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, table);
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, int family, u_int fibnum)
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;
360 /* Save metadata associated with this routing table. */
361 rh->rib_family = family;
362 rh->rib_fibnum = fibnum;
364 rh->rib_vnet = curvnet;
370 /* Finally, set base callbacks */
371 rh->rnh_addaddr = rn_addroute;
372 rh->rnh_deladdr = rn_delete;
373 rh->rnh_matchaddr = rn_match;
374 rh->rnh_lookup = rn_lookup;
375 rh->rnh_walktree = rn_walktree;
376 rh->rnh_walktree_from = rn_walktree_from;
382 rt_freeentry(struct radix_node *rn, void *arg)
384 struct radix_head * const rnh = arg;
385 struct radix_node *x;
387 x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh);
394 rt_table_destroy(struct rib_head *rh)
397 rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head);
399 /* Assume table is already empty */
400 RIB_LOCK_DESTROY(rh);
405 #ifndef _SYS_SYSPROTO_H_
411 sys_setfib(struct thread *td, struct setfib_args *uap)
413 if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs)
415 td->td_proc->p_fibnum = uap->fibnum;
420 * Packet routing routines.
423 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum)
427 if ((rt = ro->ro_rt) != NULL) {
428 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
433 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum);
435 RT_UNLOCK(ro->ro_rt);
439 * Look up the route that matches the address given
440 * Or, at least try.. Create a cloned route if needed.
442 * The returned route, if any, is locked.
445 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
448 return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB));
452 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags,
457 struct radix_node *rn;
458 struct rtentry *newrt;
459 struct rt_addrinfo info;
460 int err = 0, msgtype = RTM_MISS;
462 KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum"));
463 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
469 * Look up the address in the table for that Address Family
471 if ((ignflags & RTF_RNH_LOCKED) == 0)
477 rn = rh->rnh_matchaddr(dst, &rh->head);
478 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
482 if ((ignflags & RTF_RNH_LOCKED) == 0)
486 } else if ((ignflags & RTF_RNH_LOCKED) == 0)
489 * Either we hit the root or could not find any match,
490 * which basically means: "cannot get there from here".
493 RTSTAT_INC(rts_unreach);
497 * If required, report the failure to the supervising
499 * For a delete, this is not an error. (report == 0)
501 bzero(&info, sizeof(info));
502 info.rti_info[RTAX_DST] = dst;
503 rt_missmsg_fib(msgtype, &info, 0, err, fibnum);
509 * Remove a reference count from an rtentry.
510 * If the count gets low enough, take it out of the routing table
513 rtfree(struct rtentry *rt)
515 struct rib_head *rnh;
517 KASSERT(rt != NULL,("%s: NULL rt", __func__));
518 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
519 KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
524 * The callers should use RTFREE_LOCKED() or RTFREE(), so
525 * we should come here exactly with the last reference.
528 if (rt->rt_refcnt > 0) {
529 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt);
534 * On last reference give the "close method" a chance
535 * to cleanup private state. This also permits (for
536 * IPv4 and IPv6) a chance to decide if the routing table
537 * entry should be purged immediately or at a later time.
538 * When an immediate purge is to happen the close routine
539 * typically calls rtexpunge which clears the RTF_UP flag
540 * on the entry so that the code below reclaims the storage.
542 if (rt->rt_refcnt == 0 && rnh->rnh_close)
543 rnh->rnh_close((struct radix_node *)rt, &rnh->head);
546 * If we are no longer "up" (and ref == 0)
547 * then we can free the resources associated
550 if ((rt->rt_flags & RTF_UP) == 0) {
551 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
554 * the rtentry must have been removed from the routing table
555 * so it is represented in rttrash.. remove that now.
559 if (rt->rt_refcnt < 0) {
560 printf("rtfree: %p not freed (neg refs)\n", rt);
565 * release references on items we hold them on..
566 * e.g other routes and ifaddrs.
569 ifa_free(rt->rt_ifa);
571 * The key is separatly alloc'd so free it (see rt_setgate()).
572 * This also frees the gateway, as they are always malloc'd
578 * and the rtentry itself of course
580 uma_zfree(V_rtzone, rt);
589 * Force a routing table entry to the specified
590 * destination to go through the given gateway.
591 * Normally called as a result of a routing redirect
592 * message from the network layer.
595 rtredirect_fib(struct sockaddr *dst,
596 struct sockaddr *gateway,
597 struct sockaddr *netmask,
599 struct sockaddr *src,
604 struct rt_addrinfo info;
606 struct rib_head *rnh;
611 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
613 error = EAFNOSUPPORT;
616 /* verify the gateway is directly reachable */
617 if ((ifa = ifa_ifwithnet(gateway, 0, fibnum)) == NULL) {
621 rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */
623 * If the redirect isn't from our current router for this dst,
624 * it's either old or wrong. If it redirects us to ourselves,
625 * we have a routing loop, perhaps as a result of an interface
626 * going down recently.
628 if (!(flags & RTF_DONE) && rt) {
629 if (!sa_equal(src, rt->rt_gateway)) {
633 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) {
638 if ((flags & RTF_GATEWAY) && ifa_ifwithaddr_check(gateway)) {
639 error = EHOSTUNREACH;
643 * Create a new entry if we just got back a wildcard entry
644 * or the lookup failed. This is necessary for hosts
645 * which use routing redirects generated by smart gateways
646 * to dynamically build the routing tables.
648 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
651 * Don't listen to the redirect if it's
652 * for a route to an interface.
654 if (rt->rt_flags & RTF_GATEWAY) {
655 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
657 * Changing from route to net => route to host.
658 * Create new route, rather than smashing route to net.
664 flags |= RTF_DYNAMIC;
665 bzero((caddr_t)&info, sizeof(info));
666 info.rti_info[RTAX_DST] = dst;
667 info.rti_info[RTAX_GATEWAY] = gateway;
668 info.rti_info[RTAX_NETMASK] = netmask;
671 info.rti_flags = flags;
672 error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum);
675 flags = rt->rt_flags;
678 RTSTAT_INC(rts_dynamic);
682 * Smash the current notion of the gateway to
683 * this destination. Should check about netmask!!!
685 if ((flags & RTF_GATEWAY) == 0)
686 rt->rt_flags &= ~RTF_GATEWAY;
687 rt->rt_flags |= RTF_MODIFIED;
688 flags |= RTF_MODIFIED;
689 RTSTAT_INC(rts_newgateway);
691 * add the key and gateway (in one malloc'd chunk).
696 rt_setgate(rt, rt_key(rt), gateway);
700 error = EHOSTUNREACH;
706 RTSTAT_INC(rts_badredirect);
707 bzero((caddr_t)&info, sizeof(info));
708 info.rti_info[RTAX_DST] = dst;
709 info.rti_info[RTAX_GATEWAY] = gateway;
710 info.rti_info[RTAX_NETMASK] = netmask;
711 info.rti_info[RTAX_AUTHOR] = src;
712 rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum);
716 * Routing table ioctl interface.
719 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
723 * If more ioctl commands are added here, make sure the proper
724 * super-user checks are being performed because it is possible for
725 * prison-root to make it this far if raw sockets have been enabled
729 /* Multicast goop, grrr... */
730 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
737 ifa_ifwithroute(int flags, const struct sockaddr *dst, struct sockaddr *gateway,
743 MPASS(in_epoch(net_epoch_preempt));
744 if ((flags & RTF_GATEWAY) == 0) {
746 * If we are adding a route to an interface,
747 * and the interface is a pt to pt link
748 * we should search for the destination
749 * as our clue to the interface. Otherwise
750 * we can use the local address.
753 if (flags & RTF_HOST)
754 ifa = ifa_ifwithdstaddr(dst, fibnum);
756 ifa = ifa_ifwithaddr(gateway);
759 * If we are adding a route to a remote net
760 * or host, the gateway may still be on the
761 * other end of a pt to pt link.
763 ifa = ifa_ifwithdstaddr(gateway, fibnum);
766 ifa = ifa_ifwithnet(gateway, 0, fibnum);
770 rt = rtalloc1_fib(gateway, 0, flags, fibnum);
774 * dismiss a gateway that is reachable only
775 * through the default router
777 switch (gateway->sa_family) {
779 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
783 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
789 if (!not_found && rt->rt_ifa != NULL) {
794 if (not_found || ifa == NULL)
797 if (ifa->ifa_addr->sa_family != dst->sa_family) {
798 struct ifaddr *oifa = ifa;
799 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
808 * Do appropriate manipulations of a routing tree given
809 * all the bits of info needed
812 rtrequest_fib(int req,
813 struct sockaddr *dst,
814 struct sockaddr *gateway,
815 struct sockaddr *netmask,
817 struct rtentry **ret_nrt,
820 struct rt_addrinfo info;
822 if (dst->sa_len == 0)
825 bzero((caddr_t)&info, sizeof(info));
826 info.rti_flags = flags;
827 info.rti_info[RTAX_DST] = dst;
828 info.rti_info[RTAX_GATEWAY] = gateway;
829 info.rti_info[RTAX_NETMASK] = netmask;
830 return rtrequest1_fib(req, &info, ret_nrt, fibnum);
835 * Copy most of @rt data into @info.
837 * If @flags contains NHR_COPY, copies dst,netmask and gw to the
838 * pointers specified by @info structure. Assume such pointers
839 * are zeroed sockaddr-like structures with sa_len field initialized
840 * to reflect size of the provided buffer. if no NHR_COPY is specified,
841 * point dst,netmask and gw @info fields to appropriate @rt values.
843 * if @flags contains NHR_REF, do refcouting on rt_ifp and rt_ifa.
845 * Returns 0 on success.
848 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags)
850 struct rt_metrics *rmx;
851 struct sockaddr *src, *dst;
854 if (flags & NHR_COPY) {
855 /* Copy destination if dst is non-zero */
857 dst = info->rti_info[RTAX_DST];
858 sa_len = src->sa_len;
860 if (src->sa_len > dst->sa_len)
862 memcpy(dst, src, src->sa_len);
863 info->rti_addrs |= RTA_DST;
866 /* Copy mask if set && dst is non-zero */
868 dst = info->rti_info[RTAX_NETMASK];
869 if (src != NULL && dst != NULL) {
872 * Radix stores different value in sa_len,
873 * assume rt_mask() to have the same length
876 if (sa_len > dst->sa_len)
878 memcpy(dst, src, src->sa_len);
879 info->rti_addrs |= RTA_NETMASK;
882 /* Copy gateway is set && dst is non-zero */
883 src = rt->rt_gateway;
884 dst = info->rti_info[RTAX_GATEWAY];
885 if ((rt->rt_flags & RTF_GATEWAY) && src != NULL && dst != NULL){
886 if (src->sa_len > dst->sa_len)
888 memcpy(dst, src, src->sa_len);
889 info->rti_addrs |= RTA_GATEWAY;
892 info->rti_info[RTAX_DST] = rt_key(rt);
893 info->rti_addrs |= RTA_DST;
894 if (rt_mask(rt) != NULL) {
895 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
896 info->rti_addrs |= RTA_NETMASK;
898 if (rt->rt_flags & RTF_GATEWAY) {
899 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
900 info->rti_addrs |= RTA_GATEWAY;
906 info->rti_mflags |= RTV_MTU;
907 rmx->rmx_mtu = rt->rt_mtu;
910 info->rti_flags = rt->rt_flags;
911 info->rti_ifp = rt->rt_ifp;
912 info->rti_ifa = rt->rt_ifa;
913 if (flags & NHR_REF) {
914 if_ref(info->rti_ifp);
915 ifa_ref(info->rti_ifa);
922 * Lookups up route entry for @dst in RIB database for fib @fibnum.
923 * Exports entry data to @info using rt_exportinfo().
925 * If @flags contains NHR_REF, refcouting is performed on rt_ifp and rt_ifa.
926 * All references can be released later by calling rib_free_info().
928 * Returns 0 on success.
929 * Returns ENOENT for lookup failure, ENOMEM for export failure.
932 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
933 uint32_t flowid, struct rt_addrinfo *info)
937 struct radix_node *rn;
941 KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
942 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
947 rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head);
948 if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
950 /* Ensure route & ifp is UP */
951 if (RT_LINK_IS_UP(rt->rt_ifp)) {
952 flags = (flags & NHR_REF) | NHR_COPY;
953 error = rt_exportinfo(rt, info, flags);
965 * Releases all references acquired by rib_lookup_info() when
966 * called with NHR_REF flags.
969 rib_free_info(struct rt_addrinfo *info)
972 ifa_free(info->rti_ifa);
973 if_rele(info->rti_ifp);
977 * Iterates over all existing fibs in system calling
978 * @setwa_f function prior to traversing each fib.
979 * Calls @wa_f function for each element in current fib.
980 * If af is not AF_UNSPEC, iterates over fibs in particular
984 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f,
987 struct rib_head *rnh;
991 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
992 /* Do we want some specific family? */
993 if (af != AF_UNSPEC) {
994 rnh = rt_tables_get_rnh(fibnum, af);
998 setwa_f(rnh, fibnum, af, arg);
1001 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
1006 for (i = 1; i <= AF_MAX; i++) {
1007 rnh = rt_tables_get_rnh(fibnum, i);
1010 if (setwa_f != NULL)
1011 setwa_f(rnh, fibnum, i, arg);
1014 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
1022 struct rt_addrinfo info;
1023 struct rib_head *rnh;
1024 struct rtentry *head;
1028 * Conditionally unlinks @rn from radix tree based
1029 * on info data passed in @arg.
1032 rt_checkdelroute(struct radix_node *rn, void *arg)
1034 struct rt_delinfo *di;
1035 struct rt_addrinfo *info;
1039 di = (struct rt_delinfo *)arg;
1040 rt = (struct rtentry *)rn;
1044 info->rti_info[RTAX_DST] = rt_key(rt);
1045 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
1046 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1048 rt = rt_unlinkrte(di->rnh, info, &error);
1050 /* Either not allowed or not matched. Skip entry */
1054 /* Entry was unlinked. Add to the list and return */
1055 rt->rt_chain = di->head;
1062 * Iterates over all existing fibs in system.
1063 * Deletes each element for which @filter_f function returned
1065 * If @af is not AF_UNSPEC, iterates over fibs in particular
1069 rt_foreach_fib_walk_del(int af, rt_filter_f_t *filter_f, void *arg)
1071 struct rib_head *rnh;
1072 struct rt_delinfo di;
1077 bzero(&di, sizeof(di));
1078 di.info.rti_filter = filter_f;
1079 di.info.rti_filterdata = arg;
1081 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1082 /* Do we want some specific family? */
1083 if (af != AF_UNSPEC) {
1091 for (i = start; i <= end; i++) {
1092 rnh = rt_tables_get_rnh(fibnum, i);
1098 rnh->rnh_walktree(&rnh->head, rt_checkdelroute, &di);
1101 if (di.head == NULL)
1104 /* We might have something to reclaim */
1105 while (di.head != NULL) {
1107 di.head = rt->rt_chain;
1108 rt->rt_chain = NULL;
1110 /* TODO std rt -> rt_addrinfo export */
1111 di.info.rti_info[RTAX_DST] = rt_key(rt);
1112 di.info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1114 rt_notifydelete(rt, &di.info);
1123 * Delete Routes for a Network Interface
1125 * Called for each routing entry via the rnh->rnh_walktree() call above
1126 * to delete all route entries referencing a detaching network interface.
1129 * rt pointer to rtentry
1130 * arg argument passed to rnh->rnh_walktree() - detaching interface
1134 * errno failed - reason indicated
1137 rt_ifdelroute(const struct rtentry *rt, void *arg)
1139 struct ifnet *ifp = arg;
1141 if (rt->rt_ifp != ifp)
1145 * Protect (sorta) against walktree recursion problems
1146 * with cloned routes
1148 if ((rt->rt_flags & RTF_UP) == 0)
1155 * Delete all remaining routes using this interface
1156 * Unfortuneatly the only way to do this is to slog through
1157 * the entire routing table looking for routes which point
1158 * to this interface...oh well...
1161 rt_flushifroutes_af(struct ifnet *ifp, int af)
1163 KASSERT((af >= 1 && af <= AF_MAX), ("%s: af %d not >= 1 and <= %d",
1164 __func__, af, AF_MAX));
1166 rt_foreach_fib_walk_del(af, rt_ifdelroute, ifp);
1170 rt_flushifroutes(struct ifnet *ifp)
1173 rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
1177 * Conditionally unlinks rtentry matching data inside @info from @rnh.
1178 * Returns unlinked, locked and referenced @rtentry on success,
1179 * Returns NULL and sets @perror to:
1180 * ESRCH - if prefix was not found,
1181 * EADDRINUSE - if trying to delete PINNED route without appropriate flag.
1182 * ENOENT - if supplied filter function returned 0 (not matched).
1184 static struct rtentry *
1185 rt_unlinkrte(struct rib_head *rnh, struct rt_addrinfo *info, int *perror)
1187 struct sockaddr *dst, *netmask;
1189 struct radix_node *rn;
1191 dst = info->rti_info[RTAX_DST];
1192 netmask = info->rti_info[RTAX_NETMASK];
1194 rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, &rnh->head);
1200 if ((info->rti_flags & RTF_PINNED) == 0) {
1201 /* Check if target route can be deleted */
1202 if (rt->rt_flags & RTF_PINNED) {
1203 *perror = EADDRINUSE;
1208 if (info->rti_filter != NULL) {
1209 if (info->rti_filter(rt, info->rti_filterdata) == 0) {
1216 * Filter function requested rte deletion.
1217 * Ease the caller work by filling in remaining info
1218 * from that particular entry.
1220 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1224 * Remove the item from the tree and return it.
1225 * Complain if it is not there and do no more processing.
1229 if (rt_mpath_capable(rnh))
1230 rn = rt_mpath_unlink(rnh, info, rt, perror);
1233 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1237 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
1238 panic ("rtrequest delete");
1243 rt->rt_flags &= ~RTF_UP;
1251 rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info)
1256 * give the protocol a chance to keep things in sync.
1259 if (ifa != NULL && ifa->ifa_rtrequest != NULL)
1260 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1263 * One more rtentry floating around that is not
1264 * linked to the routing table. rttrash will be decremented
1265 * when RTFREE(rt) is eventually called.
1272 * These (questionable) definitions of apparent local variables apply
1273 * to the next two functions. XXXXXX!!!
1275 #define dst info->rti_info[RTAX_DST]
1276 #define gateway info->rti_info[RTAX_GATEWAY]
1277 #define netmask info->rti_info[RTAX_NETMASK]
1278 #define ifaaddr info->rti_info[RTAX_IFA]
1279 #define ifpaddr info->rti_info[RTAX_IFP]
1280 #define flags info->rti_flags
1283 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined,
1284 * it will be referenced so the caller must free it.
1286 * Assume basic consistency checks are executed by callers:
1287 * RTAX_DST exists, if RTF_GATEWAY is set, RTAX_GATEWAY exists as well.
1290 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
1292 struct epoch_tracker et;
1296 * ifp may be specified by sockaddr_dl
1297 * when protocol address is ambiguous.
1300 needref = (info->rti_ifa == NULL);
1301 NET_EPOCH_ENTER(et);
1303 /* If we have interface specified by the ifindex in the address, use it */
1304 if (info->rti_ifp == NULL && ifpaddr != NULL &&
1305 ifpaddr->sa_family == AF_LINK) {
1306 const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)ifpaddr;
1307 if (sdl->sdl_index != 0)
1308 info->rti_ifp = ifnet_byindex(sdl->sdl_index);
1311 * If we have source address specified, try to find it
1312 * TODO: avoid enumerating all ifas on all interfaces.
1314 if (info->rti_ifa == NULL && ifaaddr != NULL)
1315 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
1316 if (info->rti_ifa == NULL) {
1317 struct sockaddr *sa;
1320 * Most common use case for the userland-supplied routes.
1322 * Choose sockaddr to select ifa.
1323 * -- if ifp is set --
1324 * Order of preference:
1326 * 2) gateway address
1327 * Note: for interface routes link-level gateway address
1328 * is specified to indicate the interface index without
1329 * specifying RTF_GATEWAY. In this case, ignore gateway
1330 * Note: gateway AF may be different from dst AF. In this case,
1332 * 3) final destination.
1333 * 4) if all of these fails, try to get at least link-level ifa.
1335 * try to lookup gateway or dst in the routing table to get ifa
1337 if (info->rti_info[RTAX_IFA] != NULL)
1338 sa = info->rti_info[RTAX_IFA];
1339 else if ((info->rti_flags & RTF_GATEWAY) != 0 &&
1340 gateway->sa_family == dst->sa_family)
1344 if (info->rti_ifp != NULL) {
1345 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
1347 if (info->rti_ifa == NULL && gateway != NULL)
1348 info->rti_ifa = ifaof_ifpforaddr(gateway, info->rti_ifp);
1349 } else if (dst != NULL && gateway != NULL)
1350 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway,
1352 else if (sa != NULL)
1353 info->rti_ifa = ifa_ifwithroute(flags, sa, sa,
1356 if (needref && info->rti_ifa != NULL) {
1357 if (info->rti_ifp == NULL)
1358 info->rti_ifp = info->rti_ifa->ifa_ifp;
1359 ifa_ref(info->rti_ifa);
1361 error = ENETUNREACH;
1367 if_updatemtu_cb(struct radix_node *rn, void *arg)
1370 struct if_mtuinfo *ifmtu;
1372 rt = (struct rtentry *)rn;
1373 ifmtu = (struct if_mtuinfo *)arg;
1375 if (rt->rt_ifp != ifmtu->ifp)
1378 if (rt->rt_mtu >= ifmtu->mtu) {
1379 /* We have to decrease mtu regardless of flags */
1380 rt->rt_mtu = ifmtu->mtu;
1385 * New MTU is bigger. Check if are allowed to alter it
1387 if ((rt->rt_flags & (RTF_FIXEDMTU | RTF_GATEWAY | RTF_HOST)) != 0) {
1390 * Skip routes with user-supplied MTU and
1391 * non-interface routes
1396 /* We are safe to update route MTU */
1397 rt->rt_mtu = ifmtu->mtu;
1403 rt_updatemtu(struct ifnet *ifp)
1405 struct if_mtuinfo ifmtu;
1406 struct rib_head *rnh;
1412 * Try to update rt_mtu for all routes using this interface
1413 * Unfortunately the only way to do this is to traverse all
1414 * routing tables in all fibs/domains.
1416 for (i = 1; i <= AF_MAX; i++) {
1417 ifmtu.mtu = if_getmtu_family(ifp, i);
1418 for (j = 0; j < rt_numfibs; j++) {
1419 rnh = rt_tables_get_rnh(j, i);
1423 rnh->rnh_walktree(&rnh->head, if_updatemtu_cb, &ifmtu);
1431 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
1432 int rt_print(char *buf, int buflen, struct rtentry *rt);
1435 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
1439 switch (s->sa_family) {
1441 paddr = &((struct sockaddr_in *)s)->sin_addr;
1444 paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
1451 if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
1454 return (strlen(buf));
1458 rt_print(char *buf, int buflen, struct rtentry *rt)
1460 struct sockaddr *addr, *mask;
1466 i = p_sockaddr(buf, buflen, addr);
1467 if (!(rt->rt_flags & RTF_HOST)) {
1469 i += p_sockaddr(buf + i, buflen - i, mask);
1472 if (rt->rt_flags & RTF_GATEWAY) {
1474 i += p_sockaddr(buf + i, buflen - i, rt->rt_gateway);
1483 * Deletes key for single-path routes, unlinks rtentry with
1484 * gateway specified in @info from multi-path routes.
1486 * Returnes unlinked entry. In case of failure, returns NULL
1487 * and sets @perror to ESRCH.
1489 static struct radix_node *
1490 rt_mpath_unlink(struct rib_head *rnh, struct rt_addrinfo *info,
1491 struct rtentry *rto, int *perror)
1494 * if we got multipath routes, we require users to specify
1495 * a matching RTAX_GATEWAY.
1497 struct rtentry *rt; // *rto = NULL;
1498 struct radix_node *rn;
1499 struct sockaddr *gw;
1501 gw = info->rti_info[RTAX_GATEWAY];
1502 rt = rt_mpath_matchgate(rto, gw);
1509 * this is the first entry in the chain
1512 rn = rn_mpath_next((struct radix_node *)rt);
1514 * there is another entry, now it's active
1519 rto->rt_flags |= RTF_UP;
1521 } else if (rt->rt_flags & RTF_GATEWAY) {
1523 * For gateway routes, we need to
1524 * make sure that we we are deleting
1525 * the correct gateway.
1526 * rt_mpath_matchgate() does not
1527 * check the case when there is only
1528 * one route in the chain.
1531 (rt->rt_gateway->sa_len != gw->sa_len ||
1532 memcmp(rt->rt_gateway, gw, gw->sa_len))) {
1539 * use the normal delete code to remove
1542 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1548 * if the entry is 2nd and on up
1550 if (rt_mpath_deldup(rto, rt) == 0)
1551 panic ("rtrequest1: rt_mpath_deldup");
1553 rn = (struct radix_node *)rt;
1559 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt,
1563 struct rtentry *rt, *rt_old;
1564 struct radix_node *rn;
1565 struct rib_head *rnh;
1567 struct sockaddr *ndst;
1568 struct sockaddr_storage mdst;
1570 KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum"));
1571 KASSERT((flags & RTF_RNH_LOCKED) == 0, ("rtrequest1_fib: locked"));
1572 switch (dst->sa_family) {
1575 /* We support multiple FIBs. */
1578 fibnum = RT_DEFAULT_FIB;
1583 * Find the correct routing tree to use for this Address Family
1585 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
1587 return (EAFNOSUPPORT);
1590 * If we are adding a host route then we don't want to put
1591 * a netmask in the tree, nor do we want to clone it.
1593 if (flags & RTF_HOST)
1599 if (dst->sa_len > sizeof(mdst))
1601 rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask);
1602 dst = (struct sockaddr *)&mdst;
1606 rt = rt_unlinkrte(rnh, info, &error);
1611 rt_notifydelete(rt, info);
1614 * If the caller wants it, then it can have it,
1615 * but it's up to it to free the rtentry as we won't be
1626 * resolve was only used for route cloning
1631 if ((flags & RTF_GATEWAY) && !gateway)
1633 if (dst && gateway && (dst->sa_family != gateway->sa_family) &&
1634 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
1637 if (info->rti_ifa == NULL) {
1638 error = rt_getifa_fib(info, fibnum);
1642 ifa_ref(info->rti_ifa);
1644 rt = uma_zalloc(V_rtzone, M_NOWAIT);
1646 ifa_free(info->rti_ifa);
1649 rt->rt_flags = RTF_UP | flags;
1650 rt->rt_fibnum = fibnum;
1652 * Add the gateway. Possibly re-malloc-ing the storage for it.
1654 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
1655 ifa_free(info->rti_ifa);
1656 uma_zfree(V_rtzone, rt);
1661 * point to the (possibly newly malloc'd) dest address.
1663 ndst = (struct sockaddr *)rt_key(rt);
1666 * make sure it contains the value we want (masked if needed).
1669 rt_maskedcopy(dst, ndst, netmask);
1671 bcopy(dst, ndst, dst->sa_len);
1674 * We use the ifa reference returned by rt_getifa_fib().
1675 * This moved from below so that rnh->rnh_addaddr() can
1676 * examine the ifa and ifa->ifa_ifp if it so desires.
1678 ifa = info->rti_ifa;
1680 rt->rt_ifp = ifa->ifa_ifp;
1683 rt_setmetrics(info, rt);
1688 /* do not permit exactly the same dst/mask/gw pair */
1689 if (rt_mpath_capable(rnh) &&
1690 rt_mpath_conflict(rnh, rt, netmask)) {
1693 ifa_free(rt->rt_ifa);
1695 uma_zfree(V_rtzone, rt);
1700 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
1701 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head, rt->rt_nodes);
1704 if (rn == NULL && (info->rti_flags & RTF_PINNED) != 0) {
1707 * Force removal and re-try addition
1708 * TODO: better multipath&pinned support
1710 struct sockaddr *info_dst = info->rti_info[RTAX_DST];
1711 info->rti_info[RTAX_DST] = ndst;
1712 /* Do not delete existing PINNED(interface) routes */
1713 info->rti_flags &= ~RTF_PINNED;
1714 rt_old = rt_unlinkrte(rnh, info, &error);
1715 info->rti_flags |= RTF_PINNED;
1716 info->rti_info[RTAX_DST] = info_dst;
1718 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head,
1727 * If it still failed to go into the tree,
1728 * then un-make it (this should be a function)
1731 ifa_free(rt->rt_ifa);
1733 uma_zfree(V_rtzone, rt);
1737 if (rt_old != NULL) {
1738 rt_notifydelete(rt_old, info);
1743 * If this protocol has something to add to this then
1744 * allow it to do that as well.
1746 if (ifa->ifa_rtrequest)
1747 ifa->ifa_rtrequest(req, rt, info);
1750 * actually return a resultant rtentry and
1751 * give the caller a single reference.
1757 rnh->rnh_gen++; /* Routing table updated */
1762 error = rtrequest1_fib_change(rnh, info, ret_nrt, fibnum);
1780 rtrequest1_fib_change(struct rib_head *rnh, struct rt_addrinfo *info,
1781 struct rtentry **ret_nrt, u_int fibnum)
1783 struct rtentry *rt = NULL;
1787 struct if_mtuinfo ifmtu;
1789 RIB_WLOCK_ASSERT(rnh);
1791 rt = (struct rtentry *)rnh->rnh_lookup(info->rti_info[RTAX_DST],
1792 info->rti_info[RTAX_NETMASK], &rnh->head);
1799 * If we got multipath routes,
1800 * we require users to specify a matching RTAX_GATEWAY.
1802 if (rt_mpath_capable(rnh)) {
1803 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
1811 rt_setmetrics(info, rt);
1814 * New gateway could require new ifaddr, ifp;
1815 * flags may also be different; ifp may be specified
1816 * by ll sockaddr when protocol address is ambiguous
1818 if (((rt->rt_flags & RTF_GATEWAY) &&
1819 info->rti_info[RTAX_GATEWAY] != NULL) ||
1820 info->rti_info[RTAX_IFP] != NULL ||
1821 (info->rti_info[RTAX_IFA] != NULL &&
1822 !sa_equal(info->rti_info[RTAX_IFA], rt->rt_ifa->ifa_addr))) {
1824 * XXX: Temporarily set RTF_RNH_LOCKED flag in the rti_flags
1825 * to avoid rlock in the ifa_ifwithroute().
1827 info->rti_flags |= RTF_RNH_LOCKED;
1828 error = rt_getifa_fib(info, fibnum);
1829 info->rti_flags &= ~RTF_RNH_LOCKED;
1830 if (info->rti_ifa != NULL)
1837 /* Check if outgoing interface has changed */
1838 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa &&
1839 rt->rt_ifa != NULL) {
1840 if (rt->rt_ifa->ifa_rtrequest != NULL)
1841 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1842 ifa_free(rt->rt_ifa);
1845 /* Update gateway address */
1846 if (info->rti_info[RTAX_GATEWAY] != NULL) {
1847 error = rt_setgate(rt, rt_key(rt), info->rti_info[RTAX_GATEWAY]);
1851 rt->rt_flags &= ~RTF_GATEWAY;
1852 rt->rt_flags |= (RTF_GATEWAY & info->rti_flags);
1855 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa) {
1856 ifa_ref(info->rti_ifa);
1857 rt->rt_ifa = info->rti_ifa;
1858 rt->rt_ifp = info->rti_ifp;
1860 /* Allow some flags to be toggled on change. */
1861 rt->rt_flags &= ~RTF_FMASK;
1862 rt->rt_flags |= info->rti_flags & RTF_FMASK;
1864 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest != NULL)
1865 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
1867 /* Alter route MTU if necessary */
1868 if (rt->rt_ifp != NULL) {
1869 family = info->rti_info[RTAX_DST]->sa_family;
1870 mtu = if_getmtu_family(rt->rt_ifp, family);
1871 /* Set default MTU */
1872 if (rt->rt_mtu == 0)
1874 if (rt->rt_mtu != mtu) {
1875 /* Check if we really need to update */
1876 ifmtu.ifp = rt->rt_ifp;
1878 if_updatemtu_cb(rt->rt_nodes, &ifmtu);
1883 * This route change may have modified the route's gateway. In that
1884 * case, any inpcbs that have cached this route need to invalidate their
1895 if (free_ifa != 0) {
1896 ifa_free(info->rti_ifa);
1897 info->rti_ifa = NULL;
1903 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt)
1906 if (info->rti_mflags & RTV_MTU) {
1907 if (info->rti_rmx->rmx_mtu != 0) {
1910 * MTU was explicitly provided by user.
1913 rt->rt_flags |= RTF_FIXEDMTU;
1917 * User explicitly sets MTU to 0.
1918 * Assume rollback to default.
1920 rt->rt_flags &= ~RTF_FIXEDMTU;
1922 rt->rt_mtu = info->rti_rmx->rmx_mtu;
1924 if (info->rti_mflags & RTV_WEIGHT)
1925 rt->rt_weight = info->rti_rmx->rmx_weight;
1926 /* Kernel -> userland timebase conversion. */
1927 if (info->rti_mflags & RTV_EXPIRE)
1928 rt->rt_expire = info->rti_rmx->rmx_expire ?
1929 info->rti_rmx->rmx_expire - time_second + time_uptime : 0;
1933 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
1935 /* XXX dst may be overwritten, can we move this to below */
1936 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
1939 * Prepare to store the gateway in rt->rt_gateway.
1940 * Both dst and gateway are stored one after the other in the same
1941 * malloc'd chunk. If we have room, we can reuse the old buffer,
1942 * rt_gateway already points to the right place.
1943 * Otherwise, malloc a new block and update the 'dst' address.
1945 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
1948 R_Malloc(new, caddr_t, dlen + glen);
1952 * XXX note, we copy from *dst and not *rt_key(rt) because
1953 * rt_setgate() can be called to initialize a newly
1954 * allocated route entry, in which case rt_key(rt) == NULL
1955 * (and also rt->rt_gateway == NULL).
1956 * Free()/free() handle a NULL argument just fine.
1958 bcopy(dst, new, dlen);
1959 R_Free(rt_key(rt)); /* free old block, if any */
1960 rt_key(rt) = (struct sockaddr *)new;
1961 rt->rt_gateway = (struct sockaddr *)(new + dlen);
1965 * Copy the new gateway value into the memory chunk.
1967 bcopy(gate, rt->rt_gateway, glen);
1973 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
1975 u_char *cp1 = (u_char *)src;
1976 u_char *cp2 = (u_char *)dst;
1977 u_char *cp3 = (u_char *)netmask;
1978 u_char *cplim = cp2 + *cp3;
1979 u_char *cplim2 = cp2 + *cp1;
1981 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1986 *cp2++ = *cp1++ & *cp3++;
1988 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1992 * Set up a routing table entry, normally
1995 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
1997 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
2000 struct sockaddr *dst;
2001 struct sockaddr *netmask;
2002 struct rtentry *rt = NULL;
2003 struct rt_addrinfo info;
2005 int startfib, endfib;
2006 char tempbuf[_SOCKADDR_TMPSIZE];
2009 struct sockaddr_dl *sdl = NULL;
2010 struct rib_head *rnh;
2012 if (flags & RTF_HOST) {
2013 dst = ifa->ifa_dstaddr;
2016 dst = ifa->ifa_addr;
2017 netmask = ifa->ifa_netmask;
2019 if (dst->sa_len == 0)
2021 switch (dst->sa_family) {
2024 /* We support multiple FIBs. */
2027 fibnum = RT_DEFAULT_FIB;
2030 if (fibnum == RT_ALL_FIBS) {
2031 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD)
2032 startfib = endfib = ifa->ifa_ifp->if_fib;
2035 endfib = rt_numfibs - 1;
2038 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
2044 * If it's a delete, check that if it exists,
2045 * it's on the correct interface or we might scrub
2046 * a route to another ifa which would
2047 * be confusing at best and possibly worse.
2049 if (cmd == RTM_DELETE) {
2051 * It's a delete, so it should already exist..
2052 * If it's a net, mask off the host bits
2053 * (Assuming we have a mask)
2054 * XXX this is kinda inet specific..
2056 if (netmask != NULL) {
2057 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
2058 dst = (struct sockaddr *)tempbuf;
2060 } else if (cmd == RTM_ADD) {
2061 sdl = (struct sockaddr_dl *)tempbuf;
2062 bzero(sdl, sizeof(struct sockaddr_dl));
2063 sdl->sdl_family = AF_LINK;
2064 sdl->sdl_len = sizeof(struct sockaddr_dl);
2065 sdl->sdl_type = ifa->ifa_ifp->if_type;
2066 sdl->sdl_index = ifa->ifa_ifp->if_index;
2069 * Now go through all the requested tables (fibs) and do the
2070 * requested action. Realistically, this will either be fib 0
2071 * for protocols that don't do multiple tables or all the
2072 * tables for those that do.
2074 for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
2075 if (cmd == RTM_DELETE) {
2076 struct radix_node *rn;
2078 * Look up an rtentry that is in the routing tree and
2079 * contains the correct info.
2081 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
2083 /* this table doesn't exist but others might */
2086 rn = rnh->rnh_lookup(dst, netmask, &rnh->head);
2088 if (rt_mpath_capable(rnh)) {
2095 * for interface route the
2096 * rt->rt_gateway is sockaddr_intf
2097 * for cloning ARP entries, so
2098 * rt_mpath_matchgate must use the
2101 rt = rt_mpath_matchgate(rt,
2108 error = (rn == NULL ||
2109 (rn->rn_flags & RNF_ROOT) ||
2110 RNTORT(rn)->rt_ifa != ifa);
2113 /* this is only an error if bad on ALL tables */
2118 * Do the actual request
2120 bzero((caddr_t)&info, sizeof(info));
2122 info.rti_flags = flags |
2123 (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED;
2124 info.rti_info[RTAX_DST] = dst;
2126 * doing this for compatibility reasons
2129 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)sdl;
2131 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
2132 info.rti_info[RTAX_NETMASK] = netmask;
2133 error = rtrequest1_fib(cmd, &info, &rt, fibnum);
2134 if (error == 0 && rt != NULL) {
2136 * notify any listening routing agents of the change
2141 * in case address alias finds the first address
2142 * e.g. ifconfig bge0 192.0.2.246/24
2143 * e.g. ifconfig bge0 192.0.2.247/24
2144 * the address set in the route is 192.0.2.246
2145 * so we need to replace it with 192.0.2.247
2147 if (memcmp(rt->rt_ifa->ifa_addr,
2148 ifa->ifa_addr, ifa->ifa_addr->sa_len)) {
2149 ifa_free(rt->rt_ifa);
2151 rt->rt_ifp = ifa->ifa_ifp;
2157 rt_newaddrmsg_fib(cmd, ifa, rt, fibnum);
2160 if (cmd == RTM_DELETE) {
2162 * If we are deleting, and we found an entry,
2163 * then it's been removed from the tree..
2164 * now throw it away.
2168 if (cmd == RTM_ADD) {
2170 * We just wanted to add it..
2171 * we don't actually need a reference.
2182 if (cmd == RTM_DELETE) {
2186 /* we only give an error if it wasn't in any table */
2187 error = ((flags & RTF_HOST) ?
2188 EHOSTUNREACH : ENETUNREACH);
2192 /* return an error if any of them failed */
2200 * Set up a routing table entry, normally
2204 rtinit(struct ifaddr *ifa, int cmd, int flags)
2206 struct sockaddr *dst;
2207 int fib = RT_DEFAULT_FIB;
2209 if (flags & RTF_HOST) {
2210 dst = ifa->ifa_dstaddr;
2212 dst = ifa->ifa_addr;
2215 switch (dst->sa_family) {
2218 /* We do support multiple FIBs. */
2222 return (rtinit1(ifa, cmd, flags, fib));
2226 * Announce interface address arrival/withdraw
2227 * Returns 0 on success.
2230 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
2233 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2234 ("unexpected cmd %d", cmd));
2235 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2236 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2238 EVENTHANDLER_DIRECT_INVOKE(rt_addrmsg, ifa, cmd);
2239 return (rtsock_addrmsg(cmd, ifa, fibnum));
2243 * Announce kernel-originated route addition/removal to rtsock based on @rt data.
2245 * @rt: valid rtentry
2246 * @ifp: target route interface
2247 * @fibnum: fib id or RT_ALL_FIBS
2249 * Returns 0 on success.
2252 rt_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs,
2256 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2257 ("unexpected cmd %d", cmd));
2259 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2260 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2262 KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
2264 return (rtsock_routemsg(cmd, rt, ifp, 0, fibnum));
2268 * Announce kernel-originated route addition/removal to rtsock based on @rt data.
2270 * @info: addrinfo structure with valid data.
2271 * @fibnum: fib id or RT_ALL_FIBS
2273 * Returns 0 on success.
2276 rt_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
2279 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE || cmd == RTM_CHANGE,
2280 ("unexpected cmd %d", cmd));
2282 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2283 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2285 KASSERT(info->rti_info[RTAX_DST] != NULL, (":%s: RTAX_DST must be supplied", __func__));
2287 return (rtsock_routemsg_info(cmd, info, fibnum));
2292 * This is called to generate messages from the routing socket
2293 * indicating a network interface has had addresses associated with it.
2296 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, struct rtentry *rt, int fibnum)
2299 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2300 ("unexpected cmd %u", cmd));
2301 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2302 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2304 if (cmd == RTM_ADD) {
2305 rt_addrmsg(cmd, ifa, fibnum);
2307 rt_routemsg(cmd, rt, ifa->ifa_ifp, 0, fibnum);
2310 rt_routemsg(cmd, rt, ifa->ifa_ifp, 0, fibnum);
2311 rt_addrmsg(cmd, ifa, fibnum);