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>
58 #include <sys/rmlock.h>
61 #include <net/if_var.h>
62 #include <net/if_dl.h>
63 #include <net/route.h>
64 #include <net/route_var.h>
68 #include <net/radix_mpath.h>
71 #include <netinet/in.h>
72 #include <netinet/ip_mroute.h>
76 #define RT_MAXFIBS UINT16_MAX
78 /* Kernel config default option. */
81 #error "ROUTETABLES defined too low"
83 #if ROUTETABLES > RT_MAXFIBS
84 #error "ROUTETABLES defined too big"
86 #define RT_NUMFIBS ROUTETABLES
87 #endif /* ROUTETABLES */
88 /* Initialize to default if not otherwise set. */
93 #if defined(INET) || defined(INET6)
95 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
100 /* This is read-only.. */
101 u_int rt_numfibs = RT_NUMFIBS;
102 SYSCTL_UINT(_net, OID_AUTO, fibs, CTLFLAG_RDTUN, &rt_numfibs, 0, "");
105 * By default add routes to all fibs for new interfaces.
106 * Once this is set to 0 then only allocate routes on interface
107 * changes for the FIB of the caller when adding a new set of addresses
108 * to an interface. XXX this is a shotgun aproach to a problem that needs
109 * a more fine grained solution.. that will come.
110 * XXX also has the problems getting the FIB from curthread which will not
111 * always work given the fib can be overridden and prefixes can be added
112 * from the network stack context.
114 VNET_DEFINE(u_int, rt_add_addr_allfibs) = 1;
115 SYSCTL_UINT(_net, OID_AUTO, add_addr_allfibs, CTLFLAG_RWTUN | CTLFLAG_VNET,
116 &VNET_NAME(rt_add_addr_allfibs), 0, "");
118 VNET_DEFINE(struct rtstat, rtstat);
119 #define V_rtstat VNET(rtstat)
121 VNET_DEFINE(struct rib_head *, rt_tables);
122 #define V_rt_tables VNET(rt_tables)
124 VNET_DEFINE(int, rttrash); /* routes not in table but not freed */
125 #define V_rttrash VNET(rttrash)
129 * Convert a 'struct radix_node *' to a 'struct rtentry *'.
130 * The operation can be done safely (in this code) because a
131 * 'struct rtentry' starts with two 'struct radix_node''s, the first
132 * one representing leaf nodes in the routing tree, which is
133 * what the code in radix.c passes us as a 'struct radix_node'.
135 * But because there are a lot of assumptions in this conversion,
136 * do not cast explicitly, but always use the macro below.
138 #define RNTORT(p) ((struct rtentry *)(p))
140 VNET_DEFINE_STATIC(uma_zone_t, rtzone); /* Routing table UMA zone. */
141 #define V_rtzone VNET(rtzone)
143 static int rtrequest1_fib_change(struct rib_head *, struct rt_addrinfo *,
144 struct rtentry **, u_int);
145 static void rt_setmetrics(const struct rt_addrinfo *, struct rtentry *);
146 static int rt_ifdelroute(const struct rtentry *rt, void *arg);
147 static struct rtentry *rt_unlinkrte(struct rib_head *rnh,
148 struct rt_addrinfo *info, int *perror);
149 static void rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info);
151 static struct radix_node *rt_mpath_unlink(struct rib_head *rnh,
152 struct rt_addrinfo *info, struct rtentry *rto, int *perror);
154 static int rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info,
163 static int if_updatemtu_cb(struct radix_node *, void *);
166 * handler for net.my_fibnum
169 sysctl_my_fibnum(SYSCTL_HANDLER_ARGS)
174 fibnum = curthread->td_proc->p_fibnum;
175 error = sysctl_handle_int(oidp, &fibnum, 0, req);
179 SYSCTL_PROC(_net, OID_AUTO, my_fibnum, CTLTYPE_INT|CTLFLAG_RD,
180 NULL, 0, &sysctl_my_fibnum, "I", "default FIB of caller");
182 static __inline struct rib_head **
183 rt_tables_get_rnh_ptr(int table, int fam)
185 struct rib_head **rnh;
187 KASSERT(table >= 0 && table < rt_numfibs, ("%s: table out of bounds.",
189 KASSERT(fam >= 0 && fam < (AF_MAX+1), ("%s: fam out of bounds.",
192 /* rnh is [fib=0][af=0]. */
193 rnh = (struct rib_head **)V_rt_tables;
194 /* Get the offset to the requested table and fam. */
195 rnh += table * (AF_MAX+1) + fam;
201 rt_tables_get_rnh(int table, int fam)
204 return (*rt_tables_get_rnh_ptr(table, fam));
208 rt_tables_get_gen(int table, int fam)
210 struct rib_head *rnh;
212 rnh = *rt_tables_get_rnh_ptr(table, fam);
213 KASSERT(rnh != NULL, ("%s: NULL rib_head pointer table %d fam %d",
214 __func__, table, fam));
215 return (rnh->rnh_gen);
220 * route initialization must occur before ip6_init2(), which happenas at
227 /* whack the tunable ints into line. */
228 if (rt_numfibs > RT_MAXFIBS)
229 rt_numfibs = RT_MAXFIBS;
233 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, NULL);
236 rtentry_zinit(void *mem, int size, int how)
238 struct rtentry *rt = mem;
240 rt->rt_pksent = counter_u64_alloc(how);
241 if (rt->rt_pksent == NULL)
250 rtentry_zfini(void *mem, int size)
252 struct rtentry *rt = mem;
255 counter_u64_free(rt->rt_pksent);
259 rtentry_ctor(void *mem, int size, void *arg, int how)
261 struct rtentry *rt = mem;
263 bzero(rt, offsetof(struct rtentry, rt_endzero));
264 counter_u64_zero(rt->rt_pksent);
271 rtentry_dtor(void *mem, int size, void *arg)
273 struct rtentry *rt = mem;
279 vnet_route_init(const void *unused __unused)
282 struct rib_head **rnh;
286 V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) *
287 sizeof(struct rib_head *), M_RTABLE, M_WAITOK|M_ZERO);
289 V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry),
290 rtentry_ctor, rtentry_dtor,
291 rtentry_zinit, rtentry_zfini, UMA_ALIGN_PTR, 0);
292 for (dom = domains; dom; dom = dom->dom_next) {
293 if (dom->dom_rtattach == NULL)
296 for (table = 0; table < rt_numfibs; table++) {
297 fam = dom->dom_family;
298 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
301 rnh = rt_tables_get_rnh_ptr(table, fam);
303 panic("%s: rnh NULL", __func__);
304 dom->dom_rtattach((void **)rnh, 0);
308 VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
313 vnet_route_uninit(const void *unused __unused)
318 struct rib_head **rnh;
320 for (dom = domains; dom; dom = dom->dom_next) {
321 if (dom->dom_rtdetach == NULL)
324 for (table = 0; table < rt_numfibs; table++) {
325 fam = dom->dom_family;
327 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
330 rnh = rt_tables_get_rnh_ptr(table, fam);
332 panic("%s: rnh NULL", __func__);
333 dom->dom_rtdetach((void **)rnh, 0);
337 free(V_rt_tables, M_RTABLE);
338 uma_zdestroy(V_rtzone);
340 VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST,
341 vnet_route_uninit, 0);
345 rt_table_init(int offset)
349 rh = malloc(sizeof(struct rib_head), M_RTABLE, M_WAITOK | M_ZERO);
351 /* TODO: These details should be hidded inside radix.c */
352 /* Init masks tree */
353 rn_inithead_internal(&rh->head, rh->rnh_nodes, offset);
354 rn_inithead_internal(&rh->rmhead.head, rh->rmhead.mask_nodes, 0);
355 rh->head.rnh_masks = &rh->rmhead;
360 /* Finally, set base callbacks */
361 rh->rnh_addaddr = rn_addroute;
362 rh->rnh_deladdr = rn_delete;
363 rh->rnh_matchaddr = rn_match;
364 rh->rnh_lookup = rn_lookup;
365 rh->rnh_walktree = rn_walktree;
366 rh->rnh_walktree_from = rn_walktree_from;
372 rt_freeentry(struct radix_node *rn, void *arg)
374 struct radix_head * const rnh = arg;
375 struct radix_node *x;
377 x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh);
384 rt_table_destroy(struct rib_head *rh)
387 rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head);
389 /* Assume table is already empty */
390 RIB_LOCK_DESTROY(rh);
395 #ifndef _SYS_SYSPROTO_H_
401 sys_setfib(struct thread *td, struct setfib_args *uap)
403 if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs)
405 td->td_proc->p_fibnum = uap->fibnum;
410 * Packet routing routines.
413 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum)
417 if ((rt = ro->ro_rt) != NULL) {
418 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
423 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum);
425 RT_UNLOCK(ro->ro_rt);
429 * Look up the route that matches the address given
430 * Or, at least try.. Create a cloned route if needed.
432 * The returned route, if any, is locked.
435 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
438 return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB));
442 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags,
447 struct radix_node *rn;
448 struct rtentry *newrt;
449 struct rt_addrinfo info;
450 int err = 0, msgtype = RTM_MISS;
452 KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum"));
453 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
459 * Look up the address in the table for that Address Family
461 if ((ignflags & RTF_RNH_LOCKED) == 0)
467 rn = rh->rnh_matchaddr(dst, &rh->head);
468 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
472 if ((ignflags & RTF_RNH_LOCKED) == 0)
476 } else if ((ignflags & RTF_RNH_LOCKED) == 0)
479 * Either we hit the root or could not find any match,
480 * which basically means: "cannot get there from here".
483 V_rtstat.rts_unreach++;
487 * If required, report the failure to the supervising
489 * For a delete, this is not an error. (report == 0)
491 bzero(&info, sizeof(info));
492 info.rti_info[RTAX_DST] = dst;
493 rt_missmsg_fib(msgtype, &info, 0, err, fibnum);
499 * Remove a reference count from an rtentry.
500 * If the count gets low enough, take it out of the routing table
503 rtfree(struct rtentry *rt)
505 struct rib_head *rnh;
507 KASSERT(rt != NULL,("%s: NULL rt", __func__));
508 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
509 KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
514 * The callers should use RTFREE_LOCKED() or RTFREE(), so
515 * we should come here exactly with the last reference.
518 if (rt->rt_refcnt > 0) {
519 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt);
524 * On last reference give the "close method" a chance
525 * to cleanup private state. This also permits (for
526 * IPv4 and IPv6) a chance to decide if the routing table
527 * entry should be purged immediately or at a later time.
528 * When an immediate purge is to happen the close routine
529 * typically calls rtexpunge which clears the RTF_UP flag
530 * on the entry so that the code below reclaims the storage.
532 if (rt->rt_refcnt == 0 && rnh->rnh_close)
533 rnh->rnh_close((struct radix_node *)rt, &rnh->head);
536 * If we are no longer "up" (and ref == 0)
537 * then we can free the resources associated
540 if ((rt->rt_flags & RTF_UP) == 0) {
541 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
544 * the rtentry must have been removed from the routing table
545 * so it is represented in rttrash.. remove that now.
549 if (rt->rt_refcnt < 0) {
550 printf("rtfree: %p not freed (neg refs)\n", rt);
555 * release references on items we hold them on..
556 * e.g other routes and ifaddrs.
559 ifa_free(rt->rt_ifa);
561 * The key is separatly alloc'd so free it (see rt_setgate()).
562 * This also frees the gateway, as they are always malloc'd
568 * and the rtentry itself of course
570 uma_zfree(V_rtzone, rt);
579 * Force a routing table entry to the specified
580 * destination to go through the given gateway.
581 * Normally called as a result of a routing redirect
582 * message from the network layer.
585 rtredirect_fib(struct sockaddr *dst,
586 struct sockaddr *gateway,
587 struct sockaddr *netmask,
589 struct sockaddr *src,
595 struct rt_addrinfo info;
596 struct epoch_tracker et;
598 struct rib_head *rnh;
602 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
604 error = EAFNOSUPPORT;
607 /* verify the gateway is directly reachable */
608 if ((ifa = ifa_ifwithnet(gateway, 0, fibnum)) == NULL) {
612 rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */
614 * If the redirect isn't from our current router for this dst,
615 * it's either old or wrong. If it redirects us to ourselves,
616 * we have a routing loop, perhaps as a result of an interface
617 * going down recently.
619 if (!(flags & RTF_DONE) && rt) {
620 if (!sa_equal(src, rt->rt_gateway)) {
624 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) {
629 if ((flags & RTF_GATEWAY) && ifa_ifwithaddr_check(gateway)) {
630 error = EHOSTUNREACH;
634 * Create a new entry if we just got back a wildcard entry
635 * or the lookup failed. This is necessary for hosts
636 * which use routing redirects generated by smart gateways
637 * to dynamically build the routing tables.
639 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
642 * Don't listen to the redirect if it's
643 * for a route to an interface.
645 if (rt->rt_flags & RTF_GATEWAY) {
646 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
648 * Changing from route to net => route to host.
649 * Create new route, rather than smashing route to net.
655 flags |= RTF_DYNAMIC;
656 bzero((caddr_t)&info, sizeof(info));
657 info.rti_info[RTAX_DST] = dst;
658 info.rti_info[RTAX_GATEWAY] = gateway;
659 info.rti_info[RTAX_NETMASK] = netmask;
662 info.rti_flags = flags;
663 error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum);
666 flags = rt->rt_flags;
669 stat = &V_rtstat.rts_dynamic;
673 * Smash the current notion of the gateway to
674 * this destination. Should check about netmask!!!
676 if ((flags & RTF_GATEWAY) == 0)
677 rt->rt_flags &= ~RTF_GATEWAY;
678 rt->rt_flags |= RTF_MODIFIED;
679 flags |= RTF_MODIFIED;
680 stat = &V_rtstat.rts_newgateway;
682 * add the key and gateway (in one malloc'd chunk).
687 rt_setgate(rt, rt_key(rt), gateway);
691 error = EHOSTUNREACH;
698 V_rtstat.rts_badredirect++;
699 else if (stat != NULL)
701 bzero((caddr_t)&info, sizeof(info));
702 info.rti_info[RTAX_DST] = dst;
703 info.rti_info[RTAX_GATEWAY] = gateway;
704 info.rti_info[RTAX_NETMASK] = netmask;
705 info.rti_info[RTAX_AUTHOR] = src;
706 rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum);
710 * Routing table ioctl interface.
713 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
717 * If more ioctl commands are added here, make sure the proper
718 * super-user checks are being performed because it is possible for
719 * prison-root to make it this far if raw sockets have been enabled
723 /* Multicast goop, grrr... */
724 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
731 ifa_ifwithroute(int flags, const struct sockaddr *dst, struct sockaddr *gateway,
737 MPASS(in_epoch(net_epoch_preempt));
738 if ((flags & RTF_GATEWAY) == 0) {
740 * If we are adding a route to an interface,
741 * and the interface is a pt to pt link
742 * we should search for the destination
743 * as our clue to the interface. Otherwise
744 * we can use the local address.
747 if (flags & RTF_HOST)
748 ifa = ifa_ifwithdstaddr(dst, fibnum);
750 ifa = ifa_ifwithaddr(gateway);
753 * If we are adding a route to a remote net
754 * or host, the gateway may still be on the
755 * other end of a pt to pt link.
757 ifa = ifa_ifwithdstaddr(gateway, fibnum);
760 ifa = ifa_ifwithnet(gateway, 0, fibnum);
764 rt = rtalloc1_fib(gateway, 0, flags, fibnum);
768 * dismiss a gateway that is reachable only
769 * through the default router
771 switch (gateway->sa_family) {
773 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
777 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
783 if (!not_found && rt->rt_ifa != NULL) {
788 if (not_found || ifa == NULL)
791 if (ifa->ifa_addr->sa_family != dst->sa_family) {
792 struct ifaddr *oifa = ifa;
793 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
802 * Do appropriate manipulations of a routing tree given
803 * all the bits of info needed
806 rtrequest_fib(int req,
807 struct sockaddr *dst,
808 struct sockaddr *gateway,
809 struct sockaddr *netmask,
811 struct rtentry **ret_nrt,
814 struct rt_addrinfo info;
816 if (dst->sa_len == 0)
819 bzero((caddr_t)&info, sizeof(info));
820 info.rti_flags = flags;
821 info.rti_info[RTAX_DST] = dst;
822 info.rti_info[RTAX_GATEWAY] = gateway;
823 info.rti_info[RTAX_NETMASK] = netmask;
824 return rtrequest1_fib(req, &info, ret_nrt, fibnum);
829 * Copy most of @rt data into @info.
831 * If @flags contains NHR_COPY, copies dst,netmask and gw to the
832 * pointers specified by @info structure. Assume such pointers
833 * are zeroed sockaddr-like structures with sa_len field initialized
834 * to reflect size of the provided buffer. if no NHR_COPY is specified,
835 * point dst,netmask and gw @info fields to appropriate @rt values.
837 * if @flags contains NHR_REF, do refcouting on rt_ifp.
839 * Returns 0 on success.
842 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags)
844 struct rt_metrics *rmx;
845 struct sockaddr *src, *dst;
848 if (flags & NHR_COPY) {
849 /* Copy destination if dst is non-zero */
851 dst = info->rti_info[RTAX_DST];
852 sa_len = src->sa_len;
854 if (src->sa_len > dst->sa_len)
856 memcpy(dst, src, src->sa_len);
857 info->rti_addrs |= RTA_DST;
860 /* Copy mask if set && dst is non-zero */
862 dst = info->rti_info[RTAX_NETMASK];
863 if (src != NULL && dst != NULL) {
866 * Radix stores different value in sa_len,
867 * assume rt_mask() to have the same length
870 if (sa_len > dst->sa_len)
872 memcpy(dst, src, src->sa_len);
873 info->rti_addrs |= RTA_NETMASK;
876 /* Copy gateway is set && dst is non-zero */
877 src = rt->rt_gateway;
878 dst = info->rti_info[RTAX_GATEWAY];
879 if ((rt->rt_flags & RTF_GATEWAY) && src != NULL && dst != NULL){
880 if (src->sa_len > dst->sa_len)
882 memcpy(dst, src, src->sa_len);
883 info->rti_addrs |= RTA_GATEWAY;
886 info->rti_info[RTAX_DST] = rt_key(rt);
887 info->rti_addrs |= RTA_DST;
888 if (rt_mask(rt) != NULL) {
889 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
890 info->rti_addrs |= RTA_NETMASK;
892 if (rt->rt_flags & RTF_GATEWAY) {
893 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
894 info->rti_addrs |= RTA_GATEWAY;
900 info->rti_mflags |= RTV_MTU;
901 rmx->rmx_mtu = rt->rt_mtu;
904 info->rti_flags = rt->rt_flags;
905 info->rti_ifp = rt->rt_ifp;
906 info->rti_ifa = rt->rt_ifa;
907 ifa_ref(info->rti_ifa);
908 if (flags & NHR_REF) {
909 /* Do 'traditional' refcouting */
910 if_ref(info->rti_ifp);
917 * Lookups up route entry for @dst in RIB database for fib @fibnum.
918 * Exports entry data to @info using rt_exportinfo().
920 * if @flags contains NHR_REF, refcouting is performed on rt_ifp.
921 * All references can be released later by calling rib_free_info()
923 * Returns 0 on success.
924 * Returns ENOENT for lookup failure, ENOMEM for export failure.
927 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
928 uint32_t flowid, struct rt_addrinfo *info)
932 struct radix_node *rn;
936 KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
937 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
942 rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head);
943 if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
945 /* Ensure route & ifp is UP */
946 if (RT_LINK_IS_UP(rt->rt_ifp)) {
947 flags = (flags & NHR_REF) | NHR_COPY;
948 error = rt_exportinfo(rt, info, flags);
960 * Releases all references acquired by rib_lookup_info() when
961 * called with NHR_REF flags.
964 rib_free_info(struct rt_addrinfo *info)
967 if_rele(info->rti_ifp);
971 * Iterates over all existing fibs in system calling
972 * @setwa_f function prior to traversing each fib.
973 * Calls @wa_f function for each element in current fib.
974 * If af is not AF_UNSPEC, iterates over fibs in particular
978 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f,
981 struct rib_head *rnh;
985 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
986 /* Do we want some specific family? */
987 if (af != AF_UNSPEC) {
988 rnh = rt_tables_get_rnh(fibnum, af);
992 setwa_f(rnh, fibnum, af, arg);
995 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
1000 for (i = 1; i <= AF_MAX; i++) {
1001 rnh = rt_tables_get_rnh(fibnum, i);
1004 if (setwa_f != NULL)
1005 setwa_f(rnh, fibnum, i, arg);
1008 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
1016 struct rt_addrinfo info;
1017 struct rib_head *rnh;
1018 struct rtentry *head;
1022 * Conditionally unlinks @rn from radix tree based
1023 * on info data passed in @arg.
1026 rt_checkdelroute(struct radix_node *rn, void *arg)
1028 struct rt_delinfo *di;
1029 struct rt_addrinfo *info;
1033 di = (struct rt_delinfo *)arg;
1034 rt = (struct rtentry *)rn;
1038 info->rti_info[RTAX_DST] = rt_key(rt);
1039 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
1040 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1042 rt = rt_unlinkrte(di->rnh, info, &error);
1044 /* Either not allowed or not matched. Skip entry */
1048 /* Entry was unlinked. Add to the list and return */
1049 rt->rt_chain = di->head;
1056 * Iterates over all existing fibs in system.
1057 * Deletes each element for which @filter_f function returned
1059 * If @af is not AF_UNSPEC, iterates over fibs in particular
1063 rt_foreach_fib_walk_del(int af, rt_filter_f_t *filter_f, void *arg)
1065 struct rib_head *rnh;
1066 struct rt_delinfo di;
1071 bzero(&di, sizeof(di));
1072 di.info.rti_filter = filter_f;
1073 di.info.rti_filterdata = arg;
1075 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1076 /* Do we want some specific family? */
1077 if (af != AF_UNSPEC) {
1085 for (i = start; i <= end; i++) {
1086 rnh = rt_tables_get_rnh(fibnum, i);
1092 rnh->rnh_walktree(&rnh->head, rt_checkdelroute, &di);
1095 if (di.head == NULL)
1098 /* We might have something to reclaim */
1099 while (di.head != NULL) {
1101 di.head = rt->rt_chain;
1102 rt->rt_chain = NULL;
1104 /* TODO std rt -> rt_addrinfo export */
1105 di.info.rti_info[RTAX_DST] = rt_key(rt);
1106 di.info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1108 rt_notifydelete(rt, &di.info);
1117 * Delete Routes for a Network Interface
1119 * Called for each routing entry via the rnh->rnh_walktree() call above
1120 * to delete all route entries referencing a detaching network interface.
1123 * rt pointer to rtentry
1124 * arg argument passed to rnh->rnh_walktree() - detaching interface
1128 * errno failed - reason indicated
1131 rt_ifdelroute(const struct rtentry *rt, void *arg)
1133 struct ifnet *ifp = arg;
1135 if (rt->rt_ifp != ifp)
1139 * Protect (sorta) against walktree recursion problems
1140 * with cloned routes
1142 if ((rt->rt_flags & RTF_UP) == 0)
1149 * Delete all remaining routes using this interface
1150 * Unfortuneatly the only way to do this is to slog through
1151 * the entire routing table looking for routes which point
1152 * to this interface...oh well...
1155 rt_flushifroutes_af(struct ifnet *ifp, int af)
1157 KASSERT((af >= 1 && af <= AF_MAX), ("%s: af %d not >= 1 and <= %d",
1158 __func__, af, AF_MAX));
1160 rt_foreach_fib_walk_del(af, rt_ifdelroute, ifp);
1164 rt_flushifroutes(struct ifnet *ifp)
1167 rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
1171 * Conditionally unlinks rtentry matching data inside @info from @rnh.
1172 * Returns unlinked, locked and referenced @rtentry on success,
1173 * Returns NULL and sets @perror to:
1174 * ESRCH - if prefix was not found,
1175 * EADDRINUSE - if trying to delete PINNED route without appropriate flag.
1176 * ENOENT - if supplied filter function returned 0 (not matched).
1178 static struct rtentry *
1179 rt_unlinkrte(struct rib_head *rnh, struct rt_addrinfo *info, int *perror)
1181 struct sockaddr *dst, *netmask;
1183 struct radix_node *rn;
1185 dst = info->rti_info[RTAX_DST];
1186 netmask = info->rti_info[RTAX_NETMASK];
1188 rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, &rnh->head);
1194 if ((info->rti_flags & RTF_PINNED) == 0) {
1195 /* Check if target route can be deleted */
1196 if (rt->rt_flags & RTF_PINNED) {
1197 *perror = EADDRINUSE;
1202 if (info->rti_filter != NULL) {
1203 if (info->rti_filter(rt, info->rti_filterdata) == 0) {
1210 * Filter function requested rte deletion.
1211 * Ease the caller work by filling in remaining info
1212 * from that particular entry.
1214 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1218 * Remove the item from the tree and return it.
1219 * Complain if it is not there and do no more processing.
1223 if (rt_mpath_capable(rnh))
1224 rn = rt_mpath_unlink(rnh, info, rt, perror);
1227 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1231 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
1232 panic ("rtrequest delete");
1237 rt->rt_flags &= ~RTF_UP;
1245 rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info)
1250 * give the protocol a chance to keep things in sync.
1253 if (ifa != NULL && ifa->ifa_rtrequest != NULL)
1254 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1257 * One more rtentry floating around that is not
1258 * linked to the routing table. rttrash will be decremented
1259 * when RTFREE(rt) is eventually called.
1266 * These (questionable) definitions of apparent local variables apply
1267 * to the next two functions. XXXXXX!!!
1269 #define dst info->rti_info[RTAX_DST]
1270 #define gateway info->rti_info[RTAX_GATEWAY]
1271 #define netmask info->rti_info[RTAX_NETMASK]
1272 #define ifaaddr info->rti_info[RTAX_IFA]
1273 #define ifpaddr info->rti_info[RTAX_IFP]
1274 #define flags info->rti_flags
1277 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined,
1278 * it will be referenced so the caller must free it.
1280 * Assume basic consistency checks are executed by callers:
1281 * RTAX_DST exists, if RTF_GATEWAY is set, RTAX_GATEWAY exists as well.
1284 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
1286 struct epoch_tracker et;
1290 * ifp may be specified by sockaddr_dl
1291 * when protocol address is ambiguous.
1294 needref = (info->rti_ifa == NULL);
1295 NET_EPOCH_ENTER(et);
1297 /* If we have interface specified by the ifindex in the address, use it */
1298 if (info->rti_ifp == NULL && ifpaddr != NULL &&
1299 ifpaddr->sa_family == AF_LINK) {
1300 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)ifpaddr;
1301 if (sdl->sdl_index != 0)
1302 info->rti_ifp = ifnet_byindex_locked(sdl->sdl_index);
1305 * If we have source address specified, try to find it
1306 * TODO: avoid enumerating all ifas on all interfaces.
1308 if (info->rti_ifa == NULL && ifaaddr != NULL)
1309 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
1310 if (info->rti_ifa == NULL) {
1311 struct sockaddr *sa;
1314 * Most common use case for the userland-supplied routes.
1316 * Choose sockaddr to select ifa.
1317 * -- if ifp is set --
1318 * Order of preference:
1320 * 2) gateway address
1321 * Note: for interface routes link-level gateway address
1322 * is specified to indicate the interface index without
1323 * specifying RTF_GATEWAY. In this case, ignore gateway
1324 * Note: gateway AF may be different from dst AF. In this case,
1326 * 3) final destination.
1327 * 4) if all of these fails, try to get at least link-level ifa.
1329 * try to lookup gateway or dst in the routing table to get ifa
1331 if (info->rti_info[RTAX_IFA] != NULL)
1332 sa = info->rti_info[RTAX_IFA];
1333 else if ((info->rti_flags & RTF_GATEWAY) != 0 &&
1334 gateway->sa_family == dst->sa_family)
1338 if (info->rti_ifp != NULL) {
1339 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
1341 if (info->rti_ifa == NULL && gateway != NULL)
1342 info->rti_ifa = ifaof_ifpforaddr(gateway, info->rti_ifp);
1343 } else if (dst != NULL && gateway != NULL)
1344 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway,
1346 else if (sa != NULL)
1347 info->rti_ifa = ifa_ifwithroute(flags, sa, sa,
1350 if (needref && info->rti_ifa != NULL) {
1351 if (info->rti_ifp == NULL)
1352 info->rti_ifp = info->rti_ifa->ifa_ifp;
1353 ifa_ref(info->rti_ifa);
1355 error = ENETUNREACH;
1361 if_updatemtu_cb(struct radix_node *rn, void *arg)
1364 struct if_mtuinfo *ifmtu;
1366 rt = (struct rtentry *)rn;
1367 ifmtu = (struct if_mtuinfo *)arg;
1369 if (rt->rt_ifp != ifmtu->ifp)
1372 if (rt->rt_mtu >= ifmtu->mtu) {
1373 /* We have to decrease mtu regardless of flags */
1374 rt->rt_mtu = ifmtu->mtu;
1379 * New MTU is bigger. Check if are allowed to alter it
1381 if ((rt->rt_flags & (RTF_FIXEDMTU | RTF_GATEWAY | RTF_HOST)) != 0) {
1384 * Skip routes with user-supplied MTU and
1385 * non-interface routes
1390 /* We are safe to update route MTU */
1391 rt->rt_mtu = ifmtu->mtu;
1397 rt_updatemtu(struct ifnet *ifp)
1399 struct if_mtuinfo ifmtu;
1400 struct rib_head *rnh;
1406 * Try to update rt_mtu for all routes using this interface
1407 * Unfortunately the only way to do this is to traverse all
1408 * routing tables in all fibs/domains.
1410 for (i = 1; i <= AF_MAX; i++) {
1411 ifmtu.mtu = if_getmtu_family(ifp, i);
1412 for (j = 0; j < rt_numfibs; j++) {
1413 rnh = rt_tables_get_rnh(j, i);
1417 rnh->rnh_walktree(&rnh->head, if_updatemtu_cb, &ifmtu);
1425 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
1426 int rt_print(char *buf, int buflen, struct rtentry *rt);
1429 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
1433 switch (s->sa_family) {
1435 paddr = &((struct sockaddr_in *)s)->sin_addr;
1438 paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
1445 if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
1448 return (strlen(buf));
1452 rt_print(char *buf, int buflen, struct rtentry *rt)
1454 struct sockaddr *addr, *mask;
1460 i = p_sockaddr(buf, buflen, addr);
1461 if (!(rt->rt_flags & RTF_HOST)) {
1463 i += p_sockaddr(buf + i, buflen - i, mask);
1466 if (rt->rt_flags & RTF_GATEWAY) {
1468 i += p_sockaddr(buf + i, buflen - i, rt->rt_gateway);
1477 * Deletes key for single-path routes, unlinks rtentry with
1478 * gateway specified in @info from multi-path routes.
1480 * Returnes unlinked entry. In case of failure, returns NULL
1481 * and sets @perror to ESRCH.
1483 static struct radix_node *
1484 rt_mpath_unlink(struct rib_head *rnh, struct rt_addrinfo *info,
1485 struct rtentry *rto, int *perror)
1488 * if we got multipath routes, we require users to specify
1489 * a matching RTAX_GATEWAY.
1491 struct rtentry *rt; // *rto = NULL;
1492 struct radix_node *rn;
1493 struct sockaddr *gw;
1495 gw = info->rti_info[RTAX_GATEWAY];
1496 rt = rt_mpath_matchgate(rto, gw);
1503 * this is the first entry in the chain
1506 rn = rn_mpath_next((struct radix_node *)rt);
1508 * there is another entry, now it's active
1513 rto->rt_flags |= RTF_UP;
1515 } else if (rt->rt_flags & RTF_GATEWAY) {
1517 * For gateway routes, we need to
1518 * make sure that we we are deleting
1519 * the correct gateway.
1520 * rt_mpath_matchgate() does not
1521 * check the case when there is only
1522 * one route in the chain.
1525 (rt->rt_gateway->sa_len != gw->sa_len ||
1526 memcmp(rt->rt_gateway, gw, gw->sa_len))) {
1533 * use the normal delete code to remove
1536 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1542 * if the entry is 2nd and on up
1544 if (rt_mpath_deldup(rto, rt) == 0)
1545 panic ("rtrequest1: rt_mpath_deldup");
1547 rn = (struct radix_node *)rt;
1553 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt,
1557 struct rtentry *rt, *rt_old;
1558 struct radix_node *rn;
1559 struct rib_head *rnh;
1561 struct sockaddr *ndst;
1562 struct sockaddr_storage mdst;
1564 KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum"));
1565 KASSERT((flags & RTF_RNH_LOCKED) == 0, ("rtrequest1_fib: locked"));
1566 switch (dst->sa_family) {
1569 /* We support multiple FIBs. */
1572 fibnum = RT_DEFAULT_FIB;
1577 * Find the correct routing tree to use for this Address Family
1579 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
1581 return (EAFNOSUPPORT);
1584 * If we are adding a host route then we don't want to put
1585 * a netmask in the tree, nor do we want to clone it.
1587 if (flags & RTF_HOST)
1593 rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask);
1594 dst = (struct sockaddr *)&mdst;
1598 rt = rt_unlinkrte(rnh, info, &error);
1603 rt_notifydelete(rt, info);
1606 * If the caller wants it, then it can have it,
1607 * but it's up to it to free the rtentry as we won't be
1618 * resolve was only used for route cloning
1623 if ((flags & RTF_GATEWAY) && !gateway)
1625 if (dst && gateway && (dst->sa_family != gateway->sa_family) &&
1626 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
1629 if (info->rti_ifa == NULL) {
1630 error = rt_getifa_fib(info, fibnum);
1634 rt = uma_zalloc(V_rtzone, M_NOWAIT);
1638 rt->rt_flags = RTF_UP | flags;
1639 rt->rt_fibnum = fibnum;
1641 * Add the gateway. Possibly re-malloc-ing the storage for it.
1643 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
1644 uma_zfree(V_rtzone, rt);
1649 * point to the (possibly newly malloc'd) dest address.
1651 ndst = (struct sockaddr *)rt_key(rt);
1654 * make sure it contains the value we want (masked if needed).
1657 rt_maskedcopy(dst, ndst, netmask);
1659 bcopy(dst, ndst, dst->sa_len);
1662 * We use the ifa reference returned by rt_getifa_fib().
1663 * This moved from below so that rnh->rnh_addaddr() can
1664 * examine the ifa and ifa->ifa_ifp if it so desires.
1666 ifa = info->rti_ifa;
1669 rt->rt_ifp = ifa->ifa_ifp;
1672 rt_setmetrics(info, rt);
1677 /* do not permit exactly the same dst/mask/gw pair */
1678 if (rt_mpath_capable(rnh) &&
1679 rt_mpath_conflict(rnh, rt, netmask)) {
1682 ifa_free(rt->rt_ifa);
1684 uma_zfree(V_rtzone, rt);
1689 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
1690 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head, rt->rt_nodes);
1693 if (rn == NULL && (info->rti_flags & RTF_PINNED) != 0) {
1696 * Force removal and re-try addition
1697 * TODO: better multipath&pinned support
1699 struct sockaddr *info_dst = info->rti_info[RTAX_DST];
1700 info->rti_info[RTAX_DST] = ndst;
1701 /* Do not delete existing PINNED(interface) routes */
1702 info->rti_flags &= ~RTF_PINNED;
1703 rt_old = rt_unlinkrte(rnh, info, &error);
1704 info->rti_flags |= RTF_PINNED;
1705 info->rti_info[RTAX_DST] = info_dst;
1707 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head,
1716 * If it still failed to go into the tree,
1717 * then un-make it (this should be a function)
1720 ifa_free(rt->rt_ifa);
1722 uma_zfree(V_rtzone, rt);
1726 if (rt_old != NULL) {
1727 rt_notifydelete(rt_old, info);
1732 * If this protocol has something to add to this then
1733 * allow it to do that as well.
1735 if (ifa->ifa_rtrequest)
1736 ifa->ifa_rtrequest(req, rt, info);
1739 * actually return a resultant rtentry and
1740 * give the caller a single reference.
1746 rnh->rnh_gen++; /* Routing table updated */
1751 error = rtrequest1_fib_change(rnh, info, ret_nrt, fibnum);
1769 rtrequest1_fib_change(struct rib_head *rnh, struct rt_addrinfo *info,
1770 struct rtentry **ret_nrt, u_int fibnum)
1772 struct rtentry *rt = NULL;
1776 struct if_mtuinfo ifmtu;
1778 RIB_WLOCK_ASSERT(rnh);
1780 rt = (struct rtentry *)rnh->rnh_lookup(info->rti_info[RTAX_DST],
1781 info->rti_info[RTAX_NETMASK], &rnh->head);
1788 * If we got multipath routes,
1789 * we require users to specify a matching RTAX_GATEWAY.
1791 if (rt_mpath_capable(rnh)) {
1792 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
1800 rt_setmetrics(info, rt);
1803 * New gateway could require new ifaddr, ifp;
1804 * flags may also be different; ifp may be specified
1805 * by ll sockaddr when protocol address is ambiguous
1807 if (((rt->rt_flags & RTF_GATEWAY) &&
1808 info->rti_info[RTAX_GATEWAY] != NULL) ||
1809 info->rti_info[RTAX_IFP] != NULL ||
1810 (info->rti_info[RTAX_IFA] != NULL &&
1811 !sa_equal(info->rti_info[RTAX_IFA], rt->rt_ifa->ifa_addr))) {
1813 * XXX: Temporarily set RTF_RNH_LOCKED flag in the rti_flags
1814 * to avoid rlock in the ifa_ifwithroute().
1816 info->rti_flags |= RTF_RNH_LOCKED;
1817 error = rt_getifa_fib(info, fibnum);
1818 info->rti_flags &= ~RTF_RNH_LOCKED;
1819 if (info->rti_ifa != NULL)
1826 /* Check if outgoing interface has changed */
1827 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa &&
1828 rt->rt_ifa != NULL) {
1829 if (rt->rt_ifa->ifa_rtrequest != NULL)
1830 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1831 ifa_free(rt->rt_ifa);
1834 /* Update gateway address */
1835 if (info->rti_info[RTAX_GATEWAY] != NULL) {
1836 error = rt_setgate(rt, rt_key(rt), info->rti_info[RTAX_GATEWAY]);
1840 rt->rt_flags &= ~RTF_GATEWAY;
1841 rt->rt_flags |= (RTF_GATEWAY & info->rti_flags);
1844 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa) {
1845 ifa_ref(info->rti_ifa);
1846 rt->rt_ifa = info->rti_ifa;
1847 rt->rt_ifp = info->rti_ifp;
1849 /* Allow some flags to be toggled on change. */
1850 rt->rt_flags &= ~RTF_FMASK;
1851 rt->rt_flags |= info->rti_flags & RTF_FMASK;
1853 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest != NULL)
1854 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
1856 /* Alter route MTU if necessary */
1857 if (rt->rt_ifp != NULL) {
1858 family = info->rti_info[RTAX_DST]->sa_family;
1859 mtu = if_getmtu_family(rt->rt_ifp, family);
1860 /* Set default MTU */
1861 if (rt->rt_mtu == 0)
1863 if (rt->rt_mtu != mtu) {
1864 /* Check if we really need to update */
1865 ifmtu.ifp = rt->rt_ifp;
1867 if_updatemtu_cb(rt->rt_nodes, &ifmtu);
1872 * This route change may have modified the route's gateway. In that
1873 * case, any inpcbs that have cached this route need to invalidate their
1884 if (free_ifa != 0) {
1885 ifa_free(info->rti_ifa);
1886 info->rti_ifa = NULL;
1892 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt)
1895 if (info->rti_mflags & RTV_MTU) {
1896 if (info->rti_rmx->rmx_mtu != 0) {
1899 * MTU was explicitly provided by user.
1902 rt->rt_flags |= RTF_FIXEDMTU;
1906 * User explicitly sets MTU to 0.
1907 * Assume rollback to default.
1909 rt->rt_flags &= ~RTF_FIXEDMTU;
1911 rt->rt_mtu = info->rti_rmx->rmx_mtu;
1913 if (info->rti_mflags & RTV_WEIGHT)
1914 rt->rt_weight = info->rti_rmx->rmx_weight;
1915 /* Kernel -> userland timebase conversion. */
1916 if (info->rti_mflags & RTV_EXPIRE)
1917 rt->rt_expire = info->rti_rmx->rmx_expire ?
1918 info->rti_rmx->rmx_expire - time_second + time_uptime : 0;
1922 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
1924 /* XXX dst may be overwritten, can we move this to below */
1925 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
1928 * Prepare to store the gateway in rt->rt_gateway.
1929 * Both dst and gateway are stored one after the other in the same
1930 * malloc'd chunk. If we have room, we can reuse the old buffer,
1931 * rt_gateway already points to the right place.
1932 * Otherwise, malloc a new block and update the 'dst' address.
1934 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
1937 R_Malloc(new, caddr_t, dlen + glen);
1941 * XXX note, we copy from *dst and not *rt_key(rt) because
1942 * rt_setgate() can be called to initialize a newly
1943 * allocated route entry, in which case rt_key(rt) == NULL
1944 * (and also rt->rt_gateway == NULL).
1945 * Free()/free() handle a NULL argument just fine.
1947 bcopy(dst, new, dlen);
1948 R_Free(rt_key(rt)); /* free old block, if any */
1949 rt_key(rt) = (struct sockaddr *)new;
1950 rt->rt_gateway = (struct sockaddr *)(new + dlen);
1954 * Copy the new gateway value into the memory chunk.
1956 bcopy(gate, rt->rt_gateway, glen);
1962 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
1964 u_char *cp1 = (u_char *)src;
1965 u_char *cp2 = (u_char *)dst;
1966 u_char *cp3 = (u_char *)netmask;
1967 u_char *cplim = cp2 + *cp3;
1968 u_char *cplim2 = cp2 + *cp1;
1970 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1975 *cp2++ = *cp1++ & *cp3++;
1977 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1981 * Set up a routing table entry, normally
1984 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
1986 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
1989 struct sockaddr *dst;
1990 struct sockaddr *netmask;
1991 struct rtentry *rt = NULL;
1992 struct rt_addrinfo info;
1994 int startfib, endfib;
1995 char tempbuf[_SOCKADDR_TMPSIZE];
1998 struct sockaddr_dl *sdl = NULL;
1999 struct rib_head *rnh;
2001 if (flags & RTF_HOST) {
2002 dst = ifa->ifa_dstaddr;
2005 dst = ifa->ifa_addr;
2006 netmask = ifa->ifa_netmask;
2008 if (dst->sa_len == 0)
2010 switch (dst->sa_family) {
2013 /* We support multiple FIBs. */
2016 fibnum = RT_DEFAULT_FIB;
2019 if (fibnum == RT_ALL_FIBS) {
2020 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD)
2021 startfib = endfib = ifa->ifa_ifp->if_fib;
2024 endfib = rt_numfibs - 1;
2027 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
2033 * If it's a delete, check that if it exists,
2034 * it's on the correct interface or we might scrub
2035 * a route to another ifa which would
2036 * be confusing at best and possibly worse.
2038 if (cmd == RTM_DELETE) {
2040 * It's a delete, so it should already exist..
2041 * If it's a net, mask off the host bits
2042 * (Assuming we have a mask)
2043 * XXX this is kinda inet specific..
2045 if (netmask != NULL) {
2046 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
2047 dst = (struct sockaddr *)tempbuf;
2049 } else if (cmd == RTM_ADD) {
2050 sdl = (struct sockaddr_dl *)tempbuf;
2051 bzero(sdl, sizeof(struct sockaddr_dl));
2052 sdl->sdl_family = AF_LINK;
2053 sdl->sdl_len = sizeof(struct sockaddr_dl);
2054 sdl->sdl_type = ifa->ifa_ifp->if_type;
2055 sdl->sdl_index = ifa->ifa_ifp->if_index;
2058 * Now go through all the requested tables (fibs) and do the
2059 * requested action. Realistically, this will either be fib 0
2060 * for protocols that don't do multiple tables or all the
2061 * tables for those that do.
2063 for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
2064 if (cmd == RTM_DELETE) {
2065 struct radix_node *rn;
2067 * Look up an rtentry that is in the routing tree and
2068 * contains the correct info.
2070 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
2072 /* this table doesn't exist but others might */
2075 rn = rnh->rnh_lookup(dst, netmask, &rnh->head);
2077 if (rt_mpath_capable(rnh)) {
2084 * for interface route the
2085 * rt->rt_gateway is sockaddr_intf
2086 * for cloning ARP entries, so
2087 * rt_mpath_matchgate must use the
2090 rt = rt_mpath_matchgate(rt,
2097 error = (rn == NULL ||
2098 (rn->rn_flags & RNF_ROOT) ||
2099 RNTORT(rn)->rt_ifa != ifa);
2102 /* this is only an error if bad on ALL tables */
2107 * Do the actual request
2109 bzero((caddr_t)&info, sizeof(info));
2112 info.rti_flags = flags |
2113 (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED;
2114 info.rti_info[RTAX_DST] = dst;
2116 * doing this for compatibility reasons
2119 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)sdl;
2121 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
2122 info.rti_info[RTAX_NETMASK] = netmask;
2123 error = rtrequest1_fib(cmd, &info, &rt, fibnum);
2125 if (error == 0 && rt != NULL) {
2127 * notify any listening routing agents of the change
2132 * in case address alias finds the first address
2133 * e.g. ifconfig bge0 192.0.2.246/24
2134 * e.g. ifconfig bge0 192.0.2.247/24
2135 * the address set in the route is 192.0.2.246
2136 * so we need to replace it with 192.0.2.247
2138 if (memcmp(rt->rt_ifa->ifa_addr,
2139 ifa->ifa_addr, ifa->ifa_addr->sa_len)) {
2140 ifa_free(rt->rt_ifa);
2142 rt->rt_ifp = ifa->ifa_ifp;
2148 rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum);
2151 if (cmd == RTM_DELETE) {
2153 * If we are deleting, and we found an entry,
2154 * then it's been removed from the tree..
2155 * now throw it away.
2159 if (cmd == RTM_ADD) {
2161 * We just wanted to add it..
2162 * we don't actually need a reference.
2173 if (cmd == RTM_DELETE) {
2177 /* we only give an error if it wasn't in any table */
2178 error = ((flags & RTF_HOST) ?
2179 EHOSTUNREACH : ENETUNREACH);
2183 /* return an error if any of them failed */
2191 * Set up a routing table entry, normally
2195 rtinit(struct ifaddr *ifa, int cmd, int flags)
2197 struct sockaddr *dst;
2198 int fib = RT_DEFAULT_FIB;
2200 if (flags & RTF_HOST) {
2201 dst = ifa->ifa_dstaddr;
2203 dst = ifa->ifa_addr;
2206 switch (dst->sa_family) {
2209 /* We do support multiple FIBs. */
2213 return (rtinit1(ifa, cmd, flags, fib));
2217 * Announce interface address arrival/withdraw
2218 * Returns 0 on success.
2221 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
2224 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2225 ("unexpected cmd %d", cmd));
2227 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2228 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2230 #if defined(INET) || defined(INET6)
2233 * notify the SCTP stack
2234 * this will only get called when an address is added/deleted
2235 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
2237 sctp_addr_change(ifa, cmd);
2240 return (rtsock_addrmsg(cmd, ifa, fibnum));
2244 * Announce route addition/removal.
2245 * Users of this function MUST validate input data BEFORE calling.
2246 * However we have to be able to handle invalid data:
2247 * if some userland app sends us "invalid" route message (invalid mask,
2248 * no dst, wrong address families, etc...) we need to pass it back
2249 * to app (and any other rtsock consumers) with rtm_errno field set to
2251 * Returns 0 on success.
2254 rt_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
2258 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2259 ("unexpected cmd %d", cmd));
2261 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2262 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2264 KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
2266 return (rtsock_routemsg(cmd, ifp, error, rt, fibnum));
2270 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
2273 rt_newaddrmsg_fib(cmd, ifa, error, rt, RT_ALL_FIBS);
2277 * This is called to generate messages from the routing socket
2278 * indicating a network interface has had addresses associated with it.
2281 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
2285 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2286 ("unexpected cmd %u", cmd));
2287 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2288 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2290 if (cmd == RTM_ADD) {
2291 rt_addrmsg(cmd, ifa, fibnum);
2293 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2296 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2297 rt_addrmsg(cmd, ifa, fibnum);