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_mrouting.h"
42 #include "opt_mpath.h"
43 #include "opt_route.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/malloc.h>
49 #include <sys/socket.h>
50 #include <sys/sysctl.h>
51 #include <sys/syslog.h>
52 #include <sys/sysproto.h>
54 #include <sys/domain.h>
55 #include <sys/eventhandler.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 /* This is read-only.. */
94 u_int rt_numfibs = RT_NUMFIBS;
95 SYSCTL_UINT(_net, OID_AUTO, fibs, CTLFLAG_RDTUN, &rt_numfibs, 0, "");
98 * By default add routes to all fibs for new interfaces.
99 * Once this is set to 0 then only allocate routes on interface
100 * changes for the FIB of the caller when adding a new set of addresses
101 * to an interface. XXX this is a shotgun aproach to a problem that needs
102 * a more fine grained solution.. that will come.
103 * XXX also has the problems getting the FIB from curthread which will not
104 * always work given the fib can be overridden and prefixes can be added
105 * from the network stack context.
107 VNET_DEFINE(u_int, rt_add_addr_allfibs) = 1;
108 SYSCTL_UINT(_net, OID_AUTO, add_addr_allfibs, CTLFLAG_RWTUN | CTLFLAG_VNET,
109 &VNET_NAME(rt_add_addr_allfibs), 0, "");
111 VNET_DEFINE(struct rtstat, rtstat);
112 #define V_rtstat VNET(rtstat)
114 VNET_DEFINE(struct rib_head *, rt_tables);
115 #define V_rt_tables VNET(rt_tables)
117 VNET_DEFINE(int, rttrash); /* routes not in table but not freed */
118 #define V_rttrash VNET(rttrash)
122 * Convert a 'struct radix_node *' to a 'struct rtentry *'.
123 * The operation can be done safely (in this code) because a
124 * 'struct rtentry' starts with two 'struct radix_node''s, the first
125 * one representing leaf nodes in the routing tree, which is
126 * what the code in radix.c passes us as a 'struct radix_node'.
128 * But because there are a lot of assumptions in this conversion,
129 * do not cast explicitly, but always use the macro below.
131 #define RNTORT(p) ((struct rtentry *)(p))
133 VNET_DEFINE_STATIC(uma_zone_t, rtzone); /* Routing table UMA zone. */
134 #define V_rtzone VNET(rtzone)
136 EVENTHANDLER_LIST_DEFINE(rt_addrmsg);
138 static int rtrequest1_fib_change(struct rib_head *, struct rt_addrinfo *,
139 struct rtentry **, u_int);
140 static void rt_setmetrics(const struct rt_addrinfo *, struct rtentry *);
141 static int rt_ifdelroute(const struct rtentry *rt, void *arg);
142 static struct rtentry *rt_unlinkrte(struct rib_head *rnh,
143 struct rt_addrinfo *info, int *perror);
144 static void rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info);
146 static struct radix_node *rt_mpath_unlink(struct rib_head *rnh,
147 struct rt_addrinfo *info, struct rtentry *rto, int *perror);
149 static int rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info,
158 static int if_updatemtu_cb(struct radix_node *, void *);
161 * handler for net.my_fibnum
164 sysctl_my_fibnum(SYSCTL_HANDLER_ARGS)
169 fibnum = curthread->td_proc->p_fibnum;
170 error = sysctl_handle_int(oidp, &fibnum, 0, req);
174 SYSCTL_PROC(_net, OID_AUTO, my_fibnum, CTLTYPE_INT|CTLFLAG_RD,
175 NULL, 0, &sysctl_my_fibnum, "I", "default FIB of caller");
177 static __inline struct rib_head **
178 rt_tables_get_rnh_ptr(int table, int fam)
180 struct rib_head **rnh;
182 KASSERT(table >= 0 && table < rt_numfibs, ("%s: table out of bounds.",
184 KASSERT(fam >= 0 && fam < (AF_MAX+1), ("%s: fam out of bounds.",
187 /* rnh is [fib=0][af=0]. */
188 rnh = (struct rib_head **)V_rt_tables;
189 /* Get the offset to the requested table and fam. */
190 rnh += table * (AF_MAX+1) + fam;
196 rt_tables_get_rnh(int table, int fam)
199 return (*rt_tables_get_rnh_ptr(table, fam));
203 rt_tables_get_gen(int table, int fam)
205 struct rib_head *rnh;
207 rnh = *rt_tables_get_rnh_ptr(table, fam);
208 KASSERT(rnh != NULL, ("%s: NULL rib_head pointer table %d fam %d",
209 __func__, table, fam));
210 return (rnh->rnh_gen);
215 * route initialization must occur before ip6_init2(), which happenas at
222 /* whack the tunable ints into line. */
223 if (rt_numfibs > RT_MAXFIBS)
224 rt_numfibs = RT_MAXFIBS;
228 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, NULL);
231 rtentry_zinit(void *mem, int size, int how)
233 struct rtentry *rt = mem;
235 rt->rt_pksent = counter_u64_alloc(how);
236 if (rt->rt_pksent == NULL)
245 rtentry_zfini(void *mem, int size)
247 struct rtentry *rt = mem;
250 counter_u64_free(rt->rt_pksent);
254 rtentry_ctor(void *mem, int size, void *arg, int how)
256 struct rtentry *rt = mem;
258 bzero(rt, offsetof(struct rtentry, rt_endzero));
259 counter_u64_zero(rt->rt_pksent);
266 rtentry_dtor(void *mem, int size, void *arg)
268 struct rtentry *rt = mem;
274 vnet_route_init(const void *unused __unused)
277 struct rib_head **rnh;
281 V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) *
282 sizeof(struct rib_head *), M_RTABLE, M_WAITOK|M_ZERO);
284 V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry),
285 rtentry_ctor, rtentry_dtor,
286 rtentry_zinit, rtentry_zfini, UMA_ALIGN_PTR, 0);
287 for (dom = domains; dom; dom = dom->dom_next) {
288 if (dom->dom_rtattach == NULL)
291 for (table = 0; table < rt_numfibs; table++) {
292 fam = dom->dom_family;
293 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
296 rnh = rt_tables_get_rnh_ptr(table, fam);
298 panic("%s: rnh NULL", __func__);
299 dom->dom_rtattach((void **)rnh, 0);
303 VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
308 vnet_route_uninit(const void *unused __unused)
313 struct rib_head **rnh;
315 for (dom = domains; dom; dom = dom->dom_next) {
316 if (dom->dom_rtdetach == NULL)
319 for (table = 0; table < rt_numfibs; table++) {
320 fam = dom->dom_family;
322 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
325 rnh = rt_tables_get_rnh_ptr(table, fam);
327 panic("%s: rnh NULL", __func__);
328 dom->dom_rtdetach((void **)rnh, 0);
332 free(V_rt_tables, M_RTABLE);
333 uma_zdestroy(V_rtzone);
335 VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST,
336 vnet_route_uninit, 0);
340 rt_table_init(int offset)
344 rh = malloc(sizeof(struct rib_head), M_RTABLE, M_WAITOK | M_ZERO);
346 /* TODO: These details should be hidded inside radix.c */
347 /* Init masks tree */
348 rn_inithead_internal(&rh->head, rh->rnh_nodes, offset);
349 rn_inithead_internal(&rh->rmhead.head, rh->rmhead.mask_nodes, 0);
350 rh->head.rnh_masks = &rh->rmhead;
355 /* Finally, set base callbacks */
356 rh->rnh_addaddr = rn_addroute;
357 rh->rnh_deladdr = rn_delete;
358 rh->rnh_matchaddr = rn_match;
359 rh->rnh_lookup = rn_lookup;
360 rh->rnh_walktree = rn_walktree;
361 rh->rnh_walktree_from = rn_walktree_from;
367 rt_freeentry(struct radix_node *rn, void *arg)
369 struct radix_head * const rnh = arg;
370 struct radix_node *x;
372 x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh);
379 rt_table_destroy(struct rib_head *rh)
382 rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head);
384 /* Assume table is already empty */
385 RIB_LOCK_DESTROY(rh);
390 #ifndef _SYS_SYSPROTO_H_
396 sys_setfib(struct thread *td, struct setfib_args *uap)
398 if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs)
400 td->td_proc->p_fibnum = uap->fibnum;
405 * Packet routing routines.
408 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum)
412 if ((rt = ro->ro_rt) != NULL) {
413 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
418 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum);
420 RT_UNLOCK(ro->ro_rt);
424 * Look up the route that matches the address given
425 * Or, at least try.. Create a cloned route if needed.
427 * The returned route, if any, is locked.
430 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
433 return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB));
437 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags,
442 struct radix_node *rn;
443 struct rtentry *newrt;
444 struct rt_addrinfo info;
445 int err = 0, msgtype = RTM_MISS;
447 KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum"));
448 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
454 * Look up the address in the table for that Address Family
456 if ((ignflags & RTF_RNH_LOCKED) == 0)
462 rn = rh->rnh_matchaddr(dst, &rh->head);
463 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
467 if ((ignflags & RTF_RNH_LOCKED) == 0)
471 } else if ((ignflags & RTF_RNH_LOCKED) == 0)
474 * Either we hit the root or could not find any match,
475 * which basically means: "cannot get there from here".
478 V_rtstat.rts_unreach++;
482 * If required, report the failure to the supervising
484 * For a delete, this is not an error. (report == 0)
486 bzero(&info, sizeof(info));
487 info.rti_info[RTAX_DST] = dst;
488 rt_missmsg_fib(msgtype, &info, 0, err, fibnum);
494 * Remove a reference count from an rtentry.
495 * If the count gets low enough, take it out of the routing table
498 rtfree(struct rtentry *rt)
500 struct rib_head *rnh;
502 KASSERT(rt != NULL,("%s: NULL rt", __func__));
503 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
504 KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
509 * The callers should use RTFREE_LOCKED() or RTFREE(), so
510 * we should come here exactly with the last reference.
513 if (rt->rt_refcnt > 0) {
514 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt);
519 * On last reference give the "close method" a chance
520 * to cleanup private state. This also permits (for
521 * IPv4 and IPv6) a chance to decide if the routing table
522 * entry should be purged immediately or at a later time.
523 * When an immediate purge is to happen the close routine
524 * typically calls rtexpunge which clears the RTF_UP flag
525 * on the entry so that the code below reclaims the storage.
527 if (rt->rt_refcnt == 0 && rnh->rnh_close)
528 rnh->rnh_close((struct radix_node *)rt, &rnh->head);
531 * If we are no longer "up" (and ref == 0)
532 * then we can free the resources associated
535 if ((rt->rt_flags & RTF_UP) == 0) {
536 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
539 * the rtentry must have been removed from the routing table
540 * so it is represented in rttrash.. remove that now.
544 if (rt->rt_refcnt < 0) {
545 printf("rtfree: %p not freed (neg refs)\n", rt);
550 * release references on items we hold them on..
551 * e.g other routes and ifaddrs.
554 ifa_free(rt->rt_ifa);
556 * The key is separatly alloc'd so free it (see rt_setgate()).
557 * This also frees the gateway, as they are always malloc'd
563 * and the rtentry itself of course
565 uma_zfree(V_rtzone, rt);
574 * Force a routing table entry to the specified
575 * destination to go through the given gateway.
576 * Normally called as a result of a routing redirect
577 * message from the network layer.
580 rtredirect_fib(struct sockaddr *dst,
581 struct sockaddr *gateway,
582 struct sockaddr *netmask,
584 struct sockaddr *src,
590 struct rt_addrinfo info;
592 struct rib_head *rnh;
596 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
598 error = EAFNOSUPPORT;
601 /* verify the gateway is directly reachable */
602 if ((ifa = ifa_ifwithnet(gateway, 0, fibnum)) == NULL) {
606 rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */
608 * If the redirect isn't from our current router for this dst,
609 * it's either old or wrong. If it redirects us to ourselves,
610 * we have a routing loop, perhaps as a result of an interface
611 * going down recently.
613 if (!(flags & RTF_DONE) && rt) {
614 if (!sa_equal(src, rt->rt_gateway)) {
618 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) {
623 if ((flags & RTF_GATEWAY) && ifa_ifwithaddr_check(gateway)) {
624 error = EHOSTUNREACH;
628 * Create a new entry if we just got back a wildcard entry
629 * or the lookup failed. This is necessary for hosts
630 * which use routing redirects generated by smart gateways
631 * to dynamically build the routing tables.
633 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
636 * Don't listen to the redirect if it's
637 * for a route to an interface.
639 if (rt->rt_flags & RTF_GATEWAY) {
640 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
642 * Changing from route to net => route to host.
643 * Create new route, rather than smashing route to net.
649 flags |= RTF_DYNAMIC;
650 bzero((caddr_t)&info, sizeof(info));
651 info.rti_info[RTAX_DST] = dst;
652 info.rti_info[RTAX_GATEWAY] = gateway;
653 info.rti_info[RTAX_NETMASK] = netmask;
656 info.rti_flags = flags;
657 error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum);
660 flags = rt->rt_flags;
663 stat = &V_rtstat.rts_dynamic;
667 * Smash the current notion of the gateway to
668 * this destination. Should check about netmask!!!
670 if ((flags & RTF_GATEWAY) == 0)
671 rt->rt_flags &= ~RTF_GATEWAY;
672 rt->rt_flags |= RTF_MODIFIED;
673 flags |= RTF_MODIFIED;
674 stat = &V_rtstat.rts_newgateway;
676 * add the key and gateway (in one malloc'd chunk).
681 rt_setgate(rt, rt_key(rt), gateway);
685 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);
704 * Routing table ioctl interface.
707 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
711 * If more ioctl commands are added here, make sure the proper
712 * super-user checks are being performed because it is possible for
713 * prison-root to make it this far if raw sockets have been enabled
717 /* Multicast goop, grrr... */
718 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
725 ifa_ifwithroute(int flags, const struct sockaddr *dst, struct sockaddr *gateway,
731 MPASS(in_epoch(net_epoch_preempt));
732 if ((flags & RTF_GATEWAY) == 0) {
734 * If we are adding a route to an interface,
735 * and the interface is a pt to pt link
736 * we should search for the destination
737 * as our clue to the interface. Otherwise
738 * we can use the local address.
741 if (flags & RTF_HOST)
742 ifa = ifa_ifwithdstaddr(dst, fibnum);
744 ifa = ifa_ifwithaddr(gateway);
747 * If we are adding a route to a remote net
748 * or host, the gateway may still be on the
749 * other end of a pt to pt link.
751 ifa = ifa_ifwithdstaddr(gateway, fibnum);
754 ifa = ifa_ifwithnet(gateway, 0, fibnum);
758 rt = rtalloc1_fib(gateway, 0, flags, fibnum);
762 * dismiss a gateway that is reachable only
763 * through the default router
765 switch (gateway->sa_family) {
767 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
771 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
777 if (!not_found && rt->rt_ifa != NULL) {
782 if (not_found || ifa == NULL)
785 if (ifa->ifa_addr->sa_family != dst->sa_family) {
786 struct ifaddr *oifa = ifa;
787 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
796 * Do appropriate manipulations of a routing tree given
797 * all the bits of info needed
800 rtrequest_fib(int req,
801 struct sockaddr *dst,
802 struct sockaddr *gateway,
803 struct sockaddr *netmask,
805 struct rtentry **ret_nrt,
808 struct rt_addrinfo info;
810 if (dst->sa_len == 0)
813 bzero((caddr_t)&info, sizeof(info));
814 info.rti_flags = flags;
815 info.rti_info[RTAX_DST] = dst;
816 info.rti_info[RTAX_GATEWAY] = gateway;
817 info.rti_info[RTAX_NETMASK] = netmask;
818 return rtrequest1_fib(req, &info, ret_nrt, fibnum);
823 * Copy most of @rt data into @info.
825 * If @flags contains NHR_COPY, copies dst,netmask and gw to the
826 * pointers specified by @info structure. Assume such pointers
827 * are zeroed sockaddr-like structures with sa_len field initialized
828 * to reflect size of the provided buffer. if no NHR_COPY is specified,
829 * point dst,netmask and gw @info fields to appropriate @rt values.
831 * if @flags contains NHR_REF, do refcouting on rt_ifp and rt_ifa.
833 * Returns 0 on success.
836 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags)
838 struct rt_metrics *rmx;
839 struct sockaddr *src, *dst;
842 if (flags & NHR_COPY) {
843 /* Copy destination if dst is non-zero */
845 dst = info->rti_info[RTAX_DST];
846 sa_len = src->sa_len;
848 if (src->sa_len > dst->sa_len)
850 memcpy(dst, src, src->sa_len);
851 info->rti_addrs |= RTA_DST;
854 /* Copy mask if set && dst is non-zero */
856 dst = info->rti_info[RTAX_NETMASK];
857 if (src != NULL && dst != NULL) {
860 * Radix stores different value in sa_len,
861 * assume rt_mask() to have the same length
864 if (sa_len > dst->sa_len)
866 memcpy(dst, src, src->sa_len);
867 info->rti_addrs |= RTA_NETMASK;
870 /* Copy gateway is set && dst is non-zero */
871 src = rt->rt_gateway;
872 dst = info->rti_info[RTAX_GATEWAY];
873 if ((rt->rt_flags & RTF_GATEWAY) && src != NULL && dst != NULL){
874 if (src->sa_len > dst->sa_len)
876 memcpy(dst, src, src->sa_len);
877 info->rti_addrs |= RTA_GATEWAY;
880 info->rti_info[RTAX_DST] = rt_key(rt);
881 info->rti_addrs |= RTA_DST;
882 if (rt_mask(rt) != NULL) {
883 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
884 info->rti_addrs |= RTA_NETMASK;
886 if (rt->rt_flags & RTF_GATEWAY) {
887 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
888 info->rti_addrs |= RTA_GATEWAY;
894 info->rti_mflags |= RTV_MTU;
895 rmx->rmx_mtu = rt->rt_mtu;
898 info->rti_flags = rt->rt_flags;
899 info->rti_ifp = rt->rt_ifp;
900 info->rti_ifa = rt->rt_ifa;
901 if (flags & NHR_REF) {
902 if_ref(info->rti_ifp);
903 ifa_ref(info->rti_ifa);
910 * Lookups up route entry for @dst in RIB database for fib @fibnum.
911 * Exports entry data to @info using rt_exportinfo().
913 * If @flags contains NHR_REF, refcouting is performed on rt_ifp and rt_ifa.
914 * All references can be released later by calling rib_free_info().
916 * Returns 0 on success.
917 * Returns ENOENT for lookup failure, ENOMEM for export failure.
920 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
921 uint32_t flowid, struct rt_addrinfo *info)
925 struct radix_node *rn;
929 KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
930 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
935 rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head);
936 if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
938 /* Ensure route & ifp is UP */
939 if (RT_LINK_IS_UP(rt->rt_ifp)) {
940 flags = (flags & NHR_REF) | NHR_COPY;
941 error = rt_exportinfo(rt, info, flags);
953 * Releases all references acquired by rib_lookup_info() when
954 * called with NHR_REF flags.
957 rib_free_info(struct rt_addrinfo *info)
960 ifa_free(info->rti_ifa);
961 if_rele(info->rti_ifp);
965 * Iterates over all existing fibs in system calling
966 * @setwa_f function prior to traversing each fib.
967 * Calls @wa_f function for each element in current fib.
968 * If af is not AF_UNSPEC, iterates over fibs in particular
972 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f,
975 struct rib_head *rnh;
979 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
980 /* Do we want some specific family? */
981 if (af != AF_UNSPEC) {
982 rnh = rt_tables_get_rnh(fibnum, af);
986 setwa_f(rnh, fibnum, af, arg);
989 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
994 for (i = 1; i <= AF_MAX; i++) {
995 rnh = rt_tables_get_rnh(fibnum, i);
999 setwa_f(rnh, fibnum, i, arg);
1002 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
1010 struct rt_addrinfo info;
1011 struct rib_head *rnh;
1012 struct rtentry *head;
1016 * Conditionally unlinks @rn from radix tree based
1017 * on info data passed in @arg.
1020 rt_checkdelroute(struct radix_node *rn, void *arg)
1022 struct rt_delinfo *di;
1023 struct rt_addrinfo *info;
1027 di = (struct rt_delinfo *)arg;
1028 rt = (struct rtentry *)rn;
1032 info->rti_info[RTAX_DST] = rt_key(rt);
1033 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
1034 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1036 rt = rt_unlinkrte(di->rnh, info, &error);
1038 /* Either not allowed or not matched. Skip entry */
1042 /* Entry was unlinked. Add to the list and return */
1043 rt->rt_chain = di->head;
1050 * Iterates over all existing fibs in system.
1051 * Deletes each element for which @filter_f function returned
1053 * If @af is not AF_UNSPEC, iterates over fibs in particular
1057 rt_foreach_fib_walk_del(int af, rt_filter_f_t *filter_f, void *arg)
1059 struct rib_head *rnh;
1060 struct rt_delinfo di;
1065 bzero(&di, sizeof(di));
1066 di.info.rti_filter = filter_f;
1067 di.info.rti_filterdata = arg;
1069 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1070 /* Do we want some specific family? */
1071 if (af != AF_UNSPEC) {
1079 for (i = start; i <= end; i++) {
1080 rnh = rt_tables_get_rnh(fibnum, i);
1086 rnh->rnh_walktree(&rnh->head, rt_checkdelroute, &di);
1089 if (di.head == NULL)
1092 /* We might have something to reclaim */
1093 while (di.head != NULL) {
1095 di.head = rt->rt_chain;
1096 rt->rt_chain = NULL;
1098 /* TODO std rt -> rt_addrinfo export */
1099 di.info.rti_info[RTAX_DST] = rt_key(rt);
1100 di.info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1102 rt_notifydelete(rt, &di.info);
1111 * Delete Routes for a Network Interface
1113 * Called for each routing entry via the rnh->rnh_walktree() call above
1114 * to delete all route entries referencing a detaching network interface.
1117 * rt pointer to rtentry
1118 * arg argument passed to rnh->rnh_walktree() - detaching interface
1122 * errno failed - reason indicated
1125 rt_ifdelroute(const struct rtentry *rt, void *arg)
1127 struct ifnet *ifp = arg;
1129 if (rt->rt_ifp != ifp)
1133 * Protect (sorta) against walktree recursion problems
1134 * with cloned routes
1136 if ((rt->rt_flags & RTF_UP) == 0)
1143 * Delete all remaining routes using this interface
1144 * Unfortuneatly the only way to do this is to slog through
1145 * the entire routing table looking for routes which point
1146 * to this interface...oh well...
1149 rt_flushifroutes_af(struct ifnet *ifp, int af)
1151 KASSERT((af >= 1 && af <= AF_MAX), ("%s: af %d not >= 1 and <= %d",
1152 __func__, af, AF_MAX));
1154 rt_foreach_fib_walk_del(af, rt_ifdelroute, ifp);
1158 rt_flushifroutes(struct ifnet *ifp)
1161 rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
1165 * Conditionally unlinks rtentry matching data inside @info from @rnh.
1166 * Returns unlinked, locked and referenced @rtentry on success,
1167 * Returns NULL and sets @perror to:
1168 * ESRCH - if prefix was not found,
1169 * EADDRINUSE - if trying to delete PINNED route without appropriate flag.
1170 * ENOENT - if supplied filter function returned 0 (not matched).
1172 static struct rtentry *
1173 rt_unlinkrte(struct rib_head *rnh, struct rt_addrinfo *info, int *perror)
1175 struct sockaddr *dst, *netmask;
1177 struct radix_node *rn;
1179 dst = info->rti_info[RTAX_DST];
1180 netmask = info->rti_info[RTAX_NETMASK];
1182 rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, &rnh->head);
1188 if ((info->rti_flags & RTF_PINNED) == 0) {
1189 /* Check if target route can be deleted */
1190 if (rt->rt_flags & RTF_PINNED) {
1191 *perror = EADDRINUSE;
1196 if (info->rti_filter != NULL) {
1197 if (info->rti_filter(rt, info->rti_filterdata) == 0) {
1204 * Filter function requested rte deletion.
1205 * Ease the caller work by filling in remaining info
1206 * from that particular entry.
1208 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1212 * Remove the item from the tree and return it.
1213 * Complain if it is not there and do no more processing.
1217 if (rt_mpath_capable(rnh))
1218 rn = rt_mpath_unlink(rnh, info, rt, perror);
1221 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1225 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
1226 panic ("rtrequest delete");
1231 rt->rt_flags &= ~RTF_UP;
1239 rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info)
1244 * give the protocol a chance to keep things in sync.
1247 if (ifa != NULL && ifa->ifa_rtrequest != NULL)
1248 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1251 * One more rtentry floating around that is not
1252 * linked to the routing table. rttrash will be decremented
1253 * when RTFREE(rt) is eventually called.
1260 * These (questionable) definitions of apparent local variables apply
1261 * to the next two functions. XXXXXX!!!
1263 #define dst info->rti_info[RTAX_DST]
1264 #define gateway info->rti_info[RTAX_GATEWAY]
1265 #define netmask info->rti_info[RTAX_NETMASK]
1266 #define ifaaddr info->rti_info[RTAX_IFA]
1267 #define ifpaddr info->rti_info[RTAX_IFP]
1268 #define flags info->rti_flags
1271 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined,
1272 * it will be referenced so the caller must free it.
1275 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
1281 * ifp may be specified by sockaddr_dl
1282 * when protocol address is ambiguous.
1285 needref = (info->rti_ifa == NULL);
1287 if (info->rti_ifp == NULL && ifpaddr != NULL &&
1288 ifpaddr->sa_family == AF_LINK &&
1289 (ifa = ifa_ifwithnet(ifpaddr, 0, fibnum)) != NULL) {
1290 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 (needref && info->rti_ifa != NULL) {
1309 if (info->rti_ifp == NULL)
1310 info->rti_ifp = info->rti_ifa->ifa_ifp;
1311 ifa_ref(info->rti_ifa);
1313 error = ENETUNREACH;
1319 if_updatemtu_cb(struct radix_node *rn, void *arg)
1322 struct if_mtuinfo *ifmtu;
1324 rt = (struct rtentry *)rn;
1325 ifmtu = (struct if_mtuinfo *)arg;
1327 if (rt->rt_ifp != ifmtu->ifp)
1330 if (rt->rt_mtu >= ifmtu->mtu) {
1331 /* We have to decrease mtu regardless of flags */
1332 rt->rt_mtu = ifmtu->mtu;
1337 * New MTU is bigger. Check if are allowed to alter it
1339 if ((rt->rt_flags & (RTF_FIXEDMTU | RTF_GATEWAY | RTF_HOST)) != 0) {
1342 * Skip routes with user-supplied MTU and
1343 * non-interface routes
1348 /* We are safe to update route MTU */
1349 rt->rt_mtu = ifmtu->mtu;
1355 rt_updatemtu(struct ifnet *ifp)
1357 struct if_mtuinfo ifmtu;
1358 struct rib_head *rnh;
1364 * Try to update rt_mtu for all routes using this interface
1365 * Unfortunately the only way to do this is to traverse all
1366 * routing tables in all fibs/domains.
1368 for (i = 1; i <= AF_MAX; i++) {
1369 ifmtu.mtu = if_getmtu_family(ifp, i);
1370 for (j = 0; j < rt_numfibs; j++) {
1371 rnh = rt_tables_get_rnh(j, i);
1375 rnh->rnh_walktree(&rnh->head, if_updatemtu_cb, &ifmtu);
1383 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
1384 int rt_print(char *buf, int buflen, struct rtentry *rt);
1387 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
1391 switch (s->sa_family) {
1393 paddr = &((struct sockaddr_in *)s)->sin_addr;
1396 paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
1403 if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
1406 return (strlen(buf));
1410 rt_print(char *buf, int buflen, struct rtentry *rt)
1412 struct sockaddr *addr, *mask;
1418 i = p_sockaddr(buf, buflen, addr);
1419 if (!(rt->rt_flags & RTF_HOST)) {
1421 i += p_sockaddr(buf + i, buflen - i, mask);
1424 if (rt->rt_flags & RTF_GATEWAY) {
1426 i += p_sockaddr(buf + i, buflen - i, rt->rt_gateway);
1435 * Deletes key for single-path routes, unlinks rtentry with
1436 * gateway specified in @info from multi-path routes.
1438 * Returnes unlinked entry. In case of failure, returns NULL
1439 * and sets @perror to ESRCH.
1441 static struct radix_node *
1442 rt_mpath_unlink(struct rib_head *rnh, struct rt_addrinfo *info,
1443 struct rtentry *rto, int *perror)
1446 * if we got multipath routes, we require users to specify
1447 * a matching RTAX_GATEWAY.
1449 struct rtentry *rt; // *rto = NULL;
1450 struct radix_node *rn;
1451 struct sockaddr *gw;
1453 gw = info->rti_info[RTAX_GATEWAY];
1454 rt = rt_mpath_matchgate(rto, gw);
1461 * this is the first entry in the chain
1464 rn = rn_mpath_next((struct radix_node *)rt);
1466 * there is another entry, now it's active
1471 rto->rt_flags |= RTF_UP;
1473 } else if (rt->rt_flags & RTF_GATEWAY) {
1475 * For gateway routes, we need to
1476 * make sure that we we are deleting
1477 * the correct gateway.
1478 * rt_mpath_matchgate() does not
1479 * check the case when there is only
1480 * one route in the chain.
1483 (rt->rt_gateway->sa_len != gw->sa_len ||
1484 memcmp(rt->rt_gateway, gw, gw->sa_len))) {
1491 * use the normal delete code to remove
1494 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1500 * if the entry is 2nd and on up
1502 if (rt_mpath_deldup(rto, rt) == 0)
1503 panic ("rtrequest1: rt_mpath_deldup");
1505 rn = (struct radix_node *)rt;
1511 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt,
1515 struct rtentry *rt, *rt_old;
1516 struct radix_node *rn;
1517 struct rib_head *rnh;
1519 struct sockaddr *ndst;
1520 struct sockaddr_storage mdst;
1522 KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum"));
1523 KASSERT((flags & RTF_RNH_LOCKED) == 0, ("rtrequest1_fib: locked"));
1524 switch (dst->sa_family) {
1527 /* We support multiple FIBs. */
1530 fibnum = RT_DEFAULT_FIB;
1535 * Find the correct routing tree to use for this Address Family
1537 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
1539 return (EAFNOSUPPORT);
1542 * If we are adding a host route then we don't want to put
1543 * a netmask in the tree, nor do we want to clone it.
1545 if (flags & RTF_HOST)
1551 if (dst->sa_len > sizeof(mdst))
1553 rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask);
1554 dst = (struct sockaddr *)&mdst;
1558 rt = rt_unlinkrte(rnh, info, &error);
1563 rt_notifydelete(rt, info);
1566 * If the caller wants it, then it can have it,
1567 * but it's up to it to free the rtentry as we won't be
1578 * resolve was only used for route cloning
1583 if ((flags & RTF_GATEWAY) && !gateway)
1585 if (dst && gateway && (dst->sa_family != gateway->sa_family) &&
1586 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
1589 if (info->rti_ifa == NULL) {
1590 error = rt_getifa_fib(info, fibnum);
1594 ifa_ref(info->rti_ifa);
1596 rt = uma_zalloc(V_rtzone, M_NOWAIT);
1598 ifa_free(info->rti_ifa);
1601 rt->rt_flags = RTF_UP | flags;
1602 rt->rt_fibnum = fibnum;
1604 * Add the gateway. Possibly re-malloc-ing the storage for it.
1606 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
1607 ifa_free(info->rti_ifa);
1608 uma_zfree(V_rtzone, rt);
1613 * point to the (possibly newly malloc'd) dest address.
1615 ndst = (struct sockaddr *)rt_key(rt);
1618 * make sure it contains the value we want (masked if needed).
1621 rt_maskedcopy(dst, ndst, netmask);
1623 bcopy(dst, ndst, dst->sa_len);
1626 * We use the ifa reference returned by rt_getifa_fib().
1627 * This moved from below so that rnh->rnh_addaddr() can
1628 * examine the ifa and ifa->ifa_ifp if it so desires.
1630 ifa = info->rti_ifa;
1632 rt->rt_ifp = ifa->ifa_ifp;
1635 rt_setmetrics(info, rt);
1640 /* do not permit exactly the same dst/mask/gw pair */
1641 if (rt_mpath_capable(rnh) &&
1642 rt_mpath_conflict(rnh, rt, netmask)) {
1645 ifa_free(rt->rt_ifa);
1647 uma_zfree(V_rtzone, rt);
1652 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
1653 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head, rt->rt_nodes);
1656 if (rn == NULL && (info->rti_flags & RTF_PINNED) != 0) {
1659 * Force removal and re-try addition
1660 * TODO: better multipath&pinned support
1662 struct sockaddr *info_dst = info->rti_info[RTAX_DST];
1663 info->rti_info[RTAX_DST] = ndst;
1664 /* Do not delete existing PINNED(interface) routes */
1665 info->rti_flags &= ~RTF_PINNED;
1666 rt_old = rt_unlinkrte(rnh, info, &error);
1667 info->rti_flags |= RTF_PINNED;
1668 info->rti_info[RTAX_DST] = info_dst;
1670 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head,
1679 * If it still failed to go into the tree,
1680 * then un-make it (this should be a function)
1683 ifa_free(rt->rt_ifa);
1685 uma_zfree(V_rtzone, rt);
1689 if (rt_old != NULL) {
1690 rt_notifydelete(rt_old, info);
1695 * If this protocol has something to add to this then
1696 * allow it to do that as well.
1698 if (ifa->ifa_rtrequest)
1699 ifa->ifa_rtrequest(req, rt, info);
1702 * actually return a resultant rtentry and
1703 * give the caller a single reference.
1709 rnh->rnh_gen++; /* Routing table updated */
1714 error = rtrequest1_fib_change(rnh, info, ret_nrt, fibnum);
1732 rtrequest1_fib_change(struct rib_head *rnh, struct rt_addrinfo *info,
1733 struct rtentry **ret_nrt, u_int fibnum)
1735 struct rtentry *rt = NULL;
1739 struct if_mtuinfo ifmtu;
1741 RIB_WLOCK_ASSERT(rnh);
1743 rt = (struct rtentry *)rnh->rnh_lookup(info->rti_info[RTAX_DST],
1744 info->rti_info[RTAX_NETMASK], &rnh->head);
1751 * If we got multipath routes,
1752 * we require users to specify a matching RTAX_GATEWAY.
1754 if (rt_mpath_capable(rnh)) {
1755 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
1763 rt_setmetrics(info, rt);
1766 * New gateway could require new ifaddr, ifp;
1767 * flags may also be different; ifp may be specified
1768 * by ll sockaddr when protocol address is ambiguous
1770 if (((rt->rt_flags & RTF_GATEWAY) &&
1771 info->rti_info[RTAX_GATEWAY] != NULL) ||
1772 info->rti_info[RTAX_IFP] != NULL ||
1773 (info->rti_info[RTAX_IFA] != NULL &&
1774 !sa_equal(info->rti_info[RTAX_IFA], rt->rt_ifa->ifa_addr))) {
1776 * XXX: Temporarily set RTF_RNH_LOCKED flag in the rti_flags
1777 * to avoid rlock in the ifa_ifwithroute().
1779 info->rti_flags |= RTF_RNH_LOCKED;
1780 error = rt_getifa_fib(info, fibnum);
1781 info->rti_flags &= ~RTF_RNH_LOCKED;
1782 if (info->rti_ifa != NULL)
1789 /* Check if outgoing interface has changed */
1790 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa &&
1791 rt->rt_ifa != NULL) {
1792 if (rt->rt_ifa->ifa_rtrequest != NULL)
1793 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1794 ifa_free(rt->rt_ifa);
1797 /* Update gateway address */
1798 if (info->rti_info[RTAX_GATEWAY] != NULL) {
1799 error = rt_setgate(rt, rt_key(rt), info->rti_info[RTAX_GATEWAY]);
1803 rt->rt_flags &= ~RTF_GATEWAY;
1804 rt->rt_flags |= (RTF_GATEWAY & info->rti_flags);
1807 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa) {
1808 ifa_ref(info->rti_ifa);
1809 rt->rt_ifa = info->rti_ifa;
1810 rt->rt_ifp = info->rti_ifp;
1812 /* Allow some flags to be toggled on change. */
1813 rt->rt_flags &= ~RTF_FMASK;
1814 rt->rt_flags |= info->rti_flags & RTF_FMASK;
1816 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest != NULL)
1817 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
1819 /* Alter route MTU if necessary */
1820 if (rt->rt_ifp != NULL) {
1821 family = info->rti_info[RTAX_DST]->sa_family;
1822 mtu = if_getmtu_family(rt->rt_ifp, family);
1823 /* Set default MTU */
1824 if (rt->rt_mtu == 0)
1826 if (rt->rt_mtu != mtu) {
1827 /* Check if we really need to update */
1828 ifmtu.ifp = rt->rt_ifp;
1830 if_updatemtu_cb(rt->rt_nodes, &ifmtu);
1835 * This route change may have modified the route's gateway. In that
1836 * case, any inpcbs that have cached this route need to invalidate their
1847 if (free_ifa != 0) {
1848 ifa_free(info->rti_ifa);
1849 info->rti_ifa = NULL;
1855 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt)
1858 if (info->rti_mflags & RTV_MTU) {
1859 if (info->rti_rmx->rmx_mtu != 0) {
1862 * MTU was explicitly provided by user.
1865 rt->rt_flags |= RTF_FIXEDMTU;
1869 * User explicitly sets MTU to 0.
1870 * Assume rollback to default.
1872 rt->rt_flags &= ~RTF_FIXEDMTU;
1874 rt->rt_mtu = info->rti_rmx->rmx_mtu;
1876 if (info->rti_mflags & RTV_WEIGHT)
1877 rt->rt_weight = info->rti_rmx->rmx_weight;
1878 /* Kernel -> userland timebase conversion. */
1879 if (info->rti_mflags & RTV_EXPIRE)
1880 rt->rt_expire = info->rti_rmx->rmx_expire ?
1881 info->rti_rmx->rmx_expire - time_second + time_uptime : 0;
1885 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
1887 /* XXX dst may be overwritten, can we move this to below */
1888 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
1891 * Prepare to store the gateway in rt->rt_gateway.
1892 * Both dst and gateway are stored one after the other in the same
1893 * malloc'd chunk. If we have room, we can reuse the old buffer,
1894 * rt_gateway already points to the right place.
1895 * Otherwise, malloc a new block and update the 'dst' address.
1897 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
1900 R_Malloc(new, caddr_t, dlen + glen);
1904 * XXX note, we copy from *dst and not *rt_key(rt) because
1905 * rt_setgate() can be called to initialize a newly
1906 * allocated route entry, in which case rt_key(rt) == NULL
1907 * (and also rt->rt_gateway == NULL).
1908 * Free()/free() handle a NULL argument just fine.
1910 bcopy(dst, new, dlen);
1911 R_Free(rt_key(rt)); /* free old block, if any */
1912 rt_key(rt) = (struct sockaddr *)new;
1913 rt->rt_gateway = (struct sockaddr *)(new + dlen);
1917 * Copy the new gateway value into the memory chunk.
1919 bcopy(gate, rt->rt_gateway, glen);
1925 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
1927 u_char *cp1 = (u_char *)src;
1928 u_char *cp2 = (u_char *)dst;
1929 u_char *cp3 = (u_char *)netmask;
1930 u_char *cplim = cp2 + *cp3;
1931 u_char *cplim2 = cp2 + *cp1;
1933 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1938 *cp2++ = *cp1++ & *cp3++;
1940 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1944 * Set up a routing table entry, normally
1947 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
1949 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
1952 struct sockaddr *dst;
1953 struct sockaddr *netmask;
1954 struct rtentry *rt = NULL;
1955 struct rt_addrinfo info;
1957 int startfib, endfib;
1958 char tempbuf[_SOCKADDR_TMPSIZE];
1961 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1962 struct rib_head *rnh;
1964 if (flags & RTF_HOST) {
1965 dst = ifa->ifa_dstaddr;
1968 dst = ifa->ifa_addr;
1969 netmask = ifa->ifa_netmask;
1971 if (dst->sa_len == 0)
1973 switch (dst->sa_family) {
1976 /* We support multiple FIBs. */
1979 fibnum = RT_DEFAULT_FIB;
1982 if (fibnum == RT_ALL_FIBS) {
1983 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD)
1984 startfib = endfib = ifa->ifa_ifp->if_fib;
1987 endfib = rt_numfibs - 1;
1990 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
1996 * If it's a delete, check that if it exists,
1997 * it's on the correct interface or we might scrub
1998 * a route to another ifa which would
1999 * be confusing at best and possibly worse.
2001 if (cmd == RTM_DELETE) {
2003 * It's a delete, so it should already exist..
2004 * If it's a net, mask off the host bits
2005 * (Assuming we have a mask)
2006 * XXX this is kinda inet specific..
2008 if (netmask != NULL) {
2009 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
2010 dst = (struct sockaddr *)tempbuf;
2014 * Now go through all the requested tables (fibs) and do the
2015 * requested action. Realistically, this will either be fib 0
2016 * for protocols that don't do multiple tables or all the
2017 * tables for those that do.
2019 for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
2020 if (cmd == RTM_DELETE) {
2021 struct radix_node *rn;
2023 * Look up an rtentry that is in the routing tree and
2024 * contains the correct info.
2026 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
2028 /* this table doesn't exist but others might */
2031 rn = rnh->rnh_lookup(dst, netmask, &rnh->head);
2033 if (rt_mpath_capable(rnh)) {
2040 * for interface route the
2041 * rt->rt_gateway is sockaddr_intf
2042 * for cloning ARP entries, so
2043 * rt_mpath_matchgate must use the
2046 rt = rt_mpath_matchgate(rt,
2053 error = (rn == NULL ||
2054 (rn->rn_flags & RNF_ROOT) ||
2055 RNTORT(rn)->rt_ifa != ifa);
2058 /* this is only an error if bad on ALL tables */
2063 * Do the actual request
2065 bzero((caddr_t)&info, sizeof(info));
2067 info.rti_flags = flags |
2068 (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED;
2069 info.rti_info[RTAX_DST] = dst;
2071 * doing this for compatibility reasons
2074 info.rti_info[RTAX_GATEWAY] =
2075 (struct sockaddr *)&null_sdl;
2077 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
2078 info.rti_info[RTAX_NETMASK] = netmask;
2079 error = rtrequest1_fib(cmd, &info, &rt, fibnum);
2080 if (error == 0 && rt != NULL) {
2082 * notify any listening routing agents of the change
2087 * in case address alias finds the first address
2088 * e.g. ifconfig bge0 192.0.2.246/24
2089 * e.g. ifconfig bge0 192.0.2.247/24
2090 * the address set in the route is 192.0.2.246
2091 * so we need to replace it with 192.0.2.247
2093 if (memcmp(rt->rt_ifa->ifa_addr,
2094 ifa->ifa_addr, ifa->ifa_addr->sa_len)) {
2095 ifa_free(rt->rt_ifa);
2097 rt->rt_ifp = ifa->ifa_ifp;
2102 * doing this for compatibility reasons
2104 if (cmd == RTM_ADD) {
2105 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
2106 rt->rt_ifp->if_type;
2107 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
2108 rt->rt_ifp->if_index;
2112 rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum);
2115 if (cmd == RTM_DELETE) {
2117 * If we are deleting, and we found an entry,
2118 * then it's been removed from the tree..
2119 * now throw it away.
2123 if (cmd == RTM_ADD) {
2125 * We just wanted to add it..
2126 * we don't actually need a reference.
2137 if (cmd == RTM_DELETE) {
2141 /* we only give an error if it wasn't in any table */
2142 error = ((flags & RTF_HOST) ?
2143 EHOSTUNREACH : ENETUNREACH);
2147 /* return an error if any of them failed */
2155 * Set up a routing table entry, normally
2159 rtinit(struct ifaddr *ifa, int cmd, int flags)
2161 struct sockaddr *dst;
2162 int fib = RT_DEFAULT_FIB;
2164 if (flags & RTF_HOST) {
2165 dst = ifa->ifa_dstaddr;
2167 dst = ifa->ifa_addr;
2170 switch (dst->sa_family) {
2173 /* We do support multiple FIBs. */
2177 return (rtinit1(ifa, cmd, flags, fib));
2181 * Announce interface address arrival/withdraw
2182 * Returns 0 on success.
2185 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
2188 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2189 ("unexpected cmd %d", cmd));
2190 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2191 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2193 EVENTHANDLER_DIRECT_INVOKE(rt_addrmsg, ifa, cmd);
2194 return (rtsock_addrmsg(cmd, ifa, fibnum));
2198 * Announce route addition/removal.
2199 * Users of this function MUST validate input data BEFORE calling.
2200 * However we have to be able to handle invalid data:
2201 * if some userland app sends us "invalid" route message (invalid mask,
2202 * no dst, wrong address families, etc...) we need to pass it back
2203 * to app (and any other rtsock consumers) with rtm_errno field set to
2205 * Returns 0 on success.
2208 rt_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
2212 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2213 ("unexpected cmd %d", cmd));
2215 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2216 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2218 KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
2220 return (rtsock_routemsg(cmd, ifp, error, rt, fibnum));
2224 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
2227 rt_newaddrmsg_fib(cmd, ifa, error, rt, RT_ALL_FIBS);
2231 * This is called to generate messages from the routing socket
2232 * indicating a network interface has had addresses associated with it.
2235 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
2239 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2240 ("unexpected cmd %u", cmd));
2241 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2242 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2244 if (cmd == RTM_ADD) {
2245 rt_addrmsg(cmd, ifa, fibnum);
2247 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2250 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2251 rt_addrmsg(cmd, ifa, fibnum);