Home | History | Annotate | Line # | Download | only in net
route.c revision 1.146
      1 /*	$NetBSD: route.c,v 1.146 2015/07/17 02:21:08 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 #include "opt_inet.h"
     94 #include "opt_route.h"
     95 
     96 #include <sys/cdefs.h>
     97 __KERNEL_RCSID(0, "$NetBSD: route.c,v 1.146 2015/07/17 02:21:08 ozaki-r 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. If success, refcnt of a returned rtentry
    359  * will be incremented. The caller has to rtfree it by itself.
    360  */
    361 struct rtentry *
    362 rtalloc1(const struct sockaddr *dst, int report)
    363 {
    364 	rtbl_t *rtbl = rt_gettable(dst->sa_family);
    365 	struct rtentry *rt;
    366 	struct rtentry *newrt = NULL;
    367 	struct rt_addrinfo info;
    368 	int  s = splsoftnet(), err = 0, msgtype = RTM_MISS;
    369 
    370 	if (rtbl != NULL && (rt = rt_matchaddr(rtbl, dst)) != NULL) {
    371 		newrt = rt;
    372 		if (report && (rt->rt_flags & RTF_CLONING)) {
    373 			err = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
    374 			    &newrt);
    375 			if (err) {
    376 				newrt = rt;
    377 				rt->rt_refcnt++;
    378 				goto miss;
    379 			}
    380 			KASSERT(newrt != NULL);
    381 			rt = newrt;
    382 			rt->rt_refcnt++;
    383 			if (rt->rt_flags & RTF_XRESOLVE) {
    384 				msgtype = RTM_RESOLVE;
    385 				goto miss;
    386 			}
    387 			/* Inform listeners of the new route */
    388 			memset(&info, 0, sizeof(info));
    389 			info.rti_info[RTAX_DST] = rt_getkey(rt);
    390 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    391 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    392 			if (rt->rt_ifp != NULL) {
    393 				info.rti_info[RTAX_IFP] =
    394 				    rt->rt_ifp->if_dl->ifa_addr;
    395 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
    396 			}
    397 			rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0);
    398 		} else
    399 			rt->rt_refcnt++;
    400 	} else {
    401 		rtstat.rts_unreach++;
    402 	miss:	if (report) {
    403 			memset((void *)&info, 0, sizeof(info));
    404 			info.rti_info[RTAX_DST] = dst;
    405 			rt_missmsg(msgtype, &info, 0, err);
    406 		}
    407 	}
    408 	splx(s);
    409 	return newrt;
    410 }
    411 
    412 void
    413 rtfree(struct rtentry *rt)
    414 {
    415 	struct ifaddr *ifa;
    416 
    417 	KASSERT(rt != NULL);
    418 	KASSERT(rt->rt_refcnt > 0);
    419 
    420 	rt->rt_refcnt--;
    421 	if (rt->rt_refcnt == 0 && (rt->rt_flags & RTF_UP) == 0) {
    422 		rt_assert_inactive(rt);
    423 		rttrash--;
    424 		rt_timer_remove_all(rt, 0);
    425 		ifa = rt->rt_ifa;
    426 		rt->rt_ifa = NULL;
    427 		ifafree(ifa);
    428 		rt->rt_ifp = NULL;
    429 		rt_destroy(rt);
    430 		pool_put(&rtentry_pool, rt);
    431 	}
    432 }
    433 
    434 /*
    435  * Force a routing table entry to the specified
    436  * destination to go through the given gateway.
    437  * Normally called as a result of a routing redirect
    438  * message from the network layer.
    439  *
    440  * N.B.: must be called at splsoftnet
    441  */
    442 void
    443 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
    444 	const struct sockaddr *netmask, int flags, const struct sockaddr *src,
    445 	struct rtentry **rtp)
    446 {
    447 	struct rtentry *rt;
    448 	int error = 0;
    449 	uint64_t *stat = NULL;
    450 	struct rt_addrinfo info;
    451 	struct ifaddr *ifa;
    452 
    453 	/* verify the gateway is directly reachable */
    454 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
    455 		error = ENETUNREACH;
    456 		goto out;
    457 	}
    458 	rt = rtalloc1(dst, 0);
    459 	/*
    460 	 * If the redirect isn't from our current router for this dst,
    461 	 * it's either old or wrong.  If it redirects us to ourselves,
    462 	 * we have a routing loop, perhaps as a result of an interface
    463 	 * going down recently.
    464 	 */
    465 	if (!(flags & RTF_DONE) && rt &&
    466 	     (sockaddr_cmp(src, rt->rt_gateway) != 0 || rt->rt_ifa != ifa))
    467 		error = EINVAL;
    468 	else if (ifa_ifwithaddr(gateway))
    469 		error = EHOSTUNREACH;
    470 	if (error)
    471 		goto done;
    472 	/*
    473 	 * Create a new entry if we just got back a wildcard entry
    474 	 * or the lookup failed.  This is necessary for hosts
    475 	 * which use routing redirects generated by smart gateways
    476 	 * to dynamically build the routing tables.
    477 	 */
    478 	if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
    479 		goto create;
    480 	/*
    481 	 * Don't listen to the redirect if it's
    482 	 * for a route to an interface.
    483 	 */
    484 	if (rt->rt_flags & RTF_GATEWAY) {
    485 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
    486 			/*
    487 			 * Changing from route to net => route to host.
    488 			 * Create new route, rather than smashing route to net.
    489 			 */
    490 		create:
    491 			if (rt != NULL)
    492 				rtfree(rt);
    493 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
    494 			memset(&info, 0, sizeof(info));
    495 			info.rti_info[RTAX_DST] = dst;
    496 			info.rti_info[RTAX_GATEWAY] = gateway;
    497 			info.rti_info[RTAX_NETMASK] = netmask;
    498 			info.rti_ifa = ifa;
    499 			info.rti_flags = flags;
    500 			rt = NULL;
    501 			error = rtrequest1(RTM_ADD, &info, &rt);
    502 			if (rt != NULL)
    503 				flags = rt->rt_flags;
    504 			stat = &rtstat.rts_dynamic;
    505 		} else {
    506 			/*
    507 			 * Smash the current notion of the gateway to
    508 			 * this destination.  Should check about netmask!!!
    509 			 */
    510 			rt->rt_flags |= RTF_MODIFIED;
    511 			flags |= RTF_MODIFIED;
    512 			stat = &rtstat.rts_newgateway;
    513 			rt_setgate(rt, gateway);
    514 		}
    515 	} else
    516 		error = EHOSTUNREACH;
    517 done:
    518 	if (rt) {
    519 		if (rtp != NULL && !error)
    520 			*rtp = rt;
    521 		else
    522 			rtfree(rt);
    523 	}
    524 out:
    525 	if (error)
    526 		rtstat.rts_badredirect++;
    527 	else if (stat != NULL)
    528 		(*stat)++;
    529 	memset(&info, 0, sizeof(info));
    530 	info.rti_info[RTAX_DST] = dst;
    531 	info.rti_info[RTAX_GATEWAY] = gateway;
    532 	info.rti_info[RTAX_NETMASK] = netmask;
    533 	info.rti_info[RTAX_AUTHOR] = src;
    534 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
    535 }
    536 
    537 /*
    538  * Delete a route and generate a message.
    539  * It doesn't free a passed rt.
    540  */
    541 static int
    542 rtdeletemsg(struct rtentry *rt)
    543 {
    544 	int error;
    545 	struct rt_addrinfo info;
    546 	struct rtentry *retrt;
    547 
    548 	/*
    549 	 * Request the new route so that the entry is not actually
    550 	 * deleted.  That will allow the information being reported to
    551 	 * be accurate (and consistent with route_output()).
    552 	 */
    553 	memset(&info, 0, sizeof(info));
    554 	info.rti_info[RTAX_DST] = rt_getkey(rt);
    555 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    556 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    557 	info.rti_flags = rt->rt_flags;
    558 	error = rtrequest1(RTM_DELETE, &info, &retrt);
    559 
    560 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
    561 
    562 	if (error == 0)
    563 		rtfree(retrt);
    564 	return error;
    565 }
    566 
    567 static int
    568 rtflushclone1(struct rtentry *rt, void *arg)
    569 {
    570 	struct rtentry *parent;
    571 
    572 	parent = (struct rtentry *)arg;
    573 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
    574 		rtdeletemsg(rt);
    575 	return 0;
    576 }
    577 
    578 static void
    579 rtflushclone(sa_family_t family, struct rtentry *parent)
    580 {
    581 
    582 #ifdef DIAGNOSTIC
    583 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
    584 		panic("rtflushclone: called with a non-cloning route");
    585 #endif
    586 	rt_walktree(family, rtflushclone1, (void *)parent);
    587 }
    588 
    589 struct ifaddr *
    590 ifa_ifwithroute(int flags, const struct sockaddr *dst,
    591 	const struct sockaddr *gateway)
    592 {
    593 	struct ifaddr *ifa;
    594 	if ((flags & RTF_GATEWAY) == 0) {
    595 		/*
    596 		 * If we are adding a route to an interface,
    597 		 * and the interface is a pt to pt link
    598 		 * we should search for the destination
    599 		 * as our clue to the interface.  Otherwise
    600 		 * we can use the local address.
    601 		 */
    602 		ifa = NULL;
    603 		if ((flags & RTF_HOST) && gateway->sa_family != AF_LINK)
    604 			ifa = ifa_ifwithdstaddr(dst);
    605 		if (ifa == NULL)
    606 			ifa = ifa_ifwithaddr(gateway);
    607 	} else {
    608 		/*
    609 		 * If we are adding a route to a remote net
    610 		 * or host, the gateway may still be on the
    611 		 * other end of a pt to pt link.
    612 		 */
    613 		ifa = ifa_ifwithdstaddr(gateway);
    614 	}
    615 	if (ifa == NULL)
    616 		ifa = ifa_ifwithnet(gateway);
    617 	if (ifa == NULL) {
    618 		struct rtentry *rt = rtalloc1(dst, 0);
    619 		if (rt == NULL)
    620 			return NULL;
    621 		ifa = rt->rt_ifa;
    622 		rtfree(rt);
    623 		if (ifa == NULL)
    624 			return NULL;
    625 	}
    626 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
    627 		struct ifaddr *oifa = ifa;
    628 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
    629 		if (ifa == NULL)
    630 			ifa = oifa;
    631 	}
    632 	return ifa;
    633 }
    634 
    635 /*
    636  * If it suceeds and ret_nrt isn't NULL, refcnt of ret_nrt is incremented.
    637  * The caller has to rtfree it by itself.
    638  */
    639 int
    640 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
    641 	const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
    642 {
    643 	struct rt_addrinfo info;
    644 
    645 	memset(&info, 0, sizeof(info));
    646 	info.rti_flags = flags;
    647 	info.rti_info[RTAX_DST] = dst;
    648 	info.rti_info[RTAX_GATEWAY] = gateway;
    649 	info.rti_info[RTAX_NETMASK] = netmask;
    650 	return rtrequest1(req, &info, ret_nrt);
    651 }
    652 
    653 /*
    654  * It's a utility function to add/remove a route to/from the routing table
    655  * and tell user processes the addition/removal on success.
    656  */
    657 int
    658 rtrequest_newmsg(const int req, const struct sockaddr *dst,
    659 	const struct sockaddr *gateway, const struct sockaddr *netmask,
    660 	const int flags)
    661 {
    662 	int error;
    663 	struct rtentry *ret_nrt = NULL;
    664 
    665 	KASSERT(req == RTM_ADD || req == RTM_DELETE);
    666 
    667 	error = rtrequest(req, dst, gateway, netmask, flags, &ret_nrt);
    668 	if (error != 0)
    669 		return error;
    670 
    671 	KASSERT(ret_nrt != NULL);
    672 
    673 	rt_newmsg(req, ret_nrt); /* tell user process */
    674 	rtfree(ret_nrt);
    675 
    676 	return 0;
    677 }
    678 
    679 int
    680 rt_getifa(struct rt_addrinfo *info)
    681 {
    682 	struct ifaddr *ifa;
    683 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
    684 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
    685 	const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
    686 	const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
    687 	int flags = info->rti_flags;
    688 
    689 	/*
    690 	 * ifp may be specified by sockaddr_dl when protocol address
    691 	 * is ambiguous
    692 	 */
    693 	if (info->rti_ifp == NULL && ifpaddr != NULL
    694 	    && ifpaddr->sa_family == AF_LINK &&
    695 	    (ifa = ifa_ifwithnet(ifpaddr)) != NULL)
    696 		info->rti_ifp = ifa->ifa_ifp;
    697 	if (info->rti_ifa == NULL && ifaaddr != NULL)
    698 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
    699 	if (info->rti_ifa == NULL) {
    700 		const struct sockaddr *sa;
    701 
    702 		sa = ifaaddr != NULL ? ifaaddr :
    703 		    (gateway != NULL ? gateway : dst);
    704 		if (sa != NULL && info->rti_ifp != NULL)
    705 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
    706 		else if (dst != NULL && gateway != NULL)
    707 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
    708 		else if (sa != NULL)
    709 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
    710 	}
    711 	if ((ifa = info->rti_ifa) == NULL)
    712 		return ENETUNREACH;
    713 	if (ifa->ifa_getifa != NULL) {
    714 		info->rti_ifa = ifa = (*ifa->ifa_getifa)(ifa, dst);
    715 		if (ifa == NULL)
    716 			return ENETUNREACH;
    717 	}
    718 	if (info->rti_ifp == NULL)
    719 		info->rti_ifp = ifa->ifa_ifp;
    720 	return 0;
    721 }
    722 
    723 /*
    724  * If it suceeds and ret_nrt isn't NULL, refcnt of ret_nrt is incremented.
    725  * The caller has to rtfree it by itself.
    726  */
    727 int
    728 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
    729 {
    730 	int s = splsoftnet();
    731 	int error = 0, rc;
    732 	struct rtentry *rt, *crt;
    733 	rtbl_t *rtbl;
    734 	struct ifaddr *ifa, *ifa2;
    735 	struct sockaddr_storage maskeddst;
    736 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
    737 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
    738 	const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
    739 	int flags = info->rti_flags;
    740 #define senderr(x) { error = x ; goto bad; }
    741 
    742 	if ((rtbl = rt_gettable(dst->sa_family)) == NULL)
    743 		senderr(ESRCH);
    744 	if (flags & RTF_HOST)
    745 		netmask = NULL;
    746 	switch (req) {
    747 	case RTM_DELETE:
    748 		if (netmask) {
    749 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    750 			    netmask);
    751 			dst = (struct sockaddr *)&maskeddst;
    752 		}
    753 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
    754 			senderr(ESRCH);
    755 		if ((rt->rt_flags & RTF_CLONING) != 0) {
    756 			/* clean up any cloned children */
    757 			rtflushclone(dst->sa_family, rt);
    758 		}
    759 		if ((rt = rt_deladdr(rtbl, dst, netmask)) == NULL)
    760 			senderr(ESRCH);
    761 		if (rt->rt_gwroute) {
    762 			rtfree(rt->rt_gwroute);
    763 			rt->rt_gwroute = NULL;
    764 		}
    765 		if (rt->rt_parent) {
    766 			rt->rt_parent->rt_refcnt--;
    767 			rt->rt_parent = NULL;
    768 		}
    769 		rt->rt_flags &= ~RTF_UP;
    770 		if ((ifa = rt->rt_ifa)) {
    771 			if (ifa->ifa_flags & IFA_ROUTE &&
    772 			    rt_ifa_connected(rt, ifa)) {
    773 				RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    774 				    "deleted IFA_ROUTE\n",
    775 				    (void *)rt->_rt_key, (void *)ifa);
    776 				ifa->ifa_flags &= ~IFA_ROUTE;
    777 			}
    778 			if (ifa->ifa_rtrequest)
    779 				ifa->ifa_rtrequest(RTM_DELETE, rt, info);
    780 		}
    781 		rttrash++;
    782 		if (ret_nrt) {
    783 			*ret_nrt = rt;
    784 			rt->rt_refcnt++;
    785 		} else if (rt->rt_refcnt <= 0) {
    786 			/* Adjust the refcount */
    787 			rt->rt_refcnt++;
    788 			rtfree(rt);
    789 		}
    790 		break;
    791 
    792 	case RTM_RESOLVE:
    793 		if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
    794 			senderr(EINVAL);
    795 		if ((rt->rt_flags & RTF_CLONING) == 0)
    796 			senderr(EINVAL);
    797 		ifa = rt->rt_ifa;
    798 		flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC);
    799 		flags |= RTF_CLONED;
    800 		gateway = rt->rt_gateway;
    801 		flags |= RTF_HOST;
    802 		goto makeroute;
    803 
    804 	case RTM_ADD:
    805 		if (info->rti_ifa == NULL && (error = rt_getifa(info)))
    806 			senderr(error);
    807 		ifa = info->rti_ifa;
    808 	makeroute:
    809 		/* Already at splsoftnet() so pool_get/pool_put are safe */
    810 		rt = pool_get(&rtentry_pool, PR_NOWAIT);
    811 		if (rt == NULL)
    812 			senderr(ENOBUFS);
    813 		memset(rt, 0, sizeof(*rt));
    814 		rt->rt_flags = RTF_UP | flags;
    815 		LIST_INIT(&rt->rt_timer);
    816 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    817 		if (rt_setkey(rt, dst, M_NOWAIT) == NULL ||
    818 		    rt_setgate(rt, gateway) != 0) {
    819 			pool_put(&rtentry_pool, rt);
    820 			senderr(ENOBUFS);
    821 		}
    822 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    823 		if (netmask) {
    824 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    825 			    netmask);
    826 			rt_setkey(rt, (struct sockaddr *)&maskeddst, M_NOWAIT);
    827 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    828 		} else {
    829 			rt_setkey(rt, dst, M_NOWAIT);
    830 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    831 		}
    832 		rt_set_ifa(rt, ifa);
    833 		if (info->rti_info[RTAX_TAG] != NULL)
    834 			rt_settag(rt, info->rti_info[RTAX_TAG]);
    835 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    836 		if (info->rti_info[RTAX_IFP] != NULL &&
    837 		    (ifa2 = ifa_ifwithnet(info->rti_info[RTAX_IFP])) != NULL &&
    838 		    ifa2->ifa_ifp != NULL)
    839 			rt->rt_ifp = ifa2->ifa_ifp;
    840 		else
    841 			rt->rt_ifp = ifa->ifa_ifp;
    842 		if (req == RTM_RESOLVE) {
    843 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
    844 			rt->rt_parent = *ret_nrt;
    845 			rt->rt_parent->rt_refcnt++;
    846 		}
    847 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    848 		rc = rt_addaddr(rtbl, rt, netmask);
    849 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    850 		if (rc != 0 && (crt = rtalloc1(rt_getkey(rt), 0)) != NULL) {
    851 			/* overwrite cloned route */
    852 			if ((crt->rt_flags & RTF_CLONED) != 0) {
    853 				rtdeletemsg(crt);
    854 				rc = rt_addaddr(rtbl, rt, netmask);
    855 			}
    856 			rtfree(crt);
    857 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    858 		}
    859 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    860 		if (rc != 0) {
    861 			ifafree(ifa);
    862 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
    863 				rtfree(rt->rt_parent);
    864 			if (rt->rt_gwroute)
    865 				rtfree(rt->rt_gwroute);
    866 			rt_destroy(rt);
    867 			pool_put(&rtentry_pool, rt);
    868 			senderr(rc);
    869 		}
    870 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    871 		if (ifa->ifa_rtrequest)
    872 			ifa->ifa_rtrequest(req, rt, info);
    873 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    874 		if (ret_nrt) {
    875 			*ret_nrt = rt;
    876 			rt->rt_refcnt++;
    877 		}
    878 		if ((rt->rt_flags & RTF_CLONING) != 0) {
    879 			/* clean up any cloned children */
    880 			rtflushclone(dst->sa_family, rt);
    881 		}
    882 		rtflushall(dst->sa_family);
    883 		break;
    884 	case RTM_GET:
    885 		if (netmask != NULL) {
    886 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    887 			    netmask);
    888 			dst = (struct sockaddr *)&maskeddst;
    889 		}
    890 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
    891 			senderr(ESRCH);
    892 		if (ret_nrt != NULL) {
    893 			*ret_nrt = rt;
    894 			rt->rt_refcnt++;
    895 		}
    896 		break;
    897 	}
    898 bad:
    899 	splx(s);
    900 	return error;
    901 }
    902 
    903 int
    904 rt_setgate(struct rtentry *rt, const struct sockaddr *gate)
    905 {
    906 	KASSERT(rt != rt->rt_gwroute);
    907 
    908 	KASSERT(rt->_rt_key != NULL);
    909 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    910 
    911 	if (rt->rt_gwroute) {
    912 		rtfree(rt->rt_gwroute);
    913 		rt->rt_gwroute = NULL;
    914 	}
    915 	KASSERT(rt->_rt_key != NULL);
    916 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    917 	if (rt->rt_gateway != NULL)
    918 		sockaddr_free(rt->rt_gateway);
    919 	KASSERT(rt->_rt_key != NULL);
    920 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    921 	if ((rt->rt_gateway = sockaddr_dup(gate, M_ZERO | M_NOWAIT)) == NULL)
    922 		return ENOMEM;
    923 	KASSERT(rt->_rt_key != NULL);
    924 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    925 
    926 	if (rt->rt_flags & RTF_GATEWAY) {
    927 		KASSERT(rt->_rt_key != NULL);
    928 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    929 		rt->rt_gwroute = rtalloc1(gate, 1);
    930 		/*
    931 		 * If we switched gateways, grab the MTU from the new
    932 		 * gateway route if the current MTU, if the current MTU is
    933 		 * greater than the MTU of gateway.
    934 		 * Note that, if the MTU of gateway is 0, we will reset the
    935 		 * MTU of the route to run PMTUD again from scratch. XXX
    936 		 */
    937 		KASSERT(rt->_rt_key != NULL);
    938 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    939 		if (rt->rt_gwroute
    940 		    && !(rt->rt_rmx.rmx_locks & RTV_MTU)
    941 		    && rt->rt_rmx.rmx_mtu
    942 		    && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
    943 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
    944 		}
    945 	}
    946 	KASSERT(rt->_rt_key != NULL);
    947 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    948 	return 0;
    949 }
    950 
    951 static void
    952 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
    953 	const struct sockaddr *netmask)
    954 {
    955 	const char *netmaskp = &netmask->sa_data[0],
    956 	           *srcp = &src->sa_data[0];
    957 	char *dstp = &dst->sa_data[0];
    958 	const char *maskend = (char *)dst + MIN(netmask->sa_len, src->sa_len);
    959 	const char *srcend = (char *)dst + src->sa_len;
    960 
    961 	dst->sa_len = src->sa_len;
    962 	dst->sa_family = src->sa_family;
    963 
    964 	while (dstp < maskend)
    965 		*dstp++ = *srcp++ & *netmaskp++;
    966 	if (dstp < srcend)
    967 		memset(dstp, 0, (size_t)(srcend - dstp));
    968 }
    969 
    970 /*
    971  * Inform the routing socket of a route change.
    972  */
    973 void
    974 rt_newmsg(int cmd, struct rtentry *rt)
    975 {
    976 	struct rt_addrinfo info;
    977 
    978 	memset((void *)&info, 0, sizeof(info));
    979 	info.rti_info[RTAX_DST] = rt_getkey(rt);
    980 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    981 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    982 	if (rt->rt_ifp) {
    983 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
    984 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
    985 	}
    986 
    987 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
    988 }
    989 
    990 /*
    991  * Set up or tear down a routing table entry, normally
    992  * for an interface.
    993  */
    994 int
    995 rtinit(struct ifaddr *ifa, int cmd, int flags)
    996 {
    997 	struct rtentry *rt;
    998 	struct sockaddr *dst, *odst;
    999 	struct sockaddr_storage maskeddst;
   1000 	struct rtentry *nrt = NULL;
   1001 	int error;
   1002 	struct rt_addrinfo info;
   1003 	struct sockaddr_dl *sdl;
   1004 	const struct sockaddr_dl *ifsdl;
   1005 
   1006 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
   1007 	if (cmd == RTM_DELETE) {
   1008 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
   1009 			/* Delete subnet route for this interface */
   1010 			odst = dst;
   1011 			dst = (struct sockaddr *)&maskeddst;
   1012 			rt_maskedcopy(odst, dst, ifa->ifa_netmask);
   1013 		}
   1014 		if ((rt = rtalloc1(dst, 0)) != NULL) {
   1015 			if (rt->rt_ifa != ifa) {
   1016 				rtfree(rt);
   1017 				return (flags & RTF_HOST) ? EHOSTUNREACH
   1018 							: ENETUNREACH;
   1019 			}
   1020 			rtfree(rt);
   1021 		}
   1022 	}
   1023 	memset(&info, 0, sizeof(info));
   1024 	info.rti_ifa = ifa;
   1025 	info.rti_flags = flags | ifa->ifa_flags;
   1026 	info.rti_info[RTAX_DST] = dst;
   1027 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
   1028 	/*
   1029 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
   1030 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
   1031 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
   1032 	 * change it to meet bsdi4 behavior.
   1033 	 */
   1034 	if (cmd != RTM_LLINFO_UPD)
   1035 		info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
   1036 	error = rtrequest1((cmd == RTM_LLINFO_UPD) ? RTM_GET : cmd, &info,
   1037 	    &nrt);
   1038 	if (error != 0 || (rt = nrt) == NULL)
   1039 		return error;
   1040 
   1041 	switch (cmd) {
   1042 	case RTM_DELETE:
   1043 		rt_newmsg(cmd, rt);
   1044 		rtfree(rt);
   1045 		break;
   1046 	case RTM_LLINFO_UPD:
   1047 		RT_DPRINTF("%s: updating%s\n", __func__,
   1048 		    ((rt->rt_flags & RTF_LLINFO) == 0) ? " (no llinfo)" : "");
   1049 
   1050 		ifsdl = ifa->ifa_ifp->if_sadl;
   1051 
   1052 		if ((rt->rt_flags & RTF_LLINFO) != 0 &&
   1053 		    (sdl = satosdl(rt->rt_gateway)) != NULL &&
   1054 		    sdl->sdl_family == AF_LINK &&
   1055 		    sockaddr_dl_setaddr(sdl, sdl->sdl_len, CLLADDR(ifsdl),
   1056 		                        ifa->ifa_ifp->if_addrlen) == NULL) {
   1057 			error = EINVAL;
   1058 			break;
   1059 		}
   1060 
   1061 		if (cmd == RTM_LLINFO_UPD && ifa->ifa_rtrequest != NULL)
   1062 			ifa->ifa_rtrequest(RTM_LLINFO_UPD, rt, &info);
   1063 		rt_newmsg(RTM_CHANGE, rt);
   1064 		rtfree(rt);
   1065 		break;
   1066 	case RTM_ADD:
   1067 		if (rt->rt_ifa != ifa) {
   1068 			printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
   1069 				rt->rt_ifa);
   1070 			if (rt->rt_ifa->ifa_rtrequest != NULL) {
   1071 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
   1072 				    &info);
   1073 			}
   1074 			rt_replace_ifa(rt, ifa);
   1075 			rt->rt_ifp = ifa->ifa_ifp;
   1076 			if (ifa->ifa_rtrequest != NULL)
   1077 				ifa->ifa_rtrequest(RTM_ADD, rt, &info);
   1078 		}
   1079 		rt_newmsg(cmd, rt);
   1080 		rtfree(rt);
   1081 		break;
   1082 	}
   1083 	return error;
   1084 }
   1085 
   1086 static const struct in_addr inmask32 = {.s_addr = INADDR_BROADCAST};
   1087 
   1088 /* Subroutine for rt_ifa_addlocal() and rt_ifa_remlocal() */
   1089 static int
   1090 rt_ifa_localrequest(int cmd, struct ifaddr *ifa)
   1091 {
   1092 	struct sockaddr *all1_sa;
   1093 	struct sockaddr_in all1_sin;
   1094 #ifdef INET6
   1095 	struct sockaddr_in6 all1_sin6;
   1096 #endif
   1097 	struct rtentry *nrt = NULL;
   1098 	int flags, e;
   1099 
   1100 	switch(ifa->ifa_addr->sa_family) {
   1101 	case AF_INET:
   1102 		sockaddr_in_init(&all1_sin, &inmask32, 0);
   1103 		all1_sa = (struct sockaddr *)&all1_sin;
   1104 		break;
   1105 #ifdef INET6
   1106 	case AF_INET6:
   1107 		sockaddr_in6_init(&all1_sin6, &in6mask128, 0, 0, 0);
   1108 		all1_sa = (struct sockaddr *)&all1_sin6;
   1109 		break;
   1110 #endif
   1111 	default:
   1112 		return 0;
   1113 	}
   1114 
   1115 	flags = RTF_UP | RTF_HOST | RTF_LOCAL;
   1116 	if (!(ifa->ifa_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)))
   1117 		flags |= RTF_LLINFO;
   1118 	e = rtrequest(cmd, ifa->ifa_addr, ifa->ifa_addr, all1_sa, flags, &nrt);
   1119 
   1120 	/* Make sure rt_ifa be equal to IFA, the second argument of the
   1121 	 * function. */
   1122 	if (cmd == RTM_ADD && nrt && ifa != nrt->rt_ifa)
   1123 		rt_replace_ifa(nrt, ifa);
   1124 
   1125 	rt_newaddrmsg(cmd, ifa, e, nrt);
   1126 	if (nrt != NULL)
   1127 		rtfree(nrt);
   1128 
   1129 	return e;
   1130 }
   1131 
   1132 /*
   1133  * Create a local route entry for the address.
   1134  * Announce the addition of the address and the route to the routing socket.
   1135  */
   1136 int
   1137 rt_ifa_addlocal(struct ifaddr *ifa)
   1138 {
   1139 	struct rtentry *rt;
   1140 	int e;
   1141 
   1142 	/* If there is no loopback entry, allocate one. */
   1143 	rt = rtalloc1(ifa->ifa_addr, 0);
   1144 	if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
   1145 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
   1146 		e = rt_ifa_localrequest(RTM_ADD, ifa);
   1147 	else {
   1148 		e = 0;
   1149 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1150 	}
   1151 	if (rt != NULL)
   1152 		rtfree(rt);
   1153 	return e;
   1154 }
   1155 
   1156 /*
   1157  * Remove the local route entry for the address.
   1158  * Announce the removal of the address and the route to the routing socket.
   1159  */
   1160 int
   1161 rt_ifa_remlocal(struct ifaddr *ifa, struct ifaddr *alt_ifa)
   1162 {
   1163 	struct rtentry *rt;
   1164 	int e = 0;
   1165 
   1166 	rt = rtalloc1(ifa->ifa_addr, 0);
   1167 
   1168 	/*
   1169 	 * Before deleting, check if a corresponding loopbacked
   1170 	 * host route surely exists.  With this check, we can avoid
   1171 	 * deleting an interface direct route whose destination is
   1172 	 * the same as the address being removed.  This can happen
   1173 	 * when removing a subnet-router anycast address on an
   1174 	 * interface attached to a shared medium.
   1175 	 */
   1176 	if (rt != NULL &&
   1177 	    (rt->rt_flags & RTF_HOST) &&
   1178 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK))
   1179 	{
   1180 		/* If we cannot replace the route's ifaddr with the equivalent
   1181 		 * ifaddr of another interface, I believe it is safest to
   1182 		 * delete the route.
   1183 		 */
   1184 		if (alt_ifa == NULL)
   1185 			e = rt_ifa_localrequest(RTM_DELETE, ifa);
   1186 		else {
   1187 			rt_replace_ifa(rt, alt_ifa);
   1188 			rt_newmsg(RTM_CHANGE, rt);
   1189 		}
   1190 	} else
   1191 		rt_newaddrmsg(RTM_DELADDR, ifa, 0, NULL);
   1192 	if (rt != NULL)
   1193 		rtfree(rt);
   1194 	return e;
   1195 }
   1196 
   1197 /*
   1198  * Route timer routines.  These routes allow functions to be called
   1199  * for various routes at any time.  This is useful in supporting
   1200  * path MTU discovery and redirect route deletion.
   1201  *
   1202  * This is similar to some BSDI internal functions, but it provides
   1203  * for multiple queues for efficiency's sake...
   1204  */
   1205 
   1206 LIST_HEAD(, rttimer_queue) rttimer_queue_head;
   1207 static int rt_init_done = 0;
   1208 
   1209 #define RTTIMER_CALLOUT(r)	do {					\
   1210 		if (r->rtt_func != NULL) {				\
   1211 			(*r->rtt_func)(r->rtt_rt, r);			\
   1212 		} else {						\
   1213 			rtrequest((int) RTM_DELETE,			\
   1214 				  rt_getkey(r->rtt_rt),			\
   1215 				  0, 0, 0, 0);				\
   1216 		}							\
   1217 	} while (/*CONSTCOND*/0)
   1218 
   1219 /*
   1220  * Some subtle order problems with domain initialization mean that
   1221  * we cannot count on this being run from rt_init before various
   1222  * protocol initializations are done.  Therefore, we make sure
   1223  * that this is run when the first queue is added...
   1224  */
   1225 
   1226 void
   1227 rt_timer_init(void)
   1228 {
   1229 	assert(rt_init_done == 0);
   1230 
   1231 	LIST_INIT(&rttimer_queue_head);
   1232 	callout_init(&rt_timer_ch, 0);
   1233 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
   1234 	rt_init_done = 1;
   1235 }
   1236 
   1237 struct rttimer_queue *
   1238 rt_timer_queue_create(u_int timeout)
   1239 {
   1240 	struct rttimer_queue *rtq;
   1241 
   1242 	if (rt_init_done == 0)
   1243 		rt_timer_init();
   1244 
   1245 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
   1246 	if (rtq == NULL)
   1247 		return NULL;
   1248 	memset(rtq, 0, sizeof(*rtq));
   1249 
   1250 	rtq->rtq_timeout = timeout;
   1251 	TAILQ_INIT(&rtq->rtq_head);
   1252 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
   1253 
   1254 	return rtq;
   1255 }
   1256 
   1257 void
   1258 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
   1259 {
   1260 
   1261 	rtq->rtq_timeout = timeout;
   1262 }
   1263 
   1264 void
   1265 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy)
   1266 {
   1267 	struct rttimer *r;
   1268 
   1269 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
   1270 		LIST_REMOVE(r, rtt_link);
   1271 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
   1272 		if (destroy)
   1273 			RTTIMER_CALLOUT(r);
   1274 		/* we are already at splsoftnet */
   1275 		pool_put(&rttimer_pool, r);
   1276 		if (rtq->rtq_count > 0)
   1277 			rtq->rtq_count--;
   1278 		else
   1279 			printf("rt_timer_queue_remove_all: "
   1280 			    "rtq_count reached 0\n");
   1281 	}
   1282 }
   1283 
   1284 void
   1285 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
   1286 {
   1287 
   1288 	rt_timer_queue_remove_all(rtq, destroy);
   1289 
   1290 	LIST_REMOVE(rtq, rtq_link);
   1291 
   1292 	/*
   1293 	 * Caller is responsible for freeing the rttimer_queue structure.
   1294 	 */
   1295 }
   1296 
   1297 unsigned long
   1298 rt_timer_count(struct rttimer_queue *rtq)
   1299 {
   1300 	return rtq->rtq_count;
   1301 }
   1302 
   1303 void
   1304 rt_timer_remove_all(struct rtentry *rt, int destroy)
   1305 {
   1306 	struct rttimer *r;
   1307 
   1308 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
   1309 		LIST_REMOVE(r, rtt_link);
   1310 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
   1311 		if (destroy)
   1312 			RTTIMER_CALLOUT(r);
   1313 		if (r->rtt_queue->rtq_count > 0)
   1314 			r->rtt_queue->rtq_count--;
   1315 		else
   1316 			printf("rt_timer_remove_all: rtq_count reached 0\n");
   1317 		/* we are already at splsoftnet */
   1318 		pool_put(&rttimer_pool, r);
   1319 	}
   1320 }
   1321 
   1322 int
   1323 rt_timer_add(struct rtentry *rt,
   1324 	void (*func)(struct rtentry *, struct rttimer *),
   1325 	struct rttimer_queue *queue)
   1326 {
   1327 	struct rttimer *r;
   1328 	int s;
   1329 
   1330 	/*
   1331 	 * If there's already a timer with this action, destroy it before
   1332 	 * we add a new one.
   1333 	 */
   1334 	LIST_FOREACH(r, &rt->rt_timer, rtt_link) {
   1335 		if (r->rtt_func == func)
   1336 			break;
   1337 	}
   1338 	if (r != NULL) {
   1339 		LIST_REMOVE(r, rtt_link);
   1340 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
   1341 		if (r->rtt_queue->rtq_count > 0)
   1342 			r->rtt_queue->rtq_count--;
   1343 		else
   1344 			printf("rt_timer_add: rtq_count reached 0\n");
   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 	r->rtt_rt = rt;
   1356 	r->rtt_time = time_uptime;
   1357 	r->rtt_func = func;
   1358 	r->rtt_queue = queue;
   1359 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
   1360 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
   1361 	r->rtt_queue->rtq_count++;
   1362 
   1363 	return 0;
   1364 }
   1365 
   1366 /* ARGSUSED */
   1367 void
   1368 rt_timer_timer(void *arg)
   1369 {
   1370 	struct rttimer_queue *rtq;
   1371 	struct rttimer *r;
   1372 	int s;
   1373 
   1374 	s = splsoftnet();
   1375 	LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) {
   1376 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
   1377 		    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
   1378 			LIST_REMOVE(r, rtt_link);
   1379 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
   1380 			RTTIMER_CALLOUT(r);
   1381 			pool_put(&rttimer_pool, r);
   1382 			if (rtq->rtq_count > 0)
   1383 				rtq->rtq_count--;
   1384 			else
   1385 				printf("rt_timer_timer: rtq_count reached 0\n");
   1386 		}
   1387 	}
   1388 	splx(s);
   1389 
   1390 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
   1391 }
   1392 
   1393 static struct rtentry *
   1394 _rtcache_init(struct route *ro, int flag)
   1395 {
   1396 	rtcache_invariants(ro);
   1397 	KASSERT(ro->_ro_rt == NULL);
   1398 
   1399 	if (rtcache_getdst(ro) == NULL)
   1400 		return NULL;
   1401 	ro->ro_invalid = false;
   1402 	if ((ro->_ro_rt = rtalloc1(rtcache_getdst(ro), flag)) != NULL)
   1403 		rtcache(ro);
   1404 
   1405 	rtcache_invariants(ro);
   1406 	return ro->_ro_rt;
   1407 }
   1408 
   1409 struct rtentry *
   1410 rtcache_init(struct route *ro)
   1411 {
   1412 	return _rtcache_init(ro, 1);
   1413 }
   1414 
   1415 struct rtentry *
   1416 rtcache_init_noclone(struct route *ro)
   1417 {
   1418 	return _rtcache_init(ro, 0);
   1419 }
   1420 
   1421 struct rtentry *
   1422 rtcache_update(struct route *ro, int clone)
   1423 {
   1424 	rtcache_clear(ro);
   1425 	return _rtcache_init(ro, clone);
   1426 }
   1427 
   1428 void
   1429 rtcache_copy(struct route *new_ro, const struct route *old_ro)
   1430 {
   1431 	struct rtentry *rt;
   1432 
   1433 	KASSERT(new_ro != old_ro);
   1434 	rtcache_invariants(new_ro);
   1435 	rtcache_invariants(old_ro);
   1436 
   1437 	if ((rt = rtcache_validate(old_ro)) != NULL)
   1438 		rt->rt_refcnt++;
   1439 
   1440 	if (rtcache_getdst(old_ro) == NULL ||
   1441 	    rtcache_setdst(new_ro, rtcache_getdst(old_ro)) != 0)
   1442 		return;
   1443 
   1444 	new_ro->ro_invalid = false;
   1445 	if ((new_ro->_ro_rt = rt) != NULL)
   1446 		rtcache(new_ro);
   1447 	rtcache_invariants(new_ro);
   1448 }
   1449 
   1450 static struct dom_rtlist invalid_routes = LIST_HEAD_INITIALIZER(dom_rtlist);
   1451 
   1452 static void
   1453 rtcache_invalidate(struct dom_rtlist *rtlist)
   1454 {
   1455 	struct route *ro;
   1456 
   1457 	while ((ro = LIST_FIRST(rtlist)) != NULL) {
   1458 		rtcache_invariants(ro);
   1459 		KASSERT(ro->_ro_rt != NULL);
   1460 		ro->ro_invalid = true;
   1461 		LIST_REMOVE(ro, ro_rtcache_next);
   1462 		LIST_INSERT_HEAD(&invalid_routes, ro, ro_rtcache_next);
   1463 		rtcache_invariants(ro);
   1464 	}
   1465 }
   1466 
   1467 static void
   1468 rtcache_clear(struct route *ro)
   1469 {
   1470 	rtcache_invariants(ro);
   1471 	if (ro->_ro_rt == NULL)
   1472 		return;
   1473 
   1474 	LIST_REMOVE(ro, ro_rtcache_next);
   1475 
   1476 	rtfree(ro->_ro_rt);
   1477 	ro->_ro_rt = NULL;
   1478 	ro->ro_invalid = false;
   1479 	rtcache_invariants(ro);
   1480 }
   1481 
   1482 struct rtentry *
   1483 rtcache_lookup2(struct route *ro, const struct sockaddr *dst, int clone,
   1484     int *hitp)
   1485 {
   1486 	const struct sockaddr *odst;
   1487 	struct rtentry *rt = NULL;
   1488 
   1489 	odst = rtcache_getdst(ro);
   1490 	if (odst == NULL)
   1491 		goto miss;
   1492 
   1493 	if (sockaddr_cmp(odst, dst) != 0) {
   1494 		rtcache_free(ro);
   1495 		goto miss;
   1496 	}
   1497 
   1498 	rt = rtcache_validate(ro);
   1499 	if (rt == NULL) {
   1500 		rtcache_clear(ro);
   1501 		goto miss;
   1502 	}
   1503 
   1504 	*hitp = 1;
   1505 	rtcache_invariants(ro);
   1506 
   1507 	return rt;
   1508 miss:
   1509 	*hitp = 0;
   1510 	if (rtcache_setdst(ro, dst) == 0)
   1511 		rt = _rtcache_init(ro, clone);
   1512 
   1513 	rtcache_invariants(ro);
   1514 
   1515 	return rt;
   1516 }
   1517 
   1518 void
   1519 rtcache_free(struct route *ro)
   1520 {
   1521 	rtcache_clear(ro);
   1522 	if (ro->ro_sa != NULL) {
   1523 		sockaddr_free(ro->ro_sa);
   1524 		ro->ro_sa = NULL;
   1525 	}
   1526 	rtcache_invariants(ro);
   1527 }
   1528 
   1529 int
   1530 rtcache_setdst(struct route *ro, const struct sockaddr *sa)
   1531 {
   1532 	KASSERT(sa != NULL);
   1533 
   1534 	rtcache_invariants(ro);
   1535 	if (ro->ro_sa != NULL) {
   1536 		if (ro->ro_sa->sa_family == sa->sa_family) {
   1537 			rtcache_clear(ro);
   1538 			sockaddr_copy(ro->ro_sa, ro->ro_sa->sa_len, sa);
   1539 			rtcache_invariants(ro);
   1540 			return 0;
   1541 		}
   1542 		/* free ro_sa, wrong family */
   1543 		rtcache_free(ro);
   1544 	}
   1545 
   1546 	KASSERT(ro->_ro_rt == NULL);
   1547 
   1548 	if ((ro->ro_sa = sockaddr_dup(sa, M_ZERO | M_NOWAIT)) == NULL) {
   1549 		rtcache_invariants(ro);
   1550 		return ENOMEM;
   1551 	}
   1552 	rtcache_invariants(ro);
   1553 	return 0;
   1554 }
   1555 
   1556 const struct sockaddr *
   1557 rt_settag(struct rtentry *rt, const struct sockaddr *tag)
   1558 {
   1559 	if (rt->rt_tag != tag) {
   1560 		if (rt->rt_tag != NULL)
   1561 			sockaddr_free(rt->rt_tag);
   1562 		rt->rt_tag = sockaddr_dup(tag, M_ZERO | M_NOWAIT);
   1563 	}
   1564 	return rt->rt_tag;
   1565 }
   1566 
   1567 struct sockaddr *
   1568 rt_gettag(struct rtentry *rt)
   1569 {
   1570 	return rt->rt_tag;
   1571 }
   1572