prelist_add failed for %s/%d on %s "
1064 			    "errno=%d, returnpr=%p\n",
1065 			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1066 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1067 			    error, newpr));
1068 			goto end; /* we should just give up in this case. */
1069 		}
1070 
1071 		/*
1072 		 * XXX: from the ND point of view, we can ignore a prefix
1073 		 * with the on-link bit being zero.  However, we need a
1074 		 * prefix structure for references from autoconfigured
1075 		 * addresses.  Thus, we explicitly make sure that the prefix
1076 		 * itself expires now.
1077 		 */
1078 		if (newpr->ndpr_raf_onlink == 0) {
1079 			newpr->ndpr_vltime = 0;
1080 			newpr->ndpr_pltime = 0;
1081 			in6_init_prefix_ltimes(newpr);
1082 		}
1083 
1084 		pr = newpr;
1085 	}
1086 
1087 	/*
1088 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1089 	 * Note that pr must be non NULL at this point.
1090 	 */
1091 
1092 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1093 	if (!new->ndpr_raf_auto)
1094 		goto end;
1095 
1096 	/*
1097 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1098 	 * nd6_ra_input.
1099 	 */
1100 
1101 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1102 	if (new->ndpr_pltime > new->ndpr_vltime) {
1103 		error = EINVAL;	/* XXX: won't be used */
1104 		goto end;
1105 	}
1106 
1107 	/*
1108 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1109 	 * an address configured by stateless autoconfiguration already in the
1110 	 * list of addresses associated with the interface, and the Valid
1111 	 * Lifetime is not 0, form an address.  We first check if we have
1112 	 * a matching prefix.
1113 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1114 	 * consider autoconfigured addresses while RFC2462 simply said
1115 	 * "address".
1116 	 */
1117 	IF_ADDR_LOCK(ifp);
1118 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1119 		struct in6_ifaddr *ifa6;
1120 		u_int32_t remaininglifetime;
1121 
1122 		if (ifa->ifa_addr->sa_family != AF_INET6)
1123 			continue;
1124 
1125 		ifa6 = (struct in6_ifaddr *)ifa;
1126 
1127 		/*
1128 		 * We only consider autoconfigured addresses as per rfc2462bis.
1129 		 */
1130 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1131 			continue;
1132 
1133 		/*
1134 		 * Spec is not clear here, but I believe we should concentrate
1135 		 * on unicast (i.e. not anycast) addresses.
1136 		 * XXX: other ia6_flags? detached or duplicated?
1137 		 */
1138 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1139 			continue;
1140 
1141 		/*
1142 		 * Ignore the address if it is not associated with a prefix
1143 		 * or is associated with a prefix that is different from this
1144 		 * one.  (pr is never NULL here)
1145 		 */
1146 		if (ifa6->ia6_ndpr != pr)
1147 			continue;
1148 
1149 		if (ia6_match == NULL) /* remember the first one */
1150 			ia6_match = ifa6;
1151 
1152 		/*
1153 		 * An already autoconfigured address matched.  Now that we
1154 		 * are sure there is at least one matched address, we can
1155 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1156 		 * "two hours" rule and the privacy extension.
1157 		 * We apply some clarifications in rfc2462bis:
1158 		 * - use remaininglifetime instead of storedlifetime as a
1159 		 *   variable name
1160 		 * - remove the dead code in the "two-hour" rule
1161 		 */
1162 #define TWOHOUR		(120*60)
1163 		lt6_tmp = ifa6->ia6_lifetime;
1164 
1165 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1166 			remaininglifetime = ND6_INFINITE_LIFETIME;
1167 		else if (time_second - ifa6->ia6_updatetime >
1168 			 lt6_tmp.ia6t_vltime) {
1169 			/*
1170 			 * The case of "invalid" address.  We should usually
1171 			 * not see this case.
1172 			 */
1173 			remaininglifetime = 0;
1174 		} else
1175 			remaininglifetime = lt6_tmp.ia6t_vltime -
1176 			    (time_second - ifa6->ia6_updatetime);
1177 
1178 		/* when not updating, keep the current stored lifetime. */
1179 		lt6_tmp.ia6t_vltime = remaininglifetime;
1180 
1181 		if (TWOHOUR < new->ndpr_vltime ||
1182 		    remaininglifetime < new->ndpr_vltime) {
1183 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1184 		} else if (remaininglifetime <= TWOHOUR) {
1185 			if (auth) {
1186 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1187 			}
1188 		} else {
1189 			/*
1190 			 * new->ndpr_vltime <= TWOHOUR &&
1191 			 * TWOHOUR < remaininglifetime
1192 			 */
1193 			lt6_tmp.ia6t_vltime = TWOHOUR;
1194 		}
1195 
1196 		/* The 2 hour rule is not imposed for preferred lifetime. */
1197 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1198 
1199 		in6_init_address_ltimes(pr, &lt6_tmp);
1200 
1201 		/*
1202 		 * We need to treat lifetimes for temporary addresses
1203 		 * differently, according to
1204 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1205 		 * we only update the lifetimes when they are in the maximum
1206 		 * intervals.
1207 		 */
1208 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1209 			u_int32_t maxvltime, maxpltime;
1210 
1211 			if (V_ip6_temp_valid_lifetime >
1212 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1213 			    V_ip6_desync_factor)) {
1214 				maxvltime = V_ip6_temp_valid_lifetime -
1215 				    (time_second - ifa6->ia6_createtime) -
1216 				    V_ip6_desync_factor;
1217 			} else
1218 				maxvltime = 0;
1219 			if (V_ip6_temp_preferred_lifetime >
1220 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1221 			    V_ip6_desync_factor)) {
1222 				maxpltime = V_ip6_temp_preferred_lifetime -
1223 				    (time_second - ifa6->ia6_createtime) -
1224 				    V_ip6_desync_factor;
1225 			} else
1226 				maxpltime = 0;
1227 
1228 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1229 			    lt6_tmp.ia6t_vltime > maxvltime) {
1230 				lt6_tmp.ia6t_vltime = maxvltime;
1231 			}
1232 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1233 			    lt6_tmp.ia6t_pltime > maxpltime) {
1234 				lt6_tmp.ia6t_pltime = maxpltime;
1235 			}
1236 		}
1237 		ifa6->ia6_lifetime = lt6_tmp;
1238 		ifa6->ia6_updatetime = time_second;
1239 	}
1240 	IF_ADDR_UNLOCK(ifp);
1241 	if (ia6_match == NULL && new->ndpr_vltime) {
1242 		int ifidlen;
1243 
1244 		/*
1245 		 * 5.5.3 (d) (continued)
1246 		 * No address matched and the valid lifetime is non-zero.
1247 		 * Create a new address.
1248 		 */
1249 
1250 		/*
1251 		 * Prefix Length check:
1252 		 * If the sum of the prefix length and interface identifier
1253 		 * length does not equal 128 bits, the Prefix Information
1254 		 * option MUST be ignored.  The length of the interface
1255 		 * identifier is defined in a separate link-type specific
1256 		 * document.
1257 		 */
1258 		ifidlen = in6_if2idlen(ifp);
1259 		if (ifidlen < 0) {
1260 			/* this should not happen, so we always log it. */
1261 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1262 			    if_name(ifp));
1263 			goto end;
1264 		}
1265 		if (ifidlen + pr->ndpr_plen != 128) {
1266 			nd6log((LOG_INFO,
1267 			    "prelist_update: invalid prefixlen "
1268 			    "%d for %s, ignored\n",
1269 			    pr->ndpr_plen, if_name(ifp)));
1270 			goto end;
1271 		}
1272 
1273 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1274 			/*
1275 			 * note that we should use pr (not new) for reference.
1276 			 */
1277 			pr->ndpr_refcnt++;
1278 			ia6->ia6_ndpr = pr;
1279 
1280 			/*
1281 			 * RFC 3041 3.3 (2).
1282 			 * When a new public address is created as described
1283 			 * in RFC2462, also create a new temporary address.
1284 			 *
1285 			 * RFC 3041 3.5.
1286 			 * When an interface connects to a new link, a new
1287 			 * randomized interface identifier should be generated
1288 			 * immediately together with a new set of temporary
1289 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1290 			 * in6_tmpifadd().
1291 			 */
1292 			if (V_ip6_use_tempaddr) {
1293 				int e;
1294 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1295 					nd6log((LOG_NOTICE, "prelist_update: "
1296 					    "failed to create a temporary "
1297 					    "address, errno=%d\n",
1298 					    e));
1299 				}
1300 			}
1301 			ifa_free(&ia6->ia_ifa);
1302 
1303 			/*
1304 			 * A newly added address might affect the status
1305 			 * of other addresses, so we check and update it.
1306 			 * XXX: what if address duplication happens?
1307 			 */
1308 			pfxlist_onlink_check();
1309 		} else {
1310 			/* just set an error. do not bark here. */
1311 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1312 		}
1313 	}
1314 
1315  end:
1316 	splx(s);
1317 	return error;
1318 }
1319 
1320 /*
1321  * A supplement function used in the on-link detection below;
1322  * detect if a given prefix has a (probably) reachable advertising router.
1323  * XXX: lengthy function name...
1324  */
1325 static struct nd_pfxrouter *
1326 find_pfxlist_reachable_router(struct nd_prefix *pr)
1327 {
1328 	struct nd_pfxrouter *pfxrtr;
1329 	struct llentry *ln;
1330 	int canreach;
1331 
1332 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr != NULL;
1333 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1334 		IF_AFDATA_LOCK(pfxrtr->router->ifp);
1335 		ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1336 		IF_AFDATA_UNLOCK(pfxrtr->router->ifp);
1337 		if (ln == NULL)
1338 			continue;
1339 		canreach = ND6_IS_LLINFO_PROBREACH(ln);
1340 		LLE_RUNLOCK(ln);
1341 		if (canreach)
1342 			break;
1343 	}
1344 	return (pfxrtr);
1345 }
1346 
1347 /*
1348  * Check if each prefix in the prefix list has at least one available router
1349  * that advertised the prefix (a router is "available" if its neighbor cache
1350  * entry is reachable or probably reachable).
1351  * If the check fails, the prefix may be off-link, because, for example,
1352  * we have moved from the network but the lifetime of the prefix has not
1353  * expired yet.  So we should not use the prefix if there is another prefix
1354  * that has an available router.
1355  * But, if there is no prefix that has an available router, we still regards
1356  * all the prefixes as on-link.  This is because we can't tell if all the
1357  * routers are simply dead or if we really moved from the network and there
1358  * is no router around us.
1359  */
1360 void
1361 pfxlist_onlink_check()
1362 {
1363 	struct nd_prefix *pr;
1364 	struct in6_ifaddr *ifa;
1365 	struct nd_defrouter *dr;
1366 	struct nd_pfxrouter *pfxrtr = NULL;
1367 
1368 	/*
1369 	 * Check if there is a prefix that has a reachable advertising
1370 	 * router.
1371 	 */
1372 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1373 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1374 			break;
1375 	}
1376 
1377 	/*
1378 	 * If we have no such prefix, check whether we still have a router
1379 	 * that does not advertise any prefixes.
1380 	 */
1381 	if (pr == NULL) {
1382 		for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
1383 		    dr = TAILQ_NEXT(dr, dr_entry)) {
1384 			struct nd_prefix *pr0;
1385 
1386 			for (pr0 = V_nd_prefix.lh_first; pr0;
1387 			    pr0 = pr0->ndpr_next) {
1388 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1389 					break;
1390 			}
1391 			if (pfxrtr != NULL)
1392 				break;
1393 		}
1394 	}
1395 	if (pr != NULL || (TAILQ_FIRST(&V_nd_defrouter) && pfxrtr == NULL)) {
1396 		/*
1397 		 * There is at least one prefix that has a reachable router,
1398 		 * or at least a router which probably does not advertise
1399 		 * any prefixes.  The latter would be the case when we move
1400 		 * to a new link where we have a router that does not provide
1401 		 * prefixes and we configure an address by hand.
1402 		 * Detach prefixes which have no reachable advertising
1403 		 * router, and attach other prefixes.
1404 		 */
1405 		for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1406 			/* XXX: a link-local prefix should never be detached */
1407 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1408 				continue;
1409 
1410 			/*
1411 			 * we aren't interested in prefixes without the L bit
1412 			 * set.
1413 			 */
1414 			if (pr->ndpr_raf_onlink == 0)
1415 				continue;
1416 
1417 			if (pr->ndpr_raf_auto == 0)
1418 				continue;
1419 
1420 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1421 			    find_pfxlist_reachable_router(pr) == NULL)
1422 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1423 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1424 			    find_pfxlist_reachable_router(pr) != 0)
1425 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1426 		}
1427 	} else {
1428 		/* there is no prefix that has a reachable router */
1429 		for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1430 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1431 				continue;
1432 
1433 			if (pr->ndpr_raf_onlink == 0)
1434 				continue;
1435 
1436 			if (pr->ndpr_raf_auto == 0)
1437 				continue;
1438 
1439 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1440 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1441 		}
1442 	}
1443 
1444 	/*
1445 	 * Remove each interface route associated with a (just) detached
1446 	 * prefix, and reinstall the interface route for a (just) attached
1447 	 * prefix.  Note that all attempt of reinstallation does not
1448 	 * necessarily success, when a same prefix is shared among multiple
1449 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1450 	 * so we don't have to care about them.
1451 	 */
1452 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1453 		int e;
1454 		char ip6buf[INET6_ADDRSTRLEN];
1455 
1456 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1457 			continue;
1458 
1459 		if (pr->ndpr_raf_onlink == 0)
1460 			continue;
1461 
1462 		if (pr->ndpr_raf_auto == 0)
1463 			continue;
1464 
1465 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1466 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1467 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1468 				nd6log((LOG_ERR,
1469 				    "pfxlist_onlink_check: failed to "
1470 				    "make %s/%d offlink, errno=%d\n",
1471 				    ip6_sprintf(ip6buf,
1472 					    &pr->ndpr_prefix.sin6_addr),
1473 					    pr->ndpr_plen, e));
1474 			}
1475 		}
1476 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1477 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1478 		    pr->ndpr_raf_onlink) {
1479 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1480 				nd6log((LOG_ERR,
1481 				    "pfxlist_onlink_check: failed to "
1482 				    "make %s/%d onlink, errno=%d\n",
1483 				    ip6_sprintf(ip6buf,
1484 					    &pr->ndpr_prefix.sin6_addr),
1485 					    pr->ndpr_plen, e));
1486 			}
1487 		}
1488 	}
1489 
1490 	/*
1491 	 * Changes on the prefix status might affect address status as well.
1492 	 * Make sure that all addresses derived from an attached prefix are
1493 	 * attached, and that all addresses derived from a detached prefix are
1494 	 * detached.  Note, however, that a manually configured address should
1495 	 * always be attached.
1496 	 * The precise detection logic is same as the one for prefixes.
1497 	 *
1498 	 * XXXRW: in6_ifaddrhead locking.
1499 	 */
1500 	TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1501 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1502 			continue;
1503 
1504 		if (ifa->ia6_ndpr == NULL) {
1505 			/*
1506 			 * This can happen when we first configure the address
1507 			 * (i.e. the address exists, but the prefix does not).
1508 			 * XXX: complicated relationships...
1509 			 */
1510 			continue;
1511 		}
1512 
1513 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1514 			break;
1515 	}
1516 	if (ifa) {
1517 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1518 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1519 				continue;
1520 
1521 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1522 				continue;
1523 
1524 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1525 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1526 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1527 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1528 					nd6_dad_start((struct ifaddr *)ifa, 0);
1529 				}
1530 			} else {
1531 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1532 			}
1533 		}
1534 	}
1535 	else {
1536 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1537 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1538 				continue;
1539 
1540 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1541 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1542 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1543 				/* Do we need a delay in this case? */
1544 				nd6_dad_start((struct ifaddr *)ifa, 0);
1545 			}
1546 		}
1547 	}
1548 }
1549 
1550 int
1551 nd6_prefix_onlink(struct nd_prefix *pr)
1552 {
1553 	struct ifaddr *ifa;
1554 	struct ifnet *ifp = pr->ndpr_ifp;
1555 	struct sockaddr_in6 mask6;
1556 	struct nd_prefix *opr;
1557 	u_long rtflags;
1558 	int error = 0;
1559 	struct radix_node_head *rnh;
1560 	struct rtentry *rt = NULL;
1561 	char ip6buf[INET6_ADDRSTRLEN];
1562 	struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1563 
1564 	/* sanity check */
1565 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1566 		nd6log((LOG_ERR,
1567 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1568 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1569 		    pr->ndpr_plen));
1570 		return (EEXIST);
1571 	}
1572 
1573 	/*
1574 	 * Add the interface route associated with the prefix.  Before
1575 	 * installing the route, check if there's the same prefix on another
1576 	 * interface, and the prefix has already installed the interface route.
1577 	 * Although such a configuration is expected to be rare, we explicitly
1578 	 * allow it.
1579 	 */
1580 	for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1581 		if (opr == pr)
1582 			continue;
1583 
1584 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1585 			continue;
1586 
1587 		if (opr->ndpr_plen == pr->ndpr_plen &&
1588 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1589 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1590 			return (0);
1591 	}
1592 
1593 	/*
1594 	 * We prefer link-local addresses as the associated interface address.
1595 	 */
1596 	/* search for a link-local addr */
1597 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1598 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1599 	if (ifa == NULL) {
1600 		/* XXX: freebsd does not have ifa_ifwithaf */
1601 		IF_ADDR_LOCK(ifp);
1602 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1603 			if (ifa->ifa_addr->sa_family == AF_INET6)
1604 				break;
1605 		}
1606 		if (ifa != NULL)
1607 			ifa_ref(ifa);
1608 		IF_ADDR_UNLOCK(ifp);
1609 		/* should we care about ia6_flags? */
1610 	}
1611 	if (ifa == NULL) {
1612 		/*
1613 		 * This can still happen, when, for example, we receive an RA
1614 		 * containing a prefix with the L bit set and the A bit clear,
1615 		 * after removing all IPv6 addresses on the receiving
1616 		 * interface.  This should, of course, be rare though.
1617 		 */
1618 		nd6log((LOG_NOTICE,
1619 		    "nd6_prefix_onlink: failed to find any ifaddr"
1620 		    " to add route for a prefix(%s/%d) on %s\n",
1621 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1622 		    pr->ndpr_plen, if_name(ifp)));
1623 		return (0);
1624 	}
1625 
1626 	/*
1627 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1628 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1629 	 */
1630 	bzero(&mask6, sizeof(mask6));
1631 	mask6.sin6_len = sizeof(mask6);
1632 	mask6.sin6_addr = pr->ndpr_mask;
1633 	rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1634 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1635 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1636 	if (error == 0) {
1637 		if (rt != NULL) /* this should be non NULL, though */ {
1638 			rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1639 			/* XXX what if rhn == NULL? */
1640 			RADIX_NODE_HEAD_LOCK(rnh);
1641 			RT_LOCK(rt);
1642 			if (!rt_setgate(rt, rt_key(rt), (struct sockaddr *)&null_sdl)) {
1643 				((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1644 					rt->rt_ifp->if_type;
1645 				((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1646 					rt->rt_ifp->if_index;
1647 			}
1648 			RADIX_NODE_HEAD_UNLOCK(rnh);
1649 			nd6_rtmsg(RTM_ADD, rt);
1650 			RT_UNLOCK(rt);
1651 		}
1652 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1653 	} else {
1654 		char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
1655 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1656 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1657 		    "errno = %d\n",
1658 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1659 		    pr->ndpr_plen, if_name(ifp),
1660 		    ip6_sprintf(ip6bufg, &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1661 		    ip6_sprintf(ip6bufm, &mask6.sin6_addr), rtflags, error));
1662 	}
1663 
1664 	if (rt != NULL) {
1665 		RT_LOCK(rt);
1666 		RT_REMREF(rt);
1667 		RT_UNLOCK(rt);
1668 	}
1669 	if (ifa != NULL)
1670 		ifa_free(ifa);
1671 
1672 	return (error);
1673 }
1674 
1675 int
1676 nd6_prefix_offlink(struct nd_prefix *pr)
1677 {
1678 	int error = 0;
1679 	struct ifnet *ifp = pr->ndpr_ifp;
1680 	struct nd_prefix *opr;
1681 	struct sockaddr_in6 sa6, mask6;
1682 	struct rtentry *rt = NULL;
1683 	char ip6buf[INET6_ADDRSTRLEN];
1684 
1685 	/* sanity check */
1686 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1687 		nd6log((LOG_ERR,
1688 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1689 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1690 		    pr->ndpr_plen));
1691 		return (EEXIST);
1692 	}
1693 
1694 	bzero(&sa6, sizeof(sa6));
1695 	sa6.sin6_family = AF_INET6;
1696 	sa6.sin6_len = sizeof(sa6);
1697 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1698 	    sizeof(struct in6_addr));
1699 	bzero(&mask6, sizeof(mask6));
1700 	mask6.sin6_family = AF_INET6;
1701 	mask6.sin6_len = sizeof(sa6);
1702 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1703 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1704 	    (struct sockaddr *)&mask6, 0, &rt);
1705 	if (error == 0) {
1706 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1707 
1708 		/* report the route deletion to the routing socket. */
1709 		if (rt != NULL)
1710 			nd6_rtmsg(RTM_DELETE, rt);
1711 
1712 		/*
1713 		 * There might be the same prefix on another interface,
1714 		 * the prefix which could not be on-link just because we have
1715 		 * the interface route (see comments in nd6_prefix_onlink).
1716 		 * If there's one, try to make the prefix on-link on the
1717 		 * interface.
1718 		 */
1719 		for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1720 			if (opr == pr)
1721 				continue;
1722 
1723 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1724 				continue;
1725 
1726 			/*
1727 			 * KAME specific: detached prefixes should not be
1728 			 * on-link.
1729 			 */
1730 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1731 				continue;
1732 
1733 			if (opr->ndpr_plen == pr->ndpr_plen &&
1734 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1735 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1736 				int e;
1737 
1738 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1739 					nd6log((LOG_ERR,
1740 					    "nd6_prefix_offlink: failed to "
1741 					    "recover a prefix %s/%d from %s "
1742 					    "to %s (errno = %d)\n",
1743 					    ip6_sprintf(ip6buf,
1744 						&opr->ndpr_prefix.sin6_addr),
1745 					    opr->ndpr_plen, if_name(ifp),
1746 					    if_name(opr->ndpr_ifp), e));
1747 				}
1748 			}
1749 		}
1750 	} else {
1751 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1752 		nd6log((LOG_ERR,
1753 		    "nd6_prefix_offlink: failed to delete route: "
1754 		    "%s/%d on %s (errno = %d)\n",
1755 		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1756 		    if_name(ifp), error));
1757 	}
1758 
1759 	if (rt != NULL) {
1760 		RTFREE(rt);
1761 	}
1762 
1763 	return (error);
1764 }
1765 
1766 static struct in6_ifaddr *
1767 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1768 {
1769 	struct ifnet *ifp = pr->ndpr_ifp;
1770 	struct ifaddr *ifa;
1771 	struct in6_aliasreq ifra;
1772 	struct in6_ifaddr *ia, *ib;
1773 	int error, plen0;
1774 	struct in6_addr mask;
1775 	int prefixlen = pr->ndpr_plen;
1776 	int updateflags;
1777 	char ip6buf[INET6_ADDRSTRLEN];
1778 
1779 	in6_prefixlen2mask(&mask, prefixlen);
1780 
1781 	/*
1782 	 * find a link-local address (will be interface ID).
1783 	 * Is it really mandatory? Theoretically, a global or a site-local
1784 	 * address can be configured without a link-local address, if we
1785 	 * have a unique interface identifier...
1786 	 *
1787 	 * it is not mandatory to have a link-local address, we can generate
1788 	 * interface identifier on the fly.  we do this because:
1789 	 * (1) it should be the easiest way to find interface identifier.
1790 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1791 	 * for multiple addresses on a single interface, and possible shortcut
1792 	 * of DAD.  we omitted DAD for this reason in the past.
1793 	 * (3) a user can prevent autoconfiguration of global address
1794 	 * by removing link-local address by hand (this is partly because we
1795 	 * don't have other way to control the use of IPv6 on an interface.
1796 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1797 	 * (4) it is easier to manage when an interface has addresses
1798 	 * with the same interface identifier, than to have multiple addresses
1799 	 * with different interface identifiers.
1800 	 */
1801 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1802 	if (ifa)
1803 		ib = (struct in6_ifaddr *)ifa;
1804 	else
1805 		return NULL;
1806 
1807 	/* prefixlen + ifidlen must be equal to 128 */
1808 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1809 	if (prefixlen != plen0) {
1810 		ifa_free(ifa);
1811 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1812 		    "(prefix=%d ifid=%d)\n",
1813 		    if_name(ifp), prefixlen, 128 - plen0));
1814 		return NULL;
1815 	}
1816 
1817 	/* make ifaddr */
1818 
1819 	bzero(&ifra, sizeof(ifra));
1820 	/*
1821 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1822 	 * for safety.
1823 	 */
1824 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1825 	ifra.ifra_addr.sin6_family = AF_INET6;
1826 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1827 	/* prefix */
1828 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1829 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1830 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1831 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1832 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1833 
1834 	/* interface ID */
1835 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1836 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1837 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1838 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1839 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1840 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1841 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1842 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1843 	ifa_free(ifa);
1844 
1845 	/* new prefix mask. */
1846 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1847 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1848 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1849 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1850 
1851 	/* lifetimes. */
1852 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1853 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1854 
1855 	/* XXX: scope zone ID? */
1856 
1857 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1858 
1859 	/*
1860 	 * Make sure that we do not have this address already.  This should
1861 	 * usually not happen, but we can still see this case, e.g., if we
1862 	 * have manually configured the exact address to be configured.
1863 	 */
1864 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1865 	    &ifra.ifra_addr.sin6_addr);
1866 	if (ifa != NULL) {
1867 		ifa_free(ifa);
1868 		/* this should be rare enough to make an explicit log */
1869 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1870 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1871 		return (NULL);
1872 	}
1873 
1874 	/*
1875 	 * Allocate ifaddr structure, link into chain, etc.
1876 	 * If we are going to create a new address upon receiving a multicasted
1877 	 * RA, we need to impose a random delay before starting DAD.
1878 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1879 	 */
1880 	updateflags = 0;
1881 	if (mcast)
1882 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1883 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1884 		nd6log((LOG_ERR,
1885 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1886 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1887 		    if_name(ifp), error));
1888 		return (NULL);	/* ifaddr must not have been allocated. */
1889 	}
1890 
1891 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1892 	/*
1893 	 * XXXRW: Assumption of non-NULLness here might not be true with
1894 	 * fine-grained locking -- should we validate it?  Or just return
1895 	 * earlier ifa rather than looking it up again?
1896 	 */
1897 	return (ia);		/* this is always non-NULL  and referenced. */
1898 }
1899 
1900 /*
1901  * ia0 - corresponding public address
1902  */
1903 int
1904 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1905 {
1906 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1907 	struct in6_ifaddr *newia, *ia;
1908 	struct in6_aliasreq ifra;
1909 	int i, error;
1910 	int trylimit = 3;	/* XXX: adhoc value */
1911 	int updateflags;
1912 	u_int32_t randid[2];
1913 	time_t vltime0, pltime0;
1914 
1915 	bzero(&ifra, sizeof(ifra));
1916 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1917 	ifra.ifra_addr = ia0->ia_addr;
1918 	/* copy prefix mask */
1919 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1920 	/* clear the old IFID */
1921 	for (i = 0; i < 4; i++) {
1922 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1923 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1924 	}
1925 
1926   again:
1927 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1928 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1929 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1930 		    "random IFID\n"));
1931 		return (EINVAL);
1932 	}
1933 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1934 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1935 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1936 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1937 
1938 	/*
1939 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1940 	 * However, we may still have a chance to see collision, because
1941 	 * there may be a time lag between generation of the ID and generation
1942 	 * of the address.  So, we'll do one more sanity check.
1943 	 */
1944 	IN6_IFADDR_RLOCK();
1945 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1946 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1947 		    &ifra.ifra_addr.sin6_addr)) {
1948 			if (trylimit-- == 0) {
1949 				IN6_IFADDR_RUNLOCK();
1950 				/*
1951 				 * Give up.  Something strange should have
1952 				 * happened.
1953 				 */
1954 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1955 				    "find a unique random IFID\n"));
1956 				return (EEXIST);
1957 			}
1958 			IN6_IFADDR_RUNLOCK();
1959 			forcegen = 1;
1960 			goto again;
1961 		}
1962 	}
1963 	IN6_IFADDR_RUNLOCK();
1964 
1965 	/*
1966 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1967          * public address or TEMP_VALID_LIFETIME.
1968 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1969          * of the public address or TEMP_PREFERRED_LIFETIME -
1970          * DESYNC_FACTOR.
1971 	 */
1972 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1973 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1974 		    (ia0->ia6_lifetime.ia6t_vltime -
1975 		    (time_second - ia0->ia6_updatetime));
1976 		if (vltime0 > V_ip6_temp_valid_lifetime)
1977 			vltime0 = V_ip6_temp_valid_lifetime;
1978 	} else
1979 		vltime0 = V_ip6_temp_valid_lifetime;
1980 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1981 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1982 		    (ia0->ia6_lifetime.ia6t_pltime -
1983 		    (time_second - ia0->ia6_updatetime));
1984 		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
1985 			pltime0 = V_ip6_temp_preferred_lifetime -
1986 			    V_ip6_desync_factor;
1987 		}
1988 	} else
1989 		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
1990 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1991 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1992 
1993 	/*
1994 	 * A temporary address is created only if this calculated Preferred
1995 	 * Lifetime is greater than REGEN_ADVANCE time units.
1996 	 */
1997 	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
1998 		return (0);
1999 
2000 	/* XXX: scope zone ID? */
2001 
2002 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2003 
2004 	/* allocate ifaddr structure, link into chain, etc. */
2005 	updateflags = 0;
2006 	if (delay)
2007 		updateflags |= IN6_IFAUPDATE_DADDELAY;
2008 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2009 		return (error);
2010 
2011 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2012 	if (newia == NULL) {	/* XXX: can it happen? */
2013 		nd6log((LOG_ERR,
2014 		    "in6_tmpifadd: ifa update succeeded, but we got "
2015 		    "no ifaddr\n"));
2016 		return (EINVAL); /* XXX */
2017 	}
2018 	newia->ia6_ndpr = ia0->ia6_ndpr;
2019 	newia->ia6_ndpr->ndpr_refcnt++;
2020 	ifa_free(&newia->ia_ifa);
2021 
2022 	/*
2023 	 * A newly added address might affect the status of other addresses.
2024 	 * XXX: when the temporary address is generated with a new public
2025 	 * address, the onlink check is redundant.  However, it would be safe
2026 	 * to do the check explicitly everywhere a new address is generated,
2027 	 * and, in fact, we surely need the check when we create a new
2028 	 * temporary address due to deprecation of an old temporary address.
2029 	 */
2030 	pfxlist_onlink_check();
2031 
2032 	return (0);
2033 }
2034 
2035 static int
2036 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2037 {
2038 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2039 		ndpr->ndpr_preferred = 0;
2040 	else
2041 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2042 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2043 		ndpr->ndpr_expire = 0;
2044 	else
2045 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2046 
2047 	return 0;
2048 }
2049 
2050 static void
2051 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2052 {
2053 	/* init ia6t_expire */
2054 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2055 		lt6->ia6t_expire = 0;
2056 	else {
2057 		lt6->ia6t_expire = time_second;
2058 		lt6->ia6t_expire += lt6->ia6t_vltime;
2059 	}
2060 
2061 	/* init ia6t_preferred */
2062 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2063 		lt6->ia6t_preferred = 0;
2064 	else {
2065 		lt6->ia6t_preferred = time_second;
2066 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2067 	}
2068 }
2069 
2070 /*
2071  * Delete all the routing table entries that use the specified gateway.
2072  * XXX: this function causes search through all entries of routing table, so
2073  * it shouldn't be called when acting as a router.
2074  */
2075 void
2076 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2077 {
2078 	struct radix_node_head *rnh;
2079 	int s = splnet();
2080 
2081 	/* We'll care only link-local addresses */
2082 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2083 		splx(s);
2084 		return;
2085 	}
2086 
2087 	rnh = rt_tables_get_rnh(0, AF_INET6);
2088 	if (rnh == NULL)
2089 		return;
2090 
2091 	RADIX_NODE_HEAD_LOCK(rnh);
2092 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2093 	RADIX_NODE_HEAD_UNLOCK(rnh);
2094 	splx(s);
2095 }
2096 
2097 static int
2098 rt6_deleteroute(struct radix_node *rn, void *arg)
2099 {
2100 #define SIN6(s)	((struct sockaddr_in6 *)s)
2101 	struct rtentry *rt = (struct rtentry *)rn;
2102 	struct in6_addr *gate = (struct in6_addr *)arg;
2103 
2104 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2105 		return (0);
2106 
2107 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2108 		return (0);
2109 	}
2110 
2111 	/*
2112 	 * Do not delete a static route.
2113 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2114 	 * 'cloned' bit instead?
2115 	 */
2116 	if ((rt->rt_flags & RTF_STATIC) != 0)
2117 		return (0);
2118 
2119 	/*
2120 	 * We delete only host route. This means, in particular, we don't
2121 	 * delete default route.
2122 	 */
2123 	if ((rt->rt_flags & RTF_HOST) == 0)
2124 		return (0);
2125 
2126 	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2127 	    rt_mask(rt), rt->rt_flags, 0));
2128 #undef SIN6
2129 }
2130 
2131 int
2132 nd6_setdefaultiface(int ifindex)
2133 {
2134 	int error = 0;
2135 
2136 	if (ifindex < 0 || V_if_index < ifindex)
2137 		return (EINVAL);
2138 	if (ifindex != 0 && !ifnet_byindex(ifindex))
2139 		return (EINVAL);
2140 
2141 	if (V_nd6_defifindex != ifindex) {
2142 		V_nd6_defifindex = ifindex;
2143 		if (V_nd6_defifindex > 0)
2144 			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2145 		else
2146 			V_nd6_defifp = NULL;
2147 
2148 		/*
2149 		 * Our current implementation assumes one-to-one maping between
2150 		 * interfaces and links, so it would be natural to use the
2151 		 * default interface as the default link.
2152 		 */
2153 		scope6_setdefault(V_nd6_defifp);
2154 	}
2155 
2156 	return (error);
2157 }
2158