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) 1982, 1986, 1988, 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. Neither the name of the University nor the names of its contributors
46 * may be used to endorse or promote products derived from this software
47 * without specific prior written permission.
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
61 * @(#)if_ether.c 8.1 (Berkeley) 6/10/93
62 * $FreeBSD: src/sys/netinet/if_ether.c,v 1.64.2.23 2003/04/11 07:23:15 fjoe Exp $
66 * Ethernet address resolution protocol.
68 * add "inuse/lock" bit (or ref. count) along with valid bit
74 #include <sys/param.h>
75 #include <sys/kernel.h>
76 #include <sys/queue.h>
77 #include <sys/sysctl.h>
78 #include <sys/systm.h>
80 #include <sys/malloc.h>
81 #include <sys/socket.h>
82 #include <sys/syslog.h>
86 #include <net/if_dl.h>
87 #include <net/if_types.h>
88 #include <net/route.h>
89 #include <net/netisr.h>
90 #include <net/if_llc.h>
92 #include <netinet/in.h>
93 #include <netinet/in_var.h>
94 #include <netinet/if_ether.h>
96 #include <sys/thread2.h>
97 #include <sys/msgport2.h>
98 #include <net/netmsg2.h>
99 #include <net/netisr2.h>
100 #include <sys/mplock2.h>
103 #include <netinet/ip_carp.h>
106 #define SIN(s) ((struct sockaddr_in *)s)
107 #define SDL(s) ((struct sockaddr_dl *)s)
109 SYSCTL_DECL(_net_link_ether
);
110 SYSCTL_NODE(_net_link_ether
, PF_INET
, inet
, CTLFLAG_RW
, 0, "");
113 static int arpt_prune
= (5*60*1); /* walk list every 5 minutes */
114 static int arpt_keep
= (20*60); /* once resolved, good for 20 more minutes */
115 static int arpt_down
= 20; /* once declared down, don't send for 20 sec */
117 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, prune_intvl
, CTLFLAG_RW
,
119 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, max_age
, CTLFLAG_RW
,
121 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, host_down_time
, CTLFLAG_RW
,
124 #define rt_expire rt_rmx.rmx_expire
127 LIST_ENTRY(llinfo_arp
) la_le
;
128 struct rtentry
*la_rt
;
129 struct mbuf
*la_hold
; /* last packet until resolved/timeout */
130 u_short la_preempt
; /* countdown for pre-expiry arps */
131 u_short la_asked
; /* #times we QUERIED following expiration */
134 static int arp_maxtries
= 5;
135 static int useloopback
= 1; /* use loopback interface for local traffic */
136 static int arp_proxyall
= 0;
137 static int arp_refresh
= 60; /* refresh arp cache ~60 (not impl yet) */
138 static int arp_restricted_match
= 0;
139 static int arp_ignore_probes
= 1;
141 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, maxtries
, CTLFLAG_RW
,
142 &arp_maxtries
, 0, "ARP resolution attempts before returning error");
143 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, useloopback
, CTLFLAG_RW
,
144 &useloopback
, 0, "Use the loopback interface for local traffic");
145 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, proxyall
, CTLFLAG_RW
,
146 &arp_proxyall
, 0, "Enable proxy ARP for all suitable requests");
147 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, restricted_match
, CTLFLAG_RW
,
148 &arp_restricted_match
, 0, "Only match against the sender");
149 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, refresh
, CTLFLAG_RW
,
150 &arp_refresh
, 0, "Preemptively refresh the ARP");
151 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, ignore_probes
, CTLFLAG_RW
,
152 &arp_ignore_probes
, 0, "Ignore ARP probes");
154 static void arp_rtrequest(int, struct rtentry
*);
155 static void arprequest(struct ifnet
*, const struct in_addr
*,
156 const struct in_addr
*, const u_char
*);
157 static void arprequest_async(struct ifnet
*, const struct in_addr
*,
158 const struct in_addr
*, const u_char
*);
159 static void arpintr(netmsg_t msg
);
160 static void arptfree(struct llinfo_arp
*);
161 static void arptimer(void *);
162 static struct llinfo_arp
*
163 arplookup(in_addr_t
, boolean_t
, boolean_t
, boolean_t
);
165 static void in_arpinput(struct mbuf
*);
166 static void in_arpreply(struct mbuf
*m
, in_addr_t
, in_addr_t
);
167 static void arp_update_msghandler(netmsg_t
);
168 static void arp_reply_msghandler(netmsg_t
);
171 struct arp_pcpu_data
{
172 struct callout timer_ch
;
173 struct netmsg_base timer_nmsg
;
174 LIST_HEAD(, llinfo_arp
) llinfo_list
;
177 static struct arp_pcpu_data arp_data
[MAXCPU
];
180 * Timeout routine. Age arp_tab entries periodically.
183 arptimer_dispatch(netmsg_t nmsg
)
185 struct llinfo_arp
*la
, *nla
;
186 struct arp_pcpu_data
*ad
= &arp_data
[mycpuid
];
190 lwkt_replymsg(&nmsg
->lmsg
, 0);
193 LIST_FOREACH_MUTABLE(la
, &ad
->llinfo_list
, la_le
, nla
) {
194 if (la
->la_rt
->rt_expire
&& la
->la_rt
->rt_expire
<= time_uptime
)
197 callout_reset(&ad
->timer_ch
, arpt_prune
* hz
, arptimer
, NULL
);
201 arptimer(void *arg __unused
)
204 struct lwkt_msg
*lmsg
= &arp_data
[cpuid
].timer_nmsg
.lmsg
;
207 if (lmsg
->ms_flags
& MSGF_DONE
)
208 lwkt_sendmsg_oncpu(netisr_cpuport(cpuid
), lmsg
);
213 * Parallel to llc_rtrequest.
215 * Called after a route is successfully added to the tree to fix-up the
216 * route and initiate arp operations if required.
219 arp_rtrequest(int req
, struct rtentry
*rt
)
221 struct sockaddr
*gate
= rt
->rt_gateway
;
222 struct llinfo_arp
*la
= rt
->rt_llinfo
;
224 struct sockaddr_dl null_sdl
= { sizeof null_sdl
, AF_LINK
};
226 if (rt
->rt_flags
& RTF_GATEWAY
)
232 * XXX: If this is a manually added route to interface
233 * such as older version of routed or gated might provide,
234 * restore cloning bit.
236 if (!(rt
->rt_flags
& RTF_HOST
) &&
237 SIN(rt_mask(rt
))->sin_addr
.s_addr
!= 0xffffffff)
238 rt
->rt_flags
|= RTF_CLONING
;
239 if (rt
->rt_flags
& RTF_CLONING
) {
241 * Case 1: This route should come from a route to iface.
243 rt_setgate(rt
, rt_key(rt
),
244 (struct sockaddr
*)&null_sdl
,
246 gate
= rt
->rt_gateway
;
247 SDL(gate
)->sdl_type
= rt
->rt_ifp
->if_type
;
248 SDL(gate
)->sdl_index
= rt
->rt_ifp
->if_index
;
249 rt
->rt_expire
= time_uptime
;
253 * Announce a new entry if requested, and only announce it
256 if ((rt
->rt_flags
& RTF_ANNOUNCE
) && mycpuid
== 0) {
257 arprequest_async(rt
->rt_ifp
,
258 &SIN(rt_key(rt
))->sin_addr
,
259 &SIN(rt_key(rt
))->sin_addr
,
264 if (gate
->sa_family
!= AF_LINK
||
265 gate
->sa_len
< sizeof(struct sockaddr_dl
)) {
266 log(LOG_DEBUG
, "arp_rtrequest: bad gateway value\n");
269 SDL(gate
)->sdl_type
= rt
->rt_ifp
->if_type
;
270 SDL(gate
)->sdl_index
= rt
->rt_ifp
->if_index
;
272 break; /* This happens on a route change */
274 * Case 2: This route may come from cloning, or a manual route
275 * add with a LL address.
277 R_Malloc(la
, struct llinfo_arp
*, sizeof *la
);
280 log(LOG_DEBUG
, "arp_rtrequest: malloc failed\n");
283 bzero(la
, sizeof *la
);
285 rt
->rt_flags
|= RTF_LLINFO
;
286 LIST_INSERT_HEAD(&arp_data
[mycpuid
].llinfo_list
, la
, la_le
);
290 * This keeps the multicast addresses from showing up
291 * in `arp -a' listings as unresolved. It's not actually
292 * functional. Then the same for broadcast.
294 if (IN_MULTICAST(ntohl(SIN(rt_key(rt
))->sin_addr
.s_addr
))) {
295 ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt
))->sin_addr
,
297 SDL(gate
)->sdl_alen
= 6;
300 if (in_broadcast(SIN(rt_key(rt
))->sin_addr
, rt
->rt_ifp
)) {
301 memcpy(LLADDR(SDL(gate
)), rt
->rt_ifp
->if_broadcastaddr
,
302 rt
->rt_ifp
->if_addrlen
);
303 SDL(gate
)->sdl_alen
= rt
->rt_ifp
->if_addrlen
;
309 * This fixes up the routing interface for local addresses.
310 * The route is adjusted to point at lo0 and the expiration
313 * NOTE: This prevents locally targetted traffic from going
314 * out the hardware interface, which is inefficient
315 * and might not work if the hardware cannot listen
316 * to its own transmitted packets. Setting
317 * net.link.ether.inet.useloopback to 0 will force
318 * packets for local addresses out the hardware (and
319 * it is expected to receive its own packet).
321 * XXX We should just be able to test RTF_LOCAL here instead
322 * of having to compare IPs.
324 if (SIN(rt_key(rt
))->sin_addr
.s_addr
==
325 (IA_SIN(rt
->rt_ifa
))->sin_addr
.s_addr
) {
327 bcopy(IF_LLADDR(rt
->rt_ifp
), LLADDR(SDL(gate
)),
328 SDL(gate
)->sdl_alen
= rt
->rt_ifp
->if_addrlen
);
337 LIST_REMOVE(la
, la_le
);
338 rt
->rt_llinfo
= NULL
;
339 rt
->rt_flags
&= ~RTF_LLINFO
;
340 if (la
->la_hold
!= NULL
)
341 m_freem(la
->la_hold
);
348 arpreq_alloc(struct ifnet
*ifp
, const struct in_addr
*sip
,
349 const struct in_addr
*tip
, const u_char
*enaddr
)
355 if ((m
= m_gethdr(M_NOWAIT
, MT_DATA
)) == NULL
)
357 m
->m_pkthdr
.rcvif
= NULL
;
359 switch (ifp
->if_type
) {
362 * This may not be correct for types not explicitly
363 * listed, but this is our best guess
366 ar_hrd
= htons(ARPHRD_ETHER
);
368 m
->m_len
= arphdr_len2(ifp
->if_addrlen
, sizeof(struct in_addr
));
369 m
->m_pkthdr
.len
= m
->m_len
;
370 MH_ALIGN(m
, m
->m_len
);
372 ah
= mtod(m
, struct arphdr
*);
377 ah
->ar_pro
= htons(ETHERTYPE_IP
);
378 ah
->ar_hln
= ifp
->if_addrlen
; /* hardware address length */
379 ah
->ar_pln
= sizeof(struct in_addr
); /* protocol address length */
380 ah
->ar_op
= htons(ARPOP_REQUEST
);
381 memcpy(ar_sha(ah
), enaddr
, ah
->ar_hln
);
382 memset(ar_tha(ah
), 0, ah
->ar_hln
);
383 memcpy(ar_spa(ah
), sip
, ah
->ar_pln
);
384 memcpy(ar_tpa(ah
), tip
, ah
->ar_pln
);
390 arpreq_send(struct ifnet
*ifp
, struct mbuf
*m
)
393 struct ether_header
*eh
;
395 switch (ifp
->if_type
) {
398 * This may not be correct for types not explicitly
399 * listed, but this is our best guess
402 eh
= (struct ether_header
*)sa
.sa_data
;
403 /* if_output() will not swap */
404 eh
->ether_type
= htons(ETHERTYPE_ARP
);
405 memcpy(eh
->ether_dhost
, ifp
->if_broadcastaddr
, ifp
->if_addrlen
);
409 sa
.sa_family
= AF_UNSPEC
;
410 sa
.sa_len
= sizeof(sa
);
411 ifp
->if_output(ifp
, m
, &sa
, NULL
);
415 arpreq_send_handler(netmsg_t msg
)
417 struct mbuf
*m
= msg
->packet
.nm_packet
;
418 struct ifnet
*ifp
= msg
->lmsg
.u
.ms_resultp
;
421 /* nmsg was embedded in the mbuf, do not reply! */
425 * Broadcast an ARP request. Caller specifies:
426 * - arp header source ip address
427 * - arp header target ip address
428 * - arp header source ethernet address
430 * NOTE: Caller MUST NOT hold ifp's serializer
433 arprequest(struct ifnet
*ifp
, const struct in_addr
*sip
,
434 const struct in_addr
*tip
, const u_char
*enaddr
)
438 if (enaddr
== NULL
) {
439 if (ifp
->if_bridge
) {
440 enaddr
= IF_LLADDR(ether_bridge_interface(ifp
));
442 enaddr
= IF_LLADDR(ifp
);
446 m
= arpreq_alloc(ifp
, sip
, tip
, enaddr
);
453 * Same as arprequest(), except:
454 * - Caller is allowed to hold ifp's serializer
455 * - Network output is done in protocol thead
458 arprequest_async(struct ifnet
*ifp
, const struct in_addr
*sip
,
459 const struct in_addr
*tip
, const u_char
*enaddr
)
462 struct netmsg_packet
*pmsg
;
464 if (enaddr
== NULL
) {
465 if (ifp
->if_bridge
) {
466 enaddr
= IF_LLADDR(ether_bridge_interface(ifp
));
468 enaddr
= IF_LLADDR(ifp
);
471 m
= arpreq_alloc(ifp
, sip
, tip
, enaddr
);
475 pmsg
= &m
->m_hdr
.mh_netmsg
;
476 netmsg_init(&pmsg
->base
, NULL
, &netisr_apanic_rport
,
477 0, arpreq_send_handler
);
479 pmsg
->base
.lmsg
.u
.ms_resultp
= ifp
;
481 lwkt_sendmsg_oncpu(netisr_cpuport(mycpuid
), &pmsg
->base
.lmsg
);
485 * Resolve an IP address into an ethernet address. If success,
486 * desten is filled in. If there is no entry in arptab,
487 * set one up and broadcast a request for the IP address.
488 * Hold onto this mbuf and resend it once the address
489 * is finally resolved. A return value of 1 indicates
490 * that desten has been filled in and the packet should be sent
491 * normally; a 0 return indicates that the packet has been
492 * taken over here, either now or for later transmission.
495 arpresolve(struct ifnet
*ifp
, struct rtentry
*rt0
, struct mbuf
*m
,
496 struct sockaddr
*dst
, u_char
*desten
)
498 struct rtentry
*rt
= NULL
;
499 struct llinfo_arp
*la
= NULL
;
500 struct sockaddr_dl
*sdl
;
502 if (m
->m_flags
& M_BCAST
) { /* broadcast */
503 memcpy(desten
, ifp
->if_broadcastaddr
, ifp
->if_addrlen
);
506 if (m
->m_flags
& M_MCAST
) {/* multicast */
507 ETHER_MAP_IP_MULTICAST(&SIN(dst
)->sin_addr
, desten
);
511 if (rt_llroute(dst
, rt0
, &rt
) != 0) {
518 la
= arplookup(SIN(dst
)->sin_addr
.s_addr
,
519 TRUE
, RTL_REPORTMSG
, FALSE
);
523 if (la
== NULL
|| rt
== NULL
) {
524 char addr
[INET_ADDRSTRLEN
];
526 log(LOG_DEBUG
, "arpresolve: can't allocate llinfo for %s%s%s\n",
527 kinet_ntoa(SIN(dst
)->sin_addr
, addr
), la
? "la" : " ",
532 sdl
= SDL(rt
->rt_gateway
);
534 * Check the address family and length is valid, the address
535 * is resolved; otherwise, try to resolve.
537 if ((rt
->rt_expire
== 0 || rt
->rt_expire
> time_uptime
) &&
538 sdl
->sdl_family
== AF_LINK
&& sdl
->sdl_alen
!= 0) {
540 * If entry has an expiry time and it is approaching,
541 * see if we need to send an ARP request within this
542 * arpt_down interval.
544 if ((rt
->rt_expire
!= 0) &&
545 (time_uptime
+ la
->la_preempt
> rt
->rt_expire
)) {
547 &SIN(rt
->rt_ifa
->ifa_addr
)->sin_addr
,
553 bcopy(LLADDR(sdl
), desten
, sdl
->sdl_alen
);
557 * If ARP is disabled or static on this interface, stop.
559 * Probably should not allocate empty llinfo struct if we are
560 * not going to be sending out an arp request.
562 if (ifp
->if_flags
& (IFF_NOARP
| IFF_STATICARP
)) {
567 * There is an arptab entry, but no ethernet address
568 * response yet. Replace the held mbuf with this
571 if (la
->la_hold
!= NULL
)
572 m_freem(la
->la_hold
);
574 if (rt
->rt_expire
|| ((rt
->rt_flags
& RTF_STATIC
) && !sdl
->sdl_alen
)) {
575 rt
->rt_flags
&= ~RTF_REJECT
;
576 if (la
->la_asked
== 0 || rt
->rt_expire
!= time_uptime
) {
577 rt
->rt_expire
= time_uptime
;
578 if (la
->la_asked
++ < arp_maxtries
) {
580 &SIN(rt
->rt_ifa
->ifa_addr
)->sin_addr
,
584 rt
->rt_flags
|= RTF_REJECT
;
585 rt
->rt_expire
+= arpt_down
;
587 la
->la_preempt
= arp_maxtries
;
595 * Common length and type checks are done here,
596 * then the protocol-specific routine is called.
599 arpintr(netmsg_t msg
)
601 struct mbuf
*m
= msg
->packet
.nm_packet
;
606 if (m
->m_len
< sizeof(struct arphdr
) &&
607 (m
= m_pullup(m
, sizeof(struct arphdr
))) == NULL
) {
608 log(LOG_ERR
, "arp: runt packet -- m_pullup failed\n");
611 ar
= mtod(m
, struct arphdr
*);
613 ar_hrd
= ntohs(ar
->ar_hrd
);
614 if (ar_hrd
!= ARPHRD_ETHER
&& ar_hrd
!= ARPHRD_IEEE802
) {
615 hexncpy((unsigned char *)&ar
->ar_hrd
, 2, hexstr
, 5, NULL
);
616 log(LOG_ERR
, "arp: unknown hardware address format (0x%s)\n",
622 if (m
->m_pkthdr
.len
< arphdr_len(ar
)) {
623 if ((m
= m_pullup(m
, arphdr_len(ar
))) == NULL
) {
624 log(LOG_ERR
, "arp: runt packet\n");
627 ar
= mtod(m
, struct arphdr
*);
630 switch (ntohs(ar
->ar_pro
)) {
638 /* msg was embedded in the mbuf, do not reply! */
643 * ARP for Internet protocols on 10 Mb/s Ethernet.
644 * Algorithm is that given in RFC 826.
645 * In addition, a sanity check is performed on the sender
646 * protocol address, to catch impersonators.
647 * We no longer handle negotiations for use of trailer protocol:
648 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
649 * along with IP replies if we wanted trailers sent to us,
650 * and also sent them in response to IP replies.
651 * This allowed either end to announce the desire to receive
653 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
654 * but formerly didn't normally send requests.
657 static int log_arp_wrong_iface
= 1;
658 static int log_arp_movements
= 1;
659 static int log_arp_permanent_modify
= 1;
660 static int log_arp_creation_failure
= 1;
662 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, log_arp_wrong_iface
, CTLFLAG_RW
,
663 &log_arp_wrong_iface
, 0,
664 "Log arp packets arriving on the wrong interface");
665 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, log_arp_movements
, CTLFLAG_RW
,
666 &log_arp_movements
, 0,
667 "Log arp replies from MACs different than the one in the cache");
668 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, log_arp_permanent_modify
, CTLFLAG_RW
,
669 &log_arp_permanent_modify
, 0,
670 "Log arp replies from MACs different than the one "
671 "in the permanent arp entry");
672 SYSCTL_INT(_net_link_ether_inet
, OID_AUTO
, log_arp_creation_failure
, CTLFLAG_RW
,
673 &log_arp_creation_failure
, 0, "Log arp creation failure");
676 * Returns non-zero if the routine updated anything.
679 arp_update_oncpu(struct mbuf
*m
, in_addr_t saddr
, boolean_t create
,
680 boolean_t generate_report
, boolean_t dologging
)
682 struct arphdr
*ah
= mtod(m
, struct arphdr
*);
683 struct ifnet
*ifp
= m
->m_pkthdr
.rcvif
;
684 struct llinfo_arp
*la
;
685 struct sockaddr_dl
*sdl
;
688 char sbuf
[INET_ADDRSTRLEN
];
689 int changed
= create
;
691 KASSERT(curthread
->td_type
== TD_TYPE_NETISR
,
692 ("arp update not in netisr"));
694 la
= arplookup(saddr
, create
, generate_report
, FALSE
);
695 if (la
&& (rt
= la
->la_rt
) && (sdl
= SDL(rt
->rt_gateway
))) {
696 struct in_addr isaddr
= { saddr
};
699 * Normally arps coming in on the wrong interface are ignored,
700 * but if we are bridging and the two interfaces belong to
701 * the same bridge, or one is a member of the bridge which
702 * is the other, then it isn't an error.
704 if (rt
->rt_ifp
!= ifp
) {
706 * (1) ifp and rt_ifp both members of same bridge
707 * (2) rt_ifp member of bridge ifp
708 * (3) ifp member of bridge rt_ifp
710 * Always replace rt_ifp with the bridge ifc.
714 if (ifp
->if_bridge
&&
715 rt
->rt_ifp
->if_bridge
== ifp
->if_bridge
) {
716 nifp
= ether_bridge_interface(ifp
);
717 } else if (rt
->rt_ifp
->if_bridge
&&
718 ether_bridge_interface(rt
->rt_ifp
) == ifp
) {
720 } else if (ifp
->if_bridge
&&
721 ether_bridge_interface(ifp
) == rt
->rt_ifp
) {
727 if ((log_arp_wrong_iface
== 1 && nifp
== NULL
) ||
728 log_arp_wrong_iface
== 2) {
729 hexncpy((u_char
*)ar_sha(ah
), ifp
->if_addrlen
,
730 hexstr
[0], HEX_NCPYLEN(ifp
->if_addrlen
), ":");
733 "but got reply from %s on %s\n",
734 kinet_ntoa(isaddr
, sbuf
),
735 rt
->rt_ifp
->if_xname
, hexstr
[0],
742 * nifp is our man! Replace rt_ifp and adjust
745 ifp
= rt
->rt_ifp
= nifp
;
746 if (sdl
->sdl_type
!= ifp
->if_type
) {
747 sdl
->sdl_type
= ifp
->if_type
;
750 if (sdl
->sdl_index
!= ifp
->if_index
) {
751 sdl
->sdl_index
= ifp
->if_index
;
756 bcmp(ar_sha(ah
), LLADDR(sdl
), sdl
->sdl_alen
)) {
758 if (rt
->rt_expire
!= 0) {
759 if (dologging
&& log_arp_movements
) {
760 hexncpy((u_char
*)LLADDR(sdl
), ifp
->if_addrlen
,
761 hexstr
[0], HEX_NCPYLEN(ifp
->if_addrlen
), ":");
762 hexncpy((u_char
*)ar_sha(ah
), ifp
->if_addrlen
,
763 hexstr
[1], HEX_NCPYLEN(ifp
->if_addrlen
), ":");
765 "arp: %s moved from %s to %s on %s\n",
766 kinet_ntoa(isaddr
, sbuf
), hexstr
[0], hexstr
[1],
770 if (dologging
&& log_arp_permanent_modify
) {
771 hexncpy((u_char
*)ar_sha(ah
), ifp
->if_addrlen
,
772 hexstr
[0], HEX_NCPYLEN(ifp
->if_addrlen
), ":");
774 "arp: %s attempts to modify "
775 "permanent entry for %s on %s\n",
776 hexstr
[0], kinet_ntoa(isaddr
, sbuf
), ifp
->if_xname
);
782 * sanity check for the address length.
783 * XXX this does not work for protocols with variable address
786 if (dologging
&& sdl
->sdl_alen
&& sdl
->sdl_alen
!= ah
->ar_hln
) {
787 hexncpy((u_char
*)ar_sha(ah
), ifp
->if_addrlen
,
788 hexstr
[0], HEX_NCPYLEN(ifp
->if_addrlen
), ":");
790 "arp from %s: new addr len %d, was %d",
791 hexstr
[0], ah
->ar_hln
, sdl
->sdl_alen
);
793 if (ifp
->if_addrlen
!= ah
->ar_hln
) {
795 hexncpy((u_char
*)ar_sha(ah
), ifp
->if_addrlen
,
796 hexstr
[0], HEX_NCPYLEN(ifp
->if_addrlen
), ":");
798 "arp from %s: addr len: new %d, i/f %d "
799 "(ignored)", hexstr
[0],
800 ah
->ar_hln
, ifp
->if_addrlen
);
804 memcpy(LLADDR(sdl
), ar_sha(ah
), sdl
->sdl_alen
= ah
->ar_hln
);
805 if (rt
->rt_expire
!= 0) {
806 if (rt
->rt_expire
!= time_uptime
+ arpt_keep
&&
807 rt
->rt_expire
!= time_uptime
+ arpt_keep
- 1) {
808 rt
->rt_expire
= time_uptime
+ arpt_keep
;
812 if (rt
->rt_flags
& RTF_REJECT
) {
813 rt
->rt_flags
&= ~RTF_REJECT
;
816 if (la
->la_asked
!= 0) {
820 if (la
->la_preempt
!= arp_maxtries
) {
821 la
->la_preempt
= arp_maxtries
;
826 * This particular cpu might have been holding an mbuf
827 * pending ARP resolution. If so, transmit the mbuf now.
829 if (la
->la_hold
!= NULL
) {
830 struct mbuf
*m
= la
->la_hold
;
833 m_adj(m
, sizeof(struct ether_header
));
834 ifp
->if_output(ifp
, m
, rt_key(rt
), rt
);
842 * Called from arpintr() - this routine is run from a single cpu.
845 in_arpinput(struct mbuf
*m
)
848 struct ifnet
*ifp
= m
->m_pkthdr
.rcvif
;
849 struct ifaddr_container
*ifac
;
850 struct in_ifaddr_container
*iac
;
851 struct in_ifaddr
*ia
= NULL
;
852 struct in_addr isaddr
, itaddr
, myaddr
;
853 uint8_t *enaddr
= NULL
;
856 char hexstr
[64], sbuf
[INET_ADDRSTRLEN
];
858 req_len
= arphdr_len2(ifp
->if_addrlen
, sizeof(struct in_addr
));
859 if (m
->m_len
< req_len
&& (m
= m_pullup(m
, req_len
)) == NULL
) {
860 log(LOG_ERR
, "in_arp: runt packet -- m_pullup failed\n");
864 ah
= mtod(m
, struct arphdr
*);
865 memcpy(&isaddr
, ar_spa(ah
), sizeof isaddr
);
866 memcpy(&itaddr
, ar_tpa(ah
), sizeof itaddr
);
869 * Check both target and sender IP addresses:
871 * If we receive the packet on the interface owning the address,
872 * then accept the address.
874 * For a bridge, we accept the address if the receive interface and
875 * the interface owning the address are on the same bridge, and
876 * use the bridge MAC as the is-at response. The bridge will be
877 * responsible for handling the packet.
879 * (0) Check target IP against CARP IPs
882 LIST_FOREACH(iac
, INADDR_HASH(itaddr
.s_addr
), ia_hash
) {
883 int is_match
= 0, is_parent
= 0;
887 /* Skip all ia's which don't match */
888 if (itaddr
.s_addr
!= ia
->ia_addr
.sin_addr
.s_addr
)
891 if (ia
->ia_ifp
->if_type
!= IFT_CARP
)
894 if (carp_parent(ia
->ia_ifp
) == ifp
)
896 if (is_parent
|| ia
->ia_ifp
== ifp
)
897 is_match
= carp_iamatch(ia
);
902 * The parent interface will also receive
903 * the ethernet broadcast packets, e.g. ARP
904 * REQUEST, so if we could find a CARP
905 * interface of the parent that could match
906 * the target IP address, we then drop the
907 * packets, which is delieverd to us through
908 * the parent interface.
919 * (1) Check target IP against our local IPs
921 LIST_FOREACH(iac
, INADDR_HASH(itaddr
.s_addr
), ia_hash
) {
924 /* Skip all ia's which don't match */
925 if (itaddr
.s_addr
!= ia
->ia_addr
.sin_addr
.s_addr
)
929 /* CARP interfaces are checked in (0) */
930 if (ia
->ia_ifp
->if_type
== IFT_CARP
)
934 if (ifp
->if_bridge
&& ia
->ia_ifp
&&
935 ifp
->if_bridge
== ia
->ia_ifp
->if_bridge
) {
936 ifp
= ether_bridge_interface(ifp
);
939 if (ia
->ia_ifp
&& ia
->ia_ifp
->if_bridge
&&
940 ether_bridge_interface(ia
->ia_ifp
) == ifp
) {
943 if (ifp
->if_bridge
&& ether_bridge_interface(ifp
) ==
947 if (ia
->ia_ifp
== ifp
) {
953 * (2) Check sender IP against our local IPs
955 LIST_FOREACH(iac
, INADDR_HASH(isaddr
.s_addr
), ia_hash
) {
958 /* Skip all ia's which don't match */
959 if (isaddr
.s_addr
!= ia
->ia_addr
.sin_addr
.s_addr
)
962 if (ifp
->if_bridge
&& ia
->ia_ifp
&&
963 ifp
->if_bridge
== ia
->ia_ifp
->if_bridge
) {
964 ifp
= ether_bridge_interface(ifp
);
967 if (ia
->ia_ifp
&& ia
->ia_ifp
->if_bridge
&&
968 ether_bridge_interface(ia
->ia_ifp
) == ifp
) {
971 if (ifp
->if_bridge
&& ether_bridge_interface(ifp
) ==
976 if (ia
->ia_ifp
== ifp
)
981 * No match, use the first inet address on the receive interface
982 * as a dummy address for the rest of the function.
984 TAILQ_FOREACH(ifac
, &ifp
->if_addrheads
[mycpuid
], ifa_link
) {
985 struct ifaddr
*ifa
= ifac
->ifa
;
987 if (ifa
->ifa_addr
&& ifa
->ifa_addr
->sa_family
== AF_INET
) {
994 * If we got here, we didn't find any suitable interface,
995 * so drop the packet.
1002 enaddr
= (uint8_t *)IF_LLADDR(ifp
);
1003 myaddr
= ia
->ia_addr
.sin_addr
;
1004 if (!bcmp(ar_sha(ah
), enaddr
, ifp
->if_addrlen
)) {
1005 m_freem(m
); /* it's from me, ignore it. */
1008 if (!bcmp(ar_sha(ah
), ifp
->if_broadcastaddr
, ifp
->if_addrlen
)) {
1010 "arp: link address is broadcast for IP address %s!\n",
1011 kinet_ntoa(isaddr
, sbuf
));
1015 if (isaddr
.s_addr
== myaddr
.s_addr
&& myaddr
.s_addr
!= 0) {
1016 hexncpy((u_char
*)ar_sha(ah
), ifp
->if_addrlen
,
1017 hexstr
, HEX_NCPYLEN(ifp
->if_addrlen
), ":");
1019 "arp: %s is using my IP address %s!\n",
1020 hexstr
, kinet_ntoa(isaddr
, sbuf
));
1024 if (ifp
->if_flags
& IFF_STATICARP
)
1028 * When arp_restricted_match is true and the ARP response is not
1029 * specifically targetted to me, ignore it. Otherwise the entry
1030 * timeout may be updated for an old MAC.
1032 if (arp_restricted_match
&& itaddr
.s_addr
!= myaddr
.s_addr
) {
1038 * Update all CPU's routing tables with this ARP packet.
1040 * However, we only need to generate rtmsg on CPU0.
1042 ASSERT_IN_NETISR(0);
1043 changed
= arp_update_oncpu(m
, isaddr
.s_addr
,
1044 itaddr
.s_addr
== myaddr
.s_addr
,
1045 RTL_REPORTMSG
, TRUE
);
1047 if (ncpus
> 1 && changed
) {
1048 struct netmsg_inarp
*msg
= &m
->m_hdr
.mh_arpmsg
;
1050 netmsg_init(&msg
->base
, NULL
, &netisr_apanic_rport
,
1051 0, arp_update_msghandler
);
1053 msg
->saddr
= isaddr
.s_addr
;
1054 msg
->taddr
= itaddr
.s_addr
;
1055 msg
->myaddr
= myaddr
.s_addr
;
1056 lwkt_sendmsg(netisr_cpuport(1), &msg
->base
.lmsg
);
1062 * Just return here; after all CPUs's routing tables are
1063 * properly updated by this ARP packet, an ARP reply will
1064 * be generated if appropriate.
1068 in_arpreply(m
, itaddr
.s_addr
, myaddr
.s_addr
);
1072 arp_reply_msghandler(netmsg_t msg
)
1074 struct netmsg_inarp
*rmsg
= (struct netmsg_inarp
*)msg
;
1076 in_arpreply(rmsg
->m
, rmsg
->taddr
, rmsg
->myaddr
);
1077 /* Don't reply this netmsg; netmsg_inarp is embedded in mbuf */
1081 arp_update_msghandler(netmsg_t msg
)
1083 struct netmsg_inarp
*rmsg
= (struct netmsg_inarp
*)msg
;
1087 * This message handler will be called on all of the APs;
1088 * no need to generate rtmsg on them.
1090 KASSERT(mycpuid
> 0, ("arp update msg on cpu%d", mycpuid
));
1091 arp_update_oncpu(rmsg
->m
, rmsg
->saddr
,
1092 rmsg
->taddr
== rmsg
->myaddr
,
1093 RTL_DONTREPORT
, FALSE
);
1095 nextcpu
= mycpuid
+ 1;
1096 if (nextcpu
< ncpus
) {
1097 lwkt_forwardmsg(netisr_cpuport(nextcpu
), &rmsg
->base
.lmsg
);
1099 struct mbuf
*m
= rmsg
->m
;
1100 in_addr_t saddr
= rmsg
->saddr
;
1101 in_addr_t taddr
= rmsg
->taddr
;
1102 in_addr_t myaddr
= rmsg
->myaddr
;
1105 * Dispatch this mbuf to netisr0 to perform ARP reply,
1107 * NOTE: netmsg_inarp is embedded in this mbuf.
1109 netmsg_init(&rmsg
->base
, NULL
, &netisr_apanic_rport
,
1110 0, arp_reply_msghandler
);
1112 rmsg
->saddr
= saddr
;
1113 rmsg
->taddr
= taddr
;
1114 rmsg
->myaddr
= myaddr
;
1115 lwkt_sendmsg(netisr_cpuport(0), &rmsg
->base
.lmsg
);
1120 in_arpreply(struct mbuf
*m
, in_addr_t taddr
, in_addr_t myaddr
)
1122 struct ifnet
*ifp
= m
->m_pkthdr
.rcvif
;
1123 const uint8_t *enaddr
;
1126 struct ether_header
*eh
;
1128 ah
= mtod(m
, struct arphdr
*);
1129 if (ntohs(ah
->ar_op
) != ARPOP_REQUEST
) {
1134 enaddr
= (const uint8_t *)IF_LLADDR(ifp
);
1135 if (taddr
== myaddr
) {
1136 /* I am the target */
1137 memcpy(ar_tha(ah
), ar_sha(ah
), ah
->ar_hln
);
1138 memcpy(ar_sha(ah
), enaddr
, ah
->ar_hln
);
1140 struct llinfo_arp
*la
;
1143 la
= arplookup(taddr
, FALSE
, RTL_DONTREPORT
, SIN_PROXY
);
1145 struct sockaddr_in sin
;
1147 char tbuf
[INET_ADDRSTRLEN
];
1150 if (!arp_proxyall
) {
1155 bzero(&sin
, sizeof sin
);
1156 sin
.sin_family
= AF_INET
;
1157 sin
.sin_len
= sizeof sin
;
1158 sin
.sin_addr
.s_addr
= taddr
;
1160 rt
= rtpurelookup((struct sockaddr
*)&sin
);
1167 * Don't send proxies for nodes on the same interface
1168 * as this one came out of, or we'll get into a fight
1169 * over who claims what Ether address.
1171 if (rt
->rt_ifp
== ifp
) {
1175 memcpy(ar_tha(ah
), ar_sha(ah
), ah
->ar_hln
);
1176 memcpy(ar_sha(ah
), enaddr
, ah
->ar_hln
);
1178 kprintf("arp: proxying for %s\n",
1179 kinet_ntoa(itaddr
, tbuf
));
1182 struct sockaddr_dl
*sdl
;
1185 memcpy(ar_tha(ah
), ar_sha(ah
), ah
->ar_hln
);
1186 sdl
= SDL(rt
->rt_gateway
);
1187 memcpy(ar_sha(ah
), LLADDR(sdl
), ah
->ar_hln
);
1191 memcpy(ar_tpa(ah
), ar_spa(ah
), ah
->ar_pln
);
1192 memcpy(ar_spa(ah
), &taddr
, ah
->ar_pln
);
1193 ah
->ar_op
= htons(ARPOP_REPLY
);
1194 ah
->ar_pro
= htons(ETHERTYPE_IP
); /* let's be sure! */
1195 switch (ifp
->if_type
) {
1198 * May not be correct for types not explictly
1199 * listed, but it is our best guess.
1202 eh
= (struct ether_header
*)sa
.sa_data
;
1203 memcpy(eh
->ether_dhost
, ar_tha(ah
), sizeof eh
->ether_dhost
);
1204 eh
->ether_type
= htons(ETHERTYPE_ARP
);
1207 sa
.sa_family
= AF_UNSPEC
;
1208 sa
.sa_len
= sizeof sa
;
1209 ifp
->if_output(ifp
, m
, &sa
, NULL
);
1215 * Free an arp entry. If the arp entry is actively referenced or represents
1216 * a static entry we only clear it back to an unresolved state, otherwise
1217 * we destroy the entry entirely.
1219 * Note that static entries are created when route add ... -interface is used
1220 * to create an interface route to a (direct) destination.
1223 arptfree(struct llinfo_arp
*la
)
1225 struct rtentry
*rt
= la
->la_rt
;
1226 struct sockaddr_dl
*sdl
;
1230 sdl
= SDL(rt
->rt_gateway
);
1232 ((rt
->rt_refcnt
> 0 && sdl
->sdl_family
== AF_LINK
) ||
1233 (rt
->rt_flags
& RTF_STATIC
))) {
1235 la
->la_preempt
= la
->la_asked
= 0;
1236 rt
->rt_flags
&= ~RTF_REJECT
;
1239 rtrequest(RTM_DELETE
, rt_key(rt
), NULL
, rt_mask(rt
), 0, NULL
);
1243 * Lookup or enter a new address in arptab.
1245 static struct llinfo_arp
*
1246 arplookup(in_addr_t addr
, boolean_t create
, boolean_t generate_report
,
1250 struct sockaddr_inarp sin
= { sizeof sin
, AF_INET
};
1251 const char *why
= NULL
;
1253 /* Check ARP probes, e.g. from Cisco switches. */
1254 if (addr
== INADDR_ANY
&& arp_ignore_probes
)
1257 sin
.sin_addr
.s_addr
= addr
;
1258 sin
.sin_other
= proxy
? SIN_PROXY
: 0;
1260 rt
= _rtlookup((struct sockaddr
*)&sin
,
1261 generate_report
, RTL_DOCLONE
);
1263 rt
= rtpurelookup((struct sockaddr
*)&sin
);
1269 if (rt
->rt_flags
& RTF_GATEWAY
)
1270 why
= "host is not on local network";
1271 else if (!(rt
->rt_flags
& RTF_LLINFO
))
1272 why
= "could not allocate llinfo";
1273 else if (rt
->rt_gateway
->sa_family
!= AF_LINK
)
1274 why
= "gateway route is not ours";
1277 if (create
&& log_arp_creation_failure
) {
1278 char abuf
[INET_ADDRSTRLEN
];
1280 log(LOG_DEBUG
, "arplookup %s failed: %s\n",
1281 kinet_ntoa(sin
.sin_addr
, abuf
), why
);
1283 if (rt
->rt_refcnt
<= 0 && (rt
->rt_flags
& RTF_WASCLONED
)) {
1284 /* No references to this route. Purge it. */
1285 rtrequest(RTM_DELETE
, rt_key(rt
), rt
->rt_gateway
,
1286 rt_mask(rt
), rt
->rt_flags
, NULL
);
1290 return (rt
->rt_llinfo
);
1294 arp_ifinit(struct ifnet
*ifp
, struct ifaddr
*ifa
)
1296 ifa
->ifa_rtrequest
= arp_rtrequest
;
1297 ifa
->ifa_flags
|= RTF_CLONING
;
1301 arp_gratuitous(struct ifnet
*ifp
, struct ifaddr
*ifa
)
1303 if (IA_SIN(ifa
)->sin_addr
.s_addr
!= INADDR_ANY
) {
1304 arprequest_async(ifp
, &IA_SIN(ifa
)->sin_addr
,
1305 &IA_SIN(ifa
)->sin_addr
, NULL
);
1310 arp_ifaddr(void *arg __unused
, struct ifnet
*ifp
,
1311 enum ifaddr_event event
, struct ifaddr
*ifa
)
1313 if (ifa
->ifa_rtrequest
!= arp_rtrequest
) /* XXX need a generic way */
1315 if (ifa
->ifa_addr
->sa_family
!= AF_INET
)
1317 if (event
== IFADDR_EVENT_DELETE
)
1321 * - CARP interfaces will take care of gratuitous ARP themselves.
1322 * - If we are the CARP interface's parent, don't send gratuitous
1323 * ARP to avoid unnecessary confusion.
1326 if (ifp
->if_type
!= IFT_CARP
&& ifp
->if_carp
== NULL
)
1329 arp_gratuitous(ifp
, ifa
);
1338 for (cpu
= 0; cpu
< ncpus
; cpu
++) {
1339 struct arp_pcpu_data
*ad
= &arp_data
[cpu
];
1341 LIST_INIT(&ad
->llinfo_list
);
1342 netmsg_init(&ad
->timer_nmsg
, NULL
, &netisr_adone_rport
,
1343 MSGF_PRIORITY
, arptimer_dispatch
);
1344 callout_init_mp(&ad
->timer_ch
);
1346 callout_reset_bycpu(&ad
->timer_ch
, hz
, arptimer
, NULL
, cpu
);
1349 netisr_register(NETISR_ARP
, arpintr
, NULL
);
1351 EVENTHANDLER_REGISTER(ifaddr_event
, arp_ifaddr
, NULL
,
1352 EVENTHANDLER_PRI_LAST
);
1355 SYSINIT(arp
, SI_SUB_PROTO_DOMAIN
, SI_ORDER_ANY
, arp_init
, 0);