2 * Copyright (c) 2004, 2005 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
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 DragonFly Project nor the names of its
16 * contributors may be used to endorse or promote products derived
17 * from this software without specific, prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
22 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
23 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
24 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
25 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
27 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
28 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
29 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * Copyright (c) 1980, 1986, 1991, 1993
35 * The Regents of the University of California. All rights reserved.
37 * Redistribution and use in source and binary forms, with or without
38 * modification, are permitted provided that the following conditions
40 * 1. Redistributions of source code must retain the above copyright
41 * notice, this list of conditions and the following disclaimer.
42 * 2. Redistributions in binary form must reproduce the above copyright
43 * notice, this list of conditions and the following disclaimer in the
44 * documentation and/or other materials provided with the distribution.
45 * 3. All advertising materials mentioning features or use of this software
46 * must display the following acknowledgement:
47 * This product includes software developed by the University of
48 * California, Berkeley and its contributors.
49 * 4. Neither the name of the University nor the names of its contributors
50 * may be used to endorse or promote products derived from this software
51 * without specific prior written permission.
53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65 * @(#)route.c 8.3 (Berkeley) 1/9/95
66 * $FreeBSD: src/sys/net/route.c,v 1.59.2.10 2003/01/17 08:04:00 ru Exp $
67 * $DragonFly: src/sys/net/route.c,v 1.38 2008/09/12 11:37:41 sephe Exp $
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/malloc.h>
77 #include <sys/socket.h>
78 #include <sys/domain.h>
79 #include <sys/kernel.h>
80 #include <sys/sysctl.h>
81 #include <sys/globaldata.h>
82 #include <sys/thread.h>
85 #include <net/route.h>
86 #include <net/netisr.h>
88 #include <netinet/in.h>
89 #include <net/ip_mroute/ip_mroute.h>
91 #include <sys/thread2.h>
92 #include <sys/msgport2.h>
93 #include <net/netmsg2.h>
96 #include <netproto/mpls/mpls.h>
99 static struct rtstatistics rtstatistics_percpu[MAXCPU];
101 #define rtstat rtstatistics_percpu[mycpuid]
103 #define rtstat rtstatistics_percpu[0]
106 struct radix_node_head *rt_tables[MAXCPU][AF_MAX+1];
107 struct lwkt_port *rt_ports[MAXCPU];
109 static void rt_maskedcopy (struct sockaddr *, struct sockaddr *,
111 static void rtable_init(void);
112 static void rtable_service_loop(void *dummy);
113 static void rtinit_rtrequest_callback(int, int, struct rt_addrinfo *,
114 struct rtentry *, void *);
117 static void rtredirect_msghandler(struct netmsg *netmsg);
118 static void rtrequest1_msghandler(struct netmsg *netmsg);
121 static int rt_setshims(struct rtentry *, struct sockaddr **);
123 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RW, 0, "Routing");
126 static int route_debug = 1;
127 SYSCTL_INT(_net_route, OID_AUTO, route_debug, CTLFLAG_RW,
128 &route_debug, 0, "");
131 int route_assert_owner_access = 0;
132 SYSCTL_INT(_net_route, OID_AUTO, assert_owner_access, CTLFLAG_RW,
133 &route_assert_owner_access, 0, "");
134 SYSCTL_INT(_net_route, OID_AUTO, remote_free_panic, CTLFLAG_RW,
135 &route_assert_owner_access, 0, ""); /* alias */
136 extern void db_print_backtrace(void);
139 * Initialize the route table(s) for protocol domains and
140 * create a helper thread which will be responsible for updating
141 * route table entries on each cpu.
149 for (cpu = 0; cpu < ncpus; ++cpu)
150 bzero(&rtstatistics_percpu[cpu], sizeof(struct rtstatistics));
151 rn_init(); /* initialize all zeroes, all ones, mask table */
153 for (cpu = 0; cpu < ncpus; cpu++) {
154 lwkt_migratecpu(cpu);
156 lwkt_create(rtable_service_loop, NULL, &rtd, NULL,
157 TDF_STOPREQ, cpu, "rtable_cpu %d", cpu);
158 rt_ports[cpu] = &rtd->td_msgport;
161 lwkt_migratecpu(origcpu);
169 SLIST_FOREACH(dom, &domains, dom_next) {
170 if (dom->dom_rtattach) {
172 (void **)&rt_tables[mycpuid][dom->dom_family],
179 * Our per-cpu table management protocol thread. All route table operations
180 * are sequentially chained through all cpus starting at cpu #0 in order to
181 * maintain duplicate route tables on each cpu. Having a spearate route
182 * table management thread allows the protocol and interrupt threads to
183 * issue route table changes.
186 rtable_service_loop(void *dummy __unused)
188 struct netmsg *netmsg;
189 thread_t td = curthread;
191 while ((netmsg = lwkt_waitport(&td->td_msgport, 0)) != NULL) {
192 netmsg->nm_dispatch(netmsg);
197 * Routing statistics.
201 sysctl_rtstatistics(SYSCTL_HANDLER_ARGS)
205 for (cpu = 0; cpu < ncpus; ++cpu) {
206 if ((error = SYSCTL_OUT(req, &rtstatistics_percpu[cpu],
207 sizeof(struct rtstatistics))))
209 if ((error = SYSCTL_IN(req, &rtstatistics_percpu[cpu],
210 sizeof(struct rtstatistics))))
216 SYSCTL_PROC(_net_route, OID_AUTO, stats, (CTLTYPE_OPAQUE|CTLFLAG_RW),
217 0, 0, sysctl_rtstatistics, "S,rtstatistics", "Routing statistics");
219 SYSCTL_STRUCT(_net_route, OID_AUTO, stats, CTLFLAG_RW, &rtstat, rtstatistics,
220 "Routing statistics");
224 * Packet routing routines.
228 * Look up and fill in the "ro_rt" rtentry field in a route structure given
229 * an address in the "ro_dst" field. Always send a report on a miss and
230 * always clone routes.
233 rtalloc(struct route *ro)
235 rtalloc_ign(ro, 0UL);
239 * Look up and fill in the "ro_rt" rtentry field in a route structure given
240 * an address in the "ro_dst" field. Always send a report on a miss and
241 * optionally clone routes when RTF_CLONING or RTF_PRCLONING are not being
245 rtalloc_ign(struct route *ro, u_long ignoreflags)
247 if (ro->ro_rt != NULL) {
248 if (ro->ro_rt->rt_ifp != NULL && ro->ro_rt->rt_flags & RTF_UP)
253 ro->ro_rt = _rtlookup(&ro->ro_dst, RTL_REPORTMSG, ignoreflags);
257 * Look up the route that matches the given "dst" address.
259 * Route lookup can have the side-effect of creating and returning
260 * a cloned route instead when "dst" matches a cloning route and the
261 * RTF_CLONING and RTF_PRCLONING flags are not being ignored.
263 * Any route returned has its reference count incremented.
266 _rtlookup(struct sockaddr *dst, boolean_t generate_report, u_long ignore)
268 struct radix_node_head *rnh = rt_tables[mycpuid][dst->sa_family];
275 * Look up route in the radix tree.
277 rt = (struct rtentry *) rnh->rnh_matchaddr((char *)dst, rnh);
282 * Handle cloning routes.
284 if ((rt->rt_flags & ~ignore & (RTF_CLONING | RTF_PRCLONING)) != 0) {
285 struct rtentry *clonedroute;
288 clonedroute = rt; /* copy in/copy out parameter */
289 error = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
290 &clonedroute); /* clone the route */
291 if (error != 0) { /* cloning failed */
293 rt_dstmsg(RTM_MISS, dst, error);
295 return (rt); /* return the uncloned route */
297 if (generate_report) {
298 if (clonedroute->rt_flags & RTF_XRESOLVE)
299 rt_dstmsg(RTM_RESOLVE, dst, 0);
301 rt_rtmsg(RTM_ADD, clonedroute,
302 clonedroute->rt_ifp, 0);
304 return (clonedroute); /* return cloned route */
308 * Increment the reference count of the matched route and return.
314 rtstat.rts_unreach++;
316 rt_dstmsg(RTM_MISS, dst, 0);
321 rtfree(struct rtentry *rt)
323 if (rt->rt_cpuid == mycpuid)
326 rtfree_remote(rt, 1);
330 rtfree_oncpu(struct rtentry *rt)
332 KKASSERT(rt->rt_cpuid == mycpuid);
333 KASSERT(rt->rt_refcnt > 0, ("rtfree: rt_refcnt %ld", rt->rt_refcnt));
336 if (rt->rt_refcnt == 0) {
337 struct radix_node_head *rnh =
338 rt_tables[mycpuid][rt_key(rt)->sa_family];
341 rnh->rnh_close((struct radix_node *)rt, rnh);
342 if (!(rt->rt_flags & RTF_UP)) {
343 /* deallocate route */
344 if (rt->rt_ifa != NULL)
346 if (rt->rt_parent != NULL)
347 RTFREE(rt->rt_parent); /* recursive call! */
355 rtfree_remote_dispatch(struct netmsg *nmsg)
357 struct lwkt_msg *lmsg = &nmsg->nm_lmsg;
358 struct rtentry *rt = lmsg->u.ms_resultp;
361 lwkt_replymsg(lmsg, 0);
365 rtfree_remote(struct rtentry *rt, int allow_panic)
368 struct lwkt_msg *lmsg;
370 KKASSERT(rt->rt_cpuid != mycpuid);
372 if (route_assert_owner_access && allow_panic) {
373 panic("rt remote free rt_cpuid %d, mycpuid %d\n",
374 rt->rt_cpuid, mycpuid);
376 kprintf("rt remote free rt_cpuid %d, mycpuid %d\n",
377 rt->rt_cpuid, mycpuid);
378 db_print_backtrace();
381 netmsg_init(&nmsg, &curthread->td_msgport, 0, rtfree_remote_dispatch);
382 lmsg = &nmsg.nm_lmsg;
383 lmsg->u.ms_resultp = rt;
385 lwkt_domsg(rtable_portfn(rt->rt_cpuid), lmsg, 0);
389 rtredirect_oncpu(struct sockaddr *dst, struct sockaddr *gateway,
390 struct sockaddr *netmask, int flags, struct sockaddr *src)
392 struct rtentry *rt = NULL;
393 struct rt_addrinfo rtinfo;
398 /* verify the gateway is directly reachable */
399 if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
405 * If the redirect isn't from our current router for this destination,
406 * it's either old or wrong.
408 if (!(flags & RTF_DONE) && /* XXX JH */
409 (rt = rtpurelookup(dst)) != NULL &&
410 (!sa_equal(src, rt->rt_gateway) || rt->rt_ifa != ifa)) {
416 * If it redirects us to ourselves, we have a routing loop,
417 * perhaps as a result of an interface going down recently.
419 if (ifa_ifwithaddr(gateway)) {
420 error = EHOSTUNREACH;
425 * Create a new entry if the lookup failed or if we got back
426 * a wildcard entry for the default route. This is necessary
427 * for hosts which use routing redirects generated by smart
428 * gateways to dynamically build the routing tables.
432 if ((rt_mask(rt) != NULL && rt_mask(rt)->sa_len < 2)) {
437 /* Ignore redirects for directly connected hosts. */
438 if (!(rt->rt_flags & RTF_GATEWAY)) {
439 error = EHOSTUNREACH;
443 if (!(rt->rt_flags & RTF_HOST) && (flags & RTF_HOST)) {
445 * Changing from a network route to a host route.
446 * Create a new host route rather than smashing the
450 flags |= RTF_GATEWAY | RTF_DYNAMIC;
451 bzero(&rtinfo, sizeof(struct rt_addrinfo));
452 rtinfo.rti_info[RTAX_DST] = dst;
453 rtinfo.rti_info[RTAX_GATEWAY] = gateway;
454 rtinfo.rti_info[RTAX_NETMASK] = netmask;
455 rtinfo.rti_flags = flags;
456 rtinfo.rti_ifa = ifa;
457 rt = NULL; /* copy-in/copy-out parameter */
458 error = rtrequest1(RTM_ADD, &rtinfo, &rt);
460 flags = rt->rt_flags;
461 stat = &rtstat.rts_dynamic;
464 * Smash the current notion of the gateway to this destination.
465 * Should check about netmask!!!
467 rt->rt_flags |= RTF_MODIFIED;
468 flags |= RTF_MODIFIED;
469 rt_setgate(rt, rt_key(rt), gateway);
471 stat = &rtstat.rts_newgateway;
479 rtstat.rts_badredirect++;
480 else if (stat != NULL)
488 struct netmsg_rtredirect {
489 struct netmsg netmsg;
490 struct sockaddr *dst;
491 struct sockaddr *gateway;
492 struct sockaddr *netmask;
494 struct sockaddr *src;
500 * Force a routing table entry to the specified
501 * destination to go through the given gateway.
502 * Normally called as a result of a routing redirect
503 * message from the network layer.
505 * N.B.: must be called at splnet
508 rtredirect(struct sockaddr *dst, struct sockaddr *gateway,
509 struct sockaddr *netmask, int flags, struct sockaddr *src)
511 struct rt_addrinfo rtinfo;
514 struct netmsg_rtredirect msg;
516 netmsg_init(&msg.netmsg, &curthread->td_msgport, 0,
517 rtredirect_msghandler);
519 msg.gateway = gateway;
520 msg.netmask = netmask;
523 error = lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
525 error = rtredirect_oncpu(dst, gateway, netmask, flags, src);
527 bzero(&rtinfo, sizeof(struct rt_addrinfo));
528 rtinfo.rti_info[RTAX_DST] = dst;
529 rtinfo.rti_info[RTAX_GATEWAY] = gateway;
530 rtinfo.rti_info[RTAX_NETMASK] = netmask;
531 rtinfo.rti_info[RTAX_AUTHOR] = src;
532 rt_missmsg(RTM_REDIRECT, &rtinfo, flags, error);
538 rtredirect_msghandler(struct netmsg *netmsg)
540 struct netmsg_rtredirect *msg = (void *)netmsg;
543 rtredirect_oncpu(msg->dst, msg->gateway, msg->netmask,
544 msg->flags, msg->src);
545 nextcpu = mycpuid + 1;
547 lwkt_forwardmsg(rtable_portfn(nextcpu), &netmsg->nm_lmsg);
549 lwkt_replymsg(&netmsg->nm_lmsg, 0);
555 * Routing table ioctl interface.
558 rtioctl(u_long req, caddr_t data, struct ucred *cred)
561 /* Multicast goop, grrr... */
562 return mrt_ioctl ? mrt_ioctl(req, data) : EOPNOTSUPP;
569 ifa_ifwithroute(int flags, struct sockaddr *dst, struct sockaddr *gateway)
573 if (!(flags & RTF_GATEWAY)) {
575 * If we are adding a route to an interface,
576 * and the interface is a point-to-point link,
577 * we should search for the destination
578 * as our clue to the interface. Otherwise
579 * we can use the local address.
582 if (flags & RTF_HOST) {
583 ifa = ifa_ifwithdstaddr(dst);
586 ifa = ifa_ifwithaddr(gateway);
589 * If we are adding a route to a remote net
590 * or host, the gateway may still be on the
591 * other end of a pt to pt link.
593 ifa = ifa_ifwithdstaddr(gateway);
596 ifa = ifa_ifwithnet(gateway);
600 rt = rtpurelookup(gateway);
604 if ((ifa = rt->rt_ifa) == NULL)
607 if (ifa->ifa_addr->sa_family != dst->sa_family) {
608 struct ifaddr *oldifa = ifa;
610 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
617 static int rt_fixdelete (struct radix_node *, void *);
618 static int rt_fixchange (struct radix_node *, void *);
622 struct radix_node_head *rnh;
626 * Set rtinfo->rti_ifa and rtinfo->rti_ifp.
629 rt_getifa(struct rt_addrinfo *rtinfo)
631 struct sockaddr *gateway = rtinfo->rti_info[RTAX_GATEWAY];
632 struct sockaddr *dst = rtinfo->rti_info[RTAX_DST];
633 struct sockaddr *ifaaddr = rtinfo->rti_info[RTAX_IFA];
634 int flags = rtinfo->rti_flags;
637 * ifp may be specified by sockaddr_dl
638 * when protocol address is ambiguous.
640 if (rtinfo->rti_ifp == NULL) {
641 struct sockaddr *ifpaddr;
643 ifpaddr = rtinfo->rti_info[RTAX_IFP];
644 if (ifpaddr != NULL && ifpaddr->sa_family == AF_LINK) {
647 ifa = ifa_ifwithnet(ifpaddr);
649 rtinfo->rti_ifp = ifa->ifa_ifp;
653 if (rtinfo->rti_ifa == NULL && ifaaddr != NULL)
654 rtinfo->rti_ifa = ifa_ifwithaddr(ifaaddr);
655 if (rtinfo->rti_ifa == NULL) {
658 sa = ifaaddr != NULL ? ifaaddr :
659 (gateway != NULL ? gateway : dst);
660 if (sa != NULL && rtinfo->rti_ifp != NULL)
661 rtinfo->rti_ifa = ifaof_ifpforaddr(sa, rtinfo->rti_ifp);
662 else if (dst != NULL && gateway != NULL)
663 rtinfo->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
665 rtinfo->rti_ifa = ifa_ifwithroute(flags, sa, sa);
667 if (rtinfo->rti_ifa == NULL)
668 return (ENETUNREACH);
670 if (rtinfo->rti_ifp == NULL)
671 rtinfo->rti_ifp = rtinfo->rti_ifa->ifa_ifp;
676 * Do appropriate manipulations of a routing tree given
677 * all the bits of info needed
682 struct sockaddr *dst,
683 struct sockaddr *gateway,
684 struct sockaddr *netmask,
686 struct rtentry **ret_nrt)
688 struct rt_addrinfo rtinfo;
690 bzero(&rtinfo, sizeof(struct rt_addrinfo));
691 rtinfo.rti_info[RTAX_DST] = dst;
692 rtinfo.rti_info[RTAX_GATEWAY] = gateway;
693 rtinfo.rti_info[RTAX_NETMASK] = netmask;
694 rtinfo.rti_flags = flags;
695 return rtrequest1(req, &rtinfo, ret_nrt);
701 struct sockaddr *dst,
702 struct sockaddr *gateway,
703 struct sockaddr *netmask,
706 struct rt_addrinfo rtinfo;
708 bzero(&rtinfo, sizeof(struct rt_addrinfo));
709 rtinfo.rti_info[RTAX_DST] = dst;
710 rtinfo.rti_info[RTAX_GATEWAY] = gateway;
711 rtinfo.rti_info[RTAX_NETMASK] = netmask;
712 rtinfo.rti_flags = flags;
713 return rtrequest1_global(req, &rtinfo, NULL, NULL);
719 struct netmsg netmsg;
721 struct rt_addrinfo *rtinfo;
722 rtrequest1_callback_func_t callback;
729 rtrequest1_global(int req, struct rt_addrinfo *rtinfo,
730 rtrequest1_callback_func_t callback, void *arg)
734 struct netmsg_rtq msg;
736 netmsg_init(&msg.netmsg, &curthread->td_msgport, 0,
737 rtrequest1_msghandler);
738 msg.netmsg.nm_lmsg.ms_error = -1;
741 msg.callback = callback;
743 error = lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
745 struct rtentry *rt = NULL;
747 error = rtrequest1(req, rtinfo, &rt);
751 callback(req, error, rtinfo, rt, arg);
757 * Handle a route table request on the current cpu. Since the route table's
758 * are supposed to be identical on each cpu, an error occuring later in the
759 * message chain is considered system-fatal.
764 rtrequest1_msghandler(struct netmsg *netmsg)
766 struct netmsg_rtq *msg = (void *)netmsg;
767 struct rtentry *rt = NULL;
771 error = rtrequest1(msg->req, msg->rtinfo, &rt);
775 msg->callback(msg->req, error, msg->rtinfo, rt, msg->arg);
778 * RTM_DELETE's are propogated even if an error occurs, since a
779 * cloned route might be undergoing deletion and cloned routes
780 * are not necessarily replicated. An overall error is returned
781 * only if no cpus have the route in question.
783 if (msg->netmsg.nm_lmsg.ms_error < 0 || error == 0)
784 msg->netmsg.nm_lmsg.ms_error = error;
786 nextcpu = mycpuid + 1;
787 if (error && msg->req != RTM_DELETE) {
789 panic("rtrequest1_msghandler: rtrequest table "
790 "error was not on cpu #0: %p", msg->rtinfo);
792 lwkt_replymsg(&msg->netmsg.nm_lmsg, error);
793 } else if (nextcpu < ncpus) {
794 lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg);
796 lwkt_replymsg(&msg->netmsg.nm_lmsg,
797 msg->netmsg.nm_lmsg.ms_error);
804 rtrequest1(int req, struct rt_addrinfo *rtinfo, struct rtentry **ret_nrt)
806 struct sockaddr *dst = rtinfo->rti_info[RTAX_DST];
808 struct radix_node *rn;
809 struct radix_node_head *rnh;
811 struct sockaddr *ndst;
814 #define gotoerr(x) { error = x ; goto bad; }
818 rt_addrinfo_print(req, rtinfo);
823 * Find the correct routing tree to use for this Address Family
825 if ((rnh = rt_tables[mycpuid][dst->sa_family]) == NULL)
826 gotoerr(EAFNOSUPPORT);
829 * If we are adding a host route then we don't want to put
830 * a netmask in the tree, nor do we want to clone it.
832 if (rtinfo->rti_flags & RTF_HOST) {
833 rtinfo->rti_info[RTAX_NETMASK] = NULL;
834 rtinfo->rti_flags &= ~(RTF_CLONING | RTF_PRCLONING);
839 /* Remove the item from the tree. */
840 rn = rnh->rnh_deladdr((char *)rtinfo->rti_info[RTAX_DST],
841 (char *)rtinfo->rti_info[RTAX_NETMASK],
845 KASSERT(!(rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)),
846 ("rnh_deladdr returned flags 0x%x", rn->rn_flags));
847 rt = (struct rtentry *)rn;
849 /* ref to prevent a deletion race */
852 /* Free any routes cloned from this one. */
853 if ((rt->rt_flags & (RTF_CLONING | RTF_PRCLONING)) &&
854 rt_mask(rt) != NULL) {
855 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt),
860 if (rt->rt_gwroute != NULL) {
861 RTFREE(rt->rt_gwroute);
862 rt->rt_gwroute = NULL;
866 * NB: RTF_UP must be set during the search above,
867 * because we might delete the last ref, causing
868 * rt to get freed prematurely.
870 rt->rt_flags &= ~RTF_UP;
874 rt_print(rtinfo, rt);
877 /* Give the protocol a chance to keep things in sync. */
878 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
879 ifa->ifa_rtrequest(RTM_DELETE, rt, rtinfo);
882 * If the caller wants it, then it can have it,
883 * but it's up to it to free the rtentry as we won't be
886 KASSERT(rt->rt_refcnt >= 0,
887 ("rtrequest1(DELETE): refcnt %ld", rt->rt_refcnt));
888 if (ret_nrt != NULL) {
889 /* leave ref intact for return */
892 /* deref / attempt to destroy */
898 if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
902 rt->rt_flags & ~(RTF_CLONING | RTF_PRCLONING | RTF_STATIC);
903 rtinfo->rti_flags |= RTF_WASCLONED;
904 rtinfo->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
905 if ((rtinfo->rti_info[RTAX_NETMASK] = rt->rt_genmask) == NULL)
906 rtinfo->rti_flags |= RTF_HOST;
907 rtinfo->rti_info[RTAX_MPLS1] = rt->rt_shim[0];
908 rtinfo->rti_info[RTAX_MPLS2] = rt->rt_shim[1];
909 rtinfo->rti_info[RTAX_MPLS3] = rt->rt_shim[2];
913 KASSERT(!(rtinfo->rti_flags & RTF_GATEWAY) ||
914 rtinfo->rti_info[RTAX_GATEWAY] != NULL,
915 ("rtrequest: GATEWAY but no gateway"));
917 if (rtinfo->rti_ifa == NULL && (error = rt_getifa(rtinfo)))
919 ifa = rtinfo->rti_ifa;
921 R_Malloc(rt, struct rtentry *, sizeof(struct rtentry));
924 bzero(rt, sizeof(struct rtentry));
925 rt->rt_flags = RTF_UP | rtinfo->rti_flags;
926 rt->rt_cpuid = mycpuid;
927 error = rt_setgate(rt, dst, rtinfo->rti_info[RTAX_GATEWAY]);
934 if (rtinfo->rti_info[RTAX_NETMASK] != NULL)
935 rt_maskedcopy(dst, ndst,
936 rtinfo->rti_info[RTAX_NETMASK]);
938 bcopy(dst, ndst, dst->sa_len);
940 if (rtinfo->rti_info[RTAX_MPLS1] != NULL)
941 rt_setshims(rt, rtinfo->rti_info);
944 * Note that we now have a reference to the ifa.
945 * This moved from below so that rnh->rnh_addaddr() can
946 * examine the ifa and ifa->ifa_ifp if it so desires.
950 rt->rt_ifp = ifa->ifa_ifp;
951 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
953 rn = rnh->rnh_addaddr((char *)ndst,
954 (char *)rtinfo->rti_info[RTAX_NETMASK],
957 struct rtentry *oldrt;
960 * We already have one of these in the tree.
961 * We do a special hack: if the old route was
962 * cloned, then we blow it away and try
963 * re-inserting the new one.
965 oldrt = rtpurelookup(ndst);
968 if (oldrt->rt_flags & RTF_WASCLONED) {
969 rtrequest(RTM_DELETE, rt_key(oldrt),
972 oldrt->rt_flags, NULL);
973 rn = rnh->rnh_addaddr((char *)ndst,
975 rtinfo->rti_info[RTAX_NETMASK],
982 * If it still failed to go into the tree,
983 * then un-make it (this should be a function).
986 if (rt->rt_gwroute != NULL)
987 rtfree(rt->rt_gwroute);
995 * If we got here from RESOLVE, then we are cloning
996 * so clone the rest, and note that we
997 * are a clone (and increment the parent's references)
999 if (req == RTM_RESOLVE) {
1000 rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
1001 rt->rt_rmx.rmx_pksent = 0; /* reset packet counter */
1002 if ((*ret_nrt)->rt_flags &
1003 (RTF_CLONING | RTF_PRCLONING)) {
1004 rt->rt_parent = *ret_nrt;
1005 (*ret_nrt)->rt_refcnt++;
1010 * if this protocol has something to add to this then
1011 * allow it to do that as well.
1013 if (ifa->ifa_rtrequest != NULL)
1014 ifa->ifa_rtrequest(req, rt, rtinfo);
1017 * We repeat the same procedure from rt_setgate() here because
1018 * it doesn't fire when we call it there because the node
1019 * hasn't been added to the tree yet.
1021 if (req == RTM_ADD && !(rt->rt_flags & RTF_HOST) &&
1022 rt_mask(rt) != NULL) {
1023 struct rtfc_arg arg = { rt, rnh };
1025 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt),
1026 (char *)rt_mask(rt),
1027 rt_fixchange, &arg);
1032 rt_print(rtinfo, rt);
1035 * Return the resulting rtentry,
1036 * increasing the number of references by one.
1038 if (ret_nrt != NULL) {
1050 kprintf("rti %p failed error %d\n", rtinfo, error);
1052 kprintf("rti %p succeeded\n", rtinfo);
1060 * Called from rtrequest(RTM_DELETE, ...) to fix up the route's ``family''
1061 * (i.e., the routes related to it by the operation of cloning). This
1062 * routine is iterated over all potential former-child-routes by way of
1063 * rnh->rnh_walktree_from() above, and those that actually are children of
1064 * the late parent (passed in as VP here) are themselves deleted.
1067 rt_fixdelete(struct radix_node *rn, void *vp)
1069 struct rtentry *rt = (struct rtentry *)rn;
1070 struct rtentry *rt0 = vp;
1072 if (rt->rt_parent == rt0 &&
1073 !(rt->rt_flags & (RTF_PINNED | RTF_CLONING | RTF_PRCLONING))) {
1074 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt),
1075 rt->rt_flags, NULL);
1081 * This routine is called from rt_setgate() to do the analogous thing for
1082 * adds and changes. There is the added complication in this case of a
1083 * middle insert; i.e., insertion of a new network route between an older
1084 * network route and (cloned) host routes. For this reason, a simple check
1085 * of rt->rt_parent is insufficient; each candidate route must be tested
1086 * against the (mask, value) of the new route (passed as before in vp)
1087 * to see if the new route matches it.
1089 * XXX - it may be possible to do fixdelete() for changes and reserve this
1090 * routine just for adds. I'm not sure why I thought it was necessary to do
1094 static int rtfcdebug = 0;
1098 rt_fixchange(struct radix_node *rn, void *vp)
1100 struct rtentry *rt = (struct rtentry *)rn;
1101 struct rtfc_arg *ap = vp;
1102 struct rtentry *rt0 = ap->rt0;
1103 struct radix_node_head *rnh = ap->rnh;
1104 u_char *xk1, *xm1, *xk2, *xmp;
1109 kprintf("rt_fixchange: rt %p, rt0 %p\n", rt, rt0);
1112 if (rt->rt_parent == NULL ||
1113 (rt->rt_flags & (RTF_PINNED | RTF_CLONING | RTF_PRCLONING))) {
1115 if (rtfcdebug) kprintf("no parent, pinned or cloning\n");
1120 if (rt->rt_parent == rt0) {
1122 if (rtfcdebug) kprintf("parent match\n");
1124 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt),
1125 rt->rt_flags, NULL);
1129 * There probably is a function somewhere which does this...
1130 * if not, there should be.
1132 len = imin(rt_key(rt0)->sa_len, rt_key(rt)->sa_len);
1134 xk1 = (u_char *)rt_key(rt0);
1135 xm1 = (u_char *)rt_mask(rt0);
1136 xk2 = (u_char *)rt_key(rt);
1138 /* avoid applying a less specific route */
1139 xmp = (u_char *)rt_mask(rt->rt_parent);
1140 mlen = rt_key(rt->rt_parent)->sa_len;
1141 if (mlen > rt_key(rt0)->sa_len) {
1144 kprintf("rt_fixchange: inserting a less "
1145 "specific route\n");
1149 for (i = rnh->rnh_treetop->rn_offset; i < mlen; i++) {
1150 if ((xmp[i] & ~(xmp[i] ^ xm1[i])) != xmp[i]) {
1153 kprintf("rt_fixchange: inserting a less "
1154 "specific route\n");
1160 for (i = rnh->rnh_treetop->rn_offset; i < len; i++) {
1161 if ((xk2[i] & xm1[i]) != xk1[i]) {
1163 if (rtfcdebug) kprintf("no match\n");
1170 * OK, this node is a clone, and matches the node currently being
1171 * changed/added under the node's mask. So, get rid of it.
1174 if (rtfcdebug) kprintf("deleting\n");
1176 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt),
1177 rt->rt_flags, NULL);
1180 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
1183 rt_setgate(struct rtentry *rt0, struct sockaddr *dst, struct sockaddr *gate)
1185 char *space, *oldspace;
1186 int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len);
1187 struct rtentry *rt = rt0;
1188 struct radix_node_head *rnh = rt_tables[mycpuid][dst->sa_family];
1191 * A host route with the destination equal to the gateway
1192 * will interfere with keeping LLINFO in the routing
1193 * table, so disallow it.
1195 if (((rt0->rt_flags & (RTF_HOST | RTF_GATEWAY | RTF_LLINFO)) ==
1196 (RTF_HOST | RTF_GATEWAY)) &&
1197 dst->sa_len == gate->sa_len &&
1198 sa_equal(dst, gate)) {
1200 * The route might already exist if this is an RTM_CHANGE
1201 * or a routing redirect, so try to delete it.
1203 if (rt_key(rt0) != NULL)
1204 rtrequest(RTM_DELETE, rt_key(rt0), rt0->rt_gateway,
1205 rt_mask(rt0), rt0->rt_flags, NULL);
1206 return EADDRNOTAVAIL;
1210 * Both dst and gateway are stored in the same malloc'ed chunk
1211 * (If I ever get my hands on....)
1212 * if we need to malloc a new chunk, then keep the old one around
1213 * till we don't need it any more.
1215 if (rt->rt_gateway == NULL || glen > ROUNDUP(rt->rt_gateway->sa_len)) {
1216 oldspace = (char *)rt_key(rt);
1217 R_Malloc(space, char *, dlen + glen);
1220 rt->rt_nodes->rn_key = space;
1222 space = (char *)rt_key(rt); /* Just use the old space. */
1226 /* Set the gateway value. */
1227 rt->rt_gateway = (struct sockaddr *)(space + dlen);
1228 bcopy(gate, rt->rt_gateway, glen);
1230 if (oldspace != NULL) {
1232 * If we allocated a new chunk, preserve the original dst.
1233 * This way, rt_setgate() really just sets the gate
1234 * and leaves the dst field alone.
1236 bcopy(dst, space, dlen);
1241 * If there is already a gwroute, it's now almost definitely wrong
1244 if (rt->rt_gwroute != NULL) {
1245 RTFREE(rt->rt_gwroute);
1246 rt->rt_gwroute = NULL;
1248 if (rt->rt_flags & RTF_GATEWAY) {
1250 * Cloning loop avoidance: In the presence of
1251 * protocol-cloning and bad configuration, it is
1252 * possible to get stuck in bottomless mutual recursion
1253 * (rtrequest rt_setgate rtlookup). We avoid this
1254 * by not allowing protocol-cloning to operate for
1255 * gateways (which is probably the correct choice
1256 * anyway), and avoid the resulting reference loops
1257 * by disallowing any route to run through itself as
1258 * a gateway. This is obviously mandatory when we
1259 * get rt->rt_output().
1261 * This breaks TTCP for hosts outside the gateway! XXX JH
1263 rt->rt_gwroute = _rtlookup(gate, RTL_REPORTMSG, RTF_PRCLONING);
1264 if (rt->rt_gwroute == rt) {
1265 rt->rt_gwroute = NULL;
1267 return EDQUOT; /* failure */
1272 * This isn't going to do anything useful for host routes, so
1273 * don't bother. Also make sure we have a reasonable mask
1274 * (we don't yet have one during adds).
1276 if (!(rt->rt_flags & RTF_HOST) && rt_mask(rt) != NULL) {
1277 struct rtfc_arg arg = { rt, rnh };
1279 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt),
1280 (char *)rt_mask(rt),
1281 rt_fixchange, &arg);
1289 struct sockaddr *src,
1290 struct sockaddr *dst,
1291 struct sockaddr *netmask)
1293 u_char *cp1 = (u_char *)src;
1294 u_char *cp2 = (u_char *)dst;
1295 u_char *cp3 = (u_char *)netmask;
1296 u_char *cplim = cp2 + *cp3;
1297 u_char *cplim2 = cp2 + *cp1;
1299 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1304 *cp2++ = *cp1++ & *cp3++;
1306 bzero(cp2, cplim2 - cp2);
1310 rt_llroute(struct sockaddr *dst, struct rtentry *rt0, struct rtentry **drt)
1312 struct rtentry *up_rt, *rt;
1314 if (!(rt0->rt_flags & RTF_UP)) {
1315 up_rt = rtlookup(dst);
1317 return (EHOSTUNREACH);
1321 if (up_rt->rt_flags & RTF_GATEWAY) {
1322 if (up_rt->rt_gwroute == NULL) {
1323 up_rt->rt_gwroute = rtlookup(up_rt->rt_gateway);
1324 if (up_rt->rt_gwroute == NULL)
1325 return (EHOSTUNREACH);
1326 } else if (!(up_rt->rt_gwroute->rt_flags & RTF_UP)) {
1327 rtfree(up_rt->rt_gwroute);
1328 up_rt->rt_gwroute = rtlookup(up_rt->rt_gateway);
1329 if (up_rt->rt_gwroute == NULL)
1330 return (EHOSTUNREACH);
1332 rt = up_rt->rt_gwroute;
1335 if (rt->rt_flags & RTF_REJECT &&
1336 (rt->rt_rmx.rmx_expire == 0 || /* rt doesn't expire */
1337 time_second < rt->rt_rmx.rmx_expire)) /* rt not expired */
1338 return (rt->rt_flags & RTF_HOST ? EHOSTDOWN : EHOSTUNREACH);
1344 rt_setshims(struct rtentry *rt, struct sockaddr **rt_shim){
1347 for (i=0; i<3; i++) {
1348 struct sockaddr *shim = rt_shim[RTAX_MPLS1 + i];
1354 shimlen = ROUNDUP(shim->sa_len);
1355 R_Malloc(rt->rt_shim[i], struct sockaddr *, shimlen);
1356 bcopy(shim, rt->rt_shim[i], shimlen);
1365 * Print out a route table entry
1368 rt_print(struct rt_addrinfo *rtinfo, struct rtentry *rn)
1370 kprintf("rti %p cpu %d route %p flags %08lx: ",
1371 rtinfo, mycpuid, rn, rn->rt_flags);
1372 sockaddr_print(rt_key(rn));
1374 sockaddr_print(rt_mask(rn));
1376 sockaddr_print(rn->rt_gateway);
1377 kprintf(" ifc \"%s\"", rn->rt_ifp ? rn->rt_ifp->if_dname : "?");
1378 kprintf(" ifa %p\n", rn->rt_ifa);
1382 rt_addrinfo_print(int cmd, struct rt_addrinfo *rti)
1388 if (cmd == RTM_DELETE && route_debug > 1)
1389 db_print_backtrace();
1403 kprintf("C%02d ", cmd);
1406 kprintf("rti %p cpu %d ", rti, mycpuid);
1407 for (i = 0; i < rti->rti_addrs; ++i) {
1408 if (rti->rti_info[i] == NULL)
1438 kprintf("(?%02d ", i);
1441 sockaddr_print(rti->rti_info[i]);
1449 sockaddr_print(struct sockaddr *sa)
1451 struct sockaddr_in *sa4;
1452 struct sockaddr_in6 *sa6;
1461 len = sa->sa_len - offsetof(struct sockaddr, sa_data[0]);
1463 switch(sa->sa_family) {
1467 switch(sa->sa_family) {
1469 sa4 = (struct sockaddr_in *)sa;
1470 kprintf("INET %d %d.%d.%d.%d",
1471 ntohs(sa4->sin_port),
1472 (ntohl(sa4->sin_addr.s_addr) >> 24) & 255,
1473 (ntohl(sa4->sin_addr.s_addr) >> 16) & 255,
1474 (ntohl(sa4->sin_addr.s_addr) >> 8) & 255,
1475 (ntohl(sa4->sin_addr.s_addr) >> 0) & 255
1479 sa6 = (struct sockaddr_in6 *)sa;
1480 kprintf("INET6 %d %04x:%04x%04x:%04x:%04x:%04x:%04x:%04x",
1481 ntohs(sa6->sin6_port),
1482 sa6->sin6_addr.s6_addr16[0],
1483 sa6->sin6_addr.s6_addr16[1],
1484 sa6->sin6_addr.s6_addr16[2],
1485 sa6->sin6_addr.s6_addr16[3],
1486 sa6->sin6_addr.s6_addr16[4],
1487 sa6->sin6_addr.s6_addr16[5],
1488 sa6->sin6_addr.s6_addr16[6],
1489 sa6->sin6_addr.s6_addr16[7]
1493 kprintf("AF%d ", sa->sa_family);
1494 while (len > 0 && sa->sa_data[len-1] == 0)
1497 for (i = 0; i < len; ++i) {
1500 kprintf("%d", (unsigned char)sa->sa_data[i]);
1510 * Set up a routing table entry, normally for an interface.
1513 rtinit(struct ifaddr *ifa, int cmd, int flags)
1515 struct sockaddr *dst, *deldst, *netmask;
1516 struct mbuf *m = NULL;
1517 struct radix_node_head *rnh;
1518 struct radix_node *rn;
1519 struct rt_addrinfo rtinfo;
1522 if (flags & RTF_HOST) {
1523 dst = ifa->ifa_dstaddr;
1526 dst = ifa->ifa_addr;
1527 netmask = ifa->ifa_netmask;
1530 * If it's a delete, check that if it exists, it's on the correct
1531 * interface or we might scrub a route to another ifa which would
1532 * be confusing at best and possibly worse.
1534 if (cmd == RTM_DELETE) {
1536 * It's a delete, so it should already exist..
1537 * If it's a net, mask off the host bits
1538 * (Assuming we have a mask)
1540 if (netmask != NULL) {
1541 m = m_get(MB_DONTWAIT, MT_SONAME);
1545 deldst = mtod(m, struct sockaddr *);
1546 rt_maskedcopy(dst, deldst, netmask);
1550 * Look up an rtentry that is in the routing tree and
1551 * contains the correct info.
1553 if ((rnh = rt_tables[mycpuid][dst->sa_family]) == NULL ||
1554 (rn = rnh->rnh_lookup((char *)dst,
1555 (char *)netmask, rnh)) == NULL ||
1556 ((struct rtentry *)rn)->rt_ifa != ifa ||
1557 !sa_equal((struct sockaddr *)rn->rn_key, dst)) {
1560 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
1566 * One would think that as we are deleting, and we know
1567 * it doesn't exist, we could just return at this point
1568 * with an "ELSE" clause, but apparently not..
1570 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
1575 * Do the actual request
1577 bzero(&rtinfo, sizeof(struct rt_addrinfo));
1578 rtinfo.rti_info[RTAX_DST] = dst;
1579 rtinfo.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
1580 rtinfo.rti_info[RTAX_NETMASK] = netmask;
1581 rtinfo.rti_flags = flags | ifa->ifa_flags;
1582 rtinfo.rti_ifa = ifa;
1583 error = rtrequest1_global(cmd, &rtinfo, rtinit_rtrequest_callback, ifa);
1590 rtinit_rtrequest_callback(int cmd, int error,
1591 struct rt_addrinfo *rtinfo, struct rtentry *rt,
1594 struct ifaddr *ifa = arg;
1596 if (error == 0 && rt) {
1599 rt_newaddrmsg(cmd, ifa, error, rt);
1602 if (cmd == RTM_DELETE) {
1603 if (rt->rt_refcnt == 0) {
1611 /* This must be before ip6_init2(), which is now SI_ORDER_MIDDLE */
1612 SYSINIT(route, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0);