xref: /netbsd-src/sys/net/route.c (revision c0179c282a5968435315a82f4128c61372c68fc3)
1 /*	$NetBSD: route.c,v 1.76 2006/11/16 01:33:40 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 1.76 2006/11/16 01:33:40 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 != NULL) {
166 		if (ro->ro_rt->rt_ifp != NULL &&
167 		    (ro->ro_rt->rt_flags & RTF_UP) != 0)
168 			return;
169 		RTFREE(ro->ro_rt);
170 	}
171 	ro->ro_rt = rtalloc1(&ro->ro_dst, 1);
172 }
173 
174 struct rtentry *
175 rtalloc1(const struct sockaddr *dst, int report)
176 {
177 	struct radix_node_head *rnh = rt_tables[dst->sa_family];
178 	struct rtentry *rt;
179 	struct radix_node *rn;
180 	struct rtentry *newrt = NULL;
181 	struct rt_addrinfo info;
182 	int  s = splsoftnet(), err = 0, msgtype = RTM_MISS;
183 
184 	if (rnh && (rn = rnh->rnh_matchaddr(dst, rnh)) &&
185 	    ((rn->rn_flags & RNF_ROOT) == 0)) {
186 		newrt = rt = (struct rtentry *)rn;
187 		if (report && (rt->rt_flags & RTF_CLONING)) {
188 			err = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
189 			    &newrt);
190 			if (err) {
191 				newrt = rt;
192 				rt->rt_refcnt++;
193 				goto miss;
194 			}
195 			KASSERT(newrt != NULL);
196 			if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) {
197 				msgtype = RTM_RESOLVE;
198 				goto miss;
199 			}
200 			/* Inform listeners of the new route */
201 			memset(&info, 0, sizeof(info));
202 			info.rti_info[RTAX_DST] = rt_key(rt);
203 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
204 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
205 			if (rt->rt_ifp != NULL) {
206 				info.rti_info[RTAX_IFP] =
207 				    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
208 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
209 			}
210 			rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0);
211 		} else
212 			rt->rt_refcnt++;
213 	} else {
214 		rtstat.rts_unreach++;
215 	miss:	if (report) {
216 			memset((caddr_t)&info, 0, sizeof(info));
217 			info.rti_info[RTAX_DST] = dst;
218 			rt_missmsg(msgtype, &info, 0, err);
219 		}
220 	}
221 	splx(s);
222 	return (newrt);
223 }
224 
225 void
226 rtfree(struct rtentry *rt)
227 {
228 	struct ifaddr *ifa;
229 
230 	if (rt == NULL)
231 		panic("rtfree");
232 	rt->rt_refcnt--;
233 	if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
234 		if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
235 			panic ("rtfree 2");
236 		rttrash--;
237 		if (rt->rt_refcnt < 0) {
238 			printf("rtfree: %p not freed (neg refs)\n", rt);
239 			return;
240 		}
241 		rt_timer_remove_all(rt, 0);
242 		ifa = rt->rt_ifa;
243 		IFAFREE(ifa);
244 		Free(rt_key(rt));
245 		pool_put(&rtentry_pool, rt);
246 	}
247 }
248 
249 void
250 ifafree(struct ifaddr *ifa)
251 {
252 
253 #ifdef DIAGNOSTIC
254 	if (ifa == NULL)
255 		panic("ifafree: null ifa");
256 	if (ifa->ifa_refcnt != 0)
257 		panic("ifafree: ifa_refcnt != 0 (%d)", ifa->ifa_refcnt);
258 #endif
259 #ifdef IFAREF_DEBUG
260 	printf("ifafree: freeing ifaddr %p\n", ifa);
261 #endif
262 	free(ifa, M_IFADDR);
263 }
264 
265 /*
266  * Force a routing table entry to the specified
267  * destination to go through the given gateway.
268  * Normally called as a result of a routing redirect
269  * message from the network layer.
270  *
271  * N.B.: must be called at splsoftnet
272  */
273 void
274 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
275 	const struct sockaddr *netmask, int flags, const struct sockaddr *src,
276 	struct rtentry **rtp)
277 {
278 	struct rtentry *rt;
279 	int error = 0;
280 	u_quad_t *stat = NULL;
281 	struct rt_addrinfo info;
282 	struct ifaddr *ifa;
283 
284 	/* verify the gateway is directly reachable */
285 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
286 		error = ENETUNREACH;
287 		goto out;
288 	}
289 	rt = rtalloc1(dst, 0);
290 	/*
291 	 * If the redirect isn't from our current router for this dst,
292 	 * it's either old or wrong.  If it redirects us to ourselves,
293 	 * we have a routing loop, perhaps as a result of an interface
294 	 * going down recently.
295 	 */
296 #define	equal(a1, a2) \
297 	((a1)->sa_len == (a2)->sa_len && \
298 	 memcmp((a1), (a2), (a1)->sa_len) == 0)
299 	if (!(flags & RTF_DONE) && rt &&
300 	     (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa))
301 		error = EINVAL;
302 	else if (ifa_ifwithaddr(gateway))
303 		error = EHOSTUNREACH;
304 	if (error)
305 		goto done;
306 	/*
307 	 * Create a new entry if we just got back a wildcard entry
308 	 * or the lookup failed.  This is necessary for hosts
309 	 * which use routing redirects generated by smart gateways
310 	 * to dynamically build the routing tables.
311 	 */
312 	if ((rt == NULL) || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
313 		goto create;
314 	/*
315 	 * Don't listen to the redirect if it's
316 	 * for a route to an interface.
317 	 */
318 	if (rt->rt_flags & RTF_GATEWAY) {
319 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
320 			/*
321 			 * Changing from route to net => route to host.
322 			 * Create new route, rather than smashing route to net.
323 			 */
324 		create:
325 			if (rt)
326 				rtfree(rt);
327 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
328 			info.rti_info[RTAX_DST] = dst;
329 			info.rti_info[RTAX_GATEWAY] = gateway;
330 			info.rti_info[RTAX_NETMASK] = netmask;
331 			info.rti_ifa = ifa;
332 			info.rti_flags = flags;
333 			rt = NULL;
334 			error = rtrequest1(RTM_ADD, &info, &rt);
335 			if (rt != NULL)
336 				flags = rt->rt_flags;
337 			stat = &rtstat.rts_dynamic;
338 		} else {
339 			/*
340 			 * Smash the current notion of the gateway to
341 			 * this destination.  Should check about netmask!!!
342 			 */
343 			rt->rt_flags |= RTF_MODIFIED;
344 			flags |= RTF_MODIFIED;
345 			stat = &rtstat.rts_newgateway;
346 			rt_setgate(rt, rt_key(rt), gateway);
347 		}
348 	} else
349 		error = EHOSTUNREACH;
350 done:
351 	if (rt) {
352 		if (rtp && !error)
353 			*rtp = rt;
354 		else
355 			rtfree(rt);
356 	}
357 out:
358 	if (error)
359 		rtstat.rts_badredirect++;
360 	else if (stat != NULL)
361 		(*stat)++;
362 	memset((caddr_t)&info, 0, sizeof(info));
363 	info.rti_info[RTAX_DST] = dst;
364 	info.rti_info[RTAX_GATEWAY] = gateway;
365 	info.rti_info[RTAX_NETMASK] = netmask;
366 	info.rti_info[RTAX_AUTHOR] = src;
367 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
368 }
369 
370 /*
371  * Delete a route and generate a message
372  */
373 static int
374 rtdeletemsg(struct rtentry *rt)
375 {
376 	int error;
377 	struct rt_addrinfo info;
378 
379 	/*
380 	 * Request the new route so that the entry is not actually
381 	 * deleted.  That will allow the information being reported to
382 	 * be accurate (and consistent with route_output()).
383 	 */
384 	memset((caddr_t)&info, 0, sizeof(info));
385 	info.rti_info[RTAX_DST] = rt_key(rt);
386 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
387 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
388 	info.rti_flags = rt->rt_flags;
389 	error = rtrequest1(RTM_DELETE, &info, &rt);
390 
391 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
392 
393 	/* Adjust the refcount */
394 	if (error == 0 && rt->rt_refcnt <= 0) {
395 		rt->rt_refcnt++;
396 		rtfree(rt);
397 	}
398 	return (error);
399 }
400 
401 static int
402 rtflushclone1(struct radix_node *rn, void *arg)
403 {
404 	struct rtentry *rt, *parent;
405 
406 	rt = (struct rtentry *)rn;
407 	parent = (struct rtentry *)arg;
408 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
409 		rtdeletemsg(rt);
410 	return 0;
411 }
412 
413 static void
414 rtflushclone(struct radix_node_head *rnh, struct rtentry *parent)
415 {
416 
417 #ifdef DIAGNOSTIC
418 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
419 		panic("rtflushclone: called with a non-cloning route");
420 	if (!rnh->rnh_walktree)
421 		panic("rtflushclone: no rnh_walktree");
422 #endif
423 	rnh->rnh_walktree(rnh, rtflushclone1, (void *)parent);
424 }
425 
426 /*
427  * Routing table ioctl interface.
428  */
429 int
430 rtioctl(u_long req, caddr_t data, struct lwp *l)
431 {
432 	return (EOPNOTSUPP);
433 }
434 
435 struct ifaddr *
436 ifa_ifwithroute(int flags, const struct sockaddr *dst,
437 	const struct sockaddr *gateway)
438 {
439 	struct ifaddr *ifa;
440 	if ((flags & RTF_GATEWAY) == 0) {
441 		/*
442 		 * If we are adding a route to an interface,
443 		 * and the interface is a pt to pt link
444 		 * we should search for the destination
445 		 * as our clue to the interface.  Otherwise
446 		 * we can use the local address.
447 		 */
448 		ifa = NULL;
449 		if (flags & RTF_HOST)
450 			ifa = ifa_ifwithdstaddr(dst);
451 		if (ifa == NULL)
452 			ifa = ifa_ifwithaddr(gateway);
453 	} else {
454 		/*
455 		 * If we are adding a route to a remote net
456 		 * or host, the gateway may still be on the
457 		 * other end of a pt to pt link.
458 		 */
459 		ifa = ifa_ifwithdstaddr(gateway);
460 	}
461 	if (ifa == NULL)
462 		ifa = ifa_ifwithnet(gateway);
463 	if (ifa == NULL) {
464 		struct rtentry *rt = rtalloc1(dst, 0);
465 		if (rt == NULL)
466 			return NULL;
467 		rt->rt_refcnt--;
468 		if ((ifa = rt->rt_ifa) == NULL)
469 			return NULL;
470 	}
471 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
472 		struct ifaddr *oifa = ifa;
473 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
474 		if (ifa == 0)
475 			ifa = oifa;
476 	}
477 	return (ifa);
478 }
479 
480 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
481 
482 int
483 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
484 	const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
485 {
486 	struct rt_addrinfo info;
487 
488 	memset(&info, 0, sizeof(info));
489 	info.rti_flags = flags;
490 	info.rti_info[RTAX_DST] = dst;
491 	info.rti_info[RTAX_GATEWAY] = gateway;
492 	info.rti_info[RTAX_NETMASK] = netmask;
493 	return rtrequest1(req, &info, ret_nrt);
494 }
495 
496 int
497 rt_getifa(struct rt_addrinfo *info)
498 {
499 	struct ifaddr *ifa;
500 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
501 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
502 	const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
503 	const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
504 	int flags = info->rti_flags;
505 
506 	/*
507 	 * ifp may be specified by sockaddr_dl when protocol address
508 	 * is ambiguous
509 	 */
510 	if (info->rti_ifp == NULL && ifpaddr != NULL
511 	    && ifpaddr->sa_family == AF_LINK &&
512 	    (ifa = ifa_ifwithnet((const struct sockaddr *)ifpaddr)) != NULL)
513 		info->rti_ifp = ifa->ifa_ifp;
514 	if (info->rti_ifa == NULL && ifaaddr != NULL)
515 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
516 	if (info->rti_ifa == NULL) {
517 		const struct sockaddr *sa;
518 
519 		sa = ifaaddr != NULL ? ifaaddr :
520 		    (gateway != NULL ? gateway : dst);
521 		if (sa != NULL && info->rti_ifp != NULL)
522 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
523 		else if (dst != NULL && gateway != NULL)
524 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
525 		else if (sa != NULL)
526 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
527 	}
528 	if ((ifa = info->rti_ifa) == NULL)
529 		return ENETUNREACH;
530 	if (ifa->ifa_getifa != NULL)
531 		info->rti_ifa = ifa = (*ifa->ifa_getifa)(ifa, dst);
532 	if (info->rti_ifp == NULL)
533 		info->rti_ifp = ifa->ifa_ifp;
534 	return 0;
535 }
536 
537 int
538 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
539 {
540 	int s = splsoftnet();
541 	int error = 0;
542 	struct rtentry *rt, *crt;
543 	struct radix_node *rn;
544 	struct radix_node_head *rnh;
545 	struct ifaddr *ifa;
546 	struct sockaddr *ndst;
547 	struct sockaddr_storage deldst;
548 	const struct sockaddr *dst = info->rti_info[RTAX_DST];
549 	const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
550 	const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
551 	int flags = info->rti_flags;
552 #define senderr(x) { error = x ; goto bad; }
553 
554 	if ((rnh = rt_tables[dst->sa_family]) == NULL)
555 		senderr(ESRCH);
556 	if (flags & RTF_HOST)
557 		netmask = NULL;
558 	switch (req) {
559 	case RTM_DELETE:
560 		if (netmask) {
561 			rt_maskedcopy(dst, (struct sockaddr *)&deldst, netmask);
562 			dst = (struct sockaddr *)&deldst;
563 		}
564 		if ((rn = rnh->rnh_lookup(dst, netmask, rnh)) == NULL)
565 			senderr(ESRCH);
566 		rt = (struct rtentry *)rn;
567 		if ((rt->rt_flags & RTF_CLONING) != 0) {
568 			/* clean up any cloned children */
569 			rtflushclone(rnh, rt);
570 		}
571 		if ((rn = rnh->rnh_deladdr(dst, netmask, rnh)) == NULL)
572 			senderr(ESRCH);
573 		if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
574 			panic ("rtrequest delete");
575 		rt = (struct rtentry *)rn;
576 		if (rt->rt_gwroute) {
577 			RTFREE(rt->rt_gwroute);
578 			rt->rt_gwroute = NULL;
579 		}
580 		if (rt->rt_parent) {
581 			rt->rt_parent->rt_refcnt--;
582 			rt->rt_parent = NULL;
583 		}
584 		rt->rt_flags &= ~RTF_UP;
585 		if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
586 			ifa->ifa_rtrequest(RTM_DELETE, rt, info);
587 		rttrash++;
588 		if (ret_nrt)
589 			*ret_nrt = rt;
590 		else if (rt->rt_refcnt <= 0) {
591 			rt->rt_refcnt++;
592 			rtfree(rt);
593 		}
594 		break;
595 
596 	case RTM_RESOLVE:
597 		if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
598 			senderr(EINVAL);
599 		if ((rt->rt_flags & RTF_CLONING) == 0)
600 			senderr(EINVAL);
601 		ifa = rt->rt_ifa;
602 		flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC);
603 		flags |= RTF_CLONED;
604 		gateway = rt->rt_gateway;
605 		if ((netmask = rt->rt_genmask) == NULL)
606 			flags |= RTF_HOST;
607 		goto makeroute;
608 
609 	case RTM_ADD:
610 		if (info->rti_ifa == NULL && (error = rt_getifa(info)))
611 			senderr(error);
612 		ifa = info->rti_ifa;
613 	makeroute:
614 		/* Already at splsoftnet() so pool_get/pool_put are safe */
615 		rt = pool_get(&rtentry_pool, PR_NOWAIT);
616 		if (rt == NULL)
617 			senderr(ENOBUFS);
618 		Bzero(rt, sizeof(*rt));
619 		rt->rt_flags = RTF_UP | flags;
620 		LIST_INIT(&rt->rt_timer);
621 		if (rt_setgate(rt, dst, gateway)) {
622 			pool_put(&rtentry_pool, rt);
623 			senderr(ENOBUFS);
624 		}
625 		ndst = rt_key(rt);
626 		if (netmask) {
627 			rt_maskedcopy(dst, ndst, netmask);
628 		} else
629 			Bcopy(dst, ndst, dst->sa_len);
630 		rt_set_ifa(rt, ifa);
631 		rt->rt_ifp = ifa->ifa_ifp;
632 		if (req == RTM_RESOLVE) {
633 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
634 			rt->rt_parent = *ret_nrt;
635 			rt->rt_parent->rt_refcnt++;
636 		}
637 		rn = rnh->rnh_addaddr(ndst, netmask, rnh, rt->rt_nodes);
638 		if (rn == NULL && (crt = rtalloc1(ndst, 0)) != NULL) {
639 			/* overwrite cloned route */
640 			if ((crt->rt_flags & RTF_CLONED) != 0) {
641 				rtdeletemsg(crt);
642 				rn = rnh->rnh_addaddr(ndst,
643 				    netmask, rnh, rt->rt_nodes);
644 			}
645 			RTFREE(crt);
646 		}
647 		if (rn == NULL) {
648 			IFAFREE(ifa);
649 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
650 				rtfree(rt->rt_parent);
651 			if (rt->rt_gwroute)
652 				rtfree(rt->rt_gwroute);
653 			Free(rt_key(rt));
654 			pool_put(&rtentry_pool, rt);
655 			senderr(EEXIST);
656 		}
657 		if (ifa->ifa_rtrequest)
658 			ifa->ifa_rtrequest(req, rt, info);
659 		if (ret_nrt) {
660 			*ret_nrt = rt;
661 			rt->rt_refcnt++;
662 		}
663 		if ((rt->rt_flags & RTF_CLONING) != 0) {
664 			/* clean up any cloned children */
665 			rtflushclone(rnh, rt);
666 		}
667 		break;
668 	}
669 bad:
670 	splx(s);
671 	return (error);
672 }
673 
674 int
675 rt_setgate( struct rtentry *rt0, const struct sockaddr *dst,
676 	const struct sockaddr *gate)
677 {
678 	char *new, *old;
679 	u_int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len);
680 	struct rtentry *rt = rt0;
681 
682 	if (rt->rt_gateway == NULL || glen > ROUNDUP(rt->rt_gateway->sa_len)) {
683 		old = (caddr_t)rt_key(rt);
684 		R_Malloc(new, caddr_t, dlen + glen);
685 		if (new == NULL)
686 			return 1;
687 		Bzero(new, dlen + glen);
688 		rt->rt_nodes->rn_key = new;
689 	} else {
690 		new = __UNCONST(rt->rt_nodes->rn_key); /*XXXUNCONST*/
691 		old = NULL;
692 	}
693 	Bcopy(gate, (rt->rt_gateway = (struct sockaddr *)(new + dlen)), glen);
694 	if (old) {
695 		Bcopy(dst, new, dlen);
696 		Free(old);
697 	}
698 	if (rt->rt_gwroute) {
699 		RTFREE(rt->rt_gwroute);
700 		rt->rt_gwroute = NULL;
701 	}
702 	if (rt->rt_flags & RTF_GATEWAY) {
703 		rt->rt_gwroute = rtalloc1(gate, 1);
704 		/*
705 		 * If we switched gateways, grab the MTU from the new
706 		 * gateway route if the current MTU, if the current MTU is
707 		 * greater than the MTU of gateway.
708 		 * Note that, if the MTU of gateway is 0, we will reset the
709 		 * MTU of the route to run PMTUD again from scratch. XXX
710 		 */
711 		if (rt->rt_gwroute
712 		    && !(rt->rt_rmx.rmx_locks & RTV_MTU)
713 		    && rt->rt_rmx.rmx_mtu
714 		    && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
715 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
716 		}
717 	}
718 	return 0;
719 }
720 
721 void
722 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
723 	const struct sockaddr *netmask)
724 {
725 	const u_char *cp1 = (const u_char *)src;
726 	u_char *cp2 = (u_char *)dst;
727 	const u_char *cp3 = (const u_char *)netmask;
728 	u_char *cplim = cp2 + *cp3;
729 	u_char *cplim2 = cp2 + *cp1;
730 
731 	*cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
732 	cp3 += 2;
733 	if (cplim > cplim2)
734 		cplim = cplim2;
735 	while (cp2 < cplim)
736 		*cp2++ = *cp1++ & *cp3++;
737 	if (cp2 < cplim2)
738 		memset(cp2, 0, (unsigned)(cplim2 - cp2));
739 }
740 
741 /*
742  * Set up or tear down a routing table entry, normally
743  * for an interface.
744  */
745 int
746 rtinit(struct ifaddr *ifa, int cmd, int flags)
747 {
748 	struct rtentry *rt;
749 	struct sockaddr *dst, *odst;
750 	struct sockaddr_storage deldst;
751 	struct rtentry *nrt = NULL;
752 	int error;
753 	struct rt_addrinfo info;
754 
755 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
756 	if (cmd == RTM_DELETE) {
757 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
758 			/* Delete subnet route for this interface */
759 			odst = dst;
760 			dst = (struct sockaddr *)&deldst;
761 			rt_maskedcopy(odst, dst, ifa->ifa_netmask);
762 		}
763 		if ((rt = rtalloc1(dst, 0)) != NULL) {
764 			rt->rt_refcnt--;
765 			if (rt->rt_ifa != ifa)
766 				return (flags & RTF_HOST ? EHOSTUNREACH
767 							: ENETUNREACH);
768 		}
769 	}
770 	memset(&info, 0, sizeof(info));
771 	info.rti_ifa = ifa;
772 	info.rti_flags = flags | ifa->ifa_flags;
773 	info.rti_info[RTAX_DST] = dst;
774 	info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
775 	/*
776 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
777 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
778 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
779 	 * change it to meet bsdi4 behavior.
780 	 */
781 	info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
782 	error = rtrequest1(cmd, &info, &nrt);
783 	if (cmd == RTM_DELETE && error == 0 && (rt = nrt)) {
784 		rt_newaddrmsg(cmd, ifa, error, nrt);
785 		if (rt->rt_refcnt <= 0) {
786 			rt->rt_refcnt++;
787 			rtfree(rt);
788 		}
789 	}
790 	if (cmd == RTM_ADD && error == 0 && (rt = nrt)) {
791 		rt->rt_refcnt--;
792 		if (rt->rt_ifa != ifa) {
793 			printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
794 				rt->rt_ifa);
795 			if (rt->rt_ifa->ifa_rtrequest)
796 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, NULL);
797 			rt_replace_ifa(rt, ifa);
798 			rt->rt_ifp = ifa->ifa_ifp;
799 			if (ifa->ifa_rtrequest)
800 				ifa->ifa_rtrequest(RTM_ADD, rt, NULL);
801 		}
802 		rt_newaddrmsg(cmd, ifa, error, nrt);
803 	}
804 	return (error);
805 }
806 
807 /*
808  * Route timer routines.  These routes allow functions to be called
809  * for various routes at any time.  This is useful in supporting
810  * path MTU discovery and redirect route deletion.
811  *
812  * This is similar to some BSDI internal functions, but it provides
813  * for multiple queues for efficiency's sake...
814  */
815 
816 LIST_HEAD(, rttimer_queue) rttimer_queue_head;
817 static int rt_init_done = 0;
818 
819 #define RTTIMER_CALLOUT(r)	do {					\
820 		if (r->rtt_func != NULL) {				\
821 			(*r->rtt_func)(r->rtt_rt, r);			\
822 		} else {						\
823 			rtrequest((int) RTM_DELETE,			\
824 				  (struct sockaddr *)rt_key(r->rtt_rt),	\
825 				  0, 0, 0, 0);				\
826 		}							\
827 	} while (/*CONSTCOND*/0)
828 
829 /*
830  * Some subtle order problems with domain initialization mean that
831  * we cannot count on this being run from rt_init before various
832  * protocol initializations are done.  Therefore, we make sure
833  * that this is run when the first queue is added...
834  */
835 
836 void
837 rt_timer_init(void)
838 {
839 	assert(rt_init_done == 0);
840 
841 	LIST_INIT(&rttimer_queue_head);
842 	callout_init(&rt_timer_ch);
843 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
844 	rt_init_done = 1;
845 }
846 
847 struct rttimer_queue *
848 rt_timer_queue_create(u_int timeout)
849 {
850 	struct rttimer_queue *rtq;
851 
852 	if (rt_init_done == 0)
853 		rt_timer_init();
854 
855 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
856 	if (rtq == NULL)
857 		return (NULL);
858 	Bzero(rtq, sizeof *rtq);
859 
860 	rtq->rtq_timeout = timeout;
861 	rtq->rtq_count = 0;
862 	TAILQ_INIT(&rtq->rtq_head);
863 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
864 
865 	return (rtq);
866 }
867 
868 void
869 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
870 {
871 
872 	rtq->rtq_timeout = timeout;
873 }
874 
875 void
876 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy)
877 {
878 	struct rttimer *r;
879 
880 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
881 		LIST_REMOVE(r, rtt_link);
882 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
883 		if (destroy)
884 			RTTIMER_CALLOUT(r);
885 		/* we are already at splsoftnet */
886 		pool_put(&rttimer_pool, r);
887 		if (rtq->rtq_count > 0)
888 			rtq->rtq_count--;
889 		else
890 			printf("rt_timer_queue_remove_all: "
891 			    "rtq_count reached 0\n");
892 	}
893 }
894 
895 void
896 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
897 {
898 
899 	rt_timer_queue_remove_all(rtq, destroy);
900 
901 	LIST_REMOVE(rtq, rtq_link);
902 
903 	/*
904 	 * Caller is responsible for freeing the rttimer_queue structure.
905 	 */
906 }
907 
908 unsigned long
909 rt_timer_count(struct rttimer_queue *rtq)
910 {
911 	return rtq->rtq_count;
912 }
913 
914 void
915 rt_timer_remove_all(struct rtentry *rt, int destroy)
916 {
917 	struct rttimer *r;
918 
919 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
920 		LIST_REMOVE(r, rtt_link);
921 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
922 		if (destroy)
923 			RTTIMER_CALLOUT(r);
924 		if (r->rtt_queue->rtq_count > 0)
925 			r->rtt_queue->rtq_count--;
926 		else
927 			printf("rt_timer_remove_all: rtq_count reached 0\n");
928 		/* we are already at splsoftnet */
929 		pool_put(&rttimer_pool, r);
930 	}
931 }
932 
933 int
934 rt_timer_add(struct rtentry *rt,
935 	void (*func)(struct rtentry *, struct rttimer *),
936 	struct rttimer_queue *queue)
937 {
938 	struct rttimer *r;
939 	int s;
940 
941 	/*
942 	 * If there's already a timer with this action, destroy it before
943 	 * we add a new one.
944 	 */
945 	for (r = LIST_FIRST(&rt->rt_timer); r != NULL;
946 	     r = LIST_NEXT(r, rtt_link)) {
947 		if (r->rtt_func == func) {
948 			LIST_REMOVE(r, rtt_link);
949 			TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
950 			if (r->rtt_queue->rtq_count > 0)
951 				r->rtt_queue->rtq_count--;
952 			else
953 				printf("rt_timer_add: rtq_count reached 0\n");
954 			s = splsoftnet();
955 			pool_put(&rttimer_pool, r);
956 			splx(s);
957 			break;  /* only one per list, so we can quit... */
958 		}
959 	}
960 
961 	s = splsoftnet();
962 	r = pool_get(&rttimer_pool, PR_NOWAIT);
963 	splx(s);
964 	if (r == NULL)
965 		return (ENOBUFS);
966 	Bzero(r, sizeof(*r));
967 
968 	r->rtt_rt = rt;
969 	r->rtt_time = time_uptime;
970 	r->rtt_func = func;
971 	r->rtt_queue = queue;
972 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
973 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
974 	r->rtt_queue->rtq_count++;
975 
976 	return (0);
977 }
978 
979 /* ARGSUSED */
980 void
981 rt_timer_timer(void *arg)
982 {
983 	struct rttimer_queue *rtq;
984 	struct rttimer *r;
985 	int s;
986 
987 	s = splsoftnet();
988 	for (rtq = LIST_FIRST(&rttimer_queue_head); rtq != NULL;
989 	     rtq = LIST_NEXT(rtq, rtq_link)) {
990 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
991 		    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
992 			LIST_REMOVE(r, rtt_link);
993 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
994 			RTTIMER_CALLOUT(r);
995 			pool_put(&rttimer_pool, r);
996 			if (rtq->rtq_count > 0)
997 				rtq->rtq_count--;
998 			else
999 				printf("rt_timer_timer: rtq_count reached 0\n");
1000 		}
1001 	}
1002 	splx(s);
1003 
1004 	callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1005 }
1006