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 1.41 2008/11/09 10:50:15 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 <sys/mplock2.h>
94 #include <net/netmsg2.h>
97 #include <netproto/mpls/mpls.h>
100 static struct rtstatistics rtstatistics_percpu[MAXCPU];
102 #define rtstat rtstatistics_percpu[mycpuid]
104 #define rtstat rtstatistics_percpu[0]
107 struct radix_node_head *rt_tables[MAXCPU][AF_MAX+1];
108 struct lwkt_port *rt_ports[MAXCPU];
110 static void rt_maskedcopy (struct sockaddr *, struct sockaddr *,
112 static void rtable_init(void);
113 static void rtable_service_loop(void *dummy);
114 static void rtinit_rtrequest_callback(int, int, struct rt_addrinfo *,
115 struct rtentry *, void *);
118 static void rtredirect_msghandler(struct netmsg *netmsg);
119 static void rtrequest1_msghandler(struct netmsg *netmsg);
121 static void rtsearch_msghandler(struct netmsg *netmsg);
123 static void rtmask_add_msghandler(struct netmsg *netmsg);
125 static int rt_setshims(struct rtentry *, struct sockaddr **);
127 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RW, 0, "Routing");
130 static int route_debug = 1;
131 SYSCTL_INT(_net_route, OID_AUTO, route_debug, CTLFLAG_RW,
132 &route_debug, 0, "");
135 int route_assert_owner_access = 0;
136 SYSCTL_INT(_net_route, OID_AUTO, assert_owner_access, CTLFLAG_RW,
137 &route_assert_owner_access, 0, "");
140 * Initialize the route table(s) for protocol domains and
141 * create a helper thread which will be responsible for updating
142 * route table entries on each cpu.
150 for (cpu = 0; cpu < ncpus; ++cpu)
151 bzero(&rtstatistics_percpu[cpu], sizeof(struct rtstatistics));
152 rn_init(); /* initialize all zeroes, all ones, mask table */
153 rtable_init(); /* call dom_rtattach() on each cpu */
155 for (cpu = 0; cpu < ncpus; cpu++) {
156 lwkt_create(rtable_service_loop, NULL, &rtd, NULL,
157 0, cpu, "rtable_cpu %d", cpu);
158 rt_ports[cpu] = &rtd->td_msgport;
163 rtable_init_oncpu(struct netmsg *nmsg)
168 SLIST_FOREACH(dom, &domains, dom_next) {
169 if (dom->dom_rtattach) {
171 (void **)&rt_tables[cpu][dom->dom_family],
175 ifnet_forwardmsg(&nmsg->nm_lmsg, cpu + 1);
183 netmsg_init(&nmsg, NULL, &curthread->td_msgport,
184 0, rtable_init_oncpu);
185 ifnet_domsg(&nmsg.nm_lmsg, 0);
189 * Our per-cpu table management protocol thread. All route table operations
190 * are sequentially chained through all cpus starting at cpu #0 in order to
191 * maintain duplicate route tables on each cpu. Having a spearate route
192 * table management thread allows the protocol and interrupt threads to
193 * issue route table changes.
196 rtable_service_loop(void *dummy __unused)
198 struct netmsg *netmsg;
199 thread_t td = curthread;
201 get_mplock(); /* XXX is this mpsafe yet? */
203 while ((netmsg = lwkt_waitport(&td->td_msgport, 0)) != NULL) {
204 netmsg->nm_dispatch(netmsg);
209 * Routing statistics.
213 sysctl_rtstatistics(SYSCTL_HANDLER_ARGS)
217 for (cpu = 0; cpu < ncpus; ++cpu) {
218 if ((error = SYSCTL_OUT(req, &rtstatistics_percpu[cpu],
219 sizeof(struct rtstatistics))))
221 if ((error = SYSCTL_IN(req, &rtstatistics_percpu[cpu],
222 sizeof(struct rtstatistics))))
228 SYSCTL_PROC(_net_route, OID_AUTO, stats, (CTLTYPE_OPAQUE|CTLFLAG_RW),
229 0, 0, sysctl_rtstatistics, "S,rtstatistics", "Routing statistics");
231 SYSCTL_STRUCT(_net_route, OID_AUTO, stats, CTLFLAG_RW, &rtstat, rtstatistics,
232 "Routing statistics");
236 * Packet routing routines.
240 * Look up and fill in the "ro_rt" rtentry field in a route structure given
241 * an address in the "ro_dst" field. Always send a report on a miss and
242 * always clone routes.
245 rtalloc(struct route *ro)
247 rtalloc_ign(ro, 0UL);
251 * Look up and fill in the "ro_rt" rtentry field in a route structure given
252 * an address in the "ro_dst" field. Always send a report on a miss and
253 * optionally clone routes when RTF_CLONING or RTF_PRCLONING are not being
257 rtalloc_ign(struct route *ro, u_long ignoreflags)
259 if (ro->ro_rt != NULL) {
260 if (ro->ro_rt->rt_ifp != NULL && ro->ro_rt->rt_flags & RTF_UP)
265 ro->ro_rt = _rtlookup(&ro->ro_dst, RTL_REPORTMSG, ignoreflags);
269 * Look up the route that matches the given "dst" address.
271 * Route lookup can have the side-effect of creating and returning
272 * a cloned route instead when "dst" matches a cloning route and the
273 * RTF_CLONING and RTF_PRCLONING flags are not being ignored.
275 * Any route returned has its reference count incremented.
278 _rtlookup(struct sockaddr *dst, boolean_t generate_report, u_long ignore)
280 struct radix_node_head *rnh = rt_tables[mycpuid][dst->sa_family];
287 * Look up route in the radix tree.
289 rt = (struct rtentry *) rnh->rnh_matchaddr((char *)dst, rnh);
294 * Handle cloning routes.
296 if ((rt->rt_flags & ~ignore & (RTF_CLONING | RTF_PRCLONING)) != 0) {
297 struct rtentry *clonedroute;
300 clonedroute = rt; /* copy in/copy out parameter */
301 error = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0,
302 &clonedroute); /* clone the route */
303 if (error != 0) { /* cloning failed */
305 rt_dstmsg(RTM_MISS, dst, error);
307 return (rt); /* return the uncloned route */
309 if (generate_report) {
310 if (clonedroute->rt_flags & RTF_XRESOLVE)
311 rt_dstmsg(RTM_RESOLVE, dst, 0);
313 rt_rtmsg(RTM_ADD, clonedroute,
314 clonedroute->rt_ifp, 0);
316 return (clonedroute); /* return cloned route */
320 * Increment the reference count of the matched route and return.
326 rtstat.rts_unreach++;
328 rt_dstmsg(RTM_MISS, dst, 0);
333 rtfree(struct rtentry *rt)
335 if (rt->rt_cpuid == mycpuid)
338 rtfree_remote(rt, 1);
342 rtfree_oncpu(struct rtentry *rt)
344 KKASSERT(rt->rt_cpuid == mycpuid);
345 KASSERT(rt->rt_refcnt > 0, ("rtfree: rt_refcnt %ld", rt->rt_refcnt));
348 if (rt->rt_refcnt == 0) {
349 struct radix_node_head *rnh =
350 rt_tables[mycpuid][rt_key(rt)->sa_family];
353 rnh->rnh_close((struct radix_node *)rt, rnh);
354 if (!(rt->rt_flags & RTF_UP)) {
355 /* deallocate route */
356 if (rt->rt_ifa != NULL)
358 if (rt->rt_parent != NULL)
359 RTFREE(rt->rt_parent); /* recursive call! */
367 rtfree_remote_dispatch(struct netmsg *nmsg)
369 struct lwkt_msg *lmsg = &nmsg->nm_lmsg;
370 struct rtentry *rt = lmsg->u.ms_resultp;
373 lwkt_replymsg(lmsg, 0);
377 rtfree_remote(struct rtentry *rt, int allow_panic)
380 struct lwkt_msg *lmsg;
382 KKASSERT(rt->rt_cpuid != mycpuid);
384 if (route_assert_owner_access && allow_panic) {
385 panic("rt remote free rt_cpuid %d, mycpuid %d\n",
386 rt->rt_cpuid, mycpuid);
388 kprintf("rt remote free rt_cpuid %d, mycpuid %d\n",
389 rt->rt_cpuid, mycpuid);
393 netmsg_init(&nmsg, NULL, &curthread->td_msgport,
394 0, rtfree_remote_dispatch);
395 lmsg = &nmsg.nm_lmsg;
396 lmsg->u.ms_resultp = rt;
398 lwkt_domsg(rtable_portfn(rt->rt_cpuid), lmsg, 0);
402 rtredirect_oncpu(struct sockaddr *dst, struct sockaddr *gateway,
403 struct sockaddr *netmask, int flags, struct sockaddr *src)
405 struct rtentry *rt = NULL;
406 struct rt_addrinfo rtinfo;
411 /* verify the gateway is directly reachable */
412 if ((ifa = ifa_ifwithnet(gateway)) == NULL) {
418 * If the redirect isn't from our current router for this destination,
419 * it's either old or wrong.
421 if (!(flags & RTF_DONE) && /* XXX JH */
422 (rt = rtpurelookup(dst)) != NULL &&
423 (!sa_equal(src, rt->rt_gateway) || rt->rt_ifa != ifa)) {
429 * If it redirects us to ourselves, we have a routing loop,
430 * perhaps as a result of an interface going down recently.
432 if (ifa_ifwithaddr(gateway)) {
433 error = EHOSTUNREACH;
438 * Create a new entry if the lookup failed or if we got back
439 * a wildcard entry for the default route. This is necessary
440 * for hosts which use routing redirects generated by smart
441 * gateways to dynamically build the routing tables.
445 if ((rt_mask(rt) != NULL && rt_mask(rt)->sa_len < 2)) {
450 /* Ignore redirects for directly connected hosts. */
451 if (!(rt->rt_flags & RTF_GATEWAY)) {
452 error = EHOSTUNREACH;
456 if (!(rt->rt_flags & RTF_HOST) && (flags & RTF_HOST)) {
458 * Changing from a network route to a host route.
459 * Create a new host route rather than smashing the
463 flags |= RTF_GATEWAY | RTF_DYNAMIC;
464 bzero(&rtinfo, sizeof(struct rt_addrinfo));
465 rtinfo.rti_info[RTAX_DST] = dst;
466 rtinfo.rti_info[RTAX_GATEWAY] = gateway;
467 rtinfo.rti_info[RTAX_NETMASK] = netmask;
468 rtinfo.rti_flags = flags;
469 rtinfo.rti_ifa = ifa;
470 rt = NULL; /* copy-in/copy-out parameter */
471 error = rtrequest1(RTM_ADD, &rtinfo, &rt);
473 flags = rt->rt_flags;
474 stat = &rtstat.rts_dynamic;
477 * Smash the current notion of the gateway to this destination.
478 * Should check about netmask!!!
480 rt->rt_flags |= RTF_MODIFIED;
481 flags |= RTF_MODIFIED;
483 /* We only need to report rtmsg on CPU0 */
484 rt_setgate(rt, rt_key(rt), gateway,
485 mycpuid == 0 ? RTL_REPORTMSG : RTL_DONTREPORT);
487 stat = &rtstat.rts_newgateway;
495 rtstat.rts_badredirect++;
496 else if (stat != NULL)
504 struct netmsg_rtredirect {
505 struct netmsg netmsg;
506 struct sockaddr *dst;
507 struct sockaddr *gateway;
508 struct sockaddr *netmask;
510 struct sockaddr *src;
516 * Force a routing table entry to the specified
517 * destination to go through the given gateway.
518 * Normally called as a result of a routing redirect
519 * message from the network layer.
521 * N.B.: must be called at splnet
524 rtredirect(struct sockaddr *dst, struct sockaddr *gateway,
525 struct sockaddr *netmask, int flags, struct sockaddr *src)
527 struct rt_addrinfo rtinfo;
530 struct netmsg_rtredirect msg;
532 netmsg_init(&msg.netmsg, NULL, &curthread->td_msgport,
533 0, rtredirect_msghandler);
535 msg.gateway = gateway;
536 msg.netmask = netmask;
539 error = lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
541 error = rtredirect_oncpu(dst, gateway, netmask, flags, src);
543 bzero(&rtinfo, sizeof(struct rt_addrinfo));
544 rtinfo.rti_info[RTAX_DST] = dst;
545 rtinfo.rti_info[RTAX_GATEWAY] = gateway;
546 rtinfo.rti_info[RTAX_NETMASK] = netmask;
547 rtinfo.rti_info[RTAX_AUTHOR] = src;
548 rt_missmsg(RTM_REDIRECT, &rtinfo, flags, error);
554 rtredirect_msghandler(struct netmsg *netmsg)
556 struct netmsg_rtredirect *msg = (void *)netmsg;
559 rtredirect_oncpu(msg->dst, msg->gateway, msg->netmask,
560 msg->flags, msg->src);
561 nextcpu = mycpuid + 1;
563 lwkt_forwardmsg(rtable_portfn(nextcpu), &netmsg->nm_lmsg);
565 lwkt_replymsg(&netmsg->nm_lmsg, 0);
571 * Routing table ioctl interface.
574 rtioctl(u_long req, caddr_t data, struct ucred *cred)
577 /* Multicast goop, grrr... */
578 return mrt_ioctl ? mrt_ioctl(req, data) : EOPNOTSUPP;
585 ifa_ifwithroute(int flags, struct sockaddr *dst, struct sockaddr *gateway)
589 if (!(flags & RTF_GATEWAY)) {
591 * If we are adding a route to an interface,
592 * and the interface is a point-to-point link,
593 * we should search for the destination
594 * as our clue to the interface. Otherwise
595 * we can use the local address.
598 if (flags & RTF_HOST) {
599 ifa = ifa_ifwithdstaddr(dst);
602 ifa = ifa_ifwithaddr(gateway);
605 * If we are adding a route to a remote net
606 * or host, the gateway may still be on the
607 * other end of a pt to pt link.
609 ifa = ifa_ifwithdstaddr(gateway);
612 ifa = ifa_ifwithnet(gateway);
616 rt = rtpurelookup(gateway);
620 if ((ifa = rt->rt_ifa) == NULL)
623 if (ifa->ifa_addr->sa_family != dst->sa_family) {
624 struct ifaddr *oldifa = ifa;
626 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
633 static int rt_fixdelete (struct radix_node *, void *);
634 static int rt_fixchange (struct radix_node *, void *);
638 struct radix_node_head *rnh;
642 * Set rtinfo->rti_ifa and rtinfo->rti_ifp.
645 rt_getifa(struct rt_addrinfo *rtinfo)
647 struct sockaddr *gateway = rtinfo->rti_info[RTAX_GATEWAY];
648 struct sockaddr *dst = rtinfo->rti_info[RTAX_DST];
649 struct sockaddr *ifaaddr = rtinfo->rti_info[RTAX_IFA];
650 int flags = rtinfo->rti_flags;
653 * ifp may be specified by sockaddr_dl
654 * when protocol address is ambiguous.
656 if (rtinfo->rti_ifp == NULL) {
657 struct sockaddr *ifpaddr;
659 ifpaddr = rtinfo->rti_info[RTAX_IFP];
660 if (ifpaddr != NULL && ifpaddr->sa_family == AF_LINK) {
663 ifa = ifa_ifwithnet(ifpaddr);
665 rtinfo->rti_ifp = ifa->ifa_ifp;
669 if (rtinfo->rti_ifa == NULL && ifaaddr != NULL)
670 rtinfo->rti_ifa = ifa_ifwithaddr(ifaaddr);
671 if (rtinfo->rti_ifa == NULL) {
674 sa = ifaaddr != NULL ? ifaaddr :
675 (gateway != NULL ? gateway : dst);
676 if (sa != NULL && rtinfo->rti_ifp != NULL)
677 rtinfo->rti_ifa = ifaof_ifpforaddr(sa, rtinfo->rti_ifp);
678 else if (dst != NULL && gateway != NULL)
679 rtinfo->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
681 rtinfo->rti_ifa = ifa_ifwithroute(flags, sa, sa);
683 if (rtinfo->rti_ifa == NULL)
684 return (ENETUNREACH);
686 if (rtinfo->rti_ifp == NULL)
687 rtinfo->rti_ifp = rtinfo->rti_ifa->ifa_ifp;
692 * Do appropriate manipulations of a routing tree given
693 * all the bits of info needed
698 struct sockaddr *dst,
699 struct sockaddr *gateway,
700 struct sockaddr *netmask,
702 struct rtentry **ret_nrt)
704 struct rt_addrinfo rtinfo;
706 bzero(&rtinfo, sizeof(struct rt_addrinfo));
707 rtinfo.rti_info[RTAX_DST] = dst;
708 rtinfo.rti_info[RTAX_GATEWAY] = gateway;
709 rtinfo.rti_info[RTAX_NETMASK] = netmask;
710 rtinfo.rti_flags = flags;
711 return rtrequest1(req, &rtinfo, ret_nrt);
717 struct sockaddr *dst,
718 struct sockaddr *gateway,
719 struct sockaddr *netmask,
722 struct rt_addrinfo rtinfo;
724 bzero(&rtinfo, sizeof(struct rt_addrinfo));
725 rtinfo.rti_info[RTAX_DST] = dst;
726 rtinfo.rti_info[RTAX_GATEWAY] = gateway;
727 rtinfo.rti_info[RTAX_NETMASK] = netmask;
728 rtinfo.rti_flags = flags;
729 return rtrequest1_global(req, &rtinfo, NULL, NULL);
735 struct netmsg netmsg;
737 struct rt_addrinfo *rtinfo;
738 rtrequest1_callback_func_t callback;
745 rtrequest1_global(int req, struct rt_addrinfo *rtinfo,
746 rtrequest1_callback_func_t callback, void *arg)
750 struct netmsg_rtq msg;
752 netmsg_init(&msg.netmsg, NULL, &curthread->td_msgport,
753 0, rtrequest1_msghandler);
754 msg.netmsg.nm_lmsg.ms_error = -1;
757 msg.callback = callback;
759 error = lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
761 struct rtentry *rt = NULL;
763 error = rtrequest1(req, rtinfo, &rt);
767 callback(req, error, rtinfo, rt, arg);
773 * Handle a route table request on the current cpu. Since the route table's
774 * are supposed to be identical on each cpu, an error occuring later in the
775 * message chain is considered system-fatal.
780 rtrequest1_msghandler(struct netmsg *netmsg)
782 struct netmsg_rtq *msg = (void *)netmsg;
783 struct rt_addrinfo rtinfo;
784 struct rtentry *rt = NULL;
789 * Copy the rtinfo. We need to make sure that the original
790 * rtinfo, which is setup by the caller, in the netmsg will
791 * _not_ be changed; else the next CPU on the netmsg forwarding
792 * path will see a different rtinfo than what this CPU has seen.
794 rtinfo = *msg->rtinfo;
796 error = rtrequest1(msg->req, &rtinfo, &rt);
800 msg->callback(msg->req, error, &rtinfo, rt, msg->arg);
803 * RTM_DELETE's are propogated even if an error occurs, since a
804 * cloned route might be undergoing deletion and cloned routes
805 * are not necessarily replicated. An overall error is returned
806 * only if no cpus have the route in question.
808 if (msg->netmsg.nm_lmsg.ms_error < 0 || error == 0)
809 msg->netmsg.nm_lmsg.ms_error = error;
811 nextcpu = mycpuid + 1;
812 if (error && msg->req != RTM_DELETE) {
814 panic("rtrequest1_msghandler: rtrequest table "
815 "error was not on cpu #0");
817 lwkt_replymsg(&msg->netmsg.nm_lmsg, error);
818 } else if (nextcpu < ncpus) {
819 lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg);
821 lwkt_replymsg(&msg->netmsg.nm_lmsg,
822 msg->netmsg.nm_lmsg.ms_error);
829 rtrequest1(int req, struct rt_addrinfo *rtinfo, struct rtentry **ret_nrt)
831 struct sockaddr *dst = rtinfo->rti_info[RTAX_DST];
833 struct radix_node *rn;
834 struct radix_node_head *rnh;
836 struct sockaddr *ndst;
840 #define gotoerr(x) { error = x ; goto bad; }
844 rt_addrinfo_print(req, rtinfo);
849 * Find the correct routing tree to use for this Address Family
851 if ((rnh = rt_tables[mycpuid][dst->sa_family]) == NULL)
852 gotoerr(EAFNOSUPPORT);
855 * If we are adding a host route then we don't want to put
856 * a netmask in the tree, nor do we want to clone it.
858 if (rtinfo->rti_flags & RTF_HOST) {
859 rtinfo->rti_info[RTAX_NETMASK] = NULL;
860 rtinfo->rti_flags &= ~(RTF_CLONING | RTF_PRCLONING);
865 /* Remove the item from the tree. */
866 rn = rnh->rnh_deladdr((char *)rtinfo->rti_info[RTAX_DST],
867 (char *)rtinfo->rti_info[RTAX_NETMASK],
871 KASSERT(!(rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)),
872 ("rnh_deladdr returned flags 0x%x", rn->rn_flags));
873 rt = (struct rtentry *)rn;
875 /* ref to prevent a deletion race */
878 /* Free any routes cloned from this one. */
879 if ((rt->rt_flags & (RTF_CLONING | RTF_PRCLONING)) &&
880 rt_mask(rt) != NULL) {
881 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt),
886 if (rt->rt_gwroute != NULL) {
887 RTFREE(rt->rt_gwroute);
888 rt->rt_gwroute = NULL;
892 * NB: RTF_UP must be set during the search above,
893 * because we might delete the last ref, causing
894 * rt to get freed prematurely.
896 rt->rt_flags &= ~RTF_UP;
900 rt_print(rtinfo, rt);
903 /* Give the protocol a chance to keep things in sync. */
904 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
905 ifa->ifa_rtrequest(RTM_DELETE, rt, rtinfo);
908 * If the caller wants it, then it can have it,
909 * but it's up to it to free the rtentry as we won't be
912 KASSERT(rt->rt_refcnt >= 0,
913 ("rtrequest1(DELETE): refcnt %ld", rt->rt_refcnt));
914 if (ret_nrt != NULL) {
915 /* leave ref intact for return */
918 /* deref / attempt to destroy */
924 if (ret_nrt == NULL || (rt = *ret_nrt) == NULL)
928 rt->rt_flags & ~(RTF_CLONING | RTF_PRCLONING | RTF_STATIC);
929 rtinfo->rti_flags |= RTF_WASCLONED;
930 rtinfo->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
931 if ((rtinfo->rti_info[RTAX_NETMASK] = rt->rt_genmask) == NULL)
932 rtinfo->rti_flags |= RTF_HOST;
933 rtinfo->rti_info[RTAX_MPLS1] = rt->rt_shim[0];
934 rtinfo->rti_info[RTAX_MPLS2] = rt->rt_shim[1];
935 rtinfo->rti_info[RTAX_MPLS3] = rt->rt_shim[2];
939 KASSERT(!(rtinfo->rti_flags & RTF_GATEWAY) ||
940 rtinfo->rti_info[RTAX_GATEWAY] != NULL,
941 ("rtrequest: GATEWAY but no gateway"));
943 if (rtinfo->rti_ifa == NULL && (error = rt_getifa(rtinfo)))
945 ifa = rtinfo->rti_ifa;
947 R_Malloc(rt, struct rtentry *, sizeof(struct rtentry));
950 bzero(rt, sizeof(struct rtentry));
951 rt->rt_flags = RTF_UP | rtinfo->rti_flags;
952 rt->rt_cpuid = mycpuid;
954 if (mycpuid != 0 && req == RTM_ADD) {
955 /* For RTM_ADD, we have already sent rtmsg on CPU0. */
956 reportmsg = RTL_DONTREPORT;
959 * For RTM_ADD, we only send rtmsg on CPU0.
960 * For RTM_RESOLVE, we always send rtmsg. XXX
962 reportmsg = RTL_REPORTMSG;
964 error = rt_setgate(rt, dst, rtinfo->rti_info[RTAX_GATEWAY],
972 if (rtinfo->rti_info[RTAX_NETMASK] != NULL)
973 rt_maskedcopy(dst, ndst,
974 rtinfo->rti_info[RTAX_NETMASK]);
976 bcopy(dst, ndst, dst->sa_len);
978 if (rtinfo->rti_info[RTAX_MPLS1] != NULL)
979 rt_setshims(rt, rtinfo->rti_info);
982 * Note that we now have a reference to the ifa.
983 * This moved from below so that rnh->rnh_addaddr() can
984 * examine the ifa and ifa->ifa_ifp if it so desires.
988 rt->rt_ifp = ifa->ifa_ifp;
989 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
991 rn = rnh->rnh_addaddr((char *)ndst,
992 (char *)rtinfo->rti_info[RTAX_NETMASK],
995 struct rtentry *oldrt;
998 * We already have one of these in the tree.
999 * We do a special hack: if the old route was
1000 * cloned, then we blow it away and try
1001 * re-inserting the new one.
1003 oldrt = rtpurelookup(ndst);
1004 if (oldrt != NULL) {
1006 if (oldrt->rt_flags & RTF_WASCLONED) {
1007 rtrequest(RTM_DELETE, rt_key(oldrt),
1010 oldrt->rt_flags, NULL);
1011 rn = rnh->rnh_addaddr((char *)ndst,
1013 rtinfo->rti_info[RTAX_NETMASK],
1020 * If it still failed to go into the tree,
1021 * then un-make it (this should be a function).
1024 if (rt->rt_gwroute != NULL)
1025 rtfree(rt->rt_gwroute);
1033 * If we got here from RESOLVE, then we are cloning
1034 * so clone the rest, and note that we
1035 * are a clone (and increment the parent's references)
1037 if (req == RTM_RESOLVE) {
1038 rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
1039 rt->rt_rmx.rmx_pksent = 0; /* reset packet counter */
1040 if ((*ret_nrt)->rt_flags &
1041 (RTF_CLONING | RTF_PRCLONING)) {
1042 rt->rt_parent = *ret_nrt;
1043 (*ret_nrt)->rt_refcnt++;
1048 * if this protocol has something to add to this then
1049 * allow it to do that as well.
1051 if (ifa->ifa_rtrequest != NULL)
1052 ifa->ifa_rtrequest(req, rt, rtinfo);
1055 * We repeat the same procedure from rt_setgate() here because
1056 * it doesn't fire when we call it there because the node
1057 * hasn't been added to the tree yet.
1059 if (req == RTM_ADD && !(rt->rt_flags & RTF_HOST) &&
1060 rt_mask(rt) != NULL) {
1061 struct rtfc_arg arg = { rt, rnh };
1063 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt),
1064 (char *)rt_mask(rt),
1065 rt_fixchange, &arg);
1070 rt_print(rtinfo, rt);
1073 * Return the resulting rtentry,
1074 * increasing the number of references by one.
1076 if (ret_nrt != NULL) {
1088 kprintf("rti %p failed error %d\n", rtinfo, error);
1090 kprintf("rti %p succeeded\n", rtinfo);
1098 * Called from rtrequest(RTM_DELETE, ...) to fix up the route's ``family''
1099 * (i.e., the routes related to it by the operation of cloning). This
1100 * routine is iterated over all potential former-child-routes by way of
1101 * rnh->rnh_walktree_from() above, and those that actually are children of
1102 * the late parent (passed in as VP here) are themselves deleted.
1105 rt_fixdelete(struct radix_node *rn, void *vp)
1107 struct rtentry *rt = (struct rtentry *)rn;
1108 struct rtentry *rt0 = vp;
1110 if (rt->rt_parent == rt0 &&
1111 !(rt->rt_flags & (RTF_PINNED | RTF_CLONING | RTF_PRCLONING))) {
1112 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt),
1113 rt->rt_flags, NULL);
1119 * This routine is called from rt_setgate() to do the analogous thing for
1120 * adds and changes. There is the added complication in this case of a
1121 * middle insert; i.e., insertion of a new network route between an older
1122 * network route and (cloned) host routes. For this reason, a simple check
1123 * of rt->rt_parent is insufficient; each candidate route must be tested
1124 * against the (mask, value) of the new route (passed as before in vp)
1125 * to see if the new route matches it.
1127 * XXX - it may be possible to do fixdelete() for changes and reserve this
1128 * routine just for adds. I'm not sure why I thought it was necessary to do
1132 static int rtfcdebug = 0;
1136 rt_fixchange(struct radix_node *rn, void *vp)
1138 struct rtentry *rt = (struct rtentry *)rn;
1139 struct rtfc_arg *ap = vp;
1140 struct rtentry *rt0 = ap->rt0;
1141 struct radix_node_head *rnh = ap->rnh;
1142 u_char *xk1, *xm1, *xk2, *xmp;
1147 kprintf("rt_fixchange: rt %p, rt0 %p\n", rt, rt0);
1150 if (rt->rt_parent == NULL ||
1151 (rt->rt_flags & (RTF_PINNED | RTF_CLONING | RTF_PRCLONING))) {
1153 if (rtfcdebug) kprintf("no parent, pinned or cloning\n");
1158 if (rt->rt_parent == rt0) {
1160 if (rtfcdebug) kprintf("parent match\n");
1162 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt),
1163 rt->rt_flags, NULL);
1167 * There probably is a function somewhere which does this...
1168 * if not, there should be.
1170 len = imin(rt_key(rt0)->sa_len, rt_key(rt)->sa_len);
1172 xk1 = (u_char *)rt_key(rt0);
1173 xm1 = (u_char *)rt_mask(rt0);
1174 xk2 = (u_char *)rt_key(rt);
1176 /* avoid applying a less specific route */
1177 xmp = (u_char *)rt_mask(rt->rt_parent);
1178 mlen = rt_key(rt->rt_parent)->sa_len;
1179 if (mlen > rt_key(rt0)->sa_len) {
1182 kprintf("rt_fixchange: inserting a less "
1183 "specific route\n");
1187 for (i = rnh->rnh_treetop->rn_offset; i < mlen; i++) {
1188 if ((xmp[i] & ~(xmp[i] ^ xm1[i])) != xmp[i]) {
1191 kprintf("rt_fixchange: inserting a less "
1192 "specific route\n");
1198 for (i = rnh->rnh_treetop->rn_offset; i < len; i++) {
1199 if ((xk2[i] & xm1[i]) != xk1[i]) {
1201 if (rtfcdebug) kprintf("no match\n");
1208 * OK, this node is a clone, and matches the node currently being
1209 * changed/added under the node's mask. So, get rid of it.
1212 if (rtfcdebug) kprintf("deleting\n");
1214 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt),
1215 rt->rt_flags, NULL);
1218 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
1221 rt_setgate(struct rtentry *rt0, struct sockaddr *dst, struct sockaddr *gate,
1222 boolean_t generate_report)
1224 char *space, *oldspace;
1225 int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len);
1226 struct rtentry *rt = rt0;
1227 struct radix_node_head *rnh = rt_tables[mycpuid][dst->sa_family];
1230 * A host route with the destination equal to the gateway
1231 * will interfere with keeping LLINFO in the routing
1232 * table, so disallow it.
1234 if (((rt0->rt_flags & (RTF_HOST | RTF_GATEWAY | RTF_LLINFO)) ==
1235 (RTF_HOST | RTF_GATEWAY)) &&
1236 dst->sa_len == gate->sa_len &&
1237 sa_equal(dst, gate)) {
1239 * The route might already exist if this is an RTM_CHANGE
1240 * or a routing redirect, so try to delete it.
1242 if (rt_key(rt0) != NULL)
1243 rtrequest(RTM_DELETE, rt_key(rt0), rt0->rt_gateway,
1244 rt_mask(rt0), rt0->rt_flags, NULL);
1245 return EADDRNOTAVAIL;
1249 * Both dst and gateway are stored in the same malloc'ed chunk
1250 * (If I ever get my hands on....)
1251 * if we need to malloc a new chunk, then keep the old one around
1252 * till we don't need it any more.
1254 if (rt->rt_gateway == NULL || glen > ROUNDUP(rt->rt_gateway->sa_len)) {
1255 oldspace = (char *)rt_key(rt);
1256 R_Malloc(space, char *, dlen + glen);
1259 rt->rt_nodes->rn_key = space;
1261 space = (char *)rt_key(rt); /* Just use the old space. */
1265 /* Set the gateway value. */
1266 rt->rt_gateway = (struct sockaddr *)(space + dlen);
1267 bcopy(gate, rt->rt_gateway, glen);
1269 if (oldspace != NULL) {
1271 * If we allocated a new chunk, preserve the original dst.
1272 * This way, rt_setgate() really just sets the gate
1273 * and leaves the dst field alone.
1275 bcopy(dst, space, dlen);
1280 * If there is already a gwroute, it's now almost definitely wrong
1283 if (rt->rt_gwroute != NULL) {
1284 RTFREE(rt->rt_gwroute);
1285 rt->rt_gwroute = NULL;
1287 if (rt->rt_flags & RTF_GATEWAY) {
1289 * Cloning loop avoidance: In the presence of
1290 * protocol-cloning and bad configuration, it is
1291 * possible to get stuck in bottomless mutual recursion
1292 * (rtrequest rt_setgate rtlookup). We avoid this
1293 * by not allowing protocol-cloning to operate for
1294 * gateways (which is probably the correct choice
1295 * anyway), and avoid the resulting reference loops
1296 * by disallowing any route to run through itself as
1297 * a gateway. This is obviously mandatory when we
1298 * get rt->rt_output().
1300 * This breaks TTCP for hosts outside the gateway! XXX JH
1302 rt->rt_gwroute = _rtlookup(gate, generate_report,
1304 if (rt->rt_gwroute == rt) {
1305 rt->rt_gwroute = NULL;
1307 return EDQUOT; /* failure */
1312 * This isn't going to do anything useful for host routes, so
1313 * don't bother. Also make sure we have a reasonable mask
1314 * (we don't yet have one during adds).
1316 if (!(rt->rt_flags & RTF_HOST) && rt_mask(rt) != NULL) {
1317 struct rtfc_arg arg = { rt, rnh };
1319 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt),
1320 (char *)rt_mask(rt),
1321 rt_fixchange, &arg);
1329 struct sockaddr *src,
1330 struct sockaddr *dst,
1331 struct sockaddr *netmask)
1333 u_char *cp1 = (u_char *)src;
1334 u_char *cp2 = (u_char *)dst;
1335 u_char *cp3 = (u_char *)netmask;
1336 u_char *cplim = cp2 + *cp3;
1337 u_char *cplim2 = cp2 + *cp1;
1339 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
1344 *cp2++ = *cp1++ & *cp3++;
1346 bzero(cp2, cplim2 - cp2);
1350 rt_llroute(struct sockaddr *dst, struct rtentry *rt0, struct rtentry **drt)
1352 struct rtentry *up_rt, *rt;
1354 if (!(rt0->rt_flags & RTF_UP)) {
1355 up_rt = rtlookup(dst);
1357 return (EHOSTUNREACH);
1361 if (up_rt->rt_flags & RTF_GATEWAY) {
1362 if (up_rt->rt_gwroute == NULL) {
1363 up_rt->rt_gwroute = rtlookup(up_rt->rt_gateway);
1364 if (up_rt->rt_gwroute == NULL)
1365 return (EHOSTUNREACH);
1366 } else if (!(up_rt->rt_gwroute->rt_flags & RTF_UP)) {
1367 rtfree(up_rt->rt_gwroute);
1368 up_rt->rt_gwroute = rtlookup(up_rt->rt_gateway);
1369 if (up_rt->rt_gwroute == NULL)
1370 return (EHOSTUNREACH);
1372 rt = up_rt->rt_gwroute;
1375 if (rt->rt_flags & RTF_REJECT &&
1376 (rt->rt_rmx.rmx_expire == 0 || /* rt doesn't expire */
1377 time_second < rt->rt_rmx.rmx_expire)) /* rt not expired */
1378 return (rt->rt_flags & RTF_HOST ? EHOSTDOWN : EHOSTUNREACH);
1384 rt_setshims(struct rtentry *rt, struct sockaddr **rt_shim){
1387 for (i=0; i<3; i++) {
1388 struct sockaddr *shim = rt_shim[RTAX_MPLS1 + i];
1394 shimlen = ROUNDUP(shim->sa_len);
1395 R_Malloc(rt->rt_shim[i], struct sockaddr *, shimlen);
1396 bcopy(shim, rt->rt_shim[i], shimlen);
1405 * Print out a route table entry
1408 rt_print(struct rt_addrinfo *rtinfo, struct rtentry *rn)
1410 kprintf("rti %p cpu %d route %p flags %08lx: ",
1411 rtinfo, mycpuid, rn, rn->rt_flags);
1412 sockaddr_print(rt_key(rn));
1414 sockaddr_print(rt_mask(rn));
1416 sockaddr_print(rn->rt_gateway);
1417 kprintf(" ifc \"%s\"", rn->rt_ifp ? rn->rt_ifp->if_dname : "?");
1418 kprintf(" ifa %p\n", rn->rt_ifa);
1422 rt_addrinfo_print(int cmd, struct rt_addrinfo *rti)
1428 if (cmd == RTM_DELETE && route_debug > 1)
1429 print_backtrace(-1);
1443 kprintf("C%02d ", cmd);
1446 kprintf("rti %p cpu %d ", rti, mycpuid);
1447 for (i = 0; i < rti->rti_addrs; ++i) {
1448 if (rti->rti_info[i] == NULL)
1478 kprintf("(?%02d ", i);
1481 sockaddr_print(rti->rti_info[i]);
1489 sockaddr_print(struct sockaddr *sa)
1491 struct sockaddr_in *sa4;
1492 struct sockaddr_in6 *sa6;
1501 len = sa->sa_len - offsetof(struct sockaddr, sa_data[0]);
1503 switch(sa->sa_family) {
1507 switch(sa->sa_family) {
1509 sa4 = (struct sockaddr_in *)sa;
1510 kprintf("INET %d %d.%d.%d.%d",
1511 ntohs(sa4->sin_port),
1512 (ntohl(sa4->sin_addr.s_addr) >> 24) & 255,
1513 (ntohl(sa4->sin_addr.s_addr) >> 16) & 255,
1514 (ntohl(sa4->sin_addr.s_addr) >> 8) & 255,
1515 (ntohl(sa4->sin_addr.s_addr) >> 0) & 255
1519 sa6 = (struct sockaddr_in6 *)sa;
1520 kprintf("INET6 %d %04x:%04x%04x:%04x:%04x:%04x:%04x:%04x",
1521 ntohs(sa6->sin6_port),
1522 sa6->sin6_addr.s6_addr16[0],
1523 sa6->sin6_addr.s6_addr16[1],
1524 sa6->sin6_addr.s6_addr16[2],
1525 sa6->sin6_addr.s6_addr16[3],
1526 sa6->sin6_addr.s6_addr16[4],
1527 sa6->sin6_addr.s6_addr16[5],
1528 sa6->sin6_addr.s6_addr16[6],
1529 sa6->sin6_addr.s6_addr16[7]
1533 kprintf("AF%d ", sa->sa_family);
1534 while (len > 0 && sa->sa_data[len-1] == 0)
1537 for (i = 0; i < len; ++i) {
1540 kprintf("%d", (unsigned char)sa->sa_data[i]);
1550 * Set up a routing table entry, normally for an interface.
1553 rtinit(struct ifaddr *ifa, int cmd, int flags)
1555 struct sockaddr *dst, *deldst, *netmask;
1556 struct mbuf *m = NULL;
1557 struct radix_node_head *rnh;
1558 struct radix_node *rn;
1559 struct rt_addrinfo rtinfo;
1562 if (flags & RTF_HOST) {
1563 dst = ifa->ifa_dstaddr;
1566 dst = ifa->ifa_addr;
1567 netmask = ifa->ifa_netmask;
1570 * If it's a delete, check that if it exists, it's on the correct
1571 * interface or we might scrub a route to another ifa which would
1572 * be confusing at best and possibly worse.
1574 if (cmd == RTM_DELETE) {
1576 * It's a delete, so it should already exist..
1577 * If it's a net, mask off the host bits
1578 * (Assuming we have a mask)
1580 if (netmask != NULL) {
1581 m = m_get(MB_DONTWAIT, MT_SONAME);
1585 deldst = mtod(m, struct sockaddr *);
1586 rt_maskedcopy(dst, deldst, netmask);
1590 * Look up an rtentry that is in the routing tree and
1591 * contains the correct info.
1593 if ((rnh = rt_tables[mycpuid][dst->sa_family]) == NULL ||
1594 (rn = rnh->rnh_lookup((char *)dst,
1595 (char *)netmask, rnh)) == NULL ||
1596 ((struct rtentry *)rn)->rt_ifa != ifa ||
1597 !sa_equal((struct sockaddr *)rn->rn_key, dst)) {
1600 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
1606 * One would think that as we are deleting, and we know
1607 * it doesn't exist, we could just return at this point
1608 * with an "ELSE" clause, but apparently not..
1610 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
1615 * Do the actual request
1617 bzero(&rtinfo, sizeof(struct rt_addrinfo));
1618 rtinfo.rti_info[RTAX_DST] = dst;
1619 rtinfo.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
1620 rtinfo.rti_info[RTAX_NETMASK] = netmask;
1621 rtinfo.rti_flags = flags | ifa->ifa_flags;
1622 rtinfo.rti_ifa = ifa;
1623 error = rtrequest1_global(cmd, &rtinfo, rtinit_rtrequest_callback, ifa);
1630 rtinit_rtrequest_callback(int cmd, int error,
1631 struct rt_addrinfo *rtinfo, struct rtentry *rt,
1634 struct ifaddr *ifa = arg;
1636 if (error == 0 && rt) {
1639 rt_newaddrmsg(cmd, ifa, error, rt);
1642 if (cmd == RTM_DELETE) {
1643 if (rt->rt_refcnt == 0) {
1652 struct netmsg netmsg;
1654 struct rt_addrinfo *rtinfo;
1655 rtsearch_callback_func_t callback;
1657 boolean_t exact_match;
1662 rtsearch_global(int req, struct rt_addrinfo *rtinfo,
1663 rtsearch_callback_func_t callback, void *arg,
1664 boolean_t exact_match)
1666 struct netmsg_rts msg;
1668 netmsg_init(&msg.netmsg, NULL, &curthread->td_msgport,
1669 0, rtsearch_msghandler);
1671 msg.rtinfo = rtinfo;
1672 msg.callback = callback;
1674 msg.exact_match = exact_match;
1676 return lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
1680 rtsearch_msghandler(struct netmsg *netmsg)
1682 struct netmsg_rts *msg = (void *)netmsg;
1683 struct rt_addrinfo rtinfo;
1684 struct radix_node_head *rnh;
1689 * Copy the rtinfo. We need to make sure that the original
1690 * rtinfo, which is setup by the caller, in the netmsg will
1691 * _not_ be changed; else the next CPU on the netmsg forwarding
1692 * path will see a different rtinfo than what this CPU has seen.
1694 rtinfo = *msg->rtinfo;
1697 * Find the correct routing tree to use for this Address Family
1699 if ((rnh = rt_tables[mycpuid][rtinfo.rti_dst->sa_family]) == NULL) {
1701 panic("partially initialized routing tables\n");
1702 lwkt_replymsg(&msg->netmsg.nm_lmsg, EAFNOSUPPORT);
1707 * Correct rtinfo for the host route searching.
1709 if (rtinfo.rti_flags & RTF_HOST) {
1710 rtinfo.rti_netmask = NULL;
1711 rtinfo.rti_flags &= ~(RTF_CLONING | RTF_PRCLONING);
1714 rt = (struct rtentry *)
1715 rnh->rnh_lookup((char *)rtinfo.rti_dst,
1716 (char *)rtinfo.rti_netmask, rnh);
1719 * If we are asked to do the "exact match", we need to make sure
1720 * that host route searching got a host route while a network
1721 * route searching got a network route.
1723 if (rt != NULL && msg->exact_match &&
1724 ((rt->rt_flags ^ rtinfo.rti_flags) & RTF_HOST))
1729 * No matching routes have been found, don't count this
1730 * as a critical error (here, we set 'error' to 0), just
1731 * keep moving on, since at least prcloned routes are not
1732 * duplicated onto each CPU.
1739 error = msg->callback(msg->req, &rtinfo, rt, msg->arg,
1743 if (error == EJUSTRETURN) {
1744 lwkt_replymsg(&msg->netmsg.nm_lmsg, 0);
1749 nextcpu = mycpuid + 1;
1751 KKASSERT(msg->found_cnt > 0);
1754 * Under following cases, unrecoverable error has
1756 * o Request is RTM_GET
1757 * o The first time that we find the route, but the
1758 * modification fails.
1760 if (msg->req != RTM_GET && msg->found_cnt > 1) {
1761 panic("rtsearch_msghandler: unrecoverable error "
1764 lwkt_replymsg(&msg->netmsg.nm_lmsg, error);
1765 } else if (nextcpu < ncpus) {
1766 lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg);
1768 if (msg->found_cnt == 0) {
1769 /* The requested route was never seen ... */
1772 lwkt_replymsg(&msg->netmsg.nm_lmsg, error);
1777 rtmask_add_global(struct sockaddr *mask)
1781 netmsg_init(&nmsg, NULL, &curthread->td_msgport,
1782 0, rtmask_add_msghandler);
1783 nmsg.nm_lmsg.u.ms_resultp = mask;
1785 return lwkt_domsg(rtable_portfn(0), &nmsg.nm_lmsg, 0);
1789 _rtmask_lookup(struct sockaddr *mask, boolean_t search)
1791 struct radix_node *n;
1793 #define clen(s) (*(u_char *)(s))
1794 n = rn_addmask((char *)mask, search, 1);
1796 mask->sa_len >= clen(n->rn_key) &&
1797 bcmp((char *)mask + 1,
1798 (char *)n->rn_key + 1, clen(n->rn_key) - 1) == 0) {
1799 return (struct sockaddr *)n->rn_key;
1807 rtmask_add_msghandler(struct netmsg *nmsg)
1809 struct lwkt_msg *lmsg = &nmsg->nm_lmsg;
1810 struct sockaddr *mask = lmsg->u.ms_resultp;
1811 int error = 0, nextcpu;
1813 if (rtmask_lookup(mask) == NULL)
1816 nextcpu = mycpuid + 1;
1817 if (!error && nextcpu < ncpus)
1818 lwkt_forwardmsg(rtable_portfn(nextcpu), lmsg);
1820 lwkt_replymsg(lmsg, error);
1823 /* This must be before ip6_init2(), which is now SI_ORDER_MIDDLE */
1824 SYSINIT(route, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0);