xref: /netbsd-src/sys/netinet6/nd6_rtr.c (revision eb961d0e02b7a46a9acfa877b02df48df6637278)
1 /*	$NetBSD: nd6_rtr.c,v 1.57 2006/03/06 20:33:52 rpaulo 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 1.57 2006/03/06 20:33:52 rpaulo 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 #define SDL(s)	((struct sockaddr_dl *)s)
66 
67 static int rtpref __P((struct nd_defrouter *));
68 static struct nd_defrouter *defrtrlist_update __P((struct nd_defrouter *));
69 static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
70     struct mbuf *, int));
71 static struct in6_ifaddr *in6_ifadd __P((struct nd_prefixctl *, int));
72 static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
73 	struct nd_defrouter *));
74 static void pfxrtr_add __P((struct nd_prefix *, struct nd_defrouter *));
75 static void pfxrtr_del __P((struct nd_pfxrouter *));
76 static struct nd_pfxrouter *find_pfxlist_reachable_router
77 	__P((struct nd_prefix *));
78 static void defrouter_delreq __P((struct nd_defrouter *));
79 static void nd6_rtmsg __P((int, struct rtentry *));
80 
81 static int in6_init_prefix_ltimes __P((struct nd_prefix *));
82 static void in6_init_address_ltimes __P((struct nd_prefix *ndpr,
83 	struct in6_addrlifetime *lt6));
84 
85 static int rt6_deleteroute __P((struct radix_node *, void *));
86 
87 extern int nd6_recalc_reachtm_interval;
88 
89 static struct ifnet *nd6_defifp;
90 int nd6_defifindex;
91 
92 int ip6_use_tempaddr = 0;
93 
94 int ip6_desync_factor;
95 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
96 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
97 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
98 
99 /* RTPREF_MEDIUM has to be 0! */
100 #define RTPREF_HIGH	1
101 #define RTPREF_MEDIUM	0
102 #define RTPREF_LOW	(-1)
103 #define RTPREF_RESERVED	(-2)
104 #define RTPREF_INVALID	(-3)	/* internal */
105 
106 /*
107  * Receive Router Solicitation Message - just for routers.
108  * Router solicitation/advertisement is mostly managed by userland program
109  * (rtadvd) so here we have no function like nd6_ra_output().
110  *
111  * Based on RFC 2461
112  */
113 void
114 nd6_rs_input(m, off, icmp6len)
115 	struct	mbuf *m;
116 	int off, icmp6len;
117 {
118 	struct ifnet *ifp = m->m_pkthdr.rcvif;
119 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
120 	struct nd_router_solicit *nd_rs;
121 	struct in6_addr saddr6 = ip6->ip6_src;
122 #if 0
123 	struct in6_addr daddr6 = ip6->ip6_dst;
124 #endif
125 	char *lladdr = NULL;
126 	int lladdrlen = 0;
127 #if 0
128 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)NULL;
129 	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)NULL;
130 	struct rtentry *rt = NULL;
131 	int is_newentry;
132 #endif
133 	union nd_opts ndopts;
134 
135 	/* If I'm not a router, ignore it. */
136 	if (ip6_accept_rtadv != 0 || !ip6_forwarding)
137 		goto freeit;
138 
139 	/* Sanity checks */
140 	if (ip6->ip6_hlim != 255) {
141 		nd6log((LOG_ERR,
142 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
143 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
144 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
145 		goto bad;
146 	}
147 
148 	/*
149 	 * Don't update the neighbor cache, if src = ::.
150 	 * This indicates that the src has no IP address assigned yet.
151 	 */
152 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
153 		goto freeit;
154 
155 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
156 	if (nd_rs == NULL) {
157 		icmp6stat.icp6s_tooshort++;
158 		return;
159 	}
160 
161 	icmp6len -= sizeof(*nd_rs);
162 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
163 	if (nd6_options(&ndopts) < 0) {
164 		nd6log((LOG_INFO,
165 		    "nd6_rs_input: invalid ND option, ignored\n"));
166 		/* nd6_options have incremented stats */
167 		goto freeit;
168 	}
169 
170 	if (ndopts.nd_opts_src_lladdr) {
171 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
172 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
173 	}
174 
175 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
176 		nd6log((LOG_INFO,
177 		    "nd6_rs_input: lladdrlen mismatch for %s "
178 		    "(if %d, RS packet %d)\n",
179 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
180 		goto bad;
181 	}
182 
183 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
184 
185  freeit:
186 	m_freem(m);
187 	return;
188 
189  bad:
190 	icmp6stat.icp6s_badrs++;
191 	m_freem(m);
192 }
193 
194 /*
195  * Receive Router Advertisement Message.
196  *
197  * Based on RFC 2461
198  * TODO: on-link bit on prefix information
199  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
200  */
201 void
202 nd6_ra_input(m, off, icmp6len)
203 	struct	mbuf *m;
204 	int off, icmp6len;
205 {
206 	struct ifnet *ifp = m->m_pkthdr.rcvif;
207 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
208 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
209 	struct nd_router_advert *nd_ra;
210 	struct in6_addr saddr6 = ip6->ip6_src;
211 #if 0
212 	struct in6_addr daddr6 = ip6->ip6_dst;
213 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
214 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
215 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
216 #endif
217 	int mcast = 0;
218 	union nd_opts ndopts;
219 	struct nd_defrouter *dr;
220 
221 	/*
222 	 * We only accept RAs only when
223 	 * the system-wide variable allows the acceptance, and
224 	 * per-interface variable allows RAs on the receiving interface.
225 	 */
226 	if (ip6_accept_rtadv == 0)
227 		goto freeit;
228 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
229 		goto freeit;
230 
231 	if (ip6->ip6_hlim != 255) {
232 		nd6log((LOG_ERR,
233 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
234 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
235 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
236 		goto bad;
237 	}
238 
239 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
240 		nd6log((LOG_ERR,
241 		    "nd6_ra_input: src %s is not link-local\n",
242 		    ip6_sprintf(&saddr6)));
243 		goto bad;
244 	}
245 
246 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
247 	if (nd_ra == NULL) {
248 		icmp6stat.icp6s_tooshort++;
249 		return;
250 	}
251 
252 	icmp6len -= sizeof(*nd_ra);
253 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
254 	if (nd6_options(&ndopts) < 0) {
255 		nd6log((LOG_INFO,
256 		    "nd6_ra_input: invalid ND option, ignored\n"));
257 		/* nd6_options have incremented stats */
258 		goto freeit;
259 	}
260 
261     {
262 	struct nd_defrouter drtr;
263 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
264 
265 	/* remember if this is a multicasted advertisement */
266 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
267 		mcast = 1;
268 
269 	memset(&drtr, 0, sizeof(drtr));
270 	drtr.rtaddr = saddr6;
271 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
272 	if (rtpref(&drtr) == RTPREF_RESERVED) {
273 		/*
274 		 * "reserved" router preference should be treated as
275 		 * 0-lifetime.  Note that rtpref() covers the case that the
276 		 * kernel is not cnofigured to support the preference
277 		 * extension.
278 		 */
279 		drtr.rtlifetime = 0;
280 	} else
281 		drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
282 	drtr.expire = time.tv_sec + drtr.rtlifetime;
283 	drtr.ifp = ifp;
284 	/* unspecified or not? (RFC 2461 6.3.4) */
285 	if (advreachable) {
286 		NTOHL(advreachable);
287 		if (advreachable <= MAX_REACHABLE_TIME &&
288 		    ndi->basereachable != advreachable) {
289 			ndi->basereachable = advreachable;
290 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
291 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
292 		}
293 	}
294 	if (nd_ra->nd_ra_retransmit)
295 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
296 	if (nd_ra->nd_ra_curhoplimit)
297 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
298 	dr = defrtrlist_update(&drtr);
299     }
300 
301 	/*
302 	 * prefix
303 	 */
304 	if (ndopts.nd_opts_pi) {
305 		struct nd_opt_hdr *pt;
306 		struct nd_opt_prefix_info *pi = NULL;
307 		struct nd_prefixctl pr;
308 
309 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
310 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
311 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
312 						(pt->nd_opt_len << 3))) {
313 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
314 				continue;
315 			pi = (struct nd_opt_prefix_info *)pt;
316 
317 			if (pi->nd_opt_pi_len != 4) {
318 				nd6log((LOG_INFO,
319 				    "nd6_ra_input: invalid option "
320 				    "len %d for prefix information option, "
321 				    "ignored\n", pi->nd_opt_pi_len));
322 				continue;
323 			}
324 
325 			if (128 < pi->nd_opt_pi_prefix_len) {
326 				nd6log((LOG_INFO,
327 				    "nd6_ra_input: invalid prefix "
328 				    "len %d for prefix information option, "
329 				    "ignored\n", pi->nd_opt_pi_prefix_len));
330 				continue;
331 			}
332 
333 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
334 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
335 				nd6log((LOG_INFO,
336 				    "nd6_ra_input: invalid prefix "
337 				    "%s, ignored\n",
338 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
339 				continue;
340 			}
341 
342 			bzero(&pr, sizeof(pr));
343 			pr.ndpr_prefix.sin6_family = AF_INET6;
344 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
345 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
346 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
347 
348 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
349 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
350 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
351 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
352 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
353 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
354 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
355 
356 			(void)prelist_update(&pr, dr, m, mcast);
357 		}
358 	}
359 
360 	/*
361 	 * MTU
362 	 */
363 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
364 		u_long mtu;
365 		u_long maxmtu;
366 
367 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
368 
369 		/* lower bound */
370 		if (mtu < IPV6_MMTU) {
371 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
372 			    "mtu=%lu sent from %s, ignoring\n",
373 			    mtu, ip6_sprintf(&ip6->ip6_src)));
374 			goto skip;
375 		}
376 
377 		/* upper bound */
378 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
379 		    ? ndi->maxmtu : ifp->if_mtu;
380 		if (mtu <= maxmtu) {
381 			int change = (ndi->linkmtu != mtu);
382 
383 			ndi->linkmtu = mtu;
384 			if (change) /* in6_maxmtu may change */
385 				in6_setmaxmtu();
386 		} else {
387 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
388 			    "mtu=%lu sent from %s; "
389 			    "exceeds maxmtu %lu, ignoring\n",
390 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
391 		}
392 	}
393 
394  skip:
395 
396 	/*
397 	 * Source link layer address
398 	 */
399     {
400 	char *lladdr = NULL;
401 	int lladdrlen = 0;
402 
403 	if (ndopts.nd_opts_src_lladdr) {
404 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
405 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
406 	}
407 
408 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
409 		nd6log((LOG_INFO,
410 		    "nd6_ra_input: lladdrlen mismatch for %s "
411 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
412 		    ifp->if_addrlen, lladdrlen - 2));
413 		goto bad;
414 	}
415 
416 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
417 
418 	/*
419 	 * Installing a link-layer address might change the state of the
420 	 * router's neighbor cache, which might also affect our on-link
421 	 * detection of adveritsed prefixes.
422 	 */
423 	pfxlist_onlink_check();
424     }
425 
426  freeit:
427 	m_freem(m);
428 	return;
429 
430  bad:
431 	icmp6stat.icp6s_badra++;
432 	m_freem(m);
433 }
434 
435 /*
436  * default router list processing sub routines
437  */
438 
439 /* tell the change to user processes watching the routing socket. */
440 static void
441 nd6_rtmsg(cmd, rt)
442 	int cmd;
443 	struct rtentry *rt;
444 {
445 	struct rt_addrinfo info;
446 
447 	bzero((caddr_t)&info, sizeof(info));
448 	info.rti_info[RTAX_DST] = rt_key(rt);
449 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
450 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
451 	if (rt->rt_ifp) {
452 		info.rti_info[RTAX_IFP] =
453 		    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
454 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
455 	}
456 
457 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
458 }
459 
460 void
461 defrouter_addreq(new)
462 	struct nd_defrouter *new;
463 {
464 	struct sockaddr_in6 def, mask, gate;
465 	struct rtentry *newrt = NULL;
466 	int s;
467 	int error;
468 
469 	memset(&def, 0, sizeof(def));
470 	memset(&mask, 0, sizeof(mask));
471 	memset(&gate, 0,sizeof(gate)); /* for safety */
472 
473 	def.sin6_len = mask.sin6_len = gate.sin6_len =
474 	    sizeof(struct sockaddr_in6);
475 	def.sin6_family = mask.sin6_family = gate.sin6_family = AF_INET6;
476 	gate.sin6_addr = new->rtaddr;
477 #ifndef SCOPEDROUTING
478 	gate.sin6_scope_id = 0;	/* XXX */
479 #endif
480 
481 	s = splsoftnet();
482 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
483 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
484 	    RTF_GATEWAY, &newrt);
485 	if (newrt) {
486 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
487 		newrt->rt_refcnt--;
488 	}
489 	if (error == 0)
490 		new->installed = 1;
491 	splx(s);
492 	return;
493 }
494 
495 struct nd_defrouter *
496 defrouter_lookup(addr, ifp)
497 	struct in6_addr *addr;
498 	struct ifnet *ifp;
499 {
500 	struct nd_defrouter *dr;
501 
502 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
503 	     dr = TAILQ_NEXT(dr, dr_entry)) {
504 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) {
505 			return (dr);
506 		}
507 	}
508 
509 	return (NULL);		/* search failed */
510 }
511 
512 void
513 defrtrlist_del(dr)
514 	struct nd_defrouter *dr;
515 {
516 	struct nd_defrouter *deldr = NULL;
517 	struct nd_prefix *pr;
518 
519 	/*
520 	 * Flush all the routing table entries that use the router
521 	 * as a next hop.
522 	 */
523 	if (!ip6_forwarding && ip6_accept_rtadv) /* XXX: better condition? */
524 		rt6_flush(&dr->rtaddr, dr->ifp);
525 
526 	if (dr->installed) {
527 		deldr = dr;
528 		defrouter_delreq(dr);
529 	}
530 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
531 
532 	/*
533 	 * Also delete all the pointers to the router in each prefix lists.
534 	 */
535 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
536 		struct nd_pfxrouter *pfxrtr;
537 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
538 			pfxrtr_del(pfxrtr);
539 	}
540 	pfxlist_onlink_check();
541 
542 	/*
543 	 * If the router is the primary one, choose a new one.
544 	 * Note that defrouter_select() will remove the current gateway
545 	 * from the routing table.
546 	 */
547 	if (deldr)
548 		defrouter_select();
549 
550 	free(dr, M_IP6NDP);
551 }
552 
553 /*
554  * Remove the default route for a given router.
555  * This is just a subroutine function for defrouter_select(), and should
556  * not be called from anywhere else.
557  */
558 static void
559 defrouter_delreq(dr)
560 	struct nd_defrouter *dr;
561 {
562 	struct sockaddr_in6 def, mask, gw;
563 	struct rtentry *oldrt = NULL;
564 
565 #ifdef DIAGNOSTIC
566 	if (dr == NULL)
567 		panic("dr == NULL in defrouter_delreq");
568 #endif
569 
570 	bzero(&def, sizeof(def));
571 	bzero(&mask, sizeof(mask));
572 	bzero(&gw, sizeof(gw));	/* for safety */
573 
574 	def.sin6_len = mask.sin6_len = gw.sin6_len =
575 	    sizeof(struct sockaddr_in6);
576 	def.sin6_family = mask.sin6_family = gw.sin6_family = AF_INET6;
577 	gw.sin6_addr = dr->rtaddr;
578 #ifndef SCOPEDROUTING
579 	gw.sin6_scope_id = 0;	/* XXX */
580 #endif
581 
582 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
583 	    (struct sockaddr *)&gw,
584 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
585 	if (oldrt) {
586 		nd6_rtmsg(RTM_DELETE, oldrt);
587 		if (oldrt->rt_refcnt <= 0) {
588 			/*
589 			 * XXX: borrowed from the RTM_DELETE case of
590 			 * rtrequest().
591 			 */
592 			oldrt->rt_refcnt++;
593 			rtfree(oldrt);
594 		}
595 	}
596 
597 	dr->installed = 0;
598 }
599 
600 /*
601  * remove all default routes from default router list
602  */
603 void
604 defrouter_reset()
605 {
606 	struct nd_defrouter *dr;
607 
608 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
609 	     dr = TAILQ_NEXT(dr, dr_entry))
610 		defrouter_delreq(dr);
611 
612 	/*
613 	 * XXX should we also nuke any default routers in the kernel, by
614 	 * going through them by rtalloc1()?
615 	 */
616 }
617 
618 /*
619  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
620  * draft-ietf-ipngwg-router-selection:
621  * 1) Routers that are reachable or probably reachable should be preferred.
622  *    If we have more than one (probably) reachable router, prefer ones
623  *    with the highest router preference.
624  * 2) When no routers on the list are known to be reachable or
625  *    probably reachable, routers SHOULD be selected in a round-robin
626  *    fashion, regardless of router preference values.
627  * 3) If the Default Router List is empty, assume that all
628  *    destinations are on-link.
629  *
630  * We assume nd_defrouter is sorted by router preference value.
631  * Since the code below covers both with and without router preference cases,
632  * we do not need to classify the cases by ifdef.
633  *
634  * At this moment, we do not try to install more than one default router,
635  * even when the multipath routing is available, because we're not sure about
636  * the benefits for stub hosts comparing to the risk of making the code
637  * complicated and the possibility of introducing bugs.
638  */
639 void
640 defrouter_select()
641 {
642 	int s = splsoftnet();
643 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
644 	struct rtentry *rt = NULL;
645 	struct llinfo_nd6 *ln = NULL;
646 
647 	/*
648 	 * This function should be called only when acting as an autoconfigured
649 	 * host.  Although the remaining part of this function is not effective
650 	 * if the node is not an autoconfigured host, we explicitly exclude
651 	 * such cases here for safety.
652 	 */
653 	if (ip6_forwarding || !ip6_accept_rtadv) {
654 		nd6log((LOG_WARNING,
655 		    "defrouter_select: called unexpectedly (forwarding=%d, "
656 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
657 		splx(s);
658 		return;
659 	}
660 
661 	/*
662 	 * Let's handle easy case (3) first:
663 	 * If default router list is empty, there's nothing to be done.
664 	 */
665 	if (!TAILQ_FIRST(&nd_defrouter)) {
666 		splx(s);
667 		return;
668 	}
669 
670 	/*
671 	 * Search for a (probably) reachable router from the list.
672 	 * We just pick up the first reachable one (if any), assuming that
673 	 * the ordering rule of the list described in defrtrlist_update().
674 	 */
675 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
676 	     dr = TAILQ_NEXT(dr, dr_entry)) {
677 		if (selected_dr == NULL &&
678 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) != NULL &&
679 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) != NULL &&
680 		    ND6_IS_LLINFO_PROBREACH(ln)) {
681 			selected_dr = dr;
682 		}
683 
684 		if (dr->installed && !installed_dr)
685 			installed_dr = dr;
686 		else if (dr->installed && installed_dr) {
687 			/* this should not happen.  warn for diagnosis. */
688 			log(LOG_ERR, "defrouter_select: more than one router"
689 			    " is installed\n");
690 		}
691 	}
692 	/*
693 	 * If none of the default routers was found to be reachable,
694 	 * round-robin the list regardless of preference.
695 	 * Otherwise, if we have an installed router, check if the selected
696 	 * (reachable) router should really be preferred to the installed one.
697 	 * We only prefer the new router when the old one is not reachable
698 	 * or when the new one has a really higher preference value.
699 	 */
700 	if (selected_dr == NULL) {
701 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
702 			selected_dr = TAILQ_FIRST(&nd_defrouter);
703 		else
704 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
705 	} else if (installed_dr &&
706 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
707 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
708 	    ND6_IS_LLINFO_PROBREACH(ln) &&
709 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
710 		selected_dr = installed_dr;
711 	}
712 
713 	/*
714 	 * If the selected router is different than the installed one,
715 	 * remove the installed router and install the selected one.
716 	 * Note that the selected router is never NULL here.
717 	 */
718 	if (installed_dr != selected_dr) {
719 		if (installed_dr)
720 			defrouter_delreq(installed_dr);
721 		defrouter_addreq(selected_dr);
722 	}
723 
724 	splx(s);
725 	return;
726 }
727 
728 /*
729  * for default router selection
730  * regards router-preference field as a 2-bit signed integer
731  */
732 static int
733 rtpref(struct nd_defrouter *dr)
734 {
735 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
736 	case ND_RA_FLAG_RTPREF_HIGH:
737 		return (RTPREF_HIGH);
738 	case ND_RA_FLAG_RTPREF_MEDIUM:
739 		return (RTPREF_MEDIUM);
740 	case ND_RA_FLAG_RTPREF_RSV:
741 		return (RTPREF_RESERVED);
742 	case ND_RA_FLAG_RTPREF_LOW:
743 		return (RTPREF_LOW);
744 	default:
745 		/*
746 		 * This case should never happen.  If it did, it would mean a
747 		 * serious bug of kernel internal.  We thus always bark here.
748 		 * Or, can we even panic?
749 		 */
750 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
751 		return (RTPREF_INVALID);
752 	}
753 	/* NOTREACHED */
754 }
755 
756 static struct nd_defrouter *
757 defrtrlist_update(new)
758 	struct nd_defrouter *new;
759 {
760 	struct nd_defrouter *dr, *n;
761 	int s = splsoftnet();
762 
763 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
764 		/* entry exists */
765 		if (new->rtlifetime == 0) {
766 			defrtrlist_del(dr);
767 			dr = NULL;
768 		} else {
769 			int oldpref = rtpref(dr);
770 
771 			/* override */
772 			dr->flags = new->flags; /* xxx flag check */
773 			dr->rtlifetime = new->rtlifetime;
774 			dr->expire = new->expire;
775 
776 			/*
777 			 * If the preference does not change, there's no need
778 			 * to sort the entries.
779 			 */
780 			if (rtpref(new) == oldpref) {
781 				splx(s);
782 				return (dr);
783 			}
784 
785 			/*
786 			 * preferred router may be changed, so relocate
787 			 * this router.
788 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
789 			 * However, since defrtrlist_del() has many side
790 			 * effects, we intentionally do so here.
791 			 * defrouter_select() below will handle routing
792 			 * changes later.
793 			 */
794 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
795 			n = dr;
796 			goto insert;
797 		}
798 		splx(s);
799 		return (dr);
800 	}
801 
802 	/* entry does not exist */
803 	if (new->rtlifetime == 0) {
804 		splx(s);
805 		return (NULL);
806 	}
807 
808 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
809 	if (n == NULL) {
810 		splx(s);
811 		return (NULL);
812 	}
813 	bzero(n, sizeof(*n));
814 	*n = *new;
815 
816 insert:
817 	/*
818 	 * Insert the new router in the Default Router List;
819 	 * The Default Router List should be in the descending order
820 	 * of router-preferece.  Routers with the same preference are
821 	 * sorted in the arriving time order.
822 	 */
823 
824 	/* insert at the end of the group */
825 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
826 	     dr = TAILQ_NEXT(dr, dr_entry)) {
827 		if (rtpref(n) > rtpref(dr))
828 			break;
829 	}
830 	if (dr)
831 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
832 	else
833 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
834 
835 	defrouter_select();
836 
837 	splx(s);
838 
839 	return (n);
840 }
841 
842 static struct nd_pfxrouter *
843 pfxrtr_lookup(pr, dr)
844 	struct nd_prefix *pr;
845 	struct nd_defrouter *dr;
846 {
847 	struct nd_pfxrouter *search;
848 
849 	for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
850 		if (search->router == dr)
851 			break;
852 	}
853 
854 	return (search);
855 }
856 
857 static void
858 pfxrtr_add(pr, dr)
859 	struct nd_prefix *pr;
860 	struct nd_defrouter *dr;
861 {
862 	struct nd_pfxrouter *new;
863 
864 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
865 	if (new == NULL)
866 		return;
867 	bzero(new, sizeof(*new));
868 	new->router = dr;
869 
870 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
871 
872 	pfxlist_onlink_check();
873 }
874 
875 static void
876 pfxrtr_del(pfr)
877 	struct nd_pfxrouter *pfr;
878 {
879 	LIST_REMOVE(pfr, pfr_entry);
880 	free(pfr, M_IP6NDP);
881 }
882 
883 struct nd_prefix *
884 nd6_prefix_lookup(key)
885 	struct nd_prefixctl *key;
886 {
887 	struct nd_prefix *search;
888 
889 	for (search = nd_prefix.lh_first; search; search = search->ndpr_next) {
890 		if (key->ndpr_ifp == search->ndpr_ifp &&
891 		    key->ndpr_plen == search->ndpr_plen &&
892 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
893 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
894 			break;
895 		}
896 	}
897 
898 	return (search);
899 }
900 
901 int
902 nd6_prelist_add(pr, dr, newp)
903 	struct nd_prefixctl *pr;
904 	struct nd_prefix **newp;
905 	struct nd_defrouter *dr;
906 {
907 	struct nd_prefix *new = NULL;
908 	int i, s;
909 	int error;
910 
911 	error = 0;
912 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
913 	if (new == NULL)
914 		return ENOMEM;
915 	bzero(new, sizeof(*new));
916 	new->ndpr_ifp = pr->ndpr_ifp;
917 	new->ndpr_prefix = pr->ndpr_prefix;
918 	new->ndpr_plen = pr->ndpr_plen;
919 	new->ndpr_vltime = pr->ndpr_vltime;
920 	new->ndpr_pltime = pr->ndpr_pltime;
921 	new->ndpr_flags = pr->ndpr_flags;
922 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
923 		free(new, M_IP6NDP);
924 		return(error);
925 	}
926 	new->ndpr_lastupdate = time.tv_sec;
927 	if (newp != NULL)
928 		*newp = new;
929 
930 	/* initialization */
931 	LIST_INIT(&new->ndpr_advrtrs);
932 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
933 	/* make prefix in the canonical form */
934 	for (i = 0; i < 4; i++)
935 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
936 		    new->ndpr_mask.s6_addr32[i];
937 
938 	s = splsoftnet();
939 	/* link ndpr_entry to nd_prefix list */
940 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
941 	splx(s);
942 
943 	/* ND_OPT_PI_FLAG_ONLINK processing */
944 	if (new->ndpr_raf_onlink) {
945 		int e;
946 
947 		if ((e = nd6_prefix_onlink(new)) != 0) {
948 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
949 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
950 			    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
951 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
952 			/* proceed anyway. XXX: is it correct? */
953 		}
954 	}
955 
956 	if (dr)
957 		pfxrtr_add(new, dr);
958 
959 	return 0;
960 }
961 
962 void
963 prelist_remove(pr)
964 	struct nd_prefix *pr;
965 {
966 	struct nd_pfxrouter *pfr, *next;
967 	int e, s;
968 
969 	/* make sure to invalidate the prefix until it is really freed. */
970 	pr->ndpr_vltime = 0;
971 	pr->ndpr_pltime = 0;
972 #if 0
973 	/*
974 	 * Though these flags are now meaningless, we'd rather keep the value
975 	 * not to confuse users when executing "ndp -p".
976 	 */
977 	pr->ndpr_raf_onlink = 0;
978 	pr->ndpr_raf_auto = 0;
979 #endif
980 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
981 	    (e = nd6_prefix_offlink(pr)) != 0) {
982 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
983 		    "on %s, errno=%d\n",
984 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
985 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
986 		/* what should we do? */
987 	}
988 
989 	if (pr->ndpr_refcnt > 0)
990 		return;		/* notice here? */
991 
992 	s = splsoftnet();
993 	/* unlink ndpr_entry from nd_prefix list */
994 	LIST_REMOVE(pr, ndpr_entry);
995 
996 	/* free list of routers that adversed the prefix */
997 	for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
998 		next = pfr->pfr_next;
999 
1000 		free(pfr, M_IP6NDP);
1001 	}
1002 	splx(s);
1003 
1004 	free(pr, M_IP6NDP);
1005 
1006 	pfxlist_onlink_check();
1007 }
1008 
1009 static int
1010 prelist_update(new, dr, m, mcast)
1011 	struct nd_prefixctl *new;
1012 	struct nd_defrouter *dr; /* may be NULL */
1013 	struct mbuf *m;
1014 	int mcast;
1015 {
1016 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
1017 	struct ifaddr *ifa;
1018 	struct ifnet *ifp = new->ndpr_ifp;
1019 	struct nd_prefix *pr;
1020 	int s = splsoftnet();
1021 	int error = 0;
1022 	int newprefix = 0;
1023 	int auth;
1024 	struct in6_addrlifetime lt6_tmp;
1025 
1026 	auth = 0;
1027 	if (m) {
1028 		/*
1029 		 * Authenticity for NA consists authentication for
1030 		 * both IP header and IP datagrams, doesn't it ?
1031 		 */
1032 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1033 		auth = (m->m_flags & M_AUTHIPHDR
1034 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
1035 #endif
1036 	}
1037 
1038 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1039 		/*
1040 		 * nd6_prefix_lookup() ensures that pr and new have the same
1041 		 * prefix on a same interface.
1042 		 */
1043 
1044 		/*
1045 		 * Update prefix information.  Note that the on-link (L) bit
1046 		 * and the autonomous (A) bit should NOT be changed from 1
1047 		 * to 0.
1048 		 */
1049 		if (new->ndpr_raf_onlink == 1)
1050 			pr->ndpr_raf_onlink = 1;
1051 		if (new->ndpr_raf_auto == 1)
1052 			pr->ndpr_raf_auto = 1;
1053 		if (new->ndpr_raf_onlink) {
1054 			pr->ndpr_vltime = new->ndpr_vltime;
1055 			pr->ndpr_pltime = new->ndpr_pltime;
1056 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1057 			pr->ndpr_lastupdate = time.tv_sec;
1058 		}
1059 
1060 		if (new->ndpr_raf_onlink &&
1061 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1062 			int e;
1063 
1064 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1065 				nd6log((LOG_ERR,
1066 				    "prelist_update: failed to make "
1067 				    "the prefix %s/%d on-link on %s "
1068 				    "(errno=%d)\n",
1069 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1070 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1071 				/* proceed anyway. XXX: is it correct? */
1072 			}
1073 		}
1074 
1075 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1076 			pfxrtr_add(pr, dr);
1077 	} else {
1078 		struct nd_prefix *newpr = NULL;
1079 
1080 		newprefix = 1;
1081 
1082 		if (new->ndpr_vltime == 0)
1083 			goto end;
1084 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1085 			goto end;
1086 
1087 		error = nd6_prelist_add(new, dr, &newpr);
1088 		if (error != 0 || newpr == NULL) {
1089 			nd6log((LOG_NOTICE, "prelist_update: "
1090 			    "nd6_prelist_add failed for %s/%d on %s "
1091 			    "errno=%d, returnpr=%p\n",
1092 			    ip6_sprintf(&new->ndpr_prefix.sin6_addr),
1093 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1094 			    error, newpr));
1095 			goto end; /* we should just give up in this case. */
1096 		}
1097 
1098 		/*
1099 		 * XXX: from the ND point of view, we can ignore a prefix
1100 		 * with the on-link bit being zero.  However, we need a
1101 		 * prefix structure for references from autoconfigured
1102 		 * addresses.  Thus, we explicitly make sure that the prefix
1103 		 * itself expires now.
1104 		 */
1105 		if (newpr->ndpr_raf_onlink == 0) {
1106 			newpr->ndpr_vltime = 0;
1107 			newpr->ndpr_pltime = 0;
1108 			in6_init_prefix_ltimes(newpr);
1109 		}
1110 
1111 		pr = newpr;
1112 	}
1113 
1114 	/*
1115 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1116 	 * Note that pr must be non NULL at this point.
1117 	 */
1118 
1119 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1120 	if (!new->ndpr_raf_auto)
1121 		goto end;
1122 
1123 	/*
1124 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1125 	 * nd6_ra_input.
1126 	 */
1127 
1128 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1129 	if (new->ndpr_pltime > new->ndpr_vltime) {
1130 		error = EINVAL;	/* XXX: won't be used */
1131 		goto end;
1132 	}
1133 
1134 	/*
1135 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1136 	 * an address configured by stateless autoconfiguration already in the
1137 	 * list of addresses associated with the interface, and the Valid
1138 	 * Lifetime is not 0, form an address.  We first check if we have
1139 	 * a matching prefix.
1140 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1141 	 * consider autoconfigured addresses while RFC2462 simply said
1142 	 * "address".
1143 	 */
1144 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
1145 	{
1146 		struct in6_ifaddr *ifa6;
1147 		u_int32_t remaininglifetime;
1148 		long time_second = time.tv_sec;
1149 
1150 		if (ifa->ifa_addr->sa_family != AF_INET6)
1151 			continue;
1152 
1153 		ifa6 = (struct in6_ifaddr *)ifa;
1154 
1155 		/*
1156 		 * We only consider autoconfigured addresses as per rfc2462bis.
1157 		 */
1158 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1159 			continue;
1160 
1161 		/*
1162 		 * Spec is not clear here, but I believe we should concentrate
1163 		 * on unicast (i.e. not anycast) addresses.
1164 		 * XXX: other ia6_flags? detached or duplicated?
1165 		 */
1166 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1167 			continue;
1168 
1169 		/*
1170 		 * Ignore the address if it is not associated with a prefix
1171 		 * or is associated with a prefix that is different from this
1172 		 * one.  (pr is never NULL here)
1173 		 */
1174 		if (ifa6->ia6_ndpr != pr)
1175 			continue;
1176 
1177 		if (ia6_match == NULL) /* remember the first one */
1178 			ia6_match = ifa6;
1179 
1180 		/*
1181 		 * An already autoconfigured address matched.  Now that we
1182 		 * are sure there is at least one matched address, we can
1183 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1184 		 * "two hours" rule and the privacy extension.
1185 		 * We apply some clarifications in rfc2462bis:
1186 		 * - use remaininglifetime instead of storedlifetime as a
1187 		 *   variable name
1188 		 * - remove the dead code in the "two-hour" rule
1189 		 */
1190 #define TWOHOUR		(120*60)
1191 		lt6_tmp = ifa6->ia6_lifetime;
1192 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1193 			remaininglifetime = ND6_INFINITE_LIFETIME;
1194 		else if (time_second - ifa6->ia6_updatetime >
1195 			 lt6_tmp.ia6t_vltime) {
1196 			/*
1197 			 * The case of "invalid" address.  We should usually
1198 			 * not see this case.
1199 			 */
1200 			remaininglifetime = 0;
1201 		} else
1202 			remaininglifetime = lt6_tmp.ia6t_vltime -
1203 			    (time_second - ifa6->ia6_updatetime);
1204 
1205 		/* when not updating, keep the current stored lifetime. */
1206 		lt6_tmp.ia6t_vltime = remaininglifetime;
1207 
1208 		if (TWOHOUR < new->ndpr_vltime ||
1209 		    remaininglifetime < new->ndpr_vltime) {
1210 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1211 		} else if (remaininglifetime <= TWOHOUR) {
1212 			if (auth)
1213 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1214 		} else {
1215 			/*
1216 			 * new->ndpr_vltime <= TWOHOUR &&
1217 			 * TWOHOUR < remaininglifetime
1218 			 */
1219 			lt6_tmp.ia6t_vltime = TWOHOUR;
1220 		}
1221 
1222 		/* The 2 hour rule is not imposed for preferred lifetime. */
1223 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1224 
1225 		in6_init_address_ltimes(pr, &lt6_tmp);
1226 
1227 		/*
1228 		 * We need to treat lifetimes for temporary addresses
1229 		 * differently, according to
1230 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1231 		 * we only update the lifetimes when they are in the maximum
1232 		 * intervals.
1233 		 */
1234 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1235 			u_int32_t maxvltime, maxpltime;
1236 
1237 			if (ip6_temp_valid_lifetime >
1238 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1239 			    ip6_desync_factor)) {
1240 				maxvltime = ip6_temp_valid_lifetime -
1241 				    (time_second - ifa6->ia6_createtime) -
1242 				    ip6_desync_factor;
1243 			} else
1244 				maxvltime = 0;
1245 			if (ip6_temp_preferred_lifetime >
1246 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1247 			    ip6_desync_factor)) {
1248 				maxpltime = ip6_temp_preferred_lifetime -
1249 				    (time_second - ifa6->ia6_createtime) -
1250 				    ip6_desync_factor;
1251 			} else
1252 				maxpltime = 0;
1253 
1254 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1255 			    lt6_tmp.ia6t_vltime > maxvltime) {
1256 				lt6_tmp.ia6t_vltime = maxvltime;
1257 			}
1258 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1259 			    lt6_tmp.ia6t_pltime > maxpltime) {
1260 				lt6_tmp.ia6t_pltime = maxpltime;
1261 			}
1262 		}
1263 
1264 		ifa6->ia6_lifetime = lt6_tmp;
1265 		ifa6->ia6_updatetime = time.tv_sec;
1266 	}
1267 	if (ia6_match == NULL && new->ndpr_vltime) {
1268 		int ifidlen;
1269 
1270 		/*
1271 		 * 5.5.3 (d) (continued)
1272 		 * No address matched and the valid lifetime is non-zero.
1273 		 * Create a new address.
1274 		 */
1275 
1276 		/*
1277 		 * Prefix Length check:
1278 		 * If the sum of the prefix length and interface identifier
1279 		 * length does not equal 128 bits, the Prefix Information
1280 		 * option MUST be ignored.  The length of the interface
1281 		 * identifier is defined in a separate link-type specific
1282 		 * document.
1283 		 */
1284 		ifidlen = in6_if2idlen(ifp);
1285 		if (ifidlen < 0) {
1286 			/* this should not happen, so we always log it. */
1287 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1288 			    if_name(ifp));
1289 			goto end;
1290 		}
1291 		if (ifidlen + pr->ndpr_plen != 128) {
1292 			nd6log((LOG_INFO,
1293 			    "prelist_update: invalid prefixlen "
1294 			    "%d for %s, ignored\n",
1295 			    pr->ndpr_plen, if_name(ifp)));
1296 			goto end;
1297 		}
1298 
1299 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1300 			/*
1301 			 * note that we should use pr (not new) for reference.
1302 			 */
1303 			pr->ndpr_refcnt++;
1304 			ia6->ia6_ndpr = pr;
1305 
1306 			/*
1307 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
1308 			 * When a new public address is created as described
1309 			 * in RFC2462, also create a new temporary address.
1310 			 *
1311 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
1312 			 * When an interface connects to a new link, a new
1313 			 * randomized interface identifier should be generated
1314 			 * immediately together with a new set of temporary
1315 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1316 			 * in6_tmpifadd().
1317 			 */
1318 			if (ip6_use_tempaddr) {
1319 				int e;
1320 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1321 					nd6log((LOG_NOTICE, "prelist_update: "
1322 					    "failed to create a temporary "
1323 					    "address, errno=%d\n",
1324 					    e));
1325 				}
1326 			}
1327 
1328 			/*
1329 			 * A newly added address might affect the status
1330 			 * of other addresses, so we check and update it.
1331 			 * XXX: what if address duplication happens?
1332 			 */
1333 			pfxlist_onlink_check();
1334 		} else {
1335 			/* just set an error. do not bark here. */
1336 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1337 		}
1338 	}
1339 
1340  end:
1341 	splx(s);
1342 	return error;
1343 }
1344 
1345 /*
1346  * A supplement function used in the on-link detection below;
1347  * detect if a given prefix has a (probably) reachable advertising router.
1348  * XXX: lengthy function name...
1349  */
1350 static struct nd_pfxrouter *
1351 find_pfxlist_reachable_router(pr)
1352 	struct nd_prefix *pr;
1353 {
1354 	struct nd_pfxrouter *pfxrtr;
1355 	struct rtentry *rt;
1356 	struct llinfo_nd6 *ln;
1357 
1358 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1359 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1360 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1361 		    pfxrtr->router->ifp)) &&
1362 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1363 		    ND6_IS_LLINFO_PROBREACH(ln))
1364 			break;	/* found */
1365 	}
1366 
1367 	return (pfxrtr);
1368 }
1369 
1370 /*
1371  * Check if each prefix in the prefix list has at least one available router
1372  * that advertised the prefix (a router is "available" if its neighbor cache
1373  * entry is reachable or probably reachable).
1374  * If the check fails, the prefix may be off-link, because, for example,
1375  * we have moved from the network but the lifetime of the prefix has not
1376  * expired yet.  So we should not use the prefix if there is another prefix
1377  * that has an available router.
1378  * But, if there is no prefix that has an available router, we still regards
1379  * all the prefixes as on-link.  This is because we can't tell if all the
1380  * routers are simply dead or if we really moved from the network and there
1381  * is no router around us.
1382  */
1383 void
1384 pfxlist_onlink_check()
1385 {
1386 	struct nd_prefix *pr;
1387 	struct in6_ifaddr *ifa;
1388 	struct nd_defrouter *dr;
1389 	struct nd_pfxrouter *pfxrtr = NULL;
1390 
1391 	/*
1392 	 * Check if there is a prefix that has a reachable advertising
1393 	 * router.
1394 	 */
1395 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1396 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1397 			break;
1398 	}
1399 	/*
1400 	 * If we have no such prefix, check whether we still have a router
1401 	 * that does not advertise any prefixes.
1402 	 */
1403 	if (pr == NULL) {
1404 		for (dr = TAILQ_FIRST(&nd_defrouter); dr;
1405 		    dr = TAILQ_NEXT(dr, dr_entry)) {
1406 			struct nd_prefix *pr0;
1407 
1408 			for (pr0 = nd_prefix.lh_first; pr0;
1409 			    pr0 = pr0->ndpr_next) {
1410 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1411 					break;
1412 			}
1413 			if (pfxrtr)
1414 				break;
1415 		}
1416 	}
1417 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
1418 		/*
1419 		 * There is at least one prefix that has a reachable router,
1420 		 * or at least a router which probably does not advertise
1421 		 * any prefixes.  The latter would be the case when we move
1422 		 * to a new link where we have a router that does not provide
1423 		 * prefixes and we configure an address by hand.
1424 		 * Detach prefixes which have no reachable advertising
1425 		 * router, and attach other prefixes.
1426 		 */
1427 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1428 			/* XXX: a link-local prefix should never be detached */
1429 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1430 				continue;
1431 
1432 			/*
1433 			 * we aren't interested in prefixes without the L bit
1434 			 * set.
1435 			 */
1436 			if (pr->ndpr_raf_onlink == 0)
1437 				continue;
1438 
1439 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1440 			    find_pfxlist_reachable_router(pr) == NULL)
1441 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1442 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1443 			    find_pfxlist_reachable_router(pr) != 0)
1444 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1445 		}
1446 	} else {
1447 		/* there is no prefix that has a reachable router */
1448 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1449 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1450 				continue;
1451 
1452 			if (pr->ndpr_raf_onlink == 0)
1453 				continue;
1454 
1455 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1456 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1457 		}
1458 	}
1459 
1460 	/*
1461 	 * Remove each interface route associated with a (just) detached
1462 	 * prefix, and reinstall the interface route for a (just) attached
1463 	 * prefix.  Note that all attempt of reinstallation does not
1464 	 * necessarily success, when a same prefix is shared among multiple
1465 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1466 	 * so we don't have to care about them.
1467 	 */
1468 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1469 		int e;
1470 
1471 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1472 			continue;
1473 
1474 		if (pr->ndpr_raf_onlink == 0)
1475 			continue;
1476 
1477 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1478 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1479 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1480 				nd6log((LOG_ERR,
1481 				    "pfxlist_onlink_check: failed to "
1482 				    "make %s/%d offlink, errno=%d\n",
1483 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1484 				    pr->ndpr_plen, e));
1485 			}
1486 		}
1487 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1488 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1489 		    pr->ndpr_raf_onlink) {
1490 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1491 				nd6log((LOG_ERR,
1492 				    "pfxlist_onlink_check: failed to "
1493 				    "make %s/%d onlink, errno=%d\n",
1494 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1495 				    pr->ndpr_plen, e));
1496 			}
1497 		}
1498 	}
1499 
1500 	/*
1501 	 * Changes on the prefix status might affect address status as well.
1502 	 * Make sure that all addresses derived from an attached prefix are
1503 	 * attached, and that all addresses derived from a detached prefix are
1504 	 * detached.  Note, however, that a manually configured address should
1505 	 * always be attached.
1506 	 * The precise detection logic is same as the one for prefixes.
1507 	 */
1508 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1509 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1510 			continue;
1511 
1512 		if (ifa->ia6_ndpr == NULL) {
1513 			/*
1514 			 * This can happen when we first configure the address
1515 			 * (i.e. the address exists, but the prefix does not).
1516 			 * XXX: complicated relationships...
1517 			 */
1518 			continue;
1519 		}
1520 
1521 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1522 			break;
1523 	}
1524 	if (ifa) {
1525 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1526 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1527 				continue;
1528 
1529 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1530 				continue;
1531 
1532 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1533 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1534 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1535 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1536 					nd6_dad_start((struct ifaddr *)ifa,
1537 					    0);
1538 				}
1539 			} else {
1540 				if ((ifa->ia6_flags & IN6_IFF_DETACHED) == 0) {
1541 					ifa->ia6_flags |= IN6_IFF_DETACHED;
1542 				}
1543 			}
1544 		}
1545 	}
1546 	else {
1547 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1548 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1549 				continue;
1550 
1551 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1552 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1553 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1554 				/* Do we need a delay in this case? */
1555 				nd6_dad_start((struct ifaddr *)ifa, 0);
1556 			}
1557 		}
1558 	}
1559 }
1560 
1561 int
1562 nd6_prefix_onlink(pr)
1563 	struct nd_prefix *pr;
1564 {
1565 	struct ifaddr *ifa;
1566 	struct ifnet *ifp = pr->ndpr_ifp;
1567 	struct sockaddr_in6 mask6;
1568 	struct nd_prefix *opr;
1569 	u_long rtflags;
1570 	int error = 0;
1571 	struct rtentry *rt = NULL;
1572 
1573 	/* sanity check */
1574 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1575 		nd6log((LOG_ERR,
1576 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1577 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1578 		return (EEXIST);
1579 	}
1580 
1581 	/*
1582 	 * Add the interface route associated with the prefix.  Before
1583 	 * installing the route, check if there's the same prefix on another
1584 	 * interface, and the prefix has already installed the interface route.
1585 	 * Although such a configuration is expected to be rare, we explicitly
1586 	 * allow it.
1587 	 */
1588 	for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1589 		if (opr == pr)
1590 			continue;
1591 
1592 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1593 			continue;
1594 
1595 		if (opr->ndpr_plen == pr->ndpr_plen &&
1596 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1597 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1598 			return (0);
1599 	}
1600 
1601 	/*
1602 	 * We prefer link-local addresses as the associated interface address.
1603 	 */
1604 	/* search for a link-local addr */
1605 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1606 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1607 	if (ifa == NULL) {
1608 		/* XXX: freebsd does not have ifa_ifwithaf */
1609 		for (ifa = ifp->if_addrlist.tqh_first;
1610 		     ifa;
1611 		     ifa = ifa->ifa_list.tqe_next)
1612 		{
1613 			if (ifa->ifa_addr->sa_family == AF_INET6)
1614 				break;
1615 		}
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(&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 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
1641 	rtflags = ifa->ifa_flags | RTF_UP;
1642 	if (nd6_need_cache(ifp)) {
1643 		/* explicitly set in case ifa_flags does not set the flag. */
1644 		rtflags |= RTF_CLONING;
1645 	} else {
1646 		/*
1647 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1648 		 */
1649 		rtflags &= ~RTF_CLONING;
1650 	}
1651 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1652 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1653 	if (error == 0) {
1654 		if (rt != NULL) /* this should be non NULL, though */
1655 			nd6_rtmsg(RTM_ADD, rt);
1656 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1657 	} else {
1658 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1659 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1660 		    "errno = %d\n",
1661 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1662 		    pr->ndpr_plen, if_name(ifp),
1663 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1664 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
1665 	}
1666 
1667 	if (rt != NULL)
1668 		rt->rt_refcnt--;
1669 
1670 	return (error);
1671 }
1672 
1673 int
1674 nd6_prefix_offlink(pr)
1675 	struct nd_prefix *pr;
1676 {
1677 	int error = 0;
1678 	struct ifnet *ifp = pr->ndpr_ifp;
1679 	struct nd_prefix *opr;
1680 	struct sockaddr_in6 sa6, mask6;
1681 	struct rtentry *rt = NULL;
1682 
1683 	/* sanity check */
1684 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1685 		nd6log((LOG_ERR,
1686 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1687 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1688 		return (EEXIST);
1689 	}
1690 
1691 	bzero(&sa6, sizeof(sa6));
1692 	sa6.sin6_family = AF_INET6;
1693 	sa6.sin6_len = sizeof(sa6);
1694 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1695 	    sizeof(struct in6_addr));
1696 	bzero(&mask6, sizeof(mask6));
1697 	mask6.sin6_family = AF_INET6;
1698 	mask6.sin6_len = sizeof(sa6);
1699 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1700 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1701 	    (struct sockaddr *)&mask6, 0, &rt);
1702 	if (error == 0) {
1703 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1704 
1705 		/* report the route deletion to the routing socket. */
1706 		if (rt != NULL)
1707 			nd6_rtmsg(RTM_DELETE, rt);
1708 
1709 		/*
1710 		 * There might be the same prefix on another interface,
1711 		 * the prefix which could not be on-link just because we have
1712 		 * the interface route (see comments in nd6_prefix_onlink).
1713 		 * If there's one, try to make the prefix on-link on the
1714 		 * interface.
1715 		 */
1716 		for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1717 			if (opr == pr)
1718 				continue;
1719 
1720 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1721 				continue;
1722 
1723 			/*
1724 			 * KAME specific: detached prefixes should not be
1725 			 * on-link.
1726 			 */
1727 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1728 				continue;
1729 
1730 			if (opr->ndpr_plen == pr->ndpr_plen &&
1731 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1732 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1733 				int e;
1734 
1735 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1736 					nd6log((LOG_ERR,
1737 					    "nd6_prefix_offlink: failed to "
1738 					    "recover a prefix %s/%d from %s "
1739 					    "to %s (errno = %d)\n",
1740 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1741 					    opr->ndpr_plen, if_name(ifp),
1742 					    if_name(opr->ndpr_ifp), e));
1743 				}
1744 			}
1745 		}
1746 	} else {
1747 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1748 		nd6log((LOG_ERR,
1749 		    "nd6_prefix_offlink: failed to delete route: "
1750 		    "%s/%d on %s (errno = %d)\n",
1751 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1752 		    error));
1753 	}
1754 
1755 	if (rt != NULL) {
1756 		if (rt->rt_refcnt <= 0) {
1757 			/* XXX: we should free the entry ourselves. */
1758 			rt->rt_refcnt++;
1759 			rtfree(rt);
1760 		}
1761 	}
1762 
1763 	return (error);
1764 }
1765 
1766 static struct in6_ifaddr *
1767 in6_ifadd(pr, mcast)
1768 	struct nd_prefixctl *pr;
1769 	int mcast;
1770 {
1771 	struct ifnet *ifp = pr->ndpr_ifp;
1772 	struct ifaddr *ifa;
1773 	struct in6_aliasreq ifra;
1774 	struct in6_ifaddr *ia, *ib;
1775 	int error, plen0;
1776 	struct in6_addr mask;
1777 	int prefixlen = pr->ndpr_plen;
1778 	int updateflags;
1779 
1780 	in6_prefixlen2mask(&mask, prefixlen);
1781 
1782 	/*
1783 	 * find a link-local address (will be interface ID).
1784 	 * Is it really mandatory? Theoretically, a global or a site-local
1785 	 * address can be configured without a link-local address, if we
1786 	 * have a unique interface identifier...
1787 	 *
1788 	 * it is not mandatory to have a link-local address, we can generate
1789 	 * interface identifier on the fly.  we do this because:
1790 	 * (1) it should be the easiest way to find interface identifier.
1791 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1792 	 * for multiple addresses on a single interface, and possible shortcut
1793 	 * of DAD.  we omitted DAD for this reason in the past.
1794 	 * (3) a user can prevent autoconfiguration of global address
1795 	 * by removing link-local address by hand (this is partly because we
1796 	 * don't have other way to control the use of IPv6 on an interface.
1797 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1798 	 * (4) it is easier to manage when an interface has addresses
1799 	 * with the same interface identifier, than to have multiple addresses
1800 	 * with different interface identifiers.
1801 	 */
1802 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1803 	if (ifa)
1804 		ib = (struct in6_ifaddr *)ifa;
1805 	else
1806 		return NULL;
1807 
1808 #if 0 /* don't care link local addr state, and always do DAD */
1809 	/* if link-local address is not eligible, do not autoconfigure. */
1810 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1811 		printf("in6_ifadd: link-local address not ready\n");
1812 		return NULL;
1813 	}
1814 #endif
1815 
1816 	/* prefixlen + ifidlen must be equal to 128 */
1817 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1818 	if (prefixlen != plen0) {
1819 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1820 		    "(prefix=%d ifid=%d)\n",
1821 		    if_name(ifp), prefixlen, 128 - plen0));
1822 		return NULL;
1823 	}
1824 
1825 	/* make ifaddr */
1826 
1827 	memset(&ifra, 0, sizeof(ifra));
1828 	/*
1829 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1830 	 * for safety.
1831 	 */
1832 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1833 	ifra.ifra_addr.sin6_family = AF_INET6;
1834 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1835 	/* prefix */
1836 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1837 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1838 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1839 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1840 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1841 
1842 	/* interface ID */
1843 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1844 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1845 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1846 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1847 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1848 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1849 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1850 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1851 
1852 	/* new prefix mask. */
1853 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1854 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1855 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1856 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1857 
1858 	/* lifetimes */
1859 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1860 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1861 
1862 	/* XXX: scope zone ID? */
1863 
1864 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1865 
1866 	/*
1867 	 * Make sure that we do not have this address already.  This should
1868 	 * usually not happen, but we can still see this case, e.g., if we
1869 	 * have manually configured the exact address to be configured.
1870 	 */
1871 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1872 		/* this should be rare enough to make an explicit log */
1873 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1874 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1875 		return (NULL);
1876 	}
1877 
1878 	/*
1879 	 * Allocate ifaddr structure, link into chain, etc.
1880 	 * If we are going to create a new address upon receiving a multicasted
1881 	 * RA, we need to impose a random delay before starting DAD.
1882 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1883 	 */
1884 	updateflags = 0;
1885 	if (mcast)
1886 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1887 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1888 		nd6log((LOG_ERR,
1889 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1890 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1891 		    error));
1892 		return (NULL);	/* ifaddr must not have been allocated. */
1893 	}
1894 
1895 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1896 
1897 	return (ia);		/* this is always non-NULL */
1898 }
1899 
1900 int
1901 in6_tmpifadd(ia0, forcegen, dad_delay)
1902 	const struct in6_ifaddr *ia0; /* corresponding public address */
1903 	int forcegen, dad_delay;
1904 {
1905 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1906 	struct in6_ifaddr *newia, *ia;
1907 	struct in6_aliasreq ifra;
1908 	int i, error;
1909 	int trylimit = 3;	/* XXX: adhoc value */
1910 	int updateflags;
1911 	u_int32_t randid[2];
1912 	u_int32_t vltime0, pltime0;
1913 	time_t time_second = (time_t)time.tv_sec;
1914 
1915 	memset(&ifra, 0, sizeof(ifra));
1916 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1917 	ifra.ifra_addr = ia0->ia_addr;
1918 	/* copy prefix mask */
1919 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1920 	/* clear the old IFID */
1921 	for (i = 0; i < 4; i++) {
1922 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1923 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1924 	}
1925 
1926   again:
1927 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1928 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1929 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1930 		    "random IFID\n"));
1931 		return (EINVAL);
1932 	}
1933 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1934 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1935 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1936 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1937 
1938 	/*
1939 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1940 	 * However, we may still have a chance to see collision, because
1941 	 * there may be a time lag between generation of the ID and generation
1942 	 * of the address.  So, we'll do one more sanity check.
1943 	 */
1944 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1945 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1946 		    &ifra.ifra_addr.sin6_addr)) {
1947 			if (trylimit-- == 0) {
1948 				/*
1949 				 * Give up.  Something strange should have
1950 				 * happened.
1951 				 */
1952 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1953 				    "find a unique random IFID\n"));
1954 				return (EEXIST);
1955 			}
1956 			forcegen = 1;
1957 			goto again;
1958 		}
1959 	}
1960 
1961 	/*
1962 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1963          * public address or TEMP_VALID_LIFETIME.
1964 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1965          * of the public address or TEMP_PREFERRED_LIFETIME -
1966          * DESYNC_FACTOR.
1967 	 */
1968 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1969 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1970 		    (ia0->ia6_lifetime.ia6t_vltime -
1971 		    (time_second - ia0->ia6_updatetime));
1972 		if (vltime0 > ip6_temp_valid_lifetime)
1973 			vltime0 = ip6_temp_valid_lifetime;
1974 	} else
1975 		vltime0 = ip6_temp_valid_lifetime;
1976 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1977 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1978 		    (ia0->ia6_lifetime.ia6t_pltime -
1979 		    (time_second - ia0->ia6_updatetime));
1980 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
1981 			pltime0 = ip6_temp_preferred_lifetime -
1982 			    ip6_desync_factor;
1983 		}
1984 	} else
1985 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
1986 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1987 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1988 
1989 	/*
1990 	 * A temporary address is created only if this calculated Preferred
1991 	 * Lifetime is greater than REGEN_ADVANCE time units.
1992 	 */
1993 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
1994 		return (0);
1995 
1996 	/* XXX: scope zone ID? */
1997 
1998 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1999 
2000 	/* allocate ifaddr structure, link into chain, etc. */
2001 	updateflags = 0;
2002 	if (dad_delay)
2003 		updateflags |= IN6_IFAUPDATE_DADDELAY;
2004 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2005 		return (error);
2006 
2007 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2008 	if (newia == NULL) {	/* XXX: can it happen? */
2009 		nd6log((LOG_ERR,
2010 		    "in6_tmpifadd: ifa update succeeded, but we got "
2011 		    "no ifaddr\n"));
2012 		return (EINVAL); /* XXX */
2013 	}
2014 	newia->ia6_ndpr = ia0->ia6_ndpr;
2015 	newia->ia6_ndpr->ndpr_refcnt++;
2016 
2017 	/*
2018 	 * A newly added address might affect the status of other addresses.
2019 	 * XXX: when the temporary address is generated with a new public
2020 	 * address, the onlink check is redundant.  However, it would be safe
2021 	 * to do the check explicitly everywhere a new address is generated,
2022 	 * and, in fact, we surely need the check when we create a new
2023 	 * temporary address due to deprecation of an old temporary address.
2024 	 */
2025 	pfxlist_onlink_check();
2026 
2027 	return (0);
2028 }
2029 
2030 static int
2031 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2032 {
2033 
2034 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
2035 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
2036 		nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
2037 		    "(%d) is greater than valid lifetime(%d)\n",
2038 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
2039 		return (EINVAL);
2040 	}
2041 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2042 		ndpr->ndpr_preferred = 0;
2043 	else
2044 		ndpr->ndpr_preferred = time.tv_sec + ndpr->ndpr_pltime;
2045 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2046 		ndpr->ndpr_expire = 0;
2047 	else
2048 		ndpr->ndpr_expire = time.tv_sec + ndpr->ndpr_vltime;
2049 
2050 	return 0;
2051 }
2052 
2053 static void
2054 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2055 {
2056 
2057 	/* Valid lifetime must not be updated unless explicitly specified. */
2058 	/* init ia6t_expire */
2059 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2060 		lt6->ia6t_expire = 0;
2061 	else {
2062 		lt6->ia6t_expire = time.tv_sec;
2063 		lt6->ia6t_expire += lt6->ia6t_vltime;
2064 	}
2065 
2066 	/* init ia6t_preferred */
2067 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2068 		lt6->ia6t_preferred = 0;
2069 	else {
2070 		lt6->ia6t_preferred = time.tv_sec;
2071 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2072 	}
2073 }
2074 
2075 /*
2076  * Delete all the routing table entries that use the specified gateway.
2077  * XXX: this function causes search through all entries of routing table, so
2078  * it shouldn't be called when acting as a router.
2079  */
2080 void
2081 rt6_flush(gateway, ifp)
2082 	struct in6_addr *gateway;
2083 	struct ifnet *ifp;	/* unused */
2084 {
2085 	struct radix_node_head *rnh = rt_tables[AF_INET6];
2086 	int s = splsoftnet();
2087 
2088 	/* We'll care only link-local addresses */
2089 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2090 		splx(s);
2091 		return;
2092 	}
2093 
2094 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2095 	splx(s);
2096 }
2097 
2098 static int
2099 rt6_deleteroute(rn, arg)
2100 	struct radix_node *rn;
2101 	void *arg;
2102 {
2103 #define SIN6(s)	((struct sockaddr_in6 *)s)
2104 	struct rtentry *rt = (struct rtentry *)rn;
2105 	struct in6_addr *gate = (struct in6_addr *)arg;
2106 
2107 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2108 		return (0);
2109 
2110 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr))
2111 		return (0);
2112 
2113 	/*
2114 	 * Do not delete a static route.
2115 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2116 	 * 'cloned' bit instead?
2117 	 */
2118 	if ((rt->rt_flags & RTF_STATIC) != 0)
2119 		return (0);
2120 
2121 	/*
2122 	 * We delete only host route. This means, in particular, we don't
2123 	 * delete default route.
2124 	 */
2125 	if ((rt->rt_flags & RTF_HOST) == 0)
2126 		return (0);
2127 
2128 	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2129 	    rt_mask(rt), rt->rt_flags, 0));
2130 #undef SIN6
2131 }
2132 
2133 int
2134 nd6_setdefaultiface(ifindex)
2135 	int ifindex;
2136 {
2137 	int error = 0;
2138 
2139 	if (ifindex < 0 || if_indexlim <= ifindex)
2140 		return (EINVAL);
2141 	if (ifindex != 0 && !ifindex2ifnet[ifindex])
2142 		return (EINVAL);
2143 
2144 	if (nd6_defifindex != ifindex) {
2145 		nd6_defifindex = ifindex;
2146 		if (nd6_defifindex > 0) {
2147 			nd6_defifp = ifindex2ifnet[nd6_defifindex];
2148 		} else
2149 			nd6_defifp = NULL;
2150 
2151 		/*
2152 		 * Our current implementation assumes one-to-one maping between
2153 		 * interfaces and links, so it would be natural to use the
2154 		 * default interface as the default link.
2155 		 */
2156 		scope6_setdefault(nd6_defifp);
2157 	}
2158 
2159 	return (error);
2160 }
2161