xref: /netbsd-src/sys/net/route.c (revision d11b170b9000ada93db553723522a63d5deac310)
1 /*	$NetBSD: route.c,v 1.130 2014/04/26 11:10:10 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 #include "opt_route.h"
94 
95 #include <sys/cdefs.h>
96 __KERNEL_RCSID(0, "$NetBSD: route.c,v 1.130 2014/04/26 11:10:10 pooka Exp $");
97 
98 #include <sys/param.h>
99 #include <sys/kmem.h>
100 #include <sys/sysctl.h>
101 #include <sys/systm.h>
102 #include <sys/callout.h>
103 #include <sys/proc.h>
104 #include <sys/mbuf.h>
105 #include <sys/socket.h>
106 #include <sys/socketvar.h>
107 #include <sys/domain.h>
108 #include <sys/protosw.h>
109 #include <sys/kernel.h>
110 #include <sys/ioctl.h>
111 #include <sys/pool.h>
112 #include <sys/kauth.h>
113 
114 #include <net/if.h>
115 #include <net/if_dl.h>
116 #include <net/route.h>
117 #include <net/raw_cb.h>
118 
119 #include <netinet/in.h>
120 #include <netinet/in_var.h>
121 
122 #ifdef RTFLUSH_DEBUG
123 #define	rtcache_debug() __predict_false(_rtcache_debug)
124 #else /* RTFLUSH_DEBUG */
125 #define	rtcache_debug() 0
126 #endif /* RTFLUSH_DEBUG */
127 
128 struct	rtstat	rtstat;
129 
130 int	rttrash;		/* routes not in table but not freed */
131 
132 struct pool rtentry_pool;
133 struct pool rttimer_pool;
134 
135 struct callout rt_timer_ch; /* callout for rt_timer_timer() */
136 
137 #ifdef RTFLUSH_DEBUG
138 static int _rtcache_debug = 0;
139 #endif /* RTFLUSH_DEBUG */
140 
141 static kauth_listener_t route_listener;
142 
143 static int rtdeletemsg(struct rtentry *);
144 static int rtflushclone1(struct rtentry *, void *);
145 static void rtflushclone(sa_family_t family, struct rtentry *);
146 
147 #ifdef RTFLUSH_DEBUG
148 static void sysctl_net_rtcache_setup(struct sysctllog **);
149 static void
150 sysctl_net_rtcache_setup(struct sysctllog **clog)
151 {
152 	const struct sysctlnode *rnode;
153 
154 	if (sysctl_createv(clog, 0, NULL, &rnode, CTLFLAG_PERMANENT,
155 	    CTLTYPE_NODE,
156 	    "rtcache", SYSCTL_DESCR("Route cache related settings"),
157 	    NULL, 0, NULL, 0, CTL_NET, CTL_CREATE, CTL_EOL) != 0)
158 		return;
159 	if (sysctl_createv(clog, 0, &rnode, &rnode,
160 	    CTLFLAG_PERMANENT|CTLFLAG_READWRITE, CTLTYPE_INT,
161 	    "debug", SYSCTL_DESCR("Debug route caches"),
162 	    NULL, 0, &_rtcache_debug, 0, CTL_CREATE, CTL_EOL) != 0)
163 		return;
164 }
165 #endif /* RTFLUSH_DEBUG */
166 
167 struct ifaddr *
168 rt_get_ifa(struct rtentry *rt)
169 {
170 	struct ifaddr *ifa;
171 
172 	if ((ifa = rt->rt_ifa) == NULL)
173 		return ifa;
174 	else if (ifa->ifa_getifa == NULL)
175 		return ifa;
176 #if 0
177 	else if (ifa->ifa_seqno != NULL && *ifa->ifa_seqno == rt->rt_ifa_seqno)
178 		return ifa;
179 #endif
180 	else {
181 		ifa = (*ifa->ifa_getifa)(ifa, rt_getkey(rt));
182 		rt_replace_ifa(rt, ifa);
183 		return ifa;
184 	}
185 }
186 
187 static void
188 rt_set_ifa1(struct rtentry *rt, struct ifaddr *ifa)
189 {
190 	rt->rt_ifa = ifa;
191 	if (ifa->ifa_seqno != NULL)
192 		rt->rt_ifa_seqno = *ifa->ifa_seqno;
193 }
194 
195 /*
196  * Is this route the connected route for the ifa?
197  */
198 static int
199 rt_ifa_connected(const struct rtentry *rt, const struct ifaddr *ifa)
200 {
201 	const struct sockaddr *key, *dst, *odst;
202 	struct sockaddr_storage maskeddst;
203 
204 	key = rt_getkey(rt);
205 	dst = rt->rt_flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
206 	if (dst == NULL ||
207 	    dst->sa_family != key->sa_family ||
208 	    dst->sa_len != key->sa_len)
209 		return 0;
210 	if ((rt->rt_flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
211 		odst = dst;
212 		dst = (struct sockaddr *)&maskeddst;
213 		rt_maskedcopy(odst, (struct sockaddr *)&maskeddst,
214 		    ifa->ifa_netmask);
215 	}
216 	return (memcmp(dst, key, dst->sa_len) == 0);
217 }
218 
219 void
220 rt_replace_ifa(struct rtentry *rt, struct ifaddr *ifa)
221 {
222 	if (rt->rt_ifa &&
223 	    rt->rt_ifa != ifa &&
224 	    rt->rt_ifa->ifa_flags & IFA_ROUTE &&
225 	    rt_ifa_connected(rt, rt->rt_ifa))
226 	{
227 		RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
228 		    "replace deleted IFA_ROUTE\n",
229 		    (void *)rt->_rt_key, (void *)rt->rt_ifa);
230 		rt->rt_ifa->ifa_flags &= ~IFA_ROUTE;
231 		if (rt_ifa_connected(rt, ifa)) {
232 			RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
233 			    "replace added IFA_ROUTE\n",
234 			    (void *)rt->_rt_key, (void *)ifa);
235 			ifa->ifa_flags |= IFA_ROUTE;
236 		}
237 	}
238 
239 	IFAREF(ifa);
240 	IFAFREE(rt->rt_ifa);
241 	rt_set_ifa1(rt, ifa);
242 }
243 
244 static void
245 rt_set_ifa(struct rtentry *rt, struct ifaddr *ifa)
246 {
247 	IFAREF(ifa);
248 	rt_set_ifa1(rt, ifa);
249 }
250 
251 static int
252 route_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
253     void *arg0, void *arg1, void *arg2, void *arg3)
254 {
255 	struct rt_msghdr *rtm;
256 	int result;
257 
258 	result = KAUTH_RESULT_DEFER;
259 	rtm = arg1;
260 
261 	if (action != KAUTH_NETWORK_ROUTE)
262 		return result;
263 
264 	if (rtm->rtm_type == RTM_GET)
265 		result = KAUTH_RESULT_ALLOW;
266 
267 	return result;
268 }
269 
270 void
271 rt_init(void)
272 {
273 
274 #ifdef RTFLUSH_DEBUG
275 	sysctl_net_rtcache_setup(NULL);
276 #endif
277 
278 	pool_init(&rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl",
279 	    NULL, IPL_SOFTNET);
280 	pool_init(&rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl",
281 	    NULL, IPL_SOFTNET);
282 
283 	rn_init();	/* initialize all zeroes, all ones, mask table */
284 	rtbl_init();
285 
286 	route_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
287 	    route_listener_cb, NULL);
288 }
289 
290 void
291 rtflushall(int family)
292 {
293 	struct domain *dom;
294 
295 	if (rtcache_debug())
296 		printf("%s: enter\n", __func__);
297 
298 	if ((dom = pffinddomain(family)) == NULL)
299 		return;
300 
301 	rtcache_invalidate(&dom->dom_rtcache);
302 }
303 
304 void
305 rtcache(struct route *ro)
306 {
307 	struct domain *dom;
308 
309 	rtcache_invariants(ro);
310 	KASSERT(ro->_ro_rt != NULL);
311 	KASSERT(ro->ro_invalid == false);
312 	KASSERT(rtcache_getdst(ro) != NULL);
313 
314 	if ((dom = pffinddomain(rtcache_getdst(ro)->sa_family)) == NULL)
315 		return;
316 
317 	LIST_INSERT_HEAD(&dom->dom_rtcache, ro, ro_rtcache_next);
318 	rtcache_invariants(ro);
319 }
320 
321 /*
322  * Packet routing routines.
323  */
324 struct rtentry *
325 rtalloc1(const struct sockaddr *dst, int report)
326 {
327 	rtbl_t *rtbl = rt_gettable(dst->sa_family);
328 	struct rtentry *rt;
329 	struct rtentry *newrt = NULL;
330 	struct rt_addrinfo info;
331 	int  s = splsoftnet(), err = 0, msgtype = RTM_MISS;
332 
333 	if (rtbl != NULL && (rt = rt_matchaddr(rtbl, dst)) != NULL) {
334 		newrt = rt;
335 		if (report && (rt->rt_flags & RTF_CLONING)) {
336 			err = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
337 			    &newrt);
338 			if (err) {
339 				newrt = rt;
340 				rt->rt_refcnt++;
341 				goto miss;
342 			}
343 			KASSERT(newrt != NULL);
344 			rt = newrt;
345 			if (rt->rt_flags & RTF_XRESOLVE) {
346 				msgtype = RTM_RESOLVE;
347 				goto miss;
348 			}
349 			/* Inform listeners of the new route */
350 			memset(&info, 0, sizeof(info));
351 			info.rti_info[RTAX_DST] = rt_getkey(rt);
352 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
353 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
354 			if (rt->rt_ifp != NULL) {
355 				info.rti_info[RTAX_IFP] =
356 				    rt->rt_ifp->if_dl->ifa_addr;
357 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
358 			}
359 			rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0);
360 		} else
361 			rt->rt_refcnt++;
362 	} else {
363 		rtstat.rts_unreach++;
364 	miss:	if (report) {
365 			memset((void *)&info, 0, sizeof(info));
366 			info.rti_info[RTAX_DST] = dst;
367 			rt_missmsg(msgtype, &info, 0, err);
368 		}
369 	}
370 	splx(s);
371 	return newrt;
372 }
373 
374 void
375 rtfree(struct rtentry *rt)
376 {
377 	struct ifaddr *ifa;
378 
379 	if (rt == NULL)
380 		panic("rtfree");
381 	rt->rt_refcnt--;
382 	if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
383 		rt_assert_inactive(rt);
384 		rttrash--;
385 		if (rt->rt_refcnt < 0) {
386 			printf("rtfree: %p not freed (neg refs)\n", rt);
387 			return;
388 		}
389 		rt_timer_remove_all(rt, 0);
390 		ifa = rt->rt_ifa;
391 		rt->rt_ifa = NULL;
392 		IFAFREE(ifa);
393 		rt->rt_ifp = NULL;
394 		rt_destroy(rt);
395 		pool_put(&rtentry_pool, rt);
396 	}
397 }
398 
399 void
400 ifafree(struct ifaddr *ifa)
401 {
402 
403 #ifdef DIAGNOSTIC
404 	if (ifa == NULL)
405 		panic("ifafree: null ifa");
406 	if (ifa->ifa_refcnt != 0)
407 		panic("ifafree: ifa_refcnt != 0 (%d)", ifa->ifa_refcnt);
408 #endif
409 #ifdef IFAREF_DEBUG
410 	printf("ifafree: freeing ifaddr %p\n", ifa);
411 #endif
412 	free(ifa, M_IFADDR);
413 }
414 
415 /*
416  * Force a routing table entry to the specified
417  * destination to go through the given gateway.
418  * Normally called as a result of a routing redirect
419  * message from the network layer.
420  *
421  * N.B.: must be called at splsoftnet
422  */
423 void
424 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
425 	const struct sockaddr *netmask, int flags, const struct sockaddr *src,
426 	struct rtentry **rtp)
427 {
428 	struct rtentry *rt;
429 	int error = 0;
430 	uint64_t *stat = NULL;
431 	struct rt_addrinfo info;
432 	struct ifaddr *ifa;
433 
434 	/* verify the gateway is directly reachable */
435 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
436 		error = ENETUNREACH;
437 		goto out;
438 	}
439 	rt = rtalloc1(dst, 0);
440 	/*
441 	 * If the redirect isn't from our current router for this dst,
442 	 * it's either old or wrong.  If it redirects us to ourselves,
443 	 * we have a routing loop, perhaps as a result of an interface
444 	 * going down recently.
445 	 */
446 	if (!(flags & RTF_DONE) && rt &&
447 	     (sockaddr_cmp(src, rt->rt_gateway) != 0 || rt->rt_ifa != ifa))
448 		error = EINVAL;
449 	else if (ifa_ifwithaddr(gateway))
450 		error = EHOSTUNREACH;
451 	if (error)
452 		goto done;
453 	/*
454 	 * Create a new entry if we just got back a wildcard entry
455 	 * or the lookup failed.  This is necessary for hosts
456 	 * which use routing redirects generated by smart gateways
457 	 * to dynamically build the routing tables.
458 	 */
459 	if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
460 		goto create;
461 	/*
462 	 * Don't listen to the redirect if it's
463 	 * for a route to an interface.
464 	 */
465 	if (rt->rt_flags & RTF_GATEWAY) {
466 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
467 			/*
468 			 * Changing from route to net => route to host.
469 			 * Create new route, rather than smashing route to net.
470 			 */
471 		create:
472 			if (rt != NULL)
473 				rtfree(rt);
474 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
475 			memset(&info, 0, sizeof(info));
476 			info.rti_info[RTAX_DST] = dst;
477 			info.rti_info[RTAX_GATEWAY] = gateway;
478 			info.rti_info[RTAX_NETMASK] = netmask;
479 			info.rti_ifa = ifa;
480 			info.rti_flags = flags;
481 			rt = NULL;
482 			error = rtrequest1(RTM_ADD, &info, &rt);
483 			if (rt != NULL)
484 				flags = rt->rt_flags;
485 			stat = &rtstat.rts_dynamic;
486 		} else {
487 			/*
488 			 * Smash the current notion of the gateway to
489 			 * this destination.  Should check about netmask!!!
490 			 */
491 			rt->rt_flags |= RTF_MODIFIED;
492 			flags |= RTF_MODIFIED;
493 			stat = &rtstat.rts_newgateway;
494 			rt_setgate(rt, gateway);
495 		}
496 	} else
497 		error = EHOSTUNREACH;
498 done:
499 	if (rt) {
500 		if (rtp != NULL && !error)
501 			*rtp = rt;
502 		else
503 			rtfree(rt);
504 	}
505 out:
506 	if (error)
507 		rtstat.rts_badredirect++;
508 	else if (stat != NULL)
509 		(*stat)++;
510 	memset(&info, 0, sizeof(info));
511 	info.rti_info[RTAX_DST] = dst;
512 	info.rti_info[RTAX_GATEWAY] = gateway;
513 	info.rti_info[RTAX_NETMASK] = netmask;
514 	info.rti_info[RTAX_AUTHOR] = src;
515 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
516 }
517 
518 /*
519  * Delete a route and generate a message
520  */
521 static int
522 rtdeletemsg(struct rtentry *rt)
523 {
524 	int error;
525 	struct rt_addrinfo info;
526 
527 	/*
528 	 * Request the new route so that the entry is not actually
529 	 * deleted.  That will allow the information being reported to
530 	 * be accurate (and consistent with route_output()).
531 	 */
532 	memset(&info, 0, sizeof(info));
533 	info.rti_info[RTAX_DST] = rt_getkey(rt);
534 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
535 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
536 	info.rti_flags = rt->rt_flags;
537 	error = rtrequest1(RTM_DELETE, &info, &rt);
538 
539 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
540 
541 	/* Adjust the refcount */
542 	if (error == 0 && rt->rt_refcnt <= 0) {
543 		rt->rt_refcnt++;
544 		rtfree(rt);
545 	}
546 	return error;
547 }
548 
549 static int
550 rtflushclone1(struct rtentry *rt, void *arg)
551 {
552 	struct rtentry *parent;
553 
554 	parent = (struct rtentry *)arg;
555 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
556 		rtdeletemsg(rt);
557 	return 0;
558 }
559 
560 static void
561 rtflushclone(sa_family_t family, struct rtentry *parent)
562 {
563 
564 #ifdef DIAGNOSTIC
565 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
566 		panic("rtflushclone: called with a non-cloning route");
567 #endif
568 	rt_walktree(family, rtflushclone1, (void *)parent);
569 }
570 
571 struct ifaddr *
572 ifa_ifwithroute(int flags, const struct sockaddr *dst,
573 	const struct sockaddr *gateway)
574 {
575 	struct ifaddr *ifa;
576 	if ((flags & RTF_GATEWAY) == 0) {
577 		/*
578 		 * If we are adding a route to an interface,
579 		 * and the interface is a pt to pt link
580 		 * we should search for the destination
581 		 * as our clue to the interface.  Otherwise
582 		 * we can use the local address.
583 		 */
584 		ifa = NULL;
585 		if ((flags & RTF_HOST) && gateway->sa_family != AF_LINK)
586 			ifa = ifa_ifwithdstaddr(dst);
587 		if (ifa == NULL)
588 			ifa = ifa_ifwithaddr(gateway);
589 	} else {
590 		/*
591 		 * If we are adding a route to a remote net
592 		 * or host, the gateway may still be on the
593 		 * other end of a pt to pt link.
594 		 */
595 		ifa = ifa_ifwithdstaddr(gateway);
596 	}
597 	if (ifa == NULL)
598 		ifa = ifa_ifwithnet(gateway);
599 	if (ifa == NULL) {
600 		struct rtentry *rt = rtalloc1(dst, 0);
601 		if (rt == NULL)
602 			return NULL;
603 		rt->rt_refcnt--;
604 		if ((ifa = rt->rt_ifa) == NULL)
605 			return NULL;
606 	}
607 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
608 		struct ifaddr *oifa = ifa;
609 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
610 		if (ifa == NULL)
611 			ifa = oifa;
612 	}
613 	return ifa;
614 }
615 
616 int
617 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
618 	const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
619 {
620 	struct rt_addrinfo info;
621 
622 	memset(&info, 0, sizeof(info));
623 	info.rti_flags = flags;
624 	info.rti_info[RTAX_DST] = dst;
625 	info.rti_info[RTAX_GATEWAY] = gateway;
626 	info.rti_info[RTAX_NETMASK] = netmask;
627 	return rtrequest1(req, &info, ret_nrt);
628 }
629 
630 int
631 rt_getifa(struct rt_addrinfo *info)
632 {
633 	struct ifaddr *ifa;
634 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
635 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
636 	const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
637 	const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
638 	int flags = info->rti_flags;
639 
640 	/*
641 	 * ifp may be specified by sockaddr_dl when protocol address
642 	 * is ambiguous
643 	 */
644 	if (info->rti_ifp == NULL && ifpaddr != NULL
645 	    && ifpaddr->sa_family == AF_LINK &&
646 	    (ifa = ifa_ifwithnet(ifpaddr)) != NULL)
647 		info->rti_ifp = ifa->ifa_ifp;
648 	if (info->rti_ifa == NULL && ifaaddr != NULL)
649 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
650 	if (info->rti_ifa == NULL) {
651 		const struct sockaddr *sa;
652 
653 		sa = ifaaddr != NULL ? ifaaddr :
654 		    (gateway != NULL ? gateway : dst);
655 		if (sa != NULL && info->rti_ifp != NULL)
656 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
657 		else if (dst != NULL && gateway != NULL)
658 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
659 		else if (sa != NULL)
660 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
661 	}
662 	if ((ifa = info->rti_ifa) == NULL)
663 		return ENETUNREACH;
664 	if (ifa->ifa_getifa != NULL)
665 		info->rti_ifa = ifa = (*ifa->ifa_getifa)(ifa, dst);
666 	if (info->rti_ifp == NULL)
667 		info->rti_ifp = ifa->ifa_ifp;
668 	return 0;
669 }
670 
671 int
672 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
673 {
674 	int s = splsoftnet();
675 	int error = 0, rc;
676 	struct rtentry *rt, *crt;
677 	rtbl_t *rtbl;
678 	struct ifaddr *ifa, *ifa2;
679 	struct sockaddr_storage maskeddst;
680 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
681 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
682 	const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
683 	int flags = info->rti_flags;
684 #define senderr(x) { error = x ; goto bad; }
685 
686 	if ((rtbl = rt_gettable(dst->sa_family)) == NULL)
687 		senderr(ESRCH);
688 	if (flags & RTF_HOST)
689 		netmask = NULL;
690 	switch (req) {
691 	case RTM_DELETE:
692 		if (netmask) {
693 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
694 			    netmask);
695 			dst = (struct sockaddr *)&maskeddst;
696 		}
697 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
698 			senderr(ESRCH);
699 		if ((rt->rt_flags & RTF_CLONING) != 0) {
700 			/* clean up any cloned children */
701 			rtflushclone(dst->sa_family, rt);
702 		}
703 		if ((rt = rt_deladdr(rtbl, dst, netmask)) == NULL)
704 			senderr(ESRCH);
705 		if (rt->rt_gwroute) {
706 			RTFREE(rt->rt_gwroute);
707 			rt->rt_gwroute = NULL;
708 		}
709 		if (rt->rt_parent) {
710 			rt->rt_parent->rt_refcnt--;
711 			rt->rt_parent = NULL;
712 		}
713 		rt->rt_flags &= ~RTF_UP;
714 		if ((ifa = rt->rt_ifa)) {
715 			if (ifa->ifa_flags & IFA_ROUTE &&
716 			    rt_ifa_connected(rt, ifa)) {
717 				RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
718 				    "deleted IFA_ROUTE\n",
719 				    (void *)rt->_rt_key, (void *)ifa);
720 				ifa->ifa_flags &= ~IFA_ROUTE;
721 			}
722 			if (ifa->ifa_rtrequest)
723 				ifa->ifa_rtrequest(RTM_DELETE, rt, info);
724 		}
725 		rttrash++;
726 		if (ret_nrt)
727 			*ret_nrt = rt;
728 		else if (rt->rt_refcnt <= 0) {
729 			rt->rt_refcnt++;
730 			rtfree(rt);
731 		}
732 		break;
733 
734 	case RTM_RESOLVE:
735 		if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
736 			senderr(EINVAL);
737 		if ((rt->rt_flags & RTF_CLONING) == 0)
738 			senderr(EINVAL);
739 		ifa = rt->rt_ifa;
740 		flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC);
741 		flags |= RTF_CLONED;
742 		gateway = rt->rt_gateway;
743 		flags |= RTF_HOST;
744 		goto makeroute;
745 
746 	case RTM_ADD:
747 		if (info->rti_ifa == NULL && (error = rt_getifa(info)))
748 			senderr(error);
749 		ifa = info->rti_ifa;
750 	makeroute:
751 		/* Already at splsoftnet() so pool_get/pool_put are safe */
752 		rt = pool_get(&rtentry_pool, PR_NOWAIT);
753 		if (rt == NULL)
754 			senderr(ENOBUFS);
755 		memset(rt, 0, sizeof(*rt));
756 		rt->rt_flags = RTF_UP | flags;
757 		LIST_INIT(&rt->rt_timer);
758 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
759 		if (rt_setkey(rt, dst, M_NOWAIT) == NULL ||
760 		    rt_setgate(rt, gateway) != 0) {
761 			pool_put(&rtentry_pool, rt);
762 			senderr(ENOBUFS);
763 		}
764 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
765 		if (netmask) {
766 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
767 			    netmask);
768 			rt_setkey(rt, (struct sockaddr *)&maskeddst, M_NOWAIT);
769 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
770 		} else {
771 			rt_setkey(rt, dst, M_NOWAIT);
772 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
773 		}
774 		rt_set_ifa(rt, ifa);
775 		if (info->rti_info[RTAX_TAG] != NULL)
776 			rt_settag(rt, info->rti_info[RTAX_TAG]);
777 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
778 		if (info->rti_info[RTAX_IFP] != NULL &&
779 		    (ifa2 = ifa_ifwithnet(info->rti_info[RTAX_IFP])) != NULL &&
780 		    ifa2->ifa_ifp != NULL)
781 			rt->rt_ifp = ifa2->ifa_ifp;
782 		else
783 			rt->rt_ifp = ifa->ifa_ifp;
784 		if (req == RTM_RESOLVE) {
785 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
786 			rt->rt_parent = *ret_nrt;
787 			rt->rt_parent->rt_refcnt++;
788 		}
789 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
790 		rc = rt_addaddr(rtbl, rt, netmask);
791 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
792 		if (rc != 0 && (crt = rtalloc1(rt_getkey(rt), 0)) != NULL) {
793 			/* overwrite cloned route */
794 			if ((crt->rt_flags & RTF_CLONED) != 0) {
795 				rtdeletemsg(crt);
796 				rc = rt_addaddr(rtbl, rt, netmask);
797 			}
798 			RTFREE(crt);
799 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
800 		}
801 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
802 		if (rc != 0) {
803 			IFAFREE(ifa);
804 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
805 				rtfree(rt->rt_parent);
806 			if (rt->rt_gwroute)
807 				rtfree(rt->rt_gwroute);
808 			rt_destroy(rt);
809 			pool_put(&rtentry_pool, rt);
810 			senderr(rc);
811 		}
812 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
813 		if (ifa->ifa_rtrequest)
814 			ifa->ifa_rtrequest(req, rt, info);
815 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
816 		if (ret_nrt) {
817 			*ret_nrt = rt;
818 			rt->rt_refcnt++;
819 		}
820 		if ((rt->rt_flags & RTF_CLONING) != 0) {
821 			/* clean up any cloned children */
822 			rtflushclone(dst->sa_family, rt);
823 		}
824 		rtflushall(dst->sa_family);
825 		break;
826 	case RTM_GET:
827 		if (netmask != NULL) {
828 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
829 			    netmask);
830 			dst = (struct sockaddr *)&maskeddst;
831 		}
832 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
833 			senderr(ESRCH);
834 		if (ret_nrt != NULL) {
835 			*ret_nrt = rt;
836 			rt->rt_refcnt++;
837 		}
838 		break;
839 	}
840 bad:
841 	splx(s);
842 	return error;
843 }
844 
845 int
846 rt_setgate(struct rtentry *rt, const struct sockaddr *gate)
847 {
848 	KASSERT(rt != rt->rt_gwroute);
849 
850 	KASSERT(rt->_rt_key != NULL);
851 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
852 
853 	if (rt->rt_gwroute) {
854 		RTFREE(rt->rt_gwroute);
855 		rt->rt_gwroute = NULL;
856 	}
857 	KASSERT(rt->_rt_key != NULL);
858 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
859 	if (rt->rt_gateway != NULL)
860 		sockaddr_free(rt->rt_gateway);
861 	KASSERT(rt->_rt_key != NULL);
862 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
863 	if ((rt->rt_gateway = sockaddr_dup(gate, M_NOWAIT)) == NULL)
864 		return ENOMEM;
865 	KASSERT(rt->_rt_key != NULL);
866 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
867 
868 	if (rt->rt_flags & RTF_GATEWAY) {
869 		KASSERT(rt->_rt_key != NULL);
870 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
871 		rt->rt_gwroute = rtalloc1(gate, 1);
872 		/*
873 		 * If we switched gateways, grab the MTU from the new
874 		 * gateway route if the current MTU, if the current MTU is
875 		 * greater than the MTU of gateway.
876 		 * Note that, if the MTU of gateway is 0, we will reset the
877 		 * MTU of the route to run PMTUD again from scratch. XXX
878 		 */
879 		KASSERT(rt->_rt_key != NULL);
880 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
881 		if (rt->rt_gwroute
882 		    && !(rt->rt_rmx.rmx_locks & RTV_MTU)
883 		    && rt->rt_rmx.rmx_mtu
884 		    && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
885 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
886 		}
887 	}
888 	KASSERT(rt->_rt_key != NULL);
889 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
890 	return 0;
891 }
892 
893 void
894 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
895 	const struct sockaddr *netmask)
896 {
897 	const char *netmaskp = &netmask->sa_data[0],
898 	           *srcp = &src->sa_data[0];
899 	char *dstp = &dst->sa_data[0];
900 	const char *maskend = (char *)dst + MIN(netmask->sa_len, src->sa_len);
901 	const char *srcend = (char *)dst + src->sa_len;
902 
903 	dst->sa_len = src->sa_len;
904 	dst->sa_family = src->sa_family;
905 
906 	while (dstp < maskend)
907 		*dstp++ = *srcp++ & *netmaskp++;
908 	if (dstp < srcend)
909 		memset(dstp, 0, (size_t)(srcend - dstp));
910 }
911 
912 /*
913  * Set up or tear down a routing table entry, normally
914  * for an interface.
915  */
916 int
917 rtinit(struct ifaddr *ifa, int cmd, int flags)
918 {
919 	struct rtentry *rt;
920 	struct sockaddr *dst, *odst;
921 	struct sockaddr_storage maskeddst;
922 	struct rtentry *nrt = NULL;
923 	int error;
924 	struct rt_addrinfo info;
925 	struct sockaddr_dl *sdl;
926 	const struct sockaddr_dl *ifsdl;
927 
928 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
929 	if (cmd == RTM_DELETE) {
930 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
931 			/* Delete subnet route for this interface */
932 			odst = dst;
933 			dst = (struct sockaddr *)&maskeddst;
934 			rt_maskedcopy(odst, dst, ifa->ifa_netmask);
935 		}
936 		if ((rt = rtalloc1(dst, 0)) != NULL) {
937 			rt->rt_refcnt--;
938 			if (rt->rt_ifa != ifa)
939 				return (flags & RTF_HOST) ? EHOSTUNREACH
940 							: ENETUNREACH;
941 		}
942 	}
943 	memset(&info, 0, sizeof(info));
944 	info.rti_ifa = ifa;
945 	info.rti_flags = flags | ifa->ifa_flags;
946 	info.rti_info[RTAX_DST] = dst;
947 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
948 	/*
949 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
950 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
951 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
952 	 * change it to meet bsdi4 behavior.
953 	 */
954 	if (cmd != RTM_LLINFO_UPD)
955 		info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
956 	error = rtrequest1((cmd == RTM_LLINFO_UPD) ? RTM_GET : cmd, &info,
957 	    &nrt);
958 	if (error != 0 || (rt = nrt) == NULL)
959 		;
960 	else switch (cmd) {
961 	case RTM_DELETE:
962 		rt_newaddrmsg(cmd, ifa, error, nrt);
963 		if (rt->rt_refcnt <= 0) {
964 			rt->rt_refcnt++;
965 			rtfree(rt);
966 		}
967 		break;
968 	case RTM_LLINFO_UPD:
969 		rt->rt_refcnt--;
970 		RT_DPRINTF("%s: updating%s\n", __func__,
971 		    ((rt->rt_flags & RTF_LLINFO) == 0) ? " (no llinfo)" : "");
972 
973 		ifsdl = ifa->ifa_ifp->if_sadl;
974 
975 		if ((rt->rt_flags & RTF_LLINFO) != 0 &&
976 		    (sdl = satosdl(rt->rt_gateway)) != NULL &&
977 		    sdl->sdl_family == AF_LINK &&
978 		    sockaddr_dl_setaddr(sdl, sdl->sdl_len, CLLADDR(ifsdl),
979 		                        ifa->ifa_ifp->if_addrlen) == NULL) {
980 			error = EINVAL;
981 			break;
982 		}
983 
984 		if (cmd == RTM_LLINFO_UPD && ifa->ifa_rtrequest != NULL)
985 			ifa->ifa_rtrequest(RTM_LLINFO_UPD, rt, &info);
986 		rt_newaddrmsg(RTM_CHANGE, ifa, error, nrt);
987 		break;
988 	case RTM_ADD:
989 		rt->rt_refcnt--;
990 		if (rt->rt_ifa != ifa) {
991 			printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
992 				rt->rt_ifa);
993 			if (rt->rt_ifa->ifa_rtrequest != NULL) {
994 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
995 				    &info);
996 			}
997 			rt_replace_ifa(rt, ifa);
998 			rt->rt_ifp = ifa->ifa_ifp;
999 			if (ifa->ifa_rtrequest != NULL)
1000 				ifa->ifa_rtrequest(RTM_ADD, rt, &info);
1001 		}
1002 		rt_newaddrmsg(cmd, ifa, error, nrt);
1003 		break;
1004 	}
1005 	return error;
1006 }
1007 
1008 /*
1009  * Route timer routines.  These routes allow functions to be called
1010  * for various routes at any time.  This is useful in supporting
1011  * path MTU discovery and redirect route deletion.
1012  *
1013  * This is similar to some BSDI internal functions, but it provides
1014  * for multiple queues for efficiency's sake...
1015  */
1016 
1017 LIST_HEAD(, rttimer_queue) rttimer_queue_head;
1018 static int rt_init_done = 0;
1019 
1020 #define RTTIMER_CALLOUT(r)	do {					\
1021 		if (r->rtt_func != NULL) {				\
1022 			(*r->rtt_func)(r->rtt_rt, r);			\
1023 		} else {						\
1024 			rtrequest((int) RTM_DELETE,			\
1025 				  rt_getkey(r->rtt_rt),			\
1026 				  0, 0, 0, 0);				\
1027 		}							\
1028 	} while (/*CONSTCOND*/0)
1029 
1030 /*
1031  * Some subtle order problems with domain initialization mean that
1032  * we cannot count on this being run from rt_init before various
1033  * protocol initializations are done.  Therefore, we make sure
1034  * that this is run when the first queue is added...
1035  */
1036 
1037 void
1038 rt_timer_init(void)
1039 {
1040 	assert(rt_init_done == 0);
1041 
1042 	LIST_INIT(&rttimer_queue_head);
1043 	callout_init(&rt_timer_ch, 0);
1044 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1045 	rt_init_done = 1;
1046 }
1047 
1048 struct rttimer_queue *
1049 rt_timer_queue_create(u_int timeout)
1050 {
1051 	struct rttimer_queue *rtq;
1052 
1053 	if (rt_init_done == 0)
1054 		rt_timer_init();
1055 
1056 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
1057 	if (rtq == NULL)
1058 		return NULL;
1059 	memset(rtq, 0, sizeof(*rtq));
1060 
1061 	rtq->rtq_timeout = timeout;
1062 	TAILQ_INIT(&rtq->rtq_head);
1063 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
1064 
1065 	return rtq;
1066 }
1067 
1068 void
1069 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
1070 {
1071 
1072 	rtq->rtq_timeout = timeout;
1073 }
1074 
1075 void
1076 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy)
1077 {
1078 	struct rttimer *r;
1079 
1080 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
1081 		LIST_REMOVE(r, rtt_link);
1082 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1083 		if (destroy)
1084 			RTTIMER_CALLOUT(r);
1085 		/* we are already at splsoftnet */
1086 		pool_put(&rttimer_pool, r);
1087 		if (rtq->rtq_count > 0)
1088 			rtq->rtq_count--;
1089 		else
1090 			printf("rt_timer_queue_remove_all: "
1091 			    "rtq_count reached 0\n");
1092 	}
1093 }
1094 
1095 void
1096 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
1097 {
1098 
1099 	rt_timer_queue_remove_all(rtq, destroy);
1100 
1101 	LIST_REMOVE(rtq, rtq_link);
1102 
1103 	/*
1104 	 * Caller is responsible for freeing the rttimer_queue structure.
1105 	 */
1106 }
1107 
1108 unsigned long
1109 rt_timer_count(struct rttimer_queue *rtq)
1110 {
1111 	return rtq->rtq_count;
1112 }
1113 
1114 void
1115 rt_timer_remove_all(struct rtentry *rt, int destroy)
1116 {
1117 	struct rttimer *r;
1118 
1119 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
1120 		LIST_REMOVE(r, rtt_link);
1121 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
1122 		if (destroy)
1123 			RTTIMER_CALLOUT(r);
1124 		if (r->rtt_queue->rtq_count > 0)
1125 			r->rtt_queue->rtq_count--;
1126 		else
1127 			printf("rt_timer_remove_all: rtq_count reached 0\n");
1128 		/* we are already at splsoftnet */
1129 		pool_put(&rttimer_pool, r);
1130 	}
1131 }
1132 
1133 int
1134 rt_timer_add(struct rtentry *rt,
1135 	void (*func)(struct rtentry *, struct rttimer *),
1136 	struct rttimer_queue *queue)
1137 {
1138 	struct rttimer *r;
1139 	int s;
1140 
1141 	/*
1142 	 * If there's already a timer with this action, destroy it before
1143 	 * we add a new one.
1144 	 */
1145 	LIST_FOREACH(r, &rt->rt_timer, rtt_link) {
1146 		if (r->rtt_func == func)
1147 			break;
1148 	}
1149 	if (r != NULL) {
1150 		LIST_REMOVE(r, rtt_link);
1151 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
1152 		if (r->rtt_queue->rtq_count > 0)
1153 			r->rtt_queue->rtq_count--;
1154 		else
1155 			printf("rt_timer_add: rtq_count reached 0\n");
1156 	} else {
1157 		s = splsoftnet();
1158 		r = pool_get(&rttimer_pool, PR_NOWAIT);
1159 		splx(s);
1160 		if (r == NULL)
1161 			return ENOBUFS;
1162 	}
1163 
1164 	memset(r, 0, sizeof(*r));
1165 
1166 	r->rtt_rt = rt;
1167 	r->rtt_time = time_uptime;
1168 	r->rtt_func = func;
1169 	r->rtt_queue = queue;
1170 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
1171 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
1172 	r->rtt_queue->rtq_count++;
1173 
1174 	return 0;
1175 }
1176 
1177 /* ARGSUSED */
1178 void
1179 rt_timer_timer(void *arg)
1180 {
1181 	struct rttimer_queue *rtq;
1182 	struct rttimer *r;
1183 	int s;
1184 
1185 	s = splsoftnet();
1186 	LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) {
1187 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
1188 		    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
1189 			LIST_REMOVE(r, rtt_link);
1190 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1191 			RTTIMER_CALLOUT(r);
1192 			pool_put(&rttimer_pool, r);
1193 			if (rtq->rtq_count > 0)
1194 				rtq->rtq_count--;
1195 			else
1196 				printf("rt_timer_timer: rtq_count reached 0\n");
1197 		}
1198 	}
1199 	splx(s);
1200 
1201 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1202 }
1203 
1204 static struct rtentry *
1205 _rtcache_init(struct route *ro, int flag)
1206 {
1207 	rtcache_invariants(ro);
1208 	KASSERT(ro->_ro_rt == NULL);
1209 
1210 	if (rtcache_getdst(ro) == NULL)
1211 		return NULL;
1212 	ro->ro_invalid = false;
1213 	if ((ro->_ro_rt = rtalloc1(rtcache_getdst(ro), flag)) != NULL)
1214 		rtcache(ro);
1215 
1216 	rtcache_invariants(ro);
1217 	return ro->_ro_rt;
1218 }
1219 
1220 struct rtentry *
1221 rtcache_init(struct route *ro)
1222 {
1223 	return _rtcache_init(ro, 1);
1224 }
1225 
1226 struct rtentry *
1227 rtcache_init_noclone(struct route *ro)
1228 {
1229 	return _rtcache_init(ro, 0);
1230 }
1231 
1232 struct rtentry *
1233 rtcache_update(struct route *ro, int clone)
1234 {
1235 	rtcache_clear(ro);
1236 	return _rtcache_init(ro, clone);
1237 }
1238 
1239 void
1240 rtcache_copy(struct route *new_ro, const struct route *old_ro)
1241 {
1242 	struct rtentry *rt;
1243 
1244 	KASSERT(new_ro != old_ro);
1245 	rtcache_invariants(new_ro);
1246 	rtcache_invariants(old_ro);
1247 
1248 	if ((rt = rtcache_validate(old_ro)) != NULL)
1249 		rt->rt_refcnt++;
1250 
1251 	if (rtcache_getdst(old_ro) == NULL ||
1252 	    rtcache_setdst(new_ro, rtcache_getdst(old_ro)) != 0)
1253 		return;
1254 
1255 	new_ro->ro_invalid = false;
1256 	if ((new_ro->_ro_rt = rt) != NULL)
1257 		rtcache(new_ro);
1258 	rtcache_invariants(new_ro);
1259 }
1260 
1261 static struct dom_rtlist invalid_routes = LIST_HEAD_INITIALIZER(dom_rtlist);
1262 
1263 void
1264 rtcache_invalidate(struct dom_rtlist *rtlist)
1265 {
1266 	struct route *ro;
1267 
1268 	while ((ro = LIST_FIRST(rtlist)) != NULL) {
1269 		rtcache_invariants(ro);
1270 		KASSERT(ro->_ro_rt != NULL);
1271 		ro->ro_invalid = true;
1272 		LIST_REMOVE(ro, ro_rtcache_next);
1273 		LIST_INSERT_HEAD(&invalid_routes, ro, ro_rtcache_next);
1274 		rtcache_invariants(ro);
1275 	}
1276 }
1277 
1278 void
1279 rtcache_clear(struct route *ro)
1280 {
1281 	rtcache_invariants(ro);
1282 	if (ro->_ro_rt == NULL)
1283 		return;
1284 
1285 	LIST_REMOVE(ro, ro_rtcache_next);
1286 
1287 	RTFREE(ro->_ro_rt);
1288 	ro->_ro_rt = NULL;
1289 	ro->ro_invalid = false;
1290 	rtcache_invariants(ro);
1291 }
1292 
1293 struct rtentry *
1294 rtcache_lookup2(struct route *ro, const struct sockaddr *dst, int clone,
1295     int *hitp)
1296 {
1297 	const struct sockaddr *odst;
1298 	struct rtentry *rt = NULL;
1299 
1300 	rtcache_invariants(ro);
1301 
1302 	odst = rtcache_getdst(ro);
1303 
1304 	if (odst == NULL)
1305 		;
1306 	else if (sockaddr_cmp(odst, dst) != 0)
1307 		rtcache_free(ro);
1308 	else if ((rt = rtcache_validate(ro)) == NULL)
1309 		rtcache_clear(ro);
1310 
1311 	if (rt == NULL) {
1312 		*hitp = 0;
1313 		if (rtcache_setdst(ro, dst) == 0)
1314 			rt = _rtcache_init(ro, clone);
1315 	} else
1316 		*hitp = 1;
1317 
1318 	rtcache_invariants(ro);
1319 
1320 	return rt;
1321 }
1322 
1323 void
1324 rtcache_free(struct route *ro)
1325 {
1326 	rtcache_clear(ro);
1327 	if (ro->ro_sa != NULL) {
1328 		sockaddr_free(ro->ro_sa);
1329 		ro->ro_sa = NULL;
1330 	}
1331 	rtcache_invariants(ro);
1332 }
1333 
1334 int
1335 rtcache_setdst(struct route *ro, const struct sockaddr *sa)
1336 {
1337 	KASSERT(sa != NULL);
1338 
1339 	rtcache_invariants(ro);
1340 	if (ro->ro_sa != NULL && ro->ro_sa->sa_family == sa->sa_family) {
1341 		rtcache_clear(ro);
1342 		if (sockaddr_copy(ro->ro_sa, ro->ro_sa->sa_len, sa) != NULL) {
1343 			rtcache_invariants(ro);
1344 			return 0;
1345 		}
1346 		sockaddr_free(ro->ro_sa);
1347 	} else if (ro->ro_sa != NULL)
1348 		rtcache_free(ro);	/* free ro_sa, wrong family */
1349 
1350 	KASSERT(ro->_ro_rt == NULL);
1351 
1352 	if ((ro->ro_sa = sockaddr_dup(sa, M_NOWAIT)) == NULL) {
1353 		rtcache_invariants(ro);
1354 		return ENOMEM;
1355 	}
1356 	rtcache_invariants(ro);
1357 	return 0;
1358 }
1359 
1360 const struct sockaddr *
1361 rt_settag(struct rtentry *rt, const struct sockaddr *tag)
1362 {
1363 	if (rt->rt_tag != tag) {
1364 		if (rt->rt_tag != NULL)
1365 			sockaddr_free(rt->rt_tag);
1366 		rt->rt_tag = sockaddr_dup(tag, M_NOWAIT);
1367 	}
1368 	return rt->rt_tag;
1369 }
1370 
1371 struct sockaddr *
1372 rt_gettag(struct rtentry *rt)
1373 {
1374 	return rt->rt_tag;
1375 }
1376