2 * Copyright (c) 1980, 1986, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)route.c 8.3.1.1 (Berkeley) 2/23/95
32 /************************************************************************
33 * Note: In this file a 'fib' is a "forwarding information base" *
34 * Which is the new name for an in kernel routing (next hop) table. *
35 ***********************************************************************/
38 #include "opt_inet6.h"
39 #include "opt_route.h"
41 #include "opt_mrouting.h"
42 #include "opt_mpath.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/malloc.h>
48 #include <sys/socket.h>
49 #include <sys/sysctl.h>
50 #include <sys/syslog.h>
51 #include <sys/sysproto.h>
53 #include <sys/domain.h>
54 #include <sys/kernel.h>
57 #include <net/if_var.h>
58 #include <net/if_dl.h>
59 #include <net/route.h>
60 #include <net/route_var.h>
62 #include <net/flowtable.h>
65 #include <net/radix_mpath.h>
68 #include <netinet/in.h>
69 #include <netinet/ip_mroute.h>
73 #define RT_MAXFIBS UINT16_MAX
75 /* Kernel config default option. */
78 #error "ROUTETABLES defined too low"
80 #if ROUTETABLES > RT_MAXFIBS
81 #error "ROUTETABLES defined too big"
83 #define RT_NUMFIBS ROUTETABLES
84 #endif /* ROUTETABLES */
85 /* Initialize to default if not otherwise set. */
90 #if defined(INET) || defined(INET6)
92 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
97 /* This is read-only.. */
98 u_int rt_numfibs = RT_NUMFIBS;
99 SYSCTL_UINT(_net, OID_AUTO, fibs, CTLFLAG_RDTUN, &rt_numfibs, 0, "");
102 * By default add routes to all fibs for new interfaces.
103 * Once this is set to 0 then only allocate routes on interface
104 * changes for the FIB of the caller when adding a new set of addresses
105 * to an interface. XXX this is a shotgun aproach to a problem that needs
106 * a more fine grained solution.. that will come.
107 * XXX also has the problems getting the FIB from curthread which will not
108 * always work given the fib can be overridden and prefixes can be added
109 * from the network stack context.
111 VNET_DEFINE(u_int, rt_add_addr_allfibs) = 1;
112 SYSCTL_UINT(_net, OID_AUTO, add_addr_allfibs, CTLFLAG_RWTUN | CTLFLAG_VNET,
113 &VNET_NAME(rt_add_addr_allfibs), 0, "");
115 VNET_DEFINE(struct rtstat, rtstat);
116 #define V_rtstat VNET(rtstat)
118 VNET_DEFINE(struct rib_head *, rt_tables);
119 #define V_rt_tables VNET(rt_tables)
121 VNET_DEFINE(int, rttrash); /* routes not in table but not freed */
122 #define V_rttrash VNET(rttrash)
126 * Convert a 'struct radix_node *' to a 'struct rtentry *'.
127 * The operation can be done safely (in this code) because a
128 * 'struct rtentry' starts with two 'struct radix_node''s, the first
129 * one representing leaf nodes in the routing tree, which is
130 * what the code in radix.c passes us as a 'struct radix_node'.
132 * But because there are a lot of assumptions in this conversion,
133 * do not cast explicitly, but always use the macro below.
135 #define RNTORT(p) ((struct rtentry *)(p))
137 static VNET_DEFINE(uma_zone_t, rtzone); /* Routing table UMA zone. */
138 #define V_rtzone VNET(rtzone)
140 static int rtrequest1_fib_change(struct rib_head *, struct rt_addrinfo *,
141 struct rtentry **, u_int);
142 static void rt_setmetrics(const struct rt_addrinfo *, struct rtentry *);
143 static int rt_ifdelroute(const struct rtentry *rt, void *arg);
144 static struct rtentry *rt_unlinkrte(struct rib_head *rnh,
145 struct rt_addrinfo *info, int *perror);
146 static void rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info);
148 static struct radix_node *rt_mpath_unlink(struct rib_head *rnh,
149 struct rt_addrinfo *info, struct rtentry *rto, int *perror);
151 static int rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info,
160 static int if_updatemtu_cb(struct radix_node *, void *);
163 * handler for net.my_fibnum
166 sysctl_my_fibnum(SYSCTL_HANDLER_ARGS)
171 fibnum = curthread->td_proc->p_fibnum;
172 error = sysctl_handle_int(oidp, &fibnum, 0, req);
176 SYSCTL_PROC(_net, OID_AUTO, my_fibnum, CTLTYPE_INT|CTLFLAG_RD,
177 NULL, 0, &sysctl_my_fibnum, "I", "default FIB of caller");
179 static __inline struct rib_head **
180 rt_tables_get_rnh_ptr(int table, int fam)
182 struct rib_head **rnh;
184 KASSERT(table >= 0 && table < rt_numfibs, ("%s: table out of bounds.",
186 KASSERT(fam >= 0 && fam < (AF_MAX+1), ("%s: fam out of bounds.",
189 /* rnh is [fib=0][af=0]. */
190 rnh = (struct rib_head **)V_rt_tables;
191 /* Get the offset to the requested table and fam. */
192 rnh += table * (AF_MAX+1) + fam;
198 rt_tables_get_rnh(int table, int fam)
201 return (*rt_tables_get_rnh_ptr(table, fam));
205 rt_tables_get_gen(int table, int fam)
207 struct rib_head *rnh;
209 rnh = *rt_tables_get_rnh_ptr(table, fam);
210 return (rnh->rnh_gen);
215 * route initialization must occur before ip6_init2(), which happenas at
222 /* whack the tunable ints into line. */
223 if (rt_numfibs > RT_MAXFIBS)
224 rt_numfibs = RT_MAXFIBS;
228 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0);
231 rtentry_zinit(void *mem, int size, int how)
233 struct rtentry *rt = mem;
235 rt->rt_pksent = counter_u64_alloc(how);
236 if (rt->rt_pksent == NULL)
245 rtentry_zfini(void *mem, int size)
247 struct rtentry *rt = mem;
250 counter_u64_free(rt->rt_pksent);
254 rtentry_ctor(void *mem, int size, void *arg, int how)
256 struct rtentry *rt = mem;
258 bzero(rt, offsetof(struct rtentry, rt_endzero));
259 counter_u64_zero(rt->rt_pksent);
266 rtentry_dtor(void *mem, int size, void *arg)
268 struct rtentry *rt = mem;
274 vnet_route_init(const void *unused __unused)
277 struct rib_head **rnh;
281 V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) *
282 sizeof(struct rib_head *), M_RTABLE, M_WAITOK|M_ZERO);
284 V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry),
285 rtentry_ctor, rtentry_dtor,
286 rtentry_zinit, rtentry_zfini, UMA_ALIGN_PTR, 0);
287 for (dom = domains; dom; dom = dom->dom_next) {
288 if (dom->dom_rtattach == NULL)
291 for (table = 0; table < rt_numfibs; table++) {
292 fam = dom->dom_family;
293 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
296 rnh = rt_tables_get_rnh_ptr(table, fam);
298 panic("%s: rnh NULL", __func__);
299 dom->dom_rtattach((void **)rnh, 0);
303 VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
308 vnet_route_uninit(const void *unused __unused)
313 struct rib_head **rnh;
315 for (dom = domains; dom; dom = dom->dom_next) {
316 if (dom->dom_rtdetach == NULL)
319 for (table = 0; table < rt_numfibs; table++) {
320 fam = dom->dom_family;
322 if (table != 0 && fam != AF_INET6 && fam != AF_INET)
325 rnh = rt_tables_get_rnh_ptr(table, fam);
327 panic("%s: rnh NULL", __func__);
328 dom->dom_rtdetach((void **)rnh, 0);
332 free(V_rt_tables, M_RTABLE);
333 uma_zdestroy(V_rtzone);
335 VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
336 vnet_route_uninit, 0);
340 rt_table_init(int offset)
344 rh = malloc(sizeof(struct rib_head), M_RTABLE, M_WAITOK | M_ZERO);
346 /* TODO: These details should be hidded inside radix.c */
347 /* Init masks tree */
348 rn_inithead_internal(&rh->head, rh->rnh_nodes, offset);
349 rn_inithead_internal(&rh->rmhead.head, rh->rmhead.mask_nodes, 0);
350 rh->head.rnh_masks = &rh->rmhead;
353 rw_init(&rh->rib_lock, "rib head lock");
355 /* Finally, set base callbacks */
356 rh->rnh_addaddr = rn_addroute;
357 rh->rnh_deladdr = rn_delete;
358 rh->rnh_matchaddr = rn_match;
359 rh->rnh_lookup = rn_lookup;
360 rh->rnh_walktree = rn_walktree;
361 rh->rnh_walktree_from = rn_walktree_from;
367 rt_freeentry(struct radix_node *rn, void *arg)
369 struct radix_head * const rnh = arg;
370 struct radix_node *x;
372 x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh);
379 rt_table_destroy(struct rib_head *rh)
382 rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head);
384 /* Assume table is already empty */
385 rw_destroy(&rh->rib_lock);
390 #ifndef _SYS_SYSPROTO_H_
396 sys_setfib(struct thread *td, struct setfib_args *uap)
398 if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs)
400 td->td_proc->p_fibnum = uap->fibnum;
405 * Packet routing routines.
408 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum)
412 if ((rt = ro->ro_rt) != NULL) {
413 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
418 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum);
420 RT_UNLOCK(ro->ro_rt);
424 * Look up the route that matches the address given
425 * Or, at least try.. Create a cloned route if needed.
427 * The returned route, if any, is locked.
430 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
433 return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB));
437 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags,
441 struct radix_node *rn;
442 struct rtentry *newrt;
443 struct rt_addrinfo info;
444 int err = 0, msgtype = RTM_MISS;
446 KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum"));
447 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
453 * Look up the address in the table for that Address Family
456 rn = rh->rnh_matchaddr(dst, &rh->head);
457 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
468 * Either we hit the root or couldn't find any match,
469 * Which basically means
470 * "caint get there frm here"
473 V_rtstat.rts_unreach++;
477 * If required, report the failure to the supervising
479 * For a delete, this is not an error. (report == 0)
481 bzero(&info, sizeof(info));
482 info.rti_info[RTAX_DST] = dst;
483 rt_missmsg_fib(msgtype, &info, 0, err, fibnum);
489 * Remove a reference count from an rtentry.
490 * If the count gets low enough, take it out of the routing table
493 rtfree(struct rtentry *rt)
495 struct rib_head *rnh;
497 KASSERT(rt != NULL,("%s: NULL rt", __func__));
498 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
499 KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
504 * The callers should use RTFREE_LOCKED() or RTFREE(), so
505 * we should come here exactly with the last reference.
508 if (rt->rt_refcnt > 0) {
509 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt);
514 * On last reference give the "close method" a chance
515 * to cleanup private state. This also permits (for
516 * IPv4 and IPv6) a chance to decide if the routing table
517 * entry should be purged immediately or at a later time.
518 * When an immediate purge is to happen the close routine
519 * typically calls rtexpunge which clears the RTF_UP flag
520 * on the entry so that the code below reclaims the storage.
522 if (rt->rt_refcnt == 0 && rnh->rnh_close)
523 rnh->rnh_close((struct radix_node *)rt, &rnh->head);
526 * If we are no longer "up" (and ref == 0)
527 * then we can free the resources associated
530 if ((rt->rt_flags & RTF_UP) == 0) {
531 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
534 * the rtentry must have been removed from the routing table
535 * so it is represented in rttrash.. remove that now.
539 if (rt->rt_refcnt < 0) {
540 printf("rtfree: %p not freed (neg refs)\n", rt);
545 * release references on items we hold them on..
546 * e.g other routes and ifaddrs.
549 ifa_free(rt->rt_ifa);
551 * The key is separatly alloc'd so free it (see rt_setgate()).
552 * This also frees the gateway, as they are always malloc'd
558 * and the rtentry itself of course
560 uma_zfree(V_rtzone, rt);
569 * Force a routing table entry to the specified
570 * destination to go through the given gateway.
571 * Normally called as a result of a routing redirect
572 * message from the network layer.
575 rtredirect_fib(struct sockaddr *dst,
576 struct sockaddr *gateway,
577 struct sockaddr *netmask,
579 struct sockaddr *src,
585 struct rt_addrinfo info;
587 struct rib_head *rnh;
590 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
592 error = EAFNOSUPPORT;
596 /* verify the gateway is directly reachable */
597 if ((ifa = ifa_ifwithnet(gateway, 0, fibnum)) == NULL) {
601 rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */
603 * If the redirect isn't from our current router for this dst,
604 * it's either old or wrong. If it redirects us to ourselves,
605 * we have a routing loop, perhaps as a result of an interface
606 * going down recently.
608 if (!(flags & RTF_DONE) && rt) {
609 if (!sa_equal(src, rt->rt_gateway)) {
613 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) {
618 if ((flags & RTF_GATEWAY) && ifa_ifwithaddr_check(gateway)) {
619 error = EHOSTUNREACH;
623 * Create a new entry if we just got back a wildcard entry
624 * or the lookup failed. This is necessary for hosts
625 * which use routing redirects generated by smart gateways
626 * to dynamically build the routing tables.
628 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
631 * Don't listen to the redirect if it's
632 * for a route to an interface.
634 if (rt->rt_flags & RTF_GATEWAY) {
635 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
637 * Changing from route to net => route to host.
638 * Create new route, rather than smashing route to net.
644 flags |= RTF_DYNAMIC;
645 bzero((caddr_t)&info, sizeof(info));
646 info.rti_info[RTAX_DST] = dst;
647 info.rti_info[RTAX_GATEWAY] = gateway;
648 info.rti_info[RTAX_NETMASK] = netmask;
650 info.rti_flags = flags;
651 error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum);
654 flags = rt->rt_flags;
657 stat = &V_rtstat.rts_dynamic;
661 * Smash the current notion of the gateway to
662 * this destination. Should check about netmask!!!
664 if ((flags & RTF_GATEWAY) == 0)
665 rt->rt_flags &= ~RTF_GATEWAY;
666 rt->rt_flags |= RTF_MODIFIED;
667 flags |= RTF_MODIFIED;
668 stat = &V_rtstat.rts_newgateway;
670 * add the key and gateway (in one malloc'd chunk).
675 rt_setgate(rt, rt_key(rt), gateway);
679 error = EHOSTUNREACH;
685 V_rtstat.rts_badredirect++;
686 else if (stat != NULL)
688 bzero((caddr_t)&info, sizeof(info));
689 info.rti_info[RTAX_DST] = dst;
690 info.rti_info[RTAX_GATEWAY] = gateway;
691 info.rti_info[RTAX_NETMASK] = netmask;
692 info.rti_info[RTAX_AUTHOR] = src;
693 rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum);
699 * Routing table ioctl interface.
702 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
706 * If more ioctl commands are added here, make sure the proper
707 * super-user checks are being performed because it is possible for
708 * prison-root to make it this far if raw sockets have been enabled
712 /* Multicast goop, grrr... */
713 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
720 ifa_ifwithroute(int flags, const struct sockaddr *dst, struct sockaddr *gateway,
726 if ((flags & RTF_GATEWAY) == 0) {
728 * If we are adding a route to an interface,
729 * and the interface is a pt to pt link
730 * we should search for the destination
731 * as our clue to the interface. Otherwise
732 * we can use the local address.
735 if (flags & RTF_HOST)
736 ifa = ifa_ifwithdstaddr(dst, fibnum);
738 ifa = ifa_ifwithaddr(gateway);
741 * If we are adding a route to a remote net
742 * or host, the gateway may still be on the
743 * other end of a pt to pt link.
745 ifa = ifa_ifwithdstaddr(gateway, fibnum);
748 ifa = ifa_ifwithnet(gateway, 0, fibnum);
750 struct rtentry *rt = rtalloc1_fib(gateway, 0, 0, fibnum);
754 * dismiss a gateway that is reachable only
755 * through the default router
757 switch (gateway->sa_family) {
759 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
763 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
769 if (!not_found && rt->rt_ifa != NULL) {
775 if (not_found || ifa == NULL)
778 if (ifa->ifa_addr->sa_family != dst->sa_family) {
779 struct ifaddr *oifa = ifa;
780 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
790 * Do appropriate manipulations of a routing tree given
791 * all the bits of info needed
794 rtrequest_fib(int req,
795 struct sockaddr *dst,
796 struct sockaddr *gateway,
797 struct sockaddr *netmask,
799 struct rtentry **ret_nrt,
802 struct rt_addrinfo info;
804 if (dst->sa_len == 0)
807 bzero((caddr_t)&info, sizeof(info));
808 info.rti_flags = flags;
809 info.rti_info[RTAX_DST] = dst;
810 info.rti_info[RTAX_GATEWAY] = gateway;
811 info.rti_info[RTAX_NETMASK] = netmask;
812 return rtrequest1_fib(req, &info, ret_nrt, fibnum);
817 * Copy most of @rt data into @info.
819 * If @flags contains NHR_COPY, copies dst,netmask and gw to the
820 * pointers specified by @info structure. Assume such pointers
821 * are zeroed sockaddr-like structures with sa_len field initialized
822 * to reflect size of the provided buffer. if no NHR_COPY is specified,
823 * point dst,netmask and gw @info fields to appropriate @rt values.
825 * if @flags contains NHR_REF, do refcouting on rt_ifp.
827 * Returns 0 on success.
830 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags)
832 struct rt_metrics *rmx;
833 struct sockaddr *src, *dst;
836 if (flags & NHR_COPY) {
837 /* Copy destination if dst is non-zero */
839 dst = info->rti_info[RTAX_DST];
840 sa_len = src->sa_len;
842 if (src->sa_len > dst->sa_len)
844 memcpy(dst, src, src->sa_len);
845 info->rti_addrs |= RTA_DST;
848 /* Copy mask if set && dst is non-zero */
850 dst = info->rti_info[RTAX_NETMASK];
851 if (src != NULL && dst != NULL) {
854 * Radix stores different value in sa_len,
855 * assume rt_mask() to have the same length
858 if (sa_len > dst->sa_len)
860 memcpy(dst, src, src->sa_len);
861 info->rti_addrs |= RTA_NETMASK;
864 /* Copy gateway is set && dst is non-zero */
865 src = rt->rt_gateway;
866 dst = info->rti_info[RTAX_GATEWAY];
867 if ((rt->rt_flags & RTF_GATEWAY) && src != NULL && dst != NULL){
868 if (src->sa_len > dst->sa_len)
870 memcpy(dst, src, src->sa_len);
871 info->rti_addrs |= RTA_GATEWAY;
874 info->rti_info[RTAX_DST] = rt_key(rt);
875 info->rti_addrs |= RTA_DST;
876 if (rt_mask(rt) != NULL) {
877 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
878 info->rti_addrs |= RTA_NETMASK;
880 if (rt->rt_flags & RTF_GATEWAY) {
881 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
882 info->rti_addrs |= RTA_GATEWAY;
888 info->rti_mflags |= RTV_MTU;
889 rmx->rmx_mtu = rt->rt_mtu;
892 info->rti_flags = rt->rt_flags;
893 info->rti_ifp = rt->rt_ifp;
894 info->rti_ifa = rt->rt_ifa;
896 if (flags & NHR_REF) {
897 /* Do 'traditional' refcouting */
898 if_ref(info->rti_ifp);
905 * Lookups up route entry for @dst in RIB database for fib @fibnum.
906 * Exports entry data to @info using rt_exportinfo().
908 * if @flags contains NHR_REF, refcouting is performed on rt_ifp.
909 * All references can be released later by calling rib_free_info()
911 * Returns 0 on success.
912 * Returns ENOENT for lookup failure, ENOMEM for export failure.
915 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
916 uint32_t flowid, struct rt_addrinfo *info)
919 struct radix_node *rn;
923 KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
924 rh = rt_tables_get_rnh(fibnum, dst->sa_family);
929 rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head);
930 if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
932 /* Ensure route & ifp is UP */
933 if (RT_LINK_IS_UP(rt->rt_ifp)) {
934 flags = (flags & NHR_REF) | NHR_COPY;
935 error = rt_exportinfo(rt, info, flags);
947 * Releases all references acquired by rib_lookup_info() when
948 * called with NHR_REF flags.
951 rib_free_info(struct rt_addrinfo *info)
954 if_rele(info->rti_ifp);
958 * Iterates over all existing fibs in system calling
959 * @setwa_f function prior to traversing each fib.
960 * Calls @wa_f function for each element in current fib.
961 * If af is not AF_UNSPEC, iterates over fibs in particular
965 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f,
968 struct rib_head *rnh;
972 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
973 /* Do we want some specific family? */
974 if (af != AF_UNSPEC) {
975 rnh = rt_tables_get_rnh(fibnum, af);
979 setwa_f(rnh, fibnum, af, arg);
982 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
987 for (i = 1; i <= AF_MAX; i++) {
988 rnh = rt_tables_get_rnh(fibnum, i);
992 setwa_f(rnh, fibnum, i, arg);
995 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
1003 struct rt_addrinfo info;
1004 struct rib_head *rnh;
1005 struct rtentry *head;
1009 * Conditionally unlinks @rn from radix tree based
1010 * on info data passed in @arg.
1013 rt_checkdelroute(struct radix_node *rn, void *arg)
1015 struct rt_delinfo *di;
1016 struct rt_addrinfo *info;
1020 di = (struct rt_delinfo *)arg;
1021 rt = (struct rtentry *)rn;
1025 info->rti_info[RTAX_DST] = rt_key(rt);
1026 info->rti_info[RTAX_NETMASK] = rt_mask(rt);
1027 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1029 rt = rt_unlinkrte(di->rnh, info, &error);
1031 /* Either not allowed or not matched. Skip entry */
1035 /* Entry was unlinked. Add to the list and return */
1036 rt->rt_chain = di->head;
1043 * Iterates over all existing fibs in system.
1044 * Deletes each element for which @filter_f function returned
1046 * If @af is not AF_UNSPEC, iterates over fibs in particular
1050 rt_foreach_fib_walk_del(int af, rt_filter_f_t *filter_f, void *arg)
1052 struct rib_head *rnh;
1053 struct rt_delinfo di;
1058 bzero(&di, sizeof(di));
1059 di.info.rti_filter = filter_f;
1060 di.info.rti_filterdata = arg;
1062 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1063 /* Do we want some specific family? */
1064 if (af != AF_UNSPEC) {
1072 for (i = start; i <= end; i++) {
1073 rnh = rt_tables_get_rnh(fibnum, i);
1079 rnh->rnh_walktree(&rnh->head, rt_checkdelroute, &di);
1082 if (di.head == NULL)
1085 /* We might have something to reclaim */
1086 while (di.head != NULL) {
1088 di.head = rt->rt_chain;
1089 rt->rt_chain = NULL;
1091 /* TODO std rt -> rt_addrinfo export */
1092 di.info.rti_info[RTAX_DST] = rt_key(rt);
1093 di.info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1095 rt_notifydelete(rt, &di.info);
1104 * Delete Routes for a Network Interface
1106 * Called for each routing entry via the rnh->rnh_walktree() call above
1107 * to delete all route entries referencing a detaching network interface.
1110 * rt pointer to rtentry
1111 * arg argument passed to rnh->rnh_walktree() - detaching interface
1115 * errno failed - reason indicated
1118 rt_ifdelroute(const struct rtentry *rt, void *arg)
1120 struct ifnet *ifp = arg;
1122 if (rt->rt_ifp != ifp)
1126 * Protect (sorta) against walktree recursion problems
1127 * with cloned routes
1129 if ((rt->rt_flags & RTF_UP) == 0)
1136 * Delete all remaining routes using this interface
1137 * Unfortuneatly the only way to do this is to slog through
1138 * the entire routing table looking for routes which point
1139 * to this interface...oh well...
1142 rt_flushifroutes(struct ifnet *ifp)
1145 rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
1149 * Conditionally unlinks rtentry matching data inside @info from @rnh.
1150 * Returns unlinked, locked and referenced @rtentry on success,
1151 * Returns NULL and sets @perror to:
1152 * ESRCH - if prefix was not found,
1153 * EADDRINUSE - if trying to delete PINNED route without appropriate flag.
1154 * ENOENT - if supplied filter function returned 0 (not matched).
1156 static struct rtentry *
1157 rt_unlinkrte(struct rib_head *rnh, struct rt_addrinfo *info, int *perror)
1159 struct sockaddr *dst, *netmask;
1161 struct radix_node *rn;
1163 dst = info->rti_info[RTAX_DST];
1164 netmask = info->rti_info[RTAX_NETMASK];
1166 rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, &rnh->head);
1172 if ((info->rti_flags & RTF_PINNED) == 0) {
1173 /* Check if target route can be deleted */
1174 if (rt->rt_flags & RTF_PINNED) {
1175 *perror = EADDRINUSE;
1180 if (info->rti_filter != NULL) {
1181 if (info->rti_filter(rt, info->rti_filterdata) == 0) {
1188 * Filter function requested rte deletion.
1189 * Ease the caller work by filling in remaining info
1190 * from that particular entry.
1192 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1196 * Remove the item from the tree and return it.
1197 * Complain if it is not there and do no more processing.
1201 if (rt_mpath_capable(rnh))
1202 rn = rt_mpath_unlink(rnh, info, rt, perror);
1205 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1209 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
1210 panic ("rtrequest delete");
1215 rt->rt_flags &= ~RTF_UP;
1223 rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info)
1228 * give the protocol a chance to keep things in sync.
1231 if (ifa != NULL && ifa->ifa_rtrequest != NULL)
1232 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1235 * One more rtentry floating around that is not
1236 * linked to the routing table. rttrash will be decremented
1237 * when RTFREE(rt) is eventually called.
1244 * These (questionable) definitions of apparent local variables apply
1245 * to the next two functions. XXXXXX!!!
1247 #define dst info->rti_info[RTAX_DST]
1248 #define gateway info->rti_info[RTAX_GATEWAY]
1249 #define netmask info->rti_info[RTAX_NETMASK]
1250 #define ifaaddr info->rti_info[RTAX_IFA]
1251 #define ifpaddr info->rti_info[RTAX_IFP]
1252 #define flags info->rti_flags
1255 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined,
1256 * it will be referenced so the caller must free it.
1259 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
1265 * ifp may be specified by sockaddr_dl
1266 * when protocol address is ambiguous.
1268 if (info->rti_ifp == NULL && ifpaddr != NULL &&
1269 ifpaddr->sa_family == AF_LINK &&
1270 (ifa = ifa_ifwithnet(ifpaddr, 0, fibnum)) != NULL) {
1271 info->rti_ifp = ifa->ifa_ifp;
1274 if (info->rti_ifa == NULL && ifaaddr != NULL)
1275 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
1276 if (info->rti_ifa == NULL) {
1277 struct sockaddr *sa;
1279 sa = ifaaddr != NULL ? ifaaddr :
1280 (gateway != NULL ? gateway : dst);
1281 if (sa != NULL && info->rti_ifp != NULL)
1282 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
1283 else if (dst != NULL && gateway != NULL)
1284 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway,
1286 else if (sa != NULL)
1287 info->rti_ifa = ifa_ifwithroute(flags, sa, sa,
1290 if ((ifa = info->rti_ifa) != NULL) {
1291 if (info->rti_ifp == NULL)
1292 info->rti_ifp = ifa->ifa_ifp;
1294 error = ENETUNREACH;
1299 if_updatemtu_cb(struct radix_node *rn, void *arg)
1302 struct if_mtuinfo *ifmtu;
1304 rt = (struct rtentry *)rn;
1305 ifmtu = (struct if_mtuinfo *)arg;
1307 if (rt->rt_ifp != ifmtu->ifp)
1310 if (rt->rt_mtu >= ifmtu->mtu) {
1311 /* We have to decrease mtu regardless of flags */
1312 rt->rt_mtu = ifmtu->mtu;
1317 * New MTU is bigger. Check if are allowed to alter it
1319 if ((rt->rt_flags & (RTF_FIXEDMTU | RTF_GATEWAY | RTF_HOST)) != 0) {
1322 * Skip routes with user-supplied MTU and
1323 * non-interface routes
1328 /* We are safe to update route MTU */
1329 rt->rt_mtu = ifmtu->mtu;
1335 rt_updatemtu(struct ifnet *ifp)
1337 struct if_mtuinfo ifmtu;
1338 struct rib_head *rnh;
1344 * Try to update rt_mtu for all routes using this interface
1345 * Unfortunately the only way to do this is to traverse all
1346 * routing tables in all fibs/domains.
1348 for (i = 1; i <= AF_MAX; i++) {
1349 ifmtu.mtu = if_getmtu_family(ifp, i);
1350 for (j = 0; j < rt_numfibs; j++) {
1351 rnh = rt_tables_get_rnh(j, i);
1355 rnh->rnh_walktree(&rnh->head, if_updatemtu_cb, &ifmtu);
1363 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
1364 int rt_print(char *buf, int buflen, struct rtentry *rt);
1367 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
1371 switch (s->sa_family) {
1373 paddr = &((struct sockaddr_in *)s)->sin_addr;
1376 paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
1383 if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
1386 return (strlen(buf));
1390 rt_print(char *buf, int buflen, struct rtentry *rt)
1392 struct sockaddr *addr, *mask;
1398 i = p_sockaddr(buf, buflen, addr);
1399 if (!(rt->rt_flags & RTF_HOST)) {
1401 i += p_sockaddr(buf + i, buflen - i, mask);
1404 if (rt->rt_flags & RTF_GATEWAY) {
1406 i += p_sockaddr(buf + i, buflen - i, rt->rt_gateway);
1415 * Deletes key for single-path routes, unlinks rtentry with
1416 * gateway specified in @info from multi-path routes.
1418 * Returnes unlinked entry. In case of failure, returns NULL
1419 * and sets @perror to ESRCH.
1421 static struct radix_node *
1422 rt_mpath_unlink(struct rib_head *rnh, struct rt_addrinfo *info,
1423 struct rtentry *rto, int *perror)
1426 * if we got multipath routes, we require users to specify
1427 * a matching RTAX_GATEWAY.
1429 struct rtentry *rt; // *rto = NULL;
1430 struct radix_node *rn;
1431 struct sockaddr *gw;
1433 gw = info->rti_info[RTAX_GATEWAY];
1434 rt = rt_mpath_matchgate(rto, gw);
1441 * this is the first entry in the chain
1444 rn = rn_mpath_next((struct radix_node *)rt);
1446 * there is another entry, now it's active
1451 rto->rt_flags |= RTF_UP;
1453 } else if (rt->rt_flags & RTF_GATEWAY) {
1455 * For gateway routes, we need to
1456 * make sure that we we are deleting
1457 * the correct gateway.
1458 * rt_mpath_matchgate() does not
1459 * check the case when there is only
1460 * one route in the chain.
1463 (rt->rt_gateway->sa_len != gw->sa_len ||
1464 memcmp(rt->rt_gateway, gw, gw->sa_len))) {
1471 * use the normal delete code to remove
1474 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
1480 * if the entry is 2nd and on up
1482 if (rt_mpath_deldup(rto, rt) == 0)
1483 panic ("rtrequest1: rt_mpath_deldup");
1485 rn = (struct radix_node *)rt;
1491 static struct rtentry *
1492 rt_flowtable_check_route(struct rib_head *rnh, struct rt_addrinfo *info)
1494 #if defined(INET6) || defined(INET)
1495 struct radix_node *rn;
1497 struct rtentry *rt0;
1500 /* "flow-table" only supports IPv6 and IPv4 at the moment. */
1501 switch (dst->sa_family) {
1508 #if defined(INET6) || defined(INET)
1509 rn = rnh->rnh_matchaddr(dst, &rnh->head);
1510 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
1511 struct sockaddr *mask;
1516 * compare mask to see if the new route is
1517 * more specific than the existing one
1524 * A host route is already present, so
1525 * leave the flow-table entries as is.
1527 if (rt0->rt_flags & RTF_HOST) {
1530 } else if (!(flags & RTF_HOST) && netmask) {
1531 mask = rt_mask(rt0);
1534 n = (u_char *)netmask;
1541 if (len == 0 || (*n < *m)) {
1547 #endif/* INET6 || INET */
1555 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt,
1559 struct rtentry *rt, *rt_old;
1561 struct rtentry *rt0;
1563 struct radix_node *rn;
1564 struct rib_head *rnh;
1566 struct sockaddr *ndst;
1567 struct sockaddr_storage mdst;
1569 KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum"));
1570 KASSERT((flags & RTF_RNH_LOCKED) == 0, ("rtrequest1_fib: locked"));
1571 switch (dst->sa_family) {
1574 /* We support multiple FIBs. */
1577 fibnum = RT_DEFAULT_FIB;
1582 * Find the correct routing tree to use for this Address Family
1584 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
1586 return (EAFNOSUPPORT);
1589 * If we are adding a host route then we don't want to put
1590 * a netmask in the tree, nor do we want to clone it.
1592 if (flags & RTF_HOST)
1598 rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask);
1599 dst = (struct sockaddr *)&mdst;
1603 rt = rt_unlinkrte(rnh, info, &error);
1608 rt_notifydelete(rt, info);
1611 * If the caller wants it, then it can have it,
1612 * but it's up to it to free the rtentry as we won't be
1623 * resolve was only used for route cloning
1628 if ((flags & RTF_GATEWAY) && !gateway)
1630 if (dst && gateway && (dst->sa_family != gateway->sa_family) &&
1631 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
1634 if (info->rti_ifa == NULL) {
1635 error = rt_getifa_fib(info, fibnum);
1639 ifa_ref(info->rti_ifa);
1640 ifa = info->rti_ifa;
1641 rt = uma_zalloc(V_rtzone, M_NOWAIT);
1646 rt->rt_flags = RTF_UP | flags;
1647 rt->rt_fibnum = fibnum;
1649 * Add the gateway. Possibly re-malloc-ing the storage for it.
1651 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
1653 uma_zfree(V_rtzone, rt);
1658 * point to the (possibly newly malloc'd) dest address.
1660 ndst = (struct sockaddr *)rt_key(rt);
1663 * make sure it contains the value we want (masked if needed).
1666 rt_maskedcopy(dst, ndst, netmask);
1668 bcopy(dst, ndst, dst->sa_len);
1671 * We use the ifa reference returned by rt_getifa_fib().
1672 * This moved from below so that rnh->rnh_addaddr() can
1673 * examine the ifa and ifa->ifa_ifp if it so desires.
1676 rt->rt_ifp = ifa->ifa_ifp;
1679 rt_setmetrics(info, rt);
1684 /* do not permit exactly the same dst/mask/gw pair */
1685 if (rt_mpath_capable(rnh) &&
1686 rt_mpath_conflict(rnh, rt, netmask)) {
1689 ifa_free(rt->rt_ifa);
1691 uma_zfree(V_rtzone, rt);
1697 rt0 = rt_flowtable_check_route(rnh, info);
1698 #endif /* FLOWTABLE */
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);
1741 else if (rt0 != NULL) {
1742 flowtable_route_flush(dst->sa_family, rt0);
1747 if (rt_old != NULL) {
1748 rt_notifydelete(rt_old, info);
1753 * If this protocol has something to add to this then
1754 * allow it to do that as well.
1756 if (ifa->ifa_rtrequest)
1757 ifa->ifa_rtrequest(req, rt, info);
1760 * actually return a resultant rtentry and
1761 * give the caller a single reference.
1767 rnh->rnh_gen++; /* Routing table updated */
1772 error = rtrequest1_fib_change(rnh, info, ret_nrt, fibnum);
1790 rtrequest1_fib_change(struct rib_head *rnh, struct rt_addrinfo *info,
1791 struct rtentry **ret_nrt, u_int fibnum)
1793 struct rtentry *rt = NULL;
1797 struct if_mtuinfo ifmtu;
1799 rt = (struct rtentry *)rnh->rnh_lookup(info->rti_info[RTAX_DST],
1800 info->rti_info[RTAX_NETMASK], &rnh->head);
1807 * If we got multipath routes,
1808 * we require users to specify a matching RTAX_GATEWAY.
1810 if (rt_mpath_capable(rnh)) {
1811 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
1819 rt_setmetrics(info, rt);
1822 * New gateway could require new ifaddr, ifp;
1823 * flags may also be different; ifp may be specified
1824 * by ll sockaddr when protocol address is ambiguous
1826 if (((rt->rt_flags & RTF_GATEWAY) &&
1827 info->rti_info[RTAX_GATEWAY] != NULL) ||
1828 info->rti_info[RTAX_IFP] != NULL ||
1829 (info->rti_info[RTAX_IFA] != NULL &&
1830 !sa_equal(info->rti_info[RTAX_IFA], rt->rt_ifa->ifa_addr))) {
1832 error = rt_getifa_fib(info, fibnum);
1833 if (info->rti_ifa != NULL)
1840 /* Check if outgoing interface has changed */
1841 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa &&
1842 rt->rt_ifa != NULL && rt->rt_ifa->ifa_rtrequest != NULL) {
1843 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1844 ifa_free(rt->rt_ifa);
1846 /* Update gateway address */
1847 if (info->rti_info[RTAX_GATEWAY] != NULL) {
1848 error = rt_setgate(rt, rt_key(rt), info->rti_info[RTAX_GATEWAY]);
1852 rt->rt_flags &= ~RTF_GATEWAY;
1853 rt->rt_flags |= (RTF_GATEWAY & info->rti_flags);
1856 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa) {
1857 ifa_ref(info->rti_ifa);
1858 rt->rt_ifa = info->rti_ifa;
1859 rt->rt_ifp = info->rti_ifp;
1861 /* Allow some flags to be toggled on change. */
1862 rt->rt_flags &= ~RTF_FMASK;
1863 rt->rt_flags |= info->rti_flags & RTF_FMASK;
1865 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest != NULL)
1866 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
1868 /* Alter route MTU if necessary */
1869 if (rt->rt_ifp != NULL) {
1870 family = info->rti_info[RTAX_DST]->sa_family;
1871 mtu = if_getmtu_family(rt->rt_ifp, family);
1872 /* Set default MTU */
1873 if (rt->rt_mtu == 0)
1875 if (rt->rt_mtu != mtu) {
1876 /* Check if we really need to update */
1877 ifmtu.ifp = rt->rt_ifp;
1879 if_updatemtu_cb(rt->rt_nodes, &ifmtu);
1890 ifa_free(info->rti_ifa);
1895 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt)
1898 if (info->rti_mflags & RTV_MTU) {
1899 if (info->rti_rmx->rmx_mtu != 0) {
1902 * MTU was explicitly provided by user.
1905 rt->rt_flags |= RTF_FIXEDMTU;
1909 * User explicitly sets MTU to 0.
1910 * Assume rollback to default.
1912 rt->rt_flags &= ~RTF_FIXEDMTU;
1914 rt->rt_mtu = info->rti_rmx->rmx_mtu;
1916 if (info->rti_mflags & RTV_WEIGHT)
1917 rt->rt_weight = info->rti_rmx->rmx_weight;
1918 /* Kernel -> userland timebase conversion. */
1919 if (info->rti_mflags & RTV_EXPIRE)
1920 rt->rt_expire = info->rti_rmx->rmx_expire ?
1921 info->rti_rmx->rmx_expire - time_second + time_uptime : 0;
1925 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
1927 /* XXX dst may be overwritten, can we move this to below */
1928 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
1931 * Prepare to store the gateway in rt->rt_gateway.
1932 * Both dst and gateway are stored one after the other in the same
1933 * malloc'd chunk. If we have room, we can reuse the old buffer,
1934 * rt_gateway already points to the right place.
1935 * Otherwise, malloc a new block and update the 'dst' address.
1937 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
1940 R_Malloc(new, caddr_t, dlen + glen);
1944 * XXX note, we copy from *dst and not *rt_key(rt) because
1945 * rt_setgate() can be called to initialize a newly
1946 * allocated route entry, in which case rt_key(rt) == NULL
1947 * (and also rt->rt_gateway == NULL).
1948 * Free()/free() handle a NULL argument just fine.
1950 bcopy(dst, new, dlen);
1951 R_Free(rt_key(rt)); /* free old block, if any */
1952 rt_key(rt) = (struct sockaddr *)new;
1953 rt->rt_gateway = (struct sockaddr *)(new + dlen);
1957 * Copy the new gateway value into the memory chunk.
1959 bcopy(gate, rt->rt_gateway, glen);
1965 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
1967 u_char *cp1 = (u_char *)src;
1968 u_char *cp2 = (u_char *)dst;
1969 u_char *cp3 = (u_char *)netmask;
1970 u_char *cplim = cp2 + *cp3;
1971 u_char *cplim2 = cp2 + *cp1;
1973 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1978 *cp2++ = *cp1++ & *cp3++;
1980 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
1984 * Set up a routing table entry, normally
1987 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
1989 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
1991 struct sockaddr *dst;
1992 struct sockaddr *netmask;
1993 struct rtentry *rt = NULL;
1994 struct rt_addrinfo info;
1996 int startfib, endfib;
1997 char tempbuf[_SOCKADDR_TMPSIZE];
2000 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
2001 struct rib_head *rnh;
2003 if (flags & RTF_HOST) {
2004 dst = ifa->ifa_dstaddr;
2007 dst = ifa->ifa_addr;
2008 netmask = ifa->ifa_netmask;
2010 if (dst->sa_len == 0)
2012 switch (dst->sa_family) {
2015 /* We support multiple FIBs. */
2018 fibnum = RT_DEFAULT_FIB;
2021 if (fibnum == RT_ALL_FIBS) {
2022 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD)
2023 startfib = endfib = ifa->ifa_ifp->if_fib;
2026 endfib = rt_numfibs - 1;
2029 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
2035 * If it's a delete, check that if it exists,
2036 * it's on the correct interface or we might scrub
2037 * a route to another ifa which would
2038 * be confusing at best and possibly worse.
2040 if (cmd == RTM_DELETE) {
2042 * It's a delete, so it should already exist..
2043 * If it's a net, mask off the host bits
2044 * (Assuming we have a mask)
2045 * XXX this is kinda inet specific..
2047 if (netmask != NULL) {
2048 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
2049 dst = (struct sockaddr *)tempbuf;
2053 * Now go through all the requested tables (fibs) and do the
2054 * requested action. Realistically, this will either be fib 0
2055 * for protocols that don't do multiple tables or all the
2056 * tables for those that do.
2058 for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
2059 if (cmd == RTM_DELETE) {
2060 struct radix_node *rn;
2062 * Look up an rtentry that is in the routing tree and
2063 * contains the correct info.
2065 rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
2067 /* this table doesn't exist but others might */
2070 rn = rnh->rnh_lookup(dst, netmask, &rnh->head);
2072 if (rt_mpath_capable(rnh)) {
2079 * for interface route the
2080 * rt->rt_gateway is sockaddr_intf
2081 * for cloning ARP entries, so
2082 * rt_mpath_matchgate must use the
2085 rt = rt_mpath_matchgate(rt,
2092 error = (rn == NULL ||
2093 (rn->rn_flags & RNF_ROOT) ||
2094 RNTORT(rn)->rt_ifa != ifa);
2097 /* this is only an error if bad on ALL tables */
2102 * Do the actual request
2104 bzero((caddr_t)&info, sizeof(info));
2106 info.rti_flags = flags |
2107 (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED;
2108 info.rti_info[RTAX_DST] = dst;
2110 * doing this for compatibility reasons
2113 info.rti_info[RTAX_GATEWAY] =
2114 (struct sockaddr *)&null_sdl;
2116 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
2117 info.rti_info[RTAX_NETMASK] = netmask;
2118 error = rtrequest1_fib(cmd, &info, &rt, fibnum);
2120 if (error == 0 && rt != NULL) {
2122 * notify any listening routing agents of the change
2127 * in case address alias finds the first address
2128 * e.g. ifconfig bge0 192.0.2.246/24
2129 * e.g. ifconfig bge0 192.0.2.247/24
2130 * the address set in the route is 192.0.2.246
2131 * so we need to replace it with 192.0.2.247
2133 if (memcmp(rt->rt_ifa->ifa_addr,
2134 ifa->ifa_addr, ifa->ifa_addr->sa_len)) {
2135 ifa_free(rt->rt_ifa);
2137 rt->rt_ifp = ifa->ifa_ifp;
2142 * doing this for compatibility reasons
2144 if (cmd == RTM_ADD) {
2145 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
2146 rt->rt_ifp->if_type;
2147 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
2148 rt->rt_ifp->if_index;
2152 rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum);
2155 if (cmd == RTM_DELETE) {
2157 * If we are deleting, and we found an entry,
2158 * then it's been removed from the tree..
2159 * now throw it away.
2163 if (cmd == RTM_ADD) {
2165 * We just wanted to add it..
2166 * we don't actually need a reference.
2177 if (cmd == RTM_DELETE) {
2181 /* we only give an error if it wasn't in any table */
2182 error = ((flags & RTF_HOST) ?
2183 EHOSTUNREACH : ENETUNREACH);
2187 /* return an error if any of them failed */
2195 * Set up a routing table entry, normally
2199 rtinit(struct ifaddr *ifa, int cmd, int flags)
2201 struct sockaddr *dst;
2202 int fib = RT_DEFAULT_FIB;
2204 if (flags & RTF_HOST) {
2205 dst = ifa->ifa_dstaddr;
2207 dst = ifa->ifa_addr;
2210 switch (dst->sa_family) {
2213 /* We do support multiple FIBs. */
2217 return (rtinit1(ifa, cmd, flags, fib));
2221 * Announce interface address arrival/withdraw
2222 * Returns 0 on success.
2225 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
2228 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2229 ("unexpected cmd %d", cmd));
2231 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2232 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2234 #if defined(INET) || defined(INET6)
2237 * notify the SCTP stack
2238 * this will only get called when an address is added/deleted
2239 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
2241 sctp_addr_change(ifa, cmd);
2244 return (rtsock_addrmsg(cmd, ifa, fibnum));
2248 * Announce route addition/removal.
2249 * Users of this function MUST validate input data BEFORE calling.
2250 * However we have to be able to handle invalid data:
2251 * if some userland app sends us "invalid" route message (invalid mask,
2252 * no dst, wrong address families, etc...) we need to pass it back
2253 * to app (and any other rtsock consumers) with rtm_errno field set to
2255 * Returns 0 on success.
2258 rt_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
2262 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2263 ("unexpected cmd %d", cmd));
2265 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2266 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2268 KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
2270 return (rtsock_routemsg(cmd, ifp, error, rt, fibnum));
2274 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
2277 rt_newaddrmsg_fib(cmd, ifa, error, rt, RT_ALL_FIBS);
2281 * This is called to generate messages from the routing socket
2282 * indicating a network interface has had addresses associated with it.
2285 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
2289 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
2290 ("unexpected cmd %u", cmd));
2291 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
2292 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
2294 if (cmd == RTM_ADD) {
2295 rt_addrmsg(cmd, ifa, fibnum);
2297 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2300 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
2301 rt_addrmsg(cmd, ifa, fibnum);