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>
102 #include <netinet/ip_carp.h>
105 #define SIN(s) ((struct sockaddr_in *)s)
106 #define SDL(s) ((struct sockaddr_dl *)s)
108 MALLOC_DEFINE(M_ARP, "arp", "ARP");
110 SYSCTL_DECL(_net_link_ether);
111 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
114 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
115 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
116 static int arpt_down = 20; /* once declared down, don't send for 20 sec */
118 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
120 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
122 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
125 #define rt_expire rt_rmx.rmx_expire
128 LIST_ENTRY(llinfo_arp) la_le;
129 struct rtentry *la_rt;
130 struct mbuf *la_hold; /* last packet until resolved/timeout */
131 u_short la_preempt; /* countdown for pre-expiry arps */
132 u_short la_asked; /* #times we QUERIED following expiration */
135 static int arp_maxtries = 5;
136 static int useloopback = 1; /* use loopback interface for local traffic */
137 static int arp_proxyall = 0;
138 static int arp_refresh = 60; /* refresh arp cache ~60 (not impl yet) */
139 static int arp_restricted_match = 0;
140 static int arp_ignore_probes = 1;
142 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
143 &arp_maxtries, 0, "ARP resolution attempts before returning error");
144 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
145 &useloopback, 0, "Use the loopback interface for local traffic");
146 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
147 &arp_proxyall, 0, "Enable proxy ARP for all suitable requests");
148 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, restricted_match, CTLFLAG_RW,
149 &arp_restricted_match, 0, "Only match against the sender");
150 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, refresh, CTLFLAG_RW,
151 &arp_refresh, 0, "Preemptively refresh the ARP");
152 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, ignore_probes, CTLFLAG_RW,
153 &arp_ignore_probes, 0, "Ignore ARP probes");
155 static void arp_rtrequest(int, struct rtentry *);
156 static void arprequest(struct ifnet *, const struct in_addr *,
157 const struct in_addr *, const u_char *);
158 static void arprequest_async(struct ifnet *, const struct in_addr *,
159 const struct in_addr *, const u_char *);
160 static void arpintr(netmsg_t msg);
161 static void arptfree(struct llinfo_arp *);
162 static void arptimer(void *);
163 static struct llinfo_arp *
164 arplookup(in_addr_t, boolean_t, boolean_t);
166 static void in_arpinput(struct mbuf *);
167 static void in_arpreply(struct mbuf *m, in_addr_t, in_addr_t);
168 static void arp_update_msghandler(netmsg_t);
169 static void arp_reply_msghandler(netmsg_t);
172 struct arp_pcpu_data {
173 LIST_HEAD(, llinfo_arp) llinfo_list;
174 struct callout timer_ch;
175 struct netmsg_base timer_nmsg;
178 static struct arp_pcpu_data *arp_data[MAXCPU];
181 * Timeout routine. Age arp_tab entries periodically.
184 arptimer_dispatch(netmsg_t nmsg)
186 struct arp_pcpu_data *ad = nmsg->lmsg.u.ms_resultp;
187 struct llinfo_arp *la, *nla;
189 ASSERT_NETISR_NCPUS(mycpuid);
193 netisr_replymsg(&nmsg->base, 0);
196 LIST_FOREACH_MUTABLE(la, &ad->llinfo_list, la_le, nla) {
197 if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_uptime)
200 callout_reset(&ad->timer_ch, arpt_prune * hz, arptimer, &ad->timer_nmsg);
206 struct netmsg_base *nm = xnm;
208 KKASSERT(mycpuid < netisr_ncpus);
211 if (nm->lmsg.ms_flags & MSGF_DONE)
212 netisr_sendmsg_oncpu(nm);
217 * Parallel to llc_rtrequest.
219 * Called after a route is successfully added to the tree to fix-up the
220 * route and initiate arp operations if required.
223 arp_rtrequest(int req, struct rtentry *rt)
225 struct sockaddr *gate = rt->rt_gateway;
226 struct llinfo_arp *la = rt->rt_llinfo;
228 struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK };
230 if (rt->rt_flags & RTF_GATEWAY)
236 * XXX: If this is a manually added route to interface
237 * such as older version of routed or gated might provide,
238 * restore cloning bit.
240 if (!(rt->rt_flags & RTF_HOST) &&
241 SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
242 rt->rt_flags |= RTF_CLONING;
243 if (rt->rt_flags & RTF_CLONING) {
245 * Case 1: This route should come from a route to iface.
247 rt_setgate(rt, rt_key(rt),
248 (struct sockaddr *)&null_sdl);
249 gate = rt->rt_gateway;
250 SDL(gate)->sdl_type = rt->rt_ifp->if_type;
251 SDL(gate)->sdl_index = rt->rt_ifp->if_index;
252 rt->rt_expire = time_uptime;
256 * Announce a new entry if requested, and only announce it
259 if ((rt->rt_flags & RTF_ANNOUNCE) && mycpuid == 0) {
260 arprequest(rt->rt_ifp,
261 &SIN(rt_key(rt))->sin_addr,
262 &SIN(rt_key(rt))->sin_addr,
267 if (gate->sa_family != AF_LINK ||
268 gate->sa_len < sizeof(struct sockaddr_dl)) {
269 log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
272 SDL(gate)->sdl_type = rt->rt_ifp->if_type;
273 SDL(gate)->sdl_index = rt->rt_ifp->if_index;
275 break; /* This happens on a route change */
277 * Case 2: This route may come from cloning, or a manual route
278 * add with a LL address.
280 R_Malloc(la, struct llinfo_arp *, sizeof *la);
283 log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
286 bzero(la, sizeof *la);
288 rt->rt_flags |= RTF_LLINFO;
289 LIST_INSERT_HEAD(&arp_data[mycpuid]->llinfo_list, la, la_le);
293 * This keeps the multicast addresses from showing up
294 * in `arp -a' listings as unresolved. It's not actually
295 * functional. Then the same for broadcast.
297 if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) {
298 ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
300 SDL(gate)->sdl_alen = 6;
303 if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
304 memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
305 rt->rt_ifp->if_addrlen);
306 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
312 * This fixes up the routing interface for local addresses.
313 * The route is adjusted to point at lo0 and the expiration
316 * NOTE: This prevents locally targetted traffic from going
317 * out the hardware interface, which is inefficient
318 * and might not work if the hardware cannot listen
319 * to its own transmitted packets. Setting
320 * net.link.ether.inet.useloopback to 0 will force
321 * packets for local addresses out the hardware (and
322 * it is expected to receive its own packet).
324 * XXX We should just be able to test RTF_LOCAL here instead
325 * of having to compare IPs.
327 if (SIN(rt_key(rt))->sin_addr.s_addr ==
328 (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
330 bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
331 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
340 LIST_REMOVE(la, la_le);
341 rt->rt_llinfo = NULL;
342 rt->rt_flags &= ~RTF_LLINFO;
343 if (la->la_hold != NULL)
344 m_freem(la->la_hold);
351 arpreq_alloc(struct ifnet *ifp, const struct in_addr *sip,
352 const struct in_addr *tip, const u_char *enaddr)
358 if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL)
360 m->m_pkthdr.rcvif = NULL;
362 switch (ifp->if_type) {
365 * This may not be correct for types not explicitly
366 * listed, but this is our best guess
369 ar_hrd = htons(ARPHRD_ETHER);
371 m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
372 m->m_pkthdr.len = m->m_len;
373 MH_ALIGN(m, m->m_len);
375 ah = mtod(m, struct arphdr *);
380 ah->ar_pro = htons(ETHERTYPE_IP);
381 ah->ar_hln = ifp->if_addrlen; /* hardware address length */
382 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */
383 ah->ar_op = htons(ARPOP_REQUEST);
384 memcpy(ar_sha(ah), enaddr, ah->ar_hln);
385 memset(ar_tha(ah), 0, ah->ar_hln);
386 memcpy(ar_spa(ah), sip, ah->ar_pln);
387 memcpy(ar_tpa(ah), tip, ah->ar_pln);
393 arpreq_send(struct ifnet *ifp, struct mbuf *m)
396 struct ether_header *eh;
398 ASSERT_NETISR_NCPUS(mycpuid);
400 switch (ifp->if_type) {
403 * This may not be correct for types not explicitly
404 * listed, but this is our best guess
407 eh = (struct ether_header *)sa.sa_data;
408 /* if_output() will not swap */
409 eh->ether_type = htons(ETHERTYPE_ARP);
410 memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen);
414 sa.sa_family = AF_UNSPEC;
415 sa.sa_len = sizeof(sa);
416 ifp->if_output(ifp, m, &sa, NULL);
420 arpreq_send_handler(netmsg_t msg)
422 struct mbuf *m = msg->packet.nm_packet;
423 struct ifnet *ifp = msg->lmsg.u.ms_resultp;
426 /* nmsg was embedded in the mbuf, do not reply! */
430 * Broadcast an ARP request. Caller specifies:
431 * - arp header source ip address
432 * - arp header target ip address
433 * - arp header source ethernet address
435 * NOTE: Caller MUST NOT hold ifp's serializer
438 arprequest(struct ifnet *ifp, const struct in_addr *sip,
439 const struct in_addr *tip, const u_char *enaddr)
443 ASSERT_NETISR_NCPUS(mycpuid);
445 if (enaddr == NULL) {
446 if (ifp->if_bridge) {
447 enaddr = IF_LLADDR(ether_bridge_interface(ifp));
449 enaddr = IF_LLADDR(ifp);
453 m = arpreq_alloc(ifp, sip, tip, enaddr);
460 * Same as arprequest(), except:
461 * - Caller is allowed to hold ifp's serializer
462 * - Network output is done in protocol thead
465 arprequest_async(struct ifnet *ifp, const struct in_addr *sip,
466 const struct in_addr *tip, const u_char *enaddr)
469 struct netmsg_packet *pmsg;
472 if (enaddr == NULL) {
473 if (ifp->if_bridge) {
474 enaddr = IF_LLADDR(ether_bridge_interface(ifp));
476 enaddr = IF_LLADDR(ifp);
479 m = arpreq_alloc(ifp, sip, tip, enaddr);
483 pmsg = &m->m_hdr.mh_netmsg;
484 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
485 0, arpreq_send_handler);
487 pmsg->base.lmsg.u.ms_resultp = ifp;
489 if (mycpuid < netisr_ncpus)
493 lwkt_sendmsg(netisr_cpuport(cpu), &pmsg->base.lmsg);
497 * Resolve an IP address into an ethernet address. If success,
498 * desten is filled in. If there is no entry in arptab,
499 * set one up and broadcast a request for the IP address.
500 * Hold onto this mbuf and resend it once the address
501 * is finally resolved. A return value of 1 indicates
502 * that desten has been filled in and the packet should be sent
503 * normally; a 0 return indicates that the packet has been
504 * taken over here, either now or for later transmission.
507 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
508 struct sockaddr *dst, u_char *desten)
510 struct rtentry *rt = NULL;
511 struct llinfo_arp *la = NULL;
512 struct sockaddr_dl *sdl;
515 if (m->m_flags & M_BCAST) { /* broadcast */
516 memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
519 if (m->m_flags & M_MCAST) {/* multicast */
520 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
524 error = rt_llroute(dst, rt0, &rt);
532 la = arplookup(SIN(dst)->sin_addr.s_addr, TRUE, FALSE);
536 if (la == NULL || rt == NULL) {
537 char addr[INET_ADDRSTRLEN];
539 log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
540 kinet_ntoa(SIN(dst)->sin_addr, addr), la ? "la" : " ",
547 * Check the address family and length is valid, the address
548 * is resolved; otherwise, try to resolve.
550 sdl = SDL(rt->rt_gateway);
551 if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) &&
552 sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
554 * If entry has an expiry time and it is approaching,
555 * see if we need to send an ARP request within this
556 * arpt_down interval.
558 if ((rt->rt_expire != 0) &&
559 (time_uptime + la->la_preempt > rt->rt_expire)) {
561 &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
567 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
572 * If ARP is disabled or static on this interface, stop.
574 * Probably should not allocate empty llinfo struct if we are
575 * not going to be sending out an arp request.
577 if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
579 return ifp->if_flags & IFF_NOARP ? ENOTSUP : EINVAL;
583 * There is an arptab entry, but no ethernet address
584 * response yet. Replace the held mbuf with this
587 if (la->la_hold != NULL)
588 m_freem(la->la_hold);
592 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
593 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
594 * if we have already sent arp_maxtries ARP requests. Retransmit the
595 * ARP request, but not faster than one request per second.
597 if (la->la_asked < arp_maxtries)
600 error = (rt != NULL && rt->rt_flags & RTF_GATEWAY) ?
601 EHOSTUNREACH : EHOSTDOWN;
603 if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) {
604 rt->rt_flags &= ~RTF_REJECT;
605 if (la->la_asked == 0 || rt->rt_expire != time_uptime) {
606 rt->rt_expire = time_uptime;
608 &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
611 if (la->la_asked++ >= arp_maxtries) {
612 rt->rt_expire += arpt_down;
613 la->la_preempt = arp_maxtries;
614 rt_rtmsg(RTM_MISS, rt, rt->rt_ifp, 0);
622 * Common length and type checks are done here,
623 * then the protocol-specific routine is called.
626 arpintr(netmsg_t msg)
628 struct mbuf *m = msg->packet.nm_packet;
633 if (m->m_len < sizeof(struct arphdr) &&
634 (m = m_pullup(m, sizeof(struct arphdr))) == NULL) {
635 log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
638 ar = mtod(m, struct arphdr *);
640 ar_hrd = ntohs(ar->ar_hrd);
641 if (ar_hrd != ARPHRD_ETHER && ar_hrd != ARPHRD_IEEE802) {
642 hexncpy((unsigned char *)&ar->ar_hrd, 2, hexstr, 5, NULL);
643 log(LOG_ERR, "arp: unknown hardware address format (0x%s)\n",
649 if (m->m_pkthdr.len < arphdr_len(ar)) {
650 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
651 log(LOG_ERR, "arp: runt packet\n");
654 ar = mtod(m, struct arphdr *);
657 switch (ntohs(ar->ar_pro)) {
665 /* msg was embedded in the mbuf, do not reply! */
670 * ARP for Internet protocols on 10 Mb/s Ethernet.
671 * Algorithm is that given in RFC 826.
672 * In addition, a sanity check is performed on the sender
673 * protocol address, to catch impersonators.
674 * We no longer handle negotiations for use of trailer protocol:
675 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
676 * along with IP replies if we wanted trailers sent to us,
677 * and also sent them in response to IP replies.
678 * This allowed either end to announce the desire to receive
680 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
681 * but formerly didn't normally send requests.
684 static int log_arp_wrong_iface = 1;
685 static int log_arp_movements = 1;
686 static int log_arp_permanent_modify = 1;
687 static int log_arp_creation_failure = 1;
689 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
690 &log_arp_wrong_iface, 0,
691 "Log arp packets arriving on the wrong interface");
692 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
693 &log_arp_movements, 0,
694 "Log arp replies from MACs different than the one in the cache");
695 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
696 &log_arp_permanent_modify, 0,
697 "Log arp replies from MACs different than the one "
698 "in the permanent arp entry");
699 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_creation_failure, CTLFLAG_RW,
700 &log_arp_creation_failure, 0, "Log arp creation failure");
703 * Returns non-zero if the routine updated anything.
706 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create,
709 struct arphdr *ah = mtod(m, struct arphdr *);
710 struct ifnet *ifp = m->m_pkthdr.rcvif;
711 struct llinfo_arp *la;
712 struct sockaddr_dl *sdl;
715 char sbuf[INET_ADDRSTRLEN];
716 int changed = create;
718 KASSERT(curthread->td_type == TD_TYPE_NETISR,
719 ("arp update not in netisr"));
721 la = arplookup(saddr, create, FALSE);
722 if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
723 struct in_addr isaddr = { saddr };
724 int rt_cmd = sdl->sdl_alen == 0 ? RTM_ADD : RTM_CHANGE;
725 bool do_rtmsg = false;
728 * Normally arps coming in on the wrong interface are ignored,
729 * but if we are bridging and the two interfaces belong to
730 * the same bridge, or one is a member of the bridge which
731 * is the other, then it isn't an error.
733 if (rt->rt_ifp != ifp) {
735 * (1) ifp and rt_ifp both members of same bridge
736 * (2) rt_ifp member of bridge ifp
737 * (3) ifp member of bridge rt_ifp
739 * Always replace rt_ifp with the bridge ifc.
743 if (ifp->if_bridge &&
744 rt->rt_ifp->if_bridge == ifp->if_bridge) {
745 nifp = ether_bridge_interface(ifp);
746 } else if (rt->rt_ifp->if_bridge &&
747 ether_bridge_interface(rt->rt_ifp) == ifp) {
749 } else if (ifp->if_bridge &&
750 ether_bridge_interface(ifp) == rt->rt_ifp) {
756 if ((log_arp_wrong_iface == 1 && nifp == NULL) ||
757 log_arp_wrong_iface == 2) {
758 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
759 hexstr[0], HEX_NCPYLEN(ifp->if_addrlen), ":");
762 "but got reply from %s on %s\n",
763 kinet_ntoa(isaddr, sbuf),
764 rt->rt_ifp->if_xname, hexstr[0],
771 * nifp is our man! Replace rt_ifp and adjust
774 ifp = rt->rt_ifp = nifp;
775 if (sdl->sdl_type != ifp->if_type) {
776 sdl->sdl_type = ifp->if_type;
780 if (sdl->sdl_index != ifp->if_index) {
781 sdl->sdl_index = ifp->if_index;
787 bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
789 if (rt->rt_expire != 0) {
790 if (dologging && log_arp_movements) {
791 hexncpy((u_char *)LLADDR(sdl), ifp->if_addrlen,
792 hexstr[0], HEX_NCPYLEN(ifp->if_addrlen), ":");
793 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
794 hexstr[1], HEX_NCPYLEN(ifp->if_addrlen), ":");
796 "arp: %s moved from %s to %s on %s\n",
797 kinet_ntoa(isaddr, sbuf), hexstr[0], hexstr[1],
801 if (dologging && log_arp_permanent_modify) {
802 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
803 hexstr[0], HEX_NCPYLEN(ifp->if_addrlen), ":");
805 "arp: %s attempts to modify "
806 "permanent entry for %s on %s\n",
807 hexstr[0], kinet_ntoa(isaddr, sbuf), ifp->if_xname);
814 * sanity check for the address length.
815 * XXX this does not work for protocols with variable address
818 if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) {
819 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
820 hexstr[0], HEX_NCPYLEN(ifp->if_addrlen), ":");
822 "arp from %s: new addr len %d, was %d",
823 hexstr[0], ah->ar_hln, sdl->sdl_alen);
825 if (ifp->if_addrlen != ah->ar_hln) {
827 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
828 hexstr[0], HEX_NCPYLEN(ifp->if_addrlen), ":");
830 "arp from %s: addr len: new %d, i/f %d "
831 "(ignored)", hexstr[0],
832 ah->ar_hln, ifp->if_addrlen);
836 if (sdl->sdl_alen == 0)
838 memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln);
839 if (rt->rt_expire != 0) {
840 if (rt->rt_expire != time_uptime + arpt_keep &&
841 rt->rt_expire != time_uptime + arpt_keep - 1) {
842 rt->rt_expire = time_uptime + arpt_keep;
846 if (rt->rt_flags & RTF_REJECT) {
847 rt->rt_flags &= ~RTF_REJECT;
850 if (la->la_asked != 0) {
854 if (la->la_preempt != arp_maxtries) {
855 la->la_preempt = arp_maxtries;
860 * This particular cpu might have been holding an mbuf
861 * pending ARP resolution. If so, transmit the mbuf now.
863 if (la->la_hold != NULL) {
864 struct mbuf *m = la->la_hold;
867 m_adj(m, sizeof(struct ether_header));
868 ifp->if_output(ifp, m, rt_key(rt), rt);
872 if (do_rtmsg && mycpuid == 0)
873 rt_rtmsg(rt_cmd, rt, rt->rt_ifp, 0);
879 * Called from arpintr() - this routine is run from a single cpu.
882 in_arpinput(struct mbuf *m)
885 struct ifnet *ifp = m->m_pkthdr.rcvif;
886 struct ifaddr_container *ifac;
887 struct in_ifaddr_container *iac;
888 struct in_ifaddr *ia = NULL;
889 struct in_addr isaddr, itaddr, myaddr;
890 uint8_t *enaddr = NULL;
893 char hexstr[64], sbuf[INET_ADDRSTRLEN];
895 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
896 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
897 log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
901 ah = mtod(m, struct arphdr *);
902 memcpy(&isaddr, ar_spa(ah), sizeof isaddr);
903 memcpy(&itaddr, ar_tpa(ah), sizeof itaddr);
906 * Check both target and sender IP addresses:
908 * If we receive the packet on the interface owning the address,
909 * then accept the address.
911 * For a bridge, we accept the address if the receive interface and
912 * the interface owning the address are on the same bridge, and
913 * use the bridge MAC as the is-at response. The bridge will be
914 * responsible for handling the packet.
916 * (0) Check target IP against CARP IPs
919 LIST_FOREACH(iac, INADDR_HASH(itaddr.s_addr), ia_hash) {
920 int is_match = 0, is_parent = 0;
924 /* Skip all ia's which don't match */
925 if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
928 if (ia->ia_ifp->if_type != IFT_CARP)
931 if (carp_parent(ia->ia_ifp) == ifp)
933 if (is_parent || ia->ia_ifp == ifp)
934 is_match = carp_iamatch(ia);
939 * The parent interface will also receive
940 * the ethernet broadcast packets, e.g. ARP
941 * REQUEST, so if we could find a CARP
942 * interface of the parent that could match
943 * the target IP address, we then drop the
944 * packets, which is delieverd to us through
945 * the parent interface.
956 * (1) Check target IP against our local IPs
958 LIST_FOREACH(iac, INADDR_HASH(itaddr.s_addr), ia_hash) {
961 /* Skip all ia's which don't match */
962 if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
966 /* CARP interfaces are checked in (0) */
967 if (ia->ia_ifp->if_type == IFT_CARP)
971 if (ifp->if_bridge && ia->ia_ifp &&
972 ifp->if_bridge == ia->ia_ifp->if_bridge) {
973 ifp = ether_bridge_interface(ifp);
976 if (ia->ia_ifp && ia->ia_ifp->if_bridge &&
977 ether_bridge_interface(ia->ia_ifp) == ifp) {
980 if (ifp->if_bridge && ether_bridge_interface(ifp) ==
984 if (ia->ia_ifp == ifp) {
990 * (2) Check sender IP against our local IPs
992 LIST_FOREACH(iac, INADDR_HASH(isaddr.s_addr), ia_hash) {
995 /* Skip all ia's which don't match */
996 if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
999 if (ifp->if_bridge && ia->ia_ifp &&
1000 ifp->if_bridge == ia->ia_ifp->if_bridge) {
1001 ifp = ether_bridge_interface(ifp);
1004 if (ia->ia_ifp && ia->ia_ifp->if_bridge &&
1005 ether_bridge_interface(ia->ia_ifp) == ifp) {
1008 if (ifp->if_bridge && ether_bridge_interface(ifp) ==
1013 if (ia->ia_ifp == ifp)
1018 * No match, use the first inet address on the receive interface
1019 * as a dummy address for the rest of the function.
1021 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1022 struct ifaddr *ifa = ifac->ifa;
1024 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
1031 * If we got here, we didn't find any suitable interface,
1032 * so drop the packet.
1039 enaddr = (uint8_t *)IF_LLADDR(ifp);
1040 myaddr = ia->ia_addr.sin_addr;
1041 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) {
1042 m_freem(m); /* it's from me, ignore it. */
1045 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
1047 "arp: link address is broadcast for IP address %s!\n",
1048 kinet_ntoa(isaddr, sbuf));
1052 if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
1053 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
1054 hexstr, HEX_NCPYLEN(ifp->if_addrlen), ":");
1056 "arp: %s is using my IP address %s!\n",
1057 hexstr, kinet_ntoa(isaddr, sbuf));
1061 if (ifp->if_flags & IFF_STATICARP)
1065 * When arp_restricted_match is true and the ARP response is not
1066 * specifically targetted to me, ignore it. Otherwise the entry
1067 * timeout may be updated for an old MAC.
1069 if (arp_restricted_match && itaddr.s_addr != myaddr.s_addr) {
1075 * Update all CPU's routing tables with this ARP packet.
1077 * However, we only need to generate rtmsg on CPU0.
1080 changed = arp_update_oncpu(m, isaddr.s_addr,
1081 itaddr.s_addr == myaddr.s_addr,
1084 if (netisr_ncpus > 1 && changed) {
1085 struct netmsg_inarp *msg = &m->m_hdr.mh_arpmsg;
1087 netmsg_init(&msg->base, NULL, &netisr_apanic_rport,
1088 0, arp_update_msghandler);
1090 msg->saddr = isaddr.s_addr;
1091 msg->taddr = itaddr.s_addr;
1092 msg->myaddr = myaddr.s_addr;
1093 lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg);
1099 * Just return here; after all CPUs's routing tables are
1100 * properly updated by this ARP packet, an ARP reply will
1101 * be generated if appropriate.
1105 in_arpreply(m, itaddr.s_addr, myaddr.s_addr);
1109 arp_reply_msghandler(netmsg_t msg)
1111 struct netmsg_inarp *rmsg = (struct netmsg_inarp *)msg;
1113 in_arpreply(rmsg->m, rmsg->taddr, rmsg->myaddr);
1114 /* Don't reply this netmsg; netmsg_inarp is embedded in mbuf */
1118 arp_update_msghandler(netmsg_t msg)
1120 struct netmsg_inarp *rmsg = (struct netmsg_inarp *)msg;
1123 ASSERT_NETISR_NCPUS(mycpuid);
1126 * This message handler will be called on all of the APs;
1127 * no need to generate rtmsg on them.
1129 KASSERT(mycpuid > 0, ("arp update msg on cpu%d", mycpuid));
1130 arp_update_oncpu(rmsg->m, rmsg->saddr,
1131 rmsg->taddr == rmsg->myaddr,
1134 nextcpu = mycpuid + 1;
1135 if (nextcpu < netisr_ncpus) {
1136 lwkt_forwardmsg(netisr_cpuport(nextcpu), &rmsg->base.lmsg);
1138 struct mbuf *m = rmsg->m;
1139 in_addr_t saddr = rmsg->saddr;
1140 in_addr_t taddr = rmsg->taddr;
1141 in_addr_t myaddr = rmsg->myaddr;
1144 * Dispatch this mbuf to netisr0 to perform ARP reply,
1146 * NOTE: netmsg_inarp is embedded in this mbuf.
1148 netmsg_init(&rmsg->base, NULL, &netisr_apanic_rport,
1149 0, arp_reply_msghandler);
1151 rmsg->saddr = saddr;
1152 rmsg->taddr = taddr;
1153 rmsg->myaddr = myaddr;
1154 lwkt_sendmsg(netisr_cpuport(0), &rmsg->base.lmsg);
1159 * Reply to an arp request
1162 in_arpreply(struct mbuf *m, in_addr_t taddr, in_addr_t myaddr)
1164 struct ifnet *ifp = m->m_pkthdr.rcvif;
1165 const uint8_t *enaddr;
1168 struct ether_header *eh;
1172 ah = mtod(m, struct arphdr *);
1173 if (ntohs(ah->ar_op) != ARPOP_REQUEST) {
1178 enaddr = (const uint8_t *)IF_LLADDR(ifp);
1179 if (taddr == myaddr) {
1180 /* I am the target */
1181 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1182 memcpy(ar_sha(ah), enaddr, ah->ar_hln);
1184 struct llinfo_arp *la;
1187 la = arplookup(taddr, FALSE, SIN_PROXY);
1189 struct sockaddr_in sin;
1191 char tbuf[INET_ADDRSTRLEN];
1194 if (!arp_proxyall) {
1199 bzero(&sin, sizeof sin);
1200 sin.sin_family = AF_INET;
1201 sin.sin_len = sizeof sin;
1202 sin.sin_addr.s_addr = taddr;
1204 rt = rtpurelookup((struct sockaddr *)&sin);
1212 * Don't send proxies for nodes on the same interface
1213 * as this one came out of, or we'll get into a fight
1214 * over who claims what Ether address.
1216 * If the rt entry is associated with a bridge, we
1217 * count it as the 'same' interface if ifp is
1218 * associated with the bridge.
1220 if (rt->rt_ifp == ifp || rt->rt_ifp == ifp->if_bridge) {
1224 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1225 memcpy(ar_sha(ah), enaddr, ah->ar_hln);
1227 kprintf("arp: proxying for %s\n",
1228 kinet_ntoa(itaddr, tbuf));
1231 struct sockaddr_dl *sdl;
1234 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1235 sdl = SDL(rt->rt_gateway);
1236 memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
1240 memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1241 memcpy(ar_spa(ah), &taddr, ah->ar_pln);
1242 ah->ar_op = htons(ARPOP_REPLY);
1243 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1244 switch (ifp->if_type) {
1247 * May not be correct for types not explictly
1248 * listed, but it is our best guess.
1251 eh = (struct ether_header *)sa.sa_data;
1252 memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost);
1253 eh->ether_type = htons(ETHERTYPE_ARP);
1256 sa.sa_family = AF_UNSPEC;
1257 sa.sa_len = sizeof sa;
1258 ifp->if_output(ifp, m, &sa, NULL);
1264 * Free an arp entry. If the arp entry is actively referenced or represents
1265 * a static entry we only clear it back to an unresolved state, otherwise
1266 * we destroy the entry entirely.
1268 * Note that static entries are created when route add ... -interface is used
1269 * to create an interface route to a (direct) destination.
1272 arptfree(struct llinfo_arp *la)
1274 struct rtentry *rt = la->la_rt;
1275 struct sockaddr_dl *sdl;
1279 sdl = SDL(rt->rt_gateway);
1281 ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) ||
1282 (rt->rt_flags & RTF_STATIC))) {
1284 la->la_preempt = la->la_asked = 0;
1285 rt->rt_flags &= ~RTF_REJECT;
1290 * ARP expiry happens under one big timer.
1291 * To avoid overflowing the route socket, don't report this.
1292 * Now that RTM_MISS is reported when an address is unresolvable
1293 * the benefit of reporting this deletion is questionable.
1295 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
1299 * Lookup or enter a new address in arptab.
1301 static struct llinfo_arp *
1302 arplookup(in_addr_t addr, boolean_t create,
1306 struct sockaddr_inarp sin = { sizeof sin, AF_INET };
1307 const char *why = NULL;
1309 /* Check ARP probes, e.g. from Cisco switches. */
1310 if (addr == INADDR_ANY && arp_ignore_probes)
1313 sin.sin_addr.s_addr = addr;
1314 sin.sin_other = proxy ? SIN_PROXY : 0;
1316 rt = rtlookup((struct sockaddr *)&sin);
1318 rt = rtpurelookup((struct sockaddr *)&sin);
1324 if (rt->rt_flags & RTF_GATEWAY)
1325 why = "host is not on local network";
1326 else if (!(rt->rt_flags & RTF_LLINFO))
1327 why = "could not allocate llinfo";
1328 else if (rt->rt_gateway->sa_family != AF_LINK)
1329 why = "gateway route is not ours";
1332 if (create && log_arp_creation_failure) {
1333 char abuf[INET_ADDRSTRLEN];
1335 log(LOG_DEBUG, "arplookup %s failed: %s\n",
1336 kinet_ntoa(sin.sin_addr, abuf), why);
1338 if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) {
1339 /* No references to this route. Purge it. */
1340 rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1341 rt_mask(rt), rt->rt_flags, NULL);
1345 return (rt->rt_llinfo);
1349 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1351 ifa->ifa_rtrequest = arp_rtrequest;
1352 ifa->ifa_flags |= RTF_CLONING;
1356 arp_gratuitous(struct ifnet *ifp, struct ifaddr *ifa)
1358 if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) {
1359 if (IN_NETISR_NCPUS(mycpuid)) {
1360 arprequest(ifp, &IA_SIN(ifa)->sin_addr,
1361 &IA_SIN(ifa)->sin_addr, NULL);
1363 arprequest_async(ifp, &IA_SIN(ifa)->sin_addr,
1364 &IA_SIN(ifa)->sin_addr, NULL);
1370 arp_ifaddr(void *arg __unused, struct ifnet *ifp,
1371 enum ifaddr_event event, struct ifaddr *ifa)
1373 if (ifa->ifa_rtrequest != arp_rtrequest) /* XXX need a generic way */
1375 if (ifa->ifa_addr->sa_family != AF_INET)
1377 if (event == IFADDR_EVENT_DELETE)
1381 * - CARP interfaces will take care of gratuitous ARP themselves.
1382 * - If we are the CARP interface's parent, don't send gratuitous
1383 * ARP to avoid unnecessary confusion.
1386 if (ifp->if_type != IFT_CARP && ifp->if_carp == NULL)
1389 arp_gratuitous(ifp, ifa);
1394 arp_init_dispatch(netmsg_t nm)
1396 struct arp_pcpu_data *ad;
1398 ASSERT_NETISR_NCPUS(mycpuid);
1400 ad = kmalloc(sizeof(*ad), M_ARP, M_WAITOK | M_ZERO);
1402 LIST_INIT(&ad->llinfo_list);
1403 callout_init_mp(&ad->timer_ch);
1404 netmsg_init(&ad->timer_nmsg, NULL, &netisr_adone_rport,
1405 MSGF_PRIORITY, arptimer_dispatch);
1406 ad->timer_nmsg.lmsg.u.ms_resultp = ad;
1408 arp_data[mycpuid] = ad;
1410 callout_reset(&ad->timer_ch, hz, arptimer, &ad->timer_nmsg);
1412 netisr_forwardmsg(&nm->base, mycpuid + 1);
1418 struct netmsg_base nm;
1420 netmsg_init(&nm, NULL, &curthread->td_msgport, 0, arp_init_dispatch);
1421 netisr_domsg_global(&nm);
1423 netisr_register(NETISR_ARP, arpintr, NULL);
1425 EVENTHANDLER_REGISTER(ifaddr_event, arp_ifaddr, NULL,
1426 EVENTHANDLER_PRI_LAST);
1428 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);