xref: /freebsd/sys/netinet6/nd6_rtr.c (revision 30d239bc4c510432e65a84fa1c14ed67a3ab1c92)
1 /*	$FreeBSD$	*/
2 /*	$KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei 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 "opt_inet.h"
34 #include "opt_inet6.h"
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/malloc.h>
39 #include <sys/mbuf.h>
40 #include <sys/socket.h>
41 #include <sys/sockio.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/errno.h>
45 #include <sys/syslog.h>
46 #include <sys/queue.h>
47 
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/if_dl.h>
51 #include <net/route.h>
52 #include <net/radix.h>
53 
54 #include <netinet/in.h>
55 #include <netinet6/in6_var.h>
56 #include <netinet6/in6_ifattach.h>
57 #include <netinet/ip6.h>
58 #include <netinet6/ip6_var.h>
59 #include <netinet6/nd6.h>
60 #include <netinet/icmp6.h>
61 #include <netinet6/scope6_var.h>
62 
63 #define SDL(s)	((struct sockaddr_dl *)s)
64 
65 static int rtpref __P((struct nd_defrouter *));
66 static struct nd_defrouter *defrtrlist_update __P((struct nd_defrouter *));
67 static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
68     struct mbuf *, int));
69 static struct in6_ifaddr *in6_ifadd __P((struct nd_prefixctl *,	int));
70 static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
71 	struct nd_defrouter *));
72 static void pfxrtr_add __P((struct nd_prefix *, struct nd_defrouter *));
73 static void pfxrtr_del __P((struct nd_pfxrouter *));
74 static struct nd_pfxrouter *find_pfxlist_reachable_router
75 	__P((struct nd_prefix *));
76 static void defrouter_delreq __P((struct nd_defrouter *));
77 static void nd6_rtmsg __P((int, struct rtentry *));
78 
79 static int in6_init_prefix_ltimes __P((struct nd_prefix *));
80 static void in6_init_address_ltimes __P((struct nd_prefix *,
81 	struct in6_addrlifetime *));
82 
83 static int rt6_deleteroute __P((struct radix_node *, void *));
84 
85 extern int nd6_recalc_reachtm_interval;
86 
87 static struct ifnet *nd6_defifp;
88 int nd6_defifindex;
89 
90 int ip6_use_tempaddr = 0;
91 
92 int ip6_desync_factor;
93 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
94 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
95 /*
96  * shorter lifetimes for debugging purposes.
97 int ip6_temp_preferred_lifetime = 800;
98 static int ip6_temp_valid_lifetime = 1800;
99 */
100 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
101 
102 /* RTPREF_MEDIUM has to be 0! */
103 #define RTPREF_HIGH	1
104 #define RTPREF_MEDIUM	0
105 #define RTPREF_LOW	(-1)
106 #define RTPREF_RESERVED	(-2)
107 #define RTPREF_INVALID	(-3)	/* internal */
108 
109 /*
110  * Receive Router Solicitation Message - just for routers.
111  * Router solicitation/advertisement is mostly managed by userland program
112  * (rtadvd) so here we have no function like nd6_ra_output().
113  *
114  * Based on RFC 2461
115  */
116 void
117 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
118 {
119 	struct ifnet *ifp = m->m_pkthdr.rcvif;
120 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
121 	struct nd_router_solicit *nd_rs;
122 	struct in6_addr saddr6 = ip6->ip6_src;
123 	char *lladdr = NULL;
124 	int lladdrlen = 0;
125 	union nd_opts ndopts;
126 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
127 
128 	/* If I'm not a router, ignore it. */
129 	if (ip6_accept_rtadv != 0 || ip6_forwarding != 1)
130 		goto freeit;
131 
132 	/* Sanity checks */
133 	if (ip6->ip6_hlim != 255) {
134 		nd6log((LOG_ERR,
135 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
136 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
137 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
138 		goto bad;
139 	}
140 
141 	/*
142 	 * Don't update the neighbor cache, if src = ::.
143 	 * This indicates that the src has no IP address assigned yet.
144 	 */
145 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
146 		goto freeit;
147 
148 #ifndef PULLDOWN_TEST
149 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
150 	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
151 #else
152 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
153 	if (nd_rs == NULL) {
154 		icmp6stat.icp6s_tooshort++;
155 		return;
156 	}
157 #endif
158 
159 	icmp6len -= sizeof(*nd_rs);
160 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
161 	if (nd6_options(&ndopts) < 0) {
162 		nd6log((LOG_INFO,
163 		    "nd6_rs_input: invalid ND option, ignored\n"));
164 		/* nd6_options have incremented stats */
165 		goto freeit;
166 	}
167 
168 	if (ndopts.nd_opts_src_lladdr) {
169 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
170 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
171 	}
172 
173 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
174 		nd6log((LOG_INFO,
175 		    "nd6_rs_input: lladdrlen mismatch for %s "
176 		    "(if %d, RS packet %d)\n",
177 		    ip6_sprintf(ip6bufs, &saddr6),
178 		    ifp->if_addrlen, lladdrlen - 2));
179 		goto bad;
180 	}
181 
182 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
183 
184  freeit:
185 	m_freem(m);
186 	return;
187 
188  bad:
189 	icmp6stat.icp6s_badrs++;
190 	m_freem(m);
191 }
192 
193 /*
194  * Receive Router Advertisement Message.
195  *
196  * Based on RFC 2461
197  * TODO: on-link bit on prefix information
198  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
199  */
200 void
201 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
202 {
203 	struct ifnet *ifp = m->m_pkthdr.rcvif;
204 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
205 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
206 	struct nd_router_advert *nd_ra;
207 	struct in6_addr saddr6 = ip6->ip6_src;
208 	int mcast = 0;
209 	union nd_opts ndopts;
210 	struct nd_defrouter *dr;
211 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
212 
213 	/*
214 	 * We only accept RAs only when
215 	 * the system-wide variable allows the acceptance, and
216 	 * per-interface variable allows RAs on the receiving interface.
217 	 */
218 	if (ip6_accept_rtadv == 0)
219 		goto freeit;
220 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
221 		goto freeit;
222 
223 	if (ip6->ip6_hlim != 255) {
224 		nd6log((LOG_ERR,
225 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
226 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
227 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
228 		goto bad;
229 	}
230 
231 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
232 		nd6log((LOG_ERR,
233 		    "nd6_ra_input: src %s is not link-local\n",
234 		    ip6_sprintf(ip6bufs, &saddr6)));
235 		goto bad;
236 	}
237 
238 #ifndef PULLDOWN_TEST
239 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
240 	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
241 #else
242 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
243 	if (nd_ra == NULL) {
244 		icmp6stat.icp6s_tooshort++;
245 		return;
246 	}
247 #endif
248 
249 	icmp6len -= sizeof(*nd_ra);
250 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
251 	if (nd6_options(&ndopts) < 0) {
252 		nd6log((LOG_INFO,
253 		    "nd6_ra_input: invalid ND option, ignored\n"));
254 		/* nd6_options have incremented stats */
255 		goto freeit;
256 	}
257 
258     {
259 	struct nd_defrouter dr0;
260 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
261 
262 	/* remember if this is a multicasted advertisement */
263 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
264 		mcast = 1;
265 
266 	bzero(&dr0, sizeof(dr0));
267 	dr0.rtaddr = saddr6;
268 	dr0.flags  = nd_ra->nd_ra_flags_reserved;
269 	dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
270 	dr0.expire = time_second + dr0.rtlifetime;
271 	dr0.ifp = ifp;
272 	/* unspecified or not? (RFC 2461 6.3.4) */
273 	if (advreachable) {
274 		advreachable = ntohl(advreachable);
275 		if (advreachable <= MAX_REACHABLE_TIME &&
276 		    ndi->basereachable != advreachable) {
277 			ndi->basereachable = advreachable;
278 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
279 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
280 		}
281 	}
282 	if (nd_ra->nd_ra_retransmit)
283 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
284 	if (nd_ra->nd_ra_curhoplimit)
285 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
286 	dr = defrtrlist_update(&dr0);
287     }
288 
289 	/*
290 	 * prefix
291 	 */
292 	if (ndopts.nd_opts_pi) {
293 		struct nd_opt_hdr *pt;
294 		struct nd_opt_prefix_info *pi = NULL;
295 		struct nd_prefixctl pr;
296 
297 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
298 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
299 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
300 						(pt->nd_opt_len << 3))) {
301 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
302 				continue;
303 			pi = (struct nd_opt_prefix_info *)pt;
304 
305 			if (pi->nd_opt_pi_len != 4) {
306 				nd6log((LOG_INFO,
307 				    "nd6_ra_input: invalid option "
308 				    "len %d for prefix information option, "
309 				    "ignored\n", pi->nd_opt_pi_len));
310 				continue;
311 			}
312 
313 			if (128 < pi->nd_opt_pi_prefix_len) {
314 				nd6log((LOG_INFO,
315 				    "nd6_ra_input: invalid prefix "
316 				    "len %d for prefix information option, "
317 				    "ignored\n", pi->nd_opt_pi_prefix_len));
318 				continue;
319 			}
320 
321 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
322 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
323 				nd6log((LOG_INFO,
324 				    "nd6_ra_input: invalid prefix "
325 				    "%s, ignored\n",
326 				    ip6_sprintf(ip6bufs,
327 					&pi->nd_opt_pi_prefix)));
328 				continue;
329 			}
330 
331 			bzero(&pr, sizeof(pr));
332 			pr.ndpr_prefix.sin6_family = AF_INET6;
333 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
334 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
335 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
336 
337 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
338 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
339 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
340 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
341 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
342 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
343 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
344 			(void)prelist_update(&pr, dr, m, mcast);
345 		}
346 	}
347 
348 	/*
349 	 * MTU
350 	 */
351 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
352 		u_long mtu;
353 		u_long maxmtu;
354 
355 		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
356 
357 		/* lower bound */
358 		if (mtu < IPV6_MMTU) {
359 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
360 			    "mtu=%lu sent from %s, ignoring\n",
361 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
362 			goto skip;
363 		}
364 
365 		/* upper bound */
366 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
367 		    ? ndi->maxmtu : ifp->if_mtu;
368 		if (mtu <= maxmtu) {
369 			int change = (ndi->linkmtu != mtu);
370 
371 			ndi->linkmtu = mtu;
372 			if (change) /* in6_maxmtu may change */
373 				in6_setmaxmtu();
374 		} else {
375 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
376 			    "mtu=%lu sent from %s; "
377 			    "exceeds maxmtu %lu, ignoring\n",
378 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
379 		}
380 	}
381 
382  skip:
383 
384 	/*
385 	 * Source link layer address
386 	 */
387     {
388 	char *lladdr = NULL;
389 	int lladdrlen = 0;
390 
391 	if (ndopts.nd_opts_src_lladdr) {
392 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
393 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
394 	}
395 
396 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
397 		nd6log((LOG_INFO,
398 		    "nd6_ra_input: lladdrlen mismatch for %s "
399 		    "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
400 		    ifp->if_addrlen, lladdrlen - 2));
401 		goto bad;
402 	}
403 
404 	nd6_cache_lladdr(ifp, &saddr6, lladdr,
405 	    lladdrlen, ND_ROUTER_ADVERT, 0);
406 
407 	/*
408 	 * Installing a link-layer address might change the state of the
409 	 * router's neighbor cache, which might also affect our on-link
410 	 * detection of adveritsed prefixes.
411 	 */
412 	pfxlist_onlink_check();
413     }
414 
415  freeit:
416 	m_freem(m);
417 	return;
418 
419  bad:
420 	icmp6stat.icp6s_badra++;
421 	m_freem(m);
422 }
423 
424 /*
425  * default router list proccessing sub routines
426  */
427 
428 /* tell the change to user processes watching the routing socket. */
429 static void
430 nd6_rtmsg(int cmd, struct rtentry *rt)
431 {
432 	struct rt_addrinfo info;
433 
434 	bzero((caddr_t)&info, sizeof(info));
435 	info.rti_info[RTAX_DST] = rt_key(rt);
436 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
437 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
438 	if (rt->rt_ifp) {
439 		info.rti_info[RTAX_IFP] =
440 		    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
441 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
442 	}
443 
444 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
445 }
446 
447 void
448 defrouter_addreq(struct nd_defrouter *new)
449 {
450 	struct sockaddr_in6 def, mask, gate;
451 	struct rtentry *newrt = NULL;
452 	int s;
453 	int error;
454 
455 	bzero(&def, sizeof(def));
456 	bzero(&mask, sizeof(mask));
457 	bzero(&gate, sizeof(gate));
458 
459 	def.sin6_len = mask.sin6_len = gate.sin6_len =
460 	    sizeof(struct sockaddr_in6);
461 	def.sin6_family = gate.sin6_family = AF_INET6;
462 	gate.sin6_addr = new->rtaddr;
463 
464 	s = splnet();
465 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
466 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
467 	    RTF_GATEWAY, &newrt);
468 	if (newrt) {
469 		RT_LOCK(newrt);
470 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
471 		RT_REMREF(newrt);
472 		RT_UNLOCK(newrt);
473 	}
474 	if (error == 0)
475 		new->installed = 1;
476 	splx(s);
477 	return;
478 }
479 
480 struct nd_defrouter *
481 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
482 {
483 	struct nd_defrouter *dr;
484 
485 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
486 	     dr = TAILQ_NEXT(dr, dr_entry)) {
487 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
488 			return (dr);
489 	}
490 
491 	return (NULL);		/* search failed */
492 }
493 
494 /*
495  * Remove the default route for a given router.
496  * This is just a subroutine function for defrouter_select(), and should
497  * not be called from anywhere else.
498  */
499 static void
500 defrouter_delreq(struct nd_defrouter *dr)
501 {
502 	struct sockaddr_in6 def, mask, gate;
503 	struct rtentry *oldrt = NULL;
504 
505 	bzero(&def, sizeof(def));
506 	bzero(&mask, sizeof(mask));
507 	bzero(&gate, sizeof(gate));
508 
509 	def.sin6_len = mask.sin6_len = gate.sin6_len =
510 	    sizeof(struct sockaddr_in6);
511 	def.sin6_family = gate.sin6_family = AF_INET6;
512 	gate.sin6_addr = dr->rtaddr;
513 
514 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
515 	    (struct sockaddr *)&gate,
516 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
517 	if (oldrt) {
518 		nd6_rtmsg(RTM_DELETE, oldrt);
519 		RTFREE(oldrt);
520 	}
521 
522 	dr->installed = 0;
523 }
524 
525 /*
526  * remove all default routes from default router list
527  */
528 void
529 defrouter_reset(void)
530 {
531 	struct nd_defrouter *dr;
532 
533 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
534 	     dr = TAILQ_NEXT(dr, dr_entry))
535 		defrouter_delreq(dr);
536 
537 	/*
538 	 * XXX should we also nuke any default routers in the kernel, by
539 	 * going through them by rtalloc1()?
540 	 */
541 }
542 
543 void
544 defrtrlist_del(struct nd_defrouter *dr)
545 {
546 	struct nd_defrouter *deldr = NULL;
547 	struct nd_prefix *pr;
548 
549 	/*
550 	 * Flush all the routing table entries that use the router
551 	 * as a next hop.
552 	 */
553 	if (!ip6_forwarding && ip6_accept_rtadv) /* XXX: better condition? */
554 		rt6_flush(&dr->rtaddr, dr->ifp);
555 
556 	if (dr->installed) {
557 		deldr = dr;
558 		defrouter_delreq(dr);
559 	}
560 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
561 
562 	/*
563 	 * Also delete all the pointers to the router in each prefix lists.
564 	 */
565 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
566 		struct nd_pfxrouter *pfxrtr;
567 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
568 			pfxrtr_del(pfxrtr);
569 	}
570 	pfxlist_onlink_check();
571 
572 	/*
573 	 * If the router is the primary one, choose a new one.
574 	 * Note that defrouter_select() will remove the current gateway
575 	 * from the routing table.
576 	 */
577 	if (deldr)
578 		defrouter_select();
579 
580 	free(dr, M_IP6NDP);
581 }
582 
583 /*
584  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
585  * draft-ietf-ipngwg-router-selection:
586  * 1) Routers that are reachable or probably reachable should be preferred.
587  *    If we have more than one (probably) reachable router, prefer ones
588  *    with the highest router preference.
589  * 2) When no routers on the list are known to be reachable or
590  *    probably reachable, routers SHOULD be selected in a round-robin
591  *    fashion, regardless of router preference values.
592  * 3) If the Default Router List is empty, assume that all
593  *    destinations are on-link.
594  *
595  * We assume nd_defrouter is sorted by router preference value.
596  * Since the code below covers both with and without router preference cases,
597  * we do not need to classify the cases by ifdef.
598  *
599  * At this moment, we do not try to install more than one default router,
600  * even when the multipath routing is available, because we're not sure about
601  * the benefits for stub hosts comparing to the risk of making the code
602  * complicated and the possibility of introducing bugs.
603  */
604 void
605 defrouter_select(void)
606 {
607 	int s = splnet();
608 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
609 	struct rtentry *rt = NULL;
610 	struct llinfo_nd6 *ln = NULL;
611 
612 	/*
613 	 * This function should be called only when acting as an autoconfigured
614 	 * host.  Although the remaining part of this function is not effective
615 	 * if the node is not an autoconfigured host, we explicitly exclude
616 	 * such cases here for safety.
617 	 */
618 	if (ip6_forwarding || !ip6_accept_rtadv) {
619 		nd6log((LOG_WARNING,
620 		    "defrouter_select: called unexpectedly (forwarding=%d, "
621 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
622 		splx(s);
623 		return;
624 	}
625 
626 	/*
627 	 * Let's handle easy case (3) first:
628 	 * If default router list is empty, there's nothing to be done.
629 	 */
630 	if (!TAILQ_FIRST(&nd_defrouter)) {
631 		splx(s);
632 		return;
633 	}
634 
635 	/*
636 	 * Search for a (probably) reachable router from the list.
637 	 * We just pick up the first reachable one (if any), assuming that
638 	 * the ordering rule of the list described in defrtrlist_update().
639 	 */
640 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
641 	     dr = TAILQ_NEXT(dr, dr_entry)) {
642 		if (selected_dr == NULL &&
643 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
644 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
645 		    ND6_IS_LLINFO_PROBREACH(ln)) {
646 			selected_dr = dr;
647 		}
648 
649 		if (dr->installed && installed_dr == NULL)
650 			installed_dr = dr;
651 		else if (dr->installed && installed_dr) {
652 			/* this should not happen.  warn for diagnosis. */
653 			log(LOG_ERR, "defrouter_select: more than one router"
654 			    " is installed\n");
655 		}
656 	}
657 	/*
658 	 * If none of the default routers was found to be reachable,
659 	 * round-robin the list regardless of preference.
660 	 * Otherwise, if we have an installed router, check if the selected
661 	 * (reachable) router should really be preferred to the installed one.
662 	 * We only prefer the new router when the old one is not reachable
663 	 * or when the new one has a really higher preference value.
664 	 */
665 	if (selected_dr == NULL) {
666 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
667 			selected_dr = TAILQ_FIRST(&nd_defrouter);
668 		else
669 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
670 	} else if (installed_dr &&
671 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
672 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
673 	    ND6_IS_LLINFO_PROBREACH(ln) &&
674 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
675 		selected_dr = installed_dr;
676 	}
677 
678 	/*
679 	 * If the selected router is different than the installed one,
680 	 * remove the installed router and install the selected one.
681 	 * Note that the selected router is never NULL here.
682 	 */
683 	if (installed_dr != selected_dr) {
684 		if (installed_dr)
685 			defrouter_delreq(installed_dr);
686 		defrouter_addreq(selected_dr);
687 	}
688 
689 	splx(s);
690 	return;
691 }
692 
693 /*
694  * for default router selection
695  * regards router-preference field as a 2-bit signed integer
696  */
697 static int
698 rtpref(struct nd_defrouter *dr)
699 {
700 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
701 	case ND_RA_FLAG_RTPREF_HIGH:
702 		return (RTPREF_HIGH);
703 	case ND_RA_FLAG_RTPREF_MEDIUM:
704 	case ND_RA_FLAG_RTPREF_RSV:
705 		return (RTPREF_MEDIUM);
706 	case ND_RA_FLAG_RTPREF_LOW:
707 		return (RTPREF_LOW);
708 	default:
709 		/*
710 		 * This case should never happen.  If it did, it would mean a
711 		 * serious bug of kernel internal.  We thus always bark here.
712 		 * Or, can we even panic?
713 		 */
714 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
715 		return (RTPREF_INVALID);
716 	}
717 	/* NOTREACHED */
718 }
719 
720 static struct nd_defrouter *
721 defrtrlist_update(struct nd_defrouter *new)
722 {
723 	struct nd_defrouter *dr, *n;
724 	int s = splnet();
725 
726 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
727 		/* entry exists */
728 		if (new->rtlifetime == 0) {
729 			defrtrlist_del(dr);
730 			dr = NULL;
731 		} else {
732 			int oldpref = rtpref(dr);
733 
734 			/* override */
735 			dr->flags = new->flags; /* xxx flag check */
736 			dr->rtlifetime = new->rtlifetime;
737 			dr->expire = new->expire;
738 
739 			/*
740 			 * If the preference does not change, there's no need
741 			 * to sort the entries.
742 			 */
743 			if (rtpref(new) == oldpref) {
744 				splx(s);
745 				return (dr);
746 			}
747 
748 			/*
749 			 * preferred router may be changed, so relocate
750 			 * this router.
751 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
752 			 * However, since defrtrlist_del() has many side
753 			 * effects, we intentionally do so here.
754 			 * defrouter_select() below will handle routing
755 			 * changes later.
756 			 */
757 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
758 			n = dr;
759 			goto insert;
760 		}
761 		splx(s);
762 		return (dr);
763 	}
764 
765 	/* entry does not exist */
766 	if (new->rtlifetime == 0) {
767 		splx(s);
768 		return (NULL);
769 	}
770 
771 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
772 	if (n == NULL) {
773 		splx(s);
774 		return (NULL);
775 	}
776 	bzero(n, sizeof(*n));
777 	*n = *new;
778 
779 insert:
780 	/*
781 	 * Insert the new router in the Default Router List;
782 	 * The Default Router List should be in the descending order
783 	 * of router-preferece.  Routers with the same preference are
784 	 * sorted in the arriving time order.
785 	 */
786 
787 	/* insert at the end of the group */
788 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
789 	     dr = TAILQ_NEXT(dr, dr_entry)) {
790 		if (rtpref(n) > rtpref(dr))
791 			break;
792 	}
793 	if (dr)
794 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
795 	else
796 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
797 
798 	defrouter_select();
799 
800 	splx(s);
801 
802 	return (n);
803 }
804 
805 static struct nd_pfxrouter *
806 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
807 {
808 	struct nd_pfxrouter *search;
809 
810 	for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
811 		if (search->router == dr)
812 			break;
813 	}
814 
815 	return (search);
816 }
817 
818 static void
819 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
820 {
821 	struct nd_pfxrouter *new;
822 
823 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
824 	if (new == NULL)
825 		return;
826 	bzero(new, sizeof(*new));
827 	new->router = dr;
828 
829 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
830 
831 	pfxlist_onlink_check();
832 }
833 
834 static void
835 pfxrtr_del(struct nd_pfxrouter *pfr)
836 {
837 	LIST_REMOVE(pfr, pfr_entry);
838 	free(pfr, M_IP6NDP);
839 }
840 
841 struct nd_prefix *
842 nd6_prefix_lookup(struct nd_prefixctl *key)
843 {
844 	struct nd_prefix *search;
845 
846 	for (search = nd_prefix.lh_first; search; search = search->ndpr_next) {
847 		if (key->ndpr_ifp == search->ndpr_ifp &&
848 		    key->ndpr_plen == search->ndpr_plen &&
849 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
850 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
851 			break;
852 		}
853 	}
854 
855 	return (search);
856 }
857 
858 int
859 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
860     struct nd_prefix **newp)
861 {
862 	struct nd_prefix *new = NULL;
863 	int error = 0;
864 	int i, s;
865 	char ip6buf[INET6_ADDRSTRLEN];
866 
867 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
868 	if (new == NULL)
869 		return(ENOMEM);
870 	bzero(new, sizeof(*new));
871 	new->ndpr_ifp = pr->ndpr_ifp;
872 	new->ndpr_prefix = pr->ndpr_prefix;
873 	new->ndpr_plen = pr->ndpr_plen;
874 	new->ndpr_vltime = pr->ndpr_vltime;
875 	new->ndpr_pltime = pr->ndpr_pltime;
876 	new->ndpr_flags = pr->ndpr_flags;
877 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
878 		free(new, M_IP6NDP);
879 		return(error);
880 	}
881 	new->ndpr_lastupdate = time_second;
882 	if (newp != NULL)
883 		*newp = new;
884 
885 	/* initialization */
886 	LIST_INIT(&new->ndpr_advrtrs);
887 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
888 	/* make prefix in the canonical form */
889 	for (i = 0; i < 4; i++)
890 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
891 		    new->ndpr_mask.s6_addr32[i];
892 
893 	s = splnet();
894 	/* link ndpr_entry to nd_prefix list */
895 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
896 	splx(s);
897 
898 	/* ND_OPT_PI_FLAG_ONLINK processing */
899 	if (new->ndpr_raf_onlink) {
900 		int e;
901 
902 		if ((e = nd6_prefix_onlink(new)) != 0) {
903 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
904 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
905 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
906 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
907 			/* proceed anyway. XXX: is it correct? */
908 		}
909 	}
910 
911 	if (dr)
912 		pfxrtr_add(new, dr);
913 
914 	return 0;
915 }
916 
917 void
918 prelist_remove(struct nd_prefix *pr)
919 {
920 	struct nd_pfxrouter *pfr, *next;
921 	int e, s;
922 	char ip6buf[INET6_ADDRSTRLEN];
923 
924 	/* make sure to invalidate the prefix until it is really freed. */
925 	pr->ndpr_vltime = 0;
926 	pr->ndpr_pltime = 0;
927 
928 	/*
929 	 * Though these flags are now meaningless, we'd rather keep the value
930 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
931 	 * when executing "ndp -p".
932 	 */
933 
934 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
935 	    (e = nd6_prefix_offlink(pr)) != 0) {
936 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
937 		    "on %s, errno=%d\n",
938 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
939 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
940 		/* what should we do? */
941 	}
942 
943 	if (pr->ndpr_refcnt > 0)
944 		return;		/* notice here? */
945 
946 	s = splnet();
947 
948 	/* unlink ndpr_entry from nd_prefix list */
949 	LIST_REMOVE(pr, ndpr_entry);
950 
951 	/* free list of routers that adversed the prefix */
952 	for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
953 		next = pfr->pfr_next;
954 
955 		free(pfr, M_IP6NDP);
956 	}
957 	splx(s);
958 
959 	free(pr, M_IP6NDP);
960 
961 	pfxlist_onlink_check();
962 }
963 
964 /*
965  * dr - may be NULL
966  */
967 
968 static int
969 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
970     struct mbuf *m, int mcast)
971 {
972 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
973 	struct ifaddr *ifa;
974 	struct ifnet *ifp = new->ndpr_ifp;
975 	struct nd_prefix *pr;
976 	int s = splnet();
977 	int error = 0;
978 	int newprefix = 0;
979 	int auth;
980 	struct in6_addrlifetime lt6_tmp;
981 	char ip6buf[INET6_ADDRSTRLEN];
982 
983 	auth = 0;
984 	if (m) {
985 		/*
986 		 * Authenticity for NA consists authentication for
987 		 * both IP header and IP datagrams, doesn't it ?
988 		 */
989 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
990 		auth = ((m->m_flags & M_AUTHIPHDR) &&
991 		    (m->m_flags & M_AUTHIPDGM));
992 #endif
993 	}
994 
995 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
996 		/*
997 		 * nd6_prefix_lookup() ensures that pr and new have the same
998 		 * prefix on a same interface.
999 		 */
1000 
1001 		/*
1002 		 * Update prefix information.  Note that the on-link (L) bit
1003 		 * and the autonomous (A) bit should NOT be changed from 1
1004 		 * to 0.
1005 		 */
1006 		if (new->ndpr_raf_onlink == 1)
1007 			pr->ndpr_raf_onlink = 1;
1008 		if (new->ndpr_raf_auto == 1)
1009 			pr->ndpr_raf_auto = 1;
1010 		if (new->ndpr_raf_onlink) {
1011 			pr->ndpr_vltime = new->ndpr_vltime;
1012 			pr->ndpr_pltime = new->ndpr_pltime;
1013 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1014 			pr->ndpr_lastupdate = time_second;
1015 		}
1016 
1017 		if (new->ndpr_raf_onlink &&
1018 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1019 			int e;
1020 
1021 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1022 				nd6log((LOG_ERR,
1023 				    "prelist_update: failed to make "
1024 				    "the prefix %s/%d on-link on %s "
1025 				    "(errno=%d)\n",
1026 				    ip6_sprintf(ip6buf,
1027 					    &pr->ndpr_prefix.sin6_addr),
1028 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1029 				/* proceed anyway. XXX: is it correct? */
1030 			}
1031 		}
1032 
1033 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1034 			pfxrtr_add(pr, dr);
1035 	} else {
1036 		struct nd_prefix *newpr = NULL;
1037 
1038 		newprefix = 1;
1039 
1040 		if (new->ndpr_vltime == 0)
1041 			goto end;
1042 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1043 			goto end;
1044 
1045 		error = nd6_prelist_add(new, dr, &newpr);
1046 		if (error != 0 || newpr == NULL) {
1047 			nd6log((LOG_NOTICE, "prelist_update: "
1048 			    "nd6_prelist_add failed for %s/%d on %s "
1049 			    "errno=%d, returnpr=%p\n",
1050 			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1051 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1052 			    error, newpr));
1053 			goto end; /* we should just give up in this case. */
1054 		}
1055 
1056 		/*
1057 		 * XXX: from the ND point of view, we can ignore a prefix
1058 		 * with the on-link bit being zero.  However, we need a
1059 		 * prefix structure for references from autoconfigured
1060 		 * addresses.  Thus, we explicitly make sure that the prefix
1061 		 * itself expires now.
1062 		 */
1063 		if (newpr->ndpr_raf_onlink == 0) {
1064 			newpr->ndpr_vltime = 0;
1065 			newpr->ndpr_pltime = 0;
1066 			in6_init_prefix_ltimes(newpr);
1067 		}
1068 
1069 		pr = newpr;
1070 	}
1071 
1072 	/*
1073 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1074 	 * Note that pr must be non NULL at this point.
1075 	 */
1076 
1077 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1078 	if (!new->ndpr_raf_auto)
1079 		goto end;
1080 
1081 	/*
1082 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1083 	 * nd6_ra_input.
1084 	 */
1085 
1086 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1087 	if (new->ndpr_pltime > new->ndpr_vltime) {
1088 		error = EINVAL;	/* XXX: won't be used */
1089 		goto end;
1090 	}
1091 
1092 	/*
1093 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1094 	 * an address configured by stateless autoconfiguration already in the
1095 	 * list of addresses associated with the interface, and the Valid
1096 	 * Lifetime is not 0, form an address.  We first check if we have
1097 	 * a matching prefix.
1098 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1099 	 * consider autoconfigured addresses while RFC2462 simply said
1100 	 * "address".
1101 	 */
1102 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1103 		struct in6_ifaddr *ifa6;
1104 		u_int32_t remaininglifetime;
1105 
1106 		if (ifa->ifa_addr->sa_family != AF_INET6)
1107 			continue;
1108 
1109 		ifa6 = (struct in6_ifaddr *)ifa;
1110 
1111 		/*
1112 		 * We only consider autoconfigured addresses as per rfc2462bis.
1113 		 */
1114 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1115 			continue;
1116 
1117 		/*
1118 		 * Spec is not clear here, but I believe we should concentrate
1119 		 * on unicast (i.e. not anycast) addresses.
1120 		 * XXX: other ia6_flags? detached or duplicated?
1121 		 */
1122 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1123 			continue;
1124 
1125 		/*
1126 		 * Ignore the address if it is not associated with a prefix
1127 		 * or is associated with a prefix that is different from this
1128 		 * one.  (pr is never NULL here)
1129 		 */
1130 		if (ifa6->ia6_ndpr != pr)
1131 			continue;
1132 
1133 		if (ia6_match == NULL) /* remember the first one */
1134 			ia6_match = ifa6;
1135 
1136 		/*
1137 		 * An already autoconfigured address matched.  Now that we
1138 		 * are sure there is at least one matched address, we can
1139 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1140 		 * "two hours" rule and the privacy extension.
1141 		 * We apply some clarifications in rfc2462bis:
1142 		 * - use remaininglifetime instead of storedlifetime as a
1143 		 *   variable name
1144 		 * - remove the dead code in the "two-hour" rule
1145 		 */
1146 #define TWOHOUR		(120*60)
1147 		lt6_tmp = ifa6->ia6_lifetime;
1148 
1149 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1150 			remaininglifetime = ND6_INFINITE_LIFETIME;
1151 		else if (time_second - ifa6->ia6_updatetime >
1152 			 lt6_tmp.ia6t_vltime) {
1153 			/*
1154 			 * The case of "invalid" address.  We should usually
1155 			 * not see this case.
1156 			 */
1157 			remaininglifetime = 0;
1158 		} else
1159 			remaininglifetime = lt6_tmp.ia6t_vltime -
1160 			    (time_second - ifa6->ia6_updatetime);
1161 
1162 		/* when not updating, keep the current stored lifetime. */
1163 		lt6_tmp.ia6t_vltime = remaininglifetime;
1164 
1165 		if (TWOHOUR < new->ndpr_vltime ||
1166 		    remaininglifetime < new->ndpr_vltime) {
1167 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1168 		} else if (remaininglifetime <= TWOHOUR) {
1169 			if (auth) {
1170 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1171 			}
1172 		} else {
1173 			/*
1174 			 * new->ndpr_vltime <= TWOHOUR &&
1175 			 * TWOHOUR < remaininglifetime
1176 			 */
1177 			lt6_tmp.ia6t_vltime = TWOHOUR;
1178 		}
1179 
1180 		/* The 2 hour rule is not imposed for preferred lifetime. */
1181 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1182 
1183 		in6_init_address_ltimes(pr, &lt6_tmp);
1184 
1185 		/*
1186 		 * We need to treat lifetimes for temporary addresses
1187 		 * differently, according to
1188 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1189 		 * we only update the lifetimes when they are in the maximum
1190 		 * intervals.
1191 		 */
1192 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1193 			u_int32_t maxvltime, maxpltime;
1194 
1195 			if (ip6_temp_valid_lifetime >
1196 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1197 			    ip6_desync_factor)) {
1198 				maxvltime = ip6_temp_valid_lifetime -
1199 				    (time_second - ifa6->ia6_createtime) -
1200 				    ip6_desync_factor;
1201 			} else
1202 				maxvltime = 0;
1203 			if (ip6_temp_preferred_lifetime >
1204 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1205 			    ip6_desync_factor)) {
1206 				maxpltime = ip6_temp_preferred_lifetime -
1207 				    (time_second - ifa6->ia6_createtime) -
1208 				    ip6_desync_factor;
1209 			} else
1210 				maxpltime = 0;
1211 
1212 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1213 			    lt6_tmp.ia6t_vltime > maxvltime) {
1214 				lt6_tmp.ia6t_vltime = maxvltime;
1215 			}
1216 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1217 			    lt6_tmp.ia6t_pltime > maxpltime) {
1218 				lt6_tmp.ia6t_pltime = maxpltime;
1219 			}
1220 		}
1221 		ifa6->ia6_lifetime = lt6_tmp;
1222 		ifa6->ia6_updatetime = time_second;
1223 	}
1224 	if (ia6_match == NULL && new->ndpr_vltime) {
1225 		int ifidlen;
1226 
1227 		/*
1228 		 * 5.5.3 (d) (continued)
1229 		 * No address matched and the valid lifetime is non-zero.
1230 		 * Create a new address.
1231 		 */
1232 
1233 		/*
1234 		 * Prefix Length check:
1235 		 * If the sum of the prefix length and interface identifier
1236 		 * length does not equal 128 bits, the Prefix Information
1237 		 * option MUST be ignored.  The length of the interface
1238 		 * identifier is defined in a separate link-type specific
1239 		 * document.
1240 		 */
1241 		ifidlen = in6_if2idlen(ifp);
1242 		if (ifidlen < 0) {
1243 			/* this should not happen, so we always log it. */
1244 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1245 			    if_name(ifp));
1246 			goto end;
1247 		}
1248 		if (ifidlen + pr->ndpr_plen != 128) {
1249 			nd6log((LOG_INFO,
1250 			    "prelist_update: invalid prefixlen "
1251 			    "%d for %s, ignored\n",
1252 			    pr->ndpr_plen, if_name(ifp)));
1253 			goto end;
1254 		}
1255 
1256 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1257 			/*
1258 			 * note that we should use pr (not new) for reference.
1259 			 */
1260 			pr->ndpr_refcnt++;
1261 			ia6->ia6_ndpr = pr;
1262 
1263 			/*
1264 			 * RFC 3041 3.3 (2).
1265 			 * When a new public address is created as described
1266 			 * in RFC2462, also create a new temporary address.
1267 			 *
1268 			 * RFC 3041 3.5.
1269 			 * When an interface connects to a new link, a new
1270 			 * randomized interface identifier should be generated
1271 			 * immediately together with a new set of temporary
1272 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1273 			 * in6_tmpifadd().
1274 			 */
1275 			if (ip6_use_tempaddr) {
1276 				int e;
1277 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1278 					nd6log((LOG_NOTICE, "prelist_update: "
1279 					    "failed to create a temporary "
1280 					    "address, errno=%d\n",
1281 					    e));
1282 				}
1283 			}
1284 
1285 			/*
1286 			 * A newly added address might affect the status
1287 			 * of other addresses, so we check and update it.
1288 			 * XXX: what if address duplication happens?
1289 			 */
1290 			pfxlist_onlink_check();
1291 		} else {
1292 			/* just set an error. do not bark here. */
1293 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1294 		}
1295 	}
1296 
1297  end:
1298 	splx(s);
1299 	return error;
1300 }
1301 
1302 /*
1303  * A supplement function used in the on-link detection below;
1304  * detect if a given prefix has a (probably) reachable advertising router.
1305  * XXX: lengthy function name...
1306  */
1307 static struct nd_pfxrouter *
1308 find_pfxlist_reachable_router(struct nd_prefix *pr)
1309 {
1310 	struct nd_pfxrouter *pfxrtr;
1311 	struct rtentry *rt;
1312 	struct llinfo_nd6 *ln;
1313 
1314 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1315 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1316 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1317 		    pfxrtr->router->ifp)) &&
1318 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1319 		    ND6_IS_LLINFO_PROBREACH(ln))
1320 			break;	/* found */
1321 	}
1322 
1323 	return (pfxrtr);
1324 }
1325 
1326 /*
1327  * Check if each prefix in the prefix list has at least one available router
1328  * that advertised the prefix (a router is "available" if its neighbor cache
1329  * entry is reachable or probably reachable).
1330  * If the check fails, the prefix may be off-link, because, for example,
1331  * we have moved from the network but the lifetime of the prefix has not
1332  * expired yet.  So we should not use the prefix if there is another prefix
1333  * that has an available router.
1334  * But, if there is no prefix that has an available router, we still regards
1335  * all the prefixes as on-link.  This is because we can't tell if all the
1336  * routers are simply dead or if we really moved from the network and there
1337  * is no router around us.
1338  */
1339 void
1340 pfxlist_onlink_check()
1341 {
1342 	struct nd_prefix *pr;
1343 	struct in6_ifaddr *ifa;
1344 	struct nd_defrouter *dr;
1345 	struct nd_pfxrouter *pfxrtr = NULL;
1346 
1347 	/*
1348 	 * Check if there is a prefix that has a reachable advertising
1349 	 * router.
1350 	 */
1351 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1352 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1353 			break;
1354 	}
1355 
1356 	/*
1357 	 * If we have no such prefix, check whether we still have a router
1358 	 * that does not advertise any prefixes.
1359 	 */
1360 	if (pr == NULL) {
1361 		for (dr = TAILQ_FIRST(&nd_defrouter); dr;
1362 		    dr = TAILQ_NEXT(dr, dr_entry)) {
1363 			struct nd_prefix *pr0;
1364 
1365 			for (pr0 = nd_prefix.lh_first; pr0;
1366 			    pr0 = pr0->ndpr_next) {
1367 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1368 					break;
1369 			}
1370 			if (pfxrtr != NULL)
1371 				break;
1372 		}
1373 	}
1374 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && pfxrtr == NULL)) {
1375 		/*
1376 		 * There is at least one prefix that has a reachable router,
1377 		 * or at least a router which probably does not advertise
1378 		 * any prefixes.  The latter would be the case when we move
1379 		 * to a new link where we have a router that does not provide
1380 		 * prefixes and we configure an address by hand.
1381 		 * Detach prefixes which have no reachable advertising
1382 		 * router, and attach other prefixes.
1383 		 */
1384 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1385 			/* XXX: a link-local prefix should never be detached */
1386 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1387 				continue;
1388 
1389 			/*
1390 			 * we aren't interested in prefixes without the L bit
1391 			 * set.
1392 			 */
1393 			if (pr->ndpr_raf_onlink == 0)
1394 				continue;
1395 
1396 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1397 			    find_pfxlist_reachable_router(pr) == NULL)
1398 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1399 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1400 			    find_pfxlist_reachable_router(pr) != 0)
1401 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1402 		}
1403 	} else {
1404 		/* there is no prefix that has a reachable router */
1405 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1406 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1407 				continue;
1408 
1409 			if (pr->ndpr_raf_onlink == 0)
1410 				continue;
1411 
1412 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1413 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1414 		}
1415 	}
1416 
1417 	/*
1418 	 * Remove each interface route associated with a (just) detached
1419 	 * prefix, and reinstall the interface route for a (just) attached
1420 	 * prefix.  Note that all attempt of reinstallation does not
1421 	 * necessarily success, when a same prefix is shared among multiple
1422 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1423 	 * so we don't have to care about them.
1424 	 */
1425 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1426 		int e;
1427 		char ip6buf[INET6_ADDRSTRLEN];
1428 
1429 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1430 			continue;
1431 
1432 		if (pr->ndpr_raf_onlink == 0)
1433 			continue;
1434 
1435 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1436 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1437 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1438 				nd6log((LOG_ERR,
1439 				    "pfxlist_onlink_check: failed to "
1440 				    "make %s/%d offlink, errno=%d\n",
1441 				    ip6_sprintf(ip6buf,
1442 					    &pr->ndpr_prefix.sin6_addr),
1443 					    pr->ndpr_plen, e));
1444 			}
1445 		}
1446 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1447 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1448 		    pr->ndpr_raf_onlink) {
1449 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1450 				nd6log((LOG_ERR,
1451 				    "pfxlist_onlink_check: failed to "
1452 				    "make %s/%d onlink, errno=%d\n",
1453 				    ip6_sprintf(ip6buf,
1454 					    &pr->ndpr_prefix.sin6_addr),
1455 					    pr->ndpr_plen, e));
1456 			}
1457 		}
1458 	}
1459 
1460 	/*
1461 	 * Changes on the prefix status might affect address status as well.
1462 	 * Make sure that all addresses derived from an attached prefix are
1463 	 * attached, and that all addresses derived from a detached prefix are
1464 	 * detached.  Note, however, that a manually configured address should
1465 	 * always be attached.
1466 	 * The precise detection logic is same as the one for prefixes.
1467 	 */
1468 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1469 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1470 			continue;
1471 
1472 		if (ifa->ia6_ndpr == NULL) {
1473 			/*
1474 			 * This can happen when we first configure the address
1475 			 * (i.e. the address exists, but the prefix does not).
1476 			 * XXX: complicated relationships...
1477 			 */
1478 			continue;
1479 		}
1480 
1481 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1482 			break;
1483 	}
1484 	if (ifa) {
1485 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1486 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1487 				continue;
1488 
1489 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1490 				continue;
1491 
1492 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1493 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1494 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1495 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1496 					nd6_dad_start((struct ifaddr *)ifa, 0);
1497 				}
1498 			} else {
1499 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1500 			}
1501 		}
1502 	}
1503 	else {
1504 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1505 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1506 				continue;
1507 
1508 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1509 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1510 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1511 				/* Do we need a delay in this case? */
1512 				nd6_dad_start((struct ifaddr *)ifa, 0);
1513 			}
1514 		}
1515 	}
1516 }
1517 
1518 int
1519 nd6_prefix_onlink(struct nd_prefix *pr)
1520 {
1521 	struct ifaddr *ifa;
1522 	struct ifnet *ifp = pr->ndpr_ifp;
1523 	struct sockaddr_in6 mask6;
1524 	struct nd_prefix *opr;
1525 	u_long rtflags;
1526 	int error = 0;
1527 	struct rtentry *rt = NULL;
1528 	char ip6buf[INET6_ADDRSTRLEN];
1529 
1530 	/* sanity check */
1531 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1532 		nd6log((LOG_ERR,
1533 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1534 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1535 		    pr->ndpr_plen));
1536 		return (EEXIST);
1537 	}
1538 
1539 	/*
1540 	 * Add the interface route associated with the prefix.  Before
1541 	 * installing the route, check if there's the same prefix on another
1542 	 * interface, and the prefix has already installed the interface route.
1543 	 * Although such a configuration is expected to be rare, we explicitly
1544 	 * allow it.
1545 	 */
1546 	for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1547 		if (opr == pr)
1548 			continue;
1549 
1550 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1551 			continue;
1552 
1553 		if (opr->ndpr_plen == pr->ndpr_plen &&
1554 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1555 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1556 			return (0);
1557 	}
1558 
1559 	/*
1560 	 * We prefer link-local addresses as the associated interface address.
1561 	 */
1562 	/* search for a link-local addr */
1563 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1564 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1565 	if (ifa == NULL) {
1566 		/* XXX: freebsd does not have ifa_ifwithaf */
1567 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1568 			if (ifa->ifa_addr->sa_family == AF_INET6)
1569 				break;
1570 		}
1571 		/* should we care about ia6_flags? */
1572 	}
1573 	if (ifa == NULL) {
1574 		/*
1575 		 * This can still happen, when, for example, we receive an RA
1576 		 * containing a prefix with the L bit set and the A bit clear,
1577 		 * after removing all IPv6 addresses on the receiving
1578 		 * interface.  This should, of course, be rare though.
1579 		 */
1580 		nd6log((LOG_NOTICE,
1581 		    "nd6_prefix_onlink: failed to find any ifaddr"
1582 		    " to add route for a prefix(%s/%d) on %s\n",
1583 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1584 		    pr->ndpr_plen, if_name(ifp)));
1585 		return (0);
1586 	}
1587 
1588 	/*
1589 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1590 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1591 	 */
1592 	bzero(&mask6, sizeof(mask6));
1593 	mask6.sin6_len = sizeof(mask6);
1594 	mask6.sin6_addr = pr->ndpr_mask;
1595 	rtflags = ifa->ifa_flags | RTF_CLONING | RTF_UP;
1596 	if (nd6_need_cache(ifp)) {
1597 		/* explicitly set in case ifa_flags does not set the flag. */
1598 		rtflags |= RTF_CLONING;
1599 	} else {
1600 		/*
1601 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1602 		 */
1603 		rtflags &= ~RTF_CLONING;
1604 	}
1605 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1606 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1607 	if (error == 0) {
1608 		if (rt != NULL) /* this should be non NULL, though */
1609 			nd6_rtmsg(RTM_ADD, rt);
1610 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1611 	} else {
1612 		char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
1613 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1614 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1615 		    "errno = %d\n",
1616 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1617 		    pr->ndpr_plen, if_name(ifp),
1618 		    ip6_sprintf(ip6bufg, &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1619 		    ip6_sprintf(ip6bufm, &mask6.sin6_addr), rtflags, error));
1620 	}
1621 
1622 	if (rt != NULL) {
1623 		RT_LOCK(rt);
1624 		RT_REMREF(rt);
1625 		RT_UNLOCK(rt);
1626 	}
1627 
1628 	return (error);
1629 }
1630 
1631 int
1632 nd6_prefix_offlink(struct nd_prefix *pr)
1633 {
1634 	int error = 0;
1635 	struct ifnet *ifp = pr->ndpr_ifp;
1636 	struct nd_prefix *opr;
1637 	struct sockaddr_in6 sa6, mask6;
1638 	struct rtentry *rt = NULL;
1639 	char ip6buf[INET6_ADDRSTRLEN];
1640 
1641 	/* sanity check */
1642 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1643 		nd6log((LOG_ERR,
1644 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1645 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1646 		    pr->ndpr_plen));
1647 		return (EEXIST);
1648 	}
1649 
1650 	bzero(&sa6, sizeof(sa6));
1651 	sa6.sin6_family = AF_INET6;
1652 	sa6.sin6_len = sizeof(sa6);
1653 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1654 	    sizeof(struct in6_addr));
1655 	bzero(&mask6, sizeof(mask6));
1656 	mask6.sin6_family = AF_INET6;
1657 	mask6.sin6_len = sizeof(sa6);
1658 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1659 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1660 	    (struct sockaddr *)&mask6, 0, &rt);
1661 	if (error == 0) {
1662 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1663 
1664 		/* report the route deletion to the routing socket. */
1665 		if (rt != NULL)
1666 			nd6_rtmsg(RTM_DELETE, rt);
1667 
1668 		/*
1669 		 * There might be the same prefix on another interface,
1670 		 * the prefix which could not be on-link just because we have
1671 		 * the interface route (see comments in nd6_prefix_onlink).
1672 		 * If there's one, try to make the prefix on-link on the
1673 		 * interface.
1674 		 */
1675 		for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1676 			if (opr == pr)
1677 				continue;
1678 
1679 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1680 				continue;
1681 
1682 			/*
1683 			 * KAME specific: detached prefixes should not be
1684 			 * on-link.
1685 			 */
1686 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1687 				continue;
1688 
1689 			if (opr->ndpr_plen == pr->ndpr_plen &&
1690 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1691 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1692 				int e;
1693 
1694 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1695 					nd6log((LOG_ERR,
1696 					    "nd6_prefix_offlink: failed to "
1697 					    "recover a prefix %s/%d from %s "
1698 					    "to %s (errno = %d)\n",
1699 					    ip6_sprintf(ip6buf,
1700 						&opr->ndpr_prefix.sin6_addr),
1701 					    opr->ndpr_plen, if_name(ifp),
1702 					    if_name(opr->ndpr_ifp), e));
1703 				}
1704 			}
1705 		}
1706 	} else {
1707 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1708 		nd6log((LOG_ERR,
1709 		    "nd6_prefix_offlink: failed to delete route: "
1710 		    "%s/%d on %s (errno = %d)\n",
1711 		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1712 		    if_name(ifp), error));
1713 	}
1714 
1715 	if (rt != NULL) {
1716 		RTFREE(rt);
1717 	}
1718 
1719 	return (error);
1720 }
1721 
1722 static struct in6_ifaddr *
1723 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1724 {
1725 	struct ifnet *ifp = pr->ndpr_ifp;
1726 	struct ifaddr *ifa;
1727 	struct in6_aliasreq ifra;
1728 	struct in6_ifaddr *ia, *ib;
1729 	int error, plen0;
1730 	struct in6_addr mask;
1731 	int prefixlen = pr->ndpr_plen;
1732 	int updateflags;
1733 	char ip6buf[INET6_ADDRSTRLEN];
1734 
1735 	in6_prefixlen2mask(&mask, prefixlen);
1736 
1737 	/*
1738 	 * find a link-local address (will be interface ID).
1739 	 * Is it really mandatory? Theoretically, a global or a site-local
1740 	 * address can be configured without a link-local address, if we
1741 	 * have a unique interface identifier...
1742 	 *
1743 	 * it is not mandatory to have a link-local address, we can generate
1744 	 * interface identifier on the fly.  we do this because:
1745 	 * (1) it should be the easiest way to find interface identifier.
1746 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1747 	 * for multiple addresses on a single interface, and possible shortcut
1748 	 * of DAD.  we omitted DAD for this reason in the past.
1749 	 * (3) a user can prevent autoconfiguration of global address
1750 	 * by removing link-local address by hand (this is partly because we
1751 	 * don't have other way to control the use of IPv6 on an interface.
1752 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1753 	 * (4) it is easier to manage when an interface has addresses
1754 	 * with the same interface identifier, than to have multiple addresses
1755 	 * with different interface identifiers.
1756 	 */
1757 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1758 	if (ifa)
1759 		ib = (struct in6_ifaddr *)ifa;
1760 	else
1761 		return NULL;
1762 
1763 	/* prefixlen + ifidlen must be equal to 128 */
1764 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1765 	if (prefixlen != plen0) {
1766 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1767 		    "(prefix=%d ifid=%d)\n",
1768 		    if_name(ifp), prefixlen, 128 - plen0));
1769 		return NULL;
1770 	}
1771 
1772 	/* make ifaddr */
1773 
1774 	bzero(&ifra, sizeof(ifra));
1775 	/*
1776 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1777 	 * for safety.
1778 	 */
1779 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1780 	ifra.ifra_addr.sin6_family = AF_INET6;
1781 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1782 	/* prefix */
1783 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1784 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1785 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1786 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1787 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1788 
1789 	/* interface ID */
1790 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1791 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1792 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1793 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1794 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1795 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1796 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1797 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1798 
1799 	/* new prefix mask. */
1800 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1801 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1802 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1803 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1804 
1805 	/* lifetimes. */
1806 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1807 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1808 
1809 	/* XXX: scope zone ID? */
1810 
1811 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1812 
1813 	/*
1814 	 * Make sure that we do not have this address already.  This should
1815 	 * usually not happen, but we can still see this case, e.g., if we
1816 	 * have manually configured the exact address to be configured.
1817 	 */
1818 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1819 		/* this should be rare enough to make an explicit log */
1820 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1821 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1822 		return (NULL);
1823 	}
1824 
1825 	/*
1826 	 * Allocate ifaddr structure, link into chain, etc.
1827 	 * If we are going to create a new address upon receiving a multicasted
1828 	 * RA, we need to impose a random delay before starting DAD.
1829 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1830 	 */
1831 	updateflags = 0;
1832 	if (mcast)
1833 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1834 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1835 		nd6log((LOG_ERR,
1836 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1837 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1838 		    if_name(ifp), error));
1839 		return (NULL);	/* ifaddr must not have been allocated. */
1840 	}
1841 
1842 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1843 
1844 	return (ia);		/* this is always non-NULL */
1845 }
1846 
1847 /*
1848  * ia0 - corresponding public address
1849  */
1850 int
1851 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1852 {
1853 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1854 	struct in6_ifaddr *newia, *ia;
1855 	struct in6_aliasreq ifra;
1856 	int i, error;
1857 	int trylimit = 3;	/* XXX: adhoc value */
1858 	int updateflags;
1859 	u_int32_t randid[2];
1860 	time_t vltime0, pltime0;
1861 
1862 	bzero(&ifra, sizeof(ifra));
1863 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1864 	ifra.ifra_addr = ia0->ia_addr;
1865 	/* copy prefix mask */
1866 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1867 	/* clear the old IFID */
1868 	for (i = 0; i < 4; i++) {
1869 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1870 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1871 	}
1872 
1873   again:
1874 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1875 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1876 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1877 		    "random IFID\n"));
1878 		return (EINVAL);
1879 	}
1880 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1881 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1882 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1883 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1884 
1885 	/*
1886 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1887 	 * However, we may still have a chance to see collision, because
1888 	 * there may be a time lag between generation of the ID and generation
1889 	 * of the address.  So, we'll do one more sanity check.
1890 	 */
1891 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1892 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1893 		    &ifra.ifra_addr.sin6_addr)) {
1894 			if (trylimit-- == 0) {
1895 				/*
1896 				 * Give up.  Something strange should have
1897 				 * happened.
1898 				 */
1899 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1900 				    "find a unique random IFID\n"));
1901 				return (EEXIST);
1902 			}
1903 			forcegen = 1;
1904 			goto again;
1905 		}
1906 	}
1907 
1908 	/*
1909 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1910          * public address or TEMP_VALID_LIFETIME.
1911 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1912          * of the public address or TEMP_PREFERRED_LIFETIME -
1913          * DESYNC_FACTOR.
1914 	 */
1915 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1916 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1917 		    (ia0->ia6_lifetime.ia6t_vltime -
1918 		    (time_second - ia0->ia6_updatetime));
1919 		if (vltime0 > ip6_temp_valid_lifetime)
1920 			vltime0 = ip6_temp_valid_lifetime;
1921 	} else
1922 		vltime0 = ip6_temp_valid_lifetime;
1923 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1924 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1925 		    (ia0->ia6_lifetime.ia6t_pltime -
1926 		    (time_second - ia0->ia6_updatetime));
1927 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
1928 			pltime0 = ip6_temp_preferred_lifetime -
1929 			    ip6_desync_factor;
1930 		}
1931 	} else
1932 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
1933 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1934 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1935 
1936 	/*
1937 	 * A temporary address is created only if this calculated Preferred
1938 	 * Lifetime is greater than REGEN_ADVANCE time units.
1939 	 */
1940 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
1941 		return (0);
1942 
1943 	/* XXX: scope zone ID? */
1944 
1945 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1946 
1947 	/* allocate ifaddr structure, link into chain, etc. */
1948 	updateflags = 0;
1949 	if (delay)
1950 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1951 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1952 		return (error);
1953 
1954 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1955 	if (newia == NULL) {	/* XXX: can it happen? */
1956 		nd6log((LOG_ERR,
1957 		    "in6_tmpifadd: ifa update succeeded, but we got "
1958 		    "no ifaddr\n"));
1959 		return (EINVAL); /* XXX */
1960 	}
1961 	newia->ia6_ndpr = ia0->ia6_ndpr;
1962 	newia->ia6_ndpr->ndpr_refcnt++;
1963 
1964 	/*
1965 	 * A newly added address might affect the status of other addresses.
1966 	 * XXX: when the temporary address is generated with a new public
1967 	 * address, the onlink check is redundant.  However, it would be safe
1968 	 * to do the check explicitly everywhere a new address is generated,
1969 	 * and, in fact, we surely need the check when we create a new
1970 	 * temporary address due to deprecation of an old temporary address.
1971 	 */
1972 	pfxlist_onlink_check();
1973 
1974 	return (0);
1975 }
1976 
1977 static int
1978 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1979 {
1980 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1981 		ndpr->ndpr_preferred = 0;
1982 	else
1983 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
1984 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1985 		ndpr->ndpr_expire = 0;
1986 	else
1987 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
1988 
1989 	return 0;
1990 }
1991 
1992 static void
1993 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
1994 {
1995 	/* init ia6t_expire */
1996 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
1997 		lt6->ia6t_expire = 0;
1998 	else {
1999 		lt6->ia6t_expire = time_second;
2000 		lt6->ia6t_expire += lt6->ia6t_vltime;
2001 	}
2002 
2003 	/* init ia6t_preferred */
2004 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2005 		lt6->ia6t_preferred = 0;
2006 	else {
2007 		lt6->ia6t_preferred = time_second;
2008 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2009 	}
2010 }
2011 
2012 /*
2013  * Delete all the routing table entries that use the specified gateway.
2014  * XXX: this function causes search through all entries of routing table, so
2015  * it shouldn't be called when acting as a router.
2016  */
2017 void
2018 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2019 {
2020 	struct radix_node_head *rnh = rt_tables[AF_INET6];
2021 	int s = splnet();
2022 
2023 	/* We'll care only link-local addresses */
2024 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2025 		splx(s);
2026 		return;
2027 	}
2028 
2029 	RADIX_NODE_HEAD_LOCK(rnh);
2030 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2031 	RADIX_NODE_HEAD_UNLOCK(rnh);
2032 	splx(s);
2033 }
2034 
2035 static int
2036 rt6_deleteroute(struct radix_node *rn, void *arg)
2037 {
2038 #define SIN6(s)	((struct sockaddr_in6 *)s)
2039 	struct rtentry *rt = (struct rtentry *)rn;
2040 	struct in6_addr *gate = (struct in6_addr *)arg;
2041 
2042 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2043 		return (0);
2044 
2045 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2046 		return (0);
2047 	}
2048 
2049 	/*
2050 	 * Do not delete a static route.
2051 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2052 	 * 'cloned' bit instead?
2053 	 */
2054 	if ((rt->rt_flags & RTF_STATIC) != 0)
2055 		return (0);
2056 
2057 	/*
2058 	 * We delete only host route. This means, in particular, we don't
2059 	 * delete default route.
2060 	 */
2061 	if ((rt->rt_flags & RTF_HOST) == 0)
2062 		return (0);
2063 
2064 	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2065 	    rt_mask(rt), rt->rt_flags, 0));
2066 #undef SIN6
2067 }
2068 
2069 int
2070 nd6_setdefaultiface(int ifindex)
2071 {
2072 	int error = 0;
2073 
2074 	if (ifindex < 0 || if_index < ifindex)
2075 		return (EINVAL);
2076 	if (ifindex != 0 && !ifnet_byindex(ifindex))
2077 		return (EINVAL);
2078 
2079 	if (nd6_defifindex != ifindex) {
2080 		nd6_defifindex = ifindex;
2081 		if (nd6_defifindex > 0)
2082 			nd6_defifp = ifnet_byindex(nd6_defifindex);
2083 		else
2084 			nd6_defifp = NULL;
2085 
2086 		/*
2087 		 * Our current implementation assumes one-to-one maping between
2088 		 * interfaces and links, so it would be natural to use the
2089 		 * default interface as the default link.
2090 		 */
2091 		scope6_setdefault(nd6_defifp);
2092 	}
2093 
2094 	return (error);
2095 }
2096