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>
60 #include <net/route_var.h>
62 #include <net/flowtable.h>
65 #include <net/radix_mpath.h>
68 #include <netinet/in.h>
69 #include <netinet/ip_mroute.h>
73 #define RT_MAXFIBS UINT16_MAX
75 /* Kernel config default option. */
78 #error "ROUTETABLES defined too low"
80 #if ROUTETABLES > RT_MAXFIBS
81 #error "ROUTETABLES defined too big"
83 #define RT_NUMFIBS ROUTETABLES
84 #endif /* ROUTETABLES */
85 /* Initialize to default if not otherwise set. */
90 #if defined(INET) || defined(INET6)
92 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
97 /* This is read-only.. */
98 u_int rt_numfibs = RT_NUMFIBS;
99 SYSCTL_UINT(_net, OID_AUTO, fibs, CTLFLAG_RDTUN, &rt_numfibs, 0, "");
102 * By default add routes to all fibs for new interfaces.
103 * Once this is set to 0 then only allocate routes on interface
104 * changes for the FIB of the caller when adding a new set of addresses
105 * to an interface. XXX this is a shotgun aproach to a problem that needs
106 * a more fine grained solution.. that will come.
107 * XXX also has the problems getting the FIB from curthread which will not
108 * always work given the fib can be overridden and prefixes can be added
109 * from the network stack context.
111 VNET_DEFINE(u_int, rt_add_addr_allfibs) = 1;
112 SYSCTL_UINT(_net, OID_AUTO, add_addr_allfibs, CTLFLAG_RWTUN | CTLFLAG_VNET,
113 &VNET_NAME(rt_add_addr_allfibs), 0, "");
115 VNET_DEFINE(struct rtstat, rtstat);
116 #define V_rtstat VNET(rtstat)
118 VNET_DEFINE(struct rib_head *, rt_tables);
119 #define V_rt_tables VNET(rt_tables)
121 VNET_DEFINE(int, rttrash); /* routes not in table but not freed */
122 #define V_rttrash VNET(rttrash)
126 * Convert a 'struct radix_node *' to a 'struct rtentry *'.
127 * The operation can be done safely (in this code) because a
128 * 'struct rtentry' starts with two 'struct radix_node''s, the first
129 * one representing leaf nodes in the routing tree, which is
130 * what the code in radix.c passes us as a 'struct radix_node'.
132 * But because there are a lot of assumptions in this conversion,
133 * do not cast explicitly, but always use the macro below.
135 #define RNTORT(p) ((struct rtentry *)(p))
137 static VNET_DEFINE(uma_zone_t, rtzone); /* Routing table UMA zone. */
138 #define V_rtzone VNET(rtzone)
140 static int rtrequest1_fib_change(struct rib_head *, struct rt_addrinfo *,
141 struct rtentry **, u_int);
142 static void rt_setmetrics(const struct rt_addrinfo *, struct rtentry *);
143 static int rt_ifdelroute(const struct rtentry *rt, void *arg);
144 static struct rtentry *rt_unlinkrte(struct rib_head *rnh,
145 struct rt_addrinfo *info, int *perror);
146 static void rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info);
148 static struct radix_node *rt_mpath_unlink(struct rib_head *rnh,
149 struct rt_addrinfo *info, struct rtentry *rto, int *perror);
151 static int rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info,
160 static int if_updatemtu_cb(struct radix_node *, void *);
163 * handler for net.my_fibnum
166 sysctl_my_fibnum(SYSCTL_HANDLER_ARGS)
171 fibnum = curthread->td_proc->p_fibnum;
172 error = sysctl_handle_int(oidp, &fibnum, 0, req);
176 SYSCTL_PROC(_net, OID_AUTO, my_fibnum, CTLTYPE_INT|CTLFLAG_RD,
177 NULL, 0, &sysctl_my_fibnum, "I", "default FIB of caller");
179 static __inline struct rib_head **
180 rt_tables_get_rnh_ptr(int table, int fam)
182 struct rib_head **rnh;
184 KASSERT(table >= 0 && table < rt_numfibs, ("%s: table out of bounds.",
186 KASSERT(fam >= 0 && fam < (AF_MAX+1), ("%s: fam out of bounds.",
189 /* rnh is [fib=0][af=0]. */
190 rnh = (struct rib_head **)V_rt_tables;
191 /* Get the offset to the requested table and fam. */
192 rnh += table * (AF_MAX+1) + fam;
198 rt_tables_get_rnh(int table, int fam)
201 return (*rt_tables_get_rnh_ptr(table, fam));
205 * route initialization must occur before ip6_init2(), which happenas at
212 /* whack the tunable ints into line. */
213 if (rt_numfibs > RT_MAXFIBS)
214 rt_numfibs = RT_MAXFIBS;
218 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0);
221 rtentry_zinit(void *mem, int size, int how)
223 struct rtentry *rt = mem;
225 rt->rt_pksent = counter_u64_alloc(how);
226 if (rt->rt_pksent == NULL)
235 rtentry_zfini(void *mem, int size)
237 struct rtentry *rt = mem;
240 counter_u64_free(rt->rt_pksent);
244 rtentry_ctor(void *mem, int size, void *arg, int how)
246 struct rtentry *rt = mem;
248 bzero(rt, offsetof(struct rtentry, rt_endzero));
249 counter_u64_zero(rt->rt_pksent);
256 rtentry_dtor(void *mem, int size, void *arg)
258 struct rtentry *rt = mem;
264 vnet_route_init(const void *unused __unused)
267 struct rib_head **rnh;
271 V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) *
272 sizeof(struct rib_head *), M_RTABLE, M_WAITOK|M_ZERO);
274 V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry),
275 rtentry_ctor, rtentry_dtor,
276 rtentry_zinit, rtentry_zfini, UMA_ALIGN_PTR, 0);
277 for (dom = domains; dom; dom = dom->dom_next) {
278 if (dom->dom_rtattach == NULL)
281 for (table = 0; table < rt_numfibs; table++) {
282 fam = dom->dom_family;
283 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
286 rnh = rt_tables_get_rnh_ptr(table, fam);
288 panic("%s: rnh NULL", __func__);
289 dom->dom_rtattach((void **)rnh, 0);
293 VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
298 vnet_route_uninit(const void *unused __unused)
303 struct rib_head **rnh;
305 for (dom = domains; dom; dom = dom->dom_next) {
306 if (dom->dom_rtdetach == NULL)
309 for (table = 0; table < rt_numfibs; table++) {
310 fam = dom->dom_family;
312 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
315 rnh = rt_tables_get_rnh_ptr(table, fam);
317 panic("%s: rnh NULL", __func__);
318 dom->dom_rtdetach((void **)rnh, 0);
322 free(V_rt_tables, M_RTABLE);
323 uma_zdestroy(V_rtzone);
325 VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
326 vnet_route_uninit, 0);
330 rt_table_init(int offset)
334 rh = malloc(sizeof(struct rib_head), M_RTABLE, M_WAITOK | M_ZERO);
336 /* TODO: These details should be hidded inside radix.c */
337 /* Init masks tree */
338 rn_inithead_internal(&rh->head, rh->rnh_nodes, offset);
339 rn_inithead_internal(&rh->rmhead.head, rh->rmhead.mask_nodes, 0);
340 rh->head.rnh_masks = &rh->rmhead;
343 rw_init(&rh->rib_lock, "rib head lock");
345 /* Finally, set base callbacks */
346 rh->rnh_addaddr = rn_addroute;
347 rh->rnh_deladdr = rn_delete;
348 rh->rnh_matchaddr = rn_match;
349 rh->rnh_lookup = rn_lookup;
350 rh->rnh_walktree = rn_walktree;
351 rh->rnh_walktree_from = rn_walktree_from;
357 rt_freeentry(struct radix_node *rn, void *arg)
359 struct radix_head * const rnh = arg;
360 struct radix_node *x;
362 x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh);
369 rt_table_destroy(struct rib_head *rh)
372 rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head);
374 /* Assume table is already empty */
375 rw_destroy(&rh->rib_lock);
380 #ifndef _SYS_SYSPROTO_H_
386 sys_setfib(struct thread *td, struct setfib_args *uap)
388 if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs)
390 td->td_proc->p_fibnum = uap->fibnum;
395 * Packet routing routines.
398 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum)
402 if ((rt = ro->ro_rt) != NULL) {
403 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
408 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum);
410 RT_UNLOCK(ro->ro_rt);
414 * Look up the route that matches the address given
415 * Or, at least try.. Create a cloned route if needed.
417 * The returned route, if any, is locked.
420 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
423 return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB));
427 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags,
431 struct radix_node *rn;
432 struct rtentry *newrt;
433 struct rt_addrinfo info;
434 int err = 0, msgtype = RTM_MISS;
436 KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum"));
437 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
443 * Look up the address in the table for that Address Family
446 rn = rh->rnh_matchaddr(dst, &rh->head);
447 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
458 * Either we hit the root or couldn't find any match,
459 * Which basically means
460 * "caint get there frm here"
463 V_rtstat.rts_unreach++;
467 * If required, report the failure to the supervising
469 * For a delete, this is not an error. (report == 0)
471 bzero(&info, sizeof(info));
472 info.rti_info[RTAX_DST] = dst;
473 rt_missmsg_fib(msgtype, &info, 0, err, fibnum);
479 * Remove a reference count from an rtentry.
480 * If the count gets low enough, take it out of the routing table
483 rtfree(struct rtentry *rt)
485 struct rib_head *rnh;
487 KASSERT(rt != NULL,("%s: NULL rt", __func__));
488 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
489 KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
494 * The callers should use RTFREE_LOCKED() or RTFREE(), so
495 * we should come here exactly with the last reference.
498 if (rt->rt_refcnt > 0) {
499 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt);
504 * On last reference give the "close method" a chance
505 * to cleanup private state. This also permits (for
506 * IPv4 and IPv6) a chance to decide if the routing table
507 * entry should be purged immediately or at a later time.
508 * When an immediate purge is to happen the close routine
509 * typically calls rtexpunge which clears the RTF_UP flag
510 * on the entry so that the code below reclaims the storage.
512 if (rt->rt_refcnt == 0 && rnh->rnh_close)
513 rnh->rnh_close((struct radix_node *)rt, &rnh->head);
516 * If we are no longer "up" (and ref == 0)
517 * then we can free the resources associated
520 if ((rt->rt_flags & RTF_UP) == 0) {
521 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
524 * the rtentry must have been removed from the routing table
525 * so it is represented in rttrash.. remove that now.
529 if (rt->rt_refcnt < 0) {
530 printf("rtfree: %p not freed (neg refs)\n", rt);
535 * release references on items we hold them on..
536 * e.g other routes and ifaddrs.
539 ifa_free(rt->rt_ifa);
541 * The key is separatly alloc'd so free it (see rt_setgate()).
542 * This also frees the gateway, as they are always malloc'd
548 * and the rtentry itself of course
550 uma_zfree(V_rtzone, rt);
559 * Force a routing table entry to the specified
560 * destination to go through the given gateway.
561 * Normally called as a result of a routing redirect
562 * message from the network layer.
565 rtredirect_fib(struct sockaddr *dst,
566 struct sockaddr *gateway,
567 struct sockaddr *netmask,
569 struct sockaddr *src,
575 struct rt_addrinfo info;
577 struct rib_head *rnh;
580 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
582 error = EAFNOSUPPORT;
586 /* verify the gateway is directly reachable */
587 if ((ifa = ifa_ifwithnet(gateway, 0, fibnum)) == NULL) {
591 rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */
593 * If the redirect isn't from our current router for this dst,
594 * it's either old or wrong. If it redirects us to ourselves,
595 * we have a routing loop, perhaps as a result of an interface
596 * going down recently.
598 if (!(flags & RTF_DONE) && rt) {
599 if (!sa_equal(src, rt->rt_gateway)) {
603 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) {
608 if ((flags & RTF_GATEWAY) && ifa_ifwithaddr_check(gateway)) {
609 error = EHOSTUNREACH;
613 * Create a new entry if we just got back a wildcard entry
614 * or the lookup failed. This is necessary for hosts
615 * which use routing redirects generated by smart gateways
616 * to dynamically build the routing tables.
618 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
621 * Don't listen to the redirect if it's
622 * for a route to an interface.
624 if (rt->rt_flags & RTF_GATEWAY) {
625 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
627 * Changing from route to net => route to host.
628 * Create new route, rather than smashing route to net.
634 flags |= RTF_DYNAMIC;
635 bzero((caddr_t)&info, sizeof(info));
636 info.rti_info[RTAX_DST] = dst;
637 info.rti_info[RTAX_GATEWAY] = gateway;
638 info.rti_info[RTAX_NETMASK] = netmask;
640 info.rti_flags = flags;
641 error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum);
644 flags = rt->rt_flags;
647 stat = &V_rtstat.rts_dynamic;
651 * Smash the current notion of the gateway to
652 * this destination. Should check about netmask!!!
654 if ((flags & RTF_GATEWAY) == 0)
655 rt->rt_flags &= ~RTF_GATEWAY;
656 rt->rt_flags |= RTF_MODIFIED;
657 flags |= RTF_MODIFIED;
658 stat = &V_rtstat.rts_newgateway;
660 * add the key and gateway (in one malloc'd chunk).
665 rt_setgate(rt, rt_key(rt), gateway);
669 error = EHOSTUNREACH;
675 V_rtstat.rts_badredirect++;
676 else if (stat != NULL)
678 bzero((caddr_t)&info, sizeof(info));
679 info.rti_info[RTAX_DST] = dst;
680 info.rti_info[RTAX_GATEWAY] = gateway;
681 info.rti_info[RTAX_NETMASK] = netmask;
682 info.rti_info[RTAX_AUTHOR] = src;
683 rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum);
689 * Routing table ioctl interface.
692 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
696 * If more ioctl commands are added here, make sure the proper
697 * super-user checks are being performed because it is possible for
698 * prison-root to make it this far if raw sockets have been enabled
702 /* Multicast goop, grrr... */
703 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
710 ifa_ifwithroute(int flags, const struct sockaddr *dst, struct sockaddr *gateway,
716 if ((flags & RTF_GATEWAY) == 0) {
718 * If we are adding a route to an interface,
719 * and the interface is a pt to pt link
720 * we should search for the destination
721 * as our clue to the interface. Otherwise
722 * we can use the local address.
725 if (flags & RTF_HOST)
726 ifa = ifa_ifwithdstaddr(dst, fibnum);
728 ifa = ifa_ifwithaddr(gateway);
731 * If we are adding a route to a remote net
732 * or host, the gateway may still be on the
733 * other end of a pt to pt link.
735 ifa = ifa_ifwithdstaddr(gateway, fibnum);
738 ifa = ifa_ifwithnet(gateway, 0, fibnum);
740 struct rtentry *rt = rtalloc1_fib(gateway, 0, 0, fibnum);
744 * dismiss a gateway that is reachable only
745 * through the default router
747 switch (gateway->sa_family) {
749 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
753 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
759 if (!not_found && rt->rt_ifa != NULL) {
765 if (not_found || ifa == NULL)
768 if (ifa->ifa_addr->sa_family != dst->sa_family) {
769 struct ifaddr *oifa = ifa;
770 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
780 * Do appropriate manipulations of a routing tree given
781 * all the bits of info needed
784 rtrequest_fib(int req,
785 struct sockaddr *dst,
786 struct sockaddr *gateway,
787 struct sockaddr *netmask,
789 struct rtentry **ret_nrt,
792 struct rt_addrinfo info;
794 if (dst->sa_len == 0)
797 bzero((caddr_t)&info, sizeof(info));
798 info.rti_flags = flags;
799 info.rti_info[RTAX_DST] = dst;
800 info.rti_info[RTAX_GATEWAY] = gateway;
801 info.rti_info[RTAX_NETMASK] = netmask;
802 return rtrequest1_fib(req, &info, ret_nrt, fibnum);
807 * Copy most of @rt data into @info.
809 * If @flags contains NHR_COPY, copies dst,netmask and gw to the
810 * pointers specified by @info structure. Assume such pointers
811 * are zeroed sockaddr-like structures with sa_len field initialized
812 * to reflect size of the provided buffer. if no NHR_COPY is specified,
813 * point dst,netmask and gw @info fields to appropriate @rt values.
815 * if @flags contains NHR_REF, do refcouting on rt_ifp.
817 * Returns 0 on success.
820 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags)
822 struct rt_metrics *rmx;
823 struct sockaddr *src, *dst;
826 if (flags & NHR_COPY) {
827 /* Copy destination if dst is non-zero */
829 dst = info->rti_info[RTAX_DST];
830 sa_len = src->sa_len;
832 if (src->sa_len > dst->sa_len)
834 memcpy(dst, src, src->sa_len);
835 info->rti_addrs |= RTA_DST;
838 /* Copy mask if set && dst is non-zero */
840 dst = info->rti_info[RTAX_NETMASK];
841 if (src != NULL && dst != NULL) {
844 * Radix stores different value in sa_len,
845 * assume rt_mask() to have the same length
848 if (sa_len > dst->sa_len)
850 memcpy(dst, src, src->sa_len);
851 info->rti_addrs |= RTA_NETMASK;
854 /* Copy gateway is set && dst is non-zero */
855 src = rt->rt_gateway;
856 dst = info->rti_info[RTAX_GATEWAY];
857 if ((rt->rt_flags & RTF_GATEWAY) && src != NULL && dst != NULL){
858 if (src->sa_len > dst->sa_len)
860 memcpy(dst, src, src->sa_len);
861 info->rti_addrs |= RTA_GATEWAY;
864 info->rti_info[RTAX_DST] = rt_key(rt);
865 info->rti_addrs |= RTA_DST;
866 if (rt_mask(rt) != NULL) {
867 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
868 info->rti_addrs |= RTA_NETMASK;
870 if (rt->rt_flags & RTF_GATEWAY) {
871 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
872 info->rti_addrs |= RTA_GATEWAY;
878 info->rti_mflags |= RTV_MTU;
879 rmx->rmx_mtu = rt->rt_mtu;
882 info->rti_flags = rt->rt_flags;
883 info->rti_ifp = rt->rt_ifp;
884 info->rti_ifa = rt->rt_ifa;
886 if (flags & NHR_REF) {
887 /* Do 'traditional' refcouting */
888 if_ref(info->rti_ifp);
895 * Lookups up route entry for @dst in RIB database for fib @fibnum.
896 * Exports entry data to @info using rt_exportinfo().
898 * if @flags contains NHR_REF, refcouting is performed on rt_ifp.
899 * All references can be released later by calling rib_free_info()
901 * Returns 0 on success.
902 * Returns ENOENT for lookup failure, ENOMEM for export failure.
905 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
906 uint32_t flowid, struct rt_addrinfo *info)
909 struct radix_node *rn;
913 KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
914 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
919 rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head);
920 if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
922 /* Ensure route & ifp is UP */
923 if (RT_LINK_IS_UP(rt->rt_ifp)) {
924 flags = (flags & NHR_REF) | NHR_COPY;
925 error = rt_exportinfo(rt, info, flags);
937 * Releases all references acquired by rib_lookup_info() when
938 * called with NHR_REF flags.
941 rib_free_info(struct rt_addrinfo *info)
944 if_rele(info->rti_ifp);
948 * Iterates over all existing fibs in system calling
949 * @setwa_f function prior to traversing each fib.
950 * Calls @wa_f function for each element in current fib.
951 * If af is not AF_UNSPEC, iterates over fibs in particular
955 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f,
958 struct rib_head *rnh;
962 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
963 /* Do we want some specific family? */
964 if (af != AF_UNSPEC) {
965 rnh = rt_tables_get_rnh(fibnum, af);
969 setwa_f(rnh, fibnum, af, arg);
972 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
977 for (i = 1; i <= AF_MAX; i++) {
978 rnh = rt_tables_get_rnh(fibnum, i);
982 setwa_f(rnh, fibnum, i, arg);
985 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
993 struct rt_addrinfo info;
994 struct rib_head *rnh;
995 struct rtentry *head;
999 * Conditionally unlinks @rn from radix tree based
1000 * on info data passed in @arg.
1003 rt_checkdelroute(struct radix_node *rn, void *arg)
1005 struct rt_delinfo *di;
1006 struct rt_addrinfo *info;
1010 di = (struct rt_delinfo *)arg;
1011 rt = (struct rtentry *)rn;
1015 info->rti_info[RTAX_DST] = rt_key(rt);
1016 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
1017 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1019 rt = rt_unlinkrte(di->rnh, info, &error);
1021 /* Either not allowed or not matched. Skip entry */
1025 /* Entry was unlinked. Add to the list and return */
1026 rt->rt_chain = di->head;
1033 * Iterates over all existing fibs in system.
1034 * Deletes each element for which @filter_f function returned
1036 * If @af is not AF_UNSPEC, iterates over fibs in particular
1040 rt_foreach_fib_walk_del(int af, rt_filter_f_t *filter_f, void *arg)
1042 struct rib_head *rnh;
1043 struct rt_delinfo di;
1048 bzero(&di, sizeof(di));
1049 di.info.rti_filter = filter_f;
1050 di.info.rti_filterdata = arg;
1052 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1053 /* Do we want some specific family? */
1054 if (af != AF_UNSPEC) {
1062 for (i = start; i <= end; i++) {
1063 rnh = rt_tables_get_rnh(fibnum, i);
1069 rnh->rnh_walktree(&rnh->head, rt_checkdelroute, &di);
1072 if (di.head == NULL)
1075 /* We might have something to reclaim */
1076 while (di.head != NULL) {
1078 di.head = rt->rt_chain;
1079 rt->rt_chain = NULL;
1081 /* TODO std rt -> rt_addrinfo export */
1082 di.info.rti_info[RTAX_DST] = rt_key(rt);
1083 di.info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1085 rt_notifydelete(rt, &di.info);
1094 * Delete Routes for a Network Interface
1096 * Called for each routing entry via the rnh->rnh_walktree() call above
1097 * to delete all route entries referencing a detaching network interface.
1100 * rt pointer to rtentry
1101 * arg argument passed to rnh->rnh_walktree() - detaching interface
1105 * errno failed - reason indicated
1108 rt_ifdelroute(const struct rtentry *rt, void *arg)
1110 struct ifnet *ifp = arg;
1112 if (rt->rt_ifp != ifp)
1116 * Protect (sorta) against walktree recursion problems
1117 * with cloned routes
1119 if ((rt->rt_flags & RTF_UP) == 0)
1126 * Delete all remaining routes using this interface
1127 * Unfortuneatly the only way to do this is to slog through
1128 * the entire routing table looking for routes which point
1129 * to this interface...oh well...
1132 rt_flushifroutes(struct ifnet *ifp)
1135 rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
1139 * Conditionally unlinks rtentry matching data inside @info from @rnh.
1140 * Returns unlinked, locked and referenced @rtentry on success,
1141 * Returns NULL and sets @perror to:
1142 * ESRCH - if prefix was not found,
1143 * EADDRINUSE - if trying to delete PINNED route without appropriate flag.
1144 * ENOENT - if supplied filter function returned 0 (not matched).
1146 static struct rtentry *
1147 rt_unlinkrte(struct rib_head *rnh, struct rt_addrinfo *info, int *perror)
1149 struct sockaddr *dst, *netmask;
1151 struct radix_node *rn;
1153 dst = info->rti_info[RTAX_DST];
1154 netmask = info->rti_info[RTAX_NETMASK];
1156 rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, &rnh->head);
1162 if ((info->rti_flags & RTF_PINNED) == 0) {
1163 /* Check if target route can be deleted */
1164 if (rt->rt_flags & RTF_PINNED) {
1165 *perror = EADDRINUSE;
1170 if (info->rti_filter != NULL) {
1171 if (info->rti_filter(rt, info->rti_filterdata) == 0) {
1178 * Filter function requested rte deletion.
1179 * Ease the caller work by filling in remaining info
1180 * from that particular entry.
1182 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1186 * Remove the item from the tree and return it.
1187 * Complain if it is not there and do no more processing.
1191 if (rt_mpath_capable(rnh))
1192 rn = rt_mpath_unlink(rnh, info, rt, perror);
1195 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1199 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
1200 panic ("rtrequest delete");
1205 rt->rt_flags &= ~RTF_UP;
1213 rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info)
1218 * give the protocol a chance to keep things in sync.
1221 if (ifa != NULL && ifa->ifa_rtrequest != NULL)
1222 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1225 * One more rtentry floating around that is not
1226 * linked to the routing table. rttrash will be decremented
1227 * when RTFREE(rt) is eventually called.
1234 * These (questionable) definitions of apparent local variables apply
1235 * to the next two functions. XXXXXX!!!
1237 #define dst info->rti_info[RTAX_DST]
1238 #define gateway info->rti_info[RTAX_GATEWAY]
1239 #define netmask info->rti_info[RTAX_NETMASK]
1240 #define ifaaddr info->rti_info[RTAX_IFA]
1241 #define ifpaddr info->rti_info[RTAX_IFP]
1242 #define flags info->rti_flags
1245 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined,
1246 * it will be referenced so the caller must free it.
1249 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
1255 * ifp may be specified by sockaddr_dl
1256 * when protocol address is ambiguous.
1258 if (info->rti_ifp == NULL && ifpaddr != NULL &&
1259 ifpaddr->sa_family == AF_LINK &&
1260 (ifa = ifa_ifwithnet(ifpaddr, 0, fibnum)) != NULL) {
1261 info->rti_ifp = ifa->ifa_ifp;
1264 if (info->rti_ifa == NULL && ifaaddr != NULL)
1265 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
1266 if (info->rti_ifa == NULL) {
1267 struct sockaddr *sa;
1269 sa = ifaaddr != NULL ? ifaaddr :
1270 (gateway != NULL ? gateway : dst);
1271 if (sa != NULL && info->rti_ifp != NULL)
1272 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
1273 else if (dst != NULL && gateway != NULL)
1274 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway,
1276 else if (sa != NULL)
1277 info->rti_ifa = ifa_ifwithroute(flags, sa, sa,
1280 if ((ifa = info->rti_ifa) != NULL) {
1281 if (info->rti_ifp == NULL)
1282 info->rti_ifp = ifa->ifa_ifp;
1284 error = ENETUNREACH;
1289 if_updatemtu_cb(struct radix_node *rn, void *arg)
1292 struct if_mtuinfo *ifmtu;
1294 rt = (struct rtentry *)rn;
1295 ifmtu = (struct if_mtuinfo *)arg;
1297 if (rt->rt_ifp != ifmtu->ifp)
1300 if (rt->rt_mtu >= ifmtu->mtu) {
1301 /* We have to decrease mtu regardless of flags */
1302 rt->rt_mtu = ifmtu->mtu;
1307 * New MTU is bigger. Check if are allowed to alter it
1309 if ((rt->rt_flags & (RTF_FIXEDMTU | RTF_GATEWAY | RTF_HOST)) != 0) {
1312 * Skip routes with user-supplied MTU and
1313 * non-interface routes
1318 /* We are safe to update route MTU */
1319 rt->rt_mtu = ifmtu->mtu;
1325 rt_updatemtu(struct ifnet *ifp)
1327 struct if_mtuinfo ifmtu;
1328 struct rib_head *rnh;
1334 * Try to update rt_mtu for all routes using this interface
1335 * Unfortunately the only way to do this is to traverse all
1336 * routing tables in all fibs/domains.
1338 for (i = 1; i <= AF_MAX; i++) {
1339 ifmtu.mtu = if_getmtu_family(ifp, i);
1340 for (j = 0; j < rt_numfibs; j++) {
1341 rnh = rt_tables_get_rnh(j, i);
1345 rnh->rnh_walktree(&rnh->head, if_updatemtu_cb, &ifmtu);
1353 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
1354 int rt_print(char *buf, int buflen, struct rtentry *rt);
1357 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
1361 switch (s->sa_family) {
1363 paddr = &((struct sockaddr_in *)s)->sin_addr;
1366 paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
1373 if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
1376 return (strlen(buf));
1380 rt_print(char *buf, int buflen, struct rtentry *rt)
1382 struct sockaddr *addr, *mask;
1388 i = p_sockaddr(buf, buflen, addr);
1389 if (!(rt->rt_flags & RTF_HOST)) {
1391 i += p_sockaddr(buf + i, buflen - i, mask);
1394 if (rt->rt_flags & RTF_GATEWAY) {
1396 i += p_sockaddr(buf + i, buflen - i, rt->rt_gateway);
1405 * Deletes key for single-path routes, unlinks rtentry with
1406 * gateway specified in @info from multi-path routes.
1408 * Returnes unlinked entry. In case of failure, returns NULL
1409 * and sets @perror to ESRCH.
1411 static struct radix_node *
1412 rt_mpath_unlink(struct rib_head *rnh, struct rt_addrinfo *info,
1413 struct rtentry *rto, int *perror)
1416 * if we got multipath routes, we require users to specify
1417 * a matching RTAX_GATEWAY.
1419 struct rtentry *rt; // *rto = NULL;
1420 struct radix_node *rn;
1421 struct sockaddr *gw;
1423 gw = info->rti_info[RTAX_GATEWAY];
1424 rt = rt_mpath_matchgate(rto, gw);
1431 * this is the first entry in the chain
1434 rn = rn_mpath_next((struct radix_node *)rt);
1436 * there is another entry, now it's active
1441 rto->rt_flags |= RTF_UP;
1443 } else if (rt->rt_flags & RTF_GATEWAY) {
1445 * For gateway routes, we need to
1446 * make sure that we we are deleting
1447 * the correct gateway.
1448 * rt_mpath_matchgate() does not
1449 * check the case when there is only
1450 * one route in the chain.
1453 (rt->rt_gateway->sa_len != gw->sa_len ||
1454 memcmp(rt->rt_gateway, gw, gw->sa_len))) {
1461 * use the normal delete code to remove
1464 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1470 * if the entry is 2nd and on up
1472 if (rt_mpath_deldup(rto, rt) == 0)
1473 panic ("rtrequest1: rt_mpath_deldup");
1475 rn = (struct radix_node *)rt;
1481 static struct rtentry *
1482 rt_flowtable_check_route(struct rib_head *rnh, struct rt_addrinfo *info)
1484 #if defined(INET6) || defined(INET)
1485 struct radix_node *rn;
1487 struct rtentry *rt0;
1490 /* "flow-table" only supports IPv6 and IPv4 at the moment. */
1491 switch (dst->sa_family) {
1498 #if defined(INET6) || defined(INET)
1499 rn = rnh->rnh_matchaddr(dst, &rnh->head);
1500 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
1501 struct sockaddr *mask;
1506 * compare mask to see if the new route is
1507 * more specific than the existing one
1514 * A host route is already present, so
1515 * leave the flow-table entries as is.
1517 if (rt0->rt_flags & RTF_HOST) {
1520 } else if (!(flags & RTF_HOST) && netmask) {
1521 mask = rt_mask(rt0);
1524 n = (u_char *)netmask;
1531 if (len == 0 || (*n < *m)) {
1537 #endif/* INET6 || INET */
1545 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt,
1549 struct rtentry *rt, *rt_old;
1551 struct rtentry *rt0;
1553 struct radix_node *rn;
1554 struct rib_head *rnh;
1556 struct sockaddr *ndst;
1557 struct sockaddr_storage mdst;
1559 KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum"));
1560 KASSERT((flags & RTF_RNH_LOCKED) == 0, ("rtrequest1_fib: locked"));
1561 switch (dst->sa_family) {
1564 /* We support multiple FIBs. */
1567 fibnum = RT_DEFAULT_FIB;
1572 * Find the correct routing tree to use for this Address Family
1574 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
1576 return (EAFNOSUPPORT);
1579 * If we are adding a host route then we don't want to put
1580 * a netmask in the tree, nor do we want to clone it.
1582 if (flags & RTF_HOST)
1588 rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask);
1589 dst = (struct sockaddr *)&mdst;
1593 rt = rt_unlinkrte(rnh, info, &error);
1598 rt_notifydelete(rt, info);
1601 * If the caller wants it, then it can have it,
1602 * but it's up to it to free the rtentry as we won't be
1613 * resolve was only used for route cloning
1618 if ((flags & RTF_GATEWAY) && !gateway)
1620 if (dst && gateway && (dst->sa_family != gateway->sa_family) &&
1621 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
1624 if (info->rti_ifa == NULL) {
1625 error = rt_getifa_fib(info, fibnum);
1629 ifa_ref(info->rti_ifa);
1630 ifa = info->rti_ifa;
1631 rt = uma_zalloc(V_rtzone, M_NOWAIT);
1636 rt->rt_flags = RTF_UP | flags;
1637 rt->rt_fibnum = fibnum;
1639 * Add the gateway. Possibly re-malloc-ing the storage for it.
1641 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
1643 uma_zfree(V_rtzone, rt);
1648 * point to the (possibly newly malloc'd) dest address.
1650 ndst = (struct sockaddr *)rt_key(rt);
1653 * make sure it contains the value we want (masked if needed).
1656 rt_maskedcopy(dst, ndst, netmask);
1658 bcopy(dst, ndst, dst->sa_len);
1661 * We use the ifa reference returned by rt_getifa_fib().
1662 * This moved from below so that rnh->rnh_addaddr() can
1663 * examine the ifa and ifa->ifa_ifp if it so desires.
1666 rt->rt_ifp = ifa->ifa_ifp;
1669 rt_setmetrics(info, rt);
1674 /* do not permit exactly the same dst/mask/gw pair */
1675 if (rt_mpath_capable(rnh) &&
1676 rt_mpath_conflict(rnh, rt, netmask)) {
1679 ifa_free(rt->rt_ifa);
1681 uma_zfree(V_rtzone, rt);
1687 rt0 = rt_flowtable_check_route(rnh, info);
1688 #endif /* FLOWTABLE */
1690 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
1691 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head, rt->rt_nodes);
1694 if (rn == NULL && (info->rti_flags & RTF_PINNED) != 0) {
1697 * Force removal and re-try addition
1698 * TODO: better multipath&pinned support
1700 struct sockaddr *info_dst = info->rti_info[RTAX_DST];
1701 info->rti_info[RTAX_DST] = ndst;
1702 /* Do not delete existing PINNED(interface) routes */
1703 info->rti_flags &= ~RTF_PINNED;
1704 rt_old = rt_unlinkrte(rnh, info, &error);
1705 info->rti_flags |= RTF_PINNED;
1706 info->rti_info[RTAX_DST] = info_dst;
1708 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head,
1717 * If it still failed to go into the tree,
1718 * then un-make it (this should be a function)
1721 ifa_free(rt->rt_ifa);
1723 uma_zfree(V_rtzone, rt);
1731 else if (rt0 != NULL) {
1732 flowtable_route_flush(dst->sa_family, rt0);
1737 if (rt_old != NULL) {
1738 rt_notifydelete(rt_old, info);
1743 * If this protocol has something to add to this then
1744 * allow it to do that as well.
1746 if (ifa->ifa_rtrequest)
1747 ifa->ifa_rtrequest(req, rt, info);
1750 * actually return a resultant rtentry and
1751 * give the caller a single reference.
1761 error = rtrequest1_fib_change(rnh, info, ret_nrt, fibnum);
1779 rtrequest1_fib_change(struct rib_head *rnh, struct rt_addrinfo *info,
1780 struct rtentry **ret_nrt, u_int fibnum)
1782 struct rtentry *rt = NULL;
1786 struct if_mtuinfo ifmtu;
1788 rt = (struct rtentry *)rnh->rnh_lookup(info->rti_info[RTAX_DST],
1789 info->rti_info[RTAX_NETMASK], &rnh->head);
1796 * If we got multipath routes,
1797 * we require users to specify a matching RTAX_GATEWAY.
1799 if (rt_mpath_capable(rnh)) {
1800 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
1808 rt_setmetrics(info, rt);
1811 * New gateway could require new ifaddr, ifp;
1812 * flags may also be different; ifp may be specified
1813 * by ll sockaddr when protocol address is ambiguous
1815 if (((rt->rt_flags & RTF_GATEWAY) &&
1816 info->rti_info[RTAX_GATEWAY] != NULL) ||
1817 info->rti_info[RTAX_IFP] != NULL ||
1818 (info->rti_info[RTAX_IFA] != NULL &&
1819 !sa_equal(info->rti_info[RTAX_IFA], rt->rt_ifa->ifa_addr))) {
1821 error = rt_getifa_fib(info, fibnum);
1822 if (info->rti_ifa != NULL)
1829 /* Check if outgoing interface has changed */
1830 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa &&
1831 rt->rt_ifa != NULL && rt->rt_ifa->ifa_rtrequest != NULL) {
1832 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1833 ifa_free(rt->rt_ifa);
1835 /* Update gateway address */
1836 if (info->rti_info[RTAX_GATEWAY] != NULL) {
1837 error = rt_setgate(rt, rt_key(rt), info->rti_info[RTAX_GATEWAY]);
1841 rt->rt_flags &= ~RTF_GATEWAY;
1842 rt->rt_flags |= (RTF_GATEWAY & info->rti_flags);
1845 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa) {
1846 ifa_ref(info->rti_ifa);
1847 rt->rt_ifa = info->rti_ifa;
1848 rt->rt_ifp = info->rti_ifp;
1850 /* Allow some flags to be toggled on change. */
1851 rt->rt_flags &= ~RTF_FMASK;
1852 rt->rt_flags |= info->rti_flags & RTF_FMASK;
1854 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest != NULL)
1855 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
1857 /* Alter route MTU if necessary */
1858 if (rt->rt_ifp != NULL) {
1859 family = info->rti_info[RTAX_DST]->sa_family;
1860 mtu = if_getmtu_family(rt->rt_ifp, family);
1861 /* Set default MTU */
1862 if (rt->rt_mtu == 0)
1864 if (rt->rt_mtu != mtu) {
1865 /* Check if we really need to update */
1866 ifmtu.ifp = rt->rt_ifp;
1868 if_updatemtu_cb(rt->rt_nodes, &ifmtu);
1879 ifa_free(info->rti_ifa);
1884 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt)
1887 if (info->rti_mflags & RTV_MTU) {
1888 if (info->rti_rmx->rmx_mtu != 0) {
1891 * MTU was explicitly provided by user.
1894 rt->rt_flags |= RTF_FIXEDMTU;
1898 * User explicitly sets MTU to 0.
1899 * Assume rollback to default.
1901 rt->rt_flags &= ~RTF_FIXEDMTU;
1903 rt->rt_mtu = info->rti_rmx->rmx_mtu;
1905 if (info->rti_mflags & RTV_WEIGHT)
1906 rt->rt_weight = info->rti_rmx->rmx_weight;
1907 /* Kernel -> userland timebase conversion. */
1908 if (info->rti_mflags & RTV_EXPIRE)
1909 rt->rt_expire = info->rti_rmx->rmx_expire ?
1910 info->rti_rmx->rmx_expire - time_second + time_uptime : 0;
1914 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
1916 /* XXX dst may be overwritten, can we move this to below */
1917 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
1920 * Prepare to store the gateway in rt->rt_gateway.
1921 * Both dst and gateway are stored one after the other in the same
1922 * malloc'd chunk. If we have room, we can reuse the old buffer,
1923 * rt_gateway already points to the right place.
1924 * Otherwise, malloc a new block and update the 'dst' address.
1926 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
1929 R_Malloc(new, caddr_t, dlen + glen);
1933 * XXX note, we copy from *dst and not *rt_key(rt) because
1934 * rt_setgate() can be called to initialize a newly
1935 * allocated route entry, in which case rt_key(rt) == NULL
1936 * (and also rt->rt_gateway == NULL).
1937 * Free()/free() handle a NULL argument just fine.
1939 bcopy(dst, new, dlen);
1940 R_Free(rt_key(rt)); /* free old block, if any */
1941 rt_key(rt) = (struct sockaddr *)new;
1942 rt->rt_gateway = (struct sockaddr *)(new + dlen);
1946 * Copy the new gateway value into the memory chunk.
1948 bcopy(gate, rt->rt_gateway, glen);
1954 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
1956 u_char *cp1 = (u_char *)src;
1957 u_char *cp2 = (u_char *)dst;
1958 u_char *cp3 = (u_char *)netmask;
1959 u_char *cplim = cp2 + *cp3;
1960 u_char *cplim2 = cp2 + *cp1;
1962 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1967 *cp2++ = *cp1++ & *cp3++;
1969 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1973 * Set up a routing table entry, normally
1976 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
1978 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
1980 struct sockaddr *dst;
1981 struct sockaddr *netmask;
1982 struct rtentry *rt = NULL;
1983 struct rt_addrinfo info;
1985 int startfib, endfib;
1986 char tempbuf[_SOCKADDR_TMPSIZE];
1989 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1990 struct rib_head *rnh;
1992 if (flags & RTF_HOST) {
1993 dst = ifa->ifa_dstaddr;
1996 dst = ifa->ifa_addr;
1997 netmask = ifa->ifa_netmask;
1999 if (dst->sa_len == 0)
2001 switch (dst->sa_family) {
2004 /* We support multiple FIBs. */
2007 fibnum = RT_DEFAULT_FIB;
2010 if (fibnum == RT_ALL_FIBS) {
2011 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD)
2012 startfib = endfib = ifa->ifa_ifp->if_fib;
2015 endfib = rt_numfibs - 1;
2018 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
2024 * If it's a delete, check that if it exists,
2025 * it's on the correct interface or we might scrub
2026 * a route to another ifa which would
2027 * be confusing at best and possibly worse.
2029 if (cmd == RTM_DELETE) {
2031 * It's a delete, so it should already exist..
2032 * If it's a net, mask off the host bits
2033 * (Assuming we have a mask)
2034 * XXX this is kinda inet specific..
2036 if (netmask != NULL) {
2037 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
2038 dst = (struct sockaddr *)tempbuf;
2042 * Now go through all the requested tables (fibs) and do the
2043 * requested action. Realistically, this will either be fib 0
2044 * for protocols that don't do multiple tables or all the
2045 * tables for those that do.
2047 for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
2048 if (cmd == RTM_DELETE) {
2049 struct radix_node *rn;
2051 * Look up an rtentry that is in the routing tree and
2052 * contains the correct info.
2054 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
2056 /* this table doesn't exist but others might */
2059 rn = rnh->rnh_lookup(dst, netmask, &rnh->head);
2061 if (rt_mpath_capable(rnh)) {
2068 * for interface route the
2069 * rt->rt_gateway is sockaddr_intf
2070 * for cloning ARP entries, so
2071 * rt_mpath_matchgate must use the
2074 rt = rt_mpath_matchgate(rt,
2081 error = (rn == NULL ||
2082 (rn->rn_flags & RNF_ROOT) ||
2083 RNTORT(rn)->rt_ifa != ifa);
2086 /* this is only an error if bad on ALL tables */
2091 * Do the actual request
2093 bzero((caddr_t)&info, sizeof(info));
2095 info.rti_flags = flags |
2096 (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED;
2097 info.rti_info[RTAX_DST] = dst;
2099 * doing this for compatibility reasons
2102 info.rti_info[RTAX_GATEWAY] =
2103 (struct sockaddr *)&null_sdl;
2105 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
2106 info.rti_info[RTAX_NETMASK] = netmask;
2107 error = rtrequest1_fib(cmd, &info, &rt, fibnum);
2109 if (error == 0 && rt != NULL) {
2111 * notify any listening routing agents of the change
2116 * in case address alias finds the first address
2117 * e.g. ifconfig bge0 192.0.2.246/24
2118 * e.g. ifconfig bge0 192.0.2.247/24
2119 * the address set in the route is 192.0.2.246
2120 * so we need to replace it with 192.0.2.247
2122 if (memcmp(rt->rt_ifa->ifa_addr,
2123 ifa->ifa_addr, ifa->ifa_addr->sa_len)) {
2124 ifa_free(rt->rt_ifa);
2126 rt->rt_ifp = ifa->ifa_ifp;
2131 * doing this for compatibility reasons
2133 if (cmd == RTM_ADD) {
2134 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
2135 rt->rt_ifp->if_type;
2136 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
2137 rt->rt_ifp->if_index;
2141 rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum);
2144 if (cmd == RTM_DELETE) {
2146 * If we are deleting, and we found an entry,
2147 * then it's been removed from the tree..
2148 * now throw it away.
2152 if (cmd == RTM_ADD) {
2154 * We just wanted to add it..
2155 * we don't actually need a reference.
2166 if (cmd == RTM_DELETE) {
2170 /* we only give an error if it wasn't in any table */
2171 error = ((flags & RTF_HOST) ?
2172 EHOSTUNREACH : ENETUNREACH);
2176 /* return an error if any of them failed */
2184 * Set up a routing table entry, normally
2188 rtinit(struct ifaddr *ifa, int cmd, int flags)
2190 struct sockaddr *dst;
2191 int fib = RT_DEFAULT_FIB;
2193 if (flags & RTF_HOST) {
2194 dst = ifa->ifa_dstaddr;
2196 dst = ifa->ifa_addr;
2199 switch (dst->sa_family) {
2202 /* We do support multiple FIBs. */
2206 return (rtinit1(ifa, cmd, flags, fib));
2210 * Announce interface address arrival/withdraw
2211 * Returns 0 on success.
2214 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
2217 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2218 ("unexpected cmd %d", cmd));
2220 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2221 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2223 #if defined(INET) || defined(INET6)
2226 * notify the SCTP stack
2227 * this will only get called when an address is added/deleted
2228 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
2230 sctp_addr_change(ifa, cmd);
2233 return (rtsock_addrmsg(cmd, ifa, fibnum));
2237 * Announce route addition/removal.
2238 * Users of this function MUST validate input data BEFORE calling.
2239 * However we have to be able to handle invalid data:
2240 * if some userland app sends us "invalid" route message (invalid mask,
2241 * no dst, wrong address families, etc...) we need to pass it back
2242 * to app (and any other rtsock consumers) with rtm_errno field set to
2244 * Returns 0 on success.
2247 rt_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
2251 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2252 ("unexpected cmd %d", cmd));
2254 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2255 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2257 KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
2259 return (rtsock_routemsg(cmd, ifp, error, rt, fibnum));
2263 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
2266 rt_newaddrmsg_fib(cmd, ifa, error, rt, RT_ALL_FIBS);
2270 * This is called to generate messages from the routing socket
2271 * indicating a network interface has had addresses associated with it.
2274 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
2278 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2279 ("unexpected cmd %u", cmd));
2280 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2281 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2283 if (cmd == RTM_ADD) {
2284 rt_addrmsg(cmd, ifa, fibnum);
2286 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2289 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2290 rt_addrmsg(cmd, ifa, fibnum);