Home | History | Annotate | Line # | Download | only in net
route.c revision 1.155
      1 /*	$NetBSD: route.c,v 1.155 2016/04/01 02:00:14 ozaki-r Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1998, 2008 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Kevin M. Lahey of the Numerical Aerospace Simulation Facility,
      9  * NASA Ames Research Center.
     10  *
     11  * Redistribution and use in source and binary forms, with or without
     12  * modification, are permitted provided that the following conditions
     13  * are met:
     14  * 1. Redistributions of source code must retain the above copyright
     15  *    notice, this list of conditions and the following disclaimer.
     16  * 2. Redistributions in binary form must reproduce the above copyright
     17  *    notice, this list of conditions and the following disclaimer in the
     18  *    documentation and/or other materials provided with the distribution.
     19  *
     20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     30  * POSSIBILITY OF SUCH DAMAGE.
     31  */
     32 
     33 /*
     34  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
     35  * All rights reserved.
     36  *
     37  * Redistribution and use in source and binary forms, with or without
     38  * modification, are permitted provided that the following conditions
     39  * are met:
     40  * 1. Redistributions of source code must retain the above copyright
     41  *    notice, this list of conditions and the following disclaimer.
     42  * 2. Redistributions in binary form must reproduce the above copyright
     43  *    notice, this list of conditions and the following disclaimer in the
     44  *    documentation and/or other materials provided with the distribution.
     45  * 3. Neither the name of the project nor the names of its contributors
     46  *    may be used to endorse or promote products derived from this software
     47  *    without specific prior written permission.
     48  *
     49  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     59  * SUCH DAMAGE.
     60  */
     61 
     62 /*
     63  * Copyright (c) 1980, 1986, 1991, 1993
     64  *	The Regents of the University of California.  All rights reserved.
     65  *
     66  * Redistribution and use in source and binary forms, with or without
     67  * modification, are permitted provided that the following conditions
     68  * are met:
     69  * 1. Redistributions of source code must retain the above copyright
     70  *    notice, this list of conditions and the following disclaimer.
     71  * 2. Redistributions in binary form must reproduce the above copyright
     72  *    notice, this list of conditions and the following disclaimer in the
     73  *    documentation and/or other materials provided with the distribution.
     74  * 3. Neither the name of the University nor the names of its contributors
     75  *    may be used to endorse or promote products derived from this software
     76  *    without specific prior written permission.
     77  *
     78  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     79  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     80  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     81  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     82  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     83  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     84  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     85  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     86  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     87  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     88  * SUCH DAMAGE.
     89  *
     90  *	@(#)route.c	8.3 (Berkeley) 1/9/95
     91  */
     92 
     93 #ifdef _KERNEL_OPT
     94 #include "opt_inet.h"
     95 #include "opt_route.h"
     96 #endif
     97 
     98 #include <sys/cdefs.h>
     99 __KERNEL_RCSID(0, "$NetBSD: route.c,v 1.155 2016/04/01 02:00:14 ozaki-r Exp $");
    100 
    101 #include <sys/param.h>
    102 #ifdef RTFLUSH_DEBUG
    103 #include <sys/sysctl.h>
    104 #endif
    105 #include <sys/systm.h>
    106 #include <sys/callout.h>
    107 #include <sys/proc.h>
    108 #include <sys/mbuf.h>
    109 #include <sys/socket.h>
    110 #include <sys/socketvar.h>
    111 #include <sys/domain.h>
    112 #include <sys/protosw.h>
    113 #include <sys/kernel.h>
    114 #include <sys/ioctl.h>
    115 #include <sys/pool.h>
    116 #include <sys/kauth.h>
    117 
    118 #include <net/if.h>
    119 #include <net/if_dl.h>
    120 #include <net/route.h>
    121 
    122 #include <netinet/in.h>
    123 #include <netinet/in_var.h>
    124 
    125 #ifdef RTFLUSH_DEBUG
    126 #define	rtcache_debug() __predict_false(_rtcache_debug)
    127 #else /* RTFLUSH_DEBUG */
    128 #define	rtcache_debug() 0
    129 #endif /* RTFLUSH_DEBUG */
    130 
    131 struct rtstat		rtstat;
    132 
    133 static int		rttrash;	/* routes not in table but not freed */
    134 
    135 static struct pool	rtentry_pool;
    136 static struct pool	rttimer_pool;
    137 
    138 static struct callout	rt_timer_ch; /* callout for rt_timer_timer() */
    139 
    140 #ifdef RTFLUSH_DEBUG
    141 static int _rtcache_debug = 0;
    142 #endif /* RTFLUSH_DEBUG */
    143 
    144 static kauth_listener_t route_listener;
    145 
    146 static int rtdeletemsg(struct rtentry *);
    147 static int rtflushclone1(struct rtentry *, void *);
    148 static void rtflushclone(sa_family_t family, struct rtentry *);
    149 static void rtflushall(int);
    150 
    151 static void rt_maskedcopy(const struct sockaddr *,
    152     struct sockaddr *, const struct sockaddr *);
    153 
    154 static void rtcache_clear(struct route *);
    155 static void rtcache_invalidate(struct dom_rtlist *);
    156 
    157 #ifdef RTFLUSH_DEBUG
    158 static void sysctl_net_rtcache_setup(struct sysctllog **);
    159 static void
    160 sysctl_net_rtcache_setup(struct sysctllog **clog)
    161 {
    162 	const struct sysctlnode *rnode;
    163 
    164 	if (sysctl_createv(clog, 0, NULL, &rnode, CTLFLAG_PERMANENT,
    165 	    CTLTYPE_NODE,
    166 	    "rtcache", SYSCTL_DESCR("Route cache related settings"),
    167 	    NULL, 0, NULL, 0, CTL_NET, CTL_CREATE, CTL_EOL) != 0)
    168 		return;
    169 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    170 	    CTLFLAG_PERMANENT|CTLFLAG_READWRITE, CTLTYPE_INT,
    171 	    "debug", SYSCTL_DESCR("Debug route caches"),
    172 	    NULL, 0, &_rtcache_debug, 0, CTL_CREATE, CTL_EOL) != 0)
    173 		return;
    174 }
    175 #endif /* RTFLUSH_DEBUG */
    176 
    177 static inline void
    178 rt_destroy(struct rtentry *rt)
    179 {
    180 	if (rt->_rt_key != NULL)
    181 		sockaddr_free(rt->_rt_key);
    182 	if (rt->rt_gateway != NULL)
    183 		sockaddr_free(rt->rt_gateway);
    184 	if (rt_gettag(rt) != NULL)
    185 		sockaddr_free(rt_gettag(rt));
    186 	rt->_rt_key = rt->rt_gateway = rt->rt_tag = NULL;
    187 }
    188 
    189 static inline const struct sockaddr *
    190 rt_setkey(struct rtentry *rt, const struct sockaddr *key, int flags)
    191 {
    192 	if (rt->_rt_key == key)
    193 		goto out;
    194 
    195 	if (rt->_rt_key != NULL)
    196 		sockaddr_free(rt->_rt_key);
    197 	rt->_rt_key = sockaddr_dup(key, flags);
    198 out:
    199 	rt->rt_nodes->rn_key = (const char *)rt->_rt_key;
    200 	return rt->_rt_key;
    201 }
    202 
    203 struct ifaddr *
    204 rt_get_ifa(struct rtentry *rt)
    205 {
    206 	struct ifaddr *ifa;
    207 
    208 	if ((ifa = rt->rt_ifa) == NULL)
    209 		return ifa;
    210 	else if (ifa->ifa_getifa == NULL)
    211 		return ifa;
    212 #if 0
    213 	else if (ifa->ifa_seqno != NULL && *ifa->ifa_seqno == rt->rt_ifa_seqno)
    214 		return ifa;
    215 #endif
    216 	else {
    217 		ifa = (*ifa->ifa_getifa)(ifa, rt_getkey(rt));
    218 		if (ifa == NULL)
    219 			return NULL;
    220 		rt_replace_ifa(rt, ifa);
    221 		return ifa;
    222 	}
    223 }
    224 
    225 static void
    226 rt_set_ifa1(struct rtentry *rt, struct ifaddr *ifa)
    227 {
    228 	rt->rt_ifa = ifa;
    229 	if (ifa->ifa_seqno != NULL)
    230 		rt->rt_ifa_seqno = *ifa->ifa_seqno;
    231 }
    232 
    233 /*
    234  * Is this route the connected route for the ifa?
    235  */
    236 static int
    237 rt_ifa_connected(const struct rtentry *rt, const struct ifaddr *ifa)
    238 {
    239 	const struct sockaddr *key, *dst, *odst;
    240 	struct sockaddr_storage maskeddst;
    241 
    242 	key = rt_getkey(rt);
    243 	dst = rt->rt_flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
    244 	if (dst == NULL ||
    245 	    dst->sa_family != key->sa_family ||
    246 	    dst->sa_len != key->sa_len)
    247 		return 0;
    248 	if ((rt->rt_flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
    249 		odst = dst;
    250 		dst = (struct sockaddr *)&maskeddst;
    251 		rt_maskedcopy(odst, (struct sockaddr *)&maskeddst,
    252 		    ifa->ifa_netmask);
    253 	}
    254 	return (memcmp(dst, key, dst->sa_len) == 0);
    255 }
    256 
    257 void
    258 rt_replace_ifa(struct rtentry *rt, struct ifaddr *ifa)
    259 {
    260 	if (rt->rt_ifa &&
    261 	    rt->rt_ifa != ifa &&
    262 	    rt->rt_ifa->ifa_flags & IFA_ROUTE &&
    263 	    rt_ifa_connected(rt, rt->rt_ifa))
    264 	{
    265 		RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    266 		    "replace deleted IFA_ROUTE\n",
    267 		    (void *)rt->_rt_key, (void *)rt->rt_ifa);
    268 		rt->rt_ifa->ifa_flags &= ~IFA_ROUTE;
    269 		if (rt_ifa_connected(rt, ifa)) {
    270 			RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    271 			    "replace added IFA_ROUTE\n",
    272 			    (void *)rt->_rt_key, (void *)ifa);
    273 			ifa->ifa_flags |= IFA_ROUTE;
    274 		}
    275 	}
    276 
    277 	ifaref(ifa);
    278 	ifafree(rt->rt_ifa);
    279 	rt_set_ifa1(rt, ifa);
    280 }
    281 
    282 static void
    283 rt_set_ifa(struct rtentry *rt, struct ifaddr *ifa)
    284 {
    285 	ifaref(ifa);
    286 	rt_set_ifa1(rt, ifa);
    287 }
    288 
    289 static int
    290 route_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
    291     void *arg0, void *arg1, void *arg2, void *arg3)
    292 {
    293 	struct rt_msghdr *rtm;
    294 	int result;
    295 
    296 	result = KAUTH_RESULT_DEFER;
    297 	rtm = arg1;
    298 
    299 	if (action != KAUTH_NETWORK_ROUTE)
    300 		return result;
    301 
    302 	if (rtm->rtm_type == RTM_GET)
    303 		result = KAUTH_RESULT_ALLOW;
    304 
    305 	return result;
    306 }
    307 
    308 void
    309 rt_init(void)
    310 {
    311 
    312 #ifdef RTFLUSH_DEBUG
    313 	sysctl_net_rtcache_setup(NULL);
    314 #endif
    315 
    316 	pool_init(&rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl",
    317 	    NULL, IPL_SOFTNET);
    318 	pool_init(&rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl",
    319 	    NULL, IPL_SOFTNET);
    320 
    321 	rn_init();	/* initialize all zeroes, all ones, mask table */
    322 	rtbl_init();
    323 
    324 	route_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
    325 	    route_listener_cb, NULL);
    326 }
    327 
    328 static void
    329 rtflushall(int family)
    330 {
    331 	struct domain *dom;
    332 
    333 	if (rtcache_debug())
    334 		printf("%s: enter\n", __func__);
    335 
    336 	if ((dom = pffinddomain(family)) == NULL)
    337 		return;
    338 
    339 	rtcache_invalidate(&dom->dom_rtcache);
    340 }
    341 
    342 static void
    343 rtcache(struct route *ro)
    344 {
    345 	struct domain *dom;
    346 
    347 	rtcache_invariants(ro);
    348 	KASSERT(ro->_ro_rt != NULL);
    349 	KASSERT(ro->ro_invalid == false);
    350 	KASSERT(rtcache_getdst(ro) != NULL);
    351 
    352 	if ((dom = pffinddomain(rtcache_getdst(ro)->sa_family)) == NULL)
    353 		return;
    354 
    355 	LIST_INSERT_HEAD(&dom->dom_rtcache, ro, ro_rtcache_next);
    356 	rtcache_invariants(ro);
    357 }
    358 
    359 /*
    360  * Packet routing routines. If success, refcnt of a returned rtentry
    361  * will be incremented. The caller has to rtfree it by itself.
    362  */
    363 struct rtentry *
    364 rtalloc1(const struct sockaddr *dst, int report)
    365 {
    366 	rtbl_t *rtbl = rt_gettable(dst->sa_family);
    367 	struct rtentry *rt;
    368 	struct rtentry *newrt = NULL;
    369 	struct rt_addrinfo info;
    370 	int  s = splsoftnet(), err = 0, msgtype = RTM_MISS;
    371 
    372 	if (rtbl != NULL && (rt = rt_matchaddr(rtbl, dst)) != NULL) {
    373 		newrt = rt;
    374 		if (report && (rt->rt_flags & RTF_CLONING)) {
    375 			err = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
    376 			    &newrt);
    377 			if (err) {
    378 				newrt = rt;
    379 				rt->rt_refcnt++;
    380 				goto miss;
    381 			}
    382 			KASSERT(newrt != NULL);
    383 			rt = newrt;
    384 			if (rt->rt_flags & RTF_XRESOLVE) {
    385 				msgtype = RTM_RESOLVE;
    386 				goto miss;
    387 			}
    388 			/* Inform listeners of the new route */
    389 			memset(&info, 0, sizeof(info));
    390 			info.rti_info[RTAX_DST] = rt_getkey(rt);
    391 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    392 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    393 			if (rt->rt_ifp != NULL) {
    394 				info.rti_info[RTAX_IFP] =
    395 				    rt->rt_ifp->if_dl->ifa_addr;
    396 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
    397 			}
    398 			rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0);
    399 		} else
    400 			rt->rt_refcnt++;
    401 	} else {
    402 		rtstat.rts_unreach++;
    403 	miss:	if (report) {
    404 			memset((void *)&info, 0, sizeof(info));
    405 			info.rti_info[RTAX_DST] = dst;
    406 			rt_missmsg(msgtype, &info, 0, err);
    407 		}
    408 	}
    409 	splx(s);
    410 	return newrt;
    411 }
    412 
    413 #ifdef DEBUG
    414 /*
    415  * Check the following constraint for each rtcache:
    416  *   if a rtcache holds a rtentry, the rtentry's refcnt is more than zero,
    417  *   i.e., the rtentry should be referenced at least by the rtcache.
    418  */
    419 static void
    420 rtcache_check_rtrefcnt(int family)
    421 {
    422 	struct domain *dom = pffinddomain(family);
    423 	struct route *ro;
    424 
    425 	if (dom == NULL)
    426 		return;
    427 
    428 	LIST_FOREACH(ro, &dom->dom_rtcache, ro_rtcache_next)
    429 		KDASSERT(ro->_ro_rt == NULL || ro->_ro_rt->rt_refcnt > 0);
    430 }
    431 #endif
    432 
    433 void
    434 rtfree(struct rtentry *rt)
    435 {
    436 	struct ifaddr *ifa;
    437 
    438 	KASSERT(rt != NULL);
    439 	KASSERT(rt->rt_refcnt > 0);
    440 
    441 	rt->rt_refcnt--;
    442 #ifdef DEBUG
    443 	if (rt_getkey(rt) != NULL)
    444 		rtcache_check_rtrefcnt(rt_getkey(rt)->sa_family);
    445 #endif
    446 	if (rt->rt_refcnt == 0 && (rt->rt_flags & RTF_UP) == 0) {
    447 		rt_assert_inactive(rt);
    448 		rttrash--;
    449 		rt_timer_remove_all(rt, 0);
    450 		ifa = rt->rt_ifa;
    451 		rt->rt_ifa = NULL;
    452 		ifafree(ifa);
    453 		rt->rt_ifp = NULL;
    454 		rt_destroy(rt);
    455 		pool_put(&rtentry_pool, rt);
    456 	}
    457 }
    458 
    459 /*
    460  * Force a routing table entry to the specified
    461  * destination to go through the given gateway.
    462  * Normally called as a result of a routing redirect
    463  * message from the network layer.
    464  *
    465  * N.B.: must be called at splsoftnet
    466  */
    467 void
    468 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
    469 	const struct sockaddr *netmask, int flags, const struct sockaddr *src,
    470 	struct rtentry **rtp)
    471 {
    472 	struct rtentry *rt;
    473 	int error = 0;
    474 	uint64_t *stat = NULL;
    475 	struct rt_addrinfo info;
    476 	struct ifaddr *ifa;
    477 
    478 	/* verify the gateway is directly reachable */
    479 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
    480 		error = ENETUNREACH;
    481 		goto out;
    482 	}
    483 	rt = rtalloc1(dst, 0);
    484 	/*
    485 	 * If the redirect isn't from our current router for this dst,
    486 	 * it's either old or wrong.  If it redirects us to ourselves,
    487 	 * we have a routing loop, perhaps as a result of an interface
    488 	 * going down recently.
    489 	 */
    490 	if (!(flags & RTF_DONE) && rt &&
    491 	     (sockaddr_cmp(src, rt->rt_gateway) != 0 || rt->rt_ifa != ifa))
    492 		error = EINVAL;
    493 	else if (ifa_ifwithaddr(gateway))
    494 		error = EHOSTUNREACH;
    495 	if (error)
    496 		goto done;
    497 	/*
    498 	 * Create a new entry if we just got back a wildcard entry
    499 	 * or the lookup failed.  This is necessary for hosts
    500 	 * which use routing redirects generated by smart gateways
    501 	 * to dynamically build the routing tables.
    502 	 */
    503 	if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
    504 		goto create;
    505 	/*
    506 	 * Don't listen to the redirect if it's
    507 	 * for a route to an interface.
    508 	 */
    509 	if (rt->rt_flags & RTF_GATEWAY) {
    510 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
    511 			/*
    512 			 * Changing from route to net => route to host.
    513 			 * Create new route, rather than smashing route to net.
    514 			 */
    515 		create:
    516 			if (rt != NULL)
    517 				rtfree(rt);
    518 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
    519 			memset(&info, 0, sizeof(info));
    520 			info.rti_info[RTAX_DST] = dst;
    521 			info.rti_info[RTAX_GATEWAY] = gateway;
    522 			info.rti_info[RTAX_NETMASK] = netmask;
    523 			info.rti_ifa = ifa;
    524 			info.rti_flags = flags;
    525 			rt = NULL;
    526 			error = rtrequest1(RTM_ADD, &info, &rt);
    527 			if (rt != NULL)
    528 				flags = rt->rt_flags;
    529 			stat = &rtstat.rts_dynamic;
    530 		} else {
    531 			/*
    532 			 * Smash the current notion of the gateway to
    533 			 * this destination.  Should check about netmask!!!
    534 			 */
    535 			rt->rt_flags |= RTF_MODIFIED;
    536 			flags |= RTF_MODIFIED;
    537 			stat = &rtstat.rts_newgateway;
    538 			rt_setgate(rt, gateway);
    539 		}
    540 	} else
    541 		error = EHOSTUNREACH;
    542 done:
    543 	if (rt) {
    544 		if (rtp != NULL && !error)
    545 			*rtp = rt;
    546 		else
    547 			rtfree(rt);
    548 	}
    549 out:
    550 	if (error)
    551 		rtstat.rts_badredirect++;
    552 	else if (stat != NULL)
    553 		(*stat)++;
    554 	memset(&info, 0, sizeof(info));
    555 	info.rti_info[RTAX_DST] = dst;
    556 	info.rti_info[RTAX_GATEWAY] = gateway;
    557 	info.rti_info[RTAX_NETMASK] = netmask;
    558 	info.rti_info[RTAX_AUTHOR] = src;
    559 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
    560 }
    561 
    562 /*
    563  * Delete a route and generate a message.
    564  * It doesn't free a passed rt.
    565  */
    566 static int
    567 rtdeletemsg(struct rtentry *rt)
    568 {
    569 	int error;
    570 	struct rt_addrinfo info;
    571 	struct rtentry *retrt;
    572 
    573 	/*
    574 	 * Request the new route so that the entry is not actually
    575 	 * deleted.  That will allow the information being reported to
    576 	 * be accurate (and consistent with route_output()).
    577 	 */
    578 	memset(&info, 0, sizeof(info));
    579 	info.rti_info[RTAX_DST] = rt_getkey(rt);
    580 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    581 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    582 	info.rti_flags = rt->rt_flags;
    583 	error = rtrequest1(RTM_DELETE, &info, &retrt);
    584 
    585 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
    586 
    587 	if (error == 0)
    588 		rtfree(retrt);
    589 	return error;
    590 }
    591 
    592 static int
    593 rtflushclone1(struct rtentry *rt, void *arg)
    594 {
    595 	struct rtentry *parent;
    596 
    597 	parent = (struct rtentry *)arg;
    598 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
    599 		rtdeletemsg(rt);
    600 	return 0;
    601 }
    602 
    603 static void
    604 rtflushclone(sa_family_t family, struct rtentry *parent)
    605 {
    606 
    607 #ifdef DIAGNOSTIC
    608 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
    609 		panic("rtflushclone: called with a non-cloning route");
    610 #endif
    611 	rt_walktree(family, rtflushclone1, (void *)parent);
    612 }
    613 
    614 struct ifaddr *
    615 ifa_ifwithroute(int flags, const struct sockaddr *dst,
    616 	const struct sockaddr *gateway)
    617 {
    618 	struct ifaddr *ifa;
    619 	if ((flags & RTF_GATEWAY) == 0) {
    620 		/*
    621 		 * If we are adding a route to an interface,
    622 		 * and the interface is a pt to pt link
    623 		 * we should search for the destination
    624 		 * as our clue to the interface.  Otherwise
    625 		 * we can use the local address.
    626 		 */
    627 		ifa = NULL;
    628 		if ((flags & RTF_HOST) && gateway->sa_family != AF_LINK)
    629 			ifa = ifa_ifwithdstaddr(dst);
    630 		if (ifa == NULL)
    631 			ifa = ifa_ifwithaddr(gateway);
    632 	} else {
    633 		/*
    634 		 * If we are adding a route to a remote net
    635 		 * or host, the gateway may still be on the
    636 		 * other end of a pt to pt link.
    637 		 */
    638 		ifa = ifa_ifwithdstaddr(gateway);
    639 	}
    640 	if (ifa == NULL)
    641 		ifa = ifa_ifwithnet(gateway);
    642 	if (ifa == NULL) {
    643 		struct rtentry *rt = rtalloc1(dst, 0);
    644 		if (rt == NULL)
    645 			return NULL;
    646 		ifa = rt->rt_ifa;
    647 		rtfree(rt);
    648 		if (ifa == NULL)
    649 			return NULL;
    650 	}
    651 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
    652 		struct ifaddr *oifa = ifa;
    653 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
    654 		if (ifa == NULL)
    655 			ifa = oifa;
    656 	}
    657 	return ifa;
    658 }
    659 
    660 /*
    661  * If it suceeds and ret_nrt isn't NULL, refcnt of ret_nrt is incremented.
    662  * The caller has to rtfree it by itself.
    663  */
    664 int
    665 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
    666 	const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
    667 {
    668 	struct rt_addrinfo info;
    669 
    670 	memset(&info, 0, sizeof(info));
    671 	info.rti_flags = flags;
    672 	info.rti_info[RTAX_DST] = dst;
    673 	info.rti_info[RTAX_GATEWAY] = gateway;
    674 	info.rti_info[RTAX_NETMASK] = netmask;
    675 	return rtrequest1(req, &info, ret_nrt);
    676 }
    677 
    678 /*
    679  * It's a utility function to add/remove a route to/from the routing table
    680  * and tell user processes the addition/removal on success.
    681  */
    682 int
    683 rtrequest_newmsg(const int req, const struct sockaddr *dst,
    684 	const struct sockaddr *gateway, const struct sockaddr *netmask,
    685 	const int flags)
    686 {
    687 	int error;
    688 	struct rtentry *ret_nrt = NULL;
    689 
    690 	KASSERT(req == RTM_ADD || req == RTM_DELETE);
    691 
    692 	error = rtrequest(req, dst, gateway, netmask, flags, &ret_nrt);
    693 	if (error != 0)
    694 		return error;
    695 
    696 	KASSERT(ret_nrt != NULL);
    697 
    698 	rt_newmsg(req, ret_nrt); /* tell user process */
    699 	rtfree(ret_nrt);
    700 
    701 	return 0;
    702 }
    703 
    704 int
    705 rt_getifa(struct rt_addrinfo *info)
    706 {
    707 	struct ifaddr *ifa;
    708 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
    709 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
    710 	const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
    711 	const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
    712 	int flags = info->rti_flags;
    713 
    714 	/*
    715 	 * ifp may be specified by sockaddr_dl when protocol address
    716 	 * is ambiguous
    717 	 */
    718 	if (info->rti_ifp == NULL && ifpaddr != NULL
    719 	    && ifpaddr->sa_family == AF_LINK &&
    720 	    (ifa = ifa_ifwithnet(ifpaddr)) != NULL)
    721 		info->rti_ifp = ifa->ifa_ifp;
    722 	if (info->rti_ifa == NULL && ifaaddr != NULL)
    723 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
    724 	if (info->rti_ifa == NULL) {
    725 		const struct sockaddr *sa;
    726 
    727 		sa = ifaaddr != NULL ? ifaaddr :
    728 		    (gateway != NULL ? gateway : dst);
    729 		if (sa != NULL && info->rti_ifp != NULL)
    730 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
    731 		else if (dst != NULL && gateway != NULL)
    732 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
    733 		else if (sa != NULL)
    734 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
    735 	}
    736 	if ((ifa = info->rti_ifa) == NULL)
    737 		return ENETUNREACH;
    738 	if (ifa->ifa_getifa != NULL) {
    739 		info->rti_ifa = ifa = (*ifa->ifa_getifa)(ifa, dst);
    740 		if (ifa == NULL)
    741 			return ENETUNREACH;
    742 	}
    743 	if (info->rti_ifp == NULL)
    744 		info->rti_ifp = ifa->ifa_ifp;
    745 	return 0;
    746 }
    747 
    748 /*
    749  * If it suceeds and ret_nrt isn't NULL, refcnt of ret_nrt is incremented.
    750  * The caller has to rtfree it by itself.
    751  */
    752 int
    753 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
    754 {
    755 	int s = splsoftnet();
    756 	int error = 0, rc;
    757 	struct rtentry *rt, *crt;
    758 	rtbl_t *rtbl;
    759 	struct ifaddr *ifa, *ifa2;
    760 	struct sockaddr_storage maskeddst;
    761 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
    762 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
    763 	const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
    764 	int flags = info->rti_flags;
    765 #define senderr(x) { error = x ; goto bad; }
    766 
    767 	if ((rtbl = rt_gettable(dst->sa_family)) == NULL)
    768 		senderr(ESRCH);
    769 	if (flags & RTF_HOST)
    770 		netmask = NULL;
    771 	switch (req) {
    772 	case RTM_DELETE:
    773 		if (netmask) {
    774 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    775 			    netmask);
    776 			dst = (struct sockaddr *)&maskeddst;
    777 		}
    778 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
    779 			senderr(ESRCH);
    780 		if ((rt->rt_flags & RTF_CLONING) != 0) {
    781 			/* clean up any cloned children */
    782 			rtflushclone(dst->sa_family, rt);
    783 		}
    784 		if ((rt = rt_deladdr(rtbl, dst, netmask)) == NULL)
    785 			senderr(ESRCH);
    786 		if (rt->rt_gwroute) {
    787 			rtfree(rt->rt_gwroute);
    788 			rt->rt_gwroute = NULL;
    789 		}
    790 		if (rt->rt_parent) {
    791 			rt->rt_parent->rt_refcnt--;
    792 			rt->rt_parent = NULL;
    793 		}
    794 		rt->rt_flags &= ~RTF_UP;
    795 		if ((ifa = rt->rt_ifa)) {
    796 			if (ifa->ifa_flags & IFA_ROUTE &&
    797 			    rt_ifa_connected(rt, ifa)) {
    798 				RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    799 				    "deleted IFA_ROUTE\n",
    800 				    (void *)rt->_rt_key, (void *)ifa);
    801 				ifa->ifa_flags &= ~IFA_ROUTE;
    802 			}
    803 			if (ifa->ifa_rtrequest)
    804 				ifa->ifa_rtrequest(RTM_DELETE, rt, info);
    805 		}
    806 		rttrash++;
    807 		if (ret_nrt) {
    808 			*ret_nrt = rt;
    809 			rt->rt_refcnt++;
    810 		} else if (rt->rt_refcnt <= 0) {
    811 			/* Adjust the refcount */
    812 			rt->rt_refcnt++;
    813 			rtfree(rt);
    814 		}
    815 		break;
    816 
    817 	case RTM_RESOLVE:
    818 		if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
    819 			senderr(EINVAL);
    820 		if ((rt->rt_flags & RTF_CLONING) == 0)
    821 			senderr(EINVAL);
    822 		ifa = rt->rt_ifa;
    823 		flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC);
    824 		flags |= RTF_CLONED;
    825 		gateway = rt->rt_gateway;
    826 		flags |= RTF_HOST;
    827 		goto makeroute;
    828 
    829 	case RTM_ADD:
    830 		if (info->rti_ifa == NULL && (error = rt_getifa(info)))
    831 			senderr(error);
    832 		ifa = info->rti_ifa;
    833 	makeroute:
    834 		/* Already at splsoftnet() so pool_get/pool_put are safe */
    835 		rt = pool_get(&rtentry_pool, PR_NOWAIT);
    836 		if (rt == NULL)
    837 			senderr(ENOBUFS);
    838 		memset(rt, 0, sizeof(*rt));
    839 		rt->rt_flags = RTF_UP | flags;
    840 		LIST_INIT(&rt->rt_timer);
    841 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    842 		if (rt_setkey(rt, dst, M_NOWAIT) == NULL ||
    843 		    rt_setgate(rt, gateway) != 0) {
    844 			pool_put(&rtentry_pool, rt);
    845 			senderr(ENOBUFS);
    846 		}
    847 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    848 		if (netmask) {
    849 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    850 			    netmask);
    851 			rt_setkey(rt, (struct sockaddr *)&maskeddst, M_NOWAIT);
    852 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    853 		} else {
    854 			rt_setkey(rt, dst, M_NOWAIT);
    855 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    856 		}
    857 		rt_set_ifa(rt, ifa);
    858 		if (info->rti_info[RTAX_TAG] != NULL)
    859 			rt_settag(rt, info->rti_info[RTAX_TAG]);
    860 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    861 		if (info->rti_info[RTAX_IFP] != NULL &&
    862 		    (ifa2 = ifa_ifwithnet(info->rti_info[RTAX_IFP])) != NULL &&
    863 		    ifa2->ifa_ifp != NULL)
    864 			rt->rt_ifp = ifa2->ifa_ifp;
    865 		else
    866 			rt->rt_ifp = ifa->ifa_ifp;
    867 		if (req == RTM_RESOLVE) {
    868 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
    869 			rt->rt_parent = *ret_nrt;
    870 			rt->rt_parent->rt_refcnt++;
    871 		}
    872 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    873 		rc = rt_addaddr(rtbl, rt, netmask);
    874 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    875 		if (rc != 0 && (crt = rtalloc1(rt_getkey(rt), 0)) != NULL) {
    876 			/* overwrite cloned route */
    877 			if ((crt->rt_flags & RTF_CLONED) != 0) {
    878 				rtdeletemsg(crt);
    879 				rc = rt_addaddr(rtbl, rt, netmask);
    880 			}
    881 			rtfree(crt);
    882 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    883 		}
    884 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    885 		if (rc != 0) {
    886 			ifafree(ifa);
    887 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
    888 				rtfree(rt->rt_parent);
    889 			if (rt->rt_gwroute)
    890 				rtfree(rt->rt_gwroute);
    891 			rt_destroy(rt);
    892 			pool_put(&rtentry_pool, rt);
    893 			senderr(rc);
    894 		}
    895 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    896 		if (ifa->ifa_rtrequest)
    897 			ifa->ifa_rtrequest(req, rt, info);
    898 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    899 		if (ret_nrt) {
    900 			*ret_nrt = rt;
    901 			rt->rt_refcnt++;
    902 		}
    903 		if ((rt->rt_flags & RTF_CLONING) != 0) {
    904 			/* clean up any cloned children */
    905 			rtflushclone(dst->sa_family, rt);
    906 		}
    907 		rtflushall(dst->sa_family);
    908 		break;
    909 	case RTM_GET:
    910 		if (netmask != NULL) {
    911 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    912 			    netmask);
    913 			dst = (struct sockaddr *)&maskeddst;
    914 		}
    915 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
    916 			senderr(ESRCH);
    917 		if (ret_nrt != NULL) {
    918 			*ret_nrt = rt;
    919 			rt->rt_refcnt++;
    920 		}
    921 		break;
    922 	}
    923 bad:
    924 	splx(s);
    925 	return error;
    926 }
    927 
    928 int
    929 rt_setgate(struct rtentry *rt, const struct sockaddr *gate)
    930 {
    931 	KASSERT(rt != rt->rt_gwroute);
    932 
    933 	KASSERT(rt->_rt_key != NULL);
    934 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    935 
    936 	if (rt->rt_gwroute) {
    937 		rtfree(rt->rt_gwroute);
    938 		rt->rt_gwroute = NULL;
    939 	}
    940 	KASSERT(rt->_rt_key != NULL);
    941 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    942 	if (rt->rt_gateway != NULL)
    943 		sockaddr_free(rt->rt_gateway);
    944 	KASSERT(rt->_rt_key != NULL);
    945 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    946 	if ((rt->rt_gateway = sockaddr_dup(gate, M_ZERO | M_NOWAIT)) == NULL)
    947 		return ENOMEM;
    948 	KASSERT(rt->_rt_key != NULL);
    949 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    950 
    951 	if (rt->rt_flags & RTF_GATEWAY) {
    952 		KASSERT(rt->_rt_key != NULL);
    953 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    954 		rt->rt_gwroute = rtalloc1(gate, 1);
    955 		/*
    956 		 * If we switched gateways, grab the MTU from the new
    957 		 * gateway route if the current MTU, if the current MTU is
    958 		 * greater than the MTU of gateway.
    959 		 * Note that, if the MTU of gateway is 0, we will reset the
    960 		 * MTU of the route to run PMTUD again from scratch. XXX
    961 		 */
    962 		KASSERT(rt->_rt_key != NULL);
    963 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    964 		if (rt->rt_gwroute
    965 		    && !(rt->rt_rmx.rmx_locks & RTV_MTU)
    966 		    && rt->rt_rmx.rmx_mtu
    967 		    && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
    968 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
    969 		}
    970 	}
    971 	KASSERT(rt->_rt_key != NULL);
    972 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    973 	return 0;
    974 }
    975 
    976 static void
    977 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
    978 	const struct sockaddr *netmask)
    979 {
    980 	const char *netmaskp = &netmask->sa_data[0],
    981 	           *srcp = &src->sa_data[0];
    982 	char *dstp = &dst->sa_data[0];
    983 	const char *maskend = (char *)dst + MIN(netmask->sa_len, src->sa_len);
    984 	const char *srcend = (char *)dst + src->sa_len;
    985 
    986 	dst->sa_len = src->sa_len;
    987 	dst->sa_family = src->sa_family;
    988 
    989 	while (dstp < maskend)
    990 		*dstp++ = *srcp++ & *netmaskp++;
    991 	if (dstp < srcend)
    992 		memset(dstp, 0, (size_t)(srcend - dstp));
    993 }
    994 
    995 /*
    996  * Inform the routing socket of a route change.
    997  */
    998 void
    999 rt_newmsg(const int cmd, const struct rtentry *rt)
   1000 {
   1001 	struct rt_addrinfo info;
   1002 
   1003 	memset((void *)&info, 0, sizeof(info));
   1004 	info.rti_info[RTAX_DST] = rt_getkey(rt);
   1005 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
   1006 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
   1007 	if (rt->rt_ifp) {
   1008 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
   1009 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
   1010 	}
   1011 
   1012 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
   1013 }
   1014 
   1015 /*
   1016  * Set up or tear down a routing table entry, normally
   1017  * for an interface.
   1018  */
   1019 int
   1020 rtinit(struct ifaddr *ifa, int cmd, int flags)
   1021 {
   1022 	struct rtentry *rt;
   1023 	struct sockaddr *dst, *odst;
   1024 	struct sockaddr_storage maskeddst;
   1025 	struct rtentry *nrt = NULL;
   1026 	int error;
   1027 	struct rt_addrinfo info;
   1028 	struct sockaddr_dl *sdl;
   1029 	const struct sockaddr_dl *ifsdl;
   1030 
   1031 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
   1032 	if (cmd == RTM_DELETE) {
   1033 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
   1034 			/* Delete subnet route for this interface */
   1035 			odst = dst;
   1036 			dst = (struct sockaddr *)&maskeddst;
   1037 			rt_maskedcopy(odst, dst, ifa->ifa_netmask);
   1038 		}
   1039 		if ((rt = rtalloc1(dst, 0)) != NULL) {
   1040 			if (rt->rt_ifa != ifa) {
   1041 				rtfree(rt);
   1042 				return (flags & RTF_HOST) ? EHOSTUNREACH
   1043 							: ENETUNREACH;
   1044 			}
   1045 			rtfree(rt);
   1046 		}
   1047 	}
   1048 	memset(&info, 0, sizeof(info));
   1049 	info.rti_ifa = ifa;
   1050 	info.rti_flags = flags | ifa->ifa_flags;
   1051 	info.rti_info[RTAX_DST] = dst;
   1052 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
   1053 	/*
   1054 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
   1055 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
   1056 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
   1057 	 * change it to meet bsdi4 behavior.
   1058 	 */
   1059 	if (cmd != RTM_LLINFO_UPD)
   1060 		info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
   1061 	error = rtrequest1((cmd == RTM_LLINFO_UPD) ? RTM_GET : cmd, &info,
   1062 	    &nrt);
   1063 	if (error != 0)
   1064 		return error;
   1065 
   1066 	rt = nrt;
   1067 	switch (cmd) {
   1068 	case RTM_DELETE:
   1069 		rt_newmsg(cmd, rt);
   1070 		break;
   1071 	case RTM_LLINFO_UPD:
   1072 		RT_DPRINTF("%s: updating%s\n", __func__,
   1073 		    ((rt->rt_flags & RTF_LLINFO) == 0) ? " (no llinfo)" : "");
   1074 
   1075 		ifsdl = ifa->ifa_ifp->if_sadl;
   1076 
   1077 		if ((rt->rt_flags & RTF_LLINFO) != 0 &&
   1078 		    (sdl = satosdl(rt->rt_gateway)) != NULL &&
   1079 		    sdl->sdl_family == AF_LINK &&
   1080 		    sockaddr_dl_setaddr(sdl, sdl->sdl_len, CLLADDR(ifsdl),
   1081 		                        ifa->ifa_ifp->if_addrlen) == NULL) {
   1082 			error = EINVAL;
   1083 			break;
   1084 		}
   1085 
   1086 		if (cmd == RTM_LLINFO_UPD && ifa->ifa_rtrequest != NULL)
   1087 			ifa->ifa_rtrequest(RTM_LLINFO_UPD, rt, &info);
   1088 		rt_newmsg(RTM_CHANGE, rt);
   1089 		break;
   1090 	case RTM_ADD:
   1091 		if (rt->rt_ifa != ifa) {
   1092 			printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
   1093 				rt->rt_ifa);
   1094 			if (rt->rt_ifa->ifa_rtrequest != NULL) {
   1095 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
   1096 				    &info);
   1097 			}
   1098 			rt_replace_ifa(rt, ifa);
   1099 			rt->rt_ifp = ifa->ifa_ifp;
   1100 			if (ifa->ifa_rtrequest != NULL)
   1101 				ifa->ifa_rtrequest(RTM_ADD, rt, &info);
   1102 		}
   1103 		rt_newmsg(cmd, rt);
   1104 		break;
   1105 	}
   1106 	rtfree(rt);
   1107 	return error;
   1108 }
   1109 
   1110 /*
   1111  * Create a local route entry for the address.
   1112  * Announce the addition of the address and the route to the routing socket.
   1113  */
   1114 int
   1115 rt_ifa_addlocal(struct ifaddr *ifa)
   1116 {
   1117 	struct rtentry *rt;
   1118 	int e;
   1119 
   1120 	/* If there is no loopback entry, allocate one. */
   1121 	rt = rtalloc1(ifa->ifa_addr, 0);
   1122 	if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
   1123 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
   1124 	{
   1125 		struct rt_addrinfo info;
   1126 		struct rtentry *nrt;
   1127 
   1128 		memset(&info, 0, sizeof(info));
   1129 		info.rti_flags = RTF_HOST | RTF_LOCAL;
   1130 		if (!(ifa->ifa_ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT)))
   1131 			info.rti_flags |= RTF_LLINFO;
   1132 		info.rti_info[RTAX_DST] = ifa->ifa_addr;
   1133 		info.rti_info[RTAX_GATEWAY] =
   1134 		    (const struct sockaddr *)ifa->ifa_ifp->if_sadl;
   1135 		info.rti_ifa = ifa;
   1136 		nrt = NULL;
   1137 		e = rtrequest1(RTM_ADD, &info, &nrt);
   1138 		if (nrt && ifa != nrt->rt_ifa)
   1139 			rt_replace_ifa(nrt, ifa);
   1140 		rt_newaddrmsg(RTM_ADD, ifa, e, nrt);
   1141 		if (nrt != NULL)
   1142 			rtfree(nrt);
   1143 	} else {
   1144 		e = 0;
   1145 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1146 	}
   1147 	if (rt != NULL)
   1148 		rtfree(rt);
   1149 	return e;
   1150 }
   1151 
   1152 /*
   1153  * Remove the local route entry for the address.
   1154  * Announce the removal of the address and the route to the routing socket.
   1155  */
   1156 int
   1157 rt_ifa_remlocal(struct ifaddr *ifa, struct ifaddr *alt_ifa)
   1158 {
   1159 	struct rtentry *rt;
   1160 	int e = 0;
   1161 
   1162 	rt = rtalloc1(ifa->ifa_addr, 0);
   1163 
   1164 	/*
   1165 	 * Before deleting, check if a corresponding loopbacked
   1166 	 * host route surely exists.  With this check, we can avoid
   1167 	 * deleting an interface direct route whose destination is
   1168 	 * the same as the address being removed.  This can happen
   1169 	 * when removing a subnet-router anycast address on an
   1170 	 * interface attached to a shared medium.
   1171 	 */
   1172 	if (rt != NULL &&
   1173 	    (rt->rt_flags & RTF_HOST) &&
   1174 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK))
   1175 	{
   1176 		/* If we cannot replace the route's ifaddr with the equivalent
   1177 		 * ifaddr of another interface, I believe it is safest to
   1178 		 * delete the route.
   1179 		 */
   1180 		if (alt_ifa == NULL) {
   1181 			e = rtdeletemsg(rt);
   1182 			rt_newaddrmsg(RTM_DELADDR, ifa, 0, NULL);
   1183 		} else {
   1184 			rt_replace_ifa(rt, alt_ifa);
   1185 			rt_newmsg(RTM_CHANGE, rt);
   1186 		}
   1187 	} else
   1188 		rt_newaddrmsg(RTM_DELADDR, ifa, 0, NULL);
   1189 	if (rt != NULL)
   1190 		rtfree(rt);
   1191 	return e;
   1192 }
   1193 
   1194 /*
   1195  * Route timer routines.  These routes allow functions to be called
   1196  * for various routes at any time.  This is useful in supporting
   1197  * path MTU discovery and redirect route deletion.
   1198  *
   1199  * This is similar to some BSDI internal functions, but it provides
   1200  * for multiple queues for efficiency's sake...
   1201  */
   1202 
   1203 LIST_HEAD(, rttimer_queue) rttimer_queue_head;
   1204 static int rt_init_done = 0;
   1205 
   1206 #define RTTIMER_CALLOUT(r)	do {					\
   1207 		if (r->rtt_func != NULL) {				\
   1208 			(*r->rtt_func)(r->rtt_rt, r);			\
   1209 		} else {						\
   1210 			rtrequest((int) RTM_DELETE,			\
   1211 				  rt_getkey(r->rtt_rt),			\
   1212 				  0, 0, 0, 0);				\
   1213 		}							\
   1214 	} while (/*CONSTCOND*/0)
   1215 
   1216 /*
   1217  * Some subtle order problems with domain initialization mean that
   1218  * we cannot count on this being run from rt_init before various
   1219  * protocol initializations are done.  Therefore, we make sure
   1220  * that this is run when the first queue is added...
   1221  */
   1222 
   1223 void
   1224 rt_timer_init(void)
   1225 {
   1226 	assert(rt_init_done == 0);
   1227 
   1228 	LIST_INIT(&rttimer_queue_head);
   1229 	callout_init(&rt_timer_ch, 0);
   1230 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
   1231 	rt_init_done = 1;
   1232 }
   1233 
   1234 struct rttimer_queue *
   1235 rt_timer_queue_create(u_int timeout)
   1236 {
   1237 	struct rttimer_queue *rtq;
   1238 
   1239 	if (rt_init_done == 0)
   1240 		rt_timer_init();
   1241 
   1242 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
   1243 	if (rtq == NULL)
   1244 		return NULL;
   1245 	memset(rtq, 0, sizeof(*rtq));
   1246 
   1247 	rtq->rtq_timeout = timeout;
   1248 	TAILQ_INIT(&rtq->rtq_head);
   1249 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
   1250 
   1251 	return rtq;
   1252 }
   1253 
   1254 void
   1255 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
   1256 {
   1257 
   1258 	rtq->rtq_timeout = timeout;
   1259 }
   1260 
   1261 void
   1262 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy)
   1263 {
   1264 	struct rttimer *r;
   1265 
   1266 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
   1267 		LIST_REMOVE(r, rtt_link);
   1268 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
   1269 		if (destroy)
   1270 			RTTIMER_CALLOUT(r);
   1271 		rtfree(r->rtt_rt);
   1272 		/* we are already at splsoftnet */
   1273 		pool_put(&rttimer_pool, r);
   1274 		if (rtq->rtq_count > 0)
   1275 			rtq->rtq_count--;
   1276 		else
   1277 			printf("rt_timer_queue_remove_all: "
   1278 			    "rtq_count reached 0\n");
   1279 	}
   1280 }
   1281 
   1282 void
   1283 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
   1284 {
   1285 
   1286 	rt_timer_queue_remove_all(rtq, destroy);
   1287 
   1288 	LIST_REMOVE(rtq, rtq_link);
   1289 
   1290 	/*
   1291 	 * Caller is responsible for freeing the rttimer_queue structure.
   1292 	 */
   1293 }
   1294 
   1295 unsigned long
   1296 rt_timer_count(struct rttimer_queue *rtq)
   1297 {
   1298 	return rtq->rtq_count;
   1299 }
   1300 
   1301 void
   1302 rt_timer_remove_all(struct rtentry *rt, int destroy)
   1303 {
   1304 	struct rttimer *r;
   1305 
   1306 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
   1307 		LIST_REMOVE(r, rtt_link);
   1308 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
   1309 		if (destroy)
   1310 			RTTIMER_CALLOUT(r);
   1311 		if (r->rtt_queue->rtq_count > 0)
   1312 			r->rtt_queue->rtq_count--;
   1313 		else
   1314 			printf("rt_timer_remove_all: rtq_count reached 0\n");
   1315 		rtfree(r->rtt_rt);
   1316 		/* we are already at splsoftnet */
   1317 		pool_put(&rttimer_pool, r);
   1318 	}
   1319 }
   1320 
   1321 int
   1322 rt_timer_add(struct rtentry *rt,
   1323 	void (*func)(struct rtentry *, struct rttimer *),
   1324 	struct rttimer_queue *queue)
   1325 {
   1326 	struct rttimer *r;
   1327 	int s;
   1328 
   1329 	/*
   1330 	 * If there's already a timer with this action, destroy it before
   1331 	 * we add a new one.
   1332 	 */
   1333 	LIST_FOREACH(r, &rt->rt_timer, rtt_link) {
   1334 		if (r->rtt_func == func)
   1335 			break;
   1336 	}
   1337 	if (r != NULL) {
   1338 		LIST_REMOVE(r, rtt_link);
   1339 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
   1340 		if (r->rtt_queue->rtq_count > 0)
   1341 			r->rtt_queue->rtq_count--;
   1342 		else
   1343 			printf("rt_timer_add: rtq_count reached 0\n");
   1344 		rtfree(r->rtt_rt);
   1345 	} else {
   1346 		s = splsoftnet();
   1347 		r = pool_get(&rttimer_pool, PR_NOWAIT);
   1348 		splx(s);
   1349 		if (r == NULL)
   1350 			return ENOBUFS;
   1351 	}
   1352 
   1353 	memset(r, 0, sizeof(*r));
   1354 
   1355 	rt->rt_refcnt++;
   1356 	r->rtt_rt = rt;
   1357 	r->rtt_time = time_uptime;
   1358 	r->rtt_func = func;
   1359 	r->rtt_queue = queue;
   1360 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
   1361 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
   1362 	r->rtt_queue->rtq_count++;
   1363 
   1364 	return 0;
   1365 }
   1366 
   1367 /* ARGSUSED */
   1368 void
   1369 rt_timer_timer(void *arg)
   1370 {
   1371 	struct rttimer_queue *rtq;
   1372 	struct rttimer *r;
   1373 	int s;
   1374 
   1375 	s = splsoftnet();
   1376 	LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) {
   1377 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
   1378 		    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
   1379 			LIST_REMOVE(r, rtt_link);
   1380 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
   1381 			RTTIMER_CALLOUT(r);
   1382 			rtfree(r->rtt_rt);
   1383 			pool_put(&rttimer_pool, r);
   1384 			if (rtq->rtq_count > 0)
   1385 				rtq->rtq_count--;
   1386 			else
   1387 				printf("rt_timer_timer: rtq_count reached 0\n");
   1388 		}
   1389 	}
   1390 	splx(s);
   1391 
   1392 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
   1393 }
   1394 
   1395 static struct rtentry *
   1396 _rtcache_init(struct route *ro, int flag)
   1397 {
   1398 	rtcache_invariants(ro);
   1399 	KASSERT(ro->_ro_rt == NULL);
   1400 
   1401 	if (rtcache_getdst(ro) == NULL)
   1402 		return NULL;
   1403 	ro->ro_invalid = false;
   1404 	if ((ro->_ro_rt = rtalloc1(rtcache_getdst(ro), flag)) != NULL)
   1405 		rtcache(ro);
   1406 
   1407 	rtcache_invariants(ro);
   1408 	return ro->_ro_rt;
   1409 }
   1410 
   1411 struct rtentry *
   1412 rtcache_init(struct route *ro)
   1413 {
   1414 	return _rtcache_init(ro, 1);
   1415 }
   1416 
   1417 struct rtentry *
   1418 rtcache_init_noclone(struct route *ro)
   1419 {
   1420 	return _rtcache_init(ro, 0);
   1421 }
   1422 
   1423 struct rtentry *
   1424 rtcache_update(struct route *ro, int clone)
   1425 {
   1426 	rtcache_clear(ro);
   1427 	return _rtcache_init(ro, clone);
   1428 }
   1429 
   1430 void
   1431 rtcache_copy(struct route *new_ro, const struct route *old_ro)
   1432 {
   1433 	struct rtentry *rt;
   1434 
   1435 	KASSERT(new_ro != old_ro);
   1436 	rtcache_invariants(new_ro);
   1437 	rtcache_invariants(old_ro);
   1438 
   1439 	if ((rt = rtcache_validate(old_ro)) != NULL)
   1440 		rt->rt_refcnt++;
   1441 
   1442 	if (rtcache_getdst(old_ro) == NULL ||
   1443 	    rtcache_setdst(new_ro, rtcache_getdst(old_ro)) != 0)
   1444 		return;
   1445 
   1446 	new_ro->ro_invalid = false;
   1447 	if ((new_ro->_ro_rt = rt) != NULL)
   1448 		rtcache(new_ro);
   1449 	rtcache_invariants(new_ro);
   1450 }
   1451 
   1452 static struct dom_rtlist invalid_routes = LIST_HEAD_INITIALIZER(dom_rtlist);
   1453 
   1454 static void
   1455 rtcache_invalidate(struct dom_rtlist *rtlist)
   1456 {
   1457 	struct route *ro;
   1458 
   1459 	while ((ro = LIST_FIRST(rtlist)) != NULL) {
   1460 		rtcache_invariants(ro);
   1461 		KASSERT(ro->_ro_rt != NULL);
   1462 		ro->ro_invalid = true;
   1463 		LIST_REMOVE(ro, ro_rtcache_next);
   1464 		LIST_INSERT_HEAD(&invalid_routes, ro, ro_rtcache_next);
   1465 		rtcache_invariants(ro);
   1466 	}
   1467 }
   1468 
   1469 static void
   1470 rtcache_clear(struct route *ro)
   1471 {
   1472 	rtcache_invariants(ro);
   1473 	if (ro->_ro_rt == NULL)
   1474 		return;
   1475 
   1476 	LIST_REMOVE(ro, ro_rtcache_next);
   1477 
   1478 	rtfree(ro->_ro_rt);
   1479 	ro->_ro_rt = NULL;
   1480 	ro->ro_invalid = false;
   1481 	rtcache_invariants(ro);
   1482 }
   1483 
   1484 struct rtentry *
   1485 rtcache_lookup2(struct route *ro, const struct sockaddr *dst, int clone,
   1486     int *hitp)
   1487 {
   1488 	const struct sockaddr *odst;
   1489 	struct rtentry *rt = NULL;
   1490 
   1491 	odst = rtcache_getdst(ro);
   1492 	if (odst == NULL)
   1493 		goto miss;
   1494 
   1495 	if (sockaddr_cmp(odst, dst) != 0) {
   1496 		rtcache_free(ro);
   1497 		goto miss;
   1498 	}
   1499 
   1500 	rt = rtcache_validate(ro);
   1501 	if (rt == NULL) {
   1502 		rtcache_clear(ro);
   1503 		goto miss;
   1504 	}
   1505 
   1506 	*hitp = 1;
   1507 	rtcache_invariants(ro);
   1508 
   1509 	return rt;
   1510 miss:
   1511 	*hitp = 0;
   1512 	if (rtcache_setdst(ro, dst) == 0)
   1513 		rt = _rtcache_init(ro, clone);
   1514 
   1515 	rtcache_invariants(ro);
   1516 
   1517 	return rt;
   1518 }
   1519 
   1520 void
   1521 rtcache_free(struct route *ro)
   1522 {
   1523 	rtcache_clear(ro);
   1524 	if (ro->ro_sa != NULL) {
   1525 		sockaddr_free(ro->ro_sa);
   1526 		ro->ro_sa = NULL;
   1527 	}
   1528 	rtcache_invariants(ro);
   1529 }
   1530 
   1531 int
   1532 rtcache_setdst(struct route *ro, const struct sockaddr *sa)
   1533 {
   1534 	KASSERT(sa != NULL);
   1535 
   1536 	rtcache_invariants(ro);
   1537 	if (ro->ro_sa != NULL) {
   1538 		if (ro->ro_sa->sa_family == sa->sa_family) {
   1539 			rtcache_clear(ro);
   1540 			sockaddr_copy(ro->ro_sa, ro->ro_sa->sa_len, sa);
   1541 			rtcache_invariants(ro);
   1542 			return 0;
   1543 		}
   1544 		/* free ro_sa, wrong family */
   1545 		rtcache_free(ro);
   1546 	}
   1547 
   1548 	KASSERT(ro->_ro_rt == NULL);
   1549 
   1550 	if ((ro->ro_sa = sockaddr_dup(sa, M_ZERO | M_NOWAIT)) == NULL) {
   1551 		rtcache_invariants(ro);
   1552 		return ENOMEM;
   1553 	}
   1554 	rtcache_invariants(ro);
   1555 	return 0;
   1556 }
   1557 
   1558 const struct sockaddr *
   1559 rt_settag(struct rtentry *rt, const struct sockaddr *tag)
   1560 {
   1561 	if (rt->rt_tag != tag) {
   1562 		if (rt->rt_tag != NULL)
   1563 			sockaddr_free(rt->rt_tag);
   1564 		rt->rt_tag = sockaddr_dup(tag, M_ZERO | M_NOWAIT);
   1565 	}
   1566 	return rt->rt_tag;
   1567 }
   1568 
   1569 struct sockaddr *
   1570 rt_gettag(struct rtentry *rt)
   1571 {
   1572 	return rt->rt_tag;
   1573 }
   1574