Home | History | Annotate | Line # | Download | only in netinet6
nd6_rtr.c revision 1.127
      1 /*	$NetBSD: nd6_rtr.c,v 1.127 2016/12/14 06:33:01 ozaki-r Exp $	*/
      2 /*	$KAME: nd6_rtr.c,v 1.95 2001/02/07 08:09:47 itojun Exp $	*/
      3 
      4 /*
      5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      6  * All rights reserved.
      7  *
      8  * Redistribution and use in source and binary forms, with or without
      9  * modification, are permitted provided that the following conditions
     10  * are met:
     11  * 1. Redistributions of source code must retain the above copyright
     12  *    notice, this list of conditions and the following disclaimer.
     13  * 2. Redistributions in binary form must reproduce the above copyright
     14  *    notice, this list of conditions and the following disclaimer in the
     15  *    documentation and/or other materials provided with the distribution.
     16  * 3. Neither the name of the project nor the names of its contributors
     17  *    may be used to endorse or promote products derived from this software
     18  *    without specific prior written permission.
     19  *
     20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     30  * SUCH DAMAGE.
     31  */
     32 
     33 #include <sys/cdefs.h>
     34 __KERNEL_RCSID(0, "$NetBSD: nd6_rtr.c,v 1.127 2016/12/14 06:33:01 ozaki-r Exp $");
     35 
     36 #include <sys/param.h>
     37 #include <sys/systm.h>
     38 #include <sys/malloc.h>
     39 #include <sys/mbuf.h>
     40 #include <sys/socket.h>
     41 #include <sys/sockio.h>
     42 #include <sys/time.h>
     43 #include <sys/kernel.h>
     44 #include <sys/errno.h>
     45 #include <sys/ioctl.h>
     46 #include <sys/syslog.h>
     47 #include <sys/cprng.h>
     48 
     49 #include <net/if.h>
     50 #include <net/if_types.h>
     51 #include <net/if_dl.h>
     52 
     53 #include <netinet/in.h>
     54 #include <netinet6/in6_var.h>
     55 #include <netinet6/in6_ifattach.h>
     56 #include <netinet/ip6.h>
     57 #include <netinet6/ip6_var.h>
     58 #include <netinet6/nd6.h>
     59 #include <netinet/icmp6.h>
     60 #include <netinet6/icmp6_private.h>
     61 #include <netinet6/scope6_var.h>
     62 
     63 #include <net/net_osdep.h>
     64 
     65 static int rtpref(struct nd_defrouter *);
     66 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
     67 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
     68     struct mbuf *, int);
     69 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int, struct psref *);
     70 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
     71 	struct nd_defrouter *);
     72 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
     73 static void pfxrtr_del(struct nd_pfxrouter *);
     74 static struct nd_pfxrouter *find_pfxlist_reachable_router
     75 	(struct nd_prefix *);
     76 
     77 static void defrouter_addreq(struct nd_defrouter *);
     78 static void defrouter_delreq(struct nd_defrouter *);
     79 
     80 static int in6_init_prefix_ltimes(struct nd_prefix *);
     81 static void in6_init_address_ltimes(struct nd_prefix *,
     82 	struct in6_addrlifetime *);
     83 static void purge_detached(struct ifnet *);
     84 
     85 static int rt6_deleteroute_matcher(struct rtentry *, void *);
     86 
     87 static int nd6_prelist_add(struct nd_prefixctl *, struct nd_defrouter *,
     88 	struct nd_prefix **);
     89 static int nd6_prefix_onlink(struct nd_prefix *);
     90 static int nd6_prefix_offlink(struct nd_prefix *);
     91 static struct nd_prefix *nd6_prefix_lookup(struct nd_prefixctl *);
     92 
     93 extern int nd6_recalc_reachtm_interval;
     94 
     95 int ip6_use_tempaddr = 0;
     96 
     97 int ip6_desync_factor;
     98 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
     99 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
    100 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
    101 
    102 int nd6_numroutes = 0;
    103 
    104 /* RTPREF_MEDIUM has to be 0! */
    105 #define RTPREF_HIGH	1
    106 #define RTPREF_MEDIUM	0
    107 #define RTPREF_LOW	(-1)
    108 #define RTPREF_RESERVED	(-2)
    109 #define RTPREF_INVALID	(-3)	/* internal */
    110 
    111 static inline bool
    112 nd6_is_llinfo_probreach(struct nd_defrouter *dr)
    113 {
    114 	struct llentry *ln = NULL;
    115 
    116 	ln = nd6_lookup(&dr->rtaddr, dr->ifp, false);
    117 	if (ln == NULL)
    118 		return false;
    119 
    120 	if (!ND6_IS_LLINFO_PROBREACH(ln)) {
    121 		LLE_RUNLOCK(ln);
    122 		return false;
    123 	}
    124 
    125 	LLE_RUNLOCK(ln);
    126 	return true;
    127 }
    128 
    129 /*
    130  * Receive Router Solicitation Message - just for routers.
    131  * Router solicitation/advertisement is mostly managed by a userland program
    132  * (rtadvd) so here we have no function like nd6_ra_output().
    133  *
    134  * Based on RFC 2461
    135  */
    136 void
    137 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
    138 {
    139 	struct ifnet *ifp;
    140 	struct nd_ifinfo *ndi;
    141 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
    142 	struct nd_router_solicit *nd_rs;
    143 	struct in6_addr saddr6 = ip6->ip6_src;
    144 	char *lladdr = NULL;
    145 	int lladdrlen = 0;
    146 	union nd_opts ndopts;
    147 	struct psref psref;
    148 
    149 	ifp = m_get_rcvif_psref(m, &psref);
    150 	if (ifp == NULL)
    151 		goto freeit;
    152 
    153 	ndi = ND_IFINFO(ifp);
    154 
    155 	/* If I'm not a router, ignore it. */
    156 	if (nd6_accepts_rtadv(ndi) || !ip6_forwarding)
    157 		goto freeit;
    158 
    159 	/* Sanity checks */
    160 	if (ip6->ip6_hlim != 255) {
    161 		nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
    162 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
    163 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
    164 		goto bad;
    165 	}
    166 
    167 	/*
    168 	 * Don't update the neighbor cache, if src = ::.
    169 	 * This indicates that the src has no IP address assigned yet.
    170 	 */
    171 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
    172 		goto freeit;
    173 
    174 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
    175 	if (nd_rs == NULL) {
    176 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
    177 		return;
    178 	}
    179 
    180 	icmp6len -= sizeof(*nd_rs);
    181 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
    182 	if (nd6_options(&ndopts) < 0) {
    183 		nd6log(LOG_INFO, "invalid ND option, ignored\n");
    184 		/* nd6_options have incremented stats */
    185 		goto freeit;
    186 	}
    187 
    188 	if (ndopts.nd_opts_src_lladdr) {
    189 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
    190 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
    191 	}
    192 
    193 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
    194 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
    195 		    "(if %d, RS packet %d)\n",
    196 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2);
    197 		goto bad;
    198 	}
    199 
    200 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
    201 
    202  freeit:
    203 	m_put_rcvif_psref(ifp, &psref);
    204 	m_freem(m);
    205 	return;
    206 
    207  bad:
    208 	ICMP6_STATINC(ICMP6_STAT_BADRS);
    209 	m_put_rcvif_psref(ifp, &psref);
    210 	m_freem(m);
    211 }
    212 
    213 /*
    214  * Receive Router Advertisement Message.
    215  *
    216  * Based on RFC 2461
    217  * TODO: on-link bit on prefix information
    218  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
    219  */
    220 void
    221 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
    222 {
    223 	struct ifnet *ifp;
    224 	struct nd_ifinfo *ndi;
    225 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
    226 	struct nd_router_advert *nd_ra;
    227 	struct in6_addr saddr6 = ip6->ip6_src;
    228 #if 0
    229 	struct in6_addr daddr6 = ip6->ip6_dst;
    230 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
    231 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
    232 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
    233 #endif
    234 	int mcast = 0;
    235 	union nd_opts ndopts;
    236 	struct nd_defrouter *dr;
    237 	struct psref psref;
    238 
    239 	ifp = m_get_rcvif_psref(m, &psref);
    240 	if (ifp == NULL)
    241 		goto freeit;
    242 
    243 	ndi = ND_IFINFO(ifp);
    244 	/*
    245 	 * We only accept RAs when
    246 	 * the system-wide variable allows the acceptance, and the
    247 	 * per-interface variable allows RAs on the receiving interface.
    248 	 */
    249 	if (!nd6_accepts_rtadv(ndi))
    250 		goto freeit;
    251 
    252 	if (ip6->ip6_hlim != 255) {
    253 		nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
    254 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
    255 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
    256 		goto bad;
    257 	}
    258 
    259 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
    260 		nd6log(LOG_ERR, "src %s is not link-local\n",
    261 		    ip6_sprintf(&saddr6));
    262 		goto bad;
    263 	}
    264 
    265 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
    266 	if (nd_ra == NULL) {
    267 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
    268 		m_put_rcvif_psref(ifp, &psref);
    269 		return;
    270 	}
    271 
    272 	icmp6len -= sizeof(*nd_ra);
    273 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
    274 	if (nd6_options(&ndopts) < 0) {
    275 		nd6log(LOG_INFO, "invalid ND option, ignored\n");
    276 		/* nd6_options have incremented stats */
    277 		goto freeit;
    278 	}
    279 
    280     {
    281 	struct nd_defrouter drtr;
    282 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
    283 
    284 	/* remember if this is a multicasted advertisement */
    285 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
    286 		mcast = 1;
    287 
    288 	memset(&drtr, 0, sizeof(drtr));
    289 	drtr.rtaddr = saddr6;
    290 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
    291 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
    292 	drtr.expire = time_uptime + drtr.rtlifetime;
    293 	drtr.ifp = ifp;
    294 	/* unspecified or not? (RFC 2461 6.3.4) */
    295 	if (advreachable) {
    296 		NTOHL(advreachable);
    297 		if (advreachable <= MAX_REACHABLE_TIME &&
    298 		    ndi->basereachable != advreachable) {
    299 			ndi->basereachable = advreachable;
    300 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
    301 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
    302 		}
    303 	}
    304 	if (nd_ra->nd_ra_retransmit)
    305 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
    306 	if (nd_ra->nd_ra_curhoplimit) {
    307 		if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
    308 			ndi->chlim = nd_ra->nd_ra_curhoplimit;
    309 		else if (ndi->chlim != nd_ra->nd_ra_curhoplimit)
    310 			log(LOG_ERR, "nd_ra_input: lower CurHopLimit sent from "
    311 			   "%s on %s (current=%d, received=%d), ignored\n",
    312 			   ip6_sprintf(&ip6->ip6_src),
    313 			   if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
    314 	}
    315 	dr = defrtrlist_update(&drtr);
    316     }
    317 
    318 	/*
    319 	 * prefix
    320 	 */
    321 	if (ndopts.nd_opts_pi) {
    322 		struct nd_opt_hdr *pt;
    323 		struct nd_opt_prefix_info *pi = NULL;
    324 		struct nd_prefixctl prc;
    325 
    326 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
    327 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
    328 		     pt = (struct nd_opt_hdr *)((char *)pt +
    329 						(pt->nd_opt_len << 3))) {
    330 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
    331 				continue;
    332 			pi = (struct nd_opt_prefix_info *)pt;
    333 
    334 			if (pi->nd_opt_pi_len != 4) {
    335 				nd6log(LOG_INFO, "invalid option "
    336 				    "len %d for prefix information option, "
    337 				    "ignored\n", pi->nd_opt_pi_len);
    338 				continue;
    339 			}
    340 
    341 			if (128 < pi->nd_opt_pi_prefix_len) {
    342 				nd6log(LOG_INFO, "invalid prefix "
    343 				    "len %d for prefix information option, "
    344 				    "ignored\n", pi->nd_opt_pi_prefix_len);
    345 				continue;
    346 			}
    347 
    348 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
    349 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
    350 				nd6log(LOG_INFO,
    351 				    "invalid prefix %s, ignored\n",
    352 				    ip6_sprintf(&pi->nd_opt_pi_prefix));
    353 				continue;
    354 			}
    355 
    356 			memset(&prc, 0, sizeof(prc));
    357 			sockaddr_in6_init(&prc.ndprc_prefix,
    358 			    &pi->nd_opt_pi_prefix, 0, 0, 0);
    359 			prc.ndprc_ifp = ifp;
    360 
    361 			prc.ndprc_raf_onlink = (pi->nd_opt_pi_flags_reserved &
    362 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
    363 			prc.ndprc_raf_auto = (pi->nd_opt_pi_flags_reserved &
    364 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
    365 			prc.ndprc_plen = pi->nd_opt_pi_prefix_len;
    366 			prc.ndprc_vltime = ntohl(pi->nd_opt_pi_valid_time);
    367 			prc.ndprc_pltime = ntohl(pi->nd_opt_pi_preferred_time);
    368 
    369 			(void)prelist_update(&prc, dr, m, mcast);
    370 		}
    371 	}
    372 
    373 	/*
    374 	 * MTU
    375 	 */
    376 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
    377 		u_long mtu;
    378 		u_long maxmtu;
    379 
    380 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
    381 
    382 		/* lower bound */
    383 		if (mtu < IPV6_MMTU) {
    384 			nd6log(LOG_INFO, "bogus mtu option "
    385 			    "mtu=%lu sent from %s, ignoring\n",
    386 			    mtu, ip6_sprintf(&ip6->ip6_src));
    387 			goto skip;
    388 		}
    389 
    390 		/* upper bound */
    391 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
    392 		    ? ndi->maxmtu : ifp->if_mtu;
    393 		if (mtu <= maxmtu) {
    394 			int change = (ndi->linkmtu != mtu);
    395 
    396 			ndi->linkmtu = mtu;
    397 			if (change) /* in6_maxmtu may change */
    398 				in6_setmaxmtu();
    399 		} else {
    400 			nd6log(LOG_INFO,
    401 			    "bogus mtu mtu=%lu sent from %s; "
    402 			    "exceeds maxmtu %lu, ignoring\n",
    403 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu);
    404 		}
    405 	}
    406 
    407  skip:
    408 
    409 	/*
    410 	 * Source link layer address
    411 	 */
    412     {
    413 	char *lladdr = NULL;
    414 	int lladdrlen = 0;
    415 
    416 	if (ndopts.nd_opts_src_lladdr) {
    417 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
    418 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
    419 	}
    420 
    421 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
    422 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
    423 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
    424 		    ifp->if_addrlen, lladdrlen - 2);
    425 		goto bad;
    426 	}
    427 
    428 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
    429 
    430 	/*
    431 	 * Installing a link-layer address might change the state of the
    432 	 * router's neighbor cache, which might also affect our on-link
    433 	 * detection of adveritsed prefixes.
    434 	 */
    435 	nd6_pfxlist_onlink_check();
    436     }
    437 
    438  freeit:
    439 	m_put_rcvif_psref(ifp, &psref);
    440 	m_freem(m);
    441 	return;
    442 
    443  bad:
    444 	ICMP6_STATINC(ICMP6_STAT_BADRA);
    445 	m_put_rcvif_psref(ifp, &psref);
    446 	m_freem(m);
    447 }
    448 
    449 /*
    450  * default router list processing sub routines
    451  */
    452 static void
    453 defrouter_addreq(struct nd_defrouter *newdr)
    454 {
    455 	union {
    456 		struct sockaddr_in6 sin6;
    457 		struct sockaddr sa;
    458 	} def, mask, gate;
    459 	int s;
    460 	int error;
    461 
    462 	memset(&def, 0, sizeof(def));
    463 	memset(&mask, 0, sizeof(mask));
    464 	memset(&gate, 0,sizeof(gate)); /* for safety */
    465 
    466 	def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len =
    467 	    sizeof(struct sockaddr_in6);
    468 	def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6;
    469 	gate.sin6.sin6_addr = newdr->rtaddr;
    470 #ifndef SCOPEDROUTING
    471 	gate.sin6.sin6_scope_id = 0;	/* XXX */
    472 #endif
    473 
    474 	s = splsoftnet();
    475 	error = rtrequest_newmsg(RTM_ADD, &def.sa, &gate.sa, &mask.sa,
    476 	    RTF_GATEWAY);
    477 	if (error == 0) {
    478 		nd6_numroutes++;
    479 		newdr->installed = 1;
    480 	}
    481 	splx(s);
    482 	return;
    483 }
    484 
    485 struct nd_defrouter *
    486 nd6_defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
    487 {
    488 	struct nd_defrouter *dr;
    489 
    490 	ND_DEFROUTER_LIST_FOREACH(dr) {
    491 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
    492 			break;
    493 	}
    494 
    495 	return dr;		/* search failed */
    496 }
    497 
    498 void
    499 nd6_defrtrlist_del(struct nd_defrouter *dr, struct in6_ifextra *ext)
    500 {
    501 	struct nd_defrouter *deldr = NULL;
    502 	struct nd_prefix *pr;
    503 	struct nd_ifinfo *ndi;
    504 
    505 	if (ext == NULL)
    506 		ext = dr->ifp->if_afdata[AF_INET6];
    507 
    508 	/* detach already in progress, can not do anything */
    509 	if (ext == NULL)
    510 		return;
    511 
    512 	ndi = ext->nd_ifinfo;
    513 
    514 	/*
    515 	 * Flush all the routing table entries that use the router
    516 	 * as a next hop.
    517 	 */
    518 	/* XXX: better condition? */
    519 	if (!ip6_forwarding && nd6_accepts_rtadv(ndi))
    520 		nd6_rt_flush(&dr->rtaddr, dr->ifp);
    521 
    522 	if (dr->installed) {
    523 		deldr = dr;
    524 		defrouter_delreq(dr);
    525 	}
    526 	ND_DEFROUTER_LIST_REMOVE(dr);
    527 
    528 	/*
    529 	 * Also delete all the pointers to the router in each prefix lists.
    530 	 */
    531 	ND_PREFIX_LIST_FOREACH(pr) {
    532 		struct nd_pfxrouter *pfxrtr;
    533 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
    534 			pfxrtr_del(pfxrtr);
    535 	}
    536 	nd6_pfxlist_onlink_check();
    537 
    538 	/*
    539 	 * If the router is the primary one, choose a new one.
    540 	 * Note that nd6_defrouter_select() will remove the current gateway
    541 	 * from the routing table.
    542 	 */
    543 	if (deldr)
    544 		nd6_defrouter_select();
    545 
    546 	ext->ndefrouters--;
    547 	if (ext->ndefrouters < 0) {
    548 		log(LOG_WARNING, "nd6_defrtrlist_del: negative count on %s\n",
    549 		    dr->ifp->if_xname);
    550 	}
    551 
    552 	free(dr, M_IP6NDP);
    553 }
    554 
    555 /*
    556  * Remove the default route for a given router.
    557  * This is just a subroutine function for nd6_defrouter_select(), and should
    558  * not be called from anywhere else.
    559  */
    560 static void
    561 defrouter_delreq(struct nd_defrouter *dr)
    562 {
    563 	union {
    564 		struct sockaddr_in6 sin6;
    565 		struct sockaddr sa;
    566 	} def, mask, gw;
    567 	int error;
    568 
    569 #ifdef DIAGNOSTIC
    570 	if (dr == NULL)
    571 		panic("dr == NULL in defrouter_delreq");
    572 #endif
    573 
    574 	memset(&def, 0, sizeof(def));
    575 	memset(&mask, 0, sizeof(mask));
    576 	memset(&gw, 0, sizeof(gw));	/* for safety */
    577 
    578 	def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len =
    579 	    sizeof(struct sockaddr_in6);
    580 	def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6;
    581 	gw.sin6.sin6_addr = dr->rtaddr;
    582 #ifndef SCOPEDROUTING
    583 	gw.sin6.sin6_scope_id = 0;	/* XXX */
    584 #endif
    585 
    586 	error = rtrequest_newmsg(RTM_DELETE, &def.sa, &gw.sa, &mask.sa,
    587 	    RTF_GATEWAY);
    588 	if (error == 0)
    589 		nd6_numroutes--;
    590 
    591 	dr->installed = 0;
    592 }
    593 
    594 /*
    595  * remove all default routes from default router list
    596  */
    597 void
    598 nd6_defrouter_reset(void)
    599 {
    600 	struct nd_defrouter *dr;
    601 
    602 	ND_DEFROUTER_LIST_FOREACH(dr)
    603 		defrouter_delreq(dr);
    604 
    605 	/*
    606 	 * XXX should we also nuke any default routers in the kernel, by
    607 	 * going through them by rtalloc1()?
    608 	 */
    609 }
    610 
    611 /*
    612  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
    613  * draft-ietf-ipngwg-router-selection:
    614  * 1) Routers that are reachable or probably reachable should be preferred.
    615  *    If we have more than one (probably) reachable router, prefer ones
    616  *    with the highest router preference.
    617  * 2) When no routers on the list are known to be reachable or
    618  *    probably reachable, routers SHOULD be selected in a round-robin
    619  *    fashion, regardless of router preference values.
    620  * 3) If the Default Router List is empty, assume that all
    621  *    destinations are on-link.
    622  *
    623  * We assume nd_defrouter is sorted by router preference value.
    624  * Since the code below covers both with and without router preference cases,
    625  * we do not need to classify the cases by ifdef.
    626  *
    627  * At this moment, we do not try to install more than one default router,
    628  * even when the multipath routing is available, because we're not sure about
    629  * the benefits for stub hosts comparing to the risk of making the code
    630  * complicated and the possibility of introducing bugs.
    631  */
    632 void
    633 nd6_defrouter_select(void)
    634 {
    635 	struct nd_ifinfo *ndi;
    636 	int s = splsoftnet();
    637 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
    638 
    639 	/*
    640 	 * This function should be called only when acting as an autoconfigured
    641 	 * host.  Although the remaining part of this function is not effective
    642 	 * if the node is not an autoconfigured host, we explicitly exclude
    643 	 * such cases here for safety.
    644 	 */
    645 	if (ip6_forwarding) {
    646 		nd6log(LOG_WARNING, "called unexpectedly (forwarding=%d, "
    647 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv);
    648 		splx(s);
    649 		return;
    650 	}
    651 
    652 	/*
    653 	 * Let's handle easy case (3) first:
    654 	 * If default router list is empty, there's nothing to be done.
    655 	 */
    656 	if (ND_DEFROUTER_LIST_EMPTY()) {
    657 		splx(s);
    658 		return;
    659 	}
    660 
    661 	/*
    662 	 * Search for a (probably) reachable router from the list.
    663 	 * We just pick up the first reachable one (if any), assuming that
    664 	 * the ordering rule of the list described in defrtrlist_update().
    665 	 */
    666 	ND_DEFROUTER_LIST_FOREACH(dr) {
    667 		ndi = ND_IFINFO(dr->ifp);
    668 		if (nd6_accepts_rtadv(ndi))
    669 			continue;
    670 
    671 		if (selected_dr == NULL &&
    672 		    nd6_is_llinfo_probreach(dr))
    673 			selected_dr = dr;
    674 
    675 		if (dr->installed && !installed_dr)
    676 			installed_dr = dr;
    677 		else if (dr->installed && installed_dr) {
    678 			/* this should not happen.  warn for diagnosis. */
    679 			log(LOG_ERR, "nd6_defrouter_select: more than one router"
    680 			    " is installed\n");
    681 		}
    682 	}
    683 	/*
    684 	 * If none of the default routers was found to be reachable,
    685 	 * round-robin the list regardless of preference.
    686 	 * Otherwise, if we have an installed router, check if the selected
    687 	 * (reachable) router should really be preferred to the installed one.
    688 	 * We only prefer the new router when the old one is not reachable
    689 	 * or when the new one has a really higher preference value.
    690 	 */
    691 	if (selected_dr == NULL) {
    692 		if (installed_dr == NULL ||
    693 		    ND_DEFROUTER_LIST_NEXT(installed_dr) == NULL)
    694 			selected_dr = ND_DEFROUTER_LIST_FIRST();
    695 		else
    696 			selected_dr = ND_DEFROUTER_LIST_NEXT(installed_dr);
    697 	} else if (installed_dr &&
    698 	    nd6_is_llinfo_probreach(installed_dr) &&
    699 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
    700 		selected_dr = installed_dr;
    701 	}
    702 
    703 	/*
    704 	 * If the selected router is different than the installed one,
    705 	 * remove the installed router and install the selected one.
    706 	 * Note that the selected router is never NULL here.
    707 	 */
    708 	if (installed_dr != selected_dr) {
    709 		if (installed_dr)
    710 			defrouter_delreq(installed_dr);
    711 		defrouter_addreq(selected_dr);
    712 	}
    713 
    714 	splx(s);
    715 	return;
    716 }
    717 
    718 /*
    719  * for default router selection
    720  * regards router-preference field as a 2-bit signed integer
    721  */
    722 static int
    723 rtpref(struct nd_defrouter *dr)
    724 {
    725 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
    726 	case ND_RA_FLAG_RTPREF_HIGH:
    727 		return (RTPREF_HIGH);
    728 	case ND_RA_FLAG_RTPREF_MEDIUM:
    729 	case ND_RA_FLAG_RTPREF_RSV:
    730 		return (RTPREF_MEDIUM);
    731 	case ND_RA_FLAG_RTPREF_LOW:
    732 		return (RTPREF_LOW);
    733 	default:
    734 		/*
    735 		 * This case should never happen.  If it did, it would mean a
    736 		 * serious bug of kernel internal.  We thus always bark here.
    737 		 * Or, can we even panic?
    738 		 */
    739 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
    740 		return (RTPREF_INVALID);
    741 	}
    742 	/* NOTREACHED */
    743 }
    744 
    745 static struct nd_defrouter *
    746 defrtrlist_update(struct nd_defrouter *newdr)
    747 {
    748 	struct nd_defrouter *dr, *n, *ret = NULL;
    749 	struct in6_ifextra *ext = newdr->ifp->if_afdata[AF_INET6];
    750 	int s = splsoftnet();
    751 
    752 	if ((dr = nd6_defrouter_lookup(&newdr->rtaddr, newdr->ifp)) != NULL) {
    753 		/* entry exists */
    754 		if (newdr->rtlifetime == 0) {
    755 			nd6_defrtrlist_del(dr, ext);
    756 			dr = NULL;
    757 		} else {
    758 			int oldpref = rtpref(dr);
    759 
    760 			/* override */
    761 			dr->flags = newdr->flags; /* xxx flag check */
    762 			dr->rtlifetime = newdr->rtlifetime;
    763 			dr->expire = newdr->expire;
    764 
    765 			/*
    766 			 * If the preference does not change, there's no need
    767 			 * to sort the entries.
    768 			 */
    769 			if (rtpref(newdr) == oldpref) {
    770 				ret = dr;
    771 				goto out;
    772 			}
    773 
    774 			/*
    775 			 * preferred router may be changed, so relocate
    776 			 * this router.
    777 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
    778 			 * However, since nd6_defrtrlist_del() has many side
    779 			 * effects, we intentionally do so here.
    780 			 * nd6_defrouter_select() below will handle routing
    781 			 * changes later.
    782 			 */
    783 			ND_DEFROUTER_LIST_REMOVE(dr);
    784 			n = dr;
    785 			goto insert;
    786 		}
    787 		ret = dr;
    788 		goto out;
    789 	}
    790 
    791 	if (ip6_maxifdefrouters >= 0 && ext->ndefrouters >= ip6_maxifdefrouters)
    792 		goto out;
    793 
    794 	/* entry does not exist */
    795 	if (newdr->rtlifetime == 0)
    796 		goto out;
    797 
    798 	if (ip6_rtadv_maxroutes <= nd6_numroutes) {
    799 		ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
    800 		goto out;
    801 	}
    802 
    803 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
    804 	if (n == NULL)
    805 		goto out;
    806 	memset(n, 0, sizeof(*n));
    807 	*n = *newdr;
    808 
    809 insert:
    810 	/*
    811 	 * Insert the new router in the Default Router List;
    812 	 * The Default Router List should be in the descending order
    813 	 * of router-preferece.  Routers with the same preference are
    814 	 * sorted in the arriving time order.
    815 	 */
    816 
    817 	/* insert at the end of the group */
    818 	ND_DEFROUTER_LIST_FOREACH(dr) {
    819 		if (rtpref(n) > rtpref(dr))
    820 			break;
    821 	}
    822 	if (dr)
    823 		ND_DEFROUTER_LIST_INSERT_BEFORE(dr, n);
    824 	else
    825 		ND_DEFROUTER_LIST_INSERT_TAIL(n);
    826 
    827 	nd6_defrouter_select();
    828 
    829 	ext->ndefrouters++;
    830 
    831 	ret = n;
    832 out:
    833 	splx(s);
    834 	return ret;
    835 }
    836 
    837 static struct nd_pfxrouter *
    838 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
    839 {
    840 	struct nd_pfxrouter *search;
    841 
    842 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
    843 		if (search->router == dr)
    844 			break;
    845 	}
    846 
    847 	return (search);
    848 }
    849 
    850 static void
    851 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
    852 {
    853 	struct nd_pfxrouter *newpfr;
    854 
    855 	newpfr = malloc(sizeof(*newpfr), M_IP6NDP, M_NOWAIT|M_ZERO);
    856 	if (newpfr == NULL)
    857 		return;
    858 	newpfr->router = dr;
    859 
    860 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, newpfr, pfr_entry);
    861 
    862 	nd6_pfxlist_onlink_check();
    863 }
    864 
    865 static void
    866 pfxrtr_del(struct nd_pfxrouter *pfr)
    867 {
    868 	LIST_REMOVE(pfr, pfr_entry);
    869 	free(pfr, M_IP6NDP);
    870 }
    871 
    872 static struct nd_prefix *
    873 nd6_prefix_lookup(struct nd_prefixctl *key)
    874 {
    875 	struct nd_prefix *search;
    876 
    877 	ND_PREFIX_LIST_FOREACH(search) {
    878 		if (key->ndprc_ifp == search->ndpr_ifp &&
    879 		    key->ndprc_plen == search->ndpr_plen &&
    880 		    in6_are_prefix_equal(&key->ndprc_prefix.sin6_addr,
    881 		    &search->ndpr_prefix.sin6_addr, key->ndprc_plen)) {
    882 			break;
    883 		}
    884 	}
    885 
    886 	return (search);
    887 }
    888 
    889 static void
    890 purge_detached(struct ifnet *ifp)
    891 {
    892 	struct nd_prefix *pr, *pr_next;
    893 	struct in6_ifaddr *ia;
    894 	struct ifaddr *ifa, *ifa_next;
    895 
    896 	ND_PREFIX_LIST_FOREACH_SAFE(pr, pr_next) {
    897 		int s;
    898 
    899 		/*
    900 		 * This function is called when we need to make more room for
    901 		 * new prefixes rather than keeping old, possibly stale ones.
    902 		 * Detached prefixes would be a good candidate; if all routers
    903 		 * that advertised the prefix expired, the prefix is also
    904 		 * probably stale.
    905 		 */
    906 		if (pr->ndpr_ifp != ifp ||
    907 		    IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
    908 		    ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
    909 		    !LIST_EMPTY(&pr->ndpr_advrtrs)))
    910 			continue;
    911 
    912 	restart:
    913 		s = pserialize_read_enter();
    914 		for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = ifa_next) {
    915 			ifa_next = IFADDR_READER_NEXT(ifa);
    916 			if (ifa->ifa_addr->sa_family != AF_INET6)
    917 				continue;
    918 			ia = (struct in6_ifaddr *)ifa;
    919 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) ==
    920 			    IN6_IFF_AUTOCONF && ia->ia6_ndpr == pr) {
    921 				pserialize_read_exit(s);
    922 				in6_purgeaddr(ifa);
    923 				goto restart;
    924 			}
    925 		}
    926 		pserialize_read_exit(s);
    927 
    928 		if (pr->ndpr_refcnt == 0)
    929 			nd6_prelist_remove(pr);
    930 	}
    931 }
    932 
    933 static int
    934 nd6_prelist_add(struct nd_prefixctl *prc, struct nd_defrouter *dr,
    935 	struct nd_prefix **newp)
    936 {
    937 	struct nd_prefix *newpr = NULL;
    938 	int i, s;
    939 	int error;
    940 	struct in6_ifextra *ext = prc->ndprc_ifp->if_afdata[AF_INET6];
    941 
    942 	if (ip6_maxifprefixes >= 0) {
    943 		if (ext->nprefixes >= ip6_maxifprefixes / 2)
    944 			purge_detached(prc->ndprc_ifp);
    945 		if (ext->nprefixes >= ip6_maxifprefixes)
    946 			return ENOMEM;
    947 	}
    948 
    949 	error = 0;
    950 	newpr = malloc(sizeof(*newpr), M_IP6NDP, M_NOWAIT|M_ZERO);
    951 	if (newpr == NULL)
    952 		return ENOMEM;
    953 	newpr->ndpr_ifp = prc->ndprc_ifp;
    954 	newpr->ndpr_prefix = prc->ndprc_prefix;
    955 	newpr->ndpr_plen = prc->ndprc_plen;
    956 	newpr->ndpr_vltime = prc->ndprc_vltime;
    957 	newpr->ndpr_pltime = prc->ndprc_pltime;
    958 	newpr->ndpr_flags = prc->ndprc_flags;
    959 	if ((error = in6_init_prefix_ltimes(newpr)) != 0) {
    960 		free(newpr, M_IP6NDP);
    961 		return(error);
    962 	}
    963 	newpr->ndpr_lastupdate = time_uptime;
    964 	if (newp != NULL)
    965 		*newp = newpr;
    966 
    967 	/* initialization */
    968 	LIST_INIT(&newpr->ndpr_advrtrs);
    969 	in6_prefixlen2mask(&newpr->ndpr_mask, newpr->ndpr_plen);
    970 	/* make prefix in the canonical form */
    971 	for (i = 0; i < 4; i++) {
    972 		newpr->ndpr_prefix.sin6_addr.s6_addr32[i] &=
    973 		    newpr->ndpr_mask.s6_addr32[i];
    974 	}
    975 
    976 	s = splsoftnet();
    977 	/* link ndpr_entry to nd_prefix list */
    978 	ND_PREFIX_LIST_INSERT_HEAD(newpr);
    979 	splx(s);
    980 
    981 	/* ND_OPT_PI_FLAG_ONLINK processing */
    982 	if (newpr->ndpr_raf_onlink) {
    983 		int e;
    984 
    985 		if ((e = nd6_prefix_onlink(newpr)) != 0) {
    986 			nd6log(LOG_ERR, "failed to make "
    987 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
    988 			    ip6_sprintf(&prc->ndprc_prefix.sin6_addr),
    989 			    prc->ndprc_plen, if_name(prc->ndprc_ifp), e);
    990 			/* proceed anyway. XXX: is it correct? */
    991 		}
    992 	}
    993 
    994 	if (dr)
    995 		pfxrtr_add(newpr, dr);
    996 
    997 	ext->nprefixes++;
    998 
    999 	return 0;
   1000 }
   1001 
   1002 void
   1003 nd6_prelist_remove(struct nd_prefix *pr)
   1004 {
   1005 	struct nd_pfxrouter *pfr, *next;
   1006 	int e, s;
   1007 	struct in6_ifextra *ext = pr->ndpr_ifp->if_afdata[AF_INET6];
   1008 
   1009 	/* make sure to invalidate the prefix until it is really freed. */
   1010 	pr->ndpr_vltime = 0;
   1011 	pr->ndpr_pltime = 0;
   1012 #if 0
   1013 	/*
   1014 	 * Though these flags are now meaningless, we'd rather keep the value
   1015 	 * not to confuse users when executing "ndp -p".
   1016 	 */
   1017 	pr->ndpr_raf_onlink = 0;
   1018 	pr->ndpr_raf_auto = 0;
   1019 #endif
   1020 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
   1021 	    (e = nd6_prefix_offlink(pr)) != 0) {
   1022 		nd6log(LOG_ERR,
   1023 		    "failed to make %s/%d offlink on %s, errno=%d\n",
   1024 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1025 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e);
   1026 		/* what should we do? */
   1027 	}
   1028 
   1029 	if (pr->ndpr_refcnt > 0)
   1030 		return;		/* notice here? */
   1031 
   1032 	s = splsoftnet();
   1033 	/* unlink ndpr_entry from nd_prefix list */
   1034 	ND_PREFIX_LIST_REMOVE(pr);
   1035 
   1036 	/* free list of routers that adversed the prefix */
   1037 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
   1038 		next = LIST_NEXT(pfr, pfr_entry);
   1039 
   1040 		free(pfr, M_IP6NDP);
   1041 	}
   1042 
   1043 	if (ext) {
   1044 		ext->nprefixes--;
   1045 		if (ext->nprefixes < 0) {
   1046 			log(LOG_WARNING, "nd6_prelist_remove: negative count on "
   1047 			    "%s\n", pr->ndpr_ifp->if_xname);
   1048 		}
   1049 	}
   1050 	splx(s);
   1051 
   1052 	free(pr, M_IP6NDP);
   1053 
   1054 	nd6_pfxlist_onlink_check();
   1055 }
   1056 
   1057 static int
   1058 prelist_update(struct nd_prefixctl *newprc,
   1059 	struct nd_defrouter *dr, /* may be NULL */
   1060 	struct mbuf *m,
   1061 	int mcast)
   1062 {
   1063 	struct in6_ifaddr *ia6_match = NULL;
   1064 	struct ifaddr *ifa;
   1065 	struct ifnet *ifp = newprc->ndprc_ifp;
   1066 	struct nd_prefix *pr;
   1067 	int s = splsoftnet();
   1068 	int error = 0;
   1069 	int auth;
   1070 	struct in6_addrlifetime lt6_tmp;
   1071 	int ss;
   1072 
   1073 	auth = 0;
   1074 	if (m) {
   1075 		/*
   1076 		 * Authenticity for NA consists authentication for
   1077 		 * both IP header and IP datagrams, doesn't it ?
   1078 		 */
   1079 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
   1080 		auth = (m->m_flags & M_AUTHIPHDR
   1081 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
   1082 #endif
   1083 	}
   1084 
   1085 	if ((pr = nd6_prefix_lookup(newprc)) != NULL) {
   1086 		/*
   1087 		 * nd6_prefix_lookup() ensures that pr and newprc have the same
   1088 		 * prefix on a same interface.
   1089 		 */
   1090 
   1091 		/*
   1092 		 * Update prefix information.  Note that the on-link (L) bit
   1093 		 * and the autonomous (A) bit should NOT be changed from 1
   1094 		 * to 0.
   1095 		 */
   1096 		if (newprc->ndprc_raf_onlink == 1)
   1097 			pr->ndpr_raf_onlink = 1;
   1098 		if (newprc->ndprc_raf_auto == 1)
   1099 			pr->ndpr_raf_auto = 1;
   1100 		if (newprc->ndprc_raf_onlink) {
   1101 			pr->ndpr_vltime = newprc->ndprc_vltime;
   1102 			pr->ndpr_pltime = newprc->ndprc_pltime;
   1103 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
   1104 			pr->ndpr_lastupdate = time_uptime;
   1105 		}
   1106 
   1107 		if (newprc->ndprc_raf_onlink &&
   1108 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
   1109 			int e;
   1110 
   1111 			if ((e = nd6_prefix_onlink(pr)) != 0) {
   1112 				nd6log(LOG_ERR,
   1113 				    "failed to make "
   1114 				    "the prefix %s/%d on-link on %s "
   1115 				    "(errno=%d)\n",
   1116 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1117 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e);
   1118 				/* proceed anyway. XXX: is it correct? */
   1119 			}
   1120 		}
   1121 
   1122 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
   1123 			pfxrtr_add(pr, dr);
   1124 	} else {
   1125 		struct nd_prefix *newpr = NULL;
   1126 
   1127 		if (newprc->ndprc_vltime == 0)
   1128 			goto end;
   1129 		if (newprc->ndprc_raf_onlink == 0 && newprc->ndprc_raf_auto == 0)
   1130 			goto end;
   1131 
   1132 		if (ip6_rtadv_maxroutes <= nd6_numroutes) {
   1133 			ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
   1134 			goto end;
   1135 		}
   1136 
   1137 		error = nd6_prelist_add(newprc, dr, &newpr);
   1138 		if (error != 0 || newpr == NULL) {
   1139 			nd6log(LOG_NOTICE,
   1140 			    "nd6_prelist_add failed for %s/%d on %s "
   1141 			    "errno=%d, returnpr=%p\n",
   1142 			    ip6_sprintf(&newprc->ndprc_prefix.sin6_addr),
   1143 			    newprc->ndprc_plen, if_name(newprc->ndprc_ifp),
   1144 			    error, newpr);
   1145 			goto end; /* we should just give up in this case. */
   1146 		}
   1147 
   1148 		/*
   1149 		 * XXX: from the ND point of view, we can ignore a prefix
   1150 		 * with the on-link bit being zero.  However, we need a
   1151 		 * prefix structure for references from autoconfigured
   1152 		 * addresses.  Thus, we explicitly make sure that the prefix
   1153 		 * itself expires now.
   1154 		 */
   1155 		if (newpr->ndpr_raf_onlink == 0) {
   1156 			newpr->ndpr_vltime = 0;
   1157 			newpr->ndpr_pltime = 0;
   1158 			in6_init_prefix_ltimes(newpr);
   1159 		}
   1160 
   1161 		pr = newpr;
   1162 	}
   1163 
   1164 	/*
   1165 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
   1166 	 * Note that pr must be non NULL at this point.
   1167 	 */
   1168 
   1169 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
   1170 	if (!newprc->ndprc_raf_auto)
   1171 		goto end;
   1172 
   1173 	/*
   1174 	 * 5.5.3 (b). the link-local prefix should have been ignored in
   1175 	 * nd6_ra_input.
   1176 	 */
   1177 
   1178 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
   1179 	if (newprc->ndprc_pltime > newprc->ndprc_vltime) {
   1180 		error = EINVAL;	/* XXX: won't be used */
   1181 		goto end;
   1182 	}
   1183 
   1184 	/*
   1185 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
   1186 	 * an address configured by stateless autoconfiguration already in the
   1187 	 * list of addresses associated with the interface, and the Valid
   1188 	 * Lifetime is not 0, form an address.  We first check if we have
   1189 	 * a matching prefix.
   1190 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
   1191 	 * consider autoconfigured addresses while RFC2462 simply said
   1192 	 * "address".
   1193 	 */
   1194 	ss = pserialize_read_enter();
   1195 	IFADDR_READER_FOREACH(ifa, ifp) {
   1196 		struct in6_ifaddr *ia6;
   1197 		u_int32_t remaininglifetime;
   1198 
   1199 		if (ifa->ifa_addr->sa_family != AF_INET6)
   1200 			continue;
   1201 
   1202 		ia6 = (struct in6_ifaddr *)ifa;
   1203 
   1204 		/*
   1205 		 * We only consider autoconfigured addresses as per rfc2462bis.
   1206 		 */
   1207 		if (!(ia6->ia6_flags & IN6_IFF_AUTOCONF))
   1208 			continue;
   1209 
   1210 		/*
   1211 		 * Spec is not clear here, but I believe we should concentrate
   1212 		 * on unicast (i.e. not anycast) addresses.
   1213 		 * XXX: other ia6_flags? detached or duplicated?
   1214 		 */
   1215 		if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0)
   1216 			continue;
   1217 
   1218 		/*
   1219 		 * Ignore the address if it is not associated with a prefix
   1220 		 * or is associated with a prefix that is different from this
   1221 		 * one.  (pr is never NULL here)
   1222 		 */
   1223 		if (ia6->ia6_ndpr != pr)
   1224 			continue;
   1225 
   1226 		if (ia6_match == NULL) /* remember the first one */
   1227 			ia6_match = ia6;
   1228 
   1229 		/*
   1230 		 * An already autoconfigured address matched.  Now that we
   1231 		 * are sure there is at least one matched address, we can
   1232 		 * proceed to 5.5.3. (e): update the lifetimes according to the
   1233 		 * "two hours" rule and the privacy extension.
   1234 		 * We apply some clarifications in rfc2462bis:
   1235 		 * - use remaininglifetime instead of storedlifetime as a
   1236 		 *   variable name
   1237 		 * - remove the dead code in the "two-hour" rule
   1238 		 */
   1239 #define TWOHOUR		(120*60)
   1240 		lt6_tmp = ia6->ia6_lifetime;
   1241 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
   1242 			remaininglifetime = ND6_INFINITE_LIFETIME;
   1243 		else if (time_uptime - ia6->ia6_updatetime >
   1244 			 lt6_tmp.ia6t_vltime) {
   1245 			/*
   1246 			 * The case of "invalid" address.  We should usually
   1247 			 * not see this case.
   1248 			 */
   1249 			remaininglifetime = 0;
   1250 		} else
   1251 			remaininglifetime = lt6_tmp.ia6t_vltime -
   1252 			    (time_uptime - ia6->ia6_updatetime);
   1253 
   1254 		/* when not updating, keep the current stored lifetime. */
   1255 		lt6_tmp.ia6t_vltime = remaininglifetime;
   1256 
   1257 		if (TWOHOUR < newprc->ndprc_vltime ||
   1258 		    remaininglifetime < newprc->ndprc_vltime) {
   1259 			lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
   1260 		} else if (remaininglifetime <= TWOHOUR) {
   1261 			if (auth)
   1262 				lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
   1263 		} else {
   1264 			/*
   1265 			 * newprc->ndprc_vltime <= TWOHOUR &&
   1266 			 * TWOHOUR < remaininglifetime
   1267 			 */
   1268 			lt6_tmp.ia6t_vltime = TWOHOUR;
   1269 		}
   1270 
   1271 		/* The 2 hour rule is not imposed for preferred lifetime. */
   1272 		lt6_tmp.ia6t_pltime = newprc->ndprc_pltime;
   1273 
   1274 		in6_init_address_ltimes(pr, &lt6_tmp);
   1275 
   1276 		/*
   1277 		 * We need to treat lifetimes for temporary addresses
   1278 		 * differently, according to
   1279 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
   1280 		 * we only update the lifetimes when they are in the maximum
   1281 		 * intervals.
   1282 		 */
   1283 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
   1284 			u_int32_t maxvltime, maxpltime;
   1285 
   1286 			if (ip6_temp_valid_lifetime >
   1287 			    (u_int32_t)((time_uptime - ia6->ia6_createtime) +
   1288 			    ip6_desync_factor)) {
   1289 				maxvltime = ip6_temp_valid_lifetime -
   1290 				    (time_uptime - ia6->ia6_createtime) -
   1291 				    ip6_desync_factor;
   1292 			} else
   1293 				maxvltime = 0;
   1294 			if (ip6_temp_preferred_lifetime >
   1295 			    (u_int32_t)((time_uptime - ia6->ia6_createtime) +
   1296 			    ip6_desync_factor)) {
   1297 				maxpltime = ip6_temp_preferred_lifetime -
   1298 				    (time_uptime - ia6->ia6_createtime) -
   1299 				    ip6_desync_factor;
   1300 			} else
   1301 				maxpltime = 0;
   1302 
   1303 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
   1304 			    lt6_tmp.ia6t_vltime > maxvltime) {
   1305 				lt6_tmp.ia6t_vltime = maxvltime;
   1306 			}
   1307 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
   1308 			    lt6_tmp.ia6t_pltime > maxpltime) {
   1309 				lt6_tmp.ia6t_pltime = maxpltime;
   1310 			}
   1311 		}
   1312 
   1313 		ia6->ia6_lifetime = lt6_tmp;
   1314 		ia6->ia6_updatetime = time_uptime;
   1315 	}
   1316 	pserialize_read_exit(ss);
   1317 
   1318 	if (ia6_match == NULL && newprc->ndprc_vltime) {
   1319 		int ifidlen;
   1320 		struct in6_ifaddr *ia6;
   1321 		struct psref psref;
   1322 
   1323 		/*
   1324 		 * 5.5.3 (d) (continued)
   1325 		 * No address matched and the valid lifetime is non-zero.
   1326 		 * Create a new address.
   1327 		 */
   1328 
   1329 		/*
   1330 		 * Prefix Length check:
   1331 		 * If the sum of the prefix length and interface identifier
   1332 		 * length does not equal 128 bits, the Prefix Information
   1333 		 * option MUST be ignored.  The length of the interface
   1334 		 * identifier is defined in a separate link-type specific
   1335 		 * document.
   1336 		 */
   1337 		ifidlen = in6_if2idlen(ifp);
   1338 		if (ifidlen < 0) {
   1339 			/* this should not happen, so we always log it. */
   1340 			log(LOG_ERR, "%s: IFID undefined (%s)\n",
   1341 			    __func__, if_name(ifp));
   1342 			goto end;
   1343 		}
   1344 		if (ifidlen + pr->ndpr_plen != 128) {
   1345 			nd6log(LOG_INFO,
   1346 			    "invalid prefixlen %d for %s, ignored\n",
   1347 			    pr->ndpr_plen, if_name(ifp));
   1348 			goto end;
   1349 		}
   1350 
   1351 		if ((ia6 = in6_ifadd(newprc, mcast, &psref)) != NULL) {
   1352 			/*
   1353 			 * note that we should use pr (not newprc) for reference.
   1354 			 */
   1355 			pr->ndpr_refcnt++;
   1356 			ia6->ia6_ndpr = pr;
   1357 
   1358 			/* toggle onlink state if the address was assigned
   1359 			 * a prefix route. */
   1360 			if (ia6->ia_flags & IFA_ROUTE)
   1361 				pr->ndpr_stateflags |= NDPRF_ONLINK;
   1362 
   1363 			/*
   1364 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
   1365 			 * When a new public address is created as described
   1366 			 * in RFC2462, also create a new temporary address.
   1367 			 *
   1368 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
   1369 			 * When an interface connects to a new link, a new
   1370 			 * randomized interface identifier should be generated
   1371 			 * immediately together with a new set of temporary
   1372 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
   1373 			 * in6_tmpifadd().
   1374 			 */
   1375 			if (ip6_use_tempaddr) {
   1376 				int e;
   1377 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
   1378 					nd6log(LOG_NOTICE,
   1379 					    "failed to create a temporary "
   1380 					    "address, errno=%d\n", e);
   1381 				}
   1382 			}
   1383 			ia6_release(ia6, &psref);
   1384 
   1385 			/*
   1386 			 * A newly added address might affect the status
   1387 			 * of other addresses, so we check and update it.
   1388 			 * XXX: what if address duplication happens?
   1389 			 */
   1390 			nd6_pfxlist_onlink_check();
   1391 		} else {
   1392 			/* just set an error. do not bark here. */
   1393 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
   1394 		}
   1395 	}
   1396 
   1397  end:
   1398 	splx(s);
   1399 	return error;
   1400 }
   1401 
   1402 /*
   1403  * A supplement function used in the on-link detection below;
   1404  * detect if a given prefix has a (probably) reachable advertising router.
   1405  * XXX: lengthy function name...
   1406  */
   1407 static struct nd_pfxrouter *
   1408 find_pfxlist_reachable_router(struct nd_prefix *pr)
   1409 {
   1410 	struct nd_pfxrouter *pfxrtr;
   1411 
   1412 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
   1413 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
   1414 		if (pfxrtr->router->ifp->if_flags & IFF_UP &&
   1415 		    pfxrtr->router->ifp->if_link_state != LINK_STATE_DOWN &&
   1416 		    nd6_is_llinfo_probreach(pfxrtr->router))
   1417 			break;	/* found */
   1418 	}
   1419 
   1420 	return (pfxrtr);
   1421 }
   1422 
   1423 /*
   1424  * Check if each prefix in the prefix list has at least one available router
   1425  * that advertised the prefix (a router is "available" if its neighbor cache
   1426  * entry is reachable or probably reachable).
   1427  * If the check fails, the prefix may be off-link, because, for example,
   1428  * we have moved from the network but the lifetime of the prefix has not
   1429  * expired yet.  So we should not use the prefix if there is another prefix
   1430  * that has an available router.
   1431  * But, if there is no prefix that has an available router, we still regards
   1432  * all the prefixes as on-link.  This is because we can't tell if all the
   1433  * routers are simply dead or if we really moved from the network and there
   1434  * is no router around us.
   1435  */
   1436 void
   1437 nd6_pfxlist_onlink_check(void)
   1438 {
   1439 	struct nd_prefix *pr;
   1440 	struct in6_ifaddr *ia;
   1441 	struct nd_defrouter *dr;
   1442 	struct nd_pfxrouter *pfxrtr = NULL;
   1443 	int s;
   1444 
   1445 	/*
   1446 	 * Check if there is a prefix that has a reachable advertising
   1447 	 * router.
   1448 	 */
   1449 	ND_PREFIX_LIST_FOREACH(pr) {
   1450 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
   1451 			break;
   1452 	}
   1453 	/*
   1454 	 * If we have no such prefix, check whether we still have a router
   1455 	 * that does not advertise any prefixes.
   1456 	 */
   1457 	if (pr == NULL) {
   1458 		ND_DEFROUTER_LIST_FOREACH(dr) {
   1459 			struct nd_prefix *pr0;
   1460 
   1461 			ND_PREFIX_LIST_FOREACH(pr0) {
   1462 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
   1463 					break;
   1464 			}
   1465 			if (pfxrtr)
   1466 				break;
   1467 		}
   1468 	}
   1469 	if (pr != NULL || (!ND_DEFROUTER_LIST_EMPTY() && !pfxrtr)) {
   1470 		/*
   1471 		 * There is at least one prefix that has a reachable router,
   1472 		 * or at least a router which probably does not advertise
   1473 		 * any prefixes.  The latter would be the case when we move
   1474 		 * to a new link where we have a router that does not provide
   1475 		 * prefixes and we configure an address by hand.
   1476 		 * Detach prefixes which have no reachable advertising
   1477 		 * router, and attach other prefixes.
   1478 		 */
   1479 		ND_PREFIX_LIST_FOREACH(pr) {
   1480 			/* XXX: a link-local prefix should never be detached */
   1481 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
   1482 				continue;
   1483 
   1484 			/*
   1485 			 * we aren't interested in prefixes without the L bit
   1486 			 * set.
   1487 			 */
   1488 			if (pr->ndpr_raf_onlink == 0)
   1489 				continue;
   1490 
   1491 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
   1492 			    find_pfxlist_reachable_router(pr) == NULL)
   1493 				pr->ndpr_stateflags |= NDPRF_DETACHED;
   1494 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
   1495 			    find_pfxlist_reachable_router(pr) != 0)
   1496 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
   1497 		}
   1498 	} else {
   1499 		/* there is no prefix that has a reachable router */
   1500 		ND_PREFIX_LIST_FOREACH(pr) {
   1501 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
   1502 				continue;
   1503 
   1504 			if (pr->ndpr_raf_onlink == 0)
   1505 				continue;
   1506 
   1507 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
   1508 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
   1509 		}
   1510 	}
   1511 
   1512 	/*
   1513 	 * Remove each interface route associated with a (just) detached
   1514 	 * prefix, and reinstall the interface route for a (just) attached
   1515 	 * prefix.  Note that all attempt of reinstallation does not
   1516 	 * necessarily success, when a same prefix is shared among multiple
   1517 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
   1518 	 * so we don't have to care about them.
   1519 	 */
   1520 	ND_PREFIX_LIST_FOREACH(pr) {
   1521 		int e;
   1522 
   1523 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
   1524 			continue;
   1525 
   1526 		if (pr->ndpr_raf_onlink == 0)
   1527 			continue;
   1528 
   1529 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
   1530 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
   1531 			if ((e = nd6_prefix_offlink(pr)) != 0) {
   1532 				nd6log(LOG_ERR,
   1533 				    "failed to make %s/%d offlink, errno=%d\n",
   1534 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1535 				    pr->ndpr_plen, e);
   1536 			}
   1537 		}
   1538 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
   1539 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
   1540 		    pr->ndpr_raf_onlink) {
   1541 			if ((e = nd6_prefix_onlink(pr)) != 0) {
   1542 				nd6log(LOG_ERR,
   1543 				    "failed to make %s/%d onlink, errno=%d\n",
   1544 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1545 				    pr->ndpr_plen, e);
   1546 			}
   1547 		}
   1548 	}
   1549 
   1550 	/*
   1551 	 * Changes on the prefix status might affect address status as well.
   1552 	 * Make sure that all addresses derived from an attached prefix are
   1553 	 * attached, and that all addresses derived from a detached prefix are
   1554 	 * detached.  Note, however, that a manually configured address should
   1555 	 * always be attached.
   1556 	 * The precise detection logic is same as the one for prefixes.
   1557 	 */
   1558 	s = pserialize_read_enter();
   1559 	IN6_ADDRLIST_READER_FOREACH(ia) {
   1560 		if (!(ia->ia6_flags & IN6_IFF_AUTOCONF))
   1561 			continue;
   1562 
   1563 		if (ia->ia6_ndpr == NULL) {
   1564 			/*
   1565 			 * This can happen when we first configure the address
   1566 			 * (i.e. the address exists, but the prefix does not).
   1567 			 * XXX: complicated relationships...
   1568 			 */
   1569 			continue;
   1570 		}
   1571 
   1572 		if (find_pfxlist_reachable_router(ia->ia6_ndpr))
   1573 			break;
   1574 	}
   1575 	pserialize_read_exit(s);
   1576 
   1577 	if (ia) {
   1578 		int bound = curlwp_bind();
   1579 
   1580 		s = pserialize_read_enter();
   1581 		IN6_ADDRLIST_READER_FOREACH(ia) {
   1582 			struct ifaddr *ifa = (struct ifaddr *)ia;
   1583 			struct psref psref;
   1584 
   1585 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
   1586 				continue;
   1587 
   1588 			if (ia->ia6_ndpr == NULL) /* XXX: see above. */
   1589 				continue;
   1590 
   1591 			ia6_acquire(ia, &psref);
   1592 			pserialize_read_exit(s);
   1593 
   1594 			if (find_pfxlist_reachable_router(ia->ia6_ndpr)) {
   1595 				if (ia->ia6_flags & IN6_IFF_DETACHED) {
   1596 					ia->ia6_flags &= ~IN6_IFF_DETACHED;
   1597 					ia->ia6_flags |= IN6_IFF_TENTATIVE;
   1598 					nd6_dad_start(ifa,
   1599 					    0);
   1600 					/* We will notify the routing socket
   1601 					 * of the DAD result, so no need to
   1602 					 * here */
   1603 				}
   1604 			} else {
   1605 				if ((ia->ia6_flags & IN6_IFF_DETACHED) == 0) {
   1606 					ia->ia6_flags |= IN6_IFF_DETACHED;
   1607 					rt_newaddrmsg(RTM_NEWADDR,
   1608 					    ifa, 0, NULL);
   1609 				}
   1610 			}
   1611 
   1612 			s = pserialize_read_enter();
   1613 			ia6_release(ia, &psref);
   1614 		}
   1615 		pserialize_read_exit(s);
   1616 		curlwp_bindx(bound);
   1617 	}
   1618 	else {
   1619 		int bound = curlwp_bind();
   1620 
   1621 		s = pserialize_read_enter();
   1622 		IN6_ADDRLIST_READER_FOREACH(ia) {
   1623 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
   1624 				continue;
   1625 
   1626 			if (ia->ia6_flags & IN6_IFF_DETACHED) {
   1627 				struct ifaddr *ifa = (struct ifaddr *)ia;
   1628 				struct psref psref;
   1629 
   1630 				ia->ia6_flags &= ~IN6_IFF_DETACHED;
   1631 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
   1632 
   1633 				ia6_acquire(ia, &psref);
   1634 				pserialize_read_exit(s);
   1635 
   1636 				/* Do we need a delay in this case? */
   1637 				nd6_dad_start(ifa, 0);
   1638 
   1639 				s = pserialize_read_enter();
   1640 				ia6_release(ia, &psref);
   1641 			}
   1642 		}
   1643 		pserialize_read_exit(s);
   1644 		curlwp_bindx(bound);
   1645 	}
   1646 }
   1647 
   1648 static int
   1649 nd6_prefix_onlink(struct nd_prefix *pr)
   1650 {
   1651 	struct ifaddr *ifa;
   1652 	struct ifnet *ifp = pr->ndpr_ifp;
   1653 	struct sockaddr_in6 mask6;
   1654 	struct nd_prefix *opr;
   1655 	u_long rtflags;
   1656 	int error = 0;
   1657 	struct psref psref;
   1658 	int bound;
   1659 
   1660 	/* sanity check */
   1661 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
   1662 		nd6log(LOG_ERR, "%s/%d is already on-link\n",
   1663 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen);
   1664 		return (EEXIST);
   1665 	}
   1666 
   1667 	/*
   1668 	 * Add the interface route associated with the prefix.  Before
   1669 	 * installing the route, check if there's the same prefix on another
   1670 	 * interface, and the prefix has already installed the interface route.
   1671 	 * Although such a configuration is expected to be rare, we explicitly
   1672 	 * allow it.
   1673 	 */
   1674 	ND_PREFIX_LIST_FOREACH(opr) {
   1675 		if (opr == pr)
   1676 			continue;
   1677 
   1678 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
   1679 			continue;
   1680 
   1681 		if (opr->ndpr_plen == pr->ndpr_plen &&
   1682 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
   1683 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
   1684 			return (0);
   1685 	}
   1686 
   1687 	/*
   1688 	 * We prefer link-local addresses as the associated interface address.
   1689 	 */
   1690 	/* search for a link-local addr */
   1691 	bound = curlwp_bind();
   1692 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal_psref(ifp,
   1693 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST, &psref);
   1694 	if (ifa == NULL) {
   1695 		int s = pserialize_read_enter();
   1696 		IFADDR_READER_FOREACH(ifa, ifp) {
   1697 			if (ifa->ifa_addr->sa_family == AF_INET6)
   1698 				break;
   1699 		}
   1700 		if (ifa != NULL)
   1701 			ifa_acquire(ifa, &psref);
   1702 		pserialize_read_exit(s);
   1703 		/* should we care about ia6_flags? */
   1704 	}
   1705 	if (ifa == NULL) {
   1706 		/*
   1707 		 * This can still happen, when, for example, we receive an RA
   1708 		 * containing a prefix with the L bit set and the A bit clear,
   1709 		 * after removing all IPv6 addresses on the receiving
   1710 		 * interface.  This should, of course, be rare though.
   1711 		 */
   1712 		nd6log(LOG_NOTICE, "failed to find any ifaddr"
   1713 		    " to add route for a prefix(%s/%d) on %s\n",
   1714 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1715 		    pr->ndpr_plen, if_name(ifp));
   1716 		curlwp_bindx(bound);
   1717 		return (0);
   1718 	}
   1719 
   1720 	/*
   1721 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
   1722 	 * ifa->ifa_rtrequest = nd6_rtrequest;
   1723 	 */
   1724 	memset(&mask6, 0, sizeof(mask6));
   1725 	mask6.sin6_family = AF_INET6;
   1726 	mask6.sin6_len = sizeof(mask6);
   1727 	mask6.sin6_addr = pr->ndpr_mask;
   1728 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
   1729 	rtflags = ifa->ifa_flags | RTF_UP;
   1730 	if (nd6_need_cache(ifp)) {
   1731 		/* explicitly set in case ifa_flags does not set the flag. */
   1732 		rtflags |= RTF_CONNECTED;
   1733 	} else {
   1734 		/*
   1735 		 * explicitly clear the cloning bit in case ifa_flags sets it.
   1736 		 */
   1737 		rtflags &= ~RTF_CONNECTED;
   1738 	}
   1739 	error = rtrequest_newmsg(RTM_ADD, sin6tosa(&pr->ndpr_prefix),
   1740 	    ifa->ifa_addr, sin6tosa(&mask6), rtflags);
   1741 	if (error == 0) {
   1742 		nd6_numroutes++;
   1743 		pr->ndpr_stateflags |= NDPRF_ONLINK;
   1744 	} else {
   1745 		nd6log(LOG_ERR, "failed to add route for a"
   1746 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
   1747 		    "errno = %d\n",
   1748 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1749 		    pr->ndpr_plen, if_name(ifp),
   1750 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
   1751 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error);
   1752 	}
   1753 	ifa_release(ifa, &psref);
   1754 	curlwp_bindx(bound);
   1755 
   1756 	return (error);
   1757 }
   1758 
   1759 static int
   1760 nd6_prefix_offlink(struct nd_prefix *pr)
   1761 {
   1762 	int error = 0;
   1763 	struct ifnet *ifp = pr->ndpr_ifp;
   1764 	struct nd_prefix *opr;
   1765 	struct sockaddr_in6 sa6, mask6;
   1766 
   1767 	/* sanity check */
   1768 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
   1769 		nd6log(LOG_ERR, "%s/%d is already off-link\n",
   1770 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen);
   1771 		return (EEXIST);
   1772 	}
   1773 
   1774 	sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
   1775 	sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
   1776 	error = rtrequest_newmsg(RTM_DELETE, sin6tosa(&sa6), NULL,
   1777 	    sin6tosa(&mask6), 0);
   1778 	if (error == 0) {
   1779 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
   1780 		nd6_numroutes--;
   1781 
   1782 		/*
   1783 		 * There might be the same prefix on another interface,
   1784 		 * the prefix which could not be on-link just because we have
   1785 		 * the interface route (see comments in nd6_prefix_onlink).
   1786 		 * If there's one, try to make the prefix on-link on the
   1787 		 * interface.
   1788 		 */
   1789 		ND_PREFIX_LIST_FOREACH(opr) {
   1790 			if (opr == pr)
   1791 				continue;
   1792 
   1793 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
   1794 				continue;
   1795 
   1796 			/*
   1797 			 * KAME specific: detached prefixes should not be
   1798 			 * on-link.
   1799 			 */
   1800 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
   1801 				continue;
   1802 
   1803 			if (opr->ndpr_plen == pr->ndpr_plen &&
   1804 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
   1805 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
   1806 				int e;
   1807 
   1808 				if ((e = nd6_prefix_onlink(opr)) != 0) {
   1809 					nd6log(LOG_ERR, "failed to "
   1810 					    "recover a prefix %s/%d from %s "
   1811 					    "to %s (errno = %d)\n",
   1812 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
   1813 					    opr->ndpr_plen, if_name(ifp),
   1814 					    if_name(opr->ndpr_ifp), e);
   1815 				}
   1816 			}
   1817 		}
   1818 	} else {
   1819 		/* XXX: can we still set the NDPRF_ONLINK flag? */
   1820 		nd6log(LOG_ERR, "failed to delete route: "
   1821 		    "%s/%d on %s (errno = %d)\n",
   1822 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
   1823 		    error);
   1824 	}
   1825 
   1826 	return error;
   1827 }
   1828 
   1829 static struct in6_ifaddr *
   1830 in6_ifadd(struct nd_prefixctl *prc, int mcast, struct psref *psref)
   1831 {
   1832 	struct ifnet *ifp = prc->ndprc_ifp;
   1833 	struct ifaddr *ifa;
   1834 	struct in6_aliasreq ifra;
   1835 	struct in6_ifaddr *ia, *ib;
   1836 	int error, plen0;
   1837 	struct in6_addr mask;
   1838 	int prefixlen = prc->ndprc_plen;
   1839 	int updateflags;
   1840 	int s;
   1841 
   1842 	in6_prefixlen2mask(&mask, prefixlen);
   1843 
   1844 	/*
   1845 	 * find a link-local address (will be interface ID).
   1846 	 * Is it really mandatory? Theoretically, a global or a site-local
   1847 	 * address can be configured without a link-local address, if we
   1848 	 * have a unique interface identifier...
   1849 	 *
   1850 	 * it is not mandatory to have a link-local address, we can generate
   1851 	 * interface identifier on the fly.  we do this because:
   1852 	 * (1) it should be the easiest way to find interface identifier.
   1853 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
   1854 	 * for multiple addresses on a single interface, and possible shortcut
   1855 	 * of DAD.  we omitted DAD for this reason in the past.
   1856 	 * (3) a user can prevent autoconfiguration of global address
   1857 	 * by removing link-local address by hand (this is partly because we
   1858 	 * don't have other way to control the use of IPv6 on an interface.
   1859 	 * this has been our design choice - cf. NRL's "ifconfig auto").
   1860 	 * (4) it is easier to manage when an interface has addresses
   1861 	 * with the same interface identifier, than to have multiple addresses
   1862 	 * with different interface identifiers.
   1863 	 */
   1864 	s = pserialize_read_enter();
   1865 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
   1866 	if (ifa)
   1867 		ib = (struct in6_ifaddr *)ifa;
   1868 	else {
   1869 		pserialize_read_exit(s);
   1870 		return NULL;
   1871 	}
   1872 
   1873 #if 0 /* don't care link local addr state, and always do DAD */
   1874 	/* if link-local address is not eligible, do not autoconfigure. */
   1875 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
   1876 		printf("in6_ifadd: link-local address not ready\n");
   1877 		return NULL;
   1878 	}
   1879 #endif
   1880 
   1881 	/* prefixlen + ifidlen must be equal to 128 */
   1882 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
   1883 	if (prefixlen != plen0) {
   1884 		nd6log(LOG_INFO, "wrong prefixlen for %s "
   1885 		    "(prefix=%d ifid=%d)\n",
   1886 		    if_name(ifp), prefixlen, 128 - plen0);
   1887 		pserialize_read_exit(s);
   1888 		return NULL;
   1889 	}
   1890 
   1891 	/* make ifaddr */
   1892 
   1893 	memset(&ifra, 0, sizeof(ifra));
   1894 	/*
   1895 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
   1896 	 * for safety.
   1897 	 */
   1898 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
   1899 	sockaddr_in6_init(&ifra.ifra_addr, &prc->ndprc_prefix.sin6_addr, 0, 0, 0);
   1900 	/* prefix */
   1901 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
   1902 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
   1903 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
   1904 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
   1905 
   1906 	/* interface ID */
   1907 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
   1908 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
   1909 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
   1910 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
   1911 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
   1912 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
   1913 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
   1914 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
   1915 	pserialize_read_exit(s);
   1916 
   1917 	/* new prefix mask. */
   1918 	sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
   1919 
   1920 	/* lifetimes */
   1921 	ifra.ifra_lifetime.ia6t_vltime = prc->ndprc_vltime;
   1922 	ifra.ifra_lifetime.ia6t_pltime = prc->ndprc_pltime;
   1923 
   1924 	/* XXX: scope zone ID? */
   1925 
   1926 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
   1927 
   1928 	/*
   1929 	 * Make sure that we do not have this address already.  This should
   1930 	 * usually not happen, but we can still see this case, e.g., if we
   1931 	 * have manually configured the exact address to be configured.
   1932 	 */
   1933 	s = pserialize_read_enter();
   1934 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
   1935 		/* this should be rare enough to make an explicit log */
   1936 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
   1937 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
   1938 		pserialize_read_exit(s);
   1939 		return (NULL);
   1940 	}
   1941 	pserialize_read_exit(s);
   1942 
   1943 	/*
   1944 	 * Allocate ifaddr structure, link into chain, etc.
   1945 	 * If we are going to create a new address upon receiving a multicasted
   1946 	 * RA, we need to impose a random delay before starting DAD.
   1947 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
   1948 	 */
   1949 	updateflags = 0;
   1950 	if (mcast)
   1951 		updateflags |= IN6_IFAUPDATE_DADDELAY;
   1952 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
   1953 		nd6log(LOG_ERR, "failed to make ifaddr %s on %s (errno=%d)\n",
   1954 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
   1955 		    error);
   1956 		return (NULL);	/* ifaddr must not have been allocated. */
   1957 	}
   1958 
   1959 	ia = in6ifa_ifpwithaddr_psref(ifp, &ifra.ifra_addr.sin6_addr, psref);
   1960 
   1961 	return (ia);		/* this is always non-NULL */
   1962 }
   1963 
   1964 int
   1965 in6_tmpifadd(
   1966 	const struct in6_ifaddr *ia0, /* corresponding public address */
   1967 	int forcegen,
   1968 	int dad_delay)
   1969 {
   1970 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
   1971 	struct in6_ifaddr *newia, *ia;
   1972 	struct in6_aliasreq ifra;
   1973 	int i, error;
   1974 	int trylimit = 3;	/* XXX: adhoc value */
   1975 	int updateflags;
   1976 	u_int32_t randid[2];
   1977 	u_int32_t vltime0, pltime0;
   1978 	int s;
   1979 
   1980 	memset(&ifra, 0, sizeof(ifra));
   1981 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
   1982 	ifra.ifra_addr = ia0->ia_addr;
   1983 	/* copy prefix mask */
   1984 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
   1985 	/* clear the old IFID */
   1986 	for (i = 0; i < 4; i++) {
   1987 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
   1988 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
   1989 	}
   1990 
   1991   again:
   1992 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
   1993 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
   1994 		nd6log(LOG_NOTICE, "failed to find a good random IFID\n");
   1995 		return (EINVAL);
   1996 	}
   1997 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
   1998 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
   1999 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
   2000 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
   2001 
   2002 	/*
   2003 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
   2004 	 * However, we may still have a chance to see collision, because
   2005 	 * there may be a time lag between generation of the ID and generation
   2006 	 * of the address.  So, we'll do one more sanity check.
   2007 	 */
   2008 	s = pserialize_read_enter();
   2009 	IN6_ADDRLIST_READER_FOREACH(ia) {
   2010 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
   2011 		    &ifra.ifra_addr.sin6_addr)) {
   2012 			pserialize_read_exit(s);
   2013 			if (trylimit-- == 0) {
   2014 				/*
   2015 				 * Give up.  Something strange should have
   2016 				 * happened.
   2017 				 */
   2018 				nd6log(LOG_NOTICE,
   2019 				    "failed to find a unique random IFID\n");
   2020 				return (EEXIST);
   2021 			}
   2022 			forcegen = 1;
   2023 			goto again;
   2024 		}
   2025 	}
   2026 	pserialize_read_exit(s);
   2027 
   2028 	/*
   2029 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
   2030          * public address or TEMP_VALID_LIFETIME.
   2031 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
   2032          * of the public address or TEMP_PREFERRED_LIFETIME -
   2033          * DESYNC_FACTOR.
   2034 	 */
   2035 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
   2036 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
   2037 		    (ia0->ia6_lifetime.ia6t_vltime -
   2038 		    (time_uptime - ia0->ia6_updatetime));
   2039 		if (vltime0 > ip6_temp_valid_lifetime)
   2040 			vltime0 = ip6_temp_valid_lifetime;
   2041 	} else
   2042 		vltime0 = ip6_temp_valid_lifetime;
   2043 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
   2044 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
   2045 		    (ia0->ia6_lifetime.ia6t_pltime -
   2046 		    (time_uptime - ia0->ia6_updatetime));
   2047 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
   2048 			pltime0 = ip6_temp_preferred_lifetime -
   2049 			    ip6_desync_factor;
   2050 		}
   2051 	} else
   2052 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
   2053 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
   2054 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
   2055 
   2056 	/*
   2057 	 * A temporary address is created only if this calculated Preferred
   2058 	 * Lifetime is greater than REGEN_ADVANCE time units.
   2059 	 */
   2060 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
   2061 		return (0);
   2062 
   2063 	/* XXX: scope zone ID? */
   2064 
   2065 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
   2066 
   2067 	/* allocate ifaddr structure, link into chain, etc. */
   2068 	updateflags = 0;
   2069 	if (dad_delay)
   2070 		updateflags |= IN6_IFAUPDATE_DADDELAY;
   2071 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
   2072 		return (error);
   2073 
   2074 	s = pserialize_read_enter();
   2075 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
   2076 	if (newia == NULL) {	/* XXX: can it happen? */
   2077 		pserialize_read_exit(s);
   2078 		nd6log(LOG_ERR,
   2079 		    "ifa update succeeded, but we got no ifaddr\n");
   2080 		return (EINVAL); /* XXX */
   2081 	}
   2082 	newia->ia6_ndpr = ia0->ia6_ndpr;
   2083 	newia->ia6_ndpr->ndpr_refcnt++;
   2084 	pserialize_read_exit(s);
   2085 
   2086 	/*
   2087 	 * A newly added address might affect the status of other addresses.
   2088 	 * XXX: when the temporary address is generated with a new public
   2089 	 * address, the onlink check is redundant.  However, it would be safe
   2090 	 * to do the check explicitly everywhere a new address is generated,
   2091 	 * and, in fact, we surely need the check when we create a new
   2092 	 * temporary address due to deprecation of an old temporary address.
   2093 	 */
   2094 	nd6_pfxlist_onlink_check();
   2095 
   2096 	return (0);
   2097 }
   2098 
   2099 static int
   2100 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
   2101 {
   2102 
   2103 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
   2104 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
   2105 		nd6log(LOG_INFO, "preferred lifetime"
   2106 		    "(%d) is greater than valid lifetime(%d)\n",
   2107 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime);
   2108 		return (EINVAL);
   2109 	}
   2110 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
   2111 		ndpr->ndpr_preferred = 0;
   2112 	else
   2113 		ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
   2114 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
   2115 		ndpr->ndpr_expire = 0;
   2116 	else
   2117 		ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
   2118 
   2119 	return 0;
   2120 }
   2121 
   2122 static void
   2123 in6_init_address_ltimes(struct nd_prefix *newpr,
   2124     struct in6_addrlifetime *lt6)
   2125 {
   2126 
   2127 	/* Valid lifetime must not be updated unless explicitly specified. */
   2128 	/* init ia6t_expire */
   2129 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
   2130 		lt6->ia6t_expire = 0;
   2131 	else {
   2132 		lt6->ia6t_expire = time_uptime;
   2133 		lt6->ia6t_expire += lt6->ia6t_vltime;
   2134 	}
   2135 
   2136 	/* init ia6t_preferred */
   2137 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
   2138 		lt6->ia6t_preferred = 0;
   2139 	else {
   2140 		lt6->ia6t_preferred = time_uptime;
   2141 		lt6->ia6t_preferred += lt6->ia6t_pltime;
   2142 	}
   2143 }
   2144 
   2145 /*
   2146  * Delete all the routing table entries that use the specified gateway.
   2147  * XXX: this function causes search through all entries of routing table, so
   2148  * it shouldn't be called when acting as a router.
   2149  */
   2150 void
   2151 nd6_rt_flush(struct in6_addr *gateway, struct ifnet *ifp)
   2152 {
   2153 	int s = splsoftnet();
   2154 
   2155 	/* We'll care only link-local addresses */
   2156 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
   2157 		splx(s);
   2158 		return;
   2159 	}
   2160 
   2161 	rt_delete_matched_entries(AF_INET6, rt6_deleteroute_matcher, gateway);
   2162 	splx(s);
   2163 }
   2164 
   2165 static int
   2166 rt6_deleteroute_matcher(struct rtentry *rt, void *arg)
   2167 {
   2168 	struct in6_addr *gate = (struct in6_addr *)arg;
   2169 
   2170 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
   2171 		return (0);
   2172 
   2173 	if (!IN6_ARE_ADDR_EQUAL(gate, &satosin6(rt->rt_gateway)->sin6_addr))
   2174 		return (0);
   2175 
   2176 	/*
   2177 	 * Do not delete a static route.
   2178 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
   2179 	 * 'cloned' bit instead?
   2180 	 */
   2181 	if ((rt->rt_flags & RTF_STATIC) != 0)
   2182 		return (0);
   2183 
   2184 	/*
   2185 	 * We delete only host route. This means, in particular, we don't
   2186 	 * delete default route.
   2187 	 */
   2188 	if ((rt->rt_flags & RTF_HOST) == 0)
   2189 		return (0);
   2190 
   2191 	return 1;
   2192 }
   2193