xref: /netbsd-src/sys/net/route.c (revision 80d9064ac03cbb6a4174695f0d5b237c8766d3d0)
1 /*	$NetBSD: route.c,v 1.133 2014/09/09 20:16:12 rmind 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 1.133 2014/09/09 20:16:12 rmind 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 static 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 	KASSERT(rt != NULL);
380 	KASSERT(rt->rt_refcnt > 0);
381 
382 	rt->rt_refcnt--;
383 	if (rt->rt_refcnt == 0 && (rt->rt_flags & RTF_UP) == 0) {
384 		rt_assert_inactive(rt);
385 		rttrash--;
386 		rt_timer_remove_all(rt, 0);
387 		ifa = rt->rt_ifa;
388 		rt->rt_ifa = NULL;
389 		ifafree(ifa);
390 		rt->rt_ifp = NULL;
391 		rt_destroy(rt);
392 		pool_put(&rtentry_pool, rt);
393 	}
394 }
395 
396 /*
397  * Force a routing table entry to the specified
398  * destination to go through the given gateway.
399  * Normally called as a result of a routing redirect
400  * message from the network layer.
401  *
402  * N.B.: must be called at splsoftnet
403  */
404 void
405 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
406 	const struct sockaddr *netmask, int flags, const struct sockaddr *src,
407 	struct rtentry **rtp)
408 {
409 	struct rtentry *rt;
410 	int error = 0;
411 	uint64_t *stat = NULL;
412 	struct rt_addrinfo info;
413 	struct ifaddr *ifa;
414 
415 	/* verify the gateway is directly reachable */
416 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
417 		error = ENETUNREACH;
418 		goto out;
419 	}
420 	rt = rtalloc1(dst, 0);
421 	/*
422 	 * If the redirect isn't from our current router for this dst,
423 	 * it's either old or wrong.  If it redirects us to ourselves,
424 	 * we have a routing loop, perhaps as a result of an interface
425 	 * going down recently.
426 	 */
427 	if (!(flags & RTF_DONE) && rt &&
428 	     (sockaddr_cmp(src, rt->rt_gateway) != 0 || rt->rt_ifa != ifa))
429 		error = EINVAL;
430 	else if (ifa_ifwithaddr(gateway))
431 		error = EHOSTUNREACH;
432 	if (error)
433 		goto done;
434 	/*
435 	 * Create a new entry if we just got back a wildcard entry
436 	 * or the lookup failed.  This is necessary for hosts
437 	 * which use routing redirects generated by smart gateways
438 	 * to dynamically build the routing tables.
439 	 */
440 	if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
441 		goto create;
442 	/*
443 	 * Don't listen to the redirect if it's
444 	 * for a route to an interface.
445 	 */
446 	if (rt->rt_flags & RTF_GATEWAY) {
447 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
448 			/*
449 			 * Changing from route to net => route to host.
450 			 * Create new route, rather than smashing route to net.
451 			 */
452 		create:
453 			if (rt != NULL)
454 				rtfree(rt);
455 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
456 			memset(&info, 0, sizeof(info));
457 			info.rti_info[RTAX_DST] = dst;
458 			info.rti_info[RTAX_GATEWAY] = gateway;
459 			info.rti_info[RTAX_NETMASK] = netmask;
460 			info.rti_ifa = ifa;
461 			info.rti_flags = flags;
462 			rt = NULL;
463 			error = rtrequest1(RTM_ADD, &info, &rt);
464 			if (rt != NULL)
465 				flags = rt->rt_flags;
466 			stat = &rtstat.rts_dynamic;
467 		} else {
468 			/*
469 			 * Smash the current notion of the gateway to
470 			 * this destination.  Should check about netmask!!!
471 			 */
472 			rt->rt_flags |= RTF_MODIFIED;
473 			flags |= RTF_MODIFIED;
474 			stat = &rtstat.rts_newgateway;
475 			rt_setgate(rt, gateway);
476 		}
477 	} else
478 		error = EHOSTUNREACH;
479 done:
480 	if (rt) {
481 		if (rtp != NULL && !error)
482 			*rtp = rt;
483 		else
484 			rtfree(rt);
485 	}
486 out:
487 	if (error)
488 		rtstat.rts_badredirect++;
489 	else if (stat != NULL)
490 		(*stat)++;
491 	memset(&info, 0, sizeof(info));
492 	info.rti_info[RTAX_DST] = dst;
493 	info.rti_info[RTAX_GATEWAY] = gateway;
494 	info.rti_info[RTAX_NETMASK] = netmask;
495 	info.rti_info[RTAX_AUTHOR] = src;
496 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
497 }
498 
499 /*
500  * Delete a route and generate a message
501  */
502 static int
503 rtdeletemsg(struct rtentry *rt)
504 {
505 	int error;
506 	struct rt_addrinfo info;
507 
508 	/*
509 	 * Request the new route so that the entry is not actually
510 	 * deleted.  That will allow the information being reported to
511 	 * be accurate (and consistent with route_output()).
512 	 */
513 	memset(&info, 0, sizeof(info));
514 	info.rti_info[RTAX_DST] = rt_getkey(rt);
515 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
516 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
517 	info.rti_flags = rt->rt_flags;
518 	error = rtrequest1(RTM_DELETE, &info, &rt);
519 
520 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
521 
522 	/* Adjust the refcount */
523 	if (error == 0 && rt->rt_refcnt <= 0) {
524 		rt->rt_refcnt++;
525 		rtfree(rt);
526 	}
527 	return error;
528 }
529 
530 static int
531 rtflushclone1(struct rtentry *rt, void *arg)
532 {
533 	struct rtentry *parent;
534 
535 	parent = (struct rtentry *)arg;
536 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
537 		rtdeletemsg(rt);
538 	return 0;
539 }
540 
541 static void
542 rtflushclone(sa_family_t family, struct rtentry *parent)
543 {
544 
545 #ifdef DIAGNOSTIC
546 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
547 		panic("rtflushclone: called with a non-cloning route");
548 #endif
549 	rt_walktree(family, rtflushclone1, (void *)parent);
550 }
551 
552 struct ifaddr *
553 ifa_ifwithroute(int flags, const struct sockaddr *dst,
554 	const struct sockaddr *gateway)
555 {
556 	struct ifaddr *ifa;
557 	if ((flags & RTF_GATEWAY) == 0) {
558 		/*
559 		 * If we are adding a route to an interface,
560 		 * and the interface is a pt to pt link
561 		 * we should search for the destination
562 		 * as our clue to the interface.  Otherwise
563 		 * we can use the local address.
564 		 */
565 		ifa = NULL;
566 		if ((flags & RTF_HOST) && gateway->sa_family != AF_LINK)
567 			ifa = ifa_ifwithdstaddr(dst);
568 		if (ifa == NULL)
569 			ifa = ifa_ifwithaddr(gateway);
570 	} else {
571 		/*
572 		 * If we are adding a route to a remote net
573 		 * or host, the gateway may still be on the
574 		 * other end of a pt to pt link.
575 		 */
576 		ifa = ifa_ifwithdstaddr(gateway);
577 	}
578 	if (ifa == NULL)
579 		ifa = ifa_ifwithnet(gateway);
580 	if (ifa == NULL) {
581 		struct rtentry *rt = rtalloc1(dst, 0);
582 		if (rt == NULL)
583 			return NULL;
584 		rt->rt_refcnt--;
585 		if ((ifa = rt->rt_ifa) == NULL)
586 			return NULL;
587 	}
588 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
589 		struct ifaddr *oifa = ifa;
590 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
591 		if (ifa == NULL)
592 			ifa = oifa;
593 	}
594 	return ifa;
595 }
596 
597 int
598 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
599 	const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
600 {
601 	struct rt_addrinfo info;
602 
603 	memset(&info, 0, sizeof(info));
604 	info.rti_flags = flags;
605 	info.rti_info[RTAX_DST] = dst;
606 	info.rti_info[RTAX_GATEWAY] = gateway;
607 	info.rti_info[RTAX_NETMASK] = netmask;
608 	return rtrequest1(req, &info, ret_nrt);
609 }
610 
611 int
612 rt_getifa(struct rt_addrinfo *info)
613 {
614 	struct ifaddr *ifa;
615 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
616 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
617 	const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
618 	const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
619 	int flags = info->rti_flags;
620 
621 	/*
622 	 * ifp may be specified by sockaddr_dl when protocol address
623 	 * is ambiguous
624 	 */
625 	if (info->rti_ifp == NULL && ifpaddr != NULL
626 	    && ifpaddr->sa_family == AF_LINK &&
627 	    (ifa = ifa_ifwithnet(ifpaddr)) != NULL)
628 		info->rti_ifp = ifa->ifa_ifp;
629 	if (info->rti_ifa == NULL && ifaaddr != NULL)
630 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
631 	if (info->rti_ifa == NULL) {
632 		const struct sockaddr *sa;
633 
634 		sa = ifaaddr != NULL ? ifaaddr :
635 		    (gateway != NULL ? gateway : dst);
636 		if (sa != NULL && info->rti_ifp != NULL)
637 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
638 		else if (dst != NULL && gateway != NULL)
639 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
640 		else if (sa != NULL)
641 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
642 	}
643 	if ((ifa = info->rti_ifa) == NULL)
644 		return ENETUNREACH;
645 	if (ifa->ifa_getifa != NULL)
646 		info->rti_ifa = ifa = (*ifa->ifa_getifa)(ifa, dst);
647 	if (info->rti_ifp == NULL)
648 		info->rti_ifp = ifa->ifa_ifp;
649 	return 0;
650 }
651 
652 int
653 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
654 {
655 	int s = splsoftnet();
656 	int error = 0, rc;
657 	struct rtentry *rt, *crt;
658 	rtbl_t *rtbl;
659 	struct ifaddr *ifa, *ifa2;
660 	struct sockaddr_storage maskeddst;
661 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
662 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
663 	const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
664 	int flags = info->rti_flags;
665 #define senderr(x) { error = x ; goto bad; }
666 
667 	if ((rtbl = rt_gettable(dst->sa_family)) == NULL)
668 		senderr(ESRCH);
669 	if (flags & RTF_HOST)
670 		netmask = NULL;
671 	switch (req) {
672 	case RTM_DELETE:
673 		if (netmask) {
674 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
675 			    netmask);
676 			dst = (struct sockaddr *)&maskeddst;
677 		}
678 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
679 			senderr(ESRCH);
680 		if ((rt->rt_flags & RTF_CLONING) != 0) {
681 			/* clean up any cloned children */
682 			rtflushclone(dst->sa_family, rt);
683 		}
684 		if ((rt = rt_deladdr(rtbl, dst, netmask)) == NULL)
685 			senderr(ESRCH);
686 		if (rt->rt_gwroute) {
687 			rtfree(rt->rt_gwroute);
688 			rt->rt_gwroute = NULL;
689 		}
690 		if (rt->rt_parent) {
691 			rt->rt_parent->rt_refcnt--;
692 			rt->rt_parent = NULL;
693 		}
694 		rt->rt_flags &= ~RTF_UP;
695 		if ((ifa = rt->rt_ifa)) {
696 			if (ifa->ifa_flags & IFA_ROUTE &&
697 			    rt_ifa_connected(rt, ifa)) {
698 				RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
699 				    "deleted IFA_ROUTE\n",
700 				    (void *)rt->_rt_key, (void *)ifa);
701 				ifa->ifa_flags &= ~IFA_ROUTE;
702 			}
703 			if (ifa->ifa_rtrequest)
704 				ifa->ifa_rtrequest(RTM_DELETE, rt, info);
705 		}
706 		rttrash++;
707 		if (ret_nrt)
708 			*ret_nrt = rt;
709 		else if (rt->rt_refcnt <= 0) {
710 			rt->rt_refcnt++;
711 			rtfree(rt);
712 		}
713 		break;
714 
715 	case RTM_RESOLVE:
716 		if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
717 			senderr(EINVAL);
718 		if ((rt->rt_flags & RTF_CLONING) == 0)
719 			senderr(EINVAL);
720 		ifa = rt->rt_ifa;
721 		flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC);
722 		flags |= RTF_CLONED;
723 		gateway = rt->rt_gateway;
724 		flags |= RTF_HOST;
725 		goto makeroute;
726 
727 	case RTM_ADD:
728 		if (info->rti_ifa == NULL && (error = rt_getifa(info)))
729 			senderr(error);
730 		ifa = info->rti_ifa;
731 	makeroute:
732 		/* Already at splsoftnet() so pool_get/pool_put are safe */
733 		rt = pool_get(&rtentry_pool, PR_NOWAIT);
734 		if (rt == NULL)
735 			senderr(ENOBUFS);
736 		memset(rt, 0, sizeof(*rt));
737 		rt->rt_flags = RTF_UP | flags;
738 		LIST_INIT(&rt->rt_timer);
739 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
740 		if (rt_setkey(rt, dst, M_NOWAIT) == NULL ||
741 		    rt_setgate(rt, gateway) != 0) {
742 			pool_put(&rtentry_pool, rt);
743 			senderr(ENOBUFS);
744 		}
745 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
746 		if (netmask) {
747 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
748 			    netmask);
749 			rt_setkey(rt, (struct sockaddr *)&maskeddst, M_NOWAIT);
750 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
751 		} else {
752 			rt_setkey(rt, dst, M_NOWAIT);
753 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
754 		}
755 		rt_set_ifa(rt, ifa);
756 		if (info->rti_info[RTAX_TAG] != NULL)
757 			rt_settag(rt, info->rti_info[RTAX_TAG]);
758 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
759 		if (info->rti_info[RTAX_IFP] != NULL &&
760 		    (ifa2 = ifa_ifwithnet(info->rti_info[RTAX_IFP])) != NULL &&
761 		    ifa2->ifa_ifp != NULL)
762 			rt->rt_ifp = ifa2->ifa_ifp;
763 		else
764 			rt->rt_ifp = ifa->ifa_ifp;
765 		if (req == RTM_RESOLVE) {
766 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
767 			rt->rt_parent = *ret_nrt;
768 			rt->rt_parent->rt_refcnt++;
769 		}
770 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
771 		rc = rt_addaddr(rtbl, rt, netmask);
772 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
773 		if (rc != 0 && (crt = rtalloc1(rt_getkey(rt), 0)) != NULL) {
774 			/* overwrite cloned route */
775 			if ((crt->rt_flags & RTF_CLONED) != 0) {
776 				rtdeletemsg(crt);
777 				rc = rt_addaddr(rtbl, rt, netmask);
778 			}
779 			rtfree(crt);
780 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
781 		}
782 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
783 		if (rc != 0) {
784 			ifafree(ifa);
785 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
786 				rtfree(rt->rt_parent);
787 			if (rt->rt_gwroute)
788 				rtfree(rt->rt_gwroute);
789 			rt_destroy(rt);
790 			pool_put(&rtentry_pool, rt);
791 			senderr(rc);
792 		}
793 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
794 		if (ifa->ifa_rtrequest)
795 			ifa->ifa_rtrequest(req, rt, info);
796 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
797 		if (ret_nrt) {
798 			*ret_nrt = rt;
799 			rt->rt_refcnt++;
800 		}
801 		if ((rt->rt_flags & RTF_CLONING) != 0) {
802 			/* clean up any cloned children */
803 			rtflushclone(dst->sa_family, rt);
804 		}
805 		rtflushall(dst->sa_family);
806 		break;
807 	case RTM_GET:
808 		if (netmask != NULL) {
809 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
810 			    netmask);
811 			dst = (struct sockaddr *)&maskeddst;
812 		}
813 		if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
814 			senderr(ESRCH);
815 		if (ret_nrt != NULL) {
816 			*ret_nrt = rt;
817 			rt->rt_refcnt++;
818 		}
819 		break;
820 	}
821 bad:
822 	splx(s);
823 	return error;
824 }
825 
826 int
827 rt_setgate(struct rtentry *rt, const struct sockaddr *gate)
828 {
829 	KASSERT(rt != rt->rt_gwroute);
830 
831 	KASSERT(rt->_rt_key != NULL);
832 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
833 
834 	if (rt->rt_gwroute) {
835 		rtfree(rt->rt_gwroute);
836 		rt->rt_gwroute = NULL;
837 	}
838 	KASSERT(rt->_rt_key != NULL);
839 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
840 	if (rt->rt_gateway != NULL)
841 		sockaddr_free(rt->rt_gateway);
842 	KASSERT(rt->_rt_key != NULL);
843 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
844 	if ((rt->rt_gateway = sockaddr_dup(gate, M_NOWAIT)) == NULL)
845 		return ENOMEM;
846 	KASSERT(rt->_rt_key != NULL);
847 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
848 
849 	if (rt->rt_flags & RTF_GATEWAY) {
850 		KASSERT(rt->_rt_key != NULL);
851 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
852 		rt->rt_gwroute = rtalloc1(gate, 1);
853 		/*
854 		 * If we switched gateways, grab the MTU from the new
855 		 * gateway route if the current MTU, if the current MTU is
856 		 * greater than the MTU of gateway.
857 		 * Note that, if the MTU of gateway is 0, we will reset the
858 		 * MTU of the route to run PMTUD again from scratch. XXX
859 		 */
860 		KASSERT(rt->_rt_key != NULL);
861 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
862 		if (rt->rt_gwroute
863 		    && !(rt->rt_rmx.rmx_locks & RTV_MTU)
864 		    && rt->rt_rmx.rmx_mtu
865 		    && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
866 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
867 		}
868 	}
869 	KASSERT(rt->_rt_key != NULL);
870 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
871 	return 0;
872 }
873 
874 void
875 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
876 	const struct sockaddr *netmask)
877 {
878 	const char *netmaskp = &netmask->sa_data[0],
879 	           *srcp = &src->sa_data[0];
880 	char *dstp = &dst->sa_data[0];
881 	const char *maskend = (char *)dst + MIN(netmask->sa_len, src->sa_len);
882 	const char *srcend = (char *)dst + src->sa_len;
883 
884 	dst->sa_len = src->sa_len;
885 	dst->sa_family = src->sa_family;
886 
887 	while (dstp < maskend)
888 		*dstp++ = *srcp++ & *netmaskp++;
889 	if (dstp < srcend)
890 		memset(dstp, 0, (size_t)(srcend - dstp));
891 }
892 
893 /*
894  * Set up or tear down a routing table entry, normally
895  * for an interface.
896  */
897 int
898 rtinit(struct ifaddr *ifa, int cmd, int flags)
899 {
900 	struct rtentry *rt;
901 	struct sockaddr *dst, *odst;
902 	struct sockaddr_storage maskeddst;
903 	struct rtentry *nrt = NULL;
904 	int error;
905 	struct rt_addrinfo info;
906 	struct sockaddr_dl *sdl;
907 	const struct sockaddr_dl *ifsdl;
908 
909 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
910 	if (cmd == RTM_DELETE) {
911 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
912 			/* Delete subnet route for this interface */
913 			odst = dst;
914 			dst = (struct sockaddr *)&maskeddst;
915 			rt_maskedcopy(odst, dst, ifa->ifa_netmask);
916 		}
917 		if ((rt = rtalloc1(dst, 0)) != NULL) {
918 			rt->rt_refcnt--;
919 			if (rt->rt_ifa != ifa)
920 				return (flags & RTF_HOST) ? EHOSTUNREACH
921 							: ENETUNREACH;
922 		}
923 	}
924 	memset(&info, 0, sizeof(info));
925 	info.rti_ifa = ifa;
926 	info.rti_flags = flags | ifa->ifa_flags;
927 	info.rti_info[RTAX_DST] = dst;
928 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
929 	/*
930 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
931 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
932 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
933 	 * change it to meet bsdi4 behavior.
934 	 */
935 	if (cmd != RTM_LLINFO_UPD)
936 		info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
937 	error = rtrequest1((cmd == RTM_LLINFO_UPD) ? RTM_GET : cmd, &info,
938 	    &nrt);
939 	if (error != 0 || (rt = nrt) == NULL)
940 		;
941 	else switch (cmd) {
942 	case RTM_DELETE:
943 		rt_newaddrmsg(cmd, ifa, error, nrt);
944 		if (rt->rt_refcnt <= 0) {
945 			rt->rt_refcnt++;
946 			rtfree(rt);
947 		}
948 		break;
949 	case RTM_LLINFO_UPD:
950 		rt->rt_refcnt--;
951 		RT_DPRINTF("%s: updating%s\n", __func__,
952 		    ((rt->rt_flags & RTF_LLINFO) == 0) ? " (no llinfo)" : "");
953 
954 		ifsdl = ifa->ifa_ifp->if_sadl;
955 
956 		if ((rt->rt_flags & RTF_LLINFO) != 0 &&
957 		    (sdl = satosdl(rt->rt_gateway)) != NULL &&
958 		    sdl->sdl_family == AF_LINK &&
959 		    sockaddr_dl_setaddr(sdl, sdl->sdl_len, CLLADDR(ifsdl),
960 		                        ifa->ifa_ifp->if_addrlen) == NULL) {
961 			error = EINVAL;
962 			break;
963 		}
964 
965 		if (cmd == RTM_LLINFO_UPD && ifa->ifa_rtrequest != NULL)
966 			ifa->ifa_rtrequest(RTM_LLINFO_UPD, rt, &info);
967 		rt_newaddrmsg(RTM_CHANGE, ifa, error, nrt);
968 		break;
969 	case RTM_ADD:
970 		rt->rt_refcnt--;
971 		if (rt->rt_ifa != ifa) {
972 			printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
973 				rt->rt_ifa);
974 			if (rt->rt_ifa->ifa_rtrequest != NULL) {
975 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
976 				    &info);
977 			}
978 			rt_replace_ifa(rt, ifa);
979 			rt->rt_ifp = ifa->ifa_ifp;
980 			if (ifa->ifa_rtrequest != NULL)
981 				ifa->ifa_rtrequest(RTM_ADD, rt, &info);
982 		}
983 		rt_newaddrmsg(cmd, ifa, error, nrt);
984 		break;
985 	}
986 	return error;
987 }
988 
989 /*
990  * Route timer routines.  These routes allow functions to be called
991  * for various routes at any time.  This is useful in supporting
992  * path MTU discovery and redirect route deletion.
993  *
994  * This is similar to some BSDI internal functions, but it provides
995  * for multiple queues for efficiency's sake...
996  */
997 
998 LIST_HEAD(, rttimer_queue) rttimer_queue_head;
999 static int rt_init_done = 0;
1000 
1001 #define RTTIMER_CALLOUT(r)	do {					\
1002 		if (r->rtt_func != NULL) {				\
1003 			(*r->rtt_func)(r->rtt_rt, r);			\
1004 		} else {						\
1005 			rtrequest((int) RTM_DELETE,			\
1006 				  rt_getkey(r->rtt_rt),			\
1007 				  0, 0, 0, 0);				\
1008 		}							\
1009 	} while (/*CONSTCOND*/0)
1010 
1011 /*
1012  * Some subtle order problems with domain initialization mean that
1013  * we cannot count on this being run from rt_init before various
1014  * protocol initializations are done.  Therefore, we make sure
1015  * that this is run when the first queue is added...
1016  */
1017 
1018 void
1019 rt_timer_init(void)
1020 {
1021 	assert(rt_init_done == 0);
1022 
1023 	LIST_INIT(&rttimer_queue_head);
1024 	callout_init(&rt_timer_ch, 0);
1025 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1026 	rt_init_done = 1;
1027 }
1028 
1029 struct rttimer_queue *
1030 rt_timer_queue_create(u_int timeout)
1031 {
1032 	struct rttimer_queue *rtq;
1033 
1034 	if (rt_init_done == 0)
1035 		rt_timer_init();
1036 
1037 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
1038 	if (rtq == NULL)
1039 		return NULL;
1040 	memset(rtq, 0, sizeof(*rtq));
1041 
1042 	rtq->rtq_timeout = timeout;
1043 	TAILQ_INIT(&rtq->rtq_head);
1044 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
1045 
1046 	return rtq;
1047 }
1048 
1049 void
1050 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
1051 {
1052 
1053 	rtq->rtq_timeout = timeout;
1054 }
1055 
1056 void
1057 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy)
1058 {
1059 	struct rttimer *r;
1060 
1061 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
1062 		LIST_REMOVE(r, rtt_link);
1063 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1064 		if (destroy)
1065 			RTTIMER_CALLOUT(r);
1066 		/* we are already at splsoftnet */
1067 		pool_put(&rttimer_pool, r);
1068 		if (rtq->rtq_count > 0)
1069 			rtq->rtq_count--;
1070 		else
1071 			printf("rt_timer_queue_remove_all: "
1072 			    "rtq_count reached 0\n");
1073 	}
1074 }
1075 
1076 void
1077 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
1078 {
1079 
1080 	rt_timer_queue_remove_all(rtq, destroy);
1081 
1082 	LIST_REMOVE(rtq, rtq_link);
1083 
1084 	/*
1085 	 * Caller is responsible for freeing the rttimer_queue structure.
1086 	 */
1087 }
1088 
1089 unsigned long
1090 rt_timer_count(struct rttimer_queue *rtq)
1091 {
1092 	return rtq->rtq_count;
1093 }
1094 
1095 void
1096 rt_timer_remove_all(struct rtentry *rt, int destroy)
1097 {
1098 	struct rttimer *r;
1099 
1100 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
1101 		LIST_REMOVE(r, rtt_link);
1102 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
1103 		if (destroy)
1104 			RTTIMER_CALLOUT(r);
1105 		if (r->rtt_queue->rtq_count > 0)
1106 			r->rtt_queue->rtq_count--;
1107 		else
1108 			printf("rt_timer_remove_all: rtq_count reached 0\n");
1109 		/* we are already at splsoftnet */
1110 		pool_put(&rttimer_pool, r);
1111 	}
1112 }
1113 
1114 int
1115 rt_timer_add(struct rtentry *rt,
1116 	void (*func)(struct rtentry *, struct rttimer *),
1117 	struct rttimer_queue *queue)
1118 {
1119 	struct rttimer *r;
1120 	int s;
1121 
1122 	/*
1123 	 * If there's already a timer with this action, destroy it before
1124 	 * we add a new one.
1125 	 */
1126 	LIST_FOREACH(r, &rt->rt_timer, rtt_link) {
1127 		if (r->rtt_func == func)
1128 			break;
1129 	}
1130 	if (r != NULL) {
1131 		LIST_REMOVE(r, rtt_link);
1132 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
1133 		if (r->rtt_queue->rtq_count > 0)
1134 			r->rtt_queue->rtq_count--;
1135 		else
1136 			printf("rt_timer_add: rtq_count reached 0\n");
1137 	} else {
1138 		s = splsoftnet();
1139 		r = pool_get(&rttimer_pool, PR_NOWAIT);
1140 		splx(s);
1141 		if (r == NULL)
1142 			return ENOBUFS;
1143 	}
1144 
1145 	memset(r, 0, sizeof(*r));
1146 
1147 	r->rtt_rt = rt;
1148 	r->rtt_time = time_uptime;
1149 	r->rtt_func = func;
1150 	r->rtt_queue = queue;
1151 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
1152 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
1153 	r->rtt_queue->rtq_count++;
1154 
1155 	return 0;
1156 }
1157 
1158 /* ARGSUSED */
1159 void
1160 rt_timer_timer(void *arg)
1161 {
1162 	struct rttimer_queue *rtq;
1163 	struct rttimer *r;
1164 	int s;
1165 
1166 	s = splsoftnet();
1167 	LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) {
1168 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
1169 		    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
1170 			LIST_REMOVE(r, rtt_link);
1171 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1172 			RTTIMER_CALLOUT(r);
1173 			pool_put(&rttimer_pool, r);
1174 			if (rtq->rtq_count > 0)
1175 				rtq->rtq_count--;
1176 			else
1177 				printf("rt_timer_timer: rtq_count reached 0\n");
1178 		}
1179 	}
1180 	splx(s);
1181 
1182 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1183 }
1184 
1185 static struct rtentry *
1186 _rtcache_init(struct route *ro, int flag)
1187 {
1188 	rtcache_invariants(ro);
1189 	KASSERT(ro->_ro_rt == NULL);
1190 
1191 	if (rtcache_getdst(ro) == NULL)
1192 		return NULL;
1193 	ro->ro_invalid = false;
1194 	if ((ro->_ro_rt = rtalloc1(rtcache_getdst(ro), flag)) != NULL)
1195 		rtcache(ro);
1196 
1197 	rtcache_invariants(ro);
1198 	return ro->_ro_rt;
1199 }
1200 
1201 struct rtentry *
1202 rtcache_init(struct route *ro)
1203 {
1204 	return _rtcache_init(ro, 1);
1205 }
1206 
1207 struct rtentry *
1208 rtcache_init_noclone(struct route *ro)
1209 {
1210 	return _rtcache_init(ro, 0);
1211 }
1212 
1213 struct rtentry *
1214 rtcache_update(struct route *ro, int clone)
1215 {
1216 	rtcache_clear(ro);
1217 	return _rtcache_init(ro, clone);
1218 }
1219 
1220 void
1221 rtcache_copy(struct route *new_ro, const struct route *old_ro)
1222 {
1223 	struct rtentry *rt;
1224 
1225 	KASSERT(new_ro != old_ro);
1226 	rtcache_invariants(new_ro);
1227 	rtcache_invariants(old_ro);
1228 
1229 	if ((rt = rtcache_validate(old_ro)) != NULL)
1230 		rt->rt_refcnt++;
1231 
1232 	if (rtcache_getdst(old_ro) == NULL ||
1233 	    rtcache_setdst(new_ro, rtcache_getdst(old_ro)) != 0)
1234 		return;
1235 
1236 	new_ro->ro_invalid = false;
1237 	if ((new_ro->_ro_rt = rt) != NULL)
1238 		rtcache(new_ro);
1239 	rtcache_invariants(new_ro);
1240 }
1241 
1242 static struct dom_rtlist invalid_routes = LIST_HEAD_INITIALIZER(dom_rtlist);
1243 
1244 void
1245 rtcache_invalidate(struct dom_rtlist *rtlist)
1246 {
1247 	struct route *ro;
1248 
1249 	while ((ro = LIST_FIRST(rtlist)) != NULL) {
1250 		rtcache_invariants(ro);
1251 		KASSERT(ro->_ro_rt != NULL);
1252 		ro->ro_invalid = true;
1253 		LIST_REMOVE(ro, ro_rtcache_next);
1254 		LIST_INSERT_HEAD(&invalid_routes, ro, ro_rtcache_next);
1255 		rtcache_invariants(ro);
1256 	}
1257 }
1258 
1259 void
1260 rtcache_clear(struct route *ro)
1261 {
1262 	rtcache_invariants(ro);
1263 	if (ro->_ro_rt == NULL)
1264 		return;
1265 
1266 	LIST_REMOVE(ro, ro_rtcache_next);
1267 
1268 	rtfree(ro->_ro_rt);
1269 	ro->_ro_rt = NULL;
1270 	ro->ro_invalid = false;
1271 	rtcache_invariants(ro);
1272 }
1273 
1274 struct rtentry *
1275 rtcache_lookup2(struct route *ro, const struct sockaddr *dst, int clone,
1276     int *hitp)
1277 {
1278 	const struct sockaddr *odst;
1279 	struct rtentry *rt = NULL;
1280 
1281 	rtcache_invariants(ro);
1282 
1283 	odst = rtcache_getdst(ro);
1284 
1285 	if (odst == NULL)
1286 		;
1287 	else if (sockaddr_cmp(odst, dst) != 0)
1288 		rtcache_free(ro);
1289 	else if ((rt = rtcache_validate(ro)) == NULL)
1290 		rtcache_clear(ro);
1291 
1292 	if (rt == NULL) {
1293 		*hitp = 0;
1294 		if (rtcache_setdst(ro, dst) == 0)
1295 			rt = _rtcache_init(ro, clone);
1296 	} else
1297 		*hitp = 1;
1298 
1299 	rtcache_invariants(ro);
1300 
1301 	return rt;
1302 }
1303 
1304 void
1305 rtcache_free(struct route *ro)
1306 {
1307 	rtcache_clear(ro);
1308 	if (ro->ro_sa != NULL) {
1309 		sockaddr_free(ro->ro_sa);
1310 		ro->ro_sa = NULL;
1311 	}
1312 	rtcache_invariants(ro);
1313 }
1314 
1315 int
1316 rtcache_setdst(struct route *ro, const struct sockaddr *sa)
1317 {
1318 	KASSERT(sa != NULL);
1319 
1320 	rtcache_invariants(ro);
1321 	if (ro->ro_sa != NULL && ro->ro_sa->sa_family == sa->sa_family) {
1322 		rtcache_clear(ro);
1323 		if (sockaddr_copy(ro->ro_sa, ro->ro_sa->sa_len, sa) != NULL) {
1324 			rtcache_invariants(ro);
1325 			return 0;
1326 		}
1327 		sockaddr_free(ro->ro_sa);
1328 	} else if (ro->ro_sa != NULL)
1329 		rtcache_free(ro);	/* free ro_sa, wrong family */
1330 
1331 	KASSERT(ro->_ro_rt == NULL);
1332 
1333 	if ((ro->ro_sa = sockaddr_dup(sa, M_NOWAIT)) == NULL) {
1334 		rtcache_invariants(ro);
1335 		return ENOMEM;
1336 	}
1337 	rtcache_invariants(ro);
1338 	return 0;
1339 }
1340 
1341 const struct sockaddr *
1342 rt_settag(struct rtentry *rt, const struct sockaddr *tag)
1343 {
1344 	if (rt->rt_tag != tag) {
1345 		if (rt->rt_tag != NULL)
1346 			sockaddr_free(rt->rt_tag);
1347 		rt->rt_tag = sockaddr_dup(tag, M_NOWAIT);
1348 	}
1349 	return rt->rt_tag;
1350 }
1351 
1352 struct sockaddr *
1353 rt_gettag(struct rtentry *rt)
1354 {
1355 	return rt->rt_tag;
1356 }
1357