xref: /netbsd-src/sys/netinet6/nd6_rtr.c (revision f17b710f3d406bee67aa39c65053114ab78297c5)
1 /*	$NetBSD: nd6_rtr.c,v 1.100 2015/06/30 06:42:06 ozaki-r 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 1.100 2015/06/30 06:42:06 ozaki-r 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 #include <sys/cprng.h>
48 
49 #include <net/if.h>
50 #include <net/if_types.h>
51 #include <net/if_dl.h>
52 #include <net/route.h>
53 #include <net/radix.h>
54 
55 #include <netinet/in.h>
56 #include <netinet6/in6_var.h>
57 #include <netinet6/in6_ifattach.h>
58 #include <netinet/ip6.h>
59 #include <netinet6/ip6_var.h>
60 #include <netinet6/nd6.h>
61 #include <netinet/icmp6.h>
62 #include <netinet6/icmp6_private.h>
63 #include <netinet6/scope6_var.h>
64 
65 #include <net/net_osdep.h>
66 
67 static int rtpref(struct nd_defrouter *);
68 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
69 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
70     struct mbuf *, int);
71 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
72 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
73 	struct nd_defrouter *);
74 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
75 static void pfxrtr_del(struct nd_pfxrouter *);
76 static struct nd_pfxrouter *find_pfxlist_reachable_router
77 	(struct nd_prefix *);
78 static void defrouter_delreq(struct nd_defrouter *);
79 
80 static int in6_init_prefix_ltimes(struct nd_prefix *);
81 static void in6_init_address_ltimes(struct nd_prefix *,
82 	struct in6_addrlifetime *);
83 static void purge_detached(struct ifnet *);
84 
85 static int rt6_deleteroute(struct rtentry *, void *);
86 
87 extern int nd6_recalc_reachtm_interval;
88 
89 static struct ifnet *nd6_defifp;
90 int nd6_defifindex;
91 
92 int ip6_use_tempaddr = 0;
93 
94 int ip6_desync_factor;
95 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
96 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
97 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
98 
99 int nd6_numroutes = 0;
100 
101 /* RTPREF_MEDIUM has to be 0! */
102 #define RTPREF_HIGH	1
103 #define RTPREF_MEDIUM	0
104 #define RTPREF_LOW	(-1)
105 #define RTPREF_RESERVED	(-2)
106 #define RTPREF_INVALID	(-3)	/* internal */
107 
108 /*
109  * Receive Router Solicitation Message - just for routers.
110  * Router solicitation/advertisement is mostly managed by a userland program
111  * (rtadvd) so here we have no function like nd6_ra_output().
112  *
113  * Based on RFC 2461
114  */
115 void
116 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
117 {
118 	struct ifnet *ifp = m->m_pkthdr.rcvif;
119 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
120 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
121 	struct nd_router_solicit *nd_rs;
122 	struct in6_addr saddr6 = ip6->ip6_src;
123 	char *lladdr = NULL;
124 	int lladdrlen = 0;
125 	union nd_opts ndopts;
126 
127 	/* If I'm not a router, ignore it. */
128 	if (nd6_accepts_rtadv(ndi) || !ip6_forwarding)
129 		goto freeit;
130 
131 	/* Sanity checks */
132 	if (ip6->ip6_hlim != 255) {
133 		nd6log((LOG_ERR,
134 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
135 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
136 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
137 		goto bad;
138 	}
139 
140 	/*
141 	 * Don't update the neighbor cache, if src = ::.
142 	 * This indicates that the src has no IP address assigned yet.
143 	 */
144 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
145 		goto freeit;
146 
147 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
148 	if (nd_rs == NULL) {
149 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
150 		return;
151 	}
152 
153 	icmp6len -= sizeof(*nd_rs);
154 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
155 	if (nd6_options(&ndopts) < 0) {
156 		nd6log((LOG_INFO,
157 		    "nd6_rs_input: invalid ND option, ignored\n"));
158 		/* nd6_options have incremented stats */
159 		goto freeit;
160 	}
161 
162 	if (ndopts.nd_opts_src_lladdr) {
163 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
164 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
165 	}
166 
167 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
168 		nd6log((LOG_INFO,
169 		    "nd6_rs_input: lladdrlen mismatch for %s "
170 		    "(if %d, RS packet %d)\n",
171 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
172 		goto bad;
173 	}
174 
175 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
176 
177  freeit:
178 	m_freem(m);
179 	return;
180 
181  bad:
182 	ICMP6_STATINC(ICMP6_STAT_BADRS);
183 	m_freem(m);
184 }
185 
186 /*
187  * Receive Router Advertisement Message.
188  *
189  * Based on RFC 2461
190  * TODO: on-link bit on prefix information
191  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
192  */
193 void
194 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
195 {
196 	struct ifnet *ifp = m->m_pkthdr.rcvif;
197 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
198 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
199 	struct nd_router_advert *nd_ra;
200 	struct in6_addr saddr6 = ip6->ip6_src;
201 #if 0
202 	struct in6_addr daddr6 = ip6->ip6_dst;
203 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
204 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
205 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
206 #endif
207 	int mcast = 0;
208 	union nd_opts ndopts;
209 	struct nd_defrouter *dr;
210 
211 	/*
212 	 * We only accept RAs when
213 	 * the system-wide variable allows the acceptance, and the
214 	 * per-interface variable allows RAs on the receiving interface.
215 	 */
216 	if (!nd6_accepts_rtadv(ndi))
217 		goto freeit;
218 
219 	if (ip6->ip6_hlim != 255) {
220 		nd6log((LOG_ERR,
221 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
222 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
223 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
224 		goto bad;
225 	}
226 
227 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
228 		nd6log((LOG_ERR,
229 		    "nd6_ra_input: src %s is not link-local\n",
230 		    ip6_sprintf(&saddr6)));
231 		goto bad;
232 	}
233 
234 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
235 	if (nd_ra == NULL) {
236 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
237 		return;
238 	}
239 
240 	icmp6len -= sizeof(*nd_ra);
241 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
242 	if (nd6_options(&ndopts) < 0) {
243 		nd6log((LOG_INFO,
244 		    "nd6_ra_input: invalid ND option, ignored\n"));
245 		/* nd6_options have incremented stats */
246 		goto freeit;
247 	}
248 
249     {
250 	struct nd_defrouter drtr;
251 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
252 
253 	/* remember if this is a multicasted advertisement */
254 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
255 		mcast = 1;
256 
257 	memset(&drtr, 0, sizeof(drtr));
258 	drtr.rtaddr = saddr6;
259 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
260 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
261 	drtr.expire = time_second + drtr.rtlifetime;
262 	drtr.ifp = ifp;
263 	/* unspecified or not? (RFC 2461 6.3.4) */
264 	if (advreachable) {
265 		NTOHL(advreachable);
266 		if (advreachable <= MAX_REACHABLE_TIME &&
267 		    ndi->basereachable != advreachable) {
268 			ndi->basereachable = advreachable;
269 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
270 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
271 		}
272 	}
273 	if (nd_ra->nd_ra_retransmit)
274 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
275 	if (nd_ra->nd_ra_curhoplimit) {
276 		if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
277 			ndi->chlim = nd_ra->nd_ra_curhoplimit;
278 		else if (ndi->chlim != nd_ra->nd_ra_curhoplimit)
279 			log(LOG_ERR, "nd_ra_input: lower CurHopLimit sent from "
280 			   "%s on %s (current=%d, received=%d), ignored\n",
281 			   ip6_sprintf(&ip6->ip6_src),
282 			   if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
283 	}
284 	dr = defrtrlist_update(&drtr);
285     }
286 
287 	/*
288 	 * prefix
289 	 */
290 	if (ndopts.nd_opts_pi) {
291 		struct nd_opt_hdr *pt;
292 		struct nd_opt_prefix_info *pi = NULL;
293 		struct nd_prefixctl prc;
294 
295 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
296 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
297 		     pt = (struct nd_opt_hdr *)((char *)pt +
298 						(pt->nd_opt_len << 3))) {
299 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
300 				continue;
301 			pi = (struct nd_opt_prefix_info *)pt;
302 
303 			if (pi->nd_opt_pi_len != 4) {
304 				nd6log((LOG_INFO,
305 				    "nd6_ra_input: invalid option "
306 				    "len %d for prefix information option, "
307 				    "ignored\n", pi->nd_opt_pi_len));
308 				continue;
309 			}
310 
311 			if (128 < pi->nd_opt_pi_prefix_len) {
312 				nd6log((LOG_INFO,
313 				    "nd6_ra_input: invalid prefix "
314 				    "len %d for prefix information option, "
315 				    "ignored\n", pi->nd_opt_pi_prefix_len));
316 				continue;
317 			}
318 
319 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
320 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
321 				nd6log((LOG_INFO,
322 				    "nd6_ra_input: invalid prefix "
323 				    "%s, ignored\n",
324 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
325 				continue;
326 			}
327 
328 			memset(&prc, 0, sizeof(prc));
329 			sockaddr_in6_init(&prc.ndprc_prefix,
330 			    &pi->nd_opt_pi_prefix, 0, 0, 0);
331 			prc.ndprc_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
332 
333 			prc.ndprc_raf_onlink = (pi->nd_opt_pi_flags_reserved &
334 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
335 			prc.ndprc_raf_auto = (pi->nd_opt_pi_flags_reserved &
336 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
337 			prc.ndprc_plen = pi->nd_opt_pi_prefix_len;
338 			prc.ndprc_vltime = ntohl(pi->nd_opt_pi_valid_time);
339 			prc.ndprc_pltime = ntohl(pi->nd_opt_pi_preferred_time);
340 
341 			(void)prelist_update(&prc, dr, m, mcast);
342 		}
343 	}
344 
345 	/*
346 	 * MTU
347 	 */
348 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
349 		u_long mtu;
350 		u_long maxmtu;
351 
352 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
353 
354 		/* lower bound */
355 		if (mtu < IPV6_MMTU) {
356 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
357 			    "mtu=%lu sent from %s, ignoring\n",
358 			    mtu, ip6_sprintf(&ip6->ip6_src)));
359 			goto skip;
360 		}
361 
362 		/* upper bound */
363 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
364 		    ? ndi->maxmtu : ifp->if_mtu;
365 		if (mtu <= maxmtu) {
366 			int change = (ndi->linkmtu != mtu);
367 
368 			ndi->linkmtu = mtu;
369 			if (change) /* in6_maxmtu may change */
370 				in6_setmaxmtu();
371 		} else {
372 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
373 			    "mtu=%lu sent from %s; "
374 			    "exceeds maxmtu %lu, ignoring\n",
375 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
376 		}
377 	}
378 
379  skip:
380 
381 	/*
382 	 * Source link layer address
383 	 */
384     {
385 	char *lladdr = NULL;
386 	int lladdrlen = 0;
387 
388 	if (ndopts.nd_opts_src_lladdr) {
389 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
390 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
391 	}
392 
393 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
394 		nd6log((LOG_INFO,
395 		    "nd6_ra_input: lladdrlen mismatch for %s "
396 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
397 		    ifp->if_addrlen, lladdrlen - 2));
398 		goto bad;
399 	}
400 
401 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
402 
403 	/*
404 	 * Installing a link-layer address might change the state of the
405 	 * router's neighbor cache, which might also affect our on-link
406 	 * detection of adveritsed prefixes.
407 	 */
408 	pfxlist_onlink_check();
409     }
410 
411  freeit:
412 	m_freem(m);
413 	return;
414 
415  bad:
416 	ICMP6_STATINC(ICMP6_STAT_BADRA);
417 	m_freem(m);
418 }
419 
420 /*
421  * default router list processing sub routines
422  */
423 void
424 defrouter_addreq(struct nd_defrouter *newdr)
425 {
426 	union {
427 		struct sockaddr_in6 sin6;
428 		struct sockaddr sa;
429 	} def, mask, gate;
430 	struct rtentry *newrt = NULL;
431 	int s;
432 	int error;
433 
434 	memset(&def, 0, sizeof(def));
435 	memset(&mask, 0, sizeof(mask));
436 	memset(&gate, 0,sizeof(gate)); /* for safety */
437 
438 	def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len =
439 	    sizeof(struct sockaddr_in6);
440 	def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6;
441 	gate.sin6.sin6_addr = newdr->rtaddr;
442 #ifndef SCOPEDROUTING
443 	gate.sin6.sin6_scope_id = 0;	/* XXX */
444 #endif
445 
446 	s = splsoftnet();
447 	error = rtrequest(RTM_ADD, &def.sa, &gate.sa, &mask.sa,
448 	    RTF_GATEWAY, &newrt);
449 	if (newrt) {
450 		rt_newmsg(RTM_ADD, newrt); /* tell user process */
451 		newrt->rt_refcnt--;
452 		nd6_numroutes++;
453 	}
454 	if (error == 0)
455 		newdr->installed = 1;
456 	splx(s);
457 	return;
458 }
459 
460 struct nd_defrouter *
461 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
462 {
463 	struct nd_defrouter *dr;
464 
465 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
466 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
467 			break;
468 	}
469 
470 	return dr;		/* search failed */
471 }
472 
473 void
474 defrtrlist_del(struct nd_defrouter *dr, struct in6_ifextra *ext)
475 {
476 	struct nd_defrouter *deldr = NULL;
477 	struct nd_prefix *pr;
478 	struct nd_ifinfo *ndi;
479 
480 	if (ext == NULL)
481 		ext = dr->ifp->if_afdata[AF_INET6];
482 
483 	/* detach already in progress, can not do anything */
484 	if (ext == NULL)
485 		return;
486 
487 	ndi = ext->nd_ifinfo;
488 
489 	/*
490 	 * Flush all the routing table entries that use the router
491 	 * as a next hop.
492 	 */
493 	/* XXX: better condition? */
494 	if (!ip6_forwarding && nd6_accepts_rtadv(ndi))
495 		rt6_flush(&dr->rtaddr, dr->ifp);
496 
497 	if (dr->installed) {
498 		deldr = dr;
499 		defrouter_delreq(dr);
500 	}
501 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
502 
503 	/*
504 	 * Also delete all the pointers to the router in each prefix lists.
505 	 */
506 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
507 		struct nd_pfxrouter *pfxrtr;
508 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
509 			pfxrtr_del(pfxrtr);
510 	}
511 	pfxlist_onlink_check();
512 
513 	/*
514 	 * If the router is the primary one, choose a new one.
515 	 * Note that defrouter_select() will remove the current gateway
516 	 * from the routing table.
517 	 */
518 	if (deldr)
519 		defrouter_select();
520 
521 	ext->ndefrouters--;
522 	if (ext->ndefrouters < 0) {
523 		log(LOG_WARNING, "defrtrlist_del: negative count on %s\n",
524 		    dr->ifp->if_xname);
525 	}
526 
527 	free(dr, M_IP6NDP);
528 }
529 
530 /*
531  * Remove the default route for a given router.
532  * This is just a subroutine function for defrouter_select(), and should
533  * not be called from anywhere else.
534  */
535 static void
536 defrouter_delreq(struct nd_defrouter *dr)
537 {
538 	union {
539 		struct sockaddr_in6 sin6;
540 		struct sockaddr sa;
541 	} def, mask, gw;
542 	struct rtentry *oldrt = NULL;
543 
544 #ifdef DIAGNOSTIC
545 	if (dr == NULL)
546 		panic("dr == NULL in defrouter_delreq");
547 #endif
548 
549 	memset(&def, 0, sizeof(def));
550 	memset(&mask, 0, sizeof(mask));
551 	memset(&gw, 0, sizeof(gw));	/* for safety */
552 
553 	def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len =
554 	    sizeof(struct sockaddr_in6);
555 	def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6;
556 	gw.sin6.sin6_addr = dr->rtaddr;
557 #ifndef SCOPEDROUTING
558 	gw.sin6.sin6_scope_id = 0;	/* XXX */
559 #endif
560 
561 	rtrequest(RTM_DELETE, &def.sa, &gw.sa, &mask.sa, RTF_GATEWAY, &oldrt);
562 	if (oldrt) {
563 		rt_newmsg(RTM_DELETE, oldrt);
564 		if (oldrt->rt_refcnt <= 0) {
565 			/*
566 			 * XXX: borrowed from the RTM_DELETE case of
567 			 * rtrequest().
568 			 */
569 			oldrt->rt_refcnt++;
570 			rtfree(oldrt);
571 		}
572 		nd6_numroutes--;
573 	}
574 
575 	dr->installed = 0;
576 }
577 
578 /*
579  * remove all default routes from default router list
580  */
581 void
582 defrouter_reset(void)
583 {
584 	struct nd_defrouter *dr;
585 
586 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
587 	     dr = TAILQ_NEXT(dr, dr_entry))
588 		defrouter_delreq(dr);
589 
590 	/*
591 	 * XXX should we also nuke any default routers in the kernel, by
592 	 * going through them by rtalloc1()?
593 	 */
594 }
595 
596 /*
597  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
598  * draft-ietf-ipngwg-router-selection:
599  * 1) Routers that are reachable or probably reachable should be preferred.
600  *    If we have more than one (probably) reachable router, prefer ones
601  *    with the highest router preference.
602  * 2) When no routers on the list are known to be reachable or
603  *    probably reachable, routers SHOULD be selected in a round-robin
604  *    fashion, regardless of router preference values.
605  * 3) If the Default Router List is empty, assume that all
606  *    destinations are on-link.
607  *
608  * We assume nd_defrouter is sorted by router preference value.
609  * Since the code below covers both with and without router preference cases,
610  * we do not need to classify the cases by ifdef.
611  *
612  * At this moment, we do not try to install more than one default router,
613  * even when the multipath routing is available, because we're not sure about
614  * the benefits for stub hosts comparing to the risk of making the code
615  * complicated and the possibility of introducing bugs.
616  */
617 void
618 defrouter_select(void)
619 {
620 	struct nd_ifinfo *ndi;
621 	int s = splsoftnet();
622 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
623 	struct rtentry *rt = NULL;
624 	struct llinfo_nd6 *ln = NULL;
625 
626 	/*
627 	 * This function should be called only when acting as an autoconfigured
628 	 * host.  Although the remaining part of this function is not effective
629 	 * if the node is not an autoconfigured host, we explicitly exclude
630 	 * such cases here for safety.
631 	 */
632 	if (ip6_forwarding) {
633 		nd6log((LOG_WARNING,
634 		    "defrouter_select: called unexpectedly (forwarding=%d, "
635 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
636 		splx(s);
637 		return;
638 	}
639 
640 	/*
641 	 * Let's handle easy case (3) first:
642 	 * If default router list is empty, there's nothing to be done.
643 	 */
644 	if (!TAILQ_FIRST(&nd_defrouter)) {
645 		splx(s);
646 		return;
647 	}
648 
649 	/*
650 	 * Search for a (probably) reachable router from the list.
651 	 * We just pick up the first reachable one (if any), assuming that
652 	 * the ordering rule of the list described in defrtrlist_update().
653 	 */
654 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
655 	     dr = TAILQ_NEXT(dr, dr_entry)) {
656 		ndi = ND_IFINFO(dr->ifp);
657 		if (nd6_accepts_rtadv(ndi))
658 			continue;
659 
660 		if (selected_dr == NULL &&
661 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) != NULL &&
662 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) != NULL &&
663 		    ND6_IS_LLINFO_PROBREACH(ln)) {
664 			selected_dr = dr;
665 		}
666 
667 		if (dr->installed && !installed_dr)
668 			installed_dr = dr;
669 		else if (dr->installed && installed_dr) {
670 			/* this should not happen.  warn for diagnosis. */
671 			log(LOG_ERR, "defrouter_select: more than one router"
672 			    " is installed\n");
673 		}
674 	}
675 	/*
676 	 * If none of the default routers was found to be reachable,
677 	 * round-robin the list regardless of preference.
678 	 * Otherwise, if we have an installed router, check if the selected
679 	 * (reachable) router should really be preferred to the installed one.
680 	 * We only prefer the new router when the old one is not reachable
681 	 * or when the new one has a really higher preference value.
682 	 */
683 	if (selected_dr == NULL) {
684 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
685 			selected_dr = TAILQ_FIRST(&nd_defrouter);
686 		else
687 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
688 	} else if (installed_dr &&
689 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
690 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
691 	    ND6_IS_LLINFO_PROBREACH(ln) &&
692 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
693 		selected_dr = installed_dr;
694 	}
695 
696 	/*
697 	 * If the selected router is different than the installed one,
698 	 * remove the installed router and install the selected one.
699 	 * Note that the selected router is never NULL here.
700 	 */
701 	if (installed_dr != selected_dr) {
702 		if (installed_dr)
703 			defrouter_delreq(installed_dr);
704 		defrouter_addreq(selected_dr);
705 	}
706 
707 	splx(s);
708 	return;
709 }
710 
711 /*
712  * for default router selection
713  * regards router-preference field as a 2-bit signed integer
714  */
715 static int
716 rtpref(struct nd_defrouter *dr)
717 {
718 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
719 	case ND_RA_FLAG_RTPREF_HIGH:
720 		return (RTPREF_HIGH);
721 	case ND_RA_FLAG_RTPREF_MEDIUM:
722 	case ND_RA_FLAG_RTPREF_RSV:
723 		return (RTPREF_MEDIUM);
724 	case ND_RA_FLAG_RTPREF_LOW:
725 		return (RTPREF_LOW);
726 	default:
727 		/*
728 		 * This case should never happen.  If it did, it would mean a
729 		 * serious bug of kernel internal.  We thus always bark here.
730 		 * Or, can we even panic?
731 		 */
732 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
733 		return (RTPREF_INVALID);
734 	}
735 	/* NOTREACHED */
736 }
737 
738 static struct nd_defrouter *
739 defrtrlist_update(struct nd_defrouter *newdr)
740 {
741 	struct nd_defrouter *dr, *n;
742 	struct in6_ifextra *ext = newdr->ifp->if_afdata[AF_INET6];
743 	int s = splsoftnet();
744 
745 	if ((dr = defrouter_lookup(&newdr->rtaddr, newdr->ifp)) != NULL) {
746 		/* entry exists */
747 		if (newdr->rtlifetime == 0) {
748 			defrtrlist_del(dr, ext);
749 			dr = NULL;
750 		} else {
751 			int oldpref = rtpref(dr);
752 
753 			/* override */
754 			dr->flags = newdr->flags; /* xxx flag check */
755 			dr->rtlifetime = newdr->rtlifetime;
756 			dr->expire = newdr->expire;
757 
758 			/*
759 			 * If the preference does not change, there's no need
760 			 * to sort the entries.
761 			 */
762 			if (rtpref(newdr) == oldpref) {
763 				splx(s);
764 				return (dr);
765 			}
766 
767 			/*
768 			 * preferred router may be changed, so relocate
769 			 * this router.
770 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
771 			 * However, since defrtrlist_del() has many side
772 			 * effects, we intentionally do so here.
773 			 * defrouter_select() below will handle routing
774 			 * changes later.
775 			 */
776 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
777 			n = dr;
778 			goto insert;
779 		}
780 		splx(s);
781 		return (dr);
782 	}
783 
784 	if (ip6_maxifdefrouters >= 0 &&
785 	    ext->ndefrouters >= ip6_maxifdefrouters) {
786 		splx(s);
787 		return (NULL);
788 	}
789 
790 	/* entry does not exist */
791 	if (newdr->rtlifetime == 0) {
792 		splx(s);
793 		return (NULL);
794 	}
795 
796 	if (ip6_rtadv_maxroutes <= nd6_numroutes) {
797 		ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
798 		splx(s);
799 		return (NULL);
800 	}
801 
802 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
803 	if (n == NULL) {
804 		splx(s);
805 		return (NULL);
806 	}
807 	memset(n, 0, sizeof(*n));
808 	*n = *newdr;
809 
810 insert:
811 	/*
812 	 * Insert the new router in the Default Router List;
813 	 * The Default Router List should be in the descending order
814 	 * of router-preferece.  Routers with the same preference are
815 	 * sorted in the arriving time order.
816 	 */
817 
818 	/* insert at the end of the group */
819 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
820 	     dr = TAILQ_NEXT(dr, dr_entry)) {
821 		if (rtpref(n) > rtpref(dr))
822 			break;
823 	}
824 	if (dr)
825 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
826 	else
827 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
828 
829 	defrouter_select();
830 
831 	ext->ndefrouters++;
832 
833 	splx(s);
834 
835 	return (n);
836 }
837 
838 static struct nd_pfxrouter *
839 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
840 {
841 	struct nd_pfxrouter *search;
842 
843 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
844 		if (search->router == dr)
845 			break;
846 	}
847 
848 	return (search);
849 }
850 
851 static void
852 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
853 {
854 	struct nd_pfxrouter *newpfr;
855 
856 	newpfr = malloc(sizeof(*newpfr), M_IP6NDP, M_NOWAIT|M_ZERO);
857 	if (newpfr == NULL)
858 		return;
859 	newpfr->router = dr;
860 
861 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, newpfr, pfr_entry);
862 
863 	pfxlist_onlink_check();
864 }
865 
866 static void
867 pfxrtr_del(struct nd_pfxrouter *pfr)
868 {
869 	LIST_REMOVE(pfr, pfr_entry);
870 	free(pfr, M_IP6NDP);
871 }
872 
873 struct nd_prefix *
874 nd6_prefix_lookup(struct nd_prefixctl *key)
875 {
876 	struct nd_prefix *search;
877 
878 	LIST_FOREACH(search, &nd_prefix, ndpr_entry) {
879 		if (key->ndprc_ifp == search->ndpr_ifp &&
880 		    key->ndprc_plen == search->ndpr_plen &&
881 		    in6_are_prefix_equal(&key->ndprc_prefix.sin6_addr,
882 		    &search->ndpr_prefix.sin6_addr, key->ndprc_plen)) {
883 			break;
884 		}
885 	}
886 
887 	return (search);
888 }
889 
890 static void
891 purge_detached(struct ifnet *ifp)
892 {
893 	struct nd_prefix *pr, *pr_next;
894 	struct in6_ifaddr *ia;
895 	struct ifaddr *ifa, *ifa_next;
896 
897 	for (pr = nd_prefix.lh_first; pr; pr = pr_next) {
898 		pr_next = pr->ndpr_next;
899 
900 		/*
901 		 * This function is called when we need to make more room for
902 		 * new prefixes rather than keeping old, possibly stale ones.
903 		 * Detached prefixes would be a good candidate; if all routers
904 		 * that advertised the prefix expired, the prefix is also
905 		 * probably stale.
906 		 */
907 		if (pr->ndpr_ifp != ifp ||
908 		    IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
909 		    ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
910 		    !LIST_EMPTY(&pr->ndpr_advrtrs)))
911 			continue;
912 
913 		IFADDR_FOREACH_SAFE(ifa, ifp, ifa_next) {
914 			if (ifa->ifa_addr->sa_family != AF_INET6)
915 				continue;
916 			ia = (struct in6_ifaddr *)ifa;
917 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) ==
918 			    IN6_IFF_AUTOCONF && ia->ia6_ndpr == pr) {
919 				in6_purgeaddr(ifa);
920 			}
921 		}
922 		if (pr->ndpr_refcnt == 0)
923 			prelist_remove(pr);
924 	}
925 }
926 int
927 nd6_prelist_add(struct nd_prefixctl *prc, struct nd_defrouter *dr,
928 	struct nd_prefix **newp)
929 {
930 	struct nd_prefix *newpr = NULL;
931 	int i, s;
932 	int error;
933 	struct in6_ifextra *ext = prc->ndprc_ifp->if_afdata[AF_INET6];
934 
935 	if (ip6_maxifprefixes >= 0) {
936 		if (ext->nprefixes >= ip6_maxifprefixes / 2)
937 			purge_detached(prc->ndprc_ifp);
938 		if (ext->nprefixes >= ip6_maxifprefixes)
939 			return ENOMEM;
940 	}
941 
942 	error = 0;
943 	newpr = malloc(sizeof(*newpr), M_IP6NDP, M_NOWAIT|M_ZERO);
944 	if (newpr == NULL)
945 		return ENOMEM;
946 	newpr->ndpr_ifp = prc->ndprc_ifp;
947 	newpr->ndpr_prefix = prc->ndprc_prefix;
948 	newpr->ndpr_plen = prc->ndprc_plen;
949 	newpr->ndpr_vltime = prc->ndprc_vltime;
950 	newpr->ndpr_pltime = prc->ndprc_pltime;
951 	newpr->ndpr_flags = prc->ndprc_flags;
952 	if ((error = in6_init_prefix_ltimes(newpr)) != 0) {
953 		free(newpr, M_IP6NDP);
954 		return(error);
955 	}
956 	newpr->ndpr_lastupdate = time_second;
957 	if (newp != NULL)
958 		*newp = newpr;
959 
960 	/* initialization */
961 	LIST_INIT(&newpr->ndpr_advrtrs);
962 	in6_prefixlen2mask(&newpr->ndpr_mask, newpr->ndpr_plen);
963 	/* make prefix in the canonical form */
964 	for (i = 0; i < 4; i++) {
965 		newpr->ndpr_prefix.sin6_addr.s6_addr32[i] &=
966 		    newpr->ndpr_mask.s6_addr32[i];
967 	}
968 
969 	s = splsoftnet();
970 	/* link ndpr_entry to nd_prefix list */
971 	LIST_INSERT_HEAD(&nd_prefix, newpr, ndpr_entry);
972 	splx(s);
973 
974 	/* ND_OPT_PI_FLAG_ONLINK processing */
975 	if (newpr->ndpr_raf_onlink) {
976 		int e;
977 
978 		if ((e = nd6_prefix_onlink(newpr)) != 0) {
979 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
980 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
981 			    ip6_sprintf(&prc->ndprc_prefix.sin6_addr),
982 			    prc->ndprc_plen, if_name(prc->ndprc_ifp), e));
983 			/* proceed anyway. XXX: is it correct? */
984 		}
985 	}
986 
987 	if (dr)
988 		pfxrtr_add(newpr, dr);
989 
990 	ext->nprefixes++;
991 
992 	return 0;
993 }
994 
995 void
996 prelist_remove(struct nd_prefix *pr)
997 {
998 	struct nd_pfxrouter *pfr, *next;
999 	int e, s;
1000 	struct in6_ifextra *ext = pr->ndpr_ifp->if_afdata[AF_INET6];
1001 
1002 	/* make sure to invalidate the prefix until it is really freed. */
1003 	pr->ndpr_vltime = 0;
1004 	pr->ndpr_pltime = 0;
1005 #if 0
1006 	/*
1007 	 * Though these flags are now meaningless, we'd rather keep the value
1008 	 * not to confuse users when executing "ndp -p".
1009 	 */
1010 	pr->ndpr_raf_onlink = 0;
1011 	pr->ndpr_raf_auto = 0;
1012 #endif
1013 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
1014 	    (e = nd6_prefix_offlink(pr)) != 0) {
1015 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
1016 		    "on %s, errno=%d\n",
1017 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1018 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1019 		/* what should we do? */
1020 	}
1021 
1022 	if (pr->ndpr_refcnt > 0)
1023 		return;		/* notice here? */
1024 
1025 	s = splsoftnet();
1026 	/* unlink ndpr_entry from nd_prefix list */
1027 	LIST_REMOVE(pr, ndpr_entry);
1028 
1029 	/* free list of routers that adversed the prefix */
1030 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
1031 		next = LIST_NEXT(pfr, pfr_entry);
1032 
1033 		free(pfr, M_IP6NDP);
1034 	}
1035 
1036 	if (ext) {
1037 		ext->nprefixes--;
1038 		if (ext->nprefixes < 0) {
1039 			log(LOG_WARNING, "prelist_remove: negative count on "
1040 			    "%s\n", pr->ndpr_ifp->if_xname);
1041 		}
1042 	}
1043 	splx(s);
1044 
1045 	free(pr, M_IP6NDP);
1046 
1047 	pfxlist_onlink_check();
1048 }
1049 
1050 static int
1051 prelist_update(struct nd_prefixctl *newprc,
1052 	struct nd_defrouter *dr, /* may be NULL */
1053 	struct mbuf *m,
1054 	int mcast)
1055 {
1056 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
1057 	struct ifaddr *ifa;
1058 	struct ifnet *ifp = newprc->ndprc_ifp;
1059 	struct nd_prefix *pr;
1060 	int s = splsoftnet();
1061 	int error = 0;
1062 	int auth;
1063 	struct in6_addrlifetime lt6_tmp;
1064 
1065 	auth = 0;
1066 	if (m) {
1067 		/*
1068 		 * Authenticity for NA consists authentication for
1069 		 * both IP header and IP datagrams, doesn't it ?
1070 		 */
1071 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1072 		auth = (m->m_flags & M_AUTHIPHDR
1073 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
1074 #endif
1075 	}
1076 
1077 	if ((pr = nd6_prefix_lookup(newprc)) != NULL) {
1078 		/*
1079 		 * nd6_prefix_lookup() ensures that pr and newprc have the same
1080 		 * prefix on a same interface.
1081 		 */
1082 
1083 		/*
1084 		 * Update prefix information.  Note that the on-link (L) bit
1085 		 * and the autonomous (A) bit should NOT be changed from 1
1086 		 * to 0.
1087 		 */
1088 		if (newprc->ndprc_raf_onlink == 1)
1089 			pr->ndpr_raf_onlink = 1;
1090 		if (newprc->ndprc_raf_auto == 1)
1091 			pr->ndpr_raf_auto = 1;
1092 		if (newprc->ndprc_raf_onlink) {
1093 			pr->ndpr_vltime = newprc->ndprc_vltime;
1094 			pr->ndpr_pltime = newprc->ndprc_pltime;
1095 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1096 			pr->ndpr_lastupdate = time_second;
1097 		}
1098 
1099 		if (newprc->ndprc_raf_onlink &&
1100 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1101 			int e;
1102 
1103 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1104 				nd6log((LOG_ERR,
1105 				    "prelist_update: failed to make "
1106 				    "the prefix %s/%d on-link on %s "
1107 				    "(errno=%d)\n",
1108 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1109 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1110 				/* proceed anyway. XXX: is it correct? */
1111 			}
1112 		}
1113 
1114 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1115 			pfxrtr_add(pr, dr);
1116 	} else {
1117 		struct nd_prefix *newpr = NULL;
1118 
1119 		if (newprc->ndprc_vltime == 0)
1120 			goto end;
1121 		if (newprc->ndprc_raf_onlink == 0 && newprc->ndprc_raf_auto == 0)
1122 			goto end;
1123 
1124 		if (ip6_rtadv_maxroutes <= nd6_numroutes) {
1125 			ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
1126 			goto end;
1127 		}
1128 
1129 		error = nd6_prelist_add(newprc, dr, &newpr);
1130 		if (error != 0 || newpr == NULL) {
1131 			nd6log((LOG_NOTICE, "prelist_update: "
1132 			    "nd6_prelist_add failed for %s/%d on %s "
1133 			    "errno=%d, returnpr=%p\n",
1134 			    ip6_sprintf(&newprc->ndprc_prefix.sin6_addr),
1135 			    newprc->ndprc_plen, if_name(newprc->ndprc_ifp),
1136 			    error, newpr));
1137 			goto end; /* we should just give up in this case. */
1138 		}
1139 
1140 		/*
1141 		 * XXX: from the ND point of view, we can ignore a prefix
1142 		 * with the on-link bit being zero.  However, we need a
1143 		 * prefix structure for references from autoconfigured
1144 		 * addresses.  Thus, we explicitly make sure that the prefix
1145 		 * itself expires now.
1146 		 */
1147 		if (newpr->ndpr_raf_onlink == 0) {
1148 			newpr->ndpr_vltime = 0;
1149 			newpr->ndpr_pltime = 0;
1150 			in6_init_prefix_ltimes(newpr);
1151 		}
1152 
1153 		pr = newpr;
1154 	}
1155 
1156 	/*
1157 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1158 	 * Note that pr must be non NULL at this point.
1159 	 */
1160 
1161 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1162 	if (!newprc->ndprc_raf_auto)
1163 		goto end;
1164 
1165 	/*
1166 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1167 	 * nd6_ra_input.
1168 	 */
1169 
1170 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1171 	if (newprc->ndprc_pltime > newprc->ndprc_vltime) {
1172 		error = EINVAL;	/* XXX: won't be used */
1173 		goto end;
1174 	}
1175 
1176 	/*
1177 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1178 	 * an address configured by stateless autoconfiguration already in the
1179 	 * list of addresses associated with the interface, and the Valid
1180 	 * Lifetime is not 0, form an address.  We first check if we have
1181 	 * a matching prefix.
1182 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1183 	 * consider autoconfigured addresses while RFC2462 simply said
1184 	 * "address".
1185 	 */
1186 	IFADDR_FOREACH(ifa, ifp) {
1187 		struct in6_ifaddr *ifa6;
1188 		u_int32_t remaininglifetime;
1189 
1190 		if (ifa->ifa_addr->sa_family != AF_INET6)
1191 			continue;
1192 
1193 		ifa6 = (struct in6_ifaddr *)ifa;
1194 
1195 		/*
1196 		 * We only consider autoconfigured addresses as per rfc2462bis.
1197 		 */
1198 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1199 			continue;
1200 
1201 		/*
1202 		 * Spec is not clear here, but I believe we should concentrate
1203 		 * on unicast (i.e. not anycast) addresses.
1204 		 * XXX: other ia6_flags? detached or duplicated?
1205 		 */
1206 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1207 			continue;
1208 
1209 		/*
1210 		 * Ignore the address if it is not associated with a prefix
1211 		 * or is associated with a prefix that is different from this
1212 		 * one.  (pr is never NULL here)
1213 		 */
1214 		if (ifa6->ia6_ndpr != pr)
1215 			continue;
1216 
1217 		if (ia6_match == NULL) /* remember the first one */
1218 			ia6_match = ifa6;
1219 
1220 		/*
1221 		 * An already autoconfigured address matched.  Now that we
1222 		 * are sure there is at least one matched address, we can
1223 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1224 		 * "two hours" rule and the privacy extension.
1225 		 * We apply some clarifications in rfc2462bis:
1226 		 * - use remaininglifetime instead of storedlifetime as a
1227 		 *   variable name
1228 		 * - remove the dead code in the "two-hour" rule
1229 		 */
1230 #define TWOHOUR		(120*60)
1231 		lt6_tmp = ifa6->ia6_lifetime;
1232 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1233 			remaininglifetime = ND6_INFINITE_LIFETIME;
1234 		else if (time_second - ifa6->ia6_updatetime >
1235 			 lt6_tmp.ia6t_vltime) {
1236 			/*
1237 			 * The case of "invalid" address.  We should usually
1238 			 * not see this case.
1239 			 */
1240 			remaininglifetime = 0;
1241 		} else
1242 			remaininglifetime = lt6_tmp.ia6t_vltime -
1243 			    (time_second - ifa6->ia6_updatetime);
1244 
1245 		/* when not updating, keep the current stored lifetime. */
1246 		lt6_tmp.ia6t_vltime = remaininglifetime;
1247 
1248 		if (TWOHOUR < newprc->ndprc_vltime ||
1249 		    remaininglifetime < newprc->ndprc_vltime) {
1250 			lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
1251 		} else if (remaininglifetime <= TWOHOUR) {
1252 			if (auth)
1253 				lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
1254 		} else {
1255 			/*
1256 			 * newprc->ndprc_vltime <= TWOHOUR &&
1257 			 * TWOHOUR < remaininglifetime
1258 			 */
1259 			lt6_tmp.ia6t_vltime = TWOHOUR;
1260 		}
1261 
1262 		/* The 2 hour rule is not imposed for preferred lifetime. */
1263 		lt6_tmp.ia6t_pltime = newprc->ndprc_pltime;
1264 
1265 		in6_init_address_ltimes(pr, &lt6_tmp);
1266 
1267 		/*
1268 		 * We need to treat lifetimes for temporary addresses
1269 		 * differently, according to
1270 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1271 		 * we only update the lifetimes when they are in the maximum
1272 		 * intervals.
1273 		 */
1274 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1275 			u_int32_t maxvltime, maxpltime;
1276 
1277 			if (ip6_temp_valid_lifetime >
1278 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1279 			    ip6_desync_factor)) {
1280 				maxvltime = ip6_temp_valid_lifetime -
1281 				    (time_second - ifa6->ia6_createtime) -
1282 				    ip6_desync_factor;
1283 			} else
1284 				maxvltime = 0;
1285 			if (ip6_temp_preferred_lifetime >
1286 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1287 			    ip6_desync_factor)) {
1288 				maxpltime = ip6_temp_preferred_lifetime -
1289 				    (time_second - ifa6->ia6_createtime) -
1290 				    ip6_desync_factor;
1291 			} else
1292 				maxpltime = 0;
1293 
1294 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1295 			    lt6_tmp.ia6t_vltime > maxvltime) {
1296 				lt6_tmp.ia6t_vltime = maxvltime;
1297 			}
1298 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1299 			    lt6_tmp.ia6t_pltime > maxpltime) {
1300 				lt6_tmp.ia6t_pltime = maxpltime;
1301 			}
1302 		}
1303 
1304 		ifa6->ia6_lifetime = lt6_tmp;
1305 		ifa6->ia6_updatetime = time_second;
1306 	}
1307 	if (ia6_match == NULL && newprc->ndprc_vltime) {
1308 		int ifidlen;
1309 
1310 		/*
1311 		 * 5.5.3 (d) (continued)
1312 		 * No address matched and the valid lifetime is non-zero.
1313 		 * Create a new address.
1314 		 */
1315 
1316 		/*
1317 		 * Prefix Length check:
1318 		 * If the sum of the prefix length and interface identifier
1319 		 * length does not equal 128 bits, the Prefix Information
1320 		 * option MUST be ignored.  The length of the interface
1321 		 * identifier is defined in a separate link-type specific
1322 		 * document.
1323 		 */
1324 		ifidlen = in6_if2idlen(ifp);
1325 		if (ifidlen < 0) {
1326 			/* this should not happen, so we always log it. */
1327 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1328 			    if_name(ifp));
1329 			goto end;
1330 		}
1331 		if (ifidlen + pr->ndpr_plen != 128) {
1332 			nd6log((LOG_INFO,
1333 			    "prelist_update: invalid prefixlen "
1334 			    "%d for %s, ignored\n",
1335 			    pr->ndpr_plen, if_name(ifp)));
1336 			goto end;
1337 		}
1338 
1339 		if ((ia6 = in6_ifadd(newprc, mcast)) != NULL) {
1340 			/*
1341 			 * note that we should use pr (not newprc) for reference.
1342 			 */
1343 			pr->ndpr_refcnt++;
1344 			ia6->ia6_ndpr = pr;
1345 
1346 			/*
1347 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
1348 			 * When a new public address is created as described
1349 			 * in RFC2462, also create a new temporary address.
1350 			 *
1351 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
1352 			 * When an interface connects to a new link, a new
1353 			 * randomized interface identifier should be generated
1354 			 * immediately together with a new set of temporary
1355 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1356 			 * in6_tmpifadd().
1357 			 */
1358 			if (ip6_use_tempaddr) {
1359 				int e;
1360 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1361 					nd6log((LOG_NOTICE, "prelist_update: "
1362 					    "failed to create a temporary "
1363 					    "address, errno=%d\n",
1364 					    e));
1365 				}
1366 			}
1367 
1368 			/*
1369 			 * A newly added address might affect the status
1370 			 * of other addresses, so we check and update it.
1371 			 * XXX: what if address duplication happens?
1372 			 */
1373 			pfxlist_onlink_check();
1374 		} else {
1375 			/* just set an error. do not bark here. */
1376 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1377 		}
1378 	}
1379 
1380  end:
1381 	splx(s);
1382 	return error;
1383 }
1384 
1385 /*
1386  * A supplement function used in the on-link detection below;
1387  * detect if a given prefix has a (probably) reachable advertising router.
1388  * XXX: lengthy function name...
1389  */
1390 static struct nd_pfxrouter *
1391 find_pfxlist_reachable_router(struct nd_prefix *pr)
1392 {
1393 	struct nd_pfxrouter *pfxrtr;
1394 	struct rtentry *rt;
1395 	struct llinfo_nd6 *ln;
1396 
1397 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1398 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1399 		if (pfxrtr->router->ifp->if_flags & IFF_UP &&
1400 		    pfxrtr->router->ifp->if_link_state != LINK_STATE_DOWN &&
1401 		    (rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1402 		    pfxrtr->router->ifp)) &&
1403 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1404 		    ND6_IS_LLINFO_PROBREACH(ln))
1405 			break;	/* found */
1406 	}
1407 
1408 	return (pfxrtr);
1409 }
1410 
1411 /*
1412  * Check if each prefix in the prefix list has at least one available router
1413  * that advertised the prefix (a router is "available" if its neighbor cache
1414  * entry is reachable or probably reachable).
1415  * If the check fails, the prefix may be off-link, because, for example,
1416  * we have moved from the network but the lifetime of the prefix has not
1417  * expired yet.  So we should not use the prefix if there is another prefix
1418  * that has an available router.
1419  * But, if there is no prefix that has an available router, we still regards
1420  * all the prefixes as on-link.  This is because we can't tell if all the
1421  * routers are simply dead or if we really moved from the network and there
1422  * is no router around us.
1423  */
1424 void
1425 pfxlist_onlink_check(void)
1426 {
1427 	struct nd_prefix *pr;
1428 	struct in6_ifaddr *ifa;
1429 	struct nd_defrouter *dr;
1430 	struct nd_pfxrouter *pfxrtr = NULL;
1431 
1432 	/*
1433 	 * Check if there is a prefix that has a reachable advertising
1434 	 * router.
1435 	 */
1436 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1437 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1438 			break;
1439 	}
1440 	/*
1441 	 * If we have no such prefix, check whether we still have a router
1442 	 * that does not advertise any prefixes.
1443 	 */
1444 	if (pr == NULL) {
1445 		TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1446 			struct nd_prefix *pr0;
1447 
1448 			LIST_FOREACH(pr0, &nd_prefix, ndpr_entry) {
1449 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1450 					break;
1451 			}
1452 			if (pfxrtr)
1453 				break;
1454 		}
1455 	}
1456 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
1457 		/*
1458 		 * There is at least one prefix that has a reachable router,
1459 		 * or at least a router which probably does not advertise
1460 		 * any prefixes.  The latter would be the case when we move
1461 		 * to a new link where we have a router that does not provide
1462 		 * prefixes and we configure an address by hand.
1463 		 * Detach prefixes which have no reachable advertising
1464 		 * router, and attach other prefixes.
1465 		 */
1466 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1467 			/* XXX: a link-local prefix should never be detached */
1468 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1469 				continue;
1470 
1471 			/*
1472 			 * we aren't interested in prefixes without the L bit
1473 			 * set.
1474 			 */
1475 			if (pr->ndpr_raf_onlink == 0)
1476 				continue;
1477 
1478 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1479 			    find_pfxlist_reachable_router(pr) == NULL)
1480 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1481 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1482 			    find_pfxlist_reachable_router(pr) != 0)
1483 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1484 		}
1485 	} else {
1486 		/* there is no prefix that has a reachable router */
1487 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1488 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1489 				continue;
1490 
1491 			if (pr->ndpr_raf_onlink == 0)
1492 				continue;
1493 
1494 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1495 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1496 		}
1497 	}
1498 
1499 	/*
1500 	 * Remove each interface route associated with a (just) detached
1501 	 * prefix, and reinstall the interface route for a (just) attached
1502 	 * prefix.  Note that all attempt of reinstallation does not
1503 	 * necessarily success, when a same prefix is shared among multiple
1504 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1505 	 * so we don't have to care about them.
1506 	 */
1507 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1508 		int e;
1509 
1510 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1511 			continue;
1512 
1513 		if (pr->ndpr_raf_onlink == 0)
1514 			continue;
1515 
1516 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1517 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1518 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1519 				nd6log((LOG_ERR,
1520 				    "pfxlist_onlink_check: failed to "
1521 				    "make %s/%d offlink, errno=%d\n",
1522 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1523 				    pr->ndpr_plen, e));
1524 			}
1525 		}
1526 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1527 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1528 		    pr->ndpr_raf_onlink) {
1529 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1530 				nd6log((LOG_ERR,
1531 				    "pfxlist_onlink_check: failed to "
1532 				    "make %s/%d onlink, errno=%d\n",
1533 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1534 				    pr->ndpr_plen, e));
1535 			}
1536 		}
1537 	}
1538 
1539 	/*
1540 	 * Changes on the prefix status might affect address status as well.
1541 	 * Make sure that all addresses derived from an attached prefix are
1542 	 * attached, and that all addresses derived from a detached prefix are
1543 	 * detached.  Note, however, that a manually configured address should
1544 	 * always be attached.
1545 	 * The precise detection logic is same as the one for prefixes.
1546 	 */
1547 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1548 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1549 			continue;
1550 
1551 		if (ifa->ia6_ndpr == NULL) {
1552 			/*
1553 			 * This can happen when we first configure the address
1554 			 * (i.e. the address exists, but the prefix does not).
1555 			 * XXX: complicated relationships...
1556 			 */
1557 			continue;
1558 		}
1559 
1560 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1561 			break;
1562 	}
1563 	if (ifa) {
1564 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1565 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1566 				continue;
1567 
1568 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1569 				continue;
1570 
1571 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1572 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1573 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1574 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1575 					nd6_dad_start((struct ifaddr *)ifa,
1576 					    0);
1577 					/* We will notify the routing socket
1578 					 * of the DAD result, so no need to
1579 					 * here */
1580 				}
1581 			} else {
1582 				if ((ifa->ia6_flags & IN6_IFF_DETACHED) == 0) {
1583 					ifa->ia6_flags |= IN6_IFF_DETACHED;
1584 					rt_newaddrmsg(RTM_NEWADDR,
1585 					    (struct ifaddr *)ifa, 0, NULL);
1586 				}
1587 			}
1588 		}
1589 	}
1590 	else {
1591 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1592 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1593 				continue;
1594 
1595 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1596 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1597 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1598 				/* Do we need a delay in this case? */
1599 				nd6_dad_start((struct ifaddr *)ifa, 0);
1600 			}
1601 		}
1602 	}
1603 }
1604 
1605 int
1606 nd6_prefix_onlink(struct nd_prefix *pr)
1607 {
1608 	struct ifaddr *ifa;
1609 	struct ifnet *ifp = pr->ndpr_ifp;
1610 	struct sockaddr_in6 mask6;
1611 	struct nd_prefix *opr;
1612 	u_long rtflags;
1613 	int error = 0;
1614 	struct rtentry *rt = NULL;
1615 
1616 	/* sanity check */
1617 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1618 		nd6log((LOG_ERR,
1619 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1620 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1621 		return (EEXIST);
1622 	}
1623 
1624 	/*
1625 	 * Add the interface route associated with the prefix.  Before
1626 	 * installing the route, check if there's the same prefix on another
1627 	 * interface, and the prefix has already installed the interface route.
1628 	 * Although such a configuration is expected to be rare, we explicitly
1629 	 * allow it.
1630 	 */
1631 	LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1632 		if (opr == pr)
1633 			continue;
1634 
1635 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1636 			continue;
1637 
1638 		if (opr->ndpr_plen == pr->ndpr_plen &&
1639 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1640 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1641 			return (0);
1642 	}
1643 
1644 	/*
1645 	 * We prefer link-local addresses as the associated interface address.
1646 	 */
1647 	/* search for a link-local addr */
1648 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1649 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1650 	if (ifa == NULL) {
1651 		/* XXX: freebsd does not have ifa_ifwithaf */
1652 		IFADDR_FOREACH(ifa, ifp) {
1653 			if (ifa->ifa_addr->sa_family == AF_INET6)
1654 				break;
1655 		}
1656 		/* should we care about ia6_flags? */
1657 	}
1658 	if (ifa == NULL) {
1659 		/*
1660 		 * This can still happen, when, for example, we receive an RA
1661 		 * containing a prefix with the L bit set and the A bit clear,
1662 		 * after removing all IPv6 addresses on the receiving
1663 		 * interface.  This should, of course, be rare though.
1664 		 */
1665 		nd6log((LOG_NOTICE,
1666 		    "nd6_prefix_onlink: failed to find any ifaddr"
1667 		    " to add route for a prefix(%s/%d) on %s\n",
1668 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1669 		    pr->ndpr_plen, if_name(ifp)));
1670 		return (0);
1671 	}
1672 
1673 	/*
1674 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1675 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1676 	 */
1677 	memset(&mask6, 0, sizeof(mask6));
1678 	mask6.sin6_family = AF_INET6;
1679 	mask6.sin6_len = sizeof(mask6);
1680 	mask6.sin6_addr = pr->ndpr_mask;
1681 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
1682 	rtflags = ifa->ifa_flags | RTF_UP;
1683 	if (nd6_need_cache(ifp)) {
1684 		/* explicitly set in case ifa_flags does not set the flag. */
1685 		rtflags |= RTF_CLONING;
1686 	} else {
1687 		/*
1688 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1689 		 */
1690 		rtflags &= ~RTF_CLONING;
1691 	}
1692 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1693 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1694 	if (error == 0) {
1695 		if (rt != NULL) { /* this should be non NULL, though */
1696 			rt_newmsg(RTM_ADD, rt);
1697 			nd6_numroutes++;
1698 		}
1699 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1700 	} else {
1701 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1702 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1703 		    "errno = %d\n",
1704 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1705 		    pr->ndpr_plen, if_name(ifp),
1706 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1707 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
1708 	}
1709 
1710 	if (rt != NULL)
1711 		rt->rt_refcnt--;
1712 
1713 	return (error);
1714 }
1715 
1716 int
1717 nd6_prefix_offlink(struct nd_prefix *pr)
1718 {
1719 	int error = 0;
1720 	struct ifnet *ifp = pr->ndpr_ifp;
1721 	struct nd_prefix *opr;
1722 	struct sockaddr_in6 sa6, mask6;
1723 	struct rtentry *rt = NULL;
1724 
1725 	/* sanity check */
1726 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1727 		nd6log((LOG_ERR,
1728 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1729 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1730 		return (EEXIST);
1731 	}
1732 
1733 	sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1734 	sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
1735 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1736 	    (struct sockaddr *)&mask6, 0, &rt);
1737 	if (error == 0) {
1738 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1739 
1740 		/* report the route deletion to the routing socket. */
1741 		if (rt != NULL) {
1742 			rt_newmsg(RTM_DELETE, rt);
1743 			nd6_numroutes--;
1744 		}
1745 
1746 		/*
1747 		 * There might be the same prefix on another interface,
1748 		 * the prefix which could not be on-link just because we have
1749 		 * the interface route (see comments in nd6_prefix_onlink).
1750 		 * If there's one, try to make the prefix on-link on the
1751 		 * interface.
1752 		 */
1753 		LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1754 			if (opr == pr)
1755 				continue;
1756 
1757 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1758 				continue;
1759 
1760 			/*
1761 			 * KAME specific: detached prefixes should not be
1762 			 * on-link.
1763 			 */
1764 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1765 				continue;
1766 
1767 			if (opr->ndpr_plen == pr->ndpr_plen &&
1768 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1769 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1770 				int e;
1771 
1772 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1773 					nd6log((LOG_ERR,
1774 					    "nd6_prefix_offlink: failed to "
1775 					    "recover a prefix %s/%d from %s "
1776 					    "to %s (errno = %d)\n",
1777 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1778 					    opr->ndpr_plen, if_name(ifp),
1779 					    if_name(opr->ndpr_ifp), e));
1780 				}
1781 			}
1782 		}
1783 	} else {
1784 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1785 		nd6log((LOG_ERR,
1786 		    "nd6_prefix_offlink: failed to delete route: "
1787 		    "%s/%d on %s (errno = %d)\n",
1788 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1789 		    error));
1790 	}
1791 
1792 	if (rt != NULL) {
1793 		if (rt->rt_refcnt <= 0) {
1794 			/* XXX: we should free the entry ourselves. */
1795 			rt->rt_refcnt++;
1796 			rtfree(rt);
1797 		}
1798 	}
1799 
1800 	return (error);
1801 }
1802 
1803 static struct in6_ifaddr *
1804 in6_ifadd(struct nd_prefixctl *prc, int mcast)
1805 {
1806 	struct ifnet *ifp = prc->ndprc_ifp;
1807 	struct ifaddr *ifa;
1808 	struct in6_aliasreq ifra;
1809 	struct in6_ifaddr *ia, *ib;
1810 	int error, plen0;
1811 	struct in6_addr mask;
1812 	int prefixlen = prc->ndprc_plen;
1813 	int updateflags;
1814 
1815 	in6_prefixlen2mask(&mask, prefixlen);
1816 
1817 	/*
1818 	 * find a link-local address (will be interface ID).
1819 	 * Is it really mandatory? Theoretically, a global or a site-local
1820 	 * address can be configured without a link-local address, if we
1821 	 * have a unique interface identifier...
1822 	 *
1823 	 * it is not mandatory to have a link-local address, we can generate
1824 	 * interface identifier on the fly.  we do this because:
1825 	 * (1) it should be the easiest way to find interface identifier.
1826 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1827 	 * for multiple addresses on a single interface, and possible shortcut
1828 	 * of DAD.  we omitted DAD for this reason in the past.
1829 	 * (3) a user can prevent autoconfiguration of global address
1830 	 * by removing link-local address by hand (this is partly because we
1831 	 * don't have other way to control the use of IPv6 on an interface.
1832 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1833 	 * (4) it is easier to manage when an interface has addresses
1834 	 * with the same interface identifier, than to have multiple addresses
1835 	 * with different interface identifiers.
1836 	 */
1837 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1838 	if (ifa)
1839 		ib = (struct in6_ifaddr *)ifa;
1840 	else
1841 		return NULL;
1842 
1843 #if 0 /* don't care link local addr state, and always do DAD */
1844 	/* if link-local address is not eligible, do not autoconfigure. */
1845 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1846 		printf("in6_ifadd: link-local address not ready\n");
1847 		return NULL;
1848 	}
1849 #endif
1850 
1851 	/* prefixlen + ifidlen must be equal to 128 */
1852 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1853 	if (prefixlen != plen0) {
1854 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1855 		    "(prefix=%d ifid=%d)\n",
1856 		    if_name(ifp), prefixlen, 128 - plen0));
1857 		return NULL;
1858 	}
1859 
1860 	/* make ifaddr */
1861 
1862 	memset(&ifra, 0, sizeof(ifra));
1863 	/*
1864 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1865 	 * for safety.
1866 	 */
1867 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1868 	sockaddr_in6_init(&ifra.ifra_addr, &prc->ndprc_prefix.sin6_addr, 0, 0, 0);
1869 	/* prefix */
1870 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1871 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1872 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1873 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1874 
1875 	/* interface ID */
1876 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1877 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1878 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1879 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1880 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1881 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1882 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1883 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1884 
1885 	/* new prefix mask. */
1886 	sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
1887 
1888 	/* lifetimes */
1889 	ifra.ifra_lifetime.ia6t_vltime = prc->ndprc_vltime;
1890 	ifra.ifra_lifetime.ia6t_pltime = prc->ndprc_pltime;
1891 
1892 	/* XXX: scope zone ID? */
1893 
1894 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1895 
1896 	/*
1897 	 * Make sure that we do not have this address already.  This should
1898 	 * usually not happen, but we can still see this case, e.g., if we
1899 	 * have manually configured the exact address to be configured.
1900 	 */
1901 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1902 		/* this should be rare enough to make an explicit log */
1903 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1904 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1905 		return (NULL);
1906 	}
1907 
1908 	/*
1909 	 * Allocate ifaddr structure, link into chain, etc.
1910 	 * If we are going to create a new address upon receiving a multicasted
1911 	 * RA, we need to impose a random delay before starting DAD.
1912 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1913 	 */
1914 	updateflags = 0;
1915 	if (mcast)
1916 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1917 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1918 		nd6log((LOG_ERR,
1919 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1920 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1921 		    error));
1922 		return (NULL);	/* ifaddr must not have been allocated. */
1923 	}
1924 
1925 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1926 
1927 	return (ia);		/* this is always non-NULL */
1928 }
1929 
1930 int
1931 in6_tmpifadd(
1932 	const struct in6_ifaddr *ia0, /* corresponding public address */
1933 	int forcegen,
1934 	int dad_delay)
1935 {
1936 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1937 	struct in6_ifaddr *newia, *ia;
1938 	struct in6_aliasreq ifra;
1939 	int i, error;
1940 	int trylimit = 3;	/* XXX: adhoc value */
1941 	int updateflags;
1942 	u_int32_t randid[2];
1943 	u_int32_t vltime0, pltime0;
1944 
1945 	memset(&ifra, 0, sizeof(ifra));
1946 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1947 	ifra.ifra_addr = ia0->ia_addr;
1948 	/* copy prefix mask */
1949 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1950 	/* clear the old IFID */
1951 	for (i = 0; i < 4; i++) {
1952 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1953 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1954 	}
1955 
1956   again:
1957 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1958 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1959 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1960 		    "random IFID\n"));
1961 		return (EINVAL);
1962 	}
1963 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1964 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1965 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1966 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1967 
1968 	/*
1969 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1970 	 * However, we may still have a chance to see collision, because
1971 	 * there may be a time lag between generation of the ID and generation
1972 	 * of the address.  So, we'll do one more sanity check.
1973 	 */
1974 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1975 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1976 		    &ifra.ifra_addr.sin6_addr)) {
1977 			if (trylimit-- == 0) {
1978 				/*
1979 				 * Give up.  Something strange should have
1980 				 * happened.
1981 				 */
1982 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1983 				    "find a unique random IFID\n"));
1984 				return (EEXIST);
1985 			}
1986 			forcegen = 1;
1987 			goto again;
1988 		}
1989 	}
1990 
1991 	/*
1992 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1993          * public address or TEMP_VALID_LIFETIME.
1994 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1995          * of the public address or TEMP_PREFERRED_LIFETIME -
1996          * DESYNC_FACTOR.
1997 	 */
1998 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1999 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2000 		    (ia0->ia6_lifetime.ia6t_vltime -
2001 		    (time_second - ia0->ia6_updatetime));
2002 		if (vltime0 > ip6_temp_valid_lifetime)
2003 			vltime0 = ip6_temp_valid_lifetime;
2004 	} else
2005 		vltime0 = ip6_temp_valid_lifetime;
2006 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2007 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2008 		    (ia0->ia6_lifetime.ia6t_pltime -
2009 		    (time_second - ia0->ia6_updatetime));
2010 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
2011 			pltime0 = ip6_temp_preferred_lifetime -
2012 			    ip6_desync_factor;
2013 		}
2014 	} else
2015 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
2016 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
2017 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
2018 
2019 	/*
2020 	 * A temporary address is created only if this calculated Preferred
2021 	 * Lifetime is greater than REGEN_ADVANCE time units.
2022 	 */
2023 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
2024 		return (0);
2025 
2026 	/* XXX: scope zone ID? */
2027 
2028 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2029 
2030 	/* allocate ifaddr structure, link into chain, etc. */
2031 	updateflags = 0;
2032 	if (dad_delay)
2033 		updateflags |= IN6_IFAUPDATE_DADDELAY;
2034 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2035 		return (error);
2036 
2037 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2038 	if (newia == NULL) {	/* XXX: can it happen? */
2039 		nd6log((LOG_ERR,
2040 		    "in6_tmpifadd: ifa update succeeded, but we got "
2041 		    "no ifaddr\n"));
2042 		return (EINVAL); /* XXX */
2043 	}
2044 	newia->ia6_ndpr = ia0->ia6_ndpr;
2045 	newia->ia6_ndpr->ndpr_refcnt++;
2046 
2047 	/*
2048 	 * A newly added address might affect the status of other addresses.
2049 	 * XXX: when the temporary address is generated with a new public
2050 	 * address, the onlink check is redundant.  However, it would be safe
2051 	 * to do the check explicitly everywhere a new address is generated,
2052 	 * and, in fact, we surely need the check when we create a new
2053 	 * temporary address due to deprecation of an old temporary address.
2054 	 */
2055 	pfxlist_onlink_check();
2056 
2057 	return (0);
2058 }
2059 
2060 static int
2061 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2062 {
2063 
2064 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
2065 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
2066 		nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
2067 		    "(%d) is greater than valid lifetime(%d)\n",
2068 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
2069 		return (EINVAL);
2070 	}
2071 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2072 		ndpr->ndpr_preferred = 0;
2073 	else
2074 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2075 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2076 		ndpr->ndpr_expire = 0;
2077 	else
2078 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2079 
2080 	return 0;
2081 }
2082 
2083 static void
2084 in6_init_address_ltimes(struct nd_prefix *newpr,
2085     struct in6_addrlifetime *lt6)
2086 {
2087 
2088 	/* Valid lifetime must not be updated unless explicitly specified. */
2089 	/* init ia6t_expire */
2090 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2091 		lt6->ia6t_expire = 0;
2092 	else {
2093 		lt6->ia6t_expire = time_second;
2094 		lt6->ia6t_expire += lt6->ia6t_vltime;
2095 	}
2096 
2097 	/* init ia6t_preferred */
2098 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2099 		lt6->ia6t_preferred = 0;
2100 	else {
2101 		lt6->ia6t_preferred = time_second;
2102 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2103 	}
2104 }
2105 
2106 /*
2107  * Delete all the routing table entries that use the specified gateway.
2108  * XXX: this function causes search through all entries of routing table, so
2109  * it shouldn't be called when acting as a router.
2110  */
2111 void
2112 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2113 {
2114 	int s = splsoftnet();
2115 
2116 	/* We'll care only link-local addresses */
2117 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2118 		splx(s);
2119 		return;
2120 	}
2121 
2122 	rt_walktree(AF_INET6, rt6_deleteroute, (void *)gateway);
2123 	splx(s);
2124 }
2125 
2126 static int
2127 rt6_deleteroute(struct rtentry *rt, void *arg)
2128 {
2129 #define SIN6(s)	((struct sockaddr_in6 *)s)
2130 	struct in6_addr *gate = (struct in6_addr *)arg;
2131 
2132 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2133 		return (0);
2134 
2135 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr))
2136 		return (0);
2137 
2138 	/*
2139 	 * Do not delete a static route.
2140 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2141 	 * 'cloned' bit instead?
2142 	 */
2143 	if ((rt->rt_flags & RTF_STATIC) != 0)
2144 		return (0);
2145 
2146 	/*
2147 	 * We delete only host route. This means, in particular, we don't
2148 	 * delete default route.
2149 	 */
2150 	if ((rt->rt_flags & RTF_HOST) == 0)
2151 		return (0);
2152 
2153 	return (rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
2154 	    rt_mask(rt), rt->rt_flags, 0));
2155 #undef SIN6
2156 }
2157 
2158 int
2159 nd6_setdefaultiface(int ifindex)
2160 {
2161 	ifnet_t *ifp;
2162 	int error = 0;
2163 
2164 	if ((ifp = if_byindex(ifindex)) == NULL) {
2165 		return EINVAL;
2166 	}
2167 	if (nd6_defifindex != ifindex) {
2168 		nd6_defifindex = ifindex;
2169 		nd6_defifp = nd6_defifindex > 0 ? ifp : NULL;
2170 
2171 		/*
2172 		 * Our current implementation assumes one-to-one maping between
2173 		 * interfaces and links, so it would be natural to use the
2174 		 * default interface as the default link.
2175 		 */
2176 		scope6_setdefault(nd6_defifp);
2177 	}
2178 
2179 	return (error);
2180 }
2181