Home | History | Annotate | Line # | Download | only in net
route.c revision 1.149
      1 /*	$NetBSD: route.c,v 1.149 2015/08/24 22:21:26 pooka Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1998, 2008 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Kevin M. Lahey of the Numerical Aerospace Simulation Facility,
      9  * NASA Ames Research Center.
     10  *
     11  * Redistribution and use in source and binary forms, with or without
     12  * modification, are permitted provided that the following conditions
     13  * are met:
     14  * 1. Redistributions of source code must retain the above copyright
     15  *    notice, this list of conditions and the following disclaimer.
     16  * 2. Redistributions in binary form must reproduce the above copyright
     17  *    notice, this list of conditions and the following disclaimer in the
     18  *    documentation and/or other materials provided with the distribution.
     19  *
     20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     30  * POSSIBILITY OF SUCH DAMAGE.
     31  */
     32 
     33 /*
     34  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
     35  * All rights reserved.
     36  *
     37  * Redistribution and use in source and binary forms, with or without
     38  * modification, are permitted provided that the following conditions
     39  * are met:
     40  * 1. Redistributions of source code must retain the above copyright
     41  *    notice, this list of conditions and the following disclaimer.
     42  * 2. Redistributions in binary form must reproduce the above copyright
     43  *    notice, this list of conditions and the following disclaimer in the
     44  *    documentation and/or other materials provided with the distribution.
     45  * 3. Neither the name of the project nor the names of its contributors
     46  *    may be used to endorse or promote products derived from this software
     47  *    without specific prior written permission.
     48  *
     49  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     59  * SUCH DAMAGE.
     60  */
     61 
     62 /*
     63  * Copyright (c) 1980, 1986, 1991, 1993
     64  *	The Regents of the University of California.  All rights reserved.
     65  *
     66  * Redistribution and use in source and binary forms, with or without
     67  * modification, are permitted provided that the following conditions
     68  * are met:
     69  * 1. Redistributions of source code must retain the above copyright
     70  *    notice, this list of conditions and the following disclaimer.
     71  * 2. Redistributions in binary form must reproduce the above copyright
     72  *    notice, this list of conditions and the following disclaimer in the
     73  *    documentation and/or other materials provided with the distribution.
     74  * 3. Neither the name of the University nor the names of its contributors
     75  *    may be used to endorse or promote products derived from this software
     76  *    without specific prior written permission.
     77  *
     78  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     79  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     80  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     81  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     82  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     83  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     84  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     85  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     86  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     87  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     88  * SUCH DAMAGE.
     89  *
     90  *	@(#)route.c	8.3 (Berkeley) 1/9/95
     91  */
     92 
     93 #ifdef _KERNEL_OPT
     94 #include "opt_inet.h"
     95 #include "opt_route.h"
     96 #endif
     97 
     98 #include <sys/cdefs.h>
     99 __KERNEL_RCSID(0, "$NetBSD: route.c,v 1.149 2015/08/24 22:21:26 pooka Exp $");
    100 
    101 #include <sys/param.h>
    102 #ifdef RTFLUSH_DEBUG
    103 #include <sys/sysctl.h>
    104 #endif
    105 #include <sys/systm.h>
    106 #include <sys/callout.h>
    107 #include <sys/proc.h>
    108 #include <sys/mbuf.h>
    109 #include <sys/socket.h>
    110 #include <sys/socketvar.h>
    111 #include <sys/domain.h>
    112 #include <sys/protosw.h>
    113 #include <sys/kernel.h>
    114 #include <sys/ioctl.h>
    115 #include <sys/pool.h>
    116 #include <sys/kauth.h>
    117 
    118 #include <net/if.h>
    119 #include <net/if_dl.h>
    120 #include <net/route.h>
    121 
    122 #include <netinet/in.h>
    123 #include <netinet/in_var.h>
    124 
    125 #ifdef RTFLUSH_DEBUG
    126 #define	rtcache_debug() __predict_false(_rtcache_debug)
    127 #else /* RTFLUSH_DEBUG */
    128 #define	rtcache_debug() 0
    129 #endif /* RTFLUSH_DEBUG */
    130 
    131 struct	rtstat	rtstat;
    132 
    133 int	rttrash;		/* routes not in table but not freed */
    134 
    135 struct pool rtentry_pool;
    136 struct pool rttimer_pool;
    137 
    138 struct callout rt_timer_ch; /* callout for rt_timer_timer() */
    139 
    140 #ifdef RTFLUSH_DEBUG
    141 static int _rtcache_debug = 0;
    142 #endif /* RTFLUSH_DEBUG */
    143 
    144 static kauth_listener_t route_listener;
    145 
    146 static int rtdeletemsg(struct rtentry *);
    147 static int rtflushclone1(struct rtentry *, void *);
    148 static void rtflushclone(sa_family_t family, struct rtentry *);
    149 static void rtflushall(int);
    150 
    151 static void rt_maskedcopy(const struct sockaddr *,
    152     struct sockaddr *, const struct sockaddr *);
    153 
    154 static void rtcache_clear(struct route *);
    155 static void rtcache_invalidate(struct dom_rtlist *);
    156 
    157 #ifdef RTFLUSH_DEBUG
    158 static void sysctl_net_rtcache_setup(struct sysctllog **);
    159 static void
    160 sysctl_net_rtcache_setup(struct sysctllog **clog)
    161 {
    162 	const struct sysctlnode *rnode;
    163 
    164 	if (sysctl_createv(clog, 0, NULL, &rnode, CTLFLAG_PERMANENT,
    165 	    CTLTYPE_NODE,
    166 	    "rtcache", SYSCTL_DESCR("Route cache related settings"),
    167 	    NULL, 0, NULL, 0, CTL_NET, CTL_CREATE, CTL_EOL) != 0)
    168 		return;
    169 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    170 	    CTLFLAG_PERMANENT|CTLFLAG_READWRITE, CTLTYPE_INT,
    171 	    "debug", SYSCTL_DESCR("Debug route caches"),
    172 	    NULL, 0, &_rtcache_debug, 0, CTL_CREATE, CTL_EOL) != 0)
    173 		return;
    174 }
    175 #endif /* RTFLUSH_DEBUG */
    176 
    177 static inline void
    178 rt_destroy(struct rtentry *rt)
    179 {
    180 	if (rt->_rt_key != NULL)
    181 		sockaddr_free(rt->_rt_key);
    182 	if (rt->rt_gateway != NULL)
    183 		sockaddr_free(rt->rt_gateway);
    184 	if (rt_gettag(rt) != NULL)
    185 		sockaddr_free(rt_gettag(rt));
    186 	rt->_rt_key = rt->rt_gateway = rt->rt_tag = NULL;
    187 }
    188 
    189 static inline const struct sockaddr *
    190 rt_setkey(struct rtentry *rt, const struct sockaddr *key, int flags)
    191 {
    192 	if (rt->_rt_key == key)
    193 		goto out;
    194 
    195 	if (rt->_rt_key != NULL)
    196 		sockaddr_free(rt->_rt_key);
    197 	rt->_rt_key = sockaddr_dup(key, flags);
    198 out:
    199 	rt->rt_nodes->rn_key = (const char *)rt->_rt_key;
    200 	return rt->_rt_key;
    201 }
    202 
    203 struct ifaddr *
    204 rt_get_ifa(struct rtentry *rt)
    205 {
    206 	struct ifaddr *ifa;
    207 
    208 	if ((ifa = rt->rt_ifa) == NULL)
    209 		return ifa;
    210 	else if (ifa->ifa_getifa == NULL)
    211 		return ifa;
    212 #if 0
    213 	else if (ifa->ifa_seqno != NULL && *ifa->ifa_seqno == rt->rt_ifa_seqno)
    214 		return ifa;
    215 #endif
    216 	else {
    217 		ifa = (*ifa->ifa_getifa)(ifa, rt_getkey(rt));
    218 		if (ifa == NULL)
    219 			return NULL;
    220 		rt_replace_ifa(rt, ifa);
    221 		return ifa;
    222 	}
    223 }
    224 
    225 static void
    226 rt_set_ifa1(struct rtentry *rt, struct ifaddr *ifa)
    227 {
    228 	rt->rt_ifa = ifa;
    229 	if (ifa->ifa_seqno != NULL)
    230 		rt->rt_ifa_seqno = *ifa->ifa_seqno;
    231 }
    232 
    233 /*
    234  * Is this route the connected route for the ifa?
    235  */
    236 static int
    237 rt_ifa_connected(const struct rtentry *rt, const struct ifaddr *ifa)
    238 {
    239 	const struct sockaddr *key, *dst, *odst;
    240 	struct sockaddr_storage maskeddst;
    241 
    242 	key = rt_getkey(rt);
    243 	dst = rt->rt_flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
    244 	if (dst == NULL ||
    245 	    dst->sa_family != key->sa_family ||
    246 	    dst->sa_len != key->sa_len)
    247 		return 0;
    248 	if ((rt->rt_flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
    249 		odst = dst;
    250 		dst = (struct sockaddr *)&maskeddst;
    251 		rt_maskedcopy(odst, (struct sockaddr *)&maskeddst,
    252 		    ifa->ifa_netmask);
    253 	}
    254 	return (memcmp(dst, key, dst->sa_len) == 0);
    255 }
    256 
    257 void
    258 rt_replace_ifa(struct rtentry *rt, struct ifaddr *ifa)
    259 {
    260 	if (rt->rt_ifa &&
    261 	    rt->rt_ifa != ifa &&
    262 	    rt->rt_ifa->ifa_flags & IFA_ROUTE &&
    263 	    rt_ifa_connected(rt, rt->rt_ifa))
    264 	{
    265 		RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    266 		    "replace deleted IFA_ROUTE\n",
    267 		    (void *)rt->_rt_key, (void *)rt->rt_ifa);
    268 		rt->rt_ifa->ifa_flags &= ~IFA_ROUTE;
    269 		if (rt_ifa_connected(rt, ifa)) {
    270 			RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    271 			    "replace added IFA_ROUTE\n",
    272 			    (void *)rt->_rt_key, (void *)ifa);
    273 			ifa->ifa_flags |= IFA_ROUTE;
    274 		}
    275 	}
    276 
    277 	ifaref(ifa);
    278 	ifafree(rt->rt_ifa);
    279 	rt_set_ifa1(rt, ifa);
    280 }
    281 
    282 static void
    283 rt_set_ifa(struct rtentry *rt, struct ifaddr *ifa)
    284 {
    285 	ifaref(ifa);
    286 	rt_set_ifa1(rt, ifa);
    287 }
    288 
    289 static int
    290 route_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
    291     void *arg0, void *arg1, void *arg2, void *arg3)
    292 {
    293 	struct rt_msghdr *rtm;
    294 	int result;
    295 
    296 	result = KAUTH_RESULT_DEFER;
    297 	rtm = arg1;
    298 
    299 	if (action != KAUTH_NETWORK_ROUTE)
    300 		return result;
    301 
    302 	if (rtm->rtm_type == RTM_GET)
    303 		result = KAUTH_RESULT_ALLOW;
    304 
    305 	return result;
    306 }
    307 
    308 void
    309 rt_init(void)
    310 {
    311 
    312 #ifdef RTFLUSH_DEBUG
    313 	sysctl_net_rtcache_setup(NULL);
    314 #endif
    315 
    316 	pool_init(&rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl",
    317 	    NULL, IPL_SOFTNET);
    318 	pool_init(&rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl",
    319 	    NULL, IPL_SOFTNET);
    320 
    321 	rn_init();	/* initialize all zeroes, all ones, mask table */
    322 	rtbl_init();
    323 
    324 	route_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
    325 	    route_listener_cb, NULL);
    326 }
    327 
    328 static void
    329 rtflushall(int family)
    330 {
    331 	struct domain *dom;
    332 
    333 	if (rtcache_debug())
    334 		printf("%s: enter\n", __func__);
    335 
    336 	if ((dom = pffinddomain(family)) == NULL)
    337 		return;
    338 
    339 	rtcache_invalidate(&dom->dom_rtcache);
    340 }
    341 
    342 static void
    343 rtcache(struct route *ro)
    344 {
    345 	struct domain *dom;
    346 
    347 	rtcache_invariants(ro);
    348 	KASSERT(ro->_ro_rt != NULL);
    349 	KASSERT(ro->ro_invalid == false);
    350 	KASSERT(rtcache_getdst(ro) != NULL);
    351 
    352 	if ((dom = pffinddomain(rtcache_getdst(ro)->sa_family)) == NULL)
    353 		return;
    354 
    355 	LIST_INSERT_HEAD(&dom->dom_rtcache, ro, ro_rtcache_next);
    356 	rtcache_invariants(ro);
    357 }
    358 
    359 /*
    360  * Packet routing routines. If success, refcnt of a returned rtentry
    361  * will be incremented. The caller has to rtfree it by itself.
    362  */
    363 struct rtentry *
    364 rtalloc1(const struct sockaddr *dst, int report)
    365 {
    366 	rtbl_t *rtbl = rt_gettable(dst->sa_family);
    367 	struct rtentry *rt;
    368 	struct rtentry *newrt = NULL;
    369 	struct rt_addrinfo info;
    370 	int  s = splsoftnet(), err = 0, msgtype = RTM_MISS;
    371 
    372 	if (rtbl != NULL && (rt = rt_matchaddr(rtbl, dst)) != NULL) {
    373 		newrt = rt;
    374 		if (report && (rt->rt_flags & RTF_CLONING)) {
    375 			err = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
    376 			    &newrt);
    377 			if (err) {
    378 				newrt = rt;
    379 				rt->rt_refcnt++;
    380 				goto miss;
    381 			}
    382 			KASSERT(newrt != NULL);
    383 			rt = newrt;
    384 			if (rt->rt_flags & RTF_XRESOLVE) {
    385 				msgtype = RTM_RESOLVE;
    386 				goto miss;
    387 			}
    388 			/* Inform listeners of the new route */
    389 			memset(&info, 0, sizeof(info));
    390 			info.rti_info[RTAX_DST] = rt_getkey(rt);
    391 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    392 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    393 			if (rt->rt_ifp != NULL) {
    394 				info.rti_info[RTAX_IFP] =
    395 				    rt->rt_ifp->if_dl->ifa_addr;
    396 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
    397 			}
    398 			rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0);
    399 		} else
    400 			rt->rt_refcnt++;
    401 	} else {
    402 		rtstat.rts_unreach++;
    403 	miss:	if (report) {
    404 			memset((void *)&info, 0, sizeof(info));
    405 			info.rti_info[RTAX_DST] = dst;
    406 			rt_missmsg(msgtype, &info, 0, err);
    407 		}
    408 	}
    409 	splx(s);
    410 	return newrt;
    411 }
    412 
    413 void
    414 rtfree(struct rtentry *rt)
    415 {
    416 	struct ifaddr *ifa;
    417 
    418 	KASSERT(rt != NULL);
    419 	KASSERT(rt->rt_refcnt > 0);
    420 
    421 	rt->rt_refcnt--;
    422 	if (rt->rt_refcnt == 0 && (rt->rt_flags & RTF_UP) == 0) {
    423 		rt_assert_inactive(rt);
    424 		rttrash--;
    425 		rt_timer_remove_all(rt, 0);
    426 		ifa = rt->rt_ifa;
    427 		rt->rt_ifa = NULL;
    428 		ifafree(ifa);
    429 		rt->rt_ifp = NULL;
    430 		rt_destroy(rt);
    431 		pool_put(&rtentry_pool, rt);
    432 	}
    433 }
    434 
    435 /*
    436  * Force a routing table entry to the specified
    437  * destination to go through the given gateway.
    438  * Normally called as a result of a routing redirect
    439  * message from the network layer.
    440  *
    441  * N.B.: must be called at splsoftnet
    442  */
    443 void
    444 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
    445 	const struct sockaddr *netmask, int flags, const struct sockaddr *src,
    446 	struct rtentry **rtp)
    447 {
    448 	struct rtentry *rt;
    449 	int error = 0;
    450 	uint64_t *stat = NULL;
    451 	struct rt_addrinfo info;
    452 	struct ifaddr *ifa;
    453 
    454 	/* verify the gateway is directly reachable */
    455 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
    456 		error = ENETUNREACH;
    457 		goto out;
    458 	}
    459 	rt = rtalloc1(dst, 0);
    460 	/*
    461 	 * If the redirect isn't from our current router for this dst,
    462 	 * it's either old or wrong.  If it redirects us to ourselves,
    463 	 * we have a routing loop, perhaps as a result of an interface
    464 	 * going down recently.
    465 	 */
    466 	if (!(flags & RTF_DONE) && rt &&
    467 	     (sockaddr_cmp(src, rt->rt_gateway) != 0 || rt->rt_ifa != ifa))
    468 		error = EINVAL;
    469 	else if (ifa_ifwithaddr(gateway))
    470 		error = EHOSTUNREACH;
    471 	if (error)
    472 		goto done;
    473 	/*
    474 	 * Create a new entry if we just got back a wildcard entry
    475 	 * or the lookup failed.  This is necessary for hosts
    476 	 * which use routing redirects generated by smart gateways
    477 	 * to dynamically build the routing tables.
    478 	 */
    479 	if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
    480 		goto create;
    481 	/*
    482 	 * Don't listen to the redirect if it's
    483 	 * for a route to an interface.
    484 	 */
    485 	if (rt->rt_flags & RTF_GATEWAY) {
    486 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
    487 			/*
    488 			 * Changing from route to net => route to host.
    489 			 * Create new route, rather than smashing route to net.
    490 			 */
    491 		create:
    492 			if (rt != NULL)
    493 				rtfree(rt);
    494 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
    495 			memset(&info, 0, sizeof(info));
    496 			info.rti_info[RTAX_DST] = dst;
    497 			info.rti_info[RTAX_GATEWAY] = gateway;
    498 			info.rti_info[RTAX_NETMASK] = netmask;
    499 			info.rti_ifa = ifa;
    500 			info.rti_flags = flags;
    501 			rt = NULL;
    502 			error = rtrequest1(RTM_ADD, &info, &rt);
    503 			if (rt != NULL)
    504 				flags = rt->rt_flags;
    505 			stat = &rtstat.rts_dynamic;
    506 		} else {
    507 			/*
    508 			 * Smash the current notion of the gateway to
    509 			 * this destination.  Should check about netmask!!!
    510 			 */
    511 			rt->rt_flags |= RTF_MODIFIED;
    512 			flags |= RTF_MODIFIED;
    513 			stat = &rtstat.rts_newgateway;
    514 			rt_setgate(rt, gateway);
    515 		}
    516 	} else
    517 		error = EHOSTUNREACH;
    518 done:
    519 	if (rt) {
    520 		if (rtp != NULL && !error)
    521 			*rtp = rt;
    522 		else
    523 			rtfree(rt);
    524 	}
    525 out:
    526 	if (error)
    527 		rtstat.rts_badredirect++;
    528 	else if (stat != NULL)
    529 		(*stat)++;
    530 	memset(&info, 0, sizeof(info));
    531 	info.rti_info[RTAX_DST] = dst;
    532 	info.rti_info[RTAX_GATEWAY] = gateway;
    533 	info.rti_info[RTAX_NETMASK] = netmask;
    534 	info.rti_info[RTAX_AUTHOR] = src;
    535 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
    536 }
    537 
    538 /*
    539  * Delete a route and generate a message.
    540  * It doesn't free a passed rt.
    541  */
    542 static int
    543 rtdeletemsg(struct rtentry *rt)
    544 {
    545 	int error;
    546 	struct rt_addrinfo info;
    547 	struct rtentry *retrt;
    548 
    549 	/*
    550 	 * Request the new route so that the entry is not actually
    551 	 * deleted.  That will allow the information being reported to
    552 	 * be accurate (and consistent with route_output()).
    553 	 */
    554 	memset(&info, 0, sizeof(info));
    555 	info.rti_info[RTAX_DST] = rt_getkey(rt);
    556 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    557 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    558 	info.rti_flags = rt->rt_flags;
    559 	error = rtrequest1(RTM_DELETE, &info, &retrt);
    560 
    561 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
    562 
    563 	if (error == 0)
    564 		rtfree(retrt);
    565 	return error;
    566 }
    567 
    568 static int
    569 rtflushclone1(struct rtentry *rt, void *arg)
    570 {
    571 	struct rtentry *parent;
    572 
    573 	parent = (struct rtentry *)arg;
    574 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
    575 		rtdeletemsg(rt);
    576 	return 0;
    577 }
    578 
    579 static void
    580 rtflushclone(sa_family_t family, struct rtentry *parent)
    581 {
    582 
    583 #ifdef DIAGNOSTIC
    584 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
    585 		panic("rtflushclone: called with a non-cloning route");
    586 #endif
    587 	rt_walktree(family, rtflushclone1, (void *)parent);
    588 }
    589 
    590 struct ifaddr *
    591 ifa_ifwithroute(int flags, const struct sockaddr *dst,
    592 	const struct sockaddr *gateway)
    593 {
    594 	struct ifaddr *ifa;
    595 	if ((flags & RTF_GATEWAY) == 0) {
    596 		/*
    597 		 * If we are adding a route to an interface,
    598 		 * and the interface is a pt to pt link
    599 		 * we should search for the destination
    600 		 * as our clue to the interface.  Otherwise
    601 		 * we can use the local address.
    602 		 */
    603 		ifa = NULL;
    604 		if ((flags & RTF_HOST) && gateway->sa_family != AF_LINK)
    605 			ifa = ifa_ifwithdstaddr(dst);
    606 		if (ifa == NULL)
    607 			ifa = ifa_ifwithaddr(gateway);
    608 	} else {
    609 		/*
    610 		 * If we are adding a route to a remote net
    611 		 * or host, the gateway may still be on the
    612 		 * other end of a pt to pt link.
    613 		 */
    614 		ifa = ifa_ifwithdstaddr(gateway);
    615 	}
    616 	if (ifa == NULL)
    617 		ifa = ifa_ifwithnet(gateway);
    618 	if (ifa == NULL) {
    619 		struct rtentry *rt = rtalloc1(dst, 0);
    620 		if (rt == NULL)
    621 			return NULL;
    622 		ifa = rt->rt_ifa;
    623 		rtfree(rt);
    624 		if (ifa == NULL)
    625 			return NULL;
    626 	}
    627 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
    628 		struct ifaddr *oifa = ifa;
    629 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
    630 		if (ifa == NULL)
    631 			ifa = oifa;
    632 	}
    633 	return ifa;
    634 }
    635 
    636 /*
    637  * If it suceeds and ret_nrt isn't NULL, refcnt of ret_nrt is incremented.
    638  * The caller has to rtfree it by itself.
    639  */
    640 int
    641 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
    642 	const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
    643 {
    644 	struct rt_addrinfo info;
    645 
    646 	memset(&info, 0, sizeof(info));
    647 	info.rti_flags = flags;
    648 	info.rti_info[RTAX_DST] = dst;
    649 	info.rti_info[RTAX_GATEWAY] = gateway;
    650 	info.rti_info[RTAX_NETMASK] = netmask;
    651 	return rtrequest1(req, &info, ret_nrt);
    652 }
    653 
    654 /*
    655  * It's a utility function to add/remove a route to/from the routing table
    656  * and tell user processes the addition/removal on success.
    657  */
    658 int
    659 rtrequest_newmsg(const int req, const struct sockaddr *dst,
    660 	const struct sockaddr *gateway, const struct sockaddr *netmask,
    661 	const int flags)
    662 {
    663 	int error;
    664 	struct rtentry *ret_nrt = NULL;
    665 
    666 	KASSERT(req == RTM_ADD || req == RTM_DELETE);
    667 
    668 	error = rtrequest(req, dst, gateway, netmask, flags, &ret_nrt);
    669 	if (error != 0)
    670 		return error;
    671 
    672 	KASSERT(ret_nrt != NULL);
    673 
    674 	rt_newmsg(req, ret_nrt); /* tell user process */
    675 	rtfree(ret_nrt);
    676 
    677 	return 0;
    678 }
    679 
    680 int
    681 rt_getifa(struct rt_addrinfo *info)
    682 {
    683 	struct ifaddr *ifa;
    684 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
    685 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
    686 	const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
    687 	const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
    688 	int flags = info->rti_flags;
    689 
    690 	/*
    691 	 * ifp may be specified by sockaddr_dl when protocol address
    692 	 * is ambiguous
    693 	 */
    694 	if (info->rti_ifp == NULL && ifpaddr != NULL
    695 	    && ifpaddr->sa_family == AF_LINK &&
    696 	    (ifa = ifa_ifwithnet(ifpaddr)) != NULL)
    697 		info->rti_ifp = ifa->ifa_ifp;
    698 	if (info->rti_ifa == NULL && ifaaddr != NULL)
    699 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
    700 	if (info->rti_ifa == NULL) {
    701 		const struct sockaddr *sa;
    702 
    703 		sa = ifaaddr != NULL ? ifaaddr :
    704 		    (gateway != NULL ? gateway : dst);
    705 		if (sa != NULL && info->rti_ifp != NULL)
    706 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
    707 		else if (dst != NULL && gateway != NULL)
    708 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
    709 		else if (sa != NULL)
    710 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
    711 	}
    712 	if ((ifa = info->rti_ifa) == NULL)
    713 		return ENETUNREACH;
    714 	if (ifa->ifa_getifa != NULL) {
    715 		info->rti_ifa = ifa = (*ifa->ifa_getifa)(ifa, dst);
    716 		if (ifa == NULL)
    717 			return ENETUNREACH;
    718 	}
    719 	if (info->rti_ifp == NULL)
    720 		info->rti_ifp = ifa->ifa_ifp;
    721 	return 0;
    722 }
    723 
    724 /*
    725  * If it suceeds and ret_nrt isn't NULL, refcnt of ret_nrt is incremented.
    726  * The caller has to rtfree it by itself.
    727  */
    728 int
    729 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
    730 {
    731 	int s = splsoftnet();
    732 	int error = 0, rc;
    733 	struct rtentry *rt, *crt;
    734 	rtbl_t *rtbl;
    735 	struct ifaddr *ifa, *ifa2;
    736 	struct sockaddr_storage maskeddst;
    737 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
    738 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
    739 	const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
    740 	int flags = info->rti_flags;
    741 #define senderr(x) { error = x ; goto bad; }
    742 
    743 	if ((rtbl = rt_gettable(dst->sa_family)) == NULL)
    744 		senderr(ESRCH);
    745 	if (flags & RTF_HOST)
    746 		netmask = NULL;
    747 	switch (req) {
    748 	case RTM_DELETE:
    749 		if (netmask) {
    750 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    751 			    netmask);
    752 			dst = (struct sockaddr *)&maskeddst;
    753 		}
    754 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
    755 			senderr(ESRCH);
    756 		if ((rt->rt_flags & RTF_CLONING) != 0) {
    757 			/* clean up any cloned children */
    758 			rtflushclone(dst->sa_family, rt);
    759 		}
    760 		if ((rt = rt_deladdr(rtbl, dst, netmask)) == NULL)
    761 			senderr(ESRCH);
    762 		if (rt->rt_gwroute) {
    763 			rtfree(rt->rt_gwroute);
    764 			rt->rt_gwroute = NULL;
    765 		}
    766 		if (rt->rt_parent) {
    767 			rt->rt_parent->rt_refcnt--;
    768 			rt->rt_parent = NULL;
    769 		}
    770 		rt->rt_flags &= ~RTF_UP;
    771 		if ((ifa = rt->rt_ifa)) {
    772 			if (ifa->ifa_flags & IFA_ROUTE &&
    773 			    rt_ifa_connected(rt, ifa)) {
    774 				RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    775 				    "deleted IFA_ROUTE\n",
    776 				    (void *)rt->_rt_key, (void *)ifa);
    777 				ifa->ifa_flags &= ~IFA_ROUTE;
    778 			}
    779 			if (ifa->ifa_rtrequest)
    780 				ifa->ifa_rtrequest(RTM_DELETE, rt, info);
    781 		}
    782 		rttrash++;
    783 		if (ret_nrt) {
    784 			*ret_nrt = rt;
    785 			rt->rt_refcnt++;
    786 		} else if (rt->rt_refcnt <= 0) {
    787 			/* Adjust the refcount */
    788 			rt->rt_refcnt++;
    789 			rtfree(rt);
    790 		}
    791 		break;
    792 
    793 	case RTM_RESOLVE:
    794 		if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
    795 			senderr(EINVAL);
    796 		if ((rt->rt_flags & RTF_CLONING) == 0)
    797 			senderr(EINVAL);
    798 		ifa = rt->rt_ifa;
    799 		flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC);
    800 		flags |= RTF_CLONED;
    801 		gateway = rt->rt_gateway;
    802 		flags |= RTF_HOST;
    803 		goto makeroute;
    804 
    805 	case RTM_ADD:
    806 		if (info->rti_ifa == NULL && (error = rt_getifa(info)))
    807 			senderr(error);
    808 		ifa = info->rti_ifa;
    809 	makeroute:
    810 		/* Already at splsoftnet() so pool_get/pool_put are safe */
    811 		rt = pool_get(&rtentry_pool, PR_NOWAIT);
    812 		if (rt == NULL)
    813 			senderr(ENOBUFS);
    814 		memset(rt, 0, sizeof(*rt));
    815 		rt->rt_flags = RTF_UP | flags;
    816 		LIST_INIT(&rt->rt_timer);
    817 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    818 		if (rt_setkey(rt, dst, M_NOWAIT) == NULL ||
    819 		    rt_setgate(rt, gateway) != 0) {
    820 			pool_put(&rtentry_pool, rt);
    821 			senderr(ENOBUFS);
    822 		}
    823 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    824 		if (netmask) {
    825 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    826 			    netmask);
    827 			rt_setkey(rt, (struct sockaddr *)&maskeddst, M_NOWAIT);
    828 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    829 		} else {
    830 			rt_setkey(rt, dst, M_NOWAIT);
    831 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    832 		}
    833 		rt_set_ifa(rt, ifa);
    834 		if (info->rti_info[RTAX_TAG] != NULL)
    835 			rt_settag(rt, info->rti_info[RTAX_TAG]);
    836 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    837 		if (info->rti_info[RTAX_IFP] != NULL &&
    838 		    (ifa2 = ifa_ifwithnet(info->rti_info[RTAX_IFP])) != NULL &&
    839 		    ifa2->ifa_ifp != NULL)
    840 			rt->rt_ifp = ifa2->ifa_ifp;
    841 		else
    842 			rt->rt_ifp = ifa->ifa_ifp;
    843 		if (req == RTM_RESOLVE) {
    844 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
    845 			rt->rt_parent = *ret_nrt;
    846 			rt->rt_parent->rt_refcnt++;
    847 		}
    848 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    849 		rc = rt_addaddr(rtbl, rt, netmask);
    850 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    851 		if (rc != 0 && (crt = rtalloc1(rt_getkey(rt), 0)) != NULL) {
    852 			/* overwrite cloned route */
    853 			if ((crt->rt_flags & RTF_CLONED) != 0) {
    854 				rtdeletemsg(crt);
    855 				rc = rt_addaddr(rtbl, rt, netmask);
    856 			}
    857 			rtfree(crt);
    858 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    859 		}
    860 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    861 		if (rc != 0) {
    862 			ifafree(ifa);
    863 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
    864 				rtfree(rt->rt_parent);
    865 			if (rt->rt_gwroute)
    866 				rtfree(rt->rt_gwroute);
    867 			rt_destroy(rt);
    868 			pool_put(&rtentry_pool, rt);
    869 			senderr(rc);
    870 		}
    871 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    872 		if (ifa->ifa_rtrequest)
    873 			ifa->ifa_rtrequest(req, rt, info);
    874 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    875 		if (ret_nrt) {
    876 			*ret_nrt = rt;
    877 			rt->rt_refcnt++;
    878 		}
    879 		if ((rt->rt_flags & RTF_CLONING) != 0) {
    880 			/* clean up any cloned children */
    881 			rtflushclone(dst->sa_family, rt);
    882 		}
    883 		rtflushall(dst->sa_family);
    884 		break;
    885 	case RTM_GET:
    886 		if (netmask != NULL) {
    887 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    888 			    netmask);
    889 			dst = (struct sockaddr *)&maskeddst;
    890 		}
    891 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
    892 			senderr(ESRCH);
    893 		if (ret_nrt != NULL) {
    894 			*ret_nrt = rt;
    895 			rt->rt_refcnt++;
    896 		}
    897 		break;
    898 	}
    899 bad:
    900 	splx(s);
    901 	return error;
    902 }
    903 
    904 int
    905 rt_setgate(struct rtentry *rt, const struct sockaddr *gate)
    906 {
    907 	KASSERT(rt != rt->rt_gwroute);
    908 
    909 	KASSERT(rt->_rt_key != NULL);
    910 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    911 
    912 	if (rt->rt_gwroute) {
    913 		rtfree(rt->rt_gwroute);
    914 		rt->rt_gwroute = NULL;
    915 	}
    916 	KASSERT(rt->_rt_key != NULL);
    917 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    918 	if (rt->rt_gateway != NULL)
    919 		sockaddr_free(rt->rt_gateway);
    920 	KASSERT(rt->_rt_key != NULL);
    921 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    922 	if ((rt->rt_gateway = sockaddr_dup(gate, M_ZERO | M_NOWAIT)) == NULL)
    923 		return ENOMEM;
    924 	KASSERT(rt->_rt_key != NULL);
    925 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    926 
    927 	if (rt->rt_flags & RTF_GATEWAY) {
    928 		KASSERT(rt->_rt_key != NULL);
    929 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    930 		rt->rt_gwroute = rtalloc1(gate, 1);
    931 		/*
    932 		 * If we switched gateways, grab the MTU from the new
    933 		 * gateway route if the current MTU, if the current MTU is
    934 		 * greater than the MTU of gateway.
    935 		 * Note that, if the MTU of gateway is 0, we will reset the
    936 		 * MTU of the route to run PMTUD again from scratch. XXX
    937 		 */
    938 		KASSERT(rt->_rt_key != NULL);
    939 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    940 		if (rt->rt_gwroute
    941 		    && !(rt->rt_rmx.rmx_locks & RTV_MTU)
    942 		    && rt->rt_rmx.rmx_mtu
    943 		    && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
    944 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
    945 		}
    946 	}
    947 	KASSERT(rt->_rt_key != NULL);
    948 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    949 	return 0;
    950 }
    951 
    952 static void
    953 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
    954 	const struct sockaddr *netmask)
    955 {
    956 	const char *netmaskp = &netmask->sa_data[0],
    957 	           *srcp = &src->sa_data[0];
    958 	char *dstp = &dst->sa_data[0];
    959 	const char *maskend = (char *)dst + MIN(netmask->sa_len, src->sa_len);
    960 	const char *srcend = (char *)dst + src->sa_len;
    961 
    962 	dst->sa_len = src->sa_len;
    963 	dst->sa_family = src->sa_family;
    964 
    965 	while (dstp < maskend)
    966 		*dstp++ = *srcp++ & *netmaskp++;
    967 	if (dstp < srcend)
    968 		memset(dstp, 0, (size_t)(srcend - dstp));
    969 }
    970 
    971 /*
    972  * Inform the routing socket of a route change.
    973  */
    974 void
    975 rt_newmsg(int cmd, struct rtentry *rt)
    976 {
    977 	struct rt_addrinfo info;
    978 
    979 	memset((void *)&info, 0, sizeof(info));
    980 	info.rti_info[RTAX_DST] = rt_getkey(rt);
    981 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    982 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    983 	if (rt->rt_ifp) {
    984 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
    985 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
    986 	}
    987 
    988 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
    989 }
    990 
    991 /*
    992  * Set up or tear down a routing table entry, normally
    993  * for an interface.
    994  */
    995 int
    996 rtinit(struct ifaddr *ifa, int cmd, int flags)
    997 {
    998 	struct rtentry *rt;
    999 	struct sockaddr *dst, *odst;
   1000 	struct sockaddr_storage maskeddst;
   1001 	struct rtentry *nrt = NULL;
   1002 	int error;
   1003 	struct rt_addrinfo info;
   1004 	struct sockaddr_dl *sdl;
   1005 	const struct sockaddr_dl *ifsdl;
   1006 
   1007 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
   1008 	if (cmd == RTM_DELETE) {
   1009 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
   1010 			/* Delete subnet route for this interface */
   1011 			odst = dst;
   1012 			dst = (struct sockaddr *)&maskeddst;
   1013 			rt_maskedcopy(odst, dst, ifa->ifa_netmask);
   1014 		}
   1015 		if ((rt = rtalloc1(dst, 0)) != NULL) {
   1016 			if (rt->rt_ifa != ifa) {
   1017 				rtfree(rt);
   1018 				return (flags & RTF_HOST) ? EHOSTUNREACH
   1019 							: ENETUNREACH;
   1020 			}
   1021 			rtfree(rt);
   1022 		}
   1023 	}
   1024 	memset(&info, 0, sizeof(info));
   1025 	info.rti_ifa = ifa;
   1026 	info.rti_flags = flags | ifa->ifa_flags;
   1027 	info.rti_info[RTAX_DST] = dst;
   1028 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
   1029 	/*
   1030 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
   1031 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
   1032 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
   1033 	 * change it to meet bsdi4 behavior.
   1034 	 */
   1035 	if (cmd != RTM_LLINFO_UPD)
   1036 		info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
   1037 	error = rtrequest1((cmd == RTM_LLINFO_UPD) ? RTM_GET : cmd, &info,
   1038 	    &nrt);
   1039 	if (error != 0 || (rt = nrt) == NULL)
   1040 		return error;
   1041 
   1042 	switch (cmd) {
   1043 	case RTM_DELETE:
   1044 		rt_newmsg(cmd, 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 		break;
   1065 	case RTM_ADD:
   1066 		if (rt->rt_ifa != ifa) {
   1067 			printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
   1068 				rt->rt_ifa);
   1069 			if (rt->rt_ifa->ifa_rtrequest != NULL) {
   1070 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
   1071 				    &info);
   1072 			}
   1073 			rt_replace_ifa(rt, ifa);
   1074 			rt->rt_ifp = ifa->ifa_ifp;
   1075 			if (ifa->ifa_rtrequest != NULL)
   1076 				ifa->ifa_rtrequest(RTM_ADD, rt, &info);
   1077 		}
   1078 		rt_newmsg(cmd, rt);
   1079 		break;
   1080 	}
   1081 	rtfree(rt);
   1082 	return error;
   1083 }
   1084 
   1085 static const struct in_addr inmask32 = {.s_addr = INADDR_BROADCAST};
   1086 
   1087 /* Subroutine for rt_ifa_addlocal() and rt_ifa_remlocal() */
   1088 static int
   1089 rt_ifa_localrequest(int cmd, struct ifaddr *ifa)
   1090 {
   1091 	struct sockaddr *all1_sa;
   1092 	struct sockaddr_in all1_sin;
   1093 #ifdef INET6
   1094 	struct sockaddr_in6 all1_sin6;
   1095 #endif
   1096 	struct rtentry *nrt = NULL;
   1097 	int flags, e;
   1098 
   1099 	switch(ifa->ifa_addr->sa_family) {
   1100 	case AF_INET:
   1101 		sockaddr_in_init(&all1_sin, &inmask32, 0);
   1102 		all1_sa = (struct sockaddr *)&all1_sin;
   1103 		break;
   1104 #ifdef INET6
   1105 	case AF_INET6:
   1106 		sockaddr_in6_init(&all1_sin6, &in6mask128, 0, 0, 0);
   1107 		all1_sa = (struct sockaddr *)&all1_sin6;
   1108 		break;
   1109 #endif
   1110 	default:
   1111 		return 0;
   1112 	}
   1113 
   1114 	flags = RTF_UP | RTF_HOST | RTF_LOCAL;
   1115 	if (!(ifa->ifa_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)))
   1116 		flags |= RTF_LLINFO;
   1117 	e = rtrequest(cmd, ifa->ifa_addr, ifa->ifa_addr, all1_sa, flags, &nrt);
   1118 
   1119 	/* Make sure rt_ifa be equal to IFA, the second argument of the
   1120 	 * function. */
   1121 	if (cmd == RTM_ADD && nrt && ifa != nrt->rt_ifa)
   1122 		rt_replace_ifa(nrt, ifa);
   1123 
   1124 	rt_newaddrmsg(cmd, ifa, e, nrt);
   1125 	if (nrt != NULL)
   1126 		rtfree(nrt);
   1127 
   1128 	return e;
   1129 }
   1130 
   1131 /*
   1132  * Create a local route entry for the address.
   1133  * Announce the addition of the address and the route to the routing socket.
   1134  */
   1135 int
   1136 rt_ifa_addlocal(struct ifaddr *ifa)
   1137 {
   1138 	struct rtentry *rt;
   1139 	int e;
   1140 
   1141 	/* If there is no loopback entry, allocate one. */
   1142 	rt = rtalloc1(ifa->ifa_addr, 0);
   1143 	if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
   1144 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
   1145 		e = rt_ifa_localrequest(RTM_ADD, ifa);
   1146 	else {
   1147 		e = 0;
   1148 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1149 	}
   1150 	if (rt != NULL)
   1151 		rtfree(rt);
   1152 	return e;
   1153 }
   1154 
   1155 /*
   1156  * Remove the local route entry for the address.
   1157  * Announce the removal of the address and the route to the routing socket.
   1158  */
   1159 int
   1160 rt_ifa_remlocal(struct ifaddr *ifa, struct ifaddr *alt_ifa)
   1161 {
   1162 	struct rtentry *rt;
   1163 	int e = 0;
   1164 
   1165 	rt = rtalloc1(ifa->ifa_addr, 0);
   1166 
   1167 	/*
   1168 	 * Before deleting, check if a corresponding loopbacked
   1169 	 * host route surely exists.  With this check, we can avoid
   1170 	 * deleting an interface direct route whose destination is
   1171 	 * the same as the address being removed.  This can happen
   1172 	 * when removing a subnet-router anycast address on an
   1173 	 * interface attached to a shared medium.
   1174 	 */
   1175 	if (rt != NULL &&
   1176 	    (rt->rt_flags & RTF_HOST) &&
   1177 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK))
   1178 	{
   1179 		/* If we cannot replace the route's ifaddr with the equivalent
   1180 		 * ifaddr of another interface, I believe it is safest to
   1181 		 * delete the route.
   1182 		 */
   1183 		if (alt_ifa == NULL)
   1184 			e = rt_ifa_localrequest(RTM_DELETE, ifa);
   1185 		else {
   1186 			rt_replace_ifa(rt, alt_ifa);
   1187 			rt_newmsg(RTM_CHANGE, rt);
   1188 		}
   1189 	} else
   1190 		rt_newaddrmsg(RTM_DELADDR, ifa, 0, NULL);
   1191 	if (rt != NULL)
   1192 		rtfree(rt);
   1193 	return e;
   1194 }
   1195 
   1196 /*
   1197  * Route timer routines.  These routes allow functions to be called
   1198  * for various routes at any time.  This is useful in supporting
   1199  * path MTU discovery and redirect route deletion.
   1200  *
   1201  * This is similar to some BSDI internal functions, but it provides
   1202  * for multiple queues for efficiency's sake...
   1203  */
   1204 
   1205 LIST_HEAD(, rttimer_queue) rttimer_queue_head;
   1206 static int rt_init_done = 0;
   1207 
   1208 #define RTTIMER_CALLOUT(r)	do {					\
   1209 		if (r->rtt_func != NULL) {				\
   1210 			(*r->rtt_func)(r->rtt_rt, r);			\
   1211 		} else {						\
   1212 			rtrequest((int) RTM_DELETE,			\
   1213 				  rt_getkey(r->rtt_rt),			\
   1214 				  0, 0, 0, 0);				\
   1215 		}							\
   1216 	} while (/*CONSTCOND*/0)
   1217 
   1218 /*
   1219  * Some subtle order problems with domain initialization mean that
   1220  * we cannot count on this being run from rt_init before various
   1221  * protocol initializations are done.  Therefore, we make sure
   1222  * that this is run when the first queue is added...
   1223  */
   1224 
   1225 void
   1226 rt_timer_init(void)
   1227 {
   1228 	assert(rt_init_done == 0);
   1229 
   1230 	LIST_INIT(&rttimer_queue_head);
   1231 	callout_init(&rt_timer_ch, 0);
   1232 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
   1233 	rt_init_done = 1;
   1234 }
   1235 
   1236 struct rttimer_queue *
   1237 rt_timer_queue_create(u_int timeout)
   1238 {
   1239 	struct rttimer_queue *rtq;
   1240 
   1241 	if (rt_init_done == 0)
   1242 		rt_timer_init();
   1243 
   1244 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
   1245 	if (rtq == NULL)
   1246 		return NULL;
   1247 	memset(rtq, 0, sizeof(*rtq));
   1248 
   1249 	rtq->rtq_timeout = timeout;
   1250 	TAILQ_INIT(&rtq->rtq_head);
   1251 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
   1252 
   1253 	return rtq;
   1254 }
   1255 
   1256 void
   1257 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
   1258 {
   1259 
   1260 	rtq->rtq_timeout = timeout;
   1261 }
   1262 
   1263 void
   1264 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy)
   1265 {
   1266 	struct rttimer *r;
   1267 
   1268 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
   1269 		LIST_REMOVE(r, rtt_link);
   1270 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
   1271 		if (destroy)
   1272 			RTTIMER_CALLOUT(r);
   1273 		/* we are already at splsoftnet */
   1274 		pool_put(&rttimer_pool, r);
   1275 		if (rtq->rtq_count > 0)
   1276 			rtq->rtq_count--;
   1277 		else
   1278 			printf("rt_timer_queue_remove_all: "
   1279 			    "rtq_count reached 0\n");
   1280 	}
   1281 }
   1282 
   1283 void
   1284 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
   1285 {
   1286 
   1287 	rt_timer_queue_remove_all(rtq, destroy);
   1288 
   1289 	LIST_REMOVE(rtq, rtq_link);
   1290 
   1291 	/*
   1292 	 * Caller is responsible for freeing the rttimer_queue structure.
   1293 	 */
   1294 }
   1295 
   1296 unsigned long
   1297 rt_timer_count(struct rttimer_queue *rtq)
   1298 {
   1299 	return rtq->rtq_count;
   1300 }
   1301 
   1302 void
   1303 rt_timer_remove_all(struct rtentry *rt, int destroy)
   1304 {
   1305 	struct rttimer *r;
   1306 
   1307 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
   1308 		LIST_REMOVE(r, rtt_link);
   1309 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
   1310 		if (destroy)
   1311 			RTTIMER_CALLOUT(r);
   1312 		if (r->rtt_queue->rtq_count > 0)
   1313 			r->rtt_queue->rtq_count--;
   1314 		else
   1315 			printf("rt_timer_remove_all: rtq_count reached 0\n");
   1316 		/* we are already at splsoftnet */
   1317 		pool_put(&rttimer_pool, r);
   1318 	}
   1319 }
   1320 
   1321 int
   1322 rt_timer_add(struct rtentry *rt,
   1323 	void (*func)(struct rtentry *, struct rttimer *),
   1324 	struct rttimer_queue *queue)
   1325 {
   1326 	struct rttimer *r;
   1327 	int s;
   1328 
   1329 	/*
   1330 	 * If there's already a timer with this action, destroy it before
   1331 	 * we add a new one.
   1332 	 */
   1333 	LIST_FOREACH(r, &rt->rt_timer, rtt_link) {
   1334 		if (r->rtt_func == func)
   1335 			break;
   1336 	}
   1337 	if (r != NULL) {
   1338 		LIST_REMOVE(r, rtt_link);
   1339 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
   1340 		if (r->rtt_queue->rtq_count > 0)
   1341 			r->rtt_queue->rtq_count--;
   1342 		else
   1343 			printf("rt_timer_add: rtq_count reached 0\n");
   1344 	} else {
   1345 		s = splsoftnet();
   1346 		r = pool_get(&rttimer_pool, PR_NOWAIT);
   1347 		splx(s);
   1348 		if (r == NULL)
   1349 			return ENOBUFS;
   1350 	}
   1351 
   1352 	memset(r, 0, sizeof(*r));
   1353 
   1354 	r->rtt_rt = rt;
   1355 	r->rtt_time = time_uptime;
   1356 	r->rtt_func = func;
   1357 	r->rtt_queue = queue;
   1358 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
   1359 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
   1360 	r->rtt_queue->rtq_count++;
   1361 
   1362 	return 0;
   1363 }
   1364 
   1365 /* ARGSUSED */
   1366 void
   1367 rt_timer_timer(void *arg)
   1368 {
   1369 	struct rttimer_queue *rtq;
   1370 	struct rttimer *r;
   1371 	int s;
   1372 
   1373 	s = splsoftnet();
   1374 	LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) {
   1375 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
   1376 		    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
   1377 			LIST_REMOVE(r, rtt_link);
   1378 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
   1379 			RTTIMER_CALLOUT(r);
   1380 			pool_put(&rttimer_pool, r);
   1381 			if (rtq->rtq_count > 0)
   1382 				rtq->rtq_count--;
   1383 			else
   1384 				printf("rt_timer_timer: rtq_count reached 0\n");
   1385 		}
   1386 	}
   1387 	splx(s);
   1388 
   1389 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
   1390 }
   1391 
   1392 static struct rtentry *
   1393 _rtcache_init(struct route *ro, int flag)
   1394 {
   1395 	rtcache_invariants(ro);
   1396 	KASSERT(ro->_ro_rt == NULL);
   1397 
   1398 	if (rtcache_getdst(ro) == NULL)
   1399 		return NULL;
   1400 	ro->ro_invalid = false;
   1401 	if ((ro->_ro_rt = rtalloc1(rtcache_getdst(ro), flag)) != NULL)
   1402 		rtcache(ro);
   1403 
   1404 	rtcache_invariants(ro);
   1405 	return ro->_ro_rt;
   1406 }
   1407 
   1408 struct rtentry *
   1409 rtcache_init(struct route *ro)
   1410 {
   1411 	return _rtcache_init(ro, 1);
   1412 }
   1413 
   1414 struct rtentry *
   1415 rtcache_init_noclone(struct route *ro)
   1416 {
   1417 	return _rtcache_init(ro, 0);
   1418 }
   1419 
   1420 struct rtentry *
   1421 rtcache_update(struct route *ro, int clone)
   1422 {
   1423 	rtcache_clear(ro);
   1424 	return _rtcache_init(ro, clone);
   1425 }
   1426 
   1427 void
   1428 rtcache_copy(struct route *new_ro, const struct route *old_ro)
   1429 {
   1430 	struct rtentry *rt;
   1431 
   1432 	KASSERT(new_ro != old_ro);
   1433 	rtcache_invariants(new_ro);
   1434 	rtcache_invariants(old_ro);
   1435 
   1436 	if ((rt = rtcache_validate(old_ro)) != NULL)
   1437 		rt->rt_refcnt++;
   1438 
   1439 	if (rtcache_getdst(old_ro) == NULL ||
   1440 	    rtcache_setdst(new_ro, rtcache_getdst(old_ro)) != 0)
   1441 		return;
   1442 
   1443 	new_ro->ro_invalid = false;
   1444 	if ((new_ro->_ro_rt = rt) != NULL)
   1445 		rtcache(new_ro);
   1446 	rtcache_invariants(new_ro);
   1447 }
   1448 
   1449 static struct dom_rtlist invalid_routes = LIST_HEAD_INITIALIZER(dom_rtlist);
   1450 
   1451 static void
   1452 rtcache_invalidate(struct dom_rtlist *rtlist)
   1453 {
   1454 	struct route *ro;
   1455 
   1456 	while ((ro = LIST_FIRST(rtlist)) != NULL) {
   1457 		rtcache_invariants(ro);
   1458 		KASSERT(ro->_ro_rt != NULL);
   1459 		ro->ro_invalid = true;
   1460 		LIST_REMOVE(ro, ro_rtcache_next);
   1461 		LIST_INSERT_HEAD(&invalid_routes, ro, ro_rtcache_next);
   1462 		rtcache_invariants(ro);
   1463 	}
   1464 }
   1465 
   1466 static void
   1467 rtcache_clear(struct route *ro)
   1468 {
   1469 	rtcache_invariants(ro);
   1470 	if (ro->_ro_rt == NULL)
   1471 		return;
   1472 
   1473 	LIST_REMOVE(ro, ro_rtcache_next);
   1474 
   1475 	rtfree(ro->_ro_rt);
   1476 	ro->_ro_rt = NULL;
   1477 	ro->ro_invalid = false;
   1478 	rtcache_invariants(ro);
   1479 }
   1480 
   1481 struct rtentry *
   1482 rtcache_lookup2(struct route *ro, const struct sockaddr *dst, int clone,
   1483     int *hitp)
   1484 {
   1485 	const struct sockaddr *odst;
   1486 	struct rtentry *rt = NULL;
   1487 
   1488 	odst = rtcache_getdst(ro);
   1489 	if (odst == NULL)
   1490 		goto miss;
   1491 
   1492 	if (sockaddr_cmp(odst, dst) != 0) {
   1493 		rtcache_free(ro);
   1494 		goto miss;
   1495 	}
   1496 
   1497 	rt = rtcache_validate(ro);
   1498 	if (rt == NULL) {
   1499 		rtcache_clear(ro);
   1500 		goto miss;
   1501 	}
   1502 
   1503 	*hitp = 1;
   1504 	rtcache_invariants(ro);
   1505 
   1506 	return rt;
   1507 miss:
   1508 	*hitp = 0;
   1509 	if (rtcache_setdst(ro, dst) == 0)
   1510 		rt = _rtcache_init(ro, clone);
   1511 
   1512 	rtcache_invariants(ro);
   1513 
   1514 	return rt;
   1515 }
   1516 
   1517 void
   1518 rtcache_free(struct route *ro)
   1519 {
   1520 	rtcache_clear(ro);
   1521 	if (ro->ro_sa != NULL) {
   1522 		sockaddr_free(ro->ro_sa);
   1523 		ro->ro_sa = NULL;
   1524 	}
   1525 	rtcache_invariants(ro);
   1526 }
   1527 
   1528 int
   1529 rtcache_setdst(struct route *ro, const struct sockaddr *sa)
   1530 {
   1531 	KASSERT(sa != NULL);
   1532 
   1533 	rtcache_invariants(ro);
   1534 	if (ro->ro_sa != NULL) {
   1535 		if (ro->ro_sa->sa_family == sa->sa_family) {
   1536 			rtcache_clear(ro);
   1537 			sockaddr_copy(ro->ro_sa, ro->ro_sa->sa_len, sa);
   1538 			rtcache_invariants(ro);
   1539 			return 0;
   1540 		}
   1541 		/* free ro_sa, wrong family */
   1542 		rtcache_free(ro);
   1543 	}
   1544 
   1545 	KASSERT(ro->_ro_rt == NULL);
   1546 
   1547 	if ((ro->ro_sa = sockaddr_dup(sa, M_ZERO | M_NOWAIT)) == NULL) {
   1548 		rtcache_invariants(ro);
   1549 		return ENOMEM;
   1550 	}
   1551 	rtcache_invariants(ro);
   1552 	return 0;
   1553 }
   1554 
   1555 const struct sockaddr *
   1556 rt_settag(struct rtentry *rt, const struct sockaddr *tag)
   1557 {
   1558 	if (rt->rt_tag != tag) {
   1559 		if (rt->rt_tag != NULL)
   1560 			sockaddr_free(rt->rt_tag);
   1561 		rt->rt_tag = sockaddr_dup(tag, M_ZERO | M_NOWAIT);
   1562 	}
   1563 	return rt->rt_tag;
   1564 }
   1565 
   1566 struct sockaddr *
   1567 rt_gettag(struct rtentry *rt)
   1568 {
   1569 	return rt->rt_tag;
   1570 }
   1571