Home | History | Annotate | Line # | Download | only in net
route.c revision 1.145
      1 /*	$NetBSD: route.c,v 1.145 2015/06/08 08:21:49 roy 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 #include "opt_inet.h"
     94 #include "opt_route.h"
     95 
     96 #include <sys/cdefs.h>
     97 __KERNEL_RCSID(0, "$NetBSD: route.c,v 1.145 2015/06/08 08:21:49 roy Exp $");
     98 
     99 #include <sys/param.h>
    100 #ifdef RTFLUSH_DEBUG
    101 #include <sys/sysctl.h>
    102 #endif
    103 #include <sys/systm.h>
    104 #include <sys/callout.h>
    105 #include <sys/proc.h>
    106 #include <sys/mbuf.h>
    107 #include <sys/socket.h>
    108 #include <sys/socketvar.h>
    109 #include <sys/domain.h>
    110 #include <sys/protosw.h>
    111 #include <sys/kernel.h>
    112 #include <sys/ioctl.h>
    113 #include <sys/pool.h>
    114 #include <sys/kauth.h>
    115 
    116 #include <net/if.h>
    117 #include <net/if_dl.h>
    118 #include <net/route.h>
    119 
    120 #include <netinet/in.h>
    121 #include <netinet/in_var.h>
    122 
    123 #ifdef RTFLUSH_DEBUG
    124 #define	rtcache_debug() __predict_false(_rtcache_debug)
    125 #else /* RTFLUSH_DEBUG */
    126 #define	rtcache_debug() 0
    127 #endif /* RTFLUSH_DEBUG */
    128 
    129 struct	rtstat	rtstat;
    130 
    131 int	rttrash;		/* routes not in table but not freed */
    132 
    133 struct pool rtentry_pool;
    134 struct pool rttimer_pool;
    135 
    136 struct callout rt_timer_ch; /* callout for rt_timer_timer() */
    137 
    138 #ifdef RTFLUSH_DEBUG
    139 static int _rtcache_debug = 0;
    140 #endif /* RTFLUSH_DEBUG */
    141 
    142 static kauth_listener_t route_listener;
    143 
    144 static int rtdeletemsg(struct rtentry *);
    145 static int rtflushclone1(struct rtentry *, void *);
    146 static void rtflushclone(sa_family_t family, struct rtentry *);
    147 static void rtflushall(int);
    148 
    149 static void rt_maskedcopy(const struct sockaddr *,
    150     struct sockaddr *, const struct sockaddr *);
    151 
    152 static void rtcache_clear(struct route *);
    153 static void rtcache_invalidate(struct dom_rtlist *);
    154 
    155 #ifdef RTFLUSH_DEBUG
    156 static void sysctl_net_rtcache_setup(struct sysctllog **);
    157 static void
    158 sysctl_net_rtcache_setup(struct sysctllog **clog)
    159 {
    160 	const struct sysctlnode *rnode;
    161 
    162 	if (sysctl_createv(clog, 0, NULL, &rnode, CTLFLAG_PERMANENT,
    163 	    CTLTYPE_NODE,
    164 	    "rtcache", SYSCTL_DESCR("Route cache related settings"),
    165 	    NULL, 0, NULL, 0, CTL_NET, CTL_CREATE, CTL_EOL) != 0)
    166 		return;
    167 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    168 	    CTLFLAG_PERMANENT|CTLFLAG_READWRITE, CTLTYPE_INT,
    169 	    "debug", SYSCTL_DESCR("Debug route caches"),
    170 	    NULL, 0, &_rtcache_debug, 0, CTL_CREATE, CTL_EOL) != 0)
    171 		return;
    172 }
    173 #endif /* RTFLUSH_DEBUG */
    174 
    175 static inline void
    176 rt_destroy(struct rtentry *rt)
    177 {
    178 	if (rt->_rt_key != NULL)
    179 		sockaddr_free(rt->_rt_key);
    180 	if (rt->rt_gateway != NULL)
    181 		sockaddr_free(rt->rt_gateway);
    182 	if (rt_gettag(rt) != NULL)
    183 		sockaddr_free(rt_gettag(rt));
    184 	rt->_rt_key = rt->rt_gateway = rt->rt_tag = NULL;
    185 }
    186 
    187 static inline const struct sockaddr *
    188 rt_setkey(struct rtentry *rt, const struct sockaddr *key, int flags)
    189 {
    190 	if (rt->_rt_key == key)
    191 		goto out;
    192 
    193 	if (rt->_rt_key != NULL)
    194 		sockaddr_free(rt->_rt_key);
    195 	rt->_rt_key = sockaddr_dup(key, flags);
    196 out:
    197 	rt->rt_nodes->rn_key = (const char *)rt->_rt_key;
    198 	return rt->_rt_key;
    199 }
    200 
    201 struct ifaddr *
    202 rt_get_ifa(struct rtentry *rt)
    203 {
    204 	struct ifaddr *ifa;
    205 
    206 	if ((ifa = rt->rt_ifa) == NULL)
    207 		return ifa;
    208 	else if (ifa->ifa_getifa == NULL)
    209 		return ifa;
    210 #if 0
    211 	else if (ifa->ifa_seqno != NULL && *ifa->ifa_seqno == rt->rt_ifa_seqno)
    212 		return ifa;
    213 #endif
    214 	else {
    215 		ifa = (*ifa->ifa_getifa)(ifa, rt_getkey(rt));
    216 		if (ifa == NULL)
    217 			return NULL;
    218 		rt_replace_ifa(rt, ifa);
    219 		return ifa;
    220 	}
    221 }
    222 
    223 static void
    224 rt_set_ifa1(struct rtentry *rt, struct ifaddr *ifa)
    225 {
    226 	rt->rt_ifa = ifa;
    227 	if (ifa->ifa_seqno != NULL)
    228 		rt->rt_ifa_seqno = *ifa->ifa_seqno;
    229 }
    230 
    231 /*
    232  * Is this route the connected route for the ifa?
    233  */
    234 static int
    235 rt_ifa_connected(const struct rtentry *rt, const struct ifaddr *ifa)
    236 {
    237 	const struct sockaddr *key, *dst, *odst;
    238 	struct sockaddr_storage maskeddst;
    239 
    240 	key = rt_getkey(rt);
    241 	dst = rt->rt_flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
    242 	if (dst == NULL ||
    243 	    dst->sa_family != key->sa_family ||
    244 	    dst->sa_len != key->sa_len)
    245 		return 0;
    246 	if ((rt->rt_flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
    247 		odst = dst;
    248 		dst = (struct sockaddr *)&maskeddst;
    249 		rt_maskedcopy(odst, (struct sockaddr *)&maskeddst,
    250 		    ifa->ifa_netmask);
    251 	}
    252 	return (memcmp(dst, key, dst->sa_len) == 0);
    253 }
    254 
    255 void
    256 rt_replace_ifa(struct rtentry *rt, struct ifaddr *ifa)
    257 {
    258 	if (rt->rt_ifa &&
    259 	    rt->rt_ifa != ifa &&
    260 	    rt->rt_ifa->ifa_flags & IFA_ROUTE &&
    261 	    rt_ifa_connected(rt, rt->rt_ifa))
    262 	{
    263 		RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    264 		    "replace deleted IFA_ROUTE\n",
    265 		    (void *)rt->_rt_key, (void *)rt->rt_ifa);
    266 		rt->rt_ifa->ifa_flags &= ~IFA_ROUTE;
    267 		if (rt_ifa_connected(rt, ifa)) {
    268 			RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    269 			    "replace added IFA_ROUTE\n",
    270 			    (void *)rt->_rt_key, (void *)ifa);
    271 			ifa->ifa_flags |= IFA_ROUTE;
    272 		}
    273 	}
    274 
    275 	ifaref(ifa);
    276 	ifafree(rt->rt_ifa);
    277 	rt_set_ifa1(rt, ifa);
    278 }
    279 
    280 static void
    281 rt_set_ifa(struct rtentry *rt, struct ifaddr *ifa)
    282 {
    283 	ifaref(ifa);
    284 	rt_set_ifa1(rt, ifa);
    285 }
    286 
    287 static int
    288 route_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
    289     void *arg0, void *arg1, void *arg2, void *arg3)
    290 {
    291 	struct rt_msghdr *rtm;
    292 	int result;
    293 
    294 	result = KAUTH_RESULT_DEFER;
    295 	rtm = arg1;
    296 
    297 	if (action != KAUTH_NETWORK_ROUTE)
    298 		return result;
    299 
    300 	if (rtm->rtm_type == RTM_GET)
    301 		result = KAUTH_RESULT_ALLOW;
    302 
    303 	return result;
    304 }
    305 
    306 void
    307 rt_init(void)
    308 {
    309 
    310 #ifdef RTFLUSH_DEBUG
    311 	sysctl_net_rtcache_setup(NULL);
    312 #endif
    313 
    314 	pool_init(&rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl",
    315 	    NULL, IPL_SOFTNET);
    316 	pool_init(&rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl",
    317 	    NULL, IPL_SOFTNET);
    318 
    319 	rn_init();	/* initialize all zeroes, all ones, mask table */
    320 	rtbl_init();
    321 
    322 	route_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
    323 	    route_listener_cb, NULL);
    324 }
    325 
    326 static void
    327 rtflushall(int family)
    328 {
    329 	struct domain *dom;
    330 
    331 	if (rtcache_debug())
    332 		printf("%s: enter\n", __func__);
    333 
    334 	if ((dom = pffinddomain(family)) == NULL)
    335 		return;
    336 
    337 	rtcache_invalidate(&dom->dom_rtcache);
    338 }
    339 
    340 static void
    341 rtcache(struct route *ro)
    342 {
    343 	struct domain *dom;
    344 
    345 	rtcache_invariants(ro);
    346 	KASSERT(ro->_ro_rt != NULL);
    347 	KASSERT(ro->ro_invalid == false);
    348 	KASSERT(rtcache_getdst(ro) != NULL);
    349 
    350 	if ((dom = pffinddomain(rtcache_getdst(ro)->sa_family)) == NULL)
    351 		return;
    352 
    353 	LIST_INSERT_HEAD(&dom->dom_rtcache, ro, ro_rtcache_next);
    354 	rtcache_invariants(ro);
    355 }
    356 
    357 /*
    358  * Packet routing routines.
    359  */
    360 struct rtentry *
    361 rtalloc1(const struct sockaddr *dst, int report)
    362 {
    363 	rtbl_t *rtbl = rt_gettable(dst->sa_family);
    364 	struct rtentry *rt;
    365 	struct rtentry *newrt = NULL;
    366 	struct rt_addrinfo info;
    367 	int  s = splsoftnet(), err = 0, msgtype = RTM_MISS;
    368 
    369 	if (rtbl != NULL && (rt = rt_matchaddr(rtbl, dst)) != NULL) {
    370 		newrt = rt;
    371 		if (report && (rt->rt_flags & RTF_CLONING)) {
    372 			err = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
    373 			    &newrt);
    374 			if (err) {
    375 				newrt = rt;
    376 				rt->rt_refcnt++;
    377 				goto miss;
    378 			}
    379 			KASSERT(newrt != NULL);
    380 			rt = newrt;
    381 			if (rt->rt_flags & RTF_XRESOLVE) {
    382 				msgtype = RTM_RESOLVE;
    383 				goto miss;
    384 			}
    385 			/* Inform listeners of the new route */
    386 			memset(&info, 0, sizeof(info));
    387 			info.rti_info[RTAX_DST] = rt_getkey(rt);
    388 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    389 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    390 			if (rt->rt_ifp != NULL) {
    391 				info.rti_info[RTAX_IFP] =
    392 				    rt->rt_ifp->if_dl->ifa_addr;
    393 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
    394 			}
    395 			rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0);
    396 		} else
    397 			rt->rt_refcnt++;
    398 	} else {
    399 		rtstat.rts_unreach++;
    400 	miss:	if (report) {
    401 			memset((void *)&info, 0, sizeof(info));
    402 			info.rti_info[RTAX_DST] = dst;
    403 			rt_missmsg(msgtype, &info, 0, err);
    404 		}
    405 	}
    406 	splx(s);
    407 	return newrt;
    408 }
    409 
    410 void
    411 rtfree(struct rtentry *rt)
    412 {
    413 	struct ifaddr *ifa;
    414 
    415 	KASSERT(rt != NULL);
    416 	KASSERT(rt->rt_refcnt > 0);
    417 
    418 	rt->rt_refcnt--;
    419 	if (rt->rt_refcnt == 0 && (rt->rt_flags & RTF_UP) == 0) {
    420 		rt_assert_inactive(rt);
    421 		rttrash--;
    422 		rt_timer_remove_all(rt, 0);
    423 		ifa = rt->rt_ifa;
    424 		rt->rt_ifa = NULL;
    425 		ifafree(ifa);
    426 		rt->rt_ifp = NULL;
    427 		rt_destroy(rt);
    428 		pool_put(&rtentry_pool, rt);
    429 	}
    430 }
    431 
    432 /*
    433  * Force a routing table entry to the specified
    434  * destination to go through the given gateway.
    435  * Normally called as a result of a routing redirect
    436  * message from the network layer.
    437  *
    438  * N.B.: must be called at splsoftnet
    439  */
    440 void
    441 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
    442 	const struct sockaddr *netmask, int flags, const struct sockaddr *src,
    443 	struct rtentry **rtp)
    444 {
    445 	struct rtentry *rt;
    446 	int error = 0;
    447 	uint64_t *stat = NULL;
    448 	struct rt_addrinfo info;
    449 	struct ifaddr *ifa;
    450 
    451 	/* verify the gateway is directly reachable */
    452 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
    453 		error = ENETUNREACH;
    454 		goto out;
    455 	}
    456 	rt = rtalloc1(dst, 0);
    457 	/*
    458 	 * If the redirect isn't from our current router for this dst,
    459 	 * it's either old or wrong.  If it redirects us to ourselves,
    460 	 * we have a routing loop, perhaps as a result of an interface
    461 	 * going down recently.
    462 	 */
    463 	if (!(flags & RTF_DONE) && rt &&
    464 	     (sockaddr_cmp(src, rt->rt_gateway) != 0 || rt->rt_ifa != ifa))
    465 		error = EINVAL;
    466 	else if (ifa_ifwithaddr(gateway))
    467 		error = EHOSTUNREACH;
    468 	if (error)
    469 		goto done;
    470 	/*
    471 	 * Create a new entry if we just got back a wildcard entry
    472 	 * or the lookup failed.  This is necessary for hosts
    473 	 * which use routing redirects generated by smart gateways
    474 	 * to dynamically build the routing tables.
    475 	 */
    476 	if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
    477 		goto create;
    478 	/*
    479 	 * Don't listen to the redirect if it's
    480 	 * for a route to an interface.
    481 	 */
    482 	if (rt->rt_flags & RTF_GATEWAY) {
    483 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
    484 			/*
    485 			 * Changing from route to net => route to host.
    486 			 * Create new route, rather than smashing route to net.
    487 			 */
    488 		create:
    489 			if (rt != NULL)
    490 				rtfree(rt);
    491 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
    492 			memset(&info, 0, sizeof(info));
    493 			info.rti_info[RTAX_DST] = dst;
    494 			info.rti_info[RTAX_GATEWAY] = gateway;
    495 			info.rti_info[RTAX_NETMASK] = netmask;
    496 			info.rti_ifa = ifa;
    497 			info.rti_flags = flags;
    498 			rt = NULL;
    499 			error = rtrequest1(RTM_ADD, &info, &rt);
    500 			if (rt != NULL)
    501 				flags = rt->rt_flags;
    502 			stat = &rtstat.rts_dynamic;
    503 		} else {
    504 			/*
    505 			 * Smash the current notion of the gateway to
    506 			 * this destination.  Should check about netmask!!!
    507 			 */
    508 			rt->rt_flags |= RTF_MODIFIED;
    509 			flags |= RTF_MODIFIED;
    510 			stat = &rtstat.rts_newgateway;
    511 			rt_setgate(rt, gateway);
    512 		}
    513 	} else
    514 		error = EHOSTUNREACH;
    515 done:
    516 	if (rt) {
    517 		if (rtp != NULL && !error)
    518 			*rtp = rt;
    519 		else
    520 			rtfree(rt);
    521 	}
    522 out:
    523 	if (error)
    524 		rtstat.rts_badredirect++;
    525 	else if (stat != NULL)
    526 		(*stat)++;
    527 	memset(&info, 0, sizeof(info));
    528 	info.rti_info[RTAX_DST] = dst;
    529 	info.rti_info[RTAX_GATEWAY] = gateway;
    530 	info.rti_info[RTAX_NETMASK] = netmask;
    531 	info.rti_info[RTAX_AUTHOR] = src;
    532 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
    533 }
    534 
    535 /*
    536  * Delete a route and generate a message
    537  */
    538 static int
    539 rtdeletemsg(struct rtentry *rt)
    540 {
    541 	int error;
    542 	struct rt_addrinfo info;
    543 
    544 	/*
    545 	 * Request the new route so that the entry is not actually
    546 	 * deleted.  That will allow the information being reported to
    547 	 * be accurate (and consistent with route_output()).
    548 	 */
    549 	memset(&info, 0, sizeof(info));
    550 	info.rti_info[RTAX_DST] = rt_getkey(rt);
    551 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    552 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    553 	info.rti_flags = rt->rt_flags;
    554 	error = rtrequest1(RTM_DELETE, &info, &rt);
    555 
    556 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
    557 
    558 	/* Adjust the refcount */
    559 	if (error == 0 && rt->rt_refcnt <= 0) {
    560 		rt->rt_refcnt++;
    561 		rtfree(rt);
    562 	}
    563 	return error;
    564 }
    565 
    566 static int
    567 rtflushclone1(struct rtentry *rt, void *arg)
    568 {
    569 	struct rtentry *parent;
    570 
    571 	parent = (struct rtentry *)arg;
    572 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
    573 		rtdeletemsg(rt);
    574 	return 0;
    575 }
    576 
    577 static void
    578 rtflushclone(sa_family_t family, struct rtentry *parent)
    579 {
    580 
    581 #ifdef DIAGNOSTIC
    582 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
    583 		panic("rtflushclone: called with a non-cloning route");
    584 #endif
    585 	rt_walktree(family, rtflushclone1, (void *)parent);
    586 }
    587 
    588 struct ifaddr *
    589 ifa_ifwithroute(int flags, const struct sockaddr *dst,
    590 	const struct sockaddr *gateway)
    591 {
    592 	struct ifaddr *ifa;
    593 	if ((flags & RTF_GATEWAY) == 0) {
    594 		/*
    595 		 * If we are adding a route to an interface,
    596 		 * and the interface is a pt to pt link
    597 		 * we should search for the destination
    598 		 * as our clue to the interface.  Otherwise
    599 		 * we can use the local address.
    600 		 */
    601 		ifa = NULL;
    602 		if ((flags & RTF_HOST) && gateway->sa_family != AF_LINK)
    603 			ifa = ifa_ifwithdstaddr(dst);
    604 		if (ifa == NULL)
    605 			ifa = ifa_ifwithaddr(gateway);
    606 	} else {
    607 		/*
    608 		 * If we are adding a route to a remote net
    609 		 * or host, the gateway may still be on the
    610 		 * other end of a pt to pt link.
    611 		 */
    612 		ifa = ifa_ifwithdstaddr(gateway);
    613 	}
    614 	if (ifa == NULL)
    615 		ifa = ifa_ifwithnet(gateway);
    616 	if (ifa == NULL) {
    617 		struct rtentry *rt = rtalloc1(dst, 0);
    618 		if (rt == NULL)
    619 			return NULL;
    620 		rt->rt_refcnt--;
    621 		if ((ifa = rt->rt_ifa) == NULL)
    622 			return NULL;
    623 	}
    624 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
    625 		struct ifaddr *oifa = ifa;
    626 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
    627 		if (ifa == NULL)
    628 			ifa = oifa;
    629 	}
    630 	return ifa;
    631 }
    632 
    633 int
    634 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
    635 	const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
    636 {
    637 	struct rt_addrinfo info;
    638 
    639 	memset(&info, 0, sizeof(info));
    640 	info.rti_flags = flags;
    641 	info.rti_info[RTAX_DST] = dst;
    642 	info.rti_info[RTAX_GATEWAY] = gateway;
    643 	info.rti_info[RTAX_NETMASK] = netmask;
    644 	return rtrequest1(req, &info, ret_nrt);
    645 }
    646 
    647 int
    648 rt_getifa(struct rt_addrinfo *info)
    649 {
    650 	struct ifaddr *ifa;
    651 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
    652 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
    653 	const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
    654 	const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
    655 	int flags = info->rti_flags;
    656 
    657 	/*
    658 	 * ifp may be specified by sockaddr_dl when protocol address
    659 	 * is ambiguous
    660 	 */
    661 	if (info->rti_ifp == NULL && ifpaddr != NULL
    662 	    && ifpaddr->sa_family == AF_LINK &&
    663 	    (ifa = ifa_ifwithnet(ifpaddr)) != NULL)
    664 		info->rti_ifp = ifa->ifa_ifp;
    665 	if (info->rti_ifa == NULL && ifaaddr != NULL)
    666 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
    667 	if (info->rti_ifa == NULL) {
    668 		const struct sockaddr *sa;
    669 
    670 		sa = ifaaddr != NULL ? ifaaddr :
    671 		    (gateway != NULL ? gateway : dst);
    672 		if (sa != NULL && info->rti_ifp != NULL)
    673 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
    674 		else if (dst != NULL && gateway != NULL)
    675 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
    676 		else if (sa != NULL)
    677 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
    678 	}
    679 	if ((ifa = info->rti_ifa) == NULL)
    680 		return ENETUNREACH;
    681 	if (ifa->ifa_getifa != NULL) {
    682 		info->rti_ifa = ifa = (*ifa->ifa_getifa)(ifa, dst);
    683 		if (ifa == NULL)
    684 			return ENETUNREACH;
    685 	}
    686 	if (info->rti_ifp == NULL)
    687 		info->rti_ifp = ifa->ifa_ifp;
    688 	return 0;
    689 }
    690 
    691 int
    692 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
    693 {
    694 	int s = splsoftnet();
    695 	int error = 0, rc;
    696 	struct rtentry *rt, *crt;
    697 	rtbl_t *rtbl;
    698 	struct ifaddr *ifa, *ifa2;
    699 	struct sockaddr_storage maskeddst;
    700 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
    701 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
    702 	const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
    703 	int flags = info->rti_flags;
    704 #define senderr(x) { error = x ; goto bad; }
    705 
    706 	if ((rtbl = rt_gettable(dst->sa_family)) == NULL)
    707 		senderr(ESRCH);
    708 	if (flags & RTF_HOST)
    709 		netmask = NULL;
    710 	switch (req) {
    711 	case RTM_DELETE:
    712 		if (netmask) {
    713 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    714 			    netmask);
    715 			dst = (struct sockaddr *)&maskeddst;
    716 		}
    717 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
    718 			senderr(ESRCH);
    719 		if ((rt->rt_flags & RTF_CLONING) != 0) {
    720 			/* clean up any cloned children */
    721 			rtflushclone(dst->sa_family, rt);
    722 		}
    723 		if ((rt = rt_deladdr(rtbl, dst, netmask)) == NULL)
    724 			senderr(ESRCH);
    725 		if (rt->rt_gwroute) {
    726 			rtfree(rt->rt_gwroute);
    727 			rt->rt_gwroute = NULL;
    728 		}
    729 		if (rt->rt_parent) {
    730 			rt->rt_parent->rt_refcnt--;
    731 			rt->rt_parent = NULL;
    732 		}
    733 		rt->rt_flags &= ~RTF_UP;
    734 		if ((ifa = rt->rt_ifa)) {
    735 			if (ifa->ifa_flags & IFA_ROUTE &&
    736 			    rt_ifa_connected(rt, ifa)) {
    737 				RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    738 				    "deleted IFA_ROUTE\n",
    739 				    (void *)rt->_rt_key, (void *)ifa);
    740 				ifa->ifa_flags &= ~IFA_ROUTE;
    741 			}
    742 			if (ifa->ifa_rtrequest)
    743 				ifa->ifa_rtrequest(RTM_DELETE, rt, info);
    744 		}
    745 		rttrash++;
    746 		if (ret_nrt)
    747 			*ret_nrt = rt;
    748 		else if (rt->rt_refcnt <= 0) {
    749 			rt->rt_refcnt++;
    750 			rtfree(rt);
    751 		}
    752 		break;
    753 
    754 	case RTM_RESOLVE:
    755 		if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
    756 			senderr(EINVAL);
    757 		if ((rt->rt_flags & RTF_CLONING) == 0)
    758 			senderr(EINVAL);
    759 		ifa = rt->rt_ifa;
    760 		flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC);
    761 		flags |= RTF_CLONED;
    762 		gateway = rt->rt_gateway;
    763 		flags |= RTF_HOST;
    764 		goto makeroute;
    765 
    766 	case RTM_ADD:
    767 		if (info->rti_ifa == NULL && (error = rt_getifa(info)))
    768 			senderr(error);
    769 		ifa = info->rti_ifa;
    770 	makeroute:
    771 		/* Already at splsoftnet() so pool_get/pool_put are safe */
    772 		rt = pool_get(&rtentry_pool, PR_NOWAIT);
    773 		if (rt == NULL)
    774 			senderr(ENOBUFS);
    775 		memset(rt, 0, sizeof(*rt));
    776 		rt->rt_flags = RTF_UP | flags;
    777 		LIST_INIT(&rt->rt_timer);
    778 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    779 		if (rt_setkey(rt, dst, M_NOWAIT) == NULL ||
    780 		    rt_setgate(rt, gateway) != 0) {
    781 			pool_put(&rtentry_pool, rt);
    782 			senderr(ENOBUFS);
    783 		}
    784 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    785 		if (netmask) {
    786 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    787 			    netmask);
    788 			rt_setkey(rt, (struct sockaddr *)&maskeddst, M_NOWAIT);
    789 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    790 		} else {
    791 			rt_setkey(rt, dst, M_NOWAIT);
    792 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    793 		}
    794 		rt_set_ifa(rt, ifa);
    795 		if (info->rti_info[RTAX_TAG] != NULL)
    796 			rt_settag(rt, info->rti_info[RTAX_TAG]);
    797 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    798 		if (info->rti_info[RTAX_IFP] != NULL &&
    799 		    (ifa2 = ifa_ifwithnet(info->rti_info[RTAX_IFP])) != NULL &&
    800 		    ifa2->ifa_ifp != NULL)
    801 			rt->rt_ifp = ifa2->ifa_ifp;
    802 		else
    803 			rt->rt_ifp = ifa->ifa_ifp;
    804 		if (req == RTM_RESOLVE) {
    805 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
    806 			rt->rt_parent = *ret_nrt;
    807 			rt->rt_parent->rt_refcnt++;
    808 		}
    809 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    810 		rc = rt_addaddr(rtbl, rt, netmask);
    811 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    812 		if (rc != 0 && (crt = rtalloc1(rt_getkey(rt), 0)) != NULL) {
    813 			/* overwrite cloned route */
    814 			if ((crt->rt_flags & RTF_CLONED) != 0) {
    815 				rtdeletemsg(crt);
    816 				rc = rt_addaddr(rtbl, rt, netmask);
    817 			}
    818 			rtfree(crt);
    819 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    820 		}
    821 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    822 		if (rc != 0) {
    823 			ifafree(ifa);
    824 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
    825 				rtfree(rt->rt_parent);
    826 			if (rt->rt_gwroute)
    827 				rtfree(rt->rt_gwroute);
    828 			rt_destroy(rt);
    829 			pool_put(&rtentry_pool, rt);
    830 			senderr(rc);
    831 		}
    832 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    833 		if (ifa->ifa_rtrequest)
    834 			ifa->ifa_rtrequest(req, rt, info);
    835 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    836 		if (ret_nrt) {
    837 			*ret_nrt = rt;
    838 			rt->rt_refcnt++;
    839 		}
    840 		if ((rt->rt_flags & RTF_CLONING) != 0) {
    841 			/* clean up any cloned children */
    842 			rtflushclone(dst->sa_family, rt);
    843 		}
    844 		rtflushall(dst->sa_family);
    845 		break;
    846 	case RTM_GET:
    847 		if (netmask != NULL) {
    848 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    849 			    netmask);
    850 			dst = (struct sockaddr *)&maskeddst;
    851 		}
    852 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
    853 			senderr(ESRCH);
    854 		if (ret_nrt != NULL) {
    855 			*ret_nrt = rt;
    856 			rt->rt_refcnt++;
    857 		}
    858 		break;
    859 	}
    860 bad:
    861 	splx(s);
    862 	return error;
    863 }
    864 
    865 int
    866 rt_setgate(struct rtentry *rt, const struct sockaddr *gate)
    867 {
    868 	KASSERT(rt != rt->rt_gwroute);
    869 
    870 	KASSERT(rt->_rt_key != NULL);
    871 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    872 
    873 	if (rt->rt_gwroute) {
    874 		rtfree(rt->rt_gwroute);
    875 		rt->rt_gwroute = NULL;
    876 	}
    877 	KASSERT(rt->_rt_key != NULL);
    878 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    879 	if (rt->rt_gateway != NULL)
    880 		sockaddr_free(rt->rt_gateway);
    881 	KASSERT(rt->_rt_key != NULL);
    882 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    883 	if ((rt->rt_gateway = sockaddr_dup(gate, M_ZERO | M_NOWAIT)) == NULL)
    884 		return ENOMEM;
    885 	KASSERT(rt->_rt_key != NULL);
    886 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    887 
    888 	if (rt->rt_flags & RTF_GATEWAY) {
    889 		KASSERT(rt->_rt_key != NULL);
    890 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    891 		rt->rt_gwroute = rtalloc1(gate, 1);
    892 		/*
    893 		 * If we switched gateways, grab the MTU from the new
    894 		 * gateway route if the current MTU, if the current MTU is
    895 		 * greater than the MTU of gateway.
    896 		 * Note that, if the MTU of gateway is 0, we will reset the
    897 		 * MTU of the route to run PMTUD again from scratch. XXX
    898 		 */
    899 		KASSERT(rt->_rt_key != NULL);
    900 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    901 		if (rt->rt_gwroute
    902 		    && !(rt->rt_rmx.rmx_locks & RTV_MTU)
    903 		    && rt->rt_rmx.rmx_mtu
    904 		    && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
    905 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
    906 		}
    907 	}
    908 	KASSERT(rt->_rt_key != NULL);
    909 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    910 	return 0;
    911 }
    912 
    913 static void
    914 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
    915 	const struct sockaddr *netmask)
    916 {
    917 	const char *netmaskp = &netmask->sa_data[0],
    918 	           *srcp = &src->sa_data[0];
    919 	char *dstp = &dst->sa_data[0];
    920 	const char *maskend = (char *)dst + MIN(netmask->sa_len, src->sa_len);
    921 	const char *srcend = (char *)dst + src->sa_len;
    922 
    923 	dst->sa_len = src->sa_len;
    924 	dst->sa_family = src->sa_family;
    925 
    926 	while (dstp < maskend)
    927 		*dstp++ = *srcp++ & *netmaskp++;
    928 	if (dstp < srcend)
    929 		memset(dstp, 0, (size_t)(srcend - dstp));
    930 }
    931 
    932 /*
    933  * Inform the routing socket of a route change.
    934  */
    935 void
    936 rt_newmsg(int cmd, struct rtentry *rt)
    937 {
    938 	struct rt_addrinfo info;
    939 
    940 	memset((void *)&info, 0, sizeof(info));
    941 	info.rti_info[RTAX_DST] = rt_getkey(rt);
    942 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    943 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    944 	if (rt->rt_ifp) {
    945 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
    946 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
    947 	}
    948 
    949 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
    950 }
    951 
    952 /*
    953  * Set up or tear down a routing table entry, normally
    954  * for an interface.
    955  */
    956 int
    957 rtinit(struct ifaddr *ifa, int cmd, int flags)
    958 {
    959 	struct rtentry *rt;
    960 	struct sockaddr *dst, *odst;
    961 	struct sockaddr_storage maskeddst;
    962 	struct rtentry *nrt = NULL;
    963 	int error;
    964 	struct rt_addrinfo info;
    965 	struct sockaddr_dl *sdl;
    966 	const struct sockaddr_dl *ifsdl;
    967 
    968 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
    969 	if (cmd == RTM_DELETE) {
    970 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
    971 			/* Delete subnet route for this interface */
    972 			odst = dst;
    973 			dst = (struct sockaddr *)&maskeddst;
    974 			rt_maskedcopy(odst, dst, ifa->ifa_netmask);
    975 		}
    976 		if ((rt = rtalloc1(dst, 0)) != NULL) {
    977 			rt->rt_refcnt--;
    978 			if (rt->rt_ifa != ifa)
    979 				return (flags & RTF_HOST) ? EHOSTUNREACH
    980 							: ENETUNREACH;
    981 		}
    982 	}
    983 	memset(&info, 0, sizeof(info));
    984 	info.rti_ifa = ifa;
    985 	info.rti_flags = flags | ifa->ifa_flags;
    986 	info.rti_info[RTAX_DST] = dst;
    987 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
    988 	/*
    989 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
    990 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
    991 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
    992 	 * change it to meet bsdi4 behavior.
    993 	 */
    994 	if (cmd != RTM_LLINFO_UPD)
    995 		info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
    996 	error = rtrequest1((cmd == RTM_LLINFO_UPD) ? RTM_GET : cmd, &info,
    997 	    &nrt);
    998 	if (error != 0 || (rt = nrt) == NULL)
    999 		;
   1000 	else switch (cmd) {
   1001 	case RTM_DELETE:
   1002 		rt_newmsg(cmd, nrt);
   1003 		if (rt->rt_refcnt <= 0) {
   1004 			rt->rt_refcnt++;
   1005 			rtfree(rt);
   1006 		}
   1007 		break;
   1008 	case RTM_LLINFO_UPD:
   1009 		rt->rt_refcnt--;
   1010 		RT_DPRINTF("%s: updating%s\n", __func__,
   1011 		    ((rt->rt_flags & RTF_LLINFO) == 0) ? " (no llinfo)" : "");
   1012 
   1013 		ifsdl = ifa->ifa_ifp->if_sadl;
   1014 
   1015 		if ((rt->rt_flags & RTF_LLINFO) != 0 &&
   1016 		    (sdl = satosdl(rt->rt_gateway)) != NULL &&
   1017 		    sdl->sdl_family == AF_LINK &&
   1018 		    sockaddr_dl_setaddr(sdl, sdl->sdl_len, CLLADDR(ifsdl),
   1019 		                        ifa->ifa_ifp->if_addrlen) == NULL) {
   1020 			error = EINVAL;
   1021 			break;
   1022 		}
   1023 
   1024 		if (cmd == RTM_LLINFO_UPD && ifa->ifa_rtrequest != NULL)
   1025 			ifa->ifa_rtrequest(RTM_LLINFO_UPD, rt, &info);
   1026 		rt_newmsg(RTM_CHANGE, nrt);
   1027 		break;
   1028 	case RTM_ADD:
   1029 		rt->rt_refcnt--;
   1030 		if (rt->rt_ifa != ifa) {
   1031 			printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
   1032 				rt->rt_ifa);
   1033 			if (rt->rt_ifa->ifa_rtrequest != NULL) {
   1034 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
   1035 				    &info);
   1036 			}
   1037 			rt_replace_ifa(rt, ifa);
   1038 			rt->rt_ifp = ifa->ifa_ifp;
   1039 			if (ifa->ifa_rtrequest != NULL)
   1040 				ifa->ifa_rtrequest(RTM_ADD, rt, &info);
   1041 		}
   1042 		rt_newmsg(cmd, nrt);
   1043 		break;
   1044 	}
   1045 	return error;
   1046 }
   1047 
   1048 static const struct in_addr inmask32 = {.s_addr = INADDR_BROADCAST};
   1049 
   1050 /* Subroutine for rt_ifa_addlocal() and rt_ifa_remlocal() */
   1051 static int
   1052 rt_ifa_localrequest(int cmd, struct ifaddr *ifa)
   1053 {
   1054 	struct sockaddr *all1_sa;
   1055 	struct sockaddr_in all1_sin;
   1056 #ifdef INET6
   1057 	struct sockaddr_in6 all1_sin6;
   1058 #endif
   1059 	struct rtentry *nrt = NULL;
   1060 	int flags, e;
   1061 
   1062 	switch(ifa->ifa_addr->sa_family) {
   1063 	case AF_INET:
   1064 		sockaddr_in_init(&all1_sin, &inmask32, 0);
   1065 		all1_sa = (struct sockaddr *)&all1_sin;
   1066 		break;
   1067 #ifdef INET6
   1068 	case AF_INET6:
   1069 		sockaddr_in6_init(&all1_sin6, &in6mask128, 0, 0, 0);
   1070 		all1_sa = (struct sockaddr *)&all1_sin6;
   1071 		break;
   1072 #endif
   1073 	default:
   1074 		return 0;
   1075 	}
   1076 
   1077 	flags = RTF_UP | RTF_HOST | RTF_LOCAL;
   1078 	if (!(ifa->ifa_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)))
   1079 		flags |= RTF_LLINFO;
   1080 	e = rtrequest(cmd, ifa->ifa_addr, ifa->ifa_addr, all1_sa, flags, &nrt);
   1081 
   1082 	/* Make sure rt_ifa be equal to IFA, the second argument of the
   1083 	 * function. */
   1084 	if (cmd == RTM_ADD && nrt && ifa != nrt->rt_ifa)
   1085 		rt_replace_ifa(nrt, ifa);
   1086 
   1087 	rt_newaddrmsg(cmd, ifa, e, nrt);
   1088 	if (nrt) {
   1089 		if (cmd == RTM_DELETE) {
   1090 			if (nrt->rt_refcnt <= 0) {
   1091 				/* XXX: we should free the entry ourselves. */
   1092 				nrt->rt_refcnt++;
   1093 				rtfree(nrt);
   1094 			}
   1095 		} else {
   1096 			/* the cmd must be RTM_ADD here */
   1097 			nrt->rt_refcnt--;
   1098 		}
   1099 	}
   1100 	return e;
   1101 }
   1102 
   1103 /*
   1104  * Create a local route entry for the address.
   1105  * Announce the addition of the address and the route to the routing socket.
   1106  */
   1107 int
   1108 rt_ifa_addlocal(struct ifaddr *ifa)
   1109 {
   1110 	struct rtentry *rt;
   1111 	int e;
   1112 
   1113 	/* If there is no loopback entry, allocate one. */
   1114 	rt = rtalloc1(ifa->ifa_addr, 0);
   1115 	if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
   1116 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
   1117 		e = rt_ifa_localrequest(RTM_ADD, ifa);
   1118 	else {
   1119 		e = 0;
   1120 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1121 	}
   1122 	if (rt != NULL)
   1123 		rt->rt_refcnt--;
   1124 	return e;
   1125 }
   1126 
   1127 /*
   1128  * Remove the local route entry for the address.
   1129  * Announce the removal of the address and the route to the routing socket.
   1130  */
   1131 int
   1132 rt_ifa_remlocal(struct ifaddr *ifa, struct ifaddr *alt_ifa)
   1133 {
   1134 	struct rtentry *rt;
   1135 	int e = 0;
   1136 
   1137 	rt = rtalloc1(ifa->ifa_addr, 0);
   1138 
   1139 	/*
   1140 	 * Before deleting, check if a corresponding loopbacked
   1141 	 * host route surely exists.  With this check, we can avoid
   1142 	 * deleting an interface direct route whose destination is
   1143 	 * the same as the address being removed.  This can happen
   1144 	 * when removing a subnet-router anycast address on an
   1145 	 * interface attached to a shared medium.
   1146 	 */
   1147 	if (rt != NULL &&
   1148 	    (rt->rt_flags & RTF_HOST) &&
   1149 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK))
   1150 	{
   1151 		/* If we cannot replace the route's ifaddr with the equivalent
   1152 		 * ifaddr of another interface, I believe it is safest to
   1153 		 * delete the route.
   1154 		 */
   1155 		if (alt_ifa == NULL)
   1156 			e = rt_ifa_localrequest(RTM_DELETE, ifa);
   1157 		else {
   1158 			rt_replace_ifa(rt, alt_ifa);
   1159 			rt_newmsg(RTM_CHANGE, rt);
   1160 		}
   1161 	} else
   1162 		rt_newaddrmsg(RTM_DELADDR, ifa, 0, NULL);
   1163 	if (rt != NULL)
   1164 		rt->rt_refcnt--;
   1165 	return e;
   1166 }
   1167 
   1168 /*
   1169  * Route timer routines.  These routes allow functions to be called
   1170  * for various routes at any time.  This is useful in supporting
   1171  * path MTU discovery and redirect route deletion.
   1172  *
   1173  * This is similar to some BSDI internal functions, but it provides
   1174  * for multiple queues for efficiency's sake...
   1175  */
   1176 
   1177 LIST_HEAD(, rttimer_queue) rttimer_queue_head;
   1178 static int rt_init_done = 0;
   1179 
   1180 #define RTTIMER_CALLOUT(r)	do {					\
   1181 		if (r->rtt_func != NULL) {				\
   1182 			(*r->rtt_func)(r->rtt_rt, r);			\
   1183 		} else {						\
   1184 			rtrequest((int) RTM_DELETE,			\
   1185 				  rt_getkey(r->rtt_rt),			\
   1186 				  0, 0, 0, 0);				\
   1187 		}							\
   1188 	} while (/*CONSTCOND*/0)
   1189 
   1190 /*
   1191  * Some subtle order problems with domain initialization mean that
   1192  * we cannot count on this being run from rt_init before various
   1193  * protocol initializations are done.  Therefore, we make sure
   1194  * that this is run when the first queue is added...
   1195  */
   1196 
   1197 void
   1198 rt_timer_init(void)
   1199 {
   1200 	assert(rt_init_done == 0);
   1201 
   1202 	LIST_INIT(&rttimer_queue_head);
   1203 	callout_init(&rt_timer_ch, 0);
   1204 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
   1205 	rt_init_done = 1;
   1206 }
   1207 
   1208 struct rttimer_queue *
   1209 rt_timer_queue_create(u_int timeout)
   1210 {
   1211 	struct rttimer_queue *rtq;
   1212 
   1213 	if (rt_init_done == 0)
   1214 		rt_timer_init();
   1215 
   1216 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
   1217 	if (rtq == NULL)
   1218 		return NULL;
   1219 	memset(rtq, 0, sizeof(*rtq));
   1220 
   1221 	rtq->rtq_timeout = timeout;
   1222 	TAILQ_INIT(&rtq->rtq_head);
   1223 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
   1224 
   1225 	return rtq;
   1226 }
   1227 
   1228 void
   1229 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
   1230 {
   1231 
   1232 	rtq->rtq_timeout = timeout;
   1233 }
   1234 
   1235 void
   1236 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy)
   1237 {
   1238 	struct rttimer *r;
   1239 
   1240 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
   1241 		LIST_REMOVE(r, rtt_link);
   1242 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
   1243 		if (destroy)
   1244 			RTTIMER_CALLOUT(r);
   1245 		/* we are already at splsoftnet */
   1246 		pool_put(&rttimer_pool, r);
   1247 		if (rtq->rtq_count > 0)
   1248 			rtq->rtq_count--;
   1249 		else
   1250 			printf("rt_timer_queue_remove_all: "
   1251 			    "rtq_count reached 0\n");
   1252 	}
   1253 }
   1254 
   1255 void
   1256 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
   1257 {
   1258 
   1259 	rt_timer_queue_remove_all(rtq, destroy);
   1260 
   1261 	LIST_REMOVE(rtq, rtq_link);
   1262 
   1263 	/*
   1264 	 * Caller is responsible for freeing the rttimer_queue structure.
   1265 	 */
   1266 }
   1267 
   1268 unsigned long
   1269 rt_timer_count(struct rttimer_queue *rtq)
   1270 {
   1271 	return rtq->rtq_count;
   1272 }
   1273 
   1274 void
   1275 rt_timer_remove_all(struct rtentry *rt, int destroy)
   1276 {
   1277 	struct rttimer *r;
   1278 
   1279 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
   1280 		LIST_REMOVE(r, rtt_link);
   1281 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
   1282 		if (destroy)
   1283 			RTTIMER_CALLOUT(r);
   1284 		if (r->rtt_queue->rtq_count > 0)
   1285 			r->rtt_queue->rtq_count--;
   1286 		else
   1287 			printf("rt_timer_remove_all: rtq_count reached 0\n");
   1288 		/* we are already at splsoftnet */
   1289 		pool_put(&rttimer_pool, r);
   1290 	}
   1291 }
   1292 
   1293 int
   1294 rt_timer_add(struct rtentry *rt,
   1295 	void (*func)(struct rtentry *, struct rttimer *),
   1296 	struct rttimer_queue *queue)
   1297 {
   1298 	struct rttimer *r;
   1299 	int s;
   1300 
   1301 	/*
   1302 	 * If there's already a timer with this action, destroy it before
   1303 	 * we add a new one.
   1304 	 */
   1305 	LIST_FOREACH(r, &rt->rt_timer, rtt_link) {
   1306 		if (r->rtt_func == func)
   1307 			break;
   1308 	}
   1309 	if (r != NULL) {
   1310 		LIST_REMOVE(r, rtt_link);
   1311 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
   1312 		if (r->rtt_queue->rtq_count > 0)
   1313 			r->rtt_queue->rtq_count--;
   1314 		else
   1315 			printf("rt_timer_add: rtq_count reached 0\n");
   1316 	} else {
   1317 		s = splsoftnet();
   1318 		r = pool_get(&rttimer_pool, PR_NOWAIT);
   1319 		splx(s);
   1320 		if (r == NULL)
   1321 			return ENOBUFS;
   1322 	}
   1323 
   1324 	memset(r, 0, sizeof(*r));
   1325 
   1326 	r->rtt_rt = rt;
   1327 	r->rtt_time = time_uptime;
   1328 	r->rtt_func = func;
   1329 	r->rtt_queue = queue;
   1330 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
   1331 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
   1332 	r->rtt_queue->rtq_count++;
   1333 
   1334 	return 0;
   1335 }
   1336 
   1337 /* ARGSUSED */
   1338 void
   1339 rt_timer_timer(void *arg)
   1340 {
   1341 	struct rttimer_queue *rtq;
   1342 	struct rttimer *r;
   1343 	int s;
   1344 
   1345 	s = splsoftnet();
   1346 	LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) {
   1347 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
   1348 		    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
   1349 			LIST_REMOVE(r, rtt_link);
   1350 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
   1351 			RTTIMER_CALLOUT(r);
   1352 			pool_put(&rttimer_pool, r);
   1353 			if (rtq->rtq_count > 0)
   1354 				rtq->rtq_count--;
   1355 			else
   1356 				printf("rt_timer_timer: rtq_count reached 0\n");
   1357 		}
   1358 	}
   1359 	splx(s);
   1360 
   1361 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
   1362 }
   1363 
   1364 static struct rtentry *
   1365 _rtcache_init(struct route *ro, int flag)
   1366 {
   1367 	rtcache_invariants(ro);
   1368 	KASSERT(ro->_ro_rt == NULL);
   1369 
   1370 	if (rtcache_getdst(ro) == NULL)
   1371 		return NULL;
   1372 	ro->ro_invalid = false;
   1373 	if ((ro->_ro_rt = rtalloc1(rtcache_getdst(ro), flag)) != NULL)
   1374 		rtcache(ro);
   1375 
   1376 	rtcache_invariants(ro);
   1377 	return ro->_ro_rt;
   1378 }
   1379 
   1380 struct rtentry *
   1381 rtcache_init(struct route *ro)
   1382 {
   1383 	return _rtcache_init(ro, 1);
   1384 }
   1385 
   1386 struct rtentry *
   1387 rtcache_init_noclone(struct route *ro)
   1388 {
   1389 	return _rtcache_init(ro, 0);
   1390 }
   1391 
   1392 struct rtentry *
   1393 rtcache_update(struct route *ro, int clone)
   1394 {
   1395 	rtcache_clear(ro);
   1396 	return _rtcache_init(ro, clone);
   1397 }
   1398 
   1399 void
   1400 rtcache_copy(struct route *new_ro, const struct route *old_ro)
   1401 {
   1402 	struct rtentry *rt;
   1403 
   1404 	KASSERT(new_ro != old_ro);
   1405 	rtcache_invariants(new_ro);
   1406 	rtcache_invariants(old_ro);
   1407 
   1408 	if ((rt = rtcache_validate(old_ro)) != NULL)
   1409 		rt->rt_refcnt++;
   1410 
   1411 	if (rtcache_getdst(old_ro) == NULL ||
   1412 	    rtcache_setdst(new_ro, rtcache_getdst(old_ro)) != 0)
   1413 		return;
   1414 
   1415 	new_ro->ro_invalid = false;
   1416 	if ((new_ro->_ro_rt = rt) != NULL)
   1417 		rtcache(new_ro);
   1418 	rtcache_invariants(new_ro);
   1419 }
   1420 
   1421 static struct dom_rtlist invalid_routes = LIST_HEAD_INITIALIZER(dom_rtlist);
   1422 
   1423 static void
   1424 rtcache_invalidate(struct dom_rtlist *rtlist)
   1425 {
   1426 	struct route *ro;
   1427 
   1428 	while ((ro = LIST_FIRST(rtlist)) != NULL) {
   1429 		rtcache_invariants(ro);
   1430 		KASSERT(ro->_ro_rt != NULL);
   1431 		ro->ro_invalid = true;
   1432 		LIST_REMOVE(ro, ro_rtcache_next);
   1433 		LIST_INSERT_HEAD(&invalid_routes, ro, ro_rtcache_next);
   1434 		rtcache_invariants(ro);
   1435 	}
   1436 }
   1437 
   1438 static void
   1439 rtcache_clear(struct route *ro)
   1440 {
   1441 	rtcache_invariants(ro);
   1442 	if (ro->_ro_rt == NULL)
   1443 		return;
   1444 
   1445 	LIST_REMOVE(ro, ro_rtcache_next);
   1446 
   1447 	rtfree(ro->_ro_rt);
   1448 	ro->_ro_rt = NULL;
   1449 	ro->ro_invalid = false;
   1450 	rtcache_invariants(ro);
   1451 }
   1452 
   1453 struct rtentry *
   1454 rtcache_lookup2(struct route *ro, const struct sockaddr *dst, int clone,
   1455     int *hitp)
   1456 {
   1457 	const struct sockaddr *odst;
   1458 	struct rtentry *rt = NULL;
   1459 
   1460 	odst = rtcache_getdst(ro);
   1461 	if (odst == NULL)
   1462 		goto miss;
   1463 
   1464 	if (sockaddr_cmp(odst, dst) != 0) {
   1465 		rtcache_free(ro);
   1466 		goto miss;
   1467 	}
   1468 
   1469 	rt = rtcache_validate(ro);
   1470 	if (rt == NULL) {
   1471 		rtcache_clear(ro);
   1472 		goto miss;
   1473 	}
   1474 
   1475 	*hitp = 1;
   1476 	rtcache_invariants(ro);
   1477 
   1478 	return rt;
   1479 miss:
   1480 	*hitp = 0;
   1481 	if (rtcache_setdst(ro, dst) == 0)
   1482 		rt = _rtcache_init(ro, clone);
   1483 
   1484 	rtcache_invariants(ro);
   1485 
   1486 	return rt;
   1487 }
   1488 
   1489 void
   1490 rtcache_free(struct route *ro)
   1491 {
   1492 	rtcache_clear(ro);
   1493 	if (ro->ro_sa != NULL) {
   1494 		sockaddr_free(ro->ro_sa);
   1495 		ro->ro_sa = NULL;
   1496 	}
   1497 	rtcache_invariants(ro);
   1498 }
   1499 
   1500 int
   1501 rtcache_setdst(struct route *ro, const struct sockaddr *sa)
   1502 {
   1503 	KASSERT(sa != NULL);
   1504 
   1505 	rtcache_invariants(ro);
   1506 	if (ro->ro_sa != NULL) {
   1507 		if (ro->ro_sa->sa_family == sa->sa_family) {
   1508 			rtcache_clear(ro);
   1509 			sockaddr_copy(ro->ro_sa, ro->ro_sa->sa_len, sa);
   1510 			rtcache_invariants(ro);
   1511 			return 0;
   1512 		}
   1513 		/* free ro_sa, wrong family */
   1514 		rtcache_free(ro);
   1515 	}
   1516 
   1517 	KASSERT(ro->_ro_rt == NULL);
   1518 
   1519 	if ((ro->ro_sa = sockaddr_dup(sa, M_ZERO | M_NOWAIT)) == NULL) {
   1520 		rtcache_invariants(ro);
   1521 		return ENOMEM;
   1522 	}
   1523 	rtcache_invariants(ro);
   1524 	return 0;
   1525 }
   1526 
   1527 const struct sockaddr *
   1528 rt_settag(struct rtentry *rt, const struct sockaddr *tag)
   1529 {
   1530 	if (rt->rt_tag != tag) {
   1531 		if (rt->rt_tag != NULL)
   1532 			sockaddr_free(rt->rt_tag);
   1533 		rt->rt_tag = sockaddr_dup(tag, M_ZERO | M_NOWAIT);
   1534 	}
   1535 	return rt->rt_tag;
   1536 }
   1537 
   1538 struct sockaddr *
   1539 rt_gettag(struct rtentry *rt)
   1540 {
   1541 	return rt->rt_tag;
   1542 }
   1543