xref: /openbsd-src/sys/netinet6/nd6_rtr.c (revision 4c1e55dc91edd6e69ccc60ce855900fbc12cf34f)
1 /*	$OpenBSD: nd6_rtr.c,v 1.59 2012/05/18 10:50:07 mikeb Exp $	*/
2 /*	$KAME: nd6_rtr.c,v 1.97 2001/02/07 11:09:13 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/param.h>
34 #include <sys/systm.h>
35 #include <sys/malloc.h>
36 #include <sys/mbuf.h>
37 #include <sys/socket.h>
38 #include <sys/sockio.h>
39 #include <sys/time.h>
40 #include <sys/kernel.h>
41 #include <sys/errno.h>
42 #include <sys/ioctl.h>
43 #include <sys/syslog.h>
44 #include <sys/queue.h>
45 #include <sys/workq.h>
46 #include <dev/rndvar.h>
47 
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/if_dl.h>
51 #include <net/route.h>
52 #include <net/radix.h>
53 
54 #include <netinet/in.h>
55 #include <netinet6/in6_var.h>
56 #include <netinet/ip6.h>
57 #include <netinet6/ip6_var.h>
58 #include <netinet6/nd6.h>
59 #include <netinet/icmp6.h>
60 
61 #include <dev/rndvar.h>
62 
63 #define SDL(s)	((struct sockaddr_dl *)s)
64 
65 int rtpref(struct nd_defrouter *);
66 struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
67 struct in6_ifaddr *in6_ifadd(struct nd_prefix *);
68 struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *, struct nd_defrouter *);
69 void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
70 void pfxrtr_del(struct nd_pfxrouter *);
71 struct nd_pfxrouter *find_pfxlist_reachable_router(struct nd_prefix *);
72 void defrouter_delreq(struct nd_defrouter *);
73 void nd6_rtmsg(int, struct rtentry *);
74 void purge_detached(struct ifnet *);
75 
76 void in6_init_address_ltimes(struct nd_prefix *, struct in6_addrlifetime *);
77 
78 int rt6_deleteroute(struct radix_node *, void *, u_int);
79 
80 void nd6_addr_add(void *, void *);
81 
82 extern int nd6_recalc_reachtm_interval;
83 
84 static struct ifnet *nd6_defifp;
85 int nd6_defifindex;
86 
87 /*
88  * Receive Router Solicitation Message - just for routers.
89  * Router solicitation/advertisement is mostly managed by userland program
90  * (rtadvd) so here we have no function like nd6_ra_output().
91  *
92  * Based on RFC 2461
93  */
94 void
95 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
96 {
97 	struct ifnet *ifp = m->m_pkthdr.rcvif;
98 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
99 	struct nd_router_solicit *nd_rs;
100 	struct in6_addr saddr6 = ip6->ip6_src;
101 #if 0
102 	struct in6_addr daddr6 = ip6->ip6_dst;
103 #endif
104 	char *lladdr = NULL;
105 	int lladdrlen = 0;
106 #if 0
107 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)NULL;
108 	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)NULL;
109 	struct rtentry *rt = NULL;
110 	int is_newentry;
111 #endif
112 	union nd_opts ndopts;
113 
114 	/* If I'm not a router, ignore it. */
115 	if (ip6_accept_rtadv != 0 || !ip6_forwarding)
116 		goto freeit;
117 
118 	/* Sanity checks */
119 	if (ip6->ip6_hlim != 255) {
120 		nd6log((LOG_ERR,
121 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
122 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
123 		    ip6_sprintf(&ip6->ip6_dst), ifp->if_xname));
124 		goto bad;
125 	}
126 
127 	/*
128 	 * Don't update the neighbor cache, if src = ::.
129 	 * This indicates that the src has no IP address assigned yet.
130 	 */
131 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
132 		goto freeit;
133 
134 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
135 	if (nd_rs == NULL) {
136 		icmp6stat.icp6s_tooshort++;
137 		return;
138 	}
139 
140 	icmp6len -= sizeof(*nd_rs);
141 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
142 	if (nd6_options(&ndopts) < 0) {
143 		nd6log((LOG_INFO,
144 		    "nd6_rs_input: invalid ND option, ignored\n"));
145 		/* nd6_options have incremented stats */
146 		goto freeit;
147 	}
148 
149 	if (ndopts.nd_opts_src_lladdr) {
150 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
151 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
152 	}
153 
154 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
155 		nd6log((LOG_INFO,
156 		    "nd6_rs_input: lladdrlen mismatch for %s "
157 		    "(if %d, RS packet %d)\n",
158 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
159 		goto bad;
160 	}
161 
162 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
163 
164  freeit:
165 	m_freem(m);
166 	return;
167 
168  bad:
169 	icmp6stat.icp6s_badrs++;
170 	m_freem(m);
171 }
172 
173 /*
174  * Receive Router Advertisement Message.
175  *
176  * Based on RFC 2461
177  * TODO: on-link bit on prefix information
178  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
179  */
180 void
181 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
182 {
183 	struct ifnet *ifp = m->m_pkthdr.rcvif;
184 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
185 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
186 	struct nd_router_advert *nd_ra;
187 	struct in6_addr saddr6 = ip6->ip6_src;
188 #if 0
189 	struct in6_addr daddr6 = ip6->ip6_dst;
190 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
191 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
192 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
193 #endif
194 	union nd_opts ndopts;
195 	struct nd_defrouter *dr;
196 
197 	/*
198 	 * We only accept RAs only when
199 	 * the system-wide variable allows the acceptance, and
200 	 * per-interface variable allows RAs on the receiving interface.
201 	 */
202 	if (ip6_accept_rtadv == 0)
203 		goto freeit;
204 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
205 		goto freeit;
206 
207 	if (ip6->ip6_hlim != 255) {
208 		nd6log((LOG_ERR,
209 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
210 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
211 		    ip6_sprintf(&ip6->ip6_dst), ifp->if_xname));
212 		goto bad;
213 	}
214 
215 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
216 		nd6log((LOG_ERR,
217 		    "nd6_ra_input: src %s is not link-local\n",
218 		    ip6_sprintf(&saddr6)));
219 		goto bad;
220 	}
221 
222 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
223 	if (nd_ra == NULL) {
224 		icmp6stat.icp6s_tooshort++;
225 		return;
226 	}
227 
228 	icmp6len -= sizeof(*nd_ra);
229 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
230 	if (nd6_options(&ndopts) < 0) {
231 		nd6log((LOG_INFO,
232 		    "nd6_ra_input: invalid ND option, ignored\n"));
233 		/* nd6_options have incremented stats */
234 		goto freeit;
235 	}
236 
237     {
238 	struct nd_defrouter dr0;
239 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
240 
241 	Bzero(&dr0, sizeof(dr0));
242 	dr0.rtaddr = saddr6;
243 	dr0.flags  = nd_ra->nd_ra_flags_reserved;
244 	dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
245 	dr0.expire = time_second + dr0.rtlifetime;
246 	dr0.ifp = ifp;
247 	/* unspecified or not? (RFC 2461 6.3.4) */
248 	if (advreachable) {
249 		NTOHL(advreachable);
250 		if (advreachable <= MAX_REACHABLE_TIME &&
251 		    ndi->basereachable != advreachable) {
252 			ndi->basereachable = advreachable;
253 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
254 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
255 		}
256 	}
257 	if (nd_ra->nd_ra_retransmit)
258 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
259 	if (nd_ra->nd_ra_curhoplimit)
260 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
261 	dr = defrtrlist_update(&dr0);
262     }
263 
264 	/*
265 	 * prefix
266 	 */
267 	if (ndopts.nd_opts_pi) {
268 		struct nd_opt_hdr *pt;
269 		struct nd_opt_prefix_info *pi = NULL;
270 		struct nd_prefix pr;
271 
272 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
273 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
274 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
275 						(pt->nd_opt_len << 3))) {
276 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
277 				continue;
278 			pi = (struct nd_opt_prefix_info *)pt;
279 
280 			if (pi->nd_opt_pi_len != 4) {
281 				nd6log((LOG_INFO,
282 				    "nd6_ra_input: invalid option "
283 				    "len %d for prefix information option, "
284 				    "ignored\n", pi->nd_opt_pi_len));
285 				continue;
286 			}
287 
288 			if (128 < pi->nd_opt_pi_prefix_len) {
289 				nd6log((LOG_INFO,
290 				    "nd6_ra_input: invalid prefix "
291 				    "len %d for prefix information option, "
292 				    "ignored\n", pi->nd_opt_pi_prefix_len));
293 				continue;
294 			}
295 
296 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
297 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
298 				nd6log((LOG_INFO,
299 				    "nd6_ra_input: invalid prefix "
300 				    "%s, ignored\n",
301 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
302 				continue;
303 			}
304 
305 			/* aggregatable unicast address, rfc2374 */
306 			if ((pi->nd_opt_pi_prefix.s6_addr8[0] & 0xe0) == 0x20
307 			 && pi->nd_opt_pi_prefix_len != 64) {
308 				nd6log((LOG_INFO,
309 				    "nd6_ra_input: invalid prefixlen "
310 				    "%d for rfc2374 prefix %s, ignored\n",
311 				    pi->nd_opt_pi_prefix_len,
312 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
313 				continue;
314 			}
315 
316 			bzero(&pr, sizeof(pr));
317 			pr.ndpr_prefix.sin6_family = AF_INET6;
318 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
319 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
320 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
321 
322 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
323 			     ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
324 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
325 			     ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
326 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
327 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
328 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
329 			pr.ndpr_lastupdate = time_second;
330 
331 			if (in6_init_prefix_ltimes(&pr))
332 				continue; /* prefix lifetime init failed */
333 
334 			(void)prelist_update(&pr, dr, m);
335 		}
336 	}
337 
338 	/*
339 	 * MTU
340 	 */
341 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
342 		u_long mtu;
343 		u_long maxmtu;
344 
345 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
346 
347 		/* lower bound */
348 		if (mtu < IPV6_MMTU) {
349 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
350 			    "mtu=%lu sent from %s, ignoring\n",
351 			    mtu, ip6_sprintf(&ip6->ip6_src)));
352 			goto skip;
353 		}
354 
355 		/* upper bound */
356 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
357 		    ? ndi->maxmtu : ifp->if_mtu;
358 		if (mtu <= maxmtu) {
359 			int change = (ndi->linkmtu != mtu);
360 
361 			ndi->linkmtu = mtu;
362 			if (change) /* in6_maxmtu may change */
363 				in6_setmaxmtu();
364 		} else {
365 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
366 			    "mtu=%lu sent from %s; "
367 			    "exceeds maxmtu %lu, ignoring\n",
368 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
369 		}
370 	}
371 
372  skip:
373 
374 	/*
375 	 * Source link layer address
376 	 */
377     {
378 	char *lladdr = NULL;
379 	int lladdrlen = 0;
380 
381 	if (ndopts.nd_opts_src_lladdr) {
382 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
383 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
384 	}
385 
386 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
387 		nd6log((LOG_INFO,
388 		    "nd6_ra_input: lladdrlen mismatch for %s "
389 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
390 		    ifp->if_addrlen, lladdrlen - 2));
391 		goto bad;
392 	}
393 
394 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
395 
396 	/*
397 	 * Installing a link-layer address might change the state of the
398 	 * router's neighbor cache, which might also affect our on-link
399 	 * detection of adveritsed prefixes.
400 	 */
401 	pfxlist_onlink_check();
402     }
403 
404  freeit:
405 	m_freem(m);
406 	return;
407 
408  bad:
409 	icmp6stat.icp6s_badra++;
410 	m_freem(m);
411 }
412 
413 /*
414  * default router list processing sub routines
415  */
416 
417 /* tell the change to user processes watching the routing socket. */
418 void
419 nd6_rtmsg(int cmd, struct rtentry *rt)
420 {
421 	struct rt_addrinfo info;
422 
423 	bzero((caddr_t)&info, sizeof(info));
424 	info.rti_info[RTAX_DST] = rt_key(rt);
425 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
426 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
427 	if (rt->rt_ifp) {
428 		info.rti_info[RTAX_IFP] =
429 		    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
430 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
431 	}
432 
433 	rt_missmsg(cmd, &info, rt->rt_flags, rt->rt_ifp, 0, 0);
434 }
435 
436 void
437 defrouter_addreq(struct nd_defrouter *new)
438 {
439 	struct rt_addrinfo info;
440 	struct sockaddr_in6 def, mask, gate;
441 	struct rtentry *newrt = NULL;
442 	int s;
443 	int error;
444 
445 	Bzero(&def, sizeof(def));
446 	Bzero(&mask, sizeof(mask));
447 	Bzero(&gate, sizeof(gate)); /* for safety */
448 	Bzero(&info, sizeof(info));
449 
450 	def.sin6_len = mask.sin6_len = gate.sin6_len =
451 	    sizeof(struct sockaddr_in6);
452 	def.sin6_family = mask.sin6_family = gate.sin6_family = AF_INET6;
453 	gate.sin6_addr = new->rtaddr;
454 	gate.sin6_scope_id = 0;	/* XXX */
455 
456 	info.rti_flags = RTF_GATEWAY;
457 	info.rti_info[RTAX_DST] = (struct sockaddr *)&def;
458 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gate;
459 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&mask;
460 
461 	s = splsoftnet();
462 	error = rtrequest1(RTM_ADD, &info, RTP_DEFAULT, &newrt,
463 	    new->ifp->if_rdomain);
464 	if (newrt) {
465 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
466 		newrt->rt_refcnt--;
467 	}
468 	if (error == 0)
469 		new->installed = 1;
470 	splx(s);
471 	return;
472 }
473 
474 struct nd_defrouter *
475 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
476 {
477 	struct nd_defrouter *dr;
478 
479 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry)
480 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
481 			return (dr);
482 
483 	return (NULL);		/* search failed */
484 }
485 
486 void
487 defrtrlist_del(struct nd_defrouter *dr)
488 {
489 	struct nd_defrouter *deldr = NULL;
490 	struct in6_ifextra *ext = dr->ifp->if_afdata[AF_INET6];
491 	struct nd_prefix *pr;
492 
493 	/*
494 	 * Flush all the routing table entries that use the router
495 	 * as a next hop.
496 	 */
497 	if (!ip6_forwarding && ip6_accept_rtadv) /* XXX: better condition? */
498 		rt6_flush(&dr->rtaddr, dr->ifp);
499 
500 	if (dr->installed) {
501 		deldr = dr;
502 		defrouter_delreq(dr);
503 	}
504 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
505 
506 	/*
507 	 * Also delete all the pointers to the router in each prefix lists.
508 	 */
509 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
510 		struct nd_pfxrouter *pfxrtr;
511 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
512 			pfxrtr_del(pfxrtr);
513 	}
514 	pfxlist_onlink_check();
515 
516 	/*
517 	 * If the router is the primary one, choose a new one.
518 	 * Note that defrouter_select() will remove the current gateway
519 	 * from the routing table.
520 	 */
521 	if (deldr)
522 		defrouter_select();
523 
524 	ext->ndefrouters--;
525 	if (ext->ndefrouters < 0) {
526 		log(LOG_WARNING, "defrtrlist_del: negative count on %s\n",
527 		    dr->ifp->if_xname);
528 	}
529 
530 	free(dr, M_IP6NDP);
531 }
532 
533 /*
534  * Remove the default route for a given router.
535  * This is just a subroutine function for defrouter_select(), and should
536  * not be called from anywhere else.
537  */
538 void
539 defrouter_delreq(struct nd_defrouter *dr)
540 {
541 	struct rt_addrinfo info;
542 	struct sockaddr_in6 def, mask, gw;
543 	struct rtentry *oldrt = NULL;
544 
545 #ifdef DIAGNOSTIC
546 	if (!dr)
547 		panic("dr == NULL in defrouter_delreq");
548 #endif
549 
550 	Bzero(&info, sizeof(info));
551 	Bzero(&def, sizeof(def));
552 	Bzero(&mask, sizeof(mask));
553 	Bzero(&gw, sizeof(gw));	/* for safety */
554 
555 	def.sin6_len = mask.sin6_len = gw.sin6_len =
556 	    sizeof(struct sockaddr_in6);
557 	def.sin6_family = mask.sin6_family = gw.sin6_family = AF_INET6;
558 	gw.sin6_addr = dr->rtaddr;
559 	gw.sin6_scope_id = 0;	/* XXX */
560 
561 	info.rti_flags = RTF_GATEWAY;
562 	info.rti_info[RTAX_DST] = (struct sockaddr *)&def;
563 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw;
564 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&mask;
565 
566 	rtrequest1(RTM_DELETE, &info, RTP_DEFAULT, &oldrt,
567 	    dr->ifp->if_rdomain);
568 	if (oldrt) {
569 		nd6_rtmsg(RTM_DELETE, oldrt);
570 		if (oldrt->rt_refcnt <= 0) {
571 			/*
572 			 * XXX: borrowed from the RTM_DELETE case of
573 			 * rtrequest1().
574 			 */
575 			oldrt->rt_refcnt++;
576 			rtfree(oldrt);
577 		}
578 	}
579 
580 	dr->installed = 0;
581 }
582 
583 /*
584  * remove all default routes from default router list
585  */
586 void
587 defrouter_reset(void)
588 {
589 	struct nd_defrouter *dr;
590 
591 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry)
592 		defrouter_delreq(dr);
593 
594 	/*
595 	 * XXX should we also nuke any default routers in the kernel, by
596 	 * going through them by rtalloc1()?
597 	 */
598 }
599 
600 /*
601  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
602  * draft-ietf-ipngwg-router-selection:
603  * 1) Routers that are reachable or probably reachable should be preferred.
604  *    If we have more than one (probably) reachable router, prefer ones
605  *    with the highest router preference.
606  * 2) When no routers on the list are known to be reachable or
607  *    probably reachable, routers SHOULD be selected in a round-robin
608  *    fashion, regardless of router preference values.
609  * 3) If the Default Router List is empty, assume that all
610  *    destinations are on-link.
611  *
612  * We assume nd_defrouter is sorted by router preference value.
613  * Since the code below covers both with and without router preference cases,
614  * we do not need to classify the cases by ifdef.
615  *
616  * At this moment, we do not try to install more than one default router,
617  * even when the multipath routing is available, because we're not sure about
618  * the benefits for stub hosts comparing to the risk of making the code
619  * complicated and the possibility of introducing bugs.
620  */
621 void
622 defrouter_select(void)
623 {
624 	int s = splsoftnet();
625 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
626 	struct rtentry *rt = NULL;
627 	struct llinfo_nd6 *ln = NULL;
628 
629 	/*
630 	 * This function should be called only when acting as an autoconfigured
631 	 * host.  Although the remaining part of this function is not effective
632 	 * if the node is not an autoconfigured host, we explicitly exclude
633 	 * such cases here for safety.
634 	 */
635 	if (ip6_forwarding || !ip6_accept_rtadv) {
636 		nd6log((LOG_WARNING,
637 		    "defrouter_select: called unexpectedly (forwarding=%d, "
638 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
639 		splx(s);
640 		return;
641 	}
642 
643 	/*
644 	 * Let's handle easy case (3) first:
645 	 * If default router list is empty, there's nothing to be done.
646 	 */
647 	if (TAILQ_EMPTY(&nd_defrouter)) {
648 		splx(s);
649 		return;
650 	}
651 
652 	/*
653 	 * Search for a (probably) reachable router from the list.
654 	 * We just pick up the first reachable one (if any), assuming that
655 	 * the ordering rule of the list described in defrtrlist_update().
656 	 */
657 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
658 		if (!selected_dr &&
659 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
660 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
661 		    ND6_IS_LLINFO_PROBREACH(ln)) {
662 			selected_dr = dr;
663 		}
664 
665 		if (dr->installed && !installed_dr)
666 			installed_dr = dr;
667 		else if (dr->installed && installed_dr) {
668 			/* this should not happen.  warn for diagnosis. */
669 			log(LOG_ERR, "defrouter_select: more than one router"
670 			    " is installed\n");
671 		}
672 	}
673 	/*
674 	 * If none of the default routers was found to be reachable,
675 	 * round-robin the list regardless of preference.
676 	 * Otherwise, if we have an installed router, check if the selected
677 	 * (reachable) router should really be preferred to the installed one.
678 	 * We only prefer the new router when the old one is not reachable
679 	 * or when the new one has a really higher preference value.
680 	 */
681 	if (!selected_dr) {
682 		if (!installed_dr || !TAILQ_NEXT(installed_dr, dr_entry))
683 			selected_dr = TAILQ_FIRST(&nd_defrouter);
684 		else
685 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
686 	} else if (installed_dr &&
687 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
688 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
689 	    ND6_IS_LLINFO_PROBREACH(ln) &&
690 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
691 		selected_dr = installed_dr;
692 	}
693 
694 	/*
695 	 * If the selected router is different than the installed one,
696 	 * remove the installed router and install the selected one.
697 	 * Note that the selected router is never NULL here.
698 	 */
699 	if (installed_dr != selected_dr) {
700 		if (installed_dr)
701 			defrouter_delreq(installed_dr);
702 		defrouter_addreq(selected_dr);
703 	}
704 
705 	splx(s);
706 	return;
707 }
708 
709 /*
710  * for default router selection
711  * regards router-preference field as a 2-bit signed integer
712  */
713 int
714 rtpref(struct nd_defrouter *dr)
715 {
716 #ifdef RTPREF
717 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
718 	case ND_RA_FLAG_RTPREF_HIGH:
719 		return RTPREF_HIGH;
720 	case ND_RA_FLAG_RTPREF_MEDIUM:
721 	case ND_RA_FLAG_RTPREF_RSV:
722 		return RTPREF_MEDIUM;
723 	case ND_RA_FLAG_RTPREF_LOW:
724 		return RTPREF_LOW;
725 	default:
726 		/*
727 		 * This case should never happen.  If it did, it would mean a
728 		 * serious bug of kernel internal.  We thus always bark here.
729 		 * Or, can we even panic?
730 		 */
731 		log(LOG_ERR, "rtpref: impossible RA flag %x", dr->flags);
732 		return RTPREF_INVALID;
733 	}
734 	/* NOTREACHED */
735 #else
736 	return 0;
737 #endif
738 }
739 
740 struct nd_defrouter *
741 defrtrlist_update(struct nd_defrouter *new)
742 {
743 	struct nd_defrouter *dr, *n;
744 	struct in6_ifextra *ext = new->ifp->if_afdata[AF_INET6];
745 	int s = splsoftnet();
746 
747 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
748 		/* entry exists */
749 		if (new->rtlifetime == 0) {
750 			defrtrlist_del(dr);
751 			dr = NULL;
752 		} else {
753 			int oldpref = rtpref(dr);
754 
755 			/* override */
756 			dr->flags = new->flags; /* xxx flag check */
757 			dr->rtlifetime = new->rtlifetime;
758 			dr->expire = new->expire;
759 
760 			if (!dr->installed)
761 				defrouter_select();
762 
763 			/*
764 			 * If the preference does not change, there's no need
765 			 * to sort the entries.
766 			 */
767 			if (rtpref(new) == oldpref) {
768 				splx(s);
769 				return (dr);
770 			}
771 
772 			/*
773 			 * preferred router may be changed, so relocate
774 			 * this router.
775 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
776 			 * However, since defrtrlist_del() has many side
777 			 * effects, we intentionally do so here.
778 			 * defrouter_select() below will handle routing
779 			 * changes later.
780 			 */
781 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
782 			n = dr;
783 			goto insert;
784 		}
785 		splx(s);
786 		return (dr);
787 	}
788 
789 	/* entry does not exist */
790 	if (new->rtlifetime == 0) {
791 		/* flush all possible redirects */
792 		if (!ip6_forwarding && ip6_accept_rtadv)
793 			rt6_flush(&new->rtaddr, new->ifp);
794 		splx(s);
795 		return (NULL);
796 	}
797 
798 	if (ip6_maxifdefrouters >= 0 &&
799 	    ext->ndefrouters >= ip6_maxifdefrouters) {
800 		splx(s);
801 		return (NULL);
802 	}
803 
804 	n = malloc(sizeof(*n), M_IP6NDP, M_NOWAIT | M_ZERO);
805 	if (n == NULL) {
806 		splx(s);
807 		return (NULL);
808 	}
809 	*n = *new;
810 
811 insert:
812 	/*
813 	 * Insert the new router in the Default Router List;
814 	 * The Default Router List should be in the descending order
815 	 * of router-preference.  Routers with the same preference are
816 	 * sorted in the arriving time order.
817 	 */
818 
819 	/* insert at the end of the group */
820 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry)
821 		if (rtpref(n) > rtpref(dr))
822 			break;
823 	if (dr)
824 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
825 	else
826 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
827 
828 	defrouter_select();
829 
830 	ext->ndefrouters++;
831 
832 	splx(s);
833 
834 	return (n);
835 }
836 
837 struct nd_pfxrouter *
838 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
839 {
840 	struct nd_pfxrouter *search;
841 
842 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
843 		if (search->router == dr)
844 			break;
845 	}
846 
847 	return (search);
848 }
849 
850 void
851 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
852 {
853 	struct nd_pfxrouter *new;
854 
855 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
856 	if (new == NULL)
857 		return;
858 	new->router = dr;
859 
860 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
861 
862 	pfxlist_onlink_check();
863 }
864 
865 void
866 pfxrtr_del(struct nd_pfxrouter *pfr)
867 {
868 	LIST_REMOVE(pfr, pfr_entry);
869 	free(pfr, M_IP6NDP);
870 }
871 
872 struct nd_prefix *
873 nd6_prefix_lookup(struct nd_prefix *pr)
874 {
875 	struct nd_prefix *search;
876 
877 	LIST_FOREACH(search, &nd_prefix, ndpr_entry) {
878 		if (pr->ndpr_ifp == search->ndpr_ifp &&
879 		    pr->ndpr_plen == search->ndpr_plen &&
880 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
881 		    &search->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
882 			break;
883 		}
884 	}
885 
886 	return (search);
887 }
888 
889 void
890 purge_detached(struct ifnet *ifp)
891 {
892 	struct nd_prefix *pr, *pr_next;
893 	struct in6_ifaddr *ia;
894 	struct ifaddr *ifa, *ifa_next;
895 
896 	for (pr = nd_prefix.lh_first; pr; pr = pr_next) {
897 		pr_next = pr->ndpr_next;
898 
899 		/*
900 		 * This function is called when we need to make more room for
901 		 * new prefixes rather than keeping old, possibly stale ones.
902 		 * Detached prefixes would be a good candidate; if all routers
903 		 * that advertised the prefix expired, the prefix is also
904 		 * probably stale.
905 		 */
906 		if (pr->ndpr_ifp != ifp ||
907 		    IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
908 		    ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
909 		    !LIST_EMPTY(&pr->ndpr_advrtrs)))
910 			continue;
911 
912 		for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa_next) {
913 			ifa_next = ifa->ifa_list.tqe_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 
927 int
928 nd6_prelist_add(struct nd_prefix *pr, struct nd_defrouter *dr,
929     struct nd_prefix **newp)
930 {
931 	struct nd_prefix *new = NULL;
932 	int i, s;
933 	struct in6_ifextra *ext = pr->ndpr_ifp->if_afdata[AF_INET6];
934 
935 	if (ip6_maxifprefixes >= 0) {
936 		if (ext->nprefixes >= ip6_maxifprefixes / 2)
937 			purge_detached(pr->ndpr_ifp);
938 		if (ext->nprefixes >= ip6_maxifprefixes)
939 			return(ENOMEM);
940 	}
941 
942 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
943 	if (new == NULL)
944 		return ENOMEM;
945 	*new = *pr;
946 	if (newp != NULL)
947 		*newp = new;
948 
949 	/* initialization */
950 	LIST_INIT(&new->ndpr_advrtrs);
951 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
952 	/* make prefix in the canonical form */
953 	for (i = 0; i < 4; i++)
954 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
955 		    new->ndpr_mask.s6_addr32[i];
956 
957 	s = splsoftnet();
958 	/* link ndpr_entry to nd_prefix list */
959 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
960 	splx(s);
961 
962 	/* ND_OPT_PI_FLAG_ONLINK processing */
963 	if (new->ndpr_raf_onlink) {
964 		int e;
965 
966 		if ((e = nd6_prefix_onlink(new)) != 0) {
967 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
968 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
969 			    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
970 			    pr->ndpr_plen, pr->ndpr_ifp->if_xname, e));
971 			/* proceed anyway. XXX: is it correct? */
972 		}
973 	}
974 
975 	if (dr)
976 		pfxrtr_add(new, dr);
977 
978 	ext->nprefixes++;
979 
980 	return 0;
981 }
982 
983 void
984 prelist_remove(struct nd_prefix *pr)
985 {
986 	struct nd_pfxrouter *pfr, *next;
987 	int e, s;
988 	struct in6_ifextra *ext = pr->ndpr_ifp->if_afdata[AF_INET6];
989 
990 	/* make sure to invalidate the prefix until it is really freed. */
991 	pr->ndpr_vltime = 0;
992 	pr->ndpr_pltime = 0;
993 #if 0
994 	/*
995 	 * Though these flags are now meaningless, we'd rather keep the value
996 	 * not to confuse users when executing "ndp -p".
997 	 */
998 	pr->ndpr_raf_onlink = 0;
999 	pr->ndpr_raf_auto = 0;
1000 #endif
1001 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
1002 	    (e = nd6_prefix_offlink(pr)) != 0) {
1003 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
1004 		    "on %s, errno=%d\n",
1005 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1006 		    pr->ndpr_plen, pr->ndpr_ifp->if_xname, e));
1007 		/* what should we do? */
1008 	}
1009 
1010 	if (pr->ndpr_refcnt > 0)
1011 		return;		/* notice here? */
1012 
1013 	s = splsoftnet();
1014 
1015 	/* unlink ndpr_entry from nd_prefix list */
1016 	LIST_REMOVE(pr, ndpr_entry);
1017 
1018 	/* free list of routers that adversed the prefix */
1019 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
1020 		next = LIST_NEXT(pfr, pfr_entry);
1021 
1022 		free(pfr, M_IP6NDP);
1023 	}
1024 
1025 	ext->nprefixes--;
1026 	if (ext->nprefixes < 0) {
1027 		log(LOG_WARNING, "prelist_remove: negative count on %s\n",
1028 		    pr->ndpr_ifp->if_xname);
1029 	}
1030 	splx(s);
1031 
1032 	free(pr, M_IP6NDP);
1033 
1034 	pfxlist_onlink_check();
1035 }
1036 
1037 /*
1038  * dr - may be NULL
1039  */
1040 
1041 int
1042 prelist_update(struct nd_prefix *new, struct nd_defrouter *dr, struct mbuf *m)
1043 {
1044 	struct in6_ifaddr *ia6_match = NULL;
1045 	struct ifaddr *ifa;
1046 	struct ifnet *ifp = new->ndpr_ifp;
1047 	struct nd_prefix *pr;
1048 	int s = splsoftnet();
1049 	int error = 0;
1050 	int tempaddr_preferred = 0;
1051 	int auth;
1052 	struct in6_addrlifetime lt6_tmp;
1053 
1054 	auth = 0;
1055 	if (m) {
1056 		/*
1057 		 * Authenticity for NA consists authentication for
1058 		 * both IP header and IP datagrams, doesn't it ?
1059 		 */
1060 		auth = ((m->m_flags & M_AUTH_AH) && (m->m_flags & M_AUTH));
1061 	}
1062 
1063 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1064 		/*
1065 		 * nd6_prefix_lookup() ensures that pr and new have the same
1066 		 * prefix on a same interface.
1067 		 */
1068 
1069 		/*
1070 		 * Update prefix information.  Note that the on-link (L) bit
1071 		 * and the autonomous (A) bit should NOT be changed from 1
1072 		 * to 0.
1073 		 */
1074 		if (new->ndpr_raf_onlink == 1)
1075 			pr->ndpr_raf_onlink = 1;
1076 		if (new->ndpr_raf_auto == 1)
1077 			pr->ndpr_raf_auto = 1;
1078 		if (new->ndpr_raf_onlink) {
1079 			pr->ndpr_vltime = new->ndpr_vltime;
1080 			pr->ndpr_pltime = new->ndpr_pltime;
1081 			pr->ndpr_preferred = new->ndpr_preferred;
1082 			pr->ndpr_expire = new->ndpr_expire;
1083 			pr->ndpr_lastupdate = new->ndpr_lastupdate;
1084 		}
1085 
1086 		if (new->ndpr_raf_onlink &&
1087 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1088 			int e;
1089 
1090 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1091 				nd6log((LOG_ERR,
1092 				    "prelist_update: failed to make "
1093 				    "the prefix %s/%d on-link on %s "
1094 				    "(errno=%d)\n",
1095 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1096 				    pr->ndpr_plen, pr->ndpr_ifp->if_xname, e));
1097 				/* proceed anyway. XXX: is it correct? */
1098 			}
1099 		}
1100 
1101 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1102 			pfxrtr_add(pr, dr);
1103 	} else {
1104 		struct nd_prefix *newpr = NULL;
1105 
1106 		if (new->ndpr_vltime == 0)
1107 			goto end;
1108 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1109 			goto end;
1110 
1111 		error = nd6_prelist_add(new, dr, &newpr);
1112 		if (error != 0 || newpr == NULL) {
1113 			nd6log((LOG_NOTICE, "prelist_update: "
1114 			    "nd6_prelist_add failed for %s/%d on %s "
1115 			    "errno=%d, returnpr=%p\n",
1116 			    ip6_sprintf(&new->ndpr_prefix.sin6_addr),
1117 			    new->ndpr_plen, new->ndpr_ifp->if_xname,
1118 			    error, newpr));
1119 			goto end; /* we should just give up in this case. */
1120 		}
1121 
1122 		/*
1123 		 * XXX: from the ND point of view, we can ignore a prefix
1124 		 * with the on-link bit being zero.  However, we need a
1125 		 * prefix structure for references from autoconfigured
1126 		 * addresses.  Thus, we explicitly make sure that the prefix
1127 		 * itself expires now.
1128 		 */
1129 		if (newpr->ndpr_raf_onlink == 0) {
1130 			newpr->ndpr_vltime = 0;
1131 			newpr->ndpr_pltime = 0;
1132 			in6_init_prefix_ltimes(newpr);
1133 		}
1134 
1135 		pr = newpr;
1136 	}
1137 
1138 	/*
1139 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1140 	 * Note that pr must be non NULL at this point.
1141 	 */
1142 
1143 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1144 	if (!new->ndpr_raf_auto)
1145 		goto end;
1146 
1147 	/*
1148 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1149 	 * nd6_ra_input.
1150 	 */
1151 
1152 	/*
1153 	 * 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime.
1154 	 * This should have been done in nd6_ra_input.
1155 	 */
1156 
1157 	/*
1158 	 * 5.5.3 (d). If the prefix advertised does not match the prefix of an
1159 	 * address already in the list, and the Valid Lifetime is not 0,
1160 	 * form an address.  Note that even a manually configured address
1161 	 * should reject autoconfiguration of a new address.
1162 	 */
1163 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1164 		struct in6_ifaddr *ifa6;
1165 		int ifa_plen;
1166 		u_int32_t storedlifetime;
1167 
1168 		if (ifa->ifa_addr->sa_family != AF_INET6)
1169 			continue;
1170 
1171 		ifa6 = (struct in6_ifaddr *)ifa;
1172 
1173 		/*
1174 		 * Spec is not clear here, but I believe we should concentrate
1175 		 * on unicast (i.e. not anycast) addresses.
1176 		 * XXX: other ia6_flags? detached or duplicated?
1177 		 */
1178 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1179 			continue;
1180 
1181 		ifa_plen = in6_mask2len(&ifa6->ia_prefixmask.sin6_addr, NULL);
1182 		if (ifa_plen != new->ndpr_plen ||
1183 		    !in6_are_prefix_equal(&ifa6->ia_addr.sin6_addr,
1184 		    &new->ndpr_prefix.sin6_addr, ifa_plen))
1185 			continue;
1186 
1187 		if (ia6_match == NULL) /* remember the first one */
1188 			ia6_match = ifa6;
1189 
1190 		if ((ifa6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1191 			continue;
1192 
1193 		/*
1194 		 * An already autoconfigured address matched.  Now that we
1195 		 * are sure there is at least one matched address, we can
1196 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1197 		 * "two hours" rule and the privacy extension.
1198 		 */
1199 #define TWOHOUR		(120*60)
1200 		/*
1201 		 * RFC2462 introduces the notion of StoredLifetime to the
1202 		 * "two hours" rule as follows:
1203 		 *   the Lifetime associated with the previously autoconfigured
1204 		 *   address.
1205 		 * Our interpretation of this definition is "the remaining
1206 		 * lifetime to expiration at the evaluation time".  One might
1207 		 * be wondering if this interpretation is really conform to the
1208 		 * RFC, because the text can read that "Lifetimes" are never
1209 		 * decreased, and our definition of the "storedlifetime" below
1210 		 * essentially reduces the "Valid Lifetime" advertised in the
1211 		 * previous RA.  But, this is due to the wording of the text,
1212 		 * and our interpretation is the same as an author's intention.
1213 		 * See the discussion in the IETF ipngwg ML in August 2001,
1214 		 * with the Subject "StoredLifetime in RFC 2462".
1215 		 */
1216 		lt6_tmp = ifa6->ia6_lifetime;
1217 
1218 		/* RFC 4941 temporary addresses (privacy extension). */
1219 		if (ifa6->ia6_flags & IN6_IFF_PRIVACY) {
1220 			/* Do we still have a non-deprecated address? */
1221 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0)
1222 				tempaddr_preferred = 1;
1223 			/* Don't extend lifetime for temporary addresses. */
1224 			if (new->ndpr_vltime >= lt6_tmp.ia6t_vltime)
1225 				continue;
1226 			if (new->ndpr_pltime >= lt6_tmp.ia6t_pltime)
1227 				continue;
1228 		}
1229 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1230 			storedlifetime = ND6_INFINITE_LIFETIME;
1231 		else if (time_second - ifa6->ia6_updatetime >
1232 			 lt6_tmp.ia6t_vltime) {
1233 			/*
1234 			 * The case of "invalid" address.  We should usually
1235 			 * not see this case.
1236 			 */
1237 			storedlifetime = 0;
1238 		} else
1239 			storedlifetime = lt6_tmp.ia6t_vltime -
1240 				(time_second - ifa6->ia6_updatetime);
1241 		if (TWOHOUR < new->ndpr_vltime ||
1242 		    storedlifetime < new->ndpr_vltime) {
1243 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1244 		} else if (storedlifetime <= TWOHOUR
1245 #if 0
1246 			   /*
1247 			    * This condition is logically redundant, so we just
1248 			    * omit it.
1249 			    * See IPng 6712, 6717, and 6721.
1250 			    */
1251 			   && new->ndpr_vltime <= storedlifetime
1252 #endif
1253 			) {
1254 			if (auth) {
1255 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1256 			}
1257 		} else {
1258 			/*
1259 			 * new->ndpr_vltime <= TWOHOUR &&
1260 			 * TWOHOUR < storedlifetime
1261 			 */
1262 			lt6_tmp.ia6t_vltime = TWOHOUR;
1263 		}
1264 
1265 		/* The 2 hour rule is not imposed for preferred lifetime. */
1266 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1267 
1268 		in6_init_address_ltimes(pr, &lt6_tmp);
1269 
1270 		ifa6->ia6_lifetime = lt6_tmp;
1271 		ifa6->ia6_updatetime = time_second;
1272 	}
1273 
1274 	if ((ia6_match == NULL ||
1275 	    (((ia6_match->ia6_flags & IN6_IFF_PRIVACY) ||
1276 	    (ifp->if_xflags & IFXF_INET6_PRIVACY)) && !tempaddr_preferred)) &&
1277 	    new->ndpr_vltime) {
1278 		/*
1279 		 * No address matched, or there is no preferred RFC 4941
1280 		 * temporary address. And the valid prefix lifetime is non-zero.
1281 		 * Create a new address in process context.
1282 		 */
1283 		pr->ndpr_refcnt++;
1284 		if (workq_add_task(NULL, 0, nd6_addr_add, pr, NULL))
1285 			pr->ndpr_refcnt--;
1286 	}
1287 
1288  end:
1289 	splx(s);
1290 	return error;
1291 }
1292 
1293 void
1294 nd6_addr_add(void *prptr, void *arg2)
1295 {
1296 	struct nd_prefix *pr = (struct nd_prefix *)prptr;
1297 	struct in6_ifaddr *ia6 = NULL;
1298 	struct ifaddr *ifa;
1299 	int ifa_plen;
1300 
1301 	/* Because prelist_update() runs in interrupt context it may run
1302 	 * again before this work queue task is run, causing multiple work
1303 	 * queue tasks to be scheduled all of which add addresses for the
1304 	 * same prefix. So check again if a non-deprecated address has already
1305 	 * been autoconfigured for this prefix. */
1306 	TAILQ_FOREACH(ifa, &pr->ndpr_ifp->if_addrlist, ifa_list) {
1307 		if (ifa->ifa_addr->sa_family != AF_INET6)
1308 			continue;
1309 
1310 		ia6 = (struct in6_ifaddr *)ifa;
1311 
1312 		/*
1313 		 * Spec is not clear here, but I believe we should concentrate
1314 		 * on unicast (i.e. not anycast) addresses.
1315 		 * XXX: other ia6_flags? detached or duplicated?
1316 		 */
1317 		if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1318 			continue;
1319 
1320 		if ((ia6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1321 			continue;
1322 
1323 		if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1324 			continue;
1325 
1326 		ifa_plen = in6_mask2len(&ia6->ia_prefixmask.sin6_addr, NULL);
1327 		if (ifa_plen == pr->ndpr_plen &&
1328 		    in6_are_prefix_equal(&ia6->ia_addr.sin6_addr,
1329 		    &pr->ndpr_prefix.sin6_addr, ifa_plen)) {
1330 			pr->ndpr_refcnt--;
1331 			return;
1332 		}
1333 	}
1334 
1335 	if ((ia6 = in6_ifadd(pr)) != NULL) {
1336 		ia6->ia6_ndpr = pr;
1337 
1338 		/*
1339 		 * A newly added address might affect the status
1340 		 * of other addresses, so we check and update it.
1341 		 * XXX: what if address duplication happens?
1342 		 */
1343 		pfxlist_onlink_check();
1344 	} else
1345 		pr->ndpr_refcnt--;
1346 }
1347 
1348 /*
1349  * A supplement function used in the on-link detection below;
1350  * detect if a given prefix has a (probably) reachable advertising router.
1351  * XXX: lengthy function name...
1352  */
1353 struct nd_pfxrouter *
1354 find_pfxlist_reachable_router(struct nd_prefix *pr)
1355 {
1356 	struct nd_pfxrouter *pfxrtr;
1357 	struct rtentry *rt;
1358 	struct llinfo_nd6 *ln;
1359 
1360 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1361 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1362 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1363 		    pfxrtr->router->ifp)) &&
1364 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1365 		    ND6_IS_LLINFO_PROBREACH(ln))
1366 			break;	/* found */
1367 	}
1368 
1369 	return (pfxrtr);
1370 }
1371 
1372 /*
1373  * Check if each prefix in the prefix list has at least one available router
1374  * that advertised the prefix (a router is "available" if its neighbor cache
1375  * entry is reachable or probably reachable).
1376  * If the check fails, the prefix may be off-link, because, for example,
1377  * we have moved from the network but the lifetime of the prefix has not
1378  * expired yet.  So we should not use the prefix if there is another prefix
1379  * that has an available router.
1380  * But, if there is no prefix that has an available router, we still regards
1381  * all the prefixes as on-link.  This is because we can't tell if all the
1382  * routers are simply dead or if we really moved from the network and there
1383  * is no router around us.
1384  */
1385 void
1386 pfxlist_onlink_check(void)
1387 {
1388 	struct nd_prefix *pr;
1389 	struct in6_ifaddr *ifa;
1390 
1391 	/*
1392 	 * Check if there is a prefix that has a reachable advertising
1393 	 * router.
1394 	 */
1395 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1396 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1397 			break;
1398 	}
1399 	if (pr != NULL || !TAILQ_EMPTY(&nd_defrouter)) {
1400 		/*
1401 		 * There is at least one prefix that has a reachable router,
1402 		 * or at least a router which probably does not advertise
1403 		 * any prefixes.  The latter would be the case when we move
1404 		 * to a new link where we have a router that does not provide
1405 		 * prefixes and we configure an address by hand.
1406 		 * Detach prefixes which have no reachable advertising
1407 		 * router, and attach other prefixes.
1408 		 */
1409 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1410 			/* XXX: a link-local prefix should never be detached */
1411 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1412 				continue;
1413 
1414 			/*
1415 			 * we aren't interested in prefixes without the L bit
1416 			 * set.
1417 			 */
1418 			if (pr->ndpr_raf_onlink == 0)
1419 				continue;
1420 
1421 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1422 			    find_pfxlist_reachable_router(pr) == NULL)
1423 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1424 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1425 			    find_pfxlist_reachable_router(pr) != 0)
1426 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1427 		}
1428 	} else {
1429 		/* there is no prefix that has a reachable router */
1430 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1431 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1432 				continue;
1433 
1434 			if (pr->ndpr_raf_onlink == 0)
1435 				continue;
1436 
1437 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1438 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1439 		}
1440 	}
1441 
1442 	/*
1443 	 * Remove each interface route associated with a (just) detached
1444 	 * prefix, and reinstall the interface route for a (just) attached
1445 	 * prefix.  Note that all attempt of reinstallation does not
1446 	 * necessarily success, when a same prefix is shared among multiple
1447 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1448 	 * so we don't have to care about them.
1449 	 */
1450 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1451 		int e;
1452 
1453 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1454 			continue;
1455 
1456 		if (pr->ndpr_raf_onlink == 0)
1457 			continue;
1458 
1459 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1460 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1461 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1462 				nd6log((LOG_ERR,
1463 				    "pfxlist_onlink_check: failed to "
1464 				    "make %s/%d offlink, errno=%d\n",
1465 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1466 				    pr->ndpr_plen, e));
1467 			}
1468 		}
1469 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1470 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1471 		    pr->ndpr_raf_onlink) {
1472 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1473 				nd6log((LOG_ERR,
1474 				    "pfxlist_onlink_check: failed to "
1475 				    "make %s/%d offlink, errno=%d\n",
1476 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1477 				    pr->ndpr_plen, e));
1478 			}
1479 		}
1480 	}
1481 
1482 	/*
1483 	 * Changes on the prefix status might affect address status as well.
1484 	 * Make sure that all addresses derived from an attached prefix are
1485 	 * attached, and that all addresses derived from a detached prefix are
1486 	 * detached.  Note, however, that a manually configured address should
1487 	 * always be attached.
1488 	 * The precise detection logic is same as the one for prefixes.
1489 	 */
1490 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1491 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1492 			continue;
1493 
1494 		if (ifa->ia6_ndpr == NULL) {
1495 			/*
1496 			 * This can happen when we first configure the address
1497 			 * (i.e. the address exists, but the prefix does not).
1498 			 * XXX: complicated relationships...
1499 			 */
1500 			continue;
1501 		}
1502 
1503 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1504 			break;
1505 	}
1506 	if (ifa) {
1507 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1508 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1509 				continue;
1510 
1511 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1512 				continue;
1513 
1514 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1515 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1516 			else
1517 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1518 		}
1519 	}
1520 	else {
1521 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1522 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1523 				continue;
1524 
1525 			ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1526 		}
1527 	}
1528 }
1529 
1530 int
1531 nd6_prefix_onlink(struct nd_prefix *pr)
1532 {
1533 	struct rt_addrinfo info;
1534 	struct ifaddr *ifa;
1535 	struct ifnet *ifp = pr->ndpr_ifp;
1536 	struct sockaddr_in6 mask6;
1537 	struct nd_prefix *opr;
1538 	u_long rtflags;
1539 	int error = 0;
1540 	struct rtentry *rt = NULL;
1541 
1542 	/* sanity check */
1543 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1544 		nd6log((LOG_ERR,
1545 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1546 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1547 		return (EEXIST);
1548 	}
1549 
1550 	/*
1551 	 * Add the interface route associated with the prefix.  Before
1552 	 * installing the route, check if there's the same prefix on another
1553 	 * interface, and the prefix has already installed the interface route.
1554 	 * Although such a configuration is expected to be rare, we explicitly
1555 	 * allow it.
1556 	 */
1557 	LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1558 		if (opr == pr)
1559 			continue;
1560 
1561 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1562 			continue;
1563 
1564 		if (opr->ndpr_plen == pr->ndpr_plen &&
1565 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1566 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1567 			return (0);
1568 	}
1569 
1570 	/*
1571 	 * We prefer link-local addresses as the associated interface address.
1572 	 */
1573 	/* search for a link-local addr */
1574 	ifa = &in6ifa_ifpforlinklocal(ifp,
1575 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST)->ia_ifa;
1576 	if (ifa == NULL) {
1577 		/* XXX: freebsd does not have ifa_ifwithaf */
1578 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1579 			if (ifa->ifa_addr->sa_family == AF_INET6)
1580 				break;
1581 		}
1582 		/* should we care about ia6_flags? */
1583 	}
1584 	if (ifa == NULL) {
1585 		/*
1586 		 * This can still happen, when, for example, we receive an RA
1587 		 * containing a prefix with the L bit set and the A bit clear,
1588 		 * after removing all IPv6 addresses on the receiving
1589 		 * interface.  This should, of course, be rare though.
1590 		 */
1591 		nd6log((LOG_NOTICE,
1592 		    "nd6_prefix_onlink: failed to find any ifaddr"
1593 		    " to add route for a prefix(%s/%d) on %s\n",
1594 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1595 		    pr->ndpr_plen, ifp->if_xname));
1596 		return (0);
1597 	}
1598 
1599 	/*
1600 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1601 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1602 	 */
1603 	bzero(&mask6, sizeof(mask6));
1604 	mask6.sin6_len = sizeof(mask6);
1605 	mask6.sin6_addr = pr->ndpr_mask;
1606 	/* rtrequest1() will probably set RTF_UP, but we're not sure. */
1607 	rtflags = ifa->ifa_flags | RTF_UP;
1608 	if (nd6_need_cache(ifp)) {
1609 		/* explicitly set in case ifa_flags does not set the flag. */
1610 		rtflags |= RTF_CLONING;
1611 	} else {
1612 		/*
1613 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1614 		 */
1615 		rtflags &= ~RTF_CLONING;
1616 	}
1617 
1618 	bzero(&info, sizeof(info));
1619 	info.rti_flags = rtflags;
1620 	info.rti_info[RTAX_DST] = (struct sockaddr *)&pr->ndpr_prefix;
1621 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
1622 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&mask6;
1623 
1624 	error = rtrequest1(RTM_ADD, &info, RTP_CONNECTED, &rt, ifp->if_rdomain);
1625 	if (error == 0) {
1626 		if (rt != NULL) /* this should be non NULL, though */
1627 			nd6_rtmsg(RTM_ADD, rt);
1628 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1629 	} else {
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(&pr->ndpr_prefix.sin6_addr),
1634 		    pr->ndpr_plen, ifp->if_xname,
1635 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1636 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
1637 	}
1638 
1639 	if (rt != NULL)
1640 		rt->rt_refcnt--;
1641 
1642 	return (error);
1643 }
1644 
1645 int
1646 nd6_prefix_offlink(struct nd_prefix *pr)
1647 {
1648 	struct rt_addrinfo info;
1649 	int error = 0;
1650 	struct ifnet *ifp = pr->ndpr_ifp;
1651 	struct nd_prefix *opr;
1652 	struct sockaddr_in6 sa6, mask6;
1653 	struct rtentry *rt = NULL;
1654 
1655 	/* sanity check */
1656 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1657 		nd6log((LOG_ERR,
1658 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1659 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1660 		return (EEXIST);
1661 	}
1662 
1663 	bzero(&sa6, sizeof(sa6));
1664 	sa6.sin6_family = AF_INET6;
1665 	sa6.sin6_len = sizeof(sa6);
1666 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1667 	    sizeof(struct in6_addr));
1668 	bzero(&mask6, sizeof(mask6));
1669 	mask6.sin6_family = AF_INET6;
1670 	mask6.sin6_len = sizeof(sa6);
1671 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1672 	bzero(&info, sizeof(info));
1673 	info.rti_info[RTAX_DST] = (struct sockaddr *)&sa6;
1674 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&mask6;
1675 	error = rtrequest1(RTM_DELETE, &info, RTP_CONNECTED, &rt,
1676 	    ifp->if_rdomain);
1677 	if (error == 0) {
1678 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1679 
1680 		/* report the route deletion to the routing socket. */
1681 		if (rt != NULL)
1682 			nd6_rtmsg(RTM_DELETE, rt);
1683 
1684 		/*
1685 		 * There might be the same prefix on another interface,
1686 		 * the prefix which could not be on-link just because we have
1687 		 * the interface route (see comments in nd6_prefix_onlink).
1688 		 * If there's one, try to make the prefix on-link on the
1689 		 * interface.
1690 		 */
1691 		LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1692 			if (opr == pr)
1693 				continue;
1694 
1695 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1696 				continue;
1697 
1698 			/*
1699 			 * KAME specific: detached prefixes should not be
1700 			 * on-link.
1701 			 */
1702 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1703 				continue;
1704 
1705 			if (opr->ndpr_plen == pr->ndpr_plen &&
1706 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1707 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1708 				int e;
1709 
1710 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1711 					nd6log((LOG_ERR,
1712 					    "nd6_prefix_offlink: failed to "
1713 					    "recover a prefix %s/%d from %s "
1714 					    "to %s (errno = %d)\n",
1715 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1716 					    opr->ndpr_plen, ifp->if_xname,
1717 					    opr->ndpr_ifp->if_xname, e));
1718 				}
1719 			}
1720 		}
1721 	} else {
1722 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1723 		nd6log((LOG_ERR,
1724 		    "nd6_prefix_offlink: failed to delete route: "
1725 		    "%s/%d on %s (errno = %d)\n",
1726 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, ifp->if_xname,
1727 		    error));
1728 	}
1729 
1730 	if (rt != NULL) {
1731 		if (rt->rt_refcnt <= 0) {
1732 			/* XXX: we should free the entry ourselves. */
1733 			rt->rt_refcnt++;
1734 			rtfree(rt);
1735 		}
1736 	}
1737 
1738 	return (error);
1739 }
1740 
1741 struct in6_ifaddr *
1742 in6_ifadd(struct nd_prefix *pr)
1743 {
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, s, plen0;
1749 	struct in6_addr mask, rand_ifid;
1750 	int prefixlen = pr->ndpr_plen;
1751 
1752 	in6_prefixlen2mask(&mask, prefixlen);
1753 
1754 	/*
1755 	 * find a link-local address (will be interface ID).
1756 	 * Is it really mandatory? Theoretically, a global or a site-local
1757 	 * address can be configured without a link-local address, if we
1758 	 * have a unique interface identifier...
1759 	 *
1760 	 * it is not mandatory to have a link-local address, we can generate
1761 	 * interface identifier on the fly.  we do this because:
1762 	 * (1) it should be the easiest way to find interface identifier.
1763 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1764 	 * for multiple addresses on a single interface, and possible shortcut
1765 	 * of DAD.  we omitted DAD for this reason in the past.
1766 	 * (3) a user can prevent autoconfiguration of global address
1767 	 * by removing link-local address by hand (this is partly because we
1768 	 * don't have other way to control the use of IPv6 on a interface.
1769 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1770 	 * (4) it is easier to manage when an interface has addresses
1771 	 * with the same interface identifier, than to have multiple addresses
1772 	 * with different interface identifiers.
1773 	 */
1774 	ifa = &in6ifa_ifpforlinklocal(ifp, 0)->ia_ifa; /* 0 is OK? */
1775 	if (ifa)
1776 		ib = (struct in6_ifaddr *)ifa;
1777 	else
1778 		return NULL;
1779 
1780 #if 0 /* don't care link local addr state, and always do DAD */
1781 	/* if link-local address is not eligible, do not autoconfigure. */
1782 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1783 		printf("in6_ifadd: link-local address not ready\n");
1784 		return NULL;
1785 	}
1786 #endif
1787 
1788 	/* prefixlen + ifidlen must be equal to 128 */
1789 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1790 	if (prefixlen != plen0) {
1791 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1792 		    "(prefix=%d ifid=%d)\n",
1793 		    ifp->if_xname, prefixlen, 128 - plen0));
1794 		return NULL;
1795 	}
1796 
1797 	/* make ifaddr */
1798 
1799 	bzero(&ifra, sizeof(ifra));
1800 	/*
1801 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1802 	 * for safety.
1803 	 */
1804 	strncpy(ifra.ifra_name, ifp->if_xname, sizeof(ifra.ifra_name));
1805 	ifra.ifra_addr.sin6_family = AF_INET6;
1806 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1807 	/* prefix */
1808 	bcopy(&pr->ndpr_prefix.sin6_addr, &ifra.ifra_addr.sin6_addr,
1809 	    sizeof(ifra.ifra_addr.sin6_addr));
1810 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1811 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1812 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1813 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1814 
1815 	/* interface ID */
1816 	if (ifp->if_xflags & IFXF_INET6_PRIVACY) {
1817 		ifra.ifra_flags |= IN6_IFF_PRIVACY;
1818 		bcopy(&pr->ndpr_prefix.sin6_addr, &rand_ifid,
1819 		    sizeof(rand_ifid));
1820 		in6_get_rand_ifid(ifp, &rand_ifid);
1821 		ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1822 		    (rand_ifid.s6_addr32[0] & ~mask.s6_addr32[0]);
1823 		ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1824 		    (rand_ifid.s6_addr32[1] & ~mask.s6_addr32[1]);
1825 		ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1826 		    (rand_ifid.s6_addr32[2] & ~mask.s6_addr32[2]);
1827 		ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1828 		    (rand_ifid.s6_addr32[3] & ~mask.s6_addr32[3]);
1829 	} else {
1830 		ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1831 		    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1832 		ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1833 		    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1834 		ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1835 		    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1836 		ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1837 		    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1838 	}
1839 
1840 	/* new prefix mask. */
1841 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1842 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1843 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1844 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1845 
1846 	/*
1847 	 * lifetime.
1848 	 * XXX: in6_init_address_ltimes would override these values later.
1849 	 * We should reconsider this logic.
1850 	 */
1851 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1852 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1853 
1854 	if (ifp->if_xflags & IFXF_INET6_PRIVACY) {
1855 	    if (ifra.ifra_lifetime.ia6t_vltime > ND6_PRIV_VALID_LIFETIME)
1856 		ifra.ifra_lifetime.ia6t_vltime = ND6_PRIV_VALID_LIFETIME;
1857 	    if (ifra.ifra_lifetime.ia6t_pltime > ND6_PRIV_PREFERRED_LIFETIME)
1858 		ifra.ifra_lifetime.ia6t_pltime = ND6_PRIV_PREFERRED_LIFETIME
1859 			- (arc4random() % ND6_PRIV_MAX_DESYNC_FACTOR);
1860 	}
1861 
1862 	/* XXX: scope zone ID? */
1863 
1864 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1865 
1866 	/* allocate ifaddr structure, link into chain, etc. */
1867 	s = splsoftnet();
1868 	error = in6_update_ifa(ifp, &ifra, NULL);
1869 	splx(s);
1870 
1871 	if (error != 0) {
1872 		nd6log((LOG_ERR,
1873 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1874 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), ifp->if_xname,
1875 		    error));
1876 		return (NULL);	/* ifaddr must not have been allocated. */
1877 	}
1878 
1879 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1880 
1881 	return (ia);		/* this is always non-NULL */
1882 }
1883 
1884 int
1885 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1886 {
1887 
1888 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
1889 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
1890 		nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
1891 		    "(%d) is greater than valid lifetime(%d)\n",
1892 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
1893 		return (EINVAL);
1894 	}
1895 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1896 		ndpr->ndpr_preferred = 0;
1897 	else
1898 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
1899 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1900 		ndpr->ndpr_expire = 0;
1901 	else
1902 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
1903 
1904 	return 0;
1905 }
1906 
1907 void
1908 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
1909 {
1910 
1911 	/* Valid lifetime must not be updated unless explicitly specified. */
1912 	/* init ia6t_expire */
1913 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
1914 		lt6->ia6t_expire = 0;
1915 	else {
1916 		lt6->ia6t_expire = time_second;
1917 		lt6->ia6t_expire += lt6->ia6t_vltime;
1918 	}
1919 
1920 	/* init ia6t_preferred */
1921 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
1922 		lt6->ia6t_preferred = 0;
1923 	else {
1924 		lt6->ia6t_preferred = time_second;
1925 		lt6->ia6t_preferred += lt6->ia6t_pltime;
1926 	}
1927 }
1928 
1929 /*
1930  * Delete all the routing table entries that use the specified gateway.
1931  * XXX: this function causes search through all entries of routing table, so
1932  * it shouldn't be called when acting as a router.
1933  */
1934 void
1935 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
1936 {
1937 	struct radix_node_head *rnh = rt_gettable(AF_INET6, 0);
1938 	int s = splsoftnet();
1939 
1940 	/* We'll care only link-local addresses */
1941 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
1942 		splx(s);
1943 		return;
1944 	}
1945 	/* XXX: hack for KAME's link-local address kludge */
1946 	gateway->s6_addr16[1] = htons(ifp->if_index);
1947 
1948 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
1949 	splx(s);
1950 }
1951 
1952 int
1953 rt6_deleteroute(struct radix_node *rn, void *arg, u_int id)
1954 {
1955 #define SIN6(s)	((struct sockaddr_in6 *)s)
1956 	struct rt_addrinfo info;
1957 	struct rtentry *rt = (struct rtentry *)rn;
1958 	struct in6_addr *gate = (struct in6_addr *)arg;
1959 
1960 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
1961 		return (0);
1962 
1963 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr))
1964 		return (0);
1965 
1966 	/*
1967 	 * Do not delete a static route.
1968 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
1969 	 * 'cloned' bit instead?
1970 	 */
1971 	if ((rt->rt_flags & RTF_STATIC) != 0)
1972 		return (0);
1973 
1974 	/*
1975 	 * We delete only host route. This means, in particular, we don't
1976 	 * delete default route.
1977 	 */
1978 	if ((rt->rt_flags & RTF_HOST) == 0)
1979 		return (0);
1980 
1981 	bzero(&info, sizeof(info));
1982 	info.rti_flags =  rt->rt_flags;
1983 	info.rti_info[RTAX_DST] = rt_key(rt);
1984 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1985 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1986 	return (rtrequest1(RTM_DELETE, &info, RTP_ANY, NULL, id));
1987 #undef SIN6
1988 }
1989 
1990 int
1991 nd6_setdefaultiface(int ifindex)
1992 {
1993 	int error = 0;
1994 
1995 	if (ifindex < 0 || if_indexlim <= ifindex)
1996 		return (EINVAL);
1997 	if (ifindex != 0 && !ifindex2ifnet[ifindex])
1998 		return (EINVAL);
1999 
2000 	if (nd6_defifindex != ifindex) {
2001 		nd6_defifindex = ifindex;
2002 		if (nd6_defifindex > 0) {
2003 			nd6_defifp = ifindex2ifnet[nd6_defifindex];
2004 		} else
2005 			nd6_defifp = NULL;
2006 	}
2007 
2008 	return (error);
2009 }
2010