2 * Copyright (c) 1980, 1986, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)route.c 8.3.1.1 (Berkeley) 2/23/95
34 #include "opt_mrouting.h"
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/malloc.h>
40 #include <sys/socket.h>
41 #include <sys/domain.h>
42 #include <sys/kernel.h>
45 #include <net/route.h>
47 #include <netinet/in.h>
48 #include <netinet/ip_mroute.h>
52 static struct rtstat rtstat;
53 struct radix_node_head *rt_tables[AF_MAX+1];
55 static int rttrash; /* routes not in table but not freed */
57 static void rt_maskedcopy(struct sockaddr *,
58 struct sockaddr *, struct sockaddr *);
59 static void rtable_init(void **);
61 /* compare two sockaddr structures */
62 #define sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
65 * Convert a 'struct radix_node *' to a 'struct rtentry *'.
66 * The operation can be done safely (in this code) because a
67 * 'struct rtentry' starts with two 'struct radix_node''s, the first
68 * one representing leaf nodes in the routing tree, which is
69 * what the code in radix.c passes us as a 'struct radix_node'.
71 * But because there are a lot of assumptions in this conversion,
72 * do not cast explicitly, but always use the macro below.
74 #define RNTORT(p) ((struct rtentry *)(p))
77 rtable_init(void **table)
80 for (dom = domains; dom; dom = dom->dom_next)
81 if (dom->dom_rtattach)
82 dom->dom_rtattach(&table[dom->dom_family],
86 static uma_zone_t rtzone; /* Routing table UMA zone. */
91 rtzone = uma_zcreate("rtentry", sizeof(struct rtentry), NULL, NULL,
92 NULL, NULL, UMA_ALIGN_PTR, 0);
93 rn_init(); /* initialize all zeroes, all ones, mask table */
94 rtable_init((void **)rt_tables);
98 * Packet routing routines.
101 rtalloc(struct route *ro)
103 rtalloc_ign(ro, 0UL);
107 rtalloc_ign(struct route *ro, u_long ignore)
111 if ((rt = ro->ro_rt) != NULL) {
112 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
117 ro->ro_rt = rtalloc1(&ro->ro_dst, 1, ignore);
119 RT_UNLOCK(ro->ro_rt);
123 * Look up the route that matches the address given
124 * Or, at least try.. Create a cloned route if needed.
126 * The returned route, if any, is locked.
129 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
131 struct radix_node_head *rnh = rt_tables[dst->sa_family];
133 struct radix_node *rn;
134 struct rtentry *newrt;
135 struct rt_addrinfo info;
137 int err = 0, msgtype = RTM_MISS;
141 * Look up the address in the table for that Address Family
144 rtstat.rts_unreach++;
147 RADIX_NODE_HEAD_LOCK(rnh);
148 if ((rn = rnh->rnh_matchaddr(dst, rnh)) &&
149 (rn->rn_flags & RNF_ROOT) == 0) {
151 * If we find it and it's not the root node, then
152 * get a reference on the rtentry associated.
154 newrt = rt = RNTORT(rn);
155 nflags = rt->rt_flags & ~ignflags;
156 if (report && (nflags & RTF_CLONING)) {
158 * We are apparently adding (report = 0 in delete).
159 * If it requires that it be cloned, do so.
160 * (This implies it wasn't a HOST route.)
162 err = rtrequest(RTM_RESOLVE, dst, NULL,
166 * If the cloning didn't succeed, maybe
167 * what we have will do. Return that.
169 newrt = rt; /* existing route */
174 KASSERT(newrt, ("no route and no error"));
176 if (newrt->rt_flags & RTF_XRESOLVE) {
178 * If the new route specifies it be
179 * externally resolved, then go do that.
181 msgtype = RTM_RESOLVE;
184 /* Inform listeners of the new route. */
185 bzero(&info, sizeof(info));
186 info.rti_info[RTAX_DST] = rt_key(newrt);
187 info.rti_info[RTAX_NETMASK] = rt_mask(newrt);
188 info.rti_info[RTAX_GATEWAY] = newrt->rt_gateway;
189 if (newrt->rt_ifp != NULL) {
190 info.rti_info[RTAX_IFP] =
191 newrt->rt_ifp->if_addr->ifa_addr;
192 info.rti_info[RTAX_IFA] = newrt->rt_ifa->ifa_addr;
194 rt_missmsg(RTM_ADD, &info, newrt->rt_flags, 0);
199 RADIX_NODE_HEAD_UNLOCK(rnh);
202 * Either we hit the root or couldn't find any match,
203 * Which basically means
204 * "caint get there frm here"
206 rtstat.rts_unreach++;
208 RADIX_NODE_HEAD_UNLOCK(rnh);
211 * If required, report the failure to the supervising
213 * For a delete, this is not an error. (report == 0)
215 bzero(&info, sizeof(info));
216 info.rti_info[RTAX_DST] = dst;
217 rt_missmsg(msgtype, &info, 0, err);
221 RT_LOCK_ASSERT(newrt);
226 * Remove a reference count from an rtentry.
227 * If the count gets low enough, take it out of the routing table
230 rtfree(struct rtentry *rt)
232 struct radix_node_head *rnh;
234 KASSERT(rt != NULL,("%s: NULL rt", __func__));
235 rnh = rt_tables[rt_key(rt)->sa_family];
236 KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
241 * The callers should use RTFREE_LOCKED() or RTFREE(), so
242 * we should come here exactly with the last reference.
245 if (rt->rt_refcnt > 0) {
246 printf("%s: %p has %lu refs\n", __func__, rt, rt->rt_refcnt);
251 * On last reference give the "close method" a chance
252 * to cleanup private state. This also permits (for
253 * IPv4 and IPv6) a chance to decide if the routing table
254 * entry should be purged immediately or at a later time.
255 * When an immediate purge is to happen the close routine
256 * typically calls rtexpunge which clears the RTF_UP flag
257 * on the entry so that the code below reclaims the storage.
259 if (rt->rt_refcnt == 0 && rnh->rnh_close)
260 rnh->rnh_close((struct radix_node *)rt, rnh);
263 * If we are no longer "up" (and ref == 0)
264 * then we can free the resources associated
267 if ((rt->rt_flags & RTF_UP) == 0) {
268 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
271 * the rtentry must have been removed from the routing table
272 * so it is represented in rttrash.. remove that now.
276 if (rt->rt_refcnt < 0) {
277 printf("rtfree: %p not freed (neg refs)\n", rt);
282 * release references on items we hold them on..
283 * e.g other routes and ifaddrs.
287 rt->rt_parent = NULL; /* NB: no refcnt on parent */
290 * The key is separatly alloc'd so free it (see rt_setgate()).
291 * This also frees the gateway, as they are always malloc'd
297 * and the rtentry itself of course
300 uma_zfree(rtzone, rt);
309 * Force a routing table entry to the specified
310 * destination to go through the given gateway.
311 * Normally called as a result of a routing redirect
312 * message from the network layer.
315 rtredirect(struct sockaddr *dst,
316 struct sockaddr *gateway,
317 struct sockaddr *netmask,
319 struct sockaddr *src)
321 struct rtentry *rt, *rt0 = NULL;
324 struct rt_addrinfo info;
327 /* verify the gateway is directly reachable */
328 if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
332 rt = rtalloc1(dst, 0, 0UL); /* NB: rt is locked */
334 * If the redirect isn't from our current router for this dst,
335 * it's either old or wrong. If it redirects us to ourselves,
336 * we have a routing loop, perhaps as a result of an interface
337 * going down recently.
339 if (!(flags & RTF_DONE) && rt &&
340 (!sa_equal(src, rt->rt_gateway) || rt->rt_ifa != ifa))
342 else if (ifa_ifwithaddr(gateway))
343 error = EHOSTUNREACH;
347 * Create a new entry if we just got back a wildcard entry
348 * or the the lookup failed. This is necessary for hosts
349 * which use routing redirects generated by smart gateways
350 * to dynamically build the routing tables.
352 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
355 * Don't listen to the redirect if it's
356 * for a route to an interface.
358 if (rt->rt_flags & RTF_GATEWAY) {
359 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
361 * Changing from route to net => route to host.
362 * Create new route, rather than smashing route to net.
368 flags |= RTF_GATEWAY | RTF_DYNAMIC;
369 bzero((caddr_t)&info, sizeof(info));
370 info.rti_info[RTAX_DST] = dst;
371 info.rti_info[RTAX_GATEWAY] = gateway;
372 info.rti_info[RTAX_NETMASK] = netmask;
374 info.rti_flags = flags;
375 error = rtrequest1(RTM_ADD, &info, &rt);
378 EVENTHANDLER_INVOKE(route_redirect_event, rt0, rt, dst);
379 flags = rt->rt_flags;
384 stat = &rtstat.rts_dynamic;
386 struct rtentry *gwrt;
389 * Smash the current notion of the gateway to
390 * this destination. Should check about netmask!!!
392 rt->rt_flags |= RTF_MODIFIED;
393 flags |= RTF_MODIFIED;
394 stat = &rtstat.rts_newgateway;
396 * add the key and gateway (in one malloc'd chunk).
398 rt_setgate(rt, rt_key(rt), gateway);
399 gwrt = rtalloc1(gateway, 1, 0);
400 EVENTHANDLER_INVOKE(route_redirect_event, rt, gwrt, dst);
404 error = EHOSTUNREACH;
410 rtstat.rts_badredirect++;
411 else if (stat != NULL)
413 bzero((caddr_t)&info, sizeof(info));
414 info.rti_info[RTAX_DST] = dst;
415 info.rti_info[RTAX_GATEWAY] = gateway;
416 info.rti_info[RTAX_NETMASK] = netmask;
417 info.rti_info[RTAX_AUTHOR] = src;
418 rt_missmsg(RTM_REDIRECT, &info, flags, error);
422 * Routing table ioctl interface.
425 rtioctl(u_long req, caddr_t data)
429 * If more ioctl commands are added here, make sure the proper
430 * super-user checks are being performed because it is possible for
431 * prison-root to make it this far if raw sockets have been enabled
435 /* Multicast goop, grrr... */
436 return mrt_ioctl ? mrt_ioctl(req, data) : EOPNOTSUPP;
443 ifa_ifwithroute(int flags, struct sockaddr *dst, struct sockaddr *gateway)
445 register struct ifaddr *ifa;
448 if ((flags & RTF_GATEWAY) == 0) {
450 * If we are adding a route to an interface,
451 * and the interface is a pt to pt link
452 * we should search for the destination
453 * as our clue to the interface. Otherwise
454 * we can use the local address.
457 if (flags & RTF_HOST)
458 ifa = ifa_ifwithdstaddr(dst);
460 ifa = ifa_ifwithaddr(gateway);
463 * If we are adding a route to a remote net
464 * or host, the gateway may still be on the
465 * other end of a pt to pt link.
467 ifa = ifa_ifwithdstaddr(gateway);
470 ifa = ifa_ifwithnet(gateway);
472 struct rtentry *rt = rtalloc1(gateway, 0, 0UL);
476 * dismiss a gateway that is reachable only
477 * through the default router
479 switch (gateway->sa_family) {
481 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
485 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
495 if ((ifa = rt->rt_ifa) == NULL)
498 if (ifa->ifa_addr->sa_family != dst->sa_family) {
499 struct ifaddr *oifa = ifa;
500 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
507 static walktree_f_t rt_fixdelete;
508 static walktree_f_t rt_fixchange;
512 struct radix_node_head *rnh;
516 * Do appropriate manipulations of a routing tree given
517 * all the bits of info needed
521 struct sockaddr *dst,
522 struct sockaddr *gateway,
523 struct sockaddr *netmask,
525 struct rtentry **ret_nrt)
527 struct rt_addrinfo info;
529 if (dst->sa_len == 0)
532 bzero((caddr_t)&info, sizeof(info));
533 info.rti_flags = flags;
534 info.rti_info[RTAX_DST] = dst;
535 info.rti_info[RTAX_GATEWAY] = gateway;
536 info.rti_info[RTAX_NETMASK] = netmask;
537 return rtrequest1(req, &info, ret_nrt);
541 * These (questionable) definitions of apparent local variables apply
542 * to the next two functions. XXXXXX!!!
544 #define dst info->rti_info[RTAX_DST]
545 #define gateway info->rti_info[RTAX_GATEWAY]
546 #define netmask info->rti_info[RTAX_NETMASK]
547 #define ifaaddr info->rti_info[RTAX_IFA]
548 #define ifpaddr info->rti_info[RTAX_IFP]
549 #define flags info->rti_flags
552 rt_getifa(struct rt_addrinfo *info)
558 * ifp may be specified by sockaddr_dl
559 * when protocol address is ambiguous.
561 if (info->rti_ifp == NULL && ifpaddr != NULL &&
562 ifpaddr->sa_family == AF_LINK &&
563 (ifa = ifa_ifwithnet(ifpaddr)) != NULL)
564 info->rti_ifp = ifa->ifa_ifp;
565 if (info->rti_ifa == NULL && ifaaddr != NULL)
566 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
567 if (info->rti_ifa == NULL) {
570 sa = ifaaddr != NULL ? ifaaddr :
571 (gateway != NULL ? gateway : dst);
572 if (sa != NULL && info->rti_ifp != NULL)
573 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
574 else if (dst != NULL && gateway != NULL)
575 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
577 info->rti_ifa = ifa_ifwithroute(flags, sa, sa);
579 if ((ifa = info->rti_ifa) != NULL) {
580 if (info->rti_ifp == NULL)
581 info->rti_ifp = ifa->ifa_ifp;
588 * Expunges references to a route that's about to be reclaimed.
589 * The route must be locked.
592 rtexpunge(struct rtentry *rt)
594 struct radix_node *rn;
595 struct radix_node_head *rnh;
602 * We cannot assume anything about the reference count
603 * because protocols call us in many situations; often
604 * before unwinding references to the table entry.
606 KASSERT(rt->rt_refcnt <= 1, ("bogus refcnt %ld", rt->rt_refcnt));
609 * Find the correct routing tree to use for this Address Family
611 rnh = rt_tables[rt_key(rt)->sa_family];
613 return (EAFNOSUPPORT);
615 RADIX_NODE_HEAD_LOCK(rnh);
618 * Remove the item from the tree; it should be there,
619 * but when callers invoke us blindly it may not (sigh).
621 rn = rnh->rnh_deladdr(rt_key(rt), rt_mask(rt), rnh);
626 KASSERT((rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)) == 0,
627 ("unexpected flags 0x%x", rn->rn_flags));
628 KASSERT(rt == RNTORT(rn),
629 ("lookup mismatch, rt %p rn %p", rt, rn));
631 rt->rt_flags &= ~RTF_UP;
634 * Now search what's left of the subtree for any cloned
635 * routes which might have been formed from this node.
637 if ((rt->rt_flags & RTF_CLONING) && rt_mask(rt))
638 rnh->rnh_walktree_from(rnh, rt_key(rt), rt_mask(rt),
642 * Remove any external references we may have.
643 * This might result in another rtentry being freed if
644 * we held its last reference.
646 if (rt->rt_gwroute) {
647 RTFREE(rt->rt_gwroute);
648 rt->rt_gwroute = NULL;
652 * Give the protocol a chance to keep things in sync.
654 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) {
655 struct rt_addrinfo info;
657 bzero((caddr_t)&info, sizeof(info));
658 info.rti_flags = rt->rt_flags;
659 info.rti_info[RTAX_DST] = rt_key(rt);
660 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
661 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
662 ifa->ifa_rtrequest(RTM_DELETE, rt, &info);
666 * one more rtentry floating around that is not
667 * linked to the routing table.
671 RADIX_NODE_HEAD_UNLOCK(rnh);
676 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
679 register struct rtentry *rt;
680 register struct radix_node *rn;
681 register struct radix_node_head *rnh;
683 struct sockaddr *ndst;
684 #define senderr(x) { error = x ; goto bad; }
687 * Find the correct routing tree to use for this Address Family
689 rnh = rt_tables[dst->sa_family];
691 return (EAFNOSUPPORT);
692 RADIX_NODE_HEAD_LOCK(rnh);
694 * If we are adding a host route then we don't want to put
695 * a netmask in the tree, nor do we want to clone it.
697 if (flags & RTF_HOST) {
699 flags &= ~RTF_CLONING;
704 * Remove the item from the tree and return it.
705 * Complain if it is not there and do no more processing.
707 rn = rnh->rnh_deladdr(dst, netmask, rnh);
710 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
711 panic ("rtrequest delete");
715 rt->rt_flags &= ~RTF_UP;
718 * Now search what's left of the subtree for any cloned
719 * routes which might have been formed from this node.
721 if ((rt->rt_flags & RTF_CLONING) &&
723 rnh->rnh_walktree_from(rnh, dst, rt_mask(rt),
728 * Remove any external references we may have.
729 * This might result in another rtentry being freed if
730 * we held its last reference.
732 if (rt->rt_gwroute) {
733 RTFREE(rt->rt_gwroute);
734 rt->rt_gwroute = NULL;
738 * give the protocol a chance to keep things in sync.
740 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
741 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
744 * One more rtentry floating around that is not
745 * linked to the routing table. rttrash will be decremented
746 * when RTFREE(rt) is eventually called.
751 * If the caller wants it, then it can have it,
752 * but it's up to it to free the rtentry as we won't be
763 if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
767 flags = rt->rt_flags &
768 ~(RTF_CLONING | RTF_STATIC);
769 flags |= RTF_WASCLONED;
770 gateway = rt->rt_gateway;
771 if ((netmask = rt->rt_genmask) == NULL)
776 if ((flags & RTF_GATEWAY) && !gateway)
778 if (dst && gateway && (dst->sa_family != gateway->sa_family) &&
779 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
782 if (info->rti_ifa == NULL && (error = rt_getifa(info)))
787 rt = uma_zalloc(rtzone, M_NOWAIT | M_ZERO);
791 rt->rt_flags = RTF_UP | flags;
793 * Add the gateway. Possibly re-malloc-ing the storage for it
794 * also add the rt_gwroute if possible.
797 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
799 uma_zfree(rtzone, rt);
804 * point to the (possibly newly malloc'd) dest address.
806 ndst = (struct sockaddr *)rt_key(rt);
809 * make sure it contains the value we want (masked if needed).
812 rt_maskedcopy(dst, ndst, netmask);
814 bcopy(dst, ndst, dst->sa_len);
817 * Note that we now have a reference to the ifa.
818 * This moved from below so that rnh->rnh_addaddr() can
819 * examine the ifa and ifa->ifa_ifp if it so desires.
823 rt->rt_ifp = ifa->ifa_ifp;
825 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
826 rn = rnh->rnh_addaddr(ndst, netmask, rnh, rt->rt_nodes);
830 * Uh-oh, we already have one of these in the tree.
831 * We do a special hack: if the route that's already
832 * there was generated by the cloning mechanism
833 * then we just blow it away and retry the insertion
836 rt2 = rtalloc1(dst, 0, 0);
837 if (rt2 && rt2->rt_parent) {
840 rn = rnh->rnh_addaddr(ndst, netmask,
843 /* undo the extra ref we got */
849 * If it still failed to go into the tree,
850 * then un-make it (this should be a function)
854 RTFREE(rt->rt_gwroute);
859 uma_zfree(rtzone, rt);
863 rt->rt_parent = NULL;
866 * If we got here from RESOLVE, then we are cloning
867 * so clone the rest, and note that we
868 * are a clone (and increment the parent's references)
870 if (req == RTM_RESOLVE) {
871 KASSERT(ret_nrt && *ret_nrt,
872 ("no route to clone from"));
873 rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
874 rt->rt_rmx.rmx_pksent = 0; /* reset packet counter */
875 if ((*ret_nrt)->rt_flags & RTF_CLONING) {
877 * NB: We do not bump the refcnt on the parent
878 * entry under the assumption that it will
879 * remain so long as we do. This is
880 * important when deleting the parent route
881 * as this operation requires traversing
882 * the tree to delete all clones and futzing
883 * with refcnts requires us to double-lock
884 * parent through this back reference.
886 rt->rt_parent = *ret_nrt;
891 * If this protocol has something to add to this then
892 * allow it to do that as well.
894 if (ifa->ifa_rtrequest)
895 ifa->ifa_rtrequest(req, rt, info);
898 * We repeat the same procedure from rt_setgate() here because
899 * it doesn't fire when we call it there because the node
900 * hasn't been added to the tree yet.
902 if (req == RTM_ADD &&
903 !(rt->rt_flags & RTF_HOST) && rt_mask(rt) != NULL) {
907 rnh->rnh_walktree_from(rnh, rt_key(rt), rt_mask(rt),
912 * actually return a resultant rtentry and
913 * give the caller a single reference.
925 RADIX_NODE_HEAD_UNLOCK(rnh);
938 * Called from rtrequest(RTM_DELETE, ...) to fix up the route's ``family''
939 * (i.e., the routes related to it by the operation of cloning). This
940 * routine is iterated over all potential former-child-routes by way of
941 * rnh->rnh_walktree_from() above, and those that actually are children of
942 * the late parent (passed in as VP here) are themselves deleted.
945 rt_fixdelete(struct radix_node *rn, void *vp)
947 struct rtentry *rt = RNTORT(rn);
948 struct rtentry *rt0 = vp;
950 if (rt->rt_parent == rt0 &&
951 !(rt->rt_flags & (RTF_PINNED | RTF_CLONING))) {
952 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt),
959 * This routine is called from rt_setgate() to do the analogous thing for
960 * adds and changes. There is the added complication in this case of a
961 * middle insert; i.e., insertion of a new network route between an older
962 * network route and (cloned) host routes. For this reason, a simple check
963 * of rt->rt_parent is insufficient; each candidate route must be tested
964 * against the (mask, value) of the new route (passed as before in vp)
965 * to see if the new route matches it.
967 * XXX - it may be possible to do fixdelete() for changes and reserve this
968 * routine just for adds. I'm not sure why I thought it was necessary to do
973 rt_fixchange(struct radix_node *rn, void *vp)
975 struct rtentry *rt = RNTORT(rn);
976 struct rtfc_arg *ap = vp;
977 struct rtentry *rt0 = ap->rt0;
978 struct radix_node_head *rnh = ap->rnh;
979 u_char *xk1, *xm1, *xk2, *xmp;
982 /* make sure we have a parent, and route is not pinned or cloning */
983 if (!rt->rt_parent ||
984 (rt->rt_flags & (RTF_PINNED | RTF_CLONING)))
987 if (rt->rt_parent == rt0) /* parent match */
990 * There probably is a function somewhere which does this...
991 * if not, there should be.
993 len = imin(rt_key(rt0)->sa_len, rt_key(rt)->sa_len);
995 xk1 = (u_char *)rt_key(rt0);
996 xm1 = (u_char *)rt_mask(rt0);
997 xk2 = (u_char *)rt_key(rt);
999 /* avoid applying a less specific route */
1000 xmp = (u_char *)rt_mask(rt->rt_parent);
1001 mlen = rt_key(rt->rt_parent)->sa_len;
1002 if (mlen > rt_key(rt0)->sa_len) /* less specific route */
1004 for (i = rnh->rnh_treetop->rn_offset; i < mlen; i++)
1005 if ((xmp[i] & ~(xmp[i] ^ xm1[i])) != xmp[i])
1006 return 0; /* less specific route */
1008 for (i = rnh->rnh_treetop->rn_offset; i < len; i++)
1009 if ((xk2[i] & xm1[i]) != xk1[i])
1010 return 0; /* no match */
1013 * OK, this node is a clone, and matches the node currently being
1014 * changed/added under the node's mask. So, get rid of it.
1017 return rtrequest(RTM_DELETE, rt_key(rt), NULL,
1018 rt_mask(rt), rt->rt_flags, NULL);
1022 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
1024 /* XXX dst may be overwritten, can we move this to below */
1025 struct radix_node_head *rnh = rt_tables[dst->sa_family];
1026 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
1032 * A host route with the destination equal to the gateway
1033 * will interfere with keeping LLINFO in the routing
1034 * table, so disallow it.
1036 if (((rt->rt_flags & (RTF_HOST|RTF_GATEWAY|RTF_LLINFO)) ==
1037 (RTF_HOST|RTF_GATEWAY)) &&
1038 dst->sa_len == gate->sa_len &&
1039 bcmp(dst, gate, dst->sa_len) == 0) {
1041 * The route might already exist if this is an RTM_CHANGE
1042 * or a routing redirect, so try to delete it.
1046 return EADDRNOTAVAIL;
1050 * Cloning loop avoidance in case of bad configuration.
1052 if (rt->rt_flags & RTF_GATEWAY) {
1053 struct rtentry *gwrt;
1055 RT_UNLOCK(rt); /* XXX workaround LOR */
1056 gwrt = rtalloc1(gate, 1, 0);
1059 return (EADDRINUSE); /* failure */
1062 * Try to reacquire the lock on rt, and if it fails,
1063 * clean state and restart from scratch.
1065 if (!RT_TRYLOCK(rt)) {
1066 RTFREE_LOCKED(gwrt);
1071 * If there is already a gwroute, then drop it. If we
1072 * are asked to replace route with itself, then do
1073 * not leak its refcounter.
1075 if (rt->rt_gwroute != NULL) {
1076 if (rt->rt_gwroute == gwrt) {
1077 RT_REMREF(rt->rt_gwroute);
1079 RTFREE(rt->rt_gwroute);
1082 if ((rt->rt_gwroute = gwrt) != NULL)
1083 RT_UNLOCK(rt->rt_gwroute);
1087 * Prepare to store the gateway in rt->rt_gateway.
1088 * Both dst and gateway are stored one after the other in the same
1089 * malloc'd chunk. If we have room, we can reuse the old buffer,
1090 * rt_gateway already points to the right place.
1091 * Otherwise, malloc a new block and update the 'dst' address.
1093 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
1096 R_Malloc(new, caddr_t, dlen + glen);
1100 * XXX note, we copy from *dst and not *rt_key(rt) because
1101 * rt_setgate() can be called to initialize a newly
1102 * allocated route entry, in which case rt_key(rt) == NULL
1103 * (and also rt->rt_gateway == NULL).
1104 * Free()/free() handle a NULL argument just fine.
1106 bcopy(dst, new, dlen);
1107 Free(rt_key(rt)); /* free old block, if any */
1108 rt_key(rt) = (struct sockaddr *)new;
1109 rt->rt_gateway = (struct sockaddr *)(new + dlen);
1113 * Copy the new gateway value into the memory chunk.
1115 bcopy(gate, rt->rt_gateway, glen);
1118 * This isn't going to do anything useful for host routes, so
1119 * don't bother. Also make sure we have a reasonable mask
1120 * (we don't yet have one during adds).
1122 if (!(rt->rt_flags & RTF_HOST) && rt_mask(rt) != 0) {
1123 struct rtfc_arg arg;
1127 RT_UNLOCK(rt); /* XXX workaround LOR */
1128 RADIX_NODE_HEAD_LOCK(rnh);
1130 rnh->rnh_walktree_from(rnh, rt_key(rt), rt_mask(rt),
1131 rt_fixchange, &arg);
1132 RADIX_NODE_HEAD_UNLOCK(rnh);
1139 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
1141 register u_char *cp1 = (u_char *)src;
1142 register u_char *cp2 = (u_char *)dst;
1143 register u_char *cp3 = (u_char *)netmask;
1144 u_char *cplim = cp2 + *cp3;
1145 u_char *cplim2 = cp2 + *cp1;
1147 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1152 *cp2++ = *cp1++ & *cp3++;
1154 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1158 * Set up a routing table entry, normally
1162 rtinit(struct ifaddr *ifa, int cmd, int flags)
1164 struct sockaddr *dst;
1165 struct sockaddr *netmask;
1166 struct mbuf *m = NULL;
1167 struct rtentry *rt = NULL;
1168 struct rt_addrinfo info;
1171 if (flags & RTF_HOST) {
1172 dst = ifa->ifa_dstaddr;
1175 dst = ifa->ifa_addr;
1176 netmask = ifa->ifa_netmask;
1178 if (dst->sa_len == 0)
1182 * If it's a delete, check that if it exists, it's on the correct
1183 * interface or we might scrub a route to another ifa which would
1184 * be confusing at best and possibly worse.
1186 if (cmd == RTM_DELETE) {
1187 struct sockaddr *deldst;
1188 struct radix_node_head *rnh;
1189 struct radix_node *rn;
1192 * It's a delete, so it should already exist..
1193 * If it's a net, mask off the host bits
1194 * (Assuming we have a mask)
1196 if (netmask != NULL) {
1197 m = m_get(M_DONTWAIT, MT_SONAME);
1200 deldst = mtod(m, struct sockaddr *);
1201 rt_maskedcopy(dst, deldst, netmask);
1205 * Look up an rtentry that is in the routing tree and
1206 * contains the correct info.
1208 if ((rnh = rt_tables[dst->sa_family]) == NULL)
1210 RADIX_NODE_HEAD_LOCK(rnh);
1211 error = ((rn = rnh->rnh_lookup(dst, netmask, rnh)) == NULL ||
1212 (rn->rn_flags & RNF_ROOT) ||
1213 RNTORT(rn)->rt_ifa != ifa ||
1214 !sa_equal((struct sockaddr *)rn->rn_key, dst));
1215 RADIX_NODE_HEAD_UNLOCK(rnh);
1220 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
1224 * Do the actual request
1226 bzero((caddr_t)&info, sizeof(info));
1228 info.rti_flags = flags | ifa->ifa_flags;
1229 info.rti_info[RTAX_DST] = dst;
1230 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
1231 info.rti_info[RTAX_NETMASK] = netmask;
1232 error = rtrequest1(cmd, &info, &rt);
1233 if (error == 0 && rt != NULL) {
1235 * notify any listening routing agents of the change
1238 rt_newaddrmsg(cmd, ifa, error, rt);
1239 if (cmd == RTM_DELETE) {
1241 * If we are deleting, and we found an entry, then
1242 * it's been removed from the tree.. now throw it away.
1246 if (cmd == RTM_ADD) {
1248 * We just wanted to add it.. we don't actually
1262 * rt_check() is invoked on each layer 2 output path, prior to
1263 * encapsulating outbound packets.
1265 * The function is mostly used to find a routing entry for the gateway,
1266 * which in some protocol families could also point to the link-level
1267 * address for the gateway itself (the side effect of revalidating the
1268 * route to the destination is rather pointless at this stage, we did it
1269 * already a moment before in the pr_output() routine to locate the ifp
1270 * and gateway to use).
1272 * When we remove the layer-3 to layer-2 mapping tables from the
1273 * routing table, this function can be removed.
1276 * *dst is the address of the NEXT HOP (which coincides with the
1277 * final destination if directly reachable);
1278 * *lrt0 points to the cached route to the final destination;
1279 * *lrt is not meaningful;
1282 * If the route is marked down try to find a new route. If the route
1283 * to the gateway is gone, try to setup a new route. Otherwise,
1284 * if the route is marked for packets to be rejected, enforce that.
1287 * *dst is unchanged;
1288 * *lrt0 points to the (possibly new) route to the final destination
1289 * *lrt points to the route to the next hop
1291 * Their values are meaningful ONLY if no error is returned.
1294 rt_check(struct rtentry **lrt, struct rtentry **lrt0, struct sockaddr *dst)
1297 struct rtentry *rt0;
1300 KASSERT(*lrt0 != NULL, ("rt_check"));
1303 /* NB: the locking here is tortuous... */
1305 if ((rt->rt_flags & RTF_UP) == 0) {
1307 rt = rtalloc1(dst, 1, 0UL);
1310 /* XXX what about if change? */
1312 return (EHOSTUNREACH);
1315 /* XXX BSD/OS checks dst->sa_family != AF_NS */
1316 if (rt->rt_flags & RTF_GATEWAY) {
1317 if (rt->rt_gwroute == NULL)
1319 rt = rt->rt_gwroute;
1320 RT_LOCK(rt); /* NB: gwroute */
1321 if ((rt->rt_flags & RTF_UP) == 0) {
1322 RTFREE_LOCKED(rt); /* unlock gwroute */
1324 rt0->rt_gwroute = NULL;
1327 rt = rtalloc1(rt->rt_gateway, 1, 0UL);
1331 return (ENETUNREACH);
1334 if (rt0->rt_gwroute != NULL)
1335 RTFREE(rt0->rt_gwroute);
1336 rt0->rt_gwroute = rt;
1339 return (EHOSTUNREACH);
1344 /* XXX why are we inspecting rmx_expire? */
1345 error = (rt->rt_flags & RTF_REJECT) &&
1346 (rt->rt_rmx.rmx_expire == 0 ||
1347 time_uptime < rt->rt_rmx.rmx_expire);
1350 return (rt == rt0 ? EHOSTDOWN : EHOSTUNREACH);
1358 /* This must be before ip6_init2(), which is now SI_ORDER_MIDDLE */
1359 SYSINIT(route, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0);