xref: /netbsd-src/sys/netinet6/nd6_rtr.c (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1 /*	$NetBSD: nd6_rtr.c,v 1.80 2009/11/06 20:41:22 dyoung Exp $	*/
2 /*	$KAME: nd6_rtr.c,v 1.95 2001/02/07 08:09:47 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: nd6_rtr.c,v 1.80 2009/11/06 20:41:22 dyoung Exp $");
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/malloc.h>
39 #include <sys/mbuf.h>
40 #include <sys/socket.h>
41 #include <sys/sockio.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/errno.h>
45 #include <sys/ioctl.h>
46 #include <sys/syslog.h>
47 
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/if_dl.h>
51 #include <net/route.h>
52 #include <net/radix.h>
53 
54 #include <netinet/in.h>
55 #include <netinet6/in6_var.h>
56 #include <netinet6/in6_ifattach.h>
57 #include <netinet/ip6.h>
58 #include <netinet6/ip6_var.h>
59 #include <netinet6/nd6.h>
60 #include <netinet/icmp6.h>
61 #include <netinet6/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 nd_ifinfo *ndi = ND_IFINFO(ifp);
117 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
118 	struct nd_router_solicit *nd_rs;
119 	struct in6_addr saddr6 = ip6->ip6_src;
120 	char *lladdr = NULL;
121 	int lladdrlen = 0;
122 	union nd_opts ndopts;
123 
124 	/* If I'm not a router, ignore it. */
125 	if (nd6_accepts_rtadv(ndi) || !ip6_forwarding)
126 		goto freeit;
127 
128 	/* Sanity checks */
129 	if (ip6->ip6_hlim != 255) {
130 		nd6log((LOG_ERR,
131 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
132 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
133 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
134 		goto bad;
135 	}
136 
137 	/*
138 	 * Don't update the neighbor cache, if src = ::.
139 	 * This indicates that the src has no IP address assigned yet.
140 	 */
141 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
142 		goto freeit;
143 
144 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
145 	if (nd_rs == NULL) {
146 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
147 		return;
148 	}
149 
150 	icmp6len -= sizeof(*nd_rs);
151 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
152 	if (nd6_options(&ndopts) < 0) {
153 		nd6log((LOG_INFO,
154 		    "nd6_rs_input: invalid ND option, ignored\n"));
155 		/* nd6_options have incremented stats */
156 		goto freeit;
157 	}
158 
159 	if (ndopts.nd_opts_src_lladdr) {
160 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
161 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
162 	}
163 
164 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
165 		nd6log((LOG_INFO,
166 		    "nd6_rs_input: lladdrlen mismatch for %s "
167 		    "(if %d, RS packet %d)\n",
168 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
169 		goto bad;
170 	}
171 
172 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
173 
174  freeit:
175 	m_freem(m);
176 	return;
177 
178  bad:
179 	ICMP6_STATINC(ICMP6_STAT_BADRS);
180 	m_freem(m);
181 }
182 
183 /*
184  * Receive Router Advertisement Message.
185  *
186  * Based on RFC 2461
187  * TODO: on-link bit on prefix information
188  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
189  */
190 void
191 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
192 {
193 	struct ifnet *ifp = m->m_pkthdr.rcvif;
194 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
195 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
196 	struct nd_router_advert *nd_ra;
197 	struct in6_addr saddr6 = ip6->ip6_src;
198 #if 0
199 	struct in6_addr daddr6 = ip6->ip6_dst;
200 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
201 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
202 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
203 #endif
204 	int mcast = 0;
205 	union nd_opts ndopts;
206 	struct nd_defrouter *dr;
207 
208 	/*
209 	 * We only accept RAs only when
210 	 * the system-wide variable allows the acceptance, and
211 	 * per-interface variable allows RAs on the receiving interface.
212 	 */
213 	if (!nd6_accepts_rtadv(ndi))
214 		goto freeit;
215 
216 	if (ip6->ip6_hlim != 255) {
217 		nd6log((LOG_ERR,
218 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
219 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
220 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
221 		goto bad;
222 	}
223 
224 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
225 		nd6log((LOG_ERR,
226 		    "nd6_ra_input: src %s is not link-local\n",
227 		    ip6_sprintf(&saddr6)));
228 		goto bad;
229 	}
230 
231 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
232 	if (nd_ra == NULL) {
233 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
234 		return;
235 	}
236 
237 	icmp6len -= sizeof(*nd_ra);
238 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
239 	if (nd6_options(&ndopts) < 0) {
240 		nd6log((LOG_INFO,
241 		    "nd6_ra_input: invalid ND option, ignored\n"));
242 		/* nd6_options have incremented stats */
243 		goto freeit;
244 	}
245 
246     {
247 	struct nd_defrouter drtr;
248 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
249 
250 	/* remember if this is a multicasted advertisement */
251 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
252 		mcast = 1;
253 
254 	memset(&drtr, 0, sizeof(drtr));
255 	drtr.rtaddr = saddr6;
256 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
257 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
258 	drtr.expire = time_second + drtr.rtlifetime;
259 	drtr.ifp = ifp;
260 	/* unspecified or not? (RFC 2461 6.3.4) */
261 	if (advreachable) {
262 		NTOHL(advreachable);
263 		if (advreachable <= MAX_REACHABLE_TIME &&
264 		    ndi->basereachable != advreachable) {
265 			ndi->basereachable = advreachable;
266 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
267 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
268 		}
269 	}
270 	if (nd_ra->nd_ra_retransmit)
271 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
272 	if (nd_ra->nd_ra_curhoplimit)
273 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
274 	dr = defrtrlist_update(&drtr);
275     }
276 
277 	/*
278 	 * prefix
279 	 */
280 	if (ndopts.nd_opts_pi) {
281 		struct nd_opt_hdr *pt;
282 		struct nd_opt_prefix_info *pi = NULL;
283 		struct nd_prefixctl pr;
284 
285 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
286 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
287 		     pt = (struct nd_opt_hdr *)((char *)pt +
288 						(pt->nd_opt_len << 3))) {
289 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
290 				continue;
291 			pi = (struct nd_opt_prefix_info *)pt;
292 
293 			if (pi->nd_opt_pi_len != 4) {
294 				nd6log((LOG_INFO,
295 				    "nd6_ra_input: invalid option "
296 				    "len %d for prefix information option, "
297 				    "ignored\n", pi->nd_opt_pi_len));
298 				continue;
299 			}
300 
301 			if (128 < pi->nd_opt_pi_prefix_len) {
302 				nd6log((LOG_INFO,
303 				    "nd6_ra_input: invalid prefix "
304 				    "len %d for prefix information option, "
305 				    "ignored\n", pi->nd_opt_pi_prefix_len));
306 				continue;
307 			}
308 
309 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
310 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
311 				nd6log((LOG_INFO,
312 				    "nd6_ra_input: invalid prefix "
313 				    "%s, ignored\n",
314 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
315 				continue;
316 			}
317 
318 			memset(&pr, 0, sizeof(pr));
319 			sockaddr_in6_init(&pr.ndpr_prefix,
320 			    &pi->nd_opt_pi_prefix, 0, 0, 0);
321 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
322 
323 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
324 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
325 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
326 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
327 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
328 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
329 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
330 
331 			(void)prelist_update(&pr, dr, m, mcast);
332 		}
333 	}
334 
335 	/*
336 	 * MTU
337 	 */
338 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
339 		u_long mtu;
340 		u_long maxmtu;
341 
342 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
343 
344 		/* lower bound */
345 		if (mtu < IPV6_MMTU) {
346 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
347 			    "mtu=%lu sent from %s, ignoring\n",
348 			    mtu, ip6_sprintf(&ip6->ip6_src)));
349 			goto skip;
350 		}
351 
352 		/* upper bound */
353 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
354 		    ? ndi->maxmtu : ifp->if_mtu;
355 		if (mtu <= maxmtu) {
356 			int change = (ndi->linkmtu != mtu);
357 
358 			ndi->linkmtu = mtu;
359 			if (change) /* in6_maxmtu may change */
360 				in6_setmaxmtu();
361 		} else {
362 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
363 			    "mtu=%lu sent from %s; "
364 			    "exceeds maxmtu %lu, ignoring\n",
365 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
366 		}
367 	}
368 
369  skip:
370 
371 	/*
372 	 * Source link layer address
373 	 */
374     {
375 	char *lladdr = NULL;
376 	int lladdrlen = 0;
377 
378 	if (ndopts.nd_opts_src_lladdr) {
379 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
380 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
381 	}
382 
383 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
384 		nd6log((LOG_INFO,
385 		    "nd6_ra_input: lladdrlen mismatch for %s "
386 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
387 		    ifp->if_addrlen, lladdrlen - 2));
388 		goto bad;
389 	}
390 
391 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
392 
393 	/*
394 	 * Installing a link-layer address might change the state of the
395 	 * router's neighbor cache, which might also affect our on-link
396 	 * detection of adveritsed prefixes.
397 	 */
398 	pfxlist_onlink_check();
399     }
400 
401  freeit:
402 	m_freem(m);
403 	return;
404 
405  bad:
406 	ICMP6_STATINC(ICMP6_STAT_BADRA);
407 	m_freem(m);
408 }
409 
410 /*
411  * default router list processing sub routines
412  */
413 
414 /* tell the change to user processes watching the routing socket. */
415 static void
416 nd6_rtmsg(int cmd, struct rtentry *rt)
417 {
418 	struct rt_addrinfo info;
419 
420 	memset((void *)&info, 0, sizeof(info));
421 	info.rti_info[RTAX_DST] = rt_getkey(rt);
422 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
423 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
424 	if (rt->rt_ifp) {
425 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
426 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
427 	}
428 
429 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
430 }
431 
432 void
433 defrouter_addreq(struct nd_defrouter *new)
434 {
435 	union {
436 		struct sockaddr_in6 sin6;
437 		struct sockaddr sa;
438 	} def, mask, gate;
439 	struct rtentry *newrt = NULL;
440 	int s;
441 	int error;
442 
443 	memset(&def, 0, sizeof(def));
444 	memset(&mask, 0, sizeof(mask));
445 	memset(&gate, 0,sizeof(gate)); /* for safety */
446 
447 	def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len =
448 	    sizeof(struct sockaddr_in6);
449 	def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6;
450 	gate.sin6.sin6_addr = new->rtaddr;
451 #ifndef SCOPEDROUTING
452 	gate.sin6.sin6_scope_id = 0;	/* XXX */
453 #endif
454 
455 	s = splsoftnet();
456 	error = rtrequest(RTM_ADD, &def.sa, &gate.sa, &mask.sa,
457 	    RTF_GATEWAY, &newrt);
458 	if (newrt) {
459 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
460 		newrt->rt_refcnt--;
461 	}
462 	if (error == 0)
463 		new->installed = 1;
464 	splx(s);
465 	return;
466 }
467 
468 struct nd_defrouter *
469 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
470 {
471 	struct nd_defrouter *dr;
472 
473 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
474 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
475 			break;
476 	}
477 
478 	return dr;		/* search failed */
479 }
480 
481 void
482 defrtrlist_del(struct nd_defrouter *dr)
483 {
484 	struct nd_ifinfo *ndi = ND_IFINFO(dr->ifp);
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 	/* XXX: better condition? */
493 	if (!ip6_forwarding && nd6_accepts_rtadv(ndi))
494 		rt6_flush(&dr->rtaddr, dr->ifp);
495 
496 	if (dr->installed) {
497 		deldr = dr;
498 		defrouter_delreq(dr);
499 	}
500 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
501 
502 	/*
503 	 * Also delete all the pointers to the router in each prefix lists.
504 	 */
505 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
506 		struct nd_pfxrouter *pfxrtr;
507 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
508 			pfxrtr_del(pfxrtr);
509 	}
510 	pfxlist_onlink_check();
511 
512 	/*
513 	 * If the router is the primary one, choose a new one.
514 	 * Note that defrouter_select() will remove the current gateway
515 	 * from the routing table.
516 	 */
517 	if (deldr)
518 		defrouter_select();
519 
520 	free(dr, M_IP6NDP);
521 }
522 
523 /*
524  * Remove the default route for a given router.
525  * This is just a subroutine function for defrouter_select(), and should
526  * not be called from anywhere else.
527  */
528 static void
529 defrouter_delreq(struct nd_defrouter *dr)
530 {
531 	union {
532 		struct sockaddr_in6 sin6;
533 		struct sockaddr sa;
534 	} def, mask, gw;
535 	struct rtentry *oldrt = NULL;
536 
537 #ifdef DIAGNOSTIC
538 	if (dr == NULL)
539 		panic("dr == NULL in defrouter_delreq");
540 #endif
541 
542 	memset(&def, 0, sizeof(def));
543 	memset(&mask, 0, sizeof(mask));
544 	memset(&gw, 0, sizeof(gw));	/* for safety */
545 
546 	def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len =
547 	    sizeof(struct sockaddr_in6);
548 	def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6;
549 	gw.sin6.sin6_addr = dr->rtaddr;
550 #ifndef SCOPEDROUTING
551 	gw.sin6.sin6_scope_id = 0;	/* XXX */
552 #endif
553 
554 	rtrequest(RTM_DELETE, &def.sa, &gw.sa, &mask.sa, RTF_GATEWAY, &oldrt);
555 	if (oldrt) {
556 		nd6_rtmsg(RTM_DELETE, oldrt);
557 		if (oldrt->rt_refcnt <= 0) {
558 			/*
559 			 * XXX: borrowed from the RTM_DELETE case of
560 			 * rtrequest().
561 			 */
562 			oldrt->rt_refcnt++;
563 			rtfree(oldrt);
564 		}
565 	}
566 
567 	dr->installed = 0;
568 }
569 
570 /*
571  * remove all default routes from default router list
572  */
573 void
574 defrouter_reset(void)
575 {
576 	struct nd_defrouter *dr;
577 
578 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
579 	     dr = TAILQ_NEXT(dr, dr_entry))
580 		defrouter_delreq(dr);
581 
582 	/*
583 	 * XXX should we also nuke any default routers in the kernel, by
584 	 * going through them by rtalloc1()?
585 	 */
586 }
587 
588 /*
589  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
590  * draft-ietf-ipngwg-router-selection:
591  * 1) Routers that are reachable or probably reachable should be preferred.
592  *    If we have more than one (probably) reachable router, prefer ones
593  *    with the highest router preference.
594  * 2) When no routers on the list are known to be reachable or
595  *    probably reachable, routers SHOULD be selected in a round-robin
596  *    fashion, regardless of router preference values.
597  * 3) If the Default Router List is empty, assume that all
598  *    destinations are on-link.
599  *
600  * We assume nd_defrouter is sorted by router preference value.
601  * Since the code below covers both with and without router preference cases,
602  * we do not need to classify the cases by ifdef.
603  *
604  * At this moment, we do not try to install more than one default router,
605  * even when the multipath routing is available, because we're not sure about
606  * the benefits for stub hosts comparing to the risk of making the code
607  * complicated and the possibility of introducing bugs.
608  */
609 void
610 defrouter_select(void)
611 {
612 	struct nd_ifinfo *ndi;
613 	int s = splsoftnet();
614 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
615 	struct rtentry *rt = NULL;
616 	struct llinfo_nd6 *ln = NULL;
617 
618 	/*
619 	 * This function should be called only when acting as an autoconfigured
620 	 * host.  Although the remaining part of this function is not effective
621 	 * if the node is not an autoconfigured host, we explicitly exclude
622 	 * such cases here for safety.
623 	 */
624 	if (ip6_forwarding) {
625 		nd6log((LOG_WARNING,
626 		    "defrouter_select: called unexpectedly (forwarding=%d, "
627 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
628 		splx(s);
629 		return;
630 	}
631 
632 	/*
633 	 * Let's handle easy case (3) first:
634 	 * If default router list is empty, there's nothing to be done.
635 	 */
636 	if (!TAILQ_FIRST(&nd_defrouter)) {
637 		splx(s);
638 		return;
639 	}
640 
641 	/*
642 	 * Search for a (probably) reachable router from the list.
643 	 * We just pick up the first reachable one (if any), assuming that
644 	 * the ordering rule of the list described in defrtrlist_update().
645 	 */
646 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
647 	     dr = TAILQ_NEXT(dr, dr_entry)) {
648 		ndi = ND_IFINFO(dr->ifp);
649 		if (nd6_accepts_rtadv(ndi))
650 			continue;
651 
652 		if (selected_dr == NULL &&
653 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) != NULL &&
654 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) != NULL &&
655 		    ND6_IS_LLINFO_PROBREACH(ln)) {
656 			selected_dr = dr;
657 		}
658 
659 		if (dr->installed && !installed_dr)
660 			installed_dr = dr;
661 		else if (dr->installed && installed_dr) {
662 			/* this should not happen.  warn for diagnosis. */
663 			log(LOG_ERR, "defrouter_select: more than one router"
664 			    " is installed\n");
665 		}
666 	}
667 	/*
668 	 * If none of the default routers was found to be reachable,
669 	 * round-robin the list regardless of preference.
670 	 * Otherwise, if we have an installed router, check if the selected
671 	 * (reachable) router should really be preferred to the installed one.
672 	 * We only prefer the new router when the old one is not reachable
673 	 * or when the new one has a really higher preference value.
674 	 */
675 	if (selected_dr == NULL) {
676 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
677 			selected_dr = TAILQ_FIRST(&nd_defrouter);
678 		else
679 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
680 	} else if (installed_dr &&
681 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
682 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
683 	    ND6_IS_LLINFO_PROBREACH(ln) &&
684 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
685 		selected_dr = installed_dr;
686 	}
687 
688 	/*
689 	 * If the selected router is different than the installed one,
690 	 * remove the installed router and install the selected one.
691 	 * Note that the selected router is never NULL here.
692 	 */
693 	if (installed_dr != selected_dr) {
694 		if (installed_dr)
695 			defrouter_delreq(installed_dr);
696 		defrouter_addreq(selected_dr);
697 	}
698 
699 	splx(s);
700 	return;
701 }
702 
703 /*
704  * for default router selection
705  * regards router-preference field as a 2-bit signed integer
706  */
707 static int
708 rtpref(struct nd_defrouter *dr)
709 {
710 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
711 	case ND_RA_FLAG_RTPREF_HIGH:
712 		return (RTPREF_HIGH);
713 	case ND_RA_FLAG_RTPREF_MEDIUM:
714 	case ND_RA_FLAG_RTPREF_RSV:
715 		return (RTPREF_MEDIUM);
716 	case ND_RA_FLAG_RTPREF_LOW:
717 		return (RTPREF_LOW);
718 	default:
719 		/*
720 		 * This case should never happen.  If it did, it would mean a
721 		 * serious bug of kernel internal.  We thus always bark here.
722 		 * Or, can we even panic?
723 		 */
724 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
725 		return (RTPREF_INVALID);
726 	}
727 	/* NOTREACHED */
728 }
729 
730 static struct nd_defrouter *
731 defrtrlist_update(struct nd_defrouter *new)
732 {
733 	struct nd_defrouter *dr, *n;
734 	int s = splsoftnet();
735 
736 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
737 		/* entry exists */
738 		if (new->rtlifetime == 0) {
739 			defrtrlist_del(dr);
740 			dr = NULL;
741 		} else {
742 			int oldpref = rtpref(dr);
743 
744 			/* override */
745 			dr->flags = new->flags; /* xxx flag check */
746 			dr->rtlifetime = new->rtlifetime;
747 			dr->expire = new->expire;
748 
749 			/*
750 			 * If the preference does not change, there's no need
751 			 * to sort the entries.
752 			 */
753 			if (rtpref(new) == oldpref) {
754 				splx(s);
755 				return (dr);
756 			}
757 
758 			/*
759 			 * preferred router may be changed, so relocate
760 			 * this router.
761 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
762 			 * However, since defrtrlist_del() has many side
763 			 * effects, we intentionally do so here.
764 			 * defrouter_select() below will handle routing
765 			 * changes later.
766 			 */
767 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
768 			n = dr;
769 			goto insert;
770 		}
771 		splx(s);
772 		return (dr);
773 	}
774 
775 	/* entry does not exist */
776 	if (new->rtlifetime == 0) {
777 		splx(s);
778 		return (NULL);
779 	}
780 
781 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
782 	if (n == NULL) {
783 		splx(s);
784 		return (NULL);
785 	}
786 	memset(n, 0, sizeof(*n));
787 	*n = *new;
788 
789 insert:
790 	/*
791 	 * Insert the new router in the Default Router List;
792 	 * The Default Router List should be in the descending order
793 	 * of router-preferece.  Routers with the same preference are
794 	 * sorted in the arriving time order.
795 	 */
796 
797 	/* insert at the end of the group */
798 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
799 	     dr = TAILQ_NEXT(dr, dr_entry)) {
800 		if (rtpref(n) > rtpref(dr))
801 			break;
802 	}
803 	if (dr)
804 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
805 	else
806 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
807 
808 	defrouter_select();
809 
810 	splx(s);
811 
812 	return (n);
813 }
814 
815 static struct nd_pfxrouter *
816 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
817 {
818 	struct nd_pfxrouter *search;
819 
820 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
821 		if (search->router == dr)
822 			break;
823 	}
824 
825 	return (search);
826 }
827 
828 static void
829 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
830 {
831 	struct nd_pfxrouter *new;
832 
833 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT|M_ZERO);
834 	if (new == NULL)
835 		return;
836 	new->router = dr;
837 
838 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
839 
840 	pfxlist_onlink_check();
841 }
842 
843 static void
844 pfxrtr_del(struct nd_pfxrouter *pfr)
845 {
846 	LIST_REMOVE(pfr, pfr_entry);
847 	free(pfr, M_IP6NDP);
848 }
849 
850 struct nd_prefix *
851 nd6_prefix_lookup(struct nd_prefixctl *key)
852 {
853 	struct nd_prefix *search;
854 
855 	LIST_FOREACH(search, &nd_prefix, ndpr_entry) {
856 		if (key->ndpr_ifp == search->ndpr_ifp &&
857 		    key->ndpr_plen == search->ndpr_plen &&
858 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
859 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
860 			break;
861 		}
862 	}
863 
864 	return (search);
865 }
866 
867 int
868 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
869 	struct nd_prefix **newp)
870 {
871 	struct nd_prefix *new = NULL;
872 	int i, s;
873 	int error;
874 
875 	error = 0;
876 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT|M_ZERO);
877 	if (new == NULL)
878 		return ENOMEM;
879 	new->ndpr_ifp = pr->ndpr_ifp;
880 	new->ndpr_prefix = pr->ndpr_prefix;
881 	new->ndpr_plen = pr->ndpr_plen;
882 	new->ndpr_vltime = pr->ndpr_vltime;
883 	new->ndpr_pltime = pr->ndpr_pltime;
884 	new->ndpr_flags = pr->ndpr_flags;
885 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
886 		free(new, M_IP6NDP);
887 		return(error);
888 	}
889 	new->ndpr_lastupdate = time_second;
890 	if (newp != NULL)
891 		*newp = new;
892 
893 	/* initialization */
894 	LIST_INIT(&new->ndpr_advrtrs);
895 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
896 	/* make prefix in the canonical form */
897 	for (i = 0; i < 4; i++)
898 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
899 		    new->ndpr_mask.s6_addr32[i];
900 
901 	s = splsoftnet();
902 	/* link ndpr_entry to nd_prefix list */
903 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
904 	splx(s);
905 
906 	/* ND_OPT_PI_FLAG_ONLINK processing */
907 	if (new->ndpr_raf_onlink) {
908 		int e;
909 
910 		if ((e = nd6_prefix_onlink(new)) != 0) {
911 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
912 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
913 			    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
914 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
915 			/* proceed anyway. XXX: is it correct? */
916 		}
917 	}
918 
919 	if (dr)
920 		pfxrtr_add(new, dr);
921 
922 	return 0;
923 }
924 
925 void
926 prelist_remove(struct nd_prefix *pr)
927 {
928 	struct nd_pfxrouter *pfr, *next;
929 	int e, s;
930 
931 	/* make sure to invalidate the prefix until it is really freed. */
932 	pr->ndpr_vltime = 0;
933 	pr->ndpr_pltime = 0;
934 #if 0
935 	/*
936 	 * Though these flags are now meaningless, we'd rather keep the value
937 	 * not to confuse users when executing "ndp -p".
938 	 */
939 	pr->ndpr_raf_onlink = 0;
940 	pr->ndpr_raf_auto = 0;
941 #endif
942 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
943 	    (e = nd6_prefix_offlink(pr)) != 0) {
944 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
945 		    "on %s, errno=%d\n",
946 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
947 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
948 		/* what should we do? */
949 	}
950 
951 	if (pr->ndpr_refcnt > 0)
952 		return;		/* notice here? */
953 
954 	s = splsoftnet();
955 	/* unlink ndpr_entry from nd_prefix list */
956 	LIST_REMOVE(pr, ndpr_entry);
957 
958 	/* free list of routers that adversed the prefix */
959 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
960 		next = LIST_NEXT(pfr, pfr_entry);
961 
962 		free(pfr, M_IP6NDP);
963 	}
964 	splx(s);
965 
966 	free(pr, M_IP6NDP);
967 
968 	pfxlist_onlink_check();
969 }
970 
971 static int
972 prelist_update(struct nd_prefixctl *new,
973 	struct nd_defrouter *dr, /* may be NULL */
974 	struct mbuf *m,
975 	int mcast)
976 {
977 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
978 	struct ifaddr *ifa;
979 	struct ifnet *ifp = new->ndpr_ifp;
980 	struct nd_prefix *pr;
981 	int s = splsoftnet();
982 	int error = 0;
983 	int newprefix = 0;
984 	int auth;
985 	struct in6_addrlifetime lt6_tmp;
986 
987 	auth = 0;
988 	if (m) {
989 		/*
990 		 * Authenticity for NA consists authentication for
991 		 * both IP header and IP datagrams, doesn't it ?
992 		 */
993 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
994 		auth = (m->m_flags & M_AUTHIPHDR
995 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
996 #endif
997 	}
998 
999 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1000 		/*
1001 		 * nd6_prefix_lookup() ensures that pr and new have the same
1002 		 * prefix on a same interface.
1003 		 */
1004 
1005 		/*
1006 		 * Update prefix information.  Note that the on-link (L) bit
1007 		 * and the autonomous (A) bit should NOT be changed from 1
1008 		 * to 0.
1009 		 */
1010 		if (new->ndpr_raf_onlink == 1)
1011 			pr->ndpr_raf_onlink = 1;
1012 		if (new->ndpr_raf_auto == 1)
1013 			pr->ndpr_raf_auto = 1;
1014 		if (new->ndpr_raf_onlink) {
1015 			pr->ndpr_vltime = new->ndpr_vltime;
1016 			pr->ndpr_pltime = new->ndpr_pltime;
1017 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1018 			pr->ndpr_lastupdate = time_second;
1019 		}
1020 
1021 		if (new->ndpr_raf_onlink &&
1022 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1023 			int e;
1024 
1025 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1026 				nd6log((LOG_ERR,
1027 				    "prelist_update: failed to make "
1028 				    "the prefix %s/%d on-link on %s "
1029 				    "(errno=%d)\n",
1030 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1031 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1032 				/* proceed anyway. XXX: is it correct? */
1033 			}
1034 		}
1035 
1036 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1037 			pfxrtr_add(pr, dr);
1038 	} else {
1039 		struct nd_prefix *newpr = NULL;
1040 
1041 		newprefix = 1;
1042 
1043 		if (new->ndpr_vltime == 0)
1044 			goto end;
1045 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1046 			goto end;
1047 
1048 		error = nd6_prelist_add(new, dr, &newpr);
1049 		if (error != 0 || newpr == NULL) {
1050 			nd6log((LOG_NOTICE, "prelist_update: "
1051 			    "nd6_prelist_add failed for %s/%d on %s "
1052 			    "errno=%d, returnpr=%p\n",
1053 			    ip6_sprintf(&new->ndpr_prefix.sin6_addr),
1054 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1055 			    error, newpr));
1056 			goto end; /* we should just give up in this case. */
1057 		}
1058 
1059 		/*
1060 		 * XXX: from the ND point of view, we can ignore a prefix
1061 		 * with the on-link bit being zero.  However, we need a
1062 		 * prefix structure for references from autoconfigured
1063 		 * addresses.  Thus, we explicitly make sure that the prefix
1064 		 * itself expires now.
1065 		 */
1066 		if (newpr->ndpr_raf_onlink == 0) {
1067 			newpr->ndpr_vltime = 0;
1068 			newpr->ndpr_pltime = 0;
1069 			in6_init_prefix_ltimes(newpr);
1070 		}
1071 
1072 		pr = newpr;
1073 	}
1074 
1075 	/*
1076 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1077 	 * Note that pr must be non NULL at this point.
1078 	 */
1079 
1080 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1081 	if (!new->ndpr_raf_auto)
1082 		goto end;
1083 
1084 	/*
1085 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1086 	 * nd6_ra_input.
1087 	 */
1088 
1089 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1090 	if (new->ndpr_pltime > new->ndpr_vltime) {
1091 		error = EINVAL;	/* XXX: won't be used */
1092 		goto end;
1093 	}
1094 
1095 	/*
1096 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1097 	 * an address configured by stateless autoconfiguration already in the
1098 	 * list of addresses associated with the interface, and the Valid
1099 	 * Lifetime is not 0, form an address.  We first check if we have
1100 	 * a matching prefix.
1101 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1102 	 * consider autoconfigured addresses while RFC2462 simply said
1103 	 * "address".
1104 	 */
1105 	IFADDR_FOREACH(ifa, ifp) {
1106 		struct in6_ifaddr *ifa6;
1107 		u_int32_t remaininglifetime;
1108 
1109 		if (ifa->ifa_addr->sa_family != AF_INET6)
1110 			continue;
1111 
1112 		ifa6 = (struct in6_ifaddr *)ifa;
1113 
1114 		/*
1115 		 * We only consider autoconfigured addresses as per rfc2462bis.
1116 		 */
1117 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1118 			continue;
1119 
1120 		/*
1121 		 * Spec is not clear here, but I believe we should concentrate
1122 		 * on unicast (i.e. not anycast) addresses.
1123 		 * XXX: other ia6_flags? detached or duplicated?
1124 		 */
1125 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1126 			continue;
1127 
1128 		/*
1129 		 * Ignore the address if it is not associated with a prefix
1130 		 * or is associated with a prefix that is different from this
1131 		 * one.  (pr is never NULL here)
1132 		 */
1133 		if (ifa6->ia6_ndpr != pr)
1134 			continue;
1135 
1136 		if (ia6_match == NULL) /* remember the first one */
1137 			ia6_match = ifa6;
1138 
1139 		/*
1140 		 * An already autoconfigured address matched.  Now that we
1141 		 * are sure there is at least one matched address, we can
1142 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1143 		 * "two hours" rule and the privacy extension.
1144 		 * We apply some clarifications in rfc2462bis:
1145 		 * - use remaininglifetime instead of storedlifetime as a
1146 		 *   variable name
1147 		 * - remove the dead code in the "two-hour" rule
1148 		 */
1149 #define TWOHOUR		(120*60)
1150 		lt6_tmp = ifa6->ia6_lifetime;
1151 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1152 			remaininglifetime = ND6_INFINITE_LIFETIME;
1153 		else if (time_second - ifa6->ia6_updatetime >
1154 			 lt6_tmp.ia6t_vltime) {
1155 			/*
1156 			 * The case of "invalid" address.  We should usually
1157 			 * not see this case.
1158 			 */
1159 			remaininglifetime = 0;
1160 		} else
1161 			remaininglifetime = lt6_tmp.ia6t_vltime -
1162 			    (time_second - ifa6->ia6_updatetime);
1163 
1164 		/* when not updating, keep the current stored lifetime. */
1165 		lt6_tmp.ia6t_vltime = remaininglifetime;
1166 
1167 		if (TWOHOUR < new->ndpr_vltime ||
1168 		    remaininglifetime < new->ndpr_vltime) {
1169 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1170 		} else if (remaininglifetime <= TWOHOUR) {
1171 			if (auth)
1172 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1173 		} else {
1174 			/*
1175 			 * new->ndpr_vltime <= TWOHOUR &&
1176 			 * TWOHOUR < remaininglifetime
1177 			 */
1178 			lt6_tmp.ia6t_vltime = TWOHOUR;
1179 		}
1180 
1181 		/* The 2 hour rule is not imposed for preferred lifetime. */
1182 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1183 
1184 		in6_init_address_ltimes(pr, &lt6_tmp);
1185 
1186 		/*
1187 		 * We need to treat lifetimes for temporary addresses
1188 		 * differently, according to
1189 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1190 		 * we only update the lifetimes when they are in the maximum
1191 		 * intervals.
1192 		 */
1193 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1194 			u_int32_t maxvltime, maxpltime;
1195 
1196 			if (ip6_temp_valid_lifetime >
1197 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1198 			    ip6_desync_factor)) {
1199 				maxvltime = ip6_temp_valid_lifetime -
1200 				    (time_second - ifa6->ia6_createtime) -
1201 				    ip6_desync_factor;
1202 			} else
1203 				maxvltime = 0;
1204 			if (ip6_temp_preferred_lifetime >
1205 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1206 			    ip6_desync_factor)) {
1207 				maxpltime = ip6_temp_preferred_lifetime -
1208 				    (time_second - ifa6->ia6_createtime) -
1209 				    ip6_desync_factor;
1210 			} else
1211 				maxpltime = 0;
1212 
1213 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1214 			    lt6_tmp.ia6t_vltime > maxvltime) {
1215 				lt6_tmp.ia6t_vltime = maxvltime;
1216 			}
1217 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1218 			    lt6_tmp.ia6t_pltime > maxpltime) {
1219 				lt6_tmp.ia6t_pltime = maxpltime;
1220 			}
1221 		}
1222 
1223 		ifa6->ia6_lifetime = lt6_tmp;
1224 		ifa6->ia6_updatetime = time_second;
1225 	}
1226 	if (ia6_match == NULL && new->ndpr_vltime) {
1227 		int ifidlen;
1228 
1229 		/*
1230 		 * 5.5.3 (d) (continued)
1231 		 * No address matched and the valid lifetime is non-zero.
1232 		 * Create a new address.
1233 		 */
1234 
1235 		/*
1236 		 * Prefix Length check:
1237 		 * If the sum of the prefix length and interface identifier
1238 		 * length does not equal 128 bits, the Prefix Information
1239 		 * option MUST be ignored.  The length of the interface
1240 		 * identifier is defined in a separate link-type specific
1241 		 * document.
1242 		 */
1243 		ifidlen = in6_if2idlen(ifp);
1244 		if (ifidlen < 0) {
1245 			/* this should not happen, so we always log it. */
1246 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1247 			    if_name(ifp));
1248 			goto end;
1249 		}
1250 		if (ifidlen + pr->ndpr_plen != 128) {
1251 			nd6log((LOG_INFO,
1252 			    "prelist_update: invalid prefixlen "
1253 			    "%d for %s, ignored\n",
1254 			    pr->ndpr_plen, if_name(ifp)));
1255 			goto end;
1256 		}
1257 
1258 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1259 			/*
1260 			 * note that we should use pr (not new) for reference.
1261 			 */
1262 			pr->ndpr_refcnt++;
1263 			ia6->ia6_ndpr = pr;
1264 
1265 			/*
1266 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
1267 			 * When a new public address is created as described
1268 			 * in RFC2462, also create a new temporary address.
1269 			 *
1270 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
1271 			 * When an interface connects to a new link, a new
1272 			 * randomized interface identifier should be generated
1273 			 * immediately together with a new set of temporary
1274 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1275 			 * in6_tmpifadd().
1276 			 */
1277 			if (ip6_use_tempaddr) {
1278 				int e;
1279 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1280 					nd6log((LOG_NOTICE, "prelist_update: "
1281 					    "failed to create a temporary "
1282 					    "address, errno=%d\n",
1283 					    e));
1284 				}
1285 			}
1286 
1287 			/*
1288 			 * A newly added address might affect the status
1289 			 * of other addresses, so we check and update it.
1290 			 * XXX: what if address duplication happens?
1291 			 */
1292 			pfxlist_onlink_check();
1293 		} else {
1294 			/* just set an error. do not bark here. */
1295 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1296 		}
1297 	}
1298 
1299  end:
1300 	splx(s);
1301 	return error;
1302 }
1303 
1304 /*
1305  * A supplement function used in the on-link detection below;
1306  * detect if a given prefix has a (probably) reachable advertising router.
1307  * XXX: lengthy function name...
1308  */
1309 static struct nd_pfxrouter *
1310 find_pfxlist_reachable_router(struct nd_prefix *pr)
1311 {
1312 	struct nd_pfxrouter *pfxrtr;
1313 	struct rtentry *rt;
1314 	struct llinfo_nd6 *ln;
1315 
1316 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1317 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1318 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1319 		    pfxrtr->router->ifp)) &&
1320 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1321 		    ND6_IS_LLINFO_PROBREACH(ln))
1322 			break;	/* found */
1323 	}
1324 
1325 	return (pfxrtr);
1326 }
1327 
1328 /*
1329  * Check if each prefix in the prefix list has at least one available router
1330  * that advertised the prefix (a router is "available" if its neighbor cache
1331  * entry is reachable or probably reachable).
1332  * If the check fails, the prefix may be off-link, because, for example,
1333  * we have moved from the network but the lifetime of the prefix has not
1334  * expired yet.  So we should not use the prefix if there is another prefix
1335  * that has an available router.
1336  * But, if there is no prefix that has an available router, we still regards
1337  * all the prefixes as on-link.  This is because we can't tell if all the
1338  * routers are simply dead or if we really moved from the network and there
1339  * is no router around us.
1340  */
1341 void
1342 pfxlist_onlink_check(void)
1343 {
1344 	struct nd_prefix *pr;
1345 	struct in6_ifaddr *ifa;
1346 	struct nd_defrouter *dr;
1347 	struct nd_pfxrouter *pfxrtr = NULL;
1348 
1349 	/*
1350 	 * Check if there is a prefix that has a reachable advertising
1351 	 * router.
1352 	 */
1353 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1354 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1355 			break;
1356 	}
1357 	/*
1358 	 * If we have no such prefix, check whether we still have a router
1359 	 * that does not advertise any prefixes.
1360 	 */
1361 	if (pr == NULL) {
1362 		TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1363 			struct nd_prefix *pr0;
1364 
1365 			LIST_FOREACH(pr0, &nd_prefix, ndpr_entry) {
1366 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1367 					break;
1368 			}
1369 			if (pfxrtr)
1370 				break;
1371 		}
1372 	}
1373 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
1374 		/*
1375 		 * There is at least one prefix that has a reachable router,
1376 		 * or at least a router which probably does not advertise
1377 		 * any prefixes.  The latter would be the case when we move
1378 		 * to a new link where we have a router that does not provide
1379 		 * prefixes and we configure an address by hand.
1380 		 * Detach prefixes which have no reachable advertising
1381 		 * router, and attach other prefixes.
1382 		 */
1383 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1384 			/* XXX: a link-local prefix should never be detached */
1385 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1386 				continue;
1387 
1388 			/*
1389 			 * we aren't interested in prefixes without the L bit
1390 			 * set.
1391 			 */
1392 			if (pr->ndpr_raf_onlink == 0)
1393 				continue;
1394 
1395 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1396 			    find_pfxlist_reachable_router(pr) == NULL)
1397 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1398 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1399 			    find_pfxlist_reachable_router(pr) != 0)
1400 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1401 		}
1402 	} else {
1403 		/* there is no prefix that has a reachable router */
1404 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1405 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1406 				continue;
1407 
1408 			if (pr->ndpr_raf_onlink == 0)
1409 				continue;
1410 
1411 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1412 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1413 		}
1414 	}
1415 
1416 	/*
1417 	 * Remove each interface route associated with a (just) detached
1418 	 * prefix, and reinstall the interface route for a (just) attached
1419 	 * prefix.  Note that all attempt of reinstallation does not
1420 	 * necessarily success, when a same prefix is shared among multiple
1421 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1422 	 * so we don't have to care about them.
1423 	 */
1424 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1425 		int e;
1426 
1427 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1428 			continue;
1429 
1430 		if (pr->ndpr_raf_onlink == 0)
1431 			continue;
1432 
1433 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1434 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1435 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1436 				nd6log((LOG_ERR,
1437 				    "pfxlist_onlink_check: failed to "
1438 				    "make %s/%d offlink, errno=%d\n",
1439 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1440 				    pr->ndpr_plen, e));
1441 			}
1442 		}
1443 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1444 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1445 		    pr->ndpr_raf_onlink) {
1446 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1447 				nd6log((LOG_ERR,
1448 				    "pfxlist_onlink_check: failed to "
1449 				    "make %s/%d onlink, errno=%d\n",
1450 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1451 				    pr->ndpr_plen, e));
1452 			}
1453 		}
1454 	}
1455 
1456 	/*
1457 	 * Changes on the prefix status might affect address status as well.
1458 	 * Make sure that all addresses derived from an attached prefix are
1459 	 * attached, and that all addresses derived from a detached prefix are
1460 	 * detached.  Note, however, that a manually configured address should
1461 	 * always be attached.
1462 	 * The precise detection logic is same as the one for prefixes.
1463 	 */
1464 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1465 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1466 			continue;
1467 
1468 		if (ifa->ia6_ndpr == NULL) {
1469 			/*
1470 			 * This can happen when we first configure the address
1471 			 * (i.e. the address exists, but the prefix does not).
1472 			 * XXX: complicated relationships...
1473 			 */
1474 			continue;
1475 		}
1476 
1477 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1478 			break;
1479 	}
1480 	if (ifa) {
1481 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1482 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1483 				continue;
1484 
1485 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1486 				continue;
1487 
1488 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1489 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1490 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1491 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1492 					nd6_dad_start((struct ifaddr *)ifa,
1493 					    0);
1494 				}
1495 			} else {
1496 				if ((ifa->ia6_flags & IN6_IFF_DETACHED) == 0) {
1497 					ifa->ia6_flags |= IN6_IFF_DETACHED;
1498 				}
1499 			}
1500 		}
1501 	}
1502 	else {
1503 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1504 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1505 				continue;
1506 
1507 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1508 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1509 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1510 				/* Do we need a delay in this case? */
1511 				nd6_dad_start((struct ifaddr *)ifa, 0);
1512 			}
1513 		}
1514 	}
1515 }
1516 
1517 int
1518 nd6_prefix_onlink(struct nd_prefix *pr)
1519 {
1520 	struct ifaddr *ifa;
1521 	struct ifnet *ifp = pr->ndpr_ifp;
1522 	struct sockaddr_in6 mask6;
1523 	struct nd_prefix *opr;
1524 	u_long rtflags;
1525 	int error = 0;
1526 	struct rtentry *rt = NULL;
1527 
1528 	/* sanity check */
1529 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1530 		nd6log((LOG_ERR,
1531 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1532 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1533 		return (EEXIST);
1534 	}
1535 
1536 	/*
1537 	 * Add the interface route associated with the prefix.  Before
1538 	 * installing the route, check if there's the same prefix on another
1539 	 * interface, and the prefix has already installed the interface route.
1540 	 * Although such a configuration is expected to be rare, we explicitly
1541 	 * allow it.
1542 	 */
1543 	LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1544 		if (opr == pr)
1545 			continue;
1546 
1547 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1548 			continue;
1549 
1550 		if (opr->ndpr_plen == pr->ndpr_plen &&
1551 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1552 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1553 			return (0);
1554 	}
1555 
1556 	/*
1557 	 * We prefer link-local addresses as the associated interface address.
1558 	 */
1559 	/* search for a link-local addr */
1560 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1561 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1562 	if (ifa == NULL) {
1563 		/* XXX: freebsd does not have ifa_ifwithaf */
1564 		IFADDR_FOREACH(ifa, ifp) {
1565 			if (ifa->ifa_addr->sa_family == AF_INET6)
1566 				break;
1567 		}
1568 		/* should we care about ia6_flags? */
1569 	}
1570 	if (ifa == NULL) {
1571 		/*
1572 		 * This can still happen, when, for example, we receive an RA
1573 		 * containing a prefix with the L bit set and the A bit clear,
1574 		 * after removing all IPv6 addresses on the receiving
1575 		 * interface.  This should, of course, be rare though.
1576 		 */
1577 		nd6log((LOG_NOTICE,
1578 		    "nd6_prefix_onlink: failed to find any ifaddr"
1579 		    " to add route for a prefix(%s/%d) on %s\n",
1580 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1581 		    pr->ndpr_plen, if_name(ifp)));
1582 		return (0);
1583 	}
1584 
1585 	/*
1586 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1587 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1588 	 */
1589 	memset(&mask6, 0, sizeof(mask6));
1590 	mask6.sin6_len = sizeof(mask6);
1591 	mask6.sin6_addr = pr->ndpr_mask;
1592 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
1593 	rtflags = ifa->ifa_flags | RTF_UP;
1594 	if (nd6_need_cache(ifp)) {
1595 		/* explicitly set in case ifa_flags does not set the flag. */
1596 		rtflags |= RTF_CLONING;
1597 	} else {
1598 		/*
1599 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1600 		 */
1601 		rtflags &= ~RTF_CLONING;
1602 	}
1603 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1604 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1605 	if (error == 0) {
1606 		if (rt != NULL) /* this should be non NULL, though */
1607 			nd6_rtmsg(RTM_ADD, rt);
1608 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1609 	} else {
1610 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1611 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1612 		    "errno = %d\n",
1613 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1614 		    pr->ndpr_plen, if_name(ifp),
1615 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1616 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
1617 	}
1618 
1619 	if (rt != NULL)
1620 		rt->rt_refcnt--;
1621 
1622 	return (error);
1623 }
1624 
1625 int
1626 nd6_prefix_offlink(struct nd_prefix *pr)
1627 {
1628 	int error = 0;
1629 	struct ifnet *ifp = pr->ndpr_ifp;
1630 	struct nd_prefix *opr;
1631 	struct sockaddr_in6 sa6, mask6;
1632 	struct rtentry *rt = NULL;
1633 
1634 	/* sanity check */
1635 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1636 		nd6log((LOG_ERR,
1637 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1638 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1639 		return (EEXIST);
1640 	}
1641 
1642 	sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1643 	sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
1644 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1645 	    (struct sockaddr *)&mask6, 0, &rt);
1646 	if (error == 0) {
1647 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1648 
1649 		/* report the route deletion to the routing socket. */
1650 		if (rt != NULL)
1651 			nd6_rtmsg(RTM_DELETE, rt);
1652 
1653 		/*
1654 		 * There might be the same prefix on another interface,
1655 		 * the prefix which could not be on-link just because we have
1656 		 * the interface route (see comments in nd6_prefix_onlink).
1657 		 * If there's one, try to make the prefix on-link on the
1658 		 * interface.
1659 		 */
1660 		LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1661 			if (opr == pr)
1662 				continue;
1663 
1664 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1665 				continue;
1666 
1667 			/*
1668 			 * KAME specific: detached prefixes should not be
1669 			 * on-link.
1670 			 */
1671 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1672 				continue;
1673 
1674 			if (opr->ndpr_plen == pr->ndpr_plen &&
1675 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1676 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1677 				int e;
1678 
1679 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1680 					nd6log((LOG_ERR,
1681 					    "nd6_prefix_offlink: failed to "
1682 					    "recover a prefix %s/%d from %s "
1683 					    "to %s (errno = %d)\n",
1684 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1685 					    opr->ndpr_plen, if_name(ifp),
1686 					    if_name(opr->ndpr_ifp), e));
1687 				}
1688 			}
1689 		}
1690 	} else {
1691 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1692 		nd6log((LOG_ERR,
1693 		    "nd6_prefix_offlink: failed to delete route: "
1694 		    "%s/%d on %s (errno = %d)\n",
1695 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1696 		    error));
1697 	}
1698 
1699 	if (rt != NULL) {
1700 		if (rt->rt_refcnt <= 0) {
1701 			/* XXX: we should free the entry ourselves. */
1702 			rt->rt_refcnt++;
1703 			rtfree(rt);
1704 		}
1705 	}
1706 
1707 	return (error);
1708 }
1709 
1710 static struct in6_ifaddr *
1711 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1712 {
1713 	struct ifnet *ifp = pr->ndpr_ifp;
1714 	struct ifaddr *ifa;
1715 	struct in6_aliasreq ifra;
1716 	struct in6_ifaddr *ia, *ib;
1717 	int error, plen0;
1718 	struct in6_addr mask;
1719 	int prefixlen = pr->ndpr_plen;
1720 	int updateflags;
1721 
1722 	in6_prefixlen2mask(&mask, prefixlen);
1723 
1724 	/*
1725 	 * find a link-local address (will be interface ID).
1726 	 * Is it really mandatory? Theoretically, a global or a site-local
1727 	 * address can be configured without a link-local address, if we
1728 	 * have a unique interface identifier...
1729 	 *
1730 	 * it is not mandatory to have a link-local address, we can generate
1731 	 * interface identifier on the fly.  we do this because:
1732 	 * (1) it should be the easiest way to find interface identifier.
1733 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1734 	 * for multiple addresses on a single interface, and possible shortcut
1735 	 * of DAD.  we omitted DAD for this reason in the past.
1736 	 * (3) a user can prevent autoconfiguration of global address
1737 	 * by removing link-local address by hand (this is partly because we
1738 	 * don't have other way to control the use of IPv6 on an interface.
1739 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1740 	 * (4) it is easier to manage when an interface has addresses
1741 	 * with the same interface identifier, than to have multiple addresses
1742 	 * with different interface identifiers.
1743 	 */
1744 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1745 	if (ifa)
1746 		ib = (struct in6_ifaddr *)ifa;
1747 	else
1748 		return NULL;
1749 
1750 #if 0 /* don't care link local addr state, and always do DAD */
1751 	/* if link-local address is not eligible, do not autoconfigure. */
1752 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1753 		printf("in6_ifadd: link-local address not ready\n");
1754 		return NULL;
1755 	}
1756 #endif
1757 
1758 	/* prefixlen + ifidlen must be equal to 128 */
1759 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1760 	if (prefixlen != plen0) {
1761 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1762 		    "(prefix=%d ifid=%d)\n",
1763 		    if_name(ifp), prefixlen, 128 - plen0));
1764 		return NULL;
1765 	}
1766 
1767 	/* make ifaddr */
1768 
1769 	memset(&ifra, 0, sizeof(ifra));
1770 	/*
1771 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1772 	 * for safety.
1773 	 */
1774 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1775 	sockaddr_in6_init(&ifra.ifra_addr, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1776 	/* prefix */
1777 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1778 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1779 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1780 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1781 
1782 	/* interface ID */
1783 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1784 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1785 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1786 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1787 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1788 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1789 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1790 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1791 
1792 	/* new prefix mask. */
1793 	sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
1794 
1795 	/* lifetimes */
1796 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1797 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1798 
1799 	/* XXX: scope zone ID? */
1800 
1801 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1802 
1803 	/*
1804 	 * Make sure that we do not have this address already.  This should
1805 	 * usually not happen, but we can still see this case, e.g., if we
1806 	 * have manually configured the exact address to be configured.
1807 	 */
1808 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1809 		/* this should be rare enough to make an explicit log */
1810 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1811 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1812 		return (NULL);
1813 	}
1814 
1815 	/*
1816 	 * Allocate ifaddr structure, link into chain, etc.
1817 	 * If we are going to create a new address upon receiving a multicasted
1818 	 * RA, we need to impose a random delay before starting DAD.
1819 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1820 	 */
1821 	updateflags = 0;
1822 	if (mcast)
1823 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1824 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1825 		nd6log((LOG_ERR,
1826 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1827 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1828 		    error));
1829 		return (NULL);	/* ifaddr must not have been allocated. */
1830 	}
1831 
1832 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1833 
1834 	return (ia);		/* this is always non-NULL */
1835 }
1836 
1837 int
1838 in6_tmpifadd(
1839 	const struct in6_ifaddr *ia0, /* corresponding public address */
1840 	int forcegen,
1841 	int dad_delay)
1842 {
1843 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1844 	struct in6_ifaddr *newia, *ia;
1845 	struct in6_aliasreq ifra;
1846 	int i, error;
1847 	int trylimit = 3;	/* XXX: adhoc value */
1848 	int updateflags;
1849 	u_int32_t randid[2];
1850 	u_int32_t vltime0, pltime0;
1851 
1852 	memset(&ifra, 0, sizeof(ifra));
1853 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1854 	ifra.ifra_addr = ia0->ia_addr;
1855 	/* copy prefix mask */
1856 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1857 	/* clear the old IFID */
1858 	for (i = 0; i < 4; i++) {
1859 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1860 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1861 	}
1862 
1863   again:
1864 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1865 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1866 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1867 		    "random IFID\n"));
1868 		return (EINVAL);
1869 	}
1870 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1871 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1872 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1873 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1874 
1875 	/*
1876 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1877 	 * However, we may still have a chance to see collision, because
1878 	 * there may be a time lag between generation of the ID and generation
1879 	 * of the address.  So, we'll do one more sanity check.
1880 	 */
1881 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1882 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1883 		    &ifra.ifra_addr.sin6_addr)) {
1884 			if (trylimit-- == 0) {
1885 				/*
1886 				 * Give up.  Something strange should have
1887 				 * happened.
1888 				 */
1889 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1890 				    "find a unique random IFID\n"));
1891 				return (EEXIST);
1892 			}
1893 			forcegen = 1;
1894 			goto again;
1895 		}
1896 	}
1897 
1898 	/*
1899 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1900          * public address or TEMP_VALID_LIFETIME.
1901 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1902          * of the public address or TEMP_PREFERRED_LIFETIME -
1903          * DESYNC_FACTOR.
1904 	 */
1905 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1906 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1907 		    (ia0->ia6_lifetime.ia6t_vltime -
1908 		    (time_second - ia0->ia6_updatetime));
1909 		if (vltime0 > ip6_temp_valid_lifetime)
1910 			vltime0 = ip6_temp_valid_lifetime;
1911 	} else
1912 		vltime0 = ip6_temp_valid_lifetime;
1913 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1914 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1915 		    (ia0->ia6_lifetime.ia6t_pltime -
1916 		    (time_second - ia0->ia6_updatetime));
1917 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
1918 			pltime0 = ip6_temp_preferred_lifetime -
1919 			    ip6_desync_factor;
1920 		}
1921 	} else
1922 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
1923 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1924 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1925 
1926 	/*
1927 	 * A temporary address is created only if this calculated Preferred
1928 	 * Lifetime is greater than REGEN_ADVANCE time units.
1929 	 */
1930 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
1931 		return (0);
1932 
1933 	/* XXX: scope zone ID? */
1934 
1935 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1936 
1937 	/* allocate ifaddr structure, link into chain, etc. */
1938 	updateflags = 0;
1939 	if (dad_delay)
1940 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1941 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1942 		return (error);
1943 
1944 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1945 	if (newia == NULL) {	/* XXX: can it happen? */
1946 		nd6log((LOG_ERR,
1947 		    "in6_tmpifadd: ifa update succeeded, but we got "
1948 		    "no ifaddr\n"));
1949 		return (EINVAL); /* XXX */
1950 	}
1951 	newia->ia6_ndpr = ia0->ia6_ndpr;
1952 	newia->ia6_ndpr->ndpr_refcnt++;
1953 
1954 	/*
1955 	 * A newly added address might affect the status of other addresses.
1956 	 * XXX: when the temporary address is generated with a new public
1957 	 * address, the onlink check is redundant.  However, it would be safe
1958 	 * to do the check explicitly everywhere a new address is generated,
1959 	 * and, in fact, we surely need the check when we create a new
1960 	 * temporary address due to deprecation of an old temporary address.
1961 	 */
1962 	pfxlist_onlink_check();
1963 
1964 	return (0);
1965 }
1966 
1967 static int
1968 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1969 {
1970 
1971 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
1972 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
1973 		nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
1974 		    "(%d) is greater than valid lifetime(%d)\n",
1975 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
1976 		return (EINVAL);
1977 	}
1978 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1979 		ndpr->ndpr_preferred = 0;
1980 	else
1981 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
1982 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1983 		ndpr->ndpr_expire = 0;
1984 	else
1985 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
1986 
1987 	return 0;
1988 }
1989 
1990 static void
1991 in6_init_address_ltimes(struct nd_prefix *new,
1992     struct in6_addrlifetime *lt6)
1993 {
1994 
1995 	/* Valid lifetime must not be updated unless explicitly specified. */
1996 	/* init ia6t_expire */
1997 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
1998 		lt6->ia6t_expire = 0;
1999 	else {
2000 		lt6->ia6t_expire = time_second;
2001 		lt6->ia6t_expire += lt6->ia6t_vltime;
2002 	}
2003 
2004 	/* init ia6t_preferred */
2005 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2006 		lt6->ia6t_preferred = 0;
2007 	else {
2008 		lt6->ia6t_preferred = time_second;
2009 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2010 	}
2011 }
2012 
2013 /*
2014  * Delete all the routing table entries that use the specified gateway.
2015  * XXX: this function causes search through all entries of routing table, so
2016  * it shouldn't be called when acting as a router.
2017  */
2018 void
2019 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2020 {
2021 	int s = splsoftnet();
2022 
2023 	/* We'll care only link-local addresses */
2024 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2025 		splx(s);
2026 		return;
2027 	}
2028 
2029 	rt_walktree(AF_INET6, rt6_deleteroute, (void *)gateway);
2030 	splx(s);
2031 }
2032 
2033 static int
2034 rt6_deleteroute(struct rtentry *rt, void *arg)
2035 {
2036 #define SIN6(s)	((struct sockaddr_in6 *)s)
2037 	struct in6_addr *gate = (struct in6_addr *)arg;
2038 
2039 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2040 		return (0);
2041 
2042 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr))
2043 		return (0);
2044 
2045 	/*
2046 	 * Do not delete a static route.
2047 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2048 	 * 'cloned' bit instead?
2049 	 */
2050 	if ((rt->rt_flags & RTF_STATIC) != 0)
2051 		return (0);
2052 
2053 	/*
2054 	 * We delete only host route. This means, in particular, we don't
2055 	 * delete default route.
2056 	 */
2057 	if ((rt->rt_flags & RTF_HOST) == 0)
2058 		return (0);
2059 
2060 	return (rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
2061 	    rt_mask(rt), rt->rt_flags, 0));
2062 #undef SIN6
2063 }
2064 
2065 int
2066 nd6_setdefaultiface(int ifindex)
2067 {
2068 	int error = 0;
2069 
2070 	if (ifindex < 0 || if_indexlim <= ifindex)
2071 		return (EINVAL);
2072 	if (ifindex != 0 && !ifindex2ifnet[ifindex])
2073 		return (EINVAL);
2074 
2075 	if (nd6_defifindex != ifindex) {
2076 		nd6_defifindex = ifindex;
2077 		if (nd6_defifindex > 0) {
2078 			nd6_defifp = ifindex2ifnet[nd6_defifindex];
2079 		} else
2080 			nd6_defifp = NULL;
2081 
2082 		/*
2083 		 * Our current implementation assumes one-to-one maping between
2084 		 * interfaces and links, so it would be natural to use the
2085 		 * default interface as the default link.
2086 		 */
2087 		scope6_setdefault(nd6_defifp);
2088 	}
2089 
2090 	return (error);
2091 }
2092