Home | History | Annotate | Line # | Download | only in net
route.c revision 1.121
      1 /*	$NetBSD: route.c,v 1.121 2009/11/03 00:30:11 dyoung 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_route.h"
     94 
     95 #include <sys/cdefs.h>
     96 __KERNEL_RCSID(0, "$NetBSD: route.c,v 1.121 2009/11/03 00:30:11 dyoung Exp $");
     97 
     98 #include <sys/param.h>
     99 #include <sys/sysctl.h>
    100 #include <sys/systm.h>
    101 #include <sys/callout.h>
    102 #include <sys/proc.h>
    103 #include <sys/mbuf.h>
    104 #include <sys/socket.h>
    105 #include <sys/socketvar.h>
    106 #include <sys/domain.h>
    107 #include <sys/protosw.h>
    108 #include <sys/kernel.h>
    109 #include <sys/ioctl.h>
    110 #include <sys/pool.h>
    111 #include <sys/kauth.h>
    112 
    113 #include <net/if.h>
    114 #include <net/if_dl.h>
    115 #include <net/route.h>
    116 #include <net/raw_cb.h>
    117 
    118 #include <netinet/in.h>
    119 #include <netinet/in_var.h>
    120 
    121 #ifdef RTFLUSH_DEBUG
    122 #define	rtcache_debug() __predict_false(_rtcache_debug)
    123 #else /* RTFLUSH_DEBUG */
    124 #define	rtcache_debug() 0
    125 #endif /* RTFLUSH_DEBUG */
    126 
    127 struct	route_cb route_cb;
    128 struct	rtstat	rtstat;
    129 struct	radix_node_head *rt_tables[AF_MAX+1];
    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 
    148 #ifdef RTFLUSH_DEBUG
    149 static void sysctl_net_rtcache_setup(struct sysctllog **);
    150 static void
    151 sysctl_net_rtcache_setup(struct sysctllog **clog)
    152 {
    153 	const struct sysctlnode *rnode;
    154 
    155 	/* XXX do not duplicate */
    156 	if (sysctl_createv(clog, 0, NULL, &rnode, CTLFLAG_PERMANENT,
    157 	    CTLTYPE_NODE, "net", NULL, NULL, 0, NULL, 0, CTL_NET, CTL_EOL) != 0)
    158 		return;
    159 	if (sysctl_createv(clog, 0, &rnode, &rnode, CTLFLAG_PERMANENT,
    160 	    CTLTYPE_NODE,
    161 	    "rtcache", SYSCTL_DESCR("Route cache related settings"),
    162 	    NULL, 0, NULL, 0, CTL_CREATE, CTL_EOL) != 0)
    163 		return;
    164 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    165 	    CTLFLAG_PERMANENT|CTLFLAG_READWRITE, CTLTYPE_INT,
    166 	    "debug", SYSCTL_DESCR("Debug route caches"),
    167 	    NULL, 0, &_rtcache_debug, 0, CTL_CREATE, CTL_EOL) != 0)
    168 		return;
    169 }
    170 #endif /* RTFLUSH_DEBUG */
    171 
    172 struct ifaddr *
    173 rt_get_ifa(struct rtentry *rt)
    174 {
    175 	struct ifaddr *ifa;
    176 
    177 	if ((ifa = rt->rt_ifa) == NULL)
    178 		return ifa;
    179 	else if (ifa->ifa_getifa == NULL)
    180 		return ifa;
    181 #if 0
    182 	else if (ifa->ifa_seqno != NULL && *ifa->ifa_seqno == rt->rt_ifa_seqno)
    183 		return ifa;
    184 #endif
    185 	else {
    186 		ifa = (*ifa->ifa_getifa)(ifa, rt_getkey(rt));
    187 		rt_replace_ifa(rt, ifa);
    188 		return ifa;
    189 	}
    190 }
    191 
    192 static void
    193 rt_set_ifa1(struct rtentry *rt, struct ifaddr *ifa)
    194 {
    195 	rt->rt_ifa = ifa;
    196 	if (ifa->ifa_seqno != NULL)
    197 		rt->rt_ifa_seqno = *ifa->ifa_seqno;
    198 }
    199 
    200 /*
    201  * Is this route the connected route for the ifa?
    202  */
    203 static int
    204 rt_ifa_connected(const struct rtentry *rt, const struct ifaddr *ifa)
    205 {
    206 	const struct sockaddr *key, *dst, *odst;
    207 	struct sockaddr_storage maskeddst;
    208 
    209 	key = rt_getkey(rt);
    210 	dst = rt->rt_flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
    211 	if (dst == NULL ||
    212 	    dst->sa_family != key->sa_family ||
    213 	    dst->sa_len != key->sa_len)
    214 		return 0;
    215 	if ((rt->rt_flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
    216 		odst = dst;
    217 		dst = (struct sockaddr *)&maskeddst;
    218 		rt_maskedcopy(odst, (struct sockaddr *)&maskeddst,
    219 		    ifa->ifa_netmask);
    220 	}
    221 	return (memcmp(dst, key, dst->sa_len) == 0);
    222 }
    223 
    224 void
    225 rt_replace_ifa(struct rtentry *rt, struct ifaddr *ifa)
    226 {
    227 	if (rt->rt_ifa &&
    228 	    rt->rt_ifa != ifa &&
    229 	    rt->rt_ifa->ifa_flags & IFA_ROUTE &&
    230 	    rt_ifa_connected(rt, rt->rt_ifa))
    231 	{
    232 		RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    233 		    "replace deleted IFA_ROUTE\n",
    234 		    (void *)rt->_rt_key, (void *)rt->rt_ifa);
    235 		rt->rt_ifa->ifa_flags &= ~IFA_ROUTE;
    236 		if (rt_ifa_connected(rt, ifa)) {
    237 			RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    238 			    "replace added IFA_ROUTE\n",
    239 			    (void *)rt->_rt_key, (void *)ifa);
    240 			ifa->ifa_flags |= IFA_ROUTE;
    241 		}
    242 	}
    243 
    244 	IFAREF(ifa);
    245 	IFAFREE(rt->rt_ifa);
    246 	rt_set_ifa1(rt, ifa);
    247 }
    248 
    249 static void
    250 rt_set_ifa(struct rtentry *rt, struct ifaddr *ifa)
    251 {
    252 	IFAREF(ifa);
    253 	rt_set_ifa1(rt, ifa);
    254 }
    255 
    256 void
    257 rtable_init(void **table)
    258 {
    259 	struct domain *dom;
    260 	DOMAIN_FOREACH(dom)
    261 		if (dom->dom_rtattach)
    262 			dom->dom_rtattach(&table[dom->dom_family],
    263 			    dom->dom_rtoffset);
    264 }
    265 
    266 static int
    267 route_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
    268     void *arg0, void *arg1, void *arg2, void *arg3)
    269 {
    270 	struct rt_msghdr *rtm;
    271 	int result;
    272 
    273 	result = KAUTH_RESULT_DEFER;
    274 	rtm = arg1;
    275 
    276 	if (action != KAUTH_NETWORK_ROUTE)
    277 		return result;
    278 
    279 	if (rtm->rtm_type == RTM_GET)
    280 		result = KAUTH_RESULT_ALLOW;
    281 
    282 	return result;
    283 }
    284 
    285 void
    286 route_init(void)
    287 {
    288 
    289 #ifdef RTFLUSH_DEBUG
    290 	sysctl_net_rtcache_setup(NULL);
    291 #endif
    292 
    293 	pool_init(&rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl",
    294 	    NULL, IPL_SOFTNET);
    295 	pool_init(&rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl",
    296 	    NULL, IPL_SOFTNET);
    297 
    298 	rt_init();
    299 	rn_init();	/* initialize all zeroes, all ones, mask table */
    300 	rtable_init((void **)rt_tables);
    301 
    302 	route_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
    303 	    route_listener_cb, NULL);
    304 }
    305 
    306 void
    307 rtflushall(int family)
    308 {
    309 	struct domain *dom;
    310 
    311 	if (rtcache_debug())
    312 		printf("%s: enter\n", __func__);
    313 
    314 	if ((dom = pffinddomain(family)) == NULL)
    315 		return;
    316 
    317 	rtcache_invalidate(&dom->dom_rtcache);
    318 }
    319 
    320 void
    321 rtcache(struct route *ro)
    322 {
    323 	struct domain *dom;
    324 
    325 	rtcache_invariants(ro);
    326 	KASSERT(ro->_ro_rt != NULL);
    327 	KASSERT(ro->ro_invalid == false);
    328 	KASSERT(rtcache_getdst(ro) != NULL);
    329 
    330 	if ((dom = pffinddomain(rtcache_getdst(ro)->sa_family)) == NULL)
    331 		return;
    332 
    333 	LIST_INSERT_HEAD(&dom->dom_rtcache, ro, ro_rtcache_next);
    334 	rtcache_invariants(ro);
    335 }
    336 
    337 /*
    338  * Packet routing routines.
    339  */
    340 struct rtentry *
    341 rtalloc1(const struct sockaddr *dst, int report)
    342 {
    343 	struct radix_node_head *rnh = rt_tables[dst->sa_family];
    344 	struct rtentry *rt;
    345 	struct radix_node *rn;
    346 	struct rtentry *newrt = NULL;
    347 	struct rt_addrinfo info;
    348 	int  s = splsoftnet(), err = 0, msgtype = RTM_MISS;
    349 
    350 	if (rnh && (rn = rnh->rnh_matchaddr(dst, rnh)) &&
    351 	    ((rn->rn_flags & RNF_ROOT) == 0)) {
    352 		newrt = rt = (struct rtentry *)rn;
    353 		if (report && (rt->rt_flags & RTF_CLONING)) {
    354 			err = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
    355 			    &newrt);
    356 			if (err) {
    357 				newrt = rt;
    358 				rt->rt_refcnt++;
    359 				goto miss;
    360 			}
    361 			KASSERT(newrt != NULL);
    362 			if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) {
    363 				msgtype = RTM_RESOLVE;
    364 				goto miss;
    365 			}
    366 			/* Inform listeners of the new route */
    367 			memset(&info, 0, sizeof(info));
    368 			info.rti_info[RTAX_DST] = rt_getkey(rt);
    369 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    370 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    371 			if (rt->rt_ifp != NULL) {
    372 				info.rti_info[RTAX_IFP] =
    373 				    rt->rt_ifp->if_dl->ifa_addr;
    374 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
    375 			}
    376 			rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0);
    377 		} else
    378 			rt->rt_refcnt++;
    379 	} else {
    380 		rtstat.rts_unreach++;
    381 	miss:	if (report) {
    382 			memset((void *)&info, 0, sizeof(info));
    383 			info.rti_info[RTAX_DST] = dst;
    384 			rt_missmsg(msgtype, &info, 0, err);
    385 		}
    386 	}
    387 	splx(s);
    388 	return newrt;
    389 }
    390 
    391 void
    392 rtfree(struct rtentry *rt)
    393 {
    394 	struct ifaddr *ifa;
    395 
    396 	if (rt == NULL)
    397 		panic("rtfree");
    398 	rt->rt_refcnt--;
    399 	if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
    400 		if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
    401 			panic ("rtfree 2");
    402 		rttrash--;
    403 		if (rt->rt_refcnt < 0) {
    404 			printf("rtfree: %p not freed (neg refs)\n", rt);
    405 			return;
    406 		}
    407 		rt_timer_remove_all(rt, 0);
    408 		ifa = rt->rt_ifa;
    409 		rt->rt_ifa = NULL;
    410 		IFAFREE(ifa);
    411 		rt->rt_ifp = NULL;
    412 		rt_destroy(rt);
    413 		pool_put(&rtentry_pool, rt);
    414 	}
    415 }
    416 
    417 void
    418 ifafree(struct ifaddr *ifa)
    419 {
    420 
    421 #ifdef DIAGNOSTIC
    422 	if (ifa == NULL)
    423 		panic("ifafree: null ifa");
    424 	if (ifa->ifa_refcnt != 0)
    425 		panic("ifafree: ifa_refcnt != 0 (%d)", ifa->ifa_refcnt);
    426 #endif
    427 #ifdef IFAREF_DEBUG
    428 	printf("ifafree: freeing ifaddr %p\n", ifa);
    429 #endif
    430 	free(ifa, M_IFADDR);
    431 }
    432 
    433 /*
    434  * Force a routing table entry to the specified
    435  * destination to go through the given gateway.
    436  * Normally called as a result of a routing redirect
    437  * message from the network layer.
    438  *
    439  * N.B.: must be called at splsoftnet
    440  */
    441 void
    442 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
    443 	const struct sockaddr *netmask, int flags, const struct sockaddr *src,
    444 	struct rtentry **rtp)
    445 {
    446 	struct rtentry *rt;
    447 	int error = 0;
    448 	uint64_t *stat = NULL;
    449 	struct rt_addrinfo info;
    450 	struct ifaddr *ifa;
    451 
    452 	/* verify the gateway is directly reachable */
    453 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
    454 		error = ENETUNREACH;
    455 		goto out;
    456 	}
    457 	rt = rtalloc1(dst, 0);
    458 	/*
    459 	 * If the redirect isn't from our current router for this dst,
    460 	 * it's either old or wrong.  If it redirects us to ourselves,
    461 	 * we have a routing loop, perhaps as a result of an interface
    462 	 * going down recently.
    463 	 */
    464 	if (!(flags & RTF_DONE) && rt &&
    465 	     (sockaddr_cmp(src, rt->rt_gateway) != 0 || rt->rt_ifa != ifa))
    466 		error = EINVAL;
    467 	else if (ifa_ifwithaddr(gateway))
    468 		error = EHOSTUNREACH;
    469 	if (error)
    470 		goto done;
    471 	/*
    472 	 * Create a new entry if we just got back a wildcard entry
    473 	 * or the lookup failed.  This is necessary for hosts
    474 	 * which use routing redirects generated by smart gateways
    475 	 * to dynamically build the routing tables.
    476 	 */
    477 	if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
    478 		goto create;
    479 	/*
    480 	 * Don't listen to the redirect if it's
    481 	 * for a route to an interface.
    482 	 */
    483 	if (rt->rt_flags & RTF_GATEWAY) {
    484 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
    485 			/*
    486 			 * Changing from route to net => route to host.
    487 			 * Create new route, rather than smashing route to net.
    488 			 */
    489 		create:
    490 			if (rt != NULL)
    491 				rtfree(rt);
    492 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
    493 			info.rti_info[RTAX_DST] = dst;
    494 			info.rti_info[RTAX_GATEWAY] = gateway;
    495 			info.rti_info[RTAX_NETMASK] = netmask;
    496 			info.rti_ifa = ifa;
    497 			info.rti_flags = flags;
    498 			rt = NULL;
    499 			error = rtrequest1(RTM_ADD, &info, &rt);
    500 			if (rt != NULL)
    501 				flags = rt->rt_flags;
    502 			stat = &rtstat.rts_dynamic;
    503 		} else {
    504 			/*
    505 			 * Smash the current notion of the gateway to
    506 			 * this destination.  Should check about netmask!!!
    507 			 */
    508 			rt->rt_flags |= RTF_MODIFIED;
    509 			flags |= RTF_MODIFIED;
    510 			stat = &rtstat.rts_newgateway;
    511 			rt_setgate(rt, gateway);
    512 		}
    513 	} else
    514 		error = EHOSTUNREACH;
    515 done:
    516 	if (rt) {
    517 		if (rtp != NULL && !error)
    518 			*rtp = rt;
    519 		else
    520 			rtfree(rt);
    521 	}
    522 out:
    523 	if (error)
    524 		rtstat.rts_badredirect++;
    525 	else if (stat != NULL)
    526 		(*stat)++;
    527 	memset(&info, 0, sizeof(info));
    528 	info.rti_info[RTAX_DST] = dst;
    529 	info.rti_info[RTAX_GATEWAY] = gateway;
    530 	info.rti_info[RTAX_NETMASK] = netmask;
    531 	info.rti_info[RTAX_AUTHOR] = src;
    532 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
    533 }
    534 
    535 /*
    536  * Delete a route and generate a message
    537  */
    538 static int
    539 rtdeletemsg(struct rtentry *rt)
    540 {
    541 	int error;
    542 	struct rt_addrinfo info;
    543 
    544 	/*
    545 	 * Request the new route so that the entry is not actually
    546 	 * deleted.  That will allow the information being reported to
    547 	 * be accurate (and consistent with route_output()).
    548 	 */
    549 	memset(&info, 0, sizeof(info));
    550 	info.rti_info[RTAX_DST] = rt_getkey(rt);
    551 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    552 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    553 	info.rti_flags = rt->rt_flags;
    554 	error = rtrequest1(RTM_DELETE, &info, &rt);
    555 
    556 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
    557 
    558 	/* Adjust the refcount */
    559 	if (error == 0 && rt->rt_refcnt <= 0) {
    560 		rt->rt_refcnt++;
    561 		rtfree(rt);
    562 	}
    563 	return error;
    564 }
    565 
    566 static int
    567 rtflushclone1(struct rtentry *rt, void *arg)
    568 {
    569 	struct rtentry *parent;
    570 
    571 	parent = (struct rtentry *)arg;
    572 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
    573 		rtdeletemsg(rt);
    574 	return 0;
    575 }
    576 
    577 static void
    578 rtflushclone(sa_family_t family, struct rtentry *parent)
    579 {
    580 
    581 #ifdef DIAGNOSTIC
    582 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
    583 		panic("rtflushclone: called with a non-cloning route");
    584 #endif
    585 	rt_walktree(family, rtflushclone1, (void *)parent);
    586 }
    587 
    588 /*
    589  * Routing table ioctl interface.
    590  */
    591 int
    592 rtioctl(u_long req, void *data, struct lwp *l)
    593 {
    594 	return EOPNOTSUPP;
    595 }
    596 
    597 struct ifaddr *
    598 ifa_ifwithroute(int flags, const struct sockaddr *dst,
    599 	const struct sockaddr *gateway)
    600 {
    601 	struct ifaddr *ifa;
    602 	if ((flags & RTF_GATEWAY) == 0) {
    603 		/*
    604 		 * If we are adding a route to an interface,
    605 		 * and the interface is a pt to pt link
    606 		 * we should search for the destination
    607 		 * as our clue to the interface.  Otherwise
    608 		 * we can use the local address.
    609 		 */
    610 		ifa = NULL;
    611 		if (flags & RTF_HOST)
    612 			ifa = ifa_ifwithdstaddr(dst);
    613 		if (ifa == NULL)
    614 			ifa = ifa_ifwithaddr(gateway);
    615 	} else {
    616 		/*
    617 		 * If we are adding a route to a remote net
    618 		 * or host, the gateway may still be on the
    619 		 * other end of a pt to pt link.
    620 		 */
    621 		ifa = ifa_ifwithdstaddr(gateway);
    622 	}
    623 	if (ifa == NULL)
    624 		ifa = ifa_ifwithnet(gateway);
    625 	if (ifa == NULL) {
    626 		struct rtentry *rt = rtalloc1(dst, 0);
    627 		if (rt == NULL)
    628 			return NULL;
    629 		rt->rt_refcnt--;
    630 		if ((ifa = rt->rt_ifa) == NULL)
    631 			return NULL;
    632 	}
    633 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
    634 		struct ifaddr *oifa = ifa;
    635 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
    636 		if (ifa == 0)
    637 			ifa = oifa;
    638 	}
    639 	return ifa;
    640 }
    641 
    642 int
    643 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
    644 	const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
    645 {
    646 	struct rt_addrinfo info;
    647 
    648 	memset(&info, 0, sizeof(info));
    649 	info.rti_flags = flags;
    650 	info.rti_info[RTAX_DST] = dst;
    651 	info.rti_info[RTAX_GATEWAY] = gateway;
    652 	info.rti_info[RTAX_NETMASK] = netmask;
    653 	return rtrequest1(req, &info, ret_nrt);
    654 }
    655 
    656 int
    657 rt_getifa(struct rt_addrinfo *info)
    658 {
    659 	struct ifaddr *ifa;
    660 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
    661 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
    662 	const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
    663 	const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
    664 	int flags = info->rti_flags;
    665 
    666 	/*
    667 	 * ifp may be specified by sockaddr_dl when protocol address
    668 	 * is ambiguous
    669 	 */
    670 	if (info->rti_ifp == NULL && ifpaddr != NULL
    671 	    && ifpaddr->sa_family == AF_LINK &&
    672 	    (ifa = ifa_ifwithnet(ifpaddr)) != NULL)
    673 		info->rti_ifp = ifa->ifa_ifp;
    674 	if (info->rti_ifa == NULL && ifaaddr != NULL)
    675 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
    676 	if (info->rti_ifa == NULL) {
    677 		const struct sockaddr *sa;
    678 
    679 		sa = ifaaddr != NULL ? ifaaddr :
    680 		    (gateway != NULL ? gateway : dst);
    681 		if (sa != NULL && info->rti_ifp != NULL)
    682 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
    683 		else if (dst != NULL && gateway != NULL)
    684 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
    685 		else if (sa != NULL)
    686 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
    687 	}
    688 	if ((ifa = info->rti_ifa) == NULL)
    689 		return ENETUNREACH;
    690 	if (ifa->ifa_getifa != NULL)
    691 		info->rti_ifa = ifa = (*ifa->ifa_getifa)(ifa, dst);
    692 	if (info->rti_ifp == NULL)
    693 		info->rti_ifp = ifa->ifa_ifp;
    694 	return 0;
    695 }
    696 
    697 int
    698 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
    699 {
    700 	int s = splsoftnet();
    701 	int error = 0;
    702 	struct rtentry *rt, *crt;
    703 	struct radix_node *rn;
    704 	struct radix_node_head *rnh;
    705 	struct ifaddr *ifa;
    706 	struct sockaddr_storage maskeddst;
    707 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
    708 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
    709 	const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
    710 	int flags = info->rti_flags;
    711 #define senderr(x) { error = x ; goto bad; }
    712 
    713 	if ((rnh = rt_tables[dst->sa_family]) == NULL)
    714 		senderr(ESRCH);
    715 	if (flags & RTF_HOST)
    716 		netmask = NULL;
    717 	switch (req) {
    718 	case RTM_DELETE:
    719 		if (netmask) {
    720 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    721 			    netmask);
    722 			dst = (struct sockaddr *)&maskeddst;
    723 		}
    724 		if ((rn = rnh->rnh_lookup(dst, netmask, rnh)) == NULL)
    725 			senderr(ESRCH);
    726 		rt = (struct rtentry *)rn;
    727 		if ((rt->rt_flags & RTF_CLONING) != 0) {
    728 			/* clean up any cloned children */
    729 			rtflushclone(dst->sa_family, rt);
    730 		}
    731 		if ((rn = rnh->rnh_deladdr(dst, netmask, rnh)) == NULL)
    732 			senderr(ESRCH);
    733 		if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
    734 			panic ("rtrequest delete");
    735 		rt = (struct rtentry *)rn;
    736 		if (rt->rt_gwroute) {
    737 			RTFREE(rt->rt_gwroute);
    738 			rt->rt_gwroute = NULL;
    739 		}
    740 		if (rt->rt_parent) {
    741 			rt->rt_parent->rt_refcnt--;
    742 			rt->rt_parent = NULL;
    743 		}
    744 		rt->rt_flags &= ~RTF_UP;
    745 		if ((ifa = rt->rt_ifa)) {
    746 			if (ifa->ifa_flags & IFA_ROUTE &&
    747 			    rt_ifa_connected(rt, ifa)) {
    748 				RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
    749 				    "deleted IFA_ROUTE\n",
    750 				    (void *)rt->_rt_key, (void *)ifa);
    751 				ifa->ifa_flags &= ~IFA_ROUTE;
    752 			}
    753 			if (ifa->ifa_rtrequest)
    754 				ifa->ifa_rtrequest(RTM_DELETE, rt, info);
    755 		}
    756 		rttrash++;
    757 		if (ret_nrt)
    758 			*ret_nrt = rt;
    759 		else if (rt->rt_refcnt <= 0) {
    760 			rt->rt_refcnt++;
    761 			rtfree(rt);
    762 		}
    763 		break;
    764 
    765 	case RTM_RESOLVE:
    766 		if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
    767 			senderr(EINVAL);
    768 		if ((rt->rt_flags & RTF_CLONING) == 0)
    769 			senderr(EINVAL);
    770 		ifa = rt->rt_ifa;
    771 		flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC);
    772 		flags |= RTF_CLONED;
    773 		gateway = rt->rt_gateway;
    774 		flags |= RTF_HOST;
    775 		goto makeroute;
    776 
    777 	case RTM_ADD:
    778 		if (info->rti_ifa == NULL && (error = rt_getifa(info)))
    779 			senderr(error);
    780 		ifa = info->rti_ifa;
    781 	makeroute:
    782 		/* Already at splsoftnet() so pool_get/pool_put are safe */
    783 		rt = pool_get(&rtentry_pool, PR_NOWAIT);
    784 		if (rt == NULL)
    785 			senderr(ENOBUFS);
    786 		memset(rt, 0, sizeof(*rt));
    787 		rt->rt_flags = RTF_UP | flags;
    788 		LIST_INIT(&rt->rt_timer);
    789 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    790 		if (rt_setkey(rt, dst, M_NOWAIT) == NULL ||
    791 		    rt_setgate(rt, gateway) != 0) {
    792 			pool_put(&rtentry_pool, rt);
    793 			senderr(ENOBUFS);
    794 		}
    795 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    796 		if (netmask) {
    797 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    798 			    netmask);
    799 			rt_setkey(rt, (struct sockaddr *)&maskeddst, M_NOWAIT);
    800 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    801 		} else {
    802 			rt_setkey(rt, dst, M_NOWAIT);
    803 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    804 		}
    805 		rt_set_ifa(rt, ifa);
    806 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    807 		rt->rt_ifp = ifa->ifa_ifp;
    808 		if (req == RTM_RESOLVE) {
    809 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
    810 			rt->rt_parent = *ret_nrt;
    811 			rt->rt_parent->rt_refcnt++;
    812 		}
    813 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    814 		rn = rnh->rnh_addaddr(rt_getkey(rt), netmask, rnh,
    815 		    rt->rt_nodes);
    816 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    817 		if (rn == NULL && (crt = rtalloc1(rt_getkey(rt), 0)) != NULL) {
    818 			/* overwrite cloned route */
    819 			if ((crt->rt_flags & RTF_CLONED) != 0) {
    820 				rtdeletemsg(crt);
    821 				rn = rnh->rnh_addaddr(rt_getkey(rt),
    822 				    netmask, rnh, rt->rt_nodes);
    823 			}
    824 			RTFREE(crt);
    825 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    826 		}
    827 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    828 		if (rn == NULL) {
    829 			IFAFREE(ifa);
    830 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
    831 				rtfree(rt->rt_parent);
    832 			if (rt->rt_gwroute)
    833 				rtfree(rt->rt_gwroute);
    834 			rt_destroy(rt);
    835 			pool_put(&rtentry_pool, rt);
    836 			senderr(EEXIST);
    837 		}
    838 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    839 		if (ifa->ifa_rtrequest)
    840 			ifa->ifa_rtrequest(req, rt, info);
    841 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    842 		if (ret_nrt) {
    843 			*ret_nrt = rt;
    844 			rt->rt_refcnt++;
    845 		}
    846 		if ((rt->rt_flags & RTF_CLONING) != 0) {
    847 			/* clean up any cloned children */
    848 			rtflushclone(dst->sa_family, rt);
    849 		}
    850 		rtflushall(dst->sa_family);
    851 		break;
    852 	case RTM_GET:
    853 		if (netmask != NULL) {
    854 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
    855 			    netmask);
    856 			dst = (struct sockaddr *)&maskeddst;
    857 		}
    858 		rn = rnh->rnh_lookup(dst, netmask, rnh);
    859 		if (rn == NULL || (rn->rn_flags & RNF_ROOT) != 0)
    860 			senderr(ESRCH);
    861 		if (ret_nrt != NULL) {
    862 			rt = (struct rtentry *)rn;
    863 			*ret_nrt = rt;
    864 			rt->rt_refcnt++;
    865 		}
    866 		break;
    867 	}
    868 bad:
    869 	splx(s);
    870 	return error;
    871 }
    872 
    873 int
    874 rt_setgate(struct rtentry *rt, const struct sockaddr *gate)
    875 {
    876 	KASSERT(rt != rt->rt_gwroute);
    877 
    878 	KASSERT(rt->_rt_key != NULL);
    879 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    880 
    881 	if (rt->rt_gwroute) {
    882 		RTFREE(rt->rt_gwroute);
    883 		rt->rt_gwroute = NULL;
    884 	}
    885 	KASSERT(rt->_rt_key != NULL);
    886 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    887 	if (rt->rt_gateway != NULL)
    888 		sockaddr_free(rt->rt_gateway);
    889 	KASSERT(rt->_rt_key != NULL);
    890 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    891 	if ((rt->rt_gateway = sockaddr_dup(gate, M_NOWAIT)) == NULL)
    892 		return ENOMEM;
    893 	KASSERT(rt->_rt_key != NULL);
    894 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    895 
    896 	if (rt->rt_flags & RTF_GATEWAY) {
    897 		KASSERT(rt->_rt_key != NULL);
    898 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    899 		rt->rt_gwroute = rtalloc1(gate, 1);
    900 		/*
    901 		 * If we switched gateways, grab the MTU from the new
    902 		 * gateway route if the current MTU, if the current MTU is
    903 		 * greater than the MTU of gateway.
    904 		 * Note that, if the MTU of gateway is 0, we will reset the
    905 		 * MTU of the route to run PMTUD again from scratch. XXX
    906 		 */
    907 		KASSERT(rt->_rt_key != NULL);
    908 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    909 		if (rt->rt_gwroute
    910 		    && !(rt->rt_rmx.rmx_locks & RTV_MTU)
    911 		    && rt->rt_rmx.rmx_mtu
    912 		    && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
    913 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
    914 		}
    915 	}
    916 	KASSERT(rt->_rt_key != NULL);
    917 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
    918 	return 0;
    919 }
    920 
    921 void
    922 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
    923 	const struct sockaddr *netmask)
    924 {
    925 	const char *netmaskp = &netmask->sa_data[0],
    926 	           *srcp = &src->sa_data[0];
    927 	char *dstp = &dst->sa_data[0];
    928 	const char *maskend = dstp + MIN(netmask->sa_len, src->sa_len);
    929 	const char *srcend = dstp + src->sa_len;
    930 
    931 	dst->sa_len = src->sa_len;
    932 	dst->sa_family = src->sa_family;
    933 
    934 	while (dstp < maskend)
    935 		*dstp++ = *srcp++ & *netmaskp++;
    936 	if (dstp < srcend)
    937 		memset(dstp, 0, (size_t)(srcend - dstp));
    938 }
    939 
    940 /*
    941  * Set up or tear down a routing table entry, normally
    942  * for an interface.
    943  */
    944 int
    945 rtinit(struct ifaddr *ifa, int cmd, int flags)
    946 {
    947 	struct rtentry *rt;
    948 	struct sockaddr *dst, *odst;
    949 	struct sockaddr_storage maskeddst;
    950 	struct rtentry *nrt = NULL;
    951 	int error;
    952 	struct rt_addrinfo info;
    953 	struct sockaddr_dl *sdl;
    954 	const struct sockaddr_dl *ifsdl;
    955 
    956 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
    957 	if (cmd == RTM_DELETE) {
    958 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
    959 			/* Delete subnet route for this interface */
    960 			odst = dst;
    961 			dst = (struct sockaddr *)&maskeddst;
    962 			rt_maskedcopy(odst, dst, ifa->ifa_netmask);
    963 		}
    964 		if ((rt = rtalloc1(dst, 0)) != NULL) {
    965 			rt->rt_refcnt--;
    966 			if (rt->rt_ifa != ifa)
    967 				return (flags & RTF_HOST) ? EHOSTUNREACH
    968 							: ENETUNREACH;
    969 		}
    970 	}
    971 	memset(&info, 0, sizeof(info));
    972 	info.rti_ifa = ifa;
    973 	info.rti_flags = flags | ifa->ifa_flags;
    974 	info.rti_info[RTAX_DST] = dst;
    975 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
    976 	/*
    977 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
    978 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
    979 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
    980 	 * change it to meet bsdi4 behavior.
    981 	 */
    982 	if (cmd != RTM_LLINFO_UPD)
    983 		info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
    984 	error = rtrequest1((cmd == RTM_LLINFO_UPD) ? RTM_GET : cmd, &info,
    985 	    &nrt);
    986 	if (error != 0 || (rt = nrt) == NULL)
    987 		;
    988 	else switch (cmd) {
    989 	case RTM_DELETE:
    990 		rt_newaddrmsg(cmd, ifa, error, nrt);
    991 		if (rt->rt_refcnt <= 0) {
    992 			rt->rt_refcnt++;
    993 			rtfree(rt);
    994 		}
    995 		break;
    996 	case RTM_LLINFO_UPD:
    997 		rt->rt_refcnt--;
    998 		RT_DPRINTF("%s: updating%s\n", __func__,
    999 		    ((rt->rt_flags & RTF_LLINFO) == 0) ? " (no llinfo)" : "");
   1000 
   1001 		ifsdl = ifa->ifa_ifp->if_sadl;
   1002 
   1003 		if ((rt->rt_flags & RTF_LLINFO) != 0 &&
   1004 		    (sdl = satosdl(rt->rt_gateway)) != NULL &&
   1005 		    sdl->sdl_family == AF_LINK &&
   1006 		    sockaddr_dl_setaddr(sdl, sdl->sdl_len, CLLADDR(ifsdl),
   1007 		                        ifa->ifa_ifp->if_addrlen) == NULL) {
   1008 			error = EINVAL;
   1009 			break;
   1010 		}
   1011 
   1012 		if (cmd == RTM_LLINFO_UPD && ifa->ifa_rtrequest != NULL)
   1013 			ifa->ifa_rtrequest(RTM_LLINFO_UPD, rt, &info);
   1014 		rt_newaddrmsg(RTM_CHANGE, ifa, error, nrt);
   1015 		break;
   1016 	case RTM_ADD:
   1017 		rt->rt_refcnt--;
   1018 		if (rt->rt_ifa != ifa) {
   1019 			printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
   1020 				rt->rt_ifa);
   1021 			if (rt->rt_ifa->ifa_rtrequest != NULL) {
   1022 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
   1023 				    &info);
   1024 			}
   1025 			rt_replace_ifa(rt, ifa);
   1026 			rt->rt_ifp = ifa->ifa_ifp;
   1027 			if (ifa->ifa_rtrequest != NULL)
   1028 				ifa->ifa_rtrequest(RTM_ADD, rt, &info);
   1029 		}
   1030 		rt_newaddrmsg(cmd, ifa, error, nrt);
   1031 		break;
   1032 	}
   1033 	return error;
   1034 }
   1035 
   1036 /*
   1037  * Route timer routines.  These routes allow functions to be called
   1038  * for various routes at any time.  This is useful in supporting
   1039  * path MTU discovery and redirect route deletion.
   1040  *
   1041  * This is similar to some BSDI internal functions, but it provides
   1042  * for multiple queues for efficiency's sake...
   1043  */
   1044 
   1045 LIST_HEAD(, rttimer_queue) rttimer_queue_head;
   1046 static int rt_init_done = 0;
   1047 
   1048 #define RTTIMER_CALLOUT(r)	do {					\
   1049 		if (r->rtt_func != NULL) {				\
   1050 			(*r->rtt_func)(r->rtt_rt, r);			\
   1051 		} else {						\
   1052 			rtrequest((int) RTM_DELETE,			\
   1053 				  rt_getkey(r->rtt_rt),			\
   1054 				  0, 0, 0, 0);				\
   1055 		}							\
   1056 	} while (/*CONSTCOND*/0)
   1057 
   1058 /*
   1059  * Some subtle order problems with domain initialization mean that
   1060  * we cannot count on this being run from rt_init before various
   1061  * protocol initializations are done.  Therefore, we make sure
   1062  * that this is run when the first queue is added...
   1063  */
   1064 
   1065 void
   1066 rt_timer_init(void)
   1067 {
   1068 	assert(rt_init_done == 0);
   1069 
   1070 	LIST_INIT(&rttimer_queue_head);
   1071 	callout_init(&rt_timer_ch, 0);
   1072 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
   1073 	rt_init_done = 1;
   1074 }
   1075 
   1076 struct rttimer_queue *
   1077 rt_timer_queue_create(u_int timeout)
   1078 {
   1079 	struct rttimer_queue *rtq;
   1080 
   1081 	if (rt_init_done == 0)
   1082 		rt_timer_init();
   1083 
   1084 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
   1085 	if (rtq == NULL)
   1086 		return NULL;
   1087 	memset(rtq, 0, sizeof(*rtq));
   1088 
   1089 	rtq->rtq_timeout = timeout;
   1090 	TAILQ_INIT(&rtq->rtq_head);
   1091 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
   1092 
   1093 	return rtq;
   1094 }
   1095 
   1096 void
   1097 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
   1098 {
   1099 
   1100 	rtq->rtq_timeout = timeout;
   1101 }
   1102 
   1103 void
   1104 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy)
   1105 {
   1106 	struct rttimer *r;
   1107 
   1108 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
   1109 		LIST_REMOVE(r, rtt_link);
   1110 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
   1111 		if (destroy)
   1112 			RTTIMER_CALLOUT(r);
   1113 		/* we are already at splsoftnet */
   1114 		pool_put(&rttimer_pool, r);
   1115 		if (rtq->rtq_count > 0)
   1116 			rtq->rtq_count--;
   1117 		else
   1118 			printf("rt_timer_queue_remove_all: "
   1119 			    "rtq_count reached 0\n");
   1120 	}
   1121 }
   1122 
   1123 void
   1124 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
   1125 {
   1126 
   1127 	rt_timer_queue_remove_all(rtq, destroy);
   1128 
   1129 	LIST_REMOVE(rtq, rtq_link);
   1130 
   1131 	/*
   1132 	 * Caller is responsible for freeing the rttimer_queue structure.
   1133 	 */
   1134 }
   1135 
   1136 unsigned long
   1137 rt_timer_count(struct rttimer_queue *rtq)
   1138 {
   1139 	return rtq->rtq_count;
   1140 }
   1141 
   1142 void
   1143 rt_timer_remove_all(struct rtentry *rt, int destroy)
   1144 {
   1145 	struct rttimer *r;
   1146 
   1147 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
   1148 		LIST_REMOVE(r, rtt_link);
   1149 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
   1150 		if (destroy)
   1151 			RTTIMER_CALLOUT(r);
   1152 		if (r->rtt_queue->rtq_count > 0)
   1153 			r->rtt_queue->rtq_count--;
   1154 		else
   1155 			printf("rt_timer_remove_all: rtq_count reached 0\n");
   1156 		/* we are already at splsoftnet */
   1157 		pool_put(&rttimer_pool, r);
   1158 	}
   1159 }
   1160 
   1161 int
   1162 rt_timer_add(struct rtentry *rt,
   1163 	void (*func)(struct rtentry *, struct rttimer *),
   1164 	struct rttimer_queue *queue)
   1165 {
   1166 	struct rttimer *r;
   1167 	int s;
   1168 
   1169 	/*
   1170 	 * If there's already a timer with this action, destroy it before
   1171 	 * we add a new one.
   1172 	 */
   1173 	LIST_FOREACH(r, &rt->rt_timer, rtt_link) {
   1174 		if (r->rtt_func == func)
   1175 			break;
   1176 	}
   1177 	if (r != NULL) {
   1178 		LIST_REMOVE(r, rtt_link);
   1179 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
   1180 		if (r->rtt_queue->rtq_count > 0)
   1181 			r->rtt_queue->rtq_count--;
   1182 		else
   1183 			printf("rt_timer_add: rtq_count reached 0\n");
   1184 	} else {
   1185 		s = splsoftnet();
   1186 		r = pool_get(&rttimer_pool, PR_NOWAIT);
   1187 		splx(s);
   1188 		if (r == NULL)
   1189 			return ENOBUFS;
   1190 	}
   1191 
   1192 	memset(r, 0, sizeof(*r));
   1193 
   1194 	r->rtt_rt = rt;
   1195 	r->rtt_time = time_uptime;
   1196 	r->rtt_func = func;
   1197 	r->rtt_queue = queue;
   1198 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
   1199 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
   1200 	r->rtt_queue->rtq_count++;
   1201 
   1202 	return 0;
   1203 }
   1204 
   1205 /* ARGSUSED */
   1206 void
   1207 rt_timer_timer(void *arg)
   1208 {
   1209 	struct rttimer_queue *rtq;
   1210 	struct rttimer *r;
   1211 	int s;
   1212 
   1213 	s = splsoftnet();
   1214 	LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) {
   1215 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
   1216 		    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
   1217 			LIST_REMOVE(r, rtt_link);
   1218 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
   1219 			RTTIMER_CALLOUT(r);
   1220 			pool_put(&rttimer_pool, r);
   1221 			if (rtq->rtq_count > 0)
   1222 				rtq->rtq_count--;
   1223 			else
   1224 				printf("rt_timer_timer: rtq_count reached 0\n");
   1225 		}
   1226 	}
   1227 	splx(s);
   1228 
   1229 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
   1230 }
   1231 
   1232 static struct rtentry *
   1233 _rtcache_init(struct route *ro, int flag)
   1234 {
   1235 	rtcache_invariants(ro);
   1236 	KASSERT(ro->_ro_rt == NULL);
   1237 
   1238 	if (rtcache_getdst(ro) == NULL)
   1239 		return NULL;
   1240 	ro->ro_invalid = false;
   1241 	if ((ro->_ro_rt = rtalloc1(rtcache_getdst(ro), flag)) != NULL)
   1242 		rtcache(ro);
   1243 
   1244 	rtcache_invariants(ro);
   1245 	return ro->_ro_rt;
   1246 }
   1247 
   1248 struct rtentry *
   1249 rtcache_init(struct route *ro)
   1250 {
   1251 	return _rtcache_init(ro, 1);
   1252 }
   1253 
   1254 struct rtentry *
   1255 rtcache_init_noclone(struct route *ro)
   1256 {
   1257 	return _rtcache_init(ro, 0);
   1258 }
   1259 
   1260 struct rtentry *
   1261 rtcache_update(struct route *ro, int clone)
   1262 {
   1263 	rtcache_clear(ro);
   1264 	return _rtcache_init(ro, clone);
   1265 }
   1266 
   1267 void
   1268 rtcache_copy(struct route *new_ro, const struct route *old_ro)
   1269 {
   1270 	struct rtentry *rt;
   1271 
   1272 	KASSERT(new_ro != old_ro);
   1273 	rtcache_invariants(new_ro);
   1274 	rtcache_invariants(old_ro);
   1275 
   1276 	if ((rt = rtcache_validate(old_ro)) != NULL)
   1277 		rt->rt_refcnt++;
   1278 
   1279 	if (rtcache_getdst(old_ro) == NULL ||
   1280 	    rtcache_setdst(new_ro, rtcache_getdst(old_ro)) != 0)
   1281 		return;
   1282 
   1283 	new_ro->ro_invalid = false;
   1284 	if ((new_ro->_ro_rt = rt) != NULL)
   1285 		rtcache(new_ro);
   1286 	rtcache_invariants(new_ro);
   1287 }
   1288 
   1289 static struct dom_rtlist invalid_routes = LIST_HEAD_INITIALIZER(dom_rtlist);
   1290 
   1291 void
   1292 rtcache_invalidate(struct dom_rtlist *rtlist)
   1293 {
   1294 	struct route *ro;
   1295 
   1296 	while ((ro = LIST_FIRST(rtlist)) != NULL) {
   1297 		rtcache_invariants(ro);
   1298 		KASSERT(ro->_ro_rt != NULL);
   1299 		ro->ro_invalid = true;
   1300 		LIST_REMOVE(ro, ro_rtcache_next);
   1301 		LIST_INSERT_HEAD(&invalid_routes, ro, ro_rtcache_next);
   1302 		rtcache_invariants(ro);
   1303 	}
   1304 }
   1305 
   1306 void
   1307 rtcache_clear(struct route *ro)
   1308 {
   1309 	rtcache_invariants(ro);
   1310 	if (ro->_ro_rt == NULL)
   1311 		return;
   1312 
   1313 	LIST_REMOVE(ro, ro_rtcache_next);
   1314 
   1315 	RTFREE(ro->_ro_rt);
   1316 	ro->_ro_rt = NULL;
   1317 	ro->ro_invalid = false;
   1318 	rtcache_invariants(ro);
   1319 }
   1320 
   1321 struct rtentry *
   1322 rtcache_lookup2(struct route *ro, const struct sockaddr *dst, int clone,
   1323     int *hitp)
   1324 {
   1325 	const struct sockaddr *odst;
   1326 	struct rtentry *rt = NULL;
   1327 
   1328 	rtcache_invariants(ro);
   1329 
   1330 	odst = rtcache_getdst(ro);
   1331 
   1332 	if (odst == NULL)
   1333 		;
   1334 	else if (sockaddr_cmp(odst, dst) != 0)
   1335 		rtcache_free(ro);
   1336 	else if ((rt = rtcache_validate(ro)) == NULL)
   1337 		rtcache_clear(ro);
   1338 
   1339 	if (rt == NULL) {
   1340 		*hitp = 0;
   1341 		if (rtcache_setdst(ro, dst) == 0)
   1342 			rt = _rtcache_init(ro, clone);
   1343 	} else
   1344 		*hitp = 1;
   1345 
   1346 	rtcache_invariants(ro);
   1347 
   1348 	return rt;
   1349 }
   1350 
   1351 void
   1352 rtcache_free(struct route *ro)
   1353 {
   1354 	rtcache_clear(ro);
   1355 	if (ro->ro_sa != NULL) {
   1356 		sockaddr_free(ro->ro_sa);
   1357 		ro->ro_sa = NULL;
   1358 	}
   1359 	rtcache_invariants(ro);
   1360 }
   1361 
   1362 int
   1363 rtcache_setdst(struct route *ro, const struct sockaddr *sa)
   1364 {
   1365 	KASSERT(sa != NULL);
   1366 
   1367 	rtcache_invariants(ro);
   1368 	if (ro->ro_sa != NULL && ro->ro_sa->sa_family == sa->sa_family) {
   1369 		rtcache_clear(ro);
   1370 		if (sockaddr_copy(ro->ro_sa, ro->ro_sa->sa_len, sa) != NULL) {
   1371 			rtcache_invariants(ro);
   1372 			return 0;
   1373 		}
   1374 		sockaddr_free(ro->ro_sa);
   1375 	} else if (ro->ro_sa != NULL)
   1376 		rtcache_free(ro);	/* free ro_sa, wrong family */
   1377 
   1378 	KASSERT(ro->_ro_rt == NULL);
   1379 
   1380 	if ((ro->ro_sa = sockaddr_dup(sa, M_NOWAIT)) == NULL) {
   1381 		rtcache_invariants(ro);
   1382 		return ENOMEM;
   1383 	}
   1384 	rtcache_invariants(ro);
   1385 	return 0;
   1386 }
   1387 
   1388 static int
   1389 rt_walktree_visitor(struct radix_node *rn, void *v)
   1390 {
   1391 	struct rtwalk *rw = (struct rtwalk *)v;
   1392 
   1393 	return (*rw->rw_f)((struct rtentry *)rn, rw->rw_v);
   1394 }
   1395 
   1396 int
   1397 rt_walktree(sa_family_t family, int (*f)(struct rtentry *, void *), void *v)
   1398 {
   1399 	struct radix_node_head *rnh = rt_tables[family];
   1400 	struct rtwalk rw;
   1401 
   1402 	if (rnh == NULL)
   1403 		return 0;
   1404 
   1405 	rw.rw_f = f;
   1406 	rw.rw_v = v;
   1407 
   1408 	return rn_walktree(rnh, rt_walktree_visitor, &rw);
   1409 }
   1410