xref: /freebsd/sys/netinet6/nd6_rtr.c (revision bfe691b2f75de2224c7ceb304ebcdef2b42d4179)
1 /*	$FreeBSD$	*/
2 /*	$KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei 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 "opt_inet.h"
34 #include "opt_inet6.h"
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/syslog.h>
46 #include <sys/queue.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/scope6_var.h>
62 
63 #define SDL(s)	((struct sockaddr_dl *)s)
64 
65 static int rtpref __P((struct nd_defrouter *));
66 static struct nd_defrouter *defrtrlist_update __P((struct nd_defrouter *));
67 static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
68     struct mbuf *, int));
69 static struct in6_ifaddr *in6_ifadd __P((struct nd_prefixctl *,	int));
70 static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
71 	struct nd_defrouter *));
72 static void pfxrtr_add __P((struct nd_prefix *, struct nd_defrouter *));
73 static void pfxrtr_del __P((struct nd_pfxrouter *));
74 static struct nd_pfxrouter *find_pfxlist_reachable_router
75 	__P((struct nd_prefix *));
76 static void defrouter_delreq __P((struct nd_defrouter *));
77 static void nd6_rtmsg __P((int, struct rtentry *));
78 
79 static int in6_init_prefix_ltimes __P((struct nd_prefix *));
80 static void in6_init_address_ltimes __P((struct nd_prefix *,
81 	struct in6_addrlifetime *));
82 
83 static int rt6_deleteroute __P((struct radix_node *, void *));
84 
85 extern int nd6_recalc_reachtm_interval;
86 
87 static struct ifnet *nd6_defifp;
88 int nd6_defifindex;
89 
90 int ip6_use_tempaddr = 0;
91 
92 int ip6_desync_factor;
93 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
94 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
95 /*
96  * shorter lifetimes for debugging purposes.
97 int ip6_temp_preferred_lifetime = 800;
98 static int ip6_temp_valid_lifetime = 1800;
99 */
100 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
101 
102 /* RTPREF_MEDIUM has to be 0! */
103 #define RTPREF_HIGH	1
104 #define RTPREF_MEDIUM	0
105 #define RTPREF_LOW	(-1)
106 #define RTPREF_RESERVED	(-2)
107 #define RTPREF_INVALID	(-3)	/* internal */
108 
109 /*
110  * Receive Router Solicitation Message - just for routers.
111  * Router solicitation/advertisement is mostly managed by userland program
112  * (rtadvd) so here we have no function like nd6_ra_output().
113  *
114  * Based on RFC 2461
115  */
116 void
117 nd6_rs_input(m, off, icmp6len)
118 	struct	mbuf *m;
119 	int off, icmp6len;
120 {
121 	struct ifnet *ifp = m->m_pkthdr.rcvif;
122 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
123 	struct nd_router_solicit *nd_rs;
124 	struct in6_addr saddr6 = ip6->ip6_src;
125 	char *lladdr = NULL;
126 	int lladdrlen = 0;
127 	union nd_opts ndopts;
128 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
129 
130 	/* If I'm not a router, ignore it. */
131 	if (ip6_accept_rtadv != 0 || ip6_forwarding != 1)
132 		goto freeit;
133 
134 	/* Sanity checks */
135 	if (ip6->ip6_hlim != 255) {
136 		nd6log((LOG_ERR,
137 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
138 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
139 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
140 		goto bad;
141 	}
142 
143 	/*
144 	 * Don't update the neighbor cache, if src = ::.
145 	 * This indicates that the src has no IP address assigned yet.
146 	 */
147 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
148 		goto freeit;
149 
150 #ifndef PULLDOWN_TEST
151 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
152 	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
153 #else
154 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
155 	if (nd_rs == NULL) {
156 		icmp6stat.icp6s_tooshort++;
157 		return;
158 	}
159 #endif
160 
161 	icmp6len -= sizeof(*nd_rs);
162 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
163 	if (nd6_options(&ndopts) < 0) {
164 		nd6log((LOG_INFO,
165 		    "nd6_rs_input: invalid ND option, ignored\n"));
166 		/* nd6_options have incremented stats */
167 		goto freeit;
168 	}
169 
170 	if (ndopts.nd_opts_src_lladdr) {
171 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
172 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
173 	}
174 
175 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
176 		nd6log((LOG_INFO,
177 		    "nd6_rs_input: lladdrlen mismatch for %s "
178 		    "(if %d, RS packet %d)\n",
179 		    ip6_sprintf(ip6bufs, &saddr6),
180 		    ifp->if_addrlen, lladdrlen - 2));
181 		goto bad;
182 	}
183 
184 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
185 
186  freeit:
187 	m_freem(m);
188 	return;
189 
190  bad:
191 	icmp6stat.icp6s_badrs++;
192 	m_freem(m);
193 }
194 
195 /*
196  * Receive Router Advertisement Message.
197  *
198  * Based on RFC 2461
199  * TODO: on-link bit on prefix information
200  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
201  */
202 void
203 nd6_ra_input(m, off, icmp6len)
204 	struct	mbuf *m;
205 	int off, icmp6len;
206 {
207 	struct ifnet *ifp = m->m_pkthdr.rcvif;
208 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
209 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
210 	struct nd_router_advert *nd_ra;
211 	struct in6_addr saddr6 = ip6->ip6_src;
212 	int mcast = 0;
213 	union nd_opts ndopts;
214 	struct nd_defrouter *dr;
215 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
216 
217 	/*
218 	 * We only accept RAs only when
219 	 * the system-wide variable allows the acceptance, and
220 	 * per-interface variable allows RAs on the receiving interface.
221 	 */
222 	if (ip6_accept_rtadv == 0)
223 		goto freeit;
224 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
225 		goto freeit;
226 
227 	if (ip6->ip6_hlim != 255) {
228 		nd6log((LOG_ERR,
229 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
230 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
231 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
232 		goto bad;
233 	}
234 
235 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
236 		nd6log((LOG_ERR,
237 		    "nd6_ra_input: src %s is not link-local\n",
238 		    ip6_sprintf(ip6bufs, &saddr6)));
239 		goto bad;
240 	}
241 
242 #ifndef PULLDOWN_TEST
243 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
244 	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
245 #else
246 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
247 	if (nd_ra == NULL) {
248 		icmp6stat.icp6s_tooshort++;
249 		return;
250 	}
251 #endif
252 
253 	icmp6len -= sizeof(*nd_ra);
254 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
255 	if (nd6_options(&ndopts) < 0) {
256 		nd6log((LOG_INFO,
257 		    "nd6_ra_input: invalid ND option, ignored\n"));
258 		/* nd6_options have incremented stats */
259 		goto freeit;
260 	}
261 
262     {
263 	struct nd_defrouter dr0;
264 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
265 
266 	/* remember if this is a multicasted advertisement */
267 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
268 		mcast = 1;
269 
270 	bzero(&dr0, sizeof(dr0));
271 	dr0.rtaddr = saddr6;
272 	dr0.flags  = nd_ra->nd_ra_flags_reserved;
273 	dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
274 	dr0.expire = time_second + dr0.rtlifetime;
275 	dr0.ifp = ifp;
276 	/* unspecified or not? (RFC 2461 6.3.4) */
277 	if (advreachable) {
278 		advreachable = ntohl(advreachable);
279 		if (advreachable <= MAX_REACHABLE_TIME &&
280 		    ndi->basereachable != advreachable) {
281 			ndi->basereachable = advreachable;
282 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
283 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
284 		}
285 	}
286 	if (nd_ra->nd_ra_retransmit)
287 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
288 	if (nd_ra->nd_ra_curhoplimit)
289 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
290 	dr = defrtrlist_update(&dr0);
291     }
292 
293 	/*
294 	 * prefix
295 	 */
296 	if (ndopts.nd_opts_pi) {
297 		struct nd_opt_hdr *pt;
298 		struct nd_opt_prefix_info *pi = NULL;
299 		struct nd_prefixctl pr;
300 
301 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
302 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
303 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
304 						(pt->nd_opt_len << 3))) {
305 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
306 				continue;
307 			pi = (struct nd_opt_prefix_info *)pt;
308 
309 			if (pi->nd_opt_pi_len != 4) {
310 				nd6log((LOG_INFO,
311 				    "nd6_ra_input: invalid option "
312 				    "len %d for prefix information option, "
313 				    "ignored\n", pi->nd_opt_pi_len));
314 				continue;
315 			}
316 
317 			if (128 < pi->nd_opt_pi_prefix_len) {
318 				nd6log((LOG_INFO,
319 				    "nd6_ra_input: invalid prefix "
320 				    "len %d for prefix information option, "
321 				    "ignored\n", pi->nd_opt_pi_prefix_len));
322 				continue;
323 			}
324 
325 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
326 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
327 				nd6log((LOG_INFO,
328 				    "nd6_ra_input: invalid prefix "
329 				    "%s, ignored\n",
330 				    ip6_sprintf(ip6bufs,
331 					&pi->nd_opt_pi_prefix)));
332 				continue;
333 			}
334 
335 			bzero(&pr, sizeof(pr));
336 			pr.ndpr_prefix.sin6_family = AF_INET6;
337 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
338 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
339 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
340 
341 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
342 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
343 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
344 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
345 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
346 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
347 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
348 			(void)prelist_update(&pr, dr, m, mcast);
349 		}
350 	}
351 
352 	/*
353 	 * MTU
354 	 */
355 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
356 		u_long mtu;
357 		u_long maxmtu;
358 
359 		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
360 
361 		/* lower bound */
362 		if (mtu < IPV6_MMTU) {
363 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
364 			    "mtu=%lu sent from %s, ignoring\n",
365 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
366 			goto skip;
367 		}
368 
369 		/* upper bound */
370 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
371 		    ? ndi->maxmtu : ifp->if_mtu;
372 		if (mtu <= maxmtu) {
373 			int change = (ndi->linkmtu != mtu);
374 
375 			ndi->linkmtu = mtu;
376 			if (change) /* in6_maxmtu may change */
377 				in6_setmaxmtu();
378 		} else {
379 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
380 			    "mtu=%lu sent from %s; "
381 			    "exceeds maxmtu %lu, ignoring\n",
382 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
383 		}
384 	}
385 
386  skip:
387 
388 	/*
389 	 * Source link layer address
390 	 */
391     {
392 	char *lladdr = NULL;
393 	int lladdrlen = 0;
394 
395 	if (ndopts.nd_opts_src_lladdr) {
396 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
397 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
398 	}
399 
400 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
401 		nd6log((LOG_INFO,
402 		    "nd6_ra_input: lladdrlen mismatch for %s "
403 		    "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
404 		    ifp->if_addrlen, lladdrlen - 2));
405 		goto bad;
406 	}
407 
408 	nd6_cache_lladdr(ifp, &saddr6, lladdr,
409 	    lladdrlen, ND_ROUTER_ADVERT, 0);
410 
411 	/*
412 	 * Installing a link-layer address might change the state of the
413 	 * router's neighbor cache, which might also affect our on-link
414 	 * detection of adveritsed prefixes.
415 	 */
416 	pfxlist_onlink_check();
417     }
418 
419  freeit:
420 	m_freem(m);
421 	return;
422 
423  bad:
424 	icmp6stat.icp6s_badra++;
425 	m_freem(m);
426 }
427 
428 /*
429  * default router list proccessing sub routines
430  */
431 
432 /* tell the change to user processes watching the routing socket. */
433 static void
434 nd6_rtmsg(cmd, rt)
435 	int cmd;
436 	struct rtentry *rt;
437 {
438 	struct rt_addrinfo info;
439 
440 	bzero((caddr_t)&info, sizeof(info));
441 	info.rti_info[RTAX_DST] = rt_key(rt);
442 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
443 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
444 	if (rt->rt_ifp) {
445 		info.rti_info[RTAX_IFP] =
446 		    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
447 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
448 	}
449 
450 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
451 }
452 
453 void
454 defrouter_addreq(new)
455 	struct nd_defrouter *new;
456 {
457 	struct sockaddr_in6 def, mask, gate;
458 	struct rtentry *newrt = NULL;
459 	int s;
460 	int error;
461 
462 	bzero(&def, sizeof(def));
463 	bzero(&mask, sizeof(mask));
464 	bzero(&gate, sizeof(gate));
465 
466 	def.sin6_len = mask.sin6_len = gate.sin6_len =
467 	    sizeof(struct sockaddr_in6);
468 	def.sin6_family = gate.sin6_family = AF_INET6;
469 	gate.sin6_addr = new->rtaddr;
470 
471 	s = splnet();
472 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
473 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
474 	    RTF_GATEWAY, &newrt);
475 	if (newrt) {
476 		RT_LOCK(newrt);
477 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
478 		RT_REMREF(newrt);
479 		RT_UNLOCK(newrt);
480 	}
481 	if (error == 0)
482 		new->installed = 1;
483 	splx(s);
484 	return;
485 }
486 
487 struct nd_defrouter *
488 defrouter_lookup(addr, ifp)
489 	struct in6_addr *addr;
490 	struct ifnet *ifp;
491 {
492 	struct nd_defrouter *dr;
493 
494 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
495 	     dr = TAILQ_NEXT(dr, dr_entry)) {
496 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
497 			return (dr);
498 	}
499 
500 	return (NULL);		/* search failed */
501 }
502 
503 /*
504  * Remove the default route for a given router.
505  * This is just a subroutine function for defrouter_select(), and should
506  * not be called from anywhere else.
507  */
508 static void
509 defrouter_delreq(dr)
510 	struct nd_defrouter *dr;
511 {
512 	struct sockaddr_in6 def, mask, gate;
513 	struct rtentry *oldrt = NULL;
514 
515 	bzero(&def, sizeof(def));
516 	bzero(&mask, sizeof(mask));
517 	bzero(&gate, sizeof(gate));
518 
519 	def.sin6_len = mask.sin6_len = gate.sin6_len =
520 	    sizeof(struct sockaddr_in6);
521 	def.sin6_family = gate.sin6_family = AF_INET6;
522 	gate.sin6_addr = dr->rtaddr;
523 
524 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
525 	    (struct sockaddr *)&gate,
526 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
527 	if (oldrt) {
528 		nd6_rtmsg(RTM_DELETE, oldrt);
529 		RTFREE(oldrt);
530 	}
531 
532 	dr->installed = 0;
533 }
534 
535 /*
536  * remove all default routes from default router list
537  */
538 void
539 defrouter_reset()
540 {
541 	struct nd_defrouter *dr;
542 
543 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
544 	     dr = TAILQ_NEXT(dr, dr_entry))
545 		defrouter_delreq(dr);
546 
547 	/*
548 	 * XXX should we also nuke any default routers in the kernel, by
549 	 * going through them by rtalloc1()?
550 	 */
551 }
552 
553 void
554 defrtrlist_del(dr)
555 	struct nd_defrouter *dr;
556 {
557 	struct nd_defrouter *deldr = NULL;
558 	struct nd_prefix *pr;
559 
560 	/*
561 	 * Flush all the routing table entries that use the router
562 	 * as a next hop.
563 	 */
564 	if (!ip6_forwarding && ip6_accept_rtadv) /* XXX: better condition? */
565 		rt6_flush(&dr->rtaddr, dr->ifp);
566 
567 	if (dr->installed) {
568 		deldr = dr;
569 		defrouter_delreq(dr);
570 	}
571 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
572 
573 	/*
574 	 * Also delete all the pointers to the router in each prefix lists.
575 	 */
576 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
577 		struct nd_pfxrouter *pfxrtr;
578 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
579 			pfxrtr_del(pfxrtr);
580 	}
581 	pfxlist_onlink_check();
582 
583 	/*
584 	 * If the router is the primary one, choose a new one.
585 	 * Note that defrouter_select() will remove the current gateway
586 	 * from the routing table.
587 	 */
588 	if (deldr)
589 		defrouter_select();
590 
591 	free(dr, M_IP6NDP);
592 }
593 
594 /*
595  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
596  * draft-ietf-ipngwg-router-selection:
597  * 1) Routers that are reachable or probably reachable should be preferred.
598  *    If we have more than one (probably) reachable router, prefer ones
599  *    with the highest router preference.
600  * 2) When no routers on the list are known to be reachable or
601  *    probably reachable, routers SHOULD be selected in a round-robin
602  *    fashion, regardless of router preference values.
603  * 3) If the Default Router List is empty, assume that all
604  *    destinations are on-link.
605  *
606  * We assume nd_defrouter is sorted by router preference value.
607  * Since the code below covers both with and without router preference cases,
608  * we do not need to classify the cases by ifdef.
609  *
610  * At this moment, we do not try to install more than one default router,
611  * even when the multipath routing is available, because we're not sure about
612  * the benefits for stub hosts comparing to the risk of making the code
613  * complicated and the possibility of introducing bugs.
614  */
615 void
616 defrouter_select()
617 {
618 	int s = splnet();
619 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
620 	struct rtentry *rt = NULL;
621 	struct llinfo_nd6 *ln = NULL;
622 
623 	/*
624 	 * This function should be called only when acting as an autoconfigured
625 	 * host.  Although the remaining part of this function is not effective
626 	 * if the node is not an autoconfigured host, we explicitly exclude
627 	 * such cases here for safety.
628 	 */
629 	if (ip6_forwarding || !ip6_accept_rtadv) {
630 		nd6log((LOG_WARNING,
631 		    "defrouter_select: called unexpectedly (forwarding=%d, "
632 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
633 		splx(s);
634 		return;
635 	}
636 
637 	/*
638 	 * Let's handle easy case (3) first:
639 	 * If default router list is empty, there's nothing to be done.
640 	 */
641 	if (!TAILQ_FIRST(&nd_defrouter)) {
642 		splx(s);
643 		return;
644 	}
645 
646 	/*
647 	 * Search for a (probably) reachable router from the list.
648 	 * We just pick up the first reachable one (if any), assuming that
649 	 * the ordering rule of the list described in defrtrlist_update().
650 	 */
651 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
652 	     dr = TAILQ_NEXT(dr, dr_entry)) {
653 		if (selected_dr == NULL &&
654 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
655 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
656 		    ND6_IS_LLINFO_PROBREACH(ln)) {
657 			selected_dr = dr;
658 		}
659 
660 		if (dr->installed && installed_dr == NULL)
661 			installed_dr = dr;
662 		else if (dr->installed && installed_dr) {
663 			/* this should not happen.  warn for diagnosis. */
664 			log(LOG_ERR, "defrouter_select: more than one router"
665 			    " is installed\n");
666 		}
667 	}
668 	/*
669 	 * If none of the default routers was found to be reachable,
670 	 * round-robin the list regardless of preference.
671 	 * Otherwise, if we have an installed router, check if the selected
672 	 * (reachable) router should really be preferred to the installed one.
673 	 * We only prefer the new router when the old one is not reachable
674 	 * or when the new one has a really higher preference value.
675 	 */
676 	if (selected_dr == NULL) {
677 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
678 			selected_dr = TAILQ_FIRST(&nd_defrouter);
679 		else
680 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
681 	} else if (installed_dr &&
682 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
683 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
684 	    ND6_IS_LLINFO_PROBREACH(ln) &&
685 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
686 		selected_dr = installed_dr;
687 	}
688 
689 	/*
690 	 * If the selected router is different than the installed one,
691 	 * remove the installed router and install the selected one.
692 	 * Note that the selected router is never NULL here.
693 	 */
694 	if (installed_dr != selected_dr) {
695 		if (installed_dr)
696 			defrouter_delreq(installed_dr);
697 		defrouter_addreq(selected_dr);
698 	}
699 
700 	splx(s);
701 	return;
702 }
703 
704 /*
705  * for default router selection
706  * regards router-preference field as a 2-bit signed integer
707  */
708 static int
709 rtpref(struct nd_defrouter *dr)
710 {
711 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
712 	case ND_RA_FLAG_RTPREF_HIGH:
713 		return (RTPREF_HIGH);
714 	case ND_RA_FLAG_RTPREF_MEDIUM:
715 	case ND_RA_FLAG_RTPREF_RSV:
716 		return (RTPREF_MEDIUM);
717 	case ND_RA_FLAG_RTPREF_LOW:
718 		return (RTPREF_LOW);
719 	default:
720 		/*
721 		 * This case should never happen.  If it did, it would mean a
722 		 * serious bug of kernel internal.  We thus always bark here.
723 		 * Or, can we even panic?
724 		 */
725 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
726 		return (RTPREF_INVALID);
727 	}
728 	/* NOTREACHED */
729 }
730 
731 static struct nd_defrouter *
732 defrtrlist_update(new)
733 	struct nd_defrouter *new;
734 {
735 	struct nd_defrouter *dr, *n;
736 	int s = splnet();
737 
738 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
739 		/* entry exists */
740 		if (new->rtlifetime == 0) {
741 			defrtrlist_del(dr);
742 			dr = NULL;
743 		} else {
744 			int oldpref = rtpref(dr);
745 
746 			/* override */
747 			dr->flags = new->flags; /* xxx flag check */
748 			dr->rtlifetime = new->rtlifetime;
749 			dr->expire = new->expire;
750 
751 			/*
752 			 * If the preference does not change, there's no need
753 			 * to sort the entries.
754 			 */
755 			if (rtpref(new) == oldpref) {
756 				splx(s);
757 				return (dr);
758 			}
759 
760 			/*
761 			 * preferred router may be changed, so relocate
762 			 * this router.
763 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
764 			 * However, since defrtrlist_del() has many side
765 			 * effects, we intentionally do so here.
766 			 * defrouter_select() below will handle routing
767 			 * changes later.
768 			 */
769 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
770 			n = dr;
771 			goto insert;
772 		}
773 		splx(s);
774 		return (dr);
775 	}
776 
777 	/* entry does not exist */
778 	if (new->rtlifetime == 0) {
779 		splx(s);
780 		return (NULL);
781 	}
782 
783 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
784 	if (n == NULL) {
785 		splx(s);
786 		return (NULL);
787 	}
788 	bzero(n, sizeof(*n));
789 	*n = *new;
790 
791 insert:
792 	/*
793 	 * Insert the new router in the Default Router List;
794 	 * The Default Router List should be in the descending order
795 	 * of router-preferece.  Routers with the same preference are
796 	 * sorted in the arriving time order.
797 	 */
798 
799 	/* insert at the end of the group */
800 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
801 	     dr = TAILQ_NEXT(dr, dr_entry)) {
802 		if (rtpref(n) > rtpref(dr))
803 			break;
804 	}
805 	if (dr)
806 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
807 	else
808 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
809 
810 	defrouter_select();
811 
812 	splx(s);
813 
814 	return (n);
815 }
816 
817 static struct nd_pfxrouter *
818 pfxrtr_lookup(pr, dr)
819 	struct nd_prefix *pr;
820 	struct nd_defrouter *dr;
821 {
822 	struct nd_pfxrouter *search;
823 
824 	for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
825 		if (search->router == dr)
826 			break;
827 	}
828 
829 	return (search);
830 }
831 
832 static void
833 pfxrtr_add(pr, dr)
834 	struct nd_prefix *pr;
835 	struct nd_defrouter *dr;
836 {
837 	struct nd_pfxrouter *new;
838 
839 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
840 	if (new == NULL)
841 		return;
842 	bzero(new, sizeof(*new));
843 	new->router = dr;
844 
845 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
846 
847 	pfxlist_onlink_check();
848 }
849 
850 static void
851 pfxrtr_del(pfr)
852 	struct nd_pfxrouter *pfr;
853 {
854 	LIST_REMOVE(pfr, pfr_entry);
855 	free(pfr, M_IP6NDP);
856 }
857 
858 struct nd_prefix *
859 nd6_prefix_lookup(key)
860 	struct nd_prefixctl *key;
861 {
862 	struct nd_prefix *search;
863 
864 	for (search = nd_prefix.lh_first; search; search = search->ndpr_next) {
865 		if (key->ndpr_ifp == search->ndpr_ifp &&
866 		    key->ndpr_plen == search->ndpr_plen &&
867 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
868 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
869 			break;
870 		}
871 	}
872 
873 	return (search);
874 }
875 
876 int
877 nd6_prelist_add(pr, dr, newp)
878 	struct nd_prefixctl *pr;
879 	struct nd_prefix **newp;
880 	struct nd_defrouter *dr;
881 {
882 	struct nd_prefix *new = NULL;
883 	int error = 0;
884 	int i, s;
885 	char ip6buf[INET6_ADDRSTRLEN];
886 
887 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
888 	if (new == NULL)
889 		return(ENOMEM);
890 	bzero(new, sizeof(*new));
891 	new->ndpr_ifp = pr->ndpr_ifp;
892 	new->ndpr_prefix = pr->ndpr_prefix;
893 	new->ndpr_plen = pr->ndpr_plen;
894 	new->ndpr_vltime = pr->ndpr_vltime;
895 	new->ndpr_pltime = pr->ndpr_pltime;
896 	new->ndpr_flags = pr->ndpr_flags;
897 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
898 		free(new, M_IP6NDP);
899 		return(error);
900 	}
901 	new->ndpr_lastupdate = time_second;
902 	if (newp != NULL)
903 		*newp = new;
904 
905 	/* initialization */
906 	LIST_INIT(&new->ndpr_advrtrs);
907 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
908 	/* make prefix in the canonical form */
909 	for (i = 0; i < 4; i++)
910 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
911 		    new->ndpr_mask.s6_addr32[i];
912 
913 	s = splnet();
914 	/* link ndpr_entry to nd_prefix list */
915 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
916 	splx(s);
917 
918 	/* ND_OPT_PI_FLAG_ONLINK processing */
919 	if (new->ndpr_raf_onlink) {
920 		int e;
921 
922 		if ((e = nd6_prefix_onlink(new)) != 0) {
923 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
924 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
925 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
926 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
927 			/* proceed anyway. XXX: is it correct? */
928 		}
929 	}
930 
931 	if (dr)
932 		pfxrtr_add(new, dr);
933 
934 	return 0;
935 }
936 
937 void
938 prelist_remove(pr)
939 	struct nd_prefix *pr;
940 {
941 	struct nd_pfxrouter *pfr, *next;
942 	int e, s;
943 	char ip6buf[INET6_ADDRSTRLEN];
944 
945 	/* make sure to invalidate the prefix until it is really freed. */
946 	pr->ndpr_vltime = 0;
947 	pr->ndpr_pltime = 0;
948 
949 	/*
950 	 * Though these flags are now meaningless, we'd rather keep the value
951 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
952 	 * when executing "ndp -p".
953 	 */
954 
955 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
956 	    (e = nd6_prefix_offlink(pr)) != 0) {
957 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
958 		    "on %s, errno=%d\n",
959 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
960 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
961 		/* what should we do? */
962 	}
963 
964 	if (pr->ndpr_refcnt > 0)
965 		return;		/* notice here? */
966 
967 	s = splnet();
968 
969 	/* unlink ndpr_entry from nd_prefix list */
970 	LIST_REMOVE(pr, ndpr_entry);
971 
972 	/* free list of routers that adversed the prefix */
973 	for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
974 		next = pfr->pfr_next;
975 
976 		free(pfr, M_IP6NDP);
977 	}
978 	splx(s);
979 
980 	free(pr, M_IP6NDP);
981 
982 	pfxlist_onlink_check();
983 }
984 
985 static int
986 prelist_update(new, dr, m, mcast)
987 	struct nd_prefixctl *new;
988 	struct nd_defrouter *dr; /* may be NULL */
989 	struct mbuf *m;
990 	int mcast;
991 {
992 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
993 	struct ifaddr *ifa;
994 	struct ifnet *ifp = new->ndpr_ifp;
995 	struct nd_prefix *pr;
996 	int s = splnet();
997 	int error = 0;
998 	int newprefix = 0;
999 	int auth;
1000 	struct in6_addrlifetime lt6_tmp;
1001 	char ip6buf[INET6_ADDRSTRLEN];
1002 
1003 	auth = 0;
1004 	if (m) {
1005 		/*
1006 		 * Authenticity for NA consists authentication for
1007 		 * both IP header and IP datagrams, doesn't it ?
1008 		 */
1009 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1010 		auth = ((m->m_flags & M_AUTHIPHDR) &&
1011 		    (m->m_flags & M_AUTHIPDGM));
1012 #endif
1013 	}
1014 
1015 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1016 		/*
1017 		 * nd6_prefix_lookup() ensures that pr and new have the same
1018 		 * prefix on a same interface.
1019 		 */
1020 
1021 		/*
1022 		 * Update prefix information.  Note that the on-link (L) bit
1023 		 * and the autonomous (A) bit should NOT be changed from 1
1024 		 * to 0.
1025 		 */
1026 		if (new->ndpr_raf_onlink == 1)
1027 			pr->ndpr_raf_onlink = 1;
1028 		if (new->ndpr_raf_auto == 1)
1029 			pr->ndpr_raf_auto = 1;
1030 		if (new->ndpr_raf_onlink) {
1031 			pr->ndpr_vltime = new->ndpr_vltime;
1032 			pr->ndpr_pltime = new->ndpr_pltime;
1033 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1034 			pr->ndpr_lastupdate = time_second;
1035 		}
1036 
1037 		if (new->ndpr_raf_onlink &&
1038 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1039 			int e;
1040 
1041 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1042 				nd6log((LOG_ERR,
1043 				    "prelist_update: failed to make "
1044 				    "the prefix %s/%d on-link on %s "
1045 				    "(errno=%d)\n",
1046 				    ip6_sprintf(ip6buf,
1047 					    &pr->ndpr_prefix.sin6_addr),
1048 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1049 				/* proceed anyway. XXX: is it correct? */
1050 			}
1051 		}
1052 
1053 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1054 			pfxrtr_add(pr, dr);
1055 	} else {
1056 		struct nd_prefix *newpr = NULL;
1057 
1058 		newprefix = 1;
1059 
1060 		if (new->ndpr_vltime == 0)
1061 			goto end;
1062 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1063 			goto end;
1064 
1065 		error = nd6_prelist_add(new, dr, &newpr);
1066 		if (error != 0 || newpr == NULL) {
1067 			nd6log((LOG_NOTICE, "prelist_update: "
1068 			    "nd6_prelist_add failed for %s/%d on %s "
1069 			    "errno=%d, returnpr=%p\n",
1070 			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1071 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1072 			    error, newpr));
1073 			goto end; /* we should just give up in this case. */
1074 		}
1075 
1076 		/*
1077 		 * XXX: from the ND point of view, we can ignore a prefix
1078 		 * with the on-link bit being zero.  However, we need a
1079 		 * prefix structure for references from autoconfigured
1080 		 * addresses.  Thus, we explicitly make sure that the prefix
1081 		 * itself expires now.
1082 		 */
1083 		if (newpr->ndpr_raf_onlink == 0) {
1084 			newpr->ndpr_vltime = 0;
1085 			newpr->ndpr_pltime = 0;
1086 			in6_init_prefix_ltimes(newpr);
1087 		}
1088 
1089 		pr = newpr;
1090 	}
1091 
1092 	/*
1093 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1094 	 * Note that pr must be non NULL at this point.
1095 	 */
1096 
1097 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1098 	if (!new->ndpr_raf_auto)
1099 		goto end;
1100 
1101 	/*
1102 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1103 	 * nd6_ra_input.
1104 	 */
1105 
1106 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1107 	if (new->ndpr_pltime > new->ndpr_vltime) {
1108 		error = EINVAL;	/* XXX: won't be used */
1109 		goto end;
1110 	}
1111 
1112  	/*
1113 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1114 	 * an address configured by stateless autoconfiguration already in the
1115 	 * list of addresses associated with the interface, and the Valid
1116 	 * Lifetime is not 0, form an address.  We first check if we have
1117 	 * a matching prefix.
1118 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1119 	 * consider autoconfigured addresses while RFC2462 simply said
1120 	 * "address".
1121 	 */
1122 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1123 		struct in6_ifaddr *ifa6;
1124 		u_int32_t remaininglifetime;
1125 
1126 		if (ifa->ifa_addr->sa_family != AF_INET6)
1127 			continue;
1128 
1129 		ifa6 = (struct in6_ifaddr *)ifa;
1130 
1131 		/*
1132 		 * We only consider autoconfigured addresses as per rfc2462bis.
1133 		 */
1134 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1135 			continue;
1136 
1137 		/*
1138 		 * Spec is not clear here, but I believe we should concentrate
1139 		 * on unicast (i.e. not anycast) addresses.
1140 		 * XXX: other ia6_flags? detached or duplicated?
1141 		 */
1142 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1143 			continue;
1144 
1145 		/*
1146 		 * Ignore the address if it is not associated with a prefix
1147 		 * or is associated with a prefix that is different from this
1148 		 * one.  (pr is never NULL here)
1149 		 */
1150 		if (ifa6->ia6_ndpr != pr)
1151 			continue;
1152 
1153 		if (ia6_match == NULL) /* remember the first one */
1154 			ia6_match = ifa6;
1155 
1156 		/*
1157 		 * An already autoconfigured address matched.  Now that we
1158 		 * are sure there is at least one matched address, we can
1159 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1160 		 * "two hours" rule and the privacy extension.
1161 		 * We apply some clarifications in rfc2462bis:
1162 		 * - use remaininglifetime instead of storedlifetime as a
1163 		 *   variable name
1164 		 * - remove the dead code in the "two-hour" rule
1165 		 */
1166 #define TWOHOUR		(120*60)
1167 		lt6_tmp = ifa6->ia6_lifetime;
1168 
1169 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1170 			remaininglifetime = ND6_INFINITE_LIFETIME;
1171 		else if (time_second - ifa6->ia6_updatetime >
1172 			 lt6_tmp.ia6t_vltime) {
1173 			/*
1174 			 * The case of "invalid" address.  We should usually
1175 			 * not see this case.
1176 			 */
1177 			remaininglifetime = 0;
1178 		} else
1179 			remaininglifetime = lt6_tmp.ia6t_vltime -
1180 			    (time_second - ifa6->ia6_updatetime);
1181 
1182 		/* when not updating, keep the current stored lifetime. */
1183 		lt6_tmp.ia6t_vltime = remaininglifetime;
1184 
1185 		if (TWOHOUR < new->ndpr_vltime ||
1186 		    remaininglifetime < new->ndpr_vltime) {
1187 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1188 		} else if (remaininglifetime <= TWOHOUR) {
1189 			if (auth) {
1190 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1191 			}
1192 		} else {
1193 			/*
1194 			 * new->ndpr_vltime <= TWOHOUR &&
1195 			 * TWOHOUR < remaininglifetime
1196 			 */
1197 			lt6_tmp.ia6t_vltime = TWOHOUR;
1198 		}
1199 
1200 		/* The 2 hour rule is not imposed for preferred lifetime. */
1201 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1202 
1203 		in6_init_address_ltimes(pr, &lt6_tmp);
1204 
1205   		/*
1206 		 * We need to treat lifetimes for temporary addresses
1207 		 * differently, according to
1208 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1209 		 * we only update the lifetimes when they are in the maximum
1210 		 * intervals.
1211   		 */
1212   		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1213 			u_int32_t maxvltime, maxpltime;
1214 
1215 			if (ip6_temp_valid_lifetime >
1216 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1217 			    ip6_desync_factor)) {
1218 				maxvltime = ip6_temp_valid_lifetime -
1219 				    (time_second - ifa6->ia6_createtime) -
1220 				    ip6_desync_factor;
1221 			} else
1222 				maxvltime = 0;
1223 			if (ip6_temp_preferred_lifetime >
1224 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
1225 			    ip6_desync_factor)) {
1226 				maxpltime = ip6_temp_preferred_lifetime -
1227 				    (time_second - ifa6->ia6_createtime) -
1228 				    ip6_desync_factor;
1229 			} else
1230 				maxpltime = 0;
1231 
1232 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1233 			    lt6_tmp.ia6t_vltime > maxvltime) {
1234 				lt6_tmp.ia6t_vltime = maxvltime;
1235 			}
1236 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1237 			    lt6_tmp.ia6t_pltime > maxpltime) {
1238 				lt6_tmp.ia6t_pltime = maxpltime;
1239 			}
1240 		}
1241 		ifa6->ia6_lifetime = lt6_tmp;
1242 		ifa6->ia6_updatetime = time_second;
1243 	}
1244 	if (ia6_match == NULL && new->ndpr_vltime) {
1245 		int ifidlen;
1246 
1247 		/*
1248 		 * 5.5.3 (d) (continued)
1249 		 * No address matched and the valid lifetime is non-zero.
1250 		 * Create a new address.
1251 		 */
1252 
1253 		/*
1254 		 * Prefix Length check:
1255 		 * If the sum of the prefix length and interface identifier
1256 		 * length does not equal 128 bits, the Prefix Information
1257 		 * option MUST be ignored.  The length of the interface
1258 		 * identifier is defined in a separate link-type specific
1259 		 * document.
1260 		 */
1261 		ifidlen = in6_if2idlen(ifp);
1262 		if (ifidlen < 0) {
1263 			/* this should not happen, so we always log it. */
1264 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1265 			    if_name(ifp));
1266 			goto end;
1267 		}
1268 		if (ifidlen + pr->ndpr_plen != 128) {
1269 			nd6log((LOG_INFO,
1270 			    "prelist_update: invalid prefixlen "
1271 			    "%d for %s, ignored\n",
1272 			    pr->ndpr_plen, if_name(ifp)));
1273 			goto end;
1274 		}
1275 
1276 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1277 			/*
1278 			 * note that we should use pr (not new) for reference.
1279 			 */
1280 			pr->ndpr_refcnt++;
1281 			ia6->ia6_ndpr = pr;
1282 
1283 			/*
1284 			 * RFC 3041 3.3 (2).
1285 			 * When a new public address is created as described
1286 			 * in RFC2462, also create a new temporary address.
1287 			 *
1288 			 * RFC 3041 3.5.
1289 			 * When an interface connects to a new link, a new
1290 			 * randomized interface identifier should be generated
1291 			 * immediately together with a new set of temporary
1292 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1293 			 * in6_tmpifadd().
1294 			 */
1295 			if (ip6_use_tempaddr) {
1296 				int e;
1297 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1298 					nd6log((LOG_NOTICE, "prelist_update: "
1299 					    "failed to create a temporary "
1300 					    "address, errno=%d\n",
1301 					    e));
1302 				}
1303 			}
1304 
1305 			/*
1306 			 * A newly added address might affect the status
1307 			 * of other addresses, so we check and update it.
1308 			 * XXX: what if address duplication happens?
1309 			 */
1310 			pfxlist_onlink_check();
1311 		} else {
1312 			/* just set an error. do not bark here. */
1313 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1314 		}
1315 	}
1316 
1317  end:
1318 	splx(s);
1319 	return error;
1320 }
1321 
1322 /*
1323  * A supplement function used in the on-link detection below;
1324  * detect if a given prefix has a (probably) reachable advertising router.
1325  * XXX: lengthy function name...
1326  */
1327 static struct nd_pfxrouter *
1328 find_pfxlist_reachable_router(pr)
1329 	struct nd_prefix *pr;
1330 {
1331 	struct nd_pfxrouter *pfxrtr;
1332 	struct rtentry *rt;
1333 	struct llinfo_nd6 *ln;
1334 
1335 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
1336 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
1337 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
1338 		    pfxrtr->router->ifp)) &&
1339 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
1340 		    ND6_IS_LLINFO_PROBREACH(ln))
1341 			break;	/* found */
1342 	}
1343 
1344 	return (pfxrtr);
1345 }
1346 
1347 /*
1348  * Check if each prefix in the prefix list has at least one available router
1349  * that advertised the prefix (a router is "available" if its neighbor cache
1350  * entry is reachable or probably reachable).
1351  * If the check fails, the prefix may be off-link, because, for example,
1352  * we have moved from the network but the lifetime of the prefix has not
1353  * expired yet.  So we should not use the prefix if there is another prefix
1354  * that has an available router.
1355  * But, if there is no prefix that has an available router, we still regards
1356  * all the prefixes as on-link.  This is because we can't tell if all the
1357  * routers are simply dead or if we really moved from the network and there
1358  * is no router around us.
1359  */
1360 void
1361 pfxlist_onlink_check()
1362 {
1363 	struct nd_prefix *pr;
1364 	struct in6_ifaddr *ifa;
1365 	struct nd_defrouter *dr;
1366 	struct nd_pfxrouter *pfxrtr = NULL;
1367 
1368 	/*
1369 	 * Check if there is a prefix that has a reachable advertising
1370 	 * router.
1371 	 */
1372 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1373 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1374 			break;
1375 	}
1376 
1377 	/*
1378 	 * If we have no such prefix, check whether we still have a router
1379 	 * that does not advertise any prefixes.
1380 	 */
1381 	if (pr == NULL) {
1382 		for (dr = TAILQ_FIRST(&nd_defrouter); dr;
1383 		    dr = TAILQ_NEXT(dr, dr_entry)) {
1384 			struct nd_prefix *pr0;
1385 
1386 			for (pr0 = nd_prefix.lh_first; pr0;
1387 			    pr0 = pr0->ndpr_next) {
1388 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1389 					break;
1390 			}
1391 			if (pfxrtr != NULL)
1392 				break;
1393 		}
1394 	}
1395 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && pfxrtr == NULL)) {
1396   		/*
1397 		 * There is at least one prefix that has a reachable router,
1398 		 * or at least a router which probably does not advertise
1399 		 * any prefixes.  The latter would be the case when we move
1400 		 * to a new link where we have a router that does not provide
1401 		 * prefixes and we configure an address by hand.
1402   		 * Detach prefixes which have no reachable advertising
1403   		 * router, and attach other prefixes.
1404   		 */
1405 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1406 			/* XXX: a link-local prefix should never be detached */
1407 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1408 				continue;
1409 
1410 			/*
1411 			 * we aren't interested in prefixes without the L bit
1412 			 * set.
1413 			 */
1414 			if (pr->ndpr_raf_onlink == 0)
1415 				continue;
1416 
1417 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1418 			    find_pfxlist_reachable_router(pr) == NULL)
1419 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1420 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1421 			    find_pfxlist_reachable_router(pr) != 0)
1422 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1423 		}
1424 	} else {
1425 		/* there is no prefix that has a reachable router */
1426 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1427 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1428 				continue;
1429 
1430 			if (pr->ndpr_raf_onlink == 0)
1431 				continue;
1432 
1433 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1434 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1435 		}
1436 	}
1437 
1438 	/*
1439 	 * Remove each interface route associated with a (just) detached
1440 	 * prefix, and reinstall the interface route for a (just) attached
1441 	 * prefix.  Note that all attempt of reinstallation does not
1442 	 * necessarily success, when a same prefix is shared among multiple
1443 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1444 	 * so we don't have to care about them.
1445 	 */
1446 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
1447 		int e;
1448 		char ip6buf[INET6_ADDRSTRLEN];
1449 
1450 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1451 			continue;
1452 
1453 		if (pr->ndpr_raf_onlink == 0)
1454 			continue;
1455 
1456 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1457 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1458 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1459 				nd6log((LOG_ERR,
1460 				    "pfxlist_onlink_check: failed to "
1461 				    "make %s/%d offlink, errno=%d\n",
1462 				    ip6_sprintf(ip6buf,
1463 					    &pr->ndpr_prefix.sin6_addr),
1464 					    pr->ndpr_plen, e));
1465 			}
1466 		}
1467 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1468 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1469 		    pr->ndpr_raf_onlink) {
1470 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1471 				nd6log((LOG_ERR,
1472 				    "pfxlist_onlink_check: failed to "
1473 				    "make %s/%d onlink, errno=%d\n",
1474 				    ip6_sprintf(ip6buf,
1475 					    &pr->ndpr_prefix.sin6_addr),
1476 					    pr->ndpr_plen, e));
1477 			}
1478 		}
1479 	}
1480 
1481 	/*
1482 	 * Changes on the prefix status might affect address status as well.
1483 	 * Make sure that all addresses derived from an attached prefix are
1484 	 * attached, and that all addresses derived from a detached prefix are
1485 	 * detached.  Note, however, that a manually configured address should
1486 	 * always be attached.
1487 	 * The precise detection logic is same as the one for prefixes.
1488 	 */
1489 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1490 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1491 			continue;
1492 
1493 		if (ifa->ia6_ndpr == NULL) {
1494 			/*
1495 			 * This can happen when we first configure the address
1496 			 * (i.e. the address exists, but the prefix does not).
1497 			 * XXX: complicated relationships...
1498 			 */
1499 			continue;
1500 		}
1501 
1502 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1503 			break;
1504 	}
1505 	if (ifa) {
1506 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1507 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1508 				continue;
1509 
1510 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1511 				continue;
1512 
1513 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1514 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1515 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1516 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1517 					nd6_dad_start((struct ifaddr *)ifa, 0);
1518 				}
1519 			} else {
1520 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1521 			}
1522 		}
1523 	}
1524 	else {
1525 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
1526 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1527 				continue;
1528 
1529 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1530 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1531 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1532 				/* Do we need a delay in this case? */
1533 				nd6_dad_start((struct ifaddr *)ifa, 0);
1534 			}
1535 		}
1536 	}
1537 }
1538 
1539 int
1540 nd6_prefix_onlink(pr)
1541 	struct nd_prefix *pr;
1542 {
1543 	struct ifaddr *ifa;
1544 	struct ifnet *ifp = pr->ndpr_ifp;
1545 	struct sockaddr_in6 mask6;
1546 	struct nd_prefix *opr;
1547 	u_long rtflags;
1548 	int error = 0;
1549 	struct rtentry *rt = NULL;
1550 	char ip6buf[INET6_ADDRSTRLEN];
1551 
1552 	/* sanity check */
1553 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1554 		nd6log((LOG_ERR,
1555 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1556 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1557 		    pr->ndpr_plen));
1558 		return (EEXIST);
1559 	}
1560 
1561 	/*
1562 	 * Add the interface route associated with the prefix.  Before
1563 	 * installing the route, check if there's the same prefix on another
1564 	 * interface, and the prefix has already installed the interface route.
1565 	 * Although such a configuration is expected to be rare, we explicitly
1566 	 * allow it.
1567 	 */
1568 	for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1569 		if (opr == pr)
1570 			continue;
1571 
1572 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1573 			continue;
1574 
1575 		if (opr->ndpr_plen == pr->ndpr_plen &&
1576 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1577 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1578 			return (0);
1579 	}
1580 
1581 	/*
1582 	 * We prefer link-local addresses as the associated interface address.
1583 	 */
1584 	/* search for a link-local addr */
1585 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1586 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1587 	if (ifa == NULL) {
1588 		/* XXX: freebsd does not have ifa_ifwithaf */
1589 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1590 			if (ifa->ifa_addr->sa_family == AF_INET6)
1591 				break;
1592 		}
1593 		/* should we care about ia6_flags? */
1594 	}
1595 	if (ifa == NULL) {
1596 		/*
1597 		 * This can still happen, when, for example, we receive an RA
1598 		 * containing a prefix with the L bit set and the A bit clear,
1599 		 * after removing all IPv6 addresses on the receiving
1600 		 * interface.  This should, of course, be rare though.
1601 		 */
1602 		nd6log((LOG_NOTICE,
1603 		    "nd6_prefix_onlink: failed to find any ifaddr"
1604 		    " to add route for a prefix(%s/%d) on %s\n",
1605 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1606 		    pr->ndpr_plen, if_name(ifp)));
1607 		return (0);
1608 	}
1609 
1610 	/*
1611 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1612 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1613 	 */
1614 	bzero(&mask6, sizeof(mask6));
1615 	mask6.sin6_len = sizeof(mask6);
1616 	mask6.sin6_addr = pr->ndpr_mask;
1617 	rtflags = ifa->ifa_flags | RTF_CLONING | RTF_UP;
1618 	if (nd6_need_cache(ifp)) {
1619 		/* explicitly set in case ifa_flags does not set the flag. */
1620 		rtflags |= RTF_CLONING;
1621 	} else {
1622 		/*
1623 		 * explicitly clear the cloning bit in case ifa_flags sets it.
1624 		 */
1625 		rtflags &= ~RTF_CLONING;
1626 	}
1627 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
1628 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
1629 	if (error == 0) {
1630 		if (rt != NULL) /* this should be non NULL, though */
1631 			nd6_rtmsg(RTM_ADD, rt);
1632 		pr->ndpr_stateflags |= NDPRF_ONLINK;
1633 	} else {
1634 		char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
1635 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
1636 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
1637 		    "errno = %d\n",
1638 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1639 		    pr->ndpr_plen, if_name(ifp),
1640 		    ip6_sprintf(ip6bufg, &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
1641 		    ip6_sprintf(ip6bufm, &mask6.sin6_addr), rtflags, error));
1642 	}
1643 
1644 	if (rt != NULL) {
1645 		RT_LOCK(rt);
1646 		RT_REMREF(rt);
1647 		RT_UNLOCK(rt);
1648 	}
1649 
1650 	return (error);
1651 }
1652 
1653 int
1654 nd6_prefix_offlink(pr)
1655 	struct nd_prefix *pr;
1656 {
1657 	int error = 0;
1658 	struct ifnet *ifp = pr->ndpr_ifp;
1659 	struct nd_prefix *opr;
1660 	struct sockaddr_in6 sa6, mask6;
1661 	struct rtentry *rt = NULL;
1662 	char ip6buf[INET6_ADDRSTRLEN];
1663 
1664 	/* sanity check */
1665 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1666 		nd6log((LOG_ERR,
1667 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1668 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1669 		    pr->ndpr_plen));
1670 		return (EEXIST);
1671 	}
1672 
1673 	bzero(&sa6, sizeof(sa6));
1674 	sa6.sin6_family = AF_INET6;
1675 	sa6.sin6_len = sizeof(sa6);
1676 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1677 	    sizeof(struct in6_addr));
1678 	bzero(&mask6, sizeof(mask6));
1679 	mask6.sin6_family = AF_INET6;
1680 	mask6.sin6_len = sizeof(sa6);
1681 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1682 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1683 	    (struct sockaddr *)&mask6, 0, &rt);
1684 	if (error == 0) {
1685 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1686 
1687 		/* report the route deletion to the routing socket. */
1688 		if (rt != NULL)
1689 			nd6_rtmsg(RTM_DELETE, rt);
1690 
1691 		/*
1692 		 * There might be the same prefix on another interface,
1693 		 * the prefix which could not be on-link just because we have
1694 		 * the interface route (see comments in nd6_prefix_onlink).
1695 		 * If there's one, try to make the prefix on-link on the
1696 		 * interface.
1697 		 */
1698 		for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
1699 			if (opr == pr)
1700 				continue;
1701 
1702 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1703 				continue;
1704 
1705 			/*
1706 			 * KAME specific: detached prefixes should not be
1707 			 * on-link.
1708 			 */
1709 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1710 				continue;
1711 
1712 			if (opr->ndpr_plen == pr->ndpr_plen &&
1713 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1714 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1715 				int e;
1716 
1717 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1718 					nd6log((LOG_ERR,
1719 					    "nd6_prefix_offlink: failed to "
1720 					    "recover a prefix %s/%d from %s "
1721 					    "to %s (errno = %d)\n",
1722 					    ip6_sprintf(ip6buf,
1723 						&opr->ndpr_prefix.sin6_addr),
1724 					    opr->ndpr_plen, if_name(ifp),
1725 					    if_name(opr->ndpr_ifp), e));
1726 				}
1727 			}
1728 		}
1729 	} else {
1730 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1731 		nd6log((LOG_ERR,
1732 		    "nd6_prefix_offlink: failed to delete route: "
1733 		    "%s/%d on %s (errno = %d)\n",
1734 		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1735 		    if_name(ifp), error));
1736 	}
1737 
1738 	if (rt != NULL) {
1739 		RTFREE(rt);
1740 	}
1741 
1742 	return (error);
1743 }
1744 
1745 static struct in6_ifaddr *
1746 in6_ifadd(pr, mcast)
1747 	struct nd_prefixctl *pr;
1748 	int mcast;
1749 {
1750 	struct ifnet *ifp = pr->ndpr_ifp;
1751 	struct ifaddr *ifa;
1752 	struct in6_aliasreq ifra;
1753 	struct in6_ifaddr *ia, *ib;
1754 	int error, plen0;
1755 	struct in6_addr mask;
1756 	int prefixlen = pr->ndpr_plen;
1757 	int updateflags;
1758 	char ip6buf[INET6_ADDRSTRLEN];
1759 
1760 	in6_prefixlen2mask(&mask, prefixlen);
1761 
1762 	/*
1763 	 * find a link-local address (will be interface ID).
1764 	 * Is it really mandatory? Theoretically, a global or a site-local
1765 	 * address can be configured without a link-local address, if we
1766 	 * have a unique interface identifier...
1767 	 *
1768 	 * it is not mandatory to have a link-local address, we can generate
1769 	 * interface identifier on the fly.  we do this because:
1770 	 * (1) it should be the easiest way to find interface identifier.
1771 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1772 	 * for multiple addresses on a single interface, and possible shortcut
1773 	 * of DAD.  we omitted DAD for this reason in the past.
1774 	 * (3) a user can prevent autoconfiguration of global address
1775 	 * by removing link-local address by hand (this is partly because we
1776 	 * don't have other way to control the use of IPv6 on an interface.
1777 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1778 	 * (4) it is easier to manage when an interface has addresses
1779 	 * with the same interface identifier, than to have multiple addresses
1780 	 * with different interface identifiers.
1781 	 */
1782 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1783 	if (ifa)
1784 		ib = (struct in6_ifaddr *)ifa;
1785 	else
1786 		return NULL;
1787 
1788 	/* prefixlen + ifidlen must be equal to 128 */
1789 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1790 	if (prefixlen != plen0) {
1791 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1792 		    "(prefix=%d ifid=%d)\n",
1793 		    if_name(ifp), prefixlen, 128 - plen0));
1794 		return NULL;
1795 	}
1796 
1797 	/* make ifaddr */
1798 
1799 	bzero(&ifra, sizeof(ifra));
1800 	/*
1801 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
1802 	 * for safety.
1803 	 */
1804 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1805 	ifra.ifra_addr.sin6_family = AF_INET6;
1806 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
1807 	/* prefix */
1808 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
1809 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
1810 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
1811 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
1812 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
1813 
1814 	/* interface ID */
1815 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1816 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1817 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1818 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1819 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1820 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1821 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1822 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1823 
1824 	/* new prefix mask. */
1825 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1826 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
1827 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
1828 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
1829 
1830 	/* lifetimes. */
1831 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1832 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1833 
1834 	/* XXX: scope zone ID? */
1835 
1836 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1837 
1838   	/*
1839 	 * Make sure that we do not have this address already.  This should
1840 	 * usually not happen, but we can still see this case, e.g., if we
1841 	 * have manually configured the exact address to be configured.
1842 	 */
1843 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
1844 		/* this should be rare enough to make an explicit log */
1845 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1846 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1847 		return (NULL);
1848 	}
1849 
1850 	/*
1851 	 * Allocate ifaddr structure, link into chain, etc.
1852 	 * If we are going to create a new address upon receiving a multicasted
1853 	 * RA, we need to impose a random delay before starting DAD.
1854 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1855 	 */
1856 	updateflags = 0;
1857 	if (mcast)
1858 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1859 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1860 		nd6log((LOG_ERR,
1861 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1862 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1863 		    if_name(ifp), error));
1864 		return (NULL);	/* ifaddr must not have been allocated. */
1865 	}
1866 
1867 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1868 
1869 	return (ia);		/* this is always non-NULL */
1870 }
1871 
1872 int
1873 in6_tmpifadd(ia0, forcegen, delay)
1874 	const struct in6_ifaddr *ia0; /* corresponding public address */
1875 	int forcegen, delay;
1876 {
1877 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1878 	struct in6_ifaddr *newia, *ia;
1879 	struct in6_aliasreq ifra;
1880 	int i, error;
1881 	int trylimit = 3;	/* XXX: adhoc value */
1882 	int updateflags;
1883 	u_int32_t randid[2];
1884 	time_t vltime0, pltime0;
1885 
1886 	bzero(&ifra, sizeof(ifra));
1887 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
1888 	ifra.ifra_addr = ia0->ia_addr;
1889 	/* copy prefix mask */
1890 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
1891 	/* clear the old IFID */
1892 	for (i = 0; i < 4; i++) {
1893 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
1894 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
1895 	}
1896 
1897   again:
1898 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1899 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1900 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1901 		    "random IFID\n"));
1902 		return (EINVAL);
1903 	}
1904 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1905 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1906 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1907 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1908 
1909   	/*
1910 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1911 	 * However, we may still have a chance to see collision, because
1912 	 * there may be a time lag between generation of the ID and generation
1913 	 * of the address.  So, we'll do one more sanity check.
1914 	 */
1915 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1916 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1917 		    &ifra.ifra_addr.sin6_addr)) {
1918   			if (trylimit-- == 0) {
1919 				/*
1920 				 * Give up.  Something strange should have
1921 				 * happened.
1922 				 */
1923 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1924 				    "find a unique random IFID\n"));
1925 				return (EEXIST);
1926 			}
1927 			forcegen = 1;
1928 			goto again;
1929 		}
1930 	}
1931 
1932 	/*
1933 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1934          * public address or TEMP_VALID_LIFETIME.
1935 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1936          * of the public address or TEMP_PREFERRED_LIFETIME -
1937          * DESYNC_FACTOR.
1938 	 */
1939 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1940 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1941 		    (ia0->ia6_lifetime.ia6t_vltime -
1942 		    (time_second - ia0->ia6_updatetime));
1943 		if (vltime0 > ip6_temp_valid_lifetime)
1944 			vltime0 = ip6_temp_valid_lifetime;
1945 	} else
1946 		vltime0 = ip6_temp_valid_lifetime;
1947 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1948 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1949 		    (ia0->ia6_lifetime.ia6t_pltime -
1950 		    (time_second - ia0->ia6_updatetime));
1951 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
1952 			pltime0 = ip6_temp_preferred_lifetime -
1953 			    ip6_desync_factor;
1954 		}
1955 	} else
1956 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
1957 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1958 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1959 
1960 	/*
1961 	 * A temporary address is created only if this calculated Preferred
1962 	 * Lifetime is greater than REGEN_ADVANCE time units.
1963 	 */
1964 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
1965 		return (0);
1966 
1967 	/* XXX: scope zone ID? */
1968 
1969 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1970 
1971 	/* allocate ifaddr structure, link into chain, etc. */
1972 	updateflags = 0;
1973 	if (delay)
1974 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1975 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
1976 		return (error);
1977 
1978 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1979 	if (newia == NULL) {	/* XXX: can it happen? */
1980 		nd6log((LOG_ERR,
1981 		    "in6_tmpifadd: ifa update succeeded, but we got "
1982 		    "no ifaddr\n"));
1983 		return (EINVAL); /* XXX */
1984 	}
1985 	newia->ia6_ndpr = ia0->ia6_ndpr;
1986 	newia->ia6_ndpr->ndpr_refcnt++;
1987 
1988 	/*
1989 	 * A newly added address might affect the status of other addresses.
1990 	 * XXX: when the temporary address is generated with a new public
1991 	 * address, the onlink check is redundant.  However, it would be safe
1992 	 * to do the check explicitly everywhere a new address is generated,
1993 	 * and, in fact, we surely need the check when we create a new
1994 	 * temporary address due to deprecation of an old temporary address.
1995 	 */
1996 	pfxlist_onlink_check();
1997 
1998 	return (0);
1999 }
2000 
2001 static int
2002 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2003 {
2004 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2005 		ndpr->ndpr_preferred = 0;
2006 	else
2007 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
2008 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2009 		ndpr->ndpr_expire = 0;
2010 	else
2011 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
2012 
2013 	return 0;
2014 }
2015 
2016 static void
2017 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2018 {
2019 	/* init ia6t_expire */
2020 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2021 		lt6->ia6t_expire = 0;
2022 	else {
2023 		lt6->ia6t_expire = time_second;
2024 		lt6->ia6t_expire += lt6->ia6t_vltime;
2025 	}
2026 
2027 	/* init ia6t_preferred */
2028 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2029 		lt6->ia6t_preferred = 0;
2030 	else {
2031 		lt6->ia6t_preferred = time_second;
2032 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2033 	}
2034 }
2035 
2036 /*
2037  * Delete all the routing table entries that use the specified gateway.
2038  * XXX: this function causes search through all entries of routing table, so
2039  * it shouldn't be called when acting as a router.
2040  */
2041 void
2042 rt6_flush(gateway, ifp)
2043 	struct in6_addr *gateway;
2044 	struct ifnet *ifp;
2045 {
2046 	struct radix_node_head *rnh = rt_tables[AF_INET6];
2047 	int s = splnet();
2048 
2049 	/* We'll care only link-local addresses */
2050 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
2051 		splx(s);
2052 		return;
2053 	}
2054 
2055 	RADIX_NODE_HEAD_LOCK(rnh);
2056 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2057 	RADIX_NODE_HEAD_UNLOCK(rnh);
2058 	splx(s);
2059 }
2060 
2061 static int
2062 rt6_deleteroute(rn, arg)
2063 	struct radix_node *rn;
2064 	void *arg;
2065 {
2066 #define SIN6(s)	((struct sockaddr_in6 *)s)
2067 	struct rtentry *rt = (struct rtentry *)rn;
2068 	struct in6_addr *gate = (struct in6_addr *)arg;
2069 
2070 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2071 		return (0);
2072 
2073 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2074 		return (0);
2075 	}
2076 
2077 	/*
2078 	 * Do not delete a static route.
2079 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2080 	 * 'cloned' bit instead?
2081 	 */
2082 	if ((rt->rt_flags & RTF_STATIC) != 0)
2083 		return (0);
2084 
2085 	/*
2086 	 * We delete only host route. This means, in particular, we don't
2087 	 * delete default route.
2088 	 */
2089 	if ((rt->rt_flags & RTF_HOST) == 0)
2090 		return (0);
2091 
2092 	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2093 	    rt_mask(rt), rt->rt_flags, 0));
2094 #undef SIN6
2095 }
2096 
2097 int
2098 nd6_setdefaultiface(ifindex)
2099 	int ifindex;
2100 {
2101 	int error = 0;
2102 
2103 	if (ifindex < 0 || if_index < ifindex)
2104 		return (EINVAL);
2105 	if (ifindex != 0 && !ifnet_byindex(ifindex))
2106 		return (EINVAL);
2107 
2108 	if (nd6_defifindex != ifindex) {
2109 		nd6_defifindex = ifindex;
2110 		if (nd6_defifindex > 0)
2111 			nd6_defifp = ifnet_byindex(nd6_defifindex);
2112 		else
2113 			nd6_defifp = NULL;
2114 
2115 		/*
2116 		 * Our current implementation assumes one-to-one maping between
2117 		 * interfaces and links, so it would be natural to use the
2118 		 * default interface as the default link.
2119 		 */
2120 		scope6_setdefault(nd6_defifp);
2121 	}
2122 
2123 	return (error);
2124 }
2125