xref: /netbsd-src/sys/net/route.c (revision 10ad5ffa714ce1a679dcc9dd8159648df2d67b5a)
1 /*	$NetBSD: route.c,v 1.117 2009/04/02 21:02:06 christos 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 1.117 2009/04/02 21:02:06 christos Exp $");
97 
98 #include <sys/param.h>
99 #include <sys/sysctl.h>
100 #include <sys/systm.h>
101 #include <sys/callout.h>
102 #include <sys/proc.h>
103 #include <sys/mbuf.h>
104 #include <sys/socket.h>
105 #include <sys/socketvar.h>
106 #include <sys/domain.h>
107 #include <sys/protosw.h>
108 #include <sys/kernel.h>
109 #include <sys/ioctl.h>
110 #include <sys/pool.h>
111 
112 #include <net/if.h>
113 #include <net/if_dl.h>
114 #include <net/route.h>
115 #include <net/raw_cb.h>
116 
117 #include <netinet/in.h>
118 #include <netinet/in_var.h>
119 
120 #ifdef RTFLUSH_DEBUG
121 #define	rtcache_debug() __predict_false(_rtcache_debug)
122 #else /* RTFLUSH_DEBUG */
123 #define	rtcache_debug() 0
124 #endif /* RTFLUSH_DEBUG */
125 
126 struct	route_cb route_cb;
127 struct	rtstat	rtstat;
128 struct	radix_node_head *rt_tables[AF_MAX+1];
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 int rtdeletemsg(struct rtentry *);
142 static int rtflushclone1(struct rtentry *, void *);
143 static void rtflushclone(sa_family_t family, struct rtentry *);
144 
145 #ifdef RTFLUSH_DEBUG
146 SYSCTL_SETUP(sysctl_net_rtcache_setup, "sysctl net.rtcache.debug setup")
147 {
148 	const struct sysctlnode *rnode;
149 
150 	/* XXX do not duplicate */
151 	if (sysctl_createv(clog, 0, NULL, &rnode, CTLFLAG_PERMANENT,
152 	    CTLTYPE_NODE, "net", NULL, NULL, 0, NULL, 0, CTL_NET, CTL_EOL) != 0)
153 		return;
154 	if (sysctl_createv(clog, 0, &rnode, &rnode, CTLFLAG_PERMANENT,
155 	    CTLTYPE_NODE,
156 	    "rtcache", SYSCTL_DESCR("Route cache related settings"),
157 	    NULL, 0, NULL, 0, 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 void
252 rtable_init(void **table)
253 {
254 	struct domain *dom;
255 	DOMAIN_FOREACH(dom)
256 		if (dom->dom_rtattach)
257 			dom->dom_rtattach(&table[dom->dom_family],
258 			    dom->dom_rtoffset);
259 }
260 
261 void
262 route_init(void)
263 {
264 
265 	pool_init(&rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl",
266 	    NULL, IPL_SOFTNET);
267 	pool_init(&rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl",
268 	    NULL, IPL_SOFTNET);
269 
270 	rt_init();
271 	rn_init();	/* initialize all zeroes, all ones, mask table */
272 	rtable_init((void **)rt_tables);
273 }
274 
275 void
276 rtflushall(int family)
277 {
278 	struct domain *dom;
279 
280 	if (rtcache_debug())
281 		printf("%s: enter\n", __func__);
282 
283 	if ((dom = pffinddomain(family)) == NULL)
284 		return;
285 
286 	rtcache_invalidate(&dom->dom_rtcache);
287 }
288 
289 void
290 rtcache(struct route *ro)
291 {
292 	struct domain *dom;
293 
294 	rtcache_invariants(ro);
295 	KASSERT(ro->_ro_rt != NULL);
296 	KASSERT(ro->ro_invalid == false);
297 	KASSERT(rtcache_getdst(ro) != NULL);
298 
299 	if ((dom = pffinddomain(rtcache_getdst(ro)->sa_family)) == NULL)
300 		return;
301 
302 	LIST_INSERT_HEAD(&dom->dom_rtcache, ro, ro_rtcache_next);
303 	rtcache_invariants(ro);
304 }
305 
306 /*
307  * Packet routing routines.
308  */
309 struct rtentry *
310 rtalloc1(const struct sockaddr *dst, int report)
311 {
312 	struct radix_node_head *rnh = rt_tables[dst->sa_family];
313 	struct rtentry *rt;
314 	struct radix_node *rn;
315 	struct rtentry *newrt = NULL;
316 	struct rt_addrinfo info;
317 	int  s = splsoftnet(), err = 0, msgtype = RTM_MISS;
318 
319 	if (rnh && (rn = rnh->rnh_matchaddr(dst, rnh)) &&
320 	    ((rn->rn_flags & RNF_ROOT) == 0)) {
321 		newrt = rt = (struct rtentry *)rn;
322 		if (report && (rt->rt_flags & RTF_CLONING)) {
323 			err = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
324 			    &newrt);
325 			if (err) {
326 				newrt = rt;
327 				rt->rt_refcnt++;
328 				goto miss;
329 			}
330 			KASSERT(newrt != NULL);
331 			if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) {
332 				msgtype = RTM_RESOLVE;
333 				goto miss;
334 			}
335 			/* Inform listeners of the new route */
336 			memset(&info, 0, sizeof(info));
337 			info.rti_info[RTAX_DST] = rt_getkey(rt);
338 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
339 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
340 			if (rt->rt_ifp != NULL) {
341 				info.rti_info[RTAX_IFP] =
342 				    rt->rt_ifp->if_dl->ifa_addr;
343 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
344 			}
345 			rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0);
346 		} else
347 			rt->rt_refcnt++;
348 	} else {
349 		rtstat.rts_unreach++;
350 	miss:	if (report) {
351 			memset((void *)&info, 0, sizeof(info));
352 			info.rti_info[RTAX_DST] = dst;
353 			rt_missmsg(msgtype, &info, 0, err);
354 		}
355 	}
356 	splx(s);
357 	return newrt;
358 }
359 
360 void
361 rtfree(struct rtentry *rt)
362 {
363 	struct ifaddr *ifa;
364 
365 	if (rt == NULL)
366 		panic("rtfree");
367 	rt->rt_refcnt--;
368 	if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
369 		if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
370 			panic ("rtfree 2");
371 		rttrash--;
372 		if (rt->rt_refcnt < 0) {
373 			printf("rtfree: %p not freed (neg refs)\n", rt);
374 			return;
375 		}
376 		rt_timer_remove_all(rt, 0);
377 		ifa = rt->rt_ifa;
378 		rt->rt_ifa = NULL;
379 		IFAFREE(ifa);
380 		rt->rt_ifp = NULL;
381 		rt_destroy(rt);
382 		pool_put(&rtentry_pool, rt);
383 	}
384 }
385 
386 void
387 ifafree(struct ifaddr *ifa)
388 {
389 
390 #ifdef DIAGNOSTIC
391 	if (ifa == NULL)
392 		panic("ifafree: null ifa");
393 	if (ifa->ifa_refcnt != 0)
394 		panic("ifafree: ifa_refcnt != 0 (%d)", ifa->ifa_refcnt);
395 #endif
396 #ifdef IFAREF_DEBUG
397 	printf("ifafree: freeing ifaddr %p\n", ifa);
398 #endif
399 	free(ifa, M_IFADDR);
400 }
401 
402 /*
403  * Force a routing table entry to the specified
404  * destination to go through the given gateway.
405  * Normally called as a result of a routing redirect
406  * message from the network layer.
407  *
408  * N.B.: must be called at splsoftnet
409  */
410 void
411 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
412 	const struct sockaddr *netmask, int flags, const struct sockaddr *src,
413 	struct rtentry **rtp)
414 {
415 	struct rtentry *rt;
416 	int error = 0;
417 	u_quad_t *stat = NULL;
418 	struct rt_addrinfo info;
419 	struct ifaddr *ifa;
420 
421 	/* verify the gateway is directly reachable */
422 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
423 		error = ENETUNREACH;
424 		goto out;
425 	}
426 	rt = rtalloc1(dst, 0);
427 	/*
428 	 * If the redirect isn't from our current router for this dst,
429 	 * it's either old or wrong.  If it redirects us to ourselves,
430 	 * we have a routing loop, perhaps as a result of an interface
431 	 * going down recently.
432 	 */
433 	if (!(flags & RTF_DONE) && rt &&
434 	     (sockaddr_cmp(src, rt->rt_gateway) != 0 || rt->rt_ifa != ifa))
435 		error = EINVAL;
436 	else if (ifa_ifwithaddr(gateway))
437 		error = EHOSTUNREACH;
438 	if (error)
439 		goto done;
440 	/*
441 	 * Create a new entry if we just got back a wildcard entry
442 	 * or the lookup failed.  This is necessary for hosts
443 	 * which use routing redirects generated by smart gateways
444 	 * to dynamically build the routing tables.
445 	 */
446 	if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
447 		goto create;
448 	/*
449 	 * Don't listen to the redirect if it's
450 	 * for a route to an interface.
451 	 */
452 	if (rt->rt_flags & RTF_GATEWAY) {
453 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
454 			/*
455 			 * Changing from route to net => route to host.
456 			 * Create new route, rather than smashing route to net.
457 			 */
458 		create:
459 			if (rt != NULL)
460 				rtfree(rt);
461 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
462 			info.rti_info[RTAX_DST] = dst;
463 			info.rti_info[RTAX_GATEWAY] = gateway;
464 			info.rti_info[RTAX_NETMASK] = netmask;
465 			info.rti_ifa = ifa;
466 			info.rti_flags = flags;
467 			rt = NULL;
468 			error = rtrequest1(RTM_ADD, &info, &rt);
469 			if (rt != NULL)
470 				flags = rt->rt_flags;
471 			stat = &rtstat.rts_dynamic;
472 		} else {
473 			/*
474 			 * Smash the current notion of the gateway to
475 			 * this destination.  Should check about netmask!!!
476 			 */
477 			rt->rt_flags |= RTF_MODIFIED;
478 			flags |= RTF_MODIFIED;
479 			stat = &rtstat.rts_newgateway;
480 			rt_setgate(rt, gateway);
481 		}
482 	} else
483 		error = EHOSTUNREACH;
484 done:
485 	if (rt) {
486 		if (rtp != NULL && !error)
487 			*rtp = rt;
488 		else
489 			rtfree(rt);
490 	}
491 out:
492 	if (error)
493 		rtstat.rts_badredirect++;
494 	else if (stat != NULL)
495 		(*stat)++;
496 	memset(&info, 0, sizeof(info));
497 	info.rti_info[RTAX_DST] = dst;
498 	info.rti_info[RTAX_GATEWAY] = gateway;
499 	info.rti_info[RTAX_NETMASK] = netmask;
500 	info.rti_info[RTAX_AUTHOR] = src;
501 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
502 }
503 
504 /*
505  * Delete a route and generate a message
506  */
507 static int
508 rtdeletemsg(struct rtentry *rt)
509 {
510 	int error;
511 	struct rt_addrinfo info;
512 
513 	/*
514 	 * Request the new route so that the entry is not actually
515 	 * deleted.  That will allow the information being reported to
516 	 * be accurate (and consistent with route_output()).
517 	 */
518 	memset(&info, 0, sizeof(info));
519 	info.rti_info[RTAX_DST] = rt_getkey(rt);
520 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
521 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
522 	info.rti_flags = rt->rt_flags;
523 	error = rtrequest1(RTM_DELETE, &info, &rt);
524 
525 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
526 
527 	/* Adjust the refcount */
528 	if (error == 0 && rt->rt_refcnt <= 0) {
529 		rt->rt_refcnt++;
530 		rtfree(rt);
531 	}
532 	return error;
533 }
534 
535 static int
536 rtflushclone1(struct rtentry *rt, void *arg)
537 {
538 	struct rtentry *parent;
539 
540 	parent = (struct rtentry *)arg;
541 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
542 		rtdeletemsg(rt);
543 	return 0;
544 }
545 
546 static void
547 rtflushclone(sa_family_t family, struct rtentry *parent)
548 {
549 
550 #ifdef DIAGNOSTIC
551 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
552 		panic("rtflushclone: called with a non-cloning route");
553 #endif
554 	rt_walktree(family, rtflushclone1, (void *)parent);
555 }
556 
557 /*
558  * Routing table ioctl interface.
559  */
560 int
561 rtioctl(u_long req, void *data, struct lwp *l)
562 {
563 	return EOPNOTSUPP;
564 }
565 
566 struct ifaddr *
567 ifa_ifwithroute(int flags, const struct sockaddr *dst,
568 	const struct sockaddr *gateway)
569 {
570 	struct ifaddr *ifa;
571 	if ((flags & RTF_GATEWAY) == 0) {
572 		/*
573 		 * If we are adding a route to an interface,
574 		 * and the interface is a pt to pt link
575 		 * we should search for the destination
576 		 * as our clue to the interface.  Otherwise
577 		 * we can use the local address.
578 		 */
579 		ifa = NULL;
580 		if (flags & RTF_HOST)
581 			ifa = ifa_ifwithdstaddr(dst);
582 		if (ifa == NULL)
583 			ifa = ifa_ifwithaddr(gateway);
584 	} else {
585 		/*
586 		 * If we are adding a route to a remote net
587 		 * or host, the gateway may still be on the
588 		 * other end of a pt to pt link.
589 		 */
590 		ifa = ifa_ifwithdstaddr(gateway);
591 	}
592 	if (ifa == NULL)
593 		ifa = ifa_ifwithnet(gateway);
594 	if (ifa == NULL) {
595 		struct rtentry *rt = rtalloc1(dst, 0);
596 		if (rt == NULL)
597 			return NULL;
598 		rt->rt_refcnt--;
599 		if ((ifa = rt->rt_ifa) == NULL)
600 			return NULL;
601 	}
602 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
603 		struct ifaddr *oifa = ifa;
604 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
605 		if (ifa == 0)
606 			ifa = oifa;
607 	}
608 	return ifa;
609 }
610 
611 int
612 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
613 	const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
614 {
615 	struct rt_addrinfo info;
616 
617 	memset(&info, 0, sizeof(info));
618 	info.rti_flags = flags;
619 	info.rti_info[RTAX_DST] = dst;
620 	info.rti_info[RTAX_GATEWAY] = gateway;
621 	info.rti_info[RTAX_NETMASK] = netmask;
622 	return rtrequest1(req, &info, ret_nrt);
623 }
624 
625 int
626 rt_getifa(struct rt_addrinfo *info)
627 {
628 	struct ifaddr *ifa;
629 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
630 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
631 	const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
632 	const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
633 	int flags = info->rti_flags;
634 
635 	/*
636 	 * ifp may be specified by sockaddr_dl when protocol address
637 	 * is ambiguous
638 	 */
639 	if (info->rti_ifp == NULL && ifpaddr != NULL
640 	    && ifpaddr->sa_family == AF_LINK &&
641 	    (ifa = ifa_ifwithnet(ifpaddr)) != NULL)
642 		info->rti_ifp = ifa->ifa_ifp;
643 	if (info->rti_ifa == NULL && ifaaddr != NULL)
644 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
645 	if (info->rti_ifa == NULL) {
646 		const struct sockaddr *sa;
647 
648 		sa = ifaaddr != NULL ? ifaaddr :
649 		    (gateway != NULL ? gateway : dst);
650 		if (sa != NULL && info->rti_ifp != NULL)
651 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
652 		else if (dst != NULL && gateway != NULL)
653 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
654 		else if (sa != NULL)
655 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
656 	}
657 	if ((ifa = info->rti_ifa) == NULL)
658 		return ENETUNREACH;
659 	if (ifa->ifa_getifa != NULL)
660 		info->rti_ifa = ifa = (*ifa->ifa_getifa)(ifa, dst);
661 	if (info->rti_ifp == NULL)
662 		info->rti_ifp = ifa->ifa_ifp;
663 	return 0;
664 }
665 
666 int
667 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
668 {
669 	int s = splsoftnet();
670 	int error = 0;
671 	struct rtentry *rt, *crt;
672 	struct radix_node *rn;
673 	struct radix_node_head *rnh;
674 	struct ifaddr *ifa;
675 	struct sockaddr_storage maskeddst;
676 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
677 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
678 	const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
679 	int flags = info->rti_flags;
680 #define senderr(x) { error = x ; goto bad; }
681 
682 	if ((rnh = rt_tables[dst->sa_family]) == NULL)
683 		senderr(ESRCH);
684 	if (flags & RTF_HOST)
685 		netmask = NULL;
686 	switch (req) {
687 	case RTM_DELETE:
688 		if (netmask) {
689 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
690 			    netmask);
691 			dst = (struct sockaddr *)&maskeddst;
692 		}
693 		if ((rn = rnh->rnh_lookup(dst, netmask, rnh)) == NULL)
694 			senderr(ESRCH);
695 		rt = (struct rtentry *)rn;
696 		if ((rt->rt_flags & RTF_CLONING) != 0) {
697 			/* clean up any cloned children */
698 			rtflushclone(dst->sa_family, rt);
699 		}
700 		if ((rn = rnh->rnh_deladdr(dst, netmask, rnh)) == NULL)
701 			senderr(ESRCH);
702 		if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
703 			panic ("rtrequest delete");
704 		rt = (struct rtentry *)rn;
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 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
776 		rt->rt_ifp = ifa->ifa_ifp;
777 		if (req == RTM_RESOLVE) {
778 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
779 			rt->rt_parent = *ret_nrt;
780 			rt->rt_parent->rt_refcnt++;
781 		}
782 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
783 		rn = rnh->rnh_addaddr(rt_getkey(rt), netmask, rnh,
784 		    rt->rt_nodes);
785 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
786 		if (rn == NULL && (crt = rtalloc1(rt_getkey(rt), 0)) != NULL) {
787 			/* overwrite cloned route */
788 			if ((crt->rt_flags & RTF_CLONED) != 0) {
789 				rtdeletemsg(crt);
790 				rn = rnh->rnh_addaddr(rt_getkey(rt),
791 				    netmask, rnh, rt->rt_nodes);
792 			}
793 			RTFREE(crt);
794 			RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
795 		}
796 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
797 		if (rn == NULL) {
798 			IFAFREE(ifa);
799 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
800 				rtfree(rt->rt_parent);
801 			if (rt->rt_gwroute)
802 				rtfree(rt->rt_gwroute);
803 			rt_destroy(rt);
804 			pool_put(&rtentry_pool, rt);
805 			senderr(EEXIST);
806 		}
807 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
808 		if (ifa->ifa_rtrequest)
809 			ifa->ifa_rtrequest(req, rt, info);
810 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
811 		if (ret_nrt) {
812 			*ret_nrt = rt;
813 			rt->rt_refcnt++;
814 		}
815 		if ((rt->rt_flags & RTF_CLONING) != 0) {
816 			/* clean up any cloned children */
817 			rtflushclone(dst->sa_family, rt);
818 		}
819 		rtflushall(dst->sa_family);
820 		break;
821 	case RTM_GET:
822 		if (netmask != NULL) {
823 			rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
824 			    netmask);
825 			dst = (struct sockaddr *)&maskeddst;
826 		}
827 		rn = rnh->rnh_lookup(dst, netmask, rnh);
828 		if (rn == NULL || (rn->rn_flags & RNF_ROOT) != 0)
829 			senderr(ESRCH);
830 		if (ret_nrt != NULL) {
831 			rt = (struct rtentry *)rn;
832 			*ret_nrt = rt;
833 			rt->rt_refcnt++;
834 		}
835 		break;
836 	}
837 bad:
838 	splx(s);
839 	return error;
840 }
841 
842 int
843 rt_setgate(struct rtentry *rt, const struct sockaddr *gate)
844 {
845 	KASSERT(rt != rt->rt_gwroute);
846 
847 	KASSERT(rt->_rt_key != NULL);
848 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
849 
850 	if (rt->rt_gwroute) {
851 		RTFREE(rt->rt_gwroute);
852 		rt->rt_gwroute = NULL;
853 	}
854 	KASSERT(rt->_rt_key != NULL);
855 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
856 	if (rt->rt_gateway != NULL)
857 		sockaddr_free(rt->rt_gateway);
858 	KASSERT(rt->_rt_key != NULL);
859 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
860 	if ((rt->rt_gateway = sockaddr_dup(gate, M_NOWAIT)) == NULL)
861 		return ENOMEM;
862 	KASSERT(rt->_rt_key != NULL);
863 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
864 
865 	if (rt->rt_flags & RTF_GATEWAY) {
866 		KASSERT(rt->_rt_key != NULL);
867 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
868 		rt->rt_gwroute = rtalloc1(gate, 1);
869 		/*
870 		 * If we switched gateways, grab the MTU from the new
871 		 * gateway route if the current MTU, if the current MTU is
872 		 * greater than the MTU of gateway.
873 		 * Note that, if the MTU of gateway is 0, we will reset the
874 		 * MTU of the route to run PMTUD again from scratch. XXX
875 		 */
876 		KASSERT(rt->_rt_key != NULL);
877 		RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
878 		if (rt->rt_gwroute
879 		    && !(rt->rt_rmx.rmx_locks & RTV_MTU)
880 		    && rt->rt_rmx.rmx_mtu
881 		    && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
882 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
883 		}
884 	}
885 	KASSERT(rt->_rt_key != NULL);
886 	RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
887 	return 0;
888 }
889 
890 void
891 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
892 	const struct sockaddr *netmask)
893 {
894 	const char *netmaskp = &netmask->sa_data[0],
895 	           *srcp = &src->sa_data[0];
896 	char *dstp = &dst->sa_data[0];
897 	const char *maskend = dstp + MIN(netmask->sa_len, src->sa_len);
898 	const char *srcend = dstp + src->sa_len;
899 
900 	dst->sa_len = src->sa_len;
901 	dst->sa_family = src->sa_family;
902 
903 	while (dstp < maskend)
904 		*dstp++ = *srcp++ & *netmaskp++;
905 	if (dstp < srcend)
906 		memset(dstp, 0, (size_t)(srcend - dstp));
907 }
908 
909 /*
910  * Set up or tear down a routing table entry, normally
911  * for an interface.
912  */
913 int
914 rtinit(struct ifaddr *ifa, int cmd, int flags)
915 {
916 	struct rtentry *rt;
917 	struct sockaddr *dst, *odst;
918 	struct sockaddr_storage maskeddst;
919 	struct rtentry *nrt = NULL;
920 	int error;
921 	struct rt_addrinfo info;
922 	struct sockaddr_dl *sdl;
923 	const struct sockaddr_dl *ifsdl;
924 
925 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
926 	if (cmd == RTM_DELETE) {
927 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
928 			/* Delete subnet route for this interface */
929 			odst = dst;
930 			dst = (struct sockaddr *)&maskeddst;
931 			rt_maskedcopy(odst, dst, ifa->ifa_netmask);
932 		}
933 		if ((rt = rtalloc1(dst, 0)) != NULL) {
934 			rt->rt_refcnt--;
935 			if (rt->rt_ifa != ifa)
936 				return (flags & RTF_HOST) ? EHOSTUNREACH
937 							: ENETUNREACH;
938 		}
939 	}
940 	memset(&info, 0, sizeof(info));
941 	info.rti_ifa = ifa;
942 	info.rti_flags = flags | ifa->ifa_flags;
943 	info.rti_info[RTAX_DST] = dst;
944 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
945 	/*
946 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
947 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
948 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
949 	 * change it to meet bsdi4 behavior.
950 	 */
951 	if (cmd != RTM_LLINFO_UPD)
952 		info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
953 	error = rtrequest1((cmd == RTM_LLINFO_UPD) ? RTM_GET : cmd, &info,
954 	    &nrt);
955 	if (error != 0 || (rt = nrt) == NULL)
956 		;
957 	else switch (cmd) {
958 	case RTM_DELETE:
959 		rt_newaddrmsg(cmd, ifa, error, nrt);
960 		if (rt->rt_refcnt <= 0) {
961 			rt->rt_refcnt++;
962 			rtfree(rt);
963 		}
964 		break;
965 	case RTM_LLINFO_UPD:
966 		rt->rt_refcnt--;
967 		RT_DPRINTF("%s: updating%s\n", __func__,
968 		    ((rt->rt_flags & RTF_LLINFO) == 0) ? " (no llinfo)" : "");
969 
970 		ifsdl = ifa->ifa_ifp->if_sadl;
971 
972 		if ((rt->rt_flags & RTF_LLINFO) != 0 &&
973 		    (sdl = satosdl(rt->rt_gateway)) != NULL &&
974 		    sdl->sdl_family == AF_LINK &&
975 		    sockaddr_dl_setaddr(sdl, sdl->sdl_len, CLLADDR(ifsdl),
976 		                        ifa->ifa_ifp->if_addrlen) == NULL) {
977 			error = EINVAL;
978 			break;
979 		}
980 
981 		if (cmd == RTM_LLINFO_UPD && ifa->ifa_rtrequest != NULL)
982 			ifa->ifa_rtrequest(RTM_LLINFO_UPD, rt, &info);
983 		rt_newaddrmsg(RTM_CHANGE, ifa, error, nrt);
984 		break;
985 	case RTM_ADD:
986 		rt->rt_refcnt--;
987 		if (rt->rt_ifa != ifa) {
988 			printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
989 				rt->rt_ifa);
990 			if (rt->rt_ifa->ifa_rtrequest != NULL) {
991 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
992 				    &info);
993 			}
994 			rt_replace_ifa(rt, ifa);
995 			rt->rt_ifp = ifa->ifa_ifp;
996 			if (ifa->ifa_rtrequest != NULL)
997 				ifa->ifa_rtrequest(RTM_ADD, rt, &info);
998 		}
999 		rt_newaddrmsg(cmd, ifa, error, nrt);
1000 		break;
1001 	}
1002 	return error;
1003 }
1004 
1005 /*
1006  * Route timer routines.  These routes allow functions to be called
1007  * for various routes at any time.  This is useful in supporting
1008  * path MTU discovery and redirect route deletion.
1009  *
1010  * This is similar to some BSDI internal functions, but it provides
1011  * for multiple queues for efficiency's sake...
1012  */
1013 
1014 LIST_HEAD(, rttimer_queue) rttimer_queue_head;
1015 static int rt_init_done = 0;
1016 
1017 #define RTTIMER_CALLOUT(r)	do {					\
1018 		if (r->rtt_func != NULL) {				\
1019 			(*r->rtt_func)(r->rtt_rt, r);			\
1020 		} else {						\
1021 			rtrequest((int) RTM_DELETE,			\
1022 				  rt_getkey(r->rtt_rt),			\
1023 				  0, 0, 0, 0);				\
1024 		}							\
1025 	} while (/*CONSTCOND*/0)
1026 
1027 /*
1028  * Some subtle order problems with domain initialization mean that
1029  * we cannot count on this being run from rt_init before various
1030  * protocol initializations are done.  Therefore, we make sure
1031  * that this is run when the first queue is added...
1032  */
1033 
1034 void
1035 rt_timer_init(void)
1036 {
1037 	assert(rt_init_done == 0);
1038 
1039 	LIST_INIT(&rttimer_queue_head);
1040 	callout_init(&rt_timer_ch, 0);
1041 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1042 	rt_init_done = 1;
1043 }
1044 
1045 struct rttimer_queue *
1046 rt_timer_queue_create(u_int timeout)
1047 {
1048 	struct rttimer_queue *rtq;
1049 
1050 	if (rt_init_done == 0)
1051 		rt_timer_init();
1052 
1053 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
1054 	if (rtq == NULL)
1055 		return NULL;
1056 	memset(rtq, 0, sizeof(*rtq));
1057 
1058 	rtq->rtq_timeout = timeout;
1059 	TAILQ_INIT(&rtq->rtq_head);
1060 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
1061 
1062 	return rtq;
1063 }
1064 
1065 void
1066 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
1067 {
1068 
1069 	rtq->rtq_timeout = timeout;
1070 }
1071 
1072 void
1073 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy)
1074 {
1075 	struct rttimer *r;
1076 
1077 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
1078 		LIST_REMOVE(r, rtt_link);
1079 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1080 		if (destroy)
1081 			RTTIMER_CALLOUT(r);
1082 		/* we are already at splsoftnet */
1083 		pool_put(&rttimer_pool, r);
1084 		if (rtq->rtq_count > 0)
1085 			rtq->rtq_count--;
1086 		else
1087 			printf("rt_timer_queue_remove_all: "
1088 			    "rtq_count reached 0\n");
1089 	}
1090 }
1091 
1092 void
1093 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
1094 {
1095 
1096 	rt_timer_queue_remove_all(rtq, destroy);
1097 
1098 	LIST_REMOVE(rtq, rtq_link);
1099 
1100 	/*
1101 	 * Caller is responsible for freeing the rttimer_queue structure.
1102 	 */
1103 }
1104 
1105 unsigned long
1106 rt_timer_count(struct rttimer_queue *rtq)
1107 {
1108 	return rtq->rtq_count;
1109 }
1110 
1111 void
1112 rt_timer_remove_all(struct rtentry *rt, int destroy)
1113 {
1114 	struct rttimer *r;
1115 
1116 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
1117 		LIST_REMOVE(r, rtt_link);
1118 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
1119 		if (destroy)
1120 			RTTIMER_CALLOUT(r);
1121 		if (r->rtt_queue->rtq_count > 0)
1122 			r->rtt_queue->rtq_count--;
1123 		else
1124 			printf("rt_timer_remove_all: rtq_count reached 0\n");
1125 		/* we are already at splsoftnet */
1126 		pool_put(&rttimer_pool, r);
1127 	}
1128 }
1129 
1130 int
1131 rt_timer_add(struct rtentry *rt,
1132 	void (*func)(struct rtentry *, struct rttimer *),
1133 	struct rttimer_queue *queue)
1134 {
1135 	struct rttimer *r;
1136 	int s;
1137 
1138 	/*
1139 	 * If there's already a timer with this action, destroy it before
1140 	 * we add a new one.
1141 	 */
1142 	LIST_FOREACH(r, &rt->rt_timer, rtt_link) {
1143 		if (r->rtt_func == func)
1144 			break;
1145 	}
1146 	if (r != NULL) {
1147 		LIST_REMOVE(r, rtt_link);
1148 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
1149 		if (r->rtt_queue->rtq_count > 0)
1150 			r->rtt_queue->rtq_count--;
1151 		else
1152 			printf("rt_timer_add: rtq_count reached 0\n");
1153 	} else {
1154 		s = splsoftnet();
1155 		r = pool_get(&rttimer_pool, PR_NOWAIT);
1156 		splx(s);
1157 		if (r == NULL)
1158 			return ENOBUFS;
1159 	}
1160 
1161 	memset(r, 0, sizeof(*r));
1162 
1163 	r->rtt_rt = rt;
1164 	r->rtt_time = time_uptime;
1165 	r->rtt_func = func;
1166 	r->rtt_queue = queue;
1167 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
1168 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
1169 	r->rtt_queue->rtq_count++;
1170 
1171 	return 0;
1172 }
1173 
1174 /* ARGSUSED */
1175 void
1176 rt_timer_timer(void *arg)
1177 {
1178 	struct rttimer_queue *rtq;
1179 	struct rttimer *r;
1180 	int s;
1181 
1182 	s = splsoftnet();
1183 	LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) {
1184 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
1185 		    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
1186 			LIST_REMOVE(r, rtt_link);
1187 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1188 			RTTIMER_CALLOUT(r);
1189 			pool_put(&rttimer_pool, r);
1190 			if (rtq->rtq_count > 0)
1191 				rtq->rtq_count--;
1192 			else
1193 				printf("rt_timer_timer: rtq_count reached 0\n");
1194 		}
1195 	}
1196 	splx(s);
1197 
1198 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1199 }
1200 
1201 static struct rtentry *
1202 _rtcache_init(struct route *ro, int flag)
1203 {
1204 	rtcache_invariants(ro);
1205 	KASSERT(ro->_ro_rt == NULL);
1206 
1207 	if (rtcache_getdst(ro) == NULL)
1208 		return NULL;
1209 	ro->ro_invalid = false;
1210 	if ((ro->_ro_rt = rtalloc1(rtcache_getdst(ro), flag)) != NULL)
1211 		rtcache(ro);
1212 
1213 	rtcache_invariants(ro);
1214 	return ro->_ro_rt;
1215 }
1216 
1217 struct rtentry *
1218 rtcache_init(struct route *ro)
1219 {
1220 	return _rtcache_init(ro, 1);
1221 }
1222 
1223 struct rtentry *
1224 rtcache_init_noclone(struct route *ro)
1225 {
1226 	return _rtcache_init(ro, 0);
1227 }
1228 
1229 struct rtentry *
1230 rtcache_update(struct route *ro, int clone)
1231 {
1232 	rtcache_clear(ro);
1233 	return _rtcache_init(ro, clone);
1234 }
1235 
1236 void
1237 rtcache_copy(struct route *new_ro, const struct route *old_ro)
1238 {
1239 	struct rtentry *rt;
1240 
1241 	KASSERT(new_ro != old_ro);
1242 	rtcache_invariants(new_ro);
1243 	rtcache_invariants(old_ro);
1244 
1245 	if ((rt = rtcache_validate(old_ro)) != NULL)
1246 		rt->rt_refcnt++;
1247 
1248 	if (rtcache_getdst(old_ro) == NULL ||
1249 	    rtcache_setdst(new_ro, rtcache_getdst(old_ro)) != 0)
1250 		return;
1251 
1252 	new_ro->ro_invalid = false;
1253 	if ((new_ro->_ro_rt = rt) != NULL)
1254 		rtcache(new_ro);
1255 	rtcache_invariants(new_ro);
1256 }
1257 
1258 static struct dom_rtlist invalid_routes = LIST_HEAD_INITIALIZER(dom_rtlist);
1259 
1260 void
1261 rtcache_invalidate(struct dom_rtlist *rtlist)
1262 {
1263 	struct route *ro;
1264 
1265 	while ((ro = LIST_FIRST(rtlist)) != NULL) {
1266 		rtcache_invariants(ro);
1267 		KASSERT(ro->_ro_rt != NULL);
1268 		ro->ro_invalid = true;
1269 		LIST_REMOVE(ro, ro_rtcache_next);
1270 		LIST_INSERT_HEAD(&invalid_routes, ro, ro_rtcache_next);
1271 		rtcache_invariants(ro);
1272 	}
1273 }
1274 
1275 void
1276 rtcache_clear(struct route *ro)
1277 {
1278 	rtcache_invariants(ro);
1279 	if (ro->_ro_rt == NULL)
1280 		return;
1281 
1282 	LIST_REMOVE(ro, ro_rtcache_next);
1283 
1284 	RTFREE(ro->_ro_rt);
1285 	ro->_ro_rt = NULL;
1286 	ro->ro_invalid = false;
1287 	rtcache_invariants(ro);
1288 }
1289 
1290 struct rtentry *
1291 rtcache_lookup2(struct route *ro, const struct sockaddr *dst, int clone,
1292     int *hitp)
1293 {
1294 	const struct sockaddr *odst;
1295 	struct rtentry *rt = NULL;
1296 
1297 	rtcache_invariants(ro);
1298 
1299 	odst = rtcache_getdst(ro);
1300 
1301 	if (odst == NULL)
1302 		;
1303 	else if (sockaddr_cmp(odst, dst) != 0)
1304 		rtcache_free(ro);
1305 	else if ((rt = rtcache_validate(ro)) == NULL)
1306 		rtcache_clear(ro);
1307 
1308 	if (rt == NULL) {
1309 		*hitp = 0;
1310 		if (rtcache_setdst(ro, dst) == 0)
1311 			rt = _rtcache_init(ro, clone);
1312 	} else
1313 		*hitp = 1;
1314 
1315 	rtcache_invariants(ro);
1316 
1317 	return rt;
1318 }
1319 
1320 void
1321 rtcache_free(struct route *ro)
1322 {
1323 	rtcache_clear(ro);
1324 	if (ro->ro_sa != NULL) {
1325 		sockaddr_free(ro->ro_sa);
1326 		ro->ro_sa = NULL;
1327 	}
1328 	rtcache_invariants(ro);
1329 }
1330 
1331 int
1332 rtcache_setdst(struct route *ro, const struct sockaddr *sa)
1333 {
1334 	KASSERT(sa != NULL);
1335 
1336 	rtcache_invariants(ro);
1337 	if (ro->ro_sa != NULL && ro->ro_sa->sa_family == sa->sa_family) {
1338 		rtcache_clear(ro);
1339 		if (sockaddr_copy(ro->ro_sa, ro->ro_sa->sa_len, sa) != NULL) {
1340 			rtcache_invariants(ro);
1341 			return 0;
1342 		}
1343 		sockaddr_free(ro->ro_sa);
1344 	} else if (ro->ro_sa != NULL)
1345 		rtcache_free(ro);	/* free ro_sa, wrong family */
1346 
1347 	KASSERT(ro->_ro_rt == NULL);
1348 
1349 	if ((ro->ro_sa = sockaddr_dup(sa, M_NOWAIT)) == NULL) {
1350 		rtcache_invariants(ro);
1351 		return ENOMEM;
1352 	}
1353 	rtcache_invariants(ro);
1354 	return 0;
1355 }
1356 
1357 static int
1358 rt_walktree_visitor(struct radix_node *rn, void *v)
1359 {
1360 	struct rtwalk *rw = (struct rtwalk *)v;
1361 
1362 	return (*rw->rw_f)((struct rtentry *)rn, rw->rw_v);
1363 }
1364 
1365 int
1366 rt_walktree(sa_family_t family, int (*f)(struct rtentry *, void *), void *v)
1367 {
1368 	struct radix_node_head *rnh = rt_tables[family];
1369 	struct rtwalk rw;
1370 
1371 	if (rnh == NULL)
1372 		return 0;
1373 
1374 	rw.rw_f = f;
1375 	rw.rw_v = v;
1376 
1377 	return rn_walktree(rnh, rt_walktree_visitor, &rw);
1378 }
1379