2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1980, 1986, 1991, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)route.c 8.3.1.1 (Berkeley) 2/23/95
34 /************************************************************************
35 * Note: In this file a 'fib' is a "forwarding information base" *
36 * Which is the new name for an in kernel routing (next hop) table. *
37 ***********************************************************************/
40 #include "opt_inet6.h"
41 #include "opt_route.h"
43 #include "opt_mrouting.h"
44 #include "opt_mpath.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/malloc.h>
50 #include <sys/socket.h>
51 #include <sys/sysctl.h>
52 #include <sys/syslog.h>
53 #include <sys/sysproto.h>
55 #include <sys/domain.h>
56 #include <sys/kernel.h>
59 #include <net/if_var.h>
60 #include <net/if_dl.h>
61 #include <net/route.h>
62 #include <net/route_var.h>
66 #include <net/radix_mpath.h>
69 #include <netinet/in.h>
70 #include <netinet/ip_mroute.h>
74 #define RT_MAXFIBS UINT16_MAX
76 /* Kernel config default option. */
79 #error "ROUTETABLES defined too low"
81 #if ROUTETABLES > RT_MAXFIBS
82 #error "ROUTETABLES defined too big"
84 #define RT_NUMFIBS ROUTETABLES
85 #endif /* ROUTETABLES */
86 /* Initialize to default if not otherwise set. */
91 #if defined(INET) || defined(INET6)
93 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
98 /* This is read-only.. */
99 u_int rt_numfibs = RT_NUMFIBS;
100 SYSCTL_UINT(_net, OID_AUTO, fibs, CTLFLAG_RDTUN, &rt_numfibs, 0, "");
103 * By default add routes to all fibs for new interfaces.
104 * Once this is set to 0 then only allocate routes on interface
105 * changes for the FIB of the caller when adding a new set of addresses
106 * to an interface. XXX this is a shotgun aproach to a problem that needs
107 * a more fine grained solution.. that will come.
108 * XXX also has the problems getting the FIB from curthread which will not
109 * always work given the fib can be overridden and prefixes can be added
110 * from the network stack context.
112 VNET_DEFINE(u_int, rt_add_addr_allfibs) = 1;
113 SYSCTL_UINT(_net, OID_AUTO, add_addr_allfibs, CTLFLAG_RWTUN | CTLFLAG_VNET,
114 &VNET_NAME(rt_add_addr_allfibs), 0, "");
116 VNET_DEFINE(struct rtstat, rtstat);
117 #define V_rtstat VNET(rtstat)
119 VNET_DEFINE(struct rib_head *, rt_tables);
120 #define V_rt_tables VNET(rt_tables)
122 VNET_DEFINE(int, rttrash); /* routes not in table but not freed */
123 #define V_rttrash VNET(rttrash)
127 * Convert a 'struct radix_node *' to a 'struct rtentry *'.
128 * The operation can be done safely (in this code) because a
129 * 'struct rtentry' starts with two 'struct radix_node''s, the first
130 * one representing leaf nodes in the routing tree, which is
131 * what the code in radix.c passes us as a 'struct radix_node'.
133 * But because there are a lot of assumptions in this conversion,
134 * do not cast explicitly, but always use the macro below.
136 #define RNTORT(p) ((struct rtentry *)(p))
138 static VNET_DEFINE(uma_zone_t, rtzone); /* Routing table UMA zone. */
139 #define V_rtzone VNET(rtzone)
141 static int rtrequest1_fib_change(struct rib_head *, struct rt_addrinfo *,
142 struct rtentry **, u_int);
143 static void rt_setmetrics(const struct rt_addrinfo *, struct rtentry *);
144 static int rt_ifdelroute(const struct rtentry *rt, void *arg);
145 static struct rtentry *rt_unlinkrte(struct rib_head *rnh,
146 struct rt_addrinfo *info, int *perror);
147 static void rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info);
149 static struct radix_node *rt_mpath_unlink(struct rib_head *rnh,
150 struct rt_addrinfo *info, struct rtentry *rto, int *perror);
152 static int rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info,
161 static int if_updatemtu_cb(struct radix_node *, void *);
164 * handler for net.my_fibnum
167 sysctl_my_fibnum(SYSCTL_HANDLER_ARGS)
172 fibnum = curthread->td_proc->p_fibnum;
173 error = sysctl_handle_int(oidp, &fibnum, 0, req);
177 SYSCTL_PROC(_net, OID_AUTO, my_fibnum, CTLTYPE_INT|CTLFLAG_RD,
178 NULL, 0, &sysctl_my_fibnum, "I", "default FIB of caller");
180 static __inline struct rib_head **
181 rt_tables_get_rnh_ptr(int table, int fam)
183 struct rib_head **rnh;
185 KASSERT(table >= 0 && table < rt_numfibs, ("%s: table out of bounds.",
187 KASSERT(fam >= 0 && fam < (AF_MAX+1), ("%s: fam out of bounds.",
190 /* rnh is [fib=0][af=0]. */
191 rnh = (struct rib_head **)V_rt_tables;
192 /* Get the offset to the requested table and fam. */
193 rnh += table * (AF_MAX+1) + fam;
199 rt_tables_get_rnh(int table, int fam)
202 return (*rt_tables_get_rnh_ptr(table, fam));
206 rt_tables_get_gen(int table, int fam)
208 struct rib_head *rnh;
210 rnh = *rt_tables_get_rnh_ptr(table, fam);
211 KASSERT(rnh != NULL, ("%s: NULL rib_head pointer table %d fam %d",
212 __func__, table, fam));
213 return (rnh->rnh_gen);
218 * route initialization must occur before ip6_init2(), which happenas at
225 /* whack the tunable ints into line. */
226 if (rt_numfibs > RT_MAXFIBS)
227 rt_numfibs = RT_MAXFIBS;
231 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0);
234 rtentry_zinit(void *mem, int size, int how)
236 struct rtentry *rt = mem;
238 rt->rt_pksent = counter_u64_alloc(how);
239 if (rt->rt_pksent == NULL)
248 rtentry_zfini(void *mem, int size)
250 struct rtentry *rt = mem;
253 counter_u64_free(rt->rt_pksent);
257 rtentry_ctor(void *mem, int size, void *arg, int how)
259 struct rtentry *rt = mem;
261 bzero(rt, offsetof(struct rtentry, rt_endzero));
262 counter_u64_zero(rt->rt_pksent);
269 rtentry_dtor(void *mem, int size, void *arg)
271 struct rtentry *rt = mem;
277 vnet_route_init(const void *unused __unused)
280 struct rib_head **rnh;
284 V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) *
285 sizeof(struct rib_head *), M_RTABLE, M_WAITOK|M_ZERO);
287 V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry),
288 rtentry_ctor, rtentry_dtor,
289 rtentry_zinit, rtentry_zfini, UMA_ALIGN_PTR, 0);
290 for (dom = domains; dom; dom = dom->dom_next) {
291 if (dom->dom_rtattach == NULL)
294 for (table = 0; table < rt_numfibs; table++) {
295 fam = dom->dom_family;
296 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
299 rnh = rt_tables_get_rnh_ptr(table, fam);
301 panic("%s: rnh NULL", __func__);
302 dom->dom_rtattach((void **)rnh, 0);
306 VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
311 vnet_route_uninit(const void *unused __unused)
316 struct rib_head **rnh;
318 for (dom = domains; dom; dom = dom->dom_next) {
319 if (dom->dom_rtdetach == NULL)
322 for (table = 0; table < rt_numfibs; table++) {
323 fam = dom->dom_family;
325 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
328 rnh = rt_tables_get_rnh_ptr(table, fam);
330 panic("%s: rnh NULL", __func__);
331 dom->dom_rtdetach((void **)rnh, 0);
335 free(V_rt_tables, M_RTABLE);
336 uma_zdestroy(V_rtzone);
338 VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST,
339 vnet_route_uninit, 0);
343 rt_table_init(int offset)
347 rh = malloc(sizeof(struct rib_head), M_RTABLE, M_WAITOK | M_ZERO);
349 /* TODO: These details should be hidded inside radix.c */
350 /* Init masks tree */
351 rn_inithead_internal(&rh->head, rh->rnh_nodes, offset);
352 rn_inithead_internal(&rh->rmhead.head, rh->rmhead.mask_nodes, 0);
353 rh->head.rnh_masks = &rh->rmhead;
358 /* Finally, set base callbacks */
359 rh->rnh_addaddr = rn_addroute;
360 rh->rnh_deladdr = rn_delete;
361 rh->rnh_matchaddr = rn_match;
362 rh->rnh_lookup = rn_lookup;
363 rh->rnh_walktree = rn_walktree;
364 rh->rnh_walktree_from = rn_walktree_from;
370 rt_freeentry(struct radix_node *rn, void *arg)
372 struct radix_head * const rnh = arg;
373 struct radix_node *x;
375 x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh);
382 rt_table_destroy(struct rib_head *rh)
385 rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head);
387 /* Assume table is already empty */
388 RIB_LOCK_DESTROY(rh);
393 #ifndef _SYS_SYSPROTO_H_
399 sys_setfib(struct thread *td, struct setfib_args *uap)
401 if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs)
403 td->td_proc->p_fibnum = uap->fibnum;
408 * Packet routing routines.
411 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum)
415 if ((rt = ro->ro_rt) != NULL) {
416 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
421 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum);
423 RT_UNLOCK(ro->ro_rt);
427 * Look up the route that matches the address given
428 * Or, at least try.. Create a cloned route if needed.
430 * The returned route, if any, is locked.
433 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
436 return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB));
440 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags,
444 struct radix_node *rn;
445 struct rtentry *newrt;
446 struct rt_addrinfo info;
447 int err = 0, msgtype = RTM_MISS;
449 KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum"));
450 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
456 * Look up the address in the table for that Address Family
458 if ((ignflags & RTF_RNH_LOCKED) == 0)
464 rn = rh->rnh_matchaddr(dst, &rh->head);
465 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
469 if ((ignflags & RTF_RNH_LOCKED) == 0)
473 } else if ((ignflags & RTF_RNH_LOCKED) == 0)
476 * Either we hit the root or could not find any match,
477 * which basically means: "cannot get there from here".
480 V_rtstat.rts_unreach++;
484 * If required, report the failure to the supervising
486 * For a delete, this is not an error. (report == 0)
488 bzero(&info, sizeof(info));
489 info.rti_info[RTAX_DST] = dst;
490 rt_missmsg_fib(msgtype, &info, 0, err, fibnum);
496 * Remove a reference count from an rtentry.
497 * If the count gets low enough, take it out of the routing table
500 rtfree(struct rtentry *rt)
502 struct rib_head *rnh;
504 KASSERT(rt != NULL,("%s: NULL rt", __func__));
505 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
506 KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
511 * The callers should use RTFREE_LOCKED() or RTFREE(), so
512 * we should come here exactly with the last reference.
515 if (rt->rt_refcnt > 0) {
516 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt);
521 * On last reference give the "close method" a chance
522 * to cleanup private state. This also permits (for
523 * IPv4 and IPv6) a chance to decide if the routing table
524 * entry should be purged immediately or at a later time.
525 * When an immediate purge is to happen the close routine
526 * typically calls rtexpunge which clears the RTF_UP flag
527 * on the entry so that the code below reclaims the storage.
529 if (rt->rt_refcnt == 0 && rnh->rnh_close)
530 rnh->rnh_close((struct radix_node *)rt, &rnh->head);
533 * If we are no longer "up" (and ref == 0)
534 * then we can free the resources associated
537 if ((rt->rt_flags & RTF_UP) == 0) {
538 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
541 * the rtentry must have been removed from the routing table
542 * so it is represented in rttrash.. remove that now.
546 if (rt->rt_refcnt < 0) {
547 printf("rtfree: %p not freed (neg refs)\n", rt);
552 * release references on items we hold them on..
553 * e.g other routes and ifaddrs.
556 ifa_free(rt->rt_ifa);
558 * The key is separatly alloc'd so free it (see rt_setgate()).
559 * This also frees the gateway, as they are always malloc'd
565 * and the rtentry itself of course
567 uma_zfree(V_rtzone, rt);
576 * Force a routing table entry to the specified
577 * destination to go through the given gateway.
578 * Normally called as a result of a routing redirect
579 * message from the network layer.
582 rtredirect_fib(struct sockaddr *dst,
583 struct sockaddr *gateway,
584 struct sockaddr *netmask,
586 struct sockaddr *src,
592 struct rt_addrinfo info;
594 struct rib_head *rnh;
597 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
599 error = EAFNOSUPPORT;
603 /* verify the gateway is directly reachable */
604 if ((ifa = ifa_ifwithnet(gateway, 0, fibnum)) == NULL) {
608 rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */
610 * If the redirect isn't from our current router for this dst,
611 * it's either old or wrong. If it redirects us to ourselves,
612 * we have a routing loop, perhaps as a result of an interface
613 * going down recently.
615 if (!(flags & RTF_DONE) && rt) {
616 if (!sa_equal(src, rt->rt_gateway)) {
620 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) {
625 if ((flags & RTF_GATEWAY) && ifa_ifwithaddr_check(gateway)) {
626 error = EHOSTUNREACH;
630 * Create a new entry if we just got back a wildcard entry
631 * or the lookup failed. This is necessary for hosts
632 * which use routing redirects generated by smart gateways
633 * to dynamically build the routing tables.
635 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
638 * Don't listen to the redirect if it's
639 * for a route to an interface.
641 if (rt->rt_flags & RTF_GATEWAY) {
642 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
644 * Changing from route to net => route to host.
645 * Create new route, rather than smashing route to net.
651 flags |= RTF_DYNAMIC;
652 bzero((caddr_t)&info, sizeof(info));
653 info.rti_info[RTAX_DST] = dst;
654 info.rti_info[RTAX_GATEWAY] = gateway;
655 info.rti_info[RTAX_NETMASK] = netmask;
657 info.rti_flags = flags;
658 error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum);
661 flags = rt->rt_flags;
664 stat = &V_rtstat.rts_dynamic;
668 * Smash the current notion of the gateway to
669 * this destination. Should check about netmask!!!
671 if ((flags & RTF_GATEWAY) == 0)
672 rt->rt_flags &= ~RTF_GATEWAY;
673 rt->rt_flags |= RTF_MODIFIED;
674 flags |= RTF_MODIFIED;
675 stat = &V_rtstat.rts_newgateway;
677 * add the key and gateway (in one malloc'd chunk).
682 rt_setgate(rt, rt_key(rt), gateway);
686 error = EHOSTUNREACH;
692 V_rtstat.rts_badredirect++;
693 else if (stat != NULL)
695 bzero((caddr_t)&info, sizeof(info));
696 info.rti_info[RTAX_DST] = dst;
697 info.rti_info[RTAX_GATEWAY] = gateway;
698 info.rti_info[RTAX_NETMASK] = netmask;
699 info.rti_info[RTAX_AUTHOR] = src;
700 rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum);
706 * Routing table ioctl interface.
709 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
713 * If more ioctl commands are added here, make sure the proper
714 * super-user checks are being performed because it is possible for
715 * prison-root to make it this far if raw sockets have been enabled
719 /* Multicast goop, grrr... */
720 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
727 ifa_ifwithroute(int flags, const struct sockaddr *dst, struct sockaddr *gateway,
733 if ((flags & RTF_GATEWAY) == 0) {
735 * If we are adding a route to an interface,
736 * and the interface is a pt to pt link
737 * we should search for the destination
738 * as our clue to the interface. Otherwise
739 * we can use the local address.
742 if (flags & RTF_HOST)
743 ifa = ifa_ifwithdstaddr(dst, fibnum);
745 ifa = ifa_ifwithaddr(gateway);
748 * If we are adding a route to a remote net
749 * or host, the gateway may still be on the
750 * other end of a pt to pt link.
752 ifa = ifa_ifwithdstaddr(gateway, fibnum);
755 ifa = ifa_ifwithnet(gateway, 0, fibnum);
759 rt = rtalloc1_fib(gateway, 0, flags, fibnum);
763 * dismiss a gateway that is reachable only
764 * through the default router
766 switch (gateway->sa_family) {
768 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
772 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
778 if (!not_found && rt->rt_ifa != NULL) {
784 if (not_found || ifa == NULL)
787 if (ifa->ifa_addr->sa_family != dst->sa_family) {
788 struct ifaddr *oifa = ifa;
789 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
799 * Do appropriate manipulations of a routing tree given
800 * all the bits of info needed
803 rtrequest_fib(int req,
804 struct sockaddr *dst,
805 struct sockaddr *gateway,
806 struct sockaddr *netmask,
808 struct rtentry **ret_nrt,
811 struct rt_addrinfo info;
813 if (dst->sa_len == 0)
816 bzero((caddr_t)&info, sizeof(info));
817 info.rti_flags = flags;
818 info.rti_info[RTAX_DST] = dst;
819 info.rti_info[RTAX_GATEWAY] = gateway;
820 info.rti_info[RTAX_NETMASK] = netmask;
821 return rtrequest1_fib(req, &info, ret_nrt, fibnum);
826 * Copy most of @rt data into @info.
828 * If @flags contains NHR_COPY, copies dst,netmask and gw to the
829 * pointers specified by @info structure. Assume such pointers
830 * are zeroed sockaddr-like structures with sa_len field initialized
831 * to reflect size of the provided buffer. if no NHR_COPY is specified,
832 * point dst,netmask and gw @info fields to appropriate @rt values.
834 * if @flags contains NHR_REF, do refcouting on rt_ifp.
836 * Returns 0 on success.
839 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags)
841 struct rt_metrics *rmx;
842 struct sockaddr *src, *dst;
845 if (flags & NHR_COPY) {
846 /* Copy destination if dst is non-zero */
848 dst = info->rti_info[RTAX_DST];
849 sa_len = src->sa_len;
851 if (src->sa_len > dst->sa_len)
853 memcpy(dst, src, src->sa_len);
854 info->rti_addrs |= RTA_DST;
857 /* Copy mask if set && dst is non-zero */
859 dst = info->rti_info[RTAX_NETMASK];
860 if (src != NULL && dst != NULL) {
863 * Radix stores different value in sa_len,
864 * assume rt_mask() to have the same length
867 if (sa_len > dst->sa_len)
869 memcpy(dst, src, src->sa_len);
870 info->rti_addrs |= RTA_NETMASK;
873 /* Copy gateway is set && dst is non-zero */
874 src = rt->rt_gateway;
875 dst = info->rti_info[RTAX_GATEWAY];
876 if ((rt->rt_flags & RTF_GATEWAY) && src != NULL && dst != NULL){
877 if (src->sa_len > dst->sa_len)
879 memcpy(dst, src, src->sa_len);
880 info->rti_addrs |= RTA_GATEWAY;
883 info->rti_info[RTAX_DST] = rt_key(rt);
884 info->rti_addrs |= RTA_DST;
885 if (rt_mask(rt) != NULL) {
886 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
887 info->rti_addrs |= RTA_NETMASK;
889 if (rt->rt_flags & RTF_GATEWAY) {
890 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
891 info->rti_addrs |= RTA_GATEWAY;
897 info->rti_mflags |= RTV_MTU;
898 rmx->rmx_mtu = rt->rt_mtu;
901 info->rti_flags = rt->rt_flags;
902 info->rti_ifp = rt->rt_ifp;
903 info->rti_ifa = rt->rt_ifa;
905 if (flags & NHR_REF) {
906 /* Do 'traditional' refcouting */
907 if_ref(info->rti_ifp);
914 * Lookups up route entry for @dst in RIB database for fib @fibnum.
915 * Exports entry data to @info using rt_exportinfo().
917 * if @flags contains NHR_REF, refcouting is performed on rt_ifp.
918 * All references can be released later by calling rib_free_info()
920 * Returns 0 on success.
921 * Returns ENOENT for lookup failure, ENOMEM for export failure.
924 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
925 uint32_t flowid, struct rt_addrinfo *info)
928 struct radix_node *rn;
932 KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
933 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
938 rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head);
939 if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
941 /* Ensure route & ifp is UP */
942 if (RT_LINK_IS_UP(rt->rt_ifp)) {
943 flags = (flags & NHR_REF) | NHR_COPY;
944 error = rt_exportinfo(rt, info, flags);
956 * Releases all references acquired by rib_lookup_info() when
957 * called with NHR_REF flags.
960 rib_free_info(struct rt_addrinfo *info)
963 if_rele(info->rti_ifp);
967 * Iterates over all existing fibs in system calling
968 * @setwa_f function prior to traversing each fib.
969 * Calls @wa_f function for each element in current fib.
970 * If af is not AF_UNSPEC, iterates over fibs in particular
974 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f,
977 struct rib_head *rnh;
981 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
982 /* Do we want some specific family? */
983 if (af != AF_UNSPEC) {
984 rnh = rt_tables_get_rnh(fibnum, af);
988 setwa_f(rnh, fibnum, af, arg);
991 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
996 for (i = 1; i <= AF_MAX; i++) {
997 rnh = rt_tables_get_rnh(fibnum, i);
1000 if (setwa_f != NULL)
1001 setwa_f(rnh, fibnum, i, arg);
1004 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
1012 struct rt_addrinfo info;
1013 struct rib_head *rnh;
1014 struct rtentry *head;
1018 * Conditionally unlinks @rn from radix tree based
1019 * on info data passed in @arg.
1022 rt_checkdelroute(struct radix_node *rn, void *arg)
1024 struct rt_delinfo *di;
1025 struct rt_addrinfo *info;
1029 di = (struct rt_delinfo *)arg;
1030 rt = (struct rtentry *)rn;
1034 info->rti_info[RTAX_DST] = rt_key(rt);
1035 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
1036 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1038 rt = rt_unlinkrte(di->rnh, info, &error);
1040 /* Either not allowed or not matched. Skip entry */
1044 /* Entry was unlinked. Add to the list and return */
1045 rt->rt_chain = di->head;
1052 * Iterates over all existing fibs in system.
1053 * Deletes each element for which @filter_f function returned
1055 * If @af is not AF_UNSPEC, iterates over fibs in particular
1059 rt_foreach_fib_walk_del(int af, rt_filter_f_t *filter_f, void *arg)
1061 struct rib_head *rnh;
1062 struct rt_delinfo di;
1067 bzero(&di, sizeof(di));
1068 di.info.rti_filter = filter_f;
1069 di.info.rti_filterdata = arg;
1071 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1072 /* Do we want some specific family? */
1073 if (af != AF_UNSPEC) {
1081 for (i = start; i <= end; i++) {
1082 rnh = rt_tables_get_rnh(fibnum, i);
1088 rnh->rnh_walktree(&rnh->head, rt_checkdelroute, &di);
1091 if (di.head == NULL)
1094 /* We might have something to reclaim */
1095 while (di.head != NULL) {
1097 di.head = rt->rt_chain;
1098 rt->rt_chain = NULL;
1100 /* TODO std rt -> rt_addrinfo export */
1101 di.info.rti_info[RTAX_DST] = rt_key(rt);
1102 di.info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1104 rt_notifydelete(rt, &di.info);
1113 * Delete Routes for a Network Interface
1115 * Called for each routing entry via the rnh->rnh_walktree() call above
1116 * to delete all route entries referencing a detaching network interface.
1119 * rt pointer to rtentry
1120 * arg argument passed to rnh->rnh_walktree() - detaching interface
1124 * errno failed - reason indicated
1127 rt_ifdelroute(const struct rtentry *rt, void *arg)
1129 struct ifnet *ifp = arg;
1131 if (rt->rt_ifp != ifp)
1135 * Protect (sorta) against walktree recursion problems
1136 * with cloned routes
1138 if ((rt->rt_flags & RTF_UP) == 0)
1145 * Delete all remaining routes using this interface
1146 * Unfortuneatly the only way to do this is to slog through
1147 * the entire routing table looking for routes which point
1148 * to this interface...oh well...
1151 rt_flushifroutes_af(struct ifnet *ifp, int af)
1153 KASSERT((af >= 1 && af <= AF_MAX), ("%s: af %d not >= 1 and <= %d",
1154 __func__, af, AF_MAX));
1156 rt_foreach_fib_walk_del(af, rt_ifdelroute, ifp);
1160 rt_flushifroutes(struct ifnet *ifp)
1163 rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
1167 * Conditionally unlinks rtentry matching data inside @info from @rnh.
1168 * Returns unlinked, locked and referenced @rtentry on success,
1169 * Returns NULL and sets @perror to:
1170 * ESRCH - if prefix was not found,
1171 * EADDRINUSE - if trying to delete PINNED route without appropriate flag.
1172 * ENOENT - if supplied filter function returned 0 (not matched).
1174 static struct rtentry *
1175 rt_unlinkrte(struct rib_head *rnh, struct rt_addrinfo *info, int *perror)
1177 struct sockaddr *dst, *netmask;
1179 struct radix_node *rn;
1181 dst = info->rti_info[RTAX_DST];
1182 netmask = info->rti_info[RTAX_NETMASK];
1184 rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, &rnh->head);
1190 if ((info->rti_flags & RTF_PINNED) == 0) {
1191 /* Check if target route can be deleted */
1192 if (rt->rt_flags & RTF_PINNED) {
1193 *perror = EADDRINUSE;
1198 if (info->rti_filter != NULL) {
1199 if (info->rti_filter(rt, info->rti_filterdata) == 0) {
1206 * Filter function requested rte deletion.
1207 * Ease the caller work by filling in remaining info
1208 * from that particular entry.
1210 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1214 * Remove the item from the tree and return it.
1215 * Complain if it is not there and do no more processing.
1219 if (rt_mpath_capable(rnh))
1220 rn = rt_mpath_unlink(rnh, info, rt, perror);
1223 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1227 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
1228 panic ("rtrequest delete");
1233 rt->rt_flags &= ~RTF_UP;
1241 rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info)
1246 * give the protocol a chance to keep things in sync.
1249 if (ifa != NULL && ifa->ifa_rtrequest != NULL)
1250 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1253 * One more rtentry floating around that is not
1254 * linked to the routing table. rttrash will be decremented
1255 * when RTFREE(rt) is eventually called.
1262 * These (questionable) definitions of apparent local variables apply
1263 * to the next two functions. XXXXXX!!!
1265 #define dst info->rti_info[RTAX_DST]
1266 #define gateway info->rti_info[RTAX_GATEWAY]
1267 #define netmask info->rti_info[RTAX_NETMASK]
1268 #define ifaaddr info->rti_info[RTAX_IFA]
1269 #define ifpaddr info->rti_info[RTAX_IFP]
1270 #define flags info->rti_flags
1273 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined,
1274 * it will be referenced so the caller must free it.
1277 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
1283 * ifp may be specified by sockaddr_dl
1284 * when protocol address is ambiguous.
1286 if (info->rti_ifp == NULL && ifpaddr != NULL &&
1287 ifpaddr->sa_family == AF_LINK &&
1288 (ifa = ifa_ifwithnet(ifpaddr, 0, fibnum)) != NULL) {
1289 info->rti_ifp = ifa->ifa_ifp;
1292 if (info->rti_ifa == NULL && ifaaddr != NULL)
1293 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
1294 if (info->rti_ifa == NULL) {
1295 struct sockaddr *sa;
1297 sa = ifaaddr != NULL ? ifaaddr :
1298 (gateway != NULL ? gateway : dst);
1299 if (sa != NULL && info->rti_ifp != NULL)
1300 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
1301 else if (dst != NULL && gateway != NULL)
1302 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway,
1304 else if (sa != NULL)
1305 info->rti_ifa = ifa_ifwithroute(flags, sa, sa,
1308 if ((ifa = info->rti_ifa) != NULL) {
1309 if (info->rti_ifp == NULL)
1310 info->rti_ifp = ifa->ifa_ifp;
1312 error = ENETUNREACH;
1317 if_updatemtu_cb(struct radix_node *rn, void *arg)
1320 struct if_mtuinfo *ifmtu;
1322 rt = (struct rtentry *)rn;
1323 ifmtu = (struct if_mtuinfo *)arg;
1325 if (rt->rt_ifp != ifmtu->ifp)
1328 if (rt->rt_mtu >= ifmtu->mtu) {
1329 /* We have to decrease mtu regardless of flags */
1330 rt->rt_mtu = ifmtu->mtu;
1335 * New MTU is bigger. Check if are allowed to alter it
1337 if ((rt->rt_flags & (RTF_FIXEDMTU | RTF_GATEWAY | RTF_HOST)) != 0) {
1340 * Skip routes with user-supplied MTU and
1341 * non-interface routes
1346 /* We are safe to update route MTU */
1347 rt->rt_mtu = ifmtu->mtu;
1353 rt_updatemtu(struct ifnet *ifp)
1355 struct if_mtuinfo ifmtu;
1356 struct rib_head *rnh;
1362 * Try to update rt_mtu for all routes using this interface
1363 * Unfortunately the only way to do this is to traverse all
1364 * routing tables in all fibs/domains.
1366 for (i = 1; i <= AF_MAX; i++) {
1367 ifmtu.mtu = if_getmtu_family(ifp, i);
1368 for (j = 0; j < rt_numfibs; j++) {
1369 rnh = rt_tables_get_rnh(j, i);
1373 rnh->rnh_walktree(&rnh->head, if_updatemtu_cb, &ifmtu);
1381 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
1382 int rt_print(char *buf, int buflen, struct rtentry *rt);
1385 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
1389 switch (s->sa_family) {
1391 paddr = &((struct sockaddr_in *)s)->sin_addr;
1394 paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
1401 if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
1404 return (strlen(buf));
1408 rt_print(char *buf, int buflen, struct rtentry *rt)
1410 struct sockaddr *addr, *mask;
1416 i = p_sockaddr(buf, buflen, addr);
1417 if (!(rt->rt_flags & RTF_HOST)) {
1419 i += p_sockaddr(buf + i, buflen - i, mask);
1422 if (rt->rt_flags & RTF_GATEWAY) {
1424 i += p_sockaddr(buf + i, buflen - i, rt->rt_gateway);
1433 * Deletes key for single-path routes, unlinks rtentry with
1434 * gateway specified in @info from multi-path routes.
1436 * Returnes unlinked entry. In case of failure, returns NULL
1437 * and sets @perror to ESRCH.
1439 static struct radix_node *
1440 rt_mpath_unlink(struct rib_head *rnh, struct rt_addrinfo *info,
1441 struct rtentry *rto, int *perror)
1444 * if we got multipath routes, we require users to specify
1445 * a matching RTAX_GATEWAY.
1447 struct rtentry *rt; // *rto = NULL;
1448 struct radix_node *rn;
1449 struct sockaddr *gw;
1451 gw = info->rti_info[RTAX_GATEWAY];
1452 rt = rt_mpath_matchgate(rto, gw);
1459 * this is the first entry in the chain
1462 rn = rn_mpath_next((struct radix_node *)rt);
1464 * there is another entry, now it's active
1469 rto->rt_flags |= RTF_UP;
1471 } else if (rt->rt_flags & RTF_GATEWAY) {
1473 * For gateway routes, we need to
1474 * make sure that we we are deleting
1475 * the correct gateway.
1476 * rt_mpath_matchgate() does not
1477 * check the case when there is only
1478 * one route in the chain.
1481 (rt->rt_gateway->sa_len != gw->sa_len ||
1482 memcmp(rt->rt_gateway, gw, gw->sa_len))) {
1489 * use the normal delete code to remove
1492 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1498 * if the entry is 2nd and on up
1500 if (rt_mpath_deldup(rto, rt) == 0)
1501 panic ("rtrequest1: rt_mpath_deldup");
1503 rn = (struct radix_node *)rt;
1509 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt,
1513 struct rtentry *rt, *rt_old;
1514 struct radix_node *rn;
1515 struct rib_head *rnh;
1517 struct sockaddr *ndst;
1518 struct sockaddr_storage mdst;
1520 KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum"));
1521 KASSERT((flags & RTF_RNH_LOCKED) == 0, ("rtrequest1_fib: locked"));
1522 switch (dst->sa_family) {
1525 /* We support multiple FIBs. */
1528 fibnum = RT_DEFAULT_FIB;
1533 * Find the correct routing tree to use for this Address Family
1535 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
1537 return (EAFNOSUPPORT);
1540 * If we are adding a host route then we don't want to put
1541 * a netmask in the tree, nor do we want to clone it.
1543 if (flags & RTF_HOST)
1549 rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask);
1550 dst = (struct sockaddr *)&mdst;
1554 rt = rt_unlinkrte(rnh, info, &error);
1559 rt_notifydelete(rt, info);
1562 * If the caller wants it, then it can have it,
1563 * but it's up to it to free the rtentry as we won't be
1574 * resolve was only used for route cloning
1579 if ((flags & RTF_GATEWAY) && !gateway)
1581 if (dst && gateway && (dst->sa_family != gateway->sa_family) &&
1582 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
1585 if (info->rti_ifa == NULL) {
1586 error = rt_getifa_fib(info, fibnum);
1590 ifa_ref(info->rti_ifa);
1591 ifa = info->rti_ifa;
1592 rt = uma_zalloc(V_rtzone, M_NOWAIT);
1597 rt->rt_flags = RTF_UP | flags;
1598 rt->rt_fibnum = fibnum;
1600 * Add the gateway. Possibly re-malloc-ing the storage for it.
1602 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
1604 uma_zfree(V_rtzone, rt);
1609 * point to the (possibly newly malloc'd) dest address.
1611 ndst = (struct sockaddr *)rt_key(rt);
1614 * make sure it contains the value we want (masked if needed).
1617 rt_maskedcopy(dst, ndst, netmask);
1619 bcopy(dst, ndst, dst->sa_len);
1622 * We use the ifa reference returned by rt_getifa_fib().
1623 * This moved from below so that rnh->rnh_addaddr() can
1624 * examine the ifa and ifa->ifa_ifp if it so desires.
1627 rt->rt_ifp = ifa->ifa_ifp;
1630 rt_setmetrics(info, rt);
1635 /* do not permit exactly the same dst/mask/gw pair */
1636 if (rt_mpath_capable(rnh) &&
1637 rt_mpath_conflict(rnh, rt, netmask)) {
1640 ifa_free(rt->rt_ifa);
1642 uma_zfree(V_rtzone, rt);
1647 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
1648 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head, rt->rt_nodes);
1651 if (rn == NULL && (info->rti_flags & RTF_PINNED) != 0) {
1654 * Force removal and re-try addition
1655 * TODO: better multipath&pinned support
1657 struct sockaddr *info_dst = info->rti_info[RTAX_DST];
1658 info->rti_info[RTAX_DST] = ndst;
1659 /* Do not delete existing PINNED(interface) routes */
1660 info->rti_flags &= ~RTF_PINNED;
1661 rt_old = rt_unlinkrte(rnh, info, &error);
1662 info->rti_flags |= RTF_PINNED;
1663 info->rti_info[RTAX_DST] = info_dst;
1665 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head,
1674 * If it still failed to go into the tree,
1675 * then un-make it (this should be a function)
1678 ifa_free(rt->rt_ifa);
1680 uma_zfree(V_rtzone, rt);
1684 if (rt_old != NULL) {
1685 rt_notifydelete(rt_old, info);
1690 * If this protocol has something to add to this then
1691 * allow it to do that as well.
1693 if (ifa->ifa_rtrequest)
1694 ifa->ifa_rtrequest(req, rt, info);
1697 * actually return a resultant rtentry and
1698 * give the caller a single reference.
1704 rnh->rnh_gen++; /* Routing table updated */
1709 error = rtrequest1_fib_change(rnh, info, ret_nrt, fibnum);
1727 rtrequest1_fib_change(struct rib_head *rnh, struct rt_addrinfo *info,
1728 struct rtentry **ret_nrt, u_int fibnum)
1730 struct rtentry *rt = NULL;
1734 struct if_mtuinfo ifmtu;
1736 rt = (struct rtentry *)rnh->rnh_lookup(info->rti_info[RTAX_DST],
1737 info->rti_info[RTAX_NETMASK], &rnh->head);
1744 * If we got multipath routes,
1745 * we require users to specify a matching RTAX_GATEWAY.
1747 if (rt_mpath_capable(rnh)) {
1748 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
1756 rt_setmetrics(info, rt);
1759 * New gateway could require new ifaddr, ifp;
1760 * flags may also be different; ifp may be specified
1761 * by ll sockaddr when protocol address is ambiguous
1763 if (((rt->rt_flags & RTF_GATEWAY) &&
1764 info->rti_info[RTAX_GATEWAY] != NULL) ||
1765 info->rti_info[RTAX_IFP] != NULL ||
1766 (info->rti_info[RTAX_IFA] != NULL &&
1767 !sa_equal(info->rti_info[RTAX_IFA], rt->rt_ifa->ifa_addr))) {
1769 * XXX: Temporarily set RTF_RNH_LOCKED flag in the rti_flags
1770 * to avoid rlock in the ifa_ifwithroute().
1772 info->rti_flags |= RTF_RNH_LOCKED;
1773 error = rt_getifa_fib(info, fibnum);
1774 info->rti_flags &= ~RTF_RNH_LOCKED;
1775 if (info->rti_ifa != NULL)
1782 /* Check if outgoing interface has changed */
1783 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa &&
1784 rt->rt_ifa != NULL) {
1785 if (rt->rt_ifa->ifa_rtrequest != NULL)
1786 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1787 ifa_free(rt->rt_ifa);
1789 /* Update gateway address */
1790 if (info->rti_info[RTAX_GATEWAY] != NULL) {
1791 error = rt_setgate(rt, rt_key(rt), info->rti_info[RTAX_GATEWAY]);
1795 rt->rt_flags &= ~RTF_GATEWAY;
1796 rt->rt_flags |= (RTF_GATEWAY & info->rti_flags);
1799 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa) {
1800 ifa_ref(info->rti_ifa);
1801 rt->rt_ifa = info->rti_ifa;
1802 rt->rt_ifp = info->rti_ifp;
1804 /* Allow some flags to be toggled on change. */
1805 rt->rt_flags &= ~RTF_FMASK;
1806 rt->rt_flags |= info->rti_flags & RTF_FMASK;
1808 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest != NULL)
1809 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
1811 /* Alter route MTU if necessary */
1812 if (rt->rt_ifp != NULL) {
1813 family = info->rti_info[RTAX_DST]->sa_family;
1814 mtu = if_getmtu_family(rt->rt_ifp, family);
1815 /* Set default MTU */
1816 if (rt->rt_mtu == 0)
1818 if (rt->rt_mtu != mtu) {
1819 /* Check if we really need to update */
1820 ifmtu.ifp = rt->rt_ifp;
1822 if_updatemtu_cb(rt->rt_nodes, &ifmtu);
1833 ifa_free(info->rti_ifa);
1838 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt)
1841 if (info->rti_mflags & RTV_MTU) {
1842 if (info->rti_rmx->rmx_mtu != 0) {
1845 * MTU was explicitly provided by user.
1848 rt->rt_flags |= RTF_FIXEDMTU;
1852 * User explicitly sets MTU to 0.
1853 * Assume rollback to default.
1855 rt->rt_flags &= ~RTF_FIXEDMTU;
1857 rt->rt_mtu = info->rti_rmx->rmx_mtu;
1859 if (info->rti_mflags & RTV_WEIGHT)
1860 rt->rt_weight = info->rti_rmx->rmx_weight;
1861 /* Kernel -> userland timebase conversion. */
1862 if (info->rti_mflags & RTV_EXPIRE)
1863 rt->rt_expire = info->rti_rmx->rmx_expire ?
1864 info->rti_rmx->rmx_expire - time_second + time_uptime : 0;
1868 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
1870 /* XXX dst may be overwritten, can we move this to below */
1871 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
1874 * Prepare to store the gateway in rt->rt_gateway.
1875 * Both dst and gateway are stored one after the other in the same
1876 * malloc'd chunk. If we have room, we can reuse the old buffer,
1877 * rt_gateway already points to the right place.
1878 * Otherwise, malloc a new block and update the 'dst' address.
1880 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
1883 R_Malloc(new, caddr_t, dlen + glen);
1887 * XXX note, we copy from *dst and not *rt_key(rt) because
1888 * rt_setgate() can be called to initialize a newly
1889 * allocated route entry, in which case rt_key(rt) == NULL
1890 * (and also rt->rt_gateway == NULL).
1891 * Free()/free() handle a NULL argument just fine.
1893 bcopy(dst, new, dlen);
1894 R_Free(rt_key(rt)); /* free old block, if any */
1895 rt_key(rt) = (struct sockaddr *)new;
1896 rt->rt_gateway = (struct sockaddr *)(new + dlen);
1900 * Copy the new gateway value into the memory chunk.
1902 bcopy(gate, rt->rt_gateway, glen);
1908 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
1910 u_char *cp1 = (u_char *)src;
1911 u_char *cp2 = (u_char *)dst;
1912 u_char *cp3 = (u_char *)netmask;
1913 u_char *cplim = cp2 + *cp3;
1914 u_char *cplim2 = cp2 + *cp1;
1916 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1921 *cp2++ = *cp1++ & *cp3++;
1923 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1927 * Set up a routing table entry, normally
1930 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
1932 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
1934 struct sockaddr *dst;
1935 struct sockaddr *netmask;
1936 struct rtentry *rt = NULL;
1937 struct rt_addrinfo info;
1939 int startfib, endfib;
1940 char tempbuf[_SOCKADDR_TMPSIZE];
1943 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1944 struct rib_head *rnh;
1946 if (flags & RTF_HOST) {
1947 dst = ifa->ifa_dstaddr;
1950 dst = ifa->ifa_addr;
1951 netmask = ifa->ifa_netmask;
1953 if (dst->sa_len == 0)
1955 switch (dst->sa_family) {
1958 /* We support multiple FIBs. */
1961 fibnum = RT_DEFAULT_FIB;
1964 if (fibnum == RT_ALL_FIBS) {
1965 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD)
1966 startfib = endfib = ifa->ifa_ifp->if_fib;
1969 endfib = rt_numfibs - 1;
1972 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
1978 * If it's a delete, check that if it exists,
1979 * it's on the correct interface or we might scrub
1980 * a route to another ifa which would
1981 * be confusing at best and possibly worse.
1983 if (cmd == RTM_DELETE) {
1985 * It's a delete, so it should already exist..
1986 * If it's a net, mask off the host bits
1987 * (Assuming we have a mask)
1988 * XXX this is kinda inet specific..
1990 if (netmask != NULL) {
1991 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
1992 dst = (struct sockaddr *)tempbuf;
1996 * Now go through all the requested tables (fibs) and do the
1997 * requested action. Realistically, this will either be fib 0
1998 * for protocols that don't do multiple tables or all the
1999 * tables for those that do.
2001 for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
2002 if (cmd == RTM_DELETE) {
2003 struct radix_node *rn;
2005 * Look up an rtentry that is in the routing tree and
2006 * contains the correct info.
2008 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
2010 /* this table doesn't exist but others might */
2013 rn = rnh->rnh_lookup(dst, netmask, &rnh->head);
2015 if (rt_mpath_capable(rnh)) {
2022 * for interface route the
2023 * rt->rt_gateway is sockaddr_intf
2024 * for cloning ARP entries, so
2025 * rt_mpath_matchgate must use the
2028 rt = rt_mpath_matchgate(rt,
2035 error = (rn == NULL ||
2036 (rn->rn_flags & RNF_ROOT) ||
2037 RNTORT(rn)->rt_ifa != ifa);
2040 /* this is only an error if bad on ALL tables */
2045 * Do the actual request
2047 bzero((caddr_t)&info, sizeof(info));
2049 info.rti_flags = flags |
2050 (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED;
2051 info.rti_info[RTAX_DST] = dst;
2053 * doing this for compatibility reasons
2056 info.rti_info[RTAX_GATEWAY] =
2057 (struct sockaddr *)&null_sdl;
2059 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
2060 info.rti_info[RTAX_NETMASK] = netmask;
2061 error = rtrequest1_fib(cmd, &info, &rt, fibnum);
2063 if (error == 0 && rt != NULL) {
2065 * notify any listening routing agents of the change
2070 * in case address alias finds the first address
2071 * e.g. ifconfig bge0 192.0.2.246/24
2072 * e.g. ifconfig bge0 192.0.2.247/24
2073 * the address set in the route is 192.0.2.246
2074 * so we need to replace it with 192.0.2.247
2076 if (memcmp(rt->rt_ifa->ifa_addr,
2077 ifa->ifa_addr, ifa->ifa_addr->sa_len)) {
2078 ifa_free(rt->rt_ifa);
2080 rt->rt_ifp = ifa->ifa_ifp;
2085 * doing this for compatibility reasons
2087 if (cmd == RTM_ADD) {
2088 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
2089 rt->rt_ifp->if_type;
2090 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
2091 rt->rt_ifp->if_index;
2095 rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum);
2098 if (cmd == RTM_DELETE) {
2100 * If we are deleting, and we found an entry,
2101 * then it's been removed from the tree..
2102 * now throw it away.
2106 if (cmd == RTM_ADD) {
2108 * We just wanted to add it..
2109 * we don't actually need a reference.
2120 if (cmd == RTM_DELETE) {
2124 /* we only give an error if it wasn't in any table */
2125 error = ((flags & RTF_HOST) ?
2126 EHOSTUNREACH : ENETUNREACH);
2130 /* return an error if any of them failed */
2138 * Set up a routing table entry, normally
2142 rtinit(struct ifaddr *ifa, int cmd, int flags)
2144 struct sockaddr *dst;
2145 int fib = RT_DEFAULT_FIB;
2147 if (flags & RTF_HOST) {
2148 dst = ifa->ifa_dstaddr;
2150 dst = ifa->ifa_addr;
2153 switch (dst->sa_family) {
2156 /* We do support multiple FIBs. */
2160 return (rtinit1(ifa, cmd, flags, fib));
2164 * Announce interface address arrival/withdraw
2165 * Returns 0 on success.
2168 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
2171 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2172 ("unexpected cmd %d", cmd));
2174 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2175 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2177 #if defined(INET) || defined(INET6)
2180 * notify the SCTP stack
2181 * this will only get called when an address is added/deleted
2182 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
2184 sctp_addr_change(ifa, cmd);
2187 return (rtsock_addrmsg(cmd, ifa, fibnum));
2191 * Announce route addition/removal.
2192 * Users of this function MUST validate input data BEFORE calling.
2193 * However we have to be able to handle invalid data:
2194 * if some userland app sends us "invalid" route message (invalid mask,
2195 * no dst, wrong address families, etc...) we need to pass it back
2196 * to app (and any other rtsock consumers) with rtm_errno field set to
2198 * Returns 0 on success.
2201 rt_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
2205 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2206 ("unexpected cmd %d", cmd));
2208 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2209 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2211 KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
2213 return (rtsock_routemsg(cmd, ifp, error, rt, fibnum));
2217 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
2220 rt_newaddrmsg_fib(cmd, ifa, error, rt, RT_ALL_FIBS);
2224 * This is called to generate messages from the routing socket
2225 * indicating a network interface has had addresses associated with it.
2228 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
2232 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2233 ("unexpected cmd %u", cmd));
2234 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2235 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2237 if (cmd == RTM_ADD) {
2238 rt_addrmsg(cmd, ifa, fibnum);
2240 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2243 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2244 rt_addrmsg(cmd, ifa, fibnum);