xref: /netbsd-src/sys/net/route.c (revision ce2c90c7c172d95d2402a5b3d96d8f8e6d138a21)
1 /*	$NetBSD: route.c,v 1.73 2006/10/12 01:32:30 christos 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 <sys/cdefs.h>
101 __KERNEL_RCSID(0, "$NetBSD: route.c,v 1.73 2006/10/12 01:32:30 christos Exp $");
102 
103 
104 #include <sys/param.h>
105 #include <sys/systm.h>
106 #include <sys/callout.h>
107 #include <sys/proc.h>
108 #include <sys/mbuf.h>
109 #include <sys/socket.h>
110 #include <sys/socketvar.h>
111 #include <sys/domain.h>
112 #include <sys/protosw.h>
113 #include <sys/kernel.h>
114 #include <sys/ioctl.h>
115 #include <sys/pool.h>
116 
117 #include <net/if.h>
118 #include <net/route.h>
119 #include <net/raw_cb.h>
120 
121 #include <netinet/in.h>
122 #include <netinet/in_var.h>
123 
124 
125 struct	route_cb route_cb;
126 struct	rtstat	rtstat;
127 struct	radix_node_head *rt_tables[AF_MAX+1];
128 
129 int	rttrash;		/* routes not in table but not freed */
130 struct	sockaddr wildcard;	/* zero valued cookie for wildcard searches */
131 
132 POOL_INIT(rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl", NULL);
133 POOL_INIT(rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl", NULL);
134 
135 struct callout rt_timer_ch; /* callout for rt_timer_timer() */
136 
137 static int rtdeletemsg(struct rtentry *);
138 static int rtflushclone1(struct radix_node *, void *);
139 static void rtflushclone(struct radix_node_head *, struct rtentry *);
140 
141 void
142 rtable_init(void **table)
143 {
144 	struct domain *dom;
145 	DOMAIN_FOREACH(dom)
146 		if (dom->dom_rtattach)
147 			dom->dom_rtattach(&table[dom->dom_family],
148 			    dom->dom_rtoffset);
149 }
150 
151 void
152 route_init(void)
153 {
154 
155 	rn_init();	/* initialize all zeroes, all ones, mask table */
156 	rtable_init((void **)rt_tables);
157 }
158 
159 /*
160  * Packet routing routines.
161  */
162 void
163 rtalloc(struct route *ro)
164 {
165 	if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP))
166 		return;				 /* XXX */
167 	ro->ro_rt = rtalloc1(&ro->ro_dst, 1);
168 }
169 
170 struct rtentry *
171 rtalloc1(const struct sockaddr *dst, int report)
172 {
173 	struct radix_node_head *rnh = rt_tables[dst->sa_family];
174 	struct rtentry *rt;
175 	struct radix_node *rn;
176 	struct rtentry *newrt = NULL;
177 	struct rt_addrinfo info;
178 	int  s = splsoftnet(), err = 0, msgtype = RTM_MISS;
179 
180 	if (rnh && (rn = rnh->rnh_matchaddr(dst, rnh)) &&
181 	    ((rn->rn_flags & RNF_ROOT) == 0)) {
182 		newrt = rt = (struct rtentry *)rn;
183 		if (report && (rt->rt_flags & RTF_CLONING)) {
184 			err = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
185 			    &newrt);
186 			if (err) {
187 				newrt = rt;
188 				rt->rt_refcnt++;
189 				goto miss;
190 			}
191 			KASSERT(newrt != NULL);
192 			if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) {
193 				msgtype = RTM_RESOLVE;
194 				goto miss;
195 			}
196 			/* Inform listeners of the new route */
197 			memset(&info, 0, sizeof(info));
198 			info.rti_info[RTAX_DST] = rt_key(rt);
199 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
200 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
201 			if (rt->rt_ifp != NULL) {
202 				info.rti_info[RTAX_IFP] =
203 				    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
204 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
205 			}
206 			rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0);
207 		} else
208 			rt->rt_refcnt++;
209 	} else {
210 		rtstat.rts_unreach++;
211 	miss:	if (report) {
212 			memset((caddr_t)&info, 0, sizeof(info));
213 			info.rti_info[RTAX_DST] = dst;
214 			rt_missmsg(msgtype, &info, 0, err);
215 		}
216 	}
217 	splx(s);
218 	return (newrt);
219 }
220 
221 void
222 rtfree(struct rtentry *rt)
223 {
224 	struct ifaddr *ifa;
225 
226 	if (rt == NULL)
227 		panic("rtfree");
228 	rt->rt_refcnt--;
229 	if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
230 		if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
231 			panic ("rtfree 2");
232 		rttrash--;
233 		if (rt->rt_refcnt < 0) {
234 			printf("rtfree: %p not freed (neg refs)\n", rt);
235 			return;
236 		}
237 		rt_timer_remove_all(rt, 0);
238 		ifa = rt->rt_ifa;
239 		IFAFREE(ifa);
240 		Free(rt_key(rt));
241 		pool_put(&rtentry_pool, rt);
242 	}
243 }
244 
245 void
246 ifafree(struct ifaddr *ifa)
247 {
248 
249 #ifdef DIAGNOSTIC
250 	if (ifa == NULL)
251 		panic("ifafree: null ifa");
252 	if (ifa->ifa_refcnt != 0)
253 		panic("ifafree: ifa_refcnt != 0 (%d)", ifa->ifa_refcnt);
254 #endif
255 #ifdef IFAREF_DEBUG
256 	printf("ifafree: freeing ifaddr %p\n", ifa);
257 #endif
258 	free(ifa, M_IFADDR);
259 }
260 
261 /*
262  * Force a routing table entry to the specified
263  * destination to go through the given gateway.
264  * Normally called as a result of a routing redirect
265  * message from the network layer.
266  *
267  * N.B.: must be called at splsoftnet
268  */
269 void
270 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
271 	const struct sockaddr *netmask, int flags, const struct sockaddr *src,
272 	struct rtentry **rtp)
273 {
274 	struct rtentry *rt;
275 	int error = 0;
276 	u_quad_t *stat = NULL;
277 	struct rt_addrinfo info;
278 	struct ifaddr *ifa;
279 
280 	/* verify the gateway is directly reachable */
281 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
282 		error = ENETUNREACH;
283 		goto out;
284 	}
285 	rt = rtalloc1(dst, 0);
286 	/*
287 	 * If the redirect isn't from our current router for this dst,
288 	 * it's either old or wrong.  If it redirects us to ourselves,
289 	 * we have a routing loop, perhaps as a result of an interface
290 	 * going down recently.
291 	 */
292 #define	equal(a1, a2) \
293 	((a1)->sa_len == (a2)->sa_len && \
294 	 memcmp((a1), (a2), (a1)->sa_len) == 0)
295 	if (!(flags & RTF_DONE) && rt &&
296 	     (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa))
297 		error = EINVAL;
298 	else if (ifa_ifwithaddr(gateway))
299 		error = EHOSTUNREACH;
300 	if (error)
301 		goto done;
302 	/*
303 	 * Create a new entry if we just got back a wildcard entry
304 	 * or the lookup failed.  This is necessary for hosts
305 	 * which use routing redirects generated by smart gateways
306 	 * to dynamically build the routing tables.
307 	 */
308 	if ((rt == NULL) || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
309 		goto create;
310 	/*
311 	 * Don't listen to the redirect if it's
312 	 * for a route to an interface.
313 	 */
314 	if (rt->rt_flags & RTF_GATEWAY) {
315 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
316 			/*
317 			 * Changing from route to net => route to host.
318 			 * Create new route, rather than smashing route to net.
319 			 */
320 		create:
321 			if (rt)
322 				rtfree(rt);
323 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
324 			info.rti_info[RTAX_DST] = dst;
325 			info.rti_info[RTAX_GATEWAY] = gateway;
326 			info.rti_info[RTAX_NETMASK] = netmask;
327 			info.rti_ifa = ifa;
328 			info.rti_flags = flags;
329 			rt = NULL;
330 			error = rtrequest1(RTM_ADD, &info, &rt);
331 			if (rt != NULL)
332 				flags = rt->rt_flags;
333 			stat = &rtstat.rts_dynamic;
334 		} else {
335 			/*
336 			 * Smash the current notion of the gateway to
337 			 * this destination.  Should check about netmask!!!
338 			 */
339 			rt->rt_flags |= RTF_MODIFIED;
340 			flags |= RTF_MODIFIED;
341 			stat = &rtstat.rts_newgateway;
342 			rt_setgate(rt, rt_key(rt), gateway);
343 		}
344 	} else
345 		error = EHOSTUNREACH;
346 done:
347 	if (rt) {
348 		if (rtp && !error)
349 			*rtp = rt;
350 		else
351 			rtfree(rt);
352 	}
353 out:
354 	if (error)
355 		rtstat.rts_badredirect++;
356 	else if (stat != NULL)
357 		(*stat)++;
358 	memset((caddr_t)&info, 0, sizeof(info));
359 	info.rti_info[RTAX_DST] = dst;
360 	info.rti_info[RTAX_GATEWAY] = gateway;
361 	info.rti_info[RTAX_NETMASK] = netmask;
362 	info.rti_info[RTAX_AUTHOR] = src;
363 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
364 }
365 
366 /*
367  * Delete a route and generate a message
368  */
369 static int
370 rtdeletemsg(struct rtentry *rt)
371 {
372 	int error;
373 	struct rt_addrinfo info;
374 
375 	/*
376 	 * Request the new route so that the entry is not actually
377 	 * deleted.  That will allow the information being reported to
378 	 * be accurate (and consistent with route_output()).
379 	 */
380 	memset((caddr_t)&info, 0, sizeof(info));
381 	info.rti_info[RTAX_DST] = rt_key(rt);
382 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
383 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
384 	info.rti_flags = rt->rt_flags;
385 	error = rtrequest1(RTM_DELETE, &info, &rt);
386 
387 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
388 
389 	/* Adjust the refcount */
390 	if (error == 0 && rt->rt_refcnt <= 0) {
391 		rt->rt_refcnt++;
392 		rtfree(rt);
393 	}
394 	return (error);
395 }
396 
397 static int
398 rtflushclone1(struct radix_node *rn, void *arg)
399 {
400 	struct rtentry *rt, *parent;
401 
402 	rt = (struct rtentry *)rn;
403 	parent = (struct rtentry *)arg;
404 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
405 		rtdeletemsg(rt);
406 	return 0;
407 }
408 
409 static void
410 rtflushclone(struct radix_node_head *rnh, struct rtentry *parent)
411 {
412 
413 #ifdef DIAGNOSTIC
414 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
415 		panic("rtflushclone: called with a non-cloning route");
416 	if (!rnh->rnh_walktree)
417 		panic("rtflushclone: no rnh_walktree");
418 #endif
419 	rnh->rnh_walktree(rnh, rtflushclone1, (void *)parent);
420 }
421 
422 /*
423  * Routing table ioctl interface.
424  */
425 int
426 rtioctl(u_long req __unused, caddr_t data __unused, struct lwp *l __unused)
427 {
428 	return (EOPNOTSUPP);
429 }
430 
431 struct ifaddr *
432 ifa_ifwithroute(int flags, const struct sockaddr *dst,
433 	const struct sockaddr *gateway)
434 {
435 	struct ifaddr *ifa;
436 	if ((flags & RTF_GATEWAY) == 0) {
437 		/*
438 		 * If we are adding a route to an interface,
439 		 * and the interface is a pt to pt link
440 		 * we should search for the destination
441 		 * as our clue to the interface.  Otherwise
442 		 * we can use the local address.
443 		 */
444 		ifa = NULL;
445 		if (flags & RTF_HOST)
446 			ifa = ifa_ifwithdstaddr(dst);
447 		if (ifa == NULL)
448 			ifa = ifa_ifwithaddr(gateway);
449 	} else {
450 		/*
451 		 * If we are adding a route to a remote net
452 		 * or host, the gateway may still be on the
453 		 * other end of a pt to pt link.
454 		 */
455 		ifa = ifa_ifwithdstaddr(gateway);
456 	}
457 	if (ifa == NULL)
458 		ifa = ifa_ifwithnet(gateway);
459 	if (ifa == NULL) {
460 		struct rtentry *rt = rtalloc1(dst, 0);
461 		if (rt == NULL)
462 			return NULL;
463 		rt->rt_refcnt--;
464 		if ((ifa = rt->rt_ifa) == NULL)
465 			return NULL;
466 	}
467 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
468 		struct ifaddr *oifa = ifa;
469 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
470 		if (ifa == 0)
471 			ifa = oifa;
472 	}
473 	return (ifa);
474 }
475 
476 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
477 
478 int
479 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
480 	const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
481 {
482 	struct rt_addrinfo info;
483 
484 	memset(&info, 0, sizeof(info));
485 	info.rti_flags = flags;
486 	info.rti_info[RTAX_DST] = dst;
487 	info.rti_info[RTAX_GATEWAY] = gateway;
488 	info.rti_info[RTAX_NETMASK] = netmask;
489 	return rtrequest1(req, &info, ret_nrt);
490 }
491 
492 int
493 rt_getifa(struct rt_addrinfo *info)
494 {
495 	struct ifaddr *ifa;
496 	int error = 0;
497 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
498 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
499 	const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
500 	const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
501 	int flags = info->rti_flags;
502 
503 	/*
504 	 * ifp may be specified by sockaddr_dl when protocol address
505 	 * is ambiguous
506 	 */
507 	if (info->rti_ifp == NULL && ifpaddr != NULL
508 	    && ifpaddr->sa_family == AF_LINK &&
509 	    (ifa = ifa_ifwithnet((const struct sockaddr *)ifpaddr)) != NULL)
510 		info->rti_ifp = ifa->ifa_ifp;
511 	if (info->rti_ifa == NULL && ifaaddr != NULL)
512 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
513 	if (info->rti_ifa == NULL) {
514 		const struct sockaddr *sa;
515 
516 		sa = ifaaddr != NULL ? ifaaddr :
517 		    (gateway != NULL ? gateway : dst);
518 		if (sa != NULL && info->rti_ifp != NULL)
519 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
520 		else if (dst != NULL && gateway != NULL)
521 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
522 		else if (sa != NULL)
523 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
524 	}
525 	if ((ifa = info->rti_ifa) != NULL) {
526 		if (info->rti_ifp == NULL)
527 			info->rti_ifp = ifa->ifa_ifp;
528 	} else
529 		error = ENETUNREACH;
530 	return (error);
531 }
532 
533 int
534 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
535 {
536 	int s = splsoftnet();
537 	int error = 0;
538 	struct rtentry *rt, *crt;
539 	struct radix_node *rn;
540 	struct radix_node_head *rnh;
541 	struct ifaddr *ifa;
542 	struct sockaddr *ndst;
543 	struct sockaddr_storage deldst;
544 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
545 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
546 	const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
547 	int flags = info->rti_flags;
548 #define senderr(x) { error = x ; goto bad; }
549 
550 	if ((rnh = rt_tables[dst->sa_family]) == NULL)
551 		senderr(ESRCH);
552 	if (flags & RTF_HOST)
553 		netmask = NULL;
554 	switch (req) {
555 	case RTM_DELETE:
556 		if (netmask) {
557 			rt_maskedcopy(dst, (struct sockaddr *)&deldst, netmask);
558 			dst = (struct sockaddr *)&deldst;
559 		}
560 		if ((rn = rnh->rnh_lookup(dst, netmask, rnh)) == NULL)
561 			senderr(ESRCH);
562 		rt = (struct rtentry *)rn;
563 		if ((rt->rt_flags & RTF_CLONING) != 0) {
564 			/* clean up any cloned children */
565 			rtflushclone(rnh, rt);
566 		}
567 		if ((rn = rnh->rnh_deladdr(dst, netmask, rnh)) == NULL)
568 			senderr(ESRCH);
569 		if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
570 			panic ("rtrequest delete");
571 		rt = (struct rtentry *)rn;
572 		if (rt->rt_gwroute) {
573 			RTFREE(rt->rt_gwroute);
574 			rt->rt_gwroute = NULL;
575 		}
576 		if (rt->rt_parent) {
577 			rt->rt_parent->rt_refcnt--;
578 			rt->rt_parent = NULL;
579 		}
580 		rt->rt_flags &= ~RTF_UP;
581 		if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
582 			ifa->ifa_rtrequest(RTM_DELETE, rt, info);
583 		rttrash++;
584 		if (ret_nrt)
585 			*ret_nrt = rt;
586 		else if (rt->rt_refcnt <= 0) {
587 			rt->rt_refcnt++;
588 			rtfree(rt);
589 		}
590 		break;
591 
592 	case RTM_RESOLVE:
593 		if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
594 			senderr(EINVAL);
595 		if ((rt->rt_flags & RTF_CLONING) == 0)
596 			senderr(EINVAL);
597 		ifa = rt->rt_ifa;
598 		flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC);
599 		flags |= RTF_CLONED;
600 		gateway = rt->rt_gateway;
601 		if ((netmask = rt->rt_genmask) == NULL)
602 			flags |= RTF_HOST;
603 		goto makeroute;
604 
605 	case RTM_ADD:
606 		if (info->rti_ifa == NULL && (error = rt_getifa(info)))
607 			senderr(error);
608 		ifa = info->rti_ifa;
609 	makeroute:
610 		/* Already at splsoftnet() so pool_get/pool_put are safe */
611 		rt = pool_get(&rtentry_pool, PR_NOWAIT);
612 		if (rt == NULL)
613 			senderr(ENOBUFS);
614 		Bzero(rt, sizeof(*rt));
615 		rt->rt_flags = RTF_UP | flags;
616 		LIST_INIT(&rt->rt_timer);
617 		if (rt_setgate(rt, dst, gateway)) {
618 			pool_put(&rtentry_pool, rt);
619 			senderr(ENOBUFS);
620 		}
621 		ndst = rt_key(rt);
622 		if (netmask) {
623 			rt_maskedcopy(dst, ndst, netmask);
624 		} else
625 			Bcopy(dst, ndst, dst->sa_len);
626 		IFAREF(ifa);
627 		rt->rt_ifa = ifa;
628 		rt->rt_ifp = ifa->ifa_ifp;
629 		if (req == RTM_RESOLVE) {
630 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
631 			rt->rt_parent = *ret_nrt;
632 			rt->rt_parent->rt_refcnt++;
633 		}
634 		rn = rnh->rnh_addaddr(ndst, netmask, rnh, rt->rt_nodes);
635 		if (rn == NULL && (crt = rtalloc1(ndst, 0)) != NULL) {
636 			/* overwrite cloned route */
637 			if ((crt->rt_flags & RTF_CLONED) != 0) {
638 				rtdeletemsg(crt);
639 				rn = rnh->rnh_addaddr(ndst,
640 				    netmask, rnh, rt->rt_nodes);
641 			}
642 			RTFREE(crt);
643 		}
644 		if (rn == NULL) {
645 			IFAFREE(ifa);
646 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
647 				rtfree(rt->rt_parent);
648 			if (rt->rt_gwroute)
649 				rtfree(rt->rt_gwroute);
650 			Free(rt_key(rt));
651 			pool_put(&rtentry_pool, rt);
652 			senderr(EEXIST);
653 		}
654 		if (ifa->ifa_rtrequest)
655 			ifa->ifa_rtrequest(req, rt, info);
656 		if (ret_nrt) {
657 			*ret_nrt = rt;
658 			rt->rt_refcnt++;
659 		}
660 		if ((rt->rt_flags & RTF_CLONING) != 0) {
661 			/* clean up any cloned children */
662 			rtflushclone(rnh, rt);
663 		}
664 		break;
665 	}
666 bad:
667 	splx(s);
668 	return (error);
669 }
670 
671 int
672 rt_setgate( struct rtentry *rt0, const struct sockaddr *dst,
673 	const struct sockaddr *gate)
674 {
675 	char *new, *old;
676 	u_int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len);
677 	struct rtentry *rt = rt0;
678 
679 	if (rt->rt_gateway == NULL || glen > ROUNDUP(rt->rt_gateway->sa_len)) {
680 		old = (caddr_t)rt_key(rt);
681 		R_Malloc(new, caddr_t, dlen + glen);
682 		if (new == NULL)
683 			return 1;
684 		Bzero(new, dlen + glen);
685 		rt->rt_nodes->rn_key = new;
686 	} else {
687 		new = __UNCONST(rt->rt_nodes->rn_key); /*XXXUNCONST*/
688 		old = NULL;
689 	}
690 	Bcopy(gate, (rt->rt_gateway = (struct sockaddr *)(new + dlen)), glen);
691 	if (old) {
692 		Bcopy(dst, new, dlen);
693 		Free(old);
694 	}
695 	if (rt->rt_gwroute) {
696 		RTFREE(rt->rt_gwroute);
697 		rt->rt_gwroute = NULL;
698 	}
699 	if (rt->rt_flags & RTF_GATEWAY) {
700 		rt->rt_gwroute = rtalloc1(gate, 1);
701 		/*
702 		 * If we switched gateways, grab the MTU from the new
703 		 * gateway route if the current MTU, if the current MTU is
704 		 * greater than the MTU of gateway.
705 		 * Note that, if the MTU of gateway is 0, we will reset the
706 		 * MTU of the route to run PMTUD again from scratch. XXX
707 		 */
708 		if (rt->rt_gwroute
709 		    && !(rt->rt_rmx.rmx_locks & RTV_MTU)
710 		    && rt->rt_rmx.rmx_mtu
711 		    && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
712 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
713 		}
714 	}
715 	return 0;
716 }
717 
718 void
719 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
720 	const struct sockaddr *netmask)
721 {
722 	const u_char *cp1 = (const u_char *)src;
723 	u_char *cp2 = (u_char *)dst;
724 	const u_char *cp3 = (const u_char *)netmask;
725 	u_char *cplim = cp2 + *cp3;
726 	u_char *cplim2 = cp2 + *cp1;
727 
728 	*cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
729 	cp3 += 2;
730 	if (cplim > cplim2)
731 		cplim = cplim2;
732 	while (cp2 < cplim)
733 		*cp2++ = *cp1++ & *cp3++;
734 	if (cp2 < cplim2)
735 		memset(cp2, 0, (unsigned)(cplim2 - cp2));
736 }
737 
738 /*
739  * Set up or tear down a routing table entry, normally
740  * for an interface.
741  */
742 int
743 rtinit(struct ifaddr *ifa, int cmd, int flags)
744 {
745 	struct rtentry *rt;
746 	struct sockaddr *dst, *odst;
747 	struct sockaddr_storage deldst;
748 	struct rtentry *nrt = NULL;
749 	int error;
750 	struct rt_addrinfo info;
751 
752 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
753 	if (cmd == RTM_DELETE) {
754 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
755 			/* Delete subnet route for this interface */
756 			odst = dst;
757 			dst = (struct sockaddr *)&deldst;
758 			rt_maskedcopy(odst, dst, ifa->ifa_netmask);
759 		}
760 		if ((rt = rtalloc1(dst, 0)) != NULL) {
761 			rt->rt_refcnt--;
762 			if (rt->rt_ifa != ifa)
763 				return (flags & RTF_HOST ? EHOSTUNREACH
764 							: ENETUNREACH);
765 		}
766 	}
767 	memset(&info, 0, sizeof(info));
768 	info.rti_ifa = ifa;
769 	info.rti_flags = flags | ifa->ifa_flags;
770 	info.rti_info[RTAX_DST] = dst;
771 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
772 	/*
773 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
774 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
775 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
776 	 * change it to meet bsdi4 behavior.
777 	 */
778 	info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
779 	error = rtrequest1(cmd, &info, &nrt);
780 	if (cmd == RTM_DELETE && error == 0 && (rt = nrt)) {
781 		rt_newaddrmsg(cmd, ifa, error, nrt);
782 		if (rt->rt_refcnt <= 0) {
783 			rt->rt_refcnt++;
784 			rtfree(rt);
785 		}
786 	}
787 	if (cmd == RTM_ADD && error == 0 && (rt = nrt)) {
788 		rt->rt_refcnt--;
789 		if (rt->rt_ifa != ifa) {
790 			printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
791 				rt->rt_ifa);
792 			if (rt->rt_ifa->ifa_rtrequest)
793 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, NULL);
794 			IFAFREE(rt->rt_ifa);
795 			rt->rt_ifa = ifa;
796 			rt->rt_ifp = ifa->ifa_ifp;
797 			IFAREF(ifa);
798 			if (ifa->ifa_rtrequest)
799 				ifa->ifa_rtrequest(RTM_ADD, rt, NULL);
800 		}
801 		rt_newaddrmsg(cmd, ifa, error, nrt);
802 	}
803 	return (error);
804 }
805 
806 /*
807  * Route timer routines.  These routes allow functions to be called
808  * for various routes at any time.  This is useful in supporting
809  * path MTU discovery and redirect route deletion.
810  *
811  * This is similar to some BSDI internal functions, but it provides
812  * for multiple queues for efficiency's sake...
813  */
814 
815 LIST_HEAD(, rttimer_queue) rttimer_queue_head;
816 static int rt_init_done = 0;
817 
818 #define RTTIMER_CALLOUT(r)	do {					\
819 		if (r->rtt_func != NULL) {				\
820 			(*r->rtt_func)(r->rtt_rt, r);			\
821 		} else {						\
822 			rtrequest((int) RTM_DELETE,			\
823 				  (struct sockaddr *)rt_key(r->rtt_rt),	\
824 				  0, 0, 0, 0);				\
825 		}							\
826 	} while (/*CONSTCOND*/0)
827 
828 /*
829  * Some subtle order problems with domain initialization mean that
830  * we cannot count on this being run from rt_init before various
831  * protocol initializations are done.  Therefore, we make sure
832  * that this is run when the first queue is added...
833  */
834 
835 void
836 rt_timer_init(void)
837 {
838 	assert(rt_init_done == 0);
839 
840 	LIST_INIT(&rttimer_queue_head);
841 	callout_init(&rt_timer_ch);
842 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
843 	rt_init_done = 1;
844 }
845 
846 struct rttimer_queue *
847 rt_timer_queue_create(u_int timeout)
848 {
849 	struct rttimer_queue *rtq;
850 
851 	if (rt_init_done == 0)
852 		rt_timer_init();
853 
854 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
855 	if (rtq == NULL)
856 		return (NULL);
857 	Bzero(rtq, sizeof *rtq);
858 
859 	rtq->rtq_timeout = timeout;
860 	rtq->rtq_count = 0;
861 	TAILQ_INIT(&rtq->rtq_head);
862 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
863 
864 	return (rtq);
865 }
866 
867 void
868 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
869 {
870 
871 	rtq->rtq_timeout = timeout;
872 }
873 
874 void
875 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy)
876 {
877 	struct rttimer *r;
878 
879 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
880 		LIST_REMOVE(r, rtt_link);
881 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
882 		if (destroy)
883 			RTTIMER_CALLOUT(r);
884 		/* we are already at splsoftnet */
885 		pool_put(&rttimer_pool, r);
886 		if (rtq->rtq_count > 0)
887 			rtq->rtq_count--;
888 		else
889 			printf("rt_timer_queue_remove_all: "
890 			    "rtq_count reached 0\n");
891 	}
892 }
893 
894 void
895 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
896 {
897 
898 	rt_timer_queue_remove_all(rtq, destroy);
899 
900 	LIST_REMOVE(rtq, rtq_link);
901 
902 	/*
903 	 * Caller is responsible for freeing the rttimer_queue structure.
904 	 */
905 }
906 
907 unsigned long
908 rt_timer_count(struct rttimer_queue *rtq)
909 {
910 	return rtq->rtq_count;
911 }
912 
913 void
914 rt_timer_remove_all(struct rtentry *rt, int destroy)
915 {
916 	struct rttimer *r;
917 
918 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
919 		LIST_REMOVE(r, rtt_link);
920 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
921 		if (destroy)
922 			RTTIMER_CALLOUT(r);
923 		if (r->rtt_queue->rtq_count > 0)
924 			r->rtt_queue->rtq_count--;
925 		else
926 			printf("rt_timer_remove_all: rtq_count reached 0\n");
927 		/* we are already at splsoftnet */
928 		pool_put(&rttimer_pool, r);
929 	}
930 }
931 
932 int
933 rt_timer_add(struct rtentry *rt,
934 	void (*func)(struct rtentry *, struct rttimer *),
935 	struct rttimer_queue *queue)
936 {
937 	struct rttimer *r;
938 	int s;
939 
940 	/*
941 	 * If there's already a timer with this action, destroy it before
942 	 * we add a new one.
943 	 */
944 	for (r = LIST_FIRST(&rt->rt_timer); r != NULL;
945 	     r = LIST_NEXT(r, rtt_link)) {
946 		if (r->rtt_func == func) {
947 			LIST_REMOVE(r, rtt_link);
948 			TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
949 			if (r->rtt_queue->rtq_count > 0)
950 				r->rtt_queue->rtq_count--;
951 			else
952 				printf("rt_timer_add: rtq_count reached 0\n");
953 			s = splsoftnet();
954 			pool_put(&rttimer_pool, r);
955 			splx(s);
956 			break;  /* only one per list, so we can quit... */
957 		}
958 	}
959 
960 	s = splsoftnet();
961 	r = pool_get(&rttimer_pool, PR_NOWAIT);
962 	splx(s);
963 	if (r == NULL)
964 		return (ENOBUFS);
965 	Bzero(r, sizeof(*r));
966 
967 	r->rtt_rt = rt;
968 	r->rtt_time = time_uptime;
969 	r->rtt_func = func;
970 	r->rtt_queue = queue;
971 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
972 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
973 	r->rtt_queue->rtq_count++;
974 
975 	return (0);
976 }
977 
978 /* ARGSUSED */
979 void
980 rt_timer_timer(void *arg __unused)
981 {
982 	struct rttimer_queue *rtq;
983 	struct rttimer *r;
984 	int s;
985 
986 	s = splsoftnet();
987 	for (rtq = LIST_FIRST(&rttimer_queue_head); rtq != NULL;
988 	     rtq = LIST_NEXT(rtq, rtq_link)) {
989 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
990 		    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
991 			LIST_REMOVE(r, rtt_link);
992 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
993 			RTTIMER_CALLOUT(r);
994 			pool_put(&rttimer_pool, r);
995 			if (rtq->rtq_count > 0)
996 				rtq->rtq_count--;
997 			else
998 				printf("rt_timer_timer: rtq_count reached 0\n");
999 		}
1000 	}
1001 	splx(s);
1002 
1003 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1004 }
1005