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 * 3. 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 rt_tables_get_gen(int table, int fam)
207 struct rib_head *rnh;
209 rnh = *rt_tables_get_rnh_ptr(table, fam);
210 KASSERT(rnh != NULL, ("%s: NULL rib_head pointer table %d fam %d",
211 __func__, table, fam));
212 return (rnh->rnh_gen);
217 * route initialization must occur before ip6_init2(), which happenas at
224 /* whack the tunable ints into line. */
225 if (rt_numfibs > RT_MAXFIBS)
226 rt_numfibs = RT_MAXFIBS;
230 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0);
233 rtentry_zinit(void *mem, int size, int how)
235 struct rtentry *rt = mem;
237 rt->rt_pksent = counter_u64_alloc(how);
238 if (rt->rt_pksent == NULL)
247 rtentry_zfini(void *mem, int size)
249 struct rtentry *rt = mem;
252 counter_u64_free(rt->rt_pksent);
256 rtentry_ctor(void *mem, int size, void *arg, int how)
258 struct rtentry *rt = mem;
260 bzero(rt, offsetof(struct rtentry, rt_endzero));
261 counter_u64_zero(rt->rt_pksent);
268 rtentry_dtor(void *mem, int size, void *arg)
270 struct rtentry *rt = mem;
276 vnet_route_init(const void *unused __unused)
279 struct rib_head **rnh;
283 V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) *
284 sizeof(struct rib_head *), M_RTABLE, M_WAITOK|M_ZERO);
286 V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry),
287 rtentry_ctor, rtentry_dtor,
288 rtentry_zinit, rtentry_zfini, UMA_ALIGN_PTR, 0);
289 for (dom = domains; dom; dom = dom->dom_next) {
290 if (dom->dom_rtattach == NULL)
293 for (table = 0; table < rt_numfibs; table++) {
294 fam = dom->dom_family;
295 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
298 rnh = rt_tables_get_rnh_ptr(table, fam);
300 panic("%s: rnh NULL", __func__);
301 dom->dom_rtattach((void **)rnh, 0);
305 VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
310 vnet_route_uninit(const void *unused __unused)
315 struct rib_head **rnh;
317 for (dom = domains; dom; dom = dom->dom_next) {
318 if (dom->dom_rtdetach == NULL)
321 for (table = 0; table < rt_numfibs; table++) {
322 fam = dom->dom_family;
324 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
327 rnh = rt_tables_get_rnh_ptr(table, fam);
329 panic("%s: rnh NULL", __func__);
330 dom->dom_rtdetach((void **)rnh, 0);
334 free(V_rt_tables, M_RTABLE);
335 uma_zdestroy(V_rtzone);
337 VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST,
338 vnet_route_uninit, 0);
342 rt_table_init(int offset)
346 rh = malloc(sizeof(struct rib_head), M_RTABLE, M_WAITOK | M_ZERO);
348 /* TODO: These details should be hidded inside radix.c */
349 /* Init masks tree */
350 rn_inithead_internal(&rh->head, rh->rnh_nodes, offset);
351 rn_inithead_internal(&rh->rmhead.head, rh->rmhead.mask_nodes, 0);
352 rh->head.rnh_masks = &rh->rmhead;
357 /* Finally, set base callbacks */
358 rh->rnh_addaddr = rn_addroute;
359 rh->rnh_deladdr = rn_delete;
360 rh->rnh_matchaddr = rn_match;
361 rh->rnh_lookup = rn_lookup;
362 rh->rnh_walktree = rn_walktree;
363 rh->rnh_walktree_from = rn_walktree_from;
369 rt_freeentry(struct radix_node *rn, void *arg)
371 struct radix_head * const rnh = arg;
372 struct radix_node *x;
374 x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh);
381 rt_table_destroy(struct rib_head *rh)
384 rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head);
386 /* Assume table is already empty */
387 RIB_LOCK_DESTROY(rh);
392 #ifndef _SYS_SYSPROTO_H_
398 sys_setfib(struct thread *td, struct setfib_args *uap)
400 if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs)
402 td->td_proc->p_fibnum = uap->fibnum;
407 * Packet routing routines.
410 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum)
414 if ((rt = ro->ro_rt) != NULL) {
415 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
420 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum);
422 RT_UNLOCK(ro->ro_rt);
426 * Look up the route that matches the address given
427 * Or, at least try.. Create a cloned route if needed.
429 * The returned route, if any, is locked.
432 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
435 return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB));
439 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags,
443 struct radix_node *rn;
444 struct rtentry *newrt;
445 struct rt_addrinfo info;
446 int err = 0, msgtype = RTM_MISS;
448 KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum"));
449 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
455 * Look up the address in the table for that Address Family
458 rn = rh->rnh_matchaddr(dst, &rh->head);
459 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
470 * Either we hit the root or could not find any match,
471 * which basically means: "cannot get there from here".
474 V_rtstat.rts_unreach++;
478 * If required, report the failure to the supervising
480 * For a delete, this is not an error. (report == 0)
482 bzero(&info, sizeof(info));
483 info.rti_info[RTAX_DST] = dst;
484 rt_missmsg_fib(msgtype, &info, 0, err, fibnum);
490 * Remove a reference count from an rtentry.
491 * If the count gets low enough, take it out of the routing table
494 rtfree(struct rtentry *rt)
496 struct rib_head *rnh;
498 KASSERT(rt != NULL,("%s: NULL rt", __func__));
499 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
500 KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
505 * The callers should use RTFREE_LOCKED() or RTFREE(), so
506 * we should come here exactly with the last reference.
509 if (rt->rt_refcnt > 0) {
510 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt);
515 * On last reference give the "close method" a chance
516 * to cleanup private state. This also permits (for
517 * IPv4 and IPv6) a chance to decide if the routing table
518 * entry should be purged immediately or at a later time.
519 * When an immediate purge is to happen the close routine
520 * typically calls rtexpunge which clears the RTF_UP flag
521 * on the entry so that the code below reclaims the storage.
523 if (rt->rt_refcnt == 0 && rnh->rnh_close)
524 rnh->rnh_close((struct radix_node *)rt, &rnh->head);
527 * If we are no longer "up" (and ref == 0)
528 * then we can free the resources associated
531 if ((rt->rt_flags & RTF_UP) == 0) {
532 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
535 * the rtentry must have been removed from the routing table
536 * so it is represented in rttrash.. remove that now.
540 if (rt->rt_refcnt < 0) {
541 printf("rtfree: %p not freed (neg refs)\n", rt);
546 * release references on items we hold them on..
547 * e.g other routes and ifaddrs.
550 ifa_free(rt->rt_ifa);
552 * The key is separatly alloc'd so free it (see rt_setgate()).
553 * This also frees the gateway, as they are always malloc'd
559 * and the rtentry itself of course
561 uma_zfree(V_rtzone, rt);
570 * Force a routing table entry to the specified
571 * destination to go through the given gateway.
572 * Normally called as a result of a routing redirect
573 * message from the network layer.
576 rtredirect_fib(struct sockaddr *dst,
577 struct sockaddr *gateway,
578 struct sockaddr *netmask,
580 struct sockaddr *src,
586 struct rt_addrinfo info;
588 struct rib_head *rnh;
591 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
593 error = EAFNOSUPPORT;
597 /* verify the gateway is directly reachable */
598 if ((ifa = ifa_ifwithnet(gateway, 0, fibnum)) == NULL) {
602 rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */
604 * If the redirect isn't from our current router for this dst,
605 * it's either old or wrong. If it redirects us to ourselves,
606 * we have a routing loop, perhaps as a result of an interface
607 * going down recently.
609 if (!(flags & RTF_DONE) && rt) {
610 if (!sa_equal(src, rt->rt_gateway)) {
614 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) {
619 if ((flags & RTF_GATEWAY) && ifa_ifwithaddr_check(gateway)) {
620 error = EHOSTUNREACH;
624 * Create a new entry if we just got back a wildcard entry
625 * or the lookup failed. This is necessary for hosts
626 * which use routing redirects generated by smart gateways
627 * to dynamically build the routing tables.
629 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
632 * Don't listen to the redirect if it's
633 * for a route to an interface.
635 if (rt->rt_flags & RTF_GATEWAY) {
636 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
638 * Changing from route to net => route to host.
639 * Create new route, rather than smashing route to net.
645 flags |= RTF_DYNAMIC;
646 bzero((caddr_t)&info, sizeof(info));
647 info.rti_info[RTAX_DST] = dst;
648 info.rti_info[RTAX_GATEWAY] = gateway;
649 info.rti_info[RTAX_NETMASK] = netmask;
651 info.rti_flags = flags;
652 error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum);
655 flags = rt->rt_flags;
658 stat = &V_rtstat.rts_dynamic;
662 * Smash the current notion of the gateway to
663 * this destination. Should check about netmask!!!
665 if ((flags & RTF_GATEWAY) == 0)
666 rt->rt_flags &= ~RTF_GATEWAY;
667 rt->rt_flags |= RTF_MODIFIED;
668 flags |= RTF_MODIFIED;
669 stat = &V_rtstat.rts_newgateway;
671 * add the key and gateway (in one malloc'd chunk).
676 rt_setgate(rt, rt_key(rt), gateway);
680 error = EHOSTUNREACH;
686 V_rtstat.rts_badredirect++;
687 else if (stat != NULL)
689 bzero((caddr_t)&info, sizeof(info));
690 info.rti_info[RTAX_DST] = dst;
691 info.rti_info[RTAX_GATEWAY] = gateway;
692 info.rti_info[RTAX_NETMASK] = netmask;
693 info.rti_info[RTAX_AUTHOR] = src;
694 rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum);
700 * Routing table ioctl interface.
703 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
707 * If more ioctl commands are added here, make sure the proper
708 * super-user checks are being performed because it is possible for
709 * prison-root to make it this far if raw sockets have been enabled
713 /* Multicast goop, grrr... */
714 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
721 ifa_ifwithroute(int flags, const struct sockaddr *dst, struct sockaddr *gateway,
727 if ((flags & RTF_GATEWAY) == 0) {
729 * If we are adding a route to an interface,
730 * and the interface is a pt to pt link
731 * we should search for the destination
732 * as our clue to the interface. Otherwise
733 * we can use the local address.
736 if (flags & RTF_HOST)
737 ifa = ifa_ifwithdstaddr(dst, fibnum);
739 ifa = ifa_ifwithaddr(gateway);
742 * If we are adding a route to a remote net
743 * or host, the gateway may still be on the
744 * other end of a pt to pt link.
746 ifa = ifa_ifwithdstaddr(gateway, fibnum);
749 ifa = ifa_ifwithnet(gateway, 0, fibnum);
751 struct rtentry *rt = rtalloc1_fib(gateway, 0, 0, fibnum);
755 * dismiss a gateway that is reachable only
756 * through the default router
758 switch (gateway->sa_family) {
760 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
764 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
770 if (!not_found && rt->rt_ifa != NULL) {
776 if (not_found || ifa == NULL)
779 if (ifa->ifa_addr->sa_family != dst->sa_family) {
780 struct ifaddr *oifa = ifa;
781 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
791 * Do appropriate manipulations of a routing tree given
792 * all the bits of info needed
795 rtrequest_fib(int req,
796 struct sockaddr *dst,
797 struct sockaddr *gateway,
798 struct sockaddr *netmask,
800 struct rtentry **ret_nrt,
803 struct rt_addrinfo info;
805 if (dst->sa_len == 0)
808 bzero((caddr_t)&info, sizeof(info));
809 info.rti_flags = flags;
810 info.rti_info[RTAX_DST] = dst;
811 info.rti_info[RTAX_GATEWAY] = gateway;
812 info.rti_info[RTAX_NETMASK] = netmask;
813 return rtrequest1_fib(req, &info, ret_nrt, fibnum);
818 * Copy most of @rt data into @info.
820 * If @flags contains NHR_COPY, copies dst,netmask and gw to the
821 * pointers specified by @info structure. Assume such pointers
822 * are zeroed sockaddr-like structures with sa_len field initialized
823 * to reflect size of the provided buffer. if no NHR_COPY is specified,
824 * point dst,netmask and gw @info fields to appropriate @rt values.
826 * if @flags contains NHR_REF, do refcouting on rt_ifp.
828 * Returns 0 on success.
831 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags)
833 struct rt_metrics *rmx;
834 struct sockaddr *src, *dst;
837 if (flags & NHR_COPY) {
838 /* Copy destination if dst is non-zero */
840 dst = info->rti_info[RTAX_DST];
841 sa_len = src->sa_len;
843 if (src->sa_len > dst->sa_len)
845 memcpy(dst, src, src->sa_len);
846 info->rti_addrs |= RTA_DST;
849 /* Copy mask if set && dst is non-zero */
851 dst = info->rti_info[RTAX_NETMASK];
852 if (src != NULL && dst != NULL) {
855 * Radix stores different value in sa_len,
856 * assume rt_mask() to have the same length
859 if (sa_len > dst->sa_len)
861 memcpy(dst, src, src->sa_len);
862 info->rti_addrs |= RTA_NETMASK;
865 /* Copy gateway is set && dst is non-zero */
866 src = rt->rt_gateway;
867 dst = info->rti_info[RTAX_GATEWAY];
868 if ((rt->rt_flags & RTF_GATEWAY) && src != NULL && dst != NULL){
869 if (src->sa_len > dst->sa_len)
871 memcpy(dst, src, src->sa_len);
872 info->rti_addrs |= RTA_GATEWAY;
875 info->rti_info[RTAX_DST] = rt_key(rt);
876 info->rti_addrs |= RTA_DST;
877 if (rt_mask(rt) != NULL) {
878 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
879 info->rti_addrs |= RTA_NETMASK;
881 if (rt->rt_flags & RTF_GATEWAY) {
882 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
883 info->rti_addrs |= RTA_GATEWAY;
889 info->rti_mflags |= RTV_MTU;
890 rmx->rmx_mtu = rt->rt_mtu;
893 info->rti_flags = rt->rt_flags;
894 info->rti_ifp = rt->rt_ifp;
895 info->rti_ifa = rt->rt_ifa;
897 if (flags & NHR_REF) {
898 /* Do 'traditional' refcouting */
899 if_ref(info->rti_ifp);
906 * Lookups up route entry for @dst in RIB database for fib @fibnum.
907 * Exports entry data to @info using rt_exportinfo().
909 * if @flags contains NHR_REF, refcouting is performed on rt_ifp.
910 * All references can be released later by calling rib_free_info()
912 * Returns 0 on success.
913 * Returns ENOENT for lookup failure, ENOMEM for export failure.
916 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
917 uint32_t flowid, struct rt_addrinfo *info)
920 struct radix_node *rn;
924 KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
925 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
930 rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head);
931 if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
933 /* Ensure route & ifp is UP */
934 if (RT_LINK_IS_UP(rt->rt_ifp)) {
935 flags = (flags & NHR_REF) | NHR_COPY;
936 error = rt_exportinfo(rt, info, flags);
948 * Releases all references acquired by rib_lookup_info() when
949 * called with NHR_REF flags.
952 rib_free_info(struct rt_addrinfo *info)
955 if_rele(info->rti_ifp);
959 * Iterates over all existing fibs in system calling
960 * @setwa_f function prior to traversing each fib.
961 * Calls @wa_f function for each element in current fib.
962 * If af is not AF_UNSPEC, iterates over fibs in particular
966 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f,
969 struct rib_head *rnh;
973 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
974 /* Do we want some specific family? */
975 if (af != AF_UNSPEC) {
976 rnh = rt_tables_get_rnh(fibnum, af);
980 setwa_f(rnh, fibnum, af, arg);
983 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
988 for (i = 1; i <= AF_MAX; i++) {
989 rnh = rt_tables_get_rnh(fibnum, i);
993 setwa_f(rnh, fibnum, i, arg);
996 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
1004 struct rt_addrinfo info;
1005 struct rib_head *rnh;
1006 struct rtentry *head;
1010 * Conditionally unlinks @rn from radix tree based
1011 * on info data passed in @arg.
1014 rt_checkdelroute(struct radix_node *rn, void *arg)
1016 struct rt_delinfo *di;
1017 struct rt_addrinfo *info;
1021 di = (struct rt_delinfo *)arg;
1022 rt = (struct rtentry *)rn;
1026 info->rti_info[RTAX_DST] = rt_key(rt);
1027 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
1028 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1030 rt = rt_unlinkrte(di->rnh, info, &error);
1032 /* Either not allowed or not matched. Skip entry */
1036 /* Entry was unlinked. Add to the list and return */
1037 rt->rt_chain = di->head;
1044 * Iterates over all existing fibs in system.
1045 * Deletes each element for which @filter_f function returned
1047 * If @af is not AF_UNSPEC, iterates over fibs in particular
1051 rt_foreach_fib_walk_del(int af, rt_filter_f_t *filter_f, void *arg)
1053 struct rib_head *rnh;
1054 struct rt_delinfo di;
1059 bzero(&di, sizeof(di));
1060 di.info.rti_filter = filter_f;
1061 di.info.rti_filterdata = arg;
1063 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1064 /* Do we want some specific family? */
1065 if (af != AF_UNSPEC) {
1073 for (i = start; i <= end; i++) {
1074 rnh = rt_tables_get_rnh(fibnum, i);
1080 rnh->rnh_walktree(&rnh->head, rt_checkdelroute, &di);
1083 if (di.head == NULL)
1086 /* We might have something to reclaim */
1087 while (di.head != NULL) {
1089 di.head = rt->rt_chain;
1090 rt->rt_chain = NULL;
1092 /* TODO std rt -> rt_addrinfo export */
1093 di.info.rti_info[RTAX_DST] = rt_key(rt);
1094 di.info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1096 rt_notifydelete(rt, &di.info);
1105 * Delete Routes for a Network Interface
1107 * Called for each routing entry via the rnh->rnh_walktree() call above
1108 * to delete all route entries referencing a detaching network interface.
1111 * rt pointer to rtentry
1112 * arg argument passed to rnh->rnh_walktree() - detaching interface
1116 * errno failed - reason indicated
1119 rt_ifdelroute(const struct rtentry *rt, void *arg)
1121 struct ifnet *ifp = arg;
1123 if (rt->rt_ifp != ifp)
1127 * Protect (sorta) against walktree recursion problems
1128 * with cloned routes
1130 if ((rt->rt_flags & RTF_UP) == 0)
1137 * Delete all remaining routes using this interface
1138 * Unfortuneatly the only way to do this is to slog through
1139 * the entire routing table looking for routes which point
1140 * to this interface...oh well...
1143 rt_flushifroutes_af(struct ifnet *ifp, int af)
1145 KASSERT((af >= 1 && af <= AF_MAX), ("%s: af %d not >= 1 and <= %d",
1146 __func__, af, AF_MAX));
1148 rt_foreach_fib_walk_del(af, rt_ifdelroute, ifp);
1152 rt_flushifroutes(struct ifnet *ifp)
1155 rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
1159 * Conditionally unlinks rtentry matching data inside @info from @rnh.
1160 * Returns unlinked, locked and referenced @rtentry on success,
1161 * Returns NULL and sets @perror to:
1162 * ESRCH - if prefix was not found,
1163 * EADDRINUSE - if trying to delete PINNED route without appropriate flag.
1164 * ENOENT - if supplied filter function returned 0 (not matched).
1166 static struct rtentry *
1167 rt_unlinkrte(struct rib_head *rnh, struct rt_addrinfo *info, int *perror)
1169 struct sockaddr *dst, *netmask;
1171 struct radix_node *rn;
1173 dst = info->rti_info[RTAX_DST];
1174 netmask = info->rti_info[RTAX_NETMASK];
1176 rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, &rnh->head);
1182 if ((info->rti_flags & RTF_PINNED) == 0) {
1183 /* Check if target route can be deleted */
1184 if (rt->rt_flags & RTF_PINNED) {
1185 *perror = EADDRINUSE;
1190 if (info->rti_filter != NULL) {
1191 if (info->rti_filter(rt, info->rti_filterdata) == 0) {
1198 * Filter function requested rte deletion.
1199 * Ease the caller work by filling in remaining info
1200 * from that particular entry.
1202 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1206 * Remove the item from the tree and return it.
1207 * Complain if it is not there and do no more processing.
1211 if (rt_mpath_capable(rnh))
1212 rn = rt_mpath_unlink(rnh, info, rt, perror);
1215 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1219 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
1220 panic ("rtrequest delete");
1225 rt->rt_flags &= ~RTF_UP;
1233 rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info)
1238 * give the protocol a chance to keep things in sync.
1241 if (ifa != NULL && ifa->ifa_rtrequest != NULL)
1242 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1245 * One more rtentry floating around that is not
1246 * linked to the routing table. rttrash will be decremented
1247 * when RTFREE(rt) is eventually called.
1254 * These (questionable) definitions of apparent local variables apply
1255 * to the next two functions. XXXXXX!!!
1257 #define dst info->rti_info[RTAX_DST]
1258 #define gateway info->rti_info[RTAX_GATEWAY]
1259 #define netmask info->rti_info[RTAX_NETMASK]
1260 #define ifaaddr info->rti_info[RTAX_IFA]
1261 #define ifpaddr info->rti_info[RTAX_IFP]
1262 #define flags info->rti_flags
1265 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined,
1266 * it will be referenced so the caller must free it.
1269 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
1275 * ifp may be specified by sockaddr_dl
1276 * when protocol address is ambiguous.
1278 if (info->rti_ifp == NULL && ifpaddr != NULL &&
1279 ifpaddr->sa_family == AF_LINK &&
1280 (ifa = ifa_ifwithnet(ifpaddr, 0, fibnum)) != NULL) {
1281 info->rti_ifp = ifa->ifa_ifp;
1284 if (info->rti_ifa == NULL && ifaaddr != NULL)
1285 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
1286 if (info->rti_ifa == NULL) {
1287 struct sockaddr *sa;
1289 sa = ifaaddr != NULL ? ifaaddr :
1290 (gateway != NULL ? gateway : dst);
1291 if (sa != NULL && info->rti_ifp != NULL)
1292 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
1293 else if (dst != NULL && gateway != NULL)
1294 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway,
1296 else if (sa != NULL)
1297 info->rti_ifa = ifa_ifwithroute(flags, sa, sa,
1300 if ((ifa = info->rti_ifa) != NULL) {
1301 if (info->rti_ifp == NULL)
1302 info->rti_ifp = ifa->ifa_ifp;
1304 error = ENETUNREACH;
1309 if_updatemtu_cb(struct radix_node *rn, void *arg)
1312 struct if_mtuinfo *ifmtu;
1314 rt = (struct rtentry *)rn;
1315 ifmtu = (struct if_mtuinfo *)arg;
1317 if (rt->rt_ifp != ifmtu->ifp)
1320 if (rt->rt_mtu >= ifmtu->mtu) {
1321 /* We have to decrease mtu regardless of flags */
1322 rt->rt_mtu = ifmtu->mtu;
1327 * New MTU is bigger. Check if are allowed to alter it
1329 if ((rt->rt_flags & (RTF_FIXEDMTU | RTF_GATEWAY | RTF_HOST)) != 0) {
1332 * Skip routes with user-supplied MTU and
1333 * non-interface routes
1338 /* We are safe to update route MTU */
1339 rt->rt_mtu = ifmtu->mtu;
1345 rt_updatemtu(struct ifnet *ifp)
1347 struct if_mtuinfo ifmtu;
1348 struct rib_head *rnh;
1354 * Try to update rt_mtu for all routes using this interface
1355 * Unfortunately the only way to do this is to traverse all
1356 * routing tables in all fibs/domains.
1358 for (i = 1; i <= AF_MAX; i++) {
1359 ifmtu.mtu = if_getmtu_family(ifp, i);
1360 for (j = 0; j < rt_numfibs; j++) {
1361 rnh = rt_tables_get_rnh(j, i);
1365 rnh->rnh_walktree(&rnh->head, if_updatemtu_cb, &ifmtu);
1373 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
1374 int rt_print(char *buf, int buflen, struct rtentry *rt);
1377 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
1381 switch (s->sa_family) {
1383 paddr = &((struct sockaddr_in *)s)->sin_addr;
1386 paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
1393 if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
1396 return (strlen(buf));
1400 rt_print(char *buf, int buflen, struct rtentry *rt)
1402 struct sockaddr *addr, *mask;
1408 i = p_sockaddr(buf, buflen, addr);
1409 if (!(rt->rt_flags & RTF_HOST)) {
1411 i += p_sockaddr(buf + i, buflen - i, mask);
1414 if (rt->rt_flags & RTF_GATEWAY) {
1416 i += p_sockaddr(buf + i, buflen - i, rt->rt_gateway);
1425 * Deletes key for single-path routes, unlinks rtentry with
1426 * gateway specified in @info from multi-path routes.
1428 * Returnes unlinked entry. In case of failure, returns NULL
1429 * and sets @perror to ESRCH.
1431 static struct radix_node *
1432 rt_mpath_unlink(struct rib_head *rnh, struct rt_addrinfo *info,
1433 struct rtentry *rto, int *perror)
1436 * if we got multipath routes, we require users to specify
1437 * a matching RTAX_GATEWAY.
1439 struct rtentry *rt; // *rto = NULL;
1440 struct radix_node *rn;
1441 struct sockaddr *gw;
1443 gw = info->rti_info[RTAX_GATEWAY];
1444 rt = rt_mpath_matchgate(rto, gw);
1451 * this is the first entry in the chain
1454 rn = rn_mpath_next((struct radix_node *)rt);
1456 * there is another entry, now it's active
1461 rto->rt_flags |= RTF_UP;
1463 } else if (rt->rt_flags & RTF_GATEWAY) {
1465 * For gateway routes, we need to
1466 * make sure that we we are deleting
1467 * the correct gateway.
1468 * rt_mpath_matchgate() does not
1469 * check the case when there is only
1470 * one route in the chain.
1473 (rt->rt_gateway->sa_len != gw->sa_len ||
1474 memcmp(rt->rt_gateway, gw, gw->sa_len))) {
1481 * use the normal delete code to remove
1484 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1490 * if the entry is 2nd and on up
1492 if (rt_mpath_deldup(rto, rt) == 0)
1493 panic ("rtrequest1: rt_mpath_deldup");
1495 rn = (struct radix_node *)rt;
1501 static struct rtentry *
1502 rt_flowtable_check_route(struct rib_head *rnh, struct rt_addrinfo *info)
1504 #if defined(INET6) || defined(INET)
1505 struct radix_node *rn;
1507 struct rtentry *rt0;
1510 /* "flow-table" only supports IPv6 and IPv4 at the moment. */
1511 switch (dst->sa_family) {
1518 #if defined(INET6) || defined(INET)
1519 rn = rnh->rnh_matchaddr(dst, &rnh->head);
1520 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
1521 struct sockaddr *mask;
1526 * compare mask to see if the new route is
1527 * more specific than the existing one
1534 * A host route is already present, so
1535 * leave the flow-table entries as is.
1537 if (rt0->rt_flags & RTF_HOST) {
1540 } else if (!(flags & RTF_HOST) && netmask) {
1541 mask = rt_mask(rt0);
1544 n = (u_char *)netmask;
1551 if (len == 0 || (*n < *m)) {
1557 #endif/* INET6 || INET */
1565 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt,
1569 struct rtentry *rt, *rt_old;
1571 struct rtentry *rt0;
1573 struct radix_node *rn;
1574 struct rib_head *rnh;
1576 struct sockaddr *ndst;
1577 struct sockaddr_storage mdst;
1579 KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum"));
1580 KASSERT((flags & RTF_RNH_LOCKED) == 0, ("rtrequest1_fib: locked"));
1581 switch (dst->sa_family) {
1584 /* We support multiple FIBs. */
1587 fibnum = RT_DEFAULT_FIB;
1592 * Find the correct routing tree to use for this Address Family
1594 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
1596 return (EAFNOSUPPORT);
1599 * If we are adding a host route then we don't want to put
1600 * a netmask in the tree, nor do we want to clone it.
1602 if (flags & RTF_HOST)
1608 rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask);
1609 dst = (struct sockaddr *)&mdst;
1613 rt = rt_unlinkrte(rnh, info, &error);
1618 rt_notifydelete(rt, info);
1621 * If the caller wants it, then it can have it,
1622 * but it's up to it to free the rtentry as we won't be
1633 * resolve was only used for route cloning
1638 if ((flags & RTF_GATEWAY) && !gateway)
1640 if (dst && gateway && (dst->sa_family != gateway->sa_family) &&
1641 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
1644 if (info->rti_ifa == NULL) {
1645 error = rt_getifa_fib(info, fibnum);
1649 ifa_ref(info->rti_ifa);
1650 ifa = info->rti_ifa;
1651 rt = uma_zalloc(V_rtzone, M_NOWAIT);
1656 rt->rt_flags = RTF_UP | flags;
1657 rt->rt_fibnum = fibnum;
1659 * Add the gateway. Possibly re-malloc-ing the storage for it.
1661 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
1663 uma_zfree(V_rtzone, rt);
1668 * point to the (possibly newly malloc'd) dest address.
1670 ndst = (struct sockaddr *)rt_key(rt);
1673 * make sure it contains the value we want (masked if needed).
1676 rt_maskedcopy(dst, ndst, netmask);
1678 bcopy(dst, ndst, dst->sa_len);
1681 * We use the ifa reference returned by rt_getifa_fib().
1682 * This moved from below so that rnh->rnh_addaddr() can
1683 * examine the ifa and ifa->ifa_ifp if it so desires.
1686 rt->rt_ifp = ifa->ifa_ifp;
1689 rt_setmetrics(info, rt);
1694 /* do not permit exactly the same dst/mask/gw pair */
1695 if (rt_mpath_capable(rnh) &&
1696 rt_mpath_conflict(rnh, rt, netmask)) {
1699 ifa_free(rt->rt_ifa);
1701 uma_zfree(V_rtzone, rt);
1707 rt0 = rt_flowtable_check_route(rnh, info);
1708 #endif /* FLOWTABLE */
1710 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
1711 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head, rt->rt_nodes);
1714 if (rn == NULL && (info->rti_flags & RTF_PINNED) != 0) {
1717 * Force removal and re-try addition
1718 * TODO: better multipath&pinned support
1720 struct sockaddr *info_dst = info->rti_info[RTAX_DST];
1721 info->rti_info[RTAX_DST] = ndst;
1722 /* Do not delete existing PINNED(interface) routes */
1723 info->rti_flags &= ~RTF_PINNED;
1724 rt_old = rt_unlinkrte(rnh, info, &error);
1725 info->rti_flags |= RTF_PINNED;
1726 info->rti_info[RTAX_DST] = info_dst;
1728 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head,
1737 * If it still failed to go into the tree,
1738 * then un-make it (this should be a function)
1741 ifa_free(rt->rt_ifa);
1743 uma_zfree(V_rtzone, rt);
1751 else if (rt0 != NULL) {
1752 flowtable_route_flush(dst->sa_family, rt0);
1757 if (rt_old != NULL) {
1758 rt_notifydelete(rt_old, info);
1763 * If this protocol has something to add to this then
1764 * allow it to do that as well.
1766 if (ifa->ifa_rtrequest)
1767 ifa->ifa_rtrequest(req, rt, info);
1770 * actually return a resultant rtentry and
1771 * give the caller a single reference.
1777 rnh->rnh_gen++; /* Routing table updated */
1782 error = rtrequest1_fib_change(rnh, info, ret_nrt, fibnum);
1800 rtrequest1_fib_change(struct rib_head *rnh, struct rt_addrinfo *info,
1801 struct rtentry **ret_nrt, u_int fibnum)
1803 struct rtentry *rt = NULL;
1807 struct if_mtuinfo ifmtu;
1809 rt = (struct rtentry *)rnh->rnh_lookup(info->rti_info[RTAX_DST],
1810 info->rti_info[RTAX_NETMASK], &rnh->head);
1817 * If we got multipath routes,
1818 * we require users to specify a matching RTAX_GATEWAY.
1820 if (rt_mpath_capable(rnh)) {
1821 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
1829 rt_setmetrics(info, rt);
1832 * New gateway could require new ifaddr, ifp;
1833 * flags may also be different; ifp may be specified
1834 * by ll sockaddr when protocol address is ambiguous
1836 if (((rt->rt_flags & RTF_GATEWAY) &&
1837 info->rti_info[RTAX_GATEWAY] != NULL) ||
1838 info->rti_info[RTAX_IFP] != NULL ||
1839 (info->rti_info[RTAX_IFA] != NULL &&
1840 !sa_equal(info->rti_info[RTAX_IFA], rt->rt_ifa->ifa_addr))) {
1842 error = rt_getifa_fib(info, fibnum);
1843 if (info->rti_ifa != NULL)
1850 /* Check if outgoing interface has changed */
1851 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa &&
1852 rt->rt_ifa != NULL && rt->rt_ifa->ifa_rtrequest != NULL) {
1853 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1854 ifa_free(rt->rt_ifa);
1856 /* Update gateway address */
1857 if (info->rti_info[RTAX_GATEWAY] != NULL) {
1858 error = rt_setgate(rt, rt_key(rt), info->rti_info[RTAX_GATEWAY]);
1862 rt->rt_flags &= ~RTF_GATEWAY;
1863 rt->rt_flags |= (RTF_GATEWAY & info->rti_flags);
1866 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa) {
1867 ifa_ref(info->rti_ifa);
1868 rt->rt_ifa = info->rti_ifa;
1869 rt->rt_ifp = info->rti_ifp;
1871 /* Allow some flags to be toggled on change. */
1872 rt->rt_flags &= ~RTF_FMASK;
1873 rt->rt_flags |= info->rti_flags & RTF_FMASK;
1875 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest != NULL)
1876 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
1878 /* Alter route MTU if necessary */
1879 if (rt->rt_ifp != NULL) {
1880 family = info->rti_info[RTAX_DST]->sa_family;
1881 mtu = if_getmtu_family(rt->rt_ifp, family);
1882 /* Set default MTU */
1883 if (rt->rt_mtu == 0)
1885 if (rt->rt_mtu != mtu) {
1886 /* Check if we really need to update */
1887 ifmtu.ifp = rt->rt_ifp;
1889 if_updatemtu_cb(rt->rt_nodes, &ifmtu);
1900 ifa_free(info->rti_ifa);
1905 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt)
1908 if (info->rti_mflags & RTV_MTU) {
1909 if (info->rti_rmx->rmx_mtu != 0) {
1912 * MTU was explicitly provided by user.
1915 rt->rt_flags |= RTF_FIXEDMTU;
1919 * User explicitly sets MTU to 0.
1920 * Assume rollback to default.
1922 rt->rt_flags &= ~RTF_FIXEDMTU;
1924 rt->rt_mtu = info->rti_rmx->rmx_mtu;
1926 if (info->rti_mflags & RTV_WEIGHT)
1927 rt->rt_weight = info->rti_rmx->rmx_weight;
1928 /* Kernel -> userland timebase conversion. */
1929 if (info->rti_mflags & RTV_EXPIRE)
1930 rt->rt_expire = info->rti_rmx->rmx_expire ?
1931 info->rti_rmx->rmx_expire - time_second + time_uptime : 0;
1935 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
1937 /* XXX dst may be overwritten, can we move this to below */
1938 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
1941 * Prepare to store the gateway in rt->rt_gateway.
1942 * Both dst and gateway are stored one after the other in the same
1943 * malloc'd chunk. If we have room, we can reuse the old buffer,
1944 * rt_gateway already points to the right place.
1945 * Otherwise, malloc a new block and update the 'dst' address.
1947 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
1950 R_Malloc(new, caddr_t, dlen + glen);
1954 * XXX note, we copy from *dst and not *rt_key(rt) because
1955 * rt_setgate() can be called to initialize a newly
1956 * allocated route entry, in which case rt_key(rt) == NULL
1957 * (and also rt->rt_gateway == NULL).
1958 * Free()/free() handle a NULL argument just fine.
1960 bcopy(dst, new, dlen);
1961 R_Free(rt_key(rt)); /* free old block, if any */
1962 rt_key(rt) = (struct sockaddr *)new;
1963 rt->rt_gateway = (struct sockaddr *)(new + dlen);
1967 * Copy the new gateway value into the memory chunk.
1969 bcopy(gate, rt->rt_gateway, glen);
1975 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
1977 u_char *cp1 = (u_char *)src;
1978 u_char *cp2 = (u_char *)dst;
1979 u_char *cp3 = (u_char *)netmask;
1980 u_char *cplim = cp2 + *cp3;
1981 u_char *cplim2 = cp2 + *cp1;
1983 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1988 *cp2++ = *cp1++ & *cp3++;
1990 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1994 * Set up a routing table entry, normally
1997 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
1999 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
2001 struct sockaddr *dst;
2002 struct sockaddr *netmask;
2003 struct rtentry *rt = NULL;
2004 struct rt_addrinfo info;
2006 int startfib, endfib;
2007 char tempbuf[_SOCKADDR_TMPSIZE];
2010 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
2011 struct rib_head *rnh;
2013 if (flags & RTF_HOST) {
2014 dst = ifa->ifa_dstaddr;
2017 dst = ifa->ifa_addr;
2018 netmask = ifa->ifa_netmask;
2020 if (dst->sa_len == 0)
2022 switch (dst->sa_family) {
2025 /* We support multiple FIBs. */
2028 fibnum = RT_DEFAULT_FIB;
2031 if (fibnum == RT_ALL_FIBS) {
2032 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD)
2033 startfib = endfib = ifa->ifa_ifp->if_fib;
2036 endfib = rt_numfibs - 1;
2039 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
2045 * If it's a delete, check that if it exists,
2046 * it's on the correct interface or we might scrub
2047 * a route to another ifa which would
2048 * be confusing at best and possibly worse.
2050 if (cmd == RTM_DELETE) {
2052 * It's a delete, so it should already exist..
2053 * If it's a net, mask off the host bits
2054 * (Assuming we have a mask)
2055 * XXX this is kinda inet specific..
2057 if (netmask != NULL) {
2058 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
2059 dst = (struct sockaddr *)tempbuf;
2063 * Now go through all the requested tables (fibs) and do the
2064 * requested action. Realistically, this will either be fib 0
2065 * for protocols that don't do multiple tables or all the
2066 * tables for those that do.
2068 for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
2069 if (cmd == RTM_DELETE) {
2070 struct radix_node *rn;
2072 * Look up an rtentry that is in the routing tree and
2073 * contains the correct info.
2075 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
2077 /* this table doesn't exist but others might */
2080 rn = rnh->rnh_lookup(dst, netmask, &rnh->head);
2082 if (rt_mpath_capable(rnh)) {
2089 * for interface route the
2090 * rt->rt_gateway is sockaddr_intf
2091 * for cloning ARP entries, so
2092 * rt_mpath_matchgate must use the
2095 rt = rt_mpath_matchgate(rt,
2102 error = (rn == NULL ||
2103 (rn->rn_flags & RNF_ROOT) ||
2104 RNTORT(rn)->rt_ifa != ifa);
2107 /* this is only an error if bad on ALL tables */
2112 * Do the actual request
2114 bzero((caddr_t)&info, sizeof(info));
2116 info.rti_flags = flags |
2117 (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED;
2118 info.rti_info[RTAX_DST] = dst;
2120 * doing this for compatibility reasons
2123 info.rti_info[RTAX_GATEWAY] =
2124 (struct sockaddr *)&null_sdl;
2126 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
2127 info.rti_info[RTAX_NETMASK] = netmask;
2128 error = rtrequest1_fib(cmd, &info, &rt, fibnum);
2130 if (error == 0 && rt != NULL) {
2132 * notify any listening routing agents of the change
2137 * in case address alias finds the first address
2138 * e.g. ifconfig bge0 192.0.2.246/24
2139 * e.g. ifconfig bge0 192.0.2.247/24
2140 * the address set in the route is 192.0.2.246
2141 * so we need to replace it with 192.0.2.247
2143 if (memcmp(rt->rt_ifa->ifa_addr,
2144 ifa->ifa_addr, ifa->ifa_addr->sa_len)) {
2145 ifa_free(rt->rt_ifa);
2147 rt->rt_ifp = ifa->ifa_ifp;
2152 * doing this for compatibility reasons
2154 if (cmd == RTM_ADD) {
2155 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
2156 rt->rt_ifp->if_type;
2157 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
2158 rt->rt_ifp->if_index;
2162 rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum);
2165 if (cmd == RTM_DELETE) {
2167 * If we are deleting, and we found an entry,
2168 * then it's been removed from the tree..
2169 * now throw it away.
2173 if (cmd == RTM_ADD) {
2175 * We just wanted to add it..
2176 * we don't actually need a reference.
2187 if (cmd == RTM_DELETE) {
2191 /* we only give an error if it wasn't in any table */
2192 error = ((flags & RTF_HOST) ?
2193 EHOSTUNREACH : ENETUNREACH);
2197 /* return an error if any of them failed */
2205 * Set up a routing table entry, normally
2209 rtinit(struct ifaddr *ifa, int cmd, int flags)
2211 struct sockaddr *dst;
2212 int fib = RT_DEFAULT_FIB;
2214 if (flags & RTF_HOST) {
2215 dst = ifa->ifa_dstaddr;
2217 dst = ifa->ifa_addr;
2220 switch (dst->sa_family) {
2223 /* We do support multiple FIBs. */
2227 return (rtinit1(ifa, cmd, flags, fib));
2231 * Announce interface address arrival/withdraw
2232 * Returns 0 on success.
2235 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
2238 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2239 ("unexpected cmd %d", cmd));
2241 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2242 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2244 #if defined(INET) || defined(INET6)
2247 * notify the SCTP stack
2248 * this will only get called when an address is added/deleted
2249 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
2251 sctp_addr_change(ifa, cmd);
2254 return (rtsock_addrmsg(cmd, ifa, fibnum));
2258 * Announce route addition/removal.
2259 * Users of this function MUST validate input data BEFORE calling.
2260 * However we have to be able to handle invalid data:
2261 * if some userland app sends us "invalid" route message (invalid mask,
2262 * no dst, wrong address families, etc...) we need to pass it back
2263 * to app (and any other rtsock consumers) with rtm_errno field set to
2265 * Returns 0 on success.
2268 rt_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
2272 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2273 ("unexpected cmd %d", cmd));
2275 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2276 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2278 KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
2280 return (rtsock_routemsg(cmd, ifp, error, rt, fibnum));
2284 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
2287 rt_newaddrmsg_fib(cmd, ifa, error, rt, RT_ALL_FIBS);
2291 * This is called to generate messages from the routing socket
2292 * indicating a network interface has had addresses associated with it.
2295 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
2299 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2300 ("unexpected cmd %u", cmd));
2301 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2302 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2304 if (cmd == RTM_ADD) {
2305 rt_addrmsg(cmd, ifa, fibnum);
2307 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2310 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2311 rt_addrmsg(cmd, ifa, fibnum);