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_inet6.h"
39 #include "opt_route.h"
41 #include "opt_mrouting.h"
42 #include "opt_mpath.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/malloc.h>
48 #include <sys/socket.h>
49 #include <sys/sysctl.h>
50 #include <sys/syslog.h>
51 #include <sys/sysproto.h>
53 #include <sys/domain.h>
54 #include <sys/kernel.h>
57 #include <net/if_var.h>
58 #include <net/if_dl.h>
59 #include <net/route.h>
61 #include <net/flowtable.h>
64 #include <net/radix_mpath.h>
67 #include <netinet/in.h>
68 #include <netinet/ip_mroute.h>
72 #define RT_MAXFIBS UINT16_MAX
74 /* Kernel config default option. */
77 #error "ROUTETABLES defined too low"
79 #if ROUTETABLES > RT_MAXFIBS
80 #error "ROUTETABLES defined too big"
82 #define RT_NUMFIBS ROUTETABLES
83 #endif /* ROUTETABLES */
84 /* Initialize to default if not otherwise set. */
89 #if defined(INET) || defined(INET6)
91 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
96 /* This is read-only.. */
97 u_int rt_numfibs = RT_NUMFIBS;
98 SYSCTL_UINT(_net, OID_AUTO, fibs, CTLFLAG_RDTUN, &rt_numfibs, 0, "");
101 * By default add routes to all fibs for new interfaces.
102 * Once this is set to 0 then only allocate routes on interface
103 * changes for the FIB of the caller when adding a new set of addresses
104 * to an interface. XXX this is a shotgun aproach to a problem that needs
105 * a more fine grained solution.. that will come.
106 * XXX also has the problems getting the FIB from curthread which will not
107 * always work given the fib can be overridden and prefixes can be added
108 * from the network stack context.
110 VNET_DEFINE(u_int, rt_add_addr_allfibs) = 1;
111 SYSCTL_UINT(_net, OID_AUTO, add_addr_allfibs, CTLFLAG_RWTUN | CTLFLAG_VNET,
112 &VNET_NAME(rt_add_addr_allfibs), 0, "");
114 VNET_DEFINE(struct rtstat, rtstat);
115 #define V_rtstat VNET(rtstat)
117 VNET_DEFINE(struct radix_node_head *, rt_tables);
118 #define V_rt_tables VNET(rt_tables)
120 VNET_DEFINE(int, rttrash); /* routes not in table but not freed */
121 #define V_rttrash VNET(rttrash)
125 * Convert a 'struct radix_node *' to a 'struct rtentry *'.
126 * The operation can be done safely (in this code) because a
127 * 'struct rtentry' starts with two 'struct radix_node''s, the first
128 * one representing leaf nodes in the routing tree, which is
129 * what the code in radix.c passes us as a 'struct radix_node'.
131 * But because there are a lot of assumptions in this conversion,
132 * do not cast explicitly, but always use the macro below.
134 #define RNTORT(p) ((struct rtentry *)(p))
136 static VNET_DEFINE(uma_zone_t, rtzone); /* Routing table UMA zone. */
137 #define V_rtzone VNET(rtzone)
139 static int rtrequest1_fib_change(struct radix_node_head *, struct rt_addrinfo *,
140 struct rtentry **, u_int);
141 static void rt_setmetrics(const struct rt_addrinfo *, struct rtentry *);
142 static int rt_ifdelroute(const struct rtentry *rt, void *arg);
143 static struct rtentry *rt_unlinkrte(struct radix_node_head *rnh,
144 struct rt_addrinfo *info, int *perror);
145 static void rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info);
147 static struct radix_node *rt_mpath_unlink(struct radix_node_head *rnh,
148 struct rt_addrinfo *info, struct rtentry *rto, int *perror);
150 static int rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info,
159 static int if_updatemtu_cb(struct radix_node *, void *);
162 * handler for net.my_fibnum
165 sysctl_my_fibnum(SYSCTL_HANDLER_ARGS)
170 fibnum = curthread->td_proc->p_fibnum;
171 error = sysctl_handle_int(oidp, &fibnum, 0, req);
175 SYSCTL_PROC(_net, OID_AUTO, my_fibnum, CTLTYPE_INT|CTLFLAG_RD,
176 NULL, 0, &sysctl_my_fibnum, "I", "default FIB of caller");
178 static __inline struct radix_node_head **
179 rt_tables_get_rnh_ptr(int table, int fam)
181 struct radix_node_head **rnh;
183 KASSERT(table >= 0 && table < rt_numfibs, ("%s: table out of bounds.",
185 KASSERT(fam >= 0 && fam < (AF_MAX+1), ("%s: fam out of bounds.",
188 /* rnh is [fib=0][af=0]. */
189 rnh = (struct radix_node_head **)V_rt_tables;
190 /* Get the offset to the requested table and fam. */
191 rnh += table * (AF_MAX+1) + fam;
196 struct radix_node_head *
197 rt_tables_get_rnh(int table, int fam)
200 return (*rt_tables_get_rnh_ptr(table, fam));
204 * route initialization must occur before ip6_init2(), which happenas at
211 /* whack the tunable ints into line. */
212 if (rt_numfibs > RT_MAXFIBS)
213 rt_numfibs = RT_MAXFIBS;
217 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0);
220 rtentry_zinit(void *mem, int size, int how)
222 struct rtentry *rt = mem;
224 rt->rt_pksent = counter_u64_alloc(how);
225 if (rt->rt_pksent == NULL)
234 rtentry_zfini(void *mem, int size)
236 struct rtentry *rt = mem;
239 counter_u64_free(rt->rt_pksent);
243 rtentry_ctor(void *mem, int size, void *arg, int how)
245 struct rtentry *rt = mem;
247 bzero(rt, offsetof(struct rtentry, rt_endzero));
248 counter_u64_zero(rt->rt_pksent);
255 rtentry_dtor(void *mem, int size, void *arg)
257 struct rtentry *rt = mem;
263 vnet_route_init(const void *unused __unused)
266 struct radix_node_head **rnh;
270 V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) *
271 sizeof(struct radix_node_head *), M_RTABLE, M_WAITOK|M_ZERO);
273 V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry),
274 rtentry_ctor, rtentry_dtor,
275 rtentry_zinit, rtentry_zfini, UMA_ALIGN_PTR, 0);
276 for (dom = domains; dom; dom = dom->dom_next) {
277 if (dom->dom_rtattach == NULL)
280 for (table = 0; table < rt_numfibs; table++) {
281 fam = dom->dom_family;
282 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
285 rnh = rt_tables_get_rnh_ptr(table, fam);
287 panic("%s: rnh NULL", __func__);
288 dom->dom_rtattach((void **)rnh, 0);
292 VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
297 vnet_route_uninit(const void *unused __unused)
302 struct radix_node_head **rnh;
304 for (dom = domains; dom; dom = dom->dom_next) {
305 if (dom->dom_rtdetach == NULL)
308 for (table = 0; table < rt_numfibs; table++) {
309 fam = dom->dom_family;
311 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
314 rnh = rt_tables_get_rnh_ptr(table, fam);
316 panic("%s: rnh NULL", __func__);
317 dom->dom_rtdetach((void **)rnh, 0);
321 free(V_rt_tables, M_RTABLE);
322 uma_zdestroy(V_rtzone);
324 VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
325 vnet_route_uninit, 0);
328 #ifndef _SYS_SYSPROTO_H_
334 sys_setfib(struct thread *td, struct setfib_args *uap)
336 if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs)
338 td->td_proc->p_fibnum = uap->fibnum;
343 * Packet routing routines.
346 rtalloc(struct route *ro)
349 rtalloc_ign_fib(ro, 0UL, RT_DEFAULT_FIB);
353 rtalloc_fib(struct route *ro, u_int fibnum)
355 rtalloc_ign_fib(ro, 0UL, fibnum);
359 rtalloc_ign(struct route *ro, u_long ignore)
363 if ((rt = ro->ro_rt) != NULL) {
364 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
369 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, RT_DEFAULT_FIB);
371 RT_UNLOCK(ro->ro_rt);
375 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum)
379 if ((rt = ro->ro_rt) != NULL) {
380 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
385 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum);
387 RT_UNLOCK(ro->ro_rt);
391 * Look up the route that matches the address given
392 * Or, at least try.. Create a cloned route if needed.
394 * The returned route, if any, is locked.
397 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
400 return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB));
404 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags,
407 struct radix_node_head *rnh;
408 struct radix_node *rn;
409 struct rtentry *newrt;
410 struct rt_addrinfo info;
411 int err = 0, msgtype = RTM_MISS;
414 KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum"));
415 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
421 * Look up the address in the table for that Address Family
423 needlock = !(ignflags & RTF_RNH_LOCKED);
425 RADIX_NODE_HEAD_RLOCK(rnh);
428 RADIX_NODE_HEAD_LOCK_ASSERT(rnh);
430 rn = rnh->rnh_matchaddr(dst, rnh);
431 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
436 RADIX_NODE_HEAD_RUNLOCK(rnh);
440 RADIX_NODE_HEAD_RUNLOCK(rnh);
443 * Either we hit the root or couldn't find any match,
444 * Which basically means
445 * "caint get there frm here"
448 V_rtstat.rts_unreach++;
452 * If required, report the failure to the supervising
454 * For a delete, this is not an error. (report == 0)
456 bzero(&info, sizeof(info));
457 info.rti_info[RTAX_DST] = dst;
458 rt_missmsg_fib(msgtype, &info, 0, err, fibnum);
462 RT_LOCK_ASSERT(newrt);
467 * Remove a reference count from an rtentry.
468 * If the count gets low enough, take it out of the routing table
471 rtfree(struct rtentry *rt)
473 struct radix_node_head *rnh;
475 KASSERT(rt != NULL,("%s: NULL rt", __func__));
476 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
477 KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
482 * The callers should use RTFREE_LOCKED() or RTFREE(), so
483 * we should come here exactly with the last reference.
486 if (rt->rt_refcnt > 0) {
487 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt);
492 * On last reference give the "close method" a chance
493 * to cleanup private state. This also permits (for
494 * IPv4 and IPv6) a chance to decide if the routing table
495 * entry should be purged immediately or at a later time.
496 * When an immediate purge is to happen the close routine
497 * typically calls rtexpunge which clears the RTF_UP flag
498 * on the entry so that the code below reclaims the storage.
500 if (rt->rt_refcnt == 0 && rnh->rnh_close)
501 rnh->rnh_close((struct radix_node *)rt, rnh);
504 * If we are no longer "up" (and ref == 0)
505 * then we can free the resources associated
508 if ((rt->rt_flags & RTF_UP) == 0) {
509 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
512 * the rtentry must have been removed from the routing table
513 * so it is represented in rttrash.. remove that now.
517 if (rt->rt_refcnt < 0) {
518 printf("rtfree: %p not freed (neg refs)\n", rt);
523 * release references on items we hold them on..
524 * e.g other routes and ifaddrs.
527 ifa_free(rt->rt_ifa);
529 * The key is separatly alloc'd so free it (see rt_setgate()).
530 * This also frees the gateway, as they are always malloc'd
536 * and the rtentry itself of course
538 uma_zfree(V_rtzone, rt);
547 * Force a routing table entry to the specified
548 * destination to go through the given gateway.
549 * Normally called as a result of a routing redirect
550 * message from the network layer.
553 rtredirect(struct sockaddr *dst,
554 struct sockaddr *gateway,
555 struct sockaddr *netmask,
557 struct sockaddr *src)
560 rtredirect_fib(dst, gateway, netmask, flags, src, RT_DEFAULT_FIB);
564 rtredirect_fib(struct sockaddr *dst,
565 struct sockaddr *gateway,
566 struct sockaddr *netmask,
568 struct sockaddr *src,
571 struct rtentry *rt, *rt0 = NULL;
574 struct rt_addrinfo info;
576 struct radix_node_head *rnh;
579 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
581 error = EAFNOSUPPORT;
585 /* verify the gateway is directly reachable */
586 if ((ifa = ifa_ifwithnet(gateway, 0, fibnum)) == NULL) {
590 rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */
592 * If the redirect isn't from our current router for this dst,
593 * it's either old or wrong. If it redirects us to ourselves,
594 * we have a routing loop, perhaps as a result of an interface
595 * going down recently.
597 if (!(flags & RTF_DONE) && rt) {
598 if (!sa_equal(src, rt->rt_gateway)) {
602 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) {
607 if ((flags & RTF_GATEWAY) && ifa_ifwithaddr_check(gateway)) {
608 error = EHOSTUNREACH;
612 * Create a new entry if we just got back a wildcard entry
613 * or the lookup failed. This is necessary for hosts
614 * which use routing redirects generated by smart gateways
615 * to dynamically build the routing tables.
617 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
620 * Don't listen to the redirect if it's
621 * for a route to an interface.
623 if (rt->rt_flags & RTF_GATEWAY) {
624 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
626 * Changing from route to net => route to host.
627 * Create new route, rather than smashing route to net.
633 flags |= RTF_DYNAMIC;
634 bzero((caddr_t)&info, sizeof(info));
635 info.rti_info[RTAX_DST] = dst;
636 info.rti_info[RTAX_GATEWAY] = gateway;
637 info.rti_info[RTAX_NETMASK] = netmask;
639 info.rti_flags = flags;
641 RT_UNLOCK(rt0); /* drop lock to avoid LOR with RNH */
642 error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum);
646 EVENTHANDLER_INVOKE(route_redirect_event, rt0, rt, dst);
647 flags = rt->rt_flags;
652 stat = &V_rtstat.rts_dynamic;
654 struct rtentry *gwrt;
657 * Smash the current notion of the gateway to
658 * this destination. Should check about netmask!!!
660 if ((flags & RTF_GATEWAY) == 0)
661 rt->rt_flags &= ~RTF_GATEWAY;
662 rt->rt_flags |= RTF_MODIFIED;
663 flags |= RTF_MODIFIED;
664 stat = &V_rtstat.rts_newgateway;
666 * add the key and gateway (in one malloc'd chunk).
669 RADIX_NODE_HEAD_LOCK(rnh);
671 rt_setgate(rt, rt_key(rt), gateway);
672 gwrt = rtalloc1(gateway, 1, RTF_RNH_LOCKED);
673 RADIX_NODE_HEAD_UNLOCK(rnh);
674 EVENTHANDLER_INVOKE(route_redirect_event, rt, gwrt, dst);
679 error = EHOSTUNREACH;
685 V_rtstat.rts_badredirect++;
686 else if (stat != NULL)
688 bzero((caddr_t)&info, sizeof(info));
689 info.rti_info[RTAX_DST] = dst;
690 info.rti_info[RTAX_GATEWAY] = gateway;
691 info.rti_info[RTAX_NETMASK] = netmask;
692 info.rti_info[RTAX_AUTHOR] = src;
693 rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum);
699 rtioctl(u_long req, caddr_t data)
702 return (rtioctl_fib(req, data, RT_DEFAULT_FIB));
706 * Routing table ioctl interface.
709 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
713 * If more ioctl commands are added here, make sure the proper
714 * super-user checks are being performed because it is possible for
715 * prison-root to make it this far if raw sockets have been enabled
719 /* Multicast goop, grrr... */
720 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
727 ifa_ifwithroute(int flags, const struct sockaddr *dst, struct sockaddr *gateway,
733 if ((flags & RTF_GATEWAY) == 0) {
735 * If we are adding a route to an interface,
736 * and the interface is a pt to pt link
737 * we should search for the destination
738 * as our clue to the interface. Otherwise
739 * we can use the local address.
742 if (flags & RTF_HOST)
743 ifa = ifa_ifwithdstaddr(dst, fibnum);
745 ifa = ifa_ifwithaddr(gateway);
748 * If we are adding a route to a remote net
749 * or host, the gateway may still be on the
750 * other end of a pt to pt link.
752 ifa = ifa_ifwithdstaddr(gateway, fibnum);
755 ifa = ifa_ifwithnet(gateway, 0, fibnum);
757 struct rtentry *rt = rtalloc1_fib(gateway, 0, RTF_RNH_LOCKED, fibnum);
761 * dismiss a gateway that is reachable only
762 * through the default router
764 switch (gateway->sa_family) {
766 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
770 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
776 if (!not_found && rt->rt_ifa != NULL) {
782 if (not_found || ifa == NULL)
785 if (ifa->ifa_addr->sa_family != dst->sa_family) {
786 struct ifaddr *oifa = ifa;
787 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
797 * Do appropriate manipulations of a routing tree given
798 * all the bits of info needed
802 struct sockaddr *dst,
803 struct sockaddr *gateway,
804 struct sockaddr *netmask,
806 struct rtentry **ret_nrt)
809 return (rtrequest_fib(req, dst, gateway, netmask, flags, ret_nrt,
814 rtrequest_fib(int req,
815 struct sockaddr *dst,
816 struct sockaddr *gateway,
817 struct sockaddr *netmask,
819 struct rtentry **ret_nrt,
822 struct rt_addrinfo info;
824 if (dst->sa_len == 0)
827 bzero((caddr_t)&info, sizeof(info));
828 info.rti_flags = flags;
829 info.rti_info[RTAX_DST] = dst;
830 info.rti_info[RTAX_GATEWAY] = gateway;
831 info.rti_info[RTAX_NETMASK] = netmask;
832 return rtrequest1_fib(req, &info, ret_nrt, fibnum);
837 * Copy most of @rt data into @info.
839 * If @flags contains NHR_COPY, copies dst,netmask and gw to the
840 * pointers specified by @info structure. Assume such pointers
841 * are zeroed sockaddr-like structures with sa_len field initialized
842 * to reflect size of the provided buffer. if no NHR_COPY is specified,
843 * point dst,netmask and gw @info fields to appropriate @rt values.
845 * if @flags contains NHR_REF, do refcouting on rt_ifp.
847 * Returns 0 on success.
850 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags)
852 struct rt_metrics *rmx;
853 struct sockaddr *src, *dst;
856 if (flags & NHR_COPY) {
857 /* Copy destination if dst is non-zero */
859 dst = info->rti_info[RTAX_DST];
860 sa_len = src->sa_len;
861 if (src != NULL && dst != NULL) {
862 if (src->sa_len > dst->sa_len)
864 memcpy(dst, src, src->sa_len);
865 info->rti_addrs |= RTA_DST;
868 /* Copy mask if set && dst is non-zero */
870 dst = info->rti_info[RTAX_NETMASK];
871 if (src != NULL && dst != NULL) {
874 * Radix stores different value in sa_len,
875 * assume rt_mask() to have the same length
878 if (sa_len > dst->sa_len)
880 memcpy(dst, src, src->sa_len);
881 info->rti_addrs |= RTA_NETMASK;
884 /* Copy gateway is set && dst is non-zero */
885 src = rt->rt_gateway;
886 dst = info->rti_info[RTAX_GATEWAY];
887 if ((rt->rt_flags & RTF_GATEWAY) && src != NULL && dst != NULL){
888 if (src->sa_len > dst->sa_len)
890 memcpy(dst, src, src->sa_len);
891 info->rti_addrs |= RTA_GATEWAY;
894 info->rti_info[RTAX_DST] = rt_key(rt);
895 info->rti_addrs |= RTA_DST;
896 if (rt_mask(rt) != NULL) {
897 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
898 info->rti_addrs |= RTA_NETMASK;
900 if (rt->rt_flags & RTF_GATEWAY) {
901 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
902 info->rti_addrs |= RTA_GATEWAY;
908 info->rti_mflags |= RTV_MTU;
909 rmx->rmx_mtu = rt->rt_mtu;
912 info->rti_flags = rt->rt_flags;
913 info->rti_ifp = rt->rt_ifp;
914 info->rti_ifa = rt->rt_ifa;
916 if (flags & NHR_REF) {
917 /* Do 'traditional' refcouting */
918 if_ref(info->rti_ifp);
925 * Lookups up route entry for @dst in RIB database for fib @fibnum.
926 * Exports entry data to @info using rt_exportinfo().
928 * if @flags contains NHR_REF, refcouting is performed on rt_ifp.
929 * All references can be released later by calling rib_free_info()
931 * Returns 0 on success.
932 * Returns ENOENT for lookup failure, ENOMEM for export failure.
935 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
936 uint32_t flowid, struct rt_addrinfo *info)
938 struct radix_node_head *rh;
939 struct radix_node *rn;
943 KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
944 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
948 RADIX_NODE_HEAD_RLOCK(rh);
949 rn = rh->rnh_matchaddr(__DECONST(void *, dst), rh);
950 if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
952 /* Ensure route & ifp is UP */
953 if (RT_LINK_IS_UP(rt->rt_ifp)) {
954 flags = (flags & NHR_REF) | NHR_COPY;
955 error = rt_exportinfo(rt, info, flags);
956 RADIX_NODE_HEAD_RUNLOCK(rh);
961 RADIX_NODE_HEAD_RUNLOCK(rh);
967 * Releases all references acquired by rib_lookup_info() when
968 * called with NHR_REF flags.
971 rib_free_info(struct rt_addrinfo *info)
974 if_rele(info->rti_ifp);
978 * Iterates over all existing fibs in system calling
979 * @setwa_f function prior to traversing each fib.
980 * Calls @wa_f function for each element in current fib.
981 * If af is not AF_UNSPEC, iterates over fibs in particular
985 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f,
988 struct radix_node_head *rnh;
992 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
993 /* Do we want some specific family? */
994 if (af != AF_UNSPEC) {
995 rnh = rt_tables_get_rnh(fibnum, af);
999 setwa_f(rnh, fibnum, af, arg);
1001 RADIX_NODE_HEAD_LOCK(rnh);
1002 rnh->rnh_walktree(rnh, (walktree_f_t *)wa_f, arg);
1003 RADIX_NODE_HEAD_UNLOCK(rnh);
1007 for (i = 1; i <= AF_MAX; i++) {
1008 rnh = rt_tables_get_rnh(fibnum, i);
1011 if (setwa_f != NULL)
1012 setwa_f(rnh, fibnum, i, arg);
1014 RADIX_NODE_HEAD_LOCK(rnh);
1015 rnh->rnh_walktree(rnh, (walktree_f_t *)wa_f, arg);
1016 RADIX_NODE_HEAD_UNLOCK(rnh);
1023 struct rt_addrinfo info;
1024 struct radix_node_head *rnh;
1025 struct rtentry *head;
1029 * Conditionally unlinks @rn from radix tree based
1030 * on info data passed in @arg.
1033 rt_checkdelroute(struct radix_node *rn, void *arg)
1035 struct rt_delinfo *di;
1036 struct rt_addrinfo *info;
1040 di = (struct rt_delinfo *)arg;
1041 rt = (struct rtentry *)rn;
1045 info->rti_info[RTAX_DST] = rt_key(rt);
1046 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
1047 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1049 rt = rt_unlinkrte(di->rnh, info, &error);
1051 /* Either not allowed or not matched. Skip entry */
1055 /* Entry was unlinked. Add to the list and return */
1056 rt->rt_chain = di->head;
1063 * Iterates over all existing fibs in system.
1064 * Deletes each element for which @filter_f function returned
1066 * If @af is not AF_UNSPEC, iterates over fibs in particular
1070 rt_foreach_fib_walk_del(int af, rt_filter_f_t *filter_f, void *arg)
1072 struct radix_node_head *rnh;
1073 struct rt_delinfo di;
1078 bzero(&di, sizeof(di));
1079 di.info.rti_filter = filter_f;
1080 di.info.rti_filterdata = arg;
1082 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1083 /* Do we want some specific family? */
1084 if (af != AF_UNSPEC) {
1092 for (i = start; i <= end; i++) {
1093 rnh = rt_tables_get_rnh(fibnum, i);
1098 RADIX_NODE_HEAD_LOCK(rnh);
1099 rnh->rnh_walktree(rnh, rt_checkdelroute, &di);
1100 RADIX_NODE_HEAD_UNLOCK(rnh);
1102 if (di.head == NULL)
1105 /* We might have something to reclaim */
1106 while (di.head != NULL) {
1108 di.head = rt->rt_chain;
1109 rt->rt_chain = NULL;
1111 /* TODO std rt -> rt_addrinfo export */
1112 di.info.rti_info[RTAX_DST] = rt_key(rt);
1113 di.info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1115 rt_notifydelete(rt, &di.info);
1124 * Delete Routes for a Network Interface
1126 * Called for each routing entry via the rnh->rnh_walktree() call above
1127 * to delete all route entries referencing a detaching network interface.
1130 * rt pointer to rtentry
1131 * arg argument passed to rnh->rnh_walktree() - detaching interface
1135 * errno failed - reason indicated
1138 rt_ifdelroute(const struct rtentry *rt, void *arg)
1140 struct ifnet *ifp = arg;
1142 if (rt->rt_ifp != ifp)
1146 * Protect (sorta) against walktree recursion problems
1147 * with cloned routes
1149 if ((rt->rt_flags & RTF_UP) == 0)
1156 * Delete all remaining routes using this interface
1157 * Unfortuneatly the only way to do this is to slog through
1158 * the entire routing table looking for routes which point
1159 * to this interface...oh well...
1162 rt_flushifroutes(struct ifnet *ifp)
1165 rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
1169 * Conditionally unlinks rtentry matching data inside @info from @rnh.
1170 * Returns unlinked, locked and referenced @rtentry on success,
1171 * Returns NULL and sets @perror to:
1172 * ESRCH - if prefix was not found,
1173 * EADDRINUSE - if trying to delete PINNED route without appropriate flag.
1174 * ENOENT - if supplied filter function returned 0 (not matched).
1176 static struct rtentry *
1177 rt_unlinkrte(struct radix_node_head *rnh, struct rt_addrinfo *info, int *perror)
1179 struct sockaddr *dst, *netmask;
1181 struct radix_node *rn;
1183 dst = info->rti_info[RTAX_DST];
1184 netmask = info->rti_info[RTAX_NETMASK];
1186 rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, rnh);
1192 if ((info->rti_flags & RTF_PINNED) == 0) {
1193 /* Check if target route can be deleted */
1194 if (rt->rt_flags & RTF_PINNED) {
1195 *perror = EADDRINUSE;
1200 if (info->rti_filter != NULL) {
1201 if (info->rti_filter(rt, info->rti_filterdata) == 0) {
1208 * Filter function requested rte deletion.
1209 * Ease the caller work by filling in remaining info
1210 * from that particular entry.
1212 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1216 * Remove the item from the tree and return it.
1217 * Complain if it is not there and do no more processing.
1221 if (rn_mpath_capable(rnh))
1222 rn = rt_mpath_unlink(rnh, info, rt, perror);
1225 rn = rnh->rnh_deladdr(dst, netmask, rnh);
1229 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
1230 panic ("rtrequest delete");
1235 rt->rt_flags &= ~RTF_UP;
1243 rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info)
1248 * give the protocol a chance to keep things in sync.
1251 if (ifa != NULL && ifa->ifa_rtrequest != NULL)
1252 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1255 * One more rtentry floating around that is not
1256 * linked to the routing table. rttrash will be decremented
1257 * when RTFREE(rt) is eventually called.
1264 * These (questionable) definitions of apparent local variables apply
1265 * to the next two functions. XXXXXX!!!
1267 #define dst info->rti_info[RTAX_DST]
1268 #define gateway info->rti_info[RTAX_GATEWAY]
1269 #define netmask info->rti_info[RTAX_NETMASK]
1270 #define ifaaddr info->rti_info[RTAX_IFA]
1271 #define ifpaddr info->rti_info[RTAX_IFP]
1272 #define flags info->rti_flags
1275 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined,
1276 * it will be referenced so the caller must free it.
1279 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
1285 * ifp may be specified by sockaddr_dl
1286 * when protocol address is ambiguous.
1288 if (info->rti_ifp == NULL && ifpaddr != NULL &&
1289 ifpaddr->sa_family == AF_LINK &&
1290 (ifa = ifa_ifwithnet(ifpaddr, 0, fibnum)) != NULL) {
1291 info->rti_ifp = ifa->ifa_ifp;
1294 if (info->rti_ifa == NULL && ifaaddr != NULL)
1295 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
1296 if (info->rti_ifa == NULL) {
1297 struct sockaddr *sa;
1299 sa = ifaaddr != NULL ? ifaaddr :
1300 (gateway != NULL ? gateway : dst);
1301 if (sa != NULL && info->rti_ifp != NULL)
1302 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
1303 else if (dst != NULL && gateway != NULL)
1304 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway,
1306 else if (sa != NULL)
1307 info->rti_ifa = ifa_ifwithroute(flags, sa, sa,
1310 if ((ifa = info->rti_ifa) != NULL) {
1311 if (info->rti_ifp == NULL)
1312 info->rti_ifp = ifa->ifa_ifp;
1314 error = ENETUNREACH;
1319 if_updatemtu_cb(struct radix_node *rn, void *arg)
1322 struct if_mtuinfo *ifmtu;
1324 rt = (struct rtentry *)rn;
1325 ifmtu = (struct if_mtuinfo *)arg;
1327 if (rt->rt_ifp != ifmtu->ifp)
1330 if (rt->rt_mtu >= ifmtu->mtu) {
1331 /* We have to decrease mtu regardless of flags */
1332 rt->rt_mtu = ifmtu->mtu;
1337 * New MTU is bigger. Check if are allowed to alter it
1339 if ((rt->rt_flags & (RTF_FIXEDMTU | RTF_GATEWAY | RTF_HOST)) != 0) {
1342 * Skip routes with user-supplied MTU and
1343 * non-interface routes
1348 /* We are safe to update route MTU */
1349 rt->rt_mtu = ifmtu->mtu;
1355 rt_updatemtu(struct ifnet *ifp)
1357 struct if_mtuinfo ifmtu;
1358 struct radix_node_head *rnh;
1364 * Try to update rt_mtu for all routes using this interface
1365 * Unfortunately the only way to do this is to traverse all
1366 * routing tables in all fibs/domains.
1368 for (i = 1; i <= AF_MAX; i++) {
1369 ifmtu.mtu = if_getmtu_family(ifp, i);
1370 for (j = 0; j < rt_numfibs; j++) {
1371 rnh = rt_tables_get_rnh(j, i);
1374 RADIX_NODE_HEAD_LOCK(rnh);
1375 rnh->rnh_walktree(rnh, if_updatemtu_cb, &ifmtu);
1376 RADIX_NODE_HEAD_UNLOCK(rnh);
1383 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
1384 int rt_print(char *buf, int buflen, struct rtentry *rt);
1387 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
1391 switch (s->sa_family) {
1393 paddr = &((struct sockaddr_in *)s)->sin_addr;
1396 paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
1403 if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
1406 return (strlen(buf));
1410 rt_print(char *buf, int buflen, struct rtentry *rt)
1412 struct sockaddr *addr, *mask;
1418 i = p_sockaddr(buf, buflen, addr);
1419 if (!(rt->rt_flags & RTF_HOST)) {
1421 i += p_sockaddr(buf + i, buflen - i, mask);
1424 if (rt->rt_flags & RTF_GATEWAY) {
1426 i += p_sockaddr(buf + i, buflen - i, rt->rt_gateway);
1435 * Deletes key for single-path routes, unlinks rtentry with
1436 * gateway specified in @info from multi-path routes.
1438 * Returnes unlinked entry. In case of failure, returns NULL
1439 * and sets @perror to ESRCH.
1441 static struct radix_node *
1442 rt_mpath_unlink(struct radix_node_head *rnh, struct rt_addrinfo *info,
1443 struct rtentry *rto, int *perror)
1446 * if we got multipath routes, we require users to specify
1447 * a matching RTAX_GATEWAY.
1449 struct rtentry *rt; // *rto = NULL;
1450 struct radix_node *rn;
1451 struct sockaddr *gw;
1453 gw = info->rti_info[RTAX_GATEWAY];
1454 rt = rt_mpath_matchgate(rto, gw);
1461 * this is the first entry in the chain
1464 rn = rn_mpath_next((struct radix_node *)rt);
1466 * there is another entry, now it's active
1471 rto->rt_flags |= RTF_UP;
1473 } else if (rt->rt_flags & RTF_GATEWAY) {
1475 * For gateway routes, we need to
1476 * make sure that we we are deleting
1477 * the correct gateway.
1478 * rt_mpath_matchgate() does not
1479 * check the case when there is only
1480 * one route in the chain.
1483 (rt->rt_gateway->sa_len != gw->sa_len ||
1484 memcmp(rt->rt_gateway, gw, gw->sa_len))) {
1491 * use the normal delete code to remove
1494 rn = rnh->rnh_deladdr(dst, netmask, rnh);
1500 * if the entry is 2nd and on up
1502 if (rt_mpath_deldup(rto, rt) == 0)
1503 panic ("rtrequest1: rt_mpath_deldup");
1505 rn = (struct radix_node *)rt;
1511 static struct rtentry *
1512 rt_flowtable_check_route(struct radix_node_head *rnh, struct rt_addrinfo *info)
1514 #if defined(INET6) || defined(INET)
1515 struct radix_node *rn;
1517 struct rtentry *rt0;
1520 /* "flow-table" only supports IPv6 and IPv4 at the moment. */
1521 switch (dst->sa_family) {
1528 #if defined(INET6) || defined(INET)
1529 rn = rnh->rnh_matchaddr(dst, rnh);
1530 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
1531 struct sockaddr *mask;
1536 * compare mask to see if the new route is
1537 * more specific than the existing one
1544 * A host route is already present, so
1545 * leave the flow-table entries as is.
1547 if (rt0->rt_flags & RTF_HOST) {
1550 } else if (!(flags & RTF_HOST) && netmask) {
1551 mask = rt_mask(rt0);
1554 n = (u_char *)netmask;
1561 if (len == 0 || (*n < *m)) {
1567 #endif/* INET6 || INET */
1575 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt,
1578 int error = 0, needlock = 0;
1579 struct rtentry *rt, *rt_old;
1581 struct rtentry *rt0;
1583 struct radix_node *rn;
1584 struct radix_node_head *rnh;
1586 struct sockaddr *ndst;
1587 struct sockaddr_storage mdst;
1588 #define senderr(x) { error = x ; goto bad; }
1590 KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum"));
1591 switch (dst->sa_family) {
1594 /* We support multiple FIBs. */
1597 fibnum = RT_DEFAULT_FIB;
1602 * Find the correct routing tree to use for this Address Family
1604 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
1606 return (EAFNOSUPPORT);
1607 needlock = ((flags & RTF_RNH_LOCKED) == 0);
1608 flags &= ~RTF_RNH_LOCKED;
1610 RADIX_NODE_HEAD_LOCK(rnh);
1612 RADIX_NODE_HEAD_LOCK_ASSERT(rnh);
1614 * If we are adding a host route then we don't want to put
1615 * a netmask in the tree, nor do we want to clone it.
1617 if (flags & RTF_HOST)
1623 rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask);
1624 dst = (struct sockaddr *)&mdst;
1627 rt = rt_unlinkrte(rnh, info, &error);
1631 rt_notifydelete(rt, info);
1634 * If the caller wants it, then it can have it,
1635 * but it's up to it to free the rtentry as we won't be
1646 * resolve was only used for route cloning
1651 if ((flags & RTF_GATEWAY) && !gateway)
1653 if (dst && gateway && (dst->sa_family != gateway->sa_family) &&
1654 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
1657 if (info->rti_ifa == NULL) {
1658 error = rt_getifa_fib(info, fibnum);
1662 ifa_ref(info->rti_ifa);
1663 ifa = info->rti_ifa;
1664 rt = uma_zalloc(V_rtzone, M_NOWAIT);
1669 rt->rt_flags = RTF_UP | flags;
1670 rt->rt_fibnum = fibnum;
1672 * Add the gateway. Possibly re-malloc-ing the storage for it.
1675 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
1677 uma_zfree(V_rtzone, rt);
1682 * point to the (possibly newly malloc'd) dest address.
1684 ndst = (struct sockaddr *)rt_key(rt);
1687 * make sure it contains the value we want (masked if needed).
1690 rt_maskedcopy(dst, ndst, netmask);
1692 bcopy(dst, ndst, dst->sa_len);
1695 * We use the ifa reference returned by rt_getifa_fib().
1696 * This moved from below so that rnh->rnh_addaddr() can
1697 * examine the ifa and ifa->ifa_ifp if it so desires.
1700 rt->rt_ifp = ifa->ifa_ifp;
1703 rt_setmetrics(info, rt);
1706 /* do not permit exactly the same dst/mask/gw pair */
1707 if (rn_mpath_capable(rnh) &&
1708 rt_mpath_conflict(rnh, rt, netmask)) {
1709 ifa_free(rt->rt_ifa);
1711 uma_zfree(V_rtzone, rt);
1717 rt0 = rt_flowtable_check_route(rnh, info);
1718 #endif /* FLOWTABLE */
1720 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
1721 rn = rnh->rnh_addaddr(ndst, netmask, rnh, rt->rt_nodes);
1724 if (rn == NULL && (info->rti_flags & RTF_PINNED) != 0) {
1727 * Force removal and re-try addition
1728 * TODO: better multipath&pinned support
1730 struct sockaddr *info_dst = info->rti_info[RTAX_DST];
1731 info->rti_info[RTAX_DST] = ndst;
1732 /* Do not delete existing PINNED(interface) routes */
1733 info->rti_flags &= ~RTF_PINNED;
1734 rt_old = rt_unlinkrte(rnh, info, &error);
1735 info->rti_flags |= RTF_PINNED;
1736 info->rti_info[RTAX_DST] = info_dst;
1738 rn = rnh->rnh_addaddr(ndst, netmask, rnh,
1746 * If it still failed to go into the tree,
1747 * then un-make it (this should be a function)
1750 ifa_free(rt->rt_ifa);
1752 uma_zfree(V_rtzone, rt);
1760 else if (rt0 != NULL) {
1761 flowtable_route_flush(dst->sa_family, rt0);
1766 if (rt_old != NULL) {
1767 rt_notifydelete(rt_old, info);
1772 * If this protocol has something to add to this then
1773 * allow it to do that as well.
1775 if (ifa->ifa_rtrequest)
1776 ifa->ifa_rtrequest(req, rt, info);
1779 * actually return a resultant rtentry and
1780 * give the caller a single reference.
1789 error = rtrequest1_fib_change(rnh, info, ret_nrt, fibnum);
1796 RADIX_NODE_HEAD_UNLOCK(rnh);
1809 rtrequest1_fib_change(struct radix_node_head *rnh, struct rt_addrinfo *info,
1810 struct rtentry **ret_nrt, u_int fibnum)
1812 struct rtentry *rt = NULL;
1816 struct if_mtuinfo ifmtu;
1818 rt = (struct rtentry *)rnh->rnh_lookup(info->rti_info[RTAX_DST],
1819 info->rti_info[RTAX_NETMASK], rnh);
1826 * If we got multipath routes,
1827 * we require users to specify a matching RTAX_GATEWAY.
1829 if (rn_mpath_capable(rnh)) {
1830 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
1838 rt_setmetrics(info, rt);
1841 * New gateway could require new ifaddr, ifp;
1842 * flags may also be different; ifp may be specified
1843 * by ll sockaddr when protocol address is ambiguous
1845 if (((rt->rt_flags & RTF_GATEWAY) &&
1846 info->rti_info[RTAX_GATEWAY] != NULL) ||
1847 info->rti_info[RTAX_IFP] != NULL ||
1848 (info->rti_info[RTAX_IFA] != NULL &&
1849 !sa_equal(info->rti_info[RTAX_IFA], rt->rt_ifa->ifa_addr))) {
1851 error = rt_getifa_fib(info, fibnum);
1852 if (info->rti_ifa != NULL)
1859 /* Check if outgoing interface has changed */
1860 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa &&
1861 rt->rt_ifa != NULL && rt->rt_ifa->ifa_rtrequest != NULL) {
1862 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1863 ifa_free(rt->rt_ifa);
1865 /* Update gateway address */
1866 if (info->rti_info[RTAX_GATEWAY] != NULL) {
1867 error = rt_setgate(rt, rt_key(rt), info->rti_info[RTAX_GATEWAY]);
1871 rt->rt_flags &= ~RTF_GATEWAY;
1872 rt->rt_flags |= (RTF_GATEWAY & info->rti_flags);
1875 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa) {
1876 ifa_ref(info->rti_ifa);
1877 rt->rt_ifa = info->rti_ifa;
1878 rt->rt_ifp = info->rti_ifp;
1880 /* Allow some flags to be toggled on change. */
1881 rt->rt_flags &= ~RTF_FMASK;
1882 rt->rt_flags |= info->rti_flags & RTF_FMASK;
1884 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest != NULL)
1885 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
1887 /* Alter route MTU if necessary */
1888 if (rt->rt_ifp != NULL) {
1889 family = info->rti_info[RTAX_DST]->sa_family;
1890 mtu = if_getmtu_family(rt->rt_ifp, family);
1891 /* Set default MTU */
1892 if (rt->rt_mtu == 0)
1894 if (rt->rt_mtu != mtu) {
1895 /* Check if we really need to update */
1896 ifmtu.ifp = rt->rt_ifp;
1898 if_updatemtu_cb(rt->rt_nodes, &ifmtu);
1909 ifa_free(info->rti_ifa);
1914 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt)
1917 if (info->rti_mflags & RTV_MTU) {
1918 if (info->rti_rmx->rmx_mtu != 0) {
1921 * MTU was explicitly provided by user.
1924 rt->rt_flags |= RTF_FIXEDMTU;
1928 * User explicitly sets MTU to 0.
1929 * Assume rollback to default.
1931 rt->rt_flags &= ~RTF_FIXEDMTU;
1933 rt->rt_mtu = info->rti_rmx->rmx_mtu;
1935 if (info->rti_mflags & RTV_WEIGHT)
1936 rt->rt_weight = info->rti_rmx->rmx_weight;
1937 /* Kernel -> userland timebase conversion. */
1938 if (info->rti_mflags & RTV_EXPIRE)
1939 rt->rt_expire = info->rti_rmx->rmx_expire ?
1940 info->rti_rmx->rmx_expire - time_second + time_uptime : 0;
1944 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
1946 /* XXX dst may be overwritten, can we move this to below */
1947 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
1949 struct radix_node_head *rnh;
1951 rnh = rt_tables_get_rnh(rt->rt_fibnum, dst->sa_family);
1955 RADIX_NODE_HEAD_LOCK_ASSERT(rnh);
1958 * Prepare to store the gateway in rt->rt_gateway.
1959 * Both dst and gateway are stored one after the other in the same
1960 * malloc'd chunk. If we have room, we can reuse the old buffer,
1961 * rt_gateway already points to the right place.
1962 * Otherwise, malloc a new block and update the 'dst' address.
1964 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
1967 R_Malloc(new, caddr_t, dlen + glen);
1971 * XXX note, we copy from *dst and not *rt_key(rt) because
1972 * rt_setgate() can be called to initialize a newly
1973 * allocated route entry, in which case rt_key(rt) == NULL
1974 * (and also rt->rt_gateway == NULL).
1975 * Free()/free() handle a NULL argument just fine.
1977 bcopy(dst, new, dlen);
1978 R_Free(rt_key(rt)); /* free old block, if any */
1979 rt_key(rt) = (struct sockaddr *)new;
1980 rt->rt_gateway = (struct sockaddr *)(new + dlen);
1984 * Copy the new gateway value into the memory chunk.
1986 bcopy(gate, rt->rt_gateway, glen);
1992 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
1994 u_char *cp1 = (u_char *)src;
1995 u_char *cp2 = (u_char *)dst;
1996 u_char *cp3 = (u_char *)netmask;
1997 u_char *cplim = cp2 + *cp3;
1998 u_char *cplim2 = cp2 + *cp1;
2000 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
2005 *cp2++ = *cp1++ & *cp3++;
2007 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
2011 * Set up a routing table entry, normally
2014 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
2016 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
2018 struct sockaddr *dst;
2019 struct sockaddr *netmask;
2020 struct rtentry *rt = NULL;
2021 struct rt_addrinfo info;
2023 int startfib, endfib;
2024 char tempbuf[_SOCKADDR_TMPSIZE];
2027 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
2028 struct radix_node_head *rnh;
2030 if (flags & RTF_HOST) {
2031 dst = ifa->ifa_dstaddr;
2034 dst = ifa->ifa_addr;
2035 netmask = ifa->ifa_netmask;
2037 if (dst->sa_len == 0)
2039 switch (dst->sa_family) {
2042 /* We support multiple FIBs. */
2045 fibnum = RT_DEFAULT_FIB;
2048 if (fibnum == RT_ALL_FIBS) {
2049 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD)
2050 startfib = endfib = ifa->ifa_ifp->if_fib;
2053 endfib = rt_numfibs - 1;
2056 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
2062 * If it's a delete, check that if it exists,
2063 * it's on the correct interface or we might scrub
2064 * a route to another ifa which would
2065 * be confusing at best and possibly worse.
2067 if (cmd == RTM_DELETE) {
2069 * It's a delete, so it should already exist..
2070 * If it's a net, mask off the host bits
2071 * (Assuming we have a mask)
2072 * XXX this is kinda inet specific..
2074 if (netmask != NULL) {
2075 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
2076 dst = (struct sockaddr *)tempbuf;
2080 * Now go through all the requested tables (fibs) and do the
2081 * requested action. Realistically, this will either be fib 0
2082 * for protocols that don't do multiple tables or all the
2083 * tables for those that do.
2085 for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
2086 if (cmd == RTM_DELETE) {
2087 struct radix_node *rn;
2089 * Look up an rtentry that is in the routing tree and
2090 * contains the correct info.
2092 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
2094 /* this table doesn't exist but others might */
2096 RADIX_NODE_HEAD_RLOCK(rnh);
2097 rn = rnh->rnh_lookup(dst, netmask, rnh);
2099 if (rn_mpath_capable(rnh)) {
2106 * for interface route the
2107 * rt->rt_gateway is sockaddr_intf
2108 * for cloning ARP entries, so
2109 * rt_mpath_matchgate must use the
2112 rt = rt_mpath_matchgate(rt,
2119 error = (rn == NULL ||
2120 (rn->rn_flags & RNF_ROOT) ||
2121 RNTORT(rn)->rt_ifa != ifa);
2122 RADIX_NODE_HEAD_RUNLOCK(rnh);
2124 /* this is only an error if bad on ALL tables */
2129 * Do the actual request
2131 bzero((caddr_t)&info, sizeof(info));
2133 info.rti_flags = flags |
2134 (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED;
2135 info.rti_info[RTAX_DST] = dst;
2137 * doing this for compatibility reasons
2140 info.rti_info[RTAX_GATEWAY] =
2141 (struct sockaddr *)&null_sdl;
2143 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
2144 info.rti_info[RTAX_NETMASK] = netmask;
2145 error = rtrequest1_fib(cmd, &info, &rt, fibnum);
2147 if (error == 0 && rt != NULL) {
2149 * notify any listening routing agents of the change
2154 * in case address alias finds the first address
2155 * e.g. ifconfig bge0 192.0.2.246/24
2156 * e.g. ifconfig bge0 192.0.2.247/24
2157 * the address set in the route is 192.0.2.246
2158 * so we need to replace it with 192.0.2.247
2160 if (memcmp(rt->rt_ifa->ifa_addr,
2161 ifa->ifa_addr, ifa->ifa_addr->sa_len)) {
2162 ifa_free(rt->rt_ifa);
2164 rt->rt_ifp = ifa->ifa_ifp;
2169 * doing this for compatibility reasons
2171 if (cmd == RTM_ADD) {
2172 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
2173 rt->rt_ifp->if_type;
2174 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
2175 rt->rt_ifp->if_index;
2179 rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum);
2182 if (cmd == RTM_DELETE) {
2184 * If we are deleting, and we found an entry,
2185 * then it's been removed from the tree..
2186 * now throw it away.
2190 if (cmd == RTM_ADD) {
2192 * We just wanted to add it..
2193 * we don't actually need a reference.
2204 if (cmd == RTM_DELETE) {
2208 /* we only give an error if it wasn't in any table */
2209 error = ((flags & RTF_HOST) ?
2210 EHOSTUNREACH : ENETUNREACH);
2214 /* return an error if any of them failed */
2222 * Set up a routing table entry, normally
2226 rtinit(struct ifaddr *ifa, int cmd, int flags)
2228 struct sockaddr *dst;
2229 int fib = RT_DEFAULT_FIB;
2231 if (flags & RTF_HOST) {
2232 dst = ifa->ifa_dstaddr;
2234 dst = ifa->ifa_addr;
2237 switch (dst->sa_family) {
2240 /* We do support multiple FIBs. */
2244 return (rtinit1(ifa, cmd, flags, fib));
2248 * Announce interface address arrival/withdraw
2249 * Returns 0 on success.
2252 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
2255 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2256 ("unexpected cmd %d", cmd));
2258 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2259 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2261 #if defined(INET) || defined(INET6)
2264 * notify the SCTP stack
2265 * this will only get called when an address is added/deleted
2266 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
2268 sctp_addr_change(ifa, cmd);
2271 return (rtsock_addrmsg(cmd, ifa, fibnum));
2275 * Announce route addition/removal.
2276 * Users of this function MUST validate input data BEFORE calling.
2277 * However we have to be able to handle invalid data:
2278 * if some userland app sends us "invalid" route message (invalid mask,
2279 * no dst, wrong address families, etc...) we need to pass it back
2280 * to app (and any other rtsock consumers) with rtm_errno field set to
2282 * Returns 0 on success.
2285 rt_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
2289 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2290 ("unexpected cmd %d", cmd));
2292 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2293 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2295 KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
2297 return (rtsock_routemsg(cmd, ifp, error, rt, fibnum));
2301 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
2304 rt_newaddrmsg_fib(cmd, ifa, error, rt, RT_ALL_FIBS);
2308 * This is called to generate messages from the routing socket
2309 * indicating a network interface has had addresses associated with it.
2312 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
2316 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2317 ("unexpected cmd %u", cmd));
2318 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2319 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2321 if (cmd == RTM_ADD) {
2322 rt_addrmsg(cmd, ifa, fibnum);
2324 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2327 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2328 rt_addrmsg(cmd, ifa, fibnum);