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 <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
);
120 static void rtsearch_msghandler(struct netmsg
*netmsg
);
122 static void rtmask_add_msghandler(struct netmsg
*netmsg
);
124 static int rt_setshims(struct rtentry
*, struct sockaddr
**);
126 SYSCTL_NODE(_net
, OID_AUTO
, route
, CTLFLAG_RW
, 0, "Routing");
129 static int route_debug
= 1;
130 SYSCTL_INT(_net_route
, OID_AUTO
, route_debug
, CTLFLAG_RW
,
131 &route_debug
, 0, "");
134 int route_assert_owner_access
= 0;
135 SYSCTL_INT(_net_route
, OID_AUTO
, assert_owner_access
, CTLFLAG_RW
,
136 &route_assert_owner_access
, 0, "");
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 */
152 rtable_init(); /* call dom_rtattach() on each cpu */
154 for (cpu
= 0; cpu
< ncpus
; cpu
++) {
155 lwkt_create(rtable_service_loop
, NULL
, &rtd
, NULL
,
156 0, cpu
, "rtable_cpu %d", cpu
);
157 rt_ports
[cpu
] = &rtd
->td_msgport
;
162 rtable_init_oncpu(struct netmsg
*nmsg
)
167 SLIST_FOREACH(dom
, &domains
, dom_next
) {
168 if (dom
->dom_rtattach
) {
170 (void **)&rt_tables
[cpu
][dom
->dom_family
],
174 ifnet_forwardmsg(&nmsg
->nm_lmsg
, cpu
+ 1);
182 netmsg_init(&nmsg
, &curthread
->td_msgport
, 0, rtable_init_oncpu
);
183 ifnet_domsg(&nmsg
.nm_lmsg
, 0);
187 * Our per-cpu table management protocol thread. All route table operations
188 * are sequentially chained through all cpus starting at cpu #0 in order to
189 * maintain duplicate route tables on each cpu. Having a spearate route
190 * table management thread allows the protocol and interrupt threads to
191 * issue route table changes.
194 rtable_service_loop(void *dummy __unused
)
196 struct netmsg
*netmsg
;
197 thread_t td
= curthread
;
199 while ((netmsg
= lwkt_waitport(&td
->td_msgport
, 0)) != NULL
) {
200 netmsg
->nm_dispatch(netmsg
);
205 * Routing statistics.
209 sysctl_rtstatistics(SYSCTL_HANDLER_ARGS
)
213 for (cpu
= 0; cpu
< ncpus
; ++cpu
) {
214 if ((error
= SYSCTL_OUT(req
, &rtstatistics_percpu
[cpu
],
215 sizeof(struct rtstatistics
))))
217 if ((error
= SYSCTL_IN(req
, &rtstatistics_percpu
[cpu
],
218 sizeof(struct rtstatistics
))))
224 SYSCTL_PROC(_net_route
, OID_AUTO
, stats
, (CTLTYPE_OPAQUE
|CTLFLAG_RW
),
225 0, 0, sysctl_rtstatistics
, "S,rtstatistics", "Routing statistics");
227 SYSCTL_STRUCT(_net_route
, OID_AUTO
, stats
, CTLFLAG_RW
, &rtstat
, rtstatistics
,
228 "Routing statistics");
232 * Packet routing routines.
236 * Look up and fill in the "ro_rt" rtentry field in a route structure given
237 * an address in the "ro_dst" field. Always send a report on a miss and
238 * always clone routes.
241 rtalloc(struct route
*ro
)
243 rtalloc_ign(ro
, 0UL);
247 * Look up and fill in the "ro_rt" rtentry field in a route structure given
248 * an address in the "ro_dst" field. Always send a report on a miss and
249 * optionally clone routes when RTF_CLONING or RTF_PRCLONING are not being
253 rtalloc_ign(struct route
*ro
, u_long ignoreflags
)
255 if (ro
->ro_rt
!= NULL
) {
256 if (ro
->ro_rt
->rt_ifp
!= NULL
&& ro
->ro_rt
->rt_flags
& RTF_UP
)
261 ro
->ro_rt
= _rtlookup(&ro
->ro_dst
, RTL_REPORTMSG
, ignoreflags
);
265 * Look up the route that matches the given "dst" address.
267 * Route lookup can have the side-effect of creating and returning
268 * a cloned route instead when "dst" matches a cloning route and the
269 * RTF_CLONING and RTF_PRCLONING flags are not being ignored.
271 * Any route returned has its reference count incremented.
274 _rtlookup(struct sockaddr
*dst
, boolean_t generate_report
, u_long ignore
)
276 struct radix_node_head
*rnh
= rt_tables
[mycpuid
][dst
->sa_family
];
283 * Look up route in the radix tree.
285 rt
= (struct rtentry
*) rnh
->rnh_matchaddr((char *)dst
, rnh
);
290 * Handle cloning routes.
292 if ((rt
->rt_flags
& ~ignore
& (RTF_CLONING
| RTF_PRCLONING
)) != 0) {
293 struct rtentry
*clonedroute
;
296 clonedroute
= rt
; /* copy in/copy out parameter */
297 error
= rtrequest(RTM_RESOLVE
, dst
, NULL
, NULL
, 0,
298 &clonedroute
); /* clone the route */
299 if (error
!= 0) { /* cloning failed */
301 rt_dstmsg(RTM_MISS
, dst
, error
);
303 return (rt
); /* return the uncloned route */
305 if (generate_report
) {
306 if (clonedroute
->rt_flags
& RTF_XRESOLVE
)
307 rt_dstmsg(RTM_RESOLVE
, dst
, 0);
309 rt_rtmsg(RTM_ADD
, clonedroute
,
310 clonedroute
->rt_ifp
, 0);
312 return (clonedroute
); /* return cloned route */
316 * Increment the reference count of the matched route and return.
322 rtstat
.rts_unreach
++;
324 rt_dstmsg(RTM_MISS
, dst
, 0);
329 rtfree(struct rtentry
*rt
)
331 if (rt
->rt_cpuid
== mycpuid
)
334 rtfree_remote(rt
, 1);
338 rtfree_oncpu(struct rtentry
*rt
)
340 KKASSERT(rt
->rt_cpuid
== mycpuid
);
341 KASSERT(rt
->rt_refcnt
> 0, ("rtfree: rt_refcnt %ld", rt
->rt_refcnt
));
344 if (rt
->rt_refcnt
== 0) {
345 struct radix_node_head
*rnh
=
346 rt_tables
[mycpuid
][rt_key(rt
)->sa_family
];
349 rnh
->rnh_close((struct radix_node
*)rt
, rnh
);
350 if (!(rt
->rt_flags
& RTF_UP
)) {
351 /* deallocate route */
352 if (rt
->rt_ifa
!= NULL
)
354 if (rt
->rt_parent
!= NULL
)
355 RTFREE(rt
->rt_parent
); /* recursive call! */
363 rtfree_remote_dispatch(struct netmsg
*nmsg
)
365 struct lwkt_msg
*lmsg
= &nmsg
->nm_lmsg
;
366 struct rtentry
*rt
= lmsg
->u
.ms_resultp
;
369 lwkt_replymsg(lmsg
, 0);
373 rtfree_remote(struct rtentry
*rt
, int allow_panic
)
376 struct lwkt_msg
*lmsg
;
378 KKASSERT(rt
->rt_cpuid
!= mycpuid
);
380 if (route_assert_owner_access
&& allow_panic
) {
381 panic("rt remote free rt_cpuid %d, mycpuid %d\n",
382 rt
->rt_cpuid
, mycpuid
);
384 kprintf("rt remote free rt_cpuid %d, mycpuid %d\n",
385 rt
->rt_cpuid
, mycpuid
);
389 netmsg_init(&nmsg
, &curthread
->td_msgport
, 0, rtfree_remote_dispatch
);
390 lmsg
= &nmsg
.nm_lmsg
;
391 lmsg
->u
.ms_resultp
= rt
;
393 lwkt_domsg(rtable_portfn(rt
->rt_cpuid
), lmsg
, 0);
397 rtredirect_oncpu(struct sockaddr
*dst
, struct sockaddr
*gateway
,
398 struct sockaddr
*netmask
, int flags
, struct sockaddr
*src
)
400 struct rtentry
*rt
= NULL
;
401 struct rt_addrinfo rtinfo
;
406 /* verify the gateway is directly reachable */
407 if ((ifa
= ifa_ifwithnet(gateway
)) == NULL
) {
413 * If the redirect isn't from our current router for this destination,
414 * it's either old or wrong.
416 if (!(flags
& RTF_DONE
) && /* XXX JH */
417 (rt
= rtpurelookup(dst
)) != NULL
&&
418 (!sa_equal(src
, rt
->rt_gateway
) || rt
->rt_ifa
!= ifa
)) {
424 * If it redirects us to ourselves, we have a routing loop,
425 * perhaps as a result of an interface going down recently.
427 if (ifa_ifwithaddr(gateway
)) {
428 error
= EHOSTUNREACH
;
433 * Create a new entry if the lookup failed or if we got back
434 * a wildcard entry for the default route. This is necessary
435 * for hosts which use routing redirects generated by smart
436 * gateways to dynamically build the routing tables.
440 if ((rt_mask(rt
) != NULL
&& rt_mask(rt
)->sa_len
< 2)) {
445 /* Ignore redirects for directly connected hosts. */
446 if (!(rt
->rt_flags
& RTF_GATEWAY
)) {
447 error
= EHOSTUNREACH
;
451 if (!(rt
->rt_flags
& RTF_HOST
) && (flags
& RTF_HOST
)) {
453 * Changing from a network route to a host route.
454 * Create a new host route rather than smashing the
458 flags
|= RTF_GATEWAY
| RTF_DYNAMIC
;
459 bzero(&rtinfo
, sizeof(struct rt_addrinfo
));
460 rtinfo
.rti_info
[RTAX_DST
] = dst
;
461 rtinfo
.rti_info
[RTAX_GATEWAY
] = gateway
;
462 rtinfo
.rti_info
[RTAX_NETMASK
] = netmask
;
463 rtinfo
.rti_flags
= flags
;
464 rtinfo
.rti_ifa
= ifa
;
465 rt
= NULL
; /* copy-in/copy-out parameter */
466 error
= rtrequest1(RTM_ADD
, &rtinfo
, &rt
);
468 flags
= rt
->rt_flags
;
469 stat
= &rtstat
.rts_dynamic
;
472 * Smash the current notion of the gateway to this destination.
473 * Should check about netmask!!!
475 rt
->rt_flags
|= RTF_MODIFIED
;
476 flags
|= RTF_MODIFIED
;
478 /* We only need to report rtmsg on CPU0 */
479 rt_setgate(rt
, rt_key(rt
), gateway
,
480 mycpuid
== 0 ? RTL_REPORTMSG
: RTL_DONTREPORT
);
482 stat
= &rtstat
.rts_newgateway
;
490 rtstat
.rts_badredirect
++;
491 else if (stat
!= NULL
)
499 struct netmsg_rtredirect
{
500 struct netmsg netmsg
;
501 struct sockaddr
*dst
;
502 struct sockaddr
*gateway
;
503 struct sockaddr
*netmask
;
505 struct sockaddr
*src
;
511 * Force a routing table entry to the specified
512 * destination to go through the given gateway.
513 * Normally called as a result of a routing redirect
514 * message from the network layer.
516 * N.B.: must be called at splnet
519 rtredirect(struct sockaddr
*dst
, struct sockaddr
*gateway
,
520 struct sockaddr
*netmask
, int flags
, struct sockaddr
*src
)
522 struct rt_addrinfo rtinfo
;
525 struct netmsg_rtredirect msg
;
527 netmsg_init(&msg
.netmsg
, &curthread
->td_msgport
, 0,
528 rtredirect_msghandler
);
530 msg
.gateway
= gateway
;
531 msg
.netmask
= netmask
;
534 error
= lwkt_domsg(rtable_portfn(0), &msg
.netmsg
.nm_lmsg
, 0);
536 error
= rtredirect_oncpu(dst
, gateway
, netmask
, flags
, src
);
538 bzero(&rtinfo
, sizeof(struct rt_addrinfo
));
539 rtinfo
.rti_info
[RTAX_DST
] = dst
;
540 rtinfo
.rti_info
[RTAX_GATEWAY
] = gateway
;
541 rtinfo
.rti_info
[RTAX_NETMASK
] = netmask
;
542 rtinfo
.rti_info
[RTAX_AUTHOR
] = src
;
543 rt_missmsg(RTM_REDIRECT
, &rtinfo
, flags
, error
);
549 rtredirect_msghandler(struct netmsg
*netmsg
)
551 struct netmsg_rtredirect
*msg
= (void *)netmsg
;
554 rtredirect_oncpu(msg
->dst
, msg
->gateway
, msg
->netmask
,
555 msg
->flags
, msg
->src
);
556 nextcpu
= mycpuid
+ 1;
558 lwkt_forwardmsg(rtable_portfn(nextcpu
), &netmsg
->nm_lmsg
);
560 lwkt_replymsg(&netmsg
->nm_lmsg
, 0);
566 * Routing table ioctl interface.
569 rtioctl(u_long req
, caddr_t data
, struct ucred
*cred
)
572 /* Multicast goop, grrr... */
573 return mrt_ioctl
? mrt_ioctl(req
, data
) : EOPNOTSUPP
;
580 ifa_ifwithroute(int flags
, struct sockaddr
*dst
, struct sockaddr
*gateway
)
584 if (!(flags
& RTF_GATEWAY
)) {
586 * If we are adding a route to an interface,
587 * and the interface is a point-to-point link,
588 * we should search for the destination
589 * as our clue to the interface. Otherwise
590 * we can use the local address.
593 if (flags
& RTF_HOST
) {
594 ifa
= ifa_ifwithdstaddr(dst
);
597 ifa
= ifa_ifwithaddr(gateway
);
600 * If we are adding a route to a remote net
601 * or host, the gateway may still be on the
602 * other end of a pt to pt link.
604 ifa
= ifa_ifwithdstaddr(gateway
);
607 ifa
= ifa_ifwithnet(gateway
);
611 rt
= rtpurelookup(gateway
);
615 if ((ifa
= rt
->rt_ifa
) == NULL
)
618 if (ifa
->ifa_addr
->sa_family
!= dst
->sa_family
) {
619 struct ifaddr
*oldifa
= ifa
;
621 ifa
= ifaof_ifpforaddr(dst
, ifa
->ifa_ifp
);
628 static int rt_fixdelete (struct radix_node
*, void *);
629 static int rt_fixchange (struct radix_node
*, void *);
633 struct radix_node_head
*rnh
;
637 * Set rtinfo->rti_ifa and rtinfo->rti_ifp.
640 rt_getifa(struct rt_addrinfo
*rtinfo
)
642 struct sockaddr
*gateway
= rtinfo
->rti_info
[RTAX_GATEWAY
];
643 struct sockaddr
*dst
= rtinfo
->rti_info
[RTAX_DST
];
644 struct sockaddr
*ifaaddr
= rtinfo
->rti_info
[RTAX_IFA
];
645 int flags
= rtinfo
->rti_flags
;
648 * ifp may be specified by sockaddr_dl
649 * when protocol address is ambiguous.
651 if (rtinfo
->rti_ifp
== NULL
) {
652 struct sockaddr
*ifpaddr
;
654 ifpaddr
= rtinfo
->rti_info
[RTAX_IFP
];
655 if (ifpaddr
!= NULL
&& ifpaddr
->sa_family
== AF_LINK
) {
658 ifa
= ifa_ifwithnet(ifpaddr
);
660 rtinfo
->rti_ifp
= ifa
->ifa_ifp
;
664 if (rtinfo
->rti_ifa
== NULL
&& ifaaddr
!= NULL
)
665 rtinfo
->rti_ifa
= ifa_ifwithaddr(ifaaddr
);
666 if (rtinfo
->rti_ifa
== NULL
) {
669 sa
= ifaaddr
!= NULL
? ifaaddr
:
670 (gateway
!= NULL
? gateway
: dst
);
671 if (sa
!= NULL
&& rtinfo
->rti_ifp
!= NULL
)
672 rtinfo
->rti_ifa
= ifaof_ifpforaddr(sa
, rtinfo
->rti_ifp
);
673 else if (dst
!= NULL
&& gateway
!= NULL
)
674 rtinfo
->rti_ifa
= ifa_ifwithroute(flags
, dst
, gateway
);
676 rtinfo
->rti_ifa
= ifa_ifwithroute(flags
, sa
, sa
);
678 if (rtinfo
->rti_ifa
== NULL
)
679 return (ENETUNREACH
);
681 if (rtinfo
->rti_ifp
== NULL
)
682 rtinfo
->rti_ifp
= rtinfo
->rti_ifa
->ifa_ifp
;
687 * Do appropriate manipulations of a routing tree given
688 * all the bits of info needed
693 struct sockaddr
*dst
,
694 struct sockaddr
*gateway
,
695 struct sockaddr
*netmask
,
697 struct rtentry
**ret_nrt
)
699 struct rt_addrinfo rtinfo
;
701 bzero(&rtinfo
, sizeof(struct rt_addrinfo
));
702 rtinfo
.rti_info
[RTAX_DST
] = dst
;
703 rtinfo
.rti_info
[RTAX_GATEWAY
] = gateway
;
704 rtinfo
.rti_info
[RTAX_NETMASK
] = netmask
;
705 rtinfo
.rti_flags
= flags
;
706 return rtrequest1(req
, &rtinfo
, ret_nrt
);
712 struct sockaddr
*dst
,
713 struct sockaddr
*gateway
,
714 struct sockaddr
*netmask
,
717 struct rt_addrinfo rtinfo
;
719 bzero(&rtinfo
, sizeof(struct rt_addrinfo
));
720 rtinfo
.rti_info
[RTAX_DST
] = dst
;
721 rtinfo
.rti_info
[RTAX_GATEWAY
] = gateway
;
722 rtinfo
.rti_info
[RTAX_NETMASK
] = netmask
;
723 rtinfo
.rti_flags
= flags
;
724 return rtrequest1_global(req
, &rtinfo
, NULL
, NULL
);
730 struct netmsg netmsg
;
732 struct rt_addrinfo
*rtinfo
;
733 rtrequest1_callback_func_t callback
;
740 rtrequest1_global(int req
, struct rt_addrinfo
*rtinfo
,
741 rtrequest1_callback_func_t callback
, void *arg
)
745 struct netmsg_rtq msg
;
747 netmsg_init(&msg
.netmsg
, &curthread
->td_msgport
, 0,
748 rtrequest1_msghandler
);
749 msg
.netmsg
.nm_lmsg
.ms_error
= -1;
752 msg
.callback
= callback
;
754 error
= lwkt_domsg(rtable_portfn(0), &msg
.netmsg
.nm_lmsg
, 0);
756 struct rtentry
*rt
= NULL
;
758 error
= rtrequest1(req
, rtinfo
, &rt
);
762 callback(req
, error
, rtinfo
, rt
, arg
);
768 * Handle a route table request on the current cpu. Since the route table's
769 * are supposed to be identical on each cpu, an error occuring later in the
770 * message chain is considered system-fatal.
775 rtrequest1_msghandler(struct netmsg
*netmsg
)
777 struct netmsg_rtq
*msg
= (void *)netmsg
;
778 struct rtentry
*rt
= NULL
;
782 error
= rtrequest1(msg
->req
, msg
->rtinfo
, &rt
);
786 msg
->callback(msg
->req
, error
, msg
->rtinfo
, rt
, msg
->arg
);
789 * RTM_DELETE's are propogated even if an error occurs, since a
790 * cloned route might be undergoing deletion and cloned routes
791 * are not necessarily replicated. An overall error is returned
792 * only if no cpus have the route in question.
794 if (msg
->netmsg
.nm_lmsg
.ms_error
< 0 || error
== 0)
795 msg
->netmsg
.nm_lmsg
.ms_error
= error
;
797 nextcpu
= mycpuid
+ 1;
798 if (error
&& msg
->req
!= RTM_DELETE
) {
800 panic("rtrequest1_msghandler: rtrequest table "
801 "error was not on cpu #0: %p", msg
->rtinfo
);
803 lwkt_replymsg(&msg
->netmsg
.nm_lmsg
, error
);
804 } else if (nextcpu
< ncpus
) {
805 lwkt_forwardmsg(rtable_portfn(nextcpu
), &msg
->netmsg
.nm_lmsg
);
807 lwkt_replymsg(&msg
->netmsg
.nm_lmsg
,
808 msg
->netmsg
.nm_lmsg
.ms_error
);
815 rtrequest1(int req
, struct rt_addrinfo
*rtinfo
, struct rtentry
**ret_nrt
)
817 struct sockaddr
*dst
= rtinfo
->rti_info
[RTAX_DST
];
819 struct radix_node
*rn
;
820 struct radix_node_head
*rnh
;
822 struct sockaddr
*ndst
;
826 #define gotoerr(x) { error = x ; goto bad; }
830 rt_addrinfo_print(req
, rtinfo
);
835 * Find the correct routing tree to use for this Address Family
837 if ((rnh
= rt_tables
[mycpuid
][dst
->sa_family
]) == NULL
)
838 gotoerr(EAFNOSUPPORT
);
841 * If we are adding a host route then we don't want to put
842 * a netmask in the tree, nor do we want to clone it.
844 if (rtinfo
->rti_flags
& RTF_HOST
) {
845 rtinfo
->rti_info
[RTAX_NETMASK
] = NULL
;
846 rtinfo
->rti_flags
&= ~(RTF_CLONING
| RTF_PRCLONING
);
851 /* Remove the item from the tree. */
852 rn
= rnh
->rnh_deladdr((char *)rtinfo
->rti_info
[RTAX_DST
],
853 (char *)rtinfo
->rti_info
[RTAX_NETMASK
],
857 KASSERT(!(rn
->rn_flags
& (RNF_ACTIVE
| RNF_ROOT
)),
858 ("rnh_deladdr returned flags 0x%x", rn
->rn_flags
));
859 rt
= (struct rtentry
*)rn
;
861 /* ref to prevent a deletion race */
864 /* Free any routes cloned from this one. */
865 if ((rt
->rt_flags
& (RTF_CLONING
| RTF_PRCLONING
)) &&
866 rt_mask(rt
) != NULL
) {
867 rnh
->rnh_walktree_from(rnh
, (char *)rt_key(rt
),
872 if (rt
->rt_gwroute
!= NULL
) {
873 RTFREE(rt
->rt_gwroute
);
874 rt
->rt_gwroute
= NULL
;
878 * NB: RTF_UP must be set during the search above,
879 * because we might delete the last ref, causing
880 * rt to get freed prematurely.
882 rt
->rt_flags
&= ~RTF_UP
;
886 rt_print(rtinfo
, rt
);
889 /* Give the protocol a chance to keep things in sync. */
890 if ((ifa
= rt
->rt_ifa
) && ifa
->ifa_rtrequest
)
891 ifa
->ifa_rtrequest(RTM_DELETE
, rt
, rtinfo
);
894 * If the caller wants it, then it can have it,
895 * but it's up to it to free the rtentry as we won't be
898 KASSERT(rt
->rt_refcnt
>= 0,
899 ("rtrequest1(DELETE): refcnt %ld", rt
->rt_refcnt
));
900 if (ret_nrt
!= NULL
) {
901 /* leave ref intact for return */
904 /* deref / attempt to destroy */
910 if (ret_nrt
== NULL
|| (rt
= *ret_nrt
) == NULL
)
914 rt
->rt_flags
& ~(RTF_CLONING
| RTF_PRCLONING
| RTF_STATIC
);
915 rtinfo
->rti_flags
|= RTF_WASCLONED
;
916 rtinfo
->rti_info
[RTAX_GATEWAY
] = rt
->rt_gateway
;
917 if ((rtinfo
->rti_info
[RTAX_NETMASK
] = rt
->rt_genmask
) == NULL
)
918 rtinfo
->rti_flags
|= RTF_HOST
;
919 rtinfo
->rti_info
[RTAX_MPLS1
] = rt
->rt_shim
[0];
920 rtinfo
->rti_info
[RTAX_MPLS2
] = rt
->rt_shim
[1];
921 rtinfo
->rti_info
[RTAX_MPLS3
] = rt
->rt_shim
[2];
925 KASSERT(!(rtinfo
->rti_flags
& RTF_GATEWAY
) ||
926 rtinfo
->rti_info
[RTAX_GATEWAY
] != NULL
,
927 ("rtrequest: GATEWAY but no gateway"));
929 if (rtinfo
->rti_ifa
== NULL
&& (error
= rt_getifa(rtinfo
)))
931 ifa
= rtinfo
->rti_ifa
;
933 R_Malloc(rt
, struct rtentry
*, sizeof(struct rtentry
));
936 bzero(rt
, sizeof(struct rtentry
));
937 rt
->rt_flags
= RTF_UP
| rtinfo
->rti_flags
;
938 rt
->rt_cpuid
= mycpuid
;
940 if (mycpuid
!= 0 && req
== RTM_ADD
) {
941 /* For RTM_ADD, we have already sent rtmsg on CPU0. */
942 reportmsg
= RTL_DONTREPORT
;
945 * For RTM_ADD, we only send rtmsg on CPU0.
946 * For RTM_RESOLVE, we always send rtmsg. XXX
948 reportmsg
= RTL_REPORTMSG
;
950 error
= rt_setgate(rt
, dst
, rtinfo
->rti_info
[RTAX_GATEWAY
],
958 if (rtinfo
->rti_info
[RTAX_NETMASK
] != NULL
)
959 rt_maskedcopy(dst
, ndst
,
960 rtinfo
->rti_info
[RTAX_NETMASK
]);
962 bcopy(dst
, ndst
, dst
->sa_len
);
964 if (rtinfo
->rti_info
[RTAX_MPLS1
] != NULL
)
965 rt_setshims(rt
, rtinfo
->rti_info
);
968 * Note that we now have a reference to the ifa.
969 * This moved from below so that rnh->rnh_addaddr() can
970 * examine the ifa and ifa->ifa_ifp if it so desires.
974 rt
->rt_ifp
= ifa
->ifa_ifp
;
975 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
977 rn
= rnh
->rnh_addaddr((char *)ndst
,
978 (char *)rtinfo
->rti_info
[RTAX_NETMASK
],
981 struct rtentry
*oldrt
;
984 * We already have one of these in the tree.
985 * We do a special hack: if the old route was
986 * cloned, then we blow it away and try
987 * re-inserting the new one.
989 oldrt
= rtpurelookup(ndst
);
992 if (oldrt
->rt_flags
& RTF_WASCLONED
) {
993 rtrequest(RTM_DELETE
, rt_key(oldrt
),
996 oldrt
->rt_flags
, NULL
);
997 rn
= rnh
->rnh_addaddr((char *)ndst
,
999 rtinfo
->rti_info
[RTAX_NETMASK
],
1006 * If it still failed to go into the tree,
1007 * then un-make it (this should be a function).
1010 if (rt
->rt_gwroute
!= NULL
)
1011 rtfree(rt
->rt_gwroute
);
1019 * If we got here from RESOLVE, then we are cloning
1020 * so clone the rest, and note that we
1021 * are a clone (and increment the parent's references)
1023 if (req
== RTM_RESOLVE
) {
1024 rt
->rt_rmx
= (*ret_nrt
)->rt_rmx
; /* copy metrics */
1025 rt
->rt_rmx
.rmx_pksent
= 0; /* reset packet counter */
1026 if ((*ret_nrt
)->rt_flags
&
1027 (RTF_CLONING
| RTF_PRCLONING
)) {
1028 rt
->rt_parent
= *ret_nrt
;
1029 (*ret_nrt
)->rt_refcnt
++;
1034 * if this protocol has something to add to this then
1035 * allow it to do that as well.
1037 if (ifa
->ifa_rtrequest
!= NULL
)
1038 ifa
->ifa_rtrequest(req
, rt
, rtinfo
);
1041 * We repeat the same procedure from rt_setgate() here because
1042 * it doesn't fire when we call it there because the node
1043 * hasn't been added to the tree yet.
1045 if (req
== RTM_ADD
&& !(rt
->rt_flags
& RTF_HOST
) &&
1046 rt_mask(rt
) != NULL
) {
1047 struct rtfc_arg arg
= { rt
, rnh
};
1049 rnh
->rnh_walktree_from(rnh
, (char *)rt_key(rt
),
1050 (char *)rt_mask(rt
),
1051 rt_fixchange
, &arg
);
1056 rt_print(rtinfo
, rt
);
1059 * Return the resulting rtentry,
1060 * increasing the number of references by one.
1062 if (ret_nrt
!= NULL
) {
1074 kprintf("rti %p failed error %d\n", rtinfo
, error
);
1076 kprintf("rti %p succeeded\n", rtinfo
);
1084 * Called from rtrequest(RTM_DELETE, ...) to fix up the route's ``family''
1085 * (i.e., the routes related to it by the operation of cloning). This
1086 * routine is iterated over all potential former-child-routes by way of
1087 * rnh->rnh_walktree_from() above, and those that actually are children of
1088 * the late parent (passed in as VP here) are themselves deleted.
1091 rt_fixdelete(struct radix_node
*rn
, void *vp
)
1093 struct rtentry
*rt
= (struct rtentry
*)rn
;
1094 struct rtentry
*rt0
= vp
;
1096 if (rt
->rt_parent
== rt0
&&
1097 !(rt
->rt_flags
& (RTF_PINNED
| RTF_CLONING
| RTF_PRCLONING
))) {
1098 return rtrequest(RTM_DELETE
, rt_key(rt
), NULL
, rt_mask(rt
),
1099 rt
->rt_flags
, NULL
);
1105 * This routine is called from rt_setgate() to do the analogous thing for
1106 * adds and changes. There is the added complication in this case of a
1107 * middle insert; i.e., insertion of a new network route between an older
1108 * network route and (cloned) host routes. For this reason, a simple check
1109 * of rt->rt_parent is insufficient; each candidate route must be tested
1110 * against the (mask, value) of the new route (passed as before in vp)
1111 * to see if the new route matches it.
1113 * XXX - it may be possible to do fixdelete() for changes and reserve this
1114 * routine just for adds. I'm not sure why I thought it was necessary to do
1118 static int rtfcdebug
= 0;
1122 rt_fixchange(struct radix_node
*rn
, void *vp
)
1124 struct rtentry
*rt
= (struct rtentry
*)rn
;
1125 struct rtfc_arg
*ap
= vp
;
1126 struct rtentry
*rt0
= ap
->rt0
;
1127 struct radix_node_head
*rnh
= ap
->rnh
;
1128 u_char
*xk1
, *xm1
, *xk2
, *xmp
;
1133 kprintf("rt_fixchange: rt %p, rt0 %p\n", rt
, rt0
);
1136 if (rt
->rt_parent
== NULL
||
1137 (rt
->rt_flags
& (RTF_PINNED
| RTF_CLONING
| RTF_PRCLONING
))) {
1139 if (rtfcdebug
) kprintf("no parent, pinned or cloning\n");
1144 if (rt
->rt_parent
== rt0
) {
1146 if (rtfcdebug
) kprintf("parent match\n");
1148 return rtrequest(RTM_DELETE
, rt_key(rt
), NULL
, rt_mask(rt
),
1149 rt
->rt_flags
, NULL
);
1153 * There probably is a function somewhere which does this...
1154 * if not, there should be.
1156 len
= imin(rt_key(rt0
)->sa_len
, rt_key(rt
)->sa_len
);
1158 xk1
= (u_char
*)rt_key(rt0
);
1159 xm1
= (u_char
*)rt_mask(rt0
);
1160 xk2
= (u_char
*)rt_key(rt
);
1162 /* avoid applying a less specific route */
1163 xmp
= (u_char
*)rt_mask(rt
->rt_parent
);
1164 mlen
= rt_key(rt
->rt_parent
)->sa_len
;
1165 if (mlen
> rt_key(rt0
)->sa_len
) {
1168 kprintf("rt_fixchange: inserting a less "
1169 "specific route\n");
1173 for (i
= rnh
->rnh_treetop
->rn_offset
; i
< mlen
; i
++) {
1174 if ((xmp
[i
] & ~(xmp
[i
] ^ xm1
[i
])) != xmp
[i
]) {
1177 kprintf("rt_fixchange: inserting a less "
1178 "specific route\n");
1184 for (i
= rnh
->rnh_treetop
->rn_offset
; i
< len
; i
++) {
1185 if ((xk2
[i
] & xm1
[i
]) != xk1
[i
]) {
1187 if (rtfcdebug
) kprintf("no match\n");
1194 * OK, this node is a clone, and matches the node currently being
1195 * changed/added under the node's mask. So, get rid of it.
1198 if (rtfcdebug
) kprintf("deleting\n");
1200 return rtrequest(RTM_DELETE
, rt_key(rt
), NULL
, rt_mask(rt
),
1201 rt
->rt_flags
, NULL
);
1204 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
1207 rt_setgate(struct rtentry
*rt0
, struct sockaddr
*dst
, struct sockaddr
*gate
,
1208 boolean_t generate_report
)
1210 char *space
, *oldspace
;
1211 int dlen
= ROUNDUP(dst
->sa_len
), glen
= ROUNDUP(gate
->sa_len
);
1212 struct rtentry
*rt
= rt0
;
1213 struct radix_node_head
*rnh
= rt_tables
[mycpuid
][dst
->sa_family
];
1216 * A host route with the destination equal to the gateway
1217 * will interfere with keeping LLINFO in the routing
1218 * table, so disallow it.
1220 if (((rt0
->rt_flags
& (RTF_HOST
| RTF_GATEWAY
| RTF_LLINFO
)) ==
1221 (RTF_HOST
| RTF_GATEWAY
)) &&
1222 dst
->sa_len
== gate
->sa_len
&&
1223 sa_equal(dst
, gate
)) {
1225 * The route might already exist if this is an RTM_CHANGE
1226 * or a routing redirect, so try to delete it.
1228 if (rt_key(rt0
) != NULL
)
1229 rtrequest(RTM_DELETE
, rt_key(rt0
), rt0
->rt_gateway
,
1230 rt_mask(rt0
), rt0
->rt_flags
, NULL
);
1231 return EADDRNOTAVAIL
;
1235 * Both dst and gateway are stored in the same malloc'ed chunk
1236 * (If I ever get my hands on....)
1237 * if we need to malloc a new chunk, then keep the old one around
1238 * till we don't need it any more.
1240 if (rt
->rt_gateway
== NULL
|| glen
> ROUNDUP(rt
->rt_gateway
->sa_len
)) {
1241 oldspace
= (char *)rt_key(rt
);
1242 R_Malloc(space
, char *, dlen
+ glen
);
1245 rt
->rt_nodes
->rn_key
= space
;
1247 space
= (char *)rt_key(rt
); /* Just use the old space. */
1251 /* Set the gateway value. */
1252 rt
->rt_gateway
= (struct sockaddr
*)(space
+ dlen
);
1253 bcopy(gate
, rt
->rt_gateway
, glen
);
1255 if (oldspace
!= NULL
) {
1257 * If we allocated a new chunk, preserve the original dst.
1258 * This way, rt_setgate() really just sets the gate
1259 * and leaves the dst field alone.
1261 bcopy(dst
, space
, dlen
);
1266 * If there is already a gwroute, it's now almost definitely wrong
1269 if (rt
->rt_gwroute
!= NULL
) {
1270 RTFREE(rt
->rt_gwroute
);
1271 rt
->rt_gwroute
= NULL
;
1273 if (rt
->rt_flags
& RTF_GATEWAY
) {
1275 * Cloning loop avoidance: In the presence of
1276 * protocol-cloning and bad configuration, it is
1277 * possible to get stuck in bottomless mutual recursion
1278 * (rtrequest rt_setgate rtlookup). We avoid this
1279 * by not allowing protocol-cloning to operate for
1280 * gateways (which is probably the correct choice
1281 * anyway), and avoid the resulting reference loops
1282 * by disallowing any route to run through itself as
1283 * a gateway. This is obviously mandatory when we
1284 * get rt->rt_output().
1286 * This breaks TTCP for hosts outside the gateway! XXX JH
1288 rt
->rt_gwroute
= _rtlookup(gate
, generate_report
,
1290 if (rt
->rt_gwroute
== rt
) {
1291 rt
->rt_gwroute
= NULL
;
1293 return EDQUOT
; /* failure */
1298 * This isn't going to do anything useful for host routes, so
1299 * don't bother. Also make sure we have a reasonable mask
1300 * (we don't yet have one during adds).
1302 if (!(rt
->rt_flags
& RTF_HOST
) && rt_mask(rt
) != NULL
) {
1303 struct rtfc_arg arg
= { rt
, rnh
};
1305 rnh
->rnh_walktree_from(rnh
, (char *)rt_key(rt
),
1306 (char *)rt_mask(rt
),
1307 rt_fixchange
, &arg
);
1315 struct sockaddr
*src
,
1316 struct sockaddr
*dst
,
1317 struct sockaddr
*netmask
)
1319 u_char
*cp1
= (u_char
*)src
;
1320 u_char
*cp2
= (u_char
*)dst
;
1321 u_char
*cp3
= (u_char
*)netmask
;
1322 u_char
*cplim
= cp2
+ *cp3
;
1323 u_char
*cplim2
= cp2
+ *cp1
;
1325 *cp2
++ = *cp1
++; *cp2
++ = *cp1
++; /* copies sa_len & sa_family */
1330 *cp2
++ = *cp1
++ & *cp3
++;
1332 bzero(cp2
, cplim2
- cp2
);
1336 rt_llroute(struct sockaddr
*dst
, struct rtentry
*rt0
, struct rtentry
**drt
)
1338 struct rtentry
*up_rt
, *rt
;
1340 if (!(rt0
->rt_flags
& RTF_UP
)) {
1341 up_rt
= rtlookup(dst
);
1343 return (EHOSTUNREACH
);
1347 if (up_rt
->rt_flags
& RTF_GATEWAY
) {
1348 if (up_rt
->rt_gwroute
== NULL
) {
1349 up_rt
->rt_gwroute
= rtlookup(up_rt
->rt_gateway
);
1350 if (up_rt
->rt_gwroute
== NULL
)
1351 return (EHOSTUNREACH
);
1352 } else if (!(up_rt
->rt_gwroute
->rt_flags
& RTF_UP
)) {
1353 rtfree(up_rt
->rt_gwroute
);
1354 up_rt
->rt_gwroute
= rtlookup(up_rt
->rt_gateway
);
1355 if (up_rt
->rt_gwroute
== NULL
)
1356 return (EHOSTUNREACH
);
1358 rt
= up_rt
->rt_gwroute
;
1361 if (rt
->rt_flags
& RTF_REJECT
&&
1362 (rt
->rt_rmx
.rmx_expire
== 0 || /* rt doesn't expire */
1363 time_second
< rt
->rt_rmx
.rmx_expire
)) /* rt not expired */
1364 return (rt
->rt_flags
& RTF_HOST
? EHOSTDOWN
: EHOSTUNREACH
);
1370 rt_setshims(struct rtentry
*rt
, struct sockaddr
**rt_shim
){
1373 for (i
=0; i
<3; i
++) {
1374 struct sockaddr
*shim
= rt_shim
[RTAX_MPLS1
+ i
];
1380 shimlen
= ROUNDUP(shim
->sa_len
);
1381 R_Malloc(rt
->rt_shim
[i
], struct sockaddr
*, shimlen
);
1382 bcopy(shim
, rt
->rt_shim
[i
], shimlen
);
1391 * Print out a route table entry
1394 rt_print(struct rt_addrinfo
*rtinfo
, struct rtentry
*rn
)
1396 kprintf("rti %p cpu %d route %p flags %08lx: ",
1397 rtinfo
, mycpuid
, rn
, rn
->rt_flags
);
1398 sockaddr_print(rt_key(rn
));
1400 sockaddr_print(rt_mask(rn
));
1402 sockaddr_print(rn
->rt_gateway
);
1403 kprintf(" ifc \"%s\"", rn
->rt_ifp
? rn
->rt_ifp
->if_dname
: "?");
1404 kprintf(" ifa %p\n", rn
->rt_ifa
);
1408 rt_addrinfo_print(int cmd
, struct rt_addrinfo
*rti
)
1414 if (cmd
== RTM_DELETE
&& route_debug
> 1)
1429 kprintf("C%02d ", cmd
);
1432 kprintf("rti %p cpu %d ", rti
, mycpuid
);
1433 for (i
= 0; i
< rti
->rti_addrs
; ++i
) {
1434 if (rti
->rti_info
[i
] == NULL
)
1464 kprintf("(?%02d ", i
);
1467 sockaddr_print(rti
->rti_info
[i
]);
1475 sockaddr_print(struct sockaddr
*sa
)
1477 struct sockaddr_in
*sa4
;
1478 struct sockaddr_in6
*sa6
;
1487 len
= sa
->sa_len
- offsetof(struct sockaddr
, sa_data
[0]);
1489 switch(sa
->sa_family
) {
1493 switch(sa
->sa_family
) {
1495 sa4
= (struct sockaddr_in
*)sa
;
1496 kprintf("INET %d %d.%d.%d.%d",
1497 ntohs(sa4
->sin_port
),
1498 (ntohl(sa4
->sin_addr
.s_addr
) >> 24) & 255,
1499 (ntohl(sa4
->sin_addr
.s_addr
) >> 16) & 255,
1500 (ntohl(sa4
->sin_addr
.s_addr
) >> 8) & 255,
1501 (ntohl(sa4
->sin_addr
.s_addr
) >> 0) & 255
1505 sa6
= (struct sockaddr_in6
*)sa
;
1506 kprintf("INET6 %d %04x:%04x%04x:%04x:%04x:%04x:%04x:%04x",
1507 ntohs(sa6
->sin6_port
),
1508 sa6
->sin6_addr
.s6_addr16
[0],
1509 sa6
->sin6_addr
.s6_addr16
[1],
1510 sa6
->sin6_addr
.s6_addr16
[2],
1511 sa6
->sin6_addr
.s6_addr16
[3],
1512 sa6
->sin6_addr
.s6_addr16
[4],
1513 sa6
->sin6_addr
.s6_addr16
[5],
1514 sa6
->sin6_addr
.s6_addr16
[6],
1515 sa6
->sin6_addr
.s6_addr16
[7]
1519 kprintf("AF%d ", sa
->sa_family
);
1520 while (len
> 0 && sa
->sa_data
[len
-1] == 0)
1523 for (i
= 0; i
< len
; ++i
) {
1526 kprintf("%d", (unsigned char)sa
->sa_data
[i
]);
1536 * Set up a routing table entry, normally for an interface.
1539 rtinit(struct ifaddr
*ifa
, int cmd
, int flags
)
1541 struct sockaddr
*dst
, *deldst
, *netmask
;
1542 struct mbuf
*m
= NULL
;
1543 struct radix_node_head
*rnh
;
1544 struct radix_node
*rn
;
1545 struct rt_addrinfo rtinfo
;
1548 if (flags
& RTF_HOST
) {
1549 dst
= ifa
->ifa_dstaddr
;
1552 dst
= ifa
->ifa_addr
;
1553 netmask
= ifa
->ifa_netmask
;
1556 * If it's a delete, check that if it exists, it's on the correct
1557 * interface or we might scrub a route to another ifa which would
1558 * be confusing at best and possibly worse.
1560 if (cmd
== RTM_DELETE
) {
1562 * It's a delete, so it should already exist..
1563 * If it's a net, mask off the host bits
1564 * (Assuming we have a mask)
1566 if (netmask
!= NULL
) {
1567 m
= m_get(MB_DONTWAIT
, MT_SONAME
);
1571 deldst
= mtod(m
, struct sockaddr
*);
1572 rt_maskedcopy(dst
, deldst
, netmask
);
1576 * Look up an rtentry that is in the routing tree and
1577 * contains the correct info.
1579 if ((rnh
= rt_tables
[mycpuid
][dst
->sa_family
]) == NULL
||
1580 (rn
= rnh
->rnh_lookup((char *)dst
,
1581 (char *)netmask
, rnh
)) == NULL
||
1582 ((struct rtentry
*)rn
)->rt_ifa
!= ifa
||
1583 !sa_equal((struct sockaddr
*)rn
->rn_key
, dst
)) {
1586 return (flags
& RTF_HOST
? EHOSTUNREACH
: ENETUNREACH
);
1592 * One would think that as we are deleting, and we know
1593 * it doesn't exist, we could just return at this point
1594 * with an "ELSE" clause, but apparently not..
1596 return (flags
& RTF_HOST
? EHOSTUNREACH
: ENETUNREACH
);
1601 * Do the actual request
1603 bzero(&rtinfo
, sizeof(struct rt_addrinfo
));
1604 rtinfo
.rti_info
[RTAX_DST
] = dst
;
1605 rtinfo
.rti_info
[RTAX_GATEWAY
] = ifa
->ifa_addr
;
1606 rtinfo
.rti_info
[RTAX_NETMASK
] = netmask
;
1607 rtinfo
.rti_flags
= flags
| ifa
->ifa_flags
;
1608 rtinfo
.rti_ifa
= ifa
;
1609 error
= rtrequest1_global(cmd
, &rtinfo
, rtinit_rtrequest_callback
, ifa
);
1616 rtinit_rtrequest_callback(int cmd
, int error
,
1617 struct rt_addrinfo
*rtinfo
, struct rtentry
*rt
,
1620 struct ifaddr
*ifa
= arg
;
1622 if (error
== 0 && rt
) {
1625 rt_newaddrmsg(cmd
, ifa
, error
, rt
);
1628 if (cmd
== RTM_DELETE
) {
1629 if (rt
->rt_refcnt
== 0) {
1638 struct netmsg netmsg
;
1640 struct rt_addrinfo
*rtinfo
;
1641 rtsearch_callback_func_t callback
;
1643 boolean_t exact_match
;
1648 rtsearch_global(int req
, struct rt_addrinfo
*rtinfo
,
1649 rtsearch_callback_func_t callback
, void *arg
,
1650 boolean_t exact_match
)
1652 struct netmsg_rts msg
;
1654 netmsg_init(&msg
.netmsg
, &curthread
->td_msgport
, 0,
1655 rtsearch_msghandler
);
1657 msg
.rtinfo
= rtinfo
;
1658 msg
.callback
= callback
;
1660 msg
.exact_match
= exact_match
;
1662 return lwkt_domsg(rtable_portfn(0), &msg
.netmsg
.nm_lmsg
, 0);
1666 rtsearch_msghandler(struct netmsg
*netmsg
)
1668 struct netmsg_rts
*msg
= (void *)netmsg
;
1669 struct rt_addrinfo
*rtinfo
= msg
->rtinfo
;
1670 struct radix_node_head
*rnh
;
1675 * Find the correct routing tree to use for this Address Family
1677 if ((rnh
= rt_tables
[mycpuid
][rtinfo
->rti_dst
->sa_family
]) == NULL
) {
1679 panic("partially initialized routing tables\n");
1680 lwkt_replymsg(&msg
->netmsg
.nm_lmsg
, EAFNOSUPPORT
);
1685 * Correct rtinfo for the host route searching.
1687 if (rtinfo
->rti_flags
& RTF_HOST
) {
1688 rtinfo
->rti_netmask
= NULL
;
1689 rtinfo
->rti_flags
&= ~(RTF_CLONING
| RTF_PRCLONING
);
1692 rt
= (struct rtentry
*)
1693 rnh
->rnh_lookup((char *)rtinfo
->rti_dst
,
1694 (char *)rtinfo
->rti_netmask
, rnh
);
1697 * If we are asked to do the "exact match", we need to make sure
1698 * that host route searching got a host route while a network
1699 * route searching got a network route.
1701 if (rt
!= NULL
&& msg
->exact_match
&&
1702 ((rt
->rt_flags
^ rtinfo
->rti_flags
) & RTF_HOST
))
1707 * No matching routes have been found, don't count this
1708 * as a critical error (here, we set 'error' to 0), just
1709 * keep moving on, since at least prcloned routes are not
1710 * duplicated onto each CPU.
1717 error
= msg
->callback(msg
->req
, msg
->rtinfo
, rt
, msg
->arg
,
1721 if (error
== EJUSTRETURN
) {
1722 lwkt_replymsg(&msg
->netmsg
.nm_lmsg
, 0);
1727 nextcpu
= mycpuid
+ 1;
1729 KKASSERT(msg
->found_cnt
> 0);
1732 * Under following cases, unrecoverable error has
1734 * o Request is RTM_GET
1735 * o The first time that we find the route, but the
1736 * modification fails.
1738 if (msg
->req
!= RTM_GET
&& msg
->found_cnt
> 1) {
1739 panic("rtsearch_msghandler: unrecoverable error "
1740 "cpu %d, rtinfo %p", mycpuid
, msg
->rtinfo
);
1742 lwkt_replymsg(&msg
->netmsg
.nm_lmsg
, error
);
1743 } else if (nextcpu
< ncpus
) {
1744 lwkt_forwardmsg(rtable_portfn(nextcpu
), &msg
->netmsg
.nm_lmsg
);
1746 if (msg
->found_cnt
== 0) {
1747 /* The requested route was never seen ... */
1750 lwkt_replymsg(&msg
->netmsg
.nm_lmsg
, error
);
1755 rtmask_add_global(struct sockaddr
*mask
)
1759 netmsg_init(&nmsg
, &curthread
->td_msgport
, 0,
1760 rtmask_add_msghandler
);
1761 nmsg
.nm_lmsg
.u
.ms_resultp
= mask
;
1763 return lwkt_domsg(rtable_portfn(0), &nmsg
.nm_lmsg
, 0);
1767 _rtmask_lookup(struct sockaddr
*mask
, boolean_t search
)
1769 struct radix_node
*n
;
1771 #define clen(s) (*(u_char *)(s))
1772 n
= rn_addmask((char *)mask
, search
, 1);
1774 mask
->sa_len
>= clen(n
->rn_key
) &&
1775 bcmp((char *)mask
+ 1,
1776 (char *)n
->rn_key
+ 1, clen(n
->rn_key
) - 1) == 0) {
1777 return (struct sockaddr
*)n
->rn_key
;
1785 rtmask_add_msghandler(struct netmsg
*nmsg
)
1787 struct lwkt_msg
*lmsg
= &nmsg
->nm_lmsg
;
1788 struct sockaddr
*mask
= lmsg
->u
.ms_resultp
;
1789 int error
= 0, nextcpu
;
1791 if (rtmask_lookup(mask
) == NULL
)
1794 nextcpu
= mycpuid
+ 1;
1795 if (!error
&& nextcpu
< ncpus
)
1796 lwkt_forwardmsg(rtable_portfn(nextcpu
), lmsg
);
1798 lwkt_replymsg(lmsg
, error
);
1801 /* This must be before ip6_init2(), which is now SI_ORDER_MIDDLE */
1802 SYSINIT(route
, SI_SUB_PROTO_DOMAIN
, SI_ORDER_THIRD
, route_init
, 0);