Deleted Added
sdiff udiff text old ( 85052 ) new ( 85074 )
full compact
1/*
2 * Copyright (c) 1980, 1986, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)route.c 8.3 (Berkeley) 1/9/95
34 * $FreeBSD: head/sys/net/route.c 85074 2001-10-17 18:07:05Z ru $
35 */
36
37#include "opt_inet.h"
38#include "opt_mrouting.h"
39
40#include <sys/param.h>
41#include <sys/systm.h>
42#include <sys/malloc.h>
43#include <sys/mbuf.h>
44#include <sys/socket.h>
45#include <sys/domain.h>
46#include <sys/kernel.h>
47
48#include <net/if.h>
49#include <net/route.h>
50
51#include <netinet/in.h>
52#include <netinet/ip_mroute.h>
53
54#define SA(p) ((struct sockaddr *)(p))
55
56struct route_cb route_cb;
57static struct rtstat rtstat;
58struct radix_node_head *rt_tables[AF_MAX+1];
59
60static int rttrash; /* routes not in table but not freed */
61
62static void rt_maskedcopy __P((struct sockaddr *,
63 struct sockaddr *, struct sockaddr *));
64static void rtable_init __P((void **));
65
66static void
67rtable_init(table)
68 void **table;
69{
70 struct domain *dom;
71 for (dom = domains; dom; dom = dom->dom_next)
72 if (dom->dom_rtattach)
73 dom->dom_rtattach(&table[dom->dom_family],
74 dom->dom_rtoffset);
75}
76
77void
78route_init()
79{
80 rn_init(); /* initialize all zeroes, all ones, mask table */
81 rtable_init((void **)rt_tables);
82}
83
84/*
85 * Packet routing routines.
86 */
87void
88rtalloc(ro)
89 register struct route *ro;
90{
91 rtalloc_ign(ro, 0UL);
92}
93
94void
95rtalloc_ign(ro, ignore)
96 register struct route *ro;
97 u_long ignore;
98{
99 struct rtentry *rt;
100 int s;
101
102 if ((rt = ro->ro_rt) != NULL) {
103 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
104 return;
105 /* XXX - We are probably always at splnet here already. */
106 s = splnet();
107 RTFREE(rt);
108 ro->ro_rt = NULL;
109 splx(s);
110 }
111 ro->ro_rt = rtalloc1(&ro->ro_dst, 1, ignore);
112}
113
114/*
115 * Look up the route that matches the address given
116 * Or, at least try.. Create a cloned route if needed.
117 */
118struct rtentry *
119rtalloc1(dst, report, ignflags)
120 register struct sockaddr *dst;
121 int report;
122 u_long ignflags;
123{
124 register struct radix_node_head *rnh = rt_tables[dst->sa_family];
125 register struct rtentry *rt;
126 register struct radix_node *rn;
127 struct rtentry *newrt = 0;
128 struct rt_addrinfo info;
129 u_long nflags;
130 int s = splnet(), err = 0, msgtype = RTM_MISS;
131
132 /*
133 * Look up the address in the table for that Address Family
134 */
135 if (rnh && (rn = rnh->rnh_matchaddr((caddr_t)dst, rnh)) &&
136 ((rn->rn_flags & RNF_ROOT) == 0)) {
137 /*
138 * If we find it and it's not the root node, then
139 * get a refernce on the rtentry associated.
140 */
141 newrt = rt = (struct rtentry *)rn;
142 nflags = rt->rt_flags & ~ignflags;
143 if (report && (nflags & (RTF_CLONING | RTF_PRCLONING))) {
144 /*
145 * We are apparently adding (report = 0 in delete).
146 * If it requires that it be cloned, do so.
147 * (This implies it wasn't a HOST route.)
148 */
149 err = rtrequest(RTM_RESOLVE, dst, SA(0),
150 SA(0), 0, &newrt);
151 if (err) {
152 /*
153 * If the cloning didn't succeed, maybe
154 * what we have will do. Return that.
155 */
156 newrt = rt;
157 rt->rt_refcnt++;
158 goto miss;
159 }
160 if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) {
161 /*
162 * If the new route specifies it be
163 * externally resolved, then go do that.
164 */
165 msgtype = RTM_RESOLVE;
166 goto miss;
167 }
168 /* Inform listeners of the new route. */
169 bzero(&info, sizeof(info));
170 info.rti_info[RTAX_DST] = rt_key(rt);
171 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
172 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
173 if (rt->rt_ifp != NULL) {
174 info.rti_info[RTAX_IFP] =
175 TAILQ_FIRST(&rt->rt_ifp->if_addrhead)->ifa_addr;
176 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
177 }
178 rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0);
179 } else
180 rt->rt_refcnt++;
181 } else {
182 /*
183 * Either we hit the root or couldn't find any match,
184 * Which basically means
185 * "caint get there frm here"
186 */
187 rtstat.rts_unreach++;
188 miss: if (report) {
189 /*
190 * If required, report the failure to the supervising
191 * Authorities.
192 * For a delete, this is not an error. (report == 0)
193 */
194 bzero((caddr_t)&info, sizeof(info));
195 info.rti_info[RTAX_DST] = dst;
196 rt_missmsg(msgtype, &info, 0, err);
197 }
198 }
199 splx(s);
200 return (newrt);
201}
202
203/*
204 * Remove a reference count from an rtentry.
205 * If the count gets low enough, take it out of the routing table
206 */
207void
208rtfree(rt)
209 register struct rtentry *rt;
210{
211 /*
212 * find the tree for that address family
213 */
214 register struct radix_node_head *rnh =
215 rt_tables[rt_key(rt)->sa_family];
216 register struct ifaddr *ifa;
217
218 if (rt == 0 || rnh == 0)
219 panic("rtfree");
220
221 /*
222 * decrement the reference count by one and if it reaches 0,
223 * and there is a close function defined, call the close function
224 */
225 rt->rt_refcnt--;
226 if(rnh->rnh_close && rt->rt_refcnt == 0) {
227 rnh->rnh_close((struct radix_node *)rt, rnh);
228 }
229
230 /*
231 * If we are no longer "up" (and ref == 0)
232 * then we can free the resources associated
233 * with the route.
234 */
235 if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
236 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
237 panic ("rtfree 2");
238 /*
239 * the rtentry must have been removed from the routing table
240 * so it is represented in rttrash.. remove that now.
241 */
242 rttrash--;
243
244#ifdef DIAGNOSTIC
245 if (rt->rt_refcnt < 0) {
246 printf("rtfree: %p not freed (neg refs)\n", rt);
247 return;
248 }
249#endif
250
251 /*
252 * release references on items we hold them on..
253 * e.g other routes and ifaddrs.
254 */
255 if((ifa = rt->rt_ifa))
256 IFAFREE(ifa);
257 if (rt->rt_parent) {
258 RTFREE(rt->rt_parent);
259 }
260
261 /*
262 * The key is separatly alloc'd so free it (see rt_setgate()).
263 * This also frees the gateway, as they are always malloc'd
264 * together.
265 */
266 Free(rt_key(rt));
267
268 /*
269 * and the rtentry itself of course
270 */
271 Free(rt);
272 }
273}
274
275void
276ifafree(ifa)
277 register struct ifaddr *ifa;
278{
279 if (ifa == NULL)
280 panic("ifafree");
281 if (ifa->ifa_refcnt == 0)
282 free(ifa, M_IFADDR);
283 else
284 ifa->ifa_refcnt--;
285}
286
287/*
288 * Force a routing table entry to the specified
289 * destination to go through the given gateway.
290 * Normally called as a result of a routing redirect
291 * message from the network layer.
292 *
293 * N.B.: must be called at splnet
294 *
295 */
296void
297rtredirect(dst, gateway, netmask, flags, src, rtp)
298 struct sockaddr *dst, *gateway, *netmask, *src;
299 int flags;
300 struct rtentry **rtp;
301{
302 struct rtentry *rt;
303 int error = 0;
304 short *stat = 0;
305 struct rt_addrinfo info;
306 struct ifaddr *ifa;
307
308 /* verify the gateway is directly reachable */
309 if ((ifa = ifa_ifwithnet(gateway)) == 0) {
310 error = ENETUNREACH;
311 goto out;
312 }
313 rt = rtalloc1(dst, 0, 0UL);
314 /*
315 * If the redirect isn't from our current router for this dst,
316 * it's either old or wrong. If it redirects us to ourselves,
317 * we have a routing loop, perhaps as a result of an interface
318 * going down recently.
319 */
320#define equal(a1, a2) (bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0)
321 if (!(flags & RTF_DONE) && rt &&
322 (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa))
323 error = EINVAL;
324 else if (ifa_ifwithaddr(gateway))
325 error = EHOSTUNREACH;
326 if (error)
327 goto done;
328 /*
329 * Create a new entry if we just got back a wildcard entry
330 * or the the lookup failed. This is necessary for hosts
331 * which use routing redirects generated by smart gateways
332 * to dynamically build the routing tables.
333 */
334 if ((rt == 0) || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
335 goto create;
336 /*
337 * Don't listen to the redirect if it's
338 * for a route to an interface.
339 */
340 if (rt->rt_flags & RTF_GATEWAY) {
341 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
342 /*
343 * Changing from route to net => route to host.
344 * Create new route, rather than smashing route to net.
345 */
346 create:
347 if (rt)
348 rtfree(rt);
349 flags |= RTF_GATEWAY | RTF_DYNAMIC;
350 bzero((caddr_t)&info, sizeof(info));
351 info.rti_info[RTAX_DST] = dst;
352 info.rti_info[RTAX_GATEWAY] = gateway;
353 info.rti_info[RTAX_NETMASK] = netmask;
354 info.rti_ifa = ifa;
355 info.rti_flags = flags;
356 rt = NULL;
357 error = rtrequest1(RTM_ADD, &info, &rt);
358 if (rt != NULL)
359 flags = rt->rt_flags;
360 stat = &rtstat.rts_dynamic;
361 } else {
362 /*
363 * Smash the current notion of the gateway to
364 * this destination. Should check about netmask!!!
365 */
366 rt->rt_flags |= RTF_MODIFIED;
367 flags |= RTF_MODIFIED;
368 stat = &rtstat.rts_newgateway;
369 /*
370 * add the key and gateway (in one malloc'd chunk).
371 */
372 rt_setgate(rt, rt_key(rt), gateway);
373 }
374 } else
375 error = EHOSTUNREACH;
376done:
377 if (rt) {
378 if (rtp && !error)
379 *rtp = rt;
380 else
381 rtfree(rt);
382 }
383out:
384 if (error)
385 rtstat.rts_badredirect++;
386 else if (stat != NULL)
387 (*stat)++;
388 bzero((caddr_t)&info, sizeof(info));
389 info.rti_info[RTAX_DST] = dst;
390 info.rti_info[RTAX_GATEWAY] = gateway;
391 info.rti_info[RTAX_NETMASK] = netmask;
392 info.rti_info[RTAX_AUTHOR] = src;
393 rt_missmsg(RTM_REDIRECT, &info, flags, error);
394}
395
396/*
397* Routing table ioctl interface.
398*/
399int
400rtioctl(req, data)
401 u_long req;
402 caddr_t data;
403{
404#ifdef INET
405 /* Multicast goop, grrr... */
406 return mrt_ioctl(req, data);
407#else /* INET */
408 return ENXIO;
409#endif /* INET */
410}
411
412struct ifaddr *
413ifa_ifwithroute(flags, dst, gateway)
414 int flags;
415 struct sockaddr *dst, *gateway;
416{
417 register struct ifaddr *ifa;
418 if ((flags & RTF_GATEWAY) == 0) {
419 /*
420 * If we are adding a route to an interface,
421 * and the interface is a pt to pt link
422 * we should search for the destination
423 * as our clue to the interface. Otherwise
424 * we can use the local address.
425 */
426 ifa = 0;
427 if (flags & RTF_HOST) {
428 ifa = ifa_ifwithdstaddr(dst);
429 }
430 if (ifa == 0)
431 ifa = ifa_ifwithaddr(gateway);
432 } else {
433 /*
434 * If we are adding a route to a remote net
435 * or host, the gateway may still be on the
436 * other end of a pt to pt link.
437 */
438 ifa = ifa_ifwithdstaddr(gateway);
439 }
440 if (ifa == 0)
441 ifa = ifa_ifwithnet(gateway);
442 if (ifa == 0) {
443 struct rtentry *rt = rtalloc1(gateway, 0, 0UL);
444 if (rt == 0)
445 return (0);
446 rt->rt_refcnt--;
447 if ((ifa = rt->rt_ifa) == 0)
448 return (0);
449 }
450 if (ifa->ifa_addr->sa_family != dst->sa_family) {
451 struct ifaddr *oifa = ifa;
452 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
453 if (ifa == 0)
454 ifa = oifa;
455 }
456 return (ifa);
457}
458
459static int rt_fixdelete __P((struct radix_node *, void *));
460static int rt_fixchange __P((struct radix_node *, void *));
461
462struct rtfc_arg {
463 struct rtentry *rt0;
464 struct radix_node_head *rnh;
465};
466
467/*
468 * Do appropriate manipulations of a routing tree given
469 * all the bits of info needed
470 */
471int
472rtrequest(req, dst, gateway, netmask, flags, ret_nrt)
473 int req, flags;
474 struct sockaddr *dst, *gateway, *netmask;
475 struct rtentry **ret_nrt;
476{
477 struct rt_addrinfo info;
478
479 bzero((caddr_t)&info, sizeof(info));
480 info.rti_flags = flags;
481 info.rti_info[RTAX_DST] = dst;
482 info.rti_info[RTAX_GATEWAY] = gateway;
483 info.rti_info[RTAX_NETMASK] = netmask;
484 return rtrequest1(req, &info, ret_nrt);
485}
486
487/*
488 * These (questionable) definitions of apparent local variables apply
489 * to the next two functions. XXXXXX!!!
490 */
491#define dst info->rti_info[RTAX_DST]
492#define gateway info->rti_info[RTAX_GATEWAY]
493#define netmask info->rti_info[RTAX_NETMASK]
494#define ifaaddr info->rti_info[RTAX_IFA]
495#define ifpaddr info->rti_info[RTAX_IFP]
496#define flags info->rti_flags
497
498int
499rt_getifa(info)
500 struct rt_addrinfo *info;
501{
502 struct ifaddr *ifa;
503 int error = 0;
504
505 /*
506 * ifp may be specified by sockaddr_dl
507 * when protocol address is ambiguous.
508 */
509 if (info->rti_ifp == NULL && ifpaddr != NULL &&
510 ifpaddr->sa_family == AF_LINK &&
511 (ifa = ifa_ifwithnet(ifpaddr)) != NULL)
512 info->rti_ifp = ifa->ifa_ifp;
513 if (info->rti_ifa == NULL && ifaaddr != NULL)
514 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
515 if (info->rti_ifa == NULL) {
516 struct sockaddr *sa;
517
518 sa = ifaaddr != NULL ? ifaaddr :
519 (gateway != NULL ? gateway : dst);
520 if (sa != NULL && info->rti_ifp != NULL)
521 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
522 else if (dst != NULL && gateway != NULL)
523 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
524 else if (sa != NULL)
525 info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
526 }
527 if ((ifa = info->rti_ifa) != NULL) {
528 if (info->rti_ifp == NULL)
529 info->rti_ifp = ifa->ifa_ifp;
530 } else
531 error = ENETUNREACH;
532 return (error);
533}
534
535int
536rtrequest1(req, info, ret_nrt)
537 int req;
538 struct rt_addrinfo *info;
539 struct rtentry **ret_nrt;
540{
541 int s = splnet(); int error = 0;
542 register struct rtentry *rt;
543 register struct radix_node *rn;
544 register struct radix_node_head *rnh;
545 struct ifaddr *ifa;
546 struct sockaddr *ndst;
547#define senderr(x) { error = x ; goto bad; }
548
549 /*
550 * Find the correct routing tree to use for this Address Family
551 */
552 if ((rnh = rt_tables[dst->sa_family]) == 0)
553 senderr(EAFNOSUPPORT);
554 /*
555 * If we are adding a host route then we don't want to put
556 * a netmask in the tree, nor do we want to clone it.
557 */
558 if (flags & RTF_HOST) {
559 netmask = 0;
560 flags &= ~(RTF_CLONING | RTF_PRCLONING);
561 }
562 switch (req) {
563 case RTM_DELETE:
564 /*
565 * Remove the item from the tree and return it.
566 * Complain if it is not there and do no more processing.
567 */
568 if ((rn = rnh->rnh_deladdr(dst, netmask, rnh)) == 0)
569 senderr(ESRCH);
570 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
571 panic ("rtrequest delete");
572 rt = (struct rtentry *)rn;
573
574 /*
575 * Now search what's left of the subtree for any cloned
576 * routes which might have been formed from this node.
577 */
578 if ((rt->rt_flags & (RTF_CLONING | RTF_PRCLONING)) &&
579 rt_mask(rt)) {
580 rnh->rnh_walktree_from(rnh, dst, rt_mask(rt),
581 rt_fixdelete, rt);
582 }
583
584 /*
585 * Remove any external references we may have.
586 * This might result in another rtentry being freed if
587 * we held its last reference.
588 */
589 if (rt->rt_gwroute) {
590 rt = rt->rt_gwroute;
591 RTFREE(rt);
592 (rt = (struct rtentry *)rn)->rt_gwroute = 0;
593 }
594
595 /*
596 * NB: RTF_UP must be set during the search above,
597 * because we might delete the last ref, causing
598 * rt to get freed prematurely.
599 * eh? then why not just add a reference?
600 * I'm not sure how RTF_UP helps matters. (JRE)
601 */
602 rt->rt_flags &= ~RTF_UP;
603
604 /*
605 * give the protocol a chance to keep things in sync.
606 */
607 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
608 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
609
610 /*
611 * one more rtentry floating around that is not
612 * linked to the routing table.
613 */
614 rttrash++;
615
616 /*
617 * If the caller wants it, then it can have it,
618 * but it's up to it to free the rtentry as we won't be
619 * doing it.
620 */
621 if (ret_nrt)
622 *ret_nrt = rt;
623 else if (rt->rt_refcnt <= 0) {
624 rt->rt_refcnt++; /* make a 1->0 transition */
625 rtfree(rt);
626 }
627 break;
628
629 case RTM_RESOLVE:
630 if (ret_nrt == 0 || (rt = *ret_nrt) == 0)
631 senderr(EINVAL);
632 ifa = rt->rt_ifa;
633 flags = rt->rt_flags &
634 ~(RTF_CLONING | RTF_PRCLONING | RTF_STATIC);
635 flags |= RTF_WASCLONED;
636 gateway = rt->rt_gateway;
637 if ((netmask = rt->rt_genmask) == 0)
638 flags |= RTF_HOST;
639 goto makeroute;
640
641 case RTM_ADD:
642 if ((flags & RTF_GATEWAY) && !gateway)
643 panic("rtrequest: GATEWAY but no gateway");
644
645 if (info->rti_ifa == NULL && (error = rt_getifa(info)))
646 senderr(error);
647 ifa = info->rti_ifa;
648
649 makeroute:
650 R_Malloc(rt, struct rtentry *, sizeof(*rt));
651 if (rt == 0)
652 senderr(ENOBUFS);
653 Bzero(rt, sizeof(*rt));
654 rt->rt_flags = RTF_UP | flags;
655 /*
656 * Add the gateway. Possibly re-malloc-ing the storage for it
657 * also add the rt_gwroute if possible.
658 */
659 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
660 Free(rt);
661 senderr(error);
662 }
663
664 /*
665 * point to the (possibly newly malloc'd) dest address.
666 */
667 ndst = rt_key(rt);
668
669 /*
670 * make sure it contains the value we want (masked if needed).
671 */
672 if (netmask) {
673 rt_maskedcopy(dst, ndst, netmask);
674 } else
675 Bcopy(dst, ndst, dst->sa_len);
676
677 /*
678 * Note that we now have a reference to the ifa.
679 * This moved from below so that rnh->rnh_addaddr() can
680 * examine the ifa and ifa->ifa_ifp if it so desires.
681 */
682 ifa->ifa_refcnt++;
683 rt->rt_ifa = ifa;
684 rt->rt_ifp = ifa->ifa_ifp;
685 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
686
687 rn = rnh->rnh_addaddr((caddr_t)ndst, (caddr_t)netmask,
688 rnh, rt->rt_nodes);
689 if (rn == 0) {
690 struct rtentry *rt2;
691 /*
692 * Uh-oh, we already have one of these in the tree.
693 * We do a special hack: if the route that's already
694 * there was generated by the protocol-cloning
695 * mechanism, then we just blow it away and retry
696 * the insertion of the new one.
697 */
698 rt2 = rtalloc1(dst, 0, RTF_PRCLONING);
699 if (rt2 && rt2->rt_parent) {
700 rtrequest(RTM_DELETE,
701 (struct sockaddr *)rt_key(rt2),
702 rt2->rt_gateway,
703 rt_mask(rt2), rt2->rt_flags, 0);
704 RTFREE(rt2);
705 rn = rnh->rnh_addaddr((caddr_t)ndst,
706 (caddr_t)netmask,
707 rnh, rt->rt_nodes);
708 } else if (rt2) {
709 /* undo the extra ref we got */
710 RTFREE(rt2);
711 }
712 }
713
714 /*
715 * If it still failed to go into the tree,
716 * then un-make it (this should be a function)
717 */
718 if (rn == 0) {
719 if (rt->rt_gwroute)
720 rtfree(rt->rt_gwroute);
721 if (rt->rt_ifa) {
722 IFAFREE(rt->rt_ifa);
723 }
724 Free(rt_key(rt));
725 Free(rt);
726 senderr(EEXIST);
727 }
728
729 rt->rt_parent = 0;
730
731 /*
732 * If we got here from RESOLVE, then we are cloning
733 * so clone the rest, and note that we
734 * are a clone (and increment the parent's references)
735 */
736 if (req == RTM_RESOLVE) {
737 rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
738 if ((*ret_nrt)->rt_flags & (RTF_CLONING | RTF_PRCLONING)) {
739 rt->rt_parent = (*ret_nrt);
740 (*ret_nrt)->rt_refcnt++;
741 }
742 }
743
744 /*
745 * if this protocol has something to add to this then
746 * allow it to do that as well.
747 */
748 if (ifa->ifa_rtrequest)
749 ifa->ifa_rtrequest(req, rt, info);
750
751 /*
752 * We repeat the same procedure from rt_setgate() here because
753 * it doesn't fire when we call it there because the node
754 * hasn't been added to the tree yet.
755 */
756 if (!(rt->rt_flags & RTF_HOST) && rt_mask(rt) != 0) {
757 struct rtfc_arg arg;
758 arg.rnh = rnh;
759 arg.rt0 = rt;
760 rnh->rnh_walktree_from(rnh, rt_key(rt), rt_mask(rt),
761 rt_fixchange, &arg);
762 }
763
764 /*
765 * actually return a resultant rtentry and
766 * give the caller a single reference.
767 */
768 if (ret_nrt) {
769 *ret_nrt = rt;
770 rt->rt_refcnt++;
771 }
772 break;
773 default:
774 error = EOPNOTSUPP;
775 }
776bad:
777 splx(s);
778 return (error);
779#undef dst
780#undef gateway
781#undef netmask
782#undef ifaaddr
783#undef ifpaddr
784#undef flags
785}
786
787/*
788 * Called from rtrequest(RTM_DELETE, ...) to fix up the route's ``family''
789 * (i.e., the routes related to it by the operation of cloning). This
790 * routine is iterated over all potential former-child-routes by way of
791 * rnh->rnh_walktree_from() above, and those that actually are children of
792 * the late parent (passed in as VP here) are themselves deleted.
793 */
794static int
795rt_fixdelete(rn, vp)
796 struct radix_node *rn;
797 void *vp;
798{
799 struct rtentry *rt = (struct rtentry *)rn;
800 struct rtentry *rt0 = vp;
801
802 if (rt->rt_parent == rt0 && !(rt->rt_flags & RTF_PINNED)) {
803 return rtrequest(RTM_DELETE, rt_key(rt),
804 (struct sockaddr *)0, rt_mask(rt),
805 rt->rt_flags, (struct rtentry **)0);
806 }
807 return 0;
808}
809
810/*
811 * This routine is called from rt_setgate() to do the analogous thing for
812 * adds and changes. There is the added complication in this case of a
813 * middle insert; i.e., insertion of a new network route between an older
814 * network route and (cloned) host routes. For this reason, a simple check
815 * of rt->rt_parent is insufficient; each candidate route must be tested
816 * against the (mask, value) of the new route (passed as before in vp)
817 * to see if the new route matches it.
818 *
819 * XXX - it may be possible to do fixdelete() for changes and reserve this
820 * routine just for adds. I'm not sure why I thought it was necessary to do
821 * changes this way.
822 */
823#ifdef DEBUG
824static int rtfcdebug = 0;
825#endif
826
827static int
828rt_fixchange(rn, vp)
829 struct radix_node *rn;
830 void *vp;
831{
832 struct rtentry *rt = (struct rtentry *)rn;
833 struct rtfc_arg *ap = vp;
834 struct rtentry *rt0 = ap->rt0;
835 struct radix_node_head *rnh = ap->rnh;
836 u_char *xk1, *xm1, *xk2, *xmp;
837 int i, len, mlen;
838
839#ifdef DEBUG
840 if (rtfcdebug)
841 printf("rt_fixchange: rt %p, rt0 %p\n", rt, rt0);
842#endif
843
844 if (!rt->rt_parent || (rt->rt_flags & RTF_PINNED)) {
845#ifdef DEBUG
846 if(rtfcdebug) printf("no parent or pinned\n");
847#endif
848 return 0;
849 }
850
851 if (rt->rt_parent == rt0) {
852#ifdef DEBUG
853 if(rtfcdebug) printf("parent match\n");
854#endif
855 return rtrequest(RTM_DELETE, rt_key(rt),
856 (struct sockaddr *)0, rt_mask(rt),
857 rt->rt_flags, (struct rtentry **)0);
858 }
859
860 /*
861 * There probably is a function somewhere which does this...
862 * if not, there should be.
863 */
864 len = imin(((struct sockaddr *)rt_key(rt0))->sa_len,
865 ((struct sockaddr *)rt_key(rt))->sa_len);
866
867 xk1 = (u_char *)rt_key(rt0);
868 xm1 = (u_char *)rt_mask(rt0);
869 xk2 = (u_char *)rt_key(rt);
870
871 /* avoid applying a less specific route */
872 xmp = (u_char *)rt_mask(rt->rt_parent);
873 mlen = ((struct sockaddr *)rt_key(rt->rt_parent))->sa_len;
874 if (mlen > ((struct sockaddr *)rt_key(rt0))->sa_len) {
875#ifdef DEBUG
876 if (rtfcdebug)
877 printf("rt_fixchange: inserting a less "
878 "specific route\n");
879#endif
880 return 0;
881 }
882 for (i = rnh->rnh_treetop->rn_offset; i < mlen; i++) {
883 if ((xmp[i] & ~(xmp[i] ^ xm1[i])) != xmp[i]) {
884#ifdef DEBUG
885 if (rtfcdebug)
886 printf("rt_fixchange: inserting a less "
887 "specific route\n");
888#endif
889 return 0;
890 }
891 }
892
893 for (i = rnh->rnh_treetop->rn_offset; i < len; i++) {
894 if ((xk2[i] & xm1[i]) != xk1[i]) {
895#ifdef DEBUG
896 if(rtfcdebug) printf("no match\n");
897#endif
898 return 0;
899 }
900 }
901
902 /*
903 * OK, this node is a clone, and matches the node currently being
904 * changed/added under the node's mask. So, get rid of it.
905 */
906#ifdef DEBUG
907 if(rtfcdebug) printf("deleting\n");
908#endif
909 return rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0,
910 rt_mask(rt), rt->rt_flags, (struct rtentry **)0);
911}
912
913#define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
914
915int
916rt_setgate(rt0, dst, gate)
917 struct rtentry *rt0;
918 struct sockaddr *dst, *gate;
919{
920 caddr_t new, old;
921 int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len);
922 register struct rtentry *rt = rt0;
923 struct radix_node_head *rnh = rt_tables[dst->sa_family];
924
925 /*
926 * A host route with the destination equal to the gateway
927 * will interfere with keeping LLINFO in the routing
928 * table, so disallow it.
929 */
930 if (((rt0->rt_flags & (RTF_HOST|RTF_GATEWAY|RTF_LLINFO)) ==
931 (RTF_HOST|RTF_GATEWAY)) &&
932 (dst->sa_len == gate->sa_len) &&
933 (bcmp(dst, gate, dst->sa_len) == 0)) {
934 /*
935 * The route might already exist if this is an RTM_CHANGE
936 * or a routing redirect, so try to delete it.
937 */
938 if (rt_key(rt0))
939 rtrequest(RTM_DELETE, (struct sockaddr *)rt_key(rt0),
940 rt0->rt_gateway, rt_mask(rt0), rt0->rt_flags, 0);
941 return EADDRNOTAVAIL;
942 }
943
944 /*
945 * Both dst and gateway are stored in the same malloc'd chunk
946 * (If I ever get my hands on....)
947 * if we need to malloc a new chunk, then keep the old one around
948 * till we don't need it any more.
949 */
950 if (rt->rt_gateway == 0 || glen > ROUNDUP(rt->rt_gateway->sa_len)) {
951 old = (caddr_t)rt_key(rt);
952 R_Malloc(new, caddr_t, dlen + glen);
953 if (new == 0)
954 return ENOBUFS;
955 rt->rt_nodes->rn_key = new;
956 } else {
957 /*
958 * otherwise just overwrite the old one
959 */
960 new = rt->rt_nodes->rn_key;
961 old = 0;
962 }
963
964 /*
965 * copy the new gateway value into the memory chunk
966 */
967 Bcopy(gate, (rt->rt_gateway = (struct sockaddr *)(new + dlen)), glen);
968
969 /*
970 * if we are replacing the chunk (or it's new) we need to
971 * replace the dst as well
972 */
973 if (old) {
974 Bcopy(dst, new, dlen);
975 Free(old);
976 }
977
978 /*
979 * If there is already a gwroute, it's now almost definitly wrong
980 * so drop it.
981 */
982 if (rt->rt_gwroute != NULL) {
983 RTFREE(rt->rt_gwroute);
984 rt->rt_gwroute = NULL;
985 }
986 /*
987 * Cloning loop avoidance:
988 * In the presence of protocol-cloning and bad configuration,
989 * it is possible to get stuck in bottomless mutual recursion
990 * (rtrequest rt_setgate rtalloc1). We avoid this by not allowing
991 * protocol-cloning to operate for gateways (which is probably the
992 * correct choice anyway), and avoid the resulting reference loops
993 * by disallowing any route to run through itself as a gateway.
994 * This is obviously mandatory when we get rt->rt_output().
995 */
996 if (rt->rt_flags & RTF_GATEWAY) {
997 rt->rt_gwroute = rtalloc1(gate, 1, RTF_PRCLONING);
998 if (rt->rt_gwroute == rt) {
999 RTFREE(rt->rt_gwroute);
1000 rt->rt_gwroute = 0;
1001 return EDQUOT; /* failure */
1002 }
1003 }
1004
1005 /*
1006 * This isn't going to do anything useful for host routes, so
1007 * don't bother. Also make sure we have a reasonable mask
1008 * (we don't yet have one during adds).
1009 */
1010 if (!(rt->rt_flags & RTF_HOST) && rt_mask(rt) != 0) {
1011 struct rtfc_arg arg;
1012 arg.rnh = rnh;
1013 arg.rt0 = rt;
1014 rnh->rnh_walktree_from(rnh, rt_key(rt), rt_mask(rt),
1015 rt_fixchange, &arg);
1016 }
1017
1018 return 0;
1019}
1020
1021static void
1022rt_maskedcopy(src, dst, netmask)
1023 struct sockaddr *src, *dst, *netmask;
1024{
1025 register u_char *cp1 = (u_char *)src;
1026 register u_char *cp2 = (u_char *)dst;
1027 register u_char *cp3 = (u_char *)netmask;
1028 u_char *cplim = cp2 + *cp3;
1029 u_char *cplim2 = cp2 + *cp1;
1030
1031 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1032 cp3 += 2;
1033 if (cplim > cplim2)
1034 cplim = cplim2;
1035 while (cp2 < cplim)
1036 *cp2++ = *cp1++ & *cp3++;
1037 if (cp2 < cplim2)
1038 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1039}
1040
1041/*
1042 * Set up a routing table entry, normally
1043 * for an interface.
1044 */
1045int
1046rtinit(ifa, cmd, flags)
1047 register struct ifaddr *ifa;
1048 int cmd, flags;
1049{
1050 register struct rtentry *rt;
1051 register struct sockaddr *dst;
1052 register struct sockaddr *deldst;
1053 struct sockaddr *netmask;
1054 struct mbuf *m = 0;
1055 struct rtentry *nrt = 0;
1056 struct radix_node_head *rnh;
1057 struct radix_node *rn;
1058 int error;
1059 struct rt_addrinfo info;
1060
1061 if (flags & RTF_HOST) {
1062 dst = ifa->ifa_dstaddr;
1063 netmask = NULL;
1064 } else {
1065 dst = ifa->ifa_addr;
1066 netmask = ifa->ifa_netmask;
1067 }
1068 /*
1069 * If it's a delete, check that if it exists, it's on the correct
1070 * interface or we might scrub a route to another ifa which would
1071 * be confusing at best and possibly worse.
1072 */
1073 if (cmd == RTM_DELETE) {
1074 /*
1075 * It's a delete, so it should already exist..
1076 * If it's a net, mask off the host bits
1077 * (Assuming we have a mask)
1078 */
1079 if (netmask != NULL) {
1080 m = m_get(M_DONTWAIT, MT_SONAME);
1081 if (m == NULL)
1082 return(ENOBUFS);
1083 deldst = mtod(m, struct sockaddr *);
1084 rt_maskedcopy(dst, deldst, netmask);
1085 dst = deldst;
1086 }
1087 /*
1088 * Look up an rtentry that is in the routing tree and
1089 * contains the correct info.
1090 */
1091 if ((rnh = rt_tables[dst->sa_family]) == NULL ||
1092 (rn = rnh->rnh_lookup(dst, netmask, rnh)) == NULL ||
1093 (rn->rn_flags & RNF_ROOT) ||
1094 ((struct rtentry *)rn)->rt_ifa != ifa ||
1095 !equal(SA(rn->rn_key), dst)) {
1096 if (m)
1097 (void) m_free(m);
1098 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
1099 }
1100 /* XXX */
1101#if 0
1102 else {
1103 /*
1104 * One would think that as we are deleting, and we know
1105 * it doesn't exist, we could just return at this point
1106 * with an "ELSE" clause, but apparently not..
1107 */
1108 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
1109 }
1110#endif
1111 }
1112 /*
1113 * Do the actual request
1114 */
1115 bzero((caddr_t)&info, sizeof(info));
1116 info.rti_ifa = ifa;
1117 info.rti_flags = flags | ifa->ifa_flags;
1118 info.rti_info[RTAX_DST] = dst;
1119 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
1120 info.rti_info[RTAX_NETMASK] = netmask;
1121 error = rtrequest1(cmd, &info, &nrt);
1122 if (error == 0 && (rt = nrt) != NULL) {
1123 /*
1124 * notify any listenning routing agents of the change
1125 */
1126 rt_newaddrmsg(cmd, ifa, error, rt);
1127 if (cmd == RTM_DELETE) {
1128 /*
1129 * If we are deleting, and we found an entry, then
1130 * it's been removed from the tree.. now throw it away.
1131 */
1132 if (rt->rt_refcnt <= 0) {
1133 rt->rt_refcnt++; /* make a 1->0 transition */
1134 rtfree(rt);
1135 }
1136 } else if (cmd == RTM_ADD) {
1137 /*
1138 * We just wanted to add it.. we don't actually
1139 * need a reference.
1140 */
1141 rt->rt_refcnt--;
1142 }
1143 }
1144 if (m)
1145 (void) m_free(m);
1146 return (error);
1147}
1148
1149/* This must be before ip6_init2(), which is now SI_ORDER_MIDDLE */
1150SYSINIT(route, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0);