1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 *    may be used to endorse or promote products derived from this software
17 *    without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 *
31 *	$KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $
32 */
33
34#include <sys/cdefs.h>
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/refcount.h>
43#include <sys/socket.h>
44#include <sys/sockio.h>
45#include <sys/time.h>
46#include <sys/kernel.h>
47#include <sys/lock.h>
48#include <sys/errno.h>
49#include <sys/rmlock.h>
50#include <sys/rwlock.h>
51#include <sys/sysctl.h>
52#include <sys/syslog.h>
53#include <sys/queue.h>
54
55#include <net/if.h>
56#include <net/if_var.h>
57#include <net/if_private.h>
58#include <net/if_types.h>
59#include <net/if_dl.h>
60#include <net/route.h>
61#include <net/route/nhop.h>
62#include <net/route/route_ctl.h>
63#include <net/radix.h>
64#include <net/vnet.h>
65
66#include <netinet/in.h>
67#include <net/if_llatbl.h>
68#include <netinet6/in6_var.h>
69#include <netinet6/in6_ifattach.h>
70#include <netinet/ip6.h>
71#include <netinet6/ip6_var.h>
72#include <netinet6/nd6.h>
73#include <netinet/icmp6.h>
74#include <netinet6/scope6_var.h>
75
76static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
77static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
78    struct mbuf *, int);
79static int nd6_prefix_onlink(struct nd_prefix *);
80
81TAILQ_HEAD(nd6_drhead, nd_defrouter);
82VNET_DEFINE_STATIC(struct nd6_drhead, nd6_defrouter);
83#define	V_nd6_defrouter			VNET(nd6_defrouter)
84
85VNET_DECLARE(int, nd6_recalc_reachtm_interval);
86#define	V_nd6_recalc_reachtm_interval	VNET(nd6_recalc_reachtm_interval)
87
88VNET_DEFINE_STATIC(struct ifnet *, nd6_defifp);
89VNET_DEFINE(int, nd6_defifindex);
90#define	V_nd6_defifp			VNET(nd6_defifp)
91
92VNET_DEFINE(int, ip6_use_tempaddr) = 0;
93
94VNET_DEFINE(int, ip6_desync_factor);
95VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
96VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
97
98VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
99
100#ifdef EXPERIMENTAL
101VNET_DEFINE_STATIC(int, nd6_ignore_ipv6_only_ra) = 1;
102#define	V_nd6_ignore_ipv6_only_ra	VNET(nd6_ignore_ipv6_only_ra)
103SYSCTL_INT(_net_inet6_icmp6, OID_AUTO,
104    nd6_ignore_ipv6_only_ra, CTLFLAG_VNET | CTLFLAG_RW,
105    &VNET_NAME(nd6_ignore_ipv6_only_ra), 0,
106    "Ignore the 'IPv6-Only flag' in RA messages in compliance with "
107    "draft-ietf-6man-ipv6only-flag");
108#endif
109
110/* RTPREF_MEDIUM has to be 0! */
111#define RTPREF_HIGH	1
112#define RTPREF_MEDIUM	0
113#define RTPREF_LOW	(-1)
114#define RTPREF_RESERVED	(-2)
115#define RTPREF_INVALID	(-3)	/* internal */
116
117static void
118defrouter_ref(struct nd_defrouter *dr)
119{
120
121	refcount_acquire(&dr->refcnt);
122}
123
124void
125defrouter_rele(struct nd_defrouter *dr)
126{
127
128	if (refcount_release(&dr->refcnt))
129		free(dr, M_IP6NDP);
130}
131
132/*
133 * Remove a router from the global list and optionally stash it in a
134 * caller-supplied queue.
135 */
136static void
137defrouter_unlink(struct nd_defrouter *dr, struct nd6_drhead *drq)
138{
139
140	ND6_WLOCK_ASSERT();
141
142	TAILQ_REMOVE(&V_nd6_defrouter, dr, dr_entry);
143	V_nd6_list_genid++;
144	if (drq != NULL)
145		TAILQ_INSERT_TAIL(drq, dr, dr_entry);
146}
147
148/*
149 * Receive Router Solicitation Message - just for routers.
150 * Router solicitation/advertisement is mostly managed by userland program
151 * (rtadvd) so here we have no function like nd6_ra_output().
152 *
153 * Based on RFC 2461
154 */
155void
156nd6_rs_input(struct mbuf *m, int off, int icmp6len)
157{
158	struct ifnet *ifp;
159	struct ip6_hdr *ip6;
160	struct nd_router_solicit *nd_rs;
161	struct in6_addr saddr6;
162	union nd_opts ndopts;
163	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
164	char *lladdr;
165	int lladdrlen;
166
167	ifp = m->m_pkthdr.rcvif;
168
169	/*
170	 * Accept RS only when V_ip6_forwarding=1 and the interface has
171	 * no ND6_IFF_ACCEPT_RTADV.
172	 */
173	if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
174		goto freeit;
175
176	/* RFC 6980: Nodes MUST silently ignore fragments */
177	if(m->m_flags & M_FRAGMENTED)
178		goto freeit;
179
180	/* Sanity checks */
181	ip6 = mtod(m, struct ip6_hdr *);
182	if (__predict_false(ip6->ip6_hlim != 255)) {
183		ICMP6STAT_INC(icp6s_invlhlim);
184		nd6log((LOG_ERR,
185		    "%s: invalid hlim (%d) from %s to %s on %s\n", __func__,
186		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
187		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
188		goto bad;
189	}
190
191	/*
192	 * Don't update the neighbor cache, if src = ::.
193	 * This indicates that the src has no IP address assigned yet.
194	 */
195	saddr6 = ip6->ip6_src;
196	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
197		goto freeit;
198
199	if (m->m_len < off + icmp6len) {
200		m = m_pullup(m, off + icmp6len);
201		if (m == NULL) {
202			IP6STAT_INC(ip6s_exthdrtoolong);
203			return;
204		}
205	}
206	ip6 = mtod(m, struct ip6_hdr *);
207	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
208
209	icmp6len -= sizeof(*nd_rs);
210	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
211	if (nd6_options(&ndopts) < 0) {
212		nd6log((LOG_INFO,
213		    "%s: invalid ND option, ignored\n", __func__));
214		/* nd6_options have incremented stats */
215		goto freeit;
216	}
217
218	lladdr = NULL;
219	lladdrlen = 0;
220	if (ndopts.nd_opts_src_lladdr) {
221		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
222		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
223	}
224
225	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
226		nd6log((LOG_INFO,
227		    "%s: lladdrlen mismatch for %s (if %d, RS packet %d)\n",
228		    __func__, ip6_sprintf(ip6bufs, &saddr6),
229		    ifp->if_addrlen, lladdrlen - 2));
230		goto bad;
231	}
232
233	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
234
235 freeit:
236	m_freem(m);
237	return;
238
239 bad:
240	ICMP6STAT_INC(icp6s_badrs);
241	m_freem(m);
242}
243
244#ifdef EXPERIMENTAL
245/*
246 * An initial update routine for draft-ietf-6man-ipv6only-flag.
247 * We need to iterate over all default routers for the given
248 * interface to see whether they are all advertising the "S"
249 * (IPv6-Only) flag.  If they do set, otherwise unset, the
250 * interface flag we later use to filter on.
251 */
252static void
253defrtr_ipv6_only_ifp(struct ifnet *ifp)
254{
255	struct nd_defrouter *dr;
256	bool ipv6_only, ipv6_only_old;
257#ifdef INET
258	struct epoch_tracker et;
259	struct ifaddr *ifa;
260	bool has_ipv4_addr;
261#endif
262
263	if (V_nd6_ignore_ipv6_only_ra != 0)
264		return;
265
266	ipv6_only = true;
267	ND6_RLOCK();
268	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
269		if (dr->ifp == ifp &&
270		    (dr->raflags & ND_RA_FLAG_IPV6_ONLY) == 0)
271			ipv6_only = false;
272	ND6_RUNLOCK();
273
274	IF_AFDATA_WLOCK(ifp);
275	ipv6_only_old = ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY;
276	IF_AFDATA_WUNLOCK(ifp);
277
278	/* If nothing changed, we have an early exit. */
279	if (ipv6_only == ipv6_only_old)
280		return;
281
282#ifdef INET
283	/*
284	 * Should we want to set the IPV6-ONLY flag, check if the
285	 * interface has a non-0/0 and non-link-local IPv4 address
286	 * configured on it.  If it has we will assume working
287	 * IPv4 operations and will clear the interface flag.
288	 */
289	has_ipv4_addr = false;
290	if (ipv6_only) {
291		NET_EPOCH_ENTER(et);
292		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
293			if (ifa->ifa_addr->sa_family != AF_INET)
294				continue;
295			if (in_canforward(
296			    satosin(ifa->ifa_addr)->sin_addr)) {
297				has_ipv4_addr = true;
298				break;
299			}
300		}
301		NET_EPOCH_EXIT(et);
302	}
303	if (ipv6_only && has_ipv4_addr) {
304		log(LOG_NOTICE, "%s rcvd RA w/ IPv6-Only flag set but has IPv4 "
305		    "configured, ignoring IPv6-Only flag.\n", ifp->if_xname);
306		ipv6_only = false;
307	}
308#endif
309
310	IF_AFDATA_WLOCK(ifp);
311	if (ipv6_only)
312		ND_IFINFO(ifp)->flags |= ND6_IFF_IPV6_ONLY;
313	else
314		ND_IFINFO(ifp)->flags &= ~ND6_IFF_IPV6_ONLY;
315	IF_AFDATA_WUNLOCK(ifp);
316
317#ifdef notyet
318	/* Send notification of flag change. */
319#endif
320}
321
322static void
323defrtr_ipv6_only_ipf_down(struct ifnet *ifp)
324{
325
326	IF_AFDATA_WLOCK(ifp);
327	ND_IFINFO(ifp)->flags &= ~ND6_IFF_IPV6_ONLY;
328	IF_AFDATA_WUNLOCK(ifp);
329}
330#endif	/* EXPERIMENTAL */
331
332void
333nd6_ifnet_link_event(void *arg __unused, struct ifnet *ifp, int linkstate)
334{
335
336	/*
337	 * XXX-BZ we might want to trigger re-evaluation of our default router
338	 * availability. E.g., on link down the default router might be
339	 * unreachable but a different interface might still have connectivity.
340	 */
341
342#ifdef EXPERIMENTAL
343	if (linkstate == LINK_STATE_DOWN)
344		defrtr_ipv6_only_ipf_down(ifp);
345#endif
346}
347
348/*
349 * Receive Router Advertisement Message.
350 *
351 * Based on RFC 2461
352 * TODO: on-link bit on prefix information
353 * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
354 */
355void
356nd6_ra_input(struct mbuf *m, int off, int icmp6len)
357{
358	struct ifnet *ifp;
359	struct nd_ifinfo *ndi;
360	struct ip6_hdr *ip6;
361	struct nd_router_advert *nd_ra;
362	struct in6_addr saddr6;
363	struct nd_defrouter *dr;
364	union nd_opts ndopts;
365	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
366	int mcast;
367
368	/*
369	 * We only accept RAs only when the per-interface flag
370	 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
371	 */
372	ifp = m->m_pkthdr.rcvif;
373	ndi = ND_IFINFO(ifp);
374	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
375		goto freeit;
376
377	/* RFC 6980: Nodes MUST silently ignore fragments */
378	if(m->m_flags & M_FRAGMENTED)
379		goto freeit;
380
381	ip6 = mtod(m, struct ip6_hdr *);
382	if (__predict_false(ip6->ip6_hlim != 255)) {
383		ICMP6STAT_INC(icp6s_invlhlim);
384		nd6log((LOG_ERR,
385		    "%s: invalid hlim (%d) from %s to %s on %s\n", __func__,
386		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
387		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
388		goto bad;
389	}
390
391	saddr6 = ip6->ip6_src;
392	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
393		nd6log((LOG_ERR,
394		    "%s: src %s is not link-local\n", __func__,
395		    ip6_sprintf(ip6bufs, &saddr6)));
396		goto bad;
397	}
398
399	if (m->m_len < off + icmp6len) {
400		m = m_pullup(m, off + icmp6len);
401		if (m == NULL) {
402			IP6STAT_INC(ip6s_exthdrtoolong);
403			return;
404		}
405	}
406	ip6 = mtod(m, struct ip6_hdr *);
407	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
408
409	icmp6len -= sizeof(*nd_ra);
410	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
411	if (nd6_options(&ndopts) < 0) {
412		nd6log((LOG_INFO,
413		    "%s: invalid ND option, ignored\n", __func__));
414		/* nd6_options have incremented stats */
415		goto freeit;
416	}
417
418	mcast = 0;
419	dr = NULL;
420    {
421	struct nd_defrouter dr0;
422	u_int32_t advreachable = nd_ra->nd_ra_reachable;
423
424	/* remember if this is a multicasted advertisement */
425	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
426		mcast = 1;
427
428	bzero(&dr0, sizeof(dr0));
429	dr0.rtaddr = saddr6;
430	dr0.raflags = nd_ra->nd_ra_flags_reserved;
431	/*
432	 * Effectively-disable routes from RA messages when
433	 * ND6_IFF_NO_RADR enabled on the receiving interface or
434	 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
435	 */
436	if (ndi->flags & ND6_IFF_NO_RADR)
437		dr0.rtlifetime = 0;
438	else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
439		dr0.rtlifetime = 0;
440	else
441		dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
442	dr0.expire = time_uptime + dr0.rtlifetime;
443	dr0.ifp = ifp;
444	/* unspecified or not? (RFC 2461 6.3.4) */
445	if (advreachable) {
446		advreachable = ntohl(advreachable);
447		if (advreachable <= MAX_REACHABLE_TIME &&
448		    ndi->basereachable != advreachable) {
449			ndi->basereachable = advreachable;
450			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
451			ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
452		}
453	}
454	if (nd_ra->nd_ra_retransmit)
455		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
456	if (nd_ra->nd_ra_curhoplimit) {
457		if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
458			ndi->chlim = nd_ra->nd_ra_curhoplimit;
459		else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) {
460			log(LOG_ERR, "RA with a lower CurHopLimit sent from "
461			    "%s on %s (current = %d, received = %d). "
462			    "Ignored.\n", ip6_sprintf(ip6bufs, &ip6->ip6_src),
463			    if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
464		}
465	}
466	dr = defrtrlist_update(&dr0);
467#ifdef EXPERIMENTAL
468	defrtr_ipv6_only_ifp(ifp);
469#endif
470    }
471
472	/*
473	 * prefix
474	 */
475	if (ndopts.nd_opts_pi) {
476		struct nd_opt_hdr *pt;
477		struct nd_opt_prefix_info *pi = NULL;
478		struct nd_prefixctl pr;
479
480		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
481		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
482		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
483						(pt->nd_opt_len << 3))) {
484			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
485				continue;
486			pi = (struct nd_opt_prefix_info *)pt;
487
488			if (pi->nd_opt_pi_len != 4) {
489				nd6log((LOG_INFO,
490				    "%s: invalid option len %d for prefix "
491				    "information option, ignored\n", __func__,
492				    pi->nd_opt_pi_len));
493				continue;
494			}
495
496			if (128 < pi->nd_opt_pi_prefix_len) {
497				nd6log((LOG_INFO,
498				    "%s: invalid prefix len %d for prefix "
499				    "information option, ignored\n", __func__,
500				    pi->nd_opt_pi_prefix_len));
501				continue;
502			}
503
504			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
505			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
506				nd6log((LOG_INFO,
507				    "%s: invalid prefix %s, ignored\n",
508				    __func__, ip6_sprintf(ip6bufs,
509					&pi->nd_opt_pi_prefix)));
510				continue;
511			}
512
513			bzero(&pr, sizeof(pr));
514			pr.ndpr_prefix.sin6_family = AF_INET6;
515			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
516			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
517			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
518
519			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
520			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
521			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
522			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
523			pr.ndpr_raf_ra_derived = 1;
524			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
525			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
526			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
527			(void)prelist_update(&pr, dr, m, mcast);
528		}
529	}
530	if (dr != NULL) {
531		defrouter_rele(dr);
532		dr = NULL;
533	}
534
535	/*
536	 * MTU
537	 */
538	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
539		u_long mtu;
540		u_long maxmtu;
541
542		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
543
544		/* lower bound */
545		if (mtu < IPV6_MMTU) {
546			nd6log((LOG_INFO, "%s: bogus mtu option mtu=%lu sent "
547			    "from %s, ignoring\n", __func__,
548			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
549			goto skip;
550		}
551
552		/* upper bound */
553		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
554		    ? ndi->maxmtu : ifp->if_mtu;
555		if (mtu <= maxmtu) {
556			int change = (ndi->linkmtu != mtu);
557
558			ndi->linkmtu = mtu;
559			if (change) {
560				/* in6_maxmtu may change */
561				in6_setmaxmtu();
562				rt_updatemtu(ifp);
563			}
564		} else {
565			nd6log((LOG_INFO, "%s: bogus mtu=%lu sent from %s; "
566			    "exceeds maxmtu %lu, ignoring\n", __func__,
567			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
568		}
569	}
570
571 skip:
572
573	/*
574	 * Source link layer address
575	 */
576    {
577	char *lladdr = NULL;
578	int lladdrlen = 0;
579
580	if (ndopts.nd_opts_src_lladdr) {
581		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
582		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
583	}
584
585	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
586		nd6log((LOG_INFO,
587		    "%s: lladdrlen mismatch for %s (if %d, RA packet %d)\n",
588		    __func__, ip6_sprintf(ip6bufs, &saddr6),
589		    ifp->if_addrlen, lladdrlen - 2));
590		goto bad;
591	}
592
593	nd6_cache_lladdr(ifp, &saddr6, lladdr,
594	    lladdrlen, ND_ROUTER_ADVERT, 0);
595
596	/*
597	 * Installing a link-layer address might change the state of the
598	 * router's neighbor cache, which might also affect our on-link
599	 * detection of adveritsed prefixes.
600	 */
601	pfxlist_onlink_check();
602    }
603
604 freeit:
605	m_freem(m);
606	return;
607
608 bad:
609	ICMP6STAT_INC(icp6s_badra);
610	m_freem(m);
611}
612
613/* PFXRTR */
614static struct nd_pfxrouter *
615pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
616{
617	struct nd_pfxrouter *search;
618
619	ND6_LOCK_ASSERT();
620
621	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
622		if (search->router == dr)
623			break;
624	}
625	return (search);
626}
627
628static void
629pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
630{
631	struct nd_pfxrouter *new;
632	bool update;
633
634	ND6_UNLOCK_ASSERT();
635
636	ND6_RLOCK();
637	if (pfxrtr_lookup(pr, dr) != NULL) {
638		ND6_RUNLOCK();
639		return;
640	}
641	ND6_RUNLOCK();
642
643	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
644	if (new == NULL)
645		return;
646	defrouter_ref(dr);
647	new->router = dr;
648
649	ND6_WLOCK();
650	if (pfxrtr_lookup(pr, dr) == NULL) {
651		LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
652		update = true;
653	} else {
654		/* We lost a race to add the reference. */
655		defrouter_rele(dr);
656		free(new, M_IP6NDP);
657		update = false;
658	}
659	ND6_WUNLOCK();
660
661	if (update)
662		pfxlist_onlink_check();
663}
664
665static void
666pfxrtr_del(struct nd_pfxrouter *pfr)
667{
668
669	ND6_WLOCK_ASSERT();
670
671	LIST_REMOVE(pfr, pfr_entry);
672	defrouter_rele(pfr->router);
673	free(pfr, M_IP6NDP);
674}
675
676/* Default router list processing sub routines. */
677static void
678defrouter_addreq(struct nd_defrouter *new)
679{
680	uint32_t fibnum = new->ifp->if_fib;
681	struct rib_cmd_info rc = {};
682	int error = 0;
683
684	NET_EPOCH_ASSERT();
685
686	struct sockaddr_in6 gw = {
687		.sin6_family = AF_INET6,
688		.sin6_len = sizeof(struct sockaddr_in6),
689		.sin6_addr = new->rtaddr,
690	};
691
692	error = rib_add_default_route(fibnum, AF_INET6, new->ifp,
693	    (struct sockaddr *)&gw, &rc);
694
695	if (error == 0) {
696		struct nhop_object *nh = nhop_select_func(rc.rc_nh_new, 0);
697		rt_routemsg(RTM_ADD, rc.rc_rt, nh, fibnum);
698		new->installed = 1;
699	}
700}
701
702/*
703 * Remove the default route for a given router.
704 * This is just a subroutine function for defrouter_select_fib(), and
705 * should not be called from anywhere else.
706 */
707static void
708defrouter_delreq(struct nd_defrouter *dr)
709{
710	uint32_t fibnum = dr->ifp->if_fib;
711	struct epoch_tracker et;
712	struct rib_cmd_info rc;
713	int error;
714
715	struct sockaddr_in6 dst = {
716		.sin6_family = AF_INET6,
717		.sin6_len = sizeof(struct sockaddr_in6),
718	};
719
720	struct sockaddr_in6 gw = {
721		.sin6_family = AF_INET6,
722		.sin6_len = sizeof(struct sockaddr_in6),
723		.sin6_addr = dr->rtaddr,
724	};
725
726	NET_EPOCH_ENTER(et);
727	error = rib_del_route_px(fibnum, (struct sockaddr *)&dst, 0,
728		    rib_match_gw, (struct sockaddr *)&gw, 0, &rc);
729	if (error == 0) {
730		struct nhop_object *nh = nhop_select_func(rc.rc_nh_old, 0);
731		rt_routemsg(RTM_DELETE, rc.rc_rt, nh, fibnum);
732	}
733	NET_EPOCH_EXIT(et);
734
735	dr->installed = 0;
736}
737
738static void
739defrouter_del(struct nd_defrouter *dr)
740{
741	struct nd_defrouter *deldr = NULL;
742	struct nd_prefix *pr;
743	struct nd_pfxrouter *pfxrtr;
744
745	ND6_UNLOCK_ASSERT();
746
747	/*
748	 * Flush all the routing table entries that use the router
749	 * as a next hop.
750	 */
751	if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
752		rt6_flush(&dr->rtaddr, dr->ifp);
753
754#ifdef EXPERIMENTAL
755	defrtr_ipv6_only_ifp(dr->ifp);
756#endif
757
758	if (dr->installed) {
759		deldr = dr;
760		defrouter_delreq(dr);
761	}
762
763	/*
764	 * Also delete all the pointers to the router in each prefix lists.
765	 */
766	ND6_WLOCK();
767	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
768		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
769			pfxrtr_del(pfxrtr);
770	}
771	ND6_WUNLOCK();
772
773	pfxlist_onlink_check();
774
775	/*
776	 * If the router is the primary one, choose a new one.
777	 * Note that defrouter_select_fib() will remove the current
778         * gateway from the routing table.
779	 */
780	if (deldr)
781		defrouter_select_fib(deldr->ifp->if_fib);
782
783	/*
784	 * Release the list reference.
785	 */
786	defrouter_rele(dr);
787}
788
789struct nd_defrouter *
790defrouter_lookup_locked(const struct in6_addr *addr, struct ifnet *ifp)
791{
792	struct nd_defrouter *dr;
793
794	ND6_LOCK_ASSERT();
795	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
796		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) {
797			defrouter_ref(dr);
798			return (dr);
799		}
800	return (NULL);
801}
802
803struct nd_defrouter *
804defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
805{
806	struct nd_defrouter *dr;
807
808	ND6_RLOCK();
809	dr = defrouter_lookup_locked(addr, ifp);
810	ND6_RUNLOCK();
811	return (dr);
812}
813
814/*
815 * Remove all default routes from default router list.
816 */
817void
818defrouter_reset(void)
819{
820	struct nd_defrouter *dr, **dra;
821	int count, i;
822
823	count = i = 0;
824
825	/*
826	 * We can't delete routes with the ND lock held, so make a copy of the
827	 * current default router list and use that when deleting routes.
828	 */
829	ND6_RLOCK();
830	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
831		count++;
832	ND6_RUNLOCK();
833
834	dra = malloc(count * sizeof(*dra), M_TEMP, M_WAITOK | M_ZERO);
835
836	ND6_RLOCK();
837	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
838		if (i == count)
839			break;
840		defrouter_ref(dr);
841		dra[i++] = dr;
842	}
843	ND6_RUNLOCK();
844
845	for (i = 0; i < count && dra[i] != NULL; i++) {
846		defrouter_delreq(dra[i]);
847		defrouter_rele(dra[i]);
848	}
849	free(dra, M_TEMP);
850
851	/*
852	 * XXX should we also nuke any default routers in the kernel, by
853	 * going through them by rtalloc1()?
854	 */
855}
856
857/*
858 * Look up a matching default router list entry and remove it. Returns true if a
859 * matching entry was found, false otherwise.
860 */
861bool
862defrouter_remove(struct in6_addr *addr, struct ifnet *ifp)
863{
864	struct nd_defrouter *dr;
865
866	ND6_WLOCK();
867	dr = defrouter_lookup_locked(addr, ifp);
868	if (dr == NULL) {
869		ND6_WUNLOCK();
870		return (false);
871	}
872
873	defrouter_unlink(dr, NULL);
874	ND6_WUNLOCK();
875	defrouter_del(dr);
876	defrouter_rele(dr);
877	return (true);
878}
879
880/*
881 * for default router selection
882 * regards router-preference field as a 2-bit signed integer
883 */
884static int
885rtpref(struct nd_defrouter *dr)
886{
887	switch (dr->raflags & ND_RA_FLAG_RTPREF_MASK) {
888	case ND_RA_FLAG_RTPREF_HIGH:
889		return (RTPREF_HIGH);
890	case ND_RA_FLAG_RTPREF_MEDIUM:
891	case ND_RA_FLAG_RTPREF_RSV:
892		return (RTPREF_MEDIUM);
893	case ND_RA_FLAG_RTPREF_LOW:
894		return (RTPREF_LOW);
895	default:
896		/*
897		 * This case should never happen.  If it did, it would mean a
898		 * serious bug of kernel internal.  We thus always bark here.
899		 * Or, can we even panic?
900		 */
901		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->raflags);
902		return (RTPREF_INVALID);
903	}
904	/* NOTREACHED */
905}
906
907static bool
908is_dr_reachable(const struct nd_defrouter *dr) {
909	struct llentry *ln = NULL;
910
911	ln = nd6_lookup(&dr->rtaddr, LLE_SF(AF_INET6, 0), dr->ifp);
912	if (ln == NULL)
913		return (false);
914	bool reachable = ND6_IS_LLINFO_PROBREACH(ln);
915	LLE_RUNLOCK(ln);
916	return reachable;
917}
918
919/*
920 * Default Router Selection according to Section 6.3.6 of RFC 2461 and
921 * draft-ietf-ipngwg-router-selection:
922 * 1) Routers that are reachable or probably reachable should be preferred.
923 *    If we have more than one (probably) reachable router, prefer ones
924 *    with the highest router preference.
925 * 2) When no routers on the list are known to be reachable or
926 *    probably reachable, routers SHOULD be selected in a round-robin
927 *    fashion, regardless of router preference values.
928 * 3) If the Default Router List is empty, assume that all
929 *    destinations are on-link.
930 *
931 * We assume nd_defrouter is sorted by router preference value.
932 * Since the code below covers both with and without router preference cases,
933 * we do not need to classify the cases by ifdef.
934 *
935 * At this moment, we do not try to install more than one default router,
936 * even when the multipath routing is available, because we're not sure about
937 * the benefits for stub hosts comparing to the risk of making the code
938 * complicated and the possibility of introducing bugs.
939 *
940 * We maintain a single list of routers for multiple FIBs, only considering one
941 * at a time based on the receiving interface's FIB. If @fibnum is RT_ALL_FIBS,
942 * we do the whole thing multiple times.
943 */
944void
945defrouter_select_fib(int fibnum)
946{
947	struct epoch_tracker et;
948	struct nd_defrouter *dr, *selected_dr, *installed_dr;
949
950	if (fibnum == RT_ALL_FIBS) {
951		for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
952			defrouter_select_fib(fibnum);
953		}
954		return;
955	}
956
957	ND6_RLOCK();
958	/*
959	 * Let's handle easy case (3) first:
960	 * If default router list is empty, there's nothing to be done.
961	 */
962	if (TAILQ_EMPTY(&V_nd6_defrouter)) {
963		ND6_RUNLOCK();
964		return;
965	}
966
967	/*
968	 * Search for a (probably) reachable router from the list.
969	 * We just pick up the first reachable one (if any), assuming that
970	 * the ordering rule of the list described in defrtrlist_update().
971	 */
972	selected_dr = installed_dr = NULL;
973	NET_EPOCH_ENTER(et);
974	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
975		if (dr->ifp->if_fib != fibnum)
976			continue;
977
978		if (selected_dr == NULL && is_dr_reachable(dr)) {
979			selected_dr = dr;
980			defrouter_ref(selected_dr);
981		}
982
983		if (dr->installed) {
984			if (installed_dr == NULL) {
985				installed_dr = dr;
986				defrouter_ref(installed_dr);
987			} else {
988				/*
989				 * this should not happen.
990				 * warn for diagnosis.
991				 */
992				log(LOG_ERR, "defrouter_select_fib: more than "
993				             "one router is installed\n");
994			}
995		}
996	}
997
998	/*
999	 * If none of the default routers was found to be reachable,
1000	 * round-robin the list regardless of preference.
1001	 * Otherwise, if we have an installed router, check if the selected
1002	 * (reachable) router should really be preferred to the installed one.
1003	 * We only prefer the new router when the old one is not reachable
1004	 * or when the new one has a really higher preference value.
1005	 */
1006	if (selected_dr == NULL) {
1007		if (installed_dr == NULL ||
1008		    TAILQ_NEXT(installed_dr, dr_entry) == NULL)
1009			dr = TAILQ_FIRST(&V_nd6_defrouter);
1010		else
1011			dr = TAILQ_NEXT(installed_dr, dr_entry);
1012
1013		/* Ensure we select a router for this FIB. */
1014		TAILQ_FOREACH_FROM(dr, &V_nd6_defrouter, dr_entry) {
1015			if (dr->ifp->if_fib == fibnum) {
1016				selected_dr = dr;
1017				defrouter_ref(selected_dr);
1018				break;
1019			}
1020		}
1021	} else if (installed_dr != NULL) {
1022		if (is_dr_reachable(installed_dr) &&
1023		    rtpref(selected_dr) <= rtpref(installed_dr)) {
1024			defrouter_rele(selected_dr);
1025			selected_dr = installed_dr;
1026		}
1027	}
1028	ND6_RUNLOCK();
1029
1030	/*
1031	 * If we selected a router for this FIB and it's different
1032	 * than the installed one, remove the installed router and
1033	 * install the selected one in its place.
1034	 */
1035	if (installed_dr != selected_dr) {
1036		if (installed_dr != NULL) {
1037			defrouter_delreq(installed_dr);
1038			defrouter_rele(installed_dr);
1039		}
1040		if (selected_dr != NULL)
1041			defrouter_addreq(selected_dr);
1042	}
1043	if (selected_dr != NULL)
1044		defrouter_rele(selected_dr);
1045	NET_EPOCH_EXIT(et);
1046}
1047
1048static struct nd_defrouter *
1049defrtrlist_update(struct nd_defrouter *new)
1050{
1051	struct nd_defrouter *dr, *n;
1052	uint64_t genid;
1053	int oldpref;
1054	bool writelocked;
1055
1056	if (new->rtlifetime == 0) {
1057		defrouter_remove(&new->rtaddr, new->ifp);
1058		return (NULL);
1059	}
1060
1061	ND6_RLOCK();
1062	writelocked = false;
1063restart:
1064	dr = defrouter_lookup_locked(&new->rtaddr, new->ifp);
1065	if (dr != NULL) {
1066		oldpref = rtpref(dr);
1067
1068		/* override */
1069		dr->raflags = new->raflags; /* XXX flag check */
1070		dr->rtlifetime = new->rtlifetime;
1071		dr->expire = new->expire;
1072
1073		/*
1074		 * If the preference does not change, there's no need
1075		 * to sort the entries. Also make sure the selected
1076		 * router is still installed in the kernel.
1077		 */
1078		if (dr->installed && rtpref(new) == oldpref) {
1079			if (writelocked)
1080				ND6_WUNLOCK();
1081			else
1082				ND6_RUNLOCK();
1083			return (dr);
1084		}
1085	}
1086
1087	/*
1088	 * The router needs to be reinserted into the default router
1089	 * list, so upgrade to a write lock. If that fails and the list
1090	 * has potentially changed while the lock was dropped, we'll
1091	 * redo the lookup with the write lock held.
1092	 */
1093	if (!writelocked) {
1094		writelocked = true;
1095		if (!ND6_TRY_UPGRADE()) {
1096			genid = V_nd6_list_genid;
1097			ND6_RUNLOCK();
1098			ND6_WLOCK();
1099			if (genid != V_nd6_list_genid)
1100				goto restart;
1101		}
1102	}
1103
1104	if (dr != NULL) {
1105		/*
1106		 * The preferred router may have changed, so relocate this
1107		 * router.
1108		 */
1109		TAILQ_REMOVE(&V_nd6_defrouter, dr, dr_entry);
1110		n = dr;
1111	} else {
1112		n = malloc(sizeof(*n), M_IP6NDP, M_NOWAIT | M_ZERO);
1113		if (n == NULL) {
1114			ND6_WUNLOCK();
1115			return (NULL);
1116		}
1117		memcpy(n, new, sizeof(*n));
1118		/* Initialize with an extra reference for the caller. */
1119		refcount_init(&n->refcnt, 2);
1120	}
1121
1122	/*
1123	 * Insert the new router in the Default Router List;
1124	 * The Default Router List should be in the descending order
1125	 * of router-preferece.  Routers with the same preference are
1126	 * sorted in the arriving time order.
1127	 */
1128
1129	/* insert at the end of the group */
1130	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
1131		if (rtpref(n) > rtpref(dr))
1132			break;
1133	}
1134	if (dr != NULL)
1135		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
1136	else
1137		TAILQ_INSERT_TAIL(&V_nd6_defrouter, n, dr_entry);
1138	V_nd6_list_genid++;
1139	ND6_WUNLOCK();
1140
1141	defrouter_select_fib(new->ifp->if_fib);
1142
1143	return (n);
1144}
1145
1146static int
1147in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1148{
1149	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1150		ndpr->ndpr_preferred = 0;
1151	else
1152		ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
1153	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1154		ndpr->ndpr_expire = 0;
1155	else
1156		ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
1157
1158	return 0;
1159}
1160
1161static void
1162in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
1163{
1164	/* init ia6t_expire */
1165	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
1166		lt6->ia6t_expire = 0;
1167	else {
1168		lt6->ia6t_expire = time_uptime;
1169		lt6->ia6t_expire += lt6->ia6t_vltime;
1170	}
1171
1172	/* init ia6t_preferred */
1173	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
1174		lt6->ia6t_preferred = 0;
1175	else {
1176		lt6->ia6t_preferred = time_uptime;
1177		lt6->ia6t_preferred += lt6->ia6t_pltime;
1178	}
1179}
1180
1181static struct in6_ifaddr *
1182in6_ifadd(struct nd_prefixctl *pr, int mcast)
1183{
1184	struct ifnet *ifp = pr->ndpr_ifp;
1185	struct ifaddr *ifa;
1186	struct in6_aliasreq ifra;
1187	struct in6_ifaddr *ia, *ib;
1188	int error, plen0;
1189	struct in6_addr mask;
1190	int prefixlen = pr->ndpr_plen;
1191	int updateflags;
1192	char ip6buf[INET6_ADDRSTRLEN];
1193
1194	in6_prefixlen2mask(&mask, prefixlen);
1195
1196	/*
1197	 * find a link-local address (will be interface ID).
1198	 * Is it really mandatory? Theoretically, a global or a site-local
1199	 * address can be configured without a link-local address, if we
1200	 * have a unique interface identifier...
1201	 *
1202	 * it is not mandatory to have a link-local address, we can generate
1203	 * interface identifier on the fly.  we do this because:
1204	 * (1) it should be the easiest way to find interface identifier.
1205	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1206	 * for multiple addresses on a single interface, and possible shortcut
1207	 * of DAD.  we omitted DAD for this reason in the past.
1208	 * (3) a user can prevent autoconfiguration of global address
1209	 * by removing link-local address by hand (this is partly because we
1210	 * don't have other way to control the use of IPv6 on an interface.
1211	 * this has been our design choice - cf. NRL's "ifconfig auto").
1212	 * (4) it is easier to manage when an interface has addresses
1213	 * with the same interface identifier, than to have multiple addresses
1214	 * with different interface identifiers.
1215	 */
1216	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1217	if (ifa)
1218		ib = (struct in6_ifaddr *)ifa;
1219	else
1220		return NULL;
1221
1222	/* prefixlen + ifidlen must be equal to 128 */
1223	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1224	if (prefixlen != plen0) {
1225		ifa_free(ifa);
1226		nd6log((LOG_INFO,
1227		    "%s: wrong prefixlen for %s (prefix=%d ifid=%d)\n",
1228		    __func__, if_name(ifp), prefixlen, 128 - plen0));
1229		return NULL;
1230	}
1231
1232	/* make ifaddr */
1233	in6_prepare_ifra(&ifra, &pr->ndpr_prefix.sin6_addr, &mask);
1234
1235	IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, &mask);
1236	/* interface ID */
1237	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1238	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1239	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1240	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1241	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1242	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1243	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1244	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1245	ifa_free(ifa);
1246
1247	/* lifetimes. */
1248	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1249	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1250
1251	/* XXX: scope zone ID? */
1252
1253	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1254
1255	/*
1256	 * Make sure that we do not have this address already.  This should
1257	 * usually not happen, but we can still see this case, e.g., if we
1258	 * have manually configured the exact address to be configured.
1259	 */
1260	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1261	    &ifra.ifra_addr.sin6_addr);
1262	if (ifa != NULL) {
1263		ifa_free(ifa);
1264		/* this should be rare enough to make an explicit log */
1265		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1266		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1267		return (NULL);
1268	}
1269
1270	/*
1271	 * Allocate ifaddr structure, link into chain, etc.
1272	 * If we are going to create a new address upon receiving a multicasted
1273	 * RA, we need to impose a random delay before starting DAD.
1274	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1275	 */
1276	updateflags = 0;
1277	if (mcast)
1278		updateflags |= IN6_IFAUPDATE_DADDELAY;
1279	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1280		nd6log((LOG_ERR,
1281		    "%s: failed to make ifaddr %s on %s (errno=%d)\n", __func__,
1282		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1283		    if_name(ifp), error));
1284		return (NULL);	/* ifaddr must not have been allocated. */
1285	}
1286
1287	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1288	/*
1289	 * XXXRW: Assumption of non-NULLness here might not be true with
1290	 * fine-grained locking -- should we validate it?  Or just return
1291	 * earlier ifa rather than looking it up again?
1292	 */
1293	return (ia);		/* this is always non-NULL  and referenced. */
1294}
1295
1296static struct nd_prefix *
1297nd6_prefix_lookup_locked(struct nd_prefixctl *key)
1298{
1299	struct nd_prefix *search;
1300
1301	ND6_LOCK_ASSERT();
1302
1303	LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
1304		if (key->ndpr_ifp == search->ndpr_ifp &&
1305		    key->ndpr_plen == search->ndpr_plen &&
1306		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
1307		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
1308			nd6_prefix_ref(search);
1309			break;
1310		}
1311	}
1312	return (search);
1313}
1314
1315struct nd_prefix *
1316nd6_prefix_lookup(struct nd_prefixctl *key)
1317{
1318	struct nd_prefix *search;
1319
1320	ND6_RLOCK();
1321	search = nd6_prefix_lookup_locked(key);
1322	ND6_RUNLOCK();
1323	return (search);
1324}
1325
1326void
1327nd6_prefix_ref(struct nd_prefix *pr)
1328{
1329
1330	refcount_acquire(&pr->ndpr_refcnt);
1331}
1332
1333void
1334nd6_prefix_rele(struct nd_prefix *pr)
1335{
1336
1337	if (refcount_release(&pr->ndpr_refcnt)) {
1338		KASSERT(LIST_EMPTY(&pr->ndpr_advrtrs),
1339		    ("prefix %p has advertising routers", pr));
1340		free(pr, M_IP6NDP);
1341	}
1342}
1343
1344int
1345nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
1346    struct nd_prefix **newp)
1347{
1348	struct nd_prefix *new;
1349	char ip6buf[INET6_ADDRSTRLEN];
1350	int error;
1351
1352	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
1353	if (new == NULL)
1354		return (ENOMEM);
1355	refcount_init(&new->ndpr_refcnt, newp != NULL ? 2 : 1);
1356	new->ndpr_ifp = pr->ndpr_ifp;
1357	new->ndpr_prefix = pr->ndpr_prefix;
1358	new->ndpr_plen = pr->ndpr_plen;
1359	new->ndpr_vltime = pr->ndpr_vltime;
1360	new->ndpr_pltime = pr->ndpr_pltime;
1361	new->ndpr_flags = pr->ndpr_flags;
1362	if ((error = in6_init_prefix_ltimes(new)) != 0) {
1363		free(new, M_IP6NDP);
1364		return (error);
1365	}
1366	new->ndpr_lastupdate = time_uptime;
1367
1368	/* initialization */
1369	LIST_INIT(&new->ndpr_advrtrs);
1370	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
1371	/* make prefix in the canonical form */
1372	IN6_MASK_ADDR(&new->ndpr_prefix.sin6_addr, &new->ndpr_mask);
1373
1374	ND6_WLOCK();
1375	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
1376	V_nd6_list_genid++;
1377	ND6_WUNLOCK();
1378
1379	/* ND_OPT_PI_FLAG_ONLINK processing */
1380	if (new->ndpr_raf_onlink) {
1381		struct epoch_tracker et;
1382
1383		ND6_ONLINK_LOCK();
1384		NET_EPOCH_ENTER(et);
1385		if ((error = nd6_prefix_onlink(new)) != 0) {
1386			nd6log((LOG_ERR, "%s: failed to make the prefix %s/%d "
1387			    "on-link on %s (errno=%d)\n", __func__,
1388			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1389			    pr->ndpr_plen, if_name(pr->ndpr_ifp), error));
1390			/* proceed anyway. XXX: is it correct? */
1391		}
1392		NET_EPOCH_EXIT(et);
1393		ND6_ONLINK_UNLOCK();
1394	}
1395
1396	if (dr != NULL)
1397		pfxrtr_add(new, dr);
1398	if (newp != NULL)
1399		*newp = new;
1400	return (0);
1401}
1402
1403/*
1404 * Remove a prefix from the prefix list and optionally stash it in a
1405 * caller-provided list.
1406 *
1407 * The ND6 lock must be held.
1408 */
1409void
1410nd6_prefix_unlink(struct nd_prefix *pr, struct nd_prhead *list)
1411{
1412
1413	ND6_WLOCK_ASSERT();
1414
1415	LIST_REMOVE(pr, ndpr_entry);
1416	V_nd6_list_genid++;
1417	if (list != NULL)
1418		LIST_INSERT_HEAD(list, pr, ndpr_entry);
1419}
1420
1421/*
1422 * Free an unlinked prefix, first marking it off-link if necessary.
1423 */
1424void
1425nd6_prefix_del(struct nd_prefix *pr)
1426{
1427	struct nd_pfxrouter *pfr, *next;
1428	int e;
1429	char ip6buf[INET6_ADDRSTRLEN];
1430
1431	KASSERT(pr->ndpr_addrcnt == 0,
1432	    ("prefix %p has referencing addresses", pr));
1433	ND6_UNLOCK_ASSERT();
1434
1435	/*
1436	 * Though these flags are now meaningless, we'd rather keep the value
1437	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
1438	 * when executing "ndp -p".
1439	 */
1440	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1441		ND6_ONLINK_LOCK();
1442		if ((e = nd6_prefix_offlink(pr)) != 0) {
1443			nd6log((LOG_ERR,
1444			    "%s: failed to make the prefix %s/%d offlink on %s "
1445			    "(errno=%d)\n", __func__,
1446			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1447			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1448			/* what should we do? */
1449		}
1450		ND6_ONLINK_UNLOCK();
1451	}
1452
1453	/* Release references to routers that have advertised this prefix. */
1454	ND6_WLOCK();
1455	LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next)
1456		pfxrtr_del(pfr);
1457	ND6_WUNLOCK();
1458
1459	nd6_prefix_rele(pr);
1460
1461	pfxlist_onlink_check();
1462}
1463
1464static int
1465prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
1466    struct mbuf *m, int mcast)
1467{
1468	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
1469	struct ifaddr *ifa;
1470	struct ifnet *ifp = new->ndpr_ifp;
1471	struct nd_prefix *pr;
1472	int error = 0;
1473	int auth;
1474	struct in6_addrlifetime lt6_tmp;
1475	char ip6buf[INET6_ADDRSTRLEN];
1476
1477	NET_EPOCH_ASSERT();
1478
1479	auth = 0;
1480	if (m) {
1481		/*
1482		 * Authenticity for NA consists authentication for
1483		 * both IP header and IP datagrams, doesn't it ?
1484		 */
1485#if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1486		auth = ((m->m_flags & M_AUTHIPHDR) &&
1487		    (m->m_flags & M_AUTHIPDGM));
1488#endif
1489	}
1490
1491	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1492		/*
1493		 * nd6_prefix_lookup() ensures that pr and new have the same
1494		 * prefix on a same interface.
1495		 */
1496
1497		/*
1498		 * Update prefix information.  Note that the on-link (L) bit
1499		 * and the autonomous (A) bit should NOT be changed from 1
1500		 * to 0.
1501		 */
1502		if (new->ndpr_raf_onlink == 1)
1503			pr->ndpr_raf_onlink = 1;
1504		if (new->ndpr_raf_auto == 1)
1505			pr->ndpr_raf_auto = 1;
1506		if (new->ndpr_raf_onlink) {
1507			pr->ndpr_vltime = new->ndpr_vltime;
1508			pr->ndpr_pltime = new->ndpr_pltime;
1509			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1510			pr->ndpr_lastupdate = time_uptime;
1511		}
1512
1513		if (new->ndpr_raf_onlink &&
1514		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1515			ND6_ONLINK_LOCK();
1516			if ((error = nd6_prefix_onlink(pr)) != 0) {
1517				nd6log((LOG_ERR,
1518				    "%s: failed to make the prefix %s/%d "
1519				    "on-link on %s (errno=%d)\n", __func__,
1520				    ip6_sprintf(ip6buf,
1521				        &pr->ndpr_prefix.sin6_addr),
1522				    pr->ndpr_plen, if_name(pr->ndpr_ifp),
1523				    error));
1524				/* proceed anyway. XXX: is it correct? */
1525			}
1526			ND6_ONLINK_UNLOCK();
1527		}
1528
1529		if (dr != NULL)
1530			pfxrtr_add(pr, dr);
1531	} else {
1532		if (new->ndpr_vltime == 0)
1533			goto end;
1534		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1535			goto end;
1536
1537		error = nd6_prelist_add(new, dr, &pr);
1538		if (error != 0) {
1539			nd6log((LOG_NOTICE, "%s: nd6_prelist_add() failed for "
1540			    "the prefix %s/%d on %s (errno=%d)\n", __func__,
1541			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1542			    new->ndpr_plen, if_name(new->ndpr_ifp), error));
1543			goto end; /* we should just give up in this case. */
1544		}
1545
1546		/*
1547		 * XXX: from the ND point of view, we can ignore a prefix
1548		 * with the on-link bit being zero.  However, we need a
1549		 * prefix structure for references from autoconfigured
1550		 * addresses.  Thus, we explicitly make sure that the prefix
1551		 * itself expires now.
1552		 */
1553		if (pr->ndpr_raf_onlink == 0) {
1554			pr->ndpr_vltime = 0;
1555			pr->ndpr_pltime = 0;
1556			in6_init_prefix_ltimes(pr);
1557		}
1558	}
1559
1560	/*
1561	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1562	 * Note that pr must be non NULL at this point.
1563	 */
1564
1565	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1566	if (!new->ndpr_raf_auto)
1567		goto end;
1568
1569	/*
1570	 * 5.5.3 (b). the link-local prefix should have been ignored in
1571	 * nd6_ra_input.
1572	 */
1573
1574	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1575	if (new->ndpr_pltime > new->ndpr_vltime) {
1576		error = EINVAL;	/* XXX: won't be used */
1577		goto end;
1578	}
1579
1580	/*
1581	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1582	 * an address configured by stateless autoconfiguration already in the
1583	 * list of addresses associated with the interface, and the Valid
1584	 * Lifetime is not 0, form an address.  We first check if we have
1585	 * a matching prefix.
1586	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1587	 * consider autoconfigured addresses while RFC2462 simply said
1588	 * "address".
1589	 */
1590	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1591		struct in6_ifaddr *ifa6;
1592		u_int32_t remaininglifetime;
1593
1594		if (ifa->ifa_addr->sa_family != AF_INET6)
1595			continue;
1596
1597		ifa6 = (struct in6_ifaddr *)ifa;
1598
1599		/*
1600		 * We only consider autoconfigured addresses as per rfc2462bis.
1601		 */
1602		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1603			continue;
1604
1605		/*
1606		 * Spec is not clear here, but I believe we should concentrate
1607		 * on unicast (i.e. not anycast) addresses.
1608		 * XXX: other ia6_flags? detached or duplicated?
1609		 */
1610		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1611			continue;
1612
1613		/*
1614		 * Ignore the address if it is not associated with a prefix
1615		 * or is associated with a prefix that is different from this
1616		 * one.  (pr is never NULL here)
1617		 */
1618		if (ifa6->ia6_ndpr != pr)
1619			continue;
1620
1621		if (ia6_match == NULL) /* remember the first one */
1622			ia6_match = ifa6;
1623
1624		/*
1625		 * An already autoconfigured address matched.  Now that we
1626		 * are sure there is at least one matched address, we can
1627		 * proceed to 5.5.3. (e): update the lifetimes according to the
1628		 * "two hours" rule and the privacy extension.
1629		 * We apply some clarifications in rfc2462bis:
1630		 * - use remaininglifetime instead of storedlifetime as a
1631		 *   variable name
1632		 * - remove the dead code in the "two-hour" rule
1633		 */
1634#define TWOHOUR		(120*60)
1635		lt6_tmp = ifa6->ia6_lifetime;
1636
1637		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1638			remaininglifetime = ND6_INFINITE_LIFETIME;
1639		else if (time_uptime - ifa6->ia6_updatetime >
1640			 lt6_tmp.ia6t_vltime) {
1641			/*
1642			 * The case of "invalid" address.  We should usually
1643			 * not see this case.
1644			 */
1645			remaininglifetime = 0;
1646		} else
1647			remaininglifetime = lt6_tmp.ia6t_vltime -
1648			    (time_uptime - ifa6->ia6_updatetime);
1649
1650		/* when not updating, keep the current stored lifetime. */
1651		lt6_tmp.ia6t_vltime = remaininglifetime;
1652
1653		if (TWOHOUR < new->ndpr_vltime ||
1654		    remaininglifetime < new->ndpr_vltime) {
1655			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1656		} else if (remaininglifetime <= TWOHOUR) {
1657			if (auth) {
1658				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1659			}
1660		} else {
1661			/*
1662			 * new->ndpr_vltime <= TWOHOUR &&
1663			 * TWOHOUR < remaininglifetime
1664			 */
1665			lt6_tmp.ia6t_vltime = TWOHOUR;
1666		}
1667
1668		/* The 2 hour rule is not imposed for preferred lifetime. */
1669		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1670
1671		in6_init_address_ltimes(pr, &lt6_tmp);
1672
1673		/*
1674		 * We need to treat lifetimes for temporary addresses
1675		 * differently, according to
1676		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1677		 * we only update the lifetimes when they are in the maximum
1678		 * intervals.
1679		 */
1680		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1681			u_int32_t maxvltime, maxpltime;
1682
1683			if (V_ip6_temp_valid_lifetime >
1684			    (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1685			    V_ip6_desync_factor)) {
1686				maxvltime = V_ip6_temp_valid_lifetime -
1687				    (time_uptime - ifa6->ia6_createtime) -
1688				    V_ip6_desync_factor;
1689			} else
1690				maxvltime = 0;
1691			if (V_ip6_temp_preferred_lifetime >
1692			    (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1693			    V_ip6_desync_factor)) {
1694				maxpltime = V_ip6_temp_preferred_lifetime -
1695				    (time_uptime - ifa6->ia6_createtime) -
1696				    V_ip6_desync_factor;
1697			} else
1698				maxpltime = 0;
1699
1700			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1701			    lt6_tmp.ia6t_vltime > maxvltime) {
1702				lt6_tmp.ia6t_vltime = maxvltime;
1703			}
1704			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1705			    lt6_tmp.ia6t_pltime > maxpltime) {
1706				lt6_tmp.ia6t_pltime = maxpltime;
1707			}
1708		}
1709		ifa6->ia6_lifetime = lt6_tmp;
1710		ifa6->ia6_updatetime = time_uptime;
1711	}
1712	if (ia6_match == NULL && new->ndpr_vltime) {
1713		int ifidlen;
1714
1715		/*
1716		 * 5.5.3 (d) (continued)
1717		 * No address matched and the valid lifetime is non-zero.
1718		 * Create a new address.
1719		 */
1720
1721		/*
1722		 * Prefix Length check:
1723		 * If the sum of the prefix length and interface identifier
1724		 * length does not equal 128 bits, the Prefix Information
1725		 * option MUST be ignored.  The length of the interface
1726		 * identifier is defined in a separate link-type specific
1727		 * document.
1728		 */
1729		ifidlen = in6_if2idlen(ifp);
1730		if (ifidlen < 0) {
1731			/* this should not happen, so we always log it. */
1732			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1733			    if_name(ifp));
1734			goto end;
1735		}
1736		if (ifidlen + pr->ndpr_plen != 128) {
1737			nd6log((LOG_INFO,
1738			    "%s: invalid prefixlen %d for %s, ignored\n",
1739			    __func__, pr->ndpr_plen, if_name(ifp)));
1740			goto end;
1741		}
1742
1743		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1744			/*
1745			 * note that we should use pr (not new) for reference.
1746			 */
1747			pr->ndpr_addrcnt++;
1748			ia6->ia6_ndpr = pr;
1749
1750			/*
1751			 * RFC 3041 3.3 (2).
1752			 * When a new public address is created as described
1753			 * in RFC2462, also create a new temporary address.
1754			 *
1755			 * RFC 3041 3.5.
1756			 * When an interface connects to a new link, a new
1757			 * randomized interface identifier should be generated
1758			 * immediately together with a new set of temporary
1759			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1760			 * in6_tmpifadd().
1761			 */
1762			if (V_ip6_use_tempaddr) {
1763				int e;
1764				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1765					nd6log((LOG_NOTICE, "%s: failed to "
1766					    "create a temporary address "
1767					    "(errno=%d)\n", __func__, e));
1768				}
1769			}
1770			ifa_free(&ia6->ia_ifa);
1771
1772			/*
1773			 * A newly added address might affect the status
1774			 * of other addresses, so we check and update it.
1775			 * XXX: what if address duplication happens?
1776			 */
1777			pfxlist_onlink_check();
1778		} else {
1779			/* just set an error. do not bark here. */
1780			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1781		}
1782	}
1783
1784end:
1785	if (pr != NULL)
1786		nd6_prefix_rele(pr);
1787	return (error);
1788}
1789
1790/*
1791 * A supplement function used in the on-link detection below;
1792 * detect if a given prefix has a (probably) reachable advertising router.
1793 * XXX: lengthy function name...
1794 */
1795static struct nd_pfxrouter *
1796find_pfxlist_reachable_router(struct nd_prefix *pr)
1797{
1798	struct epoch_tracker et;
1799	struct nd_pfxrouter *pfxrtr;
1800
1801	ND6_LOCK_ASSERT();
1802
1803	NET_EPOCH_ENTER(et);
1804	LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1805		if (is_dr_reachable(pfxrtr->router))
1806			break;
1807	}
1808	NET_EPOCH_EXIT(et);
1809	return (pfxrtr);
1810}
1811
1812/*
1813 * Check if each prefix in the prefix list has at least one available router
1814 * that advertised the prefix (a router is "available" if its neighbor cache
1815 * entry is reachable or probably reachable).
1816 * If the check fails, the prefix may be off-link, because, for example,
1817 * we have moved from the network but the lifetime of the prefix has not
1818 * expired yet.  So we should not use the prefix if there is another prefix
1819 * that has an available router.
1820 * But, if there is no prefix that has an available router, we still regard
1821 * all the prefixes as on-link.  This is because we can't tell if all the
1822 * routers are simply dead or if we really moved from the network and there
1823 * is no router around us.
1824 */
1825void
1826pfxlist_onlink_check(void)
1827{
1828	struct nd_prefix *pr;
1829	struct in6_ifaddr *ifa;
1830	struct nd_defrouter *dr;
1831	struct nd_pfxrouter *pfxrtr = NULL;
1832	struct rm_priotracker in6_ifa_tracker;
1833	uint64_t genid;
1834	uint32_t flags;
1835
1836	ND6_ONLINK_LOCK();
1837	ND6_RLOCK();
1838
1839	/*
1840	 * Check if there is a prefix that has a reachable advertising
1841	 * router.
1842	 */
1843	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1844		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1845			break;
1846	}
1847
1848	/*
1849	 * If we have no such prefix, check whether we still have a router
1850	 * that does not advertise any prefixes.
1851	 */
1852	if (pr == NULL) {
1853		TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
1854			struct nd_prefix *pr0;
1855
1856			LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1857				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1858					break;
1859			}
1860			if (pfxrtr != NULL)
1861				break;
1862		}
1863	}
1864	if (pr != NULL || (!TAILQ_EMPTY(&V_nd6_defrouter) && pfxrtr == NULL)) {
1865		/*
1866		 * There is at least one prefix that has a reachable router,
1867		 * or at least a router which probably does not advertise
1868		 * any prefixes.  The latter would be the case when we move
1869		 * to a new link where we have a router that does not provide
1870		 * prefixes and we configure an address by hand.
1871		 * Detach prefixes which have no reachable advertising
1872		 * router, and attach other prefixes.
1873		 */
1874		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1875			/* XXX: a link-local prefix should never be detached */
1876			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1877			    pr->ndpr_raf_onlink == 0 ||
1878			    pr->ndpr_raf_auto == 0)
1879				continue;
1880
1881			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1882			    find_pfxlist_reachable_router(pr) == NULL)
1883				pr->ndpr_stateflags |= NDPRF_DETACHED;
1884			else if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1885			    find_pfxlist_reachable_router(pr) != NULL)
1886				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1887		}
1888	} else {
1889		/* there is no prefix that has a reachable router */
1890		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1891			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1892			    pr->ndpr_raf_onlink == 0 ||
1893			    pr->ndpr_raf_auto == 0)
1894				continue;
1895			pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1896		}
1897	}
1898
1899	/*
1900	 * Remove each interface route associated with a (just) detached
1901	 * prefix, and reinstall the interface route for a (just) attached
1902	 * prefix.  Note that all attempt of reinstallation does not
1903	 * necessarily success, when a same prefix is shared among multiple
1904	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1905	 * so we don't have to care about them.
1906	 */
1907restart:
1908	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1909		char ip6buf[INET6_ADDRSTRLEN];
1910		int e;
1911
1912		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1913		    pr->ndpr_raf_onlink == 0 ||
1914		    pr->ndpr_raf_auto == 0)
1915			continue;
1916
1917		flags = pr->ndpr_stateflags & (NDPRF_DETACHED | NDPRF_ONLINK);
1918		if (flags == 0 || flags == (NDPRF_DETACHED | NDPRF_ONLINK)) {
1919			genid = V_nd6_list_genid;
1920			ND6_RUNLOCK();
1921			if ((flags & NDPRF_ONLINK) != 0 &&
1922			    (e = nd6_prefix_offlink(pr)) != 0) {
1923				nd6log((LOG_ERR,
1924				    "%s: failed to make %s/%d offlink "
1925				    "(errno=%d)\n", __func__,
1926				    ip6_sprintf(ip6buf,
1927					    &pr->ndpr_prefix.sin6_addr),
1928					    pr->ndpr_plen, e));
1929			} else if ((flags & NDPRF_ONLINK) == 0 &&
1930			    (e = nd6_prefix_onlink(pr)) != 0) {
1931				nd6log((LOG_ERR,
1932				    "%s: failed to make %s/%d onlink "
1933				    "(errno=%d)\n", __func__,
1934				    ip6_sprintf(ip6buf,
1935					    &pr->ndpr_prefix.sin6_addr),
1936					    pr->ndpr_plen, e));
1937			}
1938			ND6_RLOCK();
1939			if (genid != V_nd6_list_genid)
1940				goto restart;
1941		}
1942	}
1943
1944	/*
1945	 * Changes on the prefix status might affect address status as well.
1946	 * Make sure that all addresses derived from an attached prefix are
1947	 * attached, and that all addresses derived from a detached prefix are
1948	 * detached.  Note, however, that a manually configured address should
1949	 * always be attached.
1950	 * The precise detection logic is same as the one for prefixes.
1951	 */
1952	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1953	CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1954		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1955			continue;
1956
1957		if (ifa->ia6_ndpr == NULL) {
1958			/*
1959			 * This can happen when we first configure the address
1960			 * (i.e. the address exists, but the prefix does not).
1961			 * XXX: complicated relationships...
1962			 */
1963			continue;
1964		}
1965
1966		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1967			break;
1968	}
1969	if (ifa) {
1970		CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1971			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1972				continue;
1973
1974			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1975				continue;
1976
1977			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1978				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1979					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1980					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1981					nd6_dad_start((struct ifaddr *)ifa, 0);
1982				}
1983			} else {
1984				ifa->ia6_flags |= IN6_IFF_DETACHED;
1985			}
1986		}
1987	} else {
1988		CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1989			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1990				continue;
1991
1992			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1993				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1994				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1995				/* Do we need a delay in this case? */
1996				nd6_dad_start((struct ifaddr *)ifa, 0);
1997			}
1998		}
1999	}
2000	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
2001	ND6_RUNLOCK();
2002	ND6_ONLINK_UNLOCK();
2003}
2004
2005/*
2006 * Add or remove interface route specified by @dst, @netmask and @ifp.
2007 * ifa can be NULL.
2008 * Returns 0 on success
2009 */
2010static int
2011nd6_prefix_rtrequest(uint32_t fibnum, int cmd, struct sockaddr_in6 *dst,
2012    struct sockaddr_in6 *netmask, struct ifnet *ifp, struct ifaddr *ifa)
2013{
2014	struct epoch_tracker et;
2015	int error;
2016
2017	/* Prepare gateway */
2018	struct sockaddr_dl_short sdl = {
2019		.sdl_family = AF_LINK,
2020		.sdl_len = sizeof(struct sockaddr_dl_short),
2021		.sdl_type = ifp->if_type,
2022		.sdl_index = ifp->if_index,
2023	};
2024
2025	struct rt_addrinfo info = {
2026		.rti_ifa = ifa,
2027		.rti_ifp = ifp,
2028		.rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
2029		.rti_info = {
2030			[RTAX_DST] = (struct sockaddr *)dst,
2031			[RTAX_NETMASK] = (struct sockaddr *)netmask,
2032			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
2033		},
2034	};
2035	/* Don't set additional per-gw filters on removal */
2036
2037	NET_EPOCH_ENTER(et);
2038	error = rib_handle_ifaddr_info(fibnum, cmd, &info);
2039	NET_EPOCH_EXIT(et);
2040	return (error);
2041}
2042
2043static int
2044nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
2045{
2046	int error;
2047
2048	struct sockaddr_in6 mask6 = {
2049		.sin6_family = AF_INET6,
2050		.sin6_len = sizeof(struct sockaddr_in6),
2051		.sin6_addr = pr->ndpr_mask,
2052	};
2053	struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL;
2054
2055	error = nd6_prefix_rtrequest(pr->ndpr_ifp->if_fib, RTM_ADD,
2056	    &pr->ndpr_prefix, pmask6, pr->ndpr_ifp, ifa);
2057	if (error == 0)
2058		pr->ndpr_stateflags |= NDPRF_ONLINK;
2059
2060	return (error);
2061}
2062
2063static int
2064nd6_prefix_onlink(struct nd_prefix *pr)
2065{
2066	struct epoch_tracker et;
2067	struct ifaddr *ifa;
2068	struct ifnet *ifp = pr->ndpr_ifp;
2069	struct nd_prefix *opr;
2070	char ip6buf[INET6_ADDRSTRLEN];
2071	int error;
2072
2073	ND6_ONLINK_LOCK_ASSERT();
2074	ND6_UNLOCK_ASSERT();
2075
2076	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0)
2077		return (EEXIST);
2078
2079	/*
2080	 * Add the interface route associated with the prefix.  Before
2081	 * installing the route, check if there's the same prefix on another
2082	 * interface, and the prefix has already installed the interface route.
2083	 * Although such a configuration is expected to be rare, we explicitly
2084	 * allow it.
2085	 */
2086	ND6_RLOCK();
2087	LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2088		if (opr == pr)
2089			continue;
2090
2091		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
2092			continue;
2093
2094		if (!V_rt_add_addr_allfibs &&
2095		    opr->ndpr_ifp->if_fib != pr->ndpr_ifp->if_fib)
2096			continue;
2097
2098		if (opr->ndpr_plen == pr->ndpr_plen &&
2099		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2100		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2101			ND6_RUNLOCK();
2102			return (0);
2103		}
2104	}
2105	ND6_RUNLOCK();
2106
2107	/*
2108	 * We prefer link-local addresses as the associated interface address.
2109	 */
2110	/* search for a link-local addr */
2111	NET_EPOCH_ENTER(et);
2112	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
2113	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
2114	if (ifa == NULL) {
2115		/* XXX: freebsd does not have ifa_ifwithaf */
2116		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2117			if (ifa->ifa_addr->sa_family == AF_INET6) {
2118				ifa_ref(ifa);
2119				break;
2120			}
2121		}
2122		/* should we care about ia6_flags? */
2123	}
2124	if (ifa == NULL) {
2125		/*
2126		 * This can still happen, when, for example, we receive an RA
2127		 * containing a prefix with the L bit set and the A bit clear,
2128		 * after removing all IPv6 addresses on the receiving
2129		 * interface.  This should, of course, be rare though.
2130		 */
2131		nd6log((LOG_NOTICE,
2132		    "%s: failed to find any ifaddr to add route for a "
2133		    "prefix(%s/%d) on %s\n", __func__,
2134		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
2135		    pr->ndpr_plen, if_name(ifp)));
2136		error = 0;
2137	} else {
2138		error = nd6_prefix_onlink_rtrequest(pr, ifa);
2139		ifa_free(ifa);
2140	}
2141	NET_EPOCH_EXIT(et);
2142
2143	return (error);
2144}
2145
2146int
2147nd6_prefix_offlink(struct nd_prefix *pr)
2148{
2149	int error = 0;
2150	struct ifnet *ifp = pr->ndpr_ifp;
2151	struct nd_prefix *opr;
2152	char ip6buf[INET6_ADDRSTRLEN];
2153	uint64_t genid;
2154	int a_failure;
2155
2156	ND6_ONLINK_LOCK_ASSERT();
2157	ND6_UNLOCK_ASSERT();
2158
2159	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0)
2160		return (EEXIST);
2161
2162	struct sockaddr_in6 mask6 = {
2163		.sin6_family = AF_INET6,
2164		.sin6_len = sizeof(struct sockaddr_in6),
2165		.sin6_addr = pr->ndpr_mask,
2166	};
2167	struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL;
2168
2169	error = nd6_prefix_rtrequest(ifp->if_fib, RTM_DELETE,
2170	    &pr->ndpr_prefix, pmask6, ifp, NULL);
2171
2172	a_failure = 1;
2173	if (error == 0) {
2174		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
2175
2176		/*
2177		 * There might be the same prefix on another interface,
2178		 * the prefix which could not be on-link just because we have
2179		 * the interface route (see comments in nd6_prefix_onlink).
2180		 * If there's one, try to make the prefix on-link on the
2181		 * interface.
2182		 */
2183		ND6_RLOCK();
2184restart:
2185		LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2186			/*
2187			 * KAME specific: detached prefixes should not be
2188			 * on-link.
2189			 */
2190			if (opr == pr || (opr->ndpr_stateflags &
2191			    (NDPRF_ONLINK | NDPRF_DETACHED)) != 0)
2192				continue;
2193
2194			if (opr->ndpr_plen == pr->ndpr_plen &&
2195			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2196			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2197				int e;
2198
2199				genid = V_nd6_list_genid;
2200				ND6_RUNLOCK();
2201				if ((e = nd6_prefix_onlink(opr)) != 0) {
2202					nd6log((LOG_ERR,
2203					    "%s: failed to recover a prefix "
2204					    "%s/%d from %s to %s (errno=%d)\n",
2205					    __func__, ip6_sprintf(ip6buf,
2206						&opr->ndpr_prefix.sin6_addr),
2207					    opr->ndpr_plen, if_name(ifp),
2208					    if_name(opr->ndpr_ifp), e));
2209				} else
2210					a_failure = 0;
2211				ND6_RLOCK();
2212				if (genid != V_nd6_list_genid)
2213					goto restart;
2214			}
2215		}
2216		ND6_RUNLOCK();
2217	} else {
2218		/* XXX: can we still set the NDPRF_ONLINK flag? */
2219		nd6log((LOG_ERR,
2220		    "%s: failed to delete route: %s/%d on %s (errno=%d)\n",
2221		    __func__, ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
2222		    pr->ndpr_plen, if_name(ifp), error));
2223	}
2224
2225	if (a_failure)
2226		lltable_prefix_free(AF_INET6,
2227		    (struct sockaddr *)&pr->ndpr_prefix,
2228		    (struct sockaddr *)&mask6, LLE_STATIC);
2229
2230	return (error);
2231}
2232
2233/*
2234 * ia0 - corresponding public address
2235 */
2236int
2237in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
2238{
2239	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
2240	struct in6_ifaddr *newia;
2241	struct in6_aliasreq ifra;
2242	int error;
2243	int trylimit = 3;	/* XXX: adhoc value */
2244	int updateflags;
2245	u_int32_t randid[2];
2246	time_t vltime0, pltime0;
2247
2248	in6_prepare_ifra(&ifra, &ia0->ia_addr.sin6_addr,
2249	    &ia0->ia_prefixmask.sin6_addr);
2250
2251	ifra.ifra_addr = ia0->ia_addr;	/* XXX: do we need this ? */
2252	/* clear the old IFID */
2253	IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr,
2254	    &ifra.ifra_prefixmask.sin6_addr);
2255
2256  again:
2257	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
2258	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
2259		nd6log((LOG_NOTICE, "%s: failed to find a good random IFID\n",
2260		    __func__));
2261		return (EINVAL);
2262	}
2263	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
2264	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
2265	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
2266	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
2267
2268	/*
2269	 * in6_get_tmpifid() quite likely provided a unique interface ID.
2270	 * However, we may still have a chance to see collision, because
2271	 * there may be a time lag between generation of the ID and generation
2272	 * of the address.  So, we'll do one more sanity check.
2273	 */
2274
2275	if (in6_localip(&ifra.ifra_addr.sin6_addr) != 0) {
2276		if (trylimit-- > 0) {
2277			forcegen = 1;
2278			goto again;
2279		}
2280
2281		/* Give up.  Something strange should have happened.  */
2282		nd6log((LOG_NOTICE, "%s: failed to find a unique random IFID\n",
2283		    __func__));
2284		return (EEXIST);
2285	}
2286
2287	/*
2288	 * The Valid Lifetime is the lower of the Valid Lifetime of the
2289         * public address or TEMP_VALID_LIFETIME.
2290	 * The Preferred Lifetime is the lower of the Preferred Lifetime
2291         * of the public address or TEMP_PREFERRED_LIFETIME -
2292         * DESYNC_FACTOR.
2293	 */
2294	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
2295		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2296		    (ia0->ia6_lifetime.ia6t_vltime -
2297		    (time_uptime - ia0->ia6_updatetime));
2298		if (vltime0 > V_ip6_temp_valid_lifetime)
2299			vltime0 = V_ip6_temp_valid_lifetime;
2300	} else
2301		vltime0 = V_ip6_temp_valid_lifetime;
2302	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2303		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2304		    (ia0->ia6_lifetime.ia6t_pltime -
2305		    (time_uptime - ia0->ia6_updatetime));
2306		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
2307			pltime0 = V_ip6_temp_preferred_lifetime -
2308			    V_ip6_desync_factor;
2309		}
2310	} else
2311		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
2312	ifra.ifra_lifetime.ia6t_vltime = vltime0;
2313	ifra.ifra_lifetime.ia6t_pltime = pltime0;
2314
2315	/*
2316	 * A temporary address is created only if this calculated Preferred
2317	 * Lifetime is greater than REGEN_ADVANCE time units.
2318	 */
2319	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2320		return (0);
2321
2322	/* XXX: scope zone ID? */
2323
2324	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2325
2326	/* allocate ifaddr structure, link into chain, etc. */
2327	updateflags = 0;
2328	if (delay)
2329		updateflags |= IN6_IFAUPDATE_DADDELAY;
2330	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2331		return (error);
2332
2333	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2334	if (newia == NULL) {	/* XXX: can it happen? */
2335		nd6log((LOG_ERR,
2336		    "%s: ifa update succeeded, but we got no ifaddr\n",
2337		    __func__));
2338		return (EINVAL); /* XXX */
2339	}
2340	newia->ia6_ndpr = ia0->ia6_ndpr;
2341	newia->ia6_ndpr->ndpr_addrcnt++;
2342	ifa_free(&newia->ia_ifa);
2343
2344	/*
2345	 * A newly added address might affect the status of other addresses.
2346	 * XXX: when the temporary address is generated with a new public
2347	 * address, the onlink check is redundant.  However, it would be safe
2348	 * to do the check explicitly everywhere a new address is generated,
2349	 * and, in fact, we surely need the check when we create a new
2350	 * temporary address due to deprecation of an old temporary address.
2351	 */
2352	pfxlist_onlink_check();
2353
2354	return (0);
2355}
2356
2357static int
2358rt6_deleteroute(const struct rtentry *rt, const struct nhop_object *nh,
2359    void *arg)
2360{
2361	struct in6_addr *gate = (struct in6_addr *)arg;
2362	int nh_rt_flags;
2363
2364	if (nh->gw_sa.sa_family != AF_INET6)
2365		return (0);
2366
2367	if (!IN6_ARE_ADDR_EQUAL(gate, &nh->gw6_sa.sin6_addr)) {
2368		return (0);
2369	}
2370
2371	/*
2372	 * Do not delete a static route.
2373	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2374	 * 'cloned' bit instead?
2375	 */
2376	nh_rt_flags = nhop_get_rtflags(nh);
2377	if ((nh_rt_flags & RTF_STATIC) != 0)
2378		return (0);
2379
2380	/*
2381	 * We delete only host route. This means, in particular, we don't
2382	 * delete default route.
2383	 */
2384	if ((nh_rt_flags & RTF_HOST) == 0)
2385		return (0);
2386
2387	return (1);
2388#undef SIN6
2389}
2390
2391/*
2392 * Delete all the routing table entries that use the specified gateway.
2393 * XXX: this function causes search through all entries of routing table, so
2394 * it shouldn't be called when acting as a router.
2395 */
2396void
2397rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2398{
2399
2400	/* We'll care only link-local addresses */
2401	if (!IN6_IS_ADDR_LINKLOCAL(gateway))
2402		return;
2403
2404	/* XXX Do we really need to walk any but the default FIB? */
2405	rib_foreach_table_walk_del(AF_INET6, rt6_deleteroute, (void *)gateway);
2406}
2407
2408int
2409nd6_setdefaultiface(int ifindex)
2410{
2411
2412	if (V_nd6_defifindex != ifindex) {
2413		V_nd6_defifindex = ifindex;
2414		if (V_nd6_defifindex != 0) {
2415			struct epoch_tracker et;
2416
2417			/*
2418			 * XXXGL: this function should use ifnet_byindex_ref!
2419			 */
2420			NET_EPOCH_ENTER(et);
2421			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2422			NET_EPOCH_EXIT(et);
2423			if (V_nd6_defifp == NULL)
2424				return (EINVAL);
2425		} else
2426			V_nd6_defifp = NULL;
2427
2428		/*
2429		 * Our current implementation assumes one-to-one mapping between
2430		 * interfaces and links, so it would be natural to use the
2431		 * default interface as the default link.
2432		 */
2433		scope6_setdefault(V_nd6_defifp);
2434	}
2435
2436	return (0);
2437}
2438
2439bool
2440nd6_defrouter_list_empty(void)
2441{
2442
2443	return (TAILQ_EMPTY(&V_nd6_defrouter));
2444}
2445
2446void
2447nd6_defrouter_timer(void)
2448{
2449	struct nd_defrouter *dr, *ndr;
2450	struct nd6_drhead drq;
2451
2452	TAILQ_INIT(&drq);
2453
2454	ND6_WLOCK();
2455	TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr)
2456		if (dr->expire && dr->expire < time_uptime)
2457			defrouter_unlink(dr, &drq);
2458	ND6_WUNLOCK();
2459
2460	while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2461		TAILQ_REMOVE(&drq, dr, dr_entry);
2462		defrouter_del(dr);
2463	}
2464}
2465
2466/*
2467 * Nuke default router list entries toward ifp.
2468 * We defer removal of default router list entries that is installed in the
2469 * routing table, in order to keep additional side effects as small as possible.
2470 */
2471void
2472nd6_defrouter_purge(struct ifnet *ifp)
2473{
2474	struct nd_defrouter *dr, *ndr;
2475	struct nd6_drhead drq;
2476
2477	TAILQ_INIT(&drq);
2478
2479	ND6_WLOCK();
2480	TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) {
2481		if (dr->installed)
2482			continue;
2483		if (dr->ifp == ifp)
2484			defrouter_unlink(dr, &drq);
2485	}
2486	TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) {
2487		if (!dr->installed)
2488			continue;
2489		if (dr->ifp == ifp)
2490			defrouter_unlink(dr, &drq);
2491	}
2492	ND6_WUNLOCK();
2493
2494	/* Delete the unlinked router objects. */
2495	while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2496		TAILQ_REMOVE(&drq, dr, dr_entry);
2497		defrouter_del(dr);
2498	}
2499}
2500
2501void
2502nd6_defrouter_flush_all(void)
2503{
2504	struct nd_defrouter *dr;
2505	struct nd6_drhead drq;
2506
2507	TAILQ_INIT(&drq);
2508
2509	ND6_WLOCK();
2510	while ((dr = TAILQ_FIRST(&V_nd6_defrouter)) != NULL)
2511		defrouter_unlink(dr, &drq);
2512	ND6_WUNLOCK();
2513
2514	while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2515		TAILQ_REMOVE(&drq, dr, dr_entry);
2516		defrouter_del(dr);
2517	}
2518}
2519
2520void
2521nd6_defrouter_init(void)
2522{
2523
2524	TAILQ_INIT(&V_nd6_defrouter);
2525}
2526
2527static int
2528nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2529{
2530	struct in6_defrouter d;
2531	struct nd_defrouter *dr;
2532	int error;
2533
2534	if (req->newptr != NULL)
2535		return (EPERM);
2536
2537	error = sysctl_wire_old_buffer(req, 0);
2538	if (error != 0)
2539		return (error);
2540
2541	bzero(&d, sizeof(d));
2542	d.rtaddr.sin6_family = AF_INET6;
2543	d.rtaddr.sin6_len = sizeof(d.rtaddr);
2544
2545	ND6_RLOCK();
2546	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
2547		d.rtaddr.sin6_addr = dr->rtaddr;
2548		error = sa6_recoverscope(&d.rtaddr);
2549		if (error != 0)
2550			break;
2551		d.flags = dr->raflags;
2552		d.rtlifetime = dr->rtlifetime;
2553		d.expire = dr->expire + (time_second - time_uptime);
2554		d.if_index = dr->ifp->if_index;
2555		error = SYSCTL_OUT(req, &d, sizeof(d));
2556		if (error != 0)
2557			break;
2558	}
2559	ND6_RUNLOCK();
2560	return (error);
2561}
2562SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2563	CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
2564	NULL, 0, nd6_sysctl_drlist, "S,in6_defrouter",
2565	"NDP default router list");
2566