xref: /netbsd-src/sys/netinet6/nd6_rtr.c (revision 1b9578b8c2c1f848eeb16dabbfd7d1f0d9fdefbd)
1 /*	$NetBSD: nd6_rtr.c,v 1.81 2011/05/24 18:07:11 spz Exp $	*/
2 /*	$KAME: nd6_rtr.c,v 1.95 2001/02/07 08:09:47 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: nd6_rtr.c,v 1.81 2011/05/24 18:07:11 spz Exp $");
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/malloc.h>
39 #include <sys/mbuf.h>
40 #include <sys/socket.h>
41 #include <sys/sockio.h>
42 #include <sys/time.h>
43 #include <sys/kernel.h>
44 #include <sys/errno.h>
45 #include <sys/ioctl.h>
46 #include <sys/syslog.h>
47 
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/if_dl.h>
51 #include <net/route.h>
52 #include <net/radix.h>
53 
54 #include <netinet/in.h>
55 #include <netinet6/in6_var.h>
56 #include <netinet6/in6_ifattach.h>
57 #include <netinet/ip6.h>
58 #include <netinet6/ip6_var.h>
59 #include <netinet6/nd6.h>
60 #include <netinet/icmp6.h>
61 #include <netinet6/icmp6_private.h>
62 #include <netinet6/scope6_var.h>
63 
64 #include <net/net_osdep.h>
65 
66 static int rtpref(struct nd_defrouter *);
67 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
68 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
69     struct mbuf *, int);
70 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
71 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
72 	struct nd_defrouter *);
73 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
74 static void pfxrtr_del(struct nd_pfxrouter *);
75 static struct nd_pfxrouter *find_pfxlist_reachable_router
76 	(struct nd_prefix *);
77 static void defrouter_delreq(struct nd_defrouter *);
78 static void nd6_rtmsg(int, struct rtentry *);
79 
80 static int in6_init_prefix_ltimes(struct nd_prefix *);
81 static void in6_init_address_ltimes(struct nd_prefix *ndpr,
82 	struct in6_addrlifetime *lt6);
83 
84 static int rt6_deleteroute(struct rtentry *, void *);
85 
86 extern int nd6_recalc_reachtm_interval;
87 
88 static struct ifnet *nd6_defifp;
89 int nd6_defifindex;
90 
91 int ip6_use_tempaddr = 0;
92 
93 int ip6_desync_factor;
94 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
95 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
96 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
97 
98 int nd6_numroutes = 0;
99 
100 /* RTPREF_MEDIUM has to be 0! */
101 #define RTPREF_HIGH	1
102 #define RTPREF_MEDIUM	0
103 #define RTPREF_LOW	(-1)
104 #define RTPREF_RESERVED	(-2)
105 #define RTPREF_INVALID	(-3)	/* internal */
106 
107 /*
108  * Receive Router Solicitation Message - just for routers.
109  * Router solicitation/advertisement is mostly managed by a userland program
110  * (rtadvd) so here we have no function like nd6_ra_output().
111  *
112  * Based on RFC 2461
113  */
114 void
115 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
116 {
117 	struct ifnet *ifp = m->m_pkthdr.rcvif;
118 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
119 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
120 	struct nd_router_solicit *nd_rs;
121 	struct in6_addr saddr6 = ip6->ip6_src;
122 	char *lladdr = NULL;
123 	int lladdrlen = 0;
124 	union nd_opts ndopts;
125 
126 	/* If I'm not a router, ignore it. */
127 	if (nd6_accepts_rtadv(ndi) || !ip6_forwarding)
128 		goto freeit;
129 
130 	/* Sanity checks */
131 	if (ip6->ip6_hlim != 255) {
132 		nd6log((LOG_ERR,
133 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
134 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
135 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
136 		goto bad;
137 	}
138 
139 	/*
140 	 * Don't update the neighbor cache, if src = ::.
141 	 * This indicates that the src has no IP address assigned yet.
142 	 */
143 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
144 		goto freeit;
145 
146 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
147 	if (nd_rs == NULL) {
148 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
149 		return;
150 	}
151 
152 	icmp6len -= sizeof(*nd_rs);
153 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
154 	if (nd6_options(&ndopts) < 0) {
155 		nd6log((LOG_INFO,
156 		    "nd6_rs_input: invalid ND option, ignored\n"));
157 		/* nd6_options have incremented stats */
158 		goto freeit;
159 	}
160 
161 	if (ndopts.nd_opts_src_lladdr) {
162 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
163 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
164 	}
165 
166 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
167 		nd6log((LOG_INFO,
168 		    "nd6_rs_input: lladdrlen mismatch for %s "
169 		    "(if %d, RS packet %d)\n",
170 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
171 		goto bad;
172 	}
173 
174 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
175 
176  freeit:
177 	m_freem(m);
178 	return;
179 
180  bad:
181 	ICMP6_STATINC(ICMP6_STAT_BADRS);
182 	m_freem(m);
183 }
184 
185 /*
186  * Receive Router Advertisement Message.
187  *
188  * Based on RFC 2461
189  * TODO: on-link bit on prefix information
190  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
191  */
192 void
193 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
194 {
195 	struct ifnet *ifp = m->m_pkthdr.rcvif;
196 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
197 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
198 	struct nd_router_advert *nd_ra;
199 	struct in6_addr saddr6 = ip6->ip6_src;
200 #if 0
201 	struct in6_addr daddr6 = ip6->ip6_dst;
202 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
203 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
204 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
205 #endif
206 	int mcast = 0;
207 	union nd_opts ndopts;
208 	struct nd_defrouter *dr;
209 
210 	/*
211 	 * We only accept RAs when
212 	 * the system-wide variable allows the acceptance, and the
213 	 * per-interface variable allows RAs on the receiving interface.
214 	 */
215 	if (!nd6_accepts_rtadv(ndi))
216 		goto freeit;
217 
218 	if (ip6->ip6_hlim != 255) {
219 		nd6log((LOG_ERR,
220 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
221 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
222 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
223 		goto bad;
224 	}
225 
226 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
227 		nd6log((LOG_ERR,
228 		    "nd6_ra_input: src %s is not link-local\n",
229 		    ip6_sprintf(&saddr6)));
230 		goto bad;
231 	}
232 
233 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
234 	if (nd_ra == NULL) {
235 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
236 		return;
237 	}
238 
239 	icmp6len -= sizeof(*nd_ra);
240 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
241 	if (nd6_options(&ndopts) < 0) {
242 		nd6log((LOG_INFO,
243 		    "nd6_ra_input: invalid ND option, ignored\n"));
244 		/* nd6_options have incremented stats */
245 		goto freeit;
246 	}
247 
248     {
249 	struct nd_defrouter drtr;
250 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
251 
252 	/* remember if this is a multicasted advertisement */
253 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
254 		mcast = 1;
255 
256 	memset(&drtr, 0, sizeof(drtr));
257 	drtr.rtaddr = saddr6;
258 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
259 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
260 	drtr.expire = time_second + drtr.rtlifetime;
261 	drtr.ifp = ifp;
262 	/* unspecified or not? (RFC 2461 6.3.4) */
263 	if (advreachable) {
264 		NTOHL(advreachable);
265 		if (advreachable <= MAX_REACHABLE_TIME &&
266 		    ndi->basereachable != advreachable) {
267 			ndi->basereachable = advreachable;
268 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
269 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
270 		}
271 	}
272 	if (nd_ra->nd_ra_retransmit)
273 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
274 	if (nd_ra->nd_ra_curhoplimit)
275 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
276 	dr = defrtrlist_update(&drtr);
277     }
278 
279 	/*
280 	 * prefix
281 	 */
282 	if (ndopts.nd_opts_pi) {
283 		struct nd_opt_hdr *pt;
284 		struct nd_opt_prefix_info *pi = NULL;
285 		struct nd_prefixctl pr;
286 
287 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
288 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
289 		     pt = (struct nd_opt_hdr *)((char *)pt +
290 						(pt->nd_opt_len << 3))) {
291 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
292 				continue;
293 			pi = (struct nd_opt_prefix_info *)pt;
294 
295 			if (pi->nd_opt_pi_len != 4) {
296 				nd6log((LOG_INFO,
297 				    "nd6_ra_input: invalid option "
298 				    "len %d for prefix information option, "
299 				    "ignored\n", pi->nd_opt_pi_len));
300 				continue;
301 			}
302 
303 			if (128 < pi->nd_opt_pi_prefix_len) {
304 				nd6log((LOG_INFO,
305 				    "nd6_ra_input: invalid prefix "
306 				    "len %d for prefix information option, "
307 				    "ignored\n", pi->nd_opt_pi_prefix_len));
308 				continue;
309 			}
310 
311 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
312 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
313 				nd6log((LOG_INFO,
314 				    "nd6_ra_input: invalid prefix "
315 				    "%s, ignored\n",
316 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
317 				continue;
318 			}
319 
320 			memset(&pr, 0, sizeof(pr));
321 			sockaddr_in6_init(&pr.ndpr_prefix,
322 			    &pi->nd_opt_pi_prefix, 0, 0, 0);
323 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
324 
325 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
326 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
327 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
328 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
329 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
330 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
331 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
332 
333 			(void)prelist_update(&pr, dr, m, mcast);
334 		}
335 	}
336 
337 	/*
338 	 * MTU
339 	 */
340 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
341 		u_long mtu;
342 		u_long maxmtu;
343 
344 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
345 
346 		/* lower bound */
347 		if (mtu < IPV6_MMTU) {
348 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
349 			    "mtu=%lu sent from %s, ignoring\n",
350 			    mtu, ip6_sprintf(&ip6->ip6_src)));
351 			goto skip;
352 		}
353 
354 		/* upper bound */
355 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
356 		    ? ndi->maxmtu : ifp->if_mtu;
357 		if (mtu <= maxmtu) {
358 			int change = (ndi->linkmtu != mtu);
359 
360 			ndi->linkmtu = mtu;
361 			if (change) /* in6_maxmtu may change */
362 				in6_setmaxmtu();
363 		} else {
364 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
365 			    "mtu=%lu sent from %s; "
366 			    "exceeds maxmtu %lu, ignoring\n",
367 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
368 		}
369 	}
370 
371  skip:
372 
373 	/*
374 	 * Source link layer address
375 	 */
376     {
377 	char *lladdr = NULL;
378 	int lladdrlen = 0;
379 
380 	if (ndopts.nd_opts_src_lladdr) {
381 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
382 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
383 	}
384 
385 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
386 		nd6log((LOG_INFO,
387 		    "nd6_ra_input: lladdrlen mismatch for %s "
388 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
389 		    ifp->if_addrlen, lladdrlen - 2));
390 		goto bad;
391 	}
392 
393 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
394 
395 	/*
396 	 * Installing a link-layer address might change the state of the
397 	 * router's neighbor cache, which might also affect our on-link
398 	 * detection of adveritsed prefixes.
399 	 */
400 	pfxlist_onlink_check();
401     }
402 
403  freeit:
404 	m_freem(m);
405 	return;
406 
407  bad:
408 	ICMP6_STATINC(ICMP6_STAT_BADRA);
409 	m_freem(m);
410 }
411 
412 /*
413  * default router list processing sub routines
414  */
415 
416 /* tell the change to user processes watching the routing socket. */
417 static void
418 nd6_rtmsg(int cmd, struct rtentry *rt)
419 {
420 	struct rt_addrinfo info;
421 
422 	memset((void *)&info, 0, sizeof(info));
423 	info.rti_info[RTAX_DST] = rt_getkey(rt);
424 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
425 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
426 	if (rt->rt_ifp) {
427 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
428 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
429 	}
430 
431 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
432 }
433 
434 void
435 defrouter_addreq(struct nd_defrouter *new)
436 {
437 	union {
438 		struct sockaddr_in6 sin6;
439 		struct sockaddr sa;
440 	} def, mask, gate;
441 	struct rtentry *newrt = NULL;
442 	int s;
443 	int error;
444 
445 	memset(&def, 0, sizeof(def));
446 	memset(&mask, 0, sizeof(mask));
447 	memset(&gate, 0,sizeof(gate)); /* for safety */
448 
449 	def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len =
450 	    sizeof(struct sockaddr_in6);
451 	def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6;
452 	gate.sin6.sin6_addr = new->rtaddr;
453 #ifndef SCOPEDROUTING
454 	gate.sin6.sin6_scope_id = 0;	/* XXX */
455 #endif
456 
457 	s = splsoftnet();
458 	error = rtrequest(RTM_ADD, &def.sa, &gate.sa, &mask.sa,
459 	    RTF_GATEWAY, &newrt);
460 	if (newrt) {
461 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
462 		newrt->rt_refcnt--;
463 		nd6_numroutes++;
464 	}
465 	if (error == 0)
466 		new->installed = 1;
467 	splx(s);
468 	return;
469 }
470 
471 struct nd_defrouter *
472 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
473 {
474 	struct nd_defrouter *dr;
475 
476 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
477 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
478 			break;
479 	}
480 
481 	return dr;		/* search failed */
482 }
483 
484 void
485 defrtrlist_del(struct nd_defrouter *dr)
486 {
487 	struct nd_ifinfo *ndi = ND_IFINFO(dr->ifp);
488 	struct nd_defrouter *deldr = NULL;
489 	struct nd_prefix *pr;
490 
491 	/*
492 	 * Flush all the routing table entries that use the router
493 	 * as a next hop.
494 	 */
495 	/* XXX: better condition? */
496 	if (!ip6_forwarding && nd6_accepts_rtadv(ndi))
497 		rt6_flush(&dr->rtaddr, dr->ifp);
498 
499 	if (dr->installed) {
500 		deldr = dr;
501 		defrouter_delreq(dr);
502 	}
503 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
504 
505 	/*
506 	 * Also delete all the pointers to the router in each prefix lists.
507 	 */
508 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
509 		struct nd_pfxrouter *pfxrtr;
510 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
511 			pfxrtr_del(pfxrtr);
512 	}
513 	pfxlist_onlink_check();
514 
515 	/*
516 	 * If the router is the primary one, choose a new one.
517 	 * Note that defrouter_select() will remove the current gateway
518 	 * from the routing table.
519 	 */
520 	if (deldr)
521 		defrouter_select();
522 
523 	free(dr, M_IP6NDP);
524 }
525 
526 /*
527  * Remove the default route for a given router.
528  * This is just a subroutine function for defrouter_select(), and should
529  * not be called from anywhere else.
530  */
531 static void
532 defrouter_delreq(struct nd_defrouter *dr)
533 {
534 	union {
535 		struct sockaddr_in6 sin6;
536 		struct sockaddr sa;
537 	} def, mask, gw;
538 	struct rtentry *oldrt = NULL;
539 
540 #ifdef DIAGNOSTIC
541 	if (dr == NULL)
542 		panic("dr == NULL in defrouter_delreq");
543 #endif
544 
545 	memset(&def, 0, sizeof(def));
546 	memset(&mask, 0, sizeof(mask));
547 	memset(&gw, 0, sizeof(gw));	/* for safety */
548 
549 	def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len =
550 	    sizeof(struct sockaddr_in6);
551 	def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6;
552 	gw.sin6.sin6_addr = dr->rtaddr;
553 #ifndef SCOPEDROUTING
554 	gw.sin6.sin6_scope_id = 0;	/* XXX */
555 #endif
556 
557 	rtrequest(RTM_DELETE, &def.sa, &gw.sa, &mask.sa, RTF_GATEWAY, &oldrt);
558 	if (oldrt) {
559 		nd6_rtmsg(RTM_DELETE, oldrt);
560 		if (oldrt->rt_refcnt <= 0) {
561 			/*
562 			 * XXX: borrowed from the RTM_DELETE case of
563 			 * rtrequest().
564 			 */
565 			oldrt->rt_refcnt++;
566 			rtfree(oldrt);
567 			nd6_numroutes--;
568 		}
569 	}
570 
571 	dr->installed = 0;
572 }
573 
574 /*
575  * remove all default routes from default router list
576  */
577 void
578 defrouter_reset(void)
579 {
580 	struct nd_defrouter *dr;
581 
582 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
583 	     dr = TAILQ_NEXT(dr, dr_entry))
584 		defrouter_delreq(dr);
585 
586 	/*
587 	 * XXX should we also nuke any default routers in the kernel, by
588 	 * going through them by rtalloc1()?
589 	 */
590 }
591 
592 /*
593  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
594  * draft-ietf-ipngwg-router-selection:
595  * 1) Routers that are reachable or probably reachable should be preferred.
596  *    If we have more than one (probably) reachable router, prefer ones
597  *    with the highest router preference.
598  * 2) When no routers on the list are known to be reachable or
599  *    probably reachable, routers SHOULD be selected in a round-robin
600  *    fashion, regardless of router preference values.
601  * 3) If the Default Router List is empty, assume that all
602  *    destinations are on-link.
603  *
604  * We assume nd_defrouter is sorted by router preference value.
605  * Since the code below covers both with and without router preference cases,
606  * we do not need to classify the cases by ifdef.
607  *
608  * At this moment, we do not try to install more than one default router,
609  * even when the multipath routing is available, because we're not sure about
610  * the benefits for stub hosts comparing to the risk of making the code
611  * complicated and the possibility of introducing bugs.
612  */
613 void
614 defrouter_select(void)
615 {
616 	struct nd_ifinfo *ndi;
617 	int s = splsoftnet();
618 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
619 	struct rtentry *rt = NULL;
620 	struct llinfo_nd6 *ln = NULL;
621 
622 	/*
623 	 * This function should be called only when acting as an autoconfigured
624 	 * host.  Although the remaining part of this function is not effective
625 	 * if the node is not an autoconfigured host, we explicitly exclude
626 	 * such cases here for safety.
627 	 */
628 	if (ip6_forwarding) {
629 		nd6log((LOG_WARNING,
630 		    "defrouter_select: called unexpectedly (forwarding=%d, "
631 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
632 		splx(s);
633 		return;
634 	}
635 
636 	/*
637 	 * Let's handle easy case (3) first:
638 	 * If default router list is empty, there's nothing to be done.
639 	 */
640 	if (!TAILQ_FIRST(&nd_defrouter)) {
641 		splx(s);
642 		return;
643 	}
644 
645 	/*
646 	 * Search for a (probably) reachable router from the list.
647 	 * We just pick up the first reachable one (if any), assuming that
648 	 * the ordering rule of the list described in defrtrlist_update().
649 	 */
650 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
651 	     dr = TAILQ_NEXT(dr, dr_entry)) {
652 		ndi = ND_IFINFO(dr->ifp);
653 		if (nd6_accepts_rtadv(ndi))
654 			continue;
655 
656 		if (selected_dr == NULL &&
657 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) != NULL &&
658 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) != NULL &&
659 		    ND6_IS_LLINFO_PROBREACH(ln)) {
660 			selected_dr = dr;
661 		}
662 
663 		if (dr->installed && !installed_dr)
664 			installed_dr = dr;
665 		else if (dr->installed && installed_dr) {
666 			/* this should not happen.  warn for diagnosis. */
667 			log(LOG_ERR, "defrouter_select: more than one router"
668 			    " is installed\n");
669 		}
670 	}
671 	/*
672 	 * If none of the default routers was found to be reachable,
673 	 * round-robin the list regardless of preference.
674 	 * Otherwise, if we have an installed router, check if the selected
675 	 * (reachable) router should really be preferred to the installed one.
676 	 * We only prefer the new router when the old one is not reachable
677 	 * or when the new one has a really higher preference value.
678 	 */
679 	if (selected_dr == NULL) {
680 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
681 			selected_dr = TAILQ_FIRST(&nd_defrouter);
682 		else
683 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
684 	} else if (installed_dr &&
685 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
686 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
687 	    ND6_IS_LLINFO_PROBREACH(ln) &&
688 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
689 		selected_dr = installed_dr;
690 	}
691 
692 	/*
693 	 * If the selected router is different than the installed one,
694 	 * remove the installed router and install the selected one.
695 	 * Note that the selected router is never NULL here.
696 	 */
697 	if (installed_dr != selected_dr) {
698 		if (installed_dr)
699 			defrouter_delreq(installed_dr);
700 		defrouter_addreq(selected_dr);
701 	}
702 
703 	splx(s);
704 	return;
705 }
706 
707 /*
708  * for default router selection
709  * regards router-preference field as a 2-bit signed integer
710  */
711 static int
712 rtpref(struct nd_defrouter *dr)
713 {
714 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
715 	case ND_RA_FLAG_RTPREF_HIGH:
716 		return (RTPREF_HIGH);
717 	case ND_RA_FLAG_RTPREF_MEDIUM:
718 	case ND_RA_FLAG_RTPREF_RSV:
719 		return (RTPREF_MEDIUM);
720 	case ND_RA_FLAG_RTPREF_LOW:
721 		return (RTPREF_LOW);
722 	default:
723 		/*
724 		 * This case should never happen.  If it did, it would mean a
725 		 * serious bug of kernel internal.  We thus always bark here.
726 		 * Or, can we even panic?
727 		 */
728 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
729 		return (RTPREF_INVALID);
730 	}
731 	/* NOTREACHED */
732 }
733 
734 static struct nd_defrouter *
735 defrtrlist_update(struct nd_defrouter *new)
736 {
737 	struct nd_defrouter *dr, *n;
738 	int s = splsoftnet();
739 
740 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
741 		/* entry exists */
742 		if (new->rtlifetime == 0) {
743 			defrtrlist_del(dr);
744 			dr = NULL;
745 		} else {
746 			int oldpref = rtpref(dr);
747 
748 			/* override */
749 			dr->flags = new->flags; /* xxx flag check */
750 			dr->rtlifetime = new->rtlifetime;
751 			dr->expire = new->expire;
752 
753 			/*
754 			 * If the preference does not change, there's no need
755 			 * to sort the entries.
756 			 */
757 			if (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(&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 	if (ip6_rtadv_maxroutes <= nd6_numroutes) {
786 		ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
787 		splx(s);
788 		return (NULL);
789 	}
790 
791 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
792 	if (n == NULL) {
793 		splx(s);
794 		return (NULL);
795 	}
796 	memset(n, 0, sizeof(*n));
797 	*n = *new;
798 
799 insert:
800 	/*
801 	 * Insert the new router in the Default Router List;
802 	 * The Default Router List should be in the descending order
803 	 * of router-preferece.  Routers with the same preference are
804 	 * sorted in the arriving time order.
805 	 */
806 
807 	/* insert at the end of the group */
808 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
809 	     dr = TAILQ_NEXT(dr, dr_entry)) {
810 		if (rtpref(n) > rtpref(dr))
811 			break;
812 	}
813 	if (dr)
814 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
815 	else
816 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
817 
818 	defrouter_select();
819 
820 	splx(s);
821 
822 	return (n);
823 }
824 
825 static struct nd_pfxrouter *
826 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
827 {
828 	struct nd_pfxrouter *search;
829 
830 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
831 		if (search->router == dr)
832 			break;
833 	}
834 
835 	return (search);
836 }
837 
838 static void
839 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
840 {
841 	struct nd_pfxrouter *new;
842 
843 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT|M_ZERO);
844 	if (new == NULL)
845 		return;
846 	new->router = dr;
847 
848 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
849 
850 	pfxlist_onlink_check();
851 }
852 
853 static void
854 pfxrtr_del(struct nd_pfxrouter *pfr)
855 {
856 	LIST_REMOVE(pfr, pfr_entry);
857 	free(pfr, M_IP6NDP);
858 }
859 
860 struct nd_prefix *
861 nd6_prefix_lookup(struct nd_prefixctl *key)
862 {
863 	struct nd_prefix *search;
864 
865 	LIST_FOREACH(search, &nd_prefix, ndpr_entry) {
866 		if (key->ndpr_ifp == search->ndpr_ifp &&
867 		    key->ndpr_plen == search->ndpr_plen &&
868 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
869 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
870 			break;
871 		}
872 	}
873 
874 	return (search);
875 }
876 
877 int
878 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
879 	struct nd_prefix **newp)
880 {
881 	struct nd_prefix *new = NULL;
882 	int i, s;
883 	int error;
884 
885 	error = 0;
886 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT|M_ZERO);
887 	if (new == NULL)
888 		return ENOMEM;
889 	new->ndpr_ifp = pr->ndpr_ifp;
890 	new->ndpr_prefix = pr->ndpr_prefix;
891 	new->ndpr_plen = pr->ndpr_plen;
892 	new->ndpr_vltime = pr->ndpr_vltime;
893 	new->ndpr_pltime = pr->ndpr_pltime;
894 	new->ndpr_flags = pr->ndpr_flags;
895 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
896 		free(new, M_IP6NDP);
897 		return(error);
898 	}
899 	new->ndpr_lastupdate = time_second;
900 	if (newp != NULL)
901 		*newp = new;
902 
903 	/* initialization */
904 	LIST_INIT(&new->ndpr_advrtrs);
905 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
906 	/* make prefix in the canonical form */
907 	for (i = 0; i < 4; i++)
908 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
909 		    new->ndpr_mask.s6_addr32[i];
910 
911 	s = splsoftnet();
912 	/* link ndpr_entry to nd_prefix list */
913 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
914 	splx(s);
915 
916 	/* ND_OPT_PI_FLAG_ONLINK processing */
917 	if (new->ndpr_raf_onlink) {
918 		int e;
919 
920 		if ((e = nd6_prefix_onlink(new)) != 0) {
921 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
922 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
923 			    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
924 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
925 			/* proceed anyway. XXX: is it correct? */
926 		}
927 	}
928 
929 	if (dr)
930 		pfxrtr_add(new, dr);
931 
932 	return 0;
933 }
934 
935 void
936 prelist_remove(struct nd_prefix *pr)
937 {
938 	struct nd_pfxrouter *pfr, *next;
939 	int e, s;
940 
941 	/* make sure to invalidate the prefix until it is really freed. */
942 	pr->ndpr_vltime = 0;
943 	pr->ndpr_pltime = 0;
944 #if 0
945 	/*
946 	 * Though these flags are now meaningless, we'd rather keep the value
947 	 * not to confuse users when executing "ndp -p".
948 	 */
949 	pr->ndpr_raf_onlink = 0;
950 	pr->ndpr_raf_auto = 0;
951 #endif
952 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
953 	    (e = nd6_prefix_offlink(pr)) != 0) {
954 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
955 		    "on %s, errno=%d\n",
956 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
957 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
958 		/* what should we do? */
959 	}
960 
961 	if (pr->ndpr_refcnt > 0)
962 		return;		/* notice here? */
963 
964 	s = splsoftnet();
965 	/* unlink ndpr_entry from nd_prefix list */
966 	LIST_REMOVE(pr, ndpr_entry);
967 
968 	/* free list of routers that adversed the prefix */
969 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
970 		next = LIST_NEXT(pfr, pfr_entry);
971 
972 		free(pfr, M_IP6NDP);
973 	}
974 	splx(s);
975 
976 	free(pr, M_IP6NDP);
977 
978 	pfxlist_onlink_check();
979 }
980 
981 static int
982 prelist_update(struct nd_prefixctl *new,
983 	struct nd_defrouter *dr, /* may be NULL */
984 	struct mbuf *m,
985 	int mcast)
986 {
987 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
988 	struct ifaddr *ifa;
989 	struct ifnet *ifp = new->ndpr_ifp;
990 	struct nd_prefix *pr;
991 	int s = splsoftnet();
992 	int error = 0;
993 	int newprefix = 0;
994 	int auth;
995 	struct in6_addrlifetime lt6_tmp;
996 
997 	auth = 0;
998 	if (m) {
999 		/*
1000 		 * Authenticity for NA consists authentication for
1001 		 * both IP header and IP datagrams, doesn't it ?
1002 		 */
1003 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1004 		auth = (m->m_flags & M_AUTHIPHDR
1005 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
1006 #endif
1007 	}
1008 
1009 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1010 		/*
1011 		 * nd6_prefix_lookup() ensures that pr and new have the same
1012 		 * prefix on a same interface.
1013 		 */
1014 
1015 		/*
1016 		 * Update prefix information.  Note that the on-link (L) bit
1017 		 * and the autonomous (A) bit should NOT be changed from 1
1018 		 * to 0.
1019 		 */
1020 		if (new->ndpr_raf_onlink == 1)
1021 			pr->ndpr_raf_onlink = 1;
1022 		if (new->ndpr_raf_auto == 1)
1023 			pr->ndpr_raf_auto = 1;
1024 		if (new->ndpr_raf_onlink) {
1025 			pr->ndpr_vltime = new->ndpr_vltime;
1026 			pr->ndpr_pltime = new->ndpr_pltime;
1027 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1028 			pr->ndpr_lastupdate = time_second;
1029 		}
1030 
1031 		if (new->ndpr_raf_onlink &&
1032 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1033 			int e;
1034 
1035 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1036 				nd6log((LOG_ERR,
1037 				    "prelist_update: failed to make "
1038 				    "the prefix %s/%d on-link on %s "
1039 				    "(errno=%d)\n",
1040 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1041 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1042 				/* proceed anyway. XXX: is it correct? */
1043 			}
1044 		}
1045 
1046 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1047 			pfxrtr_add(pr, dr);
1048 	} else {
1049 		struct nd_prefix *newpr = NULL;
1050 
1051 		newprefix = 1;
1052 
1053 		if (new->ndpr_vltime == 0)
1054 			goto end;
1055 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1056 			goto end;
1057 
1058 		if (ip6_rtadv_maxroutes <= nd6_numroutes) {
1059 			ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
1060 			goto end;
1061 		}
1062 
1063 		error = nd6_prelist_add(new, dr, &newpr);
1064 		if (error != 0 || newpr == NULL) {
1065 			nd6log((LOG_NOTICE, "prelist_update: "
1066 			    "nd6_prelist_add failed for %s/%d on %s "
1067 			    "errno=%d, returnpr=%p\n",
1068 			    ip6_sprintf(&new->ndpr_prefix.sin6_addr),
1069 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1070 			    error, newpr));
1071 			goto end; /* we should just give up in this case. */
1072 		}
1073 
1074 		/*
1075 		 * XXX: from the ND point of view, we can ignore a prefix
1076 		 * with the on-link bit being zero.  However, we need a
1077 		 * prefix structure for references from autoconfigured
1078 		 * addresses.  Thus, we explicitly make sure that the prefix
1079 		 * itself expires now.
1080 		 */
1081 		if (newpr->ndpr_raf_onlink == 0) {
1082 			newpr->ndpr_vltime = 0;
1083 			newpr->ndpr_pltime = 0;
1084 			in6_init_prefix_ltimes(newpr);
1085 		}
1086 
1087 		pr = newpr;
1088 	}
1089 
1090 	/*
1091 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1092 	 * Note that pr must be non NULL at this point.
1093 	 */
1094 
1095 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1096 	if (!new->ndpr_raf_auto)
1097 		goto end;
1098 
1099 	/*
1100 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1101 	 * nd6_ra_input.
1102 	 */
1103 
1104 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1105 	if (new->ndpr_pltime > new->ndpr_vltime) {
1106 		error = EINVAL;	/* XXX: won't be used */
1107 		goto end;
1108 	}
1109 
1110 	/*
1111 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1112 	 * an address configured by stateless autoconfiguration already in the
1113 	 * list of addresses associated with the interface, and the Valid
1114 	 * Lifetime is not 0, form an address.  We first check if we have
1115 	 * a matching prefix.
1116 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1117 	 * consider autoconfigured addresses while RFC2462 simply said
1118 	 * "address".
1119 	 */
1120 	IFADDR_FOREACH(ifa, ifp) {
1121 		struct in6_ifaddr *ifa6;
1122 		u_int32_t remaininglifetime;
1123 
1124 		if (ifa->ifa_addr->sa_family != AF_INET6)
1125 			continue;
1126 
1127 		ifa6 = (struct in6_ifaddr *)ifa;
1128 
1129 		/*
1130 		 * We only consider autoconfigured addresses as per rfc2462bis.
1131 		 */
1132 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1133 			continue;
1134 
1135 		/*
1136 		 * Spec is not clear here, but I believe we should concentrate
1137 		 * on unicast (i.e. not anycast) addresses.
1138 		 * XXX: other ia6_flags? detached or duplicated?
1139 		 */
1140 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1141 			continue;
1142 
1143 		/*
1144 		 * Ignore the address if it is not associated with a prefix
1145 		 * or is associated with a prefix that is different from this
1146 		 * one.  (pr is never NULL here)
1147 		 */
1148 		if (ifa6->ia6_ndpr != pr)
1149 			continue;
1150 
1151 		if (ia6_match == NULL) /* remember the first one */
1152 			ia6_match = ifa6;
1153 
1154 		/*
1155 		 * An already autoconfigured address matched.  Now that we
1156 		 * are sure there is at least one matched address, we can
1157 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1158 		 * "two hours" rule and the privacy extension.
1159 		 * We apply some clarifications in rfc2462bis:
1160 		 * - use remaininglifetime instead of storedlifetime as a
1161 		 *   variable name
1162 		 * - remove the dead code in the "two-hour" rule
1163 		 */
1164 #define TWOHOUR		(120*60)
1165 		lt6_tmp = ifa6->ia6_lifetime;
1166 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1167 			remaininglifetime = ND6_INFINITE_LIFETIME;
1168 		else if (time_second - ifa6->ia6_updatetime >
1169 			 lt6_tmp.ia6t_vltime) {
1170 			/*
1171 			 * The case of "invalid" address.  We should usually
1172 			 * not see this case.
1173 			 */
1174 			remaininglifetime = 0;
1175 		} else
1176 			remaininglifetime = lt6_tmp.ia6t_vltime -
1177 			    (time_second - ifa6->ia6_updatetime);
1178 
1179 		/* when not updating, keep the current stored lifetime. */
1180 		lt6_tmp.ia6t_vltime = remaininglifetime;
1181 
1182 		if (TWOHOUR < new->ndpr_vltime ||
1183 		    remaininglifetime < new->ndpr_vltime) {
1184 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1185 		} else if (remaininglifetime <= TWOHOUR) {
1186 			if (auth)
1187 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1188 		} else {
1189 			/*
1190 			 * new->ndpr_vltime <= TWOHOUR &&
1191 			 * TWOHOUR < remaininglifetime
1192 			 */
1193 			lt6_tmp.ia6t_vltime = TWOHOUR;
1194 		}
1195 
1196 		/* The 2 hour rule is not imposed for preferred lifetime. */
1197 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1198 
1199 		in6_init_address_ltimes(pr, &lt6_tmp);
1200 
1201 		/*
1202 		 * We need to treat lifetimes for temporary addresses
1203 		 * differently, according to
1204 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1205 		 * we only update the lifetimes when they are in the maximum
1206 		 * intervals.
1207 		 */
1208 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1209 			u_int32_t maxvltime, maxpltime;
1210 
1211 			if (ip6_temp_valid_lifetime >
1212 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1213 			    ip6_desync_factor)) {
1214 				maxvltime = ip6_temp_valid_lifetime -
1215 				    (time_second - ifa6->ia6_createtime) -
1216 				    ip6_desync_factor;
1217 			} else
1218 				maxvltime = 0;
1219 			if (ip6_temp_preferred_lifetime >
1220 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1221 			    ip6_desync_factor)) {
1222 				maxpltime = ip6_temp_preferred_lifetime -
1223 				    (time_second - ifa6->ia6_createtime) -
1224 				    ip6_desync_factor;
1225 			} else
1226 				maxpltime = 0;
1227 
1228 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1229 			    lt6_tmp.ia6t_vltime > maxvltime) {
1230 				lt6_tmp.ia6t_vltime = maxvltime;
1231 			}
1232 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1233 			    lt6_tmp.ia6t_pltime > maxpltime) {
1234 				lt6_tmp.ia6t_pltime = maxpltime;
1235 			}
1236 		}
1237 
1238 		ifa6->ia6_lifetime = lt6_tmp;
1239 		ifa6->ia6_updatetime = time_second;
1240 	}
1241 	if (ia6_match == NULL && new->ndpr_vltime) {
1242 		int ifidlen;
1243 
1244 		/*
1245 		 * 5.5.3 (d) (continued)
1246 		 * No address matched and the valid lifetime is non-zero.
1247 		 * Create a new address.
1248 		 */
1249 
1250 		/*
1251 		 * Prefix Length check:
1252 		 * If the sum of the prefix length and interface identifier
1253 		 * length does not equal 128 bits, the Prefix Information
1254 		 * option MUST be ignored.  The length of the interface
1255 		 * identifier is defined in a separate link-type specific
1256 		 * document.
1257 		 */
1258 		ifidlen = in6_if2idlen(ifp);
1259 		if (ifidlen < 0) {
1260 			/* this should not happen, so we always log it. */
1261 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1262 			    if_name(ifp));
1263 			goto end;
1264 		}
1265 		if (ifidlen + pr->ndpr_plen != 128) {
1266 			nd6log((LOG_INFO,
1267 			    "prelist_update: invalid prefixlen "
1268 			    "%d for %s, ignored\n",
1269 			    pr->ndpr_plen, if_name(ifp)));
1270 			goto end;
1271 		}
1272 
1273 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1274 			/*
1275 			 * note that we should use pr (not new) for reference.
1276 			 */
1277 			pr->ndpr_refcnt++;
1278 			ia6->ia6_ndpr = pr;
1279 
1280 			/*
1281 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
1282 			 * When a new public address is created as described
1283 			 * in RFC2462, also create a new temporary address.
1284 			 *
1285 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
1286 			 * When an interface connects to a new link, a new
1287 			 * randomized interface identifier should be generated
1288 			 * immediately together with a new set of temporary
1289 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1290 			 * in6_tmpifadd().
1291 			 */
1292 			if (ip6_use_tempaddr) {
1293 				int e;
1294 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1295 					nd6log((LOG_NOTICE, "prelist_update: "
1296 					    "failed to create a temporary "
1297 					    "address, errno=%d\n",
1298 					    e));
1299 				}
1300 			}
1301 
1302 			/*
1303 			 * A newly added address might affect the status
1304 			 * of other addresses, so we check and update it.
1305 			 * XXX: what if address duplication happens?
1306 			 */
1307 			pfxlist_onlink_check();
1308 		} else {
1309 			/* just set an error. do not bark here. */
1310 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1311 		}
1312 	}
1313 
1314  end:
1315 	splx(s);
1316 	return error;
1317 }
1318 
1319 /*
1320  * A supplement function used in the on-link detection below;
1321  * detect if a given prefix has a (probably) reachable advertising router.
1322  * XXX: lengthy function name...
1323  */
1324 static struct nd_pfxrouter *
1325 find_pfxlist_reachable_router(struct nd_prefix *pr)
1326 {
1327 	struct nd_pfxrouter *pfxrtr;
1328 	struct rtentry *rt;
1329 	struct llinfo_nd6 *ln;
1330 
1331 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1332 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1333 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1334 		    pfxrtr->router->ifp)) &&
1335 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1336 		    ND6_IS_LLINFO_PROBREACH(ln))
1337 			break;	/* found */
1338 	}
1339 
1340 	return (pfxrtr);
1341 }
1342 
1343 /*
1344  * Check if each prefix in the prefix list has at least one available router
1345  * that advertised the prefix (a router is "available" if its neighbor cache
1346  * entry is reachable or probably reachable).
1347  * If the check fails, the prefix may be off-link, because, for example,
1348  * we have moved from the network but the lifetime of the prefix has not
1349  * expired yet.  So we should not use the prefix if there is another prefix
1350  * that has an available router.
1351  * But, if there is no prefix that has an available router, we still regards
1352  * all the prefixes as on-link.  This is because we can't tell if all the
1353  * routers are simply dead or if we really moved from the network and there
1354  * is no router around us.
1355  */
1356 void
1357 pfxlist_onlink_check(void)
1358 {
1359 	struct nd_prefix *pr;
1360 	struct in6_ifaddr *ifa;
1361 	struct nd_defrouter *dr;
1362 	struct nd_pfxrouter *pfxrtr = NULL;
1363 
1364 	/*
1365 	 * Check if there is a prefix that has a reachable advertising
1366 	 * router.
1367 	 */
1368 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1369 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1370 			break;
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, &nd_defrouter, dr_entry) {
1378 			struct nd_prefix *pr0;
1379 
1380 			LIST_FOREACH(pr0, &nd_prefix, ndpr_entry) {
1381 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1382 					break;
1383 			}
1384 			if (pfxrtr)
1385 				break;
1386 		}
1387 	}
1388 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
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, &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_stateflags & NDPRF_DETACHED) == 0 &&
1411 			    find_pfxlist_reachable_router(pr) == NULL)
1412 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1413 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1414 			    find_pfxlist_reachable_router(pr) != 0)
1415 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1416 		}
1417 	} else {
1418 		/* there is no prefix that has a reachable router */
1419 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1420 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1421 				continue;
1422 
1423 			if (pr->ndpr_raf_onlink == 0)
1424 				continue;
1425 
1426 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1427 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1428 		}
1429 	}
1430 
1431 	/*
1432 	 * Remove each interface route associated with a (just) detached
1433 	 * prefix, and reinstall the interface route for a (just) attached
1434 	 * prefix.  Note that all attempt of reinstallation does not
1435 	 * necessarily success, when a same prefix is shared among multiple
1436 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1437 	 * so we don't have to care about them.
1438 	 */
1439 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
1440 		int e;
1441 
1442 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1443 			continue;
1444 
1445 		if (pr->ndpr_raf_onlink == 0)
1446 			continue;
1447 
1448 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1449 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1450 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1451 				nd6log((LOG_ERR,
1452 				    "pfxlist_onlink_check: failed to "
1453 				    "make %s/%d offlink, errno=%d\n",
1454 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1455 				    pr->ndpr_plen, e));
1456 			}
1457 		}
1458 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1459 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1460 		    pr->ndpr_raf_onlink) {
1461 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1462 				nd6log((LOG_ERR,
1463 				    "pfxlist_onlink_check: failed to "
1464 				    "make %s/%d onlink, errno=%d\n",
1465 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1466 				    pr->ndpr_plen, e));
1467 			}
1468 		}
1469 	}
1470 
1471 	/*
1472 	 * Changes on the prefix status might affect address status as well.
1473 	 * Make sure that all addresses derived from an attached prefix are
1474 	 * attached, and that all addresses derived from a detached prefix are
1475 	 * detached.  Note, however, that a manually configured address should
1476 	 * always be attached.
1477 	 * The precise detection logic is same as the one for prefixes.
1478 	 */
1479 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1480 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1481 			continue;
1482 
1483 		if (ifa->ia6_ndpr == NULL) {
1484 			/*
1485 			 * This can happen when we first configure the address
1486 			 * (i.e. the address exists, but the prefix does not).
1487 			 * XXX: complicated relationships...
1488 			 */
1489 			continue;
1490 		}
1491 
1492 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1493 			break;
1494 	}
1495 	if (ifa) {
1496 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1497 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1498 				continue;
1499 
1500 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1501 				continue;
1502 
1503 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1504 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1505 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1506 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1507 					nd6_dad_start((struct ifaddr *)ifa,
1508 					    0);
1509 				}
1510 			} else {
1511 				if ((ifa->ia6_flags & IN6_IFF_DETACHED) == 0) {
1512 					ifa->ia6_flags |= IN6_IFF_DETACHED;
1513 				}
1514 			}
1515 		}
1516 	}
1517 	else {
1518 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1519 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1520 				continue;
1521 
1522 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1523 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1524 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1525 				/* Do we need a delay in this case? */
1526 				nd6_dad_start((struct ifaddr *)ifa, 0);
1527 			}
1528 		}
1529 	}
1530 }
1531 
1532 int
1533 nd6_prefix_onlink(struct nd_prefix *pr)
1534 {
1535 	struct ifaddr *ifa;
1536 	struct ifnet *ifp = pr->ndpr_ifp;
1537 	struct sockaddr_in6 mask6;
1538 	struct nd_prefix *opr;
1539 	u_long rtflags;
1540 	int error = 0;
1541 	struct rtentry *rt = NULL;
1542 
1543 	/* sanity check */
1544 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1545 		nd6log((LOG_ERR,
1546 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1547 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1548 		return (EEXIST);
1549 	}
1550 
1551 	/*
1552 	 * Add the interface route associated with the prefix.  Before
1553 	 * installing the route, check if there's the same prefix on another
1554 	 * interface, and the prefix has already installed the interface route.
1555 	 * Although such a configuration is expected to be rare, we explicitly
1556 	 * allow it.
1557 	 */
1558 	LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1559 		if (opr == pr)
1560 			continue;
1561 
1562 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1563 			continue;
1564 
1565 		if (opr->ndpr_plen == pr->ndpr_plen &&
1566 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1567 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1568 			return (0);
1569 	}
1570 
1571 	/*
1572 	 * We prefer link-local addresses as the associated interface address.
1573 	 */
1574 	/* search for a link-local addr */
1575 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1576 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1577 	if (ifa == NULL) {
1578 		/* XXX: freebsd does not have ifa_ifwithaf */
1579 		IFADDR_FOREACH(ifa, ifp) {
1580 			if (ifa->ifa_addr->sa_family == AF_INET6)
1581 				break;
1582 		}
1583 		/* should we care about ia6_flags? */
1584 	}
1585 	if (ifa == NULL) {
1586 		/*
1587 		 * This can still happen, when, for example, we receive an RA
1588 		 * containing a prefix with the L bit set and the A bit clear,
1589 		 * after removing all IPv6 addresses on the receiving
1590 		 * interface.  This should, of course, be rare though.
1591 		 */
1592 		nd6log((LOG_NOTICE,
1593 		    "nd6_prefix_onlink: failed to find any ifaddr"
1594 		    " to add route for a prefix(%s/%d) on %s\n",
1595 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1596 		    pr->ndpr_plen, if_name(ifp)));
1597 		return (0);
1598 	}
1599 
1600 	/*
1601 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1602 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1603 	 */
1604 	memset(&mask6, 0, sizeof(mask6));
1605 	mask6.sin6_len = sizeof(mask6);
1606 	mask6.sin6_addr = pr->ndpr_mask;
1607 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
1608 	rtflags = ifa->ifa_flags | RTF_UP;
1609 	if (nd6_need_cache(ifp)) {
1610 		/* explicitly set in case ifa_flags does not set the flag. */
1611 		rtflags |= RTF_CLONING;
1612 	} else {
1613 		/*
1614 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1615 		 */
1616 		rtflags &= ~RTF_CLONING;
1617 	}
1618 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1619 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1620 	if (error == 0) {
1621 		if (rt != NULL) { /* this should be non NULL, though */
1622 			nd6_rtmsg(RTM_ADD, rt);
1623 			nd6_numroutes++;
1624 		}
1625 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1626 	} else {
1627 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1628 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1629 		    "errno = %d\n",
1630 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
1631 		    pr->ndpr_plen, if_name(ifp),
1632 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1633 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
1634 	}
1635 
1636 	if (rt != NULL)
1637 		rt->rt_refcnt--;
1638 
1639 	return (error);
1640 }
1641 
1642 int
1643 nd6_prefix_offlink(struct nd_prefix *pr)
1644 {
1645 	int error = 0;
1646 	struct ifnet *ifp = pr->ndpr_ifp;
1647 	struct nd_prefix *opr;
1648 	struct sockaddr_in6 sa6, mask6;
1649 	struct rtentry *rt = NULL;
1650 
1651 	/* sanity check */
1652 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1653 		nd6log((LOG_ERR,
1654 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1655 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
1656 		return (EEXIST);
1657 	}
1658 
1659 	sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1660 	sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
1661 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1662 	    (struct sockaddr *)&mask6, 0, &rt);
1663 	if (error == 0) {
1664 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1665 
1666 		/* report the route deletion to the routing socket. */
1667 		if (rt != NULL) {
1668 			nd6_rtmsg(RTM_DELETE, rt);
1669 			nd6_numroutes--;
1670 		}
1671 
1672 		/*
1673 		 * There might be the same prefix on another interface,
1674 		 * the prefix which could not be on-link just because we have
1675 		 * the interface route (see comments in nd6_prefix_onlink).
1676 		 * If there's one, try to make the prefix on-link on the
1677 		 * interface.
1678 		 */
1679 		LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
1680 			if (opr == pr)
1681 				continue;
1682 
1683 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1684 				continue;
1685 
1686 			/*
1687 			 * KAME specific: detached prefixes should not be
1688 			 * on-link.
1689 			 */
1690 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1691 				continue;
1692 
1693 			if (opr->ndpr_plen == pr->ndpr_plen &&
1694 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1695 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1696 				int e;
1697 
1698 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1699 					nd6log((LOG_ERR,
1700 					    "nd6_prefix_offlink: failed to "
1701 					    "recover a prefix %s/%d from %s "
1702 					    "to %s (errno = %d)\n",
1703 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
1704 					    opr->ndpr_plen, if_name(ifp),
1705 					    if_name(opr->ndpr_ifp), e));
1706 				}
1707 			}
1708 		}
1709 	} else {
1710 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1711 		nd6log((LOG_ERR,
1712 		    "nd6_prefix_offlink: failed to delete route: "
1713 		    "%s/%d on %s (errno = %d)\n",
1714 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
1715 		    error));
1716 	}
1717 
1718 	if (rt != NULL) {
1719 		if (rt->rt_refcnt <= 0) {
1720 			/* XXX: we should free the entry ourselves. */
1721 			rt->rt_refcnt++;
1722 			rtfree(rt);
1723 		}
1724 	}
1725 
1726 	return (error);
1727 }
1728 
1729 static struct in6_ifaddr *
1730 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1731 {
1732 	struct ifnet *ifp = pr->ndpr_ifp;
1733 	struct ifaddr *ifa;
1734 	struct in6_aliasreq ifra;
1735 	struct in6_ifaddr *ia, *ib;
1736 	int error, plen0;
1737 	struct in6_addr mask;
1738 	int prefixlen = pr->ndpr_plen;
1739 	int updateflags;
1740 
1741 	in6_prefixlen2mask(&mask, prefixlen);
1742 
1743 	/*
1744 	 * find a link-local address (will be interface ID).
1745 	 * Is it really mandatory? Theoretically, a global or a site-local
1746 	 * address can be configured without a link-local address, if we
1747 	 * have a unique interface identifier...
1748 	 *
1749 	 * it is not mandatory to have a link-local address, we can generate
1750 	 * interface identifier on the fly.  we do this because:
1751 	 * (1) it should be the easiest way to find interface identifier.
1752 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1753 	 * for multiple addresses on a single interface, and possible shortcut
1754 	 * of DAD.  we omitted DAD for this reason in the past.
1755 	 * (3) a user can prevent autoconfiguration of global address
1756 	 * by removing link-local address by hand (this is partly because we
1757 	 * don't have other way to control the use of IPv6 on an interface.
1758 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1759 	 * (4) it is easier to manage when an interface has addresses
1760 	 * with the same interface identifier, than to have multiple addresses
1761 	 * with different interface identifiers.
1762 	 */
1763 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1764 	if (ifa)
1765 		ib = (struct in6_ifaddr *)ifa;
1766 	else
1767 		return NULL;
1768 
1769 #if 0 /* don't care link local addr state, and always do DAD */
1770 	/* if link-local address is not eligible, do not autoconfigure. */
1771 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
1772 		printf("in6_ifadd: link-local address not ready\n");
1773 		return NULL;
1774 	}
1775 #endif
1776 
1777 	/* prefixlen + ifidlen must be equal to 128 */
1778 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1779 	if (prefixlen != plen0) {
1780 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1781 		    "(prefix=%d ifid=%d)\n",
1782 		    if_name(ifp), prefixlen, 128 - plen0));
1783 		return NULL;
1784 	}
1785 
1786 	/* make ifaddr */
1787 
1788 	memset(&ifra, 0, sizeof(ifra));
1789 	/*
1790 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1791 	 * for safety.
1792 	 */
1793 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1794 	sockaddr_in6_init(&ifra.ifra_addr, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
1795 	/* prefix */
1796 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1797 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1798 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1799 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1800 
1801 	/* interface ID */
1802 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1803 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1804 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1805 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1806 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1807 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1808 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1809 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1810 
1811 	/* new prefix mask. */
1812 	sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
1813 
1814 	/* lifetimes */
1815 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1816 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1817 
1818 	/* XXX: scope zone ID? */
1819 
1820 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1821 
1822 	/*
1823 	 * Make sure that we do not have this address already.  This should
1824 	 * usually not happen, but we can still see this case, e.g., if we
1825 	 * have manually configured the exact address to be configured.
1826 	 */
1827 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1828 		/* this should be rare enough to make an explicit log */
1829 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1830 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
1831 		return (NULL);
1832 	}
1833 
1834 	/*
1835 	 * Allocate ifaddr structure, link into chain, etc.
1836 	 * If we are going to create a new address upon receiving a multicasted
1837 	 * RA, we need to impose a random delay before starting DAD.
1838 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1839 	 */
1840 	updateflags = 0;
1841 	if (mcast)
1842 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1843 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1844 		nd6log((LOG_ERR,
1845 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1846 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
1847 		    error));
1848 		return (NULL);	/* ifaddr must not have been allocated. */
1849 	}
1850 
1851 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1852 
1853 	return (ia);		/* this is always non-NULL */
1854 }
1855 
1856 int
1857 in6_tmpifadd(
1858 	const struct in6_ifaddr *ia0, /* corresponding public address */
1859 	int forcegen,
1860 	int dad_delay)
1861 {
1862 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1863 	struct in6_ifaddr *newia, *ia;
1864 	struct in6_aliasreq ifra;
1865 	int i, error;
1866 	int trylimit = 3;	/* XXX: adhoc value */
1867 	int updateflags;
1868 	u_int32_t randid[2];
1869 	u_int32_t vltime0, pltime0;
1870 
1871 	memset(&ifra, 0, sizeof(ifra));
1872 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1873 	ifra.ifra_addr = ia0->ia_addr;
1874 	/* copy prefix mask */
1875 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1876 	/* clear the old IFID */
1877 	for (i = 0; i < 4; i++) {
1878 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1879 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1880 	}
1881 
1882   again:
1883 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1884 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1885 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1886 		    "random IFID\n"));
1887 		return (EINVAL);
1888 	}
1889 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1890 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1891 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1892 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1893 
1894 	/*
1895 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1896 	 * However, we may still have a chance to see collision, because
1897 	 * there may be a time lag between generation of the ID and generation
1898 	 * of the address.  So, we'll do one more sanity check.
1899 	 */
1900 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1901 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1902 		    &ifra.ifra_addr.sin6_addr)) {
1903 			if (trylimit-- == 0) {
1904 				/*
1905 				 * Give up.  Something strange should have
1906 				 * happened.
1907 				 */
1908 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1909 				    "find a unique random IFID\n"));
1910 				return (EEXIST);
1911 			}
1912 			forcegen = 1;
1913 			goto again;
1914 		}
1915 	}
1916 
1917 	/*
1918 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1919          * public address or TEMP_VALID_LIFETIME.
1920 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1921          * of the public address or TEMP_PREFERRED_LIFETIME -
1922          * DESYNC_FACTOR.
1923 	 */
1924 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1925 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1926 		    (ia0->ia6_lifetime.ia6t_vltime -
1927 		    (time_second - ia0->ia6_updatetime));
1928 		if (vltime0 > ip6_temp_valid_lifetime)
1929 			vltime0 = ip6_temp_valid_lifetime;
1930 	} else
1931 		vltime0 = ip6_temp_valid_lifetime;
1932 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1933 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1934 		    (ia0->ia6_lifetime.ia6t_pltime -
1935 		    (time_second - ia0->ia6_updatetime));
1936 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
1937 			pltime0 = ip6_temp_preferred_lifetime -
1938 			    ip6_desync_factor;
1939 		}
1940 	} else
1941 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
1942 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1943 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1944 
1945 	/*
1946 	 * A temporary address is created only if this calculated Preferred
1947 	 * Lifetime is greater than REGEN_ADVANCE time units.
1948 	 */
1949 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
1950 		return (0);
1951 
1952 	/* XXX: scope zone ID? */
1953 
1954 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1955 
1956 	/* allocate ifaddr structure, link into chain, etc. */
1957 	updateflags = 0;
1958 	if (dad_delay)
1959 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1960 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1961 		return (error);
1962 
1963 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1964 	if (newia == NULL) {	/* XXX: can it happen? */
1965 		nd6log((LOG_ERR,
1966 		    "in6_tmpifadd: ifa update succeeded, but we got "
1967 		    "no ifaddr\n"));
1968 		return (EINVAL); /* XXX */
1969 	}
1970 	newia->ia6_ndpr = ia0->ia6_ndpr;
1971 	newia->ia6_ndpr->ndpr_refcnt++;
1972 
1973 	/*
1974 	 * A newly added address might affect the status of other addresses.
1975 	 * XXX: when the temporary address is generated with a new public
1976 	 * address, the onlink check is redundant.  However, it would be safe
1977 	 * to do the check explicitly everywhere a new address is generated,
1978 	 * and, in fact, we surely need the check when we create a new
1979 	 * temporary address due to deprecation of an old temporary address.
1980 	 */
1981 	pfxlist_onlink_check();
1982 
1983 	return (0);
1984 }
1985 
1986 static int
1987 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1988 {
1989 
1990 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
1991 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
1992 		nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
1993 		    "(%d) is greater than valid lifetime(%d)\n",
1994 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
1995 		return (EINVAL);
1996 	}
1997 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1998 		ndpr->ndpr_preferred = 0;
1999 	else
2000 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2001 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2002 		ndpr->ndpr_expire = 0;
2003 	else
2004 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2005 
2006 	return 0;
2007 }
2008 
2009 static void
2010 in6_init_address_ltimes(struct nd_prefix *new,
2011     struct in6_addrlifetime *lt6)
2012 {
2013 
2014 	/* Valid lifetime must not be updated unless explicitly specified. */
2015 	/* init ia6t_expire */
2016 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2017 		lt6->ia6t_expire = 0;
2018 	else {
2019 		lt6->ia6t_expire = time_second;
2020 		lt6->ia6t_expire += lt6->ia6t_vltime;
2021 	}
2022 
2023 	/* init ia6t_preferred */
2024 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2025 		lt6->ia6t_preferred = 0;
2026 	else {
2027 		lt6->ia6t_preferred = time_second;
2028 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2029 	}
2030 }
2031 
2032 /*
2033  * Delete all the routing table entries that use the specified gateway.
2034  * XXX: this function causes search through all entries of routing table, so
2035  * it shouldn't be called when acting as a router.
2036  */
2037 void
2038 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2039 {
2040 	int s = splsoftnet();
2041 
2042 	/* We'll care only link-local addresses */
2043 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2044 		splx(s);
2045 		return;
2046 	}
2047 
2048 	rt_walktree(AF_INET6, rt6_deleteroute, (void *)gateway);
2049 	splx(s);
2050 }
2051 
2052 static int
2053 rt6_deleteroute(struct rtentry *rt, void *arg)
2054 {
2055 #define SIN6(s)	((struct sockaddr_in6 *)s)
2056 	struct in6_addr *gate = (struct in6_addr *)arg;
2057 
2058 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2059 		return (0);
2060 
2061 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr))
2062 		return (0);
2063 
2064 	/*
2065 	 * Do not delete a static route.
2066 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2067 	 * 'cloned' bit instead?
2068 	 */
2069 	if ((rt->rt_flags & RTF_STATIC) != 0)
2070 		return (0);
2071 
2072 	/*
2073 	 * We delete only host route. This means, in particular, we don't
2074 	 * delete default route.
2075 	 */
2076 	if ((rt->rt_flags & RTF_HOST) == 0)
2077 		return (0);
2078 
2079 	return (rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
2080 	    rt_mask(rt), rt->rt_flags, 0));
2081 #undef SIN6
2082 }
2083 
2084 int
2085 nd6_setdefaultiface(int ifindex)
2086 {
2087 	int error = 0;
2088 
2089 	if (ifindex < 0 || if_indexlim <= ifindex)
2090 		return (EINVAL);
2091 	if (ifindex != 0 && !ifindex2ifnet[ifindex])
2092 		return (EINVAL);
2093 
2094 	if (nd6_defifindex != ifindex) {
2095 		nd6_defifindex = ifindex;
2096 		if (nd6_defifindex > 0) {
2097 			nd6_defifp = ifindex2ifnet[nd6_defifindex];
2098 		} else
2099 			nd6_defifp = NULL;
2100 
2101 		/*
2102 		 * Our current implementation assumes one-to-one maping between
2103 		 * interfaces and links, so it would be natural to use the
2104 		 * default interface as the default link.
2105 		 */
2106 		scope6_setdefault(nd6_defifp);
2107 	}
2108 
2109 	return (error);
2110 }
2111