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/route_ctl.h>
65 #include <net/route/route_var.h>
66 #include <net/route/nhop.h>
70 #include <net/radix_mpath.h>
73 #include <netinet/in.h>
74 #include <netinet/ip_mroute.h>
77 * By default add routes to all fibs for new interfaces.
78 * Once this is set to 0 then only allocate routes on interface
79 * changes for the FIB of the caller when adding a new set of addresses
80 * to an interface. XXX this is a shotgun aproach to a problem that needs
81 * a more fine grained solution.. that will come.
82 * XXX also has the problems getting the FIB from curthread which will not
83 * always work given the fib can be overridden and prefixes can be added
84 * from the network stack context.
86 VNET_DEFINE(u_int, rt_add_addr_allfibs) = 1;
87 SYSCTL_UINT(_net, OID_AUTO, add_addr_allfibs, CTLFLAG_RWTUN | CTLFLAG_VNET,
88 &VNET_NAME(rt_add_addr_allfibs), 0, "");
90 VNET_PCPUSTAT_DEFINE(struct rtstat, rtstat);
92 VNET_PCPUSTAT_SYSINIT(rtstat);
94 VNET_PCPUSTAT_SYSUNINIT(rtstat);
97 EVENTHANDLER_LIST_DEFINE(rt_addrmsg);
99 static int rt_ifdelroute(const struct rtentry *rt, const struct nhop_object *,
101 static int rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info,
105 * route initialization must occur before ip6_init2(), which happenas at
114 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, NULL);
117 rt_table_init(int offset, int family, u_int fibnum)
121 rh = malloc(sizeof(struct rib_head), M_RTABLE, M_WAITOK | M_ZERO);
123 /* TODO: These details should be hidded inside radix.c */
124 /* Init masks tree */
125 rn_inithead_internal(&rh->head, rh->rnh_nodes, offset);
126 rn_inithead_internal(&rh->rmhead.head, rh->rmhead.mask_nodes, 0);
127 rh->head.rnh_masks = &rh->rmhead;
129 /* Save metadata associated with this routing table. */
130 rh->rib_family = family;
131 rh->rib_fibnum = fibnum;
133 rh->rib_vnet = curvnet;
143 /* Init subscription system */
144 rib_init_subscriptions(rh);
146 /* Finally, set base callbacks */
147 rh->rnh_addaddr = rn_addroute;
148 rh->rnh_deladdr = rn_delete;
149 rh->rnh_matchaddr = rn_match;
150 rh->rnh_lookup = rn_lookup;
151 rh->rnh_walktree = rn_walktree;
152 rh->rnh_walktree_from = rn_walktree_from;
158 rt_freeentry(struct radix_node *rn, void *arg)
160 struct radix_head * const rnh = arg;
161 struct radix_node *x;
163 x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh);
170 rt_table_destroy(struct rib_head *rh)
173 tmproutes_destroy(rh);
175 rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head);
177 nhops_destroy_rib(rh);
179 rib_destroy_subscriptions(rh);
181 /* Assume table is already empty */
182 RIB_LOCK_DESTROY(rh);
187 * Adds a temporal redirect entry to the routing table.
188 * @fibnum: fib number
189 * @dst: destination to install redirect to
190 * @gateway: gateway to go via
191 * @author: sockaddr of originating router, can be NULL
192 * @ifp: interface to use for the redirected route
193 * @flags: set of flags to add. Allowed: RTF_GATEWAY
194 * @lifetime_sec: time in seconds to expire this redirect.
196 * Retuns 0 on success, errno otherwise.
199 rib_add_redirect(u_int fibnum, struct sockaddr *dst, struct sockaddr *gateway,
200 struct sockaddr *author, struct ifnet *ifp, int flags, int lifetime_sec)
202 struct rib_cmd_info rc;
204 struct rt_addrinfo info;
205 struct rt_metrics rti_rmx;
210 if (rt_tables_get_rnh(fibnum, dst->sa_family) == NULL)
211 return (EAFNOSUPPORT);
213 /* Verify the allowed flag mask. */
214 KASSERT(((flags & ~(RTF_GATEWAY)) == 0),
215 ("invalid redirect flags: %x", flags));
216 flags |= RTF_HOST | RTF_DYNAMIC;
218 /* Get the best ifa for the given interface and gateway. */
219 if ((ifa = ifaof_ifpforaddr(gateway, ifp)) == NULL)
220 return (ENETUNREACH);
223 bzero(&info, sizeof(info));
224 info.rti_info[RTAX_DST] = dst;
225 info.rti_info[RTAX_GATEWAY] = gateway;
228 info.rti_flags = flags;
230 /* Setup route metrics to define expire time. */
231 bzero(&rti_rmx, sizeof(rti_rmx));
232 /* Set expire time as absolute. */
233 rti_rmx.rmx_expire = lifetime_sec + time_second;
234 info.rti_mflags |= RTV_EXPIRE;
235 info.rti_rmx = &rti_rmx;
237 error = rib_action(fibnum, RTM_ADD, &info, &rc);
241 /* TODO: add per-fib redirect stats. */
245 RTSTAT_INC(rts_dynamic);
247 /* Send notification of a route addition to userland. */
248 bzero(&info, sizeof(info));
249 info.rti_info[RTAX_DST] = dst;
250 info.rti_info[RTAX_GATEWAY] = gateway;
251 info.rti_info[RTAX_AUTHOR] = author;
252 rt_missmsg_fib(RTM_REDIRECT, &info, flags | RTF_UP, error, fibnum);
258 * Routing table ioctl interface.
261 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
265 * If more ioctl commands are added here, make sure the proper
266 * super-user checks are being performed because it is possible for
267 * prison-root to make it this far if raw sockets have been enabled
271 /* Multicast goop, grrr... */
272 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
279 ifa_ifwithroute(int flags, const struct sockaddr *dst,
280 const struct sockaddr *gateway, u_int fibnum)
285 if ((flags & RTF_GATEWAY) == 0) {
287 * If we are adding a route to an interface,
288 * and the interface is a pt to pt link
289 * we should search for the destination
290 * as our clue to the interface. Otherwise
291 * we can use the local address.
294 if (flags & RTF_HOST)
295 ifa = ifa_ifwithdstaddr(dst, fibnum);
297 ifa = ifa_ifwithaddr(gateway);
300 * If we are adding a route to a remote net
301 * or host, the gateway may still be on the
302 * other end of a pt to pt link.
304 ifa = ifa_ifwithdstaddr(gateway, fibnum);
307 ifa = ifa_ifwithnet(gateway, 0, fibnum);
309 struct nhop_object *nh;
311 nh = rib_lookup(fibnum, gateway, NHR_NONE, 0);
314 * dismiss a gateway that is reachable only
315 * through the default router
317 if ((nh == NULL) || (nh->nh_flags & NHF_DEFAULT))
321 if (ifa->ifa_addr->sa_family != dst->sa_family) {
322 struct ifaddr *oifa = ifa;
323 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
333 * Copy most of @rt data into @info.
335 * If @flags contains NHR_COPY, copies dst,netmask and gw to the
336 * pointers specified by @info structure. Assume such pointers
337 * are zeroed sockaddr-like structures with sa_len field initialized
338 * to reflect size of the provided buffer. if no NHR_COPY is specified,
339 * point dst,netmask and gw @info fields to appropriate @rt values.
341 * if @flags contains NHR_REF, do refcouting on rt_ifp and rt_ifa.
343 * Returns 0 on success.
346 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags)
348 struct rt_metrics *rmx;
349 struct sockaddr *src, *dst;
350 struct nhop_object *nh;
354 if (flags & NHR_COPY) {
355 /* Copy destination if dst is non-zero */
357 dst = info->rti_info[RTAX_DST];
358 sa_len = src->sa_len;
360 if (src->sa_len > dst->sa_len)
362 memcpy(dst, src, src->sa_len);
363 info->rti_addrs |= RTA_DST;
366 /* Copy mask if set && dst is non-zero */
368 dst = info->rti_info[RTAX_NETMASK];
369 if (src != NULL && dst != NULL) {
372 * Radix stores different value in sa_len,
373 * assume rt_mask() to have the same length
376 if (sa_len > dst->sa_len)
378 memcpy(dst, src, src->sa_len);
379 info->rti_addrs |= RTA_NETMASK;
382 /* Copy gateway is set && dst is non-zero */
384 dst = info->rti_info[RTAX_GATEWAY];
385 if ((nhop_get_rtflags(nh) & RTF_GATEWAY) &&
386 src != NULL && dst != NULL) {
387 if (src->sa_len > dst->sa_len)
389 memcpy(dst, src, src->sa_len);
390 info->rti_addrs |= RTA_GATEWAY;
393 info->rti_info[RTAX_DST] = rt_key(rt);
394 info->rti_addrs |= RTA_DST;
395 if (rt_mask(rt) != NULL) {
396 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
397 info->rti_addrs |= RTA_NETMASK;
399 if (nhop_get_rtflags(nh) & RTF_GATEWAY) {
400 info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
401 info->rti_addrs |= RTA_GATEWAY;
407 info->rti_mflags |= RTV_MTU;
408 rmx->rmx_mtu = nh->nh_mtu;
411 info->rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
412 info->rti_ifp = nh->nh_ifp;
413 info->rti_ifa = nh->nh_ifa;
414 if (flags & NHR_REF) {
415 if_ref(info->rti_ifp);
416 ifa_ref(info->rti_ifa);
423 * Lookups up route entry for @dst in RIB database for fib @fibnum.
424 * Exports entry data to @info using rt_exportinfo().
426 * If @flags contains NHR_REF, refcouting is performed on rt_ifp and rt_ifa.
427 * All references can be released later by calling rib_free_info().
429 * Returns 0 on success.
430 * Returns ENOENT for lookup failure, ENOMEM for export failure.
433 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
434 uint32_t flowid, struct rt_addrinfo *info)
438 struct radix_node *rn;
442 KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
443 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
448 rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head);
449 if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
451 /* Ensure route & ifp is UP */
452 if (RT_LINK_IS_UP(rt->rt_nhop->nh_ifp)) {
453 flags = (flags & NHR_REF) | NHR_COPY;
454 error = rt_exportinfo(rt, info, flags);
466 * Releases all references acquired by rib_lookup_info() when
467 * called with NHR_REF flags.
470 rib_free_info(struct rt_addrinfo *info)
473 ifa_free(info->rti_ifa);
474 if_rele(info->rti_ifp);
478 * Iterates over all existing fibs in system calling
479 * @setwa_f function prior to traversing each fib.
480 * Calls @wa_f function for each element in current fib.
481 * If af is not AF_UNSPEC, iterates over fibs in particular
485 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f,
488 struct rib_head *rnh;
492 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
493 /* Do we want some specific family? */
494 if (af != AF_UNSPEC) {
495 rnh = rt_tables_get_rnh(fibnum, af);
499 setwa_f(rnh, fibnum, af, arg);
502 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
507 for (i = 1; i <= AF_MAX; i++) {
508 rnh = rt_tables_get_rnh(fibnum, i);
512 setwa_f(rnh, fibnum, i, arg);
515 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
522 * Iterates over all existing fibs in system and deletes each element
523 * for which @filter_f function returns non-zero value.
524 * If @family is not AF_UNSPEC, iterates over fibs in particular
528 rt_foreach_fib_walk_del(int family, rt_filter_f_t *filter_f, void *arg)
533 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
534 /* Do we want some specific family? */
535 if (family != AF_UNSPEC) {
543 for (i = start; i <= end; i++) {
544 if (rt_tables_get_rnh(fibnum, i) == NULL)
547 rib_walk_del(fibnum, i, filter_f, arg, 0);
553 * Delete Routes for a Network Interface
555 * Called for each routing entry via the rnh->rnh_walktree() call above
556 * to delete all route entries referencing a detaching network interface.
559 * rt pointer to rtentry
561 * arg argument passed to rnh->rnh_walktree() - detaching interface
565 * errno failed - reason indicated
568 rt_ifdelroute(const struct rtentry *rt, const struct nhop_object *nh, void *arg)
570 struct ifnet *ifp = arg;
572 if (nh->nh_ifp != ifp)
576 * Protect (sorta) against walktree recursion problems
579 if ((rt->rte_flags & RTF_UP) == 0)
586 * Delete all remaining routes using this interface
587 * Unfortuneatly the only way to do this is to slog through
588 * the entire routing table looking for routes which point
589 * to this interface...oh well...
592 rt_flushifroutes_af(struct ifnet *ifp, int af)
594 KASSERT((af >= 1 && af <= AF_MAX), ("%s: af %d not >= 1 and <= %d",
595 __func__, af, AF_MAX));
597 rt_foreach_fib_walk_del(af, rt_ifdelroute, ifp);
601 rt_flushifroutes(struct ifnet *ifp)
604 rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
608 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined,
609 * it will be referenced so the caller must free it.
611 * Assume basic consistency checks are executed by callers:
612 * RTAX_DST exists, if RTF_GATEWAY is set, RTAX_GATEWAY exists as well.
615 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
617 const struct sockaddr *dst, *gateway, *ifpaddr, *ifaaddr;
618 struct epoch_tracker et;
619 int needref, error, flags;
621 dst = info->rti_info[RTAX_DST];
622 gateway = info->rti_info[RTAX_GATEWAY];
623 ifpaddr = info->rti_info[RTAX_IFP];
624 ifaaddr = info->rti_info[RTAX_IFA];
625 flags = info->rti_flags;
628 * ifp may be specified by sockaddr_dl
629 * when protocol address is ambiguous.
632 needref = (info->rti_ifa == NULL);
635 /* If we have interface specified by the ifindex in the address, use it */
636 if (info->rti_ifp == NULL && ifpaddr != NULL &&
637 ifpaddr->sa_family == AF_LINK) {
638 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)ifpaddr;
639 if (sdl->sdl_index != 0)
640 info->rti_ifp = ifnet_byindex(sdl->sdl_index);
643 * If we have source address specified, try to find it
644 * TODO: avoid enumerating all ifas on all interfaces.
646 if (info->rti_ifa == NULL && ifaaddr != NULL)
647 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
648 if (info->rti_ifa == NULL) {
649 const struct sockaddr *sa;
652 * Most common use case for the userland-supplied routes.
654 * Choose sockaddr to select ifa.
655 * -- if ifp is set --
656 * Order of preference:
659 * Note: for interface routes link-level gateway address
660 * is specified to indicate the interface index without
661 * specifying RTF_GATEWAY. In this case, ignore gateway
662 * Note: gateway AF may be different from dst AF. In this case,
664 * 3) final destination.
665 * 4) if all of these fails, try to get at least link-level ifa.
667 * try to lookup gateway or dst in the routing table to get ifa
669 if (info->rti_info[RTAX_IFA] != NULL)
670 sa = info->rti_info[RTAX_IFA];
671 else if ((info->rti_flags & RTF_GATEWAY) != 0 &&
672 gateway->sa_family == dst->sa_family)
676 if (info->rti_ifp != NULL) {
677 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
679 if (info->rti_ifa == NULL && gateway != NULL)
680 info->rti_ifa = ifaof_ifpforaddr(gateway, info->rti_ifp);
681 } else if (dst != NULL && gateway != NULL)
682 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway,
685 info->rti_ifa = ifa_ifwithroute(flags, sa, sa,
688 if (needref && info->rti_ifa != NULL) {
689 if (info->rti_ifp == NULL)
690 info->rti_ifp = info->rti_ifa->ifa_ifp;
691 ifa_ref(info->rti_ifa);
699 rt_updatemtu(struct ifnet *ifp)
701 struct rib_head *rnh;
706 * Try to update rt_mtu for all routes using this interface
707 * Unfortunately the only way to do this is to traverse all
708 * routing tables in all fibs/domains.
710 for (i = 1; i <= AF_MAX; i++) {
711 mtu = if_getmtu_family(ifp, i);
712 for (j = 0; j < rt_numfibs; j++) {
713 rnh = rt_tables_get_rnh(j, i);
716 nhops_update_ifmtu(rnh, ifp, mtu);
723 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
724 int rt_print(char *buf, int buflen, struct rtentry *rt);
727 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
731 switch (s->sa_family) {
733 paddr = &((struct sockaddr_in *)s)->sin_addr;
736 paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
743 if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
746 return (strlen(buf));
750 rt_print(char *buf, int buflen, struct rtentry *rt)
752 struct sockaddr *addr, *mask;
758 i = p_sockaddr(buf, buflen, addr);
759 if (!(rt->rt_flags & RTF_HOST)) {
761 i += p_sockaddr(buf + i, buflen - i, mask);
764 if (rt->rt_flags & RTF_GATEWAY) {
766 i += p_sockaddr(buf + i, buflen - i, &rt->rt_nhop->gw_sa);
775 * Deletes key for single-path routes, unlinks rtentry with
776 * gateway specified in @info from multi-path routes.
778 * Returnes unlinked entry. In case of failure, returns NULL
779 * and sets @perror to ESRCH.
782 rt_mpath_unlink(struct rib_head *rnh, struct rt_addrinfo *info,
783 struct rtentry *rto, int *perror)
786 * if we got multipath routes, we require users to specify
787 * a matching RTAX_GATEWAY.
789 struct rtentry *rt; // *rto = NULL;
790 struct radix_node *rn;
793 gw = info->rti_info[RTAX_GATEWAY];
794 rt = rt_mpath_matchgate(rto, gw);
801 * this is the first entry in the chain
804 rn = rn_mpath_next((struct radix_node *)rt);
806 * there is another entry, now it's active
810 rto->rte_flags |= RTF_UP;
811 } else if (rt->rte_flags & RTF_GATEWAY) {
813 * For gateway routes, we need to
814 * make sure that we we are deleting
815 * the correct gateway.
816 * rt_mpath_matchgate() does not
817 * check the case when there is only
818 * one route in the chain.
821 (rt->rt_nhop->gw_sa.sa_len != gw->sa_len ||
822 memcmp(&rt->rt_nhop->gw_sa, gw, gw->sa_len))) {
829 * use the normal delete code to remove
832 rn = rnh->rnh_deladdr(info->rti_info[RTAX_DST],
833 info->rti_info[RTAX_NETMASK],
844 * if the entry is 2nd and on up
846 if (rt_mpath_deldup(rto, rt) == 0)
847 panic ("rtrequest1: rt_mpath_deldup");
849 rn = (struct radix_node *)rt;
855 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt)
858 if (info->rti_mflags & RTV_WEIGHT)
859 rt->rt_weight = info->rti_rmx->rmx_weight;
860 /* Kernel -> userland timebase conversion. */
861 if (info->rti_mflags & RTV_EXPIRE)
862 rt->rt_expire = info->rti_rmx->rmx_expire ?
863 info->rti_rmx->rmx_expire - time_second + time_uptime : 0;
867 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
869 u_char *cp1 = (u_char *)src;
870 u_char *cp2 = (u_char *)dst;
871 u_char *cp3 = (u_char *)netmask;
872 u_char *cplim = cp2 + *cp3;
873 u_char *cplim2 = cp2 + *cp1;
875 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
880 *cp2++ = *cp1++ & *cp3++;
882 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
886 * Set up a routing table entry, normally
889 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
891 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
894 struct epoch_tracker et;
895 struct sockaddr *dst;
896 struct sockaddr *netmask;
897 struct rib_cmd_info rc;
898 struct rt_addrinfo info;
900 int startfib, endfib;
901 char tempbuf[_SOCKADDR_TMPSIZE];
904 struct sockaddr_dl_short *sdl = NULL;
905 struct rib_head *rnh;
907 if (flags & RTF_HOST) {
908 dst = ifa->ifa_dstaddr;
912 netmask = ifa->ifa_netmask;
914 if (dst->sa_len == 0)
916 switch (dst->sa_family) {
919 /* We support multiple FIBs. */
922 fibnum = RT_DEFAULT_FIB;
925 if (fibnum == RT_ALL_FIBS) {
926 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD)
927 startfib = endfib = ifa->ifa_ifp->if_fib;
930 endfib = rt_numfibs - 1;
933 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
939 * If it's a delete, check that if it exists,
940 * it's on the correct interface or we might scrub
941 * a route to another ifa which would
942 * be confusing at best and possibly worse.
944 if (cmd == RTM_DELETE) {
946 * It's a delete, so it should already exist..
947 * If it's a net, mask off the host bits
948 * (Assuming we have a mask)
949 * XXX this is kinda inet specific..
951 if (netmask != NULL) {
952 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
953 dst = (struct sockaddr *)tempbuf;
955 } else if (cmd == RTM_ADD) {
956 sdl = (struct sockaddr_dl_short *)tempbuf;
957 bzero(sdl, sizeof(struct sockaddr_dl_short));
958 sdl->sdl_family = AF_LINK;
959 sdl->sdl_len = sizeof(struct sockaddr_dl_short);
960 sdl->sdl_type = ifa->ifa_ifp->if_type;
961 sdl->sdl_index = ifa->ifa_ifp->if_index;
964 * Now go through all the requested tables (fibs) and do the
965 * requested action. Realistically, this will either be fib 0
966 * for protocols that don't do multiple tables or all the
967 * tables for those that do.
969 for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
970 if (cmd == RTM_DELETE) {
971 struct radix_node *rn;
973 * Look up an rtentry that is in the routing tree and
974 * contains the correct info.
976 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
978 /* this table doesn't exist but others might */
981 rn = rnh->rnh_lookup(dst, netmask, &rnh->head);
983 if (rt_mpath_capable(rnh)) {
988 struct rtentry *rt = RNTORT(rn);
990 * for interface route the gateway
991 * gateway is sockaddr_dl, so
992 * rt_mpath_matchgate must use the
995 rt = rt_mpath_matchgate(rt,
1002 error = (rn == NULL ||
1003 (rn->rn_flags & RNF_ROOT) ||
1004 RNTORT(rn)->rt_nhop->nh_ifa != ifa);
1007 /* this is only an error if bad on ALL tables */
1012 * Do the actual request
1014 bzero((caddr_t)&info, sizeof(info));
1016 info.rti_flags = flags |
1017 (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED;
1018 info.rti_info[RTAX_DST] = dst;
1020 * doing this for compatibility reasons
1023 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)sdl;
1025 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
1026 info.rti_info[RTAX_NETMASK] = netmask;
1027 NET_EPOCH_ENTER(et);
1028 error = rib_action(fibnum, cmd, &info, &rc);
1029 if (error == 0 && rc.rc_rt != NULL) {
1031 * notify any listening routing agents of the change
1034 /* TODO: interface routes/aliases */
1035 rt_newaddrmsg_fib(cmd, ifa, rc.rc_rt, fibnum);
1042 if (cmd == RTM_DELETE) {
1046 /* we only give an error if it wasn't in any table */
1047 error = ((flags & RTF_HOST) ?
1048 EHOSTUNREACH : ENETUNREACH);
1052 /* return an error if any of them failed */
1060 * Set up a routing table entry, normally
1064 rtinit(struct ifaddr *ifa, int cmd, int flags)
1066 struct sockaddr *dst;
1067 int fib = RT_DEFAULT_FIB;
1069 if (flags & RTF_HOST) {
1070 dst = ifa->ifa_dstaddr;
1072 dst = ifa->ifa_addr;
1075 switch (dst->sa_family) {
1078 /* We do support multiple FIBs. */
1082 return (rtinit1(ifa, cmd, flags, fib));
1086 * Announce interface address arrival/withdraw
1087 * Returns 0 on success.
1090 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1093 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1094 ("unexpected cmd %d", cmd));
1095 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
1096 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
1098 EVENTHANDLER_DIRECT_INVOKE(rt_addrmsg, ifa, cmd);
1099 return (rtsock_addrmsg(cmd, ifa, fibnum));
1103 * Announce kernel-originated route addition/removal to rtsock based on @rt data.
1105 * @rt: valid rtentry
1106 * @ifp: target route interface
1107 * @fibnum: fib id or RT_ALL_FIBS
1109 * Returns 0 on success.
1112 rt_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs,
1116 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1117 ("unexpected cmd %d", cmd));
1119 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
1120 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
1122 KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
1124 return (rtsock_routemsg(cmd, rt, ifp, 0, fibnum));
1128 * Announce kernel-originated route addition/removal to rtsock based on @rt data.
1130 * @info: addrinfo structure with valid data.
1131 * @fibnum: fib id or RT_ALL_FIBS
1133 * Returns 0 on success.
1136 rt_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1139 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE || cmd == RTM_CHANGE,
1140 ("unexpected cmd %d", cmd));
1142 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
1143 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
1145 KASSERT(info->rti_info[RTAX_DST] != NULL, (":%s: RTAX_DST must be supplied", __func__));
1147 return (rtsock_routemsg_info(cmd, info, fibnum));
1152 * This is called to generate messages from the routing socket
1153 * indicating a network interface has had addresses associated with it.
1156 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, struct rtentry *rt, int fibnum)
1159 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1160 ("unexpected cmd %u", cmd));
1161 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
1162 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
1164 if (cmd == RTM_ADD) {
1165 rt_addrmsg(cmd, ifa, fibnum);
1167 rt_routemsg(cmd, rt, ifa->ifa_ifp, 0, fibnum);
1170 rt_routemsg(cmd, rt, ifa->ifa_ifp, 0, fibnum);
1171 rt_addrmsg(cmd, ifa, fibnum);