xref: /netbsd-src/sys/netinet6/nd6_rtr.c (revision 274254cdae52594c1aa480a736aef78313d15c9c)
1 /*	$NetBSD: nd6_rtr.c,v 1.78 2009/03/18 16:00:23 cegger 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 1.78 2009/03/18 16:00:23 cegger 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/icmp6_private.h>
62 #include <netinet6/scope6_var.h>
63 
64 #include <net/net_osdep.h>
65 
66 static int rtpref(struct nd_defrouter *);
67 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
68 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
69     struct mbuf *, int);
70 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
71 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
72 	struct nd_defrouter *);
73 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
74 static void pfxrtr_del(struct nd_pfxrouter *);
75 static struct nd_pfxrouter *find_pfxlist_reachable_router
76 	(struct nd_prefix *);
77 static void defrouter_delreq(struct nd_defrouter *);
78 static void nd6_rtmsg(int, struct rtentry *);
79 
80 static int in6_init_prefix_ltimes(struct nd_prefix *);
81 static void in6_init_address_ltimes(struct nd_prefix *ndpr,
82 	struct in6_addrlifetime *lt6);
83 
84 static int rt6_deleteroute(struct rtentry *, void *);
85 
86 extern int nd6_recalc_reachtm_interval;
87 
88 static struct ifnet *nd6_defifp;
89 int nd6_defifindex;
90 
91 int ip6_use_tempaddr = 0;
92 
93 int ip6_desync_factor;
94 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
95 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
96 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
97 
98 /* RTPREF_MEDIUM has to be 0! */
99 #define RTPREF_HIGH	1
100 #define RTPREF_MEDIUM	0
101 #define RTPREF_LOW	(-1)
102 #define RTPREF_RESERVED	(-2)
103 #define RTPREF_INVALID	(-3)	/* internal */
104 
105 /*
106  * Receive Router Solicitation Message - just for routers.
107  * Router solicitation/advertisement is mostly managed by userland program
108  * (rtadvd) so here we have no function like nd6_ra_output().
109  *
110  * Based on RFC 2461
111  */
112 void
113 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
114 {
115 	struct ifnet *ifp = m->m_pkthdr.rcvif;
116 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
117 	struct nd_router_solicit *nd_rs;
118 	struct in6_addr saddr6 = ip6->ip6_src;
119 	char *lladdr = NULL;
120 	int lladdrlen = 0;
121 	union nd_opts ndopts;
122 
123 	/* If I'm not a router, ignore it. */
124 	if (ip6_accept_rtadv != 0 || !ip6_forwarding)
125 		goto freeit;
126 
127 	/* Sanity checks */
128 	if (ip6->ip6_hlim != 255) {
129 		nd6log((LOG_ERR,
130 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
131 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
132 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
133 		goto bad;
134 	}
135 
136 	/*
137 	 * Don't update the neighbor cache, if src = ::.
138 	 * This indicates that the src has no IP address assigned yet.
139 	 */
140 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
141 		goto freeit;
142 
143 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
144 	if (nd_rs == NULL) {
145 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
146 		return;
147 	}
148 
149 	icmp6len -= sizeof(*nd_rs);
150 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
151 	if (nd6_options(&ndopts) < 0) {
152 		nd6log((LOG_INFO,
153 		    "nd6_rs_input: invalid ND option, ignored\n"));
154 		/* nd6_options have incremented stats */
155 		goto freeit;
156 	}
157 
158 	if (ndopts.nd_opts_src_lladdr) {
159 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
160 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
161 	}
162 
163 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
164 		nd6log((LOG_INFO,
165 		    "nd6_rs_input: lladdrlen mismatch for %s "
166 		    "(if %d, RS packet %d)\n",
167 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
168 		goto bad;
169 	}
170 
171 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
172 
173  freeit:
174 	m_freem(m);
175 	return;
176 
177  bad:
178 	ICMP6_STATINC(ICMP6_STAT_BADRS);
179 	m_freem(m);
180 }
181 
182 /*
183  * Receive Router Advertisement Message.
184  *
185  * Based on RFC 2461
186  * TODO: on-link bit on prefix information
187  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
188  */
189 void
190 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
191 {
192 	struct ifnet *ifp = m->m_pkthdr.rcvif;
193 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
194 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
195 	struct nd_router_advert *nd_ra;
196 	struct in6_addr saddr6 = ip6->ip6_src;
197 #if 0
198 	struct in6_addr daddr6 = ip6->ip6_dst;
199 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
200 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
201 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
202 #endif
203 	int mcast = 0;
204 	union nd_opts ndopts;
205 	struct nd_defrouter *dr;
206 
207 	/*
208 	 * We only accept RAs only when
209 	 * the system-wide variable allows the acceptance, and
210 	 * per-interface variable allows RAs on the receiving interface.
211 	 */
212 	if (ip6_accept_rtadv == 0)
213 		goto freeit;
214 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
215 		goto freeit;
216 
217 	if (ip6->ip6_hlim != 255) {
218 		nd6log((LOG_ERR,
219 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
220 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
221 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
222 		goto bad;
223 	}
224 
225 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
226 		nd6log((LOG_ERR,
227 		    "nd6_ra_input: src %s is not link-local\n",
228 		    ip6_sprintf(&saddr6)));
229 		goto bad;
230 	}
231 
232 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
233 	if (nd_ra == NULL) {
234 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
235 		return;
236 	}
237 
238 	icmp6len -= sizeof(*nd_ra);
239 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
240 	if (nd6_options(&ndopts) < 0) {
241 		nd6log((LOG_INFO,
242 		    "nd6_ra_input: invalid ND option, ignored\n"));
243 		/* nd6_options have incremented stats */
244 		goto freeit;
245 	}
246 
247     {
248 	struct nd_defrouter drtr;
249 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
250 
251 	/* remember if this is a multicasted advertisement */
252 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
253 		mcast = 1;
254 
255 	memset(&drtr, 0, sizeof(drtr));
256 	drtr.rtaddr = saddr6;
257 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
258 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
259 	drtr.expire = time_second + drtr.rtlifetime;
260 	drtr.ifp = ifp;
261 	/* unspecified or not? (RFC 2461 6.3.4) */
262 	if (advreachable) {
263 		NTOHL(advreachable);
264 		if (advreachable <= MAX_REACHABLE_TIME &&
265 		    ndi->basereachable != advreachable) {
266 			ndi->basereachable = advreachable;
267 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
268 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
269 		}
270 	}
271 	if (nd_ra->nd_ra_retransmit)
272 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
273 	if (nd_ra->nd_ra_curhoplimit)
274 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
275 	dr = defrtrlist_update(&drtr);
276     }
277 
278 	/*
279 	 * prefix
280 	 */
281 	if (ndopts.nd_opts_pi) {
282 		struct nd_opt_hdr *pt;
283 		struct nd_opt_prefix_info *pi = NULL;
284 		struct nd_prefixctl pr;
285 
286 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
287 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
288 		     pt = (struct nd_opt_hdr *)((char *)pt +
289 						(pt->nd_opt_len << 3))) {
290 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
291 				continue;
292 			pi = (struct nd_opt_prefix_info *)pt;
293 
294 			if (pi->nd_opt_pi_len != 4) {
295 				nd6log((LOG_INFO,
296 				    "nd6_ra_input: invalid option "
297 				    "len %d for prefix information option, "
298 				    "ignored\n", pi->nd_opt_pi_len));
299 				continue;
300 			}
301 
302 			if (128 < pi->nd_opt_pi_prefix_len) {
303 				nd6log((LOG_INFO,
304 				    "nd6_ra_input: invalid prefix "
305 				    "len %d for prefix information option, "
306 				    "ignored\n", pi->nd_opt_pi_prefix_len));
307 				continue;
308 			}
309 
310 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
311 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
312 				nd6log((LOG_INFO,
313 				    "nd6_ra_input: invalid prefix "
314 				    "%s, ignored\n",
315 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
316 				continue;
317 			}
318 
319 			memset(&pr, 0, sizeof(pr));
320 			sockaddr_in6_init(&pr.ndpr_prefix,
321 			    &pi->nd_opt_pi_prefix, 0, 0, 0);
322 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
323 
324 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
325 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
326 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
327 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
328 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
329 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
330 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
331 
332 			(void)prelist_update(&pr, dr, m, mcast);
333 		}
334 	}
335 
336 	/*
337 	 * MTU
338 	 */
339 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
340 		u_long mtu;
341 		u_long maxmtu;
342 
343 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
344 
345 		/* lower bound */
346 		if (mtu < IPV6_MMTU) {
347 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
348 			    "mtu=%lu sent from %s, ignoring\n",
349 			    mtu, ip6_sprintf(&ip6->ip6_src)));
350 			goto skip;
351 		}
352 
353 		/* upper bound */
354 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
355 		    ? ndi->maxmtu : ifp->if_mtu;
356 		if (mtu <= maxmtu) {
357 			int change = (ndi->linkmtu != mtu);
358 
359 			ndi->linkmtu = mtu;
360 			if (change) /* in6_maxmtu may change */
361 				in6_setmaxmtu();
362 		} else {
363 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
364 			    "mtu=%lu sent from %s; "
365 			    "exceeds maxmtu %lu, ignoring\n",
366 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
367 		}
368 	}
369 
370  skip:
371 
372 	/*
373 	 * Source link layer address
374 	 */
375     {
376 	char *lladdr = NULL;
377 	int lladdrlen = 0;
378 
379 	if (ndopts.nd_opts_src_lladdr) {
380 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
381 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
382 	}
383 
384 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
385 		nd6log((LOG_INFO,
386 		    "nd6_ra_input: lladdrlen mismatch for %s "
387 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
388 		    ifp->if_addrlen, lladdrlen - 2));
389 		goto bad;
390 	}
391 
392 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
393 
394 	/*
395 	 * Installing a link-layer address might change the state of the
396 	 * router's neighbor cache, which might also affect our on-link
397 	 * detection of adveritsed prefixes.
398 	 */
399 	pfxlist_onlink_check();
400     }
401 
402  freeit:
403 	m_freem(m);
404 	return;
405 
406  bad:
407 	ICMP6_STATINC(ICMP6_STAT_BADRA);
408 	m_freem(m);
409 }
410 
411 /*
412  * default router list processing sub routines
413  */
414 
415 /* tell the change to user processes watching the routing socket. */
416 static void
417 nd6_rtmsg(int cmd, struct rtentry *rt)
418 {
419 	struct rt_addrinfo info;
420 
421 	memset((void *)&info, 0, sizeof(info));
422 	info.rti_info[RTAX_DST] = rt_getkey(rt);
423 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
424 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
425 	if (rt->rt_ifp) {
426 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->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 	union {
437 		struct sockaddr_in6 sin6;
438 		struct sockaddr sa;
439 	} def, mask, gate;
440 	struct rtentry *newrt = NULL;
441 	int s;
442 	int error;
443 
444 	memset(&def, 0, sizeof(def));
445 	memset(&mask, 0, sizeof(mask));
446 	memset(&gate, 0,sizeof(gate)); /* for safety */
447 
448 	def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len =
449 	    sizeof(struct sockaddr_in6);
450 	def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6;
451 	gate.sin6.sin6_addr = new->rtaddr;
452 #ifndef SCOPEDROUTING
453 	gate.sin6.sin6_scope_id = 0;	/* XXX */
454 #endif
455 
456 	s = splsoftnet();
457 	error = rtrequest(RTM_ADD, &def.sa, &gate.sa, &mask.sa,
458 	    RTF_GATEWAY, &newrt);
459 	if (newrt) {
460 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
461 		newrt->rt_refcnt--;
462 	}
463 	if (error == 0)
464 		new->installed = 1;
465 	splx(s);
466 	return;
467 }
468 
469 struct nd_defrouter *
470 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
471 {
472 	struct nd_defrouter *dr;
473 
474 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
475 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
476 			break;
477 	}
478 
479 	return dr;		/* search failed */
480 }
481 
482 void
483 defrtrlist_del(struct nd_defrouter *dr)
484 {
485 	struct nd_defrouter *deldr = NULL;
486 	struct nd_prefix *pr;
487 
488 	/*
489 	 * Flush all the routing table entries that use the router
490 	 * as a next hop.
491 	 */
492 	if (!ip6_forwarding && ip6_accept_rtadv) /* XXX: better condition? */
493 		rt6_flush(&dr->rtaddr, dr->ifp);
494 
495 	if (dr->installed) {
496 		deldr = dr;
497 		defrouter_delreq(dr);
498 	}
499 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
500 
501 	/*
502 	 * Also delete all the pointers to the router in each prefix lists.
503 	 */
504 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
505 		struct nd_pfxrouter *pfxrtr;
506 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
507 			pfxrtr_del(pfxrtr);
508 	}
509 	pfxlist_onlink_check();
510 
511 	/*
512 	 * If the router is the primary one, choose a new one.
513 	 * Note that defrouter_select() will remove the current gateway
514 	 * from the routing table.
515 	 */
516 	if (deldr)
517 		defrouter_select();
518 
519 	free(dr, M_IP6NDP);
520 }
521 
522 /*
523  * Remove the default route for a given router.
524  * This is just a subroutine function for defrouter_select(), and should
525  * not be called from anywhere else.
526  */
527 static void
528 defrouter_delreq(struct nd_defrouter *dr)
529 {
530 	union {
531 		struct sockaddr_in6 sin6;
532 		struct sockaddr sa;
533 	} def, mask, gw;
534 	struct rtentry *oldrt = NULL;
535 
536 #ifdef DIAGNOSTIC
537 	if (dr == NULL)
538 		panic("dr == NULL in defrouter_delreq");
539 #endif
540 
541 	memset(&def, 0, sizeof(def));
542 	memset(&mask, 0, sizeof(mask));
543 	memset(&gw, 0, sizeof(gw));	/* for safety */
544 
545 	def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len =
546 	    sizeof(struct sockaddr_in6);
547 	def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6;
548 	gw.sin6.sin6_addr = dr->rtaddr;
549 #ifndef SCOPEDROUTING
550 	gw.sin6.sin6_scope_id = 0;	/* XXX */
551 #endif
552 
553 	rtrequest(RTM_DELETE, &def.sa, &gw.sa, &mask.sa, RTF_GATEWAY, &oldrt);
554 	if (oldrt) {
555 		nd6_rtmsg(RTM_DELETE, oldrt);
556 		if (oldrt->rt_refcnt <= 0) {
557 			/*
558 			 * XXX: borrowed from the RTM_DELETE case of
559 			 * rtrequest().
560 			 */
561 			oldrt->rt_refcnt++;
562 			rtfree(oldrt);
563 		}
564 	}
565 
566 	dr->installed = 0;
567 }
568 
569 /*
570  * remove all default routes from default router list
571  */
572 void
573 defrouter_reset(void)
574 {
575 	struct nd_defrouter *dr;
576 
577 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
578 	     dr = TAILQ_NEXT(dr, dr_entry))
579 		defrouter_delreq(dr);
580 
581 	/*
582 	 * XXX should we also nuke any default routers in the kernel, by
583 	 * going through them by rtalloc1()?
584 	 */
585 }
586 
587 /*
588  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
589  * draft-ietf-ipngwg-router-selection:
590  * 1) Routers that are reachable or probably reachable should be preferred.
591  *    If we have more than one (probably) reachable router, prefer ones
592  *    with the highest router preference.
593  * 2) When no routers on the list are known to be reachable or
594  *    probably reachable, routers SHOULD be selected in a round-robin
595  *    fashion, regardless of router preference values.
596  * 3) If the Default Router List is empty, assume that all
597  *    destinations are on-link.
598  *
599  * We assume nd_defrouter is sorted by router preference value.
600  * Since the code below covers both with and without router preference cases,
601  * we do not need to classify the cases by ifdef.
602  *
603  * At this moment, we do not try to install more than one default router,
604  * even when the multipath routing is available, because we're not sure about
605  * the benefits for stub hosts comparing to the risk of making the code
606  * complicated and the possibility of introducing bugs.
607  */
608 void
609 defrouter_select(void)
610 {
611 	int s = splsoftnet();
612 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
613 	struct rtentry *rt = NULL;
614 	struct llinfo_nd6 *ln = NULL;
615 
616 	/*
617 	 * This function should be called only when acting as an autoconfigured
618 	 * host.  Although the remaining part of this function is not effective
619 	 * if the node is not an autoconfigured host, we explicitly exclude
620 	 * such cases here for safety.
621 	 */
622 	if (ip6_forwarding || !ip6_accept_rtadv) {
623 		nd6log((LOG_WARNING,
624 		    "defrouter_select: called unexpectedly (forwarding=%d, "
625 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
626 		splx(s);
627 		return;
628 	}
629 
630 	/*
631 	 * Let's handle easy case (3) first:
632 	 * If default router list is empty, there's nothing to be done.
633 	 */
634 	if (!TAILQ_FIRST(&nd_defrouter)) {
635 		splx(s);
636 		return;
637 	}
638 
639 	/*
640 	 * Search for a (probably) reachable router from the list.
641 	 * We just pick up the first reachable one (if any), assuming that
642 	 * the ordering rule of the list described in defrtrlist_update().
643 	 */
644 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
645 	     dr = TAILQ_NEXT(dr, dr_entry)) {
646 		if (selected_dr == NULL &&
647 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) != NULL &&
648 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) != NULL &&
649 		    ND6_IS_LLINFO_PROBREACH(ln)) {
650 			selected_dr = dr;
651 		}
652 
653 		if (dr->installed && !installed_dr)
654 			installed_dr = dr;
655 		else if (dr->installed && installed_dr) {
656 			/* this should not happen.  warn for diagnosis. */
657 			log(LOG_ERR, "defrouter_select: more than one router"
658 			    " is installed\n");
659 		}
660 	}
661 	/*
662 	 * If none of the default routers was found to be reachable,
663 	 * round-robin the list regardless of preference.
664 	 * Otherwise, if we have an installed router, check if the selected
665 	 * (reachable) router should really be preferred to the installed one.
666 	 * We only prefer the new router when the old one is not reachable
667 	 * or when the new one has a really higher preference value.
668 	 */
669 	if (selected_dr == NULL) {
670 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
671 			selected_dr = TAILQ_FIRST(&nd_defrouter);
672 		else
673 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
674 	} else if (installed_dr &&
675 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
676 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
677 	    ND6_IS_LLINFO_PROBREACH(ln) &&
678 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
679 		selected_dr = installed_dr;
680 	}
681 
682 	/*
683 	 * If the selected router is different than the installed one,
684 	 * remove the installed router and install the selected one.
685 	 * Note that the selected router is never NULL here.
686 	 */
687 	if (installed_dr != selected_dr) {
688 		if (installed_dr)
689 			defrouter_delreq(installed_dr);
690 		defrouter_addreq(selected_dr);
691 	}
692 
693 	splx(s);
694 	return;
695 }
696 
697 /*
698  * for default router selection
699  * regards router-preference field as a 2-bit signed integer
700  */
701 static int
702 rtpref(struct nd_defrouter *dr)
703 {
704 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
705 	case ND_RA_FLAG_RTPREF_HIGH:
706 		return (RTPREF_HIGH);
707 	case ND_RA_FLAG_RTPREF_MEDIUM:
708 	case ND_RA_FLAG_RTPREF_RSV:
709 		return (RTPREF_MEDIUM);
710 	case ND_RA_FLAG_RTPREF_LOW:
711 		return (RTPREF_LOW);
712 	default:
713 		/*
714 		 * This case should never happen.  If it did, it would mean a
715 		 * serious bug of kernel internal.  We thus always bark here.
716 		 * Or, can we even panic?
717 		 */
718 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
719 		return (RTPREF_INVALID);
720 	}
721 	/* NOTREACHED */
722 }
723 
724 static struct nd_defrouter *
725 defrtrlist_update(struct nd_defrouter *new)
726 {
727 	struct nd_defrouter *dr, *n;
728 	int s = splsoftnet();
729 
730 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
731 		/* entry exists */
732 		if (new->rtlifetime == 0) {
733 			defrtrlist_del(dr);
734 			dr = NULL;
735 		} else {
736 			int oldpref = rtpref(dr);
737 
738 			/* override */
739 			dr->flags = new->flags; /* xxx flag check */
740 			dr->rtlifetime = new->rtlifetime;
741 			dr->expire = new->expire;
742 
743 			/*
744 			 * If the preference does not change, there's no need
745 			 * to sort the entries.
746 			 */
747 			if (rtpref(new) == oldpref) {
748 				splx(s);
749 				return (dr);
750 			}
751 
752 			/*
753 			 * preferred router may be changed, so relocate
754 			 * this router.
755 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
756 			 * However, since defrtrlist_del() has many side
757 			 * effects, we intentionally do so here.
758 			 * defrouter_select() below will handle routing
759 			 * changes later.
760 			 */
761 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
762 			n = dr;
763 			goto insert;
764 		}
765 		splx(s);
766 		return (dr);
767 	}
768 
769 	/* entry does not exist */
770 	if (new->rtlifetime == 0) {
771 		splx(s);
772 		return (NULL);
773 	}
774 
775 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
776 	if (n == NULL) {
777 		splx(s);
778 		return (NULL);
779 	}
780 	memset(n, 0, sizeof(*n));
781 	*n = *new;
782 
783 insert:
784 	/*
785 	 * Insert the new router in the Default Router List;
786 	 * The Default Router List should be in the descending order
787 	 * of router-preferece.  Routers with the same preference are
788 	 * sorted in the arriving time order.
789 	 */
790 
791 	/* insert at the end of the group */
792 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
793 	     dr = TAILQ_NEXT(dr, dr_entry)) {
794 		if (rtpref(n) > rtpref(dr))
795 			break;
796 	}
797 	if (dr)
798 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
799 	else
800 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
801 
802 	defrouter_select();
803 
804 	splx(s);
805 
806 	return (n);
807 }
808 
809 static struct nd_pfxrouter *
810 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
811 {
812 	struct nd_pfxrouter *search;
813 
814 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
815 		if (search->router == dr)
816 			break;
817 	}
818 
819 	return (search);
820 }
821 
822 static void
823 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
824 {
825 	struct nd_pfxrouter *new;
826 
827 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT|M_ZERO);
828 	if (new == NULL)
829 		return;
830 	new->router = dr;
831 
832 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
833 
834 	pfxlist_onlink_check();
835 }
836 
837 static void
838 pfxrtr_del(struct nd_pfxrouter *pfr)
839 {
840 	LIST_REMOVE(pfr, pfr_entry);
841 	free(pfr, M_IP6NDP);
842 }
843 
844 struct nd_prefix *
845 nd6_prefix_lookup(struct nd_prefixctl *key)
846 {
847 	struct nd_prefix *search;
848 
849 	LIST_FOREACH(search, &nd_prefix, ndpr_entry) {
850 		if (key->ndpr_ifp == search->ndpr_ifp &&
851 		    key->ndpr_plen == search->ndpr_plen &&
852 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
853 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
854 			break;
855 		}
856 	}
857 
858 	return (search);
859 }
860 
861 int
862 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
863 	struct nd_prefix **newp)
864 {
865 	struct nd_prefix *new = NULL;
866 	int i, s;
867 	int error;
868 
869 	error = 0;
870 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT|M_ZERO);
871 	if (new == NULL)
872 		return ENOMEM;
873 	new->ndpr_ifp = pr->ndpr_ifp;
874 	new->ndpr_prefix = pr->ndpr_prefix;
875 	new->ndpr_plen = pr->ndpr_plen;
876 	new->ndpr_vltime = pr->ndpr_vltime;
877 	new->ndpr_pltime = pr->ndpr_pltime;
878 	new->ndpr_flags = pr->ndpr_flags;
879 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
880 		free(new, M_IP6NDP);
881 		return(error);
882 	}
883 	new->ndpr_lastupdate = time_second;
884 	if (newp != NULL)
885 		*newp = new;
886 
887 	/* initialization */
888 	LIST_INIT(&new->ndpr_advrtrs);
889 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
890 	/* make prefix in the canonical form */
891 	for (i = 0; i < 4; i++)
892 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
893 		    new->ndpr_mask.s6_addr32[i];
894 
895 	s = splsoftnet();
896 	/* link ndpr_entry to nd_prefix list */
897 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
898 	splx(s);
899 
900 	/* ND_OPT_PI_FLAG_ONLINK processing */
901 	if (new->ndpr_raf_onlink) {
902 		int e;
903 
904 		if ((e = nd6_prefix_onlink(new)) != 0) {
905 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
906 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
907 			    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
908 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
909 			/* proceed anyway. XXX: is it correct? */
910 		}
911 	}
912 
913 	if (dr)
914 		pfxrtr_add(new, dr);
915 
916 	return 0;
917 }
918 
919 void
920 prelist_remove(struct nd_prefix *pr)
921 {
922 	struct nd_pfxrouter *pfr, *next;
923 	int e, s;
924 
925 	/* make sure to invalidate the prefix until it is really freed. */
926 	pr->ndpr_vltime = 0;
927 	pr->ndpr_pltime = 0;
928 #if 0
929 	/*
930 	 * Though these flags are now meaningless, we'd rather keep the value
931 	 * not to confuse users when executing "ndp -p".
932 	 */
933 	pr->ndpr_raf_onlink = 0;
934 	pr->ndpr_raf_auto = 0;
935 #endif
936 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
937 	    (e = nd6_prefix_offlink(pr)) != 0) {
938 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
939 		    "on %s, errno=%d\n",
940 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
941 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
942 		/* what should we do? */
943 	}
944 
945 	if (pr->ndpr_refcnt > 0)
946 		return;		/* notice here? */
947 
948 	s = splsoftnet();
949 	/* unlink ndpr_entry from nd_prefix list */
950 	LIST_REMOVE(pr, ndpr_entry);
951 
952 	/* free list of routers that adversed the prefix */
953 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
954 		next = LIST_NEXT(pfr, pfr_entry);
955 
956 		free(pfr, M_IP6NDP);
957 	}
958 	splx(s);
959 
960 	free(pr, M_IP6NDP);
961 
962 	pfxlist_onlink_check();
963 }
964 
965 static int
966 prelist_update(struct nd_prefixctl *new,
967 	struct nd_defrouter *dr, /* may be NULL */
968 	struct mbuf *m,
969 	int mcast)
970 {
971 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
972 	struct ifaddr *ifa;
973 	struct ifnet *ifp = new->ndpr_ifp;
974 	struct nd_prefix *pr;
975 	int s = splsoftnet();
976 	int error = 0;
977 	int newprefix = 0;
978 	int auth;
979 	struct in6_addrlifetime lt6_tmp;
980 
981 	auth = 0;
982 	if (m) {
983 		/*
984 		 * Authenticity for NA consists authentication for
985 		 * both IP header and IP datagrams, doesn't it ?
986 		 */
987 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
988 		auth = (m->m_flags & M_AUTHIPHDR
989 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
990 #endif
991 	}
992 
993 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
994 		/*
995 		 * nd6_prefix_lookup() ensures that pr and new have the same
996 		 * prefix on a same interface.
997 		 */
998 
999 		/*
1000 		 * Update prefix information.  Note that the on-link (L) bit
1001 		 * and the autonomous (A) bit should NOT be changed from 1
1002 		 * to 0.
1003 		 */
1004 		if (new->ndpr_raf_onlink == 1)
1005 			pr->ndpr_raf_onlink = 1;
1006 		if (new->ndpr_raf_auto == 1)
1007 			pr->ndpr_raf_auto = 1;
1008 		if (new->ndpr_raf_onlink) {
1009 			pr->ndpr_vltime = new->ndpr_vltime;
1010 			pr->ndpr_pltime = new->ndpr_pltime;
1011 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1012 			pr->ndpr_lastupdate = time_second;
1013 		}
1014 
1015 		if (new->ndpr_raf_onlink &&
1016 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1017 			int e;
1018 
1019 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1020 				nd6log((LOG_ERR,
1021 				    "prelist_update: failed to make "
1022 				    "the prefix %s/%d on-link on %s "
1023 				    "(errno=%d)\n",
1024 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1025 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1026 				/* proceed anyway. XXX: is it correct? */
1027 			}
1028 		}
1029 
1030 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1031 			pfxrtr_add(pr, dr);
1032 	} else {
1033 		struct nd_prefix *newpr = NULL;
1034 
1035 		newprefix = 1;
1036 
1037 		if (new->ndpr_vltime == 0)
1038 			goto end;
1039 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1040 			goto end;
1041 
1042 		error = nd6_prelist_add(new, dr, &newpr);
1043 		if (error != 0 || newpr == NULL) {
1044 			nd6log((LOG_NOTICE, "prelist_update: "
1045 			    "nd6_prelist_add failed for %s/%d on %s "
1046 			    "errno=%d, returnpr=%p\n",
1047 			    ip6_sprintf(&new->ndpr_prefix.sin6_addr),
1048 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1049 			    error, newpr));
1050 			goto end; /* we should just give up in this case. */
1051 		}
1052 
1053 		/*
1054 		 * XXX: from the ND point of view, we can ignore a prefix
1055 		 * with the on-link bit being zero.  However, we need a
1056 		 * prefix structure for references from autoconfigured
1057 		 * addresses.  Thus, we explicitly make sure that the prefix
1058 		 * itself expires now.
1059 		 */
1060 		if (newpr->ndpr_raf_onlink == 0) {
1061 			newpr->ndpr_vltime = 0;
1062 			newpr->ndpr_pltime = 0;
1063 			in6_init_prefix_ltimes(newpr);
1064 		}
1065 
1066 		pr = newpr;
1067 	}
1068 
1069 	/*
1070 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1071 	 * Note that pr must be non NULL at this point.
1072 	 */
1073 
1074 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1075 	if (!new->ndpr_raf_auto)
1076 		goto end;
1077 
1078 	/*
1079 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1080 	 * nd6_ra_input.
1081 	 */
1082 
1083 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1084 	if (new->ndpr_pltime > new->ndpr_vltime) {
1085 		error = EINVAL;	/* XXX: won't be used */
1086 		goto end;
1087 	}
1088 
1089 	/*
1090 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1091 	 * an address configured by stateless autoconfiguration already in the
1092 	 * list of addresses associated with the interface, and the Valid
1093 	 * Lifetime is not 0, form an address.  We first check if we have
1094 	 * a matching prefix.
1095 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1096 	 * consider autoconfigured addresses while RFC2462 simply said
1097 	 * "address".
1098 	 */
1099 	IFADDR_FOREACH(ifa, ifp) {
1100 		struct in6_ifaddr *ifa6;
1101 		u_int32_t remaininglifetime;
1102 
1103 		if (ifa->ifa_addr->sa_family != AF_INET6)
1104 			continue;
1105 
1106 		ifa6 = (struct in6_ifaddr *)ifa;
1107 
1108 		/*
1109 		 * We only consider autoconfigured addresses as per rfc2462bis.
1110 		 */
1111 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1112 			continue;
1113 
1114 		/*
1115 		 * Spec is not clear here, but I believe we should concentrate
1116 		 * on unicast (i.e. not anycast) addresses.
1117 		 * XXX: other ia6_flags? detached or duplicated?
1118 		 */
1119 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1120 			continue;
1121 
1122 		/*
1123 		 * Ignore the address if it is not associated with a prefix
1124 		 * or is associated with a prefix that is different from this
1125 		 * one.  (pr is never NULL here)
1126 		 */
1127 		if (ifa6->ia6_ndpr != pr)
1128 			continue;
1129 
1130 		if (ia6_match == NULL) /* remember the first one */
1131 			ia6_match = ifa6;
1132 
1133 		/*
1134 		 * An already autoconfigured address matched.  Now that we
1135 		 * are sure there is at least one matched address, we can
1136 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1137 		 * "two hours" rule and the privacy extension.
1138 		 * We apply some clarifications in rfc2462bis:
1139 		 * - use remaininglifetime instead of storedlifetime as a
1140 		 *   variable name
1141 		 * - remove the dead code in the "two-hour" rule
1142 		 */
1143 #define TWOHOUR		(120*60)
1144 		lt6_tmp = ifa6->ia6_lifetime;
1145 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1146 			remaininglifetime = ND6_INFINITE_LIFETIME;
1147 		else if (time_second - ifa6->ia6_updatetime >
1148 			 lt6_tmp.ia6t_vltime) {
1149 			/*
1150 			 * The case of "invalid" address.  We should usually
1151 			 * not see this case.
1152 			 */
1153 			remaininglifetime = 0;
1154 		} else
1155 			remaininglifetime = lt6_tmp.ia6t_vltime -
1156 			    (time_second - ifa6->ia6_updatetime);
1157 
1158 		/* when not updating, keep the current stored lifetime. */
1159 		lt6_tmp.ia6t_vltime = remaininglifetime;
1160 
1161 		if (TWOHOUR < new->ndpr_vltime ||
1162 		    remaininglifetime < new->ndpr_vltime) {
1163 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1164 		} else if (remaininglifetime <= TWOHOUR) {
1165 			if (auth)
1166 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1167 		} else {
1168 			/*
1169 			 * new->ndpr_vltime <= TWOHOUR &&
1170 			 * TWOHOUR < remaininglifetime
1171 			 */
1172 			lt6_tmp.ia6t_vltime = TWOHOUR;
1173 		}
1174 
1175 		/* The 2 hour rule is not imposed for preferred lifetime. */
1176 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1177 
1178 		in6_init_address_ltimes(pr, &lt6_tmp);
1179 
1180 		/*
1181 		 * We need to treat lifetimes for temporary addresses
1182 		 * differently, according to
1183 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1184 		 * we only update the lifetimes when they are in the maximum
1185 		 * intervals.
1186 		 */
1187 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1188 			u_int32_t maxvltime, maxpltime;
1189 
1190 			if (ip6_temp_valid_lifetime >
1191 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1192 			    ip6_desync_factor)) {
1193 				maxvltime = ip6_temp_valid_lifetime -
1194 				    (time_second - ifa6->ia6_createtime) -
1195 				    ip6_desync_factor;
1196 			} else
1197 				maxvltime = 0;
1198 			if (ip6_temp_preferred_lifetime >
1199 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1200 			    ip6_desync_factor)) {
1201 				maxpltime = ip6_temp_preferred_lifetime -
1202 				    (time_second - ifa6->ia6_createtime) -
1203 				    ip6_desync_factor;
1204 			} else
1205 				maxpltime = 0;
1206 
1207 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1208 			    lt6_tmp.ia6t_vltime > maxvltime) {
1209 				lt6_tmp.ia6t_vltime = maxvltime;
1210 			}
1211 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1212 			    lt6_tmp.ia6t_pltime > maxpltime) {
1213 				lt6_tmp.ia6t_pltime = maxpltime;
1214 			}
1215 		}
1216 
1217 		ifa6->ia6_lifetime = lt6_tmp;
1218 		ifa6->ia6_updatetime = time_second;
1219 	}
1220 	if (ia6_match == NULL && new->ndpr_vltime) {
1221 		int ifidlen;
1222 
1223 		/*
1224 		 * 5.5.3 (d) (continued)
1225 		 * No address matched and the valid lifetime is non-zero.
1226 		 * Create a new address.
1227 		 */
1228 
1229 		/*
1230 		 * Prefix Length check:
1231 		 * If the sum of the prefix length and interface identifier
1232 		 * length does not equal 128 bits, the Prefix Information
1233 		 * option MUST be ignored.  The length of the interface
1234 		 * identifier is defined in a separate link-type specific
1235 		 * document.
1236 		 */
1237 		ifidlen = in6_if2idlen(ifp);
1238 		if (ifidlen < 0) {
1239 			/* this should not happen, so we always log it. */
1240 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1241 			    if_name(ifp));
1242 			goto end;
1243 		}
1244 		if (ifidlen + pr->ndpr_plen != 128) {
1245 			nd6log((LOG_INFO,
1246 			    "prelist_update: invalid prefixlen "
1247 			    "%d for %s, ignored\n",
1248 			    pr->ndpr_plen, if_name(ifp)));
1249 			goto end;
1250 		}
1251 
1252 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1253 			/*
1254 			 * note that we should use pr (not new) for reference.
1255 			 */
1256 			pr->ndpr_refcnt++;
1257 			ia6->ia6_ndpr = pr;
1258 
1259 			/*
1260 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
1261 			 * When a new public address is created as described
1262 			 * in RFC2462, also create a new temporary address.
1263 			 *
1264 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
1265 			 * When an interface connects to a new link, a new
1266 			 * randomized interface identifier should be generated
1267 			 * immediately together with a new set of temporary
1268 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1269 			 * in6_tmpifadd().
1270 			 */
1271 			if (ip6_use_tempaddr) {
1272 				int e;
1273 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1274 					nd6log((LOG_NOTICE, "prelist_update: "
1275 					    "failed to create a temporary "
1276 					    "address, errno=%d\n",
1277 					    e));
1278 				}
1279 			}
1280 
1281 			/*
1282 			 * A newly added address might affect the status
1283 			 * of other addresses, so we check and update it.
1284 			 * XXX: what if address duplication happens?
1285 			 */
1286 			pfxlist_onlink_check();
1287 		} else {
1288 			/* just set an error. do not bark here. */
1289 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1290 		}
1291 	}
1292 
1293  end:
1294 	splx(s);
1295 	return error;
1296 }
1297 
1298 /*
1299  * A supplement function used in the on-link detection below;
1300  * detect if a given prefix has a (probably) reachable advertising router.
1301  * XXX: lengthy function name...
1302  */
1303 static struct nd_pfxrouter *
1304 find_pfxlist_reachable_router(struct nd_prefix *pr)
1305 {
1306 	struct nd_pfxrouter *pfxrtr;
1307 	struct rtentry *rt;
1308 	struct llinfo_nd6 *ln;
1309 
1310 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1311 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1312 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1313 		    pfxrtr->router->ifp)) &&
1314 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1315 		    ND6_IS_LLINFO_PROBREACH(ln))
1316 			break;	/* found */
1317 	}
1318 
1319 	return (pfxrtr);
1320 }
1321 
1322 /*
1323  * Check if each prefix in the prefix list has at least one available router
1324  * that advertised the prefix (a router is "available" if its neighbor cache
1325  * entry is reachable or probably reachable).
1326  * If the check fails, the prefix may be off-link, because, for example,
1327  * we have moved from the network but the lifetime of the prefix has not
1328  * expired yet.  So we should not use the prefix if there is another prefix
1329  * that has an available router.
1330  * But, if there is no prefix that has an available router, we still regards
1331  * all the prefixes as on-link.  This is because we can't tell if all the
1332  * routers are simply dead or if we really moved from the network and there
1333  * is no router around us.
1334  */
1335 void
1336 pfxlist_onlink_check(void)
1337 {
1338 	struct nd_prefix *pr;
1339 	struct in6_ifaddr *ifa;
1340 	struct nd_defrouter *dr;
1341 	struct nd_pfxrouter *pfxrtr = NULL;
1342 
1343 	/*
1344 	 * Check if there is a prefix that has a reachable advertising
1345 	 * router.
1346 	 */
1347 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1348 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1349 			break;
1350 	}
1351 	/*
1352 	 * If we have no such prefix, check whether we still have a router
1353 	 * that does not advertise any prefixes.
1354 	 */
1355 	if (pr == NULL) {
1356 		TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1357 			struct nd_prefix *pr0;
1358 
1359 			LIST_FOREACH(pr0, &nd_prefix, ndpr_entry) {
1360 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1361 					break;
1362 			}
1363 			if (pfxrtr)
1364 				break;
1365 		}
1366 	}
1367 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
1368 		/*
1369 		 * There is at least one prefix that has a reachable router,
1370 		 * or at least a router which probably does not advertise
1371 		 * any prefixes.  The latter would be the case when we move
1372 		 * to a new link where we have a router that does not provide
1373 		 * prefixes and we configure an address by hand.
1374 		 * Detach prefixes which have no reachable advertising
1375 		 * router, and attach other prefixes.
1376 		 */
1377 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1378 			/* XXX: a link-local prefix should never be detached */
1379 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1380 				continue;
1381 
1382 			/*
1383 			 * we aren't interested in prefixes without the L bit
1384 			 * set.
1385 			 */
1386 			if (pr->ndpr_raf_onlink == 0)
1387 				continue;
1388 
1389 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1390 			    find_pfxlist_reachable_router(pr) == NULL)
1391 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1392 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1393 			    find_pfxlist_reachable_router(pr) != 0)
1394 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1395 		}
1396 	} else {
1397 		/* there is no prefix that has a reachable router */
1398 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1399 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1400 				continue;
1401 
1402 			if (pr->ndpr_raf_onlink == 0)
1403 				continue;
1404 
1405 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1406 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1407 		}
1408 	}
1409 
1410 	/*
1411 	 * Remove each interface route associated with a (just) detached
1412 	 * prefix, and reinstall the interface route for a (just) attached
1413 	 * prefix.  Note that all attempt of reinstallation does not
1414 	 * necessarily success, when a same prefix is shared among multiple
1415 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1416 	 * so we don't have to care about them.
1417 	 */
1418 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1419 		int e;
1420 
1421 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1422 			continue;
1423 
1424 		if (pr->ndpr_raf_onlink == 0)
1425 			continue;
1426 
1427 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1428 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1429 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1430 				nd6log((LOG_ERR,
1431 				    "pfxlist_onlink_check: failed to "
1432 				    "make %s/%d offlink, errno=%d\n",
1433 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1434 				    pr->ndpr_plen, e));
1435 			}
1436 		}
1437 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1438 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1439 		    pr->ndpr_raf_onlink) {
1440 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1441 				nd6log((LOG_ERR,
1442 				    "pfxlist_onlink_check: failed to "
1443 				    "make %s/%d onlink, errno=%d\n",
1444 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1445 				    pr->ndpr_plen, e));
1446 			}
1447 		}
1448 	}
1449 
1450 	/*
1451 	 * Changes on the prefix status might affect address status as well.
1452 	 * Make sure that all addresses derived from an attached prefix are
1453 	 * attached, and that all addresses derived from a detached prefix are
1454 	 * detached.  Note, however, that a manually configured address should
1455 	 * always be attached.
1456 	 * The precise detection logic is same as the one for prefixes.
1457 	 */
1458 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1459 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1460 			continue;
1461 
1462 		if (ifa->ia6_ndpr == NULL) {
1463 			/*
1464 			 * This can happen when we first configure the address
1465 			 * (i.e. the address exists, but the prefix does not).
1466 			 * XXX: complicated relationships...
1467 			 */
1468 			continue;
1469 		}
1470 
1471 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1472 			break;
1473 	}
1474 	if (ifa) {
1475 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1476 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1477 				continue;
1478 
1479 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1480 				continue;
1481 
1482 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1483 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1484 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1485 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1486 					nd6_dad_start((struct ifaddr *)ifa,
1487 					    0);
1488 				}
1489 			} else {
1490 				if ((ifa->ia6_flags & IN6_IFF_DETACHED) == 0) {
1491 					ifa->ia6_flags |= IN6_IFF_DETACHED;
1492 				}
1493 			}
1494 		}
1495 	}
1496 	else {
1497 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1498 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1499 				continue;
1500 
1501 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1502 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1503 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1504 				/* Do we need a delay in this case? */
1505 				nd6_dad_start((struct ifaddr *)ifa, 0);
1506 			}
1507 		}
1508 	}
1509 }
1510 
1511 int
1512 nd6_prefix_onlink(struct nd_prefix *pr)
1513 {
1514 	struct ifaddr *ifa;
1515 	struct ifnet *ifp = pr->ndpr_ifp;
1516 	struct sockaddr_in6 mask6;
1517 	struct nd_prefix *opr;
1518 	u_long rtflags;
1519 	int error = 0;
1520 	struct rtentry *rt = NULL;
1521 
1522 	/* sanity check */
1523 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1524 		nd6log((LOG_ERR,
1525 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1526 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1527 		return (EEXIST);
1528 	}
1529 
1530 	/*
1531 	 * Add the interface route associated with the prefix.  Before
1532 	 * installing the route, check if there's the same prefix on another
1533 	 * interface, and the prefix has already installed the interface route.
1534 	 * Although such a configuration is expected to be rare, we explicitly
1535 	 * allow it.
1536 	 */
1537 	LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1538 		if (opr == pr)
1539 			continue;
1540 
1541 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1542 			continue;
1543 
1544 		if (opr->ndpr_plen == pr->ndpr_plen &&
1545 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1546 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1547 			return (0);
1548 	}
1549 
1550 	/*
1551 	 * We prefer link-local addresses as the associated interface address.
1552 	 */
1553 	/* search for a link-local addr */
1554 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1555 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1556 	if (ifa == NULL) {
1557 		/* XXX: freebsd does not have ifa_ifwithaf */
1558 		IFADDR_FOREACH(ifa, ifp) {
1559 			if (ifa->ifa_addr->sa_family == AF_INET6)
1560 				break;
1561 		}
1562 		/* should we care about ia6_flags? */
1563 	}
1564 	if (ifa == NULL) {
1565 		/*
1566 		 * This can still happen, when, for example, we receive an RA
1567 		 * containing a prefix with the L bit set and the A bit clear,
1568 		 * after removing all IPv6 addresses on the receiving
1569 		 * interface.  This should, of course, be rare though.
1570 		 */
1571 		nd6log((LOG_NOTICE,
1572 		    "nd6_prefix_onlink: failed to find any ifaddr"
1573 		    " to add route for a prefix(%s/%d) on %s\n",
1574 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1575 		    pr->ndpr_plen, if_name(ifp)));
1576 		return (0);
1577 	}
1578 
1579 	/*
1580 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1581 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1582 	 */
1583 	memset(&mask6, 0, sizeof(mask6));
1584 	mask6.sin6_len = sizeof(mask6);
1585 	mask6.sin6_addr = pr->ndpr_mask;
1586 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
1587 	rtflags = ifa->ifa_flags | RTF_UP;
1588 	if (nd6_need_cache(ifp)) {
1589 		/* explicitly set in case ifa_flags does not set the flag. */
1590 		rtflags |= RTF_CLONING;
1591 	} else {
1592 		/*
1593 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1594 		 */
1595 		rtflags &= ~RTF_CLONING;
1596 	}
1597 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1598 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1599 	if (error == 0) {
1600 		if (rt != NULL) /* this should be non NULL, though */
1601 			nd6_rtmsg(RTM_ADD, rt);
1602 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1603 	} else {
1604 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1605 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1606 		    "errno = %d\n",
1607 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1608 		    pr->ndpr_plen, if_name(ifp),
1609 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1610 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
1611 	}
1612 
1613 	if (rt != NULL)
1614 		rt->rt_refcnt--;
1615 
1616 	return (error);
1617 }
1618 
1619 int
1620 nd6_prefix_offlink(struct nd_prefix *pr)
1621 {
1622 	int error = 0;
1623 	struct ifnet *ifp = pr->ndpr_ifp;
1624 	struct nd_prefix *opr;
1625 	struct sockaddr_in6 sa6, mask6;
1626 	struct rtentry *rt = NULL;
1627 
1628 	/* sanity check */
1629 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1630 		nd6log((LOG_ERR,
1631 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1632 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1633 		return (EEXIST);
1634 	}
1635 
1636 	sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1637 	sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
1638 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1639 	    (struct sockaddr *)&mask6, 0, &rt);
1640 	if (error == 0) {
1641 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1642 
1643 		/* report the route deletion to the routing socket. */
1644 		if (rt != NULL)
1645 			nd6_rtmsg(RTM_DELETE, rt);
1646 
1647 		/*
1648 		 * There might be the same prefix on another interface,
1649 		 * the prefix which could not be on-link just because we have
1650 		 * the interface route (see comments in nd6_prefix_onlink).
1651 		 * If there's one, try to make the prefix on-link on the
1652 		 * interface.
1653 		 */
1654 		LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1655 			if (opr == pr)
1656 				continue;
1657 
1658 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1659 				continue;
1660 
1661 			/*
1662 			 * KAME specific: detached prefixes should not be
1663 			 * on-link.
1664 			 */
1665 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1666 				continue;
1667 
1668 			if (opr->ndpr_plen == pr->ndpr_plen &&
1669 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1670 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1671 				int e;
1672 
1673 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1674 					nd6log((LOG_ERR,
1675 					    "nd6_prefix_offlink: failed to "
1676 					    "recover a prefix %s/%d from %s "
1677 					    "to %s (errno = %d)\n",
1678 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1679 					    opr->ndpr_plen, if_name(ifp),
1680 					    if_name(opr->ndpr_ifp), e));
1681 				}
1682 			}
1683 		}
1684 	} else {
1685 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1686 		nd6log((LOG_ERR,
1687 		    "nd6_prefix_offlink: failed to delete route: "
1688 		    "%s/%d on %s (errno = %d)\n",
1689 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1690 		    error));
1691 	}
1692 
1693 	if (rt != NULL) {
1694 		if (rt->rt_refcnt <= 0) {
1695 			/* XXX: we should free the entry ourselves. */
1696 			rt->rt_refcnt++;
1697 			rtfree(rt);
1698 		}
1699 	}
1700 
1701 	return (error);
1702 }
1703 
1704 static struct in6_ifaddr *
1705 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1706 {
1707 	struct ifnet *ifp = pr->ndpr_ifp;
1708 	struct ifaddr *ifa;
1709 	struct in6_aliasreq ifra;
1710 	struct in6_ifaddr *ia, *ib;
1711 	int error, plen0;
1712 	struct in6_addr mask;
1713 	int prefixlen = pr->ndpr_plen;
1714 	int updateflags;
1715 
1716 	in6_prefixlen2mask(&mask, prefixlen);
1717 
1718 	/*
1719 	 * find a link-local address (will be interface ID).
1720 	 * Is it really mandatory? Theoretically, a global or a site-local
1721 	 * address can be configured without a link-local address, if we
1722 	 * have a unique interface identifier...
1723 	 *
1724 	 * it is not mandatory to have a link-local address, we can generate
1725 	 * interface identifier on the fly.  we do this because:
1726 	 * (1) it should be the easiest way to find interface identifier.
1727 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1728 	 * for multiple addresses on a single interface, and possible shortcut
1729 	 * of DAD.  we omitted DAD for this reason in the past.
1730 	 * (3) a user can prevent autoconfiguration of global address
1731 	 * by removing link-local address by hand (this is partly because we
1732 	 * don't have other way to control the use of IPv6 on an interface.
1733 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1734 	 * (4) it is easier to manage when an interface has addresses
1735 	 * with the same interface identifier, than to have multiple addresses
1736 	 * with different interface identifiers.
1737 	 */
1738 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1739 	if (ifa)
1740 		ib = (struct in6_ifaddr *)ifa;
1741 	else
1742 		return NULL;
1743 
1744 #if 0 /* don't care link local addr state, and always do DAD */
1745 	/* if link-local address is not eligible, do not autoconfigure. */
1746 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1747 		printf("in6_ifadd: link-local address not ready\n");
1748 		return NULL;
1749 	}
1750 #endif
1751 
1752 	/* prefixlen + ifidlen must be equal to 128 */
1753 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1754 	if (prefixlen != plen0) {
1755 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1756 		    "(prefix=%d ifid=%d)\n",
1757 		    if_name(ifp), prefixlen, 128 - plen0));
1758 		return NULL;
1759 	}
1760 
1761 	/* make ifaddr */
1762 
1763 	memset(&ifra, 0, sizeof(ifra));
1764 	/*
1765 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1766 	 * for safety.
1767 	 */
1768 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1769 	sockaddr_in6_init(&ifra.ifra_addr, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1770 	/* prefix */
1771 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1772 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1773 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1774 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1775 
1776 	/* interface ID */
1777 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1778 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1779 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1780 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1781 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1782 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1783 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1784 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1785 
1786 	/* new prefix mask. */
1787 	sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
1788 
1789 	/* lifetimes */
1790 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1791 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1792 
1793 	/* XXX: scope zone ID? */
1794 
1795 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1796 
1797 	/*
1798 	 * Make sure that we do not have this address already.  This should
1799 	 * usually not happen, but we can still see this case, e.g., if we
1800 	 * have manually configured the exact address to be configured.
1801 	 */
1802 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1803 		/* this should be rare enough to make an explicit log */
1804 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1805 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1806 		return (NULL);
1807 	}
1808 
1809 	/*
1810 	 * Allocate ifaddr structure, link into chain, etc.
1811 	 * If we are going to create a new address upon receiving a multicasted
1812 	 * RA, we need to impose a random delay before starting DAD.
1813 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1814 	 */
1815 	updateflags = 0;
1816 	if (mcast)
1817 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1818 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1819 		nd6log((LOG_ERR,
1820 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1821 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1822 		    error));
1823 		return (NULL);	/* ifaddr must not have been allocated. */
1824 	}
1825 
1826 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1827 
1828 	return (ia);		/* this is always non-NULL */
1829 }
1830 
1831 int
1832 in6_tmpifadd(
1833 	const struct in6_ifaddr *ia0, /* corresponding public address */
1834 	int forcegen,
1835 	int dad_delay)
1836 {
1837 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1838 	struct in6_ifaddr *newia, *ia;
1839 	struct in6_aliasreq ifra;
1840 	int i, error;
1841 	int trylimit = 3;	/* XXX: adhoc value */
1842 	int updateflags;
1843 	u_int32_t randid[2];
1844 	u_int32_t vltime0, pltime0;
1845 
1846 	memset(&ifra, 0, sizeof(ifra));
1847 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1848 	ifra.ifra_addr = ia0->ia_addr;
1849 	/* copy prefix mask */
1850 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1851 	/* clear the old IFID */
1852 	for (i = 0; i < 4; i++) {
1853 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1854 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1855 	}
1856 
1857   again:
1858 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1859 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1860 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1861 		    "random IFID\n"));
1862 		return (EINVAL);
1863 	}
1864 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1865 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1866 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1867 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1868 
1869 	/*
1870 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1871 	 * However, we may still have a chance to see collision, because
1872 	 * there may be a time lag between generation of the ID and generation
1873 	 * of the address.  So, we'll do one more sanity check.
1874 	 */
1875 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1876 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1877 		    &ifra.ifra_addr.sin6_addr)) {
1878 			if (trylimit-- == 0) {
1879 				/*
1880 				 * Give up.  Something strange should have
1881 				 * happened.
1882 				 */
1883 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1884 				    "find a unique random IFID\n"));
1885 				return (EEXIST);
1886 			}
1887 			forcegen = 1;
1888 			goto again;
1889 		}
1890 	}
1891 
1892 	/*
1893 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1894          * public address or TEMP_VALID_LIFETIME.
1895 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1896          * of the public address or TEMP_PREFERRED_LIFETIME -
1897          * DESYNC_FACTOR.
1898 	 */
1899 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1900 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1901 		    (ia0->ia6_lifetime.ia6t_vltime -
1902 		    (time_second - ia0->ia6_updatetime));
1903 		if (vltime0 > ip6_temp_valid_lifetime)
1904 			vltime0 = ip6_temp_valid_lifetime;
1905 	} else
1906 		vltime0 = ip6_temp_valid_lifetime;
1907 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1908 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1909 		    (ia0->ia6_lifetime.ia6t_pltime -
1910 		    (time_second - ia0->ia6_updatetime));
1911 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
1912 			pltime0 = ip6_temp_preferred_lifetime -
1913 			    ip6_desync_factor;
1914 		}
1915 	} else
1916 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
1917 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1918 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1919 
1920 	/*
1921 	 * A temporary address is created only if this calculated Preferred
1922 	 * Lifetime is greater than REGEN_ADVANCE time units.
1923 	 */
1924 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
1925 		return (0);
1926 
1927 	/* XXX: scope zone ID? */
1928 
1929 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1930 
1931 	/* allocate ifaddr structure, link into chain, etc. */
1932 	updateflags = 0;
1933 	if (dad_delay)
1934 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1935 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1936 		return (error);
1937 
1938 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1939 	if (newia == NULL) {	/* XXX: can it happen? */
1940 		nd6log((LOG_ERR,
1941 		    "in6_tmpifadd: ifa update succeeded, but we got "
1942 		    "no ifaddr\n"));
1943 		return (EINVAL); /* XXX */
1944 	}
1945 	newia->ia6_ndpr = ia0->ia6_ndpr;
1946 	newia->ia6_ndpr->ndpr_refcnt++;
1947 
1948 	/*
1949 	 * A newly added address might affect the status of other addresses.
1950 	 * XXX: when the temporary address is generated with a new public
1951 	 * address, the onlink check is redundant.  However, it would be safe
1952 	 * to do the check explicitly everywhere a new address is generated,
1953 	 * and, in fact, we surely need the check when we create a new
1954 	 * temporary address due to deprecation of an old temporary address.
1955 	 */
1956 	pfxlist_onlink_check();
1957 
1958 	return (0);
1959 }
1960 
1961 static int
1962 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1963 {
1964 
1965 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
1966 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
1967 		nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
1968 		    "(%d) is greater than valid lifetime(%d)\n",
1969 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
1970 		return (EINVAL);
1971 	}
1972 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1973 		ndpr->ndpr_preferred = 0;
1974 	else
1975 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
1976 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1977 		ndpr->ndpr_expire = 0;
1978 	else
1979 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
1980 
1981 	return 0;
1982 }
1983 
1984 static void
1985 in6_init_address_ltimes(struct nd_prefix *new,
1986     struct in6_addrlifetime *lt6)
1987 {
1988 
1989 	/* Valid lifetime must not be updated unless explicitly specified. */
1990 	/* init ia6t_expire */
1991 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
1992 		lt6->ia6t_expire = 0;
1993 	else {
1994 		lt6->ia6t_expire = time_second;
1995 		lt6->ia6t_expire += lt6->ia6t_vltime;
1996 	}
1997 
1998 	/* init ia6t_preferred */
1999 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2000 		lt6->ia6t_preferred = 0;
2001 	else {
2002 		lt6->ia6t_preferred = time_second;
2003 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2004 	}
2005 }
2006 
2007 /*
2008  * Delete all the routing table entries that use the specified gateway.
2009  * XXX: this function causes search through all entries of routing table, so
2010  * it shouldn't be called when acting as a router.
2011  */
2012 void
2013 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2014 {
2015 	int s = splsoftnet();
2016 
2017 	/* We'll care only link-local addresses */
2018 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2019 		splx(s);
2020 		return;
2021 	}
2022 
2023 	rt_walktree(AF_INET6, rt6_deleteroute, (void *)gateway);
2024 	splx(s);
2025 }
2026 
2027 static int
2028 rt6_deleteroute(struct rtentry *rt, void *arg)
2029 {
2030 #define SIN6(s)	((struct sockaddr_in6 *)s)
2031 	struct in6_addr *gate = (struct in6_addr *)arg;
2032 
2033 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2034 		return (0);
2035 
2036 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr))
2037 		return (0);
2038 
2039 	/*
2040 	 * Do not delete a static route.
2041 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2042 	 * 'cloned' bit instead?
2043 	 */
2044 	if ((rt->rt_flags & RTF_STATIC) != 0)
2045 		return (0);
2046 
2047 	/*
2048 	 * We delete only host route. This means, in particular, we don't
2049 	 * delete default route.
2050 	 */
2051 	if ((rt->rt_flags & RTF_HOST) == 0)
2052 		return (0);
2053 
2054 	return (rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
2055 	    rt_mask(rt), rt->rt_flags, 0));
2056 #undef SIN6
2057 }
2058 
2059 int
2060 nd6_setdefaultiface(int ifindex)
2061 {
2062 	int error = 0;
2063 
2064 	if (ifindex < 0 || if_indexlim <= ifindex)
2065 		return (EINVAL);
2066 	if (ifindex != 0 && !ifindex2ifnet[ifindex])
2067 		return (EINVAL);
2068 
2069 	if (nd6_defifindex != ifindex) {
2070 		nd6_defifindex = ifindex;
2071 		if (nd6_defifindex > 0) {
2072 			nd6_defifp = ifindex2ifnet[nd6_defifindex];
2073 		} else
2074 			nd6_defifp = NULL;
2075 
2076 		/*
2077 		 * Our current implementation assumes one-to-one maping between
2078 		 * interfaces and links, so it would be natural to use the
2079 		 * default interface as the default link.
2080 		 */
2081 		scope6_setdefault(nd6_defifp);
2082 	}
2083 
2084 	return (error);
2085 }
2086