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 LIST_HEAD(, llinfo_arp) llinfo_arp_list[MAXCPU];
136 static int arp_maxtries = 5;
137 static int useloopback = 1; /* use loopback interface for local traffic */
138 static int arp_proxyall = 0;
139 static int arp_refresh = 60; /* refresh arp cache ~60 (not impl yet) */
140 static int arp_restricted_match = 0;
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");
153 static void arp_rtrequest(int, struct rtentry *);
154 static void arprequest(struct ifnet *, const struct in_addr *,
155 const struct in_addr *, const u_char *);
156 static void arprequest_async(struct ifnet *, const struct in_addr *,
157 const struct in_addr *, const u_char *);
158 static void arpintr(netmsg_t msg);
159 static void arptfree(struct llinfo_arp *);
160 static void arptimer(void *);
161 static struct llinfo_arp *
162 arplookup(in_addr_t, boolean_t, boolean_t, boolean_t);
164 static void in_arpinput(struct mbuf *);
165 static void in_arpreply(struct mbuf *m, in_addr_t, in_addr_t);
166 static void arp_update_msghandler(netmsg_t);
167 static void arp_reply_msghandler(netmsg_t);
170 struct arptimer_ctx {
171 struct callout timer_ch;
172 struct netmsg_base timer_nmsg;
176 static struct arptimer_ctx arptimer_context[MAXCPU];
179 * Timeout routine. Age arp_tab entries periodically.
182 arptimer_dispatch(netmsg_t nmsg)
184 struct llinfo_arp *la, *nla;
189 lwkt_replymsg(&nmsg->lmsg, 0);
192 LIST_FOREACH_MUTABLE(la, &llinfo_arp_list[cpuid], la_le, nla) {
193 if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_uptime)
196 callout_reset(&arptimer_context[cpuid].timer_ch, arpt_prune * hz,
201 arptimer(void *arg __unused)
204 struct lwkt_msg *lmsg = &arptimer_context[cpuid].timer_nmsg.lmsg;
207 if (lmsg->ms_flags & MSGF_DONE)
208 lwkt_sendmsg(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 (__predict_false(!arptimer_context[mycpuid].timer_inited)) {
227 struct arptimer_ctx *ctx = &arptimer_context[mycpuid];
229 ctx->timer_inited = TRUE;
230 netmsg_init(&ctx->timer_nmsg, NULL, &netisr_adone_rport, 0,
232 callout_init_mp(&ctx->timer_ch);
233 callout_reset(&ctx->timer_ch, hz, arptimer, NULL);
235 if (rt->rt_flags & RTF_GATEWAY)
241 * XXX: If this is a manually added route to interface
242 * such as older version of routed or gated might provide,
243 * restore cloning bit.
245 if (!(rt->rt_flags & RTF_HOST) &&
246 SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
247 rt->rt_flags |= RTF_CLONING;
248 if (rt->rt_flags & RTF_CLONING) {
250 * Case 1: This route should come from a route to iface.
252 rt_setgate(rt, rt_key(rt),
253 (struct sockaddr *)&null_sdl,
255 gate = rt->rt_gateway;
256 SDL(gate)->sdl_type = rt->rt_ifp->if_type;
257 SDL(gate)->sdl_index = rt->rt_ifp->if_index;
258 rt->rt_expire = time_uptime;
261 /* Announce a new entry if requested. */
262 if (rt->rt_flags & RTF_ANNOUNCE) {
263 arprequest_async(rt->rt_ifp,
264 &SIN(rt_key(rt))->sin_addr,
265 &SIN(rt_key(rt))->sin_addr,
270 if (gate->sa_family != AF_LINK ||
271 gate->sa_len < sizeof(struct sockaddr_dl)) {
272 log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
275 SDL(gate)->sdl_type = rt->rt_ifp->if_type;
276 SDL(gate)->sdl_index = rt->rt_ifp->if_index;
278 break; /* This happens on a route change */
280 * Case 2: This route may come from cloning, or a manual route
281 * add with a LL address.
283 R_Malloc(la, struct llinfo_arp *, sizeof *la);
286 log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
289 bzero(la, sizeof *la);
291 rt->rt_flags |= RTF_LLINFO;
292 LIST_INSERT_HEAD(&llinfo_arp_list[mycpuid], la, la_le);
296 * This keeps the multicast addresses from showing up
297 * in `arp -a' listings as unresolved. It's not actually
298 * functional. Then the same for broadcast.
300 if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) {
301 ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
303 SDL(gate)->sdl_alen = 6;
306 if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
307 memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
308 rt->rt_ifp->if_addrlen);
309 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
315 * This fixes up the routing interface for local addresses.
316 * The route is adjusted to point at lo0 and the expiration
319 * NOTE: This prevents locally targetted traffic from going
320 * out the hardware interface, which is inefficient
321 * and might not work if the hardware cannot listen
322 * to its own transmitted packets. Setting
323 * net.link.ether.inet.useloopback to 0 will force
324 * packets for local addresses out the hardware (and
325 * it is expected to receive its own packet).
327 * XXX We should just be able to test RTF_LOCAL here instead
328 * of having to compare IPs.
330 if (SIN(rt_key(rt))->sin_addr.s_addr ==
331 (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
333 bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
334 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
343 LIST_REMOVE(la, la_le);
344 rt->rt_llinfo = NULL;
345 rt->rt_flags &= ~RTF_LLINFO;
346 if (la->la_hold != NULL)
347 m_freem(la->la_hold);
354 arpreq_alloc(struct ifnet *ifp, const struct in_addr *sip,
355 const struct in_addr *tip, const u_char *enaddr)
361 if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL)
363 m->m_pkthdr.rcvif = NULL;
365 switch (ifp->if_type) {
368 * This may not be correct for types not explicitly
369 * listed, but this is our best guess
372 ar_hrd = htons(ARPHRD_ETHER);
374 m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
375 m->m_pkthdr.len = m->m_len;
376 MH_ALIGN(m, m->m_len);
378 ah = mtod(m, struct arphdr *);
383 ah->ar_pro = htons(ETHERTYPE_IP);
384 ah->ar_hln = ifp->if_addrlen; /* hardware address length */
385 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */
386 ah->ar_op = htons(ARPOP_REQUEST);
387 memcpy(ar_sha(ah), enaddr, ah->ar_hln);
388 memset(ar_tha(ah), 0, ah->ar_hln);
389 memcpy(ar_spa(ah), sip, ah->ar_pln);
390 memcpy(ar_tpa(ah), tip, ah->ar_pln);
396 arpreq_send(struct ifnet *ifp, struct mbuf *m)
399 struct ether_header *eh;
401 switch (ifp->if_type) {
404 * This may not be correct for types not explicitly
405 * listed, but this is our best guess
408 eh = (struct ether_header *)sa.sa_data;
409 /* if_output() will not swap */
410 eh->ether_type = htons(ETHERTYPE_ARP);
411 memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen);
415 sa.sa_family = AF_UNSPEC;
416 sa.sa_len = sizeof(sa);
417 ifp->if_output(ifp, m, &sa, NULL);
421 arpreq_send_handler(netmsg_t msg)
423 struct mbuf *m = msg->packet.nm_packet;
424 struct ifnet *ifp = msg->lmsg.u.ms_resultp;
427 /* nmsg was embedded in the mbuf, do not reply! */
431 * Broadcast an ARP request. Caller specifies:
432 * - arp header source ip address
433 * - arp header target ip address
434 * - arp header source ethernet address
436 * NOTE: Caller MUST NOT hold ifp's serializer
439 arprequest(struct ifnet *ifp, const struct in_addr *sip,
440 const struct in_addr *tip, const u_char *enaddr)
444 if (enaddr == NULL) {
445 if (ifp->if_bridge) {
446 enaddr = IF_LLADDR(ether_bridge_interface(ifp));
448 enaddr = IF_LLADDR(ifp);
452 m = arpreq_alloc(ifp, sip, tip, enaddr);
459 * Same as arprequest(), except:
460 * - Caller is allowed to hold ifp's serializer
461 * - Network output is done in protocol thead
464 arprequest_async(struct ifnet *ifp, const struct in_addr *sip,
465 const struct in_addr *tip, const u_char *enaddr)
468 struct netmsg_packet *pmsg;
470 if (enaddr == NULL) {
471 if (ifp->if_bridge) {
472 enaddr = IF_LLADDR(ether_bridge_interface(ifp));
474 enaddr = IF_LLADDR(ifp);
477 m = arpreq_alloc(ifp, sip, tip, enaddr);
481 pmsg = &m->m_hdr.mh_netmsg;
482 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
483 0, arpreq_send_handler);
485 pmsg->base.lmsg.u.ms_resultp = ifp;
487 lwkt_sendmsg(netisr_cpuport(mycpuid), &pmsg->base.lmsg);
491 * Resolve an IP address into an ethernet address. If success,
492 * desten is filled in. If there is no entry in arptab,
493 * set one up and broadcast a request for the IP address.
494 * Hold onto this mbuf and resend it once the address
495 * is finally resolved. A return value of 1 indicates
496 * that desten has been filled in and the packet should be sent
497 * normally; a 0 return indicates that the packet has been
498 * taken over here, either now or for later transmission.
501 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
502 struct sockaddr *dst, u_char *desten)
504 struct rtentry *rt = NULL;
505 struct llinfo_arp *la = NULL;
506 struct sockaddr_dl *sdl;
508 if (m->m_flags & M_BCAST) { /* broadcast */
509 memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
512 if (m->m_flags & M_MCAST) {/* multicast */
513 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
517 if (rt_llroute(dst, rt0, &rt) != 0) {
524 la = arplookup(SIN(dst)->sin_addr.s_addr,
525 TRUE, RTL_REPORTMSG, FALSE);
529 if (la == NULL || rt == NULL) {
530 log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
531 inet_ntoa(SIN(dst)->sin_addr), la ? "la" : " ",
536 sdl = SDL(rt->rt_gateway);
538 * Check the address family and length is valid, the address
539 * is resolved; otherwise, try to resolve.
541 if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) &&
542 sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
544 * If entry has an expiry time and it is approaching,
545 * see if we need to send an ARP request within this
546 * arpt_down interval.
548 if ((rt->rt_expire != 0) &&
549 (time_uptime + la->la_preempt > rt->rt_expire)) {
551 &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
557 bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
561 * If ARP is disabled or static on this interface, stop.
563 * Probably should not allocate empty llinfo struct if we are
564 * not going to be sending out an arp request.
566 if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) {
571 * There is an arptab entry, but no ethernet address
572 * response yet. Replace the held mbuf with this
575 if (la->la_hold != NULL)
576 m_freem(la->la_hold);
578 if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) {
579 rt->rt_flags &= ~RTF_REJECT;
580 if (la->la_asked == 0 || rt->rt_expire != time_uptime) {
581 rt->rt_expire = time_uptime;
582 if (la->la_asked++ < arp_maxtries) {
584 &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
588 rt->rt_flags |= RTF_REJECT;
589 rt->rt_expire += arpt_down;
591 la->la_preempt = arp_maxtries;
599 * Common length and type checks are done here,
600 * then the protocol-specific routine is called.
603 arpintr(netmsg_t msg)
605 struct mbuf *m = msg->packet.nm_packet;
610 if (m->m_len < sizeof(struct arphdr) &&
611 (m = m_pullup(m, sizeof(struct arphdr))) == NULL) {
612 log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
615 ar = mtod(m, struct arphdr *);
617 ar_hrd = ntohs(ar->ar_hrd);
618 if (ar_hrd != ARPHRD_ETHER && ar_hrd != ARPHRD_IEEE802) {
619 hexncpy((unsigned char *)&ar->ar_hrd, 2, hexstr, 5, NULL);
620 log(LOG_ERR, "arp: unknown hardware address format (0x%s)\n",
626 if (m->m_pkthdr.len < arphdr_len(ar)) {
627 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
628 log(LOG_ERR, "arp: runt packet\n");
631 ar = mtod(m, struct arphdr *);
634 switch (ntohs(ar->ar_pro)) {
642 /* msg was embedded in the mbuf, do not reply! */
647 * ARP for Internet protocols on 10 Mb/s Ethernet.
648 * Algorithm is that given in RFC 826.
649 * In addition, a sanity check is performed on the sender
650 * protocol address, to catch impersonators.
651 * We no longer handle negotiations for use of trailer protocol:
652 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
653 * along with IP replies if we wanted trailers sent to us,
654 * and also sent them in response to IP replies.
655 * This allowed either end to announce the desire to receive
657 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
658 * but formerly didn't normally send requests.
661 static int log_arp_wrong_iface = 1;
662 static int log_arp_movements = 1;
663 static int log_arp_permanent_modify = 1;
665 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
666 &log_arp_wrong_iface, 0,
667 "Log arp packets arriving on the wrong interface");
668 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
669 &log_arp_movements, 0,
670 "Log arp replies from MACs different than the one in the cache");
671 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
672 &log_arp_permanent_modify, 0,
673 "Log arp replies from MACs different than the one "
674 "in the permanent arp entry");
677 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create,
678 boolean_t generate_report, boolean_t dologging)
680 struct arphdr *ah = mtod(m, struct arphdr *);
681 struct ifnet *ifp = m->m_pkthdr.rcvif;
682 struct llinfo_arp *la;
683 struct sockaddr_dl *sdl;
687 KASSERT(curthread->td_type == TD_TYPE_NETISR,
688 ("arp update not in netisr"));
690 la = arplookup(saddr, create, generate_report, FALSE);
691 if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
692 struct in_addr isaddr = { saddr };
695 * Normally arps coming in on the wrong interface are ignored,
696 * but if we are bridging and the two interfaces belong to
697 * the same bridge, or one is a member of the bridge which
698 * is the other, then it isn't an error.
700 if (rt->rt_ifp != ifp) {
702 * (1) ifp and rt_ifp both members of same bridge
703 * (2) rt_ifp member of bridge ifp
704 * (3) ifp member of bridge rt_ifp
706 * Always replace rt_ifp with the bridge ifc.
710 if (ifp->if_bridge &&
711 rt->rt_ifp->if_bridge == ifp->if_bridge) {
712 nifp = ether_bridge_interface(ifp);
713 } else if (rt->rt_ifp->if_bridge &&
714 ether_bridge_interface(rt->rt_ifp) == ifp) {
716 } else if (ifp->if_bridge &&
717 ether_bridge_interface(ifp) == rt->rt_ifp) {
723 if ((log_arp_wrong_iface == 1 && nifp == NULL) ||
724 log_arp_wrong_iface == 2) {
725 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
726 hexstr[0], HEX_NCPYLEN(ifp->if_addrlen), ":");
729 "but got reply from %s on %s\n",
731 rt->rt_ifp->if_xname, hexstr[0],
738 * nifp is our man! Replace rt_ifp and adjust
741 ifp = rt->rt_ifp = nifp;
742 sdl->sdl_type = ifp->if_type;
743 sdl->sdl_index = ifp->if_index;
746 bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
747 if (rt->rt_expire != 0) {
748 if (dologging && log_arp_movements) {
749 hexncpy((u_char *)LLADDR(sdl), ifp->if_addrlen,
750 hexstr[0], HEX_NCPYLEN(ifp->if_addrlen), ":");
751 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
752 hexstr[1], HEX_NCPYLEN(ifp->if_addrlen), ":");
754 "arp: %s moved from %s to %s on %s\n",
755 inet_ntoa(isaddr), hexstr[0], hexstr[1],
759 if (dologging && log_arp_permanent_modify) {
760 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
761 hexstr[0], HEX_NCPYLEN(ifp->if_addrlen), ":");
763 "arp: %s attempts to modify "
764 "permanent entry for %s on %s\n",
765 hexstr[0], inet_ntoa(isaddr), ifp->if_xname);
771 * sanity check for the address length.
772 * XXX this does not work for protocols with variable address
775 if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) {
776 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
777 hexstr[0], HEX_NCPYLEN(ifp->if_addrlen), ":");
779 "arp from %s: new addr len %d, was %d",
780 hexstr[0], ah->ar_hln, sdl->sdl_alen);
782 if (ifp->if_addrlen != ah->ar_hln) {
784 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
785 hexstr[0], HEX_NCPYLEN(ifp->if_addrlen), ":");
787 "arp from %s: addr len: new %d, i/f %d "
788 "(ignored)", hexstr[0],
789 ah->ar_hln, ifp->if_addrlen);
793 memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln);
794 if (rt->rt_expire != 0) {
795 rt->rt_expire = time_uptime + arpt_keep;
797 rt->rt_flags &= ~RTF_REJECT;
799 la->la_preempt = arp_maxtries;
802 * This particular cpu might have been holding an mbuf
803 * pending ARP resolution. If so, transmit the mbuf now.
805 if (la->la_hold != NULL) {
806 struct mbuf *m = la->la_hold;
809 m_adj(m, sizeof(struct ether_header));
810 ifp->if_output(ifp, m, rt_key(rt), rt);
816 * Called from arpintr() - this routine is run from a single cpu.
819 in_arpinput(struct mbuf *m)
822 struct ifnet *ifp = m->m_pkthdr.rcvif;
823 struct ifaddr_container *ifac;
824 struct in_ifaddr_container *iac;
825 struct in_ifaddr *ia = NULL;
826 struct in_addr isaddr, itaddr, myaddr;
827 uint8_t *enaddr = NULL;
831 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
832 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
833 log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
837 ah = mtod(m, struct arphdr *);
838 memcpy(&isaddr, ar_spa(ah), sizeof isaddr);
839 memcpy(&itaddr, ar_tpa(ah), sizeof itaddr);
842 * Check both target and sender IP addresses:
844 * If we receive the packet on the interface owning the address,
845 * then accept the address.
847 * For a bridge, we accept the address if the receive interface and
848 * the interface owning the address are on the same bridge, and
849 * use the bridge MAC as the is-at response. The bridge will be
850 * responsible for handling the packet.
852 * (0) Check target IP against CARP IPs
855 LIST_FOREACH(iac, INADDR_HASH(itaddr.s_addr), ia_hash) {
856 int is_match = 0, is_parent = 0;
860 /* Skip all ia's which don't match */
861 if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
864 if (ia->ia_ifp->if_type != IFT_CARP)
867 if (carp_parent(ia->ia_ifp) == ifp)
869 if (is_parent || ia->ia_ifp == ifp)
870 is_match = carp_iamatch(ia);
875 * The parent interface will also receive
876 * the ethernet broadcast packets, e.g. ARP
877 * REQUEST, so if we could find a CARP
878 * interface of the parent that could match
879 * the target IP address, we then drop the
880 * packets, which is delieverd to us through
881 * the parent interface.
892 * (1) Check target IP against our local IPs
894 LIST_FOREACH(iac, INADDR_HASH(itaddr.s_addr), ia_hash) {
897 /* Skip all ia's which don't match */
898 if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
902 /* CARP interfaces are checked in (0) */
903 if (ia->ia_ifp->if_type == IFT_CARP)
907 if (ifp->if_bridge && ia->ia_ifp &&
908 ifp->if_bridge == ia->ia_ifp->if_bridge) {
909 ifp = ether_bridge_interface(ifp);
912 if (ia->ia_ifp && ia->ia_ifp->if_bridge &&
913 ether_bridge_interface(ia->ia_ifp) == ifp) {
916 if (ifp->if_bridge && ether_bridge_interface(ifp) ==
920 if (ia->ia_ifp == ifp) {
926 * (2) Check sender IP against our local IPs
928 LIST_FOREACH(iac, INADDR_HASH(isaddr.s_addr), ia_hash) {
931 /* Skip all ia's which don't match */
932 if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
935 if (ifp->if_bridge && ia->ia_ifp &&
936 ifp->if_bridge == ia->ia_ifp->if_bridge) {
937 ifp = ether_bridge_interface(ifp);
940 if (ia->ia_ifp && ia->ia_ifp->if_bridge &&
941 ether_bridge_interface(ia->ia_ifp) == ifp) {
944 if (ifp->if_bridge && ether_bridge_interface(ifp) ==
949 if (ia->ia_ifp == ifp)
954 * No match, use the first inet address on the receive interface
955 * as a dummy address for the rest of the function.
957 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
958 struct ifaddr *ifa = ifac->ifa;
960 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
967 * If we got here, we didn't find any suitable interface,
968 * so drop the packet.
975 enaddr = (uint8_t *)IF_LLADDR(ifp);
976 myaddr = ia->ia_addr.sin_addr;
977 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) {
978 m_freem(m); /* it's from me, ignore it. */
981 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
983 "arp: link address is broadcast for IP address %s!\n",
988 if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
989 hexncpy((u_char *)ar_sha(ah), ifp->if_addrlen,
990 hexstr, HEX_NCPYLEN(ifp->if_addrlen), ":");
992 "arp: %s is using my IP address %s!\n",
993 hexstr, inet_ntoa(isaddr));
997 if (ifp->if_flags & IFF_STATICARP)
1001 * When arp_restricted_match is true and the ARP response is not
1002 * specifically targetted to me, ignore it. Otherwise the entry
1003 * timeout may be updated for an old MAC.
1005 if (arp_restricted_match && itaddr.s_addr != myaddr.s_addr) {
1011 * Update all CPU's routing tables with this ARP packet.
1013 * However, we only need to generate rtmsg on CPU0.
1015 KASSERT(&curthread->td_msgport == netisr_cpuport(0),
1016 ("arp input not in netisr0, but on cpu%d", mycpuid));
1017 arp_update_oncpu(m, isaddr.s_addr, itaddr.s_addr == myaddr.s_addr,
1018 RTL_REPORTMSG, TRUE);
1021 struct netmsg_inarp *msg = &m->m_hdr.mh_arpmsg;
1023 netmsg_init(&msg->base, NULL, &netisr_apanic_rport,
1024 0, arp_update_msghandler);
1026 msg->saddr = isaddr.s_addr;
1027 msg->taddr = itaddr.s_addr;
1028 msg->myaddr = myaddr.s_addr;
1029 lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg);
1035 * Just return here; after all CPUs's routing tables are
1036 * properly updated by this ARP packet, an ARP reply will
1037 * be generated if appropriate.
1041 in_arpreply(m, itaddr.s_addr, myaddr.s_addr);
1045 arp_reply_msghandler(netmsg_t msg)
1047 struct netmsg_inarp *rmsg = (struct netmsg_inarp *)msg;
1049 in_arpreply(rmsg->m, rmsg->taddr, rmsg->myaddr);
1050 /* Don't reply this netmsg; netmsg_inarp is embedded in mbuf */
1054 arp_update_msghandler(netmsg_t msg)
1056 struct netmsg_inarp *rmsg = (struct netmsg_inarp *)msg;
1060 * This message handler will be called on all of the APs;
1061 * no need to generate rtmsg on them.
1063 KASSERT(mycpuid > 0, ("arp update msg on cpu%d", mycpuid));
1064 arp_update_oncpu(rmsg->m, rmsg->saddr, rmsg->taddr == rmsg->myaddr,
1065 RTL_DONTREPORT, FALSE);
1067 nextcpu = mycpuid + 1;
1068 if (nextcpu < ncpus) {
1069 lwkt_forwardmsg(netisr_cpuport(nextcpu), &rmsg->base.lmsg);
1071 struct mbuf *m = rmsg->m;
1072 in_addr_t saddr = rmsg->saddr;
1073 in_addr_t taddr = rmsg->taddr;
1074 in_addr_t myaddr = rmsg->myaddr;
1077 * Dispatch this mbuf to netisr0 to perform ARP reply,
1079 * NOTE: netmsg_inarp is embedded in this mbuf.
1081 netmsg_init(&rmsg->base, NULL, &netisr_apanic_rport,
1082 0, arp_reply_msghandler);
1084 rmsg->saddr = saddr;
1085 rmsg->taddr = taddr;
1086 rmsg->myaddr = myaddr;
1087 lwkt_sendmsg(netisr_cpuport(0), &rmsg->base.lmsg);
1092 in_arpreply(struct mbuf *m, in_addr_t taddr, in_addr_t myaddr)
1094 struct ifnet *ifp = m->m_pkthdr.rcvif;
1095 const uint8_t *enaddr;
1098 struct ether_header *eh;
1100 ah = mtod(m, struct arphdr *);
1101 if (ntohs(ah->ar_op) != ARPOP_REQUEST) {
1106 enaddr = (const uint8_t *)IF_LLADDR(ifp);
1107 if (taddr == myaddr) {
1108 /* I am the target */
1109 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1110 memcpy(ar_sha(ah), enaddr, ah->ar_hln);
1112 struct llinfo_arp *la;
1115 la = arplookup(taddr, FALSE, RTL_DONTREPORT, SIN_PROXY);
1117 struct sockaddr_in sin;
1119 if (!arp_proxyall) {
1124 bzero(&sin, sizeof sin);
1125 sin.sin_family = AF_INET;
1126 sin.sin_len = sizeof sin;
1127 sin.sin_addr.s_addr = taddr;
1129 rt = rtpurelookup((struct sockaddr *)&sin);
1136 * Don't send proxies for nodes on the same interface
1137 * as this one came out of, or we'll get into a fight
1138 * over who claims what Ether address.
1140 if (rt->rt_ifp == ifp) {
1144 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1145 memcpy(ar_sha(ah), enaddr, ah->ar_hln);
1147 kprintf("arp: proxying for %s\n", inet_ntoa(itaddr));
1150 struct sockaddr_dl *sdl;
1153 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1154 sdl = SDL(rt->rt_gateway);
1155 memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
1159 memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1160 memcpy(ar_spa(ah), &taddr, ah->ar_pln);
1161 ah->ar_op = htons(ARPOP_REPLY);
1162 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1163 switch (ifp->if_type) {
1166 * May not be correct for types not explictly
1167 * listed, but it is our best guess.
1170 eh = (struct ether_header *)sa.sa_data;
1171 memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost);
1172 eh->ether_type = htons(ETHERTYPE_ARP);
1175 sa.sa_family = AF_UNSPEC;
1176 sa.sa_len = sizeof sa;
1177 ifp->if_output(ifp, m, &sa, NULL);
1183 * Free an arp entry. If the arp entry is actively referenced or represents
1184 * a static entry we only clear it back to an unresolved state, otherwise
1185 * we destroy the entry entirely.
1187 * Note that static entries are created when route add ... -interface is used
1188 * to create an interface route to a (direct) destination.
1191 arptfree(struct llinfo_arp *la)
1193 struct rtentry *rt = la->la_rt;
1194 struct sockaddr_dl *sdl;
1198 sdl = SDL(rt->rt_gateway);
1200 ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) ||
1201 (rt->rt_flags & RTF_STATIC))) {
1203 la->la_preempt = la->la_asked = 0;
1204 rt->rt_flags &= ~RTF_REJECT;
1207 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
1211 * Lookup or enter a new address in arptab.
1213 static struct llinfo_arp *
1214 arplookup(in_addr_t addr, boolean_t create, boolean_t generate_report,
1218 struct sockaddr_inarp sin = { sizeof sin, AF_INET };
1219 const char *why = NULL;
1221 sin.sin_addr.s_addr = addr;
1222 sin.sin_other = proxy ? SIN_PROXY : 0;
1224 rt = _rtlookup((struct sockaddr *)&sin,
1225 generate_report, RTL_DOCLONE);
1227 rt = rtpurelookup((struct sockaddr *)&sin);
1233 if (rt->rt_flags & RTF_GATEWAY)
1234 why = "host is not on local network";
1235 else if (!(rt->rt_flags & RTF_LLINFO))
1236 why = "could not allocate llinfo";
1237 else if (rt->rt_gateway->sa_family != AF_LINK)
1238 why = "gateway route is not ours";
1242 log(LOG_DEBUG, "arplookup %s failed: %s\n",
1243 inet_ntoa(sin.sin_addr), why);
1245 if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) {
1246 /* No references to this route. Purge it. */
1247 rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1248 rt_mask(rt), rt->rt_flags, NULL);
1252 return (rt->rt_llinfo);
1256 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1258 ifa->ifa_rtrequest = arp_rtrequest;
1259 ifa->ifa_flags |= RTF_CLONING;
1263 arp_gratuitous(struct ifnet *ifp, struct ifaddr *ifa)
1265 if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) {
1266 arprequest_async(ifp, &IA_SIN(ifa)->sin_addr,
1267 &IA_SIN(ifa)->sin_addr, NULL);
1272 arp_ifaddr(void *arg __unused, struct ifnet *ifp,
1273 enum ifaddr_event event, struct ifaddr *ifa)
1275 if (ifa->ifa_rtrequest != arp_rtrequest) /* XXX need a generic way */
1277 if (ifa->ifa_addr->sa_family != AF_INET)
1279 if (event == IFADDR_EVENT_DELETE)
1283 * - CARP interfaces will take care of gratuitous ARP themselves.
1284 * - If we are the CARP interface's parent, don't send gratuitous
1285 * ARP to avoid unnecessary confusion.
1288 if (ifp->if_type != IFT_CARP && ifp->if_carp == NULL)
1291 arp_gratuitous(ifp, ifa);
1300 for (cpu = 0; cpu < ncpus2; cpu++)
1301 LIST_INIT(&llinfo_arp_list[cpu]);
1303 netisr_register(NETISR_ARP, arpintr, NULL);
1305 EVENTHANDLER_REGISTER(ifaddr_event, arp_ifaddr, NULL,
1306 EVENTHANDLER_PRI_LAST);
1309 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);