xref: /freebsd/sys/netinet6/nd6_rtr.c (revision 5861f9665471e98e544f6fa3ce73c4912229ff82)
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 #include <sys/vimage.h>
52 
53 #include <net/if.h>
54 #include <net/if_types.h>
55 #include <net/if_dl.h>
56 #include <net/route.h>
57 #include <net/radix.h>
58 #include <net/vnet.h>
59 
60 #include <netinet/in.h>
61 #include <net/if_llatbl.h>
62 #include <netinet6/in6_var.h>
63 #include <netinet6/in6_ifattach.h>
64 #include <netinet/ip6.h>
65 #include <netinet6/ip6_var.h>
66 #include <netinet6/nd6.h>
67 #include <netinet/icmp6.h>
68 #include <netinet6/scope6_var.h>
69 #include <netinet6/vinet6.h>
70 
71 static int rtpref(struct nd_defrouter *);
72 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
73 static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
74     struct mbuf *, int));
75 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *,	int);
76 static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
77 	struct nd_defrouter *));
78 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
79 static void pfxrtr_del(struct nd_pfxrouter *);
80 static struct nd_pfxrouter *find_pfxlist_reachable_router
81 (struct nd_prefix *);
82 static void defrouter_delreq(struct nd_defrouter *);
83 static void nd6_rtmsg(int, struct rtentry *);
84 
85 static int in6_init_prefix_ltimes(struct nd_prefix *);
86 static void in6_init_address_ltimes __P((struct nd_prefix *,
87 	struct in6_addrlifetime *));
88 
89 static int rt6_deleteroute(struct radix_node *, void *);
90 
91 #ifdef VIMAGE_GLOBALS
92 extern int nd6_recalc_reachtm_interval;
93 
94 static struct ifnet *nd6_defifp;
95 int nd6_defifindex;
96 
97 int ip6_use_tempaddr;
98 int ip6_desync_factor;
99 u_int32_t ip6_temp_preferred_lifetime;
100 u_int32_t ip6_temp_valid_lifetime;
101 int ip6_temp_regen_advance;
102 #endif
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 	INIT_VNET_INET6(curvnet);
122 	struct ifnet *ifp = m->m_pkthdr.rcvif;
123 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
124 	struct nd_router_solicit *nd_rs;
125 	struct in6_addr saddr6 = ip6->ip6_src;
126 	char *lladdr = NULL;
127 	int lladdrlen = 0;
128 	union nd_opts ndopts;
129 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
130 
131 	/* If I'm not a router, ignore it. */
132 	if (V_ip6_accept_rtadv != 0 || V_ip6_forwarding != 1)
133 		goto freeit;
134 
135 	/* Sanity checks */
136 	if (ip6->ip6_hlim != 255) {
137 		nd6log((LOG_ERR,
138 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
139 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
140 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
141 		goto bad;
142 	}
143 
144 	/*
145 	 * Don't update the neighbor cache, if src = ::.
146 	 * This indicates that the src has no IP address assigned yet.
147 	 */
148 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
149 		goto freeit;
150 
151 #ifndef PULLDOWN_TEST
152 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
153 	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
154 #else
155 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
156 	if (nd_rs == NULL) {
157 		ICMP6STAT_INC(icp6s_tooshort);
158 		return;
159 	}
160 #endif
161 
162 	icmp6len -= sizeof(*nd_rs);
163 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
164 	if (nd6_options(&ndopts) < 0) {
165 		nd6log((LOG_INFO,
166 		    "nd6_rs_input: invalid ND option, ignored\n"));
167 		/* nd6_options have incremented stats */
168 		goto freeit;
169 	}
170 
171 	if (ndopts.nd_opts_src_lladdr) {
172 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
173 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
174 	}
175 
176 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
177 		nd6log((LOG_INFO,
178 		    "nd6_rs_input: lladdrlen mismatch for %s "
179 		    "(if %d, RS packet %d)\n",
180 		    ip6_sprintf(ip6bufs, &saddr6),
181 		    ifp->if_addrlen, lladdrlen - 2));
182 		goto bad;
183 	}
184 
185 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
186 
187  freeit:
188 	m_freem(m);
189 	return;
190 
191  bad:
192 	ICMP6STAT_INC(icp6s_badrs);
193 	m_freem(m);
194 }
195 
196 /*
197  * Receive Router Advertisement Message.
198  *
199  * Based on RFC 2461
200  * TODO: on-link bit on prefix information
201  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
202  */
203 void
204 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
205 {
206 	INIT_VNET_INET6(curvnet);
207 	struct ifnet *ifp = m->m_pkthdr.rcvif;
208 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
209 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
210 	struct nd_router_advert *nd_ra;
211 	struct in6_addr saddr6 = ip6->ip6_src;
212 	int mcast = 0;
213 	union nd_opts ndopts;
214 	struct nd_defrouter *dr;
215 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
216 
217 	/*
218 	 * We only accept RAs only when
219 	 * the system-wide variable allows the acceptance, and
220 	 * per-interface variable allows RAs on the receiving interface.
221 	 */
222 	if (V_ip6_accept_rtadv == 0)
223 		goto freeit;
224 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
225 		goto freeit;
226 
227 	if (ip6->ip6_hlim != 255) {
228 		nd6log((LOG_ERR,
229 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
230 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
231 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
232 		goto bad;
233 	}
234 
235 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
236 		nd6log((LOG_ERR,
237 		    "nd6_ra_input: src %s is not link-local\n",
238 		    ip6_sprintf(ip6bufs, &saddr6)));
239 		goto bad;
240 	}
241 
242 #ifndef PULLDOWN_TEST
243 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
244 	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
245 #else
246 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
247 	if (nd_ra == NULL) {
248 		ICMP6STAT_INC(icp6s_tooshort);
249 		return;
250 	}
251 #endif
252 
253 	icmp6len -= sizeof(*nd_ra);
254 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
255 	if (nd6_options(&ndopts) < 0) {
256 		nd6log((LOG_INFO,
257 		    "nd6_ra_input: invalid ND option, ignored\n"));
258 		/* nd6_options have incremented stats */
259 		goto freeit;
260 	}
261 
262     {
263 	struct nd_defrouter dr0;
264 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
265 
266 	/* remember if this is a multicasted advertisement */
267 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
268 		mcast = 1;
269 
270 	bzero(&dr0, sizeof(dr0));
271 	dr0.rtaddr = saddr6;
272 	dr0.flags  = nd_ra->nd_ra_flags_reserved;
273 	dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
274 	dr0.expire = time_second + dr0.rtlifetime;
275 	dr0.ifp = ifp;
276 	/* unspecified or not? (RFC 2461 6.3.4) */
277 	if (advreachable) {
278 		advreachable = ntohl(advreachable);
279 		if (advreachable <= MAX_REACHABLE_TIME &&
280 		    ndi->basereachable != advreachable) {
281 			ndi->basereachable = advreachable;
282 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
283 			ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
284 		}
285 	}
286 	if (nd_ra->nd_ra_retransmit)
287 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
288 	if (nd_ra->nd_ra_curhoplimit)
289 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
290 	dr = defrtrlist_update(&dr0);
291     }
292 
293 	/*
294 	 * prefix
295 	 */
296 	if (ndopts.nd_opts_pi) {
297 		struct nd_opt_hdr *pt;
298 		struct nd_opt_prefix_info *pi = NULL;
299 		struct nd_prefixctl pr;
300 
301 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
302 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
303 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
304 						(pt->nd_opt_len << 3))) {
305 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
306 				continue;
307 			pi = (struct nd_opt_prefix_info *)pt;
308 
309 			if (pi->nd_opt_pi_len != 4) {
310 				nd6log((LOG_INFO,
311 				    "nd6_ra_input: invalid option "
312 				    "len %d for prefix information option, "
313 				    "ignored\n", pi->nd_opt_pi_len));
314 				continue;
315 			}
316 
317 			if (128 < pi->nd_opt_pi_prefix_len) {
318 				nd6log((LOG_INFO,
319 				    "nd6_ra_input: invalid prefix "
320 				    "len %d for prefix information option, "
321 				    "ignored\n", pi->nd_opt_pi_prefix_len));
322 				continue;
323 			}
324 
325 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
326 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
327 				nd6log((LOG_INFO,
328 				    "nd6_ra_input: invalid prefix "
329 				    "%s, ignored\n",
330 				    ip6_sprintf(ip6bufs,
331 					&pi->nd_opt_pi_prefix)));
332 				continue;
333 			}
334 
335 			bzero(&pr, sizeof(pr));
336 			pr.ndpr_prefix.sin6_family = AF_INET6;
337 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
338 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
339 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
340 
341 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
342 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
343 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
344 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
345 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
346 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
347 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
348 			(void)prelist_update(&pr, dr, m, mcast);
349 		}
350 	}
351 
352 	/*
353 	 * MTU
354 	 */
355 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
356 		u_long mtu;
357 		u_long maxmtu;
358 
359 		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
360 
361 		/* lower bound */
362 		if (mtu < IPV6_MMTU) {
363 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
364 			    "mtu=%lu sent from %s, ignoring\n",
365 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
366 			goto skip;
367 		}
368 
369 		/* upper bound */
370 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
371 		    ? ndi->maxmtu : ifp->if_mtu;
372 		if (mtu <= maxmtu) {
373 			int change = (ndi->linkmtu != mtu);
374 
375 			ndi->linkmtu = mtu;
376 			if (change) /* in6_maxmtu may change */
377 				in6_setmaxmtu();
378 		} else {
379 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
380 			    "mtu=%lu sent from %s; "
381 			    "exceeds maxmtu %lu, ignoring\n",
382 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
383 		}
384 	}
385 
386  skip:
387 
388 	/*
389 	 * Source link layer address
390 	 */
391     {
392 	char *lladdr = NULL;
393 	int lladdrlen = 0;
394 
395 	if (ndopts.nd_opts_src_lladdr) {
396 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
397 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
398 	}
399 
400 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
401 		nd6log((LOG_INFO,
402 		    "nd6_ra_input: lladdrlen mismatch for %s "
403 		    "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
404 		    ifp->if_addrlen, lladdrlen - 2));
405 		goto bad;
406 	}
407 
408 	nd6_cache_lladdr(ifp, &saddr6, lladdr,
409 	    lladdrlen, ND_ROUTER_ADVERT, 0);
410 
411 	/*
412 	 * Installing a link-layer address might change the state of the
413 	 * router's neighbor cache, which might also affect our on-link
414 	 * detection of adveritsed prefixes.
415 	 */
416 	pfxlist_onlink_check();
417     }
418 
419  freeit:
420 	m_freem(m);
421 	return;
422 
423  bad:
424 	ICMP6STAT_INC(icp6s_badra);
425 	m_freem(m);
426 }
427 
428 /*
429  * default router list proccessing sub routines
430  */
431 
432 /* tell the change to user processes watching the routing socket. */
433 static void
434 nd6_rtmsg(int cmd, struct rtentry *rt)
435 {
436 	struct rt_addrinfo info;
437 	struct ifnet *ifp;
438 	struct ifaddr *ifa;
439 
440 	bzero((caddr_t)&info, sizeof(info));
441 	info.rti_info[RTAX_DST] = rt_key(rt);
442 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
443 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
444 	ifp = rt->rt_ifp;
445 	if (ifp != NULL) {
446 		IF_ADDR_LOCK(ifp);
447 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
448 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
449 		ifa_ref(ifa);
450 		IF_ADDR_UNLOCK(ifp);
451 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
452 	} else
453 		ifa = NULL;
454 
455 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
456 	if (ifa != NULL)
457 		ifa_free(ifa);
458 }
459 
460 void
461 defrouter_addreq(struct nd_defrouter *new)
462 {
463 	struct sockaddr_in6 def, mask, gate;
464 	struct rtentry *newrt = NULL;
465 	int s;
466 	int error;
467 
468 	bzero(&def, sizeof(def));
469 	bzero(&mask, sizeof(mask));
470 	bzero(&gate, sizeof(gate));
471 
472 	def.sin6_len = mask.sin6_len = gate.sin6_len =
473 	    sizeof(struct sockaddr_in6);
474 	def.sin6_family = gate.sin6_family = AF_INET6;
475 	gate.sin6_addr = new->rtaddr;
476 
477 	s = splnet();
478 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
479 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
480 	    RTF_GATEWAY, &newrt);
481 	if (newrt) {
482 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
483 		RTFREE(newrt);
484 	}
485 	if (error == 0)
486 		new->installed = 1;
487 	splx(s);
488 	return;
489 }
490 
491 struct nd_defrouter *
492 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
493 {
494 	INIT_VNET_INET6(ifp->if_vnet);
495 	struct nd_defrouter *dr;
496 
497 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
498 	     dr = TAILQ_NEXT(dr, dr_entry)) {
499 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
500 			return (dr);
501 	}
502 
503 	return (NULL);		/* search failed */
504 }
505 
506 /*
507  * Remove the default route for a given router.
508  * This is just a subroutine function for defrouter_select(), and should
509  * not be called from anywhere else.
510  */
511 static void
512 defrouter_delreq(struct nd_defrouter *dr)
513 {
514 	struct sockaddr_in6 def, mask, gate;
515 	struct rtentry *oldrt = NULL;
516 
517 	bzero(&def, sizeof(def));
518 	bzero(&mask, sizeof(mask));
519 	bzero(&gate, sizeof(gate));
520 
521 	def.sin6_len = mask.sin6_len = gate.sin6_len =
522 	    sizeof(struct sockaddr_in6);
523 	def.sin6_family = gate.sin6_family = AF_INET6;
524 	gate.sin6_addr = dr->rtaddr;
525 
526 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
527 	    (struct sockaddr *)&gate,
528 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
529 	if (oldrt) {
530 		nd6_rtmsg(RTM_DELETE, oldrt);
531 		RTFREE(oldrt);
532 	}
533 
534 	dr->installed = 0;
535 }
536 
537 /*
538  * remove all default routes from default router list
539  */
540 void
541 defrouter_reset(void)
542 {
543 	INIT_VNET_INET6(curvnet);
544 	struct nd_defrouter *dr;
545 
546 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
547 	     dr = TAILQ_NEXT(dr, dr_entry))
548 		defrouter_delreq(dr);
549 
550 	/*
551 	 * XXX should we also nuke any default routers in the kernel, by
552 	 * going through them by rtalloc1()?
553 	 */
554 }
555 
556 void
557 defrtrlist_del(struct nd_defrouter *dr)
558 {
559 	INIT_VNET_INET6(curvnet);
560 	struct nd_defrouter *deldr = NULL;
561 	struct nd_prefix *pr;
562 
563 	/*
564 	 * Flush all the routing table entries that use the router
565 	 * as a next hop.
566 	 */
567 	if (!V_ip6_forwarding && V_ip6_accept_rtadv) /* XXX: better condition? */
568 		rt6_flush(&dr->rtaddr, dr->ifp);
569 
570 	if (dr->installed) {
571 		deldr = dr;
572 		defrouter_delreq(dr);
573 	}
574 	TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
575 
576 	/*
577 	 * Also delete all the pointers to the router in each prefix lists.
578 	 */
579 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
580 		struct nd_pfxrouter *pfxrtr;
581 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
582 			pfxrtr_del(pfxrtr);
583 	}
584 	pfxlist_onlink_check();
585 
586 	/*
587 	 * If the router is the primary one, choose a new one.
588 	 * Note that defrouter_select() will remove the current gateway
589 	 * from the routing table.
590 	 */
591 	if (deldr)
592 		defrouter_select();
593 
594 	free(dr, M_IP6NDP);
595 }
596 
597 /*
598  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
599  * draft-ietf-ipngwg-router-selection:
600  * 1) Routers that are reachable or probably reachable should be preferred.
601  *    If we have more than one (probably) reachable router, prefer ones
602  *    with the highest router preference.
603  * 2) When no routers on the list are known to be reachable or
604  *    probably reachable, routers SHOULD be selected in a round-robin
605  *    fashion, regardless of router preference values.
606  * 3) If the Default Router List is empty, assume that all
607  *    destinations are on-link.
608  *
609  * We assume nd_defrouter is sorted by router preference value.
610  * Since the code below covers both with and without router preference cases,
611  * we do not need to classify the cases by ifdef.
612  *
613  * At this moment, we do not try to install more than one default router,
614  * even when the multipath routing is available, because we're not sure about
615  * the benefits for stub hosts comparing to the risk of making the code
616  * complicated and the possibility of introducing bugs.
617  */
618 void
619 defrouter_select(void)
620 {
621 	INIT_VNET_INET6(curvnet);
622 	int s = splnet();
623 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
624 	struct llentry *ln = NULL;
625 
626 	/*
627 	 * This function should be called only when acting as an autoconfigured
628 	 * host.  Although the remaining part of this function is not effective
629 	 * if the node is not an autoconfigured host, we explicitly exclude
630 	 * such cases here for safety.
631 	 */
632 	if (V_ip6_forwarding || !V_ip6_accept_rtadv) {
633 		nd6log((LOG_WARNING,
634 		    "defrouter_select: called unexpectedly (forwarding=%d, "
635 		    "accept_rtadv=%d)\n", V_ip6_forwarding, V_ip6_accept_rtadv));
636 		splx(s);
637 		return;
638 	}
639 
640 	/*
641 	 * Let's handle easy case (3) first:
642 	 * If default router list is empty, there's nothing to be done.
643 	 */
644 	if (!TAILQ_FIRST(&V_nd_defrouter)) {
645 		splx(s);
646 		return;
647 	}
648 
649 	/*
650 	 * Search for a (probably) reachable router from the list.
651 	 * We just pick up the first reachable one (if any), assuming that
652 	 * the ordering rule of the list described in defrtrlist_update().
653 	 */
654 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
655 	     dr = TAILQ_NEXT(dr, dr_entry)) {
656 		IF_AFDATA_LOCK(dr->ifp);
657 		if (selected_dr == NULL &&
658 		    (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
659 		    ND6_IS_LLINFO_PROBREACH(ln)) {
660 			selected_dr = dr;
661 		}
662 		IF_AFDATA_UNLOCK(dr->ifp);
663 		if (ln != NULL) {
664 			LLE_RUNLOCK(ln);
665 			ln = NULL;
666 		}
667 
668 		if (dr->installed && installed_dr == NULL)
669 			installed_dr = dr;
670 		else if (dr->installed && installed_dr) {
671 			/* this should not happen.  warn for diagnosis. */
672 			log(LOG_ERR, "defrouter_select: more than one router"
673 			    " is installed\n");
674 		}
675 	}
676 	/*
677 	 * If none of the default routers was found to be reachable,
678 	 * round-robin the list regardless of preference.
679 	 * Otherwise, if we have an installed router, check if the selected
680 	 * (reachable) router should really be preferred to the installed one.
681 	 * We only prefer the new router when the old one is not reachable
682 	 * or when the new one has a really higher preference value.
683 	 */
684 	if (selected_dr == NULL) {
685 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
686 			selected_dr = TAILQ_FIRST(&V_nd_defrouter);
687 		else
688 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
689 	} else if (installed_dr) {
690 		IF_AFDATA_LOCK(installed_dr->ifp);
691 		if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
692 		    ND6_IS_LLINFO_PROBREACH(ln) &&
693 		    rtpref(selected_dr) <= rtpref(installed_dr)) {
694 			selected_dr = installed_dr;
695 		}
696 		IF_AFDATA_UNLOCK(installed_dr->ifp);
697 		if (ln != NULL)
698 			LLE_RUNLOCK(ln);
699 	}
700 
701 	/*
702 	 * If the selected router is different than the installed one,
703 	 * remove the installed router and install the selected one.
704 	 * Note that the selected router is never NULL here.
705 	 */
706 	if (installed_dr != selected_dr) {
707 		if (installed_dr)
708 			defrouter_delreq(installed_dr);
709 		defrouter_addreq(selected_dr);
710 	}
711 
712 	splx(s);
713 	return;
714 }
715 
716 /*
717  * for default router selection
718  * regards router-preference field as a 2-bit signed integer
719  */
720 static int
721 rtpref(struct nd_defrouter *dr)
722 {
723 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
724 	case ND_RA_FLAG_RTPREF_HIGH:
725 		return (RTPREF_HIGH);
726 	case ND_RA_FLAG_RTPREF_MEDIUM:
727 	case ND_RA_FLAG_RTPREF_RSV:
728 		return (RTPREF_MEDIUM);
729 	case ND_RA_FLAG_RTPREF_LOW:
730 		return (RTPREF_LOW);
731 	default:
732 		/*
733 		 * This case should never happen.  If it did, it would mean a
734 		 * serious bug of kernel internal.  We thus always bark here.
735 		 * Or, can we even panic?
736 		 */
737 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
738 		return (RTPREF_INVALID);
739 	}
740 	/* NOTREACHED */
741 }
742 
743 static struct nd_defrouter *
744 defrtrlist_update(struct nd_defrouter *new)
745 {
746 	INIT_VNET_INET6(curvnet);
747 	struct nd_defrouter *dr, *n;
748 	int s = splnet();
749 
750 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
751 		/* entry exists */
752 		if (new->rtlifetime == 0) {
753 			defrtrlist_del(dr);
754 			dr = NULL;
755 		} else {
756 			int oldpref = rtpref(dr);
757 
758 			/* override */
759 			dr->flags = new->flags; /* xxx flag check */
760 			dr->rtlifetime = new->rtlifetime;
761 			dr->expire = new->expire;
762 
763 			/*
764 			 * If the preference does not change, there's no need
765 			 * to sort the entries.
766 			 */
767 			if (rtpref(new) == oldpref) {
768 				splx(s);
769 				return (dr);
770 			}
771 
772 			/*
773 			 * preferred router may be changed, so relocate
774 			 * this router.
775 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
776 			 * However, since defrtrlist_del() has many side
777 			 * effects, we intentionally do so here.
778 			 * defrouter_select() below will handle routing
779 			 * changes later.
780 			 */
781 			TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
782 			n = dr;
783 			goto insert;
784 		}
785 		splx(s);
786 		return (dr);
787 	}
788 
789 	/* entry does not exist */
790 	if (new->rtlifetime == 0) {
791 		splx(s);
792 		return (NULL);
793 	}
794 
795 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
796 	if (n == NULL) {
797 		splx(s);
798 		return (NULL);
799 	}
800 	bzero(n, sizeof(*n));
801 	*n = *new;
802 
803 insert:
804 	/*
805 	 * Insert the new router in the Default Router List;
806 	 * The Default Router List should be in the descending order
807 	 * of router-preferece.  Routers with the same preference are
808 	 * sorted in the arriving time order.
809 	 */
810 
811 	/* insert at the end of the group */
812 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
813 	     dr = TAILQ_NEXT(dr, dr_entry)) {
814 		if (rtpref(n) > rtpref(dr))
815 			break;
816 	}
817 	if (dr)
818 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
819 	else
820 		TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
821 
822 	defrouter_select();
823 
824 	splx(s);
825 
826 	return (n);
827 }
828 
829 static struct nd_pfxrouter *
830 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
831 {
832 	struct nd_pfxrouter *search;
833 
834 	for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
835 		if (search->router == dr)
836 			break;
837 	}
838 
839 	return (search);
840 }
841 
842 static void
843 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
844 {
845 	struct nd_pfxrouter *new;
846 
847 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
848 	if (new == NULL)
849 		return;
850 	bzero(new, sizeof(*new));
851 	new->router = dr;
852 
853 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
854 
855 	pfxlist_onlink_check();
856 }
857 
858 static void
859 pfxrtr_del(struct nd_pfxrouter *pfr)
860 {
861 	LIST_REMOVE(pfr, pfr_entry);
862 	free(pfr, M_IP6NDP);
863 }
864 
865 struct nd_prefix *
866 nd6_prefix_lookup(struct nd_prefixctl *key)
867 {
868 	INIT_VNET_INET6(curvnet);
869 	struct nd_prefix *search;
870 
871 	for (search = V_nd_prefix.lh_first;
872 	    search; search = search->ndpr_next) {
873 		if (key->ndpr_ifp == search->ndpr_ifp &&
874 		    key->ndpr_plen == search->ndpr_plen &&
875 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
876 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
877 			break;
878 		}
879 	}
880 
881 	return (search);
882 }
883 
884 int
885 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
886     struct nd_prefix **newp)
887 {
888 	INIT_VNET_INET6(curvnet);
889 	struct nd_prefix *new = NULL;
890 	int error = 0;
891 	int i, s;
892 	char ip6buf[INET6_ADDRSTRLEN];
893 
894 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
895 	if (new == NULL)
896 		return(ENOMEM);
897 	bzero(new, sizeof(*new));
898 	new->ndpr_ifp = pr->ndpr_ifp;
899 	new->ndpr_prefix = pr->ndpr_prefix;
900 	new->ndpr_plen = pr->ndpr_plen;
901 	new->ndpr_vltime = pr->ndpr_vltime;
902 	new->ndpr_pltime = pr->ndpr_pltime;
903 	new->ndpr_flags = pr->ndpr_flags;
904 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
905 		free(new, M_IP6NDP);
906 		return(error);
907 	}
908 	new->ndpr_lastupdate = time_second;
909 	if (newp != NULL)
910 		*newp = new;
911 
912 	/* initialization */
913 	LIST_INIT(&new->ndpr_advrtrs);
914 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
915 	/* make prefix in the canonical form */
916 	for (i = 0; i < 4; i++)
917 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
918 		    new->ndpr_mask.s6_addr32[i];
919 
920 	s = splnet();
921 	/* link ndpr_entry to nd_prefix list */
922 	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
923 	splx(s);
924 
925 	/* ND_OPT_PI_FLAG_ONLINK processing */
926 	if (new->ndpr_raf_onlink) {
927 		int e;
928 
929 		if ((e = nd6_prefix_onlink(new)) != 0) {
930 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
931 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
932 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
933 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
934 			/* proceed anyway. XXX: is it correct? */
935 		}
936 	}
937 
938 	if (dr)
939 		pfxrtr_add(new, dr);
940 
941 	return 0;
942 }
943 
944 void
945 prelist_remove(struct nd_prefix *pr)
946 {
947 	INIT_VNET_INET6(curvnet);
948 	struct nd_pfxrouter *pfr, *next;
949 	int e, s;
950 	char ip6buf[INET6_ADDRSTRLEN];
951 
952 	/* make sure to invalidate the prefix until it is really freed. */
953 	pr->ndpr_vltime = 0;
954 	pr->ndpr_pltime = 0;
955 
956 	/*
957 	 * Though these flags are now meaningless, we'd rather keep the value
958 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
959 	 * when executing "ndp -p".
960 	 */
961 
962 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
963 	    (e = nd6_prefix_offlink(pr)) != 0) {
964 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
965 		    "on %s, errno=%d\n",
966 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
967 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
968 		/* what should we do? */
969 	}
970 
971 	if (pr->ndpr_refcnt > 0)
972 		return;		/* notice here? */
973 
974 	s = splnet();
975 
976 	/* unlink ndpr_entry from nd_prefix list */
977 	LIST_REMOVE(pr, ndpr_entry);
978 
979 	/* free list of routers that adversed the prefix */
980 	for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
981 		next = pfr->pfr_next;
982 
983 		free(pfr, M_IP6NDP);
984 	}
985 	splx(s);
986 
987 	free(pr, M_IP6NDP);
988 
989 	pfxlist_onlink_check();
990 }
991 
992 /*
993  * dr - may be NULL
994  */
995 
996 static int
997 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
998     struct mbuf *m, int mcast)
999 {
1000 	INIT_VNET_INET6(curvnet);
1001 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
1002 	struct ifaddr *ifa;
1003 	struct ifnet *ifp = new->ndpr_ifp;
1004 	struct nd_prefix *pr;
1005 	int s = splnet();
1006 	int error = 0;
1007 	int newprefix = 0;
1008 	int auth;
1009 	struct in6_addrlifetime lt6_tmp;
1010 	char ip6buf[INET6_ADDRSTRLEN];
1011 
1012 	auth = 0;
1013 	if (m) {
1014 		/*
1015 		 * Authenticity for NA consists authentication for
1016 		 * both IP header and IP datagrams, doesn't it ?
1017 		 */
1018 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1019 		auth = ((m->m_flags & M_AUTHIPHDR) &&
1020 		    (m->m_flags & M_AUTHIPDGM));
1021 #endif
1022 	}
1023 
1024 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1025 		/*
1026 		 * nd6_prefix_lookup() ensures that pr and new have the same
1027 		 * prefix on a same interface.
1028 		 */
1029 
1030 		/*
1031 		 * Update prefix information.  Note that the on-link (L) bit
1032 		 * and the autonomous (A) bit should NOT be changed from 1
1033 		 * to 0.
1034 		 */
1035 		if (new->ndpr_raf_onlink == 1)
1036 			pr->ndpr_raf_onlink = 1;
1037 		if (new->ndpr_raf_auto == 1)
1038 			pr->ndpr_raf_auto = 1;
1039 		if (new->ndpr_raf_onlink) {
1040 			pr->ndpr_vltime = new->ndpr_vltime;
1041 			pr->ndpr_pltime = new->ndpr_pltime;
1042 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1043 			pr->ndpr_lastupdate = time_second;
1044 		}
1045 
1046 		if (new->ndpr_raf_onlink &&
1047 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1048 			int e;
1049 
1050 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1051 				nd6log((LOG_ERR,
1052 				    "prelist_update: failed to make "
1053 				    "the prefix %s/%d on-link on %s "
1054 				    "(errno=%d)\n",
1055 				    ip6_sprintf(ip6buf,
1056 					    &pr->ndpr_prefix.sin6_addr),
1057 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1058 				/* proceed anyway. XXX: is it correct? */
1059 			}
1060 		}
1061 
1062 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1063 			pfxrtr_add(pr, dr);
1064 	} else {
1065 		struct nd_prefix *newpr = NULL;
1066 
1067 		newprefix = 1;
1068 
1069 		if (new->ndpr_vltime == 0)
1070 			goto end;
1071 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1072 			goto end;
1073 
1074 		error = nd6_prelist_add(new, dr, &newpr);
1075 		if (error != 0 || newpr == NULL) {
1076 			nd6log((LOG_NOTICE, "prelist_update: "
1077 			    "nd6_prelist_add failed for %s/%d on %s "
1078 			    "errno=%d, returnpr=%p\n",
1079 			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1080 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1081 			    error, newpr));
1082 			goto end; /* we should just give up in this case. */
1083 		}
1084 
1085 		/*
1086 		 * XXX: from the ND point of view, we can ignore a prefix
1087 		 * with the on-link bit being zero.  However, we need a
1088 		 * prefix structure for references from autoconfigured
1089 		 * addresses.  Thus, we explicitly make sure that the prefix
1090 		 * itself expires now.
1091 		 */
1092 		if (newpr->ndpr_raf_onlink == 0) {
1093 			newpr->ndpr_vltime = 0;
1094 			newpr->ndpr_pltime = 0;
1095 			in6_init_prefix_ltimes(newpr);
1096 		}
1097 
1098 		pr = newpr;
1099 	}
1100 
1101 	/*
1102 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1103 	 * Note that pr must be non NULL at this point.
1104 	 */
1105 
1106 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1107 	if (!new->ndpr_raf_auto)
1108 		goto end;
1109 
1110 	/*
1111 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1112 	 * nd6_ra_input.
1113 	 */
1114 
1115 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1116 	if (new->ndpr_pltime > new->ndpr_vltime) {
1117 		error = EINVAL;	/* XXX: won't be used */
1118 		goto end;
1119 	}
1120 
1121 	/*
1122 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1123 	 * an address configured by stateless autoconfiguration already in the
1124 	 * list of addresses associated with the interface, and the Valid
1125 	 * Lifetime is not 0, form an address.  We first check if we have
1126 	 * a matching prefix.
1127 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1128 	 * consider autoconfigured addresses while RFC2462 simply said
1129 	 * "address".
1130 	 */
1131 	IF_ADDR_LOCK(ifp);
1132 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1133 		struct in6_ifaddr *ifa6;
1134 		u_int32_t remaininglifetime;
1135 
1136 		if (ifa->ifa_addr->sa_family != AF_INET6)
1137 			continue;
1138 
1139 		ifa6 = (struct in6_ifaddr *)ifa;
1140 
1141 		/*
1142 		 * We only consider autoconfigured addresses as per rfc2462bis.
1143 		 */
1144 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1145 			continue;
1146 
1147 		/*
1148 		 * Spec is not clear here, but I believe we should concentrate
1149 		 * on unicast (i.e. not anycast) addresses.
1150 		 * XXX: other ia6_flags? detached or duplicated?
1151 		 */
1152 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1153 			continue;
1154 
1155 		/*
1156 		 * Ignore the address if it is not associated with a prefix
1157 		 * or is associated with a prefix that is different from this
1158 		 * one.  (pr is never NULL here)
1159 		 */
1160 		if (ifa6->ia6_ndpr != pr)
1161 			continue;
1162 
1163 		if (ia6_match == NULL) /* remember the first one */
1164 			ia6_match = ifa6;
1165 
1166 		/*
1167 		 * An already autoconfigured address matched.  Now that we
1168 		 * are sure there is at least one matched address, we can
1169 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1170 		 * "two hours" rule and the privacy extension.
1171 		 * We apply some clarifications in rfc2462bis:
1172 		 * - use remaininglifetime instead of storedlifetime as a
1173 		 *   variable name
1174 		 * - remove the dead code in the "two-hour" rule
1175 		 */
1176 #define TWOHOUR		(120*60)
1177 		lt6_tmp = ifa6->ia6_lifetime;
1178 
1179 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1180 			remaininglifetime = ND6_INFINITE_LIFETIME;
1181 		else if (time_second - ifa6->ia6_updatetime >
1182 			 lt6_tmp.ia6t_vltime) {
1183 			/*
1184 			 * The case of "invalid" address.  We should usually
1185 			 * not see this case.
1186 			 */
1187 			remaininglifetime = 0;
1188 		} else
1189 			remaininglifetime = lt6_tmp.ia6t_vltime -
1190 			    (time_second - ifa6->ia6_updatetime);
1191 
1192 		/* when not updating, keep the current stored lifetime. */
1193 		lt6_tmp.ia6t_vltime = remaininglifetime;
1194 
1195 		if (TWOHOUR < new->ndpr_vltime ||
1196 		    remaininglifetime < new->ndpr_vltime) {
1197 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1198 		} else if (remaininglifetime <= TWOHOUR) {
1199 			if (auth) {
1200 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1201 			}
1202 		} else {
1203 			/*
1204 			 * new->ndpr_vltime <= TWOHOUR &&
1205 			 * TWOHOUR < remaininglifetime
1206 			 */
1207 			lt6_tmp.ia6t_vltime = TWOHOUR;
1208 		}
1209 
1210 		/* The 2 hour rule is not imposed for preferred lifetime. */
1211 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1212 
1213 		in6_init_address_ltimes(pr, &lt6_tmp);
1214 
1215 		/*
1216 		 * We need to treat lifetimes for temporary addresses
1217 		 * differently, according to
1218 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1219 		 * we only update the lifetimes when they are in the maximum
1220 		 * intervals.
1221 		 */
1222 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1223 			u_int32_t maxvltime, maxpltime;
1224 
1225 			if (V_ip6_temp_valid_lifetime >
1226 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1227 			    V_ip6_desync_factor)) {
1228 				maxvltime = V_ip6_temp_valid_lifetime -
1229 				    (time_second - ifa6->ia6_createtime) -
1230 				    V_ip6_desync_factor;
1231 			} else
1232 				maxvltime = 0;
1233 			if (V_ip6_temp_preferred_lifetime >
1234 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1235 			    V_ip6_desync_factor)) {
1236 				maxpltime = V_ip6_temp_preferred_lifetime -
1237 				    (time_second - ifa6->ia6_createtime) -
1238 				    V_ip6_desync_factor;
1239 			} else
1240 				maxpltime = 0;
1241 
1242 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1243 			    lt6_tmp.ia6t_vltime > maxvltime) {
1244 				lt6_tmp.ia6t_vltime = maxvltime;
1245 			}
1246 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1247 			    lt6_tmp.ia6t_pltime > maxpltime) {
1248 				lt6_tmp.ia6t_pltime = maxpltime;
1249 			}
1250 		}
1251 		ifa6->ia6_lifetime = lt6_tmp;
1252 		ifa6->ia6_updatetime = time_second;
1253 	}
1254 	IF_ADDR_UNLOCK(ifp);
1255 	if (ia6_match == NULL && new->ndpr_vltime) {
1256 		int ifidlen;
1257 
1258 		/*
1259 		 * 5.5.3 (d) (continued)
1260 		 * No address matched and the valid lifetime is non-zero.
1261 		 * Create a new address.
1262 		 */
1263 
1264 		/*
1265 		 * Prefix Length check:
1266 		 * If the sum of the prefix length and interface identifier
1267 		 * length does not equal 128 bits, the Prefix Information
1268 		 * option MUST be ignored.  The length of the interface
1269 		 * identifier is defined in a separate link-type specific
1270 		 * document.
1271 		 */
1272 		ifidlen = in6_if2idlen(ifp);
1273 		if (ifidlen < 0) {
1274 			/* this should not happen, so we always log it. */
1275 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1276 			    if_name(ifp));
1277 			goto end;
1278 		}
1279 		if (ifidlen + pr->ndpr_plen != 128) {
1280 			nd6log((LOG_INFO,
1281 			    "prelist_update: invalid prefixlen "
1282 			    "%d for %s, ignored\n",
1283 			    pr->ndpr_plen, if_name(ifp)));
1284 			goto end;
1285 		}
1286 
1287 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1288 			/*
1289 			 * note that we should use pr (not new) for reference.
1290 			 */
1291 			pr->ndpr_refcnt++;
1292 			ia6->ia6_ndpr = pr;
1293 
1294 			/*
1295 			 * RFC 3041 3.3 (2).
1296 			 * When a new public address is created as described
1297 			 * in RFC2462, also create a new temporary address.
1298 			 *
1299 			 * RFC 3041 3.5.
1300 			 * When an interface connects to a new link, a new
1301 			 * randomized interface identifier should be generated
1302 			 * immediately together with a new set of temporary
1303 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1304 			 * in6_tmpifadd().
1305 			 */
1306 			if (V_ip6_use_tempaddr) {
1307 				int e;
1308 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1309 					nd6log((LOG_NOTICE, "prelist_update: "
1310 					    "failed to create a temporary "
1311 					    "address, errno=%d\n",
1312 					    e));
1313 				}
1314 			}
1315 			ifa_free(&ia6->ia_ifa);
1316 
1317 			/*
1318 			 * A newly added address might affect the status
1319 			 * of other addresses, so we check and update it.
1320 			 * XXX: what if address duplication happens?
1321 			 */
1322 			pfxlist_onlink_check();
1323 		} else {
1324 			/* just set an error. do not bark here. */
1325 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1326 		}
1327 	}
1328 
1329  end:
1330 	splx(s);
1331 	return error;
1332 }
1333 
1334 /*
1335  * A supplement function used in the on-link detection below;
1336  * detect if a given prefix has a (probably) reachable advertising router.
1337  * XXX: lengthy function name...
1338  */
1339 static struct nd_pfxrouter *
1340 find_pfxlist_reachable_router(struct nd_prefix *pr)
1341 {
1342 	struct nd_pfxrouter *pfxrtr;
1343 	struct llentry *ln;
1344 	int canreach;
1345 
1346 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr != NULL;
1347 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1348 		IF_AFDATA_LOCK(pfxrtr->router->ifp);
1349 		ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1350 		IF_AFDATA_UNLOCK(pfxrtr->router->ifp);
1351 		if (ln == NULL)
1352 			continue;
1353 		canreach = ND6_IS_LLINFO_PROBREACH(ln);
1354 		LLE_RUNLOCK(ln);
1355 		if (canreach)
1356 			break;
1357 	}
1358 	return (pfxrtr);
1359 }
1360 
1361 /*
1362  * Check if each prefix in the prefix list has at least one available router
1363  * that advertised the prefix (a router is "available" if its neighbor cache
1364  * entry is reachable or probably reachable).
1365  * If the check fails, the prefix may be off-link, because, for example,
1366  * we have moved from the network but the lifetime of the prefix has not
1367  * expired yet.  So we should not use the prefix if there is another prefix
1368  * that has an available router.
1369  * But, if there is no prefix that has an available router, we still regards
1370  * all the prefixes as on-link.  This is because we can't tell if all the
1371  * routers are simply dead or if we really moved from the network and there
1372  * is no router around us.
1373  */
1374 void
1375 pfxlist_onlink_check()
1376 {
1377 	INIT_VNET_INET6(curvnet);
1378 	struct nd_prefix *pr;
1379 	struct in6_ifaddr *ifa;
1380 	struct nd_defrouter *dr;
1381 	struct nd_pfxrouter *pfxrtr = NULL;
1382 
1383 	/*
1384 	 * Check if there is a prefix that has a reachable advertising
1385 	 * router.
1386 	 */
1387 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1388 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1389 			break;
1390 	}
1391 
1392 	/*
1393 	 * If we have no such prefix, check whether we still have a router
1394 	 * that does not advertise any prefixes.
1395 	 */
1396 	if (pr == NULL) {
1397 		for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
1398 		    dr = TAILQ_NEXT(dr, dr_entry)) {
1399 			struct nd_prefix *pr0;
1400 
1401 			for (pr0 = V_nd_prefix.lh_first; pr0;
1402 			    pr0 = pr0->ndpr_next) {
1403 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1404 					break;
1405 			}
1406 			if (pfxrtr != NULL)
1407 				break;
1408 		}
1409 	}
1410 	if (pr != NULL || (TAILQ_FIRST(&V_nd_defrouter) && pfxrtr == NULL)) {
1411 		/*
1412 		 * There is at least one prefix that has a reachable router,
1413 		 * or at least a router which probably does not advertise
1414 		 * any prefixes.  The latter would be the case when we move
1415 		 * to a new link where we have a router that does not provide
1416 		 * prefixes and we configure an address by hand.
1417 		 * Detach prefixes which have no reachable advertising
1418 		 * router, and attach other prefixes.
1419 		 */
1420 		for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1421 			/* XXX: a link-local prefix should never be detached */
1422 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1423 				continue;
1424 
1425 			/*
1426 			 * we aren't interested in prefixes without the L bit
1427 			 * set.
1428 			 */
1429 			if (pr->ndpr_raf_onlink == 0)
1430 				continue;
1431 
1432 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1433 			    find_pfxlist_reachable_router(pr) == NULL)
1434 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1435 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1436 			    find_pfxlist_reachable_router(pr) != 0)
1437 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1438 		}
1439 	} else {
1440 		/* there is no prefix that has a reachable router */
1441 		for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
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_DETACHED;
1450 		}
1451 	}
1452 
1453 	/*
1454 	 * Remove each interface route associated with a (just) detached
1455 	 * prefix, and reinstall the interface route for a (just) attached
1456 	 * prefix.  Note that all attempt of reinstallation does not
1457 	 * necessarily success, when a same prefix is shared among multiple
1458 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1459 	 * so we don't have to care about them.
1460 	 */
1461 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1462 		int e;
1463 		char ip6buf[INET6_ADDRSTRLEN];
1464 
1465 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1466 			continue;
1467 
1468 		if (pr->ndpr_raf_onlink == 0)
1469 			continue;
1470 
1471 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1472 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1473 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1474 				nd6log((LOG_ERR,
1475 				    "pfxlist_onlink_check: failed to "
1476 				    "make %s/%d offlink, errno=%d\n",
1477 				    ip6_sprintf(ip6buf,
1478 					    &pr->ndpr_prefix.sin6_addr),
1479 					    pr->ndpr_plen, e));
1480 			}
1481 		}
1482 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1483 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1484 		    pr->ndpr_raf_onlink) {
1485 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1486 				nd6log((LOG_ERR,
1487 				    "pfxlist_onlink_check: failed to "
1488 				    "make %s/%d onlink, errno=%d\n",
1489 				    ip6_sprintf(ip6buf,
1490 					    &pr->ndpr_prefix.sin6_addr),
1491 					    pr->ndpr_plen, e));
1492 			}
1493 		}
1494 	}
1495 
1496 	/*
1497 	 * Changes on the prefix status might affect address status as well.
1498 	 * Make sure that all addresses derived from an attached prefix are
1499 	 * attached, and that all addresses derived from a detached prefix are
1500 	 * detached.  Note, however, that a manually configured address should
1501 	 * always be attached.
1502 	 * The precise detection logic is same as the one for prefixes.
1503 	 *
1504 	 * XXXRW: in6_ifaddrhead locking.
1505 	 */
1506 	TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1507 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1508 			continue;
1509 
1510 		if (ifa->ia6_ndpr == NULL) {
1511 			/*
1512 			 * This can happen when we first configure the address
1513 			 * (i.e. the address exists, but the prefix does not).
1514 			 * XXX: complicated relationships...
1515 			 */
1516 			continue;
1517 		}
1518 
1519 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1520 			break;
1521 	}
1522 	if (ifa) {
1523 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1524 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1525 				continue;
1526 
1527 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1528 				continue;
1529 
1530 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1531 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1532 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1533 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1534 					nd6_dad_start((struct ifaddr *)ifa, 0);
1535 				}
1536 			} else {
1537 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1538 			}
1539 		}
1540 	}
1541 	else {
1542 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1543 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1544 				continue;
1545 
1546 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1547 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1548 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1549 				/* Do we need a delay in this case? */
1550 				nd6_dad_start((struct ifaddr *)ifa, 0);
1551 			}
1552 		}
1553 	}
1554 }
1555 
1556 int
1557 nd6_prefix_onlink(struct nd_prefix *pr)
1558 {
1559 	INIT_VNET_INET6(curvnet);
1560 	struct ifaddr *ifa;
1561 	struct ifnet *ifp = pr->ndpr_ifp;
1562 	struct sockaddr_in6 mask6;
1563 	struct nd_prefix *opr;
1564 	u_long rtflags;
1565 	int error = 0;
1566 	struct radix_node_head *rnh;
1567 	struct rtentry *rt = NULL;
1568 	char ip6buf[INET6_ADDRSTRLEN];
1569 	struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1570 
1571 	/* sanity check */
1572 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1573 		nd6log((LOG_ERR,
1574 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1575 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1576 		    pr->ndpr_plen));
1577 		return (EEXIST);
1578 	}
1579 
1580 	/*
1581 	 * Add the interface route associated with the prefix.  Before
1582 	 * installing the route, check if there's the same prefix on another
1583 	 * interface, and the prefix has already installed the interface route.
1584 	 * Although such a configuration is expected to be rare, we explicitly
1585 	 * allow it.
1586 	 */
1587 	for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1588 		if (opr == pr)
1589 			continue;
1590 
1591 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1592 			continue;
1593 
1594 		if (opr->ndpr_plen == pr->ndpr_plen &&
1595 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1596 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1597 			return (0);
1598 	}
1599 
1600 	/*
1601 	 * We prefer link-local addresses as the associated interface address.
1602 	 */
1603 	/* search for a link-local addr */
1604 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1605 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1606 	if (ifa == NULL) {
1607 		/* XXX: freebsd does not have ifa_ifwithaf */
1608 		IF_ADDR_LOCK(ifp);
1609 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1610 			if (ifa->ifa_addr->sa_family == AF_INET6)
1611 				break;
1612 		}
1613 		if (ifa != NULL)
1614 			ifa_ref(ifa);
1615 		IF_ADDR_UNLOCK(ifp);
1616 		/* should we care about ia6_flags? */
1617 	}
1618 	if (ifa == NULL) {
1619 		/*
1620 		 * This can still happen, when, for example, we receive an RA
1621 		 * containing a prefix with the L bit set and the A bit clear,
1622 		 * after removing all IPv6 addresses on the receiving
1623 		 * interface.  This should, of course, be rare though.
1624 		 */
1625 		nd6log((LOG_NOTICE,
1626 		    "nd6_prefix_onlink: failed to find any ifaddr"
1627 		    " to add route for a prefix(%s/%d) on %s\n",
1628 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1629 		    pr->ndpr_plen, if_name(ifp)));
1630 		return (0);
1631 	}
1632 
1633 	/*
1634 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1635 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1636 	 */
1637 	bzero(&mask6, sizeof(mask6));
1638 	mask6.sin6_len = sizeof(mask6);
1639 	mask6.sin6_addr = pr->ndpr_mask;
1640 	rtflags = ifa->ifa_flags | RTF_UP;
1641 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1642 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1643 	if (error == 0) {
1644 		if (rt != NULL) /* this should be non NULL, though */ {
1645 			rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1646 			/* XXX what if rhn == NULL? */
1647 			RADIX_NODE_HEAD_LOCK(rnh);
1648 			RT_LOCK(rt);
1649 			if (!rt_setgate(rt, rt_key(rt), (struct sockaddr *)&null_sdl)) {
1650 				((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1651 					rt->rt_ifp->if_type;
1652 				((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1653 					rt->rt_ifp->if_index;
1654 			}
1655 			RADIX_NODE_HEAD_UNLOCK(rnh);
1656 			nd6_rtmsg(RTM_ADD, rt);
1657 			RT_UNLOCK(rt);
1658 		}
1659 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1660 	} else {
1661 		char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
1662 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1663 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1664 		    "errno = %d\n",
1665 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1666 		    pr->ndpr_plen, if_name(ifp),
1667 		    ip6_sprintf(ip6bufg, &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1668 		    ip6_sprintf(ip6bufm, &mask6.sin6_addr), rtflags, error));
1669 	}
1670 
1671 	if (rt != NULL) {
1672 		RT_LOCK(rt);
1673 		RT_REMREF(rt);
1674 		RT_UNLOCK(rt);
1675 	}
1676 	if (ifa != NULL)
1677 		ifa_free(ifa);
1678 
1679 	return (error);
1680 }
1681 
1682 int
1683 nd6_prefix_offlink(struct nd_prefix *pr)
1684 {
1685 	INIT_VNET_INET6(curvnet);
1686 	int error = 0;
1687 	struct ifnet *ifp = pr->ndpr_ifp;
1688 	struct nd_prefix *opr;
1689 	struct sockaddr_in6 sa6, mask6;
1690 	struct rtentry *rt = NULL;
1691 	char ip6buf[INET6_ADDRSTRLEN];
1692 
1693 	/* sanity check */
1694 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1695 		nd6log((LOG_ERR,
1696 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1697 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1698 		    pr->ndpr_plen));
1699 		return (EEXIST);
1700 	}
1701 
1702 	bzero(&sa6, sizeof(sa6));
1703 	sa6.sin6_family = AF_INET6;
1704 	sa6.sin6_len = sizeof(sa6);
1705 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1706 	    sizeof(struct in6_addr));
1707 	bzero(&mask6, sizeof(mask6));
1708 	mask6.sin6_family = AF_INET6;
1709 	mask6.sin6_len = sizeof(sa6);
1710 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1711 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1712 	    (struct sockaddr *)&mask6, 0, &rt);
1713 	if (error == 0) {
1714 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1715 
1716 		/* report the route deletion to the routing socket. */
1717 		if (rt != NULL)
1718 			nd6_rtmsg(RTM_DELETE, rt);
1719 
1720 		/*
1721 		 * There might be the same prefix on another interface,
1722 		 * the prefix which could not be on-link just because we have
1723 		 * the interface route (see comments in nd6_prefix_onlink).
1724 		 * If there's one, try to make the prefix on-link on the
1725 		 * interface.
1726 		 */
1727 		for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1728 			if (opr == pr)
1729 				continue;
1730 
1731 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1732 				continue;
1733 
1734 			/*
1735 			 * KAME specific: detached prefixes should not be
1736 			 * on-link.
1737 			 */
1738 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1739 				continue;
1740 
1741 			if (opr->ndpr_plen == pr->ndpr_plen &&
1742 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1743 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1744 				int e;
1745 
1746 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1747 					nd6log((LOG_ERR,
1748 					    "nd6_prefix_offlink: failed to "
1749 					    "recover a prefix %s/%d from %s "
1750 					    "to %s (errno = %d)\n",
1751 					    ip6_sprintf(ip6buf,
1752 						&opr->ndpr_prefix.sin6_addr),
1753 					    opr->ndpr_plen, if_name(ifp),
1754 					    if_name(opr->ndpr_ifp), e));
1755 				}
1756 			}
1757 		}
1758 	} else {
1759 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1760 		nd6log((LOG_ERR,
1761 		    "nd6_prefix_offlink: failed to delete route: "
1762 		    "%s/%d on %s (errno = %d)\n",
1763 		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1764 		    if_name(ifp), error));
1765 	}
1766 
1767 	if (rt != NULL) {
1768 		RTFREE(rt);
1769 	}
1770 
1771 	return (error);
1772 }
1773 
1774 static struct in6_ifaddr *
1775 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1776 {
1777 	INIT_VNET_INET6(curvnet);
1778 	struct ifnet *ifp = pr->ndpr_ifp;
1779 	struct ifaddr *ifa;
1780 	struct in6_aliasreq ifra;
1781 	struct in6_ifaddr *ia, *ib;
1782 	int error, plen0;
1783 	struct in6_addr mask;
1784 	int prefixlen = pr->ndpr_plen;
1785 	int updateflags;
1786 	char ip6buf[INET6_ADDRSTRLEN];
1787 
1788 	in6_prefixlen2mask(&mask, prefixlen);
1789 
1790 	/*
1791 	 * find a link-local address (will be interface ID).
1792 	 * Is it really mandatory? Theoretically, a global or a site-local
1793 	 * address can be configured without a link-local address, if we
1794 	 * have a unique interface identifier...
1795 	 *
1796 	 * it is not mandatory to have a link-local address, we can generate
1797 	 * interface identifier on the fly.  we do this because:
1798 	 * (1) it should be the easiest way to find interface identifier.
1799 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1800 	 * for multiple addresses on a single interface, and possible shortcut
1801 	 * of DAD.  we omitted DAD for this reason in the past.
1802 	 * (3) a user can prevent autoconfiguration of global address
1803 	 * by removing link-local address by hand (this is partly because we
1804 	 * don't have other way to control the use of IPv6 on an interface.
1805 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1806 	 * (4) it is easier to manage when an interface has addresses
1807 	 * with the same interface identifier, than to have multiple addresses
1808 	 * with different interface identifiers.
1809 	 */
1810 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1811 	if (ifa)
1812 		ib = (struct in6_ifaddr *)ifa;
1813 	else
1814 		return NULL;
1815 
1816 	/* prefixlen + ifidlen must be equal to 128 */
1817 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1818 	if (prefixlen != plen0) {
1819 		ifa_free(ifa);
1820 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1821 		    "(prefix=%d ifid=%d)\n",
1822 		    if_name(ifp), prefixlen, 128 - plen0));
1823 		return NULL;
1824 	}
1825 
1826 	/* make ifaddr */
1827 
1828 	bzero(&ifra, sizeof(ifra));
1829 	/*
1830 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1831 	 * for safety.
1832 	 */
1833 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1834 	ifra.ifra_addr.sin6_family = AF_INET6;
1835 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1836 	/* prefix */
1837 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1838 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1839 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1840 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1841 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1842 
1843 	/* interface ID */
1844 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1845 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1846 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1847 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1848 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1849 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1850 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1851 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1852 	ifa_free(ifa);
1853 
1854 	/* new prefix mask. */
1855 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1856 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1857 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1858 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1859 
1860 	/* lifetimes. */
1861 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1862 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1863 
1864 	/* XXX: scope zone ID? */
1865 
1866 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1867 
1868 	/*
1869 	 * Make sure that we do not have this address already.  This should
1870 	 * usually not happen, but we can still see this case, e.g., if we
1871 	 * have manually configured the exact address to be configured.
1872 	 */
1873 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1874 	    &ifra.ifra_addr.sin6_addr);
1875 	if (ifa != NULL) {
1876 		ifa_free(ifa);
1877 		/* this should be rare enough to make an explicit log */
1878 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1879 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1880 		return (NULL);
1881 	}
1882 
1883 	/*
1884 	 * Allocate ifaddr structure, link into chain, etc.
1885 	 * If we are going to create a new address upon receiving a multicasted
1886 	 * RA, we need to impose a random delay before starting DAD.
1887 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1888 	 */
1889 	updateflags = 0;
1890 	if (mcast)
1891 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1892 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1893 		nd6log((LOG_ERR,
1894 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1895 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1896 		    if_name(ifp), error));
1897 		return (NULL);	/* ifaddr must not have been allocated. */
1898 	}
1899 
1900 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1901 	/*
1902 	 * XXXRW: Assumption of non-NULLness here might not be true with
1903 	 * fine-grained locking -- should we validate it?  Or just return
1904 	 * earlier ifa rather than looking it up again?
1905 	 */
1906 	return (ia);		/* this is always non-NULL  and referenced. */
1907 }
1908 
1909 /*
1910  * ia0 - corresponding public address
1911  */
1912 int
1913 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1914 {
1915 	INIT_VNET_INET6(curvnet);
1916 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1917 	struct in6_ifaddr *newia, *ia;
1918 	struct in6_aliasreq ifra;
1919 	int i, error;
1920 	int trylimit = 3;	/* XXX: adhoc value */
1921 	int updateflags;
1922 	u_int32_t randid[2];
1923 	time_t vltime0, pltime0;
1924 
1925 	bzero(&ifra, sizeof(ifra));
1926 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1927 	ifra.ifra_addr = ia0->ia_addr;
1928 	/* copy prefix mask */
1929 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1930 	/* clear the old IFID */
1931 	for (i = 0; i < 4; i++) {
1932 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1933 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1934 	}
1935 
1936   again:
1937 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1938 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1939 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1940 		    "random IFID\n"));
1941 		return (EINVAL);
1942 	}
1943 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1944 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1945 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1946 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1947 
1948 	/*
1949 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1950 	 * However, we may still have a chance to see collision, because
1951 	 * there may be a time lag between generation of the ID and generation
1952 	 * of the address.  So, we'll do one more sanity check.
1953 	 */
1954 	IN6_IFADDR_RLOCK();
1955 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1956 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1957 		    &ifra.ifra_addr.sin6_addr)) {
1958 			if (trylimit-- == 0) {
1959 				IN6_IFADDR_RUNLOCK();
1960 				/*
1961 				 * Give up.  Something strange should have
1962 				 * happened.
1963 				 */
1964 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1965 				    "find a unique random IFID\n"));
1966 				return (EEXIST);
1967 			}
1968 			IN6_IFADDR_RUNLOCK();
1969 			forcegen = 1;
1970 			goto again;
1971 		}
1972 	}
1973 	IN6_IFADDR_RUNLOCK();
1974 
1975 	/*
1976 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1977          * public address or TEMP_VALID_LIFETIME.
1978 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1979          * of the public address or TEMP_PREFERRED_LIFETIME -
1980          * DESYNC_FACTOR.
1981 	 */
1982 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1983 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1984 		    (ia0->ia6_lifetime.ia6t_vltime -
1985 		    (time_second - ia0->ia6_updatetime));
1986 		if (vltime0 > V_ip6_temp_valid_lifetime)
1987 			vltime0 = V_ip6_temp_valid_lifetime;
1988 	} else
1989 		vltime0 = V_ip6_temp_valid_lifetime;
1990 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1991 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1992 		    (ia0->ia6_lifetime.ia6t_pltime -
1993 		    (time_second - ia0->ia6_updatetime));
1994 		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
1995 			pltime0 = V_ip6_temp_preferred_lifetime -
1996 			    V_ip6_desync_factor;
1997 		}
1998 	} else
1999 		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
2000 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
2001 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
2002 
2003 	/*
2004 	 * A temporary address is created only if this calculated Preferred
2005 	 * Lifetime is greater than REGEN_ADVANCE time units.
2006 	 */
2007 	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2008 		return (0);
2009 
2010 	/* XXX: scope zone ID? */
2011 
2012 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2013 
2014 	/* allocate ifaddr structure, link into chain, etc. */
2015 	updateflags = 0;
2016 	if (delay)
2017 		updateflags |= IN6_IFAUPDATE_DADDELAY;
2018 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2019 		return (error);
2020 
2021 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2022 	if (newia == NULL) {	/* XXX: can it happen? */
2023 		nd6log((LOG_ERR,
2024 		    "in6_tmpifadd: ifa update succeeded, but we got "
2025 		    "no ifaddr\n"));
2026 		return (EINVAL); /* XXX */
2027 	}
2028 	newia->ia6_ndpr = ia0->ia6_ndpr;
2029 	newia->ia6_ndpr->ndpr_refcnt++;
2030 	ifa_free(&newia->ia_ifa);
2031 
2032 	/*
2033 	 * A newly added address might affect the status of other addresses.
2034 	 * XXX: when the temporary address is generated with a new public
2035 	 * address, the onlink check is redundant.  However, it would be safe
2036 	 * to do the check explicitly everywhere a new address is generated,
2037 	 * and, in fact, we surely need the check when we create a new
2038 	 * temporary address due to deprecation of an old temporary address.
2039 	 */
2040 	pfxlist_onlink_check();
2041 
2042 	return (0);
2043 }
2044 
2045 static int
2046 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2047 {
2048 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2049 		ndpr->ndpr_preferred = 0;
2050 	else
2051 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2052 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2053 		ndpr->ndpr_expire = 0;
2054 	else
2055 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2056 
2057 	return 0;
2058 }
2059 
2060 static void
2061 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2062 {
2063 	/* init ia6t_expire */
2064 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2065 		lt6->ia6t_expire = 0;
2066 	else {
2067 		lt6->ia6t_expire = time_second;
2068 		lt6->ia6t_expire += lt6->ia6t_vltime;
2069 	}
2070 
2071 	/* init ia6t_preferred */
2072 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2073 		lt6->ia6t_preferred = 0;
2074 	else {
2075 		lt6->ia6t_preferred = time_second;
2076 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2077 	}
2078 }
2079 
2080 /*
2081  * Delete all the routing table entries that use the specified gateway.
2082  * XXX: this function causes search through all entries of routing table, so
2083  * it shouldn't be called when acting as a router.
2084  */
2085 void
2086 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2087 {
2088 	struct radix_node_head *rnh;
2089 	int s = splnet();
2090 
2091 	/* We'll care only link-local addresses */
2092 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2093 		splx(s);
2094 		return;
2095 	}
2096 
2097 	rnh = rt_tables_get_rnh(0, AF_INET6);
2098 	if (rnh == NULL)
2099 		return;
2100 
2101 	RADIX_NODE_HEAD_LOCK(rnh);
2102 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2103 	RADIX_NODE_HEAD_UNLOCK(rnh);
2104 	splx(s);
2105 }
2106 
2107 static int
2108 rt6_deleteroute(struct radix_node *rn, void *arg)
2109 {
2110 #define SIN6(s)	((struct sockaddr_in6 *)s)
2111 	struct rtentry *rt = (struct rtentry *)rn;
2112 	struct in6_addr *gate = (struct in6_addr *)arg;
2113 
2114 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2115 		return (0);
2116 
2117 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2118 		return (0);
2119 	}
2120 
2121 	/*
2122 	 * Do not delete a static route.
2123 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2124 	 * 'cloned' bit instead?
2125 	 */
2126 	if ((rt->rt_flags & RTF_STATIC) != 0)
2127 		return (0);
2128 
2129 	/*
2130 	 * We delete only host route. This means, in particular, we don't
2131 	 * delete default route.
2132 	 */
2133 	if ((rt->rt_flags & RTF_HOST) == 0)
2134 		return (0);
2135 
2136 	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2137 	    rt_mask(rt), rt->rt_flags, 0));
2138 #undef SIN6
2139 }
2140 
2141 int
2142 nd6_setdefaultiface(int ifindex)
2143 {
2144 	INIT_VNET_NET(curvnet);
2145 	INIT_VNET_INET6(curvnet);
2146 	int error = 0;
2147 
2148 	if (ifindex < 0 || V_if_index < ifindex)
2149 		return (EINVAL);
2150 	if (ifindex != 0 && !ifnet_byindex(ifindex))
2151 		return (EINVAL);
2152 
2153 	if (V_nd6_defifindex != ifindex) {
2154 		V_nd6_defifindex = ifindex;
2155 		if (V_nd6_defifindex > 0)
2156 			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2157 		else
2158 			V_nd6_defifp = NULL;
2159 
2160 		/*
2161 		 * Our current implementation assumes one-to-one maping between
2162 		 * interfaces and links, so it would be natural to use the
2163 		 * default interface as the default link.
2164 		 */
2165 		scope6_setdefault(V_nd6_defifp);
2166 	}
2167 
2168 	return (error);
2169 }
2170