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 1.35 2008/07/07 22:02:10 nant 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, "");
132 * Initialize the route table(s) for protocol domains and
133 * create a helper thread which will be responsible for updating
134 * route table entries on each cpu.
142 for (cpu
= 0; cpu
< ncpus
; ++cpu
)
143 bzero(&rtstatistics_percpu
[cpu
], sizeof(struct rtstatistics
));
144 rn_init(); /* initialize all zeroes, all ones, mask table */
146 for (cpu
= 0; cpu
< ncpus
; cpu
++) {
147 lwkt_migratecpu(cpu
);
149 lwkt_create(rtable_service_loop
, NULL
, &rtd
, NULL
,
150 TDF_STOPREQ
, cpu
, "rtable_cpu %d", cpu
);
151 rt_ports
[cpu
] = &rtd
->td_msgport
;
154 lwkt_migratecpu(origcpu
);
162 SLIST_FOREACH(dom
, &domains
, dom_next
) {
163 if (dom
->dom_rtattach
) {
165 (void **)&rt_tables
[mycpuid
][dom
->dom_family
],
172 * Our per-cpu table management protocol thread. All route table operations
173 * are sequentially chained through all cpus starting at cpu #0 in order to
174 * maintain duplicate route tables on each cpu. Having a spearate route
175 * table management thread allows the protocol and interrupt threads to
176 * issue route table changes.
179 rtable_service_loop(void *dummy __unused
)
181 struct netmsg
*netmsg
;
182 thread_t td
= curthread
;
184 while ((netmsg
= lwkt_waitport(&td
->td_msgport
, 0)) != NULL
) {
185 netmsg
->nm_dispatch(netmsg
);
190 * Routing statistics.
194 sysctl_rtstatistics(SYSCTL_HANDLER_ARGS
)
198 for (cpu
= 0; cpu
< ncpus
; ++cpu
) {
199 if ((error
= SYSCTL_OUT(req
, &rtstatistics_percpu
[cpu
],
200 sizeof(struct rtstatistics
))))
202 if ((error
= SYSCTL_IN(req
, &rtstatistics_percpu
[cpu
],
203 sizeof(struct rtstatistics
))))
209 SYSCTL_PROC(_net_route
, OID_AUTO
, stats
, (CTLTYPE_OPAQUE
|CTLFLAG_RW
),
210 0, 0, sysctl_rtstatistics
, "S,rtstatistics", "Routing statistics");
212 SYSCTL_STRUCT(_net_route
, OID_AUTO
, stats
, CTLFLAG_RW
, &rtstat
, rtstatistics
,
213 "Routing statistics");
217 * Packet routing routines.
221 * Look up and fill in the "ro_rt" rtentry field in a route structure given
222 * an address in the "ro_dst" field. Always send a report on a miss and
223 * always clone routes.
226 rtalloc(struct route
*ro
)
228 rtalloc_ign(ro
, 0UL);
232 * Look up and fill in the "ro_rt" rtentry field in a route structure given
233 * an address in the "ro_dst" field. Always send a report on a miss and
234 * optionally clone routes when RTF_CLONING or RTF_PRCLONING are not being
238 rtalloc_ign(struct route
*ro
, u_long ignoreflags
)
240 if (ro
->ro_rt
!= NULL
) {
241 if (ro
->ro_rt
->rt_ifp
!= NULL
&& ro
->ro_rt
->rt_flags
& RTF_UP
)
246 ro
->ro_rt
= _rtlookup(&ro
->ro_dst
, RTL_REPORTMSG
, ignoreflags
);
250 * Look up the route that matches the given "dst" address.
252 * Route lookup can have the side-effect of creating and returning
253 * a cloned route instead when "dst" matches a cloning route and the
254 * RTF_CLONING and RTF_PRCLONING flags are not being ignored.
256 * Any route returned has its reference count incremented.
259 _rtlookup(struct sockaddr
*dst
, boolean_t generate_report
, u_long ignore
)
261 struct radix_node_head
*rnh
= rt_tables
[mycpuid
][dst
->sa_family
];
268 * Look up route in the radix tree.
270 rt
= (struct rtentry
*) rnh
->rnh_matchaddr((char *)dst
, rnh
);
275 * Handle cloning routes.
277 if ((rt
->rt_flags
& ~ignore
& (RTF_CLONING
| RTF_PRCLONING
)) != 0) {
278 struct rtentry
*clonedroute
;
281 clonedroute
= rt
; /* copy in/copy out parameter */
282 error
= rtrequest(RTM_RESOLVE
, dst
, NULL
, NULL
, 0,
283 &clonedroute
); /* clone the route */
284 if (error
!= 0) { /* cloning failed */
286 rt_dstmsg(RTM_MISS
, dst
, error
);
288 return (rt
); /* return the uncloned route */
290 if (generate_report
) {
291 if (clonedroute
->rt_flags
& RTF_XRESOLVE
)
292 rt_dstmsg(RTM_RESOLVE
, dst
, 0);
294 rt_rtmsg(RTM_ADD
, clonedroute
,
295 clonedroute
->rt_ifp
, 0);
297 return (clonedroute
); /* return cloned route */
301 * Increment the reference count of the matched route and return.
307 rtstat
.rts_unreach
++;
309 rt_dstmsg(RTM_MISS
, dst
, 0);
314 rtfree(struct rtentry
*rt
)
316 KASSERT(rt
->rt_refcnt
> 0, ("rtfree: rt_refcnt %ld", rt
->rt_refcnt
));
319 if (rt
->rt_refcnt
== 0) {
320 struct radix_node_head
*rnh
=
321 rt_tables
[mycpuid
][rt_key(rt
)->sa_family
];
324 rnh
->rnh_close((struct radix_node
*)rt
, rnh
);
325 if (!(rt
->rt_flags
& RTF_UP
)) {
326 /* deallocate route */
327 if (rt
->rt_ifa
!= NULL
)
329 if (rt
->rt_parent
!= NULL
)
330 RTFREE(rt
->rt_parent
); /* recursive call! */
338 rtredirect_oncpu(struct sockaddr
*dst
, struct sockaddr
*gateway
,
339 struct sockaddr
*netmask
, int flags
, struct sockaddr
*src
)
341 struct rtentry
*rt
= NULL
;
342 struct rt_addrinfo rtinfo
;
347 /* verify the gateway is directly reachable */
348 if ((ifa
= ifa_ifwithnet(gateway
)) == NULL
) {
354 * If the redirect isn't from our current router for this destination,
355 * it's either old or wrong.
357 if (!(flags
& RTF_DONE
) && /* XXX JH */
358 (rt
= rtpurelookup(dst
)) != NULL
&&
359 (!sa_equal(src
, rt
->rt_gateway
) || rt
->rt_ifa
!= ifa
)) {
365 * If it redirects us to ourselves, we have a routing loop,
366 * perhaps as a result of an interface going down recently.
368 if (ifa_ifwithaddr(gateway
)) {
369 error
= EHOSTUNREACH
;
374 * Create a new entry if the lookup failed or if we got back
375 * a wildcard entry for the default route. This is necessary
376 * for hosts which use routing redirects generated by smart
377 * gateways to dynamically build the routing tables.
381 if ((rt_mask(rt
) != NULL
&& rt_mask(rt
)->sa_len
< 2)) {
386 /* Ignore redirects for directly connected hosts. */
387 if (!(rt
->rt_flags
& RTF_GATEWAY
)) {
388 error
= EHOSTUNREACH
;
392 if (!(rt
->rt_flags
& RTF_HOST
) && (flags
& RTF_HOST
)) {
394 * Changing from a network route to a host route.
395 * Create a new host route rather than smashing the
399 flags
|= RTF_GATEWAY
| RTF_DYNAMIC
;
400 bzero(&rtinfo
, sizeof(struct rt_addrinfo
));
401 rtinfo
.rti_info
[RTAX_DST
] = dst
;
402 rtinfo
.rti_info
[RTAX_GATEWAY
] = gateway
;
403 rtinfo
.rti_info
[RTAX_NETMASK
] = netmask
;
404 rtinfo
.rti_flags
= flags
;
405 rtinfo
.rti_ifa
= ifa
;
406 rt
= NULL
; /* copy-in/copy-out parameter */
407 error
= rtrequest1(RTM_ADD
, &rtinfo
, &rt
);
409 flags
= rt
->rt_flags
;
410 stat
= &rtstat
.rts_dynamic
;
413 * Smash the current notion of the gateway to this destination.
414 * Should check about netmask!!!
416 rt
->rt_flags
|= RTF_MODIFIED
;
417 flags
|= RTF_MODIFIED
;
418 rt_setgate(rt
, rt_key(rt
), gateway
);
420 stat
= &rtstat
.rts_newgateway
;
428 rtstat
.rts_badredirect
++;
429 else if (stat
!= NULL
)
437 struct netmsg_rtredirect
{
438 struct netmsg netmsg
;
439 struct sockaddr
*dst
;
440 struct sockaddr
*gateway
;
441 struct sockaddr
*netmask
;
443 struct sockaddr
*src
;
449 * Force a routing table entry to the specified
450 * destination to go through the given gateway.
451 * Normally called as a result of a routing redirect
452 * message from the network layer.
454 * N.B.: must be called at splnet
457 rtredirect(struct sockaddr
*dst
, struct sockaddr
*gateway
,
458 struct sockaddr
*netmask
, int flags
, struct sockaddr
*src
)
460 struct rt_addrinfo rtinfo
;
463 struct netmsg_rtredirect msg
;
465 netmsg_init(&msg
.netmsg
, &curthread
->td_msgport
, 0,
466 rtredirect_msghandler
);
468 msg
.gateway
= gateway
;
469 msg
.netmask
= netmask
;
472 error
= lwkt_domsg(rtable_portfn(0), &msg
.netmsg
.nm_lmsg
, 0);
474 error
= rtredirect_oncpu(dst
, gateway
, netmask
, flags
, src
);
476 bzero(&rtinfo
, sizeof(struct rt_addrinfo
));
477 rtinfo
.rti_info
[RTAX_DST
] = dst
;
478 rtinfo
.rti_info
[RTAX_GATEWAY
] = gateway
;
479 rtinfo
.rti_info
[RTAX_NETMASK
] = netmask
;
480 rtinfo
.rti_info
[RTAX_AUTHOR
] = src
;
481 rt_missmsg(RTM_REDIRECT
, &rtinfo
, flags
, error
);
487 rtredirect_msghandler(struct netmsg
*netmsg
)
489 struct netmsg_rtredirect
*msg
= (void *)netmsg
;
492 rtredirect_oncpu(msg
->dst
, msg
->gateway
, msg
->netmask
,
493 msg
->flags
, msg
->src
);
494 nextcpu
= mycpuid
+ 1;
496 lwkt_forwardmsg(rtable_portfn(nextcpu
), &netmsg
->nm_lmsg
);
498 lwkt_replymsg(&netmsg
->nm_lmsg
, 0);
504 * Routing table ioctl interface.
507 rtioctl(u_long req
, caddr_t data
, struct ucred
*cred
)
510 /* Multicast goop, grrr... */
511 return mrt_ioctl
? mrt_ioctl(req
, data
) : EOPNOTSUPP
;
518 ifa_ifwithroute(int flags
, struct sockaddr
*dst
, struct sockaddr
*gateway
)
522 if (!(flags
& RTF_GATEWAY
)) {
524 * If we are adding a route to an interface,
525 * and the interface is a point-to-point link,
526 * we should search for the destination
527 * as our clue to the interface. Otherwise
528 * we can use the local address.
531 if (flags
& RTF_HOST
) {
532 ifa
= ifa_ifwithdstaddr(dst
);
535 ifa
= ifa_ifwithaddr(gateway
);
538 * If we are adding a route to a remote net
539 * or host, the gateway may still be on the
540 * other end of a pt to pt link.
542 ifa
= ifa_ifwithdstaddr(gateway
);
545 ifa
= ifa_ifwithnet(gateway
);
549 rt
= rtpurelookup(gateway
);
553 if ((ifa
= rt
->rt_ifa
) == NULL
)
556 if (ifa
->ifa_addr
->sa_family
!= dst
->sa_family
) {
557 struct ifaddr
*oldifa
= ifa
;
559 ifa
= ifaof_ifpforaddr(dst
, ifa
->ifa_ifp
);
566 static int rt_fixdelete (struct radix_node
*, void *);
567 static int rt_fixchange (struct radix_node
*, void *);
571 struct radix_node_head
*rnh
;
575 * Set rtinfo->rti_ifa and rtinfo->rti_ifp.
578 rt_getifa(struct rt_addrinfo
*rtinfo
)
580 struct sockaddr
*gateway
= rtinfo
->rti_info
[RTAX_GATEWAY
];
581 struct sockaddr
*dst
= rtinfo
->rti_info
[RTAX_DST
];
582 struct sockaddr
*ifaaddr
= rtinfo
->rti_info
[RTAX_IFA
];
583 int flags
= rtinfo
->rti_flags
;
586 * ifp may be specified by sockaddr_dl
587 * when protocol address is ambiguous.
589 if (rtinfo
->rti_ifp
== NULL
) {
590 struct sockaddr
*ifpaddr
;
592 ifpaddr
= rtinfo
->rti_info
[RTAX_IFP
];
593 if (ifpaddr
!= NULL
&& ifpaddr
->sa_family
== AF_LINK
) {
596 ifa
= ifa_ifwithnet(ifpaddr
);
598 rtinfo
->rti_ifp
= ifa
->ifa_ifp
;
602 if (rtinfo
->rti_ifa
== NULL
&& ifaaddr
!= NULL
)
603 rtinfo
->rti_ifa
= ifa_ifwithaddr(ifaaddr
);
604 if (rtinfo
->rti_ifa
== NULL
) {
607 sa
= ifaaddr
!= NULL
? ifaaddr
:
608 (gateway
!= NULL
? gateway
: dst
);
609 if (sa
!= NULL
&& rtinfo
->rti_ifp
!= NULL
)
610 rtinfo
->rti_ifa
= ifaof_ifpforaddr(sa
, rtinfo
->rti_ifp
);
611 else if (dst
!= NULL
&& gateway
!= NULL
)
612 rtinfo
->rti_ifa
= ifa_ifwithroute(flags
, dst
, gateway
);
614 rtinfo
->rti_ifa
= ifa_ifwithroute(flags
, sa
, sa
);
616 if (rtinfo
->rti_ifa
== NULL
)
617 return (ENETUNREACH
);
619 if (rtinfo
->rti_ifp
== NULL
)
620 rtinfo
->rti_ifp
= rtinfo
->rti_ifa
->ifa_ifp
;
625 * Do appropriate manipulations of a routing tree given
626 * all the bits of info needed
631 struct sockaddr
*dst
,
632 struct sockaddr
*gateway
,
633 struct sockaddr
*netmask
,
635 struct rtentry
**ret_nrt
)
637 struct rt_addrinfo rtinfo
;
639 bzero(&rtinfo
, sizeof(struct rt_addrinfo
));
640 rtinfo
.rti_info
[RTAX_DST
] = dst
;
641 rtinfo
.rti_info
[RTAX_GATEWAY
] = gateway
;
642 rtinfo
.rti_info
[RTAX_NETMASK
] = netmask
;
643 rtinfo
.rti_flags
= flags
;
644 return rtrequest1(req
, &rtinfo
, ret_nrt
);
650 struct sockaddr
*dst
,
651 struct sockaddr
*gateway
,
652 struct sockaddr
*netmask
,
655 struct rt_addrinfo rtinfo
;
657 bzero(&rtinfo
, sizeof(struct rt_addrinfo
));
658 rtinfo
.rti_info
[RTAX_DST
] = dst
;
659 rtinfo
.rti_info
[RTAX_GATEWAY
] = gateway
;
660 rtinfo
.rti_info
[RTAX_NETMASK
] = netmask
;
661 rtinfo
.rti_flags
= flags
;
662 return rtrequest1_global(req
, &rtinfo
, NULL
, NULL
);
668 struct netmsg netmsg
;
670 struct rt_addrinfo
*rtinfo
;
671 rtrequest1_callback_func_t callback
;
678 rtrequest1_global(int req
, struct rt_addrinfo
*rtinfo
,
679 rtrequest1_callback_func_t callback
, void *arg
)
683 struct netmsg_rtq msg
;
685 netmsg_init(&msg
.netmsg
, &curthread
->td_msgport
, 0,
686 rtrequest1_msghandler
);
687 msg
.netmsg
.nm_lmsg
.ms_error
= -1;
690 msg
.callback
= callback
;
692 error
= lwkt_domsg(rtable_portfn(0), &msg
.netmsg
.nm_lmsg
, 0);
694 struct rtentry
*rt
= NULL
;
696 error
= rtrequest1(req
, rtinfo
, &rt
);
700 callback(req
, error
, rtinfo
, rt
, arg
);
706 * Handle a route table request on the current cpu. Since the route table's
707 * are supposed to be identical on each cpu, an error occuring later in the
708 * message chain is considered system-fatal.
713 rtrequest1_msghandler(struct netmsg
*netmsg
)
715 struct netmsg_rtq
*msg
= (void *)netmsg
;
716 struct rtentry
*rt
= NULL
;
720 error
= rtrequest1(msg
->req
, msg
->rtinfo
, &rt
);
724 msg
->callback(msg
->req
, error
, msg
->rtinfo
, rt
, msg
->arg
);
727 * RTM_DELETE's are propogated even if an error occurs, since a
728 * cloned route might be undergoing deletion and cloned routes
729 * are not necessarily replicated. An overall error is returned
730 * only if no cpus have the route in question.
732 if (msg
->netmsg
.nm_lmsg
.ms_error
< 0 || error
== 0)
733 msg
->netmsg
.nm_lmsg
.ms_error
= error
;
735 nextcpu
= mycpuid
+ 1;
736 if (error
&& msg
->req
!= RTM_DELETE
) {
738 panic("rtrequest1_msghandler: rtrequest table "
739 "error was not on cpu #0: %p", msg
->rtinfo
);
741 lwkt_replymsg(&msg
->netmsg
.nm_lmsg
, error
);
742 } else if (nextcpu
< ncpus
) {
743 lwkt_forwardmsg(rtable_portfn(nextcpu
), &msg
->netmsg
.nm_lmsg
);
745 lwkt_replymsg(&msg
->netmsg
.nm_lmsg
,
746 msg
->netmsg
.nm_lmsg
.ms_error
);
753 rtrequest1(int req
, struct rt_addrinfo
*rtinfo
, struct rtentry
**ret_nrt
)
755 struct sockaddr
*dst
= rtinfo
->rti_info
[RTAX_DST
];
757 struct radix_node
*rn
;
758 struct radix_node_head
*rnh
;
760 struct sockaddr
*ndst
;
763 #define gotoerr(x) { error = x ; goto bad; }
767 rt_addrinfo_print(req
, rtinfo
);
772 * Find the correct routing tree to use for this Address Family
774 if ((rnh
= rt_tables
[mycpuid
][dst
->sa_family
]) == NULL
)
775 gotoerr(EAFNOSUPPORT
);
778 * If we are adding a host route then we don't want to put
779 * a netmask in the tree, nor do we want to clone it.
781 if (rtinfo
->rti_flags
& RTF_HOST
) {
782 rtinfo
->rti_info
[RTAX_NETMASK
] = NULL
;
783 rtinfo
->rti_flags
&= ~(RTF_CLONING
| RTF_PRCLONING
);
788 /* Remove the item from the tree. */
789 rn
= rnh
->rnh_deladdr((char *)rtinfo
->rti_info
[RTAX_DST
],
790 (char *)rtinfo
->rti_info
[RTAX_NETMASK
],
794 KASSERT(!(rn
->rn_flags
& (RNF_ACTIVE
| RNF_ROOT
)),
795 ("rnh_deladdr returned flags 0x%x", rn
->rn_flags
));
796 rt
= (struct rtentry
*)rn
;
798 /* ref to prevent a deletion race */
801 /* Free any routes cloned from this one. */
802 if ((rt
->rt_flags
& (RTF_CLONING
| RTF_PRCLONING
)) &&
803 rt_mask(rt
) != NULL
) {
804 rnh
->rnh_walktree_from(rnh
, (char *)rt_key(rt
),
809 if (rt
->rt_gwroute
!= NULL
) {
810 RTFREE(rt
->rt_gwroute
);
811 rt
->rt_gwroute
= NULL
;
815 * NB: RTF_UP must be set during the search above,
816 * because we might delete the last ref, causing
817 * rt to get freed prematurely.
819 rt
->rt_flags
&= ~RTF_UP
;
823 rt_print(rtinfo
, rt
);
826 /* Give the protocol a chance to keep things in sync. */
827 if ((ifa
= rt
->rt_ifa
) && ifa
->ifa_rtrequest
)
828 ifa
->ifa_rtrequest(RTM_DELETE
, rt
, rtinfo
);
831 * If the caller wants it, then it can have it,
832 * but it's up to it to free the rtentry as we won't be
835 KASSERT(rt
->rt_refcnt
>= 0,
836 ("rtrequest1(DELETE): refcnt %ld", rt
->rt_refcnt
));
837 if (ret_nrt
!= NULL
) {
838 /* leave ref intact for return */
841 /* deref / attempt to destroy */
847 if (ret_nrt
== NULL
|| (rt
= *ret_nrt
) == NULL
)
851 rt
->rt_flags
& ~(RTF_CLONING
| RTF_PRCLONING
| RTF_STATIC
);
852 rtinfo
->rti_flags
|= RTF_WASCLONED
;
853 rtinfo
->rti_info
[RTAX_GATEWAY
] = rt
->rt_gateway
;
854 if ((rtinfo
->rti_info
[RTAX_NETMASK
] = rt
->rt_genmask
) == NULL
)
855 rtinfo
->rti_flags
|= RTF_HOST
;
856 rtinfo
->rti_info
[RTAX_MPLS1
] = rt
->rt_shim
[0];
857 rtinfo
->rti_info
[RTAX_MPLS2
] = rt
->rt_shim
[1];
858 rtinfo
->rti_info
[RTAX_MPLS3
] = rt
->rt_shim
[2];
862 KASSERT(!(rtinfo
->rti_flags
& RTF_GATEWAY
) ||
863 rtinfo
->rti_info
[RTAX_GATEWAY
] != NULL
,
864 ("rtrequest: GATEWAY but no gateway"));
866 if (rtinfo
->rti_ifa
== NULL
&& (error
= rt_getifa(rtinfo
)))
868 ifa
= rtinfo
->rti_ifa
;
870 R_Malloc(rt
, struct rtentry
*, sizeof(struct rtentry
));
873 bzero(rt
, sizeof(struct rtentry
));
874 rt
->rt_flags
= RTF_UP
| rtinfo
->rti_flags
;
875 rt
->rt_cpuid
= mycpuid
;
876 error
= rt_setgate(rt
, dst
, rtinfo
->rti_info
[RTAX_GATEWAY
]);
883 if (rtinfo
->rti_info
[RTAX_NETMASK
] != NULL
)
884 rt_maskedcopy(dst
, ndst
,
885 rtinfo
->rti_info
[RTAX_NETMASK
]);
887 bcopy(dst
, ndst
, dst
->sa_len
);
889 if (rtinfo
->rti_info
[RTAX_MPLS1
] != NULL
)
890 rt_setshims(rt
, rtinfo
->rti_info
);
893 * Note that we now have a reference to the ifa.
894 * This moved from below so that rnh->rnh_addaddr() can
895 * examine the ifa and ifa->ifa_ifp if it so desires.
899 rt
->rt_ifp
= ifa
->ifa_ifp
;
900 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
902 rn
= rnh
->rnh_addaddr((char *)ndst
,
903 (char *)rtinfo
->rti_info
[RTAX_NETMASK
],
906 struct rtentry
*oldrt
;
909 * We already have one of these in the tree.
910 * We do a special hack: if the old route was
911 * cloned, then we blow it away and try
912 * re-inserting the new one.
914 oldrt
= rtpurelookup(ndst
);
917 if (oldrt
->rt_flags
& RTF_WASCLONED
) {
918 rtrequest(RTM_DELETE
, rt_key(oldrt
),
921 oldrt
->rt_flags
, NULL
);
922 rn
= rnh
->rnh_addaddr((char *)ndst
,
924 rtinfo
->rti_info
[RTAX_NETMASK
],
931 * If it still failed to go into the tree,
932 * then un-make it (this should be a function).
935 if (rt
->rt_gwroute
!= NULL
)
936 rtfree(rt
->rt_gwroute
);
944 * If we got here from RESOLVE, then we are cloning
945 * so clone the rest, and note that we
946 * are a clone (and increment the parent's references)
948 if (req
== RTM_RESOLVE
) {
949 rt
->rt_rmx
= (*ret_nrt
)->rt_rmx
; /* copy metrics */
950 rt
->rt_rmx
.rmx_pksent
= 0; /* reset packet counter */
951 if ((*ret_nrt
)->rt_flags
&
952 (RTF_CLONING
| RTF_PRCLONING
)) {
953 rt
->rt_parent
= *ret_nrt
;
954 (*ret_nrt
)->rt_refcnt
++;
959 * if this protocol has something to add to this then
960 * allow it to do that as well.
962 if (ifa
->ifa_rtrequest
!= NULL
)
963 ifa
->ifa_rtrequest(req
, rt
, rtinfo
);
966 * We repeat the same procedure from rt_setgate() here because
967 * it doesn't fire when we call it there because the node
968 * hasn't been added to the tree yet.
970 if (req
== RTM_ADD
&& !(rt
->rt_flags
& RTF_HOST
) &&
971 rt_mask(rt
) != NULL
) {
972 struct rtfc_arg arg
= { rt
, rnh
};
974 rnh
->rnh_walktree_from(rnh
, (char *)rt_key(rt
),
981 rt_print(rtinfo
, rt
);
984 * Return the resulting rtentry,
985 * increasing the number of references by one.
987 if (ret_nrt
!= NULL
) {
999 kprintf("rti %p failed error %d\n", rtinfo
, error
);
1001 kprintf("rti %p succeeded\n", rtinfo
);
1009 * Called from rtrequest(RTM_DELETE, ...) to fix up the route's ``family''
1010 * (i.e., the routes related to it by the operation of cloning). This
1011 * routine is iterated over all potential former-child-routes by way of
1012 * rnh->rnh_walktree_from() above, and those that actually are children of
1013 * the late parent (passed in as VP here) are themselves deleted.
1016 rt_fixdelete(struct radix_node
*rn
, void *vp
)
1018 struct rtentry
*rt
= (struct rtentry
*)rn
;
1019 struct rtentry
*rt0
= vp
;
1021 if (rt
->rt_parent
== rt0
&&
1022 !(rt
->rt_flags
& (RTF_PINNED
| RTF_CLONING
| RTF_PRCLONING
))) {
1023 return rtrequest(RTM_DELETE
, rt_key(rt
), NULL
, rt_mask(rt
),
1024 rt
->rt_flags
, NULL
);
1030 * This routine is called from rt_setgate() to do the analogous thing for
1031 * adds and changes. There is the added complication in this case of a
1032 * middle insert; i.e., insertion of a new network route between an older
1033 * network route and (cloned) host routes. For this reason, a simple check
1034 * of rt->rt_parent is insufficient; each candidate route must be tested
1035 * against the (mask, value) of the new route (passed as before in vp)
1036 * to see if the new route matches it.
1038 * XXX - it may be possible to do fixdelete() for changes and reserve this
1039 * routine just for adds. I'm not sure why I thought it was necessary to do
1043 static int rtfcdebug
= 0;
1047 rt_fixchange(struct radix_node
*rn
, void *vp
)
1049 struct rtentry
*rt
= (struct rtentry
*)rn
;
1050 struct rtfc_arg
*ap
= vp
;
1051 struct rtentry
*rt0
= ap
->rt0
;
1052 struct radix_node_head
*rnh
= ap
->rnh
;
1053 u_char
*xk1
, *xm1
, *xk2
, *xmp
;
1058 kprintf("rt_fixchange: rt %p, rt0 %p\n", rt
, rt0
);
1061 if (rt
->rt_parent
== NULL
||
1062 (rt
->rt_flags
& (RTF_PINNED
| RTF_CLONING
| RTF_PRCLONING
))) {
1064 if (rtfcdebug
) kprintf("no parent, pinned or cloning\n");
1069 if (rt
->rt_parent
== rt0
) {
1071 if (rtfcdebug
) kprintf("parent match\n");
1073 return rtrequest(RTM_DELETE
, rt_key(rt
), NULL
, rt_mask(rt
),
1074 rt
->rt_flags
, NULL
);
1078 * There probably is a function somewhere which does this...
1079 * if not, there should be.
1081 len
= imin(rt_key(rt0
)->sa_len
, rt_key(rt
)->sa_len
);
1083 xk1
= (u_char
*)rt_key(rt0
);
1084 xm1
= (u_char
*)rt_mask(rt0
);
1085 xk2
= (u_char
*)rt_key(rt
);
1087 /* avoid applying a less specific route */
1088 xmp
= (u_char
*)rt_mask(rt
->rt_parent
);
1089 mlen
= rt_key(rt
->rt_parent
)->sa_len
;
1090 if (mlen
> rt_key(rt0
)->sa_len
) {
1093 kprintf("rt_fixchange: inserting a less "
1094 "specific route\n");
1098 for (i
= rnh
->rnh_treetop
->rn_offset
; i
< mlen
; i
++) {
1099 if ((xmp
[i
] & ~(xmp
[i
] ^ xm1
[i
])) != xmp
[i
]) {
1102 kprintf("rt_fixchange: inserting a less "
1103 "specific route\n");
1109 for (i
= rnh
->rnh_treetop
->rn_offset
; i
< len
; i
++) {
1110 if ((xk2
[i
] & xm1
[i
]) != xk1
[i
]) {
1112 if (rtfcdebug
) kprintf("no match\n");
1119 * OK, this node is a clone, and matches the node currently being
1120 * changed/added under the node's mask. So, get rid of it.
1123 if (rtfcdebug
) kprintf("deleting\n");
1125 return rtrequest(RTM_DELETE
, rt_key(rt
), NULL
, rt_mask(rt
),
1126 rt
->rt_flags
, NULL
);
1129 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
1132 rt_setgate(struct rtentry
*rt0
, struct sockaddr
*dst
, struct sockaddr
*gate
)
1134 char *space
, *oldspace
;
1135 int dlen
= ROUNDUP(dst
->sa_len
), glen
= ROUNDUP(gate
->sa_len
);
1136 struct rtentry
*rt
= rt0
;
1137 struct radix_node_head
*rnh
= rt_tables
[mycpuid
][dst
->sa_family
];
1140 * A host route with the destination equal to the gateway
1141 * will interfere with keeping LLINFO in the routing
1142 * table, so disallow it.
1144 if (((rt0
->rt_flags
& (RTF_HOST
| RTF_GATEWAY
| RTF_LLINFO
)) ==
1145 (RTF_HOST
| RTF_GATEWAY
)) &&
1146 dst
->sa_len
== gate
->sa_len
&&
1147 sa_equal(dst
, gate
)) {
1149 * The route might already exist if this is an RTM_CHANGE
1150 * or a routing redirect, so try to delete it.
1152 if (rt_key(rt0
) != NULL
)
1153 rtrequest(RTM_DELETE
, rt_key(rt0
), rt0
->rt_gateway
,
1154 rt_mask(rt0
), rt0
->rt_flags
, NULL
);
1155 return EADDRNOTAVAIL
;
1159 * Both dst and gateway are stored in the same malloc'ed chunk
1160 * (If I ever get my hands on....)
1161 * if we need to malloc a new chunk, then keep the old one around
1162 * till we don't need it any more.
1164 if (rt
->rt_gateway
== NULL
|| glen
> ROUNDUP(rt
->rt_gateway
->sa_len
)) {
1165 oldspace
= (char *)rt_key(rt
);
1166 R_Malloc(space
, char *, dlen
+ glen
);
1169 rt
->rt_nodes
->rn_key
= space
;
1171 space
= (char *)rt_key(rt
); /* Just use the old space. */
1175 /* Set the gateway value. */
1176 rt
->rt_gateway
= (struct sockaddr
*)(space
+ dlen
);
1177 bcopy(gate
, rt
->rt_gateway
, glen
);
1179 if (oldspace
!= NULL
) {
1181 * If we allocated a new chunk, preserve the original dst.
1182 * This way, rt_setgate() really just sets the gate
1183 * and leaves the dst field alone.
1185 bcopy(dst
, space
, dlen
);
1190 * If there is already a gwroute, it's now almost definitely wrong
1193 if (rt
->rt_gwroute
!= NULL
) {
1194 RTFREE(rt
->rt_gwroute
);
1195 rt
->rt_gwroute
= NULL
;
1197 if (rt
->rt_flags
& RTF_GATEWAY
) {
1199 * Cloning loop avoidance: In the presence of
1200 * protocol-cloning and bad configuration, it is
1201 * possible to get stuck in bottomless mutual recursion
1202 * (rtrequest rt_setgate rtlookup). We avoid this
1203 * by not allowing protocol-cloning to operate for
1204 * gateways (which is probably the correct choice
1205 * anyway), and avoid the resulting reference loops
1206 * by disallowing any route to run through itself as
1207 * a gateway. This is obviously mandatory when we
1208 * get rt->rt_output().
1210 * This breaks TTCP for hosts outside the gateway! XXX JH
1212 rt
->rt_gwroute
= _rtlookup(gate
, RTL_REPORTMSG
, RTF_PRCLONING
);
1213 if (rt
->rt_gwroute
== rt
) {
1214 rt
->rt_gwroute
= NULL
;
1216 return EDQUOT
; /* failure */
1221 * This isn't going to do anything useful for host routes, so
1222 * don't bother. Also make sure we have a reasonable mask
1223 * (we don't yet have one during adds).
1225 if (!(rt
->rt_flags
& RTF_HOST
) && rt_mask(rt
) != NULL
) {
1226 struct rtfc_arg arg
= { rt
, rnh
};
1228 rnh
->rnh_walktree_from(rnh
, (char *)rt_key(rt
),
1229 (char *)rt_mask(rt
),
1230 rt_fixchange
, &arg
);
1238 struct sockaddr
*src
,
1239 struct sockaddr
*dst
,
1240 struct sockaddr
*netmask
)
1242 u_char
*cp1
= (u_char
*)src
;
1243 u_char
*cp2
= (u_char
*)dst
;
1244 u_char
*cp3
= (u_char
*)netmask
;
1245 u_char
*cplim
= cp2
+ *cp3
;
1246 u_char
*cplim2
= cp2
+ *cp1
;
1248 *cp2
++ = *cp1
++; *cp2
++ = *cp1
++; /* copies sa_len & sa_family */
1253 *cp2
++ = *cp1
++ & *cp3
++;
1255 bzero(cp2
, cplim2
- cp2
);
1259 rt_llroute(struct sockaddr
*dst
, struct rtentry
*rt0
, struct rtentry
**drt
)
1261 struct rtentry
*up_rt
, *rt
;
1263 if (!(rt0
->rt_flags
& RTF_UP
)) {
1264 up_rt
= rtlookup(dst
);
1266 return (EHOSTUNREACH
);
1270 if (up_rt
->rt_flags
& RTF_GATEWAY
) {
1271 if (up_rt
->rt_gwroute
== NULL
) {
1272 up_rt
->rt_gwroute
= rtlookup(up_rt
->rt_gateway
);
1273 if (up_rt
->rt_gwroute
== NULL
)
1274 return (EHOSTUNREACH
);
1275 } else if (!(up_rt
->rt_gwroute
->rt_flags
& RTF_UP
)) {
1276 rtfree(up_rt
->rt_gwroute
);
1277 up_rt
->rt_gwroute
= rtlookup(up_rt
->rt_gateway
);
1278 if (up_rt
->rt_gwroute
== NULL
)
1279 return (EHOSTUNREACH
);
1281 rt
= up_rt
->rt_gwroute
;
1284 if (rt
->rt_flags
& RTF_REJECT
&&
1285 (rt
->rt_rmx
.rmx_expire
== 0 || /* rt doesn't expire */
1286 time_second
< rt
->rt_rmx
.rmx_expire
)) /* rt not expired */
1287 return (rt
->rt_flags
& RTF_HOST
? EHOSTDOWN
: EHOSTUNREACH
);
1293 rt_setshims(struct rtentry
*rt
, struct sockaddr
**rt_shim
){
1296 for (i
=0; i
<3; i
++) {
1297 struct sockaddr
*shim
= rt_shim
[RTAX_MPLS1
+ i
];
1303 shimlen
= ROUNDUP(shim
->sa_len
);
1304 R_Malloc(rt
->rt_shim
[i
], struct sockaddr_mpls
*, shimlen
);
1305 bcopy(shim
, rt
->rt_shim
[i
], shimlen
);
1314 * Print out a route table entry
1317 rt_print(struct rt_addrinfo
*rtinfo
, struct rtentry
*rn
)
1319 kprintf("rti %p cpu %d route %p flags %08lx: ",
1320 rtinfo
, mycpuid
, rn
, rn
->rt_flags
);
1321 sockaddr_print(rt_key(rn
));
1323 sockaddr_print(rt_mask(rn
));
1325 sockaddr_print(rn
->rt_gateway
);
1326 kprintf(" ifc \"%s\"", rn
->rt_ifp
? rn
->rt_ifp
->if_dname
: "?");
1327 kprintf(" ifa %p\n", rn
->rt_ifa
);
1331 rt_addrinfo_print(int cmd
, struct rt_addrinfo
*rti
)
1337 if (cmd
== RTM_DELETE
&& route_debug
> 1)
1338 db_print_backtrace();
1352 kprintf("C%02d ", cmd
);
1355 kprintf("rti %p cpu %d ", rti
, mycpuid
);
1356 for (i
= 0; i
< rti
->rti_addrs
; ++i
) {
1357 if (rti
->rti_info
[i
] == NULL
)
1387 kprintf("(?%02d ", i
);
1390 sockaddr_print(rti
->rti_info
[i
]);
1398 sockaddr_print(struct sockaddr
*sa
)
1400 struct sockaddr_in
*sa4
;
1401 struct sockaddr_in6
*sa6
;
1410 len
= sa
->sa_len
- offsetof(struct sockaddr
, sa_data
[0]);
1412 switch(sa
->sa_family
) {
1416 switch(sa
->sa_family
) {
1418 sa4
= (struct sockaddr_in
*)sa
;
1419 kprintf("INET %d %d.%d.%d.%d",
1420 ntohs(sa4
->sin_port
),
1421 (ntohl(sa4
->sin_addr
.s_addr
) >> 24) & 255,
1422 (ntohl(sa4
->sin_addr
.s_addr
) >> 16) & 255,
1423 (ntohl(sa4
->sin_addr
.s_addr
) >> 8) & 255,
1424 (ntohl(sa4
->sin_addr
.s_addr
) >> 0) & 255
1428 sa6
= (struct sockaddr_in6
*)sa
;
1429 kprintf("INET6 %d %04x:%04x%04x:%04x:%04x:%04x:%04x:%04x",
1430 ntohs(sa6
->sin6_port
),
1431 sa6
->sin6_addr
.s6_addr16
[0],
1432 sa6
->sin6_addr
.s6_addr16
[1],
1433 sa6
->sin6_addr
.s6_addr16
[2],
1434 sa6
->sin6_addr
.s6_addr16
[3],
1435 sa6
->sin6_addr
.s6_addr16
[4],
1436 sa6
->sin6_addr
.s6_addr16
[5],
1437 sa6
->sin6_addr
.s6_addr16
[6],
1438 sa6
->sin6_addr
.s6_addr16
[7]
1442 kprintf("AF%d ", sa
->sa_family
);
1443 while (len
> 0 && sa
->sa_data
[len
-1] == 0)
1446 for (i
= 0; i
< len
; ++i
) {
1449 kprintf("%d", (unsigned char)sa
->sa_data
[i
]);
1459 * Set up a routing table entry, normally for an interface.
1462 rtinit(struct ifaddr
*ifa
, int cmd
, int flags
)
1464 struct sockaddr
*dst
, *deldst
, *netmask
;
1465 struct mbuf
*m
= NULL
;
1466 struct radix_node_head
*rnh
;
1467 struct radix_node
*rn
;
1468 struct rt_addrinfo rtinfo
;
1471 if (flags
& RTF_HOST
) {
1472 dst
= ifa
->ifa_dstaddr
;
1475 dst
= ifa
->ifa_addr
;
1476 netmask
= ifa
->ifa_netmask
;
1479 * If it's a delete, check that if it exists, it's on the correct
1480 * interface or we might scrub a route to another ifa which would
1481 * be confusing at best and possibly worse.
1483 if (cmd
== RTM_DELETE
) {
1485 * It's a delete, so it should already exist..
1486 * If it's a net, mask off the host bits
1487 * (Assuming we have a mask)
1489 if (netmask
!= NULL
) {
1490 m
= m_get(MB_DONTWAIT
, MT_SONAME
);
1494 deldst
= mtod(m
, struct sockaddr
*);
1495 rt_maskedcopy(dst
, deldst
, netmask
);
1499 * Look up an rtentry that is in the routing tree and
1500 * contains the correct info.
1502 if ((rnh
= rt_tables
[mycpuid
][dst
->sa_family
]) == NULL
||
1503 (rn
= rnh
->rnh_lookup((char *)dst
,
1504 (char *)netmask
, rnh
)) == NULL
||
1505 ((struct rtentry
*)rn
)->rt_ifa
!= ifa
||
1506 !sa_equal((struct sockaddr
*)rn
->rn_key
, dst
)) {
1509 return (flags
& RTF_HOST
? EHOSTUNREACH
: ENETUNREACH
);
1515 * One would think that as we are deleting, and we know
1516 * it doesn't exist, we could just return at this point
1517 * with an "ELSE" clause, but apparently not..
1519 return (flags
& RTF_HOST
? EHOSTUNREACH
: ENETUNREACH
);
1524 * Do the actual request
1526 bzero(&rtinfo
, sizeof(struct rt_addrinfo
));
1527 rtinfo
.rti_info
[RTAX_DST
] = dst
;
1528 rtinfo
.rti_info
[RTAX_GATEWAY
] = ifa
->ifa_addr
;
1529 rtinfo
.rti_info
[RTAX_NETMASK
] = netmask
;
1530 rtinfo
.rti_flags
= flags
| ifa
->ifa_flags
;
1531 rtinfo
.rti_ifa
= ifa
;
1532 error
= rtrequest1_global(cmd
, &rtinfo
, rtinit_rtrequest_callback
, ifa
);
1539 rtinit_rtrequest_callback(int cmd
, int error
,
1540 struct rt_addrinfo
*rtinfo
, struct rtentry
*rt
,
1543 struct ifaddr
*ifa
= arg
;
1545 if (error
== 0 && rt
) {
1548 rt_newaddrmsg(cmd
, ifa
, error
, rt
);
1551 if (cmd
== RTM_DELETE
) {
1552 if (rt
->rt_refcnt
== 0) {
1560 /* This must be before ip6_init2(), which is now SI_ORDER_MIDDLE */
1561 SYSINIT(route
, SI_SUB_PROTO_DOMAIN
, SI_ORDER_THIRD
, route_init
, 0);