xref: /freebsd/sys/netinet6/nd6_rtr.c (revision 6472ac3d8a86336899b6cfb789a4cd9897e3fab5)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/socket.h>
43 #include <sys/sockio.h>
44 #include <sys/time.h>
45 #include <sys/kernel.h>
46 #include <sys/lock.h>
47 #include <sys/errno.h>
48 #include <sys/rwlock.h>
49 #include <sys/syslog.h>
50 #include <sys/queue.h>
51 
52 #include <net/if.h>
53 #include <net/if_types.h>
54 #include <net/if_dl.h>
55 #include <net/route.h>
56 #include <net/radix.h>
57 #include <net/vnet.h>
58 
59 #include <netinet/in.h>
60 #include <net/if_llatbl.h>
61 #include <netinet6/in6_var.h>
62 #include <netinet6/in6_ifattach.h>
63 #include <netinet/ip6.h>
64 #include <netinet6/ip6_var.h>
65 #include <netinet6/nd6.h>
66 #include <netinet/icmp6.h>
67 #include <netinet6/scope6_var.h>
68 
69 static int rtpref(struct nd_defrouter *);
70 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
71 static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
72     struct mbuf *, int));
73 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *,	int);
74 static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
75 	struct nd_defrouter *));
76 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
77 static void pfxrtr_del(struct nd_pfxrouter *);
78 static struct nd_pfxrouter *find_pfxlist_reachable_router
79 (struct nd_prefix *);
80 static void defrouter_delreq(struct nd_defrouter *);
81 static void nd6_rtmsg(int, struct rtentry *);
82 
83 static int in6_init_prefix_ltimes(struct nd_prefix *);
84 static void in6_init_address_ltimes __P((struct nd_prefix *,
85 	struct in6_addrlifetime *));
86 
87 static int rt6_deleteroute(struct radix_node *, void *);
88 
89 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
90 #define	V_nd6_recalc_reachtm_interval	VNET(nd6_recalc_reachtm_interval)
91 
92 static VNET_DEFINE(struct ifnet *, nd6_defifp);
93 VNET_DEFINE(int, nd6_defifindex);
94 #define	V_nd6_defifp			VNET(nd6_defifp)
95 
96 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
97 
98 VNET_DEFINE(int, ip6_desync_factor);
99 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
100 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
101 
102 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
103 
104 /* RTPREF_MEDIUM has to be 0! */
105 #define RTPREF_HIGH	1
106 #define RTPREF_MEDIUM	0
107 #define RTPREF_LOW	(-1)
108 #define RTPREF_RESERVED	(-2)
109 #define RTPREF_INVALID	(-3)	/* internal */
110 
111 /*
112  * Receive Router Solicitation Message - just for routers.
113  * Router solicitation/advertisement is mostly managed by userland program
114  * (rtadvd) so here we have no function like nd6_ra_output().
115  *
116  * Based on RFC 2461
117  */
118 void
119 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
120 {
121 	struct ifnet *ifp = m->m_pkthdr.rcvif;
122 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
123 	struct nd_router_solicit *nd_rs;
124 	struct in6_addr saddr6 = ip6->ip6_src;
125 	char *lladdr = NULL;
126 	int lladdrlen = 0;
127 	union nd_opts ndopts;
128 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
129 
130 	/*
131 	 * Accept RS only when V_ip6_forwarding=1 and the interface has
132 	 * no ND6_IFF_ACCEPT_RTADV.
133 	 */
134 	if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
135 		goto freeit;
136 
137 	/* Sanity checks */
138 	if (ip6->ip6_hlim != 255) {
139 		nd6log((LOG_ERR,
140 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
141 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
142 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
143 		goto bad;
144 	}
145 
146 	/*
147 	 * Don't update the neighbor cache, if src = ::.
148 	 * This indicates that the src has no IP address assigned yet.
149 	 */
150 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
151 		goto freeit;
152 
153 #ifndef PULLDOWN_TEST
154 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
155 	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
156 #else
157 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
158 	if (nd_rs == NULL) {
159 		ICMP6STAT_INC(icp6s_tooshort);
160 		return;
161 	}
162 #endif
163 
164 	icmp6len -= sizeof(*nd_rs);
165 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
166 	if (nd6_options(&ndopts) < 0) {
167 		nd6log((LOG_INFO,
168 		    "nd6_rs_input: invalid ND option, ignored\n"));
169 		/* nd6_options have incremented stats */
170 		goto freeit;
171 	}
172 
173 	if (ndopts.nd_opts_src_lladdr) {
174 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
175 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
176 	}
177 
178 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
179 		nd6log((LOG_INFO,
180 		    "nd6_rs_input: lladdrlen mismatch for %s "
181 		    "(if %d, RS packet %d)\n",
182 		    ip6_sprintf(ip6bufs, &saddr6),
183 		    ifp->if_addrlen, lladdrlen - 2));
184 		goto bad;
185 	}
186 
187 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
188 
189  freeit:
190 	m_freem(m);
191 	return;
192 
193  bad:
194 	ICMP6STAT_INC(icp6s_badrs);
195 	m_freem(m);
196 }
197 
198 /*
199  * Receive Router Advertisement Message.
200  *
201  * Based on RFC 2461
202  * TODO: on-link bit on prefix information
203  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
204  */
205 void
206 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
207 {
208 	struct ifnet *ifp = m->m_pkthdr.rcvif;
209 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
210 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
211 	struct nd_router_advert *nd_ra;
212 	struct in6_addr saddr6 = ip6->ip6_src;
213 	int mcast = 0;
214 	union nd_opts ndopts;
215 	struct nd_defrouter *dr;
216 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
217 
218 	/*
219 	 * We only accept RAs only when the per-interface flag
220 	 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
221 	 */
222 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
223 		goto freeit;
224 
225 	if (ip6->ip6_hlim != 255) {
226 		nd6log((LOG_ERR,
227 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
228 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
229 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
230 		goto bad;
231 	}
232 
233 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
234 		nd6log((LOG_ERR,
235 		    "nd6_ra_input: src %s is not link-local\n",
236 		    ip6_sprintf(ip6bufs, &saddr6)));
237 		goto bad;
238 	}
239 
240 #ifndef PULLDOWN_TEST
241 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
242 	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
243 #else
244 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
245 	if (nd_ra == NULL) {
246 		ICMP6STAT_INC(icp6s_tooshort);
247 		return;
248 	}
249 #endif
250 
251 	icmp6len -= sizeof(*nd_ra);
252 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
253 	if (nd6_options(&ndopts) < 0) {
254 		nd6log((LOG_INFO,
255 		    "nd6_ra_input: invalid ND option, ignored\n"));
256 		/* nd6_options have incremented stats */
257 		goto freeit;
258 	}
259 
260     {
261 	struct nd_defrouter dr0;
262 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
263 
264 	/* remember if this is a multicasted advertisement */
265 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
266 		mcast = 1;
267 
268 	bzero(&dr0, sizeof(dr0));
269 	dr0.rtaddr = saddr6;
270 	dr0.flags  = nd_ra->nd_ra_flags_reserved;
271 	/*
272 	 * Effectively-disable routes from RA messages when
273 	 * ND6_IFF_NO_RADR enabled on the receiving interface or
274 	 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
275 	 */
276 	if (ndi->flags & ND6_IFF_NO_RADR)
277 		dr0.rtlifetime = 0;
278 	else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
279 		dr0.rtlifetime = 0;
280 	else
281 		dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
282 	dr0.expire = time_second + dr0.rtlifetime;
283 	dr0.ifp = ifp;
284 	/* unspecified or not? (RFC 2461 6.3.4) */
285 	if (advreachable) {
286 		advreachable = ntohl(advreachable);
287 		if (advreachable <= MAX_REACHABLE_TIME &&
288 		    ndi->basereachable != advreachable) {
289 			ndi->basereachable = advreachable;
290 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
291 			ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
292 		}
293 	}
294 	if (nd_ra->nd_ra_retransmit)
295 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
296 	if (nd_ra->nd_ra_curhoplimit)
297 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
298 	dr = defrtrlist_update(&dr0);
299     }
300 
301 	/*
302 	 * prefix
303 	 */
304 	if (ndopts.nd_opts_pi) {
305 		struct nd_opt_hdr *pt;
306 		struct nd_opt_prefix_info *pi = NULL;
307 		struct nd_prefixctl pr;
308 
309 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
310 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
311 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
312 						(pt->nd_opt_len << 3))) {
313 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
314 				continue;
315 			pi = (struct nd_opt_prefix_info *)pt;
316 
317 			if (pi->nd_opt_pi_len != 4) {
318 				nd6log((LOG_INFO,
319 				    "nd6_ra_input: invalid option "
320 				    "len %d for prefix information option, "
321 				    "ignored\n", pi->nd_opt_pi_len));
322 				continue;
323 			}
324 
325 			if (128 < pi->nd_opt_pi_prefix_len) {
326 				nd6log((LOG_INFO,
327 				    "nd6_ra_input: invalid prefix "
328 				    "len %d for prefix information option, "
329 				    "ignored\n", pi->nd_opt_pi_prefix_len));
330 				continue;
331 			}
332 
333 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
334 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
335 				nd6log((LOG_INFO,
336 				    "nd6_ra_input: invalid prefix "
337 				    "%s, ignored\n",
338 				    ip6_sprintf(ip6bufs,
339 					&pi->nd_opt_pi_prefix)));
340 				continue;
341 			}
342 
343 			bzero(&pr, sizeof(pr));
344 			pr.ndpr_prefix.sin6_family = AF_INET6;
345 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
346 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
347 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
348 
349 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
350 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
351 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
352 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
353 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
354 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
355 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
356 			(void)prelist_update(&pr, dr, m, mcast);
357 		}
358 	}
359 
360 	/*
361 	 * MTU
362 	 */
363 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
364 		u_long mtu;
365 		u_long maxmtu;
366 
367 		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
368 
369 		/* lower bound */
370 		if (mtu < IPV6_MMTU) {
371 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
372 			    "mtu=%lu sent from %s, ignoring\n",
373 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
374 			goto skip;
375 		}
376 
377 		/* upper bound */
378 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
379 		    ? ndi->maxmtu : ifp->if_mtu;
380 		if (mtu <= maxmtu) {
381 			int change = (ndi->linkmtu != mtu);
382 
383 			ndi->linkmtu = mtu;
384 			if (change) /* in6_maxmtu may change */
385 				in6_setmaxmtu();
386 		} else {
387 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
388 			    "mtu=%lu sent from %s; "
389 			    "exceeds maxmtu %lu, ignoring\n",
390 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
391 		}
392 	}
393 
394  skip:
395 
396 	/*
397 	 * Source link layer address
398 	 */
399     {
400 	char *lladdr = NULL;
401 	int lladdrlen = 0;
402 
403 	if (ndopts.nd_opts_src_lladdr) {
404 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
405 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
406 	}
407 
408 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
409 		nd6log((LOG_INFO,
410 		    "nd6_ra_input: lladdrlen mismatch for %s "
411 		    "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
412 		    ifp->if_addrlen, lladdrlen - 2));
413 		goto bad;
414 	}
415 
416 	nd6_cache_lladdr(ifp, &saddr6, lladdr,
417 	    lladdrlen, ND_ROUTER_ADVERT, 0);
418 
419 	/*
420 	 * Installing a link-layer address might change the state of the
421 	 * router's neighbor cache, which might also affect our on-link
422 	 * detection of adveritsed prefixes.
423 	 */
424 	pfxlist_onlink_check();
425     }
426 
427  freeit:
428 	m_freem(m);
429 	return;
430 
431  bad:
432 	ICMP6STAT_INC(icp6s_badra);
433 	m_freem(m);
434 }
435 
436 /*
437  * default router list proccessing sub routines
438  */
439 
440 /* tell the change to user processes watching the routing socket. */
441 static void
442 nd6_rtmsg(int cmd, struct rtentry *rt)
443 {
444 	struct rt_addrinfo info;
445 	struct ifnet *ifp;
446 	struct ifaddr *ifa;
447 
448 	bzero((caddr_t)&info, sizeof(info));
449 	info.rti_info[RTAX_DST] = rt_key(rt);
450 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
451 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
452 	ifp = rt->rt_ifp;
453 	if (ifp != NULL) {
454 		IF_ADDR_LOCK(ifp);
455 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
456 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
457 		ifa_ref(ifa);
458 		IF_ADDR_UNLOCK(ifp);
459 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
460 	} else
461 		ifa = NULL;
462 
463 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
464 	if (ifa != NULL)
465 		ifa_free(ifa);
466 }
467 
468 void
469 defrouter_addreq(struct nd_defrouter *new)
470 {
471 	struct sockaddr_in6 def, mask, gate;
472 	struct rtentry *newrt = NULL;
473 	int s;
474 	int error;
475 
476 	bzero(&def, sizeof(def));
477 	bzero(&mask, sizeof(mask));
478 	bzero(&gate, sizeof(gate));
479 
480 	def.sin6_len = mask.sin6_len = gate.sin6_len =
481 	    sizeof(struct sockaddr_in6);
482 	def.sin6_family = gate.sin6_family = AF_INET6;
483 	gate.sin6_addr = new->rtaddr;
484 
485 	s = splnet();
486 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
487 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
488 	    RTF_GATEWAY, &newrt);
489 	if (newrt) {
490 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
491 		RTFREE(newrt);
492 	}
493 	if (error == 0)
494 		new->installed = 1;
495 	splx(s);
496 	return;
497 }
498 
499 struct nd_defrouter *
500 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
501 {
502 	struct nd_defrouter *dr;
503 
504 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
505 	     dr = TAILQ_NEXT(dr, dr_entry)) {
506 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
507 			return (dr);
508 	}
509 
510 	return (NULL);		/* search failed */
511 }
512 
513 /*
514  * Remove the default route for a given router.
515  * This is just a subroutine function for defrouter_select(), and should
516  * not be called from anywhere else.
517  */
518 static void
519 defrouter_delreq(struct nd_defrouter *dr)
520 {
521 	struct sockaddr_in6 def, mask, gate;
522 	struct rtentry *oldrt = NULL;
523 
524 	bzero(&def, sizeof(def));
525 	bzero(&mask, sizeof(mask));
526 	bzero(&gate, sizeof(gate));
527 
528 	def.sin6_len = mask.sin6_len = gate.sin6_len =
529 	    sizeof(struct sockaddr_in6);
530 	def.sin6_family = gate.sin6_family = AF_INET6;
531 	gate.sin6_addr = dr->rtaddr;
532 
533 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
534 	    (struct sockaddr *)&gate,
535 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
536 	if (oldrt) {
537 		nd6_rtmsg(RTM_DELETE, oldrt);
538 		RTFREE(oldrt);
539 	}
540 
541 	dr->installed = 0;
542 }
543 
544 /*
545  * remove all default routes from default router list
546  */
547 void
548 defrouter_reset(void)
549 {
550 	struct nd_defrouter *dr;
551 
552 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
553 	     dr = TAILQ_NEXT(dr, dr_entry))
554 		defrouter_delreq(dr);
555 
556 	/*
557 	 * XXX should we also nuke any default routers in the kernel, by
558 	 * going through them by rtalloc1()?
559 	 */
560 }
561 
562 void
563 defrtrlist_del(struct nd_defrouter *dr)
564 {
565 	struct nd_defrouter *deldr = NULL;
566 	struct nd_prefix *pr;
567 
568 	/*
569 	 * Flush all the routing table entries that use the router
570 	 * as a next hop.
571 	 */
572 	if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
573 		rt6_flush(&dr->rtaddr, dr->ifp);
574 
575 	if (dr->installed) {
576 		deldr = dr;
577 		defrouter_delreq(dr);
578 	}
579 	TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
580 
581 	/*
582 	 * Also delete all the pointers to the router in each prefix lists.
583 	 */
584 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
585 		struct nd_pfxrouter *pfxrtr;
586 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
587 			pfxrtr_del(pfxrtr);
588 	}
589 	pfxlist_onlink_check();
590 
591 	/*
592 	 * If the router is the primary one, choose a new one.
593 	 * Note that defrouter_select() will remove the current gateway
594 	 * from the routing table.
595 	 */
596 	if (deldr)
597 		defrouter_select();
598 
599 	free(dr, M_IP6NDP);
600 }
601 
602 /*
603  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
604  * draft-ietf-ipngwg-router-selection:
605  * 1) Routers that are reachable or probably reachable should be preferred.
606  *    If we have more than one (probably) reachable router, prefer ones
607  *    with the highest router preference.
608  * 2) When no routers on the list are known to be reachable or
609  *    probably reachable, routers SHOULD be selected in a round-robin
610  *    fashion, regardless of router preference values.
611  * 3) If the Default Router List is empty, assume that all
612  *    destinations are on-link.
613  *
614  * We assume nd_defrouter is sorted by router preference value.
615  * Since the code below covers both with and without router preference cases,
616  * we do not need to classify the cases by ifdef.
617  *
618  * At this moment, we do not try to install more than one default router,
619  * even when the multipath routing is available, because we're not sure about
620  * the benefits for stub hosts comparing to the risk of making the code
621  * complicated and the possibility of introducing bugs.
622  */
623 void
624 defrouter_select(void)
625 {
626 	int s = splnet();
627 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
628 	struct llentry *ln = NULL;
629 
630 	/*
631 	 * Let's handle easy case (3) first:
632 	 * If default router list is empty, there's nothing to be done.
633 	 */
634 	if (!TAILQ_FIRST(&V_nd_defrouter)) {
635 		splx(s);
636 		return;
637 	}
638 
639 	/*
640 	 * Search for a (probably) reachable router from the list.
641 	 * We just pick up the first reachable one (if any), assuming that
642 	 * the ordering rule of the list described in defrtrlist_update().
643 	 */
644 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
645 	     dr = TAILQ_NEXT(dr, dr_entry)) {
646 		IF_AFDATA_LOCK(dr->ifp);
647 		if (selected_dr == NULL &&
648 		    (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
649 		    ND6_IS_LLINFO_PROBREACH(ln)) {
650 			selected_dr = dr;
651 		}
652 		IF_AFDATA_UNLOCK(dr->ifp);
653 		if (ln != NULL) {
654 			LLE_RUNLOCK(ln);
655 			ln = NULL;
656 		}
657 
658 		if (dr->installed && installed_dr == NULL)
659 			installed_dr = dr;
660 		else if (dr->installed && installed_dr) {
661 			/* this should not happen.  warn for diagnosis. */
662 			log(LOG_ERR, "defrouter_select: more than one router"
663 			    " is installed\n");
664 		}
665 	}
666 	/*
667 	 * If none of the default routers was found to be reachable,
668 	 * round-robin the list regardless of preference.
669 	 * Otherwise, if we have an installed router, check if the selected
670 	 * (reachable) router should really be preferred to the installed one.
671 	 * We only prefer the new router when the old one is not reachable
672 	 * or when the new one has a really higher preference value.
673 	 */
674 	if (selected_dr == NULL) {
675 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
676 			selected_dr = TAILQ_FIRST(&V_nd_defrouter);
677 		else
678 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
679 	} else if (installed_dr) {
680 		IF_AFDATA_LOCK(installed_dr->ifp);
681 		if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
682 		    ND6_IS_LLINFO_PROBREACH(ln) &&
683 		    rtpref(selected_dr) <= rtpref(installed_dr)) {
684 			selected_dr = installed_dr;
685 		}
686 		IF_AFDATA_UNLOCK(installed_dr->ifp);
687 		if (ln != NULL)
688 			LLE_RUNLOCK(ln);
689 	}
690 
691 	/*
692 	 * If the selected router is different than the installed one,
693 	 * remove the installed router and install the selected one.
694 	 * Note that the selected router is never NULL here.
695 	 */
696 	if (installed_dr != selected_dr) {
697 		if (installed_dr)
698 			defrouter_delreq(installed_dr);
699 		defrouter_addreq(selected_dr);
700 	}
701 
702 	splx(s);
703 	return;
704 }
705 
706 /*
707  * for default router selection
708  * regards router-preference field as a 2-bit signed integer
709  */
710 static int
711 rtpref(struct nd_defrouter *dr)
712 {
713 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
714 	case ND_RA_FLAG_RTPREF_HIGH:
715 		return (RTPREF_HIGH);
716 	case ND_RA_FLAG_RTPREF_MEDIUM:
717 	case ND_RA_FLAG_RTPREF_RSV:
718 		return (RTPREF_MEDIUM);
719 	case ND_RA_FLAG_RTPREF_LOW:
720 		return (RTPREF_LOW);
721 	default:
722 		/*
723 		 * This case should never happen.  If it did, it would mean a
724 		 * serious bug of kernel internal.  We thus always bark here.
725 		 * Or, can we even panic?
726 		 */
727 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
728 		return (RTPREF_INVALID);
729 	}
730 	/* NOTREACHED */
731 }
732 
733 static struct nd_defrouter *
734 defrtrlist_update(struct nd_defrouter *new)
735 {
736 	struct nd_defrouter *dr, *n;
737 	int s = splnet();
738 
739 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
740 		/* entry exists */
741 		if (new->rtlifetime == 0) {
742 			defrtrlist_del(dr);
743 			dr = NULL;
744 		} else {
745 			int oldpref = rtpref(dr);
746 
747 			/* override */
748 			dr->flags = new->flags; /* xxx flag check */
749 			dr->rtlifetime = new->rtlifetime;
750 			dr->expire = new->expire;
751 
752 			/*
753 			 * If the preference does not change, there's no need
754 			 * to sort the entries.
755 			 */
756 			if (rtpref(new) == oldpref) {
757 				splx(s);
758 				return (dr);
759 			}
760 
761 			/*
762 			 * preferred router may be changed, so relocate
763 			 * this router.
764 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
765 			 * However, since defrtrlist_del() has many side
766 			 * effects, we intentionally do so here.
767 			 * defrouter_select() below will handle routing
768 			 * changes later.
769 			 */
770 			TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
771 			n = dr;
772 			goto insert;
773 		}
774 		splx(s);
775 		return (dr);
776 	}
777 
778 	/* entry does not exist */
779 	if (new->rtlifetime == 0) {
780 		splx(s);
781 		return (NULL);
782 	}
783 
784 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
785 	if (n == NULL) {
786 		splx(s);
787 		return (NULL);
788 	}
789 	bzero(n, sizeof(*n));
790 	*n = *new;
791 
792 insert:
793 	/*
794 	 * Insert the new router in the Default Router List;
795 	 * The Default Router List should be in the descending order
796 	 * of router-preferece.  Routers with the same preference are
797 	 * sorted in the arriving time order.
798 	 */
799 
800 	/* insert at the end of the group */
801 	for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
802 	     dr = TAILQ_NEXT(dr, dr_entry)) {
803 		if (rtpref(n) > rtpref(dr))
804 			break;
805 	}
806 	if (dr)
807 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
808 	else
809 		TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
810 
811 	defrouter_select();
812 
813 	splx(s);
814 
815 	return (n);
816 }
817 
818 static struct nd_pfxrouter *
819 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
820 {
821 	struct nd_pfxrouter *search;
822 
823 	for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
824 		if (search->router == dr)
825 			break;
826 	}
827 
828 	return (search);
829 }
830 
831 static void
832 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
833 {
834 	struct nd_pfxrouter *new;
835 
836 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
837 	if (new == NULL)
838 		return;
839 	bzero(new, sizeof(*new));
840 	new->router = dr;
841 
842 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
843 
844 	pfxlist_onlink_check();
845 }
846 
847 static void
848 pfxrtr_del(struct nd_pfxrouter *pfr)
849 {
850 	LIST_REMOVE(pfr, pfr_entry);
851 	free(pfr, M_IP6NDP);
852 }
853 
854 struct nd_prefix *
855 nd6_prefix_lookup(struct nd_prefixctl *key)
856 {
857 	struct nd_prefix *search;
858 
859 	for (search = V_nd_prefix.lh_first;
860 	    search; search = search->ndpr_next) {
861 		if (key->ndpr_ifp == search->ndpr_ifp &&
862 		    key->ndpr_plen == search->ndpr_plen &&
863 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
864 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
865 			break;
866 		}
867 	}
868 
869 	return (search);
870 }
871 
872 int
873 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
874     struct nd_prefix **newp)
875 {
876 	struct nd_prefix *new = NULL;
877 	int error = 0;
878 	int i, s;
879 	char ip6buf[INET6_ADDRSTRLEN];
880 
881 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
882 	if (new == NULL)
883 		return(ENOMEM);
884 	bzero(new, sizeof(*new));
885 	new->ndpr_ifp = pr->ndpr_ifp;
886 	new->ndpr_prefix = pr->ndpr_prefix;
887 	new->ndpr_plen = pr->ndpr_plen;
888 	new->ndpr_vltime = pr->ndpr_vltime;
889 	new->ndpr_pltime = pr->ndpr_pltime;
890 	new->ndpr_flags = pr->ndpr_flags;
891 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
892 		free(new, M_IP6NDP);
893 		return(error);
894 	}
895 	new->ndpr_lastupdate = time_second;
896 	if (newp != NULL)
897 		*newp = new;
898 
899 	/* initialization */
900 	LIST_INIT(&new->ndpr_advrtrs);
901 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
902 	/* make prefix in the canonical form */
903 	for (i = 0; i < 4; i++)
904 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
905 		    new->ndpr_mask.s6_addr32[i];
906 
907 	s = splnet();
908 	/* link ndpr_entry to nd_prefix list */
909 	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
910 	splx(s);
911 
912 	/* ND_OPT_PI_FLAG_ONLINK processing */
913 	if (new->ndpr_raf_onlink) {
914 		int e;
915 
916 		if ((e = nd6_prefix_onlink(new)) != 0) {
917 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
918 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
919 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
920 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
921 			/* proceed anyway. XXX: is it correct? */
922 		}
923 	}
924 
925 	if (dr)
926 		pfxrtr_add(new, dr);
927 
928 	return 0;
929 }
930 
931 void
932 prelist_remove(struct nd_prefix *pr)
933 {
934 	struct nd_pfxrouter *pfr, *next;
935 	int e, s;
936 	char ip6buf[INET6_ADDRSTRLEN];
937 
938 	/* make sure to invalidate the prefix until it is really freed. */
939 	pr->ndpr_vltime = 0;
940 	pr->ndpr_pltime = 0;
941 
942 	/*
943 	 * Though these flags are now meaningless, we'd rather keep the value
944 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
945 	 * when executing "ndp -p".
946 	 */
947 
948 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
949 	    (e = nd6_prefix_offlink(pr)) != 0) {
950 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
951 		    "on %s, errno=%d\n",
952 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
953 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
954 		/* what should we do? */
955 	}
956 
957 	if (pr->ndpr_refcnt > 0)
958 		return;		/* notice here? */
959 
960 	s = splnet();
961 
962 	/* unlink ndpr_entry from nd_prefix list */
963 	LIST_REMOVE(pr, ndpr_entry);
964 
965 	/* free list of routers that adversed the prefix */
966 	for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
967 		next = pfr->pfr_next;
968 
969 		free(pfr, M_IP6NDP);
970 	}
971 	splx(s);
972 
973 	free(pr, M_IP6NDP);
974 
975 	pfxlist_onlink_check();
976 }
977 
978 /*
979  * dr - may be NULL
980  */
981 
982 static int
983 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
984     struct mbuf *m, int mcast)
985 {
986 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
987 	struct ifaddr *ifa;
988 	struct ifnet *ifp = new->ndpr_ifp;
989 	struct nd_prefix *pr;
990 	int s = splnet();
991 	int error = 0;
992 	int newprefix = 0;
993 	int auth;
994 	struct in6_addrlifetime lt6_tmp;
995 	char ip6buf[INET6_ADDRSTRLEN];
996 
997 	auth = 0;
998 	if (m) {
999 		/*
1000 		 * Authenticity for NA consists authentication for
1001 		 * both IP header and IP datagrams, doesn't it ?
1002 		 */
1003 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1004 		auth = ((m->m_flags & M_AUTHIPHDR) &&
1005 		    (m->m_flags & M_AUTHIPDGM));
1006 #endif
1007 	}
1008 
1009 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1010 		/*
1011 		 * nd6_prefix_lookup() ensures that pr and new have the same
1012 		 * prefix on a same interface.
1013 		 */
1014 
1015 		/*
1016 		 * Update prefix information.  Note that the on-link (L) bit
1017 		 * and the autonomous (A) bit should NOT be changed from 1
1018 		 * to 0.
1019 		 */
1020 		if (new->ndpr_raf_onlink == 1)
1021 			pr->ndpr_raf_onlink = 1;
1022 		if (new->ndpr_raf_auto == 1)
1023 			pr->ndpr_raf_auto = 1;
1024 		if (new->ndpr_raf_onlink) {
1025 			pr->ndpr_vltime = new->ndpr_vltime;
1026 			pr->ndpr_pltime = new->ndpr_pltime;
1027 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1028 			pr->ndpr_lastupdate = time_second;
1029 		}
1030 
1031 		if (new->ndpr_raf_onlink &&
1032 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1033 			int e;
1034 
1035 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1036 				nd6log((LOG_ERR,
1037 				    "prelist_update: failed to make "
1038 				    "the prefix %s/%d on-link on %s "
1039 				    "(errno=%d)\n",
1040 				    ip6_sprintf(ip6buf,
1041 					    &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 		error = nd6_prelist_add(new, dr, &newpr);
1060 		if (error != 0 || newpr == NULL) {
1061 			nd6log((LOG_NOTICE, "prelist_update: "
1062 			    "nd6_prelist_add failed for %s/%d on %s "
1063 			    "errno=%d, returnpr=%p\n",
1064 			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1065 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1066 			    error, newpr));
1067 			goto end; /* we should just give up in this case. */
1068 		}
1069 
1070 		/*
1071 		 * XXX: from the ND point of view, we can ignore a prefix
1072 		 * with the on-link bit being zero.  However, we need a
1073 		 * prefix structure for references from autoconfigured
1074 		 * addresses.  Thus, we explicitly make sure that the prefix
1075 		 * itself expires now.
1076 		 */
1077 		if (newpr->ndpr_raf_onlink == 0) {
1078 			newpr->ndpr_vltime = 0;
1079 			newpr->ndpr_pltime = 0;
1080 			in6_init_prefix_ltimes(newpr);
1081 		}
1082 
1083 		pr = newpr;
1084 	}
1085 
1086 	/*
1087 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1088 	 * Note that pr must be non NULL at this point.
1089 	 */
1090 
1091 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1092 	if (!new->ndpr_raf_auto)
1093 		goto end;
1094 
1095 	/*
1096 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1097 	 * nd6_ra_input.
1098 	 */
1099 
1100 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1101 	if (new->ndpr_pltime > new->ndpr_vltime) {
1102 		error = EINVAL;	/* XXX: won't be used */
1103 		goto end;
1104 	}
1105 
1106 	/*
1107 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1108 	 * an address configured by stateless autoconfiguration already in the
1109 	 * list of addresses associated with the interface, and the Valid
1110 	 * Lifetime is not 0, form an address.  We first check if we have
1111 	 * a matching prefix.
1112 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1113 	 * consider autoconfigured addresses while RFC2462 simply said
1114 	 * "address".
1115 	 */
1116 	IF_ADDR_LOCK(ifp);
1117 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1118 		struct in6_ifaddr *ifa6;
1119 		u_int32_t remaininglifetime;
1120 
1121 		if (ifa->ifa_addr->sa_family != AF_INET6)
1122 			continue;
1123 
1124 		ifa6 = (struct in6_ifaddr *)ifa;
1125 
1126 		/*
1127 		 * We only consider autoconfigured addresses as per rfc2462bis.
1128 		 */
1129 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1130 			continue;
1131 
1132 		/*
1133 		 * Spec is not clear here, but I believe we should concentrate
1134 		 * on unicast (i.e. not anycast) addresses.
1135 		 * XXX: other ia6_flags? detached or duplicated?
1136 		 */
1137 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1138 			continue;
1139 
1140 		/*
1141 		 * Ignore the address if it is not associated with a prefix
1142 		 * or is associated with a prefix that is different from this
1143 		 * one.  (pr is never NULL here)
1144 		 */
1145 		if (ifa6->ia6_ndpr != pr)
1146 			continue;
1147 
1148 		if (ia6_match == NULL) /* remember the first one */
1149 			ia6_match = ifa6;
1150 
1151 		/*
1152 		 * An already autoconfigured address matched.  Now that we
1153 		 * are sure there is at least one matched address, we can
1154 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1155 		 * "two hours" rule and the privacy extension.
1156 		 * We apply some clarifications in rfc2462bis:
1157 		 * - use remaininglifetime instead of storedlifetime as a
1158 		 *   variable name
1159 		 * - remove the dead code in the "two-hour" rule
1160 		 */
1161 #define TWOHOUR		(120*60)
1162 		lt6_tmp = ifa6->ia6_lifetime;
1163 
1164 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1165 			remaininglifetime = ND6_INFINITE_LIFETIME;
1166 		else if (time_second - ifa6->ia6_updatetime >
1167 			 lt6_tmp.ia6t_vltime) {
1168 			/*
1169 			 * The case of "invalid" address.  We should usually
1170 			 * not see this case.
1171 			 */
1172 			remaininglifetime = 0;
1173 		} else
1174 			remaininglifetime = lt6_tmp.ia6t_vltime -
1175 			    (time_second - ifa6->ia6_updatetime);
1176 
1177 		/* when not updating, keep the current stored lifetime. */
1178 		lt6_tmp.ia6t_vltime = remaininglifetime;
1179 
1180 		if (TWOHOUR < new->ndpr_vltime ||
1181 		    remaininglifetime < new->ndpr_vltime) {
1182 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1183 		} else if (remaininglifetime <= TWOHOUR) {
1184 			if (auth) {
1185 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1186 			}
1187 		} else {
1188 			/*
1189 			 * new->ndpr_vltime <= TWOHOUR &&
1190 			 * TWOHOUR < remaininglifetime
1191 			 */
1192 			lt6_tmp.ia6t_vltime = TWOHOUR;
1193 		}
1194 
1195 		/* The 2 hour rule is not imposed for preferred lifetime. */
1196 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1197 
1198 		in6_init_address_ltimes(pr, &lt6_tmp);
1199 
1200 		/*
1201 		 * We need to treat lifetimes for temporary addresses
1202 		 * differently, according to
1203 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1204 		 * we only update the lifetimes when they are in the maximum
1205 		 * intervals.
1206 		 */
1207 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1208 			u_int32_t maxvltime, maxpltime;
1209 
1210 			if (V_ip6_temp_valid_lifetime >
1211 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1212 			    V_ip6_desync_factor)) {
1213 				maxvltime = V_ip6_temp_valid_lifetime -
1214 				    (time_second - ifa6->ia6_createtime) -
1215 				    V_ip6_desync_factor;
1216 			} else
1217 				maxvltime = 0;
1218 			if (V_ip6_temp_preferred_lifetime >
1219 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1220 			    V_ip6_desync_factor)) {
1221 				maxpltime = V_ip6_temp_preferred_lifetime -
1222 				    (time_second - ifa6->ia6_createtime) -
1223 				    V_ip6_desync_factor;
1224 			} else
1225 				maxpltime = 0;
1226 
1227 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1228 			    lt6_tmp.ia6t_vltime > maxvltime) {
1229 				lt6_tmp.ia6t_vltime = maxvltime;
1230 			}
1231 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1232 			    lt6_tmp.ia6t_pltime > maxpltime) {
1233 				lt6_tmp.ia6t_pltime = maxpltime;
1234 			}
1235 		}
1236 		ifa6->ia6_lifetime = lt6_tmp;
1237 		ifa6->ia6_updatetime = time_second;
1238 	}
1239 	IF_ADDR_UNLOCK(ifp);
1240 	if (ia6_match == NULL && new->ndpr_vltime) {
1241 		int ifidlen;
1242 
1243 		/*
1244 		 * 5.5.3 (d) (continued)
1245 		 * No address matched and the valid lifetime is non-zero.
1246 		 * Create a new address.
1247 		 */
1248 
1249 		/*
1250 		 * Prefix Length check:
1251 		 * If the sum of the prefix length and interface identifier
1252 		 * length does not equal 128 bits, the Prefix Information
1253 		 * option MUST be ignored.  The length of the interface
1254 		 * identifier is defined in a separate link-type specific
1255 		 * document.
1256 		 */
1257 		ifidlen = in6_if2idlen(ifp);
1258 		if (ifidlen < 0) {
1259 			/* this should not happen, so we always log it. */
1260 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1261 			    if_name(ifp));
1262 			goto end;
1263 		}
1264 		if (ifidlen + pr->ndpr_plen != 128) {
1265 			nd6log((LOG_INFO,
1266 			    "prelist_update: invalid prefixlen "
1267 			    "%d for %s, ignored\n",
1268 			    pr->ndpr_plen, if_name(ifp)));
1269 			goto end;
1270 		}
1271 
1272 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1273 			/*
1274 			 * note that we should use pr (not new) for reference.
1275 			 */
1276 			pr->ndpr_refcnt++;
1277 			ia6->ia6_ndpr = pr;
1278 
1279 			/*
1280 			 * RFC 3041 3.3 (2).
1281 			 * When a new public address is created as described
1282 			 * in RFC2462, also create a new temporary address.
1283 			 *
1284 			 * RFC 3041 3.5.
1285 			 * When an interface connects to a new link, a new
1286 			 * randomized interface identifier should be generated
1287 			 * immediately together with a new set of temporary
1288 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1289 			 * in6_tmpifadd().
1290 			 */
1291 			if (V_ip6_use_tempaddr) {
1292 				int e;
1293 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1294 					nd6log((LOG_NOTICE, "prelist_update: "
1295 					    "failed to create a temporary "
1296 					    "address, errno=%d\n",
1297 					    e));
1298 				}
1299 			}
1300 			ifa_free(&ia6->ia_ifa);
1301 
1302 			/*
1303 			 * A newly added address might affect the status
1304 			 * of other addresses, so we check and update it.
1305 			 * XXX: what if address duplication happens?
1306 			 */
1307 			pfxlist_onlink_check();
1308 		} else {
1309 			/* just set an error. do not bark here. */
1310 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1311 		}
1312 	}
1313 
1314  end:
1315 	splx(s);
1316 	return error;
1317 }
1318 
1319 /*
1320  * A supplement function used in the on-link detection below;
1321  * detect if a given prefix has a (probably) reachable advertising router.
1322  * XXX: lengthy function name...
1323  */
1324 static struct nd_pfxrouter *
1325 find_pfxlist_reachable_router(struct nd_prefix *pr)
1326 {
1327 	struct nd_pfxrouter *pfxrtr;
1328 	struct llentry *ln;
1329 	int canreach;
1330 
1331 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr != NULL;
1332 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1333 		IF_AFDATA_LOCK(pfxrtr->router->ifp);
1334 		ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1335 		IF_AFDATA_UNLOCK(pfxrtr->router->ifp);
1336 		if (ln == NULL)
1337 			continue;
1338 		canreach = ND6_IS_LLINFO_PROBREACH(ln);
1339 		LLE_RUNLOCK(ln);
1340 		if (canreach)
1341 			break;
1342 	}
1343 	return (pfxrtr);
1344 }
1345 
1346 /*
1347  * Check if each prefix in the prefix list has at least one available router
1348  * that advertised the prefix (a router is "available" if its neighbor cache
1349  * entry is reachable or probably reachable).
1350  * If the check fails, the prefix may be off-link, because, for example,
1351  * we have moved from the network but the lifetime of the prefix has not
1352  * expired yet.  So we should not use the prefix if there is another prefix
1353  * that has an available router.
1354  * But, if there is no prefix that has an available router, we still regards
1355  * all the prefixes as on-link.  This is because we can't tell if all the
1356  * routers are simply dead or if we really moved from the network and there
1357  * is no router around us.
1358  */
1359 void
1360 pfxlist_onlink_check()
1361 {
1362 	struct nd_prefix *pr;
1363 	struct in6_ifaddr *ifa;
1364 	struct nd_defrouter *dr;
1365 	struct nd_pfxrouter *pfxrtr = NULL;
1366 
1367 	/*
1368 	 * Check if there is a prefix that has a reachable advertising
1369 	 * router.
1370 	 */
1371 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1372 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1373 			break;
1374 	}
1375 
1376 	/*
1377 	 * If we have no such prefix, check whether we still have a router
1378 	 * that does not advertise any prefixes.
1379 	 */
1380 	if (pr == NULL) {
1381 		for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
1382 		    dr = TAILQ_NEXT(dr, dr_entry)) {
1383 			struct nd_prefix *pr0;
1384 
1385 			for (pr0 = V_nd_prefix.lh_first; pr0;
1386 			    pr0 = pr0->ndpr_next) {
1387 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1388 					break;
1389 			}
1390 			if (pfxrtr != NULL)
1391 				break;
1392 		}
1393 	}
1394 	if (pr != NULL || (TAILQ_FIRST(&V_nd_defrouter) && pfxrtr == NULL)) {
1395 		/*
1396 		 * There is at least one prefix that has a reachable router,
1397 		 * or at least a router which probably does not advertise
1398 		 * any prefixes.  The latter would be the case when we move
1399 		 * to a new link where we have a router that does not provide
1400 		 * prefixes and we configure an address by hand.
1401 		 * Detach prefixes which have no reachable advertising
1402 		 * router, and attach other prefixes.
1403 		 */
1404 		for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1405 			/* XXX: a link-local prefix should never be detached */
1406 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1407 				continue;
1408 
1409 			/*
1410 			 * we aren't interested in prefixes without the L bit
1411 			 * set.
1412 			 */
1413 			if (pr->ndpr_raf_onlink == 0)
1414 				continue;
1415 
1416 			if (pr->ndpr_raf_auto == 0)
1417 				continue;
1418 
1419 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1420 			    find_pfxlist_reachable_router(pr) == NULL)
1421 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1422 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1423 			    find_pfxlist_reachable_router(pr) != 0)
1424 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1425 		}
1426 	} else {
1427 		/* there is no prefix that has a reachable router */
1428 		for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
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_raf_auto == 0)
1436 				continue;
1437 
1438 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1439 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1440 		}
1441 	}
1442 
1443 	/*
1444 	 * Remove each interface route associated with a (just) detached
1445 	 * prefix, and reinstall the interface route for a (just) attached
1446 	 * prefix.  Note that all attempt of reinstallation does not
1447 	 * necessarily success, when a same prefix is shared among multiple
1448 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1449 	 * so we don't have to care about them.
1450 	 */
1451 	for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1452 		int e;
1453 		char ip6buf[INET6_ADDRSTRLEN];
1454 
1455 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1456 			continue;
1457 
1458 		if (pr->ndpr_raf_onlink == 0)
1459 			continue;
1460 
1461 		if (pr->ndpr_raf_auto == 0)
1462 			continue;
1463 
1464 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1465 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1466 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1467 				nd6log((LOG_ERR,
1468 				    "pfxlist_onlink_check: failed to "
1469 				    "make %s/%d offlink, errno=%d\n",
1470 				    ip6_sprintf(ip6buf,
1471 					    &pr->ndpr_prefix.sin6_addr),
1472 					    pr->ndpr_plen, e));
1473 			}
1474 		}
1475 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1476 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1477 		    pr->ndpr_raf_onlink) {
1478 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1479 				nd6log((LOG_ERR,
1480 				    "pfxlist_onlink_check: failed to "
1481 				    "make %s/%d onlink, errno=%d\n",
1482 				    ip6_sprintf(ip6buf,
1483 					    &pr->ndpr_prefix.sin6_addr),
1484 					    pr->ndpr_plen, e));
1485 			}
1486 		}
1487 	}
1488 
1489 	/*
1490 	 * Changes on the prefix status might affect address status as well.
1491 	 * Make sure that all addresses derived from an attached prefix are
1492 	 * attached, and that all addresses derived from a detached prefix are
1493 	 * detached.  Note, however, that a manually configured address should
1494 	 * always be attached.
1495 	 * The precise detection logic is same as the one for prefixes.
1496 	 *
1497 	 * XXXRW: in6_ifaddrhead locking.
1498 	 */
1499 	TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1500 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1501 			continue;
1502 
1503 		if (ifa->ia6_ndpr == NULL) {
1504 			/*
1505 			 * This can happen when we first configure the address
1506 			 * (i.e. the address exists, but the prefix does not).
1507 			 * XXX: complicated relationships...
1508 			 */
1509 			continue;
1510 		}
1511 
1512 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1513 			break;
1514 	}
1515 	if (ifa) {
1516 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1517 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1518 				continue;
1519 
1520 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1521 				continue;
1522 
1523 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1524 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1525 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1526 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1527 					nd6_dad_start((struct ifaddr *)ifa, 0);
1528 				}
1529 			} else {
1530 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1531 			}
1532 		}
1533 	}
1534 	else {
1535 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1536 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1537 				continue;
1538 
1539 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1540 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1541 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1542 				/* Do we need a delay in this case? */
1543 				nd6_dad_start((struct ifaddr *)ifa, 0);
1544 			}
1545 		}
1546 	}
1547 }
1548 
1549 int
1550 nd6_prefix_onlink(struct nd_prefix *pr)
1551 {
1552 	struct ifaddr *ifa;
1553 	struct ifnet *ifp = pr->ndpr_ifp;
1554 	struct sockaddr_in6 mask6;
1555 	struct nd_prefix *opr;
1556 	u_long rtflags;
1557 	int error = 0;
1558 	struct radix_node_head *rnh;
1559 	struct rtentry *rt = NULL;
1560 	char ip6buf[INET6_ADDRSTRLEN];
1561 	struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1562 
1563 	/* sanity check */
1564 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1565 		nd6log((LOG_ERR,
1566 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1567 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1568 		    pr->ndpr_plen));
1569 		return (EEXIST);
1570 	}
1571 
1572 	/*
1573 	 * Add the interface route associated with the prefix.  Before
1574 	 * installing the route, check if there's the same prefix on another
1575 	 * interface, and the prefix has already installed the interface route.
1576 	 * Although such a configuration is expected to be rare, we explicitly
1577 	 * allow it.
1578 	 */
1579 	for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1580 		if (opr == pr)
1581 			continue;
1582 
1583 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1584 			continue;
1585 
1586 		if (opr->ndpr_plen == pr->ndpr_plen &&
1587 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1588 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1589 			return (0);
1590 	}
1591 
1592 	/*
1593 	 * We prefer link-local addresses as the associated interface address.
1594 	 */
1595 	/* search for a link-local addr */
1596 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1597 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1598 	if (ifa == NULL) {
1599 		/* XXX: freebsd does not have ifa_ifwithaf */
1600 		IF_ADDR_LOCK(ifp);
1601 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1602 			if (ifa->ifa_addr->sa_family == AF_INET6)
1603 				break;
1604 		}
1605 		if (ifa != NULL)
1606 			ifa_ref(ifa);
1607 		IF_ADDR_UNLOCK(ifp);
1608 		/* should we care about ia6_flags? */
1609 	}
1610 	if (ifa == NULL) {
1611 		/*
1612 		 * This can still happen, when, for example, we receive an RA
1613 		 * containing a prefix with the L bit set and the A bit clear,
1614 		 * after removing all IPv6 addresses on the receiving
1615 		 * interface.  This should, of course, be rare though.
1616 		 */
1617 		nd6log((LOG_NOTICE,
1618 		    "nd6_prefix_onlink: failed to find any ifaddr"
1619 		    " to add route for a prefix(%s/%d) on %s\n",
1620 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1621 		    pr->ndpr_plen, if_name(ifp)));
1622 		return (0);
1623 	}
1624 
1625 	/*
1626 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1627 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1628 	 */
1629 	bzero(&mask6, sizeof(mask6));
1630 	mask6.sin6_len = sizeof(mask6);
1631 	mask6.sin6_addr = pr->ndpr_mask;
1632 	rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1633 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1634 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1635 	if (error == 0) {
1636 		if (rt != NULL) /* this should be non NULL, though */ {
1637 			rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1638 			/* XXX what if rhn == NULL? */
1639 			RADIX_NODE_HEAD_LOCK(rnh);
1640 			RT_LOCK(rt);
1641 			if (!rt_setgate(rt, rt_key(rt), (struct sockaddr *)&null_sdl)) {
1642 				((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1643 					rt->rt_ifp->if_type;
1644 				((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1645 					rt->rt_ifp->if_index;
1646 			}
1647 			RADIX_NODE_HEAD_UNLOCK(rnh);
1648 			nd6_rtmsg(RTM_ADD, rt);
1649 			RT_UNLOCK(rt);
1650 		}
1651 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1652 	} else {
1653 		char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
1654 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1655 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1656 		    "errno = %d\n",
1657 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1658 		    pr->ndpr_plen, if_name(ifp),
1659 		    ip6_sprintf(ip6bufg, &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1660 		    ip6_sprintf(ip6bufm, &mask6.sin6_addr), rtflags, error));
1661 	}
1662 
1663 	if (rt != NULL) {
1664 		RT_LOCK(rt);
1665 		RT_REMREF(rt);
1666 		RT_UNLOCK(rt);
1667 	}
1668 	if (ifa != NULL)
1669 		ifa_free(ifa);
1670 
1671 	return (error);
1672 }
1673 
1674 int
1675 nd6_prefix_offlink(struct nd_prefix *pr)
1676 {
1677 	int error = 0;
1678 	struct ifnet *ifp = pr->ndpr_ifp;
1679 	struct nd_prefix *opr;
1680 	struct sockaddr_in6 sa6, mask6;
1681 	struct rtentry *rt = NULL;
1682 	char ip6buf[INET6_ADDRSTRLEN];
1683 
1684 	/* sanity check */
1685 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1686 		nd6log((LOG_ERR,
1687 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1688 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1689 		    pr->ndpr_plen));
1690 		return (EEXIST);
1691 	}
1692 
1693 	bzero(&sa6, sizeof(sa6));
1694 	sa6.sin6_family = AF_INET6;
1695 	sa6.sin6_len = sizeof(sa6);
1696 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1697 	    sizeof(struct in6_addr));
1698 	bzero(&mask6, sizeof(mask6));
1699 	mask6.sin6_family = AF_INET6;
1700 	mask6.sin6_len = sizeof(sa6);
1701 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1702 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1703 	    (struct sockaddr *)&mask6, 0, &rt);
1704 	if (error == 0) {
1705 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1706 
1707 		/* report the route deletion to the routing socket. */
1708 		if (rt != NULL)
1709 			nd6_rtmsg(RTM_DELETE, rt);
1710 
1711 		/*
1712 		 * There might be the same prefix on another interface,
1713 		 * the prefix which could not be on-link just because we have
1714 		 * the interface route (see comments in nd6_prefix_onlink).
1715 		 * If there's one, try to make the prefix on-link on the
1716 		 * interface.
1717 		 */
1718 		for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1719 			if (opr == pr)
1720 				continue;
1721 
1722 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1723 				continue;
1724 
1725 			/*
1726 			 * KAME specific: detached prefixes should not be
1727 			 * on-link.
1728 			 */
1729 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1730 				continue;
1731 
1732 			if (opr->ndpr_plen == pr->ndpr_plen &&
1733 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1734 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1735 				int e;
1736 
1737 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1738 					nd6log((LOG_ERR,
1739 					    "nd6_prefix_offlink: failed to "
1740 					    "recover a prefix %s/%d from %s "
1741 					    "to %s (errno = %d)\n",
1742 					    ip6_sprintf(ip6buf,
1743 						&opr->ndpr_prefix.sin6_addr),
1744 					    opr->ndpr_plen, if_name(ifp),
1745 					    if_name(opr->ndpr_ifp), e));
1746 				}
1747 			}
1748 		}
1749 	} else {
1750 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1751 		nd6log((LOG_ERR,
1752 		    "nd6_prefix_offlink: failed to delete route: "
1753 		    "%s/%d on %s (errno = %d)\n",
1754 		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1755 		    if_name(ifp), error));
1756 	}
1757 
1758 	if (rt != NULL) {
1759 		RTFREE(rt);
1760 	}
1761 
1762 	return (error);
1763 }
1764 
1765 static struct in6_ifaddr *
1766 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1767 {
1768 	struct ifnet *ifp = pr->ndpr_ifp;
1769 	struct ifaddr *ifa;
1770 	struct in6_aliasreq ifra;
1771 	struct in6_ifaddr *ia, *ib;
1772 	int error, plen0;
1773 	struct in6_addr mask;
1774 	int prefixlen = pr->ndpr_plen;
1775 	int updateflags;
1776 	char ip6buf[INET6_ADDRSTRLEN];
1777 
1778 	in6_prefixlen2mask(&mask, prefixlen);
1779 
1780 	/*
1781 	 * find a link-local address (will be interface ID).
1782 	 * Is it really mandatory? Theoretically, a global or a site-local
1783 	 * address can be configured without a link-local address, if we
1784 	 * have a unique interface identifier...
1785 	 *
1786 	 * it is not mandatory to have a link-local address, we can generate
1787 	 * interface identifier on the fly.  we do this because:
1788 	 * (1) it should be the easiest way to find interface identifier.
1789 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1790 	 * for multiple addresses on a single interface, and possible shortcut
1791 	 * of DAD.  we omitted DAD for this reason in the past.
1792 	 * (3) a user can prevent autoconfiguration of global address
1793 	 * by removing link-local address by hand (this is partly because we
1794 	 * don't have other way to control the use of IPv6 on an interface.
1795 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1796 	 * (4) it is easier to manage when an interface has addresses
1797 	 * with the same interface identifier, than to have multiple addresses
1798 	 * with different interface identifiers.
1799 	 */
1800 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1801 	if (ifa)
1802 		ib = (struct in6_ifaddr *)ifa;
1803 	else
1804 		return NULL;
1805 
1806 	/* prefixlen + ifidlen must be equal to 128 */
1807 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1808 	if (prefixlen != plen0) {
1809 		ifa_free(ifa);
1810 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1811 		    "(prefix=%d ifid=%d)\n",
1812 		    if_name(ifp), prefixlen, 128 - plen0));
1813 		return NULL;
1814 	}
1815 
1816 	/* make ifaddr */
1817 
1818 	bzero(&ifra, sizeof(ifra));
1819 	/*
1820 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1821 	 * for safety.
1822 	 */
1823 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1824 	ifra.ifra_addr.sin6_family = AF_INET6;
1825 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1826 	/* prefix */
1827 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1828 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1829 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1830 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1831 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1832 
1833 	/* interface ID */
1834 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1835 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1836 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1837 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1838 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1839 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1840 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1841 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1842 	ifa_free(ifa);
1843 
1844 	/* new prefix mask. */
1845 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1846 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1847 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1848 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1849 
1850 	/* lifetimes. */
1851 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1852 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1853 
1854 	/* XXX: scope zone ID? */
1855 
1856 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1857 
1858 	/*
1859 	 * Make sure that we do not have this address already.  This should
1860 	 * usually not happen, but we can still see this case, e.g., if we
1861 	 * have manually configured the exact address to be configured.
1862 	 */
1863 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1864 	    &ifra.ifra_addr.sin6_addr);
1865 	if (ifa != NULL) {
1866 		ifa_free(ifa);
1867 		/* this should be rare enough to make an explicit log */
1868 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1869 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1870 		return (NULL);
1871 	}
1872 
1873 	/*
1874 	 * Allocate ifaddr structure, link into chain, etc.
1875 	 * If we are going to create a new address upon receiving a multicasted
1876 	 * RA, we need to impose a random delay before starting DAD.
1877 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1878 	 */
1879 	updateflags = 0;
1880 	if (mcast)
1881 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1882 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1883 		nd6log((LOG_ERR,
1884 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1885 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1886 		    if_name(ifp), error));
1887 		return (NULL);	/* ifaddr must not have been allocated. */
1888 	}
1889 
1890 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1891 	/*
1892 	 * XXXRW: Assumption of non-NULLness here might not be true with
1893 	 * fine-grained locking -- should we validate it?  Or just return
1894 	 * earlier ifa rather than looking it up again?
1895 	 */
1896 	return (ia);		/* this is always non-NULL  and referenced. */
1897 }
1898 
1899 /*
1900  * ia0 - corresponding public address
1901  */
1902 int
1903 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1904 {
1905 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1906 	struct in6_ifaddr *newia, *ia;
1907 	struct in6_aliasreq ifra;
1908 	int i, error;
1909 	int trylimit = 3;	/* XXX: adhoc value */
1910 	int updateflags;
1911 	u_int32_t randid[2];
1912 	time_t vltime0, pltime0;
1913 
1914 	bzero(&ifra, sizeof(ifra));
1915 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1916 	ifra.ifra_addr = ia0->ia_addr;
1917 	/* copy prefix mask */
1918 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1919 	/* clear the old IFID */
1920 	for (i = 0; i < 4; i++) {
1921 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1922 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1923 	}
1924 
1925   again:
1926 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1927 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1928 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1929 		    "random IFID\n"));
1930 		return (EINVAL);
1931 	}
1932 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1933 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1934 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1935 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1936 
1937 	/*
1938 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1939 	 * However, we may still have a chance to see collision, because
1940 	 * there may be a time lag between generation of the ID and generation
1941 	 * of the address.  So, we'll do one more sanity check.
1942 	 */
1943 	IN6_IFADDR_RLOCK();
1944 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1945 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1946 		    &ifra.ifra_addr.sin6_addr)) {
1947 			if (trylimit-- == 0) {
1948 				IN6_IFADDR_RUNLOCK();
1949 				/*
1950 				 * Give up.  Something strange should have
1951 				 * happened.
1952 				 */
1953 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1954 				    "find a unique random IFID\n"));
1955 				return (EEXIST);
1956 			}
1957 			IN6_IFADDR_RUNLOCK();
1958 			forcegen = 1;
1959 			goto again;
1960 		}
1961 	}
1962 	IN6_IFADDR_RUNLOCK();
1963 
1964 	/*
1965 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1966          * public address or TEMP_VALID_LIFETIME.
1967 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1968          * of the public address or TEMP_PREFERRED_LIFETIME -
1969          * DESYNC_FACTOR.
1970 	 */
1971 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1972 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1973 		    (ia0->ia6_lifetime.ia6t_vltime -
1974 		    (time_second - ia0->ia6_updatetime));
1975 		if (vltime0 > V_ip6_temp_valid_lifetime)
1976 			vltime0 = V_ip6_temp_valid_lifetime;
1977 	} else
1978 		vltime0 = V_ip6_temp_valid_lifetime;
1979 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1980 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1981 		    (ia0->ia6_lifetime.ia6t_pltime -
1982 		    (time_second - ia0->ia6_updatetime));
1983 		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
1984 			pltime0 = V_ip6_temp_preferred_lifetime -
1985 			    V_ip6_desync_factor;
1986 		}
1987 	} else
1988 		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
1989 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1990 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1991 
1992 	/*
1993 	 * A temporary address is created only if this calculated Preferred
1994 	 * Lifetime is greater than REGEN_ADVANCE time units.
1995 	 */
1996 	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
1997 		return (0);
1998 
1999 	/* XXX: scope zone ID? */
2000 
2001 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2002 
2003 	/* allocate ifaddr structure, link into chain, etc. */
2004 	updateflags = 0;
2005 	if (delay)
2006 		updateflags |= IN6_IFAUPDATE_DADDELAY;
2007 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2008 		return (error);
2009 
2010 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2011 	if (newia == NULL) {	/* XXX: can it happen? */
2012 		nd6log((LOG_ERR,
2013 		    "in6_tmpifadd: ifa update succeeded, but we got "
2014 		    "no ifaddr\n"));
2015 		return (EINVAL); /* XXX */
2016 	}
2017 	newia->ia6_ndpr = ia0->ia6_ndpr;
2018 	newia->ia6_ndpr->ndpr_refcnt++;
2019 	ifa_free(&newia->ia_ifa);
2020 
2021 	/*
2022 	 * A newly added address might affect the status of other addresses.
2023 	 * XXX: when the temporary address is generated with a new public
2024 	 * address, the onlink check is redundant.  However, it would be safe
2025 	 * to do the check explicitly everywhere a new address is generated,
2026 	 * and, in fact, we surely need the check when we create a new
2027 	 * temporary address due to deprecation of an old temporary address.
2028 	 */
2029 	pfxlist_onlink_check();
2030 
2031 	return (0);
2032 }
2033 
2034 static int
2035 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2036 {
2037 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2038 		ndpr->ndpr_preferred = 0;
2039 	else
2040 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2041 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2042 		ndpr->ndpr_expire = 0;
2043 	else
2044 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2045 
2046 	return 0;
2047 }
2048 
2049 static void
2050 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2051 {
2052 	/* init ia6t_expire */
2053 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2054 		lt6->ia6t_expire = 0;
2055 	else {
2056 		lt6->ia6t_expire = time_second;
2057 		lt6->ia6t_expire += lt6->ia6t_vltime;
2058 	}
2059 
2060 	/* init ia6t_preferred */
2061 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2062 		lt6->ia6t_preferred = 0;
2063 	else {
2064 		lt6->ia6t_preferred = time_second;
2065 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2066 	}
2067 }
2068 
2069 /*
2070  * Delete all the routing table entries that use the specified gateway.
2071  * XXX: this function causes search through all entries of routing table, so
2072  * it shouldn't be called when acting as a router.
2073  */
2074 void
2075 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2076 {
2077 	struct radix_node_head *rnh;
2078 	int s = splnet();
2079 
2080 	/* We'll care only link-local addresses */
2081 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2082 		splx(s);
2083 		return;
2084 	}
2085 
2086 	rnh = rt_tables_get_rnh(0, AF_INET6);
2087 	if (rnh == NULL)
2088 		return;
2089 
2090 	RADIX_NODE_HEAD_LOCK(rnh);
2091 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2092 	RADIX_NODE_HEAD_UNLOCK(rnh);
2093 	splx(s);
2094 }
2095 
2096 static int
2097 rt6_deleteroute(struct radix_node *rn, void *arg)
2098 {
2099 #define SIN6(s)	((struct sockaddr_in6 *)s)
2100 	struct rtentry *rt = (struct rtentry *)rn;
2101 	struct in6_addr *gate = (struct in6_addr *)arg;
2102 
2103 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2104 		return (0);
2105 
2106 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2107 		return (0);
2108 	}
2109 
2110 	/*
2111 	 * Do not delete a static route.
2112 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2113 	 * 'cloned' bit instead?
2114 	 */
2115 	if ((rt->rt_flags & RTF_STATIC) != 0)
2116 		return (0);
2117 
2118 	/*
2119 	 * We delete only host route. This means, in particular, we don't
2120 	 * delete default route.
2121 	 */
2122 	if ((rt->rt_flags & RTF_HOST) == 0)
2123 		return (0);
2124 
2125 	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2126 	    rt_mask(rt), rt->rt_flags, 0));
2127 #undef SIN6
2128 }
2129 
2130 int
2131 nd6_setdefaultiface(int ifindex)
2132 {
2133 	int error = 0;
2134 
2135 	if (ifindex < 0 || V_if_index < ifindex)
2136 		return (EINVAL);
2137 	if (ifindex != 0 && !ifnet_byindex(ifindex))
2138 		return (EINVAL);
2139 
2140 	if (V_nd6_defifindex != ifindex) {
2141 		V_nd6_defifindex = ifindex;
2142 		if (V_nd6_defifindex > 0)
2143 			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2144 		else
2145 			V_nd6_defifp = NULL;
2146 
2147 		/*
2148 		 * Our current implementation assumes one-to-one maping between
2149 		 * interfaces and links, so it would be natural to use the
2150 		 * default interface as the default link.
2151 		 */
2152 		scope6_setdefault(V_nd6_defifp);
2153 	}
2154 
2155 	return (error);
2156 }
2157