xref: /freebsd/sys/netinet6/nd6_rtr.c (revision a9148abd9da5db2f1c682fb17bed791845fc41c9)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/socket.h>
43 #include <sys/sockio.h>
44 #include <sys/time.h>
45 #include <sys/kernel.h>
46 #include <sys/errno.h>
47 #include <sys/syslog.h>
48 #include <sys/queue.h>
49 #include <sys/vimage.h>
50 
51 #include <net/if.h>
52 #include <net/if_types.h>
53 #include <net/if_dl.h>
54 #include <net/route.h>
55 #include <net/radix.h>
56 
57 #include <netinet/in.h>
58 #include <netinet6/in6_var.h>
59 #include <netinet6/in6_ifattach.h>
60 #include <netinet/ip6.h>
61 #include <netinet6/ip6_var.h>
62 #include <netinet6/nd6.h>
63 #include <netinet/icmp6.h>
64 #include <netinet6/scope6_var.h>
65 
66 #define SDL(s)	((struct sockaddr_dl *)s)
67 
68 static int rtpref(struct nd_defrouter *);
69 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
70 static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
71     struct mbuf *, int));
72 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *,	int);
73 static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
74 	struct nd_defrouter *));
75 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
76 static void pfxrtr_del(struct nd_pfxrouter *);
77 static struct nd_pfxrouter *find_pfxlist_reachable_router
78 (struct nd_prefix *);
79 static void defrouter_delreq(struct nd_defrouter *);
80 static void nd6_rtmsg(int, struct rtentry *);
81 
82 static int in6_init_prefix_ltimes(struct nd_prefix *);
83 static void in6_init_address_ltimes __P((struct nd_prefix *,
84 	struct in6_addrlifetime *));
85 
86 static int rt6_deleteroute(struct radix_node *, void *);
87 
88 extern int nd6_recalc_reachtm_interval;
89 
90 static struct ifnet *nd6_defifp;
91 int nd6_defifindex;
92 
93 int ip6_use_tempaddr = 0;
94 
95 int ip6_desync_factor;
96 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
97 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
98 /*
99  * shorter lifetimes for debugging purposes.
100 int ip6_temp_preferred_lifetime = 800;
101 static int ip6_temp_valid_lifetime = 1800;
102 */
103 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
104 
105 /* RTPREF_MEDIUM has to be 0! */
106 #define RTPREF_HIGH	1
107 #define RTPREF_MEDIUM	0
108 #define RTPREF_LOW	(-1)
109 #define RTPREF_RESERVED	(-2)
110 #define RTPREF_INVALID	(-3)	/* internal */
111 
112 /*
113  * Receive Router Solicitation Message - just for routers.
114  * Router solicitation/advertisement is mostly managed by userland program
115  * (rtadvd) so here we have no function like nd6_ra_output().
116  *
117  * Based on RFC 2461
118  */
119 void
120 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
121 {
122 	INIT_VNET_INET6(curvnet);
123 	struct ifnet *ifp = m->m_pkthdr.rcvif;
124 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
125 	struct nd_router_solicit *nd_rs;
126 	struct in6_addr saddr6 = ip6->ip6_src;
127 	char *lladdr = NULL;
128 	int lladdrlen = 0;
129 	union nd_opts ndopts;
130 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
131 
132 	/* If I'm not a router, ignore it. */
133 	if (V_ip6_accept_rtadv != 0 || V_ip6_forwarding != 1)
134 		goto freeit;
135 
136 	/* Sanity checks */
137 	if (ip6->ip6_hlim != 255) {
138 		nd6log((LOG_ERR,
139 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
140 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
141 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
142 		goto bad;
143 	}
144 
145 	/*
146 	 * Don't update the neighbor cache, if src = ::.
147 	 * This indicates that the src has no IP address assigned yet.
148 	 */
149 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
150 		goto freeit;
151 
152 #ifndef PULLDOWN_TEST
153 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
154 	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
155 #else
156 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
157 	if (nd_rs == NULL) {
158 		V_icmp6stat.icp6s_tooshort++;
159 		return;
160 	}
161 #endif
162 
163 	icmp6len -= sizeof(*nd_rs);
164 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
165 	if (nd6_options(&ndopts) < 0) {
166 		nd6log((LOG_INFO,
167 		    "nd6_rs_input: invalid ND option, ignored\n"));
168 		/* nd6_options have incremented stats */
169 		goto freeit;
170 	}
171 
172 	if (ndopts.nd_opts_src_lladdr) {
173 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
174 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
175 	}
176 
177 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
178 		nd6log((LOG_INFO,
179 		    "nd6_rs_input: lladdrlen mismatch for %s "
180 		    "(if %d, RS packet %d)\n",
181 		    ip6_sprintf(ip6bufs, &saddr6),
182 		    ifp->if_addrlen, lladdrlen - 2));
183 		goto bad;
184 	}
185 
186 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
187 
188  freeit:
189 	m_freem(m);
190 	return;
191 
192  bad:
193 	V_icmp6stat.icp6s_badrs++;
194 	m_freem(m);
195 }
196 
197 /*
198  * Receive Router Advertisement Message.
199  *
200  * Based on RFC 2461
201  * TODO: on-link bit on prefix information
202  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
203  */
204 void
205 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
206 {
207 	INIT_VNET_INET6(curvnet);
208 	struct ifnet *ifp = m->m_pkthdr.rcvif;
209 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
210 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
211 	struct nd_router_advert *nd_ra;
212 	struct in6_addr saddr6 = ip6->ip6_src;
213 	int mcast = 0;
214 	union nd_opts ndopts;
215 	struct nd_defrouter *dr;
216 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
217 
218 	/*
219 	 * We only accept RAs only when
220 	 * the system-wide variable allows the acceptance, and
221 	 * per-interface variable allows RAs on the receiving interface.
222 	 */
223 	if (V_ip6_accept_rtadv == 0)
224 		goto freeit;
225 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
226 		goto freeit;
227 
228 	if (ip6->ip6_hlim != 255) {
229 		nd6log((LOG_ERR,
230 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
231 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
232 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
233 		goto bad;
234 	}
235 
236 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
237 		nd6log((LOG_ERR,
238 		    "nd6_ra_input: src %s is not link-local\n",
239 		    ip6_sprintf(ip6bufs, &saddr6)));
240 		goto bad;
241 	}
242 
243 #ifndef PULLDOWN_TEST
244 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
245 	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
246 #else
247 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
248 	if (nd_ra == NULL) {
249 		V_icmp6stat.icp6s_tooshort++;
250 		return;
251 	}
252 #endif
253 
254 	icmp6len -= sizeof(*nd_ra);
255 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
256 	if (nd6_options(&ndopts) < 0) {
257 		nd6log((LOG_INFO,
258 		    "nd6_ra_input: invalid ND option, ignored\n"));
259 		/* nd6_options have incremented stats */
260 		goto freeit;
261 	}
262 
263     {
264 	struct nd_defrouter dr0;
265 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
266 
267 	/* remember if this is a multicasted advertisement */
268 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
269 		mcast = 1;
270 
271 	bzero(&dr0, sizeof(dr0));
272 	dr0.rtaddr = saddr6;
273 	dr0.flags  = nd_ra->nd_ra_flags_reserved;
274 	dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
275 	dr0.expire = time_second + dr0.rtlifetime;
276 	dr0.ifp = ifp;
277 	/* unspecified or not? (RFC 2461 6.3.4) */
278 	if (advreachable) {
279 		advreachable = ntohl(advreachable);
280 		if (advreachable <= MAX_REACHABLE_TIME &&
281 		    ndi->basereachable != advreachable) {
282 			ndi->basereachable = advreachable;
283 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
284 			ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
285 		}
286 	}
287 	if (nd_ra->nd_ra_retransmit)
288 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
289 	if (nd_ra->nd_ra_curhoplimit)
290 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
291 	dr = defrtrlist_update(&dr0);
292     }
293 
294 	/*
295 	 * prefix
296 	 */
297 	if (ndopts.nd_opts_pi) {
298 		struct nd_opt_hdr *pt;
299 		struct nd_opt_prefix_info *pi = NULL;
300 		struct nd_prefixctl pr;
301 
302 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
303 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
304 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
305 						(pt->nd_opt_len << 3))) {
306 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
307 				continue;
308 			pi = (struct nd_opt_prefix_info *)pt;
309 
310 			if (pi->nd_opt_pi_len != 4) {
311 				nd6log((LOG_INFO,
312 				    "nd6_ra_input: invalid option "
313 				    "len %d for prefix information option, "
314 				    "ignored\n", pi->nd_opt_pi_len));
315 				continue;
316 			}
317 
318 			if (128 < pi->nd_opt_pi_prefix_len) {
319 				nd6log((LOG_INFO,
320 				    "nd6_ra_input: invalid prefix "
321 				    "len %d for prefix information option, "
322 				    "ignored\n", pi->nd_opt_pi_prefix_len));
323 				continue;
324 			}
325 
326 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
327 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
328 				nd6log((LOG_INFO,
329 				    "nd6_ra_input: invalid prefix "
330 				    "%s, ignored\n",
331 				    ip6_sprintf(ip6bufs,
332 					&pi->nd_opt_pi_prefix)));
333 				continue;
334 			}
335 
336 			bzero(&pr, sizeof(pr));
337 			pr.ndpr_prefix.sin6_family = AF_INET6;
338 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
339 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
340 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
341 
342 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
343 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
344 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
345 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
346 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
347 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
348 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
349 			(void)prelist_update(&pr, dr, m, mcast);
350 		}
351 	}
352 
353 	/*
354 	 * MTU
355 	 */
356 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
357 		u_long mtu;
358 		u_long maxmtu;
359 
360 		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
361 
362 		/* lower bound */
363 		if (mtu < IPV6_MMTU) {
364 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
365 			    "mtu=%lu sent from %s, ignoring\n",
366 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
367 			goto skip;
368 		}
369 
370 		/* upper bound */
371 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
372 		    ? ndi->maxmtu : ifp->if_mtu;
373 		if (mtu <= maxmtu) {
374 			int change = (ndi->linkmtu != mtu);
375 
376 			ndi->linkmtu = mtu;
377 			if (change) /* in6_maxmtu may change */
378 				in6_setmaxmtu();
379 		} else {
380 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
381 			    "mtu=%lu sent from %s; "
382 			    "exceeds maxmtu %lu, ignoring\n",
383 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
384 		}
385 	}
386 
387  skip:
388 
389 	/*
390 	 * Source link layer address
391 	 */
392     {
393 	char *lladdr = NULL;
394 	int lladdrlen = 0;
395 
396 	if (ndopts.nd_opts_src_lladdr) {
397 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
398 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
399 	}
400 
401 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
402 		nd6log((LOG_INFO,
403 		    "nd6_ra_input: lladdrlen mismatch for %s "
404 		    "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
405 		    ifp->if_addrlen, lladdrlen - 2));
406 		goto bad;
407 	}
408 
409 	nd6_cache_lladdr(ifp, &saddr6, lladdr,
410 	    lladdrlen, ND_ROUTER_ADVERT, 0);
411 
412 	/*
413 	 * Installing a link-layer address might change the state of the
414 	 * router's neighbor cache, which might also affect our on-link
415 	 * detection of adveritsed prefixes.
416 	 */
417 	pfxlist_onlink_check();
418     }
419 
420  freeit:
421 	m_freem(m);
422 	return;
423 
424  bad:
425 	V_icmp6stat.icp6s_badra++;
426 	m_freem(m);
427 }
428 
429 /*
430  * default router list proccessing sub routines
431  */
432 
433 /* tell the change to user processes watching the routing socket. */
434 static void
435 nd6_rtmsg(int cmd, struct rtentry *rt)
436 {
437 	struct rt_addrinfo info;
438 
439 	bzero((caddr_t)&info, sizeof(info));
440 	info.rti_info[RTAX_DST] = rt_key(rt);
441 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
442 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
443 	if (rt->rt_ifp) {
444 		info.rti_info[RTAX_IFP] =
445 		    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
446 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
447 	}
448 
449 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
450 }
451 
452 void
453 defrouter_addreq(struct nd_defrouter *new)
454 {
455 	struct sockaddr_in6 def, mask, gate;
456 	struct rtentry *newrt = NULL;
457 	int s;
458 	int error;
459 
460 	bzero(&def, sizeof(def));
461 	bzero(&mask, sizeof(mask));
462 	bzero(&gate, sizeof(gate));
463 
464 	def.sin6_len = mask.sin6_len = gate.sin6_len =
465 	    sizeof(struct sockaddr_in6);
466 	def.sin6_family = gate.sin6_family = AF_INET6;
467 	gate.sin6_addr = new->rtaddr;
468 
469 	s = splnet();
470 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
471 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
472 	    RTF_GATEWAY, &newrt);
473 	if (newrt) {
474 		RT_LOCK(newrt);
475 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
476 		RT_REMREF(newrt);
477 		RT_UNLOCK(newrt);
478 	}
479 	if (error == 0)
480 		new->installed = 1;
481 	splx(s);
482 	return;
483 }
484 
485 struct nd_defrouter *
486 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
487 {
488 	INIT_VNET_INET6(ifp->if_vnet);
489 	struct nd_defrouter *dr;
490 
491 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
492 	     dr = TAILQ_NEXT(dr, dr_entry)) {
493 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
494 			return (dr);
495 	}
496 
497 	return (NULL);		/* search failed */
498 }
499 
500 /*
501  * Remove the default route for a given router.
502  * This is just a subroutine function for defrouter_select(), and should
503  * not be called from anywhere else.
504  */
505 static void
506 defrouter_delreq(struct nd_defrouter *dr)
507 {
508 	struct sockaddr_in6 def, mask, gate;
509 	struct rtentry *oldrt = NULL;
510 
511 	bzero(&def, sizeof(def));
512 	bzero(&mask, sizeof(mask));
513 	bzero(&gate, sizeof(gate));
514 
515 	def.sin6_len = mask.sin6_len = gate.sin6_len =
516 	    sizeof(struct sockaddr_in6);
517 	def.sin6_family = gate.sin6_family = AF_INET6;
518 	gate.sin6_addr = dr->rtaddr;
519 
520 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
521 	    (struct sockaddr *)&gate,
522 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
523 	if (oldrt) {
524 		nd6_rtmsg(RTM_DELETE, oldrt);
525 		RTFREE(oldrt);
526 	}
527 
528 	dr->installed = 0;
529 }
530 
531 /*
532  * remove all default routes from default router list
533  */
534 void
535 defrouter_reset(void)
536 {
537 	INIT_VNET_INET6(curvnet);
538 	struct nd_defrouter *dr;
539 
540 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
541 	     dr = TAILQ_NEXT(dr, dr_entry))
542 		defrouter_delreq(dr);
543 
544 	/*
545 	 * XXX should we also nuke any default routers in the kernel, by
546 	 * going through them by rtalloc1()?
547 	 */
548 }
549 
550 void
551 defrtrlist_del(struct nd_defrouter *dr)
552 {
553 	INIT_VNET_INET6(curvnet);
554 	struct nd_defrouter *deldr = NULL;
555 	struct nd_prefix *pr;
556 
557 	/*
558 	 * Flush all the routing table entries that use the router
559 	 * as a next hop.
560 	 */
561 	if (!V_ip6_forwarding && V_ip6_accept_rtadv) /* XXX: better condition? */
562 		rt6_flush(&dr->rtaddr, dr->ifp);
563 
564 	if (dr->installed) {
565 		deldr = dr;
566 		defrouter_delreq(dr);
567 	}
568 	TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
569 
570 	/*
571 	 * Also delete all the pointers to the router in each prefix lists.
572 	 */
573 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
574 		struct nd_pfxrouter *pfxrtr;
575 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
576 			pfxrtr_del(pfxrtr);
577 	}
578 	pfxlist_onlink_check();
579 
580 	/*
581 	 * If the router is the primary one, choose a new one.
582 	 * Note that defrouter_select() will remove the current gateway
583 	 * from the routing table.
584 	 */
585 	if (deldr)
586 		defrouter_select();
587 
588 	free(dr, M_IP6NDP);
589 }
590 
591 /*
592  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
593  * draft-ietf-ipngwg-router-selection:
594  * 1) Routers that are reachable or probably reachable should be preferred.
595  *    If we have more than one (probably) reachable router, prefer ones
596  *    with the highest router preference.
597  * 2) When no routers on the list are known to be reachable or
598  *    probably reachable, routers SHOULD be selected in a round-robin
599  *    fashion, regardless of router preference values.
600  * 3) If the Default Router List is empty, assume that all
601  *    destinations are on-link.
602  *
603  * We assume nd_defrouter is sorted by router preference value.
604  * Since the code below covers both with and without router preference cases,
605  * we do not need to classify the cases by ifdef.
606  *
607  * At this moment, we do not try to install more than one default router,
608  * even when the multipath routing is available, because we're not sure about
609  * the benefits for stub hosts comparing to the risk of making the code
610  * complicated and the possibility of introducing bugs.
611  */
612 void
613 defrouter_select(void)
614 {
615 	INIT_VNET_INET6(curvnet);
616 	int s = splnet();
617 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
618 	struct rtentry *rt = NULL;
619 	struct llinfo_nd6 *ln = NULL;
620 
621 	/*
622 	 * This function should be called only when acting as an autoconfigured
623 	 * host.  Although the remaining part of this function is not effective
624 	 * if the node is not an autoconfigured host, we explicitly exclude
625 	 * such cases here for safety.
626 	 */
627 	if (V_ip6_forwarding || !V_ip6_accept_rtadv) {
628 		nd6log((LOG_WARNING,
629 		    "defrouter_select: called unexpectedly (forwarding=%d, "
630 		    "accept_rtadv=%d)\n", V_ip6_forwarding, V_ip6_accept_rtadv));
631 		splx(s);
632 		return;
633 	}
634 
635 	/*
636 	 * Let's handle easy case (3) first:
637 	 * If default router list is empty, there's nothing to be done.
638 	 */
639 	if (!TAILQ_FIRST(&V_nd_defrouter)) {
640 		splx(s);
641 		return;
642 	}
643 
644 	/*
645 	 * Search for a (probably) reachable router from the list.
646 	 * We just pick up the first reachable one (if any), assuming that
647 	 * the ordering rule of the list described in defrtrlist_update().
648 	 */
649 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
650 	     dr = TAILQ_NEXT(dr, dr_entry)) {
651 		if (selected_dr == NULL &&
652 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
653 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
654 		    ND6_IS_LLINFO_PROBREACH(ln)) {
655 			selected_dr = dr;
656 		}
657 
658 		if (dr->installed && installed_dr == NULL)
659 			installed_dr = dr;
660 		else if (dr->installed && installed_dr) {
661 			/* this should not happen.  warn for diagnosis. */
662 			log(LOG_ERR, "defrouter_select: more than one router"
663 			    " is installed\n");
664 		}
665 	}
666 	/*
667 	 * If none of the default routers was found to be reachable,
668 	 * round-robin the list regardless of preference.
669 	 * Otherwise, if we have an installed router, check if the selected
670 	 * (reachable) router should really be preferred to the installed one.
671 	 * We only prefer the new router when the old one is not reachable
672 	 * or when the new one has a really higher preference value.
673 	 */
674 	if (selected_dr == NULL) {
675 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
676 			selected_dr = TAILQ_FIRST(&V_nd_defrouter);
677 		else
678 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
679 	} else if (installed_dr &&
680 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
681 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
682 	    ND6_IS_LLINFO_PROBREACH(ln) &&
683 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
684 		selected_dr = installed_dr;
685 	}
686 
687 	/*
688 	 * If the selected router is different than the installed one,
689 	 * remove the installed router and install the selected one.
690 	 * Note that the selected router is never NULL here.
691 	 */
692 	if (installed_dr != selected_dr) {
693 		if (installed_dr)
694 			defrouter_delreq(installed_dr);
695 		defrouter_addreq(selected_dr);
696 	}
697 
698 	splx(s);
699 	return;
700 }
701 
702 /*
703  * for default router selection
704  * regards router-preference field as a 2-bit signed integer
705  */
706 static int
707 rtpref(struct nd_defrouter *dr)
708 {
709 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
710 	case ND_RA_FLAG_RTPREF_HIGH:
711 		return (RTPREF_HIGH);
712 	case ND_RA_FLAG_RTPREF_MEDIUM:
713 	case ND_RA_FLAG_RTPREF_RSV:
714 		return (RTPREF_MEDIUM);
715 	case ND_RA_FLAG_RTPREF_LOW:
716 		return (RTPREF_LOW);
717 	default:
718 		/*
719 		 * This case should never happen.  If it did, it would mean a
720 		 * serious bug of kernel internal.  We thus always bark here.
721 		 * Or, can we even panic?
722 		 */
723 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
724 		return (RTPREF_INVALID);
725 	}
726 	/* NOTREACHED */
727 }
728 
729 static struct nd_defrouter *
730 defrtrlist_update(struct nd_defrouter *new)
731 {
732 	INIT_VNET_INET6(curvnet);
733 	struct nd_defrouter *dr, *n;
734 	int s = splnet();
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(&V_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 	bzero(n, 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(&V_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(&V_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 	for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
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 = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
834 	if (new == NULL)
835 		return;
836 	bzero(new, sizeof(*new));
837 	new->router = dr;
838 
839 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
840 
841 	pfxlist_onlink_check();
842 }
843 
844 static void
845 pfxrtr_del(struct nd_pfxrouter *pfr)
846 {
847 	LIST_REMOVE(pfr, pfr_entry);
848 	free(pfr, M_IP6NDP);
849 }
850 
851 struct nd_prefix *
852 nd6_prefix_lookup(struct nd_prefixctl *key)
853 {
854 	INIT_VNET_INET6(curvnet);
855 	struct nd_prefix *search;
856 
857 	for (search = V_nd_prefix.lh_first;
858 	    search; search = search->ndpr_next) {
859 		if (key->ndpr_ifp == search->ndpr_ifp &&
860 		    key->ndpr_plen == search->ndpr_plen &&
861 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
862 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
863 			break;
864 		}
865 	}
866 
867 	return (search);
868 }
869 
870 int
871 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
872     struct nd_prefix **newp)
873 {
874 	INIT_VNET_INET6(curvnet);
875 	struct nd_prefix *new = NULL;
876 	int error = 0;
877 	int i, s;
878 	char ip6buf[INET6_ADDRSTRLEN];
879 
880 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
881 	if (new == NULL)
882 		return(ENOMEM);
883 	bzero(new, sizeof(*new));
884 	new->ndpr_ifp = pr->ndpr_ifp;
885 	new->ndpr_prefix = pr->ndpr_prefix;
886 	new->ndpr_plen = pr->ndpr_plen;
887 	new->ndpr_vltime = pr->ndpr_vltime;
888 	new->ndpr_pltime = pr->ndpr_pltime;
889 	new->ndpr_flags = pr->ndpr_flags;
890 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
891 		free(new, M_IP6NDP);
892 		return(error);
893 	}
894 	new->ndpr_lastupdate = time_second;
895 	if (newp != NULL)
896 		*newp = new;
897 
898 	/* initialization */
899 	LIST_INIT(&new->ndpr_advrtrs);
900 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
901 	/* make prefix in the canonical form */
902 	for (i = 0; i < 4; i++)
903 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
904 		    new->ndpr_mask.s6_addr32[i];
905 
906 	s = splnet();
907 	/* link ndpr_entry to nd_prefix list */
908 	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
909 	splx(s);
910 
911 	/* ND_OPT_PI_FLAG_ONLINK processing */
912 	if (new->ndpr_raf_onlink) {
913 		int e;
914 
915 		if ((e = nd6_prefix_onlink(new)) != 0) {
916 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
917 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
918 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
919 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
920 			/* proceed anyway. XXX: is it correct? */
921 		}
922 	}
923 
924 	if (dr)
925 		pfxrtr_add(new, dr);
926 
927 	return 0;
928 }
929 
930 void
931 prelist_remove(struct nd_prefix *pr)
932 {
933 	INIT_VNET_INET6(curvnet);
934 	struct nd_pfxrouter *pfr, *next;
935 	int e, s;
936 	char ip6buf[INET6_ADDRSTRLEN];
937 
938 	/* make sure to invalidate the prefix until it is really freed. */
939 	pr->ndpr_vltime = 0;
940 	pr->ndpr_pltime = 0;
941 
942 	/*
943 	 * Though these flags are now meaningless, we'd rather keep the value
944 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
945 	 * when executing "ndp -p".
946 	 */
947 
948 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
949 	    (e = nd6_prefix_offlink(pr)) != 0) {
950 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
951 		    "on %s, errno=%d\n",
952 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
953 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
954 		/* what should we do? */
955 	}
956 
957 	if (pr->ndpr_refcnt > 0)
958 		return;		/* notice here? */
959 
960 	s = splnet();
961 
962 	/* unlink ndpr_entry from nd_prefix list */
963 	LIST_REMOVE(pr, ndpr_entry);
964 
965 	/* free list of routers that adversed the prefix */
966 	for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
967 		next = pfr->pfr_next;
968 
969 		free(pfr, M_IP6NDP);
970 	}
971 	splx(s);
972 
973 	free(pr, M_IP6NDP);
974 
975 	pfxlist_onlink_check();
976 }
977 
978 /*
979  * dr - may be NULL
980  */
981 
982 static int
983 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
984     struct mbuf *m, int mcast)
985 {
986 	INIT_VNET_INET6(curvnet);
987 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
988 	struct ifaddr *ifa;
989 	struct ifnet *ifp = new->ndpr_ifp;
990 	struct nd_prefix *pr;
991 	int s = splnet();
992 	int error = 0;
993 	int newprefix = 0;
994 	int auth;
995 	struct in6_addrlifetime lt6_tmp;
996 	char ip6buf[INET6_ADDRSTRLEN];
997 
998 	auth = 0;
999 	if (m) {
1000 		/*
1001 		 * Authenticity for NA consists authentication for
1002 		 * both IP header and IP datagrams, doesn't it ?
1003 		 */
1004 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1005 		auth = ((m->m_flags & M_AUTHIPHDR) &&
1006 		    (m->m_flags & M_AUTHIPDGM));
1007 #endif
1008 	}
1009 
1010 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1011 		/*
1012 		 * nd6_prefix_lookup() ensures that pr and new have the same
1013 		 * prefix on a same interface.
1014 		 */
1015 
1016 		/*
1017 		 * Update prefix information.  Note that the on-link (L) bit
1018 		 * and the autonomous (A) bit should NOT be changed from 1
1019 		 * to 0.
1020 		 */
1021 		if (new->ndpr_raf_onlink == 1)
1022 			pr->ndpr_raf_onlink = 1;
1023 		if (new->ndpr_raf_auto == 1)
1024 			pr->ndpr_raf_auto = 1;
1025 		if (new->ndpr_raf_onlink) {
1026 			pr->ndpr_vltime = new->ndpr_vltime;
1027 			pr->ndpr_pltime = new->ndpr_pltime;
1028 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1029 			pr->ndpr_lastupdate = time_second;
1030 		}
1031 
1032 		if (new->ndpr_raf_onlink &&
1033 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1034 			int e;
1035 
1036 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1037 				nd6log((LOG_ERR,
1038 				    "prelist_update: failed to make "
1039 				    "the prefix %s/%d on-link on %s "
1040 				    "(errno=%d)\n",
1041 				    ip6_sprintf(ip6buf,
1042 					    &pr->ndpr_prefix.sin6_addr),
1043 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1044 				/* proceed anyway. XXX: is it correct? */
1045 			}
1046 		}
1047 
1048 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1049 			pfxrtr_add(pr, dr);
1050 	} else {
1051 		struct nd_prefix *newpr = NULL;
1052 
1053 		newprefix = 1;
1054 
1055 		if (new->ndpr_vltime == 0)
1056 			goto end;
1057 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1058 			goto end;
1059 
1060 		error = nd6_prelist_add(new, dr, &newpr);
1061 		if (error != 0 || newpr == NULL) {
1062 			nd6log((LOG_NOTICE, "prelist_update: "
1063 			    "nd6_prelist_add failed for %s/%d on %s "
1064 			    "errno=%d, returnpr=%p\n",
1065 			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1066 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1067 			    error, newpr));
1068 			goto end; /* we should just give up in this case. */
1069 		}
1070 
1071 		/*
1072 		 * XXX: from the ND point of view, we can ignore a prefix
1073 		 * with the on-link bit being zero.  However, we need a
1074 		 * prefix structure for references from autoconfigured
1075 		 * addresses.  Thus, we explicitly make sure that the prefix
1076 		 * itself expires now.
1077 		 */
1078 		if (newpr->ndpr_raf_onlink == 0) {
1079 			newpr->ndpr_vltime = 0;
1080 			newpr->ndpr_pltime = 0;
1081 			in6_init_prefix_ltimes(newpr);
1082 		}
1083 
1084 		pr = newpr;
1085 	}
1086 
1087 	/*
1088 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1089 	 * Note that pr must be non NULL at this point.
1090 	 */
1091 
1092 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1093 	if (!new->ndpr_raf_auto)
1094 		goto end;
1095 
1096 	/*
1097 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1098 	 * nd6_ra_input.
1099 	 */
1100 
1101 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1102 	if (new->ndpr_pltime > new->ndpr_vltime) {
1103 		error = EINVAL;	/* XXX: won't be used */
1104 		goto end;
1105 	}
1106 
1107 	/*
1108 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1109 	 * an address configured by stateless autoconfiguration already in the
1110 	 * list of addresses associated with the interface, and the Valid
1111 	 * Lifetime is not 0, form an address.  We first check if we have
1112 	 * a matching prefix.
1113 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1114 	 * consider autoconfigured addresses while RFC2462 simply said
1115 	 * "address".
1116 	 */
1117 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1118 		struct in6_ifaddr *ifa6;
1119 		u_int32_t remaininglifetime;
1120 
1121 		if (ifa->ifa_addr->sa_family != AF_INET6)
1122 			continue;
1123 
1124 		ifa6 = (struct in6_ifaddr *)ifa;
1125 
1126 		/*
1127 		 * We only consider autoconfigured addresses as per rfc2462bis.
1128 		 */
1129 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1130 			continue;
1131 
1132 		/*
1133 		 * Spec is not clear here, but I believe we should concentrate
1134 		 * on unicast (i.e. not anycast) addresses.
1135 		 * XXX: other ia6_flags? detached or duplicated?
1136 		 */
1137 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1138 			continue;
1139 
1140 		/*
1141 		 * Ignore the address if it is not associated with a prefix
1142 		 * or is associated with a prefix that is different from this
1143 		 * one.  (pr is never NULL here)
1144 		 */
1145 		if (ifa6->ia6_ndpr != pr)
1146 			continue;
1147 
1148 		if (ia6_match == NULL) /* remember the first one */
1149 			ia6_match = ifa6;
1150 
1151 		/*
1152 		 * An already autoconfigured address matched.  Now that we
1153 		 * are sure there is at least one matched address, we can
1154 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1155 		 * "two hours" rule and the privacy extension.
1156 		 * We apply some clarifications in rfc2462bis:
1157 		 * - use remaininglifetime instead of storedlifetime as a
1158 		 *   variable name
1159 		 * - remove the dead code in the "two-hour" rule
1160 		 */
1161 #define TWOHOUR		(120*60)
1162 		lt6_tmp = ifa6->ia6_lifetime;
1163 
1164 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1165 			remaininglifetime = ND6_INFINITE_LIFETIME;
1166 		else if (time_second - ifa6->ia6_updatetime >
1167 			 lt6_tmp.ia6t_vltime) {
1168 			/*
1169 			 * The case of "invalid" address.  We should usually
1170 			 * not see this case.
1171 			 */
1172 			remaininglifetime = 0;
1173 		} else
1174 			remaininglifetime = lt6_tmp.ia6t_vltime -
1175 			    (time_second - ifa6->ia6_updatetime);
1176 
1177 		/* when not updating, keep the current stored lifetime. */
1178 		lt6_tmp.ia6t_vltime = remaininglifetime;
1179 
1180 		if (TWOHOUR < new->ndpr_vltime ||
1181 		    remaininglifetime < new->ndpr_vltime) {
1182 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1183 		} else if (remaininglifetime <= TWOHOUR) {
1184 			if (auth) {
1185 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1186 			}
1187 		} else {
1188 			/*
1189 			 * new->ndpr_vltime <= TWOHOUR &&
1190 			 * TWOHOUR < remaininglifetime
1191 			 */
1192 			lt6_tmp.ia6t_vltime = TWOHOUR;
1193 		}
1194 
1195 		/* The 2 hour rule is not imposed for preferred lifetime. */
1196 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1197 
1198 		in6_init_address_ltimes(pr, &lt6_tmp);
1199 
1200 		/*
1201 		 * We need to treat lifetimes for temporary addresses
1202 		 * differently, according to
1203 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1204 		 * we only update the lifetimes when they are in the maximum
1205 		 * intervals.
1206 		 */
1207 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1208 			u_int32_t maxvltime, maxpltime;
1209 
1210 			if (V_ip6_temp_valid_lifetime >
1211 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1212 			    V_ip6_desync_factor)) {
1213 				maxvltime = V_ip6_temp_valid_lifetime -
1214 				    (time_second - ifa6->ia6_createtime) -
1215 				    V_ip6_desync_factor;
1216 			} else
1217 				maxvltime = 0;
1218 			if (V_ip6_temp_preferred_lifetime >
1219 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1220 			    V_ip6_desync_factor)) {
1221 				maxpltime = V_ip6_temp_preferred_lifetime -
1222 				    (time_second - ifa6->ia6_createtime) -
1223 				    V_ip6_desync_factor;
1224 			} else
1225 				maxpltime = 0;
1226 
1227 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1228 			    lt6_tmp.ia6t_vltime > maxvltime) {
1229 				lt6_tmp.ia6t_vltime = maxvltime;
1230 			}
1231 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1232 			    lt6_tmp.ia6t_pltime > maxpltime) {
1233 				lt6_tmp.ia6t_pltime = maxpltime;
1234 			}
1235 		}
1236 		ifa6->ia6_lifetime = lt6_tmp;
1237 		ifa6->ia6_updatetime = time_second;
1238 	}
1239 	if (ia6_match == NULL && new->ndpr_vltime) {
1240 		int ifidlen;
1241 
1242 		/*
1243 		 * 5.5.3 (d) (continued)
1244 		 * No address matched and the valid lifetime is non-zero.
1245 		 * Create a new address.
1246 		 */
1247 
1248 		/*
1249 		 * Prefix Length check:
1250 		 * If the sum of the prefix length and interface identifier
1251 		 * length does not equal 128 bits, the Prefix Information
1252 		 * option MUST be ignored.  The length of the interface
1253 		 * identifier is defined in a separate link-type specific
1254 		 * document.
1255 		 */
1256 		ifidlen = in6_if2idlen(ifp);
1257 		if (ifidlen < 0) {
1258 			/* this should not happen, so we always log it. */
1259 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1260 			    if_name(ifp));
1261 			goto end;
1262 		}
1263 		if (ifidlen + pr->ndpr_plen != 128) {
1264 			nd6log((LOG_INFO,
1265 			    "prelist_update: invalid prefixlen "
1266 			    "%d for %s, ignored\n",
1267 			    pr->ndpr_plen, if_name(ifp)));
1268 			goto end;
1269 		}
1270 
1271 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1272 			/*
1273 			 * note that we should use pr (not new) for reference.
1274 			 */
1275 			pr->ndpr_refcnt++;
1276 			ia6->ia6_ndpr = pr;
1277 
1278 			/*
1279 			 * RFC 3041 3.3 (2).
1280 			 * When a new public address is created as described
1281 			 * in RFC2462, also create a new temporary address.
1282 			 *
1283 			 * RFC 3041 3.5.
1284 			 * When an interface connects to a new link, a new
1285 			 * randomized interface identifier should be generated
1286 			 * immediately together with a new set of temporary
1287 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1288 			 * in6_tmpifadd().
1289 			 */
1290 			if (V_ip6_use_tempaddr) {
1291 				int e;
1292 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1293 					nd6log((LOG_NOTICE, "prelist_update: "
1294 					    "failed to create a temporary "
1295 					    "address, errno=%d\n",
1296 					    e));
1297 				}
1298 			}
1299 
1300 			/*
1301 			 * A newly added address might affect the status
1302 			 * of other addresses, so we check and update it.
1303 			 * XXX: what if address duplication happens?
1304 			 */
1305 			pfxlist_onlink_check();
1306 		} else {
1307 			/* just set an error. do not bark here. */
1308 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1309 		}
1310 	}
1311 
1312  end:
1313 	splx(s);
1314 	return error;
1315 }
1316 
1317 /*
1318  * A supplement function used in the on-link detection below;
1319  * detect if a given prefix has a (probably) reachable advertising router.
1320  * XXX: lengthy function name...
1321  */
1322 static struct nd_pfxrouter *
1323 find_pfxlist_reachable_router(struct nd_prefix *pr)
1324 {
1325 	struct nd_pfxrouter *pfxrtr;
1326 	struct rtentry *rt;
1327 	struct llinfo_nd6 *ln;
1328 
1329 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1330 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1331 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1332 		    pfxrtr->router->ifp)) &&
1333 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1334 		    ND6_IS_LLINFO_PROBREACH(ln))
1335 			break;	/* found */
1336 	}
1337 
1338 	return (pfxrtr);
1339 }
1340 
1341 /*
1342  * Check if each prefix in the prefix list has at least one available router
1343  * that advertised the prefix (a router is "available" if its neighbor cache
1344  * entry is reachable or probably reachable).
1345  * If the check fails, the prefix may be off-link, because, for example,
1346  * we have moved from the network but the lifetime of the prefix has not
1347  * expired yet.  So we should not use the prefix if there is another prefix
1348  * that has an available router.
1349  * But, if there is no prefix that has an available router, we still regards
1350  * all the prefixes as on-link.  This is because we can't tell if all the
1351  * routers are simply dead or if we really moved from the network and there
1352  * is no router around us.
1353  */
1354 void
1355 pfxlist_onlink_check()
1356 {
1357 	INIT_VNET_INET6(curvnet);
1358 	struct nd_prefix *pr;
1359 	struct in6_ifaddr *ifa;
1360 	struct nd_defrouter *dr;
1361 	struct nd_pfxrouter *pfxrtr = NULL;
1362 
1363 	/*
1364 	 * Check if there is a prefix that has a reachable advertising
1365 	 * router.
1366 	 */
1367 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1368 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1369 			break;
1370 	}
1371 
1372 	/*
1373 	 * If we have no such prefix, check whether we still have a router
1374 	 * that does not advertise any prefixes.
1375 	 */
1376 	if (pr == NULL) {
1377 		for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
1378 		    dr = TAILQ_NEXT(dr, dr_entry)) {
1379 			struct nd_prefix *pr0;
1380 
1381 			for (pr0 = V_nd_prefix.lh_first; pr0;
1382 			    pr0 = pr0->ndpr_next) {
1383 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1384 					break;
1385 			}
1386 			if (pfxrtr != NULL)
1387 				break;
1388 		}
1389 	}
1390 	if (pr != NULL || (TAILQ_FIRST(&V_nd_defrouter) && pfxrtr == NULL)) {
1391 		/*
1392 		 * There is at least one prefix that has a reachable router,
1393 		 * or at least a router which probably does not advertise
1394 		 * any prefixes.  The latter would be the case when we move
1395 		 * to a new link where we have a router that does not provide
1396 		 * prefixes and we configure an address by hand.
1397 		 * Detach prefixes which have no reachable advertising
1398 		 * router, and attach other prefixes.
1399 		 */
1400 		for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1401 			/* XXX: a link-local prefix should never be detached */
1402 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1403 				continue;
1404 
1405 			/*
1406 			 * we aren't interested in prefixes without the L bit
1407 			 * set.
1408 			 */
1409 			if (pr->ndpr_raf_onlink == 0)
1410 				continue;
1411 
1412 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1413 			    find_pfxlist_reachable_router(pr) == NULL)
1414 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1415 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1416 			    find_pfxlist_reachable_router(pr) != 0)
1417 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1418 		}
1419 	} else {
1420 		/* there is no prefix that has a reachable router */
1421 		for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1422 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1423 				continue;
1424 
1425 			if (pr->ndpr_raf_onlink == 0)
1426 				continue;
1427 
1428 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1429 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1430 		}
1431 	}
1432 
1433 	/*
1434 	 * Remove each interface route associated with a (just) detached
1435 	 * prefix, and reinstall the interface route for a (just) attached
1436 	 * prefix.  Note that all attempt of reinstallation does not
1437 	 * necessarily success, when a same prefix is shared among multiple
1438 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1439 	 * so we don't have to care about them.
1440 	 */
1441 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1442 		int e;
1443 		char ip6buf[INET6_ADDRSTRLEN];
1444 
1445 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1446 			continue;
1447 
1448 		if (pr->ndpr_raf_onlink == 0)
1449 			continue;
1450 
1451 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1452 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1453 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1454 				nd6log((LOG_ERR,
1455 				    "pfxlist_onlink_check: failed to "
1456 				    "make %s/%d offlink, errno=%d\n",
1457 				    ip6_sprintf(ip6buf,
1458 					    &pr->ndpr_prefix.sin6_addr),
1459 					    pr->ndpr_plen, e));
1460 			}
1461 		}
1462 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1463 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1464 		    pr->ndpr_raf_onlink) {
1465 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1466 				nd6log((LOG_ERR,
1467 				    "pfxlist_onlink_check: failed to "
1468 				    "make %s/%d onlink, errno=%d\n",
1469 				    ip6_sprintf(ip6buf,
1470 					    &pr->ndpr_prefix.sin6_addr),
1471 					    pr->ndpr_plen, e));
1472 			}
1473 		}
1474 	}
1475 
1476 	/*
1477 	 * Changes on the prefix status might affect address status as well.
1478 	 * Make sure that all addresses derived from an attached prefix are
1479 	 * attached, and that all addresses derived from a detached prefix are
1480 	 * detached.  Note, however, that a manually configured address should
1481 	 * always be attached.
1482 	 * The precise detection logic is same as the one for prefixes.
1483 	 */
1484 	for (ifa = V_in6_ifaddr; ifa; ifa = ifa->ia_next) {
1485 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1486 			continue;
1487 
1488 		if (ifa->ia6_ndpr == NULL) {
1489 			/*
1490 			 * This can happen when we first configure the address
1491 			 * (i.e. the address exists, but the prefix does not).
1492 			 * XXX: complicated relationships...
1493 			 */
1494 			continue;
1495 		}
1496 
1497 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1498 			break;
1499 	}
1500 	if (ifa) {
1501 		for (ifa = V_in6_ifaddr; ifa; ifa = ifa->ia_next) {
1502 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1503 				continue;
1504 
1505 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1506 				continue;
1507 
1508 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1509 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1510 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1511 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1512 					nd6_dad_start((struct ifaddr *)ifa, 0);
1513 				}
1514 			} else {
1515 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1516 			}
1517 		}
1518 	}
1519 	else {
1520 		for (ifa = V_in6_ifaddr; ifa; ifa = ifa->ia_next) {
1521 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1522 				continue;
1523 
1524 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1525 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1526 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1527 				/* Do we need a delay in this case? */
1528 				nd6_dad_start((struct ifaddr *)ifa, 0);
1529 			}
1530 		}
1531 	}
1532 }
1533 
1534 int
1535 nd6_prefix_onlink(struct nd_prefix *pr)
1536 {
1537 	INIT_VNET_INET6(curvnet);
1538 	struct ifaddr *ifa;
1539 	struct ifnet *ifp = pr->ndpr_ifp;
1540 	struct sockaddr_in6 mask6;
1541 	struct nd_prefix *opr;
1542 	u_long rtflags;
1543 	int error = 0;
1544 	struct rtentry *rt = NULL;
1545 	char ip6buf[INET6_ADDRSTRLEN];
1546 
1547 	/* sanity check */
1548 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1549 		nd6log((LOG_ERR,
1550 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1551 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1552 		    pr->ndpr_plen));
1553 		return (EEXIST);
1554 	}
1555 
1556 	/*
1557 	 * Add the interface route associated with the prefix.  Before
1558 	 * installing the route, check if there's the same prefix on another
1559 	 * interface, and the prefix has already installed the interface route.
1560 	 * Although such a configuration is expected to be rare, we explicitly
1561 	 * allow it.
1562 	 */
1563 	for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1564 		if (opr == pr)
1565 			continue;
1566 
1567 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1568 			continue;
1569 
1570 		if (opr->ndpr_plen == pr->ndpr_plen &&
1571 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1572 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1573 			return (0);
1574 	}
1575 
1576 	/*
1577 	 * We prefer link-local addresses as the associated interface address.
1578 	 */
1579 	/* search for a link-local addr */
1580 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1581 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1582 	if (ifa == NULL) {
1583 		/* XXX: freebsd does not have ifa_ifwithaf */
1584 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1585 			if (ifa->ifa_addr->sa_family == AF_INET6)
1586 				break;
1587 		}
1588 		/* should we care about ia6_flags? */
1589 	}
1590 	if (ifa == NULL) {
1591 		/*
1592 		 * This can still happen, when, for example, we receive an RA
1593 		 * containing a prefix with the L bit set and the A bit clear,
1594 		 * after removing all IPv6 addresses on the receiving
1595 		 * interface.  This should, of course, be rare though.
1596 		 */
1597 		nd6log((LOG_NOTICE,
1598 		    "nd6_prefix_onlink: failed to find any ifaddr"
1599 		    " to add route for a prefix(%s/%d) on %s\n",
1600 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1601 		    pr->ndpr_plen, if_name(ifp)));
1602 		return (0);
1603 	}
1604 
1605 	/*
1606 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1607 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1608 	 */
1609 	bzero(&mask6, sizeof(mask6));
1610 	mask6.sin6_len = sizeof(mask6);
1611 	mask6.sin6_addr = pr->ndpr_mask;
1612 	rtflags = ifa->ifa_flags | RTF_CLONING | RTF_UP;
1613 	if (nd6_need_cache(ifp)) {
1614 		/* explicitly set in case ifa_flags does not set the flag. */
1615 		rtflags |= RTF_CLONING;
1616 	} else {
1617 		/*
1618 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1619 		 */
1620 		rtflags &= ~RTF_CLONING;
1621 	}
1622 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1623 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1624 	if (error == 0) {
1625 		if (rt != NULL) /* this should be non NULL, though */
1626 			nd6_rtmsg(RTM_ADD, rt);
1627 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1628 	} else {
1629 		char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
1630 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1631 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1632 		    "errno = %d\n",
1633 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1634 		    pr->ndpr_plen, if_name(ifp),
1635 		    ip6_sprintf(ip6bufg, &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1636 		    ip6_sprintf(ip6bufm, &mask6.sin6_addr), rtflags, error));
1637 	}
1638 
1639 	if (rt != NULL) {
1640 		RT_LOCK(rt);
1641 		RT_REMREF(rt);
1642 		RT_UNLOCK(rt);
1643 	}
1644 
1645 	return (error);
1646 }
1647 
1648 int
1649 nd6_prefix_offlink(struct nd_prefix *pr)
1650 {
1651 	INIT_VNET_INET6(curvnet);
1652 	int error = 0;
1653 	struct ifnet *ifp = pr->ndpr_ifp;
1654 	struct nd_prefix *opr;
1655 	struct sockaddr_in6 sa6, mask6;
1656 	struct rtentry *rt = NULL;
1657 	char ip6buf[INET6_ADDRSTRLEN];
1658 
1659 	/* sanity check */
1660 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1661 		nd6log((LOG_ERR,
1662 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1663 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1664 		    pr->ndpr_plen));
1665 		return (EEXIST);
1666 	}
1667 
1668 	bzero(&sa6, sizeof(sa6));
1669 	sa6.sin6_family = AF_INET6;
1670 	sa6.sin6_len = sizeof(sa6);
1671 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1672 	    sizeof(struct in6_addr));
1673 	bzero(&mask6, sizeof(mask6));
1674 	mask6.sin6_family = AF_INET6;
1675 	mask6.sin6_len = sizeof(sa6);
1676 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1677 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1678 	    (struct sockaddr *)&mask6, 0, &rt);
1679 	if (error == 0) {
1680 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1681 
1682 		/* report the route deletion to the routing socket. */
1683 		if (rt != NULL)
1684 			nd6_rtmsg(RTM_DELETE, rt);
1685 
1686 		/*
1687 		 * There might be the same prefix on another interface,
1688 		 * the prefix which could not be on-link just because we have
1689 		 * the interface route (see comments in nd6_prefix_onlink).
1690 		 * If there's one, try to make the prefix on-link on the
1691 		 * interface.
1692 		 */
1693 		for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1694 			if (opr == pr)
1695 				continue;
1696 
1697 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1698 				continue;
1699 
1700 			/*
1701 			 * KAME specific: detached prefixes should not be
1702 			 * on-link.
1703 			 */
1704 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1705 				continue;
1706 
1707 			if (opr->ndpr_plen == pr->ndpr_plen &&
1708 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1709 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1710 				int e;
1711 
1712 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1713 					nd6log((LOG_ERR,
1714 					    "nd6_prefix_offlink: failed to "
1715 					    "recover a prefix %s/%d from %s "
1716 					    "to %s (errno = %d)\n",
1717 					    ip6_sprintf(ip6buf,
1718 						&opr->ndpr_prefix.sin6_addr),
1719 					    opr->ndpr_plen, if_name(ifp),
1720 					    if_name(opr->ndpr_ifp), e));
1721 				}
1722 			}
1723 		}
1724 	} else {
1725 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1726 		nd6log((LOG_ERR,
1727 		    "nd6_prefix_offlink: failed to delete route: "
1728 		    "%s/%d on %s (errno = %d)\n",
1729 		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1730 		    if_name(ifp), error));
1731 	}
1732 
1733 	if (rt != NULL) {
1734 		RTFREE(rt);
1735 	}
1736 
1737 	return (error);
1738 }
1739 
1740 static struct in6_ifaddr *
1741 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1742 {
1743 	INIT_VNET_INET6(curvnet);
1744 	struct ifnet *ifp = pr->ndpr_ifp;
1745 	struct ifaddr *ifa;
1746 	struct in6_aliasreq ifra;
1747 	struct in6_ifaddr *ia, *ib;
1748 	int error, plen0;
1749 	struct in6_addr mask;
1750 	int prefixlen = pr->ndpr_plen;
1751 	int updateflags;
1752 	char ip6buf[INET6_ADDRSTRLEN];
1753 
1754 	in6_prefixlen2mask(&mask, prefixlen);
1755 
1756 	/*
1757 	 * find a link-local address (will be interface ID).
1758 	 * Is it really mandatory? Theoretically, a global or a site-local
1759 	 * address can be configured without a link-local address, if we
1760 	 * have a unique interface identifier...
1761 	 *
1762 	 * it is not mandatory to have a link-local address, we can generate
1763 	 * interface identifier on the fly.  we do this because:
1764 	 * (1) it should be the easiest way to find interface identifier.
1765 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1766 	 * for multiple addresses on a single interface, and possible shortcut
1767 	 * of DAD.  we omitted DAD for this reason in the past.
1768 	 * (3) a user can prevent autoconfiguration of global address
1769 	 * by removing link-local address by hand (this is partly because we
1770 	 * don't have other way to control the use of IPv6 on an interface.
1771 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1772 	 * (4) it is easier to manage when an interface has addresses
1773 	 * with the same interface identifier, than to have multiple addresses
1774 	 * with different interface identifiers.
1775 	 */
1776 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1777 	if (ifa)
1778 		ib = (struct in6_ifaddr *)ifa;
1779 	else
1780 		return NULL;
1781 
1782 	/* prefixlen + ifidlen must be equal to 128 */
1783 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1784 	if (prefixlen != plen0) {
1785 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1786 		    "(prefix=%d ifid=%d)\n",
1787 		    if_name(ifp), prefixlen, 128 - plen0));
1788 		return NULL;
1789 	}
1790 
1791 	/* make ifaddr */
1792 
1793 	bzero(&ifra, sizeof(ifra));
1794 	/*
1795 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1796 	 * for safety.
1797 	 */
1798 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1799 	ifra.ifra_addr.sin6_family = AF_INET6;
1800 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1801 	/* prefix */
1802 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1803 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1804 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1805 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1806 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1807 
1808 	/* interface ID */
1809 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1810 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1811 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1812 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1813 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1814 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1815 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1816 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1817 
1818 	/* new prefix mask. */
1819 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1820 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1821 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1822 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1823 
1824 	/* lifetimes. */
1825 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1826 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1827 
1828 	/* XXX: scope zone ID? */
1829 
1830 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1831 
1832 	/*
1833 	 * Make sure that we do not have this address already.  This should
1834 	 * usually not happen, but we can still see this case, e.g., if we
1835 	 * have manually configured the exact address to be configured.
1836 	 */
1837 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1838 		/* this should be rare enough to make an explicit log */
1839 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1840 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1841 		return (NULL);
1842 	}
1843 
1844 	/*
1845 	 * Allocate ifaddr structure, link into chain, etc.
1846 	 * If we are going to create a new address upon receiving a multicasted
1847 	 * RA, we need to impose a random delay before starting DAD.
1848 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1849 	 */
1850 	updateflags = 0;
1851 	if (mcast)
1852 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1853 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1854 		nd6log((LOG_ERR,
1855 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1856 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1857 		    if_name(ifp), error));
1858 		return (NULL);	/* ifaddr must not have been allocated. */
1859 	}
1860 
1861 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1862 
1863 	return (ia);		/* this is always non-NULL */
1864 }
1865 
1866 /*
1867  * ia0 - corresponding public address
1868  */
1869 int
1870 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1871 {
1872 	INIT_VNET_INET6(curvnet);
1873 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1874 	struct in6_ifaddr *newia, *ia;
1875 	struct in6_aliasreq ifra;
1876 	int i, error;
1877 	int trylimit = 3;	/* XXX: adhoc value */
1878 	int updateflags;
1879 	u_int32_t randid[2];
1880 	time_t vltime0, pltime0;
1881 
1882 	bzero(&ifra, sizeof(ifra));
1883 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1884 	ifra.ifra_addr = ia0->ia_addr;
1885 	/* copy prefix mask */
1886 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1887 	/* clear the old IFID */
1888 	for (i = 0; i < 4; i++) {
1889 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1890 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1891 	}
1892 
1893   again:
1894 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1895 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1896 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1897 		    "random IFID\n"));
1898 		return (EINVAL);
1899 	}
1900 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1901 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1902 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1903 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1904 
1905 	/*
1906 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1907 	 * However, we may still have a chance to see collision, because
1908 	 * there may be a time lag between generation of the ID and generation
1909 	 * of the address.  So, we'll do one more sanity check.
1910 	 */
1911 	for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) {
1912 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1913 		    &ifra.ifra_addr.sin6_addr)) {
1914 			if (trylimit-- == 0) {
1915 				/*
1916 				 * Give up.  Something strange should have
1917 				 * happened.
1918 				 */
1919 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1920 				    "find a unique random IFID\n"));
1921 				return (EEXIST);
1922 			}
1923 			forcegen = 1;
1924 			goto again;
1925 		}
1926 	}
1927 
1928 	/*
1929 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1930          * public address or TEMP_VALID_LIFETIME.
1931 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1932          * of the public address or TEMP_PREFERRED_LIFETIME -
1933          * DESYNC_FACTOR.
1934 	 */
1935 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1936 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1937 		    (ia0->ia6_lifetime.ia6t_vltime -
1938 		    (time_second - ia0->ia6_updatetime));
1939 		if (vltime0 > V_ip6_temp_valid_lifetime)
1940 			vltime0 = V_ip6_temp_valid_lifetime;
1941 	} else
1942 		vltime0 = V_ip6_temp_valid_lifetime;
1943 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1944 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1945 		    (ia0->ia6_lifetime.ia6t_pltime -
1946 		    (time_second - ia0->ia6_updatetime));
1947 		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
1948 			pltime0 = V_ip6_temp_preferred_lifetime -
1949 			    V_ip6_desync_factor;
1950 		}
1951 	} else
1952 		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
1953 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1954 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1955 
1956 	/*
1957 	 * A temporary address is created only if this calculated Preferred
1958 	 * Lifetime is greater than REGEN_ADVANCE time units.
1959 	 */
1960 	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
1961 		return (0);
1962 
1963 	/* XXX: scope zone ID? */
1964 
1965 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1966 
1967 	/* allocate ifaddr structure, link into chain, etc. */
1968 	updateflags = 0;
1969 	if (delay)
1970 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1971 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1972 		return (error);
1973 
1974 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1975 	if (newia == NULL) {	/* XXX: can it happen? */
1976 		nd6log((LOG_ERR,
1977 		    "in6_tmpifadd: ifa update succeeded, but we got "
1978 		    "no ifaddr\n"));
1979 		return (EINVAL); /* XXX */
1980 	}
1981 	newia->ia6_ndpr = ia0->ia6_ndpr;
1982 	newia->ia6_ndpr->ndpr_refcnt++;
1983 
1984 	/*
1985 	 * A newly added address might affect the status of other addresses.
1986 	 * XXX: when the temporary address is generated with a new public
1987 	 * address, the onlink check is redundant.  However, it would be safe
1988 	 * to do the check explicitly everywhere a new address is generated,
1989 	 * and, in fact, we surely need the check when we create a new
1990 	 * temporary address due to deprecation of an old temporary address.
1991 	 */
1992 	pfxlist_onlink_check();
1993 
1994 	return (0);
1995 }
1996 
1997 static int
1998 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1999 {
2000 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2001 		ndpr->ndpr_preferred = 0;
2002 	else
2003 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2004 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2005 		ndpr->ndpr_expire = 0;
2006 	else
2007 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2008 
2009 	return 0;
2010 }
2011 
2012 static void
2013 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2014 {
2015 	/* init ia6t_expire */
2016 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2017 		lt6->ia6t_expire = 0;
2018 	else {
2019 		lt6->ia6t_expire = time_second;
2020 		lt6->ia6t_expire += lt6->ia6t_vltime;
2021 	}
2022 
2023 	/* init ia6t_preferred */
2024 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2025 		lt6->ia6t_preferred = 0;
2026 	else {
2027 		lt6->ia6t_preferred = time_second;
2028 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2029 	}
2030 }
2031 
2032 /*
2033  * Delete all the routing table entries that use the specified gateway.
2034  * XXX: this function causes search through all entries of routing table, so
2035  * it shouldn't be called when acting as a router.
2036  */
2037 void
2038 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2039 {
2040 	INIT_VNET_NET(curvnet);
2041 	struct radix_node_head *rnh = V_rt_tables[0][AF_INET6];
2042 	int s = splnet();
2043 
2044 	/* We'll care only link-local addresses */
2045 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2046 		splx(s);
2047 		return;
2048 	}
2049 
2050 	RADIX_NODE_HEAD_LOCK(rnh);
2051 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2052 	RADIX_NODE_HEAD_UNLOCK(rnh);
2053 	splx(s);
2054 }
2055 
2056 static int
2057 rt6_deleteroute(struct radix_node *rn, void *arg)
2058 {
2059 #define SIN6(s)	((struct sockaddr_in6 *)s)
2060 	struct rtentry *rt = (struct rtentry *)rn;
2061 	struct in6_addr *gate = (struct in6_addr *)arg;
2062 
2063 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2064 		return (0);
2065 
2066 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2067 		return (0);
2068 	}
2069 
2070 	/*
2071 	 * Do not delete a static route.
2072 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2073 	 * 'cloned' bit instead?
2074 	 */
2075 	if ((rt->rt_flags & RTF_STATIC) != 0)
2076 		return (0);
2077 
2078 	/*
2079 	 * We delete only host route. This means, in particular, we don't
2080 	 * delete default route.
2081 	 */
2082 	if ((rt->rt_flags & RTF_HOST) == 0)
2083 		return (0);
2084 
2085 	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2086 	    rt_mask(rt), rt->rt_flags, 0));
2087 #undef SIN6
2088 }
2089 
2090 int
2091 nd6_setdefaultiface(int ifindex)
2092 {
2093 	INIT_VNET_NET(curvnet);
2094 	INIT_VNET_INET6(curvnet);
2095 	int error = 0;
2096 
2097 	if (ifindex < 0 || V_if_index < ifindex)
2098 		return (EINVAL);
2099 	if (ifindex != 0 && !ifnet_byindex(ifindex))
2100 		return (EINVAL);
2101 
2102 	if (V_nd6_defifindex != ifindex) {
2103 		V_nd6_defifindex = ifindex;
2104 		if (V_nd6_defifindex > 0)
2105 			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2106 		else
2107 			V_nd6_defifp = NULL;
2108 
2109 		/*
2110 		 * Our current implementation assumes one-to-one maping between
2111 		 * interfaces and links, so it would be natural to use the
2112 		 * default interface as the default link.
2113 		 */
2114 		scope6_setdefault(V_nd6_defifp);
2115 	}
2116 
2117 	return (error);
2118 }
2119