xref: /freebsd/sys/netinet6/nd6_rtr.c (revision 67897cfb90605ac431a8d78beca50602708173a4)
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(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(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(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_fib(cmd, &info, rt->rt_flags, 0, rt->rt_fibnum);
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 error;
477 
478 	bzero(&def, sizeof(def));
479 	bzero(&mask, sizeof(mask));
480 	bzero(&gate, sizeof(gate));
481 
482 	def.sin6_len = mask.sin6_len = gate.sin6_len =
483 	    sizeof(struct sockaddr_in6);
484 	def.sin6_family = gate.sin6_family = AF_INET6;
485 	gate.sin6_addr = new->rtaddr;
486 
487 	error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&def,
488 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
489 	    RTF_GATEWAY, &newrt, RT_DEFAULT_FIB);
490 	if (newrt) {
491 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
492 		RTFREE(newrt);
493 	}
494 	if (error == 0)
495 		new->installed = 1;
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 	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
505 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
506 			return (dr);
507 	}
508 
509 	return (NULL);		/* search failed */
510 }
511 
512 /*
513  * Remove the default route for a given router.
514  * This is just a subroutine function for defrouter_select(), and should
515  * not be called from anywhere else.
516  */
517 static void
518 defrouter_delreq(struct nd_defrouter *dr)
519 {
520 	struct sockaddr_in6 def, mask, gate;
521 	struct rtentry *oldrt = NULL;
522 
523 	bzero(&def, sizeof(def));
524 	bzero(&mask, sizeof(mask));
525 	bzero(&gate, sizeof(gate));
526 
527 	def.sin6_len = mask.sin6_len = gate.sin6_len =
528 	    sizeof(struct sockaddr_in6);
529 	def.sin6_family = gate.sin6_family = AF_INET6;
530 	gate.sin6_addr = dr->rtaddr;
531 
532 	in6_rtrequest(RTM_DELETE, (struct sockaddr *)&def,
533 	    (struct sockaddr *)&gate,
534 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, RT_DEFAULT_FIB);
535 	if (oldrt) {
536 		nd6_rtmsg(RTM_DELETE, oldrt);
537 		RTFREE(oldrt);
538 	}
539 
540 	dr->installed = 0;
541 }
542 
543 /*
544  * remove all default routes from default router list
545  */
546 void
547 defrouter_reset(void)
548 {
549 	struct nd_defrouter *dr;
550 
551 	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry)
552 		defrouter_delreq(dr);
553 
554 	/*
555 	 * XXX should we also nuke any default routers in the kernel, by
556 	 * going through them by rtalloc1()?
557 	 */
558 }
559 
560 void
561 defrtrlist_del(struct nd_defrouter *dr)
562 {
563 	struct nd_defrouter *deldr = NULL;
564 	struct nd_prefix *pr;
565 
566 	/*
567 	 * Flush all the routing table entries that use the router
568 	 * as a next hop.
569 	 */
570 	if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
571 		rt6_flush(&dr->rtaddr, dr->ifp);
572 
573 	if (dr->installed) {
574 		deldr = dr;
575 		defrouter_delreq(dr);
576 	}
577 	TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
578 
579 	/*
580 	 * Also delete all the pointers to the router in each prefix lists.
581 	 */
582 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
583 		struct nd_pfxrouter *pfxrtr;
584 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
585 			pfxrtr_del(pfxrtr);
586 	}
587 	pfxlist_onlink_check();
588 
589 	/*
590 	 * If the router is the primary one, choose a new one.
591 	 * Note that defrouter_select() will remove the current gateway
592 	 * from the routing table.
593 	 */
594 	if (deldr)
595 		defrouter_select();
596 
597 	free(dr, M_IP6NDP);
598 }
599 
600 /*
601  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
602  * draft-ietf-ipngwg-router-selection:
603  * 1) Routers that are reachable or probably reachable should be preferred.
604  *    If we have more than one (probably) reachable router, prefer ones
605  *    with the highest router preference.
606  * 2) When no routers on the list are known to be reachable or
607  *    probably reachable, routers SHOULD be selected in a round-robin
608  *    fashion, regardless of router preference values.
609  * 3) If the Default Router List is empty, assume that all
610  *    destinations are on-link.
611  *
612  * We assume nd_defrouter is sorted by router preference value.
613  * Since the code below covers both with and without router preference cases,
614  * we do not need to classify the cases by ifdef.
615  *
616  * At this moment, we do not try to install more than one default router,
617  * even when the multipath routing is available, because we're not sure about
618  * the benefits for stub hosts comparing to the risk of making the code
619  * complicated and the possibility of introducing bugs.
620  */
621 void
622 defrouter_select(void)
623 {
624 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
625 	struct llentry *ln = NULL;
626 
627 	/*
628 	 * Let's handle easy case (3) first:
629 	 * If default router list is empty, there's nothing to be done.
630 	 */
631 	if (TAILQ_EMPTY(&V_nd_defrouter))
632 		return;
633 
634 	/*
635 	 * Search for a (probably) reachable router from the list.
636 	 * We just pick up the first reachable one (if any), assuming that
637 	 * the ordering rule of the list described in defrtrlist_update().
638 	 */
639 	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
640 		IF_AFDATA_RLOCK(dr->ifp);
641 		if (selected_dr == NULL &&
642 		    (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
643 		    ND6_IS_LLINFO_PROBREACH(ln)) {
644 			selected_dr = dr;
645 		}
646 		IF_AFDATA_RUNLOCK(dr->ifp);
647 		if (ln != NULL) {
648 			LLE_RUNLOCK(ln);
649 			ln = NULL;
650 		}
651 
652 		if (dr->installed && installed_dr == NULL)
653 			installed_dr = dr;
654 		else if (dr->installed && installed_dr) {
655 			/* this should not happen.  warn for diagnosis. */
656 			log(LOG_ERR, "defrouter_select: more than one router"
657 			    " is installed\n");
658 		}
659 	}
660 	/*
661 	 * If none of the default routers was found to be reachable,
662 	 * round-robin the list regardless of preference.
663 	 * Otherwise, if we have an installed router, check if the selected
664 	 * (reachable) router should really be preferred to the installed one.
665 	 * We only prefer the new router when the old one is not reachable
666 	 * or when the new one has a really higher preference value.
667 	 */
668 	if (selected_dr == NULL) {
669 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
670 			selected_dr = TAILQ_FIRST(&V_nd_defrouter);
671 		else
672 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
673 	} else if (installed_dr) {
674 		IF_AFDATA_RLOCK(installed_dr->ifp);
675 		if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
676 		    ND6_IS_LLINFO_PROBREACH(ln) &&
677 		    rtpref(selected_dr) <= rtpref(installed_dr)) {
678 			selected_dr = installed_dr;
679 		}
680 		IF_AFDATA_RUNLOCK(installed_dr->ifp);
681 		if (ln != NULL)
682 			LLE_RUNLOCK(ln);
683 	}
684 
685 	/*
686 	 * If the selected router is different than the installed one,
687 	 * remove the installed router and install the selected one.
688 	 * Note that the selected router is never NULL here.
689 	 */
690 	if (installed_dr != selected_dr) {
691 		if (installed_dr)
692 			defrouter_delreq(installed_dr);
693 		defrouter_addreq(selected_dr);
694 	}
695 
696 	return;
697 }
698 
699 /*
700  * for default router selection
701  * regards router-preference field as a 2-bit signed integer
702  */
703 static int
704 rtpref(struct nd_defrouter *dr)
705 {
706 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
707 	case ND_RA_FLAG_RTPREF_HIGH:
708 		return (RTPREF_HIGH);
709 	case ND_RA_FLAG_RTPREF_MEDIUM:
710 	case ND_RA_FLAG_RTPREF_RSV:
711 		return (RTPREF_MEDIUM);
712 	case ND_RA_FLAG_RTPREF_LOW:
713 		return (RTPREF_LOW);
714 	default:
715 		/*
716 		 * This case should never happen.  If it did, it would mean a
717 		 * serious bug of kernel internal.  We thus always bark here.
718 		 * Or, can we even panic?
719 		 */
720 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
721 		return (RTPREF_INVALID);
722 	}
723 	/* NOTREACHED */
724 }
725 
726 static struct nd_defrouter *
727 defrtrlist_update(struct nd_defrouter *new)
728 {
729 	struct nd_defrouter *dr, *n;
730 
731 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
732 		/* entry exists */
733 		if (new->rtlifetime == 0) {
734 			defrtrlist_del(dr);
735 			dr = NULL;
736 		} else {
737 			int oldpref = rtpref(dr);
738 
739 			/* override */
740 			dr->flags = new->flags; /* xxx flag check */
741 			dr->rtlifetime = new->rtlifetime;
742 			dr->expire = new->expire;
743 
744 			/*
745 			 * If the preference does not change, there's no need
746 			 * to sort the entries. Also make sure the selected
747 			 * router is still installed in the kernel.
748 			 */
749 			if (dr->installed && rtpref(new) == oldpref)
750 				return (dr);
751 
752 			/*
753 			 * preferred router may be changed, so relocate
754 			 * this router.
755 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
756 			 * However, since defrtrlist_del() has many side
757 			 * effects, we intentionally do so here.
758 			 * defrouter_select() below will handle routing
759 			 * changes later.
760 			 */
761 			TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
762 			n = dr;
763 			goto insert;
764 		}
765 		return (dr);
766 	}
767 
768 	/* entry does not exist */
769 	if (new->rtlifetime == 0)
770 		return (NULL);
771 
772 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
773 	if (n == NULL)
774 		return (NULL);
775 	bzero(n, sizeof(*n));
776 	*n = *new;
777 
778 insert:
779 	/*
780 	 * Insert the new router in the Default Router List;
781 	 * The Default Router List should be in the descending order
782 	 * of router-preferece.  Routers with the same preference are
783 	 * sorted in the arriving time order.
784 	 */
785 
786 	/* insert at the end of the group */
787 	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
788 		if (rtpref(n) > rtpref(dr))
789 			break;
790 	}
791 	if (dr)
792 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
793 	else
794 		TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
795 
796 	defrouter_select();
797 
798 	return (n);
799 }
800 
801 static struct nd_pfxrouter *
802 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
803 {
804 	struct nd_pfxrouter *search;
805 
806 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
807 		if (search->router == dr)
808 			break;
809 	}
810 
811 	return (search);
812 }
813 
814 static void
815 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
816 {
817 	struct nd_pfxrouter *new;
818 
819 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
820 	if (new == NULL)
821 		return;
822 	bzero(new, sizeof(*new));
823 	new->router = dr;
824 
825 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
826 
827 	pfxlist_onlink_check();
828 }
829 
830 static void
831 pfxrtr_del(struct nd_pfxrouter *pfr)
832 {
833 	LIST_REMOVE(pfr, pfr_entry);
834 	free(pfr, M_IP6NDP);
835 }
836 
837 struct nd_prefix *
838 nd6_prefix_lookup(struct nd_prefixctl *key)
839 {
840 	struct nd_prefix *search;
841 
842 	LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
843 		if (key->ndpr_ifp == search->ndpr_ifp &&
844 		    key->ndpr_plen == search->ndpr_plen &&
845 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
846 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
847 			break;
848 		}
849 	}
850 
851 	return (search);
852 }
853 
854 int
855 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
856     struct nd_prefix **newp)
857 {
858 	struct nd_prefix *new = NULL;
859 	int error = 0;
860 	int i;
861 	char ip6buf[INET6_ADDRSTRLEN];
862 
863 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
864 	if (new == NULL)
865 		return(ENOMEM);
866 	bzero(new, sizeof(*new));
867 	new->ndpr_ifp = pr->ndpr_ifp;
868 	new->ndpr_prefix = pr->ndpr_prefix;
869 	new->ndpr_plen = pr->ndpr_plen;
870 	new->ndpr_vltime = pr->ndpr_vltime;
871 	new->ndpr_pltime = pr->ndpr_pltime;
872 	new->ndpr_flags = pr->ndpr_flags;
873 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
874 		free(new, M_IP6NDP);
875 		return(error);
876 	}
877 	new->ndpr_lastupdate = time_second;
878 	if (newp != NULL)
879 		*newp = new;
880 
881 	/* initialization */
882 	LIST_INIT(&new->ndpr_advrtrs);
883 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
884 	/* make prefix in the canonical form */
885 	for (i = 0; i < 4; i++)
886 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
887 		    new->ndpr_mask.s6_addr32[i];
888 
889 	/* link ndpr_entry to nd_prefix list */
890 	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
891 
892 	/* ND_OPT_PI_FLAG_ONLINK processing */
893 	if (new->ndpr_raf_onlink) {
894 		int e;
895 
896 		if ((e = nd6_prefix_onlink(new)) != 0) {
897 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
898 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
899 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
900 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
901 			/* proceed anyway. XXX: is it correct? */
902 		}
903 	}
904 
905 	if (dr)
906 		pfxrtr_add(new, dr);
907 
908 	return 0;
909 }
910 
911 void
912 prelist_remove(struct nd_prefix *pr)
913 {
914 	struct nd_pfxrouter *pfr, *next;
915 	int e;
916 	char ip6buf[INET6_ADDRSTRLEN];
917 
918 	/* make sure to invalidate the prefix until it is really freed. */
919 	pr->ndpr_vltime = 0;
920 	pr->ndpr_pltime = 0;
921 
922 	/*
923 	 * Though these flags are now meaningless, we'd rather keep the value
924 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
925 	 * when executing "ndp -p".
926 	 */
927 
928 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
929 	    (e = nd6_prefix_offlink(pr)) != 0) {
930 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
931 		    "on %s, errno=%d\n",
932 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
933 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
934 		/* what should we do? */
935 	}
936 
937 	if (pr->ndpr_refcnt > 0)
938 		return;		/* notice here? */
939 
940 	/* unlink ndpr_entry from nd_prefix list */
941 	LIST_REMOVE(pr, ndpr_entry);
942 
943 	/* free list of routers that adversed the prefix */
944 	LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next) {
945 		free(pfr, M_IP6NDP);
946 	}
947 	free(pr, M_IP6NDP);
948 
949 	pfxlist_onlink_check();
950 }
951 
952 /*
953  * dr - may be NULL
954  */
955 
956 static int
957 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
958     struct mbuf *m, int mcast)
959 {
960 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
961 	struct ifaddr *ifa;
962 	struct ifnet *ifp = new->ndpr_ifp;
963 	struct nd_prefix *pr;
964 	int error = 0;
965 	int newprefix = 0;
966 	int auth;
967 	struct in6_addrlifetime lt6_tmp;
968 	char ip6buf[INET6_ADDRSTRLEN];
969 
970 	auth = 0;
971 	if (m) {
972 		/*
973 		 * Authenticity for NA consists authentication for
974 		 * both IP header and IP datagrams, doesn't it ?
975 		 */
976 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
977 		auth = ((m->m_flags & M_AUTHIPHDR) &&
978 		    (m->m_flags & M_AUTHIPDGM));
979 #endif
980 	}
981 
982 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
983 		/*
984 		 * nd6_prefix_lookup() ensures that pr and new have the same
985 		 * prefix on a same interface.
986 		 */
987 
988 		/*
989 		 * Update prefix information.  Note that the on-link (L) bit
990 		 * and the autonomous (A) bit should NOT be changed from 1
991 		 * to 0.
992 		 */
993 		if (new->ndpr_raf_onlink == 1)
994 			pr->ndpr_raf_onlink = 1;
995 		if (new->ndpr_raf_auto == 1)
996 			pr->ndpr_raf_auto = 1;
997 		if (new->ndpr_raf_onlink) {
998 			pr->ndpr_vltime = new->ndpr_vltime;
999 			pr->ndpr_pltime = new->ndpr_pltime;
1000 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1001 			pr->ndpr_lastupdate = time_second;
1002 		}
1003 
1004 		if (new->ndpr_raf_onlink &&
1005 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1006 			int e;
1007 
1008 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1009 				nd6log((LOG_ERR,
1010 				    "prelist_update: failed to make "
1011 				    "the prefix %s/%d on-link on %s "
1012 				    "(errno=%d)\n",
1013 				    ip6_sprintf(ip6buf,
1014 					    &pr->ndpr_prefix.sin6_addr),
1015 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1016 				/* proceed anyway. XXX: is it correct? */
1017 			}
1018 		}
1019 
1020 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1021 			pfxrtr_add(pr, dr);
1022 	} else {
1023 		struct nd_prefix *newpr = NULL;
1024 
1025 		newprefix = 1;
1026 
1027 		if (new->ndpr_vltime == 0)
1028 			goto end;
1029 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1030 			goto end;
1031 
1032 		error = nd6_prelist_add(new, dr, &newpr);
1033 		if (error != 0 || newpr == NULL) {
1034 			nd6log((LOG_NOTICE, "prelist_update: "
1035 			    "nd6_prelist_add failed for %s/%d on %s "
1036 			    "errno=%d, returnpr=%p\n",
1037 			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1038 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1039 			    error, newpr));
1040 			goto end; /* we should just give up in this case. */
1041 		}
1042 
1043 		/*
1044 		 * XXX: from the ND point of view, we can ignore a prefix
1045 		 * with the on-link bit being zero.  However, we need a
1046 		 * prefix structure for references from autoconfigured
1047 		 * addresses.  Thus, we explicitly make sure that the prefix
1048 		 * itself expires now.
1049 		 */
1050 		if (newpr->ndpr_raf_onlink == 0) {
1051 			newpr->ndpr_vltime = 0;
1052 			newpr->ndpr_pltime = 0;
1053 			in6_init_prefix_ltimes(newpr);
1054 		}
1055 
1056 		pr = newpr;
1057 	}
1058 
1059 	/*
1060 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1061 	 * Note that pr must be non NULL at this point.
1062 	 */
1063 
1064 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1065 	if (!new->ndpr_raf_auto)
1066 		goto end;
1067 
1068 	/*
1069 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1070 	 * nd6_ra_input.
1071 	 */
1072 
1073 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1074 	if (new->ndpr_pltime > new->ndpr_vltime) {
1075 		error = EINVAL;	/* XXX: won't be used */
1076 		goto end;
1077 	}
1078 
1079 	/*
1080 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1081 	 * an address configured by stateless autoconfiguration already in the
1082 	 * list of addresses associated with the interface, and the Valid
1083 	 * Lifetime is not 0, form an address.  We first check if we have
1084 	 * a matching prefix.
1085 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1086 	 * consider autoconfigured addresses while RFC2462 simply said
1087 	 * "address".
1088 	 */
1089 	IF_ADDR_RLOCK(ifp);
1090 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1091 		struct in6_ifaddr *ifa6;
1092 		u_int32_t remaininglifetime;
1093 
1094 		if (ifa->ifa_addr->sa_family != AF_INET6)
1095 			continue;
1096 
1097 		ifa6 = (struct in6_ifaddr *)ifa;
1098 
1099 		/*
1100 		 * We only consider autoconfigured addresses as per rfc2462bis.
1101 		 */
1102 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1103 			continue;
1104 
1105 		/*
1106 		 * Spec is not clear here, but I believe we should concentrate
1107 		 * on unicast (i.e. not anycast) addresses.
1108 		 * XXX: other ia6_flags? detached or duplicated?
1109 		 */
1110 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1111 			continue;
1112 
1113 		/*
1114 		 * Ignore the address if it is not associated with a prefix
1115 		 * or is associated with a prefix that is different from this
1116 		 * one.  (pr is never NULL here)
1117 		 */
1118 		if (ifa6->ia6_ndpr != pr)
1119 			continue;
1120 
1121 		if (ia6_match == NULL) /* remember the first one */
1122 			ia6_match = ifa6;
1123 
1124 		/*
1125 		 * An already autoconfigured address matched.  Now that we
1126 		 * are sure there is at least one matched address, we can
1127 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1128 		 * "two hours" rule and the privacy extension.
1129 		 * We apply some clarifications in rfc2462bis:
1130 		 * - use remaininglifetime instead of storedlifetime as a
1131 		 *   variable name
1132 		 * - remove the dead code in the "two-hour" rule
1133 		 */
1134 #define TWOHOUR		(120*60)
1135 		lt6_tmp = ifa6->ia6_lifetime;
1136 
1137 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1138 			remaininglifetime = ND6_INFINITE_LIFETIME;
1139 		else if (time_second - ifa6->ia6_updatetime >
1140 			 lt6_tmp.ia6t_vltime) {
1141 			/*
1142 			 * The case of "invalid" address.  We should usually
1143 			 * not see this case.
1144 			 */
1145 			remaininglifetime = 0;
1146 		} else
1147 			remaininglifetime = lt6_tmp.ia6t_vltime -
1148 			    (time_second - ifa6->ia6_updatetime);
1149 
1150 		/* when not updating, keep the current stored lifetime. */
1151 		lt6_tmp.ia6t_vltime = remaininglifetime;
1152 
1153 		if (TWOHOUR < new->ndpr_vltime ||
1154 		    remaininglifetime < new->ndpr_vltime) {
1155 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1156 		} else if (remaininglifetime <= TWOHOUR) {
1157 			if (auth) {
1158 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1159 			}
1160 		} else {
1161 			/*
1162 			 * new->ndpr_vltime <= TWOHOUR &&
1163 			 * TWOHOUR < remaininglifetime
1164 			 */
1165 			lt6_tmp.ia6t_vltime = TWOHOUR;
1166 		}
1167 
1168 		/* The 2 hour rule is not imposed for preferred lifetime. */
1169 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1170 
1171 		in6_init_address_ltimes(pr, &lt6_tmp);
1172 
1173 		/*
1174 		 * We need to treat lifetimes for temporary addresses
1175 		 * differently, according to
1176 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1177 		 * we only update the lifetimes when they are in the maximum
1178 		 * intervals.
1179 		 */
1180 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1181 			u_int32_t maxvltime, maxpltime;
1182 
1183 			if (V_ip6_temp_valid_lifetime >
1184 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1185 			    V_ip6_desync_factor)) {
1186 				maxvltime = V_ip6_temp_valid_lifetime -
1187 				    (time_second - ifa6->ia6_createtime) -
1188 				    V_ip6_desync_factor;
1189 			} else
1190 				maxvltime = 0;
1191 			if (V_ip6_temp_preferred_lifetime >
1192 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1193 			    V_ip6_desync_factor)) {
1194 				maxpltime = V_ip6_temp_preferred_lifetime -
1195 				    (time_second - ifa6->ia6_createtime) -
1196 				    V_ip6_desync_factor;
1197 			} else
1198 				maxpltime = 0;
1199 
1200 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1201 			    lt6_tmp.ia6t_vltime > maxvltime) {
1202 				lt6_tmp.ia6t_vltime = maxvltime;
1203 			}
1204 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1205 			    lt6_tmp.ia6t_pltime > maxpltime) {
1206 				lt6_tmp.ia6t_pltime = maxpltime;
1207 			}
1208 		}
1209 		ifa6->ia6_lifetime = lt6_tmp;
1210 		ifa6->ia6_updatetime = time_second;
1211 	}
1212 	IF_ADDR_RUNLOCK(ifp);
1213 	if (ia6_match == NULL && new->ndpr_vltime) {
1214 		int ifidlen;
1215 
1216 		/*
1217 		 * 5.5.3 (d) (continued)
1218 		 * No address matched and the valid lifetime is non-zero.
1219 		 * Create a new address.
1220 		 */
1221 
1222 		/*
1223 		 * Prefix Length check:
1224 		 * If the sum of the prefix length and interface identifier
1225 		 * length does not equal 128 bits, the Prefix Information
1226 		 * option MUST be ignored.  The length of the interface
1227 		 * identifier is defined in a separate link-type specific
1228 		 * document.
1229 		 */
1230 		ifidlen = in6_if2idlen(ifp);
1231 		if (ifidlen < 0) {
1232 			/* this should not happen, so we always log it. */
1233 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1234 			    if_name(ifp));
1235 			goto end;
1236 		}
1237 		if (ifidlen + pr->ndpr_plen != 128) {
1238 			nd6log((LOG_INFO,
1239 			    "prelist_update: invalid prefixlen "
1240 			    "%d for %s, ignored\n",
1241 			    pr->ndpr_plen, if_name(ifp)));
1242 			goto end;
1243 		}
1244 
1245 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1246 			/*
1247 			 * note that we should use pr (not new) for reference.
1248 			 */
1249 			pr->ndpr_refcnt++;
1250 			ia6->ia6_ndpr = pr;
1251 
1252 			/*
1253 			 * RFC 3041 3.3 (2).
1254 			 * When a new public address is created as described
1255 			 * in RFC2462, also create a new temporary address.
1256 			 *
1257 			 * RFC 3041 3.5.
1258 			 * When an interface connects to a new link, a new
1259 			 * randomized interface identifier should be generated
1260 			 * immediately together with a new set of temporary
1261 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1262 			 * in6_tmpifadd().
1263 			 */
1264 			if (V_ip6_use_tempaddr) {
1265 				int e;
1266 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1267 					nd6log((LOG_NOTICE, "prelist_update: "
1268 					    "failed to create a temporary "
1269 					    "address, errno=%d\n",
1270 					    e));
1271 				}
1272 			}
1273 			ifa_free(&ia6->ia_ifa);
1274 
1275 			/*
1276 			 * A newly added address might affect the status
1277 			 * of other addresses, so we check and update it.
1278 			 * XXX: what if address duplication happens?
1279 			 */
1280 			pfxlist_onlink_check();
1281 		} else {
1282 			/* just set an error. do not bark here. */
1283 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1284 		}
1285 	}
1286 
1287  end:
1288 	return error;
1289 }
1290 
1291 /*
1292  * A supplement function used in the on-link detection below;
1293  * detect if a given prefix has a (probably) reachable advertising router.
1294  * XXX: lengthy function name...
1295  */
1296 static struct nd_pfxrouter *
1297 find_pfxlist_reachable_router(struct nd_prefix *pr)
1298 {
1299 	struct nd_pfxrouter *pfxrtr;
1300 	struct llentry *ln;
1301 	int canreach;
1302 
1303 	LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1304 		IF_AFDATA_RLOCK(pfxrtr->router->ifp);
1305 		ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1306 		IF_AFDATA_RUNLOCK(pfxrtr->router->ifp);
1307 		if (ln == NULL)
1308 			continue;
1309 		canreach = ND6_IS_LLINFO_PROBREACH(ln);
1310 		LLE_RUNLOCK(ln);
1311 		if (canreach)
1312 			break;
1313 	}
1314 	return (pfxrtr);
1315 }
1316 
1317 /*
1318  * Check if each prefix in the prefix list has at least one available router
1319  * that advertised the prefix (a router is "available" if its neighbor cache
1320  * entry is reachable or probably reachable).
1321  * If the check fails, the prefix may be off-link, because, for example,
1322  * we have moved from the network but the lifetime of the prefix has not
1323  * expired yet.  So we should not use the prefix if there is another prefix
1324  * that has an available router.
1325  * But, if there is no prefix that has an available router, we still regards
1326  * all the prefixes as on-link.  This is because we can't tell if all the
1327  * routers are simply dead or if we really moved from the network and there
1328  * is no router around us.
1329  */
1330 void
1331 pfxlist_onlink_check()
1332 {
1333 	struct nd_prefix *pr;
1334 	struct in6_ifaddr *ifa;
1335 	struct nd_defrouter *dr;
1336 	struct nd_pfxrouter *pfxrtr = NULL;
1337 
1338 	/*
1339 	 * Check if there is a prefix that has a reachable advertising
1340 	 * router.
1341 	 */
1342 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1343 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1344 			break;
1345 	}
1346 
1347 	/*
1348 	 * If we have no such prefix, check whether we still have a router
1349 	 * that does not advertise any prefixes.
1350 	 */
1351 	if (pr == NULL) {
1352 		TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1353 			struct nd_prefix *pr0;
1354 
1355 			LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1356 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1357 					break;
1358 			}
1359 			if (pfxrtr != NULL)
1360 				break;
1361 		}
1362 	}
1363 	if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) {
1364 		/*
1365 		 * There is at least one prefix that has a reachable router,
1366 		 * or at least a router which probably does not advertise
1367 		 * any prefixes.  The latter would be the case when we move
1368 		 * to a new link where we have a router that does not provide
1369 		 * prefixes and we configure an address by hand.
1370 		 * Detach prefixes which have no reachable advertising
1371 		 * router, and attach other prefixes.
1372 		 */
1373 		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1374 			/* XXX: a link-local prefix should never be detached */
1375 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1376 				continue;
1377 
1378 			/*
1379 			 * we aren't interested in prefixes without the L bit
1380 			 * set.
1381 			 */
1382 			if (pr->ndpr_raf_onlink == 0)
1383 				continue;
1384 
1385 			if (pr->ndpr_raf_auto == 0)
1386 				continue;
1387 
1388 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1389 			    find_pfxlist_reachable_router(pr) == NULL)
1390 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1391 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1392 			    find_pfxlist_reachable_router(pr) != 0)
1393 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1394 		}
1395 	} else {
1396 		/* there is no prefix that has a reachable router */
1397 		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1398 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1399 				continue;
1400 
1401 			if (pr->ndpr_raf_onlink == 0)
1402 				continue;
1403 
1404 			if (pr->ndpr_raf_auto == 0)
1405 				continue;
1406 
1407 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1408 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1409 		}
1410 	}
1411 
1412 	/*
1413 	 * Remove each interface route associated with a (just) detached
1414 	 * prefix, and reinstall the interface route for a (just) attached
1415 	 * prefix.  Note that all attempt of reinstallation does not
1416 	 * necessarily success, when a same prefix is shared among multiple
1417 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1418 	 * so we don't have to care about them.
1419 	 */
1420 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1421 		int e;
1422 		char ip6buf[INET6_ADDRSTRLEN];
1423 
1424 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1425 			continue;
1426 
1427 		if (pr->ndpr_raf_onlink == 0)
1428 			continue;
1429 
1430 		if (pr->ndpr_raf_auto == 0)
1431 			continue;
1432 
1433 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1434 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1435 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1436 				nd6log((LOG_ERR,
1437 				    "pfxlist_onlink_check: failed to "
1438 				    "make %s/%d offlink, errno=%d\n",
1439 				    ip6_sprintf(ip6buf,
1440 					    &pr->ndpr_prefix.sin6_addr),
1441 					    pr->ndpr_plen, e));
1442 			}
1443 		}
1444 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1445 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1446 		    pr->ndpr_raf_onlink) {
1447 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1448 				nd6log((LOG_ERR,
1449 				    "pfxlist_onlink_check: failed to "
1450 				    "make %s/%d onlink, errno=%d\n",
1451 				    ip6_sprintf(ip6buf,
1452 					    &pr->ndpr_prefix.sin6_addr),
1453 					    pr->ndpr_plen, e));
1454 			}
1455 		}
1456 	}
1457 
1458 	/*
1459 	 * Changes on the prefix status might affect address status as well.
1460 	 * Make sure that all addresses derived from an attached prefix are
1461 	 * attached, and that all addresses derived from a detached prefix are
1462 	 * detached.  Note, however, that a manually configured address should
1463 	 * always be attached.
1464 	 * The precise detection logic is same as the one for prefixes.
1465 	 *
1466 	 * XXXRW: in6_ifaddrhead locking.
1467 	 */
1468 	TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1469 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1470 			continue;
1471 
1472 		if (ifa->ia6_ndpr == NULL) {
1473 			/*
1474 			 * This can happen when we first configure the address
1475 			 * (i.e. the address exists, but the prefix does not).
1476 			 * XXX: complicated relationships...
1477 			 */
1478 			continue;
1479 		}
1480 
1481 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1482 			break;
1483 	}
1484 	if (ifa) {
1485 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1486 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1487 				continue;
1488 
1489 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1490 				continue;
1491 
1492 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1493 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1494 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1495 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1496 					nd6_dad_start((struct ifaddr *)ifa, 0);
1497 				}
1498 			} else {
1499 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1500 			}
1501 		}
1502 	}
1503 	else {
1504 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1505 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1506 				continue;
1507 
1508 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1509 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1510 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1511 				/* Do we need a delay in this case? */
1512 				nd6_dad_start((struct ifaddr *)ifa, 0);
1513 			}
1514 		}
1515 	}
1516 }
1517 
1518 static int
1519 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
1520 {
1521 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1522 	struct radix_node_head *rnh;
1523 	struct rtentry *rt;
1524 	struct sockaddr_in6 mask6;
1525 	u_long rtflags;
1526 	int error, a_failure, fibnum;
1527 
1528 	/*
1529 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1530 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1531 	 */
1532 	bzero(&mask6, sizeof(mask6));
1533 	mask6.sin6_len = sizeof(mask6);
1534 	mask6.sin6_addr = pr->ndpr_mask;
1535 	rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1536 
1537 	a_failure = 0;
1538 	for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1539 
1540 		rt = NULL;
1541 		error = in6_rtrequest(RTM_ADD,
1542 		    (struct sockaddr *)&pr->ndpr_prefix, ifa->ifa_addr,
1543 		    (struct sockaddr *)&mask6, rtflags, &rt, fibnum);
1544 		if (error == 0) {
1545 			KASSERT(rt != NULL, ("%s: in6_rtrequest return no "
1546 			    "error(%d) but rt is NULL, pr=%p, ifa=%p", __func__,
1547 			    error, pr, ifa));
1548 
1549 			rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1550 			/* XXX what if rhn == NULL? */
1551 			RADIX_NODE_HEAD_LOCK(rnh);
1552 			RT_LOCK(rt);
1553 			if (rt_setgate(rt, rt_key(rt),
1554 			    (struct sockaddr *)&null_sdl) == 0) {
1555 				struct sockaddr_dl *dl;
1556 
1557 				dl = (struct sockaddr_dl *)rt->rt_gateway;
1558 				dl->sdl_type = rt->rt_ifp->if_type;
1559 				dl->sdl_index = rt->rt_ifp->if_index;
1560 			}
1561 			RADIX_NODE_HEAD_UNLOCK(rnh);
1562 			nd6_rtmsg(RTM_ADD, rt);
1563 			RT_UNLOCK(rt);
1564 			pr->ndpr_stateflags |= NDPRF_ONLINK;
1565 		} else {
1566 			char ip6buf[INET6_ADDRSTRLEN];
1567 			char ip6bufg[INET6_ADDRSTRLEN];
1568 			char ip6bufm[INET6_ADDRSTRLEN];
1569 			struct sockaddr_in6 *sin6;
1570 
1571 			sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1572 			nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add "
1573 			    "route for a prefix (%s/%d) on %s, gw=%s, mask=%s, "
1574 			    "flags=%lx errno = %d\n",
1575 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1576 			    pr->ndpr_plen, if_name(pr->ndpr_ifp),
1577 			    ip6_sprintf(ip6bufg, &sin6->sin6_addr),
1578 			    ip6_sprintf(ip6bufm, &mask6.sin6_addr),
1579 			    rtflags, error));
1580 
1581 			/* Save last error to return, see rtinit(). */
1582 			a_failure = error;
1583 		}
1584 
1585 		if (rt != NULL) {
1586 			RT_LOCK(rt);
1587 			RT_REMREF(rt);
1588 			RT_UNLOCK(rt);
1589 		}
1590 	}
1591 
1592 	/* Return the last error we got. */
1593 	return (a_failure);
1594 }
1595 
1596 static int
1597 nd6_prefix_onlink(struct nd_prefix *pr)
1598 {
1599 	struct ifaddr *ifa;
1600 	struct ifnet *ifp = pr->ndpr_ifp;
1601 	struct nd_prefix *opr;
1602 	int error = 0;
1603 	char ip6buf[INET6_ADDRSTRLEN];
1604 
1605 	/* sanity check */
1606 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1607 		nd6log((LOG_ERR,
1608 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1609 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1610 		    pr->ndpr_plen));
1611 		return (EEXIST);
1612 	}
1613 
1614 	/*
1615 	 * Add the interface route associated with the prefix.  Before
1616 	 * installing the route, check if there's the same prefix on another
1617 	 * interface, and the prefix has already installed the interface route.
1618 	 * Although such a configuration is expected to be rare, we explicitly
1619 	 * allow it.
1620 	 */
1621 	LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1622 		if (opr == pr)
1623 			continue;
1624 
1625 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1626 			continue;
1627 
1628 		if (opr->ndpr_plen == pr->ndpr_plen &&
1629 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1630 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1631 			return (0);
1632 	}
1633 
1634 	/*
1635 	 * We prefer link-local addresses as the associated interface address.
1636 	 */
1637 	/* search for a link-local addr */
1638 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1639 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1640 	if (ifa == NULL) {
1641 		/* XXX: freebsd does not have ifa_ifwithaf */
1642 		IF_ADDR_RLOCK(ifp);
1643 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1644 			if (ifa->ifa_addr->sa_family == AF_INET6)
1645 				break;
1646 		}
1647 		if (ifa != NULL)
1648 			ifa_ref(ifa);
1649 		IF_ADDR_RUNLOCK(ifp);
1650 		/* should we care about ia6_flags? */
1651 	}
1652 	if (ifa == NULL) {
1653 		/*
1654 		 * This can still happen, when, for example, we receive an RA
1655 		 * containing a prefix with the L bit set and the A bit clear,
1656 		 * after removing all IPv6 addresses on the receiving
1657 		 * interface.  This should, of course, be rare though.
1658 		 */
1659 		nd6log((LOG_NOTICE,
1660 		    "nd6_prefix_onlink: failed to find any ifaddr"
1661 		    " to add route for a prefix(%s/%d) on %s\n",
1662 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1663 		    pr->ndpr_plen, if_name(ifp)));
1664 		return (0);
1665 	}
1666 
1667 	error = nd6_prefix_onlink_rtrequest(pr, ifa);
1668 
1669 	if (ifa != NULL)
1670 		ifa_free(ifa);
1671 
1672 	return (error);
1673 }
1674 
1675 static int
1676 nd6_prefix_offlink(struct nd_prefix *pr)
1677 {
1678 	int error = 0;
1679 	struct ifnet *ifp = pr->ndpr_ifp;
1680 	struct nd_prefix *opr;
1681 	struct sockaddr_in6 sa6, mask6;
1682 	struct rtentry *rt;
1683 	char ip6buf[INET6_ADDRSTRLEN];
1684 	int fibnum, a_failure;
1685 
1686 	/* sanity check */
1687 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1688 		nd6log((LOG_ERR,
1689 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1690 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1691 		    pr->ndpr_plen));
1692 		return (EEXIST);
1693 	}
1694 
1695 	bzero(&sa6, sizeof(sa6));
1696 	sa6.sin6_family = AF_INET6;
1697 	sa6.sin6_len = sizeof(sa6);
1698 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1699 	    sizeof(struct in6_addr));
1700 	bzero(&mask6, sizeof(mask6));
1701 	mask6.sin6_family = AF_INET6;
1702 	mask6.sin6_len = sizeof(sa6);
1703 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1704 
1705 	a_failure = 0;
1706 	for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1707 		rt = NULL;
1708 		error = in6_rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1709 		    (struct sockaddr *)&mask6, 0, &rt, fibnum);
1710 		if (error == 0) {
1711 			/* report the route deletion to the routing socket. */
1712 			if (rt != NULL)
1713 				nd6_rtmsg(RTM_DELETE, rt);
1714 		} else {
1715 			/* Save last error to return, see rtinit(). */
1716 			a_failure = error;
1717 		}
1718 		if (rt != NULL) {
1719 			RTFREE(rt);
1720 		}
1721 	}
1722 	error = a_failure;
1723 	if (error == 0) {
1724 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1725 
1726 		/*
1727 		 * There might be the same prefix on another interface,
1728 		 * the prefix which could not be on-link just because we have
1729 		 * the interface route (see comments in nd6_prefix_onlink).
1730 		 * If there's one, try to make the prefix on-link on the
1731 		 * interface.
1732 		 */
1733 		LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1734 			if (opr == pr)
1735 				continue;
1736 
1737 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1738 				continue;
1739 
1740 			/*
1741 			 * KAME specific: detached prefixes should not be
1742 			 * on-link.
1743 			 */
1744 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1745 				continue;
1746 
1747 			if (opr->ndpr_plen == pr->ndpr_plen &&
1748 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1749 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1750 				int e;
1751 
1752 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1753 					nd6log((LOG_ERR,
1754 					    "nd6_prefix_offlink: failed to "
1755 					    "recover a prefix %s/%d from %s "
1756 					    "to %s (errno = %d)\n",
1757 					    ip6_sprintf(ip6buf,
1758 						&opr->ndpr_prefix.sin6_addr),
1759 					    opr->ndpr_plen, if_name(ifp),
1760 					    if_name(opr->ndpr_ifp), e));
1761 				}
1762 			}
1763 		}
1764 	} else {
1765 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1766 		nd6log((LOG_ERR,
1767 		    "nd6_prefix_offlink: failed to delete route: "
1768 		    "%s/%d on %s (errno = %d)\n",
1769 		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1770 		    if_name(ifp), error));
1771 	}
1772 
1773 	return (error);
1774 }
1775 
1776 static struct in6_ifaddr *
1777 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1778 {
1779 	struct ifnet *ifp = pr->ndpr_ifp;
1780 	struct ifaddr *ifa;
1781 	struct in6_aliasreq ifra;
1782 	struct in6_ifaddr *ia, *ib;
1783 	int error, plen0;
1784 	struct in6_addr mask;
1785 	int prefixlen = pr->ndpr_plen;
1786 	int updateflags;
1787 	char ip6buf[INET6_ADDRSTRLEN];
1788 
1789 	in6_prefixlen2mask(&mask, prefixlen);
1790 
1791 	/*
1792 	 * find a link-local address (will be interface ID).
1793 	 * Is it really mandatory? Theoretically, a global or a site-local
1794 	 * address can be configured without a link-local address, if we
1795 	 * have a unique interface identifier...
1796 	 *
1797 	 * it is not mandatory to have a link-local address, we can generate
1798 	 * interface identifier on the fly.  we do this because:
1799 	 * (1) it should be the easiest way to find interface identifier.
1800 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1801 	 * for multiple addresses on a single interface, and possible shortcut
1802 	 * of DAD.  we omitted DAD for this reason in the past.
1803 	 * (3) a user can prevent autoconfiguration of global address
1804 	 * by removing link-local address by hand (this is partly because we
1805 	 * don't have other way to control the use of IPv6 on an interface.
1806 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1807 	 * (4) it is easier to manage when an interface has addresses
1808 	 * with the same interface identifier, than to have multiple addresses
1809 	 * with different interface identifiers.
1810 	 */
1811 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1812 	if (ifa)
1813 		ib = (struct in6_ifaddr *)ifa;
1814 	else
1815 		return NULL;
1816 
1817 	/* prefixlen + ifidlen must be equal to 128 */
1818 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1819 	if (prefixlen != plen0) {
1820 		ifa_free(ifa);
1821 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1822 		    "(prefix=%d ifid=%d)\n",
1823 		    if_name(ifp), prefixlen, 128 - plen0));
1824 		return NULL;
1825 	}
1826 
1827 	/* make ifaddr */
1828 
1829 	bzero(&ifra, sizeof(ifra));
1830 	/*
1831 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1832 	 * for safety.
1833 	 */
1834 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1835 	ifra.ifra_addr.sin6_family = AF_INET6;
1836 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1837 	/* prefix */
1838 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1839 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1840 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1841 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1842 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1843 
1844 	/* interface ID */
1845 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1846 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1847 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1848 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1849 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1850 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1851 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1852 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1853 	ifa_free(ifa);
1854 
1855 	/* new prefix mask. */
1856 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1857 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1858 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1859 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1860 
1861 	/* lifetimes. */
1862 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1863 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1864 
1865 	/* XXX: scope zone ID? */
1866 
1867 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1868 
1869 	/*
1870 	 * Make sure that we do not have this address already.  This should
1871 	 * usually not happen, but we can still see this case, e.g., if we
1872 	 * have manually configured the exact address to be configured.
1873 	 */
1874 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1875 	    &ifra.ifra_addr.sin6_addr);
1876 	if (ifa != NULL) {
1877 		ifa_free(ifa);
1878 		/* this should be rare enough to make an explicit log */
1879 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1880 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1881 		return (NULL);
1882 	}
1883 
1884 	/*
1885 	 * Allocate ifaddr structure, link into chain, etc.
1886 	 * If we are going to create a new address upon receiving a multicasted
1887 	 * RA, we need to impose a random delay before starting DAD.
1888 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1889 	 */
1890 	updateflags = 0;
1891 	if (mcast)
1892 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1893 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1894 		nd6log((LOG_ERR,
1895 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1896 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1897 		    if_name(ifp), error));
1898 		return (NULL);	/* ifaddr must not have been allocated. */
1899 	}
1900 
1901 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1902 	/*
1903 	 * XXXRW: Assumption of non-NULLness here might not be true with
1904 	 * fine-grained locking -- should we validate it?  Or just return
1905 	 * earlier ifa rather than looking it up again?
1906 	 */
1907 	return (ia);		/* this is always non-NULL  and referenced. */
1908 }
1909 
1910 /*
1911  * ia0 - corresponding public address
1912  */
1913 int
1914 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1915 {
1916 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1917 	struct in6_ifaddr *newia, *ia;
1918 	struct in6_aliasreq ifra;
1919 	int i, error;
1920 	int trylimit = 3;	/* XXX: adhoc value */
1921 	int updateflags;
1922 	u_int32_t randid[2];
1923 	time_t vltime0, pltime0;
1924 
1925 	bzero(&ifra, sizeof(ifra));
1926 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1927 	ifra.ifra_addr = ia0->ia_addr;
1928 	/* copy prefix mask */
1929 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1930 	/* clear the old IFID */
1931 	for (i = 0; i < 4; i++) {
1932 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1933 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1934 	}
1935 
1936   again:
1937 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1938 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1939 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1940 		    "random IFID\n"));
1941 		return (EINVAL);
1942 	}
1943 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1944 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1945 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1946 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1947 
1948 	/*
1949 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1950 	 * However, we may still have a chance to see collision, because
1951 	 * there may be a time lag between generation of the ID and generation
1952 	 * of the address.  So, we'll do one more sanity check.
1953 	 */
1954 	IN6_IFADDR_RLOCK();
1955 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1956 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1957 		    &ifra.ifra_addr.sin6_addr)) {
1958 			if (trylimit-- == 0) {
1959 				IN6_IFADDR_RUNLOCK();
1960 				/*
1961 				 * Give up.  Something strange should have
1962 				 * happened.
1963 				 */
1964 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1965 				    "find a unique random IFID\n"));
1966 				return (EEXIST);
1967 			}
1968 			IN6_IFADDR_RUNLOCK();
1969 			forcegen = 1;
1970 			goto again;
1971 		}
1972 	}
1973 	IN6_IFADDR_RUNLOCK();
1974 
1975 	/*
1976 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1977          * public address or TEMP_VALID_LIFETIME.
1978 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1979          * of the public address or TEMP_PREFERRED_LIFETIME -
1980          * DESYNC_FACTOR.
1981 	 */
1982 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1983 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1984 		    (ia0->ia6_lifetime.ia6t_vltime -
1985 		    (time_second - ia0->ia6_updatetime));
1986 		if (vltime0 > V_ip6_temp_valid_lifetime)
1987 			vltime0 = V_ip6_temp_valid_lifetime;
1988 	} else
1989 		vltime0 = V_ip6_temp_valid_lifetime;
1990 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1991 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1992 		    (ia0->ia6_lifetime.ia6t_pltime -
1993 		    (time_second - ia0->ia6_updatetime));
1994 		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
1995 			pltime0 = V_ip6_temp_preferred_lifetime -
1996 			    V_ip6_desync_factor;
1997 		}
1998 	} else
1999 		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
2000 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
2001 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
2002 
2003 	/*
2004 	 * A temporary address is created only if this calculated Preferred
2005 	 * Lifetime is greater than REGEN_ADVANCE time units.
2006 	 */
2007 	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2008 		return (0);
2009 
2010 	/* XXX: scope zone ID? */
2011 
2012 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2013 
2014 	/* allocate ifaddr structure, link into chain, etc. */
2015 	updateflags = 0;
2016 	if (delay)
2017 		updateflags |= IN6_IFAUPDATE_DADDELAY;
2018 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2019 		return (error);
2020 
2021 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2022 	if (newia == NULL) {	/* XXX: can it happen? */
2023 		nd6log((LOG_ERR,
2024 		    "in6_tmpifadd: ifa update succeeded, but we got "
2025 		    "no ifaddr\n"));
2026 		return (EINVAL); /* XXX */
2027 	}
2028 	newia->ia6_ndpr = ia0->ia6_ndpr;
2029 	newia->ia6_ndpr->ndpr_refcnt++;
2030 	ifa_free(&newia->ia_ifa);
2031 
2032 	/*
2033 	 * A newly added address might affect the status of other addresses.
2034 	 * XXX: when the temporary address is generated with a new public
2035 	 * address, the onlink check is redundant.  However, it would be safe
2036 	 * to do the check explicitly everywhere a new address is generated,
2037 	 * and, in fact, we surely need the check when we create a new
2038 	 * temporary address due to deprecation of an old temporary address.
2039 	 */
2040 	pfxlist_onlink_check();
2041 
2042 	return (0);
2043 }
2044 
2045 static int
2046 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2047 {
2048 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2049 		ndpr->ndpr_preferred = 0;
2050 	else
2051 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2052 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2053 		ndpr->ndpr_expire = 0;
2054 	else
2055 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2056 
2057 	return 0;
2058 }
2059 
2060 static void
2061 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2062 {
2063 	/* init ia6t_expire */
2064 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2065 		lt6->ia6t_expire = 0;
2066 	else {
2067 		lt6->ia6t_expire = time_second;
2068 		lt6->ia6t_expire += lt6->ia6t_vltime;
2069 	}
2070 
2071 	/* init ia6t_preferred */
2072 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2073 		lt6->ia6t_preferred = 0;
2074 	else {
2075 		lt6->ia6t_preferred = time_second;
2076 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2077 	}
2078 }
2079 
2080 /*
2081  * Delete all the routing table entries that use the specified gateway.
2082  * XXX: this function causes search through all entries of routing table, so
2083  * it shouldn't be called when acting as a router.
2084  */
2085 void
2086 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2087 {
2088 	struct radix_node_head *rnh;
2089 	u_int fibnum;
2090 
2091 	/* We'll care only link-local addresses */
2092 	if (!IN6_IS_ADDR_LINKLOCAL(gateway))
2093 		return;
2094 
2095 	/* XXX Do we really need to walk any but the default FIB? */
2096 	for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
2097 		rnh = rt_tables_get_rnh(fibnum, AF_INET6);
2098 		if (rnh == NULL)
2099 			continue;
2100 
2101 		RADIX_NODE_HEAD_LOCK(rnh);
2102 		rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2103 		RADIX_NODE_HEAD_UNLOCK(rnh);
2104 	}
2105 }
2106 
2107 static int
2108 rt6_deleteroute(struct radix_node *rn, void *arg)
2109 {
2110 #define SIN6(s)	((struct sockaddr_in6 *)s)
2111 	struct rtentry *rt = (struct rtentry *)rn;
2112 	struct in6_addr *gate = (struct in6_addr *)arg;
2113 
2114 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2115 		return (0);
2116 
2117 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2118 		return (0);
2119 	}
2120 
2121 	/*
2122 	 * Do not delete a static route.
2123 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2124 	 * 'cloned' bit instead?
2125 	 */
2126 	if ((rt->rt_flags & RTF_STATIC) != 0)
2127 		return (0);
2128 
2129 	/*
2130 	 * We delete only host route. This means, in particular, we don't
2131 	 * delete default route.
2132 	 */
2133 	if ((rt->rt_flags & RTF_HOST) == 0)
2134 		return (0);
2135 
2136 	return (in6_rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2137 	    rt_mask(rt), rt->rt_flags, NULL, rt->rt_fibnum));
2138 #undef SIN6
2139 }
2140 
2141 int
2142 nd6_setdefaultiface(int ifindex)
2143 {
2144 	int error = 0;
2145 
2146 	if (ifindex < 0 || V_if_index < ifindex)
2147 		return (EINVAL);
2148 	if (ifindex != 0 && !ifnet_byindex(ifindex))
2149 		return (EINVAL);
2150 
2151 	if (V_nd6_defifindex != ifindex) {
2152 		V_nd6_defifindex = ifindex;
2153 		if (V_nd6_defifindex > 0)
2154 			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2155 		else
2156 			V_nd6_defifp = NULL;
2157 
2158 		/*
2159 		 * Our current implementation assumes one-to-one maping between
2160 		 * interfaces and links, so it would be natural to use the
2161 		 * default interface as the default link.
2162 		 */
2163 		scope6_setdefault(V_nd6_defifp);
2164 	}
2165 
2166 	return (error);
2167 }
2168