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 85052 2001-10-17 11:10:55Z 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 } else
169 rt->rt_refcnt++;
170 } else {
171 /*
172 * Either we hit the root or couldn't find any match,
173 * Which basically means
174 * "caint get there frm here"
175 */
176 rtstat.rts_unreach++;
177 miss: if (report) {
178 /*
179 * If required, report the failure to the supervising
180 * Authorities.
181 * For a delete, this is not an error. (report == 0)
182 */
183 bzero((caddr_t)&info, sizeof(info));
184 info.rti_info[RTAX_DST] = dst;
185 rt_missmsg(msgtype, &info, 0, err);
186 }
187 }
188 splx(s);
189 return (newrt);
190}
191
192/*
193 * Remove a reference count from an rtentry.
194 * If the count gets low enough, take it out of the routing table
195 */
196void
197rtfree(rt)
198 register struct rtentry *rt;
199{
200 /*
201 * find the tree for that address family
202 */
203 register struct radix_node_head *rnh =
204 rt_tables[rt_key(rt)->sa_family];
205 register struct ifaddr *ifa;
206
207 if (rt == 0 || rnh == 0)
208 panic("rtfree");
209
210 /*
211 * decrement the reference count by one and if it reaches 0,
212 * and there is a close function defined, call the close function
213 */
214 rt->rt_refcnt--;
215 if(rnh->rnh_close && rt->rt_refcnt == 0) {
216 rnh->rnh_close((struct radix_node *)rt, rnh);
217 }
218
219 /*
220 * If we are no longer "up" (and ref == 0)
221 * then we can free the resources associated
222 * with the route.
223 */
224 if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
225 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
226 panic ("rtfree 2");
227 /*
228 * the rtentry must have been removed from the routing table
229 * so it is represented in rttrash.. remove that now.
230 */
231 rttrash--;
232
233#ifdef DIAGNOSTIC
234 if (rt->rt_refcnt < 0) {
235 printf("rtfree: %p not freed (neg refs)\n", rt);
236 return;
237 }
238#endif
239
240 /*
241 * release references on items we hold them on..
242 * e.g other routes and ifaddrs.
243 */
244 if((ifa = rt->rt_ifa))
245 IFAFREE(ifa);
246 if (rt->rt_parent) {
247 RTFREE(rt->rt_parent);
248 }
249
250 /*
251 * The key is separatly alloc'd so free it (see rt_setgate()).
252 * This also frees the gateway, as they are always malloc'd
253 * together.
254 */
255 Free(rt_key(rt));
256
257 /*
258 * and the rtentry itself of course
259 */
260 Free(rt);
261 }
262}
263
264void
265ifafree(ifa)
266 register struct ifaddr *ifa;
267{
268 if (ifa == NULL)
269 panic("ifafree");
270 if (ifa->ifa_refcnt == 0)
271 free(ifa, M_IFADDR);
272 else
273 ifa->ifa_refcnt--;
274}
275
276/*
277 * Force a routing table entry to the specified
278 * destination to go through the given gateway.
279 * Normally called as a result of a routing redirect
280 * message from the network layer.
281 *
282 * N.B.: must be called at splnet
283 *
284 */
285void
286rtredirect(dst, gateway, netmask, flags, src, rtp)
287 struct sockaddr *dst, *gateway, *netmask, *src;
288 int flags;
289 struct rtentry **rtp;
290{
291 register struct rtentry *rt;
292 int error = 0;
293 short *stat = 0;
294 struct rt_addrinfo info;
295 struct ifaddr *ifa;
296
297 /* verify the gateway is directly reachable */
298 if ((ifa = ifa_ifwithnet(gateway)) == 0) {
299 error = ENETUNREACH;
300 goto out;
301 }
302 rt = rtalloc1(dst, 0, 0UL);
303 /*
304 * If the redirect isn't from our current router for this dst,
305 * it's either old or wrong. If it redirects us to ourselves,
306 * we have a routing loop, perhaps as a result of an interface
307 * going down recently.
308 */
309#define equal(a1, a2) (bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0)
310 if (!(flags & RTF_DONE) && rt &&
311 (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa))
312 error = EINVAL;
313 else if (ifa_ifwithaddr(gateway))
314 error = EHOSTUNREACH;
315 if (error)
316 goto done;
317 /*
318 * Create a new entry if we just got back a wildcard entry
319 * or the the lookup failed. This is necessary for hosts
320 * which use routing redirects generated by smart gateways
321 * to dynamically build the routing tables.
322 */
323 if ((rt == 0) || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
324 goto create;
325 /*
326 * Don't listen to the redirect if it's
327 * for a route to an interface.
328 */
329 if (rt->rt_flags & RTF_GATEWAY) {
330 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
331 /*
332 * Changing from route to net => route to host.
333 * Create new route, rather than smashing route to net.
334 */
335 create:
336 flags |= RTF_GATEWAY | RTF_DYNAMIC;
337 error = rtrequest((int)RTM_ADD, dst, gateway,
338 netmask, flags,
339 (struct rtentry **)0);
340 stat = &rtstat.rts_dynamic;
341 } else {
342 /*
343 * Smash the current notion of the gateway to
344 * this destination. Should check about netmask!!!
345 */
346 rt->rt_flags |= RTF_MODIFIED;
347 flags |= RTF_MODIFIED;
348 stat = &rtstat.rts_newgateway;
349 /*
350 * add the key and gateway (in one malloc'd chunk).
351 */
352 rt_setgate(rt, rt_key(rt), gateway);
353 }
354 } else
355 error = EHOSTUNREACH;
356done:
357 if (rt) {
358 if (rtp && !error)
359 *rtp = rt;
360 else
361 rtfree(rt);
362 }
363out:
364 if (error)
365 rtstat.rts_badredirect++;
366 else if (stat != NULL)
367 (*stat)++;
368 bzero((caddr_t)&info, sizeof(info));
369 info.rti_info[RTAX_DST] = dst;
370 info.rti_info[RTAX_GATEWAY] = gateway;
371 info.rti_info[RTAX_NETMASK] = netmask;
372 info.rti_info[RTAX_AUTHOR] = src;
373 rt_missmsg(RTM_REDIRECT, &info, flags, error);
374}
375
376/*
377* Routing table ioctl interface.
378*/
379int
380rtioctl(req, data)
381 u_long req;
382 caddr_t data;
383{
384#ifdef INET
385 /* Multicast goop, grrr... */
386 return mrt_ioctl(req, data);
387#else /* INET */
388 return ENXIO;
389#endif /* INET */
390}
391
392struct ifaddr *
393ifa_ifwithroute(flags, dst, gateway)
394 int flags;
395 struct sockaddr *dst, *gateway;
396{
397 register struct ifaddr *ifa;
398 if ((flags & RTF_GATEWAY) == 0) {
399 /*
400 * If we are adding a route to an interface,
401 * and the interface is a pt to pt link
402 * we should search for the destination
403 * as our clue to the interface. Otherwise
404 * we can use the local address.
405 */
406 ifa = 0;
407 if (flags & RTF_HOST) {
408 ifa = ifa_ifwithdstaddr(dst);
409 }
410 if (ifa == 0)
411 ifa = ifa_ifwithaddr(gateway);
412 } else {
413 /*
414 * If we are adding a route to a remote net
415 * or host, the gateway may still be on the
416 * other end of a pt to pt link.
417 */
418 ifa = ifa_ifwithdstaddr(gateway);
419 }
420 if (ifa == 0)
421 ifa = ifa_ifwithnet(gateway);
422 if (ifa == 0) {
423 struct rtentry *rt = rtalloc1(gateway, 0, 0UL);
424 if (rt == 0)
425 return (0);
426 rt->rt_refcnt--;
427 if ((ifa = rt->rt_ifa) == 0)
428 return (0);
429 }
430 if (ifa->ifa_addr->sa_family != dst->sa_family) {
431 struct ifaddr *oifa = ifa;
432 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
433 if (ifa == 0)
434 ifa = oifa;
435 }
436 return (ifa);
437}
438
439#define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
440
441static int rt_fixdelete __P((struct radix_node *, void *));
442static int rt_fixchange __P((struct radix_node *, void *));
443
444struct rtfc_arg {
445 struct rtentry *rt0;
446 struct radix_node_head *rnh;
447};
448
449/*
450 * Do appropriate manipulations of a routing tree given
451 * all the bits of info needed
452 */
453int
454rtrequest(req, dst, gateway, netmask, flags, ret_nrt)
455 int req, flags;
456 struct sockaddr *dst, *gateway, *netmask;
457 struct rtentry **ret_nrt;
458{
459 int s = splnet(); int error = 0;
460 register struct rtentry *rt;
461 register struct radix_node *rn;
462 register struct radix_node_head *rnh;
463 struct ifaddr *ifa;
464 struct sockaddr *ndst;
465#define senderr(x) { error = x ; goto bad; }
466
467 /*
468 * Find the correct routing tree to use for this Address Family
469 */
470 if ((rnh = rt_tables[dst->sa_family]) == 0)
471 senderr(ESRCH);
472 /*
473 * If we are adding a host route then we don't want to put
474 * a netmask in the tree, nor do we want to clone it.
475 */
476 if (flags & RTF_HOST) {
477 netmask = 0;
478 flags &= ~(RTF_CLONING | RTF_PRCLONING);
479 }
480 switch (req) {
481 case RTM_DELETE:
482 /*
483 * Remove the item from the tree and return it.
484 * Complain if it is not there and do no more processing.
485 */
486 if ((rn = rnh->rnh_deladdr(dst, netmask, rnh)) == 0)
487 senderr(ESRCH);
488 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
489 panic ("rtrequest delete");
490 rt = (struct rtentry *)rn;
491
492 /*
493 * Now search what's left of the subtree for any cloned
494 * routes which might have been formed from this node.
495 */
496 if ((rt->rt_flags & (RTF_CLONING | RTF_PRCLONING)) &&
497 rt_mask(rt)) {
498 rnh->rnh_walktree_from(rnh, dst, rt_mask(rt),
499 rt_fixdelete, rt);
500 }
501
502 /*
503 * Remove any external references we may have.
504 * This might result in another rtentry being freed if
505 * we held its last reference.
506 */
507 if (rt->rt_gwroute) {
508 rt = rt->rt_gwroute;
509 RTFREE(rt);
510 (rt = (struct rtentry *)rn)->rt_gwroute = 0;
511 }
512
513 /*
514 * NB: RTF_UP must be set during the search above,
515 * because we might delete the last ref, causing
516 * rt to get freed prematurely.
517 * eh? then why not just add a reference?
518 * I'm not sure how RTF_UP helps matters. (JRE)
519 */
520 rt->rt_flags &= ~RTF_UP;
521
522 /*
523 * give the protocol a chance to keep things in sync.
524 */
525 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
526 ifa->ifa_rtrequest(RTM_DELETE, rt, SA(0));
527
528 /*
529 * one more rtentry floating around that is not
530 * linked to the routing table.
531 */
532 rttrash++;
533
534 /*
535 * If the caller wants it, then it can have it,
536 * but it's up to it to free the rtentry as we won't be
537 * doing it.
538 */
539 if (ret_nrt)
540 *ret_nrt = rt;
541 else if (rt->rt_refcnt <= 0) {
542 rt->rt_refcnt++; /* make a 1->0 transition */
543 rtfree(rt);
544 }
545 break;
546
547 case RTM_RESOLVE:
548 if (ret_nrt == 0 || (rt = *ret_nrt) == 0)
549 senderr(EINVAL);
550 ifa = rt->rt_ifa;
551 flags = rt->rt_flags &
552 ~(RTF_CLONING | RTF_PRCLONING | RTF_STATIC);
553 flags |= RTF_WASCLONED;
554 gateway = rt->rt_gateway;
555 if ((netmask = rt->rt_genmask) == 0)
556 flags |= RTF_HOST;
557 goto makeroute;
558
559 case RTM_ADD:
560 if ((flags & RTF_GATEWAY) && !gateway)
561 panic("rtrequest: GATEWAY but no gateway");
562
563 if ((ifa = ifa_ifwithroute(flags, dst, gateway)) == 0)
564 senderr(ENETUNREACH);
565
566 makeroute:
567 R_Malloc(rt, struct rtentry *, sizeof(*rt));
568 if (rt == 0)
569 senderr(ENOBUFS);
570 Bzero(rt, sizeof(*rt));
571 rt->rt_flags = RTF_UP | flags;
572 /*
573 * Add the gateway. Possibly re-malloc-ing the storage for it
574 * also add the rt_gwroute if possible.
575 */
576 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
577 Free(rt);
578 senderr(error);
579 }
580
581 /*
582 * point to the (possibly newly malloc'd) dest address.
583 */
584 ndst = rt_key(rt);
585
586 /*
587 * make sure it contains the value we want (masked if needed).
588 */
589 if (netmask) {
590 rt_maskedcopy(dst, ndst, netmask);
591 } else
592 Bcopy(dst, ndst, dst->sa_len);
593
594 /*
595 * Note that we now have a reference to the ifa.
596 * This moved from below so that rnh->rnh_addaddr() can
597 * examine the ifa and ifa->ifa_ifp if it so desires.
598 */
599 ifa->ifa_refcnt++;
600 rt->rt_ifa = ifa;
601 rt->rt_ifp = ifa->ifa_ifp;
602 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
603
604 rn = rnh->rnh_addaddr((caddr_t)ndst, (caddr_t)netmask,
605 rnh, rt->rt_nodes);
606 if (rn == 0) {
607 struct rtentry *rt2;
608 /*
609 * Uh-oh, we already have one of these in the tree.
610 * We do a special hack: if the route that's already
611 * there was generated by the protocol-cloning
612 * mechanism, then we just blow it away and retry
613 * the insertion of the new one.
614 */
615 rt2 = rtalloc1(dst, 0, RTF_PRCLONING);
616 if (rt2 && rt2->rt_parent) {
617 rtrequest(RTM_DELETE,
618 (struct sockaddr *)rt_key(rt2),
619 rt2->rt_gateway,
620 rt_mask(rt2), rt2->rt_flags, 0);
621 RTFREE(rt2);
622 rn = rnh->rnh_addaddr((caddr_t)ndst,
623 (caddr_t)netmask,
624 rnh, rt->rt_nodes);
625 } else if (rt2) {
626 /* undo the extra ref we got */
627 RTFREE(rt2);
628 }
629 }
630
631 /*
632 * If it still failed to go into the tree,
633 * then un-make it (this should be a function)
634 */
635 if (rn == 0) {
636 if (rt->rt_gwroute)
637 rtfree(rt->rt_gwroute);
638 if (rt->rt_ifa) {
639 IFAFREE(rt->rt_ifa);
640 }
641 Free(rt_key(rt));
642 Free(rt);
643 senderr(EEXIST);
644 }
645
646 rt->rt_parent = 0;
647
648 /*
649 * If we got here from RESOLVE, then we are cloning
650 * so clone the rest, and note that we
651 * are a clone (and increment the parent's references)
652 */
653 if (req == RTM_RESOLVE) {
654 rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
655 if ((*ret_nrt)->rt_flags & (RTF_CLONING | RTF_PRCLONING)) {
656 rt->rt_parent = (*ret_nrt);
657 (*ret_nrt)->rt_refcnt++;
658 }
659 }
660
661 /*
662 * if this protocol has something to add to this then
663 * allow it to do that as well.
664 */
665 if (ifa->ifa_rtrequest)
666 ifa->ifa_rtrequest(req, rt, SA(ret_nrt ? *ret_nrt : 0));
667
668 /*
669 * We repeat the same procedure from rt_setgate() here because
670 * it doesn't fire when we call it there because the node
671 * hasn't been added to the tree yet.
672 */
673 if (!(rt->rt_flags & RTF_HOST) && rt_mask(rt) != 0) {
674 struct rtfc_arg arg;
675 arg.rnh = rnh;
676 arg.rt0 = rt;
677 rnh->rnh_walktree_from(rnh, rt_key(rt), rt_mask(rt),
678 rt_fixchange, &arg);
679 }
680
681 /*
682 * actually return a resultant rtentry and
683 * give the caller a single reference.
684 */
685 if (ret_nrt) {
686 *ret_nrt = rt;
687 rt->rt_refcnt++;
688 }
689 break;
690 }
691bad:
692 splx(s);
693 return (error);
694}
695
696/*
697 * Called from rtrequest(RTM_DELETE, ...) to fix up the route's ``family''
698 * (i.e., the routes related to it by the operation of cloning). This
699 * routine is iterated over all potential former-child-routes by way of
700 * rnh->rnh_walktree_from() above, and those that actually are children of
701 * the late parent (passed in as VP here) are themselves deleted.
702 */
703static int
704rt_fixdelete(rn, vp)
705 struct radix_node *rn;
706 void *vp;
707{
708 struct rtentry *rt = (struct rtentry *)rn;
709 struct rtentry *rt0 = vp;
710
711 if (rt->rt_parent == rt0 && !(rt->rt_flags & RTF_PINNED)) {
712 return rtrequest(RTM_DELETE, rt_key(rt),
713 (struct sockaddr *)0, rt_mask(rt),
714 rt->rt_flags, (struct rtentry **)0);
715 }
716 return 0;
717}
718
719/*
720 * This routine is called from rt_setgate() to do the analogous thing for
721 * adds and changes. There is the added complication in this case of a
722 * middle insert; i.e., insertion of a new network route between an older
723 * network route and (cloned) host routes. For this reason, a simple check
724 * of rt->rt_parent is insufficient; each candidate route must be tested
725 * against the (mask, value) of the new route (passed as before in vp)
726 * to see if the new route matches it.
727 *
728 * XXX - it may be possible to do fixdelete() for changes and reserve this
729 * routine just for adds. I'm not sure why I thought it was necessary to do
730 * changes this way.
731 */
732#ifdef DEBUG
733static int rtfcdebug = 0;
734#endif
735
736static int
737rt_fixchange(rn, vp)
738 struct radix_node *rn;
739 void *vp;
740{
741 struct rtentry *rt = (struct rtentry *)rn;
742 struct rtfc_arg *ap = vp;
743 struct rtentry *rt0 = ap->rt0;
744 struct radix_node_head *rnh = ap->rnh;
745 u_char *xk1, *xm1, *xk2, *xmp;
746 int i, len, mlen;
747
748#ifdef DEBUG
749 if (rtfcdebug)
750 printf("rt_fixchange: rt %p, rt0 %p\n", rt, rt0);
751#endif
752
753 if (!rt->rt_parent || (rt->rt_flags & RTF_PINNED)) {
754#ifdef DEBUG
755 if(rtfcdebug) printf("no parent or pinned\n");
756#endif
757 return 0;
758 }
759
760 if (rt->rt_parent == rt0) {
761#ifdef DEBUG
762 if(rtfcdebug) printf("parent match\n");
763#endif
764 return rtrequest(RTM_DELETE, rt_key(rt),
765 (struct sockaddr *)0, rt_mask(rt),
766 rt->rt_flags, (struct rtentry **)0);
767 }
768
769 /*
770 * There probably is a function somewhere which does this...
771 * if not, there should be.
772 */
773 len = imin(((struct sockaddr *)rt_key(rt0))->sa_len,
774 ((struct sockaddr *)rt_key(rt))->sa_len);
775
776 xk1 = (u_char *)rt_key(rt0);
777 xm1 = (u_char *)rt_mask(rt0);
778 xk2 = (u_char *)rt_key(rt);
779
780 /* avoid applying a less specific route */
781 xmp = (u_char *)rt_mask(rt->rt_parent);
782 mlen = ((struct sockaddr *)rt_key(rt->rt_parent))->sa_len;
783 if (mlen > ((struct sockaddr *)rt_key(rt0))->sa_len) {
784#ifdef DEBUG
785 if (rtfcdebug)
786 printf("rt_fixchange: inserting a less "
787 "specific route\n");
788#endif
789 return 0;
790 }
791 for (i = rnh->rnh_treetop->rn_offset; i < mlen; i++) {
792 if ((xmp[i] & ~(xmp[i] ^ xm1[i])) != xmp[i]) {
793#ifdef DEBUG
794 if (rtfcdebug)
795 printf("rt_fixchange: inserting a less "
796 "specific route\n");
797#endif
798 return 0;
799 }
800 }
801
802 for (i = rnh->rnh_treetop->rn_offset; i < len; i++) {
803 if ((xk2[i] & xm1[i]) != xk1[i]) {
804#ifdef DEBUG
805 if(rtfcdebug) printf("no match\n");
806#endif
807 return 0;
808 }
809 }
810
811 /*
812 * OK, this node is a clone, and matches the node currently being
813 * changed/added under the node's mask. So, get rid of it.
814 */
815#ifdef DEBUG
816 if(rtfcdebug) printf("deleting\n");
817#endif
818 return rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0,
819 rt_mask(rt), rt->rt_flags, (struct rtentry **)0);
820}
821
822int
823rt_setgate(rt0, dst, gate)
824 struct rtentry *rt0;
825 struct sockaddr *dst, *gate;
826{
827 caddr_t new, old;
828 int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len);
829 register struct rtentry *rt = rt0;
830 struct radix_node_head *rnh = rt_tables[dst->sa_family];
831
832 /*
833 * A host route with the destination equal to the gateway
834 * will interfere with keeping LLINFO in the routing
835 * table, so disallow it.
836 */
837 if (((rt0->rt_flags & (RTF_HOST|RTF_GATEWAY|RTF_LLINFO)) ==
838 (RTF_HOST|RTF_GATEWAY)) &&
839 (dst->sa_len == gate->sa_len) &&
840 (bcmp(dst, gate, dst->sa_len) == 0)) {
841 /*
842 * The route might already exist if this is an RTM_CHANGE
843 * or a routing redirect, so try to delete it.
844 */
845 if (rt_key(rt0))
846 rtrequest(RTM_DELETE, (struct sockaddr *)rt_key(rt0),
847 rt0->rt_gateway, rt_mask(rt0), rt0->rt_flags, 0);
848 return EADDRNOTAVAIL;
849 }
850
851 /*
852 * Both dst and gateway are stored in the same malloc'd chunk
853 * (If I ever get my hands on....)
854 * if we need to malloc a new chunk, then keep the old one around
855 * till we don't need it any more.
856 */
857 if (rt->rt_gateway == 0 || glen > ROUNDUP(rt->rt_gateway->sa_len)) {
858 old = (caddr_t)rt_key(rt);
859 R_Malloc(new, caddr_t, dlen + glen);
860 if (new == 0)
861 return ENOBUFS;
862 rt->rt_nodes->rn_key = new;
863 } else {
864 /*
865 * otherwise just overwrite the old one
866 */
867 new = rt->rt_nodes->rn_key;
868 old = 0;
869 }
870
871 /*
872 * copy the new gateway value into the memory chunk
873 */
874 Bcopy(gate, (rt->rt_gateway = (struct sockaddr *)(new + dlen)), glen);
875
876 /*
877 * if we are replacing the chunk (or it's new) we need to
878 * replace the dst as well
879 */
880 if (old) {
881 Bcopy(dst, new, dlen);
882 Free(old);
883 }
884
885 /*
886 * If there is already a gwroute, it's now almost definitly wrong
887 * so drop it.
888 */
889 if (rt->rt_gwroute) {
890 rt = rt->rt_gwroute; RTFREE(rt);
891 rt = rt0; rt->rt_gwroute = 0;
892 }
893 /*
894 * Cloning loop avoidance:
895 * In the presence of protocol-cloning and bad configuration,
896 * it is possible to get stuck in bottomless mutual recursion
897 * (rtrequest rt_setgate rtalloc1). We avoid this by not allowing
898 * protocol-cloning to operate for gateways (which is probably the
899 * correct choice anyway), and avoid the resulting reference loops
900 * by disallowing any route to run through itself as a gateway.
901 * This is obviously mandatory when we get rt->rt_output().
902 */
903 if (rt->rt_flags & RTF_GATEWAY) {
904 rt->rt_gwroute = rtalloc1(gate, 1, RTF_PRCLONING);
905 if (rt->rt_gwroute == rt) {
906 RTFREE(rt->rt_gwroute);
907 rt->rt_gwroute = 0;
908 return EDQUOT; /* failure */
909 }
910 }
911
912 /*
913 * This isn't going to do anything useful for host routes, so
914 * don't bother. Also make sure we have a reasonable mask
915 * (we don't yet have one during adds).
916 */
917 if (!(rt->rt_flags & RTF_HOST) && rt_mask(rt) != 0) {
918 struct rtfc_arg arg;
919 arg.rnh = rnh;
920 arg.rt0 = rt;
921 rnh->rnh_walktree_from(rnh, rt_key(rt), rt_mask(rt),
922 rt_fixchange, &arg);
923 }
924
925 return 0;
926}
927
928static void
929rt_maskedcopy(src, dst, netmask)
930 struct sockaddr *src, *dst, *netmask;
931{
932 register u_char *cp1 = (u_char *)src;
933 register u_char *cp2 = (u_char *)dst;
934 register u_char *cp3 = (u_char *)netmask;
935 u_char *cplim = cp2 + *cp3;
936 u_char *cplim2 = cp2 + *cp1;
937
938 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
939 cp3 += 2;
940 if (cplim > cplim2)
941 cplim = cplim2;
942 while (cp2 < cplim)
943 *cp2++ = *cp1++ & *cp3++;
944 if (cp2 < cplim2)
945 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
946}
947
948/*
949 * Set up a routing table entry, normally
950 * for an interface.
951 */
952int
953rtinit(ifa, cmd, flags)
954 register struct ifaddr *ifa;
955 int cmd, flags;
956{
957 register struct rtentry *rt;
958 register struct sockaddr *dst;
959 register struct sockaddr *deldst;
960 struct mbuf *m = 0;
961 struct rtentry *nrt = 0;
962 int error;
963
964 dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
965 /*
966 * If it's a delete, check that if it exists, it's on the correct
967 * interface or we might scrub a route to another ifa which would
968 * be confusing at best and possibly worse.
969 */
970 if (cmd == RTM_DELETE) {
971 /*
972 * It's a delete, so it should already exist..
973 * If it's a net, mask off the host bits
974 * (Assuming we have a mask)
975 */
976 if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
977 m = m_get(M_DONTWAIT, MT_SONAME);
978 if (m == NULL)
979 return(ENOBUFS);
980 deldst = mtod(m, struct sockaddr *);
981 rt_maskedcopy(dst, deldst, ifa->ifa_netmask);
982 dst = deldst;
983 }
984 /*
985 * Get an rtentry that is in the routing tree and
986 * contains the correct info. (if this fails, can't get there).
987 * We set "report" to FALSE so that if it doesn't exist,
988 * it doesn't report an error or clone a route, etc. etc.
989 */
990 rt = rtalloc1(dst, 0, 0UL);
991 if (rt) {
992 /*
993 * Ok so we found the rtentry. it has an extra reference
994 * for us at this stage. we won't need that so
995 * lop that off now.
996 */
997 rt->rt_refcnt--;
998 if (rt->rt_ifa != ifa) {
999 /*
1000 * If the interface in the rtentry doesn't match
1001 * the interface we are using, then we don't
1002 * want to delete it, so return an error.
1003 * This seems to be the only point of
1004 * this whole RTM_DELETE clause.
1005 */
1006 if (m)
1007 (void) m_free(m);
1008 return (flags & RTF_HOST ? EHOSTUNREACH
1009 : ENETUNREACH);
1010 }
1011 }
1012 /* XXX */
1013#if 0
1014 else {
1015 /*
1016 * One would think that as we are deleting, and we know
1017 * it doesn't exist, we could just return at this point
1018 * with an "ELSE" clause, but apparently not..
1019 */
1020 return (flags & RTF_HOST ? EHOSTUNREACH
1021 : ENETUNREACH);
1022 }
1023#endif
1024 }
1025 /*
1026 * Do the actual request
1027 */
1028 error = rtrequest(cmd, dst, ifa->ifa_addr, ifa->ifa_netmask,
1029 flags | ifa->ifa_flags, &nrt);
1030 if (m)
1031 (void) m_free(m);
1032 /*
1033 * If we are deleting, and we found an entry, then
1034 * it's been removed from the tree.. now throw it away.
1035 */
1036 if (cmd == RTM_DELETE && error == 0 && (rt = nrt)) {
1037 /*
1038 * notify any listenning routing agents of the change
1039 */
1040 rt_newaddrmsg(cmd, ifa, error, nrt);
1041 if (rt->rt_refcnt <= 0) {
1042 rt->rt_refcnt++; /* need a 1->0 transition to free */
1043 rtfree(rt);
1044 }
1045 }
1046
1047 /*
1048 * We are adding, and we have a returned routing entry.
1049 * We need to sanity check the result.
1050 */
1051 if (cmd == RTM_ADD && error == 0 && (rt = nrt)) {
1052 /*
1053 * We just wanted to add it.. we don't actually need a reference
1054 */
1055 rt->rt_refcnt--;
1056 /*
1057 * If it came back with an unexpected interface, then it must
1058 * have already existed or something. (XXX)
1059 */
1060 if (rt->rt_ifa != ifa) {
1061 if (!(rt->rt_ifa->ifa_ifp->if_flags &
1062 (IFF_POINTOPOINT|IFF_LOOPBACK)))
1063 printf("rtinit: wrong ifa (%p) was (%p)\n",
1064 ifa, rt->rt_ifa);
1065 /*
1066 * Ask that the protocol in question
1067 * remove anything it has associated with
1068 * this route and ifaddr.
1069 */
1070 if (rt->rt_ifa->ifa_rtrequest)
1071 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, SA(0));
1072 /*
1073 * Remove the reference to its ifaddr.
1074 */
1075 IFAFREE(rt->rt_ifa);
1076 /*
1077 * And substitute in references to the ifaddr
1078 * we are adding.
1079 */
1080 rt->rt_ifa = ifa;
1081 rt->rt_ifp = ifa->ifa_ifp;
1082 rt->rt_rmx.rmx_mtu = ifa->ifa_ifp->if_mtu; /*XXX*/
1083 ifa->ifa_refcnt++;
1084 /*
1085 * Now ask the protocol to check if it needs
1086 * any special processing in its new form.
1087 */
1088 if (ifa->ifa_rtrequest)
1089 ifa->ifa_rtrequest(RTM_ADD, rt, SA(0));
1090 }
1091 /*
1092 * notify any listenning routing agents of the change
1093 */
1094 rt_newaddrmsg(cmd, ifa, error, nrt);
1095 }
1096 return (error);
1097}
1098
1099/* This must be before ip6_init2(), which is now SI_ORDER_MIDDLE */
1100SYSINIT(route, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0);