xref: /freebsd/sys/netinet6/nd6_rtr.c (revision b9e5884ef786a6c0fd203064baa07ef308025707)
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 __FBSDID("$FreeBSD$");
36 
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/refcount.h>
45 #include <sys/socket.h>
46 #include <sys/sockio.h>
47 #include <sys/time.h>
48 #include <sys/kernel.h>
49 #include <sys/lock.h>
50 #include <sys/errno.h>
51 #include <sys/rmlock.h>
52 #include <sys/rwlock.h>
53 #include <sys/sysctl.h>
54 #include <sys/syslog.h>
55 #include <sys/queue.h>
56 
57 #include <net/if.h>
58 #include <net/if_var.h>
59 #include <net/if_types.h>
60 #include <net/if_dl.h>
61 #include <net/route.h>
62 #include <net/route/nhop.h>
63 #include <net/route/route_ctl.h>
64 #include <net/radix.h>
65 #include <net/vnet.h>
66 
67 #include <netinet/in.h>
68 #include <net/if_llatbl.h>
69 #include <netinet6/in6_var.h>
70 #include <netinet6/in6_ifattach.h>
71 #include <netinet/ip6.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/nd6.h>
74 #include <netinet/icmp6.h>
75 #include <netinet6/scope6_var.h>
76 
77 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
78 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
79     struct mbuf *, int);
80 static int nd6_prefix_onlink(struct nd_prefix *);
81 
82 TAILQ_HEAD(nd6_drhead, nd_defrouter);
83 VNET_DEFINE_STATIC(struct nd6_drhead, nd6_defrouter);
84 #define	V_nd6_defrouter			VNET(nd6_defrouter)
85 
86 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
87 #define	V_nd6_recalc_reachtm_interval	VNET(nd6_recalc_reachtm_interval)
88 
89 VNET_DEFINE_STATIC(struct ifnet *, nd6_defifp);
90 VNET_DEFINE(int, nd6_defifindex);
91 #define	V_nd6_defifp			VNET(nd6_defifp)
92 
93 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
94 
95 VNET_DEFINE(int, ip6_desync_factor);
96 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
97 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
98 
99 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
100 
101 #ifdef EXPERIMENTAL
102 VNET_DEFINE(int, nd6_ignore_ipv6_only_ra) = 1;
103 #endif
104 
105 SYSCTL_DECL(_net_inet6_icmp6);
106 
107 /* RTPREF_MEDIUM has to be 0! */
108 #define RTPREF_HIGH	1
109 #define RTPREF_MEDIUM	0
110 #define RTPREF_LOW	(-1)
111 #define RTPREF_RESERVED	(-2)
112 #define RTPREF_INVALID	(-3)	/* internal */
113 
114 static void
115 defrouter_ref(struct nd_defrouter *dr)
116 {
117 
118 	refcount_acquire(&dr->refcnt);
119 }
120 
121 void
122 defrouter_rele(struct nd_defrouter *dr)
123 {
124 
125 	if (refcount_release(&dr->refcnt))
126 		free(dr, M_IP6NDP);
127 }
128 
129 /*
130  * Remove a router from the global list and optionally stash it in a
131  * caller-supplied queue.
132  */
133 static void
134 defrouter_unlink(struct nd_defrouter *dr, struct nd6_drhead *drq)
135 {
136 
137 	ND6_WLOCK_ASSERT();
138 
139 	TAILQ_REMOVE(&V_nd6_defrouter, dr, dr_entry);
140 	V_nd6_list_genid++;
141 	if (drq != NULL)
142 		TAILQ_INSERT_TAIL(drq, dr, dr_entry);
143 }
144 
145 /*
146  * Receive Router Solicitation Message - just for routers.
147  * Router solicitation/advertisement is mostly managed by userland program
148  * (rtadvd) so here we have no function like nd6_ra_output().
149  *
150  * Based on RFC 2461
151  */
152 void
153 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
154 {
155 	struct ifnet *ifp;
156 	struct ip6_hdr *ip6;
157 	struct nd_router_solicit *nd_rs;
158 	struct in6_addr saddr6;
159 	union nd_opts ndopts;
160 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
161 	char *lladdr;
162 	int lladdrlen;
163 
164 	ifp = m->m_pkthdr.rcvif;
165 
166 	/*
167 	 * Accept RS only when V_ip6_forwarding=1 and the interface has
168 	 * no ND6_IFF_ACCEPT_RTADV.
169 	 */
170 	if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
171 		goto freeit;
172 
173 	/* RFC 6980: Nodes MUST silently ignore fragments */
174 	if(m->m_flags & M_FRAGMENTED)
175 		goto freeit;
176 
177 	/* Sanity checks */
178 	ip6 = mtod(m, struct ip6_hdr *);
179 	if (__predict_false(ip6->ip6_hlim != 255)) {
180 		ICMP6STAT_INC(icp6s_invlhlim);
181 		nd6log((LOG_ERR,
182 		    "%s: invalid hlim (%d) from %s to %s on %s\n", __func__,
183 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
184 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
185 		goto bad;
186 	}
187 
188 	/*
189 	 * Don't update the neighbor cache, if src = ::.
190 	 * This indicates that the src has no IP address assigned yet.
191 	 */
192 	saddr6 = ip6->ip6_src;
193 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
194 		goto freeit;
195 
196 	if (m->m_len < off + icmp6len) {
197 		m = m_pullup(m, off + icmp6len);
198 		if (m == NULL) {
199 			IP6STAT_INC(ip6s_exthdrtoolong);
200 			return;
201 		}
202 	}
203 	ip6 = mtod(m, struct ip6_hdr *);
204 	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
205 
206 	icmp6len -= sizeof(*nd_rs);
207 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
208 	if (nd6_options(&ndopts) < 0) {
209 		nd6log((LOG_INFO,
210 		    "%s: invalid ND option, ignored\n", __func__));
211 		/* nd6_options have incremented stats */
212 		goto freeit;
213 	}
214 
215 	lladdr = NULL;
216 	lladdrlen = 0;
217 	if (ndopts.nd_opts_src_lladdr) {
218 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
219 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
220 	}
221 
222 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
223 		nd6log((LOG_INFO,
224 		    "%s: lladdrlen mismatch for %s (if %d, RS packet %d)\n",
225 		    __func__, ip6_sprintf(ip6bufs, &saddr6),
226 		    ifp->if_addrlen, lladdrlen - 2));
227 		goto bad;
228 	}
229 
230 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
231 
232  freeit:
233 	m_freem(m);
234 	return;
235 
236  bad:
237 	ICMP6STAT_INC(icp6s_badrs);
238 	m_freem(m);
239 }
240 
241 #ifdef EXPERIMENTAL
242 /*
243  * An initial update routine for draft-ietf-6man-ipv6only-flag.
244  * We need to iterate over all default routers for the given
245  * interface to see whether they are all advertising the "S"
246  * (IPv6-Only) flag.  If they do set, otherwise unset, the
247  * interface flag we later use to filter on.
248  */
249 static void
250 defrtr_ipv6_only_ifp(struct ifnet *ifp)
251 {
252 	struct nd_defrouter *dr;
253 	bool ipv6_only, ipv6_only_old;
254 #ifdef INET
255 	struct epoch_tracker et;
256 	struct ifaddr *ifa;
257 	bool has_ipv4_addr;
258 #endif
259 
260 	if (V_nd6_ignore_ipv6_only_ra != 0)
261 		return;
262 
263 	ipv6_only = true;
264 	ND6_RLOCK();
265 	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
266 		if (dr->ifp == ifp &&
267 		    (dr->raflags & ND_RA_FLAG_IPV6_ONLY) == 0)
268 			ipv6_only = false;
269 	ND6_RUNLOCK();
270 
271 	IF_AFDATA_WLOCK(ifp);
272 	ipv6_only_old = ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY;
273 	IF_AFDATA_WUNLOCK(ifp);
274 
275 	/* If nothing changed, we have an early exit. */
276 	if (ipv6_only == ipv6_only_old)
277 		return;
278 
279 #ifdef INET
280 	/*
281 	 * Should we want to set the IPV6-ONLY flag, check if the
282 	 * interface has a non-0/0 and non-link-local IPv4 address
283 	 * configured on it.  If it has we will assume working
284 	 * IPv4 operations and will clear the interface flag.
285 	 */
286 	has_ipv4_addr = false;
287 	if (ipv6_only) {
288 		NET_EPOCH_ENTER(et);
289 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
290 			if (ifa->ifa_addr->sa_family != AF_INET)
291 				continue;
292 			if (in_canforward(
293 			    satosin(ifa->ifa_addr)->sin_addr)) {
294 				has_ipv4_addr = true;
295 				break;
296 			}
297 		}
298 		NET_EPOCH_EXIT(et);
299 	}
300 	if (ipv6_only && has_ipv4_addr) {
301 		log(LOG_NOTICE, "%s rcvd RA w/ IPv6-Only flag set but has IPv4 "
302 		    "configured, ignoring IPv6-Only flag.\n", ifp->if_xname);
303 		ipv6_only = false;
304 	}
305 #endif
306 
307 	IF_AFDATA_WLOCK(ifp);
308 	if (ipv6_only)
309 		ND_IFINFO(ifp)->flags |= ND6_IFF_IPV6_ONLY;
310 	else
311 		ND_IFINFO(ifp)->flags &= ~ND6_IFF_IPV6_ONLY;
312 	IF_AFDATA_WUNLOCK(ifp);
313 
314 #ifdef notyet
315 	/* Send notification of flag change. */
316 #endif
317 }
318 
319 static void
320 defrtr_ipv6_only_ipf_down(struct ifnet *ifp)
321 {
322 
323 	IF_AFDATA_WLOCK(ifp);
324 	ND_IFINFO(ifp)->flags &= ~ND6_IFF_IPV6_ONLY;
325 	IF_AFDATA_WUNLOCK(ifp);
326 }
327 #endif	/* EXPERIMENTAL */
328 
329 void
330 nd6_ifnet_link_event(void *arg __unused, struct ifnet *ifp, int linkstate)
331 {
332 
333 	/*
334 	 * XXX-BZ we might want to trigger re-evaluation of our default router
335 	 * availability. E.g., on link down the default router might be
336 	 * unreachable but a different interface might still have connectivity.
337 	 */
338 
339 #ifdef EXPERIMENTAL
340 	if (linkstate == LINK_STATE_DOWN)
341 		defrtr_ipv6_only_ipf_down(ifp);
342 #endif
343 }
344 
345 /*
346  * Receive Router Advertisement Message.
347  *
348  * Based on RFC 2461
349  * TODO: on-link bit on prefix information
350  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
351  */
352 void
353 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
354 {
355 	struct ifnet *ifp;
356 	struct nd_ifinfo *ndi;
357 	struct ip6_hdr *ip6;
358 	struct nd_router_advert *nd_ra;
359 	struct in6_addr saddr6;
360 	struct nd_defrouter *dr;
361 	union nd_opts ndopts;
362 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
363 	int mcast;
364 
365 	/*
366 	 * We only accept RAs only when the per-interface flag
367 	 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
368 	 */
369 	ifp = m->m_pkthdr.rcvif;
370 	ndi = ND_IFINFO(ifp);
371 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
372 		goto freeit;
373 
374 	/* RFC 6980: Nodes MUST silently ignore fragments */
375 	if(m->m_flags & M_FRAGMENTED)
376 		goto freeit;
377 
378 	ip6 = mtod(m, struct ip6_hdr *);
379 	if (__predict_false(ip6->ip6_hlim != 255)) {
380 		ICMP6STAT_INC(icp6s_invlhlim);
381 		nd6log((LOG_ERR,
382 		    "%s: invalid hlim (%d) from %s to %s on %s\n", __func__,
383 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
384 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
385 		goto bad;
386 	}
387 
388 	saddr6 = ip6->ip6_src;
389 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
390 		nd6log((LOG_ERR,
391 		    "%s: src %s is not link-local\n", __func__,
392 		    ip6_sprintf(ip6bufs, &saddr6)));
393 		goto bad;
394 	}
395 
396 	if (m->m_len < off + icmp6len) {
397 		m = m_pullup(m, off + icmp6len);
398 		if (m == NULL) {
399 			IP6STAT_INC(ip6s_exthdrtoolong);
400 			return;
401 		}
402 	}
403 	ip6 = mtod(m, struct ip6_hdr *);
404 	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
405 
406 	icmp6len -= sizeof(*nd_ra);
407 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
408 	if (nd6_options(&ndopts) < 0) {
409 		nd6log((LOG_INFO,
410 		    "%s: invalid ND option, ignored\n", __func__));
411 		/* nd6_options have incremented stats */
412 		goto freeit;
413 	}
414 
415 	mcast = 0;
416 	dr = NULL;
417     {
418 	struct nd_defrouter dr0;
419 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
420 
421 	/* remember if this is a multicasted advertisement */
422 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
423 		mcast = 1;
424 
425 	bzero(&dr0, sizeof(dr0));
426 	dr0.rtaddr = saddr6;
427 	dr0.raflags = nd_ra->nd_ra_flags_reserved;
428 	/*
429 	 * Effectively-disable routes from RA messages when
430 	 * ND6_IFF_NO_RADR enabled on the receiving interface or
431 	 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
432 	 */
433 	if (ndi->flags & ND6_IFF_NO_RADR)
434 		dr0.rtlifetime = 0;
435 	else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
436 		dr0.rtlifetime = 0;
437 	else
438 		dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
439 	dr0.expire = time_uptime + dr0.rtlifetime;
440 	dr0.ifp = ifp;
441 	/* unspecified or not? (RFC 2461 6.3.4) */
442 	if (advreachable) {
443 		advreachable = ntohl(advreachable);
444 		if (advreachable <= MAX_REACHABLE_TIME &&
445 		    ndi->basereachable != advreachable) {
446 			ndi->basereachable = advreachable;
447 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
448 			ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
449 		}
450 	}
451 	if (nd_ra->nd_ra_retransmit)
452 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
453 	if (nd_ra->nd_ra_curhoplimit) {
454 		if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
455 			ndi->chlim = nd_ra->nd_ra_curhoplimit;
456 		else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) {
457 			log(LOG_ERR, "RA with a lower CurHopLimit sent from "
458 			    "%s on %s (current = %d, received = %d). "
459 			    "Ignored.\n", ip6_sprintf(ip6bufs, &ip6->ip6_src),
460 			    if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
461 		}
462 	}
463 	dr = defrtrlist_update(&dr0);
464 #ifdef EXPERIMENTAL
465 	defrtr_ipv6_only_ifp(ifp);
466 #endif
467     }
468 
469 	/*
470 	 * prefix
471 	 */
472 	if (ndopts.nd_opts_pi) {
473 		struct nd_opt_hdr *pt;
474 		struct nd_opt_prefix_info *pi = NULL;
475 		struct nd_prefixctl pr;
476 
477 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
478 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
479 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
480 						(pt->nd_opt_len << 3))) {
481 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
482 				continue;
483 			pi = (struct nd_opt_prefix_info *)pt;
484 
485 			if (pi->nd_opt_pi_len != 4) {
486 				nd6log((LOG_INFO,
487 				    "%s: invalid option len %d for prefix "
488 				    "information option, ignored\n", __func__,
489 				    pi->nd_opt_pi_len));
490 				continue;
491 			}
492 
493 			if (128 < pi->nd_opt_pi_prefix_len) {
494 				nd6log((LOG_INFO,
495 				    "%s: invalid prefix len %d for prefix "
496 				    "information option, ignored\n", __func__,
497 				    pi->nd_opt_pi_prefix_len));
498 				continue;
499 			}
500 
501 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
502 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
503 				nd6log((LOG_INFO,
504 				    "%s: invalid prefix %s, ignored\n",
505 				    __func__, ip6_sprintf(ip6bufs,
506 					&pi->nd_opt_pi_prefix)));
507 				continue;
508 			}
509 
510 			bzero(&pr, sizeof(pr));
511 			pr.ndpr_prefix.sin6_family = AF_INET6;
512 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
513 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
514 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
515 
516 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
517 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
518 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
519 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
520 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
521 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
522 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
523 			(void)prelist_update(&pr, dr, m, mcast);
524 		}
525 	}
526 	if (dr != NULL) {
527 		defrouter_rele(dr);
528 		dr = NULL;
529 	}
530 
531 	/*
532 	 * MTU
533 	 */
534 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
535 		u_long mtu;
536 		u_long maxmtu;
537 
538 		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
539 
540 		/* lower bound */
541 		if (mtu < IPV6_MMTU) {
542 			nd6log((LOG_INFO, "%s: bogus mtu option mtu=%lu sent "
543 			    "from %s, ignoring\n", __func__,
544 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
545 			goto skip;
546 		}
547 
548 		/* upper bound */
549 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
550 		    ? ndi->maxmtu : ifp->if_mtu;
551 		if (mtu <= maxmtu) {
552 			int change = (ndi->linkmtu != mtu);
553 
554 			ndi->linkmtu = mtu;
555 			if (change) {
556 				/* in6_maxmtu may change */
557 				in6_setmaxmtu();
558 				rt_updatemtu(ifp);
559 			}
560 		} else {
561 			nd6log((LOG_INFO, "%s: bogus mtu=%lu sent from %s; "
562 			    "exceeds maxmtu %lu, ignoring\n", __func__,
563 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
564 		}
565 	}
566 
567  skip:
568 
569 	/*
570 	 * Source link layer address
571 	 */
572     {
573 	char *lladdr = NULL;
574 	int lladdrlen = 0;
575 
576 	if (ndopts.nd_opts_src_lladdr) {
577 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
578 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
579 	}
580 
581 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
582 		nd6log((LOG_INFO,
583 		    "%s: lladdrlen mismatch for %s (if %d, RA packet %d)\n",
584 		    __func__, ip6_sprintf(ip6bufs, &saddr6),
585 		    ifp->if_addrlen, lladdrlen - 2));
586 		goto bad;
587 	}
588 
589 	nd6_cache_lladdr(ifp, &saddr6, lladdr,
590 	    lladdrlen, ND_ROUTER_ADVERT, 0);
591 
592 	/*
593 	 * Installing a link-layer address might change the state of the
594 	 * router's neighbor cache, which might also affect our on-link
595 	 * detection of adveritsed prefixes.
596 	 */
597 	pfxlist_onlink_check();
598     }
599 
600  freeit:
601 	m_freem(m);
602 	return;
603 
604  bad:
605 	ICMP6STAT_INC(icp6s_badra);
606 	m_freem(m);
607 }
608 
609 /* PFXRTR */
610 static struct nd_pfxrouter *
611 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
612 {
613 	struct nd_pfxrouter *search;
614 
615 	ND6_LOCK_ASSERT();
616 
617 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
618 		if (search->router == dr)
619 			break;
620 	}
621 	return (search);
622 }
623 
624 static void
625 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
626 {
627 	struct nd_pfxrouter *new;
628 	bool update;
629 
630 	ND6_UNLOCK_ASSERT();
631 
632 	ND6_RLOCK();
633 	if (pfxrtr_lookup(pr, dr) != NULL) {
634 		ND6_RUNLOCK();
635 		return;
636 	}
637 	ND6_RUNLOCK();
638 
639 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
640 	if (new == NULL)
641 		return;
642 	defrouter_ref(dr);
643 	new->router = dr;
644 
645 	ND6_WLOCK();
646 	if (pfxrtr_lookup(pr, dr) == NULL) {
647 		LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
648 		update = true;
649 	} else {
650 		/* We lost a race to add the reference. */
651 		defrouter_rele(dr);
652 		free(new, M_IP6NDP);
653 		update = false;
654 	}
655 	ND6_WUNLOCK();
656 
657 	if (update)
658 		pfxlist_onlink_check();
659 }
660 
661 static void
662 pfxrtr_del(struct nd_pfxrouter *pfr)
663 {
664 
665 	ND6_WLOCK_ASSERT();
666 
667 	LIST_REMOVE(pfr, pfr_entry);
668 	defrouter_rele(pfr->router);
669 	free(pfr, M_IP6NDP);
670 }
671 
672 /* Default router list processing sub routines. */
673 static void
674 defrouter_addreq(struct nd_defrouter *new)
675 {
676 	struct sockaddr_in6 def, mask, gate;
677 	struct rt_addrinfo info;
678 	struct rib_cmd_info rc;
679 	unsigned int fibnum;
680 	int error;
681 
682 	bzero(&def, sizeof(def));
683 	bzero(&mask, sizeof(mask));
684 	bzero(&gate, sizeof(gate));
685 
686 	def.sin6_len = mask.sin6_len = gate.sin6_len =
687 	    sizeof(struct sockaddr_in6);
688 	def.sin6_family = gate.sin6_family = AF_INET6;
689 	gate.sin6_addr = new->rtaddr;
690 	fibnum = new->ifp->if_fib;
691 
692 	bzero((caddr_t)&info, sizeof(info));
693 	info.rti_flags = RTF_GATEWAY;
694 	info.rti_info[RTAX_DST] = (struct sockaddr *)&def;
695 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gate;
696 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&mask;
697 
698 	NET_EPOCH_ASSERT();
699 	error = rib_action(fibnum, RTM_ADD, &info, &rc);
700 	if (error == 0) {
701 		struct nhop_object *nh = nhop_select_func(rc.rc_nh_new, 0);
702 		rt_routemsg(RTM_ADD, rc.rc_rt, nh, fibnum);
703 		new->installed = 1;
704 	}
705 }
706 
707 /*
708  * Remove the default route for a given router.
709  * This is just a subroutine function for defrouter_select_fib(), and
710  * should not be called from anywhere else.
711  */
712 static void
713 defrouter_delreq(struct nd_defrouter *dr)
714 {
715 	struct sockaddr_in6 def, mask, gate;
716 	struct rt_addrinfo info;
717 	struct rib_cmd_info rc;
718 	struct epoch_tracker et;
719 	unsigned int fibnum;
720 	int error;
721 
722 	bzero(&def, sizeof(def));
723 	bzero(&mask, sizeof(mask));
724 	bzero(&gate, sizeof(gate));
725 
726 	def.sin6_len = mask.sin6_len = gate.sin6_len =
727 	    sizeof(struct sockaddr_in6);
728 	def.sin6_family = gate.sin6_family = AF_INET6;
729 	gate.sin6_addr = dr->rtaddr;
730 	fibnum = dr->ifp->if_fib;
731 
732 	bzero((caddr_t)&info, sizeof(info));
733 	info.rti_flags = RTF_GATEWAY;
734 	info.rti_info[RTAX_DST] = (struct sockaddr *)&def;
735 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gate;
736 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&mask;
737 
738 	NET_EPOCH_ENTER(et);
739 	error = rib_action(fibnum, RTM_DELETE, &info, &rc);
740 	if (error == 0) {
741 		struct nhop_object *nh = nhop_select_func(rc.rc_nh_old, 0);
742 		rt_routemsg(RTM_DELETE, rc.rc_rt, nh, fibnum);
743 	}
744 	NET_EPOCH_EXIT(et);
745 
746 	dr->installed = 0;
747 }
748 
749 static void
750 defrouter_del(struct nd_defrouter *dr)
751 {
752 	struct nd_defrouter *deldr = NULL;
753 	struct nd_prefix *pr;
754 	struct nd_pfxrouter *pfxrtr;
755 
756 	ND6_UNLOCK_ASSERT();
757 
758 	/*
759 	 * Flush all the routing table entries that use the router
760 	 * as a next hop.
761 	 */
762 	if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
763 		rt6_flush(&dr->rtaddr, dr->ifp);
764 
765 #ifdef EXPERIMENTAL
766 	defrtr_ipv6_only_ifp(dr->ifp);
767 #endif
768 
769 	if (dr->installed) {
770 		deldr = dr;
771 		defrouter_delreq(dr);
772 	}
773 
774 	/*
775 	 * Also delete all the pointers to the router in each prefix lists.
776 	 */
777 	ND6_WLOCK();
778 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
779 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
780 			pfxrtr_del(pfxrtr);
781 	}
782 	ND6_WUNLOCK();
783 
784 	pfxlist_onlink_check();
785 
786 	/*
787 	 * If the router is the primary one, choose a new one.
788 	 * Note that defrouter_select_fib() will remove the current
789          * gateway from the routing table.
790 	 */
791 	if (deldr)
792 		defrouter_select_fib(deldr->ifp->if_fib);
793 
794 	/*
795 	 * Release the list reference.
796 	 */
797 	defrouter_rele(dr);
798 }
799 
800 struct nd_defrouter *
801 defrouter_lookup_locked(const struct in6_addr *addr, struct ifnet *ifp)
802 {
803 	struct nd_defrouter *dr;
804 
805 	ND6_LOCK_ASSERT();
806 	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
807 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) {
808 			defrouter_ref(dr);
809 			return (dr);
810 		}
811 	return (NULL);
812 }
813 
814 struct nd_defrouter *
815 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
816 {
817 	struct nd_defrouter *dr;
818 
819 	ND6_RLOCK();
820 	dr = defrouter_lookup_locked(addr, ifp);
821 	ND6_RUNLOCK();
822 	return (dr);
823 }
824 
825 /*
826  * Remove all default routes from default router list.
827  */
828 void
829 defrouter_reset(void)
830 {
831 	struct nd_defrouter *dr, **dra;
832 	int count, i;
833 
834 	count = i = 0;
835 
836 	/*
837 	 * We can't delete routes with the ND lock held, so make a copy of the
838 	 * current default router list and use that when deleting routes.
839 	 */
840 	ND6_RLOCK();
841 	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry)
842 		count++;
843 	ND6_RUNLOCK();
844 
845 	dra = malloc(count * sizeof(*dra), M_TEMP, M_WAITOK | M_ZERO);
846 
847 	ND6_RLOCK();
848 	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
849 		if (i == count)
850 			break;
851 		defrouter_ref(dr);
852 		dra[i++] = dr;
853 	}
854 	ND6_RUNLOCK();
855 
856 	for (i = 0; i < count && dra[i] != NULL; i++) {
857 		defrouter_delreq(dra[i]);
858 		defrouter_rele(dra[i]);
859 	}
860 	free(dra, M_TEMP);
861 
862 	/*
863 	 * XXX should we also nuke any default routers in the kernel, by
864 	 * going through them by rtalloc1()?
865 	 */
866 }
867 
868 /*
869  * Look up a matching default router list entry and remove it. Returns true if a
870  * matching entry was found, false otherwise.
871  */
872 bool
873 defrouter_remove(struct in6_addr *addr, struct ifnet *ifp)
874 {
875 	struct nd_defrouter *dr;
876 
877 	ND6_WLOCK();
878 	dr = defrouter_lookup_locked(addr, ifp);
879 	if (dr == NULL) {
880 		ND6_WUNLOCK();
881 		return (false);
882 	}
883 
884 	defrouter_unlink(dr, NULL);
885 	ND6_WUNLOCK();
886 	defrouter_del(dr);
887 	defrouter_rele(dr);
888 	return (true);
889 }
890 
891 /*
892  * for default router selection
893  * regards router-preference field as a 2-bit signed integer
894  */
895 static int
896 rtpref(struct nd_defrouter *dr)
897 {
898 	switch (dr->raflags & ND_RA_FLAG_RTPREF_MASK) {
899 	case ND_RA_FLAG_RTPREF_HIGH:
900 		return (RTPREF_HIGH);
901 	case ND_RA_FLAG_RTPREF_MEDIUM:
902 	case ND_RA_FLAG_RTPREF_RSV:
903 		return (RTPREF_MEDIUM);
904 	case ND_RA_FLAG_RTPREF_LOW:
905 		return (RTPREF_LOW);
906 	default:
907 		/*
908 		 * This case should never happen.  If it did, it would mean a
909 		 * serious bug of kernel internal.  We thus always bark here.
910 		 * Or, can we even panic?
911 		 */
912 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->raflags);
913 		return (RTPREF_INVALID);
914 	}
915 	/* NOTREACHED */
916 }
917 
918 /*
919  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
920  * draft-ietf-ipngwg-router-selection:
921  * 1) Routers that are reachable or probably reachable should be preferred.
922  *    If we have more than one (probably) reachable router, prefer ones
923  *    with the highest router preference.
924  * 2) When no routers on the list are known to be reachable or
925  *    probably reachable, routers SHOULD be selected in a round-robin
926  *    fashion, regardless of router preference values.
927  * 3) If the Default Router List is empty, assume that all
928  *    destinations are on-link.
929  *
930  * We assume nd_defrouter is sorted by router preference value.
931  * Since the code below covers both with and without router preference cases,
932  * we do not need to classify the cases by ifdef.
933  *
934  * At this moment, we do not try to install more than one default router,
935  * even when the multipath routing is available, because we're not sure about
936  * the benefits for stub hosts comparing to the risk of making the code
937  * complicated and the possibility of introducing bugs.
938  *
939  * We maintain a single list of routers for multiple FIBs, only considering one
940  * at a time based on the receiving interface's FIB. If @fibnum is RT_ALL_FIBS,
941  * we do the whole thing multiple times.
942  */
943 void
944 defrouter_select_fib(int fibnum)
945 {
946 	struct epoch_tracker et;
947 	struct nd_defrouter *dr, *selected_dr, *installed_dr;
948 	struct llentry *ln = NULL;
949 
950 	if (fibnum == RT_ALL_FIBS) {
951 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
952 			defrouter_select_fib(fibnum);
953 		}
954 	}
955 
956 	ND6_RLOCK();
957 	/*
958 	 * Let's handle easy case (3) first:
959 	 * If default router list is empty, there's nothing to be done.
960 	 */
961 	if (TAILQ_EMPTY(&V_nd6_defrouter)) {
962 		ND6_RUNLOCK();
963 		return;
964 	}
965 
966 	/*
967 	 * Search for a (probably) reachable router from the list.
968 	 * We just pick up the first reachable one (if any), assuming that
969 	 * the ordering rule of the list described in defrtrlist_update().
970 	 */
971 	selected_dr = installed_dr = NULL;
972 	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
973 		NET_EPOCH_ENTER(et);
974 		if (selected_dr == NULL && dr->ifp->if_fib == fibnum &&
975 		    (ln = nd6_lookup(&dr->rtaddr, LLE_SF(AF_INET6, 0), dr->ifp)) &&
976 		    ND6_IS_LLINFO_PROBREACH(ln)) {
977 			selected_dr = dr;
978 			defrouter_ref(selected_dr);
979 		}
980 		NET_EPOCH_EXIT(et);
981 		if (ln != NULL) {
982 			LLE_RUNLOCK(ln);
983 			ln = NULL;
984 		}
985 
986 		if (dr->installed && dr->ifp->if_fib == fibnum) {
987 			if (installed_dr == NULL) {
988 				installed_dr = dr;
989 				defrouter_ref(installed_dr);
990 			} else {
991 				/*
992 				 * this should not happen.
993 				 * warn for diagnosis.
994 				 */
995 				log(LOG_ERR, "defrouter_select_fib: more than "
996 				             "one router is installed\n");
997 			}
998 		}
999 	}
1000 	/*
1001 	 * If none of the default routers was found to be reachable,
1002 	 * round-robin the list regardless of preference.
1003 	 * Otherwise, if we have an installed router, check if the selected
1004 	 * (reachable) router should really be preferred to the installed one.
1005 	 * We only prefer the new router when the old one is not reachable
1006 	 * or when the new one has a really higher preference value.
1007 	 */
1008 	if (selected_dr == NULL) {
1009 		if (installed_dr == NULL ||
1010 		    TAILQ_NEXT(installed_dr, dr_entry) == NULL)
1011 			dr = TAILQ_FIRST(&V_nd6_defrouter);
1012 		else
1013 			dr = TAILQ_NEXT(installed_dr, dr_entry);
1014 
1015 		/* Ensure we select a router for this FIB. */
1016 		TAILQ_FOREACH_FROM(dr, &V_nd6_defrouter, dr_entry) {
1017 			if (dr->ifp->if_fib == fibnum) {
1018 				selected_dr = dr;
1019 				defrouter_ref(selected_dr);
1020 				break;
1021 			}
1022 		}
1023 	} else if (installed_dr != NULL) {
1024 		NET_EPOCH_ENTER(et);
1025 		if ((ln = nd6_lookup(&installed_dr->rtaddr, 0,
1026 		                     installed_dr->ifp)) &&
1027 		    ND6_IS_LLINFO_PROBREACH(ln) &&
1028 		    installed_dr->ifp->if_fib == fibnum &&
1029 		    rtpref(selected_dr) <= rtpref(installed_dr)) {
1030 			defrouter_rele(selected_dr);
1031 			selected_dr = installed_dr;
1032 		}
1033 		NET_EPOCH_EXIT(et);
1034 		if (ln != NULL)
1035 			LLE_RUNLOCK(ln);
1036 	}
1037 	ND6_RUNLOCK();
1038 
1039 	NET_EPOCH_ENTER(et);
1040 	/*
1041 	 * If we selected a router for this FIB and it's different
1042 	 * than the installed one, remove the installed router and
1043 	 * install the selected one in its place.
1044 	 */
1045 	if (installed_dr != selected_dr) {
1046 		if (installed_dr != NULL) {
1047 			defrouter_delreq(installed_dr);
1048 			defrouter_rele(installed_dr);
1049 		}
1050 		if (selected_dr != NULL)
1051 			defrouter_addreq(selected_dr);
1052 	}
1053 	if (selected_dr != NULL)
1054 		defrouter_rele(selected_dr);
1055 	NET_EPOCH_EXIT(et);
1056 }
1057 
1058 static struct nd_defrouter *
1059 defrtrlist_update(struct nd_defrouter *new)
1060 {
1061 	struct nd_defrouter *dr, *n;
1062 	uint64_t genid;
1063 	int oldpref;
1064 	bool writelocked;
1065 
1066 	if (new->rtlifetime == 0) {
1067 		defrouter_remove(&new->rtaddr, new->ifp);
1068 		return (NULL);
1069 	}
1070 
1071 	ND6_RLOCK();
1072 	writelocked = false;
1073 restart:
1074 	dr = defrouter_lookup_locked(&new->rtaddr, new->ifp);
1075 	if (dr != NULL) {
1076 		oldpref = rtpref(dr);
1077 
1078 		/* override */
1079 		dr->raflags = new->raflags; /* XXX flag check */
1080 		dr->rtlifetime = new->rtlifetime;
1081 		dr->expire = new->expire;
1082 
1083 		/*
1084 		 * If the preference does not change, there's no need
1085 		 * to sort the entries. Also make sure the selected
1086 		 * router is still installed in the kernel.
1087 		 */
1088 		if (dr->installed && rtpref(new) == oldpref) {
1089 			if (writelocked)
1090 				ND6_WUNLOCK();
1091 			else
1092 				ND6_RUNLOCK();
1093 			return (dr);
1094 		}
1095 	}
1096 
1097 	/*
1098 	 * The router needs to be reinserted into the default router
1099 	 * list, so upgrade to a write lock. If that fails and the list
1100 	 * has potentially changed while the lock was dropped, we'll
1101 	 * redo the lookup with the write lock held.
1102 	 */
1103 	if (!writelocked) {
1104 		writelocked = true;
1105 		if (!ND6_TRY_UPGRADE()) {
1106 			genid = V_nd6_list_genid;
1107 			ND6_RUNLOCK();
1108 			ND6_WLOCK();
1109 			if (genid != V_nd6_list_genid)
1110 				goto restart;
1111 		}
1112 	}
1113 
1114 	if (dr != NULL) {
1115 		/*
1116 		 * The preferred router may have changed, so relocate this
1117 		 * router.
1118 		 */
1119 		TAILQ_REMOVE(&V_nd6_defrouter, dr, dr_entry);
1120 		n = dr;
1121 	} else {
1122 		n = malloc(sizeof(*n), M_IP6NDP, M_NOWAIT | M_ZERO);
1123 		if (n == NULL) {
1124 			ND6_WUNLOCK();
1125 			return (NULL);
1126 		}
1127 		memcpy(n, new, sizeof(*n));
1128 		/* Initialize with an extra reference for the caller. */
1129 		refcount_init(&n->refcnt, 2);
1130 	}
1131 
1132 	/*
1133 	 * Insert the new router in the Default Router List;
1134 	 * The Default Router List should be in the descending order
1135 	 * of router-preferece.  Routers with the same preference are
1136 	 * sorted in the arriving time order.
1137 	 */
1138 
1139 	/* insert at the end of the group */
1140 	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
1141 		if (rtpref(n) > rtpref(dr))
1142 			break;
1143 	}
1144 	if (dr != NULL)
1145 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
1146 	else
1147 		TAILQ_INSERT_TAIL(&V_nd6_defrouter, n, dr_entry);
1148 	V_nd6_list_genid++;
1149 	ND6_WUNLOCK();
1150 
1151 	defrouter_select_fib(new->ifp->if_fib);
1152 
1153 	return (n);
1154 }
1155 
1156 static int
1157 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
1158 {
1159 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
1160 		ndpr->ndpr_preferred = 0;
1161 	else
1162 		ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
1163 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
1164 		ndpr->ndpr_expire = 0;
1165 	else
1166 		ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
1167 
1168 	return 0;
1169 }
1170 
1171 static void
1172 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
1173 {
1174 	/* init ia6t_expire */
1175 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
1176 		lt6->ia6t_expire = 0;
1177 	else {
1178 		lt6->ia6t_expire = time_uptime;
1179 		lt6->ia6t_expire += lt6->ia6t_vltime;
1180 	}
1181 
1182 	/* init ia6t_preferred */
1183 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
1184 		lt6->ia6t_preferred = 0;
1185 	else {
1186 		lt6->ia6t_preferred = time_uptime;
1187 		lt6->ia6t_preferred += lt6->ia6t_pltime;
1188 	}
1189 }
1190 
1191 static struct in6_ifaddr *
1192 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1193 {
1194 	struct ifnet *ifp = pr->ndpr_ifp;
1195 	struct ifaddr *ifa;
1196 	struct in6_aliasreq ifra;
1197 	struct in6_ifaddr *ia, *ib;
1198 	int error, plen0;
1199 	struct in6_addr mask;
1200 	int prefixlen = pr->ndpr_plen;
1201 	int updateflags;
1202 	char ip6buf[INET6_ADDRSTRLEN];
1203 
1204 	in6_prefixlen2mask(&mask, prefixlen);
1205 
1206 	/*
1207 	 * find a link-local address (will be interface ID).
1208 	 * Is it really mandatory? Theoretically, a global or a site-local
1209 	 * address can be configured without a link-local address, if we
1210 	 * have a unique interface identifier...
1211 	 *
1212 	 * it is not mandatory to have a link-local address, we can generate
1213 	 * interface identifier on the fly.  we do this because:
1214 	 * (1) it should be the easiest way to find interface identifier.
1215 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1216 	 * for multiple addresses on a single interface, and possible shortcut
1217 	 * of DAD.  we omitted DAD for this reason in the past.
1218 	 * (3) a user can prevent autoconfiguration of global address
1219 	 * by removing link-local address by hand (this is partly because we
1220 	 * don't have other way to control the use of IPv6 on an interface.
1221 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1222 	 * (4) it is easier to manage when an interface has addresses
1223 	 * with the same interface identifier, than to have multiple addresses
1224 	 * with different interface identifiers.
1225 	 */
1226 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1227 	if (ifa)
1228 		ib = (struct in6_ifaddr *)ifa;
1229 	else
1230 		return NULL;
1231 
1232 	/* prefixlen + ifidlen must be equal to 128 */
1233 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1234 	if (prefixlen != plen0) {
1235 		ifa_free(ifa);
1236 		nd6log((LOG_INFO,
1237 		    "%s: wrong prefixlen for %s (prefix=%d ifid=%d)\n",
1238 		    __func__, if_name(ifp), prefixlen, 128 - plen0));
1239 		return NULL;
1240 	}
1241 
1242 	/* make ifaddr */
1243 	in6_prepare_ifra(&ifra, &pr->ndpr_prefix.sin6_addr, &mask);
1244 
1245 	IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, &mask);
1246 	/* interface ID */
1247 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1248 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1249 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1250 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1251 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1252 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1253 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1254 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1255 	ifa_free(ifa);
1256 
1257 	/* lifetimes. */
1258 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1259 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1260 
1261 	/* XXX: scope zone ID? */
1262 
1263 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1264 
1265 	/*
1266 	 * Make sure that we do not have this address already.  This should
1267 	 * usually not happen, but we can still see this case, e.g., if we
1268 	 * have manually configured the exact address to be configured.
1269 	 */
1270 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1271 	    &ifra.ifra_addr.sin6_addr);
1272 	if (ifa != NULL) {
1273 		ifa_free(ifa);
1274 		/* this should be rare enough to make an explicit log */
1275 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1276 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1277 		return (NULL);
1278 	}
1279 
1280 	/*
1281 	 * Allocate ifaddr structure, link into chain, etc.
1282 	 * If we are going to create a new address upon receiving a multicasted
1283 	 * RA, we need to impose a random delay before starting DAD.
1284 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1285 	 */
1286 	updateflags = 0;
1287 	if (mcast)
1288 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1289 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1290 		nd6log((LOG_ERR,
1291 		    "%s: failed to make ifaddr %s on %s (errno=%d)\n", __func__,
1292 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1293 		    if_name(ifp), error));
1294 		return (NULL);	/* ifaddr must not have been allocated. */
1295 	}
1296 
1297 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1298 	/*
1299 	 * XXXRW: Assumption of non-NULLness here might not be true with
1300 	 * fine-grained locking -- should we validate it?  Or just return
1301 	 * earlier ifa rather than looking it up again?
1302 	 */
1303 	return (ia);		/* this is always non-NULL  and referenced. */
1304 }
1305 
1306 static struct nd_prefix *
1307 nd6_prefix_lookup_locked(struct nd_prefixctl *key)
1308 {
1309 	struct nd_prefix *search;
1310 
1311 	ND6_LOCK_ASSERT();
1312 
1313 	LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
1314 		if (key->ndpr_ifp == search->ndpr_ifp &&
1315 		    key->ndpr_plen == search->ndpr_plen &&
1316 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
1317 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
1318 			nd6_prefix_ref(search);
1319 			break;
1320 		}
1321 	}
1322 	return (search);
1323 }
1324 
1325 struct nd_prefix *
1326 nd6_prefix_lookup(struct nd_prefixctl *key)
1327 {
1328 	struct nd_prefix *search;
1329 
1330 	ND6_RLOCK();
1331 	search = nd6_prefix_lookup_locked(key);
1332 	ND6_RUNLOCK();
1333 	return (search);
1334 }
1335 
1336 void
1337 nd6_prefix_ref(struct nd_prefix *pr)
1338 {
1339 
1340 	refcount_acquire(&pr->ndpr_refcnt);
1341 }
1342 
1343 void
1344 nd6_prefix_rele(struct nd_prefix *pr)
1345 {
1346 
1347 	if (refcount_release(&pr->ndpr_refcnt)) {
1348 		KASSERT(LIST_EMPTY(&pr->ndpr_advrtrs),
1349 		    ("prefix %p has advertising routers", pr));
1350 		free(pr, M_IP6NDP);
1351 	}
1352 }
1353 
1354 int
1355 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
1356     struct nd_prefix **newp)
1357 {
1358 	struct nd_prefix *new;
1359 	char ip6buf[INET6_ADDRSTRLEN];
1360 	int error;
1361 
1362 	new = malloc(sizeof(*new), M_IP6NDP, M_NOWAIT | M_ZERO);
1363 	if (new == NULL)
1364 		return (ENOMEM);
1365 	refcount_init(&new->ndpr_refcnt, newp != NULL ? 2 : 1);
1366 	new->ndpr_ifp = pr->ndpr_ifp;
1367 	new->ndpr_prefix = pr->ndpr_prefix;
1368 	new->ndpr_plen = pr->ndpr_plen;
1369 	new->ndpr_vltime = pr->ndpr_vltime;
1370 	new->ndpr_pltime = pr->ndpr_pltime;
1371 	new->ndpr_flags = pr->ndpr_flags;
1372 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
1373 		free(new, M_IP6NDP);
1374 		return (error);
1375 	}
1376 	new->ndpr_lastupdate = time_uptime;
1377 
1378 	/* initialization */
1379 	LIST_INIT(&new->ndpr_advrtrs);
1380 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
1381 	/* make prefix in the canonical form */
1382 	IN6_MASK_ADDR(&new->ndpr_prefix.sin6_addr, &new->ndpr_mask);
1383 
1384 	ND6_WLOCK();
1385 	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
1386 	V_nd6_list_genid++;
1387 	ND6_WUNLOCK();
1388 
1389 	/* ND_OPT_PI_FLAG_ONLINK processing */
1390 	if (new->ndpr_raf_onlink) {
1391 		struct epoch_tracker et;
1392 
1393 		ND6_ONLINK_LOCK();
1394 		NET_EPOCH_ENTER(et);
1395 		if ((error = nd6_prefix_onlink(new)) != 0) {
1396 			nd6log((LOG_ERR, "%s: failed to make the prefix %s/%d "
1397 			    "on-link on %s (errno=%d)\n", __func__,
1398 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1399 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), error));
1400 			/* proceed anyway. XXX: is it correct? */
1401 		}
1402 		NET_EPOCH_EXIT(et);
1403 		ND6_ONLINK_UNLOCK();
1404 	}
1405 
1406 	if (dr != NULL)
1407 		pfxrtr_add(new, dr);
1408 	if (newp != NULL)
1409 		*newp = new;
1410 	return (0);
1411 }
1412 
1413 /*
1414  * Remove a prefix from the prefix list and optionally stash it in a
1415  * caller-provided list.
1416  *
1417  * The ND6 lock must be held.
1418  */
1419 void
1420 nd6_prefix_unlink(struct nd_prefix *pr, struct nd_prhead *list)
1421 {
1422 
1423 	ND6_WLOCK_ASSERT();
1424 
1425 	LIST_REMOVE(pr, ndpr_entry);
1426 	V_nd6_list_genid++;
1427 	if (list != NULL)
1428 		LIST_INSERT_HEAD(list, pr, ndpr_entry);
1429 }
1430 
1431 /*
1432  * Free an unlinked prefix, first marking it off-link if necessary.
1433  */
1434 void
1435 nd6_prefix_del(struct nd_prefix *pr)
1436 {
1437 	struct nd_pfxrouter *pfr, *next;
1438 	int e;
1439 	char ip6buf[INET6_ADDRSTRLEN];
1440 
1441 	KASSERT(pr->ndpr_addrcnt == 0,
1442 	    ("prefix %p has referencing addresses", pr));
1443 	ND6_UNLOCK_ASSERT();
1444 
1445 	/*
1446 	 * Though these flags are now meaningless, we'd rather keep the value
1447 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
1448 	 * when executing "ndp -p".
1449 	 */
1450 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1451 		ND6_ONLINK_LOCK();
1452 		if ((e = nd6_prefix_offlink(pr)) != 0) {
1453 			nd6log((LOG_ERR,
1454 			    "%s: failed to make the prefix %s/%d offlink on %s "
1455 			    "(errno=%d)\n", __func__,
1456 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1457 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1458 			/* what should we do? */
1459 		}
1460 		ND6_ONLINK_UNLOCK();
1461 	}
1462 
1463 	/* Release references to routers that have advertised this prefix. */
1464 	ND6_WLOCK();
1465 	LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next)
1466 		pfxrtr_del(pfr);
1467 	ND6_WUNLOCK();
1468 
1469 	nd6_prefix_rele(pr);
1470 
1471 	pfxlist_onlink_check();
1472 }
1473 
1474 static int
1475 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
1476     struct mbuf *m, int mcast)
1477 {
1478 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
1479 	struct ifaddr *ifa;
1480 	struct ifnet *ifp = new->ndpr_ifp;
1481 	struct nd_prefix *pr;
1482 	int error = 0;
1483 	int auth;
1484 	struct in6_addrlifetime lt6_tmp;
1485 	char ip6buf[INET6_ADDRSTRLEN];
1486 
1487 	NET_EPOCH_ASSERT();
1488 
1489 	auth = 0;
1490 	if (m) {
1491 		/*
1492 		 * Authenticity for NA consists authentication for
1493 		 * both IP header and IP datagrams, doesn't it ?
1494 		 */
1495 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
1496 		auth = ((m->m_flags & M_AUTHIPHDR) &&
1497 		    (m->m_flags & M_AUTHIPDGM));
1498 #endif
1499 	}
1500 
1501 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
1502 		/*
1503 		 * nd6_prefix_lookup() ensures that pr and new have the same
1504 		 * prefix on a same interface.
1505 		 */
1506 
1507 		/*
1508 		 * Update prefix information.  Note that the on-link (L) bit
1509 		 * and the autonomous (A) bit should NOT be changed from 1
1510 		 * to 0.
1511 		 */
1512 		if (new->ndpr_raf_onlink == 1)
1513 			pr->ndpr_raf_onlink = 1;
1514 		if (new->ndpr_raf_auto == 1)
1515 			pr->ndpr_raf_auto = 1;
1516 		if (new->ndpr_raf_onlink) {
1517 			pr->ndpr_vltime = new->ndpr_vltime;
1518 			pr->ndpr_pltime = new->ndpr_pltime;
1519 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1520 			pr->ndpr_lastupdate = time_uptime;
1521 		}
1522 
1523 		if (new->ndpr_raf_onlink &&
1524 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1525 			ND6_ONLINK_LOCK();
1526 			if ((error = nd6_prefix_onlink(pr)) != 0) {
1527 				nd6log((LOG_ERR,
1528 				    "%s: failed to make the prefix %s/%d "
1529 				    "on-link on %s (errno=%d)\n", __func__,
1530 				    ip6_sprintf(ip6buf,
1531 				        &pr->ndpr_prefix.sin6_addr),
1532 				    pr->ndpr_plen, if_name(pr->ndpr_ifp),
1533 				    error));
1534 				/* proceed anyway. XXX: is it correct? */
1535 			}
1536 			ND6_ONLINK_UNLOCK();
1537 		}
1538 
1539 		if (dr != NULL)
1540 			pfxrtr_add(pr, dr);
1541 	} else {
1542 		if (new->ndpr_vltime == 0)
1543 			goto end;
1544 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1545 			goto end;
1546 
1547 		error = nd6_prelist_add(new, dr, &pr);
1548 		if (error != 0) {
1549 			nd6log((LOG_NOTICE, "%s: nd6_prelist_add() failed for "
1550 			    "the prefix %s/%d on %s (errno=%d)\n", __func__,
1551 			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1552 			    new->ndpr_plen, if_name(new->ndpr_ifp), error));
1553 			goto end; /* we should just give up in this case. */
1554 		}
1555 
1556 		/*
1557 		 * XXX: from the ND point of view, we can ignore a prefix
1558 		 * with the on-link bit being zero.  However, we need a
1559 		 * prefix structure for references from autoconfigured
1560 		 * addresses.  Thus, we explicitly make sure that the prefix
1561 		 * itself expires now.
1562 		 */
1563 		if (pr->ndpr_raf_onlink == 0) {
1564 			pr->ndpr_vltime = 0;
1565 			pr->ndpr_pltime = 0;
1566 			in6_init_prefix_ltimes(pr);
1567 		}
1568 	}
1569 
1570 	/*
1571 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1572 	 * Note that pr must be non NULL at this point.
1573 	 */
1574 
1575 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1576 	if (!new->ndpr_raf_auto)
1577 		goto end;
1578 
1579 	/*
1580 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1581 	 * nd6_ra_input.
1582 	 */
1583 
1584 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1585 	if (new->ndpr_pltime > new->ndpr_vltime) {
1586 		error = EINVAL;	/* XXX: won't be used */
1587 		goto end;
1588 	}
1589 
1590 	/*
1591 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1592 	 * an address configured by stateless autoconfiguration already in the
1593 	 * list of addresses associated with the interface, and the Valid
1594 	 * Lifetime is not 0, form an address.  We first check if we have
1595 	 * a matching prefix.
1596 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1597 	 * consider autoconfigured addresses while RFC2462 simply said
1598 	 * "address".
1599 	 */
1600 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1601 		struct in6_ifaddr *ifa6;
1602 		u_int32_t remaininglifetime;
1603 
1604 		if (ifa->ifa_addr->sa_family != AF_INET6)
1605 			continue;
1606 
1607 		ifa6 = (struct in6_ifaddr *)ifa;
1608 
1609 		/*
1610 		 * We only consider autoconfigured addresses as per rfc2462bis.
1611 		 */
1612 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1613 			continue;
1614 
1615 		/*
1616 		 * Spec is not clear here, but I believe we should concentrate
1617 		 * on unicast (i.e. not anycast) addresses.
1618 		 * XXX: other ia6_flags? detached or duplicated?
1619 		 */
1620 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1621 			continue;
1622 
1623 		/*
1624 		 * Ignore the address if it is not associated with a prefix
1625 		 * or is associated with a prefix that is different from this
1626 		 * one.  (pr is never NULL here)
1627 		 */
1628 		if (ifa6->ia6_ndpr != pr)
1629 			continue;
1630 
1631 		if (ia6_match == NULL) /* remember the first one */
1632 			ia6_match = ifa6;
1633 
1634 		/*
1635 		 * An already autoconfigured address matched.  Now that we
1636 		 * are sure there is at least one matched address, we can
1637 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1638 		 * "two hours" rule and the privacy extension.
1639 		 * We apply some clarifications in rfc2462bis:
1640 		 * - use remaininglifetime instead of storedlifetime as a
1641 		 *   variable name
1642 		 * - remove the dead code in the "two-hour" rule
1643 		 */
1644 #define TWOHOUR		(120*60)
1645 		lt6_tmp = ifa6->ia6_lifetime;
1646 
1647 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1648 			remaininglifetime = ND6_INFINITE_LIFETIME;
1649 		else if (time_uptime - ifa6->ia6_updatetime >
1650 			 lt6_tmp.ia6t_vltime) {
1651 			/*
1652 			 * The case of "invalid" address.  We should usually
1653 			 * not see this case.
1654 			 */
1655 			remaininglifetime = 0;
1656 		} else
1657 			remaininglifetime = lt6_tmp.ia6t_vltime -
1658 			    (time_uptime - ifa6->ia6_updatetime);
1659 
1660 		/* when not updating, keep the current stored lifetime. */
1661 		lt6_tmp.ia6t_vltime = remaininglifetime;
1662 
1663 		if (TWOHOUR < new->ndpr_vltime ||
1664 		    remaininglifetime < new->ndpr_vltime) {
1665 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1666 		} else if (remaininglifetime <= TWOHOUR) {
1667 			if (auth) {
1668 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1669 			}
1670 		} else {
1671 			/*
1672 			 * new->ndpr_vltime <= TWOHOUR &&
1673 			 * TWOHOUR < remaininglifetime
1674 			 */
1675 			lt6_tmp.ia6t_vltime = TWOHOUR;
1676 		}
1677 
1678 		/* The 2 hour rule is not imposed for preferred lifetime. */
1679 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1680 
1681 		in6_init_address_ltimes(pr, &lt6_tmp);
1682 
1683 		/*
1684 		 * We need to treat lifetimes for temporary addresses
1685 		 * differently, according to
1686 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1687 		 * we only update the lifetimes when they are in the maximum
1688 		 * intervals.
1689 		 */
1690 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1691 			u_int32_t maxvltime, maxpltime;
1692 
1693 			if (V_ip6_temp_valid_lifetime >
1694 			    (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1695 			    V_ip6_desync_factor)) {
1696 				maxvltime = V_ip6_temp_valid_lifetime -
1697 				    (time_uptime - ifa6->ia6_createtime) -
1698 				    V_ip6_desync_factor;
1699 			} else
1700 				maxvltime = 0;
1701 			if (V_ip6_temp_preferred_lifetime >
1702 			    (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1703 			    V_ip6_desync_factor)) {
1704 				maxpltime = V_ip6_temp_preferred_lifetime -
1705 				    (time_uptime - ifa6->ia6_createtime) -
1706 				    V_ip6_desync_factor;
1707 			} else
1708 				maxpltime = 0;
1709 
1710 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1711 			    lt6_tmp.ia6t_vltime > maxvltime) {
1712 				lt6_tmp.ia6t_vltime = maxvltime;
1713 			}
1714 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1715 			    lt6_tmp.ia6t_pltime > maxpltime) {
1716 				lt6_tmp.ia6t_pltime = maxpltime;
1717 			}
1718 		}
1719 		ifa6->ia6_lifetime = lt6_tmp;
1720 		ifa6->ia6_updatetime = time_uptime;
1721 	}
1722 	if (ia6_match == NULL && new->ndpr_vltime) {
1723 		int ifidlen;
1724 
1725 		/*
1726 		 * 5.5.3 (d) (continued)
1727 		 * No address matched and the valid lifetime is non-zero.
1728 		 * Create a new address.
1729 		 */
1730 
1731 		/*
1732 		 * Prefix Length check:
1733 		 * If the sum of the prefix length and interface identifier
1734 		 * length does not equal 128 bits, the Prefix Information
1735 		 * option MUST be ignored.  The length of the interface
1736 		 * identifier is defined in a separate link-type specific
1737 		 * document.
1738 		 */
1739 		ifidlen = in6_if2idlen(ifp);
1740 		if (ifidlen < 0) {
1741 			/* this should not happen, so we always log it. */
1742 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1743 			    if_name(ifp));
1744 			goto end;
1745 		}
1746 		if (ifidlen + pr->ndpr_plen != 128) {
1747 			nd6log((LOG_INFO,
1748 			    "%s: invalid prefixlen %d for %s, ignored\n",
1749 			    __func__, pr->ndpr_plen, if_name(ifp)));
1750 			goto end;
1751 		}
1752 
1753 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1754 			/*
1755 			 * note that we should use pr (not new) for reference.
1756 			 */
1757 			pr->ndpr_addrcnt++;
1758 			ia6->ia6_ndpr = pr;
1759 
1760 			/*
1761 			 * RFC 3041 3.3 (2).
1762 			 * When a new public address is created as described
1763 			 * in RFC2462, also create a new temporary address.
1764 			 *
1765 			 * RFC 3041 3.5.
1766 			 * When an interface connects to a new link, a new
1767 			 * randomized interface identifier should be generated
1768 			 * immediately together with a new set of temporary
1769 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1770 			 * in6_tmpifadd().
1771 			 */
1772 			if (V_ip6_use_tempaddr) {
1773 				int e;
1774 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1775 					nd6log((LOG_NOTICE, "%s: failed to "
1776 					    "create a temporary address "
1777 					    "(errno=%d)\n", __func__, e));
1778 				}
1779 			}
1780 			ifa_free(&ia6->ia_ifa);
1781 
1782 			/*
1783 			 * A newly added address might affect the status
1784 			 * of other addresses, so we check and update it.
1785 			 * XXX: what if address duplication happens?
1786 			 */
1787 			pfxlist_onlink_check();
1788 		} else {
1789 			/* just set an error. do not bark here. */
1790 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1791 		}
1792 	}
1793 
1794 end:
1795 	if (pr != NULL)
1796 		nd6_prefix_rele(pr);
1797 	return (error);
1798 }
1799 
1800 /*
1801  * A supplement function used in the on-link detection below;
1802  * detect if a given prefix has a (probably) reachable advertising router.
1803  * XXX: lengthy function name...
1804  */
1805 static struct nd_pfxrouter *
1806 find_pfxlist_reachable_router(struct nd_prefix *pr)
1807 {
1808 	struct epoch_tracker et;
1809 	struct nd_pfxrouter *pfxrtr;
1810 	struct llentry *ln;
1811 	int canreach;
1812 
1813 	ND6_LOCK_ASSERT();
1814 
1815 	NET_EPOCH_ENTER(et);
1816 	LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1817 		ln = nd6_lookup(&pfxrtr->router->rtaddr, LLE_SF(AF_INET6, 0),
1818 		    pfxrtr->router->ifp);
1819 		if (ln == NULL)
1820 			continue;
1821 		canreach = ND6_IS_LLINFO_PROBREACH(ln);
1822 		LLE_RUNLOCK(ln);
1823 		if (canreach)
1824 			break;
1825 	}
1826 	NET_EPOCH_EXIT(et);
1827 	return (pfxrtr);
1828 }
1829 
1830 /*
1831  * Check if each prefix in the prefix list has at least one available router
1832  * that advertised the prefix (a router is "available" if its neighbor cache
1833  * entry is reachable or probably reachable).
1834  * If the check fails, the prefix may be off-link, because, for example,
1835  * we have moved from the network but the lifetime of the prefix has not
1836  * expired yet.  So we should not use the prefix if there is another prefix
1837  * that has an available router.
1838  * But, if there is no prefix that has an available router, we still regard
1839  * all the prefixes as on-link.  This is because we can't tell if all the
1840  * routers are simply dead or if we really moved from the network and there
1841  * is no router around us.
1842  */
1843 void
1844 pfxlist_onlink_check(void)
1845 {
1846 	struct nd_prefix *pr;
1847 	struct in6_ifaddr *ifa;
1848 	struct nd_defrouter *dr;
1849 	struct nd_pfxrouter *pfxrtr = NULL;
1850 	struct rm_priotracker in6_ifa_tracker;
1851 	uint64_t genid;
1852 	uint32_t flags;
1853 
1854 	ND6_ONLINK_LOCK();
1855 	ND6_RLOCK();
1856 
1857 	/*
1858 	 * Check if there is a prefix that has a reachable advertising
1859 	 * router.
1860 	 */
1861 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1862 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1863 			break;
1864 	}
1865 
1866 	/*
1867 	 * If we have no such prefix, check whether we still have a router
1868 	 * that does not advertise any prefixes.
1869 	 */
1870 	if (pr == NULL) {
1871 		TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
1872 			struct nd_prefix *pr0;
1873 
1874 			LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1875 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1876 					break;
1877 			}
1878 			if (pfxrtr != NULL)
1879 				break;
1880 		}
1881 	}
1882 	if (pr != NULL || (!TAILQ_EMPTY(&V_nd6_defrouter) && pfxrtr == NULL)) {
1883 		/*
1884 		 * There is at least one prefix that has a reachable router,
1885 		 * or at least a router which probably does not advertise
1886 		 * any prefixes.  The latter would be the case when we move
1887 		 * to a new link where we have a router that does not provide
1888 		 * prefixes and we configure an address by hand.
1889 		 * Detach prefixes which have no reachable advertising
1890 		 * router, and attach other prefixes.
1891 		 */
1892 		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1893 			/* XXX: a link-local prefix should never be detached */
1894 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1895 			    pr->ndpr_raf_onlink == 0 ||
1896 			    pr->ndpr_raf_auto == 0)
1897 				continue;
1898 
1899 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1900 			    find_pfxlist_reachable_router(pr) == NULL)
1901 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1902 			else if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1903 			    find_pfxlist_reachable_router(pr) != NULL)
1904 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1905 		}
1906 	} else {
1907 		/* there is no prefix that has a reachable router */
1908 		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1909 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1910 			    pr->ndpr_raf_onlink == 0 ||
1911 			    pr->ndpr_raf_auto == 0)
1912 				continue;
1913 			pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1914 		}
1915 	}
1916 
1917 	/*
1918 	 * Remove each interface route associated with a (just) detached
1919 	 * prefix, and reinstall the interface route for a (just) attached
1920 	 * prefix.  Note that all attempt of reinstallation does not
1921 	 * necessarily success, when a same prefix is shared among multiple
1922 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1923 	 * so we don't have to care about them.
1924 	 */
1925 restart:
1926 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1927 		char ip6buf[INET6_ADDRSTRLEN];
1928 		int e;
1929 
1930 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
1931 		    pr->ndpr_raf_onlink == 0 ||
1932 		    pr->ndpr_raf_auto == 0)
1933 			continue;
1934 
1935 		flags = pr->ndpr_stateflags & (NDPRF_DETACHED | NDPRF_ONLINK);
1936 		if (flags == 0 || flags == (NDPRF_DETACHED | NDPRF_ONLINK)) {
1937 			genid = V_nd6_list_genid;
1938 			ND6_RUNLOCK();
1939 			if ((flags & NDPRF_ONLINK) != 0 &&
1940 			    (e = nd6_prefix_offlink(pr)) != 0) {
1941 				nd6log((LOG_ERR,
1942 				    "%s: failed to make %s/%d offlink "
1943 				    "(errno=%d)\n", __func__,
1944 				    ip6_sprintf(ip6buf,
1945 					    &pr->ndpr_prefix.sin6_addr),
1946 					    pr->ndpr_plen, e));
1947 			} else if ((flags & NDPRF_ONLINK) == 0 &&
1948 			    (e = nd6_prefix_onlink(pr)) != 0) {
1949 				nd6log((LOG_ERR,
1950 				    "%s: failed to make %s/%d onlink "
1951 				    "(errno=%d)\n", __func__,
1952 				    ip6_sprintf(ip6buf,
1953 					    &pr->ndpr_prefix.sin6_addr),
1954 					    pr->ndpr_plen, e));
1955 			}
1956 			ND6_RLOCK();
1957 			if (genid != V_nd6_list_genid)
1958 				goto restart;
1959 		}
1960 	}
1961 
1962 	/*
1963 	 * Changes on the prefix status might affect address status as well.
1964 	 * Make sure that all addresses derived from an attached prefix are
1965 	 * attached, and that all addresses derived from a detached prefix are
1966 	 * detached.  Note, however, that a manually configured address should
1967 	 * always be attached.
1968 	 * The precise detection logic is same as the one for prefixes.
1969 	 */
1970 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1971 	CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1972 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1973 			continue;
1974 
1975 		if (ifa->ia6_ndpr == NULL) {
1976 			/*
1977 			 * This can happen when we first configure the address
1978 			 * (i.e. the address exists, but the prefix does not).
1979 			 * XXX: complicated relationships...
1980 			 */
1981 			continue;
1982 		}
1983 
1984 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1985 			break;
1986 	}
1987 	if (ifa) {
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_ndpr == NULL) /* XXX: see above. */
1993 				continue;
1994 
1995 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1996 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1997 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1998 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1999 					nd6_dad_start((struct ifaddr *)ifa, 0);
2000 				}
2001 			} else {
2002 				ifa->ia6_flags |= IN6_IFF_DETACHED;
2003 			}
2004 		}
2005 	} else {
2006 		CK_STAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
2007 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
2008 				continue;
2009 
2010 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
2011 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
2012 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
2013 				/* Do we need a delay in this case? */
2014 				nd6_dad_start((struct ifaddr *)ifa, 0);
2015 			}
2016 		}
2017 	}
2018 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
2019 	ND6_RUNLOCK();
2020 	ND6_ONLINK_UNLOCK();
2021 }
2022 
2023 /*
2024  * Add or remove interface route specified by @dst, @netmask and @ifp.
2025  * ifa can be NULL.
2026  * Returns 0 on success
2027  */
2028 static int
2029 nd6_prefix_rtrequest(uint32_t fibnum, int cmd, struct sockaddr_in6 *dst,
2030     struct sockaddr_in6 *netmask, struct ifnet *ifp, struct ifaddr *ifa)
2031 {
2032 	struct epoch_tracker et;
2033 	int error;
2034 
2035 	/* Prepare gateway */
2036 	struct sockaddr_dl_short sdl = {
2037 		.sdl_family = AF_LINK,
2038 		.sdl_len = sizeof(struct sockaddr_dl_short),
2039 		.sdl_type = ifp->if_type,
2040 		.sdl_index = ifp->if_index,
2041 	};
2042 
2043 	struct rt_addrinfo info = {
2044 		.rti_ifa = ifa,
2045 		.rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
2046 		.rti_info = {
2047 			[RTAX_DST] = (struct sockaddr *)dst,
2048 			[RTAX_NETMASK] = (struct sockaddr *)netmask,
2049 			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
2050 		},
2051 	};
2052 	/* Don't set additional per-gw filters on removal */
2053 
2054 	NET_EPOCH_ENTER(et);
2055 	error = rib_handle_ifaddr_info(fibnum, cmd, &info);
2056 	NET_EPOCH_EXIT(et);
2057 	return (error);
2058 }
2059 
2060 static int
2061 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
2062 {
2063 	int error;
2064 
2065 	struct sockaddr_in6 mask6 = {
2066 		.sin6_family = AF_INET6,
2067 		.sin6_len = sizeof(struct sockaddr_in6),
2068 		.sin6_addr = pr->ndpr_mask,
2069 	};
2070 	struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL;
2071 
2072 	error = nd6_prefix_rtrequest(pr->ndpr_ifp->if_fib, RTM_ADD,
2073 	    &pr->ndpr_prefix, pmask6, pr->ndpr_ifp, ifa);
2074 	if (error == 0)
2075 		pr->ndpr_stateflags |= NDPRF_ONLINK;
2076 
2077 	return (error);
2078 }
2079 
2080 static int
2081 nd6_prefix_onlink(struct nd_prefix *pr)
2082 {
2083 	struct epoch_tracker et;
2084 	struct ifaddr *ifa;
2085 	struct ifnet *ifp = pr->ndpr_ifp;
2086 	struct nd_prefix *opr;
2087 	char ip6buf[INET6_ADDRSTRLEN];
2088 	int error;
2089 
2090 	ND6_ONLINK_LOCK_ASSERT();
2091 	ND6_UNLOCK_ASSERT();
2092 
2093 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0)
2094 		return (EEXIST);
2095 
2096 	/*
2097 	 * Add the interface route associated with the prefix.  Before
2098 	 * installing the route, check if there's the same prefix on another
2099 	 * interface, and the prefix has already installed the interface route.
2100 	 * Although such a configuration is expected to be rare, we explicitly
2101 	 * allow it.
2102 	 */
2103 	ND6_RLOCK();
2104 	LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2105 		if (opr == pr)
2106 			continue;
2107 
2108 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
2109 			continue;
2110 
2111 		if (!V_rt_add_addr_allfibs &&
2112 		    opr->ndpr_ifp->if_fib != pr->ndpr_ifp->if_fib)
2113 			continue;
2114 
2115 		if (opr->ndpr_plen == pr->ndpr_plen &&
2116 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2117 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2118 			ND6_RUNLOCK();
2119 			return (0);
2120 		}
2121 	}
2122 	ND6_RUNLOCK();
2123 
2124 	/*
2125 	 * We prefer link-local addresses as the associated interface address.
2126 	 */
2127 	/* search for a link-local addr */
2128 	NET_EPOCH_ENTER(et);
2129 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
2130 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
2131 	if (ifa == NULL) {
2132 		/* XXX: freebsd does not have ifa_ifwithaf */
2133 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2134 			if (ifa->ifa_addr->sa_family == AF_INET6) {
2135 				ifa_ref(ifa);
2136 				break;
2137 			}
2138 		}
2139 		/* should we care about ia6_flags? */
2140 	}
2141 	if (ifa == NULL) {
2142 		/*
2143 		 * This can still happen, when, for example, we receive an RA
2144 		 * containing a prefix with the L bit set and the A bit clear,
2145 		 * after removing all IPv6 addresses on the receiving
2146 		 * interface.  This should, of course, be rare though.
2147 		 */
2148 		nd6log((LOG_NOTICE,
2149 		    "%s: failed to find any ifaddr to add route for a "
2150 		    "prefix(%s/%d) on %s\n", __func__,
2151 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
2152 		    pr->ndpr_plen, if_name(ifp)));
2153 		error = 0;
2154 	} else {
2155 		error = nd6_prefix_onlink_rtrequest(pr, ifa);
2156 		ifa_free(ifa);
2157 	}
2158 	NET_EPOCH_EXIT(et);
2159 
2160 	return (error);
2161 }
2162 
2163 int
2164 nd6_prefix_offlink(struct nd_prefix *pr)
2165 {
2166 	int error = 0;
2167 	struct ifnet *ifp = pr->ndpr_ifp;
2168 	struct nd_prefix *opr;
2169 	char ip6buf[INET6_ADDRSTRLEN];
2170 	uint64_t genid;
2171 	int a_failure;
2172 
2173 	ND6_ONLINK_LOCK_ASSERT();
2174 	ND6_UNLOCK_ASSERT();
2175 
2176 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0)
2177 		return (EEXIST);
2178 
2179 	struct sockaddr_in6 mask6 = {
2180 		.sin6_family = AF_INET6,
2181 		.sin6_len = sizeof(struct sockaddr_in6),
2182 		.sin6_addr = pr->ndpr_mask,
2183 	};
2184 	struct sockaddr_in6 *pmask6 = (pr->ndpr_plen != 128) ? &mask6 : NULL;
2185 
2186 	error = nd6_prefix_rtrequest(ifp->if_fib, RTM_DELETE,
2187 	    &pr->ndpr_prefix, pmask6, ifp, NULL);
2188 
2189 	a_failure = 1;
2190 	if (error == 0) {
2191 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
2192 
2193 		/*
2194 		 * There might be the same prefix on another interface,
2195 		 * the prefix which could not be on-link just because we have
2196 		 * the interface route (see comments in nd6_prefix_onlink).
2197 		 * If there's one, try to make the prefix on-link on the
2198 		 * interface.
2199 		 */
2200 		ND6_RLOCK();
2201 restart:
2202 		LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
2203 			/*
2204 			 * KAME specific: detached prefixes should not be
2205 			 * on-link.
2206 			 */
2207 			if (opr == pr || (opr->ndpr_stateflags &
2208 			    (NDPRF_ONLINK | NDPRF_DETACHED)) != 0)
2209 				continue;
2210 
2211 			if (opr->ndpr_plen == pr->ndpr_plen &&
2212 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
2213 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
2214 				int e;
2215 
2216 				genid = V_nd6_list_genid;
2217 				ND6_RUNLOCK();
2218 				if ((e = nd6_prefix_onlink(opr)) != 0) {
2219 					nd6log((LOG_ERR,
2220 					    "%s: failed to recover a prefix "
2221 					    "%s/%d from %s to %s (errno=%d)\n",
2222 					    __func__, ip6_sprintf(ip6buf,
2223 						&opr->ndpr_prefix.sin6_addr),
2224 					    opr->ndpr_plen, if_name(ifp),
2225 					    if_name(opr->ndpr_ifp), e));
2226 				} else
2227 					a_failure = 0;
2228 				ND6_RLOCK();
2229 				if (genid != V_nd6_list_genid)
2230 					goto restart;
2231 			}
2232 		}
2233 		ND6_RUNLOCK();
2234 	} else {
2235 		/* XXX: can we still set the NDPRF_ONLINK flag? */
2236 		nd6log((LOG_ERR,
2237 		    "%s: failed to delete route: %s/%d on %s (errno=%d)\n",
2238 		    __func__, ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
2239 		    pr->ndpr_plen, if_name(ifp), error));
2240 	}
2241 
2242 	if (a_failure)
2243 		lltable_prefix_free(AF_INET6,
2244 		    (struct sockaddr *)&pr->ndpr_prefix,
2245 		    (struct sockaddr *)&mask6, LLE_STATIC);
2246 
2247 	return (error);
2248 }
2249 
2250 /*
2251  * ia0 - corresponding public address
2252  */
2253 int
2254 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
2255 {
2256 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
2257 	struct in6_ifaddr *newia;
2258 	struct in6_aliasreq ifra;
2259 	int error;
2260 	int trylimit = 3;	/* XXX: adhoc value */
2261 	int updateflags;
2262 	u_int32_t randid[2];
2263 	time_t vltime0, pltime0;
2264 
2265 	in6_prepare_ifra(&ifra, &ia0->ia_addr.sin6_addr,
2266 	    &ia0->ia_prefixmask.sin6_addr);
2267 
2268 	ifra.ifra_addr = ia0->ia_addr;	/* XXX: do we need this ? */
2269 	/* clear the old IFID */
2270 	IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr,
2271 	    &ifra.ifra_prefixmask.sin6_addr);
2272 
2273   again:
2274 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
2275 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
2276 		nd6log((LOG_NOTICE, "%s: failed to find a good random IFID\n",
2277 		    __func__));
2278 		return (EINVAL);
2279 	}
2280 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
2281 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
2282 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
2283 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
2284 
2285 	/*
2286 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
2287 	 * However, we may still have a chance to see collision, because
2288 	 * there may be a time lag between generation of the ID and generation
2289 	 * of the address.  So, we'll do one more sanity check.
2290 	 */
2291 
2292 	if (in6_localip(&ifra.ifra_addr.sin6_addr) != 0) {
2293 		if (trylimit-- > 0) {
2294 			forcegen = 1;
2295 			goto again;
2296 		}
2297 
2298 		/* Give up.  Something strange should have happened.  */
2299 		nd6log((LOG_NOTICE, "%s: failed to find a unique random IFID\n",
2300 		    __func__));
2301 		return (EEXIST);
2302 	}
2303 
2304 	/*
2305 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
2306          * public address or TEMP_VALID_LIFETIME.
2307 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
2308          * of the public address or TEMP_PREFERRED_LIFETIME -
2309          * DESYNC_FACTOR.
2310 	 */
2311 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
2312 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
2313 		    (ia0->ia6_lifetime.ia6t_vltime -
2314 		    (time_uptime - ia0->ia6_updatetime));
2315 		if (vltime0 > V_ip6_temp_valid_lifetime)
2316 			vltime0 = V_ip6_temp_valid_lifetime;
2317 	} else
2318 		vltime0 = V_ip6_temp_valid_lifetime;
2319 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
2320 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
2321 		    (ia0->ia6_lifetime.ia6t_pltime -
2322 		    (time_uptime - ia0->ia6_updatetime));
2323 		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
2324 			pltime0 = V_ip6_temp_preferred_lifetime -
2325 			    V_ip6_desync_factor;
2326 		}
2327 	} else
2328 		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
2329 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
2330 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
2331 
2332 	/*
2333 	 * A temporary address is created only if this calculated Preferred
2334 	 * Lifetime is greater than REGEN_ADVANCE time units.
2335 	 */
2336 	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
2337 		return (0);
2338 
2339 	/* XXX: scope zone ID? */
2340 
2341 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
2342 
2343 	/* allocate ifaddr structure, link into chain, etc. */
2344 	updateflags = 0;
2345 	if (delay)
2346 		updateflags |= IN6_IFAUPDATE_DADDELAY;
2347 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2348 		return (error);
2349 
2350 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2351 	if (newia == NULL) {	/* XXX: can it happen? */
2352 		nd6log((LOG_ERR,
2353 		    "%s: ifa update succeeded, but we got no ifaddr\n",
2354 		    __func__));
2355 		return (EINVAL); /* XXX */
2356 	}
2357 	newia->ia6_ndpr = ia0->ia6_ndpr;
2358 	newia->ia6_ndpr->ndpr_addrcnt++;
2359 	ifa_free(&newia->ia_ifa);
2360 
2361 	/*
2362 	 * A newly added address might affect the status of other addresses.
2363 	 * XXX: when the temporary address is generated with a new public
2364 	 * address, the onlink check is redundant.  However, it would be safe
2365 	 * to do the check explicitly everywhere a new address is generated,
2366 	 * and, in fact, we surely need the check when we create a new
2367 	 * temporary address due to deprecation of an old temporary address.
2368 	 */
2369 	pfxlist_onlink_check();
2370 
2371 	return (0);
2372 }
2373 
2374 static int
2375 rt6_deleteroute(const struct rtentry *rt, const struct nhop_object *nh,
2376     void *arg)
2377 {
2378 	struct in6_addr *gate = (struct in6_addr *)arg;
2379 	int nh_rt_flags;
2380 
2381 	if (nh->gw_sa.sa_family != AF_INET6)
2382 		return (0);
2383 
2384 	if (!IN6_ARE_ADDR_EQUAL(gate, &nh->gw6_sa.sin6_addr)) {
2385 		return (0);
2386 	}
2387 
2388 	/*
2389 	 * Do not delete a static route.
2390 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2391 	 * 'cloned' bit instead?
2392 	 */
2393 	nh_rt_flags = nhop_get_rtflags(nh);
2394 	if ((nh_rt_flags & RTF_STATIC) != 0)
2395 		return (0);
2396 
2397 	/*
2398 	 * We delete only host route. This means, in particular, we don't
2399 	 * delete default route.
2400 	 */
2401 	if ((nh_rt_flags & RTF_HOST) == 0)
2402 		return (0);
2403 
2404 	return (1);
2405 #undef SIN6
2406 }
2407 
2408 /*
2409  * Delete all the routing table entries that use the specified gateway.
2410  * XXX: this function causes search through all entries of routing table, so
2411  * it shouldn't be called when acting as a router.
2412  */
2413 void
2414 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2415 {
2416 
2417 	/* We'll care only link-local addresses */
2418 	if (!IN6_IS_ADDR_LINKLOCAL(gateway))
2419 		return;
2420 
2421 	/* XXX Do we really need to walk any but the default FIB? */
2422 	rib_foreach_table_walk_del(AF_INET6, rt6_deleteroute, (void *)gateway);
2423 }
2424 
2425 int
2426 nd6_setdefaultiface(int ifindex)
2427 {
2428 
2429 	if (V_nd6_defifindex != ifindex) {
2430 		V_nd6_defifindex = ifindex;
2431 		if (V_nd6_defifindex != 0) {
2432 			struct epoch_tracker et;
2433 
2434 			/*
2435 			 * XXXGL: this function should use ifnet_byindex_ref!
2436 			 */
2437 			NET_EPOCH_ENTER(et);
2438 			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2439 			NET_EPOCH_EXIT(et);
2440 			if (V_nd6_defifp == NULL)
2441 				return (EINVAL);
2442 		} else
2443 			V_nd6_defifp = NULL;
2444 
2445 		/*
2446 		 * Our current implementation assumes one-to-one mapping between
2447 		 * interfaces and links, so it would be natural to use the
2448 		 * default interface as the default link.
2449 		 */
2450 		scope6_setdefault(V_nd6_defifp);
2451 	}
2452 
2453 	return (0);
2454 }
2455 
2456 bool
2457 nd6_defrouter_list_empty(void)
2458 {
2459 
2460 	return (TAILQ_EMPTY(&V_nd6_defrouter));
2461 }
2462 
2463 void
2464 nd6_defrouter_timer(void)
2465 {
2466 	struct nd_defrouter *dr, *ndr;
2467 	struct nd6_drhead drq;
2468 
2469 	TAILQ_INIT(&drq);
2470 
2471 	ND6_WLOCK();
2472 	TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr)
2473 		if (dr->expire && dr->expire < time_uptime)
2474 			defrouter_unlink(dr, &drq);
2475 	ND6_WUNLOCK();
2476 
2477 	while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2478 		TAILQ_REMOVE(&drq, dr, dr_entry);
2479 		defrouter_del(dr);
2480 	}
2481 }
2482 
2483 /*
2484  * Nuke default router list entries toward ifp.
2485  * We defer removal of default router list entries that is installed in the
2486  * routing table, in order to keep additional side effects as small as possible.
2487  */
2488 void
2489 nd6_defrouter_purge(struct ifnet *ifp)
2490 {
2491 	struct nd_defrouter *dr, *ndr;
2492 	struct nd6_drhead drq;
2493 
2494 	TAILQ_INIT(&drq);
2495 
2496 	ND6_WLOCK();
2497 	TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) {
2498 		if (dr->installed)
2499 			continue;
2500 		if (dr->ifp == ifp)
2501 			defrouter_unlink(dr, &drq);
2502 	}
2503 	TAILQ_FOREACH_SAFE(dr, &V_nd6_defrouter, dr_entry, ndr) {
2504 		if (!dr->installed)
2505 			continue;
2506 		if (dr->ifp == ifp)
2507 			defrouter_unlink(dr, &drq);
2508 	}
2509 	ND6_WUNLOCK();
2510 
2511 	/* Delete the unlinked router objects. */
2512 	while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2513 		TAILQ_REMOVE(&drq, dr, dr_entry);
2514 		defrouter_del(dr);
2515 	}
2516 }
2517 
2518 void
2519 nd6_defrouter_flush_all(void)
2520 {
2521 	struct nd_defrouter *dr;
2522 	struct nd6_drhead drq;
2523 
2524 	TAILQ_INIT(&drq);
2525 
2526 	ND6_WLOCK();
2527 	while ((dr = TAILQ_FIRST(&V_nd6_defrouter)) != NULL)
2528 		defrouter_unlink(dr, &drq);
2529 	ND6_WUNLOCK();
2530 
2531 	while ((dr = TAILQ_FIRST(&drq)) != NULL) {
2532 		TAILQ_REMOVE(&drq, dr, dr_entry);
2533 		defrouter_del(dr);
2534 	}
2535 }
2536 
2537 void
2538 nd6_defrouter_init(void)
2539 {
2540 
2541 	TAILQ_INIT(&V_nd6_defrouter);
2542 }
2543 
2544 static int
2545 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
2546 {
2547 	struct in6_defrouter d;
2548 	struct nd_defrouter *dr;
2549 	int error;
2550 
2551 	if (req->newptr != NULL)
2552 		return (EPERM);
2553 
2554 	error = sysctl_wire_old_buffer(req, 0);
2555 	if (error != 0)
2556 		return (error);
2557 
2558 	bzero(&d, sizeof(d));
2559 	d.rtaddr.sin6_family = AF_INET6;
2560 	d.rtaddr.sin6_len = sizeof(d.rtaddr);
2561 
2562 	ND6_RLOCK();
2563 	TAILQ_FOREACH(dr, &V_nd6_defrouter, dr_entry) {
2564 		d.rtaddr.sin6_addr = dr->rtaddr;
2565 		error = sa6_recoverscope(&d.rtaddr);
2566 		if (error != 0)
2567 			break;
2568 		d.flags = dr->raflags;
2569 		d.rtlifetime = dr->rtlifetime;
2570 		d.expire = dr->expire + (time_second - time_uptime);
2571 		d.if_index = dr->ifp->if_index;
2572 		error = SYSCTL_OUT(req, &d, sizeof(d));
2573 		if (error != 0)
2574 			break;
2575 	}
2576 	ND6_RUNLOCK();
2577 	return (error);
2578 }
2579 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
2580 	CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
2581 	NULL, 0, nd6_sysctl_drlist, "S,in6_defrouter",
2582 	"NDP default router list");
2583