nd6_rtr.c revision 260860
1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	$KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $
30 */
31
32#include <sys/cdefs.h>
33__FBSDID("$FreeBSD: head/sys/netinet6/nd6_rtr.c 260860 2014-01-18 20:32:59Z melifaro $");
34
35#include "opt_inet.h"
36#include "opt_inet6.h"
37
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/malloc.h>
41#include <sys/mbuf.h>
42#include <sys/socket.h>
43#include <sys/sockio.h>
44#include <sys/time.h>
45#include <sys/kernel.h>
46#include <sys/lock.h>
47#include <sys/errno.h>
48#include <sys/rwlock.h>
49#include <sys/syslog.h>
50#include <sys/queue.h>
51
52#include <net/if.h>
53#include <net/if_var.h>
54#include <net/if_types.h>
55#include <net/if_dl.h>
56#include <net/route.h>
57#include <net/radix.h>
58#include <net/vnet.h>
59
60#include <netinet/in.h>
61#include <net/if_llatbl.h>
62#include <netinet6/in6_var.h>
63#include <netinet6/in6_ifattach.h>
64#include <netinet/ip6.h>
65#include <netinet6/ip6_var.h>
66#include <netinet6/nd6.h>
67#include <netinet/icmp6.h>
68#include <netinet6/scope6_var.h>
69
70static int rtpref(struct nd_defrouter *);
71static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
72static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
73    struct mbuf *, int);
74static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
75static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
76	struct nd_defrouter *);
77static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
78static void pfxrtr_del(struct nd_pfxrouter *);
79static struct nd_pfxrouter *find_pfxlist_reachable_router
80(struct nd_prefix *);
81static void defrouter_delreq(struct nd_defrouter *);
82static void nd6_rtmsg(int, struct rtentry *);
83
84static int in6_init_prefix_ltimes(struct nd_prefix *);
85static void in6_init_address_ltimes(struct nd_prefix *,
86	struct in6_addrlifetime *);
87
88static int nd6_prefix_onlink(struct nd_prefix *);
89static int nd6_prefix_offlink(struct nd_prefix *);
90
91static int rt6_deleteroute(struct radix_node *, void *);
92
93VNET_DECLARE(int, nd6_recalc_reachtm_interval);
94#define	V_nd6_recalc_reachtm_interval	VNET(nd6_recalc_reachtm_interval)
95
96static VNET_DEFINE(struct ifnet *, nd6_defifp);
97VNET_DEFINE(int, nd6_defifindex);
98#define	V_nd6_defifp			VNET(nd6_defifp)
99
100VNET_DEFINE(int, ip6_use_tempaddr) = 0;
101
102VNET_DEFINE(int, ip6_desync_factor);
103VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
104VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
105
106VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
107
108/* RTPREF_MEDIUM has to be 0! */
109#define RTPREF_HIGH	1
110#define RTPREF_MEDIUM	0
111#define RTPREF_LOW	(-1)
112#define RTPREF_RESERVED	(-2)
113#define RTPREF_INVALID	(-3)	/* internal */
114
115/*
116 * Receive Router Solicitation Message - just for routers.
117 * Router solicitation/advertisement is mostly managed by userland program
118 * (rtadvd) so here we have no function like nd6_ra_output().
119 *
120 * Based on RFC 2461
121 */
122void
123nd6_rs_input(struct mbuf *m, int off, int icmp6len)
124{
125	struct ifnet *ifp = m->m_pkthdr.rcvif;
126	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
127	struct nd_router_solicit *nd_rs;
128	struct in6_addr saddr6 = ip6->ip6_src;
129	char *lladdr = NULL;
130	int lladdrlen = 0;
131	union nd_opts ndopts;
132	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
133
134	/*
135	 * Accept RS only when V_ip6_forwarding=1 and the interface has
136	 * no ND6_IFF_ACCEPT_RTADV.
137	 */
138	if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
139		goto freeit;
140
141	/* Sanity checks */
142	if (ip6->ip6_hlim != 255) {
143		nd6log((LOG_ERR,
144		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
145		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
146		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
147		goto bad;
148	}
149
150	/*
151	 * Don't update the neighbor cache, if src = ::.
152	 * This indicates that the src has no IP address assigned yet.
153	 */
154	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
155		goto freeit;
156
157#ifndef PULLDOWN_TEST
158	IP6_EXTHDR_CHECK(m, off, icmp6len,);
159	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
160#else
161	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
162	if (nd_rs == NULL) {
163		ICMP6STAT_INC(icp6s_tooshort);
164		return;
165	}
166#endif
167
168	icmp6len -= sizeof(*nd_rs);
169	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
170	if (nd6_options(&ndopts) < 0) {
171		nd6log((LOG_INFO,
172		    "nd6_rs_input: invalid ND option, ignored\n"));
173		/* nd6_options have incremented stats */
174		goto freeit;
175	}
176
177	if (ndopts.nd_opts_src_lladdr) {
178		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
179		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
180	}
181
182	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
183		nd6log((LOG_INFO,
184		    "nd6_rs_input: lladdrlen mismatch for %s "
185		    "(if %d, RS packet %d)\n",
186		    ip6_sprintf(ip6bufs, &saddr6),
187		    ifp->if_addrlen, lladdrlen - 2));
188		goto bad;
189	}
190
191	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
192
193 freeit:
194	m_freem(m);
195	return;
196
197 bad:
198	ICMP6STAT_INC(icp6s_badrs);
199	m_freem(m);
200}
201
202/*
203 * Receive Router Advertisement Message.
204 *
205 * Based on RFC 2461
206 * TODO: on-link bit on prefix information
207 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
208 */
209void
210nd6_ra_input(struct mbuf *m, int off, int icmp6len)
211{
212	struct ifnet *ifp = m->m_pkthdr.rcvif;
213	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
214	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
215	struct nd_router_advert *nd_ra;
216	struct in6_addr saddr6 = ip6->ip6_src;
217	int mcast = 0;
218	union nd_opts ndopts;
219	struct nd_defrouter *dr;
220	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
221
222	/*
223	 * We only accept RAs only when the per-interface flag
224	 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
225	 */
226	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
227		goto freeit;
228
229	if (ip6->ip6_hlim != 255) {
230		nd6log((LOG_ERR,
231		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
232		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
233		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
234		goto bad;
235	}
236
237	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
238		nd6log((LOG_ERR,
239		    "nd6_ra_input: src %s is not link-local\n",
240		    ip6_sprintf(ip6bufs, &saddr6)));
241		goto bad;
242	}
243
244#ifndef PULLDOWN_TEST
245	IP6_EXTHDR_CHECK(m, off, icmp6len,);
246	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
247#else
248	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
249	if (nd_ra == NULL) {
250		ICMP6STAT_INC(icp6s_tooshort);
251		return;
252	}
253#endif
254
255	icmp6len -= sizeof(*nd_ra);
256	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
257	if (nd6_options(&ndopts) < 0) {
258		nd6log((LOG_INFO,
259		    "nd6_ra_input: invalid ND option, ignored\n"));
260		/* nd6_options have incremented stats */
261		goto freeit;
262	}
263
264    {
265	struct nd_defrouter dr0;
266	u_int32_t advreachable = nd_ra->nd_ra_reachable;
267
268	/* remember if this is a multicasted advertisement */
269	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
270		mcast = 1;
271
272	bzero(&dr0, sizeof(dr0));
273	dr0.rtaddr = saddr6;
274	dr0.flags  = nd_ra->nd_ra_flags_reserved;
275	/*
276	 * Effectively-disable routes from RA messages when
277	 * ND6_IFF_NO_RADR enabled on the receiving interface or
278	 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
279	 */
280	if (ndi->flags & ND6_IFF_NO_RADR)
281		dr0.rtlifetime = 0;
282	else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
283		dr0.rtlifetime = 0;
284	else
285		dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
286	dr0.expire = time_uptime + dr0.rtlifetime;
287	dr0.ifp = ifp;
288	/* unspecified or not? (RFC 2461 6.3.4) */
289	if (advreachable) {
290		advreachable = ntohl(advreachable);
291		if (advreachable <= MAX_REACHABLE_TIME &&
292		    ndi->basereachable != advreachable) {
293			ndi->basereachable = advreachable;
294			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
295			ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
296		}
297	}
298	if (nd_ra->nd_ra_retransmit)
299		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
300	if (nd_ra->nd_ra_curhoplimit)
301		ndi->chlim = nd_ra->nd_ra_curhoplimit;
302	dr = defrtrlist_update(&dr0);
303    }
304
305	/*
306	 * prefix
307	 */
308	if (ndopts.nd_opts_pi) {
309		struct nd_opt_hdr *pt;
310		struct nd_opt_prefix_info *pi = NULL;
311		struct nd_prefixctl pr;
312
313		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
314		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
315		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
316						(pt->nd_opt_len << 3))) {
317			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
318				continue;
319			pi = (struct nd_opt_prefix_info *)pt;
320
321			if (pi->nd_opt_pi_len != 4) {
322				nd6log((LOG_INFO,
323				    "nd6_ra_input: invalid option "
324				    "len %d for prefix information option, "
325				    "ignored\n", pi->nd_opt_pi_len));
326				continue;
327			}
328
329			if (128 < pi->nd_opt_pi_prefix_len) {
330				nd6log((LOG_INFO,
331				    "nd6_ra_input: invalid prefix "
332				    "len %d for prefix information option, "
333				    "ignored\n", pi->nd_opt_pi_prefix_len));
334				continue;
335			}
336
337			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
338			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
339				nd6log((LOG_INFO,
340				    "nd6_ra_input: invalid prefix "
341				    "%s, ignored\n",
342				    ip6_sprintf(ip6bufs,
343					&pi->nd_opt_pi_prefix)));
344				continue;
345			}
346
347			bzero(&pr, sizeof(pr));
348			pr.ndpr_prefix.sin6_family = AF_INET6;
349			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
350			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
351			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
352
353			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
354			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
355			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
356			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
357			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
358			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
359			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
360			(void)prelist_update(&pr, dr, m, mcast);
361		}
362	}
363
364	/*
365	 * MTU
366	 */
367	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
368		u_long mtu;
369		u_long maxmtu;
370
371		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
372
373		/* lower bound */
374		if (mtu < IPV6_MMTU) {
375			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
376			    "mtu=%lu sent from %s, ignoring\n",
377			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
378			goto skip;
379		}
380
381		/* upper bound */
382		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
383		    ? ndi->maxmtu : ifp->if_mtu;
384		if (mtu <= maxmtu) {
385			int change = (ndi->linkmtu != mtu);
386
387			ndi->linkmtu = mtu;
388			if (change) /* in6_maxmtu may change */
389				in6_setmaxmtu();
390		} else {
391			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
392			    "mtu=%lu sent from %s; "
393			    "exceeds maxmtu %lu, ignoring\n",
394			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
395		}
396	}
397
398 skip:
399
400	/*
401	 * Source link layer address
402	 */
403    {
404	char *lladdr = NULL;
405	int lladdrlen = 0;
406
407	if (ndopts.nd_opts_src_lladdr) {
408		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
409		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
410	}
411
412	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
413		nd6log((LOG_INFO,
414		    "nd6_ra_input: lladdrlen mismatch for %s "
415		    "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
416		    ifp->if_addrlen, lladdrlen - 2));
417		goto bad;
418	}
419
420	nd6_cache_lladdr(ifp, &saddr6, lladdr,
421	    lladdrlen, ND_ROUTER_ADVERT, 0);
422
423	/*
424	 * Installing a link-layer address might change the state of the
425	 * router's neighbor cache, which might also affect our on-link
426	 * detection of adveritsed prefixes.
427	 */
428	pfxlist_onlink_check();
429    }
430
431 freeit:
432	m_freem(m);
433	return;
434
435 bad:
436	ICMP6STAT_INC(icp6s_badra);
437	m_freem(m);
438}
439
440/*
441 * default router list proccessing sub routines
442 */
443
444/* tell the change to user processes watching the routing socket. */
445static void
446nd6_rtmsg(int cmd, struct rtentry *rt)
447{
448	struct rt_addrinfo info;
449	struct ifnet *ifp;
450	struct ifaddr *ifa;
451
452	bzero((caddr_t)&info, sizeof(info));
453	info.rti_info[RTAX_DST] = rt_key(rt);
454	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
455	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
456	ifp = rt->rt_ifp;
457	if (ifp != NULL) {
458		IF_ADDR_RLOCK(ifp);
459		ifa = TAILQ_FIRST(&ifp->if_addrhead);
460		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
461		ifa_ref(ifa);
462		IF_ADDR_RUNLOCK(ifp);
463		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
464	} else
465		ifa = NULL;
466
467	rt_missmsg_fib(cmd, &info, rt->rt_flags, 0, rt->rt_fibnum);
468	if (ifa != NULL)
469		ifa_free(ifa);
470}
471
472static void
473defrouter_addreq(struct nd_defrouter *new)
474{
475	struct sockaddr_in6 def, mask, gate;
476	struct rtentry *newrt = NULL;
477	int error;
478
479	bzero(&def, sizeof(def));
480	bzero(&mask, sizeof(mask));
481	bzero(&gate, sizeof(gate));
482
483	def.sin6_len = mask.sin6_len = gate.sin6_len =
484	    sizeof(struct sockaddr_in6);
485	def.sin6_family = gate.sin6_family = AF_INET6;
486	gate.sin6_addr = new->rtaddr;
487
488	error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&def,
489	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
490	    RTF_GATEWAY, &newrt, RT_DEFAULT_FIB);
491	if (newrt) {
492		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
493		RTFREE(newrt);
494	}
495	if (error == 0)
496		new->installed = 1;
497	return;
498}
499
500struct nd_defrouter *
501defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
502{
503	struct nd_defrouter *dr;
504
505	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
506		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
507			return (dr);
508	}
509
510	return (NULL);		/* search failed */
511}
512
513/*
514 * Remove the default route for a given router.
515 * This is just a subroutine function for defrouter_select(), and should
516 * not be called from anywhere else.
517 */
518static void
519defrouter_delreq(struct nd_defrouter *dr)
520{
521	struct sockaddr_in6 def, mask, gate;
522	struct rtentry *oldrt = NULL;
523
524	bzero(&def, sizeof(def));
525	bzero(&mask, sizeof(mask));
526	bzero(&gate, sizeof(gate));
527
528	def.sin6_len = mask.sin6_len = gate.sin6_len =
529	    sizeof(struct sockaddr_in6);
530	def.sin6_family = gate.sin6_family = AF_INET6;
531	gate.sin6_addr = dr->rtaddr;
532
533	in6_rtrequest(RTM_DELETE, (struct sockaddr *)&def,
534	    (struct sockaddr *)&gate,
535	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, RT_DEFAULT_FIB);
536	if (oldrt) {
537		nd6_rtmsg(RTM_DELETE, oldrt);
538		RTFREE(oldrt);
539	}
540
541	dr->installed = 0;
542}
543
544/*
545 * remove all default routes from default router list
546 */
547void
548defrouter_reset(void)
549{
550	struct nd_defrouter *dr;
551
552	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry)
553		defrouter_delreq(dr);
554
555	/*
556	 * XXX should we also nuke any default routers in the kernel, by
557	 * going through them by rtalloc1()?
558	 */
559}
560
561void
562defrtrlist_del(struct nd_defrouter *dr)
563{
564	struct nd_defrouter *deldr = NULL;
565	struct nd_prefix *pr;
566
567	/*
568	 * Flush all the routing table entries that use the router
569	 * as a next hop.
570	 */
571	if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
572		rt6_flush(&dr->rtaddr, dr->ifp);
573
574	if (dr->installed) {
575		deldr = dr;
576		defrouter_delreq(dr);
577	}
578	TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
579
580	/*
581	 * Also delete all the pointers to the router in each prefix lists.
582	 */
583	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
584		struct nd_pfxrouter *pfxrtr;
585		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
586			pfxrtr_del(pfxrtr);
587	}
588	pfxlist_onlink_check();
589
590	/*
591	 * If the router is the primary one, choose a new one.
592	 * Note that defrouter_select() will remove the current gateway
593	 * from the routing table.
594	 */
595	if (deldr)
596		defrouter_select();
597
598	free(dr, M_IP6NDP);
599}
600
601/*
602 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
603 * draft-ietf-ipngwg-router-selection:
604 * 1) Routers that are reachable or probably reachable should be preferred.
605 *    If we have more than one (probably) reachable router, prefer ones
606 *    with the highest router preference.
607 * 2) When no routers on the list are known to be reachable or
608 *    probably reachable, routers SHOULD be selected in a round-robin
609 *    fashion, regardless of router preference values.
610 * 3) If the Default Router List is empty, assume that all
611 *    destinations are on-link.
612 *
613 * We assume nd_defrouter is sorted by router preference value.
614 * Since the code below covers both with and without router preference cases,
615 * we do not need to classify the cases by ifdef.
616 *
617 * At this moment, we do not try to install more than one default router,
618 * even when the multipath routing is available, because we're not sure about
619 * the benefits for stub hosts comparing to the risk of making the code
620 * complicated and the possibility of introducing bugs.
621 */
622void
623defrouter_select(void)
624{
625	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
626	struct llentry *ln = NULL;
627
628	/*
629	 * Let's handle easy case (3) first:
630	 * If default router list is empty, there's nothing to be done.
631	 */
632	if (TAILQ_EMPTY(&V_nd_defrouter))
633		return;
634
635	/*
636	 * Search for a (probably) reachable router from the list.
637	 * We just pick up the first reachable one (if any), assuming that
638	 * the ordering rule of the list described in defrtrlist_update().
639	 */
640	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
641		IF_AFDATA_RLOCK(dr->ifp);
642		if (selected_dr == NULL &&
643		    (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
644		    ND6_IS_LLINFO_PROBREACH(ln)) {
645			selected_dr = dr;
646		}
647		IF_AFDATA_RUNLOCK(dr->ifp);
648		if (ln != NULL) {
649			LLE_RUNLOCK(ln);
650			ln = NULL;
651		}
652
653		if (dr->installed && installed_dr == NULL)
654			installed_dr = dr;
655		else if (dr->installed && installed_dr) {
656			/* this should not happen.  warn for diagnosis. */
657			log(LOG_ERR, "defrouter_select: more than one router"
658			    " is installed\n");
659		}
660	}
661	/*
662	 * If none of the default routers was found to be reachable,
663	 * round-robin the list regardless of preference.
664	 * Otherwise, if we have an installed router, check if the selected
665	 * (reachable) router should really be preferred to the installed one.
666	 * We only prefer the new router when the old one is not reachable
667	 * or when the new one has a really higher preference value.
668	 */
669	if (selected_dr == NULL) {
670		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
671			selected_dr = TAILQ_FIRST(&V_nd_defrouter);
672		else
673			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
674	} else if (installed_dr) {
675		IF_AFDATA_RLOCK(installed_dr->ifp);
676		if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
677		    ND6_IS_LLINFO_PROBREACH(ln) &&
678		    rtpref(selected_dr) <= rtpref(installed_dr)) {
679			selected_dr = installed_dr;
680		}
681		IF_AFDATA_RUNLOCK(installed_dr->ifp);
682		if (ln != NULL)
683			LLE_RUNLOCK(ln);
684	}
685
686	/*
687	 * If the selected router is different than the installed one,
688	 * remove the installed router and install the selected one.
689	 * Note that the selected router is never NULL here.
690	 */
691	if (installed_dr != selected_dr) {
692		if (installed_dr)
693			defrouter_delreq(installed_dr);
694		defrouter_addreq(selected_dr);
695	}
696
697	return;
698}
699
700/*
701 * for default router selection
702 * regards router-preference field as a 2-bit signed integer
703 */
704static int
705rtpref(struct nd_defrouter *dr)
706{
707	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
708	case ND_RA_FLAG_RTPREF_HIGH:
709		return (RTPREF_HIGH);
710	case ND_RA_FLAG_RTPREF_MEDIUM:
711	case ND_RA_FLAG_RTPREF_RSV:
712		return (RTPREF_MEDIUM);
713	case ND_RA_FLAG_RTPREF_LOW:
714		return (RTPREF_LOW);
715	default:
716		/*
717		 * This case should never happen.  If it did, it would mean a
718		 * serious bug of kernel internal.  We thus always bark here.
719		 * Or, can we even panic?
720		 */
721		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
722		return (RTPREF_INVALID);
723	}
724	/* NOTREACHED */
725}
726
727static struct nd_defrouter *
728defrtrlist_update(struct nd_defrouter *new)
729{
730	struct nd_defrouter *dr, *n;
731
732	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
733		/* entry exists */
734		if (new->rtlifetime == 0) {
735			defrtrlist_del(dr);
736			dr = NULL;
737		} else {
738			int oldpref = rtpref(dr);
739
740			/* override */
741			dr->flags = new->flags; /* xxx flag check */
742			dr->rtlifetime = new->rtlifetime;
743			dr->expire = new->expire;
744
745			/*
746			 * If the preference does not change, there's no need
747			 * to sort the entries. Also make sure the selected
748			 * router is still installed in the kernel.
749			 */
750			if (dr->installed && rtpref(new) == oldpref)
751				return (dr);
752
753			/*
754			 * preferred router may be changed, so relocate
755			 * this router.
756			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
757			 * However, since defrtrlist_del() has many side
758			 * effects, we intentionally do so here.
759			 * defrouter_select() below will handle routing
760			 * changes later.
761			 */
762			TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
763			n = dr;
764			goto insert;
765		}
766		return (dr);
767	}
768
769	/* entry does not exist */
770	if (new->rtlifetime == 0)
771		return (NULL);
772
773	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
774	if (n == NULL)
775		return (NULL);
776	bzero(n, sizeof(*n));
777	*n = *new;
778
779insert:
780	/*
781	 * Insert the new router in the Default Router List;
782	 * The Default Router List should be in the descending order
783	 * of router-preferece.  Routers with the same preference are
784	 * sorted in the arriving time order.
785	 */
786
787	/* insert at the end of the group */
788	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
789		if (rtpref(n) > rtpref(dr))
790			break;
791	}
792	if (dr)
793		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
794	else
795		TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
796
797	defrouter_select();
798
799	return (n);
800}
801
802static struct nd_pfxrouter *
803pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
804{
805	struct nd_pfxrouter *search;
806
807	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
808		if (search->router == dr)
809			break;
810	}
811
812	return (search);
813}
814
815static void
816pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
817{
818	struct nd_pfxrouter *new;
819
820	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
821	if (new == NULL)
822		return;
823	bzero(new, sizeof(*new));
824	new->router = dr;
825
826	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
827
828	pfxlist_onlink_check();
829}
830
831static void
832pfxrtr_del(struct nd_pfxrouter *pfr)
833{
834	LIST_REMOVE(pfr, pfr_entry);
835	free(pfr, M_IP6NDP);
836}
837
838struct nd_prefix *
839nd6_prefix_lookup(struct nd_prefixctl *key)
840{
841	struct nd_prefix *search;
842
843	LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
844		if (key->ndpr_ifp == search->ndpr_ifp &&
845		    key->ndpr_plen == search->ndpr_plen &&
846		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
847		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
848			break;
849		}
850	}
851
852	return (search);
853}
854
855int
856nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
857    struct nd_prefix **newp)
858{
859	struct nd_prefix *new = NULL;
860	int error = 0;
861	char ip6buf[INET6_ADDRSTRLEN];
862
863	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
864	if (new == NULL)
865		return(ENOMEM);
866	bzero(new, sizeof(*new));
867	new->ndpr_ifp = pr->ndpr_ifp;
868	new->ndpr_prefix = pr->ndpr_prefix;
869	new->ndpr_plen = pr->ndpr_plen;
870	new->ndpr_vltime = pr->ndpr_vltime;
871	new->ndpr_pltime = pr->ndpr_pltime;
872	new->ndpr_flags = pr->ndpr_flags;
873	if ((error = in6_init_prefix_ltimes(new)) != 0) {
874		free(new, M_IP6NDP);
875		return(error);
876	}
877	new->ndpr_lastupdate = time_uptime;
878	if (newp != NULL)
879		*newp = new;
880
881	/* initialization */
882	LIST_INIT(&new->ndpr_advrtrs);
883	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
884	/* make prefix in the canonical form */
885	IN6_MASK_ADDR(&new->ndpr_prefix.sin6_addr, &new->ndpr_mask);
886
887	/* link ndpr_entry to nd_prefix list */
888	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
889
890	/* ND_OPT_PI_FLAG_ONLINK processing */
891	if (new->ndpr_raf_onlink) {
892		int e;
893
894		if ((e = nd6_prefix_onlink(new)) != 0) {
895			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
896			    "the prefix %s/%d on-link on %s (errno=%d)\n",
897			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
898			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
899			/* proceed anyway. XXX: is it correct? */
900		}
901	}
902
903	if (dr)
904		pfxrtr_add(new, dr);
905
906	return 0;
907}
908
909void
910prelist_remove(struct nd_prefix *pr)
911{
912	struct nd_pfxrouter *pfr, *next;
913	int e;
914	char ip6buf[INET6_ADDRSTRLEN];
915
916	/* make sure to invalidate the prefix until it is really freed. */
917	pr->ndpr_vltime = 0;
918	pr->ndpr_pltime = 0;
919
920	/*
921	 * Though these flags are now meaningless, we'd rather keep the value
922	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
923	 * when executing "ndp -p".
924	 */
925
926	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
927	    (e = nd6_prefix_offlink(pr)) != 0) {
928		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
929		    "on %s, errno=%d\n",
930		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
931		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
932		/* what should we do? */
933	}
934
935	if (pr->ndpr_refcnt > 0)
936		return;		/* notice here? */
937
938	/* unlink ndpr_entry from nd_prefix list */
939	LIST_REMOVE(pr, ndpr_entry);
940
941	/* free list of routers that adversed the prefix */
942	LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next) {
943		free(pfr, M_IP6NDP);
944	}
945	free(pr, M_IP6NDP);
946
947	pfxlist_onlink_check();
948}
949
950/*
951 * dr - may be NULL
952 */
953
954static int
955prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
956    struct mbuf *m, int mcast)
957{
958	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
959	struct ifaddr *ifa;
960	struct ifnet *ifp = new->ndpr_ifp;
961	struct nd_prefix *pr;
962	int error = 0;
963	int newprefix = 0;
964	int auth;
965	struct in6_addrlifetime lt6_tmp;
966	char ip6buf[INET6_ADDRSTRLEN];
967
968	auth = 0;
969	if (m) {
970		/*
971		 * Authenticity for NA consists authentication for
972		 * both IP header and IP datagrams, doesn't it ?
973		 */
974#if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
975		auth = ((m->m_flags & M_AUTHIPHDR) &&
976		    (m->m_flags & M_AUTHIPDGM));
977#endif
978	}
979
980	if ((pr = nd6_prefix_lookup(new)) != NULL) {
981		/*
982		 * nd6_prefix_lookup() ensures that pr and new have the same
983		 * prefix on a same interface.
984		 */
985
986		/*
987		 * Update prefix information.  Note that the on-link (L) bit
988		 * and the autonomous (A) bit should NOT be changed from 1
989		 * to 0.
990		 */
991		if (new->ndpr_raf_onlink == 1)
992			pr->ndpr_raf_onlink = 1;
993		if (new->ndpr_raf_auto == 1)
994			pr->ndpr_raf_auto = 1;
995		if (new->ndpr_raf_onlink) {
996			pr->ndpr_vltime = new->ndpr_vltime;
997			pr->ndpr_pltime = new->ndpr_pltime;
998			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
999			pr->ndpr_lastupdate = time_uptime;
1000		}
1001
1002		if (new->ndpr_raf_onlink &&
1003		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1004			int e;
1005
1006			if ((e = nd6_prefix_onlink(pr)) != 0) {
1007				nd6log((LOG_ERR,
1008				    "prelist_update: failed to make "
1009				    "the prefix %s/%d on-link on %s "
1010				    "(errno=%d)\n",
1011				    ip6_sprintf(ip6buf,
1012					    &pr->ndpr_prefix.sin6_addr),
1013				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1014				/* proceed anyway. XXX: is it correct? */
1015			}
1016		}
1017
1018		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1019			pfxrtr_add(pr, dr);
1020	} else {
1021		struct nd_prefix *newpr = NULL;
1022
1023		newprefix = 1;
1024
1025		if (new->ndpr_vltime == 0)
1026			goto end;
1027		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1028			goto end;
1029
1030		error = nd6_prelist_add(new, dr, &newpr);
1031		if (error != 0 || newpr == NULL) {
1032			nd6log((LOG_NOTICE, "prelist_update: "
1033			    "nd6_prelist_add failed for %s/%d on %s "
1034			    "errno=%d, returnpr=%p\n",
1035			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1036			    new->ndpr_plen, if_name(new->ndpr_ifp),
1037			    error, newpr));
1038			goto end; /* we should just give up in this case. */
1039		}
1040
1041		/*
1042		 * XXX: from the ND point of view, we can ignore a prefix
1043		 * with the on-link bit being zero.  However, we need a
1044		 * prefix structure for references from autoconfigured
1045		 * addresses.  Thus, we explicitly make sure that the prefix
1046		 * itself expires now.
1047		 */
1048		if (newpr->ndpr_raf_onlink == 0) {
1049			newpr->ndpr_vltime = 0;
1050			newpr->ndpr_pltime = 0;
1051			in6_init_prefix_ltimes(newpr);
1052		}
1053
1054		pr = newpr;
1055	}
1056
1057	/*
1058	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1059	 * Note that pr must be non NULL at this point.
1060	 */
1061
1062	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1063	if (!new->ndpr_raf_auto)
1064		goto end;
1065
1066	/*
1067	 * 5.5.3 (b). the link-local prefix should have been ignored in
1068	 * nd6_ra_input.
1069	 */
1070
1071	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1072	if (new->ndpr_pltime > new->ndpr_vltime) {
1073		error = EINVAL;	/* XXX: won't be used */
1074		goto end;
1075	}
1076
1077	/*
1078	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1079	 * an address configured by stateless autoconfiguration already in the
1080	 * list of addresses associated with the interface, and the Valid
1081	 * Lifetime is not 0, form an address.  We first check if we have
1082	 * a matching prefix.
1083	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1084	 * consider autoconfigured addresses while RFC2462 simply said
1085	 * "address".
1086	 */
1087	IF_ADDR_RLOCK(ifp);
1088	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1089		struct in6_ifaddr *ifa6;
1090		u_int32_t remaininglifetime;
1091
1092		if (ifa->ifa_addr->sa_family != AF_INET6)
1093			continue;
1094
1095		ifa6 = (struct in6_ifaddr *)ifa;
1096
1097		/*
1098		 * We only consider autoconfigured addresses as per rfc2462bis.
1099		 */
1100		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1101			continue;
1102
1103		/*
1104		 * Spec is not clear here, but I believe we should concentrate
1105		 * on unicast (i.e. not anycast) addresses.
1106		 * XXX: other ia6_flags? detached or duplicated?
1107		 */
1108		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1109			continue;
1110
1111		/*
1112		 * Ignore the address if it is not associated with a prefix
1113		 * or is associated with a prefix that is different from this
1114		 * one.  (pr is never NULL here)
1115		 */
1116		if (ifa6->ia6_ndpr != pr)
1117			continue;
1118
1119		if (ia6_match == NULL) /* remember the first one */
1120			ia6_match = ifa6;
1121
1122		/*
1123		 * An already autoconfigured address matched.  Now that we
1124		 * are sure there is at least one matched address, we can
1125		 * proceed to 5.5.3. (e): update the lifetimes according to the
1126		 * "two hours" rule and the privacy extension.
1127		 * We apply some clarifications in rfc2462bis:
1128		 * - use remaininglifetime instead of storedlifetime as a
1129		 *   variable name
1130		 * - remove the dead code in the "two-hour" rule
1131		 */
1132#define TWOHOUR		(120*60)
1133		lt6_tmp = ifa6->ia6_lifetime;
1134
1135		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1136			remaininglifetime = ND6_INFINITE_LIFETIME;
1137		else if (time_uptime - ifa6->ia6_updatetime >
1138			 lt6_tmp.ia6t_vltime) {
1139			/*
1140			 * The case of "invalid" address.  We should usually
1141			 * not see this case.
1142			 */
1143			remaininglifetime = 0;
1144		} else
1145			remaininglifetime = lt6_tmp.ia6t_vltime -
1146			    (time_uptime - ifa6->ia6_updatetime);
1147
1148		/* when not updating, keep the current stored lifetime. */
1149		lt6_tmp.ia6t_vltime = remaininglifetime;
1150
1151		if (TWOHOUR < new->ndpr_vltime ||
1152		    remaininglifetime < new->ndpr_vltime) {
1153			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1154		} else if (remaininglifetime <= TWOHOUR) {
1155			if (auth) {
1156				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1157			}
1158		} else {
1159			/*
1160			 * new->ndpr_vltime <= TWOHOUR &&
1161			 * TWOHOUR < remaininglifetime
1162			 */
1163			lt6_tmp.ia6t_vltime = TWOHOUR;
1164		}
1165
1166		/* The 2 hour rule is not imposed for preferred lifetime. */
1167		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1168
1169		in6_init_address_ltimes(pr, &lt6_tmp);
1170
1171		/*
1172		 * We need to treat lifetimes for temporary addresses
1173		 * differently, according to
1174		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1175		 * we only update the lifetimes when they are in the maximum
1176		 * intervals.
1177		 */
1178		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1179			u_int32_t maxvltime, maxpltime;
1180
1181			if (V_ip6_temp_valid_lifetime >
1182			    (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1183			    V_ip6_desync_factor)) {
1184				maxvltime = V_ip6_temp_valid_lifetime -
1185				    (time_uptime - ifa6->ia6_createtime) -
1186				    V_ip6_desync_factor;
1187			} else
1188				maxvltime = 0;
1189			if (V_ip6_temp_preferred_lifetime >
1190			    (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1191			    V_ip6_desync_factor)) {
1192				maxpltime = V_ip6_temp_preferred_lifetime -
1193				    (time_uptime - ifa6->ia6_createtime) -
1194				    V_ip6_desync_factor;
1195			} else
1196				maxpltime = 0;
1197
1198			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1199			    lt6_tmp.ia6t_vltime > maxvltime) {
1200				lt6_tmp.ia6t_vltime = maxvltime;
1201			}
1202			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1203			    lt6_tmp.ia6t_pltime > maxpltime) {
1204				lt6_tmp.ia6t_pltime = maxpltime;
1205			}
1206		}
1207		ifa6->ia6_lifetime = lt6_tmp;
1208		ifa6->ia6_updatetime = time_uptime;
1209	}
1210	IF_ADDR_RUNLOCK(ifp);
1211	if (ia6_match == NULL && new->ndpr_vltime) {
1212		int ifidlen;
1213
1214		/*
1215		 * 5.5.3 (d) (continued)
1216		 * No address matched and the valid lifetime is non-zero.
1217		 * Create a new address.
1218		 */
1219
1220		/*
1221		 * Prefix Length check:
1222		 * If the sum of the prefix length and interface identifier
1223		 * length does not equal 128 bits, the Prefix Information
1224		 * option MUST be ignored.  The length of the interface
1225		 * identifier is defined in a separate link-type specific
1226		 * document.
1227		 */
1228		ifidlen = in6_if2idlen(ifp);
1229		if (ifidlen < 0) {
1230			/* this should not happen, so we always log it. */
1231			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1232			    if_name(ifp));
1233			goto end;
1234		}
1235		if (ifidlen + pr->ndpr_plen != 128) {
1236			nd6log((LOG_INFO,
1237			    "prelist_update: invalid prefixlen "
1238			    "%d for %s, ignored\n",
1239			    pr->ndpr_plen, if_name(ifp)));
1240			goto end;
1241		}
1242
1243		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1244			/*
1245			 * note that we should use pr (not new) for reference.
1246			 */
1247			pr->ndpr_refcnt++;
1248			ia6->ia6_ndpr = pr;
1249
1250			/*
1251			 * RFC 3041 3.3 (2).
1252			 * When a new public address is created as described
1253			 * in RFC2462, also create a new temporary address.
1254			 *
1255			 * RFC 3041 3.5.
1256			 * When an interface connects to a new link, a new
1257			 * randomized interface identifier should be generated
1258			 * immediately together with a new set of temporary
1259			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1260			 * in6_tmpifadd().
1261			 */
1262			if (V_ip6_use_tempaddr) {
1263				int e;
1264				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1265					nd6log((LOG_NOTICE, "prelist_update: "
1266					    "failed to create a temporary "
1267					    "address, errno=%d\n",
1268					    e));
1269				}
1270			}
1271			ifa_free(&ia6->ia_ifa);
1272
1273			/*
1274			 * A newly added address might affect the status
1275			 * of other addresses, so we check and update it.
1276			 * XXX: what if address duplication happens?
1277			 */
1278			pfxlist_onlink_check();
1279		} else {
1280			/* just set an error. do not bark here. */
1281			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1282		}
1283	}
1284
1285 end:
1286	return error;
1287}
1288
1289/*
1290 * A supplement function used in the on-link detection below;
1291 * detect if a given prefix has a (probably) reachable advertising router.
1292 * XXX: lengthy function name...
1293 */
1294static struct nd_pfxrouter *
1295find_pfxlist_reachable_router(struct nd_prefix *pr)
1296{
1297	struct nd_pfxrouter *pfxrtr;
1298	struct llentry *ln;
1299	int canreach;
1300
1301	LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1302		IF_AFDATA_RLOCK(pfxrtr->router->ifp);
1303		ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1304		IF_AFDATA_RUNLOCK(pfxrtr->router->ifp);
1305		if (ln == NULL)
1306			continue;
1307		canreach = ND6_IS_LLINFO_PROBREACH(ln);
1308		LLE_RUNLOCK(ln);
1309		if (canreach)
1310			break;
1311	}
1312	return (pfxrtr);
1313}
1314
1315/*
1316 * Check if each prefix in the prefix list has at least one available router
1317 * that advertised the prefix (a router is "available" if its neighbor cache
1318 * entry is reachable or probably reachable).
1319 * If the check fails, the prefix may be off-link, because, for example,
1320 * we have moved from the network but the lifetime of the prefix has not
1321 * expired yet.  So we should not use the prefix if there is another prefix
1322 * that has an available router.
1323 * But, if there is no prefix that has an available router, we still regards
1324 * all the prefixes as on-link.  This is because we can't tell if all the
1325 * routers are simply dead or if we really moved from the network and there
1326 * is no router around us.
1327 */
1328void
1329pfxlist_onlink_check()
1330{
1331	struct nd_prefix *pr;
1332	struct in6_ifaddr *ifa;
1333	struct nd_defrouter *dr;
1334	struct nd_pfxrouter *pfxrtr = NULL;
1335
1336	/*
1337	 * Check if there is a prefix that has a reachable advertising
1338	 * router.
1339	 */
1340	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1341		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1342			break;
1343	}
1344
1345	/*
1346	 * If we have no such prefix, check whether we still have a router
1347	 * that does not advertise any prefixes.
1348	 */
1349	if (pr == NULL) {
1350		TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1351			struct nd_prefix *pr0;
1352
1353			LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1354				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1355					break;
1356			}
1357			if (pfxrtr != NULL)
1358				break;
1359		}
1360	}
1361	if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) {
1362		/*
1363		 * There is at least one prefix that has a reachable router,
1364		 * or at least a router which probably does not advertise
1365		 * any prefixes.  The latter would be the case when we move
1366		 * to a new link where we have a router that does not provide
1367		 * prefixes and we configure an address by hand.
1368		 * Detach prefixes which have no reachable advertising
1369		 * router, and attach other prefixes.
1370		 */
1371		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1372			/* XXX: a link-local prefix should never be detached */
1373			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1374				continue;
1375
1376			/*
1377			 * we aren't interested in prefixes without the L bit
1378			 * set.
1379			 */
1380			if (pr->ndpr_raf_onlink == 0)
1381				continue;
1382
1383			if (pr->ndpr_raf_auto == 0)
1384				continue;
1385
1386			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1387			    find_pfxlist_reachable_router(pr) == NULL)
1388				pr->ndpr_stateflags |= NDPRF_DETACHED;
1389			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1390			    find_pfxlist_reachable_router(pr) != 0)
1391				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1392		}
1393	} else {
1394		/* there is no prefix that has a reachable router */
1395		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1396			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1397				continue;
1398
1399			if (pr->ndpr_raf_onlink == 0)
1400				continue;
1401
1402			if (pr->ndpr_raf_auto == 0)
1403				continue;
1404
1405			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1406				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1407		}
1408	}
1409
1410	/*
1411	 * Remove each interface route associated with a (just) detached
1412	 * prefix, and reinstall the interface route for a (just) attached
1413	 * prefix.  Note that all attempt of reinstallation does not
1414	 * necessarily success, when a same prefix is shared among multiple
1415	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1416	 * so we don't have to care about them.
1417	 */
1418	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1419		int e;
1420		char ip6buf[INET6_ADDRSTRLEN];
1421
1422		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1423			continue;
1424
1425		if (pr->ndpr_raf_onlink == 0)
1426			continue;
1427
1428		if (pr->ndpr_raf_auto == 0)
1429			continue;
1430
1431		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1432		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1433			if ((e = nd6_prefix_offlink(pr)) != 0) {
1434				nd6log((LOG_ERR,
1435				    "pfxlist_onlink_check: failed to "
1436				    "make %s/%d offlink, errno=%d\n",
1437				    ip6_sprintf(ip6buf,
1438					    &pr->ndpr_prefix.sin6_addr),
1439					    pr->ndpr_plen, e));
1440			}
1441		}
1442		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1443		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1444		    pr->ndpr_raf_onlink) {
1445			if ((e = nd6_prefix_onlink(pr)) != 0) {
1446				nd6log((LOG_ERR,
1447				    "pfxlist_onlink_check: failed to "
1448				    "make %s/%d onlink, errno=%d\n",
1449				    ip6_sprintf(ip6buf,
1450					    &pr->ndpr_prefix.sin6_addr),
1451					    pr->ndpr_plen, e));
1452			}
1453		}
1454	}
1455
1456	/*
1457	 * Changes on the prefix status might affect address status as well.
1458	 * Make sure that all addresses derived from an attached prefix are
1459	 * attached, and that all addresses derived from a detached prefix are
1460	 * detached.  Note, however, that a manually configured address should
1461	 * always be attached.
1462	 * The precise detection logic is same as the one for prefixes.
1463	 *
1464	 * XXXRW: in6_ifaddrhead locking.
1465	 */
1466	TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1467		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1468			continue;
1469
1470		if (ifa->ia6_ndpr == NULL) {
1471			/*
1472			 * This can happen when we first configure the address
1473			 * (i.e. the address exists, but the prefix does not).
1474			 * XXX: complicated relationships...
1475			 */
1476			continue;
1477		}
1478
1479		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1480			break;
1481	}
1482	if (ifa) {
1483		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1484			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1485				continue;
1486
1487			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1488				continue;
1489
1490			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1491				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1492					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1493					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1494					nd6_dad_start((struct ifaddr *)ifa, 0);
1495				}
1496			} else {
1497				ifa->ia6_flags |= IN6_IFF_DETACHED;
1498			}
1499		}
1500	}
1501	else {
1502		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1503			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1504				continue;
1505
1506			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1507				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1508				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1509				/* Do we need a delay in this case? */
1510				nd6_dad_start((struct ifaddr *)ifa, 0);
1511			}
1512		}
1513	}
1514}
1515
1516static int
1517nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
1518{
1519	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1520	struct radix_node_head *rnh;
1521	struct rtentry *rt;
1522	struct sockaddr_in6 mask6;
1523	u_long rtflags;
1524	int error, a_failure, fibnum;
1525
1526	/*
1527	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1528	 * ifa->ifa_rtrequest = nd6_rtrequest;
1529	 */
1530	bzero(&mask6, sizeof(mask6));
1531	mask6.sin6_len = sizeof(mask6);
1532	mask6.sin6_addr = pr->ndpr_mask;
1533	rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1534
1535	a_failure = 0;
1536	for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1537
1538		rt = NULL;
1539		error = in6_rtrequest(RTM_ADD,
1540		    (struct sockaddr *)&pr->ndpr_prefix, ifa->ifa_addr,
1541		    (struct sockaddr *)&mask6, rtflags, &rt, fibnum);
1542		if (error == 0) {
1543			KASSERT(rt != NULL, ("%s: in6_rtrequest return no "
1544			    "error(%d) but rt is NULL, pr=%p, ifa=%p", __func__,
1545			    error, pr, ifa));
1546
1547			rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1548			/* XXX what if rhn == NULL? */
1549			RADIX_NODE_HEAD_LOCK(rnh);
1550			RT_LOCK(rt);
1551			if (rt_setgate(rt, rt_key(rt),
1552			    (struct sockaddr *)&null_sdl) == 0) {
1553				struct sockaddr_dl *dl;
1554
1555				dl = (struct sockaddr_dl *)rt->rt_gateway;
1556				dl->sdl_type = rt->rt_ifp->if_type;
1557				dl->sdl_index = rt->rt_ifp->if_index;
1558			}
1559			RADIX_NODE_HEAD_UNLOCK(rnh);
1560			nd6_rtmsg(RTM_ADD, rt);
1561			RT_UNLOCK(rt);
1562			pr->ndpr_stateflags |= NDPRF_ONLINK;
1563		} else {
1564			char ip6buf[INET6_ADDRSTRLEN];
1565			char ip6bufg[INET6_ADDRSTRLEN];
1566			char ip6bufm[INET6_ADDRSTRLEN];
1567			struct sockaddr_in6 *sin6;
1568
1569			sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1570			nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add "
1571			    "route for a prefix (%s/%d) on %s, gw=%s, mask=%s, "
1572			    "flags=%lx errno = %d\n",
1573			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1574			    pr->ndpr_plen, if_name(pr->ndpr_ifp),
1575			    ip6_sprintf(ip6bufg, &sin6->sin6_addr),
1576			    ip6_sprintf(ip6bufm, &mask6.sin6_addr),
1577			    rtflags, error));
1578
1579			/* Save last error to return, see rtinit(). */
1580			a_failure = error;
1581		}
1582
1583		if (rt != NULL) {
1584			RT_LOCK(rt);
1585			RT_REMREF(rt);
1586			RT_UNLOCK(rt);
1587		}
1588	}
1589
1590	/* Return the last error we got. */
1591	return (a_failure);
1592}
1593
1594static int
1595nd6_prefix_onlink(struct nd_prefix *pr)
1596{
1597	struct ifaddr *ifa;
1598	struct ifnet *ifp = pr->ndpr_ifp;
1599	struct nd_prefix *opr;
1600	int error = 0;
1601	char ip6buf[INET6_ADDRSTRLEN];
1602
1603	/* sanity check */
1604	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1605		nd6log((LOG_ERR,
1606		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1607		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1608		    pr->ndpr_plen));
1609		return (EEXIST);
1610	}
1611
1612	/*
1613	 * Add the interface route associated with the prefix.  Before
1614	 * installing the route, check if there's the same prefix on another
1615	 * interface, and the prefix has already installed the interface route.
1616	 * Although such a configuration is expected to be rare, we explicitly
1617	 * allow it.
1618	 */
1619	LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1620		if (opr == pr)
1621			continue;
1622
1623		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1624			continue;
1625
1626		if (opr->ndpr_plen == pr->ndpr_plen &&
1627		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1628		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1629			return (0);
1630	}
1631
1632	/*
1633	 * We prefer link-local addresses as the associated interface address.
1634	 */
1635	/* search for a link-local addr */
1636	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1637	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1638	if (ifa == NULL) {
1639		/* XXX: freebsd does not have ifa_ifwithaf */
1640		IF_ADDR_RLOCK(ifp);
1641		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1642			if (ifa->ifa_addr->sa_family == AF_INET6)
1643				break;
1644		}
1645		if (ifa != NULL)
1646			ifa_ref(ifa);
1647		IF_ADDR_RUNLOCK(ifp);
1648		/* should we care about ia6_flags? */
1649	}
1650	if (ifa == NULL) {
1651		/*
1652		 * This can still happen, when, for example, we receive an RA
1653		 * containing a prefix with the L bit set and the A bit clear,
1654		 * after removing all IPv6 addresses on the receiving
1655		 * interface.  This should, of course, be rare though.
1656		 */
1657		nd6log((LOG_NOTICE,
1658		    "nd6_prefix_onlink: failed to find any ifaddr"
1659		    " to add route for a prefix(%s/%d) on %s\n",
1660		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1661		    pr->ndpr_plen, if_name(ifp)));
1662		return (0);
1663	}
1664
1665	error = nd6_prefix_onlink_rtrequest(pr, ifa);
1666
1667	if (ifa != NULL)
1668		ifa_free(ifa);
1669
1670	return (error);
1671}
1672
1673static int
1674nd6_prefix_offlink(struct nd_prefix *pr)
1675{
1676	int error = 0;
1677	struct ifnet *ifp = pr->ndpr_ifp;
1678	struct nd_prefix *opr;
1679	struct sockaddr_in6 sa6, mask6;
1680	struct rtentry *rt;
1681	char ip6buf[INET6_ADDRSTRLEN];
1682	int fibnum, a_failure;
1683
1684	/* sanity check */
1685	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1686		nd6log((LOG_ERR,
1687		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1688		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1689		    pr->ndpr_plen));
1690		return (EEXIST);
1691	}
1692
1693	bzero(&sa6, sizeof(sa6));
1694	sa6.sin6_family = AF_INET6;
1695	sa6.sin6_len = sizeof(sa6);
1696	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1697	    sizeof(struct in6_addr));
1698	bzero(&mask6, sizeof(mask6));
1699	mask6.sin6_family = AF_INET6;
1700	mask6.sin6_len = sizeof(sa6);
1701	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1702
1703	a_failure = 0;
1704	for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1705		rt = NULL;
1706		error = in6_rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1707		    (struct sockaddr *)&mask6, 0, &rt, fibnum);
1708		if (error == 0) {
1709			/* report the route deletion to the routing socket. */
1710			if (rt != NULL)
1711				nd6_rtmsg(RTM_DELETE, rt);
1712		} else {
1713			/* Save last error to return, see rtinit(). */
1714			a_failure = error;
1715		}
1716		if (rt != NULL) {
1717			RTFREE(rt);
1718		}
1719	}
1720	error = a_failure;
1721	a_failure = 1;
1722	if (error == 0) {
1723		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1724
1725		/*
1726		 * There might be the same prefix on another interface,
1727		 * the prefix which could not be on-link just because we have
1728		 * the interface route (see comments in nd6_prefix_onlink).
1729		 * If there's one, try to make the prefix on-link on the
1730		 * interface.
1731		 */
1732		LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1733			if (opr == pr)
1734				continue;
1735
1736			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1737				continue;
1738
1739			/*
1740			 * KAME specific: detached prefixes should not be
1741			 * on-link.
1742			 */
1743			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1744				continue;
1745
1746			if (opr->ndpr_plen == pr->ndpr_plen &&
1747			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1748			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1749				int e;
1750
1751				if ((e = nd6_prefix_onlink(opr)) != 0) {
1752					nd6log((LOG_ERR,
1753					    "nd6_prefix_offlink: failed to "
1754					    "recover a prefix %s/%d from %s "
1755					    "to %s (errno = %d)\n",
1756					    ip6_sprintf(ip6buf,
1757						&opr->ndpr_prefix.sin6_addr),
1758					    opr->ndpr_plen, if_name(ifp),
1759					    if_name(opr->ndpr_ifp), e));
1760				} else
1761					a_failure = 0;
1762			}
1763		}
1764	} else {
1765		/* XXX: can we still set the NDPRF_ONLINK flag? */
1766		nd6log((LOG_ERR,
1767		    "nd6_prefix_offlink: failed to delete route: "
1768		    "%s/%d on %s (errno = %d)\n",
1769		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1770		    if_name(ifp), error));
1771	}
1772
1773	if (a_failure)
1774		lltable_prefix_free(AF_INET6, (struct sockaddr *)&sa6,
1775		    (struct sockaddr *)&mask6, LLE_STATIC);
1776
1777	return (error);
1778}
1779
1780static struct in6_ifaddr *
1781in6_ifadd(struct nd_prefixctl *pr, int mcast)
1782{
1783	struct ifnet *ifp = pr->ndpr_ifp;
1784	struct ifaddr *ifa;
1785	struct in6_aliasreq ifra;
1786	struct in6_ifaddr *ia, *ib;
1787	int error, plen0;
1788	struct in6_addr mask;
1789	int prefixlen = pr->ndpr_plen;
1790	int updateflags;
1791	char ip6buf[INET6_ADDRSTRLEN];
1792
1793	in6_prefixlen2mask(&mask, prefixlen);
1794
1795	/*
1796	 * find a link-local address (will be interface ID).
1797	 * Is it really mandatory? Theoretically, a global or a site-local
1798	 * address can be configured without a link-local address, if we
1799	 * have a unique interface identifier...
1800	 *
1801	 * it is not mandatory to have a link-local address, we can generate
1802	 * interface identifier on the fly.  we do this because:
1803	 * (1) it should be the easiest way to find interface identifier.
1804	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1805	 * for multiple addresses on a single interface, and possible shortcut
1806	 * of DAD.  we omitted DAD for this reason in the past.
1807	 * (3) a user can prevent autoconfiguration of global address
1808	 * by removing link-local address by hand (this is partly because we
1809	 * don't have other way to control the use of IPv6 on an interface.
1810	 * this has been our design choice - cf. NRL's "ifconfig auto").
1811	 * (4) it is easier to manage when an interface has addresses
1812	 * with the same interface identifier, than to have multiple addresses
1813	 * with different interface identifiers.
1814	 */
1815	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1816	if (ifa)
1817		ib = (struct in6_ifaddr *)ifa;
1818	else
1819		return NULL;
1820
1821	/* prefixlen + ifidlen must be equal to 128 */
1822	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1823	if (prefixlen != plen0) {
1824		ifa_free(ifa);
1825		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1826		    "(prefix=%d ifid=%d)\n",
1827		    if_name(ifp), prefixlen, 128 - plen0));
1828		return NULL;
1829	}
1830
1831	/* make ifaddr */
1832	in6_prepare_ifra(&ifra, &pr->ndpr_prefix.sin6_addr, &mask);
1833
1834	IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, &mask);
1835	/* interface ID */
1836	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1837	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1838	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1839	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1840	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1841	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1842	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1843	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1844	ifa_free(ifa);
1845
1846	/* lifetimes. */
1847	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1848	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1849
1850	/* XXX: scope zone ID? */
1851
1852	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1853
1854	/*
1855	 * Make sure that we do not have this address already.  This should
1856	 * usually not happen, but we can still see this case, e.g., if we
1857	 * have manually configured the exact address to be configured.
1858	 */
1859	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1860	    &ifra.ifra_addr.sin6_addr);
1861	if (ifa != NULL) {
1862		ifa_free(ifa);
1863		/* this should be rare enough to make an explicit log */
1864		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1865		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1866		return (NULL);
1867	}
1868
1869	/*
1870	 * Allocate ifaddr structure, link into chain, etc.
1871	 * If we are going to create a new address upon receiving a multicasted
1872	 * RA, we need to impose a random delay before starting DAD.
1873	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1874	 */
1875	updateflags = 0;
1876	if (mcast)
1877		updateflags |= IN6_IFAUPDATE_DADDELAY;
1878	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1879		nd6log((LOG_ERR,
1880		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1881		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1882		    if_name(ifp), error));
1883		return (NULL);	/* ifaddr must not have been allocated. */
1884	}
1885
1886	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1887	/*
1888	 * XXXRW: Assumption of non-NULLness here might not be true with
1889	 * fine-grained locking -- should we validate it?  Or just return
1890	 * earlier ifa rather than looking it up again?
1891	 */
1892	return (ia);		/* this is always non-NULL  and referenced. */
1893}
1894
1895/*
1896 * ia0 - corresponding public address
1897 */
1898int
1899in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1900{
1901	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1902	struct in6_ifaddr *newia, *ia;
1903	struct in6_aliasreq ifra;
1904	int error;
1905	int trylimit = 3;	/* XXX: adhoc value */
1906	int updateflags;
1907	u_int32_t randid[2];
1908	time_t vltime0, pltime0;
1909
1910	in6_prepare_ifra(&ifra, &ia0->ia_addr.sin6_addr,
1911	    &ia0->ia_prefixmask.sin6_addr);
1912
1913	ifra.ifra_addr = ia0->ia_addr;	/* XXX: do we need this ? */
1914	/* clear the old IFID */
1915	IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr,
1916	    &ifra.ifra_prefixmask.sin6_addr);
1917
1918  again:
1919	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1920	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1921		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1922		    "random IFID\n"));
1923		return (EINVAL);
1924	}
1925	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1926	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1927	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1928	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1929
1930	/*
1931	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1932	 * However, we may still have a chance to see collision, because
1933	 * there may be a time lag between generation of the ID and generation
1934	 * of the address.  So, we'll do one more sanity check.
1935	 */
1936	IN6_IFADDR_RLOCK();
1937	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1938		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1939		    &ifra.ifra_addr.sin6_addr)) {
1940			if (trylimit-- == 0) {
1941				IN6_IFADDR_RUNLOCK();
1942				/*
1943				 * Give up.  Something strange should have
1944				 * happened.
1945				 */
1946				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1947				    "find a unique random IFID\n"));
1948				return (EEXIST);
1949			}
1950			IN6_IFADDR_RUNLOCK();
1951			forcegen = 1;
1952			goto again;
1953		}
1954	}
1955	IN6_IFADDR_RUNLOCK();
1956
1957	/*
1958	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1959         * public address or TEMP_VALID_LIFETIME.
1960	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1961         * of the public address or TEMP_PREFERRED_LIFETIME -
1962         * DESYNC_FACTOR.
1963	 */
1964	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1965		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1966		    (ia0->ia6_lifetime.ia6t_vltime -
1967		    (time_uptime - ia0->ia6_updatetime));
1968		if (vltime0 > V_ip6_temp_valid_lifetime)
1969			vltime0 = V_ip6_temp_valid_lifetime;
1970	} else
1971		vltime0 = V_ip6_temp_valid_lifetime;
1972	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1973		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1974		    (ia0->ia6_lifetime.ia6t_pltime -
1975		    (time_uptime - ia0->ia6_updatetime));
1976		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
1977			pltime0 = V_ip6_temp_preferred_lifetime -
1978			    V_ip6_desync_factor;
1979		}
1980	} else
1981		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
1982	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1983	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1984
1985	/*
1986	 * A temporary address is created only if this calculated Preferred
1987	 * Lifetime is greater than REGEN_ADVANCE time units.
1988	 */
1989	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
1990		return (0);
1991
1992	/* XXX: scope zone ID? */
1993
1994	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1995
1996	/* allocate ifaddr structure, link into chain, etc. */
1997	updateflags = 0;
1998	if (delay)
1999		updateflags |= IN6_IFAUPDATE_DADDELAY;
2000	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2001		return (error);
2002
2003	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2004	if (newia == NULL) {	/* XXX: can it happen? */
2005		nd6log((LOG_ERR,
2006		    "in6_tmpifadd: ifa update succeeded, but we got "
2007		    "no ifaddr\n"));
2008		return (EINVAL); /* XXX */
2009	}
2010	newia->ia6_ndpr = ia0->ia6_ndpr;
2011	newia->ia6_ndpr->ndpr_refcnt++;
2012	ifa_free(&newia->ia_ifa);
2013
2014	/*
2015	 * A newly added address might affect the status of other addresses.
2016	 * XXX: when the temporary address is generated with a new public
2017	 * address, the onlink check is redundant.  However, it would be safe
2018	 * to do the check explicitly everywhere a new address is generated,
2019	 * and, in fact, we surely need the check when we create a new
2020	 * temporary address due to deprecation of an old temporary address.
2021	 */
2022	pfxlist_onlink_check();
2023
2024	return (0);
2025}
2026
2027static int
2028in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2029{
2030	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2031		ndpr->ndpr_preferred = 0;
2032	else
2033		ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
2034	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2035		ndpr->ndpr_expire = 0;
2036	else
2037		ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
2038
2039	return 0;
2040}
2041
2042static void
2043in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2044{
2045	/* init ia6t_expire */
2046	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2047		lt6->ia6t_expire = 0;
2048	else {
2049		lt6->ia6t_expire = time_uptime;
2050		lt6->ia6t_expire += lt6->ia6t_vltime;
2051	}
2052
2053	/* init ia6t_preferred */
2054	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2055		lt6->ia6t_preferred = 0;
2056	else {
2057		lt6->ia6t_preferred = time_uptime;
2058		lt6->ia6t_preferred += lt6->ia6t_pltime;
2059	}
2060}
2061
2062/*
2063 * Delete all the routing table entries that use the specified gateway.
2064 * XXX: this function causes search through all entries of routing table, so
2065 * it shouldn't be called when acting as a router.
2066 */
2067void
2068rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2069{
2070	struct radix_node_head *rnh;
2071	u_int fibnum;
2072
2073	/* We'll care only link-local addresses */
2074	if (!IN6_IS_ADDR_LINKLOCAL(gateway))
2075		return;
2076
2077	/* XXX Do we really need to walk any but the default FIB? */
2078	for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
2079		rnh = rt_tables_get_rnh(fibnum, AF_INET6);
2080		if (rnh == NULL)
2081			continue;
2082
2083		RADIX_NODE_HEAD_LOCK(rnh);
2084		rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
2085		RADIX_NODE_HEAD_UNLOCK(rnh);
2086	}
2087}
2088
2089static int
2090rt6_deleteroute(struct radix_node *rn, void *arg)
2091{
2092#define SIN6(s)	((struct sockaddr_in6 *)s)
2093	struct rtentry *rt = (struct rtentry *)rn;
2094	struct in6_addr *gate = (struct in6_addr *)arg;
2095
2096	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2097		return (0);
2098
2099	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2100		return (0);
2101	}
2102
2103	/*
2104	 * Do not delete a static route.
2105	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2106	 * 'cloned' bit instead?
2107	 */
2108	if ((rt->rt_flags & RTF_STATIC) != 0)
2109		return (0);
2110
2111	/*
2112	 * We delete only host route. This means, in particular, we don't
2113	 * delete default route.
2114	 */
2115	if ((rt->rt_flags & RTF_HOST) == 0)
2116		return (0);
2117
2118	return (in6_rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2119	    rt_mask(rt), rt->rt_flags, NULL, rt->rt_fibnum));
2120#undef SIN6
2121}
2122
2123int
2124nd6_setdefaultiface(int ifindex)
2125{
2126	int error = 0;
2127
2128	if (ifindex < 0 || V_if_index < ifindex)
2129		return (EINVAL);
2130	if (ifindex != 0 && !ifnet_byindex(ifindex))
2131		return (EINVAL);
2132
2133	if (V_nd6_defifindex != ifindex) {
2134		V_nd6_defifindex = ifindex;
2135		if (V_nd6_defifindex > 0)
2136			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2137		else
2138			V_nd6_defifp = NULL;
2139
2140		/*
2141		 * Our current implementation assumes one-to-one maping between
2142		 * interfaces and links, so it would be natural to use the
2143		 * default interface as the default link.
2144		 */
2145		scope6_setdefault(V_nd6_defifp);
2146	}
2147
2148	return (error);
2149}
2150