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
32 /************************************************************************
33 * Note: In this file a 'fib' is a "forwarding information base" *
34 * Which is the new name for an in kernel routing (next hop) table. *
35 ***********************************************************************/
38 #include "opt_route.h"
39 #include "opt_mrouting.h"
40 #include "opt_mpath.h"
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/syslog.h>
45 #include <sys/malloc.h>
47 #include <sys/socket.h>
48 #include <sys/sysctl.h>
49 #include <sys/syslog.h>
50 #include <sys/sysproto.h>
52 #include <sys/domain.h>
53 #include <sys/kernel.h>
56 #include <net/if_dl.h>
57 #include <net/route.h>
59 #include <net/flowtable.h>
62 #include <net/radix_mpath.h>
65 #include <netinet/in.h>
66 #include <netinet/ip_mroute.h>
70 u_int rt_numfibs = RT_NUMFIBS;
71 SYSCTL_UINT(_net, OID_AUTO, fibs, CTLFLAG_RD, &rt_numfibs, 0, "");
73 * Allow the boot code to allow LESS than RT_MAXFIBS to be used.
74 * We can't do more because storage is statically allocated for now.
75 * (for compatibility reasons.. this will change).
77 TUNABLE_INT("net.fibs", &rt_numfibs);
80 * By default add routes to all fibs for new interfaces.
81 * Once this is set to 0 then only allocate routes on interface
82 * changes for the FIB of the caller when adding a new set of addresses
83 * to an interface. XXX this is a shotgun aproach to a problem that needs
84 * a more fine grained solution.. that will come.
86 u_int rt_add_addr_allfibs = 1;
87 SYSCTL_UINT(_net, OID_AUTO, add_addr_allfibs, CTLFLAG_RW,
88 &rt_add_addr_allfibs, 0, "");
89 TUNABLE_INT("net.add_addr_allfibs", &rt_add_addr_allfibs);
91 VNET_DEFINE(struct rtstat, rtstat);
92 #define V_rtstat VNET(rtstat)
94 VNET_DEFINE(struct radix_node_head *, rt_tables);
95 #define V_rt_tables VNET(rt_tables)
97 VNET_DEFINE(int, rttrash); /* routes not in table but not freed */
98 #define V_rttrash VNET(rttrash)
101 /* compare two sockaddr structures */
102 #define sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
105 * Convert a 'struct radix_node *' to a 'struct rtentry *'.
106 * The operation can be done safely (in this code) because a
107 * 'struct rtentry' starts with two 'struct radix_node''s, the first
108 * one representing leaf nodes in the routing tree, which is
109 * what the code in radix.c passes us as a 'struct radix_node'.
111 * But because there are a lot of assumptions in this conversion,
112 * do not cast explicitly, but always use the macro below.
114 #define RNTORT(p) ((struct rtentry *)(p))
116 static VNET_DEFINE(uma_zone_t, rtzone); /* Routing table UMA zone. */
117 #define V_rtzone VNET(rtzone)
120 * handler for net.my_fibnum
123 sysctl_my_fibnum(SYSCTL_HANDLER_ARGS)
128 fibnum = curthread->td_proc->p_fibnum;
129 error = sysctl_handle_int(oidp, &fibnum, 0, req);
133 SYSCTL_PROC(_net, OID_AUTO, my_fibnum, CTLTYPE_INT|CTLFLAG_RD,
134 NULL, 0, &sysctl_my_fibnum, "I", "default FIB of caller");
136 static __inline struct radix_node_head **
137 rt_tables_get_rnh_ptr(int table, int fam)
139 struct radix_node_head **rnh;
141 KASSERT(table >= 0 && table < rt_numfibs, ("%s: table out of bounds.",
143 KASSERT(fam >= 0 && fam < (AF_MAX+1), ("%s: fam out of bounds.",
146 /* rnh is [fib=0][af=0]. */
147 rnh = (struct radix_node_head **)V_rt_tables;
148 /* Get the offset to the requested table and fam. */
149 rnh += table * (AF_MAX+1) + fam;
154 struct radix_node_head *
155 rt_tables_get_rnh(int table, int fam)
158 return (*rt_tables_get_rnh_ptr(table, fam));
162 * route initialization must occur before ip6_init2(), which happenas at
171 /* whack the tunable ints into line. */
172 if (rt_numfibs > RT_MAXFIBS)
173 rt_numfibs = RT_MAXFIBS;
177 for (dom = domains; dom; dom = dom->dom_next)
178 if (dom->dom_maxrtkey > max_keylen)
179 max_keylen = dom->dom_maxrtkey;
181 rn_init(max_keylen); /* init all zeroes, all ones, mask table */
183 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0);
186 vnet_route_init(const void *unused __unused)
189 struct radix_node_head **rnh;
193 V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) *
194 sizeof(struct radix_node_head *), M_RTABLE, M_WAITOK|M_ZERO);
196 V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry), NULL, NULL,
197 NULL, NULL, UMA_ALIGN_PTR, 0);
198 for (dom = domains; dom; dom = dom->dom_next) {
199 if (dom->dom_rtattach) {
200 for (table = 0; table < rt_numfibs; table++) {
201 if ( (fam = dom->dom_family) == AF_INET ||
203 /* for now only AF_INET has > 1 table */
205 * rtattach will be also called
206 * from vfs_export.c but the
208 * (only for AF_INET and AF_INET6
209 * which don't need it anyhow)
211 rnh = rt_tables_get_rnh_ptr(table, fam);
213 panic("%s: rnh NULL", __func__);
214 dom->dom_rtattach((void **)rnh,
223 VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
228 vnet_route_uninit(const void *unused __unused)
233 struct radix_node_head **rnh;
235 for (dom = domains; dom; dom = dom->dom_next) {
236 if (dom->dom_rtdetach) {
237 for (table = 0; table < rt_numfibs; table++) {
238 if ( (fam = dom->dom_family) == AF_INET ||
240 /* For now only AF_INET has > 1 tbl. */
241 rnh = rt_tables_get_rnh_ptr(table, fam);
243 panic("%s: rnh NULL", __func__);
244 dom->dom_rtdetach((void **)rnh,
253 VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
254 vnet_route_uninit, 0);
257 #ifndef _SYS_SYSPROTO_H_
263 sys_setfib(struct thread *td, struct setfib_args *uap)
265 if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs)
267 td->td_proc->p_fibnum = uap->fibnum;
272 * Packet routing routines.
275 rtalloc(struct route *ro)
277 rtalloc_ign_fib(ro, 0UL, 0);
281 rtalloc_fib(struct route *ro, u_int fibnum)
283 rtalloc_ign_fib(ro, 0UL, fibnum);
287 rtalloc_ign(struct route *ro, u_long ignore)
291 if ((rt = ro->ro_rt) != NULL) {
292 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
297 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, 0);
299 RT_UNLOCK(ro->ro_rt);
303 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum)
307 if ((rt = ro->ro_rt) != NULL) {
308 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
313 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum);
315 RT_UNLOCK(ro->ro_rt);
319 * Look up the route that matches the address given
320 * Or, at least try.. Create a cloned route if needed.
322 * The returned route, if any, is locked.
325 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
327 return (rtalloc1_fib(dst, report, ignflags, 0));
331 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags,
334 struct radix_node_head *rnh;
335 struct radix_node *rn;
336 struct rtentry *newrt;
337 struct rt_addrinfo info;
338 int err = 0, msgtype = RTM_MISS;
341 KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum"));
342 if (dst->sa_family != AF_INET) /* Only INET supports > 1 fib now */
344 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
350 * Look up the address in the table for that Address Family
352 needlock = !(ignflags & RTF_RNH_LOCKED);
354 RADIX_NODE_HEAD_RLOCK(rnh);
357 RADIX_NODE_HEAD_LOCK_ASSERT(rnh);
359 rn = rnh->rnh_matchaddr(dst, rnh);
360 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
365 RADIX_NODE_HEAD_RUNLOCK(rnh);
369 RADIX_NODE_HEAD_RUNLOCK(rnh);
372 * Either we hit the root or couldn't find any match,
373 * Which basically means
374 * "caint get there frm here"
377 V_rtstat.rts_unreach++;
381 * If required, report the failure to the supervising
383 * For a delete, this is not an error. (report == 0)
385 bzero(&info, sizeof(info));
386 info.rti_info[RTAX_DST] = dst;
387 rt_missmsg_fib(msgtype, &info, 0, err, fibnum);
391 RT_LOCK_ASSERT(newrt);
396 * Remove a reference count from an rtentry.
397 * If the count gets low enough, take it out of the routing table
400 rtfree(struct rtentry *rt)
402 struct radix_node_head *rnh;
404 KASSERT(rt != NULL,("%s: NULL rt", __func__));
405 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
406 KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
411 * The callers should use RTFREE_LOCKED() or RTFREE(), so
412 * we should come here exactly with the last reference.
415 if (rt->rt_refcnt > 0) {
416 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt);
421 * On last reference give the "close method" a chance
422 * to cleanup private state. This also permits (for
423 * IPv4 and IPv6) a chance to decide if the routing table
424 * entry should be purged immediately or at a later time.
425 * When an immediate purge is to happen the close routine
426 * typically calls rtexpunge which clears the RTF_UP flag
427 * on the entry so that the code below reclaims the storage.
429 if (rt->rt_refcnt == 0 && rnh->rnh_close)
430 rnh->rnh_close((struct radix_node *)rt, rnh);
433 * If we are no longer "up" (and ref == 0)
434 * then we can free the resources associated
437 if ((rt->rt_flags & RTF_UP) == 0) {
438 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
441 * the rtentry must have been removed from the routing table
442 * so it is represented in rttrash.. remove that now.
446 if (rt->rt_refcnt < 0) {
447 printf("rtfree: %p not freed (neg refs)\n", rt);
452 * release references on items we hold them on..
453 * e.g other routes and ifaddrs.
456 ifa_free(rt->rt_ifa);
458 * The key is separatly alloc'd so free it (see rt_setgate()).
459 * This also frees the gateway, as they are always malloc'd
465 * and the rtentry itself of course
468 uma_zfree(V_rtzone, rt);
477 * Force a routing table entry to the specified
478 * destination to go through the given gateway.
479 * Normally called as a result of a routing redirect
480 * message from the network layer.
483 rtredirect(struct sockaddr *dst,
484 struct sockaddr *gateway,
485 struct sockaddr *netmask,
487 struct sockaddr *src)
489 rtredirect_fib(dst, gateway, netmask, flags, src, 0);
493 rtredirect_fib(struct sockaddr *dst,
494 struct sockaddr *gateway,
495 struct sockaddr *netmask,
497 struct sockaddr *src,
500 struct rtentry *rt, *rt0 = NULL;
503 struct rt_addrinfo info;
505 struct radix_node_head *rnh;
508 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
510 error = EAFNOSUPPORT;
514 /* verify the gateway is directly reachable */
515 if ((ifa = ifa_ifwithnet(gateway, 0)) == NULL) {
519 rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */
521 * If the redirect isn't from our current router for this dst,
522 * it's either old or wrong. If it redirects us to ourselves,
523 * we have a routing loop, perhaps as a result of an interface
524 * going down recently.
526 if (!(flags & RTF_DONE) && rt &&
527 (!sa_equal(src, rt->rt_gateway) || rt->rt_ifa != ifa))
529 else if (ifa_ifwithaddr_check(gateway))
530 error = EHOSTUNREACH;
534 * Create a new entry if we just got back a wildcard entry
535 * or the lookup failed. This is necessary for hosts
536 * which use routing redirects generated by smart gateways
537 * to dynamically build the routing tables.
539 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
542 * Don't listen to the redirect if it's
543 * for a route to an interface.
545 if (rt->rt_flags & RTF_GATEWAY) {
546 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
548 * Changing from route to net => route to host.
549 * Create new route, rather than smashing route to net.
555 flags |= RTF_GATEWAY | RTF_DYNAMIC;
556 bzero((caddr_t)&info, sizeof(info));
557 info.rti_info[RTAX_DST] = dst;
558 info.rti_info[RTAX_GATEWAY] = gateway;
559 info.rti_info[RTAX_NETMASK] = netmask;
561 info.rti_flags = flags;
563 RT_UNLOCK(rt0); /* drop lock to avoid LOR with RNH */
564 error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum);
568 EVENTHANDLER_INVOKE(route_redirect_event, rt0, rt, dst);
569 flags = rt->rt_flags;
574 stat = &V_rtstat.rts_dynamic;
576 struct rtentry *gwrt;
579 * Smash the current notion of the gateway to
580 * this destination. Should check about netmask!!!
582 rt->rt_flags |= RTF_MODIFIED;
583 flags |= RTF_MODIFIED;
584 stat = &V_rtstat.rts_newgateway;
586 * add the key and gateway (in one malloc'd chunk).
589 RADIX_NODE_HEAD_LOCK(rnh);
591 rt_setgate(rt, rt_key(rt), gateway);
592 gwrt = rtalloc1(gateway, 1, RTF_RNH_LOCKED);
593 RADIX_NODE_HEAD_UNLOCK(rnh);
594 EVENTHANDLER_INVOKE(route_redirect_event, rt, gwrt, dst);
598 error = EHOSTUNREACH;
604 V_rtstat.rts_badredirect++;
605 else if (stat != NULL)
607 bzero((caddr_t)&info, sizeof(info));
608 info.rti_info[RTAX_DST] = dst;
609 info.rti_info[RTAX_GATEWAY] = gateway;
610 info.rti_info[RTAX_NETMASK] = netmask;
611 info.rti_info[RTAX_AUTHOR] = src;
612 rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum);
618 rtioctl(u_long req, caddr_t data)
620 return (rtioctl_fib(req, data, 0));
624 * Routing table ioctl interface.
627 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
631 * If more ioctl commands are added here, make sure the proper
632 * super-user checks are being performed because it is possible for
633 * prison-root to make it this far if raw sockets have been enabled
637 /* Multicast goop, grrr... */
638 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
645 * For both ifa_ifwithroute() routines, 'ifa' is returned referenced.
648 ifa_ifwithroute(int flags, struct sockaddr *dst, struct sockaddr *gateway)
650 return (ifa_ifwithroute_fib(flags, dst, gateway, 0));
654 ifa_ifwithroute_fib(int flags, struct sockaddr *dst, struct sockaddr *gateway,
657 register struct ifaddr *ifa;
660 if ((flags & RTF_GATEWAY) == 0) {
662 * If we are adding a route to an interface,
663 * and the interface is a pt to pt link
664 * we should search for the destination
665 * as our clue to the interface. Otherwise
666 * we can use the local address.
669 if (flags & RTF_HOST)
670 ifa = ifa_ifwithdstaddr(dst);
672 ifa = ifa_ifwithaddr(gateway);
675 * If we are adding a route to a remote net
676 * or host, the gateway may still be on the
677 * other end of a pt to pt link.
679 ifa = ifa_ifwithdstaddr(gateway);
682 ifa = ifa_ifwithnet(gateway, 0);
684 struct rtentry *rt = rtalloc1_fib(gateway, 0, RTF_RNH_LOCKED, fibnum);
688 * dismiss a gateway that is reachable only
689 * through the default router
691 switch (gateway->sa_family) {
693 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
697 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
703 if (!not_found && rt->rt_ifa != NULL) {
709 if (not_found || ifa == NULL)
712 if (ifa->ifa_addr->sa_family != dst->sa_family) {
713 struct ifaddr *oifa = ifa;
714 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
724 * Do appropriate manipulations of a routing tree given
725 * all the bits of info needed
729 struct sockaddr *dst,
730 struct sockaddr *gateway,
731 struct sockaddr *netmask,
733 struct rtentry **ret_nrt)
735 return (rtrequest_fib(req, dst, gateway, netmask, flags, ret_nrt, 0));
739 rtrequest_fib(int req,
740 struct sockaddr *dst,
741 struct sockaddr *gateway,
742 struct sockaddr *netmask,
744 struct rtentry **ret_nrt,
747 struct rt_addrinfo info;
749 if (dst->sa_len == 0)
752 bzero((caddr_t)&info, sizeof(info));
753 info.rti_flags = flags;
754 info.rti_info[RTAX_DST] = dst;
755 info.rti_info[RTAX_GATEWAY] = gateway;
756 info.rti_info[RTAX_NETMASK] = netmask;
757 return rtrequest1_fib(req, &info, ret_nrt, fibnum);
761 * These (questionable) definitions of apparent local variables apply
762 * to the next two functions. XXXXXX!!!
764 #define dst info->rti_info[RTAX_DST]
765 #define gateway info->rti_info[RTAX_GATEWAY]
766 #define netmask info->rti_info[RTAX_NETMASK]
767 #define ifaaddr info->rti_info[RTAX_IFA]
768 #define ifpaddr info->rti_info[RTAX_IFP]
769 #define flags info->rti_flags
772 rt_getifa(struct rt_addrinfo *info)
774 return (rt_getifa_fib(info, 0));
778 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined,
779 * it will be referenced so the caller must free it.
782 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
788 * ifp may be specified by sockaddr_dl
789 * when protocol address is ambiguous.
791 if (info->rti_ifp == NULL && ifpaddr != NULL &&
792 ifpaddr->sa_family == AF_LINK &&
793 (ifa = ifa_ifwithnet(ifpaddr, 0)) != NULL) {
794 info->rti_ifp = ifa->ifa_ifp;
797 if (info->rti_ifa == NULL && ifaaddr != NULL)
798 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
799 if (info->rti_ifa == NULL) {
802 sa = ifaaddr != NULL ? ifaaddr :
803 (gateway != NULL ? gateway : dst);
804 if (sa != NULL && info->rti_ifp != NULL)
805 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
806 else if (dst != NULL && gateway != NULL)
807 info->rti_ifa = ifa_ifwithroute_fib(flags, dst, gateway,
810 info->rti_ifa = ifa_ifwithroute_fib(flags, sa, sa,
813 if ((ifa = info->rti_ifa) != NULL) {
814 if (info->rti_ifp == NULL)
815 info->rti_ifp = ifa->ifa_ifp;
822 * Expunges references to a route that's about to be reclaimed.
823 * The route must be locked.
826 rtexpunge(struct rtentry *rt)
828 #if !defined(RADIX_MPATH)
829 struct radix_node *rn;
831 struct rt_addrinfo info;
835 struct radix_node_head *rnh;
840 * Find the correct routing tree to use for this Address Family
842 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
845 return (EAFNOSUPPORT);
846 RADIX_NODE_HEAD_LOCK_ASSERT(rnh);
850 bzero(&info, sizeof(info));
851 info.rti_ifp = rt->rt_ifp;
852 info.rti_flags = RTF_RNH_LOCKED;
853 info.rti_info[RTAX_DST] = rt_key(rt);
854 info.rti_info[RTAX_GATEWAY] = rt->rt_ifa->ifa_addr;
857 error = rtrequest1_fib(RTM_DELETE, &info, &rt0, fib);
859 if (error == 0 && rt0 != NULL) {
862 } else if (error != 0) {
868 * Remove the item from the tree; it should be there,
869 * but when callers invoke us blindly it may not (sigh).
871 rn = rnh->rnh_deladdr(rt_key(rt), rt_mask(rt), rnh);
876 KASSERT((rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)) == 0,
877 ("unexpected flags 0x%x", rn->rn_flags));
878 KASSERT(rt == RNTORT(rn),
879 ("lookup mismatch, rt %p rn %p", rt, rn));
880 #endif /* RADIX_MPATH */
882 rt->rt_flags &= ~RTF_UP;
885 * Give the protocol a chance to keep things in sync.
887 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) {
888 struct rt_addrinfo info;
890 bzero((caddr_t)&info, sizeof(info));
891 info.rti_flags = rt->rt_flags;
892 info.rti_info[RTAX_DST] = rt_key(rt);
893 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
894 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
895 ifa->ifa_rtrequest(RTM_DELETE, rt, &info);
899 * one more rtentry floating around that is not
900 * linked to the routing table.
903 #if !defined(RADIX_MPATH)
911 rn_mpath_update(int req, struct rt_addrinfo *info,
912 struct radix_node_head *rnh, struct rtentry **ret_nrt)
915 * if we got multipath routes, we require users to specify
916 * a matching RTAX_GATEWAY.
918 struct rtentry *rt, *rto = NULL;
919 register struct radix_node *rn;
922 rn = rnh->rnh_matchaddr(dst, rnh);
925 rto = rt = RNTORT(rn);
926 rt = rt_mpath_matchgate(rt, gateway);
930 * this is the first entry in the chain
933 rn = rn_mpath_next((struct radix_node *)rt);
935 * there is another entry, now it's active
940 rto->rt_flags |= RTF_UP;
942 } else if (rt->rt_flags & RTF_GATEWAY) {
944 * For gateway routes, we need to
945 * make sure that we we are deleting
946 * the correct gateway.
947 * rt_mpath_matchgate() does not
948 * check the case when there is only
949 * one route in the chain.
952 (rt->rt_gateway->sa_len != gateway->sa_len ||
953 memcmp(rt->rt_gateway, gateway, gateway->sa_len)))
957 * remove from tree before returning it
960 rn = rnh->rnh_deladdr(dst, netmask, rnh);
961 KASSERT(rt == RNTORT(rn), ("radix node disappeared"));
967 * use the normal delete code to remove
970 if (req != RTM_DELETE)
978 * if the entry is 2nd and on up
980 if ((req == RTM_DELETE) && !rt_mpath_deldup(rto, rt))
981 panic ("rtrequest1: rt_mpath_deldup");
985 if (req == RTM_DELETE) {
986 rt->rt_flags &= ~RTF_UP;
988 * One more rtentry floating around that is not
989 * linked to the routing table. rttrash will be decremented
990 * when RTFREE(rt) is eventually called.
996 if (req != RTM_DELETE)
997 panic("unrecognized request %d", req);
1001 * If the caller wants it, then it can have it,
1002 * but it's up to it to free the rtentry as we won't be
1016 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt,
1019 int error = 0, needlock = 0;
1020 register struct rtentry *rt;
1022 register struct rtentry *rt0;
1024 register struct radix_node *rn;
1025 register struct radix_node_head *rnh;
1027 struct sockaddr *ndst;
1028 struct sockaddr_storage mdst;
1029 #define senderr(x) { error = x ; goto bad; }
1031 KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum"));
1032 if (dst->sa_family != AF_INET) /* Only INET supports > 1 fib now */
1035 * Find the correct routing tree to use for this Address Family
1037 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
1039 return (EAFNOSUPPORT);
1040 needlock = ((flags & RTF_RNH_LOCKED) == 0);
1041 flags &= ~RTF_RNH_LOCKED;
1043 RADIX_NODE_HEAD_LOCK(rnh);
1045 RADIX_NODE_HEAD_LOCK_ASSERT(rnh);
1047 * If we are adding a host route then we don't want to put
1048 * a netmask in the tree, nor do we want to clone it.
1050 if (flags & RTF_HOST)
1056 rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask);
1057 dst = (struct sockaddr *)&mdst;
1060 if (rn_mpath_capable(rnh)) {
1061 error = rn_mpath_update(req, info, rnh, ret_nrt);
1063 * "bad" holds true for the success case
1066 if (error != ENOENT)
1072 * Remove the item from the tree and return it.
1073 * Complain if it is not there and do no more processing.
1075 rn = rnh->rnh_deladdr(dst, netmask, rnh);
1078 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
1079 panic ("rtrequest delete");
1083 rt->rt_flags &= ~RTF_UP;
1086 * give the protocol a chance to keep things in sync.
1088 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
1089 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1092 * One more rtentry floating around that is not
1093 * linked to the routing table. rttrash will be decremented
1094 * when RTFREE(rt) is eventually called.
1099 * If the caller wants it, then it can have it,
1100 * but it's up to it to free the rtentry as we won't be
1111 * resolve was only used for route cloning
1116 if ((flags & RTF_GATEWAY) && !gateway)
1118 if (dst && gateway && (dst->sa_family != gateway->sa_family) &&
1119 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
1122 if (info->rti_ifa == NULL) {
1123 error = rt_getifa_fib(info, fibnum);
1127 ifa_ref(info->rti_ifa);
1128 ifa = info->rti_ifa;
1129 rt = uma_zalloc(V_rtzone, M_NOWAIT | M_ZERO);
1136 rt->rt_flags = RTF_UP | flags;
1137 rt->rt_fibnum = fibnum;
1139 * Add the gateway. Possibly re-malloc-ing the storage for it
1143 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
1144 RT_LOCK_DESTROY(rt);
1147 uma_zfree(V_rtzone, rt);
1152 * point to the (possibly newly malloc'd) dest address.
1154 ndst = (struct sockaddr *)rt_key(rt);
1157 * make sure it contains the value we want (masked if needed).
1160 rt_maskedcopy(dst, ndst, netmask);
1162 bcopy(dst, ndst, dst->sa_len);
1165 * We use the ifa reference returned by rt_getifa_fib().
1166 * This moved from below so that rnh->rnh_addaddr() can
1167 * examine the ifa and ifa->ifa_ifp if it so desires.
1170 rt->rt_ifp = ifa->ifa_ifp;
1171 rt->rt_rmx.rmx_weight = 1;
1174 /* do not permit exactly the same dst/mask/gw pair */
1175 if (rn_mpath_capable(rnh) &&
1176 rt_mpath_conflict(rnh, rt, netmask)) {
1178 ifa_free(rt->rt_ifa);
1181 RT_LOCK_DESTROY(rt);
1182 uma_zfree(V_rtzone, rt);
1190 * "flow-table" only support IPv4 at the moment.
1191 * XXX-BZ as of r205066 it would support IPv6.
1194 if (dst->sa_family == AF_INET) {
1195 rn = rnh->rnh_matchaddr(dst, rnh);
1196 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
1197 struct sockaddr *mask;
1202 * compare mask to see if the new route is
1203 * more specific than the existing one
1210 * A host route is already present, so
1211 * leave the flow-table entries as is.
1213 if (rt0->rt_flags & RTF_HOST) {
1216 } else if (!(flags & RTF_HOST) && netmask) {
1217 mask = rt_mask(rt0);
1220 n = (u_char *)netmask;
1227 if (len == 0 || (*n < *m)) {
1237 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
1238 rn = rnh->rnh_addaddr(ndst, netmask, rnh, rt->rt_nodes);
1240 * If it still failed to go into the tree,
1241 * then un-make it (this should be a function)
1245 ifa_free(rt->rt_ifa);
1247 RT_LOCK_DESTROY(rt);
1248 uma_zfree(V_rtzone, rt);
1256 else if (rt0 != NULL) {
1258 flowtable_route_flush(V_ip_ft, rt0);
1265 * If this protocol has something to add to this then
1266 * allow it to do that as well.
1268 if (ifa->ifa_rtrequest)
1269 ifa->ifa_rtrequest(req, rt, info);
1272 * actually return a resultant rtentry and
1273 * give the caller a single reference.
1286 RADIX_NODE_HEAD_UNLOCK(rnh);
1299 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
1301 /* XXX dst may be overwritten, can we move this to below */
1302 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
1304 struct radix_node_head *rnh;
1306 rnh = rt_tables_get_rnh(rt->rt_fibnum, dst->sa_family);
1310 RADIX_NODE_HEAD_LOCK_ASSERT(rnh);
1313 * Prepare to store the gateway in rt->rt_gateway.
1314 * Both dst and gateway are stored one after the other in the same
1315 * malloc'd chunk. If we have room, we can reuse the old buffer,
1316 * rt_gateway already points to the right place.
1317 * Otherwise, malloc a new block and update the 'dst' address.
1319 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
1322 R_Malloc(new, caddr_t, dlen + glen);
1326 * XXX note, we copy from *dst and not *rt_key(rt) because
1327 * rt_setgate() can be called to initialize a newly
1328 * allocated route entry, in which case rt_key(rt) == NULL
1329 * (and also rt->rt_gateway == NULL).
1330 * Free()/free() handle a NULL argument just fine.
1332 bcopy(dst, new, dlen);
1333 Free(rt_key(rt)); /* free old block, if any */
1334 rt_key(rt) = (struct sockaddr *)new;
1335 rt->rt_gateway = (struct sockaddr *)(new + dlen);
1339 * Copy the new gateway value into the memory chunk.
1341 bcopy(gate, rt->rt_gateway, glen);
1347 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
1349 register u_char *cp1 = (u_char *)src;
1350 register u_char *cp2 = (u_char *)dst;
1351 register u_char *cp3 = (u_char *)netmask;
1352 u_char *cplim = cp2 + *cp3;
1353 u_char *cplim2 = cp2 + *cp1;
1355 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1360 *cp2++ = *cp1++ & *cp3++;
1362 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1366 * Set up a routing table entry, normally
1369 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
1371 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
1373 struct sockaddr *dst;
1374 struct sockaddr *netmask;
1375 struct rtentry *rt = NULL;
1376 struct rt_addrinfo info;
1378 int startfib, endfib;
1379 char tempbuf[_SOCKADDR_TMPSIZE];
1382 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1384 if (flags & RTF_HOST) {
1385 dst = ifa->ifa_dstaddr;
1388 dst = ifa->ifa_addr;
1389 netmask = ifa->ifa_netmask;
1391 if ( dst->sa_family != AF_INET)
1394 if (rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD) {
1395 startfib = endfib = curthread->td_proc->p_fibnum;
1398 endfib = rt_numfibs - 1;
1401 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
1405 if (dst->sa_len == 0)
1409 * If it's a delete, check that if it exists,
1410 * it's on the correct interface or we might scrub
1411 * a route to another ifa which would
1412 * be confusing at best and possibly worse.
1414 if (cmd == RTM_DELETE) {
1416 * It's a delete, so it should already exist..
1417 * If it's a net, mask off the host bits
1418 * (Assuming we have a mask)
1419 * XXX this is kinda inet specific..
1421 if (netmask != NULL) {
1422 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
1423 dst = (struct sockaddr *)tempbuf;
1427 * Now go through all the requested tables (fibs) and do the
1428 * requested action. Realistically, this will either be fib 0
1429 * for protocols that don't do multiple tables or all the
1430 * tables for those that do. XXX For this version only AF_INET.
1431 * When that changes code should be refactored to protocol
1432 * independent parts and protocol dependent parts.
1434 for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
1435 if (cmd == RTM_DELETE) {
1436 struct radix_node_head *rnh;
1437 struct radix_node *rn;
1439 * Look up an rtentry that is in the routing tree and
1440 * contains the correct info.
1442 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
1444 /* this table doesn't exist but others might */
1446 RADIX_NODE_HEAD_LOCK(rnh);
1448 if (rn_mpath_capable(rnh)) {
1450 rn = rnh->rnh_matchaddr(dst, rnh);
1456 * for interface route the
1457 * rt->rt_gateway is sockaddr_intf
1458 * for cloning ARP entries, so
1459 * rt_mpath_matchgate must use the
1462 rt = rt_mpath_matchgate(rt,
1470 rn = rnh->rnh_lookup(dst, netmask, rnh);
1471 error = (rn == NULL ||
1472 (rn->rn_flags & RNF_ROOT) ||
1473 RNTORT(rn)->rt_ifa != ifa ||
1474 !sa_equal((struct sockaddr *)rn->rn_key, dst));
1475 RADIX_NODE_HEAD_UNLOCK(rnh);
1477 /* this is only an error if bad on ALL tables */
1482 * Do the actual request
1484 bzero((caddr_t)&info, sizeof(info));
1486 info.rti_flags = flags | (ifa->ifa_flags & ~IFA_RTSELF);
1487 info.rti_info[RTAX_DST] = dst;
1489 * doing this for compatibility reasons
1492 info.rti_info[RTAX_GATEWAY] =
1493 (struct sockaddr *)&null_sdl;
1495 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
1496 info.rti_info[RTAX_NETMASK] = netmask;
1497 error = rtrequest1_fib(cmd, &info, &rt, fibnum);
1498 if (error == 0 && rt != NULL) {
1500 * notify any listening routing agents of the change
1505 * in case address alias finds the first address
1506 * e.g. ifconfig bge0 192.103.54.246/24
1507 * e.g. ifconfig bge0 192.103.54.247/24
1508 * the address set in the route is 192.103.54.246
1509 * so we need to replace it with 192.103.54.247
1511 if (memcmp(rt->rt_ifa->ifa_addr,
1512 ifa->ifa_addr, ifa->ifa_addr->sa_len)) {
1513 ifa_free(rt->rt_ifa);
1515 rt->rt_ifp = ifa->ifa_ifp;
1520 * doing this for compatibility reasons
1522 if (cmd == RTM_ADD) {
1523 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
1524 rt->rt_ifp->if_type;
1525 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1526 rt->rt_ifp->if_index;
1530 rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum);
1533 if (cmd == RTM_DELETE) {
1535 * If we are deleting, and we found an entry,
1536 * then it's been removed from the tree..
1537 * now throw it away.
1541 if (cmd == RTM_ADD) {
1543 * We just wanted to add it..
1544 * we don't actually need a reference.
1555 if (cmd == RTM_DELETE) {
1559 /* we only give an error if it wasn't in any table */
1560 error = ((flags & RTF_HOST) ?
1561 EHOSTUNREACH : ENETUNREACH);
1565 /* return an error if any of them failed */
1572 /* special one for inet internal use. may not use. */
1574 rtinit_fib(struct ifaddr *ifa, int cmd, int flags)
1576 return (rtinit1(ifa, cmd, flags, -1));
1580 * Set up a routing table entry, normally
1584 rtinit(struct ifaddr *ifa, int cmd, int flags)
1586 struct sockaddr *dst;
1589 if (flags & RTF_HOST) {
1590 dst = ifa->ifa_dstaddr;
1592 dst = ifa->ifa_addr;
1595 if (dst->sa_family == AF_INET)
1597 return (rtinit1(ifa, cmd, flags, fib));