xref: /netbsd-src/sys/netinet6/nd6_rtr.c (revision b757af438b42b93f8c6571f026d8b8ef3eaf5fc9)
1 /*	$NetBSD: nd6_rtr.c,v 1.82 2011/11/19 22:51:29 tls Exp $	*/
2 /*	$KAME: nd6_rtr.c,v 1.95 2001/02/07 08:09:47 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: nd6_rtr.c,v 1.82 2011/11/19 22:51:29 tls Exp $");
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/malloc.h>
39 #include <sys/mbuf.h>
40 #include <sys/socket.h>
41 #include <sys/sockio.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/errno.h>
45 #include <sys/ioctl.h>
46 #include <sys/syslog.h>
47 #include <sys/cprng.h>
48 
49 #include <net/if.h>
50 #include <net/if_types.h>
51 #include <net/if_dl.h>
52 #include <net/route.h>
53 #include <net/radix.h>
54 
55 #include <netinet/in.h>
56 #include <netinet6/in6_var.h>
57 #include <netinet6/in6_ifattach.h>
58 #include <netinet/ip6.h>
59 #include <netinet6/ip6_var.h>
60 #include <netinet6/nd6.h>
61 #include <netinet/icmp6.h>
62 #include <netinet6/icmp6_private.h>
63 #include <netinet6/scope6_var.h>
64 
65 #include <net/net_osdep.h>
66 
67 static int rtpref(struct nd_defrouter *);
68 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
69 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
70     struct mbuf *, int);
71 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
72 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
73 	struct nd_defrouter *);
74 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
75 static void pfxrtr_del(struct nd_pfxrouter *);
76 static struct nd_pfxrouter *find_pfxlist_reachable_router
77 	(struct nd_prefix *);
78 static void defrouter_delreq(struct nd_defrouter *);
79 static void nd6_rtmsg(int, struct rtentry *);
80 
81 static int in6_init_prefix_ltimes(struct nd_prefix *);
82 static void in6_init_address_ltimes(struct nd_prefix *ndpr,
83 	struct in6_addrlifetime *lt6);
84 
85 static int rt6_deleteroute(struct rtentry *, void *);
86 
87 extern int nd6_recalc_reachtm_interval;
88 
89 static struct ifnet *nd6_defifp;
90 int nd6_defifindex;
91 
92 int ip6_use_tempaddr = 0;
93 
94 int ip6_desync_factor;
95 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
96 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
97 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
98 
99 int nd6_numroutes = 0;
100 
101 /* RTPREF_MEDIUM has to be 0! */
102 #define RTPREF_HIGH	1
103 #define RTPREF_MEDIUM	0
104 #define RTPREF_LOW	(-1)
105 #define RTPREF_RESERVED	(-2)
106 #define RTPREF_INVALID	(-3)	/* internal */
107 
108 /*
109  * Receive Router Solicitation Message - just for routers.
110  * Router solicitation/advertisement is mostly managed by a userland program
111  * (rtadvd) so here we have no function like nd6_ra_output().
112  *
113  * Based on RFC 2461
114  */
115 void
116 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
117 {
118 	struct ifnet *ifp = m->m_pkthdr.rcvif;
119 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
120 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
121 	struct nd_router_solicit *nd_rs;
122 	struct in6_addr saddr6 = ip6->ip6_src;
123 	char *lladdr = NULL;
124 	int lladdrlen = 0;
125 	union nd_opts ndopts;
126 
127 	/* If I'm not a router, ignore it. */
128 	if (nd6_accepts_rtadv(ndi) || !ip6_forwarding)
129 		goto freeit;
130 
131 	/* Sanity checks */
132 	if (ip6->ip6_hlim != 255) {
133 		nd6log((LOG_ERR,
134 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
135 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
136 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
137 		goto bad;
138 	}
139 
140 	/*
141 	 * Don't update the neighbor cache, if src = ::.
142 	 * This indicates that the src has no IP address assigned yet.
143 	 */
144 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
145 		goto freeit;
146 
147 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
148 	if (nd_rs == NULL) {
149 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
150 		return;
151 	}
152 
153 	icmp6len -= sizeof(*nd_rs);
154 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
155 	if (nd6_options(&ndopts) < 0) {
156 		nd6log((LOG_INFO,
157 		    "nd6_rs_input: invalid ND option, ignored\n"));
158 		/* nd6_options have incremented stats */
159 		goto freeit;
160 	}
161 
162 	if (ndopts.nd_opts_src_lladdr) {
163 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
164 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
165 	}
166 
167 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
168 		nd6log((LOG_INFO,
169 		    "nd6_rs_input: lladdrlen mismatch for %s "
170 		    "(if %d, RS packet %d)\n",
171 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
172 		goto bad;
173 	}
174 
175 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
176 
177  freeit:
178 	m_freem(m);
179 	return;
180 
181  bad:
182 	ICMP6_STATINC(ICMP6_STAT_BADRS);
183 	m_freem(m);
184 }
185 
186 /*
187  * Receive Router Advertisement Message.
188  *
189  * Based on RFC 2461
190  * TODO: on-link bit on prefix information
191  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
192  */
193 void
194 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
195 {
196 	struct ifnet *ifp = m->m_pkthdr.rcvif;
197 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
198 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
199 	struct nd_router_advert *nd_ra;
200 	struct in6_addr saddr6 = ip6->ip6_src;
201 #if 0
202 	struct in6_addr daddr6 = ip6->ip6_dst;
203 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
204 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
205 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
206 #endif
207 	int mcast = 0;
208 	union nd_opts ndopts;
209 	struct nd_defrouter *dr;
210 
211 	/*
212 	 * We only accept RAs when
213 	 * the system-wide variable allows the acceptance, and the
214 	 * per-interface variable allows RAs on the receiving interface.
215 	 */
216 	if (!nd6_accepts_rtadv(ndi))
217 		goto freeit;
218 
219 	if (ip6->ip6_hlim != 255) {
220 		nd6log((LOG_ERR,
221 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
222 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
223 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
224 		goto bad;
225 	}
226 
227 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
228 		nd6log((LOG_ERR,
229 		    "nd6_ra_input: src %s is not link-local\n",
230 		    ip6_sprintf(&saddr6)));
231 		goto bad;
232 	}
233 
234 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
235 	if (nd_ra == NULL) {
236 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
237 		return;
238 	}
239 
240 	icmp6len -= sizeof(*nd_ra);
241 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
242 	if (nd6_options(&ndopts) < 0) {
243 		nd6log((LOG_INFO,
244 		    "nd6_ra_input: invalid ND option, ignored\n"));
245 		/* nd6_options have incremented stats */
246 		goto freeit;
247 	}
248 
249     {
250 	struct nd_defrouter drtr;
251 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
252 
253 	/* remember if this is a multicasted advertisement */
254 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
255 		mcast = 1;
256 
257 	memset(&drtr, 0, sizeof(drtr));
258 	drtr.rtaddr = saddr6;
259 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
260 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
261 	drtr.expire = time_second + drtr.rtlifetime;
262 	drtr.ifp = ifp;
263 	/* unspecified or not? (RFC 2461 6.3.4) */
264 	if (advreachable) {
265 		NTOHL(advreachable);
266 		if (advreachable <= MAX_REACHABLE_TIME &&
267 		    ndi->basereachable != advreachable) {
268 			ndi->basereachable = advreachable;
269 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
270 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
271 		}
272 	}
273 	if (nd_ra->nd_ra_retransmit)
274 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
275 	if (nd_ra->nd_ra_curhoplimit)
276 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
277 	dr = defrtrlist_update(&drtr);
278     }
279 
280 	/*
281 	 * prefix
282 	 */
283 	if (ndopts.nd_opts_pi) {
284 		struct nd_opt_hdr *pt;
285 		struct nd_opt_prefix_info *pi = NULL;
286 		struct nd_prefixctl pr;
287 
288 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
289 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
290 		     pt = (struct nd_opt_hdr *)((char *)pt +
291 						(pt->nd_opt_len << 3))) {
292 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
293 				continue;
294 			pi = (struct nd_opt_prefix_info *)pt;
295 
296 			if (pi->nd_opt_pi_len != 4) {
297 				nd6log((LOG_INFO,
298 				    "nd6_ra_input: invalid option "
299 				    "len %d for prefix information option, "
300 				    "ignored\n", pi->nd_opt_pi_len));
301 				continue;
302 			}
303 
304 			if (128 < pi->nd_opt_pi_prefix_len) {
305 				nd6log((LOG_INFO,
306 				    "nd6_ra_input: invalid prefix "
307 				    "len %d for prefix information option, "
308 				    "ignored\n", pi->nd_opt_pi_prefix_len));
309 				continue;
310 			}
311 
312 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
313 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
314 				nd6log((LOG_INFO,
315 				    "nd6_ra_input: invalid prefix "
316 				    "%s, ignored\n",
317 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
318 				continue;
319 			}
320 
321 			memset(&pr, 0, sizeof(pr));
322 			sockaddr_in6_init(&pr.ndpr_prefix,
323 			    &pi->nd_opt_pi_prefix, 0, 0, 0);
324 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
325 
326 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
327 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
328 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
329 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
330 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
331 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
332 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
333 
334 			(void)prelist_update(&pr, dr, m, mcast);
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 	ICMP6_STATINC(ICMP6_STAT_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 static void
419 nd6_rtmsg(int cmd, struct rtentry *rt)
420 {
421 	struct rt_addrinfo info;
422 
423 	memset((void *)&info, 0, sizeof(info));
424 	info.rti_info[RTAX_DST] = rt_getkey(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] = rt->rt_ifp->if_dl->ifa_addr;
429 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
430 	}
431 
432 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
433 }
434 
435 void
436 defrouter_addreq(struct nd_defrouter *new)
437 {
438 	union {
439 		struct sockaddr_in6 sin6;
440 		struct sockaddr sa;
441 	} def, mask, gate;
442 	struct rtentry *newrt = NULL;
443 	int s;
444 	int error;
445 
446 	memset(&def, 0, sizeof(def));
447 	memset(&mask, 0, sizeof(mask));
448 	memset(&gate, 0,sizeof(gate)); /* for safety */
449 
450 	def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len =
451 	    sizeof(struct sockaddr_in6);
452 	def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6;
453 	gate.sin6.sin6_addr = new->rtaddr;
454 #ifndef SCOPEDROUTING
455 	gate.sin6.sin6_scope_id = 0;	/* XXX */
456 #endif
457 
458 	s = splsoftnet();
459 	error = rtrequest(RTM_ADD, &def.sa, &gate.sa, &mask.sa,
460 	    RTF_GATEWAY, &newrt);
461 	if (newrt) {
462 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
463 		newrt->rt_refcnt--;
464 		nd6_numroutes++;
465 	}
466 	if (error == 0)
467 		new->installed = 1;
468 	splx(s);
469 	return;
470 }
471 
472 struct nd_defrouter *
473 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
474 {
475 	struct nd_defrouter *dr;
476 
477 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
478 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
479 			break;
480 	}
481 
482 	return dr;		/* search failed */
483 }
484 
485 void
486 defrtrlist_del(struct nd_defrouter *dr)
487 {
488 	struct nd_ifinfo *ndi = ND_IFINFO(dr->ifp);
489 	struct nd_defrouter *deldr = NULL;
490 	struct nd_prefix *pr;
491 
492 	/*
493 	 * Flush all the routing table entries that use the router
494 	 * as a next hop.
495 	 */
496 	/* XXX: better condition? */
497 	if (!ip6_forwarding && nd6_accepts_rtadv(ndi))
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 	free(dr, M_IP6NDP);
525 }
526 
527 /*
528  * Remove the default route for a given router.
529  * This is just a subroutine function for defrouter_select(), and should
530  * not be called from anywhere else.
531  */
532 static void
533 defrouter_delreq(struct nd_defrouter *dr)
534 {
535 	union {
536 		struct sockaddr_in6 sin6;
537 		struct sockaddr sa;
538 	} def, mask, gw;
539 	struct rtentry *oldrt = NULL;
540 
541 #ifdef DIAGNOSTIC
542 	if (dr == NULL)
543 		panic("dr == NULL in defrouter_delreq");
544 #endif
545 
546 	memset(&def, 0, sizeof(def));
547 	memset(&mask, 0, sizeof(mask));
548 	memset(&gw, 0, sizeof(gw));	/* for safety */
549 
550 	def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len =
551 	    sizeof(struct sockaddr_in6);
552 	def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6;
553 	gw.sin6.sin6_addr = dr->rtaddr;
554 #ifndef SCOPEDROUTING
555 	gw.sin6.sin6_scope_id = 0;	/* XXX */
556 #endif
557 
558 	rtrequest(RTM_DELETE, &def.sa, &gw.sa, &mask.sa, RTF_GATEWAY, &oldrt);
559 	if (oldrt) {
560 		nd6_rtmsg(RTM_DELETE, oldrt);
561 		if (oldrt->rt_refcnt <= 0) {
562 			/*
563 			 * XXX: borrowed from the RTM_DELETE case of
564 			 * rtrequest().
565 			 */
566 			oldrt->rt_refcnt++;
567 			rtfree(oldrt);
568 			nd6_numroutes--;
569 		}
570 	}
571 
572 	dr->installed = 0;
573 }
574 
575 /*
576  * remove all default routes from default router list
577  */
578 void
579 defrouter_reset(void)
580 {
581 	struct nd_defrouter *dr;
582 
583 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
584 	     dr = TAILQ_NEXT(dr, dr_entry))
585 		defrouter_delreq(dr);
586 
587 	/*
588 	 * XXX should we also nuke any default routers in the kernel, by
589 	 * going through them by rtalloc1()?
590 	 */
591 }
592 
593 /*
594  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
595  * draft-ietf-ipngwg-router-selection:
596  * 1) Routers that are reachable or probably reachable should be preferred.
597  *    If we have more than one (probably) reachable router, prefer ones
598  *    with the highest router preference.
599  * 2) When no routers on the list are known to be reachable or
600  *    probably reachable, routers SHOULD be selected in a round-robin
601  *    fashion, regardless of router preference values.
602  * 3) If the Default Router List is empty, assume that all
603  *    destinations are on-link.
604  *
605  * We assume nd_defrouter is sorted by router preference value.
606  * Since the code below covers both with and without router preference cases,
607  * we do not need to classify the cases by ifdef.
608  *
609  * At this moment, we do not try to install more than one default router,
610  * even when the multipath routing is available, because we're not sure about
611  * the benefits for stub hosts comparing to the risk of making the code
612  * complicated and the possibility of introducing bugs.
613  */
614 void
615 defrouter_select(void)
616 {
617 	struct nd_ifinfo *ndi;
618 	int s = splsoftnet();
619 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
620 	struct rtentry *rt = NULL;
621 	struct llinfo_nd6 *ln = NULL;
622 
623 	/*
624 	 * This function should be called only when acting as an autoconfigured
625 	 * host.  Although the remaining part of this function is not effective
626 	 * if the node is not an autoconfigured host, we explicitly exclude
627 	 * such cases here for safety.
628 	 */
629 	if (ip6_forwarding) {
630 		nd6log((LOG_WARNING,
631 		    "defrouter_select: called unexpectedly (forwarding=%d, "
632 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
633 		splx(s);
634 		return;
635 	}
636 
637 	/*
638 	 * Let's handle easy case (3) first:
639 	 * If default router list is empty, there's nothing to be done.
640 	 */
641 	if (!TAILQ_FIRST(&nd_defrouter)) {
642 		splx(s);
643 		return;
644 	}
645 
646 	/*
647 	 * Search for a (probably) reachable router from the list.
648 	 * We just pick up the first reachable one (if any), assuming that
649 	 * the ordering rule of the list described in defrtrlist_update().
650 	 */
651 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
652 	     dr = TAILQ_NEXT(dr, dr_entry)) {
653 		ndi = ND_IFINFO(dr->ifp);
654 		if (nd6_accepts_rtadv(ndi))
655 			continue;
656 
657 		if (selected_dr == NULL &&
658 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) != NULL &&
659 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) != NULL &&
660 		    ND6_IS_LLINFO_PROBREACH(ln)) {
661 			selected_dr = dr;
662 		}
663 
664 		if (dr->installed && !installed_dr)
665 			installed_dr = dr;
666 		else if (dr->installed && installed_dr) {
667 			/* this should not happen.  warn for diagnosis. */
668 			log(LOG_ERR, "defrouter_select: more than one router"
669 			    " is installed\n");
670 		}
671 	}
672 	/*
673 	 * If none of the default routers was found to be reachable,
674 	 * round-robin the list regardless of preference.
675 	 * Otherwise, if we have an installed router, check if the selected
676 	 * (reachable) router should really be preferred to the installed one.
677 	 * We only prefer the new router when the old one is not reachable
678 	 * or when the new one has a really higher preference value.
679 	 */
680 	if (selected_dr == NULL) {
681 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
682 			selected_dr = TAILQ_FIRST(&nd_defrouter);
683 		else
684 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
685 	} else if (installed_dr &&
686 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
687 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
688 	    ND6_IS_LLINFO_PROBREACH(ln) &&
689 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
690 		selected_dr = installed_dr;
691 	}
692 
693 	/*
694 	 * If the selected router is different than the installed one,
695 	 * remove the installed router and install the selected one.
696 	 * Note that the selected router is never NULL here.
697 	 */
698 	if (installed_dr != selected_dr) {
699 		if (installed_dr)
700 			defrouter_delreq(installed_dr);
701 		defrouter_addreq(selected_dr);
702 	}
703 
704 	splx(s);
705 	return;
706 }
707 
708 /*
709  * for default router selection
710  * regards router-preference field as a 2-bit signed integer
711  */
712 static int
713 rtpref(struct nd_defrouter *dr)
714 {
715 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
716 	case ND_RA_FLAG_RTPREF_HIGH:
717 		return (RTPREF_HIGH);
718 	case ND_RA_FLAG_RTPREF_MEDIUM:
719 	case ND_RA_FLAG_RTPREF_RSV:
720 		return (RTPREF_MEDIUM);
721 	case ND_RA_FLAG_RTPREF_LOW:
722 		return (RTPREF_LOW);
723 	default:
724 		/*
725 		 * This case should never happen.  If it did, it would mean a
726 		 * serious bug of kernel internal.  We thus always bark here.
727 		 * Or, can we even panic?
728 		 */
729 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
730 		return (RTPREF_INVALID);
731 	}
732 	/* NOTREACHED */
733 }
734 
735 static struct nd_defrouter *
736 defrtrlist_update(struct nd_defrouter *new)
737 {
738 	struct nd_defrouter *dr, *n;
739 	int s = splsoftnet();
740 
741 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
742 		/* entry exists */
743 		if (new->rtlifetime == 0) {
744 			defrtrlist_del(dr);
745 			dr = NULL;
746 		} else {
747 			int oldpref = rtpref(dr);
748 
749 			/* override */
750 			dr->flags = new->flags; /* xxx flag check */
751 			dr->rtlifetime = new->rtlifetime;
752 			dr->expire = new->expire;
753 
754 			/*
755 			 * If the preference does not change, there's no need
756 			 * to sort the entries.
757 			 */
758 			if (rtpref(new) == oldpref) {
759 				splx(s);
760 				return (dr);
761 			}
762 
763 			/*
764 			 * preferred router may be changed, so relocate
765 			 * this router.
766 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
767 			 * However, since defrtrlist_del() has many side
768 			 * effects, we intentionally do so here.
769 			 * defrouter_select() below will handle routing
770 			 * changes later.
771 			 */
772 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
773 			n = dr;
774 			goto insert;
775 		}
776 		splx(s);
777 		return (dr);
778 	}
779 
780 	/* entry does not exist */
781 	if (new->rtlifetime == 0) {
782 		splx(s);
783 		return (NULL);
784 	}
785 
786 	if (ip6_rtadv_maxroutes <= nd6_numroutes) {
787 		ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
788 		splx(s);
789 		return (NULL);
790 	}
791 
792 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
793 	if (n == NULL) {
794 		splx(s);
795 		return (NULL);
796 	}
797 	memset(n, 0, sizeof(*n));
798 	*n = *new;
799 
800 insert:
801 	/*
802 	 * Insert the new router in the Default Router List;
803 	 * The Default Router List should be in the descending order
804 	 * of router-preferece.  Routers with the same preference are
805 	 * sorted in the arriving time order.
806 	 */
807 
808 	/* insert at the end of the group */
809 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
810 	     dr = TAILQ_NEXT(dr, dr_entry)) {
811 		if (rtpref(n) > rtpref(dr))
812 			break;
813 	}
814 	if (dr)
815 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
816 	else
817 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
818 
819 	defrouter_select();
820 
821 	splx(s);
822 
823 	return (n);
824 }
825 
826 static struct nd_pfxrouter *
827 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
828 {
829 	struct nd_pfxrouter *search;
830 
831 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
832 		if (search->router == dr)
833 			break;
834 	}
835 
836 	return (search);
837 }
838 
839 static void
840 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
841 {
842 	struct nd_pfxrouter *new;
843 
844 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT|M_ZERO);
845 	if (new == NULL)
846 		return;
847 	new->router = dr;
848 
849 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
850 
851 	pfxlist_onlink_check();
852 }
853 
854 static void
855 pfxrtr_del(struct nd_pfxrouter *pfr)
856 {
857 	LIST_REMOVE(pfr, pfr_entry);
858 	free(pfr, M_IP6NDP);
859 }
860 
861 struct nd_prefix *
862 nd6_prefix_lookup(struct nd_prefixctl *key)
863 {
864 	struct nd_prefix *search;
865 
866 	LIST_FOREACH(search, &nd_prefix, ndpr_entry) {
867 		if (key->ndpr_ifp == search->ndpr_ifp &&
868 		    key->ndpr_plen == search->ndpr_plen &&
869 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
870 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
871 			break;
872 		}
873 	}
874 
875 	return (search);
876 }
877 
878 int
879 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
880 	struct nd_prefix **newp)
881 {
882 	struct nd_prefix *new = NULL;
883 	int i, s;
884 	int error;
885 
886 	error = 0;
887 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT|M_ZERO);
888 	if (new == NULL)
889 		return ENOMEM;
890 	new->ndpr_ifp = pr->ndpr_ifp;
891 	new->ndpr_prefix = pr->ndpr_prefix;
892 	new->ndpr_plen = pr->ndpr_plen;
893 	new->ndpr_vltime = pr->ndpr_vltime;
894 	new->ndpr_pltime = pr->ndpr_pltime;
895 	new->ndpr_flags = pr->ndpr_flags;
896 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
897 		free(new, M_IP6NDP);
898 		return(error);
899 	}
900 	new->ndpr_lastupdate = time_second;
901 	if (newp != NULL)
902 		*newp = new;
903 
904 	/* initialization */
905 	LIST_INIT(&new->ndpr_advrtrs);
906 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
907 	/* make prefix in the canonical form */
908 	for (i = 0; i < 4; i++)
909 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
910 		    new->ndpr_mask.s6_addr32[i];
911 
912 	s = splsoftnet();
913 	/* link ndpr_entry to nd_prefix list */
914 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
915 	splx(s);
916 
917 	/* ND_OPT_PI_FLAG_ONLINK processing */
918 	if (new->ndpr_raf_onlink) {
919 		int e;
920 
921 		if ((e = nd6_prefix_onlink(new)) != 0) {
922 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
923 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
924 			    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
925 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
926 			/* proceed anyway. XXX: is it correct? */
927 		}
928 	}
929 
930 	if (dr)
931 		pfxrtr_add(new, dr);
932 
933 	return 0;
934 }
935 
936 void
937 prelist_remove(struct nd_prefix *pr)
938 {
939 	struct nd_pfxrouter *pfr, *next;
940 	int e, s;
941 
942 	/* make sure to invalidate the prefix until it is really freed. */
943 	pr->ndpr_vltime = 0;
944 	pr->ndpr_pltime = 0;
945 #if 0
946 	/*
947 	 * Though these flags are now meaningless, we'd rather keep the value
948 	 * not to confuse users when executing "ndp -p".
949 	 */
950 	pr->ndpr_raf_onlink = 0;
951 	pr->ndpr_raf_auto = 0;
952 #endif
953 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
954 	    (e = nd6_prefix_offlink(pr)) != 0) {
955 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
956 		    "on %s, errno=%d\n",
957 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
958 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
959 		/* what should we do? */
960 	}
961 
962 	if (pr->ndpr_refcnt > 0)
963 		return;		/* notice here? */
964 
965 	s = splsoftnet();
966 	/* unlink ndpr_entry from nd_prefix list */
967 	LIST_REMOVE(pr, ndpr_entry);
968 
969 	/* free list of routers that adversed the prefix */
970 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
971 		next = LIST_NEXT(pfr, pfr_entry);
972 
973 		free(pfr, M_IP6NDP);
974 	}
975 	splx(s);
976 
977 	free(pr, M_IP6NDP);
978 
979 	pfxlist_onlink_check();
980 }
981 
982 static int
983 prelist_update(struct nd_prefixctl *new,
984 	struct nd_defrouter *dr, /* may be NULL */
985 	struct mbuf *m,
986 	int mcast)
987 {
988 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
989 	struct ifaddr *ifa;
990 	struct ifnet *ifp = new->ndpr_ifp;
991 	struct nd_prefix *pr;
992 	int s = splsoftnet();
993 	int error = 0;
994 	int newprefix = 0;
995 	int auth;
996 	struct in6_addrlifetime lt6_tmp;
997 
998 	auth = 0;
999 	if (m) {
1000 		/*
1001 		 * Authenticity for NA consists authentication for
1002 		 * both IP header and IP datagrams, doesn't it ?
1003 		 */
1004 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1005 		auth = (m->m_flags & M_AUTHIPHDR
1006 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
1007 #endif
1008 	}
1009 
1010 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1011 		/*
1012 		 * nd6_prefix_lookup() ensures that pr and new have the same
1013 		 * prefix on a same interface.
1014 		 */
1015 
1016 		/*
1017 		 * Update prefix information.  Note that the on-link (L) bit
1018 		 * and the autonomous (A) bit should NOT be changed from 1
1019 		 * to 0.
1020 		 */
1021 		if (new->ndpr_raf_onlink == 1)
1022 			pr->ndpr_raf_onlink = 1;
1023 		if (new->ndpr_raf_auto == 1)
1024 			pr->ndpr_raf_auto = 1;
1025 		if (new->ndpr_raf_onlink) {
1026 			pr->ndpr_vltime = new->ndpr_vltime;
1027 			pr->ndpr_pltime = new->ndpr_pltime;
1028 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1029 			pr->ndpr_lastupdate = time_second;
1030 		}
1031 
1032 		if (new->ndpr_raf_onlink &&
1033 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1034 			int e;
1035 
1036 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1037 				nd6log((LOG_ERR,
1038 				    "prelist_update: failed to make "
1039 				    "the prefix %s/%d on-link on %s "
1040 				    "(errno=%d)\n",
1041 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1042 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1043 				/* proceed anyway. XXX: is it correct? */
1044 			}
1045 		}
1046 
1047 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1048 			pfxrtr_add(pr, dr);
1049 	} else {
1050 		struct nd_prefix *newpr = NULL;
1051 
1052 		newprefix = 1;
1053 
1054 		if (new->ndpr_vltime == 0)
1055 			goto end;
1056 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1057 			goto end;
1058 
1059 		if (ip6_rtadv_maxroutes <= nd6_numroutes) {
1060 			ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
1061 			goto end;
1062 		}
1063 
1064 		error = nd6_prelist_add(new, dr, &newpr);
1065 		if (error != 0 || newpr == NULL) {
1066 			nd6log((LOG_NOTICE, "prelist_update: "
1067 			    "nd6_prelist_add failed for %s/%d on %s "
1068 			    "errno=%d, returnpr=%p\n",
1069 			    ip6_sprintf(&new->ndpr_prefix.sin6_addr),
1070 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1071 			    error, newpr));
1072 			goto end; /* we should just give up in this case. */
1073 		}
1074 
1075 		/*
1076 		 * XXX: from the ND point of view, we can ignore a prefix
1077 		 * with the on-link bit being zero.  However, we need a
1078 		 * prefix structure for references from autoconfigured
1079 		 * addresses.  Thus, we explicitly make sure that the prefix
1080 		 * itself expires now.
1081 		 */
1082 		if (newpr->ndpr_raf_onlink == 0) {
1083 			newpr->ndpr_vltime = 0;
1084 			newpr->ndpr_pltime = 0;
1085 			in6_init_prefix_ltimes(newpr);
1086 		}
1087 
1088 		pr = newpr;
1089 	}
1090 
1091 	/*
1092 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1093 	 * Note that pr must be non NULL at this point.
1094 	 */
1095 
1096 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1097 	if (!new->ndpr_raf_auto)
1098 		goto end;
1099 
1100 	/*
1101 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1102 	 * nd6_ra_input.
1103 	 */
1104 
1105 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1106 	if (new->ndpr_pltime > new->ndpr_vltime) {
1107 		error = EINVAL;	/* XXX: won't be used */
1108 		goto end;
1109 	}
1110 
1111 	/*
1112 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1113 	 * an address configured by stateless autoconfiguration already in the
1114 	 * list of addresses associated with the interface, and the Valid
1115 	 * Lifetime is not 0, form an address.  We first check if we have
1116 	 * a matching prefix.
1117 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1118 	 * consider autoconfigured addresses while RFC2462 simply said
1119 	 * "address".
1120 	 */
1121 	IFADDR_FOREACH(ifa, ifp) {
1122 		struct in6_ifaddr *ifa6;
1123 		u_int32_t remaininglifetime;
1124 
1125 		if (ifa->ifa_addr->sa_family != AF_INET6)
1126 			continue;
1127 
1128 		ifa6 = (struct in6_ifaddr *)ifa;
1129 
1130 		/*
1131 		 * We only consider autoconfigured addresses as per rfc2462bis.
1132 		 */
1133 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1134 			continue;
1135 
1136 		/*
1137 		 * Spec is not clear here, but I believe we should concentrate
1138 		 * on unicast (i.e. not anycast) addresses.
1139 		 * XXX: other ia6_flags? detached or duplicated?
1140 		 */
1141 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1142 			continue;
1143 
1144 		/*
1145 		 * Ignore the address if it is not associated with a prefix
1146 		 * or is associated with a prefix that is different from this
1147 		 * one.  (pr is never NULL here)
1148 		 */
1149 		if (ifa6->ia6_ndpr != pr)
1150 			continue;
1151 
1152 		if (ia6_match == NULL) /* remember the first one */
1153 			ia6_match = ifa6;
1154 
1155 		/*
1156 		 * An already autoconfigured address matched.  Now that we
1157 		 * are sure there is at least one matched address, we can
1158 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1159 		 * "two hours" rule and the privacy extension.
1160 		 * We apply some clarifications in rfc2462bis:
1161 		 * - use remaininglifetime instead of storedlifetime as a
1162 		 *   variable name
1163 		 * - remove the dead code in the "two-hour" rule
1164 		 */
1165 #define TWOHOUR		(120*60)
1166 		lt6_tmp = ifa6->ia6_lifetime;
1167 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1168 			remaininglifetime = ND6_INFINITE_LIFETIME;
1169 		else if (time_second - ifa6->ia6_updatetime >
1170 			 lt6_tmp.ia6t_vltime) {
1171 			/*
1172 			 * The case of "invalid" address.  We should usually
1173 			 * not see this case.
1174 			 */
1175 			remaininglifetime = 0;
1176 		} else
1177 			remaininglifetime = lt6_tmp.ia6t_vltime -
1178 			    (time_second - ifa6->ia6_updatetime);
1179 
1180 		/* when not updating, keep the current stored lifetime. */
1181 		lt6_tmp.ia6t_vltime = remaininglifetime;
1182 
1183 		if (TWOHOUR < new->ndpr_vltime ||
1184 		    remaininglifetime < new->ndpr_vltime) {
1185 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1186 		} else if (remaininglifetime <= TWOHOUR) {
1187 			if (auth)
1188 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1189 		} else {
1190 			/*
1191 			 * new->ndpr_vltime <= TWOHOUR &&
1192 			 * TWOHOUR < remaininglifetime
1193 			 */
1194 			lt6_tmp.ia6t_vltime = TWOHOUR;
1195 		}
1196 
1197 		/* The 2 hour rule is not imposed for preferred lifetime. */
1198 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1199 
1200 		in6_init_address_ltimes(pr, &lt6_tmp);
1201 
1202 		/*
1203 		 * We need to treat lifetimes for temporary addresses
1204 		 * differently, according to
1205 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1206 		 * we only update the lifetimes when they are in the maximum
1207 		 * intervals.
1208 		 */
1209 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1210 			u_int32_t maxvltime, maxpltime;
1211 
1212 			if (ip6_temp_valid_lifetime >
1213 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1214 			    ip6_desync_factor)) {
1215 				maxvltime = ip6_temp_valid_lifetime -
1216 				    (time_second - ifa6->ia6_createtime) -
1217 				    ip6_desync_factor;
1218 			} else
1219 				maxvltime = 0;
1220 			if (ip6_temp_preferred_lifetime >
1221 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1222 			    ip6_desync_factor)) {
1223 				maxpltime = ip6_temp_preferred_lifetime -
1224 				    (time_second - ifa6->ia6_createtime) -
1225 				    ip6_desync_factor;
1226 			} else
1227 				maxpltime = 0;
1228 
1229 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1230 			    lt6_tmp.ia6t_vltime > maxvltime) {
1231 				lt6_tmp.ia6t_vltime = maxvltime;
1232 			}
1233 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1234 			    lt6_tmp.ia6t_pltime > maxpltime) {
1235 				lt6_tmp.ia6t_pltime = maxpltime;
1236 			}
1237 		}
1238 
1239 		ifa6->ia6_lifetime = lt6_tmp;
1240 		ifa6->ia6_updatetime = time_second;
1241 	}
1242 	if (ia6_match == NULL && new->ndpr_vltime) {
1243 		int ifidlen;
1244 
1245 		/*
1246 		 * 5.5.3 (d) (continued)
1247 		 * No address matched and the valid lifetime is non-zero.
1248 		 * Create a new address.
1249 		 */
1250 
1251 		/*
1252 		 * Prefix Length check:
1253 		 * If the sum of the prefix length and interface identifier
1254 		 * length does not equal 128 bits, the Prefix Information
1255 		 * option MUST be ignored.  The length of the interface
1256 		 * identifier is defined in a separate link-type specific
1257 		 * document.
1258 		 */
1259 		ifidlen = in6_if2idlen(ifp);
1260 		if (ifidlen < 0) {
1261 			/* this should not happen, so we always log it. */
1262 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1263 			    if_name(ifp));
1264 			goto end;
1265 		}
1266 		if (ifidlen + pr->ndpr_plen != 128) {
1267 			nd6log((LOG_INFO,
1268 			    "prelist_update: invalid prefixlen "
1269 			    "%d for %s, ignored\n",
1270 			    pr->ndpr_plen, if_name(ifp)));
1271 			goto end;
1272 		}
1273 
1274 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1275 			/*
1276 			 * note that we should use pr (not new) for reference.
1277 			 */
1278 			pr->ndpr_refcnt++;
1279 			ia6->ia6_ndpr = pr;
1280 
1281 			/*
1282 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
1283 			 * When a new public address is created as described
1284 			 * in RFC2462, also create a new temporary address.
1285 			 *
1286 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
1287 			 * When an interface connects to a new link, a new
1288 			 * randomized interface identifier should be generated
1289 			 * immediately together with a new set of temporary
1290 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1291 			 * in6_tmpifadd().
1292 			 */
1293 			if (ip6_use_tempaddr) {
1294 				int e;
1295 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1296 					nd6log((LOG_NOTICE, "prelist_update: "
1297 					    "failed to create a temporary "
1298 					    "address, errno=%d\n",
1299 					    e));
1300 				}
1301 			}
1302 
1303 			/*
1304 			 * A newly added address might affect the status
1305 			 * of other addresses, so we check and update it.
1306 			 * XXX: what if address duplication happens?
1307 			 */
1308 			pfxlist_onlink_check();
1309 		} else {
1310 			/* just set an error. do not bark here. */
1311 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1312 		}
1313 	}
1314 
1315  end:
1316 	splx(s);
1317 	return error;
1318 }
1319 
1320 /*
1321  * A supplement function used in the on-link detection below;
1322  * detect if a given prefix has a (probably) reachable advertising router.
1323  * XXX: lengthy function name...
1324  */
1325 static struct nd_pfxrouter *
1326 find_pfxlist_reachable_router(struct nd_prefix *pr)
1327 {
1328 	struct nd_pfxrouter *pfxrtr;
1329 	struct rtentry *rt;
1330 	struct llinfo_nd6 *ln;
1331 
1332 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1333 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1334 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1335 		    pfxrtr->router->ifp)) &&
1336 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1337 		    ND6_IS_LLINFO_PROBREACH(ln))
1338 			break;	/* found */
1339 	}
1340 
1341 	return (pfxrtr);
1342 }
1343 
1344 /*
1345  * Check if each prefix in the prefix list has at least one available router
1346  * that advertised the prefix (a router is "available" if its neighbor cache
1347  * entry is reachable or probably reachable).
1348  * If the check fails, the prefix may be off-link, because, for example,
1349  * we have moved from the network but the lifetime of the prefix has not
1350  * expired yet.  So we should not use the prefix if there is another prefix
1351  * that has an available router.
1352  * But, if there is no prefix that has an available router, we still regards
1353  * all the prefixes as on-link.  This is because we can't tell if all the
1354  * routers are simply dead or if we really moved from the network and there
1355  * is no router around us.
1356  */
1357 void
1358 pfxlist_onlink_check(void)
1359 {
1360 	struct nd_prefix *pr;
1361 	struct in6_ifaddr *ifa;
1362 	struct nd_defrouter *dr;
1363 	struct nd_pfxrouter *pfxrtr = NULL;
1364 
1365 	/*
1366 	 * Check if there is a prefix that has a reachable advertising
1367 	 * router.
1368 	 */
1369 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1370 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1371 			break;
1372 	}
1373 	/*
1374 	 * If we have no such prefix, check whether we still have a router
1375 	 * that does not advertise any prefixes.
1376 	 */
1377 	if (pr == NULL) {
1378 		TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
1379 			struct nd_prefix *pr0;
1380 
1381 			LIST_FOREACH(pr0, &nd_prefix, ndpr_entry) {
1382 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1383 					break;
1384 			}
1385 			if (pfxrtr)
1386 				break;
1387 		}
1388 	}
1389 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
1390 		/*
1391 		 * There is at least one prefix that has a reachable router,
1392 		 * or at least a router which probably does not advertise
1393 		 * any prefixes.  The latter would be the case when we move
1394 		 * to a new link where we have a router that does not provide
1395 		 * prefixes and we configure an address by hand.
1396 		 * Detach prefixes which have no reachable advertising
1397 		 * router, and attach other prefixes.
1398 		 */
1399 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1400 			/* XXX: a link-local prefix should never be detached */
1401 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1402 				continue;
1403 
1404 			/*
1405 			 * we aren't interested in prefixes without the L bit
1406 			 * set.
1407 			 */
1408 			if (pr->ndpr_raf_onlink == 0)
1409 				continue;
1410 
1411 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1412 			    find_pfxlist_reachable_router(pr) == NULL)
1413 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1414 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1415 			    find_pfxlist_reachable_router(pr) != 0)
1416 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1417 		}
1418 	} else {
1419 		/* there is no prefix that has a reachable router */
1420 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1421 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1422 				continue;
1423 
1424 			if (pr->ndpr_raf_onlink == 0)
1425 				continue;
1426 
1427 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1428 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1429 		}
1430 	}
1431 
1432 	/*
1433 	 * Remove each interface route associated with a (just) detached
1434 	 * prefix, and reinstall the interface route for a (just) attached
1435 	 * prefix.  Note that all attempt of reinstallation does not
1436 	 * necessarily success, when a same prefix is shared among multiple
1437 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1438 	 * so we don't have to care about them.
1439 	 */
1440 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1441 		int e;
1442 
1443 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1444 			continue;
1445 
1446 		if (pr->ndpr_raf_onlink == 0)
1447 			continue;
1448 
1449 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1450 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1451 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1452 				nd6log((LOG_ERR,
1453 				    "pfxlist_onlink_check: failed to "
1454 				    "make %s/%d offlink, errno=%d\n",
1455 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1456 				    pr->ndpr_plen, e));
1457 			}
1458 		}
1459 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1460 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1461 		    pr->ndpr_raf_onlink) {
1462 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1463 				nd6log((LOG_ERR,
1464 				    "pfxlist_onlink_check: failed to "
1465 				    "make %s/%d onlink, errno=%d\n",
1466 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1467 				    pr->ndpr_plen, e));
1468 			}
1469 		}
1470 	}
1471 
1472 	/*
1473 	 * Changes on the prefix status might affect address status as well.
1474 	 * Make sure that all addresses derived from an attached prefix are
1475 	 * attached, and that all addresses derived from a detached prefix are
1476 	 * detached.  Note, however, that a manually configured address should
1477 	 * always be attached.
1478 	 * The precise detection logic is same as the one for prefixes.
1479 	 */
1480 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1481 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1482 			continue;
1483 
1484 		if (ifa->ia6_ndpr == NULL) {
1485 			/*
1486 			 * This can happen when we first configure the address
1487 			 * (i.e. the address exists, but the prefix does not).
1488 			 * XXX: complicated relationships...
1489 			 */
1490 			continue;
1491 		}
1492 
1493 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1494 			break;
1495 	}
1496 	if (ifa) {
1497 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1498 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1499 				continue;
1500 
1501 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1502 				continue;
1503 
1504 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1505 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1506 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1507 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1508 					nd6_dad_start((struct ifaddr *)ifa,
1509 					    0);
1510 				}
1511 			} else {
1512 				if ((ifa->ia6_flags & IN6_IFF_DETACHED) == 0) {
1513 					ifa->ia6_flags |= IN6_IFF_DETACHED;
1514 				}
1515 			}
1516 		}
1517 	}
1518 	else {
1519 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1520 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1521 				continue;
1522 
1523 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1524 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1525 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1526 				/* Do we need a delay in this case? */
1527 				nd6_dad_start((struct ifaddr *)ifa, 0);
1528 			}
1529 		}
1530 	}
1531 }
1532 
1533 int
1534 nd6_prefix_onlink(struct nd_prefix *pr)
1535 {
1536 	struct ifaddr *ifa;
1537 	struct ifnet *ifp = pr->ndpr_ifp;
1538 	struct sockaddr_in6 mask6;
1539 	struct nd_prefix *opr;
1540 	u_long rtflags;
1541 	int error = 0;
1542 	struct rtentry *rt = NULL;
1543 
1544 	/* sanity check */
1545 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1546 		nd6log((LOG_ERR,
1547 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1548 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1549 		return (EEXIST);
1550 	}
1551 
1552 	/*
1553 	 * Add the interface route associated with the prefix.  Before
1554 	 * installing the route, check if there's the same prefix on another
1555 	 * interface, and the prefix has already installed the interface route.
1556 	 * Although such a configuration is expected to be rare, we explicitly
1557 	 * allow it.
1558 	 */
1559 	LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1560 		if (opr == pr)
1561 			continue;
1562 
1563 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1564 			continue;
1565 
1566 		if (opr->ndpr_plen == pr->ndpr_plen &&
1567 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1568 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1569 			return (0);
1570 	}
1571 
1572 	/*
1573 	 * We prefer link-local addresses as the associated interface address.
1574 	 */
1575 	/* search for a link-local addr */
1576 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1577 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1578 	if (ifa == NULL) {
1579 		/* XXX: freebsd does not have ifa_ifwithaf */
1580 		IFADDR_FOREACH(ifa, ifp) {
1581 			if (ifa->ifa_addr->sa_family == AF_INET6)
1582 				break;
1583 		}
1584 		/* should we care about ia6_flags? */
1585 	}
1586 	if (ifa == NULL) {
1587 		/*
1588 		 * This can still happen, when, for example, we receive an RA
1589 		 * containing a prefix with the L bit set and the A bit clear,
1590 		 * after removing all IPv6 addresses on the receiving
1591 		 * interface.  This should, of course, be rare though.
1592 		 */
1593 		nd6log((LOG_NOTICE,
1594 		    "nd6_prefix_onlink: failed to find any ifaddr"
1595 		    " to add route for a prefix(%s/%d) on %s\n",
1596 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1597 		    pr->ndpr_plen, if_name(ifp)));
1598 		return (0);
1599 	}
1600 
1601 	/*
1602 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1603 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1604 	 */
1605 	memset(&mask6, 0, sizeof(mask6));
1606 	mask6.sin6_len = sizeof(mask6);
1607 	mask6.sin6_addr = pr->ndpr_mask;
1608 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
1609 	rtflags = ifa->ifa_flags | RTF_UP;
1610 	if (nd6_need_cache(ifp)) {
1611 		/* explicitly set in case ifa_flags does not set the flag. */
1612 		rtflags |= RTF_CLONING;
1613 	} else {
1614 		/*
1615 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1616 		 */
1617 		rtflags &= ~RTF_CLONING;
1618 	}
1619 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1620 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1621 	if (error == 0) {
1622 		if (rt != NULL) { /* this should be non NULL, though */
1623 			nd6_rtmsg(RTM_ADD, rt);
1624 			nd6_numroutes++;
1625 		}
1626 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1627 	} else {
1628 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1629 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1630 		    "errno = %d\n",
1631 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1632 		    pr->ndpr_plen, if_name(ifp),
1633 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1634 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
1635 	}
1636 
1637 	if (rt != NULL)
1638 		rt->rt_refcnt--;
1639 
1640 	return (error);
1641 }
1642 
1643 int
1644 nd6_prefix_offlink(struct nd_prefix *pr)
1645 {
1646 	int error = 0;
1647 	struct ifnet *ifp = pr->ndpr_ifp;
1648 	struct nd_prefix *opr;
1649 	struct sockaddr_in6 sa6, mask6;
1650 	struct rtentry *rt = NULL;
1651 
1652 	/* sanity check */
1653 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1654 		nd6log((LOG_ERR,
1655 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1656 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1657 		return (EEXIST);
1658 	}
1659 
1660 	sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1661 	sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
1662 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1663 	    (struct sockaddr *)&mask6, 0, &rt);
1664 	if (error == 0) {
1665 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1666 
1667 		/* report the route deletion to the routing socket. */
1668 		if (rt != NULL) {
1669 			nd6_rtmsg(RTM_DELETE, rt);
1670 			nd6_numroutes--;
1671 		}
1672 
1673 		/*
1674 		 * There might be the same prefix on another interface,
1675 		 * the prefix which could not be on-link just because we have
1676 		 * the interface route (see comments in nd6_prefix_onlink).
1677 		 * If there's one, try to make the prefix on-link on the
1678 		 * interface.
1679 		 */
1680 		LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1681 			if (opr == pr)
1682 				continue;
1683 
1684 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1685 				continue;
1686 
1687 			/*
1688 			 * KAME specific: detached prefixes should not be
1689 			 * on-link.
1690 			 */
1691 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1692 				continue;
1693 
1694 			if (opr->ndpr_plen == pr->ndpr_plen &&
1695 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1696 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1697 				int e;
1698 
1699 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1700 					nd6log((LOG_ERR,
1701 					    "nd6_prefix_offlink: failed to "
1702 					    "recover a prefix %s/%d from %s "
1703 					    "to %s (errno = %d)\n",
1704 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1705 					    opr->ndpr_plen, if_name(ifp),
1706 					    if_name(opr->ndpr_ifp), e));
1707 				}
1708 			}
1709 		}
1710 	} else {
1711 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1712 		nd6log((LOG_ERR,
1713 		    "nd6_prefix_offlink: failed to delete route: "
1714 		    "%s/%d on %s (errno = %d)\n",
1715 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1716 		    error));
1717 	}
1718 
1719 	if (rt != NULL) {
1720 		if (rt->rt_refcnt <= 0) {
1721 			/* XXX: we should free the entry ourselves. */
1722 			rt->rt_refcnt++;
1723 			rtfree(rt);
1724 		}
1725 	}
1726 
1727 	return (error);
1728 }
1729 
1730 static struct in6_ifaddr *
1731 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1732 {
1733 	struct ifnet *ifp = pr->ndpr_ifp;
1734 	struct ifaddr *ifa;
1735 	struct in6_aliasreq ifra;
1736 	struct in6_ifaddr *ia, *ib;
1737 	int error, plen0;
1738 	struct in6_addr mask;
1739 	int prefixlen = pr->ndpr_plen;
1740 	int updateflags;
1741 
1742 	in6_prefixlen2mask(&mask, prefixlen);
1743 
1744 	/*
1745 	 * find a link-local address (will be interface ID).
1746 	 * Is it really mandatory? Theoretically, a global or a site-local
1747 	 * address can be configured without a link-local address, if we
1748 	 * have a unique interface identifier...
1749 	 *
1750 	 * it is not mandatory to have a link-local address, we can generate
1751 	 * interface identifier on the fly.  we do this because:
1752 	 * (1) it should be the easiest way to find interface identifier.
1753 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1754 	 * for multiple addresses on a single interface, and possible shortcut
1755 	 * of DAD.  we omitted DAD for this reason in the past.
1756 	 * (3) a user can prevent autoconfiguration of global address
1757 	 * by removing link-local address by hand (this is partly because we
1758 	 * don't have other way to control the use of IPv6 on an interface.
1759 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1760 	 * (4) it is easier to manage when an interface has addresses
1761 	 * with the same interface identifier, than to have multiple addresses
1762 	 * with different interface identifiers.
1763 	 */
1764 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1765 	if (ifa)
1766 		ib = (struct in6_ifaddr *)ifa;
1767 	else
1768 		return NULL;
1769 
1770 #if 0 /* don't care link local addr state, and always do DAD */
1771 	/* if link-local address is not eligible, do not autoconfigure. */
1772 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1773 		printf("in6_ifadd: link-local address not ready\n");
1774 		return NULL;
1775 	}
1776 #endif
1777 
1778 	/* prefixlen + ifidlen must be equal to 128 */
1779 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1780 	if (prefixlen != plen0) {
1781 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1782 		    "(prefix=%d ifid=%d)\n",
1783 		    if_name(ifp), prefixlen, 128 - plen0));
1784 		return NULL;
1785 	}
1786 
1787 	/* make ifaddr */
1788 
1789 	memset(&ifra, 0, sizeof(ifra));
1790 	/*
1791 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1792 	 * for safety.
1793 	 */
1794 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1795 	sockaddr_in6_init(&ifra.ifra_addr, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1796 	/* prefix */
1797 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1798 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1799 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1800 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1801 
1802 	/* interface ID */
1803 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1804 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1805 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1806 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1807 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1808 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1809 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1810 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1811 
1812 	/* new prefix mask. */
1813 	sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
1814 
1815 	/* lifetimes */
1816 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1817 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1818 
1819 	/* XXX: scope zone ID? */
1820 
1821 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1822 
1823 	/*
1824 	 * Make sure that we do not have this address already.  This should
1825 	 * usually not happen, but we can still see this case, e.g., if we
1826 	 * have manually configured the exact address to be configured.
1827 	 */
1828 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1829 		/* this should be rare enough to make an explicit log */
1830 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1831 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1832 		return (NULL);
1833 	}
1834 
1835 	/*
1836 	 * Allocate ifaddr structure, link into chain, etc.
1837 	 * If we are going to create a new address upon receiving a multicasted
1838 	 * RA, we need to impose a random delay before starting DAD.
1839 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1840 	 */
1841 	updateflags = 0;
1842 	if (mcast)
1843 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1844 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1845 		nd6log((LOG_ERR,
1846 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1847 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1848 		    error));
1849 		return (NULL);	/* ifaddr must not have been allocated. */
1850 	}
1851 
1852 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1853 
1854 	return (ia);		/* this is always non-NULL */
1855 }
1856 
1857 int
1858 in6_tmpifadd(
1859 	const struct in6_ifaddr *ia0, /* corresponding public address */
1860 	int forcegen,
1861 	int dad_delay)
1862 {
1863 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1864 	struct in6_ifaddr *newia, *ia;
1865 	struct in6_aliasreq ifra;
1866 	int i, error;
1867 	int trylimit = 3;	/* XXX: adhoc value */
1868 	int updateflags;
1869 	u_int32_t randid[2];
1870 	u_int32_t vltime0, pltime0;
1871 
1872 	memset(&ifra, 0, sizeof(ifra));
1873 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1874 	ifra.ifra_addr = ia0->ia_addr;
1875 	/* copy prefix mask */
1876 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1877 	/* clear the old IFID */
1878 	for (i = 0; i < 4; i++) {
1879 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1880 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1881 	}
1882 
1883   again:
1884 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1885 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1886 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1887 		    "random IFID\n"));
1888 		return (EINVAL);
1889 	}
1890 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1891 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1892 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1893 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1894 
1895 	/*
1896 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1897 	 * However, we may still have a chance to see collision, because
1898 	 * there may be a time lag between generation of the ID and generation
1899 	 * of the address.  So, we'll do one more sanity check.
1900 	 */
1901 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1902 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1903 		    &ifra.ifra_addr.sin6_addr)) {
1904 			if (trylimit-- == 0) {
1905 				/*
1906 				 * Give up.  Something strange should have
1907 				 * happened.
1908 				 */
1909 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1910 				    "find a unique random IFID\n"));
1911 				return (EEXIST);
1912 			}
1913 			forcegen = 1;
1914 			goto again;
1915 		}
1916 	}
1917 
1918 	/*
1919 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1920          * public address or TEMP_VALID_LIFETIME.
1921 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1922          * of the public address or TEMP_PREFERRED_LIFETIME -
1923          * DESYNC_FACTOR.
1924 	 */
1925 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1926 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1927 		    (ia0->ia6_lifetime.ia6t_vltime -
1928 		    (time_second - ia0->ia6_updatetime));
1929 		if (vltime0 > ip6_temp_valid_lifetime)
1930 			vltime0 = ip6_temp_valid_lifetime;
1931 	} else
1932 		vltime0 = ip6_temp_valid_lifetime;
1933 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1934 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1935 		    (ia0->ia6_lifetime.ia6t_pltime -
1936 		    (time_second - ia0->ia6_updatetime));
1937 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
1938 			pltime0 = ip6_temp_preferred_lifetime -
1939 			    ip6_desync_factor;
1940 		}
1941 	} else
1942 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
1943 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1944 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1945 
1946 	/*
1947 	 * A temporary address is created only if this calculated Preferred
1948 	 * Lifetime is greater than REGEN_ADVANCE time units.
1949 	 */
1950 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
1951 		return (0);
1952 
1953 	/* XXX: scope zone ID? */
1954 
1955 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1956 
1957 	/* allocate ifaddr structure, link into chain, etc. */
1958 	updateflags = 0;
1959 	if (dad_delay)
1960 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1961 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1962 		return (error);
1963 
1964 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1965 	if (newia == NULL) {	/* XXX: can it happen? */
1966 		nd6log((LOG_ERR,
1967 		    "in6_tmpifadd: ifa update succeeded, but we got "
1968 		    "no ifaddr\n"));
1969 		return (EINVAL); /* XXX */
1970 	}
1971 	newia->ia6_ndpr = ia0->ia6_ndpr;
1972 	newia->ia6_ndpr->ndpr_refcnt++;
1973 
1974 	/*
1975 	 * A newly added address might affect the status of other addresses.
1976 	 * XXX: when the temporary address is generated with a new public
1977 	 * address, the onlink check is redundant.  However, it would be safe
1978 	 * to do the check explicitly everywhere a new address is generated,
1979 	 * and, in fact, we surely need the check when we create a new
1980 	 * temporary address due to deprecation of an old temporary address.
1981 	 */
1982 	pfxlist_onlink_check();
1983 
1984 	return (0);
1985 }
1986 
1987 static int
1988 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1989 {
1990 
1991 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
1992 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
1993 		nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
1994 		    "(%d) is greater than valid lifetime(%d)\n",
1995 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
1996 		return (EINVAL);
1997 	}
1998 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1999 		ndpr->ndpr_preferred = 0;
2000 	else
2001 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2002 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2003 		ndpr->ndpr_expire = 0;
2004 	else
2005 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2006 
2007 	return 0;
2008 }
2009 
2010 static void
2011 in6_init_address_ltimes(struct nd_prefix *new,
2012     struct in6_addrlifetime *lt6)
2013 {
2014 
2015 	/* Valid lifetime must not be updated unless explicitly specified. */
2016 	/* init ia6t_expire */
2017 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2018 		lt6->ia6t_expire = 0;
2019 	else {
2020 		lt6->ia6t_expire = time_second;
2021 		lt6->ia6t_expire += lt6->ia6t_vltime;
2022 	}
2023 
2024 	/* init ia6t_preferred */
2025 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2026 		lt6->ia6t_preferred = 0;
2027 	else {
2028 		lt6->ia6t_preferred = time_second;
2029 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2030 	}
2031 }
2032 
2033 /*
2034  * Delete all the routing table entries that use the specified gateway.
2035  * XXX: this function causes search through all entries of routing table, so
2036  * it shouldn't be called when acting as a router.
2037  */
2038 void
2039 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2040 {
2041 	int s = splsoftnet();
2042 
2043 	/* We'll care only link-local addresses */
2044 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2045 		splx(s);
2046 		return;
2047 	}
2048 
2049 	rt_walktree(AF_INET6, rt6_deleteroute, (void *)gateway);
2050 	splx(s);
2051 }
2052 
2053 static int
2054 rt6_deleteroute(struct rtentry *rt, void *arg)
2055 {
2056 #define SIN6(s)	((struct sockaddr_in6 *)s)
2057 	struct in6_addr *gate = (struct in6_addr *)arg;
2058 
2059 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2060 		return (0);
2061 
2062 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr))
2063 		return (0);
2064 
2065 	/*
2066 	 * Do not delete a static route.
2067 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2068 	 * 'cloned' bit instead?
2069 	 */
2070 	if ((rt->rt_flags & RTF_STATIC) != 0)
2071 		return (0);
2072 
2073 	/*
2074 	 * We delete only host route. This means, in particular, we don't
2075 	 * delete default route.
2076 	 */
2077 	if ((rt->rt_flags & RTF_HOST) == 0)
2078 		return (0);
2079 
2080 	return (rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
2081 	    rt_mask(rt), rt->rt_flags, 0));
2082 #undef SIN6
2083 }
2084 
2085 int
2086 nd6_setdefaultiface(int ifindex)
2087 {
2088 	int error = 0;
2089 
2090 	if (ifindex < 0 || if_indexlim <= ifindex)
2091 		return (EINVAL);
2092 	if (ifindex != 0 && !ifindex2ifnet[ifindex])
2093 		return (EINVAL);
2094 
2095 	if (nd6_defifindex != ifindex) {
2096 		nd6_defifindex = ifindex;
2097 		if (nd6_defifindex > 0) {
2098 			nd6_defifp = ifindex2ifnet[nd6_defifindex];
2099 		} else
2100 			nd6_defifp = NULL;
2101 
2102 		/*
2103 		 * Our current implementation assumes one-to-one maping between
2104 		 * interfaces and links, so it would be natural to use the
2105 		 * default interface as the default link.
2106 		 */
2107 		scope6_setdefault(nd6_defifp);
2108 	}
2109 
2110 	return (error);
2111 }
2112