xref: /netbsd-src/sys/netinet6/nd6_rtr.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: nd6_rtr.c,v 1.71 2007/12/05 23:00:58 dyoung Exp $	*/
2 /*	$KAME: nd6_rtr.c,v 1.95 2001/02/07 08:09:47 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: nd6_rtr.c,v 1.71 2007/12/05 23:00:58 dyoung Exp $");
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/malloc.h>
39 #include <sys/mbuf.h>
40 #include <sys/socket.h>
41 #include <sys/sockio.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/errno.h>
45 #include <sys/ioctl.h>
46 #include <sys/syslog.h>
47 
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/if_dl.h>
51 #include <net/route.h>
52 #include <net/radix.h>
53 
54 #include <netinet/in.h>
55 #include <netinet6/in6_var.h>
56 #include <netinet6/in6_ifattach.h>
57 #include <netinet/ip6.h>
58 #include <netinet6/ip6_var.h>
59 #include <netinet6/nd6.h>
60 #include <netinet/icmp6.h>
61 #include <netinet6/scope6_var.h>
62 
63 #include <net/net_osdep.h>
64 
65 static int rtpref(struct nd_defrouter *);
66 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
67 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
68     struct mbuf *, int);
69 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
70 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
71 	struct nd_defrouter *);
72 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
73 static void pfxrtr_del(struct nd_pfxrouter *);
74 static struct nd_pfxrouter *find_pfxlist_reachable_router
75 	(struct nd_prefix *);
76 static void defrouter_delreq(struct nd_defrouter *);
77 static void nd6_rtmsg(int, struct rtentry *);
78 
79 static int in6_init_prefix_ltimes(struct nd_prefix *);
80 static void in6_init_address_ltimes(struct nd_prefix *ndpr,
81 	struct in6_addrlifetime *lt6);
82 
83 static int rt6_deleteroute(struct rtentry *, void *);
84 
85 extern int nd6_recalc_reachtm_interval;
86 
87 static struct ifnet *nd6_defifp;
88 int nd6_defifindex;
89 
90 int ip6_use_tempaddr = 0;
91 
92 int ip6_desync_factor;
93 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
94 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
95 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
96 
97 /* RTPREF_MEDIUM has to be 0! */
98 #define RTPREF_HIGH	1
99 #define RTPREF_MEDIUM	0
100 #define RTPREF_LOW	(-1)
101 #define RTPREF_RESERVED	(-2)
102 #define RTPREF_INVALID	(-3)	/* internal */
103 
104 /*
105  * Receive Router Solicitation Message - just for routers.
106  * Router solicitation/advertisement is mostly managed by userland program
107  * (rtadvd) so here we have no function like nd6_ra_output().
108  *
109  * Based on RFC 2461
110  */
111 void
112 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
113 {
114 	struct ifnet *ifp = m->m_pkthdr.rcvif;
115 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
116 	struct nd_router_solicit *nd_rs;
117 	struct in6_addr saddr6 = ip6->ip6_src;
118 	char *lladdr = NULL;
119 	int lladdrlen = 0;
120 	union nd_opts ndopts;
121 
122 	/* If I'm not a router, ignore it. */
123 	if (ip6_accept_rtadv != 0 || !ip6_forwarding)
124 		goto freeit;
125 
126 	/* Sanity checks */
127 	if (ip6->ip6_hlim != 255) {
128 		nd6log((LOG_ERR,
129 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
130 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
131 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
132 		goto bad;
133 	}
134 
135 	/*
136 	 * Don't update the neighbor cache, if src = ::.
137 	 * This indicates that the src has no IP address assigned yet.
138 	 */
139 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
140 		goto freeit;
141 
142 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
143 	if (nd_rs == NULL) {
144 		icmp6stat.icp6s_tooshort++;
145 		return;
146 	}
147 
148 	icmp6len -= sizeof(*nd_rs);
149 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
150 	if (nd6_options(&ndopts) < 0) {
151 		nd6log((LOG_INFO,
152 		    "nd6_rs_input: invalid ND option, ignored\n"));
153 		/* nd6_options have incremented stats */
154 		goto freeit;
155 	}
156 
157 	if (ndopts.nd_opts_src_lladdr) {
158 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
159 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
160 	}
161 
162 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
163 		nd6log((LOG_INFO,
164 		    "nd6_rs_input: lladdrlen mismatch for %s "
165 		    "(if %d, RS packet %d)\n",
166 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
167 		goto bad;
168 	}
169 
170 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
171 
172  freeit:
173 	m_freem(m);
174 	return;
175 
176  bad:
177 	icmp6stat.icp6s_badrs++;
178 	m_freem(m);
179 }
180 
181 /*
182  * Receive Router Advertisement Message.
183  *
184  * Based on RFC 2461
185  * TODO: on-link bit on prefix information
186  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
187  */
188 void
189 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
190 {
191 	struct ifnet *ifp = m->m_pkthdr.rcvif;
192 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
193 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
194 	struct nd_router_advert *nd_ra;
195 	struct in6_addr saddr6 = ip6->ip6_src;
196 #if 0
197 	struct in6_addr daddr6 = ip6->ip6_dst;
198 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
199 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
200 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
201 #endif
202 	int mcast = 0;
203 	union nd_opts ndopts;
204 	struct nd_defrouter *dr;
205 
206 	/*
207 	 * We only accept RAs only when
208 	 * the system-wide variable allows the acceptance, and
209 	 * per-interface variable allows RAs on the receiving interface.
210 	 */
211 	if (ip6_accept_rtadv == 0)
212 		goto freeit;
213 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
214 		goto freeit;
215 
216 	if (ip6->ip6_hlim != 255) {
217 		nd6log((LOG_ERR,
218 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
219 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
220 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
221 		goto bad;
222 	}
223 
224 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
225 		nd6log((LOG_ERR,
226 		    "nd6_ra_input: src %s is not link-local\n",
227 		    ip6_sprintf(&saddr6)));
228 		goto bad;
229 	}
230 
231 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
232 	if (nd_ra == NULL) {
233 		icmp6stat.icp6s_tooshort++;
234 		return;
235 	}
236 
237 	icmp6len -= sizeof(*nd_ra);
238 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
239 	if (nd6_options(&ndopts) < 0) {
240 		nd6log((LOG_INFO,
241 		    "nd6_ra_input: invalid ND option, ignored\n"));
242 		/* nd6_options have incremented stats */
243 		goto freeit;
244 	}
245 
246     {
247 	struct nd_defrouter drtr;
248 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
249 
250 	/* remember if this is a multicasted advertisement */
251 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
252 		mcast = 1;
253 
254 	memset(&drtr, 0, sizeof(drtr));
255 	drtr.rtaddr = saddr6;
256 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
257 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
258 	drtr.expire = time_second + drtr.rtlifetime;
259 	drtr.ifp = ifp;
260 	/* unspecified or not? (RFC 2461 6.3.4) */
261 	if (advreachable) {
262 		NTOHL(advreachable);
263 		if (advreachable <= MAX_REACHABLE_TIME &&
264 		    ndi->basereachable != advreachable) {
265 			ndi->basereachable = advreachable;
266 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
267 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
268 		}
269 	}
270 	if (nd_ra->nd_ra_retransmit)
271 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
272 	if (nd_ra->nd_ra_curhoplimit)
273 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
274 	dr = defrtrlist_update(&drtr);
275     }
276 
277 	/*
278 	 * prefix
279 	 */
280 	if (ndopts.nd_opts_pi) {
281 		struct nd_opt_hdr *pt;
282 		struct nd_opt_prefix_info *pi = NULL;
283 		struct nd_prefixctl pr;
284 
285 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
286 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
287 		     pt = (struct nd_opt_hdr *)((char *)pt +
288 						(pt->nd_opt_len << 3))) {
289 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
290 				continue;
291 			pi = (struct nd_opt_prefix_info *)pt;
292 
293 			if (pi->nd_opt_pi_len != 4) {
294 				nd6log((LOG_INFO,
295 				    "nd6_ra_input: invalid option "
296 				    "len %d for prefix information option, "
297 				    "ignored\n", pi->nd_opt_pi_len));
298 				continue;
299 			}
300 
301 			if (128 < pi->nd_opt_pi_prefix_len) {
302 				nd6log((LOG_INFO,
303 				    "nd6_ra_input: invalid prefix "
304 				    "len %d for prefix information option, "
305 				    "ignored\n", pi->nd_opt_pi_prefix_len));
306 				continue;
307 			}
308 
309 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
310 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
311 				nd6log((LOG_INFO,
312 				    "nd6_ra_input: invalid prefix "
313 				    "%s, ignored\n",
314 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
315 				continue;
316 			}
317 
318 			bzero(&pr, sizeof(pr));
319 			sockaddr_in6_init(&pr.ndpr_prefix,
320 			    &pi->nd_opt_pi_prefix, 0, 0, 0);
321 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
322 
323 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
324 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
325 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
326 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
327 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
328 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
329 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
330 
331 			(void)prelist_update(&pr, dr, m, mcast);
332 		}
333 	}
334 
335 	/*
336 	 * MTU
337 	 */
338 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
339 		u_long mtu;
340 		u_long maxmtu;
341 
342 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
343 
344 		/* lower bound */
345 		if (mtu < IPV6_MMTU) {
346 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
347 			    "mtu=%lu sent from %s, ignoring\n",
348 			    mtu, ip6_sprintf(&ip6->ip6_src)));
349 			goto skip;
350 		}
351 
352 		/* upper bound */
353 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
354 		    ? ndi->maxmtu : ifp->if_mtu;
355 		if (mtu <= maxmtu) {
356 			int change = (ndi->linkmtu != mtu);
357 
358 			ndi->linkmtu = mtu;
359 			if (change) /* in6_maxmtu may change */
360 				in6_setmaxmtu();
361 		} else {
362 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
363 			    "mtu=%lu sent from %s; "
364 			    "exceeds maxmtu %lu, ignoring\n",
365 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
366 		}
367 	}
368 
369  skip:
370 
371 	/*
372 	 * Source link layer address
373 	 */
374     {
375 	char *lladdr = NULL;
376 	int lladdrlen = 0;
377 
378 	if (ndopts.nd_opts_src_lladdr) {
379 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
380 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
381 	}
382 
383 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
384 		nd6log((LOG_INFO,
385 		    "nd6_ra_input: lladdrlen mismatch for %s "
386 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
387 		    ifp->if_addrlen, lladdrlen - 2));
388 		goto bad;
389 	}
390 
391 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
392 
393 	/*
394 	 * Installing a link-layer address might change the state of the
395 	 * router's neighbor cache, which might also affect our on-link
396 	 * detection of adveritsed prefixes.
397 	 */
398 	pfxlist_onlink_check();
399     }
400 
401  freeit:
402 	m_freem(m);
403 	return;
404 
405  bad:
406 	icmp6stat.icp6s_badra++;
407 	m_freem(m);
408 }
409 
410 /*
411  * default router list processing sub routines
412  */
413 
414 /* tell the change to user processes watching the routing socket. */
415 static void
416 nd6_rtmsg(int cmd, struct rtentry *rt)
417 {
418 	struct rt_addrinfo info;
419 
420 	bzero((void *)&info, sizeof(info));
421 	info.rti_info[RTAX_DST] = rt_getkey(rt);
422 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
423 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
424 	if (rt->rt_ifp) {
425 		info.rti_info[RTAX_IFP] =
426 		    IFADDR_FIRST(rt->rt_ifp)->ifa_addr;
427 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
428 	}
429 
430 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
431 }
432 
433 void
434 defrouter_addreq(struct nd_defrouter *new)
435 {
436 	struct sockaddr_in6 def, mask, gate;
437 	struct rtentry *newrt = NULL;
438 	int s;
439 	int error;
440 
441 	memset(&def, 0, sizeof(def));
442 	memset(&mask, 0, sizeof(mask));
443 	memset(&gate, 0,sizeof(gate)); /* for safety */
444 
445 	def.sin6_len = mask.sin6_len = gate.sin6_len =
446 	    sizeof(struct sockaddr_in6);
447 	def.sin6_family = mask.sin6_family = gate.sin6_family = AF_INET6;
448 	gate.sin6_addr = new->rtaddr;
449 #ifndef SCOPEDROUTING
450 	gate.sin6_scope_id = 0;	/* XXX */
451 #endif
452 
453 	s = splsoftnet();
454 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
455 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
456 	    RTF_GATEWAY, &newrt);
457 	if (newrt) {
458 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
459 		newrt->rt_refcnt--;
460 	}
461 	if (error == 0)
462 		new->installed = 1;
463 	splx(s);
464 	return;
465 }
466 
467 struct nd_defrouter *
468 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
469 {
470 	struct nd_defrouter *dr;
471 
472 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
473 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
474 			break;
475 	}
476 
477 	return dr;		/* search failed */
478 }
479 
480 void
481 defrtrlist_del(struct nd_defrouter *dr)
482 {
483 	struct nd_defrouter *deldr = NULL;
484 	struct nd_prefix *pr;
485 
486 	/*
487 	 * Flush all the routing table entries that use the router
488 	 * as a next hop.
489 	 */
490 	if (!ip6_forwarding && ip6_accept_rtadv) /* XXX: better condition? */
491 		rt6_flush(&dr->rtaddr, dr->ifp);
492 
493 	if (dr->installed) {
494 		deldr = dr;
495 		defrouter_delreq(dr);
496 	}
497 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
498 
499 	/*
500 	 * Also delete all the pointers to the router in each prefix lists.
501 	 */
502 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
503 		struct nd_pfxrouter *pfxrtr;
504 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
505 			pfxrtr_del(pfxrtr);
506 	}
507 	pfxlist_onlink_check();
508 
509 	/*
510 	 * If the router is the primary one, choose a new one.
511 	 * Note that defrouter_select() will remove the current gateway
512 	 * from the routing table.
513 	 */
514 	if (deldr)
515 		defrouter_select();
516 
517 	free(dr, M_IP6NDP);
518 }
519 
520 /*
521  * Remove the default route for a given router.
522  * This is just a subroutine function for defrouter_select(), and should
523  * not be called from anywhere else.
524  */
525 static void
526 defrouter_delreq(struct nd_defrouter *dr)
527 {
528 	struct sockaddr_in6 def, mask, gw;
529 	struct rtentry *oldrt = NULL;
530 
531 #ifdef DIAGNOSTIC
532 	if (dr == NULL)
533 		panic("dr == NULL in defrouter_delreq");
534 #endif
535 
536 	bzero(&def, sizeof(def));
537 	bzero(&mask, sizeof(mask));
538 	bzero(&gw, sizeof(gw));	/* for safety */
539 
540 	def.sin6_len = mask.sin6_len = gw.sin6_len =
541 	    sizeof(struct sockaddr_in6);
542 	def.sin6_family = mask.sin6_family = gw.sin6_family = AF_INET6;
543 	gw.sin6_addr = dr->rtaddr;
544 #ifndef SCOPEDROUTING
545 	gw.sin6_scope_id = 0;	/* XXX */
546 #endif
547 
548 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
549 	    (struct sockaddr *)&gw,
550 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
551 	if (oldrt) {
552 		nd6_rtmsg(RTM_DELETE, oldrt);
553 		if (oldrt->rt_refcnt <= 0) {
554 			/*
555 			 * XXX: borrowed from the RTM_DELETE case of
556 			 * rtrequest().
557 			 */
558 			oldrt->rt_refcnt++;
559 			rtfree(oldrt);
560 		}
561 	}
562 
563 	dr->installed = 0;
564 }
565 
566 /*
567  * remove all default routes from default router list
568  */
569 void
570 defrouter_reset()
571 {
572 	struct nd_defrouter *dr;
573 
574 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
575 	     dr = TAILQ_NEXT(dr, dr_entry))
576 		defrouter_delreq(dr);
577 
578 	/*
579 	 * XXX should we also nuke any default routers in the kernel, by
580 	 * going through them by rtalloc1()?
581 	 */
582 }
583 
584 /*
585  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
586  * draft-ietf-ipngwg-router-selection:
587  * 1) Routers that are reachable or probably reachable should be preferred.
588  *    If we have more than one (probably) reachable router, prefer ones
589  *    with the highest router preference.
590  * 2) When no routers on the list are known to be reachable or
591  *    probably reachable, routers SHOULD be selected in a round-robin
592  *    fashion, regardless of router preference values.
593  * 3) If the Default Router List is empty, assume that all
594  *    destinations are on-link.
595  *
596  * We assume nd_defrouter is sorted by router preference value.
597  * Since the code below covers both with and without router preference cases,
598  * we do not need to classify the cases by ifdef.
599  *
600  * At this moment, we do not try to install more than one default router,
601  * even when the multipath routing is available, because we're not sure about
602  * the benefits for stub hosts comparing to the risk of making the code
603  * complicated and the possibility of introducing bugs.
604  */
605 void
606 defrouter_select()
607 {
608 	int s = splsoftnet();
609 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
610 	struct rtentry *rt = NULL;
611 	struct llinfo_nd6 *ln = NULL;
612 
613 	/*
614 	 * This function should be called only when acting as an autoconfigured
615 	 * host.  Although the remaining part of this function is not effective
616 	 * if the node is not an autoconfigured host, we explicitly exclude
617 	 * such cases here for safety.
618 	 */
619 	if (ip6_forwarding || !ip6_accept_rtadv) {
620 		nd6log((LOG_WARNING,
621 		    "defrouter_select: called unexpectedly (forwarding=%d, "
622 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
623 		splx(s);
624 		return;
625 	}
626 
627 	/*
628 	 * Let's handle easy case (3) first:
629 	 * If default router list is empty, there's nothing to be done.
630 	 */
631 	if (!TAILQ_FIRST(&nd_defrouter)) {
632 		splx(s);
633 		return;
634 	}
635 
636 	/*
637 	 * Search for a (probably) reachable router from the list.
638 	 * We just pick up the first reachable one (if any), assuming that
639 	 * the ordering rule of the list described in defrtrlist_update().
640 	 */
641 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
642 	     dr = TAILQ_NEXT(dr, dr_entry)) {
643 		if (selected_dr == NULL &&
644 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) != NULL &&
645 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) != NULL &&
646 		    ND6_IS_LLINFO_PROBREACH(ln)) {
647 			selected_dr = dr;
648 		}
649 
650 		if (dr->installed && !installed_dr)
651 			installed_dr = dr;
652 		else if (dr->installed && installed_dr) {
653 			/* this should not happen.  warn for diagnosis. */
654 			log(LOG_ERR, "defrouter_select: more than one router"
655 			    " is installed\n");
656 		}
657 	}
658 	/*
659 	 * If none of the default routers was found to be reachable,
660 	 * round-robin the list regardless of preference.
661 	 * Otherwise, if we have an installed router, check if the selected
662 	 * (reachable) router should really be preferred to the installed one.
663 	 * We only prefer the new router when the old one is not reachable
664 	 * or when the new one has a really higher preference value.
665 	 */
666 	if (selected_dr == NULL) {
667 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
668 			selected_dr = TAILQ_FIRST(&nd_defrouter);
669 		else
670 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
671 	} else if (installed_dr &&
672 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
673 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
674 	    ND6_IS_LLINFO_PROBREACH(ln) &&
675 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
676 		selected_dr = installed_dr;
677 	}
678 
679 	/*
680 	 * If the selected router is different than the installed one,
681 	 * remove the installed router and install the selected one.
682 	 * Note that the selected router is never NULL here.
683 	 */
684 	if (installed_dr != selected_dr) {
685 		if (installed_dr)
686 			defrouter_delreq(installed_dr);
687 		defrouter_addreq(selected_dr);
688 	}
689 
690 	splx(s);
691 	return;
692 }
693 
694 /*
695  * for default router selection
696  * regards router-preference field as a 2-bit signed integer
697  */
698 static int
699 rtpref(struct nd_defrouter *dr)
700 {
701 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
702 	case ND_RA_FLAG_RTPREF_HIGH:
703 		return (RTPREF_HIGH);
704 	case ND_RA_FLAG_RTPREF_MEDIUM:
705 	case ND_RA_FLAG_RTPREF_RSV:
706 		return (RTPREF_MEDIUM);
707 	case ND_RA_FLAG_RTPREF_LOW:
708 		return (RTPREF_LOW);
709 	default:
710 		/*
711 		 * This case should never happen.  If it did, it would mean a
712 		 * serious bug of kernel internal.  We thus always bark here.
713 		 * Or, can we even panic?
714 		 */
715 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
716 		return (RTPREF_INVALID);
717 	}
718 	/* NOTREACHED */
719 }
720 
721 static struct nd_defrouter *
722 defrtrlist_update(struct nd_defrouter *new)
723 {
724 	struct nd_defrouter *dr, *n;
725 	int s = splsoftnet();
726 
727 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
728 		/* entry exists */
729 		if (new->rtlifetime == 0) {
730 			defrtrlist_del(dr);
731 			dr = NULL;
732 		} else {
733 			int oldpref = rtpref(dr);
734 
735 			/* override */
736 			dr->flags = new->flags; /* xxx flag check */
737 			dr->rtlifetime = new->rtlifetime;
738 			dr->expire = new->expire;
739 
740 			/*
741 			 * If the preference does not change, there's no need
742 			 * to sort the entries.
743 			 */
744 			if (rtpref(new) == oldpref) {
745 				splx(s);
746 				return (dr);
747 			}
748 
749 			/*
750 			 * preferred router may be changed, so relocate
751 			 * this router.
752 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
753 			 * However, since defrtrlist_del() has many side
754 			 * effects, we intentionally do so here.
755 			 * defrouter_select() below will handle routing
756 			 * changes later.
757 			 */
758 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
759 			n = dr;
760 			goto insert;
761 		}
762 		splx(s);
763 		return (dr);
764 	}
765 
766 	/* entry does not exist */
767 	if (new->rtlifetime == 0) {
768 		splx(s);
769 		return (NULL);
770 	}
771 
772 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
773 	if (n == NULL) {
774 		splx(s);
775 		return (NULL);
776 	}
777 	bzero(n, sizeof(*n));
778 	*n = *new;
779 
780 insert:
781 	/*
782 	 * Insert the new router in the Default Router List;
783 	 * The Default Router List should be in the descending order
784 	 * of router-preferece.  Routers with the same preference are
785 	 * sorted in the arriving time order.
786 	 */
787 
788 	/* insert at the end of the group */
789 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
790 	     dr = TAILQ_NEXT(dr, dr_entry)) {
791 		if (rtpref(n) > rtpref(dr))
792 			break;
793 	}
794 	if (dr)
795 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
796 	else
797 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
798 
799 	defrouter_select();
800 
801 	splx(s);
802 
803 	return (n);
804 }
805 
806 static struct nd_pfxrouter *
807 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
808 {
809 	struct nd_pfxrouter *search;
810 
811 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
812 		if (search->router == dr)
813 			break;
814 	}
815 
816 	return (search);
817 }
818 
819 static void
820 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
821 {
822 	struct nd_pfxrouter *new;
823 
824 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
825 	if (new == NULL)
826 		return;
827 	bzero(new, sizeof(*new));
828 	new->router = dr;
829 
830 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
831 
832 	pfxlist_onlink_check();
833 }
834 
835 static void
836 pfxrtr_del(struct nd_pfxrouter *pfr)
837 {
838 	LIST_REMOVE(pfr, pfr_entry);
839 	free(pfr, M_IP6NDP);
840 }
841 
842 struct nd_prefix *
843 nd6_prefix_lookup(struct nd_prefixctl *key)
844 {
845 	struct nd_prefix *search;
846 
847 	LIST_FOREACH(search, &nd_prefix, ndpr_entry) {
848 		if (key->ndpr_ifp == search->ndpr_ifp &&
849 		    key->ndpr_plen == search->ndpr_plen &&
850 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
851 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
852 			break;
853 		}
854 	}
855 
856 	return (search);
857 }
858 
859 int
860 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
861 	struct nd_prefix **newp)
862 {
863 	struct nd_prefix *new = NULL;
864 	int i, s;
865 	int error;
866 
867 	error = 0;
868 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
869 	if (new == NULL)
870 		return ENOMEM;
871 	bzero(new, sizeof(*new));
872 	new->ndpr_ifp = pr->ndpr_ifp;
873 	new->ndpr_prefix = pr->ndpr_prefix;
874 	new->ndpr_plen = pr->ndpr_plen;
875 	new->ndpr_vltime = pr->ndpr_vltime;
876 	new->ndpr_pltime = pr->ndpr_pltime;
877 	new->ndpr_flags = pr->ndpr_flags;
878 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
879 		free(new, M_IP6NDP);
880 		return(error);
881 	}
882 	new->ndpr_lastupdate = time_second;
883 	if (newp != NULL)
884 		*newp = new;
885 
886 	/* initialization */
887 	LIST_INIT(&new->ndpr_advrtrs);
888 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
889 	/* make prefix in the canonical form */
890 	for (i = 0; i < 4; i++)
891 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
892 		    new->ndpr_mask.s6_addr32[i];
893 
894 	s = splsoftnet();
895 	/* link ndpr_entry to nd_prefix list */
896 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
897 	splx(s);
898 
899 	/* ND_OPT_PI_FLAG_ONLINK processing */
900 	if (new->ndpr_raf_onlink) {
901 		int e;
902 
903 		if ((e = nd6_prefix_onlink(new)) != 0) {
904 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
905 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
906 			    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
907 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
908 			/* proceed anyway. XXX: is it correct? */
909 		}
910 	}
911 
912 	if (dr)
913 		pfxrtr_add(new, dr);
914 
915 	return 0;
916 }
917 
918 void
919 prelist_remove(struct nd_prefix *pr)
920 {
921 	struct nd_pfxrouter *pfr, *next;
922 	int e, s;
923 
924 	/* make sure to invalidate the prefix until it is really freed. */
925 	pr->ndpr_vltime = 0;
926 	pr->ndpr_pltime = 0;
927 #if 0
928 	/*
929 	 * Though these flags are now meaningless, we'd rather keep the value
930 	 * not to confuse users when executing "ndp -p".
931 	 */
932 	pr->ndpr_raf_onlink = 0;
933 	pr->ndpr_raf_auto = 0;
934 #endif
935 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
936 	    (e = nd6_prefix_offlink(pr)) != 0) {
937 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
938 		    "on %s, errno=%d\n",
939 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
940 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
941 		/* what should we do? */
942 	}
943 
944 	if (pr->ndpr_refcnt > 0)
945 		return;		/* notice here? */
946 
947 	s = splsoftnet();
948 	/* unlink ndpr_entry from nd_prefix list */
949 	LIST_REMOVE(pr, ndpr_entry);
950 
951 	/* free list of routers that adversed the prefix */
952 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
953 		next = LIST_NEXT(pfr, pfr_entry);
954 
955 		free(pfr, M_IP6NDP);
956 	}
957 	splx(s);
958 
959 	free(pr, M_IP6NDP);
960 
961 	pfxlist_onlink_check();
962 }
963 
964 static int
965 prelist_update(struct nd_prefixctl *new,
966 	struct nd_defrouter *dr, /* may be NULL */
967 	struct mbuf *m,
968 	int mcast)
969 {
970 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
971 	struct ifaddr *ifa;
972 	struct ifnet *ifp = new->ndpr_ifp;
973 	struct nd_prefix *pr;
974 	int s = splsoftnet();
975 	int error = 0;
976 	int newprefix = 0;
977 	int auth;
978 	struct in6_addrlifetime lt6_tmp;
979 
980 	auth = 0;
981 	if (m) {
982 		/*
983 		 * Authenticity for NA consists authentication for
984 		 * both IP header and IP datagrams, doesn't it ?
985 		 */
986 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
987 		auth = (m->m_flags & M_AUTHIPHDR
988 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
989 #endif
990 	}
991 
992 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
993 		/*
994 		 * nd6_prefix_lookup() ensures that pr and new have the same
995 		 * prefix on a same interface.
996 		 */
997 
998 		/*
999 		 * Update prefix information.  Note that the on-link (L) bit
1000 		 * and the autonomous (A) bit should NOT be changed from 1
1001 		 * to 0.
1002 		 */
1003 		if (new->ndpr_raf_onlink == 1)
1004 			pr->ndpr_raf_onlink = 1;
1005 		if (new->ndpr_raf_auto == 1)
1006 			pr->ndpr_raf_auto = 1;
1007 		if (new->ndpr_raf_onlink) {
1008 			pr->ndpr_vltime = new->ndpr_vltime;
1009 			pr->ndpr_pltime = new->ndpr_pltime;
1010 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1011 			pr->ndpr_lastupdate = time_second;
1012 		}
1013 
1014 		if (new->ndpr_raf_onlink &&
1015 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1016 			int e;
1017 
1018 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1019 				nd6log((LOG_ERR,
1020 				    "prelist_update: failed to make "
1021 				    "the prefix %s/%d on-link on %s "
1022 				    "(errno=%d)\n",
1023 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1024 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1025 				/* proceed anyway. XXX: is it correct? */
1026 			}
1027 		}
1028 
1029 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1030 			pfxrtr_add(pr, dr);
1031 	} else {
1032 		struct nd_prefix *newpr = NULL;
1033 
1034 		newprefix = 1;
1035 
1036 		if (new->ndpr_vltime == 0)
1037 			goto end;
1038 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1039 			goto end;
1040 
1041 		error = nd6_prelist_add(new, dr, &newpr);
1042 		if (error != 0 || newpr == NULL) {
1043 			nd6log((LOG_NOTICE, "prelist_update: "
1044 			    "nd6_prelist_add failed for %s/%d on %s "
1045 			    "errno=%d, returnpr=%p\n",
1046 			    ip6_sprintf(&new->ndpr_prefix.sin6_addr),
1047 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1048 			    error, newpr));
1049 			goto end; /* we should just give up in this case. */
1050 		}
1051 
1052 		/*
1053 		 * XXX: from the ND point of view, we can ignore a prefix
1054 		 * with the on-link bit being zero.  However, we need a
1055 		 * prefix structure for references from autoconfigured
1056 		 * addresses.  Thus, we explicitly make sure that the prefix
1057 		 * itself expires now.
1058 		 */
1059 		if (newpr->ndpr_raf_onlink == 0) {
1060 			newpr->ndpr_vltime = 0;
1061 			newpr->ndpr_pltime = 0;
1062 			in6_init_prefix_ltimes(newpr);
1063 		}
1064 
1065 		pr = newpr;
1066 	}
1067 
1068 	/*
1069 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1070 	 * Note that pr must be non NULL at this point.
1071 	 */
1072 
1073 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1074 	if (!new->ndpr_raf_auto)
1075 		goto end;
1076 
1077 	/*
1078 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1079 	 * nd6_ra_input.
1080 	 */
1081 
1082 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1083 	if (new->ndpr_pltime > new->ndpr_vltime) {
1084 		error = EINVAL;	/* XXX: won't be used */
1085 		goto end;
1086 	}
1087 
1088 	/*
1089 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1090 	 * an address configured by stateless autoconfiguration already in the
1091 	 * list of addresses associated with the interface, and the Valid
1092 	 * Lifetime is not 0, form an address.  We first check if we have
1093 	 * a matching prefix.
1094 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1095 	 * consider autoconfigured addresses while RFC2462 simply said
1096 	 * "address".
1097 	 */
1098 	IFADDR_FOREACH(ifa, ifp) {
1099 		struct in6_ifaddr *ifa6;
1100 		u_int32_t remaininglifetime;
1101 
1102 		if (ifa->ifa_addr->sa_family != AF_INET6)
1103 			continue;
1104 
1105 		ifa6 = (struct in6_ifaddr *)ifa;
1106 
1107 		/*
1108 		 * We only consider autoconfigured addresses as per rfc2462bis.
1109 		 */
1110 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1111 			continue;
1112 
1113 		/*
1114 		 * Spec is not clear here, but I believe we should concentrate
1115 		 * on unicast (i.e. not anycast) addresses.
1116 		 * XXX: other ia6_flags? detached or duplicated?
1117 		 */
1118 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1119 			continue;
1120 
1121 		/*
1122 		 * Ignore the address if it is not associated with a prefix
1123 		 * or is associated with a prefix that is different from this
1124 		 * one.  (pr is never NULL here)
1125 		 */
1126 		if (ifa6->ia6_ndpr != pr)
1127 			continue;
1128 
1129 		if (ia6_match == NULL) /* remember the first one */
1130 			ia6_match = ifa6;
1131 
1132 		/*
1133 		 * An already autoconfigured address matched.  Now that we
1134 		 * are sure there is at least one matched address, we can
1135 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1136 		 * "two hours" rule and the privacy extension.
1137 		 * We apply some clarifications in rfc2462bis:
1138 		 * - use remaininglifetime instead of storedlifetime as a
1139 		 *   variable name
1140 		 * - remove the dead code in the "two-hour" rule
1141 		 */
1142 #define TWOHOUR		(120*60)
1143 		lt6_tmp = ifa6->ia6_lifetime;
1144 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1145 			remaininglifetime = ND6_INFINITE_LIFETIME;
1146 		else if (time_second - ifa6->ia6_updatetime >
1147 			 lt6_tmp.ia6t_vltime) {
1148 			/*
1149 			 * The case of "invalid" address.  We should usually
1150 			 * not see this case.
1151 			 */
1152 			remaininglifetime = 0;
1153 		} else
1154 			remaininglifetime = lt6_tmp.ia6t_vltime -
1155 			    (time_second - ifa6->ia6_updatetime);
1156 
1157 		/* when not updating, keep the current stored lifetime. */
1158 		lt6_tmp.ia6t_vltime = remaininglifetime;
1159 
1160 		if (TWOHOUR < new->ndpr_vltime ||
1161 		    remaininglifetime < new->ndpr_vltime) {
1162 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1163 		} else if (remaininglifetime <= TWOHOUR) {
1164 			if (auth)
1165 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1166 		} else {
1167 			/*
1168 			 * new->ndpr_vltime <= TWOHOUR &&
1169 			 * TWOHOUR < remaininglifetime
1170 			 */
1171 			lt6_tmp.ia6t_vltime = TWOHOUR;
1172 		}
1173 
1174 		/* The 2 hour rule is not imposed for preferred lifetime. */
1175 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1176 
1177 		in6_init_address_ltimes(pr, &lt6_tmp);
1178 
1179 		/*
1180 		 * We need to treat lifetimes for temporary addresses
1181 		 * differently, according to
1182 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1183 		 * we only update the lifetimes when they are in the maximum
1184 		 * intervals.
1185 		 */
1186 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1187 			u_int32_t maxvltime, maxpltime;
1188 
1189 			if (ip6_temp_valid_lifetime >
1190 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1191 			    ip6_desync_factor)) {
1192 				maxvltime = ip6_temp_valid_lifetime -
1193 				    (time_second - ifa6->ia6_createtime) -
1194 				    ip6_desync_factor;
1195 			} else
1196 				maxvltime = 0;
1197 			if (ip6_temp_preferred_lifetime >
1198 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1199 			    ip6_desync_factor)) {
1200 				maxpltime = ip6_temp_preferred_lifetime -
1201 				    (time_second - ifa6->ia6_createtime) -
1202 				    ip6_desync_factor;
1203 			} else
1204 				maxpltime = 0;
1205 
1206 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1207 			    lt6_tmp.ia6t_vltime > maxvltime) {
1208 				lt6_tmp.ia6t_vltime = maxvltime;
1209 			}
1210 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1211 			    lt6_tmp.ia6t_pltime > maxpltime) {
1212 				lt6_tmp.ia6t_pltime = maxpltime;
1213 			}
1214 		}
1215 
1216 		ifa6->ia6_lifetime = lt6_tmp;
1217 		ifa6->ia6_updatetime = time_second;
1218 	}
1219 	if (ia6_match == NULL && new->ndpr_vltime) {
1220 		int ifidlen;
1221 
1222 		/*
1223 		 * 5.5.3 (d) (continued)
1224 		 * No address matched and the valid lifetime is non-zero.
1225 		 * Create a new address.
1226 		 */
1227 
1228 		/*
1229 		 * Prefix Length check:
1230 		 * If the sum of the prefix length and interface identifier
1231 		 * length does not equal 128 bits, the Prefix Information
1232 		 * option MUST be ignored.  The length of the interface
1233 		 * identifier is defined in a separate link-type specific
1234 		 * document.
1235 		 */
1236 		ifidlen = in6_if2idlen(ifp);
1237 		if (ifidlen < 0) {
1238 			/* this should not happen, so we always log it. */
1239 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1240 			    if_name(ifp));
1241 			goto end;
1242 		}
1243 		if (ifidlen + pr->ndpr_plen != 128) {
1244 			nd6log((LOG_INFO,
1245 			    "prelist_update: invalid prefixlen "
1246 			    "%d for %s, ignored\n",
1247 			    pr->ndpr_plen, if_name(ifp)));
1248 			goto end;
1249 		}
1250 
1251 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1252 			/*
1253 			 * note that we should use pr (not new) for reference.
1254 			 */
1255 			pr->ndpr_refcnt++;
1256 			ia6->ia6_ndpr = pr;
1257 
1258 			/*
1259 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
1260 			 * When a new public address is created as described
1261 			 * in RFC2462, also create a new temporary address.
1262 			 *
1263 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
1264 			 * When an interface connects to a new link, a new
1265 			 * randomized interface identifier should be generated
1266 			 * immediately together with a new set of temporary
1267 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1268 			 * in6_tmpifadd().
1269 			 */
1270 			if (ip6_use_tempaddr) {
1271 				int e;
1272 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1273 					nd6log((LOG_NOTICE, "prelist_update: "
1274 					    "failed to create a temporary "
1275 					    "address, errno=%d\n",
1276 					    e));
1277 				}
1278 			}
1279 
1280 			/*
1281 			 * A newly added address might affect the status
1282 			 * of other addresses, so we check and update it.
1283 			 * XXX: what if address duplication happens?
1284 			 */
1285 			pfxlist_onlink_check();
1286 		} else {
1287 			/* just set an error. do not bark here. */
1288 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1289 		}
1290 	}
1291 
1292  end:
1293 	splx(s);
1294 	return error;
1295 }
1296 
1297 /*
1298  * A supplement function used in the on-link detection below;
1299  * detect if a given prefix has a (probably) reachable advertising router.
1300  * XXX: lengthy function name...
1301  */
1302 static struct nd_pfxrouter *
1303 find_pfxlist_reachable_router(struct nd_prefix *pr)
1304 {
1305 	struct nd_pfxrouter *pfxrtr;
1306 	struct rtentry *rt;
1307 	struct llinfo_nd6 *ln;
1308 
1309 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1310 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1311 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1312 		    pfxrtr->router->ifp)) &&
1313 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1314 		    ND6_IS_LLINFO_PROBREACH(ln))
1315 			break;	/* found */
1316 	}
1317 
1318 	return (pfxrtr);
1319 }
1320 
1321 /*
1322  * Check if each prefix in the prefix list has at least one available router
1323  * that advertised the prefix (a router is "available" if its neighbor cache
1324  * entry is reachable or probably reachable).
1325  * If the check fails, the prefix may be off-link, because, for example,
1326  * we have moved from the network but the lifetime of the prefix has not
1327  * expired yet.  So we should not use the prefix if there is another prefix
1328  * that has an available router.
1329  * But, if there is no prefix that has an available router, we still regards
1330  * all the prefixes as on-link.  This is because we can't tell if all the
1331  * routers are simply dead or if we really moved from the network and there
1332  * is no router around us.
1333  */
1334 void
1335 pfxlist_onlink_check()
1336 {
1337 	struct nd_prefix *pr;
1338 	struct in6_ifaddr *ifa;
1339 	struct nd_defrouter *dr;
1340 	struct nd_pfxrouter *pfxrtr = NULL;
1341 
1342 	/*
1343 	 * Check if there is a prefix that has a reachable advertising
1344 	 * router.
1345 	 */
1346 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1347 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1348 			break;
1349 	}
1350 	/*
1351 	 * If we have no such prefix, check whether we still have a router
1352 	 * that does not advertise any prefixes.
1353 	 */
1354 	if (pr == NULL) {
1355 		TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1356 			struct nd_prefix *pr0;
1357 
1358 			LIST_FOREACH(pr0, &nd_prefix, ndpr_entry) {
1359 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1360 					break;
1361 			}
1362 			if (pfxrtr)
1363 				break;
1364 		}
1365 	}
1366 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
1367 		/*
1368 		 * There is at least one prefix that has a reachable router,
1369 		 * or at least a router which probably does not advertise
1370 		 * any prefixes.  The latter would be the case when we move
1371 		 * to a new link where we have a router that does not provide
1372 		 * prefixes and we configure an address by hand.
1373 		 * Detach prefixes which have no reachable advertising
1374 		 * router, and attach other prefixes.
1375 		 */
1376 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1377 			/* XXX: a link-local prefix should never be detached */
1378 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1379 				continue;
1380 
1381 			/*
1382 			 * we aren't interested in prefixes without the L bit
1383 			 * set.
1384 			 */
1385 			if (pr->ndpr_raf_onlink == 0)
1386 				continue;
1387 
1388 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1389 			    find_pfxlist_reachable_router(pr) == NULL)
1390 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1391 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1392 			    find_pfxlist_reachable_router(pr) != 0)
1393 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1394 		}
1395 	} else {
1396 		/* there is no prefix that has a reachable router */
1397 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1398 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1399 				continue;
1400 
1401 			if (pr->ndpr_raf_onlink == 0)
1402 				continue;
1403 
1404 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1405 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1406 		}
1407 	}
1408 
1409 	/*
1410 	 * Remove each interface route associated with a (just) detached
1411 	 * prefix, and reinstall the interface route for a (just) attached
1412 	 * prefix.  Note that all attempt of reinstallation does not
1413 	 * necessarily success, when a same prefix is shared among multiple
1414 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1415 	 * so we don't have to care about them.
1416 	 */
1417 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1418 		int e;
1419 
1420 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1421 			continue;
1422 
1423 		if (pr->ndpr_raf_onlink == 0)
1424 			continue;
1425 
1426 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1427 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1428 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1429 				nd6log((LOG_ERR,
1430 				    "pfxlist_onlink_check: failed to "
1431 				    "make %s/%d offlink, errno=%d\n",
1432 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1433 				    pr->ndpr_plen, e));
1434 			}
1435 		}
1436 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1437 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1438 		    pr->ndpr_raf_onlink) {
1439 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1440 				nd6log((LOG_ERR,
1441 				    "pfxlist_onlink_check: failed to "
1442 				    "make %s/%d onlink, errno=%d\n",
1443 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1444 				    pr->ndpr_plen, e));
1445 			}
1446 		}
1447 	}
1448 
1449 	/*
1450 	 * Changes on the prefix status might affect address status as well.
1451 	 * Make sure that all addresses derived from an attached prefix are
1452 	 * attached, and that all addresses derived from a detached prefix are
1453 	 * detached.  Note, however, that a manually configured address should
1454 	 * always be attached.
1455 	 * The precise detection logic is same as the one for prefixes.
1456 	 */
1457 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1458 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1459 			continue;
1460 
1461 		if (ifa->ia6_ndpr == NULL) {
1462 			/*
1463 			 * This can happen when we first configure the address
1464 			 * (i.e. the address exists, but the prefix does not).
1465 			 * XXX: complicated relationships...
1466 			 */
1467 			continue;
1468 		}
1469 
1470 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1471 			break;
1472 	}
1473 	if (ifa) {
1474 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1475 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1476 				continue;
1477 
1478 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1479 				continue;
1480 
1481 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1482 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1483 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1484 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1485 					nd6_dad_start((struct ifaddr *)ifa,
1486 					    0);
1487 				}
1488 			} else {
1489 				if ((ifa->ia6_flags & IN6_IFF_DETACHED) == 0) {
1490 					ifa->ia6_flags |= IN6_IFF_DETACHED;
1491 				}
1492 			}
1493 		}
1494 	}
1495 	else {
1496 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1497 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1498 				continue;
1499 
1500 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1501 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1502 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1503 				/* Do we need a delay in this case? */
1504 				nd6_dad_start((struct ifaddr *)ifa, 0);
1505 			}
1506 		}
1507 	}
1508 }
1509 
1510 int
1511 nd6_prefix_onlink(struct nd_prefix *pr)
1512 {
1513 	struct ifaddr *ifa;
1514 	struct ifnet *ifp = pr->ndpr_ifp;
1515 	struct sockaddr_in6 mask6;
1516 	struct nd_prefix *opr;
1517 	u_long rtflags;
1518 	int error = 0;
1519 	struct rtentry *rt = NULL;
1520 
1521 	/* sanity check */
1522 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1523 		nd6log((LOG_ERR,
1524 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1525 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1526 		return (EEXIST);
1527 	}
1528 
1529 	/*
1530 	 * Add the interface route associated with the prefix.  Before
1531 	 * installing the route, check if there's the same prefix on another
1532 	 * interface, and the prefix has already installed the interface route.
1533 	 * Although such a configuration is expected to be rare, we explicitly
1534 	 * allow it.
1535 	 */
1536 	LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1537 		if (opr == pr)
1538 			continue;
1539 
1540 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1541 			continue;
1542 
1543 		if (opr->ndpr_plen == pr->ndpr_plen &&
1544 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1545 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1546 			return (0);
1547 	}
1548 
1549 	/*
1550 	 * We prefer link-local addresses as the associated interface address.
1551 	 */
1552 	/* search for a link-local addr */
1553 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1554 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1555 	if (ifa == NULL) {
1556 		/* XXX: freebsd does not have ifa_ifwithaf */
1557 		IFADDR_FOREACH(ifa, ifp) {
1558 			if (ifa->ifa_addr->sa_family == AF_INET6)
1559 				break;
1560 		}
1561 		/* should we care about ia6_flags? */
1562 	}
1563 	if (ifa == NULL) {
1564 		/*
1565 		 * This can still happen, when, for example, we receive an RA
1566 		 * containing a prefix with the L bit set and the A bit clear,
1567 		 * after removing all IPv6 addresses on the receiving
1568 		 * interface.  This should, of course, be rare though.
1569 		 */
1570 		nd6log((LOG_NOTICE,
1571 		    "nd6_prefix_onlink: failed to find any ifaddr"
1572 		    " to add route for a prefix(%s/%d) on %s\n",
1573 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1574 		    pr->ndpr_plen, if_name(ifp)));
1575 		return (0);
1576 	}
1577 
1578 	/*
1579 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1580 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1581 	 */
1582 	bzero(&mask6, sizeof(mask6));
1583 	mask6.sin6_len = sizeof(mask6);
1584 	mask6.sin6_addr = pr->ndpr_mask;
1585 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
1586 	rtflags = ifa->ifa_flags | RTF_UP;
1587 	if (nd6_need_cache(ifp)) {
1588 		/* explicitly set in case ifa_flags does not set the flag. */
1589 		rtflags |= RTF_CLONING;
1590 	} else {
1591 		/*
1592 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1593 		 */
1594 		rtflags &= ~RTF_CLONING;
1595 	}
1596 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1597 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1598 	if (error == 0) {
1599 		if (rt != NULL) /* this should be non NULL, though */
1600 			nd6_rtmsg(RTM_ADD, rt);
1601 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1602 	} else {
1603 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1604 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1605 		    "errno = %d\n",
1606 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1607 		    pr->ndpr_plen, if_name(ifp),
1608 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1609 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
1610 	}
1611 
1612 	if (rt != NULL)
1613 		rt->rt_refcnt--;
1614 
1615 	return (error);
1616 }
1617 
1618 int
1619 nd6_prefix_offlink(struct nd_prefix *pr)
1620 {
1621 	int error = 0;
1622 	struct ifnet *ifp = pr->ndpr_ifp;
1623 	struct nd_prefix *opr;
1624 	struct sockaddr_in6 sa6, mask6;
1625 	struct rtentry *rt = NULL;
1626 
1627 	/* sanity check */
1628 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1629 		nd6log((LOG_ERR,
1630 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1631 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1632 		return (EEXIST);
1633 	}
1634 
1635 	sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1636 	sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
1637 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1638 	    (struct sockaddr *)&mask6, 0, &rt);
1639 	if (error == 0) {
1640 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1641 
1642 		/* report the route deletion to the routing socket. */
1643 		if (rt != NULL)
1644 			nd6_rtmsg(RTM_DELETE, rt);
1645 
1646 		/*
1647 		 * There might be the same prefix on another interface,
1648 		 * the prefix which could not be on-link just because we have
1649 		 * the interface route (see comments in nd6_prefix_onlink).
1650 		 * If there's one, try to make the prefix on-link on the
1651 		 * interface.
1652 		 */
1653 		LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1654 			if (opr == pr)
1655 				continue;
1656 
1657 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1658 				continue;
1659 
1660 			/*
1661 			 * KAME specific: detached prefixes should not be
1662 			 * on-link.
1663 			 */
1664 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1665 				continue;
1666 
1667 			if (opr->ndpr_plen == pr->ndpr_plen &&
1668 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1669 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1670 				int e;
1671 
1672 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1673 					nd6log((LOG_ERR,
1674 					    "nd6_prefix_offlink: failed to "
1675 					    "recover a prefix %s/%d from %s "
1676 					    "to %s (errno = %d)\n",
1677 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1678 					    opr->ndpr_plen, if_name(ifp),
1679 					    if_name(opr->ndpr_ifp), e));
1680 				}
1681 			}
1682 		}
1683 	} else {
1684 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1685 		nd6log((LOG_ERR,
1686 		    "nd6_prefix_offlink: failed to delete route: "
1687 		    "%s/%d on %s (errno = %d)\n",
1688 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1689 		    error));
1690 	}
1691 
1692 	if (rt != NULL) {
1693 		if (rt->rt_refcnt <= 0) {
1694 			/* XXX: we should free the entry ourselves. */
1695 			rt->rt_refcnt++;
1696 			rtfree(rt);
1697 		}
1698 	}
1699 
1700 	return (error);
1701 }
1702 
1703 static struct in6_ifaddr *
1704 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1705 {
1706 	struct ifnet *ifp = pr->ndpr_ifp;
1707 	struct ifaddr *ifa;
1708 	struct in6_aliasreq ifra;
1709 	struct in6_ifaddr *ia, *ib;
1710 	int error, plen0;
1711 	struct in6_addr mask;
1712 	int prefixlen = pr->ndpr_plen;
1713 	int updateflags;
1714 
1715 	in6_prefixlen2mask(&mask, prefixlen);
1716 
1717 	/*
1718 	 * find a link-local address (will be interface ID).
1719 	 * Is it really mandatory? Theoretically, a global or a site-local
1720 	 * address can be configured without a link-local address, if we
1721 	 * have a unique interface identifier...
1722 	 *
1723 	 * it is not mandatory to have a link-local address, we can generate
1724 	 * interface identifier on the fly.  we do this because:
1725 	 * (1) it should be the easiest way to find interface identifier.
1726 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1727 	 * for multiple addresses on a single interface, and possible shortcut
1728 	 * of DAD.  we omitted DAD for this reason in the past.
1729 	 * (3) a user can prevent autoconfiguration of global address
1730 	 * by removing link-local address by hand (this is partly because we
1731 	 * don't have other way to control the use of IPv6 on an interface.
1732 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1733 	 * (4) it is easier to manage when an interface has addresses
1734 	 * with the same interface identifier, than to have multiple addresses
1735 	 * with different interface identifiers.
1736 	 */
1737 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1738 	if (ifa)
1739 		ib = (struct in6_ifaddr *)ifa;
1740 	else
1741 		return NULL;
1742 
1743 #if 0 /* don't care link local addr state, and always do DAD */
1744 	/* if link-local address is not eligible, do not autoconfigure. */
1745 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1746 		printf("in6_ifadd: link-local address not ready\n");
1747 		return NULL;
1748 	}
1749 #endif
1750 
1751 	/* prefixlen + ifidlen must be equal to 128 */
1752 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1753 	if (prefixlen != plen0) {
1754 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1755 		    "(prefix=%d ifid=%d)\n",
1756 		    if_name(ifp), prefixlen, 128 - plen0));
1757 		return NULL;
1758 	}
1759 
1760 	/* make ifaddr */
1761 
1762 	memset(&ifra, 0, sizeof(ifra));
1763 	/*
1764 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1765 	 * for safety.
1766 	 */
1767 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1768 	sockaddr_in6_init(&ifra.ifra_addr, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1769 	/* prefix */
1770 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1771 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1772 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1773 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1774 
1775 	/* interface ID */
1776 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1777 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1778 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1779 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1780 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1781 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1782 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1783 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1784 
1785 	/* new prefix mask. */
1786 	sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
1787 
1788 	/* lifetimes */
1789 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1790 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1791 
1792 	/* XXX: scope zone ID? */
1793 
1794 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1795 
1796 	/*
1797 	 * Make sure that we do not have this address already.  This should
1798 	 * usually not happen, but we can still see this case, e.g., if we
1799 	 * have manually configured the exact address to be configured.
1800 	 */
1801 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1802 		/* this should be rare enough to make an explicit log */
1803 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1804 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1805 		return (NULL);
1806 	}
1807 
1808 	/*
1809 	 * Allocate ifaddr structure, link into chain, etc.
1810 	 * If we are going to create a new address upon receiving a multicasted
1811 	 * RA, we need to impose a random delay before starting DAD.
1812 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1813 	 */
1814 	updateflags = 0;
1815 	if (mcast)
1816 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1817 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1818 		nd6log((LOG_ERR,
1819 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1820 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1821 		    error));
1822 		return (NULL);	/* ifaddr must not have been allocated. */
1823 	}
1824 
1825 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1826 
1827 	return (ia);		/* this is always non-NULL */
1828 }
1829 
1830 int
1831 in6_tmpifadd(
1832 	const struct in6_ifaddr *ia0, /* corresponding public address */
1833 	int forcegen,
1834 	int dad_delay)
1835 {
1836 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1837 	struct in6_ifaddr *newia, *ia;
1838 	struct in6_aliasreq ifra;
1839 	int i, error;
1840 	int trylimit = 3;	/* XXX: adhoc value */
1841 	int updateflags;
1842 	u_int32_t randid[2];
1843 	u_int32_t vltime0, pltime0;
1844 
1845 	memset(&ifra, 0, sizeof(ifra));
1846 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1847 	ifra.ifra_addr = ia0->ia_addr;
1848 	/* copy prefix mask */
1849 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1850 	/* clear the old IFID */
1851 	for (i = 0; i < 4; i++) {
1852 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1853 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1854 	}
1855 
1856   again:
1857 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1858 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1859 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1860 		    "random IFID\n"));
1861 		return (EINVAL);
1862 	}
1863 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1864 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1865 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1866 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1867 
1868 	/*
1869 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1870 	 * However, we may still have a chance to see collision, because
1871 	 * there may be a time lag between generation of the ID and generation
1872 	 * of the address.  So, we'll do one more sanity check.
1873 	 */
1874 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1875 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1876 		    &ifra.ifra_addr.sin6_addr)) {
1877 			if (trylimit-- == 0) {
1878 				/*
1879 				 * Give up.  Something strange should have
1880 				 * happened.
1881 				 */
1882 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1883 				    "find a unique random IFID\n"));
1884 				return (EEXIST);
1885 			}
1886 			forcegen = 1;
1887 			goto again;
1888 		}
1889 	}
1890 
1891 	/*
1892 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1893          * public address or TEMP_VALID_LIFETIME.
1894 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1895          * of the public address or TEMP_PREFERRED_LIFETIME -
1896          * DESYNC_FACTOR.
1897 	 */
1898 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1899 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1900 		    (ia0->ia6_lifetime.ia6t_vltime -
1901 		    (time_second - ia0->ia6_updatetime));
1902 		if (vltime0 > ip6_temp_valid_lifetime)
1903 			vltime0 = ip6_temp_valid_lifetime;
1904 	} else
1905 		vltime0 = ip6_temp_valid_lifetime;
1906 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1907 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1908 		    (ia0->ia6_lifetime.ia6t_pltime -
1909 		    (time_second - ia0->ia6_updatetime));
1910 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
1911 			pltime0 = ip6_temp_preferred_lifetime -
1912 			    ip6_desync_factor;
1913 		}
1914 	} else
1915 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
1916 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1917 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1918 
1919 	/*
1920 	 * A temporary address is created only if this calculated Preferred
1921 	 * Lifetime is greater than REGEN_ADVANCE time units.
1922 	 */
1923 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
1924 		return (0);
1925 
1926 	/* XXX: scope zone ID? */
1927 
1928 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1929 
1930 	/* allocate ifaddr structure, link into chain, etc. */
1931 	updateflags = 0;
1932 	if (dad_delay)
1933 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1934 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1935 		return (error);
1936 
1937 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1938 	if (newia == NULL) {	/* XXX: can it happen? */
1939 		nd6log((LOG_ERR,
1940 		    "in6_tmpifadd: ifa update succeeded, but we got "
1941 		    "no ifaddr\n"));
1942 		return (EINVAL); /* XXX */
1943 	}
1944 	newia->ia6_ndpr = ia0->ia6_ndpr;
1945 	newia->ia6_ndpr->ndpr_refcnt++;
1946 
1947 	/*
1948 	 * A newly added address might affect the status of other addresses.
1949 	 * XXX: when the temporary address is generated with a new public
1950 	 * address, the onlink check is redundant.  However, it would be safe
1951 	 * to do the check explicitly everywhere a new address is generated,
1952 	 * and, in fact, we surely need the check when we create a new
1953 	 * temporary address due to deprecation of an old temporary address.
1954 	 */
1955 	pfxlist_onlink_check();
1956 
1957 	return (0);
1958 }
1959 
1960 static int
1961 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1962 {
1963 
1964 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
1965 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
1966 		nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
1967 		    "(%d) is greater than valid lifetime(%d)\n",
1968 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
1969 		return (EINVAL);
1970 	}
1971 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1972 		ndpr->ndpr_preferred = 0;
1973 	else
1974 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
1975 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1976 		ndpr->ndpr_expire = 0;
1977 	else
1978 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
1979 
1980 	return 0;
1981 }
1982 
1983 static void
1984 in6_init_address_ltimes(struct nd_prefix *new,
1985     struct in6_addrlifetime *lt6)
1986 {
1987 
1988 	/* Valid lifetime must not be updated unless explicitly specified. */
1989 	/* init ia6t_expire */
1990 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
1991 		lt6->ia6t_expire = 0;
1992 	else {
1993 		lt6->ia6t_expire = time_second;
1994 		lt6->ia6t_expire += lt6->ia6t_vltime;
1995 	}
1996 
1997 	/* init ia6t_preferred */
1998 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
1999 		lt6->ia6t_preferred = 0;
2000 	else {
2001 		lt6->ia6t_preferred = time_second;
2002 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2003 	}
2004 }
2005 
2006 /*
2007  * Delete all the routing table entries that use the specified gateway.
2008  * XXX: this function causes search through all entries of routing table, so
2009  * it shouldn't be called when acting as a router.
2010  */
2011 void
2012 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2013 {
2014 	int s = splsoftnet();
2015 
2016 	/* We'll care only link-local addresses */
2017 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2018 		splx(s);
2019 		return;
2020 	}
2021 
2022 	rt_walktree(AF_INET6, rt6_deleteroute, (void *)gateway);
2023 	splx(s);
2024 }
2025 
2026 static int
2027 rt6_deleteroute(struct rtentry *rt, void *arg)
2028 {
2029 #define SIN6(s)	((struct sockaddr_in6 *)s)
2030 	struct in6_addr *gate = (struct in6_addr *)arg;
2031 
2032 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2033 		return (0);
2034 
2035 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr))
2036 		return (0);
2037 
2038 	/*
2039 	 * Do not delete a static route.
2040 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2041 	 * 'cloned' bit instead?
2042 	 */
2043 	if ((rt->rt_flags & RTF_STATIC) != 0)
2044 		return (0);
2045 
2046 	/*
2047 	 * We delete only host route. This means, in particular, we don't
2048 	 * delete default route.
2049 	 */
2050 	if ((rt->rt_flags & RTF_HOST) == 0)
2051 		return (0);
2052 
2053 	return (rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
2054 	    rt_mask(rt), rt->rt_flags, 0));
2055 #undef SIN6
2056 }
2057 
2058 int
2059 nd6_setdefaultiface(int ifindex)
2060 {
2061 	int error = 0;
2062 
2063 	if (ifindex < 0 || if_indexlim <= ifindex)
2064 		return (EINVAL);
2065 	if (ifindex != 0 && !ifindex2ifnet[ifindex])
2066 		return (EINVAL);
2067 
2068 	if (nd6_defifindex != ifindex) {
2069 		nd6_defifindex = ifindex;
2070 		if (nd6_defifindex > 0) {
2071 			nd6_defifp = ifindex2ifnet[nd6_defifindex];
2072 		} else
2073 			nd6_defifp = NULL;
2074 
2075 		/*
2076 		 * Our current implementation assumes one-to-one maping between
2077 		 * interfaces and links, so it would be natural to use the
2078 		 * default interface as the default link.
2079 		 */
2080 		scope6_setdefault(nd6_defifp);
2081 	}
2082 
2083 	return (error);
2084 }
2085