Home | History | Annotate | Line # | Download | only in netinet6
nd6_rtr.c revision 1.106
      1 /*	$NetBSD: nd6_rtr.c,v 1.106 2016/04/01 06:33:19 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 1.106 2016/04/01 06:33:19 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 #include <net/route.h>
     53 #include <net/radix.h>
     54 
     55 #include <netinet/in.h>
     56 #include <netinet6/in6_var.h>
     57 #include <netinet6/in6_ifattach.h>
     58 #include <netinet/ip6.h>
     59 #include <netinet6/ip6_var.h>
     60 #include <netinet6/nd6.h>
     61 #include <netinet/icmp6.h>
     62 #include <netinet6/icmp6_private.h>
     63 #include <netinet6/scope6_var.h>
     64 
     65 #include <net/net_osdep.h>
     66 
     67 static int rtpref(struct nd_defrouter *);
     68 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
     69 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
     70     struct mbuf *, int);
     71 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
     72 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
     73 	struct nd_defrouter *);
     74 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
     75 static void pfxrtr_del(struct nd_pfxrouter *);
     76 static struct nd_pfxrouter *find_pfxlist_reachable_router
     77 	(struct nd_prefix *);
     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(struct rtentry *, void *);
     86 
     87 extern int nd6_recalc_reachtm_interval;
     88 
     89 static struct ifnet *nd6_defifp;
     90 int nd6_defifindex;
     91 
     92 int ip6_use_tempaddr = 0;
     93 
     94 int ip6_desync_factor;
     95 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
     96 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
     97 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
     98 
     99 int nd6_numroutes = 0;
    100 
    101 /* RTPREF_MEDIUM has to be 0! */
    102 #define RTPREF_HIGH	1
    103 #define RTPREF_MEDIUM	0
    104 #define RTPREF_LOW	(-1)
    105 #define RTPREF_RESERVED	(-2)
    106 #define RTPREF_INVALID	(-3)	/* internal */
    107 
    108 static inline bool
    109 nd6_is_llinfo_probreach(struct nd_defrouter *dr)
    110 {
    111 	struct rtentry *rt = NULL;
    112 	struct llentry *ln = NULL;
    113 
    114 	rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp);
    115 	if (rt == NULL)
    116 		return false;
    117 	ln = rt->rt_llinfo;
    118 	rtfree(rt);
    119 	if (ln == NULL || !ND6_IS_LLINFO_PROBREACH(ln))
    120 		return false;
    121 
    122 	return true;
    123 }
    124 
    125 /*
    126  * Receive Router Solicitation Message - just for routers.
    127  * Router solicitation/advertisement is mostly managed by a userland program
    128  * (rtadvd) so here we have no function like nd6_ra_output().
    129  *
    130  * Based on RFC 2461
    131  */
    132 void
    133 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
    134 {
    135 	struct ifnet *ifp = m->m_pkthdr.rcvif;
    136 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
    137 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
    138 	struct nd_router_solicit *nd_rs;
    139 	struct in6_addr saddr6 = ip6->ip6_src;
    140 	char *lladdr = NULL;
    141 	int lladdrlen = 0;
    142 	union nd_opts ndopts;
    143 
    144 	/* If I'm not a router, ignore it. */
    145 	if (nd6_accepts_rtadv(ndi) || !ip6_forwarding)
    146 		goto freeit;
    147 
    148 	/* Sanity checks */
    149 	if (ip6->ip6_hlim != 255) {
    150 		nd6log((LOG_ERR,
    151 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
    152 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
    153 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
    154 		goto bad;
    155 	}
    156 
    157 	/*
    158 	 * Don't update the neighbor cache, if src = ::.
    159 	 * This indicates that the src has no IP address assigned yet.
    160 	 */
    161 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
    162 		goto freeit;
    163 
    164 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
    165 	if (nd_rs == NULL) {
    166 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
    167 		return;
    168 	}
    169 
    170 	icmp6len -= sizeof(*nd_rs);
    171 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
    172 	if (nd6_options(&ndopts) < 0) {
    173 		nd6log((LOG_INFO,
    174 		    "nd6_rs_input: invalid ND option, ignored\n"));
    175 		/* nd6_options have incremented stats */
    176 		goto freeit;
    177 	}
    178 
    179 	if (ndopts.nd_opts_src_lladdr) {
    180 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
    181 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
    182 	}
    183 
    184 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
    185 		nd6log((LOG_INFO,
    186 		    "nd6_rs_input: lladdrlen mismatch for %s "
    187 		    "(if %d, RS packet %d)\n",
    188 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
    189 		goto bad;
    190 	}
    191 
    192 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
    193 
    194  freeit:
    195 	m_freem(m);
    196 	return;
    197 
    198  bad:
    199 	ICMP6_STATINC(ICMP6_STAT_BADRS);
    200 	m_freem(m);
    201 }
    202 
    203 /*
    204  * Receive Router Advertisement Message.
    205  *
    206  * Based on RFC 2461
    207  * TODO: on-link bit on prefix information
    208  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
    209  */
    210 void
    211 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
    212 {
    213 	struct ifnet *ifp = m->m_pkthdr.rcvif;
    214 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
    215 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
    216 	struct nd_router_advert *nd_ra;
    217 	struct in6_addr saddr6 = ip6->ip6_src;
    218 #if 0
    219 	struct in6_addr daddr6 = ip6->ip6_dst;
    220 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
    221 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
    222 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
    223 #endif
    224 	int mcast = 0;
    225 	union nd_opts ndopts;
    226 	struct nd_defrouter *dr;
    227 
    228 	/*
    229 	 * We only accept RAs when
    230 	 * the system-wide variable allows the acceptance, and the
    231 	 * per-interface variable allows RAs on the receiving interface.
    232 	 */
    233 	if (!nd6_accepts_rtadv(ndi))
    234 		goto freeit;
    235 
    236 	if (ip6->ip6_hlim != 255) {
    237 		nd6log((LOG_ERR,
    238 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
    239 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
    240 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
    241 		goto bad;
    242 	}
    243 
    244 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
    245 		nd6log((LOG_ERR,
    246 		    "nd6_ra_input: src %s is not link-local\n",
    247 		    ip6_sprintf(&saddr6)));
    248 		goto bad;
    249 	}
    250 
    251 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
    252 	if (nd_ra == NULL) {
    253 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
    254 		return;
    255 	}
    256 
    257 	icmp6len -= sizeof(*nd_ra);
    258 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
    259 	if (nd6_options(&ndopts) < 0) {
    260 		nd6log((LOG_INFO,
    261 		    "nd6_ra_input: invalid ND option, ignored\n"));
    262 		/* nd6_options have incremented stats */
    263 		goto freeit;
    264 	}
    265 
    266     {
    267 	struct nd_defrouter drtr;
    268 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
    269 
    270 	/* remember if this is a multicasted advertisement */
    271 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
    272 		mcast = 1;
    273 
    274 	memset(&drtr, 0, sizeof(drtr));
    275 	drtr.rtaddr = saddr6;
    276 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
    277 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
    278 	drtr.expire = time_uptime + drtr.rtlifetime;
    279 	drtr.ifp = ifp;
    280 	/* unspecified or not? (RFC 2461 6.3.4) */
    281 	if (advreachable) {
    282 		NTOHL(advreachable);
    283 		if (advreachable <= MAX_REACHABLE_TIME &&
    284 		    ndi->basereachable != advreachable) {
    285 			ndi->basereachable = advreachable;
    286 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
    287 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
    288 		}
    289 	}
    290 	if (nd_ra->nd_ra_retransmit)
    291 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
    292 	if (nd_ra->nd_ra_curhoplimit) {
    293 		if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
    294 			ndi->chlim = nd_ra->nd_ra_curhoplimit;
    295 		else if (ndi->chlim != nd_ra->nd_ra_curhoplimit)
    296 			log(LOG_ERR, "nd_ra_input: lower CurHopLimit sent from "
    297 			   "%s on %s (current=%d, received=%d), ignored\n",
    298 			   ip6_sprintf(&ip6->ip6_src),
    299 			   if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
    300 	}
    301 	dr = defrtrlist_update(&drtr);
    302     }
    303 
    304 	/*
    305 	 * prefix
    306 	 */
    307 	if (ndopts.nd_opts_pi) {
    308 		struct nd_opt_hdr *pt;
    309 		struct nd_opt_prefix_info *pi = NULL;
    310 		struct nd_prefixctl prc;
    311 
    312 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
    313 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
    314 		     pt = (struct nd_opt_hdr *)((char *)pt +
    315 						(pt->nd_opt_len << 3))) {
    316 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
    317 				continue;
    318 			pi = (struct nd_opt_prefix_info *)pt;
    319 
    320 			if (pi->nd_opt_pi_len != 4) {
    321 				nd6log((LOG_INFO,
    322 				    "nd6_ra_input: invalid option "
    323 				    "len %d for prefix information option, "
    324 				    "ignored\n", pi->nd_opt_pi_len));
    325 				continue;
    326 			}
    327 
    328 			if (128 < pi->nd_opt_pi_prefix_len) {
    329 				nd6log((LOG_INFO,
    330 				    "nd6_ra_input: invalid prefix "
    331 				    "len %d for prefix information option, "
    332 				    "ignored\n", pi->nd_opt_pi_prefix_len));
    333 				continue;
    334 			}
    335 
    336 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
    337 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
    338 				nd6log((LOG_INFO,
    339 				    "nd6_ra_input: invalid prefix "
    340 				    "%s, ignored\n",
    341 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
    342 				continue;
    343 			}
    344 
    345 			memset(&prc, 0, sizeof(prc));
    346 			sockaddr_in6_init(&prc.ndprc_prefix,
    347 			    &pi->nd_opt_pi_prefix, 0, 0, 0);
    348 			prc.ndprc_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
    349 
    350 			prc.ndprc_raf_onlink = (pi->nd_opt_pi_flags_reserved &
    351 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
    352 			prc.ndprc_raf_auto = (pi->nd_opt_pi_flags_reserved &
    353 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
    354 			prc.ndprc_plen = pi->nd_opt_pi_prefix_len;
    355 			prc.ndprc_vltime = ntohl(pi->nd_opt_pi_valid_time);
    356 			prc.ndprc_pltime = ntohl(pi->nd_opt_pi_preferred_time);
    357 
    358 			(void)prelist_update(&prc, dr, m, mcast);
    359 		}
    360 	}
    361 
    362 	/*
    363 	 * MTU
    364 	 */
    365 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
    366 		u_long mtu;
    367 		u_long maxmtu;
    368 
    369 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
    370 
    371 		/* lower bound */
    372 		if (mtu < IPV6_MMTU) {
    373 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
    374 			    "mtu=%lu sent from %s, ignoring\n",
    375 			    mtu, ip6_sprintf(&ip6->ip6_src)));
    376 			goto skip;
    377 		}
    378 
    379 		/* upper bound */
    380 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
    381 		    ? ndi->maxmtu : ifp->if_mtu;
    382 		if (mtu <= maxmtu) {
    383 			int change = (ndi->linkmtu != mtu);
    384 
    385 			ndi->linkmtu = mtu;
    386 			if (change) /* in6_maxmtu may change */
    387 				in6_setmaxmtu();
    388 		} else {
    389 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
    390 			    "mtu=%lu sent from %s; "
    391 			    "exceeds maxmtu %lu, ignoring\n",
    392 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
    393 		}
    394 	}
    395 
    396  skip:
    397 
    398 	/*
    399 	 * Source link layer address
    400 	 */
    401     {
    402 	char *lladdr = NULL;
    403 	int lladdrlen = 0;
    404 
    405 	if (ndopts.nd_opts_src_lladdr) {
    406 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
    407 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
    408 	}
    409 
    410 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
    411 		nd6log((LOG_INFO,
    412 		    "nd6_ra_input: lladdrlen mismatch for %s "
    413 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
    414 		    ifp->if_addrlen, lladdrlen - 2));
    415 		goto bad;
    416 	}
    417 
    418 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
    419 
    420 	/*
    421 	 * Installing a link-layer address might change the state of the
    422 	 * router's neighbor cache, which might also affect our on-link
    423 	 * detection of adveritsed prefixes.
    424 	 */
    425 	pfxlist_onlink_check();
    426     }
    427 
    428  freeit:
    429 	m_freem(m);
    430 	return;
    431 
    432  bad:
    433 	ICMP6_STATINC(ICMP6_STAT_BADRA);
    434 	m_freem(m);
    435 }
    436 
    437 /*
    438  * default router list processing sub routines
    439  */
    440 void
    441 defrouter_addreq(struct nd_defrouter *newdr)
    442 {
    443 	union {
    444 		struct sockaddr_in6 sin6;
    445 		struct sockaddr sa;
    446 	} def, mask, gate;
    447 	int s;
    448 	int error;
    449 
    450 	memset(&def, 0, sizeof(def));
    451 	memset(&mask, 0, sizeof(mask));
    452 	memset(&gate, 0,sizeof(gate)); /* for safety */
    453 
    454 	def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len =
    455 	    sizeof(struct sockaddr_in6);
    456 	def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6;
    457 	gate.sin6.sin6_addr = newdr->rtaddr;
    458 #ifndef SCOPEDROUTING
    459 	gate.sin6.sin6_scope_id = 0;	/* XXX */
    460 #endif
    461 
    462 	s = splsoftnet();
    463 	error = rtrequest_newmsg(RTM_ADD, &def.sa, &gate.sa, &mask.sa,
    464 	    RTF_GATEWAY);
    465 	if (error == 0) {
    466 		nd6_numroutes++;
    467 		newdr->installed = 1;
    468 	}
    469 	splx(s);
    470 	return;
    471 }
    472 
    473 struct nd_defrouter *
    474 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
    475 {
    476 	struct nd_defrouter *dr;
    477 
    478 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
    479 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
    480 			break;
    481 	}
    482 
    483 	return dr;		/* search failed */
    484 }
    485 
    486 void
    487 defrtrlist_del(struct nd_defrouter *dr, struct in6_ifextra *ext)
    488 {
    489 	struct nd_defrouter *deldr = NULL;
    490 	struct nd_prefix *pr;
    491 	struct nd_ifinfo *ndi;
    492 
    493 	if (ext == NULL)
    494 		ext = dr->ifp->if_afdata[AF_INET6];
    495 
    496 	/* detach already in progress, can not do anything */
    497 	if (ext == NULL)
    498 		return;
    499 
    500 	ndi = ext->nd_ifinfo;
    501 
    502 	/*
    503 	 * Flush all the routing table entries that use the router
    504 	 * as a next hop.
    505 	 */
    506 	/* XXX: better condition? */
    507 	if (!ip6_forwarding && nd6_accepts_rtadv(ndi))
    508 		rt6_flush(&dr->rtaddr, dr->ifp);
    509 
    510 	if (dr->installed) {
    511 		deldr = dr;
    512 		defrouter_delreq(dr);
    513 	}
    514 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
    515 
    516 	/*
    517 	 * Also delete all the pointers to the router in each prefix lists.
    518 	 */
    519 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
    520 		struct nd_pfxrouter *pfxrtr;
    521 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
    522 			pfxrtr_del(pfxrtr);
    523 	}
    524 	pfxlist_onlink_check();
    525 
    526 	/*
    527 	 * If the router is the primary one, choose a new one.
    528 	 * Note that defrouter_select() will remove the current gateway
    529 	 * from the routing table.
    530 	 */
    531 	if (deldr)
    532 		defrouter_select();
    533 
    534 	ext->ndefrouters--;
    535 	if (ext->ndefrouters < 0) {
    536 		log(LOG_WARNING, "defrtrlist_del: negative count on %s\n",
    537 		    dr->ifp->if_xname);
    538 	}
    539 
    540 	free(dr, M_IP6NDP);
    541 }
    542 
    543 /*
    544  * Remove the default route for a given router.
    545  * This is just a subroutine function for defrouter_select(), and should
    546  * not be called from anywhere else.
    547  */
    548 static void
    549 defrouter_delreq(struct nd_defrouter *dr)
    550 {
    551 	union {
    552 		struct sockaddr_in6 sin6;
    553 		struct sockaddr sa;
    554 	} def, mask, gw;
    555 	int error;
    556 
    557 #ifdef DIAGNOSTIC
    558 	if (dr == NULL)
    559 		panic("dr == NULL in defrouter_delreq");
    560 #endif
    561 
    562 	memset(&def, 0, sizeof(def));
    563 	memset(&mask, 0, sizeof(mask));
    564 	memset(&gw, 0, sizeof(gw));	/* for safety */
    565 
    566 	def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len =
    567 	    sizeof(struct sockaddr_in6);
    568 	def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6;
    569 	gw.sin6.sin6_addr = dr->rtaddr;
    570 #ifndef SCOPEDROUTING
    571 	gw.sin6.sin6_scope_id = 0;	/* XXX */
    572 #endif
    573 
    574 	error = rtrequest_newmsg(RTM_DELETE, &def.sa, &gw.sa, &mask.sa,
    575 	    RTF_GATEWAY);
    576 	if (error == 0)
    577 		nd6_numroutes--;
    578 
    579 	dr->installed = 0;
    580 }
    581 
    582 /*
    583  * remove all default routes from default router list
    584  */
    585 void
    586 defrouter_reset(void)
    587 {
    588 	struct nd_defrouter *dr;
    589 
    590 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
    591 	     dr = TAILQ_NEXT(dr, dr_entry))
    592 		defrouter_delreq(dr);
    593 
    594 	/*
    595 	 * XXX should we also nuke any default routers in the kernel, by
    596 	 * going through them by rtalloc1()?
    597 	 */
    598 }
    599 
    600 /*
    601  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
    602  * draft-ietf-ipngwg-router-selection:
    603  * 1) Routers that are reachable or probably reachable should be preferred.
    604  *    If we have more than one (probably) reachable router, prefer ones
    605  *    with the highest router preference.
    606  * 2) When no routers on the list are known to be reachable or
    607  *    probably reachable, routers SHOULD be selected in a round-robin
    608  *    fashion, regardless of router preference values.
    609  * 3) If the Default Router List is empty, assume that all
    610  *    destinations are on-link.
    611  *
    612  * We assume nd_defrouter is sorted by router preference value.
    613  * Since the code below covers both with and without router preference cases,
    614  * we do not need to classify the cases by ifdef.
    615  *
    616  * At this moment, we do not try to install more than one default router,
    617  * even when the multipath routing is available, because we're not sure about
    618  * the benefits for stub hosts comparing to the risk of making the code
    619  * complicated and the possibility of introducing bugs.
    620  */
    621 void
    622 defrouter_select(void)
    623 {
    624 	struct nd_ifinfo *ndi;
    625 	int s = splsoftnet();
    626 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
    627 
    628 	/*
    629 	 * This function should be called only when acting as an autoconfigured
    630 	 * host.  Although the remaining part of this function is not effective
    631 	 * if the node is not an autoconfigured host, we explicitly exclude
    632 	 * such cases here for safety.
    633 	 */
    634 	if (ip6_forwarding) {
    635 		nd6log((LOG_WARNING,
    636 		    "defrouter_select: called unexpectedly (forwarding=%d, "
    637 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
    638 		splx(s);
    639 		return;
    640 	}
    641 
    642 	/*
    643 	 * Let's handle easy case (3) first:
    644 	 * If default router list is empty, there's nothing to be done.
    645 	 */
    646 	if (!TAILQ_FIRST(&nd_defrouter)) {
    647 		splx(s);
    648 		return;
    649 	}
    650 
    651 	/*
    652 	 * Search for a (probably) reachable router from the list.
    653 	 * We just pick up the first reachable one (if any), assuming that
    654 	 * the ordering rule of the list described in defrtrlist_update().
    655 	 */
    656 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
    657 	     dr = TAILQ_NEXT(dr, dr_entry)) {
    658 		ndi = ND_IFINFO(dr->ifp);
    659 		if (nd6_accepts_rtadv(ndi))
    660 			continue;
    661 
    662 		if (selected_dr == NULL &&
    663 		    nd6_is_llinfo_probreach(dr))
    664 			selected_dr = dr;
    665 
    666 		if (dr->installed && !installed_dr)
    667 			installed_dr = dr;
    668 		else if (dr->installed && installed_dr) {
    669 			/* this should not happen.  warn for diagnosis. */
    670 			log(LOG_ERR, "defrouter_select: more than one router"
    671 			    " is installed\n");
    672 		}
    673 	}
    674 	/*
    675 	 * If none of the default routers was found to be reachable,
    676 	 * round-robin the list regardless of preference.
    677 	 * Otherwise, if we have an installed router, check if the selected
    678 	 * (reachable) router should really be preferred to the installed one.
    679 	 * We only prefer the new router when the old one is not reachable
    680 	 * or when the new one has a really higher preference value.
    681 	 */
    682 	if (selected_dr == NULL) {
    683 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
    684 			selected_dr = TAILQ_FIRST(&nd_defrouter);
    685 		else
    686 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
    687 	} else if (installed_dr &&
    688 	    nd6_is_llinfo_probreach(installed_dr) &&
    689 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
    690 		selected_dr = installed_dr;
    691 	}
    692 
    693 	/*
    694 	 * If the selected router is different than the installed one,
    695 	 * remove the installed router and install the selected one.
    696 	 * Note that the selected router is never NULL here.
    697 	 */
    698 	if (installed_dr != selected_dr) {
    699 		if (installed_dr)
    700 			defrouter_delreq(installed_dr);
    701 		defrouter_addreq(selected_dr);
    702 	}
    703 
    704 	splx(s);
    705 	return;
    706 }
    707 
    708 /*
    709  * for default router selection
    710  * regards router-preference field as a 2-bit signed integer
    711  */
    712 static int
    713 rtpref(struct nd_defrouter *dr)
    714 {
    715 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
    716 	case ND_RA_FLAG_RTPREF_HIGH:
    717 		return (RTPREF_HIGH);
    718 	case ND_RA_FLAG_RTPREF_MEDIUM:
    719 	case ND_RA_FLAG_RTPREF_RSV:
    720 		return (RTPREF_MEDIUM);
    721 	case ND_RA_FLAG_RTPREF_LOW:
    722 		return (RTPREF_LOW);
    723 	default:
    724 		/*
    725 		 * This case should never happen.  If it did, it would mean a
    726 		 * serious bug of kernel internal.  We thus always bark here.
    727 		 * Or, can we even panic?
    728 		 */
    729 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
    730 		return (RTPREF_INVALID);
    731 	}
    732 	/* NOTREACHED */
    733 }
    734 
    735 static struct nd_defrouter *
    736 defrtrlist_update(struct nd_defrouter *newdr)
    737 {
    738 	struct nd_defrouter *dr, *n;
    739 	struct in6_ifextra *ext = newdr->ifp->if_afdata[AF_INET6];
    740 	int s = splsoftnet();
    741 
    742 	if ((dr = defrouter_lookup(&newdr->rtaddr, newdr->ifp)) != NULL) {
    743 		/* entry exists */
    744 		if (newdr->rtlifetime == 0) {
    745 			defrtrlist_del(dr, ext);
    746 			dr = NULL;
    747 		} else {
    748 			int oldpref = rtpref(dr);
    749 
    750 			/* override */
    751 			dr->flags = newdr->flags; /* xxx flag check */
    752 			dr->rtlifetime = newdr->rtlifetime;
    753 			dr->expire = newdr->expire;
    754 
    755 			/*
    756 			 * If the preference does not change, there's no need
    757 			 * to sort the entries.
    758 			 */
    759 			if (rtpref(newdr) == oldpref) {
    760 				splx(s);
    761 				return (dr);
    762 			}
    763 
    764 			/*
    765 			 * preferred router may be changed, so relocate
    766 			 * this router.
    767 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
    768 			 * However, since defrtrlist_del() has many side
    769 			 * effects, we intentionally do so here.
    770 			 * defrouter_select() below will handle routing
    771 			 * changes later.
    772 			 */
    773 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
    774 			n = dr;
    775 			goto insert;
    776 		}
    777 		splx(s);
    778 		return (dr);
    779 	}
    780 
    781 	if (ip6_maxifdefrouters >= 0 &&
    782 	    ext->ndefrouters >= ip6_maxifdefrouters) {
    783 		splx(s);
    784 		return (NULL);
    785 	}
    786 
    787 	/* entry does not exist */
    788 	if (newdr->rtlifetime == 0) {
    789 		splx(s);
    790 		return (NULL);
    791 	}
    792 
    793 	if (ip6_rtadv_maxroutes <= nd6_numroutes) {
    794 		ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
    795 		splx(s);
    796 		return (NULL);
    797 	}
    798 
    799 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
    800 	if (n == NULL) {
    801 		splx(s);
    802 		return (NULL);
    803 	}
    804 	memset(n, 0, sizeof(*n));
    805 	*n = *newdr;
    806 
    807 insert:
    808 	/*
    809 	 * Insert the new router in the Default Router List;
    810 	 * The Default Router List should be in the descending order
    811 	 * of router-preferece.  Routers with the same preference are
    812 	 * sorted in the arriving time order.
    813 	 */
    814 
    815 	/* insert at the end of the group */
    816 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
    817 	     dr = TAILQ_NEXT(dr, dr_entry)) {
    818 		if (rtpref(n) > rtpref(dr))
    819 			break;
    820 	}
    821 	if (dr)
    822 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
    823 	else
    824 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
    825 
    826 	defrouter_select();
    827 
    828 	ext->ndefrouters++;
    829 
    830 	splx(s);
    831 
    832 	return (n);
    833 }
    834 
    835 static struct nd_pfxrouter *
    836 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
    837 {
    838 	struct nd_pfxrouter *search;
    839 
    840 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
    841 		if (search->router == dr)
    842 			break;
    843 	}
    844 
    845 	return (search);
    846 }
    847 
    848 static void
    849 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
    850 {
    851 	struct nd_pfxrouter *newpfr;
    852 
    853 	newpfr = malloc(sizeof(*newpfr), M_IP6NDP, M_NOWAIT|M_ZERO);
    854 	if (newpfr == NULL)
    855 		return;
    856 	newpfr->router = dr;
    857 
    858 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, newpfr, pfr_entry);
    859 
    860 	pfxlist_onlink_check();
    861 }
    862 
    863 static void
    864 pfxrtr_del(struct nd_pfxrouter *pfr)
    865 {
    866 	LIST_REMOVE(pfr, pfr_entry);
    867 	free(pfr, M_IP6NDP);
    868 }
    869 
    870 struct nd_prefix *
    871 nd6_prefix_lookup(struct nd_prefixctl *key)
    872 {
    873 	struct nd_prefix *search;
    874 
    875 	LIST_FOREACH(search, &nd_prefix, ndpr_entry) {
    876 		if (key->ndprc_ifp == search->ndpr_ifp &&
    877 		    key->ndprc_plen == search->ndpr_plen &&
    878 		    in6_are_prefix_equal(&key->ndprc_prefix.sin6_addr,
    879 		    &search->ndpr_prefix.sin6_addr, key->ndprc_plen)) {
    880 			break;
    881 		}
    882 	}
    883 
    884 	return (search);
    885 }
    886 
    887 static void
    888 purge_detached(struct ifnet *ifp)
    889 {
    890 	struct nd_prefix *pr, *pr_next;
    891 	struct in6_ifaddr *ia;
    892 	struct ifaddr *ifa, *ifa_next;
    893 
    894 	for (pr = nd_prefix.lh_first; pr; pr = pr_next) {
    895 		pr_next = pr->ndpr_next;
    896 
    897 		/*
    898 		 * This function is called when we need to make more room for
    899 		 * new prefixes rather than keeping old, possibly stale ones.
    900 		 * Detached prefixes would be a good candidate; if all routers
    901 		 * that advertised the prefix expired, the prefix is also
    902 		 * probably stale.
    903 		 */
    904 		if (pr->ndpr_ifp != ifp ||
    905 		    IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
    906 		    ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
    907 		    !LIST_EMPTY(&pr->ndpr_advrtrs)))
    908 			continue;
    909 
    910 		IFADDR_FOREACH_SAFE(ifa, ifp, ifa_next) {
    911 			if (ifa->ifa_addr->sa_family != AF_INET6)
    912 				continue;
    913 			ia = (struct in6_ifaddr *)ifa;
    914 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) ==
    915 			    IN6_IFF_AUTOCONF && ia->ia6_ndpr == pr) {
    916 				in6_purgeaddr(ifa);
    917 			}
    918 		}
    919 		if (pr->ndpr_refcnt == 0)
    920 			prelist_remove(pr);
    921 	}
    922 }
    923 int
    924 nd6_prelist_add(struct nd_prefixctl *prc, struct nd_defrouter *dr,
    925 	struct nd_prefix **newp)
    926 {
    927 	struct nd_prefix *newpr = NULL;
    928 	int i, s;
    929 	int error;
    930 	struct in6_ifextra *ext = prc->ndprc_ifp->if_afdata[AF_INET6];
    931 
    932 	if (ip6_maxifprefixes >= 0) {
    933 		if (ext->nprefixes >= ip6_maxifprefixes / 2)
    934 			purge_detached(prc->ndprc_ifp);
    935 		if (ext->nprefixes >= ip6_maxifprefixes)
    936 			return ENOMEM;
    937 	}
    938 
    939 	error = 0;
    940 	newpr = malloc(sizeof(*newpr), M_IP6NDP, M_NOWAIT|M_ZERO);
    941 	if (newpr == NULL)
    942 		return ENOMEM;
    943 	newpr->ndpr_ifp = prc->ndprc_ifp;
    944 	newpr->ndpr_prefix = prc->ndprc_prefix;
    945 	newpr->ndpr_plen = prc->ndprc_plen;
    946 	newpr->ndpr_vltime = prc->ndprc_vltime;
    947 	newpr->ndpr_pltime = prc->ndprc_pltime;
    948 	newpr->ndpr_flags = prc->ndprc_flags;
    949 	if ((error = in6_init_prefix_ltimes(newpr)) != 0) {
    950 		free(newpr, M_IP6NDP);
    951 		return(error);
    952 	}
    953 	newpr->ndpr_lastupdate = time_uptime;
    954 	if (newp != NULL)
    955 		*newp = newpr;
    956 
    957 	/* initialization */
    958 	LIST_INIT(&newpr->ndpr_advrtrs);
    959 	in6_prefixlen2mask(&newpr->ndpr_mask, newpr->ndpr_plen);
    960 	/* make prefix in the canonical form */
    961 	for (i = 0; i < 4; i++) {
    962 		newpr->ndpr_prefix.sin6_addr.s6_addr32[i] &=
    963 		    newpr->ndpr_mask.s6_addr32[i];
    964 	}
    965 
    966 	s = splsoftnet();
    967 	/* link ndpr_entry to nd_prefix list */
    968 	LIST_INSERT_HEAD(&nd_prefix, newpr, ndpr_entry);
    969 	splx(s);
    970 
    971 	/* ND_OPT_PI_FLAG_ONLINK processing */
    972 	if (newpr->ndpr_raf_onlink) {
    973 		int e;
    974 
    975 		if ((e = nd6_prefix_onlink(newpr)) != 0) {
    976 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
    977 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
    978 			    ip6_sprintf(&prc->ndprc_prefix.sin6_addr),
    979 			    prc->ndprc_plen, if_name(prc->ndprc_ifp), e));
    980 			/* proceed anyway. XXX: is it correct? */
    981 		}
    982 	}
    983 
    984 	if (dr)
    985 		pfxrtr_add(newpr, dr);
    986 
    987 	ext->nprefixes++;
    988 
    989 	return 0;
    990 }
    991 
    992 void
    993 prelist_remove(struct nd_prefix *pr)
    994 {
    995 	struct nd_pfxrouter *pfr, *next;
    996 	int e, s;
    997 	struct in6_ifextra *ext = pr->ndpr_ifp->if_afdata[AF_INET6];
    998 
    999 	/* make sure to invalidate the prefix until it is really freed. */
   1000 	pr->ndpr_vltime = 0;
   1001 	pr->ndpr_pltime = 0;
   1002 #if 0
   1003 	/*
   1004 	 * Though these flags are now meaningless, we'd rather keep the value
   1005 	 * not to confuse users when executing "ndp -p".
   1006 	 */
   1007 	pr->ndpr_raf_onlink = 0;
   1008 	pr->ndpr_raf_auto = 0;
   1009 #endif
   1010 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
   1011 	    (e = nd6_prefix_offlink(pr)) != 0) {
   1012 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
   1013 		    "on %s, errno=%d\n",
   1014 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1015 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
   1016 		/* what should we do? */
   1017 	}
   1018 
   1019 	if (pr->ndpr_refcnt > 0)
   1020 		return;		/* notice here? */
   1021 
   1022 	s = splsoftnet();
   1023 	/* unlink ndpr_entry from nd_prefix list */
   1024 	LIST_REMOVE(pr, ndpr_entry);
   1025 
   1026 	/* free list of routers that adversed the prefix */
   1027 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
   1028 		next = LIST_NEXT(pfr, pfr_entry);
   1029 
   1030 		free(pfr, M_IP6NDP);
   1031 	}
   1032 
   1033 	if (ext) {
   1034 		ext->nprefixes--;
   1035 		if (ext->nprefixes < 0) {
   1036 			log(LOG_WARNING, "prelist_remove: negative count on "
   1037 			    "%s\n", pr->ndpr_ifp->if_xname);
   1038 		}
   1039 	}
   1040 	splx(s);
   1041 
   1042 	free(pr, M_IP6NDP);
   1043 
   1044 	pfxlist_onlink_check();
   1045 }
   1046 
   1047 static int
   1048 prelist_update(struct nd_prefixctl *newprc,
   1049 	struct nd_defrouter *dr, /* may be NULL */
   1050 	struct mbuf *m,
   1051 	int mcast)
   1052 {
   1053 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
   1054 	struct ifaddr *ifa;
   1055 	struct ifnet *ifp = newprc->ndprc_ifp;
   1056 	struct nd_prefix *pr;
   1057 	int s = splsoftnet();
   1058 	int error = 0;
   1059 	int auth;
   1060 	struct in6_addrlifetime lt6_tmp;
   1061 
   1062 	auth = 0;
   1063 	if (m) {
   1064 		/*
   1065 		 * Authenticity for NA consists authentication for
   1066 		 * both IP header and IP datagrams, doesn't it ?
   1067 		 */
   1068 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
   1069 		auth = (m->m_flags & M_AUTHIPHDR
   1070 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
   1071 #endif
   1072 	}
   1073 
   1074 	if ((pr = nd6_prefix_lookup(newprc)) != NULL) {
   1075 		/*
   1076 		 * nd6_prefix_lookup() ensures that pr and newprc have the same
   1077 		 * prefix on a same interface.
   1078 		 */
   1079 
   1080 		/*
   1081 		 * Update prefix information.  Note that the on-link (L) bit
   1082 		 * and the autonomous (A) bit should NOT be changed from 1
   1083 		 * to 0.
   1084 		 */
   1085 		if (newprc->ndprc_raf_onlink == 1)
   1086 			pr->ndpr_raf_onlink = 1;
   1087 		if (newprc->ndprc_raf_auto == 1)
   1088 			pr->ndpr_raf_auto = 1;
   1089 		if (newprc->ndprc_raf_onlink) {
   1090 			pr->ndpr_vltime = newprc->ndprc_vltime;
   1091 			pr->ndpr_pltime = newprc->ndprc_pltime;
   1092 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
   1093 			pr->ndpr_lastupdate = time_uptime;
   1094 		}
   1095 
   1096 		if (newprc->ndprc_raf_onlink &&
   1097 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
   1098 			int e;
   1099 
   1100 			if ((e = nd6_prefix_onlink(pr)) != 0) {
   1101 				nd6log((LOG_ERR,
   1102 				    "%s: failed to make "
   1103 				    "the prefix %s/%d on-link on %s "
   1104 				    "(errno=%d)\n", __func__,
   1105 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1106 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
   1107 				/* proceed anyway. XXX: is it correct? */
   1108 			}
   1109 		}
   1110 
   1111 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
   1112 			pfxrtr_add(pr, dr);
   1113 	} else {
   1114 		struct nd_prefix *newpr = NULL;
   1115 
   1116 		if (newprc->ndprc_vltime == 0)
   1117 			goto end;
   1118 		if (newprc->ndprc_raf_onlink == 0 && newprc->ndprc_raf_auto == 0)
   1119 			goto end;
   1120 
   1121 		if (ip6_rtadv_maxroutes <= nd6_numroutes) {
   1122 			ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
   1123 			goto end;
   1124 		}
   1125 
   1126 		error = nd6_prelist_add(newprc, dr, &newpr);
   1127 		if (error != 0 || newpr == NULL) {
   1128 			nd6log((LOG_NOTICE,
   1129 			    "%s: nd6_prelist_add failed for %s/%d on %s "
   1130 			    "errno=%d, returnpr=%p\n", __func__,
   1131 			    ip6_sprintf(&newprc->ndprc_prefix.sin6_addr),
   1132 			    newprc->ndprc_plen, if_name(newprc->ndprc_ifp),
   1133 			    error, newpr));
   1134 			goto end; /* we should just give up in this case. */
   1135 		}
   1136 
   1137 		/*
   1138 		 * XXX: from the ND point of view, we can ignore a prefix
   1139 		 * with the on-link bit being zero.  However, we need a
   1140 		 * prefix structure for references from autoconfigured
   1141 		 * addresses.  Thus, we explicitly make sure that the prefix
   1142 		 * itself expires now.
   1143 		 */
   1144 		if (newpr->ndpr_raf_onlink == 0) {
   1145 			newpr->ndpr_vltime = 0;
   1146 			newpr->ndpr_pltime = 0;
   1147 			in6_init_prefix_ltimes(newpr);
   1148 		}
   1149 
   1150 		pr = newpr;
   1151 	}
   1152 
   1153 	/*
   1154 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
   1155 	 * Note that pr must be non NULL at this point.
   1156 	 */
   1157 
   1158 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
   1159 	if (!newprc->ndprc_raf_auto)
   1160 		goto end;
   1161 
   1162 	/*
   1163 	 * 5.5.3 (b). the link-local prefix should have been ignored in
   1164 	 * nd6_ra_input.
   1165 	 */
   1166 
   1167 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
   1168 	if (newprc->ndprc_pltime > newprc->ndprc_vltime) {
   1169 		error = EINVAL;	/* XXX: won't be used */
   1170 		goto end;
   1171 	}
   1172 
   1173 	/*
   1174 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
   1175 	 * an address configured by stateless autoconfiguration already in the
   1176 	 * list of addresses associated with the interface, and the Valid
   1177 	 * Lifetime is not 0, form an address.  We first check if we have
   1178 	 * a matching prefix.
   1179 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
   1180 	 * consider autoconfigured addresses while RFC2462 simply said
   1181 	 * "address".
   1182 	 */
   1183 	IFADDR_FOREACH(ifa, ifp) {
   1184 		struct in6_ifaddr *ifa6;
   1185 		u_int32_t remaininglifetime;
   1186 
   1187 		if (ifa->ifa_addr->sa_family != AF_INET6)
   1188 			continue;
   1189 
   1190 		ifa6 = (struct in6_ifaddr *)ifa;
   1191 
   1192 		/*
   1193 		 * We only consider autoconfigured addresses as per rfc2462bis.
   1194 		 */
   1195 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
   1196 			continue;
   1197 
   1198 		/*
   1199 		 * Spec is not clear here, but I believe we should concentrate
   1200 		 * on unicast (i.e. not anycast) addresses.
   1201 		 * XXX: other ia6_flags? detached or duplicated?
   1202 		 */
   1203 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
   1204 			continue;
   1205 
   1206 		/*
   1207 		 * Ignore the address if it is not associated with a prefix
   1208 		 * or is associated with a prefix that is different from this
   1209 		 * one.  (pr is never NULL here)
   1210 		 */
   1211 		if (ifa6->ia6_ndpr != pr)
   1212 			continue;
   1213 
   1214 		if (ia6_match == NULL) /* remember the first one */
   1215 			ia6_match = ifa6;
   1216 
   1217 		/*
   1218 		 * An already autoconfigured address matched.  Now that we
   1219 		 * are sure there is at least one matched address, we can
   1220 		 * proceed to 5.5.3. (e): update the lifetimes according to the
   1221 		 * "two hours" rule and the privacy extension.
   1222 		 * We apply some clarifications in rfc2462bis:
   1223 		 * - use remaininglifetime instead of storedlifetime as a
   1224 		 *   variable name
   1225 		 * - remove the dead code in the "two-hour" rule
   1226 		 */
   1227 #define TWOHOUR		(120*60)
   1228 		lt6_tmp = ifa6->ia6_lifetime;
   1229 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
   1230 			remaininglifetime = ND6_INFINITE_LIFETIME;
   1231 		else if (time_uptime - ifa6->ia6_updatetime >
   1232 			 lt6_tmp.ia6t_vltime) {
   1233 			/*
   1234 			 * The case of "invalid" address.  We should usually
   1235 			 * not see this case.
   1236 			 */
   1237 			remaininglifetime = 0;
   1238 		} else
   1239 			remaininglifetime = lt6_tmp.ia6t_vltime -
   1240 			    (time_uptime - ifa6->ia6_updatetime);
   1241 
   1242 		/* when not updating, keep the current stored lifetime. */
   1243 		lt6_tmp.ia6t_vltime = remaininglifetime;
   1244 
   1245 		if (TWOHOUR < newprc->ndprc_vltime ||
   1246 		    remaininglifetime < newprc->ndprc_vltime) {
   1247 			lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
   1248 		} else if (remaininglifetime <= TWOHOUR) {
   1249 			if (auth)
   1250 				lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
   1251 		} else {
   1252 			/*
   1253 			 * newprc->ndprc_vltime <= TWOHOUR &&
   1254 			 * TWOHOUR < remaininglifetime
   1255 			 */
   1256 			lt6_tmp.ia6t_vltime = TWOHOUR;
   1257 		}
   1258 
   1259 		/* The 2 hour rule is not imposed for preferred lifetime. */
   1260 		lt6_tmp.ia6t_pltime = newprc->ndprc_pltime;
   1261 
   1262 		in6_init_address_ltimes(pr, &lt6_tmp);
   1263 
   1264 		/*
   1265 		 * We need to treat lifetimes for temporary addresses
   1266 		 * differently, according to
   1267 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
   1268 		 * we only update the lifetimes when they are in the maximum
   1269 		 * intervals.
   1270 		 */
   1271 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
   1272 			u_int32_t maxvltime, maxpltime;
   1273 
   1274 			if (ip6_temp_valid_lifetime >
   1275 			    (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
   1276 			    ip6_desync_factor)) {
   1277 				maxvltime = ip6_temp_valid_lifetime -
   1278 				    (time_uptime - ifa6->ia6_createtime) -
   1279 				    ip6_desync_factor;
   1280 			} else
   1281 				maxvltime = 0;
   1282 			if (ip6_temp_preferred_lifetime >
   1283 			    (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
   1284 			    ip6_desync_factor)) {
   1285 				maxpltime = ip6_temp_preferred_lifetime -
   1286 				    (time_uptime - ifa6->ia6_createtime) -
   1287 				    ip6_desync_factor;
   1288 			} else
   1289 				maxpltime = 0;
   1290 
   1291 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
   1292 			    lt6_tmp.ia6t_vltime > maxvltime) {
   1293 				lt6_tmp.ia6t_vltime = maxvltime;
   1294 			}
   1295 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
   1296 			    lt6_tmp.ia6t_pltime > maxpltime) {
   1297 				lt6_tmp.ia6t_pltime = maxpltime;
   1298 			}
   1299 		}
   1300 
   1301 		ifa6->ia6_lifetime = lt6_tmp;
   1302 		ifa6->ia6_updatetime = time_uptime;
   1303 	}
   1304 	if (ia6_match == NULL && newprc->ndprc_vltime) {
   1305 		int ifidlen;
   1306 
   1307 		/*
   1308 		 * 5.5.3 (d) (continued)
   1309 		 * No address matched and the valid lifetime is non-zero.
   1310 		 * Create a new address.
   1311 		 */
   1312 
   1313 		/*
   1314 		 * Prefix Length check:
   1315 		 * If the sum of the prefix length and interface identifier
   1316 		 * length does not equal 128 bits, the Prefix Information
   1317 		 * option MUST be ignored.  The length of the interface
   1318 		 * identifier is defined in a separate link-type specific
   1319 		 * document.
   1320 		 */
   1321 		ifidlen = in6_if2idlen(ifp);
   1322 		if (ifidlen < 0) {
   1323 			/* this should not happen, so we always log it. */
   1324 			log(LOG_ERR, "%s: IFID undefined (%s)\n",
   1325 			    __func__, if_name(ifp));
   1326 			goto end;
   1327 		}
   1328 		if (ifidlen + pr->ndpr_plen != 128) {
   1329 			nd6log((LOG_INFO,
   1330 			    "%s: invalid prefixlen %d for %s, ignored\n",
   1331 			    __func__, pr->ndpr_plen, if_name(ifp)));
   1332 			goto end;
   1333 		}
   1334 
   1335 		if ((ia6 = in6_ifadd(newprc, mcast)) != NULL) {
   1336 			/*
   1337 			 * note that we should use pr (not newprc) for reference.
   1338 			 */
   1339 			pr->ndpr_refcnt++;
   1340 			ia6->ia6_ndpr = pr;
   1341 
   1342 			/*
   1343 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
   1344 			 * When a new public address is created as described
   1345 			 * in RFC2462, also create a new temporary address.
   1346 			 *
   1347 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
   1348 			 * When an interface connects to a new link, a new
   1349 			 * randomized interface identifier should be generated
   1350 			 * immediately together with a new set of temporary
   1351 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
   1352 			 * in6_tmpifadd().
   1353 			 */
   1354 			if (ip6_use_tempaddr) {
   1355 				int e;
   1356 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
   1357 					nd6log((LOG_NOTICE,
   1358 					    "%s: failed to create a temporary "
   1359 					    "address, errno=%d\n", __func__,
   1360 					    e));
   1361 				}
   1362 			}
   1363 
   1364 			/*
   1365 			 * A newly added address might affect the status
   1366 			 * of other addresses, so we check and update it.
   1367 			 * XXX: what if address duplication happens?
   1368 			 */
   1369 			pfxlist_onlink_check();
   1370 		} else {
   1371 			/* just set an error. do not bark here. */
   1372 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
   1373 		}
   1374 	}
   1375 
   1376  end:
   1377 	splx(s);
   1378 	return error;
   1379 }
   1380 
   1381 /*
   1382  * A supplement function used in the on-link detection below;
   1383  * detect if a given prefix has a (probably) reachable advertising router.
   1384  * XXX: lengthy function name...
   1385  */
   1386 static struct nd_pfxrouter *
   1387 find_pfxlist_reachable_router(struct nd_prefix *pr)
   1388 {
   1389 	struct nd_pfxrouter *pfxrtr;
   1390 
   1391 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
   1392 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
   1393 		if (pfxrtr->router->ifp->if_flags & IFF_UP &&
   1394 		    pfxrtr->router->ifp->if_link_state != LINK_STATE_DOWN &&
   1395 		    nd6_is_llinfo_probreach(pfxrtr->router))
   1396 			break;	/* found */
   1397 	}
   1398 
   1399 	return (pfxrtr);
   1400 }
   1401 
   1402 /*
   1403  * Check if each prefix in the prefix list has at least one available router
   1404  * that advertised the prefix (a router is "available" if its neighbor cache
   1405  * entry is reachable or probably reachable).
   1406  * If the check fails, the prefix may be off-link, because, for example,
   1407  * we have moved from the network but the lifetime of the prefix has not
   1408  * expired yet.  So we should not use the prefix if there is another prefix
   1409  * that has an available router.
   1410  * But, if there is no prefix that has an available router, we still regards
   1411  * all the prefixes as on-link.  This is because we can't tell if all the
   1412  * routers are simply dead or if we really moved from the network and there
   1413  * is no router around us.
   1414  */
   1415 void
   1416 pfxlist_onlink_check(void)
   1417 {
   1418 	struct nd_prefix *pr;
   1419 	struct in6_ifaddr *ifa;
   1420 	struct nd_defrouter *dr;
   1421 	struct nd_pfxrouter *pfxrtr = NULL;
   1422 
   1423 	/*
   1424 	 * Check if there is a prefix that has a reachable advertising
   1425 	 * router.
   1426 	 */
   1427 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
   1428 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
   1429 			break;
   1430 	}
   1431 	/*
   1432 	 * If we have no such prefix, check whether we still have a router
   1433 	 * that does not advertise any prefixes.
   1434 	 */
   1435 	if (pr == NULL) {
   1436 		TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
   1437 			struct nd_prefix *pr0;
   1438 
   1439 			LIST_FOREACH(pr0, &nd_prefix, ndpr_entry) {
   1440 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
   1441 					break;
   1442 			}
   1443 			if (pfxrtr)
   1444 				break;
   1445 		}
   1446 	}
   1447 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
   1448 		/*
   1449 		 * There is at least one prefix that has a reachable router,
   1450 		 * or at least a router which probably does not advertise
   1451 		 * any prefixes.  The latter would be the case when we move
   1452 		 * to a new link where we have a router that does not provide
   1453 		 * prefixes and we configure an address by hand.
   1454 		 * Detach prefixes which have no reachable advertising
   1455 		 * router, and attach other prefixes.
   1456 		 */
   1457 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
   1458 			/* XXX: a link-local prefix should never be detached */
   1459 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
   1460 				continue;
   1461 
   1462 			/*
   1463 			 * we aren't interested in prefixes without the L bit
   1464 			 * set.
   1465 			 */
   1466 			if (pr->ndpr_raf_onlink == 0)
   1467 				continue;
   1468 
   1469 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
   1470 			    find_pfxlist_reachable_router(pr) == NULL)
   1471 				pr->ndpr_stateflags |= NDPRF_DETACHED;
   1472 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
   1473 			    find_pfxlist_reachable_router(pr) != 0)
   1474 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
   1475 		}
   1476 	} else {
   1477 		/* there is no prefix that has a reachable router */
   1478 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
   1479 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
   1480 				continue;
   1481 
   1482 			if (pr->ndpr_raf_onlink == 0)
   1483 				continue;
   1484 
   1485 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
   1486 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
   1487 		}
   1488 	}
   1489 
   1490 	/*
   1491 	 * Remove each interface route associated with a (just) detached
   1492 	 * prefix, and reinstall the interface route for a (just) attached
   1493 	 * prefix.  Note that all attempt of reinstallation does not
   1494 	 * necessarily success, when a same prefix is shared among multiple
   1495 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
   1496 	 * so we don't have to care about them.
   1497 	 */
   1498 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
   1499 		int e;
   1500 
   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_ONLINK) != 0) {
   1509 			if ((e = nd6_prefix_offlink(pr)) != 0) {
   1510 				nd6log((LOG_ERR,
   1511 				    "pfxlist_onlink_check: failed to "
   1512 				    "make %s/%d offlink, errno=%d\n",
   1513 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1514 				    pr->ndpr_plen, e));
   1515 			}
   1516 		}
   1517 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
   1518 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
   1519 		    pr->ndpr_raf_onlink) {
   1520 			if ((e = nd6_prefix_onlink(pr)) != 0) {
   1521 				nd6log((LOG_ERR,
   1522 				    "pfxlist_onlink_check: failed to "
   1523 				    "make %s/%d onlink, errno=%d\n",
   1524 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1525 				    pr->ndpr_plen, e));
   1526 			}
   1527 		}
   1528 	}
   1529 
   1530 	/*
   1531 	 * Changes on the prefix status might affect address status as well.
   1532 	 * Make sure that all addresses derived from an attached prefix are
   1533 	 * attached, and that all addresses derived from a detached prefix are
   1534 	 * detached.  Note, however, that a manually configured address should
   1535 	 * always be attached.
   1536 	 * The precise detection logic is same as the one for prefixes.
   1537 	 */
   1538 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
   1539 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
   1540 			continue;
   1541 
   1542 		if (ifa->ia6_ndpr == NULL) {
   1543 			/*
   1544 			 * This can happen when we first configure the address
   1545 			 * (i.e. the address exists, but the prefix does not).
   1546 			 * XXX: complicated relationships...
   1547 			 */
   1548 			continue;
   1549 		}
   1550 
   1551 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
   1552 			break;
   1553 	}
   1554 	if (ifa) {
   1555 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
   1556 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
   1557 				continue;
   1558 
   1559 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
   1560 				continue;
   1561 
   1562 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
   1563 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
   1564 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
   1565 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
   1566 					nd6_dad_start((struct ifaddr *)ifa,
   1567 					    0);
   1568 					/* We will notify the routing socket
   1569 					 * of the DAD result, so no need to
   1570 					 * here */
   1571 				}
   1572 			} else {
   1573 				if ((ifa->ia6_flags & IN6_IFF_DETACHED) == 0) {
   1574 					ifa->ia6_flags |= IN6_IFF_DETACHED;
   1575 					rt_newaddrmsg(RTM_NEWADDR,
   1576 					    (struct ifaddr *)ifa, 0, NULL);
   1577 				}
   1578 			}
   1579 		}
   1580 	}
   1581 	else {
   1582 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
   1583 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
   1584 				continue;
   1585 
   1586 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
   1587 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
   1588 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
   1589 				/* Do we need a delay in this case? */
   1590 				nd6_dad_start((struct ifaddr *)ifa, 0);
   1591 			}
   1592 		}
   1593 	}
   1594 }
   1595 
   1596 int
   1597 nd6_prefix_onlink(struct nd_prefix *pr)
   1598 {
   1599 	struct ifaddr *ifa;
   1600 	struct ifnet *ifp = pr->ndpr_ifp;
   1601 	struct sockaddr_in6 mask6;
   1602 	struct nd_prefix *opr;
   1603 	u_long rtflags;
   1604 	int error = 0;
   1605 
   1606 	/* sanity check */
   1607 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
   1608 		nd6log((LOG_ERR,
   1609 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
   1610 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
   1611 		return (EEXIST);
   1612 	}
   1613 
   1614 	/*
   1615 	 * Add the interface route associated with the prefix.  Before
   1616 	 * installing the route, check if there's the same prefix on another
   1617 	 * interface, and the prefix has already installed the interface route.
   1618 	 * Although such a configuration is expected to be rare, we explicitly
   1619 	 * allow it.
   1620 	 */
   1621 	LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
   1622 		if (opr == pr)
   1623 			continue;
   1624 
   1625 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
   1626 			continue;
   1627 
   1628 		if (opr->ndpr_plen == pr->ndpr_plen &&
   1629 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
   1630 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
   1631 			return (0);
   1632 	}
   1633 
   1634 	/*
   1635 	 * We prefer link-local addresses as the associated interface address.
   1636 	 */
   1637 	/* search for a link-local addr */
   1638 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
   1639 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
   1640 	if (ifa == NULL) {
   1641 		/* XXX: freebsd does not have ifa_ifwithaf */
   1642 		IFADDR_FOREACH(ifa, ifp) {
   1643 			if (ifa->ifa_addr->sa_family == AF_INET6)
   1644 				break;
   1645 		}
   1646 		/* should we care about ia6_flags? */
   1647 	}
   1648 	if (ifa == NULL) {
   1649 		/*
   1650 		 * This can still happen, when, for example, we receive an RA
   1651 		 * containing a prefix with the L bit set and the A bit clear,
   1652 		 * after removing all IPv6 addresses on the receiving
   1653 		 * interface.  This should, of course, be rare though.
   1654 		 */
   1655 		nd6log((LOG_NOTICE,
   1656 		    "nd6_prefix_onlink: failed to find any ifaddr"
   1657 		    " to add route for a prefix(%s/%d) on %s\n",
   1658 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1659 		    pr->ndpr_plen, if_name(ifp)));
   1660 		return (0);
   1661 	}
   1662 
   1663 	/*
   1664 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
   1665 	 * ifa->ifa_rtrequest = nd6_rtrequest;
   1666 	 */
   1667 	memset(&mask6, 0, sizeof(mask6));
   1668 	mask6.sin6_family = AF_INET6;
   1669 	mask6.sin6_len = sizeof(mask6);
   1670 	mask6.sin6_addr = pr->ndpr_mask;
   1671 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
   1672 	rtflags = ifa->ifa_flags | RTF_UP;
   1673 	if (nd6_need_cache(ifp)) {
   1674 		/* explicitly set in case ifa_flags does not set the flag. */
   1675 		rtflags |= RTF_CLONING;
   1676 	} else {
   1677 		/*
   1678 		 * explicitly clear the cloning bit in case ifa_flags sets it.
   1679 		 */
   1680 		rtflags &= ~RTF_CLONING;
   1681 	}
   1682 	error = rtrequest_newmsg(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
   1683 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags);
   1684 	if (error == 0) {
   1685 		nd6_numroutes++;
   1686 		pr->ndpr_stateflags |= NDPRF_ONLINK;
   1687 	} else {
   1688 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
   1689 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
   1690 		    "errno = %d\n",
   1691 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1692 		    pr->ndpr_plen, if_name(ifp),
   1693 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
   1694 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
   1695 	}
   1696 
   1697 	return (error);
   1698 }
   1699 
   1700 int
   1701 nd6_prefix_offlink(struct nd_prefix *pr)
   1702 {
   1703 	int error = 0;
   1704 	struct ifnet *ifp = pr->ndpr_ifp;
   1705 	struct nd_prefix *opr;
   1706 	struct sockaddr_in6 sa6, mask6;
   1707 
   1708 	/* sanity check */
   1709 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
   1710 		nd6log((LOG_ERR,
   1711 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
   1712 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
   1713 		return (EEXIST);
   1714 	}
   1715 
   1716 	sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
   1717 	sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
   1718 	error = rtrequest_newmsg(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
   1719 	    (struct sockaddr *)&mask6, 0);
   1720 	if (error == 0) {
   1721 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
   1722 		nd6_numroutes--;
   1723 
   1724 		/*
   1725 		 * There might be the same prefix on another interface,
   1726 		 * the prefix which could not be on-link just because we have
   1727 		 * the interface route (see comments in nd6_prefix_onlink).
   1728 		 * If there's one, try to make the prefix on-link on the
   1729 		 * interface.
   1730 		 */
   1731 		LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
   1732 			if (opr == pr)
   1733 				continue;
   1734 
   1735 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
   1736 				continue;
   1737 
   1738 			/*
   1739 			 * KAME specific: detached prefixes should not be
   1740 			 * on-link.
   1741 			 */
   1742 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
   1743 				continue;
   1744 
   1745 			if (opr->ndpr_plen == pr->ndpr_plen &&
   1746 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
   1747 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
   1748 				int e;
   1749 
   1750 				if ((e = nd6_prefix_onlink(opr)) != 0) {
   1751 					nd6log((LOG_ERR,
   1752 					    "nd6_prefix_offlink: failed to "
   1753 					    "recover a prefix %s/%d from %s "
   1754 					    "to %s (errno = %d)\n",
   1755 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
   1756 					    opr->ndpr_plen, if_name(ifp),
   1757 					    if_name(opr->ndpr_ifp), e));
   1758 				}
   1759 			}
   1760 		}
   1761 	} else {
   1762 		/* XXX: can we still set the NDPRF_ONLINK flag? */
   1763 		nd6log((LOG_ERR,
   1764 		    "nd6_prefix_offlink: failed to delete route: "
   1765 		    "%s/%d on %s (errno = %d)\n",
   1766 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
   1767 		    error));
   1768 	}
   1769 
   1770 	return error;
   1771 }
   1772 
   1773 static struct in6_ifaddr *
   1774 in6_ifadd(struct nd_prefixctl *prc, int mcast)
   1775 {
   1776 	struct ifnet *ifp = prc->ndprc_ifp;
   1777 	struct ifaddr *ifa;
   1778 	struct in6_aliasreq ifra;
   1779 	struct in6_ifaddr *ia, *ib;
   1780 	int error, plen0;
   1781 	struct in6_addr mask;
   1782 	int prefixlen = prc->ndprc_plen;
   1783 	int updateflags;
   1784 
   1785 	in6_prefixlen2mask(&mask, prefixlen);
   1786 
   1787 	/*
   1788 	 * find a link-local address (will be interface ID).
   1789 	 * Is it really mandatory? Theoretically, a global or a site-local
   1790 	 * address can be configured without a link-local address, if we
   1791 	 * have a unique interface identifier...
   1792 	 *
   1793 	 * it is not mandatory to have a link-local address, we can generate
   1794 	 * interface identifier on the fly.  we do this because:
   1795 	 * (1) it should be the easiest way to find interface identifier.
   1796 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
   1797 	 * for multiple addresses on a single interface, and possible shortcut
   1798 	 * of DAD.  we omitted DAD for this reason in the past.
   1799 	 * (3) a user can prevent autoconfiguration of global address
   1800 	 * by removing link-local address by hand (this is partly because we
   1801 	 * don't have other way to control the use of IPv6 on an interface.
   1802 	 * this has been our design choice - cf. NRL's "ifconfig auto").
   1803 	 * (4) it is easier to manage when an interface has addresses
   1804 	 * with the same interface identifier, than to have multiple addresses
   1805 	 * with different interface identifiers.
   1806 	 */
   1807 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
   1808 	if (ifa)
   1809 		ib = (struct in6_ifaddr *)ifa;
   1810 	else
   1811 		return NULL;
   1812 
   1813 #if 0 /* don't care link local addr state, and always do DAD */
   1814 	/* if link-local address is not eligible, do not autoconfigure. */
   1815 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
   1816 		printf("in6_ifadd: link-local address not ready\n");
   1817 		return NULL;
   1818 	}
   1819 #endif
   1820 
   1821 	/* prefixlen + ifidlen must be equal to 128 */
   1822 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
   1823 	if (prefixlen != plen0) {
   1824 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
   1825 		    "(prefix=%d ifid=%d)\n",
   1826 		    if_name(ifp), prefixlen, 128 - plen0));
   1827 		return NULL;
   1828 	}
   1829 
   1830 	/* make ifaddr */
   1831 
   1832 	memset(&ifra, 0, sizeof(ifra));
   1833 	/*
   1834 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
   1835 	 * for safety.
   1836 	 */
   1837 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
   1838 	sockaddr_in6_init(&ifra.ifra_addr, &prc->ndprc_prefix.sin6_addr, 0, 0, 0);
   1839 	/* prefix */
   1840 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
   1841 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
   1842 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
   1843 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
   1844 
   1845 	/* interface ID */
   1846 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
   1847 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
   1848 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
   1849 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
   1850 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
   1851 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
   1852 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
   1853 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
   1854 
   1855 	/* new prefix mask. */
   1856 	sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
   1857 
   1858 	/* lifetimes */
   1859 	ifra.ifra_lifetime.ia6t_vltime = prc->ndprc_vltime;
   1860 	ifra.ifra_lifetime.ia6t_pltime = prc->ndprc_pltime;
   1861 
   1862 	/* XXX: scope zone ID? */
   1863 
   1864 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
   1865 
   1866 	/*
   1867 	 * Make sure that we do not have this address already.  This should
   1868 	 * usually not happen, but we can still see this case, e.g., if we
   1869 	 * have manually configured the exact address to be configured.
   1870 	 */
   1871 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
   1872 		/* this should be rare enough to make an explicit log */
   1873 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
   1874 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
   1875 		return (NULL);
   1876 	}
   1877 
   1878 	/*
   1879 	 * Allocate ifaddr structure, link into chain, etc.
   1880 	 * If we are going to create a new address upon receiving a multicasted
   1881 	 * RA, we need to impose a random delay before starting DAD.
   1882 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
   1883 	 */
   1884 	updateflags = 0;
   1885 	if (mcast)
   1886 		updateflags |= IN6_IFAUPDATE_DADDELAY;
   1887 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
   1888 		nd6log((LOG_ERR,
   1889 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
   1890 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
   1891 		    error));
   1892 		return (NULL);	/* ifaddr must not have been allocated. */
   1893 	}
   1894 
   1895 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
   1896 
   1897 	return (ia);		/* this is always non-NULL */
   1898 }
   1899 
   1900 int
   1901 in6_tmpifadd(
   1902 	const struct in6_ifaddr *ia0, /* corresponding public address */
   1903 	int forcegen,
   1904 	int dad_delay)
   1905 {
   1906 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
   1907 	struct in6_ifaddr *newia, *ia;
   1908 	struct in6_aliasreq ifra;
   1909 	int i, error;
   1910 	int trylimit = 3;	/* XXX: adhoc value */
   1911 	int updateflags;
   1912 	u_int32_t randid[2];
   1913 	u_int32_t vltime0, pltime0;
   1914 
   1915 	memset(&ifra, 0, sizeof(ifra));
   1916 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
   1917 	ifra.ifra_addr = ia0->ia_addr;
   1918 	/* copy prefix mask */
   1919 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
   1920 	/* clear the old IFID */
   1921 	for (i = 0; i < 4; i++) {
   1922 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
   1923 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
   1924 	}
   1925 
   1926   again:
   1927 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
   1928 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
   1929 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
   1930 		    "random IFID\n"));
   1931 		return (EINVAL);
   1932 	}
   1933 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
   1934 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
   1935 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
   1936 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
   1937 
   1938 	/*
   1939 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
   1940 	 * However, we may still have a chance to see collision, because
   1941 	 * there may be a time lag between generation of the ID and generation
   1942 	 * of the address.  So, we'll do one more sanity check.
   1943 	 */
   1944 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
   1945 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
   1946 		    &ifra.ifra_addr.sin6_addr)) {
   1947 			if (trylimit-- == 0) {
   1948 				/*
   1949 				 * Give up.  Something strange should have
   1950 				 * happened.
   1951 				 */
   1952 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
   1953 				    "find a unique random IFID\n"));
   1954 				return (EEXIST);
   1955 			}
   1956 			forcegen = 1;
   1957 			goto again;
   1958 		}
   1959 	}
   1960 
   1961 	/*
   1962 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
   1963          * public address or TEMP_VALID_LIFETIME.
   1964 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
   1965          * of the public address or TEMP_PREFERRED_LIFETIME -
   1966          * DESYNC_FACTOR.
   1967 	 */
   1968 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
   1969 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
   1970 		    (ia0->ia6_lifetime.ia6t_vltime -
   1971 		    (time_uptime - ia0->ia6_updatetime));
   1972 		if (vltime0 > ip6_temp_valid_lifetime)
   1973 			vltime0 = ip6_temp_valid_lifetime;
   1974 	} else
   1975 		vltime0 = ip6_temp_valid_lifetime;
   1976 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
   1977 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
   1978 		    (ia0->ia6_lifetime.ia6t_pltime -
   1979 		    (time_uptime - ia0->ia6_updatetime));
   1980 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
   1981 			pltime0 = ip6_temp_preferred_lifetime -
   1982 			    ip6_desync_factor;
   1983 		}
   1984 	} else
   1985 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
   1986 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
   1987 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
   1988 
   1989 	/*
   1990 	 * A temporary address is created only if this calculated Preferred
   1991 	 * Lifetime is greater than REGEN_ADVANCE time units.
   1992 	 */
   1993 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
   1994 		return (0);
   1995 
   1996 	/* XXX: scope zone ID? */
   1997 
   1998 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
   1999 
   2000 	/* allocate ifaddr structure, link into chain, etc. */
   2001 	updateflags = 0;
   2002 	if (dad_delay)
   2003 		updateflags |= IN6_IFAUPDATE_DADDELAY;
   2004 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
   2005 		return (error);
   2006 
   2007 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
   2008 	if (newia == NULL) {	/* XXX: can it happen? */
   2009 		nd6log((LOG_ERR,
   2010 		    "in6_tmpifadd: ifa update succeeded, but we got "
   2011 		    "no ifaddr\n"));
   2012 		return (EINVAL); /* XXX */
   2013 	}
   2014 	newia->ia6_ndpr = ia0->ia6_ndpr;
   2015 	newia->ia6_ndpr->ndpr_refcnt++;
   2016 
   2017 	/*
   2018 	 * A newly added address might affect the status of other addresses.
   2019 	 * XXX: when the temporary address is generated with a new public
   2020 	 * address, the onlink check is redundant.  However, it would be safe
   2021 	 * to do the check explicitly everywhere a new address is generated,
   2022 	 * and, in fact, we surely need the check when we create a new
   2023 	 * temporary address due to deprecation of an old temporary address.
   2024 	 */
   2025 	pfxlist_onlink_check();
   2026 
   2027 	return (0);
   2028 }
   2029 
   2030 static int
   2031 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
   2032 {
   2033 
   2034 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
   2035 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
   2036 		nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
   2037 		    "(%d) is greater than valid lifetime(%d)\n",
   2038 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
   2039 		return (EINVAL);
   2040 	}
   2041 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
   2042 		ndpr->ndpr_preferred = 0;
   2043 	else
   2044 		ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
   2045 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
   2046 		ndpr->ndpr_expire = 0;
   2047 	else
   2048 		ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
   2049 
   2050 	return 0;
   2051 }
   2052 
   2053 static void
   2054 in6_init_address_ltimes(struct nd_prefix *newpr,
   2055     struct in6_addrlifetime *lt6)
   2056 {
   2057 
   2058 	/* Valid lifetime must not be updated unless explicitly specified. */
   2059 	/* init ia6t_expire */
   2060 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
   2061 		lt6->ia6t_expire = 0;
   2062 	else {
   2063 		lt6->ia6t_expire = time_uptime;
   2064 		lt6->ia6t_expire += lt6->ia6t_vltime;
   2065 	}
   2066 
   2067 	/* init ia6t_preferred */
   2068 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
   2069 		lt6->ia6t_preferred = 0;
   2070 	else {
   2071 		lt6->ia6t_preferred = time_uptime;
   2072 		lt6->ia6t_preferred += lt6->ia6t_pltime;
   2073 	}
   2074 }
   2075 
   2076 /*
   2077  * Delete all the routing table entries that use the specified gateway.
   2078  * XXX: this function causes search through all entries of routing table, so
   2079  * it shouldn't be called when acting as a router.
   2080  */
   2081 void
   2082 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
   2083 {
   2084 	int s = splsoftnet();
   2085 
   2086 	/* We'll care only link-local addresses */
   2087 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
   2088 		splx(s);
   2089 		return;
   2090 	}
   2091 
   2092 	rt_walktree(AF_INET6, rt6_deleteroute, (void *)gateway);
   2093 	splx(s);
   2094 }
   2095 
   2096 static int
   2097 rt6_deleteroute(struct rtentry *rt, void *arg)
   2098 {
   2099 	struct in6_addr *gate = (struct in6_addr *)arg;
   2100 
   2101 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
   2102 		return (0);
   2103 
   2104 	if (!IN6_ARE_ADDR_EQUAL(gate, &satosin6(rt->rt_gateway)->sin6_addr))
   2105 		return (0);
   2106 
   2107 	/*
   2108 	 * Do not delete a static route.
   2109 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
   2110 	 * 'cloned' bit instead?
   2111 	 */
   2112 	if ((rt->rt_flags & RTF_STATIC) != 0)
   2113 		return (0);
   2114 
   2115 	/*
   2116 	 * We delete only host route. This means, in particular, we don't
   2117 	 * delete default route.
   2118 	 */
   2119 	if ((rt->rt_flags & RTF_HOST) == 0)
   2120 		return (0);
   2121 
   2122 	return (rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
   2123 	    rt_mask(rt), rt->rt_flags, NULL));
   2124 }
   2125 
   2126 int
   2127 nd6_setdefaultiface(int ifindex)
   2128 {
   2129 	ifnet_t *ifp;
   2130 	int error = 0;
   2131 
   2132 	if ((ifp = if_byindex(ifindex)) == NULL) {
   2133 		return EINVAL;
   2134 	}
   2135 	if (nd6_defifindex != ifindex) {
   2136 		nd6_defifindex = ifindex;
   2137 		nd6_defifp = nd6_defifindex > 0 ? ifp : NULL;
   2138 
   2139 		/*
   2140 		 * Our current implementation assumes one-to-one maping between
   2141 		 * interfaces and links, so it would be natural to use the
   2142 		 * default interface as the default link.
   2143 		 */
   2144 		scope6_setdefault(nd6_defifp);
   2145 	}
   2146 
   2147 	return (error);
   2148 }
   2149