xref: /freebsd/sys/netinet6/nd6_rtr.c (revision 70e0bbedef95258a4dadc996d641a9bebd3f107d)
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 nd6_prefix_onlink(struct nd_prefix *);
88 static int nd6_prefix_offlink(struct nd_prefix *);
89 
90 static int rt6_deleteroute(struct radix_node *, void *);
91 
92 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
93 #define	V_nd6_recalc_reachtm_interval	VNET(nd6_recalc_reachtm_interval)
94 
95 static VNET_DEFINE(struct ifnet *, nd6_defifp);
96 VNET_DEFINE(int, nd6_defifindex);
97 #define	V_nd6_defifp			VNET(nd6_defifp)
98 
99 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
100 
101 VNET_DEFINE(int, ip6_desync_factor);
102 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
103 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
104 
105 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
106 
107 /* RTPREF_MEDIUM has to be 0! */
108 #define RTPREF_HIGH	1
109 #define RTPREF_MEDIUM	0
110 #define RTPREF_LOW	(-1)
111 #define RTPREF_RESERVED	(-2)
112 #define RTPREF_INVALID	(-3)	/* internal */
113 
114 /*
115  * Receive Router Solicitation Message - just for routers.
116  * Router solicitation/advertisement is mostly managed by userland program
117  * (rtadvd) so here we have no function like nd6_ra_output().
118  *
119  * Based on RFC 2461
120  */
121 void
122 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
123 {
124 	struct ifnet *ifp = m->m_pkthdr.rcvif;
125 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
126 	struct nd_router_solicit *nd_rs;
127 	struct in6_addr saddr6 = ip6->ip6_src;
128 	char *lladdr = NULL;
129 	int lladdrlen = 0;
130 	union nd_opts ndopts;
131 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
132 
133 	/*
134 	 * Accept RS only when V_ip6_forwarding=1 and the interface has
135 	 * no ND6_IFF_ACCEPT_RTADV.
136 	 */
137 	if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
138 		goto freeit;
139 
140 	/* Sanity checks */
141 	if (ip6->ip6_hlim != 255) {
142 		nd6log((LOG_ERR,
143 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
144 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
145 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
146 		goto bad;
147 	}
148 
149 	/*
150 	 * Don't update the neighbor cache, if src = ::.
151 	 * This indicates that the src has no IP address assigned yet.
152 	 */
153 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
154 		goto freeit;
155 
156 #ifndef PULLDOWN_TEST
157 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
158 	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
159 #else
160 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
161 	if (nd_rs == NULL) {
162 		ICMP6STAT_INC(icp6s_tooshort);
163 		return;
164 	}
165 #endif
166 
167 	icmp6len -= sizeof(*nd_rs);
168 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
169 	if (nd6_options(&ndopts) < 0) {
170 		nd6log((LOG_INFO,
171 		    "nd6_rs_input: invalid ND option, ignored\n"));
172 		/* nd6_options have incremented stats */
173 		goto freeit;
174 	}
175 
176 	if (ndopts.nd_opts_src_lladdr) {
177 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
178 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
179 	}
180 
181 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
182 		nd6log((LOG_INFO,
183 		    "nd6_rs_input: lladdrlen mismatch for %s "
184 		    "(if %d, RS packet %d)\n",
185 		    ip6_sprintf(ip6bufs, &saddr6),
186 		    ifp->if_addrlen, lladdrlen - 2));
187 		goto bad;
188 	}
189 
190 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
191 
192  freeit:
193 	m_freem(m);
194 	return;
195 
196  bad:
197 	ICMP6STAT_INC(icp6s_badrs);
198 	m_freem(m);
199 }
200 
201 /*
202  * Receive Router Advertisement Message.
203  *
204  * Based on RFC 2461
205  * TODO: on-link bit on prefix information
206  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
207  */
208 void
209 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
210 {
211 	struct ifnet *ifp = m->m_pkthdr.rcvif;
212 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
213 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
214 	struct nd_router_advert *nd_ra;
215 	struct in6_addr saddr6 = ip6->ip6_src;
216 	int mcast = 0;
217 	union nd_opts ndopts;
218 	struct nd_defrouter *dr;
219 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
220 
221 	/*
222 	 * We only accept RAs only when the per-interface flag
223 	 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
224 	 */
225 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
226 		goto freeit;
227 
228 	if (ip6->ip6_hlim != 255) {
229 		nd6log((LOG_ERR,
230 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
231 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
232 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
233 		goto bad;
234 	}
235 
236 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
237 		nd6log((LOG_ERR,
238 		    "nd6_ra_input: src %s is not link-local\n",
239 		    ip6_sprintf(ip6bufs, &saddr6)));
240 		goto bad;
241 	}
242 
243 #ifndef PULLDOWN_TEST
244 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
245 	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
246 #else
247 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
248 	if (nd_ra == NULL) {
249 		ICMP6STAT_INC(icp6s_tooshort);
250 		return;
251 	}
252 #endif
253 
254 	icmp6len -= sizeof(*nd_ra);
255 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
256 	if (nd6_options(&ndopts) < 0) {
257 		nd6log((LOG_INFO,
258 		    "nd6_ra_input: invalid ND option, ignored\n"));
259 		/* nd6_options have incremented stats */
260 		goto freeit;
261 	}
262 
263     {
264 	struct nd_defrouter dr0;
265 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
266 
267 	/* remember if this is a multicasted advertisement */
268 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
269 		mcast = 1;
270 
271 	bzero(&dr0, sizeof(dr0));
272 	dr0.rtaddr = saddr6;
273 	dr0.flags  = nd_ra->nd_ra_flags_reserved;
274 	/*
275 	 * Effectively-disable routes from RA messages when
276 	 * ND6_IFF_NO_RADR enabled on the receiving interface or
277 	 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
278 	 */
279 	if (ndi->flags & ND6_IFF_NO_RADR)
280 		dr0.rtlifetime = 0;
281 	else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
282 		dr0.rtlifetime = 0;
283 	else
284 		dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
285 	dr0.expire = time_second + dr0.rtlifetime;
286 	dr0.ifp = ifp;
287 	/* unspecified or not? (RFC 2461 6.3.4) */
288 	if (advreachable) {
289 		advreachable = ntohl(advreachable);
290 		if (advreachable <= MAX_REACHABLE_TIME &&
291 		    ndi->basereachable != advreachable) {
292 			ndi->basereachable = advreachable;
293 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
294 			ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
295 		}
296 	}
297 	if (nd_ra->nd_ra_retransmit)
298 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
299 	if (nd_ra->nd_ra_curhoplimit)
300 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
301 	dr = defrtrlist_update(&dr0);
302     }
303 
304 	/*
305 	 * prefix
306 	 */
307 	if (ndopts.nd_opts_pi) {
308 		struct nd_opt_hdr *pt;
309 		struct nd_opt_prefix_info *pi = NULL;
310 		struct nd_prefixctl pr;
311 
312 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
313 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
314 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
315 						(pt->nd_opt_len << 3))) {
316 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
317 				continue;
318 			pi = (struct nd_opt_prefix_info *)pt;
319 
320 			if (pi->nd_opt_pi_len != 4) {
321 				nd6log((LOG_INFO,
322 				    "nd6_ra_input: invalid option "
323 				    "len %d for prefix information option, "
324 				    "ignored\n", pi->nd_opt_pi_len));
325 				continue;
326 			}
327 
328 			if (128 < pi->nd_opt_pi_prefix_len) {
329 				nd6log((LOG_INFO,
330 				    "nd6_ra_input: invalid prefix "
331 				    "len %d for prefix information option, "
332 				    "ignored\n", pi->nd_opt_pi_prefix_len));
333 				continue;
334 			}
335 
336 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
337 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
338 				nd6log((LOG_INFO,
339 				    "nd6_ra_input: invalid prefix "
340 				    "%s, ignored\n",
341 				    ip6_sprintf(ip6bufs,
342 					&pi->nd_opt_pi_prefix)));
343 				continue;
344 			}
345 
346 			bzero(&pr, sizeof(pr));
347 			pr.ndpr_prefix.sin6_family = AF_INET6;
348 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
349 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
350 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
351 
352 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
353 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
354 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
355 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
356 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
357 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
358 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
359 			(void)prelist_update(&pr, dr, m, mcast);
360 		}
361 	}
362 
363 	/*
364 	 * MTU
365 	 */
366 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
367 		u_long mtu;
368 		u_long maxmtu;
369 
370 		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
371 
372 		/* lower bound */
373 		if (mtu < IPV6_MMTU) {
374 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
375 			    "mtu=%lu sent from %s, ignoring\n",
376 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
377 			goto skip;
378 		}
379 
380 		/* upper bound */
381 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
382 		    ? ndi->maxmtu : ifp->if_mtu;
383 		if (mtu <= maxmtu) {
384 			int change = (ndi->linkmtu != mtu);
385 
386 			ndi->linkmtu = mtu;
387 			if (change) /* in6_maxmtu may change */
388 				in6_setmaxmtu();
389 		} else {
390 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
391 			    "mtu=%lu sent from %s; "
392 			    "exceeds maxmtu %lu, ignoring\n",
393 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
394 		}
395 	}
396 
397  skip:
398 
399 	/*
400 	 * Source link layer address
401 	 */
402     {
403 	char *lladdr = NULL;
404 	int lladdrlen = 0;
405 
406 	if (ndopts.nd_opts_src_lladdr) {
407 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
408 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
409 	}
410 
411 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
412 		nd6log((LOG_INFO,
413 		    "nd6_ra_input: lladdrlen mismatch for %s "
414 		    "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
415 		    ifp->if_addrlen, lladdrlen - 2));
416 		goto bad;
417 	}
418 
419 	nd6_cache_lladdr(ifp, &saddr6, lladdr,
420 	    lladdrlen, ND_ROUTER_ADVERT, 0);
421 
422 	/*
423 	 * Installing a link-layer address might change the state of the
424 	 * router's neighbor cache, which might also affect our on-link
425 	 * detection of adveritsed prefixes.
426 	 */
427 	pfxlist_onlink_check();
428     }
429 
430  freeit:
431 	m_freem(m);
432 	return;
433 
434  bad:
435 	ICMP6STAT_INC(icp6s_badra);
436 	m_freem(m);
437 }
438 
439 /*
440  * default router list proccessing sub routines
441  */
442 
443 /* tell the change to user processes watching the routing socket. */
444 static void
445 nd6_rtmsg(int cmd, struct rtentry *rt)
446 {
447 	struct rt_addrinfo info;
448 	struct ifnet *ifp;
449 	struct ifaddr *ifa;
450 
451 	bzero((caddr_t)&info, sizeof(info));
452 	info.rti_info[RTAX_DST] = rt_key(rt);
453 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
454 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
455 	ifp = rt->rt_ifp;
456 	if (ifp != NULL) {
457 		IF_ADDR_RLOCK(ifp);
458 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
459 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
460 		ifa_ref(ifa);
461 		IF_ADDR_RUNLOCK(ifp);
462 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
463 	} else
464 		ifa = NULL;
465 
466 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
467 	if (ifa != NULL)
468 		ifa_free(ifa);
469 }
470 
471 static void
472 defrouter_addreq(struct nd_defrouter *new)
473 {
474 	struct sockaddr_in6 def, mask, gate;
475 	struct rtentry *newrt = NULL;
476 	int s;
477 	int error;
478 
479 	bzero(&def, sizeof(def));
480 	bzero(&mask, sizeof(mask));
481 	bzero(&gate, sizeof(gate));
482 
483 	def.sin6_len = mask.sin6_len = gate.sin6_len =
484 	    sizeof(struct sockaddr_in6);
485 	def.sin6_family = gate.sin6_family = AF_INET6;
486 	gate.sin6_addr = new->rtaddr;
487 
488 	s = splnet();
489 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
490 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
491 	    RTF_GATEWAY, &newrt);
492 	if (newrt) {
493 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
494 		RTFREE(newrt);
495 	}
496 	if (error == 0)
497 		new->installed = 1;
498 	splx(s);
499 	return;
500 }
501 
502 struct nd_defrouter *
503 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
504 {
505 	struct nd_defrouter *dr;
506 
507 	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
508 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
509 			return (dr);
510 	}
511 
512 	return (NULL);		/* search failed */
513 }
514 
515 /*
516  * Remove the default route for a given router.
517  * This is just a subroutine function for defrouter_select(), and should
518  * not be called from anywhere else.
519  */
520 static void
521 defrouter_delreq(struct nd_defrouter *dr)
522 {
523 	struct sockaddr_in6 def, mask, gate;
524 	struct rtentry *oldrt = NULL;
525 
526 	bzero(&def, sizeof(def));
527 	bzero(&mask, sizeof(mask));
528 	bzero(&gate, sizeof(gate));
529 
530 	def.sin6_len = mask.sin6_len = gate.sin6_len =
531 	    sizeof(struct sockaddr_in6);
532 	def.sin6_family = gate.sin6_family = AF_INET6;
533 	gate.sin6_addr = dr->rtaddr;
534 
535 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
536 	    (struct sockaddr *)&gate,
537 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
538 	if (oldrt) {
539 		nd6_rtmsg(RTM_DELETE, oldrt);
540 		RTFREE(oldrt);
541 	}
542 
543 	dr->installed = 0;
544 }
545 
546 /*
547  * remove all default routes from default router list
548  */
549 void
550 defrouter_reset(void)
551 {
552 	struct nd_defrouter *dr;
553 
554 	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry)
555 		defrouter_delreq(dr);
556 
557 	/*
558 	 * XXX should we also nuke any default routers in the kernel, by
559 	 * going through them by rtalloc1()?
560 	 */
561 }
562 
563 void
564 defrtrlist_del(struct nd_defrouter *dr)
565 {
566 	struct nd_defrouter *deldr = NULL;
567 	struct nd_prefix *pr;
568 
569 	/*
570 	 * Flush all the routing table entries that use the router
571 	 * as a next hop.
572 	 */
573 	if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
574 		rt6_flush(&dr->rtaddr, dr->ifp);
575 
576 	if (dr->installed) {
577 		deldr = dr;
578 		defrouter_delreq(dr);
579 	}
580 	TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
581 
582 	/*
583 	 * Also delete all the pointers to the router in each prefix lists.
584 	 */
585 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
586 		struct nd_pfxrouter *pfxrtr;
587 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
588 			pfxrtr_del(pfxrtr);
589 	}
590 	pfxlist_onlink_check();
591 
592 	/*
593 	 * If the router is the primary one, choose a new one.
594 	 * Note that defrouter_select() will remove the current gateway
595 	 * from the routing table.
596 	 */
597 	if (deldr)
598 		defrouter_select();
599 
600 	free(dr, M_IP6NDP);
601 }
602 
603 /*
604  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
605  * draft-ietf-ipngwg-router-selection:
606  * 1) Routers that are reachable or probably reachable should be preferred.
607  *    If we have more than one (probably) reachable router, prefer ones
608  *    with the highest router preference.
609  * 2) When no routers on the list are known to be reachable or
610  *    probably reachable, routers SHOULD be selected in a round-robin
611  *    fashion, regardless of router preference values.
612  * 3) If the Default Router List is empty, assume that all
613  *    destinations are on-link.
614  *
615  * We assume nd_defrouter is sorted by router preference value.
616  * Since the code below covers both with and without router preference cases,
617  * we do not need to classify the cases by ifdef.
618  *
619  * At this moment, we do not try to install more than one default router,
620  * even when the multipath routing is available, because we're not sure about
621  * the benefits for stub hosts comparing to the risk of making the code
622  * complicated and the possibility of introducing bugs.
623  */
624 void
625 defrouter_select(void)
626 {
627 	int s = splnet();
628 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
629 	struct llentry *ln = NULL;
630 
631 	/*
632 	 * Let's handle easy case (3) first:
633 	 * If default router list is empty, there's nothing to be done.
634 	 */
635 	if (TAILQ_EMPTY(&V_nd_defrouter)) {
636 		splx(s);
637 		return;
638 	}
639 
640 	/*
641 	 * Search for a (probably) reachable router from the list.
642 	 * We just pick up the first reachable one (if any), assuming that
643 	 * the ordering rule of the list described in defrtrlist_update().
644 	 */
645 	TAILQ_FOREACH(dr, &V_nd_defrouter, 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. Also make sure the selected
755 			 * router is still installed in the kernel.
756 			 */
757 			if (dr->installed && rtpref(new) == oldpref) {
758 				splx(s);
759 				return (dr);
760 			}
761 
762 			/*
763 			 * preferred router may be changed, so relocate
764 			 * this router.
765 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
766 			 * However, since defrtrlist_del() has many side
767 			 * effects, we intentionally do so here.
768 			 * defrouter_select() below will handle routing
769 			 * changes later.
770 			 */
771 			TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
772 			n = dr;
773 			goto insert;
774 		}
775 		splx(s);
776 		return (dr);
777 	}
778 
779 	/* entry does not exist */
780 	if (new->rtlifetime == 0) {
781 		splx(s);
782 		return (NULL);
783 	}
784 
785 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
786 	if (n == NULL) {
787 		splx(s);
788 		return (NULL);
789 	}
790 	bzero(n, sizeof(*n));
791 	*n = *new;
792 
793 insert:
794 	/*
795 	 * Insert the new router in the Default Router List;
796 	 * The Default Router List should be in the descending order
797 	 * of router-preferece.  Routers with the same preference are
798 	 * sorted in the arriving time order.
799 	 */
800 
801 	/* insert at the end of the group */
802 	TAILQ_FOREACH(dr, &V_nd_defrouter, 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 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
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 	LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
860 		if (key->ndpr_ifp == search->ndpr_ifp &&
861 		    key->ndpr_plen == search->ndpr_plen &&
862 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
863 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
864 			break;
865 		}
866 	}
867 
868 	return (search);
869 }
870 
871 int
872 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
873     struct nd_prefix **newp)
874 {
875 	struct nd_prefix *new = NULL;
876 	int error = 0;
877 	int i, s;
878 	char ip6buf[INET6_ADDRSTRLEN];
879 
880 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
881 	if (new == NULL)
882 		return(ENOMEM);
883 	bzero(new, sizeof(*new));
884 	new->ndpr_ifp = pr->ndpr_ifp;
885 	new->ndpr_prefix = pr->ndpr_prefix;
886 	new->ndpr_plen = pr->ndpr_plen;
887 	new->ndpr_vltime = pr->ndpr_vltime;
888 	new->ndpr_pltime = pr->ndpr_pltime;
889 	new->ndpr_flags = pr->ndpr_flags;
890 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
891 		free(new, M_IP6NDP);
892 		return(error);
893 	}
894 	new->ndpr_lastupdate = time_second;
895 	if (newp != NULL)
896 		*newp = new;
897 
898 	/* initialization */
899 	LIST_INIT(&new->ndpr_advrtrs);
900 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
901 	/* make prefix in the canonical form */
902 	for (i = 0; i < 4; i++)
903 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
904 		    new->ndpr_mask.s6_addr32[i];
905 
906 	s = splnet();
907 	/* link ndpr_entry to nd_prefix list */
908 	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
909 	splx(s);
910 
911 	/* ND_OPT_PI_FLAG_ONLINK processing */
912 	if (new->ndpr_raf_onlink) {
913 		int e;
914 
915 		if ((e = nd6_prefix_onlink(new)) != 0) {
916 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
917 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
918 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
919 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
920 			/* proceed anyway. XXX: is it correct? */
921 		}
922 	}
923 
924 	if (dr)
925 		pfxrtr_add(new, dr);
926 
927 	return 0;
928 }
929 
930 void
931 prelist_remove(struct nd_prefix *pr)
932 {
933 	struct nd_pfxrouter *pfr, *next;
934 	int e, s;
935 	char ip6buf[INET6_ADDRSTRLEN];
936 
937 	/* make sure to invalidate the prefix until it is really freed. */
938 	pr->ndpr_vltime = 0;
939 	pr->ndpr_pltime = 0;
940 
941 	/*
942 	 * Though these flags are now meaningless, we'd rather keep the value
943 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
944 	 * when executing "ndp -p".
945 	 */
946 
947 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
948 	    (e = nd6_prefix_offlink(pr)) != 0) {
949 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
950 		    "on %s, errno=%d\n",
951 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
952 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
953 		/* what should we do? */
954 	}
955 
956 	if (pr->ndpr_refcnt > 0)
957 		return;		/* notice here? */
958 
959 	s = splnet();
960 
961 	/* unlink ndpr_entry from nd_prefix list */
962 	LIST_REMOVE(pr, ndpr_entry);
963 
964 	/* free list of routers that adversed the prefix */
965 	LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next) {
966 		free(pfr, M_IP6NDP);
967 	}
968 	splx(s);
969 
970 	free(pr, M_IP6NDP);
971 
972 	pfxlist_onlink_check();
973 }
974 
975 /*
976  * dr - may be NULL
977  */
978 
979 static int
980 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
981     struct mbuf *m, int mcast)
982 {
983 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
984 	struct ifaddr *ifa;
985 	struct ifnet *ifp = new->ndpr_ifp;
986 	struct nd_prefix *pr;
987 	int s = splnet();
988 	int error = 0;
989 	int newprefix = 0;
990 	int auth;
991 	struct in6_addrlifetime lt6_tmp;
992 	char ip6buf[INET6_ADDRSTRLEN];
993 
994 	auth = 0;
995 	if (m) {
996 		/*
997 		 * Authenticity for NA consists authentication for
998 		 * both IP header and IP datagrams, doesn't it ?
999 		 */
1000 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1001 		auth = ((m->m_flags & M_AUTHIPHDR) &&
1002 		    (m->m_flags & M_AUTHIPDGM));
1003 #endif
1004 	}
1005 
1006 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1007 		/*
1008 		 * nd6_prefix_lookup() ensures that pr and new have the same
1009 		 * prefix on a same interface.
1010 		 */
1011 
1012 		/*
1013 		 * Update prefix information.  Note that the on-link (L) bit
1014 		 * and the autonomous (A) bit should NOT be changed from 1
1015 		 * to 0.
1016 		 */
1017 		if (new->ndpr_raf_onlink == 1)
1018 			pr->ndpr_raf_onlink = 1;
1019 		if (new->ndpr_raf_auto == 1)
1020 			pr->ndpr_raf_auto = 1;
1021 		if (new->ndpr_raf_onlink) {
1022 			pr->ndpr_vltime = new->ndpr_vltime;
1023 			pr->ndpr_pltime = new->ndpr_pltime;
1024 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1025 			pr->ndpr_lastupdate = time_second;
1026 		}
1027 
1028 		if (new->ndpr_raf_onlink &&
1029 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1030 			int e;
1031 
1032 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1033 				nd6log((LOG_ERR,
1034 				    "prelist_update: failed to make "
1035 				    "the prefix %s/%d on-link on %s "
1036 				    "(errno=%d)\n",
1037 				    ip6_sprintf(ip6buf,
1038 					    &pr->ndpr_prefix.sin6_addr),
1039 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1040 				/* proceed anyway. XXX: is it correct? */
1041 			}
1042 		}
1043 
1044 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1045 			pfxrtr_add(pr, dr);
1046 	} else {
1047 		struct nd_prefix *newpr = NULL;
1048 
1049 		newprefix = 1;
1050 
1051 		if (new->ndpr_vltime == 0)
1052 			goto end;
1053 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1054 			goto end;
1055 
1056 		error = nd6_prelist_add(new, dr, &newpr);
1057 		if (error != 0 || newpr == NULL) {
1058 			nd6log((LOG_NOTICE, "prelist_update: "
1059 			    "nd6_prelist_add failed for %s/%d on %s "
1060 			    "errno=%d, returnpr=%p\n",
1061 			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1062 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1063 			    error, newpr));
1064 			goto end; /* we should just give up in this case. */
1065 		}
1066 
1067 		/*
1068 		 * XXX: from the ND point of view, we can ignore a prefix
1069 		 * with the on-link bit being zero.  However, we need a
1070 		 * prefix structure for references from autoconfigured
1071 		 * addresses.  Thus, we explicitly make sure that the prefix
1072 		 * itself expires now.
1073 		 */
1074 		if (newpr->ndpr_raf_onlink == 0) {
1075 			newpr->ndpr_vltime = 0;
1076 			newpr->ndpr_pltime = 0;
1077 			in6_init_prefix_ltimes(newpr);
1078 		}
1079 
1080 		pr = newpr;
1081 	}
1082 
1083 	/*
1084 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1085 	 * Note that pr must be non NULL at this point.
1086 	 */
1087 
1088 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1089 	if (!new->ndpr_raf_auto)
1090 		goto end;
1091 
1092 	/*
1093 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1094 	 * nd6_ra_input.
1095 	 */
1096 
1097 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1098 	if (new->ndpr_pltime > new->ndpr_vltime) {
1099 		error = EINVAL;	/* XXX: won't be used */
1100 		goto end;
1101 	}
1102 
1103 	/*
1104 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1105 	 * an address configured by stateless autoconfiguration already in the
1106 	 * list of addresses associated with the interface, and the Valid
1107 	 * Lifetime is not 0, form an address.  We first check if we have
1108 	 * a matching prefix.
1109 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1110 	 * consider autoconfigured addresses while RFC2462 simply said
1111 	 * "address".
1112 	 */
1113 	IF_ADDR_RLOCK(ifp);
1114 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1115 		struct in6_ifaddr *ifa6;
1116 		u_int32_t remaininglifetime;
1117 
1118 		if (ifa->ifa_addr->sa_family != AF_INET6)
1119 			continue;
1120 
1121 		ifa6 = (struct in6_ifaddr *)ifa;
1122 
1123 		/*
1124 		 * We only consider autoconfigured addresses as per rfc2462bis.
1125 		 */
1126 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1127 			continue;
1128 
1129 		/*
1130 		 * Spec is not clear here, but I believe we should concentrate
1131 		 * on unicast (i.e. not anycast) addresses.
1132 		 * XXX: other ia6_flags? detached or duplicated?
1133 		 */
1134 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1135 			continue;
1136 
1137 		/*
1138 		 * Ignore the address if it is not associated with a prefix
1139 		 * or is associated with a prefix that is different from this
1140 		 * one.  (pr is never NULL here)
1141 		 */
1142 		if (ifa6->ia6_ndpr != pr)
1143 			continue;
1144 
1145 		if (ia6_match == NULL) /* remember the first one */
1146 			ia6_match = ifa6;
1147 
1148 		/*
1149 		 * An already autoconfigured address matched.  Now that we
1150 		 * are sure there is at least one matched address, we can
1151 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1152 		 * "two hours" rule and the privacy extension.
1153 		 * We apply some clarifications in rfc2462bis:
1154 		 * - use remaininglifetime instead of storedlifetime as a
1155 		 *   variable name
1156 		 * - remove the dead code in the "two-hour" rule
1157 		 */
1158 #define TWOHOUR		(120*60)
1159 		lt6_tmp = ifa6->ia6_lifetime;
1160 
1161 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1162 			remaininglifetime = ND6_INFINITE_LIFETIME;
1163 		else if (time_second - ifa6->ia6_updatetime >
1164 			 lt6_tmp.ia6t_vltime) {
1165 			/*
1166 			 * The case of "invalid" address.  We should usually
1167 			 * not see this case.
1168 			 */
1169 			remaininglifetime = 0;
1170 		} else
1171 			remaininglifetime = lt6_tmp.ia6t_vltime -
1172 			    (time_second - ifa6->ia6_updatetime);
1173 
1174 		/* when not updating, keep the current stored lifetime. */
1175 		lt6_tmp.ia6t_vltime = remaininglifetime;
1176 
1177 		if (TWOHOUR < new->ndpr_vltime ||
1178 		    remaininglifetime < new->ndpr_vltime) {
1179 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1180 		} else if (remaininglifetime <= TWOHOUR) {
1181 			if (auth) {
1182 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1183 			}
1184 		} else {
1185 			/*
1186 			 * new->ndpr_vltime <= TWOHOUR &&
1187 			 * TWOHOUR < remaininglifetime
1188 			 */
1189 			lt6_tmp.ia6t_vltime = TWOHOUR;
1190 		}
1191 
1192 		/* The 2 hour rule is not imposed for preferred lifetime. */
1193 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1194 
1195 		in6_init_address_ltimes(pr, &lt6_tmp);
1196 
1197 		/*
1198 		 * We need to treat lifetimes for temporary addresses
1199 		 * differently, according to
1200 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1201 		 * we only update the lifetimes when they are in the maximum
1202 		 * intervals.
1203 		 */
1204 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1205 			u_int32_t maxvltime, maxpltime;
1206 
1207 			if (V_ip6_temp_valid_lifetime >
1208 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1209 			    V_ip6_desync_factor)) {
1210 				maxvltime = V_ip6_temp_valid_lifetime -
1211 				    (time_second - ifa6->ia6_createtime) -
1212 				    V_ip6_desync_factor;
1213 			} else
1214 				maxvltime = 0;
1215 			if (V_ip6_temp_preferred_lifetime >
1216 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1217 			    V_ip6_desync_factor)) {
1218 				maxpltime = V_ip6_temp_preferred_lifetime -
1219 				    (time_second - ifa6->ia6_createtime) -
1220 				    V_ip6_desync_factor;
1221 			} else
1222 				maxpltime = 0;
1223 
1224 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1225 			    lt6_tmp.ia6t_vltime > maxvltime) {
1226 				lt6_tmp.ia6t_vltime = maxvltime;
1227 			}
1228 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1229 			    lt6_tmp.ia6t_pltime > maxpltime) {
1230 				lt6_tmp.ia6t_pltime = maxpltime;
1231 			}
1232 		}
1233 		ifa6->ia6_lifetime = lt6_tmp;
1234 		ifa6->ia6_updatetime = time_second;
1235 	}
1236 	IF_ADDR_RUNLOCK(ifp);
1237 	if (ia6_match == NULL && new->ndpr_vltime) {
1238 		int ifidlen;
1239 
1240 		/*
1241 		 * 5.5.3 (d) (continued)
1242 		 * No address matched and the valid lifetime is non-zero.
1243 		 * Create a new address.
1244 		 */
1245 
1246 		/*
1247 		 * Prefix Length check:
1248 		 * If the sum of the prefix length and interface identifier
1249 		 * length does not equal 128 bits, the Prefix Information
1250 		 * option MUST be ignored.  The length of the interface
1251 		 * identifier is defined in a separate link-type specific
1252 		 * document.
1253 		 */
1254 		ifidlen = in6_if2idlen(ifp);
1255 		if (ifidlen < 0) {
1256 			/* this should not happen, so we always log it. */
1257 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1258 			    if_name(ifp));
1259 			goto end;
1260 		}
1261 		if (ifidlen + pr->ndpr_plen != 128) {
1262 			nd6log((LOG_INFO,
1263 			    "prelist_update: invalid prefixlen "
1264 			    "%d for %s, ignored\n",
1265 			    pr->ndpr_plen, if_name(ifp)));
1266 			goto end;
1267 		}
1268 
1269 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1270 			/*
1271 			 * note that we should use pr (not new) for reference.
1272 			 */
1273 			pr->ndpr_refcnt++;
1274 			ia6->ia6_ndpr = pr;
1275 
1276 			/*
1277 			 * RFC 3041 3.3 (2).
1278 			 * When a new public address is created as described
1279 			 * in RFC2462, also create a new temporary address.
1280 			 *
1281 			 * RFC 3041 3.5.
1282 			 * When an interface connects to a new link, a new
1283 			 * randomized interface identifier should be generated
1284 			 * immediately together with a new set of temporary
1285 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1286 			 * in6_tmpifadd().
1287 			 */
1288 			if (V_ip6_use_tempaddr) {
1289 				int e;
1290 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1291 					nd6log((LOG_NOTICE, "prelist_update: "
1292 					    "failed to create a temporary "
1293 					    "address, errno=%d\n",
1294 					    e));
1295 				}
1296 			}
1297 			ifa_free(&ia6->ia_ifa);
1298 
1299 			/*
1300 			 * A newly added address might affect the status
1301 			 * of other addresses, so we check and update it.
1302 			 * XXX: what if address duplication happens?
1303 			 */
1304 			pfxlist_onlink_check();
1305 		} else {
1306 			/* just set an error. do not bark here. */
1307 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1308 		}
1309 	}
1310 
1311  end:
1312 	splx(s);
1313 	return error;
1314 }
1315 
1316 /*
1317  * A supplement function used in the on-link detection below;
1318  * detect if a given prefix has a (probably) reachable advertising router.
1319  * XXX: lengthy function name...
1320  */
1321 static struct nd_pfxrouter *
1322 find_pfxlist_reachable_router(struct nd_prefix *pr)
1323 {
1324 	struct nd_pfxrouter *pfxrtr;
1325 	struct llentry *ln;
1326 	int canreach;
1327 
1328 	LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1329 		IF_AFDATA_LOCK(pfxrtr->router->ifp);
1330 		ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1331 		IF_AFDATA_UNLOCK(pfxrtr->router->ifp);
1332 		if (ln == NULL)
1333 			continue;
1334 		canreach = ND6_IS_LLINFO_PROBREACH(ln);
1335 		LLE_RUNLOCK(ln);
1336 		if (canreach)
1337 			break;
1338 	}
1339 	return (pfxrtr);
1340 }
1341 
1342 /*
1343  * Check if each prefix in the prefix list has at least one available router
1344  * that advertised the prefix (a router is "available" if its neighbor cache
1345  * entry is reachable or probably reachable).
1346  * If the check fails, the prefix may be off-link, because, for example,
1347  * we have moved from the network but the lifetime of the prefix has not
1348  * expired yet.  So we should not use the prefix if there is another prefix
1349  * that has an available router.
1350  * But, if there is no prefix that has an available router, we still regards
1351  * all the prefixes as on-link.  This is because we can't tell if all the
1352  * routers are simply dead or if we really moved from the network and there
1353  * is no router around us.
1354  */
1355 void
1356 pfxlist_onlink_check()
1357 {
1358 	struct nd_prefix *pr;
1359 	struct in6_ifaddr *ifa;
1360 	struct nd_defrouter *dr;
1361 	struct nd_pfxrouter *pfxrtr = NULL;
1362 
1363 	/*
1364 	 * Check if there is a prefix that has a reachable advertising
1365 	 * router.
1366 	 */
1367 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1368 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1369 			break;
1370 	}
1371 
1372 	/*
1373 	 * If we have no such prefix, check whether we still have a router
1374 	 * that does not advertise any prefixes.
1375 	 */
1376 	if (pr == NULL) {
1377 		TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1378 			struct nd_prefix *pr0;
1379 
1380 			LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1381 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1382 					break;
1383 			}
1384 			if (pfxrtr != NULL)
1385 				break;
1386 		}
1387 	}
1388 	if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) {
1389 		/*
1390 		 * There is at least one prefix that has a reachable router,
1391 		 * or at least a router which probably does not advertise
1392 		 * any prefixes.  The latter would be the case when we move
1393 		 * to a new link where we have a router that does not provide
1394 		 * prefixes and we configure an address by hand.
1395 		 * Detach prefixes which have no reachable advertising
1396 		 * router, and attach other prefixes.
1397 		 */
1398 		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1399 			/* XXX: a link-local prefix should never be detached */
1400 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1401 				continue;
1402 
1403 			/*
1404 			 * we aren't interested in prefixes without the L bit
1405 			 * set.
1406 			 */
1407 			if (pr->ndpr_raf_onlink == 0)
1408 				continue;
1409 
1410 			if (pr->ndpr_raf_auto == 0)
1411 				continue;
1412 
1413 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1414 			    find_pfxlist_reachable_router(pr) == NULL)
1415 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1416 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1417 			    find_pfxlist_reachable_router(pr) != 0)
1418 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1419 		}
1420 	} else {
1421 		/* there is no prefix that has a reachable router */
1422 		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1423 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1424 				continue;
1425 
1426 			if (pr->ndpr_raf_onlink == 0)
1427 				continue;
1428 
1429 			if (pr->ndpr_raf_auto == 0)
1430 				continue;
1431 
1432 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1433 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1434 		}
1435 	}
1436 
1437 	/*
1438 	 * Remove each interface route associated with a (just) detached
1439 	 * prefix, and reinstall the interface route for a (just) attached
1440 	 * prefix.  Note that all attempt of reinstallation does not
1441 	 * necessarily success, when a same prefix is shared among multiple
1442 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1443 	 * so we don't have to care about them.
1444 	 */
1445 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1446 		int e;
1447 		char ip6buf[INET6_ADDRSTRLEN];
1448 
1449 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1450 			continue;
1451 
1452 		if (pr->ndpr_raf_onlink == 0)
1453 			continue;
1454 
1455 		if (pr->ndpr_raf_auto == 0)
1456 			continue;
1457 
1458 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1459 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1460 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1461 				nd6log((LOG_ERR,
1462 				    "pfxlist_onlink_check: failed to "
1463 				    "make %s/%d offlink, errno=%d\n",
1464 				    ip6_sprintf(ip6buf,
1465 					    &pr->ndpr_prefix.sin6_addr),
1466 					    pr->ndpr_plen, e));
1467 			}
1468 		}
1469 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1470 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1471 		    pr->ndpr_raf_onlink) {
1472 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1473 				nd6log((LOG_ERR,
1474 				    "pfxlist_onlink_check: failed to "
1475 				    "make %s/%d onlink, errno=%d\n",
1476 				    ip6_sprintf(ip6buf,
1477 					    &pr->ndpr_prefix.sin6_addr),
1478 					    pr->ndpr_plen, e));
1479 			}
1480 		}
1481 	}
1482 
1483 	/*
1484 	 * Changes on the prefix status might affect address status as well.
1485 	 * Make sure that all addresses derived from an attached prefix are
1486 	 * attached, and that all addresses derived from a detached prefix are
1487 	 * detached.  Note, however, that a manually configured address should
1488 	 * always be attached.
1489 	 * The precise detection logic is same as the one for prefixes.
1490 	 *
1491 	 * XXXRW: in6_ifaddrhead locking.
1492 	 */
1493 	TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1494 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1495 			continue;
1496 
1497 		if (ifa->ia6_ndpr == NULL) {
1498 			/*
1499 			 * This can happen when we first configure the address
1500 			 * (i.e. the address exists, but the prefix does not).
1501 			 * XXX: complicated relationships...
1502 			 */
1503 			continue;
1504 		}
1505 
1506 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1507 			break;
1508 	}
1509 	if (ifa) {
1510 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1511 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1512 				continue;
1513 
1514 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1515 				continue;
1516 
1517 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1518 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1519 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1520 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1521 					nd6_dad_start((struct ifaddr *)ifa, 0);
1522 				}
1523 			} else {
1524 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1525 			}
1526 		}
1527 	}
1528 	else {
1529 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1530 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1531 				continue;
1532 
1533 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1534 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1535 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1536 				/* Do we need a delay in this case? */
1537 				nd6_dad_start((struct ifaddr *)ifa, 0);
1538 			}
1539 		}
1540 	}
1541 }
1542 
1543 static int
1544 nd6_prefix_onlink(struct nd_prefix *pr)
1545 {
1546 	struct ifaddr *ifa;
1547 	struct ifnet *ifp = pr->ndpr_ifp;
1548 	struct sockaddr_in6 mask6;
1549 	struct nd_prefix *opr;
1550 	u_long rtflags;
1551 	int error = 0;
1552 	struct radix_node_head *rnh;
1553 	struct rtentry *rt = NULL;
1554 	char ip6buf[INET6_ADDRSTRLEN];
1555 	struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1556 
1557 	/* sanity check */
1558 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1559 		nd6log((LOG_ERR,
1560 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1561 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1562 		    pr->ndpr_plen));
1563 		return (EEXIST);
1564 	}
1565 
1566 	/*
1567 	 * Add the interface route associated with the prefix.  Before
1568 	 * installing the route, check if there's the same prefix on another
1569 	 * interface, and the prefix has already installed the interface route.
1570 	 * Although such a configuration is expected to be rare, we explicitly
1571 	 * allow it.
1572 	 */
1573 	LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1574 		if (opr == pr)
1575 			continue;
1576 
1577 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1578 			continue;
1579 
1580 		if (opr->ndpr_plen == pr->ndpr_plen &&
1581 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1582 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1583 			return (0);
1584 	}
1585 
1586 	/*
1587 	 * We prefer link-local addresses as the associated interface address.
1588 	 */
1589 	/* search for a link-local addr */
1590 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1591 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1592 	if (ifa == NULL) {
1593 		/* XXX: freebsd does not have ifa_ifwithaf */
1594 		IF_ADDR_RLOCK(ifp);
1595 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1596 			if (ifa->ifa_addr->sa_family == AF_INET6)
1597 				break;
1598 		}
1599 		if (ifa != NULL)
1600 			ifa_ref(ifa);
1601 		IF_ADDR_RUNLOCK(ifp);
1602 		/* should we care about ia6_flags? */
1603 	}
1604 	if (ifa == NULL) {
1605 		/*
1606 		 * This can still happen, when, for example, we receive an RA
1607 		 * containing a prefix with the L bit set and the A bit clear,
1608 		 * after removing all IPv6 addresses on the receiving
1609 		 * interface.  This should, of course, be rare though.
1610 		 */
1611 		nd6log((LOG_NOTICE,
1612 		    "nd6_prefix_onlink: failed to find any ifaddr"
1613 		    " to add route for a prefix(%s/%d) on %s\n",
1614 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1615 		    pr->ndpr_plen, if_name(ifp)));
1616 		return (0);
1617 	}
1618 
1619 	/*
1620 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1621 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1622 	 */
1623 	bzero(&mask6, sizeof(mask6));
1624 	mask6.sin6_len = sizeof(mask6);
1625 	mask6.sin6_addr = pr->ndpr_mask;
1626 	rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1627 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1628 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1629 	if (error == 0) {
1630 		if (rt != NULL) /* this should be non NULL, though */ {
1631 			rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1632 			/* XXX what if rhn == NULL? */
1633 			RADIX_NODE_HEAD_LOCK(rnh);
1634 			RT_LOCK(rt);
1635 			if (!rt_setgate(rt, rt_key(rt), (struct sockaddr *)&null_sdl)) {
1636 				((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1637 					rt->rt_ifp->if_type;
1638 				((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1639 					rt->rt_ifp->if_index;
1640 			}
1641 			RADIX_NODE_HEAD_UNLOCK(rnh);
1642 			nd6_rtmsg(RTM_ADD, rt);
1643 			RT_UNLOCK(rt);
1644 		}
1645 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1646 	} else {
1647 		char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
1648 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1649 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1650 		    "errno = %d\n",
1651 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1652 		    pr->ndpr_plen, if_name(ifp),
1653 		    ip6_sprintf(ip6bufg, &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1654 		    ip6_sprintf(ip6bufm, &mask6.sin6_addr), rtflags, error));
1655 	}
1656 
1657 	if (rt != NULL) {
1658 		RT_LOCK(rt);
1659 		RT_REMREF(rt);
1660 		RT_UNLOCK(rt);
1661 	}
1662 	if (ifa != NULL)
1663 		ifa_free(ifa);
1664 
1665 	return (error);
1666 }
1667 
1668 static int
1669 nd6_prefix_offlink(struct nd_prefix *pr)
1670 {
1671 	int error = 0;
1672 	struct ifnet *ifp = pr->ndpr_ifp;
1673 	struct nd_prefix *opr;
1674 	struct sockaddr_in6 sa6, mask6;
1675 	struct rtentry *rt = NULL;
1676 	char ip6buf[INET6_ADDRSTRLEN];
1677 
1678 	/* sanity check */
1679 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1680 		nd6log((LOG_ERR,
1681 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1682 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1683 		    pr->ndpr_plen));
1684 		return (EEXIST);
1685 	}
1686 
1687 	bzero(&sa6, sizeof(sa6));
1688 	sa6.sin6_family = AF_INET6;
1689 	sa6.sin6_len = sizeof(sa6);
1690 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1691 	    sizeof(struct in6_addr));
1692 	bzero(&mask6, sizeof(mask6));
1693 	mask6.sin6_family = AF_INET6;
1694 	mask6.sin6_len = sizeof(sa6);
1695 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1696 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1697 	    (struct sockaddr *)&mask6, 0, &rt);
1698 	if (error == 0) {
1699 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1700 
1701 		/* report the route deletion to the routing socket. */
1702 		if (rt != NULL)
1703 			nd6_rtmsg(RTM_DELETE, rt);
1704 
1705 		/*
1706 		 * There might be the same prefix on another interface,
1707 		 * the prefix which could not be on-link just because we have
1708 		 * the interface route (see comments in nd6_prefix_onlink).
1709 		 * If there's one, try to make the prefix on-link on the
1710 		 * interface.
1711 		 */
1712 		LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1713 			if (opr == pr)
1714 				continue;
1715 
1716 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1717 				continue;
1718 
1719 			/*
1720 			 * KAME specific: detached prefixes should not be
1721 			 * on-link.
1722 			 */
1723 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1724 				continue;
1725 
1726 			if (opr->ndpr_plen == pr->ndpr_plen &&
1727 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1728 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1729 				int e;
1730 
1731 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1732 					nd6log((LOG_ERR,
1733 					    "nd6_prefix_offlink: failed to "
1734 					    "recover a prefix %s/%d from %s "
1735 					    "to %s (errno = %d)\n",
1736 					    ip6_sprintf(ip6buf,
1737 						&opr->ndpr_prefix.sin6_addr),
1738 					    opr->ndpr_plen, if_name(ifp),
1739 					    if_name(opr->ndpr_ifp), e));
1740 				}
1741 			}
1742 		}
1743 	} else {
1744 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1745 		nd6log((LOG_ERR,
1746 		    "nd6_prefix_offlink: failed to delete route: "
1747 		    "%s/%d on %s (errno = %d)\n",
1748 		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1749 		    if_name(ifp), error));
1750 	}
1751 
1752 	if (rt != NULL) {
1753 		RTFREE(rt);
1754 	}
1755 
1756 	return (error);
1757 }
1758 
1759 static struct in6_ifaddr *
1760 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1761 {
1762 	struct ifnet *ifp = pr->ndpr_ifp;
1763 	struct ifaddr *ifa;
1764 	struct in6_aliasreq ifra;
1765 	struct in6_ifaddr *ia, *ib;
1766 	int error, plen0;
1767 	struct in6_addr mask;
1768 	int prefixlen = pr->ndpr_plen;
1769 	int updateflags;
1770 	char ip6buf[INET6_ADDRSTRLEN];
1771 
1772 	in6_prefixlen2mask(&mask, prefixlen);
1773 
1774 	/*
1775 	 * find a link-local address (will be interface ID).
1776 	 * Is it really mandatory? Theoretically, a global or a site-local
1777 	 * address can be configured without a link-local address, if we
1778 	 * have a unique interface identifier...
1779 	 *
1780 	 * it is not mandatory to have a link-local address, we can generate
1781 	 * interface identifier on the fly.  we do this because:
1782 	 * (1) it should be the easiest way to find interface identifier.
1783 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1784 	 * for multiple addresses on a single interface, and possible shortcut
1785 	 * of DAD.  we omitted DAD for this reason in the past.
1786 	 * (3) a user can prevent autoconfiguration of global address
1787 	 * by removing link-local address by hand (this is partly because we
1788 	 * don't have other way to control the use of IPv6 on an interface.
1789 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1790 	 * (4) it is easier to manage when an interface has addresses
1791 	 * with the same interface identifier, than to have multiple addresses
1792 	 * with different interface identifiers.
1793 	 */
1794 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1795 	if (ifa)
1796 		ib = (struct in6_ifaddr *)ifa;
1797 	else
1798 		return NULL;
1799 
1800 	/* prefixlen + ifidlen must be equal to 128 */
1801 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1802 	if (prefixlen != plen0) {
1803 		ifa_free(ifa);
1804 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1805 		    "(prefix=%d ifid=%d)\n",
1806 		    if_name(ifp), prefixlen, 128 - plen0));
1807 		return NULL;
1808 	}
1809 
1810 	/* make ifaddr */
1811 
1812 	bzero(&ifra, sizeof(ifra));
1813 	/*
1814 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1815 	 * for safety.
1816 	 */
1817 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1818 	ifra.ifra_addr.sin6_family = AF_INET6;
1819 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1820 	/* prefix */
1821 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1822 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1823 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1824 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1825 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1826 
1827 	/* interface ID */
1828 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1829 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1830 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1831 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1832 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1833 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1834 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1835 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1836 	ifa_free(ifa);
1837 
1838 	/* new prefix mask. */
1839 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1840 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1841 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1842 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1843 
1844 	/* lifetimes. */
1845 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1846 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1847 
1848 	/* XXX: scope zone ID? */
1849 
1850 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1851 
1852 	/*
1853 	 * Make sure that we do not have this address already.  This should
1854 	 * usually not happen, but we can still see this case, e.g., if we
1855 	 * have manually configured the exact address to be configured.
1856 	 */
1857 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1858 	    &ifra.ifra_addr.sin6_addr);
1859 	if (ifa != NULL) {
1860 		ifa_free(ifa);
1861 		/* this should be rare enough to make an explicit log */
1862 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1863 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1864 		return (NULL);
1865 	}
1866 
1867 	/*
1868 	 * Allocate ifaddr structure, link into chain, etc.
1869 	 * If we are going to create a new address upon receiving a multicasted
1870 	 * RA, we need to impose a random delay before starting DAD.
1871 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1872 	 */
1873 	updateflags = 0;
1874 	if (mcast)
1875 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1876 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1877 		nd6log((LOG_ERR,
1878 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1879 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1880 		    if_name(ifp), error));
1881 		return (NULL);	/* ifaddr must not have been allocated. */
1882 	}
1883 
1884 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1885 	/*
1886 	 * XXXRW: Assumption of non-NULLness here might not be true with
1887 	 * fine-grained locking -- should we validate it?  Or just return
1888 	 * earlier ifa rather than looking it up again?
1889 	 */
1890 	return (ia);		/* this is always non-NULL  and referenced. */
1891 }
1892 
1893 /*
1894  * ia0 - corresponding public address
1895  */
1896 int
1897 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1898 {
1899 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1900 	struct in6_ifaddr *newia, *ia;
1901 	struct in6_aliasreq ifra;
1902 	int i, error;
1903 	int trylimit = 3;	/* XXX: adhoc value */
1904 	int updateflags;
1905 	u_int32_t randid[2];
1906 	time_t vltime0, pltime0;
1907 
1908 	bzero(&ifra, sizeof(ifra));
1909 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1910 	ifra.ifra_addr = ia0->ia_addr;
1911 	/* copy prefix mask */
1912 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1913 	/* clear the old IFID */
1914 	for (i = 0; i < 4; i++) {
1915 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1916 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1917 	}
1918 
1919   again:
1920 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1921 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1922 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1923 		    "random IFID\n"));
1924 		return (EINVAL);
1925 	}
1926 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1927 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1928 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1929 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1930 
1931 	/*
1932 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1933 	 * However, we may still have a chance to see collision, because
1934 	 * there may be a time lag between generation of the ID and generation
1935 	 * of the address.  So, we'll do one more sanity check.
1936 	 */
1937 	IN6_IFADDR_RLOCK();
1938 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1939 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1940 		    &ifra.ifra_addr.sin6_addr)) {
1941 			if (trylimit-- == 0) {
1942 				IN6_IFADDR_RUNLOCK();
1943 				/*
1944 				 * Give up.  Something strange should have
1945 				 * happened.
1946 				 */
1947 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1948 				    "find a unique random IFID\n"));
1949 				return (EEXIST);
1950 			}
1951 			IN6_IFADDR_RUNLOCK();
1952 			forcegen = 1;
1953 			goto again;
1954 		}
1955 	}
1956 	IN6_IFADDR_RUNLOCK();
1957 
1958 	/*
1959 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1960          * public address or TEMP_VALID_LIFETIME.
1961 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1962          * of the public address or TEMP_PREFERRED_LIFETIME -
1963          * DESYNC_FACTOR.
1964 	 */
1965 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1966 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1967 		    (ia0->ia6_lifetime.ia6t_vltime -
1968 		    (time_second - ia0->ia6_updatetime));
1969 		if (vltime0 > V_ip6_temp_valid_lifetime)
1970 			vltime0 = V_ip6_temp_valid_lifetime;
1971 	} else
1972 		vltime0 = V_ip6_temp_valid_lifetime;
1973 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1974 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1975 		    (ia0->ia6_lifetime.ia6t_pltime -
1976 		    (time_second - ia0->ia6_updatetime));
1977 		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
1978 			pltime0 = V_ip6_temp_preferred_lifetime -
1979 			    V_ip6_desync_factor;
1980 		}
1981 	} else
1982 		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
1983 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1984 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1985 
1986 	/*
1987 	 * A temporary address is created only if this calculated Preferred
1988 	 * Lifetime is greater than REGEN_ADVANCE time units.
1989 	 */
1990 	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
1991 		return (0);
1992 
1993 	/* XXX: scope zone ID? */
1994 
1995 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1996 
1997 	/* allocate ifaddr structure, link into chain, etc. */
1998 	updateflags = 0;
1999 	if (delay)
2000 		updateflags |= IN6_IFAUPDATE_DADDELAY;
2001 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2002 		return (error);
2003 
2004 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2005 	if (newia == NULL) {	/* XXX: can it happen? */
2006 		nd6log((LOG_ERR,
2007 		    "in6_tmpifadd: ifa update succeeded, but we got "
2008 		    "no ifaddr\n"));
2009 		return (EINVAL); /* XXX */
2010 	}
2011 	newia->ia6_ndpr = ia0->ia6_ndpr;
2012 	newia->ia6_ndpr->ndpr_refcnt++;
2013 	ifa_free(&newia->ia_ifa);
2014 
2015 	/*
2016 	 * A newly added address might affect the status of other addresses.
2017 	 * XXX: when the temporary address is generated with a new public
2018 	 * address, the onlink check is redundant.  However, it would be safe
2019 	 * to do the check explicitly everywhere a new address is generated,
2020 	 * and, in fact, we surely need the check when we create a new
2021 	 * temporary address due to deprecation of an old temporary address.
2022 	 */
2023 	pfxlist_onlink_check();
2024 
2025 	return (0);
2026 }
2027 
2028 static int
2029 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2030 {
2031 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2032 		ndpr->ndpr_preferred = 0;
2033 	else
2034 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2035 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2036 		ndpr->ndpr_expire = 0;
2037 	else
2038 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2039 
2040 	return 0;
2041 }
2042 
2043 static void
2044 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2045 {
2046 	/* init ia6t_expire */
2047 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2048 		lt6->ia6t_expire = 0;
2049 	else {
2050 		lt6->ia6t_expire = time_second;
2051 		lt6->ia6t_expire += lt6->ia6t_vltime;
2052 	}
2053 
2054 	/* init ia6t_preferred */
2055 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2056 		lt6->ia6t_preferred = 0;
2057 	else {
2058 		lt6->ia6t_preferred = time_second;
2059 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2060 	}
2061 }
2062 
2063 /*
2064  * Delete all the routing table entries that use the specified gateway.
2065  * XXX: this function causes search through all entries of routing table, so
2066  * it shouldn't be called when acting as a router.
2067  */
2068 void
2069 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2070 {
2071 	struct radix_node_head *rnh;
2072 	int s = splnet();
2073 
2074 	/* We'll care only link-local addresses */
2075 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2076 		splx(s);
2077 		return;
2078 	}
2079 
2080 	rnh = rt_tables_get_rnh(0, AF_INET6);
2081 	if (rnh == NULL)
2082 		return;
2083 
2084 	RADIX_NODE_HEAD_LOCK(rnh);
2085 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2086 	RADIX_NODE_HEAD_UNLOCK(rnh);
2087 	splx(s);
2088 }
2089 
2090 static int
2091 rt6_deleteroute(struct radix_node *rn, void *arg)
2092 {
2093 #define SIN6(s)	((struct sockaddr_in6 *)s)
2094 	struct rtentry *rt = (struct rtentry *)rn;
2095 	struct in6_addr *gate = (struct in6_addr *)arg;
2096 
2097 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2098 		return (0);
2099 
2100 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2101 		return (0);
2102 	}
2103 
2104 	/*
2105 	 * Do not delete a static route.
2106 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2107 	 * 'cloned' bit instead?
2108 	 */
2109 	if ((rt->rt_flags & RTF_STATIC) != 0)
2110 		return (0);
2111 
2112 	/*
2113 	 * We delete only host route. This means, in particular, we don't
2114 	 * delete default route.
2115 	 */
2116 	if ((rt->rt_flags & RTF_HOST) == 0)
2117 		return (0);
2118 
2119 	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2120 	    rt_mask(rt), rt->rt_flags, 0));
2121 #undef SIN6
2122 }
2123 
2124 int
2125 nd6_setdefaultiface(int ifindex)
2126 {
2127 	int error = 0;
2128 
2129 	if (ifindex < 0 || V_if_index < ifindex)
2130 		return (EINVAL);
2131 	if (ifindex != 0 && !ifnet_byindex(ifindex))
2132 		return (EINVAL);
2133 
2134 	if (V_nd6_defifindex != ifindex) {
2135 		V_nd6_defifindex = ifindex;
2136 		if (V_nd6_defifindex > 0)
2137 			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2138 		else
2139 			V_nd6_defifp = NULL;
2140 
2141 		/*
2142 		 * Our current implementation assumes one-to-one maping between
2143 		 * interfaces and links, so it would be natural to use the
2144 		 * default interface as the default link.
2145 		 */
2146 		scope6_setdefault(V_nd6_defifp);
2147 	}
2148 
2149 	return (error);
2150 }
2151