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_PCPUSTAT_DEFINE_STATIC(struct rtstat, rtstat);
112 #define RTSTAT_ADD(name, val) \
113 VNET_PCPUSTAT_ADD(struct rtstat, rtstat, name, (val))
114 #define RTSTAT_INC(name) RTSTAT_ADD(name, 1)
116 VNET_PCPUSTAT_SYSINIT(rtstat);
118 VNET_PCPUSTAT_SYSUNINIT(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 EVENTHANDLER_LIST_DEFINE(rt_addrmsg);
145 static int rt_getifa_fib(struct rt_addrinfo *, u_int);
146 static int rtrequest1_fib_change(struct rib_head *, struct rt_addrinfo *,
147 struct rtentry **, u_int);
148 static void rt_setmetrics(const struct rt_addrinfo *, struct rtentry *);
149 static int rt_ifdelroute(const struct rtentry *rt, void *arg);
150 static struct rtentry *rt_unlinkrte(struct rib_head *rnh,
151 struct rt_addrinfo *info, int *perror);
152 static void rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info);
154 static struct radix_node *rt_mpath_unlink(struct rib_head *rnh,
155 struct rt_addrinfo *info, struct rtentry *rto, int *perror);
157 static int rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info,
166 static int if_updatemtu_cb(struct radix_node *, void *);
169 * handler for net.my_fibnum
172 sysctl_my_fibnum(SYSCTL_HANDLER_ARGS)
177 fibnum = curthread->td_proc->p_fibnum;
178 error = sysctl_handle_int(oidp, &fibnum, 0, req);
182 SYSCTL_PROC(_net, OID_AUTO, my_fibnum, CTLTYPE_INT|CTLFLAG_RD,
183 NULL, 0, &sysctl_my_fibnum, "I", "default FIB of caller");
185 static __inline struct rib_head **
186 rt_tables_get_rnh_ptr(int table, int fam)
188 struct rib_head **rnh;
190 KASSERT(table >= 0 && table < rt_numfibs, ("%s: table out of bounds.",
192 KASSERT(fam >= 0 && fam < (AF_MAX+1), ("%s: fam out of bounds.",
195 /* rnh is [fib=0][af=0]. */
196 rnh = (struct rib_head **)V_rt_tables;
197 /* Get the offset to the requested table and fam. */
198 rnh += table * (AF_MAX+1) + fam;
204 rt_tables_get_rnh(int table, int fam)
207 return (*rt_tables_get_rnh_ptr(table, fam));
211 rt_tables_get_gen(int table, int fam)
213 struct rib_head *rnh;
215 rnh = *rt_tables_get_rnh_ptr(table, fam);
216 KASSERT(rnh != NULL, ("%s: NULL rib_head pointer table %d fam %d",
217 __func__, table, fam));
218 return (rnh->rnh_gen);
223 * route initialization must occur before ip6_init2(), which happenas at
230 /* whack the tunable ints into line. */
231 if (rt_numfibs > RT_MAXFIBS)
232 rt_numfibs = RT_MAXFIBS;
236 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, NULL);
239 rtentry_zinit(void *mem, int size, int how)
241 struct rtentry *rt = mem;
243 rt->rt_pksent = counter_u64_alloc(how);
244 if (rt->rt_pksent == NULL)
253 rtentry_zfini(void *mem, int size)
255 struct rtentry *rt = mem;
258 counter_u64_free(rt->rt_pksent);
262 rtentry_ctor(void *mem, int size, void *arg, int how)
264 struct rtentry *rt = mem;
266 bzero(rt, offsetof(struct rtentry, rt_endzero));
267 counter_u64_zero(rt->rt_pksent);
274 rtentry_dtor(void *mem, int size, void *arg)
276 struct rtentry *rt = mem;
282 vnet_route_init(const void *unused __unused)
285 struct rib_head **rnh;
289 V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) *
290 sizeof(struct rib_head *), M_RTABLE, M_WAITOK|M_ZERO);
292 V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry),
293 rtentry_ctor, rtentry_dtor,
294 rtentry_zinit, rtentry_zfini, UMA_ALIGN_PTR, 0);
295 for (dom = domains; dom; dom = dom->dom_next) {
296 if (dom->dom_rtattach == NULL)
299 for (table = 0; table < rt_numfibs; table++) {
300 fam = dom->dom_family;
301 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
304 rnh = rt_tables_get_rnh_ptr(table, fam);
306 panic("%s: rnh NULL", __func__);
307 dom->dom_rtattach((void **)rnh, 0);
311 VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
316 vnet_route_uninit(const void *unused __unused)
321 struct rib_head **rnh;
323 for (dom = domains; dom; dom = dom->dom_next) {
324 if (dom->dom_rtdetach == NULL)
327 for (table = 0; table < rt_numfibs; table++) {
328 fam = dom->dom_family;
330 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
333 rnh = rt_tables_get_rnh_ptr(table, fam);
335 panic("%s: rnh NULL", __func__);
336 dom->dom_rtdetach((void **)rnh, 0);
340 free(V_rt_tables, M_RTABLE);
341 uma_zdestroy(V_rtzone);
343 VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST,
344 vnet_route_uninit, 0);
348 rt_table_init(int offset)
352 rh = malloc(sizeof(struct rib_head), M_RTABLE, M_WAITOK | M_ZERO);
354 /* TODO: These details should be hidded inside radix.c */
355 /* Init masks tree */
356 rn_inithead_internal(&rh->head, rh->rnh_nodes, offset);
357 rn_inithead_internal(&rh->rmhead.head, rh->rmhead.mask_nodes, 0);
358 rh->head.rnh_masks = &rh->rmhead;
363 /* Finally, set base callbacks */
364 rh->rnh_addaddr = rn_addroute;
365 rh->rnh_deladdr = rn_delete;
366 rh->rnh_matchaddr = rn_match;
367 rh->rnh_lookup = rn_lookup;
368 rh->rnh_walktree = rn_walktree;
369 rh->rnh_walktree_from = rn_walktree_from;
375 rt_freeentry(struct radix_node *rn, void *arg)
377 struct radix_head * const rnh = arg;
378 struct radix_node *x;
380 x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh);
387 rt_table_destroy(struct rib_head *rh)
390 rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head);
392 /* Assume table is already empty */
393 RIB_LOCK_DESTROY(rh);
398 #ifndef _SYS_SYSPROTO_H_
404 sys_setfib(struct thread *td, struct setfib_args *uap)
406 if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs)
408 td->td_proc->p_fibnum = uap->fibnum;
413 * Packet routing routines.
416 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum)
420 if ((rt = ro->ro_rt) != NULL) {
421 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
426 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum);
428 RT_UNLOCK(ro->ro_rt);
432 * Look up the route that matches the address given
433 * Or, at least try.. Create a cloned route if needed.
435 * The returned route, if any, is locked.
438 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
441 return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB));
445 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags,
450 struct radix_node *rn;
451 struct rtentry *newrt;
452 struct rt_addrinfo info;
453 int err = 0, msgtype = RTM_MISS;
455 KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum"));
456 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
462 * Look up the address in the table for that Address Family
464 if ((ignflags & RTF_RNH_LOCKED) == 0)
470 rn = rh->rnh_matchaddr(dst, &rh->head);
471 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
475 if ((ignflags & RTF_RNH_LOCKED) == 0)
479 } else if ((ignflags & RTF_RNH_LOCKED) == 0)
482 * Either we hit the root or could not find any match,
483 * which basically means: "cannot get there from here".
486 RTSTAT_INC(rts_unreach);
490 * If required, report the failure to the supervising
492 * For a delete, this is not an error. (report == 0)
494 bzero(&info, sizeof(info));
495 info.rti_info[RTAX_DST] = dst;
496 rt_missmsg_fib(msgtype, &info, 0, err, fibnum);
502 * Remove a reference count from an rtentry.
503 * If the count gets low enough, take it out of the routing table
506 rtfree(struct rtentry *rt)
508 struct rib_head *rnh;
510 KASSERT(rt != NULL,("%s: NULL rt", __func__));
511 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
512 KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
517 * The callers should use RTFREE_LOCKED() or RTFREE(), so
518 * we should come here exactly with the last reference.
521 if (rt->rt_refcnt > 0) {
522 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt);
527 * On last reference give the "close method" a chance
528 * to cleanup private state. This also permits (for
529 * IPv4 and IPv6) a chance to decide if the routing table
530 * entry should be purged immediately or at a later time.
531 * When an immediate purge is to happen the close routine
532 * typically calls rtexpunge which clears the RTF_UP flag
533 * on the entry so that the code below reclaims the storage.
535 if (rt->rt_refcnt == 0 && rnh->rnh_close)
536 rnh->rnh_close((struct radix_node *)rt, &rnh->head);
539 * If we are no longer "up" (and ref == 0)
540 * then we can free the resources associated
543 if ((rt->rt_flags & RTF_UP) == 0) {
544 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
547 * the rtentry must have been removed from the routing table
548 * so it is represented in rttrash.. remove that now.
552 if (rt->rt_refcnt < 0) {
553 printf("rtfree: %p not freed (neg refs)\n", rt);
558 * release references on items we hold them on..
559 * e.g other routes and ifaddrs.
562 ifa_free(rt->rt_ifa);
564 * The key is separatly alloc'd so free it (see rt_setgate()).
565 * This also frees the gateway, as they are always malloc'd
571 * and the rtentry itself of course
573 uma_zfree(V_rtzone, rt);
582 * Force a routing table entry to the specified
583 * destination to go through the given gateway.
584 * Normally called as a result of a routing redirect
585 * message from the network layer.
588 rtredirect_fib(struct sockaddr *dst,
589 struct sockaddr *gateway,
590 struct sockaddr *netmask,
592 struct sockaddr *src,
597 struct rt_addrinfo info;
599 struct rib_head *rnh;
604 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
606 error = EAFNOSUPPORT;
609 /* verify the gateway is directly reachable */
610 if ((ifa = ifa_ifwithnet(gateway, 0, fibnum)) == NULL) {
614 rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */
616 * If the redirect isn't from our current router for this dst,
617 * it's either old or wrong. If it redirects us to ourselves,
618 * we have a routing loop, perhaps as a result of an interface
619 * going down recently.
621 if (!(flags & RTF_DONE) && rt) {
622 if (!sa_equal(src, rt->rt_gateway)) {
626 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) {
631 if ((flags & RTF_GATEWAY) && ifa_ifwithaddr_check(gateway)) {
632 error = EHOSTUNREACH;
636 * Create a new entry if we just got back a wildcard entry
637 * or the lookup failed. This is necessary for hosts
638 * which use routing redirects generated by smart gateways
639 * to dynamically build the routing tables.
641 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
644 * Don't listen to the redirect if it's
645 * for a route to an interface.
647 if (rt->rt_flags & RTF_GATEWAY) {
648 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
650 * Changing from route to net => route to host.
651 * Create new route, rather than smashing route to net.
657 flags |= RTF_DYNAMIC;
658 bzero((caddr_t)&info, sizeof(info));
659 info.rti_info[RTAX_DST] = dst;
660 info.rti_info[RTAX_GATEWAY] = gateway;
661 info.rti_info[RTAX_NETMASK] = netmask;
664 info.rti_flags = flags;
665 error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum);
668 flags = rt->rt_flags;
671 RTSTAT_INC(rts_dynamic);
675 * Smash the current notion of the gateway to
676 * this destination. Should check about netmask!!!
678 if ((flags & RTF_GATEWAY) == 0)
679 rt->rt_flags &= ~RTF_GATEWAY;
680 rt->rt_flags |= RTF_MODIFIED;
681 flags |= RTF_MODIFIED;
682 RTSTAT_INC(rts_newgateway);
684 * add the key and gateway (in one malloc'd chunk).
689 rt_setgate(rt, rt_key(rt), gateway);
693 error = EHOSTUNREACH;
699 RTSTAT_INC(rts_badredirect);
700 bzero((caddr_t)&info, sizeof(info));
701 info.rti_info[RTAX_DST] = dst;
702 info.rti_info[RTAX_GATEWAY] = gateway;
703 info.rti_info[RTAX_NETMASK] = netmask;
704 info.rti_info[RTAX_AUTHOR] = src;
705 rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum);
709 * Routing table ioctl interface.
712 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
716 * If more ioctl commands are added here, make sure the proper
717 * super-user checks are being performed because it is possible for
718 * prison-root to make it this far if raw sockets have been enabled
722 /* Multicast goop, grrr... */
723 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
730 ifa_ifwithroute(int flags, const struct sockaddr *dst, struct sockaddr *gateway,
736 MPASS(in_epoch(net_epoch_preempt));
737 if ((flags & RTF_GATEWAY) == 0) {
739 * If we are adding a route to an interface,
740 * and the interface is a pt to pt link
741 * we should search for the destination
742 * as our clue to the interface. Otherwise
743 * we can use the local address.
746 if (flags & RTF_HOST)
747 ifa = ifa_ifwithdstaddr(dst, fibnum);
749 ifa = ifa_ifwithaddr(gateway);
752 * If we are adding a route to a remote net
753 * or host, the gateway may still be on the
754 * other end of a pt to pt link.
756 ifa = ifa_ifwithdstaddr(gateway, fibnum);
759 ifa = ifa_ifwithnet(gateway, 0, fibnum);
763 rt = rtalloc1_fib(gateway, 0, flags, fibnum);
767 * dismiss a gateway that is reachable only
768 * through the default router
770 switch (gateway->sa_family) {
772 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
776 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
782 if (!not_found && rt->rt_ifa != NULL) {
787 if (not_found || ifa == NULL)
790 if (ifa->ifa_addr->sa_family != dst->sa_family) {
791 struct ifaddr *oifa = ifa;
792 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
801 * Do appropriate manipulations of a routing tree given
802 * all the bits of info needed
805 rtrequest_fib(int req,
806 struct sockaddr *dst,
807 struct sockaddr *gateway,
808 struct sockaddr *netmask,
810 struct rtentry **ret_nrt,
813 struct rt_addrinfo info;
815 if (dst->sa_len == 0)
818 bzero((caddr_t)&info, sizeof(info));
819 info.rti_flags = flags;
820 info.rti_info[RTAX_DST] = dst;
821 info.rti_info[RTAX_GATEWAY] = gateway;
822 info.rti_info[RTAX_NETMASK] = netmask;
823 return rtrequest1_fib(req, &info, ret_nrt, fibnum);
828 * Copy most of @rt data into @info.
830 * If @flags contains NHR_COPY, copies dst,netmask and gw to the
831 * pointers specified by @info structure. Assume such pointers
832 * are zeroed sockaddr-like structures with sa_len field initialized
833 * to reflect size of the provided buffer. if no NHR_COPY is specified,
834 * point dst,netmask and gw @info fields to appropriate @rt values.
836 * if @flags contains NHR_REF, do refcouting on rt_ifp.
838 * Returns 0 on success.
841 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags)
843 struct rt_metrics *rmx;
844 struct sockaddr *src, *dst;
847 if (flags & NHR_COPY) {
848 /* Copy destination if dst is non-zero */
850 dst = info->rti_info[RTAX_DST];
851 sa_len = src->sa_len;
853 if (src->sa_len > dst->sa_len)
855 memcpy(dst, src, src->sa_len);
856 info->rti_addrs |= RTA_DST;
859 /* Copy mask if set && dst is non-zero */
861 dst = info->rti_info[RTAX_NETMASK];
862 if (src != NULL && dst != NULL) {
865 * Radix stores different value in sa_len,
866 * assume rt_mask() to have the same length
869 if (sa_len > dst->sa_len)
871 memcpy(dst, src, src->sa_len);
872 info->rti_addrs |= RTA_NETMASK;
875 /* Copy gateway is set && dst is non-zero */
876 src = rt->rt_gateway;
877 dst = info->rti_info[RTAX_GATEWAY];
878 if ((rt->rt_flags & RTF_GATEWAY) && src != NULL && dst != NULL){
879 if (src->sa_len > dst->sa_len)
881 memcpy(dst, src, src->sa_len);
882 info->rti_addrs |= RTA_GATEWAY;
885 info->rti_info[RTAX_DST] = rt_key(rt);
886 info->rti_addrs |= RTA_DST;
887 if (rt_mask(rt) != NULL) {
888 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
889 info->rti_addrs |= RTA_NETMASK;
891 if (rt->rt_flags & RTF_GATEWAY) {
892 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
893 info->rti_addrs |= RTA_GATEWAY;
899 info->rti_mflags |= RTV_MTU;
900 rmx->rmx_mtu = rt->rt_mtu;
903 info->rti_flags = rt->rt_flags;
904 info->rti_ifp = rt->rt_ifp;
905 info->rti_ifa = rt->rt_ifa;
906 ifa_ref(info->rti_ifa);
907 if (flags & NHR_REF) {
908 /* Do 'traditional' refcouting */
909 if_ref(info->rti_ifp);
916 * Lookups up route entry for @dst in RIB database for fib @fibnum.
917 * Exports entry data to @info using rt_exportinfo().
919 * if @flags contains NHR_REF, refcouting is performed on rt_ifp.
920 * All references can be released later by calling rib_free_info()
922 * Returns 0 on success.
923 * Returns ENOENT for lookup failure, ENOMEM for export failure.
926 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
927 uint32_t flowid, struct rt_addrinfo *info)
931 struct radix_node *rn;
935 KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
936 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
941 rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head);
942 if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
944 /* Ensure route & ifp is UP */
945 if (RT_LINK_IS_UP(rt->rt_ifp)) {
946 flags = (flags & NHR_REF) | NHR_COPY;
947 error = rt_exportinfo(rt, info, flags);
959 * Releases all references acquired by rib_lookup_info() when
960 * called with NHR_REF flags.
963 rib_free_info(struct rt_addrinfo *info)
966 if_rele(info->rti_ifp);
970 * Iterates over all existing fibs in system calling
971 * @setwa_f function prior to traversing each fib.
972 * Calls @wa_f function for each element in current fib.
973 * If af is not AF_UNSPEC, iterates over fibs in particular
977 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f,
980 struct rib_head *rnh;
984 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
985 /* Do we want some specific family? */
986 if (af != AF_UNSPEC) {
987 rnh = rt_tables_get_rnh(fibnum, af);
991 setwa_f(rnh, fibnum, af, arg);
994 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
999 for (i = 1; i <= AF_MAX; i++) {
1000 rnh = rt_tables_get_rnh(fibnum, i);
1003 if (setwa_f != NULL)
1004 setwa_f(rnh, fibnum, i, arg);
1007 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
1015 struct rt_addrinfo info;
1016 struct rib_head *rnh;
1017 struct rtentry *head;
1021 * Conditionally unlinks @rn from radix tree based
1022 * on info data passed in @arg.
1025 rt_checkdelroute(struct radix_node *rn, void *arg)
1027 struct rt_delinfo *di;
1028 struct rt_addrinfo *info;
1032 di = (struct rt_delinfo *)arg;
1033 rt = (struct rtentry *)rn;
1037 info->rti_info[RTAX_DST] = rt_key(rt);
1038 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
1039 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1041 rt = rt_unlinkrte(di->rnh, info, &error);
1043 /* Either not allowed or not matched. Skip entry */
1047 /* Entry was unlinked. Add to the list and return */
1048 rt->rt_chain = di->head;
1055 * Iterates over all existing fibs in system.
1056 * Deletes each element for which @filter_f function returned
1058 * If @af is not AF_UNSPEC, iterates over fibs in particular
1062 rt_foreach_fib_walk_del(int af, rt_filter_f_t *filter_f, void *arg)
1064 struct rib_head *rnh;
1065 struct rt_delinfo di;
1070 bzero(&di, sizeof(di));
1071 di.info.rti_filter = filter_f;
1072 di.info.rti_filterdata = arg;
1074 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1075 /* Do we want some specific family? */
1076 if (af != AF_UNSPEC) {
1084 for (i = start; i <= end; i++) {
1085 rnh = rt_tables_get_rnh(fibnum, i);
1091 rnh->rnh_walktree(&rnh->head, rt_checkdelroute, &di);
1094 if (di.head == NULL)
1097 /* We might have something to reclaim */
1098 while (di.head != NULL) {
1100 di.head = rt->rt_chain;
1101 rt->rt_chain = NULL;
1103 /* TODO std rt -> rt_addrinfo export */
1104 di.info.rti_info[RTAX_DST] = rt_key(rt);
1105 di.info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1107 rt_notifydelete(rt, &di.info);
1116 * Delete Routes for a Network Interface
1118 * Called for each routing entry via the rnh->rnh_walktree() call above
1119 * to delete all route entries referencing a detaching network interface.
1122 * rt pointer to rtentry
1123 * arg argument passed to rnh->rnh_walktree() - detaching interface
1127 * errno failed - reason indicated
1130 rt_ifdelroute(const struct rtentry *rt, void *arg)
1132 struct ifnet *ifp = arg;
1134 if (rt->rt_ifp != ifp)
1138 * Protect (sorta) against walktree recursion problems
1139 * with cloned routes
1141 if ((rt->rt_flags & RTF_UP) == 0)
1148 * Delete all remaining routes using this interface
1149 * Unfortuneatly the only way to do this is to slog through
1150 * the entire routing table looking for routes which point
1151 * to this interface...oh well...
1154 rt_flushifroutes_af(struct ifnet *ifp, int af)
1156 KASSERT((af >= 1 && af <= AF_MAX), ("%s: af %d not >= 1 and <= %d",
1157 __func__, af, AF_MAX));
1159 rt_foreach_fib_walk_del(af, rt_ifdelroute, ifp);
1163 rt_flushifroutes(struct ifnet *ifp)
1166 rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
1170 * Conditionally unlinks rtentry matching data inside @info from @rnh.
1171 * Returns unlinked, locked and referenced @rtentry on success,
1172 * Returns NULL and sets @perror to:
1173 * ESRCH - if prefix was not found,
1174 * EADDRINUSE - if trying to delete PINNED route without appropriate flag.
1175 * ENOENT - if supplied filter function returned 0 (not matched).
1177 static struct rtentry *
1178 rt_unlinkrte(struct rib_head *rnh, struct rt_addrinfo *info, int *perror)
1180 struct sockaddr *dst, *netmask;
1182 struct radix_node *rn;
1184 dst = info->rti_info[RTAX_DST];
1185 netmask = info->rti_info[RTAX_NETMASK];
1187 rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, &rnh->head);
1193 if ((info->rti_flags & RTF_PINNED) == 0) {
1194 /* Check if target route can be deleted */
1195 if (rt->rt_flags & RTF_PINNED) {
1196 *perror = EADDRINUSE;
1201 if (info->rti_filter != NULL) {
1202 if (info->rti_filter(rt, info->rti_filterdata) == 0) {
1209 * Filter function requested rte deletion.
1210 * Ease the caller work by filling in remaining info
1211 * from that particular entry.
1213 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1217 * Remove the item from the tree and return it.
1218 * Complain if it is not there and do no more processing.
1222 if (rt_mpath_capable(rnh))
1223 rn = rt_mpath_unlink(rnh, info, rt, perror);
1226 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1230 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
1231 panic ("rtrequest delete");
1236 rt->rt_flags &= ~RTF_UP;
1244 rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info)
1249 * give the protocol a chance to keep things in sync.
1252 if (ifa != NULL && ifa->ifa_rtrequest != NULL)
1253 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1256 * One more rtentry floating around that is not
1257 * linked to the routing table. rttrash will be decremented
1258 * when RTFREE(rt) is eventually called.
1265 * These (questionable) definitions of apparent local variables apply
1266 * to the next two functions. XXXXXX!!!
1268 #define dst info->rti_info[RTAX_DST]
1269 #define gateway info->rti_info[RTAX_GATEWAY]
1270 #define netmask info->rti_info[RTAX_NETMASK]
1271 #define ifaaddr info->rti_info[RTAX_IFA]
1272 #define ifpaddr info->rti_info[RTAX_IFP]
1273 #define flags info->rti_flags
1276 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined,
1277 * it will be referenced so the caller must free it.
1279 * Assume basic consistency checks are executed by callers:
1280 * RTAX_DST exists, if RTF_GATEWAY is set, RTAX_GATEWAY exists as well.
1283 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
1285 struct epoch_tracker et;
1289 * ifp may be specified by sockaddr_dl
1290 * when protocol address is ambiguous.
1293 needref = (info->rti_ifa == NULL);
1294 NET_EPOCH_ENTER(et);
1296 /* If we have interface specified by the ifindex in the address, use it */
1297 if (info->rti_ifp == NULL && ifpaddr != NULL &&
1298 ifpaddr->sa_family == AF_LINK) {
1299 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)ifpaddr;
1300 if (sdl->sdl_index != 0)
1301 info->rti_ifp = ifnet_byindex(sdl->sdl_index);
1304 * If we have source address specified, try to find it
1305 * TODO: avoid enumerating all ifas on all interfaces.
1307 if (info->rti_ifa == NULL && ifaaddr != NULL)
1308 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
1309 if (info->rti_ifa == NULL) {
1310 struct sockaddr *sa;
1313 * Most common use case for the userland-supplied routes.
1315 * Choose sockaddr to select ifa.
1316 * -- if ifp is set --
1317 * Order of preference:
1319 * 2) gateway address
1320 * Note: for interface routes link-level gateway address
1321 * is specified to indicate the interface index without
1322 * specifying RTF_GATEWAY. In this case, ignore gateway
1323 * Note: gateway AF may be different from dst AF. In this case,
1325 * 3) final destination.
1326 * 4) if all of these fails, try to get at least link-level ifa.
1328 * try to lookup gateway or dst in the routing table to get ifa
1330 if (info->rti_info[RTAX_IFA] != NULL)
1331 sa = info->rti_info[RTAX_IFA];
1332 else if ((info->rti_flags & RTF_GATEWAY) != 0 &&
1333 gateway->sa_family == dst->sa_family)
1337 if (info->rti_ifp != NULL) {
1338 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
1340 if (info->rti_ifa == NULL && gateway != NULL)
1341 info->rti_ifa = ifaof_ifpforaddr(gateway, info->rti_ifp);
1342 } else if (dst != NULL && gateway != NULL)
1343 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway,
1345 else if (sa != NULL)
1346 info->rti_ifa = ifa_ifwithroute(flags, sa, sa,
1349 if (needref && info->rti_ifa != NULL) {
1350 if (info->rti_ifp == NULL)
1351 info->rti_ifp = info->rti_ifa->ifa_ifp;
1352 ifa_ref(info->rti_ifa);
1354 error = ENETUNREACH;
1360 if_updatemtu_cb(struct radix_node *rn, void *arg)
1363 struct if_mtuinfo *ifmtu;
1365 rt = (struct rtentry *)rn;
1366 ifmtu = (struct if_mtuinfo *)arg;
1368 if (rt->rt_ifp != ifmtu->ifp)
1371 if (rt->rt_mtu >= ifmtu->mtu) {
1372 /* We have to decrease mtu regardless of flags */
1373 rt->rt_mtu = ifmtu->mtu;
1378 * New MTU is bigger. Check if are allowed to alter it
1380 if ((rt->rt_flags & (RTF_FIXEDMTU | RTF_GATEWAY | RTF_HOST)) != 0) {
1383 * Skip routes with user-supplied MTU and
1384 * non-interface routes
1389 /* We are safe to update route MTU */
1390 rt->rt_mtu = ifmtu->mtu;
1396 rt_updatemtu(struct ifnet *ifp)
1398 struct if_mtuinfo ifmtu;
1399 struct rib_head *rnh;
1405 * Try to update rt_mtu for all routes using this interface
1406 * Unfortunately the only way to do this is to traverse all
1407 * routing tables in all fibs/domains.
1409 for (i = 1; i <= AF_MAX; i++) {
1410 ifmtu.mtu = if_getmtu_family(ifp, i);
1411 for (j = 0; j < rt_numfibs; j++) {
1412 rnh = rt_tables_get_rnh(j, i);
1416 rnh->rnh_walktree(&rnh->head, if_updatemtu_cb, &ifmtu);
1424 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
1425 int rt_print(char *buf, int buflen, struct rtentry *rt);
1428 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
1432 switch (s->sa_family) {
1434 paddr = &((struct sockaddr_in *)s)->sin_addr;
1437 paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
1444 if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
1447 return (strlen(buf));
1451 rt_print(char *buf, int buflen, struct rtentry *rt)
1453 struct sockaddr *addr, *mask;
1459 i = p_sockaddr(buf, buflen, addr);
1460 if (!(rt->rt_flags & RTF_HOST)) {
1462 i += p_sockaddr(buf + i, buflen - i, mask);
1465 if (rt->rt_flags & RTF_GATEWAY) {
1467 i += p_sockaddr(buf + i, buflen - i, rt->rt_gateway);
1476 * Deletes key for single-path routes, unlinks rtentry with
1477 * gateway specified in @info from multi-path routes.
1479 * Returnes unlinked entry. In case of failure, returns NULL
1480 * and sets @perror to ESRCH.
1482 static struct radix_node *
1483 rt_mpath_unlink(struct rib_head *rnh, struct rt_addrinfo *info,
1484 struct rtentry *rto, int *perror)
1487 * if we got multipath routes, we require users to specify
1488 * a matching RTAX_GATEWAY.
1490 struct rtentry *rt; // *rto = NULL;
1491 struct radix_node *rn;
1492 struct sockaddr *gw;
1494 gw = info->rti_info[RTAX_GATEWAY];
1495 rt = rt_mpath_matchgate(rto, gw);
1502 * this is the first entry in the chain
1505 rn = rn_mpath_next((struct radix_node *)rt);
1507 * there is another entry, now it's active
1512 rto->rt_flags |= RTF_UP;
1514 } else if (rt->rt_flags & RTF_GATEWAY) {
1516 * For gateway routes, we need to
1517 * make sure that we we are deleting
1518 * the correct gateway.
1519 * rt_mpath_matchgate() does not
1520 * check the case when there is only
1521 * one route in the chain.
1524 (rt->rt_gateway->sa_len != gw->sa_len ||
1525 memcmp(rt->rt_gateway, gw, gw->sa_len))) {
1532 * use the normal delete code to remove
1535 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1541 * if the entry is 2nd and on up
1543 if (rt_mpath_deldup(rto, rt) == 0)
1544 panic ("rtrequest1: rt_mpath_deldup");
1546 rn = (struct radix_node *)rt;
1552 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt,
1556 struct rtentry *rt, *rt_old;
1557 struct radix_node *rn;
1558 struct rib_head *rnh;
1560 struct sockaddr *ndst;
1561 struct sockaddr_storage mdst;
1563 KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum"));
1564 KASSERT((flags & RTF_RNH_LOCKED) == 0, ("rtrequest1_fib: locked"));
1565 switch (dst->sa_family) {
1568 /* We support multiple FIBs. */
1571 fibnum = RT_DEFAULT_FIB;
1576 * Find the correct routing tree to use for this Address Family
1578 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
1580 return (EAFNOSUPPORT);
1583 * If we are adding a host route then we don't want to put
1584 * a netmask in the tree, nor do we want to clone it.
1586 if (flags & RTF_HOST)
1592 if (dst->sa_len > sizeof(mdst))
1594 rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask);
1595 dst = (struct sockaddr *)&mdst;
1599 rt = rt_unlinkrte(rnh, info, &error);
1604 rt_notifydelete(rt, info);
1607 * If the caller wants it, then it can have it,
1608 * but it's up to it to free the rtentry as we won't be
1619 * resolve was only used for route cloning
1624 if ((flags & RTF_GATEWAY) && !gateway)
1626 if (dst && gateway && (dst->sa_family != gateway->sa_family) &&
1627 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
1630 if (info->rti_ifa == NULL) {
1631 error = rt_getifa_fib(info, fibnum);
1635 rt = uma_zalloc(V_rtzone, M_NOWAIT);
1639 rt->rt_flags = RTF_UP | flags;
1640 rt->rt_fibnum = fibnum;
1642 * Add the gateway. Possibly re-malloc-ing the storage for it.
1644 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
1645 uma_zfree(V_rtzone, rt);
1650 * point to the (possibly newly malloc'd) dest address.
1652 ndst = (struct sockaddr *)rt_key(rt);
1655 * make sure it contains the value we want (masked if needed).
1658 rt_maskedcopy(dst, ndst, netmask);
1660 bcopy(dst, ndst, dst->sa_len);
1663 * We use the ifa reference returned by rt_getifa_fib().
1664 * This moved from below so that rnh->rnh_addaddr() can
1665 * examine the ifa and ifa->ifa_ifp if it so desires.
1667 ifa = info->rti_ifa;
1670 rt->rt_ifp = ifa->ifa_ifp;
1673 rt_setmetrics(info, rt);
1678 /* do not permit exactly the same dst/mask/gw pair */
1679 if (rt_mpath_capable(rnh) &&
1680 rt_mpath_conflict(rnh, rt, netmask)) {
1683 ifa_free(rt->rt_ifa);
1685 uma_zfree(V_rtzone, rt);
1690 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
1691 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head, rt->rt_nodes);
1694 if (rn == NULL && (info->rti_flags & RTF_PINNED) != 0) {
1697 * Force removal and re-try addition
1698 * TODO: better multipath&pinned support
1700 struct sockaddr *info_dst = info->rti_info[RTAX_DST];
1701 info->rti_info[RTAX_DST] = ndst;
1702 /* Do not delete existing PINNED(interface) routes */
1703 info->rti_flags &= ~RTF_PINNED;
1704 rt_old = rt_unlinkrte(rnh, info, &error);
1705 info->rti_flags |= RTF_PINNED;
1706 info->rti_info[RTAX_DST] = info_dst;
1708 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head,
1717 * If it still failed to go into the tree,
1718 * then un-make it (this should be a function)
1721 ifa_free(rt->rt_ifa);
1723 uma_zfree(V_rtzone, rt);
1727 if (rt_old != NULL) {
1728 rt_notifydelete(rt_old, info);
1733 * If this protocol has something to add to this then
1734 * allow it to do that as well.
1736 if (ifa->ifa_rtrequest)
1737 ifa->ifa_rtrequest(req, rt, info);
1740 * actually return a resultant rtentry and
1741 * give the caller a single reference.
1747 rnh->rnh_gen++; /* Routing table updated */
1752 error = rtrequest1_fib_change(rnh, info, ret_nrt, fibnum);
1770 rtrequest1_fib_change(struct rib_head *rnh, struct rt_addrinfo *info,
1771 struct rtentry **ret_nrt, u_int fibnum)
1773 struct rtentry *rt = NULL;
1777 struct if_mtuinfo ifmtu;
1779 RIB_WLOCK_ASSERT(rnh);
1781 rt = (struct rtentry *)rnh->rnh_lookup(info->rti_info[RTAX_DST],
1782 info->rti_info[RTAX_NETMASK], &rnh->head);
1789 * If we got multipath routes,
1790 * we require users to specify a matching RTAX_GATEWAY.
1792 if (rt_mpath_capable(rnh)) {
1793 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
1801 rt_setmetrics(info, rt);
1804 * New gateway could require new ifaddr, ifp;
1805 * flags may also be different; ifp may be specified
1806 * by ll sockaddr when protocol address is ambiguous
1808 if (((rt->rt_flags & RTF_GATEWAY) &&
1809 info->rti_info[RTAX_GATEWAY] != NULL) ||
1810 info->rti_info[RTAX_IFP] != NULL ||
1811 (info->rti_info[RTAX_IFA] != NULL &&
1812 !sa_equal(info->rti_info[RTAX_IFA], rt->rt_ifa->ifa_addr))) {
1814 * XXX: Temporarily set RTF_RNH_LOCKED flag in the rti_flags
1815 * to avoid rlock in the ifa_ifwithroute().
1817 info->rti_flags |= RTF_RNH_LOCKED;
1818 error = rt_getifa_fib(info, fibnum);
1819 info->rti_flags &= ~RTF_RNH_LOCKED;
1820 if (info->rti_ifa != NULL)
1827 /* Check if outgoing interface has changed */
1828 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa &&
1829 rt->rt_ifa != NULL) {
1830 if (rt->rt_ifa->ifa_rtrequest != NULL)
1831 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1832 ifa_free(rt->rt_ifa);
1835 /* Update gateway address */
1836 if (info->rti_info[RTAX_GATEWAY] != NULL) {
1837 error = rt_setgate(rt, rt_key(rt), info->rti_info[RTAX_GATEWAY]);
1841 rt->rt_flags &= ~RTF_GATEWAY;
1842 rt->rt_flags |= (RTF_GATEWAY & info->rti_flags);
1845 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa) {
1846 ifa_ref(info->rti_ifa);
1847 rt->rt_ifa = info->rti_ifa;
1848 rt->rt_ifp = info->rti_ifp;
1850 /* Allow some flags to be toggled on change. */
1851 rt->rt_flags &= ~RTF_FMASK;
1852 rt->rt_flags |= info->rti_flags & RTF_FMASK;
1854 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest != NULL)
1855 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
1857 /* Alter route MTU if necessary */
1858 if (rt->rt_ifp != NULL) {
1859 family = info->rti_info[RTAX_DST]->sa_family;
1860 mtu = if_getmtu_family(rt->rt_ifp, family);
1861 /* Set default MTU */
1862 if (rt->rt_mtu == 0)
1864 if (rt->rt_mtu != mtu) {
1865 /* Check if we really need to update */
1866 ifmtu.ifp = rt->rt_ifp;
1868 if_updatemtu_cb(rt->rt_nodes, &ifmtu);
1873 * This route change may have modified the route's gateway. In that
1874 * case, any inpcbs that have cached this route need to invalidate their
1885 if (free_ifa != 0) {
1886 ifa_free(info->rti_ifa);
1887 info->rti_ifa = NULL;
1893 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt)
1896 if (info->rti_mflags & RTV_MTU) {
1897 if (info->rti_rmx->rmx_mtu != 0) {
1900 * MTU was explicitly provided by user.
1903 rt->rt_flags |= RTF_FIXEDMTU;
1907 * User explicitly sets MTU to 0.
1908 * Assume rollback to default.
1910 rt->rt_flags &= ~RTF_FIXEDMTU;
1912 rt->rt_mtu = info->rti_rmx->rmx_mtu;
1914 if (info->rti_mflags & RTV_WEIGHT)
1915 rt->rt_weight = info->rti_rmx->rmx_weight;
1916 /* Kernel -> userland timebase conversion. */
1917 if (info->rti_mflags & RTV_EXPIRE)
1918 rt->rt_expire = info->rti_rmx->rmx_expire ?
1919 info->rti_rmx->rmx_expire - time_second + time_uptime : 0;
1923 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
1925 /* XXX dst may be overwritten, can we move this to below */
1926 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
1929 * Prepare to store the gateway in rt->rt_gateway.
1930 * Both dst and gateway are stored one after the other in the same
1931 * malloc'd chunk. If we have room, we can reuse the old buffer,
1932 * rt_gateway already points to the right place.
1933 * Otherwise, malloc a new block and update the 'dst' address.
1935 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
1938 R_Malloc(new, caddr_t, dlen + glen);
1942 * XXX note, we copy from *dst and not *rt_key(rt) because
1943 * rt_setgate() can be called to initialize a newly
1944 * allocated route entry, in which case rt_key(rt) == NULL
1945 * (and also rt->rt_gateway == NULL).
1946 * Free()/free() handle a NULL argument just fine.
1948 bcopy(dst, new, dlen);
1949 R_Free(rt_key(rt)); /* free old block, if any */
1950 rt_key(rt) = (struct sockaddr *)new;
1951 rt->rt_gateway = (struct sockaddr *)(new + dlen);
1955 * Copy the new gateway value into the memory chunk.
1957 bcopy(gate, rt->rt_gateway, glen);
1963 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
1965 u_char *cp1 = (u_char *)src;
1966 u_char *cp2 = (u_char *)dst;
1967 u_char *cp3 = (u_char *)netmask;
1968 u_char *cplim = cp2 + *cp3;
1969 u_char *cplim2 = cp2 + *cp1;
1971 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1976 *cp2++ = *cp1++ & *cp3++;
1978 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1982 * Set up a routing table entry, normally
1985 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
1987 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
1990 struct sockaddr *dst;
1991 struct sockaddr *netmask;
1992 struct rtentry *rt = NULL;
1993 struct rt_addrinfo info;
1995 int startfib, endfib;
1996 char tempbuf[_SOCKADDR_TMPSIZE];
1999 struct sockaddr_dl *sdl = NULL;
2000 struct rib_head *rnh;
2002 if (flags & RTF_HOST) {
2003 dst = ifa->ifa_dstaddr;
2006 dst = ifa->ifa_addr;
2007 netmask = ifa->ifa_netmask;
2009 if (dst->sa_len == 0)
2011 switch (dst->sa_family) {
2014 /* We support multiple FIBs. */
2017 fibnum = RT_DEFAULT_FIB;
2020 if (fibnum == RT_ALL_FIBS) {
2021 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD)
2022 startfib = endfib = ifa->ifa_ifp->if_fib;
2025 endfib = rt_numfibs - 1;
2028 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
2034 * If it's a delete, check that if it exists,
2035 * it's on the correct interface or we might scrub
2036 * a route to another ifa which would
2037 * be confusing at best and possibly worse.
2039 if (cmd == RTM_DELETE) {
2041 * It's a delete, so it should already exist..
2042 * If it's a net, mask off the host bits
2043 * (Assuming we have a mask)
2044 * XXX this is kinda inet specific..
2046 if (netmask != NULL) {
2047 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
2048 dst = (struct sockaddr *)tempbuf;
2050 } else if (cmd == RTM_ADD) {
2051 sdl = (struct sockaddr_dl *)tempbuf;
2052 bzero(sdl, sizeof(struct sockaddr_dl));
2053 sdl->sdl_family = AF_LINK;
2054 sdl->sdl_len = sizeof(struct sockaddr_dl);
2055 sdl->sdl_type = ifa->ifa_ifp->if_type;
2056 sdl->sdl_index = ifa->ifa_ifp->if_index;
2059 * Now go through all the requested tables (fibs) and do the
2060 * requested action. Realistically, this will either be fib 0
2061 * for protocols that don't do multiple tables or all the
2062 * tables for those that do.
2064 for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
2065 if (cmd == RTM_DELETE) {
2066 struct radix_node *rn;
2068 * Look up an rtentry that is in the routing tree and
2069 * contains the correct info.
2071 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
2073 /* this table doesn't exist but others might */
2076 rn = rnh->rnh_lookup(dst, netmask, &rnh->head);
2078 if (rt_mpath_capable(rnh)) {
2085 * for interface route the
2086 * rt->rt_gateway is sockaddr_intf
2087 * for cloning ARP entries, so
2088 * rt_mpath_matchgate must use the
2091 rt = rt_mpath_matchgate(rt,
2098 error = (rn == NULL ||
2099 (rn->rn_flags & RNF_ROOT) ||
2100 RNTORT(rn)->rt_ifa != ifa);
2103 /* this is only an error if bad on ALL tables */
2108 * Do the actual request
2110 bzero((caddr_t)&info, sizeof(info));
2113 info.rti_flags = flags |
2114 (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED;
2115 info.rti_info[RTAX_DST] = dst;
2117 * doing this for compatibility reasons
2120 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)sdl;
2122 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
2123 info.rti_info[RTAX_NETMASK] = netmask;
2124 error = rtrequest1_fib(cmd, &info, &rt, fibnum);
2126 if (error == 0 && rt != NULL) {
2128 * notify any listening routing agents of the change
2133 * in case address alias finds the first address
2134 * e.g. ifconfig bge0 192.0.2.246/24
2135 * e.g. ifconfig bge0 192.0.2.247/24
2136 * the address set in the route is 192.0.2.246
2137 * so we need to replace it with 192.0.2.247
2139 if (memcmp(rt->rt_ifa->ifa_addr,
2140 ifa->ifa_addr, ifa->ifa_addr->sa_len)) {
2141 ifa_free(rt->rt_ifa);
2143 rt->rt_ifp = ifa->ifa_ifp;
2149 rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum);
2152 if (cmd == RTM_DELETE) {
2154 * If we are deleting, and we found an entry,
2155 * then it's been removed from the tree..
2156 * now throw it away.
2160 if (cmd == RTM_ADD) {
2162 * We just wanted to add it..
2163 * we don't actually need a reference.
2174 if (cmd == RTM_DELETE) {
2178 /* we only give an error if it wasn't in any table */
2179 error = ((flags & RTF_HOST) ?
2180 EHOSTUNREACH : ENETUNREACH);
2184 /* return an error if any of them failed */
2192 * Set up a routing table entry, normally
2196 rtinit(struct ifaddr *ifa, int cmd, int flags)
2198 struct sockaddr *dst;
2199 int fib = RT_DEFAULT_FIB;
2201 if (flags & RTF_HOST) {
2202 dst = ifa->ifa_dstaddr;
2204 dst = ifa->ifa_addr;
2207 switch (dst->sa_family) {
2210 /* We do support multiple FIBs. */
2214 return (rtinit1(ifa, cmd, flags, fib));
2218 * Announce interface address arrival/withdraw
2219 * Returns 0 on success.
2222 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
2225 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2226 ("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 EVENTHANDLER_DIRECT_INVOKE(rt_addrmsg, ifa, cmd);
2231 return (rtsock_addrmsg(cmd, ifa, fibnum));
2235 * Announce route addition/removal.
2236 * Users of this function MUST validate input data BEFORE calling.
2237 * However we have to be able to handle invalid data:
2238 * if some userland app sends us "invalid" route message (invalid mask,
2239 * no dst, wrong address families, etc...) we need to pass it back
2240 * to app (and any other rtsock consumers) with rtm_errno field set to
2242 * Returns 0 on success.
2245 rt_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
2249 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2250 ("unexpected cmd %d", cmd));
2252 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2253 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2255 KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
2257 return (rtsock_routemsg(cmd, ifp, error, rt, fibnum));
2261 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
2264 rt_newaddrmsg_fib(cmd, ifa, error, rt, RT_ALL_FIBS);
2268 * This is called to generate messages from the routing socket
2269 * indicating a network interface has had addresses associated with it.
2272 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
2276 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2277 ("unexpected cmd %u", cmd));
2278 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2279 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2281 if (cmd == RTM_ADD) {
2282 rt_addrmsg(cmd, ifa, fibnum);
2284 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2287 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2288 rt_addrmsg(cmd, ifa, fibnum);