xref: /netbsd-src/sys/net/route.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: route.c,v 1.98 2007/10/10 22:14:38 dyoung Exp $	*/
2 
3 /*-
4  * Copyright (c) 1998 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  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
42  * All rights reserved.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  * 3. Neither the name of the project nor the names of its contributors
53  *    may be used to endorse or promote products derived from this software
54  *    without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
57  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
58  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
59  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
60  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
61  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
62  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
63  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
64  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
65  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66  * SUCH DAMAGE.
67  */
68 
69 /*
70  * Copyright (c) 1980, 1986, 1991, 1993
71  *	The Regents of the University of California.  All rights reserved.
72  *
73  * Redistribution and use in source and binary forms, with or without
74  * modification, are permitted provided that the following conditions
75  * are met:
76  * 1. Redistributions of source code must retain the above copyright
77  *    notice, this list of conditions and the following disclaimer.
78  * 2. Redistributions in binary form must reproduce the above copyright
79  *    notice, this list of conditions and the following disclaimer in the
80  *    documentation and/or other materials provided with the distribution.
81  * 3. Neither the name of the University nor the names of its contributors
82  *    may be used to endorse or promote products derived from this software
83  *    without specific prior written permission.
84  *
85  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
86  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
87  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
88  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
89  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
90  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
91  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
92  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
93  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
94  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
95  * SUCH DAMAGE.
96  *
97  *	@(#)route.c	8.3 (Berkeley) 1/9/95
98  */
99 
100 #include "opt_route.h"
101 
102 #include <sys/cdefs.h>
103 __KERNEL_RCSID(0, "$NetBSD: route.c,v 1.98 2007/10/10 22:14:38 dyoung Exp $");
104 
105 #include <sys/param.h>
106 #include <sys/sysctl.h>
107 #include <sys/systm.h>
108 #include <sys/callout.h>
109 #include <sys/proc.h>
110 #include <sys/mbuf.h>
111 #include <sys/socket.h>
112 #include <sys/socketvar.h>
113 #include <sys/domain.h>
114 #include <sys/protosw.h>
115 #include <sys/kernel.h>
116 #include <sys/ioctl.h>
117 #include <sys/pool.h>
118 
119 #include <net/if.h>
120 #include <net/route.h>
121 #include <net/raw_cb.h>
122 
123 #include <netinet/in.h>
124 #include <netinet/in_var.h>
125 
126 #ifdef RTFLUSH_DEBUG
127 #define	rtcache_debug() __predict_false(_rtcache_debug)
128 #else /* RTFLUSH_DEBUG */
129 #define	rtcache_debug() 0
130 #endif /* RTFLUSH_DEBUG */
131 
132 struct	route_cb route_cb;
133 struct	rtstat	rtstat;
134 struct	radix_node_head *rt_tables[AF_MAX+1];
135 
136 int	rttrash;		/* routes not in table but not freed */
137 struct	sockaddr wildcard;	/* zero valued cookie for wildcard searches */
138 
139 POOL_INIT(rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl", NULL,
140     IPL_SOFTNET);
141 POOL_INIT(rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl", NULL,
142     IPL_SOFTNET);
143 
144 struct callout rt_timer_ch; /* callout for rt_timer_timer() */
145 
146 #ifdef RTFLUSH_DEBUG
147 static int _rtcache_debug = 0;
148 #endif /* RTFLUSH_DEBUG */
149 
150 static int rtdeletemsg(struct rtentry *);
151 static int rtflushclone1(struct rtentry *, void *);
152 static void rtflushclone(sa_family_t family, struct rtentry *);
153 
154 #ifdef RTFLUSH_DEBUG
155 SYSCTL_SETUP(sysctl_net_rtcache_setup, "sysctl net.rtcache.debug setup")
156 {
157 	const struct sysctlnode *rnode;
158 
159 	/* XXX do not duplicate */
160 	if (sysctl_createv(clog, 0, NULL, &rnode, CTLFLAG_PERMANENT,
161 	    CTLTYPE_NODE, "net", NULL, NULL, 0, NULL, 0, CTL_NET, CTL_EOL) != 0)
162 		return;
163 	if (sysctl_createv(clog, 0, &rnode, &rnode, CTLFLAG_PERMANENT,
164 	    CTLTYPE_NODE,
165 	    "rtcache", SYSCTL_DESCR("Route cache related settings"),
166 	    NULL, 0, NULL, 0, CTL_CREATE, CTL_EOL) != 0)
167 		return;
168 	if (sysctl_createv(clog, 0, &rnode, &rnode,
169 	    CTLFLAG_PERMANENT|CTLFLAG_READWRITE, CTLTYPE_INT,
170 	    "debug", SYSCTL_DESCR("Debug route caches"),
171 	    NULL, 0, &_rtcache_debug, 0, CTL_CREATE, CTL_EOL) != 0)
172 		return;
173 }
174 #endif /* RTFLUSH_DEBUG */
175 
176 struct ifaddr *
177 rt_get_ifa(struct rtentry *rt)
178 {
179 	struct ifaddr *ifa;
180 
181 	if ((ifa = rt->rt_ifa) == NULL)
182 		return ifa;
183 	else if (ifa->ifa_getifa == NULL)
184 		return ifa;
185 #if 0
186 	else if (ifa->ifa_seqno != NULL && *ifa->ifa_seqno == rt->rt_ifa_seqno)
187 		return ifa;
188 #endif
189 	else {
190 		ifa = (*ifa->ifa_getifa)(ifa, rt_getkey(rt));
191 		rt_replace_ifa(rt, ifa);
192 		return ifa;
193 	}
194 }
195 
196 static void
197 rt_set_ifa1(struct rtentry *rt, struct ifaddr *ifa)
198 {
199 	rt->rt_ifa = ifa;
200 	if (ifa->ifa_seqno != NULL)
201 		rt->rt_ifa_seqno = *ifa->ifa_seqno;
202 }
203 
204 void
205 rt_replace_ifa(struct rtentry *rt, struct ifaddr *ifa)
206 {
207 	IFAREF(ifa);
208 	IFAFREE(rt->rt_ifa);
209 	rt_set_ifa1(rt, ifa);
210 }
211 
212 static void
213 rt_set_ifa(struct rtentry *rt, struct ifaddr *ifa)
214 {
215 	IFAREF(ifa);
216 	rt_set_ifa1(rt, ifa);
217 }
218 
219 void
220 rtable_init(void **table)
221 {
222 	struct domain *dom;
223 	DOMAIN_FOREACH(dom)
224 		if (dom->dom_rtattach)
225 			dom->dom_rtattach(&table[dom->dom_family],
226 			    dom->dom_rtoffset);
227 }
228 
229 void
230 route_init(void)
231 {
232 
233 	rn_init();	/* initialize all zeroes, all ones, mask table */
234 	rtable_init((void **)rt_tables);
235 }
236 
237 void
238 rtflushall(int family)
239 {
240 	int s;
241 	struct domain *dom;
242 	struct route *ro;
243 
244 	if (rtcache_debug())
245 		printf("%s: enter\n", __func__);
246 
247 	if ((dom = pffinddomain(family)) == NULL)
248 		return;
249 
250 	s = splnet();
251 	while ((ro = LIST_FIRST(&dom->dom_rtcache)) != NULL) {
252 		KASSERT(ro->ro_rt != NULL);
253 		rtcache_clear(ro);
254 	}
255 	splx(s);
256 }
257 
258 void
259 rtflush(struct route *ro)
260 {
261 	int s = splnet();
262 	KASSERT(ro->ro_rt != NULL);
263 	KASSERT(rtcache_getdst(ro) != NULL);
264 
265 	RTFREE(ro->ro_rt);
266 	ro->ro_rt = NULL;
267 
268 	LIST_REMOVE(ro, ro_rtcache_next);
269 	splx(s);
270 }
271 
272 void
273 rtcache(struct route *ro)
274 {
275 	int s;
276 	struct domain *dom;
277 
278 	KASSERT(ro->ro_rt != NULL);
279 	KASSERT(rtcache_getdst(ro) != NULL);
280 
281 	if ((dom = pffinddomain(rtcache_getdst(ro)->sa_family)) == NULL)
282 		return;
283 
284 	s = splnet();
285 	LIST_INSERT_HEAD(&dom->dom_rtcache, ro, ro_rtcache_next);
286 	splx(s);
287 }
288 
289 /*
290  * Packet routing routines.
291  */
292 void
293 rtalloc(struct route *ro)
294 {
295 	if (ro->ro_rt != NULL) {
296 		if (ro->ro_rt->rt_ifp != NULL &&
297 		    (ro->ro_rt->rt_flags & RTF_UP) != 0)
298 			return;
299 		rtflush(ro);
300 	}
301 	if (rtcache_getdst(ro) == NULL ||
302 	    (ro->ro_rt = rtalloc1(rtcache_getdst(ro), 1)) == NULL)
303 		return;
304 	rtcache(ro);
305 }
306 
307 struct rtentry *
308 rtalloc1(const struct sockaddr *dst, int report)
309 {
310 	struct radix_node_head *rnh = rt_tables[dst->sa_family];
311 	struct rtentry *rt;
312 	struct radix_node *rn;
313 	struct rtentry *newrt = NULL;
314 	struct rt_addrinfo info;
315 	int  s = splsoftnet(), err = 0, msgtype = RTM_MISS;
316 
317 	if (rnh && (rn = rnh->rnh_matchaddr(dst, rnh)) &&
318 	    ((rn->rn_flags & RNF_ROOT) == 0)) {
319 		newrt = rt = (struct rtentry *)rn;
320 		if (report && (rt->rt_flags & RTF_CLONING)) {
321 			err = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
322 			    &newrt);
323 			if (err) {
324 				newrt = rt;
325 				rt->rt_refcnt++;
326 				goto miss;
327 			}
328 			KASSERT(newrt != NULL);
329 			if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) {
330 				msgtype = RTM_RESOLVE;
331 				goto miss;
332 			}
333 			/* Inform listeners of the new route */
334 			memset(&info, 0, sizeof(info));
335 			info.rti_info[RTAX_DST] = rt_getkey(rt);
336 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
337 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
338 			if (rt->rt_ifp != NULL) {
339 				info.rti_info[RTAX_IFP] =
340 				    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
341 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
342 			}
343 			rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0);
344 		} else
345 			rt->rt_refcnt++;
346 	} else {
347 		rtstat.rts_unreach++;
348 	miss:	if (report) {
349 			memset((void *)&info, 0, sizeof(info));
350 			info.rti_info[RTAX_DST] = dst;
351 			rt_missmsg(msgtype, &info, 0, err);
352 		}
353 	}
354 	splx(s);
355 	return newrt;
356 }
357 
358 void
359 rtfree(struct rtentry *rt)
360 {
361 	struct ifaddr *ifa;
362 
363 	if (rt == NULL)
364 		panic("rtfree");
365 	rt->rt_refcnt--;
366 	if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
367 		if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
368 			panic ("rtfree 2");
369 		rttrash--;
370 		if (rt->rt_refcnt < 0) {
371 			printf("rtfree: %p not freed (neg refs)\n", rt);
372 			return;
373 		}
374 		rt_timer_remove_all(rt, 0);
375 		ifa = rt->rt_ifa;
376 		rt->rt_ifa = NULL;
377 		IFAFREE(ifa);
378 		rt->rt_ifp = NULL;
379 		rt_destroy(rt);
380 		pool_put(&rtentry_pool, rt);
381 	}
382 }
383 
384 void
385 ifafree(struct ifaddr *ifa)
386 {
387 
388 #ifdef DIAGNOSTIC
389 	if (ifa == NULL)
390 		panic("ifafree: null ifa");
391 	if (ifa->ifa_refcnt != 0)
392 		panic("ifafree: ifa_refcnt != 0 (%d)", ifa->ifa_refcnt);
393 #endif
394 #ifdef IFAREF_DEBUG
395 	printf("ifafree: freeing ifaddr %p\n", ifa);
396 #endif
397 	free(ifa, M_IFADDR);
398 }
399 
400 static inline int
401 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
402 {
403 	return sockaddr_cmp(sa1, sa2) == 0;
404 }
405 
406 /*
407  * Force a routing table entry to the specified
408  * destination to go through the given gateway.
409  * Normally called as a result of a routing redirect
410  * message from the network layer.
411  *
412  * N.B.: must be called at splsoftnet
413  */
414 void
415 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
416 	const struct sockaddr *netmask, int flags, const struct sockaddr *src,
417 	struct rtentry **rtp)
418 {
419 	struct rtentry *rt;
420 	int error = 0;
421 	u_quad_t *stat = NULL;
422 	struct rt_addrinfo info;
423 	struct ifaddr *ifa;
424 
425 	/* verify the gateway is directly reachable */
426 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
427 		error = ENETUNREACH;
428 		goto out;
429 	}
430 	rt = rtalloc1(dst, 0);
431 	/*
432 	 * If the redirect isn't from our current router for this dst,
433 	 * it's either old or wrong.  If it redirects us to ourselves,
434 	 * we have a routing loop, perhaps as a result of an interface
435 	 * going down recently.
436 	 */
437 	if (!(flags & RTF_DONE) && rt &&
438 	     (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa))
439 		error = EINVAL;
440 	else if (ifa_ifwithaddr(gateway))
441 		error = EHOSTUNREACH;
442 	if (error)
443 		goto done;
444 	/*
445 	 * Create a new entry if we just got back a wildcard entry
446 	 * or the lookup failed.  This is necessary for hosts
447 	 * which use routing redirects generated by smart gateways
448 	 * to dynamically build the routing tables.
449 	 */
450 	if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
451 		goto create;
452 	/*
453 	 * Don't listen to the redirect if it's
454 	 * for a route to an interface.
455 	 */
456 	if (rt->rt_flags & RTF_GATEWAY) {
457 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
458 			/*
459 			 * Changing from route to net => route to host.
460 			 * Create new route, rather than smashing route to net.
461 			 */
462 		create:
463 			if (rt != NULL)
464 				rtfree(rt);
465 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
466 			info.rti_info[RTAX_DST] = dst;
467 			info.rti_info[RTAX_GATEWAY] = gateway;
468 			info.rti_info[RTAX_NETMASK] = netmask;
469 			info.rti_ifa = ifa;
470 			info.rti_flags = flags;
471 			rt = NULL;
472 			error = rtrequest1(RTM_ADD, &info, &rt);
473 			if (rt != NULL)
474 				flags = rt->rt_flags;
475 			stat = &rtstat.rts_dynamic;
476 		} else {
477 			/*
478 			 * Smash the current notion of the gateway to
479 			 * this destination.  Should check about netmask!!!
480 			 */
481 			rt->rt_flags |= RTF_MODIFIED;
482 			flags |= RTF_MODIFIED;
483 			stat = &rtstat.rts_newgateway;
484 			rt_setgate(rt, gateway);
485 		}
486 	} else
487 		error = EHOSTUNREACH;
488 done:
489 	if (rt) {
490 		if (rtp != NULL && !error)
491 			*rtp = rt;
492 		else
493 			rtfree(rt);
494 	}
495 out:
496 	if (error)
497 		rtstat.rts_badredirect++;
498 	else if (stat != NULL)
499 		(*stat)++;
500 	memset(&info, 0, sizeof(info));
501 	info.rti_info[RTAX_DST] = dst;
502 	info.rti_info[RTAX_GATEWAY] = gateway;
503 	info.rti_info[RTAX_NETMASK] = netmask;
504 	info.rti_info[RTAX_AUTHOR] = src;
505 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
506 }
507 
508 /*
509  * Delete a route and generate a message
510  */
511 static int
512 rtdeletemsg(struct rtentry *rt)
513 {
514 	int error;
515 	struct rt_addrinfo info;
516 
517 	/*
518 	 * Request the new route so that the entry is not actually
519 	 * deleted.  That will allow the information being reported to
520 	 * be accurate (and consistent with route_output()).
521 	 */
522 	memset(&info, 0, sizeof(info));
523 	info.rti_info[RTAX_DST] = rt_getkey(rt);
524 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
525 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
526 	info.rti_flags = rt->rt_flags;
527 	error = rtrequest1(RTM_DELETE, &info, &rt);
528 
529 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
530 
531 	/* Adjust the refcount */
532 	if (error == 0 && rt->rt_refcnt <= 0) {
533 		rt->rt_refcnt++;
534 		rtfree(rt);
535 	}
536 	return error;
537 }
538 
539 static int
540 rtflushclone1(struct rtentry *rt, void *arg)
541 {
542 	struct rtentry *parent;
543 
544 	parent = (struct rtentry *)arg;
545 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
546 		rtdeletemsg(rt);
547 	return 0;
548 }
549 
550 static void
551 rtflushclone(sa_family_t family, struct rtentry *parent)
552 {
553 
554 #ifdef DIAGNOSTIC
555 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
556 		panic("rtflushclone: called with a non-cloning route");
557 #endif
558 	rt_walktree(family, rtflushclone1, (void *)parent);
559 }
560 
561 /*
562  * Routing table ioctl interface.
563  */
564 int
565 rtioctl(u_long req, void *data, struct lwp *l)
566 {
567 	return EOPNOTSUPP;
568 }
569 
570 struct ifaddr *
571 ifa_ifwithroute(int flags, const struct sockaddr *dst,
572 	const struct sockaddr *gateway)
573 {
574 	struct ifaddr *ifa;
575 	if ((flags & RTF_GATEWAY) == 0) {
576 		/*
577 		 * If we are adding a route to an interface,
578 		 * and the interface is a pt to pt link
579 		 * we should search for the destination
580 		 * as our clue to the interface.  Otherwise
581 		 * we can use the local address.
582 		 */
583 		ifa = NULL;
584 		if (flags & RTF_HOST)
585 			ifa = ifa_ifwithdstaddr(dst);
586 		if (ifa == NULL)
587 			ifa = ifa_ifwithaddr(gateway);
588 	} else {
589 		/*
590 		 * If we are adding a route to a remote net
591 		 * or host, the gateway may still be on the
592 		 * other end of a pt to pt link.
593 		 */
594 		ifa = ifa_ifwithdstaddr(gateway);
595 	}
596 	if (ifa == NULL)
597 		ifa = ifa_ifwithnet(gateway);
598 	if (ifa == NULL) {
599 		struct rtentry *rt = rtalloc1(dst, 0);
600 		if (rt == NULL)
601 			return NULL;
602 		rt->rt_refcnt--;
603 		if ((ifa = rt->rt_ifa) == NULL)
604 			return NULL;
605 	}
606 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
607 		struct ifaddr *oifa = ifa;
608 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
609 		if (ifa == 0)
610 			ifa = oifa;
611 	}
612 	return ifa;
613 }
614 
615 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
616 
617 int
618 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
619 	const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
620 {
621 	struct rt_addrinfo info;
622 
623 	memset(&info, 0, sizeof(info));
624 	info.rti_flags = flags;
625 	info.rti_info[RTAX_DST] = dst;
626 	info.rti_info[RTAX_GATEWAY] = gateway;
627 	info.rti_info[RTAX_NETMASK] = netmask;
628 	return rtrequest1(req, &info, ret_nrt);
629 }
630 
631 int
632 rt_getifa(struct rt_addrinfo *info)
633 {
634 	struct ifaddr *ifa;
635 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
636 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
637 	const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
638 	const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
639 	int flags = info->rti_flags;
640 
641 	/*
642 	 * ifp may be specified by sockaddr_dl when protocol address
643 	 * is ambiguous
644 	 */
645 	if (info->rti_ifp == NULL && ifpaddr != NULL
646 	    && ifpaddr->sa_family == AF_LINK &&
647 	    (ifa = ifa_ifwithnet((const struct sockaddr *)ifpaddr)) != NULL)
648 		info->rti_ifp = ifa->ifa_ifp;
649 	if (info->rti_ifa == NULL && ifaaddr != NULL)
650 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
651 	if (info->rti_ifa == NULL) {
652 		const struct sockaddr *sa;
653 
654 		sa = ifaaddr != NULL ? ifaaddr :
655 		    (gateway != NULL ? gateway : dst);
656 		if (sa != NULL && info->rti_ifp != NULL)
657 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
658 		else if (dst != NULL && gateway != NULL)
659 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
660 		else if (sa != NULL)
661 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
662 	}
663 	if ((ifa = info->rti_ifa) == NULL)
664 		return ENETUNREACH;
665 	if (ifa->ifa_getifa != NULL)
666 		info->rti_ifa = ifa = (*ifa->ifa_getifa)(ifa, dst);
667 	if (info->rti_ifp == NULL)
668 		info->rti_ifp = ifa->ifa_ifp;
669 	return 0;
670 }
671 
672 int
673 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
674 {
675 	int s = splsoftnet();
676 	int error = 0;
677 	struct rtentry *rt, *crt;
678 	struct radix_node *rn;
679 	struct radix_node_head *rnh;
680 	struct ifaddr *ifa;
681 	struct sockaddr_storage maskeddst;
682 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
683 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
684 	const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
685 	int flags = info->rti_flags;
686 #define senderr(x) { error = x ; goto bad; }
687 
688 	if ((rnh = rt_tables[dst->sa_family]) == NULL)
689 		senderr(ESRCH);
690 	if (flags & RTF_HOST)
691 		netmask = NULL;
692 	switch (req) {
693 	case RTM_DELETE:
694 		if (netmask) {
695 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
696 			    netmask);
697 			dst = (struct sockaddr *)&maskeddst;
698 		}
699 		if ((rn = rnh->rnh_lookup(dst, netmask, rnh)) == NULL)
700 			senderr(ESRCH);
701 		rt = (struct rtentry *)rn;
702 		if ((rt->rt_flags & RTF_CLONING) != 0) {
703 			/* clean up any cloned children */
704 			rtflushclone(dst->sa_family, rt);
705 		}
706 		if ((rn = rnh->rnh_deladdr(dst, netmask, rnh)) == NULL)
707 			senderr(ESRCH);
708 		if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
709 			panic ("rtrequest delete");
710 		rt = (struct rtentry *)rn;
711 		if (rt->rt_gwroute) {
712 			RTFREE(rt->rt_gwroute);
713 			rt->rt_gwroute = NULL;
714 		}
715 		if (rt->rt_parent) {
716 			rt->rt_parent->rt_refcnt--;
717 			rt->rt_parent = NULL;
718 		}
719 		rt->rt_flags &= ~RTF_UP;
720 		if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
721 			ifa->ifa_rtrequest(RTM_DELETE, rt, info);
722 		rttrash++;
723 		if (ret_nrt)
724 			*ret_nrt = rt;
725 		else if (rt->rt_refcnt <= 0) {
726 			rt->rt_refcnt++;
727 			rtfree(rt);
728 		}
729 		break;
730 
731 	case RTM_RESOLVE:
732 		if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
733 			senderr(EINVAL);
734 		if ((rt->rt_flags & RTF_CLONING) == 0)
735 			senderr(EINVAL);
736 		ifa = rt->rt_ifa;
737 		flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC);
738 		flags |= RTF_CLONED;
739 		gateway = rt->rt_gateway;
740 		flags |= RTF_HOST;
741 		goto makeroute;
742 
743 	case RTM_ADD:
744 		if (info->rti_ifa == NULL && (error = rt_getifa(info)))
745 			senderr(error);
746 		ifa = info->rti_ifa;
747 	makeroute:
748 		/* Already at splsoftnet() so pool_get/pool_put are safe */
749 		rt = pool_get(&rtentry_pool, PR_NOWAIT);
750 		if (rt == NULL)
751 			senderr(ENOBUFS);
752 		Bzero(rt, sizeof(*rt));
753 		rt->rt_flags = RTF_UP | flags;
754 		LIST_INIT(&rt->rt_timer);
755 		RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__, __LINE__,
756 		    (void *)rt->_rt_key);
757 		if (rt_setkey(rt, dst, M_NOWAIT) == NULL ||
758 		    rt_setgate(rt, gateway) != 0) {
759 			pool_put(&rtentry_pool, rt);
760 			senderr(ENOBUFS);
761 		}
762 		RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__, __LINE__,
763 		    (void *)rt->_rt_key);
764 		if (netmask) {
765 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
766 			    netmask);
767 			rt_setkey(rt, (struct sockaddr *)&maskeddst, M_NOWAIT);
768 			RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
769 			    __LINE__, (void *)rt->_rt_key);
770 		} else {
771 			rt_setkey(rt, dst, M_NOWAIT);
772 			RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
773 			    __LINE__, (void *)rt->_rt_key);
774 		}
775 		rt_set_ifa(rt, ifa);
776 		RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
777 		    __LINE__, (void *)rt->_rt_key);
778 		rt->rt_ifp = ifa->ifa_ifp;
779 		if (req == RTM_RESOLVE) {
780 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
781 			rt->rt_parent = *ret_nrt;
782 			rt->rt_parent->rt_refcnt++;
783 		}
784 		RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
785 		    __LINE__, (void *)rt->_rt_key);
786 		rn = rnh->rnh_addaddr(rt_getkey(rt), netmask, rnh,
787 		    rt->rt_nodes);
788 		RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
789 		    __LINE__, (void *)rt->_rt_key);
790 		if (rn == NULL && (crt = rtalloc1(rt_getkey(rt), 0)) != NULL) {
791 			/* overwrite cloned route */
792 			if ((crt->rt_flags & RTF_CLONED) != 0) {
793 				rtdeletemsg(crt);
794 				rn = rnh->rnh_addaddr(rt_getkey(rt),
795 				    netmask, rnh, rt->rt_nodes);
796 			}
797 			RTFREE(crt);
798 			RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
799 			    __LINE__, (void *)rt->_rt_key);
800 		}
801 		RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
802 		    __LINE__, (void *)rt->_rt_key);
803 		if (rn == NULL) {
804 			IFAFREE(ifa);
805 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
806 				rtfree(rt->rt_parent);
807 			if (rt->rt_gwroute)
808 				rtfree(rt->rt_gwroute);
809 			rt_destroy(rt);
810 			pool_put(&rtentry_pool, rt);
811 			senderr(EEXIST);
812 		}
813 		RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
814 		    __LINE__, (void *)rt->_rt_key);
815 		if (ifa->ifa_rtrequest)
816 			ifa->ifa_rtrequest(req, rt, info);
817 		RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
818 		    __LINE__, (void *)rt->_rt_key);
819 		if (ret_nrt) {
820 			*ret_nrt = rt;
821 			rt->rt_refcnt++;
822 		}
823 		if ((rt->rt_flags & RTF_CLONING) != 0) {
824 			/* clean up any cloned children */
825 			rtflushclone(dst->sa_family, rt);
826 		}
827 		rtflushall(dst->sa_family);
828 		break;
829 	case RTM_GET:
830 		if (netmask != NULL) {
831 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
832 			    netmask);
833 			dst = (struct sockaddr *)&maskeddst;
834 		}
835 		rn = rnh->rnh_lookup(dst, netmask, rnh);
836 		if (rn == NULL || (rn->rn_flags & RNF_ROOT) != 0)
837 			senderr(ESRCH);
838 		if (ret_nrt != NULL) {
839 			rt = (struct rtentry *)rn;
840 			*ret_nrt = rt;
841 			rt->rt_refcnt++;
842 		}
843 		break;
844 	}
845 bad:
846 	splx(s);
847 	return error;
848 }
849 
850 int
851 rt_setgate(struct rtentry *rt, const struct sockaddr *gate)
852 {
853 	KASSERT(rt != rt->rt_gwroute);
854 
855 	KASSERT(rt->_rt_key != NULL);
856 	RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
857 	    __LINE__, (void *)rt->_rt_key);
858 
859 	if (rt->rt_gwroute) {
860 		RTFREE(rt->rt_gwroute);
861 		rt->rt_gwroute = NULL;
862 	}
863 	KASSERT(rt->_rt_key != NULL);
864 	RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
865 	    __LINE__, (void *)rt->_rt_key);
866 	if (rt->rt_gateway != NULL)
867 		sockaddr_free(rt->rt_gateway);
868 	KASSERT(rt->_rt_key != NULL);
869 	RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
870 	    __LINE__, (void *)rt->_rt_key);
871 	if ((rt->rt_gateway = sockaddr_dup(gate, M_NOWAIT)) == NULL)
872 		return ENOMEM;
873 	KASSERT(rt->_rt_key != NULL);
874 	RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
875 	    __LINE__, (void *)rt->_rt_key);
876 
877 	if (rt->rt_flags & RTF_GATEWAY) {
878 		KASSERT(rt->_rt_key != NULL);
879 		RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
880 		    __LINE__, (void *)rt->_rt_key);
881 		rt->rt_gwroute = rtalloc1(gate, 1);
882 		/*
883 		 * If we switched gateways, grab the MTU from the new
884 		 * gateway route if the current MTU, if the current MTU is
885 		 * greater than the MTU of gateway.
886 		 * Note that, if the MTU of gateway is 0, we will reset the
887 		 * MTU of the route to run PMTUD again from scratch. XXX
888 		 */
889 		KASSERT(rt->_rt_key != NULL);
890 		RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
891 		    __LINE__, (void *)rt->_rt_key);
892 		if (rt->rt_gwroute
893 		    && !(rt->rt_rmx.rmx_locks & RTV_MTU)
894 		    && rt->rt_rmx.rmx_mtu
895 		    && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
896 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
897 		}
898 	}
899 	KASSERT(rt->_rt_key != NULL);
900 	RT_DPRINTF("%s l.%d: rt->_rt_key = %p\n", __func__,
901 	    __LINE__, (void *)rt->_rt_key);
902 	return 0;
903 }
904 
905 void
906 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
907 	const struct sockaddr *netmask)
908 {
909 	const char *netmaskp = &netmask->sa_data[0],
910 	           *srcp = &src->sa_data[0];
911 	char *dstp = &dst->sa_data[0];
912 	const char *maskend = dstp + MIN(netmask->sa_len, src->sa_len);
913 	const char *srcend = dstp + src->sa_len;
914 
915 	dst->sa_len = src->sa_len;
916 	dst->sa_family = src->sa_family;
917 
918 	while (dstp < maskend)
919 		*dstp++ = *srcp++ & *netmaskp++;
920 	if (dstp < srcend)
921 		memset(dstp, 0, (size_t)(srcend - dstp));
922 }
923 
924 /*
925  * Set up or tear down a routing table entry, normally
926  * for an interface.
927  */
928 int
929 rtinit(struct ifaddr *ifa, int cmd, int flags)
930 {
931 	struct rtentry *rt;
932 	struct sockaddr *dst, *odst;
933 	struct sockaddr_storage maskeddst;
934 	struct rtentry *nrt = NULL;
935 	int error;
936 	struct rt_addrinfo info;
937 
938 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
939 	if (cmd == RTM_DELETE) {
940 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
941 			/* Delete subnet route for this interface */
942 			odst = dst;
943 			dst = (struct sockaddr *)&maskeddst;
944 			rt_maskedcopy(odst, dst, ifa->ifa_netmask);
945 		}
946 		if ((rt = rtalloc1(dst, 0)) != NULL) {
947 			rt->rt_refcnt--;
948 			if (rt->rt_ifa != ifa)
949 				return (flags & RTF_HOST) ? EHOSTUNREACH
950 							: ENETUNREACH;
951 		}
952 	}
953 	memset(&info, 0, sizeof(info));
954 	info.rti_ifa = ifa;
955 	info.rti_flags = flags | ifa->ifa_flags;
956 	info.rti_info[RTAX_DST] = dst;
957 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
958 	/*
959 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
960 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
961 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
962 	 * change it to meet bsdi4 behavior.
963 	 */
964 	info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
965 	error = rtrequest1(cmd, &info, &nrt);
966 	if (cmd == RTM_DELETE && error == 0 && (rt = nrt)) {
967 		rt_newaddrmsg(cmd, ifa, error, nrt);
968 		if (rt->rt_refcnt <= 0) {
969 			rt->rt_refcnt++;
970 			rtfree(rt);
971 		}
972 	}
973 	if (cmd == RTM_ADD && error == 0 && (rt = nrt)) {
974 		rt->rt_refcnt--;
975 		if (rt->rt_ifa != ifa) {
976 			printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
977 				rt->rt_ifa);
978 			if (rt->rt_ifa->ifa_rtrequest)
979 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, NULL);
980 			rt_replace_ifa(rt, ifa);
981 			rt->rt_ifp = ifa->ifa_ifp;
982 			if (ifa->ifa_rtrequest)
983 				ifa->ifa_rtrequest(RTM_ADD, rt, NULL);
984 		}
985 		rt_newaddrmsg(cmd, ifa, error, nrt);
986 	}
987 	return error;
988 }
989 
990 /*
991  * Route timer routines.  These routes allow functions to be called
992  * for various routes at any time.  This is useful in supporting
993  * path MTU discovery and redirect route deletion.
994  *
995  * This is similar to some BSDI internal functions, but it provides
996  * for multiple queues for efficiency's sake...
997  */
998 
999 LIST_HEAD(, rttimer_queue) rttimer_queue_head;
1000 static int rt_init_done = 0;
1001 
1002 #define RTTIMER_CALLOUT(r)	do {					\
1003 		if (r->rtt_func != NULL) {				\
1004 			(*r->rtt_func)(r->rtt_rt, r);			\
1005 		} else {						\
1006 			rtrequest((int) RTM_DELETE,			\
1007 				  rt_getkey(r->rtt_rt),			\
1008 				  0, 0, 0, 0);				\
1009 		}							\
1010 	} while (/*CONSTCOND*/0)
1011 
1012 /*
1013  * Some subtle order problems with domain initialization mean that
1014  * we cannot count on this being run from rt_init before various
1015  * protocol initializations are done.  Therefore, we make sure
1016  * that this is run when the first queue is added...
1017  */
1018 
1019 void
1020 rt_timer_init(void)
1021 {
1022 	assert(rt_init_done == 0);
1023 
1024 	LIST_INIT(&rttimer_queue_head);
1025 	callout_init(&rt_timer_ch, 0);
1026 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1027 	rt_init_done = 1;
1028 }
1029 
1030 struct rttimer_queue *
1031 rt_timer_queue_create(u_int timeout)
1032 {
1033 	struct rttimer_queue *rtq;
1034 
1035 	if (rt_init_done == 0)
1036 		rt_timer_init();
1037 
1038 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
1039 	if (rtq == NULL)
1040 		return NULL;
1041 	Bzero(rtq, sizeof *rtq);
1042 
1043 	rtq->rtq_timeout = timeout;
1044 	rtq->rtq_count = 0;
1045 	TAILQ_INIT(&rtq->rtq_head);
1046 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
1047 
1048 	return rtq;
1049 }
1050 
1051 void
1052 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
1053 {
1054 
1055 	rtq->rtq_timeout = timeout;
1056 }
1057 
1058 void
1059 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy)
1060 {
1061 	struct rttimer *r;
1062 
1063 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
1064 		LIST_REMOVE(r, rtt_link);
1065 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1066 		if (destroy)
1067 			RTTIMER_CALLOUT(r);
1068 		/* we are already at splsoftnet */
1069 		pool_put(&rttimer_pool, r);
1070 		if (rtq->rtq_count > 0)
1071 			rtq->rtq_count--;
1072 		else
1073 			printf("rt_timer_queue_remove_all: "
1074 			    "rtq_count reached 0\n");
1075 	}
1076 }
1077 
1078 void
1079 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
1080 {
1081 
1082 	rt_timer_queue_remove_all(rtq, destroy);
1083 
1084 	LIST_REMOVE(rtq, rtq_link);
1085 
1086 	/*
1087 	 * Caller is responsible for freeing the rttimer_queue structure.
1088 	 */
1089 }
1090 
1091 unsigned long
1092 rt_timer_count(struct rttimer_queue *rtq)
1093 {
1094 	return rtq->rtq_count;
1095 }
1096 
1097 void
1098 rt_timer_remove_all(struct rtentry *rt, int destroy)
1099 {
1100 	struct rttimer *r;
1101 
1102 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
1103 		LIST_REMOVE(r, rtt_link);
1104 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
1105 		if (destroy)
1106 			RTTIMER_CALLOUT(r);
1107 		if (r->rtt_queue->rtq_count > 0)
1108 			r->rtt_queue->rtq_count--;
1109 		else
1110 			printf("rt_timer_remove_all: rtq_count reached 0\n");
1111 		/* we are already at splsoftnet */
1112 		pool_put(&rttimer_pool, r);
1113 	}
1114 }
1115 
1116 int
1117 rt_timer_add(struct rtentry *rt,
1118 	void (*func)(struct rtentry *, struct rttimer *),
1119 	struct rttimer_queue *queue)
1120 {
1121 	struct rttimer *r;
1122 	int s;
1123 
1124 	/*
1125 	 * If there's already a timer with this action, destroy it before
1126 	 * we add a new one.
1127 	 */
1128 	LIST_FOREACH(r, &rt->rt_timer, rtt_link) {
1129 		if (r->rtt_func == func)
1130 			break;
1131 	}
1132 	if (r != NULL) {
1133 		LIST_REMOVE(r, rtt_link);
1134 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
1135 		if (r->rtt_queue->rtq_count > 0)
1136 			r->rtt_queue->rtq_count--;
1137 		else
1138 			printf("rt_timer_add: rtq_count reached 0\n");
1139 	} else {
1140 		s = splsoftnet();
1141 		r = pool_get(&rttimer_pool, PR_NOWAIT);
1142 		splx(s);
1143 		if (r == NULL)
1144 			return ENOBUFS;
1145 	}
1146 
1147 	memset(r, 0, sizeof(*r));
1148 
1149 	r->rtt_rt = rt;
1150 	r->rtt_time = time_uptime;
1151 	r->rtt_func = func;
1152 	r->rtt_queue = queue;
1153 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
1154 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
1155 	r->rtt_queue->rtq_count++;
1156 
1157 	return 0;
1158 }
1159 
1160 /* ARGSUSED */
1161 void
1162 rt_timer_timer(void *arg)
1163 {
1164 	struct rttimer_queue *rtq;
1165 	struct rttimer *r;
1166 	int s;
1167 
1168 	s = splsoftnet();
1169 	LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) {
1170 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
1171 		    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
1172 			LIST_REMOVE(r, rtt_link);
1173 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1174 			RTTIMER_CALLOUT(r);
1175 			pool_put(&rttimer_pool, r);
1176 			if (rtq->rtq_count > 0)
1177 				rtq->rtq_count--;
1178 			else
1179 				printf("rt_timer_timer: rtq_count reached 0\n");
1180 		}
1181 	}
1182 	splx(s);
1183 
1184 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1185 }
1186 
1187 #ifdef RTCACHE_DEBUG
1188 #ifndef	RTCACHE_DEBUG_SIZE
1189 #define	RTCACHE_DEBUG_SIZE (1024 * 1024)
1190 #endif
1191 static const char *cache_caller[RTCACHE_DEBUG_SIZE];
1192 static struct route *cache_entry[RTCACHE_DEBUG_SIZE];
1193 size_t cache_cur;
1194 #endif
1195 
1196 #ifdef RTCACHE_DEBUG
1197 static void
1198 _rtcache_init_debug(const char *caller, struct route *ro, int flag)
1199 #else
1200 static void
1201 _rtcache_init(struct route *ro, int flag)
1202 #endif
1203 {
1204 #ifdef RTCACHE_DEBUG
1205 	size_t i;
1206 	for (i = 0; i < cache_cur; ++i) {
1207 		if (cache_entry[i] == ro)
1208 			panic("Reinit of route %p, initialised from %s", ro, cache_caller[i]);
1209 	}
1210 #endif
1211 
1212 	if (rtcache_getdst(ro) == NULL)
1213 		return;
1214 	ro->ro_rt = rtalloc1(rtcache_getdst(ro), flag);
1215 	if (ro->ro_rt != NULL) {
1216 #ifdef RTCACHE_DEBUG
1217 		if (cache_cur == RTCACHE_DEBUG_SIZE)
1218 			panic("Route cache debug overflow");
1219 		cache_caller[cache_cur] = caller;
1220 		cache_entry[cache_cur] = ro;
1221 		++cache_cur;
1222 #endif
1223 		rtcache(ro);
1224 	}
1225 }
1226 
1227 #ifdef RTCACHE_DEBUG
1228 void
1229 rtcache_init_debug(const char *caller, struct route *ro)
1230 {
1231 	_rtcache_init_debug(caller, ro, 1);
1232 }
1233 
1234 void
1235 rtcache_init_noclone_debug(const char *caller, struct route *ro)
1236 {
1237 	_rtcache_init_debug(caller, ro, 0);
1238 }
1239 
1240 void
1241 rtcache_update(struct route *ro, int clone)
1242 {
1243 	rtcache_clear(ro);
1244 	_rtcache_init_debug(__func__, ro, clone);
1245 }
1246 #else
1247 void
1248 rtcache_init(struct route *ro)
1249 {
1250 	_rtcache_init(ro, 1);
1251 }
1252 
1253 void
1254 rtcache_init_noclone(struct route *ro)
1255 {
1256 	_rtcache_init(ro, 0);
1257 }
1258 
1259 void
1260 rtcache_update(struct route *ro, int clone)
1261 {
1262 	rtcache_clear(ro);
1263 	_rtcache_init(ro, clone);
1264 }
1265 #endif
1266 
1267 #ifdef RTCACHE_DEBUG
1268 void
1269 rtcache_copy_debug(const char *caller, struct route *new_ro, const struct route *old_ro)
1270 #else
1271 void
1272 rtcache_copy(struct route *new_ro, const struct route *old_ro)
1273 #endif
1274 {
1275 	/* XXX i doubt this DTRT any longer --dyoung */
1276 #ifdef RTCACHE_DEBUG
1277 	size_t i;
1278 
1279 	for (i = 0; i < cache_cur; ++i) {
1280 		if (cache_entry[i] == new_ro)
1281 			panic("Copy to initalised route %p (before %s)", new_ro, cache_caller[i]);
1282 	}
1283 #endif
1284 
1285 	if (rtcache_getdst(old_ro) == NULL ||
1286 	    rtcache_setdst(new_ro, rtcache_getdst(old_ro)) != 0)
1287 		return;
1288 	new_ro->ro_rt = old_ro->ro_rt;
1289 	if (new_ro->ro_rt != NULL) {
1290 #ifdef RTCACHE_DEBUG
1291 		if (cache_cur == RTCACHE_DEBUG_SIZE)
1292 			panic("Route cache debug overflow");
1293 		cache_caller[cache_cur] = caller;
1294 		cache_entry[cache_cur] = new_ro;
1295 		++cache_cur;
1296 #endif
1297 		rtcache(new_ro);
1298 		++new_ro->ro_rt->rt_refcnt;
1299 	}
1300 }
1301 
1302 void
1303 rtcache_clear(struct route *ro)
1304 {
1305 #ifdef RTCACHE_DEBUG
1306 	size_t j, i = cache_cur;
1307 	for (i = j = 0; i < cache_cur; ++i, ++j) {
1308 		if (cache_entry[i] == ro) {
1309 			if (ro->ro_rt == NULL)
1310 				panic("Route cache manipulated (allocated by %s)", cache_caller[i]);
1311 			--j;
1312 		} else {
1313 			cache_caller[j] = cache_caller[i];
1314 			cache_entry[j] = cache_entry[i];
1315 		}
1316 	}
1317 	if (ro->ro_rt != NULL) {
1318 		if (i != j + 1)
1319 			panic("Wrong entries after rtcache_free: %zu (expected %zu)", j, i - 1);
1320 		--cache_cur;
1321 	}
1322 #endif
1323 
1324 	if (ro->ro_rt != NULL)
1325 		rtflush(ro);
1326 	ro->ro_rt = NULL;
1327 }
1328 
1329 struct rtentry *
1330 rtcache_lookup2(struct route *ro, const struct sockaddr *dst, int clone,
1331     int *hitp)
1332 {
1333 	const struct sockaddr *odst;
1334 
1335 	odst = rtcache_getdst(ro);
1336 
1337 	if (odst == NULL)
1338 		;
1339 	else if (sockaddr_cmp(odst, dst) != 0)
1340 		rtcache_free(ro);
1341 	else if (rtcache_down(ro))
1342 		rtcache_clear(ro);
1343 
1344 	if (ro->ro_rt == NULL) {
1345 		*hitp = 0;
1346 		rtcache_setdst(ro, dst);
1347 		_rtcache_init(ro, clone);
1348 	} else
1349 		*hitp = 1;
1350 
1351 	return ro->ro_rt;
1352 }
1353 
1354 void
1355 rtcache_free(struct route *ro)
1356 {
1357 	rtcache_clear(ro);
1358 	if (ro->ro_sa != NULL) {
1359 		sockaddr_free(ro->ro_sa);
1360 		ro->ro_sa = NULL;
1361 	}
1362 }
1363 
1364 int
1365 rtcache_setdst(struct route *ro, const struct sockaddr *sa)
1366 {
1367 	KASSERT(sa != NULL);
1368 
1369 	if (ro->ro_sa != NULL && ro->ro_sa->sa_family == sa->sa_family) {
1370 		rtcache_clear(ro);
1371 		if (sockaddr_copy(ro->ro_sa, ro->ro_sa->sa_len, sa) != NULL)
1372 			return 0;
1373 		sockaddr_free(ro->ro_sa);
1374 	} else if (ro->ro_sa != NULL)
1375 		rtcache_free(ro);	/* free ro_sa, wrong family */
1376 
1377 	if ((ro->ro_sa = sockaddr_dup(sa, M_NOWAIT)) == NULL)
1378 		return ENOMEM;
1379 	return 0;
1380 }
1381 
1382 static int
1383 rt_walktree_visitor(struct radix_node *rn, void *v)
1384 {
1385 	struct rtwalk *rw = (struct rtwalk *)v;
1386 
1387 	return (*rw->rw_f)((struct rtentry *)rn, rw->rw_v);
1388 }
1389 
1390 int
1391 rt_walktree(sa_family_t family, int (*f)(struct rtentry *, void *), void *v)
1392 {
1393 	struct radix_node_head *rnh = rt_tables[family];
1394 	struct rtwalk rw;
1395 
1396 	if (rnh == NULL)
1397 		return 0;
1398 
1399 	rw.rw_f = f;
1400 	rw.rw_v = v;
1401 
1402 	return rn_walktree(rnh, rt_walktree_visitor, &rw);
1403 }
1404