xref: /openbsd-src/sys/net/route.c (revision 2b0358df1d88d06ef4139321dd05bd5e05d91eaf)
1 /*	$OpenBSD: route.c,v 1.105 2009/03/15 19:40:41 miod Exp $	*/
2 /*	$NetBSD: route.c,v 1.14 1996/02/13 22:00:46 christos Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1980, 1986, 1991, 1993
35  *	The Regents of the University of California.  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 University 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 REGENTS 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 REGENTS 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  *	@(#)route.c	8.2 (Berkeley) 11/15/93
62  */
63 
64 /*
65  *	@(#)COPYRIGHT	1.1 (NRL) 17 January 1995
66  *
67  * NRL grants permission for redistribution and use in source and binary
68  * forms, with or without modification, of the software and documentation
69  * created at NRL provided that the following conditions are met:
70  *
71  * 1. Redistributions of source code must retain the above copyright
72  *    notice, this list of conditions and the following disclaimer.
73  * 2. Redistributions in binary form must reproduce the above copyright
74  *    notice, this list of conditions and the following disclaimer in the
75  *    documentation and/or other materials provided with the distribution.
76  * 3. All advertising materials mentioning features or use of this software
77  *    must display the following acknowledgements:
78  * 	This product includes software developed by the University of
79  * 	California, Berkeley and its contributors.
80  * 	This product includes software developed at the Information
81  * 	Technology Division, US Naval Research Laboratory.
82  * 4. Neither the name of the NRL nor the names of its contributors
83  *    may be used to endorse or promote products derived from this software
84  *    without specific prior written permission.
85  *
86  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
87  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
88  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
89  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
90  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
91  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
92  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
93  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
94  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
95  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
96  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
97  *
98  * The views and conclusions contained in the software and documentation
99  * are those of the authors and should not be interpreted as representing
100  * official policies, either expressed or implied, of the US Naval
101  * Research Laboratory (NRL).
102  */
103 
104 #include <sys/param.h>
105 #include <sys/systm.h>
106 #include <sys/proc.h>
107 #include <sys/mbuf.h>
108 #include <sys/socket.h>
109 #include <sys/socketvar.h>
110 #include <sys/domain.h>
111 #include <sys/protosw.h>
112 #include <sys/ioctl.h>
113 #include <sys/kernel.h>
114 #include <sys/queue.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 #ifdef MPLS
125 #include <netmpls/mpls.h>
126 #endif
127 
128 #ifdef IPSEC
129 #include <netinet/ip_ipsp.h>
130 #include <net/if_enc.h>
131 
132 struct ifaddr	*encap_findgwifa(struct sockaddr *);
133 #endif
134 
135 #define	SA(p) ((struct sockaddr *)(p))
136 
137 struct	route_cb	   route_cb;
138 struct	rtstat		   rtstat;
139 struct	radix_node_head	***rt_tables;
140 u_int8_t		   af2rtafidx[AF_MAX+1];
141 u_int8_t		   rtafidx_max;
142 u_int			   rtbl_id_max = 0;
143 
144 int			rttrash;	/* routes not in table but not freed */
145 
146 struct pool		rtentry_pool;	/* pool for rtentry structures */
147 struct pool		rttimer_pool;	/* pool for rttimer structures */
148 
149 int	rtable_init(struct radix_node_head ***);
150 int	okaytoclone(u_int, int);
151 int	rtflushclone1(struct radix_node *, void *);
152 void	rtflushclone(struct radix_node_head *, struct rtentry *);
153 int	rt_if_remove_rtdelete(struct radix_node *, void *);
154 #ifndef SMALL_KERNEL
155 int	rt_if_linkstate_change(struct radix_node *, void *);
156 #endif
157 
158 #define	LABELID_MAX	50000
159 
160 struct rt_label {
161 	TAILQ_ENTRY(rt_label)	rtl_entry;
162 	char			rtl_name[RTLABEL_LEN];
163 	u_int16_t		rtl_id;
164 	int			rtl_ref;
165 };
166 
167 TAILQ_HEAD(rt_labels, rt_label)	rt_labels = TAILQ_HEAD_INITIALIZER(rt_labels);
168 
169 #ifdef IPSEC
170 struct ifaddr *
171 encap_findgwifa(struct sockaddr *gw)
172 {
173 	return (TAILQ_FIRST(&encif[0].sc_if.if_addrlist));
174 }
175 #endif
176 
177 int
178 rtable_init(struct radix_node_head ***table)
179 {
180 	void		**p;
181 	struct domain	 *dom;
182 
183 	if ((p = malloc(sizeof(void *) * (rtafidx_max + 1), M_RTABLE,
184 	    M_NOWAIT|M_ZERO)) == NULL)
185 		return (-1);
186 
187 	/* 2nd pass: attach */
188 	for (dom = domains; dom != NULL; dom = dom->dom_next)
189 		if (dom->dom_rtattach)
190 			dom->dom_rtattach(&p[af2rtafidx[dom->dom_family]],
191 			    dom->dom_rtoffset);
192 
193 	*table = (struct radix_node_head **)p;
194 	return (0);
195 }
196 
197 void
198 route_init()
199 {
200 	struct domain	 *dom;
201 
202 	pool_init(&rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl",
203 	    NULL);
204 	rn_init();	/* initialize all zeroes, all ones, mask table */
205 
206 	bzero(af2rtafidx, sizeof(af2rtafidx));
207 	rtafidx_max = 1;	/* must have NULL at index 0, so start at 1 */
208 
209 	/* find out how many tables to allocate */
210 	for (dom = domains; dom != NULL; dom = dom->dom_next)
211 		if (dom->dom_rtattach)
212 			af2rtafidx[dom->dom_family] = rtafidx_max++;
213 
214 	if (rtable_add(0) == -1)
215 		panic("route_init rtable_add");
216 }
217 
218 int
219 rtable_add(u_int id)	/* must be called at splsoftnet */
220 {
221 	void	*p;
222 
223 	if (id > RT_TABLEID_MAX)
224 		return (-1);
225 
226 	if (id == 0 || id > rtbl_id_max) {
227 		size_t	newlen = sizeof(void *) * (id+1);
228 
229 		if ((p = malloc(newlen, M_RTABLE, M_NOWAIT|M_ZERO)) == NULL)
230 			return (-1);
231 		if (id > 0) {
232 			bcopy(rt_tables, p, sizeof(void *) * (rtbl_id_max+1));
233 			free(rt_tables, M_RTABLE);
234 		}
235 		rt_tables = p;
236 		rtbl_id_max = id;
237 	}
238 
239 	if (rt_tables[id] != NULL)	/* already exists */
240 		return (-1);
241 
242 	return (rtable_init(&rt_tables[id]));
243 }
244 
245 int
246 rtable_exists(u_int id)	/* verify table with that ID exists */
247 {
248 	if (id > RT_TABLEID_MAX)
249 		return (0);
250 
251 	if (id > rtbl_id_max)
252 		return (0);
253 
254 	if (rt_tables[id] == NULL)	/* should not happen */
255 		return (0);
256 
257 	return (1);
258 }
259 
260 #include "pf.h"
261 #if NPF > 0
262 void
263 rtalloc_noclone(struct route *ro, int howstrict)
264 {
265 	if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP))
266 		return;		/* XXX */
267 	ro->ro_rt = rtalloc2(&ro->ro_dst, 1, howstrict);
268 }
269 
270 int
271 okaytoclone(u_int flags, int howstrict)
272 {
273 	if (howstrict == ALL_CLONING)
274 		return (1);
275 	if (howstrict == ONNET_CLONING && !(flags & RTF_GATEWAY))
276 		return (1);
277 	return (0);
278 }
279 
280 struct rtentry *
281 rtalloc2(struct sockaddr *dst, int report, int howstrict)
282 {
283 	struct radix_node_head	*rnh;
284 	struct rtentry		*rt;
285 	struct radix_node	*rn;
286 	struct rtentry		*newrt = 0;
287 	struct rt_addrinfo	 info;
288 	int			 s = splnet(), err = 0, msgtype = RTM_MISS;
289 
290 	bzero(&info, sizeof(info));
291 	info.rti_info[RTAX_DST] = dst;
292 
293 	rnh = rt_gettable(dst->sa_family, 0);
294 	if (rnh && (rn = rnh->rnh_matchaddr((caddr_t)dst, rnh)) &&
295 	    ((rn->rn_flags & RNF_ROOT) == 0)) {
296 		newrt = rt = (struct rtentry *)rn;
297 		if (report && (rt->rt_flags & RTF_CLONING) &&
298 		    okaytoclone(rt->rt_flags, howstrict)) {
299 			err = rtrequest1(RTM_RESOLVE, &info, RTP_DEFAULT,
300 			    &newrt, 0);
301 			if (err) {
302 				newrt = rt;
303 				rt->rt_refcnt++;
304 				goto miss;
305 			}
306 			if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) {
307 				msgtype = RTM_RESOLVE;
308 				goto miss;
309 			}
310 		} else
311 			rt->rt_refcnt++;
312 	} else {
313 		rtstat.rts_unreach++;
314 miss:
315 		if (report) {
316 			rt_missmsg(msgtype, &info, 0, NULL, err, 0);
317 		}
318 	}
319 	splx(s);
320 	return (newrt);
321 }
322 #endif /* NPF > 0 */
323 
324 /*
325  * Packet routing routines.
326  */
327 void
328 rtalloc(struct route *ro)
329 {
330 	if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP))
331 		return;				 /* XXX */
332 	ro->ro_rt = rtalloc1(&ro->ro_dst, 1, 0);
333 }
334 
335 struct rtentry *
336 rtalloc1(struct sockaddr *dst, int report, u_int tableid)
337 {
338 	struct radix_node_head	*rnh;
339 	struct rtentry		*rt;
340 	struct radix_node	*rn;
341 	struct rtentry		*newrt = 0;
342 	struct rt_addrinfo	 info;
343 	int			 s = splsoftnet(), err = 0, msgtype = RTM_MISS;
344 
345 	bzero(&info, sizeof(info));
346 	info.rti_info[RTAX_DST] = dst;
347 
348 	rnh = rt_gettable(dst->sa_family, tableid);
349 	if (rnh && (rn = rnh->rnh_matchaddr((caddr_t)dst, rnh)) &&
350 	    ((rn->rn_flags & RNF_ROOT) == 0)) {
351 		newrt = rt = (struct rtentry *)rn;
352 		if (report && (rt->rt_flags & RTF_CLONING)) {
353 			err = rtrequest1(RTM_RESOLVE, &info, RTP_DEFAULT,
354 			    &newrt, tableid);
355 			if (err) {
356 				newrt = rt;
357 				rt->rt_refcnt++;
358 				goto miss;
359 			}
360 			if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) {
361 				msgtype = RTM_RESOLVE;
362 				goto miss;
363 			}
364 			/* Inform listeners of the new route */
365 			bzero(&info, sizeof(info));
366 			info.rti_info[RTAX_DST] = rt_key(rt);
367 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
368 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
369 			if (rt->rt_ifp != NULL) {
370 				info.rti_info[RTAX_IFP] =
371 				    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
372 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
373 			}
374 			rt_missmsg(RTM_ADD, &info, rt->rt_flags,
375 			    rt->rt_ifp, 0, tableid);
376 		} else
377 			rt->rt_refcnt++;
378 	} else {
379 		if (dst->sa_family != PF_KEY)
380 			rtstat.rts_unreach++;
381 	/*
382 	 * IP encapsulation does lots of lookups where we don't need nor want
383 	 * the RTM_MISSes that would be generated.  It causes RTM_MISS storms
384 	 * sent upward breaking user-level routing queries.
385 	 */
386 miss:
387 		if (report && dst->sa_family != PF_KEY) {
388 			bzero((caddr_t)&info, sizeof(info));
389 			info.rti_info[RTAX_DST] = dst;
390 			rt_missmsg(msgtype, &info, 0, NULL, err, tableid);
391 		}
392 	}
393 	splx(s);
394 	return (newrt);
395 }
396 
397 void
398 rtfree(struct rtentry *rt)
399 {
400 	struct ifaddr	*ifa;
401 
402 	if (rt == NULL)
403 		panic("rtfree");
404 
405 	rt->rt_refcnt--;
406 
407 	if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
408 		if (rt->rt_refcnt == 0 && (rt->rt_nodes->rn_flags & RNF_ACTIVE))
409 			return; /* route still active but currently down */
410 		if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
411 			panic("rtfree 2");
412 		rttrash--;
413 		if (rt->rt_refcnt < 0) {
414 			printf("rtfree: %p not freed (neg refs)\n", rt);
415 			return;
416 		}
417 		rt_timer_remove_all(rt);
418 		ifa = rt->rt_ifa;
419 		if (ifa)
420 			IFAFREE(ifa);
421 		rtlabel_unref(rt->rt_labelid);
422 #ifdef MPLS
423 		if (rt->rt_flags & RTF_MPLS)
424 			free(rt->rt_llinfo, M_TEMP);
425 #endif
426 		Free(rt_key(rt));
427 		pool_put(&rtentry_pool, rt);
428 	}
429 }
430 
431 void
432 ifafree(struct ifaddr *ifa)
433 {
434 	if (ifa == NULL)
435 		panic("ifafree");
436 	if (ifa->ifa_refcnt == 0)
437 		free(ifa, M_IFADDR);
438 	else
439 		ifa->ifa_refcnt--;
440 }
441 
442 /*
443  * Force a routing table entry to the specified
444  * destination to go through the given gateway.
445  * Normally called as a result of a routing redirect
446  * message from the network layer.
447  *
448  * N.B.: must be called at splsoftnet
449  */
450 void
451 rtredirect(struct sockaddr *dst, struct sockaddr *gateway,
452     struct sockaddr *netmask, int flags, struct sockaddr *src,
453     struct rtentry **rtp)
454 {
455 	struct rtentry		*rt;
456 	int			 error = 0;
457 	u_int32_t		*stat = NULL;
458 	struct rt_addrinfo	 info;
459 	struct ifaddr		*ifa;
460 	struct ifnet		*ifp = NULL;
461 
462 	splsoftassert(IPL_SOFTNET);
463 
464 	/* verify the gateway is directly reachable */
465 	if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
466 		error = ENETUNREACH;
467 		goto out;
468 	}
469 	ifp = ifa->ifa_ifp;
470 	rt = rtalloc1(dst, 0, 0);
471 	/*
472 	 * If the redirect isn't from our current router for this dst,
473 	 * it's either old or wrong.  If it redirects us to ourselves,
474 	 * we have a routing loop, perhaps as a result of an interface
475 	 * going down recently.
476 	 */
477 #define	equal(a1, a2) \
478 	((a1)->sa_len == (a2)->sa_len && \
479 	 bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0)
480 	if (!(flags & RTF_DONE) && rt &&
481 	     (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa))
482 		error = EINVAL;
483 	else if (ifa_ifwithaddr(gateway) != NULL)
484 		error = EHOSTUNREACH;
485 	if (error)
486 		goto done;
487 	/*
488 	 * Create a new entry if we just got back a wildcard entry
489 	 * or the lookup failed.  This is necessary for hosts
490 	 * which use routing redirects generated by smart gateways
491 	 * to dynamically build the routing tables.
492 	 */
493 	if ((rt == NULL) || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
494 		goto create;
495 	/*
496 	 * Don't listen to the redirect if it's
497 	 * for a route to an interface.
498 	 */
499 	if (rt->rt_flags & RTF_GATEWAY) {
500 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
501 			/*
502 			 * Changing from route to net => route to host.
503 			 * Create new route, rather than smashing route to net.
504 			 */
505 create:
506 			if (rt)
507 				rtfree(rt);
508 			flags |= RTF_GATEWAY | RTF_DYNAMIC;
509 			bzero(&info, sizeof(info));
510 			info.rti_info[RTAX_DST] = dst;
511 			info.rti_info[RTAX_GATEWAY] = gateway;
512 			info.rti_info[RTAX_NETMASK] = netmask;
513 			info.rti_ifa = ifa;
514 			info.rti_flags = flags;
515 			rt = NULL;
516 			error = rtrequest1(RTM_ADD, &info, RTP_DEFAULT, &rt, 0);
517 			if (rt != NULL)
518 				flags = rt->rt_flags;
519 			stat = &rtstat.rts_dynamic;
520 		} else {
521 			/*
522 			 * Smash the current notion of the gateway to
523 			 * this destination.  Should check about netmask!!!
524 			 */
525 			rt->rt_flags |= RTF_MODIFIED;
526 			flags |= RTF_MODIFIED;
527 			stat = &rtstat.rts_newgateway;
528 			rt_setgate(rt, rt_key(rt), gateway, 0);
529 		}
530 	} else
531 		error = EHOSTUNREACH;
532 done:
533 	if (rt) {
534 		if (rtp && !error)
535 			*rtp = rt;
536 		else
537 			rtfree(rt);
538 	}
539 out:
540 	if (error)
541 		rtstat.rts_badredirect++;
542 	else if (stat != NULL)
543 		(*stat)++;
544 	bzero((caddr_t)&info, sizeof(info));
545 	info.rti_info[RTAX_DST] = dst;
546 	info.rti_info[RTAX_GATEWAY] = gateway;
547 	info.rti_info[RTAX_NETMASK] = netmask;
548 	info.rti_info[RTAX_AUTHOR] = src;
549 	rt_missmsg(RTM_REDIRECT, &info, flags, ifp, error, 0);
550 }
551 
552 /*
553  * Delete a route and generate a message
554  */
555 int
556 rtdeletemsg(struct rtentry *rt, u_int tableid)
557 {
558 	int			error;
559 	struct rt_addrinfo	info;
560 	struct ifnet		*ifp;
561 
562 	/*
563 	 * Request the new route so that the entry is not actually
564 	 * deleted.  That will allow the information being reported to
565 	 * be accurate (and consistent with route_output()).
566 	 */
567 	bzero((caddr_t)&info, sizeof(info));
568 	info.rti_info[RTAX_DST] = rt_key(rt);
569 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
570 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
571 	info.rti_flags = rt->rt_flags;
572 	ifp = rt->rt_ifp;
573 	error = rtrequest1(RTM_DELETE, &info, rt->rt_priority, &rt, tableid);
574 
575 	rt_missmsg(RTM_DELETE, &info, info.rti_flags, ifp, error, tableid);
576 
577 	/* Adjust the refcount */
578 	if (error == 0 && rt->rt_refcnt <= 0) {
579 		rt->rt_refcnt++;
580 		rtfree(rt);
581 	}
582 	return (error);
583 }
584 
585 int
586 rtflushclone1(struct radix_node *rn, void *arg)
587 {
588 	struct rtentry	*rt, *parent;
589 
590 	rt = (struct rtentry *)rn;
591 	parent = (struct rtentry *)arg;
592 	if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent)
593 		rtdeletemsg(rt, 0);
594 	return 0;
595 }
596 
597 void
598 rtflushclone(struct radix_node_head *rnh, struct rtentry *parent)
599 {
600 
601 #ifdef DIAGNOSTIC
602 	if (!parent || (parent->rt_flags & RTF_CLONING) == 0)
603 		panic("rtflushclone: called with a non-cloning route");
604 	if (!rnh->rnh_walktree)
605 		panic("rtflushclone: no rnh_walktree");
606 #endif
607 	rnh->rnh_walktree(rnh, rtflushclone1, (void *)parent);
608 }
609 
610 int
611 rtioctl(u_long req, caddr_t data, struct proc *p)
612 {
613 	return (EOPNOTSUPP);
614 }
615 
616 struct ifaddr *
617 ifa_ifwithroute(int flags, struct sockaddr *dst, struct sockaddr *gateway)
618 {
619 	struct ifaddr	*ifa;
620 
621 #ifdef IPSEC
622 	/*
623 	 * If the destination is a PF_KEY address, we'll look
624 	 * for the existence of a encap interface number or address
625 	 * in the options list of the gateway. By default, we'll return
626 	 * enc0.
627 	 */
628 	if (dst && (dst->sa_family == PF_KEY))
629 		return (encap_findgwifa(gateway));
630 #endif
631 
632 	if ((flags & RTF_GATEWAY) == 0) {
633 		/*
634 		 * If we are adding a route to an interface,
635 		 * and the interface is a pt to pt link
636 		 * we should search for the destination
637 		 * as our clue to the interface.  Otherwise
638 		 * we can use the local address.
639 		 */
640 		ifa = NULL;
641 		if (flags & RTF_HOST)
642 			ifa = ifa_ifwithdstaddr(dst);
643 		if (ifa == NULL)
644 			ifa = ifa_ifwithaddr(gateway);
645 	} else {
646 		/*
647 		 * If we are adding a route to a remote net
648 		 * or host, the gateway may still be on the
649 		 * other end of a pt to pt link.
650 		 */
651 		ifa = ifa_ifwithdstaddr(gateway);
652 	}
653 	if (ifa == NULL)
654 		ifa = ifa_ifwithnet(gateway);
655 	if (ifa == NULL) {
656 		struct rtentry	*rt = rtalloc1(gateway, 0, 0);
657 		if (rt == NULL)
658 			return (NULL);
659 		rt->rt_refcnt--;
660 		/* The gateway must be local if the same address family. */
661 		if ((rt->rt_flags & RTF_GATEWAY) &&
662 		    rt_key(rt)->sa_family == dst->sa_family)
663 			return (0);
664 		if ((ifa = rt->rt_ifa) == NULL)
665 			return (NULL);
666 	}
667 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
668 		struct ifaddr	*oifa = ifa;
669 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
670 		if (ifa == NULL)
671 			ifa = oifa;
672 	}
673 	return (ifa);
674 }
675 
676 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
677 
678 int
679 rt_getifa(struct rt_addrinfo *info)
680 {
681 	struct ifaddr	*ifa;
682 	int		 error = 0;
683 
684 	/*
685 	 * ifp may be specified by sockaddr_dl when protocol address
686 	 * is ambiguous
687 	 */
688 	if (info->rti_ifp == NULL && info->rti_info[RTAX_IFP] != NULL
689 	    && info->rti_info[RTAX_IFP]->sa_family == AF_LINK &&
690 	    (ifa = ifa_ifwithnet((struct sockaddr *)info->rti_info[RTAX_IFP]))
691 	    != NULL)
692 		info->rti_ifp = ifa->ifa_ifp;
693 
694 	if (info->rti_ifa == NULL && info->rti_info[RTAX_IFA] != NULL)
695 		info->rti_ifa = ifa_ifwithaddr(info->rti_info[RTAX_IFA]);
696 
697 	if (info->rti_ifa == NULL) {
698 		struct sockaddr	*sa;
699 
700 		if ((sa = info->rti_info[RTAX_IFA]) == NULL)
701 			if ((sa = info->rti_info[RTAX_GATEWAY]) == NULL)
702 				sa = info->rti_info[RTAX_DST];
703 
704 		if (sa != NULL && info->rti_ifp != NULL)
705 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
706 		else if (info->rti_info[RTAX_DST] != NULL &&
707 		    info->rti_info[RTAX_GATEWAY] != NULL)
708 			info->rti_ifa = ifa_ifwithroute(info->rti_flags,
709 			    info->rti_info[RTAX_DST],
710 			    info->rti_info[RTAX_GATEWAY]);
711 		else if (sa != NULL)
712 			info->rti_ifa = ifa_ifwithroute(info->rti_flags,
713 			    sa, sa);
714 	}
715 	if ((ifa = info->rti_ifa) != NULL) {
716 		if (info->rti_ifp == NULL)
717 			info->rti_ifp = ifa->ifa_ifp;
718 	} else
719 		error = ENETUNREACH;
720 	return (error);
721 }
722 
723 int
724 rtrequest1(int req, struct rt_addrinfo *info, u_int8_t prio,
725     struct rtentry **ret_nrt, u_int tableid)
726 {
727 	int			 s = splsoftnet(); int error = 0;
728 	struct rtentry		*rt, *crt;
729 	struct radix_node	*rn;
730 	struct radix_node_head	*rnh;
731 	struct ifaddr		*ifa;
732 	struct sockaddr		*ndst;
733 	struct sockaddr_rtlabel	*sa_rl;
734 #ifdef MPLS
735 	struct sockaddr_mpls	*sa_mpls;
736 #endif
737 #define senderr(x) { error = x ; goto bad; }
738 
739 	if ((rnh = rt_gettable(info->rti_info[RTAX_DST]->sa_family, tableid)) ==
740 	    NULL)
741 		senderr(EAFNOSUPPORT);
742 	if (info->rti_flags & RTF_HOST)
743 		info->rti_info[RTAX_NETMASK] = NULL;
744 	switch (req) {
745 	case RTM_DELETE:
746 		if ((rn = rnh->rnh_lookup(info->rti_info[RTAX_DST],
747 		    info->rti_info[RTAX_NETMASK], rnh)) == NULL)
748 			senderr(ESRCH);
749 		rt = (struct rtentry *)rn;
750 #ifndef SMALL_KERNEL
751 		/*
752 		 * if we got multipath routes, we require users to specify
753 		 * a matching RTAX_GATEWAY.
754 		 */
755 		if (rn_mpath_capable(rnh)) {
756 			rt = rt_mpath_matchgate(rt,
757 			    info->rti_info[RTAX_GATEWAY], prio);
758 			rn = (struct radix_node *)rt;
759 			if (!rt)
760 				senderr(ESRCH);
761 		}
762 #endif
763 		if ((rn = rnh->rnh_deladdr(info->rti_info[RTAX_DST],
764 		    info->rti_info[RTAX_NETMASK], rnh, rn)) == NULL)
765 			senderr(ESRCH);
766 		rt = (struct rtentry *)rn;
767 
768 		/* clean up any cloned children */
769 		if ((rt->rt_flags & RTF_CLONING) != 0)
770 			rtflushclone(rnh, rt);
771 
772 		if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
773 			panic ("rtrequest delete");
774 
775 		if (rt->rt_gwroute) {
776 			rt = rt->rt_gwroute; RTFREE(rt);
777 			(rt = (struct rtentry *)rn)->rt_gwroute = NULL;
778 		}
779 
780 		if (rt->rt_parent) {
781 			rt->rt_parent->rt_refcnt--;
782 			rt->rt_parent = NULL;
783 		}
784 
785 #ifndef SMALL_KERNEL
786 		if (rn_mpath_capable(rnh)) {
787 			if ((rn = rnh->rnh_lookup(info->rti_info[RTAX_DST],
788 			    info->rti_info[RTAX_NETMASK], rnh)) != NULL &&
789 			    rn_mpath_next(rn, 0) == NULL)
790 				((struct rtentry *)rn)->rt_flags &= ~RTF_MPATH;
791 		}
792 #endif
793 
794 		rt->rt_flags &= ~RTF_UP;
795 		if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
796 			ifa->ifa_rtrequest(RTM_DELETE, rt, info);
797 		rttrash++;
798 
799 		if (ret_nrt)
800 			*ret_nrt = rt;
801 		else if (rt->rt_refcnt <= 0) {
802 			rt->rt_refcnt++;
803 			rtfree(rt);
804 		}
805 		break;
806 
807 	case RTM_RESOLVE:
808 		if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
809 			senderr(EINVAL);
810 		if ((rt->rt_flags & RTF_CLONING) == 0)
811 			senderr(EINVAL);
812 		ifa = rt->rt_ifa;
813 		info->rti_flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC);
814 		info->rti_flags |= RTF_CLONED;
815 		info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
816 		if ((info->rti_info[RTAX_NETMASK] = rt->rt_genmask) == NULL)
817 			info->rti_flags |= RTF_HOST;
818 		goto makeroute;
819 
820 	case RTM_ADD:
821 		if (info->rti_ifa == 0 && (error = rt_getifa(info)))
822 			senderr(error);
823 		ifa = info->rti_ifa;
824 makeroute:
825 		rt = pool_get(&rtentry_pool, PR_NOWAIT);
826 		if (rt == NULL)
827 			senderr(ENOBUFS);
828 		Bzero(rt, sizeof(*rt));
829 
830 		rt->rt_flags = info->rti_flags;
831 
832 		if (prio == 0)
833 			prio = ifa->ifa_ifp->if_priority + RTP_STATIC;
834 		rt->rt_priority = prio;	/* init routing priority */
835 #if 0
836 		if ((LINK_STATE_IS_UP(ifa->ifa_ifp->if_link_state) ||
837 		    ifa->ifa_ifp->if_link_state == LINK_STATE_UNKNOWN) &&
838 		    ifa->ifa_ifp->if_flags & IFF_UP)
839 			rt->rt_flags |= RTF_UP;
840 		else {
841 			rt->rt_flags &= ~RTF_UP;
842 			rt->rt_priority |= RTP_DOWN;
843 		}
844 #endif
845 		LIST_INIT(&rt->rt_timer);
846 		if (rt_setgate(rt, info->rti_info[RTAX_DST],
847 		    info->rti_info[RTAX_GATEWAY], tableid)) {
848 			pool_put(&rtentry_pool, rt);
849 			senderr(ENOBUFS);
850 		}
851 		ndst = rt_key(rt);
852 		if (info->rti_info[RTAX_NETMASK] != NULL) {
853 			rt_maskedcopy(info->rti_info[RTAX_DST], ndst,
854 			    info->rti_info[RTAX_NETMASK]);
855 		} else
856 			Bcopy(info->rti_info[RTAX_DST], ndst,
857 			    info->rti_info[RTAX_DST]->sa_len);
858 #ifndef SMALL_KERNEL
859 		/* do not permit exactly the same dst/mask/gw pair */
860 		if (rn_mpath_capable(rnh) &&
861 		    rt_mpath_conflict(rnh, rt, info->rti_info[RTAX_NETMASK],
862 		    info->rti_flags & RTF_MPATH)) {
863 			if (rt->rt_gwroute)
864 				rtfree(rt->rt_gwroute);
865 			Free(rt_key(rt));
866 			pool_put(&rtentry_pool, rt);
867 			senderr(EEXIST);
868 		}
869 #endif
870 
871 		if (info->rti_info[RTAX_LABEL] != NULL) {
872 			sa_rl = (struct sockaddr_rtlabel *)
873 			    info->rti_info[RTAX_LABEL];
874 			rt->rt_labelid = rtlabel_name2id(sa_rl->sr_label);
875 		}
876 
877 #ifdef MPLS
878 		/* We have to allocate additional space for MPLS infos */
879 		if (info->rti_info[RTAX_SRC] != NULL ||
880 		    info->rti_info[RTAX_DST]->sa_family == AF_MPLS) {
881 			struct rt_mpls *rt_mpls;
882 
883 			sa_mpls = (struct sockaddr_mpls *)
884 			    info->rti_info[RTAX_SRC];
885 
886 			rt->rt_llinfo = (caddr_t)malloc(sizeof(struct rt_mpls),
887 			    M_TEMP, M_NOWAIT|M_ZERO);
888 
889 			if (rt->rt_llinfo == NULL) {
890 				if (rt->rt_gwroute)
891 					rtfree(rt->rt_gwroute);
892 				Free(rt_key(rt));
893 				pool_put(&rtentry_pool, rt);
894 				senderr(ENOMEM);
895 			}
896 
897 			rt_mpls = (struct rt_mpls *)rt->rt_llinfo;
898 
899 			if (sa_mpls != NULL)
900 				rt_mpls->mpls_label = sa_mpls->smpls_label;
901 
902 			rt_mpls->mpls_operation = info->rti_mpls;
903 
904 			/* XXX: set experimental bits */
905 
906 			rt->rt_flags |= RTF_MPLS;
907 		}
908 #endif
909 
910 		ifa->ifa_refcnt++;
911 		rt->rt_ifa = ifa;
912 		rt->rt_ifp = ifa->ifa_ifp;
913 		if (req == RTM_RESOLVE) {
914 			/*
915 			 * Copy both metrics and a back pointer to the cloned
916 			 * route's parent.
917 			 */
918 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
919 			rt->rt_priority = (*ret_nrt)->rt_priority;
920 			rt->rt_parent = *ret_nrt;	 /* Back ptr. to parent. */
921 			rt->rt_parent->rt_refcnt++;
922 		}
923 		rn = rnh->rnh_addaddr((caddr_t)ndst,
924 		    (caddr_t)info->rti_info[RTAX_NETMASK], rnh, rt->rt_nodes,
925 		    rt->rt_priority);
926 		if (rn == NULL && (crt = rtalloc1(ndst, 0, tableid)) != NULL) {
927 			/* overwrite cloned route */
928 			if ((crt->rt_flags & RTF_CLONED) != 0) {
929 				rtdeletemsg(crt, tableid);
930 				rn = rnh->rnh_addaddr((caddr_t)ndst,
931 				    (caddr_t)info->rti_info[RTAX_NETMASK],
932 				    rnh, rt->rt_nodes, rt->rt_priority);
933 			}
934 			RTFREE(crt);
935 		}
936 		if (rn == 0) {
937 			IFAFREE(ifa);
938 			if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent)
939 				rtfree(rt->rt_parent);
940 			if (rt->rt_gwroute)
941 				rtfree(rt->rt_gwroute);
942 			Free(rt_key(rt));
943 			pool_put(&rtentry_pool, rt);
944 			senderr(EEXIST);
945 		}
946 
947 #ifndef SMALL_KERNEL
948 		if (rn_mpath_capable(rnh) &&
949 		    (rn = rnh->rnh_lookup(info->rti_info[RTAX_DST],
950 		    info->rti_info[RTAX_NETMASK], rnh)) != NULL &&
951 		    (rn = rn_mpath_prio(rn, prio)) != NULL) {
952 			if (rn_mpath_next(rn, 0) == NULL)
953 				((struct rtentry *)rn)->rt_flags &= ~RTF_MPATH;
954 			else
955 				((struct rtentry *)rn)->rt_flags |= RTF_MPATH;
956 		}
957 #endif
958 
959 		if (ifa->ifa_rtrequest)
960 			ifa->ifa_rtrequest(req, rt, info);
961 		if (ret_nrt) {
962 			*ret_nrt = rt;
963 			rt->rt_refcnt++;
964 		}
965 		if ((rt->rt_flags & RTF_CLONING) != 0) {
966 			/* clean up any cloned children */
967 			rtflushclone(rnh, rt);
968 		}
969 
970 		if_group_routechange(info->rti_info[RTAX_DST],
971 			info->rti_info[RTAX_NETMASK]);
972 		break;
973 	}
974 bad:
975 	splx(s);
976 	return (error);
977 }
978 
979 int
980 rt_setgate(struct rtentry *rt0, struct sockaddr *dst, struct sockaddr *gate,
981     u_int tableid)
982 {
983 	caddr_t	new, old;
984 	int	dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len);
985 	struct rtentry	*rt = rt0;
986 
987 	if (rt->rt_gateway == NULL || glen > ROUNDUP(rt->rt_gateway->sa_len)) {
988 		old = (caddr_t)rt_key(rt);
989 		R_Malloc(new, caddr_t, dlen + glen);
990 		if (new == NULL)
991 			return 1;
992 		rt->rt_nodes->rn_key = new;
993 	} else {
994 		new = rt->rt_nodes->rn_key;
995 		old = NULL;
996 	}
997 	Bcopy(gate, (rt->rt_gateway = (struct sockaddr *)(new + dlen)), glen);
998 	if (old) {
999 		Bcopy(dst, new, dlen);
1000 		Free(old);
1001 	}
1002 	if (rt->rt_gwroute != NULL) {
1003 		rt = rt->rt_gwroute;
1004 		RTFREE(rt);
1005 		rt = rt0;
1006 		rt->rt_gwroute = NULL;
1007 	}
1008 	if (rt->rt_flags & RTF_GATEWAY) {
1009 		rt->rt_gwroute = rtalloc1(gate, 1, tableid);
1010 		/*
1011 		 * If we switched gateways, grab the MTU from the new
1012 		 * gateway route if the current MTU is 0 or greater
1013 		 * than the MTU of gateway.
1014 		 * Note that, if the MTU of gateway is 0, we will reset the
1015 		 * MTU of the route to run PMTUD again from scratch. XXX
1016 		 */
1017 		if (rt->rt_gwroute && !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1018 		    rt->rt_rmx.rmx_mtu &&
1019 		    rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) {
1020 			rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu;
1021 		}
1022 	}
1023 	return (0);
1024 }
1025 
1026 void
1027 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst,
1028     struct sockaddr *netmask)
1029 {
1030 	u_char	*cp1 = (u_char *)src;
1031 	u_char	*cp2 = (u_char *)dst;
1032 	u_char	*cp3 = (u_char *)netmask;
1033 	u_char	*cplim = cp2 + *cp3;
1034 	u_char	*cplim2 = cp2 + *cp1;
1035 
1036 	*cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1037 	cp3 += 2;
1038 	if (cplim > cplim2)
1039 		cplim = cplim2;
1040 	while (cp2 < cplim)
1041 		*cp2++ = *cp1++ & *cp3++;
1042 	if (cp2 < cplim2)
1043 		bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1044 }
1045 
1046 /*
1047  * Set up a routing table entry, normally
1048  * for an interface.
1049  */
1050 int
1051 rtinit(struct ifaddr *ifa, int cmd, int flags)
1052 {
1053 	struct rtentry		*rt;
1054 	struct sockaddr		*dst, *deldst;
1055 	struct mbuf		*m = NULL;
1056 	struct rtentry		*nrt = NULL;
1057 	int			 error;
1058 	struct rt_addrinfo	 info;
1059 	struct sockaddr_rtlabel	 sa_rl;
1060 	const char		*label;
1061 
1062 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
1063 	if (cmd == RTM_DELETE) {
1064 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
1065 			m = m_get(M_DONTWAIT, MT_SONAME);
1066 			if (m == NULL)
1067 				return (ENOBUFS);
1068 			deldst = mtod(m, struct sockaddr *);
1069 			rt_maskedcopy(dst, deldst, ifa->ifa_netmask);
1070 			dst = deldst;
1071 		}
1072 		if ((rt = rtalloc1(dst, 0, 0)) != NULL) {
1073 			rt->rt_refcnt--;
1074 			if (rt->rt_ifa != ifa) {
1075 				if (m != NULL)
1076 					(void) m_free(m);
1077 				return (flags & RTF_HOST ? EHOSTUNREACH
1078 							: ENETUNREACH);
1079 			}
1080 		}
1081 	}
1082 	bzero(&info, sizeof(info));
1083 	info.rti_ifa = ifa;
1084 	info.rti_flags = flags | ifa->ifa_flags;
1085 	info.rti_info[RTAX_DST] = dst;
1086 	if (cmd == RTM_ADD)
1087 		info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
1088 	if (ifa->ifa_ifp->if_rtlabelid &&
1089 	    (label = rtlabel_id2name(ifa->ifa_ifp->if_rtlabelid)) != NULL) {
1090 		bzero(&sa_rl, sizeof(sa_rl));
1091 		sa_rl.sr_len = sizeof(sa_rl);
1092 		sa_rl.sr_family = AF_UNSPEC;
1093 		strlcpy(sa_rl.sr_label, label, sizeof(sa_rl.sr_label));
1094 		info.rti_info[RTAX_LABEL] = (struct sockaddr *)&sa_rl;
1095 	}
1096 
1097 	/*
1098 	 * XXX here, it seems that we are assuming that ifa_netmask is NULL
1099 	 * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
1100 	 * variable) when RTF_HOST is 1.  still not sure if i can safely
1101 	 * change it to meet bsdi4 behavior.
1102 	 */
1103 	info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1104 	error = rtrequest1(cmd, &info, RTP_CONNECTED, &nrt, 0);
1105 	if (cmd == RTM_DELETE && error == 0 && (rt = nrt) != NULL) {
1106 		rt_newaddrmsg(cmd, ifa, error, nrt);
1107 		if (rt->rt_refcnt <= 0) {
1108 			rt->rt_refcnt++;
1109 			rtfree(rt);
1110 		}
1111 	}
1112 	if (cmd == RTM_ADD && error == 0 && (rt = nrt) != NULL) {
1113 		rt->rt_refcnt--;
1114 		if (rt->rt_ifa != ifa) {
1115 			printf("rtinit: wrong ifa (%p) was (%p)\n",
1116 			    ifa, rt->rt_ifa);
1117 			if (rt->rt_ifa->ifa_rtrequest)
1118 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, NULL);
1119 			IFAFREE(rt->rt_ifa);
1120 			rt->rt_ifa = ifa;
1121 			rt->rt_ifp = ifa->ifa_ifp;
1122 			ifa->ifa_refcnt++;
1123 			if (ifa->ifa_rtrequest)
1124 				ifa->ifa_rtrequest(RTM_ADD, rt, NULL);
1125 		}
1126 		rt_newaddrmsg(cmd, ifa, error, nrt);
1127 	}
1128 	return (error);
1129 }
1130 
1131 /*
1132  * Route timer routines.  These routes allow functions to be called
1133  * for various routes at any time.  This is useful in supporting
1134  * path MTU discovery and redirect route deletion.
1135  *
1136  * This is similar to some BSDI internal functions, but it provides
1137  * for multiple queues for efficiency's sake...
1138  */
1139 
1140 LIST_HEAD(, rttimer_queue)	rttimer_queue_head;
1141 static int			rt_init_done = 0;
1142 
1143 #define RTTIMER_CALLOUT(r)	{				\
1144 	if (r->rtt_func != NULL) {				\
1145 		(*r->rtt_func)(r->rtt_rt, r);			\
1146 	} else {						\
1147 		struct rt_addrinfo info;			\
1148 		bzero(&info, sizeof(info));			\
1149 		info.rti_info[RTAX_DST] = rt_key(r->rtt_rt);	\
1150 		rtrequest1(RTM_DELETE, &info,			\
1151 		    r->rtt_rt->rt_priority, NULL, 0 /* XXX */);	\
1152 	}							\
1153 }
1154 
1155 /*
1156  * Some subtle order problems with domain initialization mean that
1157  * we cannot count on this being run from rt_init before various
1158  * protocol initializations are done.  Therefore, we make sure
1159  * that this is run when the first queue is added...
1160  */
1161 
1162 void
1163 rt_timer_init()
1164 {
1165 	static struct timeout	rt_timer_timeout;
1166 
1167 	if (rt_init_done)
1168 		panic("rt_timer_init: already initialized");
1169 
1170 	pool_init(&rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl",
1171 	    NULL);
1172 
1173 	LIST_INIT(&rttimer_queue_head);
1174 	timeout_set(&rt_timer_timeout, rt_timer_timer, &rt_timer_timeout);
1175 	timeout_add_sec(&rt_timer_timeout, 1);
1176 	rt_init_done = 1;
1177 }
1178 
1179 struct rttimer_queue *
1180 rt_timer_queue_create(u_int timeout)
1181 {
1182 	struct rttimer_queue	*rtq;
1183 
1184 	if (rt_init_done == 0)
1185 		rt_timer_init();
1186 
1187 	R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
1188 	if (rtq == NULL)
1189 		return (NULL);
1190 	Bzero(rtq, sizeof *rtq);
1191 
1192 	rtq->rtq_timeout = timeout;
1193 	rtq->rtq_count = 0;
1194 	TAILQ_INIT(&rtq->rtq_head);
1195 	LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
1196 
1197 	return (rtq);
1198 }
1199 
1200 void
1201 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
1202 {
1203 	rtq->rtq_timeout = timeout;
1204 }
1205 
1206 void
1207 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy)
1208 {
1209 	struct rttimer	*r;
1210 
1211 	while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
1212 		LIST_REMOVE(r, rtt_link);
1213 		TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1214 		if (destroy)
1215 			RTTIMER_CALLOUT(r);
1216 		pool_put(&rttimer_pool, r);
1217 		if (rtq->rtq_count > 0)
1218 			rtq->rtq_count--;
1219 		else
1220 			printf("rt_timer_queue_destroy: rtq_count reached 0\n");
1221 	}
1222 
1223 	LIST_REMOVE(rtq, rtq_link);
1224 
1225 	/*
1226 	 * Caller is responsible for freeing the rttimer_queue structure.
1227 	 */
1228 }
1229 
1230 unsigned long
1231 rt_timer_count(struct rttimer_queue *rtq)
1232 {
1233 	return (rtq->rtq_count);
1234 }
1235 
1236 void
1237 rt_timer_remove_all(struct rtentry *rt)
1238 {
1239 	struct rttimer	*r;
1240 
1241 	while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
1242 		LIST_REMOVE(r, rtt_link);
1243 		TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
1244 		if (r->rtt_queue->rtq_count > 0)
1245 			r->rtt_queue->rtq_count--;
1246 		else
1247 			printf("rt_timer_remove_all: rtq_count reached 0\n");
1248 		pool_put(&rttimer_pool, r);
1249 	}
1250 }
1251 
1252 int
1253 rt_timer_add(struct rtentry *rt, void (*func)(struct rtentry *,
1254     struct rttimer *), struct rttimer_queue *queue)
1255 {
1256 	struct rttimer	*r;
1257 	long		 current_time;
1258 
1259 	current_time = time_uptime;
1260 	rt->rt_rmx.rmx_expire = time_second + queue->rtq_timeout;
1261 
1262 	/*
1263 	 * If there's already a timer with this action, destroy it before
1264 	 * we add a new one.
1265 	 */
1266 	for (r = LIST_FIRST(&rt->rt_timer); r != NULL;
1267 	     r = LIST_NEXT(r, rtt_link)) {
1268 		if (r->rtt_func == func) {
1269 			LIST_REMOVE(r, rtt_link);
1270 			TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
1271 			if (r->rtt_queue->rtq_count > 0)
1272 				r->rtt_queue->rtq_count--;
1273 			else
1274 				printf("rt_timer_add: rtq_count reached 0\n");
1275 			pool_put(&rttimer_pool, r);
1276 			break;  /* only one per list, so we can quit... */
1277 		}
1278 	}
1279 
1280 	r = pool_get(&rttimer_pool, PR_NOWAIT);
1281 	if (r == NULL)
1282 		return (ENOBUFS);
1283 	Bzero(r, sizeof(*r));
1284 
1285 	r->rtt_rt = rt;
1286 	r->rtt_time = current_time;
1287 	r->rtt_func = func;
1288 	r->rtt_queue = queue;
1289 	LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
1290 	TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
1291 	r->rtt_queue->rtq_count++;
1292 
1293 	return (0);
1294 }
1295 
1296 struct radix_node_head *
1297 rt_gettable(sa_family_t af, u_int id)
1298 {
1299 	return (rt_tables[id] ? rt_tables[id][af2rtafidx[af]] : NULL);
1300 }
1301 
1302 struct radix_node *
1303 rt_lookup(struct sockaddr *dst, struct sockaddr *mask, int tableid)
1304 {
1305 	struct radix_node_head	*rnh;
1306 
1307 	if ((rnh = rt_gettable(dst->sa_family, tableid)) == NULL)
1308 		return (NULL);
1309 
1310 	return (rnh->rnh_lookup(dst, mask, rnh));
1311 }
1312 
1313 /* ARGSUSED */
1314 void
1315 rt_timer_timer(void *arg)
1316 {
1317 	struct timeout		*to = (struct timeout *)arg;
1318 	struct rttimer_queue	*rtq;
1319 	struct rttimer		*r;
1320 	long			 current_time;
1321 	int			 s;
1322 
1323 	current_time = time_uptime;
1324 
1325 	s = splsoftnet();
1326 	for (rtq = LIST_FIRST(&rttimer_queue_head); rtq != NULL;
1327 	     rtq = LIST_NEXT(rtq, rtq_link)) {
1328 		while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
1329 		    (r->rtt_time + rtq->rtq_timeout) < current_time) {
1330 			LIST_REMOVE(r, rtt_link);
1331 			TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1332 			RTTIMER_CALLOUT(r);
1333 			pool_put(&rttimer_pool, r);
1334 			if (rtq->rtq_count > 0)
1335 				rtq->rtq_count--;
1336 			else
1337 				printf("rt_timer_timer: rtq_count reached 0\n");
1338 		}
1339 	}
1340 	splx(s);
1341 
1342 	timeout_add_sec(to, 1);
1343 }
1344 
1345 u_int16_t
1346 rtlabel_name2id(char *name)
1347 {
1348 	struct rt_label		*label, *p = NULL;
1349 	u_int16_t		 new_id = 1;
1350 
1351 	if (!name[0])
1352 		return (0);
1353 
1354 	TAILQ_FOREACH(label, &rt_labels, rtl_entry)
1355 		if (strcmp(name, label->rtl_name) == 0) {
1356 			label->rtl_ref++;
1357 			return (label->rtl_id);
1358 		}
1359 
1360 	/*
1361 	 * to avoid fragmentation, we do a linear search from the beginning
1362 	 * and take the first free slot we find. if there is none or the list
1363 	 * is empty, append a new entry at the end.
1364 	 */
1365 
1366 	if (!TAILQ_EMPTY(&rt_labels))
1367 		for (p = TAILQ_FIRST(&rt_labels); p != NULL &&
1368 		    p->rtl_id == new_id; p = TAILQ_NEXT(p, rtl_entry))
1369 			new_id = p->rtl_id + 1;
1370 
1371 	if (new_id > LABELID_MAX)
1372 		return (0);
1373 
1374 	label = malloc(sizeof(*label), M_TEMP, M_NOWAIT|M_ZERO);
1375 	if (label == NULL)
1376 		return (0);
1377 	strlcpy(label->rtl_name, name, sizeof(label->rtl_name));
1378 	label->rtl_id = new_id;
1379 	label->rtl_ref++;
1380 
1381 	if (p != NULL)	/* insert new entry before p */
1382 		TAILQ_INSERT_BEFORE(p, label, rtl_entry);
1383 	else		/* either list empty or no free slot in between */
1384 		TAILQ_INSERT_TAIL(&rt_labels, label, rtl_entry);
1385 
1386 	return (label->rtl_id);
1387 }
1388 
1389 const char *
1390 rtlabel_id2name(u_int16_t id)
1391 {
1392 	struct rt_label	*label;
1393 
1394 	TAILQ_FOREACH(label, &rt_labels, rtl_entry)
1395 		if (label->rtl_id == id)
1396 			return (label->rtl_name);
1397 
1398 	return (NULL);
1399 }
1400 
1401 void
1402 rtlabel_unref(u_int16_t id)
1403 {
1404 	struct rt_label	*p, *next;
1405 
1406 	if (id == 0)
1407 		return;
1408 
1409 	for (p = TAILQ_FIRST(&rt_labels); p != NULL; p = next) {
1410 		next = TAILQ_NEXT(p, rtl_entry);
1411 		if (id == p->rtl_id) {
1412 			if (--p->rtl_ref == 0) {
1413 				TAILQ_REMOVE(&rt_labels, p, rtl_entry);
1414 				free(p, M_TEMP);
1415 			}
1416 			break;
1417 		}
1418 	}
1419 }
1420 
1421 void
1422 rt_if_remove(struct ifnet *ifp)
1423 {
1424 	int			 i;
1425 	struct radix_node_head	*rnh;
1426 
1427 	for (i = 1; i <= AF_MAX; i++)
1428 		if ((rnh = rt_gettable(i, 0)) != NULL)
1429 			while ((*rnh->rnh_walktree)(rnh,
1430 			    rt_if_remove_rtdelete, ifp) == EAGAIN)
1431 				;	/* nothing */
1432 }
1433 
1434 /*
1435  * Note that deleting a RTF_CLONING route can trigger the
1436  * deletion of more entries, so we need to cancel the walk
1437  * and return EAGAIN.  The caller should restart the walk
1438  * as long as EAGAIN is returned.
1439  */
1440 int
1441 rt_if_remove_rtdelete(struct radix_node *rn, void *vifp)
1442 {
1443 	struct ifnet	*ifp = vifp;
1444 	struct rtentry	*rt = (struct rtentry *)rn;
1445 
1446 	if (rt->rt_ifp == ifp) {
1447 		int	cloning = (rt->rt_flags & RTF_CLONING);
1448 
1449 		if (rtdeletemsg(rt, 0) == 0 && cloning)
1450 			return (EAGAIN);
1451 	}
1452 
1453 	/*
1454 	 * XXX There should be no need to check for rt_ifa belonging to this
1455 	 * interface, because then rt_ifp is set, right?
1456 	 */
1457 
1458 	return (0);
1459 }
1460 
1461 #ifndef SMALL_KERNEL
1462 void
1463 rt_if_track(struct ifnet *ifp)
1464 {
1465 	struct radix_node_head *rnh;
1466 	int i;
1467 	u_int tid;
1468 
1469 	if (rt_tables == NULL)
1470 		return;
1471 
1472 	for (tid = 0; tid <= rtbl_id_max; tid++) {
1473 		for (i = 1; i <= AF_MAX; i++) {
1474 			if ((rnh = rt_gettable(i, tid)) != NULL) {
1475 				if (!rn_mpath_capable(rnh))
1476 					continue;
1477 				while ((*rnh->rnh_walktree)(rnh,
1478 				    rt_if_linkstate_change, ifp) == EAGAIN)
1479 					;	/* nothing */
1480 			}
1481 		}
1482 	}
1483 }
1484 
1485 int
1486 rt_if_linkstate_change(struct radix_node *rn, void *arg)
1487 {
1488 	struct ifnet *ifp = arg;
1489 	struct rtentry *rt = (struct rtentry *)rn;
1490 
1491 	if (rt->rt_ifp == ifp) {
1492 		if ((LINK_STATE_IS_UP(ifp->if_link_state) ||
1493 		    ifp->if_link_state == LINK_STATE_UNKNOWN) &&
1494 		    ifp->if_flags & IFF_UP) {
1495 			if (!(rt->rt_flags & RTF_UP)) {
1496 				/* bring route up */
1497 				rt->rt_flags |= RTF_UP;
1498 				rn_mpath_reprio(rn, rt->rt_priority & RTP_MASK);
1499 			}
1500 		} else {
1501 			if (rt->rt_flags & RTF_UP) {
1502 				/* take route done */
1503 				rt->rt_flags &= ~RTF_UP;
1504 				rn_mpath_reprio(rn, rt->rt_priority | RTP_DOWN);
1505 			}
1506 		}
1507 		if_group_routechange(rt_key(rt), rt_mask(rt));
1508 	}
1509 
1510 	return (0);
1511 }
1512 #endif
1513